]>
Commit | Line | Data |
---|---|---|
1da177e4 LT |
1 | /* |
2 | * linux/ipc/sem.c | |
3 | * Copyright (C) 1992 Krishna Balasubramanian | |
4 | * Copyright (C) 1995 Eric Schenk, Bruno Haible | |
5 | * | |
1da177e4 LT |
6 | * /proc/sysvipc/sem support (c) 1999 Dragos Acostachioaie <dragos@iname.com> |
7 | * | |
8 | * SMP-threaded, sysctl's added | |
624dffcb | 9 | * (c) 1999 Manfred Spraul <manfred@colorfullife.com> |
1da177e4 | 10 | * Enforced range limit on SEM_UNDO |
046c6884 | 11 | * (c) 2001 Red Hat Inc |
1da177e4 LT |
12 | * Lockless wakeup |
13 | * (c) 2003 Manfred Spraul <manfred@colorfullife.com> | |
c5cf6359 MS |
14 | * Further wakeup optimizations, documentation |
15 | * (c) 2010 Manfred Spraul <manfred@colorfullife.com> | |
073115d6 SG |
16 | * |
17 | * support for audit of ipc object properties and permission changes | |
18 | * Dustin Kirkland <dustin.kirkland@us.ibm.com> | |
e3893534 KK |
19 | * |
20 | * namespaces support | |
21 | * OpenVZ, SWsoft Inc. | |
22 | * Pavel Emelianov <xemul@openvz.org> | |
c5cf6359 MS |
23 | * |
24 | * Implementation notes: (May 2010) | |
25 | * This file implements System V semaphores. | |
26 | * | |
27 | * User space visible behavior: | |
28 | * - FIFO ordering for semop() operations (just FIFO, not starvation | |
29 | * protection) | |
30 | * - multiple semaphore operations that alter the same semaphore in | |
31 | * one semop() are handled. | |
32 | * - sem_ctime (time of last semctl()) is updated in the IPC_SET, SETVAL and | |
33 | * SETALL calls. | |
34 | * - two Linux specific semctl() commands: SEM_STAT, SEM_INFO. | |
35 | * - undo adjustments at process exit are limited to 0..SEMVMX. | |
36 | * - namespace are supported. | |
37 | * - SEMMSL, SEMMNS, SEMOPM and SEMMNI can be configured at runtine by writing | |
38 | * to /proc/sys/kernel/sem. | |
39 | * - statistics about the usage are reported in /proc/sysvipc/sem. | |
40 | * | |
41 | * Internals: | |
42 | * - scalability: | |
43 | * - all global variables are read-mostly. | |
44 | * - semop() calls and semctl(RMID) are synchronized by RCU. | |
45 | * - most operations do write operations (actually: spin_lock calls) to | |
46 | * the per-semaphore array structure. | |
47 | * Thus: Perfect SMP scaling between independent semaphore arrays. | |
48 | * If multiple semaphores in one array are used, then cache line | |
49 | * trashing on the semaphore array spinlock will limit the scaling. | |
50 | * - semncnt and semzcnt are calculated on demand in count_semncnt() and | |
51 | * count_semzcnt() | |
52 | * - the task that performs a successful semop() scans the list of all | |
53 | * sleeping tasks and completes any pending operations that can be fulfilled. | |
54 | * Semaphores are actively given to waiting tasks (necessary for FIFO). | |
55 | * (see update_queue()) | |
56 | * - To improve the scalability, the actual wake-up calls are performed after | |
57 | * dropping all locks. (see wake_up_sem_queue_prepare(), | |
58 | * wake_up_sem_queue_do()) | |
59 | * - All work is done by the waker, the woken up task does not have to do | |
60 | * anything - not even acquiring a lock or dropping a refcount. | |
61 | * - A woken up task may not even touch the semaphore array anymore, it may | |
62 | * have been destroyed already by a semctl(RMID). | |
63 | * - The synchronizations between wake-ups due to a timeout/signal and a | |
64 | * wake-up due to a completed semaphore operation is achieved by using an | |
65 | * intermediate state (IN_WAKEUP). | |
66 | * - UNDO values are stored in an array (one per process and per | |
67 | * semaphore array, lazily allocated). For backwards compatibility, multiple | |
68 | * modes for the UNDO variables are supported (per process, per thread) | |
69 | * (see copy_semundo, CLONE_SYSVSEM) | |
70 | * - There are two lists of the pending operations: a per-array list | |
71 | * and per-semaphore list (stored in the array). This allows to achieve FIFO | |
72 | * ordering without always scanning all pending operations. | |
73 | * The worst-case behavior is nevertheless O(N^2) for N wakeups. | |
1da177e4 LT |
74 | */ |
75 | ||
1da177e4 LT |
76 | #include <linux/slab.h> |
77 | #include <linux/spinlock.h> | |
78 | #include <linux/init.h> | |
79 | #include <linux/proc_fs.h> | |
80 | #include <linux/time.h> | |
1da177e4 LT |
81 | #include <linux/security.h> |
82 | #include <linux/syscalls.h> | |
83 | #include <linux/audit.h> | |
c59ede7b | 84 | #include <linux/capability.h> |
19b4946c | 85 | #include <linux/seq_file.h> |
3e148c79 | 86 | #include <linux/rwsem.h> |
e3893534 | 87 | #include <linux/nsproxy.h> |
ae5e1b22 | 88 | #include <linux/ipc_namespace.h> |
5f921ae9 | 89 | |
1da177e4 LT |
90 | #include <asm/uaccess.h> |
91 | #include "util.h" | |
92 | ||
e57940d7 MS |
93 | /* One semaphore structure for each semaphore in the system. */ |
94 | struct sem { | |
95 | int semval; /* current value */ | |
96 | int sempid; /* pid of last operation */ | |
6062a8dc | 97 | spinlock_t lock; /* spinlock for fine-grained semtimedop */ |
1a82e9e1 MS |
98 | struct list_head pending_alter; /* pending single-sop operations */ |
99 | /* that alter the semaphore */ | |
100 | struct list_head pending_const; /* pending single-sop operations */ | |
101 | /* that do not alter the semaphore*/ | |
f5c936c0 | 102 | } ____cacheline_aligned_in_smp; |
e57940d7 MS |
103 | |
104 | /* One queue for each sleeping process in the system. */ | |
105 | struct sem_queue { | |
e57940d7 MS |
106 | struct list_head list; /* queue of pending operations */ |
107 | struct task_struct *sleeper; /* this process */ | |
108 | struct sem_undo *undo; /* undo structure */ | |
109 | int pid; /* process id of requesting process */ | |
110 | int status; /* completion status of operation */ | |
111 | struct sembuf *sops; /* array of pending operations */ | |
112 | int nsops; /* number of operations */ | |
113 | int alter; /* does *sops alter the array? */ | |
114 | }; | |
115 | ||
116 | /* Each task has a list of undo requests. They are executed automatically | |
117 | * when the process exits. | |
118 | */ | |
119 | struct sem_undo { | |
120 | struct list_head list_proc; /* per-process list: * | |
121 | * all undos from one process | |
122 | * rcu protected */ | |
123 | struct rcu_head rcu; /* rcu struct for sem_undo */ | |
124 | struct sem_undo_list *ulp; /* back ptr to sem_undo_list */ | |
125 | struct list_head list_id; /* per semaphore array list: | |
126 | * all undos for one array */ | |
127 | int semid; /* semaphore set identifier */ | |
128 | short *semadj; /* array of adjustments */ | |
129 | /* one per semaphore */ | |
130 | }; | |
131 | ||
132 | /* sem_undo_list controls shared access to the list of sem_undo structures | |
133 | * that may be shared among all a CLONE_SYSVSEM task group. | |
134 | */ | |
135 | struct sem_undo_list { | |
136 | atomic_t refcnt; | |
137 | spinlock_t lock; | |
138 | struct list_head list_proc; | |
139 | }; | |
140 | ||
141 | ||
ed2ddbf8 | 142 | #define sem_ids(ns) ((ns)->ids[IPC_SEM_IDS]) |
e3893534 | 143 | |
1b531f21 | 144 | #define sem_checkid(sma, semid) ipc_checkid(&sma->sem_perm, semid) |
1da177e4 | 145 | |
7748dbfa | 146 | static int newary(struct ipc_namespace *, struct ipc_params *); |
01b8b07a | 147 | static void freeary(struct ipc_namespace *, struct kern_ipc_perm *); |
1da177e4 | 148 | #ifdef CONFIG_PROC_FS |
19b4946c | 149 | static int sysvipc_sem_proc_show(struct seq_file *s, void *it); |
1da177e4 LT |
150 | #endif |
151 | ||
152 | #define SEMMSL_FAST 256 /* 512 bytes on stack */ | |
153 | #define SEMOPM_FAST 64 /* ~ 372 bytes on stack */ | |
154 | ||
155 | /* | |
156 | * linked list protection: | |
157 | * sem_undo.id_next, | |
1a82e9e1 | 158 | * sem_array.pending{_alter,_cont}, |
1da177e4 LT |
159 | * sem_array.sem_undo: sem_lock() for read/write |
160 | * sem_undo.proc_next: only "current" is allowed to read/write that field. | |
161 | * | |
162 | */ | |
163 | ||
e3893534 KK |
164 | #define sc_semmsl sem_ctls[0] |
165 | #define sc_semmns sem_ctls[1] | |
166 | #define sc_semopm sem_ctls[2] | |
167 | #define sc_semmni sem_ctls[3] | |
168 | ||
ed2ddbf8 | 169 | void sem_init_ns(struct ipc_namespace *ns) |
e3893534 | 170 | { |
e3893534 KK |
171 | ns->sc_semmsl = SEMMSL; |
172 | ns->sc_semmns = SEMMNS; | |
173 | ns->sc_semopm = SEMOPM; | |
174 | ns->sc_semmni = SEMMNI; | |
175 | ns->used_sems = 0; | |
ed2ddbf8 | 176 | ipc_init_ids(&ns->ids[IPC_SEM_IDS]); |
e3893534 KK |
177 | } |
178 | ||
ae5e1b22 | 179 | #ifdef CONFIG_IPC_NS |
e3893534 KK |
180 | void sem_exit_ns(struct ipc_namespace *ns) |
181 | { | |
01b8b07a | 182 | free_ipcs(ns, &sem_ids(ns), freeary); |
7d6feeb2 | 183 | idr_destroy(&ns->ids[IPC_SEM_IDS].ipcs_idr); |
e3893534 | 184 | } |
ae5e1b22 | 185 | #endif |
1da177e4 LT |
186 | |
187 | void __init sem_init (void) | |
188 | { | |
ed2ddbf8 | 189 | sem_init_ns(&init_ipc_ns); |
19b4946c MW |
190 | ipc_init_proc_interface("sysvipc/sem", |
191 | " key semid perms nsems uid gid cuid cgid otime ctime\n", | |
e3893534 | 192 | IPC_SEM_IDS, sysvipc_sem_proc_show); |
1da177e4 LT |
193 | } |
194 | ||
f269f40a MS |
195 | /** |
196 | * unmerge_queues - unmerge queues, if possible. | |
197 | * @sma: semaphore array | |
198 | * | |
199 | * The function unmerges the wait queues if complex_count is 0. | |
200 | * It must be called prior to dropping the global semaphore array lock. | |
201 | */ | |
202 | static void unmerge_queues(struct sem_array *sma) | |
203 | { | |
204 | struct sem_queue *q, *tq; | |
205 | ||
206 | /* complex operations still around? */ | |
207 | if (sma->complex_count) | |
208 | return; | |
209 | /* | |
210 | * We will switch back to simple mode. | |
211 | * Move all pending operation back into the per-semaphore | |
212 | * queues. | |
213 | */ | |
