1 /*- 2 * SPDX-License-Identifier: BSD-3-Clause 3 * 4 * Copyright (c) 1999-2004 Poul-Henning Kamp 5 * Copyright (c) 1999 Michael Smith 6 * Copyright (c) 1989, 1993 7 * The Regents of the University of California. All rights reserved. 8 * (c) UNIX System Laboratories, Inc. 9 * All or some portions of this file are derived from material licensed 10 * to the University of California by American Telephone and Telegraph 11 * Co. or Unix System Laboratories, Inc. and are reproduced herein with 12 * the permission of UNIX System Laboratories, Inc. 13 * 14 * Redistribution and use in source and binary forms, with or without 15 * modification, are permitted provided that the following conditions 16 * are met: 17 * 1. Redistributions of source code must retain the above copyright 18 * notice, this list of conditions and the following disclaimer. 19 * 2. Redistributions in binary form must reproduce the above copyright 20 * notice, this list of conditions and the following disclaimer in the 21 * documentation and/or other materials provided with the distribution. 22 * 3. Neither the name of the University nor the names of its contributors 23 * may be used to endorse or promote products derived from this software 24 * without specific prior written permission. 25 * 26 * THIS SOFTWARE IS PROVIDED BY THE AUTHOR AND CONTRIBUTORS ``AS IS'' AND 27 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE 28 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE 29 * ARE DISCLAIMED. IN NO EVENT SHALL THE AUTHOR OR CONTRIBUTORS BE LIABLE 30 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL 31 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS 32 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) 33 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT 34 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY 35 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF 36 * SUCH DAMAGE. 37 */ 38 39 #include <sys/cdefs.h> 40 __FBSDID("$FreeBSD$"); 41 42 #include <sys/param.h> 43 #include <sys/conf.h> 44 #include <sys/eventhandler.h> 45 #include <sys/fcntl.h> 46 #include <sys/jail.h> 47 #include <sys/kernel.h> 48 #include <sys/libkern.h> 49 #include <sys/malloc.h> 50 #include <sys/mount.h> 51 #include <sys/mutex.h> 52 #include <sys/namei.h> 53 #include <sys/priv.h> 54 #include <sys/proc.h> 55 #include <sys/filedesc.h> 56 #include <sys/reboot.h> 57 #include <sys/sbuf.h> 58 #include <sys/syscallsubr.h> 59 #include <sys/sysproto.h> 60 #include <sys/sx.h> 61 #include <sys/sysctl.h> 62 #include <sys/sysent.h> 63 #include <sys/systm.h> 64 #include <sys/vnode.h> 65 #include <vm/uma.h> 66 67 #include <geom/geom.h> 68 69 #include <machine/stdarg.h> 70 71 #include <security/audit/audit.h> 72 #include <security/mac/mac_framework.h> 73 74 #define VFS_MOUNTARG_SIZE_MAX (1024 * 64) 75 76 static int vfs_domount(struct thread *td, const char *fstype, char *fspath, 77 uint64_t fsflags, struct vfsoptlist **optlist); 78 static void free_mntarg(struct mntarg *ma); 79 80 static int usermount = 0; 81 SYSCTL_INT(_vfs, OID_AUTO, usermount, CTLFLAG_RW, &usermount, 0, 82 "Unprivileged users may mount and unmount file systems"); 83 84 static bool default_autoro = false; 85 SYSCTL_BOOL(_vfs, OID_AUTO, default_autoro, CTLFLAG_RW, &default_autoro, 0, 86 "Retry failed r/w mount as r/o if no explicit ro/rw option is specified"); 87 88 MALLOC_DEFINE(M_MOUNT, "mount", "vfs mount structure"); 89 MALLOC_DEFINE(M_STATFS, "statfs", "statfs structure"); 90 static uma_zone_t mount_zone; 91 92 /* List of mounted filesystems. */ 93 struct mntlist mountlist = TAILQ_HEAD_INITIALIZER(mountlist); 94 95 /* For any iteration/modification of mountlist */ 96 struct mtx mountlist_mtx; 97 MTX_SYSINIT(mountlist, &mountlist_mtx, "mountlist", MTX_DEF); 98 99 EVENTHANDLER_LIST_DEFINE(vfs_mounted); 100 EVENTHANDLER_LIST_DEFINE(vfs_unmounted); 101 102 /* 103 * Global opts, taken by all filesystems 104 */ 105 static const char *global_opts[] = { 106 "errmsg", 107 "fstype", 108 "fspath", 109 "ro", 110 "rw", 111 "nosuid", 112 "noexec", 113 NULL 114 }; 115 116 static int 117 mount_init(void *mem, int size, int flags) 118 { 119 struct mount *mp; 120 121 mp = (struct mount *)mem; 122 mtx_init(&mp->mnt_mtx, "struct mount mtx", NULL, MTX_DEF); 123 mtx_init(&mp->mnt_listmtx, "struct mount vlist mtx", NULL, MTX_DEF); 124 lockinit(&mp->mnt_explock, PVFS, "explock", 0, 0); 125 return (0); 126 } 127 128 static void 129 mount_fini(void *mem, int size) 130 { 131 struct mount *mp; 132 133 mp = (struct mount *)mem; 134 lockdestroy(&mp->mnt_explock); 135 mtx_destroy(&mp->mnt_listmtx); 136 mtx_destroy(&mp->mnt_mtx); 137 } 138 139 static void 140 vfs_mount_init(void *dummy __unused) 141 { 142 143 mount_zone = uma_zcreate("Mountpoints", sizeof(struct mount), NULL, 144 NULL, mount_init, mount_fini, UMA_ALIGN_PTR, UMA_ZONE_NOFREE); 145 } 146 SYSINIT(vfs_mount, SI_SUB_VFS, SI_ORDER_ANY, vfs_mount_init, NULL); 147 148 /* 149 * --------------------------------------------------------------------- 150 * Functions for building and sanitizing the mount options 151 */ 152 153 /* Remove one mount option. */ 154 static void 155 vfs_freeopt(struct vfsoptlist *opts, struct vfsopt *opt) 156 { 157 158 TAILQ_REMOVE(opts, opt, link); 159 free(opt->name, M_MOUNT); 160 if (opt->value != NULL) 161 free(opt->value, M_MOUNT); 162 free(opt, M_MOUNT); 163 } 164 165 /* Release all resources related to the mount options. */ 166 void 167 vfs_freeopts(struct vfsoptlist *opts) 168 { 169 struct vfsopt *opt; 170 171 while (!TAILQ_EMPTY(opts)) { 172 opt = TAILQ_FIRST(opts); 173 vfs_freeopt(opts, opt); 174 } 175 free(opts, M_MOUNT); 176 } 177 178 void 179 vfs_deleteopt(struct vfsoptlist *opts, const char *name) 180 { 181 struct vfsopt *opt, *temp; 182 183 if (opts == NULL) 184 return; 185 TAILQ_FOREACH_SAFE(opt, opts, link, temp) { 186 if (strcmp(opt->name, name) == 0) 187 vfs_freeopt(opts, opt); 188 } 189 } 190 191 static int 192 vfs_isopt_ro(const char *opt) 193 { 194 195 if (strcmp(opt, "ro") == 0 || strcmp(opt, "rdonly") == 0 || 196 strcmp(opt, "norw") == 0) 197 return (1); 198 return (0); 199 } 200 201 static int 202 vfs_isopt_rw(const char *opt) 203 { 204 205 if (strcmp(opt, "rw") == 0 || strcmp(opt, "noro") == 0) 206 return (1); 207 return (0); 208 } 209 210 /* 211 * Check if options are equal (with or without the "no" prefix). 