[PATCH] drm/sched: avoid killing parent entity on child SIGKILL v3

[PATCH] drm/sched: avoid killing parent entity on child SIGKILL v3

[Christian König]

From: David Rosca <david.rosca@amd.com>

The DRM scheduler tracks who last uses an entity and when that process
is killed blocks all further submissions to that entity.

The problem is that we didn't track who initially created an entity, so
when a process accidently leaked its file descriptor to a child and
that child got killed, we killed the parent's entities.

Avoid that and instead initialize the entities last user on entity
creation. This also allows to drop the extra NULL check.

v2: still use cmpxchg
v3: improve the commit message

Signed-off-by: David Rosca <david.rosca@amd.com>
Signed-off-by: Christian König <christian.koenig@amd.com>
Closes: https://gitlab.freedesktop.org/drm/amd/-/issues/4568
Reviewed-by: Alex Deucher <alexander.deucher@amd.com>
CC: stable@vger.kernel.org
---
 drivers/gpu/drm/scheduler/sched_entity.c | 3 ++-
 1 file changed, 2 insertions(+), 1 deletion(-)

diff --git a/drivers/gpu/drm/scheduler/sched_entity.c b/drivers/gpu/drm/scheduler/sched_entity.c
index 5a4697f636f2..3e2f83dc3f24 100644
--- a/drivers/gpu/drm/scheduler/sched_entity.c
+++ b/drivers/gpu/drm/scheduler/sched_entity.c
@@ -70,6 +70,7 @@ int drm_sched_entity_init(struct drm_sched_entity *entity,
 	entity->guilty = guilty;
 	entity->num_sched_list = num_sched_list;
 	entity->priority = priority;
+	entity->last_user = current->group_leader;
 	/*
 	 * It's perfectly valid to initialize an entity without having a valid
 	 * scheduler attached. It's just not valid to use the scheduler before it
@@ -302,7 +303,7 @@ long drm_sched_entity_flush(struct drm_sched_entity *entity, long timeout)
 
 	/* For a killed process disallow further enqueueing of jobs. */
 	last_user = cmpxchg(&entity->last_user, current->group_leader, NULL);
-	if ((!last_user || last_user == current->group_leader) &&
+	if (last_user == current->group_leader &&
 	    (current->flags & PF_EXITING) && (current->exit_code == SIGKILL))
 		drm_sched_entity_kill(entity);
 
-- 
2.43.0

Re: [PATCH] drm/sched: avoid killing parent entity on child SIGKILL v3

[Philipp Stanner]

On Wed, 2025-10-15 at 16:01 +0200, Christian König wrote:
> From: David Rosca <david.rosca@amd.com>
> 
> The DRM scheduler tracks who last uses an entity and when that process
> is killed blocks all further submissions to that entity.
> 
> The problem is that we didn't track who initially created an entity, so
> when a process accidently leaked its file descriptor to a child and
> that child got killed, we killed the parent's entities.
> 
> Avoid that and instead initialize the entities last user on entity
> creation. This also allows to drop the extra NULL check.
> 
> v2: still use cmpxchg
> v3: improve the commit message

For the future, commit messages in the patche's comment body are to be
preferred since it's common kernel style. Same applies to the patch
version in the title, which should be in [PATCH v3].

But that's just a nit. More important:

> 
> Signed-off-by: David Rosca <david.rosca@amd.com>
> Signed-off-by: Christian König <christian.koenig@amd.com>
> Closes: https://gitlab.freedesktop.org/drm/amd/-/issues/4568

Should this have a Fixes: ?

> Reviewed-by: Alex Deucher <alexander.deucher@amd.com>
> CC: stable@vger.kernel.org

So we want it in drm-misc-fixes, don't we?


P.