214 | list_for_each_entry_safe(q, tq, &sma->pending_alter, list) { | |
215 | struct sem *curr; | |
216 | curr = &sma->sem_base[q->sops[0].sem_num]; | |
217 | ||
218 | list_add_tail(&q->list, &curr->pending_alter); | |
219 | } | |
220 | INIT_LIST_HEAD(&sma->pending_alter); | |
221 | } | |
222 | ||
223 | /** | |
224 | * merge_queues - Merge single semop queues into global queue | |
225 | * @sma: semaphore array | |
226 | * | |
227 | * This function merges all per-semaphore queues into the global queue. | |
228 | * It is necessary to achieve FIFO ordering for the pending single-sop | |
229 | * operations when a multi-semop operation must sleep. | |
230 | * Only the alter operations must be moved, the const operations can stay. | |
231 | */ | |
232 | static void merge_queues(struct sem_array *sma) | |
233 | { | |
234 | int i; | |
235 | for (i = 0; i < sma->sem_nsems; i++) { | |
236 | struct sem *sem = sma->sem_base + i; | |
237 | ||
238 | list_splice_init(&sem->pending_alter, &sma->pending_alter); | |
239 | } | |
240 | } | |
241 | ||
6062a8dc RR |
242 | /* |
243 | * If the request contains only one semaphore operation, and there are | |
244 | * no complex transactions pending, lock only the semaphore involved. | |
245 | * Otherwise, lock the entire semaphore array, since we either have | |
246 | * multiple semaphores in our own semops, or we need to look at | |
247 | * semaphores from other pending complex operations. | |
248 | * | |
249 | * Carefully guard against sma->complex_count changing between zero | |
250 | * and non-zero while we are spinning for the lock. The value of | |
251 | * sma->complex_count cannot change while we are holding the lock, | |
252 | * so sem_unlock should be fine. | |
253 | * | |
254 | * The global lock path checks that all the local locks have been released, | |
255 | * checking each local lock once. This means that the local lock paths | |
256 | * cannot start their critical sections while the global lock is held. | |
257 | */ | |
258 | static inline int sem_lock(struct sem_array *sma, struct sembuf *sops, | |
259 | int nsops) | |
260 | { | |
261 | int locknum; | |
262 | again: | |
263 | if (nsops == 1 && !sma->complex_count) { | |
264 | struct sem *sem = sma->sem_base + sops->sem_num; | |
265 | ||
266 | /* Lock just the semaphore we are interested in. */ | |
267 | spin_lock(&sem->lock); | |
268 | ||
269 | /* | |
270 | * If sma->complex_count was set while we were spinning, | |
271 | * we may need to look at things we did not lock here. | |
272 | */ | |
273 | if (unlikely(sma->complex_count)) { | |
274 | spin_unlock(&sem->lock); | |
275 | goto lock_array; | |
276 | } | |
277 | ||
278 | /* | |
279 | * Another process is holding the global lock on the | |
280 | * sem_array; we cannot enter our critical section, | |
281 | * but have to wait for the global lock to be released. | |
282 | */ | |
283 | if (unlikely(spin_is_locked(&sma->sem_perm.lock))) { | |
284 | spin_unlock(&sem->lock); | |
285 | spin_unlock_wait(&sma->sem_perm.lock); | |
286 | goto again; | |
287 | } | |
288 | ||
289 | locknum = sops->sem_num; | |
290 | } else { | |
291 | int i; | |
292 | /* | |
293 | * Lock the semaphore array, and wait for all of the | |
294 | * individual semaphore locks to go away. The code | |
295 | * above ensures no new single-lock holders will enter | |
296 | * their critical section while the array lock is held. | |
297 | */ | |
298 | lock_array: | |
cf9d5d78 | 299 | ipc_lock_object(&sma->sem_perm); |
6062a8dc RR |
300 | for (i = 0; i < sma->sem_nsems; i++) { |
301 | struct sem *sem = sma->sem_base + i; | |
302 | spin_unlock_wait(&sem->lock); | |
303 | } | |
304 | locknum = -1; | |
305 | } | |
306 | return locknum; | |
307 | } | |
308 | ||
309 | static inline void sem_unlock(struct sem_array *sma, int locknum) | |
310 | { | |
311 | if (locknum == -1) { | |
f269f40a | 312 | unmerge_queues(sma); |
cf9d5d78 | 313 | ipc_unlock_object(&sma->sem_perm); |
6062a8dc RR |
314 | } else { |
315 | struct sem *sem = sma->sem_base + locknum; | |
316 | spin_unlock(&sem->lock); | |
317 | } | |
6062a8dc RR |
318 | } |
319 | ||
3e148c79 ND |
320 | /* |
321 | * sem_lock_(check_) routines are called in the paths where the rw_mutex | |
322 | * is not held. | |
321310ce LT |
323 | * |
324 | * The caller holds the RCU read lock. | |
3e148c79 | 325 | */ |
6062a8dc RR |
326 | static inline struct sem_array *sem_obtain_lock(struct ipc_namespace *ns, |
327 | int id, struct sembuf *sops, int nsops, int *locknum) | |
023a5355 | 328 | { |
c460b662 RR |
329 | struct kern_ipc_perm *ipcp; |
330 | struct sem_array *sma; | |
03f02c76 | 331 | |
c460b662 | 332 | ipcp = ipc_obtain_object(&sem_ids(ns), id); |
321310ce LT |
333 | if (IS_ERR(ipcp)) |
334 | return ERR_CAST(ipcp); | |
b1ed88b4 | 335 | |
6062a8dc RR |
336 | sma = container_of(ipcp, struct sem_array, sem_perm); |
337 | *locknum = sem_lock(sma, sops, nsops); | |
c460b662 RR |
338 | |
339 | /* ipc_rmid() may have already freed the ID while sem_lock | |
340 | * was spinning: verify that the structure is still valid | |
341 | */ | |
342 | if (!ipcp->deleted) | |
343 | return container_of(ipcp, struct sem_array, sem_perm); | |
344 | ||
6062a8dc | 345 | sem_unlock(sma, *locknum); |
321310ce | 346 | return ERR_PTR(-EINVAL); |
023a5355 ND |
347 | } |
348 | ||
16df3674 DB |
349 | static inline struct sem_array *sem_obtain_object(struct ipc_namespace *ns, int id) |
350 | { | |
351 | struct kern_ipc_perm *ipcp = ipc_obtain_object(&sem_ids(ns), id); | |
352 | ||
353 | if (IS_ERR(ipcp)) | |
354 | return ERR_CAST(ipcp); | |
355 | ||
356 | return container_of(ipcp, struct sem_array, sem_perm); | |
357 | } | |
358 | ||
16df3674 DB |
359 | static inline struct sem_array *sem_obtain_object_check(struct ipc_namespace *ns, |
360 | int id) | |
361 | { | |
362 | struct kern_ipc_perm *ipcp = ipc_obtain_object_check(&sem_ids(ns), id); | |
363 | ||
364 | if (IS_ERR(ipcp)) | |
365 | return ERR_CAST(ipcp); | |
b1ed88b4 | 366 | |
03f02c76 | 367 | return container_of(ipcp, struct sem_array, sem_perm); |
023a5355 ND |
368 | } |
369 | ||
6ff37972 PP |
370 | static inline void sem_lock_and_putref(struct sem_array *sma) |
371 | { | |
6062a8dc | 372 | sem_lock(sma, NULL, -1); |
6ff37972 PP |
373 | ipc_rcu_putref(sma); |
374 | } | |
375 | ||
6ff37972 PP |
376 | static inline void sem_putref(struct sem_array *sma) |
377 | { | |
73b29505 | 378 | ipc_rcu_putref(sma); |
6ff37972 PP |
379 | } |
380 | ||
7ca7e564 ND |
381 | static inline void sem_rmid(struct ipc_namespace *ns, struct sem_array *s) |
382 | { | |
383 | ipc_rmid(&sem_ids(ns), &s->sem_perm); | |
384 | } | |
385 | ||
1da177e4 LT |
386 | /* |
387 | * Lockless wakeup algorithm: | |
388 | * Without the check/retry algorithm a lockless wakeup is possible: | |
389 | * - queue.status is initialized to -EINTR before blocking. | |
390 | * - wakeup is performed by | |
1a82e9e1 | 391 | * * unlinking the queue entry from the pending list |
1da177e4 LT |
392 | * * setting queue.status to IN_WAKEUP |
393 | * This is the notification for the blocked thread that a | |
394 | * result value is imminent. | |
395 | * * call wake_up_process | |
396 | * * set queue.status to the final value. | |
397 | * - the previously blocked thread checks queue.status: | |
398 | * * if it's IN_WAKEUP, then it must wait until the value changes | |
399 | * * if it's not -EINTR, then the operation was completed by | |
400 | * update_queue. semtimedop can return queue.status without | |
5f921ae9 | 401 | * performing any operation on the sem array. |
1da177e4 LT |
402 | * * otherwise it must acquire the spinlock and check what's up. |
403 | * | |
404 | * The two-stage algorithm is necessary to protect against the following | |
405 | * races: | |
406 | * - if queue.status is set after wake_up_process, then the woken up idle | |
407 | * thread could race forward and try (and fail) to acquire sma->lock | |
408 | * before update_queue had a chance to set queue.status | |
409 | * - if queue.status is written before wake_up_process and if the | |
410 | * blocked process is woken up by a signal between writing | |
411 | * queue.status and the wake_up_process, then the woken up | |
412 | * process could return from semtimedop and die by calling | |
413 | * sys_exit before wake_up_process is called. Then wake_up_process | |
414 | * will oops, because the task structure is already invalid. | |
415 | * (yes, this happened on s390 with sysv msg). | |
416 | * | |
417 | */ | |
418 | #define IN_WAKEUP 1 | |
419 | ||
f4566f04 ND |
420 | /** |
421 | * newary - Create a new semaphore set | |
422 | * @ns: namespace | |
423 | * @params: ptr to the structure that contains key, semflg and nsems | |
424 | * | |
3e148c79 | 425 | * Called with sem_ids.rw_mutex held (as a writer) |
f4566f04 ND |
426 | */ |
427 | ||
7748dbfa | 428 | static int newary(struct ipc_namespace *ns, struct ipc_params *params) |
1da177e4 LT |
429 | { |
430 | int id; | |
431 | int retval; | |
432 | struct sem_array *sma; | |
433 | int size; | |
7748dbfa ND |
434 | key_t key = params->key; |
435 | int nsems = params->u.nsems; | |
436 | int semflg = params->flg; | |
b97e820f | 437 | int i; |
1da177e4 LT |
438 | |
439 | if (!nsems) | |
440 | return -EINVAL; | |
e3893534 | 441 | if (ns->used_sems + nsems > ns->sc_semmns) |
1da177e4 LT |
442 | return -ENOSPC; |
443 | ||
444 | size = sizeof (*sma) + nsems * sizeof (struct sem); | |
445 | sma = ipc_rcu_alloc(size); | |
446 | if (!sma) { | |
447 | return -ENOMEM; | |
448 | } | |
449 | memset (sma, 0, size); | |
450 | ||
451 | sma->sem_perm.mode = (semflg & S_IRWXUGO); | |
452 | sma->sem_perm.key = key; | |
453 | ||
454 | sma->sem_perm.security = NULL; | |
455 | retval = security_sem_alloc(sma); | |
456 | if (retval) { | |
457 | ipc_rcu_putref(sma); | |
458 | return retval; | |
459 | } | |
