sys/kern/vfs_conf.c

   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  *      $DragonFly: src/sys/kern/vfs_conf.c,v 1.34 2008/05/24 19:08:28 dillon Exp $
  30  */
  31
  32 /*
  33  * Locate and mount the root filesystem.
  34  *
  35  * The root filesystem is detailed in the kernel environment variable
  36  * vfs.root.mountfrom, which is expected to be in the general format
  37  *
  38  * <vfsname>:[<path>]
  39  * vfsname   := the name of a VFS known to the kernel and capable
  40  *              of being mounted as root
  41  * path      := disk device name or other data used by the filesystem
  42  *              to locate its physical store
  43  *
  44  */
  45
  46 #include "opt_rootdevname.h"
  47
  48 #include <sys/param.h>
  49 #include <sys/kernel.h>
  50 #include <sys/systm.h>
  51 #include <sys/proc.h>
  52 #include <sys/vnode.h>
  53 #include <sys/mount.h>
  54 #include <sys/malloc.h>
  55 #include <sys/reboot.h>
  56 #include <sys/diskslice.h>
  57 #include <sys/conf.h>
  58 #include <sys/cons.h>
  59 #include <sys/device.h>
  60 #include <sys/disk.h>
  61 #include <sys/namecache.h>
  62 #include <sys/paths.h>
  63 #include <sys/thread2.h>
  64 #include <sys/nlookup.h>
  65 #include <vfs/devfs/devfs.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",
  86         "cd9660:acd0",
  87         "cd9660:cd1",
  88         "cd9660:acd1",
  89         NULL
  90 };
  91
  92 int vfs_mountroot_devfs(void);
  93 static void     vfs_mountroot(void *junk);
  94 static int      vfs_mountroot_try(const char *mountfrom);
  95 static int      vfs_mountroot_ask(void);
  96 static int      getline(char *cp, int limit);
  97
  98 /* legacy find-root code */
  99 char            *rootdevnames[2] = {NULL, NULL};
 100 static int      setrootbyname(char *name);
 101
 102 SYSINIT(mountroot, SI_SUB_MOUNT_ROOT, SI_ORDER_SECOND, vfs_mountroot, NULL);
 103
 104 /*
 105  * Find and mount the root filesystem
 106  */
 107 static void
 108 vfs_mountroot(void *junk)
 109 {
 110         int     i;
 111         cdev_t  save_rootdev = rootdev;
 112
 113         /*
 114          * Make sure all disk devices created so far have also been probed,
 115          * and also make sure that the newly created device nodes for
 116          * probed disks are ready, too.
 117          */
 118         disk_config(NULL);
 119         devfs_config();
 120
 121         /*
 122          * The root filesystem information is compiled in, and we are
 123          * booted with instructions to use it.
 124          */
 125 #ifdef ROOTDEVNAME
 126         if ((boothowto & RB_DFLTROOT) &&
 127             !vfs_mountroot_try(ROOTDEVNAME))
 128                 return;
 129 #endif
 130         /*
 131          * We are booted with instructions to prompt for the root filesystem,
 132          * or to use the compiled-in default when it doesn't exist.
 133          */
 134         if (boothowto & (RB_DFLTROOT | RB_ASKNAME)) {
 135                 if (!vfs_mountroot_ask())
 136                         return;
 137         }
 138
 139         /*
 140          * We've been given the generic "use CDROM as root" flag.  This is
 141          * necessary because one media may be used in many different
 142          * devices, so we need to search for them.
 143          */
 144         if (boothowto & RB_CDROM) {
 145                 for (i = 0; cdrom_rootdevnames[i] != NULL; i++) {
 146                         if (!vfs_mountroot_try(cdrom_rootdevnames[i]))
 147                                 return;
 148                 }
 149         }
 150
 151         /*
 152          * Try to use the value read by the loader from /etc/fstab, or
 153          * supplied via some other means.  This is the preferred
 154          * mechanism.
 155          */
 156         if (!vfs_mountroot_try(kgetenv("vfs.root.mountfrom")))
 157                 return;
 158
 159         /*
 160          * If a vfs set rootdev, try it (XXX VINUM HACK!)
