[PATCH] drm/i915/pmu: Check if pmu is closed before stopping event

[PATCH] drm/i915/pmu: Check if pmu is closed before stopping event

[Umesh Nerlige Ramappa]

When the driver unbinds, pmu is unregistered and i915->uabi_engines is
set to RB_ROOT. Due to this, when i915 PMU tries to stop the engine
events, it issues a warn_on because engine lookup fails.

All perf hooks are taking care of this using a pmu->closed flag that is
set when PMU unregisters. The stop event seems to have been left out.

Check for pmu->closed in pmu_event_stop as well.

Based on discussion here -
https://patchwork.freedesktop.org/patch/492079/?series=105790&rev=2

v2: s/is/if/ in commit title
v3: Add fixes tag and cc stable

Cc: <stable@vger.kernel.org> # v5.11+
Fixes: b00bccb3f0bb ("drm/i915/pmu: Handle PCI unbind")
Signed-off-by: Umesh Nerlige Ramappa <umesh.nerlige.ramappa@intel.com>
Reviewed-by: Tvrtko Ursulin <tvrtko.ursulin@intel.com>
---
 drivers/gpu/drm/i915/i915_pmu.c | 9 +++++++++
 1 file changed, 9 insertions(+)

diff --git a/drivers/gpu/drm/i915/i915_pmu.c b/drivers/gpu/drm/i915/i915_pmu.c
index 108b675088ba..f861863eb7c1 100644
--- a/drivers/gpu/drm/i915/i915_pmu.c
+++ b/drivers/gpu/drm/i915/i915_pmu.c
@@ -831,9 +831,18 @@ static void i915_pmu_event_start(struct perf_event *event, int flags)
 
 static void i915_pmu_event_stop(struct perf_event *event, int flags)
 {
+	struct drm_i915_private *i915 =
+		container_of(event->pmu, typeof(*i915), pmu.base);
+	struct i915_pmu *pmu = &i915->pmu;
+
+	if (pmu->closed)
+		goto out;
+
 	if (flags & PERF_EF_UPDATE)
 		i915_pmu_event_read(event);
 	i915_pmu_disable(event);
+
+out:
 	event->hw.state = PERF_HES_STOPPED;
 }
 
-- 
2.38.1

Re: [Intel-gfx] [PATCH] drm/i915/pmu: Check if pmu is closed before stopping event

[Andi Shyti]

Hi Umesh,

On Fri, Oct 20, 2023 at 08:24:41AM -0700, Umesh Nerlige Ramappa wrote:
> When the driver unbinds, pmu is unregistered and i915->uabi_engines is
> set to RB_ROOT. Due to this, when i915 PMU tries to stop the engine
> events, it issues a warn_on because engine lookup fails.
> 
> All perf hooks are taking care of this using a pmu->closed flag that is
> set when PMU unregisters. The stop event seems to have been left out.
> 
> Check for pmu->closed in pmu_event_stop as well.
> 
> Based on discussion here -
> https://patchwork.freedesktop.org/patch/492079/?series=105790&rev=2
> 
> v2: s/is/if/ in commit title
> v3: Add fixes tag and cc stable
> 
> Cc: <stable@vger.kernel.org> # v5.11+
> Fixes: b00bccb3f0bb ("drm/i915/pmu: Handle PCI unbind")
> Signed-off-by: Umesh Nerlige Ramappa <umesh.nerlige.ramappa@intel.com>
> Reviewed-by: Tvrtko Ursulin <tvrtko.ursulin@intel.com>

argh! 4th time that this patch has been sent. Please next time
use:

   git format-patch -v <version number>

Reviewed-by: Andi Shyti <andi.shyti@linux.intel.com>

No need to resend :-)

Andi

> ---
>  drivers/gpu/drm/i915/i915_pmu.c | 9 +++++++++
>  1 file changed, 9 insertions(+)
> 
> diff --git a/drivers/gpu/drm/i915/i915_pmu.c b/drivers/gpu/drm/i915/i915_pmu.c
> index 108b675088ba..f861863eb7c1 100644
> --- a/drivers/gpu/drm/i915/i915_pmu.c
> +++ b/drivers/gpu/drm/i915/i915_pmu.c
> @@ -831,9 +831,18 @@ static void i915_pmu_event_start(struct perf_event *event, int flags)
>  
>  static void i915_pmu_event_stop(struct perf_event *event, int flags)
>  {
> +	struct drm_i915_private *i915 =
> +		container_of(event->pmu, typeof(*i915), pmu.base);
> +	struct i915_pmu *pmu = &i915->pmu;
> +
> +	if (pmu->closed)
> +		goto out;
> +
>  	if (flags & PERF_EF_UPDATE)
>  		i915_pmu_event_read(event);
>  	i915_pmu_disable(event);
> +
> +out:
>  	event->hw.state = PERF_HES_STOPPED;
>  }
>  
> -- 
> 2.38.1

