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 /* 107 * Find and mount the root filesystem 108 */ 109 static void 110 vfs_mountroot(void *junk) 111 { 112 cdev_t save_rootdev = rootdev; 113 int i; 114 int dummy; 115 116 /* 117 * Make sure all disk devices created so far have also been probed, 118 * and also make sure that the newly created device nodes for 119 * probed disks are ready, too. 120 * 121 * Messages can fly around here so get good synchronization 122 * coverage. 123 * 124 * XXX - Delay an additional 2 seconds to help drivers which pickup 125 * devices asynchronously and are not caught by CAM's initial 126 * probe. 127 */ 128 sync_devs(); 129 tsleep(&dummy, 0, "syncer", hz*2); 130 131 132 /* 133 * The root filesystem information is compiled in, and we are 134 * booted with instructions to use it. 135 */ 136 #ifdef ROOTDEVNAME 137 if ((boothowto & RB_DFLTROOT) && 138 !vfs_mountroot_try(ROOTDEVNAME)) 139 return; 140 #endif 141 /* 142 * We are booted with instructions to prompt for the root filesystem, 143 * or to use the compiled-in default when it doesn't exist. 144 */ 145 if (boothowto & (RB_DFLTROOT | RB_ASKNAME)) { 146 if (!vfs_mountroot_ask()) 147 return; 148 } 149 150 /* 151 * We've been given the generic "use CDROM as root" flag. This is 152 * necessary because one media may be used in many different 153 * devices, so we need to search for them. 154 */ 155 if (boothowto & RB_CDROM) { 156 for (i = 0; cdrom_rootdevnames[i] != NULL; i++) { 157 if (!vfs_mountroot_try(cdrom_rootdevnames[i])) 158 return; 159 } 160 } 161 162 /* 163 * Try to use the value read by the loader from /etc/fstab, or 164 * supplied via some other means. This is the preferred 165 * mechanism. 166 */ 167 if (!vfs_mountroot_try(kgetenv("vfs.root.mountfrom"))) 168 return; 169 170 /* 171 * If a vfs set rootdev, try it (XXX VINUM HACK!) 172 */ 173 if (save_rootdev != NULL) { 174 rootdev = save_rootdev; 175 if (!vfs_mountroot_try("")) 176 return; 177 } 178 179 /* 180 * Try values that may have been computed by the machine-dependant 181 * legacy code. 182 */ 183 if (rootdevnames[0] && !vfs_mountroot_try(rootdevnames[0])) 184 return; 185 if (rootdevnames[1] && !vfs_mountroot_try(rootdevnames[1])) 186 return; 187 188 /* 189 * If we have a compiled-in default, and haven't already tried it, try 190 * it now. 191 */ 192 #ifdef ROOTDEVNAME 193 if (!(boothowto & RB_DFLTROOT)) 194 if (!vfs_mountroot_try(ROOTDEVNAME)) 195 return; 196 #endif 197 198 /* 199 * Everything so far has failed, prompt on the console if we haven't 200 * already tried that. 201 */ 202 if (!