1 /* 2 * Copyright (c) 1989, 1993 3 * The Regents of the University of California. All rights reserved. 4 * (c) UNIX System Laboratories, Inc. 5 * All or some portions of this file are derived from material licensed 6 * to the University of California by American Telephone and Telegraph 7 * Co. or Unix System Laboratories, Inc. and are reproduced herein with 8 * the permission of UNIX System Laboratories, Inc. 9 * 10 * Redistribution and use in source and binary forms, with or without 11 * modification, are permitted provided that the following conditions 12 * are met: 13 * 1. Redistributions of source code must retain the above copyright 14 * notice, this list of conditions and the following disclaimer. 15 * 2. Redistributions in binary form must reproduce the above copyright 16 * notice, this list of conditions and the following disclaimer in the 17 * documentation and/or other materials provided with the distribution. 18 * 3. All advertising materials mentioning features or use of this software 19 * must display the following acknowledgement: 20 * This product includes software developed by the University of 21 * California, Berkeley and its contributors. 22 * 4. 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 REGENTS 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 REGENTS 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 * Copyright (c) 1999 Michael Smith 39 * All rights reserved. 40 * Copyright (c) 1999 Poul-Henning Kamp 41 * All rights reserved. 42 * 43 * Redistribution and use in source and binary forms, with or without 44 * modification, are permitted provided that the following conditions 45 * are met: 46 * 1. Redistributions of source code must retain the above copyright 47 * notice, this list of conditions and the following disclaimer. 48 * 2. Redistributions in binary form must reproduce the above copyright 49 * notice, this list of conditions and the following disclaimer in the 50 * documentation and/or other materials provided with the distribution. 51 * 52 * THIS SOFTWARE IS PROVIDED BY THE AUTHOR AND CONTRIBUTORS ``AS IS'' AND 53 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE 54 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE 55 * ARE DISCLAIMED. IN NO EVENT SHALL THE AUTHOR OR CONTRIBUTORS BE LIABLE 56 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL 57 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS 58 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) 59 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT 60 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY 61 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF 62 * SUCH DAMAGE. 63 * 64 * $FreeBSD$ 65 */ 66 67 #include <sys/param.h> 68 #include <sys/conf.h> 69 #include <sys/cons.h> 70 #include <sys/kernel.h> 71 #include <sys/linker.h> 72 #include <sys/mac.h> 73 #include <sys/malloc.h> 74 #include <sys/mount.h> 75 #include <sys/mutex.h> 76 #include <sys/namei.h> 77 #include <sys/proc.h> 78 #include <sys/filedesc.h> 79 #include <sys/reboot.h> 80 #include <sys/sysproto.h> 81 #include <sys/sx.h> 82 #include <sys/sysctl.h> 83 #include <sys/sysent.h> 84 #include <sys/systm.h> 85 #include <sys/vnode.h> 86 87 #include <geom/geom.h> 88 89 #include <machine/stdarg.h> 90 91 #include "opt_rootdevname.h" 92 #include "opt_ddb.h" 93 #include "opt_mac.h" 94 95 #ifdef DDB 96 #include <ddb/ddb.h> 97 #endif 98 99 #define ROOTNAME "root_device" 100 101 static void checkdirs(struct vnode *olddp, struct vnode *newdp); 102 static int vfs_nmount(struct thread *td, int, struct uio *); 103 static int vfs_mountroot_try(char *mountfrom); 104 static int vfs_mountroot_ask(void); 105 static void gets(char *cp); 106 107 static int usermount = 0; /* if 1, non-root can mount fs. */ 108 SYSCTL_INT(_vfs, OID_AUTO, usermount, CTLFLAG_RW, &usermount, 0, ""); 109 110 MALLOC_DEFINE(M_MOUNT, "mount", "vfs mount structure"); 111 112 /* List of mounted filesystems. */ 113 struct mntlist mountlist = TAILQ_HEAD_INITIALIZER(mountlist); 114 115 /* For any iteration/modification of mountlist */ 116 struct mtx mountlist_mtx; 117 118 /* For any iteration/modification of mnt_vnodelist */ 119 struct mtx mntvnode_mtx; 120 121 /* 122 * The vnode of the system's root (/ in the filesystem, without chroot 123 * active.) 124 */ 125 struct vnode *rootvnode; 126 127 /* 128 * The root filesystem is detailed in the kernel environment variable 129 * vfs.root.mountfrom, which is expected to be in the general format 130 * 131 * <vfsname>:[<path>] 132 * vfsname := the name of a VFS known to the kernel and capable 133 * of being mounted as root 134 * path := disk device name or other data used by the filesystem 135 * to locate its physical store 136 */ 137 138 /* 139 * The root specifiers we will try if RB_CDROM is specified. 140 */ 141 static char *cdrom_rootdevnames[] = { 142 "cd9660:cd0a", 143 "cd9660:acd0a", 144 "cd9660:wcd0a", 145 NULL 146 }; 147 148 /* legacy find-root code */ 149 char *rootdevnames[2] = {NULL, NULL}; 150 static int setrootbyname(char *name); 151 dev_t rootdev = NODEV; 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 #ifdef INVARIANTS 163 else if (opt->len != 0) 164 panic("%s: mount option with NULL value but length != 0", 165 __func__); 166 #endif 167 free(opt, M_MOUNT); 168 } 169 170 /* Release all resources related to the mount options. */ 171 static void 172 vfs_freeopts(struct vfsoptlist *opts) 173 { 174 struct vfsopt *opt; 175 176 while (!TAILQ_EMPTY(opts)) { 177 opt = TAILQ_FIRST(opts); 178 vfs_freeopt(opts, opt); 179 } 180 free(opts, M_MOUNT); 181 } 182 183 /* 184 * If a mount option is specified several times, 185 * (with or without the "no" prefix) only keep 186 * the last