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 *);
75 * ipc_init - initialise ipc subsystem
77 * The various sysv ipc resources (semaphores, messages and shared
78 * memory) are initialised.
80 * A callback routine is registered into the memory hotplug notifier
81 * chain: since msgmni scales to lowmem this callback routine will be
82 * called upon successful memory add / remove to recompute msmgni.
84 static int __init
ipc_init(void)
91 device_initcall(ipc_init
);
94 * ipc_init_ids - initialise ipc identifiers
95 * @ids: ipc identifier set
97 * Set up the sequence range to use for the ipc identifier range (limited
98 * below IPCMNI) then initialise the ids idr.
100 void ipc_init_ids(struct ipc_ids
*ids
)
105 init_rwsem(&ids
->rwsem
);
106 idr_init(&ids
->ipcs_idr
);
109 #ifdef CONFIG_PROC_FS
110 static const struct file_operations sysvipc_proc_fops
;
112 * ipc_init_proc_interface - create a proc interface for sysipc types using a seq_file interface.
113 * @path: Path in procfs
114 * @header: Banner to be printed at the beginning of the file.
115 * @ids: ipc id table to iterate.
116 * @show: show routine.
118 void __init
ipc_init_proc_interface(const char *path
, const char *header
,
119 int ids
, int (*show
)(struct seq_file
*, void *))
121 struct proc_dir_entry
*pde
;
122 struct ipc_proc_iface
*iface
;
124 iface
= kmalloc(sizeof(*iface
), GFP_KERNEL
);
128 iface
->header
= header
;
132 pde
= proc_create_data(path
,
133 S_IRUGO
, /* world readable */
134 NULL
, /* parent dir */
143 * ipc_findkey - find a key in an ipc identifier set
144 * @ids: ipc identifier set
147 * Returns the locked pointer to the ipc structure if found or NULL
148 * otherwise. If key is found ipc points to the owning ipc structure
150 * Called with ipc_ids.rwsem held.
152 static struct kern_ipc_perm
*ipc_findkey(struct ipc_ids
*ids
, key_t key
)
154 struct kern_ipc_perm
*ipc
;
158 for (total
= 0, next_id
= 0; total
< ids
->in_use
; next_id
++) {
159 ipc
= idr_find(&ids
->ipcs_idr
, next_id
);
164 if (ipc
->key
!= key
) {
170 ipc_lock_object(ipc
);
178 * ipc_get_maxid - get the last assigned id
179 * @ids: ipc identifier set
181 * Called with ipc_ids.rwsem held.
183 int ipc_get_maxid(struct ipc_ids
*ids
)
185 struct kern_ipc_perm
*ipc
;
189 if (ids
->in_use
== 0)
192 if (ids
->in_use
== IPCMNI
)
195 /* Look for the last assigned id */
197 for (id
= 0; id
< IPCMNI
&& total
< ids
->in_use
; id
++) {
198 ipc
= idr_find(&ids
->ipcs_idr
, id
);
208 * ipc_addid - add an ipc identifier
209 * @ids: ipc identifier set
210 * @new: new ipc permission set
211 * @size: limit for the number of used ids
213 * Add an entry 'new' to the ipc ids idr. The permissions object is
214 * initialised and the first free entry is set up and the id assigned
215 * is returned. The 'new' entry is returned in a locked state on success.
216 * On failure the entry is not locked and a negative err-code is returned.
218 * Called with writer ipc_ids.rwsem held.
220 int ipc_addid(struct ipc_ids
*ids
, struct kern_ipc_perm
*new, int size
)
225 int next_id
= ids
->next_id
;
230 if (ids
->in_use
>= size
)
233 idr_preload(GFP_KERNEL
);
235 spin_lock_init(&new->lock
);
236 new->deleted
= false;
238 spin_lock(&new->lock
);
240 current_euid_egid(&euid
, &egid
);
241 new->cuid
= new->uid
= euid
;
242 new->gid
= new->cgid
= egid
;
244 id
= idr_alloc(&ids
->ipcs_idr
, new,
245 (next_id
< 0) ? 0 : ipcid_to_idx(next_id
), 0,
249 spin_unlock(&new->lock
);
257 new->seq
= ids
->seq
++;
258 if (ids
->seq
> IPCID_SEQ_MAX
)
261 new->seq
= ipcid_to_seqx(next_id
);
265 new->id
= ipc_buildid(id
, new->seq
);
270 * ipcget_new - create a new ipc object
272 * @ids: ipc identifier set
273 * @ops: the actual creation routine to call
274 * @params: its parameters
276 * This routine is called by sys_msgget, sys_semget() and sys_shmget()
277 * when the key is IPC_PRIVATE.
