1 // SPDX-License-Identifier: GPL-2.0
4 * Copyright (C) 1992 Krishna Balasubramanian
6 * Removed all the remaining kerneld mess
7 * Catch the -EFAULT stuff properly
8 * Use GFP_KERNEL for messages as in 1.2
9 * Fixed up the unchecked user space derefs
10 * Copyright (C) 1998 Alan Cox & Andi Kleen
12 * /proc/sysvipc/msg support (c) 1999 Dragos Acostachioaie <dragos@iname.com>
14 * mostly rewritten, threaded and wake-one semantics added
15 * MSGMAX limit removed, sysctl's added
16 * (c) 1999 Manfred Spraul <manfred@colorfullife.com>
18 * support for audit of ipc object properties and permission changes
19 * Dustin Kirkland <dustin.kirkland@us.ibm.com>
23 * Pavel Emelianov <xemul@openvz.org>
26 #include <linux/capability.h>
27 #include <linux/msg.h>
28 #include <linux/spinlock.h>
29 #include <linux/init.h>
31 #include <linux/proc_fs.h>
32 #include <linux/list.h>
33 #include <linux/security.h>
34 #include <linux/sched/wake_q.h>
35 #include <linux/syscalls.h>
36 #include <linux/audit.h>
37 #include <linux/seq_file.h>
38 #include <linux/rwsem.h>
39 #include <linux/nsproxy.h>
40 #include <linux/ipc_namespace.h>
42 #include <asm/current.h>
43 #include <linux/uaccess.h>
46 /* one msg_receiver structure for each sleeping receiver */
48 struct list_head r_list
;
49 struct task_struct
*r_tsk
;
55 struct msg_msg
*r_msg
;
58 /* one msg_sender for each sleeping sender */
60 struct list_head list
;
61 struct task_struct
*tsk
;
66 #define SEARCH_EQUAL 2
67 #define SEARCH_NOTEQUAL 3
68 #define SEARCH_LESSEQUAL 4
69 #define SEARCH_NUMBER 5
71 #define msg_ids(ns) ((ns)->ids[IPC_MSG_IDS])
73 static inline struct msg_queue
*msq_obtain_object(struct ipc_namespace
*ns
, int id
)
75 struct kern_ipc_perm
*ipcp
= ipc_obtain_object_idr(&msg_ids(ns
), id
);
78 return ERR_CAST(ipcp
);
80 return container_of(ipcp
, struct msg_queue
, q_perm
);
83 static inline struct msg_queue
*msq_obtain_object_check(struct ipc_namespace
*ns
,
86 struct kern_ipc_perm
*ipcp
= ipc_obtain_object_check(&msg_ids(ns
), id
);
89 return ERR_CAST(ipcp
);
91 return container_of(ipcp
, struct msg_queue
, q_perm
);
94 static inline void msg_rmid(struct ipc_namespace
*ns
, struct msg_queue
*s
)
96 ipc_rmid(&msg_ids(ns
), &s
->q_perm
);
99 static void msg_rcu_free(struct rcu_head
*head
)
101 struct kern_ipc_perm
*p
= container_of(head
, struct kern_ipc_perm
, rcu
);
102 struct msg_queue
*msq
= container_of(p
, struct msg_queue
, q_perm
);
104 security_msg_queue_free(&msq
->q_perm
);
109 * newque - Create a new msg queue
111 * @params: ptr to the structure that contains the key and msgflg
113 * Called with msg_ids.rwsem held (writer)
115 static int newque(struct ipc_namespace
*ns
, struct ipc_params
*params
)
117 struct msg_queue
*msq
;
119 key_t key
= params
->key
;
120 int msgflg
= params
->flg
;
122 msq
= kvmalloc(sizeof(*msq
), GFP_KERNEL
);
126 msq
->q_perm
.mode
= msgflg
& S_IRWXUGO
;
127 msq
->q_perm
.key
= key
;
129 msq
->q_perm
.security
= NULL
;
130 retval
= security_msg_queue_alloc(&msq
->q_perm
);
136 msq
