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_flags & VFCF_SBDRY) != 0 && 812 vfsp->vfc_vfsops_sd->vfs_cmount == NULL) || 813 ((vfsp->vfc_flags & VFCF_SBDRY) == 0 && 814 vfsp->vfc_vfsops->vfs_cmount == NULL)) 815 return (EOPNOTSUPP); 816 817 ma = mount_argsu(ma, "fstype", uap->type, MFSNAMELEN); 818 ma = mount_argsu(ma, "fspath", uap->path, MNAMELEN); 819 ma = mount_argb(ma, flags & MNT_RDONLY, "noro"); 820 ma = mount_argb(ma, !(flags & MNT_NOSUID), "nosuid"); 821 ma = mount_argb(ma, !(flags & MNT_NOEXEC), "noexec"); 822 823 if ((vfsp->vfc_flags & VFCF_SBDRY) != 0) 824 return (vfsp->vfc_vfsops_sd->vfs_cmount(ma, uap->data, flags)); 825 return (vfsp->vfc_vfsops->vfs_cmount(ma, uap->data, flags)); 826 } 827 828 /* 829 * vfs_domount_first(): first file system mount (not update) 830 */ 831 static int 832 vfs_domount_first( 833 struct thread *td, /* Calling thread. */ 834 struct vfsconf *vfsp, /* File system type. */ 835 char *fspath, /* Mount path. */ 836 struct vnode *vp, /* Vnode to be covered. */ 837 uint64_t fsflags, /* Flags common to all filesystems. */ 838 struct vfsoptlist **optlist /* Options local to the filesystem. */ 839 ) 840 { 841 struct vattr va; 842 struct mount *mp; 843 struct vnode *newdp; 844 int error; 845 846 ASSERT_VOP_ELOCKED(vp, __func__); 847 KASSERT((fsflags & MNT_UPDATE) == 0, ("MNT_UPDATE shouldn't be here")); 848 849 /* 850 * If the jail of the calling thread lacks permission for this type of 851 * file system, deny immediately. 852 */ 853 if (jailed(td->td_ucred) && !prison_allow(td->td_ucred, 854 vfsp->vfc_prison_flag)) { 855 vput(vp); 856 return (EPERM); 857 } 858 859 /* 860 * If the user is not root, ensure that they own the directory 861 * onto which we are attempting to mount. 862 */ 863 error = VOP_GETATTR(vp, &va, td->td_ucred); 864 if (error == 0 && va.va_uid != td->td_ucred->cr_uid) 865 error = priv_check_cred(td->td_ucred, PRIV_VFS_ADMIN, 0); 866 if (error == 0) 867 error = vinvalbuf(vp, V_SAVE, 0, 0); 868 if (error == 0 && vp->v_type != VDIR) 869 error = ENOTDIR; 870 if (error == 0) { 871 VI_LOCK(vp); 872 if ((vp->v_iflag & VI_MOUNT) == 0 && vp->v_mountedhere == NULL) 873 vp->v_iflag |= VI_MOUNT; 874 else 875 error = EBUSY; 876 VI_UNLOCK(vp); 877 } 878 if (error != 0) { 879 vput(vp); 880 return (error); 881 } 882 VOP_UNLOCK(vp, 0); 883 884 /* Allocate and initialize the filesystem. */ 885 mp = vfs_mount_alloc(vp, vfsp, fspath, td->td_ucred); 886 /* XXXMAC: pass to vfs_mount_alloc? */ 887 mp->mnt_optnew = *optlist; 888 /* Set the mount level flags. */ 889 mp->mnt_flag = (fsflags & (MNT_UPDATEMASK | MNT_ROOTFS | MNT_RDONLY)); 890 891 /* 892 * Mount the filesystem. 893 * XXX The final recipients of VFS_MOUNT just overwrite the ndp they 894 * get. No freeing of cn_pnbuf. 895 */ 896 error = VFS_MOUNT(mp); 897 if (error != 0) { 898 vfs_unbusy(mp); 899 mp->mnt_vnodecovered = NULL; 900 vfs_mount_destroy(mp); 901 VI_LOCK(vp); 902 vp->v_iflag &= ~VI_MOUNT; 903 VI_UNLOCK(vp); 904 vrele(vp); 905 return (error); 906 } 907 908 if (mp->mnt_opt != NULL) 909 vfs_freeopts(mp->mnt_opt); 910 mp->mnt_opt = mp->mnt_optnew; 911 *optlist = NULL; 912 (void)VFS_STATFS(mp, &mp->mnt_stat); 913 914 /* 915 * Prevent external consumers of mount options from reading mnt_optnew. 916 */ 917 mp->mnt_optnew = NULL; 918 919 MNT_ILOCK(mp); 920 if ((mp->mnt_flag & MNT_ASYNC) != 0 && 921 (mp->mnt_kern_flag & MNTK_NOASYNC) == 0) 922 mp->mnt_kern_flag |= MNTK_ASYNC; 923 else 924 mp->mnt_kern_flag &= ~MNTK_ASYNC; 925 MNT_IUNLOCK(mp); 926 927 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 928 cache_purge(vp); 929 VI_LOCK(vp); 930 vp->v_iflag &= ~VI_MOUNT; 931 VI_UNLOCK(vp); 932 vp->v_mountedhere = mp; 933 /* Place the new filesystem at the end of the mount list. */ 934 mtx_lock(&mountlist_mtx); 935 TAILQ_INSERT_TAIL(&mountlist, mp, mnt_list); 936 mtx_unlock(&mountlist_mtx); 937 vfs_event_signal(NULL, VQ_MOUNT, 0); 938 if (VFS_ROOT(mp, LK_EXCLUSIVE, &newdp)) 939 panic("mount: lost mount"); 940 VOP_UNLOCK(vp, 0); 941 EVENTHANDLER_DIRECT_INVOKE(vfs_mounted, mp, newdp, td); 942 VOP_UNLOCK(newdp, 0); 943 mountcheckdirs(vp, newdp); 944 vrele(newdp); 945 if ((mp->mnt_flag & MNT_RDONLY) == 0) 946 vfs_allocate_syncvnode(mp); 947 vfs_unbusy(mp); 948 return (0); 949 } 950 951 /* 952 * vfs_domount_update(): update of mounted file system 953 */ 954 static int 955 vfs_domount_update( 956 struct thread *td, /* Calling thread. */ 957 struct vnode *vp, /* Mount point vnode. */ 958 uint64_t fsflags, /* Flags common to all filesystems. */ 959 struct vfsoptlist **optlist /* Options local to the filesystem. */ 960 ) 961 { 962 struct export_args export; 963 void *bufp; 964 struct mount *mp; 965 int error, export_error, len; 966 uint64_t flag; 967 968 ASSERT_VOP_ELOCKED(vp, __func__); 969 KASSERT((fsflags & MNT_UPDATE) != 0, ("MNT_UPDATE should be here")); 970 mp = vp->v_mount; 971 972 if ((vp->v_vflag & VV_ROOT) == 0) { 973 if (vfs_copyopt(*optlist, "export", &export, sizeof(export)) 974 == 0) 975 error = EXDEV; 976 else 977 error = EINVAL; 978 vput(vp); 979 return (error); 980 } 981 982 /* 983 * We only allow the filesystem to be reloaded if it 984 * is currently mounted read-only. 985 */ 986 flag = mp->mnt_flag; 987 if ((fsflags & MNT_RELOAD) != 0 && (flag & MNT_RDONLY) == 0) { 988 vput(vp); 989 return (EOPNOTSUPP); /* Needs translation */ 990 } 991 /* 992 * Only privileged root, or (if MNT_USER is set) the user that 993 * did the original mount is permitted to update it. 994 */ 995 error = vfs_suser(mp, td); 996 if (error != 0) { 997 vput(vp); 998 return (error); 999 } 1000 if (vfs_busy(mp, MBF_NOWAIT)) { 1001 vput(vp); 1002 return (EBUSY); 1003 } 1004 VI_LOCK(vp); 1005 if ((vp->v_iflag & VI_MOUNT) != 0 || vp->v_mountedhere != NULL) { 1006 VI_UNLOCK(vp); 1007 vfs_unbusy(mp); 1008 vput(vp); 1009 return (EBUSY); 1010 } 1011 vp->v_iflag |= VI_MOUNT; 1012 VI_UNLOCK(vp); 1013 VOP_UNLOCK(vp, 0); 1014 1015 MNT_ILOCK(mp); 1016 if ((mp->mnt_kern_flag & MNTK_UNMOUNT) != 0) { 1017 MNT_IUNLOCK(mp); 1018 error = EBUSY; 1019 goto end; 1020 } 1021 mp->mnt_flag &= ~MNT_UPDATEMASK; 1022 mp->mnt_flag |= fsflags & (MNT_RELOAD | MNT_FORCE | MNT_UPDATE | 1023 MNT_SNAPSHOT | MNT_ROOTFS | MNT_UPDATEMASK | MNT_RDONLY); 1024 if ((mp->mnt_flag & MNT_ASYNC) == 0) 1025 mp->mnt_kern_flag &= ~MNTK_ASYNC; 1026 MNT_IUNLOCK(mp); 1027 mp->mnt_optnew = *optlist; 1028 vfs_mergeopts(mp->mnt_optnew, mp->mnt_opt); 1029 1030 /* 1031 * Mount the filesystem. 1032 * XXX The final recipients of VFS_MOUNT just overwrite the ndp they 1033 * get. No freeing of cn_pnbuf. 1034 */ 1035 error = VFS_MOUNT(mp); 1036 1037 export_error = 0; 1038 /* Process the export option. */ 1039 if (error == 0 && vfs_getopt(mp->mnt_optnew, "export", &bufp, 1040 &len) == 0) { 1041 /* Assume that there is only 1 ABI for each length. */ 1042 switch (len) { 1043 case (sizeof(struct oexport_args)): 1044 bzero(&export, sizeof(export)); 1045 /* FALLTHROUGH */ 1046 case (sizeof(export)): 1047 bcopy(bufp, &export, len); 1048 export_error = vfs_export(mp, &export); 1049 break; 1050 default: 1051 export_error = EINVAL; 1052 break; 1053 } 1054 } 1055 1056 MNT_ILOCK(mp); 1057 if (error == 0) { 1058 mp->mnt_flag &= ~(MNT_UPDATE | MNT_RELOAD | MNT_FORCE | 1059 MNT_SNAPSHOT); 1060 } else { 1061 /* 1062 * If we fail, restore old mount flags. MNT_QUOTA is special, 1063 * because it is not part of MNT_UPDATEMASK, but it could have 1064 * changed in the meantime if quotactl(2) was called. 1065 * All in all we want current value of MNT_QUOTA, not the old 1066 * one. 1067 */ 1068 mp->mnt_flag = (mp->mnt_flag & MNT_QUOTA) | (flag & ~MNT_QUOTA); 1069 } 1070 if ((mp->mnt_flag & MNT_ASYNC) != 0 && 1071 (mp->mnt_kern_flag & MNTK_NOASYNC) == 0) 1072 mp->mnt_kern_flag |= MNTK_ASYNC; 1073 else 1074 mp->mnt_kern_flag &= ~MNTK_ASYNC; 1075 MNT_IUNLOCK(mp); 1076 1077 if (error != 0) 1078 goto end; 1079 1080 if (mp->mnt_opt != NULL) 1081 vfs_freeopts(mp->mnt_opt); 1082 mp->mnt_opt = mp->mnt_optnew; 1083 *optlist = NULL; 1084 (void)VFS_STATFS(mp, &mp->mnt_stat); 1085 /* 1086 * Prevent external consumers of mount options from reading 1087 * mnt_optnew. 