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>
37 #include <asm/unistd.h>
41 struct ipc_proc_iface
{
45 int (*show
)(struct seq_file
*, void *);
48 struct ipc_namespace init_ipc_ns
= {
50 .refcount
= ATOMIC_INIT(2),
54 static struct ipc_namespace
*clone_ipc_ns(struct ipc_namespace
*old_ns
)
57 struct ipc_namespace
*ns
;
60 ns
= kmalloc(sizeof(struct ipc_namespace
), GFP_KERNEL
);
64 err
= sem_init_ns(ns
);
67 err
= msg_init_ns(ns
);
70 err
= shm_init_ns(ns
);
87 struct ipc_namespace
*copy_ipcs(unsigned long flags
, struct ipc_namespace
*ns
)
89 struct ipc_namespace
*new_ns
;
94 if (!(flags
& CLONE_NEWIPC
))
97 new_ns
= clone_ipc_ns(ns
);
103 void free_ipc_ns(struct kref
*kref
)
105 struct ipc_namespace
*ns
;
107 ns
= container_of(kref
, struct ipc_namespace
, kref
);
115 * ipc_init - initialise IPC subsystem
117 * The various system5 IPC resources (semaphores, messages and shared
118 * memory) are initialised
121 static int __init
ipc_init(void)
128 __initcall(ipc_init
);
131 * ipc_init_ids - initialise IPC identifiers
132 * @ids: Identifier set
134 * Set up the sequence range to use for the ipc identifier range (limited
135 * below IPCMNI) then initialise the ids idr.
138 void ipc_init_ids(struct ipc_ids
*ids
)
140 init_rwsem(&ids
->rw_mutex
);
145 int seq_limit
= INT_MAX
/SEQ_MULTIPLIER
;
146 if(seq_limit
> USHRT_MAX
)
147 ids
->seq_max
= USHRT_MAX
;
149 ids
->seq_max
= seq_limit
;
152 idr_init(&ids
->ipcs_idr
);
155 #ifdef CONFIG_PROC_FS
156 static const struct file_operations sysvipc_proc_fops
;
158 * ipc_init_proc_interface - Create a proc interface for sysipc types using a seq_file interface.
159 * @path: Path in procfs
160 * @header: Banner to be printed at the beginning of the file.
161 * @ids: ipc id table to iterate.
162 * @show: show routine.
164 void __init
ipc_init_proc_interface(const char *path
, const char *header
,
165 int ids
, int (*show
)(struct seq_file
*, void *))
167 struct proc_dir_entry
*pde
;
168 struct ipc_proc_iface
*iface
;
170 iface
= kmalloc(sizeof(*iface
), GFP_KERNEL
);
174 iface
->header
= header
;
178 pde
= create_proc_entry(path
,
179 S_IRUGO
, /* world readable */
180 NULL
/* parent dir */);
183 pde
->proc_fops
= &sysvipc_proc_fops
;
191 * ipc_findkey - find a key in an ipc identifier set
192 * @ids: Identifier set
193 * @key: The key to find
195 * Requires ipc_ids.rw_mutex locked.
196 * Returns the LOCKED pointer to the ipc structure if found or NULL
198 * If key is found ipc points to the owning ipc structure
201 static struct kern_ipc_perm
*ipc_findkey(struct ipc_ids
*ids
, key_t key
)
203 struct kern_ipc_perm
*ipc
;
207 for (total
= 0, next_id
= 0; total
< ids
->in_use
; next_id
++) {
208 ipc
= idr_find(&ids
->ipcs_idr
, next_id
);
213 if (ipc
->key
!= key
) {
218 ipc_lock_by_ptr(ipc
);
226 * ipc_get_maxid - get the last assigned id
227 * @ids: IPC identifier set
229 * Called with ipc_ids.rw_mutex held.
232 int ipc_get_maxid(struct ipc_ids
*ids
)
234 struct kern_ipc_perm
*ipc
;
238 if (ids
->in_use
== 0)
241 if (ids
->in_use
== IPCMNI
)
244 /* Look for the last assigned id */
246 for (id
= 0; id
< IPCMNI
&& total
< ids
->in_use
; id
++) {
247 ipc
= idr_find(&ids
->ipcs_idr
, id
);
257 * ipc_addid - add an IPC identifier
258 * @ids: IPC identifier set
259 * @new: new IPC permission set
260 * @size: limit for the number of used ids
262 * Add an entry 'new' to the IPC ids idr. The permissions object is
263 * initialised and the first free entry is set up and the id assigned
264 * is returned. The 'new' entry is returned in a locked state on success.