->q_stime
= msq
->q_rtime
= 0;
137 msq
->q_ctime
= ktime_get_real_seconds();
138 msq
->q_cbytes
= msq
->q_qnum
= 0;
139 msq
->q_qbytes
= ns
->msg_ctlmnb
;
140 msq
->q_lspid
= msq
->q_lrpid
= 0;
141 INIT_LIST_HEAD(&msq
->q_messages
);
142 INIT_LIST_HEAD(&msq
->q_receivers
);
143 INIT_LIST_HEAD(&msq
->q_senders
);
145 /* ipc_addid() locks msq upon success. */
146 retval
= ipc_addid(&msg_ids(ns
), &msq
->q_perm
, ns
->msg_ctlmni
);
148 call_rcu(&msq
->q_perm
.rcu
, msg_rcu_free
);
152 ipc_unlock_object(&msq
->q_perm
);
155 return msq
->q_perm
.id
;
158 static inline bool msg_fits_inqueue(struct msg_queue
*msq
, size_t msgsz
)
160 return msgsz
+ msq
->q_cbytes
<= msq
->q_qbytes
&&
161 1 + msq
->q_qnum
<= msq
->q_qbytes
;
164 static inline void ss_add(struct msg_queue
*msq
,
165 struct msg_sender
*mss
, size_t msgsz
)
169 __set_current_state(TASK_INTERRUPTIBLE
);
170 list_add_tail(&mss
->list
, &msq
->q_senders
);
173 static inline void ss_del(struct msg_sender
*mss
)
176 list_del(&mss
->list
);
179 static void ss_wakeup(struct msg_queue
*msq
,
180 struct wake_q_head
*wake_q
, bool kill
)
182 struct msg_sender
*mss
, *t
;
183 struct task_struct
*stop_tsk
= NULL
;
184 struct list_head
*h
= &msq
->q_senders
;
186 list_for_each_entry_safe(mss
, t
, h
, list
) {
188 mss
->list
.next
= NULL
;
191 * Stop at the first task we don't wakeup,
192 * we've already iterated the original
195 else if (stop_tsk
== mss
->tsk
)
198 * We are not in an EIDRM scenario here, therefore
199 * verify that we really need to wakeup the task.
200 * To maintain current semantics and wakeup order,
201 * move the sender to the tail on behalf of the
204 else if (!msg_fits_inqueue(msq
, mss
->msgsz
)) {
208 list_move_tail(&mss
->list
, &msq
->q_senders
);
212 wake_q_add(wake_q
, mss
->tsk
);
216 static void expunge_all(struct msg_queue
*msq
, int res
,
217 struct wake_q_head
*wake_q
)
219 struct msg_receiver
*msr
, *t
;
221 list_for_each_entry_safe(msr
, t
, &msq
->q_receivers
, r_list
) {
222 wake_q_add(wake_q
, msr
->r_tsk
);
223 WRITE_ONCE(msr
->r_msg
, ERR_PTR(res
));
228 * freeque() wakes up waiters on the sender and receiver waiting queue,
229 * removes the message queue from message queue ID IDR, and cleans up all the
230 * messages associated with this queue.
232 * msg_ids.rwsem (writer) and the spinlock for this message queue are held
233 * before freeque() is called. msg_ids.rwsem remains locked on exit.
235 static void freeque(struct ipc_namespace
*ns
, struct kern_ipc_perm
*ipcp
)
237 struct msg_msg
*msg
, *t
;
238 struct msg_queue
*msq
= container_of(ipcp
, struct msg_queue
, q_perm
);
239 DEFINE_WAKE_Q(wake_q
);
241 expunge_all(msq
, -EIDRM
, &wake_q
);
242 ss_wakeup(msq
, &wake_q
, true);
244 ipc_unlock_object(&msq
->q_perm
);
248 list_for_each_entry_safe(msg
, t
, &msq
->q_messages
, m_list
) {
249 atomic_dec(&ns
->msg_hdrs
);
252 atomic_sub(msq
->q_cbytes
, &ns
->msg_bytes
);
253 ipc_rcu_putref(&msq
->q_perm
, msg_rcu_free
);
257 * Called with msg_ids.rwsem and ipcp locked.