460 | ||
e3893534 | 461 | id = ipc_addid(&sem_ids(ns), &sma->sem_perm, ns->sc_semmni); |
283bb7fa | 462 | if (id < 0) { |
1da177e4 LT |
463 | security_sem_free(sma); |
464 | ipc_rcu_putref(sma); | |
283bb7fa | 465 | return id; |
1da177e4 | 466 | } |
e3893534 | 467 | ns->used_sems += nsems; |
1da177e4 LT |
468 | |
469 | sma->sem_base = (struct sem *) &sma[1]; | |
b97e820f | 470 | |
6062a8dc | 471 | for (i = 0; i < nsems; i++) { |
1a82e9e1 MS |
472 | INIT_LIST_HEAD(&sma->sem_base[i].pending_alter); |
473 | INIT_LIST_HEAD(&sma->sem_base[i].pending_const); | |
6062a8dc RR |
474 | spin_lock_init(&sma->sem_base[i].lock); |
475 | } | |
b97e820f MS |
476 | |
477 | sma->complex_count = 0; | |
1a82e9e1 MS |
478 | INIT_LIST_HEAD(&sma->pending_alter); |
479 | INIT_LIST_HEAD(&sma->pending_const); | |
4daa28f6 | 480 | INIT_LIST_HEAD(&sma->list_id); |
1da177e4 LT |
481 | sma->sem_nsems = nsems; |
482 | sma->sem_ctime = get_seconds(); | |
6062a8dc | 483 | sem_unlock(sma, -1); |
6d49dab8 | 484 | rcu_read_unlock(); |
1da177e4 | 485 | |
7ca7e564 | 486 | return sma->sem_perm.id; |
1da177e4 LT |
487 | } |
488 | ||
7748dbfa | 489 | |
f4566f04 | 490 | /* |
3e148c79 | 491 | * Called with sem_ids.rw_mutex and ipcp locked. |
f4566f04 | 492 | */ |
03f02c76 | 493 | static inline int sem_security(struct kern_ipc_perm *ipcp, int semflg) |
7748dbfa | 494 | { |
03f02c76 ND |
495 | struct sem_array *sma; |
496 | ||
497 | sma = container_of(ipcp, struct sem_array, sem_perm); | |
498 | return security_sem_associate(sma, semflg); | |
7748dbfa ND |
499 | } |
500 | ||
f4566f04 | 501 | /* |
3e148c79 | 502 | * Called with sem_ids.rw_mutex and ipcp locked. |
f4566f04 | 503 | */ |
03f02c76 ND |
504 | static inline int sem_more_checks(struct kern_ipc_perm *ipcp, |
505 | struct ipc_params *params) | |
7748dbfa | 506 | { |
03f02c76 ND |
507 | struct sem_array *sma; |
508 | ||
509 | sma = container_of(ipcp, struct sem_array, sem_perm); | |
510 | if (params->u.nsems > sma->sem_nsems) | |
7748dbfa ND |
511 | return -EINVAL; |
512 | ||
513 | return 0; | |
514 | } | |
515 | ||
d5460c99 | 516 | SYSCALL_DEFINE3(semget, key_t, key, int, nsems, int, semflg) |
1da177e4 | 517 | { |
e3893534 | 518 | struct ipc_namespace *ns; |
7748dbfa ND |
519 | struct ipc_ops sem_ops; |
520 | struct ipc_params sem_params; | |
e3893534 KK |
521 | |
522 | ns = current->nsproxy->ipc_ns; | |
1da177e4 | 523 | |
e3893534 | 524 | if (nsems < 0 || nsems > ns->sc_semmsl) |
1da177e4 | 525 | return -EINVAL; |
7ca7e564 | 526 | |
7748dbfa ND |
527 | sem_ops.getnew = newary; |
528 | sem_ops.associate = sem_security; | |
529 | sem_ops.more_checks = sem_more_checks; | |
530 | ||
531 | sem_params.key = key; | |
532 | sem_params.flg = semflg; | |
533 | sem_params.u.nsems = nsems; | |
1da177e4 | 534 | |
7748dbfa | 535 | return ipcget(ns, &sem_ids(ns), &sem_ops, &sem_params); |
1da177e4 LT |
536 | } |
537 | ||
1da177e4 LT |
538 | /* |
539 | * Determine whether a sequence of semaphore operations would succeed | |
540 | * all at once. Return 0 if yes, 1 if need to sleep, else return error code. | |
541 | */ | |
542 | ||
543 | static int try_atomic_semop (struct sem_array * sma, struct sembuf * sops, | |
544 | int nsops, struct sem_undo *un, int pid) | |
545 | { | |
546 | int result, sem_op; | |
547 | struct sembuf *sop; | |
548 | struct sem * curr; | |
549 | ||
550 | for (sop = sops; sop < sops + nsops; sop++) { | |
551 | curr = sma->sem_base + sop->sem_num; | |
552 | sem_op = sop->sem_op; | |
553 | result = curr->semval; | |
554 | ||
555 | if (!sem_op && result) | |
556 | goto would_block; | |
557 | ||
558 | result += sem_op; | |
559 | if (result < 0) | |
560 | goto would_block; | |
561 | if (result > SEMVMX) | |
562 | goto out_of_range; | |
563 | if (sop->sem_flg & SEM_UNDO) { | |
564 | int undo = un->semadj[sop->sem_num] - sem_op; | |
565 | /* | |
566 | * Exceeding the undo range is an error. | |
567 | */ | |
568 | if (undo < (-SEMAEM - 1) || undo > SEMAEM) | |
569 | goto out_of_range; | |
570 | } | |
571 | curr->semval = result; | |
572 | } | |
573 | ||
574 | sop--; | |
575 | while (sop >= sops) { | |
576 | sma->sem_base[sop->sem_num].sempid = pid; | |
577 | if (sop->sem_flg & SEM_UNDO) | |
578 | un->semadj[sop->sem_num] -= sop->sem_op; | |
579 | sop--; | |
580 | } | |
581 | ||
1da177e4 LT |
582 | return 0; |
583 | ||
584 | out_of_range: | |
585 | result = -ERANGE; | |
586 | goto undo; | |
587 | ||
588 | would_block: | |
589 | if (sop->sem_flg & IPC_NOWAIT) | |
590 | result = -EAGAIN; | |
591 | else | |
592 | result = 1; | |
593 | ||
594 | undo: | |
595 | sop--; | |
596 | while (sop >= sops) { | |
597 | sma->sem_base[sop->sem_num].semval -= sop->sem_op; | |
598 | sop--; | |
599 | } | |
600 | ||
601 | return result; | |
602 | } | |
603 | ||
0a2b9d4c MS |
604 | /** wake_up_sem_queue_prepare(q, error): Prepare wake-up |
605 | * @q: queue entry that must be signaled | |
606 | * @error: Error value for the signal | |
607 | * | |
608 | * Prepare the wake-up of the queue entry q. | |
d4212093 | 609 | */ |
0a2b9d4c MS |
610 | static void wake_up_sem_queue_prepare(struct list_head *pt, |
611 | struct sem_queue *q, int error) | |
d4212093 | 612 | { |
0a2b9d4c MS |
613 | if (list_empty(pt)) { |
614 | /* | |
615 | * Hold preempt off so that we don't get preempted and have the | |
616 | * wakee busy-wait until we're scheduled back on. | |
617 | */ | |
618 | preempt_disable(); | |
619 | } | |
d4212093 | 620 | q->status = IN_WAKEUP; |
0a2b9d4c MS |
621 | q->pid = error; |
622 | ||
9f1bc2c9 | 623 | list_add_tail(&q->list, pt); |
0a2b9d4c MS |
624 | } |
625 | ||
626 | /** | |
627 | * wake_up_sem_queue_do(pt) - do the actual wake-up | |
628 | * @pt: list of tasks to be woken up | |
629 | * | |
630 | * Do the actual wake-up. | |
631 | * The function is called without any locks held, thus the semaphore array | |
632 | * could be destroyed already and the tasks can disappear as soon as the | |
633 | * status is set to the actual return code. | |
634 | */ | |
635 | static void wake_up_sem_queue_do(struct list_head *pt) | |
636 | { | |
637 | struct sem_queue *q, *t; | |
638 | int did_something; | |
639 | ||
640 | did_something = !list_empty(pt); | |
9f1bc2c9 | 641 | list_for_each_entry_safe(q, t, pt, list) { |
0a2b9d4c MS |
642 | wake_up_process(q->sleeper); |
643 | /* q can disappear immediately after writing q->status. */ | |
644 | smp_wmb(); | |
645 | q->status = q->pid; | |
646 | } | |
647 | if (did_something) | |
648 | preempt_enable(); | |
d4212093 NP |
649 | } |
650 | ||
b97e820f MS |
651 | static void unlink_queue(struct sem_array *sma, struct sem_queue *q) |
652 | { | |
653 | list_del(&q->list); | |
9f1bc2c9 | 654 | if (q->nsops > 1) |
b97e820f MS |
655 | sma->complex_count--; |
656 | } | |
657 | ||
fd5db422 MS |
658 | /** check_restart(sma, q) |
659 | * @sma: semaphore array | |
660 | * @q: the operation that just completed | |
661 | * | |
662 | * update_queue is O(N^2) when it restarts scanning the whole queue of | |
663 | * waiting operations. Therefore this function checks if the restart is | |
664 | * really necessary. It is called after a previously waiting operation | |
1a82e9e1 MS |
665 | * modified the array. |
666 | * Note that wait-for-zero operations are handled without restart. | |
fd5db422 MS |
667 | */ |
668 | static int check_restart(struct sem_array *sma, struct sem_queue *q) | |
669 | { | |
1a82e9e1 MS |
670 | /* pending complex alter operations are too difficult to analyse */ |
671 | if (!list_empty(&sma->pending_alter)) | |
fd5db422 MS |
672 | return 1; |
673 | ||
674 | /* we were a sleeping complex operation. Too difficult */ | |
675 | if (q->nsops > 1) | |
676 | return 1; | |
677 | ||
1a82e9e1 MS |
678 | /* It is impossible that someone waits for the new value: |
679 | * - complex operations always restart. | |
680 | * - wait-for-zero are handled seperately. | |
681 | * - q is a previously sleeping simple operation that | |
682 | * altered the array. It must be a decrement, because | |
683 | * simple increments never sleep. | |
684 | * - If there are older (higher priority) decrements | |
685 | * in the queue, then they have observed the original | |
686 | * semval value and couldn't proceed. The operation | |
687 | * decremented to value - thus they won't proceed either. | |
688 | */ | |
689 | return 0; | |
690 | } | |
fd5db422 | 691 | |
1a82e9e1 MS |
692 | /** |
693 | * wake_const_ops(sma, semnum, pt) - Wake up non-alter tasks | |
694 | * @sma: semaphore array. | |
695 | * @semnum: semaphore that was modified. | |
696 | * @pt: list head for the tasks that must be woken up. | |
697 | * | |
698 | * wake_const_ops must be called after a semaphore in a semaphore array | |
699 | * was set to 0. If complex const operations are pending, wake_const_ops must | |
700 | * be called with semnum = -1, as well as with the number of each modified | |
701 | * semaphore. | |
702 | * The tasks that must be woken up are added to @pt. The return code | |
703 | * is stored in q->pid. | |
704 | * The function returns 1 if at least one operation was completed successfully. | |
705 | */ | |
706 | static int wake_const_ops(struct sem_array *sma, int semnum, | |
707 | struct list_head *pt) | |
708 | { | |
709 | struct sem_queue *q; | |
710 | struct list_head *walk; | |
711 | struct list_head *pending_list; | |
712 | int semop_completed = 0; | |
713 | ||
714 | if (semnum == -1) | |
715 | pending_list = &sma->pending_const; | |
716 | else | |
717 | pending_list = &sma->sem_base[semnum].pending_const; | |
fd5db422 | 718 | |
1a82e9e1 MS |
719 | walk = pending_list->next; |
720 | while (walk != pending_list) { | |
721 | int error; | |
722 | ||
723 | q = container_of(walk, struct sem_queue, list); | |
724 | walk = walk->next; | |
725 | ||
726 | error = try_atomic_semop(sma, q->sops, q->nsops, | |
