3 * Copyright (C) 1992 Krishna Balasubramanian
4 * Copyright (C) 1995 Eric Schenk, Bruno Haible
6 * /proc/sysvipc/sem support (c) 1999 Dragos Acostachioaie <dragos@iname.com>
8 * SMP-threaded, sysctl's added
9 * (c) 1999 Manfred Spraul <manfred@colorfullife.com>
10 * Enforced range limit on SEM_UNDO
11 * (c) 2001 Red Hat Inc
13 * (c) 2003 Manfred Spraul <manfred@colorfullife.com>
14 * Further wakeup optimizations, documentation
15 * (c) 2010 Manfred Spraul <manfred@colorfullife.com>
17 * support for audit of ipc object properties and permission changes
18 * Dustin Kirkland <dustin.kirkland@us.ibm.com>
22 * Pavel Emelianov <xemul@openvz.org>
24 * Implementation notes: (May 2010)
25 * This file implements System V semaphores.
27 * User space visible behavior:
28 * - FIFO ordering for semop() operations (just FIFO, not starvation
30 * - multiple semaphore operations that alter the same semaphore in
31 * one semop() are handled.
32 * - sem_ctime (time of last semctl()) is updated in the IPC_SET, SETVAL and
34 * - two Linux specific semctl() commands: SEM_STAT, SEM_INFO.
35 * - undo adjustments at process exit are limited to 0..SEMVMX.
36 * - namespace are supported.
37 * - SEMMSL, SEMMNS, SEMOPM and SEMMNI can be configured at runtine by writing
38 * to /proc/sys/kernel/sem.
39 * - statistics about the usage are reported in /proc/sysvipc/sem.
43 * - all global variables are read-mostly.
44 * - semop() calls and semctl(RMID) are synchronized by RCU.
45 * - most operations do write operations (actually: spin_lock calls) to
46 * the per-semaphore array structure.
47 * Thus: Perfect SMP scaling between independent semaphore arrays.
48 * If multiple semaphores in one array are used, then cache line
49 * trashing on the semaphore array spinlock will limit the scaling.
50 * - semncnt and semzcnt are calculated on demand in count_semncnt() and
52 * - the task that performs a successful semop() scans the list of all
53 * sleeping tasks and completes any pending operations that can be fulfilled.
54 * Semaphores are actively given to waiting tasks (necessary for FIFO).
55 * (see update_queue())
56 * - To improve the scalability, the actual wake-up calls are performed after
57 * dropping all locks. (see wake_up_sem_queue_prepare(),
58 * wake_up_sem_queue_do())
59 * - All work is done by the waker, the woken up task does not have to do
60 * anything - not even acquiring a lock or dropping a refcount.
61 * - A woken up task may not even touch the semaphore array anymore, it may
62 * have been destroyed already by a semctl(RMID).
63 * - The synchronizations between wake-ups due to a timeout/signal and a
64 * wake-up due to a completed semaphore operation is achieved by using an
65 * intermediate state (IN_WAKEUP).
66 * - UNDO values are stored in an array (one per process and per
67 * semaphore array, lazily allocated). For backwards compatibility, multiple
68 * modes for the UNDO variables are supported (per process, per thread)
69 * (see copy_semundo, CLONE_SYSVSEM)
70 * - There are two lists of the pending operations: a per-array list
71 * and per-semaphore list (stored in the array). This allows to achieve FIFO
72 * ordering without always scanning all pending operations.
73 * The worst-case behavior is nevertheless O(N^2) for N wakeups.
76 #include <linux/slab.h>
77 #include <linux/spinlock.h>
78 #include <linux/init.h>
79 #include <linux/proc_fs.h>
80 #include <linux/time.h>
81 #include <linux/security.h>
82 #include <linux/syscalls.h>
83 #include <linux/audit.h>
84 #include <linux/capability.h>
85 #include <linux/seq_file.h>
86 #include <linux/rwsem.h>
87 #include <linux/nsproxy.h>
88 #include <linux/ipc_namespace.h>
90 #include <asm/uaccess.h>
93 /* One semaphore structure for each semaphore in the system. */
95 int semval; /* current value */
96 int sempid; /* pid of last operation */
97 struct list_head sem_pending; /* pending single-sop operations */
100 /* One queue for each sleeping process in the system. */
102 struct list_head list; /* queue of pending operations */
103 struct task_struct *sleeper; /* this process */
104 struct sem_undo *undo; /* undo structure */
105 int pid; /* process id of requesting process */
106 int status; /* completion status of operation */
107 struct sembuf *sops; /* array of pending operations */
108 int nsops; /* number of operations */
109 int alter; /* does *sops alter the array? */
112 /* Each task has a list of undo requests. They are executed automatically
113 * when the process exits.
116 struct list_head list_proc; /* per-process list: *
117 * all undos from one process
119 struct rcu_head rcu; /* rcu struct for sem_undo */
120 struct sem_undo_list *ulp; /* back ptr to sem_undo_list */
121 struct list_head list_id; /* per semaphore array list:
122 * all undos for one array */
123 int semid; /* semaphore set identifier */
124 short *semadj; /* array of adjustments */
125 /* one per semaphore */
128 /* sem_undo_list controls shared access to the list of sem_undo structures
129 * that may be shared among all a CLONE_SYSVSEM task group.
131 struct sem_undo_list {
134 struct list_head list_proc;
138 #define sem_ids(ns) ((ns)->ids[IPC_SEM_IDS])
140 #define sem_unlock(sma) ipc_unlock(&(sma)->sem_perm)
141 #define sem_checkid(sma, semid) ipc_checkid(&sma->sem_perm, semid)
143 static int newary(struct ipc_namespace *, struct ipc_params *);
144 static void freeary(struct ipc_namespace *, struct kern_ipc_perm *);
145 #ifdef CONFIG_PROC_FS
146 static int sysvipc_sem_proc_show(struct seq_file *s, void *it);
149 #define SEMMSL_FAST 256 /* 512 bytes on stack */
150 #define SEMOPM_FAST 64 /* ~ 372 bytes on stack */
153 * linked list protection:
155 * sem_array.sem_pending{,last},
156 * sem_array.sem_undo: sem_lock() for read/write
157 * sem_undo.proc_next: only "current" is allowed to read/write that field.