259 static inline int msg_security(struct kern_ipc_perm
*ipcp
, int msgflg
)
261 return security_msg_queue_associate(ipcp
, msgflg
);
264 SYSCALL_DEFINE2(msgget
, key_t
, key
, int, msgflg
)
266 struct ipc_namespace
*ns
;
267 static const struct ipc_ops msg_ops
= {
269 .associate
= msg_security
,
271 struct ipc_params msg_params
;
273 ns
= current
->nsproxy
->ipc_ns
;
275 msg_params
.key
= key
;
276 msg_params
.flg
= msgflg
;
278 return ipcget(ns
, &msg_ids(ns
), &msg_ops
, &msg_params
);
281 static inline unsigned long
282 copy_msqid_to_user(void __user
*buf
, struct msqid64_ds
*in
, int version
)
286 return copy_to_user(buf
, in
, sizeof(*in
));
291 memset(&out
, 0, sizeof(out
));
293 ipc64_perm_to_ipc_perm(&in
->msg_perm
, &out
.msg_perm
);
295 out
.msg_stime
= in
->msg_stime
;
296 out
.msg_rtime
= in
->msg_rtime
;
297 out
.msg_ctime
= in
->msg_ctime
;
299 if (in
->msg_cbytes
> USHRT_MAX
)
300 out
.msg_cbytes
= USHRT_MAX
;
302 out
.msg_cbytes
= in
->msg_cbytes
;
303 out
.msg_lcbytes
= in
->msg_cbytes
;
305 if (in
->msg_qnum
> USHRT_MAX
)
306 out
.msg_qnum
= USHRT_MAX
;
308 out
.msg_qnum
= in
->msg_qnum
;
310 if (in
->msg_qbytes
> USHRT_MAX
)
311 out
.msg_qbytes
= USHRT_MAX
;
313 out
.msg_qbytes
= in
->msg_qbytes
;
314 out
.msg_lqbytes
= in
->msg_qbytes
;
316 out
.msg_lspid
= in
->msg_lspid
;
317 out
.msg_lrpid
= in
->msg_lrpid
;
319 return copy_to_user(buf
, &out
, sizeof(out
));
326 static inline unsigned long
327 copy_msqid_from_user(struct msqid64_ds
*out
, void __user
*buf
, int version
)
331 if (copy_from_user(out
, buf
, sizeof(*out
)))
336 struct msqid_ds tbuf_old
;
338 if (copy_from_user(&tbuf_old
, buf
, sizeof(tbuf_old
)))
341 out
->msg_perm
.uid
= tbuf_old
.msg_perm
.uid
;
342 out
->msg_perm
.gid
= tbuf_old
.msg_perm
.gid
;
343 out
->msg_perm
.mode
= tbuf_old
.msg_perm
.mode
;
345 if (tbuf_old
.msg_qbytes
== 0)
346 out
->msg_qbytes
= tbuf_old
.msg_lqbytes
;
348 out
->msg_qbytes
= tbuf_old
.msg_qbytes
;
358 * This function handles some msgctl commands which require the rwsem
359 * to be held in write mode.
360 * NOTE: no locks must be held, the rwsem is taken inside this function.
362 static int msgctl_down(struct ipc_namespace
*ns
, int msqid
, int cmd
,
363 struct msqid64_ds
*msqid64
)
365 struct kern_ipc_perm
*ipcp
;
366 struct msg_queue
*msq
;
369 down_write(&msg_ids(ns
).rwsem
);
372 ipcp
= ipcctl_pre_down_nolock(ns
, &msg_ids(ns
), msqid
, cmd
,
373 &msqid64
->msg_perm
, msqid64
->msg_qbytes
);
379 msq
= container_of(ipcp
, struct msg_queue
, q_perm
);
381 err
= security_msg_queue_msgctl(&msq
->q_perm
, cmd
);
387 ipc_lock_object(&msq
->q_perm
);
388 /* freeque unlocks the ipc object and rcu */
393 DEFINE_WAKE_Q(wake_q
);
395 if (msqid64
->msg_qbytes
> ns
->msg_ctlmnb
&&
396 !capable(CAP_SYS_RESOURCE
)) {
401 ipc_lock_object(&msq
->q_perm
);
402 err
= ipc_update_perm(&msqid64
->msg_perm
, ipcp
);
406 msq
->q_qbytes
= msqid64
->msg_qbytes
;
408 msq
->q_ctime
= ktime_get_real_seconds();
410 * Sleeping receivers might be excluded by
411 * stricter permissions.
413 expunge_all(msq
, -EAGAIN
, &wake_q
);
415 * Sleeping senders might be able to send
416 * due to a larger queue size.
418 ss_wakeup(msq
, &wake_q
, false);
419 ipc_unlock_object(&msq
->q_perm
);
430 ipc_unlock_object(&msq
->q_perm
);
434 up_write(&msg_ids(ns
).rwsem
);
438 static int msgctl_info(struct ipc_namespace
*ns
, int msqid
,
439 int cmd
, struct msginfo
*msginfo
)
445 * We must not return kernel stack data.
446 * due to padding, it's not enough
447 * to set all member fields.