212 */ 213 static int 214 vfs_equalopts(const char *opt1, const char *opt2) 215 { 216 char *p; 217 218 /* "opt" vs. "opt" or "noopt" vs. "noopt" */ 219 if (strcmp(opt1, opt2) == 0) 220 return (1); 221 /* "noopt" vs. "opt" */ 222 if (strncmp(opt1, "no", 2) == 0 && strcmp(opt1 + 2, opt2) == 0) 223 return (1); 224 /* "opt" vs. "noopt" */ 225 if (strncmp(opt2, "no", 2) == 0 && strcmp(opt1, opt2 + 2) == 0) 226 return (1); 227 while ((p = strchr(opt1, '.')) != NULL && 228 !strncmp(opt1, opt2, ++p - opt1)) { 229 opt2 += p - opt1; 230 opt1 = p; 231 /* "foo.noopt" vs. "foo.opt" */ 232 if (strncmp(opt1, "no", 2) == 0 && strcmp(opt1 + 2, opt2) == 0) 233 return (1); 234 /* "foo.opt" vs. "foo.noopt" */ 235 if (strncmp(opt2, "no", 2) == 0 && strcmp(opt1, opt2 + 2) == 0) 236 return (1); 237 } 238 /* "ro" / "rdonly" / "norw" / "rw" / "noro" */ 239 if ((vfs_isopt_ro(opt1) || vfs_isopt_rw(opt1)) && 240 (vfs_isopt_ro(opt2) || vfs_isopt_rw(opt2))) 241 return (1); 242 return (0); 243 } 244 245 /* 246 * If a mount option is specified several times, 247 * (with or without the "no" prefix) only keep 248 * the last occurrence of it. 249 */ 250 static void 251 vfs_sanitizeopts(struct vfsoptlist *opts) 252 { 253 struct vfsopt *opt, *opt2, *tmp; 254 255 TAILQ_FOREACH_REVERSE(opt, opts, vfsoptlist, link) { 256 opt2 = TAILQ_PREV(opt, vfsoptlist, link); 257 while (opt2 != NULL) { 258 if (vfs_equalopts(opt->name, opt2->name)) { 259 tmp = TAILQ_PREV(opt2, vfsoptlist, link); 260 vfs_freeopt(opts, opt2); 261 opt2 = tmp; 262 } else { 263 opt2 = TAILQ_PREV(opt2, vfsoptlist, link); 264 } 265 } 266 } 267 } 268 269 /* 270 * Build a linked list of mount options from a struct uio. 271 */ 272 int 273 vfs_buildopts(struct uio *auio, struct vfsoptlist **options) 274 { 275 struct vfsoptlist *opts; 276 struct vfsopt *opt; 277 size_t memused, namelen, optlen; 278 unsigned int i, iovcnt; 279 int error; 280 281 opts = malloc(sizeof(struct vfsoptlist), M_MOUNT, M_WAITOK); 282 TAILQ_INIT(opts); 283 memused = 0; 284 iovcnt = auio->uio_iovcnt; 285 for (i = 0; i < iovcnt; i += 2) { 286 namelen = auio->uio_iov[i].iov_len; 287 optlen = auio->uio_iov[i + 1].iov_len; 288 memused += sizeof(struct vfsopt) + optlen + namelen; 289 /* 290 * Avoid consuming too much memory, and attempts to overflow 291 * memused. 292 */ 293 if (memused > VFS_MOUNTARG_SIZE_MAX || 294 optlen > VFS_MOUNTARG_SIZE_MAX || 295 namelen > VFS_MOUNTARG_SIZE_MAX) { 296 error = EINVAL; 297 goto bad; 298 } 299 300 opt = malloc(sizeof(struct vfsopt), M_MOUNT, M_WAITOK); 301 opt->name = malloc(namelen, M_MOUNT, M_WAITOK); 302 opt->value = NULL; 303 opt->len = 0; 304 opt->pos = i / 2; 305 opt->seen = 0; 306 307 /* 308 * Do this early, so jumps to "bad" will free the current 309 * option. 310 */ 311 TAILQ_INSERT_TAIL(opts, opt, link); 312 313 if (auio->uio_segflg == UIO_SYSSPACE) { 314 bcopy(auio->uio_iov[i].iov_base, opt->name, namelen); 315 } else { 316 error = copyin(auio->uio_iov[i].iov_base, opt->name, 317 namelen); 318 if (error) 319 goto bad; 320 } 321 /* Ensure names are null-terminated strings. */ 322 if (namelen == 0 || opt->name[namelen - 1] != '\0') { 323 error = EINVAL; 324 goto bad; 325 } 326 if (optlen != 0) { 327 opt->len = optlen; 328 opt->value = malloc(optlen, M_MOUNT, M_WAITOK); 329 if (auio->uio_segflg == UIO_SYSSPACE) { 330 bcopy(auio->uio_iov[i + 1].iov_base, opt->value, 331 optlen); 332 } else { 333 error = copyin(auio->uio_iov[i + 1].iov_base, 334 opt->value, optlen); 335 if (error) 336 goto bad; 337 } 338 } 339 } 340 vfs_sanitizeopts(opts); 341 *options = opts; 342 return (0); 343 bad: 344 vfs_freeopts(opts); 345 return (error); 346 } 347 348 /* 349 * Merge the old mount options with the new ones passed 350 * in the MNT_UPDATE case. 351 * 352 * XXX: This function will keep a "nofoo" option in the new 353 * options. E.g, if the option's canonical name is "foo", 354 * "nofoo" ends up in the mount point's active options. 355 */ 356 static void 357 vfs_mergeopts(struct vfsoptlist *toopts, struct vfsoptlist *oldopts) 358 { 359 struct vfsopt *opt, *new; 360 361 TAILQ_FOREACH(opt, oldopts, link) { 362 new = malloc(sizeof(struct vfsopt), M_MOUNT, M_WAITOK); 363 new->name = strdup(opt->name, M_MOUNT); 364 if (opt->len != 0) { 365 new->value = malloc(opt->len, M_MOUNT, M_WAITOK); 366 bcopy(opt->value, new->value, opt->len); 367 } else 368 new->value = NULL; 369 new->len = opt->len; 370 new->seen = opt->seen; 371 TAILQ_INSERT_HEAD(toopts, new, link); 372 } 373 vfs_sanitizeopts(toopts); 374 } 375 376 /* 377 * Mount a filesystem. 378 */ 379 #ifndef _SYS_SYSPROTO_H_ 380 struct nmount_args { 381 struct iovec *iovp; 382 unsigned int iovcnt; 383 int flags; 384 }; 385 #endif 386 int 387 sys_nmount(struct thread *td, struct nmount_args *uap) 388 { 389 struct uio *auio; 390 int error; 391 u_int iovcnt; 392 uint64_t flags; 393 394 /* 395 * Mount flags are now 64-bits. On 32-bit archtectures only 396 * 32-bits are passed in, but from here on everything handles 397 * 64-bit flags correctly. 398 */ 399 flags = uap->flags; 400 401 AUDIT_ARG_FFLAGS(flags); 402 CTR4(KTR_VFS, "%s: iovp %p with iovcnt %d and flags %d", __func__, 403 uap->iovp, uap->iovcnt, flags); 404 405 /* 406 * Filter out MNT_ROOTFS. We do not want clients of nmount() in 407 * userspace to set this flag, but we must filter it out if we want 408 * MNT_UPDATE on the root file system to work. 409 * MNT_ROOTFS should only be set by the kernel when mounting its 410 * root file system. 411 */ 412 flags &= ~MNT_ROOTFS; 413 414 iovcnt = uap->iovcnt; 415 /* 416 * Check that we have an even number of iovec's 417 * and that we have at least two options. 418 */ 419 if ((iovcnt & 1) || (iovcnt < 4)) { 420 CTR2(KTR_VFS, "%s: failed for invalid iovcnt %d", __func__, 421 uap->iovcnt); 422 return (EINVAL); 423 } 424 425 error = copyinuio(uap->iovp, iovcnt, &auio); 426 if (error) { 427 CTR2(KTR_VFS, "%s: failed for invalid uio op with %d errno", 428 __func__, error); 429 return (error); 430 } 431 error = vfs_donmount(td, flags, auio); 432 433 free(auio, M_IOV); 434 return (error); 435 } 436 437 /* 438 * --------------------------------------------------------------------- 439 * Various utility functions 440 */ 441 442 void 443 vfs_ref(struct mount *mp) 444 { 445 446 CTR2(KTR_VFS, "%s: mp %p", __func__, mp); 447 MNT_ILOCK(mp); 448 MNT_REF(mp); 449 MNT_IUNLOCK(mp); 450 } 451 452 void 453 vfs_rel(struct mount *mp) 454 { 455 456 CTR2(KTR_VFS, "%s: mp %p", __func__, mp); 457 MNT_ILOCK(mp); 458 MNT_REL(mp); 459 MNT_IUNLOCK(mp); 460 } 461 462 /* 463 * Allocate and initialize the mount point struct. 464 */ 465 struct mount * 466 vfs_mount_alloc(struct vnode *vp, struct vfsconf *vfsp, const char *fspath, 467 struct ucred *cred) 468 { 469 struct mount *mp; 470 471 mp = uma_zalloc(mount_zone, M_WAITOK); 472 bzero(&mp->mnt_startzero, 473 __rangeof(struct mount, mnt_startzero, mnt_endzero)); 474 TAILQ_INIT(&mp->mnt_nvnodelist); 475 mp->mnt_nvnodelistsize = 0; 476 TAILQ_INIT(&mp->mnt_activevnodelist); 477 mp->mnt_activevnodelistsize = 0; 478 TAILQ_INIT(&mp->mnt_tmpfreevnodelist); 479 mp->mnt_tmpfreevnodelistsize = 0; 480 mp->mnt_ref = 0; 481 (void) vfs_busy(mp, MBF_NOWAIT); 482 atomic_add_acq_int(&vfsp->vfc_refcount, 1); 483 mp->mnt_op = vfsp->vfc_vfsops; 484 mp->mnt_vfc = vfsp; 485 mp->mnt_stat.f_type = vfsp->vfc_typenum; 486 mp->mnt_gen++; 487 strlcpy(mp->mnt_stat.f_fstypename, vfsp->vfc_name, MFSNAMELEN); 488 mp->mnt_vnodecovered = vp; 489 mp->mnt_cred = crdup(cred); 490 mp->mnt_stat.f_owner = cred->cr_uid; 491 strlcpy(mp->mnt_stat.f_mntonname, fspath, MNAMELEN); 492 mp->mnt_iosize_max = DFLTPHYS; 493 #ifdef MAC 494 mac_mount_init(mp); 495 mac_mount_create(cred, mp); 496 #endif 497 arc4rand(&mp->mnt_hashseed, sizeof mp->mnt_hashseed, 0); 498 TAILQ_INIT(&mp->mnt_uppers); 499 return (mp); 500 } 501 502 /* 503 * Destroy the mount struct previously allocated by vfs_mount_alloc(). 