]>
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*/ | |
d12e1e50 | 102 | time_t sem_otime; /* candidate for sem_otime */ |
f5c936c0 | 103 | } ____cacheline_aligned_in_smp; |
e57940d7 MS |
104 | |
105 | /* One queue for each sleeping process in the system. */ | |
106 | struct sem_queue { | |
e57940d7 MS |
107 | struct list_head list; /* queue of pending operations */ |
108 | struct task_struct *sleeper; /* this process */ | |
109 | struct sem_undo *undo; /* undo structure */ | |
110 | int pid; /* process id of requesting process */ | |
111 | int status; /* completion status of operation */ | |
112 | struct sembuf *sops; /* array of pending operations */ | |
113 | int nsops; /* number of operations */ | |
114 | int alter; /* does *sops alter the array? */ | |
115 | }; | |
116 | ||
117 | /* Each task has a list of undo requests. They are executed automatically | |
118 | * when the process exits. | |
119 | */ | |
120 | struct sem_undo { | |
121 | struct list_head list_proc; /* per-process list: * | |
122 | * all undos from one process | |
123 | * rcu protected */ | |
124 | struct rcu_head rcu; /* rcu struct for sem_undo */ | |
125 | struct sem_undo_list *ulp; /* back ptr to sem_undo_list */ | |
126 | struct list_head list_id; /* per semaphore array list: | |
127 | * all undos for one array */ | |
128 | int semid; /* semaphore set identifier */ | |
129 | short *semadj; /* array of adjustments */ | |
130 | /* one per semaphore */ | |
131 | }; | |
132 | ||
133 | /* sem_undo_list controls shared access to the list of sem_undo structures | |
134 | * that may be shared among all a CLONE_SYSVSEM task group. | |
135 | */ | |
136 | struct sem_undo_list { | |
137 | atomic_t refcnt; | |
138 | spinlock_t lock; | |
139 | struct list_head list_proc; | |
140 | }; | |
141 | ||
142 | ||
ed2ddbf8 | 143 | #define sem_ids(ns) ((ns)->ids[IPC_SEM_IDS]) |
e3893534 | 144 | |
1b531f21 | 145 | #define sem_checkid(sma, semid) ipc_checkid(&sma->sem_perm, semid) |
1da177e4 | 146 | |
7748dbfa | 147 | static int newary(struct ipc_namespace *, struct ipc_params *); |
01b8b07a | 148 | static void freeary(struct ipc_namespace *, struct kern_ipc_perm *); |
1da177e4 | 149 | #ifdef CONFIG_PROC_FS |
19b4946c | 150 | static int sysvipc_sem_proc_show(struct seq_file *s, void *it); |
1da177e4 LT |
151 | #endif |
152 | ||
153 | #define SEMMSL_FAST 256 /* 512 bytes on stack */ | |
154 | #define SEMOPM_FAST 64 /* ~ 372 bytes on stack */ | |
155 | ||
156 | /* | |
157 | * linked list protection: | |
158 | * sem_undo.id_next, | |
1a82e9e1 | 159 | * sem_array.pending{_alter,_cont}, |
1da177e4 LT |
160 | * sem_array.sem_undo: sem_lock() for read/write |
161 | * sem_undo.proc_next: only "current" is allowed to read/write that field. | |
162 | * | |
163 | */ | |
164 | ||
e3893534 KK |
165 | #define sc_semmsl sem_ctls[0] |
166 | #define sc_semmns sem_ctls[1] | |
167 | #define sc_semopm sem_ctls[2] | |
168 | #define sc_semmni sem_ctls[3] | |
169 | ||
ed2ddbf8 | 170 | void sem_init_ns(struct ipc_namespace *ns) |
e3893534 | 171 | { |
e3893534 KK |
172 | ns->sc_semmsl = SEMMSL; |
173 | ns->sc_semmns = SEMMNS; | |
174 | ns->sc_semopm = SEMOPM; | |
175 | ns->sc_semmni = SEMMNI; | |
176 | ns->used_sems = 0; | |
ed2ddbf8 | 177 | ipc_init_ids(&ns->ids[IPC_SEM_IDS]); |
e3893534 KK |
178 | } |
179 | ||
ae5e1b22 | 180 | #ifdef CONFIG_IPC_NS |
e3893534 KK |
181 | void sem_exit_ns(struct ipc_namespace *ns) |
182 | { | |
01b8b07a | 183 | free_ipcs(ns, &sem_ids(ns), freeary); |
7d6feeb2 | 184 | idr_destroy(&ns->ids[IPC_SEM_IDS].ipcs_idr); |
e3893534 | 185 | } |
ae5e1b22 | 186 | #endif |
1da177e4 LT |
187 | |
188 | void __init sem_init (void) | |
189 | { | |
ed2ddbf8 | 190 | sem_init_ns(&init_ipc_ns); |
19b4946c MW |
191 | ipc_init_proc_interface("sysvipc/sem", |
192 | " key semid perms nsems uid gid cuid cgid otime ctime\n", | |
e3893534 | 193 | IPC_SEM_IDS, sysvipc_sem_proc_show); |
1da177e4 LT |
194 | } |
195 | ||
f269f40a MS |
196 | /** |
197 | * unmerge_queues - unmerge queues, if possible. | |
198 | * @sma: semaphore array | |
199 | * | |
200 | * The function unmerges the wait queues if complex_count is 0. | |
201 | * It must be called prior to dropping the global semaphore array lock. | |
202 | */ | |
203 | static void unmerge_queues(struct sem_array *sma) | |
204 | { | |
205 | struct sem_queue *q, *tq; | |
206 | ||
207 | /* complex operations still around? */ | |
208 | if (sma->complex_count) | |
209 | return; | |
210 | /* | |
211 | * We will switch back to simple mode. | |
212 | * Move all pending operation back into the per-semaphore | |
213 | * queues. | |
214 | */ | |
215 | list_for_each_entry_safe(q, tq, &sma->pending_alter, list) { | |
216 | struct sem *curr; | |
217 | curr = &sma->sem_base[q->sops[0].sem_num]; | |
218 | ||
219 | list_add_tail(&q->list, &curr->pending_alter); | |
220 | } | |
221 | INIT_LIST_HEAD(&sma->pending_alter); | |
222 | } | |
223 | ||
224 | /** | |
225 | * merge_queues - Merge single semop queues into global queue | |
226 | * @sma: semaphore array | |
227 | * | |
228 | * This function merges all per-semaphore queues into the global queue. | |
229 | * It is necessary to achieve FIFO ordering for the pending single-sop | |
230 | * operations when a multi-semop operation must sleep. | |
231 | * Only the alter operations must be moved, the const operations can stay. | |
232 | */ | |
233 | static void merge_queues(struct sem_array *sma) | |
234 | { | |
235 | int i; | |
236 | for (i = 0; i < sma->sem_nsems; i++) { | |
237 | struct sem *sem = sma->sem_base + i; | |
238 | ||
239 | list_splice_init(&sem->pending_alter, &sma->pending_alter); | |
240 | } | |
241 | } | |
242 | ||
6062a8dc RR |
243 | /* |
244 | * If the request contains only one semaphore operation, and there are | |
245 | * no complex transactions pending, lock only the semaphore involved. | |
246 | * Otherwise, lock the entire semaphore array, since we either have | |
247 | * multiple semaphores in our own semops, or we need to look at | |
248 | * semaphores from other pending complex operations. | |
249 | * | |
250 | * Carefully guard against sma->complex_count changing between zero | |
251 | * and non-zero while we are spinning for the lock. The value of | |
252 | * sma->complex_count cannot change while we are holding the lock, | |
253 | * so sem_unlock should be fine. | |
254 | * | |
255 | * The global lock path checks that all the local locks have been released, | |
256 | * checking each local lock once. This means that the local lock paths | |
257 | * cannot start their critical sections while the global lock is held. | |
258 | */ | |
259 | static inline int sem_lock(struct sem_array *sma, struct sembuf *sops, | |
260 | int nsops) | |
261 | { | |
262 | int locknum; | |
263 | again: | |
264 | if (nsops == 1 && !sma->complex_count) { | |
265 | struct sem *sem = sma->sem_base + sops->sem_num; | |
266 | ||
267 | /* Lock just the semaphore we are interested in. */ | |
268 | spin_lock(&sem->lock); | |
269 | ||
270 | /* | |
271 | * If sma->complex_count was set while we were spinning, | |
272 | * we may need to look at things we did not lock here. | |
273 | */ | |
274 | if (unlikely(sma->complex_count)) { | |
275 | spin_unlock(&sem->lock); | |
276 | goto lock_array; | |
277 | } | |
278 | ||
279 | /* | |
280 | * Another process is holding the global lock on the | |
281 | * sem_array; we cannot enter our critical section, | |
282 | * but have to wait for the global lock to be released. | |
283 | */ | |
284 | if (unlikely(spin_is_locked(&sma->sem_perm.lock))) { | |
285 | spin_unlock(&sem->lock); | |
286 | spin_unlock_wait(&sma->sem_perm.lock); | |
287 | goto again; | |
288 | } | |
289 | ||
290 | locknum = sops->sem_num; | |
291 | } else { | |
292 | int i; | |
293 | /* | |
294 | * Lock the semaphore array, and wait for all of the | |
295 | * individual semaphore locks to go away. The code | |
296 | * above ensures no new single-lock holders will enter | |
297 | * their critical section while the array lock is held. | |
298 | */ | |
299 | lock_array: | |
cf9d5d78 | 300 | ipc_lock_object(&sma->sem_perm); |
6062a8dc RR |
301 | for (i = 0; i < sma->sem_nsems; i++) { |
302 | struct sem *sem = sma->sem_base + i; | |
303 | spin_unlock_wait(&sem->lock); | |
304 | } | |
305 | locknum = -1; | |
306 | } | |
307 | return locknum; | |
308 | } | |
309 | ||
310 | static inline void sem_unlock(struct sem_array *sma, int locknum) | |
311 | { | |
312 | if (locknum == -1) { | |
f269f40a | 313 | unmerge_queues(sma); |
cf9d5d78 | 314 | ipc_unlock_object(&sma->sem_perm); |
6062a8dc RR |
315 | } else { |
316 | struct sem *sem = sma->sem_base + locknum; | |
317 | spin_unlock(&sem->lock); | |
318 | } | |
6062a8dc RR |
319 | } |
320 | ||
3e148c79 ND |
321 | /* |
322 | * sem_lock_(check_) routines are called in the paths where the rw_mutex | |
323 | * is not held. | |
321310ce LT |
324 | * |
325 | * The caller holds the RCU read lock. | |
3e148c79 | 326 | */ |
6062a8dc RR |
327 | static inline struct sem_array *sem_obtain_lock(struct ipc_namespace *ns, |
328 | int id, struct sembuf *sops, int nsops, int *locknum) | |
023a5355 | 329 | { |
c460b662 RR |
330 | struct kern_ipc_perm *ipcp; |
331 | struct sem_array *sma; | |
03f02c76 | 332 | |