727 | q->undo, q->pid); | |
728 | ||
729 | if (error <= 0) { | |
730 | /* operation completed, remove from queue & wakeup */ | |
731 | ||
732 | unlink_queue(sma, q); | |
733 | ||
734 | wake_up_sem_queue_prepare(pt, q, error); | |
735 | if (error == 0) | |
736 | semop_completed = 1; | |
737 | } | |
738 | } | |
739 | return semop_completed; | |
740 | } | |
741 | ||
742 | /** | |
743 | * do_smart_wakeup_zero(sma, sops, nsops, pt) - wakeup all wait for zero tasks | |
744 | * @sma: semaphore array | |
745 | * @sops: operations that were performed | |
746 | * @nsops: number of operations | |
747 | * @pt: list head of the tasks that must be woken up. | |
748 | * | |
749 | * do_smart_wakeup_zero() checks all required queue for wait-for-zero | |
750 | * operations, based on the actual changes that were performed on the | |
751 | * semaphore array. | |
752 | * The function returns 1 if at least one operation was completed successfully. | |
753 | */ | |
754 | static int do_smart_wakeup_zero(struct sem_array *sma, struct sembuf *sops, | |
755 | int nsops, struct list_head *pt) | |
756 | { | |
757 | int i; | |
758 | int semop_completed = 0; | |
759 | int got_zero = 0; | |
760 | ||
761 | /* first: the per-semaphore queues, if known */ | |
762 | if (sops) { | |
763 | for (i = 0; i < nsops; i++) { | |
764 | int num = sops[i].sem_num; | |
765 | ||
766 | if (sma->sem_base[num].semval == 0) { | |
767 | got_zero = 1; | |
768 | semop_completed |= wake_const_ops(sma, num, pt); | |
769 | } | |
770 | } | |
771 | } else { | |
772 | /* | |
773 | * No sops means modified semaphores not known. | |
774 | * Assume all were changed. | |
fd5db422 | 775 | */ |
1a82e9e1 MS |
776 | for (i = 0; i < sma->sem_nsems; i++) { |
777 | if (sma->sem_base[i].semval == 0) { | |
778 | got_zero = 1; | |
779 | semop_completed |= wake_const_ops(sma, i, pt); | |
780 | } | |
781 | } | |
fd5db422 MS |
782 | } |
783 | /* | |
1a82e9e1 MS |
784 | * If one of the modified semaphores got 0, |
785 | * then check the global queue, too. | |
fd5db422 | 786 | */ |
1a82e9e1 MS |
787 | if (got_zero) |
788 | semop_completed |= wake_const_ops(sma, -1, pt); | |
fd5db422 | 789 | |
1a82e9e1 | 790 | return semop_completed; |
fd5db422 MS |
791 | } |
792 | ||
636c6be8 MS |
793 | |
794 | /** | |
795 | * update_queue(sma, semnum): Look for tasks that can be completed. | |
796 | * @sma: semaphore array. | |
797 | * @semnum: semaphore that was modified. | |
0a2b9d4c | 798 | * @pt: list head for the tasks that must be woken up. |
636c6be8 MS |
799 | * |
800 | * update_queue must be called after a semaphore in a semaphore array | |
9f1bc2c9 RR |
801 | * was modified. If multiple semaphores were modified, update_queue must |
802 | * be called with semnum = -1, as well as with the number of each modified | |
803 | * semaphore. | |
0a2b9d4c MS |
804 | * The tasks that must be woken up are added to @pt. The return code |
805 | * is stored in q->pid. | |
1a82e9e1 MS |
806 | * The function internally checks if const operations can now succeed. |
807 | * | |
0a2b9d4c | 808 | * The function return 1 if at least one semop was completed successfully. |
1da177e4 | 809 | */ |
0a2b9d4c | 810 | static int update_queue(struct sem_array *sma, int semnum, struct list_head *pt) |
1da177e4 | 811 | { |
636c6be8 MS |
812 | struct sem_queue *q; |
813 | struct list_head *walk; | |
814 | struct list_head *pending_list; | |
0a2b9d4c | 815 | int semop_completed = 0; |
636c6be8 | 816 | |
9f1bc2c9 | 817 | if (semnum == -1) |
1a82e9e1 | 818 | pending_list = &sma->pending_alter; |
9f1bc2c9 | 819 | else |
1a82e9e1 | 820 | pending_list = &sma->sem_base[semnum].pending_alter; |
9cad200c NP |
821 | |
822 | again: | |
636c6be8 MS |
823 | walk = pending_list->next; |
824 | while (walk != pending_list) { | |
fd5db422 | 825 | int error, restart; |
636c6be8 | 826 | |
9f1bc2c9 | 827 | q = container_of(walk, struct sem_queue, list); |
636c6be8 | 828 | walk = walk->next; |
1da177e4 | 829 | |
d987f8b2 MS |
830 | /* If we are scanning the single sop, per-semaphore list of |
831 | * one semaphore and that semaphore is 0, then it is not | |
1a82e9e1 | 832 | * necessary to scan further: simple increments |
d987f8b2 MS |
833 | * that affect only one entry succeed immediately and cannot |
834 | * be in the per semaphore pending queue, and decrements | |
835 | * cannot be successful if the value is already 0. | |
836 | */ | |
1a82e9e1 | 837 | if (semnum != -1 && sma->sem_base[semnum].semval == 0) |
d987f8b2 MS |
838 | break; |
839 | ||
1da177e4 LT |
840 | error = try_atomic_semop(sma, q->sops, q->nsops, |
841 | q->undo, q->pid); | |
842 | ||
843 | /* Does q->sleeper still need to sleep? */ | |
9cad200c NP |
844 | if (error > 0) |
845 | continue; | |
846 | ||
b97e820f | 847 | unlink_queue(sma, q); |
9cad200c | 848 | |
0a2b9d4c | 849 | if (error) { |
fd5db422 | 850 | restart = 0; |
0a2b9d4c MS |
851 | } else { |
852 | semop_completed = 1; | |
1a82e9e1 | 853 | do_smart_wakeup_zero(sma, q->sops, q->nsops, pt); |
fd5db422 | 854 | restart = check_restart(sma, q); |
0a2b9d4c | 855 | } |
fd5db422 | 856 | |
0a2b9d4c | 857 | wake_up_sem_queue_prepare(pt, q, error); |
fd5db422 | 858 | if (restart) |
9cad200c | 859 | goto again; |
1da177e4 | 860 | } |
0a2b9d4c | 861 | return semop_completed; |
1da177e4 LT |
862 | } |
863 | ||
0a2b9d4c MS |
864 | /** |
865 | * do_smart_update(sma, sops, nsops, otime, pt) - optimized update_queue | |
fd5db422 MS |
866 | * @sma: semaphore array |
867 | * @sops: operations that were performed | |
868 | * @nsops: number of operations | |
0a2b9d4c MS |
869 | * @otime: force setting otime |
870 | * @pt: list head of the tasks that must be woken up. | |
fd5db422 | 871 | * |
1a82e9e1 MS |
872 | * do_smart_update() does the required calls to update_queue and wakeup_zero, |
873 | * based on the actual changes that were performed on the semaphore array. | |
0a2b9d4c MS |
874 | * Note that the function does not do the actual wake-up: the caller is |
875 | * responsible for calling wake_up_sem_queue_do(@pt). | |
876 | * It is safe to perform this call after dropping all locks. | |
fd5db422 | 877 | */ |
0a2b9d4c MS |
878 | static void do_smart_update(struct sem_array *sma, struct sembuf *sops, int nsops, |
879 | int otime, struct list_head *pt) | |
fd5db422 MS |
880 | { |
881 | int i; | |
882 | ||
1a82e9e1 MS |
883 | otime |= do_smart_wakeup_zero(sma, sops, nsops, pt); |
884 | ||
f269f40a MS |
885 | if (!list_empty(&sma->pending_alter)) { |
886 | /* semaphore array uses the global queue - just process it. */ | |
887 | otime |= update_queue(sma, -1, pt); | |
888 | } else { | |
889 | if (!sops) { | |
890 | /* | |
891 | * No sops, thus the modified semaphores are not | |
892 | * known. Check all. | |
893 | */ | |
894 | for (i = 0; i < sma->sem_nsems; i++) | |
895 | otime |= update_queue(sma, i, pt); | |
896 | } else { | |
897 | /* | |
898 | * Check the semaphores that were increased: | |
899 | * - No complex ops, thus all sleeping ops are | |
900 | * decrease. | |
901 | * - if we decreased the value, then any sleeping | |
902 | * semaphore ops wont be able to run: If the | |
903 | * previous value was too small, then the new | |
904 | * value will be too small, too. | |
905 | */ | |
906 | for (i = 0; i < nsops; i++) { | |
907 | if (sops[i].sem_op > 0) { | |
908 | otime |= update_queue(sma, | |
909 | sops[i].sem_num, pt); | |
910 | } | |
ab465df9 | 911 | } |
9f1bc2c9 | 912 | } |
fd5db422 | 913 | } |
0a2b9d4c MS |
914 | if (otime) |
915 | sma->sem_otime = get_seconds(); | |
fd5db422 MS |
916 | } |
917 | ||
918 | ||
1da177e4 LT |
919 | /* The following counts are associated to each semaphore: |
920 | * semncnt number of tasks waiting on semval being nonzero | |
921 | * semzcnt number of tasks waiting on semval being zero | |
922 | * This model assumes that a task waits on exactly one semaphore. | |
923 | * Since semaphore operations are to be performed atomically, tasks actually | |
924 | * wait on a whole sequence of semaphores simultaneously. | |
925 | * The counts we return here are a rough approximation, but still | |
926 | * warrant that semncnt+semzcnt>0 if the task is on the pending queue. | |
927 | */ | |
928 | static int count_semncnt (struct sem_array * sma, ushort semnum) | |
929 | { | |
930 | int semncnt; | |
931 | struct sem_queue * q; | |
932 | ||
933 | semncnt = 0; | |
1a82e9e1 | 934 | list_for_each_entry(q, &sma->sem_base[semnum].pending_alter, list) { |
de2657f9 RR |
935 | struct sembuf * sops = q->sops; |
936 | BUG_ON(sops->sem_num != semnum); | |
937 | if ((sops->sem_op < 0) && !(sops->sem_flg & IPC_NOWAIT)) | |
938 | semncnt++; | |
939 | } | |
940 | ||
1a82e9e1 | 941 | list_for_each_entry(q, &sma->pending_alter, list) { |
1da177e4 LT |
942 | struct sembuf * sops = q->sops; |
943 | int nsops = q->nsops; | |
944 | int i; | |
945 | for (i = 0; i < nsops; i++) | |
946 | if (sops[i].sem_num == semnum | |
947 | && (sops[i].sem_op < 0) | |
948 | && !(sops[i].sem_flg & IPC_NOWAIT)) | |
949 | semncnt++; | |
950 | } | |
951 | return semncnt; | |
952 | } | |
a1193f8e | 953 | |
1da177e4 LT |
954 | static int count_semzcnt (struct sem_array * sma, ushort semnum) |
955 | { | |
956 | int semzcnt; | |
957 | struct sem_queue * q; | |
958 | ||
959 | semzcnt = 0; | |
1a82e9e1 | 960 | list_for_each_entry(q, &sma->sem_base[semnum].pending_const, list) { |
ebc2e5e6 RR |
961 | struct sembuf * sops = q->sops; |
962 | BUG_ON(sops->sem_num != semnum); | |
963 | if ((sops->sem_op == 0) && !(sops->sem_flg & IPC_NOWAIT)) | |
964 | semzcnt++; | |
965 | } | |
966 | ||
1a82e9e1 | 967 | list_for_each_entry(q, &sma->pending_const, list) { |
1da177e4 LT |
968 | struct sembuf * sops = q->sops; |
969 | int nsops = q->nsops; | |
970 | int i; | |
971 | for (i = 0; i < nsops; i++) | |
972 | if (sops[i].sem_num == semnum | |
