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)
145 init_rwsem(&ids->rwsem);
146 idr_init(&ids->ipcs_idr);
149 #ifdef CONFIG_PROC_FS
150 static const struct file_operations sysvipc_proc_fops;
152 * ipc_init_proc_interface - create a proc interface for sysipc types using a seq_file interface.
153 * @path: Path in procfs
154 * @header: Banner to be printed at the beginning of the file.
155 * @ids: ipc id table to iterate.
156 * @show: show routine.
158 void __init ipc_init_proc_interface(const char *path, const char *header,
159 int ids, int (*show)(struct seq_file *, void *))
161 struct proc_dir_entry *pde;
162 struct ipc_proc_iface *iface;
164 iface = kmalloc(sizeof(*iface), GFP_KERNEL);
168 iface->header = header;
172 pde = proc_create_data(path,
173 S_IRUGO, /* world readable */
174 NULL, /* parent dir */
183 * ipc_findkey - find a key in an ipc identifier set
184 * @ids: ipc identifier set
187 * Returns the locked pointer to the ipc structure if found or NULL
188 * otherwise. If key is found ipc points to the owning ipc structure
190 * Called with ipc_ids.rwsem held.
192 static struct kern_ipc_perm *ipc_findkey(struct ipc_ids *ids, key_t key)
194 struct kern_ipc_perm *ipc;
198 for (total = 0, next_id = 0; total < ids->in_use; next_id++) {
199 ipc = idr_find(&ids->ipcs_idr, next_id);
204 if (ipc->key != key) {
210 ipc_lock_object(ipc);
218 * ipc_get_maxid - get the last assigned id
219 * @ids: ipc identifier set
221 * Called with ipc_ids.rwsem held.
223 int ipc_get_maxid(struct ipc_ids *ids)
225 struct kern_ipc_perm *ipc;
229 if (ids->in_use == 0)
232 if (ids->in_use == IPCMNI)
235 /* Look for the last assigned id */
237 for (id = 0; id < IPCMNI && total < ids->in_use; id++) {
238 ipc = idr_find(&ids->ipcs_idr, id);
248 * ipc_addid - add an ipc identifier
249 * @ids: ipc identifier set
250 * @new: new ipc permission set
251 * @size: limit for the number of used ids
253 * Add an entry 'new' to the ipc ids idr. The permissions object is
254 * initialised and the first free entry is set up and the id assigned
255 * is returned. The 'new' entry is returned in a locked state on success.
256 * On failure the entry is not locked and a negative err-code is returned.
258 * Called with writer ipc_ids.rwsem held.
260 int ipc_addid(struct ipc_ids *ids, struct kern_ipc_perm *new, int size)
265 int next_id = ids->next_id;
270 if (ids->in_use >= size)
273 idr_preload(GFP_KERNEL);
275 spin_lock_init(&new->lock);
276 new->deleted = false;
278 spin_lock(&new->lock);
280 id = idr_alloc(&ids->ipcs_idr, new,
281 (next_id < 0) ? 0 : ipcid_to_idx(next_id), 0,
285 spin_unlock(&new->lock);
292 current_euid_egid(&euid, &egid);
293 new->cuid = new->uid = euid;
294 new->gid = new->cgid = egid;
297 new->seq = ids->seq++;
298 if (ids->seq > IPCID_SEQ_MAX)
301 new->seq = ipcid_to_seqx(next_id);
305 new->id = ipc_buildid(id, new->seq);
310 * ipcget_new - create a new ipc object
312 * @ids: ipc identifer set
313 * @ops: the actual creation routine to call
314 * @params: its parameters
316 * This routine is called by sys_msgget, sys_semget() and sys_shmget()
317 * when the key is IPC_PRIVATE.
