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>
36 #include <asm/unistd.h>
40 struct ipc_proc_iface
{
44 int (*show
)(struct seq_file
*, void *);
47 struct ipc_namespace init_ipc_ns
= {
49 .refcount
= ATOMIC_INIT(2),
53 static struct ipc_namespace
*clone_ipc_ns(struct ipc_namespace
*old_ns
)
56 struct ipc_namespace
*ns
;
59 ns
= kmalloc(sizeof(struct ipc_namespace
), GFP_KERNEL
);
63 err
= sem_init_ns(ns
);
66 err
= msg_init_ns(ns
);
69 err
= shm_init_ns(ns
);
86 struct ipc_namespace
*copy_ipcs(unsigned long flags
, struct ipc_namespace
*ns
)
88 struct ipc_namespace
*new_ns
;
93 if (!(flags
& CLONE_NEWIPC
))
96 new_ns
= clone_ipc_ns(ns
);
102 void free_ipc_ns(struct kref
*kref
)
104 struct ipc_namespace
*ns
;
106 ns
= container_of(kref
, struct ipc_namespace
, kref
);
114 * ipc_init - initialise IPC subsystem
116 * The various system5 IPC resources (semaphores, messages and shared
117 * memory) are initialised
120 static int __init
ipc_init(void)
127 __initcall(ipc_init
);
130 * ipc_init_ids - initialise IPC identifiers
131 * @ids: Identifier set
133 * Set up the sequence range to use for the ipc identifier range (limited
134 * below IPCMNI) then initialise the ids idr.
137 void ipc_init_ids(struct ipc_ids
*ids
)
139 mutex_init(&ids
->mutex
);
144 int seq_limit
= INT_MAX
/SEQ_MULTIPLIER
;
145 if(seq_limit
> USHRT_MAX
)
146 ids
->seq_max
= USHRT_MAX
;
148 ids
->seq_max
= seq_limit
;
151 idr_init(&ids
->ipcs_idr
);
154 #ifdef CONFIG_PROC_FS
155 static const struct file_operations sysvipc_proc_fops
;
157 * ipc_init_proc_interface - Create a proc interface for sysipc types using a seq_file interface.
158 * @path: Path in procfs
159 * @header: Banner to be printed at the beginning of the file.
160 * @ids: ipc id table to iterate.
161 * @show: show routine.
163 void __init
ipc_init_proc_interface(const char *path
, const char *header
,
164 int ids
, int (*show
)(struct seq_file
*, void *))
166 struct proc_dir_entry
*pde
;
167 struct ipc_proc_iface
*iface
;
169 iface
= kmalloc(sizeof(*iface
), GFP_KERNEL
);
173 iface
->header
= header
;
177 pde
= create_proc_entry(path
,
178 S_IRUGO
, /* world readable */
179 NULL
/* parent dir */);
182 pde
->proc_fops
= &sysvipc_proc_fops
;
190 * ipc_findkey - find a key in an ipc identifier set
191 * @ids: Identifier set
192 * @key: The key to find
194 * Requires ipc_ids.mutex locked.
195 * Returns the LOCKED pointer to the ipc structure if found or NULL
197 * If key is found ipc contains its ipc structure
200 static struct kern_ipc_perm
*ipc_findkey(struct ipc_ids
*ids
, key_t key
)
202 struct kern_ipc_perm
*ipc
;
206 for (total
= 0, next_id
= 0; total
< ids
->in_use
; next_id
++) {
207 ipc
= idr_find(&ids
->ipcs_idr
, next_id
);
212 if (ipc
->key
!= key
) {
217 ipc_lock_by_ptr(ipc
);
225 * ipc_get_maxid - get the last assigned id
226 * @ids: IPC identifier set
228 * Called with ipc_ids.mutex held.
231 int ipc_get_maxid(struct ipc_ids
*ids
)
233 struct kern_ipc_perm
*ipc
;
237 if (ids
->in_use
== 0)
240 if (ids
->in_use
== IPCMNI
)
243 /* Look for the last assigned id */
245 for (id
= 0; id
< IPCMNI
&& total
< ids
->in_use
; id
++) {
246 ipc
= idr_find(&ids
->ipcs_idr
, id
);
256 * ipc_addid - add an IPC identifier
257 * @ids: IPC identifier set
258 * @new: new IPC permission set
259 * @size: limit for the number of used ids
261 * Add an entry 'new' to the IPC idr. The permissions object is
262 * initialised and the first free entry is set up and the id assigned
263 * is returned. The list is returned in a locked state on success.
