]>
Commit | Line | Data |
---|---|---|
b2441318 | 1 | // SPDX-License-Identifier: GPL-2.0 |
1da177e4 LT |
2 | /* |
3 | * linux/ipc/sem.c | |
4 | * Copyright (C) 1992 Krishna Balasubramanian | |
5 | * Copyright (C) 1995 Eric Schenk, Bruno Haible | |
6 | * | |
1da177e4 LT |
7 | * /proc/sysvipc/sem support (c) 1999 Dragos Acostachioaie <dragos@iname.com> |
8 | * | |
9 | * SMP-threaded, sysctl's added | |
624dffcb | 10 | * (c) 1999 Manfred Spraul <manfred@colorfullife.com> |
1da177e4 | 11 | * Enforced range limit on SEM_UNDO |
046c6884 | 12 | * (c) 2001 Red Hat Inc |
1da177e4 LT |
13 | * Lockless wakeup |
14 | * (c) 2003 Manfred Spraul <manfred@colorfullife.com> | |
9ae949fa | 15 | * (c) 2016 Davidlohr Bueso <dave@stgolabs.net> |
c5cf6359 MS |
16 | * Further wakeup optimizations, documentation |
17 | * (c) 2010 Manfred Spraul <manfred@colorfullife.com> | |
073115d6 SG |
18 | * |
19 | * support for audit of ipc object properties and permission changes | |
20 | * Dustin Kirkland <dustin.kirkland@us.ibm.com> | |
e3893534 KK |
21 | * |
22 | * namespaces support | |
23 | * OpenVZ, SWsoft Inc. | |
24 | * Pavel Emelianov <xemul@openvz.org> | |
c5cf6359 MS |
25 | * |
26 | * Implementation notes: (May 2010) | |
27 | * This file implements System V semaphores. | |
28 | * | |
29 | * User space visible behavior: | |
30 | * - FIFO ordering for semop() operations (just FIFO, not starvation | |
31 | * protection) | |
32 | * - multiple semaphore operations that alter the same semaphore in | |
33 | * one semop() are handled. | |
34 | * - sem_ctime (time of last semctl()) is updated in the IPC_SET, SETVAL and | |
35 | * SETALL calls. | |
36 | * - two Linux specific semctl() commands: SEM_STAT, SEM_INFO. | |
37 | * - undo adjustments at process exit are limited to 0..SEMVMX. | |
38 | * - namespace are supported. | |
39 | * - SEMMSL, SEMMNS, SEMOPM and SEMMNI can be configured at runtine by writing | |
40 | * to /proc/sys/kernel/sem. | |
41 | * - statistics about the usage are reported in /proc/sysvipc/sem. | |
42 | * | |
43 | * Internals: | |
44 | * - scalability: | |
45 | * - all global variables are read-mostly. | |
46 | * - semop() calls and semctl(RMID) are synchronized by RCU. | |
47 | * - most operations do write operations (actually: spin_lock calls) to | |
48 | * the per-semaphore array structure. | |
49 | * Thus: Perfect SMP scaling between independent semaphore arrays. | |
50 | * If multiple semaphores in one array are used, then cache line | |
51 | * trashing on the semaphore array spinlock will limit the scaling. | |
2f2ed41d | 52 | * - semncnt and semzcnt are calculated on demand in count_semcnt() |
c5cf6359 MS |
53 | * - the task that performs a successful semop() scans the list of all |
54 | * sleeping tasks and completes any pending operations that can be fulfilled. | |
55 | * Semaphores are actively given to waiting tasks (necessary for FIFO). | |
56 | * (see update_queue()) | |
57 | * - To improve the scalability, the actual wake-up calls are performed after | |
9ae949fa | 58 | * dropping all locks. (see wake_up_sem_queue_prepare()) |
c5cf6359 MS |
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). | |
c5cf6359 MS |
63 | * - UNDO values are stored in an array (one per process and per |
64 | * semaphore array, lazily allocated). For backwards compatibility, multiple | |
65 | * modes for the UNDO variables are supported (per process, per thread) | |
66 | * (see copy_semundo, CLONE_SYSVSEM) | |
67 | * - There are two lists of the pending operations: a per-array list | |
68 | * and per-semaphore list (stored in the array). This allows to achieve FIFO | |
69 | * ordering without always scanning all pending operations. | |
70 | * The worst-case behavior is nevertheless O(N^2) for N wakeups. | |
1da177e4 LT |
71 | */ |
72 | ||
b0d17578 | 73 | #include <linux/compat.h> |
1da177e4 LT |
74 | #include <linux/slab.h> |
75 | #include <linux/spinlock.h> | |
76 | #include <linux/init.h> | |
77 | #include <linux/proc_fs.h> | |
78 | #include <linux/time.h> | |
1da177e4 LT |
79 | #include <linux/security.h> |
80 | #include <linux/syscalls.h> | |
81 | #include <linux/audit.h> | |
c59ede7b | 82 | #include <linux/capability.h> |
19b4946c | 83 | #include <linux/seq_file.h> |
3e148c79 | 84 | #include <linux/rwsem.h> |
e3893534 | 85 | #include <linux/nsproxy.h> |
ae5e1b22 | 86 | #include <linux/ipc_namespace.h> |
84f001e1 | 87 | #include <linux/sched/wake_q.h> |
5f921ae9 | 88 | |
7153e402 | 89 | #include <linux/uaccess.h> |
1da177e4 LT |
90 | #include "util.h" |
91 | ||
1a5c1349 EB |
92 | /* One semaphore structure for each semaphore in the system. */ |
93 | struct sem { | |
94 | int semval; /* current value */ | |
95 | /* | |
96 | * PID of the process that last modified the semaphore. For | |
97 | * Linux, specifically these are: | |
98 | * - semop | |
99 | * - semctl, via SETVAL and SETALL. | |
100 | * - at task exit when performing undo adjustments (see exit_sem). | |
101 | */ | |
51d6f263 | 102 | struct pid *sempid; |
1a5c1349 EB |
103 | spinlock_t lock; /* spinlock for fine-grained semtimedop */ |
104 | struct list_head pending_alter; /* pending single-sop operations */ | |
105 | /* that alter the semaphore */ | |
106 | struct list_head pending_const; /* pending single-sop operations */ | |
107 | /* that do not alter the semaphore*/ | |
2a70b787 | 108 | time64_t sem_otime; /* candidate for sem_otime */ |
1a5c1349 EB |
109 | } ____cacheline_aligned_in_smp; |
110 | ||
111 | /* One sem_array data structure for each set of semaphores in the system. */ | |
112 | struct sem_array { | |
113 | struct kern_ipc_perm sem_perm; /* permissions .. see ipc.h */ | |
114 | time64_t sem_ctime; /* create/last semctl() time */ | |
115 | struct list_head pending_alter; /* pending operations */ | |
116 | /* that alter the array */ | |
117 | struct list_head pending_const; /* pending complex operations */ | |
118 | /* that do not alter semvals */ | |
119 | struct list_head list_id; /* undo requests on this array */ | |
120 | int sem_nsems; /* no. of semaphores in array */ | |
121 | int complex_count; /* pending complex operations */ | |
122 | unsigned int use_global_lock;/* >0: global lock required */ | |
123 | ||
124 | struct sem sems[]; | |
125 | } __randomize_layout; | |
e57940d7 MS |
126 | |
127 | /* One queue for each sleeping process in the system. */ | |
128 | struct sem_queue { | |
e57940d7 MS |
129 | struct list_head list; /* queue of pending operations */ |
130 | struct task_struct *sleeper; /* this process */ | |
131 | struct sem_undo *undo; /* undo structure */ | |
51d6f263 | 132 | struct pid *pid; /* process id of requesting process */ |
e57940d7 MS |
133 | int status; /* completion status of operation */ |
134 | struct sembuf *sops; /* array of pending operations */ | |
ed247b7c | 135 | struct sembuf *blocking; /* the operation that blocked */ |
e57940d7 | 136 | int nsops; /* number of operations */ |
4ce33ec2 DB |
137 | bool alter; /* does *sops alter the array? */ |
138 | bool dupsop; /* sops on more than one sem_num */ | |
e57940d7 MS |
139 | }; |
140 | ||
141 | /* Each task has a list of undo requests. They are executed automatically | |
142 | * when the process exits. | |
143 | */ | |
144 | struct sem_undo { | |
145 | struct list_head list_proc; /* per-process list: * | |
146 | * all undos from one process | |
147 | * rcu protected */ | |
148 | struct rcu_head rcu; /* rcu struct for sem_undo */ | |
149 | struct sem_undo_list *ulp; /* back ptr to sem_undo_list */ | |
150 | struct list_head list_id; /* per semaphore array list: | |
151 | * all undos for one array */ | |
152 | int semid; /* semaphore set identifier */ | |
153 | short *semadj; /* array of adjustments */ | |
154 | /* one per semaphore */ | |
155 | }; | |
156 | ||
157 | /* sem_undo_list controls shared access to the list of sem_undo structures | |
158 | * that may be shared among all a CLONE_SYSVSEM task group. | |
159 | */ | |
160 | struct sem_undo_list { | |
f74370b8 | 161 | refcount_t refcnt; |
e57940d7 MS |
162 | spinlock_t lock; |
163 | struct list_head list_proc; | |
164 | }; | |
165 | ||
166 | ||
ed2ddbf8 | 167 | #define sem_ids(ns) ((ns)->ids[IPC_SEM_IDS]) |
e3893534 | 168 | |
7748dbfa | 169 | static int newary(struct ipc_namespace *, struct ipc_params *); |
01b8b07a | 170 | static void freeary(struct ipc_namespace *, struct kern_ipc_perm *); |
1da177e4 | 171 | #ifdef CONFIG_PROC_FS |
19b4946c | 172 | static int sysvipc_sem_proc_show(struct seq_file *s, void *it); |
1da177e4 LT |
173 | #endif |
174 | ||
175 | #define SEMMSL_FAST 256 /* 512 bytes on stack */ | |
176 | #define SEMOPM_FAST 64 /* ~ 372 bytes on stack */ | |
177 | ||
9de5ab8a MS |
178 | /* |
179 | * Switching from the mode suitable for simple ops | |
180 | * to the mode for complex ops is costly. Therefore: | |
181 | * use some hysteresis | |
182 | */ | |
183 | #define USE_GLOBAL_LOCK_HYSTERESIS 10 | |
184 | ||
1da177e4 | 185 | /* |
758a6ba3 | 186 | * Locking: |
5864a2fd | 187 | * a) global sem_lock() for read/write |
1da177e4 | 188 | * sem_undo.id_next, |
758a6ba3 | 189 | * sem_array.complex_count, |
5864a2fd MS |
190 | * sem_array.pending{_alter,_const}, |
191 | * sem_array.sem_undo | |
46c0a8ca | 192 | * |
5864a2fd | 193 | * b) global or semaphore sem_lock() for read/write: |
1a233956 | 194 | * sem_array.sems[i].pending_{const,alter}: |
5864a2fd MS |
195 | * |
196 | * c) special: | |
197 | * sem_undo_list.list_proc: | |
198 | * * undo_list->lock for write | |
199 | * * rcu for read | |
9de5ab8a MS |
200 | * use_global_lock: |
201 | * * global sem_lock() for write | |
202 | * * either local or global sem_lock() for read. | |
203 | * | |
204 | * Memory ordering: | |
205 | * Most ordering is enforced by using spin_lock() and spin_unlock(). | |
206 | * The special case is use_global_lock: | |
207 | * Setting it from non-zero to 0 is a RELEASE, this is ensured by | |
208 | * using smp_store_release(). | |
209 | * Testing if it is non-zero is an ACQUIRE, this is ensured by using | |
210 | * smp_load_acquire(). | |
211 | * Setting it from 0 to non-zero must be ordered with regards to | |
212 | * this smp_load_acquire(), this is guaranteed because the smp_load_acquire() | |
213 | * is inside a spin_lock() and after a write from 0 to non-zero a | |
214 | * spin_lock()+spin_unlock() is done. | |
1da177e4 LT |
215 | */ |
216 | ||
e3893534 KK |
217 | #define sc_semmsl sem_ctls[0] |
218 | #define sc_semmns sem_ctls[1] | |
219 | #define sc_semopm sem_ctls[2] | |
220 | #define sc_semmni sem_ctls[3] | |
221 | ||
0cfb6aee | 222 | int sem_init_ns(struct ipc_namespace *ns) |
e3893534 | 223 | { |
e3893534 KK |
224 | ns->sc_semmsl = SEMMSL; |
225 | ns->sc_semmns = SEMMNS; | |
226 | ns->sc_semopm = SEMOPM; | |
227 | ns->sc_semmni = SEMMNI; | |
228 | ns->used_sems = 0; | |
0cfb6aee | 229 | return ipc_init_ids(&ns->ids[IPC_SEM_IDS]); |
e3893534 KK |
230 | } |
231 | ||
ae5e1b22 | 232 | #ifdef CONFIG_IPC_NS |
e3893534 KK |
233 | void sem_exit_ns(struct ipc_namespace *ns) |
234 | { | |
01b8b07a | 235 | free_ipcs(ns, &sem_ids(ns), freeary); |
7d6feeb2 | 236 | idr_destroy(&ns->ids[IPC_SEM_IDS].ipcs_idr); |
0cfb6aee | 237 | rhashtable_destroy(&ns->ids[IPC_SEM_IDS].key_ht); |
e3893534 | 238 | } |
ae5e1b22 | 239 | #endif |
1da177e4 | 240 | |
0cfb6aee | 241 | int __init sem_init(void) |
1da177e4 | 242 | { |
0cfb6aee GK |
243 | const int err = sem_init_ns(&init_ipc_ns); |
244 | ||
19b4946c MW |
245 | ipc_init_proc_interface("sysvipc/sem", |
246 | " key semid perms nsems uid gid cuid cgid otime ctime\n", | |
e3893534 | 247 | IPC_SEM_IDS, sysvipc_sem_proc_show); |
0cfb6aee | 248 | return err; |
1da177e4 LT |
249 | } |
250 | ||
f269f40a MS |
251 | /** |
252 | * unmerge_queues - unmerge queues, if possible. | |
253 | * @sma: semaphore array | |
254 | * | |
255 | * The function unmerges the wait queues if complex_count is 0. | |
