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:
20 * when doing ipc id lookups, take the ids->rwsem
22 * obtain the ipc object (kern_ipc_perm)
23 * perform security, capabilities, auditing and permission checks, etc.
24 * acquire the ipc lock (kern_ipc_perm.lock) throught ipc_lock_object()
25 * perform data updates (ie: SET, RMID, LOCK/UNLOCK commands)
29 #include <linux/shm.h>
30 #include <linux/init.h>
31 #include <linux/msg.h>
32 #include <linux/vmalloc.h>
33 #include <linux/slab.h>
34 #include <linux/notifier.h>
35 #include <linux/capability.h>
36 #include <linux/highuid.h>
37 #include <linux/security.h>
38 #include <linux/rcupdate.h>
39 #include <linux/workqueue.h>
40 #include <linux/seq_file.h>
41 #include <linux/proc_fs.h>
42 #include <linux/audit.h>
43 #include <linux/nsproxy.h>
44 #include <linux/rwsem.h>
45 #include <linux/memory.h>
46 #include <linux/ipc_namespace.h>
48 #include <asm/unistd.h>
52 struct ipc_proc_iface {
56 int (*show)(struct seq_file *, void *);
59 static void ipc_memory_notifier(struct work_struct *work)
61 ipcns_notify(IPCNS_MEMCHANGED);
64 static int ipc_memory_callback(struct notifier_block *self,
65 unsigned long action, void *arg)
67 static DECLARE_WORK(ipc_memory_wq, ipc_memory_notifier);
70 case MEM_ONLINE: /* memory successfully brought online */
71 case MEM_OFFLINE: /* or offline: it's time to recompute msgmni */
73 * This is done by invoking the ipcns notifier chain with the
74 * IPC_MEMCHANGED event.
75 * In order not to keep the lock on the hotplug memory chain
76 * for too long, queue a work item that will, when waken up,
77 * activate the ipcns notification chain.
78 * No need to keep several ipc work items on the queue.
80 if (!work_pending(&ipc_memory_wq))
81 schedule_work(&ipc_memory_wq);
83 case MEM_GOING_ONLINE:
84 case MEM_GOING_OFFLINE:
85 case MEM_CANCEL_ONLINE:
86 case MEM_CANCEL_OFFLINE:
94 static struct notifier_block ipc_memory_nb = {
95 .notifier_call = ipc_memory_callback,
96 .priority = IPC_CALLBACK_PRI,
100 * ipc_init - initialise IPC subsystem
102 * The various system5 IPC resources (semaphores, messages and shared
103 * memory) are initialised
104 * A callback routine is registered into the memory hotplug notifier
105 * chain: since msgmni scales to lowmem this callback routine will be
106 * called upon successful memory add / remove to recompute msmgni.
109 static int __init ipc_init(void)
114 register_hotmemory_notifier(&ipc_memory_nb);
115 register_ipcns_notifier(&init_ipc_ns);
118 __initcall(ipc_init);
121 * ipc_init_ids - initialise IPC identifiers
122 * @ids: Identifier set
124 * Set up the sequence range to use for the ipc identifier range (limited
125 * below IPCMNI) then initialise the ids idr.
128 void ipc_init_ids(struct ipc_ids *ids)
130 init_rwsem(&ids->rwsem);
136 int seq_limit = INT_MAX/SEQ_MULTIPLIER;
137 if (seq_limit > USHRT_MAX)
138 ids->seq_max = USHRT_MAX;
140 ids->seq_max = seq_limit;
143 idr_init(&ids->ipcs_idr);
146 #ifdef CONFIG_PROC_FS
147 static const struct file_operations sysvipc_proc_fops;
149 * ipc_init_proc_interface - Create a proc interface for sysipc types using a seq_file interface.
150 * @path: Path in procfs
151 * @header: Banner to be printed at the beginning of the file.
152 * @ids: ipc id table to iterate.
153 * @show: show routine.
155 void __init ipc_init_proc_interface(const char *path, const char *header,
156 int ids, int (*show)(struct seq_file *, void *))
158 struct proc_dir_entry *pde;
159 struct ipc_proc_iface *iface;
161 iface = kmalloc(sizeof(*iface), GFP_KERNEL);
165 iface->header = header;
169 pde = proc_create_data(path,
170 S_IRUGO, /* world readable */
171 NULL, /* parent dir */
181 * ipc_findkey - find a key in an ipc identifier set
182 * @ids: Identifier set
183 * @key: The key to find
185 * Requires ipc_ids.rwsem locked.