319 static int ipcget_new(struct ipc_namespace *ns, struct ipc_ids *ids,
320 struct ipc_ops *ops, struct ipc_params *params)
324 down_write(&ids->rwsem);
325 err = ops->getnew(ns, params);
326 up_write(&ids->rwsem);
331 * ipc_check_perms - check security and permissions for an ipc object
333 * @ipcp: ipc permission set
334 * @ops: the actual security routine to call
335 * @params: its parameters
337 * This routine is called by sys_msgget(), sys_semget() and sys_shmget()
338 * when the key is not IPC_PRIVATE and that key already exists in the
341 * On success, the ipc id is returned.
343 * It is called with ipc_ids.rwsem and ipcp->lock held.
345 static int ipc_check_perms(struct ipc_namespace *ns,
346 struct kern_ipc_perm *ipcp,
348 struct ipc_params *params)
352 if (ipcperms(ns, ipcp, params->flg))
355 err = ops->associate(ipcp, params->flg);
364 * ipcget_public - get an ipc object or create a new one
366 * @ids: ipc identifer set
367 * @ops: the actual creation routine to call
368 * @params: its parameters
370 * This routine is called by sys_msgget, sys_semget() and sys_shmget()
371 * when the key is not IPC_PRIVATE.
372 * It adds a new entry if the key is not found and does some permission
373 * / security checkings if the key is found.
375 * On success, the ipc id is returned.
377 static int ipcget_public(struct ipc_namespace *ns, struct ipc_ids *ids,
378 struct ipc_ops *ops, struct ipc_params *params)
380 struct kern_ipc_perm *ipcp;
381 int flg = params->flg;
385 * Take the lock as a writer since we are potentially going to add
386 * a new entry + read locks are not "upgradable"
388 down_write(&ids->rwsem);
389 ipcp = ipc_findkey(ids, params->key);
392 if (!(flg & IPC_CREAT))
395 err = ops->getnew(ns, params);
397 /* ipc object has been locked by ipc_findkey() */
399 if (flg & IPC_CREAT && flg & IPC_EXCL)
403 if (ops->more_checks)
404 err = ops->more_checks(ipcp, params);
407 * ipc_check_perms returns the IPC id on
410 err = ipc_check_perms(ns, ipcp, ops, params);
414 up_write(&ids->rwsem);
421 * ipc_rmid - remove an ipc identifier
422 * @ids: ipc identifier set
423 * @ipcp: ipc perm structure containing the identifier to remove
425 * ipc_ids.rwsem (as a writer) and the spinlock for this ID are held
426 * before this function is called, and remain locked on the exit.
428 void ipc_rmid(struct ipc_ids *ids, struct kern_ipc_perm *ipcp)
430 int lid = ipcid_to_idx(ipcp->id);
432 idr_remove(&ids->ipcs_idr, lid);
434 ipcp->deleted = true;
438 * ipc_alloc - allocate ipc space
439 * @size: size desired
441 * Allocate memory from the appropriate pools and return a pointer to it.
442 * NULL is returned if the allocation fails
444 void *ipc_alloc(int size)
447 if (size > PAGE_SIZE)
450 out = kmalloc(size, GFP_KERNEL);
455 * ipc_free - free ipc space
456 * @ptr: pointer returned by ipc_alloc
457 * @size: size of block
459 * Free a block created with ipc_alloc(). The caller must know the size
460 * used in the allocation call.
462 void ipc_free(void *ptr, int size)
464 if (size > PAGE_SIZE)
471 * ipc_rcu_alloc - allocate ipc and rcu space
472 * @size: size desired
474 * Allocate memory for the rcu header structure + the object.
475 * Returns the pointer to the object or NULL upon failure.
