vfs_mount.c revision 127473
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 65#include <sys/cdefs.h> 66__FBSDID("$FreeBSD: head/sys/kern/vfs_mount.c 127473 2004-03-27 08:09:00Z pjd $"); 67 68#include <sys/param.h> 69#include <sys/conf.h> 70#include <sys/cons.h> 71#include <sys/jail.h> 72#include <sys/kernel.h> 73#include <sys/linker.h> 74#include <sys/mac.h> 75#include <sys/malloc.h> 76#include <sys/mount.h> 77#include <sys/mutex.h> 78#include <sys/namei.h> 79#include <sys/proc.h> 80#include <sys/filedesc.h> 81#include <sys/reboot.h> 82#include <sys/sysproto.h> 83#include <sys/sx.h> 84#include <sys/sysctl.h> 85#include <sys/sysent.h> 86#include <sys/systm.h> 87#include <sys/vnode.h> 88 89#include <geom/geom.h> 90 91#include <machine/stdarg.h> 92 93#include "opt_rootdevname.h" 94#include "opt_ddb.h" 95#include "opt_mac.h" 96 97#ifdef DDB 98#include <ddb/ddb.h> 99#endif 100 101#define ROOTNAME "root_device" 102#define VFS_MOUNTARG_SIZE_MAX (1024 * 64) 103 104static void checkdirs(struct vnode *olddp, struct vnode *newdp); 105static void gets(char *cp); 106static int vfs_domount(struct thread *td, const char *fstype, 107 char *fspath, int fsflags, void *fsdata, int compat); 108static int vfs_mount_alloc(struct vnode *dvp, struct vfsconf *vfsp, 109 const char *fspath, struct thread *td, struct mount **mpp); 110static int vfs_mountroot_ask(void); 111static int vfs_mountroot_try(char *mountfrom); 112 113static int usermount = 0; /* if 1, non-root can mount fs. */ 114SYSCTL_INT(_vfs, OID_AUTO, usermount, CTLFLAG_RW, &usermount, 0, ""); 115 116MALLOC_DEFINE(M_MOUNT, "mount", "vfs mount structure"); 117 118/* List of mounted filesystems. */ 119struct mntlist mountlist = TAILQ_HEAD_INITIALIZER(mountlist); 120 121/* For any iteration/modification of mountlist */ 122struct mtx mountlist_mtx; 123 124/* 125 * The vnode of the system's root (/ in the filesystem, without chroot 126 * active.) 127 */ 128struct vnode *rootvnode; 129 130/* 131 * The root filesystem is detailed in the kernel environment variable 132 * vfs.root.mountfrom, which is expected to be in the general format 133 * 134 * <vfsname>:[<path>] 135 * vfsname := the name of a VFS known to the kernel and capable 136 * of being mounted as root 137 * path := disk device name or other data used by the filesystem 138 * to locate its physical store 139 */ 140 141/* 142 * The root specifiers we will try if RB_CDROM is specified. 143 */ 144static char *cdrom_rootdevnames[] = { 145 "cd9660:cd0", 146 "cd9660:acd0", 147 NULL 148}; 149 150/* legacy find-root code */ 151char *rootdevnames[2] = {NULL, NULL}; 152static int setrootbyname(char *name); 153dev_t rootdev = NODEV; 154 155/* 156 * Has to be dynamic as the value of rootdev can change; however, it can't 157 * change after the root is mounted, so a user process can't access this 158 * sysctl until after the value is unchangeable. 159 */ 160static int 161sysctl_rootdev(SYSCTL_HANDLER_ARGS) 162{ 163 int error; 164 165 /* _RD prevents this from happening. */ 166 KASSERT(req->newptr == NULL, ("Attempt to change root device name")); 167 168 if (rootdev != NODEV) 169 error = sysctl_handle_string(oidp, rootdev->si_name, 0, req); 170 else 171 error = sysctl_handle_string(oidp, "", 0, req); 172 173 return (error); 174} 175 176SYSCTL_PROC(_kern, OID_AUTO, rootdev, CTLTYPE_STRING | CTLFLAG_RD, 177 0, 0, sysctl_rootdev, "A", "Root file system device"); 178 179/* Remove one mount option. */ 180static void 181vfs_freeopt(struct vfsoptlist *opts, struct vfsopt *opt) 182{ 183 184 TAILQ_REMOVE(opts, opt, link); 185 free(opt->name, M_MOUNT); 186 if (opt->value != NULL) 187 free(opt->value, M_MOUNT); 188#ifdef INVARIANTS 189 else if (opt->len != 0) 190 panic("%s: mount option with NULL value but length != 0", 191 __func__); 192#endif 193 free(opt, M_MOUNT); 194} 195 196/* Release all resources related to the mount options. */ 197static void 198vfs_freeopts(struct vfsoptlist *opts) 199{ 200 struct vfsopt *opt; 201 202 while (!TAILQ_EMPTY(opts)) { 203 opt = TAILQ_FIRST(opts); 204 vfs_freeopt(opts, opt); 205 } 206 free(opts, M_MOUNT); 207} 208 209/* 210 * If a mount option is specified several times, 211 * (with or without the "no" prefix) only keep 212 * the last occurence of it. 213 */ 214static void 215vfs_sanitizeopts(struct vfsoptlist *opts) 216{ 217 struct vfsopt *opt, *opt2, *tmp; 218 int noopt; 219 220 TAILQ_FOREACH_REVERSE(opt, opts, vfsoptlist, link) { 221 if (strncmp(opt->name, "no", 2) == 0) 222 noopt = 1; 223 else 224 noopt = 0; 225 opt2 = TAILQ_PREV(opt, vfsoptlist, link); 226 while (opt2 != NULL) { 227 if (strcmp(opt2->name, opt->name) == 0 || 228 (noopt && strcmp(opt->name + 2, opt2->name) == 0) || 229 (!noopt && strncmp(opt2->name, "no", 2) == 0 && 230 strcmp(opt2->name + 2, opt->name) == 0)) { 231 tmp = TAILQ_PREV(opt2, vfsoptlist, link); 232 