1 /*-
2 * SPDX-License-Identifier: BSD-2-Clause-FreeBSD
3 *
4 * Copyright (c) 2005 David Xu <davidxu@freebsd.org>
5 * Copyright (c) 2016-2017 Robert N. M. Watson
6 * All rights reserved.
7 *
8 * Portions of this software were developed by BAE Systems, the University of
9 * Cambridge Computer Laboratory, and Memorial University under DARPA/AFRL
10 * contract FA8650-15-C-7558 ("CADETS"), as part of the DARPA Transparent
11 * Computing (TC) research program.
12 *
13 * Redistribution and use in source and binary forms, with or without
14 * modification, are permitted provided that the following conditions
15 * are met:
16 * 1. Redistributions of source code must retain the above copyright
17 * notice, this list of conditions and the following disclaimer.
18 * 2. Redistributions in binary form must reproduce the above copyright
19 * notice, this list of conditions and the following disclaimer in the
20 * documentation and/or other materials provided with the distribution.
21 *
22 * THIS SOFTWARE IS PROVIDED BY THE AUTHOR AND CONTRIBUTORS ``AS IS'' AND
23 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
24 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
25 * ARE DISCLAIMED. IN NO EVENT SHALL THE AUTHOR OR CONTRIBUTORS BE LIABLE
26 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
27 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
28 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
29 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
30 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
31 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
32 * SUCH DAMAGE.
33 *
34 */
35
36 /*
37 * POSIX message queue implementation.
38 *
39 * 1) A mqueue filesystem can be mounted, each message queue appears
40 * in mounted directory, user can change queue's permission and
41 * ownership, or remove a queue. Manually creating a file in the
42 * directory causes a message queue to be created in the kernel with
43 * default message queue attributes applied and same name used, this
44 * method is not advocated since mq_open syscall allows user to specify
45 * different attributes. Also the file system can be mounted multiple
46 * times at different mount points but shows same contents.
47 *
48 * 2) Standard POSIX message queue API. The syscalls do not use vfs layer,
49 * but directly operate on internal data structure, this allows user to
50 * use the IPC facility without having to mount mqueue file system.
51 */
52
53 #include <sys/cdefs.h>
54 __FBSDID("$FreeBSD$");
55
56 #include "opt_capsicum.h"
57
58 #include <sys/param.h>
59 #include <sys/kernel.h>
60 #include <sys/systm.h>
61 #include <sys/limits.h>
62 #include <sys/malloc.h>
63 #include <sys/buf.h>
64 #include <sys/capsicum.h>
65 #include <sys/dirent.h>
66 #include <sys/event.h>
67 #include <sys/eventhandler.h>
68 #include <sys/fcntl.h>
69 #include <sys/file.h>
70 #include <sys/filedesc.h>
71 #include <sys/jail.h>
72 #include <sys/lock.h>
73 #include <sys/module.h>
74 #include <sys/mount.h>
75 #include <sys/mqueue.h>
76 #include <sys/mutex.h>
77 #include <sys/namei.h>
78 #include <sys/posix4.h>
79 #include <sys/poll.h>
80 #include <sys/priv.h>
81 #include <sys/proc.h>
82 #include <sys/queue.h>
83 #include <sys/sysproto.h>
84 #include <sys/stat.h>
85 #include <sys/syscall.h>
86 #include <sys/syscallsubr.h>
87 #include <sys/sysent.h>
88 #include <sys/sx.h>
89 #include <sys/sysctl.h>
90 #include <sys/taskqueue.h>
91 #include <sys/unistd.h>
92 #include <sys/user.h>
93 #include <sys/vnode.h>
94 #include <machine/atomic.h>
95
96 #include <security/audit/audit.h>
97
98 FEATURE(p1003_1b_mqueue, "POSIX P1003.1B message queues support");
99
100 /*
101 * Limits and constants
102 */
103 #define MQFS_NAMELEN NAME_MAX
104 #define MQFS_DELEN (8 + MQFS_NAMELEN)
105
106 /* node types */
107 typedef enum {
108 mqfstype_none = 0,
109 mqfstype_root,
110 mqfstype_dir,
111 mqfstype_this,
112 mqfstype_parent,
113 mqfstype_file,
114 mqfstype_symlink,
115 } mqfs_type_t;
116
117 struct mqfs_node;
118
119 /*
120 * mqfs_info: describes a mqfs instance
121 */
122 struct mqfs_info {
123 struct sx mi_lock;
124 struct mqfs_node *mi_root;
125 struct unrhdr *mi_unrhdr;
126 };
127
128 struct mqfs_vdata {
129 LIST_ENTRY(mqfs_vdata) mv_link;
130 struct mqfs_node *mv_node;
131 struct vnode *mv_vnode;
132 struct task mv_task;
133 };
134
135 /*
136 * mqfs_node: describes a node (file or directory) within a mqfs
137 */
138 struct mqfs_node {
139 char mn_name[MQFS_NAMELEN+1];
140 struct mqfs_info *mn_info;
141 struct mqfs_node *mn_parent;
142 LIST_HEAD(,mqfs_node) mn_children;
143 LIST_ENTRY(mqfs_node) mn_sibling;
144 LIST_HEAD(,mqfs_vdata) mn_vnodes;
145 const void *mn_pr_root;
146 int mn_refcount;
147 mqfs_type_t mn_type;
148 int mn_deleted;
149 uint32_t mn_fileno;
150 void *mn_data;
151 struct timespec mn_birth;
152 struct timespec mn_ctime;
153 struct timespec mn_atime;
154 struct timespec mn_mtime;
155 uid_t mn_uid;
156 gid_t mn_gid;
157 int mn_mode;
158 };
159
160 #define VTON(vp) (((struct mqfs_vdata *)((vp)->v_data))->mv_node)
161 #define VTOMQ(vp) ((struct mqueue *)(VTON(vp)->mn_data))
162 #define VFSTOMQFS(m) ((struct mqfs_info *)((m)->mnt_data))
163 #define FPTOMQ(fp) ((struct mqueue *)(((struct mqfs_node *) \
164 (fp)->f_data)->mn_data))
165
166 TAILQ_HEAD(msgq, mqueue_msg);
167
168 struct mqueue;
169
170 struct mqueue_notifier {
171 LIST_ENTRY(mqueue_notifier) nt_link;
172 struct sigevent nt_sigev;
173 ksiginfo_t nt_ksi;
174 struct proc *nt_proc;
175 };
176
177 struct mqueue {
178 struct mtx mq_mutex;
179 int mq_flags;
180 long mq_maxmsg;
181 long mq_msgsize;
182 long mq_curmsgs;
183 long mq_totalbytes;
184 struct msgq mq_msgq;
185 int mq_receivers;
186 int mq_senders;
187 struct selinfo mq_rsel;
188 struct selinfo mq_wsel;
189 struct mqueue_notifier *mq_notifier;
190 };
191
192 #define MQ_RSEL 0x01
193 #define MQ_WSEL 0x02
194
195 struct mqueue_msg {
196 TAILQ_ENTRY(mqueue_msg) msg_link;
197 unsigned int msg_prio;
198 unsigned int msg_size;
199 /* following real data... */
200 };
201
202 static SYSCTL_NODE(_kern, OID_AUTO, mqueue, CTLFLAG_RW | CTLFLAG_MPSAFE, 0,
203 "POSIX real time message queue");
204
205 static int default_maxmsg = 10;
206 static int default_msgsize = 1024;
207
208 static int maxmsg = 100;
209 SYSCTL_INT(_kern_mqueue, OID_AUTO, maxmsg, CTLFLAG_RW,
210 &maxmsg, 0, "Default maximum messages in queue");
211 static int maxmsgsize = 16384;
212 SYSCTL_INT(_kern_mqueue, OID_AUTO, maxmsgsize, CTLFLAG_RW,
213 &maxmsgsize, 0, "Default maximum message size");
214 static int maxmq = 100;
215 SYSCTL_INT(_kern_mqueue, OID_AUTO, maxmq, CTLFLAG_RW,
216 &maxmq, 0, "maximum message queues");
217 static int curmq = 0;
218 SYSCTL_INT(_kern_mqueue, OID_AUTO, curmq, CTLFLAG_RW,
219 &curmq, 0, "current message queue number");
220 static int unloadable = 0;
221 static MALLOC_DEFINE(M_MQUEUEDATA, "mqdata", "mqueue data");
222
223 static eventhandler_tag exit_tag;
224
225 /* Only one instance per-system */
226 static struct mqfs_info mqfs_data;
227 static uma_zone_t mqnode_zone;
228 static uma_zone_t mqueue_zone;
229 static uma_zone_t mvdata_zone;
230 static uma_zone_t mqnoti_zone;
231 static struct vop_vector mqfs_vnodeops;
232 static struct fileops mqueueops;
233 static unsigned mqfs_osd_jail_slot;
234
235 /*
236 * Directory structure construction and manipulation
237 */
238 #ifdef notyet
239 static struct mqfs_node *mqfs_create_dir(struct mqfs_node *parent,
240 const char *name, int namelen, struct ucred *cred, int mode);
241 static struct mqfs_node *mqfs_create_link(struct mqfs_node *parent,
242 const char *name, int namelen, struct ucred *cred, int mode);
243 #endif
244
245 static struct mqfs_node *mqfs_create_file(struct mqfs_node *parent,
246 const char *name, int namelen, struct ucred *cred, int mode);
247 static int mqfs_destroy(struct mqfs_node *mn);
248 static void mqfs_fileno_alloc(struct mqfs_info *mi, struct mqfs_node *mn);
249 static void mqfs_fileno_free(struct mqfs_info *mi, struct mqfs_node *mn);
250 static int mqfs_allocv(struct mount *mp, struct vnode **vpp, struct mqfs_node *pn);
251 static int mqfs_prison_remove(void *obj, void *data);
252
253 /*
254 * Message queue construction and maniplation
255 */
256 static struct mqueue *mqueue_alloc(const struct mq_attr *attr);
257 static void mqueue_free(struct mqueue *mq);
258 static int mqueue_send(struct mqueue *mq, const char *msg_ptr,
259 size_t msg_len, unsigned msg_prio, int waitok,
260 const struct timespec *abs_timeout);
261 static int mqueue_receive(struct mqueue *mq, char *msg_ptr,
262 size_t msg_len, unsigned *msg_prio, int waitok,
263 const struct timespec *abs_timeout);
264 static int _mqueue_send(struct mqueue *mq, struct mqueue_msg *msg,
265 int timo);
266 static int _mqueue_recv(struct mqueue *mq, struct mqueue_msg **msg,
267 int timo);
268 static void mqueue_send_notification(struct mqueue *mq);
269 static void mqueue_fdclose(struct thread *td, int fd, struct file *fp);
270 static void mq_proc_exit(void *arg, struct proc *p);
271
272 /*
273 * kqueue filters
274 */
275 static void filt_mqdetach(struct knote *kn);
276 static int filt_mqread(struct knote *kn, long hint);
277 static int filt_mqwrite(struct knote *kn, long hint);
278
279 struct filterops mq_rfiltops = {
280 .f_isfd = 1,
281 .f_detach = filt_mqdetach,
282 .f_event = filt_mqread,
283 };
284 struct filterops mq_wfiltops = {
285 .f_isfd = 1,
286 .f_detach = filt_mqdetach,
287 .f_event = filt_mqwrite,
288 };
289
290 /*
291 * Initialize fileno bitmap
292 */
293 static void
294 mqfs_fileno_init(struct mqfs_info *mi)
295 {
296 struct unrhdr *up;
297
298 up = new_unrhdr(1, INT_MAX, NULL);
299 mi->mi_unrhdr = up;
300 }
301
302 /*
303 * Tear down fileno bitmap
304 */
305 static void
306 mqfs_fileno_uninit(struct mqfs_info *mi)
307 {
308 struct unrhdr *up;
309
310 up = mi->mi_unrhdr;
311 mi->mi_unrhdr = NULL;
312 delete_unrhdr(up);
313 }
314
315 /*
316 * Allocate a file number
317 */
318 static void
319 mqfs_fileno_alloc(struct mqfs_info *mi, struct mqfs_node *mn)
320 {
321 /* make sure our parent has a file number */
322 if (mn->mn_parent && !mn->mn_parent->mn_fileno)
323 mqfs_fileno_alloc(mi, mn->mn_parent);
324
325 switch (mn->mn_type) {
326 case mqfstype_root:
327 case mqfstype_dir:
328 case mqfstype_file:
329 case mqfstype_symlink:
330 mn->mn_fileno = alloc_unr(mi->mi_unrhdr);
331 break;
332 case mqfstype_this:
333 KASSERT(mn->mn_parent != NULL,
334 ("mqfstype_this node has no parent"));
335 mn->mn_fileno = mn->mn_parent->mn_fileno;
336 break;
337 case mqfstype_parent:
338 KASSERT(mn->mn_parent != NULL,
