drm/xe: keep process name fallback for diagnostics
Subsystems (2)
PATCH 1/1 drm/xe: keep process name fallback for diagnostics
xe_file_open() stores task->comm in xef->process_name with kstrdup(), but this diagnostic string is optional and the file open continues if the allocation fails.
The timeout and devcoredump paths initialize process_name to the safe "no process" fallback, then overwrite it whenever q->vm->xef exists. If the kstrdup() failed, the fallback is replaced with NULL and the NULL pointer is passed to string handling in diagnostic paths.
Keep the fallback process name unless xef->process_name was actually allocated. Still report the stored pid when the xe file is available.
--- drivers/gpu/drm/xe/xe_devcoredump.c | 3 ++- drivers/gpu/drm/xe/xe_guc_submit.c | 3 ++- 2 files changed, 4 insertions(+), 2 deletions(-)
drivers/gpu/drm/xe/xe_devcoredump.c +2 -1
@@ -343,7 +343,8 @@ static void devcoredump_snapshot(struct xe_devcoredump *coredump, ss->boot_time =ktime_get_boottime ();if (q->vm && q->vm->xef) { - process_name = q->vm->xef->process_name; +if (q->vm->xef->process_name) + process_name = q->vm->xef->process_name; ss->pid = q->vm->xef->pid; }
drivers/gpu/drm/xe/xe_guc_submit.c +2 -1
@@ -1189,7 +1189,8 @@ trigger_reset: }if (q->vm && q->vm->xef) { - process_name = q->vm->xef->process_name; +if (q->vm->xef->process_name) + process_name = q->vm->xef->process_name; pid = q->vm->xef->pid; }
Signature · 1 line
2.51.0
[email protected]
AI reviewThank you for your contribution! Sashiko AI review found 1 potential issue(s) to consider:
Thank you for your contribution! Sashiko AI review found 1 potential issue(s) to consider:
Pre-existing issues:
- [High] Lockless access to task->comm via kstrdup() creates a TOCTOU race leading to a non-null-terminated string and an out-of-bounds read. --
commit f329e0d91af948bfcb07b90ac00f44ede0a2f8fc
Author: Ruoyu Wang <[email protected]>
drm/xe: keep process name fallback for diagnostics
This commit addresses an issue where diagnostic paths would use a NULL pointer for a process name if allocation failed during file open. It keeps the "no process" fallback unless the process name was successfully allocated.
[ ... ]
drivers/gpu/drm/xe/xe_guc_submit.c guc_exec_queue_timedout_job(struct drm_sched_job *drm_job) · line 1596
Quoted 13 lines
drivers/gpu/drm/xe/xe_guc_submit.c +2
-1@@ -1590,7 +1590,8 @@ guc_exec_queue_timedout_job(struct drm_sched_job *drm_job) }if (q->vm && q->vm->xef) { - process_name = q->vm->xef->process_name; +if (q->vm->xef->process_name) + process_name = q->vm->xef->process_name; pid = q->vm->xef->pid; }
[Severity: High]
This is a pre-existing issue, but does using this string in formatting functions risk an out-of-bounds memory read?
Looking at xe_file_open() in drivers/gpu/drm/xe/xe_device.c, task->comm is directly passed to kstrdup() without holding task_lock(task):
xef->process_name = kstrdup(task->comm, GFP_KERNEL);
Since task->comm can be modified concurrently by userspace via prctl(PR_SET_NAME), kstrdup() can compute a short length during its internal strlen() call, but copy characters from a longer replacement string during memcpy(). This results in a heap buffer missing a null terminator.
When process_name is subsequently passed to xe_gt_notice() further down in this function and printed via the %s format specifier:
xe_gt_notice(guc_to_gt(guc),
"Timedout job: seqno=%u, lrc_seqno=%u, guc_id=%d, flags=0x%lx in %s [%d]",
xe_sched_job_seqno(job), xe_sched_job_lrc_seqno(job),
q->guc->id, q->flags, process_name, pid);Can this trigger an out-of-bounds read of adjacent slab memory, potentially leading to KASAN panics or kernel memory disclosure?
Could the string copying in xe_file_open() be safely changed to use get_task_comm() instead?
Signature · 1 line
Sashiko AI review · https://sashiko.dev/#/patchset/[email protected]?part=1
Patchwork
CI✗ LGCI.VerificationFailed: failure for drm/xe: keep process name fallback for diagnostics
In reply to Ruoyu Wang== Series Details ==
== Series Details ==
Series: drm/xe: keep process name fallback for diagnostics
URL : https://patchwork.freedesktop.org/series/168256/
State : failure
== Summary ==
Series author address '[email protected]' is not on the allowlist, which prevents CI from being automatically triggered. If you want CI to run for this series, ask Patchwork project owners to click 'retest' on the series in Patchwork. Exception occurred during validation, bailing out!
Build URL: http://intel-gfx-ci-public.igk.intel.com:8080/job/xe_pw_trigger/1185485/ (on master)