256 | * It must be called prior to dropping the global semaphore array lock. | |
257 | */ | |
258 | static void unmerge_queues(struct sem_array *sma) | |
259 | { | |
260 | struct sem_queue *q, *tq; | |
261 | ||
262 | /* complex operations still around? */ | |
263 | if (sma->complex_count) | |
264 | return; | |
265 | /* | |
266 | * We will switch back to simple mode. | |
267 | * Move all pending operation back into the per-semaphore | |
268 | * queues. | |
269 | */ | |
270 | list_for_each_entry_safe(q, tq, &sma->pending_alter, list) { | |
271 | struct sem *curr; | |
1a233956 | 272 | curr = &sma->sems[q->sops[0].sem_num]; |
f269f40a MS |
273 | |
274 | list_add_tail(&q->list, &curr->pending_alter); | |
275 | } | |
276 | INIT_LIST_HEAD(&sma->pending_alter); | |
277 | } | |
278 | ||
279 | /** | |
8001c858 | 280 | * merge_queues - merge single semop queues into global queue |
f269f40a MS |
281 | * @sma: semaphore array |
282 | * | |
283 | * This function merges all per-semaphore queues into the global queue. | |
284 | * It is necessary to achieve FIFO ordering for the pending single-sop | |
285 | * operations when a multi-semop operation must sleep. | |
286 | * Only the alter operations must be moved, the const operations can stay. | |
287 | */ | |
288 | static void merge_queues(struct sem_array *sma) | |
289 | { | |
290 | int i; | |
291 | for (i = 0; i < sma->sem_nsems; i++) { | |
1a233956 | 292 | struct sem *sem = &sma->sems[i]; |
f269f40a MS |
293 | |
294 | list_splice_init(&sem->pending_alter, &sma->pending_alter); | |
295 | } | |
296 | } | |
297 | ||
53dad6d3 DB |
298 | static void sem_rcu_free(struct rcu_head *head) |
299 | { | |
dba4cdd3 MS |
300 | struct kern_ipc_perm *p = container_of(head, struct kern_ipc_perm, rcu); |
301 | struct sem_array *sma = container_of(p, struct sem_array, sem_perm); | |
53dad6d3 | 302 | |
aefad959 | 303 | security_sem_free(&sma->sem_perm); |
e2029dfe | 304 | kvfree(sma); |
53dad6d3 DB |
305 | } |
306 | ||
5e9d5275 | 307 | /* |
5864a2fd | 308 | * Enter the mode suitable for non-simple operations: |
5e9d5275 | 309 | * Caller must own sem_perm.lock. |
5e9d5275 | 310 | */ |
5864a2fd | 311 | static void complexmode_enter(struct sem_array *sma) |
5e9d5275 MS |
312 | { |
313 | int i; | |
314 | struct sem *sem; | |
315 | ||
9de5ab8a MS |
316 | if (sma->use_global_lock > 0) { |
317 | /* | |
318 | * We are already in global lock mode. | |
319 | * Nothing to do, just reset the | |
320 | * counter until we return to simple mode. | |
321 | */ | |
322 | sma->use_global_lock = USE_GLOBAL_LOCK_HYSTERESIS; | |
6d07b68c MS |
323 | return; |
324 | } | |
9de5ab8a | 325 | sma->use_global_lock = USE_GLOBAL_LOCK_HYSTERESIS; |
5864a2fd | 326 | |
5e9d5275 | 327 | for (i = 0; i < sma->sem_nsems; i++) { |
1a233956 | 328 | sem = &sma->sems[i]; |
27d7be18 MS |
329 | spin_lock(&sem->lock); |
330 | spin_unlock(&sem->lock); | |
5e9d5275 | 331 | } |
5864a2fd MS |
332 | } |
333 | ||
334 | /* | |
335 | * Try to leave the mode that disallows simple operations: | |
336 | * Caller must own sem_perm.lock. | |
337 | */ | |
338 | static void complexmode_tryleave(struct sem_array *sma) | |
339 | { | |
340 | if (sma->complex_count) { | |
341 | /* Complex ops are sleeping. | |
342 | * We must stay in complex mode | |
343 | */ | |
344 | return; | |
345 | } | |
9de5ab8a MS |
346 | if (sma->use_global_lock == 1) { |
347 | /* | |
348 | * Immediately after setting use_global_lock to 0, | |
349 | * a simple op can start. Thus: all memory writes | |
350 | * performed by the current operation must be visible | |
351 | * before we set use_global_lock to 0. | |
352 | */ | |
353 | smp_store_release(&sma->use_global_lock, 0); | |
354 | } else { | |
355 | sma->use_global_lock--; | |
356 | } | |
5e9d5275 MS |
357 | } |
358 | ||
5864a2fd | 359 | #define SEM_GLOBAL_LOCK (-1) |
6062a8dc RR |
360 | /* |
361 | * If the request contains only one semaphore operation, and there are | |
362 | * no complex transactions pending, lock only the semaphore involved. | |
363 | * Otherwise, lock the entire semaphore array, since we either have | |
364 | * multiple semaphores in our own semops, or we need to look at | |
365 | * semaphores from other pending complex operations. | |
6062a8dc RR |
366 | */ |
367 | static inline int sem_lock(struct sem_array *sma, struct sembuf *sops, | |
368 | int nsops) | |
369 | { | |
5e9d5275 | 370 | struct sem *sem; |
6062a8dc | 371 | |
5e9d5275 MS |
372 | if (nsops != 1) { |
373 | /* Complex operation - acquire a full lock */ | |
374 | ipc_lock_object(&sma->sem_perm); | |
6062a8dc | 375 | |
5864a2fd MS |
376 | /* Prevent parallel simple ops */ |
377 | complexmode_enter(sma); | |
378 | return SEM_GLOBAL_LOCK; | |
5e9d5275 MS |
379 | } |
380 | ||
381 | /* | |
382 | * Only one semaphore affected - try to optimize locking. | |
5864a2fd MS |
383 | * Optimized locking is possible if no complex operation |
384 | * is either enqueued or processed right now. | |
385 | * | |
9de5ab8a | 386 | * Both facts are tracked by use_global_mode. |
5e9d5275 | 387 | */ |
1a233956 | 388 | sem = &sma->sems[sops->sem_num]; |
6062a8dc | 389 | |
5864a2fd | 390 | /* |
9de5ab8a | 391 | * Initial check for use_global_lock. Just an optimization, |
5864a2fd MS |
392 | * no locking, no memory barrier. |
393 | */ | |
9de5ab8a | 394 | if (!sma->use_global_lock) { |
6062a8dc | 395 | /* |
5e9d5275 MS |
396 | * It appears that no complex operation is around. |
397 | * Acquire the per-semaphore lock. | |
6062a8dc | 398 | */ |
5e9d5275 MS |
399 | spin_lock(&sem->lock); |
400 | ||
9de5ab8a MS |
401 | /* pairs with smp_store_release() */ |
402 | if (!smp_load_acquire(&sma->use_global_lock)) { | |
5864a2fd MS |
403 | /* fast path successful! */ |
404 | return sops->sem_num; | |
6062a8dc | 405 | } |
5e9d5275 MS |
406 | spin_unlock(&sem->lock); |
407 | } | |
408 | ||
409 | /* slow path: acquire the full lock */ | |
410 | ipc_lock_object(&sma->sem_perm); | |
6062a8dc | 411 | |
9de5ab8a MS |
412 | if (sma->use_global_lock == 0) { |
413 | /* | |
414 | * The use_global_lock mode ended while we waited for | |
415 | * sma->sem_perm.lock. Thus we must switch to locking | |
416 | * with sem->lock. | |
417 | * Unlike in the fast path, there is no need to recheck | |
418 | * sma->use_global_lock after we have acquired sem->lock: | |
419 | * We own sma->sem_perm.lock, thus use_global_lock cannot | |
420 | * change. | |
5e9d5275 MS |
421 | */ |
422 | spin_lock(&sem->lock); | |
9de5ab8a | 423 | |
5e9d5275 MS |
424 | ipc_unlock_object(&sma->sem_perm); |
425 | return sops->sem_num; | |
6062a8dc | 426 | } else { |
9de5ab8a MS |
427 | /* |
428 | * Not a false alarm, thus continue to use the global lock | |
429 | * mode. No need for complexmode_enter(), this was done by | |
430 | * the caller that has set use_global_mode to non-zero. | |
6062a8dc | 431 | */ |
5864a2fd | 432 | return SEM_GLOBAL_LOCK; |
6062a8dc | 433 | } |
6062a8dc RR |
434 | } |
435 | ||
436 | static inline void sem_unlock(struct sem_array *sma, int locknum) | |
437 | { | |
5864a2fd | 438 | if (locknum == SEM_GLOBAL_LOCK) { |
f269f40a | 439 | unmerge_queues(sma); |
5864a2fd | 440 | complexmode_tryleave(sma); |
cf9d5d78 | 441 | ipc_unlock_object(&sma->sem_perm); |
6062a8dc | 442 | } else { |
1a233956 | 443 | struct sem *sem = &sma->sems[locknum]; |
6062a8dc RR |
444 | spin_unlock(&sem->lock); |
445 | } | |
6062a8dc RR |
446 | } |
447 | ||
3e148c79 | 448 | /* |
d9a605e4 | 449 | * sem_lock_(check_) routines are called in the paths where the rwsem |
3e148c79 | 450 | * is not held. |
321310ce LT |
451 | * |
452 | * The caller holds the RCU read lock. | |
3e148c79 | 453 | */ |
16df3674 DB |
454 | static inline struct sem_array *sem_obtain_object(struct ipc_namespace *ns, int id) |
455 | { | |
55b7ae50 | 456 | struct kern_ipc_perm *ipcp = ipc_obtain_object_idr(&sem_ids(ns), id); |
16df3674 DB |
457 | |
458 | if (IS_ERR(ipcp)) | |
459 | return ERR_CAST(ipcp); | |
460 | ||
461 | return container_of(ipcp, struct sem_array, sem_perm); | |
462 | } | |
463 | ||
16df3674 DB |
464 | static inline struct sem_array *sem_obtain_object_check(struct ipc_namespace *ns, |
465 | int id) | |
466 | { | |
467 | struct kern_ipc_perm *ipcp = ipc_obtain_object_check(&sem_ids(ns), id); | |
468 | ||
469 | if (IS_ERR(ipcp)) | |
470 | return ERR_CAST(ipcp); | |
b1ed88b4 | 471 | |
03f02c76 | 472 | return container_of(ipcp, struct sem_array, sem_perm); |
023a5355 ND |
473 | } |
474 | ||
6ff37972 PP |
475 | static inline void sem_lock_and_putref(struct sem_array *sma) |
476 | { | |
6062a8dc | 477 | sem_lock(sma, NULL, -1); |
dba4cdd3 | 478 | ipc_rcu_putref(&sma->sem_perm, sem_rcu_free); |
6ff37972 PP |
479 | } |
480 | ||
7ca7e564 ND |
481 | static inline void sem_rmid(struct ipc_namespace *ns, struct sem_array *s) |
482 | { | |
483 | ipc_rmid(&sem_ids(ns), &s->sem_perm); | |
484 | } | |
485 | ||
101ede01 KC |
486 | static struct sem_array *sem_alloc(size_t nsems) |
487 | { | |
488 | struct sem_array *sma; | |
489 | size_t size; | |
490 | ||
491 | if (nsems > (INT_MAX - sizeof(*sma)) / sizeof(sma->sems[0])) | |
492 | return NULL; | |
493 | ||
494 | size = sizeof(*sma) + nsems * sizeof(sma->sems[0]); | |
495 | sma = kvmalloc(size, GFP_KERNEL); | |
496 | if (unlikely(!sma)) | |
497 | return NULL; | |
498 | ||
499 | memset(sma, 0, size); | |
101ede01 KC |
500 | |
501 | return sma; | |
502 | } | |
503 | ||
f4566f04 ND |
504 | /** |
505 | * newary - Create a new semaphore set | |
506 | * @ns: namespace | |
507 | * @params: ptr to the structure that contains key, semflg and nsems | |
508 | * | |
d9a605e4 | 509 | * Called with sem_ids.rwsem held (as a writer) |
f4566f04 | 510 | */ |
7748dbfa | 511 | static int newary(struct ipc_namespace *ns, struct ipc_params *params) |
1da177e4 | 512 | { |
1da177e4 LT |
513 | int retval; |
514 | struct sem_array *sma; | |
7748dbfa ND |
515 | key_t key = params->key; |
516 | int nsems = params->u.nsems; | |
517 | int semflg = params->flg; | |
b97e820f | 518 | int i; |
1da177e4 LT |
519 | |
520 | if (!nsems) | |
521 | return -EINVAL; | |
e3893534 | 522 | if (ns->used_sems + nsems > ns->sc_semmns) |
1da177e4 LT |
523 | return -ENOSPC; |
524 | ||
101ede01 | 525 | sma = sem_alloc(nsems); |
3ab08fe2 | 526 | if (!sma) |
1da177e4 | 527 | return -ENOMEM; |
3ab08fe2 | 528 | |
1da177e4 LT |
529 | sma->sem_perm.mode = (semflg & S_IRWXUGO); |
530 | sma->sem_perm.key = key; | |
531 | ||
532 | sma->sem_perm.security = NULL; | |
aefad959 | 533 | retval = security_sem_alloc(&sma->sem_perm); |
1da177e4 | 534 | if (retval) { |
e2029dfe | 535 | kvfree(sma); |
1da177e4 LT |
536 | return retval; |
537 | } | |
538 | ||
6062a8dc | 539 | for (i = 0; i < nsems; i++) { |
1a233956 MS |
540 | INIT_LIST_HEAD(&sma->sems[i].pending_alter); |
541 | INIT_LIST_HEAD(&sma->sems[i].pending_const); | |
542 | spin_lock_init(&sma->sems[i].lock); | |
6062a8dc | 543 | } |
b97e820f MS |
544 | |
545 | sma->complex_count = 0; | |
9de5ab8a | 546 | sma->use_global_lock = USE_GLOBAL_LOCK_HYSTERESIS; |
1a82e9e1 MS |
547 | INIT_LIST_HEAD(&sma->pending_alter); |
548 | INIT_LIST_HEAD(&sma->pending_const); | |
4daa28f6 | 549 | INIT_LIST_HEAD(&sma->list_id); |
1da177e4 | 550 | sma->sem_nsems = nsems; |
e54d02b2 | 551 | sma->sem_ctime = ktime_get_real_seconds(); |
e8577d1f | 552 | |
39c96a1b | 553 | /* ipc_addid() locks sma upon success. */ |
2ec55f80 MS |
554 | retval = ipc_addid(&sem_ids(ns), &sma->sem_perm, ns->sc_semmni); |
555 | if (retval < 0) { | |
556 | call_rcu(&sma->sem_perm.rcu, sem_rcu_free); | |
557 | return retval; | |
e8577d1f MS |
558 | } |
559 | ns->used_sems += nsems; | |
560 | ||
6062a8dc | 561 | sem_unlock(sma, -1); |
6d49dab8 | 562 | rcu_read_unlock(); |
1da177e4 | 563 | |
7ca7e564 | 564 | return sma->sem_perm.id; |
1da177e4 LT |
565 | } |
566 | ||
7748dbfa | 567 | |
f4566f04 | 568 | /* |
d9a605e4 | 569 | * Called with sem_ids.rwsem and ipcp locked. |
