3 * Copyright (C) 1992 Krishna Balasubramanian
5 * Removed all the remaining kerneld mess
6 * Catch the -EFAULT stuff properly
7 * Use GFP_KERNEL for messages as in 1.2
8 * Fixed up the unchecked user space derefs
9 * Copyright (C) 1998 Alan Cox & Andi Kleen
11 * /proc/sysvipc/msg support (c) 1999 Dragos Acostachioaie <dragos@iname.com>
13 * mostly rewritten, threaded and wake-one semantics added
14 * MSGMAX limit removed, sysctl's added
15 * (c) 1999 Manfred Spraul <manfreds@colorfullife.com>
18 #include <linux/config.h>
19 #include <linux/slab.h>
20 #include <linux/msg.h>
21 #include <linux/spinlock.h>
22 #include <linux/init.h>
23 #include <linux/proc_fs.h>
24 #include <linux/list.h>
25 #include <asm/uaccess.h>
29 int msg_ctlmax = MSGMAX;
30 int msg_ctlmnb = MSGMNB;
31 int msg_ctlmni = MSGMNI;
33 /* one msg_receiver structure for each sleeping receiver */
35 struct list_head r_list;
36 struct task_struct* r_tsk;
42 struct msg_msg* volatile r_msg;
45 /* one msg_sender for each sleeping sender */
47 struct list_head list;
48 struct task_struct* tsk;
52 struct msg_msgseg* next;
53 /* the next part of the message follows immediately */
55 /* one msg_msg structure for each message */
57 struct list_head m_list;
59 int m_ts; /* message text size */
60 struct msg_msgseg* next;
61 /* the actual message follows immediately */
64 #define DATALEN_MSG (PAGE_SIZE-sizeof(struct msg_msg))
65 #define DATALEN_SEG (PAGE_SIZE-sizeof(struct msg_msgseg))
67 /* one msq_queue structure for each present queue on the system */
69 struct kern_ipc_perm q_perm;
70 time_t q_stime; /* last msgsnd time */
71 time_t q_rtime; /* last msgrcv time */
72 time_t q_ctime; /* last change time */
73 unsigned long q_cbytes; /* current number of bytes on queue */
74 unsigned long q_qnum; /* number of messages in queue */
75 unsigned long q_qbytes; /* max number of bytes on queue */
76 pid_t q_lspid; /* pid of last msgsnd */
77 pid_t q_lrpid; /* last receive pid */
79 struct list_head q_messages;
80 struct list_head q_receivers;
81 struct list_head q_senders;
85 #define SEARCH_EQUAL 2
86 #define SEARCH_NOTEQUAL 3
87 #define SEARCH_LESSEQUAL 4
89 static atomic_t msg_bytes = ATOMIC_INIT(0);
90 static atomic_t msg_hdrs = ATOMIC_INIT(0);
92 static struct ipc_ids msg_ids;
94 #define msg_lock(id) ((struct msg_queue*)ipc_lock(&msg_ids,id))
95 #define msg_unlock(id) ipc_unlock(&msg_ids,id)
96 #define msg_rmid(id) ((struct msg_queue*)ipc_rmid(&msg_ids,id))
97 #define msg_checkid(msq, msgid) \
98 ipc_checkid(&msg_ids,&msq->q_perm,msgid)
99 #define msg_buildid(id, seq) \
100 ipc_buildid(&msg_ids, id, seq)
102 static void freeque (int id);