vfs_freeopt(opts, opt2); 233 opt2 = tmp; 234 } else { 235 opt2 = TAILQ_PREV(opt2, vfsoptlist, link); 236 } 237 } 238 } 239} 240 241/* 242 * Build a linked list of mount options from a struct uio. 243 */ 244static int 245vfs_buildopts(struct uio *auio, struct vfsoptlist **options) 246{ 247 struct vfsoptlist *opts; 248 struct vfsopt *opt; 249 size_t memused; 250 unsigned int i, iovcnt; 251 int error, namelen, optlen; 252 253 opts = malloc(sizeof(struct vfsoptlist), M_MOUNT, M_WAITOK); 254 TAILQ_INIT(opts); 255 memused = 0; 256 iovcnt = auio->uio_iovcnt; 257 for (i = 0; i < iovcnt; i += 2) { 258 opt = malloc(sizeof(struct vfsopt), M_MOUNT, M_WAITOK); 259 namelen = auio->uio_iov[i].iov_len; 260 optlen = auio->uio_iov[i + 1].iov_len; 261 opt->name = malloc(namelen, M_MOUNT, M_WAITOK); 262 opt->value = NULL; 263 opt->len = optlen; 264 265 /* 266 * Do this early, so jumps to "bad" will free the current 267 * option. 268 */ 269 TAILQ_INSERT_TAIL(opts, opt, link); 270 memused += sizeof(struct vfsopt) + optlen + namelen; 271 272 /* 273 * Avoid consuming too much memory, and attempts to overflow 274 * memused. 275 */ 276 if (memused > VFS_MOUNTARG_SIZE_MAX || 277 optlen > VFS_MOUNTARG_SIZE_MAX || 278 namelen > VFS_MOUNTARG_SIZE_MAX) { 279 error = EINVAL; 280 goto bad; 281 } 282 283 if (auio->uio_segflg == UIO_SYSSPACE) { 284 bcopy(auio->uio_iov[i].iov_base, opt->name, namelen); 285 } else { 286 error = copyin(auio->uio_iov[i].iov_base, opt->name, 287 namelen); 288 if (error) 289 goto bad; 290 } 291 /* Ensure names are null-terminated strings. */ 292 if (opt->name[namelen - 1] != '\0') { 293 error = EINVAL; 294 goto bad; 295 } 296 if (optlen != 0) { 297 opt->value = malloc(optlen, M_MOUNT, M_WAITOK); 298 if (auio->uio_segflg == UIO_SYSSPACE) { 299 bcopy(auio->uio_iov[i + 1].iov_base, opt->value, 300 optlen); 301 } else { 302 error = copyin(auio->uio_iov[i + 1].iov_base, 303 opt->value, optlen); 304 if (error) 305 goto bad; 306 } 307 } 308 } 309 vfs_sanitizeopts(opts); 310 *options = opts; 311 return (0); 312bad: 313 vfs_freeopts(opts); 314 return (error); 315} 316 317/* 318 * Merge the old mount options with the new ones passed 319 * in the MNT_UPDATE case. 320 */ 321static void 322vfs_mergeopts(struct vfsoptlist *toopts, struct vfsoptlist *opts) 323{ 324 struct vfsopt *opt, *opt2, *new; 325 326 TAILQ_FOREACH(opt, opts, link) { 327 /* 328 * Check that this option hasn't been redefined 329 * nor cancelled with a "no" mount option. 330 */ 331 opt2 = TAILQ_FIRST(toopts); 332 while (opt2 != NULL) { 333 if (strcmp(opt2->name, opt->name) == 0) 334 goto next; 335 if (strncmp(opt2->name, "no", 2) == 0 && 336 strcmp(opt2->name + 2, opt->name) == 0) { 337 vfs_freeopt(toopts, opt2); 338 goto next; 339 } 340 opt2 = TAILQ_NEXT(opt2, link); 341 } 342 /* We want this option, duplicate it. */ 343 new = malloc(sizeof(struct vfsopt), M_MOUNT, M_WAITOK); 344 new->name = malloc(strlen(opt->name) + 1, M_MOUNT, M_WAITOK); 345 strcpy(new->name, opt->name); 346 if (opt->len != 0) { 347 new->value = malloc(opt->len, M_MOUNT, M_WAITOK); 348 bcopy(opt->value, new->value, opt->len); 349 } else { 350 new->value = NULL; 351 } 352 new->len = opt->len; 353 TAILQ_INSERT_TAIL(toopts, new, link); 354next: 355 continue; 356 } 357} 358 359/* 360 * New mount API. 361 */ 362int 363nmount(td, uap) 364 struct thread *td; 365 struct nmount_args /* { 366 struct iovec *iovp; 367 unsigned int iovcnt; 368 int flags; 369 } */ *uap; 370{ 371 struct uio auio; 372 struct iovec *iov, *needfree; 373 struct iovec aiov[UIO_SMALLIOV]; 374 unsigned int i; 375 int error; 376 u_int iovlen, iovcnt; 377 378 iovcnt = uap->iovcnt; 379 iovlen = iovcnt * sizeof (struct iovec); 380 /* 381 * Check that we have an even number of iovec's 382 * and that we have at least two options. 383 */ 384 if ((iovcnt & 1) || (iovcnt < 4) || (iovcnt > UIO_MAXIOV)) 385 return (EINVAL); 386 387 if (iovcnt > UIO_SMALLIOV) { 388 MALLOC(iov, struct iovec *, iovlen, M_IOV, M_WAITOK); 389 needfree = iov; 390 } else { 391 iov = aiov; 392 needfree = NULL; 393 } 394 auio.uio_iov = iov; 395 auio.uio_iovcnt = iovcnt; 396 auio.uio_segflg = UIO_USERSPACE; 397 if ((error = copyin(uap->iovp, iov, iovlen))) 398 goto finish; 399 400 for (i = 0; i < iovcnt; i++) { 401 if (iov->iov_len > MMAXOPTIONLEN) { 402 error = EINVAL; 403 goto finish; 404 } 405 iov++; 406 } 407 error = vfs_nmount(td, uap->flags, &auio); 408finish: 409 if (needfree != NULL) 410 free(needfree, M_TEMP); 411 return (error); 412} 413 414int 415kernel_mount(iovp, iovcnt, flags) 416 struct iovec *iovp; 417 unsigned int iovcnt; 418 int flags; 419{ 420 struct uio auio; 421 int error; 422 423 /* 424 * Check that we have an even number of iovec's 425 * and that we have at least two options. 