vfs_mountroot.c revision 287964
1/*- 2 * Copyright (c) 2010 Marcel Moolenaar 3 * Copyright (c) 1999-2004 Poul-Henning Kamp 4 * Copyright (c) 1999 Michael Smith 5 * Copyright (c) 1989, 1993 6 * The Regents of the University of California. All rights reserved. 7 * (c) UNIX System Laboratories, Inc. 8 * All or some portions of this file are derived from material licensed 9 * to the University of California by American Telephone and Telegraph 10 * Co. or Unix System Laboratories, Inc. and are reproduced herein with 11 * the permission of UNIX System Laboratories, Inc. 12 * 13 * Redistribution and use in source and binary forms, with or without 14 * modification, are permitted provided that the following conditions 15 * are met: 16 * 1. Redistributions of source code must retain the above copyright 17 * notice, this list of conditions and the following disclaimer. 18 * 2. Redistributions in binary form must reproduce the above copyright 19 * notice, this list of conditions and the following disclaimer in the 20 * documentation and/or other materials provided with the distribution. 21 * 4. Neither the name of the University nor the names of its contributors 22 * may be used to endorse or promote products derived from this software 23 * without specific prior written permission. 24 * 25 * THIS SOFTWARE IS PROVIDED BY THE AUTHOR AND CONTRIBUTORS ``AS IS'' AND 26 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE 27 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE 28 * ARE DISCLAIMED. IN NO EVENT SHALL THE AUTHOR OR CONTRIBUTORS BE LIABLE 29 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL 30 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS 31 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) 32 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT 33 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY 34 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF 35 * SUCH DAMAGE. 36 */ 37 38#include "opt_rootdevname.h" 39 40#include <sys/cdefs.h> 41__FBSDID("$FreeBSD: head/sys/kern/vfs_mountroot.c 287964 2015-09-18 17:32:22Z trasz $"); 42 43#include <sys/param.h> 44#include <sys/conf.h> 45#include <sys/cons.h> 46#include <sys/fcntl.h> 47#include <sys/jail.h> 48#include <sys/kernel.h> 49#include <sys/malloc.h> 50#include <sys/mdioctl.h> 51#include <sys/mount.h> 52#include <sys/mutex.h> 53#include <sys/namei.h> 54#include <sys/priv.h> 55#include <sys/proc.h> 56#include <sys/filedesc.h> 57#include <sys/reboot.h> 58#include <sys/sbuf.h> 59#include <sys/stat.h> 60#include <sys/syscallsubr.h> 61#include <sys/sysproto.h> 62#include <sys/sx.h> 63#include <sys/sysctl.h> 64#include <sys/sysent.h> 65#include <sys/systm.h> 66#include <sys/vnode.h> 67 68#include <geom/geom.h> 69 70/* 71 * The root filesystem is detailed in the kernel environment variable 72 * vfs.root.mountfrom, which is expected to be in the general format 73 * 74 * <vfsname>:[<path>][ <vfsname>:[<path>] ...] 75 * vfsname := the name of a VFS known to the kernel and capable 76 * of being mounted as root 77 * path := disk device name or other data used by the filesystem 78 * to locate its physical store 79 * 80 * If the environment variable vfs.root.mountfrom is a space separated list, 81 * each list element is tried in turn and the root filesystem will be mounted 82 * from the first one that suceeds. 