diff options
Diffstat (limited to 'arch/mips/kernel')
-rw-r--r-- | arch/mips/kernel/.gitignore | 1 | ||||
-rw-r--r-- | arch/mips/kernel/Makefile | 1 | ||||
-rw-r--r-- | arch/mips/kernel/cevt-r4k.c | 173 | ||||
-rw-r--r-- | arch/mips/kernel/cevt-smtc.c | 321 | ||||
-rw-r--r-- | arch/mips/kernel/cpu-probe.c | 26 | ||||
-rw-r--r-- | arch/mips/kernel/entry.S | 10 | ||||
-rw-r--r-- | arch/mips/kernel/genex.S | 41 | ||||
-rw-r--r-- | arch/mips/kernel/kgdb.c | 10 | ||||
-rw-r--r-- | arch/mips/kernel/mips-mt-fpaff.c | 2 | ||||
-rw-r--r-- | arch/mips/kernel/process.c | 19 | ||||
-rw-r--r-- | arch/mips/kernel/ptrace.c | 2 | ||||
-rw-r--r-- | arch/mips/kernel/scall32-o32.S | 6 | ||||
-rw-r--r-- | arch/mips/kernel/scall64-64.S | 6 | ||||
-rw-r--r-- | arch/mips/kernel/scall64-n32.S | 6 | ||||
-rw-r--r-- | arch/mips/kernel/scall64-o32.S | 6 | ||||
-rw-r--r-- | arch/mips/kernel/setup.c | 75 | ||||
-rw-r--r-- | arch/mips/kernel/smtc.c | 260 | ||||
-rw-r--r-- | arch/mips/kernel/traps.c | 46 | ||||
-rw-r--r-- | arch/mips/kernel/vmlinux.lds.S | 1 |
19 files changed, 670 insertions, 342 deletions
diff --git a/arch/mips/kernel/.gitignore b/arch/mips/kernel/.gitignore new file mode 100644 index 000000000000..c5f676c3c224 --- /dev/null +++ b/arch/mips/kernel/.gitignore @@ -0,0 +1 @@ +vmlinux.lds diff --git a/arch/mips/kernel/Makefile b/arch/mips/kernel/Makefile index 706f93974797..25775cb54000 100644 --- a/arch/mips/kernel/Makefile +++ b/arch/mips/kernel/Makefile @@ -10,6 +10,7 @@ obj-y += cpu-probe.o branch.o entry.o genex.o irq.o process.o \ obj-$(CONFIG_CEVT_BCM1480) += cevt-bcm1480.o obj-$(CONFIG_CEVT_R4K) += cevt-r4k.o +obj-$(CONFIG_MIPS_MT_SMTC) += cevt-smtc.o obj-$(CONFIG_CEVT_DS1287) += cevt-ds1287.o obj-$(CONFIG_CEVT_GT641XX) += cevt-gt641xx.o obj-$(CONFIG_CEVT_SB1250) += cevt-sb1250.o diff --git a/arch/mips/kernel/cevt-r4k.c b/arch/mips/kernel/cevt-r4k.c index 24a2d907aa0d..4a4c59f2737a 100644 --- a/arch/mips/kernel/cevt-r4k.c +++ b/arch/mips/kernel/cevt-r4k.c @@ -12,6 +12,14 @@ #include <asm/smtc_ipi.h> #include <asm/time.h> +#include <asm/cevt-r4k.h> + +/* + * The SMTC Kernel for the 34K, 1004K, et. al. replaces several + * of these routines with SMTC-specific variants. + */ + +#ifndef CONFIG_MIPS_MT_SMTC static int mips_next_event(unsigned long delta, struct clock_event_device *evt) @@ -19,60 +27,27 @@ static int mips_next_event(unsigned long delta, unsigned int cnt; int res; -#ifdef CONFIG_MIPS_MT_SMTC - { - unsigned long flags, vpflags; - local_irq_save(flags); - vpflags = dvpe(); -#endif cnt = read_c0_count(); cnt += delta; write_c0_compare(cnt); res = ((int)(read_c0_count() - cnt) > 0) ? -ETIME : 0; -#ifdef CONFIG_MIPS_MT_SMTC - evpe(vpflags); - local_irq_restore(flags); - } -#endif return res; } -static void mips_set_mode(enum clock_event_mode mode, - struct clock_event_device *evt) +#endif /* CONFIG_MIPS_MT_SMTC */ + +void mips_set_clock_mode(enum clock_event_mode mode, + struct clock_event_device *evt) { /* Nothing to do ... */ } -static DEFINE_PER_CPU(struct clock_event_device, mips_clockevent_device); -static int cp0_timer_irq_installed; +DEFINE_PER_CPU(struct clock_event_device, mips_clockevent_device); +int cp0_timer_irq_installed; -/* - * Timer ack for an R4k-compatible timer of a known frequency. - */ -static void c0_timer_ack(void) -{ - write_c0_compare(read_c0_compare()); -} +#ifndef CONFIG_MIPS_MT_SMTC -/* - * Possibly handle a performance counter interrupt. - * Return true if the timer interrupt should not be checked - */ -static inline int handle_perf_irq(int r2) -{ - /* - * The performance counter overflow interrupt may be shared with the - * timer interrupt (cp0_perfcount_irq < 0). If it is and a - * performance counter has overflowed (perf_irq() == IRQ_HANDLED) - * and we can't reliably determine if a counter interrupt has also - * happened (!r2) then don't check for a timer interrupt. - */ - return (cp0_perfcount_irq < 0) && - perf_irq() == IRQ_HANDLED && - !r2; -} - -static irqreturn_t c0_compare_interrupt(int irq, void *dev_id) +irqreturn_t c0_compare_interrupt(int irq, void *dev_id) { const int r2 = cpu_has_mips_r2; struct clock_event_device *cd; @@ -93,12 +68,8 @@ static irqreturn_t c0_compare_interrupt(int irq, void *dev_id) * interrupt. Being the paranoiacs we are we check anyway. */ if (!r2 || (read_c0_cause() & (1 << 30))) { - c0_timer_ack(); -#ifdef CONFIG_MIPS_MT_SMTC - if (cpu_data[cpu].vpe_id) - goto out; - cpu = 0; -#endif + /* Clear Count/Compare Interrupt */ + write_c0_compare(read_c0_compare()); cd = &per_cpu(mips_clockevent_device, cpu); cd->event_handler(cd); } @@ -107,65 +78,16 @@ out: return IRQ_HANDLED; } -static struct irqaction c0_compare_irqaction = { +#endif /* Not CONFIG_MIPS_MT_SMTC */ + +struct irqaction c0_compare_irqaction = { .handler = c0_compare_interrupt, -#ifdef CONFIG_MIPS_MT_SMTC - .flags = IRQF_DISABLED, -#else .flags = IRQF_DISABLED | IRQF_PERCPU, -#endif .name = "timer", }; -#ifdef CONFIG_MIPS_MT_SMTC -DEFINE_PER_CPU(struct clock_event_device, smtc_dummy_clockevent_device); - -static void smtc_set_mode(enum clock_event_mode mode, - struct clock_event_device *evt) -{ -} - -static void mips_broadcast(cpumask_t mask) -{ - unsigned int cpu; - - for_each_cpu_mask(cpu, mask) - smtc_send_ipi(cpu, SMTC_CLOCK_TICK, 0); -} - -static void setup_smtc_dummy_clockevent_device(void) -{ - //uint64_t mips_freq = mips_hpt_^frequency; - unsigned int cpu = smp_processor_id(); - struct clock_event_device *cd; - cd = &per_cpu(smtc_dummy_clockevent_device, cpu); - - cd->name = "SMTC"; - cd->features = CLOCK_EVT_FEAT_DUMMY; - - /* Calculate the min / max delta */ - cd->mult = 0; //div_sc((unsigned long) mips_freq, NSEC_PER_SEC, 32); - cd->shift = 0; //32; - cd->max_delta_ns = 0; //clockevent_delta2ns(0x7fffffff, cd); - cd->min_delta_ns = 0; //clockevent_delta2ns(0x30, cd); - - cd->rating = 200; - cd->irq = 17; //-1; -// if (cpu) -// cd->cpumask = CPU_MASK_ALL; // cpumask_of_cpu(cpu); -// else - cd->cpumask = cpumask_of_cpu(cpu); - - cd->set_mode = smtc_set_mode; - - cd->broadcast = mips_broadcast; - - clockevents_register_device(cd); -} -#endif - -static void mips_event_handler(struct clock_event_device *dev) +void mips_event_handler(struct clock_event_device *dev) { } @@ -177,7 +99,23 @@ static int c0_compare_int_pending(void) return (read_c0_cause() >> cp0_compare_irq) & 0x100; } -static int c0_compare_int_usable(void) +/* + * Compare interrupt can be routed and latched outside the core, + * so a single execution hazard barrier may not be enough to give + * it time to clear as seen in the Cause register. 