> ---
>  drivers/gpu/drm/scheduler/sched_entity.c | 3 ++-
>  1 file changed, 2 insertions(+), 1 deletion(-)
> 
> diff --git a/drivers/gpu/drm/scheduler/sched_entity.c b/drivers/gpu/drm/scheduler/sched_entity.c
> index 5a4697f636f2..3e2f83dc3f24 100644
> --- a/drivers/gpu/drm/scheduler/sched_entity.c
> +++ b/drivers/gpu/drm/scheduler/sched_entity.c
> @@ -70,6 +70,7 @@ int drm_sched_entity_init(struct drm_sched_entity *entity,
>  	entity->guilty = guilty;
>  	entity->num_sched_list = num_sched_list;
>  	entity->priority = priority;
> +	entity->last_user = current->group_leader;
>  	/*
>  	 * It's perfectly valid to initialize an entity without having a valid
>  	 * scheduler attached. It's just not valid to use the scheduler before it
> @@ -302,7 +303,7 @@ long drm_sched_entity_flush(struct drm_sched_entity *entity, long timeout)
>  
>  	/* For a killed process disallow further enqueueing of jobs. */
>  	last_user = cmpxchg(&entity->last_user, current->group_leader, NULL);
> -	if ((!last_user || last_user == current->group_leader) &&
> +	if (last_user == current->group_leader &&
>  	    (current->flags & PF_EXITING) && (current->exit_code == SIGKILL))
>  		drm_sched_entity_kill(entity);
>  

Re: [PATCH] drm/sched: avoid killing parent entity on child SIGKILL v3

[Christian König]

On 16.10.25 14:31, Philipp Stanner wrote:
> On Wed, 2025-10-15 at 16:01 +0200, Christian König wrote:
>> From: David Rosca <david.rosca@amd.com>
>>
>> The DRM scheduler tracks who last uses an entity and when that process
>> is killed blocks all further submissions to that entity.
>>
>> The problem is that we didn't track who initially created an entity, so
>> when a process accidently leaked its file descriptor to a child and
>> that child got killed, we killed the parent's entities.
>>
>> Avoid that and instead initialize the entities last user on entity
>> creation. This also allows to drop the extra NULL check.
>>
>> v2: still use cmpxchg
>> v3: improve the commit message
> 
> For the future, commit messages in the patche's comment body are to be
> preferred since it's common kernel style. Same applies to the patch
> version in the title, which should be in [PATCH v3].

Ah, just forgotten about it!

> 
> But that's just a nit. More important:
> 
>>
>> Signed-off-by: David Rosca <david.rosca@amd.com>
>> Signed-off-by: Christian König <christian.koenig@amd.com>
>> Closes: https://gitlab.freedesktop.org/drm/amd/-/issues/4568
> 
> Should this have a Fixes: ?

No, I've actually removed that because the patch which made it obvious that something is wrong here is correct.

It's just that this seems to be incorrect ever since we added the code.

> 
>> Reviewed-by: Alex Deucher <alexander.deucher@amd.com>
>> CC: stable@vger.kernel.org
> 
> So we want it in drm-misc-fixes, don't we?

Yes, the patch is based on drm-misc-fixes. I can push it when you give me an rb.

Alternatively you can push it yourself, whatever you prefer.

Regards,
Christian.

> 
> 
> P.
> 
>> ---
>>  drivers/gpu/drm/scheduler/sched_entity.c | 3 ++-
>>  1 file changed, 2 insertions(+), 1 deletion(-)
>>
>> diff --git a/drivers/gpu/drm/scheduler/sched_entity.c b/drivers/gpu/drm/scheduler/sched_entity.c
>> index 5a4697f636f2..3e2f83dc3f24 100644
>> --- a/drivers/gpu/drm/scheduler/sched_entity.c
>> +++ b/drivers/gpu/drm/scheduler/sched_entity.c
>> @@ -70,6 +70,7 @@ int drm_sched_entity_init(struct drm_sched_entity *entity,
>>  	entity->guilty = guilty;
>>  	entity->num_sched_list = num_sched_list;
>>  	entity->priority = priority;
>> +	entity->last_user = current->group_leader;
>>  	/*
>>  	 * It's perfectly valid to initialize an entity without having a valid
>>  	 * scheduler attached. It's just not valid to use the scheduler before it
>> @@ -302,7 +303,7 @@ long drm_sched_entity_flush(struct drm_sched_entity *entity, long timeout)
>>  
>>  	/* For a killed process disallow further enqueueing of jobs. */
>>  	last_user = cmpxchg(&entity->last_user, current->group_leader, NULL);
>> -	if ((!last_user || last_user == current->group_leader) &&
>> +	if (last_user == current->group_leader &&
>>  	    (current->flags & PF_EXITING) && (current->exit_code == SIGKILL))
>>  		drm_sched_entity_kill(entity);
>>  
> 

Re: [PATCH] drm/sched: avoid killing parent entity on child SIGKILL v3

[Philipp Stanner]