449 err
= security_msg_queue_msgctl(NULL
, cmd
);
453 memset(msginfo
, 0, sizeof(*msginfo
));
454 msginfo
->msgmni
= ns
->msg_ctlmni
;
455 msginfo
->msgmax
= ns
->msg_ctlmax
;
456 msginfo
->msgmnb
= ns
->msg_ctlmnb
;
457 msginfo
->msgssz
= MSGSSZ
;
458 msginfo
->msgseg
= MSGSEG
;
459 down_read(&msg_ids(ns
).rwsem
);
460 if (cmd
== MSG_INFO
) {
461 msginfo
->msgpool
= msg_ids(ns
).in_use
;
462 msginfo
->msgmap
= atomic_read(&ns
->msg_hdrs
);
463 msginfo
->msgtql
= atomic_read(&ns
->msg_bytes
);
465 msginfo
->msgmap
= MSGMAP
;
466 msginfo
->msgpool
= MSGPOOL
;
467 msginfo
->msgtql
= MSGTQL
;
469 max_id
= ipc_get_maxid(&msg_ids(ns
));
470 up_read(&msg_ids(ns
).rwsem
);
471 return (max_id
< 0) ? 0 : max_id
;
474 static int msgctl_stat(struct ipc_namespace
*ns
, int msqid
,
475 int cmd
, struct msqid64_ds
*p
)
477 struct msg_queue
*msq
;
481 memset(p
, 0, sizeof(*p
));
484 if (cmd
== MSG_STAT
) {
485 msq
= msq_obtain_object(ns
, msqid
);
492 msq
= msq_obtain_object_check(ns
, msqid
);
500 if (ipcperms(ns
, &msq
->q_perm
, S_IRUGO
))
503 err
= security_msg_queue_msgctl(&msq
->q_perm
, cmd
);
507 ipc_lock_object(&msq
->q_perm
);
509 if (!ipc_valid_object(&msq
->q_perm
)) {
510 ipc_unlock_object(&msq
->q_perm
);
515 kernel_to_ipc64_perm(&msq
->q_perm
, &p
->msg_perm
);
516 p
->msg_stime
= msq
->q_stime
;
517 p
->msg_rtime
= msq
->q_rtime
;
518 p
->msg_ctime
= msq
->q_ctime
;
519 p
->msg_cbytes
= msq
->q_cbytes
;
520 p
->msg_qnum
= msq
->q_qnum
;
521 p
->msg_qbytes
= msq
->q_qbytes
;
522 p
->msg_lspid
= msq
->q_lspid
;
523 p
->msg_lrpid
= msq
->q_lrpid
;
525 ipc_unlock_object(&msq
->q_perm
);
534 SYSCALL_DEFINE3(msgctl
, int, msqid
, int, cmd
, struct msqid_ds __user
*, buf
)
537 struct ipc_namespace
*ns
;
538 struct msqid64_ds msqid64
;
541 if (msqid
< 0 || cmd
< 0)
544 version
= ipc_parse_version(&cmd
);
545 ns
= current
->nsproxy
->ipc_ns
;
550 struct msginfo msginfo
;
551 err
= msgctl_info(ns
, msqid
, cmd
, &msginfo
);
554 if (copy_to_user(buf
, &msginfo
, sizeof(struct msginfo
)))
558 case MSG_STAT
: /* msqid is an index rather than a msg queue id */
560 err
= msgctl_stat(ns
, msqid
, cmd
, &msqid64
);
563 if (copy_msqid_to_user(buf
, &msqid64
, version
))
567 if (copy_msqid_from_user(&msqid64
, buf
, version
))
571 return msgctl_down(ns
, msqid
, cmd
, &msqid64
);
579 struct compat_msqid_ds
{
580 struct compat_ipc_perm msg_perm
;
581 compat_uptr_t msg_first
;
582 compat_uptr_t msg_last
;
583 compat_time_t msg_stime
;
584 compat_time_t msg_rtime
;
585 compat_time_t msg_ctime
;
586 compat_ulong_t msg_lcbytes
;
587 compat_ulong_t msg_lqbytes
;
588 unsigned short msg_cbytes
;
589 unsigned short msg_qnum
;
590 unsigned short msg_qbytes
;
591 compat_ipc_pid_t msg_lspid
;
592 compat_ipc_pid_t msg_lrpid
;
595 static int copy_compat_msqid_from_user(struct msqid64_ds
*out
, void __user
*buf
,
598 memset(out
, 0, sizeof(*out
));
599 if (version
== IPC_64
) {
600 struct compat_msqid64_ds __user
*p
= buf
;
601 if (get_compat_ipc64_perm(&out
