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>
21 #include <linux/shm.h>
22 #include <linux/init.h>
23 #include <linux/msg.h>
24 #include <linux/vmalloc.h>
25 #include <linux/slab.h>
26 #include <linux/notifier.h>
27 #include <linux/capability.h>
28 #include <linux/highuid.h>
29 #include <linux/security.h>
30 #include <linux/rcupdate.h>
31 #include <linux/workqueue.h>
32 #include <linux/seq_file.h>
33 #include <linux/proc_fs.h>
34 #include <linux/audit.h>
35 #include <linux/nsproxy.h>
36 #include <linux/rwsem.h>
37 #include <linux/memory.h>
38 #include <linux/ipc_namespace.h>
40 #include <asm/unistd.h>
44 struct ipc_proc_iface {
48 int (*show)(struct seq_file *, void *);
51 static void ipc_memory_notifier(struct work_struct *work)
53 ipcns_notify(IPCNS_MEMCHANGED);
56 static int ipc_memory_callback(struct notifier_block *self,
57 unsigned long action, void *arg)
59 static DECLARE_WORK(ipc_memory_wq, ipc_memory_notifier);
62 case MEM_ONLINE: /* memory successfully brought online */
63 case MEM_OFFLINE: /* or offline: it's time to recompute msgmni */
65 * This is done by invoking the ipcns notifier chain with the
66 * IPC_MEMCHANGED event.
67 * In order not to keep the lock on the hotplug memory chain
68 * for too long, queue a work item that will, when waken up,
69 * activate the ipcns notification chain.
70 * No need to keep several ipc work items on the queue.
72 if (!work_pending(&ipc_memory_wq))
73 schedule_work(&ipc_memory_wq);
75 case MEM_GOING_ONLINE:
76 case MEM_GOING_OFFLINE:
77 case MEM_CANCEL_ONLINE:
78 case MEM_CANCEL_OFFLINE:
86 static struct notifier_block ipc_memory_nb = {
87 .notifier_call = ipc_memory_callback,
88 .priority = IPC_CALLBACK_PRI,
92 * ipc_init - initialise IPC subsystem
94 * The various system5 IPC resources (semaphores, messages and shared
95 * memory) are initialised
96 * A callback routine is registered into the memory hotplug notifier
97 * chain: since msgmni scales to lowmem this callback routine will be
98 * called upon successful memory add / remove to recompute msmgni.
101 static int __init ipc_init(void)
106 register_hotmemory_notifier(&ipc_memory_nb);
107 register_ipcns_notifier(&init_ipc_ns);
110 __initcall(ipc_init);
113 * ipc_init_ids - initialise IPC identifiers
114 * @ids: Identifier set
116 * Set up the sequence range to use for the ipc identifier range (limited
117 * below IPCMNI) then initialise the ids idr.
120 void ipc_init_ids(struct ipc_ids *ids)
122 init_rwsem(&ids->rw_mutex);
128 int seq_limit = INT_MAX/SEQ_MULTIPLIER;
129 if (seq_limit > USHRT_MAX)
130 ids->seq_max = USHRT_MAX;
132 ids->seq_max = seq_limit;
135 idr_init(&ids->ipcs_idr);
138 #ifdef CONFIG_PROC_FS
139 static const struct file_operations sysvipc_proc_fops;
141 * ipc_init_proc_interface - Create a proc interface for sysipc types using a seq_file interface.
142 * @path: Path in procfs
143 * @header: Banner to be printed at the beginning of the file.
144 * @ids: ipc id table to iterate.
145 * @show: show routine.
147 void __init ipc_init_proc_interface(const char *path, const char *header,
148 int ids, int (*show)(struct seq_file *, void *))
150 struct proc_dir_entry *pde;
151 struct ipc_proc_iface *iface;
153 iface = kmalloc(sizeof(*iface), GFP_KERNEL);
157 iface->header = header;
161 pde = proc_create_data(path,
162 S_IRUGO, /* world readable */
163 NULL, /* parent dir */
173 * ipc_findkey - find a key in an ipc identifier set
174 * @ids: Identifier set
175 * @key: The key to find
177 * Requires ipc_ids.rw_mutex locked.
