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/capability.h>
27 #include <linux/highuid.h>
28 #include <linux/security.h>
29 #include <linux/rcupdate.h>
30 #include <linux/workqueue.h>
31 #include <linux/seq_file.h>
32 #include <linux/proc_fs.h>
33 #include <linux/audit.h>
34 #include <linux/nsproxy.h>
35 #include <linux/rwsem.h>
36 #include <linux/ipc_namespace.h>
38 #include <asm/unistd.h>
42 struct ipc_proc_iface {
46 int (*show)(struct seq_file *, void *);
49 struct ipc_namespace init_ipc_ns = {
51 .refcount = ATOMIC_INIT(2),
55 atomic_t nr_ipc_ns = ATOMIC_INIT(1);
59 * ipc_init - initialise IPC subsystem
61 * The various system5 IPC resources (semaphores, messages and shared
62 * memory) are initialised
65 static int __init ipc_init(void)
75 * ipc_init_ids - initialise IPC identifiers
76 * @ids: Identifier set
78 * Set up the sequence range to use for the ipc identifier range (limited
79 * below IPCMNI) then initialise the ids idr.
82 void ipc_init_ids(struct ipc_ids *ids)
84 init_rwsem(&ids->rw_mutex);
89 int seq_limit = INT_MAX/SEQ_MULTIPLIER;
90 if(seq_limit > USHRT_MAX)
91 ids->seq_max = USHRT_MAX;
93 ids->seq_max = seq_limit;
96 idr_init(&ids->ipcs_idr);
100 static const struct file_operations sysvipc_proc_fops;
102 * ipc_init_proc_interface - Create a proc interface for sysipc types using a seq_file interface.
103 * @path: Path in procfs
104 * @header: Banner to be printed at the beginning of the file.
105 * @ids: ipc id table to iterate.
106 * @show: show routine.
108 void __init ipc_init_proc_interface(const char *path, const char *header,
109 int ids, int (*show)(struct seq_file *, void *))
111 struct proc_dir_entry *pde;
112 struct ipc_proc_iface *iface;
114 iface = kmalloc(sizeof(*iface), GFP_KERNEL);
118 iface->header = header;
122 pde = create_proc_entry(path,
123 S_IRUGO, /* world readable */
124 NULL /* parent dir */);
127 pde->proc_fops = &sysvipc_proc_fops;
135 * ipc_findkey - find a key in an ipc identifier set
136 * @ids: Identifier set
137 * @key: The key to find
139 * Requires ipc_ids.rw_mutex locked.
140 * Returns the LOCKED pointer to the ipc structure if found or NULL
142 * If key is found ipc points to the owning ipc structure
145 static struct kern_ipc_perm *ipc_findkey(struct ipc_ids *ids, key_t key)
147 struct kern_ipc_perm *ipc;
151 for (total = 0, next_id = 0; total < ids->in_use; next_id++) {
152 ipc = idr_find(&ids->ipcs_idr, next_id);
157 if (ipc->key != key) {
162 ipc_lock_by_ptr(ipc);
170 * ipc_get_maxid - get the last assigned id
171 * @ids: IPC identifier set
173 * Called with ipc_ids.rw_mutex held.
176 int ipc_get_maxid(struct ipc_ids *ids)
178 struct kern_ipc_perm *ipc;
182 if (ids->in_use == 0)
185 if (ids->in_use == IPCMNI)
188 /* Look for the last assigned id */
190 for (id = 0; id < IPCMNI && total < ids->in_use; id++) {
191 ipc = idr_find(&ids->ipcs_idr, id);
201 * ipc_addid - add an IPC identifier
202 * @ids: IPC identifier set
203 * @new: new IPC permission set
204 * @size: limit for the number of used ids
206 * Add an entry 'new' to the IPC ids idr. The permissions object is
207 * initialised and the first free entry is set up and the id assigned
208 * is returned. The 'new' entry is returned in a locked state on success.
209 * On failure the entry is not locked and a negative err-code is returned.
211 * Called with ipc_ids.rw_mutex held as a writer.