83 * 84 * The environment variable vfs.root.mountfrom.options is a comma delimited 85 * set of string mount options. These mount options must be parseable 86 * by nmount() in the kernel. 87 */ 88 89static int parse_mount(char **); 90static struct mntarg *parse_mountroot_options(struct mntarg *, const char *); 91 92/* 93 * The vnode of the system's root (/ in the filesystem, without chroot 94 * active.) 95 */ 96struct vnode *rootvnode; 97 98/* 99 * Mount of the system's /dev. 100 */ 101struct mount *rootdevmp; 102 103char *rootdevnames[2] = {NULL, NULL}; 104 105struct mtx root_holds_mtx; 106MTX_SYSINIT(root_holds, &root_holds_mtx, "root_holds", MTX_DEF); 107 108struct root_hold_token { 109 const char *who; 110 LIST_ENTRY(root_hold_token) list; 111}; 112 113static LIST_HEAD(, root_hold_token) root_holds = 114 LIST_HEAD_INITIALIZER(root_holds); 115 116enum action { 117 A_CONTINUE, 118 A_PANIC, 119 A_REBOOT, 120 A_RETRY 121}; 122 123static enum action root_mount_onfail = A_CONTINUE; 124 125static int root_mount_mddev; 126static int root_mount_complete; 127 128/* By default wait up to 3 seconds for devices to appear. */ 129static int root_mount_timeout = 3; 130TUNABLE_INT("vfs.mountroot.timeout", &root_mount_timeout); 131 132struct root_hold_token * 133root_mount_hold(const char *identifier) 134{ 135 struct root_hold_token *h; 136 137 if (root_mounted()) 138 return (NULL); 139 140 h = malloc(sizeof *h, M_DEVBUF, M_ZERO | M_WAITOK); 141 h->who = identifier; 142 mtx_lock(&root_holds_mtx); 143 LIST_INSERT_HEAD(&root_holds, h, list); 144 mtx_unlock(&root_holds_mtx); 145 return (h); 146} 147 148void 149root_mount_rel(struct root_hold_token *h) 150{ 151 152 if (h == NULL) 153 return; 154 mtx_lock(&root_holds_mtx); 155 LIST_REMOVE(h, list); 156 wakeup(&root_holds); 157 mtx_unlock(&root_holds_mtx); 158 free(h, M_DEVBUF); 159} 160 161int 162root_mounted(void) 163{ 164 165 /* No mutex is acquired here because int stores are atomic. */ 166 return (root_mount_complete); 167} 168 169void 170root_mount_wait(void) 171{ 172 173 /* 174 * Panic on an obvious deadlock - the function can't be called from 175 * a thread which is doing the whole SYSINIT stuff. 176 */ 177 KASSERT(curthread->td_proc->p_pid != 0, 178 ("root_mount_wait: cannot be called from the swapper thread")); 179 mtx_lock(&root_holds_mtx); 180 while (!root_mount_complete) { 181 msleep(&root_mount_complete, &root_holds_mtx, PZERO, "rootwait", 182 hz); 183 } 184 mtx_unlock(&root_holds_mtx); 185} 186 187static void 188set_rootvnode(void) 189{ 190 struct proc *p; 191 192 if (VFS_ROOT(TAILQ_FIRST(&mountlist), LK_EXCLUSIVE, &rootvnode)) 193 panic("Cannot find root vnode"); 194 195 VOP_UNLOCK(rootvnode, 0); 196 197 p = curthread->td_proc; 198 FILEDESC_XLOCK(p->p_fd); 199 200 if (p->p_fd->fd_cdir != NULL) 201 vrele(p->p_fd->fd_cdir); 202 p->p_fd->fd_cdir = rootvnode; 203 VREF(rootvnode); 204 205 if (p->p_fd->fd_rdir != NULL) 206 vrele(p->p_fd->fd_rdir); 207 p->p_fd->fd_rdir = rootvnode; 208 VREF(rootvnode); 209 210 FILEDESC_XUNLOCK(p->p_fd); 211} 212 213static int 214vfs_mountroot_devfs(struct thread *td, struct mount **mpp) 215{ 216 struct vfsoptlist *opts; 217 struct vfsconf *vfsp; 218 struct mount *mp; 219 int error; 220 221 *mpp = NULL; 222 223 if (rootdevmp != NULL) { 224 /* 225 * Already have /dev; this happens during rerooting. 