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) {
209 ipc_lock_object(ipc);
217 * ipc_get_maxid - get the last assigned id
218 * @ids: IPC identifier set
220 * 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);
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 > ids->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
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.
429 void ipc_rmid(struct ipc_ids *ids, struct kern_ipc_perm *ipcp)
431 int lid = ipcid_to_idx(ipcp->id);
433 idr_remove(&ids->ipcs_idr, lid);
443 * ipc_alloc - allocate ipc space
444 * @size: size desired
446 * Allocate memory from the appropriate pools and return a pointer to it.
447 * NULL is returned if the allocation fails
450 void *ipc_alloc(int size)
456 out = kmalloc(size, GFP_KERNEL);
461 * ipc_free - free ipc space
462 * @ptr: pointer returned by ipc_alloc
463 * @size: size of block
465 * Free a block created with ipc_alloc(). The caller must know the size
466 * used in the allocation call.
469 void ipc_free(void* ptr, int size)
480 } ____cacheline_aligned_in_smp;
483 * ipc_rcu_alloc - allocate ipc and rcu space
484 * @size: size desired
486 * Allocate memory for the rcu header structure + the object.
487 * Returns the pointer to the object or NULL upon failure.
489 void *ipc_rcu_alloc(int size)
492 * We prepend the allocation with the rcu struct
494 struct ipc_rcu *out = ipc_alloc(sizeof(struct ipc_rcu) + size);
497 atomic_set(&out->refcount, 1);
501 int ipc_rcu_getref(void *ptr)
503 struct ipc_rcu *p = ((struct ipc_rcu *)ptr) - 1;
505 return atomic_inc_not_zero(&p->refcount);
509 * ipc_schedule_free - free ipc + rcu space
510 * @head: RCU callback structure for queued work
512 static void ipc_schedule_free(struct rcu_head *head)
514 vfree(container_of(head, struct ipc_rcu, rcu));
517 void ipc_rcu_putref(void *ptr)
519 struct ipc_rcu *p = ((struct ipc_rcu *)ptr) - 1;
521 if (!atomic_dec_and_test(&p->refcount))
524 if (is_vmalloc_addr(ptr)) {
525 call_rcu(&p->rcu, ipc_schedule_free);
532 * ipcperms - check IPC permissions
534 * @ipcp: IPC permission set
535 * @flag: desired permission set.
537 * Check user, group, other permissions for access
538 * to ipc resources. return 0 if allowed
540 * @flag will most probably be 0 or S_...UGO from <linux/stat.h>
543 int ipcperms(struct ipc_namespace *ns, struct kern_ipc_perm *ipcp, short flag)
545 kuid_t euid = current_euid();
546 int requested_mode, granted_mode;
549 requested_mode = (flag >> 6) | (flag >> 3) | flag;
550 granted_mode = ipcp->mode;
551 if (uid_eq(euid, ipcp->cuid) ||
552 uid_eq(euid, ipcp->uid))
554 else if (in_group_p(ipcp->cgid) || in_group_p(ipcp->gid))
556 /* is there some bit set in requested_mode but not in granted_mode? */
557 if ((requested_mode & ~granted_mode & 0007) &&
558 !ns_capable(ns->user_ns, CAP_IPC_OWNER))
561 return security_ipc_permission(ipcp, flag);
565 * Functions to convert between the kern_ipc_perm structure and the
566 * old/new ipc_perm structures
570 * kernel_to_ipc64_perm - convert kernel ipc permissions to user
571 * @in: kernel permissions
572 * @out: new style IPC permissions
574 * Turn the kernel object @in into a set of permissions descriptions
575 * for returning to userspace (@out).
579 void kernel_to_ipc64_perm (struct kern_ipc_perm *in, struct ipc64_perm *out)
582 out->uid = from_kuid_munged(current_user_ns(), in->uid);
583 out->gid = from_kgid_munged(current_user_ns(), in->gid);
584 out->cuid = from_kuid_munged(current_user_ns(), in->cuid);
585 out->cgid = from_kgid_munged(current_user_ns(), in->cgid);
586 out->mode = in->mode;
591 * ipc64_perm_to_ipc_perm - convert new ipc permissions to old
592 * @in: new style IPC permissions
593 * @out: old style IPC permissions
595 * Turn the new style permissions object @in into a compatibility
596 * object and store it into the @out pointer.