426 */ 427 if ((iovcnt & 1) || (iovcnt < 4)) 428 return (EINVAL); 429 430 auio.uio_iov = iovp; 431 auio.uio_iovcnt = iovcnt; 432 auio.uio_segflg = UIO_SYSSPACE; 433 434 error = vfs_nmount(curthread, flags, &auio); 435 return (error); 436} 437 438int 439kernel_vmount(int flags, ...) 440{ 441 struct iovec *iovp; 442 struct uio auio; 443 va_list ap; 444 unsigned int iovcnt, iovlen, len; 445 const char *cp; 446 char *buf, *pos; 447 size_t n; 448 int error, i; 449 450 len = 0; 451 va_start(ap, flags); 452 for (iovcnt = 0; (cp = va_arg(ap, const char *)) != NULL; iovcnt++) 453 len += strlen(cp) + 1; 454 va_end(ap); 455 456 if (iovcnt < 4 || iovcnt & 1) 457 return (EINVAL); 458 459 iovlen = iovcnt * sizeof (struct iovec); 460 MALLOC(iovp, struct iovec *, iovlen, M_MOUNT, M_WAITOK); 461 MALLOC(buf, char *, len, M_MOUNT, M_WAITOK); 462 pos = buf; 463 va_start(ap, flags); 464 for (i = 0; i < iovcnt; i++) { 465 cp = va_arg(ap, const char *); 466 copystr(cp, pos, len - (pos - buf), &n); 467 iovp[i].iov_base = pos; 468 iovp[i].iov_len = n; 469 pos += n; 470 } 471 va_end(ap); 472 473 auio.uio_iov = iovp; 474 auio.uio_iovcnt = iovcnt; 475 auio.uio_segflg = UIO_SYSSPACE; 476 477 error = vfs_nmount(curthread, flags, &auio); 478 FREE(iovp, M_MOUNT); 479 FREE(buf, M_MOUNT); 480 return (error); 481} 482 483/* 484 * Allocate and initialize the mount point struct. 485 */ 486static int 487vfs_mount_alloc(struct vnode *vp, struct vfsconf *vfsp, 488 const char *fspath, struct thread *td, struct mount **mpp) 489{ 490 struct mount *mp; 491 492 mp = malloc(sizeof(struct mount), M_MOUNT, M_WAITOK | M_ZERO); 493 TAILQ_INIT(&mp->mnt_nvnodelist); 494 mp->mnt_nvnodelistsize = 0; 495 mtx_init(&mp->mnt_mtx, "struct mount mtx", NULL, MTX_DEF); 496 lockinit(&mp->mnt_lock, PVFS, "vfslock", 0, LK_NOPAUSE); 497 vfs_busy(mp, LK_NOWAIT, 0, td); 498 mp->mnt_op = vfsp->vfc_vfsops; 499 mp->mnt_vfc = vfsp; 500 vfsp->vfc_refcount++; 501 mp->mnt_stat.f_type = vfsp->vfc_typenum; 502 mp->mnt_flag |= vfsp->vfc_flags & MNT_VISFLAGMASK; 503 strlcpy(mp->mnt_stat.f_fstypename, vfsp->vfc_name, MFSNAMELEN); 504 mp->mnt_vnodecovered = vp; 505 mp->mnt_cred = crdup(td->td_ucred); 506 mp->mnt_stat.f_owner = td->td_ucred->cr_uid; 507 strlcpy(mp->mnt_stat.f_mntonname, fspath, MNAMELEN); 508 mp->mnt_iosize_max = DFLTPHYS; 509#ifdef MAC 510 mac_init_mount(mp); 511 mac_create_mount(td->td_ucred, mp); 512#endif 513 *mpp = mp; 514 return (0); 515} 516 517/* 518 * Destroy the mount struct previously allocated by vfs_mount_alloc(). 519 */ 520void 521vfs_mount_destroy(struct mount *mp, struct thread *td) 522{ 523 524 mp->mnt_vfc->vfc_refcount--; 525 if (!TAILQ_EMPTY(&mp->mnt_nvnodelist)) 526 panic("unmount: dangling vnode"); 527 vfs_unbusy(mp,td); 528 lockdestroy(&mp->mnt_lock); 529 mtx_destroy(&mp->mnt_mtx); 530 if (mp->mnt_kern_flag & MNTK_MWAIT) 531 wakeup(mp); 532#ifdef MAC 533 mac_destroy_mount(mp); 534#endif 535 if (mp->mnt_opt != NULL) 536 vfs_freeopts(mp->mnt_opt); 537 crfree(mp->mnt_cred); 538 free(mp, M_MOUNT); 539} 540 541int 542vfs_nmount(struct thread *td, int fsflags, struct uio *fsoptions) 543{ 544 struct vfsoptlist *optlist; 545 char *fstype, *fspath; 546 int error, fstypelen, fspathlen; 547 548 error = vfs_buildopts(fsoptions, &optlist); 549 if (error) 550 return (error); 551 552 /* 553 * We need these two options before the others, 554 * and they are mandatory for any filesystem. 555 * Ensure they are NUL terminated as well. 556 */ 557 fstypelen = 0; 558 error = vfs_getopt(optlist, "fstype", (void **)&fstype, &fstypelen); 559 if (error || fstype[fstypelen - 1] != '\0') { 560 error = EINVAL; 561 goto bail; 562 } 563 fspathlen = 0; 564 error = vfs_getopt(optlist, "fspath", (void **)&fspath, &fspathlen); 565 if (error || fspath[fspathlen - 1] != '\0') { 566 error = EINVAL; 567 goto bail; 568 } 569 570 /* 571 * Be ultra-paranoid about making sure the type and fspath 572 * variables will fit in our mp buffers, including the 573 * terminating NUL. 574 */ 575 if (fstypelen >= MFSNAMELEN - 1 || fspathlen >= MNAMELEN - 1) { 576 error = ENAMETOOLONG; 577 goto bail; 578 } 579 580 error = vfs_domount(td, fstype, fspath, fsflags, optlist, 0); 581bail: 582 if (error) 583 vfs_freeopts(optlist); 584 return (error); 585} 586 587/* 588 * Old mount API. 589 */ 590#ifndef _SYS_SYSPROTO_H_ 591struct mount_args { 592 char *type; 593 char *path; 594 int flags; 595 caddr_t data; 596}; 597#endif 598/* ARGSUSED */ 599int 600mount(td, uap) 601 struct thread *td; 602 struct mount_args /* { 603 char *type; 604 char *path; 605 int flags; 606 caddr_t data; 607 } */ *uap; 608{ 609 char *fstype; 610 char *fspath; 611 int error; 612 613 fstype = malloc(MFSNAMELEN, M_TEMP, M_WAITOK); 614 fspath = malloc(MNAMELEN, M_TEMP, M_WAITOK); 615 616 /* 617 * vfs_mount() actually takes a kernel string for `type' and 618 * `path' now, so extract them. 619 */ 620 error = copyinstr(uap->type, fstype, MFSNAMELEN, NULL); 621 if (error == 0) 622 error = copyinstr(uap->path, fspath, MNAMELEN, NULL); 623 if (error == 0) 624 error = vfs_domount(td, fstype, fspath, uap->flags, 625 uap->data, 1); 626 free(fstype, M_TEMP); 627 free(fspath, M_TEMP); 628 return (error); 629} 630 631/* 632 * vfs_mount(): actually attempt a filesystem mount. 