f4566f04 | 570 | */ |
03f02c76 ND |
571 | static inline int sem_more_checks(struct kern_ipc_perm *ipcp, |
572 | struct ipc_params *params) | |
7748dbfa | 573 | { |
03f02c76 ND |
574 | struct sem_array *sma; |
575 | ||
576 | sma = container_of(ipcp, struct sem_array, sem_perm); | |
577 | if (params->u.nsems > sma->sem_nsems) | |
7748dbfa ND |
578 | return -EINVAL; |
579 | ||
580 | return 0; | |
581 | } | |
582 | ||
69894718 | 583 | long ksys_semget(key_t key, int nsems, int semflg) |
1da177e4 | 584 | { |
e3893534 | 585 | struct ipc_namespace *ns; |
eb66ec44 MK |
586 | static const struct ipc_ops sem_ops = { |
587 | .getnew = newary, | |
50ab44b1 | 588 | .associate = security_sem_associate, |
eb66ec44 MK |
589 | .more_checks = sem_more_checks, |
590 | }; | |
7748dbfa | 591 | struct ipc_params sem_params; |
e3893534 KK |
592 | |
593 | ns = current->nsproxy->ipc_ns; | |
1da177e4 | 594 | |
e3893534 | 595 | if (nsems < 0 || nsems > ns->sc_semmsl) |
1da177e4 | 596 | return -EINVAL; |
7ca7e564 | 597 | |
7748dbfa ND |
598 | sem_params.key = key; |
599 | sem_params.flg = semflg; | |
600 | sem_params.u.nsems = nsems; | |
1da177e4 | 601 | |
7748dbfa | 602 | return ipcget(ns, &sem_ids(ns), &sem_ops, &sem_params); |
1da177e4 LT |
603 | } |
604 | ||
69894718 DB |
605 | SYSCALL_DEFINE3(semget, key_t, key, int, nsems, int, semflg) |
606 | { | |
607 | return ksys_semget(key, nsems, semflg); | |
608 | } | |
609 | ||
78f5009c | 610 | /** |
4ce33ec2 DB |
611 | * perform_atomic_semop[_slow] - Attempt to perform semaphore |
612 | * operations on a given array. | |
758a6ba3 | 613 | * @sma: semaphore array |
d198cd6d | 614 | * @q: struct sem_queue that describes the operation |
758a6ba3 | 615 | * |
4ce33ec2 DB |
616 | * Caller blocking are as follows, based the value |
617 | * indicated by the semaphore operation (sem_op): | |
618 | * | |
619 | * (1) >0 never blocks. | |
620 | * (2) 0 (wait-for-zero operation): semval is non-zero. | |
621 | * (3) <0 attempting to decrement semval to a value smaller than zero. | |
622 | * | |
758a6ba3 MS |
623 | * Returns 0 if the operation was possible. |
624 | * Returns 1 if the operation is impossible, the caller must sleep. | |
4ce33ec2 | 625 | * Returns <0 for error codes. |
1da177e4 | 626 | */ |
4ce33ec2 | 627 | static int perform_atomic_semop_slow(struct sem_array *sma, struct sem_queue *q) |
1da177e4 | 628 | { |
51d6f263 EB |
629 | int result, sem_op, nsops; |
630 | struct pid *pid; | |
1da177e4 | 631 | struct sembuf *sop; |
239521f3 | 632 | struct sem *curr; |
d198cd6d MS |
633 | struct sembuf *sops; |
634 | struct sem_undo *un; | |
635 | ||
636 | sops = q->sops; | |
637 | nsops = q->nsops; | |
638 | un = q->undo; | |
1da177e4 LT |
639 | |
640 | for (sop = sops; sop < sops + nsops; sop++) { | |
1a233956 | 641 | curr = &sma->sems[sop->sem_num]; |
1da177e4 LT |
642 | sem_op = sop->sem_op; |
643 | result = curr->semval; | |
78f5009c | 644 | |
1da177e4 LT |
645 | if (!sem_op && result) |
646 | goto would_block; | |
647 | ||
648 | result += sem_op; | |
649 | if (result < 0) | |
650 | goto would_block; | |
651 | if (result > SEMVMX) | |
652 | goto out_of_range; | |
78f5009c | 653 | |
1da177e4 LT |
654 | if (sop->sem_flg & SEM_UNDO) { |
655 | int undo = un->semadj[sop->sem_num] - sem_op; | |
78f5009c | 656 | /* Exceeding the undo range is an error. */ |
1da177e4 LT |
657 | if (undo < (-SEMAEM - 1) || undo > SEMAEM) |
658 | goto out_of_range; | |
78f5009c | 659 | un->semadj[sop->sem_num] = undo; |
1da177e4 | 660 | } |
78f5009c | 661 | |
1da177e4 LT |
662 | curr->semval = result; |
663 | } | |
664 | ||
665 | sop--; | |
d198cd6d | 666 | pid = q->pid; |
1da177e4 | 667 | while (sop >= sops) { |
51d6f263 | 668 | ipc_update_pid(&sma->sems[sop->sem_num].sempid, pid); |
1da177e4 LT |
669 | sop--; |
670 | } | |
78f5009c | 671 | |
1da177e4 LT |
672 | return 0; |
673 | ||
674 | out_of_range: | |
675 | result = -ERANGE; | |
676 | goto undo; | |
677 | ||
678 | would_block: | |
ed247b7c MS |
679 | q->blocking = sop; |
680 | ||
1da177e4 LT |
681 | if (sop->sem_flg & IPC_NOWAIT) |
682 | result = -EAGAIN; | |
683 | else | |
684 | result = 1; | |
685 | ||
686 | undo: | |
687 | sop--; | |
688 | while (sop >= sops) { | |
78f5009c | 689 | sem_op = sop->sem_op; |
1a233956 | 690 | sma->sems[sop->sem_num].semval -= sem_op; |
78f5009c PM |
691 | if (sop->sem_flg & SEM_UNDO) |
692 | un->semadj[sop->sem_num] += sem_op; | |
1da177e4 LT |
693 | sop--; |
694 | } | |
695 | ||
696 | return result; | |
697 | } | |
698 | ||
4ce33ec2 DB |
699 | static int perform_atomic_semop(struct sem_array *sma, struct sem_queue *q) |
700 | { | |
701 | int result, sem_op, nsops; | |
702 | struct sembuf *sop; | |
703 | struct sem *curr; | |
704 | struct sembuf *sops; | |
705 | struct sem_undo *un; | |
706 | ||
707 | sops = q->sops; | |
708 | nsops = q->nsops; | |
709 | un = q->undo; | |
710 | ||
711 | if (unlikely(q->dupsop)) | |
712 | return perform_atomic_semop_slow(sma, q); | |
713 | ||
714 | /* | |
715 | * We scan the semaphore set twice, first to ensure that the entire | |
716 | * operation can succeed, therefore avoiding any pointless writes | |
717 | * to shared memory and having to undo such changes in order to block | |
718 | * until the operations can go through. | |
719 | */ | |
720 | for (sop = sops; sop < sops + nsops; sop++) { | |
1a233956 | 721 | curr = &sma->sems[sop->sem_num]; |
4ce33ec2 DB |
722 | sem_op = sop->sem_op; |
723 | result = curr->semval; | |
724 | ||
725 | if (!sem_op && result) | |
726 | goto would_block; /* wait-for-zero */ | |
727 | ||
728 | result += sem_op; | |
729 | if (result < 0) | |
730 | goto would_block; | |
731 | ||
732 | if (result > SEMVMX) | |
733 | return -ERANGE; | |
734 | ||
735 | if (sop->sem_flg & SEM_UNDO) { | |
736 | int undo = un->semadj[sop->sem_num] - sem_op; | |
737 | ||
738 | /* Exceeding the undo range is an error. */ | |
739 | if (undo < (-SEMAEM - 1) || undo > SEMAEM) | |
740 | return -ERANGE; | |
741 | } | |
742 | } | |
743 | ||
744 | for (sop = sops; sop < sops + nsops; sop++) { | |
1a233956 | 745 | curr = &sma->sems[sop->sem_num]; |
4ce33ec2 DB |
746 | sem_op = sop->sem_op; |
747 | result = curr->semval; | |
748 | ||
749 | if (sop->sem_flg & SEM_UNDO) { | |
750 | int undo = un->semadj[sop->sem_num] - sem_op; | |
751 | ||
752 | un->semadj[sop->sem_num] = undo; | |
753 | } | |
754 | curr->semval += sem_op; | |
51d6f263 | 755 | ipc_update_pid(&curr->sempid, q->pid); |
4ce33ec2 DB |
756 | } |
757 | ||
758 | return 0; | |
759 | ||
760 | would_block: | |
761 | q->blocking = sop; | |
762 | return sop->sem_flg & IPC_NOWAIT ? -EAGAIN : 1; | |
763 | } | |
764 | ||
9ae949fa DB |
765 | static inline void wake_up_sem_queue_prepare(struct sem_queue *q, int error, |
766 | struct wake_q_head *wake_q) | |
0a2b9d4c | 767 | { |
9ae949fa DB |
768 | wake_q_add(wake_q, q->sleeper); |
769 | /* | |
770 | * Rely on the above implicit barrier, such that we can | |
771 | * ensure that we hold reference to the task before setting | |
772 | * q->status. Otherwise we could race with do_exit if the | |
773 | * task is awoken by an external event before calling | |
774 | * wake_up_process(). | |
775 | */ | |
776 | WRITE_ONCE(q->status, error); | |
d4212093 NP |
777 | } |
778 | ||
b97e820f MS |
779 | static void unlink_queue(struct sem_array *sma, struct sem_queue *q) |
780 | { | |
781 | list_del(&q->list); | |
9f1bc2c9 | 782 | if (q->nsops > 1) |
b97e820f MS |
783 | sma->complex_count--; |
784 | } | |
785 | ||
fd5db422 MS |
786 | /** check_restart(sma, q) |
787 | * @sma: semaphore array | |
788 | * @q: the operation that just completed | |
789 | * | |
790 | * update_queue is O(N^2) when it restarts scanning the whole queue of | |
791 | * waiting operations. Therefore this function checks if the restart is | |
792 | * really necessary. It is called after a previously waiting operation | |
1a82e9e1 MS |
793 | * modified the array. |
794 | * Note that wait-for-zero operations are handled without restart. | |
fd5db422 | 795 | */ |
4663d3e8 | 796 | static inline int check_restart(struct sem_array *sma, struct sem_queue *q) |
fd5db422 | 797 | { |
1a82e9e1 MS |
798 | /* pending complex alter operations are too difficult to analyse */ |
799 | if (!list_empty(&sma->pending_alter)) | |
fd5db422 MS |
800 | return 1; |
801 | ||
802 | /* we were a sleeping complex operation. Too difficult */ | |
803 | if (q->nsops > 1) | |
804 | return 1; | |
805 | ||
1a82e9e1 MS |
806 | /* It is impossible that someone waits for the new value: |
807 | * - complex operations always restart. | |
808 | * - wait-for-zero are handled seperately. | |
809 | * - q is a previously sleeping simple operation that | |
810 | * altered the array. It must be a decrement, because | |
811 | * simple increments never sleep. | |
812 | * - If there are older (higher priority) decrements | |
813 | * in the queue, then they have observed the original | |
814 | * semval value and couldn't proceed. The operation | |
815 | * decremented to value - thus they won't proceed either. | |
816 | */ | |
817 | return 0; | |
818 | } | |
fd5db422 | 819 | |
1a82e9e1 | 820 | /** |
8001c858 | 821 | * wake_const_ops - wake up non-alter tasks |
1a82e9e1 MS |
822 | * @sma: semaphore array. |
823 | * @semnum: semaphore that was modified. | |
9ae949fa | 824 | * @wake_q: lockless wake-queue head. |
1a82e9e1 MS |
825 | * |
826 | * wake_const_ops must be called after a semaphore in a semaphore array | |
827 | * was set to 0. If complex const operations are pending, wake_const_ops must | |
828 | * be called with semnum = -1, as well as with the number of each modified | |
829 | * semaphore. | |
9ae949fa | 830 | * The tasks that must be woken up are added to @wake_q. The return code |
1a82e9e1 MS |
831 | * is stored in q->pid. |
832 | * The function returns 1 if at least one operation was completed successfully. | |
833 | */ | |
834 | static int wake_const_ops(struct sem_array *sma, int semnum, | |
9ae949fa | 835 | struct wake_q_head *wake_q) |
1a82e9e1 | 836 | { |
f150f02c | 837 | struct sem_queue *q, *tmp; |
1a82e9e1 MS |
838 | struct list_head *pending_list; |
839 | int semop_completed = 0; | |
840 | ||
841 | if (semnum == -1) | |
842 | pending_list = &sma->pending_const; | |
843 | else | |
1a233956 | 844 | pending_list = &sma->sems[semnum].pending_const; |
fd5db422 | 845 | |
f150f02c DB |
846 | list_for_each_entry_safe(q, tmp, pending_list, list) { |
847 | int error = perform_atomic_semop(sma, q); | |
1a82e9e1 | 848 | |
f150f02c DB |
849 | if (error > 0) |
850 | continue; | |
851 | /* operation completed, remove from queue & wakeup */ | |
852 | unlink_queue(sma, q); | |
1a82e9e1 | 853 | |
f150f02c DB |
854 | wake_up_sem_queue_prepare(q, error, wake_q); |
855 | if (error == 0) | |
856 | semop_completed = 1; | |
1a82e9e1 | 857 | } |
f150f02c | 858 | |
1a82e9e1 MS |
859 | return semop_completed; |
860 | } | |
861 | ||
862 | /** | |
8001c858 | 863 | * do_smart_wakeup_zero - wakeup all wait for zero tasks |
1a82e9e1 MS |
864 | * @sma: semaphore array |
865 | * @sops: operations that were performed | |
866 | * @nsops: number of operations | |
9ae949fa | 867 | * @wake_q: lockless wake-queue head |
1a82e9e1 | 868 | * |
8001c858 DB |
869 | * Checks all required queue for wait-for-zero operations, based |
870 | * on the actual changes that were performed on the semaphore array. | |
1a82e9e1 MS |
871 | * The function returns 1 if at least one operation was completed successfully. |
872 | */ | |
873 | static int do_smart_wakeup_zero(struct sem_array *sma, struct sembuf *sops, | |
9ae949fa | 874 | int nsops, struct wake_q_head *wake_q) |
1a82e9e1 MS |
875 | { |
876 | int i; | |
877 | int semop_completed = 0; | |
878 | int got_zero = 0; | |
879 | ||
880 | /* first: the per-semaphore queues, if known */ | |
881 | if (sops) { | |
882 | for (i = 0; i < nsops; i++) { | |