599 void ipc64_perm_to_ipc_perm (struct ipc64_perm *in, struct ipc_perm *out)
602 SET_UID(out->uid, in->uid);
603 SET_GID(out->gid, in->gid);
604 SET_UID(out->cuid, in->cuid);
605 SET_GID(out->cgid, in->cgid);
606 out->mode = in->mode;
612 * @ids: ipc identifier set
613 * @id: ipc id to look for
615 * Look for an id in the ipc ids idr and return associated ipc object.
617 * Call inside the RCU critical section.
618 * The ipc object is *not* locked on exit.
620 struct kern_ipc_perm *ipc_obtain_object(struct ipc_ids *ids, int id)
622 struct kern_ipc_perm *out;
623 int lid = ipcid_to_idx(id);
625 out = idr_find(&ids->ipcs_idr, lid);
627 return ERR_PTR(-EINVAL);
633 * ipc_lock - Lock an ipc structure without rwsem held
634 * @ids: IPC identifier set
635 * @id: ipc id to look for
637 * Look for an id in the ipc ids idr and lock the associated ipc object.
639 * The ipc object is locked on successful exit.
641 struct kern_ipc_perm *ipc_lock(struct ipc_ids *ids, int id)
643 struct kern_ipc_perm *out;
646 out = ipc_obtain_object(ids, id);
650 spin_lock(&out->lock);
652 /* ipc_rmid() may have already freed the ID while ipc_lock
653 * was spinning: here verify that the structure is still valid
658 spin_unlock(&out->lock);
659 out = ERR_PTR(-EINVAL);
666 * ipc_obtain_object_check
667 * @ids: ipc identifier set
668 * @id: ipc id to look for
670 * Similar to ipc_obtain_object() but also checks
671 * the ipc object reference counter.
673 * Call inside the RCU critical section.
674 * The ipc object is *not* locked on exit.
676 struct kern_ipc_perm *ipc_obtain_object_check(struct ipc_ids *ids, int id)
678 struct kern_ipc_perm *out = ipc_obtain_object(ids, id);
683 if (ipc_checkid(out, id))
684 return ERR_PTR(-EIDRM);
690 * ipcget - Common sys_*get() code
692 * @ids : IPC identifier set
693 * @ops : operations to be called on ipc object creation, permission checks
695 * @params : the parameters needed by the previous operations.
697 * Common routine called by sys_msgget(), sys_semget() and sys_shmget().
699 int ipcget(struct ipc_namespace *ns, struct ipc_ids *ids,
700 struct ipc_ops *ops, struct ipc_params *params)
702 if (params->key == IPC_PRIVATE)
703 return ipcget_new(ns, ids, ops, params);
705 return ipcget_public(ns, ids, ops, params);
709 * ipc_update_perm - update the permissions of an IPC.
710 * @in: the permission given as input.
711 * @out: the permission of the ipc to set.
713 int ipc_update_perm(struct ipc64_perm *in, struct kern_ipc_perm *out)
715 kuid_t uid = make_kuid(current_user_ns(), in->uid);
716 kgid_t gid = make_kgid(current_user_ns(), in->gid);
717 if (!uid_valid(uid) || !gid_valid(gid))
722 out->mode = (out->mode & ~S_IRWXUGO)
723 | (in->mode & S_IRWXUGO);
729 * ipcctl_pre_down_nolock - retrieve an ipc and check permissions for some IPC_XXX cmd
730 * @ns: the ipc namespace
731 * @ids: the table of ids where to look for the ipc
732 * @id: the id of the ipc to retrieve
733 * @cmd: the cmd to check
734 * @perm: the permission to set
735 * @extra_perm: one extra permission parameter used by msq
737 * This function does some common audit and permissions check for some IPC_XXX
738 * cmd and is called from semctl_down, shmctl_down and msgctl_down.
739 * It must be called without any lock held and
740 * - retrieves the ipc with the given id in the given table.
741 * - performs some audit and permission check, depending on the given cmd
742 * - returns a pointer to the ipc object or otherwise, the corresponding error.
744 * Call holding the both the rwsem and the rcu read lock.
746 struct kern_ipc_perm *ipcctl_pre_down_nolock(struct ipc_namespace *ns,
747 struct ipc_ids *ids, int id, int cmd,
748 struct ipc64_perm *perm, int extra_perm)
752 struct kern_ipc_perm *ipcp;
754 ipcp = ipc_obtain_object_check(ids, id);
762 audit_ipc_set_perm(extra_perm, perm->uid,
763 perm->gid, perm->mode);
765 euid = current_euid();
766 if (uid_eq(euid, ipcp->cuid) || uid_eq(euid, ipcp->uid) ||
767 ns_capable(ns->user_ns, CAP_SYS_ADMIN))
768 return ipcp; /* successful lookup */
773 #ifdef CONFIG_ARCH_WANT_IPC_PARSE_VERSION
777 * ipc_parse_version - IPC call version
778 * @cmd: pointer to command
780 * Return IPC_64 for new style IPC and IPC_OLD for old style IPC.
