gpstate_timer_handler() uses synchronous smp_call to set the pstate
on the requested core. This causes the below hard lockup:
[c000003fe566b320] [c0000000001d5340] smp_call_function_single+0x110/0x180 (unreliable)
[c000003fe566b390] [c0000000001d55e0] smp_call_function_any+0x180/0x250
[c000003fe566b3f0] [c000000000acd3e8] gpstate_timer_handler+0x1e8/0x580
[c000003fe566b4a0] [c0000000001b46b0] call_timer_fn+0x50/0x1c0
[c000003fe566b520] [c0000000001b4958] expire_timers+0x138/0x1f0
[c000003fe566b590] [c0000000001b4bf8] run_timer_softirq+0x1e8/0x270
[c000003fe566b630] [c000000000d0d6c8] __do_softirq+0x158/0x3e4
[c000003fe566b710] [c000000000114be8] irq_exit+0xe8/0x120
[c000003fe566b730] [c000000000024d0c] timer_interrupt+0x9c/0xe0
[c000003fe566b760] [c000000000009014] decrementer_common+0x114/0x120
-- interrupt: 901 at doorbell_global_ipi+0x34/0x50
LR = arch_send_call_function_ipi_mask+0x120/0x130
[c000003fe566ba50] [c00000000004876c]
arch_send_call_function_ipi_mask+0x4c/0x130
[c000003fe566ba90] [c0000000001d59f0] smp_call_function_many+0x340/0x450
[c000003fe566bb00] [c000000000075f18] pmdp_invalidate+0x98/0xe0
[c000003fe566bb30] [c0000000003a1120] change_huge_pmd+0xe0/0x270
[c000003fe566bba0] [c000000000349278] change_protection_range+0xb88/0xe40
[c000003fe566bcf0] [c0000000003496c0] mprotect_fixup+0x140/0x340
[c000003fe566bdb0] [c000000000349a74] SyS_mprotect+0x1b4/0x350
[c000003fe566be30] [c00000000000b184] system_call+0x58/0x6c
One way to avoid this is removing the smp-call. We can ensure that the timer
always runs on one of the policy-cpus. If the timer gets migrated to a
cpu outside the policy then re-queue it back on the policy->cpus. This way
we can get rid of the smp-call which was being used to set the pstate
on the policy->cpus.
Fixes: 7bc54b652f13 (timers, cpufreq/powernv: Initialize the gpstate timer as pinned)
Cc: <[email protected]> [4.8+]
Reported-by: Nicholas Piggin <[email protected]>
Reported-by: Pridhiviraj Paidipeddi <[email protected]>
Signed-off-by: Shilpasri G Bhat <[email protected]>
---
Changes from V2:
- Remove the check for active policy while requeing the migrated timer
Changes from V1:
- Remove smp_call in the pstate handler.
drivers/cpufreq/powernv-cpufreq.c | 14 +++++++++++---
1 file changed, 11 insertions(+), 3 deletions(-)
diff --git a/drivers/cpufreq/powernv-cpufreq.c b/drivers/cpufreq/powernv-cpufreq.c
index 71f8682..e368e1f 100644
--- a/drivers/cpufreq/powernv-cpufreq.c
+++ b/drivers/cpufreq/powernv-cpufreq.c
@@ -679,6 +679,16 @@ void gpstate_timer_handler(struct timer_list *t)
if (!spin_trylock(&gpstates->gpstate_lock))
return;
+ /*
+ * If the timer has migrated to the different cpu then bring
+ * it back to one of the policy->cpus
+ */
+ if (!cpumask_test_cpu(raw_smp_processor_id(), policy->cpus)) {
+ gpstates->timer.expires = jiffies + msecs_to_jiffies(1);
+ add_timer_on(&gpstates->timer, cpumask_first(policy->cpus));
+ spin_unlock(&gpstates->gpstate_lock);
+ return;
+ }
/*
* If PMCR was last updated was using fast_swtich then
@@ -718,10 +728,8 @@ void gpstate_timer_handler(struct timer_list *t)
if (gpstate_idx != gpstates->last_lpstate_idx)
queue_gpstate_timer(gpstates);
+ set_pstate(&freq_data);
