FAILED: patch "[PATCH] drm/v3d: Fix use-after-free of CPU job query arrays on error" failed to apply to 6.12-stable tree
Subsystems (4)
[email protected]
maintainerThe patch below does not apply to the 6.12-stable tree.
If someone wants it applied there, or to any other stable or longterm tree, then please email the backport, including the original git commit id to <[email protected]>.
To reproduce the conflict and resubmit, you may use the following commands:
git fetch https://git.kernel.org/pub/scm/linux/kernel/git/stable/linux.git/ linux-6.12.y
git checkout FETCH_HEAD
git cherry-pick -x b0fe80c0b9250b35e2211bf3117e7aca814a21b0
# <resolve conflicts, build, test, etc.>
git commit -s
git send-email --to '<[email protected]>' --in-reply-to '2026052855-portable-pardon-850e@gregkh' --subject-prefix 'PATCH 6.12.y' HEAD^..
Possible dependencies:
thanks,
greg k-h
------------------ original commit in Linus's tree ------------------
From b0fe80c0b9250b35e2211bf3117e7aca814a21b0 Mon Sep 17 00:00:00 2001
From: =?UTF-8?q?Ma=C3=ADra=20Canal?= <[email protected]>
Date: Fri, 15 May 2026 12:07:14 -0300
Subject: [PATCH] drm/v3d: Fix use-after-free of CPU job query arrays on error
path
MIME-Version: 1.0
Content-Type: text/plain; charset=UTF-8
Content-Transfer-Encoding: 8bit
The CPU job ioctl's fail label calls kvfree() on cpu_job's timestamp and performance query arrays after v3d_job_cleanup(), which drops the job's last reference and frees cpu_job. Reading cpu_job at that point is a use-after-free. Also, on the early v3d_job_init() failure path, it is a NULL dereference, since v3d_job_deallocate() zeroes the local pointer.
In the success path, the arrays are released from the scheduler's .free_job callback, but on the error path, they are freed manually, as the job was never pushed to the scheduler. While the success path deals with this correctly, the fail path doesn't.
On top of that, the manual kvfree() calls only free the array storage; they don't drm_syncobj_put() the per-query syncobjs that v3d_timestamp_query_info_free() and v3d_performance_query_info_free() release on the success path. So the same fail path that triggers the use-after-free also leaks one syncobj reference per query.
Unify the CPU job teardown into the CPU job's kref destructor, mirroring v3d_render_job_free(). The scheduler's .free_job slot reverts to the generic v3d_sched_job_free() and the fail label drops the manual kvfree() calls, leaving a single teardown path that is reached from both the scheduler and the ioctl error path. That removes the use-after-free, the NULL dereference, and the syncobj leak by construction.
drivers/gpu/drm/v3d/v3d_sched.c +1 -15
@@ -125,20 +125,6 @@ v3d_performance_query_info_free(struct v3d_performance_query_info *query_info, } } -static void -v3d_cpu_job_free (struct drm_sched_job *sched_job) -{ -struct v3d_cpu_job *job =to_cpu_job (sched_job); - -v3d_timestamp_query_info_free (&job->timestamp_query, - job->timestamp_query.count); - -v3d_performance_query_info_free (&job->performance_query, - job->performance_query.count); - -v3d_job_cleanup (&job->base); -} -static void v3d_switch_perfmon (struct v3d_dev *v3d,struct v3d_job *job) {
@@ -830,7 +816,7 @@ static const struct drm_sched_backend_ops v3d_cache_clean_sched_ops = {static const struct drm_sched_backend_ops v3d_cpu_sched_ops = { .run_job = v3d_cpu_job_run, - .free_job = v3d_cpu_job_free + .free_job = v3d_sched_job_free }; static int
drivers/gpu/drm/v3d/v3d_submit.c +16 -3
