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 <sys/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",   /* 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
 219         /*
 220          * Lookup the requested path and extract the nch and vnode.
 221          */
 222         error = nlookup_init_raw(&nd,
 223              "/dev", UIO_SYSSPACE, NLC_FOLLOW,
 224              cred, &rootnch);
 225
 226         if (error == 0) {
 227                 devfs_debug(DEVFS_DEBUG_DEBUG, "vfs_mountroot_devfs: nlookup_init is ok...\n");
 228                 if ((error = nlookup(&nd)) == 0) {
 229                         devfs_debug(DEVFS_DEBUG_DEBUG, "vfs_mountroot_devfs: nlookup is ok...\n");
 230                         if (nd.nl_nch.ncp->nc_vp == NULL) {
 231                                 devfs_debug(DEVFS_DEBUG_SHOW, "vfs_mountroot_devfs: nlookup: simply not found\n");
 232                                 error = ENOENT;
 233                         }
 234                 }
 235         }
 236         if (error) {
 237                 nlookup_done(&nd);
 238                 devfs_debug(DEVFS_DEBUG_SHOW, "vfs_mountroot_devfs: nlookup failed, error: %d\n", error);
 239                 return (error);
 240         }
 241
 242         /*
 243          * Extract the locked+refd ncp and cleanup the nd structure
 244          */
 245         nch = nd.nl_nch;
 246         cache_zero(&nd.nl_nch);
 247         nlookup_done(&nd);
 248
 249         /*
 250          * now we have the locked ref'd nch and unreferenced vnode.
 251          */
 252         vp = nch.ncp->nc_vp;
 253         if ((error = vget(vp, LK_EXCLUSIVE)) != 0) {
 254                 cache_put(&nch);
 255                 devfs_debug(DEVFS_DEBUG_SHOW, "vfs_mountroot_devfs: vget failed\n");
 256                 return (error);
 257         }
 258         cache_unlock(&nch);
 259
 260         if ((error = vinvalbuf(vp, V_SAVE, 0, 0)) != 0) {
 261                 cache_drop(&nch);
 262                 vput(vp);
 263                 devfs_debug(DEVFS_DEBUG_SHOW, "vfs_mountroot_devfs: vinvalbuf failed\n");
 264                 return (error);
 265         }
 266         if (vp->v_type != VDIR) {
 267                 cache_drop(&nch);
 268                 vput(vp);
 269                 devfs_debug(DEVFS_DEBUG_SHOW, "vfs_mountroot_devfs: vp is not VDIR\n");
 270                 return (ENOTDIR);
 271         }
 272
 273         vfsp = vfsconf_find_by_name("devfs");
 274         vp->v_flag |= VMOUNT;
 275
 276         /*
 277          * Allocate and initialize the filesystem.
 278          */
 279         mp = kmalloc(sizeof(struct mount), M_MOUNT, M_ZERO|M_WAITOK);
 280         TAILQ_INIT(&mp->mnt_nvnodelist);
 281         TAILQ_INIT(&mp->mnt_reservedvnlist);
 282         TAILQ_INIT(&mp->mnt_jlist);
 283         mp->mnt_nvnodelistsize = 0;
 284         lockinit(&mp->mnt_lock, "vfslock", 0, 0);
 285         vfs_busy(mp, LK_NOWAIT);
 286         mp->mnt_op = vfsp->vfc_vfsops;
 287         mp->mnt_vfc = vfsp;
 288         vfsp->vfc_refcount++;
 289         mp->mnt_stat.f_type = vfsp->vfc_typenum;
 290         mp->mnt_flag |= vfsp->vfc_flags & MNT_VISFLAGMASK;
 291         strncpy(mp->mnt_stat.f_fstypename, vfsp->vfc_name, MFSNAMELEN);
 292         mp->mnt_stat.f_owner = cred->cr_uid;
 293         mp->mnt_iosize_max = DFLTPHYS;
 294         vn_unlock(vp);
 295
 296         /*
 297          * Mount the filesystem.
