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),
56 * ipc_init - initialise IPC subsystem
58 * The various system5 IPC resources (semaphores, messages and shared
59 * memory) are initialised
62 static int __init
ipc_init(void)
72 * ipc_init_ids - initialise IPC identifiers
73 * @ids: Identifier set
75 * Set up the sequence range to use for the ipc identifier range (limited
76 * below IPCMNI) then initialise the ids idr.
79 void ipc_init_ids(struct ipc_ids
*ids
)
81 init_rwsem(&ids
->rw_mutex
);
86 int seq_limit
= INT_MAX
/SEQ_MULTIPLIER
;
87 if(seq_limit
> USHRT_MAX
)
88 ids
->seq_max
= USHRT_MAX
;
90 ids
->seq_max
= seq_limit
;
93 idr_init(&ids
->ipcs_idr
);
97 static const struct file_operations sysvipc_proc_fops
;
99 * ipc_init_proc_interface - Create a proc interface for sysipc types using a seq_file interface.
100 * @path: Path in procfs
101 * @header: Banner to be printed at the beginning of the file.
102 * @ids: ipc id table to iterate.
103 * @show: show routine.
105 void __init
ipc_init_proc_interface(const char *path
, const char *header
,
106 int ids
, int (*show
)(struct seq_file
*, void *))
108 struct proc_dir_entry
*pde
;
109 struct ipc_proc_iface
*iface
;
111 iface
= kmalloc(sizeof(*iface
), GFP_KERNEL
);
115 iface
->header
= header
;
119 pde
= create_proc_entry(path
,
120 S_IRUGO
, /* world readable */
121 NULL
/* parent dir */);
124 pde
->proc_fops
= &sysvipc_proc_fops
;
132 * ipc_findkey - find a key in an ipc identifier set
133 * @ids: Identifier set
134 * @key: The key to find
136 * Requires ipc_ids.rw_mutex locked.
137 * Returns the LOCKED pointer to the ipc structure if found or NULL
139 * If key is found ipc points to the owning ipc structure
142 static struct kern_ipc_perm
*ipc_findkey(struct ipc_ids
*ids
, key_t key
)
144 struct kern_ipc_perm
*ipc
;
148 for (total
= 0, next_id
= 0; total
< ids
->in_use
; next_id
++) {
149 ipc
= idr_find(&ids
->ipcs_idr
, next_id
);
154 if (ipc
->key
!= key
) {
159 ipc_lock_by_ptr(ipc
);
167 * ipc_get_maxid - get the last assigned id
168 * @ids: IPC identifier set
170 * Called with ipc_ids.rw_mutex held.
173 int ipc_get_maxid(struct ipc_ids
*ids
)
175 struct kern_ipc_perm
*ipc
;
179 if (ids
->in_use
== 0)
182 if (ids
->in_use
== IPCMNI
)
185 /* Look for the last assigned id */
187 for (id
= 0; id
< IPCMNI
&& total
< ids
->in_use
; id
++) {
188 ipc
= idr_find(&ids
->ipcs_idr
, id
);
198 * ipc_addid - add an IPC identifier
199 * @ids: IPC identifier set
200 * @new: new IPC permission set
201 * @size: limit for the number of used ids
203 * Add an entry 'new' to the IPC ids idr. The permissions object is
204 * initialised and the first free entry is set up and the id assigned
205 * is returned. The 'new' entry is returned in a locked state on success.
206 * On failure the entry is not locked and a negative err-code is returned.
208 * Called with ipc_ids.rw_mutex held as a writer.
211 int ipc_addid(struct ipc_ids
* ids
, struct kern_ipc_perm
* new, int size
)
218 if (ids
->in_use
>= size
)
221 err
= idr_get_new(&ids
->ipcs_idr
, new, &id
);
227 new->cuid
= new->uid
= current
->euid
;
228 new->gid
= new->cgid
= current
->egid
;
230 new->seq
= ids
->seq
++;
231 if(ids
->seq
> ids
->seq_max
)
234 spin_lock_init(&new->lock
);
237 spin_lock(&new->lock
);
242 * ipcget_new - create a new ipc object
244 * @ids: IPC identifer set
245 * @ops: the actual creation routine to call
246 * @params: its parameters
248 * This routine is called by sys_msgget, sys_semget() and sys_shmget()
249 * when the key is IPC_PRIVATE.
