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.
93 * No need to keep several ipc work items on the queue.
95 if (!work_pending(&ipc_memory_wq))
96 schedule_work(&ipc_memory_wq);
98 case MEM_GOING_ONLINE:
99 case MEM_GOING_OFFLINE:
100 case MEM_CANCEL_ONLINE:
101 case MEM_CANCEL_OFFLINE:
109 static struct notifier_block ipc_memory_nb = {
110 .notifier_call = ipc_memory_callback,
111 .priority = IPC_CALLBACK_PRI,
115 * ipc_init - initialise IPC subsystem
117 * The various system5 IPC resources (semaphores, messages and shared
118 * memory) are initialised
119 * A callback routine is registered into the memory hotplug notifier
120 * chain: since msgmni scales to lowmem this callback routine will be
121 * called upon successful memory add / remove to recompute msmgni.
124 static int __init ipc_init(void)
129 register_hotmemory_notifier(&ipc_memory_nb);
130 register_ipcns_notifier(&init_ipc_ns);
133 __initcall(ipc_init);
136 * ipc_init_ids - initialise IPC identifiers
137 * @ids: Identifier set
139 * Set up the sequence range to use for the ipc identifier range (limited
140 * below IPCMNI) then initialise the ids idr.
143 void ipc_init_ids(struct ipc_ids *ids)
145 init_rwsem(&ids->rwsem);
151 int seq_limit = INT_MAX/SEQ_MULTIPLIER;
152 if (seq_limit > USHRT_MAX)
153 ids->seq_max = USHRT_MAX;
155 ids->seq_max = seq_limit;
158 idr_init(&ids->ipcs_idr);
161 #ifdef CONFIG_PROC_FS
162 static const struct file_operations sysvipc_proc_fops;
164 * ipc_init_proc_interface - Create a proc interface for sysipc types using a seq_file interface.
165 * @path: Path in procfs
166 * @header: Banner to be printed at the beginning of the file.
167 * @ids: ipc id table to iterate.
168 * @show: show routine.
170 void __init ipc_init_proc_interface(const char *path, const char *header,
171 int ids, int (*show)(struct seq_file *, void *))
173 struct proc_dir_entry *pde;
174 struct ipc_proc_iface *iface;
176 iface = kmalloc(sizeof(*iface), GFP_KERNEL);
180 iface->header = header;
184 pde = proc_create_data(path,
185 S_IRUGO, /* world readable */
186 NULL, /* parent dir */
196 * ipc_findkey - find a key in an ipc identifier set
197 * @ids: Identifier set
198 * @key: The key to find
200 * Requires ipc_ids.rwsem locked.
201 * Returns the LOCKED pointer to the ipc structure if found or NULL
203 * If key is found ipc points to the owning ipc structure
206 static struct kern_ipc_perm *ipc_findkey(struct ipc_ids *ids, key_t key)
208 struct kern_ipc_perm *ipc;
212 for (total = 0, next_id = 0; total < ids->in_use; next_id++) {
213 ipc = idr_find(&ids->ipcs_idr, next_id);
218 if (ipc->key != key) {
224 ipc_lock_object(ipc);
232 * ipc_get_maxid - get the last assigned id
233 * @ids: IPC identifier set
235 * Called with ipc_ids.rwsem held.
238 int ipc_get_maxid(struct ipc_ids *ids)
240 struct kern_ipc_perm *ipc;
244 if (ids->in_use == 0)
247 if (ids->in_use == IPCMNI)
250 /* Look for the last assigned id */
252 for (id = 0; id < IPCMNI && total < ids->in_use; id++) {
253 ipc = idr_find(&ids->ipcs_idr, id);
263 * ipc_addid - add an IPC identifier
264 * @ids: IPC identifier set
265 * @new: new IPC permission set
266 * @size: limit for the number of used ids
268 * Add an entry 'new' to the IPC ids idr. The permissions object is
269 * initialised and the first free entry is set up and the id assigned
270 * is returned. The 'new' entry is returned in a locked state on success.
271 * On failure the entry is not locked and a negative err-code is returned.
273 * Called with writer ipc_ids.rwsem held.