214 int ipc_addid(struct ipc_ids* ids, struct kern_ipc_perm* new, int size)
221 if (ids->in_use >= size)
224 err = idr_get_new(&ids->ipcs_idr, new, &id);
230 new->cuid = new->uid = current->euid;
231 new->gid = new->cgid = current->egid;
233 new->seq = ids->seq++;
234 if(ids->seq > ids->seq_max)
237 new->id = ipc_buildid(id, new->seq);
238 spin_lock_init(&new->lock);
241 spin_lock(&new->lock);
246 * ipcget_new - create a new ipc object
248 * @ids: IPC identifer set
249 * @ops: the actual creation routine to call
250 * @params: its parameters
252 * This routine is called by sys_msgget, sys_semget() and sys_shmget()
253 * when the key is IPC_PRIVATE.
255 static int ipcget_new(struct ipc_namespace *ns, struct ipc_ids *ids,
256 struct ipc_ops *ops, struct ipc_params *params)
260 err = idr_pre_get(&ids->ipcs_idr, GFP_KERNEL);
265 down_write(&ids->rw_mutex);
266 err = ops->getnew(ns, params);
267 up_write(&ids->rw_mutex);
276 * ipc_check_perms - check security and permissions for an IPC
277 * @ipcp: ipc permission set
278 * @ops: the actual security routine to call
279 * @params: its parameters
281 * This routine is called by sys_msgget(), sys_semget() and sys_shmget()
282 * when the key is not IPC_PRIVATE and that key already exists in the
285 * On success, the IPC id is returned.
287 * It is called with ipc_ids.rw_mutex and ipcp->lock held.
289 static int ipc_check_perms(struct kern_ipc_perm *ipcp, struct ipc_ops *ops,
290 struct ipc_params *params)
294 if (ipcperms(ipcp, params->flg))
297 err = ops->associate(ipcp, params->flg);
306 * ipcget_public - get an ipc object or create a new one
308 * @ids: IPC identifer set
309 * @ops: the actual creation routine to call
310 * @params: its parameters
312 * This routine is called by sys_msgget, sys_semget() and sys_shmget()
313 * when the key is not IPC_PRIVATE.
314 * It adds a new entry if the key is not found and does some permission
315 * / security checkings if the key is found.
317 * On success, the ipc id is returned.
319 static int ipcget_public(struct ipc_namespace *ns, struct ipc_ids *ids,
320 struct ipc_ops *ops, struct ipc_params *params)
322 struct kern_ipc_perm *ipcp;
323 int flg = params->flg;
326 err = idr_pre_get(&ids->ipcs_idr, GFP_KERNEL);
329 * Take the lock as a writer since we are potentially going to add
330 * a new entry + read locks are not "upgradable"
332 down_write(&ids->rw_mutex);
333 ipcp = ipc_findkey(ids, params->key);
336 if (!(flg & IPC_CREAT))
341 err = ops->getnew(ns, params);
343 /* ipc object has been locked by ipc_findkey() */
345 if (flg & IPC_CREAT && flg & IPC_EXCL)
349 if (ops->more_checks)
350 err = ops->more_checks(ipcp, params);
353 * ipc_check_perms returns the IPC id on
356 err = ipc_check_perms(ipcp, ops, params);
360 up_write(&ids->rw_mutex);
370 * ipc_rmid - remove an IPC identifier
371 * @ids: IPC identifier set
372 * @ipcp: ipc perm structure containing the identifier to remove
374 * ipc_ids.rw_mutex (as a writer) and the spinlock for this ID are held
375 * before this function is called, and remain locked on the exit.
378 void ipc_rmid(struct ipc_ids *ids, struct kern_ipc_perm *ipcp)
380 int lid = ipcid_to_idx(ipcp->id);
382 idr_remove(&ids->ipcs_idr, lid);
392 * ipc_alloc - allocate ipc space
393 * @size: size desired
395 * Allocate memory from the appropriate pools and return a pointer to it.
396 * NULL is returned if the allocation fails
399 void* ipc_alloc(int size)
405 out = kmalloc(size, GFP_KERNEL);
410 * ipc_free - free ipc space
411 * @ptr: pointer returned by ipc_alloc
412 * @size: size of block
414 * Free a block created with ipc_alloc(). The caller must know the size
415 * used in the allocation call.
418 void ipc_free(void* ptr, int size)
428 * There are three headers that are prepended to the actual allocation:
429 * - during use: ipc_rcu_hdr.
430 * - during the rcu grace period: ipc_rcu_grace.
431 * - [only if vmalloc]: ipc_rcu_sched.
432 * Their lifetime doesn't overlap, thus the headers share the same memory.