264 * On failure the list is not locked and -1 is returned.
266 * Called with ipc_ids.mutex held.
269 int ipc_addid(struct ipc_ids
* ids
, struct kern_ipc_perm
* new, int size
)
274 * rcu_dereference()() is not needed here since
275 * ipc_ids.mutex is held
280 if (ids
->in_use
>= size
)
283 err
= idr_get_new(&ids
->ipcs_idr
, new, &id
);
289 new->cuid
= new->uid
= current
->euid
;
290 new->gid
= new->cgid
= current
->egid
;
292 new->seq
= ids
->seq
++;
293 if(ids
->seq
> ids
->seq_max
)
296 spin_lock_init(&new->lock
);
299 spin_lock(&new->lock
);
304 * ipcget_new - create a new ipc object
306 * @ids: identifer set
307 * @ops: the actual creation routine to call
308 * @params: its parameters
310 * This routine is called sys_msgget, sys_semget() and sys_shmget() when
311 * the key is IPC_PRIVATE
313 int ipcget_new(struct ipc_namespace
*ns
, struct ipc_ids
*ids
,
314 struct ipc_ops
*ops
, struct ipc_params
*params
)
318 err
= idr_pre_get(&ids
->ipcs_idr
, GFP_KERNEL
);
323 mutex_lock(&ids
->mutex
);
324 err
= ops
->getnew(ns
, params
);
325 mutex_unlock(&ids
->mutex
);
331 * ipc_check_perms - check security and permissions for an IPC
332 * @ipcp: ipc permission set
333 * @ids: identifer set
334 * @ops: the actual security routine to call
335 * @params: its parameters
337 static int ipc_check_perms(struct kern_ipc_perm
*ipcp
, struct ipc_ops
*ops
,
338 struct ipc_params
*params
)
342 if (ipcperms(ipcp
, params
->flg
))
345 err
= ops
->associate(ipcp
, params
->flg
);
354 * ipcget_public - get an ipc object or create a new one
356 * @ids: identifer set
357 * @ops: the actual creation routine to call
358 * @params: its parameters
360 * This routine is called sys_msgget, sys_semget() and sys_shmget() when
361 * the key is not IPC_PRIVATE
363 int ipcget_public(struct ipc_namespace
*ns
, struct ipc_ids
*ids
,
364 struct ipc_ops
*ops
, struct ipc_params
*params
)
366 struct kern_ipc_perm
*ipcp
;
367 int flg
= params
->flg
;
370 err
= idr_pre_get(&ids
->ipcs_idr
, GFP_KERNEL
);
372 mutex_lock(&ids
->mutex
);
373 ipcp
= ipc_findkey(ids
, params
->key
);
376 if (!(flg
& IPC_CREAT
))
381 err
= ops
->getnew(ns
, params
);
383 /* ipc object has been locked by ipc_findkey() */
385 if (flg
& IPC_CREAT
&& flg
& IPC_EXCL
)
389 if (ops
->more_checks
)
390 err
= ops
->more_checks(ipcp
, params
);
392 err
= ipc_check_perms(ipcp
, ops
, params
);
396 mutex_unlock(&ids
->mutex
);
403 * ipc_rmid - remove an IPC identifier
404 * @ids: identifier set
405 * @id: ipc perm structure containing the identifier to remove
407 * The identifier must be valid, and in use. The kernel will panic if
408 * fed an invalid identifier. The entry is removed and internal
409 * variables recomputed.
410 * ipc_ids.mutex and the spinlock for this ID are held before this
411 * function is called, and remain locked on the exit.
414 void ipc_rmid(struct ipc_ids
*ids
, struct kern_ipc_perm
*ipcp
)
416 int lid
= ipcid_to_idx(ipcp
->id
);
418 idr_remove(&ids
->ipcs_idr
, lid
);
428 * ipc_alloc - allocate ipc space
429 * @size: size desired
431 * Allocate memory from the appropriate pools and return a pointer to it.