504 */ 505 void 506 vfs_mount_destroy(struct mount *mp) 507 { 508 509 MNT_ILOCK(mp); 510 mp->mnt_kern_flag |= MNTK_REFEXPIRE; 511 if (mp->mnt_kern_flag & MNTK_MWAIT) { 512 mp->mnt_kern_flag &= ~MNTK_MWAIT; 513 wakeup(mp); 514 } 515 while (mp->mnt_ref) 516 msleep(mp, MNT_MTX(mp), PVFS, "mntref", 0); 517 KASSERT(mp->mnt_ref == 0, 518 ("%s: invalid refcount in the drain path @ %s:%d", __func__, 519 __FILE__, __LINE__)); 520 if (mp->mnt_writeopcount != 0) 521 panic("vfs_mount_destroy: nonzero writeopcount"); 522 if (mp->mnt_secondary_writes != 0) 523 panic("vfs_mount_destroy: nonzero secondary_writes"); 524 atomic_subtract_rel_int(&mp->mnt_vfc->vfc_refcount, 1); 525 if (!TAILQ_EMPTY(&mp->mnt_nvnodelist)) { 526 struct vnode *vp; 527 528 TAILQ_FOREACH(vp, &mp->mnt_nvnodelist, v_nmntvnodes) 529 vn_printf(vp, "dangling vnode "); 530 panic("unmount: dangling vnode"); 531 } 532 KASSERT(TAILQ_EMPTY(&mp->mnt_uppers), ("mnt_uppers")); 533 if (mp->mnt_nvnodelistsize != 0) 534 panic("vfs_mount_destroy: nonzero nvnodelistsize"); 535 if (mp->mnt_activevnodelistsize != 0) 536 panic("vfs_mount_destroy: nonzero activevnodelistsize"); 537 if (mp->mnt_tmpfreevnodelistsize != 0) 538 panic("vfs_mount_destroy: nonzero tmpfreevnodelistsize"); 539 if (mp->mnt_lockref != 0) 540 panic("vfs_mount_destroy: nonzero lock refcount"); 541 MNT_IUNLOCK(mp); 542 if (mp->mnt_vnodecovered != NULL) 543 vrele(mp->mnt_vnodecovered); 544 #ifdef MAC 545 mac_mount_destroy(mp); 546 #endif 547 if (mp->mnt_opt != NULL) 548 vfs_freeopts(mp->mnt_opt); 549 crfree(mp->mnt_cred); 550 uma_zfree(mount_zone, mp); 551 } 552 553 static bool 554 vfs_should_downgrade_to_ro_mount(uint64_t fsflags, int error) 555 { 556 /* This is an upgrade of an exisiting mount. */ 557 if ((fsflags & MNT_UPDATE) != 0) 558 return (false); 559 /* This is already an R/O mount. */ 560 if ((fsflags & MNT_RDONLY) != 0) 561 return (false); 562 563 switch (error) { 564 case ENODEV: /* generic, geom, ... */ 565 case EACCES: /* cam/scsi, ... */ 566 case EROFS: /* md, mmcsd, ... */ 567 /* 568 * These errors can be returned by the storage layer to signal 569 * that the media is read-only. No harm in the R/O mount 570 * attempt if the error was returned for some other reason. 571 */ 572 return (true); 573 default: 574 return (false); 575 } 576 } 577 578 int 579 vfs_donmount(struct thread *td, uint64_t fsflags, struct uio *fsoptions) 580 { 581 struct vfsoptlist *optlist; 582 struct vfsopt *opt, *tmp_opt; 583 char *fstype, *fspath, *errmsg; 584 int error, fstypelen, fspathlen, errmsg_len, errmsg_pos; 585 bool autoro; 586 587 errmsg = fspath = NULL; 588 errmsg_len = fspathlen = 0; 589 errmsg_pos = -1; 590 autoro = default_autoro; 591 592 error = vfs_buildopts(fsoptions, &optlist); 593 if (error) 594 return (error); 595 596 if (vfs_getopt(optlist, "errmsg", (void **)&errmsg, &errmsg_len) == 0) 597 errmsg_pos = vfs_getopt_pos(optlist, "errmsg"); 598 599 /* 600 * We need these two options before the others, 601 * and they are mandatory for any filesystem. 602 * Ensure they are NUL terminated as well. 603 */ 604 fstypelen = 0; 605 error = vfs_getopt(optlist, "fstype", (void **)&fstype, &fstypelen); 606 if (error || fstype[fstypelen - 1] != '\0') { 607 error = EINVAL; 608 if (errmsg != NULL) 609 strncpy(errmsg, "Invalid fstype", errmsg_len); 610 goto bail; 611 } 612 fspathlen = 0; 613 error = vfs_getopt(optlist, "fspath", (void **)&fspath, &fspathlen); 614 if (error || fspath[fspathlen - 1] != '\0') { 615 error = EINVAL; 616 if (errmsg != NULL) 617 strncpy(errmsg, "Invalid fspath", errmsg_len); 618 goto bail; 619 } 620 621 /* 622 * We need to see if we have the "update" option 623 * before we call vfs_domount(), since vfs_domount() has special 624 * logic based on MNT_UPDATE. This is very important 625 * when we want to update the root filesystem. 626 */ 627 TAILQ_FOREACH_SAFE(opt, optlist, link, tmp_opt) { 628 if (strcmp(opt->name, "update") == 0) { 629 fsflags |= MNT_UPDATE; 630 vfs_freeopt(optlist, opt); 631 } 632 else if (strcmp(opt->name, "async") == 0) 633 fsflags |= MNT_ASYNC; 634 else if (strcmp(opt->name, "force") == 0) { 635 fsflags |= MNT_FORCE; 636 vfs_freeopt(optlist, opt); 637 } 638 else if (strcmp(opt->name, "reload") == 0) { 639 fsflags |= MNT_RELOAD; 640 vfs_freeopt(optlist, opt); 641 } 642 else if (strcmp(opt->name, "multilabel") == 0) 643 fsflags |= MNT_MULTILABEL; 644 else if (strcmp(opt->name, "noasync") == 0) 645 fsflags &= ~MNT_ASYNC; 646 else if (strcmp(opt->name, "noatime") == 0) 647 fsflags |= MNT_NOATIME; 648 else if (strcmp(opt->name, "atime") == 0) { 649 free(opt->name, M_MOUNT); 650 opt->name = strdup("nonoatime", M_MOUNT); 651 } 652 else if (strcmp(opt->name, "noclusterr") == 0) 653 fsflags |= MNT_NOCLUSTERR; 654 else if (strcmp(opt->name, "clusterr") == 0) { 655 free(opt->name, M_MOUNT); 656 opt->name = strdup("nonoclusterr", M_MOUNT); 657 } 658 else if (strcmp(opt->name, "noclusterw") == 0) 659 fsflags |= MNT_NOCLUSTERW; 660 else if (strcmp(opt->name, "clusterw") == 0) { 661 free(opt->name, M_MOUNT); 662 opt->name = strdup("nonoclusterw", M_MOUNT); 663 } 664 else if (strcmp(opt->name, "noexec") == 0) 665 fsflags |= MNT_NOEXEC; 666 else if (strcmp(opt->name, "exec") == 0) { 667 free(opt->name, M_MOUNT); 668 opt->name = strdup("nonoexec", M_MOUNT); 669 } 670 else if (strcmp(opt->name, "nosuid") == 0) 671 fsflags |= MNT_NOSUID; 672 else if (strcmp(opt->name, "suid") == 0) { 673 free(opt->name, M_MOUNT); 674 opt->name = strdup("nonosuid", M_MOUNT); 675 } 676 else if (strcmp(opt->name, "nosymfollow") == 0) 677 fsflags |= MNT_NOSYMFOLLOW; 678 else if (strcmp(opt->name, "symfollow") == 0) { 679 free(opt->name, M_MOUNT); 680 opt->name = strdup("nonosymfollow", M_MOUNT); 681 } 682 else if (strcmp(opt->name, "noro") == 0) { 683 fsflags &= ~MNT_RDONLY; 684 autoro = false; 685 } 686 else if (strcmp(opt->name, "rw") == 0) { 687 fsflags &= ~MNT_RDONLY; 688 autoro = false; 689 } 690 else if (strcmp(opt->name, "ro") == 0) { 691 fsflags |= MNT_RDONLY; 692 autoro = false; 693 } 694 else if (strcmp(opt->name, "rdonly") == 0) { 695 free(opt->name, M_MOUNT); 696 opt->name = strdup("ro", M_MOUNT); 697 fsflags |= MNT_RDONLY; 698 autoro = false; 699 } 700 else if (strcmp(opt->name, "autoro") == 0) { 701 vfs_freeopt(optlist, opt); 702 autoro = true; 703 } 704 else if (strcmp(opt->name, "suiddir") == 0) 705 fsflags |= MNT_SUIDDIR; 706 else if (strcmp(opt->name, "sync") == 0) 707 fsflags |= MNT_SYNCHRONOUS; 708 else if (strcmp(opt->name, "union") == 0) 709 fsflags |= MNT_UNION; 710 else if (strcmp(opt->name, "automounted") == 0) { 711 fsflags |= MNT_AUTOMOUNTED; 712 vfs_freeopt(optlist, opt); 713 } 714 } 715 716 /* 717 * Be ultra-paranoid about making sure the type and fspath 718 * variables will fit in our mp buffers, including the 719 * terminating NUL. 720 */ 721 if (fstypelen > MFSNAMELEN || fspathlen > MNAMELEN) { 722 error = ENAMETOOLONG; 723 goto bail; 724 } 725 726 error = vfs_domount(td, fstype, fspath, fsflags, &optlist); 727 728 /* 729 * See if we can mount in the read-only mode if the error code suggests 730 * that it could be possible and the mount options allow for that. 731 * Never try it if "[no]{ro|rw}" has been explicitly requested and not 732 * overridden by "autoro". 