433 * Unlike a normal union, they are right-aligned, thus some container_of
434 * forward/backward casting is necessary:
447 /* "void *" makes sure alignment of following data is sane. */
453 struct work_struct work;
454 /* "void *" makes sure alignment of following data is sane. */
458 #define HDRLEN_KMALLOC (sizeof(struct ipc_rcu_grace) > sizeof(struct ipc_rcu_hdr) ? \
459 sizeof(struct ipc_rcu_grace) : sizeof(struct ipc_rcu_hdr))
460 #define HDRLEN_VMALLOC (sizeof(struct ipc_rcu_sched) > HDRLEN_KMALLOC ? \
461 sizeof(struct ipc_rcu_sched) : HDRLEN_KMALLOC)
463 static inline int rcu_use_vmalloc(int size)
465 /* Too big for a single page? */
466 if (HDRLEN_KMALLOC + size > PAGE_SIZE)
472 * ipc_rcu_alloc - allocate ipc and rcu space
473 * @size: size desired
475 * Allocate memory for the rcu header structure + the object.
476 * Returns the pointer to the object.
477 * NULL is returned if the allocation fails.
480 void* ipc_rcu_alloc(int size)
484 * We prepend the allocation with the rcu struct, and
485 * workqueue if necessary (for vmalloc).
487 if (rcu_use_vmalloc(size)) {
488 out = vmalloc(HDRLEN_VMALLOC + size);
490 out += HDRLEN_VMALLOC;
491 container_of(out, struct ipc_rcu_hdr, data)->is_vmalloc = 1;
492 container_of(out, struct ipc_rcu_hdr, data)->refcount = 1;
495 out = kmalloc(HDRLEN_KMALLOC + size, GFP_KERNEL);
497 out += HDRLEN_KMALLOC;
498 container_of(out, struct ipc_rcu_hdr, data)->is_vmalloc = 0;
499 container_of(out, struct ipc_rcu_hdr, data)->refcount = 1;
506 void ipc_rcu_getref(void *ptr)
508 container_of(ptr, struct ipc_rcu_hdr, data)->refcount++;
511 static void ipc_do_vfree(struct work_struct *work)
513 vfree(container_of(work, struct ipc_rcu_sched, work));
517 * ipc_schedule_free - free ipc + rcu space
518 * @head: RCU callback structure for queued work
520 * Since RCU callback function is called in bh,
521 * we need to defer the vfree to schedule_work().
523 static void ipc_schedule_free(struct rcu_head *head)
525 struct ipc_rcu_grace *grace;
526 struct ipc_rcu_sched *sched;
528 grace = container_of(head, struct ipc_rcu_grace, rcu);
529 sched = container_of(&(grace->data[0]), struct ipc_rcu_sched,
532 INIT_WORK(&sched->work, ipc_do_vfree);
533 schedule_work(&sched->work);
537 * ipc_immediate_free - free ipc + rcu space
538 * @head: RCU callback structure that contains pointer to be freed
540 * Free from the RCU callback context.
542 static void ipc_immediate_free(struct rcu_head *head)
544 struct ipc_rcu_grace *free =
545 container_of(head, struct ipc_rcu_grace, rcu);
549 void ipc_rcu_putref(void *ptr)
551 if (--container_of(ptr, struct ipc_rcu_hdr, data)->refcount > 0)
554 if (container_of(ptr, struct ipc_rcu_hdr, data)->is_vmalloc) {
555 call_rcu(&container_of(ptr, struct ipc_rcu_grace, data)->rcu,
558 call_rcu(&container_of(ptr, struct ipc_rcu_grace, data)->rcu,
564 * ipcperms - check IPC permissions
565 * @ipcp: IPC permission set
566 * @flag: desired permission set.
568 * Check user, group, other permissions for access
569 * to ipc resources. return 0 if allowed
572 int ipcperms (struct kern_ipc_perm *ipcp, short flag)
573 { /* flag will most probably be 0 or S_...UGO from <linux/stat.h> */
574 int requested_mode, granted_mode, err;
576 if (unlikely((err = audit_ipc_obj(ipcp))))
578 requested_mode = (flag >> 6) | (flag >> 3) | flag;
579 granted_mode = ipcp->mode;
580 if (current->euid == ipcp->cuid || current->euid == ipcp->uid)
582 else if (in_group_p(ipcp->cgid) || in_group_p(ipcp->gid))
584 /* is there some bit set in requested_mode but not in granted_mode? */
585 if ((requested_mode & ~granted_mode & 0007) &&
586 !capable(CAP_IPC_OWNER))
589 return security_ipc_permission(ipcp, flag);
593 * Functions to convert between the kern_ipc_perm structure and the
594 * old/new ipc_perm structures
598 * kernel_to_ipc64_perm - convert kernel ipc permissions to user
599 * @in: kernel permissions
600 * @out: new style IPC permissions
602 * Turn the kernel object @in into a set of permissions descriptions
603 * for returning to userspace (@out).