432 * NULL is returned if the allocation fails
435 void* ipc_alloc(int size
)
441 out
= kmalloc(size
, GFP_KERNEL
);
446 * ipc_free - free ipc space
447 * @ptr: pointer returned by ipc_alloc
448 * @size: size of block
450 * Free a block created with ipc_alloc(). The caller must know the size
451 * used in the allocation call.
454 void ipc_free(void* ptr
, int size
)
464 * There are three headers that are prepended to the actual allocation:
465 * - during use: ipc_rcu_hdr.
466 * - during the rcu grace period: ipc_rcu_grace.
467 * - [only if vmalloc]: ipc_rcu_sched.
468 * Their lifetime doesn't overlap, thus the headers share the same memory.
469 * Unlike a normal union, they are right-aligned, thus some container_of
470 * forward/backward casting is necessary:
483 /* "void *" makes sure alignment of following data is sane. */
489 struct work_struct work
;
490 /* "void *" makes sure alignment of following data is sane. */
494 #define HDRLEN_KMALLOC (sizeof(struct ipc_rcu_grace) > sizeof(struct ipc_rcu_hdr) ? \
495 sizeof(struct ipc_rcu_grace) : sizeof(struct ipc_rcu_hdr))
496 #define HDRLEN_VMALLOC (sizeof(struct ipc_rcu_sched) > HDRLEN_KMALLOC ? \
497 sizeof(struct ipc_rcu_sched) : HDRLEN_KMALLOC)
499 static inline int rcu_use_vmalloc(int size
)
501 /* Too big for a single page? */
502 if (HDRLEN_KMALLOC
+ size
> PAGE_SIZE
)
508 * ipc_rcu_alloc - allocate ipc and rcu space
509 * @size: size desired
511 * Allocate memory for the rcu header structure + the object.
512 * Returns the pointer to the object.
513 * NULL is returned if the allocation fails.
516 void* ipc_rcu_alloc(int size
)
520 * We prepend the allocation with the rcu struct, and
521 * workqueue if necessary (for vmalloc).
523 if (rcu_use_vmalloc(size
)) {
524 out
= vmalloc(HDRLEN_VMALLOC
+ size
);
526 out
+= HDRLEN_VMALLOC
;
527 container_of(out
, struct ipc_rcu_hdr
, data
)->is_vmalloc
= 1;
528 container_of(out
, struct ipc_rcu_hdr
, data
)->refcount
= 1;
531 out
= kmalloc(HDRLEN_KMALLOC
+ size
, GFP_KERNEL
);
533 out
+= HDRLEN_KMALLOC
;
534 container_of(out
, struct ipc_rcu_hdr
, data
)->is_vmalloc
= 0;
535 container_of(out
, struct ipc_rcu_hdr
, data
)->refcount
= 1;
542 void ipc_rcu_getref(void *ptr
)
544 container_of(ptr
, struct ipc_rcu_hdr
, data
)->refcount
++;
547 static void ipc_do_vfree(struct work_struct
*work
)
549 vfree(container_of(work
, struct ipc_rcu_sched
, work
));
553 * ipc_schedule_free - free ipc + rcu space
554 * @head: RCU callback structure for queued work
556 * Since RCU callback function is called in bh,
557 * we need to defer the vfree to schedule_work().
559 static void ipc_schedule_free(struct rcu_head
*head
)
561 struct ipc_rcu_grace
*grace
=
562 container_of(head
, struct ipc_rcu_grace
, rcu
);
563 struct ipc_rcu_sched
*sched
=
564 container_of(&(grace
->data
[0]), struct ipc_rcu_sched
, data
[0]);
566 INIT_WORK(&sched
->work
, ipc_do_vfree
);
567 schedule_work(&sched
->work
);
571 * ipc_immediate_free - free ipc + rcu space
572 * @head: RCU callback structure that contains pointer to be freed
574 * Free from the RCU callback context.