178 * Returns the LOCKED pointer to the ipc structure if found or NULL
180 * If key is found ipc points to the owning ipc structure
183 static struct kern_ipc_perm *ipc_findkey(struct ipc_ids *ids, key_t key)
185 struct kern_ipc_perm *ipc;
189 for (total = 0, next_id = 0; total < ids->in_use; next_id++) {
190 ipc = idr_find(&ids->ipcs_idr, next_id);
195 if (ipc->key != key) {
200 ipc_lock_by_ptr(ipc);
208 * ipc_get_maxid - get the last assigned id
209 * @ids: IPC identifier set
211 * Called with ipc_ids.rw_mutex held.
214 int ipc_get_maxid(struct ipc_ids *ids)
216 struct kern_ipc_perm *ipc;
220 if (ids->in_use == 0)
223 if (ids->in_use == IPCMNI)
226 /* Look for the last assigned id */
228 for (id = 0; id < IPCMNI && total < ids->in_use; id++) {
229 ipc = idr_find(&ids->ipcs_idr, id);
239 * ipc_addid - add an IPC identifier
240 * @ids: IPC identifier set
241 * @new: new IPC permission set
242 * @size: limit for the number of used ids
244 * Add an entry 'new' to the IPC ids idr. The permissions object is
245 * initialised and the first free entry is set up and the id assigned
246 * is returned. The 'new' entry is returned in a locked state on success.
247 * On failure the entry is not locked and a negative err-code is returned.
249 * Called with writer ipc_ids.rw_mutex held.
251 int ipc_addid(struct ipc_ids* ids, struct kern_ipc_perm* new, int size)
256 int next_id = ids->next_id;
261 if (ids->in_use >= size)
264 idr_preload(GFP_KERNEL);
266 spin_lock_init(&new->lock);
269 spin_lock(&new->lock);
271 id = idr_alloc(&ids->ipcs_idr, new,
272 (next_id < 0) ? 0 : ipcid_to_idx(next_id), 0,
276 spin_unlock(&new->lock);
283 current_euid_egid(&euid, &egid);
284 new->cuid = new->uid = euid;
285 new->gid = new->cgid = egid;
288 new->seq = ids->seq++;
289 if (ids->seq > ids->seq_max)
292 new->seq = ipcid_to_seqx(next_id);
296 new->id = ipc_buildid(id, new->seq);
301 * ipcget_new - create a new ipc object
303 * @ids: IPC identifer set
304 * @ops: the actual creation routine to call
305 * @params: its parameters
307 * This routine is called by sys_msgget, sys_semget() and sys_shmget()
308 * when the key is IPC_PRIVATE.
310 static int ipcget_new(struct ipc_namespace *ns, struct ipc_ids *ids,
311 struct ipc_ops *ops, struct ipc_params *params)
315 down_write(&ids->rw_mutex);
316 err = ops->getnew(ns, params);
317 up_write(&ids->rw_mutex);
322 * ipc_check_perms - check security and permissions for an IPC
324 * @ipcp: ipc permission set
325 * @ops: the actual security routine to call
326 * @params: its parameters
328 * This routine is called by sys_msgget(), sys_semget() and sys_shmget()
329 * when the key is not IPC_PRIVATE and that key already exists in the
332 * On success, the IPC id is returned.
334 * It is called with ipc_ids.rw_mutex and ipcp->lock held.
336 static int ipc_check_perms(struct ipc_namespace *ns,
337 struct kern_ipc_perm *ipcp,
339 struct ipc_params *params)
343 if (ipcperms(ns, ipcp, params->flg))
346 err = ops->associate(ipcp, params->flg);
355 * ipcget_public - get an ipc object or create a new one
357 * @ids: IPC identifer set
358 * @ops: the actual creation routine to call
359 * @params: its parameters
361 * This routine is called by sys_msgget, sys_semget() and sys_shmget()
362 * when the key is not IPC_PRIVATE.
363 * It adds a new entry if the key is not found and does some permission
364 * / security checkings if the key is found.
366 * On success, the ipc id is returned.