633 * 634 * This routine is designed to be a "generic" entry point for routines 635 * that wish to mount a filesystem. All parameters except `fsdata' are 636 * pointers into kernel space. `fsdata' is currently still a pointer 637 * into userspace. 638 */ 639int 640vfs_mount(td, fstype, fspath, fsflags, fsdata) 641 struct thread *td; 642 const char *fstype; 643 char *fspath; 644 int fsflags; 645 void *fsdata; 646{ 647 648 return (vfs_domount(td, fstype, fspath, fsflags, fsdata, 1)); 649} 650 651/* 652 * vfs_domount(): actually attempt a filesystem mount. 653 */ 654static int 655vfs_domount( 656 struct thread *td, /* Flags common to all filesystems. */ 657 const char *fstype, /* Filesystem type. */ 658 char *fspath, /* Mount path. */ 659 int fsflags, /* Flags common to all filesystems. */ 660 void *fsdata, /* Options local to the filesystem. */ 661 int compat /* Invocation from compat syscall. */ 662 ) 663{ 664 linker_file_t lf; 665 struct vnode *vp; 666 struct mount *mp; 667 struct vfsconf *vfsp; 668 int error, flag = 0, kern_flag = 0; 669 struct vattr va; 670 struct nameidata nd; 671 672 /* 673 * Be ultra-paranoid about making sure the type and fspath 674 * variables will fit in our mp buffers, including the 675 * terminating NUL. 676 */ 677 if (strlen(fstype) >= MFSNAMELEN || strlen(fspath) >= MNAMELEN) 678 return (ENAMETOOLONG); 679 680 /* mount(2) is not permitted inside the jail. */ 681 if (jailed(td->td_ucred)) 682 return (EPERM); 683 684 if (usermount == 0) { 685 error = suser(td); 686 if (error) 687 return (error); 688 } 689 690 /* 691 * Do not allow NFS export or MNT_SUIDDIR by unprivileged users. 692 */ 693 if (fsflags & (MNT_EXPORTED | MNT_SUIDDIR)) { 694 error = suser(td); 695 if (error) 696 return (error); 697 } 698 /* 699 * Silently enforce MNT_NODEV, MNT_NOSUID and MNT_USER for 700 * unprivileged users. 701 */ 702 if (suser(td) != 0) 703 fsflags |= MNT_NODEV | MNT_NOSUID | MNT_USER; 704 /* 705 * Get vnode to be covered 706 */ 707 NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF, UIO_SYSSPACE, fspath, td); 708 if ((error = namei(&nd)) != 0) 709 return (error); 710 NDFREE(&nd, NDF_ONLY_PNBUF); 711 vp = nd.ni_vp; 712 if (fsflags & MNT_UPDATE) { 713 if ((vp->v_vflag & VV_ROOT) == 0) { 714 vput(vp); 715 return (EINVAL); 716 } 717 mp = vp->v_mount; 718 flag = mp->mnt_flag; 719 kern_flag = mp->mnt_kern_flag; 720 /* 721 * We only allow the filesystem to be reloaded if it 722 * is currently mounted read-only. 723 */ 724 if ((fsflags & MNT_RELOAD) && 725 ((mp->mnt_flag & MNT_RDONLY) == 0)) { 726 vput(vp); 727 return (EOPNOTSUPP); /* Needs translation */ 728 } 729 /* 730 * Only root, or the user that did the original mount is 731 * permitted to update it. 732 */ 733 if ((mp->mnt_flag & MNT_USER) != 0) { 734 if (mp->mnt_cred->cr_uid != td->td_ucred->cr_uid) { 735 if ((error = suser(td)) != 0) { 736 vput(vp); 737 return (error); 738 } 739 } 740 } else { 741 if ((error = suser(td)) != 0) { 742 vput(vp); 743 return (error); 744 } 745 } 746 if (vfs_busy(mp, LK_NOWAIT, 0, td)) { 747 vput(vp); 748 return (EBUSY); 749 } 750 VI_LOCK(vp); 751 if ((vp->v_iflag & VI_MOUNT) != 0 || 752 vp->v_mountedhere != NULL) { 753 VI_UNLOCK(vp); 754 vfs_unbusy(mp, td); 755 vput(vp); 756 return (EBUSY); 757 } 758 vp->v_iflag |= VI_MOUNT; 759 VI_UNLOCK(vp); 760 mp->mnt_flag |= fsflags & 761 (MNT_RELOAD | MNT_FORCE | MNT_UPDATE | MNT_SNAPSHOT); 762 VOP_UNLOCK(vp, 0, td); 763 if (compat == 0) { 764 mp->mnt_optnew = fsdata; 765 vfs_mergeopts(mp->mnt_optnew, mp->mnt_opt); 766 } 767 goto update; 768 } 769 /* 770 * If the user is not root, ensure that they own the directory 771 * onto which we are attempting to mount. 772 */ 773 error = VOP_GETATTR(vp, &va, td->td_ucred, td); 774 if (error) { 775 vput(vp); 776 return (error); 777 } 778 if (va.va_uid != td->td_ucred->cr_uid) { 779 error = suser(td); 780 if (error) { 781 vput(vp); 782 return (error); 783 } 784 } 785 if ((error = vinvalbuf(vp, V_SAVE, td->td_ucred, td, 0, 0)) != 0) { 786 vput(vp); 787 return (error); 788 } 789 if (vp->v_type != VDIR) { 790 vput(vp); 791 return (ENOTDIR); 792 } 793 for (vfsp = vfsconf; vfsp; vfsp = vfsp->vfc_next) 794 if (!strcmp(vfsp->vfc_name, fstype)) 795 break; 796 if (vfsp == NULL) { 797 /* Only load modules for root (very important!). */ 798 error = suser(td); 799 if (error) { 800 vput(vp); 801 return (error); 802 } 803 error = securelevel_gt(td->td_ucred, 0); 804 if (error) { 805 vput(vp); 806 return (error); 807 } 808 error = linker_load_module(NULL, fstype, NULL, NULL, &lf); 809 if (error || lf == NULL) { 810 vput(vp); 811 if (lf == NULL) 812 error = ENODEV; 813 return (error); 814 } 815 lf->userrefs++; 816 /* Look up again to see if the VFS was loaded. */ 817 for (vfsp = vfsconf; vfsp; vfsp = vfsp->vfc_next) 818 if (!strcmp(vfsp->vfc_name, fstype)) 819 break; 820 if (vfsp == NULL) { 821 lf->userrefs--; 822 linker_file_unload(lf); 823 vput(vp); 824 return (ENODEV); 825 } 826 } 827 VI_LOCK(vp); 828 if ((vp->v_iflag & VI_MOUNT) != 0 || 829 vp->v_mountedhere != NULL) { 830 VI_UNLOCK(vp); 831 vput(vp); 832 return (EBUSY); 833 } 834 vp->v_iflag |= VI_MOUNT; 835 VI_UNLOCK(vp); 836 837 /* 838 * Allocate and initialize the filesystem. 