On Wed, 2024-04-17 at 09:37 +1000, Ben Skeggs wrote:
> - zero reason to do otherwise
to do _what_ otherwise? that refers to the title, it seems. But one
could describe why it was ever done that way (older architecture? bug?
mistake?) in the first place.
The commit messages in this entire series a
On Wed, 2024-05-22 at 08:53 +0900, FUJITA Tomonori wrote:
> Hi,
> Thanks for working on the firmware API!
>
> On Mon, 20 May 2024 19:24:19 +0200
> Danilo Krummrich wrote:
>
> > Add an abstraction around the kernels firmware API to request
> > firmware
> > images. The abstraction provides functio
On Tue, 2024-06-04 at 10:27 -0500, Blake McBride wrote:
> Greetings,
>
> I have used the nouveau driver with my Nvidia card on Linux. Works
> fine. However, my problem has to do with running LLM on my GPU with
> your driver. My impression is, it doesn't work. Am I correct?
Yo,
"it doesn't wo
7;s "timeout"
parameter. The actual "hang_limit" parameter is directly set to 0.
Define a new variable and rename the existing one to make naming
congruent with the function API.
Signed-off-by: Philipp Stanner
---
drivers/gpu/drm/nouveau/nouveau_sched.c | 5 +++--
1 file
7;s "timeout"
parameter. The actual "hang_limit" parameter is directly set to 0.
Rename "job_hang_limit" to "timeout".
Signed-off-by: Philipp Stanner
---
Changes in v2:
- Remove variable "hang_limit". (Danilo)
---
drivers/gpu/drm/nouveau/nouveau_sc
On Tue, 2023-09-19 at 06:21 +1000, Ben Skeggs wrote:
> From: Ben Skeggs
>
> RM (and GH100) support 8 NVJPG instances.
I don't think the commit message provides enough information.
Instinctively I would read the RM as "remove", thus "remove support
[for] 8 NVJPG instances" ???
Two sentences (wit
On Tue, 2023-09-19 at 06:21 +1000, Ben Skeggs wrote:
> From: Ben Skeggs
>
> - (temporarily) disable if GSP-RM detected, will be added later
disable _what_?
The other commit messages at least briefly name the component. This one
should as well.
Furthermore, I'd say that the wording should be som
On Tue, 2023-09-19 at 06:21 +1000, Ben Skeggs wrote:
> From: Ben Skeggs
>
> - preparation for GSP-RM, which has massive FW images
> - based on a patch by Dave Airlie
Probably more canonical to use one of the standard phrases such as
Suggested-by
>
> Signed-off-by: Ben Skeggs
> ---
> .../drm
On Tue, 2023-09-19 at 06:21 +1000, Ben Skeggs wrote:
> From: Ben Skeggs
>
> Adds the plumbing to start making RM control calls, and initialises
> objects to represent internal RM objects provided to us during init.
>
> These will be used by subsequent patches.
>
> Signed-off-by: Ben Skeggs
> -
On Tue, 2023-09-19 at 06:21 +1000, Ben Skeggs wrote:
> From: Ben Skeggs
>
> Adds support for allocating VIDEO_ENCODER classes from RM.
>
> Signed-off-by: Ben Skeggs
> ---
> drivers/gpu/drm/nouveau/include/nvif/class.h | 4 +
> .../drm/nouveau/include/nvkm/engine/nvenc.h | 2 +
> .../535
On Tue, 2023-11-07 at 17:34 +0300, Dan Carpenter wrote:
> Hello Ben Skeggs,
Hi,
FYI, Ben is not maintaining Nouveau anymore. The MAINTAINERS file has
been updated in that regard.
P.
>
> The patch 176fdcbddfd2: "drm/nouveau/gsp/r535: add support for
> booting GSP-RM" from Sep 19, 2023 (linux-ne
On Mon, 2024-11-11 at 13:41 +1000, Dave Airlie wrote:
> From: Dave Airlie
>
> The upper layer transfer functions expect EBUSY as a return
> for when retries should be done.
>
> Fix the AUX error translation, but also check for both errors
> in a few places.
>
> Fixes: eb284f4b3781 ("drm/nouveau
_init()").
Introduce a new struct for the scheduler init parameters and port all
users.
