From 5168ae50a66e3ff7184c2b16d661bd6d70367e50 Mon Sep 17 00:00:00 2001 From: Steven Rostedt Date: Thu, 3 Jun 2010 09:36:50 -0400 Subject: tracing: Remove ftrace_preempt_disable/enable The ftrace_preempt_disable/enable functions were to address a recursive race caused by the function tracer. The function tracer traces all functions which makes it easily susceptible to recursion. One area was preempt_enable(). This would call the scheduler and the schedulre would call the function tracer and loop. (So was it thought). The ftrace_preempt_disable/enable was made to protect against recursion inside the scheduler by storing the NEED_RESCHED flag. If it was set before the ftrace_preempt_disable() it would not call schedule on ftrace_preempt_enable(), thinking that if it was set before then it would have already scheduled unless it was already in the scheduler. This worked fine except in the case of SMP, where another task would set the NEED_RESCHED flag for a task on another CPU, and then kick off an IPI to trigger it. This could cause the NEED_RESCHED to be saved at ftrace_preempt_disable() but the IPI to arrive in the the preempt disabled section. The ftrace_preempt_enable() would not call the scheduler because the flag was already set before entring the section. This bug would cause a missed preemption check and cause lower latencies. Investigating further, I found that the recusion caused by the function tracer was not due to schedule(), but due to preempt_schedule(). Now that preempt_schedule is completely annotated with notrace, the recusion no longer is an issue. Reported-by: Thomas Gleixner Signed-off-by: Steven Rostedt --- kernel/trace/trace_clock.c | 5 ++--- 1 file changed, 2 insertions(+), 3 deletions(-) (limited to 'kernel/trace/trace_clock.c') diff --git a/kernel/trace/trace_clock.c b/kernel/trace/trace_clock.c index 9d589d8dcd1..52fda6c04ac 100644 --- a/kernel/trace/trace_clock.c +++ b/kernel/trace/trace_clock.c @@ -32,16 +32,15 @@ u64 notrace trace_clock_local(void) { u64 clock; - int resched; /* * sched_clock() is an architecture implemented, fast, scalable, * lockless clock. It is not guaranteed to be coherent across * CPUs, nor across CPU idle events. */ - resched = ftrace_preempt_disable(); + preempt_disable_notrace(); clock = sched_clock(); - ftrace_preempt_enable(resched); + preempt_enable_notrace(); return clock; } -- cgit v1.2.3-70-g09d2 From c676329abb2b8359d9a5d734dec0c81779823fd6 Mon Sep 17 00:00:00 2001 From: Peter Zijlstra Date: Tue, 25 May 2010 10:48:51 +0200 Subject: sched_clock: Add local_clock() API and improve documentation For people who otherwise get to write: cpu_clock(smp_processor_id()), there is now: local_clock(). Also, as per suggestion from Andrew, provide some documentation on the various clock interfaces, and minimize the unsigned long long vs u64 mess. Signed-off-by: Peter Zijlstra Cc: Andrew Morton Cc: Linus Torvalds Cc: Jens Axboe LKML-Reference: <1275052414.1645.52.camel@laptop> Signed-off-by: Ingo Molnar --- arch/parisc/kernel/ftrace.c | 4 +- include/linux/sched.h | 37 ++++++++++-------- kernel/lockdep.c | 2 +- kernel/perf_event.c | 2 +- kernel/rcutorture.c | 3 +- kernel/sched.c | 2 +- kernel/sched_clock.c | 95 ++++++++++++++++++++++++++++++++++++++++----- kernel/trace/trace_clock.c | 2 +- 8 files changed, 113 insertions(+), 34 deletions(-) (limited to 'kernel/trace/trace_clock.c') diff --git a/arch/parisc/kernel/ftrace.c b/arch/parisc/kernel/ftrace.c index 9877372ffdb..5beb97bafbb 100644 --- a/arch/parisc/kernel/ftrace.c +++ b/arch/parisc/kernel/ftrace.c @@ -82,7 +82,7 @@ unsigned long ftrace_return_to_handler(unsigned long retval0, unsigned long ret; pop_return_trace(&trace, &ret); - trace.rettime = cpu_clock(raw_smp_processor_id()); + trace.rettime = local_clock(); ftrace_graph_return(&trace); if (unlikely(!ret)) { @@ -126,7 +126,7 @@ void prepare_ftrace_return(unsigned long *parent, unsigned long self_addr) return; } - calltime = cpu_clock(raw_smp_processor_id()); + calltime = local_clock(); if (push_return_trace(old, calltime, self_addr, &trace.depth) == -EBUSY) { diff --git a/include/linux/sched.h b/include/linux/sched.h index edc3dd168d8..c2d4316a04b 