4 time the + * pipeline depth seems reasonably conservative, and empirically + * works better in configurations with high CPU/bus clock ratios. + */ + +#define compare_change_hazard() \ + do { \ + irq_disable_hazard(); \ + irq_disable_hazard(); \ + irq_disable_hazard(); \ + irq_disable_hazard(); \ + } while (0) + +int c0_compare_int_usable(void) { unsigned int delta; unsigned int cnt; @@ -187,7 +125,7 @@ static int c0_compare_int_usable(void) */ if (c0_compare_int_pending()) { write_c0_compare(read_c0_count()); - irq_disable_hazard(); + compare_change_hazard(); if (c0_compare_int_pending()) return 0; } @@ -196,7 +134,7 @@ static int c0_compare_int_usable(void) cnt = read_c0_count(); cnt += delta; write_c0_compare(cnt); - irq_disable_hazard(); + compare_change_hazard(); if ((int)(read_c0_count() - cnt) < 0) break; /* increase delta if the timer was already expired */ @@ -205,11 +143,12 @@ static int c0_compare_int_usable(void) while ((int)(read_c0_count() - cnt) <= 0) ; /* Wait for expiry */ + compare_change_hazard(); if (!c0_compare_int_pending()) return 0; write_c0_compare(read_c0_count()); - irq_disable_hazard(); + compare_change_hazard(); if (c0_compare_int_pending()) return 0; @@ -219,6 +158,8 @@ static int c0_compare_int_usable(void) return 1; } +#ifndef CONFIG_MIPS_MT_SMTC + int __cpuinit mips_clockevent_init(void) { uint64_t mips_freq = mips_hpt_frequency; @@ -229,17 +170,6 @@ int __cpuinit mips_clockevent_init(void) if (!cpu_has_counter || !mips_hpt_frequency) return -ENXIO; -#ifdef CONFIG_MIPS_MT_SMTC - setup_smtc_dummy_clockevent_device(); - - /* - * On SMTC we only register VPE0's compare interrupt as clockevent - * device. - */ - if (cpu) - return 0; -#endif - if (!c0_compare_int_usable()) return -ENXIO; @@ -265,13 +195,9 @@ int __cpuinit mips_clockevent_init(void) cd->rating = 300; cd->irq = irq; -#ifdef CONFIG_MIPS_MT_SMTC - cd->cpumask = CPU_MASK_ALL; -#else cd->cpumask = cpumask_of_cpu(cpu); -#endif cd->set_next_event = mips_next_event; - cd->set_mode = mips_set_mode; + cd->set_mode = mips_set_clock_mode; cd->event_handler = mips_event_handler; clockevents_register_device(cd); @@ -281,12 +207,9 @@ int __cpuinit mips_clockevent_init(void) cp0_timer_irq_installed = 1; -#ifdef CONFIG_MIPS_MT_SMTC -#define CPUCTR_IMASKBIT (0x100 << cp0_compare_irq) - setup_irq_smtc(irq, &c0_compare_irqaction, CPUCTR_IMASKBIT); -#else setup_irq(irq, &c0_compare_irqaction); -#endif return 0; } + +#endif /* Not CONFIG_MIPS_MT_SMTC */ diff --git a/arch/mips/kernel/cevt-smtc.c b/arch/mips/kernel/cevt-smtc.c new file mode 100644 index 000000000000..5162fe4b5952 --- /dev/null +++ b/arch/mips/kernel/cevt-smtc.c @@ -0,0 +1,321 @@ +/* + * This file is subject to the terms and conditions of the GNU General Public + * License. See the file "COPYING" in the main directory of this archive + * for more details. + * + * Copyright (C) 2007 MIPS Technologies, Inc. + * Copyright (C) 2007 Ralf Baechle <ralf@linux-mips.org> + * Copyright (C) 2008 Kevin D. Kissell, Paralogos sarl + */ +#include <linux/clockchips.h> +#include <linux/interrupt.h> +#include <linux/percpu.h> + +#include <asm/smtc_ipi.h> +#include <asm/time.h> +#include <asm/cevt-r4k.h> + +/* + * Variant clock event timer support for SMTC on MIPS 34K, 1004K + * or other MIPS MT cores. + * + * Notes on SMTC Support: + * + * SMTC has multiple microthread TCs pretending to be Linux CPUs. + * But there's only one Count/Compare pair per VPE, and Compare + * interrupts are taken opportunisitically by available TCs + * bound to the VPE with the Count register. The new timer + * framework provides for global broadcasts, but we really + * want VPE-level multicasts for best behavior. So instead + * of invoking the high-level clock-event broadcast code, + * this version of SMTC support uses the historical SMTC + * multicast mechanisms "under the hood", appearing to the + * generic clock layer as if the interrupts are per-CPU. + * + * The approach taken here is to maintain a set of NR_CPUS + * virtual timers, and track which "CPU" needs to be alerted + * at each event. + * + * It's unlikely that we'll see a MIPS MT core with more than + * 2 VPEs, but we *know* that we won't need to handle more + * VPEs than we have "CPUs". So NCPUs arrays of NCPUs elements + * is always going to be overkill, but always going to be enough. + */ + +unsigned long smtc_nexttime[NR_CPUS][NR_CPUS]; +static int smtc_nextinvpe[NR_CPUS]; + +/* + * Timestamps stored are absolute values to be programmed + * into Count register. Valid timestamps will never be zero. + * If a Zero Count value is actually calculated, it is converted + * to be a 1, which will introduce 1 or two CPU cycles of error + * roughly once every four billion events, which at 1000 HZ means + * about once every 50 days. If that's actually a problem, one + * could alternate squashing 0 to 1 and to -1. + */ + +#define MAKEVALID(x) (((x) == 0L) ? 1L : (x)) +#define ISVALID(x) ((x) != 0L) + +/* + * Time comparison is subtle, as it's really truncated + * modular arithmetic. + */ + +#define IS_SOONER(a, b, reference) \ + (((a) - (unsigned long)(reference)) < ((b) - (unsigned long)(reference))) + +/* + * CATCHUP_INCREMENT, used when the function falls behind the counter. + * Could be an increasing function instead of a constant; + */ + +#define CATCHUP_INCREMENT 64 + +static int mips_next_event(unsigned long delta, + struct clock_event_device *evt) +{ + unsigned long flags; + unsigned int mtflags; + unsigned long timestamp, reference, previous; + unsigned long nextcomp = 0L; + int vpe = current_cpu_data.vpe_id; + int cpu = smp_processor_id(); + local_irq_save(flags); + mtflags = dmt(); + + /* + * Maintain the per-TC virtual timer + * and program the per-VPE shared Count register + * as appropriate here... + */ + reference = (unsigned long)read_c0_count(); + timestamp = MAKEVALID(reference + delta); + /* + * To really model the clock, we have to catch the case + * where the current next-in-VPE timestamp is the old + * timestamp for the calling CPE, but the new value is + * in fact later. In that