spin_unlock(&gpstates->gpstate_lock);
-
- /* Timer may get migrated to a different cpu on cpu hot unplug */
- smp_call_function_any(policy->cpus, set_pstate, &freq_data, 1);
}
/*
--
1.8.3.1
On Wed, 25 Apr 2018 16:29:31 +0530
Shilpasri G Bhat <[email protected]> wrote:
> gpstate_timer_handler() uses synchronous smp_call to set the pstate
> on the requested core. This causes the below hard lockup:
>
> [c000003fe566b320] [c0000000001d5340] smp_call_function_single+0x110/0x180 (unreliable)
> [c000003fe566b390] [c0000000001d55e0] smp_call_function_any+0x180/0x250
> [c000003fe566b3f0] [c000000000acd3e8] gpstate_timer_handler+0x1e8/0x580
> [c000003fe566b4a0] [c0000000001b46b0] call_timer_fn+0x50/0x1c0
> [c000003fe566b520] [c0000000001b4958] expire_timers+0x138/0x1f0
> [c000003fe566b590] [c0000000001b4bf8] run_timer_softirq+0x1e8/0x270
> [c000003fe566b630] [c000000000d0d6c8] __do_softirq+0x158/0x3e4
> [c000003fe566b710] [c000000000114be8] irq_exit+0xe8/0x120
> [c000003fe566b730] [c000000000024d0c] timer_interrupt+0x9c/0xe0
> [c000003fe566b760] [c000000000009014] decrementer_common+0x114/0x120
> -- interrupt: 901 at doorbell_global_ipi+0x34/0x50
> LR = arch_send_call_function_ipi_mask+0x120/0x130
> [c000003fe566ba50] [c00000000004876c]
> arch_send_call_function_ipi_mask+0x4c/0x130
> [c000003fe566ba90] [c0000000001d59f0] smp_call_function_many+0x340/0x450
> [c000003fe566bb00] [c000000000075f18] pmdp_invalidate+0x98/0xe0
> [c000003fe566bb30] [c0000000003a1120] change_huge_pmd+0xe0/0x270
> [c000003fe566bba0] [c000000000349278] change_protection_range+0xb88/0xe40
> [c000003fe566bcf0] [c0000000003496c0] mprotect_fixup+0x140/0x340
> [c000003fe566bdb0] [c000000000349a74] SyS_mprotect+0x1b4/0x350
> [c000003fe566be30] [c00000000000b184] system_call+0x58/0x6c
>
> One way to avoid this is removing the smp-call. We can ensure that the timer
> always runs on one of the policy-cpus. If the timer gets migrated to a
> cpu outside the policy then re-queue it back on the policy->cpus. This way
> we can get rid of the smp-call which was being used to set the pstate
> on the policy->cpus.
>
> Fixes: 7bc54b652f13 (timers, cpufreq/powernv: Initialize the gpstate timer as pinned)
> Cc: <[email protected]> [4.8+]
> Reported-by: Nicholas Piggin <[email protected]>
> Reported-by: Pridhiviraj Paidipeddi <[email protected]>
> Signed-off-by: Shilpasri G Bhat <[email protected]>
Thanks, this looks good to me. I don't know the code though, so
Acked-by: Nicholas Piggin <[email protected]>
> ---
> Changes from V2:
> - Remove the check for active policy while requeing the migrated timer
> Changes from V1:
> - Remove smp_call in the pstate handler.
>
> drivers/cpufreq/powernv-cpufreq.c | 14 +++++++++++---
> 1 file changed, 11 insertions(+), 3 deletions(-)
>
> diff --git a/drivers/cpufreq/powernv-cpufreq.c b/drivers/cpufreq/powernv-cpufreq.c
> index 71f8682..e368e1f 100644
> --- a/drivers/cpufreq/powernv-cpufreq.c
> +++ b/drivers/cpufreq/powernv-cpufreq.c
> @@ -679,6 +679,16 @@ void gpstate_timer_handler(struct timer_list *t)
>
> if (!spin_trylock(&gpstates->gpstate_lock))
> return;
I still think it would be good to do something about the trylock failure.