973 | && (sops[i].sem_op == 0) | |
974 | && !(sops[i].sem_flg & IPC_NOWAIT)) | |
975 | semzcnt++; | |
976 | } | |
977 | return semzcnt; | |
978 | } | |
979 | ||
3e148c79 ND |
980 | /* Free a semaphore set. freeary() is called with sem_ids.rw_mutex locked |
981 | * as a writer and the spinlock for this semaphore set hold. sem_ids.rw_mutex | |
982 | * remains locked on exit. | |
1da177e4 | 983 | */ |
01b8b07a | 984 | static void freeary(struct ipc_namespace *ns, struct kern_ipc_perm *ipcp) |
1da177e4 | 985 | { |
380af1b3 MS |
986 | struct sem_undo *un, *tu; |
987 | struct sem_queue *q, *tq; | |
01b8b07a | 988 | struct sem_array *sma = container_of(ipcp, struct sem_array, sem_perm); |
0a2b9d4c | 989 | struct list_head tasks; |
9f1bc2c9 | 990 | int i; |
1da177e4 | 991 | |
380af1b3 | 992 | /* Free the existing undo structures for this semaphore set. */ |
cf9d5d78 | 993 | ipc_assert_locked_object(&sma->sem_perm); |
380af1b3 MS |
994 | list_for_each_entry_safe(un, tu, &sma->list_id, list_id) { |
995 | list_del(&un->list_id); | |
996 | spin_lock(&un->ulp->lock); | |
1da177e4 | 997 | un->semid = -1; |
380af1b3 MS |
998 | list_del_rcu(&un->list_proc); |
999 | spin_unlock(&un->ulp->lock); | |
693a8b6e | 1000 | kfree_rcu(un, rcu); |
380af1b3 | 1001 | } |
1da177e4 LT |
1002 | |
1003 | /* Wake up all pending processes and let them fail with EIDRM. */ | |
0a2b9d4c | 1004 | INIT_LIST_HEAD(&tasks); |
1a82e9e1 MS |
1005 | list_for_each_entry_safe(q, tq, &sma->pending_const, list) { |
1006 | unlink_queue(sma, q); | |
1007 | wake_up_sem_queue_prepare(&tasks, q, -EIDRM); | |
1008 | } | |
1009 | ||
1010 | list_for_each_entry_safe(q, tq, &sma->pending_alter, list) { | |
b97e820f | 1011 | unlink_queue(sma, q); |
0a2b9d4c | 1012 | wake_up_sem_queue_prepare(&tasks, q, -EIDRM); |
1da177e4 | 1013 | } |
9f1bc2c9 RR |
1014 | for (i = 0; i < sma->sem_nsems; i++) { |
1015 | struct sem *sem = sma->sem_base + i; | |
1a82e9e1 MS |
1016 | list_for_each_entry_safe(q, tq, &sem->pending_const, list) { |
1017 | unlink_queue(sma, q); | |
1018 | wake_up_sem_queue_prepare(&tasks, q, -EIDRM); | |
1019 | } | |
1020 | list_for_each_entry_safe(q, tq, &sem->pending_alter, list) { | |
9f1bc2c9 RR |
1021 | unlink_queue(sma, q); |
1022 | wake_up_sem_queue_prepare(&tasks, q, -EIDRM); | |
1023 | } | |
1024 | } | |
1da177e4 | 1025 | |
7ca7e564 ND |
1026 | /* Remove the semaphore set from the IDR */ |
1027 | sem_rmid(ns, sma); | |
6062a8dc | 1028 | sem_unlock(sma, -1); |
6d49dab8 | 1029 | rcu_read_unlock(); |
1da177e4 | 1030 | |
0a2b9d4c | 1031 | wake_up_sem_queue_do(&tasks); |
e3893534 | 1032 | ns->used_sems -= sma->sem_nsems; |
1da177e4 LT |
1033 | security_sem_free(sma); |
1034 | ipc_rcu_putref(sma); | |
1035 | } | |
1036 | ||
1037 | static unsigned long copy_semid_to_user(void __user *buf, struct semid64_ds *in, int version) | |
1038 | { | |
1039 | switch(version) { | |
1040 | case IPC_64: | |
1041 | return copy_to_user(buf, in, sizeof(*in)); | |
1042 | case IPC_OLD: | |
1043 | { | |
1044 | struct semid_ds out; | |
1045 | ||
982f7c2b DR |
1046 | memset(&out, 0, sizeof(out)); |
1047 | ||
1da177e4 LT |
1048 | ipc64_perm_to_ipc_perm(&in->sem_perm, &out.sem_perm); |
1049 | ||
1050 | out.sem_otime = in->sem_otime; | |
1051 | out.sem_ctime = in->sem_ctime; | |
1052 | out.sem_nsems = in->sem_nsems; | |
1053 | ||
1054 | return copy_to_user(buf, &out, sizeof(out)); | |
1055 | } | |
1056 | default: | |
1057 | return -EINVAL; | |
1058 | } | |
1059 | } | |
1060 | ||
4b9fcb0e | 1061 | static int semctl_nolock(struct ipc_namespace *ns, int semid, |
e1fd1f49 | 1062 | int cmd, int version, void __user *p) |
1da177e4 | 1063 | { |
e5cc9c7b | 1064 | int err; |
1da177e4 LT |
1065 | struct sem_array *sma; |
1066 | ||
1067 | switch(cmd) { | |
1068 | case IPC_INFO: | |
1069 | case SEM_INFO: | |
1070 | { | |
1071 | struct seminfo seminfo; | |
1072 | int max_id; | |
1073 | ||
1074 | err = security_sem_semctl(NULL, cmd); | |
1075 | if (err) | |
1076 | return err; | |
1077 | ||
1078 | memset(&seminfo,0,sizeof(seminfo)); | |
e3893534 KK |
1079 | seminfo.semmni = ns->sc_semmni; |
1080 | seminfo.semmns = ns->sc_semmns; | |
1081 | seminfo.semmsl = ns->sc_semmsl; | |
1082 | seminfo.semopm = ns->sc_semopm; | |
1da177e4 LT |
1083 | seminfo.semvmx = SEMVMX; |
1084 | seminfo.semmnu = SEMMNU; | |
1085 | seminfo.semmap = SEMMAP; | |
1086 | seminfo.semume = SEMUME; | |
3e148c79 | 1087 | down_read(&sem_ids(ns).rw_mutex); |
1da177e4 | 1088 | if (cmd == SEM_INFO) { |
e3893534 KK |
1089 | seminfo.semusz = sem_ids(ns).in_use; |
1090 | seminfo.semaem = ns->used_sems; | |
1da177e4 LT |
1091 | } else { |
1092 | seminfo.semusz = SEMUSZ; | |
1093 | seminfo.semaem = SEMAEM; | |
1094 | } | |
7ca7e564 | 1095 | max_id = ipc_get_maxid(&sem_ids(ns)); |
3e148c79 | 1096 | up_read(&sem_ids(ns).rw_mutex); |
e1fd1f49 | 1097 | if (copy_to_user(p, &seminfo, sizeof(struct seminfo))) |
1da177e4 LT |
1098 | return -EFAULT; |
1099 | return (max_id < 0) ? 0: max_id; | |
1100 | } | |
4b9fcb0e | 1101 | case IPC_STAT: |
1da177e4 LT |
1102 | case SEM_STAT: |
1103 | { | |
1104 | struct semid64_ds tbuf; | |
16df3674 DB |
1105 | int id = 0; |
1106 | ||
1107 | memset(&tbuf, 0, sizeof(tbuf)); | |
1da177e4 | 1108 | |
941b0304 | 1109 | rcu_read_lock(); |
4b9fcb0e | 1110 | if (cmd == SEM_STAT) { |
16df3674 DB |
1111 | sma = sem_obtain_object(ns, semid); |
1112 | if (IS_ERR(sma)) { | |
1113 | err = PTR_ERR(sma); | |
1114 | goto out_unlock; | |
1115 | } | |
4b9fcb0e PP |
1116 | id = sma->sem_perm.id; |
1117 | } else { | |
16df3674 DB |
1118 | sma = sem_obtain_object_check(ns, semid); |
1119 | if (IS_ERR(sma)) { | |
1120 | err = PTR_ERR(sma); | |
1121 | goto out_unlock; | |
1122 | } | |
4b9fcb0e | 1123 | } |
1da177e4 LT |
1124 | |
1125 | err = -EACCES; | |
b0e77598 | 1126 | if (ipcperms(ns, &sma->sem_perm, S_IRUGO)) |
1da177e4 LT |
1127 | goto out_unlock; |
1128 | ||
1129 | err = security_sem_semctl(sma, cmd); | |
1130 | if (err) | |
1131 | goto out_unlock; | |
1132 | ||
1da177e4 LT |
1133 | kernel_to_ipc64_perm(&sma->sem_perm, &tbuf.sem_perm); |
1134 | tbuf.sem_otime = sma->sem_otime; | |
1135 | tbuf.sem_ctime = sma->sem_ctime; | |
1136 | tbuf.sem_nsems = sma->sem_nsems; | |
16df3674 | 1137 | rcu_read_unlock(); |
e1fd1f49 | 1138 | if (copy_semid_to_user(p, &tbuf, version)) |
1da177e4 LT |
1139 | return -EFAULT; |
1140 | return id; | |
1141 | } | |
1142 | default: | |
1143 | return -EINVAL; | |
1144 | } | |
1da177e4 | 1145 | out_unlock: |
16df3674 | 1146 | rcu_read_unlock(); |
1da177e4 LT |
1147 | return err; |
1148 | } | |
1149 | ||
e1fd1f49 AV |
1150 | static int semctl_setval(struct ipc_namespace *ns, int semid, int semnum, |
1151 | unsigned long arg) | |
1152 | { | |
1153 | struct sem_undo *un; | |
1154 | struct sem_array *sma; | |
1155 | struct sem* curr; | |
1156 | int err; | |
e1fd1f49 AV |
1157 | struct list_head tasks; |
1158 | int val; | |
1159 | #if defined(CONFIG_64BIT) && defined(__BIG_ENDIAN) | |
1160 | /* big-endian 64bit */ | |
1161 | val = arg >> 32; | |
1162 | #else | |
1163 | /* 32bit or little-endian 64bit */ | |
1164 | val = arg; | |
1165 | #endif | |
1166 | ||
6062a8dc RR |
1167 | if (val > SEMVMX || val < 0) |
1168 | return -ERANGE; | |
e1fd1f49 AV |
1169 | |
1170 | INIT_LIST_HEAD(&tasks); | |
e1fd1f49 | 1171 | |
6062a8dc RR |
1172 | rcu_read_lock(); |
1173 | sma = sem_obtain_object_check(ns, semid); | |
1174 | if (IS_ERR(sma)) { | |
1175 | rcu_read_unlock(); | |
1176 | return PTR_ERR(sma); | |
1177 | } | |
1178 | ||
1179 | if (semnum < 0 || semnum >= sma->sem_nsems) { | |
1180 | rcu_read_unlock(); | |
1181 | return -EINVAL; | |
1182 | } | |
1183 | ||
1184 | ||
1185 | if (ipcperms(ns, &sma->sem_perm, S_IWUGO)) { | |
1186 | rcu_read_unlock(); | |
1187 | return -EACCES; | |
1188 | } | |
e1fd1f49 AV |
1189 | |
1190 | err = security_sem_semctl(sma, SETVAL); | |
6062a8dc RR |
1191 | if (err) { |
1192 | rcu_read_unlock(); | |
1193 | return -EACCES; | |
1194 | } | |
e1fd1f49 | 1195 | |
6062a8dc | 1196 | sem_lock(sma, NULL, -1); |
e1fd1f49 AV |
1197 | |
1198 | curr = &sma->sem_base[semnum]; | |
1199 | ||
cf9d5d78 | 1200 | ipc_assert_locked_object(&sma->sem_perm); |
e1fd1f49 AV |
1201 | list_for_each_entry(un, &sma->list_id, list_id) |
1202 | un->semadj[semnum] = 0; | |
1203 | ||
1204 | curr->semval = val; | |
1205 | curr->sempid = task_tgid_vnr(current); | |
1206 | sma->sem_ctime = get_seconds(); | |
1207 | /* maybe some queued-up processes were waiting for this */ | |
1208 | do_smart_update(sma, NULL, 0, 0, &tasks); | |
6062a8dc | 1209 | sem_unlock(sma, -1); |
6d49dab8 | 1210 | rcu_read_unlock(); |
e1fd1f49 | 1211 | wake_up_sem_queue_do(&tasks); |
6062a8dc | 1212 | return 0; |
e1fd1f49 AV |
1213 | } |
1214 | ||
e3893534 | 1215 | static int semctl_main(struct ipc_namespace *ns, int semid, int semnum, |
e1fd1f49 | 1216 | int cmd, void __user *p) |
1da177e4 LT |
1217 | { |
1218 | struct sem_array *sma; | |
1219 | struct sem* curr; | |
16df3674 | 1220 | int err, nsems; |
1da177e4 LT |
1221 | ushort fast_sem_io[SEMMSL_FAST]; |
1222 | ushort* sem_io = fast_sem_io; | |
0a2b9d4c | 1223 | struct list_head tasks; |
1da177e4 | 1224 | |
16df3674 DB |
1225 | INIT_LIST_HEAD(&tasks); |
1226 | ||
1227 | rcu_read_lock(); | |
1228 | sma = sem_obtain_object_check(ns, semid); | |
1229 | if (IS_ERR(sma)) { | |
1230 | rcu_read_unlock(); | |
023a5355 | 1231 | return PTR_ERR(sma); |
16df3674 | 1232 | } |
1da177e4 LT |
1233 | |
1234 | nsems = sma->sem_nsems; | |
1235 | ||
1da177e4 | 1236 | err = -EACCES; |
c728b9c8 LT |
1237 | if (ipcperms(ns, &sma->sem_perm, cmd == SETALL ? S_IWUGO : S_IRUGO)) |
1238 | goto out_rcu_wakeup; | |
1da177e4 LT |
1239 | |
1240 | err = security_sem_semctl(sma, cmd); | |
c728b9c8 LT |
1241 | if (err) |
1242 | goto out_rcu_wakeup; | |
1da177e4 LT |
1243 | |
1244 | err = -EACCES; | |