161 #define sc_semmsl sem_ctls[0]
162 #define sc_semmns sem_ctls[1]
163 #define sc_semopm sem_ctls[2]
164 #define sc_semmni sem_ctls[3]
166 void sem_init_ns(struct ipc_namespace *ns)
168 ns->sc_semmsl = SEMMSL;
169 ns->sc_semmns = SEMMNS;
170 ns->sc_semopm = SEMOPM;
171 ns->sc_semmni = SEMMNI;
173 ipc_init_ids(&ns->ids[IPC_SEM_IDS]);
177 void sem_exit_ns(struct ipc_namespace *ns)
179 free_ipcs(ns, &sem_ids(ns), freeary);
180 idr_destroy(&ns->ids[IPC_SEM_IDS].ipcs_idr);
184 void __init sem_init (void)
186 sem_init_ns(&init_ipc_ns);
187 ipc_init_proc_interface("sysvipc/sem",
188 " key semid perms nsems uid gid cuid cgid otime ctime\n",
189 IPC_SEM_IDS, sysvipc_sem_proc_show);
193 * sem_lock_(check_) routines are called in the paths where the rw_mutex
196 static inline struct sem_array *sem_obtain_lock(struct ipc_namespace *ns, int id)
198 struct kern_ipc_perm *ipcp;
199 struct sem_array *sma;
202 ipcp = ipc_obtain_object(&sem_ids(ns), id);
204 sma = ERR_CAST(ipcp);
208 spin_lock(&ipcp->lock);
210 /* ipc_rmid() may have already freed the ID while sem_lock
211 * was spinning: verify that the structure is still valid
214 return container_of(ipcp, struct sem_array, sem_perm);
216 spin_unlock(&ipcp->lock);
217 sma = ERR_PTR(-EINVAL);
223 static inline struct sem_array *sem_obtain_object(struct ipc_namespace *ns, int id)
225 struct kern_ipc_perm *ipcp = ipc_obtain_object(&sem_ids(ns), id);
228 return ERR_CAST(ipcp);
230 return container_of(ipcp, struct sem_array, sem_perm);
233 static inline struct sem_array *sem_lock_check(struct ipc_namespace *ns,
236 struct kern_ipc_perm *ipcp = ipc_lock_check(&sem_ids(ns), id);
239 return ERR_CAST(ipcp);
241 return container_of(ipcp, struct sem_array, sem_perm);
244 static inline struct sem_array *sem_obtain_object_check(struct ipc_namespace *ns,
247 struct kern_ipc_perm *ipcp = ipc_obtain_object_check(&sem_ids(ns), id);
250 return ERR_CAST(ipcp);
252 return container_of(ipcp, struct sem_array, sem_perm);
255 static inline void sem_lock_and_putref(struct sem_array *sma)
257 ipc_lock_by_ptr(&sma->sem_perm);
261 static inline void sem_getref_and_unlock(struct sem_array *sma)
264 ipc_unlock(&(sma)->sem_perm);
267 static inline void sem_putref(struct sem_array *sma)
269 ipc_lock_by_ptr(&sma->sem_perm);
271 ipc_unlock(&(sma)->sem_perm);
275 * Call inside the rcu read section.
277 static inline void sem_getref(struct sem_array *sma)
279 spin_lock(&(sma)->sem_perm.lock);
281 ipc_unlock(&(sma)->sem_perm);
284 static inline void sem_rmid(struct ipc_namespace *ns, struct sem_array *s)
286 ipc_rmid(&sem_ids(ns), &s->sem_perm);
290 * Lockless wakeup algorithm:
291 * Without the check/retry algorithm a lockless wakeup is possible:
292 * - queue.status is initialized to -EINTR before blocking.
293 * - wakeup is performed by
294 * * unlinking the queue entry from sma->sem_pending
295 * * setting queue.status to IN_WAKEUP
296 * This is the notification for the blocked thread that a
297 * result value is imminent.
298 * * call wake_up_process
299 * * set queue.status to the final value.
300 * - the previously blocked thread checks queue.status:
301 * * if it's IN_WAKEUP, then it must wait until the value changes
302 * * if it's not -EINTR, then the operation was completed by
303 * update_queue. semtimedop can return queue.status without
304 * performing any operation on the sem array.
305 * * otherwise it must acquire the spinlock and check what's up.
307 * The two-stage algorithm is necessary to protect against the following
309 * - if queue.status is set after wake_up_process, then the woken up idle
310 * thread could race forward and try (and fail) to acquire sma->lock
311 * before update_queue had a chance to set queue.status
312 * - if queue.status is written before wake_up_process and if the
313 * blocked process is woken up by a signal between writing
314 * queue.status and the wake_up_process, then the woken up
315 * process could return from semtimedop and die by calling
316 * sys_exit before wake_up_process is called. Then wake_up_process
317 * will oops, because the task structure is already invalid.
318 * (yes, this happened on s390 with sysv msg).