186 * Returns the LOCKED pointer to the ipc structure if found or NULL
188 * If key is found ipc points to the owning ipc structure
191 static struct kern_ipc_perm *ipc_findkey(struct ipc_ids *ids, key_t key)
193 struct kern_ipc_perm *ipc;
197 for (total = 0, next_id = 0; total < ids->in_use; next_id++) {
198 ipc = idr_find(&ids->ipcs_idr, next_id);
203 if (ipc->key != key) {
208 ipc_lock_by_ptr(ipc);
216 * ipc_get_maxid - get the last assigned id
217 * @ids: IPC identifier set
219 * Called with ipc_ids.rwsem held.
222 int ipc_get_maxid(struct ipc_ids *ids)
224 struct kern_ipc_perm *ipc;
228 if (ids->in_use == 0)
231 if (ids->in_use == IPCMNI)
234 /* Look for the last assigned id */
236 for (id = 0; id < IPCMNI && total < ids->in_use; id++) {
237 ipc = idr_find(&ids->ipcs_idr, id);
247 * ipc_addid - add an IPC identifier
248 * @ids: IPC identifier set
249 * @new: new IPC permission set
250 * @size: limit for the number of used ids
252 * Add an entry 'new' to the IPC ids idr. The permissions object is
253 * initialised and the first free entry is set up and the id assigned
254 * is returned. The 'new' entry is returned in a locked state on success.
255 * On failure the entry is not locked and a negative err-code is returned.
257 * Called with writer ipc_ids.rwsem held.
259 int ipc_addid(struct ipc_ids* ids, struct kern_ipc_perm* new, int size)
264 int next_id = ids->next_id;
269 if (ids->in_use >= size)
272 idr_preload(GFP_KERNEL);
274 spin_lock_init(&new->lock);
277 spin_lock(&new->lock);
279 id = idr_alloc(&ids->ipcs_idr, new,
280 (next_id < 0) ? 0 : ipcid_to_idx(next_id), 0,
284 spin_unlock(&new->lock);
291 current_euid_egid(&euid, &egid);
292 new->cuid = new->uid = euid;
293 new->gid = new->cgid = egid;
296 new->seq = ids->seq++;
297 if (ids->seq > ids->seq_max)
300 new->seq = ipcid_to_seqx(next_id);
304 new->id = ipc_buildid(id, new->seq);
309 * ipcget_new - create a new ipc object
311 * @ids: IPC identifer set
312 * @ops: the actual creation routine to call
313 * @params: its parameters
315 * This routine is called by sys_msgget, sys_semget() and sys_shmget()
316 * when the key is IPC_PRIVATE.
318 static int ipcget_new(struct ipc_namespace *ns, struct ipc_ids *ids,
319 struct ipc_ops *ops, struct ipc_params *params)
323 down_write(&ids->rwsem);
324 err = ops->getnew(ns, params);
325 up_write(&ids->rwsem);
330 * ipc_check_perms - check security and permissions for an IPC
332 * @ipcp: ipc permission set
333 * @ops: the actual security routine to call
334 * @params: its parameters
336 * This routine is called by sys_msgget(), sys_semget() and sys_shmget()
337 * when the key is not IPC_PRIVATE and that key already exists in the
340 * On success, the IPC id is returned.
342 * It is called with ipc_ids.rwsem and ipcp->lock held.
344 static int ipc_check_perms(struct ipc_namespace *ns,
345 struct kern_ipc_perm *ipcp,
347 struct ipc_params *params)
351 if (ipcperms(ns, ipcp, params->flg))
354 err = ops->associate(ipcp, params->flg);
363 * ipcget_public - get an ipc object or create a new one
365 * @ids: IPC identifer set
366 * @ops: the actual creation routine to call
367 * @params: its parameters
369 * This routine is called by sys_msgget, sys_semget() and sys_shmget()
370 * when the key is not IPC_PRIVATE.
371 * It adds a new entry if the key is not found and does some permission
372 * / security checkings if the key is found.
374 * On success, the ipc id is returned.
376 static int ipcget_public(struct ipc_namespace *ns, struct ipc_ids *ids,
377 struct ipc_ops *ops, struct ipc_params *params)
379 struct kern_ipc_perm *ipcp;
380 int flg = params->flg;
384 * Take the lock as a writer since we are potentially going to add
385 * a new entry + read locks are not "upgradable"
387 down_write(&ids->rwsem);
388 ipcp = ipc_findkey(ids, params->key);
391 if (!(flg & IPC_CREAT))
394 err = ops->getnew(ns, params);
396 /* ipc object has been locked by ipc_findkey() */
398 if (flg & IPC_CREAT && flg & IPC_EXCL)
402 if (ops->more_checks)
403 err = ops->more_checks(ipcp, params);
406 * ipc_check_perms returns the IPC id on
409 err = ipc_check_perms(ns, ipcp, ops, params);
413 up_write(&ids->rwsem);
420 * ipc_rmid - remove an IPC identifier
421 * @ids: IPC identifier set
422 * @ipcp: ipc perm structure containing the identifier to remove
424 * ipc_ids.rwsem (as a writer) and the spinlock for this ID are held
425 * 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);
442 * ipc_alloc - allocate ipc space
443 * @size: size desired
445 * Allocate memory from the appropriate pools and return a pointer to it.
