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/malloc.h> 73 #include <sys/mac.h> 74 #include <sys/mount.h> 75 #include <sys/mutex.h> 76 #include <sys/namei.h> 77 #include <sys/proc.h> 78 #include <sys/reboot.h> 79 #include <sys/sysproto.h> 80 #include <sys/sx.h> 81 #include <sys/sysctl.h> 82 #include <sys/sysent.h> 83 #include <sys/systm.h> 84 #include <sys/vnode.h> 85 86 #include <machine/stdarg.h> 87 88 #include "opt_rootdevname.h" 89 #include "opt_ddb.h" 90 #include "opt_mac.h" 91 92 #ifdef DDB 93 #include <ddb/ddb.h> 94 #endif 95 96 #define ROOTNAME "root_device" 97 98 static void checkdirs(struct vnode *olddp, struct vnode *newdp); 99 static int vfs_nmount(struct thread *td, int, struct uio *); 100 static int vfs_mountroot_try(char *mountfrom); 101 static int vfs_mountroot_ask(void); 102 static void gets(char *cp); 103 104 static int usermount = 0; /* if 1, non-root can mount fs. */ 105 SYSCTL_INT(_vfs, OID_AUTO, usermount, CTLFLAG_RW, &usermount, 0, ""); 106 107 MALLOC_DEFINE(M_MOUNT, "mount", "vfs mount structure"); 108 109 /* List of mounted filesystems. */ 110 struct mntlist mountlist = TAILQ_HEAD_INITIALIZER(mountlist); 111 112 /* For any iteration/modification of mountlist */ 113 struct mtx mountlist_mtx; 114 115 /* For any iteration/modification of mnt_vnodelist */ 116 struct mtx mntvnode_mtx; 117 118 /* 119 * The vnode of the system's root (/ in the filesystem, without chroot 120 * active.) 121 */ 122 struct vnode *rootvnode; 123 124 /* 125 * The root filesystem is detailed in the kernel environment variable 126 * vfs.root.mountfrom, which is expected to be in the general format 127 * 128 * <vfsname>:[<path>] 129 * vfsname := the name of a VFS known to the kernel and capable 130 * of being mounted as root 131 * path := disk device name or other data used by the filesystem 132 * to locate its physical store 133 */ 134 135 /* 136 * The root specifiers we will try if RB_CDROM is specified. 137 */ 138 static char *cdrom_rootdevnames[] = { 139 "cd9660:cd0a", 140 "cd9660:acd0a", 141 "cd9660:wcd0a", 142 NULL 143 }; 144 145 /* legacy find-root code */ 146 char *rootdevnames[2] = {NULL, NULL}; 147 static int setrootbyname(char *name); 148 dev_t rootdev = NODEV; 149 150 /* Remove one mount option. */ 151 static void 152 vfs_freeopt(struct vfsoptlist *opts, struct vfsopt *opt) 153 { 154 155 TAILQ_REMOVE(opts, opt, link); 156 free(opt->name, M_MOUNT); 157 if (opt->value != NULL) 158 free(opt->value, M_MOUNT); 159 #ifdef INVARIANTS 160 else if (opt->len != 0) 161 panic("%s: mount option with NULL value but length != 0", 162 __func__); 163 #endif 164 free(opt, M_MOUNT); 165 } 166 167 /* Release all resources related to the mount options. */ 168 static void 169 vfs_freeopts(struct vfsoptlist *opts) 170 { 171 struct vfsopt *opt; 172 173 while (!TAILQ_EMPTY(opts)) { 174 opt = TAILQ_FIRST(opts); 175 vfs_freeopt(opts, opt); 176 } 177 free(opts, M_MOUNT); 178 } 179 180 /* 181 * If a mount option is specified several times, 182 * (with or without the "no" prefix) only keep 183 * the last occurence of it. 184 */ 185 static void 186 vfs_sanitizeopts(struct vfsoptlist *opts) 187 { 188 struct vfsopt *opt, *opt2, *tmp; 189 int noopt; 190 191 TAILQ_FOREACH_REVERSE(opt, opts, vfsoptlist, link) { 192 if (strncmp(opt->name, "no", 2) == 0) 193 noopt = 1; 194 else 195 noopt = 0; 196 opt2 = TAILQ_PREV(opt, vfsoptlist, link); 197 while (opt2 != NULL) { 198 if (strcmp(opt2->name, opt->name) == 0 || 199 (noopt && strcmp(opt->name + 2, opt2->name) == 0) || 200 (!noopt && strncmp(opt2->name, "no", 2) == 0 && 201 strcmp(opt2->name + 2, opt->name) == 0)) { 202 tmp = TAILQ_PREV(opt2, vfsoptlist, link); 203 vfs_freeopt(opts, opt2); 204 opt2 = tmp; 205 } else { 206 opt2 = TAILQ_PREV(opt2, vfsoptlist, link); 207 } 208 } 209 } 210 } 211 212 /* 213 * Build a linked list of mount options from a struct uio. 214 */ 215 static int 216 vfs_buildopts(struct uio *auio, struct vfsoptlist **options) 217 { 218 struct vfsoptlist *opts; 219 struct vfsopt *opt; 220 unsigned int i, iovcnt; 221 int error, namelen, optlen; 222 223 iovcnt = auio->uio_iovcnt; 224 opts = malloc(sizeof(struct vfsoptlist), M_MOUNT, M_WAITOK); 225 TAILQ_INIT(opts); 226 for (i = 0; i < iovcnt; i += 2) { 227 opt = malloc(sizeof(struct vfsopt), M_MOUNT, M_WAITOK); 228 namelen = auio->uio_iov[i].iov_len; 229 optlen = auio->uio_iov[i + 1].iov_len; 230 opt->name = malloc(namelen, M_MOUNT, M_WAITOK); 231 opt->value = NULL; 232 if (auio->uio_segflg == UIO_SYSSPACE) { 233 bcopy(auio->uio_iov[i].iov_base, opt->name, namelen); 234 } else { 235 error = copyin(auio->uio_iov[i].iov_base, opt->name, 236 namelen); 237 if (error) 238 goto bad; 239 } 240 opt->len = optlen; 241 if (optlen != 0) { 242 opt->value = malloc(optlen, M_MOUNT, M_WAITOK); 243 if (auio->uio_segflg == UIO_SYSSPACE) { 244 bcopy(auio->uio_iov[i + 1].iov_base, opt->value, 245 optlen); 246 } else { 247 error = copyin(auio->uio_iov[i + 1].iov_base, 248 opt->value, optlen); 249 if (error) 250 goto bad; 251 } 252 } 253 TAILQ_INSERT_TAIL(opts, opt, link); 254 } 255 vfs_sanitizeopts(opts); 256 *options = opts; 257 return (0); 258 bad: 259 vfs_freeopts(opts); 260 return (error); 261 } 262 263 /* 264 * Merge the old mount options with the new ones passed 265 * in the MNT_UPDATE case. 266 */ 267 static void 268 vfs_mergeopts(struct vfsoptlist *toopts, struct vfsoptlist *opts) 269 { 270 struct vfsopt *opt, *opt2, *new; 271 272 TAILQ_FOREACH(opt, opts, link) { 273 /* 274 * Check that this option hasn't been redefined 275 * nor cancelled with a "no" mount option. 