occurence of it. 187 */ 188 static void 189 vfs_sanitizeopts(struct vfsoptlist *opts) 190 { 191 struct vfsopt *opt, *opt2, *tmp; 192 int noopt; 193 194 TAILQ_FOREACH_REVERSE(opt, opts, vfsoptlist, link) { 195 if (strncmp(opt->name, "no", 2) == 0) 196 noopt = 1; 197 else 198 noopt = 0; 199 opt2 = TAILQ_PREV(opt, vfsoptlist, link); 200 while (opt2 != NULL) { 201 if (strcmp(opt2->name, opt->name) == 0 || 202 (noopt && strcmp(opt->name + 2, opt2->name) == 0) || 203 (!noopt && strncmp(opt2->name, "no", 2) == 0 && 204 strcmp(opt2->name + 2, opt->name) == 0)) { 205 tmp = TAILQ_PREV(opt2, vfsoptlist, link); 206 vfs_freeopt(opts, opt2); 207 opt2 = tmp; 208 } else { 209 opt2 = TAILQ_PREV(opt2, vfsoptlist, link); 210 } 211 } 212 } 213 } 214 215 /* 216 * Build a linked list of mount options from a struct uio. 217 */ 218 static int 219 vfs_buildopts(struct uio *auio, struct vfsoptlist **options) 220 { 221 struct vfsoptlist *opts; 222 struct vfsopt *opt; 223 unsigned int i, iovcnt; 224 int error, namelen, optlen; 225 226 iovcnt = auio->uio_iovcnt; 227 opts = malloc(sizeof(struct vfsoptlist), M_MOUNT, 0); 228 TAILQ_INIT(opts); 229 for (i = 0; i < iovcnt; i += 2) { 230 opt = malloc(sizeof(struct vfsopt), M_MOUNT, 0); 231 namelen = auio->uio_iov[i].iov_len; 232 optlen = auio->uio_iov[i + 1].iov_len; 233 opt->name = malloc(namelen, M_MOUNT, 0); 234 opt->value = NULL; 235 if (auio->uio_segflg == UIO_SYSSPACE) { 236 bcopy(auio->uio_iov[i].iov_base, opt->name, namelen); 237 } else { 238 error = copyin(auio->uio_iov[i].iov_base, opt->name, 239 namelen); 240 if (error) 241 goto bad; 242 } 243 opt->len = optlen; 244 if (optlen != 0) { 245 opt->value = malloc(optlen, M_MOUNT, 0); 246 if (auio->uio_segflg == UIO_SYSSPACE) { 247 bcopy(auio->uio_iov[i + 1].iov_base, opt->value, 248 optlen); 249 } else { 250 error = copyin(auio->uio_iov[i + 1].iov_base, 251 opt->value, optlen); 252 if (error) 253 goto bad; 254 } 255 } 256 TAILQ_INSERT_TAIL(opts, opt, link); 257 } 258 vfs_sanitizeopts(opts); 259 *options = opts; 260 return (0); 261 bad: 262 vfs_freeopts(opts); 263 return (error); 264 } 265 266 /* 267 * Merge the old mount options with the new ones passed 268 * in the MNT_UPDATE case. 269 */ 270 static void 271 vfs_mergeopts(struct vfsoptlist *toopts, struct vfsoptlist *opts) 272 { 273 struct vfsopt *opt, *opt2, *new; 274 275 TAILQ_FOREACH(opt, opts, link) { 276 /* 277 * Check that this option hasn't been redefined 278 * nor cancelled with a "no" mount option. 279 */ 280 opt2 = TAILQ_FIRST(toopts); 281 while (opt2 != NULL) { 282 if (strcmp(opt2->name, opt->name) == 0) 283 goto next; 284 if (strncmp(opt2->name, "no", 2) == 0 && 285 strcmp(opt2->name + 2, opt->name) == 0) { 286 vfs_freeopt(toopts, opt2); 287 goto next; 288 } 289 opt2 = TAILQ_NEXT(opt2, link); 290 } 291 /* We want this option, duplicate it. */ 292 new = malloc(sizeof(struct vfsopt), M_MOUNT, 0); 293 new->name = malloc(strlen(opt->name) + 1, M_MOUNT, 0); 294 strcpy(new->name, opt->name); 295 if (opt->len != 0) { 296 new->value = malloc(opt->len, M_MOUNT, 0); 297 bcopy(opt->value, new->value, opt->len); 298 } else { 299 new->value = NULL; 300 } 301 new->len = opt->len; 302 TAILQ_INSERT_TAIL(toopts, new, link); 303 next: 304 continue; 305 } 306 } 307 308 /* 309 * New mount API. 310 */ 311 int 312 nmount(td, uap) 313 struct thread *td; 314 struct nmount_args /* { 315 struct iovec *iovp; 316 unsigned int iovcnt; 317 int flags; 318 } */ *uap; 319 { 320 struct uio auio; 321 struct iovec *iov, *needfree; 322 struct iovec aiov[UIO_SMALLIOV]; 323 unsigned int i; 324 int error; 325 u_int iovlen, iovcnt; 326 327 iovcnt = uap->iovcnt; 328 iovlen = iovcnt * sizeof (struct iovec); 329 /* 330 * Check that we have an even number of iovec's 331 * and that we have at least two options. 332 */ 333 if ((iovcnt & 1) || (iovcnt < 4) || (iovcnt > UIO_MAXIOV)) 334 return (EINVAL); 335 336 if (iovcnt > UIO_SMALLIOV) { 337 MALLOC(iov, struct iovec *, iovlen, M_IOV, 0); 338 needfree = iov; 339 } else { 340 iov = aiov; 341 needfree = NULL; 342 } 343 auio.uio_iov = iov; 344 auio.uio_iovcnt = iovcnt; 345 auio.uio_segflg = UIO_USERSPACE; 346 if ((error = copyin(uap->iovp, iov, iovlen))) 347 goto finish; 348 349 for (i = 0; i < iovcnt; i++) { 350 if (iov->iov_len > MMAXOPTIONLEN) { 351 error = EINVAL; 352 goto finish; 353 } 354 iov++; 355 } 356 error = vfs_nmount(td, uap->flags, &auio); 357 finish: 358 if (needfree != NULL) 359 free(needfree, M_TEMP); 360 return (error); 361 } 362 363 int 364 kernel_mount(iovp, iovcnt, flags) 365 struct iovec *iovp; 366 unsigned int iovcnt; 367 int flags; 368 { 369 struct uio auio; 370 int error; 371 372 /* 373 * Check that we have an even number of iovec's 374 * and that we have at least two options. 375 */ 376 if ((iovcnt & 1) || (iovcnt < 4)) 377 return (EINVAL); 378 379 auio.uio_iov = iovp; 380 auio.uio_iovcnt = iovcnt; 381 auio.uio_segflg = UIO_SYSSPACE; 382 383 error = vfs_nmount(curthread, flags, &auio); 384 return (error); 385 } 386 387 int 388 kernel_vmount(int flags, ...) 389 { 390 struct iovec *iovp; 391 struct uio auio; 392 va_list ap; 393 unsigned int iovcnt, iovlen, len; 394 const char *cp; 395 char *buf, *pos; 396 size_t n; 397 int error, i; 398 399 len = 0; 400 va_start(ap, flags); 401 for (iovcnt = 0; (cp = va_arg(ap, const char *)) != NULL; iovcnt++) 402 len += strlen(cp) + 1; 403 va_end(ap); 404 405 if (iovcnt < 4 || iovcnt & 1) 406 return (EINVAL); 407 408 iovlen = iovcnt * sizeof (struct iovec); 409 MALLOC(iovp, struct iovec *, iovlen, M_MOUNT, 0); 410 MALLOC(buf, char *, len, M_MOUNT, 0); 411 pos = buf; 412 va_start(ap, flags); 413 for (i = 0; i < iovcnt; i++) { 414 cp = va_arg(ap, const char *); 415 copystr(cp, pos, len - (pos - buf), &n); 416 iovp[i].iov_base = pos; 417 iovp[i].iov_len = n; 418 pos += n; 419 } 420 va_end(ap); 421 422 auio.uio_iov = iovp; 423 auio.uio_iovcnt = iovcnt; 424 auio.uio_segflg = UIO_SYSSPACE; 425 426 error = vfs_nmount(curthread, flags, &auio); 427 FREE(iovp, M_MOUNT); 428 FREE(buf, M_MOUNT); 429 return (error); 430 } 431 432 /* 433 * vfs_nmount(): actually attempt a filesystem mount. 