883 | int num = sops[i].sem_num; | |
884 | ||
1a233956 | 885 | if (sma->sems[num].semval == 0) { |
1a82e9e1 | 886 | got_zero = 1; |
9ae949fa | 887 | semop_completed |= wake_const_ops(sma, num, wake_q); |
1a82e9e1 MS |
888 | } |
889 | } | |
890 | } else { | |
891 | /* | |
892 | * No sops means modified semaphores not known. | |
893 | * Assume all were changed. | |
fd5db422 | 894 | */ |
1a82e9e1 | 895 | for (i = 0; i < sma->sem_nsems; i++) { |
1a233956 | 896 | if (sma->sems[i].semval == 0) { |
1a82e9e1 | 897 | got_zero = 1; |
9ae949fa | 898 | semop_completed |= wake_const_ops(sma, i, wake_q); |
1a82e9e1 MS |
899 | } |
900 | } | |
fd5db422 MS |
901 | } |
902 | /* | |
1a82e9e1 MS |
903 | * If one of the modified semaphores got 0, |
904 | * then check the global queue, too. | |
fd5db422 | 905 | */ |
1a82e9e1 | 906 | if (got_zero) |
9ae949fa | 907 | semop_completed |= wake_const_ops(sma, -1, wake_q); |
fd5db422 | 908 | |
1a82e9e1 | 909 | return semop_completed; |
fd5db422 MS |
910 | } |
911 | ||
636c6be8 MS |
912 | |
913 | /** | |
8001c858 | 914 | * update_queue - look for tasks that can be completed. |
636c6be8 MS |
915 | * @sma: semaphore array. |
916 | * @semnum: semaphore that was modified. | |
9ae949fa | 917 | * @wake_q: lockless wake-queue head. |
636c6be8 MS |
918 | * |
919 | * update_queue must be called after a semaphore in a semaphore array | |
9f1bc2c9 RR |
920 | * was modified. If multiple semaphores were modified, update_queue must |
921 | * be called with semnum = -1, as well as with the number of each modified | |
922 | * semaphore. | |
9ae949fa | 923 | * The tasks that must be woken up are added to @wake_q. The return code |
0a2b9d4c | 924 | * is stored in q->pid. |
1a82e9e1 MS |
925 | * The function internally checks if const operations can now succeed. |
926 | * | |
0a2b9d4c | 927 | * The function return 1 if at least one semop was completed successfully. |
1da177e4 | 928 | */ |
9ae949fa | 929 | static int update_queue(struct sem_array *sma, int semnum, struct wake_q_head *wake_q) |
1da177e4 | 930 | { |
f150f02c | 931 | struct sem_queue *q, *tmp; |
636c6be8 | 932 | struct list_head *pending_list; |
0a2b9d4c | 933 | int semop_completed = 0; |
636c6be8 | 934 | |
9f1bc2c9 | 935 | if (semnum == -1) |
1a82e9e1 | 936 | pending_list = &sma->pending_alter; |
9f1bc2c9 | 937 | else |
1a233956 | 938 | pending_list = &sma->sems[semnum].pending_alter; |
9cad200c NP |
939 | |
940 | again: | |
f150f02c | 941 | list_for_each_entry_safe(q, tmp, pending_list, list) { |
fd5db422 | 942 | int error, restart; |
636c6be8 | 943 | |
d987f8b2 MS |
944 | /* If we are scanning the single sop, per-semaphore list of |
945 | * one semaphore and that semaphore is 0, then it is not | |
1a82e9e1 | 946 | * necessary to scan further: simple increments |
d987f8b2 MS |
947 | * that affect only one entry succeed immediately and cannot |
948 | * be in the per semaphore pending queue, and decrements | |
949 | * cannot be successful if the value is already 0. | |
950 | */ | |
1a233956 | 951 | if (semnum != -1 && sma->sems[semnum].semval == 0) |
d987f8b2 MS |
952 | break; |
953 | ||
d198cd6d | 954 | error = perform_atomic_semop(sma, q); |
1da177e4 LT |
955 | |
956 | /* Does q->sleeper still need to sleep? */ | |
9cad200c NP |
957 | if (error > 0) |
958 | continue; | |
959 | ||
b97e820f | 960 | unlink_queue(sma, q); |
9cad200c | 961 | |
0a2b9d4c | 962 | if (error) { |
fd5db422 | 963 | restart = 0; |
0a2b9d4c MS |
964 | } else { |
965 | semop_completed = 1; | |
9ae949fa | 966 | do_smart_wakeup_zero(sma, q->sops, q->nsops, wake_q); |
fd5db422 | 967 | restart = check_restart(sma, q); |
0a2b9d4c | 968 | } |
fd5db422 | 969 | |
9ae949fa | 970 | wake_up_sem_queue_prepare(q, error, wake_q); |
fd5db422 | 971 | if (restart) |
9cad200c | 972 | goto again; |
1da177e4 | 973 | } |
0a2b9d4c | 974 | return semop_completed; |
1da177e4 LT |
975 | } |
976 | ||
0e8c6656 | 977 | /** |
8001c858 | 978 | * set_semotime - set sem_otime |
0e8c6656 MS |
979 | * @sma: semaphore array |
980 | * @sops: operations that modified the array, may be NULL | |
981 | * | |
982 | * sem_otime is replicated to avoid cache line trashing. | |
983 | * This function sets one instance to the current time. | |
984 | */ | |
985 | static void set_semotime(struct sem_array *sma, struct sembuf *sops) | |
986 | { | |
987 | if (sops == NULL) { | |
2a70b787 | 988 | sma->sems[0].sem_otime = ktime_get_real_seconds(); |
0e8c6656 | 989 | } else { |
1a233956 | 990 | sma->sems[sops[0].sem_num].sem_otime = |
2a70b787 | 991 | ktime_get_real_seconds(); |
0e8c6656 MS |
992 | } |
993 | } | |
994 | ||
0a2b9d4c | 995 | /** |
8001c858 | 996 | * do_smart_update - optimized update_queue |
fd5db422 MS |
997 | * @sma: semaphore array |
998 | * @sops: operations that were performed | |
999 | * @nsops: number of operations | |
0a2b9d4c | 1000 | * @otime: force setting otime |
9ae949fa | 1001 | * @wake_q: lockless wake-queue head |
fd5db422 | 1002 | * |
1a82e9e1 MS |
1003 | * do_smart_update() does the required calls to update_queue and wakeup_zero, |
1004 | * based on the actual changes that were performed on the semaphore array. | |
0a2b9d4c | 1005 | * Note that the function does not do the actual wake-up: the caller is |
9ae949fa | 1006 | * responsible for calling wake_up_q(). |
0a2b9d4c | 1007 | * It is safe to perform this call after dropping all locks. |
fd5db422 | 1008 | */ |
0a2b9d4c | 1009 | static void do_smart_update(struct sem_array *sma, struct sembuf *sops, int nsops, |
9ae949fa | 1010 | int otime, struct wake_q_head *wake_q) |
fd5db422 MS |
1011 | { |
1012 | int i; | |
1013 | ||
9ae949fa | 1014 | otime |= do_smart_wakeup_zero(sma, sops, nsops, wake_q); |
1a82e9e1 | 1015 | |
f269f40a MS |
1016 | if (!list_empty(&sma->pending_alter)) { |
1017 | /* semaphore array uses the global queue - just process it. */ | |
9ae949fa | 1018 | otime |= update_queue(sma, -1, wake_q); |
f269f40a MS |
1019 | } else { |
1020 | if (!sops) { | |
1021 | /* | |
1022 | * No sops, thus the modified semaphores are not | |
1023 | * known. Check all. | |
1024 | */ | |
1025 | for (i = 0; i < sma->sem_nsems; i++) | |
9ae949fa | 1026 | otime |= update_queue(sma, i, wake_q); |
f269f40a MS |
1027 | } else { |
1028 | /* | |
1029 | * Check the semaphores that were increased: | |
1030 | * - No complex ops, thus all sleeping ops are | |
1031 | * decrease. | |
1032 | * - if we decreased the value, then any sleeping | |
1033 | * semaphore ops wont be able to run: If the | |
1034 | * previous value was too small, then the new | |
1035 | * value will be too small, too. | |
1036 | */ | |
1037 | for (i = 0; i < nsops; i++) { | |
1038 | if (sops[i].sem_op > 0) { | |
1039 | otime |= update_queue(sma, | |
9ae949fa | 1040 | sops[i].sem_num, wake_q); |
f269f40a | 1041 | } |
ab465df9 | 1042 | } |
9f1bc2c9 | 1043 | } |
fd5db422 | 1044 | } |
0e8c6656 MS |
1045 | if (otime) |
1046 | set_semotime(sma, sops); | |
fd5db422 MS |
1047 | } |
1048 | ||
2f2ed41d | 1049 | /* |
b220c57a | 1050 | * check_qop: Test if a queued operation sleeps on the semaphore semnum |
2f2ed41d MS |
1051 | */ |
1052 | static int check_qop(struct sem_array *sma, int semnum, struct sem_queue *q, | |
1053 | bool count_zero) | |
1054 | { | |
b220c57a | 1055 | struct sembuf *sop = q->blocking; |
2f2ed41d | 1056 | |
9b44ee2e MS |
1057 | /* |
1058 | * Linux always (since 0.99.10) reported a task as sleeping on all | |
1059 | * semaphores. This violates SUS, therefore it was changed to the | |
1060 | * standard compliant behavior. | |
1061 | * Give the administrators a chance to notice that an application | |
1062 | * might misbehave because it relies on the Linux behavior. | |
1063 | */ | |
1064 | pr_info_once("semctl(GETNCNT/GETZCNT) is since 3.16 Single Unix Specification compliant.\n" | |
1065 | "The task %s (%d) triggered the difference, watch for misbehavior.\n", | |
1066 | current->comm, task_pid_nr(current)); | |
1067 | ||
b220c57a MS |
1068 | if (sop->sem_num != semnum) |
1069 | return 0; | |
2f2ed41d | 1070 | |
b220c57a MS |
1071 | if (count_zero && sop->sem_op == 0) |
1072 | return 1; | |
1073 | if (!count_zero && sop->sem_op < 0) | |
1074 | return 1; | |
1075 | ||
1076 | return 0; | |
2f2ed41d MS |
1077 | } |
1078 | ||
1da177e4 LT |
1079 | /* The following counts are associated to each semaphore: |
1080 | * semncnt number of tasks waiting on semval being nonzero | |
1081 | * semzcnt number of tasks waiting on semval being zero | |
b220c57a MS |
1082 | * |
1083 | * Per definition, a task waits only on the semaphore of the first semop | |
1084 | * that cannot proceed, even if additional operation would block, too. | |
1da177e4 | 1085 | */ |
2f2ed41d MS |
1086 | static int count_semcnt(struct sem_array *sma, ushort semnum, |
1087 | bool count_zero) | |
1da177e4 | 1088 | { |
2f2ed41d | 1089 | struct list_head *l; |
239521f3 | 1090 | struct sem_queue *q; |
2f2ed41d | 1091 | int semcnt; |
1da177e4 | 1092 | |
2f2ed41d MS |
1093 | semcnt = 0; |
1094 | /* First: check the simple operations. They are easy to evaluate */ | |
1095 | if (count_zero) | |
1a233956 | 1096 | l = &sma->sems[semnum].pending_const; |
2f2ed41d | 1097 | else |
1a233956 | 1098 | l = &sma->sems[semnum].pending_alter; |
1da177e4 | 1099 | |
2f2ed41d MS |
1100 | list_for_each_entry(q, l, list) { |
1101 | /* all task on a per-semaphore list sleep on exactly | |
1102 | * that semaphore | |
1103 | */ | |
1104 | semcnt++; | |
ebc2e5e6 RR |
1105 | } |
1106 | ||
2f2ed41d | 1107 | /* Then: check the complex operations. */ |
1994862d | 1108 | list_for_each_entry(q, &sma->pending_alter, list) { |
2f2ed41d MS |
1109 | semcnt += check_qop(sma, semnum, q, count_zero); |
1110 | } | |
1111 | if (count_zero) { | |
1112 | list_for_each_entry(q, &sma->pending_const, list) { | |
1113 | semcnt += check_qop(sma, semnum, q, count_zero); | |
1114 | } | |
1994862d | 1115 | } |
2f2ed41d | 1116 | return semcnt; |
1da177e4 LT |
1117 | } |
1118 | ||
d9a605e4 DB |
1119 | /* Free a semaphore set. freeary() is called with sem_ids.rwsem locked |
1120 | * as a writer and the spinlock for this semaphore set hold. sem_ids.rwsem | |
3e148c79 | 1121 | * remains locked on exit. |
1da177e4 | 1122 | */ |
01b8b07a | 1123 | static void freeary(struct ipc_namespace *ns, struct kern_ipc_perm *ipcp) |
1da177e4 | 1124 | { |
380af1b3 MS |
1125 | struct sem_undo *un, *tu; |
1126 | struct sem_queue *q, *tq; | |
01b8b07a | 1127 | struct sem_array *sma = container_of(ipcp, struct sem_array, sem_perm); |
9f1bc2c9 | 1128 | int i; |
9ae949fa | 1129 | DEFINE_WAKE_Q(wake_q); |
1da177e4 | 1130 | |
380af1b3 | 1131 | /* Free the existing undo structures for this semaphore set. */ |
cf9d5d78 | 1132 | ipc_assert_locked_object(&sma->sem_perm); |
380af1b3 MS |
1133 | list_for_each_entry_safe(un, tu, &sma->list_id, list_id) { |
1134 | list_del(&un->list_id); | |
1135 | spin_lock(&un->ulp->lock); | |
1da177e4 | 1136 | un->semid = -1; |
380af1b3 MS |
1137 | list_del_rcu(&un->list_proc); |
1138 | spin_unlock(&un->ulp->lock); | |
693a8b6e | 1139 | kfree_rcu(un, rcu); |
380af1b3 | 1140 | } |
1da177e4 LT |
1141 | |
1142 | /* Wake up all pending processes and let them fail with EIDRM. */ | |
1a82e9e1 MS |
1143 | list_for_each_entry_safe(q, tq, &sma->pending_const, list) { |
1144 | unlink_queue(sma, q); | |
9ae949fa | 1145 | wake_up_sem_queue_prepare(q, -EIDRM, &wake_q); |
1a82e9e1 MS |
1146 | } |
1147 | ||
1148 | list_for_each_entry_safe(q, tq, &sma->pending_alter, list) { | |
b97e820f | 1149 | unlink_queue(sma, q); |
9ae949fa | 1150 | wake_up_sem_queue_prepare(q, -EIDRM, &wake_q); |
1da177e4 | 1151 | } |
9f1bc2c9 | 1152 | for (i = 0; i < sma->sem_nsems; i++) { |
1a233956 | 1153 | struct sem *sem = &sma->sems[i]; |
1a82e9e1 MS |
1154 | list_for_each_entry_safe(q, tq, &sem->pending_const, list) { |
1155 | unlink_queue(sma, q); | |
9ae949fa | 1156 | wake_up_sem_queue_prepare(q, -EIDRM, &wake_q); |
