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 user is not root, ensure that they own the directory 847 * onto which we are attempting to mount. 848 */ 849 error = VOP_GETATTR(vp, &va, td->td_ucred); 850 if (error == 0 && va.va_uid != td->td_ucred->cr_uid) 851 error = priv_check_cred(td->td_ucred, PRIV_VFS_ADMIN, 0); 852 if (error == 0) 853 error = vinvalbuf(vp, V_SAVE, 0, 0); 854 if (error == 0 && vp->v_type != VDIR) 855 error = ENOTDIR; 856 if (error == 0) { 857 VI_LOCK(vp); 858 if ((vp->v_iflag & VI_MOUNT) == 0 && vp->v_mountedhere == NULL) 859 vp->v_iflag |= VI_MOUNT; 860 else 861 error = EBUSY; 862 VI_UNLOCK(vp); 863 } 864 if (error != 0) { 865 vput(vp); 866 return (error); 867 } 868 VOP_UNLOCK(vp, 0); 869 870 /* Allocate and initialize the filesystem. */ 871 mp = vfs_mount_alloc(vp, vfsp, fspath, td->td_ucred); 872 /* XXXMAC: pass to vfs_mount_alloc? */ 873 mp->mnt_optnew = *optlist; 874 /* Set the mount level flags. */ 875 mp->mnt_flag = (fsflags & (MNT_UPDATEMASK | MNT_ROOTFS | MNT_RDONLY)); 876 877 /* 878 * Mount the filesystem. 879 * XXX The final recipients of VFS_MOUNT just overwrite the ndp they 880 * get. No freeing of cn_pnbuf. 881 */ 882 error = VFS_MOUNT(mp); 883 if (error != 0) { 884 vfs_unbusy(mp); 885 mp->mnt_vnodecovered = NULL; 886 vfs_mount_destroy(mp); 887 VI_LOCK(vp); 888 vp->v_iflag &= ~VI_MOUNT; 889 VI_UNLOCK(vp); 890 vrele(vp); 891 return (error); 892 } 893 894 if (mp->mnt_opt != NULL) 895 vfs_freeopts(mp->mnt_opt); 896 mp->mnt_opt = mp->mnt_optnew; 897 *optlist = NULL; 898 (void)VFS_STATFS(mp, &mp->mnt_stat); 899 900 /* 901 * Prevent external consumers of mount options from reading mnt_optnew. 902 */ 903 mp->mnt_optnew = NULL; 904 905 MNT_ILOCK(mp); 906 if ((mp->mnt_flag & MNT_ASYNC) != 0 && 907 (mp->mnt_kern_flag & MNTK_NOASYNC) == 0) 908 mp->mnt_kern_flag |= MNTK_ASYNC; 909 else 910 mp->mnt_kern_flag &= ~MNTK_ASYNC; 911 MNT_IUNLOCK(mp); 912 913 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 914 cache_purge(vp); 915 VI_LOCK(vp); 916 vp->v_iflag &= ~VI_MOUNT; 917 VI_UNLOCK(vp); 918 vp->v_mountedhere = mp; 919 /* Place the new filesystem at the end of the mount list. */ 920 mtx_lock(&mountlist_mtx); 921 TAILQ_INSERT_TAIL(&mountlist, mp, mnt_list); 922 mtx_unlock(&mountlist_mtx); 923 vfs_event_signal(NULL, VQ_MOUNT, 0); 924 if (VFS_ROOT(mp, LK_EXCLUSIVE, &newdp)) 925 panic("mount: lost mount"); 926 VOP_UNLOCK(vp, 0); 927 EVENTHANDLER_DIRECT_INVOKE(vfs_mounted, mp, newdp, td); 928 VOP_UNLOCK(newdp, 0); 929 mountcheckdirs(vp, newdp); 930 vrele(newdp); 931 if ((mp->mnt_flag & MNT_RDONLY) == 0) 932 vfs_allocate_syncvnode(mp); 933 vfs_unbusy(mp); 934 return (0); 935 } 936 937 /* 938 * vfs_domount_update(): update of mounted file system 939 */ 940 static int 941 vfs_domount_update( 942 struct thread *td, /* Calling thread. */ 943 struct vnode *vp, /* Mount point vnode. */ 944 uint64_t fsflags, /* Flags common to all filesystems. */ 945 struct vfsoptlist **optlist /* Options local to the filesystem. */ 946 ) 947 { 948 struct export_args export; 949 void *bufp; 950 struct mount *mp; 951 int error, export_error, len; 952 uint64_t flag; 953 954 ASSERT_VOP_ELOCKED(vp, __func__); 955 KASSERT((fsflags & MNT_UPDATE) != 0, ("MNT_UPDATE should be here")); 956 mp = vp->v_mount; 957 958 if ((vp->v_vflag & VV_ROOT) == 0) { 959 if (vfs_copyopt(*optlist, "export", &export, sizeof(export)) 960 == 0) 961 error = EXDEV; 962 else 963 error = EINVAL; 964 vput(vp); 965 return (error); 966 } 967 968 /* 969 * We only allow the filesystem to be reloaded if it 970 * is currently mounted read-only. 971 */ 972 flag = mp->mnt_flag; 973 if ((fsflags & MNT_RELOAD) != 0 && (flag & MNT_RDONLY) == 0) { 974 vput(vp); 975 return (EOPNOTSUPP); /* Needs translation */ 976 } 977 /* 978 * Only privileged root, or (if MNT_USER is set) the user that 979 * did the original mount is permitted to update it. 980 */ 981 error = vfs_suser(mp, td); 982 if (error != 0) { 983 vput(vp); 