 161          */
 162         if (save_rootdev != NULL) {
 163                 rootdev = save_rootdev;
 164                 if (!vfs_mountroot_try(""))
 165                         return;
 166         }
 167
 168         /*
 169          * Try values that may have been computed by the machine-dependant
 170          * legacy code.
 171          */
 172         if (rootdevnames[0] && !vfs_mountroot_try(rootdevnames[0]))
 173                 return;
 174         if (rootdevnames[1] && !vfs_mountroot_try(rootdevnames[1]))
 175                 return;
 176
 177         /*
 178          * If we have a compiled-in default, and haven't already tried it, try
 179          * it now.
 180          */
 181 #ifdef ROOTDEVNAME
 182         if (!(boothowto & RB_DFLTROOT))
 183                 if (!vfs_mountroot_try(ROOTDEVNAME))
 184                         return;
 185 #endif
 186
 187         /*
 188          * Everything so far has failed, prompt on the console if we haven't
 189          * already tried that.
 190          */
 191         if (!(boothowto & (RB_DFLTROOT | RB_ASKNAME)) && !vfs_mountroot_ask())
 192                 return;
 193         panic("Root mount failed, startup aborted.");
 194 }
 195
 196
 197 int
 198 vfs_mountroot_devfs(void)
 199 {
 200         struct vnode *vp;
 201         struct nchandle nch;
 202         struct nlookupdata nd;
 203         struct mount *mp;
 204         struct vfsconf *vfsp;
 205         int error;
 206         struct ucred *cred = proc0.p_ucred;
 207
 208         /*
 209          * Lookup the requested path and extract the nch and vnode.
 210          */
 211         error = nlookup_init_raw(&nd,
 212              "/dev", UIO_SYSSPACE, NLC_FOLLOW,
 213              cred, &rootnch);
 214
 215         if (error == 0) {
 216                 devfs_debug(DEVFS_DEBUG_DEBUG, "vfs_mountroot_devfs: nlookup_init is ok...\n");
 217                 if ((error = nlookup(&nd)) == 0) {
 218                         devfs_debug(DEVFS_DEBUG_DEBUG, "vfs_mountroot_devfs: nlookup is ok...\n");
 219                         if (nd.nl_nch.ncp->nc_vp == NULL) {
 220                                 devfs_debug(DEVFS_DEBUG_SHOW, "vfs_mountroot_devfs: nlookup: simply not found\n");
 221                                 error = ENOENT;
 222                         }
 223                 }
 224         }
 225         if (error) {
 226                 nlookup_done(&nd);
 227                 devfs_debug(DEVFS_DEBUG_SHOW, "vfs_mountroot_devfs: nlookup failed, error: %d\n", error);
 228                 return (error);
 229         }
 230
 231         /*
 232          * Extract the locked+refd ncp and cleanup the nd structure
 233          */
 234         nch = nd.nl_nch;
 235         cache_zero(&nd.nl_nch);
 236         nlookup_done(&nd);
 237
 238         /*
 239          * now we have the locked ref'd nch and unreferenced vnode.
 240          */
 241         vp = nch.ncp->nc_vp;
 242         if ((error = vget(vp, LK_EXCLUSIVE)) != 0) {
 243                 cache_put(&nch);
 244                 devfs_debug(DEVFS_DEBUG_SHOW, "vfs_mountroot_devfs: vget failed\n");
 245                 return (error);
 246         }
 247         cache_unlock(&nch);
 248
 249         if ((error = vinvalbuf(vp, V_SAVE, 0, 0)) != 0) {
 250                 cache_drop(&nch);
 251                 vput(vp);
 252                 devfs_debug(DEVFS_DEBUG_SHOW, "vfs_mountroot_devfs: vinvalbuf failed\n");
 253                 return (error);
 254         }
 255         if (vp->v_type != VDIR) {
 256                 cache_drop(&nch);
 257                 vput(vp);
 258                 devfs_debug(DEVFS_DEBUG_SHOW, "vfs_mountroot_devfs: vp is not VDIR\n");
 259                 return (ENOTDIR);
 260         }
 261
 262         vfsp = vfsconf_find_by_name("devfs");
 263         vp->v_flag |= VMOUNT;
 264
 265         /*
 266          * Allocate and initialize the filesystem.