case, we have to do a full + * scan and discover the new next-in-VPE CPU id and + * timestamp. + */ + previous = smtc_nexttime[vpe][cpu]; + if (cpu == smtc_nextinvpe[vpe] && ISVALID(previous) + && IS_SOONER(previous, timestamp, reference)) { + int i; + int soonest = cpu; + + /* + * Update timestamp array here, so that new + * value gets considered along with those of + * other virtual CPUs on the VPE. + */ + smtc_nexttime[vpe][cpu] = timestamp; + for_each_online_cpu(i) { + if (ISVALID(smtc_nexttime[vpe][i]) + && IS_SOONER(smtc_nexttime[vpe][i], + smtc_nexttime[vpe][soonest], reference)) { + soonest = i; + } + } + smtc_nextinvpe[vpe] = soonest; + nextcomp = smtc_nexttime[vpe][soonest]; + /* + * Otherwise, we don't have to process the whole array rank, + * we just have to see if the event horizon has gotten closer. + */ + } else { + if (!ISVALID(smtc_nexttime[vpe][smtc_nextinvpe[vpe]]) || + IS_SOONER(timestamp, + smtc_nexttime[vpe][smtc_nextinvpe[vpe]], reference)) { + smtc_nextinvpe[vpe] = cpu; + nextcomp = timestamp; + } + /* + * Since next-in-VPE may me the same as the executing + * virtual CPU, we update the array *after* checking + * its value. + */ + smtc_nexttime[vpe][cpu] = timestamp; + } + + /* + * It may be that, in fact, we don't need to update Compare, + * but if we do, we want to make sure we didn't fall into + * a crack just behind Count. + */ + if (ISVALID(nextcomp)) { + write_c0_compare(nextcomp); + ehb(); + /* + * We never return an error, we just make sure + * that we trigger the handlers as quickly as + * we can if we fell behind. + */ + while ((nextcomp - (unsigned long)read_c0_count()) + > (unsigned long)LONG_MAX) { + nextcomp += CATCHUP_INCREMENT; + write_c0_compare(nextcomp); + ehb(); + } + } + emt(mtflags); + local_irq_restore(flags); + return 0; +} + + +void smtc_distribute_timer(int vpe) +{ + unsigned long flags; + unsigned int mtflags; + int cpu; + struct clock_event_device *cd; + unsigned long nextstamp = 0L; + unsigned long reference; + + +repeat: + for_each_online_cpu(cpu) { + /* + * Find virtual CPUs within the current VPE who have + * unserviced timer requests whose time is now past. + */ + local_irq_save(flags); + mtflags = dmt(); + if (cpu_data[cpu].vpe_id == vpe && + ISVALID(smtc_nexttime[vpe][cpu])) { + reference = (unsigned long)read_c0_count(); + if ((smtc_nexttime[vpe][cpu] - reference) + > (unsigned long)LONG_MAX) { + smtc_nexttime[vpe][cpu] = 0L; + emt(mtflags); + local_irq_restore(flags); + /* + * We don't send IPIs to ourself. + */ + if (cpu != smp_processor_id()) { + smtc_send_ipi(cpu, SMTC_CLOCK_TICK, 0); + } else { + cd = &per_cpu(mips_clockevent_device, cpu); + cd->event_handler(cd); + } + } else { + /* Local to VPE but Valid Time not yet reached. */ + if (!ISVALID(nextstamp) || + IS_SOONER(smtc_nexttime[vpe][cpu], nextstamp, + reference)) { + smtc_nextinvpe[vpe] = cpu; + nextstamp = smtc_nexttime[vpe][cpu]; + } + emt(mtflags); + local_irq_restore(flags); + } + } else { + emt(mtflags); + local_irq_restore(flags); + + } + } + /* Reprogram for interrupt at next soonest timestamp for VPE */ + if (ISVALID(nextstamp)) { + write_c0_compare(nextstamp); + ehb(); + if ((nextstamp - (unsigned long)read_c0_count()) + > (unsigned long)LONG_MAX) + goto repeat; + } +} + + +irqreturn_t c0_compare_interrupt(int irq, void *dev_id) +{ + int cpu = smp_processor_id(); + + /* If we're running SMTC, we've got MIPS MT and therefore MIPS32R2 */ + handle_perf_irq(1); + + if (read_c0_cause() & (1 << 30)) { + /* Clear Count/Compare Interrupt */ + write_c0_compare(read_c0_compare()); + smtc_distribute_timer(cpu_data[cpu].vpe_id); + } + return IRQ_HANDLED; +} + + +int __cpuinit mips_clockevent_init(void) +{ + uint64_t mips_freq = mips_hpt_frequency; + unsigned int cpu = smp_processor_id(); + struct clock_event_device *cd; + unsigned int irq; + int i; + int j; + + if (!cpu_has_counter || !mips_hpt_frequency) + return -ENXIO; + if (cpu == 0) { + for (i = 0; i < num_possible_cpus(); i++) { + smtc_nextinvpe[i] = 0; + for (j = 0; j < num_possible_cpus(); j++) + smtc_nexttime[i][j] = 0L; + } + /* + * SMTC also can't have the usablility test + * run by secondary TCs once Compare is in use. + */ + if (!c0_compare_int_usable()) + return -ENXIO; + } + + /* + * With vectored interrupts things are getting platform specific. + * get_c0_compare_int is a hook to allow a platform to return the + * interrupt number of it's liking. + */ + irq = MIPS_CPU_IRQ_BASE + cp0_compare_irq; + if (get_c0_compare_int) + irq = get_c0_compare_int(); + + cd = &per_cpu(mips_clockevent_device, cpu); + + cd->name = "MIPS"; + cd->features = CLOCK_EVT_FEAT_ONESHOT; + + /* Calculate the min / max delta */ + cd->mult = div_sc((unsigned long) mips_freq, NSEC_PER_SEC, 32); + cd->shift = 32; + cd->max_delta_ns = clockevent_delta2ns(0x7fffffff, cd); + cd->min_delta_ns = clockevent_delta2ns(0x300, cd); + + cd->rating = 300; + cd->irq = irq; + cd->cpumask = cpumask_of_cpu(cpu); + cd->set_next_event = mips_next_event; + cd->set_mode = mips_set_clock_mode; + cd->event_handler = mips_event_handler; + + clockevents_register_device(cd); + + /* + * On SMTC we only want to do the data structure + * initialization and IRQ setup once. + */ + if (cpu) + return 0; + /* + * And we need the hwmask associated with the c0_compare + * vector to be initialized. + */ + irq_hwmask[irq] = (0x100 << cp0_compare_irq); + if (cp0_timer_irq_installed) + return 0; + + cp0_timer_irq_installed = 1; + + setup_irq(irq, &c0_compare_irqaction); + + return 0; +} diff --git a/arch/mips/kernel/cpu-probe.c b/arch/mips/kernel/cpu-probe.c index 335a6ae3d594..e621fda8ab37 100644 --- a/arch/mips/kernel/cpu-probe.c +++ b/arch/mips/kernel/cpu-probe.c @@ -45,18 +45,7 @@ static void r39xx_wait(void) local_irq_enable(); } -/* - * There is a race when WAIT instruction executed with interrupt - * enabled. - * But it is implementation-dependent wheter the pipelie restarts when - * a non-enabled interrupt is requested. - */ -static void r4k_wait(void) -{ - __asm__(" .set mips3 \n" - " wait \n" - " .set mips0 \n"); -} +extern void r4k_wait(void); /* * This variant is preferable as it allows testing need_resched and going to @@ -65,14 +54,18 @@ static void r4k_wait(void) * interrupt is requested" restriction in the MIPS32/MIPS64 architecture makes * using this version a gamble. */ -static void r4k_wait_irqoff(void) +void r4k_wait_irqoff(void) { local_irq_disable(); if (!need_resched()) - __asm__(" .set mips3 \n" + __asm__(" .set push \n" + " .set mips3 \n" " wait \n" - " .set mips0 \n"); + " .set pop \n"); local_irq_enable(); + __asm__(" .globl __pastwait \n" + "__pastwait: \n"); + return; } /* @@ -128,7 +121,7 @@ static int __init wait_disable(char *s) __setup("nowait", wait_disable); -static inline void check_wait(void) +void __init check_wait(void) { struct cpuinfo_mips *c = ¤t_cpu_data; @@ -242,7 +235,6 @@ static inline void check_errata(void) void __init check_bugs32(void) { - check_wait(); check_errata(); } diff --git a/arch/mips/kernel/entry.S b/arch/mips/kernel/entry.S index e29598ae939d..ffa331029e08 100644 --- a/arch/mips/kernel/entry.S +++ b/arch/mips/kernel/entry.S @@ -79,11 +79,6 @@ FEXPORT(syscall_exit) FEXPORT(restore_all) # restore full frame #ifdef CONFIG_MIPS_MT_SMTC -/* Detect and execute deferred IPI "interrupts" */ - LONG_L s0, TI_REGS($28) - LONG_S sp, TI_REGS($28) - jal deferred_smtc_ipi - LONG_S s0, TI_REGS($28) #ifdef CONFIG_MIPS_MT_SMTC_IM_BACKSTOP /* Re-arm any temporarily masked interrupts not explicitly "acked" */ mfc0 v0, CP0_TCSTATUS @@ -112,6 +107,11 @@ FEXPORT(restore_all) # restore full frame xor t0, t0, t3 mtc0 t0, CP0_TCCONTEXT #endif /* CONFIG_MIPS_MT_SMTC_IM_BACKSTOP */ +/* Detect and execute deferred IPI "interrupts" */ + LONG_L s0, TI_REGS($28) + LONG_S sp, TI_REGS($28) + jal deferred_smtc_ipi + LONG_S s0, TI_REGS($28) #endif /* CONFIG_MIPS_MT_SMTC */ .set noat RESTORE_TEMP diff --git a/arch/mips/kernel/genex.S b/arch/mips/kernel/genex.S index c6ada98ee042..01dcbe38fa01 100644 --- a/arch/mips/kernel/genex.S +++ b/arch/mips/kernel/genex.S @@ -20,6 +20,7 @@ #include <asm/stackframe.h> #include <asm/war.h> #include <asm/page.h> +#include <asm/thread_info.h> #define PANIC_PIC(msg) \ .set push; \ @@ -126,7 +127,42 @@ handle_vcei: __FINIT + .align 5 /* 32 byte rollback region */ +LEAF(r4k_wait) + .set push + .set noreorder + /* start of rollback region */ + LONG_L t0, TI_FLAGS($28) + nop + andi t0, _TIF_NEED_RESCHED + bnez t0, 1f + nop + nop + nop + .set mips3 + wait + /* end of rollback region (the region size must be power of two) */ + .set pop +1: + jr ra + END(r4k_wait) + + .macro BUILD_ROLLBACK_PROLOGUE handler + FEXPORT(rollback_\handler) + .set push + .set noat + MFC0 k0, CP0_EPC + PTR_LA k1, r4k_wait + ori k0, 0x1f /* 32 byte rollback region */ + xori k0, 0x1f + bne k0, k1, 9f + MTC0 k0, CP0_EPC +9: + .set pop + .endm + .align 5 +BUILD_ROLLBACK_PROLOGUE handle_int NESTED(handle_int, PT_SIZE, sp) #ifdef CONFIG_TRACE_IRQFLAGS /* @@ -201,6 +237,7 @@ NESTED(except_vec_ejtag_debug, 0, sp) * This prototype is copied to ebase + n*IntCtl.VS and patched * to invoke the handler */ +BUILD_ROLLBACK_PROLOGUE except_vec_vi NESTED(except_vec_vi, 0, sp) SAVE_SOME SAVE_AT @@ -245,8 +282,8 @@ NESTED(except_vec_vi_handler, 0, sp) and t0, a0, t1 #ifdef CONFIG_MIPS_MT_SMTC_IM_BACKSTOP mfc0 t2, CP0_TCCONTEXT - or t0, t0, t2 - mtc0 t0, CP0_TCCONTEXT + or t2, t0, t2 + mtc0 t2, CP0_TCCONTEXT #endif /* CONFIG_MIPS_MT_SMTC_IM_BACKSTOP */ xor t1, t1, t0 mtc0 t1, CP0_STATUS diff --git a/arch/mips/kernel/kgdb.c b/arch/mips/kernel/kgdb.c index c5a8b2d21ca4..6e152c80cd4a 100644 --- a/arch/mips/kernel/kgdb.c +++ b/arch/mips/kernel/kgdb.c @@ -62,13 +62,13 @@ void arch_kgdb_breakpoint(void) static void kgdb_call_nmi_hook(void *ignored) { - kgdb_nmicallback(raw_smp_processor_id(), (void *)0); + kgdb_nmicallback(raw_smp_processor_id(), NULL); } void kgdb_roundup_cpus(unsigned long flags) { local_irq_enable(); - smp_call_function(kgdb_call_nmi_hook, NULL, NULL); + smp_call_function(kgdb_call_nmi_hook, NULL, 0); local_irq_disable(); } @@ -190,9 +190,6 @@ static int kgdb_mips_notify(struct notifier_block *self, unsigned long cmd, struct pt_regs *regs = args->regs; int trap = (regs->cp0_cause & 0x7c) >> 2; - if (fixup_exception(regs)) - return NOTIFY_DONE; - /* Userpace events, ignore. */ if (user_mode(regs)) return NOTIFY_DONE; @@ -239,8 +236,7 @@ int kgdb_arch_handle_exception(int vector, int signo, int err_code, atomic_set(&kgdb_cpu_doing_single_step, -1); if (remcom_in_buffer[0] == 's') - if (kgdb_contthread) - atomic_set(&kgdb_cpu_doing_single_step, cpu); + atomic_set(&kgdb_cpu_doing_single_step, cpu); return 0; } diff --git a/arch/mips/kernel/mips-mt-fpaff.c b/arch/mips/kernel/mips-mt-fpaff.c index df4d3f2f740c..dc9eb72ed9de 100644 --- a/arch/mips/kernel/mips-mt-fpaff.c +++ b/arch/mips/kernel/mips-mt-fpaff.c @@ -159,7 +159,7 @@ __setup("fpaff=", fpaff_thresh); /* * FPU Use Factor empirically derived from experiments on 34K */ -#define FPUSEFACTOR 333 +#define FPUSEFACTOR 2000 static __init int mt_fp_affinity_init(void) { diff --git a/arch/mips/kernel/process.c b/arch/mips/kernel/process.c index b16facd9ea8e..22fc19bbe87f 100644 --- a/arch/mips/kernel/process.c +++ b/arch/mips/kernel/process.c @@ -55,7 +55,7 @@ void __noreturn cpu_idle(void) while (1) { tick_nohz_stop_sched_tick(1); while (!need_resched()) { -#ifdef CONFIG_SMTC_IDLE_HOOK_DEBUG +#ifdef CONFIG_MIPS_MT_SMTC extern void smtc_idle_loop_hook(void); smtc_idle_loop_hook(); @@ -145,17 +145,18 @@ int copy_thread(int nr, unsigned long clone_flags, unsigned long usp, */ p->thread.cp0_status = read_c0_status() & ~(ST0_CU2|ST0_CU1); childregs->cp0_status &= ~(ST0_CU2|ST0_CU1); - clear_tsk_thread_flag(p, TIF_USEDFPU); -#ifdef CONFIG_MIPS_MT_FPAFF +#ifdef CONFIG_MIPS_MT_SMTC /* - * FPU affinity support is cleaner if we track the - * user-visible CPU affinity from the very beginning. - * The generic cpus_allowed mask will already have - * been copied from the parent before copy_thread - * is invoked. + * SMTC restores TCStatus after Status, and the CU bits + * are aliased there. */ - p->thread.user_cpus_allowed = p->cpus_allowed; + childregs->cp0_tcstatus &= ~(ST0_CU2|ST0_CU1); +#endif + clear_tsk_thread_flag(p, TIF_USEDFPU); + +#ifdef CONFIG_MIPS_MT_FPAFF + clear_tsk_thread_flag(p, TIF_FPUBOUND); #endif /* CONFIG_MIPS_MT_FPAFF */ if (clone_flags & CLONE_SETTLS) diff --git a/arch/mips/kernel/ptrace.c b/arch/mips/kernel/ptrace.c index 35234b92b9a5..96ffc9c6d194 100644 --- a/arch/mips/kernel/ptrace.c +++ b/arch/mips/kernel/ptrace.c @@ -238,7 +238,7 @@ long arch_ptrace(struct task_struct *child, long request, long