(boothowto & (RB_DFLTROOT | RB_ASKNAME)) && !vfs_mountroot_ask()) 203 return; 204 panic("Root mount failed, startup aborted."); 205 } 206 207 208 int 209 vfs_mountroot_devfs(void) 210 { 211 struct vnode *vp; 212 struct nchandle nch; 213 struct nlookupdata nd; 214 struct mount *mp; 215 struct vfsconf *vfsp; 216 int error; 217 struct ucred *cred = proc0.p_ucred; 218 const char *devfs_path, *init_chroot; 219 char *dev_malloced = NULL; 220 221 if ((init_chroot = kgetenv("init_chroot")) != NULL) { 222 size_t l; 223 224 l = strlen(init_chroot) + sizeof("/dev"); 225 dev_malloced = kmalloc(l, M_MOUNT, M_WAITOK); 226 ksnprintf(dev_malloced, l, "%s/dev", init_chroot); 227 devfs_path = dev_malloced; 228 } else { 229 devfs_path = "/dev"; 230 } 231 /* 232 * Lookup the requested path and extract the nch and vnode. 233 */ 234 error = nlookup_init_raw(&nd, 235 devfs_path, UIO_SYSSPACE, NLC_FOLLOW, 236 cred, &rootnch); 237 238 if (error == 0) { 239 devfs_debug(DEVFS_DEBUG_DEBUG, "vfs_mountroot_devfs: nlookup_init is ok...\n"); 240 if ((error = nlookup(&nd)) == 0) { 241 devfs_debug(DEVFS_DEBUG_DEBUG, "vfs_mountroot_devfs: nlookup is ok...\n"); 242 if (nd.nl_nch.ncp->nc_vp == NULL) { 243 devfs_debug(DEVFS_DEBUG_SHOW, "vfs_mountroot_devfs: nlookup: simply not found\n"); 244 error = ENOENT; 245 } 246 } 247 } 248 if (dev_malloced != NULL) 249 kfree(dev_malloced, M_MOUNT), dev_malloced = NULL; 250 devfs_path = NULL; 251 if (error) { 252 nlookup_done(&nd); 253 devfs_debug(DEVFS_DEBUG_SHOW, "vfs_mountroot_devfs: nlookup failed, error: %d\n", error); 254 return (error); 255 } 256 257 /* 258 * Extract the locked+refd ncp and cleanup the nd structure 259 */ 260 nch = nd.nl_nch; 261 cache_zero(&nd.nl_nch); 262 nlookup_done(&nd); 263 264 /* 265 * now we have the locked ref'd nch and unreferenced vnode. 266 */ 267 vp = nch.ncp->nc_vp; 268 if ((error = vget(vp, LK_EXCLUSIVE)) != 0) { 269 cache_put(&nch); 270 devfs_debug(DEVFS_DEBUG_SHOW, "vfs_mountroot_devfs: vget failed\n"); 271 return (error); 272 } 273 cache_unlock(&nch); 274 275 if ((error = vinvalbuf(vp, V_SAVE, 0, 0)) != 0) { 276 cache_drop(&nch); 277 vput(vp); 278 devfs_debug(DEVFS_DEBUG_SHOW, "vfs_mountroot_devfs: vinvalbuf failed\n"); 279 return (error); 280 } 281 if (vp->v_type != VDIR) { 282 cache_drop(&nch); 283 vput(vp); 284 devfs_debug(DEVFS_DEBUG_SHOW, "vfs_mountroot_devfs: vp is not VDIR\n"); 285 return (ENOTDIR); 286 } 287 288 vfsp = vfsconf_find_by_name("devfs"); 289 vsetflags(vp, VMOUNT); 290 291 /* 292 * Allocate and initialize the filesystem. 293 */ 294 mp = kmalloc(sizeof(struct mount), M_MOUNT, M_ZERO|M_WAITOK); 295 mount_init(mp); 296 vfs_busy(mp, LK_NOWAIT); 297 mp->mnt_op = vfsp->vfc_vfsops; 298 mp->mnt_vfc = vfsp; 299 vfsp->vfc_refcount++; 300 mp->mnt_stat.f_type = vfsp->vfc_typenum; 301 mp->mnt_flag |= vfsp->vfc_flags & MNT_VISFLAGMASK; 302 strncpy(mp->mnt_stat.f_fstypename, vfsp->vfc_name, MFSNAMELEN); 303 mp->mnt_stat.f_owner = cred->cr_uid; 304 vn_unlock(vp); 305 306 /* 307 * Mount the filesystem. 308 */ 309 error = VFS_MOUNT(mp, "/dev", NULL, cred); 310 311 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 312 313 /* 314 * Put the new filesystem on the mount list after root. The mount 315 * point gets its own mnt_ncmountpt (unless the VFS already set one 316 * up) which represents the root of the mount. The lookup code 317 * detects the mount point going forward and checks the root of 318 * the mount going backwards. 