[PATCH] drm/scheduler: Fix mem leak when last_scheduled signaled

Qianyi Liu liuqianyi125 at gmail.com
Tue Feb 25 02:58:57 UTC 2025


Thank you for your patient reply. Let's first clarify the issue and send a new
patch if necessary.

As soon as it enters the drm_sched_entity_kill function, the entity
->last_stchedled reference count is incremented by 1. If there are still jobs in
the current entity, it will enter the while loop, assuming there is only one job
left. If entity ->last_stcheduled has already been signaled, it will enter
drm_sched_detity_kill_jobs_cb, but because null is passed in, the
last_stcheduled reference count will not be correctly reduced by 1.

Because the prev pointer has been updated to &s_sense ->defined, the
dma_ffence_put in the last line only reduces the reference count of s_fence
->finished. The reference count of entity ->last_stcheduled was not reduced by
1, causing a memory leak.

We should subtract 1 from the reference count of the prev when dma_ fence_ add_
callback fails, which is called balance.

Best Regards.
QianYi.


More information about the dri-devel mailing list