From d631097577f6fe027f4903f62eabced6445d19bf Mon Sep 17 00:00:00 2001 From: Ilya Dryomov Date: Mon, 31 Oct 2011 11:07:42 +0200 Subject: tracing: fix event_subsystem ref counting Fix a bug introduced by e9dbfae5, which prevents event_subsystem from ever being released. Ref_count was added to keep track of subsystem users, not for counting events. Subsystem is created with ref_count = 1, so there is no need to increment it for every event, we have nr_events for that. Fix this by touching ref_count only when we actually have a new user - subsystem_open(). Cc: stable@vger.kernel.org Signed-off-by: Ilya Dryomov Link: http://lkml.kernel.org/r/1320052062-7846-1-git-send-email-idryomov@gmail.com Signed-off-by: Steven Rostedt --- kernel/trace/trace_events.c | 1 - 1 file changed, 1 deletion(-) diff --git a/kernel/trace/trace_events.c b/kernel/trace/trace_events.c index 581876f9f387..c212a7f934ec 100644 --- a/kernel/trace/trace_events.c +++ b/kernel/trace/trace_events.c @@ -1078,7 +1078,6 @@ event_subsystem_dir(const char *name, struct dentry *d_events) /* First see if we did not already create this dir */ list_for_each_entry(system, &event_subsystems, list) { if (strcmp(system->name, name) == 0) { - __get_system(system); system->nr_events++; return system->entry; } -- cgit v1.2.3 From ed0449af5373abd766c79fbf83254bebc996bd23 Mon Sep 17 00:00:00 2001 From: Li Zefan Date: Tue, 1 Nov 2011 09:09:35 +0800 Subject: tracing: Restore system filter behavior Though not all events have field 'prev_pid', it was allowed to do this: # echo 'prev_pid == 100' > events/sched/filter but commit 75b8e98263fdb0bfbdeba60d4db463259f1fe8a2 (tracing/filter: Swap entire filter of events) broke it without any reason. Link: http://lkml.kernel.org/r/4EAF46CF.8040408@cn.fujitsu.com Signed-off-by: Li Zefan Signed-off-by: Steven Rostedt --- include/linux/ftrace_event.h | 2 ++ kernel/trace/trace_events_filter.c | 7 ++++++- 2 files changed, 8 insertions(+), 1 deletion(-) diff --git a/include/linux/ftrace_event.h b/include/linux/ftrace_event.h index 96efa6794ea5..c3da42dd22ba 100644 --- a/include/linux/ftrace_event.h +++ b/include/linux/ftrace_event.h @@ -172,6 +172,7 @@ enum { TRACE_EVENT_FL_FILTERED_BIT, TRACE_EVENT_FL_RECORDED_CMD_BIT, TRACE_EVENT_FL_CAP_ANY_BIT, + TRACE_EVENT_FL_NO_SET_FILTER_BIT, }; enum { @@ -179,6 +180,7 @@ enum { TRACE_EVENT_FL_FILTERED = (1 << TRACE_EVENT_FL_FILTERED_BIT), TRACE_EVENT_FL_RECORDED_CMD = (1 << TRACE_EVENT_FL_RECORDED_CMD_BIT), TRACE_EVENT_FL_CAP_ANY = (1 << TRACE_EVENT_FL_CAP_ANY_BIT), + TRACE_EVENT_FL_NO_SET_FILTER = (1 << TRACE_EVENT_FL_NO_SET_FILTER_BIT), }; struct ftrace_event_call { diff --git a/kernel/trace/trace_events_filter.c b/kernel/trace/trace_events_filter.c index 816d3d074979..86040d9c1acc 100644 --- a/kernel/trace/trace_events_filter.c +++ b/kernel/trace/trace_events_filter.c @@ -1649,7 +1649,9 @@ static int replace_system_preds(struct event_subsystem *system, */ err = replace_preds(call, NULL, ps, filter_string, true); if (err) - goto fail; + call->flags |= TRACE_EVENT_FL_NO_SET_FILTER; + else + call->flags &= ~TRACE_EVENT_FL_NO_SET_FILTER; } list_for_each_entry(call, &ftrace_events, list) { @@ -1658,6 +1660,9 @@ static int replace_system_preds(struct event_subsystem *system, if (strcmp(call->class->system, system->name) != 0) continue; + if (call->flags & TRACE_EVENT_FL_NO_SET_FILTER) + continue; + filter_item = kzalloc(sizeof(*filter_item), GFP_KERNEL); if (!filter_item) goto fail_mem; -- cgit v1.2.3 From 159a80b2142df709416ab369113de7d511c48331 Mon Sep 17 00:00:00 2001 From: Robert Richter Date: Tue, 11 Oct 2011 19:39:16 +0200 Subject: oprofile, x86: Add kernel parameter oprofile.cpu_type=timer We need this to better test x86 NMI timer mode. Otherwise it is very hard to setup systems with NMI timer enabled, but we have this e.g. in virtual machine environments. Signed-off-by: Robert Richter --- Documentation/kernel-parameters.txt | 3 +++ arch/x86/oprofile/nmi_int.c | 27 +++++++++++++++++++++------ 2 files changed, 24 insertions(+), 6 deletions(-) diff --git a/Documentation/kernel-parameters.txt b/Documentation/kernel-parameters.txt index 854ed5ca7e3f..f7735a125f4b 100644 --- a/Documentation/kernel-parameters.txt +++ b/Documentation/kernel-parameters.txt @@ -1848,6 +1848,9 @@ bytes respectively. Such letter suffixes can also be entirely omitted. arch_perfmon: [X86] Force use of architectural perfmon on Intel CPUs instead of the CPU specific event set. + timer: [X86] Force use of architectural NMI + timer mode (see also oprofile.timer + for generic hr timer mode) oops=panic Always panic on oopses. Default is to just kill the process, but there is a small probability of diff --git a/arch/x86/oprofile/nmi_int.c b/arch/x86/oprofile/nmi_int.c index 68894fdc034b..7ca4d43e8988 100644 --- a/arch/x86/oprofile/nmi_int.c +++ b/arch/x86/oprofile/nmi_int.c @@ -613,24 +613,36 @@ static int __init p4_init(char **cpu_type) return 0; } -static int force_arch_perfmon; -static int force_cpu_type(const char *str, struct kernel_param *kp) +enum __force_cpu_type { + reserved = 0, /* do not force */ + timer, + arch_perfmon, +}; + +static int force_cpu_type; + +static int set_cpu_type(const char *str, struct kernel_param *kp) { - if (!strcmp(str, "arch_perfmon")) { - force_arch_perfmon = 1; + if (!strcmp(str, "timer")) { + force_cpu_type = timer; + printk(KERN_INFO "oprofile: forcing NMI timer mode\n"); + } else if (!strcmp(str, "arch_perfmon")) { + force_cpu_type = arch_perfmon; printk(KERN_INFO "oprofile: forcing architectural perfmon\n"); + } else { + force_cpu_type = 0; } return 0; } -module_param_call(cpu_type, force_cpu_type, NULL, NULL, 0); +module_param_call(cpu_type, set_cpu_type, NULL, NULL, 0); static int __init ppro_init(char **cpu_type) { __u8 cpu_model = boot_cpu_data.x86_model; struct op_x86_model_spec *spec = &op_ppro_spec; /* default */ - if (force_arch_perfmon && cpu_has_arch_perfmon) + if (force_cpu_type == arch_perfmon && cpu_has_arch_perfmon) return 0; /* @@ -697,6 +709,9 @@ int __init op_nmi_init(struct oprofile_operations *ops) if (!cpu_has_apic) return -ENODEV; + if (force_cpu_type == timer) + return -ENODEV; + switch (vendor) { case X86_VENDOR_AMD: /* Needs to be at least an Athlon (or hammer in 32bit mode) */ -- cgit v1.2.3 From 75c43a20b220f885c39ffa7cdbbb1191e257a9a9 Mon Sep 17 00:00:00 2001 From: Robert Richter Date: Fri, 14 Oct 2011 15:46:10 +0200 Subject: oprofile: Remove exit function for timer mode Remove exit functions by moving init/exit code to oprofile's setup/ shutdown functions. Doing so the oprofile module exit code will be easier and less error-prone. Signed-off-by: Robert Richter --- drivers/oprofile/oprof.c | 8 ++------ drivers/oprofile/timer_int.c | 30 +++++++++++++++--------------- 2 files changed, 17 insertions(+), 21 deletions(-) diff --git a/drivers/oprofile/oprof.c b/drivers/oprofile/oprof.c index f8c752e408a6..f7cd06967aed 100644 --- a/drivers/oprofile/oprof.c +++ b/drivers/oprofile/oprof.c @@ -262,9 +262,7 @@ static int __init oprofile_init(void) return 0; /* failed */ - if (timer_mode) - oprofile_timer_exit(); - else + if (!timer_mode) oprofile_arch_exit(); return err; @@ -274,9 +272,7 @@ static int __init oprofile_init(void) static void __exit oprofile_exit(void) { oprofilefs_unregister(); - if (timer_mode) - oprofile_timer_exit(); - else + if (!timer_mode) oprofile_arch_exit(); } diff --git a/drivers/oprofile/timer_int.c b/drivers/oprofile/timer_int.c index 878fba126582..93404f72dfa8 100644 --- a/drivers/oprofile/timer_int.c +++ b/drivers/oprofile/timer_int.c @@ -97,24 +97,24 @@ static struct notifier_block __refdata oprofile_cpu_notifier = { .notifier_call = oprofile_cpu_notify, }; -int oprofile_timer_init(struct oprofile_operations *ops) +static int oprofile_hrtimer_setup(void) { - int rc; - - rc = register_hotcpu_notifier(&oprofile_cpu_notifier); - if (rc) - return rc; - ops->create_files = NULL; - ops->setup = NULL; - ops->shutdown = NULL; - ops->start = oprofile_hrtimer_start; - ops->stop = oprofile_hrtimer_stop; - ops->cpu_type = "timer"; - printk(KERN_INFO "oprofile: using timer interrupt.\n"); - return 0; + return register_hotcpu_notifier(&oprofile_cpu_notifier); } -void oprofile_timer_exit(void) +static void oprofile_hrtimer_shutdown(void) { unregister_hotcpu_notifier(&oprofile_cpu_notifier); } + +int oprofile_timer_init(struct oprofile_operations *ops) +{ + ops->create_files = NULL; + ops->setup = oprofile_hrtimer_setup; + ops->shutdown = oprofile_hrtimer_shutdown; + ops->start = oprofile_hrtimer_start; + ops->stop = oprofile_hrtimer_stop; + ops->cpu_type = "timer"; + printk(KERN_INFO "oprofile: using timer interrupt.\n"); + return 0; +} -- cgit v1.2.3 From dcfce4a095932e6e95d83ad982be3609947963bc Mon Sep 17 00:00:00 2001 From: Robert Richter Date: Tue, 11 Oct 2011 17:11:08 +0200 Subject: oprofile, x86: Reimplement nmi timer mode using perf event The legacy x86 nmi watchdog code was removed with the implementation of the perf based nmi watchdog. This broke Oprofile's nmi timer mode. To run nmi timer mode we relied on a continuous ticking nmi source which the nmi watchdog provided. The nmi tick was no longer available and current watchdog can not be used anymore since it runs with very long periods in the range of seconds. This patch reimplements the nmi timer mode using a perf counter nmi source. V2: * removing pr_info() * fix undefined reference to `__udivdi3' for 32 bit build * fix section mismatch of .cpuinit.data:nmi_timer_cpu_nb * removed nmi timer setup in arch/x86 * implemented function stubs for op_nmi_init/exit() * made code more readable in oprofile_init() V3: * fix architectural initialization in oprofile_init() * fix CONFIG_OPROFILE_NMI_TIMER dependencies Acked-by: Peter Zijlstra Signed-off-by: Robert Richter --- arch/Kconfig | 4 + arch/x86/oprofile/Makefile | 3 +- arch/x86/oprofile/init.c | 30 ++----- arch/x86/oprofile/nmi_timer_int.c | 66 --------------- drivers/oprofile/nmi_timer_int.c | 173 ++++++++++++++++++++++++++++++++++++++ drivers/oprofile/oprof.c | 24 +++--- drivers/oprofile/oprof.h | 9 ++ kernel/events/core.c | 2 + 8 files changed, 208 insertions(+), 103 deletions(-) delete mode 100644 arch/x86/oprofile/nmi_timer_int.c create mode 100644 drivers/oprofile/nmi_timer_int.c diff --git a/arch/Kconfig b/arch/Kconfig index 4b0669cbb3b0..2505740b81d2 100644 --- a/arch/Kconfig +++ b/arch/Kconfig @@ -30,6 +30,10 @@ config OPROFILE_EVENT_MULTIPLEX config HAVE_OPROFILE bool +config OPROFILE_NMI_TIMER + def_bool y + depends on PERF_EVENTS && HAVE_PERF_EVENTS_NMI + config KPROBES bool "Kprobes" depends on MODULES diff --git a/arch/x86/oprofile/Makefile b/arch/x86/oprofile/Makefile index 446902b2a6b6..1599f568f0e2 100644 --- a/arch/x86/oprofile/Makefile +++ b/arch/x86/oprofile/Makefile @@ -4,9 +4,8 @@ DRIVER_OBJS = $(addprefix ../../../drivers/oprofile/, \ oprof.o cpu_buffer.o buffer_sync.o \ event_buffer.o oprofile_files.o \ oprofilefs.o oprofile_stats.o \ - timer_int.o ) + timer_int.o nmi_timer_int.o ) oprofile-y := $(DRIVER_OBJS) init.o backtrace.o oprofile-$(CONFIG_X86_LOCAL_APIC) += nmi_int.o op_model_amd.o \ op_model_ppro.o op_model_p4.o -oprofile-$(CONFIG_X86_IO_APIC) += nmi_timer_int.o diff --git a/arch/x86/oprofile/init.c b/arch/x86/oprofile/init.c index f148cf652678..9e138d00ad36 100644 --- a/arch/x86/oprofile/init.c +++ b/arch/x86/oprofile/init.c @@ -16,37 +16,23 @@ * with the NMI mode driver. */ +#ifdef CONFIG_X86_LOCAL_APIC extern int op_nmi_init(struct oprofile_operations *ops); -extern int op_nmi_timer_init(struct oprofile_operations *ops); extern void op_nmi_exit(void); -extern void x86_backtrace(struct pt_regs * const regs, unsigned int depth); +#else +static int op_nmi_init(struct oprofile_operations *ops) { return -ENODEV; } +static void op_nmi_exit(void) { } +#endif -static int nmi_timer; +extern void x86_backtrace(struct pt_regs * const regs, unsigned int depth); int __init oprofile_arch_init(struct oprofile_operations *ops) { - int ret; - - ret = -ENODEV; - -#ifdef CONFIG_X86_LOCAL_APIC - ret = op_nmi_init(ops); -#endif - nmi_timer = (ret != 0); -#ifdef CONFIG_X86_IO_APIC - if (nmi_timer) - ret = op_nmi_timer_init(ops); -#endif ops->backtrace = x86_backtrace; - - return ret; + return op_nmi_init(ops); } - void oprofile_arch_exit(void) { -#ifdef CONFIG_X86_LOCAL_APIC - if (!nmi_timer) - op_nmi_exit(); -#endif + op_nmi_exit(); } diff --git a/arch/x86/oprofile/nmi_timer_int.c b/arch/x86/oprofile/nmi_timer_int.c deleted file mode 100644 index 720bf5a53c51..000000000000 --- a/arch/x86/oprofile/nmi_timer_int.c +++ /dev/null @@ -1,66 +0,0 @@ -/** - * @file nmi_timer_int.c - * - * @remark Copyright 2003 OProfile authors - * @remark Read the file COPYING - * - * @author Zwane Mwaikambo - */ - -#include -#include -#include -#include -#include -#include - -#include -#include -#include - -static int profile_timer_exceptions_notify(struct notifier_block *self, - unsigned long val, void *data) -{ - struct die_args *args = (struct die_args *)data; - int ret = NOTIFY_DONE; - - switch (val) { - case DIE_NMI: - oprofile_add_sample(args->regs, 0); - ret = NOTIFY_STOP; - break; - default: - break; - } - return ret; -} - -static struct notifier_block profile_timer_exceptions_nb = { - .notifier_call = profile_timer_exceptions_notify, - .next = NULL, - .priority = NMI_LOW_PRIOR, -}; - -static int timer_start(void) -{ - if (register_die_notifier(&profile_timer_exceptions_nb)) - return 1; - return 0; -} - - -static void timer_stop(void) -{ - unregister_die_notifier(&profile_timer_exceptions_nb); - synchronize_sched(); /* Allow already-started NMIs to complete. */ -} - - -int __init op_nmi_timer_init(struct oprofile_operations *ops) -{ - ops->start = timer_start; - ops->stop = timer_stop; - ops->cpu_type = "timer"; - printk(KERN_INFO "oprofile: using NMI timer interrupt.\n"); - return 0; -} diff --git a/drivers/oprofile/nmi_timer_int.c b/drivers/oprofile/nmi_timer_int.c new file mode 100644 index 000000000000..76f1c9357f39 --- /dev/null +++ b/drivers/oprofile/nmi_timer_int.c @@ -0,0 +1,173 @@ +/** + * @file nmi_timer_int.c + * + * @remark Copyright 2011 Advanced Micro Devices, Inc. + * + * @author Robert Richter + */ + +#include +#include +#include +#include +#include + +#ifdef CONFIG_OPROFILE_NMI_TIMER + +static DEFINE_PER_CPU(struct perf_event *, nmi_timer_events); +static int ctr_running; + +static struct perf_event_attr nmi_timer_attr = { + .type = PERF_TYPE_HARDWARE, + .config = PERF_COUNT_HW_CPU_CYCLES, + .size = sizeof(struct perf_event_attr), + .pinned = 1, + .disabled = 1, +}; + +static void nmi_timer_callback(struct perf_event *event, + struct perf_sample_data *data, + struct pt_regs *regs) +{ + event->hw.interrupts = 0; /* don't throttle interrupts */ + oprofile_add_sample(regs, 0); +} + +static int nmi_timer_start_cpu(int cpu) +{ + struct perf_event *event = per_cpu(nmi_timer_events, cpu); + + if (!event) { + event = perf_event_create_kernel_counter(&nmi_timer_attr, cpu, NULL, + nmi_timer_callback, NULL); + if (IS_ERR(event)) + return PTR_ERR(event); + per_cpu(nmi_timer_events, cpu) = event; + } + + if (event && ctr_running) + perf_event_enable(event); + + return 0; +} + +static void nmi_timer_stop_cpu(int cpu) +{ + struct perf_event *event = per_cpu(nmi_timer_events, cpu); + + if (event && ctr_running) + perf_event_disable(event); +} + +static int nmi_timer_cpu_notifier(struct notifier_block *b, unsigned long action, + void *data) +{ + int cpu = (unsigned long)data; + switch (action) { + case CPU_DOWN_FAILED: + case CPU_ONLINE: + nmi_timer_start_cpu(cpu); + break; + case CPU_DOWN_PREPARE: + nmi_timer_stop_cpu(cpu); + break; + } + return NOTIFY_DONE; +} + +static struct notifier_block nmi_timer_cpu_nb = { + .notifier_call = nmi_timer_cpu_notifier +}; + +static int nmi_timer_start(void) +{ + int cpu; + + get_online_cpus(); + ctr_running = 1; + for_each_online_cpu(cpu) + nmi_timer_start_cpu(cpu); + put_online_cpus(); + + return 0; +} + +static void nmi_timer_stop(void) +{ + int cpu; + + get_online_cpus(); + for_each_online_cpu(cpu) + nmi_timer_stop_cpu(cpu); + ctr_running = 0; + put_online_cpus(); +} + +static void nmi_timer_shutdown(void) +{ + struct perf_event *event; + int cpu; + + get_online_cpus(); + unregister_cpu_notifier(&nmi_timer_cpu_nb); + for_each_possible_cpu(cpu) { + event = per_cpu(nmi_timer_events, cpu); + if (!event) + continue; + perf_event_disable(event); + per_cpu(nmi_timer_events, cpu) = NULL; + perf_event_release_kernel(event); + } + + put_online_cpus(); +} + +static int nmi_timer_setup(void) +{ + int cpu, err; + u64 period; + + /* clock cycles per tick: */ + period = (u64)cpu_khz * 1000; + do_div(period, HZ); + nmi_timer_attr.sample_period = period; + + get_online_cpus(); + err = register_cpu_notifier(&nmi_timer_cpu_nb); + if (err) + goto out; + /* can't attach events to offline cpus: */ + for_each_online_cpu(cpu) { + err = nmi_timer_start_cpu(cpu); + if (err) + break; + } + if (err) + nmi_timer_shutdown(); +out: + put_online_cpus(); + return err; +} + +int __init op_nmi_timer_init(struct oprofile_operations *ops) +{ + int err = 0; + + err = nmi_timer_setup(); + if (err) + return err; + nmi_timer_shutdown(); /* only check, don't alloc */ + + ops->create_files = NULL; + ops->setup = nmi_timer_setup; + ops->shutdown = nmi_timer_shutdown; + ops->start = nmi_timer_start; + ops->stop = nmi_timer_stop; + ops->cpu_type = "timer"; + + printk(KERN_INFO "oprofile: using NMI timer interrupt.\n"); + + return 0; +} + +#endif diff --git a/drivers/oprofile/oprof.c b/drivers/oprofile/oprof.c index f7cd06967aed..ed2c3ec07024 100644 --- a/drivers/oprofile/oprof.c +++ b/drivers/oprofile/oprof.c @@ -246,26 +246,24 @@ static int __init oprofile_init(void) int err; /* always init architecture to setup backtrace support */ + timer_mode = 0; err = oprofile_arch_init(&oprofile_ops); + if (!err) { + if (!timer && !oprofilefs_register()) + return 0; + oprofile_arch_exit(); + } - timer_mode = err || timer; /* fall back to timer mode on errors */ - if (timer_mode) { - if (!err) - oprofile_arch_exit(); + /* setup timer mode: */ + timer_mode = 1; + /* no nmi timer mode if oprofile.timer is set */ + if (timer || op_nmi_timer_init(&oprofile_ops)) { err = oprofile_timer_init(&oprofile_ops); if (err) return err; } - err = oprofilefs_register(); - if (!err) - return 0; - - /* failed */ - if (!timer_mode) - oprofile_arch_exit(); - - return err; + return oprofilefs_register(); } diff --git a/drivers/oprofile/oprof.h b/drivers/oprofile/oprof.h index 177b73de5e5f..769fb0fcac44 100644 --- a/drivers/oprofile/oprof.h +++ b/drivers/oprofile/oprof.h @@ -36,6 +36,15 @@ struct dentry; void oprofile_create_files(struct super_block *sb, struct dentry *root); int oprofile_timer_init(struct oprofile_operations *ops); void oprofile_timer_exit(void); +#ifdef CONFIG_OPROFILE_NMI_TIMER +int op_nmi_timer_init(struct oprofile_operations *ops); +#else +static inline int op_nmi_timer_init(struct oprofile_operations *ops) +{ + return -ENODEV; +} +#endif + int oprofile_set_ulong(unsigned long *addr, unsigned long val); int oprofile_set_timeout(unsigned long time); diff --git a/kernel/events/core.c b/kernel/events/core.c index d1a1bee35228..d2e28bdd523a 100644 --- a/kernel/events/core.c +++ b/kernel/events/core.c @@ -1322,6 +1322,7 @@ retry: } raw_spin_unlock_irq(&ctx->lock); } +EXPORT_SYMBOL_GPL(perf_event_disable); static void perf_set_shadow_time(struct perf_event *event, struct perf_event_context *ctx, @@ -1806,6 +1807,7 @@ retry: out: raw_spin_unlock_irq(&ctx->lock); } +EXPORT_SYMBOL_GPL(perf_event_enable); int perf_event_refresh(struct perf_event *event, int refresh) { -- cgit v1.2.3 From 49aa29513ec995f201664cf6eee36e5326ed38bf Mon Sep 17 00:00:00 2001 From: Steven Rostedt Date: Wed, 2 Nov 2011 16:46:46 -0400 Subject: tracing: Add boiler plate for subsystem filter The system filter can be used to set multiple event filters that exist within the system. But currently it displays the last filter written that does not necessarily correspond to the filters within the system. The system filter itself is not used to filter any events. The system filter is just a means to set filters of the events within it. Because this causes an ambiguous state when the system filter reads a filter string but the events within the system have different strings it is best to just show a boiler plate: ### global filter ### # Use this to set filters for multiple events. # Only events with the given fields will be affected. # If no events are modified, an error message will be displayed here. If an error occurs while writing to the system filter, the system filter will replace the boiler plate with the error message as it currently does. Signed-off-by: Steven Rostedt --- kernel/trace/trace_events_filter.c | 26 +++++++++++++++++++------- 1 file changed, 19 insertions(+), 7 deletions(-) diff --git a/kernel/trace/trace_events_filter.c b/kernel/trace/trace_events_filter.c index 86040d9c1acc..fdc6d22d406b 100644 --- a/kernel/trace/trace_events_filter.c +++ b/kernel/trace/trace_events_filter.c @@ -27,6 +27,12 @@ #include "trace.h" #include "trace_output.h" +#define DEFAULT_SYS_FILTER_MESSAGE \ + "### global filter ###\n" \ + "# Use this to set filters for multiple events.\n" \ + "# Only events with the given fields will be affected.\n" \ + "# If no events are modified, an error message will be displayed here" + enum filter_op_ids { OP_OR, @@ -646,7 +652,7 @@ void print_subsystem_event_filter(struct event_subsystem *system, if (filter && filter->filter_string) trace_seq_printf(s, "%s\n", filter->filter_string); else - trace_seq_printf(s, "none\n"); + trace_seq_printf(s, DEFAULT_SYS_FILTER_MESSAGE "\n"); mutex_unlock(&event_mutex); } @@ -1838,7 +1844,10 @@ int apply_subsystem_event_filter(struct event_subsystem *system, if (!filter) goto out; - replace_filter_string(filter, filter_string); + /* System filters just show a default message */ + kfree(filter->filter_string); + filter->filter_string = NULL; + /* * No event actually uses the system filter * we can free it without synchronize_sched(). @@ -1848,14 +1857,12 @@ int apply_subsystem_event_filter(struct event_subsystem *system, parse_init(ps, filter_ops, filter_string); err = filter_parse(ps); - if (err) { - append_filter_err(ps, system->filter); - goto out; - } + if (err) + goto err_filter; err = replace_system_preds(system, ps, filter_string); if (err) - append_filter_err(ps, system->filter); + goto err_filter; out: filter_opstack_clear(ps); @@ -1865,6 +1872,11 @@ out_unlock: mutex_unlock(&event_mutex); return err; + +err_filter: + replace_filter_string(filter, filter_string); + append_filter_err(ps, system->filter); + goto out; } #ifdef CONFIG_PERF_EVENTS -- cgit v1.2.3 From 3890c136357284cb0656f9dd0e62286995ad32e9 Mon Sep 17 00:00:00 2001 From: Li Zefan Date: Fri, 28 Oct 2011 15:48:36 +0800 Subject: tracing: update Documentation on max preds limit Preds is no longer stored in a fixed-size stack. The max preds now is 2^14, and is way more than enough, so don't bother to document it. Link: http://lkml.kernel.org/r/4EAA5E54.1080102@cn.fujitsu.com Signed-off-by: Li Zefan Signed-off-by: Steven Rostedt --- Documentation/trace/events.txt | 2 -- 1 file changed, 2 deletions(-) diff --git a/Documentation/trace/events.txt b/Documentation/trace/events.txt index b510564aac7e..bb24c2a0e870 100644 --- a/Documentation/trace/events.txt +++ b/Documentation/trace/events.txt @@ -191,8 +191,6 @@ And for string fields they are: Currently, only exact string matches are supported. -Currently, the maximum number of predicates in a filter is 16. - 5.2 Setting filters ------------------- -- cgit v1.2.3 From e5e78d08f3ab3094783b8df08a5b6d1d1a56a58f Mon Sep 17 00:00:00 2001 From: Steven Rostedt Date: Wed, 2 Nov 2011 20:24:16 -0400 Subject: lockdep: Show subclass in pretty print of lockdep output The pretty print of the lockdep debug splat uses just the lock name to show how the locking scenario happens. But when it comes to nesting locks, the output becomes confusing which takes away the point of the pretty printing of the lock scenario. Without displaying the subclass info, we get the following output: Possible unsafe locking scenario: CPU0 CPU1 ---- ---- lock(slock-AF_INET); lock(slock-AF_INET); lock(slock-AF_INET); lock(slock-AF_INET); *** DEADLOCK *** The above looks more of a A->A locking bug than a A->B B->A. By adding the subclass to the output, we can see what really happened: other info that might help us debug this: Possible unsafe locking scenario: CPU0 CPU1 ---- ---- lock(slock-AF_INET); lock(slock-AF_INET/1); lock(slock-AF_INET); lock(slock-AF_INET/1); *** DEADLOCK *** This bug was discovered while tracking down a real bug caught by lockdep. Link: http://lkml.kernel.org/r/20111025202049.GB25043@hostway.ca Cc: Peter Zijlstra Reported-by: Thomas Gleixner Tested-by: Simon Kirby Signed-off-by: Steven Rostedt --- kernel/lockdep.c | 30 +++++++++++++----------------- 1 file changed, 13 insertions(+), 17 deletions(-) diff --git a/kernel/lockdep.c b/kernel/lockdep.c index 91d67ce3a8d5..6bd915df5fd3 100644 --- a/kernel/lockdep.c +++ b/kernel/lockdep.c @@ -490,36 +490,32 @@ void get_usage_chars(struct lock_class *class, char usage[LOCK_USAGE_CHARS]) usage[i] = '\0'; } -static int __print_lock_name(struct lock_class *class) +static void __print_lock_name(struct lock_class *class) { char str[KSYM_NAME_LEN]; const char *name; - name = class->name; - if (!name) - name = __get_key_name(class->key, str); - - return printk("%s", name); -} - -static void print_lock_name(struct lock_class *class) -{ - char str[KSYM_NAME_LEN], usage[LOCK_USAGE_CHARS]; - const char *name; - - get_usage_chars(class, usage); - name = class->name; if (!name) { name = __get_key_name(class->key, str); - printk(" (%s", name); + printk("%s", name); } else { - printk(" (%s", name); + printk("%s", name); if (class->name_version > 1) printk("#%d", class->name_version); if (class->subclass) printk("/%d", class->subclass); } +} + +static void print_lock_name(struct lock_class *class) +{ + char usage[LOCK_USAGE_CHARS]; + + get_usage_chars(class, usage); + + printk(" ("); + __print_lock_name(class); printk("){%s}", usage); } -- cgit v1.2.3 From 8ee3c92b7f2751c392be2d8fc360a410480b8757 Mon Sep 17 00:00:00 2001 From: Steven Rostedt Date: Fri, 4 Nov 2011 10:45:23 -0400 Subject: ftrace: Remove force undef config value left for testing A forced undef of a config value was used for testing and was accidently left in during the final commit. This causes x86 to run slower than needed while running function tracing as well as causes the function graph selftest to fail when DYNMAIC_FTRACE is not set. This is because the code in MCOUNT expects the ftrace code to be processed with the config value set that happened to be forced not set. The forced config option was left in by: commit 6331c28c962561aee59e5a493b7556a4bb585957 ftrace: Fix dynamic selftest failure on some archs Link: http://lkml.kernel.org/r/20111102150255.GA6973@debian Cc: stable@vger.kernel.org Reported-by: Rabin Vincent Signed-off-by: Steven Rostedt --- kernel/trace/ftrace.c | 1 - 1 file changed, 1 deletion(-) diff --git a/kernel/trace/ftrace.c b/kernel/trace/ftrace.c index 077d85387908..0fcc6caead1c 100644 --- a/kernel/trace/ftrace.c +++ b/kernel/trace/ftrace.c @@ -151,7 +151,6 @@ void clear_ftrace_function(void) ftrace_pid_function = ftrace_stub; } -#undef CONFIG_HAVE_FUNCTION_TRACE_MCOUNT_TEST #ifndef CONFIG_HAVE_FUNCTION_TRACE_MCOUNT_TEST /* * For those archs that do not test ftrace_trace_stop in their -- cgit v1.2.3 From c8452afb7426f7e21388492f40227582e3e83879 Mon Sep 17 00:00:00 2001 From: Gleb Natapov Date: Tue, 18 Oct 2011 19:55:51 +0200 Subject: jump_label: jump_label_inc may return before the code is patched If cpu A calls jump_label_inc() just after atomic_add_return() is called by cpu B, atomic_inc_not_zero() will return value greater then zero and jump_label_inc() will return to a caller before jump_label_update() finishes its job on cpu B. Link: http://lkml.kernel.org/r/20111018175551.GH17571@redhat.com Cc: stable@vger.kernel.org Cc: Peter Zijlstra Acked-by: Jason Baron Signed-off-by: Gleb Natapov Signed-off-by: Steven Rostedt --- kernel/jump_label.c | 3 ++- 1 file changed, 2 insertions(+), 1 deletion(-) diff --git a/kernel/jump_label.c b/kernel/jump_label.c index a8ce45097f3d..e6f1f24ad577 100644 --- a/kernel/jump_label.c +++ b/kernel/jump_label.c @@ -66,8 +66,9 @@ void jump_label_inc(struct jump_label_key *key) return; jump_label_lock(); - if (atomic_add_return(1, &key->enabled) == 1) + if (atomic_read(&key->enabled) == 0) jump_label_update(key, JUMP_LABEL_ENABLE); + atomic_inc(&key->enabled); jump_label_unlock(); } -- cgit v1.2.3 From 49908a1b25d448d68fd26faca260e1850201575f Mon Sep 17 00:00:00 2001 From: Steven Rostedt Date: Fri, 4 Nov 2011 16:32:25 -0400 Subject: perf: Fix parsing of __print_flags() in TP_printk() A update is made to the sched:sched_switch event that adds some logic to the first parameter of the __print_flags() that shows the state of tasks. This change cause perf to fail parsing the flags. A simple fix is needed to have the parser be able to process ops within the argument. Cc: stable@vger.kernel.org Reported-by: Andrew Vagin Signed-off-by: Steven Rostedt --- tools/perf/util/trace-event-parse.c | 2 ++ 1 file changed, 2 insertions(+) diff --git a/tools/perf/util/trace-event-parse.c b/tools/perf/util/trace-event-parse.c index 0a7ed5b5e281..6c164dc9ee95 100644 --- a/tools/perf/util/trace-event-parse.c +++ b/tools/perf/util/trace-event-parse.c @@ -1537,6 +1537,8 @@ process_flags(struct event *event, struct print_arg *arg, char **tok) field = malloc_or_die(sizeof(*field)); type = process_arg(event, field, &token); + while (type == EVENT_OP) + type = process_op(event, field, &token); if (test_type_token(type, token, EVENT_DELIM, ",")) goto out_free; -- cgit v1.2.3 From d4d34b981a5327eec956c6cb4cce397ce6f57279 Mon Sep 17 00:00:00 2001 From: Steven Rostedt Date: Fri, 4 Nov 2011 20:32:39 -0400 Subject: ftrace: Fix hash record accounting bug If the set_ftrace_filter is cleared by writing just whitespace to it, then the filter hash refcounts will be decremented but not updated. This causes two bugs: 1) No functions will be enabled for tracing when they all should be 2) If the users clears the set_ftrace_filter twice, it will crash ftrace: ------------[ cut here ]------------ WARNING: at /home/rostedt/work/git/linux-trace.git/kernel/trace/ftrace.c:1384 __ftrace_hash_rec_update.part.27+0x157/0x1a7() Modules linked in: Pid: 2330, comm: bash Not tainted 3.1.0-test+ #32 Call Trace: [] warn_slowpath_common+0x83/0x9b [] warn_slowpath_null+0x1a/0x1c [] __ftrace_hash_rec_update.part.27+0x157/0x1a7 [] ? ftrace_regex_release+0xa7/0x10f [] ? kfree+0xe5/0x115 [] ftrace_hash_move+0x2e/0x151 [] ftrace_regex_release+0xba/0x10f [] fput+0xfd/0x1c2 [] filp_close+0x6d/0x78 [] sys_dup3+0x197/0x1c1 [] sys_dup2+0x4f/0x54 [] system_call_fastpath+0x16/0x1b ---[ end trace 77a3a7ee73794a02 ]--- Link: http://lkml.kernel.org/r/20111101141420.GA4918@debian Reported-by: Rabin Vincent Signed-off-by: Steven Rostedt --- kernel/trace/ftrace.c | 4 +++- 1 file changed, 3 insertions(+), 1 deletion(-) diff --git a/kernel/trace/ftrace.c b/kernel/trace/ftrace.c index 0fcc6caead1c..7caa4505508d 100644 --- a/kernel/trace/ftrace.c +++ b/kernel/trace/ftrace.c @@ -1210,7 +1210,9 @@ ftrace_hash_move(struct ftrace_ops *ops, int enable, if (!src->count) { free_ftrace_hash_rcu(*dst); rcu_assign_pointer(*dst, EMPTY_HASH); - return 0; + /* still need to update the function records */ + ret = 0; + goto out; } /* -- cgit v1.2.3 From 7e9a49ef542610609144d1afcd516dc3fafac4d6 Mon Sep 17 00:00:00 2001 From: Jiri Olsa Date: Mon, 7 Nov 2011 16:08:49 +0100 Subject: tracing/latency: Fix header output for latency tracers In case the the graph tracer (CONFIG_FUNCTION_GRAPH_TRACER) or even the function tracer (CONFIG_FUNCTION_TRACER) are not set, the latency tracers do not display proper latency header. The involved/fixed latency tracers are: wakeup_rt wakeup preemptirqsoff preemptoff irqsoff The patch adds proper handling of tracer configuration options for latency tracers, and displaying correct header info accordingly. * The current output (for wakeup tracer) with both graph and function tracers disabled is: # tracer: wakeup # -0 0d.h5 1us+: 0:120:R + [000] 7: 0:R watchdog/0 -0 0d.h5 3us+: ttwu_do_activate.clone.1 <-try_to_wake_up ... * The fixed output is: # tracer: wakeup # # wakeup latency trace v1.1.5 on 3.1.0-tip+ # -------------------------------------------------------------------- # latency: 55 us, #4/4, CPU#0 | (M:preempt VP:0, KP:0, SP:0 HP:0 #P:2) # ----------------- # | task: migration/0-6 (uid:0 nice:0 policy:1 rt_prio:99) # ----------------- # # _------=> CPU# # / _-----=> irqs-off # | / _----=> need-resched # || / _---=> hardirq/softirq # ||| / _--=> preempt-depth # |||| / delay # cmd pid ||||| time | caller # \ / ||||| \ | / cat-1129 0d..4 1us : 1129:120:R + [000] 6: 0:R migration/0 cat-1129 0d..4 2us+: ttwu_do_activate.clone.1 <-try_to_wake_up * The current output (for wakeup tracer) with only function tracer enabled is: # tracer: wakeup # cat-1140 0d..4 1us+: 1140:120:R + [000] 6: 0:R migration/0 cat-1140 0d..4 2us : ttwu_do_activate.clone.1 <-try_to_wake_up * The fixed output is: # tracer: wakeup # # wakeup latency trace v1.1.5 on 3.1.0-tip+ # -------------------------------------------------------------------- # latency: 207 us, #109/109, CPU#1 | (M:preempt VP:0, KP:0, SP:0 HP:0 #P:2) # ----------------- # | task: watchdog/1-12 (uid:0 nice:0 policy:1 rt_prio:99) # ----------------- # # _------=> CPU# # / _-----=> irqs-off # | / _----=> need-resched # || / _---=> hardirq/softirq # ||| / _--=> preempt-depth # |||| / delay # cmd pid ||||| time | caller # \ / ||||| \ | / -0 1d.h5 1us+: 0:120:R + [001] 12: 0:R watchdog/1 -0 1d.h5 3us : ttwu_do_activate.clone.1 <-try_to_wake_up Link: http://lkml.kernel.org/r/20111107150849.GE1807@m.brq.redhat.com Cc: Frederic Weisbecker Cc: Ingo Molnar Signed-off-by: Jiri Olsa Signed-off-by: Steven Rostedt --- kernel/trace/trace.c | 15 +++++++++++++++ kernel/trace/trace.h | 1 + kernel/trace/trace_irqsoff.c | 13 ++++++++++++- kernel/trace/trace_sched_wakeup.c | 13 ++++++++++++- 4 files changed, 40 insertions(+), 2 deletions(-) diff --git a/kernel/trace/trace.c b/kernel/trace/trace.c index b24a72d35008..b296186eb93a 100644 --- a/kernel/trace/trace.c +++ b/kernel/trace/trace.c @@ -2140,6 +2140,21 @@ enum print_line_t print_trace_line(struct trace_iterator *iter) return print_trace_fmt(iter); } +void trace_latency_header(struct seq_file *m) +{ + struct trace_iterator *iter = m->private; + + /* print nothing if the buffers are empty */ + if (trace_empty(iter)) + return; + + if (iter->iter_flags & TRACE_FILE_LAT_FMT) + print_trace_header(m, iter); + + if (!(trace_flags & TRACE_ITER_VERBOSE)) + print_lat_help_header(m); +} + void trace_default_header(struct seq_file *m) { struct trace_iterator *iter = m->private; diff --git a/kernel/trace/trace.h b/kernel/trace/trace.h index 092e1f8d18dc..f8ec2291b522 100644 --- a/kernel/trace/trace.h +++ b/kernel/trace/trace.h @@ -370,6 +370,7 @@ void trace_graph_function(struct trace_array *tr, unsigned long ip, unsigned long parent_ip, unsigned long flags, int pc); +void trace_latency_header(struct seq_file *m); void trace_default_header(struct seq_file *m); void print_trace_header(struct seq_file *m, struct trace_iterator *iter); int trace_empty(struct trace_iterator *iter); diff --git a/kernel/trace/trace_irqsoff.c b/kernel/trace/trace_irqsoff.c index a1a3359996a7..a248c686b2b8 100644 --- a/kernel/trace/trace_irqsoff.c +++ b/kernel/trace/trace_irqsoff.c @@ -280,9 +280,20 @@ static enum print_line_t irqsoff_print_line(struct trace_iterator *iter) } static void irqsoff_graph_return(struct ftrace_graph_ret *trace) { } -static void irqsoff_print_header(struct seq_file *s) { } static void irqsoff_trace_open(struct trace_iterator *iter) { } static void irqsoff_trace_close(struct trace_iterator *iter) { } + +#ifdef CONFIG_FUNCTION_TRACER +static void irqsoff_print_header(struct seq_file *s) +{ + trace_default_header(s); +} +#else +static void irqsoff_print_header(struct seq_file *s) +{ + trace_latency_header(s); +} +#endif /* CONFIG_FUNCTION_TRACER */ #endif /* CONFIG_FUNCTION_GRAPH_TRACER */ /* diff --git a/kernel/trace/trace_sched_wakeup.c b/kernel/trace/trace_sched_wakeup.c index e4a70c0c71b6..ff791ea48b57 100644 --- a/kernel/trace/trace_sched_wakeup.c +++ b/kernel/trace/trace_sched_wakeup.c @@ -280,9 +280,20 @@ static enum print_line_t wakeup_print_line(struct trace_iterator *iter) } static void wakeup_graph_return(struct ftrace_graph_ret *trace) { } -static void wakeup_print_header(struct seq_file *s) { } static void wakeup_trace_open(struct trace_iterator *iter) { } static void wakeup_trace_close(struct trace_iterator *iter) { } + +#ifdef CONFIG_FUNCTION_TRACER +static void wakeup_print_header(struct seq_file *s) +{ + trace_default_header(s); +} +#else +static void wakeup_print_header(struct seq_file *s) +{ + trace_latency_header(s); +} +#endif /* CONFIG_FUNCTION_TRACER */ #endif /* CONFIG_FUNCTION_GRAPH_TRACER */ /* -- cgit v1.2.3 From 1ec454baf1245df4fdb5dae728da3363630ce6de Mon Sep 17 00:00:00 2001 From: Masami Hiramatsu Date: Thu, 20 Oct 2011 23:01:09 +0900 Subject: x86, perf: Add a build-time sanity test to the x86 decoder Add a sanity test of x86 insn decoder against a stream of randomly generated input, at build time. This test is also able to reproduce any bug that might trigger by allowing the passing of random-seed and iteration-number to the test, or by passing input which has invalid byte code. Changes in V2: - Code cleanup. - Show how to reproduce the error by insn_sanity test. Signed-off-by: Masami Hiramatsu Cc: acme@redhat.com Cc: ming.m.lin@intel.com Cc: robert.richter@amd.com Cc: ravitillo@lbl.gov Cc: yrl.pp-manager.tt@hitachi.com Cc: Andi Kleen Cc: Peter Zijlstra Cc: Stephane Eranian Cc: Andi Kleen Cc: Peter Zijlstra Link: http://lkml.kernel.org/r/20111020140109.20938.92572.stgit@localhost.localdomain Signed-off-by: Ingo Molnar --- arch/x86/include/asm/insn.h | 7 ++ arch/x86/tools/Makefile | 10 +- arch/x86/tools/insn_sanity.c | 275 +++++++++++++++++++++++++++++++++++++++++++ 3 files changed, 291 insertions(+), 1 deletion(-) create mode 100644 arch/x86/tools/insn_sanity.c diff --git a/arch/x86/include/asm/insn.h b/arch/x86/include/asm/insn.h index 88c765e16410..74df3f1eddfd 100644 --- a/arch/x86/include/asm/insn.h +++ b/arch/x86/include/asm/insn.h @@ -137,6 +137,13 @@ static inline int insn_is_avx(struct insn *insn) return (insn->vex_prefix.value != 0); } +/* Ensure this instruction is decoded completely */ +static inline int insn_complete(struct insn *insn) +{ + return insn->opcode.got && insn->modrm.got && insn->sib.got && + insn->displacement.got && insn->immediate.got; +} + static inline insn_byte_t insn_vex_m_bits(struct insn *insn) { if (insn->vex_prefix.nbytes == 2) /* 2 bytes VEX */ diff --git a/arch/x86/tools/Makefile b/arch/x86/tools/Makefile index f82082677337..3255c3df67f4 100644 --- a/arch/x86/tools/Makefile +++ b/arch/x86/tools/Makefile @@ -18,14 +18,22 @@ chkobjdump = $(srctree)/arch/x86/tools/chkobjdump.awk quiet_cmd_posttest = TEST $@ cmd_posttest = ($(OBJDUMP) -v | $(AWK) -f $(chkobjdump)) || $(OBJDUMP) -d -j .text $(objtree)/vmlinux | $(AWK) -f $(distill_awk) | $(obj)/test_get_len $(posttest_64bit) $(posttest_verbose) -posttest: $(obj)/test_get_len vmlinux +quiet_cmd_sanitytest = TEST $@ + cmd_sanitytest = $(obj)/insn_sanity $(posttest_64bit) -m 1000000 + +posttest: $(obj)/test_get_len vmlinux $(obj)/insn_sanity $(call cmd,posttest) + $(call cmd,sanitytest) hostprogs-y := test_get_len +hostprogs-y := insn_sanity # -I needed for generated C source and C source which in the kernel tree. HOSTCFLAGS_test_get_len.o := -Wall -I$(objtree)/arch/x86/lib/ -I$(srctree)/arch/x86/include/ -I$(srctree)/arch/x86/lib/ -I$(srctree)/include/ +HOSTCFLAGS_insn_sanity.o := -Wall -I$(objtree)/arch/x86/lib/ -I$(srctree)/arch/x86/include/ -I$(srctree)/arch/x86/lib/ -I$(srctree)/include/ + # Dependencies are also needed. $(obj)/test_get_len.o: $(srctree)/arch/x86/lib/insn.c $(srctree)/arch/x86/lib/inat.c $(srctree)/arch/x86/include/asm/inat_types.h $(srctree)/arch/x86/include/asm/inat.h $(srctree)/arch/x86/include/asm/insn.h $(objtree)/arch/x86/lib/inat-tables.c +$(obj)/insn_sanity.o: $(srctree)/arch/x86/lib/insn.c $(srctree)/arch/x86/lib/inat.c $(srctree)/arch/x86/include/asm/inat_types.h $(srctree)/arch/x86/include/asm/inat.h $(srctree)/arch/x86/include/asm/insn.h $(objtree)/arch/x86/lib/inat-tables.c diff --git a/arch/x86/tools/insn_sanity.c b/arch/x86/tools/insn_sanity.c new file mode 100644 index 000000000000..334d9de7d0ca --- /dev/null +++ b/arch/x86/tools/insn_sanity.c @@ -0,0 +1,275 @@ +/* + * x86 decoder sanity test - based on test_get_insn.c + * + * This program is free software; you can redistribute it and/or modify + * it under the terms of the GNU General Public License as published by + * the Free Software Foundation; either version 2 of the License, or + * (at your option) any later version. + * + * This program is distributed in the hope that it will be useful, + * but WITHOUT ANY WARRANTY; without even the implied warranty of + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the + * GNU General Public License for more details. + * + * You should have received a copy of the GNU General Public License + * along with this program; if not, write to the Free Software + * Foundation, Inc., 59 Temple Place - Suite 330, Boston, MA 02111-1307, USA. + * + * Copyright (C) IBM Corporation, 2009 + * Copyright (C) Hitachi, Ltd., 2011 + */ + +#include +#include +#include +#include +#include +#include +#include +#include + +#define unlikely(cond) (cond) +#define ARRAY_SIZE(a) (sizeof(a)/sizeof(a[0])) + +#include +#include +#include + +/* + * Test of instruction analysis against tampering. + * Feed random binary to instruction decoder and ensure not to + * access out-of-instruction-buffer. + */ + +#define DEFAULT_MAX_ITER 10000 +#define INSN_NOP 0x90 + +static const char *prog; /* Program name */ +static int verbose; /* Verbosity */ +static int x86_64; /* x86-64 bit mode flag */ +static unsigned int seed; /* Random seed */ +static unsigned long iter_start; /* Start of iteration number */ +static unsigned long iter_end = DEFAULT_MAX_ITER; /* End of iteration number */ +static FILE *input_file; /* Input file name */ + +static void usage(const char *err) +{ + if (err) + fprintf(stderr, "Error: %s\n\n", err); + fprintf(stderr, "Usage: %s [-y|-n|-v] [-s seed[,no]] [-m max] [-i input]\n", prog); + fprintf(stderr, "\t-y 64bit mode\n"); + fprintf(stderr, "\t-n 32bit mode\n"); + fprintf(stderr, "\t-v Verbose mode\n"); + fprintf(stderr, "\t-s Give a random seed (and iteration number)\n"); + fprintf(stderr, "\t-m Give a maximum iteration number\n"); + fprintf(stderr, "\t-i Give an input file with decoded binary\n"); + exit(1); +} + +static void dump_field(FILE *fp, const char *name, const char *indent, + struct insn_field *field) +{ + fprintf(fp, "%s.%s = {\n", indent, name); + fprintf(fp, "%s\t.value = %d, bytes[] = {%x, %x, %x, %x},\n", + indent, field->value, field->bytes[0], field->bytes[1], + field->bytes[2], field->bytes[3]); + fprintf(fp, "%s\t.got = %d, .nbytes = %d},\n", indent, + field->got, field->nbytes); +} + +static void dump_insn(FILE *fp, struct insn *insn) +{ + fprintf(fp, "Instruction = {\n"); + dump_field(fp, "prefixes", "\t", &insn->prefixes); + dump_field(fp, "rex_prefix", "\t", &insn->rex_prefix); + dump_field(fp, "vex_prefix", "\t", &insn->vex_prefix); + dump_field(fp, "opcode", "\t", &insn->opcode); + dump_field(fp, "modrm", "\t", &insn->modrm); + dump_field(fp, "sib", "\t", &insn->sib); + dump_field(fp, "displacement", "\t", &insn->displacement); + dump_field(fp, "immediate1", "\t", &insn->immediate1); + dump_field(fp, "immediate2", "\t", &insn->immediate2); + fprintf(fp, "\t.attr = %x, .opnd_bytes = %d, .addr_bytes = %d,\n", + insn->attr, insn->opnd_bytes, insn->addr_bytes); + fprintf(fp, "\t.length = %d, .x86_64 = %d, .kaddr = %p}\n", + insn->length, insn->x86_64, insn->kaddr); +} + +static void dump_stream(FILE *fp, const char *msg, unsigned long nr_iter, + unsigned char *insn_buf, struct insn *insn) +{ + int i; + + fprintf(fp, "%s:\n", msg); + + dump_insn(stderr, insn); + + fprintf(fp, "You can reproduce this with below command(s);\n"); + + /* Input a decoded instruction sequence directly */ + fprintf(fp, " $ echo "); + for (i = 0; i < MAX_INSN_SIZE; i++) + fprintf(fp, " %02x", insn_buf[i]); + fprintf(fp, " | %s -i -\n", prog); + + if (!input_file) { + fprintf(fp, "Or \n"); + /* Give a seed and iteration number */ + fprintf(fp, " $ %s -s 0x%x,%lu\n", prog, seed, nr_iter); + } +} + +static void init_random_seed(void) +{ + int fd; + + fd = open("/dev/urandom", O_RDONLY); + if (fd < 0) + goto fail; + + if (read(fd, &seed, sizeof(seed)) != sizeof(seed)) + goto fail; + + close(fd); + return; +fail: + usage("Failed to open /dev/urandom"); +} + +/* Read given instruction sequence from the input file */ +static int read_next_insn(unsigned char *insn_buf) +{ + char buf[256] = "", *tmp; + int i; + + tmp = fgets(buf, ARRAY_SIZE(buf), input_file); + if (tmp == NULL || feof(input_file)) + return 0; + + for (i = 0; i < MAX_INSN_SIZE; i++) { + insn_buf[i] = (unsigned char)strtoul(tmp, &tmp, 16); + if (*tmp != ' ') + break; + } + + return i; +} + +static int generate_insn(unsigned char *insn_buf) +{ + int i; + + if (input_file) + return read_next_insn(insn_buf); + + /* Fills buffer with random binary up to MAX_INSN_SIZE */ + for (i = 0; i < MAX_INSN_SIZE - 1; i += 2) + *(unsigned short *)(&insn_buf[i]) = random() & 0xffff; + + while (i < MAX_INSN_SIZE) + insn_buf[i++] = random() & 0xff; + + return i; +} + +static void parse_args(int argc, char **argv) +{ + int c; + char *tmp = NULL; + int set_seed = 0; + + prog = argv[0]; + while ((c = getopt(argc, argv, "ynvs:m:i:")) != -1) { + switch (c) { + case 'y': + x86_64 = 1; + break; + case 'n': + x86_64 = 0; + break; + case 'v': + verbose = 1; + break; + case 'i': + if (strcmp("-", optarg) == 0) + input_file = stdin; + else + input_file = fopen(optarg, "r"); + if (!input_file) + usage("Failed to open input file"); + break; + case 's': + seed = (unsigned int)strtoul(optarg, &tmp, 0); + if (*tmp == ',') { + optarg = tmp + 1; + iter_start = strtoul(optarg, &tmp, 0); + } + if (*tmp != '\0' || tmp == optarg) + usage("Failed to parse seed"); + set_seed = 1; + break; + case 'm': + iter_end = strtoul(optarg, &tmp, 0); + if (*tmp != '\0' || tmp == optarg) + usage("Failed to parse max_iter"); + break; + default: + usage(NULL); + } + } + + /* Check errors */ + if (iter_end < iter_start) + usage("Max iteration number must be bigger than iter-num"); + + if (set_seed && input_file) + usage("Don't use input file (-i) with random seed (-s)"); + + /* Initialize random seed */ + if (!input_file) { + if (!set_seed) /* No seed is given */ + init_random_seed(); + srand(seed); + } +} + +int main(int argc, char **argv) +{ + struct insn insn; + int insns = 0; + int errors = 0; + unsigned long i; + unsigned char insn_buf[MAX_INSN_SIZE * 2]; + + parse_args(argc, argv); + + /* Prepare stop bytes with NOPs */ + memset(insn_buf + MAX_INSN_SIZE, INSN_NOP, MAX_INSN_SIZE); + + for (i = 0; i < iter_end; i++) { + if (generate_insn(insn_buf) <= 0) + break; + + if (i < iter_start) /* Skip to given iteration number */ + continue; + + /* Decode an instruction */ + insn_init(&insn, insn_buf, x86_64); + insn_get_length(&insn); + + if (verbose && !insn_complete(&insn)) + dump_stream(stdout, "Info: Found an undecodable input", i, insn_buf, &insn); + + if (insn.next_byte <= insn.kaddr || + insn.kaddr + MAX_INSN_SIZE < insn.next_byte) { + /* Access out-of-range memory */ + dump_stream(stdout, "Error: Found an access violation", i, insn_buf, &insn); + errors++; + } + insns++; + } + + fprintf(stdout, "%s: decoded and checked %d %s instructions with %d errors (seed:0x%x)\n", (errors) ? "Failure" : "Success", insns, (input_file) ? "given" : "random", errors, seed); + + return errors ? 1 : 0; +} -- cgit v1.2.3 From 9251f904f95175b4a1d8cbc0449e748f9edd7629 Mon Sep 17 00:00:00 2001 From: Borislav Petkov Date: Sun, 16 Oct 2011 17:15:04 +0200 Subject: perf: Carve out callchain functionality Split the callchain code from the perf events core into a new kernel/events/callchain.c file. This simplifies a bit the big core.c Signed-off-by: Borislav Petkov Cc: Arnaldo Carvalho de Melo Cc: Stephane Eranian [keep ctx recursion handling inline and use internal headers] Signed-off-by: Frederic Weisbecker Signed-off-by: Peter Zijlstra Link: http://lkml.kernel.org/r/1318778104-17152-1-git-send-email-fweisbec@gmail.com Signed-off-by: Ingo Molnar --- kernel/events/Makefile | 2 +- kernel/events/callchain.c | 191 ++++++++++++++++++++++++++++++++++++++++++ kernel/events/core.c | 209 ---------------------------------------------- kernel/events/internal.h | 39 ++++++++- 4 files changed, 230 insertions(+), 211 deletions(-) create mode 100644 kernel/events/callchain.c diff --git a/kernel/events/Makefile b/kernel/events/Makefile index 89e5e8aa4c36..22d901f9caf4 100644 --- a/kernel/events/Makefile +++ b/kernel/events/Makefile @@ -2,5 +2,5 @@ ifdef CONFIG_FUNCTION_TRACER CFLAGS_REMOVE_core.o = -pg endif -obj-y := core.o ring_buffer.o +obj-y := core.o ring_buffer.o callchain.o obj-$(CONFIG_HAVE_HW_BREAKPOINT) += hw_breakpoint.o diff --git a/kernel/events/callchain.c b/kernel/events/callchain.c new file mode 100644 index 000000000000..057e24b665cf --- /dev/null +++ b/kernel/events/callchain.c @@ -0,0 +1,191 @@ +/* + * Performance events callchain code, extracted from core.c: + * + * Copyright (C) 2008 Thomas Gleixner + * Copyright (C) 2008-2011 Red Hat, Inc., Ingo Molnar + * Copyright (C) 2008-2011 Red Hat, Inc., Peter Zijlstra + * Copyright © 2009 Paul Mackerras, IBM Corp. + * + * For licensing details see kernel-base/COPYING + */ + +#include +#include +#include "internal.h" + +struct callchain_cpus_entries { + struct rcu_head rcu_head; + struct perf_callchain_entry *cpu_entries[0]; +}; + +static DEFINE_PER_CPU(int, callchain_recursion[PERF_NR_CONTEXTS]); +static atomic_t nr_callchain_events; +static DEFINE_MUTEX(callchain_mutex); +static struct callchain_cpus_entries *callchain_cpus_entries; + + +__weak void perf_callchain_kernel(struct perf_callchain_entry *entry, + struct pt_regs *regs) +{ +} + +__weak void perf_callchain_user(struct perf_callchain_entry *entry, + struct pt_regs *regs) +{ +} + +static void release_callchain_buffers_rcu(struct rcu_head *head) +{ + struct callchain_cpus_entries *entries; + int cpu; + + entries = container_of(head, struct callchain_cpus_entries, rcu_head); + + for_each_possible_cpu(cpu) + kfree(entries->cpu_entries[cpu]); + + kfree(entries); +} + +static void release_callchain_buffers(void) +{ + struct callchain_cpus_entries *entries; + + entries = callchain_cpus_entries; + rcu_assign_pointer(callchain_cpus_entries, NULL); + call_rcu(&entries->rcu_head, release_callchain_buffers_rcu); +} + +static int alloc_callchain_buffers(void) +{ + int cpu; + int size; + struct callchain_cpus_entries *entries; + + /* + * We can't use the percpu allocation API for data that can be + * accessed from NMI. Use a temporary manual per cpu allocation + * until that gets sorted out. + */ + size = offsetof(struct callchain_cpus_entries, cpu_entries[nr_cpu_ids]); + + entries = kzalloc(size, GFP_KERNEL); + if (!entries) + return -ENOMEM; + + size = sizeof(struct perf_callchain_entry) * PERF_NR_CONTEXTS; + + for_each_possible_cpu(cpu) { + entries->cpu_entries[cpu] = kmalloc_node(size, GFP_KERNEL, + cpu_to_node(cpu)); + if (!entries->cpu_entries[cpu]) + goto fail; + } + + rcu_assign_pointer(callchain_cpus_entries, entries); + + return 0; + +fail: + for_each_possible_cpu(cpu) + kfree(entries->cpu_entries[cpu]); + kfree(entries); + + return -ENOMEM; +} + +int get_callchain_buffers(void) +{ + int err = 0; + int count; + + mutex_lock(&callchain_mutex); + + count = atomic_inc_return(&nr_callchain_events); + if (WARN_ON_ONCE(count < 1)) { + err = -EINVAL; + goto exit; + } + + if (count > 1) { + /* If the allocation failed, give up */ + if (!callchain_cpus_entries) + err = -ENOMEM; + goto exit; + } + + err = alloc_callchain_buffers(); + if (err) + release_callchain_buffers(); +exit: + mutex_unlock(&callchain_mutex); + + return err; +} + +void put_callchain_buffers(void) +{ + if (atomic_dec_and_mutex_lock(&nr_callchain_events, &callchain_mutex)) { + release_callchain_buffers(); + mutex_unlock(&callchain_mutex); + } +} + +static struct perf_callchain_entry *get_callchain_entry(int *rctx) +{ + int cpu; + struct callchain_cpus_entries *entries; + + *rctx = get_recursion_context(__get_cpu_var(callchain_recursion)); + if (*rctx == -1) + return NULL; + + entries = rcu_dereference(callchain_cpus_entries); + if (!entries) + return NULL; + + cpu = smp_processor_id(); + + return &entries->cpu_entries[cpu][*rctx]; +} + +static void +put_callchain_entry(int rctx) +{ + put_recursion_context(__get_cpu_var(callchain_recursion), rctx); +} + +struct perf_callchain_entry *perf_callchain(struct pt_regs *regs) +{ + int rctx; + struct perf_callchain_entry *entry; + + + entry = get_callchain_entry(&rctx); + if (rctx == -1) + return NULL; + + if (!entry) + goto exit_put; + + entry->nr = 0; + + if (!user_mode(regs)) { + perf_callchain_store(entry, PERF_CONTEXT_KERNEL); + perf_callchain_kernel(entry, regs); + if (current->mm) + regs = task_pt_regs(current); + else + regs = NULL; + } + + if (regs) { + perf_callchain_store(entry, PERF_CONTEXT_USER); + perf_callchain_user(entry, regs); + } + +exit_put: + put_callchain_entry(rctx); + + return entry; +} diff --git a/kernel/events/core.c b/kernel/events/core.c index 0e8457da6f95..eadac69265fc 100644 --- a/kernel/events/core.c +++ b/kernel/events/core.c @@ -2569,215 +2569,6 @@ static u64 perf_event_read(struct perf_event *event) return perf_event_count(event); } -/* - * Callchain support - */ - -struct callchain_cpus_entries { - struct rcu_head rcu_head; - struct perf_callchain_entry *cpu_entries[0]; -}; - -static DEFINE_PER_CPU(int, callchain_recursion[PERF_NR_CONTEXTS]); -static atomic_t nr_callchain_events; -static DEFINE_MUTEX(callchain_mutex); -struct callchain_cpus_entries *callchain_cpus_entries; - - -__weak void perf_callchain_kernel(struct perf_callchain_entry *entry, - struct pt_regs *regs) -{ -} - -__weak void perf_callchain_user(struct perf_callchain_entry *entry, - struct pt_regs *regs) -{ -} - -static void release_callchain_buffers_rcu(struct rcu_head *head) -{ - struct callchain_cpus_entries *entries; - int cpu; - - entries = container_of(head, struct callchain_cpus_entries, rcu_head); - - for_each_possible_cpu(cpu) - kfree(entries->cpu_entries[cpu]); - - kfree(entries); -} - -static void release_callchain_buffers(void) -{ - struct callchain_cpus_entries *entries; - - entries = callchain_cpus_entries; - rcu_assign_pointer(callchain_cpus_entries, NULL); - call_rcu(&entries->rcu_head, release_callchain_buffers_rcu); -} - -static int alloc_callchain_buffers(void) -{ - int cpu; - int size; - struct callchain_cpus_entries *entries; - - /* - * We can't use the percpu allocation API for data that can be - * accessed from NMI. Use a temporary manual per cpu allocation - * until that gets sorted out. - */ - size = offsetof(struct callchain_cpus_entries, cpu_entries[nr_cpu_ids]); - - entries = kzalloc(size, GFP_KERNEL); - if (!entries) - return -ENOMEM; - - size = sizeof(struct perf_callchain_entry) * PERF_NR_CONTEXTS; - - for_each_possible_cpu(cpu) { - entries->cpu_entries[cpu] = kmalloc_node(size, GFP_KERNEL, - cpu_to_node(cpu)); - if (!entries->cpu_entries[cpu]) - goto fail; - } - - rcu_assign_pointer(callchain_cpus_entries, entries); - - return 0; - -fail: - for_each_possible_cpu(cpu) - kfree(entries->cpu_entries[cpu]); - kfree(entries); - - return -ENOMEM; -} - -static int get_callchain_buffers(void) -{ - int err = 0; - int count; - - mutex_lock(&callchain_mutex); - - count = atomic_inc_return(&nr_callchain_events); - if (WARN_ON_ONCE(count < 1)) { - err = -EINVAL; - goto exit; - } - - if (count > 1) { - /* If the allocation failed, give up */ - if (!callchain_cpus_entries) - err = -ENOMEM; - goto exit; - } - - err = alloc_callchain_buffers(); - if (err) - release_callchain_buffers(); -exit: - mutex_unlock(&callchain_mutex); - - return err; -} - -static void put_callchain_buffers(void) -{ - if (atomic_dec_and_mutex_lock(&nr_callchain_events, &callchain_mutex)) { - release_callchain_buffers(); - mutex_unlock(&callchain_mutex); - } -} - -static int get_recursion_context(int *recursion) -{ - int rctx; - - if (in_nmi()) - rctx = 3; - else if (in_irq()) - rctx = 2; - else if (in_softirq()) - rctx = 1; - else - rctx = 0; - - if (recursion[rctx]) - return -1; - - recursion[rctx]++; - barrier(); - - return rctx; -} - -static inline void put_recursion_context(int *recursion, int rctx) -{ - barrier(); - recursion[rctx]--; -} - -static struct perf_callchain_entry *get_callchain_entry(int *rctx) -{ - int cpu; - struct callchain_cpus_entries *entries; - - *rctx = get_recursion_context(__get_cpu_var(callchain_recursion)); - if (*rctx == -1) - return NULL; - - entries = rcu_dereference(callchain_cpus_entries); - if (!entries) - return NULL; - - cpu = smp_processor_id(); - - return &entries->cpu_entries[cpu][*rctx]; -} - -static void -put_callchain_entry(int rctx) -{ - put_recursion_context(__get_cpu_var(callchain_recursion), rctx); -} - -static struct perf_callchain_entry *perf_callchain(struct pt_regs *regs) -{ - int rctx; - struct perf_callchain_entry *entry; - - - entry = get_callchain_entry(&rctx); - if (rctx == -1) - return NULL; - - if (!entry) - goto exit_put; - - entry->nr = 0; - - if (!user_mode(regs)) { - perf_callchain_store(entry, PERF_CONTEXT_KERNEL); - perf_callchain_kernel(entry, regs); - if (current->mm) - regs = task_pt_regs(current); - else - regs = NULL; - } - - if (regs) { - perf_callchain_store(entry, PERF_CONTEXT_USER); - perf_callchain_user(entry, regs); - } - -exit_put: - put_callchain_entry(rctx); - - return entry; -} - /* * Initialize the perf_event context in a task_struct: */ diff --git a/kernel/events/internal.h b/kernel/events/internal.h index 09097dd8116c..be4a43f6de4f 100644 --- a/kernel/events/internal.h +++ b/kernel/events/internal.h @@ -1,6 +1,10 @@ #ifndef _KERNEL_EVENTS_INTERNAL_H #define _KERNEL_EVENTS_INTERNAL_H +#include + +/* Buffer handling */ + #define RING_BUFFER_WRITABLE 0x01 struct ring_buffer { @@ -64,7 +68,7 @@ static inline int page_order(struct ring_buffer *rb) } #endif -static unsigned long perf_data_size(struct ring_buffer *rb) +static inline unsigned long perf_data_size(struct ring_buffer *rb) { return rb->nr_pages << (PAGE_SHIFT + page_order(rb)); } @@ -93,4 +97,37 @@ __output_copy(struct perf_output_handle *handle, } while (len); } +/* Callchain handling */ +extern struct perf_callchain_entry *perf_callchain(struct pt_regs *regs); +extern int get_callchain_buffers(void); +extern void put_callchain_buffers(void); + +static inline int get_recursion_context(int *recursion) +{ + int rctx; + + if (in_nmi()) + rctx = 3; + else if (in_irq()) + rctx = 2; + else if (in_softirq()) + rctx = 1; + else + rctx = 0; + + if (recursion[rctx]) + return -1; + + recursion[rctx]++; + barrier(); + + return rctx; +} + +static inline void put_recursion_context(int *recursion, int rctx) +{ + barrier(); + recursion[rctx]--; +} + #endif /* _KERNEL_EVENTS_INTERNAL_H */ -- cgit v1.2.3 From 5d81e5cfb37a174e8ddc0413e2e70cdf05807ace Mon Sep 17 00:00:00 2001 From: Andrew Vagin Date: Mon, 7 Nov 2011 15:54:12 +0300 Subject: events: Don't divide events if it has field period This patch solves the following problem: Now some samples may be lost due to throttling. The number of samples is restricted by sysctl_perf_event_sample_rate/HZ. A trace event is divided on some samples according to event's period. I don't sure, that we should generate more than one sample on each trace event. I think the better way to use SAMPLE_PERIOD. E.g.: I want to trace when a process sleeps. I created a process, which sleeps for 1ms and for 4ms. perf got 100 events in both cases. swapper 0 [000] 1141.371830: sched_stat_sleep: comm=foo pid=1801 delay=1386750 [ns] swapper 0 [000] 1141.369444: sched_stat_sleep: comm=foo pid=1801 delay=4499585 [ns] In the first case a kernel want to send 4499585 events and in the second case it wants to send 1386750 events. perf-reports shows that process sleeps in both places equal time. It's bug. With this patch kernel generates one event on each "sleep" and the time slice is saved in the field "period". Perf knows how handle it. Signed-off-by: Andrew Vagin Signed-off-by: Peter Zijlstra Link: http://lkml.kernel.org/r/1320670457-2633428-3-git-send-email-avagin@openvz.org Signed-off-by: Ingo Molnar --- kernel/events/core.c | 7 ++++++- 1 file changed, 6 insertions(+), 1 deletion(-) diff --git a/kernel/events/core.c b/kernel/events/core.c index eadac69265fc..8d9dea56c262 100644 --- a/kernel/events/core.c +++ b/kernel/events/core.c @@ -4528,7 +4528,6 @@ static void perf_swevent_overflow(struct perf_event *event, u64 overflow, struct hw_perf_event *hwc = &event->hw; int throttle = 0; - data->period = event->hw.last_period; if (!overflow) overflow = perf_swevent_set_period(event); @@ -4562,6 +4561,12 @@ static void perf_swevent_event(struct perf_event *event, u64 nr, if (!is_sampling_event(event)) return; + if ((event->attr.sample_type & PERF_SAMPLE_PERIOD) && !event->attr.freq) { + data->period = nr; + return perf_swevent_overflow(event, 1, data, regs); + } else + data->period = event->hw.last_period; + if (nr == 1 && hwc->sample_period == 1 && !event->attr.freq) return perf_swevent_overflow(event, 1, data, regs); -- cgit v1.2.3 From 77271ce4b2c0df0a76ad1cbb6a95b07e1f88c1ea Mon Sep 17 00:00:00 2001 From: Steven Rostedt Date: Thu, 17 Nov 2011 09:34:33 -0500 Subject: tracing: Add irq, preempt-count and need resched info to default trace output People keep asking how to get the preempt count, irq, and need resched info and we keep telling them to enable the latency format. Some developers think that traces without this info is completely useless, and for a lot of tasks it is useless. The first option was to enable the latency trace as the default format, but the header for the latency format is pretty useless for most tracers and it also does the timestamp in straight microseconds from the time the trace started. This is sometimes more difficult to read as the default trace is seconds from the start of boot up. Latency format: # tracer: nop # # nop latency trace v1.1.5 on 3.2.0-rc1-test+ # -------------------------------------------------------------------- # latency: 0 us, #159771/64234230, CPU#1 | (M:preempt VP:0, KP:0, SP:0 HP:0 #P:4) # ----------------- # | task: -0 (uid:0 nice:0 policy:0 rt_prio:0) # ----------------- # # _------=> CPU# # / _-----=> irqs-off # | / _----=> need-resched # || / _---=> hardirq/softirq # ||| / _--=> preempt-depth # |||| / delay # cmd pid ||||| time | caller # \ / ||||| \ | / migratio-6 0...2 41778231us+: rcu_note_context_switch <-__schedule migratio-6 0...2 41778233us : trace_rcu_utilization <-rcu_note_context_switch migratio-6 0...2 41778235us+: rcu_sched_qs <-rcu_note_context_switch migratio-6 0d..2 41778236us+: rcu_preempt_qs <-rcu_note_context_switch migratio-6 0...2 41778238us : trace_rcu_utilization <-rcu_note_context_switch migratio-6 0...2 41778239us+: debug_lockdep_rcu_enabled <-__schedule default format: # tracer: nop # # TASK-PID CPU# TIMESTAMP FUNCTION # | | | | | migration/0-6 [000] 50.025810: rcu_note_context_switch <-__schedule migration/0-6 [000] 50.025812: trace_rcu_utilization <-rcu_note_context_switch migration/0-6 [000] 50.025813: rcu_sched_qs <-rcu_note_context_switch migration/0-6 [000] 50.025815: rcu_preempt_qs <-rcu_note_context_switch migration/0-6 [000] 50.025817: trace_rcu_utilization <-rcu_note_context_switch migration/0-6 [000] 50.025818: debug_lockdep_rcu_enabled <-__schedule migration/0-6 [000] 50.025820: debug_lockdep_rcu_enabled <-__schedule The latency format header has latency information that is pretty meaningless for most tracers. Although some of the header is useful, and we can add that later to the default format as well. What is really useful with the latency format is the irqs-off, need-resched hard/softirq context and the preempt count. This commit adds the option irq-info which is on by default that adds this information: # tracer: nop # # _-----=> irqs-off # / _----=> need-resched # | / _---=> hardirq/softirq # || / _--=> preempt-depth # ||| / delay # TASK-PID CPU# |||| TIMESTAMP FUNCTION # | | | |||| | | -0 [000] d..2 49.309305: cpuidle_get_driver <-cpuidle_idle_call -0 [000] d..2 49.309307: mwait_idle <-cpu_idle -0 [000] d..2 49.309309: need_resched <-mwait_idle -0 [000] d..2 49.309310: test_ti_thread_flag <-need_resched -0 [000] d..2 49.309312: trace_power_start.constprop.13 <-mwait_idle -0 [000] d..2 49.309313: trace_cpu_idle <-mwait_idle -0 [000] d..2 49.309315: need_resched <-mwait_idle If a user wants the old format, they can disable the 'irq-info' option: # tracer: nop # # TASK-PID CPU# TIMESTAMP FUNCTION # | | | | | -0 [000] 49.309305: cpuidle_get_driver <-cpuidle_idle_call -0 [000] 49.309307: mwait_idle <-cpu_idle -0 [000] 49.309309: need_resched <-mwait_idle -0 [000] 49.309310: test_ti_thread_flag <-need_resched -0 [000] 49.309312: trace_power_start.constprop.13 <-mwait_idle -0 [000] 49.309313: trace_cpu_idle <-mwait_idle -0 [000] 49.309315: need_resched <-mwait_idle Requested-by: Thomas Gleixner Signed-off-by: Steven Rostedt --- kernel/trace/trace.c | 24 ++++++++++++++++++++---- kernel/trace/trace.h | 1 + kernel/trace/trace_output.c | 16 ++++++++++++++-- 3 files changed, 35 insertions(+), 6 deletions(-) diff --git a/kernel/trace/trace.c b/kernel/trace/trace.c index 9e158cc84ced..fa5cee639962 100644 --- a/kernel/trace/trace.c +++ b/kernel/trace/trace.c @@ -338,7 +338,8 @@ static DECLARE_WAIT_QUEUE_HEAD(trace_wait); /* trace_flags holds trace_options default values */ unsigned long trace_flags = TRACE_ITER_PRINT_PARENT | TRACE_ITER_PRINTK | TRACE_ITER_ANNOTATE | TRACE_ITER_CONTEXT_INFO | TRACE_ITER_SLEEP_TIME | - TRACE_ITER_GRAPH_TIME | TRACE_ITER_RECORD_CMD | TRACE_ITER_OVERWRITE; + TRACE_ITER_GRAPH_TIME | TRACE_ITER_RECORD_CMD | TRACE_ITER_OVERWRITE | + TRACE_ITER_IRQ_INFO; static int trace_stop_count; static DEFINE_RAW_SPINLOCK(tracing_start_lock); @@ -426,6 +427,7 @@ static const char *trace_options[] = { "record-cmd", "overwrite", "disable_on_free", + "irq-info", NULL }; @@ -1857,10 +1859,20 @@ static void print_lat_help_header(struct seq_file *m) static void print_func_help_header(struct seq_file *m) { - seq_puts(m, "# TASK-PID CPU# TIMESTAMP FUNCTION\n"); + seq_puts(m, "# TASK-PID CPU# TIMESTAMP FUNCTION\n"); seq_puts(m, "# | | | | |\n"); } +static void print_func_help_header_irq(struct seq_file *m) +{ + seq_puts(m, "# _-----=> irqs-off\n"); + seq_puts(m, "# / _----=> need-resched\n"); + seq_puts(m, "# | / _---=> hardirq/softirq\n"); + seq_puts(m, "# || / _--=> preempt-depth\n"); + seq_puts(m, "# ||| / delay\n"); + seq_puts(m, "# TASK-PID CPU# |||| TIMESTAMP FUNCTION\n"); + seq_puts(m, "# | | | |||| | |\n"); +} void print_trace_header(struct seq_file *m, struct trace_iterator *iter) @@ -2170,8 +2182,12 @@ void trace_default_header(struct seq_file *m) if (!(trace_flags & TRACE_ITER_VERBOSE)) print_lat_help_header(m); } else { - if (!(trace_flags & TRACE_ITER_VERBOSE)) - print_func_help_header(m); + if (!(trace_flags & TRACE_ITER_VERBOSE)) { + if (trace_flags & TRACE_ITER_IRQ_INFO) + print_func_help_header_irq(m); + else + print_func_help_header(m); + } } } diff --git a/kernel/trace/trace.h b/kernel/trace/trace.h index f8ec2291b522..2c2657462ac3 100644 --- a/kernel/trace/trace.h +++ b/kernel/trace/trace.h @@ -655,6 +655,7 @@ enum trace_iterator_flags { TRACE_ITER_RECORD_CMD = 0x100000, TRACE_ITER_OVERWRITE = 0x200000, TRACE_ITER_STOP_ON_FREE = 0x400000, + TRACE_ITER_IRQ_INFO = 0x800000, }; /* diff --git a/kernel/trace/trace_output.c b/kernel/trace/trace_output.c index 51999309a6cf..0d6ff3555942 100644 --- a/kernel/trace/trace_output.c +++ b/kernel/trace/trace_output.c @@ -627,11 +627,23 @@ int trace_print_context(struct trace_iterator *iter) unsigned long usec_rem = do_div(t, USEC_PER_SEC); unsigned long secs = (unsigned long)t; char comm[TASK_COMM_LEN]; + int ret; trace_find_cmdline(entry->pid, comm); - return trace_seq_printf(s, "%16s-%-5d [%03d] %5lu.%06lu: ", - comm, entry->pid, iter->cpu, secs, usec_rem); + ret = trace_seq_printf(s, "%16s-%-5d [%03d] ", + comm, entry->pid, iter->cpu); + if (!ret) + return 0; + + if (trace_flags & TRACE_ITER_IRQ_INFO) { + ret = trace_print_lat_fmt(s, entry); + if (!ret) + return 0; + } + + return trace_seq_printf(s, " %5lu.%06lu: ", + secs, usec_rem); } int trace_print_lat_context(struct trace_iterator *iter) -- cgit v1.2.3 From 39eaf7ef884dcc44f7ff1bac803ca2a1dcf43544 Mon Sep 17 00:00:00 2001 From: Steven Rostedt Date: Thu, 17 Nov 2011 10:35:16 -0500 Subject: tracing: Add entries in buffer and total entries to default output header Knowing the number of event entries in the ring buffer compared to the total number that were written is useful information. The latency format gives this information and there's no reason that the default format does not. This information is now added to the default header, along with the number of online CPUs: # tracer: nop # # entries-in-buffer/entries-written: 159836/64690869 #P:4 # # _-----=> irqs-off # / _----=> need-resched # | / _---=> hardirq/softirq # || / _--=> preempt-depth # ||| / delay # TASK-PID CPU# |||| TIMESTAMP FUNCTION # | | | |||| | | -0 [000] ...2 49.442971: local_touch_nmi <-cpu_idle -0 [000] d..2 49.442973: enter_idle <-cpu_idle -0 [000] d..2 49.442974: atomic_notifier_call_chain <-enter_idle -0 [000] d..2 49.442976: __atomic_notifier_call_chain <-atomic_notifier The above shows that the trace contains 159836 entries, but 64690869 were written. One could figure out that there were 64531033 entries that were dropped. Signed-off-by: Steven Rostedt --- kernel/trace/trace.c | 72 ++++++++++++++++++++++++++++++++++------------------ 1 file changed, 47 insertions(+), 25 deletions(-) diff --git a/kernel/trace/trace.c b/kernel/trace/trace.c index fa5cee639962..7392070ffc39 100644 --- a/kernel/trace/trace.c +++ b/kernel/trace/trace.c @@ -1845,6 +1845,33 @@ static void s_stop(struct seq_file *m, void *p) trace_event_read_unlock(); } +static void +get_total_entries(struct trace_array *tr, unsigned long *total, unsigned long *entries) +{ + unsigned long count; + int cpu; + + *total = 0; + *entries = 0; + + for_each_tracing_cpu(cpu) { + count = ring_buffer_entries_cpu(tr->buffer, cpu); + /* + * If this buffer has skipped entries, then we hold all + * entries for the trace and we need to ignore the + * ones before the time stamp. + */ + if (tr->data[cpu]->skipped_entries) { + count -= tr->data[cpu]->skipped_entries; + /* total is the same as the entries */ + *total += count; + } else + *total += count + + ring_buffer_overrun_cpu(tr->buffer, cpu); + *entries += count; + } +} + static void print_lat_help_header(struct seq_file *m) { seq_puts(m, "# _------=> CPU# \n"); @@ -1857,14 +1884,27 @@ static void print_lat_help_header(struct seq_file *m) seq_puts(m, "# \\ / ||||| \\ | / \n"); } -static void print_func_help_header(struct seq_file *m) +static void print_event_info(struct trace_array *tr, struct seq_file *m) { + unsigned long total; + unsigned long entries; + + get_total_entries(tr, &total, &entries); + seq_printf(m, "# entries-in-buffer/entries-written: %lu/%lu #P:%d\n", + entries, total, num_online_cpus()); + seq_puts(m, "#\n"); +} + +static void print_func_help_header(struct trace_array *tr, struct seq_file *m) +{ + print_event_info(tr, m); seq_puts(m, "# TASK-PID CPU# TIMESTAMP FUNCTION\n"); seq_puts(m, "# | | | | |\n"); } -static void print_func_help_header_irq(struct seq_file *m) +static void print_func_help_header_irq(struct trace_array *tr, struct seq_file *m) { + print_event_info(tr, m); seq_puts(m, "# _-----=> irqs-off\n"); seq_puts(m, "# / _----=> need-resched\n"); seq_puts(m, "# | / _---=> hardirq/softirq\n"); @@ -1881,32 +1921,14 @@ print_trace_header(struct seq_file *m, struct trace_iterator *iter) struct trace_array *tr = iter->tr; struct trace_array_cpu *data = tr->data[tr->cpu]; struct tracer *type = current_trace; - unsigned long entries = 0; - unsigned long total = 0; - unsigned long count; + unsigned long entries; + unsigned long total; const char *name = "preemption"; - int cpu; if (type) name = type->name; - - for_each_tracing_cpu(cpu) { - count = ring_buffer_entries_cpu(tr->buffer, cpu); - /* - * If this buffer has skipped entries, then we hold all - * entries for the trace and we need to ignore the - * ones before the time stamp. - */ - if (tr->data[cpu]->skipped_entries) { - count -= tr->data[cpu]->skipped_entries; - /* total is the same as the entries */ - total += count; - } else - total += count + - ring_buffer_overrun_cpu(tr->buffer, cpu); - entries += count; - } + get_total_entries(tr, &total, &entries); seq_printf(m, "# %s latency trace v1.1.5 on %s\n", name, UTS_RELEASE); @@ -2184,9 +2206,9 @@ void trace_default_header(struct seq_file *m) } else { if (!(trace_flags & TRACE_ITER_VERBOSE)) { if (trace_flags & TRACE_ITER_IRQ_INFO) - print_func_help_header_irq(m); + print_func_help_header_irq(iter->tr, m); else - print_func_help_header(m); + print_func_help_header(iter->tr, m); } } } -- cgit v1.2.3 From c168fbfb93a1c4044287858c6784f0bd1f6cfe33 Mon Sep 17 00:00:00 2001 From: Arnaldo Carvalho de Melo Date: Wed, 16 Nov 2011 12:55:59 -0200 Subject: perf tools: Eliminate duplicate code and use PATH_MAX consistently No need for multiple definitions for STR() and die(), also use SuSv2's PATH_MAX instead of adding MAX_PATH. Cc: David Ahern Cc: Frederic Weisbecker Cc: Mike Galbraith Cc: Paul Mackerras Cc: Peter Zijlstra Cc: Stephane Eranian Link: http://lkml.kernel.org/n/tip-qpujjkw7u0bf0tr4wt55cr9y@git.kernel.org Signed-off-by: Arnaldo Carvalho de Melo --- tools/perf/builtin-probe.c | 1 - tools/perf/util/cgroup.c | 15 +++++++-------- tools/perf/util/debugfs.c | 12 ++++++------ tools/perf/util/debugfs.h | 29 +++++++++-------------------- tools/perf/util/probe-finder.h | 1 - tools/perf/util/trace-event-info.c | 28 ++-------------------------- 6 files changed, 24 insertions(+), 62 deletions(-) diff --git a/tools/perf/builtin-probe.c b/tools/perf/builtin-probe.c index 710ae3d0a489..59d43abfbfec 100644 --- a/tools/perf/builtin-probe.c +++ b/tools/perf/builtin-probe.c @@ -46,7 +46,6 @@ #define DEFAULT_VAR_FILTER "!__k???tab_* & !__crc_*" #define DEFAULT_FUNC_FILTER "!_*" -#define MAX_PATH_LEN 256 /* Session management structure */ static struct { diff --git a/tools/perf/util/cgroup.c b/tools/perf/util/cgroup.c index 96bee5c46008..dbe2f16b1a1a 100644 --- a/tools/perf/util/cgroup.c +++ b/tools/perf/util/cgroup.c @@ -3,7 +3,6 @@ #include "parse-options.h" #include "evsel.h" #include "cgroup.h" -#include "debugfs.h" /* MAX_PATH, STR() */ #include "evlist.h" int nr_cgroups; @@ -12,7 +11,7 @@ static int cgroupfs_find_mountpoint(char *buf, size_t maxlen) { FILE *fp; - char mountpoint[MAX_PATH+1], tokens[MAX_PATH+1], type[MAX_PATH+1]; + char mountpoint[PATH_MAX + 1], tokens[PATH_MAX + 1], type[PATH_MAX + 1]; char *token, *saved_ptr = NULL; int found = 0; @@ -25,8 +24,8 @@ cgroupfs_find_mountpoint(char *buf, size_t maxlen) * and inspect every cgroupfs mount point to find one that has * perf_event subsystem */ - while (fscanf(fp, "%*s %"STR(MAX_PATH)"s %"STR(MAX_PATH)"s %" - STR(MAX_PATH)"s %*d %*d\n", + while (fscanf(fp, "%*s %"STR(PATH_MAX)"s %"STR(PATH_MAX)"s %" + STR(PATH_MAX)"s %*d %*d\n", mountpoint, type, tokens) == 3) { if (!strcmp(type, "cgroup")) { @@ -57,15 +56,15 @@ cgroupfs_find_mountpoint(char *buf, size_t maxlen) static int open_cgroup(char *name) { - char path[MAX_PATH+1]; - char mnt[MAX_PATH+1]; + char path[PATH_MAX + 1]; + char mnt[PATH_MAX + 1]; int fd; - if (cgroupfs_find_mountpoint(mnt, MAX_PATH+1)) + if (cgroupfs_find_mountpoint(mnt, PATH_MAX + 1)) return -1; - snprintf(path, MAX_PATH, "%s/%s", mnt, name); + snprintf(path, PATH_MAX, "%s/%s", mnt, name); fd = open(path, O_RDONLY); if (fd == -1) diff --git a/tools/perf/util/debugfs.c b/tools/perf/util/debugfs.c index a88fefc0cc0a..680be3460e86 100644 --- a/tools/perf/util/debugfs.c +++ b/tools/perf/util/debugfs.c @@ -2,8 +2,10 @@ #include "debugfs.h" #include "cache.h" +#include + static int debugfs_premounted; -static char debugfs_mountpoint[MAX_PATH+1]; +static char debugfs_mountpoint[PATH_MAX + 1]; static const char *debugfs_known_mountpoints[] = { "/sys/kernel/debug/", @@ -64,9 +66,7 @@ const char *debugfs_find_mountpoint(void) if (fp == NULL) die("Can't open /proc/mounts for read"); - while (fscanf(fp, "%*s %" - STR(MAX_PATH) - "s %99s %*s %*d %*d\n", + while (fscanf(fp, "%*s %" STR(PATH_MAX) "s %99s %*s %*d %*d\n", debugfs_mountpoint, type) == 2) { if (strcmp(type, "debugfs") == 0) break; @@ -158,7 +158,7 @@ int debugfs_umount(void) int debugfs_write(const char *entry, const char *value) { - char path[MAX_PATH+1]; + char path[PATH_MAX + 1]; int ret, count; int fd; @@ -203,7 +203,7 @@ int debugfs_write(const char *entry, const char *value) */ int debugfs_read(const char *entry, char *buffer, size_t size) { - char path[MAX_PATH+1]; + char path[PATH_MAX + 1]; int ret; int fd; diff --git a/tools/perf/util/debugfs.h b/tools/perf/util/debugfs.h index 83a02879745f..8cd3fa0af880 100644 --- a/tools/perf/util/debugfs.h +++ b/tools/perf/util/debugfs.h @@ -1,25 +1,14 @@ #ifndef __DEBUGFS_H__ #define __DEBUGFS_H__ -#include - -#ifndef MAX_PATH -# define MAX_PATH 256 -#endif - -#ifndef STR -# define _STR(x) #x -# define STR(x) _STR(x) -#endif - -extern const char *debugfs_find_mountpoint(void); -extern int debugfs_valid_mountpoint(const char *debugfs); -extern int debugfs_valid_entry(const char *path); -extern char *debugfs_mount(const char *mountpoint); -extern int debugfs_umount(void); -extern int debugfs_write(const char *entry, const char *value); -extern int debugfs_read(const char *entry, char *buffer, size_t size); -extern void debugfs_force_cleanup(void); -extern int debugfs_make_path(const char *element, char *buffer, int size); +const char *debugfs_find_mountpoint(void); +int debugfs_valid_mountpoint(const char *debugfs); +int debugfs_valid_entry(const char *path); +char *debugfs_mount(const char *mountpoint); +int debugfs_umount(void); +int debugfs_write(const char *entry, const char *value); +int debugfs_read(const char *entry, char *buffer, size_t size); +void debugfs_force_cleanup(void); +int debugfs_make_path(const char *element, char *buffer, int size); #endif /* __DEBUGFS_H__ */ diff --git a/tools/perf/util/probe-finder.h b/tools/perf/util/probe-finder.h index 1132c8f0ce89..17e94d0c36f9 100644 --- a/tools/perf/util/probe-finder.h +++ b/tools/perf/util/probe-finder.h @@ -5,7 +5,6 @@ #include "util.h" #include "probe-event.h" -#define MAX_PATH_LEN 256 #define MAX_PROBE_BUFFER 1024 #define MAX_PROBES 128 diff --git a/tools/perf/util/trace-event-info.c b/tools/perf/util/trace-event-info.c index d2655f08bcc0..ac6830d8292b 100644 --- a/tools/perf/util/trace-event-info.c +++ b/tools/perf/util/trace-event-info.c @@ -18,7 +18,8 @@ * * ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ */ -#define _GNU_SOURCE +#include +#include "util.h" #include #include #include @@ -31,7 +32,6 @@ #include #include #include -#include #include #include #include @@ -44,10 +44,6 @@ #define VERSION "0.5" -#define _STR(x) #x -#define STR(x) _STR(x) -#define MAX_PATH 256 - #define TRACE_CTRL "tracing_on" #define TRACE "trace" #define AVAILABLE "available_tracers" @@ -73,26 +69,6 @@ struct events { }; - -static void die(const char *fmt, ...) -{ - va_list ap; - int ret = errno; - - if (errno) - perror("perf"); - else - ret = -1; - - va_start(ap, fmt); - fprintf(stderr, " "); - vfprintf(stderr, fmt, ap); - va_end(ap); - - fprintf(stderr, "\n"); - exit(ret); -} - void *malloc_or_die(unsigned int size) { void *data; -- cgit v1.2.3 From ebf294bf4f147aff29df5a16bfb0f8ebca15feaa Mon Sep 17 00:00:00 2001 From: Arnaldo Carvalho de Melo Date: Wed, 16 Nov 2011 14:03:07 -0200 Subject: perf tools: Simplify debugfs mountpoint handling code We don't need to have two PATH_MAX char sized arrays holding it, just one in util/debugfs.c will do. Also rename debugfs_path to tracing_events_path, as it is not the path to debugfs, that is debugfs_mountpoint. Both are now accessible. This will allow accessing this code in the perf python binding without having to drag in perf.c and util/parse-events.c. The defaults for these variables are the canonical "/sys/kernel/debug" and "/sys/kernel/debug/tracing/events/", removing the need for simple tools to call debugfs_mount(NULL). Cc: David Ahern Cc: Frederic Weisbecker Cc: Mike Galbraith Cc: Paul Mackerras Cc: Peter Zijlstra Cc: Stephane Eranian Link: http://lkml.kernel.org/n/tip-ug9jvtjrsqbluuhqqxpvg30f@git.kernel.org Signed-off-by: Arnaldo Carvalho de Melo --- tools/perf/builtin-test.c | 3 ++- tools/perf/perf.c | 33 ++++----------------------------- tools/perf/util/debugfs.c | 25 ++++++++++++++++++++----- tools/perf/util/debugfs.h | 4 ++++ tools/perf/util/parse-events.c | 28 +++++++++++++--------------- tools/perf/util/parse-events.h | 1 - 6 files changed, 43 insertions(+), 51 deletions(-) diff --git a/tools/perf/builtin-test.c b/tools/perf/builtin-test.c index 831d1baeac37..77d68bfb79da 100644 --- a/tools/perf/builtin-test.c +++ b/tools/perf/builtin-test.c @@ -7,6 +7,7 @@ #include "util/cache.h" #include "util/debug.h" +#include "util/debugfs.h" #include "util/evlist.h" #include "util/parse-options.h" #include "util/parse-events.h" @@ -247,7 +248,7 @@ static int trace_event__id(const char *evname) if (asprintf(&filename, "%s/syscalls/%s/id", - debugfs_path, evname) < 0) + tracing_events_path, evname) < 0) return -1; fd = open(filename, O_RDONLY); diff --git a/tools/perf/perf.c b/tools/perf/perf.c index 73d0cac8b67e..2b2e225a4d4c 100644 --- a/tools/perf/perf.c +++ b/tools/perf/perf.c @@ -29,8 +29,6 @@ struct pager_config { int val; }; -static char debugfs_mntpt[MAXPATHLEN]; - static int pager_command_config(const char *var, const char *value, void *data) { struct pager_config *c = data; @@ -81,15 +79,6 @@ static void commit_pager_choice(void) } } -static void set_debugfs_path(void) -{ - char *path; - - path = getenv(PERF_DEBUGFS_ENVIRONMENT); - snprintf(debugfs_path, MAXPATHLEN, "%s/%s", path ?: debugfs_mntpt, - "tracing/events"); -} - static int handle_options(const char ***argv, int *argc, int *envchanged) { int handled = 0; @@ -161,15 +150,14 @@ static int handle_options(const char ***argv, int *argc, int *envchanged) fprintf(stderr, "No directory given for --debugfs-dir.