The patch titled
highres/dyntick: prevent xtime lock contention
has been added to the -mm tree. Its filename is
highres-dyntick-prevent-xtime-lock-contention.patch
*** Remember to use Documentation/SubmitChecklist when testing your code ***
See http://www.zip.com.au/~akpm/linux/patches/stuff/added-to-mm.txt to find
out what to do about this
------------------------------------------------------
Subject: highres/dyntick: prevent xtime lock contention
From: Thomas Gleixner <tglx@xxxxxxxxxxxxx>
While the !highres/!dyntick code assigns the duty of the do_timer() call to
one specific CPU, this was dropped in the highres/dyntick part during
development.
Steven Rostedt discovered the xtime lock contention on highres/dyntick due
to several CPUs trying to update jiffies.
Add the single CPU assignement back. In the dyntick case this needs to be
handled carefully, as the CPU which has the do_timer() duty must drop the
assignement and let it be grabbed by another CPU, which is active.
Otherwise the do_timer() calls would not happen during the long sleep.
Signed-off-by: Thomas Gleixner <tglx@xxxxxxxxxxxxx>
Acked-by: Ingo Molnar <mingo@xxxxxxx>
Cc: Steven Rostedt <rostedt@xxxxxxxxxxx>
Signed-off-by: Andrew Morton <akpm@xxxxxxxxxxxxxxxxxxxx>
---
kernel/time/tick-common.c | 8 +++++-
kernel/time/tick-internal.h | 1
kernel/time/tick-sched.c | 42 ++++++++++++++++++++++++++++++++--
3 files changed, 48 insertions(+), 3 deletions(-)
diff -puN kernel/time/tick-common.c~highres-dyntick-prevent-xtime-lock-contention kernel/time/tick-common.c
--- a/kernel/time/tick-common.c~highres-dyntick-prevent-xtime-lock-contention
+++ a/kernel/time/tick-common.c
@@ -31,7 +31,7 @@ DEFINE_PER_CPU(struct tick_device, tick_
*/
ktime_t tick_next_period;
ktime_t tick_period;
-static int tick_do_timer_cpu = -1;
+int tick_do_timer_cpu __read_mostly = -1;
DEFINE_SPINLOCK(tick_device_lock);
/*
@@ -295,6 +295,12 @@ static void tick_shutdown(unsigned int *
clockevents_exchange_device(dev, NULL);
td->evtdev = NULL;
}
+ /* Transfer the do_timer job away from this cpu */
+ if (*cpup == tick_do_timer_cpu) {
+ int cpu = first_cpu(cpu_online_map);
+
+ tick_do_timer_cpu = (cpu != NR_CPUS) ? cpu : -1;
+ }
spin_unlock_irqrestore(&tick_device_lock, flags);
}
diff -puN kernel/time/tick-internal.h~highres-dyntick-prevent-xtime-lock-contention kernel/time/tick-internal.h
--- a/kernel/time/tick-internal.h~highres-dyntick-prevent-xtime-lock-contention
+++ a/kernel/time/tick-internal.h
@@ -5,6 +5,7 @@ DECLARE_PER_CPU(struct tick_device, tick
extern spinlock_t tick_device_lock;
extern ktime_t tick_next_period;
extern ktime_t tick_period;
+extern int tick_do_timer_cpu __read_mostly;
extern void tick_setup_periodic(struct clock_event_device *dev, int broadcast);
extern void tick_handle_periodic(struct clock_event_device *dev);
diff -puN kernel/time/tick-sched.c~highres-dyntick-prevent-xtime-lock-contention kernel/time/tick-sched.c
--- a/kernel/time/tick-sched.c~highres-dyntick-prevent-xtime-lock-contention
+++ a/kernel/time/tick-sched.c
@@ -221,6 +221,18 @@ void tick_nohz_stop_sched_tick(void)
ts->tick_stopped = 1;
ts->idle_jiffies = last_jiffies;
}
+
+ /*
+ * If this cpu is the one which updates jiffies, then
+ * give up the assignment and let it be taken by the
+ * cpu which runs the tick timer next, which might be
+ * this cpu as well. If we don't drop this here the
+ * jiffies might be stale and do_timer() never
+ * invoked.