477 void *ipc_rcu_alloc(int size)
480 * We prepend the allocation with the rcu struct
482 struct ipc_rcu *out = ipc_alloc(sizeof(struct ipc_rcu) + size);
485 atomic_set(&out->refcount, 1);
489 int ipc_rcu_getref(void *ptr)
491 struct ipc_rcu *p = ((struct ipc_rcu *)ptr) - 1;
493 return atomic_inc_not_zero(&p->refcount);
496 void ipc_rcu_putref(void *ptr, void (*func)(struct rcu_head *head))
498 struct ipc_rcu *p = ((struct ipc_rcu *)ptr) - 1;
500 if (!atomic_dec_and_test(&p->refcount))
503 call_rcu(&p->rcu, func);
506 void ipc_rcu_free(struct rcu_head *head)
508 struct ipc_rcu *p = container_of(head, struct ipc_rcu, rcu);
510 if (is_vmalloc_addr(p))
517 * ipcperms - check ipc permissions
519 * @ipcp: ipc permission set
520 * @flag: desired permission set
522 * Check user, group, other permissions for access
523 * to ipc resources. return 0 if allowed
525 * @flag will most probably be 0 or S_...UGO from <linux/stat.h>
527 int ipcperms(struct ipc_namespace *ns, struct kern_ipc_perm *ipcp, short flag)
529 kuid_t euid = current_euid();
530 int requested_mode, granted_mode;
533 requested_mode = (flag >> 6) | (flag >> 3) | flag;
534 granted_mode = ipcp->mode;
535 if (uid_eq(euid, ipcp->cuid) ||
536 uid_eq(euid, ipcp->uid))
538 else if (in_group_p(ipcp->cgid) || in_group_p(ipcp->gid))
540 /* is there some bit set in requested_mode but not in granted_mode? */
541 if ((requested_mode & ~granted_mode & 0007) &&
542 !ns_capable(ns->user_ns, CAP_IPC_OWNER))
545 return security_ipc_permission(ipcp, flag);
549 * Functions to convert between the kern_ipc_perm structure and the
550 * old/new ipc_perm structures
554 * kernel_to_ipc64_perm - convert kernel ipc permissions to user
555 * @in: kernel permissions
556 * @out: new style ipc permissions
558 * Turn the kernel object @in into a set of permissions descriptions
559 * for returning to userspace (@out).
561 void kernel_to_ipc64_perm(struct kern_ipc_perm *in, struct ipc64_perm *out)
564 out->uid = from_kuid_munged(current_user_ns(), in->uid);
565 out->gid = from_kgid_munged(current_user_ns(), in->gid);
566 out->cuid = from_kuid_munged(current_user_ns(), in->cuid);
567 out->cgid = from_kgid_munged(current_user_ns(), in->cgid);
568 out->mode = in->mode;
573 * ipc64_perm_to_ipc_perm - convert new ipc permissions to old
574 * @in: new style ipc permissions
575 * @out: old style ipc permissions
577 * Turn the new style permissions object @in into a compatibility
578 * object and store it into the @out pointer.
580 void ipc64_perm_to_ipc_perm(struct ipc64_perm *in, struct ipc_perm *out)
583 SET_UID(out->uid, in->uid);
584 SET_GID(out->gid, in->gid);
585 SET_UID(out->cuid, in->cuid);
586 SET_GID(out->cgid, in->cgid);
587 out->mode = in->mode;
593 * @ids: ipc identifier set
594 * @id: ipc id to look for
596 * Look for an id in the ipc ids idr and return associated ipc object.
598 * Call inside the RCU critical section.
599 * The ipc object is *not* locked on exit.
601 struct kern_ipc_perm *ipc_obtain_object(struct ipc_ids *ids, int id)
603 struct kern_ipc_perm *out;
604 int lid = ipcid_to_idx(id);
606 out = idr_find(&ids->ipcs_idr, lid);
608 return ERR_PTR(-EINVAL);
614 * ipc_lock - lock an ipc structure without rwsem held
615 * @ids: ipc identifier set
616 * @id: ipc id to look for
618 * Look for an id in the ipc ids idr and lock the associated ipc object.
620 * The ipc object is locked on successful exit.
