1 /*- 2 * Copyright (c) 1999 Michael Smith 3 * All rights reserved. 4 * Copyright (c) 1999 Poul-Henning Kamp 5 * All rights reserved. 6 * 7 * Redistribution and use in source and binary forms, with or without 8 * modification, are permitted provided that the following conditions 9 * are met: 10 * 1. Redistributions of source code must retain the above copyright 11 * notice, this list of conditions and the following disclaimer. 12 * 2. Redistributions in binary form must reproduce the above copyright 13 * notice, this list of conditions and the following disclaimer in the 14 * documentation and/or other materials provided with the distribution. 15 * 16 * THIS SOFTWARE IS PROVIDED BY THE AUTHOR AND CONTRIBUTORS ``AS IS'' AND 17 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE 18 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE 19 * ARE DISCLAIMED. IN NO EVENT SHALL THE AUTHOR OR CONTRIBUTORS BE LIABLE 20 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL 21 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS 22 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) 23 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT 24 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY 25 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF 26 * SUCH DAMAGE. 27 * 28 * $FreeBSD: src/sys/kern/vfs_conf.c,v 1.49.2.5 2003/01/07 11:56:53 joerg Exp $ 29 */ 30 31 /* 32 * Locate and mount the root filesystem. 33 * 34 * The root filesystem is detailed in the kernel environment variable 35 * vfs.root.mountfrom, which is expected to be in the general format 36 * 37 * <vfsname>:[<path>] 38 * vfsname := the name of a VFS known to the kernel and capable 39 * of being mounted as root 40 * path := disk device name or other data used by the filesystem 41 * to locate its physical store 42 * 43 */ 44 45 #include "opt_rootdevname.h" 46 47 #include <sys/param.h> 48 #include <sys/kernel.h> 49 #include <sys/systm.h> 50 #include <sys/proc.h> 51 #include <sys/vnode.h> 52 #include <sys/mount.h> 53 #include <sys/malloc.h> 54 #include <sys/reboot.h> 55 #include <sys/diskslice.h> 56 #include <sys/conf.h> 57 #include <sys/cons.h> 58 #include <sys/device.h> 59 #include <sys/disk.h> 60 #include <sys/namecache.h> 61 #include <sys/paths.h> 62 #include <sys/thread2.h> 63 #include <sys/nlookup.h> 64 #include <sys/devfs.h> 65 #include <sys/sysctl.h> 66 67 #include "opt_ddb.h" 68 #ifdef DDB 69 #include <ddb/ddb.h> 70 #endif 71 72 MALLOC_DEFINE(M_MOUNT, "mount", "vfs mount structure"); 73 74 #define ROOTNAME "root_device" 75 76 struct vnode *rootvnode; 77 struct nchandle rootnch; 78 79 /* 80 * The root specifiers we will try if RB_CDROM is specified. Note that 81 * with DEVFS we do not use the compatibility slice's whole-disk 'c' 82 * partition. Instead we just use the whole disk, e.g. cd0 or cd0s0. 