226 */ 227 error = vfs_busy(rootdevmp, 0); 228 if (error != 0) 229 return (error); 230 *mpp = rootdevmp; 231 } else { 232 vfsp = vfs_byname("devfs"); 233 KASSERT(vfsp != NULL, ("Could not find devfs by name")); 234 if (vfsp == NULL) 235 return (ENOENT); 236 237 mp = vfs_mount_alloc(NULLVP, vfsp, "/dev", td->td_ucred); 238 239 error = VFS_MOUNT(mp); 240 KASSERT(error == 0, ("VFS_MOUNT(devfs) failed %d", error)); 241 if (error) 242 return (error); 243 244 opts = malloc(sizeof(struct vfsoptlist), M_MOUNT, M_WAITOK); 245 TAILQ_INIT(opts); 246 mp->mnt_opt = opts; 247 248 mtx_lock(&mountlist_mtx); 249 TAILQ_INSERT_HEAD(&mountlist, mp, mnt_list); 250 mtx_unlock(&mountlist_mtx); 251 252 *mpp = mp; 253 rootdevmp = mp; 254 } 255 256 set_rootvnode(); 257 258 error = kern_symlinkat(td, "/", AT_FDCWD, "dev", UIO_SYSSPACE); 259 if (error) 260 printf("kern_symlink /dev -> / returns %d\n", error); 261 262 return (error); 263} 264 265static int 266vfs_mountroot_shuffle(struct thread *td, struct mount *mpdevfs) 267{ 268 struct nameidata nd; 269 struct mount *mporoot, *mpnroot; 270 struct vnode *vp, *vporoot, *vpdevfs; 271 char *fspath; 272 int error; 273 274 mpnroot = TAILQ_NEXT(mpdevfs, mnt_list); 275 276 /* Shuffle the mountlist. */ 277 mtx_lock(&mountlist_mtx); 278 mporoot = TAILQ_FIRST(&mountlist); 279 TAILQ_REMOVE(&mountlist, mpdevfs, mnt_list); 280 if (mporoot != mpdevfs) { 281 TAILQ_REMOVE(&mountlist, mpnroot, mnt_list); 282 TAILQ_INSERT_HEAD(&mountlist, mpnroot, mnt_list); 283 } 284 TAILQ_INSERT_TAIL(&mountlist, mpdevfs, mnt_list); 285 mtx_unlock(&mountlist_mtx); 286 287 cache_purgevfs(mporoot); 288 if (mporoot != mpdevfs) 289 cache_purgevfs(mpdevfs); 290 291 VFS_ROOT(mporoot, LK_EXCLUSIVE, &vporoot); 292 293 VI_LOCK(vporoot); 294 vporoot->v_iflag &= ~VI_MOUNT; 295 VI_UNLOCK(vporoot); 296 vporoot->v_mountedhere = NULL; 297 mporoot->mnt_flag &= ~MNT_ROOTFS; 298 mporoot->mnt_vnodecovered = NULL; 299 vput(vporoot); 300 301 /* Set up the new rootvnode, and purge the cache */ 302 mpnroot->mnt_vnodecovered = NULL; 303 set_rootvnode(); 304 cache_purgevfs(rootvnode->v_mount); 305 306 if (mporoot != mpdevfs) { 307 /* Remount old root under /.mount or /mnt */ 308 fspath = "/.mount"; 309 NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF, UIO_SYSSPACE, 310 fspath, td); 311 error = namei(&nd); 312 if (error) { 313 NDFREE(&nd, NDF_ONLY_PNBUF); 314 fspath = "/mnt"; 315 NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF, UIO_SYSSPACE, 316 fspath, td); 317 error = namei(&nd); 318 } 319 if (!error) { 320 vp = nd.ni_vp; 321 error = (vp->v_type == VDIR) ? 0 : ENOTDIR; 322 if (!error) 323 error = vinvalbuf(vp, V_SAVE, 0, 0); 324 if (!error) { 325 cache_purge(vp); 326 mporoot->mnt_vnodecovered = vp; 327 vp->v_mountedhere = mporoot; 328 strlcpy(mporoot->mnt_stat.f_mntonname, 329 fspath, MNAMELEN); 330 VOP_UNLOCK(vp, 0); 331 } else 332 vput(vp); 333 } 334 NDFREE(&nd, NDF_ONLY_PNBUF); 335 336 if (error && bootverbose) 337 printf("mountroot: unable to remount previous root " 338 "under /.mount or /mnt (error %d).\n", error); 339 } 340 341 /* Remount devfs under /dev */ 342 NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF, UIO_SYSSPACE, "/dev", td); 343 error = namei(&nd); 344 if (!error) { 345 vp = nd.ni_vp; 346 error = (vp->v_type == VDIR) ? 