324 * newary - Create a new semaphore set
326 * @params: ptr to the structure that contains key, semflg and nsems
328 * Called with sem_ids.rw_mutex held (as a writer)
331 static int newary(struct ipc_namespace *ns, struct ipc_params *params)
335 struct sem_array *sma;
337 key_t key = params->key;
338 int nsems = params->u.nsems;
339 int semflg = params->flg;
344 if (ns->used_sems + nsems > ns->sc_semmns)
347 size = sizeof (*sma) + nsems * sizeof (struct sem);
348 sma = ipc_rcu_alloc(size);
352 memset (sma, 0, size);
354 sma->sem_perm.mode = (semflg & S_IRWXUGO);
355 sma->sem_perm.key = key;
357 sma->sem_perm.security = NULL;
358 retval = security_sem_alloc(sma);
364 id = ipc_addid(&sem_ids(ns), &sma->sem_perm, ns->sc_semmni);
366 security_sem_free(sma);
370 ns->used_sems += nsems;
372 sma->sem_base = (struct sem *) &sma[1];
374 for (i = 0; i < nsems; i++)
375 INIT_LIST_HEAD(&sma->sem_base[i].sem_pending);
377 sma->complex_count = 0;
378 INIT_LIST_HEAD(&sma->sem_pending);
379 INIT_LIST_HEAD(&sma->list_id);
380 sma->sem_nsems = nsems;
381 sma->sem_ctime = get_seconds();
384 return sma->sem_perm.id;
389 * Called with sem_ids.rw_mutex and ipcp locked.
391 static inline int sem_security(struct kern_ipc_perm *ipcp, int semflg)
393 struct sem_array *sma;
395 sma = container_of(ipcp, struct sem_array, sem_perm);
396 return security_sem_associate(sma, semflg);
400 * Called with sem_ids.rw_mutex and ipcp locked.
402 static inline int sem_more_checks(struct kern_ipc_perm *ipcp,
403 struct ipc_params *params)
405 struct sem_array *sma;
407 sma = container_of(ipcp, struct sem_array, sem_perm);
408 if (params->u.nsems > sma->sem_nsems)
414 SYSCALL_DEFINE3(semget, key_t, key, int, nsems, int, semflg)
416 struct ipc_namespace *ns;
417 struct ipc_ops sem_ops;
418 struct ipc_params sem_params;
420 ns = current->nsproxy->ipc_ns;
422 if (nsems < 0 || nsems > ns->sc_semmsl)
425 sem_ops.getnew = newary;
426 sem_ops.associate = sem_security;
427 sem_ops.more_checks = sem_more_checks;
429 sem_params.key = key;
430 sem_params.flg = semflg;
431 sem_params.u.nsems = nsems;
433 return ipcget(ns, &sem_ids(ns), &sem_ops, &sem_params);
437 * Determine whether a sequence of semaphore operations would succeed
438 * all at once. Return 0 if yes, 1 if need to sleep, else return error code.
441 static int try_atomic_semop (struct sem_array * sma, struct sembuf * sops,
442 int nsops, struct sem_undo *un, int pid)
448 for (sop = sops; sop < sops + nsops; sop++) {
449 curr = sma->sem_base + sop->sem_num;
450 sem_op = sop->sem_op;
451 result = curr->semval;
453 if (!sem_op && result)
461 if (sop->sem_flg & SEM_UNDO) {
462 int undo = un->semadj[sop->sem_num] - sem_op;
464 * Exceeding the undo range is an error.
466 if (undo < (-SEMAEM - 1) || undo > SEMAEM)
469 curr->semval = result;
473 while (sop >= sops) {
474 sma->sem_base[sop->sem_num].sempid = pid;
475 if (sop->sem_flg & SEM_UNDO)
476 un->semadj[sop->sem_num] -= sop->sem_op;
487 if (sop->sem_flg & IPC_NOWAIT)
494 while (sop >= sops) {
495 sma->sem_base[sop->sem_num].semval -= sop->sem_op;
502 /** wake_up_sem_queue_prepare(q, error): Prepare wake-up
503 * @q: queue entry that must be signaled
504 * @error: Error value for the signal
506 * Prepare the wake-up of the queue entry q.
508 static void wake_up_sem_queue_prepare(struct list_head *pt,
509 struct sem_queue *q, int error)
511 if (list_empty(pt)) {
513 * Hold preempt off so that we don't get preempted and have the
514 * wakee busy-wait until we're scheduled back on.
518 q->status = IN_WAKEUP;
521 list_add_tail(&q->list, pt);
525 * wake_up_sem_queue_do(pt) - do the actual wake-up
526 * @pt: list of tasks to be woken up
528 * Do the actual wake-up.
529 * The function is called without any locks held, thus the semaphore array
530 * could be destroyed already and the tasks can disappear as soon as the
531 * status is set to the actual return code.
533 static void wake_up_sem_queue_do(struct list_head *pt)
535 struct sem_queue *q, *t;
538 did_something = !list_empty(pt);
539 list_for_each_entry_safe(q, t, pt, list) {
540 wake_up_process(q->sleeper);
541 /* q can disappear immediately after writing q->status. */
549 static void unlink_queue(struct sem_array *sma, struct sem_queue *q)
553 sma->complex_count--;
556 /** check_restart(sma, q)
557 * @sma: semaphore array
558 * @q: the operation that just completed
560 * update_queue is O(N^2) when it restarts scanning the whole queue of
561 * waiting operations. Therefore this function checks if the restart is
562 * really necessary. It is called after a previously waiting operation
565 static int check_restart(struct sem_array *sma, struct sem_queue *q)
570 /* if the operation didn't modify the array, then no restart */
574 /* pending complex operations are too difficult to analyse */
575 if (sma->complex_count)
578 /* we were a sleeping complex operation. Too difficult */
582 curr = sma->sem_base + q->sops[0].sem_num;
584 /* No-one waits on this queue */
585 if (list_empty(&curr->sem_pending))
588 /* the new semaphore value */
590 /* It is impossible that someone waits for the new value:
591 * - q is a previously sleeping simple operation that
592 * altered the array. It must be a decrement, because
593 * simple increments never sleep.
594 * - The value is not 0, thus wait-for-zero won't proceed.
595 * - If there are older (higher priority) decrements
596 * in the queue, then they have observed the original
597 * semval value and couldn't proceed. The operation
598 * decremented to value - thus they won't proceed either.