251 static int ipcget_new(struct ipc_namespace
*ns
, struct ipc_ids
*ids
,
252 struct ipc_ops
*ops
, struct ipc_params
*params
)
256 err
= idr_pre_get(&ids
->ipcs_idr
, GFP_KERNEL
);
261 down_write(&ids
->rw_mutex
);
262 err
= ops
->getnew(ns
, params
);
263 up_write(&ids
->rw_mutex
);
272 * ipc_check_perms - check security and permissions for an IPC
273 * @ipcp: ipc permission set
274 * @ops: the actual security routine to call
275 * @params: its parameters
277 * This routine is called by sys_msgget(), sys_semget() and sys_shmget()
278 * when the key is not IPC_PRIVATE and that key already exists in the
281 * On success, the IPC id is returned.
283 * It is called with ipc_ids.rw_mutex and ipcp->lock held.
285 static int ipc_check_perms(struct kern_ipc_perm
*ipcp
, struct ipc_ops
*ops
,
286 struct ipc_params
*params
)
290 if (ipcperms(ipcp
, params
->flg
))
293 err
= ops
->associate(ipcp
, params
->flg
);
302 * ipcget_public - get an ipc object or create a new one
304 * @ids: IPC identifer set
305 * @ops: the actual creation routine to call
306 * @params: its parameters
308 * This routine is called by sys_msgget, sys_semget() and sys_shmget()
309 * when the key is not IPC_PRIVATE.
310 * It adds a new entry if the key is not found and does some permission
311 * / security checkings if the key is found.
313 * On success, the ipc id is returned.
315 static int ipcget_public(struct ipc_namespace
*ns
, struct ipc_ids
*ids
,
316 struct ipc_ops
*ops
, struct ipc_params
*params
)
318 struct kern_ipc_perm
*ipcp
;
319 int flg
= params
->flg
;
322 err
= idr_pre_get(&ids
->ipcs_idr
, GFP_KERNEL
);
325 * Take the lock as a writer since we are potentially going to add
326 * a new entry + read locks are not "upgradable"
328 down_write(&ids
->rw_mutex
);
329 ipcp
= ipc_findkey(ids
, params
->key
);
332 if (!(flg
& IPC_CREAT
))
337 err
= ops
->getnew(ns
, params
);
339 /* ipc object has been locked by ipc_findkey() */
341 if (flg
& IPC_CREAT
&& flg
& IPC_EXCL
)
345 if (ops
->more_checks
)
346 err
= ops
->more_checks(ipcp
, params
);
349 * ipc_check_perms returns the IPC id on
352 err
= ipc_check_perms(ipcp
, ops
, params
);
356 up_write(&ids
->rw_mutex
);
366 * ipc_rmid - remove an IPC identifier
367 * @ids: IPC identifier set
368 * @ipcp: ipc perm structure containing the identifier to remove
370 * ipc_ids.rw_mutex (as a writer) and the spinlock for this ID are held
371 * before this function is called, and remain locked on the exit.
374 void ipc_rmid(struct ipc_ids
*ids
, struct kern_ipc_perm
*ipcp
)
376 int lid
= ipcid_to_idx(ipcp
->id
);
378 idr_remove(&ids
->ipcs_idr
, lid
);
388 * ipc_alloc - allocate ipc space
389 * @size: size desired
391 * Allocate memory from the appropriate pools and return a pointer to it.
392 * NULL is returned if the allocation fails
395 void* ipc_alloc(int size
)
401 out
= kmalloc(size
, GFP_KERNEL
);
406 * ipc_free - free ipc space
407 * @ptr: pointer returned by ipc_alloc
408 * @size: size of block
410 * Free a block created with ipc_alloc(). The caller must know the size
411 * used in the allocation call.
414 void ipc_free(void* ptr
, int size
)
424 * There are three headers that are prepended to the actual allocation:
425 * - during use: ipc_rcu_hdr.
426 * - during the rcu grace period: ipc_rcu_grace.
427 * - [only if vmalloc]: ipc_rcu_sched.
428 * Their lifetime doesn't overlap, thus the headers share the same memory.
429 * Unlike a normal union, they are right-aligned, thus some container_of
430 * forward/backward casting is necessary:
443 /* "void *" makes sure alignment of following data is sane. */
449 struct work_struct work
;
450 /* "void *" makes sure alignment of following data is sane. */
454 #define HDRLEN_KMALLOC (sizeof(struct ipc_rcu_grace) > sizeof(struct ipc_rcu_hdr) ? \
455 sizeof(struct ipc_rcu_grace) : sizeof(struct ipc_rcu_hdr))
456 #define HDRLEN_VMALLOC (sizeof(struct ipc_rcu_sched) > HDRLEN_KMALLOC ? \
457 sizeof(struct ipc_rcu_sched) : HDRLEN_KMALLOC)
459 static inline int rcu_use_vmalloc(int size
)
461 /* Too big for a single page? */
462 if (HDRLEN_KMALLOC
+ size
> PAGE_SIZE
)
468 * ipc_rcu_alloc - allocate ipc and rcu space
469 * @size: size desired
471 * Allocate memory for the rcu header structure + the object.
