PT-2026-152651
CVE-2026-99772
·
تاريخ النشر
2026-11-20
·
تاريخ التحديث
2026-11-20
الوصف من
الدرجة
N/A
غير متوفرلا تتوفر حاليًا تقييمات أو مقاييس للخطورة. سنحدّث المعلومات المقابلة في الصفحة فور توفرها.
In the Linux kernel, the following vulnerability has been resolved:
drm/drm crtc: fix race with dma fence signal() in ::get driver name()
Since commit 541c8f2468b9 ("dma-buf: detach fence ops on signal v3"),
I'm seeing the BUG ON() triggering in drm crtc's fence to crtc() via
drm crtc fence get driver name() regularly:
Call trace:
panic+0x58/0x5c
die+0x160/0x178
bug brk handler+0x70/0xa4
call el1 break hook+0x3c/0x1a0
do el1 brk64+0x24/0x74
el1 brk64+0x34/0x54
el1h 64 sync handler+0x80/0xfc
el1h 64 sync+0x84/0x88
drm crtc fence get driver name+0x60/0x68 (P)
sync file get name+0x184/0x45c
sync file ioctl+0x404/0xf70
arm64 sys ioctl+0x124/0x1dc
This looks to be caused by a code flow similar to the following:
+++ snip +++
thread A thread B
ioctl(SYNC IOC FILE INFO)
sync file ioctl()
sync file get name()dma fence signal timestamp locked() dma fence driver name()
ops = rcu dereference(fence->ops)
if (!dma fence test signaled flag())
ops->get driver name(fence) i.e.
drm crtc fence get driver name()
test and set bit(SIGNALED)
RCU INIT POINTER(fence->ops, NULL)
drm crtc fence get driver name()
BUG ON(rcu access pointer(fence->ops)
!= &drm crtc fence ops)
+++ snap +++
In other words, a fence's ops pointer can now change to NULL on signal,
which wasn't the case before the culprit commit and hence the BUG ON()
triggers.
Simply drop the BUG ON() as it's usage of fence ops is invalid now, and
because using BUG ON() and friends to take down the system is an
unacceptable way to handle a failure.
Note that the adjacent drm crtc fence get timeline name() has the same
problem and is fixed by this patch as well.
وجدت خطأً في الوصف؟ لديك ما تضيفه؟ لا تتردد في مراسلتنا 👾