275 int ipc_addid(struct ipc_ids* ids, struct kern_ipc_perm* new, int size)
280 int next_id = ids->next_id;
285 if (ids->in_use >= size)
288 idr_preload(GFP_KERNEL);
290 spin_lock_init(&new->lock);
293 spin_lock(&new->lock);
295 id = idr_alloc(&ids->ipcs_idr, new,
296 (next_id < 0) ? 0 : ipcid_to_idx(next_id), 0,
300 spin_unlock(&new->lock);
307 current_euid_egid(&euid, &egid);
308 new->cuid = new->uid = euid;
309 new->gid = new->cgid = egid;
312 new->seq = ids->seq++;
313 if (ids->seq > ids->seq_max)
316 new->seq = ipcid_to_seqx(next_id);
320 new->id = ipc_buildid(id, new->seq);
325 * ipcget_new - create a new ipc object
327 * @ids: IPC identifer set
328 * @ops: the actual creation routine to call
329 * @params: its parameters
331 * This routine is called by sys_msgget, sys_semget() and sys_shmget()
332 * when the key is IPC_PRIVATE.
334 static int ipcget_new(struct ipc_namespace *ns, struct ipc_ids *ids,
335 struct ipc_ops *ops, struct ipc_params *params)
339 down_write(&ids->rwsem);
340 err = ops->getnew(ns, params);
341 up_write(&ids->rwsem);
346 * ipc_check_perms - check security and permissions for an IPC
348 * @ipcp: ipc permission set
349 * @ops: the actual security routine to call
350 * @params: its parameters
352 * This routine is called by sys_msgget(), sys_semget() and sys_shmget()
353 * when the key is not IPC_PRIVATE and that key already exists in the
356 * On success, the IPC id is returned.
358 * It is called with ipc_ids.rwsem and ipcp->lock held.
360 static int ipc_check_perms(struct ipc_namespace *ns,
361 struct kern_ipc_perm *ipcp,
363 struct ipc_params *params)
367 if (ipcperms(ns, ipcp, params->flg))
370 err = ops->associate(ipcp, params->flg);
379 * ipcget_public - get an ipc object or create a new one
381 * @ids: IPC identifer set
382 * @ops: the actual creation routine to call
383 * @params: its parameters
385 * This routine is called by sys_msgget, sys_semget() and sys_shmget()
386 * when the key is not IPC_PRIVATE.
387 * It adds a new entry if the key is not found and does some permission
388 * / security checkings if the key is found.
390 * On success, the ipc id is returned.
392 static int ipcget_public(struct ipc_namespace *ns, struct ipc_ids *ids,
393 struct ipc_ops *ops, struct ipc_params *params)
395 struct kern_ipc_perm *ipcp;
396 int flg = params->flg;
400 * Take the lock as a writer since we are potentially going to add
401 * a new entry + read locks are not "upgradable"
403 down_write(&ids->rwsem);
404 ipcp = ipc_findkey(ids, params->key);
407 if (!(flg & IPC_CREAT))
410 err = ops->getnew(ns, params);
412 /* ipc object has been locked by ipc_findkey() */
414 if (flg & IPC_CREAT && flg & IPC_EXCL)
418 if (ops->more_checks)
419 err = ops->more_checks(ipcp, params);
422 * ipc_check_perms returns the IPC id on
425 err = ipc_check_perms(ns, ipcp, ops, params);
429 up_write(&ids->rwsem);
436 * ipc_rmid - remove an IPC identifier
437 * @ids: IPC identifier set
438 * @ipcp: ipc perm structure containing the identifier to remove
440 * ipc_ids.rwsem (as a writer) and the spinlock for this ID are held
441 * before this function is called, and remain locked on the exit.
444 void ipc_rmid(struct ipc_ids *ids, struct kern_ipc_perm *ipcp)
446 int lid = ipcid_to_idx(ipcp->id);
448 idr_remove(&ids->ipcs_idr, lid);
458 * ipc_alloc - allocate ipc space
459 * @size: size desired
461 * Allocate memory from the appropriate pools and return a pointer to it.
462 * NULL is returned if the allocation fails
465 void *ipc_alloc(int size)
471 out = kmalloc(size, GFP_KERNEL);
476 * ipc_free - free ipc space
477 * @ptr: pointer returned by ipc_alloc
478 * @size: size of block
480 * Free a block created with ipc_alloc(). The caller must know the size
481 * used in the allocation call.