733 */ 734 if (autoro && vfs_should_downgrade_to_ro_mount(fsflags, error)) { 735 printf("%s: R/W mount failed, possibly R/O media," 736 " trying R/O mount\n", __func__); 737 fsflags |= MNT_RDONLY; 738 error = vfs_domount(td, fstype, fspath, fsflags, &optlist); 739 } 740 bail: 741 /* copyout the errmsg */ 742 if (errmsg_pos != -1 && ((2 * errmsg_pos + 1) < fsoptions->uio_iovcnt) 743 && errmsg_len > 0 && errmsg != NULL) { 744 if (fsoptions->uio_segflg == UIO_SYSSPACE) { 745 bcopy(errmsg, 746 fsoptions->uio_iov[2 * errmsg_pos + 1].iov_base, 747 fsoptions->uio_iov[2 * errmsg_pos + 1].iov_len); 748 } else { 749 copyout(errmsg, 750 fsoptions->uio_iov[2 * errmsg_pos + 1].iov_base, 751 fsoptions->uio_iov[2 * errmsg_pos + 1].iov_len); 752 } 753 } 754 755 if (optlist != NULL) 756 vfs_freeopts(optlist); 757 return (error); 758 } 759 760 /* 761 * Old mount API. 762 */ 763 #ifndef _SYS_SYSPROTO_H_ 764 struct mount_args { 765 char *type; 766 char *path; 767 int flags; 768 caddr_t data; 769 }; 770 #endif 771 /* ARGSUSED */ 772 int 773 sys_mount(struct thread *td, struct mount_args *uap) 774 { 775 char *fstype; 776 struct vfsconf *vfsp = NULL; 777 struct mntarg *ma = NULL; 778 uint64_t flags; 779 int error; 780 781 /* 782 * Mount flags are now 64-bits. On 32-bit architectures only 783 * 32-bits are passed in, but from here on everything handles 784 * 64-bit flags correctly. 785 */ 786 flags = uap->flags; 787 788 AUDIT_ARG_FFLAGS(flags); 789 790 /* 791 * Filter out MNT_ROOTFS. We do not want clients of mount() in 792 * userspace to set this flag, but we must filter it out if we want 793 * MNT_UPDATE on the root file system to work. 794 * MNT_ROOTFS should only be set by the kernel when mounting its 795 * root file system. 796 */ 797 flags &= ~MNT_ROOTFS; 798 799 fstype = malloc(MFSNAMELEN, M_TEMP, M_WAITOK); 800 error = copyinstr(uap->type, fstype, MFSNAMELEN, NULL); 801 if (error) { 802 free(fstype, M_TEMP); 803 return (error); 804 } 805 806 AUDIT_ARG_TEXT(fstype); 807 vfsp = vfs_byname_kld(fstype, td, &error); 808 free(fstype, M_TEMP); 809 if (vfsp == NULL) 810 return (ENOENT); 811 if (vfsp->vfc_vfsops->vfs_cmount == NULL) 812 return (EOPNOTSUPP); 813 814 ma = mount_argsu(ma, "fstype", uap->type, MFSNAMELEN); 815 ma = mount_argsu(ma, "fspath", uap->path, MNAMELEN); 816 ma = mount_argb(ma, flags & MNT_RDONLY, "noro"); 817 ma = mount_argb(ma, !(flags & MNT_NOSUID), "nosuid"); 818 ma = mount_argb(ma, !(flags & MNT_NOEXEC), "noexec"); 819 820 error = vfsp->vfc_vfsops->vfs_cmount(ma, uap->data, flags); 821 return (error); 822 } 823 824 /* 825 * vfs_domount_first(): first file system mount (not update) 826 */ 827 static int 828 vfs_domount_first( 829 struct thread *td, /* Calling thread. */ 830 struct vfsconf *vfsp, /* File system type. */ 831 char *fspath, /* Mount path. */ 832 struct vnode *vp, /* Vnode to be covered. */ 833 uint64_t fsflags, /* Flags common to all filesystems. */ 834 struct vfsoptlist **optlist /* Options local to the filesystem. */ 835 ) 836 { 837 struct vattr va; 838 struct mount *mp; 839 struct vnode *newdp; 840 int error; 841 842 ASSERT_VOP_ELOCKED(vp, __func__); 843 KASSERT((fsflags & MNT_UPDATE) == 0, ("MNT_UPDATE shouldn't be here")); 844 845 /* 846 * If the jail of the calling thread lacks permission for this type of 847 * file system, deny immediately. 848 */ 849 if (jailed(td->td_ucred) && !prison_allow(td->td_ucred, 850 vfsp->vfc_prison_flag)) { 851 vput(vp); 852 return (EPERM); 853 } 854 855 /* 856 * If the user is not root, ensure that they own the directory 857 * onto which we are attempting to mount. 858 */ 859 error = VOP_GETATTR(vp, &va, td->td_ucred); 860 if (error == 0 && va.va_uid != td->td_ucred->cr_uid) 861 error = priv_check_cred(td->td_ucred, PRIV_VFS_ADMIN, 0); 862 if (error == 0) 863 error = vinvalbuf(vp, V_SAVE, 0, 0); 864 if (error == 0 && vp->v_type != VDIR) 865 error = ENOTDIR; 866 if (error == 0) { 867 VI_LOCK(vp); 868 if ((vp->v_iflag & VI_MOUNT) == 0 && vp->v_mountedhere == NULL) 869 vp->v_iflag |= VI_MOUNT; 870 else 871 error = EBUSY; 872 VI_UNLOCK(vp); 873 } 874 if (error != 0) { 875 vput(vp); 876 return (error); 877 } 878 VOP_UNLOCK(vp, 0); 879 880 /* Allocate and initialize the filesystem. */ 881 mp = vfs_mount_alloc(vp, vfsp, fspath, td->td_ucred); 882 /* XXXMAC: pass to vfs_mount_alloc? */ 883 mp->mnt_optnew = *optlist; 884 /* Set the mount level flags. */ 885 mp->mnt_flag = (fsflags & (MNT_UPDATEMASK | MNT_ROOTFS | MNT_RDONLY)); 886 887 /* 888 * Mount the filesystem. 889 * XXX The final recipients of VFS_MOUNT just overwrite the ndp they 890 * get. No freeing of cn_pnbuf. 891 */ 892 error = VFS_MOUNT(mp); 893 if (error != 0) { 894 vfs_unbusy(mp); 895 mp->mnt_vnodecovered = NULL; 896 vfs_mount_destroy(mp); 897 VI_LOCK(vp); 898 vp->v_iflag &= ~VI_MOUNT; 899 VI_UNLOCK(vp); 900 vrele(vp); 901 return (error); 902 } 903 904 if (mp->mnt_opt != NULL) 905 vfs_freeopts(mp->mnt_opt); 906 mp->mnt_opt = mp->mnt_optnew; 907 *optlist = NULL; 908 (void)VFS_STATFS(mp, &mp->mnt_stat); 909 910 /* 911 * Prevent external consumers of mount options from reading mnt_optnew. 912 */ 913 mp->mnt_optnew = NULL; 914 915 MNT_ILOCK(mp); 916 if ((mp->mnt_flag & MNT_ASYNC) != 0 && 917 (mp->mnt_kern_flag & MNTK_NOASYNC) == 0) 918 mp->mnt_kern_flag |= MNTK_ASYNC; 919 else 920 mp->mnt_kern_flag &= ~MNTK_ASYNC; 921 MNT_IUNLOCK(mp); 922 923 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 924 cache_purge(vp); 925 VI_LOCK(vp); 926 vp->v_iflag &= ~VI_MOUNT; 927 VI_UNLOCK(vp); 928 vp->v_mountedhere = mp; 929 /* Place the new filesystem at the end of the mount list. */ 930 mtx_lock(&mountlist_mtx); 931 TAILQ_INSERT_TAIL(&mountlist, mp, mnt_list); 932 mtx_unlock(&mountlist_mtx); 933 vfs_event_signal(NULL, VQ_MOUNT, 0); 934 if (VFS_ROOT(mp, LK_EXCLUSIVE, &newdp)) 935 panic("mount: lost mount"); 936 VOP_UNLOCK(vp, 0); 937 EVENTHANDLER_DIRECT_INVOKE(vfs_mounted, mp, newdp, td); 938 VOP_UNLOCK(newdp, 0); 939 mountcheckdirs(vp, newdp); 940 vrele(newdp); 941 if ((mp->mnt_flag & MNT_RDONLY) == 0) 942 vfs_allocate_syncvnode(mp); 943 vfs_unbusy(mp); 944 return (0); 945 } 946 947 /* 948 * vfs_domount_update(): update of mounted file system 949 */ 950 static int 951 vfs_domount_update( 952 struct thread *td, /* Calling thread. */ 953 struct vnode *vp, /* Mount point vnode. */ 954 uint64_t fsflags, /* Flags common to all filesystems. */ 955 struct vfsoptlist **optlist /* Options local to the filesystem. */ 956 ) 957 { 958 struct export_args export; 959 void *bufp; 960 struct mount *mp; 961 int error, export_error, len; 962 uint64_t flag; 963 964 ASSERT_VOP_ELOCKED(vp, __func__); 965 KASSERT((fsflags & MNT_UPDATE) != 0, ("MNT_UPDATE should be here")); 966 mp = vp->v_mount; 967 968 if ((vp->v_vflag & VV_ROOT) == 