It may be rare, but if it happens it could stop the timer and lead to
some rare unpredictable behaviour? Not for this patch, but while you're
looking at the code it would be good to consider it. Just queueing up
another timer seems like it should be enough.
> + /*
> + * If the timer has migrated to the different cpu then bring
> + * it back to one of the policy->cpus
> + */
> + if (!cpumask_test_cpu(raw_smp_processor_id(), policy->cpus)) {
> + gpstates->timer.expires = jiffies + msecs_to_jiffies(1);
> + add_timer_on(&gpstates->timer, cpumask_first(policy->cpus));
> + spin_unlock(&gpstates->gpstate_lock);
> + return;
> + }
Really small nitpick, but you could use cpumask_any there.
Thanks,
Nick
>
> /*
> * If PMCR was last updated was using fast_swtich then
> @@ -718,10 +728,8 @@ void gpstate_timer_handler(struct timer_list *t)
> if (gpstate_idx != gpstates->last_lpstate_idx)
> queue_gpstate_timer(gpstates);
>
> + set_pstate(&freq_data);
> spin_unlock(&gpstates->gpstate_lock);
> -
> - /* Timer may get migrated to a different cpu on cpu hot unplug */
> - smp_call_function_any(policy->cpus, set_pstate, &freq_data, 1);
> }
On 25-04-18, 16:29, Shilpasri G Bhat wrote:
> gpstate_timer_handler() uses synchronous smp_call to set the pstate
> on the requested core. This causes the below hard lockup:
>
> [c000003fe566b320] [c0000000001d5340] smp_call_function_single+0x110/0x180 (unreliable)
> [c000003fe566b390] [c0000000001d55e0] smp_call_function_any+0x180/0x250
> [c000003fe566b3f0] [c000000000acd3e8] gpstate_timer_handler+0x1e8/0x580
> [c000003fe566b4a0] [c0000000001b46b0] call_timer_fn+0x50/0x1c0
> [c000003fe566b520] [c0000000001b4958] expire_timers+0x138/0x1f0
> [c000003fe566b590] [c0000000001b4bf8] run_timer_softirq+0x1e8/0x270
> [c000003fe566b630] [c000000000d0d6c8] __do_softirq+0x158/0x3e4
> [c000003fe566b710] [c000000000114be8] irq_exit+0xe8/0x120
> [c000003fe566b730] [c000000000024d0c] timer_interrupt+0x9c/0xe0
> [c000003fe566b760] [c000000000009014] decrementer_common+0x114/0x120
> -- interrupt: 901 at doorbell_global_ipi+0x34/0x50
> LR = arch_send_call_function_ipi_mask+0x120/0x130
> [c000003fe566ba50] [c00000000004876c]
> arch_send_call_function_ipi_mask+0x4c/0x130
> [c000003fe566ba90] [c0000000001d59f0] smp_call_function_many+0x340/0x450
> [c000003fe566bb00] [c000000000075f18] pmdp_invalidate+0x98/0xe0
> [c000003fe566bb30] [c0000000003a1120] change_huge_pmd+0xe0/0x270
> [c000003fe566bba0] [c000000000349278] change_protection_range+0xb88/0xe40
> [c000003fe566bcf0] [c0000000003496c0] mprotect_fixup+0x140/0x340
> [c000003fe566bdb0] [c000000000349a74] SyS_mprotect+0x1b4/0x350
> [c000003fe566be30] [c00000000000b184] system_call+0x58/0x6c
>
> One way to avoid this is removing the smp-call. We can ensure that the timer
> always runs on one of the policy-cpus. If the timer gets migrated to a
> cpu outside the policy then re-queue it back on the policy->cpus. This way
> we can get rid of the smp-call which was being used to set the pstate
> on the policy->cpus.
>
> Fixes: 7bc54b652f13 (timers, cpufreq/powernv: Initialize the gpstate timer as pinned)
> Cc: <[email protected]> [4.8+]
> Reported-by: Nicholas Piggin <[email protected]>
> Reported-by: Pridhiviraj Paidipeddi <[email protected]>
> Signed-off-by: Shilpasri G Bhat <[email protected]>
> ---
> Changes from V2:
> - Remove the check for active policy while requeing the migrated timer
> Changes from V1:
> - Remove smp_call in the pstate handler.