576 static void ipc_immediate_free(struct rcu_head
*head
)
578 struct ipc_rcu_grace
*free
=
579 container_of(head
, struct ipc_rcu_grace
, rcu
);
583 void ipc_rcu_putref(void *ptr
)
585 if (--container_of(ptr
, struct ipc_rcu_hdr
, data
)->refcount
> 0)
588 if (container_of(ptr
, struct ipc_rcu_hdr
, data
)->is_vmalloc
) {
589 call_rcu(&container_of(ptr
, struct ipc_rcu_grace
, data
)->rcu
,
592 call_rcu(&container_of(ptr
, struct ipc_rcu_grace
, data
)->rcu
,
598 * ipcperms - check IPC permissions
599 * @ipcp: IPC permission set
600 * @flag: desired permission set.
602 * Check user, group, other permissions for access
603 * to ipc resources. return 0 if allowed
606 int ipcperms (struct kern_ipc_perm
*ipcp
, short flag
)
607 { /* flag will most probably be 0 or S_...UGO from <linux/stat.h> */
608 int requested_mode
, granted_mode
, err
;
610 if (unlikely((err
= audit_ipc_obj(ipcp
))))
612 requested_mode
= (flag
>> 6) | (flag
>> 3) | flag
;
613 granted_mode
= ipcp
->mode
;
614 if (current
->euid
== ipcp
->cuid
|| current
->euid
== ipcp
->uid
)
616 else if (in_group_p(ipcp
->cgid
) || in_group_p(ipcp
->gid
))
618 /* is there some bit set in requested_mode but not in granted_mode? */
619 if ((requested_mode
& ~granted_mode
& 0007) &&
620 !capable(CAP_IPC_OWNER
))
623 return security_ipc_permission(ipcp
, flag
);
627 * Functions to convert between the kern_ipc_perm structure and the
628 * old/new ipc_perm structures
632 * kernel_to_ipc64_perm - convert kernel ipc permissions to user
633 * @in: kernel permissions
634 * @out: new style IPC permissions
636 * Turn the kernel object @in into a set of permissions descriptions
637 * for returning to userspace (@out).
641 void kernel_to_ipc64_perm (struct kern_ipc_perm
*in
, struct ipc64_perm
*out
)
646 out
->cuid
= in
->cuid
;
647 out
->cgid
= in
->cgid
;
648 out
->mode
= in
->mode
;
653 * ipc64_perm_to_ipc_perm - convert old ipc permissions to new
654 * @in: new style IPC permissions
655 * @out: old style IPC permissions
657 * Turn the new style permissions object @in into a compatibility
658 * object and store it into the @out pointer.
661 void ipc64_perm_to_ipc_perm (struct ipc64_perm
*in
, struct ipc_perm
*out
)
664 SET_UID(out
->uid
, in
->uid
);
665 SET_GID(out
->gid
, in
->gid
);
666 SET_UID(out
->cuid
, in
->cuid
);
667 SET_GID(out
->cgid
, in
->cgid
);
668 out
->mode
= in
->mode
;
672 struct kern_ipc_perm
*ipc_lock(struct ipc_ids
*ids
, int id
)
674 struct kern_ipc_perm
*out
;
675 int lid
= ipcid_to_idx(id
);
678 out
= idr_find(&ids
->ipcs_idr
, lid
);
681 return ERR_PTR(-EINVAL
);
684 spin_lock(&out
->lock
);
686 /* ipc_rmid() may have already freed the ID while ipc_lock
687 * was spinning: here verify that the structure is still valid
690 spin_unlock(&out
->lock
);
692 return ERR_PTR(-EINVAL
);
698 int ipc_buildid(struct ipc_ids
* ids
, int id
, int seq
)
700 return SEQ_MULTIPLIER
*seq
+ id
;
703 #ifdef __ARCH_WANT_IPC_PARSE_VERSION
707 * ipc_parse_version - IPC call version
708 * @cmd: pointer to command
710 * Return IPC_64 for new style IPC and IPC_OLD for old style IPC.
711 * The @cmd value is turned from an encoding command and version into
712 * just the command code.
715 int ipc_parse_version (int *cmd
)
725 #endif /* __ARCH_WANT_IPC_PARSE_VERSION */
727 #ifdef CONFIG_PROC_FS
728 struct ipc_proc_iter
{
729 struct ipc_namespace
*ns
;
730 struct ipc_proc_iface
*iface
;
734 * This routine locks the ipc structure found at least at position pos.