339 ("mqfstype_parent node has no parent"));
340 if (mn->mn_parent == mi->mi_root) {
341 mn->mn_fileno = mn->mn_parent->mn_fileno;
342 break;
343 }
344 KASSERT(mn->mn_parent->mn_parent != NULL,
345 ("mqfstype_parent node has no grandparent"));
346 mn->mn_fileno = mn->mn_parent->mn_parent->mn_fileno;
347 break;
348 default:
349 KASSERT(0,
350 ("mqfs_fileno_alloc() called for unknown type node: %d",
351 mn->mn_type));
352 break;
353 }
354 }
355
356 /*
357 * Release a file number
358 */
359 static void
360 mqfs_fileno_free(struct mqfs_info *mi, struct mqfs_node *mn)
361 {
362 switch (mn->mn_type) {
363 case mqfstype_root:
364 case mqfstype_dir:
365 case mqfstype_file:
366 case mqfstype_symlink:
367 free_unr(mi->mi_unrhdr, mn->mn_fileno);
368 break;
369 case mqfstype_this:
370 case mqfstype_parent:
371 /* ignore these, as they don't "own" their file number */
372 break;
373 default:
374 KASSERT(0,
375 ("mqfs_fileno_free() called for unknown type node: %d",
376 mn->mn_type));
377 break;
378 }
379 }
380
381 static __inline struct mqfs_node *
382 mqnode_alloc(void)
383 {
384 return uma_zalloc(mqnode_zone, M_WAITOK | M_ZERO);
385 }
386
387 static __inline void
388 mqnode_free(struct mqfs_node *node)
389 {
390 uma_zfree(mqnode_zone, node);
391 }
392
393 static __inline void
394 mqnode_addref(struct mqfs_node *node)
395 {
396 atomic_add_int(&node->mn_refcount, 1);
397 }
398
399 static __inline void
400 mqnode_release(struct mqfs_node *node)
401 {
402 struct mqfs_info *mqfs;
403 int old, exp;
404
405 mqfs = node->mn_info;
406 old = atomic_fetchadd_int(&node->mn_refcount, -1);
407 if (node->mn_type == mqfstype_dir ||
408 node->mn_type == mqfstype_root)
409 exp = 3; /* include . and .. */
410 else
411 exp = 1;
412 if (old == exp) {
413 int locked = sx_xlocked(&mqfs->mi_lock);
414 if (!locked)
415 sx_xlock(&mqfs->mi_lock);
416 mqfs_destroy(node);
417 if (!locked)
418 sx_xunlock(&mqfs->mi_lock);
419 }
420 }
421
422 /*
423 * Add a node to a directory
424 */
425 static int
426 mqfs_add_node(struct mqfs_node *parent, struct mqfs_node *node)
427 {
428 KASSERT(parent != NULL, ("%s(): parent is NULL", __func__));
429 KASSERT(parent->mn_info != NULL,
430 ("%s(): parent has no mn_info", __func__));
431 KASSERT(parent->mn_type == mqfstype_dir ||
432 parent->mn_type == mqfstype_root,
433 ("%s(): parent is not a directory", __func__));
434
435 node->mn_info = parent->mn_info;
436 node->mn_parent = parent;
437 LIST_INIT(&node->mn_children);
438 LIST_INIT(&node->mn_vnodes);
439 LIST_INSERT_HEAD(&parent->mn_children, node, mn_sibling);
440 mqnode_addref(parent);
441 return (0);
442 }
443
444 static struct mqfs_node *
445 mqfs_create_node(const char *name, int namelen, struct ucred *cred, int mode,
446 int nodetype)
447 {
448 struct mqfs_node *node;
449
450 node = mqnode_alloc();
451 strncpy(node->mn_name, name, namelen);
452 node->mn_pr_root = cred->cr_prison->pr_root;
453 node->mn_type = nodetype;
454 node->mn_refcount = 1;
455 vfs_timestamp(&node->mn_birth);
456 node->mn_ctime = node->mn_atime = node->mn_mtime
457 = node->mn_birth;
458 node->mn_uid = cred->cr_uid;
459 node->mn_gid = cred->cr_gid;
460 node->mn_mode = mode;
461 return (node);
462 }
463
464 /*
465 * Create a file
466 */
467 static struct mqfs_node *
468 mqfs_create_file(struct mqfs_node *parent, const char *name, int namelen,
469 struct ucred *cred, int mode)
470 {
471 struct mqfs_node *node;
472
473 node = mqfs_create_node(name, namelen, cred, mode, mqfstype_file);
474 if (mqfs_add_node(parent, node) != 0) {
475 mqnode_free(node);
476 return (NULL);
477 }
478 return (node);
479 }
480
481 /*
482 * Add . and .. to a directory
483 */
484 static int
485 mqfs_fixup_dir(struct mqfs_node *parent)
486 {
487 struct mqfs_node *dir;
488
489 dir = mqnode_alloc();
490 dir->mn_name[0] = '.';
491 dir->mn_type = mqfstype_this;
492 dir->mn_refcount = 1;
493 if (mqfs_add_node(parent, dir) != 0) {
494 mqnode_free(dir);
495 return (-1);
496 }
497
498 dir = mqnode_alloc();
499 dir->mn_name[0] = dir->mn_name[1] = '.';
500 dir->mn_type = mqfstype_parent;
501 dir->mn_refcount = 1;
502
503 if (mqfs_add_node(parent, dir) != 0) {
504 mqnode_free(dir);
505 return (-1);
506 }
507
508 return (0);
509 }
510
511 #ifdef notyet
512
513 /*
514 * Create a directory
515 */
516 static struct mqfs_node *
517 mqfs_create_dir(struct mqfs_node *parent, const char *name, int namelen,
518 struct ucred *cred, int mode)
519 {
520 struct mqfs_node *node;
521
522 node = mqfs_create_node(name, namelen, cred, mode, mqfstype_dir);
523 if (mqfs_add_node(parent, node) != 0) {
524 mqnode_free(node);
525 return (NULL);
526 }
527
528 if (mqfs_fixup_dir(node) != 0) {
529 mqfs_destroy(node);
530 return (NULL);
531 }
532 return (node);
533 }
534
535 /*
536 * Create a symlink
537 */
538 static struct mqfs_node *
539 mqfs_create_link(struct mqfs_node *parent, const char *name, int namelen,
540 struct ucred *cred, int mode)
541 {
542 struct mqfs_node *node;
543
544 node = mqfs_create_node(name, namelen, cred, mode, mqfstype_symlink);
545 if (mqfs_add_node(parent, node) != 0) {
546 mqnode_free(node);
547 return (NULL);
548 }
549 return (node);
550 }
551
552 #endif
553
554 /*
555 * Destroy a node or a tree of nodes
556 */
557 static int
558 mqfs_destroy(struct mqfs_node *node)
559 {
560 struct mqfs_node *parent;
561
562 KASSERT(node != NULL,
563 ("%s(): node is NULL", __func__));
564 KASSERT(node->mn_info != NULL,
565 ("%s(): node has no mn_info", __func__));
566
567 /* destroy children */
568 if (node->mn_type == mqfstype_dir || node->mn_type == mqfstype_root)
569 while (! LIST_EMPTY(&node->mn_children))
570 mqfs_destroy(LIST_FIRST(&node->mn_children));
571
572 /* unlink from parent */
573 if ((parent = node->mn_parent) != NULL) {
574 KASSERT(parent->mn_info == node->mn_info,
575 ("%s(): parent has different mn_info", __func__));
576 LIST_REMOVE(node, mn_sibling);
577 }
578
579 if (node->mn_fileno != 0)
580 mqfs_fileno_free(node->mn_info, node);
581 if (node->mn_data != NULL)
582 mqueue_free(node->mn_data);
583 mqnode_free(node);
584 return (0);
585 }
586
587 /*
588 * Mount a mqfs instance
589 */
590 static int
591 mqfs_mount(struct mount *mp)
592 {
593 struct statfs *sbp;
594
595 if (mp->mnt_flag & MNT_UPDATE)
596 return (EOPNOTSUPP);
597
598 mp->mnt_data = &mqfs_data;
599 MNT_ILOCK(mp);
600 mp->mnt_flag |= MNT_LOCAL;
601 MNT_IUNLOCK(mp);
602 vfs_getnewfsid(mp);
603
604 sbp = &mp->mnt_stat;
605 vfs_mountedfrom(mp, "mqueue");
606 sbp->f_bsize = PAGE_SIZE;
607 sbp->f_iosize = PAGE_SIZE;
608 sbp->f_blocks = 1;
609 sbp->f_bfree = 0;
610 sbp->f_bavail = 0;
611 sbp->f_files = 1;
612 sbp->f_ffree = 0;
613 return (0);
614 }
615
616 /*
617 * Unmount a mqfs instance
618 */
619 static int
620 mqfs_unmount(struct mount *mp, int mntflags)
621 {
622 int error;
623
624 error = vflush(mp, 0, (mntflags & MNT_FORCE) ? FORCECLOSE : 0,
625 curthread);
626 return (error);
627 }
628
629 /*
630 * Return a root vnode
631 */
632 static int
633 mqfs_root(struct mount *mp, int flags, struct vnode **vpp)
634 {
635 struct mqfs_info *mqfs;
636 int ret;
637
638 mqfs = VFSTOMQFS(mp);
639 ret = mqfs_allocv(mp, vpp, mqfs->mi_root);
640 return (ret);
641 }
642
643 /*
644 * Return filesystem stats
645 */
646 static int
647 mqfs_statfs(struct mount *mp, struct statfs *sbp)
648 {
649 /* XXX update statistics */
650 return (0);
651 }
652
653 /*
654 * Initialize a mqfs instance
655 */
656 static int
657 mqfs_init(struct vfsconf *vfc)
658 {
659 struct mqfs_node *root;
660 struct mqfs_info *mi;
661 osd_method_t methods[PR_MAXMETHOD] = {
662 [PR_METHOD_REMOVE] = mqfs_prison_remove,
663 };
664
665 mqnode_zone = uma_zcreate("mqnode", sizeof(struct mqfs_node),
666 NULL, NULL, NULL, NULL, UMA_ALIGN_PTR, 0);
667 mqueue_zone = uma_zcreate("mqueue", sizeof(struct mqueue),
668 NULL, NULL, NULL, NULL, UMA_ALIGN_PTR, 0);
669 mvdata_zone = uma_zcreate("mvdata",
670 sizeof(struct mqfs_vdata), NULL, NULL, NULL,
671 NULL, UMA_ALIGN_PTR, 0);
672 mqnoti_zone = uma_zcreate("mqnotifier", sizeof(struct mqueue_notifier),
673 NULL, NULL, NULL, NULL, UMA_ALIGN_PTR, 0);
674 mi = &mqfs_data;
675 sx_init(&mi->mi_lock, "mqfs lock");
676 /* set up the root diretory */
677 root = mqfs_create_node("/", 1, curthread->td_ucred, 01777,
678 mqfstype_root);
679 root->mn_info = mi;
680 LIST_INIT(&root->mn_children);
681 LIST_INIT(&root->mn_vnodes);
682 mi->mi_root = root;
683 mqfs_fileno_init(mi);
684 mqfs_fileno_alloc(mi, root);
685 mqfs_fixup_dir(root);
686 exit_tag = EVENTHANDLER_REGISTER(process_exit, mq_proc_exit, NULL,
687 EVENTHANDLER_PRI_ANY);
688 mq_fdclose = mqueue_fdclose;
689 p31b_setcfg(CTL_P1003_1B_MESSAGE_PASSING, _POSIX_MESSAGE_PASSING);
690 mqfs_osd_jail_slot = osd_jail_register(NULL, methods);
691 return (0);
692 }
693
694 /*
695 * Destroy a mqfs instance
696 */
697 static int
698 mqfs_uninit(struct vfsconf *vfc)
699 {
700 struct mqfs_info *mi;
701
702 if (!unloadable)
703 return (EOPNOTSUPP);
704 osd_jail_deregister(mqfs_osd_jail_slot);
705 EVENTHANDLER_DEREGISTER(process_exit, exit_tag);
706 mi = &mqfs_data;
707 mqfs_destroy(mi->mi_root);
708 mi->mi_root = NULL;
709 mqfs_fileno_uninit(mi);
710 sx_destroy(&mi->mi_lock);
711 uma_zdestroy(mqnode_zone);
712 uma_zdestroy(mqueue_zone);
713 uma_zdestroy(mvdata_zone);
714 uma_zdestroy(mqnoti_zone);
715 return (0);
716 }
717
718 /*
719 * task routine
720 */
721 static void
722 do_recycle(void *context, int pending __unused)
723 {
724 struct vnode *vp = (struct vnode *)context;
725
726 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
727 vrecycle(vp);
728 VOP_UNLOCK(vp);
729 vdrop(vp);
730 }
731
732 /*
733 * Allocate a vnode
734 */
735 static int
736 mqfs_allocv(struct mount *mp, struct vnode **vpp, struct mqfs_node *pn)
737 {
738 struct mqfs_vdata *vd;
739 struct mqfs_info *mqfs;
740 struct vnode *newvpp;
741 int error;
742
743 mqfs = pn->mn_info;
744 *vpp = NULL;
745 sx_xlock(&mqfs->mi_lock);
746 LIST_FOREACH(vd, &pn->mn_vnodes, mv_link) {
747 if (vd->mv_vnode->v_mount == mp) {
748 vhold(vd->mv_vnode);
749 break;
750 }
751 }
752
753 if (vd != NULL) {
754 found:
755 *vpp = vd->mv_vnode;
756 sx_xunlock(&mqfs->mi_lock);
757 error = vget(*vpp, LK_RETRY | LK_EXCLUSIVE);
758 vdrop(*vpp);
759 return (error);
760 }
761 sx_xunlock(&mqfs->mi_lock);
762
763 error = getnewvnode("mqueue", mp, &mqfs_vnodeops, &newvpp);
764 if (error)
765 return (error);
766 vn_lock(newvpp, LK_EXCLUSIVE | LK_RETRY);
767 error = insmntque(newvpp, mp);
768 if (error != 0)
769 return (error);
770
771 sx_xlock(&mqfs->mi_lock);
772 /*
773 * Check if it has already been allocated
774 * while we were blocked.