Signed-off-by: Philipp Stanner
Reviewed-by: Liviu Dudau
Acked-by: Matthew Brost # for Xe
Reviewed-by: Boris Brezillon # for Panfrost and
Panthor
Reviewed-by: Christian Gmeiner # for Etnaviv
Reviewed
me in
nouveau_sched_init()").
Introduce a new struct for the scheduler init parameters and port all
users.
Signed-off-by: Philipp Stanner
Reviewed-by: Liviu Dudau
Acked-by: Matthew Brost # for Xe
Reviewed-by: Boris Brezillon # for Panfrost and
Panthor
Reviewed-by: Christian Gmeiner # for Etnavi
On Wed, 2025-01-22 at 15:34 +0100, Christian König wrote:
> Am 22.01.25 um 15:08 schrieb Philipp Stanner:
> > drm_sched_init() has a great many parameters and upcoming new
> > functionality for the scheduler might add even more. Generally, the
> > great number of parameters re
On Wed, 2025-01-22 at 16:06 +0100, Christian König wrote:
> Am 22.01.25 um 15:48 schrieb Philipp Stanner:
> > On Wed, 2025-01-22 at 15:34 +0100, Christian König wrote:
> > > Am 22.01.25 um 15:08 schrieb Philipp Stanner:
> > > > drm_sched_init() has a great ma
me in
nouveau_sched_init()").
Introduce a new struct for the scheduler init parameters and port all
users.
Signed-off-by: Philipp Stanner
---
Howdy,
I have a patch-series in the pipe that will add a `flags` argument to
drm_sched_init(). I thought it would be wise to first rework the API as
detailed in this p
On Thu, 2025-01-23 at 09:10 +0100, Philipp Stanner wrote:
> On Wed, 2025-01-22 at 19:07 -0300, Maíra Canal wrote:
> > Hi Philipp,
> >
> > On 22/01/25 11:08, Philipp Stanner wrote:
> > > drm_sched_init() has a great many parameters and upcoming new
> > > fu
t; > On Wed, 22 Jan 2025 15:08:20 +0100
> > > > Philipp Stanner wrote:
> > > >
> > > > > --- a/drivers/gpu/drm/panthor/panthor_sched.c
> > > > > +++ b/drivers/gpu/drm/panthor/panthor_sched.c
> > > > > @@ -3272,6 +3272,7 @@ group_creat
On Wed, 2025-01-22 at 19:07 -0300, Maíra Canal wrote:
> Hi Philipp,
>
> On 22/01/25 11:08, Philipp Stanner wrote:
> > drm_sched_init() has a great many parameters and upcoming new
> > functionality for the scheduler might add even more. Generally, the
> > great n
me in
nouveau_sched_init()").
Introduce a new struct for the scheduler init parameters and port all
users.
Signed-off-by: Philipp Stanner
---
Changes in v2:
- Point out that the hang-limit is deprecated. (Christian)
- Initialize the structs to 0 at declaration. (Planet Earth)
- Don't set stuff
reliable, clean scheduler API.
Philipp
Philipp Stanner (5):
drm/sched: Fix teardown leaks with waitqueue
drm/sched: Prevent teardown waitque from blocking too long
drm/sched: Warn if pending list is not empty
drm/nouveau: Add new callback for scheduler teardown
drm/nouveau: Remove wait
callback is not implemented.
Suggested-by: Danilo Krummrich
Signed-off-by: Philipp Stanner
---
drivers/gpu/drm/scheduler/sched_main.c | 47 +-
include/drm/gpu_scheduler.h| 11 ++
2 files changed, 42 insertions(+), 16 deletions(-)
diff --git a/drivers/gpu
From: Philipp Stanner
The GPU scheduler currently does not ensure that its pending_list is
empty before performing various other teardown tasks in
drm_sched_fini().
If there are still jobs in the pending_list, this is problematic because
after scheduler teardown, no one will call
There is a new callback for always tearing the scheduler down in a
leak-free, deadlock-free manner.
Port Nouveau as its first user by providing the scheduler with a
callback that ensures the fence context gets killed in drm_sched_fini().
Signed-off-by: Philipp Stanner
---
drivers/gpu/drm
drm_sched_fini() can leak jobs under certain circumstances.
Warn if that happens.
