CVE-2026-80777: futex/pi: Plug private futex exec() race
In the Linux kernel, the following vulnerability has been resolved:
futex/pi: Plug private futex exec() race
The check for private futexes whether the waiter's mm, which is stored in the futexkey and copied into the pistate, is the same as the owner's mm is not sufficient for exec(). exec() has a gap where the mm check fails to give the correct answer:
exec() ... execreleasemm() futexexecrelease() tsk::futex::exitstate = EXITING; cleanuprobustlist(); 1) tsk::futex::exitstate = OK; ... oldmm = tsk::mm; 2) tsk::mm = ->mm;
Between #1 and #2 the check for the mm is wrong as that mm is about to be swapped out and eventually freed.
Plug this gap by:
1) Setting tsk::futex::exitstate to FUTEXSTATEDEAD in futexexecrelease()
2) Setting tsk::futex::exitstate to FUTEXSTATEOK after the mm has been switched.
From a futex point of view the task is dead after it finished the robust list cleanup up to the point where it sets the state to OK again.
Affected Software
Event History
Frequently Asked Questions
Which workloads are relevant to this issue?
The issue concerns Linux kernel handling of private priority-inheritance futexes when a task performs exec(). It is relevant where private PI futex activity can overlap the exec() transition.
What is the fix intended to prevent?
The fix treats the task as dead to futex handling after robust-list cleanup during exec(), until the task's mm has been switched. This closes the interval in which comparing the waiter and owner mm values can produce an incorrect result.