600 BUG_ON(q->sops[0].sem_op >= 0);
604 * semval is 0. Check if there are wait-for-zero semops.
605 * They must be the first entries in the per-semaphore queue
607 h = list_first_entry(&curr->sem_pending, struct sem_queue, list);
608 BUG_ON(h->nsops != 1);
609 BUG_ON(h->sops[0].sem_num != q->sops[0].sem_num);
611 /* Yes, there is a wait-for-zero semop. Restart */
612 if (h->sops[0].sem_op == 0)
615 /* Again - no-one is waiting for the new value. */
621 * update_queue(sma, semnum): Look for tasks that can be completed.
622 * @sma: semaphore array.
623 * @semnum: semaphore that was modified.
624 * @pt: list head for the tasks that must be woken up.
626 * update_queue must be called after a semaphore in a semaphore array
627 * was modified. If multiple semaphores were modified, update_queue must
628 * be called with semnum = -1, as well as with the number of each modified
630 * The tasks that must be woken up are added to @pt. The return code
631 * is stored in q->pid.
632 * The function return 1 if at least one semop was completed successfully.
634 static int update_queue(struct sem_array *sma, int semnum, struct list_head *pt)
637 struct list_head *walk;
638 struct list_head *pending_list;
639 int semop_completed = 0;
642 pending_list = &sma->sem_pending;
644 pending_list = &sma->sem_base[semnum].sem_pending;
647 walk = pending_list->next;
648 while (walk != pending_list) {
651 q = container_of(walk, struct sem_queue, list);
654 /* If we are scanning the single sop, per-semaphore list of
655 * one semaphore and that semaphore is 0, then it is not
656 * necessary to scan the "alter" entries: simple increments
657 * that affect only one entry succeed immediately and cannot
658 * be in the per semaphore pending queue, and decrements
659 * cannot be successful if the value is already 0.
661 if (semnum != -1 && sma->sem_base[semnum].semval == 0 &&
665 error = try_atomic_semop(sma, q->sops, q->nsops,
668 /* Does q->sleeper still need to sleep? */
672 unlink_queue(sma, q);
678 restart = check_restart(sma, q);
681 wake_up_sem_queue_prepare(pt, q, error);
685 return semop_completed;
689 * do_smart_update(sma, sops, nsops, otime, pt) - optimized update_queue
690 * @sma: semaphore array
691 * @sops: operations that were performed
692 * @nsops: number of operations
693 * @otime: force setting otime
694 * @pt: list head of the tasks that must be woken up.
696 * do_smart_update() does the required called to update_queue, based on the
697 * actual changes that were performed on the semaphore array.
698 * Note that the function does not do the actual wake-up: the caller is
699 * responsible for calling wake_up_sem_queue_do(@pt).
700 * It is safe to perform this call after dropping all locks.
702 static void do_smart_update(struct sem_array *sma, struct sembuf *sops, int nsops,
703 int otime, struct list_head *pt)
707 if (sma->complex_count || sops == NULL) {
708 if (update_queue(sma, -1, pt))
713 /* No semops; something special is going on. */
714 for (i = 0; i < sma->sem_nsems; i++) {
715 if (update_queue(sma, i, pt))
721 /* Check the semaphores that were modified. */
722 for (i = 0; i < nsops; i++) {
723 if (sops[i].sem_op > 0 ||
724 (sops[i].sem_op < 0 &&
725 sma->sem_base[sops[i].sem_num].semval == 0))
726 if (update_queue(sma, sops[i].sem_num, pt))
731 sma->sem_otime = get_seconds();
735 /* The following counts are associated to each semaphore:
736 * semncnt number of tasks waiting on semval being nonzero
737 * semzcnt number of tasks waiting on semval being zero
738 * This model assumes that a task waits on exactly one semaphore.
739 * Since semaphore operations are to be performed atomically, tasks actually
740 * wait on a whole sequence of semaphores simultaneously.
741 * The counts we return here are a rough approximation, but still
742 * warrant that semncnt+semzcnt>0 if the task is on the pending queue.
744 static int count_semncnt (struct sem_array * sma, ushort semnum)
747 struct sem_queue * q;
750 list_for_each_entry(q, &sma->sem_pending, list) {
751 struct sembuf * sops = q->sops;
752 int nsops = q->nsops;
754 for (i = 0; i < nsops; i++)
755 if (sops[i].sem_num == semnum
756 && (sops[i].sem_op < 0)
757 && !(sops[i].sem_flg & IPC_NOWAIT))
763 static int count_semzcnt (struct sem_array * sma, ushort semnum)
766 struct sem_queue * q;
769 list_for_each_entry(q, &sma->sem_pending, list) {
770 struct sembuf * sops = q->sops;
771 int nsops = q->nsops;
773 for (i = 0; i < nsops; i++)
774 if (sops[i].sem_num == semnum
775 && (sops[i].sem_op == 0)
776 && !(sops[i].sem_flg & IPC_NOWAIT))
782 /* Free a semaphore set. freeary() is called with sem_ids.rw_mutex locked
783 * as a writer and the spinlock for this semaphore set hold. sem_ids.rw_mutex
784 * remains locked on exit.