\n"); usage(perf_usage_string); } - strncpy(debugfs_mntpt, (*argv)[1], MAXPATHLEN); - debugfs_mntpt[MAXPATHLEN - 1] = '\0'; + debugfs_set_path((*argv)[1]); if (envchanged) *envchanged = 1; (*argv)++; (*argc)--; } else if (!prefixcmp(cmd, CMD_DEBUGFS_DIR)) { - strncpy(debugfs_mntpt, cmd + strlen(CMD_DEBUGFS_DIR), MAXPATHLEN); - debugfs_mntpt[MAXPATHLEN - 1] = '\0'; + debugfs_set_path(cmd + strlen(CMD_DEBUGFS_DIR)); + fprintf(stderr, "dir: %s\n", debugfs_mountpoint); if (envchanged) *envchanged = 1; } else { @@ -281,7 +269,6 @@ static int run_builtin(struct cmd_struct *p, int argc, const char **argv) if (use_pager == -1 && p->option & USE_PAGER) use_pager = 1; commit_pager_choice(); - set_debugfs_path(); status = p->fn(argc, argv, prefix); exit_browser(status); @@ -416,17 +403,6 @@ static int run_argv(int *argcp, const char ***argv) return done_alias; } -/* mini /proc/mounts parser: searching for "^blah /mount/point debugfs" */ -static void get_debugfs_mntpt(void) -{ - const char *path = debugfs_mount(NULL); - - if (path) - strncpy(debugfs_mntpt, path, sizeof(debugfs_mntpt)); - else - debugfs_mntpt[0] = '\0'; -} - static void pthread__block_sigwinch(void) { sigset_t set; @@ -453,7 +429,7 @@ int main(int argc, const char **argv) if (!cmd) cmd = "perf-help"; /* get debugfs mount point from /proc/mounts */ - get_debugfs_mntpt(); + debugfs_mount(NULL); /* * "perf-xxxx" is the same as "perf xxxx", but we obviously: * @@ -476,7 +452,6 @@ int main(int argc, const char **argv) argc--; handle_options(&argv, &argc, NULL); commit_pager_choice(); - set_debugfs_path(); set_buildid_dir(); if (argc > 0) { diff --git a/tools/perf/util/debugfs.c b/tools/perf/util/debugfs.c index 680be3460e86..ffc35e748e89 100644 --- a/tools/perf/util/debugfs.c +++ b/tools/perf/util/debugfs.c @@ -2,10 +2,12 @@ #include "debugfs.h" #include "cache.h" +#include #include static int debugfs_premounted; -static char debugfs_mountpoint[PATH_MAX + 1]; +char debugfs_mountpoint[PATH_MAX + 1] = "/sys/kernel/debug"; +char tracing_events_path[PATH_MAX + 1] = "/sys/kernel/debug/tracing/events"; static const char *debugfs_known_mountpoints[] = { "/sys/kernel/debug/", @@ -64,7 +66,7 @@ const char *debugfs_find_mountpoint(void) /* give up and parse /proc/mounts */ fp = fopen("/proc/mounts", "r"); if (fp == NULL) - die("Can't open /proc/mounts for read"); + return NULL; while (fscanf(fp, "%*s %" STR(PATH_MAX) "s %99s %*s %*d %*d\n", debugfs_mountpoint, type) == 2) { @@ -106,6 +108,12 @@ int debugfs_valid_entry(const char *path) return 0; } +static void debugfs_set_tracing_events_path(const char *mountpoint) +{ + snprintf(tracing_events_path, sizeof(tracing_events_path), "%s/%s", + mountpoint, "tracing/events"); +} + /* mount the debugfs somewhere if it's not mounted */ char *debugfs_mount(const char *mountpoint) @@ -113,7 +121,7 @@ char *debugfs_mount(const char *mountpoint) /* see if it's already mounted */ if (debugfs_find_mountpoint()) { debugfs_premounted = 1; - return debugfs_mountpoint; + goto out; } /* if not mounted and no argument */ @@ -129,12 +137,19 @@ char *debugfs_mount(const char *mountpoint) return NULL; /* save the mountpoint */ - strncpy(debugfs_mountpoint, mountpoint, sizeof(debugfs_mountpoint)); debugfs_found = 1; - + strncpy(debugfs_mountpoint, mountpoint, sizeof(debugfs_mountpoint)); +out: + debugfs_set_tracing_events_path(debugfs_mountpoint); return debugfs_mountpoint; } +void debugfs_set_path(const char *mountpoint) +{ + snprintf(debugfs_mountpoint, sizeof(debugfs_mountpoint), "%s", mountpoint); + debugfs_set_tracing_events_path(mountpoint); +} + /* umount the debugfs */ int debugfs_umount(void) diff --git a/tools/perf/util/debugfs.h b/tools/perf/util/debugfs.h index 8cd3fa0af880..4a878f735eb0 100644 --- a/tools/perf/util/debugfs.h +++ b/tools/perf/util/debugfs.h @@ -6,9 +6,13 @@ int debugfs_valid_mountpoint(const char *debugfs); int debugfs_valid_entry(const char *path); char *debugfs_mount(const char *mountpoint); int debugfs_umount(void); +void debugfs_set_path(const char *mountpoint); int debugfs_write(const char *entry, const char *value); int debugfs_read(const char *entry, char *buffer, size_t size); void debugfs_force_cleanup(void); int debugfs_make_path(const char *element, char *buffer, int size); +extern char debugfs_mountpoint[]; +extern char tracing_events_path[]; + #endif /* __DEBUGFS_H__ */ diff --git a/tools/perf/util/parse-events.c b/tools/perf/util/parse-events.c index 928918b796b2..586ab3fe60f8 100644 --- a/tools/perf/util/parse-events.c +++ b/tools/perf/util/parse-events.c @@ -25,8 +25,6 @@ enum event_result { EVT_HANDLED_ALL }; -char debugfs_path[MAXPATHLEN]; - #define CHW(x) .type = PERF_TYPE_HARDWARE, .config = PERF_COUNT_HW_##x #define CSW(x) .type = PERF_TYPE_SOFTWARE, .config = PERF_COUNT_SW_##x @@ -140,7 +138,7 @@ static int tp_event_has_id(struct dirent *sys_dir, struct dirent *evt_dir) char evt_path[MAXPATHLEN]; int fd; - snprintf(evt_path, MAXPATHLEN, "%s/%s/%s/id", debugfs_path, + snprintf(evt_path, MAXPATHLEN, "%s/%s/%s/id", tracing_events_path, sys_dir->d_name, evt_dir->d_name); fd = open(evt_path, O_RDONLY); if (fd < 0) @@ -171,16 +169,16 @@ struct tracepoint_path *tracepoint_id_to_path(u64 config) char evt_path[MAXPATHLEN]; char dir_path[MAXPATHLEN]; - if (debugfs_valid_mountpoint(debugfs_path)) + if (debugfs_valid_mountpoint(tracing_events_path)) return NULL; - sys_dir = opendir(debugfs_path); + sys_dir = opendir(tracing_events_path); if (!sys_dir) return NULL; for_each_subsystem(sys_dir, sys_dirent, sys_next) { - snprintf(dir_path, MAXPATHLEN, "%s/%s", debugfs_path, + snprintf(dir_path, MAXPATHLEN, "%s/%s", tracing_events_path, sys_dirent.d_name); evt_dir = opendir(dir_path); if (!evt_dir) @@ -447,7 +445,7 @@ parse_single_tracepoint_event(char *sys_name, u64 id; int fd; - snprintf(evt_path, MAXPATHLEN, "%s/%s/%s/id", debugfs_path, + snprintf(evt_path, MAXPATHLEN, "%s/%s/%s/id", tracing_events_path, sys_name, evt_name); fd = open(evt_path, O_RDONLY); @@ -485,7 +483,7 @@ parse_multiple_tracepoint_event(struct perf_evlist *evlist, char *sys_name, struct dirent *evt_ent; DIR *evt_dir; - snprintf(evt_path, MAXPATHLEN, "%s/%s", debugfs_path, sys_name); + snprintf(evt_path, MAXPATHLEN, "%s/%s", tracing_events_path, sys_name); evt_dir = opendir(evt_path); if (!evt_dir) { @@ -528,7 +526,7 @@ parse_tracepoint_event(struct perf_evlist *evlist, const char **strp, char sys_name[MAX_EVENT_LENGTH]; unsigned int sys_length, evt_length; - if (debugfs_valid_mountpoint(debugfs_path)) + if (debugfs_valid_mountpoint(tracing_events_path)) return 0; evt_name = strchr(*strp, ':'); @@ -920,10 +918,10 @@ void print_tracepoint_events(const char *subsys_glob, const char *event_glob) char evt_path[MAXPATHLEN]; char dir_path[MAXPATHLEN]; - if (debugfs_valid_mountpoint(debugfs_path)) + if (debugfs_valid_mountpoint(tracing_events_path)) return; - sys_dir = opendir(debugfs_path); + sys_dir = opendir(tracing_events_path); if (!sys_dir) return; @@ -932,7 +930,7 @@ void print_tracepoint_events(const char *subsys_glob, const char *event_glob) !strglobmatch(sys_dirent.d_name, subsys_glob)) continue; - snprintf(dir_path, MAXPATHLEN, "%s/%s", debugfs_path, + snprintf(dir_path, MAXPATHLEN, "%s/%s", tracing_events_path, sys_dirent.d_name); evt_dir = opendir(dir_path); if (!evt_dir) @@ -964,16 +962,16 @@ int is_valid_tracepoint(const char *event_string) char evt_path[MAXPATHLEN]; char dir_path[MAXPATHLEN]; - if (debugfs_valid_mountpoint(debugfs_path)) + if (debugfs_valid_mountpoint(tracing_events_path)) return 0; - sys_dir = opendir(debugfs_path); + sys_dir = opendir(tracing_events_path); if (!sys_dir) return 0; for_each_subsystem(sys_dir, sys_dirent, sys_next) { - snprintf(dir_path, MAXPATHLEN, "%s/%s", debugfs_path, + snprintf(dir_path, MAXPATHLEN, "%s/%s", tracing_events_path, sys_dirent.d_name); evt_dir = opendir(dir_path); if (!evt_dir) diff --git a/tools/perf/util/parse-events.h b/tools/perf/util/parse-events.h index 2f8e375e038d..7e0cbe75d5f1 100644 --- a/tools/perf/util/parse-events.h +++ b/tools/perf/util/parse-events.h @@ -39,7 +39,6 @@ void print_tracepoint_events(const char *subsys_glob, const char *event_glob); int print_hwcache_events(const char *event_glob); extern int is_valid_tracepoint(const char *event_string); -extern char debugfs_path[]; extern int valid_debugfs_mount(const char *debugfs); #endif /* __PERF_PARSE_EVENTS_H */ -- cgit v1.2.3 From 50d08e47bc04eb05502f5c86b70bbd19ef1c2778 Mon Sep 17 00:00:00 2001 From: Arnaldo Carvalho de Melo Date: Fri, 4 Nov 2011 09:10:59 -0200 Subject: perf evlist: Introduce perf_evlist__add_attrs Replacing the open coded equivalents in 'perf stat'. Cc: David Ahern Cc: Frederic Weisbecker Cc: Mike Galbraith Cc: Paul Mackerras Cc: Peter Zijlstra Cc: Stephane Eranian Link: http://lkml.kernel.org/n/tip-1btwadnf2tds2g07hsccsdse@git.kernel.org Signed-off-by: Arnaldo Carvalho de Melo --- tools/perf/builtin-stat.c | 40 +++++++--------------------------------- tools/perf/util/evlist.c | 34 ++++++++++++++++++++++++++++++++++ tools/perf/util/evlist.h | 7 +++++++ 3 files changed, 48 insertions(+), 33 deletions(-) diff --git a/tools/perf/builtin-stat.c b/tools/perf/builtin-stat.c index 7d98676808d8..227befbecec8 100644 --- a/tools/perf/builtin-stat.c +++ b/tools/perf/builtin-stat.c @@ -1107,22 +1107,13 @@ static const struct option options[] = { */ static int add_default_attributes(void) { - struct perf_evsel *pos; - size_t attr_nr = 0; - size_t c; - /* Set attrs if no event is selected and !null_run: */ if (null_run) return 0; if (!evsel_list->nr_entries) { - for (c = 0; c < ARRAY_SIZE(default_attrs); c++) { - pos = perf_evsel__new(default_attrs + c, c + attr_nr); - if (pos == NULL) - return -1; - perf_evlist__add(evsel_list, pos); - } - attr_nr += c; + if (perf_evlist__add_attrs_array(evsel_list, default_attrs) < 0) + return -1; } /* Detailed events get appended to the event list: */ @@ -1131,38 +1122,21 @@ static int add_default_attributes(void) return 0; /* Append detailed run extra attributes: */ - for (c = 0; c < ARRAY_SIZE(detailed_attrs); c++) { - pos = perf_evsel__new(detailed_attrs + c, c + attr_nr); - if (pos == NULL) - return -1; - perf_evlist__add(evsel_list, pos); - } - attr_nr += c; + if (perf_evlist__add_attrs_array(evsel_list, detailed_attrs) < 0) + return -1; if (detailed_run < 2) return 0; /* Append very detailed run extra attributes: */ - for (c = 0; c < ARRAY_SIZE(very_detailed_attrs); c++) { - pos = perf_evsel__new(very_detailed_attrs + c, c + attr_nr); - if (pos == NULL) - return -1; - perf_evlist__add(evsel_list, pos); - } + if (perf_evlist__add_attrs_array(evsel_list, very_detailed_attrs) < 0) + return -1; if (detailed_run < 3) return 0; /* Append very, very detailed run extra attributes: */ - for (c = 0; c < ARRAY_SIZE(very_very_detailed_attrs); c++) { - pos = perf_evsel__new(very_very_detailed_attrs + c, c + attr_nr); - if (pos == NULL) - return -1; - perf_evlist__add(evsel_list, pos); - } - - - return 0; + return perf_evlist__add_attrs_array(evsel_list, very_very_detailed_attrs); } int cmd_stat(int argc, const char **argv, const char *prefix __used) diff --git a/tools/perf/util/evlist.c b/tools/perf/util/evlist.c index fbb4b4ab9cc6..58aa1e0092bd 100644 --- a/tools/perf/util/evlist.c +++ b/tools/perf/util/evlist.c @@ -13,6 +13,8 @@ #include "evsel.h" #include "util.h" +#include "parse-events.h" + #include #include @@ -76,6 +78,14 @@ void perf_evlist__add(struct perf_evlist *evlist, struct perf_evsel *entry) ++evlist->nr_entries; } +static void perf_evlist__splice_list_tail(struct perf_evlist *evlist, + struct list_head *list, + int nr_entries) +{ + list_splice_tail(list, &evlist->entries); + evlist->nr_entries += nr_entries; +} + int perf_evlist__add_default(struct perf_evlist *evlist) { struct perf_event_attr attr = { @@ -100,6 +110,30 @@ error: return -ENOMEM; } +int perf_evlist__add_attrs(struct perf_evlist *evlist, + struct perf_event_attr *attrs, size_t nr_attrs) +{ + struct perf_evsel *evsel, *n; + LIST_HEAD(head); + size_t i; + + for (i = 0; i < nr_attrs; i++) { + evsel = perf_evsel__new(attrs + i, evlist->nr_entries + i); + if (evsel == NULL) + goto out_delete_partial_list; + list_add_tail(&evsel->node, &head); + } + + perf_evlist__splice_list_tail(evlist, &head, nr_attrs); + + return 0; + +out_delete_partial_list: + list_for_each_entry_safe(evsel, n, &head, node) + perf_evsel__delete(evsel); + return -1; +} + void perf_evlist__disable(struct perf_evlist *evlist) { int cpu, thread; diff --git a/tools/perf/util/evlist.h b/tools/perf/util/evlist.h index 1779ffef7828..57d91ff2c56a 100644 --- a/tools/perf/util/evlist.h +++ b/tools/perf/util/evlist.h @@ -2,8 +2,10 @@ #define __PERF_EVLIST_H 1 #include +#include #include "../perf.h" #include "event.h" +#include "util.h" struct pollfd; struct thread_map; @@ -39,6 +41,11 @@ void perf_evlist__delete(struct perf_evlist *evlist); void perf_evlist__add(struct perf_evlist *evlist, struct perf_evsel *entry); int perf_evlist__add_default(struct perf_evlist *evlist); +int perf_evlist__add_attrs(struct perf_evlist *evlist, + struct perf_event_attr *attrs, size_t nr_attrs); + +#define perf_evlist__add_attrs_array(evlist, array) \ + perf_evlist__add_attrs(evlist, array, ARRAY_SIZE(array)) void perf_evlist__id_add(struct perf_evlist *evlist, struct perf_evsel *evsel, int cpu, int thread, u64 id); -- cgit v1.2.3 From a8c9ae18d810e1ae12b6ec960907e9af63171d3a Mon Sep 17 00:00:00 2001 From: Arnaldo Carvalho de Melo Date: Sat, 5 Nov 2011 08:41:51 -0200 Subject: perf evlist: Introduce add_tracepoints method Convenient way of asking for tracepoint events to be added to an existing evlist. Cc: David Ahern Cc: Frederic Weisbecker Cc: Mike Galbraith Cc: Paul Mackerras Cc: Peter Zijlstra Cc: Stephane Eranian Link: http://lkml.kernel.org/n/tip-0ylj4wrg54791u0baqb9swbb@git.kernel.org Signed-off-by: Arnaldo Carvalho de Melo --- tools/perf/util/evlist.c | 57 +++++++++++++++++++++++++++++++++++++++++++++++- tools/perf/util/evlist.h | 5 +++++ tools/perf/util/setup.py | 3 ++- 3 files changed, 63 insertions(+), 2 deletions(-) diff --git a/tools/perf/util/evlist.c b/tools/perf/util/evlist.c index 58aa1e0092bd..3bc5a287a9f9 100644 --- a/tools/perf/util/evlist.c +++ b/tools/perf/util/evlist.c @@ -6,12 +6,13 @@ * * Released under the GPL v2. (and only v2, not any later version) */ +#include "util.h" +#include "debugfs.h" #include #include "cpumap.h" #include "thread_map.h" #include "evlist.h" #include "evsel.h" -#include "util.h" #include "parse-events.h" @@ -134,6 +135,60 @@ out_delete_partial_list: return -1; } +static int trace_event__id(const char *evname) +{ + char *filename, *colon; + int err = -1, fd; + + if (asprintf(&filename, "%s/%s/id", tracing_events_path, evname) < 0) + return -1; + + colon = strrchr(filename, ':'); + if (colon != NULL) + *colon = '/'; + + fd = open(filename, O_RDONLY); + if (fd >= 0) { + char id[16]; + if (read(fd, id, sizeof(id)) > 0) + err = atoi(id); + close(fd); + } + + free(filename); + return err; +} + +int perf_evlist__add_tracepoints(struct perf_evlist *evlist, + const char *tracepoints[], + size_t nr_tracepoints) +{ + int err; + size_t i; + struct perf_event_attr *attrs = zalloc(nr_tracepoints * sizeof(*attrs)); + + if (attrs == NULL) + return -1; + + for (i = 0; i < nr_tracepoints; i++) { + err = trace_event__id(tracepoints[i]); + + if (err < 0) + goto out_free_attrs; + + attrs[i].type = PERF_TYPE_TRACEPOINT; + attrs[i].config = err; + attrs[i].sample_type = (PERF_SAMPLE_RAW | PERF_SAMPLE_TIME | + PERF_SAMPLE_CPU); + attrs[i].sample_period = 1; + } + + err = perf_evlist__add_attrs(evlist, attrs, nr_tracepoints); +out_free_attrs: + free(attrs); + return err; +} + void perf_evlist__disable(struct perf_evlist *evlist) { int cpu, thread; diff --git a/tools/perf/util/evlist.h b/tools/perf/util/evlist.h index 57d91ff2c56a..ec71c82935bd 100644 --- a/tools/perf/util/evlist.h +++ b/tools/perf/util/evlist.h @@ -43,10 +43,15 @@ void perf_evlist__add(struct perf_evlist *evlist, struct perf_evsel *entry); int perf_evlist__add_default(struct perf_evlist *evlist); int perf_evlist__add_attrs(struct perf_evlist *evlist, struct perf_event_attr *attrs, size_t nr_attrs); +int perf_evlist__add_tracepoints(struct perf_evlist *evlist, + const char *tracepoints[], size_t nr_tracepoints); #define perf_evlist__add_attrs_array(evlist, array) \ perf_evlist__add_attrs(evlist, array, ARRAY_SIZE(array)) +#define perf_evlist__add_tracepoints_array(evlist, array) \ + perf_evlist__add_tracepoints(evlist, array, ARRAY_SIZE(array)) + void perf_evlist__id_add(struct perf_evlist *evlist, struct perf_evsel *evsel, int cpu, int thread, u64 id); diff --git a/tools/perf/util/setup.py b/tools/perf/util/setup.py index 95d370074928..36d4c5619575 100644 --- a/tools/perf/util/setup.py +++ b/tools/perf/util/setup.py @@ -27,7 +27,8 @@ build_tmp = getenv('PYTHON_EXTBUILD_TMP') perf = Extension('perf', sources = ['util/python.c', 'util/ctype.c', 'util/evlist.c', 'util/evsel.c', 'util/cpumap.c', 'util/thread_map.c', - 'util/util.c', 'util/xyarray.c', 'util/cgroup.c'], + 'util/util.c', 'util/xyarray.c', 'util/cgroup.c', + 'util/debugfs.c'], include_dirs = ['util/include'], extra_compile_args = cflags, ) -- cgit v1.2.3 From 0f82ebc452f921590e216b28eee0b41f5e434a48 Mon Sep 17 00:00:00 2001 From: Arnaldo Carvalho de Melo Date: Tue, 8 Nov 2011 14:41:57 -0200 Subject: perf evsel: Introduce config attr method Out of the code in 'perf record', so that we can share option parsing, etc. Eventually will be used by 'perf top', but first 'trace' will use it. Cc: David Ahern Cc: Frederic Weisbecker Cc: Mike Galbraith Cc: Paul Mackerras Cc: Peter Zijlstra Cc: Stephane Eranian Link: http://lkml.kernel.org/n/tip-hzjqsgnte1esk90ytq0ap98v@git.kernel.org Signed-off-by: Arnaldo Carvalho de Melo --- tools/perf/builtin-record.c | 180 ++++++++++++-------------------------------- tools/perf/perf.h | 20 +++++ tools/perf/util/evlist.c | 16 ++++ tools/perf/util/evlist.h | 4 + tools/perf/util/evsel.c | 70 +++++++++++++++++ tools/perf/util/evsel.h | 4 + 6 files changed, 162 insertions(+), 132 deletions(-) diff --git a/tools/perf/builtin-record.c b/tools/perf/builtin-record.c index 6ab58cc99d53..c3ac5415c097 100644 --- a/tools/perf/builtin-record.c +++ b/tools/perf/builtin-record.c @@ -35,32 +35,24 @@ enum write_mode_t { WRITE_APPEND }; -static u64 user_interval = ULLONG_MAX; -static u64 default_interval = 0; +static struct perf_record_opts record_opts = { + .target_pid = -1, + .target_tid = -1, + .user_freq = UINT_MAX, + .user_interval = ULLONG_MAX, + .freq = 1000, + .sample_id_all_avail = true, +}; static unsigned int page_size; static unsigned int mmap_pages = UINT_MAX; -static unsigned int user_freq = UINT_MAX; -static int freq = 1000; static int output; static int pipe_output = 0; static const char *output_name = NULL; static bool group = false; static int realtime_prio = 0; -static bool nodelay = false; -static bool raw_samples = false; -static bool sample_id_all_avail = true; -static bool system_wide = false; -static pid_t target_pid = -1; -static pid_t target_tid = -1; static pid_t child_pid = -1; -static bool no_inherit = false; static enum write_mode_t write_mode = WRITE_FORCE; -static bool call_graph = false; -static bool inherit_stat = false; -static bool no_samples = false; -static bool sample_address = false; -static bool sample_time = false; static bool no_buildid = false; static bool no_buildid_cache = false; static struct perf_evlist *evsel_list; @@ -72,7 +64,6 @@ static int file_new = 1; static off_t post_processing_offset; static struct perf_session *session; -static const char *cpu_list; static const char *progname; static void advance_output(size_t size) @@ -169,78 +160,6 @@ static void sig_atexit(void) kill(getpid(), signr); } -static void config_attr(struct perf_evsel *evsel, struct perf_evlist *evlist) -{ - struct perf_event_attr *attr = &evsel->attr; - int track = !evsel->idx; /* only the first counter needs these */ - - attr->disabled = 1; - attr->inherit = !no_inherit; - attr->read_format = PERF_FORMAT_TOTAL_TIME_ENABLED | - PERF_FORMAT_TOTAL_TIME_RUNNING | - PERF_FORMAT_ID; - - attr->sample_type |= PERF_SAMPLE_IP | PERF_SAMPLE_TID; - - if (evlist->nr_entries > 1) - attr->sample_type |= PERF_SAMPLE_ID; - - /* - * We default some events to a 1 default interval. But keep - * it a weak assumption overridable by the user. - */ - if (!attr->sample_period || (user_freq != UINT_MAX && - user_interval != ULLONG_MAX)) { - if (freq) { - attr->sample_type |= PERF_SAMPLE_PERIOD; - attr->freq = 1; - attr->sample_freq = freq; - } else { - attr->sample_period = default_interval; - } - } - - if (no_samples) - attr->sample_freq = 0; - - if (inherit_stat) - attr->inherit_stat = 1; - - if (sample_address) { - attr->sample_type |= PERF_SAMPLE_ADDR; - attr->mmap_data = track; - } - - if (call_graph) - attr->sample_type |= PERF_SAMPLE_CALLCHAIN; - - if (system_wide) - attr->sample_type |= PERF_SAMPLE_CPU; - - if (sample_id_all_avail && - (sample_time || system_wide || !no_inherit || cpu_list)) - attr->sample_type |= PERF_SAMPLE_TIME; - - if (raw_samples) { - attr->sample_type |= PERF_SAMPLE_TIME; - attr->sample_type |= PERF_SAMPLE_RAW; - attr->sample_type |= PERF_SAMPLE_CPU; - } - - if (nodelay) { - attr->watermark = 0; - attr->wakeup_events = 1; - } - - attr->mmap = track; - attr->comm = track; - - if (target_pid == -1 && target_tid == -1 && !system_wide) { - attr->disabled = 1; - attr->enable_on_exec = 1; - } -} - static bool perf_evlist__equal(struct perf_evlist *evlist, struct perf_evlist *other) { @@ -264,11 +183,10 @@ static void open_counters(struct perf_evlist *evlist) { struct perf_evsel *pos, *first; - if (evlist->cpus->map[0] < 0) - no_inherit = true; - first = list_entry(evlist->entries.next, struct perf_evsel, node); + perf_evlist__config_attrs(evlist, &record_opts); + list_for_each_entry(pos, &evlist->entries, node) { struct perf_event_attr *attr = &pos->attr; struct xyarray *group_fd = NULL; @@ -288,10 +206,8 @@ static void open_counters(struct perf_evlist *evlist) if (group && pos != first) group_fd = first->fd; - - config_attr(pos, evlist); retry_sample_id: - attr->sample_id_all = sample_id_all_avail ? 1 : 0; + attr->sample_id_all = record_opts.sample_id_all_avail ? 1 : 0; try_again: if (perf_evsel__open(pos, evlist->cpus, evlist->threads, group, group_fd) < 0) { @@ -300,15 +216,15 @@ try_again: if (err == EPERM || err == EACCES) { ui__error_paranoid(); exit(EXIT_FAILURE); - } else if (err == ENODEV && cpu_list) { + } else if (err == ENODEV && record_opts.cpu_list) { die("No such device - did you specify" " an out-of-range profile CPU?\n"); - } else if (err == EINVAL && sample_id_all_avail) { + } else if (err == EINVAL && record_opts.sample_id_all_avail) { /* * Old kernel, no attr->sample_id_type_all field */ - sample_id_all_avail = false; - if (!sample_time && !raw_samples && !time_needed) + record_opts.sample_id_all_avail = false; + if (!record_opts.sample_time && !record_opts.raw_samples && !time_needed) attr->sample_type &= ~PERF_SAMPLE_TIME; goto retry_sample_id; @@ -482,13 +398,13 @@ static int __cmd_record(int argc, const char **argv) if (!output_name) { if (!fstat(STDOUT_FILENO, &st) && S_ISFIFO(st.st_mode)) - pipe_output = 1; + pipe_output = true; else output_name = "perf.data"; } if (output_name) { if (!strcmp(output_name, "-")) - pipe_output = 1; + pipe_output = true; else if (!stat(output_name, &st) && st.st_size) { if (write_mode == WRITE_FORCE) { char oldname[PATH_MAX]; @@ -592,7 +508,7 @@ static int __cmd_record(int argc, const char **argv) exit(-1); } - if (!system_wide && target_tid == -1 && target_pid == -1) + if (!record_opts.system_wide && record_opts.target_tid == -1 && record_opts.target_pid == -1) evsel_list->threads->map[0] = child_pid; close(child_ready_pipe[1]); @@ -689,7 +605,7 @@ static int __cmd_record(int argc, const char **argv) perf_session__process_machines(session, perf_event__synthesize_guest_os); - if (!system_wide) + if (!record_opts.system_wide) perf_event__synthesize_thread_map(evsel_list->threads, process_synthesized_event, session); @@ -766,44 +682,44 @@ const struct option record_options[] = { parse_events_option), OPT_CALLBACK(0, "filter", &evsel_list, "filter", "event filter", parse_filter), - OPT_INTEGER('p', "pid", &target_pid, + OPT_INTEGER('p', "pid", &record_opts.target_pid, "record events on existing process id"), - OPT_INTEGER('t', "tid", &target_tid, + OPT_INTEGER('t', "tid", &record_opts.target_tid, "record events on existing thread id"), OPT_INTEGER('r', "realtime", &realtime_prio, "collect data with this RT SCHED_FIFO priority"), - OPT_BOOLEAN('D', "no-delay", &nodelay, + OPT_BOOLEAN('D', "no-delay", &record_opts.no_delay, "collect data without buffering"), - OPT_BOOLEAN('R', "raw-samples", &raw_samples, + OPT_BOOLEAN('R', "raw-samples", &record_opts.raw_samples, "collect raw sample records from all opened counters"), - OPT_BOOLEAN('a', "all-cpus", &system_wide, + OPT_BOOLEAN('a', "all-cpus", &record_opts.system_wide, "system-wide collection from all CPUs"), OPT_BOOLEAN('A', "append", &append_file, "append to the output file to do incremental profiling"), - OPT_STRING('C', "cpu", &cpu_list, "cpu", + OPT_STRING('C', "cpu", &record_opts.cpu_list, "cpu", "list of cpus to monitor"), OPT_BOOLEAN('f', "force", &force, "overwrite existing data file (deprecated)"), - OPT_U64('c', "count", &user_interval, "event period to sample"), + OPT_U64('c', "count", &record_opts.user_interval, "event period to sample"), OPT_STRING('o', "output", &output_name, "file", "output file name"), - OPT_BOOLEAN('i', "no-inherit", &no_inherit, + OPT_BOOLEAN('i', "no-inherit", &record_opts.no_inherit, "child tasks do not inherit counters"), - OPT_UINTEGER('F', "freq", &user_freq, "profile at this frequency"), + OPT_UINTEGER('F', "freq", &record_opts.user_freq, "profile at this frequency"), OPT_UINTEGER('m', "mmap-pages", &mmap_pages, "number of mmap data pages"), OPT_BOOLEAN(0, "group", &group, "put the counters into a counter group"), - OPT_BOOLEAN('g', "call-graph", &call_graph, + OPT_BOOLEAN('g', "call-graph", &record_opts.call_graph, "do call-graph (stack chain/backtrace) recording"), OPT_INCR('v', "verbose", &verbose, "be more verbose (show counter open errors, etc)"), OPT_BOOLEAN('q', "quiet", &quiet, "don't print any message"), - OPT_BOOLEAN('s', "stat", &inherit_stat, + OPT_BOOLEAN('s', "stat", &record_opts.inherit_stat, "per thread counts"), - OPT_BOOLEAN('d', "data", &sample_address, + OPT_BOOLEAN('d', "data", &record_opts.sample_address, "Sample addresses"), - OPT_BOOLEAN('T', "timestamp", &sample_time, "Sample timestamps"), - OPT_BOOLEAN('n', "no-samples", &no_samples, + OPT_BOOLEAN('T', "timestamp", &record_opts.sample_time, "Sample timestamps"), + OPT_BOOLEAN('n', "no-samples", &record_opts.no_samples, "don't sample"), OPT_BOOLEAN('N', "no-buildid-cache", &no_buildid_cache, "do not update the buildid cache"), @@ -828,8 +744,8 @@ int cmd_record(int argc, const char **argv, const char *prefix __used) argc = parse_options(argc, argv, record_options, record_usage, PARSE_OPT_STOP_AT_NON_OPTION); - if (!argc && target_pid == -1 && target_tid == -1 && - !system_wide && !cpu_list) + if (!argc && record_opts.target_pid == -1 && record_opts.target_tid == -1 && + !record_opts.system_wide && !record_opts.cpu_list) usage_with_options(record_usage, record_options); if (force && append_file) { @@ -842,7 +758,7 @@ int cmd_record(int argc, const char **argv, const char *prefix __used) write_mode = WRITE_FORCE; } - if (nr_cgroups && !system_wide) { + if (nr_cgroups && !record_opts.system_wide) { fprintf(stderr, "cgroup monitoring only available in" " system-wide mode\n"); usage_with_options(record_usage, record_options); @@ -869,11 +785,11 @@ int cmd_record(int argc, const char **argv, const char *prefix __used) goto out_symbol_exit; } - if (target_pid != -1) - target_tid = target_pid; + if (record_opts.target_pid != -1) + record_opts.target_tid = record_opts.target_pid; - if (perf_evlist__create_maps(evsel_list, target_pid, - target_tid, cpu_list) < 0) + if (perf_evlist__create_maps(evsel_list, record_opts.target_pid, + record_opts.target_tid, record_opts.cpu_list) < 0) usage_with_options(record_usage, record_options); list_for_each_entry(pos, &evsel_list->entries, node) { @@ -887,18 +803,18 @@ int cmd_record(int argc, const char **argv, const char *prefix __used) if (perf_evlist__alloc_pollfd(evsel_list) < 0) goto out_free_fd; - if (user_interval != ULLONG_MAX) - default_interval = user_interval; - if (user_freq != UINT_MAX) - freq = user_freq; + if (record_opts.user_interval != ULLONG_MAX) + record_opts.default_interval = record_opts.user_interval; + if (record_opts.user_freq != UINT_MAX) + record_opts.freq = record_opts.user_freq; /* * User specified count overrides default frequency. */ - if (default_interval) - freq = 0; - else if (freq) { - default_interval = freq; + if (record_opts.default_interval) + record_opts.freq = 0; + else if (record_opts.freq) { + record_opts.default_interval = record_opts.freq; } else { fprintf(stderr, "frequency and count are zero, aborting\n"); err = -EINVAL; diff --git a/tools/perf/perf.h b/tools/perf/perf.h index 914c895510f7..eb6a13881887 100644 --- a/tools/perf/perf.h +++ b/tools/perf/perf.h @@ -185,4 +185,24 @@ extern const char perf_version_string[]; void pthread__unblock_sigwinch(void); +struct perf_record_opts { + pid_t target_pid; + pid_t target_tid; + bool call_graph; + bool inherit_stat; + bool no_delay; + bool no_inherit; + bool no_samples; + bool raw_samples; + bool sample_address; + bool sample_time; + bool sample_id_all_avail; + bool system_wide; + unsigned int freq; + unsigned int user_freq; + u64 default_interval; + u64 user_interval; + const char *cpu_list; +}; + #endif diff --git a/tools/perf/util/evlist.c b/tools/perf/util/evlist.c index 3bc5a287a9f9..b774341e797f 100644 --- a/tools/perf/util/evlist.c +++ b/tools/perf/util/evlist.c @@ -46,6 +46,22 @@ struct perf_evlist *perf_evlist__new(struct cpu_map *cpus, return evlist; } +void perf_evlist__config_attrs(struct perf_evlist *evlist, + struct perf_record_opts *opts) +{ + struct perf_evsel *evsel; + + if (evlist->cpus->map[0] < 0) + opts->no_inherit = true; + + list_for_each_entry(evsel, &evlist->entries, node) { + perf_evsel__config(evsel, opts); + + if (evlist->nr_entries > 1) + evsel->attr.sample_type |= PERF_SAMPLE_ID; + } +} + static void perf_evlist__purge(struct perf_evlist *evlist) { struct perf_evsel *pos, *n; diff --git a/tools/perf/util/evlist.h b/tools/perf/util/evlist.h index ec71c82935bd..231c06f8286b 100644 --- a/tools/perf/util/evlist.h +++ b/tools/perf/util/evlist.h @@ -10,6 +10,7 @@ struct pollfd; struct thread_map; struct cpu_map; +struct perf_record_opts; #define PERF_EVLIST__HLIST_BITS 8 #define PERF_EVLIST__HLIST_SIZE (1 << PERF_EVLIST__HLIST_BITS) @@ -64,6 +65,9 @@ union perf_event *perf_evlist__mmap_read(struct perf_evlist *self, int idx); int perf_evlist__open(struct perf_evlist *evlist, bool group); +void perf_evlist__config_attrs(struct perf_evlist *evlist, + struct perf_record_opts *opts); + int perf_evlist__alloc_mmap(struct perf_evlist *evlist); int perf_evlist__mmap(struct perf_evlist *evlist, int pages, bool overwrite); void perf_evlist__munmap(struct perf_evlist *evlist); diff --git a/tools/perf/util/evsel.c b/tools/perf/util/evsel.c index e42626422587..b38eaa34b28e 100644 --- a/tools/perf/util/evsel.c +++ b/tools/perf/util/evsel.c @@ -53,6 +53,76 @@ struct perf_evsel *perf_evsel__new(struct perf_event_attr *attr, int idx) return evsel; } +void perf_evsel__config(struct perf_evsel *evsel, struct perf_record_opts *opts) +{ + struct perf_event_attr *attr = &evsel->attr; + int track = !evsel->idx; /* only the first counter needs these */ + + attr->sample_id_all = opts->sample_id_all_avail ? 1 : 0; + attr->inherit = !opts->no_inherit; + attr->read_format = PERF_FORMAT_TOTAL_TIME_ENABLED | + PERF_FORMAT_TOTAL_TIME_RUNNING | + PERF_FORMAT_ID; + + attr->sample_type |= PERF_SAMPLE_IP | PERF_SAMPLE_TID; + + /* + * We default some events to a 1 default interval. But keep + * it a weak assumption overridable by the user. + */ + if (!attr->sample_period || (opts->user_freq != UINT_MAX && + opts->user_interval != ULLONG_MAX)) { + if (opts->freq) { + attr->sample_type |= PERF_SAMPLE_PERIOD; + attr->freq = 1; + attr->sample_freq = opts->freq; + } else { + attr->sample_period = opts->default_interval; + } + } + + if (opts->no_samples) + attr->sample_freq = 0; + + if (opts->inherit_stat) + attr->inherit_stat = 1; + + if (opts->sample_address) { + attr->sample_type |= PERF_SAMPLE_ADDR; + attr->mmap_data = track; + } + + if (opts->call_graph) + attr->sample_type |= PERF_SAMPLE_CALLCHAIN; + + if (opts->system_wide) + attr->sample_type |= PERF_SAMPLE_CPU; + + if (opts->sample_id_all_avail && + (opts->sample_time || opts->system_wide || + !opts->no_inherit || opts->cpu_list)) + attr->sample_type |= PERF_SAMPLE_TIME; + + if (opts->raw_samples) { + attr->sample_type |= PERF_SAMPLE_TIME; + attr->sample_type |= PERF_SAMPLE_RAW; + attr->sample_type |= PERF_SAMPLE_CPU; + } + + if (opts->no_delay) { + attr->watermark = 0; + attr->wakeup_events = 1; + } + + attr->mmap = track; + attr->comm = track; + + if (opts->target_pid == -1 && opts->target_tid == -1 && !opts->system_wide) { + attr->disabled = 1; + attr->enable_on_exec = 1; + } +} + int perf_evsel__alloc_fd(struct perf_evsel *evsel, int ncpus, int nthreads) { int cpu, thread; diff --git a/tools/perf/util/evsel.h b/tools/perf/util/evsel.h index b1d15e6f7ae3..6421c07f5015 100644 --- a/tools/perf/util/evsel.h +++ b/tools/perf/util/evsel.h @@ -67,6 +67,7 @@ struct perf_evsel { struct cpu_map; struct thread_map; struct perf_evlist; +struct perf_record_opts; struct perf_evsel *perf_evsel__new(struct perf_event_attr *attr, int idx); void perf_evsel__init(struct perf_evsel *evsel, @@ -74,6 +75,9 @@ void perf_evsel__init(struct perf_evsel *evsel, void perf_evsel__exit(struct perf_evsel *evsel); void perf_evsel__delete(struct perf_evsel *evsel); +void perf_evsel__config(struct perf_evsel *evsel, + struct perf_record_opts *opts); + int perf_evsel__alloc_fd(struct perf_evsel *evsel, int ncpus, int nthreads); int perf_evsel__alloc_id(struct perf_evsel *evsel, int ncpus, int nthreads); int perf_evsel__alloc_counts(struct perf_evsel *evsel, int ncpus); -- cgit v1.2.3 From 35b9d88ecd8c5fb720ba0dd325262f356d0b03e7 Mon Sep 17 00:00:00 2001 From: Arnaldo Carvalho de Melo Date: Wed, 9 Nov 2011 08:47:15 -0200 Subject: perf evlist: Introduce {prepare,start}_workload refactored from 'perf record' So that we can easily start a workload in other tools. Cc: David Ahern Cc: Frederic Weisbecker Cc: Mike Galbraith Cc: Paul Mackerras Cc: Peter Zijlstra Cc: Stephane Eranian Link: http://lkml.kernel.org/n/tip-zdsksd4aphu0nltg2lpwsw3x@git.kernel.org Signed-off-by: Arnaldo Carvalho de Melo --- tools/perf/builtin-record.c | 81 ++++++-------------------------------- tools/perf/perf.h | 1 + tools/perf/util/evlist.c | 96 +++++++++++++++++++++++++++++++++++++++++++++ tools/perf/util/evlist.h | 10 +++++ 4 files changed, 120 insertions(+), 68 deletions(-) diff --git a/tools/perf/builtin-record.c b/tools/perf/builtin-record.c index c3ac5415c097..4799195ed246 100644 --- a/tools/perf/builtin-record.c +++ b/tools/perf/builtin-record.c @@ -47,11 +47,9 @@ static struct perf_record_opts record_opts = { static unsigned int page_size; static unsigned int mmap_pages = UINT_MAX; static int output; -static int pipe_output = 0; static const char *output_name = NULL; static bool group = false; static int realtime_prio = 0; -static pid_t child_pid = -1; static enum write_mode_t write_mode = WRITE_FORCE; static bool no_buildid = false; static bool no_buildid_cache = false; @@ -144,9 +142,9 @@ static void sig_atexit(void) { int status; - if (child_pid > 0) { + if (evsel_list->workload.pid > 0) { if (!child_finished) - kill(child_pid, SIGTERM); + kill(evsel_list->workload.pid, SIGTERM); wait(&status); if (WIFSIGNALED(status)) @@ -304,7 +302,7 @@ static int process_buildids(void) static void atexit_header(void) { - if (!pipe_output) { + if (!record_opts.pipe_output) { session->header.data_size += bytes_written; if (!no_buildid) @@ -377,9 +375,7 @@ static int __cmd_record(int argc, const char **argv) int flags; int err; unsigned long waking = 0; - int child_ready_pipe[2], go_pipe[2]; const bool forks = argc > 0; - char buf; struct machine *machine; progname = argv[0]; @@ -391,20 +387,15 @@ static int __cmd_record(int argc, const char **argv) signal(SIGINT, sig_handler); signal(SIGUSR1, sig_handler); - if (forks && (pipe(child_ready_pipe) < 0 || pipe(go_pipe) < 0)) { - perror("failed to create pipes"); - exit(-1); - } - if (!output_name) { if (!fstat(STDOUT_FILENO, &st) && S_ISFIFO(st.st_mode)) - pipe_output = true; + record_opts.pipe_output = true; else output_name = "perf.data"; } if (output_name) { if (!strcmp(output_name, "-")) - pipe_output = true; + record_opts.pipe_output = true; else if (!stat(output_name, &st) && st.st_size) { if (write_mode == WRITE_FORCE) { char oldname[PATH_MAX]; @@ -424,7 +415,7 @@ static int __cmd_record(int argc, const char **argv) else flags |= O_TRUNC; - if (pipe_output) + if (record_opts.pipe_output) output = STDOUT_FILENO; else output = open(output_name, flags, S_IRUSR | S_IWUSR); @@ -470,57 +461,11 @@ static int __cmd_record(int argc, const char **argv) mmap_pages = (512 * 1024) / page_size; if (forks) { - child_pid = fork(); - if (child_pid < 0) { - perror("failed to fork"); - exit(-1); - } - - if (!child_pid) { - if (pipe_output) - dup2(2, 1); - close(child_ready_pipe[0]); - close(go_pipe[1]); - fcntl(go_pipe[0], F_SETFD, FD_CLOEXEC); - - /* - * Do a dummy execvp to get the PLT entry resolved, - * so we avoid the resolver overhead on the real - * execvp call. - */ - execvp("", (char **)argv); - - /* - * Tell the parent we're ready to go - */ - close(child_ready_pipe[1]); - - /* - * Wait until the parent tells us to go. - */ - if (read(go_pipe[0], &buf, 1) == -1) - perror("unable to read pipe"); - - execvp(argv[0], (char **)argv); - - perror(argv[0]); - kill(getppid(), SIGUSR1); - exit(-1); - } - - if (!record_opts.system_wide && record_opts.target_tid == -1 && record_opts.target_pid == -1) - evsel_list->threads->map[0] = child_pid; - - close(child_ready_pipe[1]); - close(go_pipe[0]); - /* - * wait for child to settle - */ - if (read(child_ready_pipe[0], &buf, 1) == -1) { - perror("unable to read pipe"); - exit(-1); + err = perf_evlist__prepare_workload(evsel_list, &record_opts, argv); + if (err < 0) { + pr_err("Couldn't run the workload!\n"); + goto out_delete_session; } - close(child_ready_pipe[0]); } open_counters(evsel_list); @@ -530,7 +475,7 @@ static int __cmd_record(int argc, const char **argv) */ atexit(atexit_header); - if (pipe_output) { + if (record_opts.pipe_output) { err = perf_header__write_pipe(output); if (err < 0) return err; @@ -543,7 +488,7 @@ static int __cmd_record(int argc, const char **argv) post_processing_offset = lseek(output, 0, SEEK_CUR); - if (pipe_output) { + if (record_opts.pipe_output) { err = perf_session__synthesize_attrs(session, process_synthesized_event); if (err < 0) { @@ -629,7 +574,7 @@ static int __cmd_record(int argc, const char **argv) * Let the child rip */ if (forks) - close(go_pipe[1]); + perf_evlist__start_workload(evsel_list); for (;;) { int hits = samples; diff --git a/tools/perf/perf.h b/tools/perf/perf.h index eb6a13881887..32ee6ca8eabd 100644 --- a/tools/perf/perf.h +++ b/tools/perf/perf.h @@ -193,6 +193,7 @@ struct perf_record_opts { bool no_delay; bool no_inherit; bool no_samples; + bool pipe_output; bool raw_samples; bool sample_address; bool sample_time; diff --git a/tools/perf/util/evlist.c b/tools/perf/util/evlist.c index b774341e797f..a472247af191 100644 --- a/tools/perf/util/evlist.c +++ b/tools/perf/util/evlist.c @@ -13,6 +13,7 @@ #include "thread_map.h" #include "evlist.h" #include "evsel.h" +#include #include "parse-events.h" @@ -33,6 +34,7 @@ void perf_evlist__init(struct perf_evlist *evlist, struct cpu_map *cpus, INIT_HLIST_HEAD(&evlist->heads[i]); INIT_LIST_HEAD(&evlist->entries); perf_evlist__set_maps(evlist, cpus, threads); + evlist->workload.pid = -1; } struct perf_evlist *perf_evlist__new(struct cpu_map *cpus, @@ -674,3 +676,97 @@ out_err: return err; } + +int perf_evlist__prepare_workload(struct perf_evlist *evlist, + struct perf_record_opts *opts, + const char *argv[]) +{ + int child_ready_pipe[2], go_pipe[2]; + char bf; + + if (pipe(child_ready_pipe) < 0) { + perror("failed to create 'ready' pipe"); + return -1; + } + + if (pipe(go_pipe) < 0) { + perror("failed to create 'go' pipe"); + goto out_close_ready_pipe; + } + + evlist->workload.pid = fork(); + if (evlist->workload.pid < 0) { + perror("failed to fork"); + goto out_close_pipes; + } + + if (!evlist->workload.pid) { + if (opts->pipe_output) + dup2(2, 1); + + close(child_ready_pipe[0]); + close(go_pipe[1]); + fcntl(go_pipe[0], F_SETFD, FD_CLOEXEC); + + /* + * Do a dummy execvp to get the PLT entry resolved, + * so we avoid the resolver overhead on the real + * execvp call. + */ + execvp("", (char **)argv); + + /* + * Tell the parent we're ready to go + */ + close(child_ready_pipe[1]); + + /* + * Wait until the parent tells us to go. + */ + if (read(go_pipe[0], &bf, 1) == -1) + perror("unable to read pipe"); + + execvp(argv[0], (char **)argv); + + perror(argv[0]); + kill(getppid(), SIGUSR1); + exit(-1); + } + + if (!opts->system_wide && opts->target_tid == -1 && opts->target_pid == -1) + evlist->threads->map[0] = evlist->workload.pid; + + close(child_ready_pipe[1]); + close(go_pipe[0]); + /* + * wait for child to settle + */ + if (read(child_ready_pipe[0], &bf, 1) == -1) { + perror("unable to read pipe"); + goto out_close_pipes; + } + + evlist->workload.cork_fd = go_pipe[1]; + close(child_ready_pipe[0]); + return 0; + +out_close_pipes: + close(go_pipe[0]); + close(go_pipe[1]); +out_close_ready_pipe: + close(child_ready_pipe[0]); + close(child_ready_pipe[1]); + return -1; +} + +int perf_evlist__start_workload(struct perf_evlist *evlist) +{ + if (evlist->workload.cork_fd > 0) { + /* + * Remove the cork, let it rip! + */ + return close(evlist->workload.cork_fd); + } + + return 0; +} diff --git a/tools/perf/util/evlist.h b/tools/perf/util/evlist.h index 231c06f8286b..07d56b3e6d61 100644 --- a/tools/perf/util/evlist.h +++ b/tools/perf/util/evlist.h @@ -6,6 +6,7 @@ #include "../perf.h" #include "event.h" #include "util.h" +#include struct pollfd; struct thread_map; @@ -22,6 +23,10 @@ struct perf_evlist { int nr_fds; int nr_mmaps; int mmap_len; + struct { + int cork_fd; + pid_t pid; + } workload; bool overwrite; union perf_event event_copy; struct perf_mmap *mmap; @@ -68,6 +73,11 @@ int perf_evlist__open(struct perf_evlist *evlist, bool group); void perf_evlist__config_attrs(struct perf_evlist *evlist, struct perf_record_opts *opts); +int perf_evlist__prepare_workload(struct perf_evlist *evlist, + struct perf_record_opts *opts, + const char *argv[]); +int perf_evlist__start_workload(struct perf_evlist *evlist); + int perf_evlist__alloc_mmap(struct perf_evlist *evlist); int perf_evlist__mmap(struct perf_evlist *evlist, int pages, bool overwrite); void perf_evlist__munmap(struct perf_evlist *evlist); -- cgit v1.2.3 From 50a682ce875f91e04c81ae5680535aba5a770bad Mon Sep 17 00:00:00 2001 From: Arnaldo Carvalho de Melo Date: Wed, 9 Nov 2011 09:10:47 -0200 Subject: perf evlist: Handle default value for 'pages' on mmap method Every tool that calls this and allows the user to override the value needs this logic. Cc: David Ahern Cc: Frederic Weisbecker Cc: Mike Galbraith Cc: Paul Mackerras Cc: Peter Zijlstra Cc: Stephane Eranian Link: http://lkml.kernel.org/n/tip-lwscxpg57xfzahz5dmdfp9uz@git.kernel.org Signed-off-by: Arnaldo Carvalho de Melo --- tools/perf/builtin-record.c | 4 ---- tools/perf/util/evlist.c | 12 +++++++++--- tools/perf/util/evlist.h | 3 ++- 3 files changed, 11 insertions(+), 8 deletions(-) diff --git a/tools/perf/builtin-record.c b/tools/perf/builtin-record.c index 4799195ed246..749862d57a83 100644 --- a/tools/perf/builtin-record.c +++ b/tools/perf/builtin-record.c @@ -456,10 +456,6 @@ static int __cmd_record(int argc, const char **argv) perf_header__set_feat(&session->header, HEADER_NUMA_TOPOLOGY); perf_header__set_feat(&session->header, HEADER_CPUID); - /* 512 kiB: default amount of unprivileged mlocked memory */ - if (mmap_pages == UINT_MAX) - mmap_pages = (512 * 1024) / page_size; - if (forks) { err = perf_evlist__prepare_workload(evsel_list, &record_opts, argv); if (err < 0) { diff --git a/tools/perf/util/evlist.c b/tools/perf/util/evlist.c index a472247af191..81237dcde1fd 100644 --- a/tools/perf/util/evlist.c +++ b/tools/perf/util/evlist.c @@ -507,14 +507,20 @@ out_unmap: * * Using perf_evlist__read_on_cpu does this automatically. */ -int perf_evlist__mmap(struct perf_evlist *evlist, int pages, bool overwrite) +int perf_evlist__mmap(struct perf_evlist *evlist, unsigned int pages, + bool overwrite) { unsigned int page_size = sysconf(_SC_PAGE_SIZE); - int mask = pages * page_size - 1; struct perf_evsel *evsel; const struct cpu_map *cpus = evlist->cpus; const struct thread_map *threads = evlist->threads; - int prot = PROT_READ | (overwrite ? 0 : PROT_WRITE); + int prot = PROT_READ | (overwrite ? 0 : PROT_WRITE), mask; + + /* 512 kiB: default amount of unprivileged mlocked memory */ + if (pages == UINT_MAX) + pages = (512 * 1024) / page_size; + + mask = pages * page_size - 1; if (evlist->mmap == NULL && perf_evlist__alloc_mmap(evlist) < 0) return -ENOMEM; diff --git a/tools/perf/util/evlist.h b/tools/perf/util/evlist.h index 07d56b3e6d61..17e9c80243ce 100644 --- a/tools/perf/util/evlist.h +++ b/tools/perf/util/evlist.h @@ -79,7 +79,8 @@ int perf_evlist__prepare_workload(struct perf_evlist *evlist, int perf_evlist__start_workload(struct perf_evlist *evlist); int perf_evlist__alloc_mmap(struct perf_evlist *evlist); -int perf_evlist__mmap(struct perf_evlist *evlist, int pages, bool overwrite); +int perf_evlist__mmap(struct perf_evlist *evlist, unsigned int pages, + bool overwrite); void perf_evlist__munmap(struct perf_evlist *evlist); void perf_evlist__disable(struct perf_evlist *evlist); -- cgit v1.2.3 From 01c2d99bcf6fc7f6ce3fe3d0fb38b124e1f127fc Mon Sep 17 00:00:00 2001 From: Arnaldo Carvalho de Melo Date: Wed, 9 Nov 2011 09:16:26 -0200 Subject: perf record: Move mmap_pages to perf_record_opts Tools being developed will need this to allow the user to override this value. Cc: David Ahern Cc: Frederic Weisbecker Cc: Mike Galbraith Cc: Paul Mackerras Cc: Peter Zijlstra Cc: Stephane Eranian Link: http://lkml.kernel.org/n/tip-zydc1yhxfm0z35fuy95bsn1l@git.kernel.org Signed-off-by: Arnaldo Carvalho de Melo --- tools/perf/builtin-record.c | 9 +++++---- tools/perf/perf.h | 1 + 2 files changed, 6 insertions(+), 4 deletions(-) diff --git a/tools/perf/builtin-record.c b/tools/perf/builtin-record.c index 749862d57a83..ffb627d40210 100644 --- a/tools/perf/builtin-record.c +++ b/tools/perf/builtin-record.c @@ -35,9 +35,10 @@ enum write_mode_t { WRITE_APPEND }; -static struct perf_record_opts record_opts = { +struct perf_record_opts record_opts = { .target_pid = -1, .target_tid = -1, + .mmap_pages = UINT_MAX, .user_freq = UINT_MAX, .user_interval = ULLONG_MAX, .freq = 1000, @@ -45,7 +46,6 @@ static struct perf_record_opts record_opts = { }; static unsigned int page_size; -static unsigned int mmap_pages = UINT_MAX; static int output; static const char *output_name = NULL; static bool group = false; @@ -272,7 +272,7 @@ try_again: exit(-1); } - if (perf_evlist__mmap(evlist, mmap_pages, false) < 0) + if (perf_evlist__mmap(evlist, record_opts.mmap_pages, false) < 0) die("failed to mmap with %d (%s)\n", errno, strerror(errno)); if (file_new) @@ -647,7 +647,8 @@ const struct option record_options[] = { OPT_BOOLEAN('i', "no-inherit", &record_opts.no_inherit, "child tasks do not inherit counters"), OPT_UINTEGER('F', "freq", &record_opts.user_freq, "profile at this frequency"), - OPT_UINTEGER('m', "mmap-pages", &mmap_pages, "number of mmap data pages"), + OPT_UINTEGER('m', "mmap-pages", &record_opts.mmap_pages, + "number of mmap data pages"), OPT_BOOLEAN(0, "group", &group, "put the counters into a counter group"), OPT_BOOLEAN('g', "call-graph", &record_opts.call_graph, diff --git a/tools/perf/perf.h b/tools/perf/perf.h index 32ee6ca8eabd..13c42f34e8b3 100644 --- a/tools/perf/perf.h +++ b/tools/perf/perf.h @@ -200,6 +200,7 @@ struct perf_record_opts { bool sample_id_all_avail; bool system_wide; unsigned int freq; + unsigned int mmap_pages; unsigned int user_freq; u64 default_interval; u64 user_interval; -- cgit v1.2.3 From b424eba27160dd19577896d4520b8eebabed919f Mon Sep 17 00:00:00 2001 From: Arnaldo Carvalho de Melo Date: Wed, 9 Nov 2011 13:24:25 -0200 Subject: perf session: Move threads to struct machine The 'machine' abstraction was introduced with 'perf kvm' where we could have samples for the host and multiple guests, but at the time we ended up keeping the list of all machines threads all in session->host_machine. Move the threads rb_tree to struct machine to separate the namespaces. Cc: David Ahern Cc: Frederic Weisbecker Cc: Mike Galbraith Cc: Paul Mackerras Cc: Peter Zijlstra Cc: Stephane Eranian Link: http://lkml.kernel.org/n/tip-mdg7sm6j3va09vtgj49gbsrp@git.kernel.org Signed-off-by: Arnaldo Carvalho de Melo --- tools/perf/util/build-id.c | 4 ++-- tools/perf/util/map.c | 4 ++++ tools/perf/util/map.h | 9 +++++++++ tools/perf/util/session.c | 47 +++++++++++++++++++++++++++++++++++++++------- tools/perf/util/session.h | 3 --- tools/perf/util/thread.c | 6 +++--- 6 files changed, 58 insertions(+), 15 deletions(-) diff --git a/tools/perf/util/build-id.c b/tools/perf/util/build-id.c index a91cd99f26ea..f2fe6ec08945 100644 --- a/tools/perf/util/build-id.c +++ b/tools/perf/util/build-id.c @@ -48,8 +48,8 @@ static int perf_event__exit_del_thread(union perf_event *event, event->fork.ppid, event->fork.ptid); if (thread) { - rb_erase(&thread->rb_node, &session->threads); - session->last_match = NULL; + rb_erase(&thread->rb_node, &session->host_machine.threads); + session->host_machine.last_match = NULL; thread__delete(thread); } diff --git a/tools/perf/util/map.c b/tools/perf/util/map.c index 78284b13e808..316aa0ab7122 100644 --- a/tools/perf/util/map.c +++ b/tools/perf/util/map.c @@ -562,6 +562,10 @@ int machine__init(struct machine *self, const char *root_dir, pid_t pid) INIT_LIST_HEAD(&self->user_dsos); INIT_LIST_HEAD(&self->kernel_dsos); + self->threads = RB_ROOT; + INIT_LIST_HEAD(&self->dead_threads); + self->last_match = NULL; + self->kmaps.machine = self; self->pid = pid; self->root_dir = strdup(root_dir); diff --git a/tools/perf/util/map.h b/tools/perf/util/map.h index 890d85545d0f..bde6835ee257 100644 --- a/tools/perf/util/map.h +++ b/tools/perf/util/map.h @@ -62,6 +62,9 @@ struct machine { struct rb_node rb_node; pid_t pid; char *root_dir; + struct rb_root threads; + struct list_head dead_threads; + struct thread *last_match; struct list_head user_dsos; struct list_head kernel_dsos; struct map_groups kmaps; @@ -190,6 +193,12 @@ struct symbol *map_groups__find_symbol_by_name(struct map_groups *mg, struct map **mapp, symbol_filter_t filter); + +struct thread *machine__findnew_thread(struct machine *machine, pid_t pid); +void machine__remove_thread(struct machine *machine, struct thread *th); + +size_t machine__fprintf(struct machine *machine, FILE *fp); + static inline struct symbol *machine__find_kernel_symbol(struct machine *self, enum map_type type, u64 addr, diff --git a/tools/perf/util/session.c b/tools/perf/util/session.c index 85c1e6b76f0a..a76666f17767 100644 --- a/tools/perf/util/session.c +++ b/tools/perf/util/session.c @@ -139,9 +139,6 @@ struct perf_session *perf_session__new(const char *filename, int mode, goto out; memcpy(self->filename, filename, len); - self->threads = RB_ROOT; - INIT_LIST_HEAD(&self->dead_threads); - self->last_match = NULL; /* * On 64bit we can mmap the data file in one go. No need for tiny mmap * slices. On 32bit we use 32MB. @@ -184,17 +181,22 @@ out_delete: return NULL; } -static void perf_session__delete_dead_threads(struct perf_session *self) +static void machine__delete_dead_threads(struct machine *machine) { struct thread *n, *t; - list_for_each_entry_safe(t, n, &self->dead_threads, node) { + list_for_each_entry_safe(t, n, &machine->dead_threads, node) { list_del(&t->node); thread__delete(t); } } -static void perf_session__delete_threads(struct perf_session *self) +static void perf_session__delete_dead_threads(struct perf_session *session) +{ + machine__delete_dead_threads(&session->host_machine); +} + +static void machine__delete_threads(struct machine *self) { struct rb_node *nd = rb_first(&self->threads); @@ -207,6 +209,11 @@ static void perf_session__delete_threads(struct perf_session *self) } } +static void perf_session__delete_threads(struct perf_session *session) +{ + machine__delete_threads(&session->host_machine); +} + void perf_session__delete(struct perf_session *self) { perf_session__destroy_kernel_maps(self); @@ -217,7 +224,7 @@ void perf_session__delete(struct perf_session *self) free(self); } -void perf_session__remove_thread(struct perf_session *self, struct thread *th) +void machine__remove_thread(struct machine *self, struct thread *th) { self->last_match = NULL; rb_erase(&th->rb_node, &self->threads); @@ -884,6 +891,11 @@ void perf_event_header__bswap(struct perf_event_header *self) self->size = bswap_16(self->size); } +struct thread *perf_session__findnew(struct perf_session *session, pid_t pid) +{ + return machine__findnew_thread(&session->host_machine, pid); +} + static struct thread *perf_session__register_idle_thread(struct perf_session *self) { struct thread *thread = perf_session__findnew(self, 0); @@ -1224,6 +1236,27 @@ size_t perf_session__fprintf_nr_events(struct perf_session *session, FILE *fp) return ret; } +size_t perf_session__fprintf(struct perf_session *session, FILE *fp) +{ + /* + * FIXME: Here we have to actually print all the machines in this + * session, not just the host... + */ + return machine__fprintf(&session->host_machine, fp); +} + +void perf_session__remove_thread(struct perf_session *session, + struct thread *th) +{ + /* + * FIXME: This one makes no sense, we need to remove the thread from + * the machine it belongs to, perf_session can have many machines, so + * doing it always on ->host_machine is wrong. Fix when auditing all + * the 'perf kvm' code. + */ + machine__remove_thread(&session->host_machine, th); +} + struct perf_evsel *perf_session__find_first_evtype(struct perf_session *session, unsigned int type) { diff --git a/tools/perf/util/session.h b/tools/perf/util/session.h index 6e393c98eb34..76d462d3bef7 100644 --- a/tools/perf/util/session.h +++ b/tools/perf/util/session.h @@ -30,9 +30,6 @@ struct perf_session { struct perf_header header; unsigned long size; unsigned long mmap_window; - struct rb_root threads; - struct list_head dead_threads; - struct thread *last_match; struct machine host_machine; struct rb_root machines; struct perf_evlist *evlist; diff --git a/tools/perf/util/thread.c b/tools/perf/util/thread.c index d5d3b22250f3..fb4b7ea6752f 100644 --- a/tools/perf/util/thread.c +++ b/tools/perf/util/thread.c @@ -61,7 +61,7 @@ static size_t thread__fprintf(struct thread *self, FILE *fp) map_groups__fprintf(&self->mg, verbose, fp); } -struct thread *perf_session__findnew(struct perf_session *self, pid_t pid) +struct thread *machine__findnew_thread(struct machine *self, pid_t pid) { struct rb_node **p = &self->threads.rb_node; struct rb_node *parent = NULL; @@ -125,12 +125,12 @@ int thread__fork(struct thread *self, struct thread *parent) return 0; } -size_t perf_session__fprintf(struct perf_session *self, FILE *fp) +size_t machine__fprintf(struct machine *machine, FILE *fp) { size_t ret = 0; struct rb_node *nd; - for (nd = rb_first(&self->threads); nd; nd = rb_next(nd)) { + for (nd = rb_first(&machine->threads); nd; nd = rb_next(nd)) { struct thread *pos = rb_entry(nd, struct thread, rb_node); ret += thread__fprintf(pos, fp); -- cgit v1.2.3 From ed80f5813fd6ecc6d74250681910a4214f699d4e Mon Sep 17 00:00:00 2001 From: Arnaldo Carvalho de Melo Date: Fri, 11 Nov 2011 15:12:56 -0200 Subject: perf record: Move 'group' to perf_event_ops Will be used in other tools to share the command line parsing code. Cc: David Ahern Cc: Frederic Weisbecker Cc: Mike Galbraith Cc: Paul Mackerras Cc: Peter Zijlstra Cc: Stephane Eranian Link: http://lkml.kernel.org/n/tip-8x0yr77r6lrd2t699s499m8n@git.kernel.org Signed-off-by: Arnaldo Carvalho de Melo --- tools/perf/builtin-record.c | 9 ++++----- tools/perf/perf.h | 1 + 2 files changed, 5 insertions(+), 5 deletions(-) diff --git a/tools/perf/builtin-record.c b/tools/perf/builtin-record.c index ffb627d40210..ba6777a147ca 100644 --- a/tools/perf/builtin-record.c +++ b/tools/perf/builtin-record.c @@ -48,7 +48,6 @@ struct perf_record_opts record_opts = { static unsigned int page_size; static int output; static const char *output_name = NULL; -static bool group = false; static int realtime_prio = 0; static enum write_mode_t write_mode = WRITE_FORCE; static bool no_buildid = false; @@ -202,13 +201,13 @@ static void open_counters(struct perf_evlist *evlist) */ bool time_needed = attr->sample_type & PERF_SAMPLE_TIME; - if (group && pos != first) + if (record_opts.group && pos != first) group_fd = first->fd; retry_sample_id: attr->sample_id_all = record_opts.sample_id_all_avail ? 1 : 0; try_again: - if (perf_evsel__open(pos, evlist->cpus, evlist->threads, group, - group_fd) < 0) { + if (perf_evsel__open(pos, evlist->cpus, evlist->threads, + record_opts.group, group_fd) < 0) { int err = errno; if (err == EPERM || err == EACCES) { @@ -649,7 +648,7 @@ const struct option record_options[] = { OPT_UINTEGER('F', "freq", &record_opts.user_freq, "profile at this frequency"), OPT_UINTEGER('m', "mmap-pages", &record_opts.mmap_pages, "number of mmap data pages"), - OPT_BOOLEAN(0, "group", &group, + OPT_BOOLEAN(0, "group", &record_opts.group, "put the counters into a counter group"), OPT_BOOLEAN('g', "call-graph", &record_opts.call_graph, "do call-graph (stack chain/backtrace) recording"), diff --git a/tools/perf/perf.h b/tools/perf/perf.h index 13c42f34e8b3..ea804f5a8cc2 100644 --- a/tools/perf/perf.h +++ b/tools/perf/perf.h @@ -189,6 +189,7 @@ struct perf_record_opts { pid_t target_pid; pid_t target_tid; bool call_graph; + bool group; bool inherit_stat; bool no_delay; bool no_inherit; -- cgit v1.2.3 From 18b552350515188a732db6ccdb81e9cefb8b58c9 Mon Sep 17 00:00:00 2001 From: Arnaldo Carvalho de Melo Date: Fri, 11 Nov 2011 22:08:07 -0200 Subject: perf ui progress: Fix divide by zero Happens in a perf.data file where one of the events had no samples. Cc: David Ahern Cc: Frederic Weisbecker Cc: Mike Galbraith Cc: Paul Mackerras Cc: Peter Zijlstra Cc: Stephane Eranian Link: http://lkml.kernel.org/n/tip-j7st3oyiotvfxqde2nc41kxb@git.kernel.org Signed-off-by: Arnaldo Carvalho de Melo --- tools/perf/util/ui/progress.c | 3 +++ 1 file changed, 3 insertions(+) diff --git a/tools/perf/util/ui/progress.c b/tools/perf/util/ui/progress.c index 295e366b6311..13aa64e50e11 100644 --- a/tools/perf/util/ui/progress.c +++ b/tools/perf/util/ui/progress.c @@ -14,6 +14,9 @@ void ui_progress__update(u64 curr, u64 total, const char *title) if (use_browser <= 0) return; + if (total == 0) + return; + ui__refresh_dimensions(true); pthread_mutex_lock(&ui__lock); y = SLtt_Screen_Rows / 2 - 2; -- cgit v1.2.3 From d04b35f8085f0d4c5c874515b8f65e7664357148 Mon Sep 17 00:00:00 2001 From: Arnaldo Carvalho de Melo Date: Fri, 11 Nov 2011 22:17:32 -0200 Subject: perf symbols: Add nr_events to symbol_conf Since symbol__alloc_hists need it, to avoid passing it around in many functions have it in the symbol_conf struct. Cc: David Ahern Cc: Frederic Weisbecker Cc: Mike Galbraith Cc: Paul Mackerras Cc: Peter Zijlstra Cc: Stephane Eranian Link: http://lkml.kernel.org/n/tip-cwv8ysvpywzjq4v3xtbd4zwv@git.kernel.org Signed-off-by: Arnaldo Carvalho de Melo --- tools/perf/builtin-annotate.c | 22 ++++++++-------------- tools/perf/builtin-report.c | 3 +-- tools/perf/builtin-top.c | 7 ++++--- tools/perf/util/annotate.c | 6 +++--- tools/perf/util/annotate.h | 5 ++--- tools/perf/util/header.c | 2 ++ tools/perf/util/hist.h | 3 +-- tools/perf/util/symbol.h | 1 + tools/perf/util/ui/browsers/annotate.c | 16 +++++++--------- tools/perf/util/ui/browsers/hists.c | 2 +- 10 files changed, 30 insertions(+), 37 deletions(-) diff --git a/tools/perf/builtin-annotate.c b/tools/perf/builtin-annotate.c index 46b4c24f338e..8b9091bce98d 100644 --- a/tools/perf/builtin-annotate.c +++ b/tools/perf/builtin-annotate.c @@ -43,10 +43,9 @@ static const char *sym_hist_filter; static const char *cpu_list; static DECLARE_BITMAP(cpu_bitmap, MAX_NR_CPUS); -static int perf_evlist__add_sample(struct perf_evlist *evlist, - struct perf_sample *sample, - struct perf_evsel *evsel, - struct addr_location *al) +static int perf_evsel__add_sample(struct perf_evsel *evsel, + struct perf_sample *sample, + struct addr_location *al) { struct hist_entry *he; int ret; @@ -69,8 +68,7 @@ static int perf_evlist__add_sample(struct perf_evlist *evlist, ret = 0; if (he->ms.sym != NULL) { struct annotation *notes = symbol__annotation(he->ms.sym); - if (notes->src == NULL && - symbol__alloc_hist(he->ms.sym, evlist->nr_entries) < 0) + if (notes->src == NULL && symbol__alloc_hist(he->ms.sym) < 0) return -ENOMEM; ret = hist_entry__inc_addr_samples(he, evsel->idx, al->addr); @@ -98,8 +96,7 @@ static int process_sample_event(union perf_event *event, if (cpu_list && !test_bit(sample->cpu, cpu_bitmap)) return 0; - if (!al.filtered && - perf_evlist__add_sample(session->evlist, sample, evsel, &al)) { + if (!al.filtered && perf_evsel__add_sample(evsel, sample, &al)) { pr_warning("problem incrementing symbol count, " "skipping event\n"); return -1; @@ -114,8 +111,7 @@ static int hist_entry__tty_annotate(struct hist_entry *he, int evidx) print_line, full_paths, 0, 0); } -static void hists__find_annotations(struct hists *self, int evidx, - int nr_events) +static void hists__find_annotations(struct hists *self, int evidx) { struct rb_node *nd = rb_first(&self->entries), *next; int key = K_RIGHT; @@ -138,8 +134,7 @@ find_next: } if (use_browser > 0) { - key = hist_entry__tui_annotate(he, evidx, nr_events, - NULL, NULL, 0); + key = hist_entry__tui_annotate(he, evidx, NULL, NULL, 0); switch (key) { case K_RIGHT: next = rb_next(nd); @@ -217,8 +212,7 @@ static int __cmd_annotate(void) total_nr_samples += nr_samples; hists__collapse_resort(hists); hists__output_resort(hists); - hists__find_annotations(hists, pos->idx, - session->evlist->nr_entries); + hists__find_annotations(hists, pos->idx); } } diff --git a/tools/perf/builtin-report.c b/tools/perf/builtin-report.c index 4d7c8340c326..758a287fc07a 100644 --- a/tools/perf/builtin-report.c +++ b/tools/perf/builtin-report.c @@ -92,8 +92,7 @@ static int perf_session__add_hist_entry(struct perf_session *session, assert(evsel != NULL); err = -ENOMEM; - if (notes->src == NULL && - symbol__alloc_hist(he->ms.sym, session->evlist->nr_entries) < 0) + if (notes->src == NULL && symbol__alloc_hist(he->ms.sym) < 0) goto out; err = hist_entry__inc_addr_samples(he, evsel->idx, al->addr); diff --git a/tools/perf/builtin-top.c b/tools/perf/builtin-top.c index c9cdedb58134..04288ee223ed 100644 --- a/tools/perf/builtin-top.c +++ b/tools/perf/builtin-top.c @@ -170,7 +170,7 @@ static int parse_source(struct hist_entry *he) pthread_mutex_lock(¬es->lock); - if (symbol__alloc_hist(sym, top.evlist->nr_entries) < 0) { + if (symbol__alloc_hist(sym) < 0) { pthread_mutex_unlock(¬es->lock); pr_err("Not enough memory for annotating '%s' symbol!\n", sym->name); @@ -210,8 +210,7 @@ static void record_precise_ip(struct hist_entry *he, int counter, u64 ip) if (pthread_mutex_trylock(¬es->lock)) return; - if (notes->src == NULL && - symbol__alloc_hist(sym, top.evlist->nr_entries) < 0) { + if (notes->src == NULL && symbol__alloc_hist(sym) < 0) { pthread_mutex_unlock(¬es->lock); pr_err("Not enough memory for annotating '%s' symbol!\n", sym->name); @@ -1215,6 +1214,8 @@ int cmd_top(int argc, const char **argv, const char *prefix __used) return -ENOMEM; } + symbol_conf.nr_events = top.evlist->nr_entries; + if (top.delay_secs < 1) top.delay_secs = 1; diff --git a/tools/perf/util/annotate.c b/tools/perf/util/annotate.c index 119e996035c8..376e643f7066 100644 --- a/tools/perf/util/annotate.c +++ b/tools/perf/util/annotate.c @@ -25,17 +25,17 @@ int symbol__annotate_init(struct map *map __used, struct symbol *sym) return 0; } -int symbol__alloc_hist(struct symbol *sym, int nevents) +int symbol__alloc_hist(struct symbol *sym) { struct annotation *notes = symbol__annotation(sym); size_t sizeof_sym_hist = (sizeof(struct sym_hist) + (sym->end - sym->start) * sizeof(u64)); - notes->src = zalloc(sizeof(*notes->src) + nevents * sizeof_sym_hist); + notes->src = zalloc(sizeof(*notes->src) + symbol_conf.nr_events * sizeof_sym_hist); if (notes->src == NULL) return -1; notes->src->sizeof_sym_hist = sizeof_sym_hist; - notes->src->nr_histograms = nevents; + notes->src->nr_histograms = symbol_conf.nr_events; INIT_LIST_HEAD(¬es->src->source); return 0; } diff --git a/tools/perf/util/annotate.h b/tools/perf/util/annotate.h index d9072523d342..efa5dc82bfae 100644 --- a/tools/perf/util/annotate.h +++ b/tools/perf/util/annotate.h @@ -72,7 +72,7 @@ static inline struct annotation *symbol__annotation(struct symbol *sym) int symbol__inc_addr_samples(struct symbol *sym, struct map *map, int evidx, u64 addr); -int symbol__alloc_hist(struct symbol *sym, int nevents); +int symbol__alloc_hist(struct symbol *sym); void symbol__annotate_zero_histograms(struct symbol *sym); int symbol__annotate(struct symbol *sym, struct map *map, size_t privsize); @@ -99,8 +99,7 @@ static inline int symbol__tui_annotate(struct symbol *sym __used, } #else int symbol__tui_annotate(struct symbol *sym, struct map *map, int evidx, - int nr_events, void(*timer)(void *arg), void *arg, - int delay_secs); + void(*timer)(void *arg), void *arg, int delay_secs); #endif extern const char *disassembler_style; diff --git a/tools/perf/util/header.c b/tools/perf/util/header.c index bcd05d05b4f0..41424a16be8e 100644 --- a/tools/perf/util/header.c +++ b/tools/perf/util/header.c @@ -2041,6 +2041,8 @@ int perf_session__read_header(struct perf_session *session, int fd) lseek(fd, tmp, SEEK_SET); } + symbol_conf.nr_events = nr_attrs; + if (f_header.event_types.size) { lseek(fd, f_header.event_types.offset, SEEK_SET); events = malloc(f_header.event_types.size); diff --git a/tools/perf/util/hist.h b/tools/perf/util/hist.h index c86c1d27bd1e..6676d558b2a7 100644 --- a/tools/perf/util/hist.h +++ b/tools/perf/util/hist.h @@ -119,7 +119,6 @@ int perf_evlist__tui_browse_hists(struct perf_evlist *evlist __used, static inline int hist_entry__tui_annotate(struct hist_entry *self __used, int evidx __used, - int nr_events __used, void(*timer)(void *arg) __used, void *arg __used, int delay_secs __used) @@ -130,7 +129,7 @@ static inline int hist_entry__tui_annotate(struct hist_entry *self __used, #define K_RIGHT -2 #else #include "ui/keysyms.h" -int hist_entry__tui_annotate(struct hist_entry *he, int evidx, int nr_events, +int hist_entry__tui_annotate(struct hist_entry *he, int evidx, void(*timer)(void *arg), void *arg, int delay_secs); int perf_evlist__tui_browse_hists(struct perf_evlist *evlist, const char *help, diff --git a/tools/perf/util/symbol.h b/tools/perf/util/symbol.h index 29f8d742e92f..123c2e14353e 100644 --- a/tools/perf/util/symbol.h +++ b/tools/perf/util/symbol.h @@ -68,6 +68,7 @@ struct strlist; struct symbol_conf { unsigned short priv_size; + unsigned short nr_events; bool try_vmlinux_path, use_modules, sort_by_name, diff --git a/tools/perf/util/ui/browsers/annotate.c b/tools/perf/util/ui/browsers/annotate.c index 0575905d1205..295a9c93f945 100644 --- a/tools/perf/util/ui/browsers/annotate.c +++ b/tools/perf/util/ui/browsers/annotate.c @@ -224,7 +224,7 @@ static bool annotate_browser__toggle_source(struct annotate_browser *browser) } static int annotate_browser__run(struct annotate_browser *self, int evidx, - int nr_events, void(*timer)(void *arg), + void(*timer)(void *arg), void *arg, int delay_secs) { struct rb_node *nd = NULL; @@ -328,8 +328,7 @@ static int annotate_browser__run(struct annotate_browser *self, int evidx, notes = symbol__annotation(target); pthread_mutex_lock(¬es->lock); - if (notes->src == NULL && - symbol__alloc_hist(target, nr_events) < 0) { + if (notes->src == NULL && symbol__alloc_hist(target) < 0) { pthread_mutex_unlock(¬es->lock); ui__warning("Not enough memory for annotating '%s' symbol!\n", target->name); @@ -337,7 +336,7 @@ static int annotate_browser__run(struct annotate_browser *self, int evidx, } pthread_mutex_unlock(¬es->lock); - symbol__tui_annotate(target, ms->map, evidx, nr_events, + symbol__tui_annotate(target, ms->map, evidx, timer, arg, delay_secs); } continue; @@ -358,15 +357,15 @@ out: return key; } -int hist_entry__tui_annotate(struct hist_entry *he, int evidx, int nr_events, +int hist_entry__tui_annotate(struct hist_entry *he, int evidx, void(*timer)(void *arg), void *arg, int delay_secs) { - return symbol__tui_annotate(he->ms.sym, he->ms.map, evidx, nr_events, + return symbol__tui_annotate(he->ms.sym, he->ms.map, evidx, timer, arg, delay_secs); } int symbol__tui_annotate(struct symbol *sym, struct map *map, int evidx, - int nr_events, void(*timer)(void *arg), void *arg, + void(*timer)(void *arg), void *arg, int delay_secs) { struct objdump_line *pos, *n; @@ -419,8 +418,7 @@ int symbol__tui_annotate(struct symbol *sym, struct map *map, int evidx, browser.b.nr_entries = browser.nr_entries; browser.b.entries = ¬es->src->source, browser.b.width += 18; /* Percentage */ - ret = annotate_browser__run(&browser, evidx, nr_events, - timer, arg, delay_secs); + ret = annotate_browser__run(&browser, evidx, timer, arg, delay_secs); list_for_each_entry_safe(pos, n, ¬es->src->source, node) { list_del(&pos->node); objdump_line__free(pos); diff --git a/tools/perf/util/ui/browsers/hists.c b/tools/perf/util/ui/browsers/hists.c index d0c94b459685..1212a386a033 100644 --- a/tools/perf/util/ui/browsers/hists.c +++ b/tools/perf/util/ui/browsers/hists.c @@ -1020,7 +1020,7 @@ do_annotate: * Don't let this be freed, say, by hists__decay_entry. */ he->used = true; - err = hist_entry__tui_annotate(he, evsel->idx, nr_events, + err = hist_entry__tui_annotate(he, evsel->idx, timer, arg, delay_secs); he->used = false; ui_browser__update_nr_entries(&browser->b, browser->hists->nr_entries); -- cgit v1.2.3 From 81e36bffad95e015af9741b5b1ee16afe08aab05 Mon Sep 17 00:00:00 2001 From: Arnaldo Carvalho de Melo Date: Fri, 11 Nov 2011 22:28:50 -0200 Subject: perf evlist: Introduce id_hdr_size method out of perf_session We will need this when not using perf_session in cases like 'perf top' and strace where no perf.data file is created nor consumed. Cc: David Ahern Cc: Frederic Weisbecker Cc: Mike Galbraith Cc: Paul Mackerras Cc: Peter Zijlstra Cc: Stephane Eranian Link: http://lkml.kernel.org/n/tip-za923wjc41q5xot5vrhuhj3j@git.kernel.org Signed-off-by: Arnaldo Carvalho de Melo --- tools/perf/util/evlist.c | 32 ++++++++++++++++++++++++++++++++ tools/perf/util/evlist.h | 1 + tools/perf/util/session.c | 29 +---------------------------- 3 files changed, 34 insertions(+), 28 deletions(-) diff --git a/tools/perf/util/evlist.c b/tools/perf/util/evlist.c index 81237dcde1fd..d44e3df13a8f 100644 --- a/tools/perf/util/evlist.c +++ b/tools/perf/util/evlist.c @@ -625,6 +625,38 @@ u64 perf_evlist__sample_type(const struct perf_evlist *evlist) return first->attr.sample_type; } +u16 perf_evlist__id_hdr_size(const struct perf_evlist *evlist) +{ + struct perf_evsel *first; + struct perf_sample *data; + u64 sample_type; + u16 size = 0; + + first = list_entry(evlist->entries.next, struct perf_evsel, node); + + if (!first->attr.sample_id_all) + goto out; + + sample_type = first->attr.sample_type; + + if (sample_type & PERF_SAMPLE_TID) + size += sizeof(data->tid) * 2; + + if (sample_type & PERF_SAMPLE_TIME) + size += sizeof(data->time); + + if (sample_type & PERF_SAMPLE_ID) + size += sizeof(data->id); + + if (sample_type & PERF_SAMPLE_STREAM_ID) + size += sizeof(data->stream_id); + + if (sample_type & PERF_SAMPLE_CPU) + size += sizeof(data->cpu) * 2; +out: + return size; +} + bool perf_evlist__valid_sample_id_all(const struct perf_evlist *evlist) { struct perf_evsel *pos, *first; diff --git a/tools/perf/util/evlist.h b/tools/perf/util/evlist.h index 17e9c80243ce..2202e7b04103 100644 --- a/tools/perf/util/evlist.h +++ b/tools/perf/util/evlist.h @@ -104,6 +104,7 @@ int perf_evlist__set_filters(struct perf_evlist *evlist); u64 perf_evlist__sample_type(const struct perf_evlist *evlist); bool perf_evlist__sample_id_all(const const struct perf_evlist *evlist); +u16 perf_evlist__id_hdr_size(const struct perf_evlist *evlist); bool perf_evlist__valid_sample_type(const struct perf_evlist *evlist); bool perf_evlist__valid_sample_id_all(const struct perf_evlist *evlist); diff --git a/tools/perf/util/session.c b/tools/perf/util/session.c index a76666f17767..675e080f66b6 100644 --- a/tools/perf/util/session.c +++ b/tools/perf/util/session.c @@ -78,39 +78,12 @@ out_close: return -1; } -static void perf_session__id_header_size(struct perf_session *session) -{ - struct perf_sample *data; - u64 sample_type = session->sample_type; - u16 size = 0; - - if (!session->sample_id_all) - goto out; - - if (sample_type & PERF_SAMPLE_TID) - size += sizeof(data->tid) * 2; - - if (sample_type & PERF_SAMPLE_TIME) - size += sizeof(data->time); - - if (sample_type & PERF_SAMPLE_ID) - size += sizeof(data->id); - - if (sample_type & PERF_SAMPLE_STREAM_ID) - size += sizeof(data->stream_id); - - if (sample_type & PERF_SAMPLE_CPU) - size += sizeof(data->cpu) * 2; -out: - session->id_hdr_size = size; -} - void perf_session__update_sample_type(struct perf_session *self) { self->sample_type = perf_evlist__sample_type(self->evlist); self->sample_size = __perf_evsel__sample_size(self->sample_type); self->sample_id_all = perf_evlist__sample_id_all(self->evlist); - perf_session__id_header_size(self); + self->id_hdr_size = perf_evlist__id_hdr_size(self->evlist); } int perf_session__create_kernel_maps(struct perf_session *self) -- cgit v1.2.3 From 10d0f086df77f3ff259b46cb501362dbaf2c7989 Mon Sep 17 00:00:00 2001 From: Arnaldo Carvalho de Melo Date: Fri, 11 Nov 2011 22:45:41 -0200 Subject: perf event: perf_event_ops->attr() manipulates only an evlist Removing another case where a perf_session is required when processing events. Cc: David Ahern Cc: Frederic Weisbecker Cc: Mike Galbraith Cc: Paul Mackerras Cc: Peter Zijlstra Cc: Stephane Eranian Link: http://lkml.kernel.org/n/tip-ug1wtjbnva4bxwknflkkrlrh@git.kernel.org Signed-off-by: Arnaldo Carvalho de Melo --- tools/perf/builtin-inject.c | 8 +++++++- tools/perf/util/header.c | 19 ++++++++----------- tools/perf/util/header.h | 2 +- tools/perf/util/session.c | 16 ++++++++++++++-- tools/perf/util/session.h | 6 ++++-- 5 files changed, 34 insertions(+), 17 deletions(-) diff --git a/tools/perf/builtin-inject.c b/tools/perf/builtin-inject.c index 8dfc12bb119b..978751ec64ce 100644 --- a/tools/perf/builtin-inject.c +++ b/tools/perf/builtin-inject.c @@ -36,6 +36,12 @@ static int perf_event__repipe_synth(union perf_event *event, return 0; } +static int perf_event__repipe_attr(union perf_event *event, + struct perf_evlist **pevlist __used) +{ + return perf_event__repipe_synth(event, NULL); +} + static int perf_event__repipe(union perf_event *event, struct perf_sample *sample __used, struct perf_session *session) @@ -182,7 +188,7 @@ struct perf_event_ops inject_ops = { .read = perf_event__repipe, .throttle = perf_event__repipe, .unthrottle = perf_event__repipe, - .attr = perf_event__repipe_synth, + .attr = perf_event__repipe_attr, .event_type = perf_event__repipe_synth, .tracing_data = perf_event__repipe_synth, .build_id = perf_event__repipe_synth, diff --git a/tools/perf/util/header.c b/tools/perf/util/header.c index 41424a16be8e..1fa97dd21200 100644 --- a/tools/perf/util/header.c +++ b/tools/perf/util/header.c @@ -2120,23 +2120,23 @@ int perf_session__synthesize_attrs(struct perf_session *session, } int perf_event__process_attr(union perf_event *event, - struct perf_session *session) + struct perf_evlist **pevlist) { unsigned int i, ids, n_ids; struct perf_evsel *evsel; + struct perf_evlist *evlist = *pevlist; - if (session->evlist == NULL) { - session->evlist = perf_evlist__new(NULL, NULL); - if (session->evlist == NULL) + if (evlist == NULL) { + *pevlist = evlist = perf_evlist__new(NULL, NULL); + if (evlist == NULL) return -ENOMEM; } - evsel = perf_evsel__new(&event->attr.attr, - session->evlist->nr_entries); + evsel = perf_evsel__new(&event->attr.attr, evlist->nr_entries); if (evsel == NULL) return -ENOMEM; - perf_evlist__add(session->evlist, evsel); + perf_evlist__add(evlist, evsel); ids = event->header.size; ids -= (void *)&event->attr.id - (void *)event; @@ -2150,12 +2150,9 @@ int perf_event__process_attr(union perf_event *event, return -ENOMEM; for (i = 0; i < n_ids; i++) { - perf_evlist__id_add(session->evlist, evsel, 0, i, - event->attr.id[i]); + perf_evlist__id_add(evlist, evsel, 0, i, event->attr.id[i]); } - perf_session__update_sample_type(session); - return 0; } diff --git a/tools/perf/util/header.h b/tools/perf/util/header.h index 3d5a742f4a2a..0a88982bc392 100644 --- a/tools/perf/util/header.h +++ b/tools/perf/util/header.h @@ -101,7 +101,7 @@ int perf_event__synthesize_attr(struct perf_event_attr *attr, u16 ids, u64 *id, struct perf_session *session); int perf_session__synthesize_attrs(struct perf_session *session, perf_event__handler_t process); -int perf_event__process_attr(union perf_event *event, struct perf_session *session); +int perf_event__process_attr(union perf_event *event, struct perf_evlist **pevlist); int perf_event__synthesize_event_type(u64 event_id, char *name, perf_event__handler_t process, diff --git a/tools/perf/util/session.c b/tools/perf/util/session.c index 675e080f66b6..6e7d5f54b37d 100644 --- a/tools/perf/util/session.c +++ b/tools/perf/util/session.c @@ -277,6 +277,13 @@ static int process_event_synth_stub(union perf_event *event __used, return 0; } +static int process_event_synth_attr_stub(union perf_event *event __used, + struct perf_evlist **pevlist __used) +{ + dump_printf(": unhandled!\n"); + return 0; +} + static int process_event_sample_stub(union perf_event *event __used, struct perf_sample *sample __used, struct perf_evsel *evsel __used, @@ -327,7 +334,7 @@ static void perf_event_ops__fill_defaults(struct perf_event_ops *handler) if (handler->unthrottle == NULL) handler->unthrottle = process_event_stub; if (handler->attr == NULL) - handler->attr = process_event_synth_stub; + handler->attr = process_event_synth_attr_stub; if (handler->event_type == NULL) handler->event_type = process_event_synth_stub; if (handler->tracing_data == NULL) @@ -794,12 +801,17 @@ static int perf_session__preprocess_sample(struct perf_session *session, static int perf_session__process_user_event(struct perf_session *session, union perf_event *event, struct perf_event_ops *ops, u64 file_offset) { + int err; + dump_event(session, event, file_offset, NULL); /* These events are processed right away */ switch (event->header.type) { case PERF_RECORD_HEADER_ATTR: - return ops->attr(event, session); + err = ops->attr(event, &session->evlist); + if (err == 0) + perf_session__update_sample_type(session); + return err; case PERF_RECORD_HEADER_EVENT_TYPE: return ops->event_type(event, session); case PERF_RECORD_HEADER_TRACING_DATA: diff --git a/tools/perf/util/session.h b/tools/perf/util/session.h index 76d462d3bef7..13bd5e0a0691 100644 --- a/tools/perf/util/session.h +++ b/tools/perf/util/session.h @@ -63,6 +63,8 @@ typedef int (*event_op)(union perf_event *self, struct perf_sample *sample, struct perf_session *session); typedef int (*event_synth_op)(union perf_event *self, struct perf_session *session); +typedef int (*event_attr_op)(union perf_event *event, + struct perf_evlist **pevlist); typedef int (*event_op2)(union perf_event *self, struct perf_session *session, struct perf_event_ops *ops); @@ -76,8 +78,8 @@ struct perf_event_ops { read, throttle, unthrottle; - event_synth_op attr, - event_type, + event_attr_op attr; + event_synth_op event_type, tracing_data, build_id; event_op2 finished_round; -- cgit v1.2.3 From 246d4ce8107ea16521384c8b2a8fcff354ef2b7c Mon Sep 17 00:00:00 2001 From: Arnaldo Carvalho de Melo Date: Fri, 11 Nov 2011 23:10:26 -0200 Subject: perf session: Remove superfluous callchain_cursor member Since we have it in evsel->hists.callchain_cursor, remove it from perf_session. One more step in disentangling several places from requiring a perf_session pointer. Cc: David Ahern Cc: Frederic Weisbecker Cc: Mike Galbraith Cc: Paul Mackerras Cc: Peter Zijlstra Cc: Stephane Eranian Link: http://lkml.kernel.org/n/tip-rxr5dj3di7ckyfmnz0naku1z@git.kernel.org Signed-off-by: Arnaldo Carvalho de Melo --- tools/perf/builtin-report.c | 5 +++-- tools/perf/builtin-script.c | 2 +- tools/perf/builtin-top.c | 4 ++-- tools/perf/util/session.c | 12 ++++++------ tools/perf/util/session.h | 5 ++--- 5 files changed, 14 insertions(+), 14 deletions(-) diff --git a/tools/perf/builtin-report.c b/tools/perf/builtin-report.c index 758a287fc07a..b7ab373b9acc 100644 --- a/tools/perf/builtin-report.c +++ b/tools/perf/builtin-report.c @@ -65,7 +65,7 @@ static int perf_session__add_hist_entry(struct perf_session *session, struct hist_entry *he; if ((sort__has_parent || symbol_conf.use_callchain) && sample->callchain) { - err = perf_session__resolve_callchain(session, al->thread, + err = perf_session__resolve_callchain(session, evsel, al->thread, sample->callchain, &parent); if (err) return err; @@ -76,7 +76,8 @@ static int perf_session__add_hist_entry(struct perf_session *session, return -ENOMEM; if (symbol_conf.use_callchain) { - err = callchain_append(he->callchain, &session->callchain_cursor, + err = callchain_append(he->callchain, + &evsel->hists.callchain_cursor, sample->period); if (err) return err; diff --git a/tools/perf/builtin-script.c b/tools/perf/builtin-script.c index 2f62a2952269..47545e9c9b27 100644 --- a/tools/perf/builtin-script.c +++ b/tools/perf/builtin-script.c @@ -384,7 +384,7 @@ static void process_event(union perf_event *event __unused, printf(" "); else printf("\n"); - perf_session__print_ip(event, sample, session, + perf_session__print_ip(event, evsel, sample, session, PRINT_FIELD(SYM), PRINT_FIELD(DSO)); } diff --git a/tools/perf/builtin-top.c b/tools/perf/builtin-top.c index 04288ee223ed..9b3bbb40d46f 100644 --- a/tools/perf/builtin-top.c +++ b/tools/perf/builtin-top.c @@ -777,7 +777,7 @@ static void perf_event__process_sample(const union perf_event *event, if ((sort__has_parent || symbol_conf.use_callchain) && sample->callchain) { - err = perf_session__resolve_callchain(session, al.thread, + err = perf_session__resolve_callchain(session, evsel, al.thread, sample->callchain, &parent); if (err) return; @@ -790,7 +790,7 @@ static void perf_event__process_sample(const union perf_event *event, } if (symbol_conf.use_callchain) { - err = callchain_append(he->callchain, &session->callchain_cursor, + err = callchain_append(he->callchain, &evsel->hists.callchain_cursor, sample->period); if (err) return; diff --git a/tools/perf/util/session.c b/tools/perf/util/session.c index 6e7d5f54b37d..734358b51ed1 100644 --- a/tools/perf/util/session.c +++ b/tools/perf/util/session.c @@ -216,7 +216,7 @@ static bool symbol__match_parent_regex(struct symbol *sym) return 0; } -int perf_session__resolve_callchain(struct perf_session *self, +int perf_session__resolve_callchain(struct perf_session *self, struct perf_evsel *evsel, struct thread *thread, struct ip_callchain *chain, struct symbol **parent) @@ -225,7 +225,7 @@ int perf_session__resolve_callchain(struct perf_session *self, unsigned int i; int err; - callchain_cursor_reset(&self->callchain_cursor); + callchain_cursor_reset(&evsel->hists.callchain_cursor); for (i = 0; i < chain->nr; i++) { u64 ip; @@ -261,7 +261,7 @@ int perf_session__resolve_callchain(struct perf_session *self, break; } - err = callchain_cursor_append(&self->callchain_cursor, + err = callchain_cursor_append(&evsel->hists.callchain_cursor, ip, al.map, al.sym); if (err) return err; @@ -1254,14 +1254,14 @@ struct perf_evsel *perf_session__find_first_evtype(struct perf_session *session, return NULL; } -void perf_session__print_ip(union perf_event *event, +void perf_session__print_ip(union perf_event *event, struct perf_evsel *evsel, struct perf_sample *sample, struct perf_session *session, int print_sym, int print_dso) { struct addr_location al; const char *symname, *dsoname; - struct callchain_cursor *cursor = &session->callchain_cursor; + struct callchain_cursor *cursor = &evsel->hists.callchain_cursor; struct callchain_cursor_node *node; if (perf_event__preprocess_sample(event, session, &al, sample, @@ -1273,7 +1273,7 @@ void perf_session__print_ip(union perf_event *event, if (symbol_conf.use_callchain && sample->callchain) { - if (perf_session__resolve_callchain(session, al.thread, + if (perf_session__resolve_callchain(session, evsel, al.thread, sample->callchain, NULL) != 0) { if (verbose) error("Failed to resolve callchain. Skipping\n"); diff --git a/tools/perf/util/session.h b/tools/perf/util/session.h index 13bd5e0a0691..d2f430367713 100644 --- a/tools/perf/util/session.h +++ b/tools/perf/util/session.h @@ -50,7 +50,6 @@ struct perf_session { int cwdlen; char *cwd; struct ordered_samples ordered_samples; - struct callchain_cursor callchain_cursor; char filename[0]; }; @@ -100,7 +99,7 @@ int __perf_session__process_events(struct perf_session *self, int perf_session__process_events(struct perf_session *self, struct perf_event_ops *event_ops); -int perf_session__resolve_callchain(struct perf_session *self, +int perf_session__resolve_callchain(struct perf_session *self, struct perf_evsel *evsel, struct thread *thread, struct ip_callchain *chain, struct symbol **parent); @@ -169,7 +168,7 @@ static inline int perf_session__parse_sample(struct perf_session *session, struct perf_evsel *perf_session__find_first_evtype(struct perf_session *session, unsigned int type); -void perf_session__print_ip(union perf_event *event, +void perf_session__print_ip(union perf_event *event, struct perf_evsel *evsel, struct perf_sample *sample, struct perf_session *session, int print_sym, int print_dso); -- cgit v1.2.3 From e3f42609628a20da92ecbc2d81053cc82c90a071 Mon Sep 17 00:00:00 2001 From: Arnaldo Carvalho de Melo Date: Wed, 16 Nov 2011 17:02:54 -0200 Subject: perf tools: Use evsel->attr.sample_type instead of session->sample_type Eventually session->sample_type will go away as we want to support multiple sample types per session, so use it from the evsel which is a step in that direction. Cc: David Ahern Cc: Frederic Weisbecker Cc: Mike Galbraith Cc: Paul Mackerras Cc: Peter Zijlstra Cc: Stephane Eranian Link: http://lkml.kernel.org/n/tip-0vwdpjcwbjezw459lw5n3ew1@git.kernel.org Signed-off-by: Arnaldo Carvalho de Melo --- tools/perf/builtin-sched.c | 5 +++-- tools/perf/builtin-timechart.c | 9 +++++---- 2 files changed, 8 insertions(+), 6 deletions(-) diff --git a/tools/perf/builtin-sched.c b/tools/perf/builtin-sched.c index 5177964943e7..d51af0beab13 100644 --- a/tools/perf/builtin-sched.c +++ b/tools/perf/builtin-sched.c @@ -3,6 +3,7 @@ #include "util/util.h" #include "util/cache.h" +#include "util/evsel.h" #include "util/symbol.h" #include "util/thread.h" #include "util/header.h" @@ -1603,12 +1604,12 @@ static void process_raw_event(union perf_event *raw_event __used, static int process_sample_event(union perf_event *event, struct perf_sample *sample, - struct perf_evsel *evsel __used, + struct perf_evsel *evsel, struct perf_session *session) { struct thread *thread; - if (!(session->sample_type & PERF_SAMPLE_RAW)) + if (!(evsel->attr.sample_type & PERF_SAMPLE_RAW)) return 0; thread = perf_session__findnew(session, sample->pid); diff --git a/tools/perf/builtin-timechart.c b/tools/perf/builtin-timechart.c index aa26f4d66d10..3fc52b1aa430 100644 --- a/tools/perf/builtin-timechart.c +++ b/tools/perf/builtin-timechart.c @@ -19,6 +19,7 @@ #include "util/color.h" #include #include "util/cache.h" +#include "util/evsel.h" #include #include "util/symbol.h" #include "util/callchain.h" @@ -488,12 +489,12 @@ static void sched_switch(int cpu, u64 timestamp, struct trace_entry *te) static int process_sample_event(union perf_event *event __used, struct perf_sample *sample, - struct perf_evsel *evsel __used, - struct perf_session *session) + struct perf_evsel *evsel, + struct perf_session *session __used) { struct trace_entry *te; - if (session->sample_type & PERF_SAMPLE_TIME) { + if (evsel->attr.sample_type & PERF_SAMPLE_TIME) { if (!first_time || first_time > sample->time) first_time = sample->time; if (last_time < sample->time) @@ -501,7 +502,7 @@ static int process_sample_event(union perf_event *event __used, } te = (void *)sample->raw_data; - if (session->sample_type & PERF_SAMPLE_RAW && sample->raw_size > 0) { + if ((evsel->attr.sample_type & PERF_SAMPLE_RAW) && sample->raw_size > 0) { char *event_str; #ifdef SUPPORT_OLD_POWER_EVENTS struct power_entry_old *peo; -- cgit v1.2.3 From fa372aae335c6dfbe808d5a728fe10cd202dde45 Mon Sep 17 00:00:00 2001 From: Arnaldo Carvalho de Melo Date: Thu, 17 Nov 2011 12:19:04 -0200 Subject: perf report: Group options in a struct Paving the way to remove these globals when we change the perf_event_ops to receive as a first parameter a pointer to a perf_event_ops that will then provide access to perf_report via container_of. Cc: David Ahern Cc: Frederic Weisbecker Cc: Mike Galbraith Cc: Paul Mackerras Cc: Peter Zijlstra Cc: Stephane Eranian Link: http://lkml.kernel.org/n/tip-2eh2vi2nb5z3tg1lvoxv09xu@git.kernel.org Signed-off-by: Arnaldo Carvalho de Melo --- tools/perf/builtin-report.c | 111 +++++++++++++++++++++++--------------------- 1 file changed, 59 insertions(+), 52 deletions(-) diff --git a/tools/perf/builtin-report.c b/tools/perf/builtin-report.c index b7ab373b9acc..5d2e819dfc40 100644 --- a/tools/perf/builtin-report.c +++ b/tools/perf/builtin-report.c @@ -35,25 +35,25 @@ #include -static char const *input_name = "perf.data"; - -static bool force, use_tui, use_stdio; -static bool hide_unresolved; -static bool dont_use_callchains; -static bool show_full_info; - -static bool show_threads; -static struct perf_read_values show_threads_values; - -static const char default_pretty_printing_style[] = "normal"; -static const char *pretty_printing_style = default_pretty_printing_style; - -static char callchain_default_opt[] = "fractal,0.5,callee"; -static bool inverted_callchain; -static symbol_filter_t annotate_init; - -static const char *cpu_list; -static DECLARE_BITMAP(cpu_bitmap, MAX_NR_CPUS); +static struct perf_report { + char const *input_name; + bool force, use_tui, use_stdio; + bool hide_unresolved; + bool dont_use_callchains; + bool show_full_info; + bool show_threads; + bool inverted_callchain; + struct perf_read_values show_threads_values; + const char *pretty_printing_style; + symbol_filter_t annotate_init; + const char *cpu_list; + DECLARE_BITMAP(cpu_bitmap, MAX_NR_CPUS); +} report = { + .input_name = "perf.data", + .pretty_printing_style = "normal", +}, *rep = &report; + +static char callchain_default_opt[] = "fractal,0.5,callee"; static int perf_session__add_hist_entry(struct perf_session *session, struct addr_location *al, @@ -114,16 +114,16 @@ static int process_sample_event(union perf_event *event, struct addr_location al; if (perf_event__preprocess_sample(event, session, &al, sample, - annotate_init) < 0) { + rep->annotate_init) < 0) { fprintf(stderr, "problem processing %d event, skipping it.\n", event->header.type); return -1; } - if (al.filtered || (hide_unresolved && al.sym == NULL)) + if (al.filtered || (rep->hide_unresolved && al.sym == NULL)) return 0; - if (cpu_list && !test_bit(sample->cpu, cpu_bitmap)) + if (rep->cpu_list && !test_bit(sample->cpu, rep->cpu_bitmap)) return 0; if (al.map != NULL) @@ -143,9 +143,9 @@ static int process_read_event(union perf_event *event, { struct perf_evsel *evsel = perf_evlist__id2evsel(session->evlist, event->read.id); - if (show_threads) { + if (rep->show_threads) { const char *name = evsel ? event_name(evsel) : "unknown"; - perf_read_values_add_value(&show_threads_values, + perf_read_values_add_value(&rep->show_threads_values, event->read.pid, event->read.tid, event->read.id, name, @@ -173,7 +173,8 @@ static int perf_session__setup_sample_type(struct perf_session *self) "you call 'perf record' without -g?\n"); return -1; } - } else if (!dont_use_callchains && callchain_param.mode != CHAIN_NONE && + } else if (!rep->dont_use_callchains && + callchain_param.mode != CHAIN_NONE && !symbol_conf.use_callchain) { symbol_conf.use_callchain = true; if (callchain_register_param(&callchain_param) < 0) { @@ -241,11 +242,11 @@ static int perf_evlist__tty_browse_hists(struct perf_evlist *evlist, parent_pattern == default_parent_pattern) { fprintf(stdout, "#\n# (%s)\n#\n", help); - if (show_threads) { - bool style = !strcmp(pretty_printing_style, "raw"); - perf_read_values_display(stdout, &show_threads_values, + if (rep->show_threads) { + bool style = !strcmp(rep->pretty_printing_style, "raw"); + perf_read_values_display(stdout, &rep->show_threads_values, style); - perf_read_values_destroy(&show_threads_values); + perf_read_values_destroy(&rep->show_threads_values); } } @@ -264,21 +265,23 @@ static int __cmd_report(void) signal(SIGINT, sig_handler); - session = perf_session__new(input_name, O_RDONLY, force, false, &event_ops); + session = perf_session__new(rep->input_name, O_RDONLY, + rep->force, false, &event_ops); if (session == NULL) return -ENOMEM; - if (cpu_list) { - ret = perf_session__cpu_bitmap(session, cpu_list, cpu_bitmap); + if (rep->cpu_list) { + ret = perf_session__cpu_bitmap(session, rep->cpu_list, + rep->cpu_bitmap); if (ret) goto out_delete; } if (use_browser <= 0) - perf_session__fprintf_info(session, stdout, show_full_info); + perf_session__fprintf_info(session, stdout, rep->show_full_info); - if (show_threads) - perf_read_values_init(&show_threads_values); + if (rep->show_threads) + perf_read_values_init(&rep->show_threads_values); ret = perf_session__setup_sample_type(session); if (ret) @@ -327,7 +330,8 @@ static int __cmd_report(void) } if (nr_samples == 0) { - ui__warning("The %s file has no samples!\n", input_name); + ui__warning("The %s file has no samples!