368 static int ipcget_public(struct ipc_namespace *ns, struct ipc_ids *ids,
369 struct ipc_ops *ops, struct ipc_params *params)
371 struct kern_ipc_perm *ipcp;
372 int flg = params->flg;
376 * Take the lock as a writer since we are potentially going to add
377 * a new entry + read locks are not "upgradable"
379 down_write(&ids->rw_mutex);
380 ipcp = ipc_findkey(ids, params->key);
383 if (!(flg & IPC_CREAT))
386 err = ops->getnew(ns, params);
388 /* ipc object has been locked by ipc_findkey() */
390 if (flg & IPC_CREAT && flg & IPC_EXCL)
394 if (ops->more_checks)
395 err = ops->more_checks(ipcp, params);
398 * ipc_check_perms returns the IPC id on
401 err = ipc_check_perms(ns, ipcp, ops, params);
405 up_write(&ids->rw_mutex);
412 * ipc_rmid - remove an IPC identifier
413 * @ids: IPC identifier set
414 * @ipcp: ipc perm structure containing the identifier to remove
416 * ipc_ids.rw_mutex (as a writer) and the spinlock for this ID are held
417 * before this function is called, and remain locked on the exit.
420 void ipc_rmid(struct ipc_ids *ids, struct kern_ipc_perm *ipcp)
422 int lid = ipcid_to_idx(ipcp->id);
424 idr_remove(&ids->ipcs_idr, lid);
434 * ipc_alloc - allocate ipc space
435 * @size: size desired
437 * Allocate memory from the appropriate pools and return a pointer to it.
438 * NULL is returned if the allocation fails
441 void *ipc_alloc(int size)
447 out = kmalloc(size, GFP_KERNEL);
452 * ipc_free - free ipc space
453 * @ptr: pointer returned by ipc_alloc
454 * @size: size of block
456 * Free a block created with ipc_alloc(). The caller must know the size
457 * used in the allocation call.
460 void ipc_free(void* ptr, int size)
471 } ____cacheline_aligned_in_smp;
474 * ipc_rcu_alloc - allocate ipc and rcu space
475 * @size: size desired
477 * Allocate memory for the rcu header structure + the object.
478 * Returns the pointer to the object or NULL upon failure.
480 void *ipc_rcu_alloc(int size)
483 * We prepend the allocation with the rcu struct
485 struct ipc_rcu *out = ipc_alloc(sizeof(struct ipc_rcu) + size);
488 atomic_set(&out->refcount, 1);
492 int ipc_rcu_getref(void *ptr)
494 struct ipc_rcu *p = ((struct ipc_rcu *)ptr) - 1;
496 return atomic_inc_not_zero(&p->refcount);
500 * ipc_schedule_free - free ipc + rcu space
501 * @head: RCU callback structure for queued work
503 static void ipc_schedule_free(struct rcu_head *head)
505 vfree(container_of(head, struct ipc_rcu, rcu));
508 void ipc_rcu_putref(void *ptr)
510 struct ipc_rcu *p = ((struct ipc_rcu *)ptr) - 1;
512 if (!atomic_dec_and_test(&p->refcount))
515 if (is_vmalloc_addr(ptr)) {
516 call_rcu(&p->rcu, ipc_schedule_free);
523 * ipcperms - check IPC permissions
525 * @ipcp: IPC permission set
526 * @flag: desired permission set.
528 * Check user, group, other permissions for access
529 * to ipc resources. return 0 if allowed
531 * @flag will most probably be 0 or S_...UGO from <linux/stat.h>
534 int ipcperms(struct ipc_namespace *ns, struct kern_ipc_perm *ipcp, short flag)
536 kuid_t euid = current_euid();
537 int requested_mode, granted_mode;
540 requested_mode = (flag >> 6) | (flag >> 3) | flag;
541 granted_mode = ipcp->mode;
542 if (uid_eq(euid, ipcp->cuid) ||
543 uid_eq(euid, ipcp->uid))
545 else if (in_group_p(ipcp->cgid) || in_group_p(ipcp->gid))
547 /* is there some bit set in requested_mode but not in granted_mode? */
548 if ((requested_mode & ~granted_mode & 0007) &&
549 !ns_capable(ns->user_ns, CAP_IPC_OWNER))
552 return security_ipc_permission(ipcp, flag);
556 * Functions to convert between the kern_ipc_perm structure and the
557 * old/new ipc_perm structures
561 * kernel_to_ipc64_perm - convert kernel ipc permissions to user
562 * @in: kernel permissions
563 * @out: new style IPC permissions
565 * Turn the kernel object @in into a set of permissions descriptions
566 * for returning to userspace (@out).