c460b662 | 333 | ipcp = ipc_obtain_object(&sem_ids(ns), id); |
321310ce LT |
334 | if (IS_ERR(ipcp)) |
335 | return ERR_CAST(ipcp); | |
b1ed88b4 | 336 | |
6062a8dc RR |
337 | sma = container_of(ipcp, struct sem_array, sem_perm); |
338 | *locknum = sem_lock(sma, sops, nsops); | |
c460b662 RR |
339 | |
340 | /* ipc_rmid() may have already freed the ID while sem_lock | |
341 | * was spinning: verify that the structure is still valid | |
342 | */ | |
343 | if (!ipcp->deleted) | |
344 | return container_of(ipcp, struct sem_array, sem_perm); | |
345 | ||
6062a8dc | 346 | sem_unlock(sma, *locknum); |
321310ce | 347 | return ERR_PTR(-EINVAL); |
023a5355 ND |
348 | } |
349 | ||
16df3674 DB |
350 | static inline struct sem_array *sem_obtain_object(struct ipc_namespace *ns, int id) |
351 | { | |
352 | struct kern_ipc_perm *ipcp = ipc_obtain_object(&sem_ids(ns), id); | |
353 | ||
354 | if (IS_ERR(ipcp)) | |
355 | return ERR_CAST(ipcp); | |
356 | ||
357 | return container_of(ipcp, struct sem_array, sem_perm); | |
358 | } | |
359 | ||
16df3674 DB |
360 | static inline struct sem_array *sem_obtain_object_check(struct ipc_namespace *ns, |
361 | int id) | |
362 | { | |
363 | struct kern_ipc_perm *ipcp = ipc_obtain_object_check(&sem_ids(ns), id); | |
364 | ||
365 | if (IS_ERR(ipcp)) | |
366 | return ERR_CAST(ipcp); | |
b1ed88b4 | 367 | |
03f02c76 | 368 | return container_of(ipcp, struct sem_array, sem_perm); |
023a5355 ND |
369 | } |
370 | ||
6ff37972 PP |
371 | static inline void sem_lock_and_putref(struct sem_array *sma) |
372 | { | |
6062a8dc | 373 | sem_lock(sma, NULL, -1); |
6ff37972 PP |
374 | ipc_rcu_putref(sma); |
375 | } | |
376 | ||
6ff37972 PP |
377 | static inline void sem_putref(struct sem_array *sma) |
378 | { | |
73b29505 | 379 | ipc_rcu_putref(sma); |
6ff37972 PP |
380 | } |
381 | ||
7ca7e564 ND |
382 | static inline void sem_rmid(struct ipc_namespace *ns, struct sem_array *s) |
383 | { | |
384 | ipc_rmid(&sem_ids(ns), &s->sem_perm); | |
385 | } | |
386 | ||
1da177e4 LT |
387 | /* |
388 | * Lockless wakeup algorithm: | |
389 | * Without the check/retry algorithm a lockless wakeup is possible: | |
390 | * - queue.status is initialized to -EINTR before blocking. | |
391 | * - wakeup is performed by | |
1a82e9e1 | 392 | * * unlinking the queue entry from the pending list |
1da177e4 LT |
393 | * * setting queue.status to IN_WAKEUP |
394 | * This is the notification for the blocked thread that a | |
395 | * result value is imminent. | |
396 | * * call wake_up_process | |
397 | * * set queue.status to the final value. | |
398 | * - the previously blocked thread checks queue.status: | |
399 | * * if it's IN_WAKEUP, then it must wait until the value changes | |
400 | * * if it's not -EINTR, then the operation was completed by | |
401 | * update_queue. semtimedop can return queue.status without | |
5f921ae9 | 402 | * performing any operation on the sem array. |
1da177e4 LT |
403 | * * otherwise it must acquire the spinlock and check what's up. |
404 | * | |
405 | * The two-stage algorithm is necessary to protect against the following | |
406 | * races: | |
407 | * - if queue.status is set after wake_up_process, then the woken up idle | |
408 | * thread could race forward and try (and fail) to acquire sma->lock | |
409 | * before update_queue had a chance to set queue.status | |
410 | * - if queue.status is written before wake_up_process and if the | |
411 | * blocked process is woken up by a signal between writing | |
412 | * queue.status and the wake_up_process, then the woken up | |
413 | * process could return from semtimedop and die by calling | |
414 | * sys_exit before wake_up_process is called. Then wake_up_process | |
415 | * will oops, because the task structure is already invalid. | |
416 | * (yes, this happened on s390 with sysv msg). | |
417 | * | |
418 | */ | |
419 | #define IN_WAKEUP 1 | |
420 | ||
f4566f04 ND |
421 | /** |
422 | * newary - Create a new semaphore set | |
423 | * @ns: namespace | |
424 | * @params: ptr to the structure that contains key, semflg and nsems | |
425 | * | |
3e148c79 | 426 | * Called with sem_ids.rw_mutex held (as a writer) |
f4566f04 ND |
427 | */ |
428 | ||
7748dbfa | 429 | static int newary(struct ipc_namespace *ns, struct ipc_params *params) |
1da177e4 LT |
430 | { |
431 | int id; | |
432 | int retval; | |
433 | struct sem_array *sma; | |
434 | int size; | |
7748dbfa ND |
435 | key_t key = params->key; |
436 | int nsems = params->u.nsems; | |
437 | int semflg = params->flg; | |
b97e820f | 438 | int i; |
1da177e4 LT |
439 | |
440 | if (!nsems) | |
441 | return -EINVAL; | |
e3893534 | 442 | if (ns->used_sems + nsems > ns->sc_semmns) |
1da177e4 LT |
443 | return -ENOSPC; |
444 | ||
445 | size = sizeof (*sma) + nsems * sizeof (struct sem); | |
446 | sma = ipc_rcu_alloc(size); | |
447 | if (!sma) { | |
448 | return -ENOMEM; | |
449 | } | |
450 | memset (sma, 0, size); | |
451 | ||
452 | sma->sem_perm.mode = (semflg & S_IRWXUGO); | |
453 | sma->sem_perm.key = key; | |
454 | ||
455 | sma->sem_perm.security = NULL; | |
456 | retval = security_sem_alloc(sma); | |
457 | if (retval) { | |
458 | ipc_rcu_putref(sma); | |
459 | return retval; | |
460 | } | |
461 | ||
e3893534 | 462 | id = ipc_addid(&sem_ids(ns), &sma->sem_perm, ns->sc_semmni); |
283bb7fa | 463 | if (id < 0) { |
1da177e4 LT |
464 | security_sem_free(sma); |
465 | ipc_rcu_putref(sma); | |
283bb7fa | 466 | return id; |
1da177e4 | 467 | } |
e3893534 | 468 | ns->used_sems += nsems; |
1da177e4 LT |
469 | |
470 | sma->sem_base = (struct sem *) &sma[1]; | |
b97e820f | 471 | |
6062a8dc | 472 | for (i = 0; i < nsems; i++) { |
1a82e9e1 MS |
473 | INIT_LIST_HEAD(&sma->sem_base[i].pending_alter); |
474 | INIT_LIST_HEAD(&sma->sem_base[i].pending_const); | |
6062a8dc RR |
475 | spin_lock_init(&sma->sem_base[i].lock); |
476 | } | |
b97e820f MS |
477 | |
478 | sma->complex_count = 0; | |
1a82e9e1 MS |
479 | INIT_LIST_HEAD(&sma->pending_alter); |
480 | INIT_LIST_HEAD(&sma->pending_const); | |
4daa28f6 | 481 | INIT_LIST_HEAD(&sma->list_id); |
1da177e4 LT |
482 | sma->sem_nsems = nsems; |
483 | sma->sem_ctime = get_seconds(); | |
6062a8dc | 484 | sem_unlock(sma, -1); |
6d49dab8 | 485 | rcu_read_unlock(); |
1da177e4 | 486 | |
7ca7e564 | 487 | return sma->sem_perm.id; |
1da177e4 LT |
488 | } |
489 | ||
7748dbfa | 490 | |
f4566f04 | 491 | /* |
3e148c79 | 492 | * Called with sem_ids.rw_mutex and ipcp locked. |
f4566f04 | 493 | */ |
03f02c76 | 494 | static inline int sem_security(struct kern_ipc_perm *ipcp, int semflg) |
7748dbfa | 495 | { |
03f02c76 ND |
496 | struct sem_array *sma; |
497 | ||
498 | sma = container_of(ipcp, struct sem_array, sem_perm); | |
499 | return security_sem_associate(sma, semflg); | |
7748dbfa ND |
500 | } |
501 | ||
f4566f04 | 502 | /* |
3e148c79 | 503 | * Called with sem_ids.rw_mutex and ipcp locked. |
f4566f04 | 504 | */ |
03f02c76 ND |
505 | static inline int sem_more_checks(struct kern_ipc_perm *ipcp, |
506 | struct ipc_params *params) | |
7748dbfa | 507 | { |
03f02c76 ND |
508 | struct sem_array *sma; |
509 | ||
510 | sma = container_of(ipcp, struct sem_array, sem_perm); | |
511 | if (params->u.nsems > sma->sem_nsems) | |
7748dbfa ND |
512 | return -EINVAL; |
513 | ||
514 | return 0; | |
515 | } | |
516 | ||
d5460c99 | 517 | SYSCALL_DEFINE3(semget, key_t, key, int, nsems, int, semflg) |
1da177e4 | 518 | { |
e3893534 | 519 | struct ipc_namespace *ns; |
7748dbfa ND |
520 | struct ipc_ops sem_ops; |
521 | struct ipc_params sem_params; | |
e3893534 KK |
522 | |
523 | ns = current->nsproxy->ipc_ns; | |
1da177e4 | 524 | |
e3893534 | 525 | if (nsems < 0 || nsems > ns->sc_semmsl) |
1da177e4 | 526 | return -EINVAL; |
7ca7e564 | 527 | |
7748dbfa ND |
528 | sem_ops.getnew = newary; |
529 | sem_ops.associate = sem_security; | |
530 | sem_ops.more_checks = sem_more_checks; | |
531 | ||
532 | sem_params.key = key; | |
533 | sem_params.flg = semflg; | |
534 | sem_params.u.nsems = nsems; | |
1da177e4 | 535 | |
7748dbfa | 536 | return ipcget(ns, &sem_ids(ns), &sem_ops, &sem_params); |
1da177e4 LT |
537 | } |
538 | ||
1da177e4 LT |
539 | /* |
540 | * Determine whether a sequence of semaphore operations would succeed | |
541 | * all at once. Return 0 if yes, 1 if need to sleep, else return error code. | |
542 | */ | |
543 | ||
544 | static int try_atomic_semop (struct sem_array * sma, struct sembuf * sops, | |
545 | int nsops, struct sem_undo *un, int pid) | |
546 | { | |
547 | int result, sem_op; | |
548 | struct sembuf *sop; | |
549 | struct sem * curr; | |
550 | ||
551 | for (sop = sops; sop < sops + nsops; sop++) { | |
552 | curr = sma->sem_base + sop->sem_num; | |
553 | sem_op = sop->sem_op; | |
554 | result = curr->semval; | |
555 | ||
556 | if (!sem_op && result) | |
557 | goto would_block; | |
558 | ||
559 | result += sem_op; | |
560 | if (result < 0) | |
561 | goto would_block; | |
562 | if (result > SEMVMX) | |
563 | goto out_of_range; | |
564 | if (sop->sem_flg & SEM_UNDO) { | |
565 | int undo = un->semadj[sop->sem_num] - sem_op; | |
566 | /* | |
567 | * Exceeding the undo range is an error. | |
568 | */ | |
569 | if (undo < (-SEMAEM - 1) || undo > SEMAEM) | |
570 | goto out_of_range; | |
571 | } | |
572 | curr->semval = result; | |
573 | } | |
574 | ||