\n", + rep->input_name); goto out_delete; } @@ -364,7 +368,7 @@ parse_callchain_opt(const struct option *opt __used, const char *arg, * --no-call-graph */ if (unset) { - dont_use_callchains = true; + rep->dont_use_callchains = true; return 0; } @@ -439,7 +443,7 @@ static const char * const report_usage[] = { }; static const struct option options[] = { - OPT_STRING('i', "input", &input_name, "file", + OPT_STRING('i', "input", &report.input_name, "file", "input file name"), OPT_INCR('v', "verbose", &verbose, "be more verbose (show symbol address, etc)"), @@ -449,17 +453,18 @@ static const struct option options[] = { "file", "vmlinux pathname"), OPT_STRING(0, "kallsyms", &symbol_conf.kallsyms_name, "file", "kallsyms pathname"), - OPT_BOOLEAN('f', "force", &force, "don't complain, do it"), + OPT_BOOLEAN('f', "force", &report.force, "don't complain, do it"), OPT_BOOLEAN('m', "modules", &symbol_conf.use_modules, "load module symbols - WARNING: use only with -k and LIVE kernel"), OPT_BOOLEAN('n', "show-nr-samples", &symbol_conf.show_nr_samples, "Show a column with the number of samples"), - OPT_BOOLEAN('T', "threads", &show_threads, + OPT_BOOLEAN('T', "threads", &report.show_threads, "Show per-thread event counters"), - OPT_STRING(0, "pretty", &pretty_printing_style, "key", + OPT_STRING(0, "pretty", &report.pretty_printing_style, "key", "pretty printing style key: normal raw"), - OPT_BOOLEAN(0, "tui", &use_tui, "Use the TUI interface"), - OPT_BOOLEAN(0, "stdio", &use_stdio, "Use the stdio interface"), + OPT_BOOLEAN(0, "tui", &report.use_tui, "Use the TUI interface"), + OPT_BOOLEAN(0, "stdio", &report.use_stdio, + "Use the stdio interface"), OPT_STRING('s', "sort", &sort_order, "key[,key2...]", "sort by key(s): pid, comm, dso, symbol, parent"), OPT_BOOLEAN(0, "showcpuutilization", &symbol_conf.show_cpu_utilization, @@ -471,7 +476,8 @@ static const struct option options[] = { OPT_CALLBACK_DEFAULT('g', "call-graph", NULL, "output_type,min_percent, call_order", "Display callchains using output_type (graph, flat, fractal, or none) , min percent threshold and callchain order. " "Default: fractal,0.5,callee", &parse_callchain_opt, callchain_default_opt), - OPT_BOOLEAN('G', "inverted", &inverted_callchain, "alias for inverted call graph"), + OPT_BOOLEAN('G', "inverted", &report.inverted_callchain, + "alias for inverted call graph"), OPT_STRING('d', "dsos", &symbol_conf.dso_list_str, "dso[,dso...]", "only consider symbols in these dsos"), OPT_STRING('C', "comms", &symbol_conf.comm_list_str, "comm[,comm...]", @@ -484,12 +490,13 @@ static const struct option options[] = { OPT_STRING('t', "field-separator", &symbol_conf.field_sep, "separator", "separator for columns, no spaces will be added between " "columns '.' is reserved."), - OPT_BOOLEAN('U', "hide-unresolved", &hide_unresolved, + OPT_BOOLEAN('U', "hide-unresolved", &report.hide_unresolved, "Only display entries resolved to a symbol"), OPT_STRING(0, "symfs", &symbol_conf.symfs, "directory", "Look for files with symbols relative to this directory"), - OPT_STRING('c', "cpu", &cpu_list, "cpu", "list of cpus to profile"), - OPT_BOOLEAN('I', "show-info", &show_full_info, + OPT_STRING('c', "cpu", &report.cpu_list, "cpu", + "list of cpus to profile"), + OPT_BOOLEAN('I', "show-info", &report.show_full_info, "Display extended information about perf.data file"), OPT_BOOLEAN(0, "source", &symbol_conf.annotate_src, "Interleave source code with assembly code (default)"), @@ -506,15 +513,15 @@ int cmd_report(int argc, const char **argv, const char *prefix __used) { argc = parse_options(argc, argv, options, report_usage, 0); - if (use_stdio) + if (report.use_stdio) use_browser = 0; - else if (use_tui) + else if (report.use_tui) use_browser = 1; - if (inverted_callchain) + if (report.inverted_callchain) callchain_param.order = ORDER_CALLER; - if (strcmp(input_name, "-") != 0) + if (strcmp(report.input_name, "-") != 0) setup_browser(true); else use_browser = 0; @@ -525,7 +532,7 @@ int cmd_report(int argc, const char **argv, const char *prefix __used) */ if (use_browser > 0) { symbol_conf.priv_size = sizeof(struct annotation); - annotate_init = symbol__annotate_init; + report.annotate_init = symbol__annotate_init; /* * For searching by name on the "Browse map details". * providing it only in verbose mode not to bloat too -- cgit v1.2.3 From 7009cc34b964939815160d7de64cf0215cdbf8bb Mon Sep 17 00:00:00 2001 From: Arnaldo Carvalho de Melo Date: Thu, 17 Nov 2011 12:33:21 -0200 Subject: perf annotate: Group options in a struct Paving the way to remove these globals when we change the perf_event_ops to receive as a first parameter a pointer to a perf_event_ops that will then provide access to perf_annotate via container_of. Cc: David Ahern Cc: Frederic Weisbecker Cc: Mike Galbraith Cc: Paul Mackerras Cc: Peter Zijlstra Cc: Stephane Eranian Link: http://lkml.kernel.org/n/tip-xduzibqrdg3h5cttmk6p5wwc@git.kernel.org Signed-off-by: Arnaldo Carvalho de Melo --- tools/perf/builtin-annotate.c | 64 ++++++++++++++++++++++--------------------- 1 file changed, 33 insertions(+), 31 deletions(-) diff --git a/tools/perf/builtin-annotate.c b/tools/perf/builtin-annotate.c index 8b9091bce98d..4f0c3d98352d 100644 --- a/tools/perf/builtin-annotate.c +++ b/tools/perf/builtin-annotate.c @@ -30,18 +30,17 @@ #include -static char const *input_name = "perf.data"; - -static bool force, use_tui, use_stdio; - -static bool full_paths; - -static bool print_line; - -static const char *sym_hist_filter; - -static const char *cpu_list; -static DECLARE_BITMAP(cpu_bitmap, MAX_NR_CPUS); +static struct perf_annotate { + char const *input_name; + bool force, use_tui, use_stdio; + bool full_paths; + bool print_line; + const char *sym_hist_filter; + const char *cpu_list; + DECLARE_BITMAP(cpu_bitmap, MAX_NR_CPUS); +} annotate = { + .input_name = "perf.data", +}, *ann = &annotate; static int perf_evsel__add_sample(struct perf_evsel *evsel, struct perf_sample *sample, @@ -50,8 +49,9 @@ static int perf_evsel__add_sample(struct perf_evsel *evsel, struct hist_entry *he; int ret; - if (sym_hist_filter != NULL && - (al->sym == NULL || strcmp(sym_hist_filter, al->sym->name) != 0)) { + if (ann->sym_hist_filter != NULL && + (al->sym == NULL || + strcmp(ann->sym_hist_filter, al->sym->name) != 0)) { /* We're only interested in a symbol named sym_hist_filter */ if (al->sym != NULL) { rb_erase(&al->sym->rb_node, @@ -93,7 +93,7 @@ static int process_sample_event(union perf_event *event, return -1; } - if (cpu_list && !test_bit(sample->cpu, cpu_bitmap)) + if (ann->cpu_list && !test_bit(sample->cpu, ann->cpu_bitmap)) return 0; if (!al.filtered && perf_evsel__add_sample(evsel, sample, &al)) { @@ -108,7 +108,7 @@ static int process_sample_event(union perf_event *event, static int hist_entry__tty_annotate(struct hist_entry *he, int evidx) { return symbol__tty_annotate(he->ms.sym, he->ms.map, evidx, - print_line, full_paths, 0, 0); + ann->print_line, ann->full_paths, 0, 0); } static void hists__find_annotations(struct hists *self, int evidx) @@ -178,12 +178,14 @@ static int __cmd_annotate(void) struct perf_evsel *pos; u64 total_nr_samples; - session = perf_session__new(input_name, O_RDONLY, force, false, &event_ops); + session = perf_session__new(ann->input_name, O_RDONLY, + ann->force, false, &event_ops); if (session == NULL) return -ENOMEM; - if (cpu_list) { - ret = perf_session__cpu_bitmap(session, cpu_list, cpu_bitmap); + if (ann->cpu_list) { + ret = perf_session__cpu_bitmap(session, ann->cpu_list, + ann->cpu_bitmap); if (ret) goto out_delete; } @@ -217,7 +219,7 @@ static int __cmd_annotate(void) } if (total_nr_samples == 0) { - ui__warning("The %s file has no samples!\n", input_name); + ui__warning("The %s file has no samples!\n", ann->input_name); goto out_delete; } out_delete: @@ -242,28 +244,28 @@ static const char * const annotate_usage[] = { }; static const struct option options[] = { - OPT_STRING('i', "input", &input_name, "file", + OPT_STRING('i', "input", &annotate.input_name, "file", "input file name"), OPT_STRING('d', "dsos", &symbol_conf.dso_list_str, "dso[,dso...]", "only consider symbols in these dsos"), - OPT_STRING('s', "symbol", &sym_hist_filter, "symbol", + OPT_STRING('s', "symbol", &annotate.sym_hist_filter, "symbol", "symbol to annotate"), - OPT_BOOLEAN('f', "force", &force, "don't complain, do it"), + OPT_BOOLEAN('f', "force", &annotate.force, "don't complain, do it"), OPT_INCR('v', "verbose", &verbose, "be more verbose (show symbol address, etc)"), OPT_BOOLEAN('D', "dump-raw-trace", &dump_trace, "dump raw trace in ASCII"), - OPT_BOOLEAN(0, "tui", &use_tui, "Use the TUI interface"), - OPT_BOOLEAN(0, "stdio", &use_stdio, "Use the stdio interface"), + OPT_BOOLEAN(0, "tui", &annotate.use_tui, "Use the TUI interface"), + OPT_BOOLEAN(0, "stdio", &annotate.use_stdio, "Use the stdio interface"), OPT_STRING('k', "vmlinux", &symbol_conf.vmlinux_name, "file", "vmlinux pathname"), OPT_BOOLEAN('m', "modules", &symbol_conf.use_modules, "load module symbols - WARNING: use only with -k and LIVE kernel"), - OPT_BOOLEAN('l', "print-line", &print_line, + OPT_BOOLEAN('l', "print-line", &annotate.print_line, "print matching source lines (may be slow)"), - OPT_BOOLEAN('P', "full-paths", &full_paths, + OPT_BOOLEAN('P', "full-paths", &annotate.full_paths, "Don't shorten the displayed pathnames"), - OPT_STRING('c', "cpu", &cpu_list, "cpu", "list of cpus to profile"), + OPT_STRING('c', "cpu", &annotate.cpu_list, "cpu", "list of cpus to profile"), OPT_STRING(0, "symfs", &symbol_conf.symfs, "directory", "Look for files with symbols relative to this directory"), OPT_BOOLEAN(0, "source", &symbol_conf.annotate_src, @@ -279,9 +281,9 @@ int cmd_annotate(int argc, const char **argv, const char *prefix __used) { argc = parse_options(argc, argv, options, annotate_usage, 0); - if (use_stdio) + if (annotate.use_stdio) use_browser = 0; - else if (use_tui) + else if (annotate.use_tui) use_browser = 1; setup_browser(true); @@ -302,7 +304,7 @@ int cmd_annotate(int argc, const char **argv, const char *prefix __used) if (argc > 1) usage_with_options(annotate_usage, options); - sym_hist_filter = argv[0]; + annotate.sym_hist_filter = argv[0]; } if (field_sep && *field_sep == '.') { -- cgit v1.2.3 From d20deb64e0490ee9442b5181bc08a62d2cadcb90 Mon Sep 17 00:00:00 2001 From: Arnaldo Carvalho de Melo Date: Fri, 25 Nov 2011 08:19:45 -0200 Subject: perf tools: Pass tool context in the the perf_event_ops functions So that we don't need to have that many globals. Next steps will remove the 'session' pointer, that in most cases is not needed. Then we can rename perf_event_ops to 'perf_tool' that better describes this class hierarchy. Cc: David Ahern Cc: Frederic Weisbecker Cc: Mike Galbraith Cc: Paul Mackerras Cc: Peter Zijlstra Cc: Stephane Eranian Link: http://lkml.kernel.org/n/tip-wp4djox7x6w1i2bab1pt4xxp@git.kernel.org Signed-off-by: Arnaldo Carvalho de Melo --- tools/perf/builtin-annotate.c | 62 ++++--- tools/perf/builtin-diff.c | 3 +- tools/perf/builtin-inject.c | 55 +++--- tools/perf/builtin-kmem.c | 3 +- tools/perf/builtin-lock.c | 3 +- tools/perf/builtin-record.c | 383 +++++++++++++++++++++++------------------ tools/perf/builtin-report.c | 97 ++++++----- tools/perf/builtin-sched.c | 3 +- tools/perf/builtin-script.c | 3 +- tools/perf/builtin-timechart.c | 12 +- tools/perf/builtin-top.c | 6 +- tools/perf/util/build-id.c | 7 +- tools/perf/util/callchain.h | 3 + tools/perf/util/event.c | 66 ++++--- tools/perf/util/event.h | 38 ++-- tools/perf/util/header.c | 36 ++-- tools/perf/util/header.h | 27 ++- tools/perf/util/session.c | 60 ++++--- tools/perf/util/session.h | 23 +-- tools/perf/util/top.h | 3 +- 20 files changed, 520 insertions(+), 373 deletions(-) diff --git a/tools/perf/builtin-annotate.c b/tools/perf/builtin-annotate.c index 4f0c3d98352d..483cb9466444 100644 --- a/tools/perf/builtin-annotate.c +++ b/tools/perf/builtin-annotate.c @@ -30,7 +30,8 @@ #include -static struct perf_annotate { +struct perf_annotate { + struct perf_event_ops ops; char const *input_name; bool force, use_tui, use_stdio; bool full_paths; @@ -38,13 +39,12 @@ static struct perf_annotate { const char *sym_hist_filter; const char *cpu_list; DECLARE_BITMAP(cpu_bitmap, MAX_NR_CPUS); -} annotate = { - .input_name = "perf.data", -}, *ann = &annotate; +}; static int perf_evsel__add_sample(struct perf_evsel *evsel, struct perf_sample *sample, - struct addr_location *al) + struct addr_location *al, + struct perf_annotate *ann) { struct hist_entry *he; int ret; @@ -79,11 +79,13 @@ static int perf_evsel__add_sample(struct perf_evsel *evsel, return ret; } -static int process_sample_event(union perf_event *event, +static int process_sample_event(struct perf_event_ops *ops, + union perf_event *event, struct perf_sample *sample, struct perf_evsel *evsel, struct perf_session *session) { + struct perf_annotate *ann = container_of(ops, struct perf_annotate, ops); struct addr_location al; if (perf_event__preprocess_sample(event, session, &al, sample, @@ -96,7 +98,7 @@ static int process_sample_event(union perf_event *event, if (ann->cpu_list && !test_bit(sample->cpu, ann->cpu_bitmap)) return 0; - if (!al.filtered && perf_evsel__add_sample(evsel, sample, &al)) { + if (!al.filtered && perf_evsel__add_sample(evsel, sample, &al, ann)) { pr_warning("problem incrementing symbol count, " "skipping event\n"); return -1; @@ -105,13 +107,15 @@ static int process_sample_event(union perf_event *event, return 0; } -static int hist_entry__tty_annotate(struct hist_entry *he, int evidx) +static int hist_entry__tty_annotate(struct hist_entry *he, int evidx, + struct perf_annotate *ann) { return symbol__tty_annotate(he->ms.sym, he->ms.map, evidx, ann->print_line, ann->full_paths, 0, 0); } -static void hists__find_annotations(struct hists *self, int evidx) +static void hists__find_annotations(struct hists *self, int evidx, + struct perf_annotate *ann) { struct rb_node *nd = rb_first(&self->entries), *next; int key = K_RIGHT; @@ -149,7 +153,7 @@ find_next: if (next != NULL) nd = next; } else { - hist_entry__tty_annotate(he, evidx); + hist_entry__tty_annotate(he, evidx, ann); nd = rb_next(nd); /* * Since we have a hist_entry per IP for the same @@ -162,16 +166,7 @@ find_next: } } -static struct perf_event_ops event_ops = { - .sample = process_sample_event, - .mmap = perf_event__process_mmap, - .comm = perf_event__process_comm, - .fork = perf_event__process_task, - .ordered_samples = true, - .ordering_requires_timestamps = true, -}; - -static int __cmd_annotate(void) +static int __cmd_annotate(struct perf_annotate *ann) { int ret; struct perf_session *session; @@ -179,7 +174,7 @@ static int __cmd_annotate(void) u64 total_nr_samples; session = perf_session__new(ann->input_name, O_RDONLY, - ann->force, false, &event_ops); + ann->force, false, &ann->ops); if (session == NULL) return -ENOMEM; @@ -190,7 +185,7 @@ static int __cmd_annotate(void) goto out_delete; } - ret = perf_session__process_events(session, &event_ops); + ret = perf_session__process_events(session, &ann->ops); if (ret) goto out_delete; @@ -214,7 +209,7 @@ static int __cmd_annotate(void) total_nr_samples += nr_samples; hists__collapse_resort(hists); hists__output_resort(hists); - hists__find_annotations(hists, pos->idx); + hists__find_annotations(hists, pos->idx, ann); } } @@ -243,7 +238,20 @@ static const char * const annotate_usage[] = { NULL }; -static const struct option options[] = { +int cmd_annotate(int argc, const char **argv, const char *prefix __used) +{ + struct perf_annotate annotate = { + .ops = { + .sample = process_sample_event, + .mmap = perf_event__process_mmap, + .comm = perf_event__process_comm, + .fork = perf_event__process_task, + .ordered_samples = true, + .ordering_requires_timestamps = true, + }, + .input_name = "perf.data", + }; + const struct option options[] = { OPT_STRING('i', "input", &annotate.input_name, "file", "input file name"), OPT_STRING('d', "dsos", &symbol_conf.dso_list_str, "dso[,dso...]", @@ -275,10 +283,8 @@ static const struct option options[] = { OPT_STRING('M', "disassembler-style", &disassembler_style, "disassembler style", "Specify disassembler style (e.g. -M intel for intel syntax)"), OPT_END() -}; + }; -int cmd_annotate(int argc, const char **argv, const char *prefix __used) -{ argc = parse_options(argc, argv, options, annotate_usage, 0); if (annotate.use_stdio) @@ -312,5 +318,5 @@ int cmd_annotate(int argc, const char **argv, const char *prefix __used) return -1; } - return __cmd_annotate(); + return __cmd_annotate(&annotate); } diff --git a/tools/perf/builtin-diff.c b/tools/perf/builtin-diff.c index b39f3a1ee7dc..9a0872f9e837 100644 --- a/tools/perf/builtin-diff.c +++ b/tools/perf/builtin-diff.c @@ -30,7 +30,8 @@ static int hists__add_entry(struct hists *self, return -ENOMEM; } -static int diff__process_sample_event(union perf_event *event, +static int diff__process_sample_event(struct perf_event_ops *ops __used, + union perf_event *event, struct perf_sample *sample, struct perf_evsel *evsel __used, struct perf_session *session) diff --git a/tools/perf/builtin-inject.c b/tools/perf/builtin-inject.c index 978751ec64ce..6ce6d80b59db 100644 --- a/tools/perf/builtin-inject.c +++ b/tools/perf/builtin-inject.c @@ -16,7 +16,8 @@ static char const *input_name = "-"; static bool inject_build_ids; -static int perf_event__repipe_synth(union perf_event *event, +static int perf_event__repipe_synth(struct perf_event_ops *ops __used, + union perf_event *event, struct perf_session *session __used) { uint32_t size; @@ -36,47 +37,57 @@ static int perf_event__repipe_synth(union perf_event *event, return 0; } +static int perf_event__repipe_tracing_data_synth(union perf_event *event, + struct perf_session *session) +{ + return perf_event__repipe_synth(NULL, event, session); +} + static int perf_event__repipe_attr(union perf_event *event, struct perf_evlist **pevlist __used) { - return perf_event__repipe_synth(event, NULL); + return perf_event__repipe_synth(NULL, event, NULL); } -static int perf_event__repipe(union perf_event *event, +static int perf_event__repipe(struct perf_event_ops *ops, + union perf_event *event, struct perf_sample *sample __used, struct perf_session *session) { - return perf_event__repipe_synth(event, session); + return perf_event__repipe_synth(ops, event, session); } -static int perf_event__repipe_sample(union perf_event *event, +static int perf_event__repipe_sample(struct perf_event_ops *ops, + union perf_event *event, struct perf_sample *sample __used, struct perf_evsel *evsel __used, struct perf_session *session) { - return perf_event__repipe_synth(event, session); + return perf_event__repipe_synth(ops, event, session); } -static int perf_event__repipe_mmap(union perf_event *event, +static int perf_event__repipe_mmap(struct perf_event_ops *ops, + union perf_event *event, struct perf_sample *sample, struct perf_session *session) { int err; - err = perf_event__process_mmap(event, sample, session); - perf_event__repipe(event, sample, session); + err = perf_event__process_mmap(ops, event, sample, session); + perf_event__repipe(ops, event, sample, session); return err; } -static int perf_event__repipe_task(union perf_event *event, +static int perf_event__repipe_task(struct perf_event_ops *ops, + union perf_event *event, struct perf_sample *sample, struct perf_session *session) { int err; - err = perf_event__process_task(event, sample, session); - perf_event__repipe(event, sample, session); + err = perf_event__process_task(ops, event, sample, session); + perf_event__repipe(ops, event, sample, session); return err; } @@ -86,7 +97,7 @@ static int perf_event__repipe_tracing_data(union perf_event *event, { int err; - perf_event__repipe_synth(event, session); + perf_event__repipe_synth(NULL, event, session); err = perf_event__process_tracing_data(event, session); return err; @@ -106,7 +117,8 @@ static int dso__read_build_id(struct dso *self) return -1; } -static int dso__inject_build_id(struct dso *self, struct perf_session *session) +static int dso__inject_build_id(struct dso *self, struct perf_event_ops *ops, + struct perf_session *session) { u16 misc = PERF_RECORD_MISC_USER; struct machine *machine; @@ -126,7 +138,7 @@ static int dso__inject_build_id(struct dso *self, struct perf_session *session) if (self->kernel) misc = PERF_RECORD_MISC_KERNEL; - err = perf_event__synthesize_build_id(self, misc, perf_event__repipe, + err = perf_event__synthesize_build_id(ops, self, misc, perf_event__repipe, machine, session); if (err) { pr_err("Can't synthesize build_id event for %s\n", self->long_name); @@ -136,7 +148,8 @@ static int dso__inject_build_id(struct dso *self, struct perf_session *session) return 0; } -static int perf_event__inject_buildid(union perf_event *event, +static int perf_event__inject_buildid(struct perf_event_ops *ops, + union perf_event *event, struct perf_sample *sample, struct perf_evsel *evsel __used, struct perf_session *session) @@ -161,7 +174,7 @@ static int perf_event__inject_buildid(union perf_event *event, if (!al.map->dso->hit) { al.map->dso->hit = 1; if (map__load(al.map, NULL) >= 0) { - dso__inject_build_id(al.map->dso, session); + dso__inject_build_id(al.map->dso, ops, session); /* * If this fails, too bad, let the other side * account this as unresolved. @@ -174,7 +187,7 @@ static int perf_event__inject_buildid(union perf_event *event, } repipe: - perf_event__repipe(event, sample, session); + perf_event__repipe(ops, event, sample, session); return 0; } @@ -189,9 +202,9 @@ struct perf_event_ops inject_ops = { .throttle = perf_event__repipe, .unthrottle = perf_event__repipe, .attr = perf_event__repipe_attr, - .event_type = perf_event__repipe_synth, - .tracing_data = perf_event__repipe_synth, - .build_id = perf_event__repipe_synth, + .event_type = perf_event__repipe_synth, + .tracing_data = perf_event__repipe_tracing_data_synth, + .build_id = perf_event__repipe_synth, }; extern volatile int session_done; diff --git a/tools/perf/builtin-kmem.c b/tools/perf/builtin-kmem.c index 225e963df105..5d01218e50e0 100644 --- a/tools/perf/builtin-kmem.c +++ b/tools/perf/builtin-kmem.c @@ -303,7 +303,8 @@ static void process_raw_event(union perf_event *raw_event __used, void *data, } } -static int process_sample_event(union perf_event *event, +static int process_sample_event(struct perf_event_ops *ops __used, + union perf_event *event, struct perf_sample *sample, struct perf_evsel *evsel __used, struct perf_session *session) diff --git a/tools/perf/builtin-lock.c b/tools/perf/builtin-lock.c index 899080ace267..f06b0a44c7cb 100644 --- a/tools/perf/builtin-lock.c +++ b/tools/perf/builtin-lock.c @@ -845,7 +845,8 @@ static void dump_info(void) die("Unknown type of information\n"); } -static int process_sample_event(union perf_event *event, +static int process_sample_event(struct perf_event_ops *ops __used, + union perf_event *event, struct perf_sample *sample, struct perf_evsel *evsel __used, struct perf_session *s) diff --git a/tools/perf/builtin-record.c b/tools/perf/builtin-record.c index ba6777a147ca..4642d38b8d19 100644 --- a/tools/perf/builtin-record.c +++ b/tools/perf/builtin-record.c @@ -35,43 +35,36 @@ enum write_mode_t { WRITE_APPEND }; -struct perf_record_opts record_opts = { - .target_pid = -1, - .target_tid = -1, - .mmap_pages = UINT_MAX, - .user_freq = UINT_MAX, - .user_interval = ULLONG_MAX, - .freq = 1000, - .sample_id_all_avail = true, +struct perf_record { + struct perf_event_ops ops; + struct perf_record_opts opts; + u64 bytes_written; + const char *output_name; + struct perf_evlist *evlist; + struct perf_session *session; + const char *progname; + int output; + unsigned int page_size; + int realtime_prio; + enum write_mode_t write_mode; + bool no_buildid; + bool no_buildid_cache; + bool force; + bool file_new; + bool append_file; + long samples; + off_t post_processing_offset; }; -static unsigned int page_size; -static int output; -static const char *output_name = NULL; -static int realtime_prio = 0; -static enum write_mode_t write_mode = WRITE_FORCE; -static bool no_buildid = false; -static bool no_buildid_cache = false; -static struct perf_evlist *evsel_list; - -static long samples = 0; -static u64 bytes_written = 0; - -static int file_new = 1; -static off_t post_processing_offset; - -static struct perf_session *session; -static const char *progname; - -static void advance_output(size_t size) +static void advance_output(struct perf_record *rec, size_t size) { - bytes_written += size; + rec->bytes_written += size; } -static void write_output(void *buf, size_t size) +static void write_output(struct perf_record *rec, void *buf, size_t size) { while (size) { - int ret = write(output, buf, size); + int ret = write(rec->output, buf, size); if (ret < 0) die("failed to write"); @@ -79,30 +72,33 @@ static void write_output(void *buf, size_t size) size -= ret; buf += ret; - bytes_written += ret; + rec->bytes_written += ret; } } -static int process_synthesized_event(union perf_event *event, +static int process_synthesized_event(struct perf_event_ops *ops, + union perf_event *event, struct perf_sample *sample __used, struct perf_session *self __used) { - write_output(event, event->header.size); + struct perf_record *rec = container_of(ops, struct perf_record, ops); + write_output(rec, event, event->header.size); return 0; } -static void mmap_read(struct perf_mmap *md) +static void perf_record__mmap_read(struct perf_record *rec, + struct perf_mmap *md) { unsigned int head = perf_mmap__read_head(md); unsigned int old = md->prev; - unsigned char *data = md->base + page_size; + unsigned char *data = md->base + rec->page_size; unsigned long size; void *buf; if (old == head) return; - samples++; + rec->samples++; size = head - old; @@ -111,14 +107,14 @@ static void mmap_read(struct perf_mmap *md) size = md->mask + 1 - (old & md->mask); old += size; - write_output(buf, size); + write_output(rec, buf, size); } buf = &data[old & md->mask]; size = head - old; old += size; - write_output(buf, size); + write_output(rec, buf, size); md->prev = old; perf_mmap__write_tail(md, old); @@ -137,17 +133,18 @@ static void sig_handler(int sig) signr = sig; } -static void sig_atexit(void) +static void perf_record__sig_exit(int exit_status __used, void *arg) { + struct perf_record *rec = arg; int status; - if (evsel_list->workload.pid > 0) { + if (rec->evlist->workload.pid > 0) { if (!child_finished) - kill(evsel_list->workload.pid, SIGTERM); + kill(rec->evlist->workload.pid, SIGTERM); wait(&status); if (WIFSIGNALED(status)) - psignal(WTERMSIG(status), progname); + psignal(WTERMSIG(status), rec->progname); } if (signr == -1 || signr == SIGUSR1) @@ -176,13 +173,16 @@ static bool perf_evlist__equal(struct perf_evlist *evlist, return true; } -static void open_counters(struct perf_evlist *evlist) +static void perf_record__open(struct perf_record *rec) { struct perf_evsel *pos, *first; + struct perf_evlist *evlist = rec->evlist; + struct perf_session *session = rec->session; + struct perf_record_opts *opts = &rec->opts; first = list_entry(evlist->entries.next, struct perf_evsel, node); - perf_evlist__config_attrs(evlist, &record_opts); + perf_evlist__config_attrs(evlist, opts); list_for_each_entry(pos, &evlist->entries, node) { struct perf_event_attr *attr = &pos->attr; @@ -201,27 +201,27 @@ static void open_counters(struct perf_evlist *evlist) */ bool time_needed = attr->sample_type & PERF_SAMPLE_TIME; - if (record_opts.group && pos != first) + if (opts->group && pos != first) group_fd = first->fd; retry_sample_id: - attr->sample_id_all = record_opts.sample_id_all_avail ? 1 : 0; + attr->sample_id_all = opts->sample_id_all_avail ? 1 : 0; try_again: if (perf_evsel__open(pos, evlist->cpus, evlist->threads, - record_opts.group, group_fd) < 0) { + opts->group, group_fd) < 0) { int err = errno; if (err == EPERM || err == EACCES) { ui__error_paranoid(); exit(EXIT_FAILURE); - } else if (err == ENODEV && record_opts.cpu_list) { + } else if (err == ENODEV && opts->cpu_list) { die("No such device - did you specify" " an out-of-range profile CPU?\n"); - } else if (err == EINVAL && record_opts.sample_id_all_avail) { + } else if (err == EINVAL && opts->sample_id_all_avail) { /* * Old kernel, no attr->sample_id_type_all field */ - record_opts.sample_id_all_avail = false; - if (!record_opts.sample_time && !record_opts.raw_samples && !time_needed) + opts->sample_id_all_avail = false; + if (!opts->sample_time && !opts->raw_samples && !time_needed) attr->sample_type &= ~PERF_SAMPLE_TIME; goto retry_sample_id; @@ -271,10 +271,10 @@ try_again: exit(-1); } - if (perf_evlist__mmap(evlist, record_opts.mmap_pages, false) < 0) + if (perf_evlist__mmap(evlist, opts->mmap_pages, false) < 0) die("failed to mmap with %d (%s)\n", errno, strerror(errno)); - if (file_new) + if (rec->file_new) session->evlist = evlist; else { if (!perf_evlist__equal(session->evlist, evlist)) { @@ -286,29 +286,32 @@ try_again: perf_session__update_sample_type(session); } -static int process_buildids(void) +static int process_buildids(struct perf_record *rec) { - u64 size = lseek(output, 0, SEEK_CUR); + u64 size = lseek(rec->output, 0, SEEK_CUR); if (size == 0) return 0; - session->fd = output; - return __perf_session__process_events(session, post_processing_offset, - size - post_processing_offset, + rec->session->fd = rec->output; + return __perf_session__process_events(rec->session, rec->post_processing_offset, + size - rec->post_processing_offset, size, &build_id__mark_dso_hit_ops); } -static void atexit_header(void) +static void perf_record__exit(int status __used, void *arg) { - if (!record_opts.pipe_output) { - session->header.data_size += bytes_written; - - if (!no_buildid) - process_buildids(); - perf_session__write_header(session, evsel_list, output, true); - perf_session__delete(session); - perf_evlist__delete(evsel_list); + struct perf_record *rec = arg; + + if (!rec->opts.pipe_output) { + rec->session->header.data_size += rec->bytes_written; + + if (!rec->no_buildid) + process_buildids(rec); + perf_session__write_header(rec->session, rec->evlist, + rec->output, true); + perf_session__delete(rec->session); + perf_evlist__delete(rec->evlist); symbol__exit(); } } @@ -316,7 +319,9 @@ static void atexit_header(void) static void perf_event__synthesize_guest_os(struct machine *machine, void *data) { int err; - struct perf_session *psession = data; + struct perf_event_ops *ops = data; + struct perf_record *rec = container_of(ops, struct perf_record, ops); + struct perf_session *psession = rec->session; if (machine__is_host(machine)) return; @@ -329,7 +334,7 @@ static void perf_event__synthesize_guest_os(struct machine *machine, void *data) *method is used to avoid symbol missing when the first addr is *in module instead of in guest kernel. */ - err = perf_event__synthesize_modules(process_synthesized_event, + err = perf_event__synthesize_modules(ops, process_synthesized_event, psession, machine); if (err < 0) pr_err("Couldn't record guest kernel [%d]'s reference" @@ -339,10 +344,10 @@ static void perf_event__synthesize_guest_os(struct machine *machine, void *data) * We use _stext for guest kernel because guest kernel's /proc/kallsyms * have no _text sometimes. */ - err = perf_event__synthesize_kernel_mmap(process_synthesized_event, + err = perf_event__synthesize_kernel_mmap(ops, process_synthesized_event, psession, machine, "_text"); if (err < 0) - err = perf_event__synthesize_kernel_mmap(process_synthesized_event, + err = perf_event__synthesize_kernel_mmap(ops, process_synthesized_event, psession, machine, "_stext"); if (err < 0) @@ -355,66 +360,71 @@ static struct perf_event_header finished_round_event = { .type = PERF_RECORD_FINISHED_ROUND, }; -static void mmap_read_all(void) +static void perf_record__mmap_read_all(struct perf_record *rec) { int i; - for (i = 0; i < evsel_list->nr_mmaps; i++) { - if (evsel_list->mmap[i].base) - mmap_read(&evsel_list->mmap[i]); + for (i = 0; i < rec->evlist->nr_mmaps; i++) { + if (rec->evlist->mmap[i].base) + perf_record__mmap_read(rec, &rec->evlist->mmap[i]); } - if (perf_header__has_feat(&session->header, HEADER_TRACE_INFO)) - write_output(&finished_round_event, sizeof(finished_round_event)); + if (perf_header__has_feat(&rec->session->header, HEADER_TRACE_INFO)) + write_output(rec, &finished_round_event, sizeof(finished_round_event)); } -static int __cmd_record(int argc, const char **argv) +static int __cmd_record(struct perf_record *rec, int argc, const char **argv) { struct stat st; int flags; - int err; + int err, output; unsigned long waking = 0; const bool forks = argc > 0; struct machine *machine; + struct perf_event_ops *ops = &rec->ops; + struct perf_record_opts *opts = &rec->opts; + struct perf_evlist *evsel_list = rec->evlist; + const char *output_name = rec->output_name; + struct perf_session *session; - progname = argv[0]; + rec->progname = argv[0]; - page_size = sysconf(_SC_PAGE_SIZE); + rec->page_size = sysconf(_SC_PAGE_SIZE); - atexit(sig_atexit); + on_exit(perf_record__sig_exit, rec); signal(SIGCHLD, sig_handler); signal(SIGINT, sig_handler); signal(SIGUSR1, sig_handler); if (!output_name) { if (!fstat(STDOUT_FILENO, &st) && S_ISFIFO(st.st_mode)) - record_opts.pipe_output = true; + opts->pipe_output = true; else - output_name = "perf.data"; + rec->output_name = output_name = "perf.data"; } if (output_name) { if (!strcmp(output_name, "-")) - record_opts.pipe_output = true; + opts->pipe_output = true; else if (!stat(output_name, &st) && st.st_size) { - if (write_mode == WRITE_FORCE) { + if (rec->write_mode == WRITE_FORCE) { char oldname[PATH_MAX]; snprintf(oldname, sizeof(oldname), "%s.old", output_name); unlink(oldname); rename(output_name, oldname); } - } else if (write_mode == WRITE_APPEND) { - write_mode = WRITE_FORCE; + } else if (rec->write_mode == WRITE_APPEND) { + rec->write_mode = WRITE_FORCE; } } flags = O_CREAT|O_RDWR; - if (write_mode == WRITE_APPEND) - file_new = 0; + if (rec->write_mode == WRITE_APPEND) + rec->file_new = 0; else flags |= O_TRUNC; - if (record_opts.pipe_output) + if (opts->pipe_output) output = STDOUT_FILENO; else output = open(output_name, flags, S_IRUSR | S_IWUSR); @@ -423,17 +433,21 @@ static int __cmd_record(int argc, const char **argv) exit(-1); } + rec->output = output; + session = perf_session__new(output_name, O_WRONLY, - write_mode == WRITE_FORCE, false, NULL); + rec->write_mode == WRITE_FORCE, false, NULL); if (session == NULL) { pr_err("Not enough memory for reading perf file header\n"); return -1; } - if (!no_buildid) + rec->session = session; + + if (!rec->no_buildid) perf_header__set_feat(&session->header, HEADER_BUILD_ID); - if (!file_new) { + if (!rec->file_new) { err = perf_session__read_header(session, output); if (err < 0) goto out_delete_session; @@ -456,42 +470,42 @@ static int __cmd_record(int argc, const char **argv) perf_header__set_feat(&session->header, HEADER_CPUID); if (forks) { - err = perf_evlist__prepare_workload(evsel_list, &record_opts, argv); + err = perf_evlist__prepare_workload(evsel_list, opts, argv); if (err < 0) { pr_err("Couldn't run the workload!\n"); goto out_delete_session; } } - open_counters(evsel_list); + perf_record__open(rec); /* - * perf_session__delete(session) will be called at atexit_header() + * perf_session__delete(session) will be called at perf_record__exit() */ - atexit(atexit_header); + on_exit(perf_record__exit, rec); - if (record_opts.pipe_output) { + if (opts->pipe_output) { err = perf_header__write_pipe(output); if (err < 0) return err; - } else if (file_new) { + } else if (rec->file_new) { err = perf_session__write_header(session, evsel_list, output, false); if (err < 0) return err; } - post_processing_offset = lseek(output, 0, SEEK_CUR); + rec->post_processing_offset = lseek(output, 0, SEEK_CUR); - if (record_opts.pipe_output) { - err = perf_session__synthesize_attrs(session, - process_synthesized_event); + if (opts->pipe_output) { + err = perf_event__synthesize_attrs(ops, session, + process_synthesized_event); if (err < 0) { pr_err("Couldn't synthesize attrs.\n"); return err; } - err = perf_event__synthesize_event_types(process_synthesized_event, + err = perf_event__synthesize_event_types(ops, process_synthesized_event, session); if (err < 0) { pr_err("Couldn't synthesize event_types.\n"); @@ -507,14 +521,14 @@ static int __cmd_record(int argc, const char **argv) * return this more properly and also * propagate errors that now are calling die() */ - err = perf_event__synthesize_tracing_data(output, evsel_list, + err = perf_event__synthesize_tracing_data(ops, output, evsel_list, process_synthesized_event, session); if (err <= 0) { pr_err("Couldn't record tracing data.\n"); return err; } - advance_output(err); + advance_output(rec, err); } } @@ -524,17 +538,17 @@ static int __cmd_record(int argc, const char **argv) return -1; } - err = perf_event__synthesize_kernel_mmap(process_synthesized_event, + err = perf_event__synthesize_kernel_mmap(ops, process_synthesized_event, session, machine, "_text"); if (err < 0) - err = perf_event__synthesize_kernel_mmap(process_synthesized_event, + err = perf_event__synthesize_kernel_mmap(ops, process_synthesized_event, session, machine, "_stext"); if (err < 0) pr_err("Couldn't record kernel reference relocation symbol\n" "Symbol resolution may be skewed if relocation was used (e.g. kexec).\n" "Check /proc/kallsyms permission or run as root.\n"); - err = perf_event__synthesize_modules(process_synthesized_event, + err = perf_event__synthesize_modules(ops, process_synthesized_event, session, machine); if (err < 0) pr_err("Couldn't record kernel module information.\n" @@ -542,21 +556,21 @@ static int __cmd_record(int argc, const char **argv) "Check /proc/modules permission or run as root.\n"); if (perf_guest) - perf_session__process_machines(session, + perf_session__process_machines(session, ops, perf_event__synthesize_guest_os); - if (!record_opts.system_wide) - perf_event__synthesize_thread_map(evsel_list->threads, + if (!opts->system_wide) + perf_event__synthesize_thread_map(ops, evsel_list->threads, process_synthesized_event, session); else - perf_event__synthesize_threads(process_synthesized_event, + perf_event__synthesize_threads(ops, process_synthesized_event, session); - if (realtime_prio) { + if (rec->realtime_prio) { struct sched_param param; - param.sched_priority = realtime_prio; + param.sched_priority = rec->realtime_prio; if (sched_setscheduler(0, SCHED_FIFO, ¶m)) { pr_err("Could not set realtime priority.\n"); exit(-1); @@ -572,11 +586,11 @@ static int __cmd_record(int argc, const char **argv) perf_evlist__start_workload(evsel_list); for (;;) { - int hits = samples; + int hits = rec->samples; - mmap_read_all(); + perf_record__mmap_read_all(rec); - if (hits == samples) { + if (hits == rec->samples) { if (done) break; err = poll(evsel_list->pollfd, evsel_list->nr_fds, -1); @@ -597,9 +611,9 @@ static int __cmd_record(int argc, const char **argv) */ fprintf(stderr, "[ perf record: Captured and wrote %.3f MB %s (~%" PRIu64 " samples) ]\n", - (double)bytes_written / 1024.0 / 1024.0, + (double)rec->bytes_written / 1024.0 / 1024.0, output_name, - bytes_written / 24); + rec->bytes_written / 24); return 0; @@ -614,59 +628,88 @@ static const char * const record_usage[] = { NULL }; -static bool force, append_file; +/* + * XXX Ideally would be local to cmd_record() and passed to a perf_record__new + * because we need to have access to it in perf_record__exit, that is called + * after cmd_record() exits, but since record_options need to be accessible to + * builtin-script, leave it here. + * + * At least we don't ouch it in all the other functions here directly. + * + * Just say no to tons of global variables, sigh. + */ +static struct perf_record record = { + .opts = { + .target_pid = -1, + .target_tid = -1, + .mmap_pages = UINT_MAX, + .user_freq = UINT_MAX, + .user_interval = ULLONG_MAX, + .freq = 1000, + .sample_id_all_avail = true, + }, + .write_mode = WRITE_FORCE, + .file_new = true, +}; +/* + * XXX Will stay a global variable till we fix builtin-script.c to stop messing + * with it and switch to use the library functions in perf_evlist that came + * from builtin-record.c, i.e. use perf_record_opts, + * perf_evlist__prepare_workload, etc instead of fork+exec'in 'perf record', + * using pipes, etc. + */ const struct option record_options[] = { - OPT_CALLBACK('e', "event", &evsel_list, "event", + OPT_CALLBACK('e', "event", &record.evlist, "event", "event selector. use 'perf list' to list available events", parse_events_option), - OPT_CALLBACK(0, "filter", &evsel_list, "filter", + OPT_CALLBACK(0, "filter", &record.evlist, "filter", "event filter", parse_filter), - OPT_INTEGER('p', "pid", &record_opts.target_pid, + OPT_INTEGER('p', "pid", &record.opts.target_pid, "record events on existing process id"), - OPT_INTEGER('t', "tid", &record_opts.target_tid, + OPT_INTEGER('t', "tid", &record.opts.target_tid, "record events on existing thread id"), - OPT_INTEGER('r', "realtime", &realtime_prio, + OPT_INTEGER('r', "realtime", &record.realtime_prio, "collect data with this RT SCHED_FIFO priority"), - OPT_BOOLEAN('D', "no-delay", &record_opts.no_delay, + OPT_BOOLEAN('D', "no-delay", &record.opts.no_delay, "collect data without buffering"), - OPT_BOOLEAN('R', "raw-samples", &record_opts.raw_samples, + OPT_BOOLEAN('R', "raw-samples", &record.opts.raw_samples, "collect raw sample records from all opened counters"), - OPT_BOOLEAN('a', "all-cpus", &record_opts.system_wide, + OPT_BOOLEAN('a', "all-cpus", &record.opts.system_wide, "system-wide collection from all CPUs"), - OPT_BOOLEAN('A', "append", &append_file, + OPT_BOOLEAN('A', "append", &record.append_file, "append to the output file to do incremental profiling"), - OPT_STRING('C', "cpu", &record_opts.cpu_list, "cpu", + OPT_STRING('C', "cpu", &record.opts.cpu_list, "cpu", "list of cpus to monitor"), - OPT_BOOLEAN('f', "force", &force, + OPT_BOOLEAN('f', "force", &record.force, "overwrite existing data file (deprecated)"), - OPT_U64('c', "count", &record_opts.user_interval, "event period to sample"), - OPT_STRING('o', "output", &output_name, "file", + OPT_U64('c', "count", &record.opts.user_interval, "event period to sample"), + OPT_STRING('o', "output", &record.output_name, "file", "output file name"), - OPT_BOOLEAN('i', "no-inherit", &record_opts.no_inherit, + OPT_BOOLEAN('i', "no-inherit", &record.opts.no_inherit, "child tasks do not inherit counters"), - OPT_UINTEGER('F', "freq", &record_opts.user_freq, "profile at this frequency"), - OPT_UINTEGER('m', "mmap-pages", &record_opts.mmap_pages, + OPT_UINTEGER('F', "freq", &record.opts.user_freq, "profile at this frequency"), + OPT_UINTEGER('m', "mmap-pages", &record.opts.mmap_pages, "number of mmap data pages"), - OPT_BOOLEAN(0, "group", &record_opts.group, + OPT_BOOLEAN(0, "group", &record.opts.group, "put the counters into a counter group"), - OPT_BOOLEAN('g', "call-graph", &record_opts.call_graph, + OPT_BOOLEAN('g', "call-graph", &record.opts.call_graph, "do call-graph (stack chain/backtrace) recording"), OPT_INCR('v', "verbose", &verbose, "be more verbose (show counter open errors, etc)"), OPT_BOOLEAN('q', "quiet", &quiet, "don't print any message"), - OPT_BOOLEAN('s', "stat", &record_opts.inherit_stat, + OPT_BOOLEAN('s', "stat", &record.opts.inherit_stat, "per thread counts"), - OPT_BOOLEAN('d', "data", &record_opts.sample_address, + OPT_BOOLEAN('d', "data", &record.opts.sample_address, "Sample addresses"), - OPT_BOOLEAN('T', "timestamp", &record_opts.sample_time, "Sample timestamps"), - OPT_BOOLEAN('n', "no-samples", &record_opts.no_samples, + OPT_BOOLEAN('T', "timestamp", &record.opts.sample_time, "Sample timestamps"), + OPT_BOOLEAN('n', "no-samples", &record.opts.no_samples, "don't sample"), - OPT_BOOLEAN('N', "no-buildid-cache", &no_buildid_cache, + OPT_BOOLEAN('N', "no-buildid-cache", &record.no_buildid_cache, "do not update the buildid cache"), - OPT_BOOLEAN('B', "no-buildid", &no_buildid, + OPT_BOOLEAN('B', "no-buildid", &record.no_buildid, "do not collect buildids in perf.data"), - OPT_CALLBACK('G', "cgroup", &evsel_list, "name", + OPT_CALLBACK('G', "cgroup", &record.evlist, "name", "monitor event in cgroup name only", parse_cgroups), OPT_END() @@ -676,6 +719,8 @@ int cmd_record(int argc, const char **argv, const char *prefix __used) { int err = -ENOMEM; struct perf_evsel *pos; + struct perf_evlist *evsel_list; + struct perf_record *rec = &record; perf_header__set_cmdline(argc, argv); @@ -683,23 +728,25 @@ int cmd_record(int argc, const char **argv, const char *prefix __used) if (evsel_list == NULL) return -ENOMEM; + rec->evlist = evsel_list; + argc = parse_options(argc, argv, record_options, record_usage, PARSE_OPT_STOP_AT_NON_OPTION); - if (!argc && record_opts.target_pid == -1 && record_opts.target_tid == -1 && - !record_opts.system_wide && !record_opts.cpu_list) + if (!argc && rec->opts.target_pid == -1 && rec->opts.target_tid == -1 && + !rec->opts.system_wide && !rec->opts.cpu_list) usage_with_options(record_usage, record_options); - if (force && append_file) { + if (rec->force && rec->append_file) { fprintf(stderr, "Can't overwrite and append at the same time." " You need to choose between -f and -A"); usage_with_options(record_usage, record_options); - } else if (append_file) { - write_mode = WRITE_APPEND; + } else if (rec->append_file) { + rec->write_mode = WRITE_APPEND; } else { - write_mode = WRITE_FORCE; + rec->write_mode = WRITE_FORCE; } - if (nr_cgroups && !record_opts.system_wide) { + if (nr_cgroups && !rec->opts.system_wide) { fprintf(stderr, "cgroup monitoring only available in" " system-wide mode\n"); usage_with_options(record_usage, record_options); @@ -717,7 +764,7 @@ int cmd_record(int argc, const char **argv, const char *prefix __used) "If some relocation was applied (e.g. kexec) symbols may be misresolved\n" "even with a suitable vmlinux or kallsyms file.\n\n"); - if (no_buildid_cache || no_buildid) + if (rec->no_buildid_cache || rec->no_buildid) disable_buildid_cache(); if (evsel_list->nr_entries == 0 && @@ -726,11 +773,11 @@ int cmd_record(int argc, const char **argv, const char *prefix __used) goto out_symbol_exit; } - if (record_opts.target_pid != -1) - record_opts.target_tid = record_opts.target_pid; + if (rec->opts.target_pid != -1) + rec->opts.target_tid = rec->opts.target_pid; - if (perf_evlist__create_maps(evsel_list, record_opts.target_pid, - record_opts.target_tid, record_opts.cpu_list) < 0) + if (perf_evlist__create_maps(evsel_list, rec->opts.target_pid, + rec->opts.target_tid, rec->opts.cpu_list) < 0) usage_with_options(record_usage, record_options); list_for_each_entry(pos, &evsel_list->entries, node) { @@ -744,25 +791,25 @@ int cmd_record(int argc, const char **argv, const char *prefix __used) if (perf_evlist__alloc_pollfd(evsel_list) < 0) goto out_free_fd; - if (record_opts.user_interval != ULLONG_MAX) - record_opts.default_interval = record_opts.user_interval; - if (record_opts.user_freq != UINT_MAX) - record_opts.freq = record_opts.user_freq; + if (rec->opts.user_interval != ULLONG_MAX) + rec->opts.default_interval = rec->opts.user_interval; + if (rec->opts.user_freq != UINT_MAX) + rec->opts.freq = rec->opts.user_freq; /* * User specified count overrides default frequency. */ - if (record_opts.default_interval) - record_opts.freq = 0; - else if (record_opts.freq) { - record_opts.default_interval = record_opts.freq; + if (rec->opts.default_interval) + rec->opts.freq = 0; + else if (rec->opts.freq) { + rec->opts.default_interval = rec->opts.freq; } else { fprintf(stderr, "frequency and count are zero, aborting\n"); err = -EINVAL; goto out_free_fd; } - err = __cmd_record(argc, argv); + err = __cmd_record(&record, argc, argv); out_free_fd: perf_evlist__delete_maps(evsel_list); out_symbol_exit: diff --git a/tools/perf/builtin-report.c b/tools/perf/builtin-report.c index 5d2e819dfc40..8795520f6e1d 100644 --- a/tools/perf/builtin-report.c +++ b/tools/perf/builtin-report.c @@ -35,7 +35,9 @@ #include -static struct perf_report { +struct perf_report { + struct perf_event_ops ops; + struct perf_session *session; char const *input_name; bool force, use_tui, use_stdio; bool hide_unresolved; @@ -48,12 +50,7 @@ static struct perf_report { symbol_filter_t annotate_init; const char *cpu_list; DECLARE_BITMAP(cpu_bitmap, MAX_NR_CPUS); -} report = { - .input_name = "perf.data", - .pretty_printing_style = "normal", -}, *rep = &report; - -static char callchain_default_opt[] = "fractal,0.5,callee"; +}; static int perf_session__add_hist_entry(struct perf_session *session, struct addr_location *al, @@ -106,11 +103,13 @@ out: } -static int process_sample_event(union perf_event *event, +static int process_sample_event(struct perf_event_ops *ops, + union perf_event *event, struct perf_sample *sample, struct perf_evsel *evsel, struct perf_session *session) { + struct perf_report *rep = container_of(ops, struct perf_report, ops); struct addr_location al; if (perf_event__preprocess_sample(event, session, &al, sample, @@ -137,10 +136,12 @@ static int process_sample_event(union perf_event *event, return 0; } -static int process_read_event(union perf_event *event, +static int process_read_event(struct perf_event_ops *ops, + union perf_event *event, struct perf_sample *sample __used, struct perf_session *session) { + struct perf_report *rep = container_of(ops, struct perf_report, ops); struct perf_evsel *evsel = perf_evlist__id2evsel(session->evlist, event->read.id); if (rep->show_threads) { @@ -159,8 +160,10 @@ static int process_read_event(union perf_event *event, return 0; } -static int perf_session__setup_sample_type(struct perf_session *self) +static int perf_report__setup_sample_type(struct perf_report *rep) { + struct perf_session *self = rep->session; + if (!(self->sample_type & PERF_SAMPLE_CALLCHAIN)) { if (sort__has_parent) { ui__warning("Selected --sort parent, but no " @@ -187,22 +190,6 @@ static int perf_session__setup_sample_type(struct perf_session *self) return 0; } -static struct perf_event_ops event_ops = { - .sample = process_sample_event, - .mmap = perf_event__process_mmap, - .comm = perf_event__process_comm, - .exit = perf_event__process_task, - .fork = perf_event__process_task, - .lost = perf_event__process_lost, - .read = process_read_event, - .attr = perf_event__process_attr, - .event_type = perf_event__process_event_type, - .tracing_data = perf_event__process_tracing_data, - .build_id = perf_event__process_build_id, - .ordered_samples = true, - .ordering_requires_timestamps = true, -}; - extern volatile int session_done; static void sig_handler(int sig __used) @@ -225,6 +212,7 @@ static size_t hists__fprintf_nr_sample_events(struct hists *self, } static int perf_evlist__tty_browse_hists(struct perf_evlist *evlist, + struct perf_report *rep, const char *help) { struct perf_evsel *pos; @@ -253,7 +241,7 @@ static int perf_evlist__tty_browse_hists(struct perf_evlist *evlist, return 0; } -static int __cmd_report(void) +static int __cmd_report(struct perf_report *rep) { int ret = -EINVAL; u64 nr_samples; @@ -266,10 +254,12 @@ static int __cmd_report(void) signal(SIGINT, sig_handler); session = perf_session__new(rep->input_name, O_RDONLY, - rep->force, false, &event_ops); + rep->force, false, &rep->ops); if (session == NULL) return -ENOMEM; + rep->session = session; + if (rep->cpu_list) { ret = perf_session__cpu_bitmap(session, rep->cpu_list, rep->cpu_bitmap); @@ -283,11 +273,11 @@ static int __cmd_report(void) if (rep->show_threads) perf_read_values_init(&rep->show_threads_values); - ret = perf_session__setup_sample_type(session); + ret = perf_report__setup_sample_type(rep); if (ret) goto out_delete; - ret = perf_session__process_events(session, &event_ops); + ret = perf_session__process_events(session, &rep->ops); if (ret) goto out_delete; @@ -339,7 +329,7 @@ static int __cmd_report(void) perf_evlist__tui_browse_hists(session->evlist, help, NULL, NULL, 0); } else - perf_evlist__tty_browse_hists(session->evlist, help); + perf_evlist__tty_browse_hists(session->evlist, rep, help); out_delete: /* @@ -358,9 +348,9 @@ out_delete: } static int -parse_callchain_opt(const struct option *opt __used, const char *arg, - int unset) +parse_callchain_opt(const struct option *opt, const char *arg, int unset) { + struct perf_report *rep = (struct perf_report *)opt->value; char *tok, *tok2; char *endptr; @@ -437,12 +427,33 @@ setup: return 0; } -static const char * const report_usage[] = { - "perf report [] ", - NULL -}; - -static const struct option options[] = { +int cmd_report(int argc, const char **argv, const char *prefix __used) +{ + char callchain_default_opt[] = "fractal,0.5,callee"; + const char * const report_usage[] = { + "perf report [] ", + NULL + }; + struct perf_report report = { + .ops = { + .sample = process_sample_event, + .mmap = perf_event__process_mmap, + .comm = perf_event__process_comm, + .exit = perf_event__process_task, + .fork = perf_event__process_task, + .lost = perf_event__process_lost, + .read = process_read_event, + .attr = perf_event__process_attr, + .event_type = perf_event__process_event_type, + .tracing_data = perf_event__process_tracing_data, + .build_id = perf_event__process_build_id, + .ordered_samples = true, + .ordering_requires_timestamps = true, + }, + .input_name = "perf.data", + .pretty_printing_style = "normal", + }; + const struct option options[] = { OPT_STRING('i', "input", &report.input_name, "file", "input file name"), OPT_INCR('v', "verbose", &verbose, @@ -473,7 +484,7 @@ static const struct option options[] = { "regex filter to identify parent, see: '--sort parent'"), OPT_BOOLEAN('x', "exclude-other", &symbol_conf.exclude_other, "Only display entries with parent-match"), - OPT_CALLBACK_DEFAULT('g', "call-graph", NULL, "output_type,min_percent, call_order", + OPT_CALLBACK_DEFAULT('g', "call-graph", &report, "output_type,min_percent, call_order", "Display callchains using output_type (graph, flat, fractal, or none) , min percent threshold and callchain order. " "Default: fractal,0.5,callee", &parse_callchain_opt, callchain_default_opt), OPT_BOOLEAN('G', "inverted", &report.inverted_callchain, @@ -507,10 +518,8 @@ static const struct option options[] = { OPT_BOOLEAN(0, "show-total-period", &symbol_conf.show_total_period, "Show a column with the sum of periods"), OPT_END() -}; + }; -int cmd_report(int argc, const char **argv, const char *prefix __used) -{ argc = parse_options(argc, argv, options, report_usage, 0); if (report.use_stdio) @@ -579,5 +588,5 @@ int cmd_report(int argc, const char **argv, const char *prefix __used) sort_entry__setup_elide(&sort_comm, symbol_conf.comm_list, "comm", stdout); sort_entry__setup_elide(&sort_sym, symbol_conf.sym_list, "symbol", stdout); - return __cmd_report(); + return __cmd_report(&report); } diff --git a/tools/perf/builtin-sched.c b/tools/perf/builtin-sched.c index d51af0beab13..b11d6283fedf 100644 --- a/tools/perf/builtin-sched.c +++ b/tools/perf/builtin-sched.c @@ -1602,7 +1602,8 @@ static void process_raw_event(union perf_event *raw_event __used, process_sched_migrate_task_event(data, session, event, cpu, timestamp, thread); } -static int process_sample_event(union perf_event *event, +static int process_sample_event(struct perf_event_ops *ops __used, + union perf_event *event, struct perf_sample *sample, struct perf_evsel *evsel, struct perf_session *session) diff --git a/tools/perf/builtin-script.c b/tools/perf/builtin-script.c index 47545e9c9b27..3b7820612ebf 100644 --- a/tools/perf/builtin-script.c +++ b/tools/perf/builtin-script.c @@ -434,7 +434,8 @@ static int cleanup_scripting(void) static char const *input_name = "perf.data"; -static int process_sample_event(union perf_event *event, +static int process_sample_event(struct perf_event_ops *ops __used, + union perf_event *event, struct perf_sample *sample, struct perf_evsel *evsel, struct perf_session *session) diff --git a/tools/perf/builtin-timechart.c b/tools/perf/builtin-timechart.c index 3fc52b1aa430..62298a0d7dc9 100644 --- a/tools/perf/builtin-timechart.c +++ b/tools/perf/builtin-timechart.c @@ -274,7 +274,8 @@ static int cpus_cstate_state[MAX_CPUS]; static u64 cpus_pstate_start_times[MAX_CPUS]; static u64 cpus_pstate_state[MAX_CPUS]; -static int process_comm_event(union perf_event *event, +static int process_comm_event(struct perf_event_ops *ops __used, + union perf_event *event, struct perf_sample *sample __used, struct perf_session *session __used) { @@ -282,7 +283,8 @@ static int process_comm_event(union perf_event *event, return 0; } -static int process_fork_event(union perf_event *event, +static int process_fork_event(struct perf_event_ops *ops __used, + union perf_event *event, struct perf_sample *sample __used, struct perf_session *session __used) { @@ -290,7 +292,8 @@ static int process_fork_event(union perf_event *event, return 0; } -static int process_exit_event(union perf_event *event, +static int process_exit_event(struct perf_event_ops *ops __used, + union perf_event *event, struct perf_sample *sample __used, struct perf_session *session __used) { @@ -487,7 +490,8 @@ static void sched_switch(int cpu, u64 timestamp, struct trace_entry *te) } -static int process_sample_event(union perf_event *event __used, +static int process_sample_event(struct perf_event_ops *ops __used, + union perf_event *event __used, struct perf_sample *sample, struct perf_evsel *evsel, struct perf_session *session __used) diff --git a/tools/perf/builtin-top.c b/tools/perf/builtin-top.c index 9b3bbb40d46f..e8e3320602bd 100644 --- a/tools/perf/builtin-top.c +++ b/tools/perf/builtin-top.c @@ -824,7 +824,7 @@ static void perf_session__mmap_read_idx(struct perf_session *self, int idx) perf_event__process_sample(event, evsel, &sample, self); else if (event->header.type < PERF_RECORD_MAX) { hists__inc_nr_events(&evsel->hists, event->header.type); - perf_event__process(event, &sample, self); + perf_event__process(&top.ops, event, &sample, self); } else ++self->hists.stats.nr_unknown_events; } @@ -966,10 +966,10 @@ static int __cmd_top(void) goto out_delete; if (top.target_tid != -1) - perf_event__synthesize_thread_map(top.evlist->threads, + perf_event__synthesize_thread_map(&top.ops, top.evlist->threads, perf_event__process, top.session); else - perf_event__synthesize_threads(perf_event__process, top.session); + perf_event__synthesize_threads(&top.ops, perf_event__process, top.session); start_counters(top.evlist); top.session->evlist = top.evlist; diff --git a/tools/perf/util/build-id.c b/tools/perf/util/build-id.c index f2fe6ec08945..0e4de1865013 100644 --- a/tools/perf/util/build-id.c +++ b/tools/perf/util/build-id.c @@ -13,8 +13,10 @@ #include "symbol.h" #include #include "debug.h" +#include "session.h" -static int build_id__mark_dso_hit(union perf_event *event, +static int build_id__mark_dso_hit(struct perf_event_ops *ops __used, + union perf_event *event, struct perf_sample *sample __used, struct perf_evsel *evsel __used, struct perf_session *session) @@ -38,7 +40,8 @@ static int build_id__mark_dso_hit(union perf_event *event, return 0; } -static int perf_event__exit_del_thread(union perf_event *event, +static int perf_event__exit_del_thread(struct perf_event_ops *ops __used, + union perf_event *event, struct perf_sample *sample __used, struct perf_session *session) { diff --git a/tools/perf/util/callchain.h b/tools/perf/util/callchain.h index 9b4ff16cac96..7f9c0f1ae3a9 100644 --- a/tools/perf/util/callchain.h +++ b/tools/perf/util/callchain.h @@ -101,6 +101,9 @@ int callchain_append(struct callchain_root *root, int callchain_merge(struct callchain_cursor *cursor, struct callchain_root *dst, struct callchain_root *src); +struct ip_callchain; +union perf_event; + bool ip_callchain__valid(struct ip_callchain *chain, const union perf_event *event); /* diff --git a/tools/perf/util/event.c b/tools/perf/util/event.c index 437f8ca679a0..4800f38c7277 100644 --- a/tools/perf/util/event.c +++ b/tools/perf/util/event.c @@ -44,7 +44,8 @@ static struct perf_sample synth_sample = { .period = 1, }; -static pid_t perf_event__synthesize_comm(union perf_event *event, pid_t pid, +static pid_t perf_event__synthesize_comm(struct perf_event_ops *ops, + union perf_event *event, pid_t pid, int full, perf_event__handler_t process, struct perf_session *session) { @@ -99,7 +100,7 @@ out_race: if (!full) { event->comm.tid = pid; - process(event, &synth_sample, session); + process(ops, event, &synth_sample, session); goto out; } @@ -117,7 +118,7 @@ out_race: event->comm.tid = pid; - process(event, &synth_sample, session); + process(ops, event, &synth_sample, session); } closedir(tasks); @@ -127,7 +128,8 @@ out: return tgid; } -static int perf_event__synthesize_mmap_events(union perf_event *event, +static int perf_event__synthesize_mmap_events(struct perf_event_ops *ops, + union perf_event *event, pid_t pid, pid_t tgid, perf_event__handler_t process, struct perf_session *session) @@ -198,7 +200,7 @@ static int perf_event__synthesize_mmap_events(union perf_event *event, event->mmap.pid = tgid; event->mmap.tid = pid; - process(event, &synth_sample, session); + process(ops, event, &synth_sample, session); } } @@ -206,7 +208,8 @@ static int perf_event__synthesize_mmap_events(union perf_event *event, return 0; } -int perf_event__synthesize_modules(perf_event__handler_t process, +int perf_event__synthesize_modules(struct perf_event_ops *ops, + perf_event__handler_t process, struct perf_session *session, struct machine *machine) { @@ -251,7 +254,7 @@ int perf_event__synthesize_modules(perf_event__handler_t process, memcpy(event->mmap.filename, pos->dso->long_name, pos->dso->long_name_len + 1); - process(event, &synth_sample, session); + process(ops, event, &synth_sample, session); } free(event); @@ -261,17 +264,19 @@ int perf_event__synthesize_modules(perf_event__handler_t process, static int __event__synthesize_thread(union perf_event *comm_event, union perf_event *mmap_event, pid_t pid, perf_event__handler_t process, + struct perf_event_ops *ops, struct perf_session *session) { - pid_t tgid = perf_event__synthesize_comm(comm_event, pid, 1, process, + pid_t tgid = perf_event__synthesize_comm(ops, comm_event, pid, 1, process, session); if (tgid == -1) return -1; - return perf_event__synthesize_mmap_events(mmap_event, pid, tgid, + return perf_event__synthesize_mmap_events(ops, mmap_event, pid, tgid, process, session); } -int perf_event__synthesize_thread_map(struct thread_map *threads, +int perf_event__synthesize_thread_map(struct perf_event_ops *ops, + struct thread_map *threads, perf_event__handler_t process, struct perf_session *session) { @@ -290,7 +295,7 @@ int perf_event__synthesize_thread_map(struct thread_map *threads, for (thread = 0; thread < threads->nr; ++thread) { if (__event__synthesize_thread(comm_event, mmap_event, threads->map[thread], - process, session)) { + process, ops, session)) { err = -1; break; } @@ -302,7 +307,8 @@ out: return err; } -int perf_event__synthesize_threads(perf_event__handler_t process, +int perf_event__synthesize_threads(struct perf_event_ops *ops, + perf_event__handler_t process, struct perf_session *session) { DIR *proc; @@ -330,7 +336,7 @@ int perf_event__synthesize_threads(perf_event__handler_t process, continue; __event__synthesize_thread(comm_event, mmap_event, pid, - process, session); + process, ops, session); } closedir(proc); @@ -365,7 +371,8 @@ static int find_symbol_cb(void *arg, const char *name, char type, return 1; } -int perf_event__synthesize_kernel_mmap(perf_event__handler_t process, +int perf_event__synthesize_kernel_mmap(struct perf_event_ops *ops, + perf_event__handler_t process, struct perf_session *session, struct machine *machine, const char *symbol_name) @@ -423,13 +430,14 @@ int perf_event__synthesize_kernel_mmap(perf_event__handler_t process, event->mmap.len = map->end - event->mmap.start; event->mmap.pid = machine->pid; - err = process(event, &synth_sample, session); + err = process(ops, event, &synth_sample, session); free(event); return err; } -int perf_event__process_comm(union perf_event *event, +int perf_event__process_comm(struct perf_event_ops *ops __used, + union perf_event *event, struct perf_sample *sample __used, struct perf_session *session) { @@ -445,7 +453,8 @@ int perf_event__process_comm(union perf_event *event, return 0; } -int perf_event__process_lost(union perf_event *event, +int perf_event__process_lost(struct perf_event_ops *ops __used, + union perf_event *event, struct perf_sample *sample __used, struct perf_session *session) { @@ -468,7 +477,8 @@ static void perf_event__set_kernel_mmap_len(union perf_event *event, maps[MAP__FUNCTION]->end = ~0ULL; } -static int perf_event__process_kernel_mmap(union perf_event *event, +static int perf_event__process_kernel_mmap(struct perf_event_ops *ops __used, + union perf_event *event, struct perf_session *session) { struct map *map; @@ -567,7 +577,8 @@ out_problem: return -1; } -int perf_event__process_mmap(union perf_event *event, +int perf_event__process_mmap(struct perf_event_ops *ops, + union perf_event *event, struct perf_sample *sample __used, struct perf_session *session) { @@ -583,7 +594,7 @@ int perf_event__process_mmap(union perf_event *event, if (cpumode == PERF_RECORD_MISC_GUEST_KERNEL || cpumode == PERF_RECORD_MISC_KERNEL) { - ret = perf_event__process_kernel_mmap(event, session); + ret = perf_event__process_kernel_mmap(ops, event, session); if (ret < 0) goto out_problem; return 0; @@ -610,7 +621,8 @@ out_problem: return 0; } -int perf_event__process_task(union perf_event *event, +int perf_event__process_task(struct perf_event_ops *ops __used, + union perf_event *event, struct perf_sample *sample __used, struct perf_session *session) { @@ -634,22 +646,22 @@ int perf_event__process_task(union perf_event *event, return 0; } -int perf_event__process(union perf_event *event, struct perf_sample *sample, - struct perf_session *session) +int perf_event__process(struct perf_event_ops *ops, union perf_event *event, + struct perf_sample *sample, struct perf_session *session) { switch (event->header.type) { case PERF_RECORD_COMM: - perf_event__process_comm(event, sample, session); + perf_event__process_comm(ops, event, sample, session); break; case PERF_RECORD_MMAP: - perf_event__process_mmap(event, sample, session); + perf_event__process_mmap(ops, event, sample, session); break; case PERF_RECORD_FORK: case PERF_RECORD_EXIT: - perf_event__process_task(event, sample, session); + perf_event__process_task(ops, event, sample, session); break; case PERF_RECORD_LOST: - perf_event__process_lost(event, sample, session); + perf_event__process_lost(ops, event, sample, session); default: break; } diff --git a/tools/perf/util/event.h b/tools/perf/util/event.h index 357a85b85248..669409d35710 100644 --- a/tools/perf/util/event.h +++ b/tools/perf/util/event.h @@ -141,38 +141,52 @@ union perf_event { void perf_event__print_totals(void); +struct perf_event_ops; struct perf_session; struct thread_map; -typedef int (*perf_event__handler_synth_t)(union perf_event *event, - struct perf_session *session); -typedef int (*perf_event__handler_t)(union perf_event *event, +typedef int (*perf_event__handler_t)(struct perf_event_ops *ops, + union perf_event *event, struct perf_sample *sample, struct perf_session *session); -int perf_event__synthesize_thread_map(struct thread_map *threads, +int perf_event__synthesize_thread_map(struct perf_event_ops *ops, + struct thread_map *threads, perf_event__handler_t process, struct perf_session *session); -int perf_event__synthesize_threads(perf_event__handler_t process, +int perf_event__synthesize_threads(struct perf_event_ops *ops, + perf_event__handler_t process, struct perf_session *session); -int perf_event__synthesize_kernel_mmap(perf_event__handler_t process, +int perf_event__synthesize_kernel_mmap(struct perf_event_ops *ops, + perf_event__handler_t process, struct perf_session *session, struct machine *machine, const char *symbol_name); -int perf_event__synthesize_modules(perf_event__handler_t process, +int perf_event__synthesize_modules(struct perf_event_ops *ops, + perf_event__handler_t process, struct perf_session *session, struct machine *machine); -int perf_event__process_comm(union perf_event *event, struct perf_sample *sample, +int perf_event__process_comm(struct perf_event_ops *ops, + union perf_event *event, + struct perf_sample *sample, struct perf_session *session); -int perf_event__process_lost(union perf_event *event, struct perf_sample *sample, +int perf_event__process_lost(struct perf_event_ops *ops, + union perf_event *event, + struct perf_sample *sample, struct perf_session *session); -int perf_event__process_mmap(union perf_event *event, struct perf_sample *sample, +int perf_event__process_mmap(struct perf_event_ops *ops, + union perf_event *event, + struct perf_sample *sample, struct perf_session *session); -int perf_event__process_task(union perf_event *event, struct perf_sample *sample, +int perf_event__process_task(struct perf_event_ops *ops, + union perf_event *event, + struct perf_sample *sample, struct perf_session *session); -int perf_event__process(union perf_event *event, struct perf_sample *sample, +int perf_event__process(struct perf_event_ops *ops, + union perf_event *event, + struct perf_sample *sample, struct perf_session *session); struct addr_location; diff --git a/tools/perf/util/header.c b/tools/perf/util/header.c index 1fa97dd21200..ab3a2b0e8f06 100644 --- a/tools/perf/util/header.c +++ b/tools/perf/util/header.c @@ -2070,7 +2070,8 @@ out_delete_evlist: return -ENOMEM; } -int perf_event__synthesize_attr(struct perf_event_attr *attr, u16 ids, u64 *id, +int perf_event__synthesize_attr(struct perf_event_ops *ops, + struct perf_event_attr *attr, u16 ids, u64 *id, perf_event__handler_t process, struct perf_session *session) { @@ -2094,21 +2095,22 @@ int perf_event__synthesize_attr(struct perf_event_attr *attr, u16 ids, u64 *id, ev->attr.header.type = PERF_RECORD_HEADER_ATTR; ev->attr.header.size = size; - err = process(ev, NULL, session); + err = process(ops, ev, NULL, session); free(ev); return err; } -int perf_session__synthesize_attrs(struct perf_session *session, +int perf_event__synthesize_attrs(struct perf_event_ops *ops, + struct perf_session *session, perf_event__handler_t process) { struct perf_evsel *attr; int err = 0; list_for_each_entry(attr, &session->evlist->entries, node) { - err = perf_event__synthesize_attr(&attr->attr, attr->ids, + err = perf_event__synthesize_attr(ops, &attr->attr, attr->ids, attr->id, process, session); if (err) { pr_debug("failed to create perf header attribute\n"); @@ -2156,7 +2158,8 @@ int perf_event__process_attr(union perf_event *event, return 0; } -int perf_event__synthesize_event_type(u64 event_id, char *name, +int perf_event__synthesize_event_type(struct perf_event_ops *ops, + u64 event_id, char *name, perf_event__handler_t process, struct perf_session *session) { @@ -2176,12 +2179,13 @@ int perf_event__synthesize_event_type(u64 event_id, char *name, ev.event_type.header.size = sizeof(ev.event_type) - (sizeof(ev.event_type.event_type.name) - size); - err = process(&ev, NULL, session); + err = process(ops, &ev, NULL, session); return err; } -int perf_event__synthesize_event_types(perf_event__handler_t process, +int perf_event__synthesize_event_types(struct perf_event_ops *ops, + perf_event__handler_t process, struct perf_session *session) { struct perf_trace_event_type *type; @@ -2190,7 +2194,7 @@ int perf_event__synthesize_event_types(perf_event__handler_t process, for (i = 0; i < event_count; i++) { type = &events[i]; - err = perf_event__synthesize_event_type(type->event_id, + err = perf_event__synthesize_event_type(ops, type->event_id, type->name, process, session); if (err) { @@ -2202,7 +2206,8 @@ int perf_event__synthesize_event_types(perf_event__handler_t process, return err; } -int perf_event__process_event_type(union perf_event *event, +int perf_event__process_event_type(struct perf_event_ops *ops __unused, + union perf_event *event, struct perf_session *session __unused) { if (perf_header__push_event(event->event_type.event_type.event_id, @@ -2212,7 +2217,8 @@ int perf_event__process_event_type(union perf_event *event, return 0; } -int perf_event__synthesize_tracing_data(int fd, struct perf_evlist *evlist, +int perf_event__synthesize_tracing_data(struct perf_event_ops *ops, int fd, + struct perf_evlist *evlist, perf_event__handler_t process, struct perf_session *session __unused) { @@ -2245,7 +2251,7 @@ int perf_event__synthesize_tracing_data(int fd, struct perf_evlist *evlist, ev.tracing_data.header.size = sizeof(ev.tracing_data); ev.tracing_data.size = aligned_size; - process(&ev, NULL, session); + process(ops, &ev, NULL, session); /* * The put function will copy all the tracing data @@ -2287,7 +2293,8 @@ int perf_event__process_tracing_data(union perf_event *event, return size_read + padding; } -int perf_event__synthesize_build_id(struct dso *pos, u16 misc, +int perf_event__synthesize_build_id(struct perf_event_ops *ops, + struct dso *pos, u16 misc, perf_event__handler_t process, struct machine *machine, struct perf_session *session) @@ -2310,12 +2317,13 @@ int perf_event__synthesize_build_id(struct dso *pos, u16 misc, ev.build_id.header.size = sizeof(ev.build_id) + len; memcpy(&ev.build_id.filename, pos->long_name, pos->long_name_len); - err = process(&ev, NULL, session); + err = process(ops, &ev, NULL, session); return err; } -int perf_event__process_build_id(union perf_event *event, +int perf_event__process_build_id(struct perf_event_ops *ops __used, + union perf_event *event, struct perf_session *session) { __event_process_build_id(&event->build_id, diff --git a/tools/perf/util/header.h b/tools/perf/util/header.h index 0a88982bc392..54dae5f09556 100644 --- a/tools/perf/util/header.h +++ b/tools/perf/util/header.h @@ -68,6 +68,7 @@ struct perf_header { }; struct perf_evlist; +struct perf_session; int perf_session__read_header(struct perf_session *session, int fd); int perf_session__write_header(struct perf_session *session, @@ -96,32 +97,40 @@ int build_id_cache__add_s(const char *sbuild_id, const char *debugdir, const char *name, bool is_kallsyms); int build_id_cache__remove_s(const char *sbuild_id, const char *debugdir); -int perf_event__synthesize_attr(struct perf_event_attr *attr, u16 ids, u64 *id, +int perf_event__synthesize_attr(struct perf_event_ops *ops, + struct perf_event_attr *attr, u16 ids, u64 *id, perf_event__handler_t process, struct perf_session *session); -int perf_session__synthesize_attrs(struct perf_session *session, - perf_event__handler_t process); +int perf_event__synthesize_attrs(struct perf_event_ops *ops, + struct perf_session *session, + perf_event__handler_t process); int perf_event__process_attr(union perf_event *event, struct perf_evlist **pevlist); -int perf_event__synthesize_event_type(u64 event_id, char *name, +int perf_event__synthesize_event_type(struct perf_event_ops *ops, + u64 event_id, char *name, perf_event__handler_t process, struct perf_session *session); -int perf_event__synthesize_event_types(perf_event__handler_t process, +int perf_event__synthesize_event_types(struct perf_event_ops *ops, + perf_event__handler_t process, struct perf_session *session); -int perf_event__process_event_type(union perf_event *event, +int perf_event__process_event_type(struct perf_event_ops *ops, + union perf_event *event, struct perf_session *session); -int perf_event__synthesize_tracing_data(int fd, struct perf_evlist *evlist, +int perf_event__synthesize_tracing_data(struct perf_event_ops *ops, + int fd, struct perf_evlist *evlist, perf_event__handler_t process, struct perf_session *session); int perf_event__process_tracing_data(union perf_event *event, struct perf_session *session); -int perf_event__synthesize_build_id(struct dso *pos, u16 misc, +int perf_event__synthesize_build_id(struct perf_event_ops *ops, + struct dso *pos, u16 misc, perf_event__handler_t process, struct machine *machine, struct perf_session *session); -int perf_event__process_build_id(union perf_event *event, +int perf_event__process_build_id(struct perf_event_ops *ops, + union perf_event *event, struct perf_session *session); /* diff --git a/tools/perf/util/session.c b/tools/perf/util/session.c index 734358b51ed1..a36023a66779 100644 --- a/tools/perf/util/session.c +++ b/tools/perf/util/session.c @@ -270,13 +270,21 @@ int perf_session__resolve_callchain(struct perf_session *self, struct perf_evsel return 0; } -static int process_event_synth_stub(union perf_event *event __used, +static int process_event_synth_stub(struct perf_event_ops *ops __used, + union perf_event *event __used, struct perf_session *session __used) { dump_printf(": unhandled!\n"); return 0; } +static int process_event_synth_tracing_data_stub(union perf_event *event __used, + struct perf_session *session __used) +{ + dump_printf(": unhandled!\n"); + return 0; +} + static int process_event_synth_attr_stub(union perf_event *event __used, struct perf_evlist **pevlist __used) { @@ -284,7 +292,8 @@ static int process_event_synth_attr_stub(union perf_event *event __used, return 0; } -static int process_event_sample_stub(union perf_event *event __used, +static int process_event_sample_stub(struct perf_event_ops *ops __used, + union perf_event *event __used, struct perf_sample *sample __used, struct perf_evsel *evsel __used, struct perf_session *session __used) @@ -293,7 +302,8 @@ static int process_event_sample_stub(union perf_event *event __used, return 0; } -static int process_event_stub(union perf_event *event __used, +static int process_event_stub(struct perf_event_ops *ops __used, + union perf_event *event __used, struct perf_sample *sample __used, struct perf_session *session __used) { @@ -301,17 +311,17 @@ static int process_event_stub(union perf_event *event __used, return 0; } -static int process_finished_round_stub(union perf_event *event __used, - struct perf_session *session __used, - struct perf_event_ops *ops __used) +static int process_finished_round_stub(struct perf_event_ops *ops __used, + union perf_event *event __used, + struct perf_session *session __used) { dump_printf(": unhandled!\n"); return 0; } -static int process_finished_round(union perf_event *event, - struct perf_session *session, - struct perf_event_ops *ops); +static int process_finished_round(struct perf_event_ops *ops, + union perf_event *event, + struct perf_session *session); static void perf_event_ops__fill_defaults(struct perf_event_ops *handler) { @@ -338,7 +348,7 @@ static void perf_event_ops__fill_defaults(struct perf_event_ops *handler) if (handler->event_type == NULL) handler->event_type = process_event_synth_stub; if (handler->tracing_data == NULL) - handler->tracing_data = process_event_synth_stub; + handler->tracing_data = process_event_synth_tracing_data_stub; if (handler->build_id == NULL) handler->build_id = process_event_synth_stub; if (handler->finished_round == NULL) { @@ -565,9 +575,9 @@ static void flush_sample_queue(struct perf_session *s, * Flush every events below timestamp 7 * etc... */ -static int process_finished_round(union perf_event *event __used, - struct perf_session *session, - struct perf_event_ops *ops) +static int process_finished_round(struct perf_event_ops *ops, + union perf_event *event __used, + struct perf_session *session) { flush_sample_queue(session, ops); session->ordered_samples.next_flush = session->ordered_samples.max_timestamp; @@ -759,23 +769,23 @@ static int perf_session_deliver_event(struct perf_session *session, ++session->hists.stats.nr_unknown_id; return -1; } - return ops->sample(event, sample, evsel, session); + return ops->sample(ops, event, sample, evsel, session); case PERF_RECORD_MMAP: - return ops->mmap(event, sample, session); + return ops->mmap(ops, event, sample, session); case PERF_RECORD_COMM: - return ops->comm(event, sample, session); + return ops->comm(ops, event, sample, session); case PERF_RECORD_FORK: - return ops->fork(event, sample, session); + return ops->fork(ops, event, sample, session); case PERF_RECORD_EXIT: - return ops->exit(event, sample, session); + return ops->exit(ops, event, sample, session); case PERF_RECORD_LOST: - return ops->lost(event, sample, session); + return ops->lost(ops, event, sample, session); case PERF_RECORD_READ: - return ops->read(event, sample, session); + return ops->read(ops, event, sample, session); case PERF_RECORD_THROTTLE: - return ops->throttle(event, sample, session); + return ops->throttle(ops, event, sample, session); case PERF_RECORD_UNTHROTTLE: - return ops->unthrottle(event, sample, session); + return ops->unthrottle(ops, event, sample, session); default: ++session->hists.stats.nr_unknown_events; return -1; @@ -813,15 +823,15 @@ static int perf_session__process_user_event(struct perf_session *session, union perf_session__update_sample_type(session); return err; case PERF_RECORD_HEADER_EVENT_TYPE: - return ops->event_type(event, session); + return ops->event_type(ops, event, session); case PERF_RECORD_HEADER_TRACING_DATA: /* setup for reading amidst mmap */ lseek(session->fd, file_offset, SEEK_SET); return ops->tracing_data(event, session); case PERF_RECORD_HEADER_BUILD_ID: - return ops->build_id(event, session); + return ops->build_id(ops, event, session); case PERF_RECORD_FINISHED_ROUND: - return ops->finished_round(event, session, ops); + return ops->finished_round(ops, event, session); default: return -EINVAL; } diff --git a/tools/perf/util/session.h b/tools/perf/util/session.h index d2f430367713..6de3d1368900 100644 --- a/tools/perf/util/session.h +++ b/tools/perf/util/session.h @@ -56,16 +56,18 @@ struct perf_session { struct perf_evsel; struct perf_event_ops; -typedef int (*event_sample)(union perf_event *event, struct perf_sample *sample, +typedef int (*event_sample)(struct perf_event_ops *ops, + union perf_event *event, struct perf_sample *sample, struct perf_evsel *evsel, struct perf_session *session); -typedef int (*event_op)(union perf_event *self, struct perf_sample *sample, +typedef int (*event_op)(struct perf_event_ops *ops, union perf_event *event, + struct perf_sample *sample, struct perf_session *session); typedef int (*event_synth_op)(union perf_event *self, struct perf_session *session); typedef int (*event_attr_op)(union perf_event *event, struct perf_evlist **pevlist); -typedef int (*event_op2)(union perf_event *self, struct perf_session *session, - struct perf_event_ops *ops); +typedef int (*event_op2)(struct perf_event_ops *ops, union perf_event *event, + struct perf_session *session); struct perf_event_ops { event_sample sample; @@ -78,10 +80,10 @@ struct perf_event_ops { throttle, unthrottle; event_attr_op attr; - event_synth_op event_type, - tracing_data, - build_id; - event_op2 finished_round; + event_synth_op tracing_data; + event_op2 event_type, + build_id, + finished_round; bool ordered_samples; bool ordering_requires_timestamps; }; @@ -142,10 +144,11 @@ struct machine *perf_session__findnew_machine(struct perf_session *self, pid_t p static inline void perf_session__process_machines(struct perf_session *self, + struct perf_event_ops *ops, machine__process_t process) { - process(&self->host_machine, self); - return machines__process(&self->machines, process, self); + process(&self->host_machine, ops); + return machines__process(&self->machines, process, ops); } size_t perf_session__fprintf_dsos(struct perf_session *self, FILE *fp); diff --git a/tools/perf/util/top.h b/tools/perf/util/top.h index 399650967958..44eda6fc6b33 100644 --- a/tools/perf/util/top.h +++ b/tools/perf/util/top.h @@ -2,14 +2,15 @@ #define __PERF_TOP_H 1 #include "types.h" +#include "session.h" #include "../perf.h" #include struct perf_evlist; struct perf_evsel; -struct perf_session; struct perf_top { + struct perf_event_ops ops; struct perf_evlist *evlist; /* * Symbols will be added here in perf_event__process_sample and will -- cgit v1.2.3 From 743eb868657bdb1b26c7b24077ca21c67c82c777 Mon Sep 17 00:00:00 2001 From: Arnaldo Carvalho de Melo Date: Mon, 28 Nov 2011 07:56:39 -0200 Subject: perf tools: Resolve machine earlier and pass it to perf_event_ops Reducing the exposure of perf_session further, so that we can use the classes in cases where no perf.data file is created. Cc: David Ahern Cc: Frederic Weisbecker Cc: Mike Galbraith Cc: Paul Mackerras Cc: Peter Zijlstra Cc: Stephane Eranian Link: http://lkml.kernel.org/n/tip-stua66dcscsezzrcdugvbmvd@git.kernel.org Signed-off-by: Arnaldo Carvalho de Melo --- tools/perf/builtin-annotate.c | 4 +- tools/perf/builtin-diff.c | 9 +- tools/perf/builtin-inject.c | 70 +++++----- tools/perf/builtin-kmem.c | 4 +- tools/perf/builtin-lock.c | 4 +- tools/perf/builtin-record.c | 38 +++--- tools/perf/builtin-report.c | 24 ++-- tools/perf/builtin-sched.c | 70 +++++----- tools/perf/builtin-script.c | 26 ++-- tools/perf/builtin-timechart.c | 8 +- tools/perf/builtin-top.c | 99 +++++++------- tools/perf/util/build-id.c | 16 +-- tools/perf/util/event.c | 151 +++++++++------------ tools/perf/util/event.h | 21 ++- tools/perf/util/header.c | 28 ++-- tools/perf/util/header.h | 16 +-- tools/perf/util/map.h | 10 ++ .../perf/util/scripting-engines/trace-event-perl.c | 4 +- .../util/scripting-engines/trace-event-python.c | 4 +- tools/perf/util/session.c | 91 +++++++------ tools/perf/util/session.h | 30 ++-- tools/perf/util/thread.h | 14 +- tools/perf/util/trace-event-scripting.c | 2 +- tools/perf/util/trace-event.h | 8 +- 24 files changed, 376 insertions(+), 375 deletions(-) diff --git a/tools/perf/builtin-annotate.c b/tools/perf/builtin-annotate.c index 483cb9466444..dff081a388bb 100644 --- a/tools/perf/builtin-annotate.c +++ b/tools/perf/builtin-annotate.c @@ -83,12 +83,12 @@ static int process_sample_event(struct perf_event_ops *ops, union perf_event *event, struct perf_sample *sample, struct perf_evsel *evsel, - struct perf_session *session) + struct machine *machine) { struct perf_annotate *ann = container_of(ops, struct perf_annotate, ops); struct addr_location al; - if (perf_event__preprocess_sample(event, session, &al, sample, + if (perf_event__preprocess_sample(event, machine, &al, sample, symbol__annotate_init) < 0) { pr_warning("problem processing %d event, skipping it.\n", event->header.type); diff --git a/tools/perf/builtin-diff.c b/tools/perf/builtin-diff.c index 9a0872f9e837..478b0aeb2a62 100644 --- a/tools/perf/builtin-diff.c +++ b/tools/perf/builtin-diff.c @@ -9,6 +9,7 @@ #include "util/debug.h" #include "util/event.h" #include "util/hist.h" +#include "util/evsel.h" #include "util/session.h" #include "util/sort.h" #include "util/symbol.h" @@ -34,11 +35,11 @@ static int diff__process_sample_event(struct perf_event_ops *ops __used, union perf_event *event, struct perf_sample *sample, struct perf_evsel *evsel __used, - struct perf_session *session) + struct machine *machine) { struct addr_location al; - if (perf_event__preprocess_sample(event, session, &al, sample, NULL) < 0) { + if (perf_event__preprocess_sample(event, machine, &al, sample, NULL) < 0) { pr_warning("problem processing %d event, skipping it.\n", event->header.type); return -1; @@ -47,12 +48,12 @@ static int diff__process_sample_event(struct perf_event_ops *ops __used, if (al.filtered || al.sym == NULL) return 0; - if (hists__add_entry(&session->hists, &al, sample->period)) { + if (hists__add_entry(&evsel->hists, &al, sample->period)) { pr_warning("problem incrementing symbol period, skipping event\n"); return -1; } - session->hists.stats.total_period += sample->period; + evsel->hists.stats.total_period += sample->period; return 0; } diff --git a/tools/perf/builtin-inject.c b/tools/perf/builtin-inject.c index 6ce6d80b59db..a5bcf81776fc 100644 --- a/tools/perf/builtin-inject.c +++ b/tools/perf/builtin-inject.c @@ -18,7 +18,7 @@ static bool inject_build_ids; static int perf_event__repipe_synth(struct perf_event_ops *ops __used, union perf_event *event, - struct perf_session *session __used) + struct machine *machine __used) { uint32_t size; void *buf = event; @@ -37,10 +37,23 @@ static int perf_event__repipe_synth(struct perf_event_ops *ops __used, return 0; } +static int perf_event__repipe_op2_synth(struct perf_event_ops *ops, + union perf_event *event, + struct perf_session *session __used) +{ + return perf_event__repipe_synth(ops, event, NULL); +} + +static int perf_event__repipe_event_type_synth(struct perf_event_ops *ops, + union perf_event *event) +{ + return perf_event__repipe_synth(ops, event, NULL); +} + static int perf_event__repipe_tracing_data_synth(union perf_event *event, - struct perf_session *session) + struct perf_session *session __used) { - return perf_event__repipe_synth(NULL, event, session); + return perf_event__repipe_synth(NULL, event, NULL); } static int perf_event__repipe_attr(union perf_event *event, @@ -52,29 +65,29 @@ static int perf_event__repipe_attr(union perf_event *event, static int perf_event__repipe(struct perf_event_ops *ops, union perf_event *event, struct perf_sample *sample __used, - struct perf_session *session) + struct machine *machine) { - return perf_event__repipe_synth(ops, event, session); + return perf_event__repipe_synth(ops, event, machine); } static int perf_event__repipe_sample(struct perf_event_ops *ops, union perf_event *event, struct perf_sample *sample __used, struct perf_evsel *evsel __used, - struct perf_session *session) + struct machine *machine) { - return perf_event__repipe_synth(ops, event, session); + return perf_event__repipe_synth(ops, event, machine); } static int perf_event__repipe_mmap(struct perf_event_ops *ops, union perf_event *event, struct perf_sample *sample, - struct perf_session *session) + struct machine *machine) { int err; - err = perf_event__process_mmap(ops, event, sample, session); - perf_event__repipe(ops, event, sample, session); + err = perf_event__process_mmap(ops, event, sample, machine); + perf_event__repipe(ops, event, sample, machine); return err; } @@ -82,12 +95,12 @@ static int perf_event__repipe_mmap(struct perf_event_ops *ops, static int perf_event__repipe_task(struct perf_event_ops *ops, union perf_event *event, struct perf_sample *sample, - struct perf_session *session) + struct machine *machine) { int err; - err = perf_event__process_task(ops, event, sample, session); - perf_event__repipe(ops, event, sample, session); + err = perf_event__process_task(ops, event, sample, machine); + perf_event__repipe(ops, event, sample, machine); return err; } @@ -97,7 +110,7 @@ static int perf_event__repipe_tracing_data(union perf_event *event, { int err; - perf_event__repipe_synth(NULL, event, session); + perf_event__repipe_synth(NULL, event, NULL); err = perf_event__process_tracing_data(event, session); return err; @@ -118,10 +131,9 @@ static int dso__read_build_id(struct dso *self) } static int dso__inject_build_id(struct dso *self, struct perf_event_ops *ops, - struct perf_session *session) + struct machine *machine) { u16 misc = PERF_RECORD_MISC_USER; - struct machine *machine; int err; if (dso__read_build_id(self) < 0) { @@ -129,17 +141,11 @@ static int dso__inject_build_id(struct dso *self, struct perf_event_ops *ops, return -1; } - machine = perf_session__find_host_machine(session); - if (machine == NULL) { - pr_err("Can't find machine for session\n"); - return -1; - } - if (self->kernel) misc = PERF_RECORD_MISC_KERNEL; err = perf_event__synthesize_build_id(ops, self, misc, perf_event__repipe, - machine, session); + machine); if (err) { pr_err("Can't synthesize build_id event for %s\n", self->long_name); return -1; @@ -152,7 +158,7 @@ static int perf_event__inject_buildid(struct perf_event_ops *ops, union perf_event *event, struct perf_sample *sample, struct perf_evsel *evsel __used, - struct perf_session *session) + struct machine *machine) { struct addr_location al; struct thread *thread; @@ -160,21 +166,21 @@ static int perf_event__inject_buildid(struct perf_event_ops *ops, cpumode = event->header.misc & PERF_RECORD_MISC_CPUMODE_MASK; - thread = perf_session__findnew(session, event->ip.pid); + thread = machine__findnew_thread(machine, event->ip.pid); if (thread == NULL) { pr_err("problem processing %d event, skipping it.