[PATCH] drm/amdgpu/vcn: drop extra cancel_delayed_work_sync()

[PATCH] drm/amdgpu/vcn: drop extra cancel_delayed_work_sync()

[Alex Deucher]

We already call this in the hw_fini() methods for all
VCN instances, so no need to call it again in
amdgpu_vcn_suspend().

Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
---
 drivers/gpu/drm/amd/amdgpu/amdgpu_vcn.c | 2 --
 1 file changed, 2 deletions(-)

diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_vcn.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_vcn.c
index 9a76e11d1c184..fd8ebf4b5a824 100644
--- a/drivers/gpu/drm/amd/amdgpu/amdgpu_vcn.c
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_vcn.c
@@ -357,8 +357,6 @@ int amdgpu_vcn_suspend(struct amdgpu_device *adev, int i)
 	if (adev->vcn.harvest_config & (1 << i))
 		return 0;
 
-	cancel_delayed_work_sync(&adev->vcn.inst[i].idle_work);
-
 	/* err_event_athub and dpc recovery will corrupt VCPU buffer, so we need to
 	 * restore fw data and clear buffer in amdgpu_vcn_resume() */
 	if (in_ras_intr || adev->pcie_reset_ctx.in_link_reset)
-- 
2.50.1

Re: [PATCH] drm/amdgpu/vcn: drop extra cancel_delayed_work_sync()

[Alex Deucher]

Ping?

On Wed, Aug 13, 2025 at 3:37 PM Alex Deucher <alexander.deucher@amd.com> wrote:
>
> We already call this in the hw_fini() methods for all
> VCN instances, so no need to call it again in
> amdgpu_vcn_suspend().
>
> Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
> ---
>  drivers/gpu/drm/amd/amdgpu/amdgpu_vcn.c | 2 --
>  1 file changed, 2 deletions(-)
>
> diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_vcn.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_vcn.c
> index 9a76e11d1c184..fd8ebf4b5a824 100644
> --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_vcn.c
> +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_vcn.c
> @@ -357,8 +357,6 @@ int amdgpu_vcn_suspend(struct amdgpu_device *adev, int i)
>         if (adev->vcn.harvest_config & (1 << i))
>                 return 0;
>
> -       cancel_delayed_work_sync(&adev->vcn.inst[i].idle_work);
> -
>         /* err_event_athub and dpc recovery will corrupt VCPU buffer, so we need to
>          * restore fw data and clear buffer in amdgpu_vcn_resume() */
>         if (in_ras_intr || adev->pcie_reset_ctx.in_link_reset)
> --
> 2.50.1
>

Re: [PATCH] drm/amdgpu/vcn: drop extra cancel_delayed_work_sync()

[Wu, David]

[-- Attachment #1: Type: text/plain, Size: 1439 bytes --]

[AMD Official Use Only - AMD Internal Distribution Only]


one thing I noticed:
   amdgpu_vcn_suspend() also gets called in vcn_v2_5_sw_fini() for example
   where there isn't cancel_delayed_work_sync() call. Does the removal work in this case?

David
On 8/20/2025 2:36 PM, Alex Deucher wrote:

Ping?

On Wed, Aug 13, 2025 at 3:37 PM Alex Deucher <alexander.deucher@amd.com><mailto:alexander.deucher@amd.com> wrote:


We already call this in the hw_fini() methods for all
VCN instances, so no need to call it again in
amdgpu_vcn_suspend().

Signed-off-by: Alex Deucher <alexander.deucher@amd.com><mailto:alexander.deucher@amd.com>
---
 drivers/gpu/drm/amd/amdgpu/amdgpu_vcn.c | 2 --
 1 file changed, 2 deletions(-)

diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_vcn.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_vcn.c
index 9a76e11d1c184..fd8ebf4b5a824 100644
--- a/drivers/gpu/drm/amd/amdgpu/amdgpu_vcn.c
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_vcn.c
@@ -357,8 +357,6 @@ int amdgpu_vcn_suspend(struct amdgpu_device *adev, int i)
        if (adev->vcn.harvest_config & (1 << i))
                return 0;

-       cancel_delayed_work_sync(&adev->vcn.inst[i].idle_work);
-
        /* err_event_athub and dpc recovery will corrupt VCPU buffer, so we need to
         * restore fw data and clear buffer in amdgpu_vcn_resume() */
        if (in_ras_intr || adev->pcie_reset_ctx.in_link_reset)
--
2.50.1




[-- Attachment #2: Type: text/html, Size: 2392 bytes --]

Re: [PATCH] drm/amdgpu/vcn: drop extra cancel_delayed_work_sync()

[Alex Deucher]

On Wed, Aug 20, 2025 at 3:29 PM Wu, David <David.Wu3@amd.com> wrote:
>
> [AMD Official Use Only - AMD Internal Distribution Only]
>
>
> one thing I noticed:
>    amdgpu_vcn_suspend() also gets called in vcn_v2_5_sw_fini() for example
>    where there isn't cancel_delayed_work_sync() call. Does the removal work in this case?

The sw callbacks are not supposed to touch any hardware, so that's
wrong to begin with.  That said, this should never end up doing
anything because we've already cancelled the delayed worker in hw_fini
and after hw_fini the hw is stopped so no work should be outstanding.