103 static int newque (key_t key, int msgflg);
104 #ifdef CONFIG_PROC_FS
105 static int sysvipc_msg_read_proc(char *buffer, char **start, off_t offset, int length, int *eof, void *data);
108 void __init msg_init (void)
110 ipc_init_ids(&msg_ids,msg_ctlmni);
112 #ifdef CONFIG_PROC_FS
113 create_proc_read_entry("sysvipc/msg", 0, 0, sysvipc_msg_read_proc, NULL);
117 static int newque (key_t key, int msgflg)
120 struct msg_queue *msq;
122 msq = (struct msg_queue *) kmalloc (sizeof (*msq), GFP_KERNEL);
125 id = ipc_addid(&msg_ids, &msq->q_perm, msg_ctlmni);
130 msq->q_perm.mode = (msgflg & S_IRWXUGO);
131 msq->q_perm.key = key;
133 msq->q_stime = msq->q_rtime = 0;
134 msq->q_ctime = CURRENT_TIME;
135 msq->q_cbytes = msq->q_qnum = 0;
136 msq->q_qbytes = msg_ctlmnb;
137 msq->q_lspid = msq->q_lrpid = 0;
138 INIT_LIST_HEAD(&msq->q_messages);
139 INIT_LIST_HEAD(&msq->q_receivers);
140 INIT_LIST_HEAD(&msq->q_senders);
143 return msg_buildid(id,msq->q_perm.seq);
146 static void free_msg(struct msg_msg* msg)
148 struct msg_msgseg* seg;
152 struct msg_msgseg* tmp = seg->next;
158 static struct msg_msg* load_msg(void* src, int len)
161 struct msg_msgseg** pseg;
166 if(alen > DATALEN_MSG)
169 msg = (struct msg_msg *) kmalloc (sizeof(*msg) + alen, GFP_KERNEL);
171 return ERR_PTR(-ENOMEM);
175 if (copy_from_user(msg+1, src, alen)) {
181 src = ((char*)src)+alen;
184 struct msg_msgseg* seg;
186 if(alen > DATALEN_SEG)
188 seg = (struct msg_msgseg *) kmalloc (sizeof(*seg) + alen, GFP_KERNEL);
195 if(copy_from_user (seg+1, src, alen)) {
201 src = ((char*)src)+alen;
210 static int store_msg(void* dest, struct msg_msg* msg, int len)
213 struct msg_msgseg *seg;
216 if(alen > DATALEN_MSG)
218 if(copy_to_user (dest, msg+1, alen))
222 dest = ((char*)dest)+alen;
226 if(alen > DATALEN_SEG)
228 if(copy_to_user (dest, seg+1, alen))
231 dest = ((char*)dest)+alen;
237 static inline void ss_add(struct msg_queue* msq, struct msg_sender* mss)
240 current->state=TASK_INTERRUPTIBLE;
241 list_add_tail(&mss->list,&msq->q_senders);
244 static inline void ss_del(struct msg_sender* mss)
246 if(mss->list.next != NULL)
247 list_del(&mss->list);
250 static void ss_wakeup(struct list_head* h, int kill)
252 struct list_head *tmp;
256 struct msg_sender* mss;
258 mss = list_entry(tmp,struct msg_sender,list);
262 wake_up_process(mss->tsk);
266 static void expunge_all(struct msg_queue* msq, int res)
268 struct list_head *tmp;
270 tmp = msq->q_receivers.next;
271 while (tmp != &msq->q_receivers) {
272 struct msg_receiver* msr;
274 msr = list_entry(tmp,struct msg_receiver,r_list);
276 msr->r_msg = ERR_PTR(res);
277 wake_up_process(msr->r_tsk);
281 static void freeque (int id)