434 */ 435 static int 436 vfs_nmount(td, fsflags, fsoptions) 437 struct thread *td; 438 int fsflags; /* Flags common to all filesystems. */ 439 struct uio *fsoptions; /* Options local to the filesystem. */ 440 { 441 linker_file_t lf; 442 struct vnode *vp; 443 struct mount *mp; 444 struct vfsconf *vfsp; 445 struct vfsoptlist *optlist; 446 char *fstype, *fspath; 447 int error, flag = 0, kern_flag = 0; 448 int fstypelen, fspathlen; 449 struct vattr va; 450 struct nameidata nd; 451 452 error = vfs_buildopts(fsoptions, &optlist); 453 if (error) 454 return (error); 455 456 /* 457 * We need these two options before the others, 458 * and they are mandatory for any filesystem. 459 * Ensure they are NUL terminated as well. 460 */ 461 fstypelen = 0; 462 error = vfs_getopt(optlist, "fstype", (void **)&fstype, &fstypelen); 463 if (error || fstype[fstypelen - 1] != '\0') { 464 error = EINVAL; 465 goto bad; 466 } 467 fspathlen = 0; 468 error = vfs_getopt(optlist, "fspath", (void **)&fspath, &fspathlen); 469 if (error || fspath[fspathlen - 1] != '\0') { 470 error = EINVAL; 471 goto bad; 472 } 473 474 /* 475 * Be ultra-paranoid about making sure the type and fspath 476 * variables will fit in our mp buffers, including the 477 * terminating NUL. 478 */ 479 if (fstypelen >= MFSNAMELEN - 1 || fspathlen >= MNAMELEN - 1) { 480 error = ENAMETOOLONG; 481 goto bad; 482 } 483 484 if (usermount == 0) { 485 error = suser(td); 486 if (error) 487 goto bad; 488 } 489 /* 490 * Do not allow NFS export by non-root users. 491 */ 492 if (fsflags & MNT_EXPORTED) { 493 error = suser(td); 494 if (error) 495 goto bad; 496 } 497 /* 498 * Silently enforce MNT_NOSUID and MNT_NODEV for non-root users. 499 */ 500 if (suser(td)) 501 fsflags |= MNT_NOSUID | MNT_NODEV; 502 /* 503 * Get vnode to be covered 504 */ 505 NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF, UIO_SYSSPACE, fspath, td); 506 if ((error = namei(&nd)) != 0) 507 goto bad; 508 NDFREE(&nd, NDF_ONLY_PNBUF); 509 vp = nd.ni_vp; 510 if (fsflags & MNT_UPDATE) { 511 if ((vp->v_vflag & VV_ROOT) == 0) { 512 vput(vp); 513 error = EINVAL; 514 goto bad; 515 } 516 mp = vp->v_mount; 517 flag = mp->mnt_flag; 518 kern_flag = mp->mnt_kern_flag; 519 /* 520 * We only allow the filesystem to be reloaded if it 521 * is currently mounted read-only. 522 */ 523 if ((fsflags & MNT_RELOAD) && 524 ((mp->mnt_flag & MNT_RDONLY) == 0)) { 525 vput(vp); 526 error = EOPNOTSUPP; /* Needs translation */ 527 goto bad; 528 } 529 /* 530 * Only root, or the user that did the original mount is 531 * permitted to update it. 532 */ 533 if (mp->mnt_cred->cr_uid != td->td_ucred->cr_uid) { 534 error = suser(td); 535 if (error) { 536 vput(vp); 537 goto bad; 538 } 539 } 540 if (vfs_busy(mp, LK_NOWAIT, 0, td)) { 541 vput(vp); 542 error = EBUSY; 543 goto bad; 544 } 545 VI_LOCK(vp); 546 if ((vp->v_iflag & VI_MOUNT) != 0 || 547 vp->v_mountedhere != NULL) { 548 VI_UNLOCK(vp); 549 vfs_unbusy(mp, td); 550 vput(vp); 551 error = EBUSY; 552 goto bad; 553 } 554 vp->v_iflag |= VI_MOUNT; 555 VI_UNLOCK(vp); 556 mp->mnt_flag |= fsflags & 557 (MNT_RELOAD | MNT_FORCE | MNT_UPDATE | MNT_SNAPSHOT); 558 VOP_UNLOCK(vp, 0, td); 559 mp->mnt_optnew = optlist; 560 vfs_mergeopts(mp->mnt_optnew, mp->mnt_opt); 561 goto update; 562 } 563 /* 564 * If the user is not root, ensure that they own the directory 565 * onto which we are attempting to mount. 566 */ 567 error = VOP_GETATTR(vp, &va, td->td_ucred, td); 568 if (error) { 569 vput(vp); 570 goto bad; 571 } 572 if (va.va_uid != td->td_ucred->cr_uid) { 573 error = suser(td); 574 if (error) { 575 vput(vp); 576 goto bad; 577 } 578 } 579 if ((error = vinvalbuf(vp, V_SAVE, td->td_ucred, td, 0, 0)) != 0) { 580 vput(vp); 581 goto bad; 582 } 583 if (vp->v_type != VDIR) { 584 vput(vp); 585 error = ENOTDIR; 586 goto bad; 587 } 588 for (vfsp = vfsconf; vfsp; vfsp = vfsp->vfc_next) 589 if (!strcmp(vfsp->vfc_name, fstype)) 590 break; 591 if (vfsp == NULL) { 592 /* Only load modules for root (very important!). */ 593 error = suser(td); 594 if (error) { 595 vput(vp); 596 goto bad; 597 } 598 error = securelevel_gt(td->td_ucred, 0); 599 if (error) { 600 vput(vp); 601 goto bad; 602 } 603 error = linker_load_module(NULL, fstype, NULL, NULL, &lf); 604 if (error || lf == NULL) { 605 vput(vp); 606 if (lf == NULL) 607 error = ENODEV; 608 goto bad; 609 } 610 lf->userrefs++; 611 /* Look up again to see if the VFS was loaded. */ 612 for (vfsp = vfsconf; vfsp; vfsp = vfsp->vfc_next) 613 if (!strcmp(vfsp->vfc_name, fstype)) 614 break; 615 if (vfsp == NULL) { 616 lf->userrefs--; 617 linker_file_unload(lf); 618 vput(vp); 619 error = ENODEV; 620 goto bad; 621 } 622 } 623 VI_LOCK(vp); 624 if ((vp->v_iflag & VI_MOUNT) != 0 || 625 vp->v_mountedhere != NULL) { 626 VI_UNLOCK(vp); 627 vput(vp); 628 error = EBUSY; 629 goto bad; 630 } 631 vp->v_iflag |= VI_MOUNT; 632 VI_UNLOCK(vp); 633 634 /* 635 * Allocate and initialize the filesystem. 