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         vsetflags(vp, VMOUNT);
 275
 276         /*
 277          * Allocate and initialize the filesystem.
 278          */
 279         mp = kmalloc(sizeof(struct mount), M_MOUNT, M_ZERO|M_WAITOK);
 280         mount_init(mp);
 281         vfs_busy(mp, LK_NOWAIT);
 282         mp->mnt_op = vfsp->vfc_vfsops;
 283         mp->mnt_vfc = vfsp;
 284         vfsp->vfc_refcount++;
 285         mp->mnt_stat.f_type = vfsp->vfc_typenum;
 286         mp->mnt_flag |= vfsp->vfc_flags & MNT_VISFLAGMASK;
 287         strncpy(mp->mnt_stat.f_fstypename, vfsp->vfc_name, MFSNAMELEN);
 288         mp->mnt_stat.f_owner = cred->cr_uid;
 289         vn_unlock(vp);
 290
 291         /*
 292          * Mount the filesystem.
 293          */
 294         error = VFS_MOUNT(mp, "/dev", NULL, cred);
 295
 296         vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
 297
 298         /*
 299          * Put the new filesystem on the mount list after root.  The mount
 300          * point gets its own mnt_ncmountpt (unless the VFS already set one
 301          * up) which represents the root of the mount.  The lookup code
 302          * detects the mount point going forward and checks the root of
 303          * the mount going backwards.
 304          *
 305          * It is not necessary to invalidate or purge the vnode underneath
 306          * because elements under the mount will be given their own glue
 307          * namecache record.
 308          */
 309         if (!error) {
 310                 if (mp->mnt_ncmountpt.ncp == NULL) {
 311                         /*
 312                          * allocate, then unlock, but leave the ref intact
 313                          */
 314                         cache_allocroot(&mp->mnt_ncmountpt, mp, NULL);
 315                         cache_unlock(&mp->mnt_ncmountpt);
 316                 }
 317                 mp->mnt_ncmounton = nch;                /* inherits ref */
 318                 nch.ncp->nc_flag |= NCF_ISMOUNTPT;
 319
 320                 /* XXX get the root of the fs and cache_setvp(mnt_ncmountpt...) */
 321                 vclrflags(vp, VMOUNT);
 322                 mountlist_insert(mp, MNTINS_LAST);
 323                 vn_unlock(vp);
 324                 //checkdirs(&mp->mnt_ncmounton, &mp->mnt_ncmountpt);
 325                 error = vfs_allocate_syncvnode(mp);
 326                 if (error) {
 327                         devfs_debug(DEVFS_DEBUG_SHOW, "vfs_mountroot_devfs: vfs_allocate_syncvnode failed\n");
 328                 }
 329                 vfs_unbusy(mp);
 330                 error = VFS_START(mp, 0);
 331                 vrele(vp);
 332         } else {
 333                 vfs_rm_vnodeops(mp, NULL, &mp->mnt_vn_coherency_ops);
 334                 vfs_rm_vnodeops(mp, NULL, &mp->mnt_vn_journal_ops);
 335                 vfs_rm_vnodeops(mp, NULL, &mp->mnt_vn_norm_ops);
 336                 vfs_rm_vnodeops(mp, NULL, &mp->mnt_vn_spec_ops);
 337                 vfs_rm_vnodeops(mp, NULL, &mp->mnt_vn_fifo_ops);
 338                 vclrflags(vp, VMOUNT);
 339                 mp->mnt_vfc->vfc_refcount--;
 340                 vfs_unbusy(mp);
 341                 kfree(mp, M_MOUNT);
 342                 cache_drop(&nch);
 343                 vput(vp);
 344                 devfs_debug(DEVFS_DEBUG_SHOW, "vfs_mountroot_devfs: mount failed\n");
 345         }
 346
 347         devfs_debug(DEVFS_DEBUG_DEBUG, "rootmount_devfs done with error: %d\n", error);
 348         return (error);
 349 }
 350
 351
 352 /*
 353  * Mount (mountfrom) as the root filesystem.