0 : ENOTDIR; 347 if (!error) 348 error = vinvalbuf(vp, V_SAVE, 0, 0); 349 if (!error) { 350 vpdevfs = mpdevfs->mnt_vnodecovered; 351 if (vpdevfs != NULL) { 352 cache_purge(vpdevfs); 353 vpdevfs->v_mountedhere = NULL; 354 vrele(vpdevfs); 355 } 356 mpdevfs->mnt_vnodecovered = vp; 357 vp->v_mountedhere = mpdevfs; 358 VOP_UNLOCK(vp, 0); 359 } else 360 vput(vp); 361 } 362 if (error && bootverbose) 363 printf("mountroot: unable to remount devfs under /dev " 364 "(error %d).\n", error); 365 NDFREE(&nd, NDF_ONLY_PNBUF); 366 367 if (mporoot == mpdevfs) { 368 vfs_unbusy(mpdevfs); 369 /* Unlink the no longer needed /dev/dev -> / symlink */ 370 error = kern_unlinkat(td, AT_FDCWD, "/dev/dev", 371 UIO_SYSSPACE, 0); 372 if (error && bootverbose) 373 printf("mountroot: unable to unlink /dev/dev " 374 "(error %d)\n", error); 375 } 376 377 return (0); 378} 379 380/* 381 * Configuration parser. 382 */ 383 384/* Parser character classes. */ 385#define CC_WHITESPACE -1 386#define CC_NONWHITESPACE -2 387 388/* Parse errors. */ 389#define PE_EOF -1 390#define PE_EOL -2 391 392static __inline int 393parse_peek(char **conf) 394{ 395 396 return (**conf); 397} 398 399static __inline void 400parse_poke(char **conf, int c) 401{ 402 403 **conf = c; 404} 405 406static __inline void 407parse_advance(char **conf) 408{ 409 410 (*conf)++; 411} 412 413static int 414parse_skipto(char **conf, int mc) 415{ 416 int c, match; 417 418 while (1) { 419 c = parse_peek(conf); 420 if (c == 0) 421 return (PE_EOF); 422 switch (mc) { 423 case CC_WHITESPACE: 424 match = (c == ' ' || c == '\t' || c == '\n') ? 1 : 0; 425 break; 426 case CC_NONWHITESPACE: 427 if (c == '\n') 428 return (PE_EOL); 429 match = (c != ' ' && c != '\t') ? 1 : 0; 430 break; 431 default: 432 match = (c == mc) ? 1 : 0; 433 break; 434 } 435 if (match) 436 break; 437 parse_advance(conf); 438 } 439 return (0); 440} 441 442static int 443parse_token(char **conf, char **tok) 444{ 445 char *p; 446 size_t len; 447 int error; 448 449 *tok = NULL; 450 error = parse_skipto(conf, CC_NONWHITESPACE); 451 if (error) 452 return (error); 453 p = *conf; 454 error = parse_skipto(conf, CC_WHITESPACE); 455 len = *conf - p; 456 *tok = malloc(len + 1, M_TEMP, M_WAITOK | M_ZERO); 457 bcopy(p, *tok, len); 458 return (0); 459} 460 461static void 462parse_dir_ask_printenv(const char *var) 463{ 464 char *val; 465 466 val = kern_getenv(var); 467 if (val != NULL) { 468 printf(" %s=%s\n", var, val); 469 freeenv(val); 470 } 471} 472 473static int 474parse_dir_ask(char **conf) 475{ 476 char name[80]; 477 char *mnt; 478 int error; 479 480 printf("\nLoader variables:\n"); 481 parse_dir_ask_printenv("vfs.root.mountfrom"); 482 parse_dir_ask_printenv("vfs.root.mountfrom.options"); 483 484 printf("\nManual root filesystem specification:\n"); 485 printf(" <fstype>:<device> [options]\n"); 486 printf(" Mount <device> using filesystem <fstype>\n"); 487 printf(" and with the specified (optional) option list.\n"); 488 printf("\n"); 489 printf(" eg. ufs:/dev/da0s1a\n"); 490 printf(" zfs:tank\n"); 491 printf(" cd9660:/dev/acd0 ro\n"); 492 printf(" (which is equivalent to: "); 493 printf("mount -t cd9660 -o ro /dev/acd0 /)\n"); 494 printf("\n"); 495 printf(" ? List valid disk boot devices\n"); 496 printf(" . Yield 1 second (for background tasks)\n"); 497 printf(" <empty line> Abort manual input\n"); 498 499 do { 500 error = EINVAL; 501 printf("\nmountroot> "); 502 cngets(name, sizeof(name), GETS_ECHO); 503 if (name[0] == '\0') 504 break; 505 if (name[0] == '?' && name[1] == '\0') { 506 printf("\nList of GEOM managed disk devices:\n "); 507 g_dev_print(); 508 continue; 509 } 510 if (name[0] == '.' && name[1] == '\0') { 511 pause("rmask", hz); 512 continue; 513 } 514 mnt = name; 515 error = parse_mount(&mnt); 516 if (error == -1) 517 printf("Invalid file system specification.