636 */ 637 mp = malloc(sizeof(struct mount), M_MOUNT, M_ZERO); 638 TAILQ_INIT(&mp->mnt_nvnodelist); 639 TAILQ_INIT(&mp->mnt_reservedvnlist); 640 lockinit(&mp->mnt_lock, PVFS, "vfslock", 0, LK_NOPAUSE); 641 (void)vfs_busy(mp, LK_NOWAIT, 0, td); 642 mp->mnt_op = vfsp->vfc_vfsops; 643 mp->mnt_vfc = vfsp; 644 vfsp->vfc_refcount++; 645 mp->mnt_stat.f_type = vfsp->vfc_typenum; 646 mp->mnt_flag |= vfsp->vfc_flags & MNT_VISFLAGMASK; 647 strlcpy(mp->mnt_stat.f_fstypename, fstype, MFSNAMELEN); 648 mp->mnt_vnodecovered = vp; 649 mp->mnt_cred = crdup(td->td_ucred); 650 mp->mnt_stat.f_owner = td->td_ucred->cr_uid; 651 strlcpy(mp->mnt_stat.f_mntonname, fspath, MNAMELEN); 652 mp->mnt_iosize_max = DFLTPHYS; 653 #ifdef MAC 654 mac_init_mount(mp); 655 mac_create_mount(td->td_ucred, mp); 656 #endif 657 VOP_UNLOCK(vp, 0, td); 658 mp->mnt_optnew = optlist; /* XXXMAC: should this be above? */ 659 660 update: 661 /* 662 * Check if the fs implements the new VFS_NMOUNT() 663 * function, since the new system call was used. 664 */ 665 if (mp->mnt_op->vfs_mount != NULL) { 666 printf("%s doesn't support the new mount syscall\n", 667 mp->mnt_vfc->vfc_name); 668 VI_LOCK(vp); 669 vp->v_iflag &= ~VI_MOUNT; 670 VI_UNLOCK(vp); 671 if (mp->mnt_flag & MNT_UPDATE) 672 vfs_unbusy(mp, td); 673 else { 674 mp->mnt_vfc->vfc_refcount--; 675 vfs_unbusy(mp, td); 676 #ifdef MAC 677 mac_destroy_mount(mp); 678 #endif 679 free(mp, M_MOUNT); 680 } 681 vrele(vp); 682 error = EOPNOTSUPP; 683 goto bad; 684 } 685 686 /* 687 * Set the mount level flags. 688 */ 689 if (fsflags & MNT_RDONLY) 690 mp->mnt_flag |= MNT_RDONLY; 691 else if (mp->mnt_flag & MNT_RDONLY) 692 mp->mnt_kern_flag |= MNTK_WANTRDWR; 693 mp->mnt_flag &=~ MNT_UPDATEMASK; 694 mp->mnt_flag |= fsflags & (MNT_UPDATEMASK | MNT_FORCE); 695 /* 696 * Mount the filesystem. 697 * XXX The final recipients of VFS_MOUNT just overwrite the ndp they 698 * get. No freeing of cn_pnbuf. 699 */ 700 error = VFS_NMOUNT(mp, &nd, td); 701 if (!error) { 702 if (mp->mnt_opt != NULL) 703 vfs_freeopts(mp->mnt_opt); 704 mp->mnt_opt = mp->mnt_optnew; 705 } 706 /* 707 * Prevent external consumers of mount 708 * options to read mnt_optnew. 709 */ 710 mp->mnt_optnew = NULL; 711 if (mp->mnt_flag & MNT_UPDATE) { 712 if (mp->mnt_kern_flag & MNTK_WANTRDWR) 713 mp->mnt_flag &= ~MNT_RDONLY; 714 mp->mnt_flag &=~ 715 (MNT_UPDATE | MNT_RELOAD | MNT_FORCE | MNT_SNAPSHOT); 716 mp->mnt_kern_flag &=~ MNTK_WANTRDWR; 717 if (error) { 718 mp->mnt_flag = flag; 719 mp->mnt_kern_flag = kern_flag; 720 } 721 if ((mp->mnt_flag & MNT_RDONLY) == 0) { 722 if (mp->mnt_syncer == NULL) 723 error = vfs_allocate_syncvnode(mp); 724 } else { 725 if (mp->mnt_syncer != NULL) 726 vrele(mp->mnt_syncer); 727 mp->mnt_syncer = NULL; 728 } 729 vfs_unbusy(mp, td); 730 VI_LOCK(vp); 731 vp->v_iflag &= ~VI_MOUNT; 732 VI_UNLOCK(vp); 733 vrele(vp); 734 return (error); 735 } 736 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY, td); 737 /* 738 * Put the new filesystem on the mount list after root. 739 */ 740 cache_purge(vp); 741 if (!error) { 742 struct vnode *newdp; 743 744 VI_LOCK(vp); 745 vp->v_iflag &= ~VI_MOUNT; 746 VI_UNLOCK(vp); 747 vp->v_mountedhere = mp; 748 mtx_lock(&mountlist_mtx); 749 TAILQ_INSERT_TAIL(&mountlist, mp, mnt_list); 750 mtx_unlock(&mountlist_mtx); 751 if (VFS_ROOT(mp, &newdp)) 752 panic("mount: lost mount"); 753 checkdirs(vp, newdp); 754 vput(newdp); 755 VOP_UNLOCK(vp, 0, td); 756 if ((mp->mnt_flag & MNT_RDONLY) == 0) 757 error = vfs_allocate_syncvnode(mp); 758 vfs_unbusy(mp, td); 759 if ((error = VFS_START(mp, 0, td)) != 0) { 760 vrele(vp); 761 goto bad; 762 } 763 } else { 764 VI_LOCK(vp); 765 vp->v_iflag &= ~VI_MOUNT; 766 VI_UNLOCK(vp); 767 mp->mnt_vfc->vfc_refcount--; 768 vfs_unbusy(mp, td); 769 #ifdef MAC 770 mac_destroy_mount(mp); 771 #endif 772 free(mp, M_MOUNT); 773 vput(vp); 774 goto bad; 775 } 776 return (0); 777 bad: 778 vfs_freeopts(optlist); 779 return (error); 780 } 781 782 /* 783 * Old mount API. 784 */ 785 #ifndef _SYS_SYSPROTO_H_ 786 struct mount_args { 787 char *type; 788 char *path; 789 int flags; 790 caddr_t data; 791 }; 792 #endif 793 /* ARGSUSED */ 794 int 795 mount(td, uap) 796 struct thread *td; 797 struct mount_args /* { 798 char *type; 799 char *path; 800 int flags; 801 caddr_t data; 802 } */ *uap; 803 { 804 char *fstype; 805 char *fspath; 806 int error; 807 808 fstype = malloc(MFSNAMELEN, M_TEMP, 0); 809 fspath = malloc(MNAMELEN, M_TEMP, 0); 810 811 /* 812 * vfs_mount() actually takes a kernel string for `type' and 813 * `path' now, so extract them. 814 */ 815 error = copyinstr(uap->type, fstype, MFSNAMELEN, NULL); 816 if (error == 0) 817 error = copyinstr(uap->path, fspath, MNAMELEN, NULL); 818 if (error == 0) 819 error = vfs_mount(td, fstype, fspath, uap->flags, uap->data); 820 free(fstype, M_TEMP); 821 free(fspath, M_TEMP); 822 return (error); 823 } 824 825 /* 826 * vfs_mount(): actually attempt a filesystem mount. 827 * 828 * This routine is designed to be a "generic" entry point for routines 829 * that wish to mount a filesystem. All parameters except `fsdata' are 830 * pointers into kernel space. `fsdata' is currently still a pointer 831 * into userspace. 832 */ 833 int 834 vfs_mount(td, fstype, fspath, fsflags, fsdata) 835 struct thread *td; 836 const char *fstype; 837 char *fspath; 838 int fsflags; 839 void *fsdata; 840 { 841 linker_file_t lf; 842 struct vnode *vp; 843 struct mount *mp; 844 struct vfsconf *vfsp; 845 int error, flag = 0, kern_flag = 0; 846 struct vattr va; 847 struct nameidata nd; 848 849 /* 850 * Be ultra-paranoid about making sure the type and fspath 851 * variables will fit in our mp buffers, including the 852 * terminating NUL. 853 */ 854 if (strlen(fstype) >= MFSNAMELEN || strlen(fspath) >= MNAMELEN) 855 return (ENAMETOOLONG); 856 857 if (usermount == 0) { 858 error = suser(td); 859 if (error) 860 return (error); 861 } 862 /* 863 * Do not allow NFS export by non-root users. 864 */ 865 if (fsflags & MNT_EXPORTED) { 866 error = suser(td); 867 if (error) 868 return (error); 869 } 870 /* 871 * Silently enforce MNT_NOSUID and MNT_NODEV for non-root users. 