319 * 320 * It is not necessary to invalidate or purge the vnode underneath 321 * because elements under the mount will be given their own glue 322 * namecache record. 323 */ 324 if (!error) { 325 if (mp->mnt_ncmountpt.ncp == NULL) { 326 /* 327 * allocate, then unlock, but leave the ref intact 328 */ 329 cache_allocroot(&mp->mnt_ncmountpt, mp, NULL); 330 cache_unlock(&mp->mnt_ncmountpt); 331 } 332 mp->mnt_ncmounton = nch; /* inherits ref */ 333 nch.ncp->nc_flag |= NCF_ISMOUNTPT; 334 335 /* XXX get the root of the fs and cache_setvp(mnt_ncmountpt...) */ 336 vclrflags(vp, VMOUNT); 337 mountlist_insert(mp, MNTINS_LAST); 338 vn_unlock(vp); 339 //checkdirs(&mp->mnt_ncmounton, &mp->mnt_ncmountpt); 340 error = vfs_allocate_syncvnode(mp); 341 if (error) { 342 devfs_debug(DEVFS_DEBUG_SHOW, "vfs_mountroot_devfs: vfs_allocate_syncvnode failed\n"); 343 } 344 vfs_unbusy(mp); 345 error = VFS_START(mp, 0); 346 vrele(vp); 347 } else { 348 vfs_rm_vnodeops(mp, NULL, &mp->mnt_vn_coherency_ops); 349 vfs_rm_vnodeops(mp, NULL, &mp->mnt_vn_journal_ops); 350 vfs_rm_vnodeops(mp, NULL, &mp->mnt_vn_norm_ops); 351 vfs_rm_vnodeops(mp, NULL, &mp->mnt_vn_spec_ops); 352 vfs_rm_vnodeops(mp, NULL, &mp->mnt_vn_fifo_ops); 353 vclrflags(vp, VMOUNT); 354 mp->mnt_vfc->vfc_refcount--; 355 vfs_unbusy(mp); 356 kfree(mp, M_MOUNT); 357 cache_drop(&nch); 358 vput(vp); 359 devfs_debug(DEVFS_DEBUG_SHOW, "vfs_mountroot_devfs: mount failed\n"); 360 } 361 362 devfs_debug(DEVFS_DEBUG_DEBUG, "rootmount_devfs done with error: %d\n", error); 363 return (error); 364 } 365 366 367 /* 368 * Mount (mountfrom) as the root filesystem. 369 */ 370 static int 371 vfs_mountroot_try(const char *mountfrom) 372 { 373 struct mount *mp; 374 char *vfsname, *devname; 375 int error; 376 char patt[32]; 377 const char *cp, *ep; 378 char *mf; 379 struct proc *p; 380 struct vnode *vp; 381 382 vfsname = NULL; 383 devname = NULL; 384 mp = NULL; 385 error = EINVAL; 386 387 if (mountfrom == NULL) 388 return(error); /* don't complain */ 389 390 crit_enter(); 391 kprintf("Mounting root from %s\n", mountfrom); 392 crit_exit(); 393 394 cp = mountfrom; 395 /* parse vfs name and devname */ 396 vfsname = kmalloc(MFSNAMELEN, M_MOUNT, M_WAITOK); 397 devname = kmalloc(MNAMELEN, M_MOUNT, M_WAITOK); 398 mf = kmalloc(MFSNAMELEN+MNAMELEN, M_MOUNT, M_WAITOK); 399 for(;;) { 400 for (ep = cp; (*ep != 0) && (*ep != ';'); ep++); 401 bzero(vfsname, MFSNAMELEN); 402 bzero(devname, MNAMELEN); 403 bzero(mf, MFSNAMELEN+MNAMELEN); 404 strncpy(mf, cp, MFSNAMELEN+MNAMELEN); 405 406 vfsname[0] = devname[0] = 0; 407 ksprintf(patt, "%%%d[a-z0-9]:%%%ds", MFSNAMELEN, MNAMELEN); 408 if (ksscanf(mf, patt, vfsname, devname) < 1) 409 goto end; 410 411 /* allocate a root mount */ 412 error = vfs_rootmountalloc(vfsname, 413 devname[0] != 0 ? devname : ROOTNAME, &mp); 414 if (error != 0) { 415 kprintf("Can't allocate root mount