On Thu, 2025-10-16 at 15:11 +0200, Christian König wrote:
> On 16.10.25 14:31, Philipp Stanner wrote:
> > On Wed, 2025-10-15 at 16:01 +0200, Christian König wrote:
> > > From: David Rosca <david.rosca@amd.com>
> > > 
> > > The DRM scheduler tracks who last uses an entity and when that process
> > > is killed blocks all further submissions to that entity.
> > > 
> > > The problem is that we didn't track who initially created an entity, so
> > > when a process accidently leaked its file descriptor to a child and
> > > that child got killed, we killed the parent's entities.
> > > 
> > > Avoid that and instead initialize the entities last user on entity
> > > creation. This also allows to drop the extra NULL check.
> > > 
> > > v2: still use cmpxchg
> > > v3: improve the commit message
> > 
> > For the future, commit messages in the patche's comment body are to be
> > preferred since it's common kernel style. Same applies to the patch
> > version in the title, which should be in [PATCH v3].
> 
> Ah, just forgotten about it!
> 
> > 
> > But that's just a nit. More important:
> > 
> > > 
> > > Signed-off-by: David Rosca <david.rosca@amd.com>
> > > Signed-off-by: Christian König <christian.koenig@amd.com>
> > > Closes: https://gitlab.freedesktop.org/drm/amd/-/issues/4568
> > 
> > Should this have a Fixes: ?
> 
> No, I've actually removed that because the patch which made it obvious that something is wrong here is correct.
> 
> It's just that this seems to be incorrect ever since we added the code.

Then we should just add the Fixes: tag for the big bang commit,
shouldn't we?

At least maintainer-tools/dim doesn't like the missing tag at all. When
trying to apply this patch it just added the following:

Signed-off-by: David Rosca <david.rosca@amd.com>
Signed-off-by: Christian König <christian.koenig@amd.com>
Closes: https://gitlab.freedesktop.org/drm/amd/-/issues/4568
Reviewed-by: Alex Deucher <alexander.deucher@amd.com>
CC: stable@vger.kernel.org
Fixes: 43bce41cf48e ("drm/scheduler: only kill entity if last user is  killed v2")
Signed-off-by: Philipp Stanner <phasta@kernel.org>
Link: https://lore.kernel.org/r/20251015140128.1470-1-christian.koenig@amd.com

and then it complains about the tag it added itself:

Applying: drm/sched: avoid killing parent entity on child SIGKILL v3
[drm-misc-fixes 2a3e82c80bd0] drm/sched: avoid killing parent entity on child SIGKILL v3
 Author: David Rosca <david.rosca@amd.com>
 Date: Wed Oct 15 16:01:28 2025 +0200
 1 file changed, 2 insertions(+), 1 deletion(-)
2a3e82c80bd0 (HEAD -> drm-misc-fixes) drm/sched: avoid killing parent entity on child SIGKILL v3
-:27: WARNING:BAD_FIXES_TAG: Please use correct Fixes: style 'Fixes: <12+ chars of sha1> ("<title line>")' - ie: 'Fixes: 43bce41cf48e ("drm/scheduler: only kill entity if last user is killed v2")'
#27: 
Fixes: 43bce41cf48e ("drm/scheduler: only kill entity if last user is  killed v2")

-:48: CHECK:UNNECESSARY_PARENTHESES: Unnecessary parentheses around 'current->exit_code == SIGKILL'
#48: FILE: drivers/gpu/drm/scheduler/sched_entity.c:306:
+	if (last_user == current->group_leader &&
 	    (current->flags & PF_EXITING) && (current->exit_code == SIGKILL))

total: 0 errors, 1 warnings, 1 checks, 15 lines checked


Which is weird..

in any case the big bang commit helps stable to apply this correctly,
too.

P.