839 */ 840 error = vfs_mount_alloc(vp, vfsp, fspath, td, &mp); 841 if (error) { 842 vput(vp); 843 return (error); 844 } 845 VOP_UNLOCK(vp, 0, td); 846 847 /* XXXMAC: pass to vfs_mount_alloc? */ 848 if (compat == 0) 849 mp->mnt_optnew = fsdata; 850update: 851 /* 852 * Check if the fs implements the type VFS_[N]MOUNT() 853 * function we are looking for. 854 */ 855 if ((compat == 0) == (mp->mnt_op->vfs_mount != NULL)) { 856 printf("%s doesn't support the %s mount syscall\n", 857 mp->mnt_vfc->vfc_name, compat? "old" : "new"); 858 VI_LOCK(vp); 859 vp->v_iflag &= ~VI_MOUNT; 860 VI_UNLOCK(vp); 861 if (mp->mnt_flag & MNT_UPDATE) 862 vfs_unbusy(mp, td); 863 else 864 vfs_mount_destroy(mp, td); 865 vrele(vp); 866 return (EOPNOTSUPP); 867 } 868 869 /* 870 * Set the mount level flags. 871 */ 872 if (fsflags & MNT_RDONLY) 873 mp->mnt_flag |= MNT_RDONLY; 874 else if (mp->mnt_flag & MNT_RDONLY) 875 mp->mnt_kern_flag |= MNTK_WANTRDWR; 876 mp->mnt_flag &=~ MNT_UPDATEMASK; 877 mp->mnt_flag |= fsflags & (MNT_UPDATEMASK | MNT_FORCE); 878 /* 879 * Mount the filesystem. 880 * XXX The final recipients of VFS_MOUNT just overwrite the ndp they 881 * get. No freeing of cn_pnbuf. 882 */ 883 error = compat? VFS_MOUNT(mp, fspath, fsdata, &nd, td) : 884 VFS_NMOUNT(mp, &nd, td); 885 if (!error) { 886 if (mp->mnt_opt != NULL) 887 vfs_freeopts(mp->mnt_opt); 888 mp->mnt_opt = mp->mnt_optnew; 889 } 890 /* 891 * Prevent external consumers of mount options from reading 892 * mnt_optnew. 893 */ 894 mp->mnt_optnew = NULL; 895 if (mp->mnt_flag & MNT_UPDATE) { 896 if (mp->mnt_kern_flag & MNTK_WANTRDWR) 897 mp->mnt_flag &= ~MNT_RDONLY; 898 mp->mnt_flag &=~ 899 (MNT_UPDATE | MNT_RELOAD | MNT_FORCE | MNT_SNAPSHOT); 900 mp->mnt_kern_flag &=~ MNTK_WANTRDWR; 901 if (error) { 902 mp->mnt_flag = flag; 903 mp->mnt_kern_flag = kern_flag; 904 } 905 if ((mp->mnt_flag & MNT_RDONLY) == 0) { 906 if (mp->mnt_syncer == NULL) 907 error = vfs_allocate_syncvnode(mp); 908 } else { 909 if (mp->mnt_syncer != NULL) 910 vrele(mp->mnt_syncer); 911 mp->mnt_syncer = NULL; 912 } 913 vfs_unbusy(mp, td); 914 VI_LOCK(vp); 915 vp->v_iflag &= ~VI_MOUNT; 916 VI_UNLOCK(vp); 917 vrele(vp); 918 return (error); 919 } 920 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY, td); 921 /* 922 * Put the new filesystem on the mount list after root. 923 */ 924 cache_purge(vp); 925 if (!error) { 926 struct vnode *newdp; 927 928 VI_LOCK(vp); 929 vp->v_iflag &= ~VI_MOUNT; 930 VI_UNLOCK(vp); 931 vp->v_mountedhere = mp; 932 mtx_lock(&mountlist_mtx); 933 TAILQ_INSERT_TAIL(&mountlist, mp, mnt_list); 934 mtx_unlock(&mountlist_mtx); 935 if (VFS_ROOT(mp, &newdp)) 936 panic("mount: lost mount"); 937 checkdirs(vp, newdp); 938 vput(newdp); 939 VOP_UNLOCK(vp, 0, td); 940 if ((mp->mnt_flag & MNT_RDONLY) == 0) 941 error = vfs_allocate_syncvnode(mp); 942 vfs_unbusy(mp, td); 943 if (error || (error = VFS_START(mp, 0, td)) != 0) 944 vrele(vp); 945 } else { 946 VI_LOCK(vp); 947 vp->v_iflag &= ~VI_MOUNT; 948 VI_UNLOCK(vp); 949 vfs_mount_destroy(mp, td); 950 vput(vp); 951 } 952 return (error); 953} 954 955/* 956 * Scan all active processes to see if any of them have a current 957 * or root directory of `olddp'. If so, replace them with the new 958 * mount point. 959 */ 960static void 961checkdirs(olddp, newdp) 962 struct vnode *olddp, *newdp; 963{ 964 struct filedesc *fdp; 965 struct proc *p; 966 int nrele; 967 968 if (vrefcnt(olddp) == 1) 969 return; 970 sx_slock(&allproc_lock); 971 LIST_FOREACH(p, &allproc, p_list) { 972 mtx_lock(&fdesc_mtx); 973 fdp = p->p_fd; 974 if (fdp == NULL) { 975 mtx_unlock(&fdesc_mtx); 976 continue; 977 } 978 nrele = 0; 979 FILEDESC_LOCK(fdp); 980 if (fdp->fd_cdir == olddp) { 981 VREF(newdp); 982 fdp->fd_cdir = newdp; 983 nrele++; 984 } 985 if (fdp->fd_rdir == olddp) { 986 VREF(newdp); 987 fdp->fd_rdir = newdp; 988 nrele++; 989 } 990 FILEDESC_UNLOCK(fdp); 991 mtx_unlock(&fdesc_mtx); 992 while (nrele--) 993 vrele(olddp); 994 } 995 sx_sunlock(&allproc_lock); 996 if (rootvnode == olddp) { 997 vrele(rootvnode); 998 VREF(newdp); 999 rootvnode = newdp; 1000 } 1001} 1002 1003/* 1004 * Unmount a filesystem. 1005 * 1006 * Note: unmount takes a path to the vnode mounted on as argument, 1007 * not special file (as before). 