Alex

>
> David
> On 8/20/2025 2:36 PM, Alex Deucher wrote:
>
> Ping?
>
> On Wed, Aug 13, 2025 at 3:37 PM Alex Deucher <alexander.deucher@amd.com> wrote:
>
> We already call this in the hw_fini() methods for all
> VCN instances, so no need to call it again in
> amdgpu_vcn_suspend().
>
> Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
> ---
>  drivers/gpu/drm/amd/amdgpu/amdgpu_vcn.c | 2 --
>  1 file changed, 2 deletions(-)
>
> diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_vcn.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_vcn.c
> index 9a76e11d1c184..fd8ebf4b5a824 100644
> --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_vcn.c
> +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_vcn.c
> @@ -357,8 +357,6 @@ int amdgpu_vcn_suspend(struct amdgpu_device *adev, int i)
>         if (adev->vcn.harvest_config & (1 << i))
>                 return 0;
>
> -       cancel_delayed_work_sync(&adev->vcn.inst[i].idle_work);
> -
>         /* err_event_athub and dpc recovery will corrupt VCPU buffer, so we need to
>          * restore fw data and clear buffer in amdgpu_vcn_resume() */
>         if (in_ras_intr || adev->pcie_reset_ctx.in_link_reset)
> --
> 2.50.1
>
>

RE: [PATCH] drm/amdgpu/vcn: drop extra cancel_delayed_work_sync()

[Liu, Leo]

[AMD Official Use Only - AMD Internal Distribution Only]

Looks reasonable to me, the patch is:
Reviewed-by: Leo Liu <leo.liu@amd.com>

> -----Original Message-----
> From: amd-gfx <amd-gfx-bounces@lists.freedesktop.org> On Behalf Of Alex
> Deucher
> Sent: August 13, 2025 3:37 PM
> To: amd-gfx@lists.freedesktop.org
> Cc: Deucher, Alexander <Alexander.Deucher@amd.com>
> Subject: [PATCH] drm/amdgpu/vcn: drop extra cancel_delayed_work_sync()
>
> We already call this in the hw_fini() methods for all VCN instances, so no need
> to call it again in amdgpu_vcn_suspend().
>
> Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
> ---
>  drivers/gpu/drm/amd/amdgpu/amdgpu_vcn.c | 2 --
>  1 file changed, 2 deletions(-)
>
> diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_vcn.c
> b/drivers/gpu/drm/amd/amdgpu/amdgpu_vcn.c
> index 9a76e11d1c184..fd8ebf4b5a824 100644
> --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_vcn.c
> +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_vcn.c
> @@ -357,8 +357,6 @@ int amdgpu_vcn_suspend(struct amdgpu_device
> *adev, int i)
>       if (adev->vcn.harvest_config & (1 << i))
>               return 0;
>
> -     cancel_delayed_work_sync(&adev->vcn.inst[i].idle_work);
> -
>       /* err_event_athub and dpc recovery will corrupt VCPU buffer, so we
> need to
>        * restore fw data and clear buffer in amdgpu_vcn_resume() */
>       if (in_ras_intr || adev->pcie_reset_ctx.in_link_reset)
> --
> 2.50.1

Re: [PATCH] drm/amdgpu/vcn: drop extra cancel_delayed_work_sync()

[David Wu]

[-- Attachment #1: Type: text/plain, Size: 2240 bytes --]

Did some tests with the sw callbacks which were called with rmmod.
It will fail if there are jobs, otherwise no need to cancel delayed work as
there is not any outstanding work and idle handler called or not did not affect
the removal of the module. insmod continues to work after that.
The case for hw_fini works as expected.
Tested-by: David (Ming Qiang) Wu<David.Wu3@amd.com>

Thanks,

David

On 2025-08-20 15:43, Alex Deucher wrote:
> On Wed, Aug 20, 2025 at 3:29 PM Wu, David<David.Wu3@amd.com> wrote:
>> [AMD Official Use Only - AMD Internal Distribution Only]
>>
>>
>> one thing I noticed:
>>     amdgpu_vcn_suspend() also gets called in vcn_v2_5_sw_fini() for example
>>     where there isn't cancel_delayed_work_sync() call. Does the removal work in this case?
> The sw callbacks are not supposed to touch any hardware, so that's
> wrong to begin with.  That said, this should never end up doing
> anything because we've already cancelled the delayed worker in hw_fini
> and after hw_fini the hw is stopped so no work should be outstanding.
>
> Alex
>
>> David
>> On 8/20/2025 2:36 PM, Alex Deucher wrote:
>>
>> Ping?
>>
>> On Wed, Aug 13, 2025 at 3:37 PM Alex Deucher<alexander.deucher@amd.com> wrote:
>>
>> We already call this in the hw_fini() methods for all
>> VCN instances, so no need to call it again in
>> amdgpu_vcn_suspend().
>>
>> Signed-off-by: Alex Deucher<alexander.deucher@amd.com>
>> ---
>>   drivers/gpu/drm/amd/amdgpu/amdgpu_vcn.c | 2 --
>>   1 file changed, 2 deletions(-)
>>
>> diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_vcn.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_vcn.c
>> index 9a76e11d1c184..fd8ebf4b5a824 100644
>> --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_vcn.c
>> +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_vcn.c
>> @@ -357,8 +357,6 @@ int amdgpu_vcn_suspend(struct amdgpu_device *adev, int i)
>>          if (adev->vcn.harvest_config & (1 << i))
>>                  return 0;
>>
>> -       cancel_delayed_work_sync(&adev->vcn.inst[i].idle_work);
>> -
>>          /* err_event_athub and dpc recovery will corrupt VCPU buffer, so we need to
>>           * restore fw data and clear buffer in amdgpu_vcn_resume() */
>>          if (in_ras_intr || adev->pcie_reset_ctx.in_link_reset)
>> --
>> 2.50.1
>>
>>

[-- Attachment #2: Type: text/html, Size: 3114 bytes --]