446 * NULL is returned if the allocation fails
449 void *ipc_alloc(int size)
455 out = kmalloc(size, GFP_KERNEL);
460 * ipc_free - free ipc space
461 * @ptr: pointer returned by ipc_alloc
462 * @size: size of block
464 * Free a block created with ipc_alloc(). The caller must know the size
465 * used in the allocation call.
468 void ipc_free(void* ptr, int size)
479 } ____cacheline_aligned_in_smp;
482 * ipc_rcu_alloc - allocate ipc and rcu space
483 * @size: size desired
485 * Allocate memory for the rcu header structure + the object.
486 * Returns the pointer to the object or NULL upon failure.
488 void *ipc_rcu_alloc(int size)
491 * We prepend the allocation with the rcu struct
493 struct ipc_rcu *out = ipc_alloc(sizeof(struct ipc_rcu) + size);
496 atomic_set(&out->refcount, 1);
500 int ipc_rcu_getref(void *ptr)
502 struct ipc_rcu *p = ((struct ipc_rcu *)ptr) - 1;
504 return atomic_inc_not_zero(&p->refcount);
508 * ipc_schedule_free - free ipc + rcu space
509 * @head: RCU callback structure for queued work
511 static void ipc_schedule_free(struct rcu_head *head)
513 vfree(container_of(head, struct ipc_rcu, rcu));
516 void ipc_rcu_putref(void *ptr)
518 struct ipc_rcu *p = ((struct ipc_rcu *)ptr) - 1;
520 if (!atomic_dec_and_test(&p->refcount))
523 if (is_vmalloc_addr(ptr)) {
524 call_rcu(&p->rcu, ipc_schedule_free);
531 * ipcperms - check IPC permissions
533 * @ipcp: IPC permission set
534 * @flag: desired permission set.
536 * Check user, group, other permissions for access
537 * to ipc resources. return 0 if allowed
539 * @flag will most probably be 0 or S_...UGO from <linux/stat.h>
542 int ipcperms(struct ipc_namespace *ns, struct kern_ipc_perm *ipcp, short flag)
544 kuid_t euid = current_euid();
545 int requested_mode, granted_mode;
548 requested_mode = (flag >> 6) | (flag >> 3) | flag;
549 granted_mode = ipcp->mode;
550 if (uid_eq(euid, ipcp->cuid) ||
551 uid_eq(euid, ipcp->uid))
553 else if (in_group_p(ipcp->cgid) || in_group_p(ipcp->gid))
555 /* is there some bit set in requested_mode but not in granted_mode? */
556 if ((requested_mode & ~granted_mode & 0007) &&
557 !ns_capable(ns->user_ns, CAP_IPC_OWNER))
560 return security_ipc_permission(ipcp, flag);
564 * Functions to convert between the kern_ipc_perm structure and the
565 * old/new ipc_perm structures
569 * kernel_to_ipc64_perm - convert kernel ipc permissions to user
570 * @in: kernel permissions
571 * @out: new style IPC permissions
573 * Turn the kernel object @in into a set of permissions descriptions
574 * for returning to userspace (@out).
578 void kernel_to_ipc64_perm (struct kern_ipc_perm *in, struct ipc64_perm *out)
581 out->uid = from_kuid_munged(current_user_ns(), in->uid);
582 out->gid = from_kgid_munged(current_user_ns(), in->gid);
583 out->cuid = from_kuid_munged(current_user_ns(), in->cuid);
584 out->cgid = from_kgid_munged(current_user_ns(), in->cgid);
585 out->mode = in->mode;
590 * ipc64_perm_to_ipc_perm - convert new ipc permissions to old
591 * @in: new style IPC permissions
592 * @out: old style IPC permissions
594 * Turn the new style permissions object @in into a compatibility
595 * object and store it into the @out pointer.
598 void ipc64_perm_to_ipc_perm (struct ipc64_perm *in, struct ipc_perm *out)
601 SET_UID(out->uid, in->uid);
602 SET_GID(out->gid, in->gid);
603 SET_UID(out->cuid, in->cuid);
604 SET_GID(out->cgid, in->cgid);
605 out->mode = in->mode;
611 * @ids: ipc identifier set
612 * @id: ipc id to look for
614 * Look for an id in the ipc ids idr and return associated ipc object.