\n"); 518 } while (error != 0); 519 520 return (error); 521} 522 523static int 524parse_dir_md(char **conf) 525{ 526 struct stat sb; 527 struct thread *td; 528 struct md_ioctl *mdio; 529 char *path, *tok; 530 int error, fd, len; 531 532 td = curthread; 533 534 error = parse_token(conf, &tok); 535 if (error) 536 return (error); 537 538 len = strlen(tok); 539 mdio = malloc(sizeof(*mdio) + len + 1, M_TEMP, M_WAITOK | M_ZERO); 540 path = (void *)(mdio + 1); 541 bcopy(tok, path, len); 542 free(tok, M_TEMP); 543 544 /* Get file status. */ 545 error = kern_statat(td, 0, AT_FDCWD, path, UIO_SYSSPACE, &sb, NULL); 546 if (error) 547 goto out; 548 549 /* Open /dev/mdctl so that we can attach/detach. */ 550 error = kern_openat(td, AT_FDCWD, "/dev/" MDCTL_NAME, UIO_SYSSPACE, 551 O_RDWR, 0); 552 if (error) 553 goto out; 554 555 fd = td->td_retval[0]; 556 mdio->md_version = MDIOVERSION; 557 mdio->md_type = MD_VNODE; 558 559 if (root_mount_mddev != -1) { 560 mdio->md_unit = root_mount_mddev; 561 DROP_GIANT(); 562 error = kern_ioctl(td, fd, MDIOCDETACH, (void *)mdio); 563 PICKUP_GIANT(); 564 /* Ignore errors. We don't care. */ 565 root_mount_mddev = -1; 566 } 567 568 mdio->md_file = (void *)(mdio + 1); 569 mdio->md_options = MD_AUTOUNIT | MD_READONLY; 570 mdio->md_mediasize = sb.st_size; 571 mdio->md_unit = 0; 572 DROP_GIANT(); 573 error = kern_ioctl(td, fd, MDIOCATTACH, (void *)mdio); 574 PICKUP_GIANT(); 575 if (error) 576 goto out; 577 578 if (mdio->md_unit > 9) { 579 printf("rootmount: too many md units\n"); 580 mdio->md_file = NULL; 581 mdio->md_options = 0; 582 mdio->md_mediasize = 0; 583 DROP_GIANT(); 584 error = kern_ioctl(td, fd, MDIOCDETACH, (void *)mdio); 585 PICKUP_GIANT(); 586 /* Ignore errors. We don't care. */ 587 error = ERANGE; 588 goto out; 589 } 590 591 root_mount_mddev = mdio->md_unit; 592 printf(MD_NAME "%u attached to %s\n", root_mount_mddev, mdio->md_file); 593 594 error = kern_close(td, fd); 595 596 out: 597 free(mdio, M_TEMP); 598 return (error); 599} 600 601static int 602parse_dir_onfail(char **conf) 603{ 604 char *action; 605 int error; 606 607 error = parse_token(conf, &action); 608 if (error) 609 return (error); 610 611 if (!strcmp(action, "continue")) 612 root_mount_onfail = A_CONTINUE; 613 else if (!strcmp(action, "panic")) 614 root_mount_onfail = A_PANIC; 615 else if (!strcmp(action, "reboot")) 616 root_mount_onfail = A_REBOOT; 617 else if (!strcmp(action, "retry")) 618 root_mount_onfail = A_RETRY; 619 else { 620 printf("rootmount: %s: unknown action\n", action); 621 error = EINVAL; 622 } 623 624 free(action, M_TEMP); 625 return (0); 626} 627 628static int 629parse_dir_timeout(char **conf) 630{ 631 char *tok, *endtok; 632 long secs; 633 int error; 634 635 error = parse_token(conf, &tok); 636 if (error) 637 return (error); 638 639 secs = strtol(tok, &endtok, 0); 640 error = (secs < 0 || *endtok != '\0') ? EINVAL : 0; 641 if (!error) 642 root_mount_timeout = secs; 643 free(tok, M_TEMP); 