 267          */
 268         mp = kmalloc(sizeof(struct mount), M_MOUNT, M_ZERO|M_WAITOK);
 269         TAILQ_INIT(&mp->mnt_nvnodelist);
 270         TAILQ_INIT(&mp->mnt_reservedvnlist);
 271         TAILQ_INIT(&mp->mnt_jlist);
 272         mp->mnt_nvnodelistsize = 0;
 273         lockinit(&mp->mnt_lock, "vfslock", 0, 0);
 274         vfs_busy(mp, LK_NOWAIT);
 275         mp->mnt_op = vfsp->vfc_vfsops;
 276         mp->mnt_vfc = vfsp;
 277         vfsp->vfc_refcount++;
 278         mp->mnt_stat.f_type = vfsp->vfc_typenum;
 279         mp->mnt_flag |= vfsp->vfc_flags & MNT_VISFLAGMASK;
 280         strncpy(mp->mnt_stat.f_fstypename, vfsp->vfc_name, MFSNAMELEN);
 281         mp->mnt_stat.f_owner = cred->cr_uid;
 282         mp->mnt_iosize_max = DFLTPHYS;
 283         vn_unlock(vp);
 284
 285         /*
 286          * Mount the filesystem.
 287          */
 288         error = VFS_MOUNT(mp, "/dev", NULL, cred);
 289
 290         vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
 291
 292         /*
 293          * Put the new filesystem on the mount list after root.  The mount
 294          * point gets its own mnt_ncmountpt (unless the VFS already set one
 295          * up) which represents the root of the mount.  The lookup code
 296          * detects the mount point going forward and checks the root of
 297          * the mount going backwards.
 298          *
 299          * It is not necessary to invalidate or purge the vnode underneath
 300          * because elements under the mount will be given their own glue
 301          * namecache record.
 302          */
 303         if (!error) {
 304                 if (mp->mnt_ncmountpt.ncp == NULL) {
 305                         /*
 306                          * allocate, then unlock, but leave the ref intact
 307                          */
 308                         cache_allocroot(&mp->mnt_ncmountpt, mp, NULL);
 309                         cache_unlock(&mp->mnt_ncmountpt);
 310                 }
 311                 mp->mnt_ncmounton = nch;                /* inherits ref */
 312                 nch.ncp->nc_flag |= NCF_ISMOUNTPT;
 313
 314                 /* XXX get the root of the fs and cache_setvp(mnt_ncmountpt...) */
 315                 vp->v_flag &= ~VMOUNT;
 316                 mountlist_insert(mp, MNTINS_LAST);
 317                 vn_unlock(vp);
 318                 //checkdirs(&mp->mnt_ncmounton, &mp->mnt_ncmountpt);
 319                 error = vfs_allocate_syncvnode(mp);
 320                 if (error) {
 321                         devfs_debug(DEVFS_DEBUG_SHOW, "vfs_mountroot_devfs: vfs_allocate_syncvnode failed\n");
 322                 }
 323                 vfs_unbusy(mp);
 324                 error = VFS_START(mp, 0);
 325                 vrele(vp);
 326         } else {
 327                 vfs_rm_vnodeops(mp, NULL, &mp->mnt_vn_coherency_ops);
 328                 vfs_rm_vnodeops(mp, NULL, &mp->mnt_vn_journal_ops);
 329                 vfs_rm_vnodeops(mp, NULL, &mp->mnt_vn_norm_ops);
 330                 vfs_rm_vnodeops(mp, NULL, &mp->mnt_vn_spec_ops);
 331                 vfs_rm_vnodeops(mp, NULL, &mp->mnt_vn_fifo_ops);
 332                 vp->v_flag &= ~VMOUNT;
 333                 mp->mnt_vfc->vfc_refcount--;
 334                 vfs_unbusy(mp);
 335                 kfree(mp, M_MOUNT);
 336                 cache_drop(&nch);
 337                 vput(vp);
 338                 devfs_debug(DEVFS_DEBUG_SHOW, "vfs_mountroot_devfs: mount failed\n");
 339         }
 340
 341         devfs_debug(DEVFS_DEBUG_DEBUG, "rootmount_devfs done with error: %d\n", error);
 342         return (error);
 343 }
 344
 345
 346 /*
 347  * Mount (mountfrom) as the root filesystem.
