ipc/sem.c: fix return code race with semop vs. semop +semctl(IPC_RMID)
authorManfred Spraul <manfred@colorfullife.com>
Wed, 2 Nov 2011 20:38:50 +0000 (13:38 -0700)
committerLinus Torvalds <torvalds@linux-foundation.org>
Wed, 2 Nov 2011 23:07:01 +0000 (16:07 -0700)
commit3c24783bb2deafaa106b7e69a97540071afc590c
tree904d59c7e197d59c472e20fc42057a5acb90f223
parent46cbc1d3981ee753518fbf9198a14f71a9f6841e
ipc/sem.c: fix return code race with semop vs. semop +semctl(IPC_RMID)

sys_semtimedop() may return -EIDRM although the semaphore operation
completed successfully:

thread 1: thread 2:
semtimedop(), sleeps
semop():
* acquires sem_lock()
semtimedop() woken up due to timeout
sem_lock() loops
* notices that thread 2 could be completed.
* performs the operations that thread 2 is sleeping on.
* marks the semaphore operation as IN_WAKEUP
* drops sem_lock(), does wakeup, sets return code to 0
* thread delayed due to interrupt, whatever
* returns to user space
* thread still delayed
semctl(IPC_RMID)
* acquires sem_lock()
* ipc_rmid(), ipcp->deleted=1
* drops sem_lock()
* thread finally continues - but seem_lock()
  now fails due to ipcp->deleted == 1
* returns -EIDRM instead of 0

The fix is trivial: Always use the return code in queue.status.

In real world, the race probably doesn't matter:
If the semaphore array is destroyed, the app is probably not interested
if the last operation succeeded or was already cancelled.

Signed-off-by: Manfred Spraul <manfred@colorfullife.com>
Cc: Thomas Gleixner <tglx@linutronix.de>
Cc: Mike Galbraith <efault@gmx.de>
Acked-by: Peter Zijlstra <a.p.zijlstra@chello.nl>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
ipc/sem.c