> 
> > 
> > > Reviewed-by: Alex Deucher <alexander.deucher@amd.com>
> > > CC: stable@vger.kernel.org
> > 
> > So we want it in drm-misc-fixes, don't we?
> 
> Yes, the patch is based on drm-misc-fixes. I can push it when you give me an rb.
> 
> Alternatively you can push it yourself, whatever you prefer.
> 
> Regards,
> Christian.
> 
> > 
> > 
> > P.
> > 
> > > ---
> > >  drivers/gpu/drm/scheduler/sched_entity.c | 3 ++-
> > >  1 file changed, 2 insertions(+), 1 deletion(-)
> > > 
> > > diff --git a/drivers/gpu/drm/scheduler/sched_entity.c b/drivers/gpu/drm/scheduler/sched_entity.c
> > > index 5a4697f636f2..3e2f83dc3f24 100644
> > > --- a/drivers/gpu/drm/scheduler/sched_entity.c
> > > +++ b/drivers/gpu/drm/scheduler/sched_entity.c
> > > @@ -70,6 +70,7 @@ int drm_sched_entity_init(struct drm_sched_entity *entity,
> > >  	entity->guilty = guilty;
> > >  	entity->num_sched_list = num_sched_list;
> > >  	entity->priority = priority;
> > > +	entity->last_user = current->group_leader;
> > >  	/*
> > >  	 * It's perfectly valid to initialize an entity without having a valid
> > >  	 * scheduler attached. It's just not valid to use the scheduler before it
> > > @@ -302,7 +303,7 @@ long drm_sched_entity_flush(struct drm_sched_entity *entity, long timeout)
> > >  
> > >  	/* For a killed process disallow further enqueueing of jobs. */
> > >  	last_user = cmpxchg(&entity->last_user, current->group_leader, NULL);
> > > -	if ((!last_user || last_user == current->group_leader) &&
> > > +	if (last_user == current->group_leader &&
> > >  	    (current->flags & PF_EXITING) && (current->exit_code == SIGKILL))
> > >  		drm_sched_entity_kill(entity);
> > >  
> > 
> 

Re: [PATCH] drm/sched: avoid killing parent entity on child SIGKILL v3

[Christian König]

On 16.10.25 19:20, Philipp Stanner wrote:
> On Thu, 2025-10-16 at 15:11 +0200, Christian König wrote:
>> On 16.10.25 14:31, Philipp Stanner wrote:
>>> On Wed, 2025-10-15 at 16:01 +0200, Christian König wrote:
>>>> From: David Rosca <david.rosca@amd.com>
>>>>
>>>> The DRM scheduler tracks who last uses an entity and when that process
>>>> is killed blocks all further submissions to that entity.
>>>>
>>>> The problem is that we didn't track who initially created an entity, so
>>>> when a process accidently leaked its file descriptor to a child and
>>>> that child got killed, we killed the parent's entities.
>>>>
>>>> Avoid that and instead initialize the entities last user on entity
>>>> creation. This also allows to drop the extra NULL check.
>>>>
>>>> v2: still use cmpxchg
>>>> v3: improve the commit message
>>>
>>> For the future, commit messages in the patche's comment body are to be
>>> preferred since it's common kernel style. Same applies to the patch
>>> version in the title, which should be in [PATCH v3].
>>
>> Ah, just forgotten about it!
>>
>>>
>>> But that's just a nit. More important:
>>>
>>>>
>>>> Signed-off-by: David Rosca <david.rosca@amd.com>
>>>> Signed-off-by: Christian König <christian.koenig@amd.com>
>>>> Closes: https://gitlab.freedesktop.org/drm/amd/-/issues/4568
>>>
>>> Should this have a Fixes: ?
>>
>> No, I've actually removed that because the patch which made it obvious that something is wrong here is correct.
>>
>> It's just that this seems to be incorrect ever since we added the code.
> 
> Then we should just add the Fixes: tag for the big bang commit,
> shouldn't we?
> 
> At least maintainer-tools/dim doesn't like the missing tag at all. When
> trying to apply this patch it just added the following:
> 
> Signed-off-by: David Rosca <david.rosca@amd.com>
> Signed-off-by: Christian König <christian.koenig@amd.com>
> Closes: https://gitlab.freedesktop.org/drm/amd/-/issues/4568
> Reviewed-by: Alex Deucher <alexander.deucher@amd.com>
> CC: stable@vger.kernel.org
> Fixes: 43bce41cf48e ("drm/scheduler: only kill entity if last user is  killed v2")
> Signed-off-by: Philipp Stanner <phasta@kernel.org>
> Link: https://lore.kernel.org/r/20251015140128.1470-1-christian.koenig@amd.com
> 
> and then it complains about the tag it added itself:
> 
> Applying: drm/sched: avoid killing parent entity on child SIGKILL v3
> [drm-misc-fixes 2a3e82c80bd0] drm/sched: avoid killing parent entity on child SIGKILL v3
>  Author: David Rosca <david.rosca@amd.com>
>  Date: Wed Oct 15 16:01:28 2025 +0200
>  1 file changed, 2 insertions(+), 1 deletion(-)
> 2a3e82c80bd0 (HEAD -> drm-misc-fixes) drm/sched: avoid killing parent entity on child SIGKILL v3
> -:27: WARNING:BAD_FIXES_TAG: Please use correct Fixes: style 'Fixes: <12+ chars of sha1> ("<title line>")' - ie: 'Fixes: 43bce41cf48e ("drm/scheduler: only kill entity if last user is killed v2")'
> #27: 
> Fixes: 43bce41cf48e ("drm/scheduler: only kill entity if last user is  killed v2")
> 
> -:48: CHECK:UNNECESSARY_PARENTHESES: Unnecessary parentheses around 'current->exit_code == SIGKILL'
> #48: FILE: drivers/gpu/drm/scheduler/sched_entity.c:306:
> +	if (last_user == current->group_leader &&
>  	    (current->flags & PF_EXITING) && (current->exit_code == SIGKILL))
> 
> total: 0 errors, 1 warnings, 1 checks, 15 lines checked
> 
> 
> Which is weird..
> 
> in any case the big bang commit helps stable to apply this correctly,
> too.