83 */ 84 static char *cdrom_rootdevnames[] = { 85 "cd9660:cd0", /* SCSI (including AHCI and SILI) */ 86 "cd9660:acd0", /* NATA */ 87 "cd9660:cd1", /* SCSI (including AHCI and SILI) */ 88 "cd9660:acd1", /* NATA */ 89 "cd9660:cd8", /* USB */ 90 "cd9660:cd9", /* USB */ 91 NULL 92 }; 93 94 int vfs_mountroot_devfs(void); 95 static void vfs_mountroot(void *junk); 96 static int vfs_mountroot_try(const char *mountfrom); 97 static int vfs_mountroot_ask(void); 98 static int getline(char *cp, int limit); 99 100 /* legacy find-root code */ 101 char *rootdevnames[2] = {NULL, NULL}; 102 static int setrootbyname(char *name); 103 104 SYSINIT(mountroot, SI_SUB_MOUNT_ROOT, SI_ORDER_SECOND, vfs_mountroot, NULL); 105 106 static int wakedelay = 2; /* delay before mounting root in seconds */ 107 TUNABLE_INT("vfs.root.wakedelay", &wakedelay); 108 109 /* 110 * Find and mount the root filesystem 111 */ 112 static void 113 vfs_mountroot(void *junk) 114 { 115 cdev_t save_rootdev = rootdev; 116 int i; 117 int dummy; 118 119 /* 120 * Make sure all disk devices created so far have also been probed, 121 * and also make sure that the newly created device nodes for 122 * probed disks are ready, too. 123 * 124 * Messages can fly around here so get good synchronization 125 * coverage. 126 * 127 * XXX - Delay some more (default: 2s) to help drivers which pickup 128 * devices asynchronously and are not caught by CAM's initial 129 * probe. 130 */ 131 sync_devs(); 132 tsleep(&dummy, 0, "syncer", hz * wakedelay); 133 134 135 /* 136 * The root filesystem information is compiled in, and we are 137 * booted with instructions to use it. 138 */ 139 #ifdef ROOTDEVNAME 140 if ((boothowto & RB_DFLTROOT) && 141 !vfs_mountroot_try(ROOTDEVNAME)) 142 return; 143 #endif 144 /* 145 * We are booted with instructions to prompt for the root filesystem, 146 * or to use the compiled-in default when it doesn't exist. 147 */ 148 if (boothowto & (RB_DFLTROOT | RB_ASKNAME)) { 149 if (!vfs_mountroot_ask()) 150 return; 151 } 152 153 /* 154 * We've been given the generic "use CDROM as root" flag. This is 155 * necessary because one media may be used in many different 156 * devices, so we need to search for them. 157 */ 158 if (boothowto & RB_CDROM) { 159 for (i = 0; cdrom_rootdevnames[i] != NULL; i++) { 160 if (!vfs_mountroot_try(cdrom_rootdevnames[i])) 161 return; 162 } 163 } 164 165 /* 166 * Try to use the value read by the loader from /etc/fstab, or 167 * supplied via some other means. This is the preferred 168 * mechanism. 169 */ 170 if (!vfs_mountroot_try(kgetenv("vfs.root.mountfrom"))) 171 return; 172 173 /* 174 * If a vfs set rootdev, try it (XXX VINUM HACK!) 175 */ 176 if (save_rootdev != NULL) { 177 rootdev = save_rootdev; 178 if (!vfs_mountroot_try("")) 179 return; 180 } 181 182 /* 183 * Try values that may have been computed by the machine-dependant 184 * legacy code. 185 */ 186 if (rootdevnames[0] && !vfs_mountroot_try(rootdevnames[0])) 187 return; 188 if (rootdevnames[1] && !vfs_mountroot_try(rootdevnames[1])) 189 return; 190 191 /* 192 * If we have a compiled-in default, and haven't already tried it, try 193 * it now. 194 */ 195 #ifdef ROOTDEVNAME 196 if (!(boothowto & RB_DFLTROOT)) 197 if (!vfs_mountroot_try(ROOTDEVNAME)) 198 return; 199 #endif 200 201 /* 202 * Everything so far has failed, prompt on the console if we haven't 203 * already tried that. 204 */ 205 if (!