0) { 969 if (vfs_copyopt(*optlist, "export", &export, sizeof(export)) 970 == 0) 971 error = EXDEV; 972 else 973 error = EINVAL; 974 vput(vp); 975 return (error); 976 } 977 978 /* 979 * We only allow the filesystem to be reloaded if it 980 * is currently mounted read-only. 981 */ 982 flag = mp->mnt_flag; 983 if ((fsflags & MNT_RELOAD) != 0 && (flag & MNT_RDONLY) == 0) { 984 vput(vp); 985 return (EOPNOTSUPP); /* Needs translation */ 986 } 987 /* 988 * Only privileged root, or (if MNT_USER is set) the user that 989 * did the original mount is permitted to update it. 990 */ 991 error = vfs_suser(mp, td); 992 if (error != 0) { 993 vput(vp); 994 return (error); 995 } 996 if (vfs_busy(mp, MBF_NOWAIT)) { 997 vput(vp); 998 return (EBUSY); 999 } 1000 VI_LOCK(vp); 1001 if ((vp->v_iflag & VI_MOUNT) != 0 || vp->v_mountedhere != NULL) { 1002 VI_UNLOCK(vp); 1003 vfs_unbusy(mp); 1004 vput(vp); 1005 return (EBUSY); 1006 } 1007 vp->v_iflag |= VI_MOUNT; 1008 VI_UNLOCK(vp); 1009 VOP_UNLOCK(vp, 0); 1010 1011 MNT_ILOCK(mp); 1012 if ((mp->mnt_kern_flag & MNTK_UNMOUNT) != 0) { 1013 MNT_IUNLOCK(mp); 1014 error = EBUSY; 1015 goto end; 1016 } 1017 mp->mnt_flag &= ~MNT_UPDATEMASK; 1018 mp->mnt_flag |= fsflags & (MNT_RELOAD | MNT_FORCE | MNT_UPDATE | 1019 MNT_SNAPSHOT | MNT_ROOTFS | MNT_UPDATEMASK | MNT_RDONLY); 1020 if ((mp->mnt_flag & MNT_ASYNC) == 0) 1021 mp->mnt_kern_flag &= ~MNTK_ASYNC; 1022 MNT_IUNLOCK(mp); 1023 mp->mnt_optnew = *optlist; 1024 vfs_mergeopts(mp->mnt_optnew, mp->mnt_opt); 1025 1026 /* 1027 * Mount the filesystem. 1028 * XXX The final recipients of VFS_MOUNT just overwrite the ndp they 1029 * get. No freeing of cn_pnbuf. 1030 */ 1031 error = VFS_MOUNT(mp); 1032 1033 export_error = 0; 1034 /* Process the export option. */ 1035 if (error == 0 && vfs_getopt(mp->mnt_optnew, "export", &bufp, 1036 &len) == 0) { 1037 /* Assume that there is only 1 ABI for each length. */ 1038 switch (len) { 1039 case (sizeof(struct oexport_args)): 1040 bzero(&export, sizeof(export)); 1041 /* FALLTHROUGH */ 1042 case (sizeof(export)): 1043 bcopy(bufp, &export, len); 1044 export_error = vfs_export(mp, &export); 1045 break; 1046 default: 1047 export_error = EINVAL; 1048 break; 1049 } 1050 } 1051 1052 MNT_ILOCK(mp); 1053 if (error == 0) { 1054 mp->mnt_flag &= ~(MNT_UPDATE | MNT_RELOAD | MNT_FORCE | 1055 MNT_SNAPSHOT); 1056 } else { 1057 /* 1058 * If we fail, restore old mount flags. MNT_QUOTA is special, 1059 * because it is not part of MNT_UPDATEMASK, but it could have 1060 * changed in the meantime if quotactl(2) was called. 1061 * All in all we want current value of MNT_QUOTA, not the old 1062 * one. 1063 */ 1064 mp->mnt_flag = (mp->mnt_flag & MNT_QUOTA) | (flag & ~MNT_QUOTA); 1065 } 1066 if ((mp->mnt_flag & MNT_ASYNC) != 0 && 1067 (mp->mnt_kern_flag & MNTK_NOASYNC) == 0) 1068 mp->mnt_kern_flag |= MNTK_ASYNC; 1069 else 1070 mp->mnt_kern_flag &= ~MNTK_ASYNC; 1071 MNT_IUNLOCK(mp); 1072 1073 if (error != 0) 1074 goto end; 1075 1076 if (mp->mnt_opt != NULL) 1077 vfs_freeopts(mp->mnt_opt); 1078 mp->mnt_opt = mp->mnt_optnew; 1079 *optlist = NULL; 1080 (void)VFS_STATFS(mp, &mp->mnt_stat); 1081 /* 1082 * Prevent external consumers of mount options from reading 1083 * mnt_optnew. 1084 */ 1085 mp->mnt_optnew = NULL; 1086 1087 if ((mp->mnt_flag & MNT_RDONLY) == 0) 1088 vfs_allocate_syncvnode(mp); 1089 else 1090 vfs_deallocate_syncvnode(mp); 1091 end: 1092 vfs_unbusy(mp); 1093 VI_LOCK(vp); 1094 vp->v_iflag &= ~VI_MOUNT; 1095 VI_UNLOCK(vp); 1096 vrele(vp); 1097 return (error != 0 ? error : export_error); 1098 } 1099 1100 /* 1101 * vfs_domount(): actually attempt a filesystem mount. 1102 */ 1103 static int 1104 vfs_domount( 1105 struct thread *td, /* Calling thread. */ 1106 const char *fstype, /* Filesystem type. */ 1107 char *fspath, /* Mount path. */ 1108 uint64_t fsflags, /* Flags common to all filesystems. */ 1109 struct vfsoptlist **optlist /* Options local to the filesystem. */ 1110 ) 1111 { 1112 struct vfsconf *vfsp; 1113 struct nameidata nd; 1114 struct vnode *vp; 1115 char *pathbuf; 1116 int error; 1117 1118 /* 1119 * Be ultra-paranoid about making sure the type and fspath 1120 * variables will fit in our mp buffers, including the 1121 * terminating NUL. 1122 */ 1123 if (strlen(fstype) >= MFSNAMELEN || strlen(fspath) >= MNAMELEN) 1124 return (ENAMETOOLONG); 1125 1126 if (jailed(td->td_ucred) || usermount == 0) { 1127 if ((error = priv_check(td, PRIV_VFS_MOUNT)) != 0) 1128 return (error); 1129 } 1130 1131 /* 1132 * Do not allow NFS export or MNT_SUIDDIR by unprivileged users. 1133 */ 1134 if (fsflags & MNT_EXPORTED) { 1135 error = priv_check(td, PRIV_VFS_MOUNT_EXPORTED); 1136 if (error) 1137 return (error); 1138 } 1139 if (fsflags & MNT_SUIDDIR) { 1140 error = priv_check(td, PRIV_VFS_MOUNT_SUIDDIR); 1141 if (error) 1142 return (error); 1143 } 1144 /* 1145 * Silently enforce MNT_NOSUID and MNT_USER for unprivileged users. 1146 */ 1147 if ((fsflags & (MNT_NOSUID | MNT_USER)) != (MNT_NOSUID | MNT_USER)) { 1148 if (priv_check(td, PRIV_VFS_MOUNT_NONUSER) != 0) 1149 fsflags |= MNT_NOSUID | MNT_USER; 1150 } 1151 1152 /* Load KLDs before we lock the covered vnode to avoid reversals. */ 1153 vfsp = NULL; 1154 if ((fsflags & MNT_UPDATE) == 0) { 1155 /* Don't try to load KLDs if we're mounting the root. */ 1156 if (fsflags & MNT_ROOTFS) 1157 vfsp = vfs_byname(fstype); 1158 else 1159 vfsp = vfs_byname_kld(fstype, td, &error); 1160 if (vfsp == NULL) 1161 return (ENODEV); 1162 } 1163 1164 /* 1165 * Get vnode to be covered or mount point's vnode in case of MNT_UPDATE. 1166 */ 1167 NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF | AUDITVNODE1, 1168 UIO_SYSSPACE, fspath, td); 1169 error = namei(&nd); 1170 if (error != 0) 1171 return (error); 1172 NDFREE(&nd, NDF_ONLY_PNBUF); 1173 vp = nd.ni_vp; 1174 if ((fsflags & MNT_UPDATE) == 0) { 1175 pathbuf = malloc(MNAMELEN, M_TEMP, M_WAITOK); 1176 strcpy(pathbuf, fspath); 1177 error = vn_path_to_global_path(td, vp, pathbuf, MNAMELEN); 1178 /* debug.disablefullpath == 1 results in ENODEV */ 1179 if (error == 0 || error == ENODEV) { 1180 error = vfs_domount_first(td, vfsp, pathbuf, vp, 1181 fsflags, optlist); 1182 } 1183 free(pathbuf, M_TEMP); 1184 } else 1185 error = vfs_domount_update(td, vp, fsflags, optlist); 1186 1187 return (error); 1188 } 1189 1190 /* 1191 * Unmount a filesystem. 1192 * 1193 * Note: unmount takes a path to the vnode mounted on as argument, not 1194 * special file (as before). 1195 */ 1196 #ifndef _SYS_SYSPROTO_H_ 1197 struct unmount_args { 1198 char *path; 1199 int flags; 1200 }; 1201 #endif 1202 /* ARGSUSED */ 1203 int 1204 sys_unmount(struct thread *td, struct unmount_args *uap) 1205 { 1206 struct nameidata nd; 1207 struct mount *mp; 1208 char *pathbuf; 1209 int error, id0, id1; 1210 1211 AUDIT_ARG_VALUE(uap->flags); 1212 if (jailed(td->td_ucred) || usermount == 0) { 1213 error = priv_check(td, PRIV_VFS_UNMOUNT); 1214 if (error) 1215 return (error); 1216 } 1217 1218 pathbuf = malloc(MNAMELEN, M_TEMP, M_WAITOK); 1219 error = copyinstr(uap->path, pathbuf, MNAMELEN, NULL); 1220 if (error) { 1221 free(pathbuf, M_TEMP); 1222 return (error); 1223 } 1224 if (uap->flags & MNT_BYFSID) { 1225 AUDIT_ARG_TEXT(pathbuf); 1226 /* Decode the filesystem ID. */ 1227 if (sscanf(pathbuf, "FSID:%d:%d", &id0, &id1) != 2) { 1228 free(pathbuf, M_TEMP); 1229 return (EINVAL); 1230 } 1231 1232 mtx_lock(&mountlist_mtx); 1233 TAILQ_FOREACH_REVERSE(mp, &mountlist, mntlist, mnt_list) { 1234 if (mp->mnt_stat.f_fsid.val[0] == id0 && 1235 mp->mnt_stat.f_fsid.val[1] == id1) { 1236 vfs_ref(mp); 1237 break; 1238 } 1239 } 1240 mtx_unlock(&mountlist_mtx); 1241 } else { 1242 /* 1243 * Try to find global path for path argument. 