786 static void freeary(struct ipc_namespace *ns, struct kern_ipc_perm *ipcp)
788 struct sem_undo *un, *tu;
789 struct sem_queue *q, *tq;
790 struct sem_array *sma = container_of(ipcp, struct sem_array, sem_perm);
791 struct list_head tasks;
794 /* Free the existing undo structures for this semaphore set. */
795 assert_spin_locked(&sma->sem_perm.lock);
796 list_for_each_entry_safe(un, tu, &sma->list_id, list_id) {
797 list_del(&un->list_id);
798 spin_lock(&un->ulp->lock);
800 list_del_rcu(&un->list_proc);
801 spin_unlock(&un->ulp->lock);
805 /* Wake up all pending processes and let them fail with EIDRM. */
806 INIT_LIST_HEAD(&tasks);
807 list_for_each_entry_safe(q, tq, &sma->sem_pending, list) {
808 unlink_queue(sma, q);
809 wake_up_sem_queue_prepare(&tasks, q, -EIDRM);
811 for (i = 0; i < sma->sem_nsems; i++) {
812 struct sem *sem = sma->sem_base + i;
813 list_for_each_entry_safe(q, tq, &sem->sem_pending, list) {
814 unlink_queue(sma, q);
815 wake_up_sem_queue_prepare(&tasks, q, -EIDRM);
819 /* Remove the semaphore set from the IDR */
823 wake_up_sem_queue_do(&tasks);
824 ns->used_sems -= sma->sem_nsems;
825 security_sem_free(sma);
829 static unsigned long copy_semid_to_user(void __user *buf, struct semid64_ds *in, int version)
833 return copy_to_user(buf, in, sizeof(*in));
838 memset(&out, 0, sizeof(out));
840 ipc64_perm_to_ipc_perm(&in->sem_perm, &out.sem_perm);
842 out.sem_otime = in->sem_otime;
843 out.sem_ctime = in->sem_ctime;
844 out.sem_nsems = in->sem_nsems;
846 return copy_to_user(buf, &out, sizeof(out));
853 static int semctl_nolock(struct ipc_namespace *ns, int semid,
854 int cmd, int version, void __user *p)
857 struct sem_array *sma;
863 struct seminfo seminfo;
866 err = security_sem_semctl(NULL, cmd);
870 memset(&seminfo,0,sizeof(seminfo));
871 seminfo.semmni = ns->sc_semmni;
872 seminfo.semmns = ns->sc_semmns;
873 seminfo.semmsl = ns->sc_semmsl;
874 seminfo.semopm = ns->sc_semopm;
875 seminfo.semvmx = SEMVMX;
876 seminfo.semmnu = SEMMNU;
877 seminfo.semmap = SEMMAP;
878 seminfo.semume = SEMUME;
879 down_read(&sem_ids(ns).rw_mutex);
880 if (cmd == SEM_INFO) {
881 seminfo.semusz = sem_ids(ns).in_use;
882 seminfo.semaem = ns->used_sems;
884 seminfo.semusz = SEMUSZ;
885 seminfo.semaem = SEMAEM;
887 max_id = ipc_get_maxid(&sem_ids(ns));
888 up_read(&sem_ids(ns).rw_mutex);
889 if (copy_to_user(p, &seminfo, sizeof(struct seminfo)))
891 return (max_id < 0) ? 0: max_id;
896 struct semid64_ds tbuf;
899 memset(&tbuf, 0, sizeof(tbuf));
901 if (cmd == SEM_STAT) {
903 sma = sem_obtain_object(ns, semid);
908 id = sma->sem_perm.id;
911 sma = sem_obtain_object_check(ns, semid);
919 if (ipcperms(ns, &sma->sem_perm, S_IRUGO))
922 err = security_sem_semctl(sma, cmd);
926 kernel_to_ipc64_perm(&sma->sem_perm, &tbuf.sem_perm);
927 tbuf.sem_otime = sma->sem_otime;
928 tbuf.sem_ctime = sma->sem_ctime;
929 tbuf.sem_nsems = sma->sem_nsems;
931 if (copy_semid_to_user(p, &tbuf, version))
943 static int semctl_setval(struct ipc_namespace *ns, int semid, int semnum,
947 struct sem_array *sma;
951 struct list_head tasks;
953 #if defined(CONFIG_64BIT) && defined(__BIG_ENDIAN)
954 /* big-endian 64bit */
957 /* 32bit or little-endian 64bit */
961 sma = sem_lock_check(ns, semid);
965 INIT_LIST_HEAD(&tasks);
966 nsems = sma->sem_nsems;
969 if (ipcperms(ns, &sma->sem_perm, S_IWUGO))
972 err = security_sem_semctl(sma, SETVAL);
977 if(semnum < 0 || semnum >= nsems)
980 curr = &sma->sem_base[semnum];
983 if (val > SEMVMX || val < 0)
986 assert_spin_locked(&sma->sem_perm.lock);
987 list_for_each_entry(un, &sma->list_id, list_id)
988 un->semadj[semnum] = 0;
991 curr->sempid = task_tgid_vnr(current);
992 sma->sem_ctime = get_seconds();
993 /* maybe some queued-up processes were waiting for this */
994 do_smart_update(sma, NULL, 0, 0, &tasks);
998 wake_up_sem_queue_do(&tasks);
1002 static int semctl_main(struct ipc_namespace *ns, int semid, int semnum,
1003 int cmd, void __user *p)
1005 struct sem_array *sma;
1008 ushort fast_sem_io[SEMMSL_FAST];
1009 ushort* sem_io = fast_sem_io;
1010 struct list_head tasks;
1012 INIT_LIST_HEAD(&tasks);
1015 sma = sem_obtain_object_check(ns, semid);
1018 return PTR_ERR(sma);
1021 nsems = sma->sem_nsems;
1024 if (ipcperms(ns, &sma->sem_perm,
1025 cmd == SETALL ? S_IWUGO : S_IRUGO)) {
1030 err = security_sem_semctl(sma, cmd);
1040 ushort __user *array = p;
1043 if(nsems > SEMMSL_FAST) {
1046 sem_io = ipc_alloc(sizeof(ushort)*nsems);
1047 if(sem_io == NULL) {
1052 sem_lock_and_putref(sma);
1053 if (sma->sem_perm.deleted) {
1060 spin_lock(&sma->sem_perm.lock);
1061 for (i = 0; i < sma->sem_nsems; i++)