Signed-off-by: Philipp Stanner
---
drivers/gpu/drm/scheduler/sched_main.c | 4
1 file changed, 4 insertions(+)
diff --git a/drivers/gpu/drm/scheduler/sched_main.c
b/drivers/gpu/drm/scheduler/sched_main.c
nouveau_sched_fence_context_kill() the waitque is not necessary anymore.
Remove the waitque.
Signed-off-by: Philipp Stanner
---
drivers/gpu/drm/nouveau/nouveau_sched.c | 20 +++-
drivers/gpu/drm/nouveau/nouveau_sched.h | 9 +++--
drivers/gpu/drm/nouveau/nouveau_uvmm.c | 8
3 files
fence")
Signed-off-by: Philipp Stanner
---
I'm not entirely sure what Fixes-Tag is appropriate. The last time the
line causing the signalled fence in the list was touched is the commit
listed above.
---
drivers/gpu/drm/nouveau/nouveau_fence.c | 41 -
drivers/
.
Remove the if-branch.
Signed-off-by: Philipp Stanner
---
drivers/gpu/drm/nouveau/nouveau_fence.c | 24 +++-
1 file changed, 11 insertions(+), 13 deletions(-)
diff --git a/drivers/gpu/drm/nouveau/nouveau_fence.c
b/drivers/gpu/drm/nouveau/nouveau_fence.c
index 33535987d8ed
drm_sched_fini() can leak jobs under certain circumstances.
Warn if that happens.
Signed-off-by: Philipp Stanner
---
drivers/gpu/drm/scheduler/sched_main.c | 4
1 file changed, 4 insertions(+)
diff --git a/drivers/gpu/drm/scheduler/sched_main.c
b/drivers/gpu/drm/scheduler/sched_main.c
a bug, or rather: the archetype of a race, since
(as Christian pointed out) nouveau_fence_update() later at some point
will remove the signaled fence (by signaling it again).
P.
Philipp Stanner (3):
drm/nouveau: Prevent signaled fences in pending list
drm/nouveau: Remove surplus if-branch
().
Signed-off-by: Philipp Stanner
---
drivers/gpu/drm/nouveau/nouveau_fence.c | 10 --
1 file changed, 8 insertions(+), 2 deletions(-)
diff --git a/drivers/gpu/drm/nouveau/nouveau_fence.c
b/drivers/gpu/drm/nouveau/nouveau_fence.c
index db6f4494405c..0d58a81b3402 100644
--- a/drivers/gpu/drm
nouveau_fence_base_is_signaled().
Cc: # 4.10+, precise commit not to be determined
Signed-off-by: Philipp Stanner
---
drivers/gpu/drm/nouveau/nouveau_fence.c | 2 +-
1 file changed, 1 insertion(+), 1 deletion(-)
diff --git a/drivers/gpu/drm/nouveau/nouveau_fence.c
b/drivers/gpu/drm/nouveau
There is a new callback for always tearing the scheduler down in a
leak-free, deadlock-free manner.
Port Nouveau as its first user by providing the scheduler with a
callback that ensures the fence context gets killed in drm_sched_fini().
Signed-off-by: Philipp Stanner
---
drivers/gpu/drm
I see two small memory leaks on a Fedora 41 desktop with a custom built
kernel @ commit:
27d4815149ba drm/sched: Group exported prototypes by object type
GPU is an RTX 5000 Ada
The leaks are there immediately after booting the machine. They don't
seem to reoccur, although I have not verified thi
t can provide users with a more
reliable, clean scheduler API.
Philipp
Philipp Stanner (5):
drm/sched: Fix teardown leaks with waitqueue
drm/sched: Prevent teardown waitque from blocking too long
drm/sched: Warn if pending list is not empty
drm/nouveau: Add new callback for scheduler te
y assumed that the TODO is not needed anymore. Besides, its content
is useless anyways since it does not specify *what* should have been
done.
Remove the TODO.
Signed-off-by: Philipp Stanner
---
drivers/gpu/drm/nouveau/nouveau_chan.h | 1 -
1 file changed, 1 deletion(-)
diff --git a/drivers/gpu
at it becomes very, very explicit when reading code that this is a
place where fences can get signaled.
This series obsoletes this patch: [2]
P.