276 */ 277 opt2 = TAILQ_FIRST(toopts); 278 while (opt2 != NULL) { 279 if (strcmp(opt2->name, opt->name) == 0) 280 goto next; 281 if (strncmp(opt2->name, "no", 2) == 0 && 282 strcmp(opt2->name + 2, opt->name) == 0) { 283 vfs_freeopt(toopts, opt2); 284 goto next; 285 } 286 opt2 = TAILQ_NEXT(opt2, link); 287 } 288 /* We want this option, duplicate it. */ 289 new = malloc(sizeof(struct vfsopt), M_MOUNT, M_WAITOK); 290 new->name = malloc(strlen(opt->name) + 1, M_MOUNT, M_WAITOK); 291 strcpy(new->name, opt->name); 292 if (opt->len != 0) { 293 new->value = malloc(opt->len, M_MOUNT, M_WAITOK); 294 bcopy(opt->value, new->value, opt->len); 295 } else { 296 new->value = NULL; 297 } 298 new->len = opt->len; 299 TAILQ_INSERT_TAIL(toopts, new, link); 300 next: 301 continue; 302 } 303 } 304 305 /* 306 * New mount API. 307 */ 308 int 309 nmount(td, uap) 310 struct thread *td; 311 struct nmount_args /* { 312 syscallarg(struct iovec *) iovp; 313 syscallarg(unsigned int) iovcnt; 314 syscallarg(int) flags; 315 } */ *uap; 316 { 317 struct uio auio; 318 struct iovec *iov, *needfree; 319 struct iovec aiov[UIO_SMALLIOV]; 320 unsigned int i; 321 int error; 322 u_int iovlen, iovcnt; 323 324 iovcnt = SCARG(uap, iovcnt); 325 iovlen = iovcnt * sizeof (struct iovec); 326 /* 327 * Check that we have an even number of iovec's 328 * and that we have at least two options. 329 */ 330 if ((iovcnt & 1) || (iovcnt < 4) || (iovcnt > UIO_MAXIOV)) 331 return (EINVAL); 332 333 if (iovcnt > UIO_SMALLIOV) { 334 MALLOC(iov, struct iovec *, iovlen, M_IOV, M_WAITOK); 335 needfree = iov; 336 } else { 337 iov = aiov; 338 needfree = NULL; 339 } 340 auio.uio_iov = iov; 341 auio.uio_iovcnt = iovcnt; 342 auio.uio_segflg = UIO_USERSPACE; 343 if ((error = copyin(uap->iovp, iov, iovlen))) 344 goto finish; 345 346 for (i = 0; i < iovcnt; i++) { 347 if (iov->iov_len > MMAXOPTIONLEN) { 348 error = EINVAL; 349 goto finish; 350 } 351 iov++; 352 } 353 error = vfs_nmount(td, SCARG(uap, flags), &auio); 354 finish: 355 if (needfree != NULL) 356 free(needfree, M_TEMP); 357 return (error); 358 } 359 360 int 361 kernel_mount(iovp, iovcnt, flags) 362 struct iovec *iovp; 363 unsigned int iovcnt; 364 int flags; 365 { 366 struct uio auio; 367 int error; 368 369 /* 370 * Check that we have an even number of iovec's 371 * and that we have at least two options. 372 */ 373 if ((iovcnt & 1) || (iovcnt < 4)) 374 return (EINVAL); 375 376 auio.uio_iov = iovp; 377 auio.uio_iovcnt = iovcnt; 378 auio.uio_segflg = UIO_SYSSPACE; 379 380 error = vfs_nmount(curthread, flags, &auio); 381 return (error); 382 } 383 384 int 385 kernel_vmount(int flags, ...) 386 { 387 struct iovec *iovp; 388 struct uio auio; 389 va_list ap; 390 unsigned int iovcnt, iovlen, len; 391 const char *cp; 392 char *buf, *pos; 393 size_t n; 394 int error, i; 395 396 len = 0; 397 va_start(ap, flags); 398 for (iovcnt = 0; (cp = va_arg(ap, const char *)) != NULL; iovcnt++) 399 len += strlen(cp) + 1; 400 va_end(ap); 401 402 if (iovcnt < 4 || iovcnt & 1) 403 return (EINVAL); 404 405 iovlen = iovcnt * sizeof (struct iovec); 406 MALLOC(iovp, struct iovec *, iovlen, M_MOUNT, M_WAITOK); 407 MALLOC(buf, char *, len, M_MOUNT, M_WAITOK); 408 pos = buf; 409 va_start(ap, flags); 410 for (i = 0; i < iovcnt; i++) { 411 cp = va_arg(ap, const char *); 412 copystr(cp, pos, len - (pos - buf), &n); 413 iovp[i].iov_base = pos; 414 iovp[i].iov_len = n; 415 pos += n; 416 } 417 va_end(ap); 418 419 auio.uio_iov = iovp; 420 auio.uio_iovcnt = iovcnt; 421 auio.uio_segflg = UIO_SYSSPACE; 422 423 error = vfs_nmount(curthread, flags, &auio); 424 FREE(iovp, M_MOUNT); 425 FREE(buf, M_MOUNT); 426 return (error); 427 } 428 429 /* 430 * vfs_nmount(): actually attempt a filesystem mount. 431 */ 432 static int 433 vfs_nmount(td, fsflags, fsoptions) 434 struct thread *td; 435 int fsflags; /* Flags common to all filesystems. */ 436 struct uio *fsoptions; /* Options local to the filesystem. */ 437 { 438 linker_file_t lf; 439 struct vnode *vp; 440 struct mount *mp; 441 struct vfsconf *vfsp; 442 struct vfsoptlist *optlist; 443 char *fstype, *fspath; 444 int error, flag = 0, kern_flag = 0; 445 int fstypelen, fspathlen; 446 struct vattr va; 447 struct nameidata nd; 448 449 error = vfs_buildopts(fsoptions, &optlist); 450 if (error) 451 return (error); 452 453 /* 454 * We need these two options before the others, 455 * and they are mandatory for any filesystem. 456 * Ensure they are NUL terminated as well. 457 */ 458 fstypelen = 0; 459 error = vfs_getopt(optlist, "fstype", (void **)&fstype, &fstypelen); 460 if (error || fstype[fstypelen - 1] != '\0') { 461 error = EINVAL; 462 goto bad; 463 } 464 fspathlen = 0; 465 error = vfs_getopt(optlist, "fspath", (void **)&fspath, &fspathlen); 466 if (error || fspath[fspathlen - 1] != '\0') { 467 error = EINVAL; 468 goto bad; 469 } 470 471 /* 472 * Be ultra-paranoid about making sure the type and fspath 473 * variables will fit in our mp buffers, including the 474 * terminating NUL. 475 */ 476 if (fstypelen >= MFSNAMELEN - 1 || fspathlen >= MNAMELEN - 1) { 477 error = ENAMETOOLONG; 478 goto bad; 479 } 480 481 if (usermount == 0) { 482 error = suser(td); 483 if (error) 484 goto bad; 485 } 486 /* 487 * Do not allow NFS export by non-root users. 