644 return (error); 645} 646 647static int 648parse_directive(char **conf) 649{ 650 char *dir; 651 int error; 652 653 error = parse_token(conf, &dir); 654 if (error) 655 return (error); 656 657 if (strcmp(dir, ".ask") == 0) 658 error = parse_dir_ask(conf); 659 else if (strcmp(dir, ".md") == 0) 660 error = parse_dir_md(conf); 661 else if (strcmp(dir, ".onfail") == 0) 662 error = parse_dir_onfail(conf); 663 else if (strcmp(dir, ".timeout") == 0) 664 error = parse_dir_timeout(conf); 665 else { 666 printf("mountroot: invalid directive `%s'\n", dir); 667 /* Ignore the rest of the line. */ 668 (void)parse_skipto(conf, '\n'); 669 error = EINVAL; 670 } 671 free(dir, M_TEMP); 672 return (error); 673} 674 675static int 676parse_mount_dev_present(const char *dev) 677{ 678 struct nameidata nd; 679 int error; 680 681 NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF, UIO_SYSSPACE, dev, curthread); 682 error = namei(&nd); 683 if (!error) 684 vput(nd.ni_vp); 685 NDFREE(&nd, NDF_ONLY_PNBUF); 686 return (error != 0) ? 0 : 1; 687} 688 689#define ERRMSGL 255 690static int 691parse_mount(char **conf) 692{ 693 char *errmsg; 694 struct mntarg *ma; 695 char *dev, *fs, *opts, *tok; 696 int delay, error, timeout; 697 698 error = parse_token(conf, &tok); 699 if (error) 700 return (error); 701 fs = tok; 702 error = parse_skipto(&tok, ':'); 703 if (error) { 704 free(fs, M_TEMP); 705 return (error); 706 } 707 parse_poke(&tok, '\0'); 708 parse_advance(&tok); 709 dev = tok; 710 711 if (root_mount_mddev != -1) { 712 /* Handle substitution for the md unit number. */ 713 tok = strstr(dev, "md#"); 714 if (tok != NULL) 715 tok[2] = '0' + root_mount_mddev; 716 } 717 718 /* Parse options. */ 719 error = parse_token(conf, &tok); 720 opts = (error == 0) ? tok : NULL; 721 722 printf("Trying to mount root from %s:%s [%s]...\n", fs, dev, 723 (opts != NULL) ? opts : ""); 724 725 errmsg = malloc(ERRMSGL, M_TEMP, M_WAITOK | M_ZERO); 726 727 if (vfs_byname(fs) == NULL) { 728 strlcpy(errmsg, "unknown file system", ERRMSGL); 729 error = ENOENT; 730 goto out; 731 } 732 733 if (strcmp(fs, "zfs") != 0 && strstr(fs, "nfs") == NULL && 734 dev[0] != '\0' && !parse_mount_dev_present(dev)) { 735 printf("mountroot: waiting for device %s ...\n", dev); 736 delay = hz / 10; 737 timeout = root_mount_timeout * hz; 738 do { 739 pause("rmdev", delay); 740 timeout -= delay; 741 } while (timeout > 0 && !parse_mount_dev_present(dev)); 742 if (timeout <= 0) { 743 error = ENODEV; 744 goto out; 745 } 746 } 747 748 ma = NULL; 749 ma = mount_arg(ma, "fstype", fs, -1); 750 ma = mount_arg(ma, "fspath", "/", -1); 751 ma = mount_arg(ma, "from", dev, -1); 752 ma = mount_arg(ma, "errmsg", errmsg, ERRMSGL); 753 ma = mount_arg(ma, "ro", NULL, 0); 754 ma = parse_mountroot_options(ma, opts); 755 error = kernel_mount(ma, MNT_ROOTFS); 756 757 out: 758 if (error) { 759 printf("Mounting from %s:%s failed with error %d", 760 fs, dev, error); 761 if (errmsg[0] != '\0') 762 printf(": %s", errmsg); 763 printf(".\n"); 764 } 765 free(fs, M_TEMP); 766 free(errmsg, M_TEMP); 767 if (opts != NULL) 768 free(opts, M_TEMP); 769 /* kernel_mount can return -1 on error. */ 770 return ((error < 0) ? EDOOFUS : error); 771} 772#undef ERRMSGL 773 774static int 775vfs_mountroot_parse(struct sbuf *sb, struct mount *mpdevfs) 776{ 777 struct mount *mp; 778 char *conf; 779 int error; 780 781 root_mount_mddev = -1; 782 783retry: 784 conf = sbuf_data(sb); 785 mp = TAILQ_NEXT(mpdevfs, mnt_list); 786 error = (mp == NULL) ? 