775 */
776 LIST_FOREACH(vd, &pn->mn_vnodes, mv_link) {
777 if (vd->mv_vnode->v_mount == mp) {
778 vhold(vd->mv_vnode);
779 sx_xunlock(&mqfs->mi_lock);
780
781 vgone(newvpp);
782 vput(newvpp);
783 goto found;
784 }
785 }
786
787 *vpp = newvpp;
788
789 vd = uma_zalloc(mvdata_zone, M_WAITOK);
790 (*vpp)->v_data = vd;
791 vd->mv_vnode = *vpp;
792 vd->mv_node = pn;
793 TASK_INIT(&vd->mv_task, 0, do_recycle, *vpp);
794 LIST_INSERT_HEAD(&pn->mn_vnodes, vd, mv_link);
795 mqnode_addref(pn);
796 switch (pn->mn_type) {
797 case mqfstype_root:
798 (*vpp)->v_vflag = VV_ROOT;
799 /* fall through */
800 case mqfstype_dir:
801 case mqfstype_this:
802 case mqfstype_parent:
803 (*vpp)->v_type = VDIR;
804 break;
805 case mqfstype_file:
806 (*vpp)->v_type = VREG;
807 break;
808 case mqfstype_symlink:
809 (*vpp)->v_type = VLNK;
810 break;
811 case mqfstype_none:
812 KASSERT(0, ("mqfs_allocf called for null node\n"));
813 default:
814 panic("%s has unexpected type: %d", pn->mn_name, pn->mn_type);
815 }
816 sx_xunlock(&mqfs->mi_lock);
817 return (0);
818 }
819
820 /*
821 * Search a directory entry
822 */
823 static struct mqfs_node *
824 mqfs_search(struct mqfs_node *pd, const char *name, int len, struct ucred *cred)
825 {
826 struct mqfs_node *pn;
827 const void *pr_root;
828
829 sx_assert(&pd->mn_info->mi_lock, SX_LOCKED);
830 pr_root = cred->cr_prison->pr_root;
831 LIST_FOREACH(pn, &pd->mn_children, mn_sibling) {
832 /* Only match names within the same prison root directory */
833 if ((pn->mn_pr_root == NULL || pn->mn_pr_root == pr_root) &&
834 strncmp(pn->mn_name, name, len) == 0 &&
835 pn->mn_name[len] == '\0')
836 return (pn);
837 }
838 return (NULL);
839 }
840
841 /*
842 * Look up a file or directory.
843 */
844 static int
845 mqfs_lookupx(struct vop_cachedlookup_args *ap)
846 {
847 struct componentname *cnp;
848 struct vnode *dvp, **vpp;
849 struct mqfs_node *pd;
850 struct mqfs_node *pn;
851 struct mqfs_info *mqfs;
852 int nameiop, flags, error, namelen;
853 char *pname;
854 struct thread *td;
855
856 cnp = ap->a_cnp;
857 vpp = ap->a_vpp;
858 dvp = ap->a_dvp;
859 pname = cnp->cn_nameptr;
860 namelen = cnp->cn_namelen;
861 td = cnp->cn_thread;
862 flags = cnp->cn_flags;
863 nameiop = cnp->cn_nameiop;
864 pd = VTON(dvp);
865 pn = NULL;
866 mqfs = pd->mn_info;
867 *vpp = NULLVP;
868
869 if (dvp->v_type != VDIR)
870 return (ENOTDIR);
871
872 error = VOP_ACCESS(dvp, VEXEC, cnp->cn_cred, cnp->cn_thread);
873 if (error)
874 return (error);
875
876 /* shortcut: check if the name is too long */
877 if (cnp->cn_namelen >= MQFS_NAMELEN)
878 return (ENOENT);
879
880 /* self */
881 if (namelen == 1 && pname[0] == '.') {
882 if ((flags & ISLASTCN) && nameiop != LOOKUP)
883 return (EINVAL);
884 pn = pd;
885 *vpp = dvp;
886 VREF(dvp);
887 return (0);
888 }
889
890 /* parent */
891 if (cnp->cn_flags & ISDOTDOT) {
892 if (dvp->v_vflag & VV_ROOT)
893 return (EIO);
894 if ((flags & ISLASTCN) && nameiop != LOOKUP)
895 return (EINVAL);
896 VOP_UNLOCK(dvp);
897 KASSERT(pd->mn_parent, ("non-root directory has no parent"));
898 pn = pd->mn_parent;
899 error = mqfs_allocv(dvp->v_mount, vpp, pn);
900 vn_lock(dvp, LK_EXCLUSIVE | LK_RETRY);
901 return (error);
902 }
903
904 /* named node */
905 sx_xlock(&mqfs->mi_lock);
906 pn = mqfs_search(pd, pname, namelen, cnp->cn_cred);
907 if (pn != NULL)
908 mqnode_addref(pn);
909 sx_xunlock(&mqfs->mi_lock);
910
911 /* found */
912 if (pn != NULL) {
913 /* DELETE */
914 if (nameiop == DELETE && (flags & ISLASTCN)) {
915 error = VOP_ACCESS(dvp, VWRITE, cnp->cn_cred, td);
916 if (error) {
917 mqnode_release(pn);
918 return (error);
919 }
920 if (*vpp == dvp) {
921 VREF(dvp);
922 *vpp = dvp;
923 mqnode_release(pn);
924 return (0);
925 }
926 }
927
928 /* allocate vnode */
929 error = mqfs_allocv(dvp->v_mount, vpp, pn);
930 mqnode_release(pn);
931 if (error == 0 && cnp->cn_flags & MAKEENTRY)
932 cache_enter(dvp, *vpp, cnp);
933 return (error);
934 }
935
936 /* not found */
937
938 /* will create a new entry in the directory ? */
939 if ((nameiop == CREATE || nameiop == RENAME) && (flags & LOCKPARENT)
940 && (flags & ISLASTCN)) {
941 error = VOP_ACCESS(dvp, VWRITE, cnp->cn_cred, td);
942 if (error)
943 return (error);
944 cnp->cn_flags |= SAVENAME;
945 return (EJUSTRETURN);
946 }
947 return (ENOENT);
948 }
949
950 #if 0
951 struct vop_lookup_args {
952 struct vop_generic_args a_gen;
953 struct vnode *a_dvp;
954 struct vnode **a_vpp;
955 struct componentname *a_cnp;
956 };
957 #endif
958
959 /*
960 * vnode lookup operation
961 */
962 static int
963 mqfs_lookup(struct vop_cachedlookup_args *ap)
964 {
965 int rc;
966
967 rc = mqfs_lookupx(ap);
968 return (rc);
969 }
970
971 #if 0
972 struct vop_create_args {
973 struct vnode *a_dvp;
974 struct vnode **a_vpp;
975 struct componentname *a_cnp;
976 struct vattr *a_vap;
977 };
978 #endif
979
980 /*
981 * vnode creation operation
982 */
983 static int
984 mqfs_create(struct vop_create_args *ap)
985 {
986 struct mqfs_info *mqfs = VFSTOMQFS(ap->a_dvp->v_mount);
987 struct componentname *cnp = ap->a_cnp;
988 struct mqfs_node *pd;
989 struct mqfs_node *pn;
990 struct mqueue *mq;
991 int error;
992
993 pd = VTON(ap->a_dvp);
994 if (pd->mn_type != mqfstype_root && pd->mn_type != mqfstype_dir)
995 return (ENOTDIR);
996 mq = mqueue_alloc(NULL);
997 if (mq == NULL)
998 return (EAGAIN);
999 sx_xlock(&mqfs->mi_lock);
1000 if ((cnp->cn_flags & HASBUF) == 0)
1001 panic("%s: no name", __func__);
1002 pn = mqfs_create_file(pd, cnp->cn_nameptr, cnp->cn_namelen,
1003 cnp->cn_cred, ap->a_vap->va_mode);
1004 if (pn == NULL) {
1005 sx_xunlock(&mqfs->mi_lock);
1006 error = ENOSPC;
1007 } else {
1008 mqnode_addref(pn);
1009 sx_xunlock(&mqfs->mi_lock);
1010 error = mqfs_allocv(ap->a_dvp->v_mount, ap->a_vpp, pn);
1011 mqnode_release(pn);
1012 if (error)
1013 mqfs_destroy(pn);
1014 else
1015 pn->mn_data = mq;
1016 }
1017 if (error)
1018 mqueue_free(mq);
1019 return (error);
1020 }
1021
1022 /*
1023 * Remove an entry
1024 */
1025 static
1026 int do_unlink(struct mqfs_node *pn, struct ucred *ucred)
1027 {
1028 struct mqfs_node *parent;
1029 struct mqfs_vdata *vd;
1030 int error = 0;
1031
1032 sx_assert(&pn->mn_info->mi_lock, SX_LOCKED);
1033
1034 if (ucred->cr_uid != pn->mn_uid &&
1035 (error = priv_check_cred(ucred, PRIV_MQ_ADMIN)) != 0)
1036 error = EACCES;
1037 else if (!pn->mn_deleted) {
1038 parent = pn->mn_parent;
1039 pn->mn_parent = NULL;
1040 pn->mn_deleted = 1;
1041 LIST_REMOVE(pn, mn_sibling);
1042 LIST_FOREACH(vd, &pn->mn_vnodes, mv_link) {
1043 cache_purge(vd->mv_vnode);
1044 vhold(vd->mv_vnode);
1045 taskqueue_enqueue(taskqueue_thread, &vd->mv_task);
1046 }
1047 mqnode_release(pn);
1048 mqnode_release(parent);
1049 } else
1050 error = ENOENT;
1051 return (error);
1052 }
1053
1054 #if 0
1055 struct vop_remove_args {
1056 struct vnode *a_dvp;
1057 struct vnode *a_vp;
1058 struct componentname *a_cnp;
1059 };
1060 #endif
1061
1062 /*
1063 * vnode removal operation
1064 */
1065 static int
1066 mqfs_remove(struct vop_remove_args *ap)
1067 {
1068 struct mqfs_info *mqfs = VFSTOMQFS(ap->a_dvp->v_mount);
1069 struct mqfs_node *pn;
1070 int error;
1071
1072 if (ap->a_vp->v_type == VDIR)
1073 return (EPERM);
1074 pn = VTON(ap->a_vp);
1075 sx_xlock(&mqfs->mi_lock);
1076 error = do_unlink(pn, ap->a_cnp->cn_cred);
1077 sx_xunlock(&mqfs->mi_lock);
1078 return (error);
1079 }
1080
1081 #if 0
1082 struct vop_inactive_args {
1083 struct vnode *a_vp;
1084 struct thread *a_td;
1085 };
1086 #endif
1087
1088 static int
1089 mqfs_inactive(struct vop_inactive_args *ap)
1090 {
1091 struct mqfs_node *pn = VTON(ap->a_vp);
1092
1093 if (pn->mn_deleted)
1094 vrecycle(ap->a_vp);
1095 return (0);
1096 }
1097
1098 #if 0
1099 struct vop_reclaim_args {
1100 struct vop_generic_args a_gen;
1101 struct vnode *a_vp;
1102 };
1103 #endif
1104
1105 static int
1106 mqfs_reclaim(struct vop_reclaim_args *ap)
1107 {
1108 struct mqfs_info *mqfs = VFSTOMQFS(ap->a_vp->v_mount);
1109 struct vnode *vp = ap->a_vp;
1110 struct mqfs_node *pn;
1111 struct mqfs_vdata *vd;
1112
1113 vd = vp->v_data;
1114 pn = vd->mv_node;
1115 sx_xlock(&mqfs->mi_lock);
1116 vp->v_data = NULL;
1117 LIST_REMOVE(vd, mv_link);
1118 uma_zfree(mvdata_zone, vd);
1119 mqnode_release(pn);
1120 sx_xunlock(&mqfs->mi_lock);
1121 return (0);
1122 }
1123
1124 #if 0
1125 struct vop_open_args {
1126 struct vop_generic_args a_gen;
1127 struct vnode *a_vp;
1128 int a_mode;
1129 struct ucred *a_cred;
1130 struct thread *a_td;
1131 struct file *a_fp;
1132 };
1133 #endif
1134
1135 static int
1136 mqfs_open(struct vop_open_args *ap)
1137 {
1138 return (0);
1139 }
1140
1141 #if 0
1142 struct vop_close_args {
1143 struct vop_generic_args a_gen;
1144 struct vnode *a_vp;
1145 int a_fflag;
1146 struct ucred *a_cred;
1147 struct thread *a_td;
1148 };
1149 #endif
1150
1151 static int
1152 mqfs_close(struct vop_close_args *ap)
1153 {
1154 return (0);
1155 }
1156
1157 #if 0