488 */ 489 if (fsflags & MNT_EXPORTED) { 490 error = suser(td); 491 if (error) 492 goto bad; 493 } 494 /* 495 * Silently enforce MNT_NOSUID and MNT_NODEV for non-root users. 496 */ 497 if (suser(td)) 498 fsflags |= MNT_NOSUID | MNT_NODEV; 499 /* 500 * Get vnode to be covered 501 */ 502 NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF, UIO_SYSSPACE, fspath, td); 503 if ((error = namei(&nd)) != 0) 504 goto bad; 505 NDFREE(&nd, NDF_ONLY_PNBUF); 506 vp = nd.ni_vp; 507 if (fsflags & MNT_UPDATE) { 508 if ((vp->v_flag & VROOT) == 0) { 509 vput(vp); 510 error = EINVAL; 511 goto bad; 512 } 513 mp = vp->v_mount; 514 flag = mp->mnt_flag; 515 kern_flag = mp->mnt_kern_flag; 516 /* 517 * We only allow the filesystem to be reloaded if it 518 * is currently mounted read-only. 519 */ 520 if ((fsflags & MNT_RELOAD) && 521 ((mp->mnt_flag & MNT_RDONLY) == 0)) { 522 vput(vp); 523 error = EOPNOTSUPP; /* Needs translation */ 524 goto bad; 525 } 526 /* 527 * Only root, or the user that did the original mount is 528 * permitted to update it. 529 */ 530 if (mp->mnt_stat.f_owner != td->td_ucred->cr_uid) { 531 error = suser(td); 532 if (error) { 533 vput(vp); 534 goto bad; 535 } 536 } 537 if (vfs_busy(mp, LK_NOWAIT, 0, td)) { 538 vput(vp); 539 error = EBUSY; 540 goto bad; 541 } 542 mtx_lock(&vp->v_interlock); 543 if ((vp->v_flag & VMOUNT) != 0 || vp->v_mountedhere != NULL) { 544 mtx_unlock(&vp->v_interlock); 545 vfs_unbusy(mp, td); 546 vput(vp); 547 error = EBUSY; 548 goto bad; 549 } 550 vp->v_flag |= VMOUNT; 551 mtx_unlock(&vp->v_interlock); 552 mp->mnt_flag |= fsflags & 553 (MNT_RELOAD | MNT_FORCE | MNT_UPDATE | MNT_SNAPSHOT); 554 VOP_UNLOCK(vp, 0, td); 555 mp->mnt_optnew = optlist; 556 vfs_mergeopts(mp->mnt_optnew, mp->mnt_opt); 557 goto update; 558 } 559 /* 560 * If the user is not root, ensure that they own the directory 561 * onto which we are attempting to mount. 562 */ 563 error = VOP_GETATTR(vp, &va, td->td_ucred, td); 564 if (error) { 565 vput(vp); 566 goto bad; 567 } 568 if (va.va_uid != td->td_ucred->cr_uid) { 569 error = suser(td); 570 if (error) { 571 vput(vp); 572 goto bad; 573 } 574 } 575 if ((error = vinvalbuf(vp, V_SAVE, td->td_ucred, td, 0, 0)) != 0) { 576 vput(vp); 577 goto bad; 578 } 579 if (vp->v_type != VDIR) { 580 vput(vp); 581 error = ENOTDIR; 582 goto bad; 583 } 584 for (vfsp = vfsconf; vfsp; vfsp = vfsp->vfc_next) 585 if (!strcmp(vfsp->vfc_name, fstype)) 586 break; 587 if (vfsp == NULL) { 588 /* Only load modules for root (very important!). */ 589 error = suser(td); 590 if (error) { 591 vput(vp); 592 goto bad; 593 } 594 error = securelevel_gt(td->td_ucred, 0); 595 if (error) { 596 vput(vp); 597 goto bad; 598 } 599 error = linker_load_file(fstype, &lf); 600 if (error || lf == NULL) { 601 vput(vp); 602 if (lf == NULL) 603 error = ENODEV; 604 goto bad; 605 } 606 lf->userrefs++; 607 /* Look up again to see if the VFS was loaded. */ 608 for (vfsp = vfsconf; vfsp; vfsp = vfsp->vfc_next) 609 if (!strcmp(vfsp->vfc_name, fstype)) 610 break; 611 if (vfsp == NULL) { 612 lf->userrefs--; 613 linker_file_unload(lf); 614 vput(vp); 615 error = ENODEV; 616 goto bad; 617 } 618 } 619 mtx_lock(&vp->v_interlock); 620 if ((vp->v_flag & VMOUNT) != 0 || 621 vp->v_mountedhere != NULL) { 622 mtx_unlock(&vp->v_interlock); 623 vput(vp); 624 error = EBUSY; 625 goto bad; 626 } 627 vp->v_flag |= VMOUNT; 628 mtx_unlock(&vp->v_interlock); 629 630 /* 631 * Allocate and initialize the filesystem. 632 */ 633 mp = malloc(sizeof(struct mount), M_MOUNT, M_WAITOK | M_ZERO); 634 TAILQ_INIT(&mp->mnt_nvnodelist); 635 TAILQ_INIT(&mp->mnt_reservedvnlist); 636 lockinit(&mp->mnt_lock, PVFS, "vfslock", 0, LK_NOPAUSE); 637 (void)vfs_busy(mp, LK_NOWAIT, 0, td); 638 mp->mnt_op = vfsp->vfc_vfsops; 639 mp->mnt_vfc = vfsp; 640 vfsp->vfc_refcount++; 641 mp->mnt_stat.f_type = vfsp->vfc_typenum; 642 mp->mnt_flag |= vfsp->vfc_flags & MNT_VISFLAGMASK; 643 strncpy(mp->mnt_stat.f_fstypename, fstype, MFSNAMELEN); 644 mp->mnt_vnodecovered = vp; 645 mp->mnt_stat.f_owner = td->td_ucred->cr_uid; 646 strncpy(mp->mnt_stat.f_mntonname, fspath, MNAMELEN); 647 mp->mnt_iosize_max = DFLTPHYS; 648 #ifdef MAC 649 mac_init_mount(mp); 650 mac_create_mount(td->td_ucred, mp); 651 #endif 652 VOP_UNLOCK(vp, 0, td); 653 mp->mnt_optnew = optlist; /* XXXMAC: should this be above? */ 654 655 update: 656 /* 657 * Check if the fs implements the new VFS_NMOUNT() 658 * function, since the new system call was used. 659 */ 660 if (mp->mnt_op->vfs_mount != NULL) { 661 printf("%s doesn't support the new mount syscall\n", 662 mp->mnt_vfc->vfc_name); 663 mtx_lock(&vp->v_interlock); 664 vp->v_flag &= ~VMOUNT; 665 mtx_unlock(&vp->v_interlock); 666 if (mp->mnt_flag & MNT_UPDATE) 667 vfs_unbusy(mp, td); 668 else { 669 mp->mnt_vfc->vfc_refcount--; 670 vfs_unbusy(mp, td); 671 #ifdef MAC 672 mac_destroy_mount(mp); 673 #endif 674 free(mp, M_MOUNT); 675 } 676 vrele(vp); 677 error = EOPNOTSUPP; 678 goto bad; 679 } 680 681 /* 682 * Set the mount level flags. 683 */ 684 if (fsflags & MNT_RDONLY) 685 mp->mnt_flag |= MNT_RDONLY; 686 else if (mp->mnt_flag & MNT_RDONLY) 687 mp->mnt_kern_flag |= MNTK_WANTRDWR; 688 mp->mnt_flag &=~ MNT_UPDATEMASK; 689 mp->mnt_flag |= fsflags & (MNT_UPDATEMASK | MNT_FORCE); 690 /* 691 * Mount the filesystem. 692 * XXX The final recipients of VFS_MOUNT just overwrite the ndp they 693 * get. No freeing of cn_pnbuf. 