265 * On failure the entry is not locked and a negative err-code is returned.
267 * Called with ipc_ids.rw_mutex held as a writer.
270 int ipc_addid(struct ipc_ids
* ids
, struct kern_ipc_perm
* new, int size
)
277 if (ids
->in_use
>= size
)
280 err
= idr_get_new(&ids
->ipcs_idr
, new, &id
);
286 new->cuid
= new->uid
= current
->euid
;
287 new->gid
= new->cgid
= current
->egid
;
289 new->seq
= ids
->seq
++;
290 if(ids
->seq
> ids
->seq_max
)
293 spin_lock_init(&new->lock
);
296 spin_lock(&new->lock
);
301 * ipcget_new - create a new ipc object
303 * @ids: IPC identifer set
304 * @ops: the actual creation routine to call
305 * @params: its parameters
307 * This routine is called by sys_msgget, sys_semget() and sys_shmget()
308 * when the key is IPC_PRIVATE.
310 int ipcget_new(struct ipc_namespace
*ns
, struct ipc_ids
*ids
,
311 struct ipc_ops
*ops
, struct ipc_params
*params
)
315 err
= idr_pre_get(&ids
->ipcs_idr
, GFP_KERNEL
);
320 down_write(&ids
->rw_mutex
);
321 err
= ops
->getnew(ns
, params
);
322 up_write(&ids
->rw_mutex
);
331 * ipc_check_perms - check security and permissions for an IPC
332 * @ipcp: ipc permission set
333 * @ops: the actual security routine to call
334 * @params: its parameters
336 * This routine is called by sys_msgget(), sys_semget() and sys_shmget()
337 * when the key is not IPC_PRIVATE and that key already exists in the
340 * On success, the IPC id is returned.
342 * It is called with ipc_ids.rw_mutex and ipcp->lock held.
344 static int ipc_check_perms(struct kern_ipc_perm
*ipcp
, struct ipc_ops
*ops
,
345 struct ipc_params
*params
)
349 if (ipcperms(ipcp
, params
->flg
))
352 err
= ops
->associate(ipcp
, params
->flg
);
361 * ipcget_public - get an ipc object or create a new one
363 * @ids: IPC identifer set
364 * @ops: the actual creation routine to call
365 * @params: its parameters
367 * This routine is called by sys_msgget, sys_semget() and sys_shmget()
368 * when the key is not IPC_PRIVATE.
369 * It adds a new entry if the key is not found and does some permission
370 * / security checkings if the key is found.
372 * On success, the ipc id is returned.
374 int ipcget_public(struct ipc_namespace
*ns
, struct ipc_ids
*ids
,
375 struct ipc_ops
*ops
, struct ipc_params
*params
)
377 struct kern_ipc_perm
*ipcp
;
378 int flg
= params
->flg
;
381 err
= idr_pre_get(&ids
->ipcs_idr
, GFP_KERNEL
);
384 * Take the lock as a writer since we are potentially going to add
385 * a new entry + read locks are not "upgradable"
387 down_write(&ids
->rw_mutex
);
388 ipcp
= ipc_findkey(ids
, params
->key
);
391 if (!(flg
& IPC_CREAT
))
396 err
= ops
->getnew(ns
, params
);
398 /* ipc object has been locked by ipc_findkey() */
400 if (flg
& IPC_CREAT
&& flg
& IPC_EXCL
)
404 if (ops
->more_checks
)
405 err
= ops
->more_checks(ipcp
, params
);
408 * ipc_check_perms returns the IPC id on
411 err
= ipc_check_perms(ipcp
, ops
, params
);
415 up_write(&ids
->rw_mutex
);
425 * ipc_rmid - remove an IPC identifier
426 * @ids: IPC identifier set
427 * @ipcp: ipc perm structure containing the identifier to remove
429 * ipc_ids.rw_mutex (as a writer) and the spinlock for this ID are held
430 * before this function is called, and remain locked on the exit.
433 void ipc_rmid(struct ipc_ids
*ids
, struct kern_ipc_perm
*ipcp
)
435 int lid
= ipcid_to_idx(ipcp
->id
);
437 idr_remove(&ids
->ipcs_idr
, lid
);
447 * ipc_alloc - allocate ipc space
448 * @size: size desired
450 * Allocate memory from the appropriate pools and return a pointer to it.
451 * NULL is returned if the allocation fails
454 void* ipc_alloc(int size
)
460 out
= kmalloc(size
, GFP_KERNEL
);
465 * ipc_free - free ipc space
466 * @ptr: pointer returned by ipc_alloc
467 * @size: size of block
469 * Free a block created with ipc_alloc(). The caller must know the size
470 * used in the allocation call.
