From 52d168e28bc11dd026b620fe1767cadde5a747cd Mon Sep 17 00:00:00 2001 From: Andi Kleen Date: Thu, 12 Feb 2009 13:39:29 +0100 Subject: [PATCH] x86, mce: switch machine check polling to per CPU timer Impact: Higher priority bug fix The machine check poller runs a single timer and then broadcasted an IPI to all CPUs to check them. This leads to unnecessary synchronization between CPUs. The original CPU running the timer has to wait potentially a long time for all other CPUs answering. This is also real time unfriendly and in general inefficient. This was especially a problem on systems with a lot of events where the poller run with a higher frequency after processing some events. There could be more and more CPU time wasted with this, to the point of significantly slowing down machines. The machine check polling is actually fully independent per CPU, so there's no reason to not just do this all with per CPU timers. This patch implements that. Also switch the poller also to use standard timers instead of work queues. It was using work queues to be able to execute a user program on a event, but mce_notify_user() handles this case now with a separate callback. So instead always run the poll code in in a standard per CPU timer, which means that in the common case of not having to execute a trigger there will be less overhead. This allows to clean up the initialization significantly, because standard timers are already up when machine checks get init'ed. No multiple initialization functions. Thanks to Thomas Gleixner for some help. Cc: thockin@google.com v2: Use del_timer_sync() on cpu shutdown and don't try to handle migrated timers. v3: Add WARN_ON for timer running on unexpected CPU Signed-off-by: Andi Kleen Acked-by: Thomas Gleixner Signed-off-by: H. Peter Anvin --- arch/x86/kernel/cpu/mcheck/mce_64.c | 68 +++++++++++++++++++---------- 1 file changed, 45 insertions(+), 23 deletions(-) diff --git a/arch/x86/kernel/cpu/mcheck/mce_64.c b/arch/x86/kernel/cpu/mcheck/mce_64.c index 18b379cf0610..3f0550d16f3c 100644 --- a/arch/x86/kernel/cpu/mcheck/mce_64.c +++ b/arch/x86/kernel/cpu/mcheck/mce_64.c @@ -353,18 +353,17 @@ void mce_log_therm_throt_event(unsigned int cpu, __u64 status) static int check_interval = 5 * 60; /* 5 minutes */ static int next_interval; /* in jiffies */ -static void mcheck_timer(struct work_struct *work); -static DECLARE_DELAYED_WORK(mcheck_work, mcheck_timer); +static void mcheck_timer(unsigned long); +static DEFINE_PER_CPU(struct timer_list, mce_timer); -static void mcheck_check_cpu(void *info) +static void mcheck_timer(unsigned long data) { + struct timer_list *t = &per_cpu(mce_timer, data); + + WARN_ON(smp_processor_id() != data); + if (mce_available(¤t_cpu_data)) do_machine_check(NULL, 0); -} - -static void mcheck_timer(struct work_struct *work) -{ - on_each_cpu(mcheck_check_cpu, NULL, 1); /* * Alert userspace if needed. If we logged an MCE, reduce the @@ -377,7 +376,8 @@ static void mcheck_timer(struct work_struct *work) (int)round_jiffies_relative(check_interval*HZ)); } - schedule_delayed_work(&mcheck_work, next_interval); + t->expires = jiffies + next_interval; + add_timer(t); } static void mce_do_trigger(struct work_struct *work) @@ -436,16 +436,11 @@ static struct notifier_block mce_idle_notifier = { static __init int periodic_mcheck_init(void) { - next_interval = check_interval * HZ; - if (next_interval) - schedule_delayed_work(&mcheck_work, - round_jiffies_relative(next_interval)); - idle_notifier_register(&mce_idle_notifier); - return 0; + idle_notifier_register(&mce_idle_notifier); + return 0; } __initcall(periodic_mcheck_init); - /* * Initialize Machine Checks for a CPU. */ @@ -515,6 +510,20 @@ static void __cpuinit mce_cpu_features(struct cpuinfo_x86 *c) } } +static void mce_init_timer(void) +{ + struct timer_list *t = &__get_cpu_var(mce_timer); + + /* data race harmless because everyone sets to the same value */ + if (!next_interval) + next_interval = check_interval * HZ; + if (!next_interval) + return; + setup_timer(t, mcheck_timer, smp_processor_id()); + t->expires = round_jiffies_relative(jiffies + next_interval); + add_timer(t); +} + /* * Called for each booted CPU to set up machine checks. * Must be called with preempt off. @@ -529,6 +538,7 @@ void __cpuinit mcheck_init(struct cpuinfo_x86 *c) mce_init(NULL); mce_cpu_features(c); + mce_init_timer(); } /* @@ -758,17 +768,19 @@ static int mce_resume(struct sys_device *dev) return 0; } +static void mce_cpu_restart(void *data) +{ + del_timer_sync(&__get_cpu_var(mce_timer)); + if (mce_available(¤t_cpu_data)) + mce_init(NULL); + mce_init_timer(); +} + /* Reinit MCEs after user configuration changes */ static void mce_restart(void) { - if (next_interval) - cancel_delayed_work(&mcheck_work); - /* Timer race is harmless here */ - on_each_cpu(mce_init, NULL, 1); next_interval = check_interval * HZ; - if (next_interval) - schedule_delayed_work(&mcheck_work, - round_jiffies_relative(next_interval)); + on_each_cpu(mce_cpu_restart, NULL, 1); } static struct sysdev_class mce_sysclass = { @@ -899,6 +911,7 @@ static int __cpuinit mce_cpu_callback(struct notifier_block *nfb, unsigned long action, void *hcpu) { unsigned int cpu = (unsigned long)hcpu; + struct timer_list *t = &per_cpu(mce_timer, cpu); switch (action) { case CPU_ONLINE: @@ -913,6 +926,15 @@ static int __cpuinit mce_cpu_callback(struct notifier_block *nfb, threshold_cpu_callback(action, cpu); mce_remove_device(cpu); break; + case CPU_DOWN_PREPARE: + case CPU_DOWN_PREPARE_FROZEN: + del_timer_sync(t); + break; + case CPU_DOWN_FAILED: + case CPU_DOWN_FAILED_FROZEN: + t->expires = round_jiffies_relative(jiffies + next_interval); + add_timer_on(t, cpu); + break; } return NOTIFY_OK; } -- 2.39.5