1008 */ 1009#ifndef _SYS_SYSPROTO_H_ 1010struct unmount_args { 1011 char *path; 1012 int flags; 1013}; 1014#endif 1015/* ARGSUSED */ 1016int 1017unmount(td, uap) 1018 struct thread *td; 1019 register struct unmount_args /* { 1020 char *path; 1021 int flags; 1022 } */ *uap; 1023{ 1024 struct mount *mp; 1025 char *pathbuf; 1026 int error, id0, id1; 1027 1028 /* unmount(2) is not permitted inside the jail. */ 1029 if (jailed(td->td_ucred)) 1030 return (EPERM); 1031 1032 if (usermount == 0) { 1033 if ((error = suser(td)) != 0) 1034 return (error); 1035 } 1036 1037 pathbuf = malloc(MNAMELEN, M_TEMP, M_WAITOK); 1038 error = copyinstr(uap->path, pathbuf, MNAMELEN, NULL); 1039 if (error) { 1040 free(pathbuf, M_TEMP); 1041 return (error); 1042 } 1043 if (uap->flags & MNT_BYFSID) { 1044 /* Decode the filesystem ID. */ 1045 if (sscanf(pathbuf, "FSID:%d:%d", &id0, &id1) != 2) { 1046 free(pathbuf, M_TEMP); 1047 return (EINVAL); 1048 } 1049 1050 mtx_lock(&mountlist_mtx); 1051 TAILQ_FOREACH_REVERSE(mp, &mountlist, mntlist, mnt_list) 1052 if (mp->mnt_stat.f_fsid.val[0] == id0 && 1053 mp->mnt_stat.f_fsid.val[1] == id1) 1054 break; 1055 mtx_unlock(&mountlist_mtx); 1056 } else { 1057 mtx_lock(&mountlist_mtx); 1058 TAILQ_FOREACH_REVERSE(mp, &mountlist, mntlist, mnt_list) 1059 if (strcmp(mp->mnt_stat.f_mntonname, pathbuf) == 0) 1060 break; 1061 mtx_unlock(&mountlist_mtx); 1062 } 1063 free(pathbuf, M_TEMP); 1064 if (mp == NULL) { 1065 /* 1066 * Previously we returned ENOENT for a nonexistent path and 1067 * EINVAL for a non-mountpoint. We cannot tell these apart 1068 * now, so in the !MNT_BYFSID case return the more likely 1069 * EINVAL for compatibility. 1070 */ 1071 return ((uap->flags & MNT_BYFSID) ? ENOENT : EINVAL); 1072 } 1073 1074 /* 1075 * Only root, or the user that did the original mount is 1076 * permitted to unmount this filesystem. 1077 */ 1078 if ((mp->mnt_flag & MNT_USER) != 0) { 1079 if (mp->mnt_cred->cr_uid != td->td_ucred->cr_uid) { 1080 if ((error = suser(td)) != 0) 1081 return (error); 1082 } 1083 } else { 1084 if ((error = suser(td)) != 0) 1085 return (error); 1086 } 1087 1088 /* 1089 * Don't allow unmounting the root filesystem. 1090 */ 1091 if (mp->mnt_flag & MNT_ROOTFS) 1092 return (EINVAL); 1093 return (dounmount(mp, uap->flags, td)); 1094} 1095 1096/* 1097 * Do the actual filesystem unmount. 1098 */ 1099int 1100dounmount(mp, flags, td) 1101 struct mount *mp; 1102 int flags; 1103 struct thread *td; 1104{ 1105 struct vnode *coveredvp, *fsrootvp; 1106 int error; 1107 int async_flag; 1108 1109 mtx_lock(&mountlist_mtx); 1110 if (mp->mnt_kern_flag & MNTK_UNMOUNT) { 1111 mtx_unlock(&mountlist_mtx); 1112 return (EBUSY); 1113 } 1114 mp->mnt_kern_flag |= MNTK_UNMOUNT; 1115 /* Allow filesystems to detect that a forced unmount is in progress. */ 1116 if (flags & MNT_FORCE) 1117 mp->mnt_kern_flag |= MNTK_UNMOUNTF; 1118 error = lockmgr(&mp->mnt_lock, LK_DRAIN | LK_INTERLOCK | 1119 ((flags & MNT_FORCE) ? 0 : LK_NOWAIT), &mountlist_mtx, td); 1120 if (error) { 1121 mp->mnt_kern_flag &= ~(MNTK_UNMOUNT | MNTK_UNMOUNTF); 1122 if (mp->mnt_kern_flag & MNTK_MWAIT) 1123 wakeup(mp); 1124 return (error); 1125 } 1126 vn_start_write(NULL, &mp, V_WAIT); 1127 1128 if (mp->mnt_flag & MNT_EXPUBLIC) 1129 vfs_setpublicfs(NULL, NULL, NULL); 1130 1131 vfs_msync(mp, MNT_WAIT); 1132 async_flag = mp->mnt_flag & MNT_ASYNC; 1133 mp->mnt_flag &=~ MNT_ASYNC; 1134 cache_purgevfs(mp); /* remove cache entries for this file sys */ 1135 if (mp->mnt_syncer != NULL) 1136 vrele(mp->mnt_syncer); 1137 /* 1138 * For forced unmounts, move process cdir/rdir refs on the fs root 1139 * vnode to the covered vnode. For non-forced unmounts we want 1140 * such references to cause an EBUSY error. 1141 */ 1142 if ((flags & MNT_FORCE) && VFS_ROOT(mp, &fsrootvp) == 0) { 1143 if (mp->mnt_vnodecovered != NULL) 1144 checkdirs(fsrootvp, mp->mnt_vnodecovered); 1145 if (fsrootvp == rootvnode) { 1146 vrele(rootvnode); 1147 rootvnode = NULL; 1148 } 1149 vput(fsrootvp); 1150 } 1151 if (((mp->mnt_flag & MNT_RDONLY) || 1152 (error = VFS_SYNC(mp, MNT_WAIT, td->td_ucred, td)) == 0) || 1153 (flags & MNT_FORCE)) { 1154 error = VFS_UNMOUNT(mp, flags, td); 1155 } 1156 vn_finished_write(mp); 1157 if (error) { 1158 /* Undo cdir/rdir and rootvnode changes made above. */ 1159 if ((flags & MNT_FORCE) && VFS_ROOT(mp, &fsrootvp) == 0) { 1160 if (mp->mnt_vnodecovered != NULL) 1161 checkdirs(mp->mnt_vnodecovered, fsrootvp); 1162 if (rootvnode == NULL) { 1163 rootvnode = fsrootvp; 1164 vref(rootvnode); 1165 } 1166 vput(fsrootvp); 1167 } 1168 if ((mp->mnt_flag & MNT_RDONLY) == 0 && mp->mnt_syncer == NULL) 1169 (void) vfs_allocate_syncvnode(mp); 1170 mtx_lock(&mountlist_mtx); 1171 mp->mnt_kern_flag &= ~(MNTK_UNMOUNT | MNTK_UNMOUNTF); 1172 mp->mnt_flag |= async_flag; 1173 lockmgr(&mp->mnt_lock, LK_RELEASE | LK_INTERLOCK, 1174 &mountlist_mtx, td); 1175 if (mp->mnt_kern_flag & MNTK_MWAIT) 1176 wakeup(mp); 1177 return (error); 1178 } 1179 mtx_lock(&mountlist_mtx); 1180 TAILQ_REMOVE(&mountlist, mp, mnt_list); 1181 if ((coveredvp = mp->mnt_vnodecovered) != NULL) 1182 coveredvp->v_mountedhere = NULL; 1183 mtx_unlock(&mountlist_mtx); 1184 vfs_mount_destroy(mp, td); 1185 if (coveredvp != NULL) 1186 vrele(coveredvp); 1187 return (0); 1188} 1189 1190/* 1191 * Lookup a filesystem type, and if found allocate and initialize 1192 * a mount structure for it. 1193 * 1194 * Devname is usually updated by mount(8) after booting. 