 348  */
 349 static int
 350 vfs_mountroot_try(const char *mountfrom)
 351 {
 352         struct mount    *mp, *mp2;
 353         char            *vfsname, *devname;
 354         int             error;
 355         char            patt[32];
 356
 357         vfsname = NULL;
 358         devname = NULL;
 359         mp      = NULL;
 360         mp2             = NULL;
 361         error   = EINVAL;
 362
 363         if (mountfrom == NULL)
 364                 return(error);          /* don't complain */
 365
 366         crit_enter();
 367         kprintf("Mounting root from %s\n", mountfrom);
 368         crit_exit();
 369
 370         /* parse vfs name and devname */
 371         vfsname = kmalloc(MFSNAMELEN, M_MOUNT, M_WAITOK);
 372         devname = kmalloc(MNAMELEN, M_MOUNT, M_WAITOK);
 373         vfsname[0] = devname[0] = 0;
 374         ksprintf(patt, "%%%d[a-z0-9]:%%%ds", MFSNAMELEN, MNAMELEN);
 375         if (ksscanf(mountfrom, patt, vfsname, devname) < 1)
 376                 goto done;
 377
 378         /* allocate a root mount */
 379         error = vfs_rootmountalloc(vfsname,
 380                         devname[0] != 0 ? devname : ROOTNAME, &mp);
 381         if (error != 0) {
 382                 kprintf("Can't allocate root mount for filesystem '%s': %d\n",
 383                        vfsname, error);
 384                 goto done;
 385         }
 386         mp->mnt_flag |= MNT_ROOTFS;
 387
 388         /* do our best to set rootdev */
 389         if ((devname[0] != 0) && setrootbyname(devname))
 390                 kprintf("setrootbyname failed\n");
 391
 392         /* If the root device is a type "memory disk", mount RW */
 393         if (rootdev != NULL && dev_is_good(rootdev) &&
 394             (dev_dflags(rootdev) & D_MEMDISK)) {
 395                 mp->mnt_flag &= ~MNT_RDONLY;
 396         }
 397
 398         error = VFS_MOUNT(mp, NULL, NULL, proc0.p_ucred);
 399
 400         if (!error) {
 401                 //kprintf("Trying vfs_mountroot_devfs!\n");
 402                 //vfs_mountroot_devfs();
 403         }
 404
 405 done:
 406         if (vfsname != NULL)
 407                 kfree(vfsname, M_MOUNT);
 408         if (devname != NULL)
 409                 kfree(devname, M_MOUNT);
 410         if (error == 0) {
 411                 /* register with list of mounted filesystems */
 412                 mountlist_insert(mp, MNTINS_FIRST);
 413
 414                 /* sanity check system clock against root fs timestamp */
 415                 inittodr(mp->mnt_time);
 416                 vfs_unbusy(mp);
 417                 if (mp->mnt_syncer == NULL) {
 418                         error = vfs_allocate_syncvnode(mp);
 419                         if (error)
 420                                 kprintf("Warning: no syncer vp for root!\n");
 421                         error = 0;
 422                 }
 423         } else {
 424                 if (mp != NULL) {
 425                         vfs_unbusy(mp);
 426                         kfree(mp, M_MOUNT);
 427                 }
 428                 kprintf("Root mount failed: %d\n", error);
 429         }
 430         return(error);
 431 }
 432
 433
 434 static void vfs_mountroot_ask_callback(cdev_t);
 435
 436 /*
 437  * Spin prompting on the console for a suitable root filesystem
 438  */
 439
 440 static int
 441 vfs_mountroot_ask(void)
 442 {
 443         char name[128];
 444         int llimit = 100;
 445
 446         kprintf("\nManual root filesystem specification:\n");
 447         kprintf("  <fstype>:<device>  Specify root (e.g. ufs:da0s1a)\n");
 448         kprintf("  ?                  List valid disk boot devices\n");
 449         kprintf("  panic              Just panic\n");
 450         kprintf("  abort              Abort manual input\n");
 451         while (llimit--) {
 452                 kprintf("\nmountroot> ");
 453
 454                 if (getline(name, 128) < 0)
 455                         break;
 456                 if (name[0] == 0) {
 457                         ;
 458                 } else if (name[0] == '?') {
 459                         kprintf("Possibly valid devices for root FS:\n");
 460                         //enumerate all disk devices
 461                         devfs_scan_callback(vfs_mountroot_ask_callback);
 462                         kprintf("\n");
 463                         continue;