\n", event->header.type); goto repipe; } - thread__find_addr_map(thread, session, cpumode, MAP__FUNCTION, - event->ip.pid, event->ip.ip, &al); + thread__find_addr_map(thread, machine, cpumode, MAP__FUNCTION, + event->ip.ip, &al); if (al.map != NULL) { if (!al.map->dso->hit) { al.map->dso->hit = 1; if (map__load(al.map, NULL) >= 0) { - dso__inject_build_id(al.map->dso, ops, session); + dso__inject_build_id(al.map->dso, ops, machine); /* * If this fails, too bad, let the other side * account this as unresolved. @@ -187,7 +193,7 @@ static int perf_event__inject_buildid(struct perf_event_ops *ops, } repipe: - perf_event__repipe(ops, event, sample, session); + perf_event__repipe(ops, event, sample, machine); return 0; } @@ -198,13 +204,13 @@ struct perf_event_ops inject_ops = { .fork = perf_event__repipe, .exit = perf_event__repipe, .lost = perf_event__repipe, - .read = perf_event__repipe, + .read = perf_event__repipe_sample, .throttle = perf_event__repipe, .unthrottle = perf_event__repipe, .attr = perf_event__repipe_attr, - .event_type = perf_event__repipe_synth, + .event_type = perf_event__repipe_event_type_synth, .tracing_data = perf_event__repipe_tracing_data_synth, - .build_id = perf_event__repipe_synth, + .build_id = perf_event__repipe_op2_synth, }; extern volatile int session_done; diff --git a/tools/perf/builtin-kmem.c b/tools/perf/builtin-kmem.c index 5d01218e50e0..27b2a15dc7b2 100644 --- a/tools/perf/builtin-kmem.c +++ b/tools/perf/builtin-kmem.c @@ -307,9 +307,9 @@ static int process_sample_event(struct perf_event_ops *ops __used, union perf_event *event, struct perf_sample *sample, struct perf_evsel *evsel __used, - struct perf_session *session) + struct machine *machine) { - struct thread *thread = perf_session__findnew(session, event->ip.pid); + struct thread *thread = machine__findnew_thread(machine, event->ip.pid); if (thread == NULL) { pr_debug("problem processing %d event, skipping it.\n", diff --git a/tools/perf/builtin-lock.c b/tools/perf/builtin-lock.c index f06b0a44c7cb..99b032adb83e 100644 --- a/tools/perf/builtin-lock.c +++ b/tools/perf/builtin-lock.c @@ -849,9 +849,9 @@ static int process_sample_event(struct perf_event_ops *ops __used, union perf_event *event, struct perf_sample *sample, struct perf_evsel *evsel __used, - struct perf_session *s) + struct machine *machine) { - struct thread *thread = perf_session__findnew(s, sample->tid); + struct thread *thread = machine__findnew_thread(machine, sample->tid); if (thread == NULL) { pr_debug("problem processing %d event, skipping it.\n", diff --git a/tools/perf/builtin-record.c b/tools/perf/builtin-record.c index 4642d38b8d19..0af598a1059f 100644 --- a/tools/perf/builtin-record.c +++ b/tools/perf/builtin-record.c @@ -79,7 +79,7 @@ static void write_output(struct perf_record *rec, void *buf, size_t size) static int process_synthesized_event(struct perf_event_ops *ops, union perf_event *event, struct perf_sample *sample __used, - struct perf_session *self __used) + struct machine *machine __used) { struct perf_record *rec = container_of(ops, struct perf_record, ops); write_output(rec, event, event->header.size); @@ -320,8 +320,6 @@ static void perf_event__synthesize_guest_os(struct machine *machine, void *data) { int err; struct perf_event_ops *ops = data; - struct perf_record *rec = container_of(ops, struct perf_record, ops); - struct perf_session *psession = rec->session; if (machine__is_host(machine)) return; @@ -335,7 +333,7 @@ static void perf_event__synthesize_guest_os(struct machine *machine, void *data) *in module instead of in guest kernel. */ err = perf_event__synthesize_modules(ops, process_synthesized_event, - psession, machine); + machine); if (err < 0) pr_err("Couldn't record guest kernel [%d]'s reference" " relocation symbol.\n", machine->pid); @@ -345,11 +343,10 @@ static void perf_event__synthesize_guest_os(struct machine *machine, void *data) * have no _text sometimes. */ err = perf_event__synthesize_kernel_mmap(ops, process_synthesized_event, - psession, machine, "_text"); + machine, "_text"); if (err < 0) err = perf_event__synthesize_kernel_mmap(ops, process_synthesized_event, - psession, machine, - "_stext"); + machine, "_stext"); if (err < 0) pr_err("Couldn't record guest kernel [%d]'s reference" " relocation symbol.\n", machine->pid); @@ -497,6 +494,12 @@ static int __cmd_record(struct perf_record *rec, int argc, const char **argv) rec->post_processing_offset = lseek(output, 0, SEEK_CUR); + machine = perf_session__find_host_machine(session); + if (!machine) { + pr_err("Couldn't find native kernel information.\n"); + return -1; + } + if (opts->pipe_output) { err = perf_event__synthesize_attrs(ops, session, process_synthesized_event); @@ -506,7 +509,7 @@ static int __cmd_record(struct perf_record *rec, int argc, const char **argv) } err = perf_event__synthesize_event_types(ops, process_synthesized_event, - session); + machine); if (err < 0) { pr_err("Couldn't synthesize event_types.\n"); return err; @@ -522,8 +525,7 @@ static int __cmd_record(struct perf_record *rec, int argc, const char **argv) * propagate errors that now are calling die() */ err = perf_event__synthesize_tracing_data(ops, output, evsel_list, - process_synthesized_event, - session); + process_synthesized_event); if (err <= 0) { pr_err("Couldn't record tracing data.\n"); return err; @@ -532,24 +534,18 @@ static int __cmd_record(struct perf_record *rec, int argc, const char **argv) } } - machine = perf_session__find_host_machine(session); - if (!machine) { - pr_err("Couldn't find native kernel information.\n"); - return -1; - } - err = perf_event__synthesize_kernel_mmap(ops, process_synthesized_event, - session, machine, "_text"); + machine, "_text"); if (err < 0) err = perf_event__synthesize_kernel_mmap(ops, process_synthesized_event, - session, machine, "_stext"); + machine, "_stext"); if (err < 0) pr_err("Couldn't record kernel reference relocation symbol\n" "Symbol resolution may be skewed if relocation was used (e.g. kexec).\n" "Check /proc/kallsyms permission or run as root.\n"); err = perf_event__synthesize_modules(ops, process_synthesized_event, - session, machine); + machine); if (err < 0) pr_err("Couldn't record kernel module information.\n" "Symbol resolution may be skewed if relocation was used (e.g. kexec).\n" @@ -562,10 +558,10 @@ static int __cmd_record(struct perf_record *rec, int argc, const char **argv) if (!opts->system_wide) perf_event__synthesize_thread_map(ops, evsel_list->threads, process_synthesized_event, - session); + machine); else perf_event__synthesize_threads(ops, process_synthesized_event, - session); + machine); if (rec->realtime_prio) { struct sched_param param; diff --git a/tools/perf/builtin-report.c b/tools/perf/builtin-report.c index 8795520f6e1d..ea64fbbdff43 100644 --- a/tools/perf/builtin-report.c +++ b/tools/perf/builtin-report.c @@ -52,18 +52,18 @@ struct perf_report { DECLARE_BITMAP(cpu_bitmap, MAX_NR_CPUS); }; -static int perf_session__add_hist_entry(struct perf_session *session, - struct addr_location *al, - struct perf_sample *sample, - struct perf_evsel *evsel) +static int perf_evsel__add_hist_entry(struct perf_evsel *evsel, + struct addr_location *al, + struct perf_sample *sample, + struct machine *machine) { struct symbol *parent = NULL; int err = 0; struct hist_entry *he; if ((sort__has_parent || symbol_conf.use_callchain) && sample->callchain) { - err = perf_session__resolve_callchain(session, evsel, al->thread, - sample->callchain, &parent); + err = machine__resolve_callchain(machine, evsel, al->thread, + sample->callchain, &parent); if (err) return err; } @@ -107,12 +107,12 @@ static int process_sample_event(struct perf_event_ops *ops, union perf_event *event, struct perf_sample *sample, struct perf_evsel *evsel, - struct perf_session *session) + struct machine *machine) { struct perf_report *rep = container_of(ops, struct perf_report, ops); struct addr_location al; - if (perf_event__preprocess_sample(event, session, &al, sample, + if (perf_event__preprocess_sample(event, machine, &al, sample, rep->annotate_init) < 0) { fprintf(stderr, "problem processing %d event, skipping it.\n", event->header.type); @@ -128,7 +128,7 @@ static int process_sample_event(struct perf_event_ops *ops, if (al.map != NULL) al.map->dso->hit = 1; - if (perf_session__add_hist_entry(session, &al, sample, evsel)) { + if (perf_evsel__add_hist_entry(evsel, &al, sample, machine)) { pr_debug("problem incrementing symbol period, skipping event\n"); return -1; } @@ -139,11 +139,11 @@ static int process_sample_event(struct perf_event_ops *ops, static int process_read_event(struct perf_event_ops *ops, union perf_event *event, struct perf_sample *sample __used, - struct perf_session *session) + struct perf_evsel *evsel, + struct machine *machine __used) { struct perf_report *rep = container_of(ops, struct perf_report, ops); - struct perf_evsel *evsel = perf_evlist__id2evsel(session->evlist, - event->read.id); + if (rep->show_threads) { const char *name = evsel ? event_name(evsel) : "unknown"; perf_read_values_add_value(&rep->show_threads_values, diff --git a/tools/perf/builtin-sched.c b/tools/perf/builtin-sched.c index b11d6283fedf..6a771f822e5d 100644 --- a/tools/perf/builtin-sched.c +++ b/tools/perf/builtin-sched.c @@ -724,21 +724,21 @@ struct trace_migrate_task_event { struct trace_sched_handler { void (*switch_event)(struct trace_switch_event *, - struct perf_session *, + struct machine *, struct event *, int cpu, u64 timestamp, struct thread *thread); void (*runtime_event)(struct trace_runtime_event *, - struct perf_session *, + struct machine *, struct event *, int cpu, u64 timestamp, struct thread *thread); void (*wakeup_event)(struct trace_wakeup_event *, - struct perf_session *, + struct machine *, struct event *, int cpu, u64 timestamp, @@ -751,7 +751,7 @@ struct trace_sched_handler { struct thread *thread); void (*migrate_task_event)(struct trace_migrate_task_event *, - struct perf_session *session, + struct machine *machine, struct event *, int cpu, u64 timestamp, @@ -761,7 +761,7 @@ struct trace_sched_handler { static void replay_wakeup_event(struct trace_wakeup_event *wakeup_event, - struct perf_session *session __used, + struct machine *machine __used, struct event *event, int cpu __used, u64 timestamp __used, @@ -788,7 +788,7 @@ static u64 cpu_last_switched[MAX_CPUS]; static void replay_switch_event(struct trace_switch_event *switch_event, - struct perf_session *session __used, + struct machine *machine __used, struct event *event, int cpu, u64 timestamp, @@ -1022,7 +1022,7 @@ add_sched_in_event(struct work_atoms *atoms, u64 timestamp) static void latency_switch_event(struct trace_switch_event *switch_event, - struct perf_session *session, + struct machine *machine, struct event *event __used, int cpu, u64 timestamp, @@ -1046,8 +1046,8 @@ latency_switch_event(struct trace_switch_event *switch_event, die("hm, delta: %" PRIu64 " < 0 ?\n", delta); - sched_out = perf_session__findnew(session, switch_event->prev_pid); - sched_in = perf_session__findnew(session, switch_event->next_pid); + sched_out = machine__findnew_thread(machine, switch_event->prev_pid); + sched_in = machine__findnew_thread(machine, switch_event->next_pid); out_events = thread_atoms_search(&atom_root, sched_out, &cmp_pid); if (!out_events) { @@ -1075,13 +1075,13 @@ latency_switch_event(struct trace_switch_event *switch_event, static void latency_runtime_event(struct trace_runtime_event *runtime_event, - struct perf_session *session, + struct machine *machine, struct event *event __used, int cpu, u64 timestamp, struct thread *this_thread __used) { - struct thread *thread = perf_session__findnew(session, runtime_event->pid); + struct thread *thread = machine__findnew_thread(machine, runtime_event->pid); struct work_atoms *atoms = thread_atoms_search(&atom_root, thread, &cmp_pid); BUG_ON(cpu >= MAX_CPUS || cpu < 0); @@ -1098,7 +1098,7 @@ latency_runtime_event(struct trace_runtime_event *runtime_event, static void latency_wakeup_event(struct trace_wakeup_event *wakeup_event, - struct perf_session *session, + struct machine *machine, struct event *__event __used, int cpu __used, u64 timestamp, @@ -1112,7 +1112,7 @@ latency_wakeup_event(struct trace_wakeup_event *wakeup_event, if (!wakeup_event->success) return; - wakee = perf_session__findnew(session, wakeup_event->pid); + wakee = machine__findnew_thread(machine, wakeup_event->pid); atoms = thread_atoms_search(&atom_root, wakee, &cmp_pid); if (!atoms) { thread_atoms_insert(wakee); @@ -1146,7 +1146,7 @@ latency_wakeup_event(struct trace_wakeup_event *wakeup_event, static void latency_migrate_task_event(struct trace_migrate_task_event *migrate_task_event, - struct perf_session *session, + struct machine *machine, struct event *__event __used, int cpu __used, u64 timestamp, @@ -1162,7 +1162,7 @@ latency_migrate_task_event(struct trace_migrate_task_event *migrate_task_event, if (profile_cpu == -1) return; - migrant = perf_session__findnew(session, migrate_task_event->pid); + migrant = machine__findnew_thread(machine, migrate_task_event->pid); atoms = thread_atoms_search(&atom_root, migrant, &cmp_pid); if (!atoms) { thread_atoms_insert(migrant); @@ -1357,7 +1357,7 @@ static void sort_lat(void) static struct trace_sched_handler *trace_handler; static void -process_sched_wakeup_event(void *data, struct perf_session *session, +process_sched_wakeup_event(void *data, struct machine *machine, struct event *event, int cpu __used, u64 timestamp __used, @@ -1374,7 +1374,7 @@ process_sched_wakeup_event(void *data, struct perf_session *session, FILL_FIELD(wakeup_event, cpu, event, data); if (trace_handler->wakeup_event) - trace_handler->wakeup_event(&wakeup_event, session, event, + trace_handler->wakeup_event(&wakeup_event, machine, event, cpu, timestamp, thread); } @@ -1393,7 +1393,7 @@ static char next_shortname2 = '0'; static void map_switch_event(struct trace_switch_event *switch_event, - struct perf_session *session, + struct machine *machine, struct event *event __used, int this_cpu, u64 timestamp, @@ -1421,8 +1421,8 @@ map_switch_event(struct trace_switch_event *switch_event, die("hm, delta: %" PRIu64 " < 0 ?\n", delta); - sched_out = perf_session__findnew(session, switch_event->prev_pid); - sched_in = perf_session__findnew(session, switch_event->next_pid); + sched_out = machine__findnew_thread(machine, switch_event->prev_pid); + sched_in = machine__findnew_thread(machine, switch_event->next_pid); curr_thread[this_cpu] = sched_in; @@ -1472,7 +1472,7 @@ map_switch_event(struct trace_switch_event *switch_event, static void -process_sched_switch_event(void *data, struct perf_session *session, +process_sched_switch_event(void *data, struct machine *machine, struct event *event, int this_cpu, u64 timestamp __used, @@ -1499,14 +1499,14 @@ process_sched_switch_event(void *data, struct perf_session *session, nr_context_switch_bugs++; } if (trace_handler->switch_event) - trace_handler->switch_event(&switch_event, session, event, + trace_handler->switch_event(&switch_event, machine, event, this_cpu, timestamp, thread); curr_pid[this_cpu] = switch_event.next_pid; } static void -process_sched_runtime_event(void *data, struct perf_session *session, +process_sched_runtime_event(void *data, struct machine *machine, struct event *event, int cpu __used, u64 timestamp __used, @@ -1520,7 +1520,7 @@ process_sched_runtime_event(void *data, struct perf_session *session, FILL_FIELD(runtime_event, vruntime, event, data); if (trace_handler->runtime_event) - trace_handler->runtime_event(&runtime_event, session, event, cpu, timestamp, thread); + trace_handler->runtime_event(&runtime_event, machine, event, cpu, timestamp, thread); } static void @@ -1555,7 +1555,7 @@ process_sched_exit_event(struct event *event, } static void -process_sched_migrate_task_event(void *data, struct perf_session *session, +process_sched_migrate_task_event(void *data, struct machine *machine, struct event *event, int cpu __used, u64 timestamp __used, @@ -1571,12 +1571,12 @@ process_sched_migrate_task_event(void *data, struct perf_session *session, FILL_FIELD(migrate_task_event, cpu, event, data); if (trace_handler->migrate_task_event) - trace_handler->migrate_task_event(&migrate_task_event, session, + trace_handler->migrate_task_event(&migrate_task_event, machine, event, cpu, timestamp, thread); } static void process_raw_event(union perf_event *raw_event __used, - struct perf_session *session, void *data, int cpu, + struct machine *machine, void *data, int cpu, u64 timestamp, struct thread *thread) { struct event *event; @@ -1587,33 +1587,33 @@ static void process_raw_event(union perf_event *raw_event __used, event = trace_find_event(type); if (!strcmp(event->name, "sched_switch")) - process_sched_switch_event(data, session, event, cpu, timestamp, thread); + process_sched_switch_event(data, machine, event, cpu, timestamp, thread); if (!strcmp(event->name, "sched_stat_runtime")) - process_sched_runtime_event(data, session, event, cpu, timestamp, thread); + process_sched_runtime_event(data, machine, event, cpu, timestamp, thread); if (!strcmp(event->name, "sched_wakeup")) - process_sched_wakeup_event(data, session, event, cpu, timestamp, thread); + process_sched_wakeup_event(data, machine, event, cpu, timestamp, thread); if (!strcmp(event->name, "sched_wakeup_new")) - process_sched_wakeup_event(data, session, event, cpu, timestamp, thread); + process_sched_wakeup_event(data, machine, event, cpu, timestamp, thread); if (!strcmp(event->name, "sched_process_fork")) process_sched_fork_event(data, event, cpu, timestamp, thread); if (!strcmp(event->name, "sched_process_exit")) process_sched_exit_event(event, cpu, timestamp, thread); if (!strcmp(event->name, "sched_migrate_task")) - process_sched_migrate_task_event(data, session, event, cpu, timestamp, thread); + process_sched_migrate_task_event(data, machine, event, cpu, timestamp, thread); } static int process_sample_event(struct perf_event_ops *ops __used, union perf_event *event, struct perf_sample *sample, struct perf_evsel *evsel, - struct perf_session *session) + struct machine *machine) { struct thread *thread; if (!(evsel->attr.sample_type & PERF_SAMPLE_RAW)) return 0; - thread = perf_session__findnew(session, sample->pid); + thread = machine__findnew_thread(machine, sample->pid); if (thread == NULL) { pr_debug("problem processing %d event, skipping it.\n", event->header.type); @@ -1625,7 +1625,7 @@ static int process_sample_event(struct perf_event_ops *ops __used, if (profile_cpu != -1 && profile_cpu != (int)sample->cpu) return 0; - process_raw_event(event, session, sample->raw_data, sample->cpu, + process_raw_event(event, machine, sample->raw_data, sample->cpu, sample->time, thread); return 0; diff --git a/tools/perf/builtin-script.c b/tools/perf/builtin-script.c index 3b7820612ebf..31a8d14e5fb7 100644 --- a/tools/perf/builtin-script.c +++ b/tools/perf/builtin-script.c @@ -315,7 +315,7 @@ static bool sample_addr_correlates_sym(struct perf_event_attr *attr) static void print_sample_addr(union perf_event *event, struct perf_sample *sample, - struct perf_session *session, + struct machine *machine, struct thread *thread, struct perf_event_attr *attr) { @@ -328,11 +328,11 @@ static void print_sample_addr(union perf_event *event, if (!sample_addr_correlates_sym(attr)) return; - thread__find_addr_map(thread, session, cpumode, MAP__FUNCTION, - event->ip.pid, sample->addr, &al); + thread__find_addr_map(thread, machine, cpumode, MAP__FUNCTION, + sample->addr, &al); if (!al.map) - thread__find_addr_map(thread, session, cpumode, MAP__VARIABLE, - event->ip.pid, sample->addr, &al); + thread__find_addr_map(thread, machine, cpumode, MAP__VARIABLE, + sample->addr, &al); al.cpu = sample->cpu; al.sym = NULL; @@ -362,7 +362,7 @@ static void print_sample_addr(union perf_event *event, static void process_event(union perf_event *event __unused, struct perf_sample *sample, struct perf_evsel *evsel, - struct perf_session *session, + struct machine *machine, struct thread *thread) { struct perf_event_attr *attr = &evsel->attr; @@ -377,15 +377,15 @@ static void process_event(union perf_event *event __unused, sample->raw_size); if (PRINT_FIELD(ADDR)) - print_sample_addr(event, sample, session, thread, attr); + print_sample_addr(event, sample, machine, thread, attr); if (PRINT_FIELD(IP)) { if (!symbol_conf.use_callchain) printf(" "); else printf("\n"); - perf_session__print_ip(event, evsel, sample, session, - PRINT_FIELD(SYM), PRINT_FIELD(DSO)); + perf_event__print_ip(event, sample, machine, evsel, + PRINT_FIELD(SYM), PRINT_FIELD(DSO)); } printf("\n"); @@ -438,9 +438,9 @@ static int process_sample_event(struct perf_event_ops *ops __used, union perf_event *event, struct perf_sample *sample, struct perf_evsel *evsel, - struct perf_session *session) + struct machine *machine) { - struct thread *thread = perf_session__findnew(session, event->ip.pid); + struct thread *thread = machine__findnew_thread(machine, event->ip.pid); if (thread == NULL) { pr_debug("problem processing %d event, skipping it.\n", @@ -462,9 +462,9 @@ static int process_sample_event(struct perf_event_ops *ops __used, if (cpu_list && !test_bit(sample->cpu, cpu_bitmap)) return 0; - scripting_ops->process_event(event, sample, evsel, session, thread); + scripting_ops->process_event(event, sample, evsel, machine, thread); - session->hists.stats.total_period += sample->period; + evsel->hists.stats.total_period += sample->period; return 0; } diff --git a/tools/perf/builtin-timechart.c b/tools/perf/builtin-timechart.c index 62298a0d7dc9..8e6539625bc1 100644 --- a/tools/perf/builtin-timechart.c +++ b/tools/perf/builtin-timechart.c @@ -277,7 +277,7 @@ static u64 cpus_pstate_state[MAX_CPUS]; static int process_comm_event(struct perf_event_ops *ops __used, union perf_event *event, struct perf_sample *sample __used, - struct perf_session *session __used) + struct machine *machine __used) { pid_set_comm(event->comm.tid, event->comm.comm); return 0; @@ -286,7 +286,7 @@ static int process_comm_event(struct perf_event_ops *ops __used, static int process_fork_event(struct perf_event_ops *ops __used, union perf_event *event, struct perf_sample *sample __used, - struct perf_session *session __used) + struct machine *machine __used) { pid_fork(event->fork.pid, event->fork.ppid, event->fork.time); return 0; @@ -295,7 +295,7 @@ static int process_fork_event(struct perf_event_ops *ops __used, static int process_exit_event(struct perf_event_ops *ops __used, union perf_event *event, struct perf_sample *sample __used, - struct perf_session *session __used) + struct machine *machine __used) { pid_exit(event->fork.pid, event->fork.time); return 0; @@ -494,7 +494,7 @@ static int process_sample_event(struct perf_event_ops *ops __used, union perf_event *event __used, struct perf_sample *sample, struct perf_evsel *evsel, - struct perf_session *session __used) + struct machine *machine __used) { struct trace_entry *te; diff --git a/tools/perf/builtin-top.c b/tools/perf/builtin-top.c index e8e3320602bd..31d497368ccf 100644 --- a/tools/perf/builtin-top.c +++ b/tools/perf/builtin-top.c @@ -258,11 +258,9 @@ out_unlock: static const char CONSOLE_CLEAR[] = ""; -static struct hist_entry * - perf_session__add_hist_entry(struct perf_session *session, - struct addr_location *al, - struct perf_sample *sample, - struct perf_evsel *evsel) +static struct hist_entry *perf_evsel__add_hist_entry(struct perf_evsel *evsel, + struct addr_location *al, + struct perf_sample *sample) { struct hist_entry *he; @@ -270,7 +268,7 @@ static struct hist_entry * if (he == NULL) return NULL; - session->hists.stats.total_period += sample->period; + evsel->hists.stats.total_period += sample->period; hists__inc_nr_events(&evsel->hists, PERF_RECORD_SAMPLE); return he; } @@ -675,44 +673,12 @@ static int symbol_filter(struct map *map __used, struct symbol *sym) static void perf_event__process_sample(const union perf_event *event, struct perf_evsel *evsel, struct perf_sample *sample, - struct perf_session *session) + struct machine *machine) { struct symbol *parent = NULL; u64 ip = event->ip.ip; struct addr_location al; - struct machine *machine; int err; - u8 origin = event->header.misc & PERF_RECORD_MISC_CPUMODE_MASK; - - ++top.samples; - - switch (origin) { - case PERF_RECORD_MISC_USER: - ++top.us_samples; - if (top.hide_user_symbols) - return; - machine = perf_session__find_host_machine(session); - break; - case PERF_RECORD_MISC_KERNEL: - ++top.kernel_samples; - if (top.hide_kernel_symbols) - return; - machine = perf_session__find_host_machine(session); - break; - case PERF_RECORD_MISC_GUEST_KERNEL: - ++top.guest_kernel_samples; - machine = perf_session__find_machine(session, event->ip.pid); - break; - case PERF_RECORD_MISC_GUEST_USER: - ++top.guest_us_samples; - /* - * TODO: we don't process guest user from host side - * except simple counting. - */ - return; - default: - return; - } if (!machine && perf_guest) { pr_err("Can't find guest [%d]'s kernel information\n", @@ -723,7 +689,7 @@ static void perf_event__process_sample(const union perf_event *event, if (event->header.misc & PERF_RECORD_MISC_EXACT_IP) top.exact_samples++; - if (perf_event__preprocess_sample(event, session, &al, sample, + if (perf_event__preprocess_sample(event, machine, &al, sample, symbol_filter) < 0 || al.filtered) return; @@ -777,13 +743,13 @@ static void perf_event__process_sample(const union perf_event *event, if ((sort__has_parent || symbol_conf.use_callchain) && sample->callchain) { - err = perf_session__resolve_callchain(session, evsel, al.thread, - sample->callchain, &parent); + err = machine__resolve_callchain(machine, evsel, al.thread, + sample->callchain, &parent); if (err) return; } - he = perf_session__add_hist_entry(session, &al, sample, evsel); + he = perf_evsel__add_hist_entry(evsel, &al, sample); if (he == NULL) { pr_err("Problem incrementing symbol period, skipping event\n"); return; @@ -808,6 +774,8 @@ static void perf_session__mmap_read_idx(struct perf_session *self, int idx) struct perf_sample sample; struct perf_evsel *evsel; union perf_event *event; + struct machine *machine; + u8 origin; int ret; while ((event = perf_evlist__mmap_read(top.evlist, idx)) != NULL) { @@ -820,11 +788,45 @@ static void perf_session__mmap_read_idx(struct perf_session *self, int idx) evsel = perf_evlist__id2evsel(self->evlist, sample.id); assert(evsel != NULL); + origin = event->header.misc & PERF_RECORD_MISC_CPUMODE_MASK; + if (event->header.type == PERF_RECORD_SAMPLE) - perf_event__process_sample(event, evsel, &sample, self); + ++top.samples; + + switch (origin) { + case PERF_RECORD_MISC_USER: + ++top.us_samples; + if (top.hide_user_symbols) + continue; + machine = perf_session__find_host_machine(self); + break; + case PERF_RECORD_MISC_KERNEL: + ++top.kernel_samples; + if (top.hide_kernel_symbols) + continue; + machine = perf_session__find_host_machine(self); + break; + case PERF_RECORD_MISC_GUEST_KERNEL: + ++top.guest_kernel_samples; + machine = perf_session__find_machine(self, event->ip.pid); + break; + case PERF_RECORD_MISC_GUEST_USER: + ++top.guest_us_samples; + /* + * TODO: we don't process guest user from host side + * except simple counting. + */ + /* Fall thru */ + default: + continue; + } + + + if (event->header.type == PERF_RECORD_SAMPLE) + perf_event__process_sample(event, evsel, &sample, machine); else if (event->header.type < PERF_RECORD_MAX) { hists__inc_nr_events(&evsel->hists, event->header.type); - perf_event__process(&top.ops, event, &sample, self); + perf_event__process(&top.ops, event, &sample, machine); } else ++self->hists.stats.nr_unknown_events; } @@ -967,10 +969,11 @@ static int __cmd_top(void) if (top.target_tid != -1) perf_event__synthesize_thread_map(&top.ops, top.evlist->threads, - perf_event__process, top.session); + perf_event__process, + &top.session->host_machine); else - perf_event__synthesize_threads(&top.ops, perf_event__process, top.session); - + perf_event__synthesize_threads(&top.ops, perf_event__process, + &top.session->host_machine); start_counters(top.evlist); top.session->evlist = top.evlist; perf_session__update_sample_type(top.session); diff --git a/tools/perf/util/build-id.c b/tools/perf/util/build-id.c index 0e4de1865013..2f84c4802aca 100644 --- a/tools/perf/util/build-id.c +++ b/tools/perf/util/build-id.c @@ -19,11 +19,11 @@ static int build_id__mark_dso_hit(struct perf_event_ops *ops __used, union perf_event *event, struct perf_sample *sample __used, struct perf_evsel *evsel __used, - struct perf_session *session) + struct machine *machine) { struct addr_location al; u8 cpumode = event->header.misc & PERF_RECORD_MISC_CPUMODE_MASK; - struct thread *thread = perf_session__findnew(session, event->ip.pid); + struct thread *thread = machine__findnew_thread(machine, event->ip.pid); if (thread == NULL) { pr_err("problem processing %d event, skipping it.\n", @@ -31,8 +31,8 @@ static int build_id__mark_dso_hit(struct perf_event_ops *ops __used, return -1; } - thread__find_addr_map(thread, session, cpumode, MAP__FUNCTION, - event->ip.pid, event->ip.ip, &al); + thread__find_addr_map(thread, machine, cpumode, MAP__FUNCTION, + event->ip.ip, &al); if (al.map != NULL) al.map->dso->hit = 1; @@ -43,16 +43,16 @@ static int build_id__mark_dso_hit(struct perf_event_ops *ops __used, static int perf_event__exit_del_thread(struct perf_event_ops *ops __used, union perf_event *event, struct perf_sample *sample __used, - struct perf_session *session) + struct machine *machine) { - struct thread *thread = perf_session__findnew(session, event->fork.tid); + struct thread *thread = machine__findnew_thread(machine, event->fork.tid); dump_printf("(%d:%d):(%d:%d)\n", event->fork.pid, event->fork.tid, event->fork.ppid, event->fork.ptid); if (thread) { - rb_erase(&thread->rb_node, &session->host_machine.threads); - session->host_machine.last_match = NULL; + rb_erase(&thread->rb_node, &machine->threads); + machine->last_match = NULL; thread__delete(thread); } diff --git a/tools/perf/util/event.c b/tools/perf/util/event.c index 4800f38c7277..0cdc811c48e2 100644 --- a/tools/perf/util/event.c +++ b/tools/perf/util/event.c @@ -1,7 +1,6 @@ #include #include "event.h" #include "debug.h" -#include "session.h" #include "sort.h" #include "string.h" #include "strlist.h" @@ -47,7 +46,7 @@ static struct perf_sample synth_sample = { static pid_t perf_event__synthesize_comm(struct perf_event_ops *ops, union perf_event *event, pid_t pid, int full, perf_event__handler_t process, - struct perf_session *session) + struct machine *machine) { char filename[PATH_MAX]; char bf[BUFSIZ]; @@ -93,14 +92,14 @@ out_race: event->comm.header.type = PERF_RECORD_COMM; size = ALIGN(size, sizeof(u64)); - memset(event->comm.comm + size, 0, session->id_hdr_size); + memset(event->comm.comm + size, 0, machine->id_hdr_size); event->comm.header.size = (sizeof(event->comm) - (sizeof(event->comm.comm) - size) + - session->id_hdr_size); + machine->id_hdr_size); if (!full) { event->comm.tid = pid; - process(ops, event, &synth_sample, session); + process(ops, event, &synth_sample, machine); goto out; } @@ -118,7 +117,7 @@ out_race: event->comm.tid = pid; - process(ops, event, &synth_sample, session); + process(ops, event, &synth_sample, machine); } closedir(tasks); @@ -132,7 +131,7 @@ static int perf_event__synthesize_mmap_events(struct perf_event_ops *ops, union perf_event *event, pid_t pid, pid_t tgid, perf_event__handler_t process, - struct perf_session *session) + struct machine *machine) { char filename[PATH_MAX]; FILE *fp; @@ -195,12 +194,12 @@ static int perf_event__synthesize_mmap_events(struct perf_event_ops *ops, event->mmap.len -= event->mmap.start; event->mmap.header.size = (sizeof(event->mmap) - (sizeof(event->mmap.filename) - size)); - memset(event->mmap.filename + size, 0, session->id_hdr_size); - event->mmap.header.size += session->id_hdr_size; + memset(event->mmap.filename + size, 0, machine->id_hdr_size); + event->mmap.header.size += machine->id_hdr_size; event->mmap.pid = tgid; event->mmap.tid = pid; - process(ops, event, &synth_sample, session); + process(ops, event, &synth_sample, machine); } } @@ -210,13 +209,12 @@ static int perf_event__synthesize_mmap_events(struct perf_event_ops *ops, int perf_event__synthesize_modules(struct perf_event_ops *ops, perf_event__handler_t process, - struct perf_session *session, struct machine *machine) { struct rb_node *nd; struct map_groups *kmaps = &machine->kmaps; union perf_event *event = zalloc((sizeof(event->mmap) + - session->id_hdr_size)); + machine->id_hdr_size)); if (event == NULL) { pr_debug("Not enough memory synthesizing mmap event " "for kernel modules\n"); @@ -246,15 +244,15 @@ int perf_event__synthesize_modules(struct perf_event_ops *ops, event->mmap.header.type = PERF_RECORD_MMAP; event->mmap.header.size = (sizeof(event->mmap) - (sizeof(event->mmap.filename) - size)); - memset(event->mmap.filename + size, 0, session->id_hdr_size); - event->mmap.header.size += session->id_hdr_size; + memset(event->mmap.filename + size, 0, machine->id_hdr_size); + event->mmap.header.size += machine->id_hdr_size; event->mmap.start = pos->start; event->mmap.len = pos->end - pos->start; event->mmap.pid = machine->pid; memcpy(event->mmap.filename, pos->dso->long_name, pos->dso->long_name_len + 1); - process(ops, event, &synth_sample, session); + process(ops, event, &synth_sample, machine); } free(event); @@ -265,29 +263,29 @@ static int __event__synthesize_thread(union perf_event *comm_event, union perf_event *mmap_event, pid_t pid, perf_event__handler_t process, struct perf_event_ops *ops, - struct perf_session *session) + struct machine *machine) { - pid_t tgid = perf_event__synthesize_comm(ops, comm_event, pid, 1, process, - session); + pid_t tgid = perf_event__synthesize_comm(ops, comm_event, pid, 1, + process, machine); if (tgid == -1) return -1; return perf_event__synthesize_mmap_events(ops, mmap_event, pid, tgid, - process, session); + process, machine); } int perf_event__synthesize_thread_map(struct perf_event_ops *ops, struct thread_map *threads, perf_event__handler_t process, - struct perf_session *session) + struct machine *machine) { union perf_event *comm_event, *mmap_event; int err = -1, thread; - comm_event = malloc(sizeof(comm_event->comm) + session->id_hdr_size); + comm_event = malloc(sizeof(comm_event->comm) + machine->id_hdr_size); if (comm_event == NULL) goto out; - mmap_event = malloc(sizeof(mmap_event->mmap) + session->id_hdr_size); + mmap_event = malloc(sizeof(mmap_event->mmap) + machine->id_hdr_size); if (mmap_event == NULL) goto out_free_comm; @@ -295,7 +293,7 @@ int perf_event__synthesize_thread_map(struct perf_event_ops *ops, for (thread = 0; thread < threads->nr; ++thread) { if (__event__synthesize_thread(comm_event, mmap_event, threads->map[thread], - process, ops, session)) { + process, ops, machine)) { err = -1; break; } @@ -309,18 +307,18 @@ out: int perf_event__synthesize_threads(struct perf_event_ops *ops, perf_event__handler_t process, - struct perf_session *session) + struct machine *machine) { DIR *proc; struct dirent dirent, *next; union perf_event *comm_event, *mmap_event; int err = -1; - comm_event = malloc(sizeof(comm_event->comm) + session->id_hdr_size); + comm_event = malloc(sizeof(comm_event->comm) + machine->id_hdr_size); if (comm_event == NULL) goto out; - mmap_event = malloc(sizeof(mmap_event->mmap) + session->id_hdr_size); + mmap_event = malloc(sizeof(mmap_event->mmap) + machine->id_hdr_size); if (mmap_event == NULL) goto out_free_comm; @@ -336,7 +334,7 @@ int perf_event__synthesize_threads(struct perf_event_ops *ops, continue; __event__synthesize_thread(comm_event, mmap_event, pid, - process, ops, session); + process, ops, machine); } closedir(proc); @@ -373,7 +371,6 @@ static int find_symbol_cb(void *arg, const char *name, char type, int perf_event__synthesize_kernel_mmap(struct perf_event_ops *ops, perf_event__handler_t process, - struct perf_session *session, struct machine *machine, const char *symbol_name) { @@ -390,7 +387,7 @@ int perf_event__synthesize_kernel_mmap(struct perf_event_ops *ops, */ struct process_symbol_args args = { .name = symbol_name, }; union perf_event *event = zalloc((sizeof(event->mmap) + - session->id_hdr_size)); + machine->id_hdr_size)); if (event == NULL) { pr_debug("Not enough memory synthesizing mmap event " "for kernel modules\n"); @@ -424,13 +421,13 @@ int perf_event__synthesize_kernel_mmap(struct perf_event_ops *ops, size = ALIGN(size, sizeof(u64)); event->mmap.header.type = PERF_RECORD_MMAP; event->mmap.header.size = (sizeof(event->mmap) - - (sizeof(event->mmap.filename) - size) + session->id_hdr_size); + (sizeof(event->mmap.filename) - size) + machine->id_hdr_size); event->mmap.pgoff = args.start; event->mmap.start = map->start; event->mmap.len = map->end - event->mmap.start; event->mmap.pid = machine->pid; - err = process(ops, event, &synth_sample, session); + err = process(ops, event, &synth_sample, machine); free(event); return err; @@ -439,9 +436,9 @@ int perf_event__synthesize_kernel_mmap(struct perf_event_ops *ops, int perf_event__process_comm(struct perf_event_ops *ops __used, union perf_event *event, struct perf_sample *sample __used, - struct perf_session *session) + struct machine *machine) { - struct thread *thread = perf_session__findnew(session, event->comm.tid); + struct thread *thread = machine__findnew_thread(machine, event->comm.tid); dump_printf(": %s:%d\n", event->comm.comm, event->comm.tid); @@ -456,11 +453,10 @@ int perf_event__process_comm(struct perf_event_ops *ops __used, int perf_event__process_lost(struct perf_event_ops *ops __used, union perf_event *event, struct perf_sample *sample __used, - struct perf_session *session) + struct machine *machine __used) { dump_printf(": id:%" PRIu64 ": lost:%" PRIu64 "\n", event->lost.id, event->lost.lost); - session->hists.stats.total_lost += event->lost.lost; return 0; } @@ -479,20 +475,13 @@ static void perf_event__set_kernel_mmap_len(union perf_event *event, static int perf_event__process_kernel_mmap(struct perf_event_ops *ops __used, union perf_event *event, - struct perf_session *session) + struct machine *machine) { struct map *map; char kmmap_prefix[PATH_MAX]; - struct machine *machine; enum dso_kernel_type kernel_type; bool is_kernel_mmap; - machine = perf_session__findnew_machine(session, event->mmap.pid); - if (!machine) { - pr_err("Can't find id %d's machine\n", event->mmap.pid); - goto out_problem; - } - machine__mmap_name(machine, kmmap_prefix, sizeof(kmmap_prefix)); if (machine__is_host(machine)) kernel_type = DSO_TYPE_KERNEL; @@ -559,9 +548,9 @@ static int perf_event__process_kernel_mmap(struct perf_event_ops *ops __used, * time /proc/sys/kernel/kptr_restrict was non zero. */ if (event->mmap.pgoff != 0) { - perf_session__set_kallsyms_ref_reloc_sym(machine->vmlinux_maps, - symbol_name, - event->mmap.pgoff); + maps__set_kallsyms_ref_reloc_sym(machine->vmlinux_maps, + symbol_name, + event->mmap.pgoff); } if (machine__is_default_guest(machine)) { @@ -580,9 +569,8 @@ out_problem: int perf_event__process_mmap(struct perf_event_ops *ops, union perf_event *event, struct perf_sample *sample __used, - struct perf_session *session) + struct machine *machine) { - struct machine *machine; struct thread *thread; struct map *map; u8 cpumode = event->header.misc & PERF_RECORD_MISC_CPUMODE_MASK; @@ -594,16 +582,13 @@ int perf_event__process_mmap(struct perf_event_ops *ops, if (cpumode == PERF_RECORD_MISC_GUEST_KERNEL || cpumode == PERF_RECORD_MISC_KERNEL) { - ret = perf_event__process_kernel_mmap(ops, event, session); + ret = perf_event__process_kernel_mmap(ops, event, machine); if (ret < 0) goto out_problem; return 0; } - machine = perf_session__find_host_machine(session); - if (machine == NULL) - goto out_problem; - thread = perf_session__findnew(session, event->mmap.pid); + thread = machine__findnew_thread(machine, event->mmap.pid); if (thread == NULL) goto out_problem; map = map__new(&machine->user_dsos, event->mmap.start, @@ -624,16 +609,16 @@ out_problem: int perf_event__process_task(struct perf_event_ops *ops __used, union perf_event *event, struct perf_sample *sample __used, - struct perf_session *session) + struct machine *machine) { - struct thread *thread = perf_session__findnew(session, event->fork.tid); - struct thread *parent = perf_session__findnew(session, event->fork.ptid); + struct thread *thread = machine__findnew_thread(machine, event->fork.tid); + struct thread *parent = machine__findnew_thread(machine, event->fork.ptid); dump_printf("(%d:%d):(%d:%d)\n", event->fork.pid, event->fork.tid, event->fork.ppid, event->fork.ptid); if (event->header.type == PERF_RECORD_EXIT) { - perf_session__remove_thread(session, thread); + machine__remove_thread(machine, thread); return 0; } @@ -647,21 +632,21 @@ int perf_event__process_task(struct perf_event_ops *ops __used, } int perf_event__process(struct perf_event_ops *ops, union perf_event *event, - struct perf_sample *sample, struct perf_session *session) + struct perf_sample *sample, struct machine *machine) { switch (event->header.type) { case PERF_RECORD_COMM: - perf_event__process_comm(ops, event, sample, session); + perf_event__process_comm(ops, event, sample, machine); break; case PERF_RECORD_MMAP: - perf_event__process_mmap(ops, event, sample, session); + perf_event__process_mmap(ops, event, sample, machine); break; case PERF_RECORD_FORK: case PERF_RECORD_EXIT: - perf_event__process_task(ops, event, sample, session); + perf_event__process_task(ops, event, sample, machine); break; case PERF_RECORD_LOST: - perf_event__process_lost(ops, event, sample, session); + perf_event__process_lost(ops, event, sample, machine); default: break; } @@ -670,36 +655,29 @@ int perf_event__process(struct perf_event_ops *ops, union perf_event *event, } void thread__find_addr_map(struct thread *self, - struct perf_session *session, u8 cpumode, - enum map_type type, pid_t pid, u64 addr, + struct machine *machine, u8 cpumode, + enum map_type type, u64 addr, struct addr_location *al) { struct map_groups *mg = &self->mg; - struct machine *machine = NULL; al->thread = self; al->addr = addr; al->cpumode = cpumode; al->filtered = false; + if (machine == NULL) { + al->map = NULL; + return; + } + if (cpumode == PERF_RECORD_MISC_KERNEL && perf_host) { al->level = 'k'; - machine = perf_session__find_host_machine(session); - if (machine == NULL) { - al->map = NULL; - return; - } mg = &machine->kmaps; } else if (cpumode == PERF_RECORD_MISC_USER && perf_host) { al->level = '.'; - machine = perf_session__find_host_machine(session); } else if (cpumode == PERF_RECORD_MISC_GUEST_KERNEL && perf_guest) { al->level = 'g'; - machine = perf_session__find_machine(session, pid); - if (machine == NULL) { - al->map = NULL; - return; - } mg = &machine->kmaps; } else { /* @@ -745,13 +723,12 @@ try_again: al->addr = al->map->map_ip(al->map, al->addr); } -void thread__find_addr_location(struct thread *self, - struct perf_session *session, u8 cpumode, - enum map_type type, pid_t pid, u64 addr, +void thread__find_addr_location(struct thread *thread, struct machine *machine, + u8 cpumode, enum map_type type, u64 addr, struct addr_location *al, symbol_filter_t filter) { - thread__find_addr_map(self, session, cpumode, type, pid, addr, al); + thread__find_addr_map(thread, machine, cpumode, type, addr, al); if (al->map != NULL) al->sym = map__find_symbol(al->map, al->addr, filter); else @@ -759,13 +736,13 @@ void thread__find_addr_location(struct thread *self, } int perf_event__preprocess_sample(const union perf_event *event, - struct perf_session *session, + struct machine *machine, struct addr_location *al, struct perf_sample *sample, symbol_filter_t filter) { u8 cpumode = event->header.misc & PERF_RECORD_MISC_CPUMODE_MASK; - struct thread *thread = perf_session__findnew(session, event->ip.pid); + struct thread *thread = machine__findnew_thread(machine, event->ip.pid); if (thread == NULL) return -1; @@ -776,18 +753,18 @@ int perf_event__preprocess_sample(const union perf_event *event, dump_printf(" ... thread: %s:%d\n", thread->comm, thread->pid); /* - * Have we already created the kernel maps for the host machine? + * Have we already created the kernel maps for this machine? * * This should have happened earlier, when we processed the kernel MMAP * events, but for older perf.data files there was no such thing, so do * it now. */ if (cpumode == PERF_RECORD_MISC_KERNEL && - session->host_machine.vmlinux_maps[MAP__FUNCTION] == NULL) - machine__create_kernel_maps(&session->host_machine); + machine->vmlinux_maps[MAP__FUNCTION] == NULL) + machine__create_kernel_maps(machine); - thread__find_addr_map(thread, session, cpumode, MAP__FUNCTION, - event->ip.pid, event->ip.ip, al); + thread__find_addr_map(thread, machine, cpumode, MAP__FUNCTION, + event->ip.ip, al); dump_printf(" ...... dso: %s\n", al->map ? al->map->dso->long_name : al->level == 'H' ? "[hypervisor]" : ""); diff --git a/tools/perf/util/event.h b/tools/perf/util/event.h index 669409d35710..1564877e8703 100644 --- a/tools/perf/util/event.h +++ b/tools/perf/util/event.h @@ -142,56 +142,53 @@ union perf_event { void perf_event__print_totals(void); struct perf_event_ops; -struct perf_session; struct thread_map; typedef int (*perf_event__handler_t)(struct perf_event_ops *ops, union perf_event *event, struct perf_sample *sample, - struct perf_session *session); + struct machine *machine); int perf_event__synthesize_thread_map(struct perf_event_ops *ops, struct thread_map *threads, perf_event__handler_t process, - struct perf_session *session); + struct machine *machine); int perf_event__synthesize_threads(struct perf_event_ops *ops, perf_event__handler_t process, - struct perf_session *session); + struct machine *machine); int perf_event__synthesize_kernel_mmap(struct perf_event_ops *ops, perf_event__handler_t process, - struct perf_session *session, struct machine *machine, const char *symbol_name); int perf_event__synthesize_modules(struct perf_event_ops *ops, perf_event__handler_t process, - struct perf_session *session, struct machine *machine); int perf_event__process_comm(struct perf_event_ops *ops, union perf_event *event, struct perf_sample *sample, - struct perf_session *session); + struct machine *machine); int perf_event__process_lost(struct perf_event_ops *ops, union perf_event *event, struct perf_sample *sample, - struct perf_session *session); + struct machine *machine); int perf_event__process_mmap(struct perf_event_ops *ops, union perf_event *event, struct perf_sample *sample, - struct perf_session *session); + struct machine *machine); int perf_event__process_task(struct perf_event_ops *ops, union perf_event *event, struct perf_sample *sample, - struct perf_session *session); + struct machine *machine); int perf_event__process(struct perf_event_ops *ops, union perf_event *event, struct perf_sample *sample, - struct perf_session *session); + struct machine *machine); struct addr_location; int perf_event__preprocess_sample(const union perf_event *self, - struct perf_session *session, + struct machine *machine, struct addr_location *al, struct perf_sample *sample, symbol_filter_t filter); diff --git a/tools/perf/util/header.c b/tools/perf/util/header.c index ab3a2b0e8f06..db280d6ca898 100644 --- a/tools/perf/util/header.c +++ b/tools/perf/util/header.c @@ -2072,8 +2072,7 @@ out_delete_evlist: int perf_event__synthesize_attr(struct perf_event_ops *ops, struct perf_event_attr *attr, u16 ids, u64 *id, - perf_event__handler_t process, - struct perf_session *session) + perf_event__handler_t process) { union perf_event *ev; size_t size; @@ -2095,7 +2094,7 @@ int perf_event__synthesize_attr(struct perf_event_ops *ops, ev->attr.header.type = PERF_RECORD_HEADER_ATTR; ev->attr.header.size = size; - err = process(ops, ev, NULL, session); + err = process(ops, ev, NULL, NULL); free(ev); @@ -2111,7 +2110,7 @@ int perf_event__synthesize_attrs(struct perf_event_ops *ops, list_for_each_entry(attr, &session->evlist->entries, node) { err = perf_event__synthesize_attr(ops, &attr->attr, attr->ids, - attr->id, process, session); + attr->id, process); if (err) { pr_debug("failed to create perf header attribute\n"); return err; @@ -2161,7 +2160,7 @@ int perf_event__process_attr(union perf_event *event, int perf_event__synthesize_event_type(struct perf_event_ops *ops, u64 event_id, char *name, perf_event__handler_t process, - struct perf_session *session) + struct machine *machine) { union perf_event ev; size_t size = 0; @@ -2179,14 +2178,14 @@ int perf_event__synthesize_event_type(struct perf_event_ops *ops, ev.event_type.header.size = sizeof(ev.event_type) - (sizeof(ev.event_type.event_type.name) - size); - err = process(ops, &ev, NULL, session); + err = process(ops, &ev, NULL, machine); return err; } int perf_event__synthesize_event_types(struct perf_event_ops *ops, perf_event__handler_t process, - struct perf_session *session) + struct machine *machine) { struct perf_trace_event_type *type; int i, err = 0; @@ -2196,7 +2195,7 @@ int perf_event__synthesize_event_types(struct perf_event_ops *ops, err = perf_event__synthesize_event_type(ops, type->event_id, type->name, process, - session); + machine); if (err) { pr_debug("failed to create perf header event type\n"); return err; @@ -2207,8 +2206,7 @@ int perf_event__synthesize_event_types(struct perf_event_ops *ops, } int perf_event__process_event_type(struct perf_event_ops *ops __unused, - union perf_event *event, - struct perf_session *session __unused) + union perf_event *event) { if (perf_header__push_event(event->event_type.event_type.event_id, event->event_type.event_type.name) < 0) @@ -2219,8 +2217,7 @@ int perf_event__process_event_type(struct perf_event_ops *ops __unused, int perf_event__synthesize_tracing_data(struct perf_event_ops *ops, int fd, struct perf_evlist *evlist, - perf_event__handler_t process, - struct perf_session *session __unused) + perf_event__handler_t process) { union perf_event ev; struct tracing_data *tdata; @@ -2251,7 +2248,7 @@ int perf_event__synthesize_tracing_data(struct perf_event_ops *ops, int fd, ev.tracing_data.header.size = sizeof(ev.tracing_data); ev.tracing_data.size = aligned_size; - process(ops, &ev, NULL, session); + process(ops, &ev, NULL, NULL); /* * The put function will copy all the tracing data @@ -2296,8 +2293,7 @@ int perf_event__process_tracing_data(union perf_event *event, int perf_event__synthesize_build_id(struct perf_event_ops *ops, struct dso *pos, u16 misc, perf_event__handler_t process, - struct machine *machine, - struct perf_session *session) + struct machine *machine) { union perf_event ev; size_t len; @@ -2317,7 +2313,7 @@ int perf_event__synthesize_build_id(struct perf_event_ops *ops, ev.build_id.header.size = sizeof(ev.build_id) + len; memcpy(&ev.build_id.filename, pos->long_name, pos->long_name_len); - err = process(ops, &ev, NULL, session); + err = process(ops, &ev, NULL, machine); return err; } diff --git a/tools/perf/util/header.h b/tools/perf/util/header.h index 54dae5f09556..a604962fc431 100644 --- a/tools/perf/util/header.h +++ b/tools/perf/util/header.h @@ -99,8 +99,7 @@ int build_id_cache__remove_s(const char *sbuild_id, const char *debugdir); int perf_event__synthesize_attr(struct perf_event_ops *ops, struct perf_event_attr *attr, u16 ids, u64 *id, - perf_event__handler_t process, - struct perf_session *session); + perf_event__handler_t process); int perf_event__synthesize_attrs(struct perf_event_ops *ops, struct perf_session *session, perf_event__handler_t process); @@ -109,26 +108,23 @@ int perf_event__process_attr(union perf_event *event, struct perf_evlist **pevli int perf_event__synthesize_event_type(struct perf_event_ops *ops, u64 event_id, char *name, perf_event__handler_t process, - struct perf_session *session); + struct machine *machine); int perf_event__synthesize_event_types(struct perf_event_ops *ops, perf_event__handler_t process, - struct perf_session *session); + struct machine *machine); int perf_event__process_event_type(struct perf_event_ops *ops, - union perf_event *event, - struct perf_session *session); + union perf_event *event); int perf_event__synthesize_tracing_data(struct perf_event_ops *ops, int fd, struct perf_evlist *evlist, - perf_event__handler_t process, - struct perf_session *session); + perf_event__handler_t process); int perf_event__process_tracing_data(union perf_event *event, struct perf_session *session); int perf_event__synthesize_build_id(struct perf_event_ops *ops, struct dso *pos, u16 misc, perf_event__handler_t process, - struct machine *machine, - struct perf_session *session); + struct machine *machine); int perf_event__process_build_id(struct perf_event_ops *ops, union perf_event *event, struct perf_session *session); diff --git a/tools/perf/util/map.h b/tools/perf/util/map.h index bde6835ee257..2b8017f8a930 100644 --- a/tools/perf/util/map.h +++ b/tools/perf/util/map.h @@ -18,9 +18,11 @@ enum map_type { extern const char *map_type__name[MAP__NR_TYPES]; struct dso; +struct ip_callchain; struct ref_reloc_sym; struct map_groups; struct machine; +struct perf_evsel; struct map { union { @@ -61,6 +63,7 @@ struct map_groups { struct machine { struct rb_node rb_node; pid_t pid; + u16 id_hdr_size; char *root_dir; struct rb_root threads; struct list_head dead_threads; @@ -151,6 +154,13 @@ int machine__init(struct machine *self, const char *root_dir, pid_t pid); void machine__exit(struct machine *self); void machine__delete(struct machine *self); +int machine__resolve_callchain(struct machine *machine, + struct perf_evsel *evsel, struct thread *thread, + struct ip_callchain *chain, + struct symbol **parent); +int maps__set_kallsyms_ref_reloc_sym(struct map **maps, const char *symbol_name, + u64 addr); + /* * Default guest kernel is defined by parameter --guestkallsyms * and --guestmodules diff --git a/tools/perf/util/scripting-engines/trace-event-perl.c b/tools/perf/util/scripting-engines/trace-event-perl.c index 74350ffb57fe..a82ce4303ff5 100644 --- a/tools/perf/util/scripting-engines/trace-event-perl.c +++ b/tools/perf/util/scripting-engines/trace-event-perl.c @@ -27,6 +27,8 @@ #include "../../perf.h" #include "../util.h" +#include "../thread.h" +#include "../event.h" #include "../trace-event.h" #include @@ -248,7 +250,7 @@ static inline struct event *find_cache_event(int type) static void perl_process_event(union perf_event *pevent __unused, struct perf_sample *sample, struct perf_evsel *evsel, - struct perf_session *session __unused, + struct machine *machine __unused, struct thread *thread) { struct format_field *field; diff --git a/tools/perf/util/scripting-engines/trace-event-python.c b/tools/perf/util/scripting-engines/trace-event-python.c index 6ccf70e8d8f2..0b2a48783172 100644 --- a/tools/perf/util/scripting-engines/trace-event-python.c +++ b/tools/perf/util/scripting-engines/trace-event-python.c @@ -29,6 +29,8 @@ #include "../../perf.h" #include "../util.h" +#include "../event.h" +#include "../thread.h" #include "../trace-event.h" PyMODINIT_FUNC initperf_trace_context(void); @@ -207,7 +209,7 @@ static inline struct event *find_cache_event(int type) static void python_process_event(union perf_event *pevent __unused, struct perf_sample *sample, struct perf_evsel *evsel __unused, - struct perf_session *session __unused, + struct machine *machine __unused, struct thread *thread) { PyObject *handler, *retval, *context, *t, *obj, *dict = NULL; diff --git a/tools/perf/util/session.c b/tools/perf/util/session.c index a36023a66779..be33606386bf 100644 --- a/tools/perf/util/session.c +++ b/tools/perf/util/session.c @@ -84,6 +84,7 @@ void perf_session__update_sample_type(struct perf_session *self) self->sample_size = __perf_evsel__sample_size(self->sample_type); self->sample_id_all = perf_evlist__sample_id_all(self->evlist); self->id_hdr_size = perf_evlist__id_hdr_size(self->evlist); + self->host_machine.id_hdr_size = self->id_hdr_size; } int perf_session__create_kernel_maps(struct perf_session *self) @@ -216,10 +217,10 @@ static bool symbol__match_parent_regex(struct symbol *sym) return 0; } -int perf_session__resolve_callchain(struct perf_session *self, struct perf_evsel *evsel, - struct thread *thread, - struct ip_callchain *chain, - struct symbol **parent) +int machine__resolve_callchain(struct machine *self, struct perf_evsel *evsel, + struct thread *thread, + struct ip_callchain *chain, + struct symbol **parent) { u8 cpumode = PERF_RECORD_MISC_USER; unsigned int i; @@ -252,7 +253,7 @@ int perf_session__resolve_callchain(struct perf_session *self, struct perf_evsel al.filtered = false; thread__find_addr_location(thread, self, cpumode, - MAP__FUNCTION, thread->pid, ip, &al, NULL); + MAP__FUNCTION, ip, &al, NULL); if (al.sym != NULL) { if (sort__has_parent && !*parent && symbol__match_parent_regex(al.sym)) @@ -270,14 +271,6 @@ int perf_session__resolve_callchain(struct perf_session *self, struct perf_evsel return 0; } -static int process_event_synth_stub(struct perf_event_ops *ops __used, - union perf_event *event __used, - struct perf_session *session __used) -{ - dump_printf(": unhandled!\n"); - return 0; -} - static int process_event_synth_tracing_data_stub(union perf_event *event __used, struct perf_session *session __used) { @@ -296,7 +289,7 @@ static int process_event_sample_stub(struct perf_event_ops *ops __used, union perf_event *event __used, struct perf_sample *sample __used, struct perf_evsel *evsel __used, - struct perf_session *session __used) + struct machine *machine __used) { dump_printf(": unhandled!\n"); return 0; @@ -305,7 +298,7 @@ static int process_event_sample_stub(struct perf_event_ops *ops __used, static int process_event_stub(struct perf_event_ops *ops __used, union perf_event *event __used, struct perf_sample *sample __used, - struct perf_session *session __used) + struct machine *machine __used) { dump_printf(": unhandled!\n"); return 0; @@ -313,7 +306,14 @@ static int process_event_stub(struct perf_event_ops *ops __used, static int process_finished_round_stub(struct perf_event_ops *ops __used, union perf_event *event __used, - struct perf_session *session __used) + struct perf_session *perf_session __used) +{ + dump_printf(": unhandled!\n"); + return 0; +} + +static int process_event_type_stub(struct perf_event_ops *ops __used, + union perf_event *event __used) { dump_printf(": unhandled!\n"); return 0; @@ -338,7 +338,7 @@ static void perf_event_ops__fill_defaults(struct perf_event_ops *handler) if (handler->lost == NULL) handler->lost = perf_event__process_lost; if (handler->read == NULL) - handler->read = process_event_stub; + handler->read = process_event_sample_stub; if (handler->throttle == NULL) handler->throttle = process_event_stub; if (handler->unthrottle == NULL) @@ -346,11 +346,11 @@ static void perf_event_ops__fill_defaults(struct perf_event_ops *handler) if (handler->attr == NULL) handler->attr = process_event_synth_attr_stub; if (handler->event_type == NULL) - handler->event_type = process_event_synth_stub; + handler->event_type = process_event_type_stub; if (handler->tracing_data == NULL) handler->tracing_data = process_event_synth_tracing_data_stub; if (handler->build_id == NULL) - handler->build_id = process_event_synth_stub; + handler->build_id = process_finished_round_stub; if (handler->finished_round == NULL) { if (handler->ordered_samples) handler->finished_round = process_finished_round; @@ -734,6 +734,18 @@ static void dump_sample(struct perf_session *session, union perf_event *event, callchain__printf(sample); } +static struct machine * + perf_session__find_machine_for_cpumode(struct perf_session *session, + union perf_event *event) +{ + const u8 cpumode = event->header.misc & PERF_RECORD_MISC_CPUMODE_MASK; + + if (cpumode == PERF_RECORD_MISC_GUEST_KERNEL && perf_guest) + return perf_session__find_machine(session, event->ip.pid); + + return perf_session__find_host_machine(session); +} + static int perf_session_deliver_event(struct perf_session *session, union perf_event *event, struct perf_sample *sample, @@ -741,6 +753,7 @@ static int perf_session_deliver_event(struct perf_session *session, u64 file_offset) { struct perf_evsel *evsel; + struct machine *machine; dump_event(session, event, file_offset, sample); @@ -762,6 +775,8 @@ static int perf_session_deliver_event(struct perf_session *session, hists__inc_nr_events(&evsel->hists, event->header.type); } + machine = perf_session__find_machine_for_cpumode(session, event); + switch (event->header.type) { case PERF_RECORD_SAMPLE: dump_sample(session, event, sample); @@ -769,23 +784,25 @@ static int perf_session_deliver_event(struct perf_session *session, ++session->hists.stats.nr_unknown_id; return -1; } - return ops->sample(ops, event, sample, evsel, session); + return ops->sample(ops, event, sample, evsel, machine); case PERF_RECORD_MMAP: - return ops->mmap(ops, event, sample, session); + return ops->mmap(ops, event, sample, machine); case PERF_RECORD_COMM: - return ops->comm(ops, event, sample, session); + return ops->comm(ops, event, sample, machine); case PERF_RECORD_FORK: - return ops->fork(ops, event, sample, session); + return ops->fork(ops, event, sample, machine); case PERF_RECORD_EXIT: - return ops->exit(ops, event, sample, session); + return ops->exit(ops, event, sample, machine); case PERF_RECORD_LOST: - return ops->lost(ops, event, sample, session); + if (ops->lost == perf_event__process_lost) + session->hists.stats.total_lost += event->lost.lost; + return ops->lost(ops, event, sample, machine); case PERF_RECORD_READ: - return ops->read(ops, event, sample, session); + return ops->read(ops, event, sample, evsel, machine); case PERF_RECORD_THROTTLE: - return ops->throttle(ops, event, sample, session); + return ops->throttle(ops, event, sample, machine); case PERF_RECORD_UNTHROTTLE: - return ops->unthrottle(ops, event, sample, session); + return ops->unthrottle(ops, event, sample, machine); default: ++session->hists.stats.nr_unknown_events; return -1; @@ -823,7 +840,7 @@ static int perf_session__process_user_event(struct perf_session *session, union perf_session__update_sample_type(session); return err; case PERF_RECORD_HEADER_EVENT_TYPE: - return ops->event_type(ops, event, session); + return ops->event_type(ops, event); case PERF_RECORD_HEADER_TRACING_DATA: /* setup for reading amidst mmap */ lseek(session->fd, file_offset, SEEK_SET); @@ -1170,9 +1187,8 @@ bool perf_session__has_traces(struct perf_session *self, const char *msg) return true; } -int perf_session__set_kallsyms_ref_reloc_sym(struct map **maps, - const char *symbol_name, - u64 addr) +int maps__set_kallsyms_ref_reloc_sym(struct map **maps, + const char *symbol_name, u64 addr) { char *bracket; enum map_type i; @@ -1264,17 +1280,16 @@ struct perf_evsel *perf_session__find_first_evtype(struct perf_session *session, return NULL; } -void perf_session__print_ip(union perf_event *event, struct perf_evsel *evsel, - struct perf_sample *sample, - struct perf_session *session, - int print_sym, int print_dso) +void perf_event__print_ip(union perf_event *event, struct perf_sample *sample, + struct machine *machine, struct perf_evsel *evsel, + int print_sym, int print_dso) { struct addr_location al; const char *symname, *dsoname; struct callchain_cursor *cursor = &evsel->hists.callchain_cursor; struct callchain_cursor_node *node; - if (perf_event__preprocess_sample(event, session, &al, sample, + if (perf_event__preprocess_sample(event, machine, &al, sample, NULL) < 0) { error("problem processing %d event, skipping it.\n", event->header.type); @@ -1283,7 +1298,7 @@ void perf_session__print_ip(union perf_event *event, struct perf_evsel *evsel, if (symbol_conf.use_callchain && sample->callchain) { - if (perf_session__resolve_callchain(session, evsel, al.thread, + if (machine__resolve_callchain(machine, evsel, al.thread, sample->callchain, NULL) != 0) { if (verbose) error("Failed to resolve callchain. Skipping\n"); diff --git a/tools/perf/util/session.h b/tools/perf/util/session.h index 6de3d1368900..1c5823c7d6dc 100644 --- a/tools/perf/util/session.h +++ b/tools/perf/util/session.h @@ -58,32 +58,34 @@ struct perf_event_ops; typedef int (*event_sample)(struct perf_event_ops *ops, union perf_event *event, struct perf_sample *sample, - struct perf_evsel *evsel, struct perf_session *session); + struct perf_evsel *evsel, struct machine *machine); typedef int (*event_op)(struct perf_event_ops *ops, union perf_event *event, struct perf_sample *sample, - struct perf_session *session); + struct machine *machine); typedef int (*event_synth_op)(union perf_event *self, struct perf_session *session); typedef int (*event_attr_op)(union perf_event *event, struct perf_evlist **pevlist); +typedef int (*event_simple_op)(struct perf_event_ops *ops, + union perf_event *event); typedef int (*event_op2)(struct perf_event_ops *ops, union perf_event *event, struct perf_session *session); struct perf_event_ops { - event_sample sample; + event_sample sample, + read; event_op mmap, comm, fork, exit, lost, - read, throttle, unthrottle; event_attr_op attr; event_synth_op tracing_data; - event_op2 event_type, - build_id, - finished_round; + event_simple_op event_type; + event_op2 finished_round, + build_id; bool ordered_samples; bool ordering_requires_timestamps; }; @@ -108,10 +110,6 @@ int perf_session__resolve_callchain(struct perf_session *self, struct perf_evsel bool perf_session__has_traces(struct perf_session *self, const char *msg); -int perf_session__set_kallsyms_ref_reloc_sym(struct map **maps, - const char *symbol_name, - u64 addr); - void mem_bswap_64(void *src, int byte_size); void perf_event__attr_swap(struct perf_event_attr *attr); @@ -151,6 +149,9 @@ void perf_session__process_machines(struct perf_session *self, return machines__process(&self->machines, process, ops); } +struct thread *perf_session__findnew(struct perf_session *self, pid_t pid); +size_t perf_session__fprintf(struct perf_session *self, FILE *fp); + size_t perf_session__fprintf_dsos(struct perf_session *self, FILE *fp); size_t perf_session__fprintf_dsos_buildid(struct perf_session *self, @@ -171,10 +172,9 @@ static inline int perf_session__parse_sample(struct perf_session *session, struct perf_evsel *perf_session__find_first_evtype(struct perf_session *session, unsigned int type); -void perf_session__print_ip(union perf_event *event, struct perf_evsel *evsel, - struct perf_sample *sample, - struct perf_session *session, - int print_sym, int print_dso); +void perf_event__print_ip(union perf_event *event, struct perf_sample *sample, + struct machine *machine, struct perf_evsel *evsel, + int print_sym, int print_dso); int perf_session__cpu_bitmap(struct perf_session *session, const char *cpu_list, unsigned long *cpu_bitmap); diff --git a/tools/perf/util/thread.h b/tools/perf/util/thread.h index e5f2401c1b5e..70c2c13ff679 100644 --- a/tools/perf/util/thread.h +++ b/tools/perf/util/thread.h @@ -18,16 +18,14 @@ struct thread { int comm_len; }; -struct perf_session; +struct machine; void thread__delete(struct thread *self); int thread__set_comm(struct thread *self, const char *comm); int thread__comm_len(struct thread *self); -struct thread *perf_session__findnew(struct perf_session *self, pid_t pid); void thread__insert_map(struct thread *self, struct map *map); int thread__fork(struct thread *self, struct thread *parent); -size_t perf_session__fprintf(struct perf_session *self, FILE *fp); static inline struct map *thread__find_map(struct thread *self, enum map_type type, u64 addr) @@ -35,14 +33,12 @@ static inline struct map *thread__find_map(struct thread *self, return self ? map_groups__find(&self->mg, type, addr) : NULL; } -void thread__find_addr_map(struct thread *self, - struct perf_session *session, u8 cpumode, - enum map_type type, pid_t pid, u64 addr, +void thread__find_addr_map(struct thread *thread, struct machine *machine, + u8 cpumode, enum map_type type, u64 addr, struct addr_location *al); -void thread__find_addr_location(struct thread *self, - struct perf_session *session, u8 cpumode, - enum map_type type, pid_t pid, u64 addr, +void thread__find_addr_location(struct thread *thread, struct machine *machine, + u8 cpumode, enum map_type type, u64 addr, struct addr_location *al, symbol_filter_t filter); #endif /* __PERF_THREAD_H */ diff --git a/tools/perf/util/trace-event-scripting.c b/tools/perf/util/trace-event-scripting.c index c9dcbec7d800..a3fdf55f317b 100644 --- a/tools/perf/util/trace-event-scripting.c +++ b/tools/perf/util/trace-event-scripting.c @@ -39,7 +39,7 @@ static int stop_script_unsupported(void) static void process_event_unsupported(union perf_event *event __unused, struct perf_sample *sample __unused, struct perf_evsel *evsel __unused, - struct perf_session *session __unused, + struct machine *machine __unused, struct thread *thread __unused) { } diff --git a/tools/perf/util/trace-event.h b/tools/perf/util/trace-event.h index a84100817649..58ae14c5baac 100644 --- a/tools/perf/util/trace-event.h +++ b/tools/perf/util/trace-event.h @@ -3,7 +3,11 @@ #include #include "parse-events.h" -#include "session.h" + +struct machine; +struct perf_sample; +union perf_event; +struct thread; #define __unused __attribute__((unused)) @@ -292,7 +296,7 @@ struct scripting_ops { void (*process_event) (union perf_event *event, struct perf_sample *sample, struct perf_evsel *evsel, - struct perf_session *session, + struct machine *machine, struct thread *thread); int (*generate_script) (const char *outfile); }; -- cgit v1.2.3 From 45694aa7702bc44d538a3bcb51bb2bb96cf190c0 Mon Sep 17 00:00:00 2001 From: Arnaldo Carvalho de Melo Date: Mon, 28 Nov 2011 08:30:20 -0200 Subject: perf tools: Rename perf_event_ops to perf_tool To better reflect that it became the base class for all tools, that must be in each tool struct and where common stuff will be put. Cc: David Ahern Cc: Frederic Weisbecker Cc: Mike Galbraith Cc: Paul Mackerras Cc: Peter Zijlstra Cc: Stephane Eranian Link: http://lkml.kernel.org/n/tip-qgpc4msetqlwr8y2k7537cxe@git.kernel.org Signed-off-by: Arnaldo Carvalho de Melo --- tools/perf/Makefile | 1 + tools/perf/builtin-annotate.c | 13 ++-- tools/perf/builtin-diff.c | 11 +-- tools/perf/builtin-inject.c | 55 +++++++------- tools/perf/builtin-kmem.c | 9 ++- tools/perf/builtin-lock.c | 5 +- tools/perf/builtin-record.c | 35 ++++----- tools/perf/builtin-report.c | 17 +++-- tools/perf/builtin-sched.c | 9 ++- tools/perf/builtin-script.c | 9 ++- tools/perf/builtin-timechart.c | 15 ++-- tools/perf/builtin-top.c | 6 +- tools/perf/util/build-id.c | 7 +- tools/perf/util/build-id.h | 2 +- tools/perf/util/event.c | 54 +++++++------- tools/perf/util/event.h | 22 +++--- tools/perf/util/header.c | 28 +++---- tools/perf/util/header.h | 16 ++-- tools/perf/util/session.c | 163 +++++++++++++++++++++-------------------- tools/perf/util/session.h | 49 ++----------- tools/perf/util/tool.h | 45 ++++++++++++ tools/perf/util/top.h | 7 +- 22 files changed, 301 insertions(+), 277 deletions(-) create mode 100644 tools/perf/util/tool.h diff --git a/tools/perf/Makefile b/tools/perf/Makefile index b98e3075646b..ac86d67b636e 100644 --- a/tools/perf/Makefile +++ b/tools/perf/Makefile @@ -278,6 +278,7 @@ LIB_H += util/strbuf.h LIB_H += util/strlist.h LIB_H += util/strfilter.h LIB_H += util/svghelper.h +LIB_H += util/tool.h LIB_H += util/run-command.h LIB_H += util/sigchain.h LIB_H += util/symbol.h diff --git a/tools/perf/builtin-annotate.c b/tools/perf/builtin-annotate.c index dff081a388bb..c01139fa4a10 100644 --- a/tools/perf/builtin-annotate.c +++ b/tools/perf/builtin-annotate.c @@ -27,11 +27,12 @@ #include "util/sort.h" #include "util/hist.h" #include "util/session.h" +#include "util/tool.h" #include struct perf_annotate { - struct perf_event_ops ops; + struct perf_tool tool; char const *input_name; bool force, use_tui, use_stdio; bool full_paths; @@ -79,13 +80,13 @@ static int perf_evsel__add_sample(struct perf_evsel *evsel, return ret; } -static int process_sample_event(struct perf_event_ops *ops, +static int process_sample_event(struct perf_tool *tool, union perf_event *event, struct perf_sample *sample, struct perf_evsel *evsel, struct machine *machine) { - struct perf_annotate *ann = container_of(ops, struct perf_annotate, ops); + struct perf_annotate *ann = container_of(tool, struct perf_annotate, tool); struct addr_location al; if (perf_event__preprocess_sample(event, machine, &al, sample, @@ -174,7 +175,7 @@ static int __cmd_annotate(struct perf_annotate *ann) u64 total_nr_samples; session = perf_session__new(ann->input_name, O_RDONLY, - ann->force, false, &ann->ops); + ann->force, false, &ann->tool); if (session == NULL) return -ENOMEM; @@ -185,7 +186,7 @@ static int __cmd_annotate(struct perf_annotate *ann) goto out_delete; } - ret = perf_session__process_events(session, &ann->ops); + ret = perf_session__process_events(session, &ann->tool); if (ret) goto out_delete; @@ -241,7 +242,7 @@ static const char * const annotate_usage[] = { int cmd_annotate(int argc, const char **argv, const char *prefix __used) { struct perf_annotate annotate = { - .ops = { + .tool = { .sample = process_sample_event, .mmap = perf_event__process_mmap, .comm = perf_event__process_comm, diff --git a/tools/perf/builtin-diff.c b/tools/perf/builtin-diff.c index 478b0aeb2a62..4f19513d7dda 100644 --- a/tools/perf/builtin-diff.c +++ b/tools/perf/builtin-diff.c @@ -11,6 +11,7 @@ #include "util/hist.h" #include "util/evsel.h" #include "util/session.h" +#include "util/tool.h" #include "util/sort.h" #include "util/symbol.h" #include "util/util.h" @@ -31,7 +32,7 @@ static int hists__add_entry(struct hists *self, return -ENOMEM; } -static int diff__process_sample_event(struct perf_event_ops *ops __used, +static int diff__process_sample_event(struct perf_tool *tool __used, union perf_event *event, struct perf_sample *sample, struct perf_evsel *evsel __used, @@ -57,7 +58,7 @@ static int diff__process_sample_event(struct perf_event_ops *ops __used, return 0; } -static struct perf_event_ops event_ops = { +static struct perf_tool perf_diff = { .sample = diff__process_sample_event, .mmap = perf_event__process_mmap, .comm = perf_event__process_comm, @@ -147,13 +148,13 @@ static int __cmd_diff(void) int ret, i; struct perf_session *session[2]; - session[0] = perf_session__new(input_old, O_RDONLY, force, false, &event_ops); - session[1] = perf_session__new(input_new, O_RDONLY, force, false, &event_ops); + session[0] = perf_session__new(input_old, O_RDONLY, force, false, &perf_diff); + session[1] = perf_session__new(input_new, O_RDONLY, force, false, &perf_diff); if (session[0] == NULL || session[1] == NULL) return -ENOMEM; for (i = 0; i < 2; ++i) { - ret = perf_session__process_events(session[i], &event_ops); + ret = perf_session__process_events(session[i], &perf_diff); if (ret) goto out_delete; } diff --git a/tools/perf/builtin-inject.c b/tools/perf/builtin-inject.c index a5bcf81776fc..09c106193e65 100644 --- a/tools/perf/builtin-inject.c +++ b/tools/perf/builtin-inject.c @@ -9,6 +9,7 @@ #include "perf.h" #include "util/session.h" +#include "util/tool.h" #include "util/debug.h" #include "util/parse-options.h" @@ -16,7 +17,7 @@ static char const *input_name = "-"; static bool inject_build_ids; -static int perf_event__repipe_synth(struct perf_event_ops *ops __used, +static int perf_event__repipe_synth(struct perf_tool *tool __used, union perf_event *event, struct machine *machine __used) { @@ -37,17 +38,17 @@ static int perf_event__repipe_synth(struct perf_event_ops *ops __used, return 0; } -static int perf_event__repipe_op2_synth(struct perf_event_ops *ops, +static int perf_event__repipe_op2_synth(struct perf_tool *tool, union perf_event *event, struct perf_session *session __used) { - return perf_event__repipe_synth(ops, event, NULL); + return perf_event__repipe_synth(tool, event, NULL); } -static int perf_event__repipe_event_type_synth(struct perf_event_ops *ops, +static int perf_event__repipe_event_type_synth(struct perf_tool *tool, union perf_event *event) { - return perf_event__repipe_synth(ops, event, NULL); + return perf_event__repipe_synth(tool, event, NULL); } static int perf_event__repipe_tracing_data_synth(union perf_event *event, @@ -62,45 +63,45 @@ static int perf_event__repipe_attr(union perf_event *event, return perf_event__repipe_synth(NULL, event, NULL); } -static int perf_event__repipe(struct perf_event_ops *ops, +static int perf_event__repipe(struct perf_tool *tool, union perf_event *event, struct perf_sample *sample __used, struct machine *machine) { - return perf_event__repipe_synth(ops, event, machine); + return perf_event__repipe_synth(tool, event, machine); } -static int perf_event__repipe_sample(struct perf_event_ops *ops, +static int perf_event__repipe_sample(struct perf_tool *tool, union perf_event *event, struct perf_sample *sample __used, struct perf_evsel *evsel __used, struct machine *machine) { - return perf_event__repipe_synth(ops, event, machine); + return perf_event__repipe_synth(tool, event, machine); } -static int perf_event__repipe_mmap(struct perf_event_ops *ops, +static int perf_event__repipe_mmap(struct perf_tool *tool, union perf_event *event, struct perf_sample *sample, struct machine *machine) { int err; - err = perf_event__process_mmap(ops, event, sample, machine); - perf_event__repipe(ops, event, sample, machine); + err = perf_event__process_mmap(tool, event, sample, machine); + perf_event__repipe(tool, event, sample, machine); return err; } -static int perf_event__repipe_task(struct perf_event_ops *ops, +static int perf_event__repipe_task(struct perf_tool *tool, union perf_event *event, struct perf_sample *sample, struct machine *machine) { int err; - err = perf_event__process_task(ops, event, sample, machine); - perf_event__repipe(ops, event, sample, machine); + err = perf_event__process_task(tool, event, sample, machine); + perf_event__repipe(tool, event, sample, machine); return err; } @@ -130,7 +131,7 @@ static int dso__read_build_id(struct dso *self) return -1; } -static int dso__inject_build_id(struct dso *self, struct perf_event_ops *ops, +static int dso__inject_build_id(struct dso *self, struct perf_tool *tool, struct machine *machine) { u16 misc = PERF_RECORD_MISC_USER; @@ -144,7 +145,7 @@ static int dso__inject_build_id(struct dso *self, struct perf_event_ops *ops, if (self->kernel) misc = PERF_RECORD_MISC_KERNEL; - err = perf_event__synthesize_build_id(ops, self, misc, perf_event__repipe, + err = perf_event__synthesize_build_id(tool, self, misc, perf_event__repipe, machine); if (err) { pr_err("Can't synthesize build_id event for %s\n", self->long_name); @@ -154,7 +155,7 @@ static int dso__inject_build_id(struct dso *self, struct perf_event_ops *ops, return 0; } -static int perf_event__inject_buildid(struct perf_event_ops *ops, +static int perf_event__inject_buildid(struct perf_tool *tool, union perf_event *event, struct perf_sample *sample, struct perf_evsel *evsel __used, @@ -180,7 +181,7 @@ static int perf_event__inject_buildid(struct perf_event_ops *ops, if (!al.map->dso->hit) { al.map->dso->hit = 1; if (map__load(al.map, NULL) >= 0) { - dso__inject_build_id(al.map->dso, ops, machine); + dso__inject_build_id(al.map->dso, tool, machine); /* * If this fails, too bad, let the other side * account this as unresolved. @@ -193,11 +194,11 @@ static int perf_event__inject_buildid(struct perf_event_ops *ops, } repipe: - perf_event__repipe(ops, event, sample, machine); + perf_event__repipe(tool, event, sample, machine); return 0; } -struct perf_event_ops inject_ops = { +struct perf_tool perf_inject = { .sample = perf_event__repipe_sample, .mmap = perf_event__repipe, .comm = perf_event__repipe, @@ -228,17 +229,17 @@ static int __cmd_inject(void) signal(SIGINT, sig_handler); if (inject_build_ids) { - inject_ops.sample = perf_event__inject_buildid; - inject_ops.mmap = perf_event__repipe_mmap; - inject_ops.fork = perf_event__repipe_task; - inject_ops.tracing_data = perf_event__repipe_tracing_data; + perf_inject.sample = perf_event__inject_buildid; + perf_inject.mmap = perf_event__repipe_mmap; + perf_inject.fork = perf_event__repipe_task; + perf_inject.tracing_data = perf_event__repipe_tracing_data; } - session = perf_session__new(input_name, O_RDONLY, false, true, &inject_ops); + session = perf_session__new(input_name, O_RDONLY, false, true, &perf_inject); if (session == NULL) return -ENOMEM; - ret = perf_session__process_events(session, &inject_ops); + ret = perf_session__process_events(session, &perf_inject); perf_session__delete(session); diff --git a/tools/perf/builtin-kmem.c b/tools/perf/builtin-kmem.c index 27b2a15dc7b2..886174e9525b 100644 --- a/tools/perf/builtin-kmem.c +++ b/tools/perf/builtin-kmem.c @@ -7,6 +7,7 @@ #include "util/thread.h" #include "util/header.h" #include "util/session.h" +#include "util/tool.h" #include "util/parse-options.h" #include "util/trace-event.h" @@ -303,7 +304,7 @@ static void process_raw_event(union perf_event *raw_event __used, void *data, } } -static int process_sample_event(struct perf_event_ops *ops __used, +static int process_sample_event(struct perf_tool *tool __used, union perf_event *event, struct perf_sample *sample, struct perf_evsel *evsel __used, @@ -325,7 +326,7 @@ static int process_sample_event(struct perf_event_ops *ops __used, return 0; } -static struct perf_event_ops event_ops = { +static struct perf_tool perf_kmem = { .sample = process_sample_event, .comm = perf_event__process_comm, .ordered_samples = true, @@ -484,7 +485,7 @@ static int __cmd_kmem(void) { int err = -EINVAL; struct perf_session *session = perf_session__new(input_name, O_RDONLY, - 0, false, &event_ops); + 0, false, &perf_kmem); if (session == NULL) return -ENOMEM; @@ -495,7 +496,7 @@ static int __cmd_kmem(void) goto out_delete; setup_pager(); - err = perf_session__process_events(session, &event_ops); + err = perf_session__process_events(session, &perf_kmem); if (err != 0) goto out_delete; sort_result(); diff --git a/tools/perf/builtin-lock.c b/tools/perf/builtin-lock.c index 99b032adb83e..4db5e5293067 100644 --- a/tools/perf/builtin-lock.c +++ b/tools/perf/builtin-lock.c @@ -12,6 +12,7 @@ #include "util/debug.h" #include "util/session.h" +#include "util/tool.h" #include #include @@ -845,7 +846,7 @@ static void dump_info(void) die("Unknown type of information\n"); } -static int process_sample_event(struct perf_event_ops *ops __used, +static int process_sample_event(struct perf_tool *tool __used, union perf_event *event, struct perf_sample *sample, struct perf_evsel *evsel __used, @@ -864,7 +865,7 @@ static int process_sample_event(struct perf_event_ops *ops __used, return 0; } -static struct perf_event_ops eops = { +static struct perf_tool eops = { .sample = process_sample_event, .comm = perf_event__process_comm, .ordered_samples = true, diff --git a/tools/perf/builtin-record.c b/tools/perf/builtin-record.c index 0af598a1059f..7d4fdaacc8ba 100644 --- a/tools/perf/builtin-record.c +++ b/tools/perf/builtin-record.c @@ -22,6 +22,7 @@ #include "util/evsel.h" #include "util/debug.h" #include "util/session.h" +#include "util/tool.h" #include "util/symbol.h" #include "util/cpumap.h" #include "util/thread_map.h" @@ -36,7 +37,7 @@ enum write_mode_t { }; struct perf_record { - struct perf_event_ops ops; + struct perf_tool tool; struct perf_record_opts opts; u64 bytes_written; const char *output_name; @@ -76,12 +77,12 @@ static void write_output(struct perf_record *rec, void *buf, size_t size) } } -static int process_synthesized_event(struct perf_event_ops *ops, +static int process_synthesized_event(struct perf_tool *tool, union perf_event *event, struct perf_sample *sample __used, struct machine *machine __used) { - struct perf_record *rec = container_of(ops, struct perf_record, ops); + struct perf_record *rec = container_of(tool, struct perf_record, tool); write_output(rec, event, event->header.size); return 0; } @@ -319,7 +320,7 @@ static void perf_record__exit(int status __used, void *arg) static void perf_event__synthesize_guest_os(struct machine *machine, void *data) { int err; - struct perf_event_ops *ops = data; + struct perf_tool *tool = data; if (machine__is_host(machine)) return; @@ -332,7 +333,7 @@ static void perf_event__synthesize_guest_os(struct machine *machine, void *data) *method is used to avoid symbol missing when the first addr is *in module instead of in guest kernel. */ - err = perf_event__synthesize_modules(ops, process_synthesized_event, + err = perf_event__synthesize_modules(tool, process_synthesized_event, machine); if (err < 0) pr_err("Couldn't record guest kernel [%d]'s reference" @@ -342,10 +343,10 @@ static void perf_event__synthesize_guest_os(struct machine *machine, void *data) * We use _stext for guest kernel because guest kernel's /proc/kallsyms * have no _text sometimes. */ - err = perf_event__synthesize_kernel_mmap(ops, process_synthesized_event, + err = perf_event__synthesize_kernel_mmap(tool, process_synthesized_event, machine, "_text"); if (err < 0) - err = perf_event__synthesize_kernel_mmap(ops, process_synthesized_event, + err = perf_event__synthesize_kernel_mmap(tool, process_synthesized_event, machine, "_stext"); if (err < 0) pr_err("Couldn't record guest kernel [%d]'s reference" @@ -378,7 +379,7 @@ static int __cmd_record(struct perf_record *rec, int argc, const char **argv) unsigned long waking = 0; const bool forks = argc > 0; struct machine *machine; - struct perf_event_ops *ops = &rec->ops; + struct perf_tool *tool = &rec->tool; struct perf_record_opts *opts = &rec->opts; struct perf_evlist *evsel_list = rec->evlist; const char *output_name = rec->output_name; @@ -501,14 +502,14 @@ static int __cmd_record(struct perf_record *rec, int argc, const char **argv) } if (opts->pipe_output) { - err = perf_event__synthesize_attrs(ops, session, + err = perf_event__synthesize_attrs(tool, session, process_synthesized_event); if (err < 0) { pr_err("Couldn't synthesize attrs.\n"); return err; } - err = perf_event__synthesize_event_types(ops, process_synthesized_event, + err = perf_event__synthesize_event_types(tool, process_synthesized_event, machine); if (err < 0) { pr_err("Couldn't synthesize event_types.\n"); @@ -524,7 +525,7 @@ static int __cmd_record(struct perf_record *rec, int argc, const char **argv) * return this more properly and also * propagate errors that now are calling die() */ - err = perf_event__synthesize_tracing_data(ops, output, evsel_list, + err = perf_event__synthesize_tracing_data(tool, output, evsel_list, process_synthesized_event); if (err <= 0) { pr_err("Couldn't record tracing data.\n"); @@ -534,17 +535,17 @@ static int __cmd_record(struct perf_record *rec, int argc, const char **argv) } } - err = perf_event__synthesize_kernel_mmap(ops, process_synthesized_event, + err = perf_event__synthesize_kernel_mmap(tool, process_synthesized_event, machine, "_text"); if (err < 0) - err = perf_event__synthesize_kernel_mmap(ops, process_synthesized_event, + err = perf_event__synthesize_kernel_mmap(tool, process_synthesized_event, machine, "_stext"); if (err < 0) pr_err("Couldn't record kernel reference relocation symbol\n" "Symbol resolution may be skewed if relocation was used (e.g. kexec).\n" "Check /proc/kallsyms permission or run as root.\n"); - err = perf_event__synthesize_modules(ops, process_synthesized_event, + err = perf_event__synthesize_modules(tool, process_synthesized_event, machine); if (err < 0) pr_err("Couldn't record kernel module information.\n" @@ -552,15 +553,15 @@ static int __cmd_record(struct perf_record *rec, int argc, const char **argv) "Check /proc/modules permission or run as root.