1088 */ 1089 mp->mnt_optnew = NULL; 1090 1091 if ((mp->mnt_flag & MNT_RDONLY) == 0) 1092 vfs_allocate_syncvnode(mp); 1093 else 1094 vfs_deallocate_syncvnode(mp); 1095 end: 1096 vfs_unbusy(mp); 1097 VI_LOCK(vp); 1098 vp->v_iflag &= ~VI_MOUNT; 1099 VI_UNLOCK(vp); 1100 vrele(vp); 1101 return (error != 0 ? error : export_error); 1102 } 1103 1104 /* 1105 * vfs_domount(): actually attempt a filesystem mount. 1106 */ 1107 static int 1108 vfs_domount( 1109 struct thread *td, /* Calling thread. */ 1110 const char *fstype, /* Filesystem type. */ 1111 char *fspath, /* Mount path. */ 1112 uint64_t fsflags, /* Flags common to all filesystems. */ 1113 struct vfsoptlist **optlist /* Options local to the filesystem. */ 1114 ) 1115 { 1116 struct vfsconf *vfsp; 1117 struct nameidata nd; 1118 struct vnode *vp; 1119 char *pathbuf; 1120 int error; 1121 1122 /* 1123 * Be ultra-paranoid about making sure the type and fspath 1124 * variables will fit in our mp buffers, including the 1125 * terminating NUL. 1126 */ 1127 if (strlen(fstype) >= MFSNAMELEN || strlen(fspath) >= MNAMELEN) 1128 return (ENAMETOOLONG); 1129 1130 if (jailed(td->td_ucred) || usermount == 0) { 1131 if ((error = priv_check(td, PRIV_VFS_MOUNT)) != 0) 1132 return (error); 1133 } 1134 1135 /* 1136 * Do not allow NFS export or MNT_SUIDDIR by unprivileged users. 1137 */ 1138 if (fsflags & MNT_EXPORTED) { 1139 error = priv_check(td, PRIV_VFS_MOUNT_EXPORTED); 1140 if (error) 1141 return (error); 1142 } 1143 if (fsflags & MNT_SUIDDIR) { 1144 error = priv_check(td, PRIV_VFS_MOUNT_SUIDDIR); 1145 if (error) 1146 return (error); 1147 } 1148 /* 1149 * Silently enforce MNT_NOSUID and MNT_USER for unprivileged users. 1150 */ 1151 if ((fsflags & (MNT_NOSUID | MNT_USER)) != (MNT_NOSUID | MNT_USER)) { 1152 if (priv_check(td, PRIV_VFS_MOUNT_NONUSER) != 0) 1153 fsflags |= MNT_NOSUID | MNT_USER; 1154 } 1155 1156 /* Load KLDs before we lock the covered vnode to avoid reversals. */ 1157 vfsp = NULL; 1158 if ((fsflags & MNT_UPDATE) == 0) { 1159 /* Don't try to load KLDs if we're mounting the root. */ 1160 if (fsflags & MNT_ROOTFS) 1161 vfsp = vfs_byname(fstype); 1162 else 1163 vfsp = vfs_byname_kld(fstype, td, &error); 1164 if (vfsp == NULL) 1165 return (ENODEV); 1166 } 1167 1168 /* 1169 * Get vnode to be covered or mount point's vnode in case of MNT_UPDATE. 1170 */ 1171 NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF | AUDITVNODE1, 1172 UIO_SYSSPACE, fspath, td); 1173 error = namei(&nd); 1174 if (error != 0) 1175 return (error); 1176 NDFREE(&nd, NDF_ONLY_PNBUF); 1177 vp = nd.ni_vp; 1178 if ((fsflags & MNT_UPDATE) == 0) { 1179 pathbuf = malloc(MNAMELEN, M_TEMP, M_WAITOK); 1180 strcpy(pathbuf, fspath); 1181 error = vn_path_to_global_path(td, vp, pathbuf, MNAMELEN); 1182 /* debug.disablefullpath == 1 results in ENODEV */ 1183 if (error == 0 || error == ENODEV) { 1184 error = vfs_domount_first(td, vfsp, pathbuf, vp, 1185 fsflags, optlist); 1186 } 1187 free(pathbuf, M_TEMP); 1188 } else 1189 error = vfs_domount_update(td, vp, fsflags, optlist); 1190 1191 return (error); 1192 } 1193 1194 /* 1195 * Unmount a filesystem. 1196 * 1197 * Note: unmount takes a path to the vnode mounted on as argument, not 1198 * special file (as before). 1199 */ 1200 #ifndef _SYS_SYSPROTO_H_ 1201 struct unmount_args { 1202 char *path; 1203 int flags; 1204 }; 1205 #endif 1206 /* ARGSUSED */ 1207 int 1208 sys_unmount(struct thread *td, struct unmount_args *uap) 1209 { 1210 struct nameidata nd; 1211 struct mount *mp; 1212 char *pathbuf; 1213 int error, id0, id1; 1214 1215 AUDIT_ARG_VALUE(uap->flags); 1216 if (jailed(td->td_ucred) || usermount == 0) { 1217 error = priv_check(td, PRIV_VFS_UNMOUNT); 1218 if (error) 1219 return (error); 1220 } 1221 1222 pathbuf = malloc(MNAMELEN, M_TEMP, M_WAITOK); 1223 error = copyinstr(uap->path, pathbuf, MNAMELEN, NULL); 1224 if (error) { 1225 free(pathbuf, M_TEMP); 1226 return (error); 1227 } 1228 if (uap->flags & MNT_BYFSID) { 1229 AUDIT_ARG_TEXT(pathbuf); 1230 /* Decode the filesystem ID. */ 1231 if (sscanf(pathbuf, "FSID:%d:%d", &id0, &id1) != 2) { 1232 free(pathbuf, M_TEMP); 1233 return (EINVAL); 1234 } 1235 1236 mtx_lock(&mountlist_mtx); 1237 TAILQ_FOREACH_REVERSE(mp, &mountlist, mntlist, mnt_list) { 1238 if (mp->mnt_stat.f_fsid.val[0] == id0 && 1239 mp->mnt_stat.f_fsid.val[1] == id1) { 1240 vfs_ref(mp); 1241 break; 1242 } 1243 } 1244 mtx_unlock(&mountlist_mtx); 1245 } else { 1246 /* 1247 * Try to find global path for path argument. 