->msg_perm
, &p
->msg_perm
))
603 if (get_user(out
->msg_qbytes
, &p
->msg_qbytes
))
606 struct compat_msqid_ds __user
*p
= buf
;
607 if (get_compat_ipc_perm(&out
->msg_perm
, &p
->msg_perm
))
609 if (get_user(out
->msg_qbytes
, &p
->msg_qbytes
))
615 static int copy_compat_msqid_to_user(void __user
*buf
, struct msqid64_ds
*in
,
618 if (version
== IPC_64
) {
619 struct compat_msqid64_ds v
;
620 memset(&v
, 0, sizeof(v
));
621 to_compat_ipc64_perm(&v
.msg_perm
, &in
->msg_perm
);
622 v
.msg_stime
= in
->msg_stime
;
623 v
.msg_rtime
= in
->msg_rtime
;
624 v
.msg_ctime
= in
->msg_ctime
;
625 v
.msg_cbytes
= in
->msg_cbytes
;
626 v
.msg_qnum
= in
->msg_qnum
;
627 v
.msg_qbytes
= in
->msg_qbytes
;
628 v
.msg_lspid
= in
->msg_lspid
;
629 v
.msg_lrpid
= in
->msg_lrpid
;
630 return copy_to_user(buf
, &v
, sizeof(v
));
632 struct compat_msqid_ds v
;
633 memset(&v
, 0, sizeof(v
));
634 to_compat_ipc_perm(&v
.msg_perm
, &in
->msg_perm
);
635 v
.msg_stime
= in
->msg_stime
;
636 v
.msg_rtime
= in
->msg_rtime
;
637 v
.msg_ctime
= in
->msg_ctime
;
638 v
.msg_cbytes
= in
->msg_cbytes
;
639 v
.msg_qnum
= in
->msg_qnum
;
640 v
.msg_qbytes
= in
->msg_qbytes
;
641 v
.msg_lspid
= in
->msg_lspid
;
642 v
.msg_lrpid
= in
->msg_lrpid
;
643 return copy_to_user(buf
, &v
, sizeof(v
));
647 COMPAT_SYSCALL_DEFINE3(msgctl
, int, msqid
, int, cmd
, void __user
*, uptr
)
649 struct ipc_namespace
*ns
;
651 struct msqid64_ds msqid64
;
652 int version
= compat_ipc_parse_version(&cmd
);
654 ns
= current
->nsproxy
->ipc_ns
;
656 if (msqid
< 0 || cmd
< 0)
659 switch (cmd
& (~IPC_64
)) {
662 struct msginfo msginfo
;
663 err
= msgctl_info(ns
, msqid
, cmd
, &msginfo
);
666 if (copy_to_user(uptr
, &msginfo
, sizeof(struct msginfo
)))
672 err
= msgctl_stat(ns
, msqid
, cmd
, &msqid64
);
675 if (copy_compat_msqid_to_user(uptr
, &msqid64
, version
))
679 if (copy_compat_msqid_from_user(&msqid64
, uptr
, version
))
683 return msgctl_down(ns
, msqid
, cmd
, &msqid64
);
690 static int testmsg(struct msg_msg
*msg
, long type
, int mode
)
696 case SEARCH_LESSEQUAL
:
697 if (msg
->m_type
<= type
)
701 if (msg
->m_type
== type
)
704 case SEARCH_NOTEQUAL
:
705 if (msg
->m_type
!= type
)
712 static inline int pipelined_send(struct msg_queue
*msq
, struct msg_msg
*msg
,
713 struct wake_q_head
*wake_q
)
715 struct msg_receiver
*msr
, *t
;
717 list_for_each_entry_safe(msr
, t
, &msq
->q_receivers
, r_list
) {
718 if (testmsg(msg
, msr
->r_msgtype
, msr
->r_mode
) &&
719 !security_msg_queue_msgrcv(&msq
->q_perm
, msg
, msr
->r_tsk
,
720 msr
->r_msgtype
, msr
->r_mode
)) {
722 list_del(&msr
->r_list
);
723 if (msr
->r_maxsize
< msg
->m_ts
) {
724 wake_q_add(wake_q
, msr
->r_tsk
);
725 WRITE_ONCE(msr
->r_msg
, ERR_PTR(-E2BIG
));
727 msq
->q_lrpid
= task_pid_vnr(msr
->r_tsk
);
728 msq
->q_rtime
= get_seconds();
730 wake_q_add(wake_q
, msr
->r_tsk
);
731 WRITE_ONCE(msr
->r_msg
, msg
);
740 static long do_msgsnd(int msqid
, long mtype
, void __user
*mtext
,