1062 sem_io[i] = sma->sem_base[i].semval;
1065 if(copy_to_user(array, sem_io, nsems*sizeof(ushort)))
1072 struct sem_undo *un;
1074 ipc_rcu_getref(sma);
1077 if(nsems > SEMMSL_FAST) {
1078 sem_io = ipc_alloc(sizeof(ushort)*nsems);
1079 if(sem_io == NULL) {
1085 if (copy_from_user (sem_io, p, nsems*sizeof(ushort))) {
1091 for (i = 0; i < nsems; i++) {
1092 if (sem_io[i] > SEMVMX) {
1098 sem_lock_and_putref(sma);
1099 if (sma->sem_perm.deleted) {
1105 for (i = 0; i < nsems; i++)
1106 sma->sem_base[i].semval = sem_io[i];
1108 assert_spin_locked(&sma->sem_perm.lock);
1109 list_for_each_entry(un, &sma->list_id, list_id) {
1110 for (i = 0; i < nsems; i++)
1113 sma->sem_ctime = get_seconds();
1114 /* maybe some queued-up processes were waiting for this */
1115 do_smart_update(sma, NULL, 0, 0, &tasks);
1119 /* GETVAL, GETPID, GETNCTN, GETZCNT: fall-through */
1122 if(semnum < 0 || semnum >= nsems)
1125 spin_lock(&sma->sem_perm.lock);
1126 curr = &sma->sem_base[semnum];
1136 err = count_semncnt(sma,semnum);
1139 err = count_semzcnt(sma,semnum);
1146 wake_up_sem_queue_do(&tasks);
1148 if(sem_io != fast_sem_io)
1149 ipc_free(sem_io, sizeof(ushort)*nsems);
1153 static inline unsigned long
1154 copy_semid_from_user(struct semid64_ds *out, void __user *buf, int version)
1158 if (copy_from_user(out, buf, sizeof(*out)))
1163 struct semid_ds tbuf_old;
1165 if(copy_from_user(&tbuf_old, buf, sizeof(tbuf_old)))
1168 out->sem_perm.uid = tbuf_old.sem_perm.uid;
1169 out->sem_perm.gid = tbuf_old.sem_perm.gid;
1170 out->sem_perm.mode = tbuf_old.sem_perm.mode;
1180 * This function handles some semctl commands which require the rw_mutex
1181 * to be held in write mode.
1182 * NOTE: no locks must be held, the rw_mutex is taken inside this function.
1184 static int semctl_down(struct ipc_namespace *ns, int semid,
1185 int cmd, int version, void __user *p)
1187 struct sem_array *sma;
1189 struct semid64_ds semid64;
1190 struct kern_ipc_perm *ipcp;
1192 if(cmd == IPC_SET) {
1193 if (copy_semid_from_user(&semid64, p, version))
1197 ipcp = ipcctl_pre_down_nolock(ns, &sem_ids(ns), semid, cmd,
1198 &semid64.sem_perm, 0);
1200 return PTR_ERR(ipcp);
1202 sma = container_of(ipcp, struct sem_array, sem_perm);
1204 err = security_sem_semctl(sma, cmd);
1212 ipc_lock_object(&sma->sem_perm);
1216 ipc_lock_object(&sma->sem_perm);
1217 err = ipc_update_perm(&semid64.sem_perm, ipcp);
1220 sma->sem_ctime = get_seconds();
1231 up_write(&sem_ids(ns).rw_mutex);
1235 SYSCALL_DEFINE4(semctl, int, semid, int, semnum, int, cmd, unsigned long, arg)
1238 struct ipc_namespace *ns;
1239 void __user *p = (void __user *)arg;
1244 version = ipc_parse_version(&cmd);
1245 ns = current->nsproxy->ipc_ns;
1252 return semctl_nolock(ns, semid, cmd, version, p);
1259 return semctl_main(ns, semid, semnum, cmd, p);
1261 return semctl_setval(ns, semid, semnum, arg);
1264 return semctl_down(ns, semid, cmd, version, p);
1270 /* If the task doesn't already have a undo_list, then allocate one
1271 * here. We guarantee there is only one thread using this undo list,
1272 * and current is THE ONE
1274 * If this allocation and assignment succeeds, but later
1275 * portions of this code fail, there is no need to free the sem_undo_list.
1276 * Just let it stay associated with the task, and it'll be freed later
1279 * This can block, so callers must hold no locks.
1281 static inline int get_undo_list(struct sem_undo_list **undo_listp)
1283 struct sem_undo_list *undo_list;
1285 undo_list = current->sysvsem.undo_list;
1287 undo_list = kzalloc(sizeof(*undo_list), GFP_KERNEL);
1288 if (undo_list == NULL)
1290 spin_lock_init(&undo_list->lock);
1291 atomic_set(&undo_list->refcnt, 1);
1292 INIT_LIST_HEAD(&undo_list->list_proc);
1294 current->sysvsem.undo_list = undo_list;
1296 *undo_listp = undo_list;
1300 static struct sem_undo *__lookup_undo(struct sem_undo_list *ulp, int semid)
1302 struct sem_undo *un;
1304 list_for_each_entry_rcu(un, &ulp->list_proc, list_proc) {
1305 if (un->semid == semid)
1311 static struct sem_undo *lookup_undo(struct sem_undo_list *ulp, int semid)
1313 struct sem_undo *un;
1315 assert_spin_locked(&ulp->lock);
1317 un = __lookup_undo(ulp, semid);
1319 list_del_rcu(&un->list_proc);
1320 list_add_rcu(&un->list_proc, &ulp->list_proc);
1326 * find_alloc_undo - Lookup (and if not present create) undo array
1328 * @semid: semaphore array id
1330 * The function looks up (and if not present creates) the undo structure.
1331 * The size of the undo structure depends on the size of the semaphore
1332 * array, thus the alloc path is not that straightforward.
