diff options
author | Nicholas Piggin <npiggin@gmail.com> | 2019-04-09 14:40:05 +1000 |
---|---|---|
committer | Michael Ellerman <mpe@ellerman.id.au> | 2019-04-30 11:31:02 +1000 |
commit | 7ae3f6e130e8dc6188b59e3b4ebc2f16e9c8d053 (patch) | |
tree | 92be9e3d4c9efa848e05c0873799c1e89630d569 /arch | |
parent | b2d3b5ee66f2a04a918cc043cec0c9ed3de58f40 (diff) | |
download | linux-7ae3f6e130e8dc6188b59e3b4ebc2f16e9c8d053.tar.bz2 |
powerpc/watchdog: Use hrtimers for per-CPU heartbeat
Using a jiffies timer creates a dependency on the tick_do_timer_cpu
incrementing jiffies. If that CPU has locked up and jiffies is not
incrementing, the watchdog heartbeat timer for all CPUs stops and
creates false positives and confusing warnings on local CPUs, and
also causes the SMP detector to stop, so the root cause is never
detected.
Fix this by using hrtimer based timers for the watchdog heartbeat,
like the generic kernel hardlockup detector.
Cc: Gautham R. Shenoy <ego@linux.vnet.ibm.com>
Reported-by: Ravikumar Bangoria <ravi.bangoria@in.ibm.com>
Signed-off-by: Nicholas Piggin <npiggin@gmail.com>
Tested-by: Ravi Bangoria <ravi.bangoria@linux.ibm.com>
Reported-by: Ravi Bangoria <ravi.bangoria@linux.ibm.com>
Reviewed-by: Gautham R. Shenoy <ego@linux.vnet.ibm.com>
Signed-off-by: Michael Ellerman <mpe@ellerman.id.au>
Diffstat (limited to 'arch')
-rw-r--r-- | arch/powerpc/kernel/watchdog.c | 81 |
1 files changed, 40 insertions, 41 deletions
diff --git a/arch/powerpc/kernel/watchdog.c b/arch/powerpc/kernel/watchdog.c index 3c6ab22a0c4e..af3c15a1d41e 100644 --- a/arch/powerpc/kernel/watchdog.c +++ b/arch/powerpc/kernel/watchdog.c @@ -77,7 +77,7 @@ static u64 wd_smp_panic_timeout_tb __read_mostly; /* panic other CPUs */ static u64 wd_timer_period_ms __read_mostly; /* interval between heartbeat */ -static DEFINE_PER_CPU(struct timer_list, wd_timer); +static DEFINE_PER_CPU(struct hrtimer, wd_hrtimer); static DEFINE_PER_CPU(u64, wd_timer_tb); /* SMP checker bits */ @@ -293,21 +293,21 @@ out: nmi_exit(); } -static void wd_timer_reset(unsigned int cpu, struct timer_list *t) -{ - t->expires = jiffies + msecs_to_jiffies(wd_timer_period_ms); - if (wd_timer_period_ms > 1000) - t->expires = __round_jiffies_up(t->expires, cpu); - add_timer_on(t, cpu); -} - -static void wd_timer_fn(struct timer_list *t) +static enum hrtimer_restart watchdog_timer_fn(struct hrtimer *hrtimer) { int cpu = smp_processor_id(); + if (!