1a82e9e1 MS |
1157 | } |
1158 | list_for_each_entry_safe(q, tq, &sem->pending_alter, list) { | |
9f1bc2c9 | 1159 | unlink_queue(sma, q); |
9ae949fa | 1160 | wake_up_sem_queue_prepare(q, -EIDRM, &wake_q); |
9f1bc2c9 | 1161 | } |
51d6f263 | 1162 | ipc_update_pid(&sem->sempid, NULL); |
9f1bc2c9 | 1163 | } |
1da177e4 | 1164 | |
7ca7e564 ND |
1165 | /* Remove the semaphore set from the IDR */ |
1166 | sem_rmid(ns, sma); | |
6062a8dc | 1167 | sem_unlock(sma, -1); |
6d49dab8 | 1168 | rcu_read_unlock(); |
1da177e4 | 1169 | |
9ae949fa | 1170 | wake_up_q(&wake_q); |
e3893534 | 1171 | ns->used_sems -= sma->sem_nsems; |
dba4cdd3 | 1172 | ipc_rcu_putref(&sma->sem_perm, sem_rcu_free); |
1da177e4 LT |
1173 | } |
1174 | ||
1175 | static unsigned long copy_semid_to_user(void __user *buf, struct semid64_ds *in, int version) | |
1176 | { | |
239521f3 | 1177 | switch (version) { |
1da177e4 LT |
1178 | case IPC_64: |
1179 | return copy_to_user(buf, in, sizeof(*in)); | |
1180 | case IPC_OLD: | |
1181 | { | |
1182 | struct semid_ds out; | |
1183 | ||
982f7c2b DR |
1184 | memset(&out, 0, sizeof(out)); |
1185 | ||
1da177e4 LT |
1186 | ipc64_perm_to_ipc_perm(&in->sem_perm, &out.sem_perm); |
1187 | ||
1188 | out.sem_otime = in->sem_otime; | |
1189 | out.sem_ctime = in->sem_ctime; | |
1190 | out.sem_nsems = in->sem_nsems; | |
1191 | ||
1192 | return copy_to_user(buf, &out, sizeof(out)); | |
1193 | } | |
1194 | default: | |
1195 | return -EINVAL; | |
1196 | } | |
1197 | } | |
1198 | ||
e54d02b2 | 1199 | static time64_t get_semotime(struct sem_array *sma) |
d12e1e50 MS |
1200 | { |
1201 | int i; | |
e54d02b2 | 1202 | time64_t res; |
d12e1e50 | 1203 | |
1a233956 | 1204 | res = sma->sems[0].sem_otime; |
d12e1e50 | 1205 | for (i = 1; i < sma->sem_nsems; i++) { |
e54d02b2 | 1206 | time64_t to = sma->sems[i].sem_otime; |
d12e1e50 MS |
1207 | |
1208 | if (to > res) | |
1209 | res = to; | |
1210 | } | |
1211 | return res; | |
1212 | } | |
1213 | ||
45a4a64a AV |
1214 | static int semctl_stat(struct ipc_namespace *ns, int semid, |
1215 | int cmd, struct semid64_ds *semid64) | |
1da177e4 | 1216 | { |
1da177e4 | 1217 | struct sem_array *sma; |
c2ab975c | 1218 | time64_t semotime; |
45a4a64a AV |
1219 | int id = 0; |
1220 | int err; | |
1da177e4 | 1221 | |
45a4a64a | 1222 | memset(semid64, 0, sizeof(*semid64)); |
46c0a8ca | 1223 | |
45a4a64a | 1224 | rcu_read_lock(); |
a280d6dc | 1225 | if (cmd == SEM_STAT || cmd == SEM_STAT_ANY) { |
45a4a64a AV |
1226 | sma = sem_obtain_object(ns, semid); |
1227 | if (IS_ERR(sma)) { | |
1228 | err = PTR_ERR(sma); | |
1229 | goto out_unlock; | |
1230 | } | |
1231 | id = sma->sem_perm.id; | |
a280d6dc | 1232 | } else { /* IPC_STAT */ |
45a4a64a AV |
1233 | sma = sem_obtain_object_check(ns, semid); |
1234 | if (IS_ERR(sma)) { | |
1235 | err = PTR_ERR(sma); | |
1236 | goto out_unlock; | |
1da177e4 | 1237 | } |
1da177e4 | 1238 | } |
1da177e4 | 1239 | |
a280d6dc DB |
1240 | /* see comment for SHM_STAT_ANY */ |
1241 | if (cmd == SEM_STAT_ANY) | |
1242 | audit_ipc_obj(&sma->sem_perm); | |
1243 | else { | |
1244 | err = -EACCES; | |
1245 | if (ipcperms(ns, &sma->sem_perm, S_IRUGO)) | |
1246 | goto out_unlock; | |
1247 | } | |
1da177e4 | 1248 | |
aefad959 | 1249 | err = security_sem_semctl(&sma->sem_perm, cmd); |
45a4a64a AV |
1250 | if (err) |
1251 | goto out_unlock; | |
1da177e4 | 1252 | |
87ad4b0d PM |
1253 | ipc_lock_object(&sma->sem_perm); |
1254 | ||
1255 | if (!ipc_valid_object(&sma->sem_perm)) { | |
1256 | ipc_unlock_object(&sma->sem_perm); | |
1257 | err = -EIDRM; | |
1258 | goto out_unlock; | |
1259 | } | |
1260 | ||
45a4a64a | 1261 | kernel_to_ipc64_perm(&sma->sem_perm, &semid64->sem_perm); |
c2ab975c AB |
1262 | semotime = get_semotime(sma); |
1263 | semid64->sem_otime = semotime; | |
45a4a64a | 1264 | semid64->sem_ctime = sma->sem_ctime; |
c2ab975c AB |
1265 | #ifndef CONFIG_64BIT |
1266 | semid64->sem_otime_high = semotime >> 32; | |
1267 | semid64->sem_ctime_high = sma->sem_ctime >> 32; | |
1268 | #endif | |
45a4a64a | 1269 | semid64->sem_nsems = sma->sem_nsems; |
87ad4b0d PM |
1270 | |
1271 | ipc_unlock_object(&sma->sem_perm); | |
45a4a64a AV |
1272 | rcu_read_unlock(); |
1273 | return id; | |
1da177e4 | 1274 | |
1da177e4 | 1275 | out_unlock: |
16df3674 | 1276 | rcu_read_unlock(); |
1da177e4 LT |
1277 | return err; |
1278 | } | |
1279 | ||
45a4a64a AV |
1280 | static int semctl_info(struct ipc_namespace *ns, int semid, |
1281 | int cmd, void __user *p) | |
1282 | { | |
1283 | struct seminfo seminfo; | |
1284 | int max_id; | |
1285 | int err; | |
1286 | ||
1287 | err = security_sem_semctl(NULL, cmd); | |
1288 | if (err) | |
1289 | return err; | |
1290 | ||
1291 | memset(&seminfo, 0, sizeof(seminfo)); | |
1292 | seminfo.semmni = ns->sc_semmni; | |
1293 | seminfo.semmns = ns->sc_semmns; | |
1294 | seminfo.semmsl = ns->sc_semmsl; | |
1295 | seminfo.semopm = ns->sc_semopm; | |
1296 | seminfo.semvmx = SEMVMX; | |
1297 | seminfo.semmnu = SEMMNU; | |
1298 | seminfo.semmap = SEMMAP; | |
1299 | seminfo.semume = SEMUME; | |
1300 | down_read(&sem_ids(ns).rwsem); | |
1301 | if (cmd == SEM_INFO) { | |
1302 | seminfo.semusz = sem_ids(ns).in_use; | |
1303 | seminfo.semaem = ns->used_sems; | |
1304 | } else { | |
1305 | seminfo.semusz = SEMUSZ; | |
1306 | seminfo.semaem = SEMAEM; | |
1307 | } | |
1308 | max_id = ipc_get_maxid(&sem_ids(ns)); | |
1309 | up_read(&sem_ids(ns).rwsem); | |
1310 | if (copy_to_user(p, &seminfo, sizeof(struct seminfo))) | |
1311 | return -EFAULT; | |
1312 | return (max_id < 0) ? 0 : max_id; | |
1313 | } | |
1314 | ||
e1fd1f49 | 1315 | static int semctl_setval(struct ipc_namespace *ns, int semid, int semnum, |
45a4a64a | 1316 | int val) |
e1fd1f49 AV |
1317 | { |
1318 | struct sem_undo *un; | |
1319 | struct sem_array *sma; | |
239521f3 | 1320 | struct sem *curr; |
45a4a64a | 1321 | int err; |
9ae949fa DB |
1322 | DEFINE_WAKE_Q(wake_q); |
1323 | ||
6062a8dc RR |
1324 | if (val > SEMVMX || val < 0) |
1325 | return -ERANGE; | |
e1fd1f49 | 1326 | |
6062a8dc RR |
1327 | rcu_read_lock(); |
1328 | sma = sem_obtain_object_check(ns, semid); | |
1329 | if (IS_ERR(sma)) { | |
1330 | rcu_read_unlock(); | |
1331 | return PTR_ERR(sma); | |
1332 | } | |
1333 | ||
1334 | if (semnum < 0 || semnum >= sma->sem_nsems) { | |
1335 | rcu_read_unlock(); | |
1336 | return -EINVAL; | |
1337 | } | |
1338 | ||
1339 | ||
1340 | if (ipcperms(ns, &sma->sem_perm, S_IWUGO)) { | |
1341 | rcu_read_unlock(); | |
1342 | return -EACCES; | |
1343 | } | |
e1fd1f49 | 1344 | |
aefad959 | 1345 | err = security_sem_semctl(&sma->sem_perm, SETVAL); |
6062a8dc RR |
1346 | if (err) { |
1347 | rcu_read_unlock(); | |
1348 | return -EACCES; | |
1349 | } | |
e1fd1f49 | 1350 | |
6062a8dc | 1351 | sem_lock(sma, NULL, -1); |
e1fd1f49 | 1352 | |
0f3d2b01 | 1353 | if (!ipc_valid_object(&sma->sem_perm)) { |
6e224f94 MS |
1354 | sem_unlock(sma, -1); |
1355 | rcu_read_unlock(); | |
1356 | return -EIDRM; | |
1357 | } | |
1358 | ||
1a233956 | 1359 | curr = &sma->sems[semnum]; |
e1fd1f49 | 1360 | |
cf9d5d78 | 1361 | ipc_assert_locked_object(&sma->sem_perm); |
e1fd1f49 AV |
1362 | list_for_each_entry(un, &sma->list_id, list_id) |
1363 | un->semadj[semnum] = 0; | |
1364 | ||
1365 | curr->semval = val; | |
51d6f263 | 1366 | ipc_update_pid(&curr->sempid, task_tgid(current)); |
e54d02b2 | 1367 | sma->sem_ctime = ktime_get_real_seconds(); |
e1fd1f49 | 1368 | /* maybe some queued-up processes were waiting for this */ |
9ae949fa | 1369 | do_smart_update(sma, NULL, 0, 0, &wake_q); |
6062a8dc | 1370 | sem_unlock(sma, -1); |
6d49dab8 | 1371 | rcu_read_unlock(); |
9ae949fa | 1372 | wake_up_q(&wake_q); |
6062a8dc | 1373 | return 0; |
e1fd1f49 AV |
1374 | } |
1375 | ||
e3893534 | 1376 | static int semctl_main(struct ipc_namespace *ns, int semid, int semnum, |
e1fd1f49 | 1377 | int cmd, void __user *p) |
1da177e4 LT |
1378 | { |
1379 | struct sem_array *sma; | |
239521f3 | 1380 | struct sem *curr; |
16df3674 | 1381 | int err, nsems; |
1da177e4 | 1382 | ushort fast_sem_io[SEMMSL_FAST]; |
239521f3 | 1383 | ushort *sem_io = fast_sem_io; |
9ae949fa | 1384 | DEFINE_WAKE_Q(wake_q); |
16df3674 DB |
1385 | |
1386 | rcu_read_lock(); | |
1387 | sma = sem_obtain_object_check(ns, semid); | |
1388 | if (IS_ERR(sma)) { | |
1389 | rcu_read_unlock(); | |
023a5355 | 1390 | return PTR_ERR(sma); |
16df3674 | 1391 | } |
1da177e4 LT |
1392 | |
1393 | nsems = sma->sem_nsems; | |
1394 | ||
1da177e4 | 1395 | err = -EACCES; |
c728b9c8 LT |
1396 | if (ipcperms(ns, &sma->sem_perm, cmd == SETALL ? S_IWUGO : S_IRUGO)) |
1397 | goto out_rcu_wakeup; | |
1da177e4 | 1398 | |
aefad959 | 1399 | err = security_sem_semctl(&sma->sem_perm, cmd); |
c728b9c8 LT |
1400 | if (err) |
1401 | goto out_rcu_wakeup; | |
1da177e4 LT |
1402 | |
1403 | err = -EACCES; | |
1404 | switch (cmd) { | |
1405 | case GETALL: | |
1406 | { | |
e1fd1f49 | 1407 | ushort __user *array = p; |
1da177e4 LT |
1408 | int i; |
1409 | ||
ce857229 | 1410 | sem_lock(sma, NULL, -1); |
0f3d2b01 | 1411 | if (!ipc_valid_object(&sma->sem_perm)) { |
6e224f94 MS |
1412 | err = -EIDRM; |
1413 | goto out_unlock; | |
1414 | } | |
239521f3 | 1415 | if (nsems > SEMMSL_FAST) { |
dba4cdd3 | 1416 | if (!ipc_rcu_getref(&sma->sem_perm)) { |
ce857229 | 1417 | err = -EIDRM; |
6e224f94 | 1418 | goto out_unlock; |
ce857229 AV |
1419 | } |
1420 | sem_unlock(sma, -1); | |
6d49dab8 | 1421 | rcu_read_unlock(); |
f8dbe8d2 KC |
1422 | sem_io = kvmalloc_array(nsems, sizeof(ushort), |
1423 | GFP_KERNEL); | |
239521f3 | 1424 | if (sem_io == NULL) { |
dba4cdd3 | 1425 | ipc_rcu_putref(&sma->sem_perm, sem_rcu_free); |
1da177e4 LT |
1426 | return -ENOMEM; |
1427 | } | |
1428 | ||
4091fd94 | 1429 | rcu_read_lock(); |
6ff37972 | 1430 | sem_lock_and_putref(sma); |
0f3d2b01 | 1431 | if (!ipc_valid_object(&sma->sem_perm)) { |
1da177e4 | 1432 | err = -EIDRM; |
6e224f94 | 1433 | goto out_unlock; |
1da177e4 | 1434 | } |
ce857229 | 1435 | } |
1da177e4 | 1436 | for (i = 0; i < sma->sem_nsems; i++) |
1a233956 | 1437 | sem_io[i] = sma->sems[i].semval; |
6062a8dc | 1438 | sem_unlock(sma, -1); |
6d49dab8 | 1439 | rcu_read_unlock(); |
1da177e4 | 1440 | err = 0; |
239521f3 | 1441 | if (copy_to_user(array, sem_io, nsems*sizeof(ushort))) |
1da177e4 LT |
1442 | err = -EFAULT; |
1443 | goto out_free; | |
1444 | } | |
1445 | case SETALL: | |
1446 | { | |
1447 | int i; | |
1448 | struct sem_undo *un; | |
1449 | ||
dba4cdd3 | 1450 | if (!ipc_rcu_getref(&sma->sem_perm)) { |
6e224f94 MS |
1451 | err = -EIDRM; |
1452 | goto out_rcu_wakeup; | |
6062a8dc | 1453 | } |
16df3674 | 1454 | rcu_read_unlock(); |
1da177e4 | 1455 | |
239521f3 | 1456 | if (nsems > SEMMSL_FAST) { |
f8dbe8d2 KC |
1457 | sem_io = kvmalloc_array(nsems, sizeof(ushort), |
1458 | GFP_KERNEL); | |
239521f3 | 1459 | if (sem_io == NULL) { |
dba4cdd3 | 1460 | ipc_rcu_putref(&sma->sem_perm, sem_rcu_free); |
1da177e4 LT |
1461 | return -ENOMEM; |
1462 | } | |
1463 | } | |
1464 | ||
239521f3 | 1465 | if (copy_from_user(sem_io, p, nsems*sizeof(ushort))) { |
dba4cdd3 | 1466 | ipc_rcu_putref(&sma->sem_perm, sem_rcu_free); |
1da177e4 LT |
1467 | err = -EFAULT; |
1468 | goto out_free; | |
1469 | } | |
1470 | ||
1471 | for (i = 0; i < nsems; i++) { | |
1472 | if (sem_io[i] > SEMVMX) { | |
dba4cdd3 | 1473 | ipc_rcu_putref(&sma->sem_perm, sem_rcu_free); |
1da177e4 LT |
1474 | err = -ERANGE; |
1475 | goto out_free; | |
1476 | } | |
1477 | } | |
4091fd94 | 1478 | rcu_read_lock(); |
6ff37972 | 1479 | sem_lock_and_putref(sma); |
0f3d2b01 | 1480 | if (!ipc_valid_object(&sma->sem_perm)) { |
1da177e4 | 1481 | err = -EIDRM; |
6e224f94 | 1482 | goto out_unlock; |
1da177e4 LT |
1483 | } |
1484 | ||
a5f4db87 | 1485 | for (i = 0; i < nsems; i++) { |
1a233956 | 1486 | sma->sems[i].semval = sem_io[i]; |
51d6f263 | 1487 | ipc_update_pid(&sma->sems[i].sempid, task_tgid(current)); |
a5f4db87 | 1488 | } |
4daa28f6 | 1489 | |
cf9d5d78 | 1490 | ipc_assert_locked_object(&sma->sem_perm); |