984 return (error); 985 } 986 if (vfs_busy(mp, MBF_NOWAIT)) { 987 vput(vp); 988 return (EBUSY); 989 } 990 VI_LOCK(vp); 991 if ((vp->v_iflag & VI_MOUNT) != 0 || vp->v_mountedhere != NULL) { 992 VI_UNLOCK(vp); 993 vfs_unbusy(mp); 994 vput(vp); 995 return (EBUSY); 996 } 997 vp->v_iflag |= VI_MOUNT; 998 VI_UNLOCK(vp); 999 VOP_UNLOCK(vp, 0); 1000 1001 MNT_ILOCK(mp); 1002 if ((mp->mnt_kern_flag & MNTK_UNMOUNT) != 0) { 1003 MNT_IUNLOCK(mp); 1004 error = EBUSY; 1005 goto end; 1006 } 1007 mp->mnt_flag &= ~MNT_UPDATEMASK; 1008 mp->mnt_flag |= fsflags & (MNT_RELOAD | MNT_FORCE | MNT_UPDATE | 1009 MNT_SNAPSHOT | MNT_ROOTFS | MNT_UPDATEMASK | MNT_RDONLY); 1010 if ((mp->mnt_flag & MNT_ASYNC) == 0) 1011 mp->mnt_kern_flag &= ~MNTK_ASYNC; 1012 MNT_IUNLOCK(mp); 1013 mp->mnt_optnew = *optlist; 1014 vfs_mergeopts(mp->mnt_optnew, mp->mnt_opt); 1015 1016 /* 1017 * Mount the filesystem. 1018 * XXX The final recipients of VFS_MOUNT just overwrite the ndp they 1019 * get. No freeing of cn_pnbuf. 1020 */ 1021 error = VFS_MOUNT(mp); 1022 1023 export_error = 0; 1024 /* Process the export option. */ 1025 if (error == 0 && vfs_getopt(mp->mnt_optnew, "export", &bufp, 1026 &len) == 0) { 1027 /* Assume that there is only 1 ABI for each length. */ 1028 switch (len) { 1029 case (sizeof(struct oexport_args)): 1030 bzero(&export, sizeof(export)); 1031 /* FALLTHROUGH */ 1032 case (sizeof(export)): 1033 bcopy(bufp, &export, len); 1034 export_error = vfs_export(mp, &export); 1035 break; 1036 default: 1037 export_error = EINVAL; 1038 break; 1039 } 1040 } 1041 1042 MNT_ILOCK(mp); 1043 if (error == 0) { 1044 mp->mnt_flag &= ~(MNT_UPDATE | MNT_RELOAD | MNT_FORCE | 1045 MNT_SNAPSHOT); 1046 } else { 1047 /* 1048 * If we fail, restore old mount flags. MNT_QUOTA is special, 1049 * because it is not part of MNT_UPDATEMASK, but it could have 1050 * changed in the meantime if quotactl(2) was called. 1051 * All in all we want current value of MNT_QUOTA, not the old 1052 * one. 1053 */ 1054 mp->mnt_flag = (mp->mnt_flag & MNT_QUOTA) | (flag & ~MNT_QUOTA); 1055 } 1056 if ((mp->mnt_flag & MNT_ASYNC) != 0 && 1057 (mp->mnt_kern_flag & MNTK_NOASYNC) == 0) 1058 mp->mnt_kern_flag |= MNTK_ASYNC; 1059 else 1060 mp->mnt_kern_flag &= ~MNTK_ASYNC; 1061 MNT_IUNLOCK(mp); 1062 1063 if (error != 0) 1064 goto end; 1065 1066 if (mp->mnt_opt != NULL) 1067 vfs_freeopts(mp->mnt_opt); 1068 mp->mnt_opt = mp->mnt_optnew; 1069 *optlist = NULL; 1070 (void)VFS_STATFS(mp, &mp->mnt_stat); 1071 /* 1072 * Prevent external consumers of mount options from reading 1073 * mnt_optnew. 1074 */ 1075 mp->mnt_optnew = NULL; 1076 1077 if ((mp->mnt_flag & MNT_RDONLY) == 0) 1078 vfs_allocate_syncvnode(mp); 1079 else 1080 vfs_deallocate_syncvnode(mp); 1081 end: 1082 vfs_unbusy(mp); 1083 VI_LOCK(vp); 1084 vp->v_iflag &= ~VI_MOUNT; 1085 VI_UNLOCK(vp); 1086 vrele(vp); 1087 return (error != 0 ? error : export_error); 1088 } 1089 1090 /* 1091 * vfs_domount(): actually attempt a filesystem mount. 1092 */ 1093 static int 1094 vfs_domount( 1095 struct thread *td, /* Calling thread. */ 1096 const char *fstype, /* Filesystem type. */ 1097 char *fspath, /* Mount path. */ 1098 uint64_t fsflags, /* Flags common to all filesystems. */ 1099 struct vfsoptlist **optlist /* Options local to the filesystem. */ 1100 ) 1101 { 1102 struct vfsconf *vfsp; 1103 struct nameidata nd; 1104 struct vnode *vp; 1105 char *pathbuf; 1106 int error; 1107 1108 /* 1109 * Be ultra-paranoid about making sure the type and fspath 1110 * variables will fit in our mp buffers, including the 1111 * terminating NUL. 1112 */ 1113 if (strlen(fstype) >= MFSNAMELEN || strlen(fspath) >= MNAMELEN) 1114 return (ENAMETOOLONG); 1115 1116 if (jailed(td->td_ucred) || usermount == 0) { 1117 if ((error = priv_check(td, PRIV_VFS_MOUNT)) != 0) 1118 return (error); 1119 } 1120 1121 /* 1122 * Do not allow NFS export or MNT_SUIDDIR by unprivileged users. 