+ */
+ if (cpu == tick_do_timer_cpu)
+ tick_do_timer_cpu = -1;
+
/*
* calculate the expiry time for the next timer wheel
* timer
@@ -338,12 +350,24 @@ static void tick_nohz_handler(struct clo
{
struct tick_sched *ts = &__get_cpu_var(tick_cpu_sched);
struct pt_regs *regs = get_irq_regs();
+ int cpu = smp_processor_id();
ktime_t now = ktime_get();
dev->next_event.tv64 = KTIME_MAX;
+ /*
+ * Check if the do_timer duty was dropped. We don't care about
+ * concurrency: This happens only when the cpu in charge went
+ * into a long sleep. If two cpus happen to assign themself to
+ * this duty, then the jiffies update is still serialized by
+ * xtime_lock.
+ */
+ if (unlikely(tick_do_timer_cpu == -1))
+ tick_do_timer_cpu = cpu;
+
/* Check, if the jiffies need an update */
- tick_do_update_jiffies64(now);
+ if (tick_do_timer_cpu == cpu)
+ tick_do_update_jiffies64(now);
/*
* When we are idle and the tick is stopped, we have to touch
@@ -431,9 +455,23 @@ static enum hrtimer_restart tick_sched_t
struct hrtimer_cpu_base *base = timer->base->cpu_base;
struct pt_regs *regs = get_irq_regs();
ktime_t now = ktime_get();
+ int cpu = smp_processor_id();
+
+#ifdef CONFIG_NO_HZ
+ /*
+ * Check if the do_timer duty was dropped. We don't care about
+ * concurrency: This happens only when the cpu in charge went
+ * into a long sleep. If two cpus happen to assign themself to
+ * this duty, then the jiffies update is still serialized by
+ * xtime_lock.
+ */
+ if (unlikely(tick_do_timer_cpu == -1))
+ tick_do_timer_cpu = cpu;
+#endif
/* Check, if the jiffies need an update */
- tick_do_update_jiffies64(now);
+ if (tick_do_timer_cpu == cpu)
+ tick_do_update_jiffies64(now);
/*
* Do not call, when we are not in irq context and have
_
Patches currently in -mm which might be from tglx@xxxxxxxxxxxxx are
origin.patch
kernel-doc-fix-plisth-comments.patch
git-arm.patch
git-ieee1394.patch
git-input.patch
vioc-replace-more-deprecated-sa_xxx-interrupt-flags.patch
i386-tsc-remove-xtime_locking-around-cpufreq-notifier.patch
optimize-timespec_trunc.patch
kernel-irq-procc-unprotected-iteration-over-the-irq-action-list-in-name_unique.patch
add-support-for-deferrable-timers-respun.patch
add-support-for-deferrable-timers-respun-tidy.patch
add-support-for-deferrable-timers-respun-fix.patch
irq-check-for-percpu-flag-only-when-adding-first-irqaction.patch
move-timekeeping-code-to-timekeepingc.patch
move-timekeeping-code-to-timekeepingc-fix.patch
ignore-stolen-time-in-the-softlockup-watchdog.patch
replace-deprecated-sa_xxx-interrupt-flags.patch
deprecate-sa_xxx-interrupt-flags-v2.patch
pad-irq_desc-to-internode-cacheline-size.patch
pad-irq_desc-to-internode-cacheline-size-fix.patch
highres-dyntick-prevent-xtime-lock-contention.patch
sched-fix-idle-load-balancing-in-softirqd-context.patch
sched-dynticks-idle-load-balancing-v3.patch
declare-struct-ktime.patch
add-irqf_irqpoll-flag-common-code.patch
add-irqf_irqpoll-flag-on-x86_64.patch
add-irqf_irqpoll-flag-on-i386.patch
add-irqf_irqpoll-flag-on-ia64.patch
add-irqf_irqpoll-flag-on-sh.patch
add-irqf_irqpoll-flag-on-arm.patch
-
To unsubscribe from this list: send the line "unsubscribe mm-commits" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at http://vger.kernel.org/majordomo-info.html