(boothowto & (RB_DFLTROOT | RB_ASKNAME)) && !vfs_mountroot_ask()) 206 return; 207 panic("Root mount failed, startup aborted."); 208 } 209 210 211 int 212 vfs_mountroot_devfs(void) 213 { 214 struct vnode *vp; 215 struct nchandle nch; 216 struct nlookupdata nd; 217 struct mount *mp; 218 struct vfsconf *vfsp; 219 int error; 220 struct ucred *cred = proc0.p_ucred; 221 const char *devfs_path, *init_chroot; 222 char *dev_malloced = NULL; 223 224 if ((init_chroot = kgetenv("init_chroot")) != NULL) { 225 size_t l; 226 227 l = strlen(init_chroot) + sizeof("/dev"); 228 dev_malloced = kmalloc(l, M_MOUNT, M_WAITOK); 229 ksnprintf(dev_malloced, l, "%s/dev", init_chroot); 230 devfs_path = dev_malloced; 231 } else { 232 devfs_path = "/dev"; 233 } 234 /* 235 * Lookup the requested path and extract the nch and vnode. 236 */ 237 error = nlookup_init_raw(&nd, 238 devfs_path, UIO_SYSSPACE, NLC_FOLLOW, 239 cred, &rootnch); 240 241 if (error == 0) { 242 devfs_debug(DEVFS_DEBUG_DEBUG, "vfs_mountroot_devfs: nlookup_init is ok...\n"); 243 if ((error = nlookup(&nd)) == 0) { 244 devfs_debug(DEVFS_DEBUG_DEBUG, "vfs_mountroot_devfs: nlookup is ok...\n"); 245 if (nd.nl_nch.ncp->nc_vp == NULL) { 246 devfs_debug(DEVFS_DEBUG_SHOW, "vfs_mountroot_devfs: nlookup: simply not found\n"); 247 error = ENOENT; 248 } 249 } 250 } 251 if (dev_malloced != NULL) 252 kfree(dev_malloced, M_MOUNT), dev_malloced = NULL; 253 devfs_path = NULL; 254 if (error) { 255 nlookup_done(&nd); 256 devfs_debug(DEVFS_DEBUG_SHOW, "vfs_mountroot_devfs: nlookup failed, error: %d\n", error); 257 return (error); 258 } 259 260 /* 261 * Extract the locked+refd ncp and cleanup the nd structure 262 */ 263 nch = nd.nl_nch; 264 cache_zero(&nd.nl_nch); 265 nlookup_done(&nd); 266 267 /* 268 * now we have the locked ref'd nch and unreferenced vnode. 269 */ 270 vp = nch.ncp->nc_vp; 271 if ((error = vget(vp, LK_EXCLUSIVE)) != 0) { 272 cache_put(&nch); 273 devfs_debug(DEVFS_DEBUG_SHOW, "vfs_mountroot_devfs: vget failed\n"); 274 return (error); 275 } 276 cache_unlock(&nch); 277 278 if ((error = vinvalbuf(vp, V_SAVE, 0, 0)) != 0) { 279 cache_drop(&nch); 280 vput(vp); 281 devfs_debug(DEVFS_DEBUG_SHOW, "vfs_mountroot_devfs: vinvalbuf failed\n"); 282 return (error); 283 } 284 if (vp->v_type != VDIR) { 285 cache_drop(&nch); 286 vput(vp); 287 devfs_debug(DEVFS_DEBUG_SHOW, "vfs_mountroot_devfs: vp is not VDIR\n"); 288 return (ENOTDIR); 289 } 290 291 vfsp = vfsconf_find_by_name("devfs"); 292 vsetflags(vp, VMOUNT); 293 294 /* 295 * Allocate and initialize the filesystem. 296 */ 297 mp = kmalloc(sizeof(struct mount), M_MOUNT, M_ZERO|M_WAITOK); 298 mount_init(mp); 299 vfs_busy(mp, LK_NOWAIT); 300 mp->mnt_op = vfsp->vfc_vfsops; 301 mp->mnt_vfc = vfsp; 302 vfsp->vfc_refcount++; 303 mp->mnt_stat.f_type = vfsp->vfc_typenum; 304 mp->mnt_flag |= vfsp->vfc_flags & MNT_VISFLAGMASK; 305 strncpy(mp->mnt_stat.f_fstypename, vfsp->vfc_name, MFSNAMELEN); 306 mp->mnt_stat.f_owner = cred->cr_uid; 307 vn_unlock(vp); 308 309 /* 310 * Mount the filesystem. 