283 struct msg_queue *msq;
284 struct list_head *tmp;
288 expunge_all(msq,-EIDRM);
289 ss_wakeup(&msq->q_senders,1);
292 tmp = msq->q_messages.next;
293 while(tmp != &msq->q_messages) {
294 struct msg_msg* msg = list_entry(tmp,struct msg_msg,m_list);
296 atomic_dec(&msg_hdrs);
299 atomic_sub(msq->q_cbytes, &msg_bytes);
303 asmlinkage long sys_msgget (key_t key, int msgflg)
305 int id, ret = -EPERM;
306 struct msg_queue *msq;
309 if (key == IPC_PRIVATE)
310 ret = newque(key, msgflg);
311 else if ((id = ipc_findkey(&msg_ids, key)) == -1) { /* key not used */
312 if (!(msgflg & IPC_CREAT))
315 ret = newque(key, msgflg);
316 } else if (msgflg & IPC_CREAT && msgflg & IPC_EXCL) {
322 if (ipcperms(&msq->q_perm, msgflg))
325 ret = msg_buildid(id, msq->q_perm.seq);
332 static inline unsigned long copy_msqid_to_user(void *buf, struct msqid64_ds *in, int version)
336 return copy_to_user (buf, in, sizeof(*in));
341 memset(&out,0,sizeof(out));
343 ipc64_perm_to_ipc_perm(&in->msg_perm, &out.msg_perm);
345 out.msg_stime = in->msg_stime;
346 out.msg_rtime = in->msg_rtime;
347 out.msg_ctime = in->msg_ctime;
349 if(in->msg_cbytes > USHRT_MAX)
350 out.msg_cbytes = USHRT_MAX;
352 out.msg_cbytes = in->msg_cbytes;
353 out.msg_lcbytes = in->msg_cbytes;
355 if(in->msg_qnum > USHRT_MAX)
356 out.msg_qnum = USHRT_MAX;
358 out.msg_qnum = in->msg_qnum;
360 if(in->msg_qbytes > USHRT_MAX)
361 out.msg_qbytes = USHRT_MAX;
363 out.msg_qbytes = in->msg_qbytes;
364 out.msg_lqbytes = in->msg_qbytes;
366 out.msg_lspid = in->msg_lspid;
367 out.msg_lrpid = in->msg_lrpid;
369 return copy_to_user (buf, &out, sizeof(out));
377 unsigned long qbytes;
383 static inline unsigned long copy_msqid_from_user(struct msq_setbuf *out, void *buf, int version)
388 struct msqid64_ds tbuf;
390 if (copy_from_user (&tbuf, buf, sizeof (tbuf)))
393 out->qbytes = tbuf.msg_qbytes;
394 out->uid = tbuf.msg_perm.uid;
395 out->gid = tbuf.msg_perm.gid;
396 out->mode = tbuf.msg_perm.mode;
402 struct msqid_ds tbuf_old;
404 if (copy_from_user (&tbuf_old, buf, sizeof (tbuf_old)))
407 out->uid = tbuf_old.msg_perm.uid;
408 out->gid = tbuf_old.msg_perm.gid;
409 out->mode = tbuf_old.msg_perm.mode;
411 if(tbuf_old.msg_qbytes == 0)
412 out->qbytes = tbuf_old.msg_lqbytes;
414 out->qbytes = tbuf_old.msg_qbytes;
423 asmlinkage long sys_msgctl (int msqid, int cmd, struct msqid_ds *buf)
426 struct msg_queue *msq;
427 struct msq_setbuf setbuf;
428 struct kern_ipc_perm *ipcp;
430 if (msqid < 0 || cmd < 0)
433 version = ipc_parse_version(&cmd);
439 struct msginfo msginfo;
443 /* We must not return kernel stack data.
444 * due to padding, it's not enough
445 * to set all member fields.