694 */ 695 error = VFS_NMOUNT(mp, &nd, td); 696 if (!error) { 697 if (mp->mnt_opt != NULL) 698 vfs_freeopts(mp->mnt_opt); 699 mp->mnt_opt = mp->mnt_optnew; 700 } 701 /* 702 * Prevent external consumers of mount 703 * options to read mnt_optnew. 704 */ 705 mp->mnt_optnew = NULL; 706 if (mp->mnt_flag & MNT_UPDATE) { 707 if (mp->mnt_kern_flag & MNTK_WANTRDWR) 708 mp->mnt_flag &= ~MNT_RDONLY; 709 mp->mnt_flag &=~ 710 (MNT_UPDATE | MNT_RELOAD | MNT_FORCE | MNT_SNAPSHOT); 711 mp->mnt_kern_flag &=~ MNTK_WANTRDWR; 712 if (error) { 713 mp->mnt_flag = flag; 714 mp->mnt_kern_flag = kern_flag; 715 } 716 if ((mp->mnt_flag & MNT_RDONLY) == 0) { 717 if (mp->mnt_syncer == NULL) 718 error = vfs_allocate_syncvnode(mp); 719 } else { 720 if (mp->mnt_syncer != NULL) 721 vrele(mp->mnt_syncer); 722 mp->mnt_syncer = NULL; 723 } 724 vfs_unbusy(mp, td); 725 mtx_lock(&vp->v_interlock); 726 vp->v_flag &= ~VMOUNT; 727 mtx_unlock(&vp->v_interlock); 728 vrele(vp); 729 return (error); 730 } 731 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY, td); 732 /* 733 * Put the new filesystem on the mount list after root. 734 */ 735 cache_purge(vp); 736 if (!error) { 737 struct vnode *newdp; 738 739 mtx_lock(&vp->v_interlock); 740 vp->v_flag &= ~VMOUNT; 741 vp->v_mountedhere = mp; 742 mtx_unlock(&vp->v_interlock); 743 mtx_lock(&mountlist_mtx); 744 TAILQ_INSERT_TAIL(&mountlist, mp, mnt_list); 745 mtx_unlock(&mountlist_mtx); 746 if (VFS_ROOT(mp, &newdp)) 747 panic("mount: lost mount"); 748 checkdirs(vp, newdp); 749 vput(newdp); 750 VOP_UNLOCK(vp, 0, td); 751 if ((mp->mnt_flag & MNT_RDONLY) == 0) 752 error = vfs_allocate_syncvnode(mp); 753 vfs_unbusy(mp, td); 754 if ((error = VFS_START(mp, 0, td)) != 0) { 755 vrele(vp); 756 goto bad; 757 } 758 } else { 759 mtx_lock(&vp->v_interlock); 760 vp->v_flag &= ~VMOUNT; 761 mtx_unlock(&vp->v_interlock); 762 mp->mnt_vfc->vfc_refcount--; 763 vfs_unbusy(mp, td); 764 #ifdef MAC 765 mac_destroy_mount(mp); 766 #endif 767 free(mp, M_MOUNT); 768 vput(vp); 769 goto bad; 770 } 771 return (0); 772 bad: 773 vfs_freeopts(optlist); 774 return (error); 775 } 776 777 /* 778 * Old mount API. 779 */ 780 #ifndef _SYS_SYSPROTO_H_ 781 struct mount_args { 782 char *type; 783 char *path; 784 int flags; 785 caddr_t data; 786 }; 787 #endif 788 /* ARGSUSED */ 789 int 790 mount(td, uap) 791 struct thread *td; 792 struct mount_args /* { 793 syscallarg(char *) type; 794 syscallarg(char *) path; 795 syscallarg(int) flags; 796 syscallarg(caddr_t) data; 797 } */ *uap; 798 { 799 char *fstype; 800 char *fspath; 801 int error; 802 803 fstype = malloc(MFSNAMELEN, M_TEMP, M_WAITOK); 804 fspath = malloc(MNAMELEN, M_TEMP, M_WAITOK); 805 806 /* 807 * vfs_mount() actually takes a kernel string for `type' and 808 * `path' now, so extract them. 809 */ 810 error = copyinstr(SCARG(uap, type), fstype, MFSNAMELEN, NULL); 811 if (error) 812 goto finish; 813 error = copyinstr(SCARG(uap, path), fspath, MNAMELEN, NULL); 814 if (error) 815 goto finish; 816 error = vfs_mount(td, fstype, fspath, SCARG(uap, flags), 817 SCARG(uap, data)); 818 finish: 819 free(fstype, M_TEMP); 820 free(fspath, M_TEMP); 821 return (error); 822 } 823 824 /* 825 * vfs_mount(): actually attempt a filesystem mount. 826 * 827 * This routine is designed to be a "generic" entry point for routines 828 * that wish to mount a filesystem. All parameters except `fsdata' are 829 * pointers into kernel space. `fsdata' is currently still a pointer 830 * into userspace. 831 */ 832 int 833 vfs_mount(td, fstype, fspath, fsflags, fsdata) 834 struct thread *td; 835 const char *fstype; 836 char *fspath; 837 int fsflags; 838 void *fsdata; 839 { 840 linker_file_t lf; 841 struct vnode *vp; 842 struct mount *mp; 843 struct vfsconf *vfsp; 844 int error, flag = 0, kern_flag = 0; 845 struct vattr va; 846 struct nameidata nd; 847 848 /* 849 * Be ultra-paranoid about making sure the type and fspath 850 * variables will fit in our mp buffers, including the 851 * terminating NUL. 852 */ 853 if (strlen(fstype) >= MFSNAMELEN || strlen(fspath) >= MNAMELEN) 854 return (ENAMETOOLONG); 855 856 if (usermount == 0) { 857 error = suser(td); 858 if (error) 859 return (error); 860 } 861 /* 862 * Do not allow NFS export by non-root users. 863 */ 864 if (fsflags & MNT_EXPORTED) { 865 error = suser(td); 866 if (error) 867 return (error); 868 } 869 /* 870 * Silently enforce MNT_NOSUID and MNT_NODEV for non-root users. 871 */ 872 if (suser(td)) 873 fsflags |= MNT_NOSUID | MNT_NODEV; 874 /* 875 * Get vnode to be covered 876 */ 877 NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF, UIO_SYSSPACE, fspath, td); 878 if ((error = namei(&nd)) != 0) 879 return (error); 880 NDFREE(&nd, NDF_ONLY_PNBUF); 881 vp = nd.ni_vp; 882 if (fsflags & MNT_UPDATE) { 883 if ((vp->v_flag & VROOT) == 0) { 884 vput(vp); 885 return (EINVAL); 886 } 887 mp = vp->v_mount; 888 flag = mp->mnt_flag; 889 kern_flag = mp->mnt_kern_flag; 890 /* 891 * We only allow the filesystem to be reloaded if it 892 * is currently mounted read-only. 893 */ 894 if ((fsflags & MNT_RELOAD) && 895 ((mp->mnt_flag & MNT_RDONLY) == 0)) { 896 vput(vp); 897 return (EOPNOTSUPP); /* Needs translation */ 898 } 899 /* 900 * Only root, or the user that did the original mount is 901 * permitted to update it. 902 */ 903 if (mp->mnt_stat.f_owner != td->td_ucred->cr_uid) { 904 error = suser(td); 905 if (error) { 906 vput(vp); 907 return (error); 908 } 909 } 910 if (vfs_busy(mp, LK_NOWAIT, 0, td)) { 911 vput(vp); 912 return (EBUSY); 913 } 914 mtx_lock(&vp->v_interlock); 915 if ((vp->v_flag & VMOUNT) != 0 || vp->v_mountedhere != NULL) { 916 mtx_unlock(&vp->v_interlock); 917 vfs_unbusy(mp, td); 918 vput(vp); 919 return (EBUSY); 920 } 921 vp->v_flag |= VMOUNT; 922 mtx_unlock(&vp->v_interlock); 923 mp->mnt_flag |= fsflags & 924 (MNT_RELOAD | MNT_FORCE | MNT_UPDATE | MNT_SNAPSHOT); 925 VOP_UNLOCK(vp, 0, td); 926 goto update; 927 } 928 /* 929 * If the user is not root, ensure that they own the directory 930 * onto which we are attempting to mount. 