484 void ipc_free(void* ptr, int size)
493 * ipc_rcu_alloc - allocate ipc and rcu space
494 * @size: size desired
496 * Allocate memory for the rcu header structure + the object.
497 * Returns the pointer to the object or NULL upon failure.
499 void *ipc_rcu_alloc(int size)
502 * We prepend the allocation with the rcu struct
504 struct ipc_rcu *out = ipc_alloc(sizeof(struct ipc_rcu) + size);
507 atomic_set(&out->refcount, 1);
511 int ipc_rcu_getref(void *ptr)
513 struct ipc_rcu *p = ((struct ipc_rcu *)ptr) - 1;
515 return atomic_inc_not_zero(&p->refcount);
518 void ipc_rcu_putref(void *ptr, void (*func)(struct rcu_head *head))
520 struct ipc_rcu *p = ((struct ipc_rcu *)ptr) - 1;
522 if (!atomic_dec_and_test(&p->refcount))
525 call_rcu(&p->rcu, func);
528 void ipc_rcu_free(struct rcu_head *head)
530 struct ipc_rcu *p = container_of(head, struct ipc_rcu, rcu);
532 if (is_vmalloc_addr(p))
539 * ipcperms - check IPC permissions
541 * @ipcp: IPC permission set
542 * @flag: desired permission set.
544 * Check user, group, other permissions for access
545 * to ipc resources. return 0 if allowed
547 * @flag will most probably be 0 or S_...UGO from <linux/stat.h>
550 int ipcperms(struct ipc_namespace *ns, struct kern_ipc_perm *ipcp, short flag)
552 kuid_t euid = current_euid();
553 int requested_mode, granted_mode;
556 requested_mode = (flag >> 6) | (flag >> 3) | flag;
557 granted_mode = ipcp->mode;
558 if (uid_eq(euid, ipcp->cuid) ||
559 uid_eq(euid, ipcp->uid))
561 else if (in_group_p(ipcp->cgid) || in_group_p(ipcp->gid))
563 /* is there some bit set in requested_mode but not in granted_mode? */
564 if ((requested_mode & ~granted_mode & 0007) &&
565 !ns_capable(ns->user_ns, CAP_IPC_OWNER))
568 return security_ipc_permission(ipcp, flag);
572 * Functions to convert between the kern_ipc_perm structure and the
573 * old/new ipc_perm structures
577 * kernel_to_ipc64_perm - convert kernel ipc permissions to user
578 * @in: kernel permissions
579 * @out: new style IPC permissions
581 * Turn the kernel object @in into a set of permissions descriptions
582 * for returning to userspace (@out).
586 void kernel_to_ipc64_perm (struct kern_ipc_perm *in, struct ipc64_perm *out)
589 out->uid = from_kuid_munged(current_user_ns(), in->uid);
590 out->gid = from_kgid_munged(current_user_ns(), in->gid);
591 out->cuid = from_kuid_munged(current_user_ns(), in->cuid);
592 out->cgid = from_kgid_munged(current_user_ns(), in->cgid);
593 out->mode = in->mode;
598 * ipc64_perm_to_ipc_perm - convert new ipc permissions to old
599 * @in: new style IPC permissions
600 * @out: old style IPC permissions
602 * Turn the new style permissions object @in into a compatibility
603 * object and store it into the @out pointer.
606 void ipc64_perm_to_ipc_perm (struct ipc64_perm *in, struct ipc_perm *out)
609 SET_UID(out->uid, in->uid);
610 SET_GID(out->gid, in->gid);
611 SET_UID(out->cuid, in->cuid);
612 SET_GID(out->cgid, in->cgid);
613 out->mode = in->mode;
619 * @ids: ipc identifier set
620 * @id: ipc id to look for
622 * Look for an id in the ipc ids idr and return associated ipc object.
624 * Call inside the RCU critical section.
625 * The ipc object is *not* locked on exit.