575 | sop--; | |
576 | while (sop >= sops) { | |
577 | sma->sem_base[sop->sem_num].sempid = pid; | |
578 | if (sop->sem_flg & SEM_UNDO) | |
579 | un->semadj[sop->sem_num] -= sop->sem_op; | |
580 | sop--; | |
581 | } | |
582 | ||
1da177e4 LT |
583 | return 0; |
584 | ||
585 | out_of_range: | |
586 | result = -ERANGE; | |
587 | goto undo; | |
588 | ||
589 | would_block: | |
590 | if (sop->sem_flg & IPC_NOWAIT) | |
591 | result = -EAGAIN; | |
592 | else | |
593 | result = 1; | |
594 | ||
595 | undo: | |
596 | sop--; | |
597 | while (sop >= sops) { | |
598 | sma->sem_base[sop->sem_num].semval -= sop->sem_op; | |
599 | sop--; | |
600 | } | |
601 | ||
602 | return result; | |
603 | } | |
604 | ||
0a2b9d4c MS |
605 | /** wake_up_sem_queue_prepare(q, error): Prepare wake-up |
606 | * @q: queue entry that must be signaled | |
607 | * @error: Error value for the signal | |
608 | * | |
609 | * Prepare the wake-up of the queue entry q. | |
d4212093 | 610 | */ |
0a2b9d4c MS |
611 | static void wake_up_sem_queue_prepare(struct list_head *pt, |
612 | struct sem_queue *q, int error) | |
d4212093 | 613 | { |
0a2b9d4c MS |
614 | if (list_empty(pt)) { |
615 | /* | |
616 | * Hold preempt off so that we don't get preempted and have the | |
617 | * wakee busy-wait until we're scheduled back on. | |
618 | */ | |
619 | preempt_disable(); | |
620 | } | |
d4212093 | 621 | q->status = IN_WAKEUP; |
0a2b9d4c MS |
622 | q->pid = error; |
623 | ||
9f1bc2c9 | 624 | list_add_tail(&q->list, pt); |
0a2b9d4c MS |
625 | } |
626 | ||
627 | /** | |
628 | * wake_up_sem_queue_do(pt) - do the actual wake-up | |
629 | * @pt: list of tasks to be woken up | |
630 | * | |
631 | * Do the actual wake-up. | |
632 | * The function is called without any locks held, thus the semaphore array | |
633 | * could be destroyed already and the tasks can disappear as soon as the | |
634 | * status is set to the actual return code. | |
635 | */ | |
636 | static void wake_up_sem_queue_do(struct list_head *pt) | |
637 | { | |
638 | struct sem_queue *q, *t; | |
639 | int did_something; | |
640 | ||
641 | did_something = !list_empty(pt); | |
9f1bc2c9 | 642 | list_for_each_entry_safe(q, t, pt, list) { |
0a2b9d4c MS |
643 | wake_up_process(q->sleeper); |
644 | /* q can disappear immediately after writing q->status. */ | |
645 | smp_wmb(); | |
646 | q->status = q->pid; | |
647 | } | |
648 | if (did_something) | |
649 | preempt_enable(); | |
d4212093 NP |
650 | } |
651 | ||
b97e820f MS |
652 | static void unlink_queue(struct sem_array *sma, struct sem_queue *q) |
653 | { | |
654 | list_del(&q->list); | |
9f1bc2c9 | 655 | if (q->nsops > 1) |
b97e820f MS |
656 | sma->complex_count--; |
657 | } | |
658 | ||
fd5db422 MS |
659 | /** check_restart(sma, q) |
660 | * @sma: semaphore array | |
661 | * @q: the operation that just completed | |
662 | * | |
663 | * update_queue is O(N^2) when it restarts scanning the whole queue of | |
664 | * waiting operations. Therefore this function checks if the restart is | |
665 | * really necessary. It is called after a previously waiting operation | |
1a82e9e1 MS |
666 | * modified the array. |
667 | * Note that wait-for-zero operations are handled without restart. | |
fd5db422 MS |
668 | */ |
669 | static int check_restart(struct sem_array *sma, struct sem_queue *q) | |
670 | { | |
1a82e9e1 MS |
671 | /* pending complex alter operations are too difficult to analyse */ |
672 | if (!list_empty(&sma->pending_alter)) | |
fd5db422 MS |
673 | return 1; |
674 | ||
675 | /* we were a sleeping complex operation. Too difficult */ | |
676 | if (q->nsops > 1) | |
677 | return 1; | |
678 | ||
1a82e9e1 MS |
679 | /* It is impossible that someone waits for the new value: |
680 | * - complex operations always restart. | |
681 | * - wait-for-zero are handled seperately. | |
682 | * - q is a previously sleeping simple operation that | |
683 | * altered the array. It must be a decrement, because | |
684 | * simple increments never sleep. | |
685 | * - If there are older (higher priority) decrements | |
686 | * in the queue, then they have observed the original | |
687 | * semval value and couldn't proceed. The operation | |
688 | * decremented to value - thus they won't proceed either. | |
689 | */ | |
690 | return 0; | |
691 | } | |
fd5db422 | 692 | |
1a82e9e1 MS |
693 | /** |
694 | * wake_const_ops(sma, semnum, pt) - Wake up non-alter tasks | |
695 | * @sma: semaphore array. | |
696 | * @semnum: semaphore that was modified. | |
697 | * @pt: list head for the tasks that must be woken up. | |
698 | * | |
699 | * wake_const_ops must be called after a semaphore in a semaphore array | |
700 | * was set to 0. If complex const operations are pending, wake_const_ops must | |
701 | * be called with semnum = -1, as well as with the number of each modified | |
702 | * semaphore. | |
703 | * The tasks that must be woken up are added to @pt. The return code | |
704 | * is stored in q->pid. | |
705 | * The function returns 1 if at least one operation was completed successfully. | |
706 | */ | |
707 | static int wake_const_ops(struct sem_array *sma, int semnum, | |
708 | struct list_head *pt) | |
709 | { | |
710 | struct sem_queue *q; | |
711 | struct list_head *walk; | |
712 | struct list_head *pending_list; | |
713 | int semop_completed = 0; | |
714 | ||
715 | if (semnum == -1) | |
716 | pending_list = &sma->pending_const; | |
717 | else | |
718 | pending_list = &sma->sem_base[semnum].pending_const; | |
fd5db422 | 719 | |
1a82e9e1 MS |
720 | walk = pending_list->next; |
721 | while (walk != pending_list) { | |
722 | int error; | |
723 | ||
724 | q = container_of(walk, struct sem_queue, list); | |
725 | walk = walk->next; | |
726 | ||
727 | error = try_atomic_semop(sma, q->sops, q->nsops, | |
728 | q->undo, q->pid); | |
729 | ||
730 | if (error <= 0) { | |
731 | /* operation completed, remove from queue & wakeup */ | |
732 | ||
733 | unlink_queue(sma, q); | |
734 | ||
735 | wake_up_sem_queue_prepare(pt, q, error); | |
736 | if (error == 0) | |
737 | semop_completed = 1; | |
738 | } | |
739 | } | |
740 | return semop_completed; | |
741 | } | |
742 | ||
743 | /** | |
744 | * do_smart_wakeup_zero(sma, sops, nsops, pt) - wakeup all wait for zero tasks | |
745 | * @sma: semaphore array | |
746 | * @sops: operations that were performed | |
747 | * @nsops: number of operations | |
748 | * @pt: list head of the tasks that must be woken up. | |
749 | * | |
750 | * do_smart_wakeup_zero() checks all required queue for wait-for-zero | |
751 | * operations, based on the actual changes that were performed on the | |
752 | * semaphore array. | |
753 | * The function returns 1 if at least one operation was completed successfully. | |
754 | */ | |
755 | static int do_smart_wakeup_zero(struct sem_array *sma, struct sembuf *sops, | |
756 | int nsops, struct list_head *pt) | |
757 | { | |
758 | int i; | |
759 | int semop_completed = 0; | |
760 | int got_zero = 0; | |
761 | ||
762 | /* first: the per-semaphore queues, if known */ | |
763 | if (sops) { | |
764 | for (i = 0; i < nsops; i++) { | |
765 | int num = sops[i].sem_num; | |
766 | ||
767 | if (sma->sem_base[num].semval == 0) { | |
768 | got_zero = 1; | |
769 | semop_completed |= wake_const_ops(sma, num, pt); | |
770 | } | |
771 | } | |
772 | } else { | |
773 | /* | |
774 | * No sops means modified semaphores not known. | |
775 | * Assume all were changed. | |
fd5db422 | 776 | */ |
1a82e9e1 MS |
777 | for (i = 0; i < sma->sem_nsems; i++) { |
778 | if (sma->sem_base[i].semval == 0) { | |
779 | got_zero = 1; | |
780 | semop_completed |= wake_const_ops(sma, i, pt); | |
781 | } | |
782 | } | |
fd5db422 MS |
783 | } |
784 | /* | |
1a82e9e1 MS |
785 | * If one of the modified semaphores got 0, |
786 | * then check the global queue, too. | |
fd5db422 | 787 | */ |
1a82e9e1 MS |
788 | if (got_zero) |
789 | semop_completed |= wake_const_ops(sma, -1, pt); | |
fd5db422 | 790 | |
1a82e9e1 | 791 | return semop_completed; |
fd5db422 MS |
792 | } |
793 | ||
636c6be8 MS |
794 | |
795 | /** | |
796 | * update_queue(sma, semnum): Look for tasks that can be completed. | |
797 | * @sma: semaphore array. | |
798 | * @semnum: semaphore that was modified. | |
0a2b9d4c | 799 | * @pt: list head for the tasks that must be woken up. |
636c6be8 MS |
800 | * |
801 | * update_queue must be called after a semaphore in a semaphore array | |
9f1bc2c9 RR |
802 | * was modified. If multiple semaphores were modified, update_queue must |
803 | * be called with semnum = -1, as well as with the number of each modified | |
804 | * semaphore. | |
0a2b9d4c MS |
805 | * The tasks that must be woken up are added to @pt. The return code |
806 | * is stored in q->pid. | |
1a82e9e1 MS |
807 | * The function internally checks if const operations can now succeed. |
808 | * | |
0a2b9d4c | 809 | * The function return 1 if at least one semop was completed successfully. |
1da177e4 | 810 | */ |
0a2b9d4c | 811 | static int update_queue(struct sem_array *sma, int semnum, struct list_head *pt) |
1da177e4 | 812 | { |
636c6be8 MS |
813 | struct sem_queue *q; |
814 | struct list_head *walk; | |
815 | struct list_head *pending_list; | |
0a2b9d4c | 816 | int semop_completed = 0; |
636c6be8 | 817 | |
9f1bc2c9 | 818 | if (semnum == -1) |