100644 --- a/include/linux/sched.h +++ b/include/linux/sched.h @@ -1791,20 +1791,23 @@ static inline int set_cpus_allowed(struct task_struct *p, cpumask_t new_mask) #endif /* - * Architectures can set this to 1 if they have specified - * CONFIG_HAVE_UNSTABLE_SCHED_CLOCK in their arch Kconfig, - * but then during bootup it turns out that sched_clock() - * is reliable after all: + * Do not use outside of architecture code which knows its limitations. + * + * sched_clock() has no promise of monotonicity or bounded drift between + * CPUs, use (which you should not) requires disabling IRQs. + * + * Please use one of the three interfaces below. */ -#ifdef CONFIG_HAVE_UNSTABLE_SCHED_CLOCK -extern int sched_clock_stable; -#endif - -/* ftrace calls sched_clock() directly */ extern unsigned long long notrace sched_clock(void); +/* + * See the comment in kernel/sched_clock.c + */ +extern u64 cpu_clock(int cpu); +extern u64 local_clock(void); +extern u64 sched_clock_cpu(int cpu); + extern void sched_clock_init(void); -extern u64 sched_clock_cpu(int cpu); #ifndef CONFIG_HAVE_UNSTABLE_SCHED_CLOCK static inline void sched_clock_tick(void) @@ -1819,17 +1822,19 @@ static inline void sched_clock_idle_wakeup_event(u64 delta_ns) { } #else +/* + * Architectures can set this to 1 if they have specified + * CONFIG_HAVE_UNSTABLE_SCHED_CLOCK in their arch Kconfig, + * but then during bootup it turns out that sched_clock() + * is reliable after all: + */ +extern int sched_clock_stable; + extern void sched_clock_tick(void); extern void sched_clock_idle_sleep_event(void); extern void sched_clock_idle_wakeup_event(u64 delta_ns); #endif -/* - * For kernel-internal use: high-speed (but slightly incorrect) per-cpu - * clock constructed from sched_clock(): - */ -extern unsigned long long cpu_clock(int cpu); - extern unsigned long long task_sched_runtime(struct task_struct *task); extern unsigned long long thread_group_sched_runtime(struct task_struct *task); diff --git a/kernel/lockdep.c b/kernel/lockdep.c index 54286798c37..f2852a51023 100644 --- a/kernel/lockdep.c +++ b/kernel/lockdep.c @@ -146,7 +146,7 @@ static DEFINE_PER_CPU(struct lock_class_stats[MAX_LOCKDEP_KEYS], static inline u64 lockstat_clock(void) { - return cpu_clock(smp_processor_id()); + return local_clock(); } static int lock_point(unsigned long points[], unsigned long ip) diff --git a/kernel/perf_event.c b/kernel/perf_event.c index 31d6afe9259..109c5ec8893 100644 --- a/kernel/perf_event.c +++ b/kernel/perf_event.c @@ -214,7 +214,7 @@ static void perf_unpin_context(struct perf_event_context *ctx) static inline u64 perf_clock(void) { - return cpu_clock(raw_smp_processor_id()); + return local_clock(); } /* diff --git a/kernel/rcutorture.c b/kernel/rcutorture.c index 6535ac8bc6a..2e2726d790b 100644 --- a/kernel/rcutorture.c +++ b/kernel/rcutorture.c @@ -239,8 +239,7 @@ static unsigned long rcu_random(struct rcu_random_state *rrsp) { if (--rrsp->rrs_count < 0) { - rrsp->rrs_state += - (unsigned long)cpu_clock(raw_smp_processor_id()); + rrsp->rrs_state += (unsigned long)local_clock(); rrsp->rrs_count = RCU_RANDOM_REFRESH; } rrsp->rrs_state = rrsp->rrs_state * RCU_RANDOM_MULT + RCU_RANDOM_ADD; diff --git a/kernel/sched.c b/kernel/sched.c index 8f351c56567..3abd8f780da 100644 --- a/kernel/sched.c +++ b/kernel/sched.c @@ -1647,7 +1647,7 @@ static void update_shares(struct sched_domain *sd) if (root_task_group_empty()) return; - now = cpu_clock(raw_smp_processor_id()); + now = local_clock(); elapsed = now - sd->last_update; if (elapsed >= (s64)(u64)sysctl_sched_shares_ratelimit) { diff --git a/kernel/sched_clock.c b/kernel/sched_clock.c index 906a0f718cb..52f1a149bfb 100644 --- a/kernel/sched_clock.c +++ b/kernel/sched_clock.c @@ -10,19 +10,55 @@ * Ingo Molnar * Guillaume Chazarain * - * Create a semi stable clock from a mixture of other events, including: - * - gtod + * + * What: + * + * cpu_clock(i) provides a fast (execution time) high resolution + * clock with bounded drift between CPUs. The value of cpu_clock(i) + * is monotonic for constant i. The timestamp returned is in nanoseconds. + * + * ######################### BIG FAT WARNING ########################## + * # when comparing cpu_clock(i) to cpu_clock(j) for i != j, time can # + * # go backwards !! # + * #################################################################### + * + * There is no strict promise about the base, although it tends to start + * at 0 on boot (but people really shouldn't rely on that). + * + * cpu_clock(i) -- can be used from any context, including NMI. + * sched_clock_cpu(i) -- must be used with local IRQs disabled (implied by NMI) + * local_clock() -- is cpu_clock() on the current cpu. + * + * How: + * + * The implementation either uses sched_clock() when + * !CONFIG_HAVE_UNSTABLE_SCHED_CLOCK, which means in that case the + * sched_clock() is assumed to provide these properties (mostly it means + * the architecture provides a globally synchronized highres time source). + * + * Otherwise it tries to create a semi stable clock from a mixture of other + * clocks, including: + * + * - GTOD (clock monotomic) * - sched_clock() * - explicit idle events * - * We use gtod as base and the unstable clock deltas. The deltas are filtered, - * making it monotonic and keeping it within an expected window. + * We use GTOD as base and use sched_clock() deltas to improve resolution. The + * deltas are filtered to provide monotonicity and keeping it within an + * expected window. * * Furthermore, explicit sleep and wakeup hooks allow us to account for time * that is otherwise invisible (TSC gets stopped). * - * The clock: sched_clock_cpu() is monotonic per cpu, and should be somewhat - * consistent between cpus (never more than 2 jiffies difference). + * + * Notes: + * + * The !IRQ-safetly of sched_clock() and sched_clock_cpu() comes from things + * like cpufreq interrupts that can change the base clock (TSC) multiplier + * and cause funny jumps in time -- although the filtering provided by + * sched_clock_cpu() should mitigate serious artifacts we cannot rely on it + * in general since for !CONFIG_HAVE_UNSTABLE_SCHED_CLOCK we fully rely on + * sched_clock(). */ #include #include @@ -170,6 +206,11 @@ again: return val; } +/* + * Similar to cpu_clock(), but requires local IRQs to be disabled. + * + * See cpu_clock(). + */ u64 sched_clock_cpu(int cpu) { struct sched_clock_data *scd; @@ -237,9 +278,19 @@ void sched_clock_idle_wakeup_event(u64 delta_ns) } EXPORT_SYMBOL_GPL(sched_clock_idle_wakeup_event); -unsigned long long cpu_clock(int cpu) +/* + * As outlined at the top, provides a fast, high resolution, nanosecond + * time source that is monotonic per cpu argument and has bounded drift + * between cpus. + * + * ######################### BIG FAT WARNING ########################## + * # when comparing cpu_clock(i) to cpu_clock(j) for i != j, time can # + * # go backwards !! # + * #################################################################### + */ +u64 cpu_clock(int cpu) { - unsigned long long clock; + u64 clock; unsigned long flags; local_irq_save(flags); @@ -249,6 +300,25 @@ unsigned long long cpu_clock(int cpu) return clock; } +/* + * Similar to cpu_clock() for the current cpu. Time will only be observed + * to be monotonic if care is taken to only compare timestampt taken on the + * same CPU. + * + * See cpu_clock(). + */ +u64 local_clock(void) +{ + u64 clock; + unsigned long flags; + + local_irq_save(flags); + clock = sched_clock_cpu(smp_processor_id()); + local_irq_restore(flags); + + return clock; +} + #else /* CONFIG_HAVE_UNSTABLE_SCHED_CLOCK */ void sched_clock_init(void) @@ -264,12 +334,17 @@ u64 sched_clock_cpu(int cpu) return sched_clock(); } - -unsigned long long cpu_clock(int cpu) +u64 cpu_clock(int cpu) { return sched_clock_cpu(cpu); } +u64 local_clock(void) +{ + return sched_clock_cpu(0); +} + #endif /* CONFIG_HAVE_UNSTABLE_SCHED_CLOCK */ EXPORT_SYMBOL_GPL(cpu_clock); +EXPORT_SYMBOL_GPL(local_clock); diff --git a/kernel/trace/trace_clock.c b/kernel/trace/trace_clock.c index 9d589d8dcd1..1723e2b8c58 100644 --- a/kernel/trace/trace_clock.c +++ b/kernel/trace/trace_clock.c @@ -56,7 +56,7 @@ u64 notrace trace_clock_local(void) */ u64 notrace trace_clock(void) { - return cpu_clock(raw_smp_processor_id()); + return local_clock(); } -- cgit v1.2.3-70-g09d2