Acked-by: Viresh Kumar <[email protected]>
--
viresh
* Shilpa Bhat <[email protected]> [2018-04-25 16:29:31]:
> gpstate_timer_handler() uses synchronous smp_call to set the pstate
> on the requested core. This causes the below hard lockup:
>
> [c000003fe566b320] [c0000000001d5340] smp_call_function_single+0x110/0x180 (unreliable)
> [c000003fe566b390] [c0000000001d55e0] smp_call_function_any+0x180/0x250
> [c000003fe566b3f0] [c000000000acd3e8] gpstate_timer_handler+0x1e8/0x580
> [c000003fe566b4a0] [c0000000001b46b0] call_timer_fn+0x50/0x1c0
> [c000003fe566b520] [c0000000001b4958] expire_timers+0x138/0x1f0
> [c000003fe566b590] [c0000000001b4bf8] run_timer_softirq+0x1e8/0x270
> [c000003fe566b630] [c000000000d0d6c8] __do_softirq+0x158/0x3e4
> [c000003fe566b710] [c000000000114be8] irq_exit+0xe8/0x120
> [c000003fe566b730] [c000000000024d0c] timer_interrupt+0x9c/0xe0
> [c000003fe566b760] [c000000000009014] decrementer_common+0x114/0x120
> -- interrupt: 901 at doorbell_global_ipi+0x34/0x50
> LR = arch_send_call_function_ipi_mask+0x120/0x130
> [c000003fe566ba50] [c00000000004876c]
> arch_send_call_function_ipi_mask+0x4c/0x130
> [c000003fe566ba90] [c0000000001d59f0] smp_call_function_many+0x340/0x450
> [c000003fe566bb00] [c000000000075f18] pmdp_invalidate+0x98/0xe0
> [c000003fe566bb30] [c0000000003a1120] change_huge_pmd+0xe0/0x270
> [c000003fe566bba0] [c000000000349278] change_protection_range+0xb88/0xe40
> [c000003fe566bcf0] [c0000000003496c0] mprotect_fixup+0x140/0x340
> [c000003fe566bdb0] [c000000000349a74] SyS_mprotect+0x1b4/0x350
> [c000003fe566be30] [c00000000000b184] system_call+0x58/0x6c
>
> One way to avoid this is removing the smp-call. We can ensure that the timer
> always runs on one of the policy-cpus. If the timer gets migrated to a
> cpu outside the policy then re-queue it back on the policy->cpus. This way
> we can get rid of the smp-call which was being used to set the pstate
> on the policy->cpus.
>
> Fixes: 7bc54b652f13 (timers, cpufreq/powernv: Initialize the gpstate timer as pinned)
> Cc: <[email protected]> [4.8+]
> Reported-by: Nicholas Piggin <[email protected]>
> Reported-by: Pridhiviraj Paidipeddi <[email protected]>
> Signed-off-by: Shilpasri G Bhat <[email protected]>
> ---
> Changes from V2:
> - Remove the check for active policy while requeing the migrated timer
> Changes from V1:
> - Remove smp_call in the pstate handler.