931 */ 932 error = VOP_GETATTR(vp, &va, td->td_ucred, td); 933 if (error) { 934 vput(vp); 935 return (error); 936 } 937 if (va.va_uid != td->td_ucred->cr_uid) { 938 error = suser(td); 939 if (error) { 940 vput(vp); 941 return (error); 942 } 943 } 944 if ((error = vinvalbuf(vp, V_SAVE, td->td_ucred, td, 0, 0)) != 0) { 945 vput(vp); 946 return (error); 947 } 948 if (vp->v_type != VDIR) { 949 vput(vp); 950 return (ENOTDIR); 951 } 952 for (vfsp = vfsconf; vfsp; vfsp = vfsp->vfc_next) 953 if (!strcmp(vfsp->vfc_name, fstype)) 954 break; 955 if (vfsp == NULL) { 956 /* Only load modules for root (very important!). */ 957 error = suser(td); 958 if (error) { 959 vput(vp); 960 return (error); 961 } 962 error = securelevel_gt(td->td_ucred, 0); 963 if (error) { 964 vput(vp); 965 return (error); 966 } 967 error = linker_load_file(fstype, &lf); 968 if (error || lf == NULL) { 969 vput(vp); 970 if (lf == NULL) 971 error = ENODEV; 972 return (error); 973 } 974 lf->userrefs++; 975 /* Look up again to see if the VFS was loaded. */ 976 for (vfsp = vfsconf; vfsp; vfsp = vfsp->vfc_next) 977 if (!strcmp(vfsp->vfc_name, fstype)) 978 break; 979 if (vfsp == NULL) { 980 lf->userrefs--; 981 linker_file_unload(lf); 982 vput(vp); 983 return (ENODEV); 984 } 985 } 986 mtx_lock(&vp->v_interlock); 987 if ((vp->v_flag & VMOUNT) != 0 || 988 vp->v_mountedhere != NULL) { 989 mtx_unlock(&vp->v_interlock); 990 vput(vp); 991 return (EBUSY); 992 } 993 vp->v_flag |= VMOUNT; 994 mtx_unlock(&vp->v_interlock); 995 996 /* 997 * Allocate and initialize the filesystem. 998 */ 999 mp = malloc(sizeof(struct mount), M_MOUNT, M_WAITOK | M_ZERO); 1000 TAILQ_INIT(&mp->mnt_nvnodelist); 1001 TAILQ_INIT(&mp->mnt_reservedvnlist); 1002 lockinit(&mp->mnt_lock, PVFS, "vfslock", 0, LK_NOPAUSE); 1003 (void)vfs_busy(mp, LK_NOWAIT, 0, td); 1004 mp->mnt_op = vfsp->vfc_vfsops; 1005 mp->mnt_vfc = vfsp; 1006 vfsp->vfc_refcount++; 1007 mp->mnt_stat.f_type = vfsp->vfc_typenum; 1008 mp->mnt_flag |= vfsp->vfc_flags & MNT_VISFLAGMASK; 1009 strncpy(mp->mnt_stat.f_fstypename, fstype, MFSNAMELEN); 1010 mp->mnt_vnodecovered = vp; 1011 mp->mnt_stat.f_owner = td->td_ucred->cr_uid; 1012 strncpy(mp->mnt_stat.f_mntonname, fspath, MNAMELEN); 1013 mp->mnt_iosize_max = DFLTPHYS; 1014 #ifdef MAC 1015 mac_init_mount(mp); 1016 mac_create_mount(td->td_ucred, mp); 1017 #endif 1018 VOP_UNLOCK(vp, 0, td); 1019 update: 1020 /* 1021 * Check if the fs implements the old VFS_MOUNT() 1022 * function, since the old system call was used. 1023 */ 1024 if (mp->mnt_op->vfs_mount == NULL) { 1025 printf("%s doesn't support the old mount syscall\n", 1026 mp->mnt_vfc->vfc_name); 1027 mtx_lock(&vp->v_interlock); 1028 vp->v_flag &= ~VMOUNT; 1029 mtx_unlock(&vp->v_interlock); 1030 if (mp->mnt_flag & MNT_UPDATE) 1031 vfs_unbusy(mp, td); 1032 else { 1033 mp->mnt_vfc->vfc_refcount--; 1034 vfs_unbusy(mp, td); 1035 #ifdef MAC 1036 mac_destroy_mount(mp); 1037 #endif 1038 free(mp, M_MOUNT); 1039 } 1040 vrele(vp); 1041 return (EOPNOTSUPP); 1042 } 1043 1044 /* 1045 * Set the mount level flags. 1046 */ 1047 if (fsflags & MNT_RDONLY) 1048 mp->mnt_flag |= MNT_RDONLY; 1049 else if (mp->mnt_flag & MNT_RDONLY) 1050 mp->mnt_kern_flag |= MNTK_WANTRDWR; 1051 mp->mnt_flag &=~ MNT_UPDATEMASK; 1052 mp->mnt_flag |= fsflags & (MNT_UPDATEMASK | MNT_FORCE); 1053 /* 1054 * Mount the filesystem. 1055 * XXX The final recipients of VFS_MOUNT just overwrite the ndp they 1056 * get. No freeing of cn_pnbuf. 1057 */ 1058 error = VFS_MOUNT(mp, fspath, fsdata, &nd, td); 1059 if (mp->mnt_flag & MNT_UPDATE) { 1060 if (mp->mnt_kern_flag & MNTK_WANTRDWR) 1061 mp->mnt_flag &= ~MNT_RDONLY; 1062 mp->mnt_flag &=~ 1063 (MNT_UPDATE | MNT_RELOAD | MNT_FORCE | MNT_SNAPSHOT); 1064 mp->mnt_kern_flag &=~ MNTK_WANTRDWR; 1065 if (error) { 1066 mp->mnt_flag = flag; 1067 mp->mnt_kern_flag = kern_flag; 1068 } 1069 if ((mp->mnt_flag & MNT_RDONLY) == 0) { 1070 if (mp->mnt_syncer == NULL) 1071 error = vfs_allocate_syncvnode(mp); 1072 } else { 1073 if (mp->mnt_syncer != NULL) 1074 vrele(mp->mnt_syncer); 1075 mp->mnt_syncer = NULL; 1076 } 1077 vfs_unbusy(mp, td); 1078 mtx_lock(&vp->v_interlock); 1079 vp->v_flag &= ~VMOUNT; 1080 mtx_unlock(&vp->v_interlock); 1081 vrele(vp); 1082 return (error); 1083 } 1084 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY, td); 1085 /* 1086 * Put the new filesystem on the mount list after root. 1087 */ 1088 cache_purge(vp); 1089 if (!error) { 1090 struct vnode *newdp; 1091 1092 mtx_lock(&vp->v_interlock); 1093 vp->v_flag &= ~VMOUNT; 1094 vp->v_mountedhere = mp; 1095 mtx_unlock(&vp->v_interlock); 1096 mtx_lock(&mountlist_mtx); 1097 TAILQ_INSERT_TAIL(&mountlist, mp, mnt_list); 1098 mtx_unlock(&mountlist_mtx); 1099 if (VFS_ROOT(mp, &newdp)) 1100 panic("mount: lost mount"); 1101 checkdirs(vp, newdp); 1102 vput(newdp); 1103 VOP_UNLOCK(vp, 0, td); 1104 if ((mp->mnt_flag & MNT_RDONLY) == 0) 1105 error = vfs_allocate_syncvnode(mp); 1106 vfs_unbusy(mp, td); 1107 if ((error = VFS_START(mp, 0, td)) != 0) 1108 vrele(vp); 1109 } else { 1110 mtx_lock(&vp->v_interlock); 1111 vp->v_flag &= ~VMOUNT; 1112 mtx_unlock(&vp->v_interlock); 1113 mp->mnt_vfc->vfc_refcount--; 1114 vfs_unbusy(mp, td); 1115 #ifdef MAC 1116 mac_destroy_mount(mp); 1117 #endif 1118 free(mp, M_MOUNT); 1119 vput(vp); 1120 } 1121 return (error); 1122 } 1123 1124 /* 1125 * Scan all active processes to see if any of them have a current 1126 * or root directory of `olddp'. If so, replace them with the new 1127 * mount point. 