472 * Returns the pointer to the object.
473 * NULL is returned if the allocation fails.
476 void* ipc_rcu_alloc(int size
)
480 * We prepend the allocation with the rcu struct, and
481 * workqueue if necessary (for vmalloc).
483 if (rcu_use_vmalloc(size
)) {
484 out
= vmalloc(HDRLEN_VMALLOC
+ size
);
486 out
+= HDRLEN_VMALLOC
;
487 container_of(out
, struct ipc_rcu_hdr
, data
)->is_vmalloc
= 1;
488 container_of(out
, struct ipc_rcu_hdr
, data
)->refcount
= 1;
491 out
= kmalloc(HDRLEN_KMALLOC
+ size
, GFP_KERNEL
);
493 out
+= HDRLEN_KMALLOC
;
494 container_of(out
, struct ipc_rcu_hdr
, data
)->is_vmalloc
= 0;
495 container_of(out
, struct ipc_rcu_hdr
, data
)->refcount
= 1;
502 void ipc_rcu_getref(void *ptr
)
504 container_of(ptr
, struct ipc_rcu_hdr
, data
)->refcount
++;
507 static void ipc_do_vfree(struct work_struct
*work
)
509 vfree(container_of(work
, struct ipc_rcu_sched
, work
));
513 * ipc_schedule_free - free ipc + rcu space
514 * @head: RCU callback structure for queued work
516 * Since RCU callback function is called in bh,
517 * we need to defer the vfree to schedule_work().
519 static void ipc_schedule_free(struct rcu_head
*head
)
521 struct ipc_rcu_grace
*grace
;
522 struct ipc_rcu_sched
*sched
;
524 grace
= container_of(head
, struct ipc_rcu_grace
, rcu
);
525 sched
= container_of(&(grace
->data
[0]), struct ipc_rcu_sched
,
528 INIT_WORK(&sched
->work
, ipc_do_vfree
);
529 schedule_work(&sched
->work
);
533 * ipc_immediate_free - free ipc + rcu space
534 * @head: RCU callback structure that contains pointer to be freed
536 * Free from the RCU callback context.
538 static void ipc_immediate_free(struct rcu_head
*head
)
540 struct ipc_rcu_grace
*free
=
541 container_of(head
, struct ipc_rcu_grace
, rcu
);
545 void ipc_rcu_putref(void *ptr
)
547 if (--container_of(ptr
, struct ipc_rcu_hdr
, data
)->refcount
> 0)
550 if (container_of(ptr
, struct ipc_rcu_hdr
, data
)->is_vmalloc
) {
551 call_rcu(&container_of(ptr
, struct ipc_rcu_grace
, data
)->rcu
,
554 call_rcu(&container_of(ptr
, struct ipc_rcu_grace
, data
)->rcu
,
560 * ipcperms - check IPC permissions
561 * @ipcp: IPC permission set
562 * @flag: desired permission set.
564 * Check user, group, other permissions for access
565 * to ipc resources. return 0 if allowed
568 int ipcperms (struct kern_ipc_perm
*ipcp
, short flag
)
569 { /* flag will most probably be 0 or S_...UGO from <linux/stat.h> */
570 int requested_mode
, granted_mode
, err
;
572 if (unlikely((err
= audit_ipc_obj(ipcp
))))
574 requested_mode
= (flag
>> 6) | (flag
>> 3) | flag
;
575 granted_mode
= ipcp
->mode
;
576 if (current
->euid
== ipcp
->cuid
|| current
->euid
== ipcp
->uid
)
578 else if (in_group_p(ipcp
->cgid
) || in_group_p(ipcp
->gid
))
580 /* is there some bit set in requested_mode but not in granted_mode? */
581 if ((requested_mode
& ~granted_mode
& 0007) &&
582 !capable(CAP_IPC_OWNER
))
585 return security_ipc_permission(ipcp
, flag
);
589 * Functions to convert between the kern_ipc_perm structure and the
590 * old/new ipc_perm structures
594 * kernel_to_ipc64_perm - convert kernel ipc permissions to user
595 * @in: kernel permissions
596 * @out: new style IPC permissions
598 * Turn the kernel object @in into a set of permissions descriptions
599 * for returning to userspace (@out).