[1] https://lore.kernel.org/all/20250403101353.42880-2-pha...@kernel.org/
[2] https://lore.kernel.org/all/20250408122217.61530-2-pha...@kernel.org/
Ph
nouveau_fence_done() uses the function to check a fence, which causes a
race.
Give the function a more obvious name.
Signed-off-by: Philipp Stanner
---
drivers/dma-buf/dma-fence-array.c | 2 +-
drivers/dma-buf/dma-fence-chain.c | 6 +++---
drivers/dma-buf/dma-
style.
Signed-off-by: Philipp Stanner
---
include/linux/dma-fence.h | 26 --
1 file changed, 20 insertions(+), 6 deletions(-)
diff --git a/include/linux/dma-fence.h b/include/linux/dma-fence.h
index dc2ad171458b..3df370b2cc7c 100644
--- a/include/linux/dma-fence.h
+++ b/in
nouveau_sched_fence_context_kill() the waitque is not necessary anymore.
Remove the waitque.
Signed-off-by: Philipp Stanner
---
drivers/gpu/drm/nouveau/nouveau_sched.c | 20 +++-
drivers/gpu/drm/nouveau/nouveau_sched.h | 9 +++--
drivers/gpu/drm/nouveau/nouveau_uvmm.c | 8
3 files
quot; before.
I've tested this with KASAN & kmemleak.
P.
Philipp Stanner (2):
drm/nouveau: Fix WARN_ON in nouveau_fence_context_kill()
drm/nouveau: nouveau_fence: Standardize list iterations
drivers/gpu/drm/nouveau/nouveau_fence.c | 12 +++-
1 file changed, 7 insertions(+), 5 deletions(-)
--
2.48.1
Fixes: ea13e5abf807 ("drm/nouveau: signal pending fences when channel has been
killed")
Suggested-by: Christian König
Signed-off-by: Philipp Stanner
---
drivers/gpu/drm/nouveau/nouveau_fence.c | 2 +-
1 file changed, 1 insertion(+), 1 deletion(-)
diff --git a/drivers/gpu/drm/nouveau/nou
nouveau_fence.c iterates over lists in a non-canonical way. Since the
operations done are just basic for-each-loops, they should be written in
the standard form.
Use for_each_safe() instead of the custom loop iterations.
Signed-off-by: Philipp Stanner
---
drivers/gpu/drm/nouveau
nouveau_fence_signal() returns a de-facto boolean to indicate when
nvif_event_block() shall be called.
The code can be made more compact and readable by calling
nvif_event_block() in nouveau_fence_update() directly.
Make those calls in nouveau_fence.c more canonical.
Signed-off-by: Philipp
In nouveau_fence_done(), a fence is checked for being signaled by
manually evaluating the base fence's bits. This can be done in a
canonical manner through dma_fence_is_signaled().
Replace the bit-check with dma_fence_is_signaled().
Signed-off-by: Philipp Stanner
---
drivers/gpu/drm/no
/
Philipp Stanner (4):
drm/nouveau: nouveau_fence: Standardize list iterations
drm/nouveau: Simplify calls to nvif_event_block()
drm/nouveau: Simplify nouveau_fence_done()
drm/nouveau: Check dma_fence in canonical way
drivers/gpu/drm/nouveau/nouveau_fence.c | 72 +++--
1 file
nouveau_fence_done() contains an if branch that checks whether a
nouveau_fence has either of the two existing nouveau_fence backend ops,
which will always evaluate to true.
Remove the surplus check.
Signed-off-by: Philipp Stanner
---
drivers/gpu/drm/nouveau/nouveau_fence.c | 24
nouveau_fence.c iterates over lists in a non-canonical way. Since the
operations done are just basic for-each-loops and list-empty checks,
they should be written in the standard form.
Use standard list operations.
Signed-off-by: Philipp Stanner
---
drivers/gpu/drm/nouveau/nouveau_fence.c | 21
callback is not implemented.
Suggested-by: Danilo Krummrich
Signed-off-by: Philipp Stanner
---
drivers/gpu/drm/scheduler/sched_main.c | 47 +-
include/drm/gpu_scheduler.h| 11 ++
2 files changed, 42 insertions(+), 16 deletions(-)
diff --git a/drivers/gpu
On Mon, 2025-04-07 at 17:22 +0200, Philipp Stanner wrote:
> From: Philipp Stanner
>
> The GPU scheduler currently does not ensure that its pending_list is
> empty before performing various other teardown tasks in
> drm_sched_fini().