1245 | switch (cmd) { | |
1246 | case GETALL: | |
1247 | { | |
e1fd1f49 | 1248 | ushort __user *array = p; |
1da177e4 LT |
1249 | int i; |
1250 | ||
ce857229 | 1251 | sem_lock(sma, NULL, -1); |
1da177e4 | 1252 | if(nsems > SEMMSL_FAST) { |
ce857229 AV |
1253 | if (!ipc_rcu_getref(sma)) { |
1254 | sem_unlock(sma, -1); | |
6d49dab8 | 1255 | rcu_read_unlock(); |
ce857229 AV |
1256 | err = -EIDRM; |
1257 | goto out_free; | |
1258 | } | |
1259 | sem_unlock(sma, -1); | |
6d49dab8 | 1260 | rcu_read_unlock(); |
1da177e4 LT |
1261 | sem_io = ipc_alloc(sizeof(ushort)*nsems); |
1262 | if(sem_io == NULL) { | |
6ff37972 | 1263 | sem_putref(sma); |
1da177e4 LT |
1264 | return -ENOMEM; |
1265 | } | |
1266 | ||
4091fd94 | 1267 | rcu_read_lock(); |
6ff37972 | 1268 | sem_lock_and_putref(sma); |
1da177e4 | 1269 | if (sma->sem_perm.deleted) { |
6062a8dc | 1270 | sem_unlock(sma, -1); |
6d49dab8 | 1271 | rcu_read_unlock(); |
1da177e4 LT |
1272 | err = -EIDRM; |
1273 | goto out_free; | |
1274 | } | |
ce857229 | 1275 | } |
1da177e4 LT |
1276 | for (i = 0; i < sma->sem_nsems; i++) |
1277 | sem_io[i] = sma->sem_base[i].semval; | |
6062a8dc | 1278 | sem_unlock(sma, -1); |
6d49dab8 | 1279 | rcu_read_unlock(); |
1da177e4 LT |
1280 | err = 0; |
1281 | if(copy_to_user(array, sem_io, nsems*sizeof(ushort))) | |
1282 | err = -EFAULT; | |
1283 | goto out_free; | |
1284 | } | |
1285 | case SETALL: | |
1286 | { | |
1287 | int i; | |
1288 | struct sem_undo *un; | |
1289 | ||
6062a8dc RR |
1290 | if (!ipc_rcu_getref(sma)) { |
1291 | rcu_read_unlock(); | |
1292 | return -EIDRM; | |
1293 | } | |
16df3674 | 1294 | rcu_read_unlock(); |
1da177e4 LT |
1295 | |
1296 | if(nsems > SEMMSL_FAST) { | |
1297 | sem_io = ipc_alloc(sizeof(ushort)*nsems); | |
1298 | if(sem_io == NULL) { | |
6ff37972 | 1299 | sem_putref(sma); |
1da177e4 LT |
1300 | return -ENOMEM; |
1301 | } | |
1302 | } | |
1303 | ||
e1fd1f49 | 1304 | if (copy_from_user (sem_io, p, nsems*sizeof(ushort))) { |
6ff37972 | 1305 | sem_putref(sma); |
1da177e4 LT |
1306 | err = -EFAULT; |
1307 | goto out_free; | |
1308 | } | |
1309 | ||
1310 | for (i = 0; i < nsems; i++) { | |
1311 | if (sem_io[i] > SEMVMX) { | |
6ff37972 | 1312 | sem_putref(sma); |
1da177e4 LT |
1313 | err = -ERANGE; |
1314 | goto out_free; | |
1315 | } | |
1316 | } | |
4091fd94 | 1317 | rcu_read_lock(); |
6ff37972 | 1318 | sem_lock_and_putref(sma); |
1da177e4 | 1319 | if (sma->sem_perm.deleted) { |
6062a8dc | 1320 | sem_unlock(sma, -1); |
6d49dab8 | 1321 | rcu_read_unlock(); |
1da177e4 LT |
1322 | err = -EIDRM; |
1323 | goto out_free; | |
1324 | } | |
1325 | ||
1326 | for (i = 0; i < nsems; i++) | |
1327 | sma->sem_base[i].semval = sem_io[i]; | |
4daa28f6 | 1328 | |
cf9d5d78 | 1329 | ipc_assert_locked_object(&sma->sem_perm); |
4daa28f6 | 1330 | list_for_each_entry(un, &sma->list_id, list_id) { |
1da177e4 LT |
1331 | for (i = 0; i < nsems; i++) |
1332 | un->semadj[i] = 0; | |
4daa28f6 | 1333 | } |
1da177e4 LT |
1334 | sma->sem_ctime = get_seconds(); |
1335 | /* maybe some queued-up processes were waiting for this */ | |
0a2b9d4c | 1336 | do_smart_update(sma, NULL, 0, 0, &tasks); |
1da177e4 LT |
1337 | err = 0; |
1338 | goto out_unlock; | |
1339 | } | |
e1fd1f49 | 1340 | /* GETVAL, GETPID, GETNCTN, GETZCNT: fall-through */ |
1da177e4 LT |
1341 | } |
1342 | err = -EINVAL; | |
c728b9c8 LT |
1343 | if (semnum < 0 || semnum >= nsems) |
1344 | goto out_rcu_wakeup; | |
1da177e4 | 1345 | |
6062a8dc | 1346 | sem_lock(sma, NULL, -1); |
1da177e4 LT |
1347 | curr = &sma->sem_base[semnum]; |
1348 | ||
1349 | switch (cmd) { | |
1350 | case GETVAL: | |
1351 | err = curr->semval; | |
1352 | goto out_unlock; | |
1353 | case GETPID: | |
1354 | err = curr->sempid; | |
1355 | goto out_unlock; | |
1356 | case GETNCNT: | |
1357 | err = count_semncnt(sma,semnum); | |
1358 | goto out_unlock; | |
1359 | case GETZCNT: | |
1360 | err = count_semzcnt(sma,semnum); | |
1361 | goto out_unlock; | |
1da177e4 | 1362 | } |
16df3674 | 1363 | |
1da177e4 | 1364 | out_unlock: |
6062a8dc | 1365 | sem_unlock(sma, -1); |
c728b9c8 | 1366 | out_rcu_wakeup: |
6d49dab8 | 1367 | rcu_read_unlock(); |
0a2b9d4c | 1368 | wake_up_sem_queue_do(&tasks); |
1da177e4 LT |
1369 | out_free: |
1370 | if(sem_io != fast_sem_io) | |
1371 | ipc_free(sem_io, sizeof(ushort)*nsems); | |
1372 | return err; | |
1373 | } | |
1374 | ||
016d7132 PP |
1375 | static inline unsigned long |
1376 | copy_semid_from_user(struct semid64_ds *out, void __user *buf, int version) | |
1da177e4 LT |
1377 | { |
1378 | switch(version) { | |
1379 | case IPC_64: | |
016d7132 | 1380 | if (copy_from_user(out, buf, sizeof(*out))) |
1da177e4 | 1381 | return -EFAULT; |
1da177e4 | 1382 | return 0; |
1da177e4 LT |
1383 | case IPC_OLD: |
1384 | { | |
1385 | struct semid_ds tbuf_old; | |
1386 | ||
1387 | if(copy_from_user(&tbuf_old, buf, sizeof(tbuf_old))) | |
1388 | return -EFAULT; | |
1389 | ||
016d7132 PP |
1390 | out->sem_perm.uid = tbuf_old.sem_perm.uid; |
1391 | out->sem_perm.gid = tbuf_old.sem_perm.gid; | |
1392 | out->sem_perm.mode = tbuf_old.sem_perm.mode; | |
1da177e4 LT |
1393 | |
1394 | return 0; | |
1395 | } | |
1396 | default: | |
1397 | return -EINVAL; | |
1398 | } | |
1399 | } | |
1400 | ||
522bb2a2 PP |
1401 | /* |
1402 | * This function handles some semctl commands which require the rw_mutex | |
1403 | * to be held in write mode. | |
1404 | * NOTE: no locks must be held, the rw_mutex is taken inside this function. | |
1405 | */ | |
21a4826a | 1406 | static int semctl_down(struct ipc_namespace *ns, int semid, |
e1fd1f49 | 1407 | int cmd, int version, void __user *p) |
1da177e4 LT |
1408 | { |
1409 | struct sem_array *sma; | |
1410 | int err; | |
016d7132 | 1411 | struct semid64_ds semid64; |
1da177e4 LT |
1412 | struct kern_ipc_perm *ipcp; |
1413 | ||
1414 | if(cmd == IPC_SET) { | |
e1fd1f49 | 1415 | if (copy_semid_from_user(&semid64, p, version)) |
1da177e4 | 1416 | return -EFAULT; |
1da177e4 | 1417 | } |
073115d6 | 1418 | |
7b4cc5d8 DB |
1419 | down_write(&sem_ids(ns).rw_mutex); |
1420 | rcu_read_lock(); | |
1421 | ||
16df3674 DB |
1422 | ipcp = ipcctl_pre_down_nolock(ns, &sem_ids(ns), semid, cmd, |
1423 | &semid64.sem_perm, 0); | |
7b4cc5d8 DB |
1424 | if (IS_ERR(ipcp)) { |
1425 | err = PTR_ERR(ipcp); | |
7b4cc5d8 DB |
1426 | goto out_unlock1; |
1427 | } | |
073115d6 | 1428 | |
a5f75e7f | 1429 | sma = container_of(ipcp, struct sem_array, sem_perm); |
1da177e4 LT |
1430 | |
1431 | err = security_sem_semctl(sma, cmd); | |
7b4cc5d8 DB |
1432 | if (err) |
1433 | goto out_unlock1; | |
1da177e4 | 1434 | |
7b4cc5d8 | 1435 | switch (cmd) { |
1da177e4 | 1436 | case IPC_RMID: |
6062a8dc | 1437 | sem_lock(sma, NULL, -1); |
7b4cc5d8 | 1438 | /* freeary unlocks the ipc object and rcu */ |
01b8b07a | 1439 | freeary(ns, ipcp); |
522bb2a2 | 1440 | goto out_up; |
1da177e4 | 1441 | case IPC_SET: |
6062a8dc | 1442 | sem_lock(sma, NULL, -1); |
1efdb69b EB |
1443 | err = ipc_update_perm(&semid64.sem_perm, ipcp); |
1444 | if (err) | |
7b4cc5d8 | 1445 | goto out_unlock0; |
1da177e4 | 1446 | sma->sem_ctime = get_seconds(); |
1da177e4 LT |
1447 | break; |
1448 | default: | |
1da177e4 | 1449 | err = -EINVAL; |
7b4cc5d8 | 1450 | goto out_unlock1; |
1da177e4 | 1451 | } |
1da177e4 | 1452 | |
7b4cc5d8 | 1453 | out_unlock0: |
6062a8dc | 1454 | sem_unlock(sma, -1); |
7b4cc5d8 | 1455 | out_unlock1: |
6d49dab8 | 1456 | rcu_read_unlock(); |
522bb2a2 PP |
1457 | out_up: |
1458 | up_write(&sem_ids(ns).rw_mutex); | |
1da177e4 LT |
1459 | return err; |
1460 | } | |
1461 | ||
e1fd1f49 | 1462 | SYSCALL_DEFINE4(semctl, int, semid, int, semnum, int, cmd, unsigned long, arg) |
1da177e4 | 1463 | { |
1da177e4 | 1464 | int version; |
e3893534 | 1465 | struct ipc_namespace *ns; |
e1fd1f49 | 1466 | void __user *p = (void __user *)arg; |
1da177e4 LT |
1467 | |
1468 | if (semid < 0) | |
1469 | return -EINVAL; | |
1470 | ||
1471 | version = ipc_parse_version(&cmd); | |
e3893534 | 1472 | ns = current->nsproxy->ipc_ns; |
1da177e4 LT |
1473 | |
1474 | switch(cmd) { | |
1475 | case IPC_INFO: | |
1476 | case SEM_INFO: | |
4b9fcb0e | 1477 | case IPC_STAT: |
1da177e4 | 1478 | case SEM_STAT: |
e1fd1f49 | 1479 | return semctl_nolock(ns, semid, cmd, version, p); |
1da177e4 LT |
1480 | case GETALL: |
1481 | case GETVAL: | |
1482 | case GETPID: | |
1483 | case GETNCNT: | |
1484 | case GETZCNT: | |
1da177e4 | 1485 | case SETALL: |
e1fd1f49 AV |
1486 | return semctl_main(ns, semid, semnum, cmd, p); |
1487 | case SETVAL: | |
1488 | return semctl_setval(ns, semid, semnum, arg); | |
1da177e4 LT |
1489 | case IPC_RMID: |
1490 | case IPC_SET: | |
e1fd1f49 | 1491 | return semctl_down(ns, semid, cmd, version, p); |
1da177e4 LT |
1492 | default: |
1493 | return -EINVAL; | |
1494 | } | |
1495 | } | |
1496 | ||
1da177e4 LT |
1497 | /* If the task doesn't already have a undo_list, then allocate one |
1498 | * here. We guarantee there is only one thread using this undo list, | |
1499 | * and current is THE ONE | |
1500 | * | |
1501 | * If this allocation and assignment succeeds, but later | |
1502 | * portions of this code fail, there is no need to free the sem_undo_list. | |
1503 | * Just let it stay associated with the task, and it'll be freed later | |
1504 | * at exit time. | |
1505 | * | |
1506 | * This can block, so callers must hold no locks. | |
1507 | */ | |
1508 | static inline int get_undo_list(struct sem_undo_list **undo_listp) | |
1509 | { | |
1510 | struct sem_undo_list *undo_list; | |
1da177e4 LT |
1511 | |
1512 | undo_list = current->sysvsem.undo_list; | |
1513 | if (!undo_list) { | |
2453a306 | 1514 | undo_list = kzalloc(sizeof(*undo_list), GFP_KERNEL); |
1da177e4 LT |
1515 | if (undo_list == NULL) |
1516 | return -ENOMEM; | |
00a5dfdb | 1517 | spin_lock_init(&undo_list->lock); |
1da177e4 | 1518 | atomic_set(&undo_list->refcnt, 1); |
4daa28f6 MS |
