In the Linux kernel, the following vulnerability has been resolved: futex: Clear stale exiting pointer in futexlockpi() retry path Fuzzying/stressing futexes triggered: WARNING: kernel/futex/core.c:825 at waitforownerexiting+0x7a/0x80, CPU#11: futexlockpis/524 When futexlockpiatomic() sees the owner is exiting, it returns -EBUSY and stores a refcounted task pointer in 'exiting'. After waitforownerexiting() consumes that reference, the local pointer is never reset to nil. Upon a retry, if futexlockpiatomic() returns a different error, the bogus pointer is passed to waitforownerexiting(). CPU0 CPU1 CPU2 futexlockpi(uaddr) // acquires the PI futex exit() futexcleanupbegin() futexstate = EXITING; futexlockpi(uaddr) futexlockpiatomic() attachtopiowner() // observes EXITING *exiting = owner; // takes ref return -EBUSY waitforownerexiting(-EBUSY, owner) puttaskstruct(); // drops ref // exiting still points to owner goto retry; futexlockpiatomic() lockpiupdateatomic() cmpxchg(uaddr) *uaddr ^= WAITERS // whatever // value changed return -EAGAIN; waitforownerexiting(-EAGAIN, exiting) // stale WARNONONCE(exiting) Fix this by resetting upon retry, essentially aligning it with requeuepi.