279 static int ipcget_new(struct ipc_namespace
*ns
, struct ipc_ids
*ids
,
280 const struct ipc_ops
*ops
, struct ipc_params
*params
)
284 down_write(&ids
->rwsem
);
285 err
= ops
->getnew(ns
, params
);
286 up_write(&ids
->rwsem
);
291 * ipc_check_perms - check security and permissions for an ipc object
293 * @ipcp: ipc permission set
294 * @ops: the actual security routine to call
295 * @params: its parameters
297 * This routine is called by sys_msgget(), sys_semget() and sys_shmget()
298 * when the key is not IPC_PRIVATE and that key already exists in the
301 * On success, the ipc id is returned.
303 * It is called with ipc_ids.rwsem and ipcp->lock held.
305 static int ipc_check_perms(struct ipc_namespace
*ns
,
306 struct kern_ipc_perm
*ipcp
,
307 const struct ipc_ops
*ops
,
308 struct ipc_params
*params
)
312 if (ipcperms(ns
, ipcp
, params
->flg
))
315 err
= ops
->associate(ipcp
, params
->flg
);
324 * ipcget_public - get an ipc object or create a new one
326 * @ids: ipc identifier set
327 * @ops: the actual creation routine to call
328 * @params: its parameters
330 * This routine is called by sys_msgget, sys_semget() and sys_shmget()
331 * when the key is not IPC_PRIVATE.
332 * It adds a new entry if the key is not found and does some permission
333 * / security checkings if the key is found.
335 * On success, the ipc id is returned.
337 static int ipcget_public(struct ipc_namespace
*ns
, struct ipc_ids
*ids
,
338 const struct ipc_ops
*ops
, struct ipc_params
*params
)
340 struct kern_ipc_perm
*ipcp
;
341 int flg
= params
->flg
;
345 * Take the lock as a writer since we are potentially going to add
346 * a new entry + read locks are not "upgradable"
348 down_write(&ids
->rwsem
);
349 ipcp
= ipc_findkey(ids
, params
->key
);
352 if (!(flg
& IPC_CREAT
))
355 err
= ops
->getnew(ns
, params
);
357 /* ipc object has been locked by ipc_findkey() */
359 if (flg
& IPC_CREAT
&& flg
& IPC_EXCL
)
363 if (ops
->more_checks
)
364 err
= ops
->more_checks(ipcp
, params
);
367 * ipc_check_perms returns the IPC id on
370 err
= ipc_check_perms(ns
, ipcp
, ops
, params
);
374 up_write(&ids
->rwsem
);
381 * ipc_rmid - remove an ipc identifier
382 * @ids: ipc identifier set
383 * @ipcp: ipc perm structure containing the identifier to remove
385 * ipc_ids.rwsem (as a writer) and the spinlock for this ID are held
386 * before this function is called, and remain locked on the exit.
388 void ipc_rmid(struct ipc_ids
*ids
, struct kern_ipc_perm
*ipcp
)
390 int lid
= ipcid_to_idx(ipcp
->id
);
392 idr_remove(&ids
->ipcs_idr
, lid
);
394 ipcp
->deleted
= true;
398 * ipc_alloc - allocate ipc space
399 * @size: size desired
401 * Allocate memory from the appropriate pools and return a pointer to it.
402 * NULL is returned if the allocation fails
404 void *ipc_alloc(int size
)
407 if (size
> PAGE_SIZE
)
410 out
= kmalloc(size
, GFP_KERNEL
);
415 * ipc_free - free ipc space
416 * @ptr: pointer returned by ipc_alloc
418 * Free a block created with ipc_alloc().
420 void ipc_free(void *ptr
)
426 * ipc_rcu_alloc - allocate ipc and rcu space
427 * @size: size desired
429 * Allocate memory for the rcu header structure + the object.
430 * Returns the pointer to the object or NULL upon failure.