\n"); if (perf_guest) - perf_session__process_machines(session, ops, + perf_session__process_machines(session, tool, perf_event__synthesize_guest_os); if (!opts->system_wide) - perf_event__synthesize_thread_map(ops, evsel_list->threads, + perf_event__synthesize_thread_map(tool, evsel_list->threads, process_synthesized_event, machine); else - perf_event__synthesize_threads(ops, process_synthesized_event, + perf_event__synthesize_threads(tool, process_synthesized_event, machine); if (rec->realtime_prio) { diff --git a/tools/perf/builtin-report.c b/tools/perf/builtin-report.c index ea64fbbdff43..eef8e423deb0 100644 --- a/tools/perf/builtin-report.c +++ b/tools/perf/builtin-report.c @@ -25,6 +25,7 @@ #include "util/evsel.h" #include "util/header.h" #include "util/session.h" +#include "util/tool.h" #include "util/parse-options.h" #include "util/parse-events.h" @@ -36,7 +37,7 @@ #include struct perf_report { - struct perf_event_ops ops; + struct perf_tool tool; struct perf_session *session; char const *input_name; bool force, use_tui, use_stdio; @@ -103,13 +104,13 @@ out: } -static int process_sample_event(struct perf_event_ops *ops, +static int process_sample_event(struct perf_tool *tool, union perf_event *event, struct perf_sample *sample, struct perf_evsel *evsel, struct machine *machine) { - struct perf_report *rep = container_of(ops, struct perf_report, ops); + struct perf_report *rep = container_of(tool, struct perf_report, tool); struct addr_location al; if (perf_event__preprocess_sample(event, machine, &al, sample, @@ -136,13 +137,13 @@ static int process_sample_event(struct perf_event_ops *ops, return 0; } -static int process_read_event(struct perf_event_ops *ops, +static int process_read_event(struct perf_tool *tool, union perf_event *event, struct perf_sample *sample __used, struct perf_evsel *evsel, struct machine *machine __used) { - struct perf_report *rep = container_of(ops, struct perf_report, ops); + struct perf_report *rep = container_of(tool, struct perf_report, tool); if (rep->show_threads) { const char *name = evsel ? event_name(evsel) : "unknown"; @@ -254,7 +255,7 @@ static int __cmd_report(struct perf_report *rep) signal(SIGINT, sig_handler); session = perf_session__new(rep->input_name, O_RDONLY, - rep->force, false, &rep->ops); + rep->force, false, &rep->tool); if (session == NULL) return -ENOMEM; @@ -277,7 +278,7 @@ static int __cmd_report(struct perf_report *rep) if (ret) goto out_delete; - ret = perf_session__process_events(session, &rep->ops); + ret = perf_session__process_events(session, &rep->tool); if (ret) goto out_delete; @@ -435,7 +436,7 @@ int cmd_report(int argc, const char **argv, const char *prefix __used) NULL }; struct perf_report report = { - .ops = { + .tool = { .sample = process_sample_event, .mmap = perf_event__process_mmap, .comm = perf_event__process_comm, diff --git a/tools/perf/builtin-sched.c b/tools/perf/builtin-sched.c index 6a771f822e5d..0ee868e6f63b 100644 --- a/tools/perf/builtin-sched.c +++ b/tools/perf/builtin-sched.c @@ -8,6 +8,7 @@ #include "util/thread.h" #include "util/header.h" #include "util/session.h" +#include "util/tool.h" #include "util/parse-options.h" #include "util/trace-event.h" @@ -1602,7 +1603,7 @@ static void process_raw_event(union perf_event *raw_event __used, process_sched_migrate_task_event(data, machine, event, cpu, timestamp, thread); } -static int process_sample_event(struct perf_event_ops *ops __used, +static int process_sample_event(struct perf_tool *tool __used, union perf_event *event, struct perf_sample *sample, struct perf_evsel *evsel, @@ -1631,7 +1632,7 @@ static int process_sample_event(struct perf_event_ops *ops __used, return 0; } -static struct perf_event_ops event_ops = { +static struct perf_tool perf_sched = { .sample = process_sample_event, .comm = perf_event__process_comm, .lost = perf_event__process_lost, @@ -1643,12 +1644,12 @@ static void read_events(bool destroy, struct perf_session **psession) { int err = -EINVAL; struct perf_session *session = perf_session__new(input_name, O_RDONLY, - 0, false, &event_ops); + 0, false, &perf_sched); if (session == NULL) die("No Memory"); if (perf_session__has_traces(session, "record -R")) { - err = perf_session__process_events(session, &event_ops); + err = perf_session__process_events(session, &perf_sched); if (err) die("Failed to process events, error %d", err); diff --git a/tools/perf/builtin-script.c b/tools/perf/builtin-script.c index 31a8d14e5fb7..5f8afc65d5f3 100644 --- a/tools/perf/builtin-script.c +++ b/tools/perf/builtin-script.c @@ -7,6 +7,7 @@ #include "util/header.h" #include "util/parse-options.h" #include "util/session.h" +#include "util/tool.h" #include "util/symbol.h" #include "util/thread.h" #include "util/trace-event.h" @@ -434,7 +435,7 @@ static int cleanup_scripting(void) static char const *input_name = "perf.data"; -static int process_sample_event(struct perf_event_ops *ops __used, +static int process_sample_event(struct perf_tool *tool __used, union perf_event *event, struct perf_sample *sample, struct perf_evsel *evsel, @@ -468,7 +469,7 @@ static int process_sample_event(struct perf_event_ops *ops __used, return 0; } -static struct perf_event_ops event_ops = { +static struct perf_tool perf_script = { .sample = process_sample_event, .mmap = perf_event__process_mmap, .comm = perf_event__process_comm, @@ -495,7 +496,7 @@ static int __cmd_script(struct perf_session *session) signal(SIGINT, sig_handler); - ret = perf_session__process_events(session, &event_ops); + ret = perf_session__process_events(session, &perf_script); if (debug_mode) pr_err("Misordered timestamps: %" PRIu64 "\n", nr_unordered); @@ -1262,7 +1263,7 @@ int cmd_script(int argc, const char **argv, const char *prefix __used) if (!script_name) setup_pager(); - session = perf_session__new(input_name, O_RDONLY, 0, false, &event_ops); + session = perf_session__new(input_name, O_RDONLY, 0, false, &perf_script); if (session == NULL) return -ENOMEM; diff --git a/tools/perf/builtin-timechart.c b/tools/perf/builtin-timechart.c index 8e6539625bc1..135376a37f97 100644 --- a/tools/perf/builtin-timechart.c +++ b/tools/perf/builtin-timechart.c @@ -32,6 +32,7 @@ #include "util/event.h" #include "util/session.h" #include "util/svghelper.h" +#include "util/tool.h" #define SUPPORT_OLD_POWER_EVENTS 1 #define PWR_EVENT_EXIT -1 @@ -274,7 +275,7 @@ static int cpus_cstate_state[MAX_CPUS]; static u64 cpus_pstate_start_times[MAX_CPUS]; static u64 cpus_pstate_state[MAX_CPUS]; -static int process_comm_event(struct perf_event_ops *ops __used, +static int process_comm_event(struct perf_tool *tool __used, union perf_event *event, struct perf_sample *sample __used, struct machine *machine __used) @@ -283,7 +284,7 @@ static int process_comm_event(struct perf_event_ops *ops __used, return 0; } -static int process_fork_event(struct perf_event_ops *ops __used, +static int process_fork_event(struct perf_tool *tool __used, union perf_event *event, struct perf_sample *sample __used, struct machine *machine __used) @@ -292,7 +293,7 @@ static int process_fork_event(struct perf_event_ops *ops __used, return 0; } -static int process_exit_event(struct perf_event_ops *ops __used, +static int process_exit_event(struct perf_tool *tool __used, union perf_event *event, struct perf_sample *sample __used, struct machine *machine __used) @@ -490,7 +491,7 @@ static void sched_switch(int cpu, u64 timestamp, struct trace_entry *te) } -static int process_sample_event(struct perf_event_ops *ops __used, +static int process_sample_event(struct perf_tool *tool __used, union perf_event *event __used, struct perf_sample *sample, struct perf_evsel *evsel, @@ -979,7 +980,7 @@ static void write_svg_file(const char *filename) svg_close(); } -static struct perf_event_ops event_ops = { +static struct perf_tool perf_timechart = { .comm = process_comm_event, .fork = process_fork_event, .exit = process_exit_event, @@ -990,7 +991,7 @@ static struct perf_event_ops event_ops = { static int __cmd_timechart(void) { struct perf_session *session = perf_session__new(input_name, O_RDONLY, - 0, false, &event_ops); + 0, false, &perf_timechart); int ret = -EINVAL; if (session == NULL) @@ -999,7 +1000,7 @@ static int __cmd_timechart(void) if (!perf_session__has_traces(session, "timechart record")) goto out_delete; - ret = perf_session__process_events(session, &event_ops); + ret = perf_session__process_events(session, &perf_timechart); if (ret) goto out_delete; diff --git a/tools/perf/builtin-top.c b/tools/perf/builtin-top.c index 31d497368ccf..42a7d96b4dbe 100644 --- a/tools/perf/builtin-top.c +++ b/tools/perf/builtin-top.c @@ -826,7 +826,7 @@ static void perf_session__mmap_read_idx(struct perf_session *self, int idx) perf_event__process_sample(event, evsel, &sample, machine); else if (event->header.type < PERF_RECORD_MAX) { hists__inc_nr_events(&evsel->hists, event->header.type); - perf_event__process(&top.ops, event, &sample, machine); + perf_event__process(&top.tool, event, &sample, machine); } else ++self->hists.stats.nr_unknown_events; } @@ -968,11 +968,11 @@ static int __cmd_top(void) goto out_delete; if (top.target_tid != -1) - perf_event__synthesize_thread_map(&top.ops, top.evlist->threads, + perf_event__synthesize_thread_map(&top.tool, top.evlist->threads, perf_event__process, &top.session->host_machine); else - perf_event__synthesize_threads(&top.ops, perf_event__process, + perf_event__synthesize_threads(&top.tool, perf_event__process, &top.session->host_machine); start_counters(top.evlist); top.session->evlist = top.evlist; diff --git a/tools/perf/util/build-id.c b/tools/perf/util/build-id.c index 2f84c4802aca..dff9c7a725f4 100644 --- a/tools/perf/util/build-id.c +++ b/tools/perf/util/build-id.c @@ -14,8 +14,9 @@ #include #include "debug.h" #include "session.h" +#include "tool.h" -static int build_id__mark_dso_hit(struct perf_event_ops *ops __used, +static int build_id__mark_dso_hit(struct perf_tool *tool __used, union perf_event *event, struct perf_sample *sample __used, struct perf_evsel *evsel __used, @@ -40,7 +41,7 @@ static int build_id__mark_dso_hit(struct perf_event_ops *ops __used, return 0; } -static int perf_event__exit_del_thread(struct perf_event_ops *ops __used, +static int perf_event__exit_del_thread(struct perf_tool *tool __used, union perf_event *event, struct perf_sample *sample __used, struct machine *machine) @@ -59,7 +60,7 @@ static int perf_event__exit_del_thread(struct perf_event_ops *ops __used, return 0; } -struct perf_event_ops build_id__mark_dso_hit_ops = { +struct perf_tool build_id__mark_dso_hit_ops = { .sample = build_id__mark_dso_hit, .mmap = perf_event__process_mmap, .fork = perf_event__process_task, diff --git a/tools/perf/util/build-id.h b/tools/perf/util/build-id.h index 5dafb00eaa06..a993ba87d996 100644 --- a/tools/perf/util/build-id.h +++ b/tools/perf/util/build-id.h @@ -3,7 +3,7 @@ #include "session.h" -extern struct perf_event_ops build_id__mark_dso_hit_ops; +extern struct perf_tool build_id__mark_dso_hit_ops; char *dso__build_id_filename(struct dso *self, char *bf, size_t size); diff --git a/tools/perf/util/event.c b/tools/perf/util/event.c index 0cdc811c48e2..0ebbe7641335 100644 --- a/tools/perf/util/event.c +++ b/tools/perf/util/event.c @@ -43,7 +43,7 @@ static struct perf_sample synth_sample = { .period = 1, }; -static pid_t perf_event__synthesize_comm(struct perf_event_ops *ops, +static pid_t perf_event__synthesize_comm(struct perf_tool *tool, union perf_event *event, pid_t pid, int full, perf_event__handler_t process, struct machine *machine) @@ -99,7 +99,7 @@ out_race: if (!full) { event->comm.tid = pid; - process(ops, event, &synth_sample, machine); + process(tool, event, &synth_sample, machine); goto out; } @@ -117,7 +117,7 @@ out_race: event->comm.tid = pid; - process(ops, event, &synth_sample, machine); + process(tool, event, &synth_sample, machine); } closedir(tasks); @@ -127,7 +127,7 @@ out: return tgid; } -static int perf_event__synthesize_mmap_events(struct perf_event_ops *ops, +static int perf_event__synthesize_mmap_events(struct perf_tool *tool, union perf_event *event, pid_t pid, pid_t tgid, perf_event__handler_t process, @@ -199,7 +199,7 @@ static int perf_event__synthesize_mmap_events(struct perf_event_ops *ops, event->mmap.pid = tgid; event->mmap.tid = pid; - process(ops, event, &synth_sample, machine); + process(tool, event, &synth_sample, machine); } } @@ -207,7 +207,7 @@ static int perf_event__synthesize_mmap_events(struct perf_event_ops *ops, return 0; } -int perf_event__synthesize_modules(struct perf_event_ops *ops, +int perf_event__synthesize_modules(struct perf_tool *tool, perf_event__handler_t process, struct machine *machine) { @@ -252,7 +252,7 @@ int perf_event__synthesize_modules(struct perf_event_ops *ops, memcpy(event->mmap.filename, pos->dso->long_name, pos->dso->long_name_len + 1); - process(ops, event, &synth_sample, machine); + process(tool, event, &synth_sample, machine); } free(event); @@ -262,18 +262,18 @@ int perf_event__synthesize_modules(struct perf_event_ops *ops, static int __event__synthesize_thread(union perf_event *comm_event, union perf_event *mmap_event, pid_t pid, perf_event__handler_t process, - struct perf_event_ops *ops, + struct perf_tool *tool, struct machine *machine) { - pid_t tgid = perf_event__synthesize_comm(ops, comm_event, pid, 1, + pid_t tgid = perf_event__synthesize_comm(tool, comm_event, pid, 1, process, machine); if (tgid == -1) return -1; - return perf_event__synthesize_mmap_events(ops, mmap_event, pid, tgid, + return perf_event__synthesize_mmap_events(tool, mmap_event, pid, tgid, process, machine); } -int perf_event__synthesize_thread_map(struct perf_event_ops *ops, +int perf_event__synthesize_thread_map(struct perf_tool *tool, struct thread_map *threads, perf_event__handler_t process, struct machine *machine) @@ -293,7 +293,7 @@ int perf_event__synthesize_thread_map(struct perf_event_ops *ops, for (thread = 0; thread < threads->nr; ++thread) { if (__event__synthesize_thread(comm_event, mmap_event, threads->map[thread], - process, ops, machine)) { + process, tool, machine)) { err = -1; break; } @@ -305,7 +305,7 @@ out: return err; } -int perf_event__synthesize_threads(struct perf_event_ops *ops, +int perf_event__synthesize_threads(struct perf_tool *tool, perf_event__handler_t process, struct machine *machine) { @@ -334,7 +334,7 @@ int perf_event__synthesize_threads(struct perf_event_ops *ops, continue; __event__synthesize_thread(comm_event, mmap_event, pid, - process, ops, machine); + process, tool, machine); } closedir(proc); @@ -369,7 +369,7 @@ static int find_symbol_cb(void *arg, const char *name, char type, return 1; } -int perf_event__synthesize_kernel_mmap(struct perf_event_ops *ops, +int perf_event__synthesize_kernel_mmap(struct perf_tool *tool, perf_event__handler_t process, struct machine *machine, const char *symbol_name) @@ -427,13 +427,13 @@ int perf_event__synthesize_kernel_mmap(struct perf_event_ops *ops, event->mmap.len = map->end - event->mmap.start; event->mmap.pid = machine->pid; - err = process(ops, event, &synth_sample, machine); + err = process(tool, event, &synth_sample, machine); free(event); return err; } -int perf_event__process_comm(struct perf_event_ops *ops __used, +int perf_event__process_comm(struct perf_tool *tool __used, union perf_event *event, struct perf_sample *sample __used, struct machine *machine) @@ -450,7 +450,7 @@ int perf_event__process_comm(struct perf_event_ops *ops __used, return 0; } -int perf_event__process_lost(struct perf_event_ops *ops __used, +int perf_event__process_lost(struct perf_tool *tool __used, union perf_event *event, struct perf_sample *sample __used, struct machine *machine __used) @@ -473,7 +473,7 @@ static void perf_event__set_kernel_mmap_len(union perf_event *event, maps[MAP__FUNCTION]->end = ~0ULL; } -static int perf_event__process_kernel_mmap(struct perf_event_ops *ops __used, +static int perf_event__process_kernel_mmap(struct perf_tool *tool __used, union perf_event *event, struct machine *machine) { @@ -566,7 +566,7 @@ out_problem: return -1; } -int perf_event__process_mmap(struct perf_event_ops *ops, +int perf_event__process_mmap(struct perf_tool *tool, union perf_event *event, struct perf_sample *sample __used, struct machine *machine) @@ -582,7 +582,7 @@ int perf_event__process_mmap(struct perf_event_ops *ops, if (cpumode == PERF_RECORD_MISC_GUEST_KERNEL || cpumode == PERF_RECORD_MISC_KERNEL) { - ret = perf_event__process_kernel_mmap(ops, event, machine); + ret = perf_event__process_kernel_mmap(tool, event, machine); if (ret < 0) goto out_problem; return 0; @@ -606,7 +606,7 @@ out_problem: return 0; } -int perf_event__process_task(struct perf_event_ops *ops __used, +int perf_event__process_task(struct perf_tool *tool __used, union perf_event *event, struct perf_sample *sample __used, struct machine *machine) @@ -631,22 +631,22 @@ int perf_event__process_task(struct perf_event_ops *ops __used, return 0; } -int perf_event__process(struct perf_event_ops *ops, union perf_event *event, +int perf_event__process(struct perf_tool *tool, union perf_event *event, struct perf_sample *sample, struct machine *machine) { switch (event->header.type) { case PERF_RECORD_COMM: - perf_event__process_comm(ops, event, sample, machine); + perf_event__process_comm(tool, event, sample, machine); break; case PERF_RECORD_MMAP: - perf_event__process_mmap(ops, event, sample, machine); + perf_event__process_mmap(tool, event, sample, machine); break; case PERF_RECORD_FORK: case PERF_RECORD_EXIT: - perf_event__process_task(ops, event, sample, machine); + perf_event__process_task(tool, event, sample, machine); break; case PERF_RECORD_LOST: - perf_event__process_lost(ops, event, sample, machine); + perf_event__process_lost(tool, event, sample, machine); default: break; } diff --git a/tools/perf/util/event.h b/tools/perf/util/event.h index 1564877e8703..d8499e7cf641 100644 --- a/tools/perf/util/event.h +++ b/tools/perf/util/event.h @@ -141,47 +141,47 @@ union perf_event { void perf_event__print_totals(void); -struct perf_event_ops; +struct perf_tool; struct thread_map; -typedef int (*perf_event__handler_t)(struct perf_event_ops *ops, +typedef int (*perf_event__handler_t)(struct perf_tool *tool, union perf_event *event, struct perf_sample *sample, struct machine *machine); -int perf_event__synthesize_thread_map(struct perf_event_ops *ops, +int perf_event__synthesize_thread_map(struct perf_tool *tool, struct thread_map *threads, perf_event__handler_t process, struct machine *machine); -int perf_event__synthesize_threads(struct perf_event_ops *ops, +int perf_event__synthesize_threads(struct perf_tool *tool, perf_event__handler_t process, struct machine *machine); -int perf_event__synthesize_kernel_mmap(struct perf_event_ops *ops, +int perf_event__synthesize_kernel_mmap(struct perf_tool *tool, perf_event__handler_t process, struct machine *machine, const char *symbol_name); -int perf_event__synthesize_modules(struct perf_event_ops *ops, +int perf_event__synthesize_modules(struct perf_tool *tool, perf_event__handler_t process, struct machine *machine); -int perf_event__process_comm(struct perf_event_ops *ops, +int perf_event__process_comm(struct perf_tool *tool, union perf_event *event, struct perf_sample *sample, struct machine *machine); -int perf_event__process_lost(struct perf_event_ops *ops, +int perf_event__process_lost(struct perf_tool *tool, union perf_event *event, struct perf_sample *sample, struct machine *machine); -int perf_event__process_mmap(struct perf_event_ops *ops, +int perf_event__process_mmap(struct perf_tool *tool, union perf_event *event, struct perf_sample *sample, struct machine *machine); -int perf_event__process_task(struct perf_event_ops *ops, +int perf_event__process_task(struct perf_tool *tool, union perf_event *event, struct perf_sample *sample, struct machine *machine); -int perf_event__process(struct perf_event_ops *ops, +int perf_event__process(struct perf_tool *tool, union perf_event *event, struct perf_sample *sample, struct machine *machine); diff --git a/tools/perf/util/header.c b/tools/perf/util/header.c index db280d6ca898..9272f3a20cac 100644 --- a/tools/perf/util/header.c +++ b/tools/perf/util/header.c @@ -2070,7 +2070,7 @@ out_delete_evlist: return -ENOMEM; } -int perf_event__synthesize_attr(struct perf_event_ops *ops, +int perf_event__synthesize_attr(struct perf_tool *tool, struct perf_event_attr *attr, u16 ids, u64 *id, perf_event__handler_t process) { @@ -2094,14 +2094,14 @@ int perf_event__synthesize_attr(struct perf_event_ops *ops, ev->attr.header.type = PERF_RECORD_HEADER_ATTR; ev->attr.header.size = size; - err = process(ops, ev, NULL, NULL); + err = process(tool, ev, NULL, NULL); free(ev); return err; } -int perf_event__synthesize_attrs(struct perf_event_ops *ops, +int perf_event__synthesize_attrs(struct perf_tool *tool, struct perf_session *session, perf_event__handler_t process) { @@ -2109,7 +2109,7 @@ int perf_event__synthesize_attrs(struct perf_event_ops *ops, int err = 0; list_for_each_entry(attr, &session->evlist->entries, node) { - err = perf_event__synthesize_attr(ops, &attr->attr, attr->ids, + err = perf_event__synthesize_attr(tool, &attr->attr, attr->ids, attr->id, process); if (err) { pr_debug("failed to create perf header attribute\n"); @@ -2157,7 +2157,7 @@ int perf_event__process_attr(union perf_event *event, return 0; } -int perf_event__synthesize_event_type(struct perf_event_ops *ops, +int perf_event__synthesize_event_type(struct perf_tool *tool, u64 event_id, char *name, perf_event__handler_t process, struct machine *machine) @@ -2178,12 +2178,12 @@ int perf_event__synthesize_event_type(struct perf_event_ops *ops, ev.event_type.header.size = sizeof(ev.event_type) - (sizeof(ev.event_type.event_type.name) - size); - err = process(ops, &ev, NULL, machine); + err = process(tool, &ev, NULL, machine); return err; } -int perf_event__synthesize_event_types(struct perf_event_ops *ops, +int perf_event__synthesize_event_types(struct perf_tool *tool, perf_event__handler_t process, struct machine *machine) { @@ -2193,7 +2193,7 @@ int perf_event__synthesize_event_types(struct perf_event_ops *ops, for (i = 0; i < event_count; i++) { type = &events[i]; - err = perf_event__synthesize_event_type(ops, type->event_id, + err = perf_event__synthesize_event_type(tool, type->event_id, type->name, process, machine); if (err) { @@ -2205,7 +2205,7 @@ int perf_event__synthesize_event_types(struct perf_event_ops *ops, return err; } -int perf_event__process_event_type(struct perf_event_ops *ops __unused, +int perf_event__process_event_type(struct perf_tool *tool __unused, union perf_event *event) { if (perf_header__push_event(event->event_type.event_type.event_id, @@ -2215,7 +2215,7 @@ int perf_event__process_event_type(struct perf_event_ops *ops __unused, return 0; } -int perf_event__synthesize_tracing_data(struct perf_event_ops *ops, int fd, +int perf_event__synthesize_tracing_data(struct perf_tool *tool, int fd, struct perf_evlist *evlist, perf_event__handler_t process) { @@ -2248,7 +2248,7 @@ int perf_event__synthesize_tracing_data(struct perf_event_ops *ops, int fd, ev.tracing_data.header.size = sizeof(ev.tracing_data); ev.tracing_data.size = aligned_size; - process(ops, &ev, NULL, NULL); + process(tool, &ev, NULL, NULL); /* * The put function will copy all the tracing data @@ -2290,7 +2290,7 @@ int perf_event__process_tracing_data(union perf_event *event, return size_read + padding; } -int perf_event__synthesize_build_id(struct perf_event_ops *ops, +int perf_event__synthesize_build_id(struct perf_tool *tool, struct dso *pos, u16 misc, perf_event__handler_t process, struct machine *machine) @@ -2313,12 +2313,12 @@ int perf_event__synthesize_build_id(struct perf_event_ops *ops, ev.build_id.header.size = sizeof(ev.build_id) + len; memcpy(&ev.build_id.filename, pos->long_name, pos->long_name_len); - err = process(ops, &ev, NULL, machine); + err = process(tool, &ev, NULL, machine); return err; } -int perf_event__process_build_id(struct perf_event_ops *ops __used, +int perf_event__process_build_id(struct perf_tool *tool __used, union perf_event *event, struct perf_session *session) { diff --git a/tools/perf/util/header.h b/tools/perf/util/header.h index a604962fc431..09365b32098e 100644 --- a/tools/perf/util/header.h +++ b/tools/perf/util/header.h @@ -97,35 +97,35 @@ int build_id_cache__add_s(const char *sbuild_id, const char *debugdir, const char *name, bool is_kallsyms); int build_id_cache__remove_s(const char *sbuild_id, const char *debugdir); -int perf_event__synthesize_attr(struct perf_event_ops *ops, +int perf_event__synthesize_attr(struct perf_tool *tool, struct perf_event_attr *attr, u16 ids, u64 *id, perf_event__handler_t process); -int perf_event__synthesize_attrs(struct perf_event_ops *ops, +int perf_event__synthesize_attrs(struct perf_tool *tool, struct perf_session *session, perf_event__handler_t process); int perf_event__process_attr(union perf_event *event, struct perf_evlist **pevlist); -int perf_event__synthesize_event_type(struct perf_event_ops *ops, +int perf_event__synthesize_event_type(struct perf_tool *tool, u64 event_id, char *name, perf_event__handler_t process, struct machine *machine); -int perf_event__synthesize_event_types(struct perf_event_ops *ops, +int perf_event__synthesize_event_types(struct perf_tool *tool, perf_event__handler_t process, struct machine *machine); -int perf_event__process_event_type(struct perf_event_ops *ops, +int perf_event__process_event_type(struct perf_tool *tool, union perf_event *event); -int perf_event__synthesize_tracing_data(struct perf_event_ops *ops, +int perf_event__synthesize_tracing_data(struct perf_tool *tool, int fd, struct perf_evlist *evlist, perf_event__handler_t process); int perf_event__process_tracing_data(union perf_event *event, struct perf_session *session); -int perf_event__synthesize_build_id(struct perf_event_ops *ops, +int perf_event__synthesize_build_id(struct perf_tool *tool, struct dso *pos, u16 misc, perf_event__handler_t process, struct machine *machine); -int perf_event__process_build_id(struct perf_event_ops *ops, +int perf_event__process_build_id(struct perf_tool *tool, union perf_event *event, struct perf_session *session); diff --git a/tools/perf/util/session.c b/tools/perf/util/session.c index be33606386bf..7d159088c4ac 100644 --- a/tools/perf/util/session.c +++ b/tools/perf/util/session.c @@ -10,6 +10,7 @@ #include "evlist.h" #include "evsel.h" #include "session.h" +#include "tool.h" #include "sort.h" #include "util.h" #include "cpumap.h" @@ -104,7 +105,7 @@ static void perf_session__destroy_kernel_maps(struct perf_session *self) struct perf_session *perf_session__new(const char *filename, int mode, bool force, bool repipe, - struct perf_event_ops *ops) + struct perf_tool *tool) { size_t len = filename ? strlen(filename) + 1 : 0; struct perf_session *self = zalloc(sizeof(*self) + len); @@ -142,10 +143,10 @@ struct perf_session *perf_session__new(const char *filename, int mode, goto out_delete; } - if (ops && ops->ordering_requires_timestamps && - ops->ordered_samples && !self->sample_id_all) { + if (tool && tool->ordering_requires_timestamps && + tool->ordered_samples && !self->sample_id_all) { dump_printf("WARNING: No sample_id_all support, falling back to unordered processing\n"); - ops->ordered_samples = false; + tool->ordered_samples = false; } out: @@ -285,7 +286,7 @@ static int process_event_synth_attr_stub(union perf_event *event __used, return 0; } -static int process_event_sample_stub(struct perf_event_ops *ops __used, +static int process_event_sample_stub(struct perf_tool *tool __used, union perf_event *event __used, struct perf_sample *sample __used, struct perf_evsel *evsel __used, @@ -295,7 +296,7 @@ static int process_event_sample_stub(struct perf_event_ops *ops __used, return 0; } -static int process_event_stub(struct perf_event_ops *ops __used, +static int process_event_stub(struct perf_tool *tool __used, union perf_event *event __used, struct perf_sample *sample __used, struct machine *machine __used) @@ -304,7 +305,7 @@ static int process_event_stub(struct perf_event_ops *ops __used, return 0; } -static int process_finished_round_stub(struct perf_event_ops *ops __used, +static int process_finished_round_stub(struct perf_tool *tool __used, union perf_event *event __used, struct perf_session *perf_session __used) { @@ -312,50 +313,50 @@ static int process_finished_round_stub(struct perf_event_ops *ops __used, return 0; } -static int process_event_type_stub(struct perf_event_ops *ops __used, +static int process_event_type_stub(struct perf_tool *tool __used, union perf_event *event __used) { dump_printf(": unhandled!\n"); return 0; } -static int process_finished_round(struct perf_event_ops *ops, +static int process_finished_round(struct perf_tool *tool, union perf_event *event, struct perf_session *session); -static void perf_event_ops__fill_defaults(struct perf_event_ops *handler) +static void perf_tool__fill_defaults(struct perf_tool *tool) { - if (handler->sample == NULL) - handler->sample = process_event_sample_stub; - if (handler->mmap == NULL) - handler->mmap = process_event_stub; - if (handler->comm == NULL) - handler->comm = process_event_stub; - if (handler->fork == NULL) - handler->fork = process_event_stub; - if (handler->exit == NULL) - handler->exit = process_event_stub; - if (handler->lost == NULL) - handler->lost = perf_event__process_lost; - if (handler->read == NULL) - handler->read = process_event_sample_stub; - if (handler->throttle == NULL) - handler->throttle = process_event_stub; - if (handler->unthrottle == NULL) - handler->unthrottle = process_event_stub; - if (handler->attr == NULL) - handler->attr = process_event_synth_attr_stub; - if (handler->event_type == NULL) - handler->event_type = process_event_type_stub; - if (handler->tracing_data == NULL) - handler->tracing_data = process_event_synth_tracing_data_stub; - if (handler->build_id == NULL) - handler->build_id = process_finished_round_stub; - if (handler->finished_round == NULL) { - if (handler->ordered_samples) - handler->finished_round = process_finished_round; + if (tool->sample == NULL) + tool->sample = process_event_sample_stub; + if (tool->mmap == NULL) + tool->mmap = process_event_stub; + if (tool->comm == NULL) + tool->comm = process_event_stub; + if (tool->fork == NULL) + tool->fork = process_event_stub; + if (tool->exit == NULL) + tool->exit = process_event_stub; + if (tool->lost == NULL) + tool->lost = perf_event__process_lost; + if (tool->read == NULL) + tool->read = process_event_sample_stub; + if (tool->throttle == NULL) + tool->throttle = process_event_stub; + if (tool->unthrottle == NULL) + tool->unthrottle = process_event_stub; + if (tool->attr == NULL) + tool->attr = process_event_synth_attr_stub; + if (tool->event_type == NULL) + tool->event_type = process_event_type_stub; + if (tool->tracing_data == NULL) + tool->tracing_data = process_event_synth_tracing_data_stub; + if (tool->build_id == NULL) + tool->build_id = process_finished_round_stub; + if (tool->finished_round == NULL) { + if (tool->ordered_samples) + tool->finished_round = process_finished_round; else - handler->finished_round = process_finished_round_stub; + tool->finished_round = process_finished_round_stub; } } @@ -487,11 +488,11 @@ static void perf_session_free_sample_buffers(struct perf_session *session) static int perf_session_deliver_event(struct perf_session *session, union perf_event *event, struct perf_sample *sample, - struct perf_event_ops *ops, + struct perf_tool *tool, u64 file_offset); static void flush_sample_queue(struct perf_session *s, - struct perf_event_ops *ops) + struct perf_tool *tool) { struct ordered_samples *os = &s->ordered_samples; struct list_head *head = &os->samples; @@ -502,7 +503,7 @@ static void flush_sample_queue(struct perf_session *s, unsigned idx = 0, progress_next = os->nr_samples / 16; int ret; - if (!ops->ordered_samples || !limit) + if (!tool->ordered_samples || !limit) return; list_for_each_entry_safe(iter, tmp, head, list) { @@ -513,7 +514,7 @@ static void flush_sample_queue(struct perf_session *s, if (ret) pr_err("Can't parse sample, err = %d\n", ret); else - perf_session_deliver_event(s, iter->event, &sample, ops, + perf_session_deliver_event(s, iter->event, &sample, tool, iter->file_offset); os->last_flush = iter->timestamp; @@ -575,11 +576,11 @@ static void flush_sample_queue(struct perf_session *s, * Flush every events below timestamp 7 * etc... */ -static int process_finished_round(struct perf_event_ops *ops, +static int process_finished_round(struct perf_tool *tool, union perf_event *event __used, struct perf_session *session) { - flush_sample_queue(session, ops); + flush_sample_queue(session, tool); session->ordered_samples.next_flush = session->ordered_samples.max_timestamp; return 0; @@ -749,7 +750,7 @@ static struct machine * static int perf_session_deliver_event(struct perf_session *session, union perf_event *event, struct perf_sample *sample, - struct perf_event_ops *ops, + struct perf_tool *tool, u64 file_offset) { struct perf_evsel *evsel; @@ -784,25 +785,25 @@ static int perf_session_deliver_event(struct perf_session *session, ++session->hists.stats.nr_unknown_id; return -1; } - return ops->sample(ops, event, sample, evsel, machine); + return tool->sample(tool, event, sample, evsel, machine); case PERF_RECORD_MMAP: - return ops->mmap(ops, event, sample, machine); + return tool->mmap(tool, event, sample, machine); case PERF_RECORD_COMM: - return ops->comm(ops, event, sample, machine); + return tool->comm(tool, event, sample, machine); case PERF_RECORD_FORK: - return ops->fork(ops, event, sample, machine); + return tool->fork(tool, event, sample, machine); case PERF_RECORD_EXIT: - return ops->exit(ops, event, sample, machine); + return tool->exit(tool, event, sample, machine); case PERF_RECORD_LOST: - if (ops->lost == perf_event__process_lost) + if (tool->lost == perf_event__process_lost) session->hists.stats.total_lost += event->lost.lost; - return ops->lost(ops, event, sample, machine); + return tool->lost(tool, event, sample, machine); case PERF_RECORD_READ: - return ops->read(ops, event, sample, evsel, machine); + return tool->read(tool, event, sample, evsel, machine); case PERF_RECORD_THROTTLE: - return ops->throttle(ops, event, sample, machine); + return tool->throttle(tool, event, sample, machine); case PERF_RECORD_UNTHROTTLE: - return ops->unthrottle(ops, event, sample, machine); + return tool->unthrottle(tool, event, sample, machine); default: ++session->hists.stats.nr_unknown_events; return -1; @@ -826,7 +827,7 @@ static int perf_session__preprocess_sample(struct perf_session *session, } static int perf_session__process_user_event(struct perf_session *session, union perf_event *event, - struct perf_event_ops *ops, u64 file_offset) + struct perf_tool *tool, u64 file_offset) { int err; @@ -835,20 +836,20 @@ static int perf_session__process_user_event(struct perf_session *session, union /* These events are processed right away */ switch (event->header.type) { case PERF_RECORD_HEADER_ATTR: - err = ops->attr(event, &session->evlist); + err = tool->attr(event, &session->evlist); if (err == 0) perf_session__update_sample_type(session); return err; case PERF_RECORD_HEADER_EVENT_TYPE: - return ops->event_type(ops, event); + return tool->event_type(tool, event); case PERF_RECORD_HEADER_TRACING_DATA: /* setup for reading amidst mmap */ lseek(session->fd, file_offset, SEEK_SET); - return ops->tracing_data(event, session); + return tool->tracing_data(event, session); case PERF_RECORD_HEADER_BUILD_ID: - return ops->build_id(ops, event, session); + return tool->build_id(tool, event, session); case PERF_RECORD_FINISHED_ROUND: - return ops->finished_round(ops, event, session); + return tool->finished_round(tool, event, session); default: return -EINVAL; } @@ -856,7 +857,7 @@ static int perf_session__process_user_event(struct perf_session *session, union static int perf_session__process_event(struct perf_session *session, union perf_event *event, - struct perf_event_ops *ops, + struct perf_tool *tool, u64 file_offset) { struct perf_sample sample; @@ -872,7 +873,7 @@ static int perf_session__process_event(struct perf_session *session, hists__inc_nr_events(&session->hists, event->header.type); if (event->header.type >= PERF_RECORD_USER_TYPE_START) - return perf_session__process_user_event(session, event, ops, file_offset); + return perf_session__process_user_event(session, event, tool, file_offset); /* * For all kernel events we get the sample data @@ -885,14 +886,14 @@ static int perf_session__process_event(struct perf_session *session, if (perf_session__preprocess_sample(session, event, &sample)) return 0; - if (ops->ordered_samples) { + if (tool->ordered_samples) { ret = perf_session_queue_event(session, event, &sample, file_offset); if (ret != -ETIME) return ret; } - return perf_session_deliver_event(session, event, &sample, ops, + return perf_session_deliver_event(session, event, &sample, tool, file_offset); } @@ -921,9 +922,9 @@ static struct thread *perf_session__register_idle_thread(struct perf_session *se } static void perf_session__warn_about_errors(const struct perf_session *session, - const struct perf_event_ops *ops) + const struct perf_tool *tool) { - if (ops->lost == perf_event__process_lost && + if (tool->lost == perf_event__process_lost && session->hists.stats.nr_events[PERF_RECORD_LOST] != 0) { ui__warning("Processed %d events and lost %d chunks!\n\n" "Check IO/CPU overload!\n\n", @@ -958,7 +959,7 @@ static void perf_session__warn_about_errors(const struct perf_session *session, volatile int session_done; static int __perf_session__process_pipe_events(struct perf_session *self, - struct perf_event_ops *ops) + struct perf_tool *tool) { union perf_event event; uint32_t size; @@ -967,7 +968,7 @@ static int __perf_session__process_pipe_events(struct perf_session *self, int err; void *p; - perf_event_ops__fill_defaults(ops); + perf_tool__fill_defaults(tool); head = 0; more: @@ -1004,7 +1005,7 @@ more: } if (size == 0 || - (skip = perf_session__process_event(self, &event, ops, head)) < 0) { + (skip = perf_session__process_event(self, &event, tool, head)) < 0) { dump_printf("%#" PRIx64 " [%#x]: skipping unknown header type: %d\n", head, event.header.size, event.header.type); /* @@ -1027,7 +1028,7 @@ more: done: err = 0; out_err: - perf_session__warn_about_errors(self, ops); + perf_session__warn_about_errors(self, tool); perf_session_free_sample_buffers(self); return err; } @@ -1058,7 +1059,7 @@ fetch_mmaped_event(struct perf_session *session, int __perf_session__process_events(struct perf_session *session, u64 data_offset, u64 data_size, - u64 file_size, struct perf_event_ops *ops) + u64 file_size, struct perf_tool *tool) { u64 head, page_offset, file_offset, file_pos, progress_next; int err, mmap_prot, mmap_flags, map_idx = 0; @@ -1067,7 +1068,7 @@ int __perf_session__process_events(struct perf_session *session, union perf_event *event; uint32_t size; - perf_event_ops__fill_defaults(ops); + perf_tool__fill_defaults(tool); page_size = sysconf(_SC_PAGESIZE); @@ -1122,7 +1123,7 @@ more: size = event->header.size; if (size == 0 || - perf_session__process_event(session, event, ops, file_pos) < 0) { + perf_session__process_event(session, event, tool, file_pos) < 0) { dump_printf("%#" PRIx64 " [%#x]: skipping unknown header type: %d\n", file_offset + head, event->header.size, event->header.type); @@ -1151,15 +1152,15 @@ more: err = 0; /* do the final flush for ordered samples */ session->ordered_samples.next_flush = ULLONG_MAX; - flush_sample_queue(session, ops); + flush_sample_queue(session, tool); out_err: - perf_session__warn_about_errors(session, ops); + perf_session__warn_about_errors(session, tool); perf_session_free_sample_buffers(session); return err; } int perf_session__process_events(struct perf_session *self, - struct perf_event_ops *ops) + struct perf_tool *tool) { int err; @@ -1170,9 +1171,9 @@ int perf_session__process_events(struct perf_session *self, err = __perf_session__process_events(self, self->header.data_offset, self->header.data_size, - self->size, ops); + self->size, tool); else - err = __perf_session__process_pipe_events(self, ops); + err = __perf_session__process_pipe_events(self, tool); return err; } diff --git a/tools/perf/util/session.h b/tools/perf/util/session.h index 1c5823c7d6dc..30e9c6b6fc3c 100644 --- a/tools/perf/util/session.h +++ b/tools/perf/util/session.h @@ -53,55 +53,20 @@ struct perf_session { char filename[0]; }; -struct perf_evsel; -struct perf_event_ops; - -typedef int (*event_sample)(struct perf_event_ops *ops, - union perf_event *event, struct perf_sample *sample, - struct perf_evsel *evsel, struct machine *machine); -typedef int (*event_op)(struct perf_event_ops *ops, union perf_event *event, - struct perf_sample *sample, - struct machine *machine); -typedef int (*event_synth_op)(union perf_event *self, - struct perf_session *session); -typedef int (*event_attr_op)(union perf_event *event, - struct perf_evlist **pevlist); -typedef int (*event_simple_op)(struct perf_event_ops *ops, - union perf_event *event); -typedef int (*event_op2)(struct perf_event_ops *ops, union perf_event *event, - struct perf_session *session); - -struct perf_event_ops { - event_sample sample, - read; - event_op mmap, - comm, - fork, - exit, - lost, - throttle, - unthrottle; - event_attr_op attr; - event_synth_op tracing_data; - event_simple_op event_type; - event_op2 finished_round, - build_id; - bool ordered_samples; - bool ordering_requires_timestamps; -}; +struct perf_tool; struct perf_session *perf_session__new(const char *filename, int mode, bool force, bool repipe, - struct perf_event_ops *ops); + struct perf_tool *tool); void perf_session__delete(struct perf_session *self); void perf_event_header__bswap(struct perf_event_header *self); int __perf_session__process_events(struct perf_session *self, u64 data_offset, u64 data_size, u64 size, - struct perf_event_ops *ops); + struct perf_tool *tool); int perf_session__process_events(struct perf_session *self, - struct perf_event_ops *event_ops); + struct perf_tool *tool); int perf_session__resolve_callchain(struct perf_session *self, struct perf_evsel *evsel, struct thread *thread, @@ -142,11 +107,11 @@ struct machine *perf_session__findnew_machine(struct perf_session *self, pid_t p static inline void perf_session__process_machines(struct perf_session *self, - struct perf_event_ops *ops, + struct perf_tool *tool, machine__process_t process) { - process(&self->host_machine, ops); - return machines__process(&self->machines, process, ops); + process(&self->host_machine, tool); + return machines__process(&self->machines, process, tool); } struct thread *perf_session__findnew(struct perf_session *self, pid_t pid); diff --git a/tools/perf/util/tool.h b/tools/perf/util/tool.h new file mode 100644 index 000000000000..89ff1b551a74 --- /dev/null +++ b/tools/perf/util/tool.h @@ -0,0 +1,45 @@ +#ifndef __PERF_TOOL_H +#define __PERF_TOOL_H + +struct perf_session; +struct perf_evsel; +struct perf_tool; +struct machine; + +typedef int (*event_sample)(struct perf_tool *tool, union perf_event *event, + struct perf_sample *sample, + struct perf_evsel *evsel, struct machine *machine); + +typedef int (*event_op)(struct perf_tool *tool, union perf_event *event, + struct perf_sample *sample, struct machine *machine); + +typedef int (*event_attr_op)(union perf_event *event, + struct perf_evlist **pevlist); +typedef int (*event_simple_op)(struct perf_tool *tool, union perf_event *event); + +typedef int (*event_synth_op)(union perf_event *event, + struct perf_session *session); + +typedef int (*event_op2)(struct perf_tool *tool, union perf_event *event, + struct perf_session *session); + +struct perf_tool { + event_sample sample, + read; + event_op mmap, + comm, + fork, + exit, + lost, + throttle, + unthrottle; + event_attr_op attr; + event_synth_op tracing_data; + event_simple_op event_type; + event_op2 finished_round, + build_id; + bool ordered_samples; + bool ordering_requires_timestamps; +}; + +#endif /* __PERF_TOOL_H */ diff --git a/tools/perf/util/top.h b/tools/perf/util/top.h index 44eda6fc6b33..40430ec5c267 100644 --- a/tools/perf/util/top.h +++ b/tools/perf/util/top.h @@ -1,16 +1,17 @@ #ifndef __PERF_TOP_H #define __PERF_TOP_H 1 +#include "tool.h" #include "types.h" -#include "session.h" -#include "../perf.h" #include +#include struct perf_evlist; struct perf_evsel; +struct perf_session; struct perf_top { - struct perf_event_ops ops; + struct perf_tool tool; struct perf_evlist *evlist; /* * Symbols will be added here in perf_event__process_sample and will -- cgit v1.2.3 From 1758af10cf13d156014035b29ff50eab3773d849 Mon Sep 17 00:00:00 2001 From: Arnaldo Carvalho de Melo Date: Mon, 28 Nov 2011 09:37:05 -0200 Subject: perf top: Stop using globals for tool state Use its 'perf_tool' base class instead. Cc: David Ahern Cc: Frederic Weisbecker Cc: Mike Galbraith Cc: Paul Mackerras Cc: Peter Zijlstra Cc: Stephane Eranian Link: http://lkml.kernel.org/n/tip-i33q40wwvk2zna8fd36ex6sm@git.kernel.org Signed-off-by: Arnaldo Carvalho de Melo --- tools/perf/builtin-top.c | 464 +++++++++++++++++++++++------------------------ tools/perf/util/top.h | 16 ++ 2 files changed, 245 insertions(+), 235 deletions(-) diff --git a/tools/perf/builtin-top.c b/tools/perf/builtin-top.c index 42a7d96b4dbe..50ff362ff012 100644 --- a/tools/perf/builtin-top.c +++ b/tools/perf/builtin-top.c @@ -64,44 +64,6 @@ #include #include -static struct perf_top top = { - .count_filter = 5, - .delay_secs = 2, - .target_pid = -1, - .target_tid = -1, - .freq = 1000, /* 1 KHz */ -}; - -static bool system_wide = false; - -static bool use_tui, use_stdio; - -static bool sort_has_symbols; - -static bool dont_use_callchains; -static char callchain_default_opt[] = "fractal,0.5,callee"; - - -static int default_interval = 0; - -static bool kptr_restrict_warned; -static bool vmlinux_warned; -static bool inherit = false; -static int realtime_prio = 0; -static bool group = false; -static bool sample_id_all_avail = true; -static unsigned int mmap_pages = 128; - -static bool dump_symtab = false; - -static struct winsize winsize; - -static const char *sym_filter = NULL; -static int sym_pcnt_filter = 5; - -/* - * Source functions - */ void get_term_dimensions(struct winsize *ws) { @@ -125,21 +87,23 @@ void get_term_dimensions(struct winsize *ws) ws->ws_col = 80; } -static void update_print_entries(struct winsize *ws) +static void perf_top__update_print_entries(struct perf_top *top) { - top.print_entries = ws->ws_row; + top->print_entries = top->winsize.ws_row; - if (top.print_entries > 9) - top.print_entries -= 9; + if (top->print_entries > 9) + top->print_entries -= 9; } -static void sig_winch_handler(int sig __used) +static void perf_top__sig_winch(int sig __used, siginfo_t *info __used, void *arg) { - get_term_dimensions(&winsize); - update_print_entries(&winsize); + struct perf_top *top = arg; + + get_term_dimensions(&top->winsize); + perf_top__update_print_entries(top); } -static int parse_source(struct hist_entry *he) +static int perf_top__parse_source(struct perf_top *top, struct hist_entry *he) { struct symbol *sym; struct annotation *notes; @@ -181,7 +145,7 @@ static int parse_source(struct hist_entry *he) err = symbol__annotate(sym, map, 0); if (err == 0) { out_assign: - top.sym_filter_entry = he; + top->sym_filter_entry = he; } pthread_mutex_unlock(¬es->lock); @@ -194,14 +158,16 @@ static void __zero_source_counters(struct hist_entry *he) symbol__annotate_zero_histograms(sym); } -static void record_precise_ip(struct hist_entry *he, int counter, u64 ip) +static void perf_top__record_precise_ip(struct perf_top *top, + struct hist_entry *he, + int counter, u64 ip) { struct annotation *notes; struct symbol *sym; if (he == NULL || he->ms.sym == NULL || - ((top.sym_filter_entry == NULL || - top.sym_filter_entry->ms.sym != he->ms.sym) && use_browser != 1)) + ((top->sym_filter_entry == NULL || + top->sym_filter_entry->ms.sym != he->ms.sym) && use_browser != 1)) return; sym = he->ms.sym; @@ -224,8 +190,9 @@ static void record_precise_ip(struct hist_entry *he, int counter, u64 ip) pthread_mutex_unlock(¬es->lock); } -static void show_details(struct hist_entry *he) +static void perf_top__show_details(struct perf_top *top) { + struct hist_entry *he = top->sym_filter_entry; struct annotation *notes; struct symbol *symbol; int more; @@ -241,15 +208,15 @@ static void show_details(struct hist_entry *he) if (notes->src == NULL) goto out_unlock; - printf("Showing %s for %s\n", event_name(top.sym_evsel), symbol->name); - printf(" Events Pcnt (>=%d%%)\n", sym_pcnt_filter); + printf("Showing %s for %s\n", event_name(top->sym_evsel), symbol->name); + printf(" Events Pcnt (>=%d%%)\n", top->sym_pcnt_filter); - more = symbol__annotate_printf(symbol, he->ms.map, top.sym_evsel->idx, - 0, sym_pcnt_filter, top.print_entries, 4); - if (top.zero) - symbol__annotate_zero_histogram(symbol, top.sym_evsel->idx); + more = symbol__annotate_printf(symbol, he->ms.map, top->sym_evsel->idx, + 0, top->sym_pcnt_filter, top->print_entries, 4); + if (top->zero) + symbol__annotate_zero_histogram(symbol, top->sym_evsel->idx); else - symbol__annotate_decay_histogram(symbol, top.sym_evsel->idx); + symbol__annotate_decay_histogram(symbol, top->sym_evsel->idx); if (more != 0) printf("%d lines not displayed, maybe increase display entries [e]\n", more); out_unlock: @@ -273,45 +240,46 @@ static struct hist_entry *perf_evsel__add_hist_entry(struct perf_evsel *evsel, return he; } -static void print_sym_table(void) +static void perf_top__print_sym_table(struct perf_top *top) { char bf[160]; int printed = 0; - const int win_width = winsize.ws_col - 1; + const int win_width = top->winsize.ws_col - 1; puts(CONSOLE_CLEAR); - perf_top__header_snprintf(&top, bf, sizeof(bf)); + perf_top__header_snprintf(top, bf, sizeof(bf)); printf("%s\n", bf); - perf_top__reset_sample_counters(&top); + perf_top__reset_sample_counters(top); printf("%-*.*s\n", win_width, win_width, graph_dotted_line); - if (top.sym_evsel->hists.stats.nr_lost_warned != - top.sym_evsel->hists.stats.nr_events[PERF_RECORD_LOST]) { - top.sym_evsel->hists.stats.nr_lost_warned = - top.sym_evsel->hists.stats.nr_events[PERF_RECORD_LOST]; + if (top->sym_evsel->hists.stats.nr_lost_warned != + top->sym_evsel->hists.stats.nr_events[PERF_RECORD_LOST]) { + top->sym_evsel->hists.stats.nr_lost_warned = + top->sym_evsel->hists.stats.nr_events[PERF_RECORD_LOST]; color_fprintf(stdout, PERF_COLOR_RED, "WARNING: LOST %d chunks, Check IO/CPU overload", - top.sym_evsel->hists.stats.nr_lost_warned); + top->sym_evsel->hists.stats.nr_lost_warned); ++printed; } - if (top.sym_filter_entry) { - show_details(top.sym_filter_entry); + if (top->sym_filter_entry) { + perf_top__show_details(top); return; } - hists__collapse_resort_threaded(&top.sym_evsel->hists); - hists__output_resort_threaded(&top.sym_evsel->hists); - hists__decay_entries_threaded(&top.sym_evsel->hists, - top.hide_user_symbols, - top.hide_kernel_symbols); - hists__output_recalc_col_len(&top.sym_evsel->hists, winsize.ws_row - 3); + hists__collapse_resort_threaded(&top->sym_evsel->hists); + hists__output_resort_threaded(&top->sym_evsel->hists); + hists__decay_entries_threaded(&top->sym_evsel->hists, + top->hide_user_symbols, + top->hide_kernel_symbols); + hists__output_recalc_col_len(&top->sym_evsel->hists, + top->winsize.ws_row - 3); putchar('\n'); - hists__fprintf(&top.sym_evsel->hists, NULL, false, false, - winsize.ws_row - 4 - printed, win_width, stdout); + hists__fprintf(&top->sym_evsel->hists, NULL, false, false, + top->winsize.ws_row - 4 - printed, win_width, stdout); } static void prompt_integer(int *target, const char *msg) @@ -349,17 +317,17 @@ static void prompt_percent(int *target, const char *msg) *target = tmp; } -static void prompt_symbol(struct hist_entry **target, const char *msg) +static void perf_top__prompt_symbol(struct perf_top *top, const char *msg) { char *buf = malloc(0), *p; - struct hist_entry *syme = *target, *n, *found = NULL; + struct hist_entry *syme = top->sym_filter_entry, *n, *found = NULL; struct rb_node *next; size_t dummy = 0; /* zero counters of active symbol */ if (syme) { __zero_source_counters(syme); - *target = NULL; + top->sym_filter_entry = NULL; } fprintf(stdout, "\n%s: ", msg); @@ -370,7 +338,7 @@ static void prompt_symbol(struct hist_entry **target, const char *msg) if (p) *p = 0; - next = rb_first(&top.sym_evsel->hists.entries); + next = rb_first(&top->sym_evsel->hists.entries); while (next) { n = rb_entry(next, struct hist_entry, rb_node); if (n->ms.sym && !strcmp(buf, n->ms.sym->name)) { @@ -385,45 +353,45 @@ static void prompt_symbol(struct hist_entry **target, const char *msg) sleep(1); return; } else - parse_source(found); + perf_top__parse_source(top, found); out_free: free(buf); } -static void print_mapped_keys(void) +static void perf_top__print_mapped_keys(struct perf_top *top) { char *name = NULL; - if (top.sym_filter_entry) { - struct symbol *sym = top.sym_filter_entry->ms.sym; + if (top->sym_filter_entry) { + struct symbol *sym = top->sym_filter_entry->ms.sym; name = sym->name; } fprintf(stdout, "\nMapped keys:\n"); - fprintf(stdout, "\t[d] display refresh delay. \t(%d)\n", top.delay_secs); - fprintf(stdout, "\t[e] display entries (lines). \t(%d)\n", top.print_entries); + fprintf(stdout, "\t[d] display refresh delay. \t(%d)\n", top->delay_secs); + fprintf(stdout, "\t[e] display entries (lines). \t(%d)\n", top->print_entries); - if (top.evlist->nr_entries > 1) - fprintf(stdout, "\t[E] active event counter. \t(%s)\n", event_name(top.sym_evsel)); + if (top->evlist->nr_entries > 1) + fprintf(stdout, "\t[E] active event counter. \t(%s)\n", event_name(top->sym_evsel)); - fprintf(stdout, "\t[f] profile display filter (count). \t(%d)\n", top.count_filter); + fprintf(stdout, "\t[f] profile display filter (count). \t(%d)\n", top->count_filter); - fprintf(stdout, "\t[F] annotate display filter (percent). \t(%d%%)\n", sym_pcnt_filter); + fprintf(stdout, "\t[F] annotate display filter (percent). \t(%d%%)\n", top->sym_pcnt_filter); fprintf(stdout, "\t[s] annotate symbol. \t(%s)\n", name?: "NULL"); fprintf(stdout, "\t[S] stop annotation.\n"); fprintf(stdout, "\t[K] hide kernel_symbols symbols. \t(%s)\n", - top.hide_kernel_symbols ? "yes" : "no"); + top->hide_kernel_symbols ? "yes" : "no"); fprintf(stdout, "\t[U] hide user symbols. \t(%s)\n", - top.hide_user_symbols ? "yes" : "no"); - fprintf(stdout, "\t[z] toggle sample zeroing. \t(%d)\n", top.zero ? 1 : 0); + top->hide_user_symbols ? "yes" : "no"); + fprintf(stdout, "\t[z] toggle sample zeroing. \t(%d)\n", top->zero ? 1 : 0); fprintf(stdout, "\t[qQ] quit.\n"); } -static int key_mapped(int c) +static int perf_top__key_mapped(struct perf_top *top, int c) { switch (c) { case 'd': @@ -439,7 +407,7 @@ static int key_mapped(int c) case 'S': return 1; case 'E': - return top.evlist->nr_entries > 1 ? 1 : 0; + return top->evlist->nr_entries > 1 ? 1 : 0; default: break; } @@ -447,13 +415,13 @@ static int key_mapped(int c) return 0; } -static void handle_keypress(int c) +static void perf_top__handle_keypress(struct perf_top *top, int c) { - if (!key_mapped(c)) { + if (!perf_top__key_mapped(top, c)) { struct pollfd stdin_poll = { .fd = 0, .events = POLLIN }; struct termios tc, save; - print_mapped_keys(); + perf_top__print_mapped_keys(top); fprintf(stdout, "\nEnter selection, or unmapped key to continue: "); fflush(stdout); @@ -468,81 +436,86 @@ static void handle_keypress(int c) c = getc(stdin); tcsetattr(0, TCSAFLUSH, &save); - if (!key_mapped(c)) + if (!perf_top__key_mapped(top, c)) return; } switch (c) { case 'd': - prompt_integer(&top.delay_secs, "Enter display delay"); - if (top.delay_secs < 1) - top.delay_secs = 1; + prompt_integer(&top->delay_secs, "Enter display delay"); + if (top->delay_secs < 1) + top->delay_secs = 1; break; case 'e': - prompt_integer(&top.print_entries, "Enter display entries (lines)"); - if (top.print_entries == 0) { - sig_winch_handler(SIGWINCH); - signal(SIGWINCH, sig_winch_handler); + prompt_integer(&top->print_entries, "Enter display entries (lines)"); + if (top->print_entries == 0) { + struct sigaction act = { + .sa_sigaction = perf_top__sig_winch, + .sa_flags = SA_SIGINFO, + }; + perf_top__sig_winch(SIGWINCH, NULL, top); + sigaction(SIGWINCH, &act, NULL); } else signal(SIGWINCH, SIG_DFL); break; case 'E': - if (top.evlist->nr_entries > 1) { + if (top->evlist->nr_entries > 1) { /* Select 0 as the default event: */ int counter = 0; fprintf(stderr, "\nAvailable events:"); - list_for_each_entry(top.sym_evsel, &top.evlist->entries, node) - fprintf(stderr, "\n\t%d %s", top.sym_evsel->idx, event_name(top.sym_evsel)); + list_for_each_entry(top->sym_evsel, &top->evlist->entries, node) + fprintf(stderr, "\n\t%d %s", top->sym_evsel->idx, event_name(top->sym_evsel)); prompt_integer(&counter, "Enter details event counter"); - if (counter >= top.evlist->nr_entries) { - top.sym_evsel = list_entry(top.evlist->entries.next, struct perf_evsel, node); - fprintf(stderr, "Sorry, no such event, using %s.\n", event_name(top.sym_evsel)); + if (counter >= top->evlist->nr_entries) { + top->sym_evsel = list_entry(top->evlist->entries.next, struct perf_evsel, node); + fprintf(stderr, "Sorry, no such event, using %s.\n", event_name(top->sym_evsel)); sleep(1); break; } - list_for_each_entry(top.sym_evsel, &top.evlist->entries, node) - if (top.sym_evsel->idx == counter) + list_for_each_entry(top->sym_evsel, &top->evlist->entries, node) + if (top->sym_evsel->idx == counter) break; } else - top.sym_evsel = list_entry(top.evlist->entries.next, struct perf_evsel, node); + top->sym_evsel = list_entry(top->evlist->entries.next, struct perf_evsel, node); break; case 'f': - prompt_integer(&top.count_filter, "Enter display event count filter"); + prompt_integer(&top->count_filter, "Enter display event count filter"); break; case 'F': - prompt_percent(&sym_pcnt_filter, "Enter details display event filter (percent)"); + prompt_percent(&top->sym_pcnt_filter, + "Enter details display event filter (percent)"); break; case 'K': - top.hide_kernel_symbols = !top.hide_kernel_symbols; + top->hide_kernel_symbols = !top->hide_kernel_symbols; break; case 'q': case 'Q': printf("exiting.\n"); - if (dump_symtab) - perf_session__fprintf_dsos(top.session, stderr); + if (top->dump_symtab) + perf_session__fprintf_dsos(top->session, stderr); exit(0); case 's': - prompt_symbol(&top.sym_filter_entry, "Enter details symbol"); + perf_top__prompt_symbol(top, "Enter details symbol"); break; case 'S': - if (!top.sym_filter_entry) + if (!top->sym_filter_entry) break; else { - struct hist_entry *syme = top.sym_filter_entry; + struct hist_entry *syme = top->sym_filter_entry; - top.sym_filter_entry = NULL; + top->sym_filter_entry = NULL; __zero_source_counters(syme); } break; case 'U': - top.hide_user_symbols = !top.hide_user_symbols; + top->hide_user_symbols = !top->hide_user_symbols; break; case 'z': - top.zero = !top.zero; + top->zero = !top->zero; break; default: break; @@ -560,28 +533,30 @@ static void perf_top__sort_new_samples(void *arg) hists__collapse_resort_threaded(&t->sym_evsel->hists); hists__output_resort_threaded(&t->sym_evsel->hists); hists__decay_entries_threaded(&t->sym_evsel->hists, - top.hide_user_symbols, - top.hide_kernel_symbols); + t->hide_user_symbols, + t->hide_kernel_symbols); } -static void *display_thread_tui(void *arg __used) +static void *display_thread_tui(void *arg) { + struct perf_top *top = arg; const char *help = "For a higher level overview, try: perf top --sort comm,dso"; - perf_top__sort_new_samples(&top); - perf_evlist__tui_browse_hists(top.evlist, help, + perf_top__sort_new_samples(top); + perf_evlist__tui_browse_hists(top->evlist, help, perf_top__sort_new_samples, - &top, top.delay_secs); + top, top->delay_secs); exit_browser(0); exit(0); return NULL; } -static void *display_thread(void *arg __used) +static void *display_thread(void *arg) { struct pollfd stdin_poll = { .fd = 0, .events = POLLIN }; struct termios tc, save; + struct perf_top *top = arg; int delay_msecs, c; tcgetattr(0, &save); @@ -592,13 +567,13 @@ static void *display_thread(void *arg __used) pthread__unblock_sigwinch(); repeat: - delay_msecs = top.delay_secs * 1000; + delay_msecs = top->delay_secs * 1000; tcsetattr(0, TCSANOW, &tc); /* trash return*/ getc(stdin); while (1) { - print_sym_table(); + perf_top__print_sym_table(top); /* * Either timeout expired or we got an EINTR due to SIGWINCH, * refresh screen in both cases. @@ -618,7 +593,7 @@ process_hotkey: c = getc(stdin); tcsetattr(0, TCSAFLUSH, &save); - handle_keypress(c); + perf_top__handle_keypress(top, c); goto repeat; return NULL; @@ -670,11 +645,13 @@ static int symbol_filter(struct map *map __used, struct symbol *sym) return 0; } -static void perf_event__process_sample(const union perf_event *event, +static void perf_event__process_sample(struct perf_tool *tool, + const union perf_event *event, struct perf_evsel *evsel, struct perf_sample *sample, struct machine *machine) { + struct perf_top *top = container_of(tool, struct perf_top, tool); struct symbol *parent = NULL; u64 ip = event->ip.ip; struct addr_location al; @@ -687,14 +664,14 @@ static void perf_event__process_sample(const union perf_event *event, } if (event->header.misc & PERF_RECORD_MISC_EXACT_IP) - top.exact_samples++; + top->exact_samples++; if (perf_event__preprocess_sample(event, machine, &al, sample, symbol_filter) < 0 || al.filtered) return; - if (!kptr_restrict_warned && + if (!top->kptr_restrict_warned && symbol_conf.kptr_restrict && al.cpumode == PERF_RECORD_MISC_KERNEL) { ui__warning( @@ -705,7 +682,7 @@ static void perf_event__process_sample(const union perf_event *event, " modules" : ""); if (use_browser <= 0) sleep(5); - kptr_restrict_warned = true; + top->kptr_restrict_warned = true; } if (al.sym == NULL) { @@ -721,7 +698,7 @@ static void perf_event__process_sample(const union perf_event *event, * --hide-kernel-symbols, even if the user specifies an * invalid --vmlinux ;-) */ - if (!kptr_restrict_warned && !vmlinux_warned && + if (!top->kptr_restrict_warned && !top->vmlinux_warned && al.map == machine->vmlinux_maps[MAP__FUNCTION] && RB_EMPTY_ROOT(&al.map->dso->symbols[MAP__FUNCTION])) { if (symbol_conf.vmlinux_name) { @@ -734,7 +711,7 @@ static void perf_event__process_sample(const union perf_event *event, if (use_browser <= 0) sleep(5); - vmlinux_warned = true; + top->vmlinux_warned = true; } } @@ -762,56 +739,57 @@ static void perf_event__process_sample(const union perf_event *event, return; } - if (sort_has_symbols) - record_precise_ip(he, evsel->idx, ip); + if (top->sort_has_symbols) + perf_top__record_precise_ip(top, he, evsel->idx, ip); } return; } -static void perf_session__mmap_read_idx(struct perf_session *self, int idx) +static void perf_top__mmap_read_idx(struct perf_top *top, int idx) { struct perf_sample sample; struct perf_evsel *evsel; + struct perf_session *session = top->session; union perf_event *event; struct machine *machine; u8 origin; int ret; - while ((event = perf_evlist__mmap_read(top.evlist, idx)) != NULL) { - ret = perf_session__parse_sample(self, event, &sample); + while ((event = perf_evlist__mmap_read(top->evlist, idx)) != NULL) { + ret = perf_session__parse_sample(session, event, &sample); if (ret) { pr_err("Can't parse sample, err = %d\n", ret); continue; } - evsel = perf_evlist__id2evsel(self->evlist, sample.id); + evsel = perf_evlist__id2evsel(session->evlist, sample.id); assert(evsel != NULL); origin = event->header.misc & PERF_RECORD_MISC_CPUMODE_MASK; if (event->header.type == PERF_RECORD_SAMPLE) - ++top.samples; + ++top->samples; switch (origin) { case PERF_RECORD_MISC_USER: - ++top.us_samples; - if (top.hide_user_symbols) + ++top->us_samples; + if (top->hide_user_symbols) continue; - machine = perf_session__find_host_machine(self); + machine = perf_session__find_host_machine(session); break; case PERF_RECORD_MISC_KERNEL: - ++top.kernel_samples; - if (top.hide_kernel_symbols) + ++top->kernel_samples; + if (top->hide_kernel_symbols) continue; - machine = perf_session__find_host_machine(self); + machine = perf_session__find_host_machine(session); break; case PERF_RECORD_MISC_GUEST_KERNEL: - ++top.guest_kernel_samples; - machine = perf_session__find_machine(self, event->ip.pid); + ++top->guest_kernel_samples; + machine = perf_session__find_machine(session, event->ip.pid); break; case PERF_RECORD_MISC_GUEST_USER: - ++top.guest_us_samples; + ++top->guest_us_samples; /* * TODO: we don't process guest user from host side * except simple counting. @@ -822,27 +800,29 @@ static void perf_session__mmap_read_idx(struct perf_session *self, int idx) } - if (event->header.type == PERF_RECORD_SAMPLE) - perf_event__process_sample(event, evsel, &sample, machine); - else if (event->header.type < PERF_RECORD_MAX) { + if (event->header.type == PERF_RECORD_SAMPLE) { + perf_event__process_sample(&top->tool, event, evsel, + &sample, machine); + } else if (event->header.type < PERF_RECORD_MAX) { hists__inc_nr_events(&evsel->hists, event->header.type); - perf_event__process(&top.tool, event, &sample, machine); + perf_event__process(&top->tool, event, &sample, machine); } else - ++self->hists.stats.nr_unknown_events; + ++session->hists.stats.nr_unknown_events; } } -static void perf_session__mmap_read(struct perf_session *self) +static void perf_top__mmap_read(struct perf_top *top) { int i; - for (i = 0; i < top.evlist->nr_mmaps; i++) - perf_session__mmap_read_idx(self, i); + for (i = 0; i < top->evlist->nr_mmaps; i++) + perf_top__mmap_read_idx(top, i); } -static void start_counters(struct perf_evlist *evlist) +static void perf_top__start_counters(struct perf_top *top) { struct perf_evsel *counter, *first; + struct perf_evlist *evlist = top->evlist; first = list_entry(evlist->entries.next, struct perf_evsel, node); @@ -850,15 +830,15 @@ static void start_counters(struct perf_evlist *evlist) struct perf_event_attr *attr = &counter->attr; struct xyarray *group_fd = NULL; - if (group && counter != first) + if (top->group && counter != first) group_fd = first->fd; attr->sample_type = PERF_SAMPLE_IP | PERF_SAMPLE_TID; - if (top.freq) { + if (top->freq) { attr->sample_type |= PERF_SAMPLE_PERIOD; attr->freq = 1; - attr->sample_freq = top.freq; + attr->sample_freq = top->freq; } if (evlist->nr_entries > 1) { @@ -871,23 +851,23 @@ static void start_counters(struct perf_evlist *evlist) attr->mmap = 1; attr->comm = 1; - attr->inherit = inherit; + attr->inherit = top->inherit; retry_sample_id: - attr->sample_id_all = sample_id_all_avail ? 1 : 0; + attr->sample_id_all = top->sample_id_all_avail ? 1 : 0; try_again: - if (perf_evsel__open(counter, top.evlist->cpus, - top.evlist->threads, group, + if (perf_evsel__open(counter, top->evlist->cpus, + top->evlist->threads, top->group, group_fd) < 0) { int err = errno; if (err == EPERM || err == EACCES) { ui__error_paranoid(); goto out_err; - } else if (err == EINVAL && sample_id_all_avail) { + } else if (err == EINVAL && top->sample_id_all_avail) { /* * Old kernel, no attr->sample_id_type_all field */ - sample_id_all_avail = false; + top->sample_id_all_avail = false; goto retry_sample_id; } /* @@ -921,7 +901,7 @@ try_again: } } - if (perf_evlist__mmap(evlist, mmap_pages, false) < 0) { + if (perf_evlist__mmap(evlist, top->mmap_pages, false) < 0) { ui__warning("Failed to mmap with %d (%s)\n", errno, strerror(errno)); goto out_err; @@ -934,14 +914,14 @@ out_err: exit(0); } -static int setup_sample_type(void) +static int perf_top__setup_sample_type(struct perf_top *top) { - if (!sort_has_symbols) { + if (!top->sort_has_symbols) { if (symbol_conf.use_callchain) { ui__warning("Selected -g but \"sym\" not present in --sort/-s."); return -EINVAL; } - } else if (!dont_use_callchains && callchain_param.mode != CHAIN_NONE) { + } else if (!top->dont_use_callchains && callchain_param.mode != CHAIN_NONE) { if (callchain_register_param(&callchain_param) < 0) { ui__warning("Can't register callchain params.\n"); return -EINVAL; @@ -951,7 +931,7 @@ static int setup_sample_type(void) return 0; } -static int __cmd_top(void) +static int __cmd_top(struct perf_top *top) { pthread_t thread; int ret; @@ -959,40 +939,40 @@ static int __cmd_top(void) * FIXME: perf_session__new should allow passing a O_MMAP, so that all this * mmap reading, etc is encapsulated in it. Use O_WRONLY for now. */ - top.session = perf_session__new(NULL, O_WRONLY, false, false, NULL); - if (top.session == NULL) + top->session = perf_session__new(NULL, O_WRONLY, false, false, NULL); + if (top->session == NULL) return -ENOMEM; - ret = setup_sample_type(); + ret = perf_top__setup_sample_type(top); if (ret) goto out_delete; - if (top.target_tid != -1) - perf_event__synthesize_thread_map(&top.tool, top.evlist->threads, + if (top->target_tid != -1) + perf_event__synthesize_thread_map(&top->tool, top->evlist->threads, perf_event__process, - &top.session->host_machine); + &top->session->host_machine); else - perf_event__synthesize_threads(&top.tool, perf_event__process, - &top.session->host_machine); - start_counters(top.evlist); - top.session->evlist = top.evlist; - perf_session__update_sample_type(top.session); + perf_event__synthesize_threads(&top->tool, perf_event__process, + &top->session->host_machine); + perf_top__start_counters(top); + top->session->evlist = top->evlist; + perf_session__update_sample_type(top->session); /* Wait for a minimal set of events before starting the snapshot */ - poll(top.evlist->pollfd, top.evlist->nr_fds, 100); + poll(top->evlist->pollfd, top->evlist->nr_fds, 100); - perf_session__mmap_read(top.session); + perf_top__mmap_read(top); if (pthread_create(&thread, NULL, (use_browser > 0 ? display_thread_tui : - display_thread), NULL)) { + display_thread), top)) { printf("Could not create display thread.\n"); exit(-1); } - if (realtime_prio) { + if (top->realtime_prio) { struct sched_param param; - param.sched_priority = realtime_prio; + param.sched_priority = top->realtime_prio; if (sched_setscheduler(0, SCHED_FIFO, ¶m)) { printf("Could not set realtime priority.\n"); exit(-1); @@ -1000,25 +980,25 @@ static int __cmd_top(void) } while (1) { - u64 hits = top.samples; + u64 hits = top->samples; - perf_session__mmap_read(top.session); + perf_top__mmap_read(top); - if (hits == top.samples) - ret = poll(top.evlist->pollfd, top.evlist->nr_fds, 100); + if (hits == top->samples) + ret = poll(top->evlist->pollfd, top->evlist->nr_fds, 100); } out_delete: - perf_session__delete(top.session); - top.session = NULL; + perf_session__delete(top->session); + top->session = NULL; return 0; } static int -parse_callchain_opt(const struct option *opt __used, const char *arg, - int unset) +parse_callchain_opt(const struct option *opt, const char *arg, int unset) { + struct perf_top *top = (struct perf_top *)opt->value; char *tok, *tok2; char *endptr; @@ -1026,7 +1006,7 @@ parse_callchain_opt(const struct option *opt __used, const char *arg, * --no-call-graph */ if (unset) { - dont_use_callchains = true; + top->dont_use_callchains = true; return 0; } @@ -1100,17 +1080,32 @@ static const char * const top_usage[] = { NULL }; -static const struct option options[] = { +int cmd_top(int argc, const char **argv, const char *prefix __used) +{ + struct perf_evsel *pos; + int status = -ENOMEM; + struct perf_top top = { + .count_filter = 5, + .delay_secs = 2, + .target_pid = -1, + .target_tid = -1, + .freq = 1000, /* 1 KHz */ + .sample_id_all_avail = true, + .mmap_pages = 128, + .sym_pcnt_filter = 5, + }; + char callchain_default_opt[] = "fractal,0.5,callee"; + const struct option options[] = { OPT_CALLBACK('e', "event", &top.evlist, "event", "event selector. use 'perf list' to list available events", parse_events_option), - OPT_INTEGER('c', "count", &default_interval, + OPT_INTEGER('c', "count", &top.default_interval, "event period to sample"), OPT_INTEGER('p', "pid", &top.target_pid, "profile events on existing process id"), OPT_INTEGER('t', "tid", &top.target_tid, "profile events on existing thread id"), - OPT_BOOLEAN('a', "all-cpus", &system_wide, + OPT_BOOLEAN('a', "all-cpus", &top.system_wide, "system-wide collection from all CPUs"), OPT_STRING('C', "cpu", &top.cpu_list, "cpu", "list of cpus to monitor"), @@ -1118,20 +1113,20 @@ static const struct option options[] = { "file", "vmlinux pathname"), OPT_BOOLEAN('K', "hide_kernel_symbols", &top.hide_kernel_symbols, "hide kernel symbols"), - OPT_UINTEGER('m', "mmap-pages", &mmap_pages, "number of mmap data pages"), - OPT_INTEGER('r', "realtime", &realtime_prio, + OPT_UINTEGER('m', "mmap-pages", &top.mmap_pages, "number of mmap data pages"), + OPT_INTEGER('r', "realtime", &top.realtime_prio, "collect data with this RT SCHED_FIFO priority"), OPT_INTEGER('d', "delay", &top.delay_secs, "number of seconds to delay between refreshes"), - OPT_BOOLEAN('D', "dump-symtab", &dump_symtab, + OPT_BOOLEAN('D', "dump-symtab", &top.dump_symtab, "dump the symbol table used for profiling"), OPT_INTEGER('f', "count-filter", &top.count_filter, "only display functions with more events than this"), - OPT_BOOLEAN('g', "group", &group, + OPT_BOOLEAN('g', "group", &top.group, "put the counters into a counter group"), - OPT_BOOLEAN('i', "inherit", &inherit, + OPT_BOOLEAN('i', "inherit", &top.inherit, "child tasks inherit counters"), - OPT_STRING(0, "sym-annotate", &sym_filter, "symbol name", + OPT_STRING(0, "sym-annotate", &top.sym_filter, "symbol name", "symbol to annotate"), OPT_BOOLEAN('z', "zero", &top.zero, "zero history across updates"), @@ -1141,15 +1136,15 @@ static const struct option options[] = { "display this many functions"), OPT_BOOLEAN('U', "hide_user_symbols", &top.hide_user_symbols, "hide user symbols"), - OPT_BOOLEAN(0, "tui", &use_tui, "Use the TUI interface"), - OPT_BOOLEAN(0, "stdio", &use_stdio, "Use the stdio interface"), + OPT_BOOLEAN(0, "tui", &top.use_tui, "Use the TUI interface"), + OPT_BOOLEAN(0, "stdio", &top.use_stdio, "Use the stdio interface"), OPT_INCR('v', "verbose", &verbose, "be more verbose (show counter open errors, etc)"), OPT_STRING('s', "sort", &sort_order, "key[,key2...]", "sort by key(s): pid, comm, dso, symbol, parent"), OPT_BOOLEAN('n', "show-nr-samples", &symbol_conf.show_nr_samples, "Show a column with the number of samples"), - OPT_CALLBACK_DEFAULT('G', "call-graph", NULL, "output_type,min_percent, call_order", + OPT_CALLBACK_DEFAULT('G', "call-graph", &top, "output_type,min_percent, call_order", "Display callchains using output_type (graph, flat, fractal, or none), min percent threshold and callchain order. " "Default: fractal,0.5,callee", &parse_callchain_opt, callchain_default_opt), @@ -1168,12 +1163,7 @@ static const struct option options[] = { OPT_STRING('M', "disassembler-style", &disassembler_style, "disassembler style", "Specify disassembler style (e.g. -M intel for intel syntax)"), OPT_END() -}; - -int cmd_top(int argc, const char **argv, const char *prefix __used) -{ - struct perf_evsel *pos; - int status = -ENOMEM; + }; top.evlist = perf_evlist__new(NULL, NULL); if (top.evlist == NULL) @@ -1190,9 +1180,9 @@ int cmd_top(int argc, const char **argv, const char *prefix __used) setup_sorting(top_usage, options); - if (use_stdio) + if (top.use_stdio) use_browser = 0; - else if (use_tui) + else if (top.use_tui) use_browser = 1; setup_browser(false); @@ -1225,10 +1215,10 @@ int cmd_top(int argc, const char **argv, const char *prefix __used) /* * User specified count overrides default frequency. */ - if (default_interval) + if (top.default_interval) top.freq = 0; else if (top.freq) { - default_interval = top.freq; + top.default_interval = top.freq; } else { fprintf(stderr, "frequency and count are zero, aborting\n"); exit(EXIT_FAILURE); @@ -1244,7 +1234,7 @@ int cmd_top(int argc, const char **argv, const char *prefix __used) if (pos->attr.sample_period) continue; - pos->attr.sample_period = default_interval; + pos->attr.sample_period = top.default_interval; } if (perf_evlist__alloc_pollfd(top.evlist) < 0 || @@ -1267,15 +1257,19 @@ int cmd_top(int argc, const char **argv, const char *prefix __used) * Avoid annotation data structures overhead when symbols aren't on the * sort list. */ - sort_has_symbols = sort_sym.list.next != NULL; + top.sort_has_symbols = sort_sym.list.next != NULL; - get_term_dimensions(&winsize); + get_term_dimensions(&top.winsize); if (top.print_entries == 0) { - update_print_entries(&winsize); - signal(SIGWINCH, sig_winch_handler); + struct sigaction act = { + .sa_sigaction = perf_top__sig_winch, + .sa_flags = SA_SIGINFO, + }; + perf_top__update_print_entries(&top); + sigaction(SIGWINCH, &act, NULL); } - status = __cmd_top(); + status = __cmd_top(&top); out_free_fd: perf_evlist__delete(top.evlist); diff --git a/tools/perf/util/top.h b/tools/perf/util/top.h index 40430ec5c267..a248f3c2c60d 100644 --- a/tools/perf/util/top.h +++ b/tools/perf/util/top.h @@ -25,10 +25,26 @@ struct perf_top { int freq; pid_t target_pid, target_tid; bool hide_kernel_symbols, hide_user_symbols, zero; + bool system_wide; + bool use_tui, use_stdio; + bool sort_has_symbols; + bool dont_use_callchains; + bool kptr_restrict_warned; + bool vmlinux_warned; + bool inherit; + bool group; + bool sample_id_all_avail; + bool dump_symtab; const char *cpu_list; struct hist_entry *sym_filter_entry; struct perf_evsel *sym_evsel; struct perf_session *session; + struct winsize winsize; + unsigned int mmap_pages; + int default_interval; + int realtime_prio; + int sym_pcnt_filter; + const char *sym_filter; }; size_t perf_top__header_snprintf(struct perf_top *top, char *bf, size_t size); -- cgit v1.2.3 From c8e6672035e84799e6167e933fafedc8e3256973 Mon Sep 17 00:00:00 2001 From: David Ahern Date: Sun, 13 Nov 2011 11:30:08 -0700 Subject: perf tools: make -C consistent across commands (for cpu list arg) Currently the meaning of -C varies by perf command: for perf-top, perf-stat, perf-record it means cpu list. For perf-report it means comm list. Then perf-annotate, perf-report and perf-script use -c for cpu list. Fix annotate, report and script to use -C for cpu list to be consistent with top, stat and record. This means report needs to use -c for comm list which does introduce a backward compatibility change. v1 -> v2 - update perf-script.txt too Cc: Frederic Weisbecker Cc: Ingo Molnar Cc: Peter Zijlstra Link: http://lkml.kernel.org/r/1321209008-7004-1-git-send-email-dsahern@gmail.com Signed-off-by: David Ahern Signed-off-by: Arnaldo Carvalho de Melo --- tools/perf/Documentation/perf-annotate.txt | 2 +- tools/perf/Documentation/perf-report.txt | 4 ++-- tools/perf/Documentation/perf-script.txt | 2 +- tools/perf/builtin-annotate.c | 2 +- tools/perf/builtin-report.c | 4 ++-- tools/perf/builtin-script.c | 2 +- 6 files changed, 8 insertions(+), 8 deletions(-) diff --git a/tools/perf/Documentation/perf-annotate.txt b/tools/perf/Documentation/perf-annotate.txt index fe6762ed56bd..476029d30621 100644 --- a/tools/perf/Documentation/perf-annotate.txt +++ b/tools/perf/Documentation/perf-annotate.txt @@ -66,7 +66,7 @@ OPTIONS used. This interfaces starts by centering on the line with more samples, TAB/UNTAB cycles through the lines with more samples. --c:: +-C:: --cpu:: Only report samples for the list of CPUs provided. Multiple CPUs can be provided as a comma-separated list with no space: 0,1. Ranges of CPUs are specified with -: 0-2. Default is to report samples on all diff --git a/tools/perf/Documentation/perf-report.txt b/tools/perf/Documentation/perf-report.txt index 212f24d672e1..dc85392a5ac7 100644 --- a/tools/perf/Documentation/perf-report.txt +++ b/tools/perf/Documentation/perf-report.txt @@ -39,7 +39,7 @@ OPTIONS -T:: --threads:: Show per-thread event counters --C:: +-c:: --comms=:: Only consider symbols in these comms. CSV that understands file://filename entries. @@ -128,7 +128,7 @@ OPTIONS --symfs=:: Look for files with symbols relative to this directory. --c:: +-C:: --cpu:: Only report samples for the list of CPUs provided. Multiple CPUs can be provided as a comma-separated list with no space: 0,1. Ranges of CPUs are specified with -: 0-2. Default is to report samples on all diff --git a/tools/perf/Documentation/perf-script.txt b/tools/perf/Documentation/perf-script.txt index dec87ecb530e..3613b0a1aff2 100644 --- a/tools/perf/Documentation/perf-script.txt +++ b/tools/perf/Documentation/perf-script.txt @@ -182,7 +182,7 @@ OPTIONS --hide-call-graph:: When printing symbols do not display call chain. --c:: +-C:: --cpu:: Only report samples for the list of CPUs provided. Multiple CPUs can be provided as a comma-separated list with no space: 0,1. Ranges of CPUs are specified with -: 0-2. Default is to report samples on all diff --git a/tools/perf/builtin-annotate.c b/tools/perf/builtin-annotate.c index c01139fa4a10..d449645c5ef1 100644 --- a/tools/perf/builtin-annotate.c +++ b/tools/perf/builtin-annotate.c @@ -274,7 +274,7 @@ int cmd_annotate(int argc, const char **argv, const char *prefix __used) "print matching source lines (may be slow)"), OPT_BOOLEAN('P', "full-paths", &annotate.full_paths, "Don't shorten the displayed pathnames"), - OPT_STRING('c', "cpu", &annotate.cpu_list, "cpu", "list of cpus to profile"), + OPT_STRING('C', "cpu", &annotate.cpu_list, "cpu", "list of cpus to profile"), OPT_STRING(0, "symfs", &symbol_conf.symfs, "directory", "Look for files with symbols relative to this directory"), OPT_BOOLEAN(0, "source", &symbol_conf.annotate_src, diff --git a/tools/perf/builtin-report.c b/tools/perf/builtin-report.c index eef8e423deb0..ece7c5d3f504 100644 --- a/tools/perf/builtin-report.c +++ b/tools/perf/builtin-report.c @@ -492,7 +492,7 @@ int cmd_report(int argc, const char **argv, const char *prefix __used) "alias for inverted call graph"), OPT_STRING('d', "dsos", &symbol_conf.dso_list_str, "dso[,dso...]", "only consider symbols in these dsos"), - OPT_STRING('C', "comms", &symbol_conf.comm_list_str, "comm[,comm...]", + OPT_STRING('c', "comms", &symbol_conf.comm_list_str, "comm[,comm...]", "only consider symbols in these comms"), OPT_STRING('S', "symbols", &symbol_conf.sym_list_str, "symbol[,symbol...]", "only consider these symbols"), @@ -506,7 +506,7 @@ int cmd_report(int argc, const char **argv, const char *prefix __used) "Only display entries resolved to a symbol"), OPT_STRING(0, "symfs", &symbol_conf.symfs, "directory", "Look for files with symbols relative to this directory"), - OPT_STRING('c', "cpu", &report.cpu_list, "cpu", + OPT_STRING('C', "cpu", &report.cpu_list, "cpu", "list of cpus to profile"), OPT_BOOLEAN('I', "show-info", &report.show_full_info, "Display extended information about perf.data file"), diff --git a/tools/perf/builtin-script.c b/tools/perf/builtin-script.c index 5f8afc65d5f3..7731a09e975c 100644 --- a/tools/perf/builtin-script.c +++ b/tools/perf/builtin-script.c @@ -1085,7 +1085,7 @@ static const struct option options[] = { OPT_CALLBACK('f', "fields", NULL, "str", "comma separated output fields prepend with 'type:'. Valid types: hw,sw,trace,raw. Fields: comm,tid,pid,time,cpu,event,trace,ip,sym,dso,addr", parse_output_fields), - OPT_STRING('c', "cpu", &cpu_list, "cpu", "list of cpus to profile"), + OPT_STRING('C', "cpu", &cpu_list, "cpu", "list of cpus to profile"), OPT_BOOLEAN('I', "show-info", &show_full_info, "display extended information from perf.data file"), OPT_END() -- cgit v1.2.3 From e7984b7bee2fca8f582f5bc2bf1e6c93420a5dd5 Mon Sep 17 00:00:00 2001 From: David Ahern Date: Mon, 21 Nov 2011 10:02:52 -0700 Subject: perf script: Add comm filtering option Allows collecting events system wide and then pulling out events for a specific task name(s). e.g, perf script -c gnome-shell,gnome-terminal Applies on top of: https://lkml.org/lkml/2011/11/13/74 v2->v3 - update Documentation v1->v2 - use comm_list from symbol_conf Cc: Frederic Weisbecker Cc: Ingo Molnar Cc: Peter Zijlstra Link: http://lkml.kernel.org/r/1321894972-24246-1-git-send-email-dsahern@gmail.com Signed-off-by: David Ahern Signed-off-by: Arnaldo Carvalho de Melo --- tools/perf/Documentation/perf-script.txt | 5 +++++ tools/perf/builtin-script.c | 12 ++++++++++++ 2 files changed, 17 insertions(+) diff --git a/tools/perf/Documentation/perf-script.txt b/tools/perf/Documentation/perf-script.txt index 3613b0a1aff2..7f61eaaf9ab8 100644 --- a/tools/perf/Documentation/perf-script.txt +++ b/tools/perf/Documentation/perf-script.txt @@ -188,6 +188,11 @@ OPTIONS CPUs are specified with -: 0-2. Default is to report samples on all CPUs. +-c:: +--comms=:: + Only display events for these comms. CSV that understands + file://filename entries. + -I:: --show-info:: Display extended information about the perf.data file. This adds diff --git a/tools/perf/builtin-script.c b/tools/perf/builtin-script.c index 7731a09e975c..619d6dcaa1d9 100644 --- a/tools/perf/builtin-script.c +++ b/tools/perf/builtin-script.c @@ -441,6 +441,7 @@ static int process_sample_event(struct perf_tool *tool __used, struct perf_evsel *evsel, struct machine *machine) { + struct addr_location al; struct thread *thread = machine__findnew_thread(machine, event->ip.pid); if (thread == NULL) { @@ -460,6 +461,15 @@ static int process_sample_event(struct perf_tool *tool __used, return 0; } + if (perf_event__preprocess_sample(event, machine, &al, sample, 0) < 0) { + pr_err("problem processing %d event, skipping it.\n", + event->header.type); + return -1; + } + + if (al.filtered) + return 0; + if (cpu_list && !test_bit(sample->cpu, cpu_bitmap)) return 0; @@ -1086,6 +1096,8 @@ static const struct option options[] = { "comma separated output fields prepend with 'type:'. Valid types: hw,sw,trace,raw. Fields: comm,tid,pid,time,cpu,event,trace,ip,sym,dso,addr", parse_output_fields), OPT_STRING('C', "cpu", &cpu_list, "cpu", "list of cpus to profile"), + OPT_STRING('c', "comms", &symbol_conf.comm_list_str, "comm[,comm...]", + "only display events for these comms"), OPT_BOOLEAN('I', "show-info", &show_full_info, "display extended information from perf.data file"), OPT_END() -- cgit v1.2.3 From ee29be625bd7b115d45eba4b0526ff3e24bf3ca0 Mon Sep 17 00:00:00 2001 From: Arnaldo Carvalho de Melo Date: Mon, 28 Nov 2011 17:57:40 -0200 Subject: perf tools: Save some loops using perf_evlist__id2evsel Since we already ask for PERF_SAMPLE_ID and use it to quickly find the associated evsel, add handler func + data to struct perf_evsel to avoid using chains of if(strcmp(event_name)) and also to avoid all the linear list searches via trace_event_find. To demonstrate the technique convert 'perf sched' to it: # perf sched record sleep 5m And then: Performance counter stats for '/tmp/oldperf sched lat': 646.929438 task-clock # 0.999 CPUs utilized 9 context-switches # 0.000 M/sec 0 CPU-migrations # 0.000 M/sec 20,901 page-faults # 0.032 M/sec 1,290,144,450 cycles # 1.994 GHz stalled-cycles-frontend stalled-cycles-backend 1,606,158,439 instructions # 1.24 insns per cycle 339,088,395 branches # 524.151 M/sec 4,550,735 branch-misses # 1.34% of all branches 0.647524759 seconds time elapsed Versus: Performance counter stats for 'perf sched lat': 473.564691 task-clock # 0.999 CPUs utilized 9 context-switches # 0.000 M/sec 0 CPU-migrations # 0.000 M/sec 20,903 page-faults # 0.044 M/sec 944,367,984 cycles # 1.994 GHz stalled-cycles-frontend stalled-cycles-backend 1,442,385,571 instructions # 1.53 insns per cycle 308,383,106 branches # 651.195 M/sec 4,481,784 branch-misses # 1.45% of all branches 0.474215751 seconds time elapsed [root@emilia ~]# Cc: David Ahern Cc: Frederic Weisbecker Cc: Mike Galbraith Cc: Paul Mackerras Cc: Peter Zijlstra Cc: Stephane Eranian Link: http://lkml.kernel.org/n/tip-1kbzpl74lwi6lavpqke2u2p3@git.kernel.org Signed-off-by: Arnaldo Carvalho de Melo --- tools/perf/builtin-sched.c | 149 ++++++++++++++++++++++----------------------- tools/perf/util/evlist.c | 42 +++++++++++++ tools/perf/util/evlist.h | 11 ++++ tools/perf/util/evsel.h | 4 ++ tools/perf/util/tool.h | 5 ++ 5 files changed, 136 insertions(+), 75 deletions(-) diff --git a/tools/perf/builtin-sched.c b/tools/perf/builtin-sched.c index 0ee868e6f63b..6284ed2317f2 100644 --- a/tools/perf/builtin-sched.c +++ b/tools/perf/builtin-sched.c @@ -2,6 +2,7 @@ #include "perf.h" #include "util/util.h" +#include "util/evlist.h" #include "util/cache.h" #include "util/evsel.h" #include "util/symbol.h" @@ -1358,12 +1359,13 @@ static void sort_lat(void) static struct trace_sched_handler *trace_handler; static void -process_sched_wakeup_event(void *data, struct machine *machine, +process_sched_wakeup_event(struct perf_tool *tool __used, struct event *event, - int cpu __used, - u64 timestamp __used, - struct thread *thread __used) + struct perf_sample *sample, + struct machine *machine, + struct thread *thread) { + void *data = sample->raw_data; struct trace_wakeup_event wakeup_event; FILL_COMMON_FIELDS(wakeup_event, event, data); @@ -1376,7 +1378,7 @@ process_sched_wakeup_event(void *data, struct machine *machine, if (trace_handler->wakeup_event) trace_handler->wakeup_event(&wakeup_event, machine, event, - cpu, timestamp, thread); + sample->cpu, sample->time, thread); } /* @@ -1471,14 +1473,15 @@ map_switch_event(struct trace_switch_event *switch_event, } } - static void -process_sched_switch_event(void *data, struct machine *machine, +process_sched_switch_event(struct perf_tool *tool __used, struct event *event, - int this_cpu, - u64 timestamp __used, - struct thread *thread __used) + struct perf_sample *sample, + struct machine *machine, + struct thread *thread) { + int this_cpu = sample->cpu; + void *data = sample->raw_data; struct trace_switch_event switch_event; FILL_COMMON_FIELDS(switch_event, event, data); @@ -1501,18 +1504,19 @@ process_sched_switch_event(void *data, struct machine *machine, } if (trace_handler->switch_event) trace_handler->switch_event(&switch_event, machine, event, - this_cpu, timestamp, thread); + this_cpu, sample->time, thread); curr_pid[this_cpu] = switch_event.next_pid; } static void -process_sched_runtime_event(void *data, struct machine *machine, - struct event *event, - int cpu __used, - u64 timestamp __used, - struct thread *thread __used) +process_sched_runtime_event(struct perf_tool *tool __used, + struct event *event, + struct perf_sample *sample, + struct machine *machine, + struct thread *thread) { + void *data = sample->raw_data; struct trace_runtime_event runtime_event; FILL_ARRAY(runtime_event, comm, event, data); @@ -1521,16 +1525,18 @@ process_sched_runtime_event(void *data, struct machine *machine, FILL_FIELD(runtime_event, vruntime, event, data); if (trace_handler->runtime_event) - trace_handler->runtime_event(&runtime_event, machine, event, cpu, timestamp, thread); + trace_handler->runtime_event(&runtime_event, machine, event, + sample->cpu, sample->time, thread); } static void -process_sched_fork_event(void *data, +process_sched_fork_event(struct perf_tool *tool __used, struct event *event, - int cpu __used, - u64 timestamp __used, - struct thread *thread __used) + struct perf_sample *sample, + struct machine *machine __used, + struct thread *thread) { + void *data = sample->raw_data; struct trace_fork_event fork_event; FILL_COMMON_FIELDS(fork_event, event, data); @@ -1542,13 +1548,14 @@ process_sched_fork_event(void *data, if (trace_handler->fork_event) trace_handler->fork_event(&fork_event, event, - cpu, timestamp, thread); + sample->cpu, sample->time, thread); } static void -process_sched_exit_event(struct event *event, - int cpu __used, - u64 timestamp __used, +process_sched_exit_event(struct perf_tool *tool __used, + struct event *event, + struct perf_sample *sample __used, + struct machine *machine __used, struct thread *thread __used) { if (verbose) @@ -1556,12 +1563,13 @@ process_sched_exit_event(struct event *event, } static void -process_sched_migrate_task_event(void *data, struct machine *machine, - struct event *event, - int cpu __used, - u64 timestamp __used, - struct thread *thread __used) +process_sched_migrate_task_event(struct perf_tool *tool __used, + struct event *event, + struct perf_sample *sample, + struct machine *machine, + struct thread *thread) { + void *data = sample->raw_data; struct trace_migrate_task_event migrate_task_event; FILL_COMMON_FIELDS(migrate_task_event, event, data); @@ -1573,67 +1581,46 @@ process_sched_migrate_task_event(void *data, struct machine *machine, if (trace_handler->migrate_task_event) trace_handler->migrate_task_event(&migrate_task_event, machine, - event, cpu, timestamp, thread); + event, sample->cpu, + sample->time, thread); } -static void process_raw_event(union perf_event *raw_event __used, - struct machine *machine, void *data, int cpu, - u64 timestamp, struct thread *thread) -{ - struct event *event; - int type; - - - type = trace_parse_common_type(data); - event = trace_find_event(type); - - if (!strcmp(event->name, "sched_switch")) - process_sched_switch_event(data, machine, event, cpu, timestamp, thread); - if (!strcmp(event->name, "sched_stat_runtime")) - process_sched_runtime_event(data, machine, event, cpu, timestamp, thread); - if (!strcmp(event->name, "sched_wakeup")) - process_sched_wakeup_event(data, machine, event, cpu, timestamp, thread); - if (!strcmp(event->name, "sched_wakeup_new")) - process_sched_wakeup_event(data, machine, event, cpu, timestamp, thread); - if (!strcmp(event->name, "sched_process_fork")) - process_sched_fork_event(data, event, cpu, timestamp, thread); - if (!strcmp(event->name, "sched_process_exit")) - process_sched_exit_event(event, cpu, timestamp, thread); - if (!strcmp(event->name, "sched_migrate_task")) - process_sched_migrate_task_event(data, machine, event, cpu, timestamp, thread); -} +typedef void (*tracepoint_handler)(struct perf_tool *tool, struct event *event, + struct perf_sample *sample, + struct machine *machine, + struct thread *thread); -static int process_sample_event(struct perf_tool *tool __used, - union perf_event *event, - struct perf_sample *sample, - struct perf_evsel *evsel, - struct machine *machine) +static int perf_sched__process_tracepoint_sample(struct perf_tool *tool, + union perf_event *event __used, + struct perf_sample *sample, + struct perf_evsel *evsel, + struct machine *machine) { - struct thread *thread; - - if (!