616 * Call inside the RCU critical section.
617 * The ipc object is *not* locked on exit.
619 struct kern_ipc_perm *ipc_obtain_object(struct ipc_ids *ids, int id)
621 struct kern_ipc_perm *out;
622 int lid = ipcid_to_idx(id);
624 out = idr_find(&ids->ipcs_idr, lid);
626 return ERR_PTR(-EINVAL);
632 * ipc_lock - Lock an ipc structure without rwsem held
633 * @ids: IPC identifier set
634 * @id: ipc id to look for
636 * Look for an id in the ipc ids idr and lock the associated ipc object.
638 * The ipc object is locked on successful exit.
640 struct kern_ipc_perm *ipc_lock(struct ipc_ids *ids, int id)
642 struct kern_ipc_perm *out;
645 out = ipc_obtain_object(ids, id);
649 spin_lock(&out->lock);
651 /* ipc_rmid() may have already freed the ID while ipc_lock
652 * was spinning: here verify that the structure is still valid
657 spin_unlock(&out->lock);
658 out = ERR_PTR(-EINVAL);
665 * ipc_obtain_object_check
666 * @ids: ipc identifier set
667 * @id: ipc id to look for
669 * Similar to ipc_obtain_object() but also checks
670 * the ipc object reference counter.
672 * Call inside the RCU critical section.
673 * The ipc object is *not* locked on exit.
675 struct kern_ipc_perm *ipc_obtain_object_check(struct ipc_ids *ids, int id)
677 struct kern_ipc_perm *out = ipc_obtain_object(ids, id);
682 if (ipc_checkid(out, id))
683 return ERR_PTR(-EIDRM);
688 struct kern_ipc_perm *ipc_lock_check(struct ipc_ids *ids, int id)
690 struct kern_ipc_perm *out;
692 out = ipc_lock(ids, id);
696 if (ipc_checkid(out, id)) {
698 return ERR_PTR(-EIDRM);
705 * ipcget - Common sys_*get() code
707 * @ids : IPC identifier set
708 * @ops : operations to be called on ipc object creation, permission checks
710 * @params : the parameters needed by the previous operations.
712 * Common routine called by sys_msgget(), sys_semget() and sys_shmget().
714 int ipcget(struct ipc_namespace *ns, struct ipc_ids *ids,
715 struct ipc_ops *ops, struct ipc_params *params)
717 if (params->key == IPC_PRIVATE)
718 return ipcget_new(ns, ids, ops, params);
720 return ipcget_public(ns, ids, ops, params);
724 * ipc_update_perm - update the permissions of an IPC.
725 * @in: the permission given as input.
726 * @out: the permission of the ipc to set.
728 int ipc_update_perm(struct ipc64_perm *in, struct kern_ipc_perm *out)
730 kuid_t uid = make_kuid(current_user_ns(), in->uid);
731 kgid_t gid = make_kgid(current_user_ns(), in->gid);
732 if (!uid_valid(uid) || !gid_valid(gid))
737 out->mode = (out->mode & ~S_IRWXUGO)
738 | (in->mode & S_IRWXUGO);
744 * ipcctl_pre_down_nolock - retrieve an ipc and check permissions for some IPC_XXX cmd
745 * @ns: the ipc namespace
746 * @ids: the table of ids where to look for the ipc
747 * @id: the id of the ipc to retrieve
748 * @cmd: the cmd to check
749 * @perm: the permission to set
750 * @extra_perm: one extra permission parameter used by msq
752 * This function does some common audit and permissions check for some IPC_XXX
753 * cmd and is called from semctl_down, shmctl_down and msgctl_down.
754 * It must be called without any lock held and
755 * - retrieves the ipc with the given id in the given table.
756 * - performs some audit and permission check, depending on the given cmd
757 * - returns a pointer to the ipc object or otherwise, the corresponding error.
759 * Call holding the both the rwsem and the rcu read lock.
761 struct kern_ipc_perm *ipcctl_pre_down_nolock(struct ipc_namespace *ns,
762 struct ipc_ids *ids, int id, int cmd,
763 struct ipc64_perm *perm, int extra_perm)
767 struct kern_ipc_perm *ipcp;
769 ipcp = ipc_obtain_object_check(ids, id);
777 audit_ipc_set_perm(extra_perm, perm->uid,
778 perm->gid, perm->mode);
780 euid = current_euid();
781 if (uid_eq(euid, ipcp->cuid) || uid_eq(euid, ipcp->uid) ||
782 ns_capable(ns->user_ns, CAP_SYS_ADMIN))
783 return ipcp; /* successful lookup */
788 #ifdef CONFIG_ARCH_WANT_IPC_PARSE_VERSION
792 * ipc_parse_version - IPC call version
793 * @cmd: pointer to command
795 * Return IPC_64 for new style IPC and IPC_OLD for old style IPC.