607 void kernel_to_ipc64_perm (struct kern_ipc_perm *in, struct ipc64_perm *out)
612 out->cuid = in->cuid;
613 out->cgid = in->cgid;
614 out->mode = in->mode;
619 * ipc64_perm_to_ipc_perm - convert new ipc permissions to old
620 * @in: new style IPC permissions
621 * @out: old style IPC permissions
623 * Turn the new style permissions object @in into a compatibility
624 * object and store it into the @out pointer.
627 void ipc64_perm_to_ipc_perm (struct ipc64_perm *in, struct ipc_perm *out)
630 SET_UID(out->uid, in->uid);
631 SET_GID(out->gid, in->gid);
632 SET_UID(out->cuid, in->cuid);
633 SET_GID(out->cgid, in->cgid);
634 out->mode = in->mode;
639 * ipc_lock - Lock an ipc structure without rw_mutex held
640 * @ids: IPC identifier set
641 * @id: ipc id to look for
643 * Look for an id in the ipc ids idr and lock the associated ipc object.
645 * The ipc object is locked on exit.
647 * This is the routine that should be called when the rw_mutex is not already
648 * held, i.e. idr tree not protected: it protects the idr tree in read mode
649 * during the idr_find().
652 struct kern_ipc_perm *ipc_lock(struct ipc_ids *ids, int id)
654 struct kern_ipc_perm *out;
655 int lid = ipcid_to_idx(id);
657 down_read(&ids->rw_mutex);
660 out = idr_find(&ids->ipcs_idr, lid);
663 up_read(&ids->rw_mutex);
664 return ERR_PTR(-EINVAL);
667 up_read(&ids->rw_mutex);
669 spin_lock(&out->lock);
671 /* ipc_rmid() may have already freed the ID while ipc_lock
672 * was spinning: here verify that the structure is still valid
675 spin_unlock(&out->lock);
677 return ERR_PTR(-EINVAL);
684 * ipc_lock_down - Lock an ipc structure with rw_sem held
685 * @ids: IPC identifier set
686 * @id: ipc id to look for
688 * Look for an id in the ipc ids idr and lock the associated ipc object.
690 * The ipc object is locked on exit.
692 * This is the routine that should be called when the rw_mutex is already
693 * held, i.e. idr tree protected.
696 struct kern_ipc_perm *ipc_lock_down(struct ipc_ids *ids, int id)
698 struct kern_ipc_perm *out;
699 int lid = ipcid_to_idx(id);
702 out = idr_find(&ids->ipcs_idr, lid);
705 return ERR_PTR(-EINVAL);
708 spin_lock(&out->lock);
711 * No need to verify that the structure is still valid since the
717 struct kern_ipc_perm *ipc_lock_check_down(struct ipc_ids *ids, int id)
719 struct kern_ipc_perm *out;
721 out = ipc_lock_down(ids, id);
725 if (ipc_checkid(out, id)) {
727 return ERR_PTR(-EIDRM);
733 struct kern_ipc_perm *ipc_lock_check(struct ipc_ids *ids, int id)
735 struct kern_ipc_perm *out;
737 out = ipc_lock(ids, id);
741 if (ipc_checkid(out, id)) {
743 return ERR_PTR(-EIDRM);
750 * ipcget - Common sys_*get() code
752 * @ids : IPC identifier set
753 * @ops : operations to be called on ipc object creation, permission checks
755 * @params : the parameters needed by the previous operations.
757 * Common routine called by sys_msgget(), sys_semget() and sys_shmget().
759 int ipcget(struct ipc_namespace *ns, struct ipc_ids *ids,
760 struct ipc_ops *ops, struct ipc_params *params)
762 if (params->key == IPC_PRIVATE)
763 return ipcget_new(ns, ids, ops, params);
765 return ipcget_public(ns, ids, ops, params);
768 #ifdef __ARCH_WANT_IPC_PARSE_VERSION
772 * ipc_parse_version - IPC call version
773 * @cmd: pointer to command
775 * Return IPC_64 for new style IPC and IPC_OLD for old style IPC.