1248 */ 1249 NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF | AUDITVNODE1, 1250 UIO_SYSSPACE, pathbuf, td); 1251 if (namei(&nd) == 0) { 1252 NDFREE(&nd, NDF_ONLY_PNBUF); 1253 error = vn_path_to_global_path(td, nd.ni_vp, pathbuf, 1254 MNAMELEN); 1255 if (error == 0 || error == ENODEV) 1256 vput(nd.ni_vp); 1257 } 1258 mtx_lock(&mountlist_mtx); 1259 TAILQ_FOREACH_REVERSE(mp, &mountlist, mntlist, mnt_list) { 1260 if (strcmp(mp->mnt_stat.f_mntonname, pathbuf) == 0) { 1261 vfs_ref(mp); 1262 break; 1263 } 1264 } 1265 mtx_unlock(&mountlist_mtx); 1266 } 1267 free(pathbuf, M_TEMP); 1268 if (mp == NULL) { 1269 /* 1270 * Previously we returned ENOENT for a nonexistent path and 1271 * EINVAL for a non-mountpoint. We cannot tell these apart 1272 * now, so in the !MNT_BYFSID case return the more likely 1273 * EINVAL for compatibility. 1274 */ 1275 return ((uap->flags & MNT_BYFSID) ? ENOENT : EINVAL); 1276 } 1277 1278 /* 1279 * Don't allow unmounting the root filesystem. 1280 */ 1281 if (mp->mnt_flag & MNT_ROOTFS) { 1282 vfs_rel(mp); 1283 return (EINVAL); 1284 } 1285 error = dounmount(mp, uap->flags, td); 1286 return (error); 1287 } 1288 1289 /* 1290 * Return error if any of the vnodes, ignoring the root vnode 1291 * and the syncer vnode, have non-zero usecount. 1292 * 1293 * This function is purely advisory - it can return false positives 1294 * and negatives. 1295 */ 1296 static int 1297 vfs_check_usecounts(struct mount *mp) 1298 { 1299 struct vnode *vp, *mvp; 1300 1301 MNT_VNODE_FOREACH_ALL(vp, mp, mvp) { 1302 if ((vp->v_vflag & VV_ROOT) == 0 && vp->v_type != VNON && 1303 vp->v_usecount != 0) { 1304 VI_UNLOCK(vp); 1305 MNT_VNODE_FOREACH_ALL_ABORT(mp, mvp); 1306 return (EBUSY); 1307 } 1308 VI_UNLOCK(vp); 1309 } 1310 1311 return (0); 1312 } 1313 1314 static void 1315 dounmount_cleanup(struct mount *mp, struct vnode *coveredvp, int mntkflags) 1316 { 1317 1318 mtx_assert(MNT_MTX(mp), MA_OWNED); 1319 mp->mnt_kern_flag &= ~mntkflags; 1320 if ((mp->mnt_kern_flag & MNTK_MWAIT) != 0) { 1321 mp->mnt_kern_flag &= ~MNTK_MWAIT; 1322 wakeup(mp); 1323 } 1324 MNT_IUNLOCK(mp); 1325 if (coveredvp != NULL) { 1326 VOP_UNLOCK(coveredvp, 0); 1327 vdrop(coveredvp); 1328 } 1329 vn_finished_write(mp); 1330 } 1331 1332 /* 1333 * Do the actual filesystem unmount. 1334 */ 1335 int 1336 dounmount(struct mount *mp, int flags, struct thread *td) 1337 { 1338 struct vnode *coveredvp; 1339 int error; 1340 uint64_t async_flag; 1341 int mnt_gen_r; 1342 1343 if ((coveredvp = mp->mnt_vnodecovered) != NULL) { 1344 mnt_gen_r = mp->mnt_gen; 1345 VI_LOCK(coveredvp); 1346 vholdl(coveredvp); 1347 vn_lock(coveredvp, LK_EXCLUSIVE | LK_INTERLOCK | LK_RETRY); 1348 /* 1349 * Check for mp being unmounted while waiting for the 1350 * covered vnode lock. 1351 */ 1352 if (coveredvp->v_mountedhere != mp || 1353 coveredvp->v_mountedhere->mnt_gen != mnt_gen_r) { 1354 VOP_UNLOCK(coveredvp, 0); 1355 vdrop(coveredvp); 1356 vfs_rel(mp); 1357 return (EBUSY); 1358 } 1359 } 1360 1361 /* 1362 * Only privileged root, or (if MNT_USER is set) the user that did the 1363 * original mount is permitted to unmount this filesystem. 1364 */ 1365 error = vfs_suser(mp, td); 1366 if (error != 0) { 1367 if (coveredvp != NULL) { 1368 VOP_UNLOCK(coveredvp, 0); 1369 vdrop(coveredvp); 1370 } 1371 vfs_rel(mp); 1372 return (error); 1373 } 1374 1375 vn_start_write(NULL, &mp, V_WAIT | V_MNTREF); 1376 MNT_ILOCK(mp); 1377 if ((mp->mnt_kern_flag & MNTK_UNMOUNT) != 0 || 1378 (mp->mnt_flag & MNT_UPDATE) != 0 || 1379 !TAILQ_EMPTY(&mp->mnt_uppers)) { 1380 dounmount_cleanup(mp, coveredvp, 0); 1381 return (EBUSY); 1382 } 1383 mp->mnt_kern_flag |= MNTK_UNMOUNT | MNTK_NOINSMNTQ; 1384 if (flags & MNT_NONBUSY) { 1385 MNT_IUNLOCK(mp); 1386 error = vfs_check_usecounts(mp); 1387 MNT_ILOCK(mp); 1388 if (error != 0) { 1389 dounmount_cleanup(mp, coveredvp, MNTK_UNMOUNT | 1390 MNTK_NOINSMNTQ); 1391 return (error); 1392 } 1393 } 1394 /* Allow filesystems to detect that a forced unmount is in progress. */ 1395 if (flags & MNT_FORCE) { 1396 mp->mnt_kern_flag |= MNTK_UNMOUNTF; 1397 MNT_IUNLOCK(mp); 1398 /* 1399 * Must be done after setting MNTK_UNMOUNTF and before 1400 * waiting for mnt_lockref to become 0. 