Skip to content

RFC drm/sched: Avoid memory leaks by canceling job-by-job

Philipp Stanner dri-devel linux-media lkml nouveau 1 version In review

Get series with b4
b4 am [email protected]

Versions newest first

  1. v1 RFC Latest6 patches 9 repliesIn review

Patches in v1 No applies recorded yet

#Subject and filesSizeRepliesApplied
1/6drm/sched: Avoid memory leaks with cancel_job() callbackdrivers/gpu/drm/scheduler/sched_main.c include/drm/gpu_scheduler.h+30 -135 replies
2/6drm/sched/tests: Implement cancel_job()drivers/gpu/drm/scheduler/tests/mock_scheduler.c drivers/gpu/drm/scheduler/tests/sched_tests.h+25 -500 replies
3/6drm/sched: Warn if pending list is not emptydrivers/gpu/drm/scheduler/sched_main.c+3 -00 replies
4/6drm/nouveau: Make fence container helper usable driver-widedrivers/gpu/drm/nouveau/nouveau_fence.c drivers/gpu/drm/nouveau/nouveau_fence.h+13 -130 replies
5/6drm/nouveau: Add new callback for scheduler teardowndrivers/gpu/drm/nouveau/nouveau_fence.c drivers/gpu/drm/nouveau/nouveau_fence.h drivers/gpu/drm/nouveau/nouveau_sched.c+30 -10 replies
6/6drm/nouveau: Remove waitque for sched teardowndrivers/gpu/drm/nouveau/nouveau_sched.c drivers/gpu/drm/nouveau/nouveau_sched.h drivers/gpu/drm/nouveau/nouveau_uvmm.c+14 -230 replies

Reviews trailers per patch

1 of 6 patches has a Reviewed-by or Acked-by.

PatchTvrtko Ursulin
1/6A-b
2/6
3/6
4/6
5/6
6/6

Trailers given in replies are filled; those carried in the patches from earlier versions are outlined with dashes.

Read the v1 thread

Cover letter [RFC PATCH 0/6] drm/sched: Avoid memory leaks by canceling job-by-job

An alternative version to [1], based on Tvrtko's suggestion from [2].

I tested this for Nouveau. Works.

I'm having, however, bigger problems properly porting the unit tests and have seen various explosions. In the process I noticed that some things in the unit tests aren't right and a bit of a larger rework will be necessary (for example, the timedout job callback must signal the timedout fence, remove it from the list and so on).

Anyways. Please comment on the general idea.

@Tvrtko: As briefly brainstormed about on IRC, if you'd be willing to take care of the unit tests patch, I could remove that one (and, maaaaybe, the warning print patch) from the series and we could merge this RFC's successor version %N once it's ready. What do you think?

P.

[1] https://lore.kernel.org/dri-devel/[email protected]/
[2] https://lore.kernel.org/dri-devel/[email protected]/

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 drm/nouveau: Add new callback for scheduler teardown drm/nouveau: Remove waitque for sched teardown

 drivers/gpu/drm/nouveau/nouveau_fence.c       | 35 +++++----
 drivers/gpu/drm/nouveau/nouveau_fence.h       |  7 ++
 drivers/gpu/drm/nouveau/nouveau_sched.c       | 35 +++++----
 drivers/gpu/drm/nouveau/nouveau_sched.h       |  9 +--
 drivers/gpu/drm/nouveau/nouveau_uvmm.c        |  8 +--
 drivers/gpu/drm/scheduler/sched_main.c        | 37 ++++++----
 .../gpu/drm/scheduler/tests/mock_scheduler.c  | 71 +++++++------------
 drivers/gpu/drm/scheduler/tests/sched_tests.h |  4 +-
 include/drm/gpu_scheduler.h                   |  9 +++
 9 files changed, 115 insertions(+), 100 deletions(-)
Signature · 1 line
2.49.0