1123 */ 1124 if (fsflags & MNT_EXPORTED) { 1125 error = priv_check(td, PRIV_VFS_MOUNT_EXPORTED); 1126 if (error) 1127 return (error); 1128 } 1129 if (fsflags & MNT_SUIDDIR) { 1130 error = priv_check(td, PRIV_VFS_MOUNT_SUIDDIR); 1131 if (error) 1132 return (error); 1133 } 1134 /* 1135 * Silently enforce MNT_NOSUID and MNT_USER for unprivileged users. 1136 */ 1137 if ((fsflags & (MNT_NOSUID | MNT_USER)) != (MNT_NOSUID | MNT_USER)) { 1138 if (priv_check(td, PRIV_VFS_MOUNT_NONUSER) != 0) 1139 fsflags |= MNT_NOSUID | MNT_USER; 1140 } 1141 1142 /* Load KLDs before we lock the covered vnode to avoid reversals. */ 1143 vfsp = NULL; 1144 if ((fsflags & MNT_UPDATE) == 0) { 1145 /* Don't try to load KLDs if we're mounting the root. */ 1146 if (fsflags & MNT_ROOTFS) 1147 vfsp = vfs_byname(fstype); 1148 else 1149 vfsp = vfs_byname_kld(fstype, td, &error); 1150 if (vfsp == NULL) 1151 return (ENODEV); 1152 if (jailed(td->td_ucred) && !(vfsp->vfc_flags & VFCF_JAIL)) 1153 return (EPERM); 1154 } 1155 1156 /* 1157 * Get vnode to be covered or mount point's vnode in case of MNT_UPDATE. 1158 */ 1159 NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF | AUDITVNODE1, 1160 UIO_SYSSPACE, fspath, td); 1161 error = namei(&nd); 1162 if (error != 0) 1163 return (error); 1164 NDFREE(&nd, NDF_ONLY_PNBUF); 1165 vp = nd.ni_vp; 1166 if ((fsflags & MNT_UPDATE) == 0) { 1167 pathbuf = malloc(MNAMELEN, M_TEMP, M_WAITOK); 1168 strcpy(pathbuf, fspath); 1169 error = vn_path_to_global_path(td, vp, pathbuf, MNAMELEN); 1170 /* debug.disablefullpath == 1 results in ENODEV */ 1171 if (error == 0 || error == ENODEV) { 1172 error = vfs_domount_first(td, vfsp, pathbuf, vp, 1173 fsflags, optlist); 1174 } 1175 free(pathbuf, M_TEMP); 1176 } else 1177 error = vfs_domount_update(td, vp, fsflags, optlist); 1178 1179 return (error); 1180 } 1181 1182 /* 1183 * Unmount a filesystem. 1184 * 1185 * Note: unmount takes a path to the vnode mounted on as argument, not 1186 * special file (as before). 1187 */ 1188 #ifndef _SYS_SYSPROTO_H_ 1189 struct unmount_args { 1190 char *path; 1191 int flags; 1192 }; 1193 #endif 1194 /* ARGSUSED */ 1195 int 1196 sys_unmount(struct thread *td, struct unmount_args *uap) 1197 { 1198 struct nameidata nd; 1199 struct mount *mp; 1200 char *pathbuf; 1201 int error, id0, id1; 1202 1203 AUDIT_ARG_VALUE(uap->flags); 1204 if (jailed(td->td_ucred) || usermount == 0) { 1205 error = priv_check(td, PRIV_VFS_UNMOUNT); 1206 if (error) 1207 return (error); 1208 } 1209 1210 pathbuf = malloc(MNAMELEN, M_TEMP, M_WAITOK); 1211 error = copyinstr(uap->path, pathbuf, MNAMELEN, NULL); 1212 if (error) { 1213 free(pathbuf, M_TEMP); 1214 return (error); 1215 } 1216 if (uap->flags & MNT_BYFSID) { 1217 AUDIT_ARG_TEXT(pathbuf); 1218 /* Decode the filesystem ID. */ 1219 if (sscanf(pathbuf, "FSID:%d:%d", &id0, &id1) != 2) { 1220 free(pathbuf, M_TEMP); 1221 return (EINVAL); 1222 } 1223 1224 mtx_lock(&mountlist_mtx); 1225 TAILQ_FOREACH_REVERSE(mp, &mountlist, mntlist, mnt_list) { 1226 if (mp->mnt_stat.f_fsid.val[0] == id0 && 1227 mp->mnt_stat.f_fsid.val[1] == id1) { 1228 vfs_ref(mp); 1229 break; 1230 } 1231 } 1232 mtx_unlock(&mountlist_mtx); 1233 } else { 1234 /* 1235 * Try to find global path for path argument. 1236 */ 1237 NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF | AUDITVNODE1, 1238 UIO_SYSSPACE, pathbuf, td); 1239 if (namei(&nd) == 0) { 1240 NDFREE(&nd, NDF_ONLY_PNBUF); 1241 error = vn_path_to_global_path(td, nd.ni_vp, pathbuf, 1242 MNAMELEN); 1243 if (error == 0 || error == ENODEV) 1244 vput(nd.ni_vp); 1245 } 1246 mtx_lock(&mountlist_mtx); 1247 TAILQ_FOREACH_REVERSE(mp, &mountlist, mntlist, mnt_list) { 1248 if (strcmp(mp->mnt_stat.f_mntonname, pathbuf) == 0) { 1249 vfs_ref(mp); 1250 break; 1251 } 1252 } 1253 mtx_unlock(&mountlist_mtx); 1254 } 1255 free(pathbuf, M_TEMP); 1256 if (mp == NULL) { 1257 /* 1258 * Previously we returned ENOENT for a nonexistent path and 1259 * EINVAL for a non-mountpoint. We cannot tell these apart 1260 * now, so in the !MNT_BYFSID case return the more likely 1261 * EINVAL for compatibility. 