>
> If there are still jobs in the
the unit tests. Remove the manual cleanup
code.
Signed-off-by: Philipp Stanner
---
.../gpu/drm/scheduler/tests/mock_scheduler.c | 34 ---
1 file changed, 21 insertions(+), 13 deletions(-)
diff --git a/drivers/gpu/drm/scheduler/tests/mock_scheduler.c
b/drivers/gpu/drm/scheduler
d, therefore, must be a relict forgotten in a previous cleanup.
Remove the TODO and accel_done.
Signed-off-by: Philipp Stanner
---
Changes in v2:
- Remove accel_done, too. (Danilo)
---
drivers/gpu/drm/nouveau/nouveau_chan.h | 2 --
drivers/gpu/drm/nouveau/nouveau_dma.h | 1 -
2 files changed
callback is not implemented.
Suggested-by: Danilo Krummrich
Signed-off-by: Philipp Stanner
---
drivers/gpu/drm/scheduler/sched_main.c | 47 +-
include/drm/gpu_scheduler.h| 11 ++
2 files changed, 42 insertions(+), 16 deletions(-)
diff --git a/drivers/gpu
There is a new callback for always tearing the scheduler down in a
leak-free, deadlock-free manner.
Port Nouveau as its first user by providing the scheduler with a
callback that ensures the fence context gets killed in drm_sched_fini().
Signed-off-by: Philipp Stanner
---
drivers/gpu/drm
nouveau_sched_fence_context_kill() the waitque is not necessary anymore.
Remove the waitque.
Signed-off-by: Philipp Stanner
---
drivers/gpu/drm/nouveau/nouveau_sched.c | 20 +++-
drivers/gpu/drm/nouveau/nouveau_sched.h | 9 +++--
drivers/gpu/drm/nouveau/nouveau_uvmm.c | 8
3 files
ks fine and
solves the problem (though we did discover an unrelated problem inside
Nouveau in the process).
It also works with the unit tests.
I'm looking forward to your input and feedback. I really hope we can
work this RFC into something that can provide users with a more
reliable, clean
From: Philipp Stanner
The GPU scheduler currently does not ensure that its pending_list is
empty before performing various other teardown tasks in
drm_sched_fini().
If there are still jobs in the pending_list, this is problematic because
after scheduler teardown, no one will call
drm_sched_fini() can leak jobs under certain circumstances.
Warn if that happens.
Signed-off-by: Philipp Stanner
---
drivers/gpu/drm/scheduler/sched_main.c | 3 +++
1 file changed, 3 insertions(+)
diff --git a/drivers/gpu/drm/scheduler/sched_main.c
b/drivers/gpu/drm/scheduler/sched_main.c
From: Philipp Stanner
The GPU scheduler currently does not ensure that its pending_list is
empty before performing various other teardown tasks in
drm_sched_fini().
If there are still jobs in the pending_list, this is problematic because
after scheduler teardown, no one will call
On Fri, 2025-03-21 at 09:56 +0100, Philipp Stanner wrote:
> I see two small memory leaks on a Fedora 41 desktop with a custom
> built
> kernel @ commit:
>
> 27d4815149ba drm/sched: Group exported prototypes by object type
>
> GPU is an RTX 5000 Ada
>
> The leaks
signalling a fence has additional effects is to add those
effects to a callback and register it on that fence.
Move the code from nouveau_fence_signal() into a dma_fence callback.
Register that callback when creating the fence.
Cc: # 4.10+
Signed-off-by: Philipp Stanner
---
Changes in v2:
- Remove
On Thu, 2025-05-22 at 14:37 +0100, Tvrtko Ursulin wrote:
>
> On 22/05/2025 09:27, Philipp Stanner wrote:
> > From: Philipp Stanner
> >
> > The GPU scheduler currently does not ensure that its pending_list
> > is
> > empty before performing various other
ovide users with a more
reliable, clean scheduler API.
Philipp
Philipp Stanner (5):
drm/sched: Fix teardown leaks with waitqueue
drm/sched/tests: Port tests to new cleanup method
drm/sched: Warn if pending list is not empty
drm/nouveau: Add new callback for scheduler teardown
drm/nouveau: Remove
From: Philipp Stanner
The GPU scheduler currently does not ensure that its pending_list is
empty before performing various other teardown tasks in
drm_sched_fini().