(evsel->attr.sample_type & PERF_SAMPLE_RAW)) - return 0; + struct thread *thread = machine__findnew_thread(machine, sample->pid); - thread = machine__findnew_thread(machine, sample->pid); if (thread == NULL) { - pr_debug("problem processing %d event, skipping it.\n", - event->header.type); + pr_debug("problem processing %s event, skipping it.\n", + evsel->name); return -1; } - dump_printf(" ... thread: %s:%d\n", thread->comm, thread->pid); + evsel->hists.stats.total_period += sample->period; + hists__inc_nr_events(&evsel->hists, PERF_RECORD_SAMPLE); - if (profile_cpu != -1 && profile_cpu != (int)sample->cpu) - return 0; + if (evsel->handler.func != NULL) { + tracepoint_handler f = evsel->handler.func; - process_raw_event(event, machine, sample->raw_data, sample->cpu, - sample->time, thread); + if (evsel->handler.data == NULL) + evsel->handler.data = trace_find_event(evsel->attr.config); + + f(tool, evsel->handler.data, sample, machine, thread); + } return 0; } static struct perf_tool perf_sched = { - .sample = process_sample_event, + .sample = perf_sched__process_tracepoint_sample, .comm = perf_event__process_comm, .lost = perf_event__process_lost, .fork = perf_event__process_task, @@ -1643,11 +1630,23 @@ static struct perf_tool perf_sched = { static void read_events(bool destroy, struct perf_session **psession) { int err = -EINVAL; + const struct perf_evsel_str_handler handlers[] = { + { "sched:sched_switch", process_sched_switch_event, }, + { "sched:sched_stat_runtime", process_sched_runtime_event, }, + { "sched:sched_wakeup", process_sched_wakeup_event, }, + { "sched:sched_wakeup_new", process_sched_wakeup_event, }, + { "sched:sched_process_fork", process_sched_fork_event, }, + { "sched:sched_process_exit", process_sched_exit_event, }, + { "sched:sched_migrate_task", process_sched_migrate_task_event, }, + }; struct perf_session *session = perf_session__new(input_name, O_RDONLY, 0, false, &perf_sched); if (session == NULL) die("No Memory"); + err = perf_evlist__set_tracepoints_handlers_array(session->evlist, handlers); + assert(err == 0); + if (perf_session__has_traces(session, "record -R")) { err = perf_session__process_events(session, &perf_sched); if (err) diff --git a/tools/perf/util/evlist.c b/tools/perf/util/evlist.c index d44e3df13a8f..b36f26fe767a 100644 --- a/tools/perf/util/evlist.c +++ b/tools/perf/util/evlist.c @@ -207,6 +207,48 @@ out_free_attrs: return err; } +static struct perf_evsel * + perf_evlist__find_tracepoint_by_id(struct perf_evlist *evlist, int id) +{ + struct perf_evsel *evsel; + + list_for_each_entry(evsel, &evlist->entries, node) { + if (evsel->attr.type == PERF_TYPE_TRACEPOINT && + (int)evsel->attr.config == id) + return evsel; + } + + return NULL; +} + +int perf_evlist__set_tracepoints_handlers(struct perf_evlist *evlist, + const struct perf_evsel_str_handler *assocs, + size_t nr_assocs) +{ + struct perf_evsel *evsel; + int err; + size_t i; + + for (i = 0; i < nr_assocs; i++) { + err = trace_event__id(assocs[i].name); + if (err < 0) + goto out; + + evsel = perf_evlist__find_tracepoint_by_id(evlist, err); + if (evsel == NULL) + continue; + + err = -EEXIST; + if (evsel->handler.func != NULL) + goto out; + evsel->handler.func = assocs[i].handler; + } + + err = 0; +out: + return err; +} + void perf_evlist__disable(struct perf_evlist *evlist) { int cpu, thread; diff --git a/tools/perf/util/evlist.h b/tools/perf/util/evlist.h index 2202e7b04103..f94ed7e0d987 100644 --- a/tools/perf/util/evlist.h +++ b/tools/perf/util/evlist.h @@ -36,6 +36,11 @@ struct perf_evlist { struct perf_evsel *selected; }; +struct perf_evsel_str_handler { + const char *name; + void *handler; +}; + struct perf_evsel; struct perf_evlist *perf_evlist__new(struct cpu_map *cpus, @@ -51,6 +56,9 @@ int perf_evlist__add_attrs(struct perf_evlist *evlist, struct perf_event_attr *attrs, size_t nr_attrs); int perf_evlist__add_tracepoints(struct perf_evlist *evlist, const char *tracepoints[], size_t nr_tracepoints); +int perf_evlist__set_tracepoints_handlers(struct perf_evlist *evlist, + const struct perf_evsel_str_handler *assocs, + size_t nr_assocs); #define perf_evlist__add_attrs_array(evlist, array) \ perf_evlist__add_attrs(evlist, array, ARRAY_SIZE(array)) @@ -58,6 +66,9 @@ int perf_evlist__add_tracepoints(struct perf_evlist *evlist, #define perf_evlist__add_tracepoints_array(evlist, array) \ perf_evlist__add_tracepoints(evlist, array, ARRAY_SIZE(array)) +#define perf_evlist__set_tracepoints_handlers_array(evlist, array) \ + perf_evlist__set_tracepoints_handlers(evlist, array, ARRAY_SIZE(array)) + void perf_evlist__id_add(struct perf_evlist *evlist, struct perf_evsel *evsel, int cpu, int thread, u64 id); diff --git a/tools/perf/util/evsel.h b/tools/perf/util/evsel.h index 6421c07f5015..326b8e4d5035 100644 --- a/tools/perf/util/evsel.h +++ b/tools/perf/util/evsel.h @@ -61,6 +61,10 @@ struct perf_evsel { off_t id_offset; }; struct cgroup_sel *cgrp; + struct { + void *func; + void *data; + } handler; bool supported; }; diff --git a/tools/perf/util/tool.h b/tools/perf/util/tool.h index 89ff1b551a74..b0e1aadba8d5 100644 --- a/tools/perf/util/tool.h +++ b/tools/perf/util/tool.h @@ -1,8 +1,13 @@ #ifndef __PERF_TOOL_H #define __PERF_TOOL_H +#include + struct perf_session; +union perf_event; +struct perf_evlist; struct perf_evsel; +struct perf_sample; struct perf_tool; struct machine; -- cgit v1.2.3 From 806fb63007447622dd61d9767b4403919737e120 Mon Sep 17 00:00:00 2001 From: Arnaldo Carvalho de Melo Date: Tue, 29 Nov 2011 08:05:52 -0200 Subject: perf evlist: Always do automatic allocation of pollfd and mmap structures At first tools were required to do that, but while writing the python bindings to simplify the API I made them auto-allocate when needed. This just makes record, stat and top use that auto allocation, simplifying them a bit. Cc: David Ahern Cc: Frederic Weisbecker Cc: Mike Galbraith Cc: Paul Mackerras Cc: Peter Zijlstra Cc: Stephane Eranian Link: http://lkml.kernel.org/n/tip-iokhcvkzzijr3keioubx8hlq@git.kernel.org Signed-off-by: Arnaldo Carvalho de Melo --- tools/perf/builtin-record.c | 6 ------ tools/perf/builtin-stat.c | 3 +-- tools/perf/builtin-top.c | 19 ++++--------------- tools/perf/util/evlist.c | 4 ++-- tools/perf/util/evlist.h | 2 -- 5 files changed, 7 insertions(+), 27 deletions(-) diff --git a/tools/perf/builtin-record.c b/tools/perf/builtin-record.c index 7d4fdaacc8ba..766fa0a91a32 100644 --- a/tools/perf/builtin-record.c +++ b/tools/perf/builtin-record.c @@ -778,16 +778,10 @@ int cmd_record(int argc, const char **argv, const char *prefix __used) usage_with_options(record_usage, record_options); list_for_each_entry(pos, &evsel_list->entries, node) { - if (perf_evsel__alloc_fd(pos, evsel_list->cpus->nr, - evsel_list->threads->nr) < 0) - goto out_free_fd; if (perf_header__push_event(pos->attr.config, event_name(pos))) goto out_free_fd; } - if (perf_evlist__alloc_pollfd(evsel_list) < 0) - goto out_free_fd; - if (rec->opts.user_interval != ULLONG_MAX) rec->opts.default_interval = rec->opts.user_interval; if (rec->opts.user_freq != UINT_MAX) diff --git a/tools/perf/builtin-stat.c b/tools/perf/builtin-stat.c index 227befbecec8..4356144f1214 100644 --- a/tools/perf/builtin-stat.c +++ b/tools/perf/builtin-stat.c @@ -1240,8 +1240,7 @@ int cmd_stat(int argc, const char **argv, const char *prefix __used) list_for_each_entry(pos, &evsel_list->entries, node) { if (perf_evsel__alloc_stat_priv(pos) < 0 || - perf_evsel__alloc_counts(pos, evsel_list->cpus->nr) < 0 || - perf_evsel__alloc_fd(pos, evsel_list->cpus->nr, evsel_list->threads->nr) < 0) + perf_evsel__alloc_counts(pos, evsel_list->cpus->nr) < 0) goto out_free_fd; } diff --git a/tools/perf/builtin-top.c b/tools/perf/builtin-top.c index 50ff362ff012..c3836b966ccf 100644 --- a/tools/perf/builtin-top.c +++ b/tools/perf/builtin-top.c @@ -1034,9 +1034,7 @@ parse_callchain_opt(const struct option *opt, const char *arg, int unset) symbol_conf.use_callchain = false; return 0; - } - - else + } else return -1; /* get the min percentage */ @@ -1225,22 +1223,13 @@ int cmd_top(int argc, const char **argv, const char *prefix __used) } list_for_each_entry(pos, &top.evlist->entries, node) { - if (perf_evsel__alloc_fd(pos, top.evlist->cpus->nr, - top.evlist->threads->nr) < 0) - goto out_free_fd; /* * Fill in the ones not specifically initialized via -c: */ - if (pos->attr.sample_period) - continue; - - pos->attr.sample_period = top.default_interval; + if (!pos->attr.sample_period) + pos->attr.sample_period = top.default_interval; } - if (perf_evlist__alloc_pollfd(top.evlist) < 0 || - perf_evlist__alloc_mmap(top.evlist) < 0) - goto out_free_fd; - top.sym_evsel = list_entry(top.evlist->entries.next, struct perf_evsel, node); symbol_conf.priv_size = sizeof(struct annotation); @@ -1270,7 +1259,7 @@ int cmd_top(int argc, const char **argv, const char *prefix __used) } status = __cmd_top(&top); -out_free_fd: + perf_evlist__delete(top.evlist); return status; diff --git a/tools/perf/util/evlist.c b/tools/perf/util/evlist.c index b36f26fe767a..8b19e7a1e881 100644 --- a/tools/perf/util/evlist.c +++ b/tools/perf/util/evlist.c @@ -275,7 +275,7 @@ void perf_evlist__enable(struct perf_evlist *evlist) } } -int perf_evlist__alloc_pollfd(struct perf_evlist *evlist) +static int perf_evlist__alloc_pollfd(struct perf_evlist *evlist) { int nfds = evlist->cpus->nr * evlist->threads->nr * evlist->nr_entries; evlist->pollfd = malloc(sizeof(struct pollfd) * nfds); @@ -431,7 +431,7 @@ void perf_evlist__munmap(struct perf_evlist *evlist) evlist->mmap = NULL; } -int perf_evlist__alloc_mmap(struct perf_evlist *evlist) +static int perf_evlist__alloc_mmap(struct perf_evlist *evlist) { evlist->nr_mmaps = evlist->cpus->nr; if (evlist->cpus->map[0] == -1) diff --git a/tools/perf/util/evlist.h b/tools/perf/util/evlist.h index f94ed7e0d987..8922aeed0467 100644 --- a/tools/perf/util/evlist.h +++ b/tools/perf/util/evlist.h @@ -72,7 +72,6 @@ int perf_evlist__set_tracepoints_handlers(struct perf_evlist *evlist, void perf_evlist__id_add(struct perf_evlist *evlist, struct perf_evsel *evsel, int cpu, int thread, u64 id); -int perf_evlist__alloc_pollfd(struct perf_evlist *evlist); void perf_evlist__add_pollfd(struct perf_evlist *evlist, int fd); struct perf_evsel *perf_evlist__id2evsel(struct perf_evlist *evlist, u64 id); @@ -89,7 +88,6 @@ int perf_evlist__prepare_workload(struct perf_evlist *evlist, const char *argv[]); int perf_evlist__start_workload(struct perf_evlist *evlist); -int perf_evlist__alloc_mmap(struct perf_evlist *evlist); int perf_evlist__mmap(struct perf_evlist *evlist, unsigned int pages, bool overwrite); void perf_evlist__munmap(struct perf_evlist *evlist); -- cgit v1.2.3 From e60770a01bd889707faaaeb794f1e278e7160458 Mon Sep 17 00:00:00 2001 From: Arnaldo Carvalho de Melo Date: Tue, 29 Nov 2011 12:52:07 -0200 Subject: perf test: Allow running just a subset of the available tests To obtain a list of available tests: [root@emilia linux]# perf test list 1: vmlinux symtab matches kallsyms 2: detect open syscall event 3: detect open syscall event on all cpus 4: read samples using the mmap interface 5: parse events tests [root@emilia linux]# To list just a subset: [root@emilia linux]# perf test list syscall 2: detect open syscall event 3: detect open syscall event on all cpus [root@emilia linux]# To run a subset: [root@emilia linux]# perf test detect 2: detect open syscall event: Ok 3: detect open syscall event on all cpus: Ok [root@emilia linux]# Specific tests can be chosen by number: [root@emilia linux]# perf test 1 3 parse 1: vmlinux symtab matches kallsyms: Ok 3: detect open syscall event on all cpus: Ok 5: parse events tests: Ok [root@emilia linux]# Now to write more tests! Suggested-by: Peter Zijlstra Cc: David Ahern Cc: Frederic Weisbecker Cc: Mike Galbraith Cc: Paul Mackerras Cc: Peter Zijlstra Cc: Stephane Eranian Link: http://lkml.kernel.org/n/tip-nqec2145qfxdgimux28aw7v8@git.kernel.org Signed-off-by: Arnaldo Carvalho de Melo --- tools/perf/Documentation/perf-test.txt | 8 +++- tools/perf/builtin-test.c | 81 +++++++++++++++++++++++++--------- 2 files changed, 67 insertions(+), 22 deletions(-) diff --git a/tools/perf/Documentation/perf-test.txt b/tools/perf/Documentation/perf-test.txt index 2c3b462f64b0..b24ac40fcd58 100644 --- a/tools/perf/Documentation/perf-test.txt +++ b/tools/perf/Documentation/perf-test.txt @@ -8,13 +8,19 @@ perf-test - Runs sanity tests. SYNOPSIS -------- [verse] -'perf test ' +'perf test [] [{list |[|]}]' DESCRIPTION ----------- This command does assorted sanity tests, initially through linked routines but also will look for a directory with more tests in the form of scripts. +To get a list of available tests use 'perf test list', specifying a test name +fragment will show all tests that have it. + +To run just specific tests, inform test name fragments or the numbers obtained +from 'perf test list'. + OPTIONS ------- -v:: diff --git a/tools/perf/builtin-test.c b/tools/perf/builtin-test.c index 77d68bfb79da..3ab27223fc67 100644 --- a/tools/perf/builtin-test.c +++ b/tools/perf/builtin-test.c @@ -15,8 +15,6 @@ #include "util/thread_map.h" #include "../../include/linux/hw_breakpoint.h" -static long page_size; - static int vmlinux_matches_kallsyms_filter(struct map *map __used, struct symbol *sym) { bool *visited = symbol__priv(sym); @@ -32,6 +30,7 @@ static int test__vmlinux_matches_kallsyms(void) struct map *kallsyms_map, *vmlinux_map; struct machine kallsyms, vmlinux; enum map_type type = MAP__FUNCTION; + long page_size = sysconf(_SC_PAGE_SIZE); struct ref_reloc_sym ref_reloc_sym = { .name = "_stext", }; /* @@ -871,41 +870,81 @@ static struct test { }, }; -static int __cmd_test(void) +static bool perf_test__matches(int curr, int argc, const char *argv[]) { - int i = 0; + int i; + + if (argc == 0) + return true; + + for (i = 0; i < argc; ++i) { + char *end; + long nr = strtoul(argv[i], &end, 10); + + if (*end == '\0') { + if (nr == curr + 1) + return true; + continue; + } - page_size = sysconf(_SC_PAGE_SIZE); + if (strstr(tests[curr].desc, argv[i])) + return true; + } + + return false; +} + +static int __cmd_test(int argc, const char *argv[]) +{ + int i = 0; while (tests[i].func) { - int err; - pr_info("%2d: %s:", i + 1, tests[i].desc); + int curr = i++, err; + + if (!perf_test__matches(curr, argc, argv)) + continue; + + pr_info("%2d: %s:", i, tests[curr].desc); pr_debug("\n--- start ---\n"); - err = tests[i].func(); - pr_debug("---- end ----\n%s:", tests[i].desc); + err = tests[curr].func(); + pr_debug("---- end ----\n%s:", tests[curr].desc); pr_info(" %s\n", err ? "FAILED!\n" : "Ok"); - ++i; } return 0; } -static const char * const test_usage[] = { - "perf test []", - NULL, -}; +static int perf_test__list(int argc, const char **argv) +{ + int i = 0; + + while (tests[i].func) { + int curr = i++; + + if (argc > 1 && !strstr(tests[curr].desc, argv[1])) + continue; + + pr_info("%2d: %s\n", i, tests[curr].desc); + } + + return 0; +} -static const struct option test_options[] = { +int cmd_test(int argc, const char **argv, const char *prefix __used) +{ + const char * const test_usage[] = { + "perf test [] [{list |[|]}]", + NULL, + }; + const struct option test_options[] = { OPT_INTEGER('v', "verbose", &verbose, "be more verbose (show symbol address, etc)"), OPT_END() -}; + }; -int cmd_test(int argc, const char **argv, const char *prefix __used) -{ argc = parse_options(argc, argv, test_options, test_usage, 0); - if (argc) - usage_with_options(test_usage, test_options); + if (argc >= 1 && !strcmp(argv[0], "list")) + return perf_test__list(argc, argv); symbol_conf.priv_size = sizeof(int); symbol_conf.sort_by_name = true; @@ -916,5 +955,5 @@ int cmd_test(int argc, const char **argv, const char *prefix __used) setup_pager(); - return __cmd_test(); + return __cmd_test(argc, argv); } -- cgit v1.2.3 From 482ad89745f7121020f6cee38aa4e894a4e7d642 Mon Sep 17 00:00:00 2001 From: Arnaldo Carvalho de Melo Date: Fri, 2 Dec 2011 11:06:37 -0200 Subject: perf event: Introduce perf_event__fprintf So that tools like 'perf test' can print the events when in verbose mode, for instance. Cc: David Ahern Cc: Frederic Weisbecker Cc: Mike Galbraith Cc: Paul Mackerras Cc: Peter Zijlstra Cc: Stephane Eranian Link: http://lkml.kernel.org/n/tip-xnovdqfi25nc48gy6604k7yp@git.kernel.org Signed-off-by: Arnaldo Carvalho de Melo --- tools/perf/util/event.c | 54 +++++++++++++++++++++++++++++++++++++++++++------ tools/perf/util/event.h | 6 ++++++ 2 files changed, 54 insertions(+), 6 deletions(-) diff --git a/tools/perf/util/event.c b/tools/perf/util/event.c index 0ebbe7641335..97c479bcb0dc 100644 --- a/tools/perf/util/event.c +++ b/tools/perf/util/event.c @@ -433,6 +433,11 @@ int perf_event__synthesize_kernel_mmap(struct perf_tool *tool, return err; } +size_t perf_event__fprintf_comm(union perf_event *event, FILE *fp) +{ + return fprintf(fp, ": %s:%d\n", event->comm.comm, event->comm.tid); +} + int perf_event__process_comm(struct perf_tool *tool __used, union perf_event *event, struct perf_sample *sample __used, @@ -440,7 +445,8 @@ int perf_event__process_comm(struct perf_tool *tool __used, { struct thread *thread = machine__findnew_thread(machine, event->comm.tid); - dump_printf(": %s:%d\n", event->comm.comm, event->comm.tid); + if (dump_trace) + perf_event__fprintf_comm(event, stdout); if (thread == NULL || thread__set_comm(thread, event->comm.comm)) { dump_printf("problem processing PERF_RECORD_COMM, skipping event.\n"); @@ -566,6 +572,13 @@ out_problem: return -1; } +size_t perf_event__fprintf_mmap(union perf_event *event, FILE *fp) +{ + return fprintf(fp, " %d/%d: [%#" PRIx64 "(%#" PRIx64 ") @ %#" PRIx64 "]: %s\n", + event->mmap.pid, event->mmap.tid, event->mmap.start, + event->mmap.len, event->mmap.pgoff, event->mmap.filename); +} + int perf_event__process_mmap(struct perf_tool *tool, union perf_event *event, struct perf_sample *sample __used, @@ -576,9 +589,8 @@ int perf_event__process_mmap(struct perf_tool *tool, u8 cpumode = event->header.misc & PERF_RECORD_MISC_CPUMODE_MASK; int ret = 0; - dump_printf(" %d/%d: [%#" PRIx64 "(%#" PRIx64 ") @ %#" PRIx64 "]: %s\n", - event->mmap.pid, event->mmap.tid, event->mmap.start, - event->mmap.len, event->mmap.pgoff, event->mmap.filename); + if (dump_trace) + perf_event__fprintf_mmap(event, stdout); if (cpumode == PERF_RECORD_MISC_GUEST_KERNEL || cpumode == PERF_RECORD_MISC_KERNEL) { @@ -606,6 +618,13 @@ out_problem: return 0; } +size_t perf_event__fprintf_task(union perf_event *event, FILE *fp) +{ + return fprintf(fp, "(%d:%d):(%d:%d)\n", + event->fork.pid, event->fork.tid, + event->fork.ppid, event->fork.ptid); +} + int perf_event__process_task(struct perf_tool *tool __used, union perf_event *event, struct perf_sample *sample __used, @@ -614,8 +633,8 @@ int perf_event__process_task(struct perf_tool *tool __used, struct thread *thread = machine__findnew_thread(machine, event->fork.tid); struct thread *parent = machine__findnew_thread(machine, event->fork.ptid); - dump_printf("(%d:%d):(%d:%d)\n", event->fork.pid, event->fork.tid, - event->fork.ppid, event->fork.ptid); + if (dump_trace) + perf_event__fprintf_task(event, stdout); if (event->header.type == PERF_RECORD_EXIT) { machine__remove_thread(machine, thread); @@ -631,6 +650,29 @@ int perf_event__process_task(struct perf_tool *tool __used, return 0; } +size_t perf_event__fprintf(union perf_event *event, FILE *fp) +{ + size_t ret = fprintf(fp, "PERF_RECORD_%s", + perf_event__name(event->header.type)); + + switch (event->header.type) { + case PERF_RECORD_COMM: + ret += perf_event__fprintf_comm(event, fp); + break; + case PERF_RECORD_FORK: + case PERF_RECORD_EXIT: + ret += perf_event__fprintf_task(event, fp); + break; + case PERF_RECORD_MMAP: + ret += perf_event__fprintf_mmap(event, fp); + break; + default: + ret += fprintf(fp, "\n"); + } + + return ret; +} + int perf_event__process(struct perf_tool *tool, union perf_event *event, struct perf_sample *sample, struct machine *machine) { diff --git a/tools/perf/util/event.h b/tools/perf/util/event.h index d8499e7cf641..0d80201ce844 100644 --- a/tools/perf/util/event.h +++ b/tools/perf/util/event.h @@ -2,6 +2,7 @@ #define __PERF_RECORD_H #include +#include #include "../perf.h" #include "map.h" @@ -199,4 +200,9 @@ int perf_event__parse_sample(const union perf_event *event, u64 type, int sample_size, bool sample_id_all, struct perf_sample *sample, bool swapped); +size_t perf_event__fprintf_comm(union perf_event *event, FILE *fp); +size_t perf_event__fprintf_mmap(union perf_event *event, FILE *fp); +size_t perf_event__fprintf_task(union perf_event *event, FILE *fp); +size_t perf_event__fprintf(union perf_event *event, FILE *fp); + #endif /* __PERF_RECORD_H */ -- cgit v1.2.3 From 3e7c439a7ce537ed662e347b9e6414d7881fb3dc Mon Sep 17 00:00:00 2001 From: Arnaldo Carvalho de Melo Date: Fri, 2 Dec 2011 11:13:50 -0200 Subject: perf test: Validate PERF_RECORD_ events and perf_sample fields This new test will validate these new routines extracted from 'perf record': - perf_evlist__config_attrs - perf_evlist__prepare_workload - perf_evlist__start_workload In addition to several other perf_evlist methods. It consists of starting a simple workload, setting up just one event to monitor ("cycles") requesting that several PERF_SAMPLE_ fields be present in all events. It then will check that the expected PERF_RECORD_ events are produced and will sanity check all its fields. Some checks performed: . PERF_SAMPLE_TIME monotonically increases. . PERF_SAMPLE_CPU is the one requested with sched_setaffinity . PERF_SAMPLE_TID and PERF_SAMPLE_PID matches the one we forked in perf_evlist__prepare_workload and that is stored in evlist->workload.pid . For the events where these fields are also present in its pre-sample_id_all fields (e.g. event->mmap.pid), that they are what is expected too. . That we get a bunch of mmaps: PATH/libcSUFFIX PATH/ldSUFFIX [vdso] PATH/sleep Example: [root@emilia ~]# taskset -c 3,4 perf test -v1 perf_sample 6: Validate PERF_RECORD_* events & perf_sample fields: --- start --- 7159480799825 3 PERF_RECORD_SAMPLE 7159480805584 3 PERF_RECORD_SAMPLE 7159480807814 3 PERF_RECORD_SAMPLE 7159480810430 3 PERF_RECORD_SAMPLE 7159480861511 3 PERF_RECORD_MMAP 8086/8086: [0x7fffffffd000(0x2000) @ 0x7fffffffd000]: //anon 7159481052516 3 PERF_RECORD_COMM: sleep:8086 7159481070188 3 PERF_RECORD_MMAP 8086/8086: [0x400000(0x6000) @ 0]: /bin/sleep 7159481077104 3 PERF_RECORD_MMAP 8086/8086: [0x3d06400000(0x221000) @ 0]: /lib64/ld-2.12.so 7159481092912 3 PERF_RECORD_MMAP 8086/8086: [0x7fff1adff000(0x1000) @ 0x7fff1adff000]: [vdso] 7159481196779 3 PERF_RECORD_MMAP 8086/8086: [0x3d06800000(0x37f000) @ 0]: /lib64/libc-2.12.so 7160481558435 3 PERF_RECORD_EXIT(8086:8086):(8086:8086) ---- end ---- Validate PERF_RECORD_* events & perf_sample fields: Ok [root@emilia ~]# Cc: David Ahern Cc: Frederic Weisbecker Cc: Mike Galbraith Cc: Paul Mackerras Cc: Peter Zijlstra Cc: Stephane Eranian Link: http://lkml.kernel.org/n/tip-svag18v2z4idas0dyz3umjpq@git.kernel.org Signed-off-by: Arnaldo Carvalho de Melo --- tools/perf/builtin-test.c | 337 ++++++++++++++++++++++++++++++++++++++++++++++ 1 file changed, 337 insertions(+) diff --git a/tools/perf/builtin-test.c b/tools/perf/builtin-test.c index 3ab27223fc67..f1e36110b101 100644 --- a/tools/perf/builtin-test.c +++ b/tools/perf/builtin-test.c @@ -841,6 +841,339 @@ static int test__parse_events(void) return ret; } + +static int sched__get_first_possible_cpu(pid_t pid, cpu_set_t **maskp, + size_t *sizep) +{ + cpu_set_t *mask; + size_t size; + int i, cpu = -1, nrcpus = 1024; +realloc: + mask = CPU_ALLOC(nrcpus); + size = CPU_ALLOC_SIZE(nrcpus); + CPU_ZERO_S(size, mask); + + if (sched_getaffinity(pid, size, mask) == -1) { + CPU_FREE(mask); + if (errno == EINVAL && nrcpus < (1024 << 8)) { + nrcpus = nrcpus << 2; + goto realloc; + } + perror("sched_getaffinity"); + return -1; + } + + for (i = 0; i < nrcpus; i++) { + if (CPU_ISSET_S(i, size, mask)) { + if (cpu == -1) { + cpu = i; + *maskp = mask; + *sizep = size; + } else + CPU_CLR_S(i, size, mask); + } + } + + if (cpu == -1) + CPU_FREE(mask); + + return cpu; +} + +static int test__PERF_RECORD(void) +{ + struct perf_record_opts opts = { + .target_pid = -1, + .target_tid = -1, + .no_delay = true, + .freq = 10, + .mmap_pages = 256, + .sample_id_all_avail = true, + }; + cpu_set_t *cpu_mask = NULL; + size_t cpu_mask_size = 0; + struct perf_evlist *evlist = perf_evlist__new(NULL, NULL); + struct perf_evsel *evsel; + struct perf_sample sample; + const char *cmd = "sleep"; + const char *argv[] = { cmd, "1", NULL, }; + char *bname; + u64 sample_type, prev_time = 0; + bool found_cmd_mmap = false, + found_libc_mmap = false, + found_vdso_mmap = false, + found_ld_mmap = false; + int err = -1, i, wakeups = 0, sample_size; + u32 cpu; + int total_events = 0, nr_events[PERF_RECORD_MAX] = { 0, }; + + if (evlist == NULL || argv == NULL) { + pr_debug("Not enough memory to create evlist\n"); + goto out; + } + + /* + * We need at least one evsel in the evlist, use the default + * one: "cycles". + */ + err = perf_evlist__add_default(evlist); + if (err < 0) { + pr_debug("Not enough memory to create evsel\n"); + goto out_delete_evlist; + } + + /* + * Create maps of threads and cpus to monitor. In this case + * we start with all threads and cpus (-1, -1) but then in + * perf_evlist__prepare_workload we'll fill in the only thread + * we're monitoring, the one forked there. + */ + err = perf_evlist__create_maps(evlist, opts.target_pid, + opts.target_tid, opts.cpu_list); + if (err < 0) { + pr_debug("Not enough memory to create thread/cpu maps\n"); + goto out_delete_evlist; + } + + /* + * Prepare the workload in argv[] to run, it'll fork it, and then wait + * for perf_evlist__start_workload() to exec it. This is done this way + * so that we have time to open the evlist (calling sys_perf_event_open + * on all the fds) and then mmap them. + */ + err = perf_evlist__prepare_workload(evlist, &opts, argv); + if (err < 0) { + pr_debug("Couldn't run the workload!\n"); + goto out_delete_evlist; + } + + /* + * Config the evsels, setting attr->comm on the first one, etc. + */ + evsel = list_entry(evlist->entries.next, struct perf_evsel, node); + evsel->attr.sample_type |= PERF_SAMPLE_CPU; + evsel->attr.sample_type |= PERF_SAMPLE_TID; + evsel->attr.sample_type |= PERF_SAMPLE_TIME; + perf_evlist__config_attrs(evlist, &opts); + + err = sched__get_first_possible_cpu(evlist->workload.pid, &cpu_mask, + &cpu_mask_size); + if (err < 0) { + pr_debug("sched__get_first_possible_cpu: %s\n", strerror(errno)); + goto out_delete_evlist; + } + + cpu = err; + + /* + * So that we can check perf_sample.cpu on all the samples. + */ + if (sched_setaffinity(evlist->workload.pid, cpu_mask_size, cpu_mask) < 0) { + pr_debug("sched_setaffinity: %s\n", strerror(errno)); + goto out_free_cpu_mask; + } + + /* + * Call sys_perf_event_open on all the fds on all the evsels, + * grouping them if asked to. + */ + err = perf_evlist__open(evlist, opts.group); + if (err < 0) { + pr_debug("perf_evlist__open: %s\n", strerror(errno)); + goto out_delete_evlist; + } + + /* + * mmap the first fd on a given CPU and ask for events for the other + * fds in the same CPU to be injected in the same mmap ring buffer + * (using ioctl(PERF_EVENT_IOC_SET_OUTPUT)). + */ + err = perf_evlist__mmap(evlist, opts.mmap_pages, false); + if (err < 0) { + pr_debug("perf_evlist__mmap: %s\n", strerror(errno)); + goto out_delete_evlist; + } + + /* + * We'll need these two to parse the PERF_SAMPLE_* fields in each + * event. + */ + sample_type = perf_evlist__sample_type(evlist); + sample_size = __perf_evsel__sample_size(sample_type); + + /* + * Now that all is properly set up, enable the events, they will + * count just on workload.pid, which will start... + */ + perf_evlist__enable(evlist); + + /* + * Now! + */ + perf_evlist__start_workload(evlist); + + err = -1; + + while (1) { + int before = total_events; + + for (i = 0; i < evlist->nr_mmaps; i++) { + union perf_event *event; + + while ((event = perf_evlist__mmap_read(evlist, i)) != NULL) { + const u32 type = event->header.type; + const char *name = perf_event__name(type); + + ++total_events; + if (type < PERF_RECORD_MAX) + nr_events[type]++; + + if (perf_event__parse_sample(event, sample_type, + sample_size, true, + &sample, false) < 0) { + if (verbose) + perf_event__fprintf(event, stderr); + pr_debug("Couldn't parse sample\n"); + goto out_err; + } + + if (verbose) { + pr_info("%" PRIu64" %d ", sample.time, sample.cpu); + perf_event__fprintf(event, stderr); + } + + if (prev_time > sample.time) { + pr_debug("%s going backwards in time, prev=%" PRIu64 ", curr=%" PRIu64 "\n", + name, prev_time, sample.time); + goto out_err; + } + + prev_time = sample.time; + + if (sample.cpu != cpu) { + pr_debug("%s with unexpected cpu, expected %d, got %d\n", + name, cpu, sample.cpu); + goto out_err; + } + + if ((pid_t)sample.pid != evlist->workload.pid) { + pr_debug("%s with unexpected pid, expected %d, got %d\n", + name, evlist->workload.pid, sample.pid); + goto out_err; + } + + if ((pid_t)sample.tid != evlist->workload.pid) { + pr_debug("%s with unexpected tid, expected %d, got %d\n", + name, evlist->workload.pid, sample.tid); + goto out_err; + } + + if ((type == PERF_RECORD_COMM || + type == PERF_RECORD_MMAP || + type == PERF_RECORD_FORK || + type == PERF_RECORD_EXIT) && + (pid_t)event->comm.pid != evlist->workload.pid) { + pr_debug("%s with unexpected pid/tid\n", name); + goto out_err; + } + + if ((type == PERF_RECORD_COMM || + type == PERF_RECORD_MMAP) && + event->comm.pid != event->comm.tid) { + pr_debug("%s with different pid/tid!\n", name); + goto out_err; + } + + switch (type) { + case PERF_RECORD_COMM: + if (strcmp(event->comm.comm, cmd)) { + pr_debug("%s with unexpected comm!\n", name); + goto out_err; + } + break; + case PERF_RECORD_EXIT: + goto found_exit; + case PERF_RECORD_MMAP: + bname = strrchr(event->mmap.filename, '/'); + if (bname != NULL) { + if (!found_cmd_mmap) + found_cmd_mmap = !strcmp(bname + 1, cmd); + if (!found_libc_mmap) + found_libc_mmap = !strncmp(bname + 1, "libc", 4); + if (!found_ld_mmap) + found_ld_mmap = !strncmp(bname + 1, "ld", 2); + } else if (!found_vdso_mmap) + found_vdso_mmap = !strcmp(event->mmap.filename, "[vdso]"); + break; + + case PERF_RECORD_SAMPLE: + /* Just ignore samples for now */ + break; + default: + pr_debug("Unexpected perf_event->header.type %d!\n", + type); + goto out_err; + } + } + } + + /* + * We don't use poll here because at least at 3.1 times the + * PERF_RECORD_{!SAMPLE} events don't honour + * perf_event_attr.wakeup_events, just PERF_EVENT_SAMPLE does. + */ + if (total_events == before && false) + poll(evlist->pollfd, evlist->nr_fds, -1); + + sleep(1); + if (++wakeups > 5) { + pr_debug("No PERF_RECORD_EXIT event!\n"); + goto out_err; + } + } + +found_exit: + if (nr_events[PERF_RECORD_COMM] > 1) { + pr_debug("Excessive number of PERF_RECORD_COMM events!\n"); + goto out_err; + } + + if (nr_events[PERF_RECORD_COMM] == 0) { + pr_debug("Missing PERF_RECORD_COMM for %s!\n", cmd); + goto out_err; + } + + if (!found_cmd_mmap) { + pr_debug("PERF_RECORD_MMAP for %s missing!\n", cmd); + goto out_err; + } + + if (!found_libc_mmap) { + pr_debug("PERF_RECORD_MMAP for %s missing!\n", "libc"); + goto out_err; + } + + if (!found_ld_mmap) { + pr_debug("PERF_RECORD_MMAP for %s missing!\n", "ld"); + goto out_err; + } + + if (!found_vdso_mmap) { + pr_debug("PERF_RECORD_MMAP for %s missing!\n", "[vdso]"); + goto out_err; + } + + err = 0; +out_err: + perf_evlist__munmap(evlist); +out_free_cpu_mask: + CPU_FREE(cpu_mask); +out_delete_evlist: + perf_evlist__delete(evlist); +out: + return err; +} + static struct test { const char *desc; int (*func)(void); @@ -865,6 +1198,10 @@ static struct test { .desc = "parse events tests", .func = test__parse_events, }, + { + .desc = "Validate PERF_RECORD_* events & perf_sample fields", + .func = test__PERF_RECORD, + }, { .func = NULL, }, -- cgit v1.2.3 From f71c49e5e985897850406a152ab99001cfb86f2a Mon Sep 17 00:00:00 2001 From: Arnaldo Carvalho de Melo Date: Fri, 2 Dec 2011 13:53:04 -0200 Subject: perf test: Soft errors shouldn't stop the "Validate PERF_RECORD_" test For errors that don't preclude checking for further errors, aka "soft" errors, just continue testing for other errors. Better coverage in verbose mode. Suggested-by: David Ahern Cc: David Ahern Cc: Frederic Weisbecker Cc: Mike Galbraith Cc: Paul Mackerras Cc: Peter Zijlstra Cc: Stephane Eranian Link: http://lkml.kernel.org/n/tip-jafcokbj26m845dsgm2hx6az@git.kernel.org Signed-off-by: Arnaldo Carvalho de Melo --- tools/perf/builtin-test.c | 45 +++++++++++++++++++++------------------------ 1 file changed, 21 insertions(+), 24 deletions(-) diff --git a/tools/perf/builtin-test.c b/tools/perf/builtin-test.c index f1e36110b101..6173f780dce0 100644 --- a/tools/perf/builtin-test.c +++ b/tools/perf/builtin-test.c @@ -903,7 +903,7 @@ static int test__PERF_RECORD(void) found_libc_mmap = false, found_vdso_mmap = false, found_ld_mmap = false; - int err = -1, i, wakeups = 0, sample_size; + int err = -1, errs = 0, i, wakeups = 0, sample_size; u32 cpu; int total_events = 0, nr_events[PERF_RECORD_MAX] = { 0, }; @@ -1012,8 +1012,6 @@ static int test__PERF_RECORD(void) */ perf_evlist__start_workload(evlist); - err = -1; - while (1) { int before = total_events; @@ -1028,9 +1026,10 @@ static int test__PERF_RECORD(void) if (type < PERF_RECORD_MAX) nr_events[type]++; - if (perf_event__parse_sample(event, sample_type, - sample_size, true, - &sample, false) < 0) { + err = perf_event__parse_sample(event, sample_type, + sample_size, true, + &sample, false); + if (err < 0) { if (verbose) perf_event__fprintf(event, stderr); pr_debug("Couldn't parse sample\n"); @@ -1045,7 +1044,7 @@ static int test__PERF_RECORD(void) if (prev_time > sample.time) { pr_debug("%s going backwards in time, prev=%" PRIu64 ", curr=%" PRIu64 "\n", name, prev_time, sample.time); - goto out_err; + ++errs; } prev_time = sample.time; @@ -1053,19 +1052,19 @@ static int test__PERF_RECORD(void) if (sample.cpu != cpu) { pr_debug("%s with unexpected cpu, expected %d, got %d\n", name, cpu, sample.cpu); - goto out_err; + ++errs; } if ((pid_t)sample.pid != evlist->workload.pid) { pr_debug("%s with unexpected pid, expected %d, got %d\n", name, evlist->workload.pid, sample.pid); - goto out_err; + ++errs; } if ((pid_t)sample.tid != evlist->workload.pid) { pr_debug("%s with unexpected tid, expected %d, got %d\n", name, evlist->workload.pid, sample.tid); - goto out_err; + ++errs; } if ((type == PERF_RECORD_COMM || @@ -1074,21 +1073,21 @@ static int test__PERF_RECORD(void) type == PERF_RECORD_EXIT) && (pid_t)event->comm.pid != evlist->workload.pid) { pr_debug("%s with unexpected pid/tid\n", name); - goto out_err; + ++errs; } if ((type == PERF_RECORD_COMM || type == PERF_RECORD_MMAP) && event->comm.pid != event->comm.tid) { pr_debug("%s with different pid/tid!\n", name); - goto out_err; + ++errs; } switch (type) { case PERF_RECORD_COMM: if (strcmp(event->comm.comm, cmd)) { pr_debug("%s with unexpected comm!\n", name); - goto out_err; + ++errs; } break; case PERF_RECORD_EXIT: @@ -1112,7 +1111,7 @@ static int test__PERF_RECORD(void) default: pr_debug("Unexpected perf_event->header.type %d!\n", type); - goto out_err; + ++errs; } } } @@ -1128,42 +1127,40 @@ static int test__PERF_RECORD(void) sleep(1); if (++wakeups > 5) { pr_debug("No PERF_RECORD_EXIT event!\n"); - goto out_err; + break; } } found_exit: if (nr_events[PERF_RECORD_COMM] > 1) { pr_debug("Excessive number of PERF_RECORD_COMM events!\n"); - goto out_err; + ++errs; } if (nr_events[PERF_RECORD_COMM] == 0) { pr_debug("Missing PERF_RECORD_COMM for %s!\n", cmd); - goto out_err; + ++errs; } if (!found_cmd_mmap) { pr_debug("PERF_RECORD_MMAP for %s missing!\n", cmd); - goto out_err; + ++errs; } if (!found_libc_mmap) { pr_debug("PERF_RECORD_MMAP for %s missing!\n", "libc"); - goto out_err; + ++errs; } if (!found_ld_mmap) { pr_debug("PERF_RECORD_MMAP for %s missing!\n", "ld"); - goto out_err; + ++errs; } if (!found_vdso_mmap) { pr_debug("PERF_RECORD_MMAP for %s missing!\n", "[vdso]"); - goto out_err; + ++errs; } - - err = 0; out_err: perf_evlist__munmap(evlist); out_free_cpu_mask: @@ -1171,7 +1168,7 @@ out_free_cpu_mask: out_delete_evlist: perf_evlist__delete(evlist); out: - return err; + return (err < 0 || errs > 0) ? -1 : 0; } static struct test { -- cgit v1.2.3 From 1056c3e916f12cdd8042ab27dfccbb3a9e871df0 Mon Sep 17 00:00:00 2001 From: Masami Hiramatsu Date: Mon, 5 Dec 2011 21:05:33 +0900 Subject: x86/tools: Fix Makefile to build all test tools Fix arch/x86/tools/Makefile to compile both test tools correctly. This bug leads build error. Signed-off-by: Masami Hiramatsu Cc: "H. Peter Anvin" Cc: yrl.pp-manager.tt@hitachi.com Link: http://lkml.kernel.org/r/20111205120533.15475.62047.stgit@cloud Signed-off-by: Ingo Molnar --- arch/x86/tools/Makefile | 3 +-- 1 file changed, 1 insertion(+), 2 deletions(-) diff --git a/arch/x86/tools/Makefile b/arch/x86/tools/Makefile index 3255c3df67f4..d511aa97533a 100644 --- a/arch/x86/tools/Makefile +++ b/arch/x86/tools/Makefile @@ -25,8 +25,7 @@ posttest: $(obj)/test_get_len vmlinux $(obj)/insn_sanity $(call cmd,posttest) $(call cmd,sanitytest) -hostprogs-y := test_get_len -hostprogs-y := insn_sanity +hostprogs-y += test_get_len insn_sanity # -I needed for generated C source and C source which in the kernel tree. HOSTCFLAGS_test_get_len.o := -Wall -I$(objtree)/arch/x86/lib/ -I$(srctree)/arch/x86/include/ -I$(srctree)/arch/x86/lib/ -I$(srctree)/include/ -- cgit v1.2.3 From 130b78b2bf16d5d89091db38374faef896360cf9 Mon Sep 17 00:00:00 2001 From: Masami Hiramatsu Date: Mon, 5 Dec 2011 21:05:39 +0900 Subject: x86: Fix instruction decoder to handle grouped AVX instructions For reducing memory usage of attribute table, x86 instruction decoder puts "Group" attribute only on "no-last-prefix" attribute table (same as vex_p == 0 case). Thus, the decoder should look no-last-prefix table first, and then only if it is not a group, move on to "with-last-prefix" table (vex_p != 0). However, current implementation, inat_get_avx_attribute() looks with-last-prefix directly. So, when decoding a grouped AVX instruction, the decoder fails to find correct group because there is no "Group" attribute on the table. This ends up with the mis-decoding of instructions, as Ingo reported in http://thread.gmane.org/gmane.linux.kernel/1214103 This patch fixes it to check no-last-prefix table first even if that is an AVX instruction, and get an attribute from "with last-prefix" table only if that is not a group. Reported-by: Ingo Molnar Signed-off-by: Masami Hiramatsu Cc: "H. Peter Anvin" Cc: yrl.pp-manager.tt@hitachi.com Link: http://lkml.kernel.org/r/20111205120539.15475.91428.stgit@cloud Signed-off-by: Ingo Molnar --- arch/x86/lib/inat.c | 9 ++++++++- arch/x86/lib/insn.c | 4 +++- 2 files changed, 11 insertions(+), 2 deletions(-) diff --git a/arch/x86/lib/inat.c b/arch/x86/lib/inat.c index 46fc4ee09fc4..88ad5fbda6e1 100644 --- a/arch/x86/lib/inat.c +++ b/arch/x86/lib/inat.c @@ -82,9 +82,16 @@ insn_attr_t inat_get_avx_attribute(insn_byte_t opcode, insn_byte_t vex_m, const insn_attr_t *table; if (vex_m > X86_VEX_M_MAX || vex_p > INAT_LSTPFX_MAX) return 0; - table = inat_avx_tables[vex_m][vex_p]; + /* At first, this checks the master table */ + table = inat_avx_tables[vex_m][0]; if (!table) return 0; + if (!inat_is_group(table[opcode]) && vex_p) { + /* If this is not a group, get attribute directly */ + table = inat_avx_tables[vex_m][vex_p]; + if (!table) + return 0; + } return table[opcode]; } diff --git a/arch/x86/lib/insn.c b/arch/x86/lib/insn.c index 374562ed6704..5a1f9f3e3fbb 100644 --- a/arch/x86/lib/insn.c +++ b/arch/x86/lib/insn.c @@ -202,7 +202,7 @@ void insn_get_opcode(struct insn *insn) m = insn_vex_m_bits(insn); p = insn_vex_p_bits(insn); insn->attr = inat_get_avx_attribute(op, m, p); - if (!inat_accept_vex(insn->attr)) + if (!inat_accept_vex(insn->attr) && !inat_is_group(insn->attr)) insn->attr = 0; /* This instruction is bad */ goto end; /* VEX has only 1 byte for opcode */ } @@ -249,6 +249,8 @@ void insn_get_modrm(struct insn *insn) pfx = insn_last_prefix(insn); insn->attr = inat_get_group_attribute(mod, pfx, insn->attr); + if (insn_is_avx(insn) && !inat_accept_vex(insn->attr)) + insn->attr = 0; /* This is bad */ } } -- cgit v1.2.3 From bfbe9015de5c78d1808cd09526b9166b2e6aa440 Mon Sep 17 00:00:00 2001 From: Masami Hiramatsu Date: Mon, 5 Dec 2011 21:05:45 +0900 Subject: x86/tools: Fix instruction decoder message output Fix instruction decoder test (insn_sanity), so that it doesn't show both info and error messages twice on same instruction. (In that case, show only error message) Signed-off-by: Masami Hiramatsu Cc: "H. Peter Anvin" Cc: yrl.pp-manager.tt@hitachi.com Link: http://lkml.kernel.org/r/20111205120545.15475.7928.stgit@cloud Signed-off-by: Ingo Molnar --- arch/x86/tools/insn_sanity.c | 7 +++---- 1 file changed, 3 insertions(+), 4 deletions(-) diff --git a/arch/x86/tools/insn_sanity.c b/arch/x86/tools/insn_sanity.c index 334d9de7d0ca..20256037ac7d 100644 --- a/arch/x86/tools/insn_sanity.c +++ b/arch/x86/tools/insn_sanity.c @@ -257,15 +257,14 @@ int main(int argc, char **argv) insn_init(&insn, insn_buf, x86_64); insn_get_length(&insn); - if (verbose && !insn_complete(&insn)) - dump_stream(stdout, "Info: Found an undecodable input", i, insn_buf, &insn); - if (insn.next_byte <= insn.kaddr || insn.kaddr + MAX_INSN_SIZE < insn.next_byte) { /* Access out-of-range memory */ dump_stream(stdout, "Error: Found an access violation", i, insn_buf, &insn); errors++; - } + } else if (verbose && !insn_complete(&insn)) + dump_stream(stdout, "Info: Found an undecodable input", i, insn_buf, &insn); + insns++; } -- cgit v1.2.3 From e70825fc51e149366ab5659bd36beb73aad187a0 Mon Sep 17 00:00:00 2001 From: Masami Hiramatsu Date: Mon, 5 Dec 2011 21:05:50 +0900 Subject: x86/tools: Fix insn_sanity message outputs Fix x86 instruction decoder test to dump all error messages to stderr and others to stdout. Signed-off-by: Masami Hiramatsu Cc: "H. Peter Anvin" Cc: yrl.pp-manager.tt@hitachi.com Link: http://lkml.kernel.org/r/20111205120550.15475.70149.stgit@cloud Signed-off-by: Ingo Molnar --- arch/x86/tools/insn_sanity.c | 4 ++-- 1 file changed, 2 insertions(+), 2 deletions(-) diff --git a/arch/x86/tools/insn_sanity.c b/arch/x86/tools/insn_sanity.c index 20256037ac7d..b6720d6b38cb 100644 --- a/arch/x86/tools/insn_sanity.c +++ b/arch/x86/tools/insn_sanity.c @@ -102,7 +102,7 @@ static void dump_stream(FILE *fp, const char *msg, unsigned long nr_iter, fprintf(fp, "%s:\n", msg); - dump_insn(stderr, insn); + dump_insn(fp, insn); fprintf(fp, "You can reproduce this with below command(s);\n"); @@ -260,7 +260,7 @@ int main(int argc, char **argv) if (insn.next_byte <= insn.kaddr || insn.kaddr + MAX_INSN_SIZE < insn.next_byte) { /* Access out-of-range memory */ - dump_stream(stdout, "Error: Found an access violation", i, insn_buf, &insn); + dump_stream(stderr, "Error: Found an access violation", i, insn_buf, &insn); errors++; } else if (verbose && !insn_complete(&insn)) dump_stream(stdout, "Info: Found an undecodable input", i, insn_buf, &insn); -- cgit v1.2.3 From a9c373d03326e98c5f05ca64a1108790d25e28a9 Mon Sep 17 00:00:00 2001 From: Masami Hiramatsu Date: Mon, 5 Dec 2011 21:05:57 +0900 Subject: x86: Update instruction decoder to support new AVX formats Since new Intel software developers manual introduces new format for AVX instruction set (including AVX2), it is important to update x86-opcode-map.txt to fit those changes. Signed-off-by: Masami Hiramatsu Cc: "H. Peter Anvin" Cc: yrl.pp-manager.tt@hitachi.com Link: http://lkml.kernel.org/r/20111205120557.15475.13236.stgit@cloud Signed-off-by: Ingo Molnar --- arch/x86/lib/x86-opcode-map.txt | 606 +++++++++++++++++++---------------- arch/x86/tools/gen-insn-attr-x86.awk | 21 +- 2 files changed, 345 insertions(+), 282 deletions(-) diff --git a/arch/x86/lib/x86-opcode-map.txt b/arch/x86/lib/x86-opcode-map.txt index a793da5e560e..5b83c51c12e0 100644 --- a/arch/x86/lib/x86-opcode-map.txt +++ b/arch/x86/lib/x86-opcode-map.txt @@ -1,5 +1,11 @@ # x86 Opcode Maps # +# This is (mostly) based on following documentations. +# - Intel(R) 64 and IA-32 Architectures Software Developer's Manual Vol.2 +# (#325383-040US, October 2011) +# - Intel(R) Advanced Vector Extensions Programming Reference +# (#319433-011,JUNE 2011). +# # # Table: table-name # Referrer: escaped-name @@ -15,10 +21,13 @@ # EndTable # # AVX Superscripts -# (VEX): this opcode can accept VEX prefix. -# (oVEX): this opcode requires VEX prefix. -# (o128): this opcode only supports 128bit VEX. -# (o256): this opcode only supports 256bit VEX. +# (v): this opcode requires VEX prefix. +# (v1): this opcode only supports 128bit VEX. +# +# Last Prefix Superscripts +# - (66): the last prefix is 0x66 +# - (F3): the last prefix is 0xF3 +# - (F2): the last prefix is 0xF2 # Table: one byte opcode @@ -199,8 +208,8 @@ a0: MOV AL,Ob a1: MOV rAX,Ov a2: MOV Ob,AL a3: MOV Ov,rAX -a4: MOVS/B Xb,Yb -a5: MOVS/W/D/Q Xv,Yv +a4: MOVS/B Yb,Xb +a5: MOVS/W/D/Q Yv,Xv a6: CMPS/B Xb,Yb a7: CMPS/W/D Xv,Yv a8: TEST AL,Ib @@ -233,8 +242,8 @@ c0: Grp2 Eb,Ib (1A) c1: Grp2 Ev,Ib (1A) c2: RETN Iw (f64) c3: RETN -c4: LES Gz,Mp (i64) | 3bytes-VEX (Prefix) -c5: LDS Gz,Mp (i64) | 2bytes-VEX (Prefix) +c4: LES Gz,Mp (i64) | VEX+2byte (Prefix) +c5: LDS Gz,Mp (i64) | VEX+1byte (Prefix) c6: Grp11 Eb,Ib (1A) c7: Grp11 Ev,Iz (1A) c8: ENTER Iw,Ib @@ -320,14 +329,19 @@ AVXcode: 1 # 3DNow! uses the last imm byte as opcode extension. 0f: 3DNow! Pq,Qq,Ib # 0x0f 0x10-0x1f -10: movups Vps,Wps (VEX) | movss Vss,Wss (F3),(VEX),(o128) | movupd Vpd,Wpd (66),(VEX) | movsd Vsd,Wsd (F2),(VEX),(o128) -11: movups Wps,Vps (VEX) | movss Wss,Vss (F3),(VEX),(o128) | movupd Wpd,Vpd (66),(VEX) | movsd Wsd,Vsd (F2),(VEX),(o128) -12: movlps Vq,Mq (VEX),(o128) | movlpd Vq,Mq (66),(VEX),(o128) | movhlps Vq,Uq (VEX),(o128) | movddup Vq,Wq (F2),(VEX) | movsldup Vq,Wq (F3),(VEX) -13: mpvlps Mq,Vq (VEX),(o128) | movlpd Mq,Vq (66),(VEX),(o128) -14: unpcklps Vps,Wq (VEX) | unpcklpd Vpd,Wq (66),(VEX) -15: unpckhps Vps,Wq (VEX) | unpckhpd Vpd,Wq (66),(VEX) -16: movhps Vq,Mq (VEX),(o128) | movhpd Vq,Mq (66),(VEX),(o128) | movlsps Vq,Uq (VEX),(o128) | movshdup Vq,Wq (F3),(VEX) -17: movhps Mq,Vq (VEX),(o128) | movhpd Mq,Vq (66),(VEX),(o128) +# NOTE: According to Intel SDM opcode map, vmovups and vmovupd has no operands +# but it actually has operands. And also, vmovss and vmovsd only accept 128bit. +# MOVSS/MOVSD has too many forms(3) on SDM. This map just shows a typical form. +# Many AVX instructions lack v1 superscript, according to Intel AVX-Prgramming +# Reference A.1 +10: vmovups Vps,Wps | vmovupd Vpd,Wpd (66) | vmovss Vx,Hx,Wss (F3),(v1) | vmovsd Vx,Hx,Wsd (F2),(v1) +11: vmovups Wps,Vps | vmovupd Wpd,Vpd (66) | vmovss Wss,Hx,Vss (F3),(v1) | vmovsd Wsd,Hx,Vsd (F2),(v1) +12: vmovlps Vq,Hq,Mq (v1) | vmovhlps Vq,Hq,Uq (v1) | vmovlpd Vq,Hq,Mq (66),(v1) | vmovsldup Vx,Wx (F3) | vmovddup Vx,Wx (F2) +13: vmovlps Mq,Vq (v1) | vmovlpd Mq,Vq (66),(v1) +14: vunpcklps Vx,Hx,Wx | vunpcklpd Vx,Hx,Wx (66) +15: vunpckhps Vx,Hx,Wx | vunpckhpd Vx,Hx,Wx (66) +16: vmovhps Vdq,Hq,Mq (v1) | vmovlhps Vdq,Hq,Uq (v1) | vmovhpd Vdq,Hq,Mq (66),(v1) | vmovshdup Vx,Wx (F3) +17: vmovhps Mq,Vq (v1) | vmovhpd Mq,Vq (66),(v1) 18: Grp16 (1A) 19: 1a: @@ -345,14 +359,14 @@ AVXcode: 1 25: 26: 27: -28: movaps Vps,Wps (VEX) | movapd Vpd,Wpd (66),(VEX) -29: movaps Wps,Vps (VEX) | movapd Wpd,Vpd (66),(VEX) -2a: cvtpi2ps Vps,Qpi | cvtsi2ss Vss,Ed/q (F3),(VEX),(o128) | cvtpi2pd Vpd,Qpi (66) | cvtsi2sd Vsd,Ed/q (F2),(VEX),(o128) -2b: movntps Mps,Vps (VEX) | movntpd Mpd,Vpd (66),(VEX) -2c: cvttps2pi Ppi,Wps | cvttss2si Gd/q,Wss (F3),(VEX),(o128) | cvttpd2pi Ppi,Wpd (66) | cvttsd2si Gd/q,Wsd (F2),(VEX),(o128) -2d: cvtps2pi Ppi,Wps | cvtss2si Gd/q,Wss (F3),(VEX),(o128) | cvtpd2pi Qpi,Wpd (66) | cvtsd2si Gd/q,Wsd (F2),(VEX),(o128) -2e: ucomiss Vss,Wss (VEX),(o128) | ucomisd Vsd,Wsd (66),(VEX),(o128) -2f: comiss Vss,Wss (VEX),(o128) | comisd Vsd,Wsd (66),(VEX),(o128) +28: vmovaps Vps,Wps | vmovapd Vpd,Wpd (66) +29: vmovaps Wps,Vps | vmovapd Wpd,Vpd (66) +2a: cvtpi2ps Vps,Qpi | cvtpi2pd Vpd,Qpi (66) | vcvtsi2ss Vss,Hss,Ey (F3),(v1) | vcvtsi2sd Vsd,Hsd,Ey (F2),(v1) +2b: vmovntps Mps,Vps | vmovntpd Mpd,Vpd (66) +2c: cvttps2pi Ppi,Wps | cvttpd2pi Ppi,Wpd (66) | vcvttss2si Gy,Wss (F3),(v1) | vcvttsd2si Gy,Wsd (F2),(v1) +2d: cvtps2pi Ppi,Wps | cvtpd2pi Qpi,Wpd (66) | vcvtss2si Gy,Wss (F3),(v1) | vcvtsd2si Gy,Wsd (F2),(v1) +2e: vucomiss Vss,Wss (v1) | vucomisd Vsd,Wsd (66),(v1) +2f: vcomiss Vss,Wss (v1) | vcomisd Vsd,Wsd (66),(v1) # 0x0f 0x30-0x3f 30: WRMSR 31: RDTSC @@ -388,65 +402,66 @@ AVXcode: 1 4e: CMOVLE/NG Gv,Ev 4f: CMOVNLE/G Gv,Ev # 0x0f 0x50-0x5f -50: movmskps Gd/q,Ups (VEX) | movmskpd Gd/q,Upd (66),(VEX) -51: sqrtps Vps,Wps (VEX) | sqrtss Vss,Wss (F3),(VEX),(o128) | sqrtpd Vpd,Wpd (66),(VEX) | sqrtsd Vsd,Wsd (F2),(VEX),(o128) -52: rsqrtps Vps,Wps (VEX) | rsqrtss Vss,Wss (F3),(VEX),(o128) -53: rcpps Vps,Wps (VEX) | rcpss Vss,Wss (F3),(VEX),(o128) -54: andps Vps,Wps (VEX) | andpd Vpd,Wpd (66),(VEX) -55: andnps Vps,Wps (VEX) | andnpd Vpd,Wpd (66),(VEX) -56: orps Vps,Wps (VEX) | orpd Vpd,Wpd (66),(VEX) -57: xorps Vps,Wps (VEX) | xorpd Vpd,Wpd (66),(VEX) -58: addps Vps,Wps (VEX) | addss Vss,Wss (F3),(VEX),(o128) | addpd Vpd,Wpd (66),(VEX) | addsd Vsd,Wsd (F2),(VEX),(o128) -59: mulps Vps,Wps (VEX) | mulss Vss,Wss (F3),(VEX),(o128) | mulpd Vpd,Wpd (66),(VEX) | mulsd Vsd,Wsd (F2),(VEX),(o128) -5a: cvtps2pd Vpd,Wps (VEX) | cvtss2sd Vsd,Wss (F3),(VEX),(o128) | cvtpd2ps Vps,Wpd (66),(VEX) | cvtsd2ss Vsd,Wsd (F2),(VEX),(o128) -5b: cvtdq2ps Vps,Wdq (VEX) | cvtps2dq Vdq,Wps (66),(VEX) | cvttps2dq Vdq,Wps (F3),(VEX) -5c: subps Vps,Wps (VEX) | subss Vss,Wss (F3),(VEX),(o128) | subpd Vpd,Wpd (66),(VEX) | subsd Vsd,Wsd (F2),(VEX),(o128) -5d: minps Vps,Wps (VEX) | minss Vss,Wss (F3),(VEX),(o128) | minpd Vpd,Wpd (66),(VEX) | minsd Vsd,Wsd (F2),(VEX),(o128) -5e: divps Vps,Wps (VEX) | divss Vss,Wss (F3),(VEX),(o128) | divpd Vpd,Wpd (66),(VEX) | divsd Vsd,Wsd (F2),(VEX),(o128) -5f: maxps Vps,Wps (VEX) | maxss Vss,Wss (F3),(VEX),(o128) | maxpd Vpd,Wpd (66),(VEX) | maxsd Vsd,Wsd (F2),(VEX),(o128) +50: vmovmskps Gy,Ups | vmovmskpd Gy,Upd (66) +51: vsqrtps Vps,Wps | vsqrtpd Vpd,Wpd (66) | vsqrtss Vss,Hss,Wss (F3),(v1) | vsqrtsd Vsd,Hsd,Wsd (F2),(v1) +52: vrsqrtps Vps,Wps | vrsqrtss Vss,Hss,Wss (F3),(v1) +53: vrcpps Vps,Wps | vrcpss Vss,Hss,Wss (F3),(v1) +54: vandps Vps,Hps,Wps | vandpd Vpd,Hpd,Wpd (66) +55: vandnps Vps,Hps,Wps | vandnpd Vpd,Hpd,Wpd (66) +56: vorps Vps,Hps,Wps | vorpd Vpd,Hpd,Wpd (66) +57: vxorps Vps,Hps,Wps | vxorpd Vpd,Hpd,Wpd (66) +58: vaddps Vps,Hps,Wps | vaddpd Vpd,Hpd,Wpd (66) | vaddss Vss,Hss,Wss (F3),(v1) | vaddsd Vsd,Hsd,Wsd (F2),(v1) +59: vmulps Vps,Hps,Wps | vmulpd Vpd,Hpd,Wpd (66) | vmulss Vss,Hss,Wss (F3),(v1) | vmulsd Vsd,Hsd,Wsd (F2),(v1) +5a: vcvtps2pd Vpd,Wps | vcvtpd2ps Vps,Wpd (66) | vcvtss2sd Vsd,Hx,Wss (F3),(v1) | vcvtsd2ss Vss,Hx,Wsd (F2),(v1) +5b: vcvtdq2ps Vps,Wdq | vcvtps2dq Vdq,Wps (66) | vcvttps2dq Vdq,Wps (F3) +5c: vsubps Vps,Hps,Wps | vsubpd Vpd,Hpd,Wpd (66) | vsubss Vss,Hss,Wss (F3),(v1) | vsubsd Vsd,Hsd,Wsd (F2),(v1) +5d: vminps Vps,Hps,Wps | vminpd Vpd,Hpd,Wpd (66) | vminss Vss,Hss,Wss (F3),(v1) | vminsd Vsd,Hsd,Wsd (F2),(v1) +5e: vdivps Vps,Hps,Wps | vdivpd Vpd,Hpd,Wpd (66) | vdivss Vss,Hss,Wss (F3),(v1) | vdivsd Vsd,Hsd,Wsd (F2),(v1) +5f: vmaxps Vps,Hps,Wps | vmaxpd Vpd,Hpd,Wpd (66) | vmaxss Vss,Hss,Wss (F3),(v1) | vmaxsd Vsd,Hsd,Wsd (F2),(v1) # 0x0f 0x60-0x6f -60: punpcklbw Pq,Qd | punpcklbw Vdq,Wdq (66),(VEX),(o128) -61: punpcklwd Pq,Qd | punpcklwd Vdq,Wdq (66),(VEX),(o128) -62: punpckldq Pq,Qd | punpckldq Vdq,Wdq (66),(VEX),(o128) -63: packsswb Pq,Qq | packsswb Vdq,Wdq (66),(VEX),(o128) -64: pcmpgtb Pq,Qq | pcmpgtb Vdq,Wdq (66),(VEX),(o128) -65: pcmpgtw Pq,Qq | pcmpgtw Vdq,Wdq (66),(VEX),(o128) -66: pcmpgtd Pq,Qq | pcmpgtd Vdq,Wdq (66),(VEX),(o128) -67: packuswb Pq,Qq | packuswb Vdq,Wdq (66),(VEX),(o128) -68: punpckhbw Pq,Qd | punpckhbw Vdq,Wdq (66),(VEX),(o128) -69: punpckhwd Pq,Qd | punpckhwd Vdq,Wdq (66),(VEX),(o128) -6a: punpckhdq Pq,Qd | punpckhdq Vdq,Wdq (66),(VEX),(o128) -6b: packssdw Pq,Qd | packssdw Vdq,Wdq (66),(VEX),(o128) -6c: punpcklqdq Vdq,Wdq (66),(VEX),(o128) -6d: punpckhqdq Vdq,Wdq (66),(VEX),(o128) -6e: movd/q/ Pd,Ed/q | movd/q Vdq,Ed/q (66),(VEX),(o128) -6f: movq Pq,Qq | movdqa Vdq,Wdq (66),(VEX) | movdqu Vdq,Wdq (F3),(VEX) +60: punpcklbw Pq,Qd | vpunpcklbw Vx,Hx,Wx (66),(v1) +61: punpcklwd Pq,Qd | vpunpcklwd Vx,Hx,Wx (66),(v1) +62: punpckldq Pq,Qd | vpunpckldq Vx,Hx,Wx (66),(v1) +63: packsswb Pq,Qq | vpacksswb Vx,Hx,Wx (66),(v1) +64: pcmpgtb Pq,Qq | vpcmpgtb Vx,Hx,Wx (66),(v1) +65: pcmpgtw Pq,Qq | vpcmpgtw Vx,Hx,Wx (66),(v1) +66: pcmpgtd Pq,Qq | vpcmpgtd Vx,Hx,Wx (66),(v1) +67: packuswb Pq,Qq | vpackuswb Vx,Hx,Wx (66),(v1) +68: punpckhbw Pq,Qd | vpunpckhbw Vx,Hx,Wx (66),(v1) +69: punpckhwd Pq,Qd | vpunpckhwd Vx,Hx,Wx (66),(v1) +6a: punpckhdq Pq,Qd | vpunpckhdq Vx,Hx,Wx (66),(v1) +6b: packssdw Pq,Qd | vpackssdw Vx,Hx,Wx (66),(v1) +6c: vpunpcklqdq Vx,Hx,Wx (66),(v1) +6d: vpunpckhqdq Vx,Hx,Wx (66),(v1) +6e: movd/q Pd,Ey | vmovd/q Vy,Ey (66),(v1) +6f: movq Pq,Qq | vmovdqa Vx,Wx (66) | vmovdqu Vx,Wx (F3) # 0x0f 0x70-0x7f -70: pshufw Pq,Qq,Ib | pshufd Vdq,Wdq,Ib (66),(VEX),(o128) | pshufhw Vdq,Wdq,Ib (F3),(VEX),(o128) | pshuflw VdqWdq,Ib (F2),(VEX),(o128) +70: pshufw Pq,Qq,Ib | vpshufd Vx,Wx,Ib (66),(v1) | vpshufhw Vx,Wx,Ib (F3),(v1) | vpshuflw Vx,Wx,Ib (F2),(v1) 71: Grp12 (1A) 72: Grp13 (1A) 73: Grp14 (1A) -74: pcmpeqb Pq,Qq | pcmpeqb Vdq,Wdq (66),(VEX),(o128) -75: pcmpeqw Pq,Qq | pcmpeqw Vdq,Wdq (66),(VEX),(o128) -76: pcmpeqd Pq,Qq | pcmpeqd Vdq,Wdq (66),(VEX),(o128) -77: emms/vzeroupper/vzeroall (VEX) -78: VMREAD Ed/q,Gd/q -79: VMWRITE Gd/q,Ed/q +74: pcmpeqb Pq,Qq | vpcmpeqb Vx,Hx,Wx (66),(v1) +75: pcmpeqw Pq,Qq | vpcmpeqw Vx,Hx,Wx (66),(v1) +76: pcmpeqd Pq,Qq | vpcmpeqd Vx,Hx,Wx (66),(v1) +# Note: Remove (v), because vzeroall and vzeroupper becomes emms without VEX. +77: emms | vzeroupper | vzeroall +78: VMREAD Ey,Gy +79: VMWRITE Gy,Ey 7a: 7b: -7c: haddps Vps,Wps (F2),(VEX) | haddpd Vpd,Wpd (66),(VEX) -7d: hsubps Vps,Wps (F2),(VEX) | hsubpd Vpd,Wpd (66),(VEX) -7e: movd/q Ed/q,Pd | movd/q Ed/q,Vdq (66),(VEX),(o128) | movq Vq,Wq (F3),(VEX),(o128) -7f: movq Qq,Pq | movdqa Wdq,Vdq (66),(VEX) | movdqu Wdq,Vdq (F3),(VEX) +7c: vhaddpd Vpd,Hpd,Wpd (66) | vhaddps Vps,Hps,Wps (F2) +7d: vhsubpd Vpd,Hpd,Wpd (66) | vhsubps Vps,Hps,Wps (F2) +7e: movd/q Ey,Pd | vmovd/q Ey,Vy (66),(v1) | vmovq Vq,Wq (F3),(v1) +7f: movq Qq,Pq | vmovdqa Wx,Vx (66) | vmovdqu Wx,Vx (F3) # 0x0f 0x80-0x8f 80: JO Jz (f64) 81: JNO Jz (f64) -82: JB/JNAE/JC Jz (f64) -83: JNB/JAE/JNC Jz (f64) -84: JZ/JE Jz (f64) -85: JNZ/JNE Jz (f64) +82: JB/JC/JNAE Jz (f64) +83: JAE/JNB/JNC Jz (f64) +84: JE/JZ Jz (f64) +85: JNE/JNZ Jz (f64) 86: JBE/JNA Jz (f64) -87: JNBE/JA Jz (f64) +87: JA/JNBE Jz (f64) 88: JS Jz (f64) 89: JNS Jz (f64) 8a: JP/JPE Jz (f64) @@ -502,18 +517,18 @@ b8: JMPE | POPCNT Gv,Ev (F3) b9: Grp10 (1A) ba: Grp8 Ev,Ib (1A) bb: BTC Ev,Gv -bc: BSF Gv,Ev -bd: BSR Gv,Ev +bc: BSF Gv,Ev | TZCNT Gv,Ev (F3) +bd: BSR Gv,Ev | LZCNT Gv,Ev (F3) be: MOVSX Gv,Eb bf: MOVSX Gv,Ew # 0x0f 0xc0-0xcf c0: XADD Eb,Gb c1: XADD Ev,Gv -c2: cmpps Vps,Wps,Ib (VEX) | cmpss Vss,Wss,Ib (F3),(VEX),(o128) | cmppd Vpd,Wpd,Ib (66),(VEX) | cmpsd Vsd,Wsd,Ib (F2),(VEX) -c3: movnti Md/q,Gd/q -c4: pinsrw Pq,Rd/q/Mw,Ib | pinsrw Vdq,Rd/q/Mw,Ib (66),(VEX),(o128) -c5: pextrw Gd,Nq,Ib | pextrw Gd,Udq,Ib (66),(VEX),(o128) -c6: shufps Vps,Wps,Ib (VEX) | shufpd Vpd,Wpd,Ib (66),(VEX) +c2: vcmpps Vps,Hps,Wps,Ib | vcmppd Vpd,Hpd,Wpd,Ib (66) | vcmpss Vss,Hss,Wss,Ib (F3),(v1) | vcmpsd Vsd,Hsd,Wsd,Ib (F2),(v1) +c3: movnti My,Gy +c4: pinsrw Pq,Ry/Mw,Ib | vpinsrw Vdq,Hdq,Ry/Mw,Ib (66),(v1) +c5: pextrw Gd,Nq,Ib | vpextrw Gd,Udq,Ib (66),(v1) +c6: vshufps Vps,Hps,Wps,Ib | vshufpd Vpd,Hpd,Wpd,Ib (66) c7: Grp9 (1A) c8: BSWAP RAX/EAX/R8/R8D c9: BSWAP RCX/ECX/R9/R9D @@ -524,55 +539,55 @@ cd: BSWAP RBP/EBP/R13/R13D ce: BSWAP RSI/ESI/R14/R14D cf: BSWAP RDI/EDI/R15/R15D # 0x0f 0xd0-0xdf -d0: addsubps Vps,Wps (F2),(VEX) | addsubpd Vpd,Wpd (66),(VEX) -d1: psrlw Pq,Qq | psrlw Vdq,Wdq (66),(VEX),(o128) -d2: psrld Pq,Qq | psrld Vdq,Wdq (66),(VEX),(o128) -d3: psrlq Pq,Qq | psrlq Vdq,Wdq (66),(VEX),(o128) -d4: paddq Pq,Qq | paddq Vdq,Wdq (66),(VEX),(o128) -d5: pmullw Pq,Qq | pmullw Vdq,Wdq (66),(VEX),(o128) -d6: movq Wq,Vq (66),(VEX),(o128) | movq2dq Vdq,Nq (F3) | movdq2q Pq,Uq (F2) -d7: pmovmskb Gd,Nq | pmovmskb Gd,Udq (66),(VEX),(o128) -d8: psubusb Pq,Qq | psubusb Vdq,Wdq (66),(VEX),(o128) -d9: psubusw Pq,Qq | psubusw Vdq,Wdq (66),(VEX),(o128) -da: pminub Pq,Qq | pminub Vdq,Wdq (66),(VEX),(o128) -db: pand Pq,Qq | pand Vdq,Wdq (66),(VEX),(o128) -dc: paddusb Pq,Qq | paddusb Vdq,Wdq (66),(VEX),(o128) -dd: paddusw Pq,Qq | paddusw Vdq,Wdq (66),(VEX),(o128) -de: pmaxub Pq,Qq | pmaxub Vdq,Wdq (66),(VEX),(o128) -df: pandn Pq,Qq | pandn Vdq,Wdq (66),(VEX),(o128) +d0: vaddsubpd Vpd,Hpd,Wpd (66) | vaddsubps Vps,Hps,Wps (F2) +d1: psrlw Pq,Qq | vpsrlw Vx,Hx,Wx (66),(v1) +d2: psrld Pq,Qq | vpsrld Vx,Hx,Wx (66),(v1) +d3: psrlq Pq,Qq | vpsrlq Vx,Hx,Wx (66),(v1) +d4: paddq Pq,Qq | vpaddq Vx,Hx,Wx (66),(v1) +d5: pmullw Pq,Qq | vpmullw Vx,Hx,Wx (66),(v1) +d6: vmovq Wq,Vq (66),(v1) | movq2dq Vdq,Nq (F3) | movdq2q Pq,Uq (F2) +d7: pmovmskb Gd,Nq | vpmovmskb Gd,Ux (66),(v1) +d8: psubusb Pq,Qq | vpsubusb Vx,Hx,Wx (66),(v1) +d9: psubusw Pq,Qq | vpsubusw Vx,Hx,Wx (66),(v1) +da: pminub Pq,Qq | vpminub Vx,Hx,Wx (66),(v1) +db: pand Pq,Qq | vpand Vx,Hx,Wx (66),(v1) +dc: paddusb Pq,Qq | vpaddusb Vx,Hx,Wx (66),(v1) +dd: paddusw Pq,Qq | vpaddusw Vx,Hx,Wx (66),(v1) +de: pmaxub Pq,Qq | vpmaxub Vx,Hx,Wx (66),(v1) +df: pandn Pq,Qq | vpandn Vx,Hx,Wx (66),(v1) # 0x0f 0xe0-0xef -e0: pavgb Pq,Qq | pavgb Vdq,Wdq (66),(VEX),(o128) -e1: psraw Pq,Qq | psraw Vdq,Wdq (66),(VEX),(o128) -e2: psrad Pq,Qq | psrad Vdq,Wdq (66),(VEX),(o128) -e3: pavgw Pq,Qq | pavgw Vdq,Wdq (66),(VEX),(o128) -e4: pmulhuw Pq,Qq | pmulhuw Vdq,Wdq (66),(VEX),(o128) -e5: pmulhw Pq,Qq | pmulhw Vdq,Wdq (66),(VEX),(o128) -e6: cvtpd2dq Vdq,Wpd (F2),(VEX) | cvttpd2dq Vdq,Wpd (66),(VEX) | cvtdq2pd Vpd,Wdq (F3),(VEX) -e7: movntq Mq,Pq | movntdq Mdq,Vdq (66),(VEX) -e8: psubsb Pq,Qq | psubsb Vdq,Wdq (66),(VEX),(o128) -e9: psubsw Pq,Qq | psubsw Vdq,Wdq (66),(VEX),(o128) -ea: pminsw Pq,Qq | pminsw Vdq,Wdq (66),(VEX),(o128) -eb: por Pq,Qq | por Vdq,Wdq (66),(VEX),(o128) -ec: paddsb Pq,Qq | paddsb Vdq,Wdq (66),(VEX),(o128) -ed: paddsw Pq,Qq | paddsw Vdq,Wdq (66),(VEX),(o128) -ee: pmaxsw Pq,Qq | pmaxsw Vdq,Wdq (66),(VEX),(o128) -ef: pxor Pq,Qq | pxor Vdq,Wdq (66),(VEX),(o128) +e0: pavgb Pq,Qq | vpavgb Vx,Hx,Wx (66),(v1) +e1: psraw Pq,Qq | vpsraw Vx,Hx,Wx (66),(v1) +e2: psrad Pq,Qq | vpsrad Vx,Hx,Wx (66),(v1) +e3: pavgw Pq,Qq | vpavgw Vx,Hx,Wx (66),(v1) +e4: pmulhuw Pq,Qq | vpmulhuw Vx,Hx,Wx (66),(v1) +e5: pmulhw Pq,Qq | vpmulhw Vx,Hx,Wx (66),(v1) +e6: vcvttpd2dq Vx,Wpd (66) | vcvtdq2pd Vx,Wdq (F3) | vcvtpd2dq Vx,Wpd (F2) +e7: movntq Mq,Pq | vmovntdq Mx,Vx (66) +e8: psubsb Pq,Qq | vpsubsb Vx,Hx,Wx (66),(v1) +e9: psubsw Pq,Qq | vpsubsw Vx,Hx,Wx (66),(v1) +ea: pminsw Pq,Qq | vpminsw Vx,Hx,Wx (66),(v1) +eb: por Pq,Qq | vpor Vx,Hx,Wx (66),(v1) +ec: paddsb Pq,Qq | vpaddsb Vx,Hx,Wx (66),(v1) +ed: paddsw Pq,Qq | vpaddsw Vx,Hx,Wx (66),(v1) +ee: pmaxsw Pq,Qq | vpmaxsw Vx,Hx,Wx (66),(v1) +ef: pxor Pq,Qq | vpxor Vx,Hx,Wx (66),(v1) # 0x0f 0xf0-0xff -f0: lddqu Vdq,Mdq (F2),(VEX) -f1: psllw Pq,Qq | psllw Vdq,Wdq (66),(VEX),(o128) -f2: pslld Pq,Qq | pslld Vdq,Wdq (66),(VEX),(o128) -f3: psllq Pq,Qq | psllq Vdq,Wdq (66),(VEX),(o128) -f4: pmuludq Pq,Qq | pmuludq Vdq,Wdq (66),(VEX),(o128) -f5: pmaddwd Pq,Qq | pmaddwd Vdq,Wdq (66),(VEX),(o128) -f6: psadbw Pq,Qq | psadbw Vdq,Wdq (66),(VEX),(o128) -f7: maskmovq Pq,Nq | maskmovdqu Vdq,Udq (66),(VEX),(o128) -f8: psubb Pq,Qq | psubb Vdq,Wdq (66),(VEX),(o128) -f9: psubw Pq,Qq | psubw Vdq,Wdq (66),(VEX),(o128) -fa: psubd Pq,Qq | psubd Vdq,Wdq (66),(VEX),(o128) -fb: psubq Pq,Qq | psubq Vdq,Wdq (66),(VEX),(o128) -fc: paddb Pq,Qq | paddb Vdq,Wdq (66),(VEX),(o128) -fd: paddw Pq,Qq | paddw Vdq,Wdq (66),(VEX),(o128) -fe: paddd Pq,Qq | paddd Vdq,Wdq (66),(VEX),(o128) +f0: vlddqu Vx,Mx (F2) +f1: psllw Pq,Qq | vpsllw Vx,Hx,Wx (66),(v1) +f2: pslld Pq,Qq | vpslld Vx,Hx,Wx (66),(v1) +f3: psllq Pq,Qq | vpsllq Vx,Hx,Wx (66),(v1) +f4: pmuludq Pq,Qq | vpmuludq Vx,Hx,Wx (66),(v1) +f5: pmaddwd Pq,Qq | vpmaddwd Vx,Hx,Wx (66),(v1) +f6: psadbw Pq,Qq | vpsadbw Vx,Hx,Wx (66),(v1) +f7: maskmovq Pq,Nq | vmaskmovdqu Vx,Ux (66),(v1) +f8: psubb Pq,Qq | vpsubb Vx,Hx,Wx (66),(v1) +f9: psubw Pq,Qq | vpsubw Vx,Hx,Wx (66),(v1) +fa: psubd Pq,Qq | vpsubd Vx,Hx,Wx (66),(v1) +fb: psubq Pq,Qq | vpsubq Vx,Hx,Wx (66),(v1) +fc: paddb Pq,Qq | vpaddb Vx,Hx,Wx (66),(v1) +fd: paddw Pq,Qq | vpaddw Vx,Hx,Wx (66),(v1) +fe: paddd Pq,Qq | vpaddd Vx,Hx,Wx (66),(v1) ff: EndTable @@ -580,155 +595,193 @@ Table: 3-byte opcode 1 (0x0f 0x38) Referrer: 3-byte escape 1 AVXcode: 2 # 0x0f 0x38 0x00-0x0f -00: pshufb Pq,Qq | pshufb Vdq,Wdq (66),(VEX),(o128) -01: phaddw Pq,Qq | phaddw Vdq,Wdq (66),(VEX),(o128) -02: phaddd Pq,Qq | phaddd Vdq,Wdq (66),(VEX),(o128) -03: phaddsw Pq,Qq | phaddsw Vdq,Wdq (66),(VEX),(o128) -04: pmaddubsw Pq,Qq | pmaddubsw Vdq,Wdq (66),(VEX),(o128) -05: phsubw Pq,Qq | phsubw Vdq,Wdq (66),(VEX),(o128) -06: phsubd Pq,Qq | phsubd Vdq,Wdq (66),(VEX),(o128) -07: phsubsw Pq,Qq | phsubsw Vdq,Wdq (66),(VEX),(o128) -08: psignb Pq,Qq | psignb Vdq,Wdq (66),(VEX),(o128) -09: psignw Pq,Qq | psignw Vdq,Wdq (66),(VEX),(o128) -0a: psignd Pq,Qq | psignd Vdq,Wdq (66),(VEX),(o128) -0b: pmulhrsw Pq,Qq | pmulhrsw Vdq,Wdq (66),(VEX),(o128) -0c: Vpermilps /r (66),(oVEX) -0d: Vpermilpd /r (66),(oVEX) -0e: vtestps /r (66),(oVEX) -0f: vtestpd /r (66),(oVEX) +00: pshufb Pq,Qq | vpshufb Vx,Hx,Wx (66),(v1) +01: phaddw Pq,Qq | vphaddw Vx,Hx,Wx (66),(v1) +02: phaddd Pq,Qq | vphaddd Vx,Hx,Wx (66),(v1) +03: phaddsw Pq,Qq | vphaddsw Vx,Hx,Wx (66),(v1) +04: pmaddubsw Pq,Qq | vpmaddubsw Vx,Hx,Wx (66),(v1) +05: phsubw Pq,Qq | vphsubw Vx,Hx,Wx (66),(v1) +06: phsubd Pq,Qq | vphsubd Vx,Hx,Wx (66),(v1) +07: phsubsw Pq,Qq | vphsubsw Vx,Hx,Wx (66),(v1) +08: psignb Pq,Qq | vpsignb Vx,Hx,Wx (66),(v1) +09: psignw Pq,Qq | vpsignw Vx,Hx,Wx (66),(v1) +0a: psignd Pq,Qq | vpsignd Vx,Hx,Wx (66),(v1) +0b: pmulhrsw Pq,Qq | vpmulhrsw Vx,Hx,Wx (66),(v1) +0c: vpermilps Vx,Hx,Wx (66),(v) +0d: vpermilpd Vx,Hx,Wx (66),(v) +0e: vtestps Vx,Wx (66),(v) +0f: vtestpd Vx,Wx (66),(v) # 0x0f 0x38 0x10-0x1f 10: pblendvb Vdq,Wdq (66) 11: 12: -13: +13: vcvtph2ps Vx,Wx,Ib (66),(v) 14: blendvps Vdq,Wdq (66) 15: blendvpd Vdq,Wdq (66) -16: -17: ptest Vdq,Wdq (66),(VEX) -18: vbroadcastss /r (66),(oVEX) -19: vbroadcastsd /r (66),(oVEX),(o256) -1a: vbroadcastf128 /r (66),(oVEX),(o256) +16: vpermps Vqq,Hqq,Wqq (66),(v) +17: vptest Vx,Wx (66) +18: vbroadcastss Vx,Wd (66),(v) +19: vbroadcastsd Vqq,Wq (66),(v) +1a: vbroadcastf128 Vqq,Mdq (66),(v) 1b: -1c: pabsb Pq,Qq | pabsb Vdq,Wdq (66),(VEX),(o128) -1d: pabsw Pq,Qq | pabsw Vdq,Wdq (66),(VEX),(o128) -1e: pabsd Pq,Qq | pabsd Vdq,Wdq (66),(VEX),(o128) +1c: pabsb Pq,Qq | vpabsb Vx,Wx (66),(v1) +1d: pabsw Pq,Qq | vpabsw Vx,Wx (66),(v1) +1e: pabsd Pq,Qq | vpabsd Vx,Wx (66),(v1) 1f: # 0x0f 0x38 0x20-0x2f -20: pmovsxbw Vdq,Udq/Mq (66),(VEX),(o128) -21: pmovsxbd Vdq,Udq/Md (66),(VEX),(o128) -22: pmovsxbq Vdq,Udq/Mw (66),(VEX),(o128) -23: pmovsxwd Vdq,Udq/Mq (66),(VEX),(o128) -24: pmovsxwq Vdq,Udq/Md (66),(VEX),(o128) -25: pmovsxdq Vdq,Udq/Mq (66),(VEX),(o128) +20: vpmovsxbw Vx,Ux/Mq (66),(v1) +21: vpmovsxbd Vx,Ux/Md (66),(v1) +22: vpmovsxbq Vx,Ux/Mw (66),(v1) +23: vpmovsxwd Vx,Ux/Mq (66),(v1) +24: vpmovsxwq Vx,Ux/Md (66),(v1) +25: vpmovsxdq Vx,Ux/Mq (66),(v1) 26: 27: -28: pmuldq Vdq,Wdq (66),(VEX),(o128) -29: pcmpeqq Vdq,Wdq (66),(VEX),(o128) -2a: movntdqa Vdq,Mdq (66),(VEX),(o128) -2b: packusdw Vdq,Wdq (66),(VEX),(o128) -2c: vmaskmovps(ld) /r (66),(oVEX) -2d: vmaskmovpd(ld) /r (66),(oVEX) -2e: vmaskmovps(st) /r (66),(oVEX) -2f: vmaskmovpd(st) /r (66),(oVEX) +28: vpmuldq Vx,Hx,Wx (66),(v1) +29: vpcmpeqq Vx,Hx,Wx (66),(v1) +2a: vmovntdqa Vx,Mx (66),(v1) +2b: vpackusdw Vx,Hx,Wx (66),(v1) +2c: vmaskmovps Vx,Hx,Mx (66),(v) +2d: vmaskmovpd Vx,Hx,Mx (66),(v) +2e: vmaskmovps Mx,Hx,Vx (66),(v) +2f: vmaskmovpd Mx,Hx,Vx (66),(v) # 0x0f 0x38 0x30-0x3f -30: pmovzxbw Vdq,Udq/Mq (66),(VEX),(o128) -31: pmovzxbd Vdq,Udq/Md (66),(VEX),(o128) -32: pmovzxbq Vdq,Udq/Mw (66),(VEX),(o128) -33: pmovzxwd Vdq,Udq/Mq (66),(VEX),(o128) -34: pmovzxwq Vdq,Udq/Md (66),(VEX),(o128) -35: pmovzxdq Vdq,Udq/Mq (66),(VEX),(o128) -36: -37: pcmpgtq Vdq,Wdq (66),(VEX),(o128) -38: pminsb Vdq,Wdq (66),(VEX),(o128) -39: pminsd Vdq,Wdq (66),(VEX),(o128) -3a: pminuw Vdq,Wdq (66),(VEX),(o128) -3b: pminud Vdq,Wdq (66),(VEX),(o128) -3c: pmaxsb Vdq,Wdq (66),(VEX),(o128) -3d: pmaxsd Vdq,Wdq (66),(VEX),(o128) -3e: pmaxuw Vdq,Wdq (66),(VEX),(o128) -3f: pmaxud Vdq,Wdq (66),(VEX),(o128) +30: vpmovzxbw Vx,Ux/Mq (66),(v1) +31: vpmovzxbd Vx,Ux/Md (66),(v1) +32: vpmovzxbq Vx,Ux/Mw (66),(v1) +33: vpmovzxwd Vx,Ux/Mq (66),(v1) +34: vpmovzxwq Vx,Ux/Md (66),(v1) +35: vpmovzxdq Vx,Ux/Mq (66),(v1) +36: vpermd Vqq,Hqq,Wqq (66),(v) +37: vpcmpgtq Vx,Hx,Wx (66),(v1) +38: vpminsb Vx,Hx,Wx (66),(v1) +39: vpminsd Vx,Hx,Wx (66),(v1) +3a: vpminuw Vx,Hx,Wx (66),(v1) +3b: vpminud Vx,Hx,Wx (66),(v1) +3c: vpmaxsb Vx,Hx,Wx (66),(v1) +3d: vpmaxsd Vx,Hx,Wx (66),(v1) +3e: vpmaxuw Vx,Hx,Wx (66),(v1) +3f: vpmaxud Vx,Hx,Wx (66),(v1) # 0x0f 0x38 0x40-0x8f -40: pmulld Vdq,Wdq (66),(VEX),(o128) -41: phminposuw Vdq,Wdq (66),(VEX),(o128) -80: INVEPT Gd/q,Mdq (66) -81: INVPID Gd/q,Mdq (66) +40: vpmulld Vx,Hx,Wx (66),(v1) +41: vphminposuw Vdq,Wdq (66),(v1) +42: +43: +44: +45: vpsrlvd/q Vx,Hx,Wx (66),(v) +46: vpsravd Vx,Hx,Wx (66),(v) +47: vpsllvd/q Vx,Hx,Wx (66),(v) +# Skip 0x48-0x57 +58: vpbroadcastd Vx,Wx (66),(v) +59: vpbroadcastq Vx,Wx (66),(v) +5a: vbroadcasti128 Vqq,Mdq (66),(v) +# Skip 0x5b-0x77 +78: vpbroadcastb Vx,Wx (66),(v) +79: vpbroadcastw Vx,Wx (66),(v) +# Skip 0x7a-0x7f +80: INVEPT Gy,Mdq (66) +81: INVPID Gy,Mdq (66) +82: INVPCID Gy,Mdq (66) +8c: vpmaskmovd/q Vx,Hx,Mx (66),(v) +8e: vpmaskmovd/q Mx,Vx,Hx (66),(v) # 0x0f 0x38 0x90-0xbf (FMA) -96: vfmaddsub132pd/ps /r (66),(VEX) -97: vfmsubadd132pd/ps /r (66),(VEX) -98: vfmadd132pd/ps /r (66),(VEX) -99: vfmadd132sd/ss /r (66),(VEX),(o128) -9a: vfmsub132pd/ps /r (66),(VEX) -9b: vfmsub132sd/ss /r (66),(VEX),(o128) -9c: vfnmadd132pd/ps /r (66),(VEX) -9d: vfnmadd132sd/ss /r (66),(VEX),(o128) -9e: vfnmsub132pd/ps /r (66),(VEX) -9f: vfnmsub132sd/ss /r (66),(VEX),(o128) -a6: vfmaddsub213pd/ps /r (66),(VEX) -a7: vfmsubadd213pd/ps /r (66),(VEX) -a8: vfmadd213pd/ps /r (66),(VEX) -a9: vfmadd213sd/ss /r (66),(VEX),(o128) -aa: vfmsub213pd/ps /r (66),(VEX) -ab: vfmsub213sd/ss /r (66),(VEX),(o128) -ac: vfnmadd213pd/ps /r (66),(VEX) -ad: vfnmadd213sd/ss /r (66),(VEX),(o128) -ae: vfnmsub213pd/ps /r (66),(VEX) -af: vfnmsub213sd/ss /r (66),(VEX),(o128) -b6: vfmaddsub231pd/ps /r (66),(VEX) -b7: vfmsubadd231pd/ps /r (66),(VEX) -b8: vfmadd231pd/ps /r (66),(VEX) -b9: vfmadd231sd/ss /r (66),(VEX),(o128) -ba: vfmsub231pd/ps /r (66),(VEX) -bb: vfmsub231sd/ss /r (66),(VEX),(o128) -bc: vfnmadd231pd/ps /r (66),(VEX) -bd: vfnmadd231sd/ss /r (66),(VEX),(o128) -be: vfnmsub231pd/ps /r (66),(VEX) -bf: vfnmsub231sd/ss /r (66),(VEX),(o128) +90: vgatherdd/q Vx,Hx,Wx (66),(v) +91: vgatherqd/q Vx,Hx,Wx (66),(v) +92: vgatherdps/d Vx,Hx,Wx (66),(v) +93: vgatherqps/d Vx,Hx,Wx (66),(v) +94: +95: +96: vfmaddsub132ps/d Vx,Hx,Wx (66),(v) +97: vfmsubadd132ps/d Vx,Hx,Wx (66),(v) +98: vfmadd132ps/d Vx,Hx,Wx (66),(v) +99: vfmadd132ss/d Vx,Hx,Wx (66),(v),(v1) +9a: vfmsub132ps/d Vx,Hx,Wx (66),(v) +9b: vfmsub132ss/d Vx,Hx,Wx (66),(v),(v1) +9c: vfnmadd132ps/d Vx,Hx,Wx (66),(v) +9d: vfnmadd132ss/d Vx,Hx,Wx (66),(v),(v1) +9e: vfnmsub132ps/d Vx,Hx,Wx (66),(v) +9f: vfnmsub132ss/d Vx,Hx,Wx (66),(v),(v1) +a6: vfmaddsub213ps/d Vx,Hx,Wx (66),(v) +a7: vfmsubadd213ps/d Vx,Hx,Wx (66),(v) +a8: vfmadd213ps/d Vx,Hx,Wx (66),(v) +a9: vfmadd213ss/d Vx,Hx,Wx (66),(v),(v1) +aa: vfmsub213ps/d Vx,Hx,Wx (66),(v) +ab: vfmsub213ss/d Vx,Hx,Wx (66),(v),(v1) +ac: vfnmadd213ps/d Vx,Hx,Wx (66),(v) +ad: vfnmadd213ss/d Vx,Hx,Wx (66),(v),(v1) +ae: vfnmsub213ps/d Vx,Hx,Wx (66),(v) +af: vfnmsub213ss/d Vx,Hx,Wx (66),(v),(v1) +b6: vfmaddsub231ps/d Vx,Hx,Wx (66),(v) +b7: vfmsubadd231ps/d Vx,Hx,Wx (66),(v) +b8: vfmadd231ps/d Vx,Hx,Wx (66),(v) +b9: vfmadd231ss/d Vx,Hx,Wx (66),(v),(v1) +ba: vfmsub231ps/d Vx,Hx,Wx (66),(v) +bb: vfmsub231ss/d Vx,Hx,Wx (66),(v),(v1) +bc: vfnmadd231ps/d Vx,Hx,Wx (66),(v) +bd: vfnmadd231ss/d Vx,Hx,Wx (66),(v),(v1) +be: vfnmsub231ps/d Vx,Hx,Wx (66),(v) +bf: vfnmsub231ss/d Vx,Hx,Wx (66),(v),(v1) # 0x0f 0x38 0xc0-0xff -db: aesimc Vdq,Wdq (66),(VEX),(o128) -dc: aesenc Vdq,Wdq (66),(VEX),(o128) -dd: aesenclast Vdq,Wdq (66),(VEX),(o128) -de: aesdec Vdq,Wdq (66),(VEX),(o128) -df: aesdeclast Vdq,Wdq (66),(VEX),(o128) -f0: MOVBE Gv,Mv | CRC32 Gd,Eb (F2) -f1: MOVBE Mv,Gv | CRC32 Gd,Ev (F2) +db: VAESIMC Vdq,Wdq (66),(v1) +dc: VAESENC Vdq,Hdq,Wdq (66),(v1) +dd: VAESENCLAST Vdq,Hdq,Wdq (66),(v1) +de: VAESDEC Vdq,Hdq,Wdq (66),(v1) +df: VAESDECLAST Vdq,Hdq,Wdq (66),(v1) +f0: MOVBE Gy,My | MOVBE Gw,Mw (66) | CRC32 Gd,Eb (F2) +f1: MOVBE My,Gy | MOVBE Mw,Gw (66) | CRC32 Gd,Ey (F2) +f3: ANDN Gy,By,Ey (v) +f4: Grp17 (1A) +f5: BZHI Gy,Ey,By (v) | PEXT Gy,By,Ey (F3),(v) | PDEP Gy,By,Ey (F2),(v) +f6: MULX By,Gy,rDX,Ey (F2),(v) +f7: BEXTR Gy,Ey,By (v) | SHLX Gy,Ey,By (66),(v) | SARX Gy,Ey,By (F3),(v) | SHRX Gy,Ey,By (F2),(v) EndTable Table: 3-byte opcode 2 (0x0f 0x3a) Referrer: 3-byte escape 2 AVXcode: 3 # 0x0f 0x3a 0x00-0xff -04: vpermilps /r,Ib (66),(oVEX) -05: vpermilpd /r,Ib (66),(oVEX) -06: vperm2f128 /r,Ib (66),(oVEX),(o256) -08: roundps Vdq,Wdq,Ib (66),(VEX) -09: roundpd Vdq,Wdq,Ib (66),(VEX) -0a: roundss Vss,Wss,Ib (66),(VEX),(o128) -0b: roundsd Vsd,Wsd,Ib (66),(VEX),(o128) -0c: blendps Vdq,Wdq,Ib (66),(VEX) -0d: blendpd Vdq,Wdq,Ib (66),(VEX) -0e: pblendw Vdq,Wdq,Ib (66),(VEX),(o128) -0f: palignr Pq,Qq,Ib | palignr Vdq,Wdq,Ib (66),(VEX),(o128) -14: pextrb Rd/Mb,Vdq,Ib (66),(VEX),(o128) -15: pextrw Rd/Mw,Vdq,Ib (66),(VEX),(o128) -16: pextrd/pextrq Ed/q,Vdq,Ib (66),(VEX),(o128) -17: extractps Ed,Vdq,Ib (66),(VEX),(o128) -18: vinsertf128 /r,Ib (66),(oVEX),(o256) -19: vextractf128 /r,Ib (66),(oVEX),(o256) -20: pinsrb Vdq,Rd/q/Mb,Ib (66),(VEX),(o128) -21: insertps Vdq,Udq/Md,Ib (66),(VEX),(o128) -22: pinsrd/pinsrq Vdq,Ed/q,Ib (66),(VEX),(o128) -40: dpps Vdq,Wdq,Ib (66),(VEX) -41: dppd Vdq,Wdq,Ib (66),(VEX),(o128) -42: mpsadbw Vdq,Wdq,Ib (66),(VEX),(o128) -44: pclmulq Vdq,Wdq,Ib (66),(VEX),(o128) -4a: vblendvps /r,Ib (66),(oVEX) -4b: vblendvpd /r,Ib (66),(oVEX) -4c: vpblendvb /r,Ib (66),(oVEX),(o128) -60: pcmpestrm Vdq,Wdq,Ib (66),(VEX),(o128) -61: pcmpestri Vdq,Wdq,Ib (66),(VEX),(o128) -62: pcmpistrm Vdq,Wdq,Ib (66),(VEX),(o128) -63: pcmpistri Vdq,Wdq,Ib (66),(VEX),(o128) -df: aeskeygenassist Vdq,Wdq,Ib (66),(VEX),(o128) +00: vpermq Vqq,Wqq,Ib (66),(v) +01: vpermpd Vqq,Wqq,Ib (66),(v) +02: vpblendd Vx,Hx,Wx,Ib (66),(v) +03: +04: vpermilps Vx,Wx,Ib (66),(v) +05: vpermilpd Vx,Wx,Ib (66),(v) +06: vperm2f128 Vqq,Hqq,Wqq,Ib (66),(v) +07: +08: vroundps Vx,Wx,Ib (66) +09: vroundpd Vx,Wx,Ib (66) +0a: vroundss Vss,Wss,Ib (66),(v1) +0b: vroundsd Vsd,Wsd,Ib (66),(v1) +0c: vblendps Vx,Hx,Wx,Ib (66) +0d: vblendpd Vx,Hx,Wx,Ib (66) +0e: vpblendw Vx,Hx,Wx,Ib (66),(v1) +0f: palignr Pq,Qq,Ib | vpalignr Vx,Hx,Wx,Ib (66),(v1) +14: vpextrb Rd/Mb,Vdq,Ib (66),(v1) +15: vpextrw Rd/Mw,Vdq,Ib (66),(v1) +16: vpextrd/q Ey,Vdq,Ib (66),(v1) +17: vextractps Ed,Vdq,Ib (66),(v1) +18: vinsertf128 Vqq,Hqq,Wqq,Ib (66),(v) +19: vextractf128 Wdq,Vqq,Ib (66),(v) +1d: vcvtps2ph Wx,Vx,Ib (66),(v) +20: vpinsrb Vdq,Hdq,Ry/Mb,Ib (66),(v1) +21: vinsertps Vdq,Hdq,Udq/Md,Ib (66),(v1) +22: vpinsrd/q Vdq,Hdq,Ey,Ib (66),(v1) +38: vinserti128 Vqq,Hqq,Wqq,Ib (66),(v) +39: vextracti128 Wdq,Vqq,Ib (66),(v) +40: vdpps Vx,Hx,Wx,Ib (66) +41: vdppd Vdq,Hdq,Wdq,Ib (66),(v1) +42: vmpsadbw Vx,Hx,Wx,Ib (66),(v1) +44: vpclmulqdq Vdq,Hdq,Wdq,Ib (66),(v1) +46: vperm2i128 Vqq,Hqq,Wqq,Ib (66),(v) +4a: vblendvps Vx,Hx,Wx,Lx (66),(v) +4b: vblendvpd Vx,Hx,Wx,Lx (66),(v) +4c: vpblendvb Vx,Hx,Wx,Lx (66),(v1) +60: vpcmpestrm Vdq,Wdq,Ib (66),(v1) +61: vpcmpestri Vdq,Wdq,Ib (66),(v1) +62: vpcmpistrm Vdq,Wdq,Ib (66),(v1) +63: vpcmpistri Vdq,Wdq,Ib (66),(v1) +df: VAESKEYGEN Vdq,Wdq,Ib (66),(v1) +f0: RORX Gy,Ey,Ib (F2),(v) EndTable GrpTable: Grp1 @@ -790,7 +843,7 @@ GrpTable: Grp5 2: CALLN Ev (f64) 3: CALLF Ep 4: JMPN Ev (f64) -5: JMPF Ep +5: JMPF Mp 6: PUSH Ev (d64) 7: EndTable @@ -807,7 +860,7 @@ EndTable GrpTable: Grp7 0: SGDT Ms | VMCALL (001),(11B) | VMLAUNCH (010),(11B) | VMRESUME (011),(11B) | VMXOFF (100),(11B) 1: SIDT Ms | MONITOR (000),(11B) | MWAIT (001) -2: LGDT Ms | XGETBV (000),(11B) | XSETBV (001),(11B) +2: LGDT Ms | XGETBV (000),(11B) | XSETBV (001),(11B) | VMFUNC (100),(11B) 3: LIDT Ms 4: SMSW Mw/Rv 5: @@ -824,44 +877,45 @@ EndTable GrpTable: Grp9 1: CMPXCHG8B/16B Mq/Mdq -6: VMPTRLD Mq | VMCLEAR Mq (66) | VMXON Mq (F3) -7: VMPTRST Mq +6: VMPTRLD Mq | VMCLEAR Mq (66) | VMXON Mq (F3) | RDRAND Rv (11B) +7: VMPTRST Mq | VMPTRST Mq (F3) EndTable GrpTable: Grp10 EndTable GrpTable: Grp11 +# Note: the operands are given by group opcode 0: MOV EndTable GrpTable: Grp12 -2: psrlw Nq,Ib (11B) | psrlw Udq,Ib (66),(11B),(VEX),(o128) -4: psraw Nq,Ib (11B) | psraw Udq,Ib (66),(11B),(VEX),(o128) -6: psllw Nq,Ib (11B) | psllw Udq,Ib (66),(11B),(VEX),(o128) +2: psrlw Nq,Ib (11B) | vpsrlw Hx,Ux,Ib (66),(11B),(v1) +4: psraw Nq,Ib (11B) | vpsraw Hx,Ux,Ib (66),(11B),(v1) +6: psllw Nq,Ib (11B) | vpsllw Hx,Ux,Ib (66),(11B),(v1) EndTable GrpTable: Grp13 -2: psrld Nq,Ib (11B) | psrld Udq,Ib (66),(11B),(VEX),(o128) -4: psrad Nq,Ib (11B) | psrad Udq,Ib (66),(11B),(VEX),(o128) -6: pslld Nq,Ib (11B) | pslld Udq,Ib (66),(11B),(VEX),(o128) +2: psrld Nq,Ib (11B) | vpsrld Hx,Ux,Ib (66),(11B),(v1) +4: psrad Nq,Ib (11B) | vpsrad Hx,Ux,Ib (66),(11B),(v1) +6: pslld Nq,Ib (11B) | vpslld Hx,Ux,Ib (66),(11B),(v1) EndTable GrpTable: Grp14 -2: psrlq Nq,Ib (11B) | psrlq Udq,Ib (66),(11B),(VEX),(o128) -3: psrldq Udq,Ib (66),(11B),(VEX),(o128) -6: psllq Nq,Ib (11B) | psllq Udq,Ib (66),(11B),(VEX),(o128) -7: pslldq Udq,Ib (66),(11B),(VEX),(o128) +2: psrlq Nq,Ib (11B) | vpsrlq Hx,Ux,Ib (66),(11B),(v1) +3: vpsrldq Hx,Ux,Ib (66),(11B),(v1) +6: psllq Nq,Ib (11B) | vpsllq Hx,Ux,Ib (66),(11B),(v1) +7: vpslldq Hx,Ux,Ib (66),(11B),(v1) EndTable GrpTable: Grp15 -0: fxsave -1: fxstor -2: ldmxcsr (VEX) -3: stmxcsr (VEX) +0: fxsave | RDFSBASE Ry (F3),(11B) +1: fxstor | RDGSBASE Ry (F3),(11B) +2: vldmxcsr Md (v1) | WRFSBASE Ry (F3),(11B) +3: vstmxcsr Md (v1) | WRGSBASE Ry (F3),(11B) 4: XSAVE 5: XRSTOR | lfence (11B) -6: mfence (11B) +6: XSAVEOPT | mfence (11B) 7: clflush | sfence (11B) EndTable @@ -872,6 +926,12 @@ GrpTable: Grp16 3: prefetch T2 EndTable +GrpTable: Grp17 +1: BLSR By,Ey (v) +2: BLSMSK By,Ey (v) +3: BLSI By,Ey (v) +EndTable + # AMD's Prefetch Group GrpTable: GrpP 0: PREFETCH diff --git a/arch/x86/tools/gen-insn-attr-x86.awk b/arch/x86/tools/gen-insn-attr-x86.awk index eaf11f52fc0b..5f6a5b6c3a15 100644 --- a/arch/x86/tools/gen-insn-attr-x86.awk +++ b/arch/x86/tools/gen-insn-attr-x86.awk @@ -47,7 +47,7 @@ BEGIN { sep_expr = "^\\|$" group_expr = "^Grp[0-9A-Za-z]+" - imm_expr = "^[IJAO][a-z]" + imm_expr = "^[IJAOL][a-z]" imm_flag["Ib"] = "INAT_MAKE_IMM(INAT_IMM_BYTE)" imm_flag["Jb"] = "INAT_MAKE_IMM(INAT_IMM_BYTE)" imm_flag["Iw"] = "INAT_MAKE_IMM(INAT_IMM_WORD)" @@ -59,6 +59,7 @@ BEGIN { imm_flag["Iv"] = "INAT_MAKE_IMM(INAT_IMM_VWORD)" imm_flag["Ob"] = "INAT_MOFFSET" imm_flag["Ov"] = "INAT_MOFFSET" + imm_flag["Lx"] = "INAT_MAKE_IMM(INAT_IMM_BYTE)" modrm_expr = "^([CDEGMNPQRSUVW/][a-z]+|NTA|T[012])" force64_expr = "\\([df]64\\)" @@ -70,8 +71,12 @@ BEGIN { lprefix3_expr = "\\(F2\\)" max_lprefix = 4 - vexok_expr = "\\(VEX\\)" - vexonly_expr = "\\(oVEX\\)" + # All opcodes starting with lower-case 'v' or with (v1) superscript + # accepts VEX prefix + vexok_opcode_expr = "^v.*" + vexok_expr = "\\(v1\\)" + # All opcodes with (v) superscript supports *only* VEX prefix + vexonly_expr = "\\(v\\)" prefix_expr = "\\(Prefix\\)" prefix_num["Operand-Size"] = "INAT_PFX_OPNDSZ" @@ -85,8 +90,8 @@ BEGIN { prefix_num["SEG=GS"] = "INAT_PFX_GS" prefix_num["SEG=SS"] = "INAT_PFX_SS" prefix_num["Address-Size"] = "INAT_PFX_ADDRSZ" - prefix_num["2bytes-VEX"] = "INAT_PFX_VEX2" - prefix_num["3bytes-VEX"] = "INAT_PFX_VEX3" + prefix_num["VEX+1byte"] = "INAT_PFX_VEX2" + prefix_num["VEX+2byte"] = "INAT_PFX_VEX3" clear_vars() } @@ -310,12 +315,10 @@ function convert_operands(count,opnd, i,j,imm,mod) if (match(opcode, fpu_expr)) flags = add_flags(flags, "INAT_MODRM") - # check VEX only code + # check VEX codes if (match(ext, vexonly_expr)) flags = add_flags(flags, "INAT_VEXOK | INAT_VEXONLY") - - # check VEX only code - if (match(ext, vexok_expr)) + else if (match(ext, vexok_expr) || match(opcode, vexok_opcode_expr)) flags = add_flags(flags, "INAT_VEXOK") # check prefixes -- cgit v1.2.3 From 9dde9dc0a81c7aeb863b35121d09011f09b4897c Mon Sep 17 00:00:00 2001 From: Masami Hiramatsu Date: Mon, 5 Dec 2011 21:06:03 +0900 Subject: x86/tools: Add decoded instruction dump mode Add instruction dump mode to insn_sanity tool for checking decoder really decoded instructions. This mode is enabled when passing double -v (-vv) to insn_sanity. It is useful for who wants to check whether the decoder can decode some instructions correctly. e.g. $ echo 0f 73 10 11 | ./insn_sanity -y -vv -i - Instruction = { .prefixes = { .value = 0, bytes[] = {0, 0, 0, 0}, .got = 1, .nbytes = 0}, .rex_prefix = { .value = 0, bytes[] = {0, 0, 0, 0}, .got = 1, .nbytes = 0}, .vex_prefix = { .value = 0, bytes[] = {0, 0, 0, 0}, .got = 1, .nbytes = 0}, .opcode = { .value = 29455, bytes[] = {f, 73, 0, 0}, .got = 1, .nbytes = 2}, .modrm = { .value = 16, bytes[] = {10, 0, 0, 0}, .got = 1, .nbytes = 1}, .sib = { .value = 0, bytes[] = {0, 0, 0, 0}, .got = 1, .nbytes = 0}, .displacement = { .value = 0, bytes[] = {0, 0, 0, 0}, .got = 1, .nbytes = 0}, .immediate1 = { .value = 17, bytes[] = {11, 0, 0, 0}, .got = 1, .nbytes = 1}, .immediate2 = { .value = 0, bytes[] = {0, 0, 0, 0}, .got = 0, .nbytes = 0}, .attr = 44800, .opnd_bytes = 4, .addr_bytes = 8, .length = 4, .x86_64 = 1, .kaddr = 0x7fff0f7d9430} Success: decoded and checked 1 given instructions with 0 errors (seed:0x0) Signed-off-by: Masami Hiramatsu Cc: "H. Peter Anvin" Cc: yrl.pp-manager.tt@hitachi.com Link: http://lkml.kernel.org/r/20111205120603.15475.91192.stgit@cloud Signed-off-by: Ingo Molnar --- arch/x86/tools/insn_sanity.c | 7 ++++--- 1 file changed, 4 insertions(+), 3 deletions(-) diff --git a/arch/x86/tools/insn_sanity.c b/arch/x86/tools/insn_sanity.c index b6720d6b38cb..cc2f8c131286 100644 --- a/arch/x86/tools/insn_sanity.c +++ b/arch/x86/tools/insn_sanity.c @@ -59,7 +59,7 @@ static void usage(const char *err) fprintf(stderr, "Usage: %s [-y|-n|-v] [-s seed[,no]] [-m max] [-i input]\n", prog); fprintf(stderr, "\t-y 64bit mode\n"); fprintf(stderr, "\t-n 32bit mode\n"); - fprintf(stderr, "\t-v Verbose mode\n"); + fprintf(stderr, "\t-v Verbosity(-vv dumps any decoded result)\n"); fprintf(stderr, "\t-s Give a random seed (and iteration number)\n"); fprintf(stderr, "\t-m Give a maximum iteration number\n"); fprintf(stderr, "\t-i Give an input file with decoded binary\n"); @@ -188,7 +188,7 @@ static void parse_args(int argc, char **argv) x86_64 = 0; break; case 'v': - verbose = 1; + verbose++; break; case 'i': if (strcmp("-", optarg) == 0) @@ -264,7 +264,8 @@ int main(int argc, char **argv) errors++; } else if (verbose && !insn_complete(&insn)) dump_stream(stdout, "Info: Found an undecodable input", i, insn_buf, &insn); - + else if (verbose >= 2) + dump_insn(stdout, &insn); insns++; } -- cgit v1.2.3 From 0f5a2601284237e2ba089389fd75d67f77626cef Mon Sep 17 00:00:00 2001 From: Peter Zijlstra Date: Wed, 16 Nov 2011 14:38:16 +0100 Subject: perf: Avoid a useless pmu_disable() in the perf-tick Gleb writes: > Currently pmu is disabled and re-enabled on each timer interrupt even > when no rotation or frequency adjustment is needed. On Intel CPU this > results in two writes into PERF_GLOBAL_CTRL MSR per tick. On bare metal > it does not cause significant slowdown, but when running perf in a virtual > machine it leads to 20% slowdown on my machine. Cure this by keeping a perf_event_context::nr_freq counter that counts the number of active events that require frequency adjustments and use this in a similar fashion to the already existing nr_events != nr_active test in perf_rotate_context(). By being able to exclude both rotation and frequency adjustments a-priory for the common case we can avoid the otherwise superfluous PMU disable. Suggested-by: Gleb Natapov Signed-off-by: Peter Zijlstra Link: http://lkml.kernel.org/n/tip-515yhoatehd3gza7we9fapaa@git.kernel.org Signed-off-by: Ingo Molnar --- include/linux/perf_event.h | 1 + kernel/events/core.c | 48 ++++++++++++++++++++++++++++++---------------- 2 files changed, 33 insertions(+), 16 deletions(-) diff --git a/include/linux/perf_event.h b/include/linux/perf_event.h index b1f89122bf6a..cb44c9e75660 100644 --- a/include/linux/perf_event.h +++ b/include/linux/perf_event.h @@ -890,6 +890,7 @@ struct perf_event_context { int nr_active; int is_active; int nr_stat; + int nr_freq; int rotate_disable; atomic_t refcount; struct task_struct *task; diff --git a/kernel/events/core.c b/kernel/events/core.c index a355ffb0b28f..b3fed52aaf20 100644 --- a/kernel/events/core.c +++ b/kernel/events/core.c @@ -1130,6 +1130,8 @@ event_sched_out(struct perf_event *event, if (!is_software_event(event)) cpuctx->active_oncpu--; ctx->nr_active--; + if (event->attr.freq && event->attr.sample_freq) + ctx->nr_freq--; if (event->attr.exclusive || !cpuctx->active_oncpu) cpuctx->exclusive = 0; } @@ -1407,6 +1409,8 @@ event_sched_in(struct perf_event *event, if (!is_software_event(event)) cpuctx->active_oncpu++; ctx->nr_active++; + if (event->attr.freq && event->attr.sample_freq) + ctx->nr_freq++; if (event->attr.exclusive) cpuctx->exclusive = 1; @@ -2329,6 +2333,9 @@ static void perf_ctx_adjust_freq(struct perf_event_context *ctx, u64 period) u64 interrupts, now; s64 delta; + if (!ctx->nr_freq) + return; + list_for_each_entry_rcu(event, &ctx->event_list, event_entry) { if (event->state != PERF_EVENT_STATE_ACTIVE) continue; @@ -2384,12 +2391,14 @@ static void perf_rotate_context(struct perf_cpu_context *cpuctx) { u64 interval = (u64)cpuctx->jiffies_interval * TICK_NSEC; struct perf_event_context *ctx = NULL; - int rotate = 0, remove = 1; + int rotate = 0, remove = 1, freq = 0; if (cpuctx->ctx.nr_events) { remove = 0; if (cpuctx->ctx.nr_events != cpuctx->ctx.nr_active) rotate = 1; + if (cpuctx->ctx.nr_freq) + freq = 1; } ctx = cpuctx->task_ctx; @@ -2397,33 +2406,40 @@ static void perf_rotate_context(struct perf_cpu_context *cpuctx) remove = 0; if (ctx->nr_events != ctx->nr_active) rotate = 1; + if (ctx->nr_freq) + freq = 1; } + if (!rotate && !freq) + goto done; + perf_ctx_lock(cpuctx, cpuctx->task_ctx); perf_pmu_disable(cpuctx->ctx.pmu); - perf_ctx_adjust_freq(&cpuctx->ctx, interval); - if (ctx) - perf_ctx_adjust_freq(ctx, interval); - if (!rotate) - goto done; + if (freq) { + perf_ctx_adjust_freq(&cpuctx->ctx, interval); + if (ctx) + perf_ctx_adjust_freq(ctx, interval); + } - cpu_ctx_sched_out(cpuctx, EVENT_FLEXIBLE); - if (ctx) - ctx_sched_out(ctx, cpuctx, EVENT_FLEXIBLE); + if (rotate) { + cpu_ctx_sched_out(cpuctx, EVENT_FLEXIBLE); + if (ctx) + ctx_sched_out(ctx, cpuctx, EVENT_FLEXIBLE); - rotate_ctx(&cpuctx->ctx); - if (ctx) - rotate_ctx(ctx); + rotate_ctx(&cpuctx->ctx); + if (ctx) + rotate_ctx(ctx); + + perf_event_sched_in(cpuctx, ctx, current); + } - perf_event_sched_in(cpuctx, ctx, current); + perf_pmu_enable(cpuctx->ctx.pmu); + perf_ctx_unlock(cpuctx, cpuctx->task_ctx); done: if (remove) list_del_init(&cpuctx->rotation_list); - - perf_pmu_enable(cpuctx->ctx.pmu); - perf_ctx_unlock(cpuctx, cpuctx->task_ctx); } void perf_event_task_tick(void) -- cgit v1.2.3 From 1e2ad28f80b4e155678259238f51edebc19e4014 Mon Sep 17 00:00:00 2001 From: Robert Richter Date: Fri, 18 Nov 2011 12:35:21 +0100 Subject: perf, x86: Implement event scheduler helper functions This patch introduces x86 perf scheduler code helper functions. We need this to later add more complex functionality to support overlapping counter constraints (next patch). The algorithm is modified so that the range of weight values is now generated from the constraints. There shouldn't be other functional changes. With the helper functions the scheduler is controlled. There are functions to initialize, traverse the event list, find unused counters etc. The scheduler keeps its own state. V3: * Added macro for_each_set_bit_cont(). * Changed functions interfaces of perf_sched_find_counter() and perf_sched_next_event() to use bool as return value. * Added some comments to make code better understandable. V4: * Fix broken event assignment if weight of the first event is not wmin (perf_sched_init()). Signed-off-by: Robert Richter Signed-off-by: Peter Zijlstra Cc: Stephane Eranian Link: http://lkml.kernel.org/r/1321616122-1533-2-git-send-email-robert.richter@amd.com Signed-off-by: Ingo Molnar --- arch/x86/kernel/cpu/perf_event.c | 185 ++++++++++++++++++++++++++++----------- include/linux/bitops.h | 10 ++- 2 files changed, 140 insertions(+), 55 deletions(-) diff --git a/arch/x86/kernel/cpu/perf_event.c b/arch/x86/kernel/cpu/perf_event.c index 2bda212a0010..5a469d3d0c66 100644 --- a/arch/x86/kernel/cpu/perf_event.c +++ b/arch/x86/kernel/cpu/perf_event.c @@ -484,18 +484,145 @@ static inline int is_x86_event(struct perf_event *event) return event->pmu == &pmu; } +/* + * Event scheduler state: + * + * Assign events iterating over all events and counters, beginning + * with events with least weights first. Keep the current iterator + * state in struct sched_state. + */ +struct sched_state { + int weight; + int event; /* event index */ + int counter; /* counter index */ + int unassigned; /* number of events to be assigned left */ + unsigned long used[BITS_TO_LONGS(X86_PMC_IDX_MAX)]; +}; + +struct perf_sched { + int max_weight; + int max_events; + struct event_constraint **constraints; + struct sched_state state; +}; + +/* + * Initialize interator that runs through all events and counters. + */ +static void perf_sched_init(struct perf_sched *sched, struct event_constraint **c, + int num, int wmin, int wmax) +{ + int idx; + + memset(sched, 0, sizeof(*sched)); + sched->max_events = num; + sched->max_weight = wmax; + sched->constraints = c; + + for (idx = 0; idx < num; idx++) { + if (c[idx]->weight == wmin) + break; + } + + sched->state.event = idx; /* start with min weight */ + sched->state.weight = wmin; + sched->state.unassigned = num; +} + +/* + * Select a counter for the current event to schedule. Return true on + * success. + */ +static bool perf_sched_find_counter(struct perf_sched *sched) +{ + struct event_constraint *c; + int idx; + + if (!sched->state.unassigned) + return false; + + if (sched->state.event >= sched->max_events) + return false; + + c = sched->constraints[sched->state.event]; + + /* Grab the first unused counter starting with idx */ + idx = sched->state.counter; + for_each_set_bit_cont(idx, c->idxmsk, X86_PMC_IDX_MAX) { + if (!__test_and_set_bit(idx, sched->state.used)) + break; + } + sched->state.counter = idx; + + if (idx >= X86_PMC_IDX_MAX) + return false; + + return true; +} + +/* + * Go through all unassigned events and find the next one to schedule. + * Take events with the least weight first. Return true on success. + */ +static bool perf_sched_next_event(struct perf_sched *sched) +{ + struct event_constraint *c; + + if (!sched->state.unassigned || !