432 void *ipc_rcu_alloc(int size
)
435 * We prepend the allocation with the rcu struct
437 struct ipc_rcu
*out
= ipc_alloc(sizeof(struct ipc_rcu
) + size
);
440 atomic_set(&out
->refcount
, 1);
444 int ipc_rcu_getref(void *ptr
)
446 struct ipc_rcu
*p
= ((struct ipc_rcu
*)ptr
) - 1;
448 return atomic_inc_not_zero(&p
->refcount
);
451 void ipc_rcu_putref(void *ptr
, void (*func
)(struct rcu_head
*head
))
453 struct ipc_rcu
*p
= ((struct ipc_rcu
*)ptr
) - 1;
455 if (!atomic_dec_and_test(&p
->refcount
))
458 call_rcu(&p
->rcu
, func
);
461 void ipc_rcu_free(struct rcu_head
*head
)
463 struct ipc_rcu
*p
= container_of(head
, struct ipc_rcu
, rcu
);
469 * ipcperms - check ipc permissions
471 * @ipcp: ipc permission set
472 * @flag: desired permission set
474 * Check user, group, other permissions for access
475 * to ipc resources. return 0 if allowed
477 * @flag will most probably be 0 or S_...UGO from <linux/stat.h>
479 int ipcperms(struct ipc_namespace
*ns
, struct kern_ipc_perm
*ipcp
, short flag
)
481 kuid_t euid
= current_euid();
482 int requested_mode
, granted_mode
;
485 requested_mode
= (flag
>> 6) | (flag
>> 3) | flag
;
486 granted_mode
= ipcp
->mode
;
487 if (uid_eq(euid
, ipcp
->cuid
) ||
488 uid_eq(euid
, ipcp
->uid
))
490 else if (in_group_p(ipcp
->cgid
) || in_group_p(ipcp
->gid
))
492 /* is there some bit set in requested_mode but not in granted_mode? */
493 if ((requested_mode
& ~granted_mode
& 0007) &&
494 !ns_capable(ns
->user_ns
, CAP_IPC_OWNER
))
497 return security_ipc_permission(ipcp
, flag
);
501 * Functions to convert between the kern_ipc_perm structure and the
502 * old/new ipc_perm structures
506 * kernel_to_ipc64_perm - convert kernel ipc permissions to user
507 * @in: kernel permissions
508 * @out: new style ipc permissions
510 * Turn the kernel object @in into a set of permissions descriptions
511 * for returning to userspace (@out).
513 void kernel_to_ipc64_perm(struct kern_ipc_perm
*in
, struct ipc64_perm
*out
)
516 out
->uid
= from_kuid_munged(current_user_ns(), in
->uid
);
517 out
->gid
= from_kgid_munged(current_user_ns(), in
->gid
);
518 out
->cuid
= from_kuid_munged(current_user_ns(), in
->cuid
);
519 out
->cgid
= from_kgid_munged(current_user_ns(), in
->cgid
);
520 out
->mode
= in
->mode
;
525 * ipc64_perm_to_ipc_perm - convert new ipc permissions to old
526 * @in: new style ipc permissions
527 * @out: old style ipc permissions
529 * Turn the new style permissions object @in into a compatibility
530 * object and store it into the @out pointer.
532 void ipc64_perm_to_ipc_perm(struct ipc64_perm
*in
, struct ipc_perm
*out
)
535 SET_UID(out
->uid
, in
->uid
);
536 SET_GID(out
->gid
, in
->gid
);
537 SET_UID(out
->cuid
, in
->cuid
);
538 SET_GID(out
->cgid
, in
->cgid
);
539 out
->mode
= in
->mode
;
545 * @ids: ipc identifier set
546 * @id: ipc id to look for
548 * Look for an id in the ipc ids idr and return associated ipc object.
550 * Call inside the RCU critical section.
551 * The ipc object is *not* locked on exit.
553 struct kern_ipc_perm
*ipc_obtain_object_idr(struct ipc_ids
*ids
, int id
)
555 struct kern_ipc_perm
*out
;
556 int lid
= ipcid_to_idx(id
);
558 out
= idr_find(&ids
->ipcs_idr
, lid
);
560 return ERR_PTR(-EINVAL
);
566 * ipc_lock - lock an ipc structure without rwsem held
567 * @ids: ipc identifier set
568 * @id: ipc id to look for
570 * Look for an id in the ipc ids idr and lock the associated ipc object.
572 * The ipc object is locked on successful exit.