1519 | INIT_LIST_HEAD(&undo_list->list_proc); |
1520 | ||
1da177e4 LT |
1521 | current->sysvsem.undo_list = undo_list; |
1522 | } | |
1523 | *undo_listp = undo_list; | |
1524 | return 0; | |
1525 | } | |
1526 | ||
bf17bb71 | 1527 | static struct sem_undo *__lookup_undo(struct sem_undo_list *ulp, int semid) |
1da177e4 | 1528 | { |
bf17bb71 | 1529 | struct sem_undo *un; |
4daa28f6 | 1530 | |
bf17bb71 NP |
1531 | list_for_each_entry_rcu(un, &ulp->list_proc, list_proc) { |
1532 | if (un->semid == semid) | |
1533 | return un; | |
1da177e4 | 1534 | } |
4daa28f6 | 1535 | return NULL; |
1da177e4 LT |
1536 | } |
1537 | ||
bf17bb71 NP |
1538 | static struct sem_undo *lookup_undo(struct sem_undo_list *ulp, int semid) |
1539 | { | |
1540 | struct sem_undo *un; | |
1541 | ||
1542 | assert_spin_locked(&ulp->lock); | |
1543 | ||
1544 | un = __lookup_undo(ulp, semid); | |
1545 | if (un) { | |
1546 | list_del_rcu(&un->list_proc); | |
1547 | list_add_rcu(&un->list_proc, &ulp->list_proc); | |
1548 | } | |
1549 | return un; | |
1550 | } | |
1551 | ||
4daa28f6 MS |
1552 | /** |
1553 | * find_alloc_undo - Lookup (and if not present create) undo array | |
1554 | * @ns: namespace | |
1555 | * @semid: semaphore array id | |
1556 | * | |
1557 | * The function looks up (and if not present creates) the undo structure. | |
1558 | * The size of the undo structure depends on the size of the semaphore | |
1559 | * array, thus the alloc path is not that straightforward. | |
380af1b3 MS |
1560 | * Lifetime-rules: sem_undo is rcu-protected, on success, the function |
1561 | * performs a rcu_read_lock(). | |
4daa28f6 MS |
1562 | */ |
1563 | static struct sem_undo *find_alloc_undo(struct ipc_namespace *ns, int semid) | |
1da177e4 LT |
1564 | { |
1565 | struct sem_array *sma; | |
1566 | struct sem_undo_list *ulp; | |
1567 | struct sem_undo *un, *new; | |
6062a8dc | 1568 | int nsems, error; |
1da177e4 LT |
1569 | |
1570 | error = get_undo_list(&ulp); | |
1571 | if (error) | |
1572 | return ERR_PTR(error); | |
1573 | ||
380af1b3 | 1574 | rcu_read_lock(); |
c530c6ac | 1575 | spin_lock(&ulp->lock); |
1da177e4 | 1576 | un = lookup_undo(ulp, semid); |
c530c6ac | 1577 | spin_unlock(&ulp->lock); |
1da177e4 LT |
1578 | if (likely(un!=NULL)) |
1579 | goto out; | |
1580 | ||
1581 | /* no undo structure around - allocate one. */ | |
4daa28f6 | 1582 | /* step 1: figure out the size of the semaphore array */ |
16df3674 DB |
1583 | sma = sem_obtain_object_check(ns, semid); |
1584 | if (IS_ERR(sma)) { | |
1585 | rcu_read_unlock(); | |
4de85cd6 | 1586 | return ERR_CAST(sma); |
16df3674 | 1587 | } |
023a5355 | 1588 | |
1da177e4 | 1589 | nsems = sma->sem_nsems; |
6062a8dc RR |
1590 | if (!ipc_rcu_getref(sma)) { |
1591 | rcu_read_unlock(); | |
1592 | un = ERR_PTR(-EIDRM); | |
1593 | goto out; | |
1594 | } | |
16df3674 | 1595 | rcu_read_unlock(); |
1da177e4 | 1596 | |
4daa28f6 | 1597 | /* step 2: allocate new undo structure */ |
4668edc3 | 1598 | new = kzalloc(sizeof(struct sem_undo) + sizeof(short)*nsems, GFP_KERNEL); |
1da177e4 | 1599 | if (!new) { |
6ff37972 | 1600 | sem_putref(sma); |
1da177e4 LT |
1601 | return ERR_PTR(-ENOMEM); |
1602 | } | |
1da177e4 | 1603 | |
380af1b3 | 1604 | /* step 3: Acquire the lock on semaphore array */ |
4091fd94 | 1605 | rcu_read_lock(); |
6ff37972 | 1606 | sem_lock_and_putref(sma); |
1da177e4 | 1607 | if (sma->sem_perm.deleted) { |
6062a8dc | 1608 | sem_unlock(sma, -1); |
6d49dab8 | 1609 | rcu_read_unlock(); |
1da177e4 LT |
1610 | kfree(new); |
1611 | un = ERR_PTR(-EIDRM); | |
1612 | goto out; | |
1613 | } | |
380af1b3 MS |
1614 | spin_lock(&ulp->lock); |
1615 | ||
1616 | /* | |
1617 | * step 4: check for races: did someone else allocate the undo struct? | |
1618 | */ | |
1619 | un = lookup_undo(ulp, semid); | |
1620 | if (un) { | |
1621 | kfree(new); | |
1622 | goto success; | |
1623 | } | |
4daa28f6 MS |
1624 | /* step 5: initialize & link new undo structure */ |
1625 | new->semadj = (short *) &new[1]; | |
380af1b3 | 1626 | new->ulp = ulp; |
4daa28f6 MS |
1627 | new->semid = semid; |
1628 | assert_spin_locked(&ulp->lock); | |
380af1b3 | 1629 | list_add_rcu(&new->list_proc, &ulp->list_proc); |
cf9d5d78 | 1630 | ipc_assert_locked_object(&sma->sem_perm); |
4daa28f6 | 1631 | list_add(&new->list_id, &sma->list_id); |
380af1b3 | 1632 | un = new; |
4daa28f6 | 1633 | |
380af1b3 | 1634 | success: |
c530c6ac | 1635 | spin_unlock(&ulp->lock); |
6062a8dc | 1636 | sem_unlock(sma, -1); |
1da177e4 LT |
1637 | out: |
1638 | return un; | |
1639 | } | |
1640 | ||
c61284e9 MS |
1641 | |
1642 | /** | |
1643 | * get_queue_result - Retrieve the result code from sem_queue | |
1644 | * @q: Pointer to queue structure | |
1645 | * | |
1646 | * Retrieve the return code from the pending queue. If IN_WAKEUP is found in | |
1647 | * q->status, then we must loop until the value is replaced with the final | |
1648 | * value: This may happen if a task is woken up by an unrelated event (e.g. | |
1649 | * signal) and in parallel the task is woken up by another task because it got | |
1650 | * the requested semaphores. | |
1651 | * | |
1652 | * The function can be called with or without holding the semaphore spinlock. | |
1653 | */ | |
1654 | static int get_queue_result(struct sem_queue *q) | |
1655 | { | |
1656 | int error; | |
1657 | ||
1658 | error = q->status; | |
1659 | while (unlikely(error == IN_WAKEUP)) { | |
1660 | cpu_relax(); | |
1661 | error = q->status; | |
1662 | } | |
1663 | ||
1664 | return error; | |
1665 | } | |
1666 | ||
1667 | ||
d5460c99 HC |
1668 | SYSCALL_DEFINE4(semtimedop, int, semid, struct sembuf __user *, tsops, |
1669 | unsigned, nsops, const struct timespec __user *, timeout) | |
1da177e4 LT |
1670 | { |
1671 | int error = -EINVAL; | |
1672 | struct sem_array *sma; | |
1673 | struct sembuf fast_sops[SEMOPM_FAST]; | |
1674 | struct sembuf* sops = fast_sops, *sop; | |
1675 | struct sem_undo *un; | |
6062a8dc | 1676 | int undos = 0, alter = 0, max, locknum; |
1da177e4 LT |
1677 | struct sem_queue queue; |
1678 | unsigned long jiffies_left = 0; | |
e3893534 | 1679 | struct ipc_namespace *ns; |
0a2b9d4c | 1680 | struct list_head tasks; |
e3893534 KK |
1681 | |
1682 | ns = current->nsproxy->ipc_ns; | |
1da177e4 LT |
1683 | |
1684 | if (nsops < 1 || semid < 0) | |
1685 | return -EINVAL; | |
e3893534 | 1686 | if (nsops > ns->sc_semopm) |
1da177e4 LT |
1687 | return -E2BIG; |
1688 | if(nsops > SEMOPM_FAST) { | |
1689 | sops = kmalloc(sizeof(*sops)*nsops,GFP_KERNEL); | |
1690 | if(sops==NULL) | |
1691 | return -ENOMEM; | |
1692 | } | |
1693 | if (copy_from_user (sops, tsops, nsops * sizeof(*tsops))) { | |
1694 | error=-EFAULT; | |
1695 | goto out_free; | |
1696 | } | |
1697 | if (timeout) { | |
1698 | struct timespec _timeout; | |
1699 | if (copy_from_user(&_timeout, timeout, sizeof(*timeout))) { | |
1700 | error = -EFAULT; | |
1701 | goto out_free; | |
1702 | } | |
1703 | if (_timeout.tv_sec < 0 || _timeout.tv_nsec < 0 || | |
1704 | _timeout.tv_nsec >= 1000000000L) { | |
1705 | error = -EINVAL; | |
1706 | goto out_free; | |
1707 | } | |
1708 | jiffies_left = timespec_to_jiffies(&_timeout); | |
1709 | } | |
1710 | max = 0; | |
1711 | for (sop = sops; sop < sops + nsops; sop++) { | |
1712 | if (sop->sem_num >= max) | |
1713 | max = sop->sem_num; | |
1714 | if (sop->sem_flg & SEM_UNDO) | |
b78755ab MS |
1715 | undos = 1; |
1716 | if (sop->sem_op != 0) | |
1da177e4 LT |
1717 | alter = 1; |
1718 | } | |
1da177e4 | 1719 | |
6062a8dc RR |
1720 | INIT_LIST_HEAD(&tasks); |
1721 | ||
1da177e4 | 1722 | if (undos) { |
6062a8dc | 1723 | /* On success, find_alloc_undo takes the rcu_read_lock */ |
4daa28f6 | 1724 | un = find_alloc_undo(ns, semid); |
1da177e4 LT |
1725 | if (IS_ERR(un)) { |
1726 | error = PTR_ERR(un); | |
1727 | goto out_free; | |
1728 | } | |
6062a8dc | 1729 | } else { |
1da177e4 | 1730 | un = NULL; |
6062a8dc RR |
1731 | rcu_read_lock(); |
1732 | } | |
1da177e4 | 1733 | |
16df3674 | 1734 | sma = sem_obtain_object_check(ns, semid); |
023a5355 | 1735 | if (IS_ERR(sma)) { |
6062a8dc | 1736 | rcu_read_unlock(); |
023a5355 | 1737 | error = PTR_ERR(sma); |
1da177e4 | 1738 | goto out_free; |
023a5355 ND |
1739 | } |
1740 | ||
16df3674 | 1741 | error = -EFBIG; |
c728b9c8 LT |
1742 | if (max >= sma->sem_nsems) |
1743 | goto out_rcu_wakeup; | |
16df3674 DB |
1744 | |
1745 | error = -EACCES; | |
c728b9c8 LT |
1746 | if (ipcperms(ns, &sma->sem_perm, alter ? S_IWUGO : S_IRUGO)) |
1747 | goto out_rcu_wakeup; | |
16df3674 DB |
1748 | |
1749 | error = security_sem_semop(sma, sops, nsops, alter); | |
c728b9c8 LT |
1750 | if (error) |
1751 | goto out_rcu_wakeup; | |
16df3674 | 1752 | |
1da177e4 | 1753 | /* |
4daa28f6 | 1754 | * semid identifiers are not unique - find_alloc_undo may have |
1da177e4 | 1755 | * allocated an undo structure, it was invalidated by an RMID |
4daa28f6 | 1756 | * and now a new array with received the same id. Check and fail. |
25985edc | 1757 | * This case can be detected checking un->semid. The existence of |
380af1b3 | 1758 | * "un" itself is guaranteed by rcu. |
1da177e4 | 1759 | */ |
4daa28f6 | 1760 | error = -EIDRM; |
6062a8dc RR |
1761 | locknum = sem_lock(sma, sops, nsops); |
1762 | if (un && un->semid == -1) | |
1763 | goto out_unlock_free; | |
4daa28f6 | 1764 | |
b488893a | 1765 | error = try_atomic_semop (sma, sops, nsops, un, task_tgid_vnr(current)); |
1da177e4 LT |
1766 | if (error <= 0) { |
1767 | if (alter && error == 0) | |
0a2b9d4c | 1768 | do_smart_update(sma, sops, nsops, 1, &tasks); |
636c6be8 | 1769 | |
1da177e4 LT |
1770 | goto out_unlock_free; |
1771 | } | |
1772 | ||
1773 | /* We need to sleep on this operation, so we put the current | |
1774 | * task into the pending queue and go to sleep. | |
1775 | */ | |
1776 | ||
1da177e4 LT |
1777 | queue.sops = sops; |
1778 | queue.nsops = nsops; | |