473 void ipc_free(void* ptr
, int size
)
483 * There are three headers that are prepended to the actual allocation:
484 * - during use: ipc_rcu_hdr.
485 * - during the rcu grace period: ipc_rcu_grace.
486 * - [only if vmalloc]: ipc_rcu_sched.
487 * Their lifetime doesn't overlap, thus the headers share the same memory.
488 * Unlike a normal union, they are right-aligned, thus some container_of
489 * forward/backward casting is necessary:
502 /* "void *" makes sure alignment of following data is sane. */
508 struct work_struct work
;
509 /* "void *" makes sure alignment of following data is sane. */
513 #define HDRLEN_KMALLOC (sizeof(struct ipc_rcu_grace) > sizeof(struct ipc_rcu_hdr) ? \
514 sizeof(struct ipc_rcu_grace) : sizeof(struct ipc_rcu_hdr))
515 #define HDRLEN_VMALLOC (sizeof(struct ipc_rcu_sched) > HDRLEN_KMALLOC ? \
516 sizeof(struct ipc_rcu_sched) : HDRLEN_KMALLOC)
518 static inline int rcu_use_vmalloc(int size
)
520 /* Too big for a single page? */
521 if (HDRLEN_KMALLOC
+ size
> PAGE_SIZE
)
527 * ipc_rcu_alloc - allocate ipc and rcu space
528 * @size: size desired
530 * Allocate memory for the rcu header structure + the object.
531 * Returns the pointer to the object.
532 * NULL is returned if the allocation fails.
535 void* ipc_rcu_alloc(int size
)
539 * We prepend the allocation with the rcu struct, and
540 * workqueue if necessary (for vmalloc).
542 if (rcu_use_vmalloc(size
)) {
543 out
= vmalloc(HDRLEN_VMALLOC
+ size
);
545 out
+= HDRLEN_VMALLOC
;
546 container_of(out
, struct ipc_rcu_hdr
, data
)->is_vmalloc
= 1;
547 container_of(out
, struct ipc_rcu_hdr
, data
)->refcount
= 1;
550 out
= kmalloc(HDRLEN_KMALLOC
+ size
, GFP_KERNEL
);
552 out
+= HDRLEN_KMALLOC
;
553 container_of(out
, struct ipc_rcu_hdr
, data
)->is_vmalloc
= 0;
554 container_of(out
, struct ipc_rcu_hdr
, data
)->refcount
= 1;
561 void ipc_rcu_getref(void *ptr
)
563 container_of(ptr
, struct ipc_rcu_hdr
, data
)->refcount
++;
566 static void ipc_do_vfree(struct work_struct
*work
)
568 vfree(container_of(work
, struct ipc_rcu_sched
, work
));
572 * ipc_schedule_free - free ipc + rcu space
573 * @head: RCU callback structure for queued work
575 * Since RCU callback function is called in bh,
576 * we need to defer the vfree to schedule_work().
578 static void ipc_schedule_free(struct rcu_head
*head
)
580 struct ipc_rcu_grace
*grace
;
581 struct ipc_rcu_sched
*sched
;
583 grace
= container_of(head
, struct ipc_rcu_grace
, rcu
);
584 sched
= container_of(&(grace
->data
[0]), struct ipc_rcu_sched
,
587 INIT_WORK(&sched
->work
, ipc_do_vfree
);
588 schedule_work(&sched
->work
);
592 * ipc_immediate_free - free ipc + rcu space
593 * @head: RCU callback structure that contains pointer to be freed
595 * Free from the RCU callback context.
597 static void ipc_immediate_free(struct rcu_head
*head
)
599 struct ipc_rcu_grace
*free
=
600 container_of(head
, struct ipc_rcu_grace
, rcu
);
604 void ipc_rcu_putref(void *ptr
)
606 if (--container_of(ptr
, struct ipc_rcu_hdr
, data
)->refcount
> 0)
609 if (container_of(ptr
, struct ipc_rcu_hdr
, data
)->is_vmalloc
) {
610 call_rcu(&container_of(ptr
, struct ipc_rcu_grace
, data
)->rcu
,
613 call_rcu(&container_of(ptr
, struct ipc_rcu_grace
, data
)->rcu
,
619 * ipcperms - check IPC permissions
620 * @ipcp: IPC permission set
621 * @flag: desired permission set.