1333 * Lifetime-rules: sem_undo is rcu-protected, on success, the function
1334 * performs a rcu_read_lock().
1336 static struct sem_undo *find_alloc_undo(struct ipc_namespace *ns, int semid)
1338 struct sem_array *sma;
1339 struct sem_undo_list *ulp;
1340 struct sem_undo *un, *new;
1344 error = get_undo_list(&ulp);
1346 return ERR_PTR(error);
1349 spin_lock(&ulp->lock);
1350 un = lookup_undo(ulp, semid);
1351 spin_unlock(&ulp->lock);
1352 if (likely(un!=NULL))
1355 /* no undo structure around - allocate one. */
1356 /* step 1: figure out the size of the semaphore array */
1357 sma = sem_obtain_object_check(ns, semid);
1360 return ERR_CAST(sma);
1363 nsems = sma->sem_nsems;
1364 ipc_rcu_getref(sma);
1367 /* step 2: allocate new undo structure */
1368 new = kzalloc(sizeof(struct sem_undo) + sizeof(short)*nsems, GFP_KERNEL);
1371 return ERR_PTR(-ENOMEM);
1374 /* step 3: Acquire the lock on semaphore array */
1375 sem_lock_and_putref(sma);
1376 if (sma->sem_perm.deleted) {
1379 un = ERR_PTR(-EIDRM);
1382 spin_lock(&ulp->lock);
1385 * step 4: check for races: did someone else allocate the undo struct?
1387 un = lookup_undo(ulp, semid);
1392 /* step 5: initialize & link new undo structure */
1393 new->semadj = (short *) &new[1];
1396 assert_spin_locked(&ulp->lock);
1397 list_add_rcu(&new->list_proc, &ulp->list_proc);
1398 assert_spin_locked(&sma->sem_perm.lock);
1399 list_add(&new->list_id, &sma->list_id);
1403 spin_unlock(&ulp->lock);
1412 * get_queue_result - Retrieve the result code from sem_queue
1413 * @q: Pointer to queue structure
1415 * Retrieve the return code from the pending queue. If IN_WAKEUP is found in
1416 * q->status, then we must loop until the value is replaced with the final
1417 * value: This may happen if a task is woken up by an unrelated event (e.g.
1418 * signal) and in parallel the task is woken up by another task because it got
1419 * the requested semaphores.
1421 * The function can be called with or without holding the semaphore spinlock.
1423 static int get_queue_result(struct sem_queue *q)
1428 while (unlikely(error == IN_WAKEUP)) {
1437 SYSCALL_DEFINE4(semtimedop, int, semid, struct sembuf __user *, tsops,
1438 unsigned, nsops, const struct timespec __user *, timeout)
1440 int error = -EINVAL;
1441 struct sem_array *sma;
1442 struct sembuf fast_sops[SEMOPM_FAST];
1443 struct sembuf* sops = fast_sops, *sop;
1444 struct sem_undo *un;
1445 int undos = 0, alter = 0, max;
1446 struct sem_queue queue;
1447 unsigned long jiffies_left = 0;
1448 struct ipc_namespace *ns;
1449 struct list_head tasks;
1451 ns = current->nsproxy->ipc_ns;
1453 if (nsops < 1 || semid < 0)
1455 if (nsops > ns->sc_semopm)
1457 if(nsops > SEMOPM_FAST) {
1458 sops = kmalloc(sizeof(*sops)*nsops,GFP_KERNEL);
1462 if (copy_from_user (sops, tsops, nsops * sizeof(*tsops))) {
1467 struct timespec _timeout;
1468 if (copy_from_user(&_timeout, timeout, sizeof(*timeout))) {
1472 if (_timeout.tv_sec < 0 || _timeout.tv_nsec < 0 ||
1473 _timeout.tv_nsec >= 1000000000L) {
1477 jiffies_left = timespec_to_jiffies(&_timeout);
1480 for (sop = sops; sop < sops + nsops; sop++) {
1481 if (sop->sem_num >= max)
1483 if (sop->sem_flg & SEM_UNDO)
1485 if (sop->sem_op != 0)
1490 un = find_alloc_undo(ns, semid);
1492 error = PTR_ERR(un);
1498 INIT_LIST_HEAD(&tasks);
1501 sma = sem_obtain_object_check(ns, semid);
1505 error = PTR_ERR(sma);
1510 if (max >= sma->sem_nsems) {
1516 if (ipcperms(ns, &sma->sem_perm, alter ? S_IWUGO : S_IRUGO)) {
1521 error = security_sem_semop(sma, sops, nsops, alter);
1528 * semid identifiers are not unique - find_alloc_undo may have
1529 * allocated an undo structure, it was invalidated by an RMID
1530 * and now a new array with received the same id. Check and fail.
1531 * This case can be detected checking un->semid. The existence of
1532 * "un" itself is guaranteed by rcu.
1535 ipc_lock_object(&sma->sem_perm);
1537 if (un->semid == -1) {
1539 goto out_unlock_free;
1542 * rcu lock can be released, "un" cannot disappear:
1543 * - sem_lock is acquired, thus IPC_RMID is
1545 * - exit_sem is impossible, it always operates on
1546 * current (or a dead task).
1553 error = try_atomic_semop (sma, sops, nsops, un, task_tgid_vnr(current));
1555 if (alter && error == 0)
1556 do_smart_update(sma, sops, nsops, 1, &tasks);
1558 goto out_unlock_free;
1561 /* We need to sleep on this operation, so we put the current
1562 * task into the pending queue and go to sleep.