741 size_t msgsz
, int msgflg
)
743 struct msg_queue
*msq
;
746 struct ipc_namespace
*ns
;
747 DEFINE_WAKE_Q(wake_q
);
749 ns
= current
->nsproxy
->ipc_ns
;
751 if (msgsz
> ns
->msg_ctlmax
|| (long) msgsz
< 0 || msqid
< 0)
756 msg
= load_msg(mtext
, msgsz
);
764 msq
= msq_obtain_object_check(ns
, msqid
);
770 ipc_lock_object(&msq
->q_perm
);
776 if (ipcperms(ns
, &msq
->q_perm
, S_IWUGO
))
779 /* raced with RMID? */
780 if (!ipc_valid_object(&msq
->q_perm
)) {
785 err
= security_msg_queue_msgsnd(&msq
->q_perm
, msg
, msgflg
);
789 if (msg_fits_inqueue(msq
, msgsz
))
792 /* queue full, wait: */
793 if (msgflg
& IPC_NOWAIT
) {
798 /* enqueue the sender and prepare to block */
799 ss_add(msq
, &s
, msgsz
);
801 if (!ipc_rcu_getref(&msq
->q_perm
)) {
806 ipc_unlock_object(&msq
->q_perm
);
811 ipc_lock_object(&msq
->q_perm
);
813 ipc_rcu_putref(&msq
->q_perm
, msg_rcu_free
);
814 /* raced with RMID? */
815 if (!ipc_valid_object(&msq
->q_perm
)) {
821 if (signal_pending(current
)) {
822 err
= -ERESTARTNOHAND
;
828 msq
->q_lspid
= task_tgid_vnr(current
);
829 msq
->q_stime
= get_seconds();
831 if (!pipelined_send(msq
, msg
, &wake_q
)) {
832 /* no one is waiting for this message, enqueue it */
833 list_add_tail(&msg
->m_list
, &msq
->q_messages
);
834 msq
->q_cbytes
+= msgsz
;
836 atomic_add(msgsz
, &ns
->msg_bytes
);
837 atomic_inc(&ns
->msg_hdrs
);
844 ipc_unlock_object(&msq
->q_perm
);
853 SYSCALL_DEFINE4(msgsnd
, int, msqid
, struct msgbuf __user
*, msgp
, size_t, msgsz
,
858 if (get_user(mtype
, &msgp
->mtype
))
860 return do_msgsnd(msqid
, mtype
, msgp
->mtext
, msgsz
, msgflg
);
865 struct compat_msgbuf
{
870 COMPAT_SYSCALL_DEFINE4(msgsnd
, int, msqid
, compat_uptr_t
, msgp
,
871 compat_ssize_t
, msgsz
, int, msgflg
)
873 struct compat_msgbuf __user
*up
= compat_ptr(msgp
);
876 if (get_user(mtype
, &up
->mtype
))
878 return do_msgsnd(msqid
, mtype
, up
->mtext
, (ssize_t
)msgsz
, msgflg
);
882 static inline int convert_mode(long *msgtyp
, int msgflg
)
884 if (msgflg
& MSG_COPY
)
885 return SEARCH_NUMBER
;
887 * find message of correct type.
888 * msgtyp = 0 => get first.
889 * msgtyp > 0 => get first message of matching type.
890 * msgtyp < 0 => get message with least type must be < abs(msgtype).
895 if (*msgtyp
== LONG_MIN
) /* -LONG_MIN is undefined */
899 return SEARCH_LESSEQUAL
;
901 if (msgflg
& MSG_EXCEPT
)
902 return SEARCH_NOTEQUAL
;
906 static long do_msg_fill(void __user
*dest
, struct msg_msg
*msg
, size_t bufsz
)
908 struct msgbuf __user
*msgp
= dest
;
911 if (put_user(msg
->m_type
, &msgp
->mtype
))
914 msgsz
= (bufsz
> msg
->m_ts
) ? msg
->m_ts
: bufsz
;
915 if (store_msg(msgp
->mtext
, msg
, msgsz
))
920 #ifdef CONFIG_CHECKPOINT_RESTORE
922 * This function creates new kernel message structure, large enough to store
923 * bufsz message bytes.
925 static inline struct msg_msg
*prepare_copy(void __user
*buf
, size_t bufsz
)
927 struct msg_msg
*copy
;
930 * Create dummy message to copy real message to.