574 struct kern_ipc_perm
*ipc_lock(struct ipc_ids
*ids
, int id
)
576 struct kern_ipc_perm
*out
;
579 out
= ipc_obtain_object_idr(ids
, id
);
583 spin_lock(&out
->lock
);
586 * ipc_rmid() may have already freed the ID while ipc_lock()
587 * was spinning: here verify that the structure is still valid.
588 * Upon races with RMID, return -EIDRM, thus indicating that
589 * the ID points to a removed identifier.
591 if (ipc_valid_object(out
))
594 spin_unlock(&out
->lock
);
595 out
= ERR_PTR(-EIDRM
);
602 * ipc_obtain_object_check
603 * @ids: ipc identifier set
604 * @id: ipc id to look for
606 * Similar to ipc_obtain_object_idr() but also checks
607 * the ipc object reference counter.
609 * Call inside the RCU critical section.
610 * The ipc object is *not* locked on exit.
612 struct kern_ipc_perm
*ipc_obtain_object_check(struct ipc_ids
*ids
, int id
)
614 struct kern_ipc_perm
*out
= ipc_obtain_object_idr(ids
, id
);
619 if (ipc_checkid(out
, id
))
620 return ERR_PTR(-EINVAL
);
626 * ipcget - Common sys_*get() code
628 * @ids: ipc identifier set
629 * @ops: operations to be called on ipc object creation, permission checks
631 * @params: the parameters needed by the previous operations.
633 * Common routine called by sys_msgget(), sys_semget() and sys_shmget().
635 int ipcget(struct ipc_namespace
*ns
, struct ipc_ids
*ids
,
636 const struct ipc_ops
*ops
, struct ipc_params
*params
)
638 if (params
->key
== IPC_PRIVATE
)
639 return ipcget_new(ns
, ids
, ops
, params
);
641 return ipcget_public(ns
, ids
, ops
, params
);
645 * ipc_update_perm - update the permissions of an ipc object
646 * @in: the permission given as input.
647 * @out: the permission of the ipc to set.
649 int ipc_update_perm(struct ipc64_perm
*in
, struct kern_ipc_perm
*out
)
651 kuid_t uid
= make_kuid(current_user_ns(), in
->uid
);
652 kgid_t gid
= make_kgid(current_user_ns(), in
->gid
);
653 if (!uid_valid(uid
) || !gid_valid(gid
))
658 out
->mode
= (out
->mode
& ~S_IRWXUGO
)
659 | (in
->mode
& S_IRWXUGO
);
665 * ipcctl_pre_down_nolock - retrieve an ipc and check permissions for some IPC_XXX cmd
667 * @ids: the table of ids where to look for the ipc
668 * @id: the id of the ipc to retrieve
669 * @cmd: the cmd to check
670 * @perm: the permission to set
671 * @extra_perm: one extra permission parameter used by msq
673 * This function does some common audit and permissions check for some IPC_XXX
674 * cmd and is called from semctl_down, shmctl_down and msgctl_down.
675 * It must be called without any lock held and
676 * - retrieves the ipc with the given id in the given table.
677 * - performs some audit and permission check, depending on the given cmd
678 * - returns a pointer to the ipc object or otherwise, the corresponding error.
680 * Call holding the both the rwsem and the rcu read lock.
682 struct kern_ipc_perm
*ipcctl_pre_down_nolock(struct ipc_namespace
*ns
,
683 struct ipc_ids
*ids
, int id
, int cmd
,
684 struct ipc64_perm
*perm
, int extra_perm
)
688 struct kern_ipc_perm
*ipcp
;
690 ipcp
= ipc_obtain_object_check(ids
, id
);
698 audit_ipc_set_perm(extra_perm
, perm
->uid
,
699 perm
->gid
, perm
->mode
);
701 euid
= current_euid();
702 if (uid_eq(euid
, ipcp
->cuid
) || uid_eq(euid
, ipcp
->uid
) ||
703 ns_capable(ns
->user_ns
, CAP_SYS_ADMIN
))
704 return ipcp
; /* successful lookup */
709 #ifdef CONFIG_ARCH_WANT_IPC_PARSE_VERSION
713 * ipc_parse_version - ipc call version
714 * @cmd: pointer to command
716 * Return IPC_64 for new style IPC and IPC_OLD for old style IPC.
717 * The @cmd value is turned from an encoding command and version into
718 * just the command code.