1128 */ 1129 static void 1130 checkdirs(olddp, newdp) 1131 struct vnode *olddp, *newdp; 1132 { 1133 struct filedesc *fdp; 1134 struct proc *p; 1135 int nrele; 1136 1137 if (olddp->v_usecount == 1) 1138 return; 1139 sx_slock(&allproc_lock); 1140 LIST_FOREACH(p, &allproc, p_list) { 1141 PROC_LOCK(p); 1142 fdp = p->p_fd; 1143 if (fdp == NULL) { 1144 PROC_UNLOCK(p); 1145 continue; 1146 } 1147 nrele = 0; 1148 FILEDESC_LOCK(fdp); 1149 if (fdp->fd_cdir == olddp) { 1150 VREF(newdp); 1151 fdp->fd_cdir = newdp; 1152 nrele++; 1153 } 1154 if (fdp->fd_rdir == olddp) { 1155 VREF(newdp); 1156 fdp->fd_rdir = newdp; 1157 nrele++; 1158 } 1159 FILEDESC_UNLOCK(fdp); 1160 PROC_UNLOCK(p); 1161 while (nrele--) 1162 vrele(olddp); 1163 } 1164 sx_sunlock(&allproc_lock); 1165 if (rootvnode == olddp) { 1166 vrele(rootvnode); 1167 VREF(newdp); 1168 rootvnode = newdp; 1169 } 1170 } 1171 1172 /* 1173 * Unmount a filesystem. 1174 * 1175 * Note: unmount takes a path to the vnode mounted on as argument, 1176 * not special file (as before). 1177 */ 1178 #ifndef _SYS_SYSPROTO_H_ 1179 struct unmount_args { 1180 char *path; 1181 int flags; 1182 }; 1183 #endif 1184 /* ARGSUSED */ 1185 int 1186 unmount(td, uap) 1187 struct thread *td; 1188 register struct unmount_args /* { 1189 syscallarg(char *) path; 1190 syscallarg(int) flags; 1191 } */ *uap; 1192 { 1193 register struct vnode *vp; 1194 struct mount *mp; 1195 int error; 1196 struct nameidata nd; 1197 1198 NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF, UIO_USERSPACE, 1199 SCARG(uap, path), td); 1200 if ((error = namei(&nd)) != 0) 1201 return (error); 1202 vp = nd.ni_vp; 1203 NDFREE(&nd, NDF_ONLY_PNBUF); 1204 mp = vp->v_mount; 1205 1206 /* 1207 * Only root, or the user that did the original mount is 1208 * permitted to unmount this filesystem. 1209 */ 1210 if (mp->mnt_stat.f_owner != td->td_ucred->cr_uid) { 1211 error = suser(td); 1212 if (error) { 1213 vput(vp); 1214 return (error); 1215 } 1216 } 1217 1218 /* 1219 * Don't allow unmounting the root filesystem. 1220 */ 1221 if (mp->mnt_flag & MNT_ROOTFS) { 1222 vput(vp); 1223 return (EINVAL); 1224 } 1225 1226 /* 1227 * Must be the root of the filesystem 1228 */ 1229 if ((vp->v_flag & VROOT) == 0) { 1230 vput(vp); 1231 return (EINVAL); 1232 } 1233 vput(vp); 1234 return (dounmount(mp, SCARG(uap, flags), td)); 1235 } 1236 1237 /* 1238 * Do the actual filesystem unmount. 1239 */ 1240 int 1241 dounmount(mp, flags, td) 1242 struct mount *mp; 1243 int flags; 1244 struct thread *td; 1245 { 1246 struct vnode *coveredvp, *fsrootvp; 1247 int error; 1248 int async_flag; 1249 1250 mtx_lock(&mountlist_mtx); 1251 if (mp->mnt_kern_flag & MNTK_UNMOUNT) { 1252 mtx_unlock(&mountlist_mtx); 1253 return (EBUSY); 1254 } 1255 mp->mnt_kern_flag |= MNTK_UNMOUNT; 1256 /* Allow filesystems to detect that a forced unmount is in progress. */ 1257 if (flags & MNT_FORCE) 1258 mp->mnt_kern_flag |= MNTK_UNMOUNTF; 1259 error = lockmgr(&mp->mnt_lock, LK_DRAIN | LK_INTERLOCK | 1260 ((flags & MNT_FORCE) ? 0 : LK_NOWAIT), &mountlist_mtx, td); 1261 if (error) { 1262 mp->mnt_kern_flag &= ~(MNTK_UNMOUNT | MNTK_UNMOUNTF); 1263 if (mp->mnt_kern_flag & MNTK_MWAIT) 1264 wakeup(mp); 1265 return (error); 1266 } 1267 vn_start_write(NULL, &mp, V_WAIT); 1268 1269 if (mp->mnt_flag & MNT_EXPUBLIC) 1270 vfs_setpublicfs(NULL, NULL, NULL); 1271 1272 vfs_msync(mp, MNT_WAIT); 1273 async_flag = mp->mnt_flag & MNT_ASYNC; 1274 mp->mnt_flag &=~ MNT_ASYNC; 1275 cache_purgevfs(mp); /* remove cache entries for this file sys */ 1276 if (mp->mnt_syncer != NULL) 1277 vrele(mp->mnt_syncer); 1278 /* Move process cdir/rdir refs on fs root to underlying vnode. */ 1279 if (VFS_ROOT(mp, &fsrootvp) == 0) { 1280 if (mp->mnt_vnodecovered != NULL) 1281 checkdirs(fsrootvp, mp->mnt_vnodecovered); 1282 if (fsrootvp == rootvnode) { 1283 vrele(rootvnode); 1284 rootvnode = NULL; 1285 } 1286 vput(fsrootvp); 1287 } 1288 if (((mp->mnt_flag & MNT_RDONLY) || 1289 (error = VFS_SYNC(mp, MNT_WAIT, td->td_ucred, td)) == 0) || 1290 (flags & MNT_FORCE)) { 1291 error = VFS_UNMOUNT(mp, flags, td); 1292 } 1293 vn_finished_write(mp); 1294 if (error) { 1295 /* Undo cdir/rdir and rootvnode changes made above. */ 1296 if (VFS_ROOT(mp, &fsrootvp) == 0) { 1297 if (mp->mnt_vnodecovered != NULL) 1298 checkdirs(mp->mnt_vnodecovered, fsrootvp); 1299 if (rootvnode == NULL) { 1300 rootvnode = fsrootvp; 1301 vref(rootvnode); 1302 } 1303 vput(fsrootvp); 1304 } 1305 if ((mp->mnt_flag & MNT_RDONLY) == 0 && mp->mnt_syncer == NULL) 1306 (void) vfs_allocate_syncvnode(mp); 1307 mtx_lock(&mountlist_mtx); 1308 mp->mnt_kern_flag &= ~(MNTK_UNMOUNT | MNTK_UNMOUNTF); 1309 mp->mnt_flag |= async_flag; 1310 lockmgr(&mp->mnt_lock, LK_RELEASE | LK_INTERLOCK, 1311 &mountlist_mtx, td); 1312 if (mp->mnt_kern_flag & MNTK_MWAIT) 1313 wakeup(mp); 1314 return (error); 1315 } 1316 mtx_lock(&mountlist_mtx); 1317 TAILQ_REMOVE(&mountlist, mp, mnt_list); 1318 if ((coveredvp = mp->mnt_vnodecovered) != NULL) 1319 coveredvp->v_mountedhere = NULL; 1320 mp->mnt_vfc->vfc_refcount--; 1321 if (!TAILQ_EMPTY(&mp->mnt_nvnodelist)) 1322 panic("unmount: dangling vnode"); 1323 lockmgr(&mp->mnt_lock, LK_RELEASE | LK_INTERLOCK, &mountlist_mtx, td); 1324 lockdestroy(&mp->mnt_lock); 1325 if (coveredvp != NULL) 1326 vrele(coveredvp); 1327 if (mp->mnt_kern_flag & MNTK_MWAIT) 1328 wakeup(mp); 1329 #ifdef MAC 1330 mac_destroy_mount(mp); 1331 #endif 1332 if (mp->mnt_op->vfs_mount == NULL) 1333 vfs_freeopts(mp->mnt_opt); 1334 free(mp, M_MOUNT); 1335 return (0); 1336 } 1337 1338 /* 1339 * Lookup a filesystem type, and if found allocate and initialize 1340 * a mount structure for it. 1341 * 1342 * Devname is usually updated by mount(8) after booting. 