1244 */ 1245 NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF | AUDITVNODE1, 1246 UIO_SYSSPACE, pathbuf, td); 1247 if (namei(&nd) == 0) { 1248 NDFREE(&nd, NDF_ONLY_PNBUF); 1249 error = vn_path_to_global_path(td, nd.ni_vp, pathbuf, 1250 MNAMELEN); 1251 if (error == 0 || error == ENODEV) 1252 vput(nd.ni_vp); 1253 } 1254 mtx_lock(&mountlist_mtx); 1255 TAILQ_FOREACH_REVERSE(mp, &mountlist, mntlist, mnt_list) { 1256 if (strcmp(mp->mnt_stat.f_mntonname, pathbuf) == 0) { 1257 vfs_ref(mp); 1258 break; 1259 } 1260 } 1261 mtx_unlock(&mountlist_mtx); 1262 } 1263 free(pathbuf, M_TEMP); 1264 if (mp == NULL) { 1265 /* 1266 * Previously we returned ENOENT for a nonexistent path and 1267 * EINVAL for a non-mountpoint. We cannot tell these apart 1268 * now, so in the !MNT_BYFSID case return the more likely 1269 * EINVAL for compatibility. 1270 */ 1271 return ((uap->flags & MNT_BYFSID) ? ENOENT : EINVAL); 1272 } 1273 1274 /* 1275 * Don't allow unmounting the root filesystem. 1276 */ 1277 if (mp->mnt_flag & MNT_ROOTFS) { 1278 vfs_rel(mp); 1279 return (EINVAL); 1280 } 1281 error = dounmount(mp, uap->flags, td); 1282 return (error); 1283 } 1284 1285 /* 1286 * Return error if any of the vnodes, ignoring the root vnode 1287 * and the syncer vnode, have non-zero usecount. 1288 * 1289 * This function is purely advisory - it can return false positives 1290 * and negatives. 1291 */ 1292 static int 1293 vfs_check_usecounts(struct mount *mp) 1294 { 1295 struct vnode *vp, *mvp; 1296 1297 MNT_VNODE_FOREACH_ALL(vp, mp, mvp) { 1298 if ((vp->v_vflag & VV_ROOT) == 0 && vp->v_type != VNON && 1299 vp->v_usecount != 0) { 1300 VI_UNLOCK(vp); 1301 MNT_VNODE_FOREACH_ALL_ABORT(mp, mvp); 1302 return (EBUSY); 1303 } 1304 VI_UNLOCK(vp); 1305 } 1306 1307 return (0); 1308 } 1309 1310 static void 1311 dounmount_cleanup(struct mount *mp, struct vnode *coveredvp, int mntkflags) 1312 { 1313 1314 mtx_assert(MNT_MTX(mp), MA_OWNED); 1315 mp->mnt_kern_flag &= ~mntkflags; 1316 if ((mp->mnt_kern_flag & MNTK_MWAIT) != 0) { 1317 mp->mnt_kern_flag &= ~MNTK_MWAIT; 1318 wakeup(mp); 1319 } 1320 MNT_IUNLOCK(mp); 1321 if (coveredvp != NULL) { 1322 VOP_UNLOCK(coveredvp, 0); 1323 vdrop(coveredvp); 1324 } 1325 vn_finished_write(mp); 1326 } 1327 1328 /* 1329 * Do the actual filesystem unmount. 1330 */ 1331 int 1332 dounmount(struct mount *mp, int flags, struct thread *td) 1333 { 1334 struct vnode *coveredvp; 1335 int error; 1336 uint64_t async_flag; 1337 int mnt_gen_r; 1338 1339 if ((coveredvp = mp->mnt_vnodecovered) != NULL) { 1340 mnt_gen_r = mp->mnt_gen; 1341 VI_LOCK(coveredvp); 1342 vholdl(coveredvp); 1343 vn_lock(coveredvp, LK_EXCLUSIVE | LK_INTERLOCK | LK_RETRY); 1344 /* 1345 * Check for mp being unmounted while waiting for the 1346 * covered vnode lock. 1347 */ 1348 if (coveredvp->v_mountedhere != mp || 1349 coveredvp->v_mountedhere->mnt_gen != mnt_gen_r) { 1350 VOP_UNLOCK(coveredvp, 0); 1351 vdrop(coveredvp); 1352 vfs_rel(mp); 1353 return (EBUSY); 1354 } 1355 } 1356 1357 /* 1358 * Only privileged root, or (if MNT_USER is set) the user that did the 1359 * original mount is permitted to unmount this filesystem. 1360 */ 1361 error = vfs_suser(mp, td); 1362 if (error != 0) { 1363 if (coveredvp != NULL) { 1364 VOP_UNLOCK(coveredvp, 0); 1365 vdrop(coveredvp); 1366 } 1367 vfs_rel(mp); 1368 return (error); 1369 } 1370 1371 vn_start_write(NULL, &mp, V_WAIT | V_MNTREF); 1372 MNT_ILOCK(mp); 1373 if ((mp->mnt_kern_flag & MNTK_UNMOUNT) != 0 || 1374 (mp->mnt_flag & MNT_UPDATE) != 0 || 1375 !TAILQ_EMPTY(&mp->mnt_uppers)) { 1376 dounmount_cleanup(mp, coveredvp, 0); 1377 return (EBUSY); 1378 } 1379 mp->mnt_kern_flag |= MNTK_UNMOUNT | MNTK_NOINSMNTQ; 1380 if (flags & MNT_NONBUSY) { 1381 MNT_IUNLOCK(mp); 1382 error = vfs_check_usecounts(mp); 1383 MNT_ILOCK(mp); 1384 if (error != 0) { 1385 dounmount_cleanup(mp, coveredvp, MNTK_UNMOUNT | 1386 MNTK_NOINSMNTQ); 1387 return (error); 1388 } 1389 } 1390 /* Allow filesystems to detect that a forced unmount is in progress. */ 1391 if (flags & MNT_FORCE) { 1392 mp->mnt_kern_flag |= MNTK_UNMOUNTF; 1393 MNT_IUNLOCK(mp); 1394 /* 1395 * Must be done after setting MNTK_UNMOUNTF and before 1396 * waiting for mnt_lockref to become 0. 1397 */ 1398 VFS_PURGE(mp); 1399 MNT_ILOCK(mp); 1400 } 1401 error = 0; 1402 if (mp->mnt_lockref) { 1403 mp->mnt_kern_flag |= MNTK_DRAINING; 1404 error = msleep(&mp->mnt_lockref, MNT_MTX(mp), PVFS, 1405 "mount drain", 0); 1406 } 1407 MNT_IUNLOCK(mp); 1408 KASSERT(mp->mnt_lockref == 0, 1409 ("%s: invalid lock refcount in the drain path @ %s:%d", 1410 __func__, __FILE__, __LINE__)); 1411 KASSERT(error == 0, 1412 ("%s: invalid return value for msleep in the drain path @ %s:%d", 1413 __func__, __FILE__, __LINE__)); 1414 1415 if (mp->mnt_flag & MNT_EXPUBLIC) 1416 vfs_setpublicfs(NULL, NULL, NULL); 1417 1418 /* 1419 * From now, we can claim that the use reference on the 1420 * coveredvp is ours, and the ref can be released only by 1421 * successfull unmount by us, or left for later unmount 1422 * attempt. The previously acquired hold reference is no 1423 * longer needed to protect the vnode from reuse. 1424 */ 1425 if (coveredvp != NULL) 1426 vdrop(coveredvp); 1427 1428 vfs_msync(mp, MNT_WAIT); 1429 MNT_ILOCK(mp); 1430 async_flag = mp->mnt_flag & MNT_ASYNC; 1431 mp->mnt_flag &= ~MNT_ASYNC; 1432 mp->mnt_kern_flag &= ~MNTK_ASYNC; 1433 MNT_IUNLOCK(mp); 1434 cache_purgevfs(mp, false); /* remove cache entries for this file sys */ 1435 vfs_deallocate_syncvnode(mp); 1436 if ((mp->mnt_flag & MNT_RDONLY) != 0 || (flags & MNT_FORCE) != 0 || 1437 (error = VFS_SYNC(mp, MNT_WAIT)) == 0) 1438 error = VFS_UNMOUNT(mp, flags); 1439 vn_finished_write(mp); 1440 /* 1441 * If we failed to flush the dirty blocks for this mount point, 1442 * undo all the cdir/rdir and rootvnode changes we made above. 1443 * Unless we failed to do so because the device is reporting that 1444 * it doesn't exist anymore. 