If there are still jobs in the pending_list, this is problematic because
after scheduler teardown, no one will call
a new error
field for the fence error.
Keep the job status as DRM_MOCK_SCHED_JOB_DONE for now, since there is
no party for which checking for a CANCELED status would be useful
currently.
Signed-off-by: Philipp Stanner
---
.../gpu/drm/scheduler/tests/mock_scheduler.c | 67
drm_sched_fini() can leak jobs under certain circumstances.
Warn if that happens.
Signed-off-by: Philipp Stanner
---
drivers/gpu/drm/scheduler/sched_main.c | 3 +++
1 file changed, 3 insertions(+)
diff --git a/drivers/gpu/drm/scheduler/sched_main.c
b/drivers/gpu/drm/scheduler/sched_main.c
nouveau_sched_fence_context_kill() the waitque is not necessary anymore.
Remove the waitque.
Signed-off-by: Philipp Stanner
---
drivers/gpu/drm/nouveau/nouveau_sched.c | 20 +++-
drivers/gpu/drm/nouveau/nouveau_sched.h | 9 +++--
drivers/gpu/drm/nouveau/nouveau_uvmm.c | 8
3 files
There is a new callback for always tearing the scheduler down in a
leak-free, deadlock-free manner.
Port Nouveau as its first user by providing the scheduler with a
callback that ensures the fence context gets killed in drm_sched_fini().
Signed-off-by: Philipp Stanner
---
drivers/gpu/drm
ed. Use it internally.
Suggested-by: Tvrtko Ursulin
Signed-off-by: Philipp Stanner
---
include/linux/dma-fence.h | 24 ++--
1 file changed, 22 insertions(+), 2 deletions(-)
diff --git a/include/linux/dma-fence.h b/include/linux/dma-fence.h
index 48b5202c531d..ac951a54a007 10
which only checks, never signals.
Signed-off-by: Philipp Stanner
---
drivers/gpu/drm/nouveau/nouveau_fence.c | 2 +-
1 file changed, 1 insertion(+), 1 deletion(-)
diff --git a/drivers/gpu/drm/nouveau/nouveau_fence.c
b/drivers/gpu/drm/nouveau/nouveau_fence.c
index d5654e26d5bc..993b3dcb5db0
hardware fence.
That should be repaired and cleaned up, but it's probably better to do
that in a separate series.
Signed-off-by: Philipp Stanner
---
.../gpu/drm/scheduler/tests/mock_scheduler.c | 71 +++
drivers/gpu/drm/scheduler/tests/sched_tests.h | 4 +-
2 files change
On Tue, 2025-06-17 at 14:00 +1000, Ben Skeggs wrote:
> If any of the ACPI calls fail, memory allocated for the input buffer
> would be leaked. Fix failure paths to free allocated memory.
>
> Also add checks to ensure the allocations succeeded in the first
> place.
>
> Reported-by: Danilo Krummri
the hardware fence associated with the
job. Afterwards, the scheduler can savely use the established free_job()
callback for freeing the job.
Implement the new backend_ops callback cancel_job().
Suggested-by: Tvrtko Ursulin
Signed-off-by: Philipp Stanner
---
drivers/gpu/drm/scheduler
drm_sched_fini() can leak jobs under certain circumstances.
Warn if that happens.
Signed-off-by: Philipp Stanner
---
drivers/gpu/drm/scheduler/sched_main.c | 3 +++
1 file changed, 3 insertions(+)
diff --git a/drivers/gpu/drm/scheduler/sched_main.c
b/drivers/gpu/drm/scheduler/sched_main.c
There is a new callback for always tearing the scheduler down in a
leak-free, deadlock-free manner.
Port Nouveau as its first user by providing the scheduler with a
callback that ensures the fence context gets killed in drm_sched_fini().