addr, long data) case FPC_EIR: { /* implementation / version register */ unsigned int flags; #ifdef CONFIG_MIPS_MT_SMTC - unsigned int irqflags; + unsigned long irqflags; unsigned int mtflags; #endif /* CONFIG_MIPS_MT_SMTC */ diff --git a/arch/mips/kernel/scall32-o32.S b/arch/mips/kernel/scall32-o32.S index fc4fd4d705e2..5e75a316f6b1 100644 --- a/arch/mips/kernel/scall32-o32.S +++ b/arch/mips/kernel/scall32-o32.S @@ -647,6 +647,12 @@ einval: li v0, -EINVAL sys sys_timerfd_create 2 sys sys_timerfd_gettime 2 sys sys_timerfd_settime 4 + sys sys_signalfd4 4 + sys sys_eventfd2 2 /* 4325 */ + sys sys_epoll_create1 1 + sys sys_dup3 3 + sys sys_pipe2 2 + sys sys_inotify_init1 1 .endm /* We pre-compute the number of _instruction_ bytes needed to diff --git a/arch/mips/kernel/scall64-64.S b/arch/mips/kernel/scall64-64.S index 2b73fd1e4528..3d58204c9d44 100644 --- a/arch/mips/kernel/scall64-64.S +++ b/arch/mips/kernel/scall64-64.S @@ -481,4 +481,10 @@ sys_call_table: PTR sys_timerfd_create /* 5280 */ PTR sys_timerfd_gettime PTR sys_timerfd_settime + PTR sys_signalfd4 + PTR sys_eventfd2 + PTR sys_epoll_create1 /* 5285 */ + PTR sys_dup3 + PTR sys_pipe2 + PTR sys_inotify_init1 .size sys_call_table,.-sys_call_table diff --git a/arch/mips/kernel/scall64-n32.S b/arch/mips/kernel/scall64-n32.S index 2654e75d2fef..da7f1b6ea0fb 100644 --- a/arch/mips/kernel/scall64-n32.S +++ b/arch/mips/kernel/scall64-n32.S @@ -407,4 +407,10 @@ EXPORT(sysn32_call_table) PTR sys_timerfd_create PTR sys_timerfd_gettime /* 5285 */ PTR sys_timerfd_settime + PTR sys_signalfd4 + PTR sys_eventfd2 + PTR sys_epoll_create1 + PTR sys_dup3 /* 5290 */ + PTR sys_pipe2 + PTR sys_inotify_init1 .size sysn32_call_table,.-sysn32_call_table diff --git a/arch/mips/kernel/scall64-o32.S b/arch/mips/kernel/scall64-o32.S index 76167bea5a70..d7cd1aac9ada 100644 --- a/arch/mips/kernel/scall64-o32.S +++ b/arch/mips/kernel/scall64-o32.S @@ -529,4 +529,10 @@ sys_call_table: PTR sys_timerfd_create PTR sys_timerfd_gettime PTR sys_timerfd_settime + PTR compat_sys_signalfd4 + PTR sys_eventfd2 /* 4325 */ + PTR sys_epoll_create1 + PTR sys_dup3 + PTR sys_pipe2 + PTR sys_inotify_init1 .size sys_call_table,.-sys_call_table diff --git a/arch/mips/kernel/setup.c b/arch/mips/kernel/setup.c index 8af84867e74d..16f8edfe5cdc 100644 --- a/arch/mips/kernel/setup.c +++ b/arch/mips/kernel/setup.c @@ -78,7 +78,7 @@ void __init add_memory_region(phys_t start, phys_t size, long type) /* Sanity check */ if (start + size < start) { - printk("Trying to add an invalid memory region, skipped\n"); + pr_warning("Trying to add an invalid memory region, skipped\n"); return; } @@ -92,7 +92,7 @@ void __init add_memory_region(phys_t start, phys_t size, long type) } if (x == BOOT_MEM_MAP_MAX) { - printk("Ooops! Too many entries in the memory map!\n"); + pr_err("Ooops! Too many entries in the memory map!\n"); return; } @@ -108,22 +108,22 @@ static void __init print_memory_map(void) const int field = 2 * sizeof(unsigned long); for (i = 0; i < boot_mem_map.nr_map; i++) { - printk(" memory: %0*Lx @ %0*Lx ", + printk(KERN_INFO " memory: %0*Lx @ %0*Lx ", field, (unsigned long long) boot_mem_map.map[i].size, field, (unsigned long long) boot_mem_map.map[i].addr); switch (boot_mem_map.map[i].type) { case BOOT_MEM_RAM: - printk("(usable)\n"); + printk(KERN_CONT "(usable)\n"); break; case BOOT_MEM_ROM_DATA: - printk("(ROM data)\n"); + printk(KERN_CONT "(ROM data)\n"); break; case BOOT_MEM_RESERVED: - printk("(reserved)\n"); + printk(KERN_CONT "(reserved)\n"); break; default: - printk("type %lu\n", boot_mem_map.map[i].type); + printk(KERN_CONT "type %lu\n", boot_mem_map.map[i].type); break; } } @@ -160,36 +160,39 @@ early_param("rd_size", rd_size_early); static unsigned long __init init_initrd(void) { unsigned long end; - u32 *initrd_header; /* * Board specific code or command line parser should have * already set up initrd_start and initrd_end. In these cases * perfom sanity checks and use them if all looks good. */ - if (initrd_start && initrd_end > initrd_start) - goto sanitize; + if (!initrd_start || initrd_end <= initrd_start) { +#ifdef CONFIG_PROBE_INITRD_HEADER + u32 *initrd_header; - /* - * See if initrd has been added to the kernel image by - * arch/mips/boot/addinitrd.c. In that case a header is - * prepended to initrd and is made up by 8 bytes. The fisrt - * word is a magic number and the second one is the size of - * initrd. Initrd start must be page aligned in any cases. - */ - initrd_header = __va(PAGE_ALIGN(__pa_symbol(&_end) + 8)) - 8; - if (initrd_header[0] != 0x494E5244) + /* + * See if initrd has been added to the kernel image by + * arch/mips/boot/addinitrd.c. In that case a header is + * prepended to initrd and is made up by 8 bytes. The first + * word is a magic number and the second one is the size of + * initrd. Initrd start must be page aligned in any cases. + */ + initrd_header = __va(PAGE_ALIGN(__pa_symbol(&_end) + 8)) - 8; + if (initrd_header[0] != 0x494E5244) + goto disable; + initrd_start = (unsigned long)(initrd_header + 2); + initrd_end = initrd_start + initrd_header[1]; +#else goto disable; - initrd_start = (unsigned long)(initrd_header + 2); - initrd_end = initrd_start + initrd_header[1]; +#endif + } -sanitize: if (initrd_start & ~PAGE_MASK) { - printk(KERN_ERR "initrd start must be page aligned\n"); + pr_err("initrd start must be page aligned\n"); goto disable; } if (initrd_start < PAGE_OFFSET) { - printk(KERN_ERR "initrd start < PAGE_OFFSET\n"); + pr_err("initrd start < PAGE_OFFSET\n"); goto disable; } @@ -221,18 +224,18 @@ static void __init finalize_initrd(void) goto disable; } if (__pa(initrd_end) > PFN_PHYS(max_low_pfn)) { - printk("Initrd extends beyond end of memory"); + printk(KERN_ERR "Initrd extends beyond end of memory"); goto disable; } reserve_bootmem(__pa(initrd_start), size, BOOTMEM_DEFAULT); initrd_below_start_ok = 1; - printk(KERN_INFO "Initial ramdisk at: 0x%lx (%lu bytes)\n", - initrd_start, size); + pr_info("Initial ramdisk at: 0x%lx (%lu bytes)\n", + initrd_start, size); return; disable: - printk(" - disabling initrd\n"); + printk(KERN_CONT " - disabling initrd\n"); initrd_start = 0; initrd_end = 0; } @@ -310,14 +313,12 @@ static void __init bootmem_init(void) if (min_low_pfn >= max_low_pfn) panic("Incorrect memory mapping !!!"); if (min_low_pfn > ARCH_PFN_OFFSET) { - printk(KERN_INFO - "Wasting %lu bytes for tracking %lu unused pages\n", - (min_low_pfn - ARCH_PFN_OFFSET) * sizeof(struct page), - min_low_pfn - ARCH_PFN_OFFSET); + pr_info("Wasting %lu bytes for tracking %lu unused pages\n", + (min_low_pfn - ARCH_PFN_OFFSET) * sizeof(struct page), + min_low_pfn - ARCH_PFN_OFFSET); } else if (min_low_pfn < ARCH_PFN_OFFSET) { - printk(KERN_INFO - "%lu free pages won't be used\n", - ARCH_PFN_OFFSET - min_low_pfn); + pr_info("%lu free pages won't be used\n", + ARCH_PFN_OFFSET - min_low_pfn); } min_low_pfn = ARCH_PFN_OFFSET; @@ -471,7 +472,7 @@ static void __init arch_mem_init(char **cmdline_p) /* call board setup routine */ plat_mem_setup(); - printk("Determined physical RAM map:\n"); + pr_info("Determined physical RAM map:\n"); print_memory_map(); strlcpy(command_line, arcs_cmdline, sizeof(command_line)); @@ -482,7 +483,7 @@ static void __init arch_mem_init(char **cmdline_p) parse_early_param(); if (usermem) { - printk("User-defined physical RAM map:\n"); + pr_info("User-defined physical RAM map:\n"); print_memory_map(); } diff --git a/arch/mips/kernel/smtc.c b/arch/mips/kernel/smtc.c index a516286532ab..897fb2b4751c 100644 --- a/arch/mips/kernel/smtc.c +++ b/arch/mips/kernel/smtc.c @@ -1,4 +1,21 @@ -/* Copyright (C) 2004 Mips Technologies, Inc */ +/* + * 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) 2004 Mips Technologies, Inc + * Copyright (C) 2008 Kevin D. Kissell + */ #include <linux/clockchips.h> #include <linux/kernel.h> @@ -21,7 +38,6 @@ #include <asm/time.h> #include <asm/addrspace.h> #include <asm/smtc.h> -#include <asm/smtc_ipi.h> #include <asm/smtc_proc.h> /* @@ -58,11 +74,6 @@ unsigned long irq_hwmask[NR_IRQS]; asiduse smtc_live_asid[MAX_SMTC_TLBS][MAX_SMTC_ASIDS]; -/* - * Clock interrupt "latch" buffers, per "CPU" - */ - -static atomic_t ipi_timer_latch[NR_CPUS]; /* * Number of InterProcessor Interrupt (IPI) message buffers to allocate @@ -70,7 +81,7 @@ static atomic_t ipi_timer_latch[NR_CPUS]; #define IPIBUF_PER_CPU 4 -static struct smtc_ipi_q IPIQ[NR_CPUS]; +struct smtc_ipi_q IPIQ[NR_CPUS]; static struct smtc_ipi_q freeIPIq; @@ -282,7 +293,7 @@ static void smtc_configure_tlb(void) * phys_cpu_present_map and the logical/physical mappings. */ -int __init mipsmt_build_cpu_map(int start_cpu_slot) +int __init smtc_build_cpu_map(int start_cpu_slot) { int i, ntcs; @@ -325,7 +336,12 @@ static void smtc_tc_setup(int vpe, int tc, int cpu) write_tc_c0_tcstatus((read_tc_c0_tcstatus() & ~(TCSTATUS_TKSU | TCSTATUS_DA | TCSTATUS_IXMT)) | TCSTATUS_A); - write_tc_c0_tccontext(0); + /* + * TCContext gets an offset from the base of the IPIQ array + * to be used in low-level code to detect the presence of + * an active IPI queue + */ + write_tc_c0_tccontext((sizeof(struct smtc_ipi_q) * cpu) << 16); /* Bind tc to vpe */ write_tc_c0_tcbind(vpe); /* In general, all TCs should have the same cpu_data indications */ @@ -336,10 +352,18 @@ static void smtc_tc_setup(int vpe, int tc, int cpu) cpu_data[cpu].options &= ~MIPS_CPU_FPU; cpu_data[cpu].vpe_id = vpe; cpu_data[cpu].tc_id = tc; + /* Multi-core SMTC hasn't been tested, but be prepared */ + cpu_data[cpu].core = (read_vpe_c0_ebase() >> 1) & 0xff; } +/* + * Tweak to get Count registes in as close a sync as possible. + * Value seems good for 34K-class cores. + */ + +#define CP0_SKEW 8 -void mipsmt_prepare_cpus(void) +void smtc_prepare_cpus(int cpus) { int i, vpe, tc, ntc, nvpe, tcpervpe[NR_CPUS], slop, cpu; unsigned long flags; @@ -363,13 +387,13 @@ void mipsmt_prepare_cpus(void) IPIQ[i].head = IPIQ[i].tail = NULL; spin_lock_init(&IPIQ[i].lock); IPIQ[i].depth = 0; - atomic_set(&ipi_timer_latch[i], 0); } /* cpu_data index starts at zero */ cpu = 0; cpu_data[cpu].vpe_id = 0; cpu_data[cpu].tc_id = 0; + cpu_data[cpu].core = (read_c0_ebase() >> 1) & 0xff; cpu++; /* Report on boot-time options */ @@ -484,7 +508,8 @@ void mipsmt_prepare_cpus(void) write_vpe_c0_compare(0); /* Propagate Config7 */ write_vpe_c0_config7(read_c0_config7()); - write_vpe_c0_count(read_c0_count()); + write_vpe_c0_count(read_c0_count() + CP0_SKEW); + ehb(); } /* enable multi-threading within VPE */ write_vpe_c0_vpecontrol(read_vpe_c0_vpecontrol() | VPECONTROL_TE); @@ -556,7 +581,7 @@ void mipsmt_prepare_cpus(void) void __cpuinit smtc_boot_secondary(int cpu, struct task_struct *idle) { extern u32 kernelsp[NR_CPUS]; - long flags; + unsigned long flags; int mtflags; LOCK_MT_PRA(); @@ -585,24 +610,22 @@ void __cpuinit smtc_boot_secondary(int cpu, struct task_struct *idle) void smtc_init_secondary(void) { - /* - * Start timer on secondary VPEs if necessary. - * plat_timer_setup has already have been invoked by init/main - * on "boot" TC. Like per_cpu_trap_init() hack, this assumes that - * SMTC init code assigns TCs consdecutively and in ascending order - * to across available VPEs. - */ - if (((read_c0_tcbind() & TCBIND_CURTC) != 0) && - ((read_c0_tcbind() & TCBIND_CURVPE) - != cpu_data[smp_processor_id() - 1].vpe_id)){ - write_c0_compare(read_c0_count() + mips_hpt_frequency/HZ); - } - local_irq_enable(); } void smtc_smp_finish(void) { + int cpu = smp_processor_id(); + + /* + * Lowest-numbered CPU per VPE starts a clock tick. + * Like per_cpu_trap_init() hack, this assumes that + * SMTC init code assigns TCs consdecutively and + * in ascending order across available VPEs. + */ + if (cpu > 0 && (cpu_data[cpu].vpe_id != cpu_data[cpu - 1].vpe_id)) + write_c0_compare(read_c0_count() + mips_hpt_frequency/HZ); + printk("TC %d going on-line as CPU %d\n", cpu_data[smp_processor_id()].tc_id, smp_processor_id()); } @@ -753,8 +776,10 @@ void smtc_send_ipi(int cpu, int type, unsigned int action) { int tcstatus; struct smtc_ipi *pipi; - long flags; + unsigned long flags; int mtflags; + unsigned long tcrestart; + extern void r4k_wait_irqoff(void), __pastwait(void); if (cpu == smp_processor_id()) { printk("Cannot Send IPI to self!