1158 struct vop_access_args {
1159 struct vop_generic_args a_gen;
1160 struct vnode *a_vp;
1161 accmode_t a_accmode;
1162 struct ucred *a_cred;
1163 struct thread *a_td;
1164 };
1165 #endif
1166
1167 /*
1168 * Verify permissions
1169 */
1170 static int
1171 mqfs_access(struct vop_access_args *ap)
1172 {
1173 struct vnode *vp = ap->a_vp;
1174 struct vattr vattr;
1175 int error;
1176
1177 error = VOP_GETATTR(vp, &vattr, ap->a_cred);
1178 if (error)
1179 return (error);
1180 error = vaccess(vp->v_type, vattr.va_mode, vattr.va_uid, vattr.va_gid,
1181 ap->a_accmode, ap->a_cred);
1182 return (error);
1183 }
1184
1185 #if 0
1186 struct vop_getattr_args {
1187 struct vop_generic_args a_gen;
1188 struct vnode *a_vp;
1189 struct vattr *a_vap;
1190 struct ucred *a_cred;
1191 };
1192 #endif
1193
1194 /*
1195 * Get file attributes
1196 */
1197 static int
1198 mqfs_getattr(struct vop_getattr_args *ap)
1199 {
1200 struct vnode *vp = ap->a_vp;
1201 struct mqfs_node *pn = VTON(vp);
1202 struct vattr *vap = ap->a_vap;
1203 int error = 0;
1204
1205 vap->va_type = vp->v_type;
1206 vap->va_mode = pn->mn_mode;
1207 vap->va_nlink = 1;
1208 vap->va_uid = pn->mn_uid;
1209 vap->va_gid = pn->mn_gid;
1210 vap->va_fsid = vp->v_mount->mnt_stat.f_fsid.val[0];
1211 vap->va_fileid = pn->mn_fileno;
1212 vap->va_size = 0;
1213 vap->va_blocksize = PAGE_SIZE;
1214 vap->va_bytes = vap->va_size = 0;
1215 vap->va_atime = pn->mn_atime;
1216 vap->va_mtime = pn->mn_mtime;
1217 vap->va_ctime = pn->mn_ctime;
1218 vap->va_birthtime = pn->mn_birth;
1219 vap->va_gen = 0;
1220 vap->va_flags = 0;
1221 vap->va_rdev = NODEV;
1222 vap->va_bytes = 0;
1223 vap->va_filerev = 0;
1224 return (error);
1225 }
1226
1227 #if 0
1228 struct vop_setattr_args {
1229 struct vop_generic_args a_gen;
1230 struct vnode *a_vp;
1231 struct vattr *a_vap;
1232 struct ucred *a_cred;
1233 };
1234 #endif
1235 /*
1236 * Set attributes
1237 */
1238 static int
1239 mqfs_setattr(struct vop_setattr_args *ap)
1240 {
1241 struct mqfs_node *pn;
1242 struct vattr *vap;
1243 struct vnode *vp;
1244 struct thread *td;
1245 int c, error;
1246 uid_t uid;
1247 gid_t gid;
1248
1249 td = curthread;
1250 vap = ap->a_vap;
1251 vp = ap->a_vp;
1252 if ((vap->va_type != VNON) ||
1253 (vap->va_nlink != VNOVAL) ||
1254 (vap->va_fsid != VNOVAL) ||
1255 (vap->va_fileid != VNOVAL) ||
1256 (vap->va_blocksize != VNOVAL) ||
1257 (vap->va_flags != VNOVAL && vap->va_flags != 0) ||
1258 (vap->va_rdev != VNOVAL) ||
1259 ((int)vap->va_bytes != VNOVAL) ||
1260 (vap->va_gen != VNOVAL)) {
1261 return (EINVAL);
1262 }
1263
1264 pn = VTON(vp);
1265
1266 error = c = 0;
1267 if (vap->va_uid == (uid_t)VNOVAL)
1268 uid = pn->mn_uid;
1269 else
1270 uid = vap->va_uid;
1271 if (vap->va_gid == (gid_t)VNOVAL)
1272 gid = pn->mn_gid;
1273 else
1274 gid = vap->va_gid;
1275
1276 if (uid != pn->mn_uid || gid != pn->mn_gid) {
1277 /*
1278 * To modify the ownership of a file, must possess VADMIN
1279 * for that file.
1280 */
1281 if ((error = VOP_ACCESS(vp, VADMIN, ap->a_cred, td)))
1282 return (error);
1283
1284 /*
1285 * XXXRW: Why is there a privilege check here: shouldn't the
1286 * check in VOP_ACCESS() be enough? Also, are the group bits
1287 * below definitely right?
1288 */
1289 if (((ap->a_cred->cr_uid != pn->mn_uid) || uid != pn->mn_uid ||
1290 (gid != pn->mn_gid && !groupmember(gid, ap->a_cred))) &&
1291 (error = priv_check(td, PRIV_MQ_ADMIN)) != 0)
1292 return (error);
1293 pn->mn_uid = uid;
1294 pn->mn_gid = gid;
1295 c = 1;
1296 }
1297
1298 if (vap->va_mode != (mode_t)VNOVAL) {
1299 if ((ap->a_cred->cr_uid != pn->mn_uid) &&
1300 (error = priv_check(td, PRIV_MQ_ADMIN)))
1301 return (error);
1302 pn->mn_mode = vap->va_mode;
1303 c = 1;
1304 }
1305
1306 if (vap->va_atime.tv_sec != VNOVAL || vap->va_mtime.tv_sec != VNOVAL) {
1307 /* See the comment in ufs_vnops::ufs_setattr(). */
1308 if ((error = VOP_ACCESS(vp, VADMIN, ap->a_cred, td)) &&
1309 ((vap->va_vaflags & VA_UTIMES_NULL) == 0 ||
1310 (error = VOP_ACCESS(vp, VWRITE, ap->a_cred, td))))
1311 return (error);
1312 if (vap->va_atime.tv_sec != VNOVAL) {
1313 pn->mn_atime = vap->va_atime;
1314 }
1315 if (vap->va_mtime.tv_sec != VNOVAL) {
1316 pn->mn_mtime = vap->va_mtime;
1317 }
1318 c = 1;
1319 }
1320 if (c) {
1321 vfs_timestamp(&pn->mn_ctime);
1322 }
1323 return (0);
1324 }
1325
1326 #if 0
1327 struct vop_read_args {
1328 struct vop_generic_args a_gen;
1329 struct vnode *a_vp;
1330 struct uio *a_uio;
1331 int a_ioflag;
1332 struct ucred *a_cred;
1333 };
1334 #endif
1335
1336 /*
1337 * Read from a file
1338 */
1339 static int
1340 mqfs_read(struct vop_read_args *ap)
1341 {
1342 char buf[80];
1343 struct vnode *vp = ap->a_vp;
1344 struct uio *uio = ap->a_uio;
1345 struct mqueue *mq;
1346 int len, error;
1347
1348 if (vp->v_type != VREG)
1349 return (EINVAL);
1350
1351 mq = VTOMQ(vp);
1352 snprintf(buf, sizeof(buf),
1353 "QSIZE:%-10ld MAXMSG:%-10ld CURMSG:%-10ld MSGSIZE:%-10ld\n",
1354 mq->mq_totalbytes,
1355 mq->mq_maxmsg,
1356 mq->mq_curmsgs,
1357 mq->mq_msgsize);
1358 buf[sizeof(buf)-1] = '\0';
1359 len = strlen(buf);
1360 error = uiomove_frombuf(buf, len, uio);
1361 return (error);
1362 }
1363
1364 #if 0
1365 struct vop_readdir_args {
1366 struct vop_generic_args a_gen;
1367 struct vnode *a_vp;
1368 struct uio *a_uio;
1369 struct ucred *a_cred;
1370 int *a_eofflag;
1371 int *a_ncookies;
1372 u_long **a_cookies;
1373 };
1374 #endif
1375
1376 /*
1377 * Return directory entries.
1378 */
1379 static int
1380 mqfs_readdir(struct vop_readdir_args *ap)
1381 {
1382 struct vnode *vp;
1383 struct mqfs_info *mi;
1384 struct mqfs_node *pd;
1385 struct mqfs_node *pn;
1386 struct dirent entry;
1387 struct uio *uio;
1388 const void *pr_root;
1389 int *tmp_ncookies = NULL;
1390 off_t offset;
1391 int error, i;
1392
1393 vp = ap->a_vp;
1394 mi = VFSTOMQFS(vp->v_mount);
1395 pd = VTON(vp);
1396 uio = ap->a_uio;
1397
1398 if (vp->v_type != VDIR)
1399 return (ENOTDIR);
1400
1401 if (uio->uio_offset < 0)
1402 return (EINVAL);
1403
1404 if (ap->a_ncookies != NULL) {
1405 tmp_ncookies = ap->a_ncookies;
1406 *ap->a_ncookies = 0;
1407 ap->a_ncookies = NULL;
1408 }
1409
1410 error = 0;
1411 offset = 0;
1412
1413 pr_root = ap->a_cred->cr_prison->pr_root;
1414 sx_xlock(&mi->mi_lock);
1415
1416 LIST_FOREACH(pn, &pd->mn_children, mn_sibling) {
1417 entry.d_reclen = sizeof(entry);
1418
1419 /*
1420 * Only show names within the same prison root directory
1421 * (or not associated with a prison, e.g. "." and "..").
1422 */
1423 if (pn->mn_pr_root != NULL && pn->mn_pr_root != pr_root)
1424 continue;
1425 if (!pn->mn_fileno)
1426 mqfs_fileno_alloc(mi, pn);
1427 entry.d_fileno = pn->mn_fileno;
1428 entry.d_off = offset + entry.d_reclen;
1429 for (i = 0; i < MQFS_NAMELEN - 1 && pn->mn_name[i] != '\0'; ++i)
1430 entry.d_name[i] = pn->mn_name[i];
1431 entry.d_namlen = i;
1432 switch (pn->mn_type) {
1433 case mqfstype_root:
1434 case mqfstype_dir:
1435 case mqfstype_this:
1436 case mqfstype_parent:
1437 entry.d_type = DT_DIR;
1438 break;
1439 case mqfstype_file:
1440 entry.d_type = DT_REG;
1441 break;
1442 case mqfstype_symlink:
1443 entry.d_type = DT_LNK;
1444 break;
1445 default:
1446 panic("%s has unexpected node type: %d", pn->mn_name,
1447 pn->mn_type);
1448 }
1449 dirent_terminate(&entry);
1450 if (entry.d_reclen > uio->uio_resid)
1451 break;
1452 if (offset >= uio->uio_offset) {
1453 error = vfs_read_dirent(ap, &entry, offset);
1454 if (error)
1455 break;
1456 }
1457 offset += entry.d_reclen;
1458 }
1459 sx_xunlock(&mi->mi_lock);
1460
1461 uio->uio_offset = offset;
1462
1463 if (tmp_ncookies != NULL)
1464 ap->a_ncookies = tmp_ncookies;
1465
1466 return (error);
1467 }
1468
1469 #ifdef notyet
1470
1471 #if 0
1472 struct vop_mkdir_args {
1473 struct vnode *a_dvp;
1474 struvt vnode **a_vpp;
1475 struvt componentname *a_cnp;
1476 struct vattr *a_vap;
1477 };
1478 #endif
1479
1480 /*
1481 * Create a directory.
1482 */
1483 static int
1484 mqfs_mkdir(struct vop_mkdir_args *ap)
1485 {
1486 struct mqfs_info *mqfs = VFSTOMQFS(ap->a_dvp->v_mount);
1487 struct componentname *cnp = ap->a_cnp;
1488 struct mqfs_node *pd = VTON(ap->a_dvp);
1489 struct mqfs_node *pn;
1490 int error;
1491
1492 if (pd->mn_type != mqfstype_root && pd->mn_type != mqfstype_dir)
1493 return (ENOTDIR);
1494 sx_xlock(&mqfs->mi_lock);
1495 if ((cnp->cn_flags & HASBUF) == 0)
1496 panic("%s: no name", __func__);
1497 pn = mqfs_create_dir(pd, cnp->cn_nameptr, cnp->cn_namelen,
1498 ap->a_vap->cn_cred, ap->a_vap->va_mode);
1499 if (pn != NULL)
1500 mqnode_addref(pn);
1501 sx_xunlock(&mqfs->mi_lock);
1502 if (pn == NULL) {
1503 error = ENOSPC;
1504 } else {
1505 error = mqfs_allocv(ap->a_dvp->v_mount, ap->a_vpp, pn);
1506 mqnode_release(pn);
1507 }
1508 return (error);
1509 }
1510
1511 #if 0
1512 struct vop_rmdir_args {
1513 struct vnode *a_dvp;
1514 struct vnode *a_vp;
1515 struct componentname *a_cnp;
1516 };
1517 #endif
1518
1519 /*
1520 * Remove a directory.