447 memset(&msginfo,0,sizeof(msginfo));
448 msginfo.msgmni = msg_ctlmni;
449 msginfo.msgmax = msg_ctlmax;
450 msginfo.msgmnb = msg_ctlmnb;
451 msginfo.msgssz = MSGSSZ;
452 msginfo.msgseg = MSGSEG;
454 if (cmd == MSG_INFO) {
455 msginfo.msgpool = msg_ids.in_use;
456 msginfo.msgmap = atomic_read(&msg_hdrs);
457 msginfo.msgtql = atomic_read(&msg_bytes);
459 msginfo.msgmap = MSGMAP;
460 msginfo.msgpool = MSGPOOL;
461 msginfo.msgtql = MSGTQL;
463 max_id = msg_ids.max_id;
465 if (copy_to_user (buf, &msginfo, sizeof(struct msginfo)))
467 return (max_id < 0) ? 0: max_id;
472 struct msqid64_ds tbuf;
476 if(cmd == MSG_STAT && msqid >= msg_ids.size)
479 memset(&tbuf,0,sizeof(tbuf));
481 msq = msg_lock(msqid);
485 if(cmd == MSG_STAT) {
486 success_return = msg_buildid(msqid, msq->q_perm.seq);
489 if (msg_checkid(msq,msqid))
494 if (ipcperms (&msq->q_perm, S_IRUGO))
497 kernel_to_ipc64_perm(&msq->q_perm, &tbuf.msg_perm);
498 tbuf.msg_stime = msq->q_stime;
499 tbuf.msg_rtime = msq->q_rtime;
500 tbuf.msg_ctime = msq->q_ctime;
501 tbuf.msg_cbytes = msq->q_cbytes;
502 tbuf.msg_qnum = msq->q_qnum;
503 tbuf.msg_qbytes = msq->q_qbytes;
504 tbuf.msg_lspid = msq->q_lspid;
505 tbuf.msg_lrpid = msq->q_lrpid;
507 if (copy_msqid_to_user(buf, &tbuf, version))
509 return success_return;
514 if (copy_msqid_from_user (&setbuf, buf, version))
524 msq = msg_lock(msqid);
530 if (msg_checkid(msq,msqid))
534 if (current->euid != ipcp->cuid &&
535 current->euid != ipcp->uid && !capable(CAP_SYS_ADMIN))
536 /* We _could_ check for CAP_CHOWN above, but we don't */
542 if (setbuf.qbytes > msg_ctlmnb && !capable(CAP_SYS_RESOURCE))
544 msq->q_qbytes = setbuf.qbytes;
546 ipcp->uid = setbuf.uid;
547 ipcp->gid = setbuf.gid;
548 ipcp->mode = (ipcp->mode & ~S_IRWXUGO) |
549 (S_IRWXUGO & setbuf.mode);
550 msq->q_ctime = CURRENT_TIME;
551 /* sleeping receivers might be excluded by
552 * stricter permissions.
554 expunge_all(msq,-EAGAIN);
555 /* sleeping senders might be able to send
556 * due to a larger queue size.
558 ss_wakeup(&msq->q_senders,0);
578 static int testmsg(struct msg_msg* msg,long type,int mode)
584 case SEARCH_LESSEQUAL:
585 if(msg->m_type <=type)
589 if(msg->m_type == type)
592 case SEARCH_NOTEQUAL:
593 if(msg->m_type != type)
600 static int inline pipelined_send(struct msg_queue* msq, struct msg_msg* msg)
602 struct list_head* tmp;
604 tmp = msq->q_receivers.next;
605 while (tmp != &msq->q_receivers) {
606 struct msg_receiver* msr;
607 msr = list_entry(tmp,struct msg_receiver,r_list);
609 if(testmsg(msg,msr->r_msgtype,msr->r_mode)) {
610 list_del(&msr->r_list);
611 if(msr->r_maxsize < msg->m_ts) {
612 msr->r_msg = ERR_PTR(-E2BIG);
613 wake_up_process(msr->r_tsk);
616 msq->q_lrpid = msr->r_tsk->pid;
617 msq->q_rtime = CURRENT_TIME;
618 wake_up_process(msr->r_tsk);
626 asmlinkage long sys_msgsnd (int msqid, struct msgbuf *msgp, size_t msgsz, int msgflg)
628 struct msg_queue *msq;
633 if (msgsz > msg_ctlmax || (long) msgsz < 0 || msqid < 0)
635 if (get_user(mtype, &msgp->mtype))
640 msg = load_msg(msgp->mtext, msgsz);
647 msq = msg_lock(msqid);
653 if (msg_checkid(msq,msqid))
654 goto out_unlock_free;
657 if (ipcperms(&msq->q_perm, S_IWUGO))
658 goto out_unlock_free;
660 if(msgsz + msq->q_cbytes > msq->q_qbytes ||
661 1 + msq->q_qnum > msq->q_qbytes) {
664 if(msgflg&IPC_NOWAIT) {
666 goto out_unlock_free;
671 current->state= TASK_RUNNING;
673 msq = msg_lock(msqid);
679 if (signal_pending(current)) {
681 goto out_unlock_free;
686 msq->q_lspid = current->pid;
687 msq->q_stime = CURRENT_TIME;
689 if(!pipelined_send(msq,msg)) {
690 /* noone is waiting for this message, enqueue it */
691 list_add_tail(&msg->m_list,&msq->q_messages);
692 msq->q_cbytes += msgsz;
694 atomic_add(msgsz,&msg_bytes);
695 atomic_inc(&msg_hdrs);
709 static int inline convert_mode(long* msgtyp, int msgflg)
712 * find message of correct type.