1a82e9e1 | 819 | pending_list = &sma->pending_alter; |
9f1bc2c9 | 820 | else |
1a82e9e1 | 821 | pending_list = &sma->sem_base[semnum].pending_alter; |
9cad200c NP |
822 | |
823 | again: | |
636c6be8 MS |
824 | walk = pending_list->next; |
825 | while (walk != pending_list) { | |
fd5db422 | 826 | int error, restart; |
636c6be8 | 827 | |
9f1bc2c9 | 828 | q = container_of(walk, struct sem_queue, list); |
636c6be8 | 829 | walk = walk->next; |
1da177e4 | 830 | |
d987f8b2 MS |
831 | /* If we are scanning the single sop, per-semaphore list of |
832 | * one semaphore and that semaphore is 0, then it is not | |
1a82e9e1 | 833 | * necessary to scan further: simple increments |
d987f8b2 MS |
834 | * that affect only one entry succeed immediately and cannot |
835 | * be in the per semaphore pending queue, and decrements | |
836 | * cannot be successful if the value is already 0. | |
837 | */ | |
1a82e9e1 | 838 | if (semnum != -1 && sma->sem_base[semnum].semval == 0) |
d987f8b2 MS |
839 | break; |
840 | ||
1da177e4 LT |
841 | error = try_atomic_semop(sma, q->sops, q->nsops, |
842 | q->undo, q->pid); | |
843 | ||
844 | /* Does q->sleeper still need to sleep? */ | |
9cad200c NP |
845 | if (error > 0) |
846 | continue; | |
847 | ||
b97e820f | 848 | unlink_queue(sma, q); |
9cad200c | 849 | |
0a2b9d4c | 850 | if (error) { |
fd5db422 | 851 | restart = 0; |
0a2b9d4c MS |
852 | } else { |
853 | semop_completed = 1; | |
1a82e9e1 | 854 | do_smart_wakeup_zero(sma, q->sops, q->nsops, pt); |
fd5db422 | 855 | restart = check_restart(sma, q); |
0a2b9d4c | 856 | } |
fd5db422 | 857 | |
0a2b9d4c | 858 | wake_up_sem_queue_prepare(pt, q, error); |
fd5db422 | 859 | if (restart) |
9cad200c | 860 | goto again; |
1da177e4 | 861 | } |
0a2b9d4c | 862 | return semop_completed; |
1da177e4 LT |
863 | } |
864 | ||
0a2b9d4c MS |
865 | /** |
866 | * do_smart_update(sma, sops, nsops, otime, pt) - optimized update_queue | |
fd5db422 MS |
867 | * @sma: semaphore array |
868 | * @sops: operations that were performed | |
869 | * @nsops: number of operations | |
0a2b9d4c MS |
870 | * @otime: force setting otime |
871 | * @pt: list head of the tasks that must be woken up. | |
fd5db422 | 872 | * |
1a82e9e1 MS |
873 | * do_smart_update() does the required calls to update_queue and wakeup_zero, |
874 | * based on the actual changes that were performed on the semaphore array. | |
0a2b9d4c MS |
875 | * Note that the function does not do the actual wake-up: the caller is |
876 | * responsible for calling wake_up_sem_queue_do(@pt). | |
877 | * It is safe to perform this call after dropping all locks. | |
fd5db422 | 878 | */ |
0a2b9d4c MS |
879 | static void do_smart_update(struct sem_array *sma, struct sembuf *sops, int nsops, |
880 | int otime, struct list_head *pt) | |
fd5db422 MS |
881 | { |
882 | int i; | |
883 | ||
1a82e9e1 MS |
884 | otime |= do_smart_wakeup_zero(sma, sops, nsops, pt); |
885 | ||
f269f40a MS |
886 | if (!list_empty(&sma->pending_alter)) { |
887 | /* semaphore array uses the global queue - just process it. */ | |
888 | otime |= update_queue(sma, -1, pt); | |
889 | } else { | |
890 | if (!sops) { | |
891 | /* | |
892 | * No sops, thus the modified semaphores are not | |
893 | * known. Check all. | |
894 | */ | |
895 | for (i = 0; i < sma->sem_nsems; i++) | |
896 | otime |= update_queue(sma, i, pt); | |
897 | } else { | |
898 | /* | |
899 | * Check the semaphores that were increased: | |
900 | * - No complex ops, thus all sleeping ops are | |
901 | * decrease. | |
902 | * - if we decreased the value, then any sleeping | |
903 | * semaphore ops wont be able to run: If the | |
904 | * previous value was too small, then the new | |
905 | * value will be too small, too. | |
906 | */ | |
907 | for (i = 0; i < nsops; i++) { | |
908 | if (sops[i].sem_op > 0) { | |
909 | otime |= update_queue(sma, | |
910 | sops[i].sem_num, pt); | |
911 | } | |
ab465df9 | 912 | } |
9f1bc2c9 | 913 | } |
fd5db422 | 914 | } |
d12e1e50 MS |
915 | if (otime) { |
916 | if (sops == NULL) { | |
917 | sma->sem_base[0].sem_otime = get_seconds(); | |
918 | } else { | |
919 | sma->sem_base[sops[0].sem_num].sem_otime = | |
920 | get_seconds(); | |
921 | } | |
922 | } | |
fd5db422 MS |
923 | } |
924 | ||
925 | ||
1da177e4 LT |
926 | /* The following counts are associated to each semaphore: |
927 | * semncnt number of tasks waiting on semval being nonzero | |
928 | * semzcnt number of tasks waiting on semval being zero | |
929 | * This model assumes that a task waits on exactly one semaphore. | |
930 | * Since semaphore operations are to be performed atomically, tasks actually | |
931 | * wait on a whole sequence of semaphores simultaneously. | |
932 | * The counts we return here are a rough approximation, but still | |
933 | * warrant that semncnt+semzcnt>0 if the task is on the pending queue. | |
934 | */ | |
935 | static int count_semncnt (struct sem_array * sma, ushort semnum) | |
936 | { | |
937 | int semncnt; | |
938 | struct sem_queue * q; | |
939 | ||
940 | semncnt = 0; | |
1a82e9e1 | 941 | list_for_each_entry(q, &sma->sem_base[semnum].pending_alter, list) { |
de2657f9 RR |
942 | struct sembuf * sops = q->sops; |
943 | BUG_ON(sops->sem_num != semnum); | |
944 | if ((sops->sem_op < 0) && !(sops->sem_flg & IPC_NOWAIT)) | |
945 | semncnt++; | |
946 | } | |
947 | ||
1a82e9e1 | 948 | list_for_each_entry(q, &sma->pending_alter, list) { |
1da177e4 LT |
949 | struct sembuf * sops = q->sops; |
950 | int nsops = q->nsops; | |
951 | int i; | |
952 | for (i = 0; i < nsops; i++) | |
953 | if (sops[i].sem_num == semnum | |
954 | && (sops[i].sem_op < 0) | |
955 | && !(sops[i].sem_flg & IPC_NOWAIT)) | |
956 | semncnt++; | |
957 | } | |
958 | return semncnt; | |
959 | } | |
a1193f8e | 960 | |
1da177e4 LT |
961 | static int count_semzcnt (struct sem_array * sma, ushort semnum) |
962 | { | |
963 | int semzcnt; | |
964 | struct sem_queue * q; | |
965 | ||
966 | semzcnt = 0; | |
1a82e9e1 | 967 | list_for_each_entry(q, &sma->sem_base[semnum].pending_const, list) { |
ebc2e5e6 RR |
968 | struct sembuf * sops = q->sops; |
969 | BUG_ON(sops->sem_num != semnum); | |
970 | if ((sops->sem_op == 0) && !(sops->sem_flg & IPC_NOWAIT)) | |
971 | semzcnt++; | |
972 | } | |
973 | ||
1a82e9e1 | 974 | list_for_each_entry(q, &sma->pending_const, list) { |
1da177e4 LT |
975 | struct sembuf * sops = q->sops; |
976 | int nsops = q->nsops; | |
977 | int i; | |
978 | for (i = 0; i < nsops; i++) | |
979 | if (sops[i].sem_num == semnum | |
980 | && (sops[i].sem_op == 0) | |
981 | && !(sops[i].sem_flg & IPC_NOWAIT)) | |
982 | semzcnt++; | |
983 | } | |
984 | return semzcnt; | |
985 | } | |
986 | ||
3e148c79 ND |
987 | /* Free a semaphore set. freeary() is called with sem_ids.rw_mutex locked |
988 | * as a writer and the spinlock for this semaphore set hold. sem_ids.rw_mutex | |
989 | * remains locked on exit. | |
1da177e4 | 990 | */ |
01b8b07a | 991 | static void freeary(struct ipc_namespace *ns, struct kern_ipc_perm *ipcp) |
1da177e4 | 992 | { |
380af1b3 MS |
993 | struct sem_undo *un, *tu; |
994 | struct sem_queue *q, *tq; | |
01b8b07a | 995 | struct sem_array *sma = container_of(ipcp, struct sem_array, sem_perm); |
0a2b9d4c | 996 | struct list_head tasks; |
9f1bc2c9 | 997 | int i; |
1da177e4 | 998 | |
380af1b3 | 999 | /* Free the existing undo structures for this semaphore set. */ |
cf9d5d78 | 1000 | ipc_assert_locked_object(&sma->sem_perm); |
380af1b3 MS |
1001 | list_for_each_entry_safe(un, tu, &sma->list_id, list_id) { |
1002 | list_del(&un->list_id); | |
1003 | spin_lock(&un->ulp->lock); | |
1da177e4 | 1004 | un->semid = -1; |
380af1b3 MS |
1005 | list_del_rcu(&un->list_proc); |
1006 | spin_unlock(&un->ulp->lock); | |
693a8b6e | 1007 | kfree_rcu(un, rcu); |
380af1b3 | 1008 | } |
1da177e4 LT |
1009 | |
1010 | /* Wake up all pending processes and let them fail with EIDRM. */ | |
0a2b9d4c | 1011 | INIT_LIST_HEAD(&tasks); |
1a82e9e1 MS |
1012 | list_for_each_entry_safe(q, tq, &sma->pending_const, list) { |
1013 | unlink_queue(sma, q); | |
1014 | wake_up_sem_queue_prepare(&tasks, q, -EIDRM); | |
1015 | } | |
1016 | ||
1017 | list_for_each_entry_safe(q, tq, &sma->pending_alter, list) { | |
b97e820f | 1018 | unlink_queue(sma, q); |
0a2b9d4c | 1019 | wake_up_sem_queue_prepare(&tasks, q, -EIDRM); |
1da177e4 | 1020 | } |
9f1bc2c9 RR |
1021 | for (i = 0; i < sma->sem_nsems; i++) { |
1022 | struct sem *sem = sma->sem_base + i; | |
1a82e9e1 MS |
1023 | list_for_each_entry_safe(q, tq, &sem->pending_const, list) { |
1024 | unlink_queue(sma, q); | |
1025 | wake_up_sem_queue_prepare(&tasks, q, -EIDRM); | |
1026 | } | |
1027 | list_for_each_entry_safe(q, tq, &sem->pending_alter, list) { | |
9f1bc2c9 RR |
1028 | unlink_queue(sma, q); |
1029 | wake_up_sem_queue_prepare(&tasks, q, -EIDRM); | |
1030 | } | |
1031 | } | |
1da177e4 | 1032 | |
7ca7e564 ND |
1033 | /* Remove the semaphore set from the IDR */ |
1034 | sem_rmid(ns, sma); | |
6062a8dc | 1035 | sem_unlock(sma, -1); |
6d49dab8 | 1036 | rcu_read_unlock(); |
1da177e4 | 1037 | |
0a2b9d4c | 1038 | wake_up_sem_queue_do(&tasks); |
e3893534 | 1039 | ns->used_sems -= sma->sem_nsems; |
1da177e4 LT |
1040 | security_sem_free(sma); |
1041 | ipc_rcu_putref(sma); | |
1042 | } | |
1043 | ||