311 */ 312 error = VFS_MOUNT(mp, "/dev", NULL, cred); 313 314 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 315 316 /* 317 * Put the new filesystem on the mount list after root. The mount 318 * point gets its own mnt_ncmountpt (unless the VFS already set one 319 * up) which represents the root of the mount. The lookup code 320 * detects the mount point going forward and checks the root of 321 * the mount going backwards. 322 * 323 * It is not necessary to invalidate or purge the vnode underneath 324 * because elements under the mount will be given their own glue 325 * namecache record. 326 */ 327 if (!error) { 328 if (mp->mnt_ncmountpt.ncp == NULL) { 329 /* 330 * allocate, then unlock, but leave the ref intact 331 */ 332 cache_allocroot(&mp->mnt_ncmountpt, mp, NULL); 333 cache_unlock(&mp->mnt_ncmountpt); 334 } 335 mp->mnt_ncmounton = nch; /* inherits ref */ 336 nch.ncp->nc_flag |= NCF_ISMOUNTPT; 337 338 /* XXX get the root of the fs and cache_setvp(mnt_ncmountpt...) */ 339 vclrflags(vp, VMOUNT); 340 mountlist_insert(mp, MNTINS_LAST); 341 vn_unlock(vp); 342 //checkdirs(&mp->mnt_ncmounton, &mp->mnt_ncmountpt); 343 error = vfs_allocate_syncvnode(mp); 344 if (error) { 345 devfs_debug(DEVFS_DEBUG_SHOW, "vfs_mountroot_devfs: vfs_allocate_syncvnode failed\n"); 346 } 347 vfs_unbusy(mp); 348 error = VFS_START(mp, 0); 349 vrele(vp); 350 } else { 351 vfs_rm_vnodeops(mp, NULL, &mp->mnt_vn_coherency_ops); 352 vfs_rm_vnodeops(mp, NULL, &mp->mnt_vn_journal_ops); 353 vfs_rm_vnodeops(mp, NULL, &mp->mnt_vn_norm_ops); 354 vfs_rm_vnodeops(mp, NULL, &mp->mnt_vn_spec_ops); 355 vfs_rm_vnodeops(mp, NULL, &mp->mnt_vn_fifo_ops); 356 vclrflags(vp, VMOUNT); 357 mp->mnt_vfc->vfc_refcount--; 358 vfs_unbusy(mp); 359 kfree(mp, M_MOUNT); 360 cache_drop(&nch); 361 vput(vp); 362 devfs_debug(DEVFS_DEBUG_SHOW, "vfs_mountroot_devfs: mount failed\n"); 363 } 364 365 devfs_debug(DEVFS_DEBUG_DEBUG, "rootmount_devfs done with error: %d\n", error); 366 return (error); 367 } 368 369 370 /* 371 * Mount (mountfrom) as the root filesystem. 372 */ 373 static int 374 vfs_mountroot_try(const char *mountfrom) 375 { 376 struct mount *mp; 377 char *vfsname, *devname; 378 int error; 379 char patt[32]; 380 const char *cp, *ep; 381 char *mf; 382 struct proc *p; 383 struct vnode *vp; 384 385 vfsname = NULL; 386 devname = NULL; 387 mp = NULL; 388 error = EINVAL; 389 390 if (mountfrom == NULL) 391 return(error); /* don't complain */ 392 393 crit_enter(); 394 kprintf("Mounting root from %s\n", mountfrom); 395 crit_exit(); 396 397 cp = mountfrom; 398 /* parse vfs name and devname */ 399 vfsname = kmalloc(MFSNAMELEN, M_MOUNT, M_WAITOK); 400 devname = kmalloc(MNAMELEN, M_MOUNT, M_WAITOK); 401 mf = kmalloc(MFSNAMELEN+MNAMELEN, M_MOUNT, M_WAITOK); 402 for(;;) { 403 for (ep = cp; (*ep != 0) && (*ep != ';'); ep++); 404 bzero(vfsname, MFSNAMELEN); 405 bzero(devname, MNAMELEN); 406 bzero(mf, MFSNAMELEN+MNAMELEN); 407 strncpy(mf, cp, MFSNAMELEN+MNAMELEN); 408 409 vfsname[0] = devname[0] = 0; 410 ksprintf(patt, "%%%d[a-z0-9]:%%%ds", MFSNAMELEN, MNAMELEN); 411 if (ksscanf(mf, patt, vfsname, devname) < 1) 