Last time I talked with Greg about it he mentioned that the Fixes tag just makes sure that this fix is applied everywhere the original patch is applied as well.

Since the original patch is so old CC: stable should be sufficient as far as I can see.

That dim complains about missing fixes tags is news to me.

Regards,
Christian.

> 
> P.
> 
>>
>>>
>>>> Reviewed-by: Alex Deucher <alexander.deucher@amd.com>
>>>> CC: stable@vger.kernel.org
>>>
>>> So we want it in drm-misc-fixes, don't we?
>>
>> Yes, the patch is based on drm-misc-fixes. I can push it when you give me an rb.
>>
>> Alternatively you can push it yourself, whatever you prefer.
>>
>> Regards,
>> Christian.
>>
>>>
>>>
>>> P.
>>>
>>>> ---
>>>>  drivers/gpu/drm/scheduler/sched_entity.c | 3 ++-
>>>>  1 file changed, 2 insertions(+), 1 deletion(-)
>>>>
>>>> diff --git a/drivers/gpu/drm/scheduler/sched_entity.c b/drivers/gpu/drm/scheduler/sched_entity.c
>>>> index 5a4697f636f2..3e2f83dc3f24 100644
>>>> --- a/drivers/gpu/drm/scheduler/sched_entity.c
>>>> +++ b/drivers/gpu/drm/scheduler/sched_entity.c
>>>> @@ -70,6 +70,7 @@ int drm_sched_entity_init(struct drm_sched_entity *entity,
>>>>  	entity->guilty = guilty;
>>>>  	entity->num_sched_list = num_sched_list;
>>>>  	entity->priority = priority;
>>>> +	entity->last_user = current->group_leader;
>>>>  	/*
>>>>  	 * It's perfectly valid to initialize an entity without having a valid
>>>>  	 * scheduler attached. It's just not valid to use the scheduler before it
>>>> @@ -302,7 +303,7 @@ long drm_sched_entity_flush(struct drm_sched_entity *entity, long timeout)
>>>>  
>>>>  	/* For a killed process disallow further enqueueing of jobs. */
>>>>  	last_user = cmpxchg(&entity->last_user, current->group_leader, NULL);
>>>> -	if ((!last_user || last_user == current->group_leader) &&
>>>> +	if (last_user == current->group_leader &&
>>>>  	    (current->flags & PF_EXITING) && (current->exit_code == SIGKILL))
>>>>  		drm_sched_entity_kill(entity);
>>>>  
>>>
>>
> 

Re: [PATCH] drm/sched: avoid killing parent entity on child SIGKILL v3

[Philipp Stanner]

On Wed, 2025-10-15 at 16:01 +0200, Christian König wrote:
> From: David Rosca <david.rosca@amd.com>
> 
> The DRM scheduler tracks who last uses an entity and when that process
> is killed blocks all further submissions to that entity.
> 
> The problem is that we didn't track who initially created an entity, so
> when a process accidently leaked its file descriptor to a child and
> that child got killed, we killed the parent's entities.
> 
> Avoid that and instead initialize the entities last user on entity
> creation. This also allows to drop the extra NULL check.
> 
> v2: still use cmpxchg
> v3: improve the commit message
> 
> Signed-off-by: David Rosca <david.rosca@amd.com>
> Signed-off-by: Christian König <christian.koenig@amd.com>
> Closes: https://gitlab.freedesktop.org/drm/amd/-/issues/4568
> Reviewed-by: Alex Deucher <alexander.deucher@amd.com>
> CC: stable@vger.kernel.org

Acked-by: Philipp Stanner <phasta@kernel.org>


Fire at will, Christian. Maybe optionally with the commit message nits
twirked in we discussed before.