1044 | static unsigned long copy_semid_to_user(void __user *buf, struct semid64_ds *in, int version) | |
1045 | { | |
1046 | switch(version) { | |
1047 | case IPC_64: | |
1048 | return copy_to_user(buf, in, sizeof(*in)); | |
1049 | case IPC_OLD: | |
1050 | { | |
1051 | struct semid_ds out; | |
1052 | ||
982f7c2b DR |
1053 | memset(&out, 0, sizeof(out)); |
1054 | ||
1da177e4 LT |
1055 | ipc64_perm_to_ipc_perm(&in->sem_perm, &out.sem_perm); |
1056 | ||
1057 | out.sem_otime = in->sem_otime; | |
1058 | out.sem_ctime = in->sem_ctime; | |
1059 | out.sem_nsems = in->sem_nsems; | |
1060 | ||
1061 | return copy_to_user(buf, &out, sizeof(out)); | |
1062 | } | |
1063 | default: | |
1064 | return -EINVAL; | |
1065 | } | |
1066 | } | |
1067 | ||
d12e1e50 MS |
1068 | static time_t get_semotime(struct sem_array *sma) |
1069 | { | |
1070 | int i; | |
1071 | time_t res; | |
1072 | ||
1073 | res = sma->sem_base[0].sem_otime; | |
1074 | for (i = 1; i < sma->sem_nsems; i++) { | |
1075 | time_t to = sma->sem_base[i].sem_otime; | |
1076 | ||
1077 | if (to > res) | |
1078 | res = to; | |
1079 | } | |
1080 | return res; | |
1081 | } | |
1082 | ||
4b9fcb0e | 1083 | static int semctl_nolock(struct ipc_namespace *ns, int semid, |
e1fd1f49 | 1084 | int cmd, int version, void __user *p) |
1da177e4 | 1085 | { |
e5cc9c7b | 1086 | int err; |
1da177e4 LT |
1087 | struct sem_array *sma; |
1088 | ||
1089 | switch(cmd) { | |
1090 | case IPC_INFO: | |
1091 | case SEM_INFO: | |
1092 | { | |
1093 | struct seminfo seminfo; | |
1094 | int max_id; | |
1095 | ||
1096 | err = security_sem_semctl(NULL, cmd); | |
1097 | if (err) | |
1098 | return err; | |
1099 | ||
1100 | memset(&seminfo,0,sizeof(seminfo)); | |
e3893534 KK |
1101 | seminfo.semmni = ns->sc_semmni; |
1102 | seminfo.semmns = ns->sc_semmns; | |
1103 | seminfo.semmsl = ns->sc_semmsl; | |
1104 | seminfo.semopm = ns->sc_semopm; | |
1da177e4 LT |
1105 | seminfo.semvmx = SEMVMX; |
1106 | seminfo.semmnu = SEMMNU; | |
1107 | seminfo.semmap = SEMMAP; | |
1108 | seminfo.semume = SEMUME; | |
3e148c79 | 1109 | down_read(&sem_ids(ns).rw_mutex); |
1da177e4 | 1110 | if (cmd == SEM_INFO) { |
e3893534 KK |
1111 | seminfo.semusz = sem_ids(ns).in_use; |
1112 | seminfo.semaem = ns->used_sems; | |
1da177e4 LT |
1113 | } else { |
1114 | seminfo.semusz = SEMUSZ; | |
1115 | seminfo.semaem = SEMAEM; | |
1116 | } | |
7ca7e564 | 1117 | max_id = ipc_get_maxid(&sem_ids(ns)); |
3e148c79 | 1118 | up_read(&sem_ids(ns).rw_mutex); |
e1fd1f49 | 1119 | if (copy_to_user(p, &seminfo, sizeof(struct seminfo))) |
1da177e4 LT |
1120 | return -EFAULT; |
1121 | return (max_id < 0) ? 0: max_id; | |
1122 | } | |
4b9fcb0e | 1123 | case IPC_STAT: |
1da177e4 LT |
1124 | case SEM_STAT: |
1125 | { | |
1126 | struct semid64_ds tbuf; | |
16df3674 DB |
1127 | int id = 0; |
1128 | ||
1129 | memset(&tbuf, 0, sizeof(tbuf)); | |
1da177e4 | 1130 | |
941b0304 | 1131 | rcu_read_lock(); |
4b9fcb0e | 1132 | if (cmd == SEM_STAT) { |
16df3674 DB |
1133 | sma = sem_obtain_object(ns, semid); |
1134 | if (IS_ERR(sma)) { | |
1135 | err = PTR_ERR(sma); | |
1136 | goto out_unlock; | |
1137 | } | |
4b9fcb0e PP |
1138 | id = sma->sem_perm.id; |
1139 | } else { | |
16df3674 DB |
1140 | sma = sem_obtain_object_check(ns, semid); |
1141 | if (IS_ERR(sma)) { | |
1142 | err = PTR_ERR(sma); | |
1143 | goto out_unlock; | |
1144 | } | |
4b9fcb0e | 1145 | } |
1da177e4 LT |
1146 | |
1147 | err = -EACCES; | |
b0e77598 | 1148 | if (ipcperms(ns, &sma->sem_perm, S_IRUGO)) |
1da177e4 LT |
1149 | goto out_unlock; |
1150 | ||
1151 | err = security_sem_semctl(sma, cmd); | |
1152 | if (err) | |
1153 | goto out_unlock; | |
1154 | ||
1da177e4 | 1155 | kernel_to_ipc64_perm(&sma->sem_perm, &tbuf.sem_perm); |
d12e1e50 MS |
1156 | tbuf.sem_otime = get_semotime(sma); |
1157 | tbuf.sem_ctime = sma->sem_ctime; | |
1158 | tbuf.sem_nsems = sma->sem_nsems; | |
16df3674 | 1159 | rcu_read_unlock(); |
e1fd1f49 | 1160 | if (copy_semid_to_user(p, &tbuf, version)) |
1da177e4 LT |
1161 | return -EFAULT; |
1162 | return id; | |
1163 | } | |
1164 | default: | |
1165 | return -EINVAL; | |
1166 | } | |
1da177e4 | 1167 | out_unlock: |
16df3674 | 1168 | rcu_read_unlock(); |
1da177e4 LT |
1169 | return err; |
1170 | } | |
1171 | ||
e1fd1f49 AV |
1172 | static int semctl_setval(struct ipc_namespace *ns, int semid, int semnum, |
1173 | unsigned long arg) | |
1174 | { | |
1175 | struct sem_undo *un; | |
1176 | struct sem_array *sma; | |
1177 | struct sem* curr; | |
1178 | int err; | |
e1fd1f49 AV |
1179 | struct list_head tasks; |
1180 | int val; | |
1181 | #if defined(CONFIG_64BIT) && defined(__BIG_ENDIAN) | |
1182 | /* big-endian 64bit */ | |
1183 | val = arg >> 32; | |
1184 | #else | |
1185 | /* 32bit or little-endian 64bit */ | |
1186 | val = arg; | |
1187 | #endif | |
1188 | ||
6062a8dc RR |
1189 | if (val > SEMVMX || val < 0) |
1190 | return -ERANGE; | |
e1fd1f49 AV |
1191 | |
1192 | INIT_LIST_HEAD(&tasks); | |
e1fd1f49 | 1193 | |
6062a8dc RR |
1194 | rcu_read_lock(); |
1195 | sma = sem_obtain_object_check(ns, semid); | |
1196 | if (IS_ERR(sma)) { | |
1197 | rcu_read_unlock(); | |
1198 | return PTR_ERR(sma); | |
1199 | } | |
1200 | ||
1201 | if (semnum < 0 || semnum >= sma->sem_nsems) { | |
1202 | rcu_read_unlock(); | |
1203 | return -EINVAL; | |
1204 | } | |
1205 | ||
1206 | ||
1207 | if (ipcperms(ns, &sma->sem_perm, S_IWUGO)) { | |
1208 | rcu_read_unlock(); | |
1209 | return -EACCES; | |
1210 | } | |
e1fd1f49 AV |
1211 | |
1212 | err = security_sem_semctl(sma, SETVAL); | |
6062a8dc RR |
1213 | if (err) { |
1214 | rcu_read_unlock(); | |
1215 | return -EACCES; | |
1216 | } | |
e1fd1f49 | 1217 | |
6062a8dc | 1218 | sem_lock(sma, NULL, -1); |
e1fd1f49 AV |
1219 | |
1220 | curr = &sma->sem_base[semnum]; | |
1221 | ||
cf9d5d78 | 1222 | ipc_assert_locked_object(&sma->sem_perm); |
e1fd1f49 AV |
1223 | list_for_each_entry(un, &sma->list_id, list_id) |
1224 | un->semadj[semnum] = 0; | |
1225 | ||
1226 | curr->semval = val; | |
1227 | curr->sempid = task_tgid_vnr(current); | |
1228 | sma->sem_ctime = get_seconds(); | |
1229 | /* maybe some queued-up processes were waiting for this */ | |
1230 | do_smart_update(sma, NULL, 0, 0, &tasks); | |
6062a8dc | 1231 | sem_unlock(sma, -1); |
6d49dab8 | 1232 | rcu_read_unlock(); |
e1fd1f49 | 1233 | wake_up_sem_queue_do(&tasks); |
6062a8dc | 1234 | return 0; |
e1fd1f49 AV |
1235 | } |
1236 | ||
e3893534 | 1237 | static int semctl_main(struct ipc_namespace *ns, int semid, int semnum, |
e1fd1f49 | 1238 | int cmd, void __user *p) |
1da177e4 LT |
1239 | { |
1240 | struct sem_array *sma; | |
1241 | struct sem* curr; | |
16df3674 | 1242 | int err, nsems; |
1da177e4 LT |
1243 | ushort fast_sem_io[SEMMSL_FAST]; |
1244 | ushort* sem_io = fast_sem_io; | |
0a2b9d4c | 1245 | struct list_head tasks; |
1da177e4 | 1246 | |
16df3674 DB |
1247 | INIT_LIST_HEAD(&tasks); |
1248 | ||
1249 | rcu_read_lock(); | |
1250 | sma = sem_obtain_object_check(ns, semid); | |
1251 | if (IS_ERR(sma)) { | |
1252 | rcu_read_unlock(); | |
023a5355 | 1253 | return PTR_ERR(sma); |
16df3674 | 1254 | } |
1da177e4 LT |
1255 | |
1256 | nsems = sma->sem_nsems; | |
1257 | ||
1da177e4 | 1258 | err = -EACCES; |
c728b9c8 LT |
1259 | if (ipcperms(ns, &sma->sem_perm, cmd == SETALL ? S_IWUGO : S_IRUGO)) |
1260 | goto out_rcu_wakeup; | |
1da177e4 LT |
1261 | |
1262 | err = security_sem_semctl(sma, cmd); | |
c728b9c8 LT |
1263 | if (err) |
1264 | goto out_rcu_wakeup; | |
1da177e4 LT |
1265 | |
1266 | err = -EACCES; | |
1267 | switch (cmd) { | |
1268 | case GETALL: | |
1269 | { | |
e1fd1f49 | 1270 | ushort __user *array = p; |
1da177e4 LT |
1271 | int i; |
1272 | ||
ce857229 | 1273 | sem_lock(sma, NULL, -1); |
1da177e4 | 1274 | if(nsems > SEMMSL_FAST) { |
ce857229 AV |
1275 | if (!ipc_rcu_getref(sma)) { |
1276 | sem_unlock(sma, -1); | |
6d49dab8 | 1277 | rcu_read_unlock(); |
ce857229 AV |
1278 | err = -EIDRM; |
1279 | goto out_free; | |
1280 | } | |
1281 | sem_unlock(sma, -1); | |
6d49dab8 | 1282 | rcu_read_unlock(); |
1da177e4 LT |
1283 | sem_io = ipc_alloc(sizeof(ushort)*nsems); |
1284 | if(sem_io == NULL) { | |
6ff37972 | 1285 | sem_putref(sma); |
1da177e4 LT |
1286 | return -ENOMEM; |
1287 | } | |
1288 | ||
4091fd94 | 1289 | rcu_read_lock(); |
6ff37972 | 1290 | sem_lock_and_putref(sma); |
1da177e4 | 1291 | if (sma->sem_perm.deleted) { |
6062a8dc | 1292 | sem_unlock(sma, -1); |
6d49dab8 | 1293 | rcu_read_unlock(); |
1da177e4 LT |
1294 | err = -EIDRM; |
1295 | goto out_free; | |
1296 | } | |
ce857229 | 1297 | } |
1da177e4 LT |
1298 | for (i = 0; i < sma->sem_nsems; i++) |
1299 | sem_io[i] = sma->sem_base[i].semval; | |
6062a8dc | 1300 | sem_unlock(sma, -1); |
6d49dab8 | 1301 | rcu_read_unlock(); |
1da177e4 LT |
1302 | err = 0; |
1303 | if(copy_to_user(array, sem_io, nsems*sizeof(ushort))) | |
1304 | err = -EFAULT; | |
1305 | goto out_free; | |
1306 | } | |
1307 | case SETALL: | |
1308 | { | |
1309 | int i; | |
1310 | struct sem_undo *un; | |
1311 | ||
6062a8dc RR |
1312 | if (!ipc_rcu_getref(sma)) { |
1313 | rcu_read_unlock(); | |
1314 | return -EIDRM; | |
1315 | } | |
16df3674 | 1316 | rcu_read_unlock(); |
1da177e4 LT |
1317 | |
1318 | if(nsems > SEMMSL_FAST) { | |
1319 | sem_io = ipc_alloc(sizeof(ushort)*nsems); | |
1320 | if(sem_io == NULL) { | |
6ff37972 | 1321 | sem_putref(sma); |