622 struct kern_ipc_perm *ipc_lock(struct ipc_ids *ids, int id)
624 struct kern_ipc_perm *out;
627 out = ipc_obtain_object(ids, id);
631 spin_lock(&out->lock);
633 /* ipc_rmid() may have already freed the ID while ipc_lock
634 * was spinning: here verify that the structure is still valid
636 if (ipc_valid_object(out))
639 spin_unlock(&out->lock);
640 out = ERR_PTR(-EINVAL);
647 * ipc_obtain_object_check
648 * @ids: ipc identifier set
649 * @id: ipc id to look for
651 * Similar to ipc_obtain_object() but also checks
652 * the ipc object reference counter.
654 * Call inside the RCU critical section.
655 * The ipc object is *not* locked on exit.
657 struct kern_ipc_perm *ipc_obtain_object_check(struct ipc_ids *ids, int id)
659 struct kern_ipc_perm *out = ipc_obtain_object(ids, id);
664 if (ipc_checkid(out, id))
665 return ERR_PTR(-EIDRM);
671 * ipcget - Common sys_*get() code
673 * @ids: ipc identifier set
674 * @ops: operations to be called on ipc object creation, permission checks
676 * @params: the parameters needed by the previous operations.
678 * Common routine called by sys_msgget(), sys_semget() and sys_shmget().
680 int ipcget(struct ipc_namespace *ns, struct ipc_ids *ids,
681 struct ipc_ops *ops, struct ipc_params *params)
683 if (params->key == IPC_PRIVATE)
684 return ipcget_new(ns, ids, ops, params);
686 return ipcget_public(ns, ids, ops, params);
690 * ipc_update_perm - update the permissions of an ipc object
691 * @in: the permission given as input.
692 * @out: the permission of the ipc to set.
694 int ipc_update_perm(struct ipc64_perm *in, struct kern_ipc_perm *out)
696 kuid_t uid = make_kuid(current_user_ns(), in->uid);
697 kgid_t gid = make_kgid(current_user_ns(), in->gid);
698 if (!uid_valid(uid) || !gid_valid(gid))
703 out->mode = (out->mode & ~S_IRWXUGO)
704 | (in->mode & S_IRWXUGO);
710 * ipcctl_pre_down_nolock - retrieve an ipc and check permissions for some IPC_XXX cmd
712 * @ids: the table of ids where to look for the ipc
713 * @id: the id of the ipc to retrieve
714 * @cmd: the cmd to check
715 * @perm: the permission to set
716 * @extra_perm: one extra permission parameter used by msq
718 * This function does some common audit and permissions check for some IPC_XXX
719 * cmd and is called from semctl_down, shmctl_down and msgctl_down.
720 * It must be called without any lock held and
721 * - retrieves the ipc with the given id in the given table.
722 * - performs some audit and permission check, depending on the given cmd
723 * - returns a pointer to the ipc object or otherwise, the corresponding error.
725 * Call holding the both the rwsem and the rcu read lock.
727 struct kern_ipc_perm *ipcctl_pre_down_nolock(struct ipc_namespace *ns,
728 struct ipc_ids *ids, int id, int cmd,
729 struct ipc64_perm *perm, int extra_perm)
733 struct kern_ipc_perm *ipcp;
735 ipcp = ipc_obtain_object_check(ids, id);
743 audit_ipc_set_perm(extra_perm, perm->uid,
744 perm->gid, perm->mode);
746 euid = current_euid();
747 if (uid_eq(euid, ipcp->cuid) || uid_eq(euid, ipcp->uid) ||
748 ns_capable(ns->user_ns, CAP_SYS_ADMIN))
749 return ipcp; /* successful lookup */
754 #ifdef CONFIG_ARCH_WANT_IPC_PARSE_VERSION
758 * ipc_parse_version - ipc call version
759 * @cmd: pointer to command
761 * Return IPC_64 for new style IPC and IPC_OLD for old style IPC.
762 * The @cmd value is turned from an encoding command and version into
763 * just the command code.
765 int ipc_parse_version(int *cmd)
775 #endif /* CONFIG_ARCH_WANT_IPC_PARSE_VERSION */
777 #ifdef CONFIG_PROC_FS
778 struct ipc_proc_iter {
779 struct ipc_namespace *ns;
780 struct ipc_proc_iface *iface;
784 * This routine locks the ipc structure found at least at position pos.