570 void kernel_to_ipc64_perm (struct kern_ipc_perm *in, struct ipc64_perm *out)
573 out->uid = from_kuid_munged(current_user_ns(), in->uid);
574 out->gid = from_kgid_munged(current_user_ns(), in->gid);
575 out->cuid = from_kuid_munged(current_user_ns(), in->cuid);
576 out->cgid = from_kgid_munged(current_user_ns(), in->cgid);
577 out->mode = in->mode;
582 * ipc64_perm_to_ipc_perm - convert new ipc permissions to old
583 * @in: new style IPC permissions
584 * @out: old style IPC permissions
586 * Turn the new style permissions object @in into a compatibility
587 * object and store it into the @out pointer.
590 void ipc64_perm_to_ipc_perm (struct ipc64_perm *in, struct ipc_perm *out)
593 SET_UID(out->uid, in->uid);
594 SET_GID(out->gid, in->gid);
595 SET_UID(out->cuid, in->cuid);
596 SET_GID(out->cgid, in->cgid);
597 out->mode = in->mode;
603 * @ids: ipc identifier set
604 * @id: ipc id to look for
606 * Look for an id in the ipc ids idr and return associated ipc object.
608 * Call inside the RCU critical section.
609 * The ipc object is *not* locked on exit.
611 struct kern_ipc_perm *ipc_obtain_object(struct ipc_ids *ids, int id)
613 struct kern_ipc_perm *out;
614 int lid = ipcid_to_idx(id);
616 out = idr_find(&ids->ipcs_idr, lid);
618 return ERR_PTR(-EINVAL);
624 * ipc_lock - Lock an ipc structure without rw_mutex held
625 * @ids: IPC identifier set
626 * @id: ipc id to look for
628 * Look for an id in the ipc ids idr and lock the associated ipc object.
630 * The ipc object is locked on successful exit.
632 struct kern_ipc_perm *ipc_lock(struct ipc_ids *ids, int id)
634 struct kern_ipc_perm *out;
637 out = ipc_obtain_object(ids, id);
641 spin_lock(&out->lock);
643 /* ipc_rmid() may have already freed the ID while ipc_lock
644 * was spinning: here verify that the structure is still valid
649 spin_unlock(&out->lock);
650 out = ERR_PTR(-EINVAL);
657 * ipc_obtain_object_check
658 * @ids: ipc identifier set
659 * @id: ipc id to look for
661 * Similar to ipc_obtain_object() but also checks
662 * the ipc object reference counter.
664 * Call inside the RCU critical section.
665 * The ipc object is *not* locked on exit.
667 struct kern_ipc_perm *ipc_obtain_object_check(struct ipc_ids *ids, int id)
669 struct kern_ipc_perm *out = ipc_obtain_object(ids, id);
674 if (ipc_checkid(out, id))
675 return ERR_PTR(-EIDRM);
680 struct kern_ipc_perm *ipc_lock_check(struct ipc_ids *ids, int id)
682 struct kern_ipc_perm *out;
684 out = ipc_lock(ids, id);
688 if (ipc_checkid(out, id)) {
690 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 - 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 the ipc with the ipc lock held in case of success
750 * or an err-code without any lock held otherwise.
752 * Call holding the both the rw_mutex and the rcu read lock.
754 struct kern_ipc_perm *ipcctl_pre_down(struct ipc_namespace *ns,
755 struct ipc_ids *ids, int id, int cmd,
756 struct ipc64_perm *perm, int extra_perm)
758 struct kern_ipc_perm *ipcp;
760 ipcp = ipcctl_pre_down_nolock(ns, ids, id, cmd, perm, extra_perm);
764 spin_lock(&ipcp->lock);
769 struct kern_ipc_perm *ipcctl_pre_down_nolock(struct ipc_namespace *ns,
770 struct ipc_ids *ids, int id, int cmd,
771 struct ipc64_perm *perm, int extra_perm)
775 struct kern_ipc_perm *ipcp;
777 ipcp = ipc_obtain_object_check(ids, id);
785 audit_ipc_set_perm(extra_perm, perm->uid,
786 perm->gid, perm->mode);
788 euid = current_euid();
789 if (uid_eq(euid, ipcp->cuid) || uid_eq(euid, ipcp->uid) ||
790 ns_capable(ns->user_ns, CAP_SYS_ADMIN))
791 return ipcp; /* successful lookup */
796 #ifdef CONFIG_ARCH_WANT_IPC_PARSE_VERSION
800 * ipc_parse_version - IPC call version
801 * @cmd: pointer to command
803 * Return IPC_64 for new style IPC and IPC_OLD for old style IPC.
