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