0 : EDOOFUS; 787 root_mount_onfail = A_CONTINUE; 788 while (mp == NULL) { 789 error = parse_skipto(&conf, CC_NONWHITESPACE); 790 if (error == PE_EOL) { 791 parse_advance(&conf); 792 continue; 793 } 794 if (error < 0) 795 break; 796 switch (parse_peek(&conf)) { 797 case '#': 798 error = parse_skipto(&conf, '\n'); 799 break; 800 case '.': 801 error = parse_directive(&conf); 802 break; 803 default: 804 error = parse_mount(&conf); 805 if (error == -1) { 806 printf("mountroot: invalid file system " 807 "specification.\n"); 808 error = 0; 809 } 810 break; 811 } 812 if (error < 0) 813 break; 814 /* Ignore any trailing garbage on the line. */ 815 if (parse_peek(&conf) != '\n') { 816 printf("mountroot: advancing to next directive...\n"); 817 (void)parse_skipto(&conf, '\n'); 818 } 819 mp = TAILQ_NEXT(mpdevfs, mnt_list); 820 } 821 if (mp != NULL) 822 return (0); 823 824 /* 825 * We failed to mount (a new) root. 826 */ 827 switch (root_mount_onfail) { 828 case A_CONTINUE: 829 break; 830 case A_PANIC: 831 panic("mountroot: unable to (re-)mount root."); 832 /* NOTREACHED */ 833 case A_RETRY: 834 goto retry; 835 case A_REBOOT: 836 kern_reboot(RB_NOSYNC); 837 /* NOTREACHED */ 838 } 839 840 return (error); 841} 842 843static void 844vfs_mountroot_conf0(struct sbuf *sb) 845{ 846 char *s, *tok, *mnt, *opt; 847 int error; 848 849 sbuf_printf(sb, ".onfail panic\n"); 850 sbuf_printf(sb, ".timeout %d\n", root_mount_timeout); 851 if (boothowto & RB_ASKNAME) 852 sbuf_printf(sb, ".ask\n"); 853#ifdef ROOTDEVNAME 854 if (boothowto & RB_DFLTROOT) 855 sbuf_printf(sb, "%s\n", ROOTDEVNAME); 856#endif 857 if (boothowto & RB_CDROM) { 858 sbuf_printf(sb, "cd9660:/dev/cd0 ro\n"); 859 sbuf_printf(sb, ".timeout 0\n"); 860 sbuf_printf(sb, "cd9660:/dev/acd0 ro\n"); 861 sbuf_printf(sb, ".timeout %d\n", root_mount_timeout); 862 } 863 s = kern_getenv("vfs.root.mountfrom"); 864 if (s != NULL) { 865 opt = kern_getenv("vfs.root.mountfrom.options"); 866 tok = s; 867 error = parse_token(&tok, &mnt); 868 while (!error) { 869 sbuf_printf(sb, "%s %s\n", mnt, 870 (opt != NULL) ? opt : ""); 871 free(mnt, M_TEMP); 872 error = parse_token(&tok, &mnt); 873 } 874 if (opt != NULL) 875 freeenv(opt); 876 freeenv(s); 877 } 878 if (rootdevnames[0] != NULL) 879 sbuf_printf(sb, "%s\n", rootdevnames[0]); 880 if (rootdevnames[1] != NULL) 881 sbuf_printf(sb, "%s\n", rootdevnames[1]); 882#ifdef ROOTDEVNAME 883 if (!(boothowto & RB_DFLTROOT)) 884 sbuf_printf(sb, "%s\n", ROOTDEVNAME); 885#endif 886 if (!(boothowto & RB_ASKNAME)) 887 sbuf_printf(sb, ".ask\n"); 888} 889 890static int 891vfs_mountroot_readconf(struct thread *td, struct sbuf *sb) 892{ 893 static char buf[128]; 894 struct nameidata nd; 895 off_t ofs; 896 ssize_t resid; 897 int error, flags, len; 898 899 NDINIT(&nd, LOOKUP, FOLLOW, UIO_SYSSPACE, "/.mount.conf", td); 900 flags = FREAD; 901 error = vn_open(&nd, &flags, 0, NULL); 902 if (error) 903 return (error); 904 905 NDFREE(&nd, NDF_ONLY_PNBUF); 906 ofs = 0; 907 len = sizeof(buf) - 1; 908 while (1) { 909 error = vn_rdwr(UIO_READ, nd.ni_vp, buf, len, ofs, 910 UIO_SYSSPACE, IO_NODELOCKED, td->td_ucred, 911 NOCRED, &resid, td); 912 if (error) 913 break; 914 if (resid == len) 915 break; 916 buf[len - resid] = 0; 917 sbuf_printf(sb, "%s", buf); 918 ofs += len - resid; 919 } 920 921 VOP_UNLOCK(nd.ni_vp, 0); 922 vn_close(nd.ni_vp, FREAD, td->td_ucred, td); 923 return (error); 924} 925 926static void 927vfs_mountroot_wait(void) 928{ 929 struct root_hold_token *h; 930 struct timeval lastfail; 931 int curfail; 932 933 curfail = 0; 934 while (1) { 935 DROP_GIANT(); 936 g_waitidle(); 937 PICKUP_GIANT(); 938 mtx_lock(&root_holds_mtx); 939 if (LIST_EMPTY(&root_holds)) { 940 mtx_unlock(&root_holds_mtx); 941 break; 942 } 943 if (ppsratecheck(&lastfail, &curfail, 1)) { 944 printf("Root mount waiting for:"); 945 LIST_FOREACH(h, &root_holds, list) 946 printf(" %s", h->who); 947 printf("\n"); 948 } 949 msleep(&root_holds, &root_holds_mtx, PZERO | PDROP, "roothold", 950 hz); 951 } 952} 953 954void 955vfs_mountroot(void) 956{ 957 struct mount *mp; 958 struct sbuf *sb; 959 struct thread *td; 960 time_t timebase; 961 int error; 962 963 td = curthread; 964 965 vfs_mountroot_wait(); 966 967 sb = sbuf_new_auto(); 968 vfs_mountroot_conf0(sb); 969 sbuf_finish(sb); 970 971 error = vfs_mountroot_devfs(td, &mp); 972 while (!error) { 973 error = vfs_mountroot_parse(sb, mp); 974 if (!error) { 975 error = vfs_mountroot_shuffle(td, mp); 976 if (!error) { 977 sbuf_clear(sb); 978 error = vfs_mountroot_readconf(td, sb); 979 sbuf_finish(sb); 980 } 981 } 982 } 983 984 sbuf_delete(sb); 985 986 /* 987 * Iterate over all currently mounted file systems and use 988 * the time stamp found to check and/or initialize the RTC. 989 * Call inittodr() only once and pass it the largest of the 990 * timestamps we encounter. 991 */ 992 timebase = 0; 993 mtx_lock(&mountlist_mtx); 994 mp = TAILQ_FIRST(&mountlist); 995 while (mp != NULL) { 996 if (mp->mnt_time > timebase) 997 timebase = mp->mnt_time; 998 mp = TAILQ_NEXT(mp, mnt_list); 999 } 1000 mtx_unlock(&mountlist_mtx); 1001 inittodr(timebase); 1002 1003 /* Keep prison0's root in sync with the global rootvnode. */ 1004 mtx_lock(&prison0.pr_mtx); 1005 prison0.pr_root = rootvnode; 1006 vref(prison0.pr_root); 1007 mtx_unlock(&prison0.pr_mtx); 1008 1009 mtx_lock(&root_holds_mtx); 1010 atomic_store_rel_int(&root_mount_complete, 1); 1011 wakeup(&root_mount_complete); 1012 mtx_unlock(&root_holds_mtx); 1013 1014 EVENTHANDLER_INVOKE(mountroot); 1015} 1016 1017static struct mntarg * 1018parse_mountroot_options(struct mntarg *ma, const char *options) 1019{ 1020 char *p; 1021 char *name, *name_arg; 1022 char *val, *val_arg; 1023 char *opts; 1024 1025 if (options == NULL || options[0] == '\0') 1026 return (ma); 1027 1028 p = opts = strdup(options, M_MOUNT); 1029 if (opts == NULL) { 1030 return (ma); 1031 } 1032 1033 while((name = strsep(&p, ",")) != NULL) { 1034 if (name[0] == '\0') 1035 break; 1036 1037 val = strchr(name, '='); 1038 if (val != NULL) { 1039 *val = '\0'; 1040 ++val; 1041 } 1042 if( strcmp(name, "rw") == 0 || 1043 strcmp(name, "noro") == 0) { 1044 /* 1045 * The first time we mount the root file system, 1046 * we need to mount 'ro', so We need to ignore 1047 * 'rw' and 'noro' mount options. 1048 */ 1049 continue; 1050 } 1051 name_arg = strdup(name, M_MOUNT); 1052 val_arg = NULL; 1053 if (val != NULL) 1054 val_arg = strdup(val, M_MOUNT); 1055 1056 ma = mount_arg(ma, name_arg, val_arg, 1057 (val_arg != NULL ? -1 : 0)); 1058 } 1059 free(opts, M_MOUNT); 1060 return (ma); 1061} 1062