1da177e4 LT |
1322 | return -ENOMEM; |
1323 | } | |
1324 | } | |
1325 | ||
e1fd1f49 | 1326 | if (copy_from_user (sem_io, p, nsems*sizeof(ushort))) { |
6ff37972 | 1327 | sem_putref(sma); |
1da177e4 LT |
1328 | err = -EFAULT; |
1329 | goto out_free; | |
1330 | } | |
1331 | ||
1332 | for (i = 0; i < nsems; i++) { | |
1333 | if (sem_io[i] > SEMVMX) { | |
6ff37972 | 1334 | sem_putref(sma); |
1da177e4 LT |
1335 | err = -ERANGE; |
1336 | goto out_free; | |
1337 | } | |
1338 | } | |
4091fd94 | 1339 | rcu_read_lock(); |
6ff37972 | 1340 | sem_lock_and_putref(sma); |
1da177e4 | 1341 | if (sma->sem_perm.deleted) { |
6062a8dc | 1342 | sem_unlock(sma, -1); |
6d49dab8 | 1343 | rcu_read_unlock(); |
1da177e4 LT |
1344 | err = -EIDRM; |
1345 | goto out_free; | |
1346 | } | |
1347 | ||
1348 | for (i = 0; i < nsems; i++) | |
1349 | sma->sem_base[i].semval = sem_io[i]; | |
4daa28f6 | 1350 | |
cf9d5d78 | 1351 | ipc_assert_locked_object(&sma->sem_perm); |
4daa28f6 | 1352 | list_for_each_entry(un, &sma->list_id, list_id) { |
1da177e4 LT |
1353 | for (i = 0; i < nsems; i++) |
1354 | un->semadj[i] = 0; | |
4daa28f6 | 1355 | } |
1da177e4 LT |
1356 | sma->sem_ctime = get_seconds(); |
1357 | /* maybe some queued-up processes were waiting for this */ | |
0a2b9d4c | 1358 | do_smart_update(sma, NULL, 0, 0, &tasks); |
1da177e4 LT |
1359 | err = 0; |
1360 | goto out_unlock; | |
1361 | } | |
e1fd1f49 | 1362 | /* GETVAL, GETPID, GETNCTN, GETZCNT: fall-through */ |
1da177e4 LT |
1363 | } |
1364 | err = -EINVAL; | |
c728b9c8 LT |
1365 | if (semnum < 0 || semnum >= nsems) |
1366 | goto out_rcu_wakeup; | |
1da177e4 | 1367 | |
6062a8dc | 1368 | sem_lock(sma, NULL, -1); |
1da177e4 LT |
1369 | curr = &sma->sem_base[semnum]; |
1370 | ||
1371 | switch (cmd) { | |
1372 | case GETVAL: | |
1373 | err = curr->semval; | |
1374 | goto out_unlock; | |
1375 | case GETPID: | |
1376 | err = curr->sempid; | |
1377 | goto out_unlock; | |
1378 | case GETNCNT: | |
1379 | err = count_semncnt(sma,semnum); | |
1380 | goto out_unlock; | |
1381 | case GETZCNT: | |
1382 | err = count_semzcnt(sma,semnum); | |
1383 | goto out_unlock; | |
1da177e4 | 1384 | } |
16df3674 | 1385 | |
1da177e4 | 1386 | out_unlock: |
6062a8dc | 1387 | sem_unlock(sma, -1); |
c728b9c8 | 1388 | out_rcu_wakeup: |
6d49dab8 | 1389 | rcu_read_unlock(); |
0a2b9d4c | 1390 | wake_up_sem_queue_do(&tasks); |
1da177e4 LT |
1391 | out_free: |
1392 | if(sem_io != fast_sem_io) | |
1393 | ipc_free(sem_io, sizeof(ushort)*nsems); | |
1394 | return err; | |
1395 | } | |
1396 | ||
016d7132 PP |
1397 | static inline unsigned long |
1398 | copy_semid_from_user(struct semid64_ds *out, void __user *buf, int version) | |
1da177e4 LT |
1399 | { |
1400 | switch(version) { | |
1401 | case IPC_64: | |
016d7132 | 1402 | if (copy_from_user(out, buf, sizeof(*out))) |
1da177e4 | 1403 | return -EFAULT; |
1da177e4 | 1404 | return 0; |
1da177e4 LT |
1405 | case IPC_OLD: |
1406 | { | |
1407 | struct semid_ds tbuf_old; | |
1408 | ||
1409 | if(copy_from_user(&tbuf_old, buf, sizeof(tbuf_old))) | |
1410 | return -EFAULT; | |
1411 | ||
016d7132 PP |
1412 | out->sem_perm.uid = tbuf_old.sem_perm.uid; |
1413 | out->sem_perm.gid = tbuf_old.sem_perm.gid; | |
1414 | out->sem_perm.mode = tbuf_old.sem_perm.mode; | |
1da177e4 LT |
1415 | |
1416 | return 0; | |
1417 | } | |
1418 | default: | |
1419 | return -EINVAL; | |
1420 | } | |
1421 | } | |
1422 | ||
522bb2a2 PP |
1423 | /* |
1424 | * This function handles some semctl commands which require the rw_mutex | |
1425 | * to be held in write mode. | |
1426 | * NOTE: no locks must be held, the rw_mutex is taken inside this function. | |
1427 | */ | |
21a4826a | 1428 | static int semctl_down(struct ipc_namespace *ns, int semid, |
e1fd1f49 | 1429 | int cmd, int version, void __user *p) |
1da177e4 LT |
1430 | { |
1431 | struct sem_array *sma; | |
1432 | int err; | |
016d7132 | 1433 | struct semid64_ds semid64; |
1da177e4 LT |
1434 | struct kern_ipc_perm *ipcp; |
1435 | ||
1436 | if(cmd == IPC_SET) { | |
e1fd1f49 | 1437 | if (copy_semid_from_user(&semid64, p, version)) |
1da177e4 | 1438 | return -EFAULT; |
1da177e4 | 1439 | } |
073115d6 | 1440 | |
7b4cc5d8 DB |
1441 | down_write(&sem_ids(ns).rw_mutex); |
1442 | rcu_read_lock(); | |
1443 | ||
16df3674 DB |
1444 | ipcp = ipcctl_pre_down_nolock(ns, &sem_ids(ns), semid, cmd, |
1445 | &semid64.sem_perm, 0); | |
7b4cc5d8 DB |
1446 | if (IS_ERR(ipcp)) { |
1447 | err = PTR_ERR(ipcp); | |
7b4cc5d8 DB |
1448 | goto out_unlock1; |
1449 | } | |
073115d6 | 1450 | |
a5f75e7f | 1451 | sma = container_of(ipcp, struct sem_array, sem_perm); |
1da177e4 LT |
1452 | |
1453 | err = security_sem_semctl(sma, cmd); | |
7b4cc5d8 DB |
1454 | if (err) |
1455 | goto out_unlock1; | |
1da177e4 | 1456 | |
7b4cc5d8 | 1457 | switch (cmd) { |
1da177e4 | 1458 | case IPC_RMID: |
6062a8dc | 1459 | sem_lock(sma, NULL, -1); |
7b4cc5d8 | 1460 | /* freeary unlocks the ipc object and rcu */ |
01b8b07a | 1461 | freeary(ns, ipcp); |
522bb2a2 | 1462 | goto out_up; |
1da177e4 | 1463 | case IPC_SET: |
6062a8dc | 1464 | sem_lock(sma, NULL, -1); |
1efdb69b EB |
1465 | err = ipc_update_perm(&semid64.sem_perm, ipcp); |
1466 | if (err) | |
7b4cc5d8 | 1467 | goto out_unlock0; |
1da177e4 | 1468 | sma->sem_ctime = get_seconds(); |
1da177e4 LT |
1469 | break; |
1470 | default: | |
1da177e4 | 1471 | err = -EINVAL; |
7b4cc5d8 | 1472 | goto out_unlock1; |
1da177e4 | 1473 | } |
1da177e4 | 1474 | |
7b4cc5d8 | 1475 | out_unlock0: |
6062a8dc | 1476 | sem_unlock(sma, -1); |
7b4cc5d8 | 1477 | out_unlock1: |
6d49dab8 | 1478 | rcu_read_unlock(); |
522bb2a2 PP |
1479 | out_up: |
1480 | up_write(&sem_ids(ns).rw_mutex); | |
1da177e4 LT |
1481 | return err; |
1482 | } | |
1483 | ||
e1fd1f49 | 1484 | SYSCALL_DEFINE4(semctl, int, semid, int, semnum, int, cmd, unsigned long, arg) |
1da177e4 | 1485 | { |
1da177e4 | 1486 | int version; |
e3893534 | 1487 | struct ipc_namespace *ns; |
e1fd1f49 | 1488 | void __user *p = (void __user *)arg; |
1da177e4 LT |
1489 | |
1490 | if (semid < 0) | |
1491 | return -EINVAL; | |
1492 | ||
1493 | version = ipc_parse_version(&cmd); | |
e3893534 | 1494 | ns = current->nsproxy->ipc_ns; |
1da177e4 LT |
1495 | |
1496 | switch(cmd) { | |
1497 | case IPC_INFO: | |
1498 | case SEM_INFO: | |
4b9fcb0e | 1499 | case IPC_STAT: |
1da177e4 | 1500 | case SEM_STAT: |
e1fd1f49 | 1501 | return semctl_nolock(ns, semid, cmd, version, p); |
1da177e4 LT |
1502 | case GETALL: |
1503 | case GETVAL: | |
1504 | case GETPID: | |
1505 | case GETNCNT: | |
1506 | case GETZCNT: | |
1da177e4 | 1507 | case SETALL: |
e1fd1f49 AV |
1508 | return semctl_main(ns, semid, semnum, cmd, p); |
1509 | case SETVAL: | |
1510 | return semctl_setval(ns, semid, semnum, arg); | |
1da177e4 LT |
1511 | case IPC_RMID: |
1512 | case IPC_SET: | |
e1fd1f49 | 1513 | return semctl_down(ns, semid, cmd, version, p); |
1da177e4 LT |
1514 | default: |
1515 | return -EINVAL; | |
1516 | } | |
1517 | } | |
1518 | ||
1da177e4 LT |
1519 | /* If the task doesn't already have a undo_list, then allocate one |
1520 | * here. We guarantee there is only one thread using this undo list, | |
1521 | * and current is THE ONE | |
1522 | * | |
1523 | * If this allocation and assignment succeeds, but later | |
1524 | * portions of this code fail, there is no need to free the sem_undo_list. | |
1525 | * Just let it stay associated with the task, and it'll be freed later | |
1526 | * at exit time. | |
1527 | * | |
1528 | * This can block, so callers must hold no locks. | |
1529 | */ | |
1530 | static inline int get_undo_list(struct sem_undo_list **undo_listp) | |
1531 | { | |
1532 | struct sem_undo_list *undo_list; | |
1da177e4 LT |
1533 | |
1534 | undo_list = current->sysvsem.undo_list; | |
1535 | if (!undo_list) { | |
2453a306 | 1536 | undo_list = kzalloc(sizeof(*undo_list), GFP_KERNEL); |
1da177e4 LT |
1537 | if (undo_list == NULL) |
1538 | return -ENOMEM; | |
00a5dfdb | 1539 | spin_lock_init(&undo_list->lock); |
1da177e4 | 1540 | atomic_set(&undo_list->refcnt, 1); |
4daa28f6 MS |
1541 | INIT_LIST_HEAD(&undo_list->list_proc); |
1542 | ||
1da177e4 LT |
1543 | current->sysvsem.undo_list = undo_list; |
1544 | } | |
1545 | *undo_listp = undo_list; | |
1546 | return 0; | |
1547 | } | |
1548 | ||
bf17bb71 | 1549 | static struct sem_undo *__lookup_undo(struct sem_undo_list *ulp, int semid) |
1da177e4 | 1550 | { |
bf17bb71 | 1551 | struct sem_undo *un; |
4daa28f6 | 1552 | |
bf17bb71 NP |
1553 | list_for_each_entry_rcu(un, &ulp->list_proc, list_proc) { |
1554 | if (un->semid == semid) | |
1555 | return un; | |
1da177e4 | 1556 | } |
4daa28f6 | 1557 | return NULL; |
1da177e4 LT |
1558 | } |
1559 | ||
bf17bb71 NP |
1560 | static struct sem_undo *lookup_undo(struct sem_undo_list *ulp, int semid) |
1561 | { | |
1562 | struct sem_undo *un; | |
1563 | ||
1564 | assert_spin_locked(&ulp->lock); | |
1565 | ||
1566 | un = __lookup_undo(ulp, semid); | |
1567 | if (un) { | |
1568 | list_del_rcu(&un->list_proc); | |
1569 | list_add_rcu(&un->list_proc, &ulp->list_proc); | |
1570 | } | |
1571 | return un; | |
1572 | } | |
1573 | ||
4daa28f6 MS |
1574 | /** |
1575 | * find_alloc_undo - Lookup (and if not present create) undo array | |
1576 | * @ns: namespace | |
1577 | * @semid: semaphore array id | |
1578 | * | |