872 */ 873 if (suser(td)) 874 fsflags |= MNT_NOSUID | MNT_NODEV; 875 /* 876 * Get vnode to be covered 877 */ 878 NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF, UIO_SYSSPACE, fspath, td); 879 if ((error = namei(&nd)) != 0) 880 return (error); 881 NDFREE(&nd, NDF_ONLY_PNBUF); 882 vp = nd.ni_vp; 883 if (fsflags & MNT_UPDATE) { 884 if ((vp->v_vflag & VV_ROOT) == 0) { 885 vput(vp); 886 return (EINVAL); 887 } 888 mp = vp->v_mount; 889 flag = mp->mnt_flag; 890 kern_flag = mp->mnt_kern_flag; 891 /* 892 * We only allow the filesystem to be reloaded if it 893 * is currently mounted read-only. 894 */ 895 if ((fsflags & MNT_RELOAD) && 896 ((mp->mnt_flag & MNT_RDONLY) == 0)) { 897 vput(vp); 898 return (EOPNOTSUPP); /* Needs translation */ 899 } 900 /* 901 * Only root, or the user that did the original mount is 902 * permitted to update it. 903 */ 904 if (mp->mnt_cred->cr_uid != td->td_ucred->cr_uid) { 905 error = suser(td); 906 if (error) { 907 vput(vp); 908 return (error); 909 } 910 } 911 if (vfs_busy(mp, LK_NOWAIT, 0, td)) { 912 vput(vp); 913 return (EBUSY); 914 } 915 VI_LOCK(vp); 916 if ((vp->v_iflag & VI_MOUNT) != 0 || 917 vp->v_mountedhere != NULL) { 918 VI_UNLOCK(vp); 919 vfs_unbusy(mp, td); 920 vput(vp); 921 return (EBUSY); 922 } 923 vp->v_iflag |= VI_MOUNT; 924 VI_UNLOCK(vp); 925 mp->mnt_flag |= fsflags & 926 (MNT_RELOAD | MNT_FORCE | MNT_UPDATE | MNT_SNAPSHOT); 927 VOP_UNLOCK(vp, 0, td); 928 goto update; 929 } 930 /* 931 * If the user is not root, ensure that they own the directory 932 * onto which we are attempting to mount. 933 */ 934 error = VOP_GETATTR(vp, &va, td->td_ucred, td); 935 if (error) { 936 vput(vp); 937 return (error); 938 } 939 if (va.va_uid != td->td_ucred->cr_uid) { 940 error = suser(td); 941 if (error) { 942 vput(vp); 943 return (error); 944 } 945 } 946 if ((error = vinvalbuf(vp, V_SAVE, td->td_ucred, td, 0, 0)) != 0) { 947 vput(vp); 948 return (error); 949 } 950 if (vp->v_type != VDIR) { 951 vput(vp); 952 return (ENOTDIR); 953 } 954 for (vfsp = vfsconf; vfsp; vfsp = vfsp->vfc_next) 955 if (!strcmp(vfsp->vfc_name, fstype)) 956 break; 957 if (vfsp == NULL) { 958 /* Only load modules for root (very important!). */ 959 error = suser(td); 960 if (error) { 961 vput(vp); 962 return (error); 963 } 964 error = securelevel_gt(td->td_ucred, 0); 965 if (error) { 966 vput(vp); 967 return (error); 968 } 969 error = linker_load_module(NULL, fstype, NULL, NULL, &lf); 970 if (error || lf == NULL) { 971 vput(vp); 972 if (lf == NULL) 973 error = ENODEV; 974 return (error); 975 } 976 lf->userrefs++; 977 /* Look up again to see if the VFS was loaded. */ 978 for (vfsp = vfsconf; vfsp; vfsp = vfsp->vfc_next) 979 if (!strcmp(vfsp->vfc_name, fstype)) 980 break; 981 if (vfsp == NULL) { 982 lf->userrefs--; 983 linker_file_unload(lf); 984 vput(vp); 985 return (ENODEV); 986 } 987 } 988 VI_LOCK(vp); 989 if ((vp->v_iflag & VI_MOUNT) != 0 || 990 vp->v_mountedhere != NULL) { 991 VI_UNLOCK(vp); 992 vput(vp); 993 return (EBUSY); 994 } 995 vp->v_iflag |= VI_MOUNT; 996 VI_UNLOCK(vp); 997 998 /* 999 * Allocate and initialize the filesystem. 1000 */ 1001 mp = malloc(sizeof(struct mount), M_MOUNT, M_ZERO); 1002 TAILQ_INIT(&mp->mnt_nvnodelist); 1003 TAILQ_INIT(&mp->mnt_reservedvnlist); 1004 lockinit(&mp->mnt_lock, PVFS, "vfslock", 0, LK_NOPAUSE); 1005 (void)vfs_busy(mp, LK_NOWAIT, 0, td); 1006 mp->mnt_op = vfsp->vfc_vfsops; 1007 mp->mnt_vfc = vfsp; 1008 vfsp->vfc_refcount++; 1009 mp->mnt_stat.f_type = vfsp->vfc_typenum; 1010 mp->mnt_flag |= vfsp->vfc_flags & MNT_VISFLAGMASK; 1011 strlcpy(mp->mnt_stat.f_fstypename, fstype, MFSNAMELEN); 1012 mp->mnt_vnodecovered = vp; 1013 mp->mnt_cred = crdup(td->td_ucred); 1014 mp->mnt_stat.f_owner = td->td_ucred->cr_uid; 1015 strlcpy(mp->mnt_stat.f_mntonname, fspath, MNAMELEN); 1016 mp->mnt_iosize_max = DFLTPHYS; 1017 #ifdef MAC 1018 mac_init_mount(mp); 1019 mac_create_mount(td->td_ucred, mp); 1020 #endif 1021 VOP_UNLOCK(vp, 0, td); 1022 update: 1023 /* 1024 * Check if the fs implements the old VFS_MOUNT() 1025 * function, since the old system call was used. 1026 */ 1027 if (mp->mnt_op->vfs_mount == NULL) { 1028 printf("%s doesn't support the old mount syscall\n", 1029 mp->mnt_vfc->vfc_name); 1030 VI_LOCK(vp); 1031 vp->v_iflag &= ~VI_MOUNT; 1032 VI_UNLOCK(vp); 1033 if (mp->mnt_flag & MNT_UPDATE) 1034 vfs_unbusy(mp, td); 1035 else { 1036 mp->mnt_vfc->vfc_refcount--; 1037 vfs_unbusy(mp, td); 1038 #ifdef MAC 1039 mac_destroy_mount(mp); 1040 #endif 1041 free(mp, M_MOUNT); 1042 } 1043 vrele(vp); 1044 return (EOPNOTSUPP); 1045 } 1046 1047 /* 1048 * Set the mount level flags. 1049 */ 1050 if (fsflags & MNT_RDONLY) 1051 mp->mnt_flag |= MNT_RDONLY; 1052 else if (mp->mnt_flag & MNT_RDONLY) 1053 mp->mnt_kern_flag |= MNTK_WANTRDWR; 1054 mp->mnt_flag &=~ MNT_UPDATEMASK; 1055 mp->mnt_flag |= fsflags & (MNT_UPDATEMASK | MNT_FORCE); 1056 /* 1057 * Mount the filesystem. 1058 * XXX The final recipients of VFS_MOUNT just overwrite the ndp they 1059 * get. No freeing of cn_pnbuf. 