aboutsummaryrefslogtreecommitdiffstats
path: root/ipc
diff options
context:
space:
mode:
authorManfred Spraul <manfred@colorfullife.com>2013-09-30 16:45:04 -0400
committerLinus Torvalds <torvalds@linux-foundation.org>2013-09-30 17:31:01 -0400
commit5e9d527591421ccdb16acb8c23662231135d8686 (patch)
treedbe055ecd0bb6e497dceab2c9c0136151d04063a /ipc
parentf6ea3adb70b20ae36277a1b0eaaf4da9f6479a28 (diff)
ipc/sem.c: fix race in sem_lock()
The exclusion of complex operations in sem_lock() is insufficient: after acquiring the per-semaphore lock, a simple op must first check that sem_perm.lock is not locked and only after that test check complex_count. The current code does it the other way around - and that creates a race. Details are below. The patch is a complete rewrite of sem_lock(), based in part on the code from Mike Galbraith. It removes all gotos and all loops and thus the risk of livelocks. I have tested the patch (together with the next one) on my i3 laptop and it didn't cause any problems. The bug is probably also present in 3.10 and 3.11, but for these kernels it might be simpler just to move the test of sma->complex_count after the spin_is_locked() test. Details of the bug: Assume: - sma->complex_count = 0. - Thread 1: semtimedop(complex op that must sleep) - Thread 2: semtimedop(simple op). Pseudo-Trace: Thread 1: sem_lock(): acquire sem_perm.lock Thread 1: sem_lock(): check for ongoing simple ops Nothing ongoing, thread 2 is still before sem_lock(). Thread 1: try_atomic_semop() <<< preempted. Thread 2: sem_lock(): static inline int sem_lock(struct sem_array *sma, struct sembuf *sops, int nsops) { int locknum; again: if (nsops == 1 && !sma->complex_count) { struct sem *sem = sma->sem_base + sops->sem_num; /* Lock just the semaphore we are interested in. */ spin_lock(&sem->lock); /* * If sma->complex_count was set while we were spinning, * we may need to look at things we did not lock here. */ if (unlikely(sma->complex_count)) { spin_unlock(&sem->lock); goto lock_array; } <<<<<<<<< <<< complex_count is still 0. <<< <<< Here it is preempted <<<<<<<<< Thread 1: try_atomic_semop() returns, notices that it must sleep. Thread 1: increases sma->complex_count. Thread 1: drops sem_perm.lock Thread 2: /* * Another process is holding the global lock on the * sem_array; we cannot enter our critical section, * but have to wait for the global lock to be released. */ if (unlikely(spin_is_locked(&sma->sem_perm.lock))) { spin_unlock(&sem->lock); spin_unlock_wait(&sma->sem_perm.lock); goto again; } <<< sem_perm.lock already dropped, thus no "goto again;" locknum = sops->sem_num; Signed-off-by: Manfred Spraul <manfred@colorfullife.com> Cc: Mike Galbraith <bitbucket@online.de> Cc: Rik van Riel <riel@redhat.com> Cc: Davidlohr Bueso <davidlohr.bueso@hp.com> Cc: <stable@vger.kernel.org> [3.10+] Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
Diffstat (limited to 'ipc')
-rw-r--r--ipc/sem.c122
1 files changed, 78 insertions, 44 deletions
diff --git a/ipc/sem.c b/ipc/sem.c
index 19c8b980d1fe..4a92c0447ad6 100644
--- a/ipc/sem.c
+++ b/ipc/sem.c
@@ -253,70 +253,104 @@ static void sem_rcu_free(struct rcu_head *head)
253} 253}
254 254
255/* 255/*
256 * Wait until all currently ongoing simple ops have completed.
257 * Caller must own sem_perm.lock.
258 * New simple ops cannot start, because simple ops first check
259 * that sem_perm.lock is free.
260 */
261static void sem_wait_array(struct sem_array *sma)
262{
263 int i;
264 struct sem *sem;
265
266 for (i = 0; i < sma->sem_nsems; i++) {
267 sem = sma->sem_base + i;
268 spin_unlock_wait(&sem->lock);
269 }
270}
271
272/*
256 * If the request contains only one semaphore operation, and there are 273 * If the request contains only one semaphore operation, and there are
257 * no complex transactions pending, lock only the semaphore involved. 274 * no complex transactions pending, lock only the semaphore involved.
258 * Otherwise, lock the entire semaphore array, since we either have 275 * Otherwise, lock the entire semaphore array, since we either have
259 * multiple semaphores in our own semops, or we need to look at 276 * multiple semaphores in our own semops, or we need to look at
260 * semaphores from other pending complex operations. 277 * semaphores from other pending complex operations.