720 int ipc_parse_version(int *cmd
)
730 #endif /* CONFIG_ARCH_WANT_IPC_PARSE_VERSION */
732 #ifdef CONFIG_PROC_FS
733 struct ipc_proc_iter
{
734 struct ipc_namespace
*ns
;
735 struct ipc_proc_iface
*iface
;
739 * This routine locks the ipc structure found at least at position pos.
741 static struct kern_ipc_perm
*sysvipc_find_ipc(struct ipc_ids
*ids
, loff_t pos
,
744 struct kern_ipc_perm
*ipc
;
748 for (id
= 0; id
< pos
&& total
< ids
->in_use
; id
++) {
749 ipc
= idr_find(&ids
->ipcs_idr
, id
);
754 if (total
>= ids
->in_use
)
757 for (; pos
< IPCMNI
; pos
++) {
758 ipc
= idr_find(&ids
->ipcs_idr
, pos
);
762 ipc_lock_object(ipc
);
767 /* Out of range - return NULL to terminate iteration */
771 static void *sysvipc_proc_next(struct seq_file
*s
, void *it
, loff_t
*pos
)
773 struct ipc_proc_iter
*iter
= s
->private;
774 struct ipc_proc_iface
*iface
= iter
->iface
;
775 struct kern_ipc_perm
*ipc
= it
;
777 /* If we had an ipc id locked before, unlock it */
778 if (ipc
&& ipc
!= SEQ_START_TOKEN
)
781 return sysvipc_find_ipc(&iter
->ns
->ids
[iface
->ids
], *pos
, pos
);
785 * File positions: pos 0 -> header, pos n -> ipc id = n - 1.
786 * SeqFile iterator: iterator value locked ipc pointer or SEQ_TOKEN_START.
788 static void *sysvipc_proc_start(struct seq_file
*s
, loff_t
*pos
)
790 struct ipc_proc_iter
*iter
= s
->private;
791 struct ipc_proc_iface
*iface
= iter
->iface
;
794 ids
= &iter
->ns
->ids
[iface
->ids
];
797 * Take the lock - this will be released by the corresponding
800 down_read(&ids
->rwsem
);
802 /* pos < 0 is invalid */
806 /* pos == 0 means header */
808 return SEQ_START_TOKEN
;
810 /* Find the (pos-1)th ipc */
811 return sysvipc_find_ipc(ids
, *pos
- 1, pos
);
814 static void sysvipc_proc_stop(struct seq_file
*s
, void *it
)
816 struct kern_ipc_perm
*ipc
= it
;
817 struct ipc_proc_iter
*iter
= s
->private;
818 struct ipc_proc_iface
*iface
= iter
->iface
;
821 /* If we had a locked structure, release it */
822 if (ipc
&& ipc
!= SEQ_START_TOKEN
)
825 ids
= &iter
->ns
->ids
[iface
->ids
];
826 /* Release the lock we took in start() */
827 up_read(&ids
->rwsem
);
830 static int sysvipc_proc_show(struct seq_file
*s
, void *it
)
832 struct ipc_proc_iter
*iter
= s
->private;
833 struct ipc_proc_iface
*iface
= iter
->iface
;
835 if (it
== SEQ_START_TOKEN
) {
836 seq_puts(s
, iface
->header
);
840 return iface
->show(s
, it
);
843 static const struct seq_operations sysvipc_proc_seqops
= {
844 .start
= sysvipc_proc_start
,
845 .stop
= sysvipc_proc_stop
,
846 .next
= sysvipc_proc_next
,
847 .show
= sysvipc_proc_show
,
850 static int sysvipc_proc_open(struct inode
*inode
, struct file
*file
)
852 struct ipc_proc_iter
*iter
;
854 iter
= __seq_open_private(file
, &sysvipc_proc_seqops
, sizeof(*iter
));
858 iter
->iface
= PDE_DATA(inode
);
859 iter
->ns
= get_ipc_ns(current
->nsproxy
->ipc_ns
);
864 static int sysvipc_proc_release(struct inode
*inode
, struct file
*file
)
866 struct seq_file
*seq
= file
->private_data
;
867 struct ipc_proc_iter
*iter
= seq
->private;
868 put_ipc_ns(iter
->ns
);
869 return seq_release_private(inode
, file
);
872 static const struct file_operations sysvipc_proc_fops
= {
873 .open
= sysvipc_proc_open
,
876 .release
= sysvipc_proc_release
,
878 #endif /* CONFIG_PROC_FS */