1401 */ 1402 VFS_PURGE(mp); 1403 MNT_ILOCK(mp); 1404 } 1405 error = 0; 1406 if (mp->mnt_lockref) { 1407 mp->mnt_kern_flag |= MNTK_DRAINING; 1408 error = msleep(&mp->mnt_lockref, MNT_MTX(mp), PVFS, 1409 "mount drain", 0); 1410 } 1411 MNT_IUNLOCK(mp); 1412 KASSERT(mp->mnt_lockref == 0, 1413 ("%s: invalid lock refcount in the drain path @ %s:%d", 1414 __func__, __FILE__, __LINE__)); 1415 KASSERT(error == 0, 1416 ("%s: invalid return value for msleep in the drain path @ %s:%d", 1417 __func__, __FILE__, __LINE__)); 1418 1419 if (mp->mnt_flag & MNT_EXPUBLIC) 1420 vfs_setpublicfs(NULL, NULL, NULL); 1421 1422 /* 1423 * From now, we can claim that the use reference on the 1424 * coveredvp is ours, and the ref can be released only by 1425 * successfull unmount by us, or left for later unmount 1426 * attempt. The previously acquired hold reference is no 1427 * longer needed to protect the vnode from reuse. 1428 */ 1429 if (coveredvp != NULL) 1430 vdrop(coveredvp); 1431 1432 vfs_msync(mp, MNT_WAIT); 1433 MNT_ILOCK(mp); 1434 async_flag = mp->mnt_flag & MNT_ASYNC; 1435 mp->mnt_flag &= ~MNT_ASYNC; 1436 mp->mnt_kern_flag &= ~MNTK_ASYNC; 1437 MNT_IUNLOCK(mp); 1438 cache_purgevfs(mp, false); /* remove cache entries for this file sys */ 1439 vfs_deallocate_syncvnode(mp); 1440 if ((mp->mnt_flag & MNT_RDONLY) != 0 || (flags & MNT_FORCE) != 0 || 1441 (error = VFS_SYNC(mp, MNT_WAIT)) == 0) 1442 error = VFS_UNMOUNT(mp, flags); 1443 vn_finished_write(mp); 1444 /* 1445 * If we failed to flush the dirty blocks for this mount point, 1446 * undo all the cdir/rdir and rootvnode changes we made above. 1447 * Unless we failed to do so because the device is reporting that 1448 * it doesn't exist anymore. 1449 */ 1450 if (error && error != ENXIO) { 1451 MNT_ILOCK(mp); 1452 mp->mnt_kern_flag &= ~MNTK_NOINSMNTQ; 1453 if ((mp->mnt_flag & MNT_RDONLY) == 0) { 1454 MNT_IUNLOCK(mp); 1455 vfs_allocate_syncvnode(mp); 1456 MNT_ILOCK(mp); 1457 } 1458 mp->mnt_kern_flag &= ~(MNTK_UNMOUNT | MNTK_UNMOUNTF); 1459 mp->mnt_flag |= async_flag; 1460 if ((mp->mnt_flag & MNT_ASYNC) != 0 && 1461 (mp->mnt_kern_flag & MNTK_NOASYNC) == 0) 1462 mp->mnt_kern_flag |= MNTK_ASYNC; 1463 if (mp->mnt_kern_flag & MNTK_MWAIT) { 1464 mp->mnt_kern_flag &= ~MNTK_MWAIT; 1465 wakeup(mp); 1466 } 1467 MNT_IUNLOCK(mp); 1468 if (coveredvp) 1469 VOP_UNLOCK(coveredvp, 0); 1470 return (error); 1471 } 1472 mtx_lock(&mountlist_mtx); 1473 TAILQ_REMOVE(&mountlist, mp, mnt_list); 1474 mtx_unlock(&mountlist_mtx); 1475 EVENTHANDLER_DIRECT_INVOKE(vfs_unmounted, mp, td); 1476 if (coveredvp != NULL) { 1477 coveredvp->v_mountedhere = NULL; 1478 VOP_UNLOCK(coveredvp, 0); 1479 } 1480 vfs_event_signal(NULL, VQ_UNMOUNT, 0); 1481 if (rootvnode != NULL && mp == rootvnode->v_mount) { 1482 vrele(rootvnode); 1483 rootvnode = NULL; 1484 } 1485 if (mp == rootdevmp) 1486 rootdevmp = NULL; 1487 vfs_mount_destroy(mp); 1488 return (0); 1489 } 1490 1491 /* 1492 * Report errors during filesystem mounting. 1493 */ 1494 void 1495 vfs_mount_error(struct mount *mp, const char *fmt, ...) 1496 { 1497 struct vfsoptlist *moptlist = mp->mnt_optnew; 1498 va_list ap; 1499 int error, len; 1500 char *errmsg; 1501 1502 error = vfs_getopt(moptlist, "errmsg", (void **)&errmsg, &len); 1503 if (error || errmsg == NULL || len <= 0) 1504 return; 1505 1506 va_start(ap, fmt); 1507 vsnprintf(errmsg, (size_t)len, fmt, ap); 1508 va_end(ap); 1509 } 1510 1511 void 1512 vfs_opterror(struct vfsoptlist *opts, const char *fmt, ...) 1513 { 1514 va_list ap; 1515 int error, len; 1516 char *errmsg; 1517 1518 error = vfs_getopt(opts, "errmsg", (void **)&errmsg, &len); 1519 if (error || errmsg == NULL || len <= 0) 1520 return; 1521 1522 va_start(ap, fmt); 1523 vsnprintf(errmsg, (size_t)len, fmt, ap); 1524 va_end(ap); 1525 } 1526 1527 /* 1528 * --------------------------------------------------------------------- 1529 * Functions for querying mount options/arguments from filesystems. 