In the Linux kernel, the following vulnerability has been resolved:
drm/sched: Fix deadlock in drmschedentitykilljobs_cb
The Mesa issue referenced below pointed out a possible deadlock:
[ 1231.611031] Possible interrupt unsafe locking scenario:
[ 1231.611033] CPU0 CPU1 [ 1231.611034] ---- ---- [ 1231.611035] lock(&xa->xalock#17); [ 1231.611038] localirqdisable(); [ 1231.611039] lock(&fence->lock); [ 1231.611041] lock(&xa->xalock#17); [ 1231.611044] <Interrupt> [ 1231.611045] lock(&fence->lock); [ 1231.611047] * DEADLOCK *
In this example, CPU0 would be any function accessing job->dependencies through the xa* functions that don't disable interrupts (eg: drmschedjobadddependency(), drmschedentitykilljobscb()).
CPU1 is executing drmschedentitykilljobscb() as a fence signalling callback so in an interrupt context. It will deadlock when trying to grab the xalock which is already held by CPU0.
Replacing all xa* usage by their xa*irq counterparts would fix this issue, but Christian pointed out another issue: dmafencesignal takes fence.lock and so does dmafenceaddcallback.
dmafencesignal() // locks f1.lock -> drmschedentitykilljobscb() -> foreach dependencies -> dmafenceaddcallback() // locks f2.lock
This will deadlock if f1 and f2 share the same spinlock.
To fix both issues, the code iterating on dependencies and re-arming them is moved out to drmschedentitykilljobs_work().
[phasta: commit message nits]
{
"osv_generated_from": "https://github.com/CVEProject/cvelistV5/tree/main/cves/2025/40xxx/CVE-2025-40329.json",
"cna_assigner": "Linux"
}