1060 */ 1061 error = VFS_MOUNT(mp, fspath, fsdata, &nd, td); 1062 if (mp->mnt_flag & MNT_UPDATE) { 1063 if (mp->mnt_kern_flag & MNTK_WANTRDWR) 1064 mp->mnt_flag &= ~MNT_RDONLY; 1065 mp->mnt_flag &=~ 1066 (MNT_UPDATE | MNT_RELOAD | MNT_FORCE | MNT_SNAPSHOT); 1067 mp->mnt_kern_flag &=~ MNTK_WANTRDWR; 1068 if (error) { 1069 mp->mnt_flag = flag; 1070 mp->mnt_kern_flag = kern_flag; 1071 } 1072 if ((mp->mnt_flag & MNT_RDONLY) == 0) { 1073 if (mp->mnt_syncer == NULL) 1074 error = vfs_allocate_syncvnode(mp); 1075 } else { 1076 if (mp->mnt_syncer != NULL) 1077 vrele(mp->mnt_syncer); 1078 mp->mnt_syncer = NULL; 1079 } 1080 vfs_unbusy(mp, td); 1081 VI_LOCK(vp); 1082 vp->v_iflag &= ~VI_MOUNT; 1083 VI_UNLOCK(vp); 1084 vrele(vp); 1085 return (error); 1086 } 1087 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY, td); 1088 /* 1089 * Put the new filesystem on the mount list after root. 1090 */ 1091 cache_purge(vp); 1092 if (!error) { 1093 struct vnode *newdp; 1094 1095 VI_LOCK(vp); 1096 vp->v_iflag &= ~VI_MOUNT; 1097 VI_UNLOCK(vp); 1098 vp->v_mountedhere = mp; 1099 mtx_lock(&mountlist_mtx); 1100 TAILQ_INSERT_TAIL(&mountlist, mp, mnt_list); 1101 mtx_unlock(&mountlist_mtx); 1102 if (VFS_ROOT(mp, &newdp)) 1103 panic("mount: lost mount"); 1104 checkdirs(vp, newdp); 1105 vput(newdp); 1106 VOP_UNLOCK(vp, 0, td); 1107 if ((mp->mnt_flag & MNT_RDONLY) == 0) 1108 error = vfs_allocate_syncvnode(mp); 1109 vfs_unbusy(mp, td); 1110 if ((error = VFS_START(mp, 0, td)) != 0) 1111 vrele(vp); 1112 } else { 1113 VI_LOCK(vp); 1114 vp->v_iflag &= ~VI_MOUNT; 1115 VI_UNLOCK(vp); 1116 mp->mnt_vfc->vfc_refcount--; 1117 vfs_unbusy(mp, td); 1118 #ifdef MAC 1119 mac_destroy_mount(mp); 1120 #endif 1121 free(mp, M_MOUNT); 1122 vput(vp); 1123 } 1124 return (error); 1125 } 1126 1127 /* 1128 * Scan all active processes to see if any of them have a current 1129 * or root directory of `olddp'. If so, replace them with the new 1130 * mount point. 1131 */ 1132 static void 1133 checkdirs(olddp, newdp) 1134 struct vnode *olddp, *newdp; 1135 { 1136 struct filedesc *fdp; 1137 struct proc *p; 1138 int nrele; 1139 1140 if (vrefcnt(olddp) == 1) 1141 return; 1142 sx_slock(&allproc_lock); 1143 LIST_FOREACH(p, &allproc, p_list) { 1144 PROC_LOCK(p); 1145 fdp = p->p_fd; 1146 if (fdp == NULL) { 1147 PROC_UNLOCK(p); 1148 continue; 1149 } 1150 nrele = 0; 1151 FILEDESC_LOCK(fdp); 1152 if (fdp->fd_cdir == olddp) { 1153 VREF(newdp); 1154 fdp->fd_cdir = newdp; 1155 nrele++; 1156 } 1157 if (fdp->fd_rdir == olddp) { 1158 VREF(newdp); 1159 fdp->fd_rdir = newdp; 1160 nrele++; 1161 } 1162 FILEDESC_UNLOCK(fdp); 1163 PROC_UNLOCK(p); 1164 while (nrele--) 1165 vrele(olddp); 1166 } 1167 sx_sunlock(&allproc_lock); 1168 if (rootvnode == olddp) { 1169 vrele(rootvnode); 1170 VREF(newdp); 1171 rootvnode = newdp; 1172 } 1173 } 1174 1175 /* 1176 * Unmount a filesystem. 1177 * 1178 * Note: unmount takes a path to the vnode mounted on as argument, 1179 * not special file (as before). 1180 */ 1181 #ifndef _SYS_SYSPROTO_H_ 1182 struct unmount_args { 1183 char *path; 1184 int flags; 1185 }; 1186 #endif 1187 /* ARGSUSED */ 1188 int 1189 unmount(td, uap) 1190 struct thread *td; 1191 register struct unmount_args /* { 1192 char *path; 1193 int flags; 1194 } */ *uap; 1195 { 1196 register struct vnode *vp; 1197 struct mount *mp; 1198 int error; 1199 struct nameidata nd; 1200 1201 NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF, UIO_USERSPACE, uap->path, td); 1202 if ((error = namei(&nd)) != 0) 1203 return (error); 1204 vp = nd.ni_vp; 1205 NDFREE(&nd, NDF_ONLY_PNBUF); 1206 mp = vp->v_mount; 1207 1208 /* 1209 * Only root, or the user that did the original mount is 1210 * permitted to unmount this filesystem. 1211 */ 1212 if (mp->mnt_cred->cr_uid != td->td_ucred->cr_uid) { 1213 error = suser(td); 1214 if (error) { 1215 vput(vp); 1216 return (error); 1217 } 1218 } 1219 1220 /* 1221 * Don't allow unmounting the root filesystem. 1222 */ 1223 if (mp->mnt_flag & MNT_ROOTFS) { 1224 vput(vp); 1225 return (EINVAL); 1226 } 1227 1228 /* 1229 * Must be the root of the filesystem 1230 */ 1231 if ((vp->v_vflag & VV_ROOT) == 0) { 1232 vput(vp); 1233 return (EINVAL); 1234 } 1235 vput(vp); 1236 return (dounmount(mp, uap->flags, td)); 1237 } 1238 1239 /* 1240 * Do the actual filesystem unmount. 1241 */ 1242 int 1243 dounmount(mp, flags, td) 1244 struct mount *mp; 1245 int flags; 1246 struct thread *td; 1247 { 1248 struct vnode *coveredvp, *fsrootvp; 1249 int error; 1250 int async_flag; 1251 1252 mtx_lock(&mountlist_mtx); 1253 if (mp->mnt_kern_flag & MNTK_UNMOUNT) { 1254 mtx_unlock(&mountlist_mtx); 1255 return (EBUSY); 1256 } 1257 mp->mnt_kern_flag |= MNTK_UNMOUNT; 1258 /* Allow filesystems to detect that a forced unmount is in progress. */ 1259 if (flags & MNT_FORCE) 1260 mp->mnt_kern_flag |= MNTK_UNMOUNTF; 1261 error = lockmgr(&mp->mnt_lock, LK_DRAIN | LK_INTERLOCK | 1262 ((flags & MNT_FORCE) ? 0 : LK_NOWAIT), &mountlist_mtx, td); 1263 if (error) { 1264 mp->mnt_kern_flag &= ~(MNTK_UNMOUNT | MNTK_UNMOUNTF); 1265 if (mp->mnt_kern_flag & MNTK_MWAIT) 1266 wakeup(mp); 1267 return (error); 1268 } 1269 vn_start_write(NULL, &mp, V_WAIT); 1270 1271 if (mp->mnt_flag & MNT_EXPUBLIC) 1272 vfs_setpublicfs(NULL, NULL, NULL); 1273 1274 vfs_msync(mp, MNT_WAIT); 1275 async_flag = mp->mnt_flag & MNT_ASYNC; 1276 mp->mnt_flag &=~ MNT_ASYNC; 1277 cache_purgevfs(mp); /* remove cache entries for this file sys */ 1278 if (mp->mnt_syncer != NULL) 1279 vrele(mp->mnt_syncer); 1280 /* Move process cdir/rdir refs on fs root to underlying vnode. */ 1281 if (VFS_ROOT(mp, &fsrootvp) == 0) { 1282 if (mp->mnt_vnodecovered != NULL) 1283 checkdirs(fsrootvp, mp->mnt_vnodecovered); 1284 if (fsrootvp == rootvnode) { 1285 vrele(rootvnode); 1286 rootvnode = NULL; 1287 } 1288 vput(fsrootvp); 1289 } 1290 if (((mp->mnt_flag & MNT_RDONLY) || 1291 (error = VFS_SYNC(mp, MNT_WAIT, td->td_ucred, td)) == 0) || 1292 (flags & MNT_FORCE)) { 1293 error = VFS_UNMOUNT(mp, flags, td); 1294 } 1295 vn_finished_write(mp); 1296 if (error) { 1297 /* Undo cdir/rdir and rootvnode changes made above. */ 1298 if (VFS_ROOT(mp, &fsrootvp) == 0) { 1299 if (mp->mnt_vnodecovered != NULL) 1300 checkdirs(mp->mnt_vnodecovered, fsrootvp); 1301 if (rootvnode == NULL) { 1302 rootvnode = fsrootvp; 1303 vref(rootvnode); 1304 } 1305 vput(fsrootvp); 1306 } 1307 if ((mp->mnt_flag & MNT_RDONLY) == 0 && mp->mnt_syncer == NULL) 1308 (void) vfs_allocate_syncvnode(mp); 