(watchdog_enabled & NMI_WATCHDOG_ENABLED)) + return HRTIMER_NORESTART; + + if (!cpumask_test_cpu(cpu, &watchdog_cpumask)) + return HRTIMER_NORESTART; + watchdog_timer_interrupt(cpu); - wd_timer_reset(cpu, t); + hrtimer_forward_now(hrtimer, ms_to_ktime(wd_timer_period_ms)); + + return HRTIMER_RESTART; } void arch_touch_nmi_watchdog(void) @@ -323,37 +323,22 @@ void arch_touch_nmi_watchdog(void) } EXPORT_SYMBOL(arch_touch_nmi_watchdog); -static void start_watchdog_timer_on(unsigned int cpu) -{ - struct timer_list *t = per_cpu_ptr(&wd_timer, cpu); - - per_cpu(wd_timer_tb, cpu) = get_tb(); - - timer_setup(t, wd_timer_fn, TIMER_PINNED); - wd_timer_reset(cpu, t); -} - -static void stop_watchdog_timer_on(unsigned int cpu) -{ - struct timer_list *t = per_cpu_ptr(&wd_timer, cpu); - - del_timer_sync(t); -} - -static int start_wd_on_cpu(unsigned int cpu) +static void start_watchdog(void *arg) { + struct hrtimer *hrtimer = this_cpu_ptr(&wd_hrtimer); + int cpu = smp_processor_id(); unsigned long flags; if (cpumask_test_cpu(cpu, &wd_cpus_enabled)) { WARN_ON(1); - return 0; + return; } if (!(watchdog_enabled & NMI_WATCHDOG_ENABLED)) - return 0; + return; if (!cpumask_test_cpu(cpu, &watchdog_cpumask)) - return 0; + return; wd_smp_lock(&flags); cpumask_set_cpu(cpu, &wd_cpus_enabled); @@ -363,27 +348,40 @@ static int start_wd_on_cpu(unsigned int cpu) } wd_smp_unlock(&flags); - start_watchdog_timer_on(cpu); + *this_cpu_ptr(&wd_timer_tb) = get_tb(); - return 0; + hrtimer_init(hrtimer, CLOCK_MONOTONIC, HRTIMER_MODE_REL); + hrtimer->function = watchdog_timer_fn; + hrtimer_start(hrtimer, ms_to_ktime(wd_timer_period_ms), + HRTIMER_MODE_REL_PINNED); } -static int stop_wd_on_cpu(unsigned int cpu) +static int start_watchdog_on_cpu(unsigned int cpu) { + return smp_call_function_single(cpu, start_watchdog, NULL, true); +} + +static void stop_watchdog(void *arg) +{ + struct hrtimer *hrtimer = this_cpu_ptr(&wd_hrtimer); + int cpu = smp_processor_id(); unsigned long flags; if (!cpumask_test_cpu(cpu, &wd_cpus_enabled)) - return 0; /* Can happen in CPU unplug case */ + return; /* Can happen in CPU unplug case */ - stop_watchdog_timer_on(cpu); + hrtimer_cancel(hrtimer); wd_smp_lock(&flags); cpumask_clear_cpu(cpu, &wd_cpus_enabled); wd_smp_unlock(&flags); wd_smp_clear_cpu_pending(cpu, get_tb()); +} - return 0; +static int stop_watchdog_on_cpu(unsigned int cpu) +{ + return smp_call_function_single(cpu, stop_watchdog, NULL, true); } static void watchdog_calc_timeouts(void) @@ -402,7 +400,7 @@ void watchdog_nmi_stop(void) int cpu; for_each_cpu(cpu, &wd_cpus_enabled) - stop_wd_on_cpu(cpu); + stop_watchdog_on_cpu(cpu); } void watchdog_nmi_start(void) @@ -411,7 +409,7 @@ void watchdog_nmi_start(void) watchdog_calc_timeouts(); for_each_cpu_and(cpu, cpu_online_mask, &watchdog_cpumask) - start_wd_on_cpu(cpu); + start_watchdog_on_cpu(cpu); } /* @@ -423,7 +421,8 @@ int __init watchdog_nmi_probe(void) err = cpuhp_setup_state_nocalls(CPUHP_AP_ONLINE_DYN, "powerpc/watchdog:online", - start_wd_on_cpu, stop_wd_on_cpu); + start_watchdog_on_cpu, + stop_watchdog_on_cpu); if (err < 0) { pr_warn("could not be initialized"); return err; |