804 * The @cmd value is turned from an encoding command and version into
805 * just the command code.
808 int ipc_parse_version (int *cmd)
818 #endif /* CONFIG_ARCH_WANT_IPC_PARSE_VERSION */
820 #ifdef CONFIG_PROC_FS
821 struct ipc_proc_iter {
822 struct ipc_namespace *ns;
823 struct ipc_proc_iface *iface;
827 * This routine locks the ipc structure found at least at position pos.
829 static struct kern_ipc_perm *sysvipc_find_ipc(struct ipc_ids *ids, loff_t pos,
832 struct kern_ipc_perm *ipc;
836 for (id = 0; id < pos && total < ids->in_use; id++) {
837 ipc = idr_find(&ids->ipcs_idr, id);
842 if (total >= ids->in_use)
845 for ( ; pos < IPCMNI; pos++) {
846 ipc = idr_find(&ids->ipcs_idr, pos);
849 ipc_lock_by_ptr(ipc);
854 /* Out of range - return NULL to terminate iteration */
858 static void *sysvipc_proc_next(struct seq_file *s, void *it, loff_t *pos)
860 struct ipc_proc_iter *iter = s->private;
861 struct ipc_proc_iface *iface = iter->iface;
862 struct kern_ipc_perm *ipc = it;
864 /* If we had an ipc id locked before, unlock it */
865 if (ipc && ipc != SEQ_START_TOKEN)
868 return sysvipc_find_ipc(&iter->ns->ids[iface->ids], *pos, pos);
872 * File positions: pos 0 -> header, pos n -> ipc id = n - 1.
873 * SeqFile iterator: iterator value locked ipc pointer or SEQ_TOKEN_START.
875 static void *sysvipc_proc_start(struct seq_file *s, loff_t *pos)
877 struct ipc_proc_iter *iter = s->private;
878 struct ipc_proc_iface *iface = iter->iface;
881 ids = &iter->ns->ids[iface->ids];
884 * Take the lock - this will be released by the corresponding
887 down_read(&ids->rw_mutex);
889 /* pos < 0 is invalid */
893 /* pos == 0 means header */
895 return SEQ_START_TOKEN;
897 /* Find the (pos-1)th ipc */
898 return sysvipc_find_ipc(ids, *pos - 1, pos);
901 static void sysvipc_proc_stop(struct seq_file *s, void *it)
903 struct kern_ipc_perm *ipc = it;
904 struct ipc_proc_iter *iter = s->private;
905 struct ipc_proc_iface *iface = iter->iface;
908 /* If we had a locked structure, release it */
909 if (ipc && ipc != SEQ_START_TOKEN)
912 ids = &iter->ns->ids[iface->ids];
913 /* Release the lock we took in start() */
914 up_read(&ids->rw_mutex);
917 static int sysvipc_proc_show(struct seq_file *s, void *it)
919 struct ipc_proc_iter *iter = s->private;
920 struct ipc_proc_iface *iface = iter->iface;
922 if (it == SEQ_START_TOKEN)
923 return seq_puts(s, iface->header);
925 return iface->show(s, it);
928 static const struct seq_operations sysvipc_proc_seqops = {
929 .start = sysvipc_proc_start,
930 .stop = sysvipc_proc_stop,
931 .next = sysvipc_proc_next,
932 .show = sysvipc_proc_show,
935 static int sysvipc_proc_open(struct inode *inode, struct file *file)
938 struct seq_file *seq;
939 struct ipc_proc_iter *iter;
942 iter = kmalloc(sizeof(*iter), GFP_KERNEL);
946 ret = seq_open(file, &sysvipc_proc_seqops);
950 seq = file->private_data;
953 iter->iface = PDE_DATA(inode);
954 iter->ns = get_ipc_ns(current->nsproxy->ipc_ns);
962 static int sysvipc_proc_release(struct inode *inode, struct file *file)
964 struct seq_file *seq = file->private_data;
965 struct ipc_proc_iter *iter = seq->private;
966 put_ipc_ns(iter->ns);
967 return seq_release_private(inode, file);
970 static const struct file_operations sysvipc_proc_fops = {
971 .open = sysvipc_proc_open,
974 .release = sysvipc_proc_release,
976 #endif /* CONFIG_PROC_FS */