796 * The @cmd value is turned from an encoding command and version into
797 * just the command code.
800 int ipc_parse_version (int *cmd)
810 #endif /* CONFIG_ARCH_WANT_IPC_PARSE_VERSION */
812 #ifdef CONFIG_PROC_FS
813 struct ipc_proc_iter {
814 struct ipc_namespace *ns;
815 struct ipc_proc_iface *iface;
819 * This routine locks the ipc structure found at least at position pos.
821 static struct kern_ipc_perm *sysvipc_find_ipc(struct ipc_ids *ids, loff_t pos,
824 struct kern_ipc_perm *ipc;
828 for (id = 0; id < pos && total < ids->in_use; id++) {
829 ipc = idr_find(&ids->ipcs_idr, id);
834 if (total >= ids->in_use)
837 for ( ; pos < IPCMNI; pos++) {
838 ipc = idr_find(&ids->ipcs_idr, pos);
841 ipc_lock_by_ptr(ipc);
846 /* Out of range - return NULL to terminate iteration */
850 static void *sysvipc_proc_next(struct seq_file *s, void *it, loff_t *pos)
852 struct ipc_proc_iter *iter = s->private;
853 struct ipc_proc_iface *iface = iter->iface;
854 struct kern_ipc_perm *ipc = it;
856 /* If we had an ipc id locked before, unlock it */
857 if (ipc && ipc != SEQ_START_TOKEN)
860 return sysvipc_find_ipc(&iter->ns->ids[iface->ids], *pos, pos);
864 * File positions: pos 0 -> header, pos n -> ipc id = n - 1.
865 * SeqFile iterator: iterator value locked ipc pointer or SEQ_TOKEN_START.
867 static void *sysvipc_proc_start(struct seq_file *s, loff_t *pos)
869 struct ipc_proc_iter *iter = s->private;
870 struct ipc_proc_iface *iface = iter->iface;
873 ids = &iter->ns->ids[iface->ids];
876 * Take the lock - this will be released by the corresponding
879 down_read(&ids->rwsem);
881 /* pos < 0 is invalid */
885 /* pos == 0 means header */
887 return SEQ_START_TOKEN;
889 /* Find the (pos-1)th ipc */
890 return sysvipc_find_ipc(ids, *pos - 1, pos);
893 static void sysvipc_proc_stop(struct seq_file *s, void *it)
895 struct kern_ipc_perm *ipc = it;
896 struct ipc_proc_iter *iter = s->private;
897 struct ipc_proc_iface *iface = iter->iface;
900 /* If we had a locked structure, release it */
901 if (ipc && ipc != SEQ_START_TOKEN)
904 ids = &iter->ns->ids[iface->ids];
905 /* Release the lock we took in start() */
906 up_read(&ids->rwsem);
909 static int sysvipc_proc_show(struct seq_file *s, void *it)
911 struct ipc_proc_iter *iter = s->private;
912 struct ipc_proc_iface *iface = iter->iface;
914 if (it == SEQ_START_TOKEN)
915 return seq_puts(s, iface->header);
917 return iface->show(s, it);
920 static const struct seq_operations sysvipc_proc_seqops = {
921 .start = sysvipc_proc_start,
922 .stop = sysvipc_proc_stop,
923 .next = sysvipc_proc_next,
924 .show = sysvipc_proc_show,
927 static int sysvipc_proc_open(struct inode *inode, struct file *file)
930 struct seq_file *seq;
931 struct ipc_proc_iter *iter;
934 iter = kmalloc(sizeof(*iter), GFP_KERNEL);
938 ret = seq_open(file, &sysvipc_proc_seqops);
942 seq = file->private_data;
945 iter->iface = PDE_DATA(inode);
946 iter->ns = get_ipc_ns(current->nsproxy->ipc_ns);
954 static int sysvipc_proc_release(struct inode *inode, struct file *file)
956 struct seq_file *seq = file->private_data;
957 struct ipc_proc_iter *iter = seq->private;
958 put_ipc_ns(iter->ns);
959 return seq_release_private(inode, file);
962 static const struct file_operations sysvipc_proc_fops = {
963 .open = sysvipc_proc_open,
966 .release = sysvipc_proc_release,
968 #endif /* CONFIG_PROC_FS */