 298          */
 299         error = VFS_MOUNT(mp, "/dev", NULL, cred);
 300
 301         vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
 302
 303         /*
 304          * Put the new filesystem on the mount list after root.  The mount
 305          * point gets its own mnt_ncmountpt (unless the VFS already set one
 306          * up) which represents the root of the mount.  The lookup code
 307          * detects the mount point going forward and checks the root of
 308          * the mount going backwards.
 309          *
 310          * It is not necessary to invalidate or purge the vnode underneath
 311          * because elements under the mount will be given their own glue
 312          * namecache record.
 313          */
 314         if (!error) {
 315                 if (mp->mnt_ncmountpt.ncp == NULL) {
 316                         /*
 317                          * allocate, then unlock, but leave the ref intact
 318                          */
 319                         cache_allocroot(&mp->mnt_ncmountpt, mp, NULL);
 320                         cache_unlock(&mp->mnt_ncmountpt);
 321                 }
 322                 mp->mnt_ncmounton = nch;                /* inherits ref */
 323                 nch.ncp->nc_flag |= NCF_ISMOUNTPT;
 324
 325                 /* XXX get the root of the fs and cache_setvp(mnt_ncmountpt...) */
 326                 vp->v_flag &= ~VMOUNT;
 327                 mountlist_insert(mp, MNTINS_LAST);
 328                 vn_unlock(vp);
 329                 //checkdirs(&mp->mnt_ncmounton, &mp->mnt_ncmountpt);
 330                 error = vfs_allocate_syncvnode(mp);
 331                 if (error) {
 332                         devfs_debug(DEVFS_DEBUG_SHOW, "vfs_mountroot_devfs: vfs_allocate_syncvnode failed\n");
 333                 }
 334                 vfs_unbusy(mp);
 335                 error = VFS_START(mp, 0);
 336                 vrele(vp);
 337         } else {
 338                 vfs_rm_vnodeops(mp, NULL, &mp->mnt_vn_coherency_ops);
 339                 vfs_rm_vnodeops(mp, NULL, &mp->mnt_vn_journal_ops);
 340                 vfs_rm_vnodeops(mp, NULL, &mp->mnt_vn_norm_ops);
 341                 vfs_rm_vnodeops(mp, NULL, &mp->mnt_vn_spec_ops);
 342                 vfs_rm_vnodeops(mp, NULL, &mp->mnt_vn_fifo_ops);
 343                 vp->v_flag &= ~VMOUNT;
 344                 mp->mnt_vfc->vfc_refcount--;
 345                 vfs_unbusy(mp);
 346                 kfree(mp, M_MOUNT);
 347                 cache_drop(&nch);
 348                 vput(vp);
 349                 devfs_debug(DEVFS_DEBUG_SHOW, "vfs_mountroot_devfs: mount failed\n");
 350         }
 351
 352         devfs_debug(DEVFS_DEBUG_DEBUG, "rootmount_devfs done with error: %d\n", error);
 353         return (error);
 354 }
 355
 356
 357 /*
 358  * Mount (mountfrom) as the root filesystem.