713 * msgtyp = 0 => get first.
714 * msgtyp > 0 => get first message of matching type.
715 * msgtyp < 0 => get message with least type must be < abs(msgtype).
721 return SEARCH_LESSEQUAL;
723 if(msgflg & MSG_EXCEPT)
724 return SEARCH_NOTEQUAL;
728 asmlinkage long sys_msgrcv (int msqid, struct msgbuf *msgp, size_t msgsz,
729 long msgtyp, int msgflg)
731 struct msg_queue *msq;
732 struct msg_receiver msr_d;
733 struct list_head* tmp;
734 struct msg_msg* msg, *found_msg;
738 if (msqid < 0 || (long) msgsz < 0)
740 mode = convert_mode(&msgtyp,msgflg);
742 msq = msg_lock(msqid);
747 if (msg_checkid(msq,msqid))
751 if (ipcperms (&msq->q_perm, S_IRUGO))
754 tmp = msq->q_messages.next;
756 while (tmp != &msq->q_messages) {
757 msg = list_entry(tmp,struct msg_msg,m_list);
758 if(testmsg(msg,msgtyp,mode)) {
760 if(mode == SEARCH_LESSEQUAL && msg->m_type != 1) {
762 msgtyp=msg->m_type-1;
772 if ((msgsz < msg->m_ts) && !(msgflg & MSG_NOERROR)) {
776 list_del(&msg->m_list);
778 msq->q_rtime = CURRENT_TIME;
779 msq->q_lrpid = current->pid;
780 msq->q_cbytes -= msg->m_ts;
781 atomic_sub(msg->m_ts,&msg_bytes);
782 atomic_dec(&msg_hdrs);
783 ss_wakeup(&msq->q_senders,0);
786 msgsz = (msgsz > msg->m_ts) ? msg->m_ts : msgsz;
787 if (put_user (msg->m_type, &msgp->mtype) ||
788 store_msg(msgp->mtext, msg, msgsz)) {
796 /* no message waiting. Prepare for pipelined
799 if (msgflg & IPC_NOWAIT) {
803 list_add_tail(&msr_d.r_list,&msq->q_receivers);
804 msr_d.r_tsk = current;
805 msr_d.r_msgtype = msgtyp;
807 if(msgflg & MSG_NOERROR)
808 msr_d.r_maxsize = INT_MAX;
810 msr_d.r_maxsize = msgsz;
811 msr_d.r_msg = ERR_PTR(-EAGAIN);
812 current->state = TASK_INTERRUPTIBLE;
816 current->state = TASK_RUNNING;
818 /* This introduces a race so we must always take
820 msg = (struct msg_msg*) msr_d.r_msg;
827 msg = (struct msg_msg*)msr_d.r_msg;
829 /* our message arived while we waited for
830 * the spinlock. Process it.
840 list_del(&msr_d.r_list);
841 if (signal_pending(current))
853 #ifdef CONFIG_PROC_FS
854 static int sysvipc_msg_read_proc(char *buffer, char **start, off_t offset, int length, int *eof, void *data)
861 len += sprintf(buffer, " key msqid perms cbytes qnum lspid lrpid uid gid cuid cgid stime rtime ctime\n");
863 for(i = 0; i <= msg_ids.max_id; i++) {
864 struct msg_queue * msq;
867 len += sprintf(buffer + len, "%10d %10d %4o %10lu %10lu %5u %5u %5u %5u %5u %5u %10lu %10lu %10lu\n",
869 msg_buildid(i,msq->q_perm.seq),
889 if(pos > offset + length)
897 *start = buffer + (offset - begin);
898 len -= (offset - begin);