ipc/sem.c: synchronize semop and semctl with IPC_RMID
Manfred Spraul [Wed, 16 Oct 2013 20:46:45 +0000 (13:46 -0700)]
commit 6e224f94597842c5eb17f1fc2208d20b6f7f7d49 upstream.

After acquiring the semlock spinlock, operations must test that the
array is still valid.

 - semctl() and exit_sem() would walk stale linked lists (ugly, but
   should be ok: all lists are empty)

 - semtimedop() would sleep forever - and if woken up due to a signal -
   access memory after free.

The patch also:
 - standardizes the tests for .deleted, so that all tests in one
   function leave the function with the same approach.
 - unconditionally tests for .deleted immediately after every call to
   sem_lock - even it it means that for semctl(GETALL), .deleted will be
   tested twice.

Both changes make the review simpler: After every sem_lock, there must
be a test of .deleted, followed by a goto to the cleanup code (if the
function uses "goto cleanup").

The only exception is semctl_down(): If sem_ids().rwsem is locked, then
the presence in ids->ipcs_idr is equivalent to !.deleted, thus no
additional test is required.

Signed-off-by: Manfred Spraul <manfred@colorfullife.com>
Cc: Mike Galbraith <efault@gmx.de>
Acked-by: Davidlohr Bueso <davidlohr@hp.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

ipc/sem.c

index 8c4f59b..db9d241 100644 (file)
--- a/ipc/sem.c
+++ b/ipc/sem.c
@@ -1282,6 +1282,12 @@ static int semctl_setval(struct ipc_namespace *ns, int semid, int semnum,
 
        sem_lock(sma, NULL, -1);
 
+       if (sma->sem_perm.deleted) {
+               sem_unlock(sma, -1);
+               rcu_read_unlock();
+               return -EIDRM;
+       }
+
        curr = &sma->sem_base[semnum];
 
        ipc_assert_locked_object(&sma->sem_perm);
@@ -1336,12 +1342,14 @@ static int semctl_main(struct ipc_namespace *ns, int semid, int semnum,
                int i;
 
                sem_lock(sma, NULL, -1);
+               if (sma->sem_perm.deleted) {
+                       err = -EIDRM;
+                       goto out_unlock;
+               }
                if(nsems > SEMMSL_FAST) {
                        if (!ipc_rcu_getref(sma)) {
-                               sem_unlock(sma, -1);
-                               rcu_read_unlock();
                                err = -EIDRM;
-                               goto out_free;
+                               goto out_unlock;
                        }
                        sem_unlock(sma, -1);
                        rcu_read_unlock();
@@ -1354,10 +1362,8 @@ static int semctl_main(struct ipc_namespace *ns, int semid, int semnum,
                        rcu_read_lock();
                        sem_lock_and_putref(sma);
                        if (sma->sem_perm.deleted) {
-                               sem_unlock(sma, -1);
-                               rcu_read_unlock();
                                err = -EIDRM;
-                               goto out_free;
+                               goto out_unlock;
                        }
                }
                for (i = 0; i < sma->sem_nsems; i++)
@@ -1375,8 +1381,8 @@ static int semctl_main(struct ipc_namespace *ns, int semid, int semnum,
                struct sem_undo *un;
 
                if (!ipc_rcu_getref(sma)) {
-                       rcu_read_unlock();
-                       return -EIDRM;
+                       err = -EIDRM;
+                       goto out_rcu_wakeup;
                }
                rcu_read_unlock();
 
@@ -1404,10 +1410,8 @@ static int semctl_main(struct ipc_namespace *ns, int semid, int semnum,
                rcu_read_lock();
                sem_lock_and_putref(sma);
                if (sma->sem_perm.deleted) {
-                       sem_unlock(sma, -1);
-                       rcu_read_unlock();
                        err = -EIDRM;
-                       goto out_free;
+                       goto out_unlock;
                }
 
                for (i = 0; i < nsems; i++)
@@ -1431,6 +1435,10 @@ static int semctl_main(struct ipc_namespace *ns, int semid, int semnum,
                goto out_rcu_wakeup;
 
        sem_lock(sma, NULL, -1);
+       if (sma->sem_perm.deleted) {
+               err = -EIDRM;
+               goto out_unlock;
+       }
        curr = &sma->sem_base[semnum];
 
        switch (cmd) {
@@ -1836,6 +1844,10 @@ SYSCALL_DEFINE4(semtimedop, int, semid, struct sembuf __user *, tsops,
        if (error)
                goto out_rcu_wakeup;
 
+       error = -EIDRM;
+       locknum = sem_lock(sma, sops, nsops);
+       if (sma->sem_perm.deleted)
+               goto out_unlock_free;
        /*
         * semid identifiers are not unique - find_alloc_undo may have
         * allocated an undo structure, it was invalidated by an RMID
@@ -1843,8 +1855,6 @@ SYSCALL_DEFINE4(semtimedop, int, semid, struct sembuf __user *, tsops,
         * This case can be detected checking un->semid. The existence of
         * "un" itself is guaranteed by rcu.
         */
-       error = -EIDRM;
-       locknum = sem_lock(sma, sops, nsops);
        if (un && un->semid == -1)
                goto out_unlock_free;
 
@@ -2057,6 +2067,12 @@ void exit_sem(struct task_struct *tsk)
                }
 
                sem_lock(sma, NULL, -1);
+               /* exit_sem raced with IPC_RMID, nothing to do */
+               if (sma->sem_perm.deleted) {
+                       sem_unlock(sma, -1);
+                       rcu_read_unlock();
+                       continue;
+               }
                un = __lookup_undo(ulp, semid);
                if (un == NULL) {
                        /* exit_sem raced with IPC_RMID+semget() that created