@@ -123,6 +123,21 @@ v3d_render_job_free(struct kref *ref)v3d_job_free (ref); } +static void +v3d_cpu_job_free (struct kref *ref) +{ + struct v3d_cpu_job *job = container_of(ref, struct v3d_cpu_job, + base.refcount); + +v3d_timestamp_query_info_free (&job->timestamp_query, + job->timestamp_query.count); + +v3d_performance_query_info_free (&job->performance_query, + job->performance_query.count); + +v3d_job_free (ref); +} +void v3d_job_cleanup (struct v3d_job *job) {if (!job)
@@ -1302,7 +1317,7 @@ v3d_submit_cpu_ioctl(struct drm_device *dev, void *data,trace_v3d_submit_cpu_ioctl (&v3d->drm, cpu_job->job_type); ret =v3d_job_init (v3d, file_priv, &cpu_job->base, - v3d_job_free,0 , &se,V3D_CPU ); + v3d_cpu_job_free,0 , &se,V3D_CPU );if (ret) {v3d_job_deallocate ((void *)&cpu_job); goto fail;
@@ -1385,8 +1400,6 @@ v3d_submit_cpu_ioctl(struct drm_device *dev, void *data,v3d_job_cleanup ((void *)csd_job);v3d_job_cleanup (clean_job);v3d_put_multisync_post_deps (&se); -kvfree (cpu_job->timestamp_query.queries); -kvfree (cpu_job->performance_query.queries);return ret; }
Sasha Levin
maintainerPATCH 1/2 drm/v3d: Synchronous operations can't timeout
From: Maíra Canal <[email protected]>
[ Upstream commit 78fe02d090d38393cc00de9d1929e59426f202a4 ]
CPU jobs and CACHE CLEAN jobs execute synchronously once the DRM scheduler starts running them. Therefore, there is no fence to wait on, neither are those jobs able to timeout.
Hence, remove the timedout_job hook from the CPU and CACHE CLEAN scheduler ops.
Stable-dep-of: b0fe80c0b925 ("drm/v3d: Fix use-after-free of CPU job query arrays on error path")
--- drivers/gpu/drm/v3d/v3d_sched.c | 8 ++++---- 1 file changed, 4 insertions(+), 4 deletions(-)
drivers/gpu/drm/v3d/v3d_sched.c +4 -4
@@ -684,6 +684,7 @@ v3d_cpu_job_run(struct drm_sched_job *sched_job)trace_v3d_cpu_job_end (&v3d->drm, job->job_type);v3d_job_update_stats (&job->base,V3D_CPU ); +/* Synchronous operation, so no fence to wait on. */ return NULL ; }
@@ -699,6 +700,7 @@ v3d_cache_clean_job_run(struct drm_sched_job *sched_job)v3d_job_update_stats (job,V3D_CACHE_CLEAN ); +/* Synchronous operation, so no fence to wait on. */ return NULL ; }
@@ -786,7 +788,7 @@ v3d_render_job_timedout(struct drm_sched_job *sched_job) }static enum drm_gpu_sched_stat -v3d_generic_job_timedout (struct drm_sched_job *sched_job) +v3d_tfu_job_timedout (struct drm_sched_job *sched_job) {struct v3d_job *job =to_v3d_job (sched_job);
@@ -827,7 +829,7 @@ static const struct drm_sched_backend_ops v3d_render_sched_ops = {static const struct drm_sched_backend_ops v3d_tfu_sched_ops = { .run_job = v3d_tfu_job_run, - .timedout_job = v3d_generic_job_timedout, + .timedout_job = v3d_tfu_job_timedout, .free_job = v3d_sched_job_free, };
@@ -839,13 +841,11 @@ static const struct drm_sched_backend_ops v3d_csd_sched_ops = {static const struct drm_sched_backend_ops v3d_cache_clean_sched_ops = { .run_job = v3d_cache_clean_job_run, - .timedout_job = v3d_generic_job_timedout, .free_job = v3d_sched_job_free };static const struct drm_sched_backend_ops v3d_cpu_sched_ops = { .run_job = v3d_cpu_job_run, - .timedout_job = v3d_generic_job_timedout, .free_job = v3d_cpu_job_free };
Signature · 1 line
2.53.0
Sasha Levin
maintainerPATCH 2/2 drm/v3d: Fix use-after-free of CPU job query arrays on error path
From: Maíra Canal <[email protected]>
[ Upstream commit b0fe80c0b9250b35e2211bf3117e7aca814a21b0 ]
The CPU job ioctl's fail label calls kvfree() on cpu_job's timestamp and performance query arrays after v3d_job_cleanup(), which drops the job's last reference and frees cpu_job. Reading cpu_job at that point is a use-after-free. Also, on the early v3d_job_init() failure path, it is a NULL dereference, since v3d_job_deallocate() zeroes the local pointer.