for filesystem '%s': %d\n", 416 vfsname, error); 417 goto end; 418 } 419 mp->mnt_flag |= MNT_ROOTFS; 420 421 /* do our best to set rootdev */ 422 if ((strcmp(vfsname, "hammer") != 0) && (devname[0] != 0) && 423 setrootbyname(devname)) 424 kprintf("setrootbyname failed\n"); 425 426 /* If the root device is a type "memory disk", mount RW */ 427 if (rootdev != NULL && dev_is_good(rootdev) && 428 (dev_dflags(rootdev) & D_MEMDISK)) { 429 mp->mnt_flag &= ~MNT_RDONLY; 430 } 431 432 error = VFS_MOUNT(mp, NULL, NULL, proc0.p_ucred); 433 434 if (!error) 435 break; 436 end: 437 if(*ep == 0) 438 break; 439 cp = ep + 1; 440 } 441 442 if (vfsname != NULL) 443 kfree(vfsname, M_MOUNT); 444 if (devname != NULL) 445 kfree(devname, M_MOUNT); 446 if (mf != NULL) 447 kfree(mf, M_MOUNT); 448 if (error == 0) { 449 /* register with list of mounted filesystems */ 450 mountlist_insert(mp, MNTINS_FIRST); 451 452 /* sanity check system clock against root fs timestamp */ 453 inittodr(mp->mnt_time); 454 455 /* Get the vnode for '/'. Set p->p_fd->fd_cdir to reference it. */ 456 mp = mountlist_boot_getfirst(); 457 if (VFS_ROOT(mp, &vp)) 458 panic("cannot find root vnode"); 459 if (mp->mnt_ncmountpt.ncp == NULL) { 460 cache_allocroot(&mp->mnt_ncmountpt, mp, vp); 461 cache_unlock(&mp->mnt_ncmountpt); /* leave ref intact */ 462 } 463 p = curproc; 464 p->p_fd->fd_cdir = vp; 465 vref(p->p_fd->fd_cdir); 466 p->p_fd->fd_rdir = vp; 467 vref(p->p_fd->fd_rdir); 468 vfs_cache_setroot(vp, cache_hold(&mp->mnt_ncmountpt)); 469 vn_unlock(vp); /* leave ref intact */ 470 cache_copy(&mp->mnt_ncmountpt, &p->p_fd->fd_ncdir); 471 cache_copy(&mp->mnt_ncmountpt, &p->p_fd->fd_nrdir); 472 473 vfs_unbusy(mp); 474 if (mp->mnt_syncer == NULL) { 475 error = vfs_allocate_syncvnode(mp); 476 if (error) 477 kprintf("Warning: no syncer vp for root!\n"); 478 error = 0; 479 } 480 VFS_START( mp, 0 ); 481 } else { 482 if (mp != NULL) { 483 vfs_unbusy(mp); 484 kfree(mp, M_MOUNT); 485 } 486 kprintf("Root mount failed: %d\n", error); 487 } 488 return(error); 489 } 490 491 492 static void 493 vfs_mountroot_ask_callback(cdev_t dev, void *arg __unused) 494 { 495 if (dev_is_good(dev) && (dev_dflags(dev) & D_DISK)) 496 kprintf(" \"%s\" ", dev->si_name); 497 } 498 499 500 /* 501 * Spin prompting on the console for a suitable root filesystem 502 */ 503 static int 504 vfs_mountroot_ask(void) 505 { 506 char name[128]; 507 int llimit = 100; 508 509 kprintf("\nManual root filesystem specification:\n"); 510 kprintf(" <fstype>:<device> Specify root (e.g. ufs:da0s1a)\n"); 511 kprintf(" ? List valid disk boot devices\n"); 512 kprintf(" panic Just panic\n"); 513 kprintf(" abort Abort manual input\n"); 514 while (llimit--) { 515 kprintf("\nmountroot> "); 516 517 if (getline(name, 128) < 0) 518 break; 519 if (name[0] == 0) { 520 ; 521 } else if (name[0] == '?') { 522 kprintf("Possibly valid devices for root FS:\n"); 523 //enumerate all disk devices 524 devfs_scan_callback(vfs_mountroot_ask_callback, NULL); 