1579 | * The function looks up (and if not present creates) the undo structure. | |
1580 | * The size of the undo structure depends on the size of the semaphore | |
1581 | * array, thus the alloc path is not that straightforward. | |
380af1b3 MS |
1582 | * Lifetime-rules: sem_undo is rcu-protected, on success, the function |
1583 | * performs a rcu_read_lock(). | |
4daa28f6 MS |
1584 | */ |
1585 | static struct sem_undo *find_alloc_undo(struct ipc_namespace *ns, int semid) | |
1da177e4 LT |
1586 | { |
1587 | struct sem_array *sma; | |
1588 | struct sem_undo_list *ulp; | |
1589 | struct sem_undo *un, *new; | |
6062a8dc | 1590 | int nsems, error; |
1da177e4 LT |
1591 | |
1592 | error = get_undo_list(&ulp); | |
1593 | if (error) | |
1594 | return ERR_PTR(error); | |
1595 | ||
380af1b3 | 1596 | rcu_read_lock(); |
c530c6ac | 1597 | spin_lock(&ulp->lock); |
1da177e4 | 1598 | un = lookup_undo(ulp, semid); |
c530c6ac | 1599 | spin_unlock(&ulp->lock); |
1da177e4 LT |
1600 | if (likely(un!=NULL)) |
1601 | goto out; | |
1602 | ||
1603 | /* no undo structure around - allocate one. */ | |
4daa28f6 | 1604 | /* step 1: figure out the size of the semaphore array */ |
16df3674 DB |
1605 | sma = sem_obtain_object_check(ns, semid); |
1606 | if (IS_ERR(sma)) { | |
1607 | rcu_read_unlock(); | |
4de85cd6 | 1608 | return ERR_CAST(sma); |
16df3674 | 1609 | } |
023a5355 | 1610 | |
1da177e4 | 1611 | nsems = sma->sem_nsems; |
6062a8dc RR |
1612 | if (!ipc_rcu_getref(sma)) { |
1613 | rcu_read_unlock(); | |
1614 | un = ERR_PTR(-EIDRM); | |
1615 | goto out; | |
1616 | } | |
16df3674 | 1617 | rcu_read_unlock(); |
1da177e4 | 1618 | |
4daa28f6 | 1619 | /* step 2: allocate new undo structure */ |
4668edc3 | 1620 | new = kzalloc(sizeof(struct sem_undo) + sizeof(short)*nsems, GFP_KERNEL); |
1da177e4 | 1621 | if (!new) { |
6ff37972 | 1622 | sem_putref(sma); |
1da177e4 LT |
1623 | return ERR_PTR(-ENOMEM); |
1624 | } | |
1da177e4 | 1625 | |
380af1b3 | 1626 | /* step 3: Acquire the lock on semaphore array */ |
4091fd94 | 1627 | rcu_read_lock(); |
6ff37972 | 1628 | sem_lock_and_putref(sma); |
1da177e4 | 1629 | if (sma->sem_perm.deleted) { |
6062a8dc | 1630 | sem_unlock(sma, -1); |
6d49dab8 | 1631 | rcu_read_unlock(); |
1da177e4 LT |
1632 | kfree(new); |
1633 | un = ERR_PTR(-EIDRM); | |
1634 | goto out; | |
1635 | } | |
380af1b3 MS |
1636 | spin_lock(&ulp->lock); |
1637 | ||
1638 | /* | |
1639 | * step 4: check for races: did someone else allocate the undo struct? | |
1640 | */ | |
1641 | un = lookup_undo(ulp, semid); | |
1642 | if (un) { | |
1643 | kfree(new); | |
1644 | goto success; | |
1645 | } | |
4daa28f6 MS |
1646 | /* step 5: initialize & link new undo structure */ |
1647 | new->semadj = (short *) &new[1]; | |
380af1b3 | 1648 | new->ulp = ulp; |
4daa28f6 MS |
1649 | new->semid = semid; |
1650 | assert_spin_locked(&ulp->lock); | |
380af1b3 | 1651 | list_add_rcu(&new->list_proc, &ulp->list_proc); |
cf9d5d78 | 1652 | ipc_assert_locked_object(&sma->sem_perm); |
4daa28f6 | 1653 | list_add(&new->list_id, &sma->list_id); |
380af1b3 | 1654 | un = new; |
4daa28f6 | 1655 | |
380af1b3 | 1656 | success: |
c530c6ac | 1657 | spin_unlock(&ulp->lock); |
6062a8dc | 1658 | sem_unlock(sma, -1); |
1da177e4 LT |
1659 | out: |
1660 | return un; | |
1661 | } | |
1662 | ||
c61284e9 MS |
1663 | |
1664 | /** | |
1665 | * get_queue_result - Retrieve the result code from sem_queue | |
1666 | * @q: Pointer to queue structure | |
1667 | * | |
1668 | * Retrieve the return code from the pending queue. If IN_WAKEUP is found in | |
1669 | * q->status, then we must loop until the value is replaced with the final | |
1670 | * value: This may happen if a task is woken up by an unrelated event (e.g. | |
1671 | * signal) and in parallel the task is woken up by another task because it got | |
1672 | * the requested semaphores. | |
1673 | * | |
1674 | * The function can be called with or without holding the semaphore spinlock. | |
1675 | */ | |
1676 | static int get_queue_result(struct sem_queue *q) | |
1677 | { | |
1678 | int error; | |
1679 | ||
1680 | error = q->status; | |
1681 | while (unlikely(error == IN_WAKEUP)) { | |
1682 | cpu_relax(); | |
1683 | error = q->status; | |
1684 | } | |
1685 | ||
1686 | return error; | |
1687 | } | |
1688 | ||
1689 | ||
d5460c99 HC |
1690 | SYSCALL_DEFINE4(semtimedop, int, semid, struct sembuf __user *, tsops, |
1691 | unsigned, nsops, const struct timespec __user *, timeout) | |
1da177e4 LT |
1692 | { |
1693 | int error = -EINVAL; | |
1694 | struct sem_array *sma; | |
1695 | struct sembuf fast_sops[SEMOPM_FAST]; | |
1696 | struct sembuf* sops = fast_sops, *sop; | |
1697 | struct sem_undo *un; | |
6062a8dc | 1698 | int undos = 0, alter = 0, max, locknum; |
1da177e4 LT |
1699 | struct sem_queue queue; |
1700 | unsigned long jiffies_left = 0; | |
e3893534 | 1701 | struct ipc_namespace *ns; |
0a2b9d4c | 1702 | struct list_head tasks; |
e3893534 KK |
1703 | |
1704 | ns = current->nsproxy->ipc_ns; | |
1da177e4 LT |
1705 | |
1706 | if (nsops < 1 || semid < 0) | |
1707 | return -EINVAL; | |
e3893534 | 1708 | if (nsops > ns->sc_semopm) |
1da177e4 LT |
1709 | return -E2BIG; |
1710 | if(nsops > SEMOPM_FAST) { | |
1711 | sops = kmalloc(sizeof(*sops)*nsops,GFP_KERNEL); | |
1712 | if(sops==NULL) | |
1713 | return -ENOMEM; | |
1714 | } | |
1715 | if (copy_from_user (sops, tsops, nsops * sizeof(*tsops))) { | |
1716 | error=-EFAULT; | |
1717 | goto out_free; | |
1718 | } | |
1719 | if (timeout) { | |
1720 | struct timespec _timeout; | |
1721 | if (copy_from_user(&_timeout, timeout, sizeof(*timeout))) { | |
1722 | error = -EFAULT; | |
1723 | goto out_free; | |
1724 | } | |
1725 | if (_timeout.tv_sec < 0 || _timeout.tv_nsec < 0 || | |
1726 | _timeout.tv_nsec >= 1000000000L) { | |
1727 | error = -EINVAL; | |
1728 | goto out_free; | |
1729 | } | |
1730 | jiffies_left = timespec_to_jiffies(&_timeout); | |
1731 | } | |
1732 | max = 0; | |
1733 | for (sop = sops; sop < sops + nsops; sop++) { | |
1734 | if (sop->sem_num >= max) | |
1735 | max = sop->sem_num; | |
1736 | if (sop->sem_flg & SEM_UNDO) | |
b78755ab MS |
1737 | undos = 1; |
1738 | if (sop->sem_op != 0) | |
1da177e4 LT |
1739 | alter = 1; |
1740 | } | |
1da177e4 | 1741 | |
6062a8dc RR |
1742 | INIT_LIST_HEAD(&tasks); |
1743 | ||
1da177e4 | 1744 | if (undos) { |
6062a8dc | 1745 | /* On success, find_alloc_undo takes the rcu_read_lock */ |
4daa28f6 | 1746 | un = find_alloc_undo(ns, semid); |
1da177e4 LT |
1747 | if (IS_ERR(un)) { |
1748 | error = PTR_ERR(un); | |
1749 | goto out_free; | |
1750 | } | |
6062a8dc | 1751 | } else { |
1da177e4 | 1752 | un = NULL; |
6062a8dc RR |
1753 | rcu_read_lock(); |
1754 | } | |
1da177e4 | 1755 | |
16df3674 | 1756 | sma = sem_obtain_object_check(ns, semid); |
023a5355 | 1757 | if (IS_ERR(sma)) { |
6062a8dc | 1758 | rcu_read_unlock(); |
023a5355 | 1759 | error = PTR_ERR(sma); |
1da177e4 | 1760 | goto out_free; |
023a5355 ND |
1761 | } |
1762 | ||
16df3674 | 1763 | error = -EFBIG; |
c728b9c8 LT |
1764 | if (max >= sma->sem_nsems) |
1765 | goto out_rcu_wakeup; | |
16df3674 DB |
1766 | |
1767 | error = -EACCES; | |
c728b9c8 LT |
1768 | if (ipcperms(ns, &sma->sem_perm, alter ? S_IWUGO : S_IRUGO)) |
1769 | goto out_rcu_wakeup; | |
16df3674 DB |
1770 | |
1771 | error = security_sem_semop(sma, sops, nsops, alter); | |
c728b9c8 LT |
1772 | if (error) |
1773 | goto out_rcu_wakeup; | |
16df3674 | 1774 | |
1da177e4 | 1775 | /* |
4daa28f6 | 1776 | * semid identifiers are not unique - find_alloc_undo may have |
1da177e4 | 1777 | * allocated an undo structure, it was invalidated by an RMID |
4daa28f6 | 1778 | * and now a new array with received the same id. Check and fail. |
25985edc | 1779 | * This case can be detected checking un->semid. The existence of |
380af1b3 | 1780 | * "un" itself is guaranteed by rcu. |
1da177e4 | 1781 | */ |
4daa28f6 | 1782 | error = -EIDRM; |
6062a8dc RR |
1783 | locknum = sem_lock(sma, sops, nsops); |
1784 | if (un && un->semid == -1) | |
1785 | goto out_unlock_free; | |
4daa28f6 | 1786 | |
b488893a | 1787 | error = try_atomic_semop (sma, sops, nsops, un, task_tgid_vnr(current)); |
1da177e4 LT |
1788 | if (error <= 0) { |
1789 | if (alter && error == 0) | |
0a2b9d4c | 1790 | do_smart_update(sma, sops, nsops, 1, &tasks); |
636c6be8 | 1791 | |
1da177e4 LT |
1792 | goto out_unlock_free; |
1793 | } | |
1794 | ||
1795 | /* We need to sleep on this operation, so we put the current | |
1796 | * task into the pending queue and go to sleep. | |
1797 | */ | |
1798 | ||
1da177e4 LT |
1799 | queue.sops = sops; |
1800 | queue.nsops = nsops; | |
1801 | queue.undo = un; | |
b488893a | 1802 | queue.pid = task_tgid_vnr(current); |
1da177e4 | 1803 | queue.alter = alter; |
1da177e4 | 1804 | |
b97e820f MS |
1805 | if (nsops == 1) { |
1806 | struct sem *curr; | |
1807 | curr = &sma->sem_base[sops->sem_num]; | |
1808 | ||
f269f40a MS |
1809 | if (alter) { |
1810 | if (sma->complex_count) { | |
1811 | list_add_tail(&queue.list, | |
1812 | &sma->pending_alter); | |
1813 | } else { | |
1814 | ||
1815 | list_add_tail(&queue.list, | |
1816 | &curr->pending_alter); | |
1817 | } | |
1818 | } else { | |
1a82e9e1 | 1819 | list_add_tail(&queue.list, &curr->pending_const); |
f269f40a | 1820 | } |
b97e820f | 1821 | } else { |