1343 */ 1344 int 1345 vfs_rootmountalloc(fstypename, devname, mpp) 1346 char *fstypename; 1347 char *devname; 1348 struct mount **mpp; 1349 { 1350 struct thread *td = curthread; /* XXX */ 1351 struct vfsconf *vfsp; 1352 struct mount *mp; 1353 1354 if (fstypename == NULL) 1355 return (ENODEV); 1356 for (vfsp = vfsconf; vfsp; vfsp = vfsp->vfc_next) 1357 if (!strcmp(vfsp->vfc_name, fstypename)) 1358 break; 1359 if (vfsp == NULL) 1360 return (ENODEV); 1361 mp = malloc((u_long)sizeof(struct mount), M_MOUNT, M_WAITOK | M_ZERO); 1362 lockinit(&mp->mnt_lock, PVFS, "vfslock", 0, LK_NOPAUSE); 1363 (void)vfs_busy(mp, LK_NOWAIT, 0, td); 1364 TAILQ_INIT(&mp->mnt_nvnodelist); 1365 TAILQ_INIT(&mp->mnt_reservedvnlist); 1366 mp->mnt_vfc = vfsp; 1367 mp->mnt_op = vfsp->vfc_vfsops; 1368 mp->mnt_flag = MNT_RDONLY; 1369 mp->mnt_vnodecovered = NULLVP; 1370 vfsp->vfc_refcount++; 1371 mp->mnt_iosize_max = DFLTPHYS; 1372 mp->mnt_stat.f_type = vfsp->vfc_typenum; 1373 mp->mnt_flag |= vfsp->vfc_flags & MNT_VISFLAGMASK; 1374 strncpy(mp->mnt_stat.f_fstypename, vfsp->vfc_name, MFSNAMELEN); 1375 mp->mnt_stat.f_mntonname[0] = '/'; 1376 mp->mnt_stat.f_mntonname[1] = 0; 1377 (void) copystr(devname, mp->mnt_stat.f_mntfromname, MNAMELEN - 1, 0); 1378 #ifdef MAC 1379 mac_init_mount(mp); 1380 mac_create_mount(td->td_ucred, mp); 1381 #endif 1382 *mpp = mp; 1383 return (0); 1384 } 1385 1386 /* 1387 * Find and mount the root filesystem 1388 */ 1389 void 1390 vfs_mountroot(void) 1391 { 1392 char *cp; 1393 int i, error; 1394 1395 /* 1396 * The root filesystem information is compiled in, and we are 1397 * booted with instructions to use it. 1398 */ 1399 #ifdef ROOTDEVNAME 1400 if ((boothowto & RB_DFLTROOT) && 1401 !vfs_mountroot_try(ROOTDEVNAME)) 1402 return; 1403 #endif 1404 /* 1405 * We are booted with instructions to prompt for the root filesystem, 1406 * or to use the compiled-in default when it doesn't exist. 1407 */ 1408 if (boothowto & (RB_DFLTROOT | RB_ASKNAME)) { 1409 if (!vfs_mountroot_ask()) 1410 return; 1411 } 1412 1413 /* 1414 * We've been given the generic "use CDROM as root" flag. This is 1415 * necessary because one media may be used in many different 1416 * devices, so we need to search for them. 1417 */ 1418 if (boothowto & RB_CDROM) { 1419 for (i = 0; cdrom_rootdevnames[i] != NULL; i++) { 1420 if (!vfs_mountroot_try(cdrom_rootdevnames[i])) 1421 return; 1422 } 1423 } 1424 1425 /* 1426 * Try to use the value read by the loader from /etc/fstab, or 1427 * supplied via some other means. This is the preferred 1428 * mechanism. 1429 */ 1430 if ((cp = getenv("vfs.root.mountfrom")) != NULL) { 1431 error = vfs_mountroot_try(cp); 1432 freeenv(cp); 1433 if (!error) 1434 return; 1435 } 1436 1437 /* 1438 * Try values that may have been computed by the machine-dependant 1439 * legacy code. 1440 */ 1441 if (!vfs_mountroot_try(rootdevnames[0])) 1442 return; 1443 if (!vfs_mountroot_try(rootdevnames[1])) 1444 return; 1445 1446 /* 1447 * If we have a compiled-in default, and haven't already tried it, try 1448 * it now. 1449 */ 1450 #ifdef ROOTDEVNAME 1451 if (!(boothowto & RB_DFLTROOT)) 1452 if (!vfs_mountroot_try(ROOTDEVNAME)) 1453 return; 1454 #endif 1455 1456 /* 1457 * Everything so far has failed, prompt on the console if we haven't 1458 * already tried that. 1459 */ 1460 if (!(boothowto & (RB_DFLTROOT | RB_ASKNAME)) && !vfs_mountroot_ask()) 1461 return; 1462 panic("Root mount failed, startup aborted."); 1463 } 1464 1465 /* 1466 * Mount (mountfrom) as the root filesystem. 1467 */ 1468 static int 1469 vfs_mountroot_try(char *mountfrom) 1470 { 1471 struct mount *mp; 1472 char *vfsname, *path; 1473 int error; 1474 char patt[32]; 1475 int s; 1476 1477 vfsname = NULL; 1478 path = NULL; 1479 mp = NULL; 1480 error = EINVAL; 1481 1482 if (mountfrom == NULL) 1483 return(error); /* don't complain */ 1484 1485 s = splcam(); /* Overkill, but annoying without it */ 1486 printf("Mounting root from %s\n", mountfrom); 1487 splx(s); 1488 1489 /* parse vfs name and path */ 1490 vfsname = malloc(MFSNAMELEN, M_MOUNT, M_WAITOK); 1491 path = malloc(MNAMELEN, M_MOUNT, M_WAITOK); 1492 vfsname[0] = path[0] = 0; 1493 sprintf(patt, "%%%d[a-z0-9]:%%%ds", MFSNAMELEN, MNAMELEN); 1494 if (sscanf(mountfrom, patt, vfsname, path) < 1) 1495 goto done; 1496 1497 /* allocate a root mount */ 1498 error = vfs_rootmountalloc(vfsname, path[0] != 0 ? path : ROOTNAME, 1499 &mp); 1500 if (error != 0) { 1501 printf("Can't allocate root mount for filesystem '%s': %d\n", 1502 vfsname, error); 1503 goto done; 1504 } 1505 mp->mnt_flag |= MNT_ROOTFS; 1506 1507 /* do our best to set rootdev */ 1508 if ((path[0] != 0) && setrootbyname(path)) 1509 printf("setrootbyname failed\n"); 1510 1511 /* If the root device is a type "memory disk", mount RW */ 1512 if (rootdev != NODEV && devsw(rootdev) && 1513 (devsw(rootdev)->d_flags & D_MEMDISK)) 1514 mp->mnt_flag &= ~MNT_RDONLY; 1515 1516 /* 1517 * Set the mount path to be something useful, because the 1518 * filesystem code isn't responsible now for initialising 1519 * f_mntonname unless they want to override the default 1520 * (which is `path'.) 