412 goto end; 413 414 /* allocate a root mount */ 415 error = vfs_rootmountalloc(vfsname, 416 devname[0] != 0 ? devname : ROOTNAME, &mp); 417 if (error != 0) { 418 kprintf("Can't allocate root mount for filesystem '%s': %d\n", 419 vfsname, error); 420 goto end; 421 } 422 mp->mnt_flag |= MNT_ROOTFS; 423 424 /* do our best to set rootdev */ 425 if ((strcmp(vfsname, "hammer") != 0) && (devname[0] != 0) && 426 setrootbyname(devname)) 427 kprintf("setrootbyname failed\n"); 428 429 /* If the root device is a type "memory disk", mount RW */ 430 if (rootdev != NULL && dev_is_good(rootdev) && 431 (dev_dflags(rootdev) & D_MEMDISK)) { 432 mp->mnt_flag &= ~MNT_RDONLY; 433 } 434 435 error = VFS_MOUNT(mp, NULL, NULL, proc0.p_ucred); 436 437 if (!error) 438 break; 439 end: 440 if(*ep == 0) 441 break; 442 cp = ep + 1; 443 } 444 445 if (vfsname != NULL) 446 kfree(vfsname, M_MOUNT); 447 if (devname != NULL) 448 kfree(devname, M_MOUNT); 449 if (mf != NULL) 450 kfree(mf, M_MOUNT); 451 if (error == 0) { 452 /* register with list of mounted filesystems */ 453 mountlist_insert(mp, MNTINS_FIRST); 454 455 /* sanity check system clock against root fs timestamp */ 456 inittodr(mp->mnt_time); 457 458 /* Get the vnode for '/'. Set p->p_fd->fd_cdir to reference it. */ 459 mp = mountlist_boot_getfirst(); 460 if (VFS_ROOT(mp, &vp)) 461 panic("cannot find root vnode"); 462 if (mp->mnt_ncmountpt.ncp == NULL) { 463 cache_allocroot(&mp->mnt_ncmountpt, mp, vp); 464 cache_unlock(&mp->mnt_ncmountpt); /* leave ref intact */ 465 } 466 p = curproc; 467 p->p_fd->fd_cdir = vp; 468 vref(p->p_fd->fd_cdir); 469 p->p_fd->fd_rdir = vp; 470 vref(p->p_fd->fd_rdir); 471 vfs_cache_setroot(vp, cache_hold(&mp->mnt_ncmountpt)); 472 vn_unlock(vp); /* leave ref intact */ 473 cache_copy(&mp->mnt_ncmountpt, &p->p_fd->fd_ncdir); 474 cache_copy(&mp->mnt_ncmountpt, &p->p_fd->fd_nrdir); 475 476 vfs_unbusy(mp); 477 if (mp->mnt_syncer == NULL) { 478 error = vfs_allocate_syncvnode(mp); 479 if (error) 480 kprintf("Warning: no syncer vp for root!\n"); 481 error = 0; 482 } 483 VFS_START( mp, 0 ); 484 } else { 485 if (mp != NULL) { 486 vfs_unbusy(mp); 487 kfree(mp, M_MOUNT); 488 } 489 kprintf("Root mount failed: %d\n", error); 490 } 491 return(error); 492 } 493 494 495 static void 496 vfs_mountroot_ask_callback(char *name, cdev_t dev, bool is_alias, 497 void *arg __unused) 498 { 499 if (!is_alias && dev_is_good(dev) && (dev_dflags(dev) & D_DISK)) 500 kprintf(" \"%s\" ", name); 501 } 502 503 504 /* 505 * Spin prompting on the console for a suitable root filesystem 506 */ 507 static int 508 vfs_mountroot_ask(void) 509 { 510 char name[128]; 511 int llimit = 100; 512 513 kprintf("\nManual root filesystem specification:\n"); 514 kprintf(" <fstype>:<device> Specify root (e.g. ufs:da0s1a)\n"); 515 kprintf(" ? List valid disk boot devices\n"); 516 kprintf(" panic Just panic\n"); 517 kprintf(" abort Abort manual input\n"); 518 while (llimit--) { 519 kprintf("\nmountroot> "); 520 521 if (getline(name, 128) < 0) 522 break; 523 if (name[0] == 0) { 524 ; 525 } else if (name[0] == '?') { 526 kprintf("Possibly