525 kprintf("\n"); 526 continue; 527 } else if (strcmp(name, "panic") == 0) { 528 panic("panic from console"); 529 } else if (strcmp(name, "abort") == 0) { 530 break; 531 } else if (vfs_mountroot_try(name) == 0) { 532 return(0); 533 } 534 } 535 return(1); 536 } 537 538 539 static int 540 getline(char *cp, int limit) 541 { 542 char *lp; 543 int c; 544 545 lp = cp; 546 for (;;) { 547 c = cngetc(); 548 549 switch (c) { 550 case -1: 551 return(-1); 552 case '\n': 553 case '\r': 554 kprintf("\n"); 555 *lp++ = '\0'; 556 return(0); 557 case '\b': 558 case '\177': 559 if (lp > cp) { 560 kprintf("\b \b"); 561 lp--; 562 } else { 563 kprintf("%c", 7); 564 } 565 continue; 566 case '#': 567 kprintf("#"); 568 lp--; 569 if (lp < cp) 570 lp = cp; 571 continue; 572 case '@': 573 case 'u' & 037: 574 lp = cp; 575 kprintf("%c", '\n'); 576 continue; 577 default: 578 if (lp - cp >= limit - 1) { 579 kprintf("%c", 7); 580 } else { 581 kprintf("%c", c); 582 *lp++ = c; 583 } 584 continue; 585 } 586 } 587 } 588 589 /* 590 * Convert a given name to the cdev_t of the disk-like device 591 * it refers to. 592 */ 593 cdev_t 594 kgetdiskbyname(const char *name) 595 { 596 cdev_t rdev; 597 598 /* 599 * Get the base name of the device 600 */ 601 if (strncmp(name, __SYS_PATH_DEV, sizeof(__SYS_PATH_DEV) - 1) == 0) 602 name += sizeof(__SYS_PATH_DEV) - 1; 603 604 /* 605 * Locate the device 606 */ 607 rdev = devfs_find_device_by_name("%s", name); 608 if (rdev == NULL) { 609 kprintf("no disk named '%s'\n", name); 610 } 611 /* 612 * FOUND DEVICE 613 */ 614 return(rdev); 615 } 616 617 /* 618 * Set rootdev to match (name), given that we expect it to 619 * refer to a disk-like device. 620 */ 621 static int 622 setrootbyname(char *name) 623 { 624 cdev_t diskdev; 625 626 diskdev = kgetdiskbyname(name); 627 if (diskdev != NULL) { 628 rootdev = diskdev; 629 return (0); 630 } 631 /* set to NULL if kgetdiskbyname() fails so that if the first rootdev is 632 * found by fails to mount and the second one isn't found, mountroot_try 633 * doesn't try again with the first one 634 */ 635 rootdev = NULL; 636 return (1); 637 } 638 639 #ifdef DDB 640 DB_SHOW_COMMAND(disk, db_getdiskbyname) 641 { 642 cdev_t dev; 643 644 if (modif[0] == '\0') { 645 db_error("usage: show disk/devicename"); 646 return; 647 } 648 dev = kgetdiskbyname(modif); 649 if (dev != NULL) 650 db_printf("cdev_t = %p\n", dev); 651 else 652 db_printf("No disk device matched.\n"); 653 } 654 #endif 655 656 static int 657 vfs_sysctl_real_root(SYSCTL_HANDLER_ARGS) 658 { 659 char *real_root; 660 size_t len; 661 int error; 662 663 real_root = kgetenv("vfs.root.realroot"); 664 665 if (real_root == NULL) 666 real_root = ""; 667 668 len = strlen(real_root) + 1; 669 670 error = sysctl_handle_string(oidp, real_root, len, req); 671 672 return error; 673 } 674 675 SYSCTL_PROC(_vfs, OID_AUTO, real_root, 676 CTLTYPE_STRING | CTLFLAG_RD, 0, 0, vfs_sysctl_real_root, 677 "A", "Real root mount string"); 678