623 * Check user, group, other permissions for access
624 * to ipc resources. return 0 if allowed
627 int ipcperms (struct kern_ipc_perm
*ipcp
, short flag
)
628 { /* flag will most probably be 0 or S_...UGO from <linux/stat.h> */
629 int requested_mode
, granted_mode
, err
;
631 if (unlikely((err
= audit_ipc_obj(ipcp
))))
633 requested_mode
= (flag
>> 6) | (flag
>> 3) | flag
;
634 granted_mode
= ipcp
->mode
;
635 if (current
->euid
== ipcp
->cuid
|| current
->euid
== ipcp
->uid
)
637 else if (in_group_p(ipcp
->cgid
) || in_group_p(ipcp
->gid
))
639 /* is there some bit set in requested_mode but not in granted_mode? */
640 if ((requested_mode
& ~granted_mode
& 0007) &&
641 !capable(CAP_IPC_OWNER
))
644 return security_ipc_permission(ipcp
, flag
);
648 * Functions to convert between the kern_ipc_perm structure and the
649 * old/new ipc_perm structures
653 * kernel_to_ipc64_perm - convert kernel ipc permissions to user
654 * @in: kernel permissions
655 * @out: new style IPC permissions
657 * Turn the kernel object @in into a set of permissions descriptions
658 * for returning to userspace (@out).
662 void kernel_to_ipc64_perm (struct kern_ipc_perm
*in
, struct ipc64_perm
*out
)
667 out
->cuid
= in
->cuid
;
668 out
->cgid
= in
->cgid
;
669 out
->mode
= in
->mode
;
674 * ipc64_perm_to_ipc_perm - convert new ipc permissions to old
675 * @in: new style IPC permissions
676 * @out: old style IPC permissions
678 * Turn the new style permissions object @in into a compatibility
679 * object and store it into the @out pointer.
682 void ipc64_perm_to_ipc_perm (struct ipc64_perm
*in
, struct ipc_perm
*out
)
685 SET_UID(out
->uid
, in
->uid
);
686 SET_GID(out
->gid
, in
->gid
);
687 SET_UID(out
->cuid
, in
->cuid
);
688 SET_GID(out
->cgid
, in
->cgid
);
689 out
->mode
= in
->mode
;
694 * ipc_lock - Lock an ipc structure without rw_mutex held
695 * @ids: IPC identifier set
696 * @id: ipc id to look for
698 * Look for an id in the ipc ids idr and lock the associated ipc object.
700 * The ipc object is locked on exit.
702 * This is the routine that should be called when the rw_mutex is not already
703 * held, i.e. idr tree not protected: it protects the idr tree in read mode
704 * during the idr_find().
707 struct kern_ipc_perm
*ipc_lock(struct ipc_ids
*ids
, int id
)
709 struct kern_ipc_perm
*out
;
710 int lid
= ipcid_to_idx(id
);
712 down_read(&ids
->rw_mutex
);
715 out
= idr_find(&ids
->ipcs_idr
, lid
);
718 up_read(&ids
->rw_mutex
);
719 return ERR_PTR(-EINVAL
);
722 up_read(&ids
->rw_mutex
);
724 spin_lock(&out
->lock
);
726 /* ipc_rmid() may have already freed the ID while ipc_lock
727 * was spinning: here verify that the structure is still valid
730 spin_unlock(&out
->lock
);
732 return ERR_PTR(-EINVAL
);
739 * ipc_lock_down - Lock an ipc structure with rw_sem held
740 * @ids: IPC identifier set
741 * @id: ipc id to look for
743 * Look for an id in the ipc ids idr and lock the associated ipc object.
745 * The ipc object is locked on exit.
747 * This is the routine that should be called when the rw_mutex is already
748 * held, i.e. idr tree protected.
751 struct kern_ipc_perm
*ipc_lock_down(struct ipc_ids
*ids
, int id
)
753 struct kern_ipc_perm
*out
;
754 int lid
= ipcid_to_idx(id
);
757 out
= idr_find(&ids
->ipcs_idr
, lid
);
760 return ERR_PTR(-EINVAL
);
763 spin_lock(&out
->lock
);
766 * No need to verify that the structure is still valid since the
772 #ifdef __ARCH_WANT_IPC_PARSE_VERSION
776 * ipc_parse_version - IPC call version
777 * @cmd: pointer to command
779 * Return IPC_64 for new style IPC and IPC_OLD for old style IPC.
780 * The @cmd value is turned from an encoding command and version into
781 * just the command code.