P.

> ---
>  drivers/gpu/drm/scheduler/sched_entity.c | 3 ++-
>  1 file changed, 2 insertions(+), 1 deletion(-)
> 
> diff --git a/drivers/gpu/drm/scheduler/sched_entity.c b/drivers/gpu/drm/scheduler/sched_entity.c
> index 5a4697f636f2..3e2f83dc3f24 100644
> --- a/drivers/gpu/drm/scheduler/sched_entity.c
> +++ b/drivers/gpu/drm/scheduler/sched_entity.c
> @@ -70,6 +70,7 @@ int drm_sched_entity_init(struct drm_sched_entity *entity,
>  	entity->guilty = guilty;
>  	entity->num_sched_list = num_sched_list;
>  	entity->priority = priority;
> +	entity->last_user = current->group_leader;
>  	/*
>  	 * It's perfectly valid to initialize an entity without having a valid
>  	 * scheduler attached. It's just not valid to use the scheduler before it
> @@ -302,7 +303,7 @@ long drm_sched_entity_flush(struct drm_sched_entity *entity, long timeout)
>  
>  	/* For a killed process disallow further enqueueing of jobs. */
>  	last_user = cmpxchg(&entity->last_user, current->group_leader, NULL);
> -	if ((!last_user || last_user == current->group_leader) &&
> +	if (last_user == current->group_leader &&
>  	    (current->flags & PF_EXITING) && (current->exit_code == SIGKILL))
>  		drm_sched_entity_kill(entity);
>  

Re: [PATCH] drm/sched: avoid killing parent entity on child SIGKILL v3

[Alex Deucher]

Looks like this hasn't landed yet.  Can someone push this?

Alex

On Fri, Oct 17, 2025 at 2:18 AM Philipp Stanner <phasta@mailbox.org> wrote:
>
> On Wed, 2025-10-15 at 16:01 +0200, Christian König wrote:
> > From: David Rosca <david.rosca@amd.com>
> >
> > The DRM scheduler tracks who last uses an entity and when that process
> > is killed blocks all further submissions to that entity.
> >
> > The problem is that we didn't track who initially created an entity, so
> > when a process accidently leaked its file descriptor to a child and
> > that child got killed, we killed the parent's entities.
> >
> > Avoid that and instead initialize the entities last user on entity
> > creation. This also allows to drop the extra NULL check.
> >
> > v2: still use cmpxchg
> > v3: improve the commit message
> >
> > Signed-off-by: David Rosca <david.rosca@amd.com>
> > Signed-off-by: Christian König <christian.koenig@amd.com>
> > Closes: https://gitlab.freedesktop.org/drm/amd/-/issues/4568
> > Reviewed-by: Alex Deucher <alexander.deucher@amd.com>
> > CC: stable@vger.kernel.org
>
> Acked-by: Philipp Stanner <phasta@kernel.org>
>
>
> Fire at will, Christian. Maybe optionally with the commit message nits
> twirked in we discussed before.
>
>
> P.
>
> > ---
> >  drivers/gpu/drm/scheduler/sched_entity.c | 3 ++-
> >  1 file changed, 2 insertions(+), 1 deletion(-)
> >
> > diff --git a/drivers/gpu/drm/scheduler/sched_entity.c b/drivers/gpu/drm/scheduler/sched_entity.c
> > index 5a4697f636f2..3e2f83dc3f24 100644
> > --- a/drivers/gpu/drm/scheduler/sched_entity.c
> > +++ b/drivers/gpu/drm/scheduler/sched_entity.c
> > @@ -70,6 +70,7 @@ int drm_sched_entity_init(struct drm_sched_entity *entity,
> >       entity->guilty = guilty;
> >       entity->num_sched_list = num_sched_list;
> >       entity->priority = priority;
> > +     entity->last_user = current->group_leader;
> >       /*
> >        * It's perfectly valid to initialize an entity without having a valid
> >        * scheduler attached. It's just not valid to use the scheduler before it
> > @@ -302,7 +303,7 @@ long drm_sched_entity_flush(struct drm_sched_entity *entity, long timeout)
> >
> >       /* For a killed process disallow further enqueueing of jobs. */
> >       last_user = cmpxchg(&entity->last_user, current->group_leader, NULL);
> > -     if ((!last_user || last_user == current->group_leader) &&
> > +     if (last_user == current->group_leader &&
> >           (current->flags & PF_EXITING) && (current->exit_code == SIGKILL))
> >               drm_sched_entity_kill(entity);
> >
>