1262 */ 1263 return ((uap->flags & MNT_BYFSID) ? ENOENT : EINVAL); 1264 } 1265 1266 /* 1267 * Don't allow unmounting the root filesystem. 1268 */ 1269 if (mp->mnt_flag & MNT_ROOTFS) { 1270 vfs_rel(mp); 1271 return (EINVAL); 1272 } 1273 error = dounmount(mp, uap->flags, td); 1274 return (error); 1275 } 1276 1277 /* 1278 * Return error if any of the vnodes, ignoring the root vnode 1279 * and the syncer vnode, have non-zero usecount. 1280 * 1281 * This function is purely advisory - it can return false positives 1282 * and negatives. 1283 */ 1284 static int 1285 vfs_check_usecounts(struct mount *mp) 1286 { 1287 struct vnode *vp, *mvp; 1288 1289 MNT_VNODE_FOREACH_ALL(vp, mp, mvp) { 1290 if ((vp->v_vflag & VV_ROOT) == 0 && vp->v_type != VNON && 1291 vp->v_usecount != 0) { 1292 VI_UNLOCK(vp); 1293 MNT_VNODE_FOREACH_ALL_ABORT(mp, mvp); 1294 return (EBUSY); 1295 } 1296 VI_UNLOCK(vp); 1297 } 1298 1299 return (0); 1300 } 1301 1302 static void 1303 dounmount_cleanup(struct mount *mp, struct vnode *coveredvp, int mntkflags) 1304 { 1305 1306 mtx_assert(MNT_MTX(mp), MA_OWNED); 1307 mp->mnt_kern_flag &= ~mntkflags; 1308 if ((mp->mnt_kern_flag & MNTK_MWAIT) != 0) { 1309 mp->mnt_kern_flag &= ~MNTK_MWAIT; 1310 wakeup(mp); 1311 } 1312 MNT_IUNLOCK(mp); 1313 if (coveredvp != NULL) { 1314 VOP_UNLOCK(coveredvp, 0); 1315 vdrop(coveredvp); 1316 } 1317 vn_finished_write(mp); 1318 } 1319 1320 /* 1321 * Do the actual filesystem unmount. 1322 */ 1323 int 1324 dounmount(struct mount *mp, int flags, struct thread *td) 1325 { 1326 struct vnode *coveredvp; 1327 int error; 1328 uint64_t async_flag; 1329 int mnt_gen_r; 1330 1331 if ((coveredvp = mp->mnt_vnodecovered) != NULL) { 1332 mnt_gen_r = mp->mnt_gen; 1333 VI_LOCK(coveredvp); 1334 vholdl(coveredvp); 1335 vn_lock(coveredvp, LK_EXCLUSIVE | LK_INTERLOCK | LK_RETRY); 1336 /* 1337 * Check for mp being unmounted while waiting for the 1338 * covered vnode lock. 1339 */ 1340 if (coveredvp->v_mountedhere != mp || 1341 coveredvp->v_mountedhere->mnt_gen != mnt_gen_r) { 1342 VOP_UNLOCK(coveredvp, 0); 1343 vdrop(coveredvp); 1344 vfs_rel(mp); 1345 return (EBUSY); 1346 } 1347 } 1348 1349 /* 1350 * Only privileged root, or (if MNT_USER is set) the user that did the 1351 * original mount is permitted to unmount this filesystem. 1352 */ 1353 error = vfs_suser(mp, td); 1354 if (error != 0) { 1355 if (coveredvp != NULL) { 1356 VOP_UNLOCK(coveredvp, 0); 1357 vdrop(coveredvp); 1358 } 1359 vfs_rel(mp); 1360 return (error); 1361 } 1362 1363 vn_start_write(NULL, &mp, V_WAIT | V_MNTREF); 1364 MNT_ILOCK(mp); 1365 if ((mp->mnt_kern_flag & MNTK_UNMOUNT) != 0 || 1366 (mp->mnt_flag & MNT_UPDATE) != 0 || 1367 !TAILQ_EMPTY(&mp->mnt_uppers)) { 1368 dounmount_cleanup(mp, coveredvp, 0); 1369 return (EBUSY); 1370 } 1371 mp->mnt_kern_flag |= MNTK_UNMOUNT | MNTK_NOINSMNTQ; 1372 if (flags & MNT_NONBUSY) { 1373 MNT_IUNLOCK(mp); 1374 error = vfs_check_usecounts(mp); 1375 MNT_ILOCK(mp); 1376 if (error != 0) { 1377 dounmount_cleanup(mp, coveredvp, MNTK_UNMOUNT | 1378 MNTK_NOINSMNTQ); 1379 return (error); 1380 } 1381 } 1382 /* Allow filesystems to detect that a forced unmount is in progress. */ 1383 if (flags & MNT_FORCE) { 1384 mp->mnt_kern_flag |= MNTK_UNMOUNTF; 1385 MNT_IUNLOCK(mp); 1386 /* 1387 * Must be done after setting MNTK_UNMOUNTF and before 1388 * waiting for mnt_lockref to become 0. 1389 */ 1390 VFS_PURGE(mp); 1391 MNT_ILOCK(mp); 1392 } 1393 error = 0; 1394 if (mp->mnt_lockref) { 1395 mp->mnt_kern_flag |= MNTK_DRAINING; 1396 error = msleep(&mp->mnt_lockref, MNT_MTX(mp), PVFS, 1397 "mount drain", 0); 1398 } 1399 MNT_IUNLOCK(mp); 1400 KASSERT(mp->mnt_lockref == 0, 1401 ("%s: invalid lock refcount in the drain path @ %s:%d", 1402 __func__, __FILE__, __LINE__)); 1403 KASSERT(error == 0, 1404 ("%s: invalid return value for msleep in the drain path @ %s:%d", 1405 __func__, __FILE__, __LINE__)); 1406 1407 if (mp->mnt_flag & MNT_EXPUBLIC) 1408 vfs_setpublicfs(NULL, NULL, NULL); 1409 1410 /* 1411 * From now, we can claim that the use reference on the 1412 * coveredvp is ours, and the ref can be released only by 1413 * successfull unmount by us, or left for later unmount 1414 * attempt. The previously acquired hold reference is no 1415 * longer needed to protect the vnode from reuse. 