3 * Copyright (C) 1992 Krishna Balasubramanian
5 * Sep 1997 - Call suser() last after "normal" permission checks so we
6 * get BSD style process accounting right.
7 * Occurs in several places in the IPC code.
8 * Chris Evans, <chris@ferret.lmh.ox.ac.uk>
9 * Nov 1999 - ipc helper functions, unified SMP locking
10 * Manfred Spraul <manfred@colorfullife.com>
11 * Oct 2002 - One lock per IPC id. RCU ipc_free for lock-free grow_ary().
12 * Mingming Cao <cmm@us.ibm.com>
13 * Mar 2006 - support for audit of ipc object properties
14 * Dustin Kirkland <dustin.kirkland@us.ibm.com>
15 * Jun 2006 - namespaces ssupport
17 * Pavel Emelianov <xemul@openvz.org>
19 * General sysv ipc locking scheme:
21 * obtain the ipc object (kern_ipc_perm) by looking up the id in an idr
23 * - perform initial checks (capabilities, auditing and permission,
25 * - perform read-only operations, such as STAT, INFO commands.
26 * acquire the ipc lock (kern_ipc_perm.lock) through
28 * - perform data updates, such as SET, RMID commands and
29 * mechanism-specific operations (semop/semtimedop,
30 * msgsnd/msgrcv, shmat/shmdt).
31 * drop the ipc lock, through ipc_unlock_object().
34 * The ids->rwsem must be taken when:
35 * - creating, removing and iterating the existing entries in ipc
37 * - iterating through files under /proc/sysvipc/
39 * Note that sems have a special fast path that avoids kern_ipc_perm.lock -
44 #include <linux/shm.h>
45 #include <linux/init.h>
46 #include <linux/msg.h>
47 #include <linux/vmalloc.h>
48 #include <linux/slab.h>
49 #include <linux/notifier.h>
50 #include <linux/capability.h>
51 #include <linux/highuid.h>
52 #include <linux/security.h>
53 #include <linux/rcupdate.h>
54 #include <linux/workqueue.h>
55 #include <linux/seq_file.h>
56 #include <linux/proc_fs.h>
57 #include <linux/audit.h>
58 #include <linux/nsproxy.h>
59 #include <linux/rwsem.h>
60 #include <linux/memory.h>
61 #include <linux/ipc_namespace.h>
63 #include <asm/unistd.h>
67 struct ipc_proc_iface {
71 int (*show)(struct seq_file *, void *);
74 static void ipc_memory_notifier(struct work_struct *work)
76 ipcns_notify(IPCNS_MEMCHANGED);
79 static int ipc_memory_callback(struct notifier_block *self,
80 unsigned long action, void *arg)
82 static DECLARE_WORK(ipc_memory_wq, ipc_memory_notifier);
85 case MEM_ONLINE: /* memory successfully brought online */
86 case MEM_OFFLINE: /* or offline: it's time to recompute msgmni */
88 * This is done by invoking the ipcns notifier chain with the
89 * IPC_MEMCHANGED event.
90 * In order not to keep the lock on the hotplug memory chain
91 * for too long, queue a work item that will, when waken up,
92 * activate the ipcns notification chain.
94 schedule_work(&ipc_memory_wq);
96 case MEM_GOING_ONLINE:
97 case MEM_GOING_OFFLINE:
98 case MEM_CANCEL_ONLINE:
99 case MEM_CANCEL_OFFLINE:
107 static struct notifier_block ipc_memory_nb = {
108 .notifier_call = ipc_memory_callback,
109 .priority = IPC_CALLBACK_PRI,
113 * ipc_init - initialise ipc subsystem
115 * The various sysv ipc resources (semaphores, messages and shared
116 * memory) are initialised.
118 * A callback routine is registered into the memory hotplug notifier
119 * chain: since msgmni scales to lowmem this callback routine will be
120 * called upon successful memory add / remove to recompute msmgni.