--sched->state.unassigned) + return false; + + do { + /* next event */ + sched->state.event++; + if (sched->state.event >= sched->max_events) { + /* next weight */ + sched->state.event = 0; + sched->state.weight++; + if (sched->state.weight > sched->max_weight) + return false; + } + c = sched->constraints[sched->state.event]; + } while (c->weight != sched->state.weight); + + sched->state.counter = 0; /* start with first counter */ + + return true; +} + +/* + * Assign a counter for each event. + */ +static int perf_assign_events(struct event_constraint **constraints, int n, + int wmin, int wmax, int *assign) +{ + struct perf_sched sched; + + perf_sched_init(&sched, constraints, n, wmin, wmax); + + do { + if (!perf_sched_find_counter(&sched)) + break; /* failed */ + if (assign) + assign[sched.state.event] = sched.state.counter; + } while (perf_sched_next_event(&sched)); + + return sched.state.unassigned; +} + int x86_schedule_events(struct cpu_hw_events *cpuc, int n, int *assign) { struct event_constraint *c, *constraints[X86_PMC_IDX_MAX]; unsigned long used_mask[BITS_TO_LONGS(X86_PMC_IDX_MAX)]; - int i, j, w, wmax, num = 0; + int i, wmin, wmax, num = 0; struct hw_perf_event *hwc; bitmap_zero(used_mask, X86_PMC_IDX_MAX); - for (i = 0; i < n; i++) { + for (i = 0, wmin = X86_PMC_IDX_MAX, wmax = 0; i < n; i++) { c = x86_pmu.get_event_constraints(cpuc, cpuc->event_list[i]); constraints[i] = c; + wmin = min(wmin, c->weight); + wmax = max(wmax, c->weight); } /* @@ -521,59 +648,11 @@ int x86_schedule_events(struct cpu_hw_events *cpuc, int n, int *assign) if (assign) assign[i] = hwc->idx; } - if (i == n) - goto done; - /* - * begin slow path - */ + /* slow path */ + if (i != n) + num = perf_assign_events(constraints, n, wmin, wmax, assign); - bitmap_zero(used_mask, X86_PMC_IDX_MAX); - - /* - * weight = number of possible counters - * - * 1 = most constrained, only works on one counter - * wmax = least constrained, works on any counter - * - * assign events to counters starting with most - * constrained events. - */ - wmax = x86_pmu.num_counters; - - /* - * when fixed event counters are present, - * wmax is incremented by 1 to account - * for one more choice - */ - if (x86_pmu.num_counters_fixed) - wmax++; - - for (w = 1, num = n; num && w <= wmax; w++) { - /* for each event */ - for (i = 0; num && i < n; i++) { - c = constraints[i]; - hwc = &cpuc->event_list[i]->hw; - - if (c->weight != w) - continue; - - for_each_set_bit(j, c->idxmsk, X86_PMC_IDX_MAX) { - if (!test_bit(j, used_mask)) - break; - } - - if (j == X86_PMC_IDX_MAX) - break; - - __set_bit(j, used_mask); - - if (assign) - assign[i] = j; - num--; - } - } -done: /* * scheduling failed or is just a simulation, * free resources if necessary diff --git a/include/linux/bitops.h b/include/linux/bitops.h index a3ef66a2a083..3c1063acb2ab 100644 --- a/include/linux/bitops.h +++ b/include/linux/bitops.h @@ -22,8 +22,14 @@ extern unsigned long __sw_hweight64(__u64 w); #include #define for_each_set_bit(bit, addr, size) \ - for ((bit) = find_first_bit((addr), (size)); \ - (bit) < (size); \ + for ((bit) = find_first_bit((addr), (size)); \ + (bit) < (size); \ + (bit) = find_next_bit((addr), (size), (bit) + 1)) + +/* same as for_each_set_bit() but use bit as value to start with */ +#define for_each_set_bit_cont(bit, addr, size) \ + for ((bit) = find_next_bit((addr), (size), (bit)); \ + (bit) < (size); \ (bit) = find_next_bit((addr), (size), (bit) + 1)) static __inline__ int get_bitmask_order(unsigned int count) -- cgit v1.2.3 From bc1738f6ee83015f090867813dcca4d690e7917c Mon Sep 17 00:00:00 2001 From: Robert Richter Date: Fri, 18 Nov 2011 12:35:22 +0100 Subject: perf, x86: Fix event scheduler for constraints with overlapping counters The current x86 event scheduler fails to resolve scheduling problems of certain combinations of events and constraints. This happens if the counter mask of such an event is not a subset of any other counter mask of a constraint with an equal or higher weight, e.g. constraints of the AMD family 15h pmu: counter mask weight amd_f15_PMC30 0x09 2 <--- overlapping counters amd_f15_PMC20 0x07 3 amd_f15_PMC53 0x38 3 The scheduler does not find then an existing solution. Here is an example: event code counter failure possible solution 0x02E PMC[3,0] 0 3 0x043 PMC[2:0] 1 0 0x045 PMC[2:0] 2 1 0x046 PMC[2:0] FAIL 2 The event scheduler may not select the correct counter in the first cycle because it needs to know which subsequent events will be scheduled. It may fail to schedule the events then. To solve this, we now save the scheduler state of events with overlapping counter counstraints. If we fail to schedule the events we rollback to those states and try to use another free counter. Constraints with overlapping counters are marked with a new introduced overlap flag. We set the overlap flag for such constraints to give the scheduler a hint which events to select for counter rescheduling. The EVENT_CONSTRAINT_OVERLAP() macro can be used for this. Care must be taken as the rescheduling algorithm is O(n!) which will increase scheduling cycles for an over-commited system dramatically. The number of such EVENT_CONSTRAINT_OVERLAP() macros and its counter masks must be kept at a minimum. Thus, the current stack is limited to 2 states to limit the number of loops the algorithm takes in the worst case. On systems with no overlapping-counter constraints, this implementation does not increase the loop count compared to the previous algorithm. V2: * Renamed redo -> overlap. * Reimplementation using perf scheduling helper functions. V3: * Added WARN_ON_ONCE() if out of save states. * Changed function interface of perf_sched_restore_state() to use bool as return value. Signed-off-by: Robert Richter Signed-off-by: Peter Zijlstra Cc: Stephane Eranian Link: http://lkml.kernel.org/r/1321616122-1533-3-git-send-email-robert.richter@amd.com Signed-off-by: Ingo Molnar --- arch/x86/kernel/cpu/perf_event.c | 45 ++++++++++++++++++++++++++++++++++-- arch/x86/kernel/cpu/perf_event.h | 30 ++++++++++++++++++++++-- arch/x86/kernel/cpu/perf_event_amd.c | 2 +- 3 files changed, 72 insertions(+), 5 deletions(-) diff --git a/arch/x86/kernel/cpu/perf_event.c b/arch/x86/kernel/cpu/perf_event.c index 5a469d3d0c66..fa6fdec5afbc 100644 --- a/arch/x86/kernel/cpu/perf_event.c +++ b/arch/x86/kernel/cpu/perf_event.c @@ -499,11 +499,16 @@ struct sched_state { unsigned long used[BITS_TO_LONGS(X86_PMC_IDX_MAX)]; }; +/* Total max is X86_PMC_IDX_MAX, but we are O(n!) limited */ +#define SCHED_STATES_MAX 2 + struct perf_sched { int max_weight; int max_events; struct event_constraint **constraints; struct sched_state state; + int saved_states; + struct sched_state saved[SCHED_STATES_MAX]; }; /* @@ -529,11 +534,34 @@ static void perf_sched_init(struct perf_sched *sched, struct event_constraint ** sched->state.unassigned = num; } +static void perf_sched_save_state(struct perf_sched *sched) +{ + if (WARN_ON_ONCE(sched->saved_states >= SCHED_STATES_MAX)) + return; + + sched->saved[sched->saved_states] = sched->state; + sched->saved_states++; +} + +static bool perf_sched_restore_state(struct perf_sched *sched) +{ + if (!sched->saved_states) + return false; + + sched->saved_states--; + sched->state = sched->saved[sched->saved_states]; + + /* continue with next counter: */ + clear_bit(sched->state.counter++, sched->state.used); + + return true; +} + /* * Select a counter for the current event to schedule. Return true on * success. */ -static bool perf_sched_find_counter(struct perf_sched *sched) +static bool __perf_sched_find_counter(struct perf_sched *sched) { struct event_constraint *c; int idx; @@ -557,6 +585,19 @@ static bool perf_sched_find_counter(struct perf_sched *sched) if (idx >= X86_PMC_IDX_MAX) return false; + if (c->overlap) + perf_sched_save_state(sched); + + return true; +} + +static bool perf_sched_find_counter(struct perf_sched *sched) +{ + while (!__perf_sched_find_counter(sched)) { + if (!perf_sched_restore_state(sched)) + return false; + } + return true; } @@ -1250,7 +1291,7 @@ static int __init init_hw_perf_events(void) unconstrained = (struct event_constraint) __EVENT_CONSTRAINT(0, (1ULL << x86_pmu.num_counters) - 1, - 0, x86_pmu.num_counters); + 0, x86_pmu.num_counters, 0); if (x86_pmu.event_constraints) { for_each_event_constraint(c, x86_pmu.event_constraints) { diff --git a/arch/x86/kernel/cpu/perf_event.h b/arch/x86/kernel/cpu/perf_event.h index b9698d40ac4b..51a985cbc12f 100644 --- a/arch/x86/kernel/cpu/perf_event.h +++ b/arch/x86/kernel/cpu/perf_event.h @@ -45,6 +45,7 @@ struct event_constraint { u64 code; u64 cmask; int weight; + int overlap; }; struct amd_nb { @@ -151,15 +152,40 @@ struct cpu_hw_events { void *kfree_on_online; }; -#define __EVENT_CONSTRAINT(c, n, m, w) {\ +#define __EVENT_CONSTRAINT(c, n, m, w, o) {\ { .idxmsk64 = (n) }, \ .code = (c), \ .cmask = (m), \ .weight = (w), \ + .overlap = (o), \ } #define EVENT_CONSTRAINT(c, n, m) \ - __EVENT_CONSTRAINT(c, n, m, HWEIGHT(n)) + __EVENT_CONSTRAINT(c, n, m, HWEIGHT(n), 0) + +/* + * The overlap flag marks event constraints with overlapping counter + * masks. This is the case if the counter mask of such an event is not + * a subset of any other counter mask of a constraint with an equal or + * higher weight, e.g.: + * + * c_overlaps = EVENT_CONSTRAINT_OVERLAP(0, 0x09, 0); + * c_another1 = EVENT_CONSTRAINT(0, 0x07, 0); + * c_another2 = EVENT_CONSTRAINT(0, 0x38, 0); + * + * The event scheduler may not select the correct counter in the first + * cycle because it needs to know which subsequent events will be + * scheduled. It may fail to schedule the events then. So we set the + * overlap flag for such constraints to give the scheduler a hint which + * events to select for counter rescheduling. + * + * Care must be taken as the rescheduling algorithm is O(n!) which + * will increase scheduling cycles for an over-commited system + * dramatically. The number of such EVENT_CONSTRAINT_OVERLAP() macros + * and its counter masks must be kept at a minimum. + */ +#define EVENT_CONSTRAINT_OVERLAP(c, n, m) \ + __EVENT_CONSTRAINT(c, n, m, HWEIGHT(n), 1) /* * Constraint on the Event code. diff --git a/arch/x86/kernel/cpu/perf_event_amd.c b/arch/x86/kernel/cpu/perf_event_amd.c index aeefd45697a2..0397b23be8e9 100644 --- a/arch/x86/kernel/cpu/perf_event_amd.c +++ b/arch/x86/kernel/cpu/perf_event_amd.c @@ -492,7 +492,7 @@ static __initconst const struct x86_pmu amd_pmu = { static struct event_constraint amd_f15_PMC0 = EVENT_CONSTRAINT(0, 0x01, 0); static struct event_constraint amd_f15_PMC20 = EVENT_CONSTRAINT(0, 0x07, 0); static struct event_constraint amd_f15_PMC3 = EVENT_CONSTRAINT(0, 0x08, 0); -static struct event_constraint amd_f15_PMC30 = EVENT_CONSTRAINT(0, 0x09, 0); +static struct event_constraint amd_f15_PMC30 = EVENT_CONSTRAINT_OVERLAP(0, 0x09, 0); static struct event_constraint amd_f15_PMC50 = EVENT_CONSTRAINT(0, 0x3F, 0); static struct event_constraint amd_f15_PMC53 = EVENT_CONSTRAINT(0, 0x38, 0); -- cgit v1.2.3 From 4defea8559bc0f97a899d94c8d19d3b8bb802bc4 Mon Sep 17 00:00:00 2001 From: Peter Zijlstra Date: Thu, 10 Nov 2011 15:15:42 +0100 Subject: perf, x86: Prefer fixed-purpose counters when scheduling This avoids a scheduling failure for cases like: cycles, cycles, instructions, instructions (on Core2) Which would end up being programmed like: PMC0, PMC1, FP-instructions, fail Because all events will have the same weight. Signed-off-by: Peter Zijlstra Link: http://lkml.kernel.org/n/tip-8tnwb92asqj7xajqqoty4gel@git.kernel.org Signed-off-by: Ingo Molnar --- arch/x86/kernel/cpu/perf_event.c | 19 ++++++++++++++----- 1 file changed, 14 insertions(+), 5 deletions(-) diff --git a/arch/x86/kernel/cpu/perf_event.c b/arch/x86/kernel/cpu/perf_event.c index fa6fdec5afbc..66f8ba9a67f9 100644 --- a/arch/x86/kernel/cpu/perf_event.c +++ b/arch/x86/kernel/cpu/perf_event.c @@ -574,16 +574,25 @@ static bool __perf_sched_find_counter(struct perf_sched *sched) c = sched->constraints[sched->state.event]; + /* Prefer fixed purpose counters */ + if (x86_pmu.num_counters_fixed) { + idx = X86_PMC_IDX_FIXED; + for_each_set_bit_cont(idx, c->idxmsk, X86_PMC_IDX_MAX) { + if (!__test_and_set_bit(idx, sched->state.used)) + goto done; + } + } /* Grab the first unused counter starting with idx */ idx = sched->state.counter; - for_each_set_bit_cont(idx, c->idxmsk, X86_PMC_IDX_MAX) { + for_each_set_bit_cont(idx, c->idxmsk, X86_PMC_IDX_FIXED) { if (!__test_and_set_bit(idx, sched->state.used)) - break; + goto done; } - sched->state.counter = idx; - if (idx >= X86_PMC_IDX_MAX) - return false; + return false; + +done: + sched->state.counter = idx; if (c->overlap) perf_sched_save_state(sched); -- cgit v1.2.3 From 1d9b482e78d3b16f97f85a82849f82db7eed3102 Mon Sep 17 00:00:00 2001 From: Peter Zijlstra Date: Wed, 23 Nov 2011 12:34:20 +0100 Subject: perf: Remove superfluous arguments Signed-off-by: Peter Zijlstra Link: http://lkml.kernel.org/n/tip-yv4o74vh90suyghccgykbnry@git.kernel.org Signed-off-by: Ingo Molnar --- kernel/events/core.c | 9 ++++----- 1 file changed, 4 insertions(+), 5 deletions(-) diff --git a/kernel/events/core.c b/kernel/events/core.c index b3fed52aaf20..eeda5403590c 100644 --- a/kernel/events/core.c +++ b/kernel/events/core.c @@ -1667,8 +1667,7 @@ retry: * Note: this works for group members as well as group leaders * since the non-leader members' sibling_lists will be empty. */ -static void __perf_event_mark_enabled(struct perf_event *event, - struct perf_event_context *ctx) +static void __perf_event_mark_enabled(struct perf_event *event) { struct perf_event *sub; u64 tstamp = perf_event_time(event); @@ -1706,7 +1705,7 @@ static int __perf_event_enable(void *info) */ perf_cgroup_set_timestamp(current, ctx); - __perf_event_mark_enabled(event, ctx); + __perf_event_mark_enabled(event); if (!event_filter_match(event)) { if (is_cgroup_event(event)) @@ -1787,7 +1786,7 @@ void perf_event_enable(struct perf_event *event) retry: if (!ctx->is_active) { - __perf_event_mark_enabled(event, ctx); + __perf_event_mark_enabled(event); goto out; } @@ -2466,7 +2465,7 @@ static int event_enable_on_exec(struct perf_event *event, if (event->state >= PERF_EVENT_STATE_INACTIVE) return 0; - __perf_event_mark_enabled(event, ctx); + __perf_event_mark_enabled(event); return 1; } -- cgit v1.2.3 From b79387ef185af2323594920923cecba5753c3817 Mon Sep 17 00:00:00 2001 From: Peter Zijlstra Date: Tue, 22 Nov 2011 11:25:43 +0100 Subject: perf: Fix enable_on_exec for sibling events Deng-Cheng Zhu reported that sibling events that were created disabled with enable_on_exec would never get enabled. Iterate all events instead of the group lists. Reported-by: Deng-Cheng Zhu Tested-by: Deng-Cheng Zhu Signed-off-by: Peter Zijlstra Link: http://lkml.kernel.org/r/1322048382.14799.41.camel@twins Signed-off-by: Ingo Molnar --- kernel/events/core.c | 8 +------- 1 file changed, 1 insertion(+), 7 deletions(-) diff --git a/kernel/events/core.c b/kernel/events/core.c index eeda5403590c..3c1541d7a53d 100644 --- a/kernel/events/core.c +++ b/kernel/events/core.c @@ -2497,13 +2497,7 @@ static void perf_event_enable_on_exec(struct perf_event_context *ctx) raw_spin_lock(&ctx->lock); task_ctx_sched_out(ctx); - list_for_each_entry(event, &ctx->pinned_groups, group_entry) { - ret = event_enable_on_exec(event, ctx); - if (ret) - enabled = 1; - } - - list_for_each_entry(event, &ctx->flexible_groups, group_entry) { + list_for_each_entry(event, &ctx->event_list, event_entry) { ret = event_enable_on_exec(event, ctx); if (ret) enabled = 1; -- cgit v1.2.3 From b202952075f62603bea9bfb6ebc6b0420db11949 Mon Sep 17 00:00:00 2001 From: Gleb Natapov Date: Sun, 27 Nov 2011 17:59:09 +0200 Subject: perf, core: Rate limit perf_sched_events jump_label patching jump_lable patching is very expensive operation that involves pausing all cpus. The patching of perf_sched_events jump_label is easily controllable from userspace by unprivileged user. When te user runs a loop like this: "while true; do perf stat -e cycles true; done" ... the performance of my test application that just increments a counter for one second drops by 4%. This is on a 16 cpu box with my test application using only one of them. An impact on a real server doing real work will be worse. Performance of KVM PMU drops nearly 50% due to jump_lable for "perf record" since KVM PMU implementation creates and destroys perf event frequently. This patch introduces a way to rate limit jump_label patching and uses it to fix the above problem. I believe that as jump_label use will spread the problem will become more common and thus solving it in a generic code is appropriate. Also fixing it in the perf code would result in moving jump_label accounting logic to perf code with all the ifdefs in case of JUMP_LABEL=n kernel. With this patch all details are nicely hidden inside jump_label code. Signed-off-by: Gleb Natapov Acked-by: Jason Baron Signed-off-by: Peter Zijlstra Link: http://lkml.kernel.org/r/20111127155909.GO2557@redhat.com Signed-off-by: Ingo Molnar --- include/linux/jump_label.h | 24 ++++++++++++++++++++++++ include/linux/perf_event.h | 6 +++--- kernel/events/core.c | 13 ++++++++----- kernel/jump_label.c | 35 +++++++++++++++++++++++++++++++++-- 4 files changed, 68 insertions(+), 10 deletions(-) diff --git a/include/linux/jump_label.h b/include/linux/jump_label.h index 388b0d425b50..a1e7f909c801 100644 --- a/include/linux/jump_label.h +++ b/include/linux/jump_label.h @@ -3,6 +3,7 @@ #include #include +#include #if defined(CC_HAVE_ASM_GOTO) && defined(CONFIG_JUMP_LABEL) @@ -14,6 +15,12 @@ struct jump_label_key { #endif }; +struct jump_label_key_deferred { + struct jump_label_key key; + unsigned long timeout; + struct delayed_work work; +}; + # include # define HAVE_JUMP_LABEL #endif /* CC_HAVE_ASM_GOTO && CONFIG_JUMP_LABEL */ @@ -51,8 +58,11 @@ extern void arch_jump_label_transform_static(struct jump_entry *entry, extern int jump_label_text_reserved(void *start, void *end); extern void jump_label_inc(struct jump_label_key *key); extern void jump_label_dec(struct jump_label_key *key); +extern void jump_label_dec_deferred(struct jump_label_key_deferred *key); extern bool jump_label_enabled(struct jump_label_key *key); extern void jump_label_apply_nops(struct module *mod); +extern void jump_label_rate_limit(struct jump_label_key_deferred *key, + unsigned long rl); #else /* !HAVE_JUMP_LABEL */ @@ -68,6 +78,10 @@ static __always_inline void jump_label_init(void) { } +struct jump_label_key_deferred { + struct jump_label_key key; +}; + static __always_inline bool static_branch(struct jump_label_key *key) { if (unlikely(atomic_read(&key->enabled))) @@ -85,6 +99,11 @@ static inline void jump_label_dec(struct jump_label_key *key) atomic_dec(&key->enabled); } +static inline void jump_label_dec_deferred(struct jump_label_key_deferred *key) +{ + jump_label_dec(&key->key); +} + static inline int jump_label_text_reserved(void *start, void *end) { return 0; @@ -102,6 +121,11 @@ static inline int jump_label_apply_nops(struct module *mod) { return 0; } + +static inline void jump_label_rate_limit(struct jump_label_key_deferred *key, + unsigned long rl) +{ +} #endif /* HAVE_JUMP_LABEL */ #endif /* _LINUX_JUMP_LABEL_H */ diff --git a/include/linux/perf_event.h b/include/linux/perf_event.h index cb44c9e75660..564769cdb473 100644 --- a/include/linux/perf_event.h +++ b/include/linux/perf_event.h @@ -1064,12 +1064,12 @@ perf_sw_event(u32 event_id, u64 nr, struct pt_regs *regs, u64 addr) } } -extern struct jump_label_key perf_sched_events; +extern struct jump_label_key_deferred perf_sched_events; static inline void perf_event_task_sched_in(struct task_struct *prev, struct task_struct *task) { - if (static_branch(&perf_sched_events)) + if (static_branch(&perf_sched_events.key)) __perf_event_task_sched_in(prev, task); } @@ -1078,7 +1078,7 @@ static inline void perf_event_task_sched_out(struct task_struct *prev, { perf_sw_event(PERF_COUNT_SW_CONTEXT_SWITCHES, 1, NULL, 0); - if (static_branch(&perf_sched_events)) + if (static_branch(&perf_sched_events.key)) __perf_event_task_sched_out(prev, next); } diff --git a/kernel/events/core.c b/kernel/events/core.c index 3c1541d7a53d..3a3b1a18f490 100644 --- a/kernel/events/core.c +++ b/kernel/events/core.c @@ -128,7 +128,7 @@ enum event_type_t { * perf_sched_events : >0 events exist * perf_cgroup_events: >0 per-cpu cgroup events exist on this cpu */ -struct jump_label_key perf_sched_events __read_mostly; +struct jump_label_key_deferred perf_sched_events __read_mostly; static DEFINE_PER_CPU(atomic_t, perf_cgroup_events); static atomic_t nr_mmap_events __read_mostly; @@ -2748,7 +2748,7 @@ static void free_event(struct perf_event *event) if (!event->parent) { if (event->attach_state & PERF_ATTACH_TASK) - jump_label_dec(&perf_sched_events); + jump_label_dec_deferred(&perf_sched_events); if (event->attr.mmap || event->attr.mmap_data) atomic_dec(&nr_mmap_events); if (event->attr.comm) @@ -2759,7 +2759,7 @@ static void free_event(struct perf_event *event) put_callchain_buffers(); if (is_cgroup_event(event)) { atomic_dec(&per_cpu(perf_cgroup_events, event->cpu)); - jump_label_dec(&perf_sched_events); + jump_label_dec_deferred(&perf_sched_events); } } @@ -5784,7 +5784,7 @@ done: if (!event->parent) { if (event->attach_state & PERF_ATTACH_TASK) - jump_label_inc(&perf_sched_events); + jump_label_inc(&perf_sched_events.key); if (event->attr.mmap || event->attr.mmap_data) atomic_inc(&nr_mmap_events); if (event->attr.comm) @@ -6022,7 +6022,7 @@ SYSCALL_DEFINE5(perf_event_open, * - that may need work on context switch */ atomic_inc(&per_cpu(perf_cgroup_events, event->cpu)); - jump_label_inc(&perf_sched_events); + jump_label_inc(&perf_sched_events.key); } /* @@ -6868,6 +6868,9 @@ void __init perf_event_init(void) ret = init_hw_breakpoint(); WARN(ret, "hw_breakpoint initialization failed with: %d", ret); + + /* do not patch jump label more than once per second */ + jump_label_rate_limit(&perf_sched_events, HZ); } static int __init perf_event_sysfs_init(void) diff --git a/kernel/jump_label.c b/kernel/jump_label.c index 66ff7109f697..51a175ab0a03 100644 --- a/kernel/jump_label.c +++ b/kernel/jump_label.c @@ -72,15 +72,46 @@ void jump_label_inc(struct jump_label_key *key) jump_label_unlock(); } -void jump_label_dec(struct jump_label_key *key) +static void __jump_label_dec(struct jump_label_key *key, + unsigned long rate_limit, struct delayed_work *work) { if (!atomic_dec_and_mutex_lock(&key->enabled, &jump_label_mutex)) return; - jump_label_update(key, JUMP_LABEL_DISABLE); + if (rate_limit) { + atomic_inc(&key->enabled); + schedule_delayed_work(work, rate_limit); + } else + jump_label_update(key, JUMP_LABEL_DISABLE); + jump_label_unlock(); } +static void jump_label_update_timeout(struct work_struct *work) +{ + struct jump_label_key_deferred *key = + container_of(work, struct jump_label_key_deferred, work.work); + __jump_label_dec(&key->key, 0, NULL); +} + +void jump_label_dec(struct jump_label_key *key) +{ + __jump_label_dec(key, 0, NULL); +} + +void jump_label_dec_deferred(struct jump_label_key_deferred *key) +{ + __jump_label_dec(&key->key, key->timeout, &key->work); +} + + +void jump_label_rate_limit(struct jump_label_key_deferred *key, + unsigned long rl) +{ + key->timeout = rl; + INIT_DELAYED_WORK(&key->work, jump_label_update_timeout); +} + static int addr_conflict(struct jump_entry *entry, void *start, void *end) { if (entry->code <= (unsigned long)end && -- cgit v1.2.3 From 9cdbe1cbac4ec318037297175587a0080acc9d11 Mon Sep 17 00:00:00 2001 From: Peter Zijlstra Date: Tue, 6 Dec 2011 17:27:29 +0100 Subject: jump_label, x86: Fix section mismatch WARNING: arch/x86/kernel/built-in.o(.text+0x4c71): Section mismatch in reference from the function arch_jump_label_transform_static() to the function .init.text:text_poke_early() The function arch_jump_label_transform_static() references the function __init text_poke_early(). This is often because arch_jump_label_transform_static lacks a __init annotation or the annotation of text_poke_early is wrong. Signed-off-by: Peter Zijlstra Cc: Jason Baron Link: http://lkml.kernel.org/n/tip-9lefe89mrvurrwpqw5h8xm8z@git.kernel.org Signed-off-by: Ingo Molnar --- arch/x86/kernel/jump_label.c | 2 +- kernel/jump_label.c | 2 +- 2 files changed, 2 insertions(+), 2 deletions(-) diff --git a/arch/x86/kernel/jump_label.c b/arch/x86/kernel/jump_label.c index ea9d5f2f13ef..2889b3d43882 100644 --- a/arch/x86/kernel/jump_label.c +++ b/arch/x86/kernel/jump_label.c @@ -50,7 +50,7 @@ void arch_jump_label_transform(struct jump_entry *entry, put_online_cpus(); } -void arch_jump_label_transform_static(struct jump_entry *entry, +__init_or_module void arch_jump_label_transform_static(struct jump_entry *entry, enum jump_label_type type) { __jump_label_transform(entry, type, text_poke_early); diff --git a/kernel/jump_label.c b/kernel/jump_label.c index 51a175ab0a03..3fb7b79c86fd 100644 --- a/kernel/jump_label.c +++ b/kernel/jump_label.c @@ -142,7 +142,7 @@ static int __jump_label_text_reserved(struct jump_entry *iter_start, * running code can override this to make the non-live update case * cheaper. */ -void __weak arch_jump_label_transform_static(struct jump_entry *entry, +void __weak __init_or_module arch_jump_label_transform_static(struct jump_entry *entry, enum jump_label_type type) { arch_jump_label_transform(entry, type); -- cgit v1.2.3 From ac99b862fb98a36929831791da31714f709c2aa8 Mon Sep 17 00:00:00 2001 From: Peter Zijlstra Date: Wed, 6 Jul 2011 14:20:14 +0200 Subject: jump_label: Provide jump_label_key initializers Provide two initializers for jump_label_key that initialize it enabled or disabled. Also modify all jump_label code to allow for jump_labels to be initialized enabled. Signed-off-by: Peter Zijlstra Cc: Jason Baron Link: http://lkml.kernel.org/n/tip-p40e3yj21b68y03z1yv825e7@git.kernel.org Signed-off-by: Ingo Molnar --- include/linux/jump_label.h | 3 +++ kernel/jump_label.c | 12 ++++++++---- 2 files changed, 11 insertions(+), 4 deletions(-) diff --git a/include/linux/jump_label.h b/include/linux/jump_label.h index a1e7f909c801..5ce8b140428f 100644 --- a/include/linux/jump_label.h +++ b/include/linux/jump_label.h @@ -128,4 +128,7 @@ static inline void jump_label_rate_limit(struct jump_label_key_deferred *key, } #endif /* HAVE_JUMP_LABEL */ +#define jump_label_key_enabled ((struct jump_label_key){ .enabled = ATOMIC_INIT(1), }) +#define jump_label_key_disabled ((struct jump_label_key){ .enabled = ATOMIC_INIT(0), }) + #endif /* _LINUX_JUMP_LABEL_H */ diff --git a/kernel/jump_label.c b/kernel/jump_label.c index 3fb7b79c86fd..30c3c7708132 100644 --- a/kernel/jump_label.c +++ b/kernel/jump_label.c @@ -248,8 +248,13 @@ void jump_label_apply_nops(struct module *mod) if (iter_start == iter_stop) return; - for (iter = iter_start; iter < iter_stop; iter++) - arch_jump_label_transform_static(iter, JUMP_LABEL_DISABLE); + for (iter = iter_start; iter < iter_stop; iter++) { + struct jump_label_key *iterk; + + iterk = (struct jump_label_key *)(unsigned long)iter->key; + arch_jump_label_transform_static(iter, jump_label_enabled(iterk) ? + JUMP_LABEL_ENABLE : JUMP_LABEL_DISABLE); + } } static int jump_label_add_module(struct module *mod) @@ -289,8 +294,7 @@ static int jump_label_add_module(struct module *mod) key->next = jlm; if (jump_label_enabled(key)) - __jump_label_update(key, iter, iter_stop, - JUMP_LABEL_ENABLE); + __jump_label_update(key, iter, iter_stop, JUMP_LABEL_ENABLE); } return 0; -- cgit v1.2.3 From ffb871bc9156ee2e5cf442f61250c5bd6aad17e3 Mon Sep 17 00:00:00 2001 From: Gleb Natapov Date: Thu, 10 Nov 2011 14:57:26 +0200 Subject: x86, perf: Disable non available architectural events Intel CPUs report non-available architectural events in cpuid leaf 0AH.EBX. Use it to disable events that are not available according to CPU. Signed-off-by: Gleb Natapov Signed-off-by: Peter Zijlstra Link: http://lkml.kernel.org/r/1320929850-10480-7-git-send-email-gleb@redhat.com Signed-off-by: Ingo Molnar --- arch/x86/include/asm/perf_event.h | 14 ++++++++++++++ arch/x86/kernel/cpu/perf_event.h | 5 +++++ arch/x86/kernel/cpu/perf_event_intel.c | 28 +++++++++++++++++++++++----- 3 files changed, 42 insertions(+), 5 deletions(-) diff --git a/arch/x86/include/asm/perf_event.h b/arch/x86/include/asm/perf_event.h index f61c62f7d5d8..c6998bc75456 100644 --- a/arch/x86/include/asm/perf_event.h +++ b/arch/x86/include/asm/perf_event.h @@ -57,6 +57,7 @@ (1 << (ARCH_PERFMON_UNHALTED_CORE_CYCLES_INDEX)) #define ARCH_PERFMON_BRANCH_MISSES_RETIRED 6 +#define ARCH_PERFMON_EVENTS_COUNT 7 /* * Intel "Architectural Performance Monitoring" CPUID @@ -72,6 +73,19 @@ union cpuid10_eax { unsigned int full; }; +union cpuid10_ebx { + struct { + unsigned int no_unhalted_core_cycles:1; + unsigned int no_instructions_retired:1; + unsigned int no_unhalted_reference_cycles:1; + unsigned int no_llc_reference:1; + unsigned int no_llc_misses:1; + unsigned int no_branch_instruction_retired:1; + unsigned int no_branch_misses_retired:1; + } split; + unsigned int full; +}; + union cpuid10_edx { struct { unsigned int num_counters_fixed:5; diff --git a/arch/x86/kernel/cpu/perf_event.h b/arch/x86/kernel/cpu/perf_event.h index 51a985cbc12f..f49c5c21085c 100644 --- a/arch/x86/kernel/cpu/perf_event.h +++ b/arch/x86/kernel/cpu/perf_event.h @@ -285,6 +285,11 @@ struct x86_pmu { int num_counters_fixed; int cntval_bits; u64 cntval_mask; + union { + unsigned long events_maskl; + unsigned long events_mask[BITS_TO_LONGS(ARCH_PERFMON_EVENTS_COUNT)]; + }; + int events_mask_len; int apic; u64 max_period; struct event_constraint * diff --git a/arch/x86/kernel/cpu/perf_event_intel.c b/arch/x86/kernel/cpu/perf_event_intel.c index 8d601b18bf9f..201156b80a37 100644 --- a/arch/x86/kernel/cpu/perf_event_intel.c +++ b/arch/x86/kernel/cpu/perf_event_intel.c @@ -1552,13 +1552,23 @@ static void intel_sandybridge_quirks(void) x86_pmu.pebs_constraints = NULL; } +static const int intel_event_id_to_hw_id[] __initconst = { + PERF_COUNT_HW_CPU_CYCLES, + PERF_COUNT_HW_INSTRUCTIONS, + PERF_COUNT_HW_BUS_CYCLES, + PERF_COUNT_HW_CACHE_REFERENCES, + PERF_COUNT_HW_CACHE_MISSES, + PERF_COUNT_HW_BRANCH_INSTRUCTIONS, + PERF_COUNT_HW_BRANCH_MISSES, +}; + __init int intel_pmu_init(void) { union cpuid10_edx edx; union cpuid10_eax eax; + union cpuid10_ebx ebx; unsigned int unused; - unsigned int ebx; - int version; + int version, bit; if (!cpu_has(&boot_cpu_data, X86_FEATURE_ARCH_PERFMON)) { switch (boot_cpu_data.x86) { @@ -1574,8 +1584,8 @@ __init int intel_pmu_init(void) * Check whether the Architectural PerfMon supports * Branch Misses Retired hw_event or not. */ - cpuid(10, &eax.full, &ebx, &unused, &edx.full); - if (eax.split.mask_length <= ARCH_PERFMON_BRANCH_MISSES_RETIRED) + cpuid(10, &eax.full, &ebx.full, &unused, &edx.full); + if (eax.split.mask_length < ARCH_PERFMON_EVENTS_COUNT) return -ENODEV; version = eax.split.version_id; @@ -1651,7 +1661,7 @@ __init int intel_pmu_init(void) /* UOPS_EXECUTED.CORE_ACTIVE_CYCLES,c=1,i=1 */ intel_perfmon_event_map[PERF_COUNT_HW_STALLED_CYCLES_BACKEND] = 0x1803fb1; - if (ebx & 0x40) { + if (ebx.split.no_branch_misses_retired) { /* * Erratum AAJ80 detected, we work it around by using * the BR_MISP_EXEC.ANY event. This will over-count @@ -1659,6 +1669,7 @@ __init int intel_pmu_init(void) * architectural event which is often completely bogus: */ intel_perfmon_event_map[PERF_COUNT_HW_BRANCH_MISSES] = 0x7f89; + ebx.split.no_branch_misses_retired = 0; pr_cont("erratum AAJ80 worked around, "); } @@ -1738,5 +1749,12 @@ __init int intel_pmu_init(void) break; } } + x86_pmu.events_maskl = ebx.full; + x86_pmu.events_mask_len = eax.split.mask_length; + + /* disable event that reported as not presend by cpuid */ + for_each_set_bit(bit, x86_pmu.events_mask, ARRAY_SIZE(intel_event_id_to_hw_id)) + intel_perfmon_event_map[intel_event_id_to_hw_id[bit]] = 0; + return 0; } -- cgit v1.2.3 From c1d6f42f1a42c721513e2f388c208e5348004f64 Mon Sep 17 00:00:00 2001 From: Peter Zijlstra Date: Tue, 6 Dec 2011 14:07:15 +0100 Subject: perf, x86: Implement arch event mask as quirk Implement the disabling of arch events as a quirk so that we can print a message along with it. This creates some visibility into the problem space and could allow us to work on adding more work-around like the AAJ80 one. Requested-by: Ingo Molnar Cc: Gleb Natapov Signed-off-by: Peter Zijlstra Link: http://lkml.kernel.org/n/tip-wcja2z48wklzu1b0nkz0a5y7@git.kernel.org Signed-off-by: Ingo Molnar --- arch/x86/kernel/cpu/perf_event.c | 5 ++- arch/x86/kernel/cpu/perf_event.h | 16 ++++++- arch/x86/kernel/cpu/perf_event_intel.c | 80 +++++++++++++++++++++------------- 3 files changed, 68 insertions(+), 33 deletions(-) diff --git a/arch/x86/kernel/cpu/perf_event.c b/arch/x86/kernel/cpu/perf_event.c index 66f8ba9a67f9..55889e0b1452 100644 --- a/arch/x86/kernel/cpu/perf_event.c +++ b/arch/x86/kernel/cpu/perf_event.c @@ -1248,6 +1248,7 @@ static void __init pmu_check_apic(void) static int __init init_hw_perf_events(void) { + struct x86_pmu_quirk *quirk; struct event_constraint *c; int err; @@ -1276,8 +1277,8 @@ static int __init init_hw_perf_events(void) pr_cont("%s PMU driver.\n", x86_pmu.name); - if (x86_pmu.quirks) - x86_pmu.quirks(); + for (quirk = x86_pmu.quirks; quirk; quirk = quirk->next) + quirk->func(); if (x86_pmu.num_counters > X86_PMC_MAX_GENERIC) { WARN(1, KERN_ERR "hw perf events %d > max(%d), clipping!", diff --git a/arch/x86/kernel/cpu/perf_event.h b/arch/x86/kernel/cpu/perf_event.h index f49c5c21085c..8944062f46e2 100644 --- a/arch/x86/kernel/cpu/perf_event.h +++ b/arch/x86/kernel/cpu/perf_event.h @@ -261,6 +261,11 @@ union perf_capabilities { u64 capabilities; }; +struct x86_pmu_quirk { + struct x86_pmu_quirk *next; + void (*func)(void); +}; + /* * struct x86_pmu - generic x86 pmu */ @@ -299,7 +304,7 @@ struct x86_pmu { void (*put_event_constraints)(struct cpu_hw_events *cpuc, struct perf_event *event); struct event_constraint *event_constraints; - void (*quirks)(void); + struct x86_pmu_quirk *quirks; int perfctr_second_write; int (*cpu_prepare)(int cpu); @@ -340,6 +345,15 @@ struct x86_pmu { struct perf_guest_switch_msr *(*guest_get_msrs)(int *nr); }; +#define x86_add_quirk(func_) \ +do { \ + static struct x86_pmu_quirk __quirk __initdata = { \ + .func = func_, \ + }; \ + __quirk.next = x86_pmu.quirks; \ + x86_pmu.quirks = &__quirk; \ +} while (0) + #define ERF_NO_HT_SHARING 1 #define ERF_HAS_RSP_1 2 diff --git a/arch/x86/kernel/cpu/perf_event_intel.c b/arch/x86/kernel/cpu/perf_event_intel.c index 201156b80a37..2c3bf53d0302 100644 --- a/arch/x86/kernel/cpu/perf_event_intel.c +++ b/arch/x86/kernel/cpu/perf_event_intel.c @@ -1519,7 +1519,7 @@ static __initconst const struct x86_pmu intel_pmu = { .guest_get_msrs = intel_guest_get_msrs, }; -static void intel_clovertown_quirks(void) +static __init void intel_clovertown_quirk(void) { /* * PEBS is unreliable due to: @@ -1545,30 +1545,61 @@ static void intel_clovertown_quirks(void) x86_pmu.pebs_constraints = NULL; } -static void intel_sandybridge_quirks(void) +static __init void intel_sandybridge_quirk(void) { printk(KERN_WARNING "PEBS disabled due to CPU errata.\n"); x86_pmu.pebs = 0; x86_pmu.pebs_constraints = NULL; } -static const int intel_event_id_to_hw_id[] __initconst = { - PERF_COUNT_HW_CPU_CYCLES, - PERF_COUNT_HW_INSTRUCTIONS, - PERF_COUNT_HW_BUS_CYCLES, - PERF_COUNT_HW_CACHE_REFERENCES, - PERF_COUNT_HW_CACHE_MISSES, - PERF_COUNT_HW_BRANCH_INSTRUCTIONS, - PERF_COUNT_HW_BRANCH_MISSES, +static const struct { int id; char *name; } intel_arch_events_map[] __initconst = { + { PERF_COUNT_HW_CPU_CYCLES, "cpu cycles" }, + { PERF_COUNT_HW_INSTRUCTIONS, "instructions" }, + { PERF_COUNT_HW_BUS_CYCLES, "bus cycles" }, + { PERF_COUNT_HW_CACHE_REFERENCES, "cache references" }, + { PERF_COUNT_HW_CACHE_MISSES, "cache misses" }, + { PERF_COUNT_HW_BRANCH_INSTRUCTIONS, "branch instructions" }, + { PERF_COUNT_HW_BRANCH_MISSES, "branch misses" }, }; +static __init void intel_arch_events_quirk(void) +{ + int bit; + + /* disable event that reported as not presend by cpuid */ + for_each_set_bit(bit, x86_pmu.events_mask, ARRAY_SIZE(intel_arch_events_map)) { + intel_perfmon_event_map[intel_arch_events_map[bit].id] = 0; + printk(KERN_WARNING "CPUID marked event: \'%s\' unavailable\n", + intel_arch_events_map[bit].name); + } +} + +static __init void intel_nehalem_quirk(void) +{ + union cpuid10_ebx ebx; + + ebx.full = x86_pmu.events_maskl; + if (ebx.split.no_branch_misses_retired) { + /* + * Erratum AAJ80 detected, we work it around by using + * the BR_MISP_EXEC.ANY event. This will over-count + * branch-misses, but it's still much better than the + * architectural event which is often completely bogus: + */ + intel_perfmon_event_map[PERF_COUNT_HW_BRANCH_MISSES] = 0x7f89; + ebx.split.no_branch_misses_retired = 0; + x86_pmu.events_maskl = ebx.full; + printk(KERN_INFO "CPU erratum AAJ80 worked around\n"); + } +} + __init int intel_pmu_init(void) { union cpuid10_edx edx; union cpuid10_eax eax; union cpuid10_ebx ebx; unsigned int unused; - int version, bit; + int version; if (!cpu_has(&boot_cpu_data, X86_FEATURE_ARCH_PERFMON)) { switch (boot_cpu_data.x86) { @@ -1599,6 +1630,9 @@ __init int intel_pmu_init(void) x86_pmu.cntval_bits = eax.split.bit_width; x86_pmu.cntval_mask = (1ULL << eax.split.bit_width) - 1; + x86_pmu.events_maskl = ebx.full; + x86_pmu.events_mask_len = eax.split.mask_length; + /* * Quirk: v2 perfmon does not report fixed-purpose events, so * assume at least 3 events: @@ -1618,6 +1652,8 @@ __init int intel_pmu_init(void) intel_ds_init(); + x86_add_quirk(intel_arch_events_quirk); /* Install first, so it runs last */ + /* * Install the hw-cache-events table: */ @@ -1627,7 +1663,7 @@ __init int intel_pmu_init(void) break; case 15: /* original 65 nm celeron/pentium/core2/xeon, "Merom"/"Conroe" */ - x86_pmu.quirks = intel_clovertown_quirks; + x86_add_quirk(intel_clovertown_quirk); case 22: /* single-core 65 nm celeron/core2solo "Merom-L"/"Conroe-L" */ case 23: /* current 45 nm celeron/core2/xeon "Penryn"/"Wolfdale" */ case 29: /* six-core 45 nm xeon "Dunnington" */ @@ -1661,18 +1697,8 @@ __init int intel_pmu_init(void) /* UOPS_EXECUTED.CORE_ACTIVE_CYCLES,c=1,i=1 */ intel_perfmon_event_map[PERF_COUNT_HW_STALLED_CYCLES_BACKEND] = 0x1803fb1; - if (ebx.split.no_branch_misses_retired) { - /* - * Erratum AAJ80 detected, we work it around by using - * the BR_MISP_EXEC.ANY event. This will over-count - * branch-misses, but it's still much better than the - * architectural event which is often completely bogus: - */ - intel_perfmon_event_map[PERF_COUNT_HW_BRANCH_MISSES] = 0x7f89; - ebx.split.no_branch_misses_retired = 0; + x86_add_quirk(intel_nehalem_quirk); - pr_cont("erratum AAJ80 worked around, "); - } pr_cont("Nehalem events, "); break; @@ -1712,7 +1738,7 @@ __init int intel_pmu_init(void) break; case 42: /* SandyBridge */ - x86_pmu.quirks = intel_sandybridge_quirks; + x86_add_quirk(intel_sandybridge_quirk); case 45: /* SandyBridge, "Romely-EP" */ memcpy(hw_cache_event_ids, snb_hw_cache_event_ids, sizeof(hw_cache_event_ids)); @@ -1749,12 +1775,6 @@ __init int intel_pmu_init(void) break; } } - x86_pmu.events_maskl = ebx.full; - x86_pmu.events_mask_len = eax.split.mask_length; - - /* disable event that reported as not presend by cpuid */ - for_each_set_bit(bit, x86_pmu.events_mask, ARRAY_SIZE(intel_event_id_to_hw_id)) - intel_perfmon_event_map[intel_event_id_to_hw_id[bit]] = 0; return 0; } -- cgit v1.2.3 From b3d9468a8bd218a695e3a0ff112cd4efd27b670a Mon Sep 17 00:00:00 2001 From: Gleb Natapov Date: Thu, 10 Nov 2011 14:57:27 +0200 Subject: perf, x86: Expose perf capability to other modules KVM needs to know perf capability to decide which PMU it can expose to a guest. Signed-off-by: Gleb Natapov Signed-off-by: Peter Zijlstra Link: http://lkml.kernel.org/r/1320929850-10480-8-git-send-email-gleb@redhat.com Signed-off-by: Ingo Molnar --- arch/x86/include/asm/perf_event.h | 15 +++++++++++++++ arch/x86/kernel/cpu/perf_event.c | 12 ++++++++++++ 2 files changed, 27 insertions(+) diff --git a/arch/x86/include/asm/perf_event.h b/arch/x86/include/asm/perf_event.h index c6998bc75456..b50e9d15aae0 100644 --- a/arch/x86/include/asm/perf_event.h +++ b/arch/x86/include/asm/perf_event.h @@ -95,6 +95,15 @@ union cpuid10_edx { unsigned int full; }; +struct x86_pmu_capability { + int version; + int num_counters_gp; + int num_counters_fixed; + int bit_width_gp; + int bit_width_fixed; + unsigned int events_mask; + int events_mask_len; +}; /* * Fixed-purpose performance events: @@ -216,6 +225,7 @@ struct perf_guest_switch_msr { }; extern struct perf_guest_switch_msr *perf_guest_get_msrs(int *nr); +extern void perf_get_x86_pmu_capability(struct x86_pmu_capability *cap); #else static inline perf_guest_switch_msr *perf_guest_get_msrs(int *nr) { @@ -223,6 +233,11 @@ static inline perf_guest_switch_msr *perf_guest_get_msrs(int *nr) return NULL; } +static inline void perf_get_x86_pmu_capability(struct x86_pmu_capability *cap) +{ + memset(cap, 0, sizeof(*cap)); +} + static inline void perf_events_lapic_init(void) { } #endif diff --git a/arch/x86/kernel/cpu/perf_event.c b/arch/x86/kernel/cpu/perf_event.c index 55889e0b1452..930fe4879542 100644 --- a/arch/x86/kernel/cpu/perf_event.c +++ b/arch/x86/kernel/cpu/perf_event.c @@ -1696,3 +1696,15 @@ unsigned long perf_misc_flags(struct pt_regs *regs) return misc; } + +void perf_get_x86_pmu_capability(struct x86_pmu_capability *cap) +{ + cap->version = x86_pmu.version; + cap->num_counters_gp = x86_pmu.num_counters; + cap->num_counters_fixed = x86_pmu.num_counters_fixed; + cap->bit_width_gp = x86_pmu.cntval_bits; + cap->bit_width_fixed = x86_pmu.cntval_bits; + cap->events_mask = (unsigned int)x86_pmu.events_maskl; + cap->events_mask_len = x86_pmu.events_mask_len; +} +EXPORT_SYMBOL_GPL(perf_get_x86_pmu_capability); -- cgit v1.2.3 From f8c852031a383ac260ae37df7ad063d42d0ed271 Mon Sep 17 00:00:00 2001 From: Robert Richter Date: Wed, 7 Dec 2011 11:16:38 +0100 Subject: oprofile: Fix oprofile_timer_exit() breakage Removing remainings of oprofile_timer_exit() completly. Signed-off-by: Robert Richter --- arch/s390/oprofile/init.c | 1 - drivers/oprofile/oprof.h | 1 - 2 files changed, 2 deletions(-) diff --git a/arch/s390/oprofile/init.c b/arch/s390/oprofile/init.c index 6efc18b5e60a..113d7cbbc065 100644 --- a/arch/s390/oprofile/init.c +++ b/arch/s390/oprofile/init.c @@ -171,7 +171,6 @@ static int oprofile_hwsampler_init(struct oprofile_operations *ops) static void oprofile_hwsampler_exit(void) { - oprofile_timer_exit(); hwsampler_shutdown(); } diff --git a/drivers/oprofile/oprof.h b/drivers/oprofile/oprof.h index 769fb0fcac44..d32ef816337c 100644 --- a/drivers/oprofile/oprof.h +++ b/drivers/oprofile/oprof.h @@ -35,7 +35,6 @@ struct dentry; void oprofile_create_files(struct super_block *sb, struct dentry *root); int oprofile_timer_init(struct oprofile_operations *ops); -void oprofile_timer_exit(void); #ifdef CONFIG_OPROFILE_NMI_TIMER int op_nmi_timer_init(struct oprofile_operations *ops); #else -- cgit v1.2.3 From dd3c4670d7fafeb18bf7542fbcfd2606fb06a4a1 Mon Sep 17 00:00:00 2001 From: Andreas Krebbel Date: Fri, 25 Nov 2011 20:03:05 +0100 Subject: oprofile, s390: Add event interface to the System z hardware sampling module With this patch the OProfile Basic Mode Sampling support for System z is enhanced with a counter file system. That way hardware sampling can be configured using the user space tools with only little modifications. With the patch by default new cpu_types (s390/z10, s390/z196) are returned in order to indicate that we are running a CPU which provides the hardware sampling facility. Existing user space tools will complain about an unknown cpu type. In order to be compatible with existing user space tools the `cpu_type' module parameter has been added. Setting the parameter to `timer' will force the module to return `timer' as cpu_type. The module will still try to use hardware sampling if available and the hwsampling virtual filesystem will be also be available for configuration. So this has a different effect than using the generic oprofile module parameter `timer=1'. If the basic mode sampling is enabled on the machine and the cpu_type=timer parameter is not used the kernel module will provide the following virtual filesystem: /dev/oprofile/0/enabled /dev/oprofile/0/event /dev/oprofile/0/count /dev/oprofile/0/unit_mask /dev/oprofile/0/kernel /dev/oprofile/0/user In the counter file system only the values of 'enabled', 'count', 'kernel', and 'user' are evaluated by the kernel module. Everything else must contain fixed values. The 'event' value only supports a single event - HWSAMPLING with value 0. The 'count' value specifies the hardware sampling rate as it is passed to the CPU measurement facility. The 'kernel' and 'user' flags can now be used to filter for samples when using hardware sampling. Additionally also the following file will be created: /dev/oprofile/timer/enabled This will always be the inverted value of /dev/oprofile/0/enabled. 0 is not accepted without hardware sampling. Signed-off-by: Andreas Krebbel Signed-off-by: Robert Richter --- Documentation/kernel-parameters.txt | 2 + arch/s390/oprofile/hwsampler.c | 7 +- arch/s390/oprofile/init.c | 372 +++++++++++++++++++++++++++++++++--- arch/s390/oprofile/op_counter.h | 23 +++ 4 files changed, 374 insertions(+), 30 deletions(-) create mode 100644 arch/s390/oprofile/op_counter.h diff --git a/Documentation/kernel-parameters.txt b/Documentation/kernel-parameters.txt index f7735a125f4b..60b98cea06fc 100644 --- a/Documentation/kernel-parameters.txt +++ b/Documentation/kernel-parameters.txt @@ -1851,6 +1851,8 @@ bytes respectively. Such letter suffixes can also be entirely omitted. timer: [X86] Force use of architectural NMI timer mode (see also oprofile.timer for generic hr timer mode) + [s390] Force legacy basic mode sampling + (report cpu_type "timer") oops=panic Always panic on oopses. Default is to just kill the process, but there is a small probability of diff --git a/arch/s390/oprofile/hwsampler.c b/arch/s390/oprofile/hwsampler.c index 4552ce40c81a..cab0fe932c08 100644 --- a/arch/s390/oprofile/hwsampler.c +++ b/arch/s390/oprofile/hwsampler.c @@ -22,6 +22,7 @@ #include #include "hwsampler.h" +#include "op_counter.h" #define MAX_NUM_SDB 511 #define MIN_NUM_SDB 1 @@ -896,6 +897,8 @@ static void add_samples_to_oprofile(unsigned int cpu, unsigned long *sdbt, if (sample_data_ptr->P == 1) { /* userspace sample */ unsigned int pid = sample_data_ptr->prim_asn; + if (!counter_config.user) + goto skip_sample; rcu_read_lock(); tsk = pid_task(find_vpid(pid), PIDTYPE_PID); if (tsk) @@ -903,6 +906,8 @@ static void add_samples_to_oprofile(unsigned int cpu, unsigned long *sdbt, rcu_read_unlock(); } else { /* kernelspace sample */ + if (!counter_config.kernel) + goto skip_sample; regs = task_pt_regs(current); } @@ -910,7 +915,7 @@ static void add_samples_to_oprofile(unsigned int cpu, unsigned long *sdbt, oprofile_add_ext_hw_sample(sample_data_ptr->ia, regs, 0, !sample_data_ptr->P, tsk); mutex_unlock(&hws_sem); - + skip_sample: sample_data_ptr++; } } diff --git a/arch/s390/oprofile/init.c b/arch/s390/oprofile/init.c index 113d7cbbc065..6cf2286d0405 100644 --- a/arch/s390/oprofile/init.c +++ b/arch/s390/oprofile/init.c @@ -2,10 +2,11 @@ * arch/s390/oprofile/init.c * * S390 Version - * Copyright (C) 2003 IBM Deutschland Entwicklung GmbH, IBM Corporation + * Copyright (C) 2002-2011 IBM Deutschland Entwicklung GmbH, IBM Corporation * Author(s): Thomas Spatzier (tspat@de.ibm.com) * Author(s): Mahesh Salgaonkar (mahesh@linux.vnet.ibm.com) * Author(s): Heinz Graalfs (graalfs@linux.vnet.ibm.com) + * Author(s): Andreas Krebbel (krebbel@linux.vnet.ibm.com) * * @remark Copyright 2002-2011 OProfile authors */ @@ -14,6 +15,8 @@ #include #include #include +#include +#include #include "../../../drivers/oprofile/oprof.h" @@ -22,6 +25,7 @@ extern void s390_backtrace(struct pt_regs * const regs, unsigned int depth); #ifdef CONFIG_64BIT #include "hwsampler.h" +#include "op_counter.h" #define DEFAULT_INTERVAL 4127518 @@ -35,16 +39,41 @@ static unsigned long oprofile_max_interval; static unsigned long oprofile_sdbt_blocks = DEFAULT_SDBT_BLOCKS; static unsigned long oprofile_sdb_blocks = DEFAULT_SDB_BLOCKS; -static int hwsampler_file; +static int hwsampler_enabled; static int hwsampler_running; /* start_mutex must be held to change */ +static int hwsampler_available; static struct oprofile_operations timer_ops; +struct op_counter_config counter_config; + +enum __force_cpu_type { + reserved = 0, /* do not force */ + timer, +}; +static int force_cpu_type; + +static int set_cpu_type(const char *str, struct kernel_param *kp) +{ + if (!strcmp(str, "timer")) { + force_cpu_type = timer; + printk(KERN_INFO "oprofile: forcing timer to be returned " + "as cpu type\n"); + } else { + force_cpu_type = 0; + } + + return 0; +} +module_param_call(cpu_type, set_cpu_type, NULL, NULL, 0); +MODULE_PARM_DESC(cpu_type, "Force legacy basic mode sampling" + "(report cpu_type \"timer\""); + static int oprofile_hwsampler_start(void) { int retval; - hwsampler_running = hwsampler_file; + hwsampler_running = hwsampler_enabled; if (!hwsampler_running) return timer_ops.start(); @@ -72,10 +101,16 @@ static void oprofile_hwsampler_stop(void) return; } +/* + * File ops used for: + * /dev/oprofile/0/enabled + * /dev/oprofile/hwsampling/hwsampler (cpu_type = timer) + */ + static ssize_t hwsampler_read(struct file *file, char __user *buf, size_t count, loff_t *offset) { - return oprofilefs_ulong_to_user(hwsampler_file, buf, count, offset); + return oprofilefs_ulong_to_user(hwsampler_enabled, buf, count, offset); } static ssize_t hwsampler_write(struct file *file, char const __user *buf, @@ -91,6 +126,9 @@ static ssize_t hwsampler_write(struct file *file, char const __user *buf, if (retval) return retval; + if (val != 0 && val != 1) + return -EINVAL; + if (oprofile_started) /* * save to do without locking as we set @@ -99,7 +137,7 @@ static ssize_t hwsampler_write(struct file *file, char const __user *buf, */ return -EBUSY; - hwsampler_file = val; + hwsampler_enabled = val; return count; } @@ -109,38 +147,311 @@ static const struct file_operations hwsampler_fops = { .write = hwsampler_write, }; +/* + * File ops used for: + * /dev/oprofile/0/count + * /dev/oprofile/hwsampling/hw_interval (cpu_type = timer) + * + * Make sure that the value is within the hardware range. + */ + +static ssize_t hw_interval_read(struct file *file, char __user *buf, + size_t count, loff_t *offset) +{ + return oprofilefs_ulong_to_user(oprofile_hw_interval, buf, + count, offset); +} + +static ssize_t hw_interval_write(struct file *file, char const __user *buf, + size_t count, loff_t *offset) +{ + unsigned long val; + int retval; + + if (*offset) + return -EINVAL; + retval = oprofilefs_ulong_from_user(&val, buf, count); + if (retval) + return retval; + if (val < oprofile_min_interval) + oprofile_hw_interval = oprofile_min_interval; + else if (val > oprofile_max_interval) + oprofile_hw_interval = oprofile_max_interval; + else + oprofile_hw_interval = val; + + return count; +} + +static const struct file_operations hw_interval_fops = { + .read = hw_interval_read, + .write = hw_interval_write, +}; + +/* + * File ops used for: + * /dev/oprofile/0/event + * Only a single event with number 0 is supported with this counter. + * + * /dev/oprofile/0/unit_mask + * This is a dummy file needed by the user space tools. + * No value other than 0 is accepted or returned. + */ + +static ssize_t hwsampler_zero_read(struct file *file, char __user *buf, + size_t count, loff_t *offset) +{ + return oprofilefs_ulong_to_user(0, buf, count, offset); +} + +static ssize_t hwsampler_zero_write(struct file *file, char const __user *buf, + size_t count, loff_t *offset) +{ + unsigned long val; + int retval; + + if (*offset) + return -EINVAL; + + retval = oprofilefs_ulong_from_user(&val, buf, count); + if (retval) + return retval; + if (val != 0) + return -EINVAL; + return count; +} + +static const struct file_operations zero_fops = { + .read = hwsampler_zero_read, + .write = hwsampler_zero_write, +}; + +/* /dev/oprofile/0/kernel file ops. */ + +static ssize_t hwsampler_kernel_read(struct file *file, char __user *buf, + size_t count, loff_t *offset) +{ + return oprofilefs_ulong_to_user(counter_config.kernel, + buf, count, offset); +} + +static ssize_t hwsampler_kernel_write(struct file *file, char const __user *buf, + size_t count, loff_t *offset) +{ + unsigned long val; + int retval; + + if (*offset) + return -EINVAL; + + retval = oprofilefs_ulong_from_user(&val, buf, count); + if (retval) + return retval; + + if (val != 0 && val != 1) + return -EINVAL; + + counter_config.kernel = val; + + return count; +} + +static const struct file_operations kernel_fops = { + .read = hwsampler_kernel_read, + .write = hwsampler_kernel_write, +}; + +/* /dev/oprofile/0/user file ops. */ + +static ssize_t hwsampler_user_read(struct file *file, char __user *buf, + size_t count, loff_t *offset) +{ + return oprofilefs_ulong_to_user(counter_config.user, + buf, count, offset); +} + +static ssize_t hwsampler_user_write(struct file *file, char const __user *buf, + size_t count, loff_t *offset) +{ + unsigned long val; + int retval; + + if (*offset) + return -EINVAL; + + retval = oprofilefs_ulong_from_user(&val, buf, count); + if (retval) + return retval; + + if (val != 0 && val != 1) + return -EINVAL; + + counter_config.user = val; + + return count; +} + +static const struct file_operations user_fops = { + .read = hwsampler_user_read, + .write = hwsampler_user_write, +}; + + +/* + * File ops used for: /dev/oprofile/timer/enabled + * The value always has to be the inverted value of hwsampler_enabled. So + * no separate variable is created. That way we do not need locking. + */ + +static ssize_t timer_enabled_read(struct file *file, char __user *buf, + size_t count, loff_t *offset) +{ + return oprofilefs_ulong_to_user(!hwsampler_enabled, buf, count, offset); +} + +static ssize_t timer_enabled_write(struct file *file, char const __user *buf, + size_t count, loff_t *offset) +{ + unsigned long val; + int retval; + + if (*offset) + return -EINVAL; + + retval = oprofilefs_ulong_from_user(&val, buf, count); + if (retval) + return retval; + + if (val != 0 && val != 1) + return -EINVAL; + + /* Timer cannot be disabled without having hardware sampling. */ + if (val == 0 && !hwsampler_available) + return -EINVAL; + + if (oprofile_started) + /* + * save to do without locking as we set + * hwsampler_running in start() when start_mutex is + * held + */ + return -EBUSY; + + hwsampler_enabled = !val; + + return count; +} + +static const struct file_operations timer_enabled_fops = { + .read = timer_enabled_read, + .write = timer_enabled_write, +}; + + static int oprofile_create_hwsampling_files(struct super_block *sb, - struct dentry *root) + struct dentry *root) { - struct dentry *hw_dir; + struct dentry *dir; + + dir = oprofilefs_mkdir(sb, root, "timer"); + if (!dir) + return -EINVAL; + + oprofilefs_create_file(sb, dir, "enabled", &timer_enabled_fops); + + if (!hwsampler_available) + return 0; /* reinitialize default values */ - hwsampler_file = 1; + hwsampler_enabled = 1; + counter_config.kernel = 1; + counter_config.user = 1; - hw_dir = oprofilefs_mkdir(sb, root, "hwsampling"); - if (!hw_dir) - return -EINVAL; + if (!force_cpu_type) { + /* + * Create the counter file system. A single virtual + * counter is created which can be used to + * enable/disable hardware sampling dynamically from + * user space. The user space will configure a single + * counter with a single event. The value of 'event' + * and 'unit_mask' are not evaluated by the kernel code + * and can only be set to 0. + */ + + dir = oprofilefs_mkdir(sb, root, "0"); + if (!dir) + return -EINVAL; - oprofilefs_create_file(sb, hw_dir, "hwsampler", &hwsampler_fops); - oprofilefs_create_ulong(sb, hw_dir, "hw_interval", - &oprofile_hw_interval); - oprofilefs_create_ro_ulong(sb, hw_dir, "hw_min_interval", - &oprofile_min_interval); - oprofilefs_create_ro_ulong(sb, hw_dir, "hw_max_interval", - &oprofile_max_interval); - oprofilefs_create_ulong(sb, hw_dir, "hw_sdbt_blocks", - &oprofile_sdbt_blocks); + oprofilefs_create_file(sb, dir, "enabled", &hwsampler_fops); + oprofilefs_create_file(sb, dir, "event", &zero_fops); + oprofilefs_create_file(sb, dir, "count", &hw_interval_fops); + oprofilefs_create_file(sb, dir, "unit_mask", &zero_fops); + oprofilefs_create_file(sb, dir, "kernel", &kernel_fops); + oprofilefs_create_file(sb, dir, "user", &user_fops); + oprofilefs_create_ulong(sb, dir, "hw_sdbt_blocks", + &oprofile_sdbt_blocks); + } else { + /* + * Hardware sampling can be used but the cpu_type is + * forced to timer in order to deal with legacy user + * space tools. The /dev/oprofile/hwsampling fs is + * provided in that case. + */ + dir = oprofilefs_mkdir(sb, root, "hwsampling"); + if (!dir) + return -EINVAL; + + oprofilefs_create_file(sb, dir, "hwsampler", + &hwsampler_fops); + oprofilefs_create_file(sb, dir, "hw_interval", + &hw_interval_fops); + oprofilefs_create_ro_ulong(sb, dir, "hw_min_interval", + &oprofile_min_interval); + oprofilefs_create_ro_ulong(sb, dir, "hw_max_interval", + &oprofile_max_interval); + oprofilefs_create_ulong(sb, dir, "hw_sdbt_blocks", + &oprofile_sdbt_blocks); + } return 0; } static int oprofile_hwsampler_init(struct oprofile_operations *ops) { + /* + * Initialize the timer mode infrastructure as well in order + * to be able to switch back dynamically. oprofile_timer_init + * is not supposed to fail. + */ + if (oprofile_timer_init(ops)) + BUG(); + + memcpy(&timer_ops, ops, sizeof(timer_ops)); + ops->create_files = oprofile_create_hwsampling_files; + + /* + * If the user space tools do not support newer cpu types, + * the force_cpu_type module parameter + * can be used to always return \"timer\" as cpu type. + */ + if (force_cpu_type != timer) { + struct cpuid id; + + get_cpu_id (&id); + + switch (id.machine) { + case 0x2097: case 0x2098: ops->cpu_type = "s390/z10"; break; + case 0x2817: case 0x2818: ops->cpu_type = "s390/z196"; break; + default: return -ENODEV; + } + } + if (hwsampler_setup()) return -ENODEV; /* - * create hwsampler files only if hwsampler_setup() succeeds. + * Query the range for the sampling interval from the + * hardware. */ oprofile_min_interval = hwsampler_query_min_interval(); if (oprofile_min_interval == 0) @@ -155,16 +466,11 @@ static int oprofile_hwsampler_init(struct oprofile_operations *ops) if (oprofile_hw_interval > oprofile_max_interval) oprofile_hw_interval = oprofile_max_interval; - if (oprofile_timer_init(ops)) - return -ENODEV; - - printk(KERN_INFO "oprofile: using hardware sampling\n"); - - memcpy(&timer_ops, ops, sizeof(timer_ops)); + printk(KERN_INFO "oprofile: System z hardware sampling " + "facility found.\n"); ops->start = oprofile_hwsampler_start; ops->stop = oprofile_hwsampler_stop; - ops->create_files = oprofile_create_hwsampling_files; return 0; } @@ -181,7 +487,15 @@ int __init oprofile_arch_init(struct oprofile_operations *ops) ops->backtrace = s390_backtrace; #ifdef CONFIG_64BIT - return oprofile_hwsampler_init(ops); + + /* + * -ENODEV is not reported to the caller. The module itself + * will use the timer mode sampling as fallback and this is + * always available. + */ + hwsampler_available = oprofile_hwsampler_init(ops) == 0; + + return 0; #else return -ENODEV; #endif diff --git a/arch/s390/oprofile/op_counter.h b/arch/s390/oprofile/op_counter.h new file mode 100644 index 000000000000..1a8d3ca09014 --- /dev/null +++ b/arch/s390/oprofile/op_counter.h @@ -0,0 +1,23 @@ +/** + * arch/s390/oprofile/op_counter.h + * + * Copyright (C) 2011 IBM Deutschland Entwicklung GmbH, IBM Corporation + * Author(s): Andreas Krebbel (krebbel@linux.vnet.ibm.com) + * + * @remark Copyright 2011 OProfile authors + */ + +#ifndef OP_COUNTER_H +#define OP_COUNTER_H + +struct op_counter_config { + /* `enabled' maps to the hwsampler_file variable. */ + /* `count' maps to the oprofile_hw_interval variable. */ + /* `event' and `unit_mask' are unused. */ + unsigned long kernel; + unsigned long user; +}; + +extern struct op_counter_config counter_config; + +#endif /* OP_COUNTER_H */ -- cgit v1.2.3 From 38efb539c13f8f173e381435cdd40463ab5d38de Mon Sep 17 00:00:00 2001 From: Robert Richter Date: Fri, 25 Nov 2011 11:38:40 +0100 Subject: perf script: Fix mem leaks and NULL pointer checks around strdup()s Fix mem leaks and missing NULL pointer checks after strdup(). And get_script_path() did not free __script_root in case of continue. Introduce a helper function get_script_root(). Cc: Ingo Molnar Link: http://lkml.kernel.org/r/1322217520-3287-1-git-send-email-robert.richter@amd.com Signed-off-by: Robert Richter Signed-off-by: Arnaldo Carvalho de Melo --- tools/perf/builtin-script.c | 48 +++++++++++++++++++++++++++------------------ 1 file changed, 29 insertions(+), 19 deletions(-) diff --git a/tools/perf/builtin-script.c b/tools/perf/builtin-script.c index 619d6dcaa1d9..ccbfd56d7549 100644 --- a/tools/perf/builtin-script.c +++ b/tools/perf/builtin-script.c @@ -693,7 +693,8 @@ static int parse_output_fields(const struct option *opt __used, type = PERF_TYPE_RAW; else { fprintf(stderr, "Invalid event type in field string.\n"); - return -EINVAL; + rc = -EINVAL; + goto out; } if (output[type].user_set) @@ -935,6 +936,24 @@ static int read_script_info(struct script_desc *desc, const char *filename) return 0; } +static char *get_script_root(struct dirent *script_dirent, const char *suffix) +{ + char *script_root, *str; + + script_root = strdup(script_dirent->d_name); + if (!script_root) + return NULL; + + str = (char *)ends_with(script_root, suffix); + if (!str) { + free(script_root); + return NULL; + } + + *str = '\0'; + return script_root; +} + static int list_available_scripts(const struct option *opt __used, const char *s __used, int unset __used) { @@ -946,7 +965,6 @@ static int list_available_scripts(const struct option *opt __used, struct script_desc *desc; char first_half[BUFSIZ]; char *script_root; - char *str; snprintf(scripts_path, MAXPATHLEN, "%s/scripts", perf_exec_path()); @@ -962,16 +980,14 @@ static int list_available_scripts(const struct option *opt __used, continue; for_each_script(lang_path, lang_dir, script_dirent, script_next) { - script_root = strdup(script_dirent.d_name); - str = (char *)ends_with(script_root, REPORT_SUFFIX); - if (str) { - *str = '\0'; + script_root = get_script_root(&script_dirent, REPORT_SUFFIX); + if (script_root) { desc = script_desc__findnew(script_root); snprintf(script_path, MAXPATHLEN, "%s/%s", lang_path, script_dirent.d_name); read_script_info(desc, script_path); + free(script_root); } - free(script_root); } } @@ -993,8 +1009,7 @@ static char *get_script_path(const char *script_root, const char *suffix) char script_path[MAXPATHLEN]; DIR *scripts_dir, *lang_dir; char lang_path[MAXPATHLEN]; - char *str, *__script_root; - char *path = NULL; + char *__script_root; snprintf(scripts_path, MAXPATHLEN, "%s/scripts", perf_exec_path()); @@ -1010,23 +1025,18 @@ static char *get_script_path(const char *script_root, const char *suffix) continue; for_each_script(lang_path, lang_dir, script_dirent, script_next) { - __script_root = strdup(script_dirent.d_name); - str = (char *)ends_with(__script_root, suffix); - if (str) { - *str = '\0'; - if (strcmp(__script_root, script_root)) - continue; + __script_root = get_script_root(&script_dirent, suffix); + if (__script_root && !strcmp(script_root, __script_root)) { + free(__script_root); snprintf(script_path, MAXPATHLEN, "%s/%s", lang_path, script_dirent.d_name); - path = strdup(script_path); - free(__script_root); - break; + return strdup(script_path); } free(__script_root); } } - return path; + return NULL; } static bool is_top_script(const char *script_path) -- cgit v1.2.3 From 317df650c588bb9091b1fa0b5d726fe485aad88e Mon Sep 17 00:00:00 2001 From: Robert Richter Date: Fri, 25 Nov 2011 15:05:25 +0100 Subject: perf script: Implement option for system-wide profiling The option is documented in man perf-script but was not yet implemented: -a Force system-wide collection. Scripts run without a normally use -a by default, while scripts run with a normally don't - this option allows the latter to be run in system-wide mode. As with perf record you now can profile in system-wide mode for the runtime of a given command, e.g.: # perf script -a syscall-counts sleep 2 Cc: Ingo Molnar Link: http://lkml.kernel.org/r/1322229925-10075-1-git-send-email-robert.richter@amd.com Signed-off-by: Robert Richter Signed-off-by: Arnaldo Carvalho de Melo --- tools/perf/builtin-script.c | 15 ++++++++++----- 1 file changed, 10 insertions(+), 5 deletions(-) diff --git a/tools/perf/builtin-script.c b/tools/perf/builtin-script.c index ccbfd56d7549..ea71c5e1a94f 100644 --- a/tools/perf/builtin-script.c +++ b/tools/perf/builtin-script.c @@ -24,6 +24,7 @@ static u64 nr_unordered; extern const struct option record_options[]; static bool no_callchain; static bool show_full_info; +static bool system_wide; static const char *cpu_list; static DECLARE_BITMAP(cpu_bitmap, MAX_NR_CPUS); @@ -1105,6 +1106,8 @@ static const struct option options[] = { OPT_CALLBACK('f', "fields", NULL, "str", "comma separated output fields prepend with 'type:'. Valid types: hw,sw,trace,raw. Fields: comm,tid,pid,time,cpu,event,trace,ip,sym,dso,addr", parse_output_fields), + OPT_BOOLEAN('a', "all-cpus", &system_wide, + "system-wide collection from all CPUs"), OPT_STRING('C', "cpu", &cpu_list, "cpu", "list of cpus to profile"), OPT_STRING('c', "comms", &symbol_conf.comm_list_str, "comm[,comm...]", "only display events for these comms"), @@ -1134,7 +1137,6 @@ int cmd_script(int argc, const char **argv, const char *prefix __used) struct perf_session *session; char *script_path = NULL; const char **__argv; - bool system_wide; int i, j, err; setup_scripting(); @@ -1202,15 +1204,17 @@ int cmd_script(int argc, const char **argv, const char *prefix __used) } if (!pid) { - system_wide = true; j = 0; dup2(live_pipe[1], 1); close(live_pipe[0]); - if (!is_top_script(argv[0])) + if (is_top_script(argv[0])) { + system_wide = true; + } else if (!system_wide) { system_wide = !have_cmd(argc - rep_args, &argv[rep_args]); + } __argv = malloc((argc + 6) * sizeof(const char *)); if (!__argv) @@ -1258,10 +1262,11 @@ int cmd_script(int argc, const char **argv, const char *prefix __used) script_path = rep_script_path; if (script_path) { - system_wide = false; j = 0; - if (rec_script_path) + if (!rec_script_path) + system_wide = false; + else if (!system_wide) system_wide = !have_cmd(argc - 1, &argv[1]); __argv = malloc((argc + 2) * sizeof(const char *)); -- cgit v1.2.3 From 74eec26facadbe6dbc0621bc862892c915c4534f Mon Sep 17 00:00:00 2001 From: Andrew Vagin Date: Mon, 28 Nov 2011 12:03:31 +0300 Subject: perf tools: Add ability to synthesize event according to a sample It's the counterpart of perf_session__parse_sample. v2: fixed mistakes found by David Ahern. v3: s/data/sample/ s/perf_event__change_sample/perf_event__synthesize_sample Reviewed-by: David Ahern Cc: Arun Sharma Cc: David Ahern Cc: Ingo Molnar Cc: Paul Mackerras Cc: Peter Zijlstra Cc: devel@openvz.org Link: http://lkml.kernel.org/r/1323266161-394927-3-git-send-email-avagin@openvz.org Signed-off-by: Andrew Vagin Signed-off-by: Arnaldo Carvalho de Melo --- tools/perf/util/event.h | 3 ++ tools/perf/util/evsel.c | 79 +++++++++++++++++++++++++++++++++++++++++++++++ tools/perf/util/session.h | 8 +++++ 3 files changed, 90 insertions(+) diff --git a/tools/perf/util/event.h b/tools/perf/util/event.h index 0d80201ce844..cbdeaad9c5e5 100644 --- a/tools/perf/util/event.h +++ b/tools/perf/util/event.h @@ -199,6 +199,9 @@ const char *perf_event__name(unsigned int id); int perf_event__parse_sample(const union perf_event *event, u64 type, int sample_size, bool sample_id_all, struct perf_sample *sample, bool swapped); +int perf_event__synthesize_sample(union perf_event *event, u64 type, + const struct perf_sample *sample, + bool swapped); size_t perf_event__fprintf_comm(union perf_event *event, FILE *fp); size_t perf_event__fprintf_mmap(union perf_event *event, FILE *fp); diff --git a/tools/perf/util/evsel.c b/tools/perf/util/evsel.c index ee68d6944e61..4a8c8b02e9cc 100644 --- a/tools/perf/util/evsel.c +++ b/tools/perf/util/evsel.c @@ -574,3 +574,82 @@ int perf_event__parse_sample(const union perf_event *event, u64 type, return 0; } + +int perf_event__synthesize_sample(union perf_event *event, u64 type, + const struct perf_sample *sample, + bool swapped) +{ + u64 *array; + + /* + * used for cross-endian analysis. See git commit 65014ab3 + * for why this goofiness is needed. + */ + union { + u64 val64; + u32 val32[2]; + } u; + + array = event->sample.array; + + if (type & PERF_SAMPLE_IP) { + event->ip.ip = sample->ip; + array++; + } + + if (type & PERF_SAMPLE_TID) { + u.val32[0] = sample->pid; + u.val32[1] = sample->tid; + if (swapped) { + /* + * Inverse of what is done in perf_event__parse_sample + */ + u.val32[0] = bswap_32(u.val32[0]); + u.val32[1] = bswap_32(u.val32[1]); + u.val64 = bswap_64(u.val64); + } + + *array = u.val64; + array++; + } + + if (type & PERF_SAMPLE_TIME) { + *array = sample->time; + array++; + } + + if (type & PERF_SAMPLE_ADDR) { + *array = sample->addr; + array++; + } + + if (type & PERF_SAMPLE_ID) { + *array = sample->id; + array++; + } + + if (type & PERF_SAMPLE_STREAM_ID) { + *array = sample->stream_id; + array++; + } + + if (type & PERF_SAMPLE_CPU) { + u.val32[0] = sample->cpu; + if (swapped) { + /* + * Inverse of what is done in perf_event__parse_sample + */ + u.val32[0] = bswap_32(u.val32[0]); + u.val64 = bswap_64(u.val64); + } + *array = u.val64; + array++; + } + + if (type & PERF_SAMPLE_PERIOD) { + *array = sample->period; + array++; + } + + return 0; +} diff --git a/tools/perf/util/session.h b/tools/perf/util/session.h index 30e9c6b6fc3c..fb696124ad61 100644 --- a/tools/perf/util/session.h +++ b/tools/perf/util/session.h @@ -134,6 +134,14 @@ static inline int perf_session__parse_sample(struct perf_session *session, session->header.needs_swap); } +static inline int perf_session__synthesize_sample(struct perf_session *session, + union perf_event *event, + const struct perf_sample *sample) +{ + return perf_event__synthesize_sample(event, session->sample_type, + sample, session->header.needs_swap); +} + struct perf_evsel *perf_session__find_first_evtype(struct perf_session *session, unsigned int type); -- cgit v1.2.3 From 3e76ac78b08479e84a3eca3fb1b3066fb8230461 Mon Sep 17 00:00:00 2001 From: Andrew Vagin Date: Tue, 20 Dec 2011 17:32:45 +0300 Subject: perf record: Add ability to record event period The problem is that when SAMPLE_PERIOD is not set, the kernel generates a number of samples in proportion to an event's period. Number of these samples may be too big and the kernel throttles all samples above a defined limit. E.g.: I want to trace when a process sleeps. I created a process which sleeps for 1ms and for 4ms. perf got 100 events in both cases. swapper 0 [000] 1141.371830: sched_stat_sleep: comm=foo pid=1801 delay=1386750 [ns] swapper 0 [000] 1141.369444: sched_stat_sleep: comm=foo pid=1801 delay=4499585 [ns] In the first case a kernel want to send 4499585 events and in the second case it wants to send 1386750 events. perf-reports shows that process sleeps in both places equal time. Instead of this we can get only one sample with an attribute period. As result we have less data transferring between kernel and user-space and we avoid throttling of samples. The patch "events: Don't divide events if it has field period" added a kernel part of this functionality. Acked-by: Arun Sharma Cc: Arun Sharma Cc: David Ahern Cc: Ingo Molnar Cc: Paul Mackerras Cc: Peter Zijlstra Cc: devel@openvz.org Link: http://lkml.kernel.org/r/1324391565-1369947-1-git-send-email-avagin@openvz.org Signed-off-by: Andrew Vagin Signed-off-by: Arnaldo Carvalho de Melo --- tools/perf/builtin-record.c | 1 + tools/perf/perf.h | 1 + tools/perf/util/evsel.c | 3 +++ 3 files changed, 5 insertions(+) diff --git a/tools/perf/builtin-record.c b/tools/perf/builtin-record.c index 766fa0a91a32..f8fd14fb62ec 100644 --- a/tools/perf/builtin-record.c +++ b/tools/perf/builtin-record.c @@ -700,6 +700,7 @@ const struct option record_options[] = { OPT_BOOLEAN('d', "data", &record.opts.sample_address, "Sample addresses"), OPT_BOOLEAN('T', "timestamp", &record.opts.sample_time, "Sample timestamps"), + OPT_BOOLEAN('P', "period", &record.opts.period, "Sample period"), OPT_BOOLEAN('n', "no-samples", &record.opts.no_samples, "don't sample"), OPT_BOOLEAN('N', "no-buildid-cache", &record.no_buildid_cache, diff --git a/tools/perf/perf.h b/tools/perf/perf.h index ea804f5a8cc2..64f8bee31ced 100644 --- a/tools/perf/perf.h +++ b/tools/perf/perf.h @@ -200,6 +200,7 @@ struct perf_record_opts { bool sample_time; bool sample_id_all_avail; bool system_wide; + bool period; unsigned int freq; unsigned int mmap_pages; unsigned int user_freq; diff --git a/tools/perf/util/evsel.c b/tools/perf/util/evsel.c index 4a8c8b02e9cc..60ad0286759e 100644 --- a/tools/perf/util/evsel.c +++ b/tools/perf/util/evsel.c @@ -108,6 +108,9 @@ void perf_evsel__config(struct perf_evsel *evsel, struct perf_record_opts *opts) if (opts->system_wide) attr->sample_type |= PERF_SAMPLE_CPU; + if (opts->period) + attr->sample_type |= PERF_SAMPLE_PERIOD; + if (opts->sample_id_all_avail && (opts->sample_time || opts->system_wide || !opts->no_inherit || opts->cpu_list)) -- cgit v1.2.3 From f3bda2c9a689b38c059f7cb2d761ff58a2996370 Mon Sep 17 00:00:00 2001 From: Robert Richter Date: Thu, 15 Dec 2011 17:32:39 +0100 Subject: perf evsel: Fix uninitialized memory access to struct perf_sample Memory in struct perf_sample is not fully initialized during parsing. Depending on sampling data some parts may left unchanged. Zero out struct perf_sample first to avoid access to uninitialized memory. Cc: Ingo Molnar Link: http://lkml.kernel.org/r/1323966762-8574-2-git-send-email-robert.richter@amd.com Signed-off-by: Robert Richter Signed-off-by: Arnaldo Carvalho de Melo --- tools/perf/util/evsel.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) diff --git a/tools/perf/util/evsel.c b/tools/perf/util/evsel.c index 60ad0286759e..667f3b78bb2c 100644 --- a/tools/perf/util/evsel.c +++ b/tools/perf/util/evsel.c @@ -460,7 +460,7 @@ int perf_event__parse_sample(const union perf_event *event, u64 type, u32 val32[2]; } u; - + memset(data, 0, sizeof(*data)); data->cpu = data->pid = data->tid = -1; data->stream_id = data->id = data->time = -1ULL; -- cgit v1.2.3 From 6581f6e35f7d0338f699fce660adb48e863f2b59 Mon Sep 17 00:00:00 2001 From: Namhyung Kim Date: Tue, 13 Dec 2011 00:16:50 +0900 Subject: perf report: Document '--call-graph' for optional print_limit argument The '--call-graph' command line option can receive undocumented optional print_limit argument. Besides, use strtoul() to parse the option since its type is u32. Cc: Ingo Molnar Cc: Paul Mackerras Cc: Peter Zijlstra Link: http://lkml.kernel.org/r/1323703017-6060-2-git-send-email-namhyung@gmail.com Signed-off-by: Namhyung Kim Signed-off-by: Arnaldo Carvalho de Melo --- tools/perf/Documentation/perf-report.txt | 5 +++-- tools/perf/builtin-report.c | 6 +++--- 2 files changed, 6 insertions(+), 5 deletions(-) diff --git a/tools/perf/Documentation/perf-report.txt b/tools/perf/Documentation/perf-report.txt index dc85392a5ac7..35af0dc8ccb4 100644 --- a/tools/perf/Documentation/perf-report.txt +++ b/tools/perf/Documentation/perf-report.txt @@ -80,9 +80,10 @@ OPTIONS --dump-raw-trace:: Dump raw trace in ASCII. --g [type,min,order]:: +-g [type,min[,limit],order]:: --call-graph:: - Display call chains using type, min percent threshold and order. + Display call chains using type, min percent threshold, optional print + limit and order. type can be either: - flat: single column, linear exposure of call chains. - graph: use a graph tree, displaying absolute overhead rates. diff --git a/tools/perf/builtin-report.c b/tools/perf/builtin-report.c index ece7c5d3f504..b2654c9fb5c6 100644 --- a/tools/perf/builtin-report.c +++ b/tools/perf/builtin-report.c @@ -407,7 +407,7 @@ parse_callchain_opt(const struct option *opt, const char *arg, int unset) goto setup; if (tok2[0] != 'c') { - callchain_param.print_limit = strtod(tok2, &endptr); + callchain_param.print_limit = strtoul(tok2, &endptr, 0); tok2 = strtok(NULL, ","); if (!tok2) goto setup; @@ -485,8 +485,8 @@ int cmd_report(int argc, const char **argv, const char *prefix __used) "regex filter to identify parent, see: '--sort parent'"), OPT_BOOLEAN('x', "exclude-other", &symbol_conf.exclude_other, "Only display entries with parent-match"), - OPT_CALLBACK_DEFAULT('g', "call-graph", &report, "output_type,min_percent, call_order", - "Display callchains using output_type (graph, flat, fractal, or none) , min percent threshold and callchain order. " + OPT_CALLBACK_DEFAULT('g', "call-graph", &report, "output_type,min_percent[,print_limit],call_order", + "Display callchains using output_type (graph, flat, fractal, or none) , min percent threshold, optional print limit and callchain order. " "Default: fractal,0.5,callee", &parse_callchain_opt, callchain_default_opt), OPT_BOOLEAN('G', "inverted", &report.inverted_callchain, "alias for inverted call graph"), -- cgit v1.2.3 From 301b195db179241da8be25f345f3c4e64960f1d5 Mon Sep 17 00:00:00 2001 From: Nelson Elhage Date: Mon, 19 Dec 2011 08:39:30 -0500 Subject: perf evlist: Fix errno value reporting on failed mmap On failure, perf_evlist__mmap_per_{cpu,thread} will try to munmap() every map that doesn't have a NULL base. This will fail with EINVAL if one of them has base == MAP_FAILED, clobbering errno, so that perf_evlist__map will return EINVAL on any failure regardless of the root cause. Fix this by resetting failed maps to a NULL base. Acked-by: Namhyung Kim Cc: Ingo Molnar Cc: Paul Mackerras Cc: Peter Zijlstra Link: http://lkml.kernel.org/r/1324301972-22740-2-git-send-email-nelhage@nelhage.com Signed-off-by: Nelson Elhage Signed-off-by: Arnaldo Carvalho de Melo --- tools/perf/util/evlist.c | 4 +++- 1 file changed, 3 insertions(+), 1 deletion(-) diff --git a/tools/perf/util/evlist.c b/tools/perf/util/evlist.c index 8b19e7a1e881..963d63dde457 100644 --- a/tools/perf/util/evlist.c +++ b/tools/perf/util/evlist.c @@ -447,8 +447,10 @@ static int __perf_evlist__mmap(struct perf_evlist *evlist, evlist->mmap[idx].mask = mask; evlist->mmap[idx].base = mmap(NULL, evlist->mmap_len, prot, MAP_SHARED, fd, 0); - if (evlist->mmap[idx].base == MAP_FAILED) + if (evlist->mmap[idx].base == MAP_FAILED) { + evlist->mmap[idx].base = NULL; return -1; + } perf_evlist__add_pollfd(evlist, fd); return 0; -- cgit v1.2.3 From 2b600f9578852d12af59420011e3dadfaa58b043 Mon Sep 17 00:00:00 2001 From: Namhyung Kim Date: Tue, 13 Dec 2011 00:16:51 +0900 Subject: perf symbols: Get rid of duplicated snprintf() The 'path' variable is set on a upper line, don't need to do it again. Cc: Ingo Molnar Cc: Paul Mackerras Cc: Peter Zijlstra Link: http://lkml.kernel.org/r/1323703017-6060-3-git-send-email-namhyung@gmail.com Signed-off-by: Namhyung Kim Signed-off-by: Arnaldo Carvalho de Melo --- tools/perf/util/symbol.c | 7 +------ 1 file changed, 1 insertion(+), 6 deletions(-) diff --git a/tools/perf/util/symbol.c b/tools/perf/util/symbol.c index 632b50c7bc26..e54b13d4c357 100644 --- a/tools/perf/util/symbol.c +++ b/tools/perf/util/symbol.c @@ -1757,7 +1757,7 @@ static int map_groups__set_modules_path_dir(struct map_groups *mg, struct stat st; /*sshfs might return bad dent->d_type, so we have to stat*/ - sprintf(path, "%s/%s", dir_name, dent->d_name); + snprintf(path, sizeof(path), "%s/%s", dir_name, dent->d_name); if (stat(path, &st)) continue; @@ -1766,8 +1766,6 @@ static int map_groups__set_modules_path_dir(struct map_groups *mg, !strcmp(dent->d_name, "..")) continue; - snprintf(path, sizeof(path), "%s/%s", - dir_name, dent->d_name); ret = map_groups__set_modules_path_dir(mg, path); if (ret < 0) goto out; @@ -1788,9 +1786,6 @@ static int map_groups__set_modules_path_dir(struct map_groups *mg, if (map == NULL) continue; - snprintf(path, sizeof(path), "%s/%s", - dir_name, dent->d_name); - long_name = strdup(path); if (long_name == NULL) { ret = -1; -- cgit v1.2.3 From d74c896b7e3250a07f7d0315eecdd2ae1a7bc3c3 Mon Sep 17 00:00:00 2001 From: Namhyung Kim Date: Tue, 13 Dec 2011 00:16:52 +0900 Subject: perf symbols: Fix error path on symbol__init() The order of freeing comm_list and dso_list should be reversed. Cc: Ingo Molnar Cc: Paul Mackerras Cc: Peter Zijlstra Link: http://lkml.kernel.org/r/1323703017-6060-4-git-send-email-namhyung@gmail.com Signed-off-by: Namhyung Kim Signed-off-by: Arnaldo Carvalho de Melo --- tools/perf/util/symbol.c | 4 ++-- 1 file changed, 2 insertions(+), 2 deletions(-) diff --git a/tools/perf/util/symbol.c b/tools/perf/util/symbol.c index e54b13d4c357..215d50f2042e 100644 --- a/tools/perf/util/symbol.c +++ b/tools/perf/util/symbol.c @@ -2604,10 +2604,10 @@ int symbol__init(void) symbol_conf.initialized = true; return 0; -out_free_dso_list: - strlist__delete(symbol_conf.dso_list); out_free_comm_list: strlist__delete(symbol_conf.comm_list); +out_free_dso_list: + strlist__delete(symbol_conf.dso_list); return -1; } -- cgit v1.2.3 From 0161d82e9b740caa90f508138d1ae1b9d981b6d3 Mon Sep 17 00:00:00 2001 From: Namhyung Kim Date: Tue, 13 Dec 2011 00:16:53 +0900 Subject: perf tools: Fix a memory leak on perf_read_values_destroy After freeing each elements of the @values->value, we should free itself too. Cc: Ingo Molnar Cc: Paul Mackerras Cc: Peter Zijlstra Link: http://lkml.kernel.org/r/1323703017-6060-5-git-send-email-namhyung@gmail.com Signed-off-by: Namhyung Kim Signed-off-by: Arnaldo Carvalho de Melo --- tools/perf/util/values.c | 1 + 1 file changed, 1 insertion(+) diff --git a/tools/perf/util/values.c b/tools/perf/util/values.c index bdd33470b235..697c8b4e59cc 100644 --- a/tools/perf/util/values.c +++ b/tools/perf/util/values.c @@ -32,6 +32,7 @@ void perf_read_values_destroy(struct perf_read_values *values) for (i = 0; i < values->threads; i++) free(values->value[i]); + free(values->value); free(values->pid); free(values->tid); free(values->counterrawid); -- cgit v1.2.3 From 5f9273d64a5ccbd3c2b4446cc8b71123ed5d6366 Mon Sep 17 00:00:00 2001 From: Namhyung Kim Date: Tue, 13 Dec 2011 22:52:03 +0900 Subject: perf tools: Remove stale git headlines from top comment These files are part of PERF not GIT although they're come from there :) Cc: Ingo Molnar Cc: Johannes Schindelin Cc: Linus Torvalds Cc: Paul Mackerras Cc: Peter Zijlstra Link: http://lkml.kernel.org/r/1323784323-2150-1-git-send-email-namhyung@gmail.com Signed-off-by: Namhyung Kim Signed-off-by: Arnaldo Carvalho de Melo --- tools/perf/util/config.c | 5 ++++- tools/perf/util/usage.c | 5 ++++- 2 files changed, 8 insertions(+), 2 deletions(-) diff --git a/tools/perf/util/config.c b/tools/perf/util/config.c index 80d9598db31a..0deac6a14b65 100644 --- a/tools/perf/util/config.c +++ b/tools/perf/util/config.c @@ -1,5 +1,8 @@ /* - * GIT - The information manager from hell + * config.c + * + * Helper functions for parsing config items. + * Originally copied from GIT source. * * Copyright (C) Linus Torvalds, 2005 * Copyright (C) Johannes Schindelin, 2005 diff --git a/tools/perf/util/usage.c b/tools/perf/util/usage.c index e16bf9a707e8..d76d1c0ff98f 100644 --- a/tools/perf/util/usage.c +++ b/tools/perf/util/usage.c @@ -1,5 +1,8 @@ /* - * GIT - The information manager from hell + * usage.c + * + * Various reporting routines. + * Originally copied from GIT source. * * Copyright (C) Linus Torvalds, 2005 */ -- cgit v1.2.3 From cb8f4e9aa37c469ddd80dda51469f327606c0118 Mon Sep 17 00:00:00 2001 From: Namhyung Kim Date: Tue, 13 Dec 2011 00:16:55 +0900 Subject: perf events: Tidy up perf_event__preprocess_sample Use local variable 'dso' to reduce typing a bit and rearrange the if condition. Also NULL check of al->map in the condition is not necessary. Cc: Ingo Molnar Cc: Paul Mackerras Cc: Peter Zijlstra Link: http://lkml.kernel.org/r/1323703017-6060-7-git-send-email-namhyung@gmail.com Signed-off-by: Namhyung Kim Signed-off-by: Arnaldo Carvalho de Melo --- tools/perf/util/event.c | 13 +++++++------ 1 file changed, 7 insertions(+), 6 deletions(-) diff --git a/tools/perf/util/event.c b/tools/perf/util/event.c index 97c479bcb0dc..b7c7f39a8f6d 100644 --- a/tools/perf/util/event.c +++ b/tools/perf/util/event.c @@ -814,13 +814,14 @@ int perf_event__preprocess_sample(const union perf_event *event, al->cpu = sample->cpu; if (al->map) { + struct dso *dso = al->map->dso; + if (symbol_conf.dso_list && - (!al->map || !al->map->dso || - !(strlist__has_entry(symbol_conf.dso_list, - al->map->dso->short_name) || - (al->map->dso->short_name != al->map->dso->long_name && - strlist__has_entry(symbol_conf.dso_list, - al->map->dso->long_name))))) + (!dso || !(strlist__has_entry(symbol_conf.dso_list, + dso->short_name) || + (dso->short_name != dso->long_name && + strlist__has_entry(symbol_conf.dso_list, + dso->long_name))))) goto out_filtered; al->sym = map__find_symbol(al->map, al->addr, filter); -- cgit v1.2.3 From 65c1e0452a3389f9b7b8c1b23305ed2922fafb2d Mon Sep 17 00:00:00 2001 From: Jiri Olsa Date: Thu, 15 Dec 2011 16:30:39 +0100 Subject: perf test: Add more automated tests for event parsing Adding automated tests for event parsing to include testing for modifier and ',' operator. Cc: Ingo Molnar Cc: Paul Mackerras Cc: Peter Zijlstra Cc: cjashfor@linux.vnet.ibm.com Link: http://lkml.kernel.org/r/1323963039-7602-4-git-send-email-jolsa@redhat.com Signed-off-by: Jiri Olsa [ committer note: Remove some tests that need group_leader & bison patchkits ] Signed-off-by: Arnaldo Carvalho de Melo --- tools/perf/builtin-test.c | 127 +++++++++++++++++++++++++++++++++++++++++++++- 1 file changed, 126 insertions(+), 1 deletion(-) diff --git a/tools/perf/builtin-test.c b/tools/perf/builtin-test.c index 6173f780dce0..2b9a7f497a20 100644 --- a/tools/perf/builtin-test.c +++ b/tools/perf/builtin-test.c @@ -603,7 +603,7 @@ out_free_threads: #define TEST_ASSERT_VAL(text, cond) \ do { \ - if (!cond) { \ + if (!