627 struct kern_ipc_perm *ipc_obtain_object(struct ipc_ids *ids, int id)
629 struct kern_ipc_perm *out;
630 int lid = ipcid_to_idx(id);
632 out = idr_find(&ids->ipcs_idr, lid);
634 return ERR_PTR(-EINVAL);
640 * ipc_lock - Lock an ipc structure without rwsem held
641 * @ids: IPC identifier set
642 * @id: ipc id to look for
644 * Look for an id in the ipc ids idr and lock the associated ipc object.
646 * The ipc object is locked on successful exit.
648 struct kern_ipc_perm *ipc_lock(struct ipc_ids *ids, int id)
650 struct kern_ipc_perm *out;
653 out = ipc_obtain_object(ids, id);
657 spin_lock(&out->lock);
659 /* ipc_rmid() may have already freed the ID while ipc_lock
660 * was spinning: here verify that the structure is still valid
665 spin_unlock(&out->lock);
666 out = ERR_PTR(-EINVAL);
673 * ipc_obtain_object_check
674 * @ids: ipc identifier set
675 * @id: ipc id to look for
677 * Similar to ipc_obtain_object() but also checks
678 * the ipc object reference counter.
680 * Call inside the RCU critical section.
681 * The ipc object is *not* locked on exit.
683 struct kern_ipc_perm *ipc_obtain_object_check(struct ipc_ids *ids, int id)
685 struct kern_ipc_perm *out = ipc_obtain_object(ids, id);
690 if (ipc_checkid(out, id))
691 return ERR_PTR(-EIDRM);
697 * ipcget - Common sys_*get() code
699 * @ids : IPC identifier set
700 * @ops : operations to be called on ipc object creation, permission checks
702 * @params : the parameters needed by the previous operations.
704 * Common routine called by sys_msgget(), sys_semget() and sys_shmget().
706 int ipcget(struct ipc_namespace *ns, struct ipc_ids *ids,
707 struct ipc_ops *ops, struct ipc_params *params)
709 if (params->key == IPC_PRIVATE)
710 return ipcget_new(ns, ids, ops, params);
712 return ipcget_public(ns, ids, ops, params);
716 * ipc_update_perm - update the permissions of an IPC.
717 * @in: the permission given as input.
718 * @out: the permission of the ipc to set.
720 int ipc_update_perm(struct ipc64_perm *in, struct kern_ipc_perm *out)
722 kuid_t uid = make_kuid(current_user_ns(), in->uid);
723 kgid_t gid = make_kgid(current_user_ns(), in->gid);
724 if (!uid_valid(uid) || !gid_valid(gid))
729 out->mode = (out->mode & ~S_IRWXUGO)
730 | (in->mode & S_IRWXUGO);
736 * ipcctl_pre_down_nolock - retrieve an ipc and check permissions for some IPC_XXX cmd
737 * @ns: the ipc namespace
738 * @ids: the table of ids where to look for the ipc
739 * @id: the id of the ipc to retrieve
740 * @cmd: the cmd to check
741 * @perm: the permission to set
742 * @extra_perm: one extra permission parameter used by msq
744 * This function does some common audit and permissions check for some IPC_XXX
745 * cmd and is called from semctl_down, shmctl_down and msgctl_down.
746 * It must be called without any lock held and
747 * - retrieves the ipc with the given id in the given table.
748 * - performs some audit and permission check, depending on the given cmd
749 * - returns a pointer to the ipc object or otherwise, the corresponding error.
751 * Call holding the both the rwsem and the rcu read lock.
753 struct kern_ipc_perm *ipcctl_pre_down_nolock(struct ipc_namespace *ns,
754 struct ipc_ids *ids, int id, int cmd,
755 struct ipc64_perm *perm, int extra_perm)
759 struct kern_ipc_perm *ipcp;
761 ipcp = ipc_obtain_object_check(ids, id);
769 audit_ipc_set_perm(extra_perm, perm->uid,
770 perm->gid, perm->mode);
772 euid = current_euid();
773 if (uid_eq(euid, ipcp->cuid) || uid_eq(euid, ipcp->uid) ||
774 ns_capable(ns->user_ns, CAP_SYS_ADMIN))
775 return ipcp; /* successful lookup */
780 #ifdef CONFIG_ARCH_WANT_IPC_PARSE_VERSION
784 * ipc_parse_version - IPC call version
785 * @cmd: pointer to command
787 * Return IPC_64 for new style IPC and IPC_OLD for old style IPC.