122 static int __init ipc_init(void)
127 register_hotmemory_notifier(&ipc_memory_nb);
128 register_ipcns_notifier(&init_ipc_ns);
131 __initcall(ipc_init);
134 * ipc_init_ids - initialise ipc identifiers
135 * @ids: ipc identifier set
137 * Set up the sequence range to use for the ipc identifier range (limited
138 * below IPCMNI) then initialise the ids idr.
140 void ipc_init_ids(struct ipc_ids *ids)
142 init_rwsem(&ids->rwsem);
148 int seq_limit = INT_MAX/SEQ_MULTIPLIER;
149 if (seq_limit > USHRT_MAX)
150 ids->seq_max = USHRT_MAX;
152 ids->seq_max = seq_limit;
155 idr_init(&ids->ipcs_idr);
158 #ifdef CONFIG_PROC_FS
159 static const struct file_operations sysvipc_proc_fops;
161 * ipc_init_proc_interface - create a proc interface for sysipc types using a seq_file interface.
162 * @path: Path in procfs
163 * @header: Banner to be printed at the beginning of the file.
164 * @ids: ipc id table to iterate.
165 * @show: show routine.
167 void __init ipc_init_proc_interface(const char *path, const char *header,
168 int ids, int (*show)(struct seq_file *, void *))
170 struct proc_dir_entry *pde;
171 struct ipc_proc_iface *iface;
173 iface = kmalloc(sizeof(*iface), GFP_KERNEL);
177 iface->header = header;
181 pde = proc_create_data(path,
182 S_IRUGO, /* world readable */
183 NULL, /* parent dir */
192 * ipc_findkey - find a key in an ipc identifier set
193 * @ids: ipc identifier set
196 * Returns the locked pointer to the ipc structure if found or NULL
197 * otherwise. If key is found ipc points to the owning ipc structure
199 * Called with ipc_ids.rwsem held.
201 static struct kern_ipc_perm *ipc_findkey(struct ipc_ids *ids, key_t key)
203 struct kern_ipc_perm *ipc;
207 for (total = 0, next_id = 0; total < ids->in_use; next_id++) {
208 ipc = idr_find(&ids->ipcs_idr, next_id);
213 if (ipc->key != key) {
219 ipc_lock_object(ipc);
227 * ipc_get_maxid - get the last assigned id
228 * @ids: ipc identifier set
230 * Called with ipc_ids.rwsem held.
232 int ipc_get_maxid(struct ipc_ids *ids)
234 struct kern_ipc_perm *ipc;
238 if (ids->in_use == 0)
241 if (ids->in_use == IPCMNI)
244 /* Look for the last assigned id */
246 for (id = 0; id < IPCMNI && total < ids->in_use; id++) {
247 ipc = idr_find(&ids->ipcs_idr, id);
257 * ipc_addid - add an ipc identifier
258 * @ids: ipc identifier set
259 * @new: new ipc permission set
260 * @size: limit for the number of used ids
262 * Add an entry 'new' to the ipc ids idr. The permissions object is
263 * initialised and the first free entry is set up and the id assigned
264 * is returned. The 'new' entry is returned in a locked state on success.
265 * On failure the entry is not locked and a negative err-code is returned.
267 * Called with writer ipc_ids.rwsem held.
269 int ipc_addid(struct ipc_ids *ids, struct kern_ipc_perm *new, int size)
274 int next_id = ids->next_id;
279 if (ids->in_use >= size)
282 idr_preload(GFP_KERNEL);
284 spin_lock_init(&new->lock);
285 new->deleted = false;
287 spin_lock(&new->lock);
289 id = idr_alloc(&ids->ipcs_idr, new,
290 (next_id < 0) ? 0 : ipcid_to_idx(next_id), 0,
294 spin_unlock(&new->lock);
301 current_euid_egid(&euid, &egid);
302 new->cuid = new->uid = euid;
303 new->gid = new->cgid = egid;
306 new->seq = ids->seq++;
307 if (ids->seq > ids->seq_max)
310 new->seq = ipcid_to_seqx(next_id);
314 new->id = ipc_buildid(id, new->seq);
319 * ipcget_new - create a new ipc object
321 * @ids: ipc identifer set
322 * @ops: the actual creation routine to call
323 * @params: its parameters
325 * This routine is called by sys_msgget, sys_semget() and sys_shmget()
326 * when the key is IPC_PRIVATE.