781 * The @cmd value is turned from an encoding command and version into
782 * just the command code.
785 int ipc_parse_version (int *cmd)
795 #endif /* CONFIG_ARCH_WANT_IPC_PARSE_VERSION */
797 #ifdef CONFIG_PROC_FS
798 struct ipc_proc_iter {
799 struct ipc_namespace *ns;
800 struct ipc_proc_iface *iface;
804 * This routine locks the ipc structure found at least at position pos.
806 static struct kern_ipc_perm *sysvipc_find_ipc(struct ipc_ids *ids, loff_t pos,
809 struct kern_ipc_perm *ipc;
813 for (id = 0; id < pos && total < ids->in_use; id++) {
814 ipc = idr_find(&ids->ipcs_idr, id);
819 if (total >= ids->in_use)
822 for ( ; pos < IPCMNI; pos++) {
823 ipc = idr_find(&ids->ipcs_idr, pos);
827 ipc_lock_object(ipc);
832 /* Out of range - return NULL to terminate iteration */
836 static void *sysvipc_proc_next(struct seq_file *s, void *it, loff_t *pos)
838 struct ipc_proc_iter *iter = s->private;
839 struct ipc_proc_iface *iface = iter->iface;
840 struct kern_ipc_perm *ipc = it;
842 /* If we had an ipc id locked before, unlock it */
843 if (ipc && ipc != SEQ_START_TOKEN)
846 return sysvipc_find_ipc(&iter->ns->ids[iface->ids], *pos, pos);
850 * File positions: pos 0 -> header, pos n -> ipc id = n - 1.
851 * SeqFile iterator: iterator value locked ipc pointer or SEQ_TOKEN_START.
853 static void *sysvipc_proc_start(struct seq_file *s, loff_t *pos)
855 struct ipc_proc_iter *iter = s->private;
856 struct ipc_proc_iface *iface = iter->iface;
859 ids = &iter->ns->ids[iface->ids];
862 * Take the lock - this will be released by the corresponding
865 down_read(&ids->rwsem);
867 /* pos < 0 is invalid */
871 /* pos == 0 means header */
873 return SEQ_START_TOKEN;
875 /* Find the (pos-1)th ipc */
876 return sysvipc_find_ipc(ids, *pos - 1, pos);
879 static void sysvipc_proc_stop(struct seq_file *s, void *it)
881 struct kern_ipc_perm *ipc = it;
882 struct ipc_proc_iter *iter = s->private;
883 struct ipc_proc_iface *iface = iter->iface;
886 /* If we had a locked structure, release it */
887 if (ipc && ipc != SEQ_START_TOKEN)
890 ids = &iter->ns->ids[iface->ids];
891 /* Release the lock we took in start() */
892 up_read(&ids->rwsem);
895 static int sysvipc_proc_show(struct seq_file *s, void *it)
897 struct ipc_proc_iter *iter = s->private;
898 struct ipc_proc_iface *iface = iter->iface;
900 if (it == SEQ_START_TOKEN)
901 return seq_puts(s, iface->header);
903 return iface->show(s, it);
906 static const struct seq_operations sysvipc_proc_seqops = {
907 .start = sysvipc_proc_start,
908 .stop = sysvipc_proc_stop,
909 .next = sysvipc_proc_next,
910 .show = sysvipc_proc_show,
913 static int sysvipc_proc_open(struct inode *inode, struct file *file)
916 struct seq_file *seq;
917 struct ipc_proc_iter *iter;
920 iter = kmalloc(sizeof(*iter), GFP_KERNEL);
924 ret = seq_open(file, &sysvipc_proc_seqops);
928 seq = file->private_data;
931 iter->iface = PDE_DATA(inode);
932 iter->ns = get_ipc_ns(current->nsproxy->ipc_ns);
940 static int sysvipc_proc_release(struct inode *inode, struct file *file)
942 struct seq_file *seq = file->private_data;
943 struct ipc_proc_iter *iter = seq->private;
944 put_ipc_ns(iter->ns);
945 return seq_release_private(inode, file);
948 static const struct file_operations sysvipc_proc_fops = {
949 .open = sysvipc_proc_open,
952 .release = sysvipc_proc_release,
954 #endif /* CONFIG_PROC_FS */