788 * The @cmd value is turned from an encoding command and version into
789 * just the command code.
792 int ipc_parse_version (int *cmd)
802 #endif /* CONFIG_ARCH_WANT_IPC_PARSE_VERSION */
804 #ifdef CONFIG_PROC_FS
805 struct ipc_proc_iter {
806 struct ipc_namespace *ns;
807 struct ipc_proc_iface *iface;
811 * This routine locks the ipc structure found at least at position pos.
813 static struct kern_ipc_perm *sysvipc_find_ipc(struct ipc_ids *ids, loff_t pos,
816 struct kern_ipc_perm *ipc;
820 for (id = 0; id < pos && total < ids->in_use; id++) {
821 ipc = idr_find(&ids->ipcs_idr, id);
826 if (total >= ids->in_use)
829 for ( ; pos < IPCMNI; pos++) {
830 ipc = idr_find(&ids->ipcs_idr, pos);
834 ipc_lock_object(ipc);
839 /* Out of range - return NULL to terminate iteration */
843 static void *sysvipc_proc_next(struct seq_file *s, void *it, loff_t *pos)
845 struct ipc_proc_iter *iter = s->private;
846 struct ipc_proc_iface *iface = iter->iface;
847 struct kern_ipc_perm *ipc = it;
849 /* If we had an ipc id locked before, unlock it */
850 if (ipc && ipc != SEQ_START_TOKEN)
853 return sysvipc_find_ipc(&iter->ns->ids[iface->ids], *pos, pos);
857 * File positions: pos 0 -> header, pos n -> ipc id = n - 1.
858 * SeqFile iterator: iterator value locked ipc pointer or SEQ_TOKEN_START.
860 static void *sysvipc_proc_start(struct seq_file *s, loff_t *pos)
862 struct ipc_proc_iter *iter = s->private;
863 struct ipc_proc_iface *iface = iter->iface;
866 ids = &iter->ns->ids[iface->ids];
869 * Take the lock - this will be released by the corresponding
872 down_read(&ids->rwsem);
874 /* pos < 0 is invalid */
878 /* pos == 0 means header */
880 return SEQ_START_TOKEN;
882 /* Find the (pos-1)th ipc */
883 return sysvipc_find_ipc(ids, *pos - 1, pos);
886 static void sysvipc_proc_stop(struct seq_file *s, void *it)
888 struct kern_ipc_perm *ipc = it;
889 struct ipc_proc_iter *iter = s->private;
890 struct ipc_proc_iface *iface = iter->iface;
893 /* If we had a locked structure, release it */
894 if (ipc && ipc != SEQ_START_TOKEN)
897 ids = &iter->ns->ids[iface->ids];
898 /* Release the lock we took in start() */
899 up_read(&ids->rwsem);
902 static int sysvipc_proc_show(struct seq_file *s, void *it)
904 struct ipc_proc_iter *iter = s->private;
905 struct ipc_proc_iface *iface = iter->iface;
907 if (it == SEQ_START_TOKEN)
908 return seq_puts(s, iface->header);
910 return iface->show(s, it);
913 static const struct seq_operations sysvipc_proc_seqops = {
914 .start = sysvipc_proc_start,
915 .stop = sysvipc_proc_stop,
916 .next = sysvipc_proc_next,
917 .show = sysvipc_proc_show,
920 static int sysvipc_proc_open(struct inode *inode, struct file *file)
923 struct seq_file *seq;
924 struct ipc_proc_iter *iter;
927 iter = kmalloc(sizeof(*iter), GFP_KERNEL);
931 ret = seq_open(file, &sysvipc_proc_seqops);
935 seq = file->private_data;
938 iter->iface = PDE_DATA(inode);
939 iter->ns = get_ipc_ns(current->nsproxy->ipc_ns);
947 static int sysvipc_proc_release(struct inode *inode, struct file *file)
949 struct seq_file *seq = file->private_data;
950 struct ipc_proc_iter *iter = seq->private;
951 put_ipc_ns(iter->ns);
952 return seq_release_private(inode, file);
955 static const struct file_operations sysvipc_proc_fops = {
956 .open = sysvipc_proc_open,
959 .release = sysvipc_proc_release,
961 #endif /* CONFIG_PROC_FS */