4daa28f6 | 1491 | list_for_each_entry(un, &sma->list_id, list_id) { |
1da177e4 LT |
1492 | for (i = 0; i < nsems; i++) |
1493 | un->semadj[i] = 0; | |
4daa28f6 | 1494 | } |
e54d02b2 | 1495 | sma->sem_ctime = ktime_get_real_seconds(); |
1da177e4 | 1496 | /* maybe some queued-up processes were waiting for this */ |
9ae949fa | 1497 | do_smart_update(sma, NULL, 0, 0, &wake_q); |
1da177e4 LT |
1498 | err = 0; |
1499 | goto out_unlock; | |
1500 | } | |
e1fd1f49 | 1501 | /* GETVAL, GETPID, GETNCTN, GETZCNT: fall-through */ |
1da177e4 LT |
1502 | } |
1503 | err = -EINVAL; | |
c728b9c8 LT |
1504 | if (semnum < 0 || semnum >= nsems) |
1505 | goto out_rcu_wakeup; | |
1da177e4 | 1506 | |
6062a8dc | 1507 | sem_lock(sma, NULL, -1); |
0f3d2b01 | 1508 | if (!ipc_valid_object(&sma->sem_perm)) { |
6e224f94 MS |
1509 | err = -EIDRM; |
1510 | goto out_unlock; | |
1511 | } | |
1a233956 | 1512 | curr = &sma->sems[semnum]; |
1da177e4 LT |
1513 | |
1514 | switch (cmd) { | |
1515 | case GETVAL: | |
1516 | err = curr->semval; | |
1517 | goto out_unlock; | |
1518 | case GETPID: | |
51d6f263 | 1519 | err = pid_vnr(curr->sempid); |
1da177e4 LT |
1520 | goto out_unlock; |
1521 | case GETNCNT: | |
2f2ed41d | 1522 | err = count_semcnt(sma, semnum, 0); |
1da177e4 LT |
1523 | goto out_unlock; |
1524 | case GETZCNT: | |
2f2ed41d | 1525 | err = count_semcnt(sma, semnum, 1); |
1da177e4 | 1526 | goto out_unlock; |
1da177e4 | 1527 | } |
16df3674 | 1528 | |
1da177e4 | 1529 | out_unlock: |
6062a8dc | 1530 | sem_unlock(sma, -1); |
c728b9c8 | 1531 | out_rcu_wakeup: |
6d49dab8 | 1532 | rcu_read_unlock(); |
9ae949fa | 1533 | wake_up_q(&wake_q); |
1da177e4 | 1534 | out_free: |
239521f3 | 1535 | if (sem_io != fast_sem_io) |
f8dbe8d2 | 1536 | kvfree(sem_io); |
1da177e4 LT |
1537 | return err; |
1538 | } | |
1539 | ||
016d7132 PP |
1540 | static inline unsigned long |
1541 | copy_semid_from_user(struct semid64_ds *out, void __user *buf, int version) | |
1da177e4 | 1542 | { |
239521f3 | 1543 | switch (version) { |
1da177e4 | 1544 | case IPC_64: |
016d7132 | 1545 | if (copy_from_user(out, buf, sizeof(*out))) |
1da177e4 | 1546 | return -EFAULT; |
1da177e4 | 1547 | return 0; |
1da177e4 LT |
1548 | case IPC_OLD: |
1549 | { | |
1550 | struct semid_ds tbuf_old; | |
1551 | ||
239521f3 | 1552 | if (copy_from_user(&tbuf_old, buf, sizeof(tbuf_old))) |
1da177e4 LT |
1553 | return -EFAULT; |
1554 | ||
016d7132 PP |
1555 | out->sem_perm.uid = tbuf_old.sem_perm.uid; |
1556 | out->sem_perm.gid = tbuf_old.sem_perm.gid; | |
1557 | out->sem_perm.mode = tbuf_old.sem_perm.mode; | |
1da177e4 LT |
1558 | |
1559 | return 0; | |
1560 | } | |
1561 | default: | |
1562 | return -EINVAL; | |
1563 | } | |
1564 | } | |
1565 | ||
522bb2a2 | 1566 | /* |
d9a605e4 | 1567 | * This function handles some semctl commands which require the rwsem |
522bb2a2 | 1568 | * to be held in write mode. |
d9a605e4 | 1569 | * NOTE: no locks must be held, the rwsem is taken inside this function. |
522bb2a2 | 1570 | */ |
21a4826a | 1571 | static int semctl_down(struct ipc_namespace *ns, int semid, |
45a4a64a | 1572 | int cmd, struct semid64_ds *semid64) |
1da177e4 LT |
1573 | { |
1574 | struct sem_array *sma; | |
1575 | int err; | |
1da177e4 LT |
1576 | struct kern_ipc_perm *ipcp; |
1577 | ||
d9a605e4 | 1578 | down_write(&sem_ids(ns).rwsem); |
7b4cc5d8 DB |
1579 | rcu_read_lock(); |
1580 | ||
16df3674 | 1581 | ipcp = ipcctl_pre_down_nolock(ns, &sem_ids(ns), semid, cmd, |
45a4a64a | 1582 | &semid64->sem_perm, 0); |
7b4cc5d8 DB |
1583 | if (IS_ERR(ipcp)) { |
1584 | err = PTR_ERR(ipcp); | |
7b4cc5d8 DB |
1585 | goto out_unlock1; |
1586 | } | |
073115d6 | 1587 | |
a5f75e7f | 1588 | sma = container_of(ipcp, struct sem_array, sem_perm); |
1da177e4 | 1589 | |
aefad959 | 1590 | err = security_sem_semctl(&sma->sem_perm, cmd); |
7b4cc5d8 DB |
1591 | if (err) |
1592 | goto out_unlock1; | |
1da177e4 | 1593 | |
7b4cc5d8 | 1594 | switch (cmd) { |
1da177e4 | 1595 | case IPC_RMID: |
6062a8dc | 1596 | sem_lock(sma, NULL, -1); |
7b4cc5d8 | 1597 | /* freeary unlocks the ipc object and rcu */ |
01b8b07a | 1598 | freeary(ns, ipcp); |
522bb2a2 | 1599 | goto out_up; |
1da177e4 | 1600 | case IPC_SET: |
6062a8dc | 1601 | sem_lock(sma, NULL, -1); |
45a4a64a | 1602 | err = ipc_update_perm(&semid64->sem_perm, ipcp); |
1efdb69b | 1603 | if (err) |
7b4cc5d8 | 1604 | goto out_unlock0; |
e54d02b2 | 1605 | sma->sem_ctime = ktime_get_real_seconds(); |
1da177e4 LT |
1606 | break; |
1607 | default: | |
1da177e4 | 1608 | err = -EINVAL; |
7b4cc5d8 | 1609 | goto out_unlock1; |
1da177e4 | 1610 | } |
1da177e4 | 1611 | |
7b4cc5d8 | 1612 | out_unlock0: |
6062a8dc | 1613 | sem_unlock(sma, -1); |
7b4cc5d8 | 1614 | out_unlock1: |
6d49dab8 | 1615 | rcu_read_unlock(); |
522bb2a2 | 1616 | out_up: |
d9a605e4 | 1617 | up_write(&sem_ids(ns).rwsem); |
1da177e4 LT |
1618 | return err; |
1619 | } | |
1620 | ||
d969c6fa | 1621 | long ksys_semctl(int semid, int semnum, int cmd, unsigned long arg) |
1da177e4 | 1622 | { |
1da177e4 | 1623 | int version; |
e3893534 | 1624 | struct ipc_namespace *ns; |
e1fd1f49 | 1625 | void __user *p = (void __user *)arg; |
45a4a64a AV |
1626 | struct semid64_ds semid64; |
1627 | int err; | |
1da177e4 LT |
1628 | |
1629 | if (semid < 0) | |
1630 | return -EINVAL; | |
1631 | ||
1632 | version = ipc_parse_version(&cmd); | |
e3893534 | 1633 | ns = current->nsproxy->ipc_ns; |
1da177e4 | 1634 | |
239521f3 | 1635 | switch (cmd) { |
1da177e4 LT |
1636 | case IPC_INFO: |
1637 | case SEM_INFO: | |
45a4a64a | 1638 | return semctl_info(ns, semid, cmd, p); |
4b9fcb0e | 1639 | case IPC_STAT: |
1da177e4 | 1640 | case SEM_STAT: |
a280d6dc | 1641 | case SEM_STAT_ANY: |
45a4a64a AV |
1642 | err = semctl_stat(ns, semid, cmd, &semid64); |
1643 | if (err < 0) | |
1644 | return err; | |
1645 | if (copy_semid_to_user(p, &semid64, version)) | |
1646 | err = -EFAULT; | |
1647 | return err; | |
1da177e4 LT |
1648 | case GETALL: |
1649 | case GETVAL: | |
1650 | case GETPID: | |
1651 | case GETNCNT: | |
1652 | case GETZCNT: | |
1da177e4 | 1653 | case SETALL: |
e1fd1f49 | 1654 | return semctl_main(ns, semid, semnum, cmd, p); |
45a4a64a AV |
1655 | case SETVAL: { |
1656 | int val; | |
1657 | #if defined(CONFIG_64BIT) && defined(__BIG_ENDIAN) | |
1658 | /* big-endian 64bit */ | |
1659 | val = arg >> 32; | |
1660 | #else | |
1661 | /* 32bit or little-endian 64bit */ | |
1662 | val = arg; | |
1663 | #endif | |
1664 | return semctl_setval(ns, semid, semnum, val); | |
1665 | } | |
1da177e4 | 1666 | case IPC_SET: |
45a4a64a AV |
1667 | if (copy_semid_from_user(&semid64, p, version)) |
1668 | return -EFAULT; | |
1669 | case IPC_RMID: | |
1670 | return semctl_down(ns, semid, cmd, &semid64); | |
1da177e4 LT |
1671 | default: |
1672 | return -EINVAL; | |
1673 | } | |
1674 | } | |
1675 | ||
d969c6fa DB |
1676 | SYSCALL_DEFINE4(semctl, int, semid, int, semnum, int, cmd, unsigned long, arg) |
1677 | { | |
1678 | return ksys_semctl(semid, semnum, cmd, arg); | |
1679 | } | |
1680 | ||
c0ebccb6 AV |
1681 | #ifdef CONFIG_COMPAT |
1682 | ||
1683 | struct compat_semid_ds { | |
1684 | struct compat_ipc_perm sem_perm; | |
1685 | compat_time_t sem_otime; | |
1686 | compat_time_t sem_ctime; | |
1687 | compat_uptr_t sem_base; | |
1688 | compat_uptr_t sem_pending; | |
1689 | compat_uptr_t sem_pending_last; | |
1690 | compat_uptr_t undo; | |
1691 | unsigned short sem_nsems; | |
1692 | }; | |
1693 | ||
1694 | static int copy_compat_semid_from_user(struct semid64_ds *out, void __user *buf, | |
1695 | int version) | |
1696 | { | |
1697 | memset(out, 0, sizeof(*out)); | |
1698 | if (version == IPC_64) { | |
6aa211e8 | 1699 | struct compat_semid64_ds __user *p = buf; |
c0ebccb6 AV |
1700 | return get_compat_ipc64_perm(&out->sem_perm, &p->sem_perm); |
1701 | } else { | |
6aa211e8 | 1702 | struct compat_semid_ds __user *p = buf; |
c0ebccb6 AV |
1703 | return get_compat_ipc_perm(&out->sem_perm, &p->sem_perm); |
1704 | } | |
1705 | } | |
1706 | ||
1707 | static int copy_compat_semid_to_user(void __user *buf, struct semid64_ds *in, | |
1708 | int version) | |
1709 | { | |
1710 | if (version == IPC_64) { | |
1711 | struct compat_semid64_ds v; | |
1712 | memset(&v, 0, sizeof(v)); | |
1713 | to_compat_ipc64_perm(&v.sem_perm, &in->sem_perm); | |
c2ab975c AB |
1714 | v.sem_otime = lower_32_bits(in->sem_otime); |
1715 | v.sem_otime_high = upper_32_bits(in->sem_otime); | |
1716 | v.sem_ctime = lower_32_bits(in->sem_ctime); | |
1717 | v.sem_ctime_high = upper_32_bits(in->sem_ctime); | |
c0ebccb6 AV |
1718 | v.sem_nsems = in->sem_nsems; |
1719 | return copy_to_user(buf, &v, sizeof(v)); | |
1720 | } else { | |
1721 | struct compat_semid_ds v; | |
1722 | memset(&v, 0, sizeof(v)); | |
1723 | to_compat_ipc_perm(&v.sem_perm, &in->sem_perm); | |
1724 | v.sem_otime = in->sem_otime; | |
1725 | v.sem_ctime = in->sem_ctime; | |
1726 | v.sem_nsems = in->sem_nsems; | |
1727 | return copy_to_user(buf, &v, sizeof(v)); | |
1728 | } | |
1729 | } | |
1730 | ||
d969c6fa | 1731 | long compat_ksys_semctl(int semid, int semnum, int cmd, int arg) |
c0ebccb6 AV |
1732 | { |
1733 | void __user *p = compat_ptr(arg); | |
1734 | struct ipc_namespace *ns; | |
1735 | struct semid64_ds semid64; | |
1736 | int version = compat_ipc_parse_version(&cmd); | |
1737 | int err; | |
1738 | ||
1739 | ns = current->nsproxy->ipc_ns; | |
1740 | ||
1741 | if (semid < 0) | |
1742 | return -EINVAL; | |
1743 | ||
1744 | switch (cmd & (~IPC_64)) { | |
1745 | case IPC_INFO: | |
1746 | case SEM_INFO: | |
1747 | return semctl_info(ns, semid, cmd, p); | |
1748 | case IPC_STAT: | |
1749 | case SEM_STAT: | |
a280d6dc | 1750 | case SEM_STAT_ANY: |
c0ebccb6 AV |
1751 | err = semctl_stat(ns, semid, cmd, &semid64); |
1752 | if (err < 0) | |
1753 | return err; | |
1754 | if (copy_compat_semid_to_user(p, &semid64, version)) | |
1755 | err = -EFAULT; | |
1756 | return err; | |
1757 | case GETVAL: | |
1758 | case GETPID: | |
1759 | case GETNCNT: | |
1760 | case GETZCNT: | |
1761 | case GETALL: | |
1da177e4 | 1762 | case SETALL: |
e1fd1f49 AV |
1763 | return semctl_main(ns, semid, semnum, cmd, p); |
1764 | case SETVAL: | |
1765 | return semctl_setval(ns, semid, semnum, arg); | |
1da177e4 | 1766 | case IPC_SET: |
c0ebccb6 AV |
1767 | if (copy_compat_semid_from_user(&semid64, p, version)) |
1768 | return -EFAULT; | |
1769 | /* fallthru */ | |
1770 | case IPC_RMID: | |
1771 | return semctl_down(ns, semid, cmd, &semid64); | |
1da177e4 LT |
1772 | default: |
1773 | return -EINVAL; | |
1774 | } | |
1775 | } | |
d969c6fa DB |
1776 | |
1777 | COMPAT_SYSCALL_DEFINE4(semctl, int, semid, int, semnum, int, cmd, int, arg) | |
1778 | { | |
1779 | return compat_ksys_semctl(semid, semnum, cmd, arg); | |
1780 | } | |
c0ebccb6 | 1781 | #endif |
1da177e4 | 1782 | |
1da177e4 LT |
1783 | /* If the task doesn't already have a undo_list, then allocate one |
1784 | * here. We guarantee there is only one thread using this undo list, | |
1785 | * and current is THE ONE | |
1786 | * | |
1787 | * If this allocation and assignment succeeds, but later | |
1788 | * portions of this code fail, there is no need to free the sem_undo_list. | |
1789 | * Just let it stay associated with the task, and it'll be freed later | |
1790 | * at exit time. | |
1791 | * | |
1792 | * This can block, so callers must hold no locks. | |
1793 | */ | |
1794 | static inline int get_undo_list(struct sem_undo_list **undo_listp) | |
1795 | { | |
1796 | struct sem_undo_list *undo_list; | |
1da177e4 LT |
1797 | |
1798 | undo_list = current->sysvsem.undo_list; | |
1799 | if (!undo_list) { | |
2453a306 | 1800 | undo_list = kzalloc(sizeof(*undo_list), GFP_KERNEL); |
1da177e4 LT |
1801 | if (undo_list == NULL) |
1802 | return -ENOMEM; | |
00a5dfdb | 1803 | spin_lock_init(&undo_list->lock); |
f74370b8 | 1804 | refcount_set(&undo_list->refcnt, 1); |
4daa28f6 MS |
1805 | INIT_LIST_HEAD(&undo_list->list_proc); |