1521 */ 1522 strncpy(mp->mnt_stat.f_mntonname, "/", MNAMELEN); 1523 1524 error = VFS_MOUNT(mp, NULL, NULL, NULL, curthread); 1525 1526 done: 1527 if (vfsname != NULL) 1528 free(vfsname, M_MOUNT); 1529 if (path != NULL) 1530 free(path, M_MOUNT); 1531 if (error != 0) { 1532 if (mp != NULL) { 1533 vfs_unbusy(mp, curthread); 1534 #ifdef MAC 1535 mac_destroy_mount(mp); 1536 #endif 1537 free(mp, M_MOUNT); 1538 } 1539 printf("Root mount failed: %d\n", error); 1540 } else { 1541 1542 /* register with list of mounted filesystems */ 1543 mtx_lock(&mountlist_mtx); 1544 TAILQ_INSERT_HEAD(&mountlist, mp, mnt_list); 1545 mtx_unlock(&mountlist_mtx); 1546 1547 /* sanity check system clock against root fs timestamp */ 1548 inittodr(mp->mnt_time); 1549 vfs_unbusy(mp, curthread); 1550 error = VFS_START(mp, 0, curthread); 1551 } 1552 return(error); 1553 } 1554 1555 /* 1556 * Spin prompting on the console for a suitable root filesystem 1557 */ 1558 static int 1559 vfs_mountroot_ask(void) 1560 { 1561 char name[128]; 1562 int i; 1563 dev_t dev; 1564 1565 for(;;) { 1566 printf("\nManual root filesystem specification:\n"); 1567 printf(" <fstype>:<device> Mount <device> using filesystem <fstype>\n"); 1568 #if defined(__i386__) || defined(__ia64__) 1569 printf(" eg. ufs:da0s1a\n"); 1570 #else 1571 printf(" eg. ufs:da0a\n"); 1572 #endif 1573 printf(" ? List valid disk boot devices\n"); 1574 printf(" <empty line> Abort manual input\n"); 1575 printf("\nmountroot> "); 1576 gets(name); 1577 if (name[0] == 0) 1578 return(1); 1579 if (name[0] == '?') { 1580 printf("Possibly valid devices for 'ufs' root:\n"); 1581 for (i = 0; i < NUMCDEVSW; i++) { 1582 dev = makedev(i, 0); 1583 if (devsw(dev) != NULL) 1584 printf(" \"%s\"", devsw(dev)->d_name); 1585 } 1586 printf("\n"); 1587 continue; 1588 } 1589 if (!vfs_mountroot_try(name)) 1590 return(0); 1591 } 1592 } 1593 1594 /* 1595 * Local helper function for vfs_mountroot_ask. 1596 */ 1597 static void 1598 gets(char *cp) 1599 { 1600 char *lp; 1601 int c; 1602 1603 lp = cp; 1604 for (;;) { 1605 printf("%c", c = cngetc() & 0177); 1606 switch (c) { 1607 case -1: 1608 case '\n': 1609 case '\r': 1610 *lp++ = '\0'; 1611 return; 1612 case '\b': 1613 case '\177': 1614 if (lp > cp) { 1615 printf(" \b"); 1616 lp--; 1617 } 1618 continue; 1619 case '#': 1620 lp--; 1621 if (lp < cp) 1622 lp = cp; 1623 continue; 1624 case '@': 1625 case 'u' & 037: 1626 lp = cp; 1627 printf("%c", '\n'); 1628 continue; 1629 default: 1630 *lp++ = c; 1631 } 1632 } 1633 } 1634 1635 /* 1636 * Convert a given name to the dev_t of the disk-like device 1637 * it refers to. 1638 */ 1639 dev_t 1640 getdiskbyname(char *name) { 1641 char *cp; 1642 dev_t dev; 1643 1644 cp = name; 1645 if (!bcmp(cp, "/dev/", 5)) 1646 cp += 5; 1647 1648 dev = NODEV; 1649 EVENTHANDLER_INVOKE(dev_clone, cp, strlen(cp), &dev); 1650 return (dev); 1651 } 1652 1653 /* 1654 * Set rootdev to match (name), given that we expect it to 1655 * refer to a disk-like device. 1656 */ 1657 static int 1658 setrootbyname(char *name) 1659 { 1660 dev_t diskdev; 1661 1662 diskdev = getdiskbyname(name); 1663 if (diskdev != NODEV) { 1664 rootdev = diskdev; 1665 return (0); 1666 } 1667 1668 return (1); 1669 } 1670 1671 /* Show the dev_t for a disk specified by name */ 1672 #ifdef DDB 1673 DB_SHOW_COMMAND(disk, db_getdiskbyname) 1674 { 1675 dev_t dev; 1676 1677 if (modif[0] == '\0') { 1678 db_error("usage: show disk/devicename"); 1679 return; 1680 } 1681 dev = getdiskbyname(modif); 1682 if (dev != NODEV) 1683 db_printf("dev_t = %p\n", dev); 1684 else 1685 db_printf("No disk device matched.\n"); 1686 } 1687 #endif 1688 1689 /* 1690 * Get a mount option by its name. 1691 * 1692 * Return 0 if the option was found, ENOENT otherwise. 1693 * If len is non-NULL it will be filled with the length 1694 * of the option. If buf is non-NULL, it will be filled 1695 * with the address of the option. 1696 */ 1697 int 1698 vfs_getopt(opts, name, buf, len) 1699 struct vfsoptlist *opts; 1700 const char *name; 1701 void **buf; 1702 int *len; 1703 { 1704 struct vfsopt *opt; 1705 1706 TAILQ_FOREACH(opt, opts, link) { 1707 if (strcmp(name, opt->name) == 0) { 1708 if (len != NULL) 1709 *len = opt->len; 1710 if (buf != NULL) 1711 *buf = opt->value; 1712 return (0); 1713 } 1714 } 1715 return (ENOENT); 1716 } 1717 1718 /* 1719 * Find and copy a mount option. 1720 * 1721 * The size of the buffer has to be specified 1722 * in len, if it is not the same length as the 1723 * mount option, EINVAL is returned. 1724 * Returns ENOENT if the option is not found. 1725 */ 1726 int 1727 vfs_copyopt(opts, name, dest, len) 1728 struct vfsoptlist *opts; 1729 const char *name; 1730 void *dest; 1731 int len; 1732 { 1733 struct vfsopt *opt; 1734 1735 TAILQ_FOREACH(opt, opts, link) { 1736 if (strcmp(name, opt->name) == 0) { 1737 if (len != opt->len) 1738 return (EINVAL); 1739 bcopy(opt->value, dest, opt->len); 1740 return (0); 1741 } 1742 } 1743 return (ENOENT); 1744 } 1745