Signed-off-by: Philipp Stanner
---
drivers/gpu/drm
ps://lore.kernel.org/dri-devel/20250418113211.69956-1-tvrtko.ursu...@igalia.com/
Philipp Stanner (6):
drm/sched: Avoid memory leaks with cancel_job() callback
drm/sched/tests: Implement cancel_job()
drm/sched: Warn if pending list is not empty
drm/nouveau: Make fence container helper usable driver-wide
: Philipp Stanner
---
drivers/gpu/drm/nouveau/nouveau_fence.c | 20 +++-
drivers/gpu/drm/nouveau/nouveau_fence.h | 6 ++
2 files changed, 13 insertions(+), 13 deletions(-)
diff --git a/drivers/gpu/drm/nouveau/nouveau_fence.c
b/drivers/gpu/drm/nouveau/nouveau_fence.c
index
nouveau_sched_fence_context_kill() the waitque is not necessary anymore.
Remove the waitque.
Signed-off-by: Philipp Stanner
---
drivers/gpu/drm/nouveau/nouveau_sched.c | 20 +++-
drivers/gpu/drm/nouveau/nouveau_sched.h | 9 +++--
drivers/gpu/drm/nouveau/nouveau_uvmm.c | 8
3 files
are still in
drm_sched.pending_list.
This series solves the leaks in a backwards-compatible manner by adding
a new, optional callback. If that callback is implemented, the scheduler
uses it to cancel all jobs from pending_list and then frees them.
Philipp Stanner (7):
drm/sched: Avoid memory
-tvrtko.ursu...@igalia.com/
Signed-off-by: Philipp Stanner
Reviewed-by: Maíra Canal
---
drivers/gpu/drm/scheduler/sched_main.c | 34 --
include/drm/gpu_scheduler.h| 18 ++
2 files changed, 39 insertions(+), 13 deletions(-)
diff --git a/drivers/gpu/drm
: Philipp Stanner
---
drivers/gpu/drm/nouveau/nouveau_fence.c | 20 +++-
drivers/gpu/drm/nouveau/nouveau_fence.h | 6 ++
2 files changed, 13 insertions(+), 13 deletions(-)
diff --git a/drivers/gpu/drm/nouveau/nouveau_fence.c
b/drivers/gpu/drm/nouveau/nouveau_fence.c
index
the code where necessary.
Signed-off-by: Philipp Stanner
---
.../gpu/drm/scheduler/tests/mock_scheduler.c | 66 +++
1 file changed, 23 insertions(+), 43 deletions(-)
diff --git a/drivers/gpu/drm/scheduler/tests/mock_scheduler.c
b/drivers/gpu/drm/scheduler/tests/mock_scheduler.c
The scheduler unit tests now provide a new callback, cancel_job(). This
callback gets used by drm_sched_fini() for all still pending jobs to
cancel them.
Implement a new unit test to test this.
Signed-off-by: Philipp Stanner
---
drivers/gpu/drm/scheduler/tests/tests_basic.c | 43
There is a new callback for always tearing the scheduler down in a
leak-free, deadlock-free manner.
Port Nouveau as its first user by providing the scheduler with a
callback that ensures the fence context gets killed in drm_sched_fini().
Signed-off-by: Philipp Stanner
---
drivers/gpu/drm
nouveau_sched_fence_context_kill() the waitque is not necessary anymore.
Remove the waitque.
Signed-off-by: Philipp Stanner
---
drivers/gpu/drm/nouveau/nouveau_sched.c | 20 +++-
drivers/gpu/drm/nouveau/nouveau_sched.h | 9 +++--
drivers/gpu/drm/nouveau/nouveau_uvmm.c | 8
3 files
drm_sched_fini() can leak jobs under certain circumstances.
Warn if that happens.
Signed-off-by: Philipp Stanner
---
drivers/gpu/drm/scheduler/sched_main.c | 3 +++
1 file changed, 3 insertions(+)
diff --git a/drivers/gpu/drm/scheduler/sched_main.c
b/drivers/gpu/drm/scheduler/sched_main.c
On Tue, 2025-06-17 at 14:00 +1000, Ben Skeggs wrote:
> If any of the ACPI calls fail, memory allocated for the input buffer
> would be leaked. Fix failure paths to free allocated memory.
>
> Also add checks to ensure the allocations succeeded in the first
> place.