603 void kernel_to_ipc64_perm (struct kern_ipc_perm
*in
, struct ipc64_perm
*out
)
608 out
->cuid
= in
->cuid
;
609 out
->cgid
= in
->cgid
;
610 out
->mode
= in
->mode
;
615 * ipc64_perm_to_ipc_perm - convert new ipc permissions to old
616 * @in: new style IPC permissions
617 * @out: old style IPC permissions
619 * Turn the new style permissions object @in into a compatibility
620 * object and store it into the @out pointer.
623 void ipc64_perm_to_ipc_perm (struct ipc64_perm
*in
, struct ipc_perm
*out
)
626 SET_UID(out
->uid
, in
->uid
);
627 SET_GID(out
->gid
, in
->gid
);
628 SET_UID(out
->cuid
, in
->cuid
);
629 SET_GID(out
->cgid
, in
->cgid
);
630 out
->mode
= in
->mode
;
635 * ipc_lock - Lock an ipc structure without rw_mutex held
636 * @ids: IPC identifier set
637 * @id: ipc id to look for
639 * Look for an id in the ipc ids idr and lock the associated ipc object.
641 * The ipc object is locked on exit.
643 * This is the routine that should be called when the rw_mutex is not already
644 * held, i.e. idr tree not protected: it protects the idr tree in read mode
645 * during the idr_find().
648 struct kern_ipc_perm
*ipc_lock(struct ipc_ids
*ids
, int id
)
650 struct kern_ipc_perm
*out
;
651 int lid
= ipcid_to_idx(id
);
653 down_read(&ids
->rw_mutex
);
656 out
= idr_find(&ids
->ipcs_idr
, lid
);
659 up_read(&ids
->rw_mutex
);
660 return ERR_PTR(-EINVAL
);
663 up_read(&ids
->rw_mutex
);
665 spin_lock(&out
->lock
);
667 /* ipc_rmid() may have already freed the ID while ipc_lock
668 * was spinning: here verify that the structure is still valid
671 spin_unlock(&out
->lock
);
673 return ERR_PTR(-EINVAL
);
680 * ipc_lock_down - Lock an ipc structure with rw_sem held
681 * @ids: IPC identifier set
682 * @id: ipc id to look for
684 * Look for an id in the ipc ids idr and lock the associated ipc object.
686 * The ipc object is locked on exit.
688 * This is the routine that should be called when the rw_mutex is already
689 * held, i.e. idr tree protected.
692 struct kern_ipc_perm
*ipc_lock_down(struct ipc_ids
*ids
, int id
)
694 struct kern_ipc_perm
*out
;
695 int lid
= ipcid_to_idx(id
);
698 out
= idr_find(&ids
->ipcs_idr
, lid
);
701 return ERR_PTR(-EINVAL
);
704 spin_lock(&out
->lock
);
707 * No need to verify that the structure is still valid since the
713 struct kern_ipc_perm
*ipc_lock_check_down(struct ipc_ids
*ids
, int id
)
715 struct kern_ipc_perm
*out
;
717 out
= ipc_lock_down(ids
, id
);
721 if (ipc_checkid(out
, id
)) {
723 return ERR_PTR(-EIDRM
);
729 struct kern_ipc_perm
*ipc_lock_check(struct ipc_ids
*ids
, int id
)
731 struct kern_ipc_perm
*out
;
733 out
= ipc_lock(ids
, id
);
737 if (ipc_checkid(out
, id
)) {
739 return ERR_PTR(-EIDRM
);
746 * ipcget - Common sys_*get() code
748 * @ids : IPC identifier set
749 * @ops : operations to be called on ipc object creation, permission checks
751 * @params : the parameters needed by the previous operations.
753 * Common routine called by sys_msgget(), sys_semget() and sys_shmget().
755 int ipcget(struct ipc_namespace
*ns
, struct ipc_ids
*ids
,
756 struct ipc_ops
*ops
, struct ipc_params
*params
)
758 if (params
->key
== IPC_PRIVATE
)
759 return ipcget_new(ns
, ids
, ops
, params
);
761 return ipcget_public(ns
, ids
, ops
, params
);
764 #ifdef __ARCH_WANT_IPC_PARSE_VERSION
768 * ipc_parse_version - IPC call version
769 * @cmd: pointer to command
771 * Return IPC_64 for new style IPC and IPC_OLD for old style IPC.
772 * The @cmd value is turned from an encoding command and version into
773 * just the command code.