f269f40a MS |
1822 | if (!sma->complex_count) |
1823 | merge_queues(sma); | |
1824 | ||
9f1bc2c9 | 1825 | if (alter) |
1a82e9e1 | 1826 | list_add_tail(&queue.list, &sma->pending_alter); |
9f1bc2c9 | 1827 | else |
1a82e9e1 MS |
1828 | list_add_tail(&queue.list, &sma->pending_const); |
1829 | ||
b97e820f MS |
1830 | sma->complex_count++; |
1831 | } | |
1832 | ||
1da177e4 LT |
1833 | queue.status = -EINTR; |
1834 | queue.sleeper = current; | |
0b0577f6 MS |
1835 | |
1836 | sleep_again: | |
1da177e4 | 1837 | current->state = TASK_INTERRUPTIBLE; |
6062a8dc | 1838 | sem_unlock(sma, locknum); |
6d49dab8 | 1839 | rcu_read_unlock(); |
1da177e4 LT |
1840 | |
1841 | if (timeout) | |
1842 | jiffies_left = schedule_timeout(jiffies_left); | |
1843 | else | |
1844 | schedule(); | |
1845 | ||
c61284e9 | 1846 | error = get_queue_result(&queue); |
1da177e4 LT |
1847 | |
1848 | if (error != -EINTR) { | |
1849 | /* fast path: update_queue already obtained all requested | |
c61284e9 MS |
1850 | * resources. |
1851 | * Perform a smp_mb(): User space could assume that semop() | |
1852 | * is a memory barrier: Without the mb(), the cpu could | |
1853 | * speculatively read in user space stale data that was | |
1854 | * overwritten by the previous owner of the semaphore. | |
1855 | */ | |
1856 | smp_mb(); | |
1857 | ||
1da177e4 LT |
1858 | goto out_free; |
1859 | } | |
1860 | ||
321310ce | 1861 | rcu_read_lock(); |
6062a8dc | 1862 | sma = sem_obtain_lock(ns, semid, sops, nsops, &locknum); |
d694ad62 MS |
1863 | |
1864 | /* | |
1865 | * Wait until it's guaranteed that no wakeup_sem_queue_do() is ongoing. | |
1866 | */ | |
1867 | error = get_queue_result(&queue); | |
1868 | ||
1869 | /* | |
1870 | * Array removed? If yes, leave without sem_unlock(). | |
1871 | */ | |
023a5355 | 1872 | if (IS_ERR(sma)) { |
321310ce | 1873 | rcu_read_unlock(); |
1da177e4 LT |
1874 | goto out_free; |
1875 | } | |
1876 | ||
c61284e9 | 1877 | |
1da177e4 | 1878 | /* |
d694ad62 MS |
1879 | * If queue.status != -EINTR we are woken up by another process. |
1880 | * Leave without unlink_queue(), but with sem_unlock(). | |
1da177e4 | 1881 | */ |
c61284e9 | 1882 | |
1da177e4 LT |
1883 | if (error != -EINTR) { |
1884 | goto out_unlock_free; | |
1885 | } | |
1886 | ||
1887 | /* | |
1888 | * If an interrupt occurred we have to clean up the queue | |
1889 | */ | |
1890 | if (timeout && jiffies_left == 0) | |
1891 | error = -EAGAIN; | |
0b0577f6 MS |
1892 | |
1893 | /* | |
1894 | * If the wakeup was spurious, just retry | |
1895 | */ | |
1896 | if (error == -EINTR && !signal_pending(current)) | |
1897 | goto sleep_again; | |
1898 | ||
b97e820f | 1899 | unlink_queue(sma, &queue); |
1da177e4 LT |
1900 | |
1901 | out_unlock_free: | |
6062a8dc | 1902 | sem_unlock(sma, locknum); |
c728b9c8 | 1903 | out_rcu_wakeup: |
6d49dab8 | 1904 | rcu_read_unlock(); |
0a2b9d4c | 1905 | wake_up_sem_queue_do(&tasks); |
1da177e4 LT |
1906 | out_free: |
1907 | if(sops != fast_sops) | |
1908 | kfree(sops); | |
1909 | return error; | |
1910 | } | |
1911 | ||
d5460c99 HC |
1912 | SYSCALL_DEFINE3(semop, int, semid, struct sembuf __user *, tsops, |
1913 | unsigned, nsops) | |
1da177e4 LT |
1914 | { |
1915 | return sys_semtimedop(semid, tsops, nsops, NULL); | |
1916 | } | |
1917 | ||
1918 | /* If CLONE_SYSVSEM is set, establish sharing of SEM_UNDO state between | |
1919 | * parent and child tasks. | |
1da177e4 LT |
1920 | */ |
1921 | ||
1922 | int copy_semundo(unsigned long clone_flags, struct task_struct *tsk) | |
1923 | { | |
1924 | struct sem_undo_list *undo_list; | |
1925 | int error; | |
1926 | ||
1927 | if (clone_flags & CLONE_SYSVSEM) { | |
1928 | error = get_undo_list(&undo_list); | |
1929 | if (error) | |
1930 | return error; | |
1da177e4 LT |
1931 | atomic_inc(&undo_list->refcnt); |
1932 | tsk->sysvsem.undo_list = undo_list; | |
1933 | } else | |
1934 | tsk->sysvsem.undo_list = NULL; | |
1935 | ||
1936 | return 0; | |
1937 | } | |
1938 | ||
1939 | /* | |
1940 | * add semadj values to semaphores, free undo structures. | |
1941 | * undo structures are not freed when semaphore arrays are destroyed | |
1942 | * so some of them may be out of date. | |
1943 | * IMPLEMENTATION NOTE: There is some confusion over whether the | |
1944 | * set of adjustments that needs to be done should be done in an atomic | |
1945 | * manner or not. That is, if we are attempting to decrement the semval | |
1946 | * should we queue up and wait until we can do so legally? | |
1947 | * The original implementation attempted to do this (queue and wait). | |
1948 | * The current implementation does not do so. The POSIX standard | |
1949 | * and SVID should be consulted to determine what behavior is mandated. | |
1950 | */ | |
1951 | void exit_sem(struct task_struct *tsk) | |
1952 | { | |
4daa28f6 | 1953 | struct sem_undo_list *ulp; |
1da177e4 | 1954 | |
4daa28f6 MS |
1955 | ulp = tsk->sysvsem.undo_list; |
1956 | if (!ulp) | |
1da177e4 | 1957 | return; |
9edff4ab | 1958 | tsk->sysvsem.undo_list = NULL; |
1da177e4 | 1959 | |
4daa28f6 | 1960 | if (!atomic_dec_and_test(&ulp->refcnt)) |
1da177e4 LT |
1961 | return; |
1962 | ||
380af1b3 | 1963 | for (;;) { |
1da177e4 | 1964 | struct sem_array *sma; |
380af1b3 | 1965 | struct sem_undo *un; |
0a2b9d4c | 1966 | struct list_head tasks; |
6062a8dc | 1967 | int semid, i; |
4daa28f6 | 1968 | |
380af1b3 | 1969 | rcu_read_lock(); |
05725f7e JP |
1970 | un = list_entry_rcu(ulp->list_proc.next, |
1971 | struct sem_undo, list_proc); | |
380af1b3 MS |
1972 | if (&un->list_proc == &ulp->list_proc) |
1973 | semid = -1; | |
1974 | else | |
1975 | semid = un->semid; | |
4daa28f6 | 1976 | |
6062a8dc RR |
1977 | if (semid == -1) { |
1978 | rcu_read_unlock(); | |
380af1b3 | 1979 | break; |
6062a8dc | 1980 | } |
1da177e4 | 1981 | |
6062a8dc | 1982 | sma = sem_obtain_object_check(tsk->nsproxy->ipc_ns, un->semid); |
380af1b3 | 1983 | /* exit_sem raced with IPC_RMID, nothing to do */ |
6062a8dc RR |
1984 | if (IS_ERR(sma)) { |
1985 | rcu_read_unlock(); | |
380af1b3 | 1986 | continue; |
6062a8dc | 1987 | } |
1da177e4 | 1988 | |
6062a8dc | 1989 | sem_lock(sma, NULL, -1); |
bf17bb71 | 1990 | un = __lookup_undo(ulp, semid); |
380af1b3 MS |
1991 | if (un == NULL) { |
1992 | /* exit_sem raced with IPC_RMID+semget() that created | |
1993 | * exactly the same semid. Nothing to do. | |
1994 | */ | |
6062a8dc | 1995 | sem_unlock(sma, -1); |
6d49dab8 | 1996 | rcu_read_unlock(); |
380af1b3 MS |
1997 | continue; |
1998 | } | |
1999 | ||
2000 | /* remove un from the linked lists */ | |
cf9d5d78 | 2001 | ipc_assert_locked_object(&sma->sem_perm); |
4daa28f6 MS |
2002 | list_del(&un->list_id); |
2003 | ||
380af1b3 MS |
2004 | spin_lock(&ulp->lock); |
2005 | list_del_rcu(&un->list_proc); | |
2006 | spin_unlock(&ulp->lock); | |
2007 | ||
4daa28f6 MS |
2008 | /* perform adjustments registered in un */ |
2009 | for (i = 0; i < sma->sem_nsems; i++) { | |
5f921ae9 | 2010 | struct sem * semaphore = &sma->sem_base[i]; |
4daa28f6 MS |
2011 | if (un->semadj[i]) { |
2012 | semaphore->semval += un->semadj[i]; | |
1da177e4 LT |
2013 | /* |
2014 | * Range checks of the new semaphore value, | |
2015 | * not defined by sus: | |
2016 | * - Some unices ignore the undo entirely | |
2017 | * (e.g. HP UX 11i 11.22, Tru64 V5.1) | |
2018 | * - some cap the value (e.g. FreeBSD caps | |
2019 | * at 0, but doesn't enforce SEMVMX) | |
2020 | * | |
2021 | * Linux caps the semaphore value, both at 0 | |
2022 | * and at SEMVMX. | |
2023 | * | |
2024 | * Manfred <manfred@colorfullife.com> | |
2025 | */ | |
5f921ae9 IM |
2026 | if (semaphore->semval < 0) |
2027 | semaphore->semval = 0; | |
2028 | if (semaphore->semval > SEMVMX) | |
2029 | semaphore->semval = SEMVMX; | |
b488893a | 2030 | semaphore->sempid = task_tgid_vnr(current); |
1da177e4 LT |
2031 | } |
2032 | } | |
1da177e4 | 2033 | /* maybe some queued-up processes were waiting for this */ |
0a2b9d4c MS |
2034 | INIT_LIST_HEAD(&tasks); |
2035 | do_smart_update(sma, NULL, 0, 1, &tasks); | |
6062a8dc | 2036 | sem_unlock(sma, -1); |
6d49dab8 | 2037 | rcu_read_unlock(); |
0a2b9d4c | 2038 | wake_up_sem_queue_do(&tasks); |
380af1b3 | 2039 | |
693a8b6e | 2040 | kfree_rcu(un, rcu); |
1da177e4 | 2041 | } |
4daa28f6 | 2042 | kfree(ulp); |
1da177e4 LT |
2043 | } |
2044 | ||
2045 | #ifdef CONFIG_PROC_FS | |
19b4946c | 2046 | static int sysvipc_sem_proc_show(struct seq_file *s, void *it) |
1da177e4 | 2047 | { |
1efdb69b | 2048 | struct user_namespace *user_ns = seq_user_ns(s); |
19b4946c | 2049 | struct sem_array *sma = it; |
d12e1e50 MS |
2050 | time_t sem_otime; |
2051 | ||
2052 | sem_otime = get_semotime(sma); | |
19b4946c MW |
2053 | |
2054 | return seq_printf(s, | |
b97e820f | 2055 | "%10d %10d %4o %10u %5u %5u %5u %5u %10lu %10lu\n", |
19b4946c | 2056 | sma->sem_perm.key, |
7ca7e564 | 2057 | sma->sem_perm.id, |
19b4946c MW |
2058 | sma->sem_perm.mode, |
2059 | sma->sem_nsems, | |
1efdb69b EB |
2060 | from_kuid_munged(user_ns, sma->sem_perm.uid), |
2061 | from_kgid_munged(user_ns, sma->sem_perm.gid), | |
2062 | from_kuid_munged(user_ns, sma->sem_perm.cuid), | |
2063 | from_kgid_munged(user_ns, sma->sem_perm.cgid), | |
d12e1e50 | 2064 | sem_otime, |
19b4946c | 2065 | sma->sem_ctime); |
1da177e4 LT |
2066 | } |
2067 | #endif |