If you've got a kmemleak trace,
On Mon, 2025-07-07 at 16:31 +0200, Danilo Krummrich wrote:
> On 7/7/25 10:27 AM, Philipp Stanner wrote:
> > On Tue, 2025-06-17 at 14:00 +1000, Ben Skeggs wrote:
> > > If any of the ACPI calls fail, memory allocated for the input buffer
> > > would be leaked. Fix fail
The scheduler unit tests now provide a new callback, cancel_job(). This
callback gets used by drm_sched_fini() for all still pending jobs to
cancel them.
Implement a new unit test to test this.
Signed-off-by: Philipp Stanner
Reviewed-by: Tvrtko Ursulin
---
drivers/gpu/drm/scheduler/tests
-tvrtko.ursu...@igalia.com/
Signed-off-by: Philipp Stanner
Reviewed-by: Maíra Canal
---
drivers/gpu/drm/scheduler/sched_main.c | 34 --
include/drm/gpu_scheduler.h| 18 ++
2 files changed, 39 insertions(+), 13 deletions(-)
diff --git a/drivers/gpu/drm
the code where necessary.
Signed-off-by: Philipp Stanner
---
.../gpu/drm/scheduler/tests/mock_scheduler.c | 68 +++
drivers/gpu/drm/scheduler/tests/sched_tests.h | 1 -
2 files changed, 25 insertions(+), 44 deletions(-)
diff --git a/drivers/gpu/drm/scheduler/tests
drm_sched_fini() can leak jobs under certain circumstances.
Warn if that happens.
Signed-off-by: Philipp Stanner
---
drivers/gpu/drm/scheduler/sched_main.c | 3 +++
1 file changed, 3 insertions(+)
diff --git a/drivers/gpu/drm/scheduler/sched_main.c
b/drivers/gpu/drm/scheduler/sched_main.c
There is a new callback for always tearing the scheduler down in a
leak-free, deadlock-free manner.
Port Nouveau as its first user by providing the scheduler with a
callback that ensures the fence context gets killed in drm_sched_fini().
Signed-off-by: Philipp Stanner
Acked-by: Danilo Krummrich
manner by adding
a new, optional callback. If that callback is implemented, the scheduler
uses it to cancel all jobs from pending_list and then frees them.
Philipp Stanner (7):
drm/sched: Avoid memory leaks with cancel_job() callback
drm/sched/tests: Implement cancel_job() callback
drm/sched/
: Philipp Stanner
Acked-by: Danilo Krummrich
---
drivers/gpu/drm/nouveau/nouveau_fence.c | 20 +++-
drivers/gpu/drm/nouveau/nouveau_fence.h | 6 ++
2 files changed, 13 insertions(+), 13 deletions(-)
diff --git a/drivers/gpu/drm/nouveau/nouveau_fence.c
b/drivers/gpu/drm
nouveau_sched_cancel_job()
the waitque is not necessary anymore.
Remove the waitque.
Signed-off-by: Philipp Stanner
Acked-by: Danilo Krummrich
---
drivers/gpu/drm/nouveau/nouveau_sched.c | 20 +++-
drivers/gpu/drm/nouveau/nouveau_sched.h | 9 +++--
drivers/gpu/drm/nouveau/nouveau_uvmm.c
-4d55-aa47-c35cd7861...@igalia.com/
Signed-off-by: Philipp Stanner
---
drivers/gpu/drm/panfrost/panfrost_job.c | 2 +-
1 file changed, 1 insertion(+), 1 deletion(-)
diff --git a/drivers/gpu/drm/panfrost/panfrost_job.c
b/drivers/gpu/drm/panfrost/panfrost_job.c
index 5657106c2f7d..15e2d505550f 10
-tvrtko.ursu...@igalia.com/
Signed-off-by: Philipp Stanner
Reviewed-by: Maíra Canal
---
drivers/gpu/drm/scheduler/sched_main.c | 34 --
include/drm/gpu_scheduler.h| 18 ++
2 files changed, 39 insertions(+), 13 deletions(-)
diff --git a/drivers/gpu/drm
the code where necessary.
Signed-off-by: Philipp Stanner
---
.../gpu/drm/scheduler/tests/mock_scheduler.c | 68 +++
drivers/gpu/drm/scheduler/tests/sched_tests.h | 1 -
2 files changed, 25 insertions(+), 44 deletions(-)
diff --git a/drivers/gpu/drm/scheduler/tests
1 - 100 of 117 matches
Mail list logo