261 *
262 * Carefully guard against sma->complex_count changing between zero
263 * and non-zero while we are spinning for the lock. The value of
264 * sma->complex_count cannot change while we are holding the lock,
265 * so sem_unlock should be fine.
266 *
267 * The global lock path checks that all the local locks have been released,
268 * checking each local lock once. This means that the local lock paths
269 * cannot start their critical sections while the global lock is held.
270 */ 278 */
271static inline int sem_lock(struct sem_array *sma, struct sembuf *sops, 279static inline int sem_lock(struct sem_array *sma, struct sembuf *sops,
272 int nsops) 280 int nsops)
273{ 281{
274 int locknum; 282 struct sem *sem;
275 again:
276 if (nsops == 1 && !sma->complex_count) {
277 struct sem *sem = sma->sem_base + sops->sem_num;
278 283
279 /* Lock just the semaphore we are interested in. */ 284 if (nsops != 1) {
280 spin_lock(&sem->lock); 285 /* Complex operation - acquire a full lock */
286 ipc_lock_object(&sma->sem_perm);
281 287
282 /* 288 /* And wait until all simple ops that are processed
283 * If sma->complex_count was set while we were spinning, 289 * right now have dropped their locks.
284 * we may need to look at things we did not lock here.
285 */ 290 */
286 if (unlikely(sma->complex_count)) { 291 sem_wait_array(sma);
287 spin_unlock(&sem->lock); 292 return -1;
288 goto lock_array; 293 }
289 } 294
295 /*
296 * Only one semaphore affected - try to optimize locking.
297 * The rules are:
298 * - optimized locking is possible if no complex operation
299 * is either enqueued or processed right now.
300 * - The test for enqueued complex ops is simple:
301 * sma->complex_count != 0
302 * - Testing for complex ops that are processed right now is
303 * a bit more difficult. Complex ops acquire the full lock
304 * and first wait that the running simple ops have completed.
305 * (see above)
306 * Thus: If we own a simple lock and the global lock is free
307 * and complex_count is now 0, then it will stay 0 and
308 * thus just locking sem->lock is sufficient.
309 */
310 sem = sma->sem_base + sops->sem_num;
290 311
312 if (sma->complex_count == 0) {
291 /* 313 /*
292 * Another process is holding the global lock on the 314 * It appears that no complex operation is around.
293 * sem_array; we cannot enter our critical section, 315 * Acquire the per-semaphore lock.
294 * but have to wait for the global lock to be released.
295 */ 316 */
296 if (unlikely(spin_is_locked(&sma->sem_perm.lock))) { 317 spin_lock(&sem->lock);
297 spin_unlock(&sem->lock); 318
298 spin_unlock_wait(&sma->sem_perm.lock); 319 /* Then check that the global lock is free */
299 goto again; 320 if (!spin_is_locked(&sma->sem_perm.lock)) {
321 /* spin_is_locked() is not a memory barrier */
322 smp_mb();
323
324 /* Now repeat the test of complex_count:
325 * It can't change anymore until we drop sem->lock.
326 * Thus: if is now 0, then it will stay 0.
327 */
328 if (sma->complex_count == 0) {
329 /* fast path successful! */
330 return sops->sem_num;
331 }
300 } 332 }
333 spin_unlock(&sem->lock);
334 }
301 335
302 locknum = sops->sem_num; 336 /* slow path: acquire the full lock */
337 ipc_lock_object(&sma->sem_perm);
338
339 if (sma->complex_count == 0) {
340 /* False alarm:
341 * There is no complex operation, thus we can switch
342 * back to the fast path.
343 */
344 spin_lock(&sem->lock);
345 ipc_unlock_object(&sma->sem_perm);
346 return sops->sem_num;
303 } else { 347 } else {
304 int i; 348 /* Not a false alarm, thus complete the sequence for a
305 /* 349 * full lock.
306 * Lock the semaphore array, and wait for all of the
307 * individual semaphore locks to go away. The code
308 * above ensures no new single-lock holders will enter
309 * their critical section while the array lock is held.
310 */ 350 */
311 lock_array: 351 sem_wait_array(sma);
312 ipc_lock_object(&sma->sem_perm); 352 return -1;
313 for (i = 0; i < sma->sem_nsems; i++) {
314 struct sem *sem = sma->sem_base + i;
315 spin_unlock_wait(&sem->lock);
316 }
317 locknum = -1;
318 } 353 }
319 return locknum;
320} 354}
321 355
322static inline void sem_unlock(struct sem_array *sma, int locknum) 356static inline void sem_unlock(struct sem_array *sma, int locknum)