1521 */
1522 static int
1523 mqfs_rmdir(struct vop_rmdir_args *ap)
1524 {
1525 struct mqfs_info *mqfs = VFSTOMQFS(ap->a_dvp->v_mount);
1526 struct mqfs_node *pn = VTON(ap->a_vp);
1527 struct mqfs_node *pt;
1528
1529 if (pn->mn_type != mqfstype_dir)
1530 return (ENOTDIR);
1531
1532 sx_xlock(&mqfs->mi_lock);
1533 if (pn->mn_deleted) {
1534 sx_xunlock(&mqfs->mi_lock);
1535 return (ENOENT);
1536 }
1537
1538 pt = LIST_FIRST(&pn->mn_children);
1539 pt = LIST_NEXT(pt, mn_sibling);
1540 pt = LIST_NEXT(pt, mn_sibling);
1541 if (pt != NULL) {
1542 sx_xunlock(&mqfs->mi_lock);
1543 return (ENOTEMPTY);
1544 }
1545 pt = pn->mn_parent;
1546 pn->mn_parent = NULL;
1547 pn->mn_deleted = 1;
1548 LIST_REMOVE(pn, mn_sibling);
1549 mqnode_release(pn);
1550 mqnode_release(pt);
1551 sx_xunlock(&mqfs->mi_lock);
1552 cache_purge(ap->a_vp);
1553 return (0);
1554 }
1555
1556 #endif /* notyet */
1557
1558 /*
1559 * See if this prison root is obsolete, and clean up associated queues if it is.
1560 */
1561 static int
1562 mqfs_prison_remove(void *obj, void *data __unused)
1563 {
1564 const struct prison *pr = obj;
1565 struct prison *tpr;
1566 struct mqfs_node *pn, *tpn;
1567 int found;
1568
1569 found = 0;
1570 TAILQ_FOREACH(tpr, &allprison, pr_list) {
1571 prison_lock(tpr);
1572 if (tpr != pr && prison_isvalid(tpr) &&
1573 tpr->pr_root == pr->pr_root)
1574 found = 1;
1575 prison_unlock(tpr);
1576 }
1577 if (!found) {
1578 /*
1579 * No jails are rooted in this directory anymore,
1580 * so no queues should be either.
1581 */
1582 sx_xlock(&mqfs_data.mi_lock);
1583 LIST_FOREACH_SAFE(pn, &mqfs_data.mi_root->mn_children,
1584 mn_sibling, tpn) {
1585 if (pn->mn_pr_root == pr->pr_root)
1586 (void)do_unlink(pn, curthread->td_ucred);
1587 }
1588 sx_xunlock(&mqfs_data.mi_lock);
1589 }
1590 return (0);
1591 }
1592
1593 /*
1594 * Allocate a message queue
1595 */
1596 static struct mqueue *
1597 mqueue_alloc(const struct mq_attr *attr)
1598 {
1599 struct mqueue *mq;
1600
1601 if (curmq >= maxmq)
1602 return (NULL);
1603 mq = uma_zalloc(mqueue_zone, M_WAITOK | M_ZERO);
1604 TAILQ_INIT(&mq->mq_msgq);
1605 if (attr != NULL) {
1606 mq->mq_maxmsg = attr->mq_maxmsg;
1607 mq->mq_msgsize = attr->mq_msgsize;
1608 } else {
1609 mq->mq_maxmsg = default_maxmsg;
1610 mq->mq_msgsize = default_msgsize;
1611 }
1612 mtx_init(&mq->mq_mutex, "mqueue lock", NULL, MTX_DEF);
1613 knlist_init_mtx(&mq->mq_rsel.si_note, &mq->mq_mutex);
1614 knlist_init_mtx(&mq->mq_wsel.si_note, &mq->mq_mutex);
1615 atomic_add_int(&curmq, 1);
1616 return (mq);
1617 }
1618
1619 /*
1620 * Destroy a message queue
1621 */
1622 static void
1623 mqueue_free(struct mqueue *mq)
1624 {
1625 struct mqueue_msg *msg;
1626
1627 while ((msg = TAILQ_FIRST(&mq->mq_msgq)) != NULL) {
1628 TAILQ_REMOVE(&mq->mq_msgq, msg, msg_link);
1629 free(msg, M_MQUEUEDATA);
1630 }
1631
1632 mtx_destroy(&mq->mq_mutex);
1633 seldrain(&mq->mq_rsel);
1634 seldrain(&mq->mq_wsel);
1635 knlist_destroy(&mq->mq_rsel.si_note);
1636 knlist_destroy(&mq->mq_wsel.si_note);
1637 uma_zfree(mqueue_zone, mq);
1638 atomic_add_int(&curmq, -1);
1639 }
1640
1641 /*
1642 * Load a message from user space
1643 */
1644 static struct mqueue_msg *
1645 mqueue_loadmsg(const char *msg_ptr, size_t msg_size, int msg_prio)
1646 {
1647 struct mqueue_msg *msg;
1648 size_t len;
1649 int error;
1650
1651 len = sizeof(struct mqueue_msg) + msg_size;
1652 msg = malloc(len, M_MQUEUEDATA, M_WAITOK);
1653 error = copyin(msg_ptr, ((char *)msg) + sizeof(struct mqueue_msg),
1654 msg_size);
1655 if (error) {
1656 free(msg, M_MQUEUEDATA);
1657 msg = NULL;
1658 } else {
1659 msg->msg_size = msg_size;
1660 msg->msg_prio = msg_prio;
1661 }
1662 return (msg);
1663 }
1664
1665 /*
1666 * Save a message to user space
1667 */
1668 static int
1669 mqueue_savemsg(struct mqueue_msg *msg, char *msg_ptr, int *msg_prio)
1670 {
1671 int error;
1672
1673 error = copyout(((char *)msg) + sizeof(*msg), msg_ptr,
1674 msg->msg_size);
1675 if (error == 0 && msg_prio != NULL)
1676 error = copyout(&msg->msg_prio, msg_prio, sizeof(int));
1677 return (error);
1678 }
1679
1680 /*
1681 * Free a message's memory
1682 */
1683 static __inline void
1684 mqueue_freemsg(struct mqueue_msg *msg)
1685 {
1686 free(msg, M_MQUEUEDATA);
1687 }
1688
1689 /*
1690 * Send a message. if waitok is false, thread will not be
1691 * blocked if there is no data in queue, otherwise, absolute
1692 * time will be checked.
1693 */
1694 int
1695 mqueue_send(struct mqueue *mq, const char *msg_ptr,
1696 size_t msg_len, unsigned msg_prio, int waitok,
1697 const struct timespec *abs_timeout)
1698 {
1699 struct mqueue_msg *msg;
1700 struct timespec ts, ts2;
1701 struct timeval tv;
1702 int error;
1703
1704 if (msg_prio >= MQ_PRIO_MAX)
1705 return (EINVAL);
1706 if (msg_len > mq->mq_msgsize)
1707 return (EMSGSIZE);
1708 msg = mqueue_loadmsg(msg_ptr, msg_len, msg_prio);
1709 if (msg == NULL)
1710 return (EFAULT);
1711
1712 /* O_NONBLOCK case */
1713 if (!waitok) {
1714 error = _mqueue_send(mq, msg, -1);
1715 if (error)
1716 goto bad;
1717 return (0);
1718 }
1719
1720 /* we allow a null timeout (wait forever) */
1721 if (abs_timeout == NULL) {
1722 error = _mqueue_send(mq, msg, 0);
1723 if (error)
1724 goto bad;
1725 return (0);
1726 }
1727
1728 /* send it before checking time */
1729 error = _mqueue_send(mq, msg, -1);
1730 if (error == 0)
1731 return (0);
1732
1733 if (error != EAGAIN)
1734 goto bad;
1735
1736 if (abs_timeout->tv_nsec >= 1000000000 || abs_timeout->tv_nsec < 0) {
1737 error = EINVAL;
1738 goto bad;
1739 }
1740 for (;;) {
1741 getnanotime(&ts);
1742 timespecsub(abs_timeout, &ts, &ts2);
1743 if (ts2.tv_sec < 0 || (ts2.tv_sec == 0 && ts2.tv_nsec <= 0)) {
1744 error = ETIMEDOUT;
1745 break;
1746 }
1747 TIMESPEC_TO_TIMEVAL(&tv, &ts2);
1748 error = _mqueue_send(mq, msg, tvtohz(&tv));
1749 if (error != ETIMEDOUT)
1750 break;
1751 }
1752 if (error == 0)
1753 return (0);
1754 bad:
1755 mqueue_freemsg(msg);
1756 return (error);
1757 }
1758
1759 /*
1760 * Common routine to send a message
1761 */
1762 static int
1763 _mqueue_send(struct mqueue *mq, struct mqueue_msg *msg, int timo)
1764 {
1765 struct mqueue_msg *msg2;
1766 int error = 0;
1767
1768 mtx_lock(&mq->mq_mutex);
1769 while (mq->mq_curmsgs >= mq->mq_maxmsg && error == 0) {
1770 if (timo < 0) {
1771 mtx_unlock(&mq->mq_mutex);
1772 return (EAGAIN);
1773 }
1774 mq->mq_senders++;
1775 error = msleep(&mq->mq_senders, &mq->mq_mutex,
1776 PCATCH, "mqsend", timo);
1777 mq->mq_senders--;
1778 if (error == EAGAIN)
1779 error = ETIMEDOUT;
1780 }
1781 if (mq->mq_curmsgs >= mq->mq_maxmsg) {
1782 mtx_unlock(&mq->mq_mutex);
1783 return (error);
1784 }
1785 error = 0;
1786 if (TAILQ_EMPTY(&mq->mq_msgq)) {
1787 TAILQ_INSERT_HEAD(&mq->mq_msgq, msg, msg_link);
1788 } else {
1789 if (msg->msg_prio <= TAILQ_LAST(&mq->mq_msgq, msgq)->msg_prio) {
1790 TAILQ_INSERT_TAIL(&mq->mq_msgq, msg, msg_link);
1791 } else {
1792 TAILQ_FOREACH(msg2, &mq->mq_msgq, msg_link) {
1793 if (msg2->msg_prio < msg->msg_prio)
1794 break;
1795 }
1796 TAILQ_INSERT_BEFORE(msg2, msg, msg_link);
1797 }
1798 }
1799 mq->mq_curmsgs++;
1800 mq->mq_totalbytes += msg->msg_size;
1801 if (mq->mq_receivers)
1802 wakeup_one(&mq->mq_receivers);
1803 else if (mq->mq_notifier != NULL)
1804 mqueue_send_notification(mq);
1805 if (mq->mq_flags & MQ_RSEL) {
1806 mq->mq_flags &= ~MQ_RSEL;
1807 selwakeup(&mq->mq_rsel);
1808 }
1809 KNOTE_LOCKED(&mq->mq_rsel.si_note, 0);
1810 mtx_unlock(&mq->mq_mutex);
1811 return (0);
1812 }
1813
1814 /*
1815 * Send realtime a signal to process which registered itself
1816 * successfully by mq_notify.
1817 */
1818 static void
1819 mqueue_send_notification(struct mqueue *mq)
1820 {
1821 struct mqueue_notifier *nt;
1822 struct thread *td;
1823 struct proc *p;
1824 int error;
1825
1826 mtx_assert(&mq->mq_mutex, MA_OWNED);
1827 nt = mq->mq_notifier;
1828 if (nt->nt_sigev.sigev_notify != SIGEV_NONE) {
1829 p = nt->nt_proc;
1830 error = sigev_findtd(p, &nt->nt_sigev, &td);
1831 if (error) {
1832 mq->mq_notifier = NULL;
1833 return;
1834 }
1835 if (!KSI_ONQ(&nt->nt_ksi)) {
1836 ksiginfo_set_sigev(&nt->nt_ksi, &nt->nt_sigev);
1837 tdsendsignal(p, td, nt->nt_ksi.ksi_signo, &nt->nt_ksi);
1838 }
1839 PROC_UNLOCK(p);
1840 }
1841 mq->mq_notifier = NULL;
1842 }
1843
1844 /*
1845 * Get a message. if waitok is false, thread will not be
1846 * blocked if there is no data in queue, otherwise, absolute
1847 * time will be checked.