932 copy
= load_msg(buf
, bufsz
);
938 static inline void free_copy(struct msg_msg
*copy
)
944 static inline struct msg_msg
*prepare_copy(void __user
*buf
, size_t bufsz
)
946 return ERR_PTR(-ENOSYS
);
949 static inline void free_copy(struct msg_msg
*copy
)
954 static struct msg_msg
*find_msg(struct msg_queue
*msq
, long *msgtyp
, int mode
)
956 struct msg_msg
*msg
, *found
= NULL
;
959 list_for_each_entry(msg
, &msq
->q_messages
, m_list
) {
960 if (testmsg(msg
, *msgtyp
, mode
) &&
961 !security_msg_queue_msgrcv(&msq
->q_perm
, msg
, current
,
963 if (mode
== SEARCH_LESSEQUAL
&& msg
->m_type
!= 1) {
964 *msgtyp
= msg
->m_type
- 1;
966 } else if (mode
== SEARCH_NUMBER
) {
967 if (*msgtyp
== count
)
975 return found
?: ERR_PTR(-EAGAIN
);
978 static long do_msgrcv(int msqid
, void __user
*buf
, size_t bufsz
, long msgtyp
, int msgflg
,
979 long (*msg_handler
)(void __user
*, struct msg_msg
*, size_t))
982 struct msg_queue
*msq
;
983 struct ipc_namespace
*ns
;
984 struct msg_msg
*msg
, *copy
= NULL
;
985 DEFINE_WAKE_Q(wake_q
);
987 ns
= current
->nsproxy
->ipc_ns
;
989 if (msqid
< 0 || (long) bufsz
< 0)
992 if (msgflg
& MSG_COPY
) {
993 if ((msgflg
& MSG_EXCEPT
) || !(msgflg
& IPC_NOWAIT
))
995 copy
= prepare_copy(buf
, min_t(size_t, bufsz
, ns
->msg_ctlmax
));
997 return PTR_ERR(copy
);
999 mode
= convert_mode(&msgtyp
, msgflg
);
1002 msq
= msq_obtain_object_check(ns
, msqid
);
1006 return PTR_ERR(msq
);
1010 struct msg_receiver msr_d
;
1012 msg
= ERR_PTR(-EACCES
);
1013 if (ipcperms(ns
, &msq
->q_perm
, S_IRUGO
))
1016 ipc_lock_object(&msq
->q_perm
);
1018 /* raced with RMID? */
1019 if (!ipc_valid_object(&msq
->q_perm
)) {
1020 msg
= ERR_PTR(-EIDRM
);
1024 msg
= find_msg(msq
, &msgtyp
, mode
);
1027 * Found a suitable message.
1028 * Unlink it from the queue.
1030 if ((bufsz
< msg
->m_ts
) && !(msgflg
& MSG_NOERROR
)) {
1031 msg
= ERR_PTR(-E2BIG
);
1035 * If we are copying, then do not unlink message and do
1036 * not update queue parameters.
1038 if (msgflg
& MSG_COPY
) {
1039 msg
= copy_msg(msg
, copy
);
1043 list_del(&msg
->m_list
);
1045 msq
->q_rtime
= get_seconds();
1046 msq
->q_lrpid
= task_tgid_vnr(current
);
1047 msq
->q_cbytes
-= msg
->m_ts
;
1048 atomic_sub(msg
->m_ts
, &ns
->msg_bytes
);
1049 atomic_dec(&ns
->msg_hdrs
);
1050 ss_wakeup(msq
, &wake_q
, false);
1055 /* No message waiting. Wait for a message */
1056 if (msgflg
& IPC_NOWAIT
) {
1057 msg
= ERR_PTR(-ENOMSG
);
1061 list_add_tail(&msr_d
.r_list
, &msq
->q_receivers
);
1062 msr_d
.r_tsk
= current
;
1063 msr_d
.r_msgtype
= msgtyp
;
1064 msr_d
.r_mode
= mode
;
1065 if (msgflg
& MSG_NOERROR
)
1066 msr_d
.r_maxsize
= INT_MAX
;
1068 msr_d
.r_maxsize
= bufsz
;
1069 msr_d
.r_msg
= ERR_PTR(-EAGAIN
);
1070 __set_current_state(TASK_INTERRUPTIBLE
);
1072 ipc_unlock_object(&msq
->q_perm
);
1077 * Lockless receive, part 1:
1078 * We don't hold a reference to the queue and getting a
1079 * reference would defeat the idea of a lockless operation,
1080 * thus the code relies on rcu to guarantee the existence of
1082 * Prior to destruction, expunge_all(-EIRDM) changes r_msg.
1083 * Thus if r_msg is -EAGAIN, then the queue not yet destroyed.