1779 | queue.undo = un; | |
b488893a | 1780 | queue.pid = task_tgid_vnr(current); |
1da177e4 | 1781 | queue.alter = alter; |
1da177e4 | 1782 | |
b97e820f MS |
1783 | if (nsops == 1) { |
1784 | struct sem *curr; | |
1785 | curr = &sma->sem_base[sops->sem_num]; | |
1786 | ||
f269f40a MS |
1787 | if (alter) { |
1788 | if (sma->complex_count) { | |
1789 | list_add_tail(&queue.list, | |
1790 | &sma->pending_alter); | |
1791 | } else { | |
1792 | ||
1793 | list_add_tail(&queue.list, | |
1794 | &curr->pending_alter); | |
1795 | } | |
1796 | } else { | |
1a82e9e1 | 1797 | list_add_tail(&queue.list, &curr->pending_const); |
f269f40a | 1798 | } |
b97e820f | 1799 | } else { |
f269f40a MS |
1800 | if (!sma->complex_count) |
1801 | merge_queues(sma); | |
1802 | ||
9f1bc2c9 | 1803 | if (alter) |
1a82e9e1 | 1804 | list_add_tail(&queue.list, &sma->pending_alter); |
9f1bc2c9 | 1805 | else |
1a82e9e1 MS |
1806 | list_add_tail(&queue.list, &sma->pending_const); |
1807 | ||
b97e820f MS |
1808 | sma->complex_count++; |
1809 | } | |
1810 | ||
1da177e4 LT |
1811 | queue.status = -EINTR; |
1812 | queue.sleeper = current; | |
0b0577f6 MS |
1813 | |
1814 | sleep_again: | |
1da177e4 | 1815 | current->state = TASK_INTERRUPTIBLE; |
6062a8dc | 1816 | sem_unlock(sma, locknum); |
6d49dab8 | 1817 | rcu_read_unlock(); |
1da177e4 LT |
1818 | |
1819 | if (timeout) | |
1820 | jiffies_left = schedule_timeout(jiffies_left); | |
1821 | else | |
1822 | schedule(); | |
1823 | ||
c61284e9 | 1824 | error = get_queue_result(&queue); |
1da177e4 LT |
1825 | |
1826 | if (error != -EINTR) { | |
1827 | /* fast path: update_queue already obtained all requested | |
c61284e9 MS |
1828 | * resources. |
1829 | * Perform a smp_mb(): User space could assume that semop() | |
1830 | * is a memory barrier: Without the mb(), the cpu could | |
1831 | * speculatively read in user space stale data that was | |
1832 | * overwritten by the previous owner of the semaphore. | |
1833 | */ | |
1834 | smp_mb(); | |
1835 | ||
1da177e4 LT |
1836 | goto out_free; |
1837 | } | |
1838 | ||
321310ce | 1839 | rcu_read_lock(); |
6062a8dc | 1840 | sma = sem_obtain_lock(ns, semid, sops, nsops, &locknum); |
d694ad62 MS |
1841 | |
1842 | /* | |
1843 | * Wait until it's guaranteed that no wakeup_sem_queue_do() is ongoing. | |
1844 | */ | |
1845 | error = get_queue_result(&queue); | |
1846 | ||
1847 | /* | |
1848 | * Array removed? If yes, leave without sem_unlock(). | |
1849 | */ | |
023a5355 | 1850 | if (IS_ERR(sma)) { |
321310ce | 1851 | rcu_read_unlock(); |
1da177e4 LT |
1852 | goto out_free; |
1853 | } | |
1854 | ||
c61284e9 | 1855 | |
1da177e4 | 1856 | /* |
d694ad62 MS |
1857 | * If queue.status != -EINTR we are woken up by another process. |
1858 | * Leave without unlink_queue(), but with sem_unlock(). | |
1da177e4 | 1859 | */ |
c61284e9 | 1860 | |
1da177e4 LT |
1861 | if (error != -EINTR) { |
1862 | goto out_unlock_free; | |
1863 | } | |
1864 | ||
1865 | /* | |
1866 | * If an interrupt occurred we have to clean up the queue | |
1867 | */ | |
1868 | if (timeout && jiffies_left == 0) | |
1869 | error = -EAGAIN; | |
0b0577f6 MS |
1870 | |
1871 | /* | |
1872 | * If the wakeup was spurious, just retry | |
1873 | */ | |
1874 | if (error == -EINTR && !signal_pending(current)) | |
1875 | goto sleep_again; | |
1876 | ||
b97e820f | 1877 | unlink_queue(sma, &queue); |
1da177e4 LT |
1878 | |
1879 | out_unlock_free: | |
6062a8dc | 1880 | sem_unlock(sma, locknum); |
c728b9c8 | 1881 | out_rcu_wakeup: |
6d49dab8 | 1882 | rcu_read_unlock(); |
0a2b9d4c | 1883 | wake_up_sem_queue_do(&tasks); |
1da177e4 LT |
1884 | out_free: |
1885 | if(sops != fast_sops) | |
1886 | kfree(sops); | |
1887 | return error; | |
1888 | } | |
1889 | ||
d5460c99 HC |
1890 | SYSCALL_DEFINE3(semop, int, semid, struct sembuf __user *, tsops, |
1891 | unsigned, nsops) | |
1da177e4 LT |
1892 | { |
1893 | return sys_semtimedop(semid, tsops, nsops, NULL); | |
1894 | } | |
1895 | ||
1896 | /* If CLONE_SYSVSEM is set, establish sharing of SEM_UNDO state between | |
1897 | * parent and child tasks. | |
1da177e4 LT |
1898 | */ |
1899 | ||
1900 | int copy_semundo(unsigned long clone_flags, struct task_struct *tsk) | |
1901 | { | |
1902 | struct sem_undo_list *undo_list; | |
1903 | int error; | |
1904 | ||
1905 | if (clone_flags & CLONE_SYSVSEM) { | |
1906 | error = get_undo_list(&undo_list); | |
1907 | if (error) | |
1908 | return error; | |
1da177e4 LT |
1909 | atomic_inc(&undo_list->refcnt); |
1910 | tsk->sysvsem.undo_list = undo_list; | |
1911 | } else | |
1912 | tsk->sysvsem.undo_list = NULL; | |
1913 | ||
1914 | return 0; | |
1915 | } | |
1916 | ||
1917 | /* | |
1918 | * add semadj values to semaphores, free undo structures. | |
1919 | * undo structures are not freed when semaphore arrays are destroyed | |
1920 | * so some of them may be out of date. | |
1921 | * IMPLEMENTATION NOTE: There is some confusion over whether the | |
1922 | * set of adjustments that needs to be done should be done in an atomic | |
1923 | * manner or not. That is, if we are attempting to decrement the semval | |
1924 | * should we queue up and wait until we can do so legally? | |
1925 | * The original implementation attempted to do this (queue and wait). | |
1926 | * The current implementation does not do so. The POSIX standard | |
1927 | * and SVID should be consulted to determine what behavior is mandated. | |
1928 | */ | |
1929 | void exit_sem(struct task_struct *tsk) | |
1930 | { | |
4daa28f6 | 1931 | struct sem_undo_list *ulp; |
1da177e4 | 1932 | |
4daa28f6 MS |
1933 | ulp = tsk->sysvsem.undo_list; |
1934 | if (!ulp) | |
1da177e4 | 1935 | return; |
9edff4ab | 1936 | tsk->sysvsem.undo_list = NULL; |
1da177e4 | 1937 | |
4daa28f6 | 1938 | if (!atomic_dec_and_test(&ulp->refcnt)) |
1da177e4 LT |
1939 | return; |
1940 | ||
380af1b3 | 1941 | for (;;) { |
1da177e4 | 1942 | struct sem_array *sma; |
380af1b3 | 1943 | struct sem_undo *un; |
0a2b9d4c | 1944 | struct list_head tasks; |
6062a8dc | 1945 | int semid, i; |
4daa28f6 | 1946 | |
380af1b3 | 1947 | rcu_read_lock(); |
05725f7e JP |
1948 | un = list_entry_rcu(ulp->list_proc.next, |
1949 | struct sem_undo, list_proc); | |
380af1b3 MS |
1950 | if (&un->list_proc == &ulp->list_proc) |
1951 | semid = -1; | |
1952 | else | |
1953 | semid = un->semid; | |
4daa28f6 | 1954 | |
6062a8dc RR |
1955 | if (semid == -1) { |
1956 | rcu_read_unlock(); | |
380af1b3 | 1957 | break; |
6062a8dc | 1958 | } |
1da177e4 | 1959 | |
6062a8dc | 1960 | sma = sem_obtain_object_check(tsk->nsproxy->ipc_ns, un->semid); |
380af1b3 | 1961 | /* exit_sem raced with IPC_RMID, nothing to do */ |
6062a8dc RR |
1962 | if (IS_ERR(sma)) { |
1963 | rcu_read_unlock(); | |
380af1b3 | 1964 | continue; |
6062a8dc | 1965 | } |
1da177e4 | 1966 | |
6062a8dc | 1967 | sem_lock(sma, NULL, -1); |
bf17bb71 | 1968 | un = __lookup_undo(ulp, semid); |
380af1b3 MS |
1969 | if (un == NULL) { |
1970 | /* exit_sem raced with IPC_RMID+semget() that created | |
1971 | * exactly the same semid. Nothing to do. | |
1972 | */ | |
6062a8dc | 1973 | sem_unlock(sma, -1); |
6d49dab8 | 1974 | rcu_read_unlock(); |
380af1b3 MS |
1975 | continue; |
1976 | } | |
1977 | ||
1978 | /* remove un from the linked lists */ | |
cf9d5d78 | 1979 | ipc_assert_locked_object(&sma->sem_perm); |
4daa28f6 MS |
1980 | list_del(&un->list_id); |
1981 | ||
380af1b3 MS |
1982 | spin_lock(&ulp->lock); |
1983 | list_del_rcu(&un->list_proc); | |
1984 | spin_unlock(&ulp->lock); | |
1985 | ||
4daa28f6 MS |
1986 | /* perform adjustments registered in un */ |
1987 | for (i = 0; i < sma->sem_nsems; i++) { | |
5f921ae9 | 1988 | struct sem * semaphore = &sma->sem_base[i]; |
4daa28f6 MS |
1989 | if (un->semadj[i]) { |
1990 | semaphore->semval += un->semadj[i]; | |
1da177e4 LT |
1991 | /* |
1992 | * Range checks of the new semaphore value, | |
1993 | * not defined by sus: | |
1994 | * - Some unices ignore the undo entirely | |
1995 | * (e.g. HP UX 11i 11.22, Tru64 V5.1) | |
1996 | * - some cap the value (e.g. FreeBSD caps | |
1997 | * at 0, but doesn't enforce SEMVMX) | |
1998 | * | |
1999 | * Linux caps the semaphore value, both at 0 | |
2000 | * and at SEMVMX. | |
2001 | * | |
2002 | * Manfred <manfred@colorfullife.com> | |
2003 | */ | |
5f921ae9 IM |
2004 | if (semaphore->semval < 0) |
2005 | semaphore->semval = 0; | |
2006 | if (semaphore->semval > SEMVMX) | |
2007 | semaphore->semval = SEMVMX; | |
b488893a | 2008 | semaphore->sempid = task_tgid_vnr(current); |
1da177e4 LT |
2009 | } |
2010 | } | |
1da177e4 | 2011 | /* maybe some queued-up processes were waiting for this */ |
0a2b9d4c MS |
2012 | INIT_LIST_HEAD(&tasks); |
2013 | do_smart_update(sma, NULL, 0, 1, &tasks); | |
6062a8dc | 2014 | sem_unlock(sma, -1); |
6d49dab8 | 2015 | rcu_read_unlock(); |
0a2b9d4c | 2016 | wake_up_sem_queue_do(&tasks); |
380af1b3 | 2017 | |
693a8b6e | 2018 | kfree_rcu(un, rcu); |
1da177e4 | 2019 | } |
4daa28f6 | 2020 | kfree(ulp); |
1da177e4 LT |
2021 | } |
2022 | ||
2023 | #ifdef CONFIG_PROC_FS | |
19b4946c | 2024 | static int sysvipc_sem_proc_show(struct seq_file *s, void *it) |
1da177e4 | 2025 | { |
1efdb69b | 2026 | struct user_namespace *user_ns = seq_user_ns(s); |
19b4946c MW |
2027 | struct sem_array *sma = it; |
2028 | ||
2029 | return seq_printf(s, | |
b97e820f | 2030 | "%10d %10d %4o %10u %5u %5u %5u %5u %10lu %10lu\n", |
19b4946c | 2031 | sma->sem_perm.key, |
7ca7e564 | 2032 | sma->sem_perm.id, |
19b4946c MW |
2033 | sma->sem_perm.mode, |
2034 | sma->sem_nsems, | |
1efdb69b EB |
2035 | from_kuid_munged(user_ns, sma->sem_perm.uid), |
2036 | from_kgid_munged(user_ns, sma->sem_perm.gid), | |
2037 | from_kuid_munged(user_ns, sma->sem_perm.cuid), | |
2038 | from_kgid_munged(user_ns, sma->sem_perm.cgid), | |
19b4946c MW |
2039 | sma->sem_otime, |
2040 | sma->sem_ctime); | |
1da177e4 LT |
2041 | } |
2042 | #endif |