Re: [PATCH] drm/sched: avoid killing parent entity on child SIGKILL v3

[Philipp Stanner]

On Tue, 2025-10-28 at 09:07 -0400, Alex Deucher wrote:
> Looks like this hasn't landed yet.  Can someone push this?

Pushed it (with fixed commit message) to drm-misc-fixes.

I had Acked it for Christian but it seems that was lost.

Thx Alex

P.

> 
> Alex
> 
> On Fri, Oct 17, 2025 at 2:18 AM Philipp Stanner <phasta@mailbox.org> wrote:
> > 
> > On Wed, 2025-10-15 at 16:01 +0200, Christian König wrote:
> > > From: David Rosca <david.rosca@amd.com>
> > > 
> > > The DRM scheduler tracks who last uses an entity and when that process
> > > is killed blocks all further submissions to that entity.
> > > 
> > > The problem is that we didn't track who initially created an entity, so
> > > when a process accidently leaked its file descriptor to a child and
> > > that child got killed, we killed the parent's entities.
> > > 
> > > Avoid that and instead initialize the entities last user on entity
> > > creation. This also allows to drop the extra NULL check.
> > > 
> > > v2: still use cmpxchg
> > > v3: improve the commit message
> > > 
> > > Signed-off-by: David Rosca <david.rosca@amd.com>
> > > Signed-off-by: Christian König <christian.koenig@amd.com>
> > > Closes: https://gitlab.freedesktop.org/drm/amd/-/issues/4568
> > > Reviewed-by: Alex Deucher <alexander.deucher@amd.com>
> > > CC: stable@vger.kernel.org
> > 
> > Acked-by: Philipp Stanner <phasta@kernel.org>
> > 
> > 
> > Fire at will, Christian. Maybe optionally with the commit message nits
> > twirked in we discussed before.
> > 
> > 
> > P.

Re: [PATCH] drm/sched: avoid killing parent entity on child SIGKILL v3

[Christian König]

On 10/28/25 14:13, Philipp Stanner wrote:
> On Tue, 2025-10-28 at 09:07 -0400, Alex Deucher wrote:
>> Looks like this hasn't landed yet.  Can someone push this?
> 
> Pushed it (with fixed commit message) to drm-misc-fixes.
> 
> I had Acked it for Christian but it seems that was lost.

Sorry, been on vacation last week and it took me till today to get through all my mails and back to this issue.

Thanks for taking care of it.

Christian.

> 
> Thx Alex
> 
> P.
> 
>>
>> Alex
>>
>> On Fri, Oct 17, 2025 at 2:18 AM Philipp Stanner <phasta@mailbox.org> wrote:
>>>
>>> On Wed, 2025-10-15 at 16:01 +0200, Christian König wrote:
>>>> From: David Rosca <david.rosca@amd.com>
>>>>
>>>> The DRM scheduler tracks who last uses an entity and when that process
>>>> is killed blocks all further submissions to that entity.
>>>>
>>>> The problem is that we didn't track who initially created an entity, so
>>>> when a process accidently leaked its file descriptor to a child and
>>>> that child got killed, we killed the parent's entities.
>>>>
>>>> Avoid that and instead initialize the entities last user on entity
>>>> creation. This also allows to drop the extra NULL check.
>>>>
>>>> v2: still use cmpxchg
>>>> v3: improve the commit message
>>>>
>>>> Signed-off-by: David Rosca <david.rosca@amd.com>
>>>> Signed-off-by: Christian König <christian.koenig@amd.com>
>>>> Closes: https://gitlab.freedesktop.org/drm/amd/-/issues/4568
>>>> Reviewed-by: Alex Deucher <alexander.deucher@amd.com>
>>>> CC: stable@vger.kernel.org
>>>
>>> Acked-by: Philipp Stanner <phasta@kernel.org>
>>>
>>>
>>> Fire at will, Christian. Maybe optionally with the commit message nits
>>> twirked in we discussed before.
>>>
>>>
>>> P.