1088 * Lockless receive, part 2:
1089 * The work in pipelined_send() and expunge_all():
1090 * - Set pointer to message
1091 * - Queue the receiver task for later wakeup
1092 * - Wake up the process after the lock is dropped.
1094 * Should the process wake up before this wakeup (due to a
1095 * signal) it will either see the message and continue ...
1097 msg
= READ_ONCE(msr_d
.r_msg
);
1098 if (msg
!= ERR_PTR(-EAGAIN
))
1102 * ... or see -EAGAIN, acquire the lock to check the message
1105 ipc_lock_object(&msq
->q_perm
);
1108 if (msg
!= ERR_PTR(-EAGAIN
))
1111 list_del(&msr_d
.r_list
);
1112 if (signal_pending(current
)) {
1113 msg
= ERR_PTR(-ERESTARTNOHAND
);
1117 ipc_unlock_object(&msq
->q_perm
);
1121 ipc_unlock_object(&msq
->q_perm
);
1127 return PTR_ERR(msg
);
1130 bufsz
= msg_handler(buf
, msg
, bufsz
);
1136 SYSCALL_DEFINE5(msgrcv
, int, msqid
, struct msgbuf __user
*, msgp
, size_t, msgsz
,
1137 long, msgtyp
, int, msgflg
)
1139 return do_msgrcv(msqid
, msgp
, msgsz
, msgtyp
, msgflg
, do_msg_fill
);
1142 #ifdef CONFIG_COMPAT
1143 static long compat_do_msg_fill(void __user
*dest
, struct msg_msg
*msg
, size_t bufsz
)
1145 struct compat_msgbuf __user
*msgp
= dest
;
1148 if (put_user(msg
->m_type
, &msgp
->mtype
))
1151 msgsz
= (bufsz
> msg
->m_ts
) ? msg
->m_ts
: bufsz
;
1152 if (store_msg(msgp
->mtext
, msg
, msgsz
))
1157 COMPAT_SYSCALL_DEFINE5(msgrcv
, int, msqid
, compat_uptr_t
, msgp
,
1158 compat_ssize_t
, msgsz
, compat_long_t
, msgtyp
, int, msgflg
)
1160 return do_msgrcv(msqid
, compat_ptr(msgp
), (ssize_t
)msgsz
, (long)msgtyp
,
1161 msgflg
, compat_do_msg_fill
);
1165 int msg_init_ns(struct ipc_namespace
*ns
)
1167 ns
->msg_ctlmax
= MSGMAX
;
1168 ns
->msg_ctlmnb
= MSGMNB
;
1169 ns
->msg_ctlmni
= MSGMNI
;
1171 atomic_set(&ns
->msg_bytes
, 0);
1172 atomic_set(&ns
->msg_hdrs
, 0);
1173 return ipc_init_ids(&ns
->ids
[IPC_MSG_IDS
]);
1176 #ifdef CONFIG_IPC_NS
1177 void msg_exit_ns(struct ipc_namespace
*ns
)
1179 free_ipcs(ns
, &msg_ids(ns
), freeque
);
1180 idr_destroy(&ns
->ids
[IPC_MSG_IDS
].ipcs_idr
);
1181 rhashtable_destroy(&ns
->ids
[IPC_MSG_IDS
].key_ht
);
1185 #ifdef CONFIG_PROC_FS
1186 static int sysvipc_msg_proc_show(struct seq_file
*s
, void *it
)
1188 struct user_namespace
*user_ns
= seq_user_ns(s
);
1189 struct kern_ipc_perm
*ipcp
= it
;
1190 struct msg_queue
*msq
= container_of(ipcp
, struct msg_queue
, q_perm
);
1193 "%10d %10d %4o %10lu %10lu %5u %5u %5u %5u %5u %5u %10llu %10llu %10llu\n",
1201 from_kuid_munged(user_ns
, msq
->q_perm
.uid
),
1202 from_kgid_munged(user_ns
, msq
->q_perm
.gid
),
1203 from_kuid_munged(user_ns
, msq
->q_perm
.cuid
),
1204 from_kgid_munged(user_ns
, msq
->q_perm
.cgid
),
1213 int __init
msg_init(void)
1215 const int err
= msg_init_ns(&init_ipc_ns
);
1217 ipc_init_proc_interface("sysvipc/msg",
1218 " key msqid perms cbytes qnum lspid lrpid uid gid cuid cgid stime rtime ctime\n",
1219 IPC_MSG_IDS
, sysvipc_msg_proc_show
);