 464                 } else if (strcmp(name, "panic") == 0) {
 465                         panic("panic from console");
 466                 } else if (strcmp(name, "abort") == 0) {
 467                         break;
 468                 } else if (vfs_mountroot_try(name) == 0) {
 469                         return(0);
 470                 }
 471         }
 472         return(1);
 473 }
 474
 475
 476 static void
 477 vfs_mountroot_ask_callback(cdev_t dev)
 478 {
 479         if (dev_is_good(dev) && (dev_dflags(dev) & D_DISK))
 480                 kprintf(" \"%s\" ", dev->si_name);
 481 }
 482
 483
 484 static int
 485 getline(char *cp, int limit)
 486 {
 487         char *lp;
 488         int c;
 489
 490         lp = cp;
 491         for (;;) {
 492                 c = cngetc();
 493
 494                 switch (c) {
 495                 case -1:
 496                         return(-1);
 497                 case '\n':
 498                 case '\r':
 499                         kprintf("\n");
 500                         *lp++ = '\0';
 501                         return(0);
 502                 case '\b':
 503                 case '\177':
 504                         if (lp > cp) {
 505                                 kprintf("\b \b");
 506                                 lp--;
 507                         } else {
 508                                 kprintf("%c", 7);
 509                         }
 510                         continue;
 511                 case '#':
 512                         kprintf("#");
 513                         lp--;
 514                         if (lp < cp)
 515                                 lp = cp;
 516                         continue;
 517                 case '@':
 518                 case 'u' & 037:
 519                         lp = cp;
 520                         kprintf("%c", '\n');
 521                         continue;
 522                 default:
 523                         if (lp - cp >= limit - 1) {
 524                                 kprintf("%c", 7);
 525                         } else {
 526                                 kprintf("%c", c);
 527                                 *lp++ = c;
 528                         }
 529                         continue;
 530                 }
 531         }
 532 }
 533
 534 /*
 535  * Convert a given name to the cdev_t of the disk-like device
 536  * it refers to.
 537  */
 538 struct kdbn_info {
 539         const char *name;
 540         int nlen;
 541         int minor;
 542         cdev_t dev;
 543 };
 544
 545
 546 cdev_t
 547 kgetdiskbyname(const char *name)
 548 {
 549         char *cp;
 550         cdev_t rdev;
 551
 552         /*
 553          * Get the base name of the device
 554          */
 555         if (strncmp(name, __SYS_PATH_DEV, sizeof(__SYS_PATH_DEV) - 1) == 0)
 556                 name += sizeof(__SYS_PATH_DEV) - 1;
 557         cp = __DECONST(char *, name);
 558
 559         /*
 560          * Locate the device
 561          */
 562         kprintf("tryroot %s\n", name);
 563         rdev = devfs_find_device_by_name(name);
 564         if (rdev == NULL) {
 565                 kprintf("no disk named '%s'\n", name);
 566         }
 567         /*
 568          * FOUND DEVICE
 569          */
 570         return(rdev);
 571 }
 572
 573 /*
 574  * Set rootdev to match (name), given that we expect it to
 575  * refer to a disk-like device.
 576  */
 577 static int
 578 setrootbyname(char *name)
 579 {
 580         cdev_t diskdev;
 581
 582         diskdev = kgetdiskbyname(name);
 583         if (diskdev != NULL) {
 584                 rootdev = diskdev;
 585                 return (0);
 586         }
 587         /* set to NULL if kgetdiskbyname() fails so that if the first rootdev is
 588          * found by fails to mount and the second one isn't found, mountroot_try
 589          * doesn't try again with the first one
 590          */
 591         rootdev = NULL;
 592         return (1);
 593 }
 594
 595 #ifdef DDB
 596 DB_SHOW_COMMAND(disk, db_getdiskbyname)
 597 {
 598         cdev_t dev;
 599
 600         if (modif[0] == '\0') {
 601                 db_error("usage: show disk/devicename");
 602                 return;
 603         }
 604         dev = kgetdiskbyname(modif);
 605         if (dev != NULL)
 606                 db_printf("cdev_t = %p\n", dev);
 607         else
 608                 db_printf("No disk device matched.\n");
 609 }
 610 #endif