1566 queue.nsops = nsops;
1568 queue.pid = task_tgid_vnr(current);
1569 queue.alter = alter;
1573 curr = &sma->sem_base[sops->sem_num];
1576 list_add_tail(&queue.list, &curr->sem_pending);
1578 list_add(&queue.list, &curr->sem_pending);
1581 list_add_tail(&queue.list, &sma->sem_pending);
1583 list_add(&queue.list, &sma->sem_pending);
1584 sma->complex_count++;
1587 queue.status = -EINTR;
1588 queue.sleeper = current;
1591 current->state = TASK_INTERRUPTIBLE;
1595 jiffies_left = schedule_timeout(jiffies_left);
1599 error = get_queue_result(&queue);
1601 if (error != -EINTR) {
1602 /* fast path: update_queue already obtained all requested
1604 * Perform a smp_mb(): User space could assume that semop()
1605 * is a memory barrier: Without the mb(), the cpu could
1606 * speculatively read in user space stale data that was
1607 * overwritten by the previous owner of the semaphore.
1614 sma = sem_obtain_lock(ns, semid);
1617 * Wait until it's guaranteed that no wakeup_sem_queue_do() is ongoing.
1619 error = get_queue_result(&queue);
1622 * Array removed? If yes, leave without sem_unlock().
1630 * If queue.status != -EINTR we are woken up by another process.
1631 * Leave without unlink_queue(), but with sem_unlock().
1634 if (error != -EINTR) {
1635 goto out_unlock_free;
1639 * If an interrupt occurred we have to clean up the queue
1641 if (timeout && jiffies_left == 0)
1645 * If the wakeup was spurious, just retry
1647 if (error == -EINTR && !signal_pending(current))
1650 unlink_queue(sma, &queue);
1655 wake_up_sem_queue_do(&tasks);
1657 if(sops != fast_sops)
1662 SYSCALL_DEFINE3(semop, int, semid, struct sembuf __user *, tsops,
1665 return sys_semtimedop(semid, tsops, nsops, NULL);
1668 /* If CLONE_SYSVSEM is set, establish sharing of SEM_UNDO state between
1669 * parent and child tasks.
1672 int copy_semundo(unsigned long clone_flags, struct task_struct *tsk)
1674 struct sem_undo_list *undo_list;
1677 if (clone_flags & CLONE_SYSVSEM) {
1678 error = get_undo_list(&undo_list);
1681 atomic_inc(&undo_list->refcnt);
1682 tsk->sysvsem.undo_list = undo_list;
1684 tsk->sysvsem.undo_list = NULL;
1690 * add semadj values to semaphores, free undo structures.
1691 * undo structures are not freed when semaphore arrays are destroyed
1692 * so some of them may be out of date.
1693 * IMPLEMENTATION NOTE: There is some confusion over whether the
1694 * set of adjustments that needs to be done should be done in an atomic
1695 * manner or not. That is, if we are attempting to decrement the semval
1696 * should we queue up and wait until we can do so legally?
1697 * The original implementation attempted to do this (queue and wait).
1698 * The current implementation does not do so. The POSIX standard
1699 * and SVID should be consulted to determine what behavior is mandated.
1701 void exit_sem(struct task_struct *tsk)
1703 struct sem_undo_list *ulp;
1705 ulp = tsk->sysvsem.undo_list;
1708 tsk->sysvsem.undo_list = NULL;
1710 if (!atomic_dec_and_test(&ulp->refcnt))
1714 struct sem_array *sma;
1715 struct sem_undo *un;
1716 struct list_head tasks;
1721 un = list_entry_rcu(ulp->list_proc.next,
1722 struct sem_undo, list_proc);
1723 if (&un->list_proc == &ulp->list_proc)
1732 sma = sem_lock_check(tsk->nsproxy->ipc_ns, un->semid);
1734 /* exit_sem raced with IPC_RMID, nothing to do */
1738 un = __lookup_undo(ulp, semid);
1740 /* exit_sem raced with IPC_RMID+semget() that created
1741 * exactly the same semid. Nothing to do.
1747 /* remove un from the linked lists */
1748 assert_spin_locked(&sma->sem_perm.lock);
1749 list_del(&un->list_id);
1751 spin_lock(&ulp->lock);
1752 list_del_rcu(&un->list_proc);
1753 spin_unlock(&ulp->lock);
1755 /* perform adjustments registered in un */
1756 for (i = 0; i < sma->sem_nsems; i++) {
1757 struct sem * semaphore = &sma->sem_base[i];
1758 if (un->semadj[i]) {
1759 semaphore->semval += un->semadj[i];
1761 * Range checks of the new semaphore value,
1762 * not defined by sus:
1763 * - Some unices ignore the undo entirely
1764 * (e.g. HP UX 11i 11.22, Tru64 V5.1)
1765 * - some cap the value (e.g. FreeBSD caps
1766 * at 0, but doesn't enforce SEMVMX)
1768 * Linux caps the semaphore value, both at 0
1771 * Manfred <manfred@colorfullife.com>
1773 if (semaphore->semval < 0)
1774 semaphore->semval = 0;
1775 if (semaphore->semval > SEMVMX)
1776 semaphore->semval = SEMVMX;
1777 semaphore->sempid = task_tgid_vnr(current);
1780 /* maybe some queued-up processes were waiting for this */
1781 INIT_LIST_HEAD(&tasks);
1782 do_smart_update(sma, NULL, 0, 1, &tasks);
1784 wake_up_sem_queue_do(&tasks);
1791 #ifdef CONFIG_PROC_FS
1792 static int sysvipc_sem_proc_show(struct seq_file *s, void *it)
1794 struct user_namespace *user_ns = seq_user_ns(s);
1795 struct sem_array *sma = it;
1797 return seq_printf(s,
1798 "%10d %10d %4o %10u %5u %5u %5u %5u %10lu %10lu\n",
1803 from_kuid_munged(user_ns, sma->sem_perm.uid),
1804 from_kgid_munged(user_ns, sma->sem_perm.gid),
1805 from_kuid_munged(user_ns, sma->sem_perm.cuid),
1806 from_kgid_munged(user_ns, sma->sem_perm.cgid),