1445 */ 1446 if (error && error != ENXIO) { 1447 MNT_ILOCK(mp); 1448 mp->mnt_kern_flag &= ~MNTK_NOINSMNTQ; 1449 if ((mp->mnt_flag & MNT_RDONLY) == 0) { 1450 MNT_IUNLOCK(mp); 1451 vfs_allocate_syncvnode(mp); 1452 MNT_ILOCK(mp); 1453 } 1454 mp->mnt_kern_flag &= ~(MNTK_UNMOUNT | MNTK_UNMOUNTF); 1455 mp->mnt_flag |= async_flag; 1456 if ((mp->mnt_flag & MNT_ASYNC) != 0 && 1457 (mp->mnt_kern_flag & MNTK_NOASYNC) == 0) 1458 mp->mnt_kern_flag |= MNTK_ASYNC; 1459 if (mp->mnt_kern_flag & MNTK_MWAIT) { 1460 mp->mnt_kern_flag &= ~MNTK_MWAIT; 1461 wakeup(mp); 1462 } 1463 MNT_IUNLOCK(mp); 1464 if (coveredvp) 1465 VOP_UNLOCK(coveredvp, 0); 1466 return (error); 1467 } 1468 mtx_lock(&mountlist_mtx); 1469 TAILQ_REMOVE(&mountlist, mp, mnt_list); 1470 mtx_unlock(&mountlist_mtx); 1471 EVENTHANDLER_DIRECT_INVOKE(vfs_unmounted, mp, td); 1472 if (coveredvp != NULL) { 1473 coveredvp->v_mountedhere = NULL; 1474 VOP_UNLOCK(coveredvp, 0); 1475 } 1476 vfs_event_signal(NULL, VQ_UNMOUNT, 0); 1477 if (rootvnode != NULL && mp == rootvnode->v_mount) { 1478 vrele(rootvnode); 1479 rootvnode = NULL; 1480 } 1481 if (mp == rootdevmp) 1482 rootdevmp = NULL; 1483 vfs_mount_destroy(mp); 1484 return (0); 1485 } 1486 1487 /* 1488 * Report errors during filesystem mounting. 1489 */ 1490 void 1491 vfs_mount_error(struct mount *mp, const char *fmt, ...) 1492 { 1493 struct vfsoptlist *moptlist = mp->mnt_optnew; 1494 va_list ap; 1495 int error, len; 1496 char *errmsg; 1497 1498 error = vfs_getopt(moptlist, "errmsg", (void **)&errmsg, &len); 1499 if (error || errmsg == NULL || len <= 0) 1500 return; 1501 1502 va_start(ap, fmt); 1503 vsnprintf(errmsg, (size_t)len, fmt, ap); 1504 va_end(ap); 1505 } 1506 1507 void 1508 vfs_opterror(struct vfsoptlist *opts, const char *fmt, ...) 1509 { 1510 va_list ap; 1511 int error, len; 1512 char *errmsg; 1513 1514 error = vfs_getopt(opts, "errmsg", (void **)&errmsg, &len); 1515 if (error || errmsg == NULL || len <= 0) 1516 return; 1517 1518 va_start(ap, fmt); 1519 vsnprintf(errmsg, (size_t)len, fmt, ap); 1520 va_end(ap); 1521 } 1522 1523 /* 1524 * --------------------------------------------------------------------- 1525 * Functions for querying mount options/arguments from filesystems. 1526 */ 1527 1528 /* 1529 * Check that no unknown options are given 1530 */ 1531 int 1532 vfs_filteropt(struct vfsoptlist *opts, const char **legal) 1533 { 1534 struct vfsopt *opt; 1535 char errmsg[255]; 1536 const char **t, *p, *q; 1537 int ret = 0; 1538 1539 TAILQ_FOREACH(opt, opts, link) { 1540 p = opt->name; 1541 q = NULL; 1542 if (p[0] == 'n' && p[1] == 'o') 1543 q = p + 2; 1544 for(t = global_opts; *t != NULL; t++) { 1545 if (strcmp(*t, p) == 0) 1546 break; 1547 if (q != NULL) { 1548 if (strcmp(*t, q) == 0) 1549 break; 1550 } 1551 } 1552 if (*t != NULL) 1553 continue; 1554 for(t = legal; *t != NULL; t++) { 1555 if (strcmp(*t, p) == 0) 1556 break; 1557 if (q != NULL) { 1558 if (strcmp(*t, q) == 0) 1559 break; 1560 } 1561 } 1562 if (*t != NULL) 1563 continue; 1564 snprintf(errmsg, sizeof(errmsg), 1565 "mount option <%s> is unknown", p); 1566 ret = EINVAL; 1567 } 1568 if (ret != 0) { 1569 TAILQ_FOREACH(opt, opts, link) { 1570 if (strcmp(opt->name, "errmsg") == 0) { 1571 strncpy((char *)opt->value, errmsg, opt->len); 1572 break; 1573 } 1574 } 1575 if (opt == NULL) 1576 printf("%s\n", errmsg); 1577 } 1578 return (ret); 1579 } 1580 1581 /* 1582 * Get a mount option by its name. 1583 * 1584 * Return 0 if the option was found, ENOENT otherwise. 1585 * If len is non-NULL it will be filled with the length 1586 * of the option. If buf is non-NULL, it will be filled 1587 * with the address of the option. 1588 */ 1589 int 1590 vfs_getopt(struct vfsoptlist *opts, const char *name, void **buf, int *len) 1591 { 1592 struct vfsopt *opt; 1593 1594 KASSERT(opts != NULL, ("vfs_getopt: caller passed 'opts' as NULL")); 1595 1596 TAILQ_FOREACH(opt, opts, link) { 1597 if (strcmp(name, opt->name) == 0) { 1598 opt->seen = 1; 1599 if (len != NULL) 1600 *len = opt->len; 1601 if (buf != NULL) 1602 *buf = opt->value; 1603 return (0); 1604 } 1605 } 1606 return (ENOENT); 1607 } 1608 1609 int 1610 vfs_getopt_pos(struct vfsoptlist *opts, const char *name) 1611 { 1612 struct vfsopt *opt; 1613 1614 if (opts == NULL) 1615 return (-1); 1616 1617 TAILQ_FOREACH(opt, opts, link) { 1618 if (strcmp(name, opt->name) == 0) { 1619 opt->seen = 1; 1620 return (opt->pos); 1621 } 1622 } 1623 return (-1); 1624 } 1625 1626 int 1627 vfs_getopt_size(struct vfsoptlist *opts, const char *name, off_t *value) 1628 { 1629 char *opt_value, *vtp; 1630 quad_t iv; 1631 int error, opt_len; 1632 1633 error = vfs_getopt(opts, name, (void **)&opt_value, &opt_len); 1634 if (error != 0) 1635 return (error); 1636 if (opt_len == 0 || opt_value == NULL) 1637 return (EINVAL); 1638 if (opt_value[0] == '\0' || opt_value[opt_len - 1] != '\0') 1639 return (EINVAL); 1640 iv = strtoq(opt_value, &vtp, 0); 1641 if (vtp == opt_value || (vtp[0] != '\0' && vtp[1] != '\0')) 1642 return (EINVAL); 1643 if (iv < 0) 1644 return (EINVAL); 1645 switch (vtp[0]) { 1646 case 't': 1647 case 'T': 1648 iv *= 1024; 1649 case 'g': 1650 case 'G': 1651 iv *= 1024; 1652 case 'm': 1653 case 'M': 1654 iv *= 1024; 1655 case 'k': 1656 case 'K': 1657 iv *= 1024; 1658 case '\0': 1659 break; 1660 default: 1661 return (EINVAL); 1662 } 1663 *value = iv; 1664 1665 return (0); 1666 } 1667 1668 char * 1669 vfs_getopts(struct vfsoptlist *opts, const char *name, int *error) 1670 { 1671 struct vfsopt *opt; 1672 1673 *error = 0; 1674 TAILQ_FOREACH(opt, opts, link) { 1675 if (strcmp(name, opt->name) != 0) 1676 continue; 1677 opt->seen = 1; 1678 if (opt->len == 0 || 1679 ((char *)opt->value)[opt->len - 1] != '\0') { 1680 *error = EINVAL; 1681 return (NULL); 1682 } 1683 return (opt->value); 1684 } 1685 *error = ENOENT; 1686 return (NULL); 1687 } 1688 1689 int 1690 vfs_flagopt(struct vfsoptlist *opts, const char *name, uint64_t *w, 1691 uint64_t val) 1692 { 1693 struct vfsopt *opt; 1694 1695 TAILQ_FOREACH(opt, opts, link) { 1696 if (strcmp(name, opt->name) == 0) { 1697 opt->seen = 1; 1698 if (w != NULL) 1699 *w |= val; 1700 return (1); 1701 } 1702 } 1703 if (w != NULL) 1704 *w &= ~val; 1705 return (0); 1706 } 1707 1708 int 1709 vfs_scanopt(struct vfsoptlist *opts, const char *name, const char *fmt, ...) 1710 { 1711 va_list ap; 1712 struct vfsopt *opt; 1713 int ret; 1714 1715 KASSERT(opts != NULL, ("vfs_getopt: caller passed 'opts' as NULL")); 1716 1717 TAILQ_FOREACH(opt, opts, link) { 1718 if (strcmp(name, opt->name) != 0) 1719 continue; 1720 opt->seen = 1; 1721 if (opt->len == 0 || opt->value == NULL) 1722 return (0); 1723 if (((char *)opt->value)[opt->len - 1] != '\0') 1724 return (0); 1725 va_start(ap, fmt); 1726 ret = vsscanf(opt->value, fmt, ap); 1727 va_end(ap); 1728 return (ret); 1729 } 1730 return (0); 1731 } 1732 1733 int 1734 vfs_setopt(struct vfsoptlist *opts, const char *name, void *value, int len) 1735 { 1736 struct vfsopt *opt; 1737 1738 TAILQ_FOREACH(opt, opts, link) { 1739 if (strcmp(name, opt->name) != 0) 1740 continue; 1741 opt->seen = 1; 1742 if (opt->value == NULL) 1743 opt->len = len; 1744 else { 1745 if (opt->len != len) 1746 return (EINVAL); 1747 bcopy(value, opt->value, len); 1748 } 1749 return (0); 1750 } 1751 return (ENOENT); 1752 } 1753 1754 int 1755 vfs_setopt_part(struct vfsoptlist *opts, const char *name, void *value, int len) 1756 { 1757 struct vfsopt *opt; 1758 1759 TAILQ_FOREACH(opt, opts, link) { 1760 if (strcmp(name, opt->name) != 0) 1761 continue; 1762 opt->seen = 1; 1763 if (opt->value == NULL) 1764 opt->len = len; 1765 else { 1766 if (opt->len < len) 1767 return (EINVAL); 1768 opt->len = len; 1769 bcopy(value, opt->value, len); 1770 } 1771 return (0); 1772 } 1773 return (ENOENT); 1774 } 1775 1776 int 1777 vfs_setopts(struct vfsoptlist *opts, const char *name, const char *value) 1778 { 1779 struct vfsopt *opt; 1780 1781 TAILQ_FOREACH(opt, opts, link) { 1782 if (strcmp(name, opt->name) != 0) 1783 continue; 1784 opt->seen = 1; 1785 if (opt->value == NULL) 1786 opt->len = strlen(value) + 1; 1787 else if (strlcpy(opt->value, value, opt->len) >= opt->len) 1788 return (EINVAL); 1789 return (0); 1790 } 1791 return (ENOENT); 1792 } 1793 1794 /* 1795 * Find and copy a mount option. 