1530 */ 1531 1532 /* 1533 * Check that no unknown options are given 1534 */ 1535 int 1536 vfs_filteropt(struct vfsoptlist *opts, const char **legal) 1537 { 1538 struct vfsopt *opt; 1539 char errmsg[255]; 1540 const char **t, *p, *q; 1541 int ret = 0; 1542 1543 TAILQ_FOREACH(opt, opts, link) { 1544 p = opt->name; 1545 q = NULL; 1546 if (p[0] == 'n' && p[1] == 'o') 1547 q = p + 2; 1548 for(t = global_opts; *t != NULL; t++) { 1549 if (strcmp(*t, p) == 0) 1550 break; 1551 if (q != NULL) { 1552 if (strcmp(*t, q) == 0) 1553 break; 1554 } 1555 } 1556 if (*t != NULL) 1557 continue; 1558 for(t = legal; *t != NULL; t++) { 1559 if (strcmp(*t, p) == 0) 1560 break; 1561 if (q != NULL) { 1562 if (strcmp(*t, q) == 0) 1563 break; 1564 } 1565 } 1566 if (*t != NULL) 1567 continue; 1568 snprintf(errmsg, sizeof(errmsg), 1569 "mount option <%s> is unknown", p); 1570 ret = EINVAL; 1571 } 1572 if (ret != 0) { 1573 TAILQ_FOREACH(opt, opts, link) { 1574 if (strcmp(opt->name, "errmsg") == 0) { 1575 strncpy((char *)opt->value, errmsg, opt->len); 1576 break; 1577 } 1578 } 1579 if (opt == NULL) 1580 printf("%s\n", errmsg); 1581 } 1582 return (ret); 1583 } 1584 1585 /* 1586 * Get a mount option by its name. 1587 * 1588 * Return 0 if the option was found, ENOENT otherwise. 1589 * If len is non-NULL it will be filled with the length 1590 * of the option. If buf is non-NULL, it will be filled 1591 * with the address of the option. 1592 */ 1593 int 1594 vfs_getopt(struct vfsoptlist *opts, const char *name, void **buf, int *len) 1595 { 1596 struct vfsopt *opt; 1597 1598 KASSERT(opts != NULL, ("vfs_getopt: caller passed 'opts' as NULL")); 1599 1600 TAILQ_FOREACH(opt, opts, link) { 1601 if (strcmp(name, opt->name) == 0) { 1602 opt->seen = 1; 1603 if (len != NULL) 1604 *len = opt->len; 1605 if (buf != NULL) 1606 *buf = opt->value; 1607 return (0); 1608 } 1609 } 1610 return (ENOENT); 1611 } 1612 1613 int 1614 vfs_getopt_pos(struct vfsoptlist *opts, const char *name) 1615 { 1616 struct vfsopt *opt; 1617 1618 if (opts == NULL) 1619 return (-1); 1620 1621 TAILQ_FOREACH(opt, opts, link) { 1622 if (strcmp(name, opt->name) == 0) { 1623 opt->seen = 1; 1624 return (opt->pos); 1625 } 1626 } 1627 return (-1); 1628 } 1629 1630 int 1631 vfs_getopt_size(struct vfsoptlist *opts, const char *name, off_t *value) 1632 { 1633 char *opt_value, *vtp; 1634 quad_t iv; 1635 int error, opt_len; 1636 1637 error = vfs_getopt(opts, name, (void **)&opt_value, &opt_len); 1638 if (error != 0) 1639 return (error); 1640 if (opt_len == 0 || opt_value == NULL) 1641 return (EINVAL); 1642 if (opt_value[0] == '\0' || opt_value[opt_len - 1] != '\0') 1643 return (EINVAL); 1644 iv = strtoq(opt_value, &vtp, 0); 1645 if (vtp == opt_value || (vtp[0] != '\0' && vtp[1] != '\0')) 1646 return (EINVAL); 1647 if (iv < 0) 1648 return (EINVAL); 1649 switch (vtp[0]) { 1650 case 't': case 'T': 1651 iv *= 1024; 1652 /* FALLTHROUGH */ 1653 case 'g': case 'G': 1654 iv *= 1024; 1655 /* FALLTHROUGH */ 1656 case 'm': case 'M': 1657 iv *= 1024; 1658 /* FALLTHROUGH */ 1659 case 'k': case 'K': 1660 iv *= 1024; 1661 case '\0': 1662 break; 1663 default: 1664 return (EINVAL); 1665 } 1666 *value = iv; 1667 1668 return (0); 1669 } 1670 1671 char * 1672 vfs_getopts(struct vfsoptlist *opts, const char *name, int *error) 1673 { 1674 struct vfsopt *opt; 1675 1676 *error = 0; 1677 TAILQ_FOREACH(opt, opts, link) { 1678 if (strcmp(name, opt->name) != 0) 1679 continue; 1680 opt->seen = 1; 1681 if (opt->len == 0 || 1682 ((char *)opt->value)[opt->len - 1] != '\0') { 1683 *error = EINVAL; 1684 return (NULL); 1685 } 1686 return (opt->value); 1687 } 1688 *error = ENOENT; 1689 return (NULL); 1690 } 1691 1692 int 1693 vfs_flagopt(struct vfsoptlist *opts, const char *name, uint64_t *w, 1694 uint64_t val) 1695 { 1696 struct vfsopt *opt; 1697 1698 TAILQ_FOREACH(opt, opts, link) { 1699 if (strcmp(name, opt->name) == 0) { 1700 opt->seen = 1; 1701 if (w != NULL) 1702 *w |= val; 1703 return (1); 1704 } 1705 } 1706 if (w != NULL) 1707 *w &= ~val; 1708 return (0); 1709 } 1710 1711 int 1712 vfs_scanopt(struct vfsoptlist *opts, const char *name, const char *fmt, ...) 