1806 | ||
1da177e4 LT |
1807 | current->sysvsem.undo_list = undo_list; |
1808 | } | |
1809 | *undo_listp = undo_list; | |
1810 | return 0; | |
1811 | } | |
1812 | ||
bf17bb71 | 1813 | static struct sem_undo *__lookup_undo(struct sem_undo_list *ulp, int semid) |
1da177e4 | 1814 | { |
bf17bb71 | 1815 | struct sem_undo *un; |
4daa28f6 | 1816 | |
bf17bb71 NP |
1817 | list_for_each_entry_rcu(un, &ulp->list_proc, list_proc) { |
1818 | if (un->semid == semid) | |
1819 | return un; | |
1da177e4 | 1820 | } |
4daa28f6 | 1821 | return NULL; |
1da177e4 LT |
1822 | } |
1823 | ||
bf17bb71 NP |
1824 | static struct sem_undo *lookup_undo(struct sem_undo_list *ulp, int semid) |
1825 | { | |
1826 | struct sem_undo *un; | |
1827 | ||
239521f3 | 1828 | assert_spin_locked(&ulp->lock); |
bf17bb71 NP |
1829 | |
1830 | un = __lookup_undo(ulp, semid); | |
1831 | if (un) { | |
1832 | list_del_rcu(&un->list_proc); | |
1833 | list_add_rcu(&un->list_proc, &ulp->list_proc); | |
1834 | } | |
1835 | return un; | |
1836 | } | |
1837 | ||
4daa28f6 | 1838 | /** |
8001c858 | 1839 | * find_alloc_undo - lookup (and if not present create) undo array |
4daa28f6 MS |
1840 | * @ns: namespace |
1841 | * @semid: semaphore array id | |
1842 | * | |
1843 | * The function looks up (and if not present creates) the undo structure. | |
1844 | * The size of the undo structure depends on the size of the semaphore | |
1845 | * array, thus the alloc path is not that straightforward. | |
380af1b3 MS |
1846 | * Lifetime-rules: sem_undo is rcu-protected, on success, the function |
1847 | * performs a rcu_read_lock(). | |
4daa28f6 MS |
1848 | */ |
1849 | static struct sem_undo *find_alloc_undo(struct ipc_namespace *ns, int semid) | |
1da177e4 LT |
1850 | { |
1851 | struct sem_array *sma; | |
1852 | struct sem_undo_list *ulp; | |
1853 | struct sem_undo *un, *new; | |
6062a8dc | 1854 | int nsems, error; |
1da177e4 LT |
1855 | |
1856 | error = get_undo_list(&ulp); | |
1857 | if (error) | |
1858 | return ERR_PTR(error); | |
1859 | ||
380af1b3 | 1860 | rcu_read_lock(); |
c530c6ac | 1861 | spin_lock(&ulp->lock); |
1da177e4 | 1862 | un = lookup_undo(ulp, semid); |
c530c6ac | 1863 | spin_unlock(&ulp->lock); |
239521f3 | 1864 | if (likely(un != NULL)) |
1da177e4 LT |
1865 | goto out; |
1866 | ||
1867 | /* no undo structure around - allocate one. */ | |
4daa28f6 | 1868 | /* step 1: figure out the size of the semaphore array */ |
16df3674 DB |
1869 | sma = sem_obtain_object_check(ns, semid); |
1870 | if (IS_ERR(sma)) { | |
1871 | rcu_read_unlock(); | |
4de85cd6 | 1872 | return ERR_CAST(sma); |
16df3674 | 1873 | } |
023a5355 | 1874 | |
1da177e4 | 1875 | nsems = sma->sem_nsems; |
dba4cdd3 | 1876 | if (!ipc_rcu_getref(&sma->sem_perm)) { |
6062a8dc RR |
1877 | rcu_read_unlock(); |
1878 | un = ERR_PTR(-EIDRM); | |
1879 | goto out; | |
1880 | } | |
16df3674 | 1881 | rcu_read_unlock(); |
1da177e4 | 1882 | |
4daa28f6 | 1883 | /* step 2: allocate new undo structure */ |
4668edc3 | 1884 | new = kzalloc(sizeof(struct sem_undo) + sizeof(short)*nsems, GFP_KERNEL); |
1da177e4 | 1885 | if (!new) { |
dba4cdd3 | 1886 | ipc_rcu_putref(&sma->sem_perm, sem_rcu_free); |
1da177e4 LT |
1887 | return ERR_PTR(-ENOMEM); |
1888 | } | |
1da177e4 | 1889 | |
380af1b3 | 1890 | /* step 3: Acquire the lock on semaphore array */ |
4091fd94 | 1891 | rcu_read_lock(); |
6ff37972 | 1892 | sem_lock_and_putref(sma); |
0f3d2b01 | 1893 | if (!ipc_valid_object(&sma->sem_perm)) { |
6062a8dc | 1894 | sem_unlock(sma, -1); |
6d49dab8 | 1895 | rcu_read_unlock(); |
1da177e4 LT |
1896 | kfree(new); |
1897 | un = ERR_PTR(-EIDRM); | |
1898 | goto out; | |
1899 | } | |
380af1b3 MS |
1900 | spin_lock(&ulp->lock); |
1901 | ||
1902 | /* | |
1903 | * step 4: check for races: did someone else allocate the undo struct? | |
1904 | */ | |
1905 | un = lookup_undo(ulp, semid); | |
1906 | if (un) { | |
1907 | kfree(new); | |
1908 | goto success; | |
1909 | } | |
4daa28f6 MS |
1910 | /* step 5: initialize & link new undo structure */ |
1911 | new->semadj = (short *) &new[1]; | |
380af1b3 | 1912 | new->ulp = ulp; |
4daa28f6 MS |
1913 | new->semid = semid; |
1914 | assert_spin_locked(&ulp->lock); | |
380af1b3 | 1915 | list_add_rcu(&new->list_proc, &ulp->list_proc); |
cf9d5d78 | 1916 | ipc_assert_locked_object(&sma->sem_perm); |
4daa28f6 | 1917 | list_add(&new->list_id, &sma->list_id); |
380af1b3 | 1918 | un = new; |
4daa28f6 | 1919 | |
380af1b3 | 1920 | success: |
c530c6ac | 1921 | spin_unlock(&ulp->lock); |
6062a8dc | 1922 | sem_unlock(sma, -1); |
1da177e4 LT |
1923 | out: |
1924 | return un; | |
1925 | } | |
1926 | ||
44ee4546 | 1927 | static long do_semtimedop(int semid, struct sembuf __user *tsops, |
3ef56dc2 | 1928 | unsigned nsops, const struct timespec64 *timeout) |
1da177e4 LT |
1929 | { |
1930 | int error = -EINVAL; | |
1931 | struct sem_array *sma; | |
1932 | struct sembuf fast_sops[SEMOPM_FAST]; | |
239521f3 | 1933 | struct sembuf *sops = fast_sops, *sop; |
1da177e4 | 1934 | struct sem_undo *un; |
4ce33ec2 DB |
1935 | int max, locknum; |
1936 | bool undos = false, alter = false, dupsop = false; | |
1da177e4 | 1937 | struct sem_queue queue; |
4ce33ec2 | 1938 | unsigned long dup = 0, jiffies_left = 0; |
e3893534 KK |
1939 | struct ipc_namespace *ns; |
1940 | ||
1941 | ns = current->nsproxy->ipc_ns; | |
1da177e4 LT |
1942 | |
1943 | if (nsops < 1 || semid < 0) | |
1944 | return -EINVAL; | |
e3893534 | 1945 | if (nsops > ns->sc_semopm) |
1da177e4 | 1946 | return -E2BIG; |
239521f3 | 1947 | if (nsops > SEMOPM_FAST) { |
344476e1 | 1948 | sops = kvmalloc_array(nsops, sizeof(*sops), GFP_KERNEL); |
239521f3 | 1949 | if (sops == NULL) |
1da177e4 LT |
1950 | return -ENOMEM; |
1951 | } | |
4ce33ec2 | 1952 | |
239521f3 MS |
1953 | if (copy_from_user(sops, tsops, nsops * sizeof(*tsops))) { |
1954 | error = -EFAULT; | |
1da177e4 LT |
1955 | goto out_free; |
1956 | } | |
4ce33ec2 | 1957 | |
1da177e4 | 1958 | if (timeout) { |
44ee4546 AV |
1959 | if (timeout->tv_sec < 0 || timeout->tv_nsec < 0 || |
1960 | timeout->tv_nsec >= 1000000000L) { | |
1da177e4 LT |
1961 | error = -EINVAL; |
1962 | goto out_free; | |
1963 | } | |
3ef56dc2 | 1964 | jiffies_left = timespec64_to_jiffies(timeout); |
1da177e4 | 1965 | } |
4ce33ec2 | 1966 | |
1da177e4 LT |
1967 | max = 0; |
1968 | for (sop = sops; sop < sops + nsops; sop++) { | |
4ce33ec2 DB |
1969 | unsigned long mask = 1ULL << ((sop->sem_num) % BITS_PER_LONG); |
1970 | ||
1da177e4 LT |
1971 | if (sop->sem_num >= max) |
1972 | max = sop->sem_num; | |
1973 | if (sop->sem_flg & SEM_UNDO) | |
4ce33ec2 DB |
1974 | undos = true; |
1975 | if (dup & mask) { | |
1976 | /* | |
1977 | * There was a previous alter access that appears | |
1978 | * to have accessed the same semaphore, thus use | |
1979 | * the dupsop logic. "appears", because the detection | |
1980 | * can only check % BITS_PER_LONG. | |
1981 | */ | |
1982 | dupsop = true; | |
1983 | } | |
1984 | if (sop->sem_op != 0) { | |
1985 | alter = true; | |
1986 | dup |= mask; | |
1987 | } | |
1da177e4 | 1988 | } |
1da177e4 | 1989 | |
1da177e4 | 1990 | if (undos) { |
6062a8dc | 1991 | /* On success, find_alloc_undo takes the rcu_read_lock */ |
4daa28f6 | 1992 | un = find_alloc_undo(ns, semid); |
1da177e4 LT |
1993 | if (IS_ERR(un)) { |
1994 | error = PTR_ERR(un); | |
1995 | goto out_free; | |
1996 | } | |
6062a8dc | 1997 | } else { |
1da177e4 | 1998 | un = NULL; |
6062a8dc RR |
1999 | rcu_read_lock(); |
2000 | } | |
1da177e4 | 2001 | |
16df3674 | 2002 | sma = sem_obtain_object_check(ns, semid); |
023a5355 | 2003 | if (IS_ERR(sma)) { |
6062a8dc | 2004 | rcu_read_unlock(); |
023a5355 | 2005 | error = PTR_ERR(sma); |
1da177e4 | 2006 | goto out_free; |
023a5355 ND |
2007 | } |
2008 | ||
16df3674 | 2009 | error = -EFBIG; |
248e7357 DB |
2010 | if (max >= sma->sem_nsems) { |
2011 | rcu_read_unlock(); | |
2012 | goto out_free; | |
2013 | } | |
16df3674 DB |
2014 | |
2015 | error = -EACCES; | |
248e7357 DB |
2016 | if (ipcperms(ns, &sma->sem_perm, alter ? S_IWUGO : S_IRUGO)) { |
2017 | rcu_read_unlock(); | |
2018 | goto out_free; | |
2019 | } | |
16df3674 | 2020 | |
aefad959 | 2021 | error = security_sem_semop(&sma->sem_perm, sops, nsops, alter); |
248e7357 DB |
2022 | if (error) { |
2023 | rcu_read_unlock(); | |
2024 | goto out_free; | |
2025 | } | |
16df3674 | 2026 | |
6e224f94 MS |
2027 | error = -EIDRM; |
2028 | locknum = sem_lock(sma, sops, nsops); | |
0f3d2b01 RA |
2029 | /* |
2030 | * We eventually might perform the following check in a lockless | |
2031 | * fashion, considering ipc_valid_object() locking constraints. | |
2032 | * If nsops == 1 and there is no contention for sem_perm.lock, then | |
2033 | * only a per-semaphore lock is held and it's OK to proceed with the | |
2034 | * check below. More details on the fine grained locking scheme | |
2035 | * entangled here and why it's RMID race safe on comments at sem_lock() | |
2036 | */ | |
2037 | if (!ipc_valid_object(&sma->sem_perm)) | |
6e224f94 | 2038 | goto out_unlock_free; |
1da177e4 | 2039 | /* |
4daa28f6 | 2040 | * semid identifiers are not unique - find_alloc_undo may have |
1da177e4 | 2041 | * allocated an undo structure, it was invalidated by an RMID |
4daa28f6 | 2042 | * and now a new array with received the same id. Check and fail. |
25985edc | 2043 | * This case can be detected checking un->semid. The existence of |
380af1b3 | 2044 | * "un" itself is guaranteed by rcu. |
1da177e4 | 2045 | */ |
6062a8dc RR |
2046 | if (un && un->semid == -1) |
2047 | goto out_unlock_free; | |
4daa28f6 | 2048 | |
d198cd6d MS |
2049 | queue.sops = sops; |
2050 | queue.nsops = nsops; | |
2051 | queue.undo = un; | |
51d6f263 | 2052 | queue.pid = task_tgid(current); |
d198cd6d | 2053 | queue.alter = alter; |
4ce33ec2 | 2054 | queue.dupsop = dupsop; |
d198cd6d MS |
2055 | |
2056 | error = perform_atomic_semop(sma, &queue); | |
9ae949fa DB |
2057 | if (error == 0) { /* non-blocking succesfull path */ |
2058 | DEFINE_WAKE_Q(wake_q); | |
2059 | ||
2060 | /* | |
2061 | * If the operation was successful, then do | |
0e8c6656 MS |
2062 | * the required updates. |
2063 | */ | |
2064 | if (alter) | |
9ae949fa | 2065 | do_smart_update(sma, sops, nsops, 1, &wake_q); |
0e8c6656 MS |
2066 | else |
2067 | set_semotime(sma, sops); | |
9ae949fa DB |
2068 | |
2069 | sem_unlock(sma, locknum); | |
2070 | rcu_read_unlock(); | |
2071 | wake_up_q(&wake_q); | |
2072 | ||
2073 | goto out_free; | |
1da177e4 | 2074 | } |
9ae949fa | 2075 | if (error < 0) /* non-blocking error path */ |
0e8c6656 | 2076 | goto out_unlock_free; |
1da177e4 | 2077 | |
9ae949fa DB |
2078 | /* |
2079 | * We need to sleep on this operation, so we put the current | |
1da177e4 LT |
2080 | * task into the pending queue and go to sleep. |
2081 | */ | |
b97e820f MS |
2082 | if (nsops == 1) { |
2083 | struct sem *curr; | |
1a233956 | 2084 | curr = &sma->sems[sops->sem_num]; |
b97e820f | 2085 | |
f269f40a MS |
2086 | if (alter) { |
2087 | if (sma->complex_count) { | |
2088 | list_add_tail(&queue.list, | |
2089 | &sma->pending_alter); | |
2090 | } else { | |
2091 | ||
2092 | list_add_tail(&queue.list, | |
2093 | &curr->pending_alter); | |
2094 | } | |
2095 | } else { | |
1a82e9e1 | 2096 | list_add_tail(&queue.list, &curr->pending_const); |
f269f40a | 2097 | } |
b97e820f | 2098 | } else { |
f269f40a MS |
2099 | if (!sma->complex_count) |
2100 | merge_queues(sma); | |
2101 | ||
9f1bc2c9 | 2102 | if (alter) |
1a82e9e1 | 2103 | list_add_tail(&queue.list, &sma->pending_alter); |
9f1bc2c9 | 2104 | else |
1a82e9e1 MS |
2105 | list_add_tail(&queue.list, &sma->pending_const); |
2106 | ||
b97e820f MS |
2107 | sma->complex_count++; |
2108 | } | |
2109 | ||
b5fa01a2 DB |
2110 | do { |
2111 | queue.status = -EINTR; | |
2112 | queue.sleeper = current; | |
0b0577f6 | 2113 | |
b5fa01a2 DB |
2114 | __set_current_state(TASK_INTERRUPTIBLE); |
2115 | sem_unlock(sma, locknum); | |