Re: [Intel-gfx] [PATCH] drm/i915/pmu: Check if pmu is closed before stopping event

[Andi Shyti]

Hi Umesh,

On Fri, Oct 20, 2023 at 08:24:41AM -0700, Umesh Nerlige Ramappa wrote:
> When the driver unbinds, pmu is unregistered and i915->uabi_engines is
> set to RB_ROOT. Due to this, when i915 PMU tries to stop the engine
> events, it issues a warn_on because engine lookup fails.
> 
> All perf hooks are taking care of this using a pmu->closed flag that is
> set when PMU unregisters. The stop event seems to have been left out.
> 
> Check for pmu->closed in pmu_event_stop as well.
> 
> Based on discussion here -
> https://patchwork.freedesktop.org/patch/492079/?series=105790&rev=2
> 
> v2: s/is/if/ in commit title
> v3: Add fixes tag and cc stable
> 
> Cc: <stable@vger.kernel.org> # v5.11+
> Fixes: b00bccb3f0bb ("drm/i915/pmu: Handle PCI unbind")
> Signed-off-by: Umesh Nerlige Ramappa <umesh.nerlige.ramappa@intel.com>
> Reviewed-by: Tvrtko Ursulin <tvrtko.ursulin@intel.com>

the failure from the shards tests looks unrelated.

Please next time don't forget to add a versioning to the patches
you are sending.

Pushed in drm-intel-gt-next.

Thank you,
Andi

Re: [Intel-gfx] [PATCH] drm/i915/pmu: Check if pmu is closed before stopping event

[Umesh Nerlige Ramappa]

On Tue, Oct 24, 2023 at 02:20:33PM +0200, Andi Shyti wrote:
>Hi Umesh,
>
>On Fri, Oct 20, 2023 at 08:24:41AM -0700, Umesh Nerlige Ramappa wrote:
>> When the driver unbinds, pmu is unregistered and i915->uabi_engines is
>> set to RB_ROOT. Due to this, when i915 PMU tries to stop the engine
>> events, it issues a warn_on because engine lookup fails.
>>
>> All perf hooks are taking care of this using a pmu->closed flag that is
>> set when PMU unregisters. The stop event seems to have been left out.
>>
>> Check for pmu->closed in pmu_event_stop as well.
>>
>> Based on discussion here -
>> https://patchwork.freedesktop.org/patch/492079/?series=105790&rev=2
>>
>> v2: s/is/if/ in commit title
>> v3: Add fixes tag and cc stable
>>
>> Cc: <stable@vger.kernel.org> # v5.11+
>> Fixes: b00bccb3f0bb ("drm/i915/pmu: Handle PCI unbind")
>> Signed-off-by: Umesh Nerlige Ramappa <umesh.nerlige.ramappa@intel.com>
>> Reviewed-by: Tvrtko Ursulin <tvrtko.ursulin@intel.com>
>
>argh! 4th time that this patch has been sent. Please next time
>use:
>
>   git format-patch -v <version number>
>
>Reviewed-by: Andi Shyti <andi.shyti@linux.intel.com>
>
>No need to resend :-)

Sorry about the versioning. Will keep inn mind next time.

Thanks,
Umesh
>
>Andi
>
>> ---
>>  drivers/gpu/drm/i915/i915_pmu.c | 9 +++++++++
>>  1 file changed, 9 insertions(+)
>>
>> diff --git a/drivers/gpu/drm/i915/i915_pmu.c b/drivers/gpu/drm/i915/i915_pmu.c
>> index 108b675088ba..f861863eb7c1 100644
>> --- a/drivers/gpu/drm/i915/i915_pmu.c
>> +++ b/drivers/gpu/drm/i915/i915_pmu.c
>> @@ -831,9 +831,18 @@ static void i915_pmu_event_start(struct perf_event *event, int flags)
>>
>>  static void i915_pmu_event_stop(struct perf_event *event, int flags)
>>  {
>> +	struct drm_i915_private *i915 =
>> +		container_of(event->pmu, typeof(*i915), pmu.base);
>> +	struct i915_pmu *pmu = &i915->pmu;
>> +
>> +	if (pmu->closed)
>> +		goto out;
>> +
>>  	if (flags & PERF_EF_UPDATE)
>>  		i915_pmu_event_read(event);
>>  	i915_pmu_disable(event);
>> +
>> +out:
>>  	event->hw.state = PERF_HES_STOPPED;
>>  }
>>
>> --
>> 2.38.1