>
> drivers/cpufreq/powernv-cpufreq.c | 14 +++++++++++---
> 1 file changed, 11 insertions(+), 3 deletions(-)
>
> diff --git a/drivers/cpufreq/powernv-cpufreq.c b/drivers/cpufreq/powernv-cpufreq.c
> index 71f8682..e368e1f 100644
> --- a/drivers/cpufreq/powernv-cpufreq.c
> +++ b/drivers/cpufreq/powernv-cpufreq.c
> @@ -679,6 +679,16 @@ void gpstate_timer_handler(struct timer_list *t)
>
> if (!spin_trylock(&gpstates->gpstate_lock))
> return;
> + /*
> + * If the timer has migrated to the different cpu then bring
> + * it back to one of the policy->cpus
> + */
> + if (!cpumask_test_cpu(raw_smp_processor_id(), policy->cpus)) {
> + gpstates->timer.expires = jiffies + msecs_to_jiffies(1);
> + add_timer_on(&gpstates->timer, cpumask_first(policy->cpus));
> + spin_unlock(&gpstates->gpstate_lock);
> + return;
> + }
>
> /*
> * If PMCR was last updated was using fast_swtich then
> @@ -718,10 +728,8 @@ void gpstate_timer_handler(struct timer_list *t)
> if (gpstate_idx != gpstates->last_lpstate_idx)
> queue_gpstate_timer(gpstates);
>
> + set_pstate(&freq_data);
> spin_unlock(&gpstates->gpstate_lock);
> -
> - /* Timer may get migrated to a different cpu on cpu hot unplug */
> - smp_call_function_any(policy->cpus, set_pstate, &freq_data, 1);
> }
Fix looks good.
Acked-by: Vaidyanathan Srinivasan <[email protected]>
On Wed, 2018-04-25 at 10:59:31 UTC, Shilpasri G Bhat wrote:
> gpstate_timer_handler() uses synchronous smp_call to set the pstate
> on the requested core. This causes the below hard lockup:
>
> [c000003fe566b320] [c0000000001d5340] smp_call_function_single+0x110/0x180 (unreliable)
> [c000003fe566b390] [c0000000001d55e0] smp_call_function_any+0x180/0x250
> [c000003fe566b3f0] [c000000000acd3e8] gpstate_timer_handler+0x1e8/0x580
> [c000003fe566b4a0] [c0000000001b46b0] call_timer_fn+0x50/0x1c0
> [c000003fe566b520] [c0000000001b4958] expire_timers+0x138/0x1f0
> [c000003fe566b590] [c0000000001b4bf8] run_timer_softirq+0x1e8/0x270
> [c000003fe566b630] [c000000000d0d6c8] __do_softirq+0x158/0x3e4
> [c000003fe566b710] [c000000000114be8] irq_exit+0xe8/0x120
> [c000003fe566b730] [c000000000024d0c] timer_interrupt+0x9c/0xe0
> [c000003fe566b760] [c000000000009014] decrementer_common+0x114/0x120
> -- interrupt: 901 at doorbell_global_ipi+0x34/0x50
> LR = arch_send_call_function_ipi_mask+0x120/0x130
> [c000003fe566ba50] [c00000000004876c]
> arch_send_call_function_ipi_mask+0x4c/0x130
> [c000003fe566ba90] [c0000000001d59f0] smp_call_function_many+0x340/0x450
> [c000003fe566bb00] [c000000000075f18] pmdp_invalidate+0x98/0xe0
> [c000003fe566bb30] [c0000000003a1120] change_huge_pmd+0xe0/0x270
> [c000003fe566bba0] [c000000000349278] change_protection_range+0xb88/0xe40
> [c000003fe566bcf0] [c0000000003496c0] mprotect_fixup+0x140/0x340
> [c000003fe566bdb0] [c000000000349a74] SyS_mprotect+0x1b4/0x350
> [c000003fe566be30] [c00000000000b184] system_call+0x58/0x6c
>
> One way to avoid this is removing the smp-call. We can ensure that the timer
> always runs on one of the policy-cpus. If the timer gets migrated to a
> cpu outside the policy then re-queue it back on the policy->cpus. This way
> we can get rid of the smp-call which was being used to set the pstate
> on the policy->cpus.
>
> Fixes: 7bc54b652f13 (timers, cpufreq/powernv: Initialize the gpstate timer as pinned)
> Cc: <[email protected]> [4.8+]
> Reported-by: Nicholas Piggin <[email protected]>
> Reported-by: Pridhiviraj Paidipeddi <[email protected]>
> Signed-off-by: Shilpasri G Bhat <[email protected]>
> Acked-by: Nicholas Piggin <[email protected]>
> Acked-by: Viresh Kumar <[email protected]>
> Acked-by: Vaidyanathan Srinivasan <[email protected]>
Applied to powerpc fixes, thanks.
https://git.kernel.org/powerpc/c/c0f7f5b6c69107ca92909512533e70
cheers