1713 { 1714 va_list ap; 1715 struct vfsopt *opt; 1716 int ret; 1717 1718 KASSERT(opts != NULL, ("vfs_getopt: caller passed 'opts' as NULL")); 1719 1720 TAILQ_FOREACH(opt, opts, link) { 1721 if (strcmp(name, opt->name) != 0) 1722 continue; 1723 opt->seen = 1; 1724 if (opt->len == 0 || opt->value == NULL) 1725 return (0); 1726 if (((char *)opt->value)[opt->len - 1] != '\0') 1727 return (0); 1728 va_start(ap, fmt); 1729 ret = vsscanf(opt->value, fmt, ap); 1730 va_end(ap); 1731 return (ret); 1732 } 1733 return (0); 1734 } 1735 1736 int 1737 vfs_setopt(struct vfsoptlist *opts, const char *name, void *value, int len) 1738 { 1739 struct vfsopt *opt; 1740 1741 TAILQ_FOREACH(opt, opts, link) { 1742 if (strcmp(name, opt->name) != 0) 1743 continue; 1744 opt->seen = 1; 1745 if (opt->value == NULL) 1746 opt->len = len; 1747 else { 1748 if (opt->len != len) 1749 return (EINVAL); 1750 bcopy(value, opt->value, len); 1751 } 1752 return (0); 1753 } 1754 return (ENOENT); 1755 } 1756 1757 int 1758 vfs_setopt_part(struct vfsoptlist *opts, const char *name, void *value, int len) 1759 { 1760 struct vfsopt *opt; 1761 1762 TAILQ_FOREACH(opt, opts, link) { 1763 if (strcmp(name, opt->name) != 0) 1764 continue; 1765 opt->seen = 1; 1766 if (opt->value == NULL) 1767 opt->len = len; 1768 else { 1769 if (opt->len < len) 1770 return (EINVAL); 1771 opt->len = len; 1772 bcopy(value, opt->value, len); 1773 } 1774 return (0); 1775 } 1776 return (ENOENT); 1777 } 1778 1779 int 1780 vfs_setopts(struct vfsoptlist *opts, const char *name, const char *value) 1781 { 1782 struct vfsopt *opt; 1783 1784 TAILQ_FOREACH(opt, opts, link) { 1785 if (strcmp(name, opt->name) != 0) 1786 continue; 1787 opt->seen = 1; 1788 if (opt->value == NULL) 1789 opt->len = strlen(value) + 1; 1790 else if (strlcpy(opt->value, value, opt->len) >= opt->len) 1791 return (EINVAL); 1792 return (0); 1793 } 1794 return (ENOENT); 1795 } 1796 1797 /* 1798 * Find and copy a mount option. 1799 * 1800 * The size of the buffer has to be specified 1801 * in len, if it is not the same length as the 1802 * mount option, EINVAL is returned. 1803 * Returns ENOENT if the option is not found. 1804 */ 1805 int 1806 vfs_copyopt(struct vfsoptlist *opts, const char *name, void *dest, int len) 1807 { 1808 struct vfsopt *opt; 1809 1810 KASSERT(opts != NULL, ("vfs_copyopt: caller passed 'opts' as NULL")); 1811 1812 TAILQ_FOREACH(opt, opts, link) { 1813 if (strcmp(name, opt->name) == 0) { 1814 opt->seen = 1; 1815 if (len != opt->len) 1816 return (EINVAL); 1817 bcopy(opt->value, dest, opt->len); 1818 return (0); 1819 } 1820 } 1821 return (ENOENT); 1822 } 1823 1824 int 1825 __vfs_statfs(struct mount *mp, struct statfs *sbp) 1826 { 1827 int error; 1828 1829 error = mp->mnt_op->vfs_statfs(mp, &mp->mnt_stat); 1830 if (sbp != &mp->mnt_stat) 1831 *sbp = mp->mnt_stat; 1832 return (error); 1833 } 1834 1835 void 1836 vfs_mountedfrom(struct mount *mp, const char *from) 1837 { 1838 1839 bzero(mp->mnt_stat.f_mntfromname, sizeof mp->mnt_stat.f_mntfromname); 1840 strlcpy(mp->mnt_stat.f_mntfromname, from, 1841 sizeof mp->mnt_stat.f_mntfromname); 1842 } 1843 1844 /* 1845 * --------------------------------------------------------------------- 1846 * This is the api for building mount args and mounting filesystems from 1847 * inside the kernel. 1848 * 1849 * The API works by accumulation of individual args. First error is 1850 * latched. 1851 * 1852 * XXX: should be documented in new manpage kernel_mount(9) 1853 */ 1854 1855 /* A memory allocation which must be freed when we are done */ 1856 struct mntaarg { 1857 SLIST_ENTRY(mntaarg) next; 1858 }; 1859 1860 /* The header for the mount arguments */ 1861 struct mntarg { 1862 struct iovec *v; 1863 int len; 1864 int error; 1865 SLIST_HEAD(, mntaarg) list; 1866 }; 1867 1868 /* 1869 * Add a boolean argument. 1870 * 1871 * flag is the boolean value. 1872 * name must start with "no". 1873 */ 1874 struct mntarg * 1875 mount_argb(struct mntarg *ma, int flag, const char *name) 1876 { 1877 1878 KASSERT(name[0] == 'n' && name[1] == 'o', 1879 ("mount_argb(...,%s): name must start with 'no'", name)); 1880 1881 return (mount_arg(ma, name + (flag ? 