2116 | rcu_read_unlock(); | |
1da177e4 | 2117 | |
b5fa01a2 DB |
2118 | if (timeout) |
2119 | jiffies_left = schedule_timeout(jiffies_left); | |
2120 | else | |
2121 | schedule(); | |
1da177e4 | 2122 | |
9ae949fa | 2123 | /* |
b5fa01a2 DB |
2124 | * fastpath: the semop has completed, either successfully or |
2125 | * not, from the syscall pov, is quite irrelevant to us at this | |
2126 | * point; we're done. | |
2127 | * | |
2128 | * We _do_ care, nonetheless, about being awoken by a signal or | |
2129 | * spuriously. The queue.status is checked again in the | |
2130 | * slowpath (aka after taking sem_lock), such that we can detect | |
2131 | * scenarios where we were awakened externally, during the | |
2132 | * window between wake_q_add() and wake_up_q(). | |
c61284e9 | 2133 | */ |
b5fa01a2 DB |
2134 | error = READ_ONCE(queue.status); |
2135 | if (error != -EINTR) { | |
2136 | /* | |
2137 | * User space could assume that semop() is a memory | |
2138 | * barrier: Without the mb(), the cpu could | |
2139 | * speculatively read in userspace stale data that was | |
2140 | * overwritten by the previous owner of the semaphore. | |
2141 | */ | |
2142 | smp_mb(); | |
2143 | goto out_free; | |
2144 | } | |
d694ad62 | 2145 | |
b5fa01a2 | 2146 | rcu_read_lock(); |
c626bc46 | 2147 | locknum = sem_lock(sma, sops, nsops); |
1da177e4 | 2148 | |
370b262c DB |
2149 | if (!ipc_valid_object(&sma->sem_perm)) |
2150 | goto out_unlock_free; | |
2151 | ||
2152 | error = READ_ONCE(queue.status); | |
1da177e4 | 2153 | |
b5fa01a2 DB |
2154 | /* |
2155 | * If queue.status != -EINTR we are woken up by another process. | |
2156 | * Leave without unlink_queue(), but with sem_unlock(). | |
2157 | */ | |
2158 | if (error != -EINTR) | |
2159 | goto out_unlock_free; | |
0b0577f6 | 2160 | |
b5fa01a2 DB |
2161 | /* |
2162 | * If an interrupt occurred we have to clean up the queue. | |
2163 | */ | |
2164 | if (timeout && jiffies_left == 0) | |
2165 | error = -EAGAIN; | |
2166 | } while (error == -EINTR && !signal_pending(current)); /* spurious */ | |
0b0577f6 | 2167 | |
b97e820f | 2168 | unlink_queue(sma, &queue); |
1da177e4 LT |
2169 | |
2170 | out_unlock_free: | |
6062a8dc | 2171 | sem_unlock(sma, locknum); |
6d49dab8 | 2172 | rcu_read_unlock(); |
1da177e4 | 2173 | out_free: |
239521f3 | 2174 | if (sops != fast_sops) |
e4243b80 | 2175 | kvfree(sops); |
1da177e4 LT |
2176 | return error; |
2177 | } | |
2178 | ||
41f4f0e2 | 2179 | long ksys_semtimedop(int semid, struct sembuf __user *tsops, |
21fc538d | 2180 | unsigned int nsops, const struct __kernel_timespec __user *timeout) |
44ee4546 AV |
2181 | { |
2182 | if (timeout) { | |
3ef56dc2 DD |
2183 | struct timespec64 ts; |
2184 | if (get_timespec64(&ts, timeout)) | |
44ee4546 AV |
2185 | return -EFAULT; |
2186 | return do_semtimedop(semid, tsops, nsops, &ts); | |
2187 | } | |
2188 | return do_semtimedop(semid, tsops, nsops, NULL); | |
2189 | } | |
2190 | ||
41f4f0e2 | 2191 | SYSCALL_DEFINE4(semtimedop, int, semid, struct sembuf __user *, tsops, |
21fc538d | 2192 | unsigned int, nsops, const struct __kernel_timespec __user *, timeout) |
41f4f0e2 DB |
2193 | { |
2194 | return ksys_semtimedop(semid, tsops, nsops, timeout); | |
2195 | } | |
2196 | ||
b0d17578 | 2197 | #ifdef CONFIG_COMPAT_32BIT_TIME |
41f4f0e2 DB |
2198 | long compat_ksys_semtimedop(int semid, struct sembuf __user *tsems, |
2199 | unsigned int nsops, | |
2200 | const struct compat_timespec __user *timeout) | |
44ee4546 AV |
2201 | { |
2202 | if (timeout) { | |
3ef56dc2 DD |
2203 | struct timespec64 ts; |
2204 | if (compat_get_timespec64(&ts, timeout)) | |
44ee4546 AV |
2205 | return -EFAULT; |
2206 | return do_semtimedop(semid, tsems, nsops, &ts); | |
2207 | } | |
2208 | return do_semtimedop(semid, tsems, nsops, NULL); | |
2209 | } | |
41f4f0e2 DB |
2210 | |
2211 | COMPAT_SYSCALL_DEFINE4(semtimedop, int, semid, struct sembuf __user *, tsems, | |
2212 | unsigned int, nsops, | |
2213 | const struct compat_timespec __user *, timeout) | |
2214 | { | |
2215 | return compat_ksys_semtimedop(semid, tsems, nsops, timeout); | |
2216 | } | |
44ee4546 AV |
2217 | #endif |
2218 | ||
d5460c99 HC |
2219 | SYSCALL_DEFINE3(semop, int, semid, struct sembuf __user *, tsops, |
2220 | unsigned, nsops) | |
1da177e4 | 2221 | { |
44ee4546 | 2222 | return do_semtimedop(semid, tsops, nsops, NULL); |
1da177e4 LT |
2223 | } |
2224 | ||
2225 | /* If CLONE_SYSVSEM is set, establish sharing of SEM_UNDO state between | |
2226 | * parent and child tasks. | |
1da177e4 LT |
2227 | */ |
2228 | ||
2229 | int copy_semundo(unsigned long clone_flags, struct task_struct *tsk) | |
2230 | { | |
2231 | struct sem_undo_list *undo_list; | |
2232 | int error; | |
2233 | ||
2234 | if (clone_flags & CLONE_SYSVSEM) { | |
2235 | error = get_undo_list(&undo_list); | |
2236 | if (error) | |
2237 | return error; | |
f74370b8 | 2238 | refcount_inc(&undo_list->refcnt); |
1da177e4 | 2239 | tsk->sysvsem.undo_list = undo_list; |
46c0a8ca | 2240 | } else |
1da177e4 LT |
2241 | tsk->sysvsem.undo_list = NULL; |
2242 | ||
2243 | return 0; | |
2244 | } | |
2245 | ||
2246 | /* | |
2247 | * add semadj values to semaphores, free undo structures. | |
2248 | * undo structures are not freed when semaphore arrays are destroyed | |
2249 | * so some of them may be out of date. | |
2250 | * IMPLEMENTATION NOTE: There is some confusion over whether the | |
2251 | * set of adjustments that needs to be done should be done in an atomic | |
2252 | * manner or not. That is, if we are attempting to decrement the semval | |
2253 | * should we queue up and wait until we can do so legally? | |
2254 | * The original implementation attempted to do this (queue and wait). | |
2255 | * The current implementation does not do so. The POSIX standard | |
2256 | * and SVID should be consulted to determine what behavior is mandated. | |
2257 | */ | |
2258 | void exit_sem(struct task_struct *tsk) | |
2259 | { | |
4daa28f6 | 2260 | struct sem_undo_list *ulp; |
1da177e4 | 2261 | |
4daa28f6 MS |
2262 | ulp = tsk->sysvsem.undo_list; |
2263 | if (!ulp) | |
1da177e4 | 2264 | return; |
9edff4ab | 2265 | tsk->sysvsem.undo_list = NULL; |
1da177e4 | 2266 | |
f74370b8 | 2267 | if (!refcount_dec_and_test(&ulp->refcnt)) |
1da177e4 LT |
2268 | return; |
2269 | ||
380af1b3 | 2270 | for (;;) { |
1da177e4 | 2271 | struct sem_array *sma; |
380af1b3 | 2272 | struct sem_undo *un; |
6062a8dc | 2273 | int semid, i; |
9ae949fa | 2274 | DEFINE_WAKE_Q(wake_q); |
4daa28f6 | 2275 | |
2a1613a5 NB |
2276 | cond_resched(); |
2277 | ||
380af1b3 | 2278 | rcu_read_lock(); |
05725f7e JP |
2279 | un = list_entry_rcu(ulp->list_proc.next, |
2280 | struct sem_undo, list_proc); | |
602b8593 HK |
2281 | if (&un->list_proc == &ulp->list_proc) { |
2282 | /* | |
2283 | * We must wait for freeary() before freeing this ulp, | |
2284 | * in case we raced with last sem_undo. There is a small | |
2285 | * possibility where we exit while freeary() didn't | |
2286 | * finish unlocking sem_undo_list. | |
2287 | */ | |
e0892e08 PM |
2288 | spin_lock(&ulp->lock); |
2289 | spin_unlock(&ulp->lock); | |
602b8593 HK |
2290 | rcu_read_unlock(); |
2291 | break; | |
2292 | } | |
2293 | spin_lock(&ulp->lock); | |
2294 | semid = un->semid; | |
2295 | spin_unlock(&ulp->lock); | |
4daa28f6 | 2296 | |
602b8593 | 2297 | /* exit_sem raced with IPC_RMID, nothing to do */ |
6062a8dc RR |
2298 | if (semid == -1) { |
2299 | rcu_read_unlock(); | |
602b8593 | 2300 | continue; |
6062a8dc | 2301 | } |
1da177e4 | 2302 | |
602b8593 | 2303 | sma = sem_obtain_object_check(tsk->nsproxy->ipc_ns, semid); |
380af1b3 | 2304 | /* exit_sem raced with IPC_RMID, nothing to do */ |
6062a8dc RR |
2305 | if (IS_ERR(sma)) { |
2306 | rcu_read_unlock(); | |
380af1b3 | 2307 | continue; |
6062a8dc | 2308 | } |
1da177e4 | 2309 | |
6062a8dc | 2310 | sem_lock(sma, NULL, -1); |
6e224f94 | 2311 | /* exit_sem raced with IPC_RMID, nothing to do */ |
0f3d2b01 | 2312 | if (!ipc_valid_object(&sma->sem_perm)) { |
6e224f94 MS |
2313 | sem_unlock(sma, -1); |
2314 | rcu_read_unlock(); | |
2315 | continue; | |
2316 | } | |
bf17bb71 | 2317 | un = __lookup_undo(ulp, semid); |
380af1b3 MS |
2318 | if (un == NULL) { |
2319 | /* exit_sem raced with IPC_RMID+semget() that created | |
2320 | * exactly the same semid. Nothing to do. | |
2321 | */ | |
6062a8dc | 2322 | sem_unlock(sma, -1); |
6d49dab8 | 2323 | rcu_read_unlock(); |
380af1b3 MS |
2324 | continue; |
2325 | } | |
2326 | ||
2327 | /* remove un from the linked lists */ | |
cf9d5d78 | 2328 | ipc_assert_locked_object(&sma->sem_perm); |
4daa28f6 MS |
2329 | list_del(&un->list_id); |
2330 | ||
a9795584 HK |
2331 | /* we are the last process using this ulp, acquiring ulp->lock |
2332 | * isn't required. Besides that, we are also protected against | |
2333 | * IPC_RMID as we hold sma->sem_perm lock now | |
2334 | */ | |
380af1b3 | 2335 | list_del_rcu(&un->list_proc); |
380af1b3 | 2336 | |
4daa28f6 MS |
2337 | /* perform adjustments registered in un */ |
2338 | for (i = 0; i < sma->sem_nsems; i++) { | |
1a233956 | 2339 | struct sem *semaphore = &sma->sems[i]; |
4daa28f6 MS |
2340 | if (un->semadj[i]) { |
2341 | semaphore->semval += un->semadj[i]; | |
1da177e4 LT |
2342 | /* |
2343 | * Range checks of the new semaphore value, | |
2344 | * not defined by sus: | |
2345 | * - Some unices ignore the undo entirely | |
2346 | * (e.g. HP UX 11i 11.22, Tru64 V5.1) | |
2347 | * - some cap the value (e.g. FreeBSD caps | |
2348 | * at 0, but doesn't enforce SEMVMX) | |
2349 | * | |
2350 | * Linux caps the semaphore value, both at 0 | |
2351 | * and at SEMVMX. | |
2352 | * | |
239521f3 | 2353 | * Manfred <manfred@colorfullife.com> |
1da177e4 | 2354 | */ |
5f921ae9 IM |
2355 | if (semaphore->semval < 0) |
2356 | semaphore->semval = 0; | |
2357 | if (semaphore->semval > SEMVMX) | |
2358 | semaphore->semval = SEMVMX; | |
51d6f263 | 2359 | ipc_update_pid(&semaphore->sempid, task_tgid(current)); |
1da177e4 LT |
2360 | } |
2361 | } | |
1da177e4 | 2362 | /* maybe some queued-up processes were waiting for this */ |
9ae949fa | 2363 | do_smart_update(sma, NULL, 0, 1, &wake_q); |
6062a8dc | 2364 | sem_unlock(sma, -1); |
6d49dab8 | 2365 | rcu_read_unlock(); |
9ae949fa | 2366 | wake_up_q(&wake_q); |
380af1b3 | 2367 | |
693a8b6e | 2368 | kfree_rcu(un, rcu); |
1da177e4 | 2369 | } |
4daa28f6 | 2370 | kfree(ulp); |
1da177e4 LT |
2371 | } |
2372 | ||
2373 | #ifdef CONFIG_PROC_FS | |
19b4946c | 2374 | static int sysvipc_sem_proc_show(struct seq_file *s, void *it) |
1da177e4 | 2375 | { |
1efdb69b | 2376 | struct user_namespace *user_ns = seq_user_ns(s); |
ade9f91b KC |
2377 | struct kern_ipc_perm *ipcp = it; |
2378 | struct sem_array *sma = container_of(ipcp, struct sem_array, sem_perm); | |
e54d02b2 | 2379 | time64_t sem_otime; |
d12e1e50 | 2380 | |
d8c63376 MS |
2381 | /* |
2382 | * The proc interface isn't aware of sem_lock(), it calls | |
2383 | * ipc_lock_object() directly (in sysvipc_find_ipc). | |
5864a2fd MS |
2384 | * In order to stay compatible with sem_lock(), we must |
2385 | * enter / leave complex_mode. | |
d8c63376 | 2386 | */ |
5864a2fd | 2387 | complexmode_enter(sma); |
d8c63376 | 2388 | |
d12e1e50 | 2389 | sem_otime = get_semotime(sma); |
19b4946c | 2390 | |
7f032d6e | 2391 | seq_printf(s, |
e54d02b2 | 2392 | "%10d %10d %4o %10u %5u %5u %5u %5u %10llu %10llu\n", |
7f032d6e JP |
2393 | sma->sem_perm.key, |
2394 | sma->sem_perm.id, | |
2395 | sma->sem_perm.mode, | |
2396 | sma->sem_nsems, | |
2397 | from_kuid_munged(user_ns, sma->sem_perm.uid), | |
2398 | from_kgid_munged(user_ns, sma->sem_perm.gid), | |
2399 | from_kuid_munged(user_ns, sma->sem_perm.cuid), | |
2400 | from_kgid_munged(user_ns, sma->sem_perm.cgid), | |
2401 | sem_otime, | |
2402 | sma->sem_ctime); | |
2403 | ||
5864a2fd MS |
2404 | complexmode_tryleave(sma); |
2405 | ||
7f032d6e | 2406 | return 0; |
1da177e4 LT |
2407 | } |
2408 | #endif |