328 static int ipcget_new(struct ipc_namespace *ns, struct ipc_ids *ids,
329 struct ipc_ops *ops, struct ipc_params *params)
333 down_write(&ids->rwsem);
334 err = ops->getnew(ns, params);
335 up_write(&ids->rwsem);
340 * ipc_check_perms - check security and permissions for an ipc object
342 * @ipcp: ipc permission set
343 * @ops: the actual security routine to call
344 * @params: its parameters
346 * This routine is called by sys_msgget(), sys_semget() and sys_shmget()
347 * when the key is not IPC_PRIVATE and that key already exists in the
350 * On success, the ipc id is returned.
352 * It is called with ipc_ids.rwsem and ipcp->lock held.
354 static int ipc_check_perms(struct ipc_namespace *ns,
355 struct kern_ipc_perm *ipcp,
357 struct ipc_params *params)
361 if (ipcperms(ns, ipcp, params->flg))
364 err = ops->associate(ipcp, params->flg);
373 * ipcget_public - get an ipc object or create a new one
375 * @ids: ipc identifer set
376 * @ops: the actual creation routine to call
377 * @params: its parameters
379 * This routine is called by sys_msgget, sys_semget() and sys_shmget()
380 * when the key is not IPC_PRIVATE.
381 * It adds a new entry if the key is not found and does some permission
382 * / security checkings if the key is found.
384 * On success, the ipc id is returned.
386 static int ipcget_public(struct ipc_namespace *ns, struct ipc_ids *ids,
387 struct ipc_ops *ops, struct ipc_params *params)
389 struct kern_ipc_perm *ipcp;
390 int flg = params->flg;
394 * Take the lock as a writer since we are potentially going to add
395 * a new entry + read locks are not "upgradable"
397 down_write(&ids->rwsem);
398 ipcp = ipc_findkey(ids, params->key);
401 if (!(flg & IPC_CREAT))
404 err = ops->getnew(ns, params);
406 /* ipc object has been locked by ipc_findkey() */
408 if (flg & IPC_CREAT && flg & IPC_EXCL)
412 if (ops->more_checks)
413 err = ops->more_checks(ipcp, params);
416 * ipc_check_perms returns the IPC id on
419 err = ipc_check_perms(ns, ipcp, ops, params);
423 up_write(&ids->rwsem);
430 * ipc_rmid - remove an ipc identifier
431 * @ids: ipc identifier set
432 * @ipcp: ipc perm structure containing the identifier to remove
434 * ipc_ids.rwsem (as a writer) and the spinlock for this ID are held
435 * before this function is called, and remain locked on the exit.
437 void ipc_rmid(struct ipc_ids *ids, struct kern_ipc_perm *ipcp)
439 int lid = ipcid_to_idx(ipcp->id);
441 idr_remove(&ids->ipcs_idr, lid);
445 ipcp->deleted = true;
451 * ipc_alloc - allocate ipc space
452 * @size: size desired
454 * Allocate memory from the appropriate pools and return a pointer to it.
455 * NULL is returned if the allocation fails
457 void *ipc_alloc(int size)
460 if (size > PAGE_SIZE)
463 out = kmalloc(size, GFP_KERNEL);
468 * ipc_free - free ipc space
469 * @ptr: pointer returned by ipc_alloc
470 * @size: size of block
472 * Free a block created with ipc_alloc(). The caller must know the size
473 * used in the allocation call.
475 void ipc_free(void *ptr, int size)
477 if (size > PAGE_SIZE)
484 * ipc_rcu_alloc - allocate ipc and rcu space
485 * @size: size desired
487 * Allocate memory for the rcu header structure + the object.
488 * Returns the pointer to the object or NULL upon failure.