 359  */
 360 static int
 361 vfs_mountroot_try(const char *mountfrom)
 362 {
 363         struct mount    *mp, *mp2;
 364         char            *vfsname, *devname;
 365         int             error;
 366         char            patt[32];
 367
 368         vfsname = NULL;
 369         devname = NULL;
 370         mp      = NULL;
 371         mp2             = NULL;
 372         error   = EINVAL;
 373
 374         if (mountfrom == NULL)
 375                 return(error);          /* don't complain */
 376
 377         crit_enter();
 378         kprintf("Mounting root from %s\n", mountfrom);
 379         crit_exit();
 380
 381         /* parse vfs name and devname */
 382         vfsname = kmalloc(MFSNAMELEN, M_MOUNT, M_WAITOK);
 383         devname = kmalloc(MNAMELEN, M_MOUNT, M_WAITOK);
 384         vfsname[0] = devname[0] = 0;
 385         ksprintf(patt, "%%%d[a-z0-9]:%%%ds", MFSNAMELEN, MNAMELEN);
 386         if (ksscanf(mountfrom, patt, vfsname, devname) < 1)
 387                 goto done;
 388
 389         /* allocate a root mount */
 390         error = vfs_rootmountalloc(vfsname,
 391                         devname[0] != 0 ? devname : ROOTNAME, &mp);
 392         if (error != 0) {
 393                 kprintf("Can't allocate root mount for filesystem '%s': %d\n",
 394                        vfsname, error);
 395                 goto done;
 396         }
 397         mp->mnt_flag |= MNT_ROOTFS;
 398
 399         /* do our best to set rootdev */
 400         if ((devname[0] != 0) && setrootbyname(devname))
 401                 kprintf("setrootbyname failed\n");
 402
 403         /* If the root device is a type "memory disk", mount RW */
 404         if (rootdev != NULL && dev_is_good(rootdev) &&
 405             (dev_dflags(rootdev) & D_MEMDISK)) {
 406                 mp->mnt_flag &= ~MNT_RDONLY;
 407         }
 408
 409         error = VFS_MOUNT(mp, NULL, NULL, proc0.p_ucred);
 410
 411         if (!error) {
 412                 //kprintf("Trying vfs_mountroot_devfs!\n");
 413                 //vfs_mountroot_devfs();
 414         }
 415
 416 done:
 417         if (vfsname != NULL)
 418                 kfree(vfsname, M_MOUNT);
 419         if (devname != NULL)
 420                 kfree(devname, M_MOUNT);
 421         if (error == 0) {
 422                 /* register with list of mounted filesystems */
 423                 mountlist_insert(mp, MNTINS_FIRST);
 424
 425                 /* sanity check system clock against root fs timestamp */
 426                 inittodr(mp->mnt_time);
 427                 vfs_unbusy(mp);
 428                 if (mp->mnt_syncer == NULL) {
 429                         error = vfs_allocate_syncvnode(mp);
 430                         if (error)
 431                                 kprintf("Warning: no syncer vp for root!\n");
 432                         error = 0;
 433                 }
 434         } else {
 435                 if (mp != NULL) {
 436                         vfs_unbusy(mp);
 437                         kfree(mp, M_MOUNT);
 438                 }
 439                 kprintf("Root mount failed: %d\n", error);
 440         }
 441         return(error);
 442 }
 443
 444
 445 static void vfs_mountroot_ask_callback(cdev_t);
 446
 447 /*
 448  * Spin prompting on the console for a suitable root filesystem
 449  */
 450
 451 static int
 452 vfs_mountroot_ask(void)
 453 {
 454         char name[128];
 455         int llimit = 100;
 456
 457         kprintf("\nManual root filesystem specification:\n");
 458         kprintf("  <fstype>:<device>  Specify root (e.g. ufs:da0s1a)\n");
 459         kprintf("  ?                  List valid disk boot devices\n");
 460         kprintf("  panic              Just panic\n");
 461         kprintf("  abort              Abort manual input\n");
 462         while (llimit--) {
 463                 kprintf("\nmountroot> ");
 464
 465                 if (getline(name, 128) < 0)
 466                         break;
 467                 if (name[0] == 0) {
 468                         ;
 469                 } else if (name[0] == '?') {
 470                         kprintf("Possibly valid devices for root FS:\n");
 471                         //enumerate all disk devices
 472                         devfs_scan_callback(vfs_mountroot_ask_callback);
 473                         kprintf("\n");
 474                         continue;