784 int ipc_parse_version (int *cmd
)
794 #endif /* __ARCH_WANT_IPC_PARSE_VERSION */
796 #ifdef CONFIG_PROC_FS
797 struct ipc_proc_iter
{
798 struct ipc_namespace
*ns
;
799 struct ipc_proc_iface
*iface
;
803 * This routine locks the ipc structure found at least at position pos.
805 struct kern_ipc_perm
*sysvipc_find_ipc(struct ipc_ids
*ids
, loff_t pos
,
808 struct kern_ipc_perm
*ipc
;
812 for (id
= 0; id
< pos
&& total
< ids
->in_use
; id
++) {
813 ipc
= idr_find(&ids
->ipcs_idr
, id
);
818 if (total
>= ids
->in_use
)
821 for ( ; pos
< IPCMNI
; pos
++) {
822 ipc
= idr_find(&ids
->ipcs_idr
, pos
);
825 ipc_lock_by_ptr(ipc
);
830 /* Out of range - return NULL to terminate iteration */
834 static void *sysvipc_proc_next(struct seq_file
*s
, void *it
, loff_t
*pos
)
836 struct ipc_proc_iter
*iter
= s
->private;
837 struct ipc_proc_iface
*iface
= iter
->iface
;
838 struct kern_ipc_perm
*ipc
= it
;
840 /* If we had an ipc id locked before, unlock it */
841 if (ipc
&& ipc
!= SEQ_START_TOKEN
)
844 return sysvipc_find_ipc(iter
->ns
->ids
[iface
->ids
], *pos
, pos
);
848 * File positions: pos 0 -> header, pos n -> ipc id = n - 1.
849 * SeqFile iterator: iterator value locked ipc pointer or SEQ_TOKEN_START.
851 static void *sysvipc_proc_start(struct seq_file
*s
, loff_t
*pos
)
853 struct ipc_proc_iter
*iter
= s
->private;
854 struct ipc_proc_iface
*iface
= iter
->iface
;
857 ids
= iter
->ns
->ids
[iface
->ids
];
860 * Take the lock - this will be released by the corresponding
863 down_read(&ids
->rw_mutex
);
865 /* pos < 0 is invalid */
869 /* pos == 0 means header */
871 return SEQ_START_TOKEN
;
873 /* Find the (pos-1)th ipc */
874 return sysvipc_find_ipc(ids
, *pos
- 1, pos
);
877 static void sysvipc_proc_stop(struct seq_file
*s
, void *it
)
879 struct kern_ipc_perm
*ipc
= it
;
880 struct ipc_proc_iter
*iter
= s
->private;
881 struct ipc_proc_iface
*iface
= iter
->iface
;
884 /* If we had a locked structure, release it */
885 if (ipc
&& ipc
!= SEQ_START_TOKEN
)
888 ids
= iter
->ns
->ids
[iface
->ids
];
889 /* Release the lock we took in start() */
890 up_read(&ids
->rw_mutex
);
893 static int sysvipc_proc_show(struct seq_file
*s
, void *it
)
895 struct ipc_proc_iter
*iter
= s
->private;
896 struct ipc_proc_iface
*iface
= iter
->iface
;
898 if (it
== SEQ_START_TOKEN
)
899 return seq_puts(s
, iface
->header
);
901 return iface
->show(s
, it
);
904 static struct seq_operations sysvipc_proc_seqops
= {
905 .start
= sysvipc_proc_start
,
906 .stop
= sysvipc_proc_stop
,
907 .next
= sysvipc_proc_next
,
908 .show
= sysvipc_proc_show
,
911 static int sysvipc_proc_open(struct inode
*inode
, struct file
*file
)
914 struct seq_file
*seq
;
915 struct ipc_proc_iter
*iter
;
918 iter
= kmalloc(sizeof(*iter
), GFP_KERNEL
);
922 ret
= seq_open(file
, &sysvipc_proc_seqops
);
926 seq
= file
->private_data
;
929 iter
->iface
= PDE(inode
)->data
;
930 iter
->ns
= get_ipc_ns(current
->nsproxy
->ipc_ns
);
938 static int sysvipc_proc_release(struct inode
*inode
, struct file
*file
)
940 struct seq_file
*seq
= file
->private_data
;
941 struct ipc_proc_iter
*iter
= seq
->private;
942 put_ipc_ns(iter
->ns
);
943 return seq_release_private(inode
, file
);
946 static const struct file_operations sysvipc_proc_fops
= {
947 .open
= sysvipc_proc_open
,
950 .release
= sysvipc_proc_release
,
952 #endif /* CONFIG_PROC_FS */