\n"); @@ -771,8 +796,6 @@ void smtc_send_ipi(int cpu, int type, unsigned int action) pipi->arg = (void *)action; pipi->dest = cpu; if (cpu_data[cpu].vpe_id != cpu_data[smp_processor_id()].vpe_id) { - if (type == SMTC_CLOCK_TICK) - atomic_inc(&ipi_timer_latch[cpu]); /* If not on same VPE, enqueue and send cross-VPE interrupt */ smtc_ipi_nq(&IPIQ[cpu], pipi); LOCK_CORE_PRA(); @@ -800,22 +823,29 @@ void smtc_send_ipi(int cpu, int type, unsigned int action) if ((tcstatus & TCSTATUS_IXMT) != 0) { /* - * Spin-waiting here can deadlock, - * so we queue the message for the target TC. + * If we're in the the irq-off version of the wait + * loop, we need to force exit from the wait and + * do a direct post of the IPI. + */ + if (cpu_wait == r4k_wait_irqoff) { + tcrestart = read_tc_c0_tcrestart(); + if (tcrestart >= (unsigned long)r4k_wait_irqoff + && tcrestart < (unsigned long)__pastwait) { + write_tc_c0_tcrestart(__pastwait); + tcstatus &= ~TCSTATUS_IXMT; + write_tc_c0_tcstatus(tcstatus); + goto postdirect; + } + } + /* + * Otherwise we queue the message for the target TC + * to pick up when he does a local_irq_restore() */ write_tc_c0_tchalt(0); UNLOCK_CORE_PRA(); - /* Try to reduce redundant timer interrupt messages */ - if (type == SMTC_CLOCK_TICK) { - if (atomic_postincrement(&ipi_timer_latch[cpu])!=0){ - smtc_ipi_nq(&freeIPIq, pipi); - return; - } - } smtc_ipi_nq(&IPIQ[cpu], pipi); } else { - if (type == SMTC_CLOCK_TICK) - atomic_inc(&ipi_timer_latch[cpu]); +postdirect: post_direct_ipi(cpu, pipi); write_tc_c0_tchalt(0); UNLOCK_CORE_PRA(); @@ -883,7 +913,7 @@ static void ipi_call_interrupt(void) smp_call_function_interrupt(); } -DECLARE_PER_CPU(struct clock_event_device, smtc_dummy_clockevent_device); +DECLARE_PER_CPU(struct clock_event_device, mips_clockevent_device); void ipi_decode(struct smtc_ipi *pipi) { @@ -891,20 +921,13 @@ void ipi_decode(struct smtc_ipi *pipi) struct clock_event_device *cd; void *arg_copy = pipi->arg; int type_copy = pipi->type; - int ticks; - smtc_ipi_nq(&freeIPIq, pipi); switch (type_copy) { case SMTC_CLOCK_TICK: irq_enter(); kstat_this_cpu.irqs[MIPS_CPU_IRQ_BASE + 1]++; - cd = &per_cpu(smtc_dummy_clockevent_device, cpu); - ticks = atomic_read(&ipi_timer_latch[cpu]); - atomic_sub(ticks, &ipi_timer_latch[cpu]); - while (ticks) { - cd->event_handler(cd); - ticks--; - } + cd = &per_cpu(mips_clockevent_device, cpu); + cd->event_handler(cd); irq_exit(); break; @@ -937,24 +960,48 @@ void ipi_decode(struct smtc_ipi *pipi) } } +/* + * Similar to smtc_ipi_replay(), but invoked from context restore, + * so it reuses the current exception frame rather than set up a + * new one with self_ipi. + */ + void deferred_smtc_ipi(void) { - struct smtc_ipi *pipi; - unsigned long flags; -/* DEBUG */ - int q = smp_processor_id(); + int cpu = smp_processor_id(); /* * Test is not atomic, but much faster than a dequeue, * and the vast majority of invocations will have a null queue. + * If irq_disabled when this was called, then any IPIs queued + * after we test last will be taken on the next irq_enable/restore. + * If interrupts were enabled, then any IPIs added after the + * last test will be taken directly. */ - if (IPIQ[q].head != NULL) { - while((pipi = smtc_ipi_dq(&IPIQ[q])) != NULL) { - /* ipi_decode() should be called with interrupts off */ - local_irq_save(flags); + + while (IPIQ[cpu].head != NULL) { + struct smtc_ipi_q *q = &IPIQ[cpu]; + struct smtc_ipi *pipi; + unsigned long flags; + + /* + * It may be possible we'll come in with interrupts + * already enabled. + */ + local_irq_save(flags); + + spin_lock(&q->lock); + pipi = __smtc_ipi_dq(q); + spin_unlock(&q->lock); + if (pipi != NULL) ipi_decode(pipi); - local_irq_restore(flags); - } + /* + * The use of the __raw_local restore isn't + * as obviously necessary here as in smtc_ipi_replay(), + * but it's more efficient, given that we're already + * running down the IPI queue. + */ + __raw_local_irq_restore(flags); } } @@ -975,7 +1022,7 @@ static irqreturn_t ipi_interrupt(int irq, void *dev_idm) struct smtc_ipi *pipi; unsigned long tcstatus; int sent; - long flags; + unsigned long flags; unsigned int mtflags; unsigned int vpflags; @@ -1066,55 +1113,53 @@ static void setup_cross_vpe_interrupts(unsigned int nvpe) /* * SMTC-specific hacks invoked from elsewhere in the kernel. - * - * smtc_ipi_replay is called from raw_local_irq_restore which is only ever - * called with interrupts disabled. We do rely on interrupts being disabled - * here because using spin_lock_irqsave()/spin_unlock_irqrestore() would - * result in a recursive call to raw_local_irq_restore(). */ -static void __smtc_ipi_replay(void) + /* + * smtc_ipi_replay is called from raw_local_irq_restore + */ + +void smtc_ipi_replay(void) { unsigned int cpu = smp_processor_id(); /* * To the extent that we've ever turned interrupts off, * we may have accumulated deferred IPIs. This is subtle. - * If we use the smtc_ipi_qdepth() macro, we'll get an - * exact number - but we'll also disable interrupts - * and create a window of failure where a new IPI gets - * queued after we test the depth but before we re-enable - * interrupts. So long as IXMT never gets set, however, * we should be OK: If we pick up something and dispatch * it here, that's great. If we see nothing, but concurrent * with this operation, another TC sends us an IPI, IXMT * is clear, and we'll handle it as a real pseudo-interrupt - * and not a pseudo-pseudo interrupt. + * and not a pseudo-pseudo interrupt. The important thing + * is to do the last check for queued message *after* the + * re-enabling of interrupts. */ - if (IPIQ[cpu].depth > 0) { - while (1) { - struct smtc_ipi_q *q = &IPIQ[cpu]; - struct smtc_ipi *pipi; - extern void self_ipi(struct smtc_ipi *); - - spin_lock(&q->lock); - pipi = __smtc_ipi_dq(q); - spin_unlock(&q->lock); - if (!pipi) - break; + while (IPIQ[cpu].head != NULL) { + struct smtc_ipi_q *q = &IPIQ[cpu]; + struct smtc_ipi *pipi; + unsigned long flags; + + /* + * It's just possible we'll come in with interrupts + * already enabled. + */ + local_irq_save(flags); + + spin_lock(&q->lock); + pipi = __smtc_ipi_dq(q); + spin_unlock(&q->lock); + /* + ** But use a raw restore here to avoid recursion. + */ + __raw_local_irq_restore(flags); + if (pipi) { self_ipi(pipi); smtc_cpu_stats[cpu].selfipis++; } } } -void smtc_ipi_replay(void) -{ - raw_local_irq_disable(); - __smtc_ipi_replay(); -} - EXPORT_SYMBOL(smtc_ipi_replay); void smtc_idle_loop_hook(void) @@ -1193,40 +1238,13 @@ void smtc_idle_loop_hook(void) } } - /* - * Now that we limit outstanding timer IPIs, check for hung TC - */ - for (tc = 0; tc < NR_CPUS; tc++) { - /* Don't check ourself - we'll dequeue IPIs just below */ - if ((tc != smp_processor_id()) && - atomic_read(&ipi_timer_latch[tc]) > timerq_limit) { - if (clock_hang_reported[tc] == 0) { - pdb_msg += sprintf(pdb_msg, - "TC %d looks hung with timer latch at %d\n", - tc, atomic_read(&ipi_timer_latch[tc])); - clock_hang_reported[tc]++; - } - } - } emt(mtflags); local_irq_restore(flags); if (pdb_msg != &id_ho_db_msg[0]) printk("CPU%d: %s", smp_processor_id(), id_ho_db_msg); #endif /* CONFIG_SMTC_IDLE_HOOK_DEBUG */ - /* - * Replay any accumulated deferred IPIs. If "Instant Replay" - * is in use, there should never be any. - */ -#ifndef CONFIG_MIPS_MT_SMTC_INSTANT_REPLAY - { - unsigned long flags; - - local_irq_save(flags); - __smtc_ipi_replay(); - local_irq_restore(flags); - } -#endif /* CONFIG_MIPS_MT_SMTC_INSTANT_REPLAY */ + smtc_ipi_replay(); } void smtc_soft_dump(void) @@ -1242,10 +1260,6 @@ void smtc_soft_dump(void) printk("%d: %ld\n", i, smtc_cpu_stats[i].selfipis); } smtc_ipi_qdump(); - printk("Timer IPI Backlogs:\n"); - for (i=0; i < NR_CPUS; i++) { - printk("%d: %d\n", i, atomic_read(&ipi_timer_latch[i])); - } printk("%d Recoveries of \"stolen\" FPU\n", atomic_read(&smtc_fpu_recoveries)); } diff --git a/arch/mips/kernel/traps.c b/arch/mips/kernel/traps.c index 426cced1e9dc..b602ac6eb47d 100644 --- a/arch/mips/kernel/traps.c +++ b/arch/mips/kernel/traps.c @@ -46,6 +46,9 @@ #include <asm/types.h> #include <asm/stacktrace.h> +extern void check_wait(void); +extern asmlinkage void r4k_wait(void); +extern asmlinkage void rollback_handle_int(void); extern asmlinkage void handle_int(void); extern asmlinkage void handle_tlbm(void); extern asmlinkage void handle_tlbl(void); @@ -373,8 +376,8 @@ void __noreturn die(const char * str, const struct pt_regs * regs) do_exit(SIGSEGV); } -extern const struct exception_table_entry __start___dbe_table[]; -extern const struct exception_table_entry __stop___dbe_table[]; +extern struct exception_table_entry __start___dbe_table[]; +extern struct exception_table_entry __stop___dbe_table[]; __asm__( " .section __dbe_table, \"a\"\n" @@ -822,8 +825,10 @@ static void mt_ase_fp_affinity(void) if (cpus_intersects(current->cpus_allowed, mt_fpu_cpumask)) { cpumask_t tmask; - cpus_and(tmask, current->thread.user_cpus_allowed, - mt_fpu_cpumask); + current->thread.user_cpus_allowed + = current->cpus_allowed; + cpus_and(tmask, current->cpus_allowed, + mt_fpu_cpumask); set_cpus_allowed(current, tmask); set_thread_flag(TIF_FPUBOUND); } @@ -1200,7 +1205,7 @@ void *set_except_vector(int n, void *addr) if (n == 0 && cpu_has_divec) { *(u32 *)(ebase + 0x200) = 0x08000000 | (0x03ffffff & (handler >> 2)); - flush_icache_range(ebase + 0x200, ebase + 0x204); + local_flush_icache_range(ebase + 0x200, ebase + 0x204); } return (void *)old_handler; } @@ -1251,6 +1256,9 @@ static void *set_vi_srs_handler(int n, vi_handler_t addr, int srs) extern char except_vec_vi, except_vec_vi_lui; extern char except_vec_vi_ori, except_vec_vi_end; + extern char rollback_except_vec_vi; + char *vec_start = (cpu_wait == r4k_wait) ? + &rollback_except_vec_vi : &except_vec_vi; #ifdef CONFIG_MIPS_MT_SMTC /* * We need to provide the SMTC vectored interrupt handler @@ -1258,11 +1266,11 @@ static void *set_vi_srs_handler(int n, vi_handler_t addr, int srs) * Status.IM bit to be masked before going there. */ extern char except_vec_vi_mori; - const int mori_offset = &except_vec_vi_mori - &except_vec_vi; + const int mori_offset = &except_vec_vi_mori - vec_start; #endif /* CONFIG_MIPS_MT_SMTC */ - const int handler_len = &except_vec_vi_end - &except_vec_vi; - const int lui_offset = &except_vec_vi_lui - &except_vec_vi; - const int ori_offset = &except_vec_vi_ori - &except_vec_vi; + const int handler_len = &except_vec_vi_end - vec_start; + const int lui_offset = &except_vec_vi_lui - vec_start; + const int ori_offset = &except_vec_vi_ori - vec_start; if (handler_len > VECTORSPACING) { /* @@ -1272,7 +1280,7 @@ static void *set_vi_srs_handler(int n, vi_handler_t addr, int srs) panic("VECTORSPACING too small"); } - memcpy(b, &except_vec_vi, handler_len); + memcpy(b, vec_start, handler_len); #ifdef CONFIG_MIPS_MT_SMTC BUG_ON(n > 7); /* Vector index %d exceeds SMTC maximum. */ @@ -1283,7 +1291,8 @@ static void *set_vi_srs_handler(int n, vi_handler_t addr, int srs) *w = (*w & 0xffff0000) | (((u32)handler >> 16) & 0xffff); w = (u32 *)(b + ori_offset); *w = (*w & 0xffff0000) | ((u32)handler & 0xffff); - flush_icache_range((unsigned long)b, (unsigned long)(b+handler_len)); + local_flush_icache_range((unsigned long)b, + (unsigned long)(b+handler_len)); } else { /* @@ -1295,7 +1304,8 @@ static void *set_vi_srs_handler(int n, vi_handler_t addr, int srs) w = (u32 *)b; *w++ = 0x08000000 | (((u32)handler >> 2) & 0x03fffff); /* j handler */ *w = 0; - flush_icache_range((unsigned long)b, (unsigned long)(b+8)); + local_flush_icache_range((unsigned long)b, + (unsigned long)(b+8)); } return (void *)old_handler; @@ -1515,7 +1525,7 @@ void __cpuinit per_cpu_trap_init(void) void __init set_handler(unsigned long offset, void *addr, unsigned long size) { memcpy((void *)(ebase + offset), addr, size); - flush_icache_range(ebase + offset, ebase + offset + size); + local_flush_icache_range(ebase + offset, ebase + offset + size); } static char panic_null_cerr[] __cpuinitdata = @@ -1552,6 +1562,10 @@ void __init trap_init(void) extern char except_vec3_generic, except_vec3_r4000; extern char except_vec4; unsigned long i; + int rollback; + + check_wait(); + rollback = (cpu_wait == r4k_wait); #if defined(CONFIG_KGDB) if (kgdb_early_setup) @@ -1616,7 +1630,7 @@ void __init trap_init(void) if (board_be_init) board_be_init(); - set_except_vector(0, handle_int); + set_except_vector(0, rollback ? rollback_handle_int : handle_int); set_except_vector(1, handle_tlbm); set_except_vector(2, handle_tlbl); set_except_vector(3, handle_tlbs); @@ -1680,6 +1694,8 @@ void __init trap_init(void) signal32_init(); #endif - flush_icache_range(ebase, ebase + 0x400); + local_flush_icache_range(ebase, ebase + 0x400); flush_tlb_handlers(); + + sort_extable(__start___dbe_table, __stop___dbe_table); } diff --git a/arch/mips/kernel/vmlinux.lds.S b/arch/mips/kernel/vmlinux.lds.S index b5470ceb418b..afb119f35682 100644 --- a/arch/mips/kernel/vmlinux.lds.S +++ b/arch/mips/kernel/vmlinux.lds.S @@ -36,6 +36,7 @@ SECTIONS SCHED_TEXT LOCK_TEXT KPROBES_TEXT + *(.text.*) *(.fixup) *(.gnu.warning) } :text = 0 |