 475                 } else if (strcmp(name, "panic") == 0) {
 476                         panic("panic from console");
 477                 } else if (strcmp(name, "abort") == 0) {
 478                         break;
 479                 } else if (vfs_mountroot_try(name) == 0) {
 480                         return(0);
 481                 }
 482         }
 483         return(1);
 484 }
 485
 486
 487 static void
 488 vfs_mountroot_ask_callback(cdev_t dev)
 489 {
 490         if (dev_is_good(dev) && (dev_dflags(dev) & D_DISK))
 491                 kprintf(" \"%s\" ", dev->si_name);
 492 }
 493
 494
 495 static int
 496 getline(char *cp, int limit)
 497 {
 498         char *lp;
 499         int c;
 500
 501         lp = cp;
 502         for (;;) {
 503                 c = cngetc();
 504
 505                 switch (c) {
 506                 case -1:
 507                         return(-1);
 508                 case '\n':
 509                 case '\r':
 510                         kprintf("\n");
 511                         *lp++ = '\0';
 512                         return(0);
 513                 case '\b':
 514                 case '\177':
 515                         if (lp > cp) {
 516                                 kprintf("\b \b");
 517                                 lp--;
 518                         } else {
 519                                 kprintf("%c", 7);
 520                         }
 521                         continue;
 522                 case '#':
 523                         kprintf("#");
 524                         lp--;
 525                         if (lp < cp)
 526                                 lp = cp;
 527                         continue;
 528                 case '@':
 529                 case 'u' & 037:
 530                         lp = cp;
 531                         kprintf("%c", '\n');
 532                         continue;
 533                 default:
 534                         if (lp - cp >= limit - 1) {
 535                                 kprintf("%c", 7);
 536                         } else {
 537                                 kprintf("%c", c);
 538                                 *lp++ = c;
 539                         }
 540                         continue;
 541                 }
 542         }
 543 }
 544
 545 /*
 546  * Convert a given name to the cdev_t of the disk-like device
 547  * it refers to.
 548  */
 549 struct kdbn_info {
 550         const char *name;
 551         int nlen;
 552         int minor;
 553         cdev_t dev;
 554 };
 555
 556
 557 cdev_t
 558 kgetdiskbyname(const char *name)
 559 {
 560         char *cp;
 561         cdev_t rdev;
 562
 563         /*
 564          * Get the base name of the device
 565          */
 566         if (strncmp(name, __SYS_PATH_DEV, sizeof(__SYS_PATH_DEV) - 1) == 0)
 567                 name += sizeof(__SYS_PATH_DEV) - 1;
 568         cp = __DECONST(char *, name);
 569
 570         /*
 571          * Locate the device
 572          */
 573         kprintf("tryroot %s\n", name);
 574         rdev = devfs_find_device_by_name(name);
 575         if (rdev == NULL) {
 576                 kprintf("no disk named '%s'\n", name);
 577         }
 578         /*
 579          * FOUND DEVICE
 580          */
 581         return(rdev);
 582 }
 583
 584 /*
 585  * Set rootdev to match (name), given that we expect it to
 586  * refer to a disk-like device.
 587  */
 588 static int
 589 setrootbyname(char *name)
 590 {
 591         cdev_t diskdev;
 592
 593         diskdev = kgetdiskbyname(name);
 594         if (diskdev != NULL) {
 595                 rootdev = diskdev;
 596                 return (0);
 597         }
 598         /* set to NULL if kgetdiskbyname() fails so that if the first rootdev is
 599          * found by fails to mount and the second one isn't found, mountroot_try
 600          * doesn't try again with the first one
 601          */
 602         rootdev = NULL;
 603         return (1);
 604 }
 605
 606 #ifdef DDB
 607 DB_SHOW_COMMAND(disk, db_getdiskbyname)
 608 {
 609         cdev_t dev;
 610
 611         if (modif[0] == '\0') {
 612                 db_error("usage: show disk/devicename");
 613                 return;
 614         }
 615         dev = kgetdiskbyname(modif);
 616         if (dev != NULL)
 617                 db_printf("cdev_t = %p\n", dev);
 618         else
 619                 db_printf("No disk device matched.\n");
 620 }
 621 #endif