776 int ipc_parse_version (int *cmd
)
786 #endif /* __ARCH_WANT_IPC_PARSE_VERSION */
788 #ifdef CONFIG_PROC_FS
789 struct ipc_proc_iter
{
790 struct ipc_namespace
*ns
;
791 struct ipc_proc_iface
*iface
;
795 * This routine locks the ipc structure found at least at position pos.
797 static struct kern_ipc_perm
*sysvipc_find_ipc(struct ipc_ids
*ids
, loff_t pos
,
800 struct kern_ipc_perm
*ipc
;
804 for (id
= 0; id
< pos
&& total
< ids
->in_use
; id
++) {
805 ipc
= idr_find(&ids
->ipcs_idr
, id
);
810 if (total
>= ids
->in_use
)
813 for ( ; pos
< IPCMNI
; pos
++) {
814 ipc
= idr_find(&ids
->ipcs_idr
, pos
);
817 ipc_lock_by_ptr(ipc
);
822 /* Out of range - return NULL to terminate iteration */
826 static void *sysvipc_proc_next(struct seq_file
*s
, void *it
, loff_t
*pos
)
828 struct ipc_proc_iter
*iter
= s
->private;
829 struct ipc_proc_iface
*iface
= iter
->iface
;
830 struct kern_ipc_perm
*ipc
= it
;
832 /* If we had an ipc id locked before, unlock it */
833 if (ipc
&& ipc
!= SEQ_START_TOKEN
)
836 return sysvipc_find_ipc(&iter
->ns
->ids
[iface
->ids
], *pos
, pos
);
840 * File positions: pos 0 -> header, pos n -> ipc id = n - 1.
841 * SeqFile iterator: iterator value locked ipc pointer or SEQ_TOKEN_START.
843 static void *sysvipc_proc_start(struct seq_file
*s
, loff_t
*pos
)
845 struct ipc_proc_iter
*iter
= s
->private;
846 struct ipc_proc_iface
*iface
= iter
->iface
;
849 ids
= &iter
->ns
->ids
[iface
->ids
];
852 * Take the lock - this will be released by the corresponding
855 down_read(&ids
->rw_mutex
);
857 /* pos < 0 is invalid */
861 /* pos == 0 means header */
863 return SEQ_START_TOKEN
;
865 /* Find the (pos-1)th ipc */
866 return sysvipc_find_ipc(ids
, *pos
- 1, pos
);
869 static void sysvipc_proc_stop(struct seq_file
*s
, void *it
)
871 struct kern_ipc_perm
*ipc
= it
;
872 struct ipc_proc_iter
*iter
= s
->private;
873 struct ipc_proc_iface
*iface
= iter
->iface
;
876 /* If we had a locked structure, release it */
877 if (ipc
&& ipc
!= SEQ_START_TOKEN
)
880 ids
= &iter
->ns
->ids
[iface
->ids
];
881 /* Release the lock we took in start() */
882 up_read(&ids
->rw_mutex
);
885 static int sysvipc_proc_show(struct seq_file
*s
, void *it
)
887 struct ipc_proc_iter
*iter
= s
->private;
888 struct ipc_proc_iface
*iface
= iter
->iface
;
890 if (it
== SEQ_START_TOKEN
)
891 return seq_puts(s
, iface
->header
);
893 return iface
->show(s
, it
);
896 static struct seq_operations sysvipc_proc_seqops
= {
897 .start
= sysvipc_proc_start
,
898 .stop
= sysvipc_proc_stop
,
899 .next
= sysvipc_proc_next
,
900 .show
= sysvipc_proc_show
,
903 static int sysvipc_proc_open(struct inode
*inode
, struct file
*file
)
906 struct seq_file
*seq
;
907 struct ipc_proc_iter
*iter
;
910 iter
= kmalloc(sizeof(*iter
), GFP_KERNEL
);
914 ret
= seq_open(file
, &sysvipc_proc_seqops
);
918 seq
= file
->private_data
;
921 iter
->iface
= PDE(inode
)->data
;
922 iter
->ns
= get_ipc_ns(current
->nsproxy
->ipc_ns
);
930 static int sysvipc_proc_release(struct inode
*inode
, struct file
*file
)
932 struct seq_file
*seq
= file
->private_data
;
933 struct ipc_proc_iter
*iter
= seq
->private;
934 put_ipc_ns(iter
->ns
);
935 return seq_release_private(inode
, file
);
938 static const struct file_operations sysvipc_proc_fops
= {
939 .open
= sysvipc_proc_open
,
942 .release
= sysvipc_proc_release
,
944 #endif /* CONFIG_PROC_FS */