1848 */
1849 int
1850 mqueue_receive(struct mqueue *mq, char *msg_ptr,
1851 size_t msg_len, unsigned *msg_prio, int waitok,
1852 const struct timespec *abs_timeout)
1853 {
1854 struct mqueue_msg *msg;
1855 struct timespec ts, ts2;
1856 struct timeval tv;
1857 int error;
1858
1859 if (msg_len < mq->mq_msgsize)
1860 return (EMSGSIZE);
1861
1862 /* O_NONBLOCK case */
1863 if (!waitok) {
1864 error = _mqueue_recv(mq, &msg, -1);
1865 if (error)
1866 return (error);
1867 goto received;
1868 }
1869
1870 /* we allow a null timeout (wait forever). */
1871 if (abs_timeout == NULL) {
1872 error = _mqueue_recv(mq, &msg, 0);
1873 if (error)
1874 return (error);
1875 goto received;
1876 }
1877
1878 /* try to get a message before checking time */
1879 error = _mqueue_recv(mq, &msg, -1);
1880 if (error == 0)
1881 goto received;
1882
1883 if (error != EAGAIN)
1884 return (error);
1885
1886 if (abs_timeout->tv_nsec >= 1000000000 || abs_timeout->tv_nsec < 0) {
1887 error = EINVAL;
1888 return (error);
1889 }
1890
1891 for (;;) {
1892 getnanotime(&ts);
1893 timespecsub(abs_timeout, &ts, &ts2);
1894 if (ts2.tv_sec < 0 || (ts2.tv_sec == 0 && ts2.tv_nsec <= 0)) {
1895 error = ETIMEDOUT;
1896 return (error);
1897 }
1898 TIMESPEC_TO_TIMEVAL(&tv, &ts2);
1899 error = _mqueue_recv(mq, &msg, tvtohz(&tv));
1900 if (error == 0)
1901 break;
1902 if (error != ETIMEDOUT)
1903 return (error);
1904 }
1905
1906 received:
1907 error = mqueue_savemsg(msg, msg_ptr, msg_prio);
1908 if (error == 0) {
1909 curthread->td_retval[0] = msg->msg_size;
1910 curthread->td_retval[1] = 0;
1911 }
1912 mqueue_freemsg(msg);
1913 return (error);
1914 }
1915
1916 /*
1917 * Common routine to receive a message
1918 */
1919 static int
1920 _mqueue_recv(struct mqueue *mq, struct mqueue_msg **msg, int timo)
1921 {
1922 int error = 0;
1923
1924 mtx_lock(&mq->mq_mutex);
1925 while ((*msg = TAILQ_FIRST(&mq->mq_msgq)) == NULL && error == 0) {
1926 if (timo < 0) {
1927 mtx_unlock(&mq->mq_mutex);
1928 return (EAGAIN);
1929 }
1930 mq->mq_receivers++;
1931 error = msleep(&mq->mq_receivers, &mq->mq_mutex,
1932 PCATCH, "mqrecv", timo);
1933 mq->mq_receivers--;
1934 if (error == EAGAIN)
1935 error = ETIMEDOUT;
1936 }
1937 if (*msg != NULL) {
1938 error = 0;
1939 TAILQ_REMOVE(&mq->mq_msgq, *msg, msg_link);
1940 mq->mq_curmsgs--;
1941 mq->mq_totalbytes -= (*msg)->msg_size;
1942 if (mq->mq_senders)
1943 wakeup_one(&mq->mq_senders);
1944 if (mq->mq_flags & MQ_WSEL) {
1945 mq->mq_flags &= ~MQ_WSEL;
1946 selwakeup(&mq->mq_wsel);
1947 }
1948 KNOTE_LOCKED(&mq->mq_wsel.si_note, 0);
1949 }
1950 if (mq->mq_notifier != NULL && mq->mq_receivers == 0 &&
1951 !TAILQ_EMPTY(&mq->mq_msgq)) {
1952 mqueue_send_notification(mq);
1953 }
1954 mtx_unlock(&mq->mq_mutex);
1955 return (error);
1956 }
1957
1958 static __inline struct mqueue_notifier *
1959 notifier_alloc(void)
1960 {
1961 return (uma_zalloc(mqnoti_zone, M_WAITOK | M_ZERO));
1962 }
1963
1964 static __inline void
1965 notifier_free(struct mqueue_notifier *p)
1966 {
1967 uma_zfree(mqnoti_zone, p);
1968 }
1969
1970 static struct mqueue_notifier *
1971 notifier_search(struct proc *p, int fd)
1972 {
1973 struct mqueue_notifier *nt;
1974
1975 LIST_FOREACH(nt, &p->p_mqnotifier, nt_link) {
1976 if (nt->nt_ksi.ksi_mqd == fd)
1977 break;
1978 }
1979 return (nt);
1980 }
1981
1982 static __inline void
1983 notifier_insert(struct proc *p, struct mqueue_notifier *nt)
1984 {
1985 LIST_INSERT_HEAD(&p->p_mqnotifier, nt, nt_link);
1986 }
1987
1988 static __inline void
1989 notifier_delete(struct proc *p, struct mqueue_notifier *nt)
1990 {
1991 LIST_REMOVE(nt, nt_link);
1992 notifier_free(nt);
1993 }
1994
1995 static void
1996 notifier_remove(struct proc *p, struct mqueue *mq, int fd)
1997 {
1998 struct mqueue_notifier *nt;
1999
2000 mtx_assert(&mq->mq_mutex, MA_OWNED);
2001 PROC_LOCK(p);
2002 nt = notifier_search(p, fd);
2003 if (nt != NULL) {
2004 if (mq->mq_notifier == nt)
2005 mq->mq_notifier = NULL;
2006 sigqueue_take(&nt->nt_ksi);
2007 notifier_delete(p, nt);
2008 }
2009 PROC_UNLOCK(p);
2010 }
2011
2012 static int
2013 kern_kmq_open(struct thread *td, const char *upath, int flags, mode_t mode,
2014 const struct mq_attr *attr)
2015 {
2016 char path[MQFS_NAMELEN + 1];
2017 struct mqfs_node *pn;
2018 struct pwddesc *pdp;
2019 struct file *fp;
2020 struct mqueue *mq;
2021 int fd, error, len, cmode;
2022
2023 AUDIT_ARG_FFLAGS(flags);
2024 AUDIT_ARG_MODE(mode);
2025
2026 pdp = td->td_proc->p_pd;
2027 cmode = (((mode & ~pdp->pd_cmask) & ALLPERMS) & ~S_ISTXT);
2028 mq = NULL;
2029 if ((flags & O_CREAT) != 0 && attr != NULL) {
2030 if (attr->mq_maxmsg <= 0 || attr->mq_maxmsg > maxmsg)
2031 return (EINVAL);
2032 if (attr->mq_msgsize <= 0 || attr->mq_msgsize > maxmsgsize)
2033 return (EINVAL);
2034 }
2035
2036 error = copyinstr(upath, path, MQFS_NAMELEN + 1, NULL);
2037 if (error)
2038 return (error);
2039
2040 /*
2041 * The first character of name must be a slash (/) character
2042 * and the remaining characters of name cannot include any slash
2043 * characters.
2044 */
2045 len = strlen(path);
2046 if (len < 2 || path[0] != '/' || strchr(path + 1, '/') != NULL)
2047 return (EINVAL);
2048 /*
2049 * "." and ".." are magic directories, populated on the fly, and cannot
2050 * be opened as queues.
2051 */
2052 if (strcmp(path, "/.") == 0 || strcmp(path, "/..") == 0)
2053 return (EINVAL);
2054 AUDIT_ARG_UPATH1_CANON(path);
2055
2056 error = falloc(td, &fp, &fd, O_CLOEXEC);
2057 if (error)
2058 return (error);
2059
2060 sx_xlock(&mqfs_data.mi_lock);
2061 pn = mqfs_search(mqfs_data.mi_root, path + 1, len - 1, td->td_ucred);
2062 if (pn == NULL) {
2063 if (!(flags & O_CREAT)) {
2064 error = ENOENT;
2065 } else {
2066 mq = mqueue_alloc(attr);
2067 if (mq == NULL) {
2068 error = ENFILE;
2069 } else {
2070 pn = mqfs_create_file(mqfs_data.mi_root,
2071 path + 1, len - 1, td->td_ucred,
2072 cmode);
2073 if (pn == NULL) {
2074 error = ENOSPC;
2075 mqueue_free(mq);
2076 }
2077 }
2078 }
2079
2080 if (error == 0) {
2081 pn->mn_data = mq;
2082 }
2083 } else {
2084 if ((flags & (O_CREAT | O_EXCL)) == (O_CREAT | O_EXCL)) {
2085 error = EEXIST;
2086 } else {
2087 accmode_t accmode = 0;
2088
2089 if (flags & FREAD)
2090 accmode |= VREAD;
2091 if (flags & FWRITE)
2092 accmode |= VWRITE;
2093 error = vaccess(VREG, pn->mn_mode, pn->mn_uid,
2094 pn->mn_gid, accmode, td->td_ucred);
2095 }
2096 }
2097
2098 if (error) {
2099 sx_xunlock(&mqfs_data.mi_lock);
2100 fdclose(td, fp, fd);
2101 fdrop(fp, td);
2102 return (error);
2103 }
2104
2105 mqnode_addref(pn);
2106 sx_xunlock(&mqfs_data.mi_lock);
2107
2108 finit(fp, flags & (FREAD | FWRITE | O_NONBLOCK), DTYPE_MQUEUE, pn,
2109 &mqueueops);
2110
2111 td->td_retval[0] = fd;
2112 fdrop(fp, td);
2113 return (0);
2114 }
2115
2116 /*
2117 * Syscall to open a message queue.
2118 */
2119 int
2120 sys_kmq_open(struct thread *td, struct kmq_open_args *uap)
2121 {
2122 struct mq_attr attr;
2123 int flags, error;
2124
2125 if ((uap->flags & O_ACCMODE) == O_ACCMODE || uap->flags & O_EXEC)
2126 return (EINVAL);
2127 flags = FFLAGS(uap->flags);
2128 if ((flags & O_CREAT) != 0 && uap->attr != NULL) {
2129 error = copyin(uap->attr, &attr, sizeof(attr));
2130 if (error)
2131 return (error);
2132 }
2133 return (kern_kmq_open(td, uap->path, flags, uap->mode,
2134 uap->attr != NULL ? &attr : NULL));
2135 }
2136
2137 /*
2138 * Syscall to unlink a message queue.
2139 */
2140 int
2141 sys_kmq_unlink(struct thread *td, struct kmq_unlink_args *uap)
2142 {
2143 char path[MQFS_NAMELEN+1];
2144 struct mqfs_node *pn;
2145 int error, len;
2146
2147 error = copyinstr(uap->path, path, MQFS_NAMELEN + 1, NULL);
2148 if (error)
2149 return (error);
2150
2151 len = strlen(path);
2152 if (len < 2 || path[0] != '/' || strchr(path + 1, '/') != NULL)
2153 return (EINVAL);
2154 if (strcmp(path, "/.") == 0 || strcmp(path, "/..") == 0)
2155 return (EINVAL);
2156 AUDIT_ARG_UPATH1_CANON(path);
2157
2158 sx_xlock(&mqfs_data.mi_lock);
2159 pn = mqfs_search(mqfs_data.mi_root, path + 1, len - 1, td->td_ucred);
2160 if (pn != NULL)
2161 error = do_unlink(pn, td->td_ucred);
2162 else
2163 error = ENOENT;
2164 sx_xunlock(&mqfs_data.mi_lock);
2165 return (error);
2166 }
2167
2168 typedef int (*_fgetf)(struct thread *, int, cap_rights_t *, struct file **);
2169
2170 /*
2171 * Get message queue by giving file slot
2172 */
2173 static int
2174 _getmq(struct thread *td, int fd, cap_rights_t *rightsp, _fgetf func,
2175 struct file **fpp, struct mqfs_node **ppn, struct mqueue **pmq)
2176 {
2177 struct mqfs_node *pn;
2178 int error;
2179
2180 error = func(td, fd, rightsp, fpp);
2181 if (error)
2182 return (error);
2183 if (&mqueueops != (*fpp)->f_ops) {
2184 fdrop(*fpp, td);
2185 return (EBADF);
2186 }
2187 pn = (*fpp)->f_data;
2188 if (ppn)
2189 *ppn = pn;
2190 if (pmq)
2191 *pmq = pn->mn_data;
2192 return (0);
2193 }
2194
2195 static __inline int
2196 getmq(struct thread *td, int fd, struct file **fpp, struct mqfs_node **ppn,
2197 struct mqueue **pmq)
2198 {
2199
2200 return _getmq(td, fd, &cap_event_rights, fget,
2201 fpp, ppn, pmq);
2202 }
2203
2204 static __inline int
2205 getmq_read(struct thread *td, int fd, struct file **fpp,
2206 struct mqfs_node **ppn, struct mqueue **pmq)
2207 {
2208
2209 return _getmq(td, fd, &cap_read_rights, fget_read,
2210 fpp, ppn, pmq);
2211 }
2212
2213 static __inline int
2214 getmq_write(struct thread *td, int fd, struct file **fpp,
2215 struct mqfs_node **ppn, struct mqueue **pmq)
2216 {
2217
2218 return _getmq(td, fd, &cap_write_rights, fget_write,
2219 fpp, ppn, pmq);
2220 }
2221
2222 static int
2223 kern_kmq_setattr(struct thread *td, int mqd, const struct mq_attr *attr,