1796 * 1797 * The size of the buffer has to be specified 1798 * in len, if it is not the same length as the 1799 * mount option, EINVAL is returned. 1800 * Returns ENOENT if the option is not found. 1801 */ 1802 int 1803 vfs_copyopt(struct vfsoptlist *opts, const char *name, void *dest, int len) 1804 { 1805 struct vfsopt *opt; 1806 1807 KASSERT(opts != NULL, ("vfs_copyopt: caller passed 'opts' as NULL")); 1808 1809 TAILQ_FOREACH(opt, opts, link) { 1810 if (strcmp(name, opt->name) == 0) { 1811 opt->seen = 1; 1812 if (len != opt->len) 1813 return (EINVAL); 1814 bcopy(opt->value, dest, opt->len); 1815 return (0); 1816 } 1817 } 1818 return (ENOENT); 1819 } 1820 1821 int 1822 __vfs_statfs(struct mount *mp, struct statfs *sbp) 1823 { 1824 int error; 1825 1826 error = mp->mnt_op->vfs_statfs(mp, &mp->mnt_stat); 1827 if (sbp != &mp->mnt_stat) 1828 *sbp = mp->mnt_stat; 1829 return (error); 1830 } 1831 1832 void 1833 vfs_mountedfrom(struct mount *mp, const char *from) 1834 { 1835 1836 bzero(mp->mnt_stat.f_mntfromname, sizeof mp->mnt_stat.f_mntfromname); 1837 strlcpy(mp->mnt_stat.f_mntfromname, from, 1838 sizeof mp->mnt_stat.f_mntfromname); 1839 } 1840 1841 /* 1842 * --------------------------------------------------------------------- 1843 * This is the api for building mount args and mounting filesystems from 1844 * inside the kernel. 1845 * 1846 * The API works by accumulation of individual args. First error is 1847 * latched. 1848 * 1849 * XXX: should be documented in new manpage kernel_mount(9) 1850 */ 1851 1852 /* A memory allocation which must be freed when we are done */ 1853 struct mntaarg { 1854 SLIST_ENTRY(mntaarg) next; 1855 }; 1856 1857 /* The header for the mount arguments */ 1858 struct mntarg { 1859 struct iovec *v; 1860 int len; 1861 int error; 1862 SLIST_HEAD(, mntaarg) list; 1863 }; 1864 1865 /* 1866 * Add a boolean argument. 1867 * 1868 * flag is the boolean value. 1869 * name must start with "no". 1870 */ 1871 struct mntarg * 1872 mount_argb(struct mntarg *ma, int flag, const char *name) 1873 { 1874 1875 KASSERT(name[0] == 'n' && name[1] == 'o', 1876 ("mount_argb(...,%s): name must start with 'no'", name)); 1877 1878 return (mount_arg(ma, name + (flag ? 2 : 0), NULL, 0)); 1879 } 1880 1881 /* 1882 * Add an argument printf style 1883 */ 1884 struct mntarg * 1885 mount_argf(struct mntarg *ma, const char *name, const char *fmt, ...) 1886 { 1887 va_list ap; 1888 struct mntaarg *maa; 1889 struct sbuf *sb; 1890 int len; 1891 1892 if (ma == NULL) { 1893 ma = malloc(sizeof *ma, M_MOUNT, M_WAITOK | M_ZERO); 1894 SLIST_INIT(&ma->list); 1895 } 1896 if (ma->error) 1897 return (ma); 1898 1899 ma->v = realloc(ma->v, sizeof *ma->v * (ma->len + 2), 1900 M_MOUNT, M_WAITOK); 1901 ma->v[ma->len].iov_base = (void *)(uintptr_t)name; 1902 ma->v[ma->len].iov_len = strlen(name) + 1; 1903 ma->len++; 1904 1905 sb = sbuf_new_auto(); 1906 va_start(ap, fmt); 1907 sbuf_vprintf(sb, fmt, ap); 1908 va_end(ap); 1909 sbuf_finish(sb); 1910 len = sbuf_len(sb) + 1; 1911 maa = malloc(sizeof *maa + len, M_MOUNT, M_WAITOK | M_ZERO); 1912 SLIST_INSERT_HEAD(&ma->list, maa, next); 1913 bcopy(sbuf_data(sb), maa + 1, len); 1914 sbuf_delete(sb); 1915 1916 ma->v[ma->len].iov_base = maa + 1; 1917 ma->v[ma->len].iov_len = len; 1918 ma->len++; 1919 1920 return (ma); 1921 } 1922 1923 /* 1924 * Add an argument which is a userland string. 1925 */ 1926 struct mntarg * 1927 mount_argsu(struct mntarg *ma, const char *name, const void *val, int len) 1928 { 1929 struct mntaarg *maa; 1930 char *tbuf; 1931 1932 if (val == NULL) 1933 return (ma); 1934 if (ma == NULL) { 1935 ma = malloc(sizeof *ma, M_MOUNT, M_WAITOK | M_ZERO); 1936 SLIST_INIT(&ma->list); 1937 } 1938 if (ma->error) 1939 return (ma); 1940 maa = malloc(sizeof *maa + len, M_MOUNT, M_WAITOK | M_ZERO); 1941 SLIST_INSERT_HEAD(&ma->list, maa, next); 1942 tbuf = (void *)(maa + 1); 1943 ma->error = copyinstr(val, tbuf, len, NULL); 1944 return (mount_arg(ma, name, tbuf, -1)); 1945 } 1946 1947 /* 1948 * Plain argument. 1949 * 1950 * If length is -1, treat value as a C string. 1951 */ 1952 struct mntarg * 1953 mount_arg(struct mntarg *ma, const char *name, const void *val, int len) 1954 { 1955 1956 if (ma == NULL) { 1957 ma = malloc(sizeof *ma, M_MOUNT, M_WAITOK | M_ZERO); 1958 SLIST_INIT(&ma->list); 1959 } 1960 if (ma->error) 1961 return (ma); 1962 1963 ma->v = realloc(ma->v, sizeof *ma->v * (ma->len + 2), 1964 M_MOUNT, M_WAITOK); 1965 ma->v[ma->len].iov_base = (void *)(uintptr_t)name; 1966 ma->v[ma->len].iov_len = strlen(name) + 1; 1967 ma->len++; 1968 1969 ma->v[ma->len].iov_base = (void *)(uintptr_t)val; 1970 if (len < 0) 1971 ma->v[ma->len].iov_len = strlen(val) + 1; 1972 else 1973 ma->v[ma->len].iov_len = len; 1974 ma->len++; 1975 return (ma); 1976 } 1977 1978 /* 1979 * Free a mntarg structure 1980 */ 1981 static void 1982 free_mntarg(struct mntarg *ma) 1983 { 1984 struct mntaarg *maa; 1985 1986 while (!SLIST_EMPTY(&ma->list)) { 1987 maa = SLIST_FIRST(&ma->list); 1988 SLIST_REMOVE_HEAD(&ma->list, next); 1989 free(maa, M_MOUNT); 1990 } 1991 free(ma->v, M_MOUNT); 1992 free(ma, M_MOUNT); 1993 } 1994 1995 /* 1996 * Mount a filesystem 1997 */ 1998 int 1999 kernel_mount(struct mntarg *ma, uint64_t flags) 2000 { 2001 struct uio auio; 2002 int error; 2003 2004 KASSERT(ma != NULL, ("kernel_mount NULL ma")); 2005 KASSERT(ma->v != NULL, ("kernel_mount NULL ma->v")); 2006 KASSERT(!(ma->len & 1), ("kernel_mount odd ma->len (%d)", ma->len)); 2007 2008 auio.uio_iov = ma->v; 2009 auio.uio_iovcnt = ma->len; 2010 auio.uio_segflg = UIO_SYSSPACE; 2011 2012 error = ma->error; 2013 if (!error) 2014 error = vfs_donmount(curthread, flags, &auio); 2015 free_mntarg(ma); 2016 return (error); 2017 } 2018 2019 /* 2020 * A printflike function to mount a filesystem. 2021 */ 2022 int 2023 kernel_vmount(int flags, ...) 2024 { 2025 struct mntarg *ma = NULL; 2026 va_list ap; 2027 const char *cp; 2028 const void *vp; 2029 int error; 2030 2031 va_start(ap, flags); 2032 for (;;) { 2033 cp = va_arg(ap, const char *); 2034 if (cp == NULL) 2035 break; 2036 vp = va_arg(ap, const void *); 2037 ma = mount_arg(ma, cp, vp, (vp != NULL ? -1 : 0)); 2038 } 2039 va_end(ap); 2040 2041 error = kernel_mount(ma, flags); 2042 return (error); 2043 } 2044 2045 void 2046 vfs_oexport_conv(const struct oexport_args *oexp, struct export_args *exp) 2047 { 2048 2049 bcopy(oexp, exp, sizeof(*oexp)); 2050 exp->ex_numsecflavors = 0; 2051 } 2052