(cond)) { \ pr_debug("FAILED %s:%d %s\n", __FILE__, __LINE__, text); \ return -1; \ } \ @@ -759,6 +759,103 @@ static int test__checkevent_breakpoint_w(struct perf_evlist *evlist) return 0; } +static int test__checkevent_tracepoint_modifier(struct perf_evlist *evlist) +{ + struct perf_evsel *evsel = list_entry(evlist->entries.next, + struct perf_evsel, node); + + TEST_ASSERT_VAL("wrong exclude_user", evsel->attr.exclude_user); + TEST_ASSERT_VAL("wrong exclude_kernel", !evsel->attr.exclude_kernel); + TEST_ASSERT_VAL("wrong exclude_hv", evsel->attr.exclude_hv); + TEST_ASSERT_VAL("wrong precise_ip", !evsel->attr.precise_ip); + + return test__checkevent_tracepoint(evlist); +} + +static int +test__checkevent_tracepoint_multi_modifier(struct perf_evlist *evlist) +{ + struct perf_evsel *evsel; + + TEST_ASSERT_VAL("wrong number of entries", evlist->nr_entries > 1); + + list_for_each_entry(evsel, &evlist->entries, node) { + TEST_ASSERT_VAL("wrong exclude_user", + !evsel->attr.exclude_user); + TEST_ASSERT_VAL("wrong exclude_kernel", + evsel->attr.exclude_kernel); + TEST_ASSERT_VAL("wrong exclude_hv", evsel->attr.exclude_hv); + TEST_ASSERT_VAL("wrong precise_ip", !evsel->attr.precise_ip); + } + + return test__checkevent_tracepoint_multi(evlist); +} + +static int test__checkevent_raw_modifier(struct perf_evlist *evlist) +{ + struct perf_evsel *evsel = list_entry(evlist->entries.next, + struct perf_evsel, node); + + TEST_ASSERT_VAL("wrong exclude_user", evsel->attr.exclude_user); + TEST_ASSERT_VAL("wrong exclude_kernel", !evsel->attr.exclude_kernel); + TEST_ASSERT_VAL("wrong exclude_hv", evsel->attr.exclude_hv); + TEST_ASSERT_VAL("wrong precise_ip", evsel->attr.precise_ip); + + return test__checkevent_raw(evlist); +} + +static int test__checkevent_numeric_modifier(struct perf_evlist *evlist) +{ + struct perf_evsel *evsel = list_entry(evlist->entries.next, + struct perf_evsel, node); + + TEST_ASSERT_VAL("wrong exclude_user", evsel->attr.exclude_user); + TEST_ASSERT_VAL("wrong exclude_kernel", evsel->attr.exclude_kernel); + TEST_ASSERT_VAL("wrong exclude_hv", !evsel->attr.exclude_hv); + TEST_ASSERT_VAL("wrong precise_ip", evsel->attr.precise_ip); + + return test__checkevent_numeric(evlist); +} + +static int test__checkevent_symbolic_name_modifier(struct perf_evlist *evlist) +{ + struct perf_evsel *evsel = list_entry(evlist->entries.next, + struct perf_evsel, node); + + TEST_ASSERT_VAL("wrong exclude_user", evsel->attr.exclude_user); + TEST_ASSERT_VAL("wrong exclude_kernel", evsel->attr.exclude_kernel); + TEST_ASSERT_VAL("wrong exclude_hv", !evsel->attr.exclude_hv); + TEST_ASSERT_VAL("wrong precise_ip", !evsel->attr.precise_ip); + + return test__checkevent_symbolic_name(evlist); +} + +static int test__checkevent_symbolic_alias_modifier(struct perf_evlist *evlist) +{ + struct perf_evsel *evsel = list_entry(evlist->entries.next, + struct perf_evsel, node); + + TEST_ASSERT_VAL("wrong exclude_user", !evsel->attr.exclude_user); + TEST_ASSERT_VAL("wrong exclude_kernel", evsel->attr.exclude_kernel); + TEST_ASSERT_VAL("wrong exclude_hv", evsel->attr.exclude_hv); + TEST_ASSERT_VAL("wrong precise_ip", !evsel->attr.precise_ip); + + return test__checkevent_symbolic_alias(evlist); +} + +static int test__checkevent_genhw_modifier(struct perf_evlist *evlist) +{ + struct perf_evsel *evsel = list_entry(evlist->entries.next, + struct perf_evsel, node); + + TEST_ASSERT_VAL("wrong exclude_user", evsel->attr.exclude_user); + TEST_ASSERT_VAL("wrong exclude_kernel", !evsel->attr.exclude_kernel); + TEST_ASSERT_VAL("wrong exclude_hv", evsel->attr.exclude_hv); + TEST_ASSERT_VAL("wrong precise_ip", evsel->attr.precise_ip); + + return test__checkevent_genhw(evlist); +} + static struct test__event_st { const char *name; __u32 type; @@ -808,6 +905,34 @@ static struct test__event_st { .name = "mem:0:w", .check = test__checkevent_breakpoint_w, }, + { + .name = "syscalls:sys_enter_open:k", + .check = test__checkevent_tracepoint_modifier, + }, + { + .name = "syscalls:*:u", + .check = test__checkevent_tracepoint_multi_modifier, + }, + { + .name = "r1:kp", + .check = test__checkevent_raw_modifier, + }, + { + .name = "1:1:hp", + .check = test__checkevent_numeric_modifier, + }, + { + .name = "instructions:h", + .check = test__checkevent_symbolic_name_modifier, + }, + { + .name = "faults:u", + .check = test__checkevent_symbolic_alias_modifier, + }, + { + .name = "L1-dcache-load-miss:kp", + .check = test__checkevent_genhw_modifier, + }, }; #define TEST__EVENTS_CNT (sizeof(test__events) / sizeof(struct test__event_st)) -- cgit v1.2.3 From cd09c0c40a971549800ce6a7e53c63f5139dd175 Mon Sep 17 00:00:00 2001 From: Stephane Eranian Date: Sun, 11 Dec 2011 00:28:51 +0100 Subject: perf events: Enable raw event support for Intel unhalted_reference_cycles event This patch adds the encoding and definitions necessary for the unhalted_reference_cycles event avaialble since Intel Core 2 processors. Signed-off-by: Stephane Eranian Signed-off-by: Peter Zijlstra Link: http://lkml.kernel.org/r/1323559734-3488-2-git-send-email-eranian@google.com Signed-off-by: Ingo Molnar --- arch/x86/include/asm/perf_event.h | 15 ++++++++------- arch/x86/kernel/cpu/perf_event.c | 8 +++++++- arch/x86/kernel/cpu/perf_event_intel.c | 15 +++++---------- 3 files changed, 20 insertions(+), 18 deletions(-) diff --git a/arch/x86/include/asm/perf_event.h b/arch/x86/include/asm/perf_event.h index b50e9d15aae0..096c975e099f 100644 --- a/arch/x86/include/asm/perf_event.h +++ b/arch/x86/include/asm/perf_event.h @@ -112,23 +112,24 @@ struct x86_pmu_capability { /* * All 3 fixed-mode PMCs are configured via this single MSR: */ -#define MSR_ARCH_PERFMON_FIXED_CTR_CTRL 0x38d +#define MSR_ARCH_PERFMON_FIXED_CTR_CTRL 0x38d /* * The counts are available in three separate MSRs: */ /* Instr_Retired.Any: */ -#define MSR_ARCH_PERFMON_FIXED_CTR0 0x309 -#define X86_PMC_IDX_FIXED_INSTRUCTIONS (X86_PMC_IDX_FIXED + 0) +#define MSR_ARCH_PERFMON_FIXED_CTR0 0x309 +#define X86_PMC_IDX_FIXED_INSTRUCTIONS (X86_PMC_IDX_FIXED + 0) /* CPU_CLK_Unhalted.Core: */ -#define MSR_ARCH_PERFMON_FIXED_CTR1 0x30a -#define X86_PMC_IDX_FIXED_CPU_CYCLES (X86_PMC_IDX_FIXED + 1) +#define MSR_ARCH_PERFMON_FIXED_CTR1 0x30a +#define X86_PMC_IDX_FIXED_CPU_CYCLES (X86_PMC_IDX_FIXED + 1) /* CPU_CLK_Unhalted.Ref: */ -#define MSR_ARCH_PERFMON_FIXED_CTR2 0x30b -#define X86_PMC_IDX_FIXED_BUS_CYCLES (X86_PMC_IDX_FIXED + 2) +#define MSR_ARCH_PERFMON_FIXED_CTR2 0x30b +#define X86_PMC_IDX_FIXED_REF_CYCLES (X86_PMC_IDX_FIXED + 2) +#define X86_PMC_MSK_FIXED_REF_CYCLES (1ULL << X86_PMC_IDX_FIXED_REF_CYCLES) /* * We model BTS tracing as another fixed-mode PMC. diff --git a/arch/x86/kernel/cpu/perf_event.c b/arch/x86/kernel/cpu/perf_event.c index 930fe4879542..5adce1040b11 100644 --- a/arch/x86/kernel/cpu/perf_event.c +++ b/arch/x86/kernel/cpu/perf_event.c @@ -1304,9 +1304,15 @@ static int __init init_hw_perf_events(void) 0, x86_pmu.num_counters, 0); if (x86_pmu.event_constraints) { + /* + * event on fixed counter2 (REF_CYCLES) only works on this + * counter, so do not extend mask to generic counters + */ for_each_event_constraint(c, x86_pmu.event_constraints) { - if (c->cmask != X86_RAW_EVENT_MASK) + if (c->cmask != X86_RAW_EVENT_MASK + || c->idxmsk64 == X86_PMC_MSK_FIXED_REF_CYCLES) { continue; + } c->idxmsk64 |= (1ULL << x86_pmu.num_counters) - 1; c->weight += x86_pmu.num_counters; diff --git a/arch/x86/kernel/cpu/perf_event_intel.c b/arch/x86/kernel/cpu/perf_event_intel.c index 2c3bf53d0302..61f865f947b3 100644 --- a/arch/x86/kernel/cpu/perf_event_intel.c +++ b/arch/x86/kernel/cpu/perf_event_intel.c @@ -45,12 +45,7 @@ static struct event_constraint intel_core2_event_constraints[] __read_mostly = { FIXED_EVENT_CONSTRAINT(0x00c0, 0), /* INST_RETIRED.ANY */ FIXED_EVENT_CONSTRAINT(0x003c, 1), /* CPU_CLK_UNHALTED.CORE */ - /* - * Core2 has Fixed Counter 2 listed as CPU_CLK_UNHALTED.REF and event - * 0x013c as CPU_CLK_UNHALTED.BUS and specifies there is a fixed - * ratio between these counters. - */ - /* FIXED_EVENT_CONSTRAINT(0x013c, 2), CPU_CLK_UNHALTED.REF */ + FIXED_EVENT_CONSTRAINT(0x0300, 2), /* CPU_CLK_UNHALTED.REF */ INTEL_EVENT_CONSTRAINT(0x10, 0x1), /* FP_COMP_OPS_EXE */ INTEL_EVENT_CONSTRAINT(0x11, 0x2), /* FP_ASSIST */ INTEL_EVENT_CONSTRAINT(0x12, 0x2), /* MUL */ @@ -68,7 +63,7 @@ static struct event_constraint intel_nehalem_event_constraints[] __read_mostly = { FIXED_EVENT_CONSTRAINT(0x00c0, 0), /* INST_RETIRED.ANY */ FIXED_EVENT_CONSTRAINT(0x003c, 1), /* CPU_CLK_UNHALTED.CORE */ - /* FIXED_EVENT_CONSTRAINT(0x013c, 2), CPU_CLK_UNHALTED.REF */ + FIXED_EVENT_CONSTRAINT(0x0300, 2), /* CPU_CLK_UNHALTED.REF */ INTEL_EVENT_CONSTRAINT(0x40, 0x3), /* L1D_CACHE_LD */ INTEL_EVENT_CONSTRAINT(0x41, 0x3), /* L1D_CACHE_ST */ INTEL_EVENT_CONSTRAINT(0x42, 0x3), /* L1D_CACHE_LOCK */ @@ -90,7 +85,7 @@ static struct event_constraint intel_westmere_event_constraints[] __read_mostly { FIXED_EVENT_CONSTRAINT(0x00c0, 0), /* INST_RETIRED.ANY */ FIXED_EVENT_CONSTRAINT(0x003c, 1), /* CPU_CLK_UNHALTED.CORE */ - /* FIXED_EVENT_CONSTRAINT(0x013c, 2), CPU_CLK_UNHALTED.REF */ + FIXED_EVENT_CONSTRAINT(0x0300, 2), /* CPU_CLK_UNHALTED.REF */ INTEL_EVENT_CONSTRAINT(0x51, 0x3), /* L1D */ INTEL_EVENT_CONSTRAINT(0x60, 0x1), /* OFFCORE_REQUESTS_OUTSTANDING */ INTEL_EVENT_CONSTRAINT(0x63, 0x3), /* CACHE_LOCK_CYCLES */ @@ -102,7 +97,7 @@ static struct event_constraint intel_snb_event_constraints[] __read_mostly = { FIXED_EVENT_CONSTRAINT(0x00c0, 0), /* INST_RETIRED.ANY */ FIXED_EVENT_CONSTRAINT(0x003c, 1), /* CPU_CLK_UNHALTED.CORE */ - /* FIXED_EVENT_CONSTRAINT(0x013c, 2), CPU_CLK_UNHALTED.REF */ + FIXED_EVENT_CONSTRAINT(0x0300, 2), /* CPU_CLK_UNHALTED.REF */ INTEL_EVENT_CONSTRAINT(0x48, 0x4), /* L1D_PEND_MISS.PENDING */ INTEL_UEVENT_CONSTRAINT(0x01c0, 0x2), /* INST_RETIRED.PREC_DIST */ INTEL_EVENT_CONSTRAINT(0xcd, 0x8), /* MEM_TRANS_RETIRED.LOAD_LATENCY */ @@ -125,7 +120,7 @@ static struct event_constraint intel_gen_event_constraints[] __read_mostly = { FIXED_EVENT_CONSTRAINT(0x00c0, 0), /* INST_RETIRED.ANY */ FIXED_EVENT_CONSTRAINT(0x003c, 1), /* CPU_CLK_UNHALTED.CORE */ - /* FIXED_EVENT_CONSTRAINT(0x013c, 2), CPU_CLK_UNHALTED.REF */ + FIXED_EVENT_CONSTRAINT(0x0300, 2), /* CPU_CLK_UNHALTED.REF */ EVENT_CONSTRAINT_END }; -- cgit v1.2.3 From c37e17497e01fc0f5d2d6feb5723b210b3ab8890 Mon Sep 17 00:00:00 2001 From: Stephane Eranian Date: Sun, 11 Dec 2011 00:28:52 +0100 Subject: perf events: Add PERF_COUNT_HW_REF_CPU_CYCLES generic PMU event This event counts the number of reference core cpu cycles. Reference means that the event increments at a constant rate which is not subject to core CPU frequency adjustments. The event may not count when the processor is in halted (low power) state. As such, it may not be equivalent to wall clock time. However, when the processor is not halted state, the event keeps a constant correlation with wall clock time. Signed-off-by: Stephane Eranian Signed-off-by: Peter Zijlstra Link: http://lkml.kernel.org/r/1323559734-3488-3-git-send-email-eranian@google.com Signed-off-by: Ingo Molnar --- include/linux/perf_event.h | 1 + 1 file changed, 1 insertion(+) diff --git a/include/linux/perf_event.h b/include/linux/perf_event.h index 564769cdb473..08855613ceb3 100644 --- a/include/linux/perf_event.h +++ b/include/linux/perf_event.h @@ -54,6 +54,7 @@ enum perf_hw_id { PERF_COUNT_HW_BUS_CYCLES = 6, PERF_COUNT_HW_STALLED_CYCLES_FRONTEND = 7, PERF_COUNT_HW_STALLED_CYCLES_BACKEND = 8, + PERF_COUNT_HW_REF_CPU_CYCLES = 9, PERF_COUNT_HW_MAX, /* non-ABI */ }; -- cgit v1.2.3 From 9c1497ea591b25d491f8e795f90a1405100b75ef Mon Sep 17 00:00:00 2001 From: Stephane Eranian Date: Sun, 11 Dec 2011 00:28:53 +0100 Subject: perf events: Add Intel x86 mapping for PERF_COUNT_HW_REF_CPU_CYCLES Add event maps for Intel x86 processors (with architected PMU v2 or later). On AMD, there is frequency scaling but no Turbo. There is no core cycle event not subject to frequency scaling, therefore we do not provide a mapping. Signed-off-by: Stephane Eranian Signed-off-by: Peter Zijlstra Link: http://lkml.kernel.org/r/1323559734-3488-4-git-send-email-eranian@google.com Signed-off-by: Ingo Molnar --- arch/x86/kernel/cpu/perf_event_intel.c | 1 + 1 file changed, 1 insertion(+) diff --git a/arch/x86/kernel/cpu/perf_event_intel.c b/arch/x86/kernel/cpu/perf_event_intel.c index 61f865f947b3..cbfaaa2475ea 100644 --- a/arch/x86/kernel/cpu/perf_event_intel.c +++ b/arch/x86/kernel/cpu/perf_event_intel.c @@ -28,6 +28,7 @@ static u64 intel_perfmon_event_map[PERF_COUNT_HW_MAX] __read_mostly = [PERF_COUNT_HW_BRANCH_INSTRUCTIONS] = 0x00c4, [PERF_COUNT_HW_BRANCH_MISSES] = 0x00c5, [PERF_COUNT_HW_BUS_CYCLES] = 0x013c, + [PERF_COUNT_HW_REF_CPU_CYCLES] = 0x0300, /* pseudo-encoding */ }; static struct event_constraint intel_core_event_constraints[] __read_mostly = -- cgit v1.2.3 From f1ac18af219835fd5b8e19c14d2dd75c55f78737 Mon Sep 17 00:00:00 2001 From: Stephane Eranian Date: Sun, 11 Dec 2011 00:28:54 +0100 Subject: perf: Add support for PERF_HW_COUNT_REF_CPU_CYCLES Add new generic hw event: ref-cycles, which maps to PERF_HW_COUNT_REF_CPUCYCLES: $ perf stat -e ref-cycles ls Signed-off-by: Stephane Eranian Signed-off-by: Peter Zijlstra Link: http://lkml.kernel.org/r/1323559734-3488-5-git-send-email-eranian@google.com Signed-off-by: Ingo Molnar --- tools/perf/util/parse-events.c | 2 ++ 1 file changed, 2 insertions(+) diff --git a/tools/perf/util/parse-events.c b/tools/perf/util/parse-events.c index 586ab3fe60f8..531c283fc0c5 100644 --- a/tools/perf/util/parse-events.c +++ b/tools/perf/util/parse-events.c @@ -38,6 +38,7 @@ static struct event_symbol event_symbols[] = { { CHW(BRANCH_INSTRUCTIONS), "branch-instructions", "branches" }, { CHW(BRANCH_MISSES), "branch-misses", "" }, { CHW(BUS_CYCLES), "bus-cycles", "" }, + { CHW(REF_CPU_CYCLES), "ref-cycles", "" }, { CSW(CPU_CLOCK), "cpu-clock", "" }, { CSW(TASK_CLOCK), "task-clock", "" }, @@ -68,6 +69,7 @@ static const char *hw_event_names[PERF_COUNT_HW_MAX] = { "bus-cycles", "stalled-cycles-frontend", "stalled-cycles-backend", + "ref-cycles", }; static const char *sw_event_names[PERF_COUNT_SW_MAX] = { -- cgit v1.2.3 From fb2baceb5a64990163e93b77ee205d0173202ee6 Mon Sep 17 00:00:00 2001 From: Namhyung Kim Date: Tue, 13 Dec 2011 00:16:56 +0900 Subject: perf report: Fix usage string perf report does not take a command from command line. Cc: Ingo Molnar Cc: Paul Mackerras Cc: Peter Zijlstra Link: http://lkml.kernel.org/r/1323703017-6060-8-git-send-email-namhyung@gmail.com Signed-off-by: Namhyung Kim Signed-off-by: Arnaldo Carvalho de Melo --- tools/perf/builtin-report.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) diff --git a/tools/perf/builtin-report.c b/tools/perf/builtin-report.c index b2654c9fb5c6..9051f6bfaa7e 100644 --- a/tools/perf/builtin-report.c +++ b/tools/perf/builtin-report.c @@ -432,7 +432,7 @@ int cmd_report(int argc, const char **argv, const char *prefix __used) { char callchain_default_opt[] = "fractal,0.5,callee"; const char * const report_usage[] = { - "perf report [] ", + "perf report []", NULL }; struct perf_report report = { -- cgit v1.2.3 From defd8d38773cf9e01c69a903d04d5895b78ee74f Mon Sep 17 00:00:00 2001 From: David Ahern Date: Thu, 22 Dec 2011 11:30:01 -0700 Subject: perf tools: Fix comm for processes with named threads perf does not properly handle monitoring of processes with named threads. For example: $ ps -C myapp -L PID LWP TTY TIME CMD 25118 25118 ? 00:00:00 myapp 25118 25119 ? 00:00:00 myapp:worker perf record -e cs -c 1 -fo /tmp/perf.data -p 25118 -- sleep 10 perf report --stdio -i /tmp/perf.data 100.00% myapp:worker [kernel.kallsyms] [k] perf_event_task_sched_out The process name is set to the name of the last thread it finds for the process. The Problem: perf-top and perf-record both create a thread_map of threads to be monitored. That map is used in perf_event__synthesize_thread_map which loops over the entries in thread_map and calls __event__synthesize_thread to generate COMM and MMAP events. __event__synthesize_thread calls perf_event__synthesize_comm which opens /proc/pid/status, reads the name of the task and its thread group id. That's all fine. The problem is that it then reads /proc/pid/task and generates COMM events for each task it finds - but using the name found in /proc/pid/status where pid is the thread of interest. The end result (looping over thread_map + synthesizing comm events for each thread each time) means the name of the last thread processed sets the name for all threads in the process - which is not good for multithreaded processes with named threads. The Fix: perf_event__synthesize_comm has an input argument (full) that decides whether to process task entries for each pid it is passed. It currently never set to 0 (perf_event__synthesize_comm has a single caller and it always passes the value 1). Let's fix that. Add the full input argument to __event__synthesize_thread which passes it to perf_event__synthesize_comm. For thread/process monitoring set full to 0 which means COMM and MMAP events are only generated for the pid passed to it. For system wide monitoring set full to 1 so that COMM events are generated for all threads in a process. Cc: Frederic Weisbecker Cc: Ingo Molnar Cc: Peter Zijlstra Link: http://lkml.kernel.org/r/1324578603-12762-2-git-send-email-dsahern@gmail.com Signed-off-by: David Ahern Signed-off-by: Arnaldo Carvalho de Melo --- tools/perf/util/event.c | 37 ++++++++++++++++++++++++++++++++----- 1 file changed, 32 insertions(+), 5 deletions(-) diff --git a/tools/perf/util/event.c b/tools/perf/util/event.c index b7c7f39a8f6d..a5787260181a 100644 --- a/tools/perf/util/event.c +++ b/tools/perf/util/event.c @@ -261,11 +261,12 @@ int perf_event__synthesize_modules(struct perf_tool *tool, static int __event__synthesize_thread(union perf_event *comm_event, union perf_event *mmap_event, - pid_t pid, perf_event__handler_t process, + pid_t pid, int full, + perf_event__handler_t process, struct perf_tool *tool, struct machine *machine) { - pid_t tgid = perf_event__synthesize_comm(tool, comm_event, pid, 1, + pid_t tgid = perf_event__synthesize_comm(tool, comm_event, pid, full, process, machine); if (tgid == -1) return -1; @@ -279,7 +280,7 @@ int perf_event__synthesize_thread_map(struct perf_tool *tool, struct machine *machine) { union perf_event *comm_event, *mmap_event; - int err = -1, thread; + int err = -1, thread, j; comm_event = malloc(sizeof(comm_event->comm) + machine->id_hdr_size); if (comm_event == NULL) @@ -292,11 +293,37 @@ int perf_event__synthesize_thread_map(struct perf_tool *tool, err = 0; for (thread = 0; thread < threads->nr; ++thread) { if (__event__synthesize_thread(comm_event, mmap_event, - threads->map[thread], + threads->map[thread], 0, process, tool, machine)) { err = -1; break; } + + /* + * comm.pid is set to thread group id by + * perf_event__synthesize_comm + */ + if ((int) comm_event->comm.pid != threads->map[thread]) { + bool need_leader = true; + + /* is thread group leader in thread_map? */ + for (j = 0; j < threads->nr; ++j) { + if ((int) comm_event->comm.pid == threads->map[j]) { + need_leader = false; + break; + } + } + + /* if not, generate events for it */ + if (need_leader && + __event__synthesize_thread(comm_event, + mmap_event, + comm_event->comm.pid, 0, + process, tool, machine)) { + err = -1; + break; + } + } } free(mmap_event); out_free_comm: @@ -333,7 +360,7 @@ int perf_event__synthesize_threads(struct perf_tool *tool, if (*end) /* only interested in proper numerical dirents */ continue; - __event__synthesize_thread(comm_event, mmap_event, pid, + __event__synthesize_thread(comm_event, mmap_event, pid, 1, process, tool, machine); } -- cgit v1.2.3 From f5faf726184a6a5ca1735f610cb97e509fce33e2 Mon Sep 17 00:00:00 2001 From: David Ahern Date: Thu, 22 Dec 2011 11:30:02 -0700 Subject: perf tools: Look up thread names for system wide profiling This handles multithreaded processes with named threads when doing system wide profiling: the comm for each thread is looked up allowing them to be different from the thread group leader. v2: - fixed sizeof arg to perf_event__get_comm_tgid Cc: Frederic Weisbecker Cc: Ingo Molnar Cc: Peter Zijlstra Link: http://lkml.kernel.org/r/1324578603-12762-3-git-send-email-dsahern@gmail.com Signed-off-by: David Ahern Signed-off-by: Arnaldo Carvalho de Melo --- tools/perf/util/event.c | 75 ++++++++++++++++++++++++++++++++++--------------- 1 file changed, 53 insertions(+), 22 deletions(-) diff --git a/tools/perf/util/event.c b/tools/perf/util/event.c index a5787260181a..73ddaf06b8e7 100644 --- a/tools/perf/util/event.c +++ b/tools/perf/util/event.c @@ -43,37 +43,27 @@ static struct perf_sample synth_sample = { .period = 1, }; -static pid_t perf_event__synthesize_comm(struct perf_tool *tool, - union perf_event *event, pid_t pid, - int full, perf_event__handler_t process, - struct machine *machine) +static pid_t perf_event__get_comm_tgid(pid_t pid, char *comm, size_t len) { char filename[PATH_MAX]; char bf[BUFSIZ]; FILE *fp; size_t size = 0; - DIR *tasks; - struct dirent dirent, *next; - pid_t tgid = 0; + pid_t tgid = -1; snprintf(filename, sizeof(filename), "/proc/%d/status", pid); fp = fopen(filename, "r"); if (fp == NULL) { -out_race: - /* - * We raced with a task exiting - just return: - */ pr_debug("couldn't open %s\n", filename); return 0; } - memset(&event->comm, 0, sizeof(event->comm)); - - while (!event->comm.comm[0] || !event->comm.pid) { + while (!comm[0] || (tgid < 0)) { if (fgets(bf, sizeof(bf), fp) == NULL) { - pr_warning("couldn't get COMM and pgid, malformed %s\n", filename); - goto out; + pr_warning("couldn't get COMM and pgid, malformed %s\n", + filename); + break; } if (memcmp(bf, "Name:", 5) == 0) { @@ -81,16 +71,46 @@ out_race: while (*name && isspace(*name)) ++name; size = strlen(name) - 1; - memcpy(event->comm.comm, name, size++); + if (size >= len) + size = len - 1; + memcpy(comm, name, size); + } else if (memcmp(bf, "Tgid:", 5) == 0) { char *tgids = bf + 5; while (*tgids && isspace(*tgids)) ++tgids; - tgid = event->comm.pid = atoi(tgids); + tgid = atoi(tgids); } } + fclose(fp); + + return tgid; +} + +static pid_t perf_event__synthesize_comm(struct perf_tool *tool, + union perf_event *event, pid_t pid, + int full, + perf_event__handler_t process, + struct machine *machine) +{ + char filename[PATH_MAX]; + size_t size; + DIR *tasks; + struct dirent dirent, *next; + pid_t tgid; + + memset(&event->comm, 0, sizeof(event->comm)); + + tgid = perf_event__get_comm_tgid(pid, event->comm.comm, + sizeof(event->comm.comm)); + if (tgid < 0) + goto out; + + event->comm.pid = tgid; event->comm.header.type = PERF_RECORD_COMM; + + size = strlen(event->comm.comm) + 1; size = ALIGN(size, sizeof(u64)); memset(event->comm.comm + size, 0, machine->id_hdr_size); event->comm.header.size = (sizeof(event->comm) - @@ -106,8 +126,10 @@ out_race: snprintf(filename, sizeof(filename), "/proc/%d/task", pid); tasks = opendir(filename); - if (tasks == NULL) - goto out_race; + if (tasks == NULL) { + pr_debug("couldn't open %s\n", filename); + return 0; + } while (!readdir_r(tasks, &dirent, &next) && next) { char *end; @@ -115,6 +137,17 @@ out_race: if (*end) continue; + /* already have tgid; jut want to update the comm */ + (void) perf_event__get_comm_tgid(pid, event->comm.comm, + sizeof(event->comm.comm)); + + size = strlen(event->comm.comm) + 1; + size = ALIGN(size, sizeof(u64)); + memset(event->comm.comm + size, 0, machine->id_hdr_size); + event->comm.header.size = (sizeof(event->comm) - + (sizeof(event->comm.comm) - size) + + machine->id_hdr_size); + event->comm.tid = pid; process(tool, event, &synth_sample, machine); @@ -122,8 +155,6 @@ out_race: closedir(tasks); out: - fclose(fp); - return tgid; } -- cgit v1.2.3 From 64aab93cdffb3967642ffab954395ae2400c0b06 Mon Sep 17 00:00:00 2001 From: David Ahern Date: Thu, 22 Dec 2011 11:30:03 -0700 Subject: perf script: look up thread using tid instead of pid This allows the thread name to be dispalyed when dumping events: myapp 25118 [000] 450385.538815: context-switches ... myapp:worker 25119 [000] 450385.538894: context-switches ... Cc: Frederic Weisbecker Cc: Ingo Molnar Cc: Peter Zijlstra Link: http://lkml.kernel.org/r/1324578603-12762-4-git-send-email-dsahern@gmail.com Signed-off-by: David Ahern Signed-off-by: Arnaldo Carvalho de Melo --- tools/perf/builtin-script.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) diff --git a/tools/perf/builtin-script.c b/tools/perf/builtin-script.c index ea71c5e1a94f..d71b745da06e 100644 --- a/tools/perf/builtin-script.c +++ b/tools/perf/builtin-script.c @@ -443,7 +443,7 @@ static int process_sample_event(struct perf_tool *tool __used, struct machine *machine) { struct addr_location al; - struct thread *thread = machine__findnew_thread(machine, event->ip.pid); + struct thread *thread = machine__findnew_thread(machine, event->ip.tid); if (thread == NULL) { pr_debug("problem processing %d event, skipping it.\n", -- cgit v1.2.3 From f41612f43be9575e1160460b08c3a760e6e27e1b Mon Sep 17 00:00:00 2001 From: Ingo Molnar Date: Fri, 23 Dec 2011 14:08:04 +0100 Subject: perf tools: Fix truncated annotation MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit I get such truncated annotation results in 'perf top': : Disassembly of section .text: â–’ : â–’ : ffffffff810966a8 : â–’ 4.94 : ffffffff810966a8: movslq %edi,%rdi â–’ 3.70 : ffffffff810966ab: mov $0x13700,%rax â–’ 0.00 : ffffffff810966b2: add -0x7e32cb00(,%rdi,8),%rax â–’ 8.64 : ffffffff810966ba: mov 0x7e0(%rax),%eax â–’ 82.72 : ffffffff810966c0: cltq â–’ Note the missing 'retq' which is there in the original function: ffffffff810966a8 : ffffffff810966a8: 48 63 ff movslq %edi,%rdi ffffffff810966ab: 48 c7 c0 00 37 01 00 mov $0x13700,%rax ffffffff810966b2: 48 03 04 fd 00 35 cd add -0x7e32cb00(,%rdi,8),%rax ffffffff810966b9: 81 ffffffff810966ba: 8b 80 e0 07 00 00 mov 0x7e0(%rax),%eax ffffffff810966c0: 48 98 cltq ffffffff810966c2: c3 retq ffffffff810966c3 : I'm using a fairly recent binutils: GNU objdump version 2.21.51.0.6-2.fc16 20110118 AFAICS the bug is simply that sym->end points to the last byte of the symbol in question - while objdump's --stop-address expects the last byte plus 1 to disassemble the full range. Cc: Peter Zijlstra Link: http://lkml.kernel.org/r/20111223130804.GA24305@elte.hu Signed-off-by: Ingo Molnar Signed-off-by: Arnaldo Carvalho de Melo --- tools/perf/util/annotate.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) diff --git a/tools/perf/util/annotate.c b/tools/perf/util/annotate.c index 376e643f7066..011ed2676604 100644 --- a/tools/perf/util/annotate.c +++ b/tools/perf/util/annotate.c @@ -334,7 +334,7 @@ fallback: disassembler_style ? "-M " : "", disassembler_style ? disassembler_style : "", map__rip_2objdump(map, sym->start), - map__rip_2objdump(map, sym->end), + map__rip_2objdump(map, sym->end+1), symbol_conf.annotate_asm_raw ? "" : "--no-show-raw", symbol_conf.annotate_src ? "-S" : "", symfs_filename, filename); -- cgit v1.2.3 From 18e6093904abfd51671ff5846c2fdaba9ebbf21b Mon Sep 17 00:00:00 2001 From: Nelson Elhage Date: Mon, 19 Dec 2011 08:39:31 -0500 Subject: perf: builtin-record: Provide advice if mmap'ing fails with EPERM. This failure is most likely due to running up against the kernel.perf_event_mlock_kb sysctl, so we can tell the user what to do to fix the issue. Cc: Ingo Molnar Cc: Paul Mackerras Cc: Peter Zijlstra Link: http://lkml.kernel.org/r/1324301972-22740-3-git-send-email-nelhage@nelhage.com Signed-off-by: Nelson Elhage Signed-off-by: Arnaldo Carvalho de Melo --- tools/perf/builtin-record.c | 9 ++++++++- 1 file changed, 8 insertions(+), 1 deletion(-) diff --git a/tools/perf/builtin-record.c b/tools/perf/builtin-record.c index f8fd14fb62ec..56bb4476e3ba 100644 --- a/tools/perf/builtin-record.c +++ b/tools/perf/builtin-record.c @@ -272,8 +272,15 @@ try_again: exit(-1); } - if (perf_evlist__mmap(evlist, opts->mmap_pages, false) < 0) + if (perf_evlist__mmap(evlist, opts->mmap_pages, false) < 0) { + if (errno == EPERM) + die("Permission error mapping pages.\n" + "Consider increasing " + "/proc/sys/kernel/perf_event_mlock_kb,\n" + "or try again with a smaller value of -m/--mmap_pages.\n" + "(current value: %d)\n", opts->mmap_pages); die("failed to mmap with %d (%s)\n", errno, strerror(errno)); + } if (rec->file_new) session->evlist = evlist; -- cgit v1.2.3 From 41d0d933494ce10eb77758a1168b08e317c42e8e Mon Sep 17 00:00:00 2001 From: Nelson Elhage Date: Mon, 19 Dec 2011 08:39:32 -0500 Subject: perf: builtin-record: Document and check that mmap_pages must be a power of two. Now that we automatically point users at it, let's provide them some guidance so that they hopefully don't just get mysterious EINVAL's from the kernel. Cc: Ingo Molnar Cc: Paul Mackerras Cc: Peter Zijlstra Link: http://lkml.kernel.org/r/1324301972-22740-4-git-send-email-nelhage@nelhage.com Signed-off-by: Nelson Elhage [ committer note: Made it work after 50a682c ] Signed-off-by: Arnaldo Carvalho de Melo --- tools/perf/Documentation/perf-record.txt | 2 +- tools/perf/builtin-record.c | 3 +++ tools/perf/util/evlist.c | 2 ++ tools/perf/util/util.h | 11 +++++++++++ 4 files changed, 17 insertions(+), 1 deletion(-) diff --git a/tools/perf/Documentation/perf-record.txt b/tools/perf/Documentation/perf-record.txt index 5a520f825295..2937f7e14bb7 100644 --- a/tools/perf/Documentation/perf-record.txt +++ b/tools/perf/Documentation/perf-record.txt @@ -89,7 +89,7 @@ OPTIONS -m:: --mmap-pages=:: - Number of mmap data pages. + Number of mmap data pages. Must be a power of two. -g:: --call-graph:: diff --git a/tools/perf/builtin-record.c b/tools/perf/builtin-record.c index 56bb4476e3ba..e873ae2dd54c 100644 --- a/tools/perf/builtin-record.c +++ b/tools/perf/builtin-record.c @@ -279,6 +279,9 @@ try_again: "/proc/sys/kernel/perf_event_mlock_kb,\n" "or try again with a smaller value of -m/--mmap_pages.\n" "(current value: %d)\n", opts->mmap_pages); + else if (!is_power_of_2(opts->mmap_pages)) + die("--mmap_pages/-m value must be a power of two."); + die("failed to mmap with %d (%s)\n", errno, strerror(errno)); } diff --git a/tools/perf/util/evlist.c b/tools/perf/util/evlist.c index 963d63dde457..fa1837088ca8 100644 --- a/tools/perf/util/evlist.c +++ b/tools/perf/util/evlist.c @@ -563,6 +563,8 @@ int perf_evlist__mmap(struct perf_evlist *evlist, unsigned int pages, /* 512 kiB: default amount of unprivileged mlocked memory */ if (pages == UINT_MAX) pages = (512 * 1024) / page_size; + else if (!is_power_of_2(pages)) + return -EINVAL; mask = pages * page_size - 1; diff --git a/tools/perf/util/util.h b/tools/perf/util/util.h index 0128906bac88..37be34dff798 100644 --- a/tools/perf/util/util.h +++ b/tools/perf/util/util.h @@ -245,4 +245,15 @@ int readn(int fd, void *buf, size_t size); #define _STR(x) #x #define STR(x) _STR(x) +/* + * Determine whether some value is a power of two, where zero is + * *not* considered a power of two. + */ + +static inline __attribute__((const)) +bool is_power_of_2(unsigned long n) +{ + return (n != 0 && ((n & (n - 1)) == 0)); +} + #endif -- cgit v1.2.3 From 8cdfa78a885d94a79205d183a611ebc4876d6f33 Mon Sep 17 00:00:00 2001 From: Robert Richter Date: Wed, 7 Dec 2011 10:02:56 +0100 Subject: perf tools: Improve macros for struct feature_ops Reducing duplication and line size by extending function names for print and write from a single name. Cc: Frederic Weisbecker Cc: Ingo Molnar Cc: Peter Zijlstra Cc: Stephane Eranian Link: http://lkml.kernel.org/r/1323248577-11268-7-git-send-email-robert.richter@amd.com Signed-off-by: Robert Richter Signed-off-by: Arnaldo Carvalho de Melo --- tools/perf/util/header.c | 40 ++++++++++++++++++++++------------------ 1 file changed, 22 insertions(+), 18 deletions(-) diff --git a/tools/perf/util/header.c b/tools/perf/util/header.c index 5b01449152ef..4c48be80dcdf 100644 --- a/tools/perf/util/header.c +++ b/tools/perf/util/header.c @@ -1065,26 +1065,30 @@ struct feature_ops { bool full_only; }; -#define FEAT_OPA(n, w, p) \ - [n] = { .name = #n, .write = w, .print = p } -#define FEAT_OPF(n, w, p) \ - [n] = { .name = #n, .write = w, .print = p, .full_only = true } +#define FEAT_OPA(n, func) \ + [n] = { .name = #n, .write = write_##func, .print = print_##func } +#define FEAT_OPF(n, func) \ + [n] = { .name = #n, .write = write_##func, .print = print_##func, .full_only = true } + +/* feature_ops not implemented: */ +#define print_trace_info NULL +#define print_build_id NULL static const struct feature_ops feat_ops[HEADER_LAST_FEATURE] = { - FEAT_OPA(HEADER_TRACE_INFO, write_trace_info, NULL), - FEAT_OPA(HEADER_BUILD_ID, write_build_id, NULL), - FEAT_OPA(HEADER_HOSTNAME, write_hostname, print_hostname), - FEAT_OPA(HEADER_OSRELEASE, write_osrelease, print_osrelease), - FEAT_OPA(HEADER_VERSION, write_version, print_version), - FEAT_OPA(HEADER_ARCH, write_arch, print_arch), - FEAT_OPA(HEADER_NRCPUS, write_nrcpus, print_nrcpus), - FEAT_OPA(HEADER_CPUDESC, write_cpudesc, print_cpudesc), - FEAT_OPA(HEADER_CPUID, write_cpuid, print_cpuid), - FEAT_OPA(HEADER_TOTAL_MEM, write_total_mem, print_total_mem), - FEAT_OPA(HEADER_EVENT_DESC, write_event_desc, print_event_desc), - FEAT_OPA(HEADER_CMDLINE, write_cmdline, print_cmdline), - FEAT_OPF(HEADER_CPU_TOPOLOGY, write_cpu_topology, print_cpu_topology), - FEAT_OPF(HEADER_NUMA_TOPOLOGY, write_numa_topology, print_numa_topology), + FEAT_OPA(HEADER_TRACE_INFO, trace_info), + FEAT_OPA(HEADER_BUILD_ID, build_id), + FEAT_OPA(HEADER_HOSTNAME, hostname), + FEAT_OPA(HEADER_OSRELEASE, osrelease), + FEAT_OPA(HEADER_VERSION, version), + FEAT_OPA(HEADER_ARCH, arch), + FEAT_OPA(HEADER_NRCPUS, nrcpus), + FEAT_OPA(HEADER_CPUDESC, cpudesc), + FEAT_OPA(HEADER_CPUID, cpuid), + FEAT_OPA(HEADER_TOTAL_MEM, total_mem), + FEAT_OPA(HEADER_EVENT_DESC, event_desc), + FEAT_OPA(HEADER_CMDLINE, cmdline), + FEAT_OPF(HEADER_CPU_TOPOLOGY, cpu_topology), + FEAT_OPF(HEADER_NUMA_TOPOLOGY, numa_topology), }; struct header_print_data { -- cgit v1.2.3 From f7a8a1336416883dc0ccd96c17c604e34de61c25 Mon Sep 17 00:00:00 2001 From: Robert Richter Date: Wed, 7 Dec 2011 10:02:51 +0100 Subject: perf tools: Continue processing header on unknown features A feature may be unknown if perf.data is created and parsed on different perf tool versions. This should not stop the header to be processed, instead continue processing it. Cc: Frederic Weisbecker Cc: Ingo Molnar Cc: Peter Zijlstra Cc: Stephane Eranian Link: http://lkml.kernel.org/r/1323248577-11268-2-git-send-email-robert.richter@amd.com Signed-off-by: Robert Richter Signed-off-by: Arnaldo Carvalho de Melo --- tools/perf/util/header.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) diff --git a/tools/perf/util/header.c b/tools/perf/util/header.c index 4c48be80dcdf..428a4a2ce350 100644 --- a/tools/perf/util/header.c +++ b/tools/perf/util/header.c @@ -1109,7 +1109,7 @@ static int perf_file_section__fprintf_info(struct perf_file_section *section, } if (feat < HEADER_TRACE_INFO || feat >= HEADER_LAST_FEATURE) { pr_warning("unknown feature %d\n", feat); - return -1; + return 0; } if (!feat_ops[feat].print) return 0; -- cgit v1.2.3 From 002c4fd92d772becf8745b9cbcebe5c95fe6dad0 Mon Sep 17 00:00:00 2001 From: Robert Richter Date: Wed, 7 Dec 2011 10:02:52 +0100 Subject: perf tools: Fix out-of-bound access to struct perf_session If filename is NULL there is an out-of-bound access to struct perf_session if it would be used with perf_session__open(). Shouldn't actually happen in current implementation as filename is always !NULL. Fixing this by always null-terminating filename. Cc: Frederic Weisbecker Cc: Ingo Molnar Cc: Peter Zijlstra Cc: Stephane Eranian Link: http://lkml.kernel.org/r/1323248577-11268-3-git-send-email-robert.richter@amd.com Signed-off-by: Robert Richter Signed-off-by: Arnaldo Carvalho de Melo --- tools/perf/util/session.c | 2 +- tools/perf/util/session.h | 2 +- 2 files changed, 2 insertions(+), 2 deletions(-) diff --git a/tools/perf/util/session.c b/tools/perf/util/session.c index d9318d8a9ba1..ea17dfb85baa 100644 --- a/tools/perf/util/session.c +++ b/tools/perf/util/session.c @@ -107,7 +107,7 @@ struct perf_session *perf_session__new(const char *filename, int mode, bool force, bool repipe, struct perf_tool *tool) { - size_t len = filename ? strlen(filename) + 1 : 0; + size_t len = filename ? strlen(filename) : 0; struct perf_session *self = zalloc(sizeof(*self) + len); if (self == NULL) diff --git a/tools/perf/util/session.h b/tools/perf/util/session.h index fb696124ad61..37bc38381fb6 100644 --- a/tools/perf/util/session.h +++ b/tools/perf/util/session.h @@ -50,7 +50,7 @@ struct perf_session { int cwdlen; char *cwd; struct ordered_samples ordered_samples; - char filename[0]; + char filename[1]; }; struct perf_tool; -- cgit v1.2.3 From 1b5495043d5bc058def21f9b66fd8feaa794eb44 Mon Sep 17 00:00:00 2001 From: Robert Richter Date: Wed, 7 Dec 2011 10:02:53 +0100 Subject: perf tools: Moving code in some files Needed for later changes. No modified functionality. Cc: Frederic Weisbecker Cc: Ingo Molnar Cc: Peter Zijlstra Cc: Stephane Eranian Link: http://lkml.kernel.org/r/1323248577-11268-4-git-send-email-robert.richter@amd.com Signed-off-by: Robert Richter Signed-off-by: Arnaldo Carvalho de Melo --- tools/perf/builtin-buildid-list.c | 36 +-- tools/perf/util/header.c | 495 +++++++++++++++++++------------------- 2 files changed, 264 insertions(+), 267 deletions(-) diff --git a/tools/perf/builtin-buildid-list.c b/tools/perf/builtin-buildid-list.c index cb690a65bf02..4895668577b5 100644 --- a/tools/perf/builtin-buildid-list.c +++ b/tools/perf/builtin-buildid-list.c @@ -39,24 +39,6 @@ static const struct option options[] = { OPT_END() }; -static int perf_session__list_build_ids(void) -{ - struct perf_session *session; - - session = perf_session__new(input_name, O_RDONLY, force, false, - &build_id__mark_dso_hit_ops); - if (session == NULL) - return -1; - - if (with_hits) - perf_session__process_events(session, &build_id__mark_dso_hit_ops); - - perf_session__fprintf_dsos_buildid(session, stdout, with_hits); - - perf_session__delete(session); - return 0; -} - static int sysfs__fprintf_build_id(FILE *fp) { u8 kallsyms_build_id[BUILD_ID_SIZE]; @@ -85,6 +67,24 @@ static int filename__fprintf_build_id(const char *name, FILE *fp) return fprintf(fp, "%s\n", sbuild_id); } +static int perf_session__list_build_ids(void) +{ + struct perf_session *session; + + session = perf_session__new(input_name, O_RDONLY, force, false, + &build_id__mark_dso_hit_ops); + if (session == NULL) + return -1; + + if (with_hits) + perf_session__process_events(session, &build_id__mark_dso_hit_ops); + + perf_session__fprintf_dsos_buildid(session, stdout, with_hits); + + perf_session__delete(session); + return 0; +} + static int __cmd_buildid_list(void) { if (show_kernel) diff --git a/tools/perf/util/header.c b/tools/perf/util/header.c index 428a4a2ce350..609d79b5fb5e 100644 --- a/tools/perf/util/header.c +++ b/tools/perf/util/header.c @@ -28,9 +28,6 @@ static struct perf_trace_event_type *events; static u32 header_argc; static const char **header_argv; -static int dsos__write_buildid_table(struct perf_header *header, int fd); -static int perf_session__cache_build_ids(struct perf_session *session); - int perf_header__push_event(u64 id, const char *name) { if (strlen(name) > MAX_EVENT_NAME) @@ -187,6 +184,252 @@ perf_header__set_cmdline(int argc, const char **argv) return 0; } +#define dsos__for_each_with_build_id(pos, head) \ + list_for_each_entry(pos, head, node) \ + if (!pos->has_build_id) \ + continue; \ + else + +static int __dsos__write_buildid_table(struct list_head *head, pid_t pid, + u16 misc, int fd) +{ + struct dso *pos; + + dsos__for_each_with_build_id(pos, head) { + int err; + struct build_id_event b; + size_t len; + + if (!pos->hit) + continue; + len = pos->long_name_len + 1; + len = ALIGN(len, NAME_ALIGN); + memset(&b, 0, sizeof(b)); + memcpy(&b.build_id, pos->build_id, sizeof(pos->build_id)); + b.pid = pid; + b.header.misc = misc; + b.header.size = sizeof(b) + len; + err = do_write(fd, &b, sizeof(b)); + if (err < 0) + return err; + err = write_padded(fd, pos->long_name, + pos->long_name_len + 1, len); + if (err < 0) + return err; + } + + return 0; +} + +static int machine__write_buildid_table(struct machine *machine, int fd) +{ + int err; + u16 kmisc = PERF_RECORD_MISC_KERNEL, + umisc = PERF_RECORD_MISC_USER; + + if (!machine__is_host(machine)) { + kmisc = PERF_RECORD_MISC_GUEST_KERNEL; + umisc = PERF_RECORD_MISC_GUEST_USER; + } + + err = __dsos__write_buildid_table(&machine->kernel_dsos, machine->pid, + kmisc, fd); + if (err == 0) + err = __dsos__write_buildid_table(&machine->user_dsos, + machine->pid, umisc, fd); + return err; +} + +static int dsos__write_buildid_table(struct perf_header *header, int fd) +{ + struct perf_session *session = container_of(header, + struct perf_session, header); + struct rb_node *nd; + int err = machine__write_buildid_table(&session->host_machine, fd); + + if (err) + return err; + + for (nd = rb_first(&session->machines); nd; nd = rb_next(nd)) { + struct machine *pos = rb_entry(nd, struct machine, rb_node); + err = machine__write_buildid_table(pos, fd); + if (err) + break; + } + return err; +} + +int build_id_cache__add_s(const char *sbuild_id, const char *debugdir, + const char *name, bool is_kallsyms) +{ + const size_t size = PATH_MAX; + char *realname, *filename = zalloc(size), + *linkname = zalloc(size), *targetname; + int len, err = -1; + + if (is_kallsyms) { + if (symbol_conf.kptr_restrict) { + pr_debug("Not caching a kptr_restrict'ed /proc/kallsyms\n"); + return 0; + } + realname = (char *)name; + } else + realname = realpath(name, NULL); + + if (realname == NULL || filename == NULL || linkname == NULL) + goto out_free; + + len = snprintf(filename, size, "%s%s%s", + debugdir, is_kallsyms ? "/" : "", realname); + if (mkdir_p(filename, 0755)) + goto out_free; + + snprintf(filename + len, sizeof(filename) - len, "/%s", sbuild_id); + + if (access(filename, F_OK)) { + if (is_kallsyms) { + if (copyfile("/proc/kallsyms", filename)) + goto out_free; + } else if (link(realname, filename) && copyfile(name, filename)) + goto out_free; + } + + len = snprintf(linkname, size, "%s/.build-id/%.2s", + debugdir, sbuild_id); + + if (access(linkname, X_OK) && mkdir_p(linkname, 0755)) + goto out_free; + + snprintf(linkname + len, size - len, "/%s", sbuild_id + 2); + targetname = filename + strlen(debugdir) - 5; + memcpy(targetname, "../..", 5); + + if (symlink(targetname, linkname) == 0) + err = 0; +out_free: + if (!is_kallsyms) + free(realname); + free(filename); + free(linkname); + return err; +} + +static int build_id_cache__add_b(const u8 *build_id, size_t build_id_size, + const char *name, const char *debugdir, + bool is_kallsyms) +{ + char sbuild_id[BUILD_ID_SIZE * 2 + 1]; + + build_id__sprintf(build_id, build_id_size, sbuild_id); + + return build_id_cache__add_s(sbuild_id, debugdir, name, is_kallsyms); +} + +int build_id_cache__remove_s(const char *sbuild_id, const char *debugdir) +{ + const size_t size = PATH_MAX; + char *filename = zalloc(size), + *linkname = zalloc(size); + int err = -1; + + if (filename == NULL || linkname == NULL) + goto out_free; + + snprintf(linkname, size, "%s/.build-id/%.2s/%s", + debugdir, sbuild_id, sbuild_id + 2); + + if (access(linkname, F_OK)) + goto out_free; + + if (readlink(linkname, filename, size - 1) < 0) + goto out_free; + + if (unlink(linkname)) + goto out_free; + + /* + * Since the link is relative, we must make it absolute: + */ + snprintf(linkname, size, "%s/.build-id/%.2s/%s", + debugdir, sbuild_id, filename); + + if (unlink(linkname)) + goto out_free; + + err = 0; +out_free: + free(filename); + free(linkname); + return err; +} + +static int dso__cache_build_id(struct dso *dso, const char *debugdir) +{ + bool is_kallsyms = dso->kernel && dso->long_name[0] != '/'; + + return build_id_cache__add_b(dso->build_id, sizeof(dso->build_id), + dso->long_name, debugdir, is_kallsyms); +} + +static int __dsos__cache_build_ids(struct list_head *head, const char *debugdir) +{ + struct dso *pos; + int err = 0; + + dsos__for_each_with_build_id(pos, head) + if (dso__cache_build_id(pos, debugdir)) + err = -1; + + return err; +} + +static int machine__cache_build_ids(struct machine *machine, const char *debugdir) +{ + int ret = __dsos__cache_build_ids(&machine->kernel_dsos, debugdir); + ret |= __dsos__cache_build_ids(&machine->user_dsos, debugdir); + return ret; +} + +static int perf_session__cache_build_ids(struct perf_session *session) +{ + struct rb_node *nd; + int ret; + char debugdir[PATH_MAX]; + + snprintf(debugdir, sizeof(debugdir), "%s", buildid_dir); + + if (mkdir(debugdir, 0755) != 0 && errno != EEXIST) + return -1; + + ret = machine__cache_build_ids(&session->host_machine, debugdir); + + for (nd = rb_first(&session->machines); nd; nd = rb_next(nd)) { + struct machine *pos = rb_entry(nd, struct machine, rb_node); + ret |= machine__cache_build_ids(pos, debugdir); + } + return ret ? -1 : 0; +} + +static bool machine__read_build_ids(struct machine *machine, bool with_hits) +{ + bool ret = __dsos__read_build_ids(&machine->kernel_dsos, with_hits); + ret |= __dsos__read_build_ids(&machine->user_dsos, with_hits); + return ret; +} + +static bool perf_session__read_build_ids(struct perf_session *session, bool with_hits) +{ + struct rb_node *nd; + bool ret = machine__read_build_ids(&session->host_machine, with_hits); + + for (nd = rb_first(&session->machines); nd; nd = rb_next(nd)) { + struct machine *pos = rb_entry(nd, struct machine, rb_node); + ret |= machine__read_build_ids(pos, with_hits); + } + + return ret; +} + static int write_trace_info(int fd, struct perf_header *h __used, struct perf_evlist *evlist) { @@ -1136,252 +1379,6 @@ int perf_header__fprintf_info(struct perf_session *session, FILE *fp, bool full) return 0; } -#define dsos__for_each_with_build_id(pos, head) \ - list_for_each_entry(pos, head, node) \ - if (!pos->has_build_id) \ - continue; \ - else - -static int __dsos__write_buildid_table(struct list_head *head, pid_t pid, - u16 misc, int fd) -{ - struct dso *pos; - - dsos__for_each_with_build_id(pos, head) { - int err; - struct build_id_event b; - size_t len; - - if (!pos->hit) - continue; - len = pos->long_name_len + 1; - len = ALIGN(len, NAME_ALIGN); - memset(&b, 0, sizeof(b)); - memcpy(&b.build_id, pos->build_id, sizeof(pos->build_id)); - b.pid = pid; - b.header.misc = misc; - b.header.size = sizeof(b) + len; - err = do_write(fd, &b, sizeof(b)); - if (err < 0) - return err; - err = write_padded(fd, pos->long_name, - pos->long_name_len + 1, len); - if (err < 0) - return err; - } - - return 0; -} - -static int machine__write_buildid_table(struct machine *machine, int fd) -{ - int err; - u16 kmisc = PERF_RECORD_MISC_KERNEL, - umisc = PERF_RECORD_MISC_USER; - - if (!machine__is_host(machine)) { - kmisc = PERF_RECORD_MISC_GUEST_KERNEL; - umisc = PERF_RECORD_MISC_GUEST_USER; - } - - err = __dsos__write_buildid_table(&machine->kernel_dsos, machine->pid, - kmisc, fd); - if (err == 0) - err = __dsos__write_buildid_table(&machine->user_dsos, - machine->pid, umisc, fd); - return err; -} - -static int dsos__write_buildid_table(struct perf_header *header, int fd) -{ - struct perf_session *session = container_of(header, - struct perf_session, header); - struct rb_node *nd; - int err = machine__write_buildid_table(&session->host_machine, fd); - - if (err) - return err; - - for (nd = rb_first(&session->machines); nd; nd = rb_next(nd)) { - struct machine *pos = rb_entry(nd, struct machine, rb_node); - err = machine__write_buildid_table(pos, fd); - if (err) - break; - } - return err; -} - -int build_id_cache__add_s(const char *sbuild_id, const char *debugdir, - const char *name, bool is_kallsyms) -{ - const size_t size = PATH_MAX; - char *realname, *filename = zalloc(size), - *linkname = zalloc(size), *targetname; - int len, err = -1; - - if (is_kallsyms) { - if (symbol_conf.kptr_restrict) { - pr_debug("Not caching a kptr_restrict'ed /proc/kallsyms\n"); - return 0; - } - realname = (char *)name; - } else - realname = realpath(name, NULL); - - if (realname == NULL || filename == NULL || linkname == NULL) - goto out_free; - - len = snprintf(filename, size, "%s%s%s", - debugdir, is_kallsyms ? "/" : "", realname); - if (mkdir_p(filename, 0755)) - goto out_free; - - snprintf(filename + len, sizeof(filename) - len, "/%s", sbuild_id); - - if (access(filename, F_OK)) { - if (is_kallsyms) { - if (copyfile("/proc/kallsyms", filename)) - goto out_free; - } else if (link(realname, filename) && copyfile(name, filename)) - goto out_free; - } - - len = snprintf(linkname, size, "%s/.build-id/%.2s", - debugdir, sbuild_id); - - if (access(linkname, X_OK) && mkdir_p(linkname, 0755)) - goto out_free; - - snprintf(linkname + len, size - len, "/%s", sbuild_id + 2); - targetname = filename + strlen(debugdir) - 5; - memcpy(targetname, "../..", 5); - - if (symlink(targetname, linkname) == 0) - err = 0; -out_free: - if (!is_kallsyms) - free(realname); - free(filename); - free(linkname); - return err; -} - -static int build_id_cache__add_b(const u8 *build_id, size_t build_id_size, - const char *name, const char *debugdir, - bool is_kallsyms) -{ - char sbuild_id[BUILD_ID_SIZE * 2 + 1]; - - build_id__sprintf(build_id, build_id_size, sbuild_id); - - return build_id_cache__add_s(sbuild_id, debugdir, name, is_kallsyms); -} - -int build_id_cache__remove_s(const char *sbuild_id, const char *debugdir) -{ - const size_t size = PATH_MAX; - char *filename = zalloc(size), - *linkname = zalloc(size); - int err = -1; - - if (filename == NULL || linkname == NULL) - goto out_free; - - snprintf(linkname, size, "%s/.build-id/%.2s/%s", - debugdir, sbuild_id, sbuild_id + 2); - - if (access(linkname, F_OK)) - goto out_free; - - if (readlink(linkname, filename, size - 1) < 0) - goto out_free; - - if (unlink(linkname)) - goto out_free; - - /* - * Since the link is relative, we must make it absolute: - */ - snprintf(linkname, size, "%s/.build-id/%.2s/%s", - debugdir, sbuild_id, filename); - - if (unlink(linkname)) - goto out_free; - - err = 0; -out_free: - free(filename); - free(linkname); - return err; -} - -static int dso__cache_build_id(struct dso *dso, const char *debugdir) -{ - bool is_kallsyms = dso->kernel && dso->long_name[0] != '/'; - - return build_id_cache__add_b(dso->build_id, sizeof(dso->build_id), - dso->long_name, debugdir, is_kallsyms); -} - -static int __dsos__cache_build_ids(struct list_head *head, const char *debugdir) -{ - struct dso *pos; - int err = 0; - - dsos__for_each_with_build_id(pos, head) - if (dso__cache_build_id(pos, debugdir)) - err = -1; - - return err; -} - -static int machine__cache_build_ids(struct machine *machine, const char *debugdir) -{ - int ret = __dsos__cache_build_ids(&machine->kernel_dsos, debugdir); - ret |= __dsos__cache_build_ids(&machine->user_dsos, debugdir); - return ret; -} - -static int perf_session__cache_build_ids(struct perf_session *session) -{ - struct rb_node *nd; - int ret; - char debugdir[PATH_MAX]; - - snprintf(debugdir, sizeof(debugdir), "%s", buildid_dir); - - if (mkdir(debugdir, 0755) != 0 && errno != EEXIST) - return -1; - - ret = machine__cache_build_ids(&session->host_machine, debugdir); - - for (nd = rb_first(&session->machines); nd; nd = rb_next(nd)) { - struct machine *pos = rb_entry(nd, struct machine, rb_node); - ret |= machine__cache_build_ids(pos, debugdir); - } - return ret ? -1 : 0; -} - -static bool machine__read_build_ids(struct machine *machine, bool with_hits) -{ - bool ret = __dsos__read_build_ids(&machine->kernel_dsos, with_hits); - ret |= __dsos__read_build_ids(&machine->user_dsos, with_hits); - return ret; -} - -static bool perf_session__read_build_ids(struct perf_session *session, bool with_hits) -{ - struct rb_node *nd; - bool ret = machine__read_build_ids(&session->host_machine, with_hits); - - for (nd = rb_first(&session->machines); nd; nd = rb_next(nd)) { - struct machine *pos = rb_entry(nd, struct machine, rb_node); - ret |= machine__read_build_ids(pos, with_hits); - } - - return ret; -} - static int do_write_feat(int fd, struct perf_header *h, int type, struct perf_file_section **p, struct perf_evlist *evlist) -- cgit v1.2.3 From efad14150a0b4429f37da7245001a8096ef7ee38 Mon Sep 17 00:00:00 2001 From: Robert Richter Date: Wed, 7 Dec 2011 10:02:54 +0100 Subject: perf report: Accept fifos as input file The default input file for perf report is not handled the same way as perf record does it for its output file. This leads to unexpected behavior of perf report, etc. E.g.: # perf record -a -e cpu-cycles sleep 2 | perf report | cat failed to open perf.data: No such file or directory (try 'perf record' first) While perf record writes to a fifo, perf report expects perf.data to be read. This patch changes this to accept fifos as input file. Applies to the following commands: perf annotate perf buildid-list perf evlist perf kmem perf lock perf report perf sched perf script perf timechart Also fixes char const* -> const char* type declaration for filename strings. v2: * Prevent potential null pointer access to input_name in builtin-report.c. Needed due to removal of patch "perf report: Setup browser if stdout is a pipe" Cc: Frederic Weisbecker Cc: Ingo Molnar Cc: Peter Zijlstra Cc: Stephane Eranian Link: http://lkml.kernel.org/r/1323248577-11268-5-git-send-email-robert.richter@amd.com Signed-off-by: Robert Richter Signed-off-by: Arnaldo Carvalho de Melo --- tools/perf/Documentation/perf-annotate.txt | 2 +- tools/perf/Documentation/perf-buildid-list.txt | 2 +- tools/perf/Documentation/perf-evlist.txt | 2 +- tools/perf/Documentation/perf-kmem.txt | 2 +- tools/perf/Documentation/perf-lock.txt | 2 +- tools/perf/Documentation/perf-report.txt | 2 +- tools/perf/Documentation/perf-sched.txt | 2 +- tools/perf/Documentation/perf-script.txt | 2 +- tools/perf/Documentation/perf-timechart.txt | 2 +- tools/perf/builtin-annotate.c | 3 +-- tools/perf/builtin-buildid-list.c | 19 ++++++++++--------- tools/perf/builtin-evlist.c | 2 +- tools/perf/builtin-kmem.c | 2 +- tools/perf/builtin-lock.c | 2 +- tools/perf/builtin-report.c | 13 ++++++++++--- tools/perf/builtin-sched.c | 2 +- tools/perf/builtin-script.c | 4 ++-- tools/perf/builtin-timechart.c | 4 ++-- tools/perf/util/session.c | 15 +++++++++++++-- 19 files changed, 51 insertions(+), 33 deletions(-) diff --git a/tools/perf/Documentation/perf-annotate.txt b/tools/perf/Documentation/perf-annotate.txt index 476029d30621..c89f9e1453f7 100644 --- a/tools/perf/Documentation/perf-annotate.txt +++ b/tools/perf/Documentation/perf-annotate.txt @@ -22,7 +22,7 @@ OPTIONS ------- -i:: --input=:: - Input file name. (default: perf.data) + Input file name. (default: perf.data unless stdin is a fifo) -d:: --dsos=:: diff --git a/tools/perf/Documentation/perf-buildid-list.txt b/tools/perf/Documentation/perf-buildid-list.txt index cc22325ffd1b..25c52efcc7f0 100644 --- a/tools/perf/Documentation/perf-buildid-list.txt +++ b/tools/perf/Documentation/perf-buildid-list.txt @@ -26,7 +26,7 @@ OPTIONS Show only DSOs with hits. -i:: --input=:: - Input file name. (default: perf.data) + Input file name. (default: perf.data unless stdin is a fifo) -f:: --force:: Don't do ownership validation. diff --git a/tools/perf/Documentation/perf-evlist.txt b/tools/perf/Documentation/perf-evlist.txt index 0cada9e053dc..0507ec7bad71 100644 --- a/tools/perf/Documentation/perf-evlist.txt +++ b/tools/perf/Documentation/perf-evlist.txt @@ -18,7 +18,7 @@ OPTIONS ------- -i:: --input=:: - Input file name. (default: perf.data) + Input file name. (default: perf.data unless stdin is a fifo) SEE ALSO -------- diff --git a/tools/perf/Documentation/perf-kmem.txt b/tools/perf/Documentation/perf-kmem.txt index a52fcde894c7..7c8fbbf3f61c 100644 --- a/tools/perf/Documentation/perf-kmem.txt +++ b/tools/perf/Documentation/perf-kmem.txt @@ -23,7 +23,7 @@ OPTIONS ------- -i :: --input=:: - Select the input file (default: perf.data) + Select the input file (default: perf.data unless stdin is a fifo) --caller:: Show per-callsite statistics diff --git a/tools/perf/Documentation/perf-lock.txt b/tools/perf/Documentation/perf-lock.txt index 4a26a2f3a6a3..d6b2a4f2108b 100644 --- a/tools/perf/Documentation/perf-lock.txt +++ b/tools/perf/Documentation/perf-lock.txt @@ -29,7 +29,7 @@ COMMON OPTIONS -i:: --input=:: - Input file name. + Input file name. (default: perf.data unless stdin is a fifo) -v:: --verbose:: diff --git a/tools/perf/Documentation/perf-report.txt b/tools/perf/Documentation/perf-report.txt index 35af0dc8ccb4..9b430e98712e 100644 --- a/tools/perf/Documentation/perf-report.txt +++ b/tools/perf/Documentation/perf-report.txt @@ -19,7 +19,7 @@ OPTIONS ------- -i:: --input=:: - Input file name. (default: perf.data) + Input file name. (default: perf.data unless stdin is a fifo) -v:: --verbose:: diff --git a/tools/perf/Documentation/perf-sched.txt b/tools/perf/Documentation/perf-sched.txt index 5b212b57f70b..8ff4df956951 100644 --- a/tools/perf/Documentation/perf-sched.txt +++ b/tools/perf/Documentation/perf-sched.txt @@ -40,7 +40,7 @@ OPTIONS ------- -i:: --input=:: - Input file name. (default: perf.data) + Input file name. (default: perf.data unless stdin is a fifo) -v:: --verbose:: diff --git a/tools/perf/Documentation/perf-script.txt b/tools/perf/Documentation/perf-script.txt index 7f61eaaf9ab8..2f6cef43da25 100644 --- a/tools/perf/Documentation/perf-script.txt +++ b/tools/perf/Documentation/perf-script.txt @@ -106,7 +106,7 @@ OPTIONS -i:: --input=:: - Input file name. + Input file name. (default: perf.data unless stdin is a fifo) -d:: --debug-mode:: diff --git a/tools/perf/Documentation/perf-timechart.txt b/tools/perf/Documentation/perf-timechart.txt index d7b79e2ba2ad..1632b0efc757 100644 --- a/tools/perf/Documentation/perf-timechart.txt +++ b/tools/perf/Documentation/perf-timechart.txt @@ -27,7 +27,7 @@ OPTIONS Select the output file (default: output.svg) -i:: --input=:: - Select the input file (default: perf.data) + Select the input file (default: perf.data unless stdin is a fifo) -w:: --width=:: Select the width of the SVG file (default: 1000) diff --git a/tools/perf/builtin-annotate.c b/tools/perf/builtin-annotate.c index d449645c5ef1..214ba7f9f577 100644 --- a/tools/perf/builtin-annotate.c +++ b/tools/perf/builtin-annotate.c @@ -215,7 +215,7 @@ static int __cmd_annotate(struct perf_annotate *ann) } if (total_nr_samples == 0) { - ui__warning("The %s file has no samples!\n", ann->input_name); + ui__warning("The %s file has no samples!\n", session->filename); goto out_delete; } out_delete: @@ -250,7 +250,6 @@ int cmd_annotate(int argc, const char **argv, const char *prefix __used) .ordered_samples = true, .ordering_requires_timestamps = true, }, - .input_name = "perf.data", }; const struct option options[] = { OPT_STRING('i', "input", &annotate.input_name, "file", diff --git a/tools/perf/builtin-buildid-list.c b/tools/perf/builtin-buildid-list.c index 4895668577b5..52480467e9ff 100644 --- a/tools/perf/builtin-buildid-list.c +++ b/tools/perf/builtin-buildid-list.c @@ -18,7 +18,7 @@ #include -static char const *input_name = "perf.data"; +static const char *input_name; static bool force; static bool show_kernel; static bool with_hits; @@ -71,16 +71,24 @@ static int perf_session__list_build_ids(void) { struct perf_session *session; + elf_version(EV_CURRENT); + session = perf_session__new(input_name, O_RDONLY, force, false, &build_id__mark_dso_hit_ops); if (session == NULL) return -1; + /* + * See if this is an ELF file first: + */ + if (filename__fprintf_build_id(session->filename, stdout)) + goto out; + if (with_hits) perf_session__process_events(session, &build_id__mark_dso_hit_ops); perf_session__fprintf_dsos_buildid(session, stdout, with_hits); - +out: perf_session__delete(session); return 0; } @@ -90,13 +98,6 @@ static int __cmd_buildid_list(void) if (show_kernel) return sysfs__fprintf_build_id(stdout); - elf_version(EV_CURRENT); - /* - * See if this is an ELF file first: - */ - if (filename__fprintf_build_id(input_name, stdout)) - return 0; - return perf_session__list_build_ids(); } diff --git a/tools/perf/builtin-evlist.c b/tools/perf/builtin-evlist.c index 4c5e9e04a41f..26760322c4f4 100644 --- a/tools/perf/builtin-evlist.c +++ b/tools/perf/builtin-evlist.c @@ -15,7 +15,7 @@ #include "util/parse-options.h" #include "util/session.h" -static char const *input_name = "perf.data"; +static const char *input_name; static int __cmd_evlist(void) { diff --git a/tools/perf/builtin-kmem.c b/tools/perf/builtin-kmem.c index 886174e9525b..fe1ad8f21961 100644 --- a/tools/perf/builtin-kmem.c +++ b/tools/perf/builtin-kmem.c @@ -19,7 +19,7 @@ struct alloc_stat; typedef int (*sort_fn_t)(struct alloc_stat *, struct alloc_stat *); -static char const *input_name = "perf.data"; +static const char *input_name; static int alloc_flag; static int caller_flag; diff --git a/tools/perf/builtin-lock.c b/tools/perf/builtin-lock.c index 4db5e5293067..2296c391d0f5 100644 --- a/tools/perf/builtin-lock.c +++ b/tools/perf/builtin-lock.c @@ -326,7 +326,7 @@ alloc_failed: die("memory allocation failed\n"); } -static char const *input_name = "perf.data"; +static const char *input_name; struct raw_event_sample { u32 size; diff --git a/tools/perf/builtin-report.c b/tools/perf/builtin-report.c index 9051f6bfaa7e..25d34d483e49 100644 --- a/tools/perf/builtin-report.c +++ b/tools/perf/builtin-report.c @@ -321,8 +321,7 @@ static int __cmd_report(struct perf_report *rep) } if (nr_samples == 0) { - ui__warning("The %s file has no samples!\n", - rep->input_name); + ui__warning("The %s file has no samples!\n", session->filename); goto out_delete; } @@ -430,6 +429,7 @@ setup: int cmd_report(int argc, const char **argv, const char *prefix __used) { + struct stat st; char callchain_default_opt[] = "fractal,0.5,callee"; const char * const report_usage[] = { "perf report []", @@ -451,7 +451,6 @@ int cmd_report(int argc, const char **argv, const char *prefix __used) .ordered_samples = true, .ordering_requires_timestamps = true, }, - .input_name = "perf.data", .pretty_printing_style = "normal", }; const struct option options[] = { @@ -531,10 +530,18 @@ int cmd_report(int argc, const char **argv, const char *prefix __used) if (report.inverted_callchain) callchain_param.order = ORDER_CALLER; + if (!report.input_name || !strlen(report.input_name)) { + if (!fstat(STDIN_FILENO, &st) && S_ISFIFO(st.st_mode)) + report.input_name = "-"; + else + report.input_name = "perf.data"; + } + if (strcmp(report.input_name, "-") != 0) setup_browser(true); else use_browser = 0; + /* * Only in the newt browser we are doing integrated annotation, * so don't allocate extra space that won't be used in the stdio diff --git a/tools/perf/builtin-sched.c b/tools/perf/builtin-sched.c index 6284ed2317f2..fb8b5f83b4a0 100644 --- a/tools/perf/builtin-sched.c +++ b/tools/perf/builtin-sched.c @@ -22,7 +22,7 @@ #include #include -static char const *input_name = "perf.data"; +static const char *input_name; static char default_sort_order[] = "avg, max, switch, runtime"; static const char *sort_order = default_sort_order; diff --git a/tools/perf/builtin-script.c b/tools/perf/builtin-script.c index d71b745da06e..3d4c0c7b576e 100644 --- a/tools/perf/builtin-script.c +++ b/tools/perf/builtin-script.c @@ -434,7 +434,7 @@ static int cleanup_scripting(void) return scripting_ops->stop_script(); } -static char const *input_name = "perf.data"; +static const char *input_name; static int process_sample_event(struct perf_tool *tool __used, union perf_event *event, @@ -1316,7 +1316,7 @@ int cmd_script(int argc, const char **argv, const char *prefix __used) return -1; } - input = open(input_name, O_RDONLY); + input = open(session->filename, O_RDONLY); /* input_name */ if (input < 0) { perror("failed to open file"); exit(-1); diff --git a/tools/perf/builtin-timechart.c b/tools/perf/builtin-timechart.c index 135376a37f97..3b75b2e21ea5 100644 --- a/tools/perf/builtin-timechart.c +++ b/tools/perf/builtin-timechart.c @@ -38,8 +38,8 @@ #define PWR_EVENT_EXIT -1 -static char const *input_name = "perf.data"; -static char const *output_name = "output.svg"; +static const char *input_name; +static const char *output_name = "output.svg"; static unsigned int numcpus; static u64 min_freq; /* Lowest CPU frequency seen */ diff --git a/tools/perf/util/session.c b/tools/perf/util/session.c index ea17dfb85baa..cc5e6be46d86 100644 --- a/tools/perf/util/session.c +++ b/tools/perf/util/session.c @@ -107,8 +107,19 @@ struct perf_session *perf_session__new(const char *filename, int mode, bool force, bool repipe, struct perf_tool *tool) { - size_t len = filename ? strlen(filename) : 0; - struct perf_session *self = zalloc(sizeof(*self) + len); + struct perf_session *self; + struct stat st; + size_t len; + + if (!filename || !strlen(filename)) { + if (!fstat(STDIN_FILENO, &st) && S_ISFIFO(st.st_mode)) + filename = "-"; + else + filename = "perf.data"; + } + + len = strlen(filename); + self = zalloc(sizeof(*self) + len); if (self == NULL) goto out; -- cgit v1.2.3 From e20960c0271f91aead94746872fd976326a703b3 Mon Sep 17 00:00:00 2001 From: Robert Richter Date: Wed, 7 Dec 2011 10:02:55 +0100 Subject: perf tools: Unify handling of features when writing feature section The features HEADER_TRACE_INFO and HEADER_BUILD_ID are handled different when writing the feature section. All other features are simply disabled on failure and writing the section goes on without returning an error. There is no reason for these special cases. This patch unifies handling of the features. This should be ok since all features can be parsed independently. Offset and size of a feature's block is stored in struct perf_file_ section right after the data block of perf.data (see perf_session__ write_header()). Thus, if a feature does not exist then other features can be processed anyway. Also moving special code for HEADER_BUILD_ID out to write_build_id(). v2: * perf record throws an error now if buildids may not be generated, which can be disabled with the --no-buildid option. Cc: Frederic Weisbecker Cc: Ingo Molnar Cc: Peter Zijlstra Cc: Stephane Eranian Link: http://lkml.kernel.org/r/1323248577-11268-6-git-send-email-robert.richter@amd.com Signed-off-by: Robert Richter Signed-off-by: Arnaldo Carvalho de Melo --- tools/perf/builtin-record.c | 7 +++++++ tools/perf/util/header.c | 14 +++++--------- 2 files changed, 12 insertions(+), 9 deletions(-) diff --git a/tools/perf/builtin-record.c b/tools/perf/builtin-record.c index e873ae2dd54c..0abfb18b911f 100644 --- a/tools/perf/builtin-record.c +++ b/tools/perf/builtin-record.c @@ -503,6 +503,13 @@ static int __cmd_record(struct perf_record *rec, int argc, const char **argv) return err; } + if (!!rec->no_buildid + && !perf_header__has_feat(&session->header, HEADER_BUILD_ID)) { + pr_err("Couldn't generating buildids. " + "Use --no-buildid to profile anyway.\n"); + return -1; + } + rec->post_processing_offset = lseek(output, 0, SEEK_CUR); machine = perf_session__find_host_machine(session); diff --git a/tools/perf/util/header.c b/tools/perf/util/header.c index 609d79b5fb5e..71326836921b 100644 --- a/tools/perf/util/header.c +++ b/tools/perf/util/header.c @@ -445,6 +445,9 @@ static int write_build_id(int fd, struct perf_header *h, session = container_of(h, struct perf_session, header); + if (!perf_session__read_build_ids(session, true)) + return -1; + err = dsos__write_buildid_table(h, fd); if (err < 0) { pr_debug("failed to write buildid table\n"); @@ -1417,10 +1420,6 @@ static int perf_header__adds_write(struct perf_header *header, session = container_of(header, struct perf_session, header); - if (perf_header__has_feat(header, HEADER_BUILD_ID && - !perf_session__read_build_ids(session, true))) - perf_header__clear_feat(header, HEADER_BUILD_ID); - nr_sections = bitmap_weight(header->adds_features, HEADER_FEAT_BITS); if (!nr_sections) return 0; @@ -1436,13 +1435,11 @@ static int perf_header__adds_write(struct perf_header *header, err = do_write_feat(fd, header, HEADER_TRACE_INFO, &p, evlist); if (err) - goto out_free; + perf_header__clear_feat(header, HEADER_TRACE_INFO); err = do_write_feat(fd, header, HEADER_BUILD_ID, &p, evlist); - if (err) { + if (err) perf_header__clear_feat(header, HEADER_BUILD_ID); - goto out_free; - } err = do_write_feat(fd, header, HEADER_HOSTNAME, &p, evlist); if (err) @@ -1500,7 +1497,6 @@ static int perf_header__adds_write(struct perf_header *header, err = do_write(fd, feat_sec, sec_size); if (err < 0) pr_debug("failed to write feature section\n"); -out_free: free(feat_sec); return err; } -- cgit v1.2.3 From b1e5a9bee3c342dd3281aef76d1be1044dd8addf Mon Sep 17 00:00:00 2001 From: Robert Richter Date: Wed, 7 Dec 2011 10:02:57 +0100 Subject: perf tools: Use for_each_set_bit() to iterate over feature flags This patch introduces the for_each_set_bit() macro and modifies feature implementation to use it. Cc: Frederic Weisbecker Cc: Ingo Molnar Cc: Peter Zijlstra Cc: Stephane Eranian Link: http://lkml.kernel.org/r/1323248577-11268-8-git-send-email-robert.richter@amd.com Signed-off-by: Robert Richter Signed-off-by: Arnaldo Carvalho de Melo --- tools/perf/util/header.c | 118 ++++++++------------------------- tools/perf/util/header.h | 6 +- tools/perf/util/include/linux/bitops.h | 118 +++++++++++++++++++++++++++++++++ 3 files changed, 149 insertions(+), 93 deletions(-) diff --git a/tools/perf/util/header.c b/tools/perf/util/header.c index 71326836921b..e509a9dea00b 100644 --- a/tools/perf/util/header.c +++ b/tools/perf/util/header.c @@ -8,6 +8,7 @@ #include #include #include +#include #include #include "evlist.h" @@ -1353,7 +1354,7 @@ static int perf_file_section__fprintf_info(struct perf_file_section *section, "%d, continuing...\n", section->offset, feat); return 0; } - if (feat < HEADER_TRACE_INFO || feat >= HEADER_LAST_FEATURE) { + if (feat >= HEADER_LAST_FEATURE) { pr_warning("unknown feature %d\n", feat); return 0; } @@ -1390,6 +1391,8 @@ static int do_write_feat(int fd, struct perf_header *h, int type, int ret = 0; if (perf_header__has_feat(h, type)) { + if (!feat_ops[type].write) + return -1; (*p)->offset = lseek(fd, 0, SEEK_CUR); @@ -1416,6 +1419,7 @@ static int perf_header__adds_write(struct perf_header *header, struct perf_file_section *feat_sec, *p; int sec_size; u64 sec_start; + int feat; int err; session = container_of(header, struct perf_session, header); @@ -1433,61 +1437,10 @@ static int perf_header__adds_write(struct perf_header *header, sec_start = header->data_offset + header->data_size; lseek(fd, sec_start + sec_size, SEEK_SET); - err = do_write_feat(fd, header, HEADER_TRACE_INFO, &p, evlist); - if (err) - perf_header__clear_feat(header, HEADER_TRACE_INFO); - - err = do_write_feat(fd, header, HEADER_BUILD_ID, &p, evlist); - if (err) - perf_header__clear_feat(header, HEADER_BUILD_ID); - - err = do_write_feat(fd, header, HEADER_HOSTNAME, &p, evlist); - if (err) - perf_header__clear_feat(header, HEADER_HOSTNAME); - - err = do_write_feat(fd, header, HEADER_OSRELEASE, &p, evlist); - if (err) - perf_header__clear_feat(header, HEADER_OSRELEASE); - - err = do_write_feat(fd, header, HEADER_VERSION, &p, evlist); - if (err) - perf_header__clear_feat(header, HEADER_VERSION); - - err = do_write_feat(fd, header, HEADER_ARCH, &p, evlist); - if (err) - perf_header__clear_feat(header, HEADER_ARCH); - - err = do_write_feat(fd, header, HEADER_NRCPUS, &p, evlist); - if (err) - perf_header__clear_feat(header, HEADER_NRCPUS); - - err = do_write_feat(fd, header, HEADER_CPUDESC, &p, evlist); - if (err) - perf_header__clear_feat(header, HEADER_CPUDESC); - - err = do_write_feat(fd, header, HEADER_CPUID, &p, evlist); - if (err) - perf_header__clear_feat(header, HEADER_CPUID); - - err = do_write_feat(fd, header, HEADER_TOTAL_MEM, &p, evlist); - if (err) - perf_header__clear_feat(header, HEADER_TOTAL_MEM); - - err = do_write_feat(fd, header, HEADER_CMDLINE, &p, evlist); - if (err) - perf_header__clear_feat(header, HEADER_CMDLINE); - - err = do_write_feat(fd, header, HEADER_EVENT_DESC, &p, evlist); - if (err) - perf_header__clear_feat(header, HEADER_EVENT_DESC); - - err = do_write_feat(fd, header, HEADER_CPU_TOPOLOGY, &p, evlist); - if (err) - perf_header__clear_feat(header, HEADER_CPU_TOPOLOGY); - - err = do_write_feat(fd, header, HEADER_NUMA_TOPOLOGY, &p, evlist); - if (err) - perf_header__clear_feat(header, HEADER_NUMA_TOPOLOGY); + for_each_set_bit(feat, header->adds_features, HEADER_FEAT_BITS) { + if (do_write_feat(fd, header, feat, &p, evlist)) + perf_header__clear_feat(header, feat); + } lseek(fd, sec_start, SEEK_SET); /* @@ -1634,20 +1587,20 @@ static int perf_header__getbuffer64(struct perf_header *header, int perf_header__process_sections(struct perf_header *header, int fd, void *data, int (*process)(struct perf_file_section *section, - struct perf_header *ph, - int feat, int fd, void *data)) + struct perf_header *ph, + int feat, int fd, void *data)) { - struct perf_file_section *feat_sec; + struct perf_file_section *feat_sec, *sec; int nr_sections; int sec_size; - int idx = 0; - int err = -1, feat = 1; + int feat; + int err; nr_sections = bitmap_weight(header->adds_features, HEADER_FEAT_BITS); if (!nr_sections) return 0; - feat_sec = calloc(sizeof(*feat_sec), nr_sections); + feat_sec = sec = calloc(sizeof(*feat_sec), nr_sections); if (!feat_sec) return -1; @@ -1655,20 +1608,16 @@ int perf_header__process_sections(struct perf_header *header, int fd, lseek(fd, header->data_offset + header->data_size, SEEK_SET); - if (perf_header__getbuffer64(header, fd, feat_sec, sec_size)) + err = perf_header__getbuffer64(header, fd, feat_sec, sec_size); + if (err < 0) goto out_free; - err = 0; - while (idx < nr_sections && feat < HEADER_LAST_FEATURE) { - if (perf_header__has_feat(header, feat)) { - struct perf_file_section *sec = &feat_sec[idx++]; - - err = process(sec, header, feat, fd, data); - if (err < 0) - break; - } - ++feat; + for_each_set_bit(feat, header->adds_features, HEADER_LAST_FEATURE) { + err = process(sec++, header, feat, fd, data); + if (err < 0) + goto out_free; } + err = 0; out_free: free(feat_sec); return err; @@ -1903,32 +1852,21 @@ static int perf_file_section__process(struct perf_file_section *section, return 0; } + if (feat >= HEADER_LAST_FEATURE) { + pr_debug("unknown feature %d, continuing...\n", feat); + return 0; + } + switch (feat) { case HEADER_TRACE_INFO: trace_report(fd, false); break; - case HEADER_BUILD_ID: if (perf_header__read_build_ids(ph, fd, section->offset, section->size)) pr_debug("Failed to read buildids, continuing...\n"); break; - - case HEADER_HOSTNAME: - case HEADER_OSRELEASE: - case HEADER_VERSION: - case HEADER_ARCH: - case HEADER_NRCPUS: - case HEADER_CPUDESC: - case HEADER_CPUID: - case HEADER_TOTAL_MEM: - case HEADER_CMDLINE: - case HEADER_EVENT_DESC: - case HEADER_CPU_TOPOLOGY: - case HEADER_NUMA_TOPOLOGY: - break; - default: - pr_debug("unknown feature %d, continuing...\n", feat); + break; } return 0; diff --git a/tools/perf/util/header.h b/tools/perf/util/header.h index 09365b32098e..ac4ec956024e 100644 --- a/tools/perf/util/header.h +++ b/tools/perf/util/header.h @@ -10,7 +10,8 @@ #include enum { - HEADER_TRACE_INFO = 1, + HEADER_RESERVED = 0, /* always cleared */ + HEADER_TRACE_INFO = 1, HEADER_BUILD_ID, HEADER_HOSTNAME, @@ -27,10 +28,9 @@ enum { HEADER_NUMA_TOPOLOGY, HEADER_LAST_FEATURE, + HEADER_FEAT_BITS = 256, }; -#define HEADER_FEAT_BITS 256 - struct perf_file_section { u64 offset; u64 size; diff --git a/tools/perf/util/include/linux/bitops.h b/tools/perf/util/include/linux/bitops.h index 305c8484f200..62cdee78db7b 100644 --- a/tools/perf/util/include/linux/bitops.h +++ b/tools/perf/util/include/linux/bitops.h @@ -9,6 +9,17 @@ #define BITS_PER_BYTE 8 #define BITS_TO_LONGS(nr) DIV_ROUND_UP(nr, BITS_PER_BYTE * sizeof(long)) +#define for_each_set_bit(bit, addr, size) \ + for ((bit) = find_first_bit((addr), (size)); \ + (bit) < (size); \ + (bit) = find_next_bit((addr), (size), (bit) + 1)) + +/* same as for_each_set_bit() but use bit as value to start with */ +#define for_each_set_bit_cont(bit, addr, size) \ + for ((bit) = find_next_bit((addr), (size), (bit)); \ + (bit) < (size); \ + (bit) = find_next_bit((addr), (size), (bit) + 1)) + static inline void set_bit(int nr, unsigned long *addr) { addr[nr / BITS_PER_LONG] |= 1UL << (nr % BITS_PER_LONG); @@ -30,4 +41,111 @@ static inline unsigned long hweight_long(unsigned long w) return sizeof(w) == 4 ? hweight32(w) : hweight64(w); } +#define BITOP_WORD(nr) ((nr) / BITS_PER_LONG) + +/** + * __ffs - find first bit in word. + * @word: The word to search + * + * Undefined if no bit exists, so code should check against 0 first. + */ +static __always_inline unsigned long __ffs(unsigned long word) +{ + int num = 0; + +#if BITS_PER_LONG == 64 + if ((word & 0xffffffff) == 0) { + num += 32; + word >>= 32; + } +#endif + if ((word & 0xffff) == 0) { + num += 16; + word >>= 16; + } + if ((word & 0xff) == 0) { + num += 8; + word >>= 8; + } + if ((word & 0xf) == 0) { + num += 4; + word >>= 4; + } + if ((word & 0x3) == 0) { + num += 2; + word >>= 2; + } + if ((word & 0x1) == 0) + num += 1; + return num; +} + +/* + * Find the first set bit in a memory region. + */ +static inline unsigned long +find_first_bit(const unsigned long *addr, unsigned long size) +{ + const unsigned long *p = addr; + unsigned long result = 0; + unsigned long tmp; + + while (size & ~(BITS_PER_LONG-1)) { + if ((tmp = *(p++))) + goto found; + result += BITS_PER_LONG; + size -= BITS_PER_LONG; + } + if (!size) + return result; + + tmp = (*p) & (~0UL >> (BITS_PER_LONG - size)); + if (tmp == 0UL) /* Are any bits set? */ + return result + size; /* Nope. */ +found: + return result + __ffs(tmp); +} + +/* + * Find the next set bit in a memory region. + */ +static inline unsigned long +find_next_bit(const unsigned long *addr, unsigned long size, unsigned long offset) +{ + const unsigned long *p = addr + BITOP_WORD(offset); + unsigned long result = offset & ~(BITS_PER_LONG-1); + unsigned long tmp; + + if (offset >= size) + return size; + size -= result; + offset %= BITS_PER_LONG; + if (offset) { + tmp = *(p++); + tmp &= (~0UL << offset); + if (size < BITS_PER_LONG) + goto found_first; + if (tmp) + goto found_middle; + size -= BITS_PER_LONG; + result += BITS_PER_LONG; + } + while (size & ~(BITS_PER_LONG-1)) { + if ((tmp = *(p++))) + goto found_middle; + result += BITS_PER_LONG; + size -= BITS_PER_LONG; + } + if (!size) + return result; + tmp = *p; + +found_first: + tmp &= (~0UL >> (BITS_PER_LONG - size)); + if (tmp == 0UL) /* Are any bits set? */ + return result + size; /* Nope. */ +found_middle: + return result + __ffs(tmp); +} + #endif -- cgit v1.2.3 From 37a058ea006de0cc24553637afa788594a975176 Mon Sep 17 00:00:00 2001 From: Robert Richter Date: Thu, 15 Dec 2011 18:23:43 +0100 Subject: perf script: Add generic perl handler to process events The current perf scripting facility only supports tracepoints. This patch implements a generic perl handler to support other events than tracepoints too. This patch introduces a function process_event() that is called by perf for each sample. The function is called with byte streams as arguments containing information about the event, its attributes, the sample and raw data. Perl's unpack() function can easily be used for byte decoding. The following is the default implementation for process_event() that can also be generated with perf script: # Packed byte string args of process_event(): # # $event: union perf_event util/event.h # $attr: struct perf_event_attr linux/perf_event.h # $sample: struct perf_sample util/event.h # $raw_data: perf_sample->raw_data util/event.h sub process_event { my ($event, $attr, $sample, $raw_data) = @_; my @event = unpack("LSS", $event); my @attr = unpack("LLQQQQQLLQQ", $attr); my @sample = unpack("QLLQQQQQLL", $sample); my @raw_data = unpack("C*", $raw_data); use Data::Dumper; print Dumper \@event, \@attr, \@sample, \@raw_data; } Cc: Ingo Molnar Cc: Peter Zijlstra Cc: Stephane Eranian Link: http://lkml.kernel.org/r/1323969824-9711-4-git-send-email-robert.richter@amd.com Signed-off-by: Robert Richter Signed-off-by: Arnaldo Carvalho de Melo --- .../perf/util/scripting-engines/trace-event-perl.c | 73 ++++++++++++++++++++-- 1 file changed, 67 insertions(+), 6 deletions(-) diff --git a/tools/perf/util/scripting-engines/trace-event-perl.c b/tools/perf/util/scripting-engines/trace-event-perl.c index a82ce4303ff5..e30749e38a9b 100644 --- a/tools/perf/util/scripting-engines/trace-event-perl.c +++ b/tools/perf/util/scripting-engines/trace-event-perl.c @@ -30,6 +30,7 @@ #include "../thread.h" #include "../event.h" #include "../trace-event.h" +#include "../evsel.h" #include #include @@ -247,11 +248,11 @@ static inline struct event *find_cache_event(int type) return event; } -static void perl_process_event(union perf_event *pevent __unused, - struct perf_sample *sample, - struct perf_evsel *evsel, - struct machine *machine __unused, - struct thread *thread) +static void perl_process_tracepoint(union perf_event *pevent __unused, + struct perf_sample *sample, + struct perf_evsel *evsel, + struct machine *machine __unused, + struct thread *thread) { struct format_field *field; static char handler[256]; @@ -267,6 +268,9 @@ static void perl_process_event(union perf_event *pevent __unused, dSP; + if (evsel->attr.type != PERF_TYPE_TRACEPOINT) + return; + type = trace_parse_common_type(data); event = find_cache_event(type); @@ -334,6 +338,42 @@ static void perl_process_event(union perf_event *pevent __unused, LEAVE; } +static void perl_process_event_generic(union perf_event *pevent __unused, + struct perf_sample *sample, + struct perf_evsel *evsel __unused, + struct machine *machine __unused, + struct thread *thread __unused) +{ + dSP; + + if (!get_cv("process_event", 0)) + return; + + ENTER; + SAVETMPS; + PUSHMARK(SP); + XPUSHs(sv_2mortal(newSVpvn((const char *)pevent, pevent->header.size))); + XPUSHs(sv_2mortal(newSVpvn((const char *)&evsel->attr, sizeof(evsel->attr)))); + XPUSHs(sv_2mortal(newSVpvn((const char *)sample, sizeof(*sample)))); + XPUSHs(sv_2mortal(newSVpvn((const char *)sample->raw_data, sample->raw_size))); + PUTBACK; + call_pv("process_event", G_SCALAR); + SPAGAIN; + PUTBACK; + FREETMPS; + LEAVE; +} + +static void perl_process_event(union perf_event *pevent, + struct perf_sample *sample, + struct perf_evsel *evsel, + struct machine *machine, + struct thread *thread) +{ + perl_process_tracepoint(pevent, sample, evsel, machine, thread); + perl_process_event_generic(pevent, sample, evsel, machine, thread); +} + static void run_start_sub(void) { dSP; /* access to Perl stack */ @@ -555,7 +595,28 @@ static int perl_generate_script(const char *outfile) fprintf(ofp, "sub print_header\n{\n" "\tmy ($event_name, $cpu, $secs, $nsecs, $pid, $comm) = @_;\n\n" "\tprintf(\"%%-20s %%5u %%05u.%%09u %%8u %%-20s \",\n\t " - "$event_name, $cpu, $secs, $nsecs, $pid, $comm);\n}"); + "$event_name, $cpu, $secs, $nsecs, $pid, $comm);\n}\n"); + + fprintf(ofp, + "\n# Packed byte string args of process_event():\n" + "#\n" + "# $event:\tunion perf_event\tutil/event.h\n" + "# $attr:\tstruct perf_event_attr\tlinux/perf_event.h\n" + "# $sample:\tstruct perf_sample\tutil/event.h\n" + "# $raw_data:\tperf_sample->raw_data\tutil/event.h\n" + "\n" + "sub process_event\n" + "{\n" + "\tmy ($event, $attr, $sample, $raw_data) = @_;\n" + "\n" + "\tmy @event\t= unpack(\"LSS\", $event);\n" + "\tmy @attr\t= unpack(\"LLQQQQQLLQQ\", $attr);\n" + "\tmy @sample\t= unpack(\"QLLQQQQQLL\", $sample);\n" + "\tmy @raw_data\t= unpack(\"C*\", $raw_data);\n" + "\n" + "\tuse Data::Dumper;\n" + "\tprint Dumper \\@event, \\@attr, \\@sample, \\@raw_data;\n" + "}\n"); fclose(ofp); -- cgit v1.2.3 From f2328062726d36e562f1458d6346b77aa048acad Mon Sep 17 00:00:00 2001 From: Ingo Molnar Date: Thu, 29 Dec 2011 21:26:17 +0100 Subject: perf tools: Fix feature-bits rework fallout, remove unused variable Cc: Arnaldo Carvalho de Melo Cc: Frederic Weisbecker Cc: Peter Zijlstra Cc: Stephane Eranian Cc: Robert Richter Link: http://lkml.kernel.org/n/tip-lfckuwbl8m1ykb7t9ydsxe4r@git.kernel.org Signed-off-by: Ingo Molnar --- tools/perf/util/header.c | 3 --- 1 file changed, 3 deletions(-) diff --git a/tools/perf/util/header.c b/tools/perf/util/header.c index e509a9dea00b..3e7e0b09c12c 100644 --- a/tools/perf/util/header.c +++ b/tools/perf/util/header.c @@ -1415,15 +1415,12 @@ static int perf_header__adds_write(struct perf_header *header, struct perf_evlist *evlist, int fd) { int nr_sections; - struct perf_session *session; struct perf_file_section *feat_sec, *p; int sec_size; u64 sec_start; int feat; int err; - session = container_of(header, struct perf_session, header); - nr_sections = bitmap_weight(header->adds_features, HEADER_FEAT_BITS); if (!nr_sections) return 0; -- cgit v1.2.3 From 29c9862f1b818bf4caa4c48a30dbe5f25c84ee08 Mon Sep 17 00:00:00 2001 From: Namhyung Kim Date: Wed, 28 Dec 2011 00:35:48 +0900 Subject: perf session: Remove impossible condition check The 'size' cannot be 0 because it was set to 8 on the above line in case it was 0 and never changed. Cc: Ingo Molnar Cc: Paul Mackerras Cc: Peter Zijlstra Link: http://lkml.kernel.org/r/1325000151-4463-1-git-send-email-namhyung@gmail.com Signed-off-by: Namhyung Kim Signed-off-by: Arnaldo Carvalho de Melo --- tools/perf/util/session.c | 3 +-- 1 file changed, 1 insertion(+), 2 deletions(-) diff --git a/tools/perf/util/session.c b/tools/perf/util/session.c index cc5e6be46d86..b5ca2558c7bb 100644 --- a/tools/perf/util/session.c +++ b/tools/perf/util/session.c @@ -1015,8 +1015,7 @@ more: } } - if (size == 0 || - (skip = perf_session__process_event(self, &event, tool, head)) < 0) { + if ((skip = perf_session__process_event(self, &event, tool, head)) < 0) { dump_printf("%#" PRIx64 " [%#x]: skipping unknown header type: %d\n", head, event.header.size, event.header.type); /* -- cgit v1.2.3 From 15e6392feec311f1e409d77e1ccfe51c1d940365 Mon Sep 17 00:00:00 2001 From: Namhyung Kim Date: Wed, 28 Dec 2011 00:35:49 +0900 Subject: perf stat: Introduce get_ratio_color() helper The get_ratio_color() returns appropriate color string based on @ratio. It helps reducing code duplication. Cc: Ingo Molnar Cc: Paul Mackerras Cc: Peter Zijlstra Link: http://lkml.kernel.org/r/1325000151-4463-2-git-send-email-namhyung@gmail.com Signed-off-by: Namhyung Kim Signed-off-by: Arnaldo Carvalho de Melo --- tools/perf/builtin-stat.c | 91 ++++++++++++++++++----------------------------- 1 file changed, 35 insertions(+), 56 deletions(-) diff --git a/tools/perf/builtin-stat.c b/tools/perf/builtin-stat.c index cc53de335ced..f5d2a63eba66 100644 --- a/tools/perf/builtin-stat.c +++ b/tools/perf/builtin-stat.c @@ -578,6 +578,33 @@ static void nsec_printout(int cpu, struct perf_evsel *evsel, double avg) avg / avg_stats(&walltime_nsecs_stats)); } +/* used for get_ratio_color() */ +enum grc_type { + GRC_STALLED_CYCLES_FE, + GRC_STALLED_CYCLES_BE, + GRC_CACHE_MISSES, + GRC_MAX_NR +}; + +static const char *get_ratio_color(enum grc_type type, double ratio) +{ + static const double grc_table[GRC_MAX_NR][3] = { + [GRC_STALLED_CYCLES_FE] = { 50.0, 30.0, 10.0 }, + [GRC_STALLED_CYCLES_BE] = { 75.0, 50.0, 20.0 }, + [GRC_CACHE_MISSES] = { 20.0, 10.0, 5.0 }, + }; + const char *color = PERF_COLOR_NORMAL; + + if (ratio > grc_table[type][0]) + color = PERF_COLOR_RED; + else if (ratio > grc_table[type][1]) + color = PERF_COLOR_MAGENTA; + else if (ratio > grc_table[type][2]) + color = PERF_COLOR_YELLOW; + + return color; +} + static void print_stalled_cycles_frontend(int cpu, struct perf_evsel *evsel __used, double avg) { double total, ratio = 0.0; @@ -588,13 +615,7 @@ static void print_stalled_cycles_frontend(int cpu, struct perf_evsel *evsel __us if (total) ratio = avg / total * 100.0; - color = PERF_COLOR_NORMAL; - if (ratio > 50.0) - color = PERF_COLOR_RED; - else if (ratio > 30.0) - color = PERF_COLOR_MAGENTA; - else if (ratio > 10.0) - color = PERF_COLOR_YELLOW; + color = get_ratio_color(GRC_STALLED_CYCLES_FE, ratio); fprintf(output, " # "); color_fprintf(output, color, "%6.2f%%", ratio); @@ -611,13 +632,7 @@ static void print_stalled_cycles_backend(int cpu, struct perf_evsel *evsel __use if (total) ratio = avg / total * 100.0; - color = PERF_COLOR_NORMAL; - if (ratio > 75.0) - color = PERF_COLOR_RED; - else if (ratio > 50.0) - color = PERF_COLOR_MAGENTA; - else if (ratio > 20.0) - color = PERF_COLOR_YELLOW; + color = get_ratio_color(GRC_STALLED_CYCLES_BE, ratio); fprintf(output, " # "); color_fprintf(output, color, "%6.2f%%", ratio); @@ -634,13 +649,7 @@ static void print_branch_misses(int cpu, struct perf_evsel *evsel __used, double if (total) ratio = avg / total * 100.0; - color = PERF_COLOR_NORMAL; - if (ratio > 20.0) - color = PERF_COLOR_RED; - else if (ratio > 10.0) - color = PERF_COLOR_MAGENTA; - else if (ratio > 5.0) - color = PERF_COLOR_YELLOW; + color = get_ratio_color(GRC_CACHE_MISSES, ratio); fprintf(output, " # "); color_fprintf(output, color, "%6.2f%%", ratio); @@ -657,13 +666,7 @@ static void print_l1_dcache_misses(int cpu, struct perf_evsel *evsel __used, dou if (total) ratio = avg / total * 100.0; - color = PERF_COLOR_NORMAL; - if (ratio > 20.0) - color = PERF_COLOR_RED; - else if (ratio > 10.0) - color = PERF_COLOR_MAGENTA; - else if (ratio > 5.0) - color = PERF_COLOR_YELLOW; + color = get_ratio_color(GRC_CACHE_MISSES, ratio); fprintf(output, " # "); color_fprintf(output, color, "%6.2f%%", ratio); @@ -680,13 +683,7 @@ static void print_l1_icache_misses(int cpu, struct perf_evsel *evsel __used, dou if (total) ratio = avg / total * 100.0; - color = PERF_COLOR_NORMAL; - if (ratio > 20.0) - color = PERF_COLOR_RED; - else if (ratio > 10.0) - color = PERF_COLOR_MAGENTA; - else if (ratio > 5.0) - color = PERF_COLOR_YELLOW; + color = get_ratio_color(GRC_CACHE_MISSES, ratio); fprintf(output, " # "); color_fprintf(output, color, "%6.2f%%", ratio); @@ -703,13 +700,7 @@ static void print_dtlb_cache_misses(int cpu, struct perf_evsel *evsel __used, do if (total) ratio = avg / total * 100.0; - color = PERF_COLOR_NORMAL; - if (ratio > 20.0) - color = PERF_COLOR_RED; - else if (ratio > 10.0) - color = PERF_COLOR_MAGENTA; - else if (ratio > 5.0) - color = PERF_COLOR_YELLOW; + color = get_ratio_color(GRC_CACHE_MISSES, ratio); fprintf(output, " # "); color_fprintf(output, color, "%6.2f%%", ratio); @@ -726,13 +717,7 @@ static void print_itlb_cache_misses(int cpu, struct perf_evsel *evsel __used, do if (total) ratio = avg / total * 100.0; - color = PERF_COLOR_NORMAL; - if (ratio > 20.0) - color = PERF_COLOR_RED; - else if (ratio > 10.0) - color = PERF_COLOR_MAGENTA; - else if (ratio > 5.0) - color = PERF_COLOR_YELLOW; + color = get_ratio_color(GRC_CACHE_MISSES, ratio); fprintf(output, " # "); color_fprintf(output, color, "%6.2f%%", ratio); @@ -749,13 +734,7 @@ static void print_ll_cache_misses(int cpu, struct perf_evsel *evsel __used, doub if (total) ratio = avg / total * 100.0; - color = PERF_COLOR_NORMAL; - if (ratio > 20.0) - color = PERF_COLOR_RED; - else if (ratio > 10.0) - color = PERF_COLOR_MAGENTA; - else if (ratio > 5.0) - color = PERF_COLOR_YELLOW; + color = get_ratio_color(GRC_CACHE_MISSES, ratio); fprintf(output, " # "); color_fprintf(output, color, "%6.2f%%", ratio); -- cgit v1.2.3 From 057a174a064f68bac042d618ce3c6ea3ccd9a8aa Mon Sep 17 00:00:00 2001 From: Namhyung Kim Date: Wed, 28 Dec 2011 00:35:50 +0900 Subject: perf top: Fix a memory leak The 'buf' should be freed when symbol wasn't found too. Cc: Ingo Molnar Cc: Paul Mackerras Cc: Peter Zijlstra Link: http://lkml.kernel.org/r/1325000151-4463-3-git-send-email-namhyung@gmail.com Signed-off-by: Namhyung Kim Signed-off-by: Arnaldo Carvalho de Melo --- tools/perf/builtin-top.c | 1 - 1 file changed, 1 deletion(-) diff --git a/tools/perf/builtin-top.c b/tools/perf/builtin-top.c index c3836b966ccf..4f81eeb99875 100644 --- a/tools/perf/builtin-top.c +++ b/tools/perf/builtin-top.c @@ -351,7 +351,6 @@ static void perf_top__prompt_symbol(struct perf_top *top, const char *msg) if (!found) { fprintf(stderr, "Sorry, %s is not active.\n", buf); sleep(1); - return; } else perf_top__parse_source(top, found); -- cgit v1.2.3 From 466e2876bcb9ddc9b92502c46689679bee7d72a0 Mon Sep 17 00:00:00 2001 From: Namhyung Kim Date: Wed, 28 Dec 2011 00:35:51 +0900 Subject: perf script: Kill script_spec__delete As script_spec__delete() frees given struct script_spec it should not be called if we failed to allocate the struct. Also it's the only caller of the function, we can get rid of the function itself. Cc: Ingo Molnar Cc: Paul Mackerras Cc: Peter Zijlstra Link: http://lkml.kernel.org/r/1325000151-4463-4-git-send-email-namhyung@gmail.com Signed-off-by: Namhyung Kim Signed-off-by: Arnaldo Carvalho de Melo --- tools/perf/builtin-script.c | 13 +------------ 1 file changed, 1 insertion(+), 12 deletions(-) diff --git a/tools/perf/builtin-script.c b/tools/perf/builtin-script.c index 3d4c0c7b576e..fd1909afcfd6 100644 --- a/tools/perf/builtin-script.c +++ b/tools/perf/builtin-script.c @@ -536,12 +536,6 @@ static struct script_spec *script_spec__new(const char *spec, return s; } -static void script_spec__delete(struct script_spec *s) -{ - free(s->spec); - free(s); -} - static void script_spec__add(struct script_spec *s) { list_add_tail(&s->node, &script_specs); @@ -567,16 +561,11 @@ static struct script_spec *script_spec__findnew(const char *spec, s = script_spec__new(spec, ops); if (!s) - goto out_delete_spec; + return NULL; script_spec__add(s); return s; - -out_delete_spec: - script_spec__delete(s); - - return NULL; } int script_spec_register(const char *spec, struct scripting_ops *ops) -- cgit v1.2.3 From 9e183426bfb52bb44bf3c443d6587e4d02478603 Mon Sep 17 00:00:00 2001 From: Joerg Roedel Date: Wed, 5 Oct 2011 14:01:19 +0200 Subject: perf kvm: Fix copy & paste error in description The --host option certainly enables host-data collection. Cc: Avi Kivity Cc: Ingo Molnar Cc: Peter Zijlstra Cc: Joerg Roedel Cc: kvm@vger.kernel.org Link: http://lkml.kernel.org/r/1317816084-18026-5-git-send-email-gleb@redhat.com Signed-off-by: Gleb Natapov Signed-off-by: Joerg Roedel Signed-off-by: Arnaldo Carvalho de Melo --- tools/perf/builtin-kvm.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) diff --git a/tools/perf/builtin-kvm.c b/tools/perf/builtin-kvm.c index 34d1e853829d..032324a76b87 100644 --- a/tools/perf/builtin-kvm.c +++ b/tools/perf/builtin-kvm.c @@ -38,7 +38,7 @@ static const struct option kvm_options[] = { OPT_BOOLEAN(0, "guest", &perf_guest, "Collect guest os data"), OPT_BOOLEAN(0, "host", &perf_host, - "Collect guest os data"), + "Collect host os data"), OPT_STRING(0, "guestmount", &symbol_conf.guestmount, "directory", "guest mount directory under which every guest os" " instance has a subdir"), -- cgit v1.2.3