1416 */ 1417 if (coveredvp != NULL) 1418 vdrop(coveredvp); 1419 1420 vfs_msync(mp, MNT_WAIT); 1421 MNT_ILOCK(mp); 1422 async_flag = mp->mnt_flag & MNT_ASYNC; 1423 mp->mnt_flag &= ~MNT_ASYNC; 1424 mp->mnt_kern_flag &= ~MNTK_ASYNC; 1425 MNT_IUNLOCK(mp); 1426 cache_purgevfs(mp, false); /* remove cache entries for this file sys */ 1427 vfs_deallocate_syncvnode(mp); 1428 if ((mp->mnt_flag & MNT_RDONLY) != 0 || (flags & MNT_FORCE) != 0 || 1429 (error = VFS_SYNC(mp, MNT_WAIT)) == 0) 1430 error = VFS_UNMOUNT(mp, flags); 1431 vn_finished_write(mp); 1432 /* 1433 * If we failed to flush the dirty blocks for this mount point, 1434 * undo all the cdir/rdir and rootvnode changes we made above. 1435 * Unless we failed to do so because the device is reporting that 1436 * it doesn't exist anymore. 1437 */ 1438 if (error && error != ENXIO) { 1439 MNT_ILOCK(mp); 1440 mp->mnt_kern_flag &= ~MNTK_NOINSMNTQ; 1441 if ((mp->mnt_flag & MNT_RDONLY) == 0) { 1442 MNT_IUNLOCK(mp); 1443 vfs_allocate_syncvnode(mp); 1444 MNT_ILOCK(mp); 1445 } 1446 mp->mnt_kern_flag &= ~(MNTK_UNMOUNT | MNTK_UNMOUNTF); 1447 mp->mnt_flag |= async_flag; 1448 if ((mp->mnt_flag & MNT_ASYNC) != 0 && 1449 (mp->mnt_kern_flag & MNTK_NOASYNC) == 0) 1450 mp->mnt_kern_flag |= MNTK_ASYNC; 1451 if (mp->mnt_kern_flag & MNTK_MWAIT) { 1452 mp->mnt_kern_flag &= ~MNTK_MWAIT; 1453 wakeup(mp); 1454 } 1455 MNT_IUNLOCK(mp); 1456 if (coveredvp) 1457 VOP_UNLOCK(coveredvp, 0); 1458 return (error); 1459 } 1460 mtx_lock(&mountlist_mtx); 1461 TAILQ_REMOVE(&mountlist, mp, mnt_list); 1462 mtx_unlock(&mountlist_mtx); 1463 EVENTHANDLER_DIRECT_INVOKE(vfs_unmounted, mp, td); 1464 if (coveredvp != NULL) { 1465 coveredvp->v_mountedhere = NULL; 1466 VOP_UNLOCK(coveredvp, 0); 1467 } 1468 vfs_event_signal(NULL, VQ_UNMOUNT, 0); 1469 if (rootvnode != NULL && mp == rootvnode->v_mount) { 1470 vrele(rootvnode); 1471 rootvnode = NULL; 1472 } 1473 if (mp == rootdevmp) 1474 rootdevmp = NULL; 1475 vfs_mount_destroy(mp); 1476 return (0); 1477 } 1478 1479 /* 1480 * Report errors during filesystem mounting. 1481 */ 1482 void 1483 vfs_mount_error(struct mount *mp, const char *fmt, ...) 1484 { 1485 struct vfsoptlist *moptlist = mp->mnt_optnew; 1486 va_list ap; 1487 int error, len; 1488 char *errmsg; 1489 1490 error = vfs_getopt(moptlist, "errmsg", (void **)&errmsg, &len); 1491 if (error || errmsg == NULL || len <= 0) 1492 return; 1493 1494 va_start(ap, fmt); 1495 vsnprintf(errmsg, (size_t)len, fmt, ap); 1496 va_end(ap); 1497 } 1498 1499 void 1500 vfs_opterror(struct vfsoptlist *opts, const char *fmt, ...) 1501 { 1502 va_list ap; 1503 int error, len; 1504 char *errmsg; 1505 1506 error = vfs_getopt(opts, "errmsg", (void **)&errmsg, &len); 1507 if (error || errmsg == NULL || len <= 0) 1508 return; 1509 1510 va_start(ap, fmt); 1511 vsnprintf(errmsg, (size_t)len, fmt, ap); 1512 va_end(ap); 1513 } 1514 1515 /* 1516 * --------------------------------------------------------------------- 1517 * Functions for querying mount options/arguments from filesystems. 1518 */ 1519 1520 /* 1521 * Check that no unknown options are given 1522 */ 1523 int 1524 vfs_filteropt(struct vfsoptlist *opts, const char **legal) 1525 { 1526 struct vfsopt *opt; 1527 char errmsg[255]; 1528 const char **t, *p, *q; 1529 int ret = 0; 1530 1531 TAILQ_FOREACH(opt, opts, link) { 1532 p = opt->name; 1533 q = NULL; 1534 if (p[0] == 'n' && p[1] == 'o') 1535 q = p + 2; 1536 for(t = global_opts; *t != NULL; t++) { 1537 if (strcmp(*t, p) == 0) 1538 break; 1539 if (q != NULL) { 1540 if (strcmp(*t, q) == 0) 1541 break; 1542 } 1543 } 1544 if (*t != NULL) 1545 continue; 1546 for(t = legal; *t != NULL; t++) { 1547 if (strcmp(*t, p) == 0) 1548 break; 1549 if (q != NULL) { 1550 if (strcmp(*t, q) == 0) 1551 break; 1552 } 1553 } 1554 if (*t != NULL) 1555 continue; 1556 snprintf(errmsg, sizeof(errmsg), 1557 "mount option <%s> is unknown", p); 1558 ret = EINVAL; 1559 } 1560 if (ret != 0) { 1561 TAILQ_FOREACH(opt, opts, link) { 1562 if (strcmp(opt->name, "errmsg") == 0) { 1563 strncpy((char *)opt->value, errmsg, opt->len); 1564 break; 1565 } 1566 } 1567 if (opt == NULL) 1568 printf("%s\n", errmsg); 1569 } 1570 return (ret); 1571 } 1572 1573 /* 1574 * Get a mount option by its name. 1575 * 1576 * Return 0 if the option was found, ENOENT otherwise. 1577 * If len is non-NULL it will be filled with the length 1578 * of the option. If buf is non-NULL, it will be filled 1579 * with the address of the option. 1580 */ 1581 int 1582 vfs_getopt(struct vfsoptlist *opts, const char *name, void **buf, int *len) 1583 { 1584 struct vfsopt *opt; 1585 1586 KASSERT(opts != NULL, ("vfs_getopt: caller passed 'opts' as NULL")); 1587 1588 TAILQ_FOREACH(opt, opts, link) { 1589 if (strcmp(name, opt->name) == 0) { 1590 opt->seen = 1; 1591 if (len != NULL) 1592 *len = opt->len; 1593 if (buf != NULL) 1594 *buf = opt->value; 1595 return (0); 1596 } 1597 } 1598 return (ENOENT); 1599 } 1600 1601 int 1602 vfs_getopt_pos(struct vfsoptlist *opts, const char *name) 1603 { 1604 struct vfsopt *opt; 1605 1606 if (opts == NULL) 1607 return (-1); 1608 1609 TAILQ_FOREACH(opt, opts, link) { 1610 if (strcmp(name, opt->name) == 0) { 1611 opt->seen = 1; 1612 return (opt->pos); 1613 } 1614 } 1615 return (-1); 1616 } 1617 1618 int 1619 vfs_getopt_size(struct vfsoptlist *opts, const char *name, off_t *value) 1620 { 1621 char *opt_value, *vtp; 1622 quad_t iv; 1623 int error, opt_len; 1624 1625 error = vfs_getopt(opts, name, (void **)&opt_value, &opt_len); 1626 if (error != 0) 1627 return (error); 1628 if (opt_len == 0 || opt_value == NULL) 1629 return (EINVAL); 1630 if (opt_value[0] == '\0' || opt_value[opt_len - 1] != '\0') 1631 return (EINVAL); 1632 iv = strtoq(opt_value, &vtp, 0); 1633 if (vtp == opt_value || (vtp[0] != '\0' && vtp[1] != '\0')) 1634 return (EINVAL); 1635 if (iv < 0) 1636 return (EINVAL); 1637 switch (vtp[0]) { 1638 case 't': 1639 case 'T': 1640 iv *= 1024; 1641 case 'g': 1642 case 'G': 1643 iv *= 1024; 1644 case 'm': 1645 case 'M': 1646 iv *= 1024; 1647 case 'k': 1648 case 'K': 1649 iv *= 1024; 1650 case '\0': 1651 break; 1652 default: 1653 return (EINVAL); 1654 } 1655 *value = iv; 1656 1657 return (0); 1658 } 1659 1660 char * 1661 vfs_getopts(struct vfsoptlist *opts, const char *name, int *error) 1662 { 1663 struct vfsopt *opt; 1664 1665 *error = 0; 1666 TAILQ_FOREACH(opt, opts, link) { 1667 if (strcmp(name, opt->name) != 0) 1668 continue; 1669 opt->seen = 1; 1670 if (opt->len == 0 || 1671 ((char *)opt->value)[opt->len - 1] != '\0') { 1672 *error = EINVAL; 1673 return (NULL); 1674 } 1675 return (opt->value); 1676 } 1677 *error = ENOENT; 1678 return (NULL); 1679 } 1680 1681 int 1682 vfs_flagopt(struct vfsoptlist *opts, const char *name, uint64_t *w, 1683 uint64_t val) 1684 { 1685 struct vfsopt *opt; 1686 1687 TAILQ_FOREACH(opt, opts, link) { 1688 if (strcmp(name, opt->name) == 0) { 1689 opt->seen = 1; 1690 if (w != NULL) 1691 *w |= val; 1692 return (1); 1693 } 1694 } 1695 if (w != NULL) 1696 *w &= ~val; 1697 return (0); 1698 } 1699 1700 int 1701 vfs_scanopt(struct vfsoptlist *opts, const char *name, const char *fmt, ...) 