2 : 0), NULL, 0)); 1882 } 1883 1884 /* 1885 * Add an argument printf style 1886 */ 1887 struct mntarg * 1888 mount_argf(struct mntarg *ma, const char *name, const char *fmt, ...) 1889 { 1890 va_list ap; 1891 struct mntaarg *maa; 1892 struct sbuf *sb; 1893 int len; 1894 1895 if (ma == NULL) { 1896 ma = malloc(sizeof *ma, M_MOUNT, M_WAITOK | M_ZERO); 1897 SLIST_INIT(&ma->list); 1898 } 1899 if (ma->error) 1900 return (ma); 1901 1902 ma->v = realloc(ma->v, sizeof *ma->v * (ma->len + 2), 1903 M_MOUNT, M_WAITOK); 1904 ma->v[ma->len].iov_base = (void *)(uintptr_t)name; 1905 ma->v[ma->len].iov_len = strlen(name) + 1; 1906 ma->len++; 1907 1908 sb = sbuf_new_auto(); 1909 va_start(ap, fmt); 1910 sbuf_vprintf(sb, fmt, ap); 1911 va_end(ap); 1912 sbuf_finish(sb); 1913 len = sbuf_len(sb) + 1; 1914 maa = malloc(sizeof *maa + len, M_MOUNT, M_WAITOK | M_ZERO); 1915 SLIST_INSERT_HEAD(&ma->list, maa, next); 1916 bcopy(sbuf_data(sb), maa + 1, len); 1917 sbuf_delete(sb); 1918 1919 ma->v[ma->len].iov_base = maa + 1; 1920 ma->v[ma->len].iov_len = len; 1921 ma->len++; 1922 1923 return (ma); 1924 } 1925 1926 /* 1927 * Add an argument which is a userland string. 1928 */ 1929 struct mntarg * 1930 mount_argsu(struct mntarg *ma, const char *name, const void *val, int len) 1931 { 1932 struct mntaarg *maa; 1933 char *tbuf; 1934 1935 if (val == NULL) 1936 return (ma); 1937 if (ma == NULL) { 1938 ma = malloc(sizeof *ma, M_MOUNT, M_WAITOK | M_ZERO); 1939 SLIST_INIT(&ma->list); 1940 } 1941 if (ma->error) 1942 return (ma); 1943 maa = malloc(sizeof *maa + len, M_MOUNT, M_WAITOK | M_ZERO); 1944 SLIST_INSERT_HEAD(&ma->list, maa, next); 1945 tbuf = (void *)(maa + 1); 1946 ma->error = copyinstr(val, tbuf, len, NULL); 1947 return (mount_arg(ma, name, tbuf, -1)); 1948 } 1949 1950 /* 1951 * Plain argument. 1952 * 1953 * If length is -1, treat value as a C string. 1954 */ 1955 struct mntarg * 1956 mount_arg(struct mntarg *ma, const char *name, const void *val, int len) 1957 { 1958 1959 if (ma == NULL) { 1960 ma = malloc(sizeof *ma, M_MOUNT, M_WAITOK | M_ZERO); 1961 SLIST_INIT(&ma->list); 1962 } 1963 if (ma->error) 1964 return (ma); 1965 1966 ma->v = realloc(ma->v, sizeof *ma->v * (ma->len + 2), 1967 M_MOUNT, M_WAITOK); 1968 ma->v[ma->len].iov_base = (void *)(uintptr_t)name; 1969 ma->v[ma->len].iov_len = strlen(name) + 1; 1970 ma->len++; 1971 1972 ma->v[ma->len].iov_base = (void *)(uintptr_t)val; 1973 if (len < 0) 1974 ma->v[ma->len].iov_len = strlen(val) + 1; 1975 else 1976 ma->v[ma->len].iov_len = len; 1977 ma->len++; 1978 return (ma); 1979 } 1980 1981 /* 1982 * Free a mntarg structure 1983 */ 1984 static void 1985 free_mntarg(struct mntarg *ma) 1986 { 1987 struct mntaarg *maa; 1988 1989 while (!SLIST_EMPTY(&ma->list)) { 1990 maa = SLIST_FIRST(&ma->list); 1991 SLIST_REMOVE_HEAD(&ma->list, next); 1992 free(maa, M_MOUNT); 1993 } 1994 free(ma->v, M_MOUNT); 1995 free(ma, M_MOUNT); 1996 } 1997 1998 /* 1999 * Mount a filesystem 2000 */ 2001 int 2002 kernel_mount(struct mntarg *ma, uint64_t flags) 2003 { 2004 struct uio auio; 2005 int error; 2006 2007 KASSERT(ma != NULL, ("kernel_mount NULL ma")); 2008 KASSERT(ma->v != NULL, ("kernel_mount NULL ma->v")); 2009 KASSERT(!(ma->len & 1), ("kernel_mount odd ma->len (%d)", ma->len)); 2010 2011 auio.uio_iov = ma->v; 2012 auio.uio_iovcnt = ma->len; 2013 auio.uio_segflg = UIO_SYSSPACE; 2014 2015 error = ma->error; 2016 if (!error) 2017 error = vfs_donmount(curthread, flags, &auio); 2018 free_mntarg(ma); 2019 return (error); 2020 } 2021 2022 /* 2023 * A printflike function to mount a filesystem. 2024 */ 2025 int 2026 kernel_vmount(int flags, ...) 2027 { 2028 struct mntarg *ma = NULL; 2029 va_list ap; 2030 const char *cp; 2031 const void *vp; 2032 int error; 2033 2034 va_start(ap, flags); 2035 for (;;) { 2036 cp = va_arg(ap, const char *); 2037 if (cp == NULL) 2038 break; 2039 vp = va_arg(ap, const void *); 2040 ma = mount_arg(ma, cp, vp, (vp != NULL ? -1 : 0)); 2041 } 2042 va_end(ap); 2043 2044 error = kernel_mount(ma, flags); 2045 return (error); 2046 } 2047 2048 void 2049 vfs_oexport_conv(const struct oexport_args *oexp, struct export_args *exp) 2050 { 2051 2052 bcopy(oexp, exp, sizeof(*oexp)); 2053 exp->ex_numsecflavors = 0; 2054 } 2055