1309 mtx_lock(&mountlist_mtx); 1310 mp->mnt_kern_flag &= ~(MNTK_UNMOUNT | MNTK_UNMOUNTF); 1311 mp->mnt_flag |= async_flag; 1312 lockmgr(&mp->mnt_lock, LK_RELEASE | LK_INTERLOCK, 1313 &mountlist_mtx, td); 1314 if (mp->mnt_kern_flag & MNTK_MWAIT) 1315 wakeup(mp); 1316 return (error); 1317 } 1318 crfree(mp->mnt_cred); 1319 mtx_lock(&mountlist_mtx); 1320 TAILQ_REMOVE(&mountlist, mp, mnt_list); 1321 if ((coveredvp = mp->mnt_vnodecovered) != NULL) 1322 coveredvp->v_mountedhere = NULL; 1323 mp->mnt_vfc->vfc_refcount--; 1324 if (!TAILQ_EMPTY(&mp->mnt_nvnodelist)) 1325 panic("unmount: dangling vnode"); 1326 lockmgr(&mp->mnt_lock, LK_RELEASE | LK_INTERLOCK, &mountlist_mtx, td); 1327 lockdestroy(&mp->mnt_lock); 1328 if (coveredvp != NULL) 1329 vrele(coveredvp); 1330 if (mp->mnt_kern_flag & MNTK_MWAIT) 1331 wakeup(mp); 1332 #ifdef MAC 1333 mac_destroy_mount(mp); 1334 #endif 1335 if (mp->mnt_op->vfs_mount == NULL) 1336 vfs_freeopts(mp->mnt_opt); 1337 free(mp, M_MOUNT); 1338 return (0); 1339 } 1340 1341 /* 1342 * Lookup a filesystem type, and if found allocate and initialize 1343 * a mount structure for it. 1344 * 1345 * Devname is usually updated by mount(8) after booting. 1346 */ 1347 int 1348 vfs_rootmountalloc(fstypename, devname, mpp) 1349 char *fstypename; 1350 char *devname; 1351 struct mount **mpp; 1352 { 1353 struct thread *td = curthread; /* XXX */ 1354 struct vfsconf *vfsp; 1355 struct mount *mp; 1356 1357 if (fstypename == NULL) 1358 return (ENODEV); 1359 for (vfsp = vfsconf; vfsp; vfsp = vfsp->vfc_next) 1360 if (!strcmp(vfsp->vfc_name, fstypename)) 1361 break; 1362 if (vfsp == NULL) 1363 return (ENODEV); 1364 mp = malloc((u_long)sizeof(struct mount), M_MOUNT, M_ZERO); 1365 lockinit(&mp->mnt_lock, PVFS, "vfslock", 0, LK_NOPAUSE); 1366 (void)vfs_busy(mp, LK_NOWAIT, 0, td); 1367 TAILQ_INIT(&mp->mnt_nvnodelist); 1368 TAILQ_INIT(&mp->mnt_reservedvnlist); 1369 mp->mnt_vfc = vfsp; 1370 mp->mnt_op = vfsp->vfc_vfsops; 1371 mp->mnt_flag = MNT_RDONLY; 1372 mp->mnt_vnodecovered = NULLVP; 1373 mp->mnt_cred = crdup(td->td_ucred); 1374 vfsp->vfc_refcount++; 1375 mp->mnt_iosize_max = DFLTPHYS; 1376 mp->mnt_stat.f_type = vfsp->vfc_typenum; 1377 mp->mnt_flag |= vfsp->vfc_flags & MNT_VISFLAGMASK; 1378 strlcpy(mp->mnt_stat.f_fstypename, vfsp->vfc_name, MFSNAMELEN); 1379 mp->mnt_stat.f_mntonname[0] = '/'; 1380 mp->mnt_stat.f_mntonname[1] = 0; 1381 strlcpy(mp->mnt_stat.f_mntfromname, devname, MNAMELEN); 1382 #ifdef MAC 1383 mac_init_mount(mp); 1384 mac_create_mount(td->td_ucred, mp); 1385 #endif 1386 *mpp = mp; 1387 return (0); 1388 } 1389 1390 /* 1391 * Find and mount the root filesystem 1392 */ 1393 void 1394 vfs_mountroot(void) 1395 { 1396 char *cp; 1397 int i, error; 1398 1399 g_waitidle(); 1400 1401 /* 1402 * The root filesystem information is compiled in, and we are 1403 * booted with instructions to use it. 1404 */ 1405 #ifdef ROOTDEVNAME 1406 if ((boothowto & RB_DFLTROOT) && 1407 !vfs_mountroot_try(ROOTDEVNAME)) 1408 return; 1409 #endif 1410 /* 1411 * We are booted with instructions to prompt for the root filesystem, 1412 * or to use the compiled-in default when it doesn't exist. 1413 */ 1414 if (boothowto & (RB_DFLTROOT | RB_ASKNAME)) { 1415 if (!vfs_mountroot_ask()) 1416 return; 1417 } 1418 1419 /* 1420 * We've been given the generic "use CDROM as root" flag. This is 1421 * necessary because one media may be used in many different 1422 * devices, so we need to search for them. 1423 */ 1424 if (boothowto & RB_CDROM) { 1425 for (i = 0; cdrom_rootdevnames[i] != NULL; i++) { 1426 if (!vfs_mountroot_try(cdrom_rootdevnames[i])) 1427 return; 1428 } 1429 } 1430 1431 /* 1432 * Try to use the value read by the loader from /etc/fstab, or 1433 * supplied via some other means. This is the preferred 1434 * mechanism. 1435 */ 1436 if ((cp = getenv("vfs.root.mountfrom")) != NULL) { 1437 error = vfs_mountroot_try(cp); 1438 freeenv(cp); 1439 if (!error) 1440 return; 1441 } 1442 1443 /* 1444 * Try values that may have been computed by the machine-dependant 1445 * legacy code. 1446 */ 1447 if (!vfs_mountroot_try(rootdevnames[0])) 1448 return; 1449 if (!vfs_mountroot_try(rootdevnames[1])) 1450 return; 1451 1452 /* 1453 * If we have a compiled-in default, and haven't already tried it, try 1454 * it now. 1455 */ 1456 #ifdef ROOTDEVNAME 1457 if (!(boothowto & RB_DFLTROOT)) 1458 if (!vfs_mountroot_try(ROOTDEVNAME)) 1459 return; 1460 #endif 1461 1462 /* 1463 * Everything so far has failed, prompt on the console if we haven't 1464 * already tried that. 1465 */ 1466 if (!(boothowto & (RB_DFLTROOT | RB_ASKNAME)) && !vfs_mountroot_ask()) 1467 return; 1468 panic("Root mount failed, startup aborted."); 1469 } 1470 1471 /* 1472 * Mount (mountfrom) as the root filesystem. 1473 */ 1474 static int 1475 vfs_mountroot_try(char *mountfrom) 1476 { 1477 struct mount *mp; 1478 char *vfsname, *path; 1479 const char *devname; 1480 int error; 1481 char patt[32]; 1482 int s; 1483 1484 vfsname = NULL; 1485 path = NULL; 1486 mp = NULL; 1487 error = EINVAL; 1488 1489 if (mountfrom == NULL) 1490 return(error); /* don't complain */ 1491 1492 s = splcam(); /* Overkill, but annoying without it */ 1493 printf("Mounting root from %s\n", mountfrom); 1494 splx(s); 1495 1496 /* parse vfs name and path */ 1497 vfsname = malloc(MFSNAMELEN, M_MOUNT, 0); 1498 path = malloc(MNAMELEN, M_MOUNT, 0); 1499 vfsname[0] = path[0] = 0; 1500 sprintf(patt, "%%%d[a-z0-9]:%%%zds", MFSNAMELEN, MNAMELEN); 1501 if (sscanf(mountfrom, patt, vfsname, path) < 1) 1502 goto done; 1503 1504 /* allocate a root mount */ 1505 error = vfs_rootmountalloc(vfsname, path[0] != 0 ? path : ROOTNAME, 1506 &mp); 1507 if (error != 0) { 1508 printf("Can't allocate root mount for filesystem '%s': %d\n", 1509 vfsname, error); 1510 goto done; 1511 } 1512 mp->mnt_flag |= MNT_ROOTFS; 1513 1514 /* do our best to set rootdev */ 1515 if ((path[0] != 0) && setrootbyname(path)) 1516 printf("setrootbyname failed\n"); 1517 1518 /* If the root device is a type "memory disk", mount RW */ 1519 if (rootdev != NODEV && devsw(rootdev) != NULL) { 1520 devname = devtoname(rootdev); 1521 if (devname[0] == 'm' && devname[1] == 'd') 1522 mp->mnt_flag &= ~MNT_RDONLY; 1523 } 1524 1525 /* 1526 * Set the mount path to be something useful, because the 1527 * filesystem code isn't responsible now for initialising 1528 * f_mntonname unless they want to override the default 1529 * (which is `path'.) 