490 void *ipc_rcu_alloc(int size)
493 * We prepend the allocation with the rcu struct
495 struct ipc_rcu *out = ipc_alloc(sizeof(struct ipc_rcu) + size);
498 atomic_set(&out->refcount, 1);
502 int ipc_rcu_getref(void *ptr)
504 struct ipc_rcu *p = ((struct ipc_rcu *)ptr) - 1;
506 return atomic_inc_not_zero(&p->refcount);
509 void ipc_rcu_putref(void *ptr, void (*func)(struct rcu_head *head))
511 struct ipc_rcu *p = ((struct ipc_rcu *)ptr) - 1;
513 if (!atomic_dec_and_test(&p->refcount))
516 call_rcu(&p->rcu, func);
519 void ipc_rcu_free(struct rcu_head *head)
521 struct ipc_rcu *p = container_of(head, struct ipc_rcu, rcu);
523 if (is_vmalloc_addr(p))
530 * ipcperms - check ipc permissions
532 * @ipcp: ipc permission set
533 * @flag: desired permission set
535 * Check user, group, other permissions for access
536 * to ipc resources. return 0 if allowed
538 * @flag will most probably be 0 or S_...UGO from <linux/stat.h>
540 int ipcperms(struct ipc_namespace *ns, struct kern_ipc_perm *ipcp, short flag)
542 kuid_t euid = current_euid();
543 int requested_mode, granted_mode;
546 requested_mode = (flag >> 6) | (flag >> 3) | flag;
547 granted_mode = ipcp->mode;
548 if (uid_eq(euid, ipcp->cuid) ||
549 uid_eq(euid, ipcp->uid))
551 else if (in_group_p(ipcp->cgid) || in_group_p(ipcp->gid))
553 /* is there some bit set in requested_mode but not in granted_mode? */
554 if ((requested_mode & ~granted_mode & 0007) &&
555 !ns_capable(ns->user_ns, CAP_IPC_OWNER))
558 return security_ipc_permission(ipcp, flag);
562 * Functions to convert between the kern_ipc_perm structure and the
563 * old/new ipc_perm structures
567 * kernel_to_ipc64_perm - convert kernel ipc permissions to user
568 * @in: kernel permissions
569 * @out: new style ipc permissions
571 * Turn the kernel object @in into a set of permissions descriptions
572 * for returning to userspace (@out).
574 void kernel_to_ipc64_perm(struct kern_ipc_perm *in, struct ipc64_perm *out)
577 out->uid = from_kuid_munged(current_user_ns(), in->uid);
578 out->gid = from_kgid_munged(current_user_ns(), in->gid);
579 out->cuid = from_kuid_munged(current_user_ns(), in->cuid);
580 out->cgid = from_kgid_munged(current_user_ns(), in->cgid);
581 out->mode = in->mode;
586 * ipc64_perm_to_ipc_perm - convert new ipc permissions to old
587 * @in: new style ipc permissions
588 * @out: old style ipc permissions
590 * Turn the new style permissions object @in into a compatibility
591 * object and store it into the @out pointer.
593 void ipc64_perm_to_ipc_perm(struct ipc64_perm *in, struct ipc_perm *out)
596 SET_UID(out->uid, in->uid);
597 SET_GID(out->gid, in->gid);
598 SET_UID(out->cuid, in->cuid);
599 SET_GID(out->cgid, in->cgid);
600 out->mode = in->mode;
606 * @ids: ipc identifier set
607 * @id: ipc id to look for
609 * Look for an id in the ipc ids idr and return associated ipc object.
611 * Call inside the RCU critical section.
612 * The ipc object is *not* locked on exit.
614 struct kern_ipc_perm *ipc_obtain_object(struct ipc_ids *ids, int id)
616 struct kern_ipc_perm *out;
617 int lid = ipcid_to_idx(id);
619 out = idr_find(&ids->ipcs_idr, lid);
621 return ERR_PTR(-EINVAL);
627 * ipc_lock - lock an ipc structure without rwsem held
628 * @ids: ipc identifier set
629 * @id: ipc id to look for
631 * Look for an id in the ipc ids idr and lock the associated ipc object.
633 * The ipc object is locked on successful exit.