2224 struct mq_attr *oattr)
2225 {
2226 struct mqueue *mq;
2227 struct file *fp;
2228 u_int oflag, flag;
2229 int error;
2230
2231 AUDIT_ARG_FD(mqd);
2232 if (attr != NULL && (attr->mq_flags & ~O_NONBLOCK) != 0)
2233 return (EINVAL);
2234 error = getmq(td, mqd, &fp, NULL, &mq);
2235 if (error)
2236 return (error);
2237 oattr->mq_maxmsg = mq->mq_maxmsg;
2238 oattr->mq_msgsize = mq->mq_msgsize;
2239 oattr->mq_curmsgs = mq->mq_curmsgs;
2240 if (attr != NULL) {
2241 do {
2242 oflag = flag = fp->f_flag;
2243 flag &= ~O_NONBLOCK;
2244 flag |= (attr->mq_flags & O_NONBLOCK);
2245 } while (atomic_cmpset_int(&fp->f_flag, oflag, flag) == 0);
2246 } else
2247 oflag = fp->f_flag;
2248 oattr->mq_flags = (O_NONBLOCK & oflag);
2249 fdrop(fp, td);
2250 return (error);
2251 }
2252
2253 int
2254 sys_kmq_setattr(struct thread *td, struct kmq_setattr_args *uap)
2255 {
2256 struct mq_attr attr, oattr;
2257 int error;
2258
2259 if (uap->attr != NULL) {
2260 error = copyin(uap->attr, &attr, sizeof(attr));
2261 if (error != 0)
2262 return (error);
2263 }
2264 error = kern_kmq_setattr(td, uap->mqd, uap->attr != NULL ? &attr : NULL,
2265 &oattr);
2266 if (error == 0 && uap->oattr != NULL) {
2267 bzero(oattr.__reserved, sizeof(oattr.__reserved));
2268 error = copyout(&oattr, uap->oattr, sizeof(oattr));
2269 }
2270 return (error);
2271 }
2272
2273 int
2274 sys_kmq_timedreceive(struct thread *td, struct kmq_timedreceive_args *uap)
2275 {
2276 struct mqueue *mq;
2277 struct file *fp;
2278 struct timespec *abs_timeout, ets;
2279 int error;
2280 int waitok;
2281
2282 AUDIT_ARG_FD(uap->mqd);
2283 error = getmq_read(td, uap->mqd, &fp, NULL, &mq);
2284 if (error)
2285 return (error);
2286 if (uap->abs_timeout != NULL) {
2287 error = copyin(uap->abs_timeout, &ets, sizeof(ets));
2288 if (error != 0)
2289 goto out;
2290 abs_timeout = &ets;
2291 } else
2292 abs_timeout = NULL;
2293 waitok = !(fp->f_flag & O_NONBLOCK);
2294 error = mqueue_receive(mq, uap->msg_ptr, uap->msg_len,
2295 uap->msg_prio, waitok, abs_timeout);
2296 out:
2297 fdrop(fp, td);
2298 return (error);
2299 }
2300
2301 int
2302 sys_kmq_timedsend(struct thread *td, struct kmq_timedsend_args *uap)
2303 {
2304 struct mqueue *mq;
2305 struct file *fp;
2306 struct timespec *abs_timeout, ets;
2307 int error, waitok;
2308
2309 AUDIT_ARG_FD(uap->mqd);
2310 error = getmq_write(td, uap->mqd, &fp, NULL, &mq);
2311 if (error)
2312 return (error);
2313 if (uap->abs_timeout != NULL) {
2314 error = copyin(uap->abs_timeout, &ets, sizeof(ets));
2315 if (error != 0)
2316 goto out;
2317 abs_timeout = &ets;
2318 } else
2319 abs_timeout = NULL;
2320 waitok = !(fp->f_flag & O_NONBLOCK);
2321 error = mqueue_send(mq, uap->msg_ptr, uap->msg_len,
2322 uap->msg_prio, waitok, abs_timeout);
2323 out:
2324 fdrop(fp, td);
2325 return (error);
2326 }
2327
2328 static int
2329 kern_kmq_notify(struct thread *td, int mqd, struct sigevent *sigev)
2330 {
2331 struct filedesc *fdp;
2332 struct proc *p;
2333 struct mqueue *mq;
2334 struct file *fp, *fp2;
2335 struct mqueue_notifier *nt, *newnt = NULL;
2336 int error;
2337
2338 AUDIT_ARG_FD(mqd);
2339 if (sigev != NULL) {
2340 if (sigev->sigev_notify != SIGEV_SIGNAL &&
2341 sigev->sigev_notify != SIGEV_THREAD_ID &&
2342 sigev->sigev_notify != SIGEV_NONE)
2343 return (EINVAL);
2344 if ((sigev->sigev_notify == SIGEV_SIGNAL ||
2345 sigev->sigev_notify == SIGEV_THREAD_ID) &&
2346 !_SIG_VALID(sigev->sigev_signo))
2347 return (EINVAL);
2348 }
2349 p = td->td_proc;
2350 fdp = td->td_proc->p_fd;
2351 error = getmq(td, mqd, &fp, NULL, &mq);
2352 if (error)
2353 return (error);
2354 again:
2355 FILEDESC_SLOCK(fdp);
2356 fp2 = fget_locked(fdp, mqd);
2357 if (fp2 == NULL) {
2358 FILEDESC_SUNLOCK(fdp);
2359 error = EBADF;
2360 goto out;
2361 }
2362 #ifdef CAPABILITIES
2363 error = cap_check(cap_rights(fdp, mqd), &cap_event_rights);
2364 if (error) {
2365 FILEDESC_SUNLOCK(fdp);
2366 goto out;
2367 }
2368 #endif
2369 if (fp2 != fp) {
2370 FILEDESC_SUNLOCK(fdp);
2371 error = EBADF;
2372 goto out;
2373 }
2374 mtx_lock(&mq->mq_mutex);
2375 FILEDESC_SUNLOCK(fdp);
2376 if (sigev != NULL) {
2377 if (mq->mq_notifier != NULL) {
2378 error = EBUSY;
2379 } else {
2380 PROC_LOCK(p);
2381 nt = notifier_search(p, mqd);
2382 if (nt == NULL) {
2383 if (newnt == NULL) {
2384 PROC_UNLOCK(p);
2385 mtx_unlock(&mq->mq_mutex);
2386 newnt = notifier_alloc();
2387 goto again;
2388 }
2389 }
2390
2391 if (nt != NULL) {
2392 sigqueue_take(&nt->nt_ksi);
2393 if (newnt != NULL) {
2394 notifier_free(newnt);
2395 newnt = NULL;
2396 }
2397 } else {
2398 nt = newnt;
2399 newnt = NULL;
2400 ksiginfo_init(&nt->nt_ksi);
2401 nt->nt_ksi.ksi_flags |= KSI_INS | KSI_EXT;
2402 nt->nt_ksi.ksi_code = SI_MESGQ;
2403 nt->nt_proc = p;
2404 nt->nt_ksi.ksi_mqd = mqd;
2405 notifier_insert(p, nt);
2406 }
2407 nt->nt_sigev = *sigev;
2408 mq->mq_notifier = nt;
2409 PROC_UNLOCK(p);
2410 /*
2411 * if there is no receivers and message queue
2412 * is not empty, we should send notification
2413 * as soon as possible.
2414 */
2415 if (mq->mq_receivers == 0 &&
2416 !TAILQ_EMPTY(&mq->mq_msgq))
2417 mqueue_send_notification(mq);
2418 }
2419 } else {
2420 notifier_remove(p, mq, mqd);
2421 }
2422 mtx_unlock(&mq->mq_mutex);
2423
2424 out:
2425 fdrop(fp, td);
2426 if (newnt != NULL)
2427 notifier_free(newnt);
2428 return (error);
2429 }
2430
2431 int
2432 sys_kmq_notify(struct thread *td, struct kmq_notify_args *uap)
2433 {
2434 struct sigevent ev, *evp;
2435 int error;
2436
2437 if (uap->sigev == NULL) {
2438 evp = NULL;
2439 } else {
2440 error = copyin(uap->sigev, &ev, sizeof(ev));
2441 if (error != 0)
2442 return (error);
2443 evp = &ev;
2444 }
2445 return (kern_kmq_notify(td, uap->mqd, evp));
2446 }
2447
2448 static void
2449 mqueue_fdclose(struct thread *td, int fd, struct file *fp)
2450 {
2451 struct mqueue *mq;
2452 #ifdef INVARIANTS
2453 struct filedesc *fdp;
2454
2455 fdp = td->td_proc->p_fd;
2456 FILEDESC_LOCK_ASSERT(fdp);
2457 #endif
2458
2459 if (fp->f_ops == &mqueueops) {
2460 mq = FPTOMQ(fp);
2461 mtx_lock(&mq->mq_mutex);
2462 notifier_remove(td->td_proc, mq, fd);
2463
2464 /* have to wakeup thread in same process */
2465 if (mq->mq_flags & MQ_RSEL) {
2466 mq->mq_flags &= ~MQ_RSEL;
2467 selwakeup(&mq->mq_rsel);
2468 }
2469 if (mq->mq_flags & MQ_WSEL) {
2470 mq->mq_flags &= ~MQ_WSEL;
2471 selwakeup(&mq->mq_wsel);
2472 }
2473 mtx_unlock(&mq->mq_mutex);
2474 }
2475 }
2476
2477 static void
2478 mq_proc_exit(void *arg __unused, struct proc *p)
2479 {
2480 struct filedesc *fdp;
2481 struct file *fp;
2482 struct mqueue *mq;
2483 int i;
2484
2485 fdp = p->p_fd;
2486 FILEDESC_SLOCK(fdp);
2487 for (i = 0; i < fdp->fd_nfiles; ++i) {
2488 fp = fget_locked(fdp, i);
2489 if (fp != NULL && fp->f_ops == &mqueueops) {
2490 mq = FPTOMQ(fp);
2491 mtx_lock(&mq->mq_mutex);
2492 notifier_remove(p, FPTOMQ(fp), i);
2493 mtx_unlock(&mq->mq_mutex);
2494 }
2495 }
2496 FILEDESC_SUNLOCK(fdp);
2497 KASSERT(LIST_EMPTY(&p->p_mqnotifier), ("mq notifiers left"));
2498 }
2499
2500 static int
2501 mqf_poll(struct file *fp, int events, struct ucred *active_cred,
2502 struct thread *td)
2503 {
2504 struct mqueue *mq = FPTOMQ(fp);
2505 int revents = 0;
2506
2507 mtx_lock(&mq->mq_mutex);
2508 if (events & (POLLIN | POLLRDNORM)) {
2509 if (mq->mq_curmsgs) {
2510 revents |= events & (POLLIN | POLLRDNORM);
2511 } else {
2512 mq->mq_flags |= MQ_RSEL;
2513 selrecord(td, &mq->mq_rsel);
2514 }
2515 }
2516 if (events & POLLOUT) {
2517 if (mq->mq_curmsgs < mq->mq_maxmsg)
2518 revents |= POLLOUT;
2519 else {
2520 mq->mq_flags |= MQ_WSEL;
2521 selrecord(td, &mq->mq_wsel);
2522 }
2523 }
2524 mtx_unlock(&mq->mq_mutex);
2525 return (revents);
2526 }
2527
2528 static int
2529 mqf_close(struct file *fp, struct thread *td)
2530 {
2531 struct mqfs_node *pn;
2532
2533 fp->f_ops = &badfileops;
2534 pn = fp->f_data;
2535 fp->f_data = NULL;
2536 sx_xlock(&mqfs_data.mi_lock);
2537 mqnode_release(pn);
2538 sx_xunlock(&mqfs_data.mi_lock);
2539 return (0);
2540 }
2541
2542 static int
2543 mqf_stat(struct file *fp, struct stat *st, struct ucred *active_cred,
2544 struct thread *td)
2545 {
2546 struct mqfs_node *pn = fp->f_data;
2547
2548 bzero(st, sizeof *st);
2549 sx_xlock(&mqfs_data.mi_lock);
2550 st->st_atim = pn->mn_atime;
2551 st->st_mtim = pn->mn_mtime;
2552 st->st_ctim = pn->mn_ctime;
2553 st->st_birthtim = pn->mn_birth;
2554 st->st_uid = pn->mn_uid;
2555 st->st_gid = pn->mn_gid;
2556 st->st_mode = S_IFIFO | pn->mn_mode;
2557 sx_xunlock(&mqfs_data.mi_lock);
2558 return (0);
2559 }
2560
2561 static int
2562 mqf_chmod(struct file *fp, mode_t mode, struct ucred *active_cred,
2563 struct thread *td)
2564 {
2565 struct mqfs_node *pn;
2566 int error;
2567
2568 error = 0;
2569 pn = fp->f_data;
2570 sx_xlock(&mqfs_data.mi_lock);
2571 error = vaccess(VREG, pn->mn_mode, pn->mn_uid, pn->mn_gid, VADMIN,
2572 active_cred);
2573 if (error != 0)
2574 goto out;
2575 pn->mn_mode = mode & ACCESSPERMS;
2576 out:
2577 sx_xunlock(&mqfs_data.mi_lock);
2578 return (error);
2579 }
2580
2581 static int
2582 mqf_chown(struct file *fp, uid_t uid, gid_t gid, struct ucred *active_cred,
2583 struct thread *td)
2584 {
2585 struct mqfs_node *pn;
2586 int error;
2587
2588 error = 0;
2589 pn = fp->f_data;
2590 sx_xlock(&mqfs_data.mi_lock);
2591 if (uid == (uid_t)-1)
2592 uid = pn->mn_uid;
2593 if (gid == (gid_t)-1)
2594 gid = pn->mn_gid;
2595 if (((uid != pn->mn_uid && uid != active_cred->cr_uid) ||
2596 (gid != pn->mn_gid && !groupmember(gid, active_cred))) &&
2597 (error = priv_check_cred(active_cred, PRIV_VFS_CHOWN)))
2598 goto out;
2599 pn->mn_uid = uid;