776 * The @cmd value is turned from an encoding command and version into
777 * just the command code.
780 int ipc_parse_version (int *cmd)
790 #endif /* __ARCH_WANT_IPC_PARSE_VERSION */
792 #ifdef CONFIG_PROC_FS
793 struct ipc_proc_iter {
794 struct ipc_namespace *ns;
795 struct ipc_proc_iface *iface;
799 * This routine locks the ipc structure found at least at position pos.
801 static struct kern_ipc_perm *sysvipc_find_ipc(struct ipc_ids *ids, loff_t pos,
804 struct kern_ipc_perm *ipc;
808 for (id = 0; id < pos && total < ids->in_use; id++) {
809 ipc = idr_find(&ids->ipcs_idr, id);
814 if (total >= ids->in_use)
817 for ( ; pos < IPCMNI; pos++) {
818 ipc = idr_find(&ids->ipcs_idr, pos);
821 ipc_lock_by_ptr(ipc);
826 /* Out of range - return NULL to terminate iteration */
830 static void *sysvipc_proc_next(struct seq_file *s, void *it, loff_t *pos)
832 struct ipc_proc_iter *iter = s->private;
833 struct ipc_proc_iface *iface = iter->iface;
834 struct kern_ipc_perm *ipc = it;
836 /* If we had an ipc id locked before, unlock it */
837 if (ipc && ipc != SEQ_START_TOKEN)
840 return sysvipc_find_ipc(&iter->ns->ids[iface->ids], *pos, pos);
844 * File positions: pos 0 -> header, pos n -> ipc id = n - 1.
845 * SeqFile iterator: iterator value locked ipc pointer or SEQ_TOKEN_START.
847 static void *sysvipc_proc_start(struct seq_file *s, loff_t *pos)
849 struct ipc_proc_iter *iter = s->private;
850 struct ipc_proc_iface *iface = iter->iface;
853 ids = &iter->ns->ids[iface->ids];
856 * Take the lock - this will be released by the corresponding
859 down_read(&ids->rw_mutex);
861 /* pos < 0 is invalid */
865 /* pos == 0 means header */
867 return SEQ_START_TOKEN;
869 /* Find the (pos-1)th ipc */
870 return sysvipc_find_ipc(ids, *pos - 1, pos);
873 static void sysvipc_proc_stop(struct seq_file *s, void *it)
875 struct kern_ipc_perm *ipc = it;
876 struct ipc_proc_iter *iter = s->private;
877 struct ipc_proc_iface *iface = iter->iface;
880 /* If we had a locked structure, release it */
881 if (ipc && ipc != SEQ_START_TOKEN)
884 ids = &iter->ns->ids[iface->ids];
885 /* Release the lock we took in start() */
886 up_read(&ids->rw_mutex);
889 static int sysvipc_proc_show(struct seq_file *s, void *it)
891 struct ipc_proc_iter *iter = s->private;
892 struct ipc_proc_iface *iface = iter->iface;
894 if (it == SEQ_START_TOKEN)
895 return seq_puts(s, iface->header);
897 return iface->show(s, it);
900 static struct seq_operations sysvipc_proc_seqops = {
901 .start = sysvipc_proc_start,
902 .stop = sysvipc_proc_stop,
903 .next = sysvipc_proc_next,
904 .show = sysvipc_proc_show,
907 static int sysvipc_proc_open(struct inode *inode, struct file *file)
910 struct seq_file *seq;
911 struct ipc_proc_iter *iter;
914 iter = kmalloc(sizeof(*iter), GFP_KERNEL);
918 ret = seq_open(file, &sysvipc_proc_seqops);
922 seq = file->private_data;
925 iter->iface = PDE(inode)->data;
926 iter->ns = get_ipc_ns(current->nsproxy->ipc_ns);
934 static int sysvipc_proc_release(struct inode *inode, struct file *file)
936 struct seq_file *seq = file->private_data;
937 struct ipc_proc_iter *iter = seq->private;
938 put_ipc_ns(iter->ns);
939 return seq_release_private(inode, file);
942 static const struct file_operations sysvipc_proc_fops = {
943 .open = sysvipc_proc_open,
946 .release = sysvipc_proc_release,
948 #endif /* CONFIG_PROC_FS */