736 struct kern_ipc_perm
*sysvipc_find_ipc(struct ipc_ids
*ids
, loff_t pos
,
739 struct kern_ipc_perm
*ipc
;
743 for (id
= 0; id
< pos
&& total
< ids
->in_use
; id
++) {
744 ipc
= idr_find(&ids
->ipcs_idr
, id
);
749 if (total
>= ids
->in_use
)
752 for ( ; pos
< IPCMNI
; pos
++) {
753 ipc
= idr_find(&ids
->ipcs_idr
, pos
);
756 ipc_lock_by_ptr(ipc
);
761 /* Out of range - return NULL to terminate iteration */
765 static void *sysvipc_proc_next(struct seq_file
*s
, void *it
, loff_t
*pos
)
767 struct ipc_proc_iter
*iter
= s
->private;
768 struct ipc_proc_iface
*iface
= iter
->iface
;
769 struct kern_ipc_perm
*ipc
= it
;
771 /* If we had an ipc id locked before, unlock it */
772 if (ipc
&& ipc
!= SEQ_START_TOKEN
)
775 return sysvipc_find_ipc(iter
->ns
->ids
[iface
->ids
], *pos
, pos
);
779 * File positions: pos 0 -> header, pos n -> ipc id + 1.
780 * SeqFile iterator: iterator value locked shp or SEQ_TOKEN_START.
782 static void *sysvipc_proc_start(struct seq_file
*s
, loff_t
*pos
)
784 struct ipc_proc_iter
*iter
= s
->private;
785 struct ipc_proc_iface
*iface
= iter
->iface
;
788 ids
= iter
->ns
->ids
[iface
->ids
];
791 * Take the lock - this will be released by the corresponding
794 mutex_lock(&ids
->mutex
);
796 /* pos < 0 is invalid */
800 /* pos == 0 means header */
802 return SEQ_START_TOKEN
;
804 /* Find the (pos-1)th ipc */
805 return sysvipc_find_ipc(ids
, *pos
- 1, pos
);
808 static void sysvipc_proc_stop(struct seq_file
*s
, void *it
)
810 struct kern_ipc_perm
*ipc
= it
;
811 struct ipc_proc_iter
*iter
= s
->private;
812 struct ipc_proc_iface
*iface
= iter
->iface
;
815 /* If we had a locked segment, release it */
816 if (ipc
&& ipc
!= SEQ_START_TOKEN
)
819 ids
= iter
->ns
->ids
[iface
->ids
];
820 /* Release the lock we took in start() */
821 mutex_unlock(&ids
->mutex
);
824 static int sysvipc_proc_show(struct seq_file
*s
, void *it
)
826 struct ipc_proc_iter
*iter
= s
->private;
827 struct ipc_proc_iface
*iface
= iter
->iface
;
829 if (it
== SEQ_START_TOKEN
)
830 return seq_puts(s
, iface
->header
);
832 return iface
->show(s
, it
);
835 static struct seq_operations sysvipc_proc_seqops
= {
836 .start
= sysvipc_proc_start
,
837 .stop
= sysvipc_proc_stop
,
838 .next
= sysvipc_proc_next
,
839 .show
= sysvipc_proc_show
,
842 static int sysvipc_proc_open(struct inode
*inode
, struct file
*file
)
845 struct seq_file
*seq
;
846 struct ipc_proc_iter
*iter
;
849 iter
= kmalloc(sizeof(*iter
), GFP_KERNEL
);
853 ret
= seq_open(file
, &sysvipc_proc_seqops
);
857 seq
= file
->private_data
;
860 iter
->iface
= PDE(inode
)->data
;
861 iter
->ns
= get_ipc_ns(current
->nsproxy
->ipc_ns
);
869 static int sysvipc_proc_release(struct inode
*inode
, struct file
*file
)
871 struct seq_file
*seq
= file
->private_data
;
872 struct ipc_proc_iter
*iter
= seq
->private;
873 put_ipc_ns(iter
->ns
);
874 return seq_release_private(inode
, file
);
877 static const struct file_operations sysvipc_proc_fops
= {
878 .open
= sysvipc_proc_open
,
881 .release
= sysvipc_proc_release
,
883 #endif /* CONFIG_PROC_FS */