2600 pn->mn_gid = gid;
2601 out:
2602 sx_xunlock(&mqfs_data.mi_lock);
2603 return (error);
2604 }
2605
2606 static int
2607 mqf_kqfilter(struct file *fp, struct knote *kn)
2608 {
2609 struct mqueue *mq = FPTOMQ(fp);
2610 int error = 0;
2611
2612 if (kn->kn_filter == EVFILT_READ) {
2613 kn->kn_fop = &mq_rfiltops;
2614 knlist_add(&mq->mq_rsel.si_note, kn, 0);
2615 } else if (kn->kn_filter == EVFILT_WRITE) {
2616 kn->kn_fop = &mq_wfiltops;
2617 knlist_add(&mq->mq_wsel.si_note, kn, 0);
2618 } else
2619 error = EINVAL;
2620 return (error);
2621 }
2622
2623 static void
2624 filt_mqdetach(struct knote *kn)
2625 {
2626 struct mqueue *mq = FPTOMQ(kn->kn_fp);
2627
2628 if (kn->kn_filter == EVFILT_READ)
2629 knlist_remove(&mq->mq_rsel.si_note, kn, 0);
2630 else if (kn->kn_filter == EVFILT_WRITE)
2631 knlist_remove(&mq->mq_wsel.si_note, kn, 0);
2632 else
2633 panic("filt_mqdetach");
2634 }
2635
2636 static int
2637 filt_mqread(struct knote *kn, long hint)
2638 {
2639 struct mqueue *mq = FPTOMQ(kn->kn_fp);
2640
2641 mtx_assert(&mq->mq_mutex, MA_OWNED);
2642 return (mq->mq_curmsgs != 0);
2643 }
2644
2645 static int
2646 filt_mqwrite(struct knote *kn, long hint)
2647 {
2648 struct mqueue *mq = FPTOMQ(kn->kn_fp);
2649
2650 mtx_assert(&mq->mq_mutex, MA_OWNED);
2651 return (mq->mq_curmsgs < mq->mq_maxmsg);
2652 }
2653
2654 static int
2655 mqf_fill_kinfo(struct file *fp, struct kinfo_file *kif, struct filedesc *fdp)
2656 {
2657
2658 kif->kf_type = KF_TYPE_MQUEUE;
2659 return (0);
2660 }
2661
2662 static struct fileops mqueueops = {
2663 .fo_read = invfo_rdwr,
2664 .fo_write = invfo_rdwr,
2665 .fo_truncate = invfo_truncate,
2666 .fo_ioctl = invfo_ioctl,
2667 .fo_poll = mqf_poll,
2668 .fo_kqfilter = mqf_kqfilter,
2669 .fo_stat = mqf_stat,
2670 .fo_close = mqf_close,
2671 .fo_chmod = mqf_chmod,
2672 .fo_chown = mqf_chown,
2673 .fo_sendfile = invfo_sendfile,
2674 .fo_fill_kinfo = mqf_fill_kinfo,
2675 .fo_flags = DFLAG_PASSABLE,
2676 };
2677
2678 static struct vop_vector mqfs_vnodeops = {
2679 .vop_default = &default_vnodeops,
2680 .vop_access = mqfs_access,
2681 .vop_cachedlookup = mqfs_lookup,
2682 .vop_lookup = vfs_cache_lookup,
2683 .vop_reclaim = mqfs_reclaim,
2684 .vop_create = mqfs_create,
2685 .vop_remove = mqfs_remove,
2686 .vop_inactive = mqfs_inactive,
2687 .vop_open = mqfs_open,
2688 .vop_close = mqfs_close,
2689 .vop_getattr = mqfs_getattr,
2690 .vop_setattr = mqfs_setattr,
2691 .vop_read = mqfs_read,
2692 .vop_write = VOP_EOPNOTSUPP,
2693 .vop_readdir = mqfs_readdir,
2694 .vop_mkdir = VOP_EOPNOTSUPP,
2695 .vop_rmdir = VOP_EOPNOTSUPP
2696 };
2697 VFS_VOP_VECTOR_REGISTER(mqfs_vnodeops);
2698
2699 static struct vfsops mqfs_vfsops = {
2700 .vfs_init = mqfs_init,
2701 .vfs_uninit = mqfs_uninit,
2702 .vfs_mount = mqfs_mount,
2703 .vfs_unmount = mqfs_unmount,
2704 .vfs_root = mqfs_root,
2705 .vfs_statfs = mqfs_statfs,
2706 };
2707
2708 static struct vfsconf mqueuefs_vfsconf = {
2709 .vfc_version = VFS_VERSION,
2710 .vfc_name = "mqueuefs",
2711 .vfc_vfsops = &mqfs_vfsops,
2712 .vfc_typenum = -1,
2713 .vfc_flags = VFCF_SYNTHETIC
2714 };
2715
2716 static struct syscall_helper_data mq_syscalls[] = {
2717 SYSCALL_INIT_HELPER(kmq_open),
2718 SYSCALL_INIT_HELPER_F(kmq_setattr, SYF_CAPENABLED),
2719 SYSCALL_INIT_HELPER_F(kmq_timedsend, SYF_CAPENABLED),
2720 SYSCALL_INIT_HELPER_F(kmq_timedreceive, SYF_CAPENABLED),
2721 SYSCALL_INIT_HELPER_F(kmq_notify, SYF_CAPENABLED),
2722 SYSCALL_INIT_HELPER(kmq_unlink),
2723 SYSCALL_INIT_LAST
2724 };
2725
2726 #ifdef COMPAT_FREEBSD32
2727 #include <compat/freebsd32/freebsd32.h>
2728 #include <compat/freebsd32/freebsd32_proto.h>
2729 #include <compat/freebsd32/freebsd32_signal.h>
2730 #include <compat/freebsd32/freebsd32_syscall.h>
2731 #include <compat/freebsd32/freebsd32_util.h>
2732
2733 static void
2734 mq_attr_from32(const struct mq_attr32 *from, struct mq_attr *to)
2735 {
2736
2737 to->mq_flags = from->mq_flags;
2738 to->mq_maxmsg = from->mq_maxmsg;
2739 to->mq_msgsize = from->mq_msgsize;
2740 to->mq_curmsgs = from->mq_curmsgs;
2741 }
2742
2743 static void
2744 mq_attr_to32(const struct mq_attr *from, struct mq_attr32 *to)
2745 {
2746
2747 to->mq_flags = from->mq_flags;
2748 to->mq_maxmsg = from->mq_maxmsg;
2749 to->mq_msgsize = from->mq_msgsize;
2750 to->mq_curmsgs = from->mq_curmsgs;
2751 }
2752
2753 int
2754 freebsd32_kmq_open(struct thread *td, struct freebsd32_kmq_open_args *uap)
2755 {
2756 struct mq_attr attr;
2757 struct mq_attr32 attr32;
2758 int flags, error;
2759
2760 if ((uap->flags & O_ACCMODE) == O_ACCMODE || uap->flags & O_EXEC)
2761 return (EINVAL);
2762 flags = FFLAGS(uap->flags);
2763 if ((flags & O_CREAT) != 0 && uap->attr != NULL) {
2764 error = copyin(uap->attr, &attr32, sizeof(attr32));
2765 if (error)
2766 return (error);
2767 mq_attr_from32(&attr32, &attr);
2768 }
2769 return (kern_kmq_open(td, uap->path, flags, uap->mode,
2770 uap->attr != NULL ? &attr : NULL));
2771 }
2772
2773 int
2774 freebsd32_kmq_setattr(struct thread *td, struct freebsd32_kmq_setattr_args *uap)
2775 {
2776 struct mq_attr attr, oattr;
2777 struct mq_attr32 attr32, oattr32;
2778 int error;
2779
2780 if (uap->attr != NULL) {
2781 error = copyin(uap->attr, &attr32, sizeof(attr32));
2782 if (error != 0)
2783 return (error);
2784 mq_attr_from32(&attr32, &attr);
2785 }
2786 error = kern_kmq_setattr(td, uap->mqd, uap->attr != NULL ? &attr : NULL,
2787 &oattr);
2788 if (error == 0 && uap->oattr != NULL) {
2789 mq_attr_to32(&oattr, &oattr32);
2790 bzero(oattr32.__reserved, sizeof(oattr32.__reserved));
2791 error = copyout(&oattr32, uap->oattr, sizeof(oattr32));
2792 }
2793 return (error);
2794 }
2795
2796 int
2797 freebsd32_kmq_timedsend(struct thread *td,
2798 struct freebsd32_kmq_timedsend_args *uap)
2799 {
2800 struct mqueue *mq;
2801 struct file *fp;
2802 struct timespec32 ets32;
2803 struct timespec *abs_timeout, ets;
2804 int error;
2805 int waitok;
2806
2807 AUDIT_ARG_FD(uap->mqd);
2808 error = getmq_write(td, uap->mqd, &fp, NULL, &mq);
2809 if (error)
2810 return (error);
2811 if (uap->abs_timeout != NULL) {
2812 error = copyin(uap->abs_timeout, &ets32, sizeof(ets32));
2813 if (error != 0)
2814 goto out;
2815 CP(ets32, ets, tv_sec);
2816 CP(ets32, ets, tv_nsec);
2817 abs_timeout = &ets;
2818 } else
2819 abs_timeout = NULL;
2820 waitok = !(fp->f_flag & O_NONBLOCK);
2821 error = mqueue_send(mq, uap->msg_ptr, uap->msg_len,
2822 uap->msg_prio, waitok, abs_timeout);
2823 out:
2824 fdrop(fp, td);
2825 return (error);
2826 }
2827
2828 int
2829 freebsd32_kmq_timedreceive(struct thread *td,
2830 struct freebsd32_kmq_timedreceive_args *uap)
2831 {
2832 struct mqueue *mq;
2833 struct file *fp;
2834 struct timespec32 ets32;
2835 struct timespec *abs_timeout, ets;
2836 int error, waitok;
2837
2838 AUDIT_ARG_FD(uap->mqd);
2839 error = getmq_read(td, uap->mqd, &fp, NULL, &mq);
2840 if (error)
2841 return (error);
2842 if (uap->abs_timeout != NULL) {
2843 error = copyin(uap->abs_timeout, &ets32, sizeof(ets32));
2844 if (error != 0)
2845 goto out;
2846 CP(ets32, ets, tv_sec);
2847 CP(ets32, ets, tv_nsec);
2848 abs_timeout = &ets;
2849 } else
2850 abs_timeout = NULL;
2851 waitok = !(fp->f_flag & O_NONBLOCK);
2852 error = mqueue_receive(mq, uap->msg_ptr, uap->msg_len,
2853 uap->msg_prio, waitok, abs_timeout);
2854 out:
2855 fdrop(fp, td);
2856 return (error);
2857 }
2858
2859 int
2860 freebsd32_kmq_notify(struct thread *td, struct freebsd32_kmq_notify_args *uap)
2861 {
2862 struct sigevent ev, *evp;
2863 struct sigevent32 ev32;
2864 int error;
2865
2866 if (uap->sigev == NULL) {
2867 evp = NULL;
2868 } else {
2869 error = copyin(uap->sigev, &ev32, sizeof(ev32));
2870 if (error != 0)
2871 return (error);
2872 error = convert_sigevent32(&ev32, &ev);
2873 if (error != 0)
2874 return (error);
2875 evp = &ev;
2876 }
2877 return (kern_kmq_notify(td, uap->mqd, evp));
2878 }
2879
2880 static struct syscall_helper_data mq32_syscalls[] = {
2881 SYSCALL32_INIT_HELPER(freebsd32_kmq_open),
2882 SYSCALL32_INIT_HELPER_F(freebsd32_kmq_setattr, SYF_CAPENABLED),
2883 SYSCALL32_INIT_HELPER_F(freebsd32_kmq_timedsend, SYF_CAPENABLED),
2884 SYSCALL32_INIT_HELPER_F(freebsd32_kmq_timedreceive, SYF_CAPENABLED),
2885 SYSCALL32_INIT_HELPER_F(freebsd32_kmq_notify, SYF_CAPENABLED),
2886 SYSCALL32_INIT_HELPER_COMPAT(kmq_unlink),
2887 SYSCALL_INIT_LAST
2888 };
2889 #endif
2890
2891 static int
2892 mqinit(void)
2893 {
2894 int error;
2895
2896 error = syscall_helper_register(mq_syscalls, SY_THR_STATIC_KLD);
2897 if (error != 0)
2898 return (error);
2899 #ifdef COMPAT_FREEBSD32
2900 error = syscall32_helper_register(mq32_syscalls, SY_THR_STATIC_KLD);
2901 if (error != 0)
2902 return (error);
2903 #endif
2904 return (0);
2905 }
2906
2907 static int
2908 mqunload(void)
2909 {
2910
2911 #ifdef COMPAT_FREEBSD32
2912 syscall32_helper_unregister(mq32_syscalls);
2913 #endif
2914 syscall_helper_unregister(mq_syscalls);
2915 return (0);
2916 }
2917
2918 static int
2919 mq_modload(struct module *module, int cmd, void *arg)
2920 {
2921 int error = 0;
2922
2923 error = vfs_modevent(module, cmd, arg);
2924 if (error != 0)
2925 return (error);
2926
2927 switch (cmd) {
2928 case MOD_LOAD:
2929 error = mqinit();
2930 if (error != 0)
2931 mqunload();
2932 break;
2933 case MOD_UNLOAD:
2934 error = mqunload();
2935 break;
2936 default:
2937 break;
2938 }
2939 return (error);
2940 }
2941
2942 static moduledata_t mqueuefs_mod = {
2943 "mqueuefs",
2944 mq_modload,
2945 &mqueuefs_vfsconf
2946 };
2947 DECLARE_MODULE(mqueuefs, mqueuefs_mod, SI_SUB_VFS, SI_ORDER_MIDDLE);
2948 MODULE_VERSION(mqueuefs, 1);
Cache object: 3945dd3670c5a8fae9c20f2775e1e801
|