635 struct kern_ipc_perm *ipc_lock(struct ipc_ids *ids, int id)
637 struct kern_ipc_perm *out;
640 out = ipc_obtain_object(ids, id);
644 spin_lock(&out->lock);
646 /* ipc_rmid() may have already freed the ID while ipc_lock
647 * was spinning: here verify that the structure is still valid
649 if (ipc_valid_object(out))
652 spin_unlock(&out->lock);
653 out = ERR_PTR(-EINVAL);
660 * ipc_obtain_object_check
661 * @ids: ipc identifier set
662 * @id: ipc id to look for
664 * Similar to ipc_obtain_object() but also checks
665 * the ipc object reference counter.
667 * Call inside the RCU critical section.
668 * The ipc object is *not* locked on exit.
670 struct kern_ipc_perm *ipc_obtain_object_check(struct ipc_ids *ids, int id)
672 struct kern_ipc_perm *out = ipc_obtain_object(ids, id);
677 if (ipc_checkid(out, id))
678 return ERR_PTR(-EIDRM);
684 * ipcget - Common sys_*get() code
686 * @ids: ipc identifier set
687 * @ops: operations to be called on ipc object creation, permission checks
689 * @params: the parameters needed by the previous operations.
691 * Common routine called by sys_msgget(), sys_semget() and sys_shmget().
693 int ipcget(struct ipc_namespace *ns, struct ipc_ids *ids,
694 struct ipc_ops *ops, struct ipc_params *params)
696 if (params->key == IPC_PRIVATE)
697 return ipcget_new(ns, ids, ops, params);
699 return ipcget_public(ns, ids, ops, params);
703 * ipc_update_perm - update the permissions of an ipc object
704 * @in: the permission given as input.
705 * @out: the permission of the ipc to set.
707 int ipc_update_perm(struct ipc64_perm *in, struct kern_ipc_perm *out)
709 kuid_t uid = make_kuid(current_user_ns(), in->uid);
710 kgid_t gid = make_kgid(current_user_ns(), in->gid);
711 if (!uid_valid(uid) || !gid_valid(gid))
716 out->mode = (out->mode & ~S_IRWXUGO)
717 | (in->mode & S_IRWXUGO);
723 * ipcctl_pre_down_nolock - retrieve an ipc and check permissions for some IPC_XXX cmd
725 * @ids: the table of ids where to look for the ipc
726 * @id: the id of the ipc to retrieve
727 * @cmd: the cmd to check
728 * @perm: the permission to set
729 * @extra_perm: one extra permission parameter used by msq
731 * This function does some common audit and permissions check for some IPC_XXX
732 * cmd and is called from semctl_down, shmctl_down and msgctl_down.
733 * It must be called without any lock held and
734 * - retrieves the ipc with the given id in the given table.
735 * - performs some audit and permission check, depending on the given cmd
736 * - returns a pointer to the ipc object or otherwise, the corresponding error.
738 * Call holding the both the rwsem and the rcu read lock.
740 struct kern_ipc_perm *ipcctl_pre_down_nolock(struct ipc_namespace *ns,
741 struct ipc_ids *ids, int id, int cmd,
742 struct ipc64_perm *perm, int extra_perm)
746 struct kern_ipc_perm *ipcp;
748 ipcp = ipc_obtain_object_check(ids, id);
756 audit_ipc_set_perm(extra_perm, perm->uid,
757 perm->gid, perm->mode);
759 euid = current_euid();
760 if (uid_eq(euid, ipcp->cuid) || uid_eq(euid, ipcp->uid) ||
761 ns_capable(ns->user_ns, CAP_SYS_ADMIN))
762 return ipcp; /* successful lookup */
767 #ifdef CONFIG_ARCH_WANT_IPC_PARSE_VERSION
771 * ipc_parse_version - ipc call version
772 * @cmd: pointer to command
774 * Return IPC_64 for new style IPC and IPC_OLD for old style IPC.
775 * The @cmd value is turned from an encoding command and version into
776 * just the command code.