1702 { 1703 va_list ap; 1704 struct vfsopt *opt; 1705 int ret; 1706 1707 KASSERT(opts != NULL, ("vfs_getopt: caller passed 'opts' as NULL")); 1708 1709 TAILQ_FOREACH(opt, opts, link) { 1710 if (strcmp(name, opt->name) != 0) 1711 continue; 1712 opt->seen = 1; 1713 if (opt->len == 0 || opt->value == NULL) 1714 return (0); 1715 if (((char *)opt->value)[opt->len - 1] != '\0') 1716 return (0); 1717 va_start(ap, fmt); 1718 ret = vsscanf(opt->value, fmt, ap); 1719 va_end(ap); 1720 return (ret); 1721 } 1722 return (0); 1723 } 1724 1725 int 1726 vfs_setopt(struct vfsoptlist *opts, const char *name, void *value, int len) 1727 { 1728 struct vfsopt *opt; 1729 1730 TAILQ_FOREACH(opt, opts, link) { 1731 if (strcmp(name, opt->name) != 0) 1732 continue; 1733 opt->seen = 1; 1734 if (opt->value == NULL) 1735 opt->len = len; 1736 else { 1737 if (opt->len != len) 1738 return (EINVAL); 1739 bcopy(value, opt->value, len); 1740 } 1741 return (0); 1742 } 1743 return (ENOENT); 1744 } 1745 1746 int 1747 vfs_setopt_part(struct vfsoptlist *opts, const char *name, void *value, int len) 1748 { 1749 struct vfsopt *opt; 1750 1751 TAILQ_FOREACH(opt, opts, link) { 1752 if (strcmp(name, opt->name) != 0) 1753 continue; 1754 opt->seen = 1; 1755 if (opt->value == NULL) 1756 opt->len = len; 1757 else { 1758 if (opt->len < len) 1759 return (EINVAL); 1760 opt->len = len; 1761 bcopy(value, opt->value, len); 1762 } 1763 return (0); 1764 } 1765 return (ENOENT); 1766 } 1767 1768 int 1769 vfs_setopts(struct vfsoptlist *opts, const char *name, const char *value) 1770 { 1771 struct vfsopt *opt; 1772 1773 TAILQ_FOREACH(opt, opts, link) { 1774 if (strcmp(name, opt->name) != 0) 1775 continue; 1776 opt->seen = 1; 1777 if (opt->value == NULL) 1778 opt->len = strlen(value) + 1; 1779 else if (strlcpy(opt->value, value, opt->len) >= opt->len) 1780 return (EINVAL); 1781 return (0); 1782 } 1783 return (ENOENT); 1784 } 1785 1786 /* 1787 * Find and copy a mount option. 1788 * 1789 * The size of the buffer has to be specified 1790 * in len, if it is not the same length as the 1791 * mount option, EINVAL is returned. 1792 * Returns ENOENT if the option is not found. 1793 */ 1794 int 1795 vfs_copyopt(struct vfsoptlist *opts, const char *name, void *dest, int len) 1796 { 1797 struct vfsopt *opt; 1798 1799 KASSERT(opts != NULL, ("vfs_copyopt: caller passed 'opts' as NULL")); 1800 1801 TAILQ_FOREACH(opt, opts, link) { 1802 if (strcmp(name, opt->name) == 0) { 1803 opt->seen = 1; 1804 if (len != opt->len) 1805 return (EINVAL); 1806 bcopy(opt->value, dest, opt->len); 1807 return (0); 1808 } 1809 } 1810 return (ENOENT); 1811 } 1812 1813 int 1814 __vfs_statfs(struct mount *mp, struct statfs *sbp) 1815 { 1816 int error; 1817 1818 error = mp->mnt_op->vfs_statfs(mp, &mp->mnt_stat); 1819 if (sbp != &mp->mnt_stat) 1820 *sbp = mp->mnt_stat; 1821 return (error); 1822 } 1823 1824 void 1825 vfs_mountedfrom(struct mount *mp, const char *from) 1826 { 1827 1828 bzero(mp->mnt_stat.f_mntfromname, sizeof mp->mnt_stat.f_mntfromname); 1829 strlcpy(mp->mnt_stat.f_mntfromname, from, 1830 sizeof mp->mnt_stat.f_mntfromname); 1831 } 1832 1833 /* 1834 * --------------------------------------------------------------------- 1835 * This is the api for building mount args and mounting filesystems from 1836 * inside the kernel. 1837 * 1838 * The API works by accumulation of individual args. First error is 1839 * latched. 1840 * 1841 * XXX: should be documented in new manpage kernel_mount(9) 1842 */ 1843 1844 /* A memory allocation which must be freed when we are done */ 1845 struct mntaarg { 1846 SLIST_ENTRY(mntaarg) next; 1847 }; 1848 1849 /* The header for the mount arguments */ 1850 struct mntarg { 1851 struct iovec *v; 1852 int len; 1853 int error; 1854 SLIST_HEAD(, mntaarg) list; 1855 }; 1856 1857 /* 1858 * Add a boolean argument. 1859 * 1860 * flag is the boolean value. 1861 * name must start with "no". 1862 */ 1863 struct mntarg * 1864 mount_argb(struct mntarg *ma, int flag, const char *name) 1865 { 1866 1867 KASSERT(name[0] == 'n' && name[1] == 'o', 1868 ("mount_argb(...,%s): name must start with 'no'", name)); 1869 1870 return (mount_arg(ma, name + (flag ? 2 : 0), NULL, 0)); 1871 } 1872 1873 /* 1874 * Add an argument printf style 1875 */ 1876 struct mntarg * 1877 mount_argf(struct mntarg *ma, const char *name, const char *fmt, ...) 