valid devices for root FS:\n"); 527 //enumerate all disk devices 528 devfs_scan_callback(vfs_mountroot_ask_callback, NULL); 529 kprintf("\n"); 530 continue; 531 } else if (strcmp(name, "panic") == 0) { 532 panic("panic from console"); 533 } else if (strcmp(name, "abort") == 0) { 534 break; 535 } else if (vfs_mountroot_try(name) == 0) { 536 return(0); 537 } 538 } 539 return(1); 540 } 541 542 543 static int 544 getline(char *cp, int limit) 545 { 546 char *lp; 547 int c; 548 549 lp = cp; 550 for (;;) { 551 c = cngetc(); 552 553 switch (c) { 554 case -1: 555 return(-1); 556 case '\n': 557 case '\r': 558 kprintf("\n"); 559 *lp++ = '\0'; 560 return(0); 561 case '\b': 562 case '\177': 563 if (lp > cp) { 564 kprintf("\b \b"); 565 lp--; 566 } else { 567 kprintf("%c", 7); 568 } 569 continue; 570 case '#': 571 kprintf("#"); 572 lp--; 573 if (lp < cp) 574 lp = cp; 575 continue; 576 case '@': 577 case 'u' & 037: 578 lp = cp; 579 kprintf("%c", '\n'); 580 continue; 581 default: 582 if (lp - cp >= limit - 1) { 583 kprintf("%c", 7); 584 } else { 585 kprintf("%c", c); 586 *lp++ = c; 587 } 588 continue; 589 } 590 } 591 } 592 593 /* 594 * Convert a given name to the cdev_t of the disk-like device 595 * it refers to. 596 */ 597 cdev_t 598 kgetdiskbyname(const char *name) 599 { 600 cdev_t rdev; 601 602 /* 603 * Get the base name of the device 604 */ 605 if (strncmp(name, __SYS_PATH_DEV, sizeof(__SYS_PATH_DEV) - 1) == 0) 606 name += sizeof(__SYS_PATH_DEV) - 1; 607 608 /* 609 * Locate the device 610 */ 611 rdev = devfs_find_device_by_name("%s", name); 612 if (rdev == NULL) { 613 kprintf("no disk named '%s'\n", name); 614 } 615 /* 616 * FOUND DEVICE 617 */ 618 return(rdev); 619 } 620 621 /* 622 * Set rootdev to match (name), given that we expect it to 623 * refer to a disk-like device. 624 */ 625 static int 626 setrootbyname(char *name) 627 { 628 cdev_t diskdev; 629 630 diskdev = kgetdiskbyname(name); 631 if (diskdev != NULL) { 632 rootdev = diskdev; 633 return (0); 634 } 635 /* set to NULL if kgetdiskbyname() fails so that if the first rootdev is 636 * found by fails to mount and the second one isn't found, mountroot_try 637 * doesn't try again with the first one 638 */ 639 rootdev = NULL; 640 return (1); 641 } 642 643 #ifdef DDB 644 DB_SHOW_COMMAND(disk, db_getdiskbyname) 645 { 646 cdev_t dev; 647 648 if (modif[0] == '\0') { 649 db_error("usage: show disk/devicename"); 650 return; 651 } 652 dev = kgetdiskbyname(modif); 653 if (dev != NULL) 654 db_printf("cdev_t = %p\n", dev); 655 else 656 db_printf("No disk device matched.\n"); 657 } 658 #endif 659 660 static int 661 vfs_sysctl_real_root(SYSCTL_HANDLER_ARGS) 662 { 663 char *real_root; 664 size_t len; 665 int error; 666 667 real_root = kgetenv("vfs.root.realroot"); 668 669 if (real_root == NULL) 670 real_root = ""; 671 672 len = strlen(real_root) + 1; 673 674 error = sysctl_handle_string(oidp, real_root, len, req); 675 676 return error; 677 } 678 679 SYSCTL_PROC(_vfs, OID_AUTO, real_root, 680 CTLTYPE_STRING | CTLFLAG_RD, 0, 0, vfs_sysctl_real_root, 681 "A", "Real root mount string"); 682