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