1530 */ 1531 strlcpy(mp->mnt_stat.f_mntonname, "/", MNAMELEN); 1532 1533 error = VFS_MOUNT(mp, NULL, NULL, NULL, curthread); 1534 1535 done: 1536 if (vfsname != NULL) 1537 free(vfsname, M_MOUNT); 1538 if (path != NULL) 1539 free(path, M_MOUNT); 1540 if (error != 0) { 1541 if (mp != NULL) { 1542 vfs_unbusy(mp, curthread); 1543 #ifdef MAC 1544 mac_destroy_mount(mp); 1545 #endif 1546 free(mp, M_MOUNT); 1547 } 1548 printf("Root mount failed: %d\n", error); 1549 } else { 1550 1551 /* register with list of mounted filesystems */ 1552 mtx_lock(&mountlist_mtx); 1553 TAILQ_INSERT_HEAD(&mountlist, mp, mnt_list); 1554 mtx_unlock(&mountlist_mtx); 1555 1556 /* sanity check system clock against root fs timestamp */ 1557 inittodr(mp->mnt_time); 1558 vfs_unbusy(mp, curthread); 1559 error = VFS_START(mp, 0, curthread); 1560 } 1561 return(error); 1562 } 1563 1564 /* 1565 * Spin prompting on the console for a suitable root filesystem 1566 */ 1567 static int 1568 vfs_mountroot_ask(void) 1569 { 1570 char name[128]; 1571 int i; 1572 dev_t dev; 1573 1574 for(;;) { 1575 printf("\nManual root filesystem specification:\n"); 1576 printf(" <fstype>:<device> Mount <device> using filesystem <fstype>\n"); 1577 #if defined(__i386__) || defined(__ia64__) 1578 printf(" eg. ufs:da0s1a\n"); 1579 #else 1580 printf(" eg. ufs:da0a\n"); 1581 #endif 1582 printf(" ? List valid disk boot devices\n"); 1583 printf(" <empty line> Abort manual input\n"); 1584 printf("\nmountroot> "); 1585 gets(name); 1586 if (name[0] == 0) 1587 return(1); 1588 if (name[0] == '?') { 1589 if (!g_dev_print()) { 1590 printf("Possibly valid devices for 'ufs' root:\n"); 1591 for (i = 0; i < NUMCDEVSW; i++) { 1592 dev = makedev(i, 0); 1593 if (devsw(dev) != NULL) 1594 printf(" \"%s\"", devsw(dev)->d_name); 1595 } 1596 } 1597 printf("\n"); 1598 continue; 1599 } 1600 if (!vfs_mountroot_try(name)) 1601 return(0); 1602 } 1603 } 1604 1605 /* 1606 * Local helper function for vfs_mountroot_ask. 1607 */ 1608 static void 1609 gets(char *cp) 1610 { 1611 char *lp; 1612 int c; 1613 1614 lp = cp; 1615 for (;;) { 1616 printf("%c", c = cngetc() & 0177); 1617 switch (c) { 1618 case -1: 1619 case '\n': 1620 case '\r': 1621 *lp++ = '\0'; 1622 return; 1623 case '\b': 1624 case '\177': 1625 if (lp > cp) { 1626 printf(" \b"); 1627 lp--; 1628 } 1629 continue; 1630 case '#': 1631 lp--; 1632 if (lp < cp) 1633 lp = cp; 1634 continue; 1635 case '@': 1636 case 'u' & 037: 1637 lp = cp; 1638 printf("%c", '\n'); 1639 continue; 1640 default: 1641 *lp++ = c; 1642 } 1643 } 1644 } 1645 1646 /* 1647 * Convert a given name to the dev_t of the disk-like device 1648 * it refers to. 1649 */ 1650 dev_t 1651 getdiskbyname(char *name) { 1652 char *cp; 1653 dev_t dev; 1654 1655 cp = name; 1656 if (!bcmp(cp, "/dev/", 5)) 1657 cp += 5; 1658 1659 dev = NODEV; 1660 EVENTHANDLER_INVOKE(dev_clone, cp, strlen(cp), &dev); 1661 return (dev); 1662 } 1663 1664 /* 1665 * Set rootdev to match (name), given that we expect it to 1666 * refer to a disk-like device. 1667 */ 1668 static int 1669 setrootbyname(char *name) 1670 { 1671 dev_t diskdev; 1672 1673 diskdev = getdiskbyname(name); 1674 if (diskdev != NODEV) { 1675 rootdev = diskdev; 1676 return (0); 1677 } 1678 1679 return (1); 1680 } 1681 1682 /* Show the dev_t for a disk specified by name */ 1683 #ifdef DDB 1684 DB_SHOW_COMMAND(disk, db_getdiskbyname) 1685 { 1686 dev_t dev; 1687 1688 if (modif[0] == '\0') { 1689 db_error("usage: show disk/devicename"); 1690 return; 1691 } 1692 dev = getdiskbyname(modif); 1693 if (dev != NODEV) 1694 db_printf("dev_t = %p\n", dev); 1695 else 1696 db_printf("No disk device matched.\n"); 1697 } 1698 #endif 1699 1700 /* 1701 * Get a mount option by its name. 1702 * 1703 * Return 0 if the option was found, ENOENT otherwise. 1704 * If len is non-NULL it will be filled with the length 1705 * of the option. If buf is non-NULL, it will be filled 1706 * with the address of the option. 1707 */ 1708 int 1709 vfs_getopt(opts, name, buf, len) 1710 struct vfsoptlist *opts; 1711 const char *name; 1712 void **buf; 1713 int *len; 1714 { 1715 struct vfsopt *opt; 1716 1717 TAILQ_FOREACH(opt, opts, link) { 1718 if (strcmp(name, opt->name) == 0) { 1719 if (len != NULL) 1720 *len = opt->len; 1721 if (buf != NULL) 1722 *buf = opt->value; 1723 return (0); 1724 } 1725 } 1726 return (ENOENT); 1727 } 1728 1729 /* 1730 * Find and copy a mount option. 1731 * 1732 * The size of the buffer has to be specified 1733 * in len, if it is not the same length as the 1734 * mount option, EINVAL is returned. 1735 * Returns ENOENT if the option is not found. 1736 */ 1737 int 1738 vfs_copyopt(opts, name, dest, len) 1739 struct vfsoptlist *opts; 1740 const char *name; 1741 void *dest; 1742 int len; 1743 { 1744 struct vfsopt *opt; 1745 1746 TAILQ_FOREACH(opt, opts, link) { 1747 if (strcmp(name, opt->name) == 0) { 1748 if (len != opt->len) 1749 return (EINVAL); 1750 bcopy(opt->value, dest, opt->len); 1751 return (0); 1752 } 1753 } 1754 return (ENOENT); 1755 } 1756