From 96a2c464de07d7c72988db851c029b204fc59108 Mon Sep 17 00:00:00 2001 From: Frederic Weisbecker Date: Sat, 1 Aug 2009 01:34:24 +0200 Subject: tracing/bkl: Add bkl ftrace events Add two events lock_kernel and unlock_kernel() to trace the bkl uses. This opens the door for userspace tools to perform statistics about the callsites that use it, dependencies with other locks (by pairing the trace with lock events), use with recursivity and so on... The {__reacquire,release}_kernel_lock() events are not traced because these are called from schedule, thus the sched events are sufficient to trace them. Example of a trace: hald-addon-stor-4152 [000] 165.875501: unlock_kernel: depth: 0, fs/block_dev.c:1358 __blkdev_put() hald-addon-stor-4152 [000] 167.832974: lock_kernel: depth: 0, fs/block_dev.c:1167 __blkdev_get() How to get the callsites that acquire it recursively: cd /debug/tracing/events/bkl echo "lock_depth > 0" > filter firefox-4951 [001] 206.276967: unlock_kernel: depth: 1, fs/reiserfs/super.c:575 reiserfs_dirty_inode() You can also filter by file and/or line. v2: Use of FILTER_PTR_STRING attribute for files and lines fields to make them traceable. Signed-off-by: Frederic Weisbecker Cc: Steven Rostedt Cc: Li Zefan --- include/trace/events/bkl.h | 61 ++++++++++++++++++++++++++++++++++++++++++++++ 1 file changed, 61 insertions(+) create mode 100644 include/trace/events/bkl.h (limited to 'include/trace/events') diff --git a/include/trace/events/bkl.h b/include/trace/events/bkl.h new file mode 100644 index 000000000000..8abd620a490e --- /dev/null +++ b/include/trace/events/bkl.h @@ -0,0 +1,61 @@ +#undef TRACE_SYSTEM +#define TRACE_SYSTEM bkl + +#if !defined(_TRACE_BKL_H) || defined(TRACE_HEADER_MULTI_READ) +#define _TRACE_BKL_H + +#include + +TRACE_EVENT(lock_kernel, + + TP_PROTO(const char *func, const char *file, int line), + + TP_ARGS(func, file, line), + + TP_STRUCT__entry( + __field( int, lock_depth ) + __field_ext( const char *, func, FILTER_PTR_STRING ) + __field_ext( const char *, file, FILTER_PTR_STRING ) + __field( int, line ) + ), + + TP_fast_assign( + /* We want to record the lock_depth after lock is acquired */ + __entry->lock_depth = current->lock_depth + 1; + __entry->func = func; + __entry->file = file; + __entry->line = line; + ), + + TP_printk("depth: %d, %s:%d %s()", __entry->lock_depth, + __entry->file, __entry->line, __entry->func) +); + +TRACE_EVENT(unlock_kernel, + + TP_PROTO(const char *func, const char *file, int line), + + TP_ARGS(func, file, line), + + TP_STRUCT__entry( + __field(int, lock_depth) + __field(const char *, func) + __field(const char *, file) + __field(int, line) + ), + + TP_fast_assign( + __entry->lock_depth = current->lock_depth; + __entry->func = func; + __entry->file = file; + __entry->line = line; + ), + + TP_printk("depth: %d, %s:%d %s()", __entry->lock_depth, + __entry->file, __entry->line, __entry->func) +); + +#endif /* _TRACE_BKL_H */ + +/* This part must be outside protection */ +#include -- cgit v1.2.3 From b8007ef7422270864eae523cb38d7522a53a94d3 Mon Sep 17 00:00:00 2001 From: Lai Jiangshan Date: Tue, 3 Nov 2009 13:45:32 +0800 Subject: tracing: Separate raw syscall from syscall tracer The current syscall tracer mixes raw syscalls and real syscalls. echo 1 > events/syscalls/enable And we get these from the output: (XXXX insteads " grep-20914 [001] 588211.446347" .. etc) XXXX: sys_read(fd: 3, buf: 80609a8, count: 7000) XXXX: sys_enter: NR 3 (3, 80609a8, 7000, a, 1000, bfce8ef8) XXXX: sys_read -> 0x138 XXXX: sys_exit: NR 3 = 312 XXXX: sys_read(fd: 3, buf: 8060ae0, count: 7000) XXXX: sys_enter: NR 3 (3, 8060ae0, 7000, a, 1000, bfce8ef8) XXXX: sys_read -> 0x138 XXXX: sys_exit: NR 3 = 312 There are 2 drawbacks here. A) two almost identical records are saved in ringbuffer when a syscall enters or exits. (4 records for every syscall) This wastes precious space in the ring buffer. B) the lines including "sys_enter/sys_exit" produces hardly any useful information for the output (no labels). The user can use this method to prevent these drawbacks: echo 1 > events/syscalls/enable echo 0 > events/syscalls/sys_enter/enable echo 0 > events/syscalls/sys_exit/enable But this is not user friendly. So we separate raw syscall from syscall tracer. After this fix applied: syscall tracer's output (echo 1 > events/syscalls/enable): XXXX: sys_read(fd: 3, buf: bfe87d88, count: 200) XXXX: sys_read -> 0x200 XXXX: sys_fstat64(fd: 3, statbuf: bfe87c98) XXXX: sys_fstat64 -> 0x0 XXXX: sys_close(fd: 3) raw syscall tracer's output (echo 1 > events/raw_syscalls/enable): XXXX: sys_enter: NR 175 (0, bf92bf18, bf92bf98, 8, b748cff4, bf92bef8) XXXX: sys_exit: NR 175 = 0 XXXX: sys_enter: NR 175 (2, bf92bf98, 0, 8, b748cff4, bf92bef8) XXXX: sys_exit: NR 175 = 0 XXXX: sys_enter: NR 3 (9, bf927f9c, 4000, b77e2518, b77dce60, bf92bff8) Signed-off-by: Lai Jiangshan LKML-Reference: <4AEFC37C.5080609@cn.fujitsu.com> Signed-off-by: Steven Rostedt --- include/trace/events/syscalls.h | 3 ++- 1 file changed, 2 insertions(+), 1 deletion(-) (limited to 'include/trace/events') diff --git a/include/trace/events/syscalls.h b/include/trace/events/syscalls.h index 397dff2dbd5a..fb726ac7caee 100644 --- a/include/trace/events/syscalls.h +++ b/include/trace/events/syscalls.h @@ -1,5 +1,6 @@ #undef TRACE_SYSTEM -#define TRACE_SYSTEM syscalls +#define TRACE_SYSTEM raw_syscalls +#define TRACE_INCLUDE_FILE syscalls #if !defined(_TRACE_EVENTS_SYSCALLS_H) || defined(TRACE_HEADER_MULTI_READ) #define _TRACE_EVENTS_SYSCALLS_H -- cgit v1.2.3