786 static struct kern_ipc_perm *sysvipc_find_ipc(struct ipc_ids *ids, loff_t pos,
789 struct kern_ipc_perm *ipc;
793 for (id = 0; id < pos && total < ids->in_use; id++) {
794 ipc = idr_find(&ids->ipcs_idr, id);
799 if (total >= ids->in_use)
802 for (; pos < IPCMNI; pos++) {
803 ipc = idr_find(&ids->ipcs_idr, pos);
807 ipc_lock_object(ipc);
812 /* Out of range - return NULL to terminate iteration */
816 static void *sysvipc_proc_next(struct seq_file *s, void *it, loff_t *pos)
818 struct ipc_proc_iter *iter = s->private;
819 struct ipc_proc_iface *iface = iter->iface;
820 struct kern_ipc_perm *ipc = it;
822 /* If we had an ipc id locked before, unlock it */
823 if (ipc && ipc != SEQ_START_TOKEN)
826 return sysvipc_find_ipc(&iter->ns->ids[iface->ids], *pos, pos);
830 * File positions: pos 0 -> header, pos n -> ipc id = n - 1.
831 * SeqFile iterator: iterator value locked ipc pointer or SEQ_TOKEN_START.
833 static void *sysvipc_proc_start(struct seq_file *s, loff_t *pos)
835 struct ipc_proc_iter *iter = s->private;
836 struct ipc_proc_iface *iface = iter->iface;
839 ids = &iter->ns->ids[iface->ids];
842 * Take the lock - this will be released by the corresponding
845 down_read(&ids->rwsem);
847 /* pos < 0 is invalid */
851 /* pos == 0 means header */
853 return SEQ_START_TOKEN;
855 /* Find the (pos-1)th ipc */
856 return sysvipc_find_ipc(ids, *pos - 1, pos);
859 static void sysvipc_proc_stop(struct seq_file *s, void *it)
861 struct kern_ipc_perm *ipc = it;
862 struct ipc_proc_iter *iter = s->private;
863 struct ipc_proc_iface *iface = iter->iface;
866 /* If we had a locked structure, release it */
867 if (ipc && ipc != SEQ_START_TOKEN)
870 ids = &iter->ns->ids[iface->ids];
871 /* Release the lock we took in start() */
872 up_read(&ids->rwsem);
875 static int sysvipc_proc_show(struct seq_file *s, void *it)
877 struct ipc_proc_iter *iter = s->private;
878 struct ipc_proc_iface *iface = iter->iface;
880 if (it == SEQ_START_TOKEN)
881 return seq_puts(s, iface->header);
883 return iface->show(s, it);
886 static const struct seq_operations sysvipc_proc_seqops = {
887 .start = sysvipc_proc_start,
888 .stop = sysvipc_proc_stop,
889 .next = sysvipc_proc_next,
890 .show = sysvipc_proc_show,
893 static int sysvipc_proc_open(struct inode *inode, struct file *file)
896 struct seq_file *seq;
897 struct ipc_proc_iter *iter;
900 iter = kmalloc(sizeof(*iter), GFP_KERNEL);
904 ret = seq_open(file, &sysvipc_proc_seqops);
910 seq = file->private_data;
913 iter->iface = PDE_DATA(inode);
914 iter->ns = get_ipc_ns(current->nsproxy->ipc_ns);
919 static int sysvipc_proc_release(struct inode *inode, struct file *file)
921 struct seq_file *seq = file->private_data;
922 struct ipc_proc_iter *iter = seq->private;
923 put_ipc_ns(iter->ns);
924 return seq_release_private(inode, file);
927 static const struct file_operations sysvipc_proc_fops = {
928 .open = sysvipc_proc_open,
931 .release = sysvipc_proc_release,
933 #endif /* CONFIG_PROC_FS */