778 int ipc_parse_version(int *cmd)
788 #endif /* CONFIG_ARCH_WANT_IPC_PARSE_VERSION */
790 #ifdef CONFIG_PROC_FS
791 struct ipc_proc_iter {
792 struct ipc_namespace *ns;
793 struct ipc_proc_iface *iface;
797 * This routine locks the ipc structure found at least at position pos.
799 static struct kern_ipc_perm *sysvipc_find_ipc(struct ipc_ids *ids, loff_t pos,
802 struct kern_ipc_perm *ipc;
806 for (id = 0; id < pos && total < ids->in_use; id++) {
807 ipc = idr_find(&ids->ipcs_idr, id);
812 if (total >= ids->in_use)
815 for (; pos < IPCMNI; pos++) {
816 ipc = idr_find(&ids->ipcs_idr, pos);
820 ipc_lock_object(ipc);
825 /* Out of range - return NULL to terminate iteration */
829 static void *sysvipc_proc_next(struct seq_file *s, void *it, loff_t *pos)
831 struct ipc_proc_iter *iter = s->private;
832 struct ipc_proc_iface *iface = iter->iface;
833 struct kern_ipc_perm *ipc = it;
835 /* If we had an ipc id locked before, unlock it */
836 if (ipc && ipc != SEQ_START_TOKEN)
839 return sysvipc_find_ipc(&iter->ns->ids[iface->ids], *pos, pos);
843 * File positions: pos 0 -> header, pos n -> ipc id = n - 1.
844 * SeqFile iterator: iterator value locked ipc pointer or SEQ_TOKEN_START.
846 static void *sysvipc_proc_start(struct seq_file *s, loff_t *pos)
848 struct ipc_proc_iter *iter = s->private;
849 struct ipc_proc_iface *iface = iter->iface;
852 ids = &iter->ns->ids[iface->ids];
855 * Take the lock - this will be released by the corresponding
858 down_read(&ids->rwsem);
860 /* pos < 0 is invalid */
864 /* pos == 0 means header */
866 return SEQ_START_TOKEN;
868 /* Find the (pos-1)th ipc */
869 return sysvipc_find_ipc(ids, *pos - 1, pos);
872 static void sysvipc_proc_stop(struct seq_file *s, void *it)
874 struct kern_ipc_perm *ipc = it;
875 struct ipc_proc_iter *iter = s->private;
876 struct ipc_proc_iface *iface = iter->iface;
879 /* If we had a locked structure, release it */
880 if (ipc && ipc != SEQ_START_TOKEN)
883 ids = &iter->ns->ids[iface->ids];
884 /* Release the lock we took in start() */
885 up_read(&ids->rwsem);
888 static int sysvipc_proc_show(struct seq_file *s, void *it)
890 struct ipc_proc_iter *iter = s->private;
891 struct ipc_proc_iface *iface = iter->iface;
893 if (it == SEQ_START_TOKEN)
894 return seq_puts(s, iface->header);
896 return iface->show(s, it);
899 static const struct seq_operations sysvipc_proc_seqops = {
900 .start = sysvipc_proc_start,
901 .stop = sysvipc_proc_stop,
902 .next = sysvipc_proc_next,
903 .show = sysvipc_proc_show,
906 static int sysvipc_proc_open(struct inode *inode, struct file *file)
909 struct seq_file *seq;
910 struct ipc_proc_iter *iter;
913 iter = kmalloc(sizeof(*iter), GFP_KERNEL);
917 ret = seq_open(file, &sysvipc_proc_seqops);
921 seq = file->private_data;
924 iter->iface = PDE_DATA(inode);
925 iter->ns = get_ipc_ns(current->nsproxy->ipc_ns);
933 static int sysvipc_proc_release(struct inode *inode, struct file *file)
935 struct seq_file *seq = file->private_data;
936 struct ipc_proc_iter *iter = seq->private;
937 put_ipc_ns(iter->ns);
938 return seq_release_private(inode, file);
941 static const struct file_operations sysvipc_proc_fops = {
942 .open = sysvipc_proc_open,
945 .release = sysvipc_proc_release,
947 #endif /* CONFIG_PROC_FS */