1195 */ 1196int 1197vfs_rootmountalloc(fstypename, devname, mpp) 1198 char *fstypename; 1199 char *devname; 1200 struct mount **mpp; 1201{ 1202 struct thread *td = curthread; /* XXX */ 1203 struct vfsconf *vfsp; 1204 struct mount *mp; 1205 int error; 1206 1207 if (fstypename == NULL) 1208 return (ENODEV); 1209 for (vfsp = vfsconf; vfsp; vfsp = vfsp->vfc_next) 1210 if (!strcmp(vfsp->vfc_name, fstypename)) 1211 break; 1212 if (vfsp == NULL) 1213 return (ENODEV); 1214 error = vfs_mount_alloc(NULLVP, vfsp, "/", td, &mp); 1215 if (error) 1216 return (error); 1217 mp->mnt_flag |= MNT_RDONLY | MNT_ROOTFS; 1218 strlcpy(mp->mnt_stat.f_mntfromname, devname, MNAMELEN); 1219 *mpp = mp; 1220 return (0); 1221} 1222 1223/* 1224 * Find and mount the root filesystem 1225 */ 1226void 1227vfs_mountroot(void) 1228{ 1229 char *cp; 1230 int error, i; 1231 1232 g_waitidle(); 1233 1234 /* 1235 * The root filesystem information is compiled in, and we are 1236 * booted with instructions to use it. 1237 */ 1238#ifdef ROOTDEVNAME 1239 if ((boothowto & RB_DFLTROOT) && !vfs_mountroot_try(ROOTDEVNAME)) 1240 return; 1241#endif 1242 /* 1243 * We are booted with instructions to prompt for the root filesystem, 1244 * or to use the compiled-in default when it doesn't exist. 1245 */ 1246 if (boothowto & (RB_DFLTROOT | RB_ASKNAME)) { 1247 if (!vfs_mountroot_ask()) 1248 return; 1249 } 1250 1251 /* 1252 * We've been given the generic "use CDROM as root" flag. This is 1253 * necessary because one media may be used in many different 1254 * devices, so we need to search for them. 1255 */ 1256 if (boothowto & RB_CDROM) { 1257 for (i = 0; cdrom_rootdevnames[i] != NULL; i++) { 1258 if (!vfs_mountroot_try(cdrom_rootdevnames[i])) 1259 return; 1260 } 1261 } 1262 1263 /* 1264 * Try to use the value read by the loader from /etc/fstab, or 1265 * supplied via some other means. This is the preferred 1266 * mechanism. 1267 */ 1268 if ((cp = getenv("vfs.root.mountfrom")) != NULL) { 1269 error = vfs_mountroot_try(cp); 1270 freeenv(cp); 1271 if (!error) 1272 return; 1273 } 1274 1275 /* 1276 * Try values that may have been computed by the machine-dependant 1277 * legacy code. 1278 */ 1279 if (!vfs_mountroot_try(rootdevnames[0])) 1280 return; 1281 if (!vfs_mountroot_try(rootdevnames[1])) 1282 return; 1283 1284 /* 1285 * If we have a compiled-in default, and haven't already tried it, try 1286 * it now. 1287 */ 1288#ifdef ROOTDEVNAME 1289 if (!(boothowto & RB_DFLTROOT)) 1290 if (!vfs_mountroot_try(ROOTDEVNAME)) 1291 return; 1292#endif 1293 1294 /* 1295 * Everything so far has failed, prompt on the console if we haven't 1296 * already tried that. 1297 */ 1298 if (!(boothowto & (RB_DFLTROOT | RB_ASKNAME)) && !vfs_mountroot_ask()) 1299 return; 1300 panic("Root mount failed, startup aborted."); 1301} 1302 1303/* 1304 * Mount (mountfrom) as the root filesystem. 1305 */ 1306static int 1307vfs_mountroot_try(char *mountfrom) 1308{ 1309 struct mount *mp; 1310 char *vfsname, *path; 1311 const char *devname; 1312 int error; 1313 char patt[32]; 1314 int s; 1315 1316 vfsname = NULL; 1317 path = NULL; 1318 mp = NULL; 1319 error = EINVAL; 1320 1321 if (mountfrom == NULL) 1322 return(error); /* don't complain */ 1323 1324 s = splcam(); /* Overkill, but annoying without it */ 1325 printf("Mounting root from %s\n", mountfrom); 1326 splx(s); 1327 1328 /* parse vfs name and path */ 1329 vfsname = malloc(MFSNAMELEN, M_MOUNT, M_WAITOK); 1330 path = malloc(MNAMELEN, M_MOUNT, M_WAITOK); 1331 vfsname[0] = path[0] = 0; 1332 sprintf(patt, "%%%d[a-z0-9]:%%%ds", MFSNAMELEN, MNAMELEN); 1333 if (sscanf(mountfrom, patt, vfsname, path) < 1) 1334 goto done; 1335 1336 /* allocate a root mount */ 1337 error = vfs_rootmountalloc(vfsname, path[0] != 0 ? path : ROOTNAME, 1338 &mp); 1339 if (error != 0) { 1340 printf("Can't allocate root mount for filesystem '%s': %d\n", 1341 vfsname, error); 1342 goto done; 1343 } 1344 1345 /* do our best to set rootdev */ 1346 if ((path[0] != 0) && setrootbyname(path)) 1347 printf("setrootbyname failed\n"); 1348 1349 /* If the root device is a type "memory disk", mount RW */ 1350 if (rootdev != NODEV && devsw(rootdev) != NULL) { 1351 devname = devtoname(rootdev); 1352 if (devname[0] == 'm' && devname[1] == 'd') 1353 mp->mnt_flag &= ~MNT_RDONLY; 1354 } 1355 1356 error = VFS_MOUNT(mp, NULL, NULL, NULL, curthread); 1357 1358done: 1359 if (vfsname != NULL) 1360 free(vfsname, M_MOUNT); 1361 if (path != NULL) 1362 free(path, M_MOUNT); 1363 if (error != 0) { 1364 if (mp != NULL) 1365 vfs_mount_destroy(mp, curthread); 1366 printf("Root mount failed: %d\n", error); 1367 } else { 1368 1369 /* register with list of mounted filesystems */ 1370 mtx_lock(&mountlist_mtx); 1371 TAILQ_INSERT_HEAD(&mountlist, mp, mnt_list); 1372 mtx_unlock(&mountlist_mtx); 1373 1374 /* sanity check system clock against root fs timestamp */ 1375 inittodr(mp->mnt_time); 1376 vfs_unbusy(mp, curthread); 1377 error = VFS_START(mp, 0, curthread); 1378 } 1379 return(error); 1380} 1381 1382/* 1383 * Spin prompting on the console for a suitable root filesystem 1384 */ 1385static int 1386vfs_mountroot_ask(void) 1387{ 1388 char name[128]; 1389 1390 for(;;) { 1391 printf("\nManual root filesystem specification:\n"); 1392 printf(" <fstype>:<device> Mount <device> using filesystem <fstype>\n"); 1393#if defined(__i386__) || defined(__ia64__) 1394 printf(" eg. ufs:da0s1a\n"); 1395#else 1396 printf(" eg. ufs:/dev/da0a\n"); 1397#endif 1398 printf(" ? List valid disk boot devices\n"); 1399 printf(" <empty line> Abort manual input\n"); 1400 printf("\nmountroot> "); 1401 gets(name); 1402 if (name[0] == 0) 1403 return(1); 1404 if (name[0] == '?') { 1405 printf("\nList of GEOM managed disk devices:\n "); 1406 g_dev_print(); 1407 continue; 1408 } 1409 if (!vfs_mountroot_try(name)) 1410 return(0); 1411 } 1412} 1413 1414/* 1415 * Local helper function for vfs_mountroot_ask. 1416 */ 1417static void 1418gets(char *cp) 1419{ 1420 char *lp; 1421 int c; 1422 1423 lp = cp; 1424 for (;;) { 1425 printf("%c", c = cngetc() & 0177); 1426 switch (c) { 1427 case -1: 1428 case '\n': 1429 case '\r': 1430 *lp++ = '\0'; 1431 return; 1432 case '\b': 1433 case '\177': 1434 if (lp > cp) { 1435 printf(" \b"); 1436 lp--; 1437 } 1438 continue; 1439 case '#': 1440 lp--; 1441 if (lp < cp) 1442 lp = cp; 1443 continue; 1444 case '@': 1445 case 'u' & 037: 1446 lp = cp; 1447 printf("%c", '\n'); 1448 continue; 1449 default: 1450 *lp++ = c; 1451 } 1452 } 1453} 1454 1455/* 1456 * Convert a given name to the dev_t of the disk-like device 1457 * it refers to. 1458 */ 1459dev_t 1460getdiskbyname(char *name) { 1461 char *cp; 1462 dev_t dev; 1463 1464 cp = name; 1465 if (!bcmp(cp, "/dev/", 5)) 1466 cp += 5; 1467 1468 dev = NODEV; 1469 EVENTHANDLER_INVOKE(dev_clone, cp, strlen(cp), &dev); 1470 return (dev); 1471} 1472 1473/* 1474 * Set rootdev to match (name), given that we expect it to 1475 * refer to a disk-like device. 1476 */ 1477static int 1478setrootbyname(char *name) 1479{ 1480 dev_t diskdev; 1481 1482 diskdev = getdiskbyname(name); 1483 if (diskdev != NODEV) { 1484 rootdev = diskdev; 1485 return (0); 1486 } 1487 1488 return (1); 1489} 1490 1491/* Show the dev_t for a disk specified by name */ 1492#ifdef DDB 1493DB_SHOW_COMMAND(disk, db_getdiskbyname) 1494{ 1495 dev_t dev; 1496 1497 if (modif[0] == '\0') { 1498 db_error("usage: show disk/devicename"); 1499 return; 1500 } 1501 dev = getdiskbyname(modif); 1502 if (dev != NODEV) 1503 db_printf("dev_t = %p\n", dev); 1504 else 1505 db_printf("No disk device matched.\n"); 1506} 1507#endif 1508 1509/* 1510 * Get a mount option by its name. 1511 * 1512 * Return 0 if the option was found, ENOENT otherwise. 1513 * If len is non-NULL it will be filled with the length 1514 * of the option. If buf is non-NULL, it will be filled 1515 * with the address of the option. 1516 */ 1517int 1518vfs_getopt(opts, name, buf, len) 1519 struct vfsoptlist *opts; 1520 const char *name; 1521 void **buf; 1522 int *len; 1523{ 1524 struct vfsopt *opt; 1525 1526 KASSERT(opts != NULL, ("vfs_getopt: caller passed 'opts' as NULL")); 1527 1528 TAILQ_FOREACH(opt, opts, link) { 1529 if (strcmp(name, opt->name) == 0) { 1530 if (len != NULL) 1531 *len = opt->len; 1532 if (buf != NULL) 1533 *buf = opt->value; 1534 return (0); 1535 } 1536 } 1537 return (ENOENT); 1538} 1539 1540/* 1541 * Find and copy a mount option. 1542 * 1543 * The size of the buffer has to be specified 1544 * in len, if it is not the same length as the 1545 * mount option, EINVAL is returned. 1546 * Returns ENOENT if the option is not found. 1547 */ 1548int 1549vfs_copyopt(opts, name, dest, len) 1550 struct vfsoptlist *opts; 1551 const char *name; 1552 void *dest; 1553 int len; 1554{ 1555 struct vfsopt *opt; 1556 1557 KASSERT(opts != NULL, ("vfs_copyopt: caller passed 'opts' as NULL")); 1558 1559 TAILQ_FOREACH(opt, opts, link) { 1560 if (strcmp(name, opt->name) == 0) { 1561 if (len != opt->len) 1562 return (EINVAL); 1563 bcopy(opt->value, dest, opt->len); 1564 return (0); 1565 } 1566 } 1567 return (ENOENT); 1568} 1569