In the success path, the arrays are released from the scheduler's .free_job callback, but on the error path, they are freed manually, as the job was never pushed to the scheduler. While the success path deals with this correctly, the fail path doesn't.
On top of that, the manual kvfree() calls only free the array storage; they don't drm_syncobj_put() the per-query syncobjs that v3d_timestamp_query_info_free() and v3d_performance_query_info_free() release on the success path. So the same fail path that triggers the use-after-free also leaks one syncobj reference per query.
Unify the CPU job teardown into the CPU job's kref destructor, mirroring v3d_render_job_free(). The scheduler's .free_job slot reverts to the generic v3d_sched_job_free() and the fail label drops the manual kvfree() calls, leaving a single teardown path that is reached from both the scheduler and the ioctl error path. That removes the use-after-free, the NULL dereference, and the syncobj leak by construction.
--- drivers/gpu/drm/v3d/v3d_sched.c | 16 +--------------- drivers/gpu/drm/v3d/v3d_submit.c | 19 ++++++++++++++++--- 2 files changed, 17 insertions(+), 18 deletions(-)
drivers/gpu/drm/v3d/v3d_sched.c +1 -15
@@ -103,20 +103,6 @@ v3d_performance_query_info_free(struct v3d_performance_query_info *query_info, } } -static void -v3d_cpu_job_free (struct drm_sched_job *sched_job) -{ -struct v3d_cpu_job *job =to_cpu_job (sched_job); - -v3d_timestamp_query_info_free (&job->timestamp_query, - job->timestamp_query.count); - -v3d_performance_query_info_free (&job->performance_query, - job->performance_query.count); - -v3d_job_cleanup (&job->base); -} -static void v3d_switch_perfmon (struct v3d_dev *v3d,struct v3d_job *job) {
@@ -846,7 +832,7 @@ static const struct drm_sched_backend_ops v3d_cache_clean_sched_ops = {static const struct drm_sched_backend_ops v3d_cpu_sched_ops = { .run_job = v3d_cpu_job_run, - .free_job = v3d_cpu_job_free + .free_job = v3d_sched_job_free };int
drivers/gpu/drm/v3d/v3d_submit.c +16 -3
@@ -118,6 +118,21 @@ v3d_render_job_free(struct kref *ref)v3d_job_free (ref); } +static void +v3d_cpu_job_free (struct kref *ref) +{ + struct v3d_cpu_job *job = container_of(ref, struct v3d_cpu_job, + base.refcount); + +v3d_timestamp_query_info_free (&job->timestamp_query, + job->timestamp_query.count); + +v3d_performance_query_info_free (&job->performance_query, + job->performance_query.count); + +v3d_job_free (ref); +} +void v3d_job_cleanup (struct v3d_job *job) {if (!job)
@@ -1310,7 +1325,7 @@ v3d_submit_cpu_ioctl(struct drm_device *dev, void *data,trace_v3d_submit_cpu_ioctl (&v3d->drm, cpu_job->job_type); ret =v3d_job_init (v3d, file_priv, &cpu_job->base, - v3d_job_free,0 , &se,V3D_CPU ); + v3d_cpu_job_free,0 , &se,V3D_CPU );if (ret) {v3d_job_deallocate ((void *)&cpu_job); goto fail;
@@ -1393,8 +1408,6 @@ v3d_submit_cpu_ioctl(struct drm_device *dev, void *data,v3d_job_cleanup ((void *)csd_job);v3d_job_cleanup (clean_job);v3d_put_multisync_post_deps (&se); -kvfree (cpu_job->timestamp_query.queries); -kvfree (cpu_job->performance_query.queries);return ret; }
Signature · 1 line
2.53.0