1878 { 1879 va_list ap; 1880 struct mntaarg *maa; 1881 struct sbuf *sb; 1882 int len; 1883 1884 if (ma == NULL) { 1885 ma = malloc(sizeof *ma, M_MOUNT, M_WAITOK | M_ZERO); 1886 SLIST_INIT(&ma->list); 1887 } 1888 if (ma->error) 1889 return (ma); 1890 1891 ma->v = realloc(ma->v, sizeof *ma->v * (ma->len + 2), 1892 M_MOUNT, M_WAITOK); 1893 ma->v[ma->len].iov_base = (void *)(uintptr_t)name; 1894 ma->v[ma->len].iov_len = strlen(name) + 1; 1895 ma->len++; 1896 1897 sb = sbuf_new_auto(); 1898 va_start(ap, fmt); 1899 sbuf_vprintf(sb, fmt, ap); 1900 va_end(ap); 1901 sbuf_finish(sb); 1902 len = sbuf_len(sb) + 1; 1903 maa = malloc(sizeof *maa + len, M_MOUNT, M_WAITOK | M_ZERO); 1904 SLIST_INSERT_HEAD(&ma->list, maa, next); 1905 bcopy(sbuf_data(sb), maa + 1, len); 1906 sbuf_delete(sb); 1907 1908 ma->v[ma->len].iov_base = maa + 1; 1909 ma->v[ma->len].iov_len = len; 1910 ma->len++; 1911 1912 return (ma); 1913 } 1914 1915 /* 1916 * Add an argument which is a userland string. 1917 */ 1918 struct mntarg * 1919 mount_argsu(struct mntarg *ma, const char *name, const void *val, int len) 1920 { 1921 struct mntaarg *maa; 1922 char *tbuf; 1923 1924 if (val == NULL) 1925 return (ma); 1926 if (ma == NULL) { 1927 ma = malloc(sizeof *ma, M_MOUNT, M_WAITOK | M_ZERO); 1928 SLIST_INIT(&ma->list); 1929 } 1930 if (ma->error) 1931 return (ma); 1932 maa = malloc(sizeof *maa + len, M_MOUNT, M_WAITOK | M_ZERO); 1933 SLIST_INSERT_HEAD(&ma->list, maa, next); 1934 tbuf = (void *)(maa + 1); 1935 ma->error = copyinstr(val, tbuf, len, NULL); 1936 return (mount_arg(ma, name, tbuf, -1)); 1937 } 1938 1939 /* 1940 * Plain argument. 1941 * 1942 * If length is -1, treat value as a C string. 1943 */ 1944 struct mntarg * 1945 mount_arg(struct mntarg *ma, const char *name, const void *val, int len) 1946 { 1947 1948 if (ma == NULL) { 1949 ma = malloc(sizeof *ma, M_MOUNT, M_WAITOK | M_ZERO); 1950 SLIST_INIT(&ma->list); 1951 } 1952 if (ma->error) 1953 return (ma); 1954 1955 ma->v = realloc(ma->v, sizeof *ma->v * (ma->len + 2), 1956 M_MOUNT, M_WAITOK); 1957 ma->v[ma->len].iov_base = (void *)(uintptr_t)name; 1958 ma->v[ma->len].iov_len = strlen(name) + 1; 1959 ma->len++; 1960 1961 ma->v[ma->len].iov_base = (void *)(uintptr_t)val; 1962 if (len < 0) 1963 ma->v[ma->len].iov_len = strlen(val) + 1; 1964 else 1965 ma->v[ma->len].iov_len = len; 1966 ma->len++; 1967 return (ma); 1968 } 1969 1970 /* 1971 * Free a mntarg structure 1972 */ 1973 static void 1974 free_mntarg(struct mntarg *ma) 1975 { 1976 struct mntaarg *maa; 1977 1978 while (!SLIST_EMPTY(&ma->list)) { 1979 maa = SLIST_FIRST(&ma->list); 1980 SLIST_REMOVE_HEAD(&ma->list, next); 1981 free(maa, M_MOUNT); 1982 } 1983 free(ma->v, M_MOUNT); 1984 free(ma, M_MOUNT); 1985 } 1986 1987 /* 1988 * Mount a filesystem 1989 */ 1990 int 1991 kernel_mount(struct mntarg *ma, uint64_t flags) 1992 { 1993 struct uio auio; 1994 int error; 1995 1996 KASSERT(ma != NULL, ("kernel_mount NULL ma")); 1997 KASSERT(ma->v != NULL, ("kernel_mount NULL ma->v")); 1998 KASSERT(!(ma->len & 1), ("kernel_mount odd ma->len (%d)", ma->len)); 1999 2000 auio.uio_iov = ma->v; 2001 auio.uio_iovcnt = ma->len; 2002 auio.uio_segflg = UIO_SYSSPACE; 2003 2004 error = ma->error; 2005 if (!error) 2006 error = vfs_donmount(curthread, flags, &auio); 2007 free_mntarg(ma); 2008 return (error); 2009 } 2010 2011 /* 2012 * A printflike function to mount a filesystem. 2013 */ 2014 int 2015 kernel_vmount(int flags, ...) 2016 { 2017 struct mntarg *ma = NULL; 2018 va_list ap; 2019 const char *cp; 2020 const void *vp; 2021 int error; 2022 2023 va_start(ap, flags); 2024 for (;;) { 2025 cp = va_arg(ap, const char *); 2026 if (cp == NULL) 2027 break; 2028 vp = va_arg(ap, const void *); 2029 ma = mount_arg(ma, cp, vp, (vp != NULL ? -1 : 0)); 2030 } 2031 va_end(ap); 2032 2033 error = kernel_mount(ma, flags); 2034 return (error); 2035 } 2036 2037 void 2038 vfs_oexport_conv(const struct oexport_args *oexp, struct export_args *exp) 2039 { 2040 2041 bcopy(oexp, exp, sizeof(*oexp)); 2042 exp->ex_numsecflavors = 0; 2043 } 2044