 354  */
 355 static int
 356 vfs_mountroot_try(const char *mountfrom)
 357 {
 358         struct mount    *mp, *mp2;
 359         char            *vfsname, *devname;
 360         int             error;
 361         char            patt[32];
 362         int             mountfromlen, len;
 363         char            *cp, *ep, *mf;
 364
 365         vfsname = NULL;
 366         devname = NULL;
 367         mp      = NULL;
 368         mp2             = NULL;
 369         error   = EINVAL;
 370
 371         if (mountfrom == NULL)
 372                 return(error);          /* don't complain */
 373
 374         crit_enter();
 375         kprintf("Mounting root from %s\n", mountfrom);
 376         crit_exit();
 377
 378         mountfromlen = strlen(mountfrom);
 379         cp = (char*)mountfrom;
 380         /* parse vfs name and devname */
 381         vfsname = kmalloc(MFSNAMELEN, M_MOUNT, M_WAITOK);
 382         devname = kmalloc(MNAMELEN, M_MOUNT, M_WAITOK);
 383         mf = kmalloc(MFSNAMELEN+MNAMELEN, M_MOUNT, M_WAITOK);
 384         for(;;) {
 385                 for (ep = cp; (*ep != 0) && (*ep != ';'); ep++);
 386                 len = ep - cp;
 387                 bzero(vfsname, MFSNAMELEN);
 388                 bzero(devname, MNAMELEN);
 389                 bzero(mf, MFSNAMELEN+MNAMELEN);
 390                 strncpy(mf, cp, MFSNAMELEN+MNAMELEN);
 391
 392                 vfsname[0] = devname[0] = 0;
 393                 ksprintf(patt, "%%%d[a-z0-9]:%%%ds", MFSNAMELEN, MNAMELEN);
 394                 if (ksscanf(mf, patt, vfsname, devname) < 1)
 395                         goto end;
 396
 397                 /* allocate a root mount */
 398                 error = vfs_rootmountalloc(vfsname,
 399                                 devname[0] != 0 ? devname : ROOTNAME, &mp);
 400                 if (error != 0) {
 401                         kprintf("Can't allocate root mount for filesystem '%s': %d\n",
 402                                vfsname, error);
 403                         goto end;
 404                 }
 405                 mp->mnt_flag |= MNT_ROOTFS;
 406
 407                 /* do our best to set rootdev */
 408                 if ((strcmp(vfsname, "hammer") != 0) && (devname[0] != 0) &&
 409                     setrootbyname(devname))
 410                         kprintf("setrootbyname failed\n");
 411
 412                 /* If the root device is a type "memory disk", mount RW */
 413                 if (rootdev != NULL && dev_is_good(rootdev) &&
 414                     (dev_dflags(rootdev) & D_MEMDISK)) {
 415                         mp->mnt_flag &= ~MNT_RDONLY;
 416                 }
 417
 418                 error = VFS_MOUNT(mp, NULL, NULL, proc0.p_ucred);
 419
 420                 if (!error)
 421                         break;
 422 end:
 423                 if(*ep == 0)
 424                         break;
 425                 cp = ep + 1;
 426         }
 427
 428         if (vfsname != NULL)
 429                 kfree(vfsname, M_MOUNT);
 430         if (devname != NULL)
 431                 kfree(devname, M_MOUNT);
 432         if (mf != NULL)
 433                 kfree(mf, M_MOUNT);
 434         if (error == 0) {
 435                 /* register with list of mounted filesystems */
 436                 mountlist_insert(mp, MNTINS_FIRST);
 437
 438                 /* sanity check system clock against root fs timestamp */
 439                 inittodr(mp->mnt_time);
 440                 vfs_unbusy(mp);
 441                 if (mp->mnt_syncer == NULL) {
 442                         error = vfs_allocate_syncvnode(mp);
 443                         if (error)
 444                                 kprintf("Warning: no syncer vp for root!\n");
 445                         error = 0;
 446                 }
 447         } else {
 448                 if (mp != NULL) {
 449                         vfs_unbusy(mp);
 450                         kfree(mp, M_MOUNT);
 451                 }
 452                 kprintf("Root mount failed: %d\n", error);
 453         }
 454         return(error);
 455 }
 456
 457
 458 static void vfs_mountroot_ask_callback(cdev_t);
 459
 460 /*
 461  * Spin prompting on the console for a suitable root filesystem
 462  */
 463
 464 static int
 465 vfs_mountroot_ask(void)
 466 {
 467         char name[128];
 468         int llimit = 100;
 469
 470         kprintf("\nManual root filesystem specification:\n");
 471         kprintf("  <fstype>:<device>  Specify root (e.g. ufs:da0s1a)\n");
 472         kprintf("  ?                  List valid disk boot devices\n");
 473         kprintf("  panic              Just panic\n");
 474         kprintf("  abort              Abort manual input\n");
 475         while (llimit--) {
 476                 kprintf("\nmountroot> ");
 477
 478                 if (getline(name, 128) < 0)
 479                         break;
 480                 if (name[0] == 0) {
 481                         ;
 482                 } else if (name[0] == '?') {
 483                         kprintf("Possibly valid devices for root FS:\n");
 484                         //enumerate all disk devices
 485                         devfs_scan_callback(vfs_mountroot_ask_callback);
 486                         kprintf("\n");
 487                         continue;
 488                 } else if (strcmp(name, "panic") == 0) {
 489                         panic("panic from console");
 490                 } else if (strcmp(name, "abort") == 0) {
 491                         break;
 492                 } else if (vfs_mountroot_try(name) == 0) {
 493                         return(0);
 494                 }
 495         }
 496         return(1);
 497 }
 498
 499
 500 static void
 501 vfs_mountroot_ask_callback(cdev_t dev)
 502 {
 503         if (dev_is_good(dev) && (dev_dflags(dev) & D_DISK))
 504                 kprintf(" \"%s\" ", dev->si_name);
 505 }
 506
 507
 508 static int
 509 getline(char *cp, int limit)
 510 {
 511         char *lp;
 512         int c;
 513
 514         lp = cp;
 515         for (;;) {
 516                 c = cngetc();
 517
 518                 switch (c) {
 519                 case -1:
 520                         return(-1);
 521                 case '\n':
 522                 case '\r':
 523                         kprintf("\n");
 524                         *lp++ = '\0';
 525                         return(0);
 526                 case '\b':
 527                 case '\177':
 528                         if (lp > cp) {
 529                                 kprintf("\b \b");
 530                                 lp--;
 531                         } else {
 532                                 kprintf("%c", 7);
 533                         }
 534                         continue;
 535                 case '#':
 536                         kprintf("#");
 537                         lp--;
 538                         if (lp < cp)
 539                                 lp = cp;
 540                         continue;
 541                 case '@':
 542                 case 'u' & 037:
 543                         lp = cp;
 544                         kprintf("%c", '\n');
 545                         continue;
 546                 default:
 547                         if (lp - cp >= limit - 1) {
 548                                 kprintf("%c", 7);
 549                         } else {
 550                                 kprintf("%c", c);
 551                                 *lp++ = c;
 552                         }
 553                         continue;
 554                 }
 555         }
 556 }
 557
 558 /*
 559  * Convert a given name to the cdev_t of the disk-like device
 560  * it refers to.
 561  */
 562 struct kdbn_info {
 563         const char *name;
 564         int nlen;
 565         int minor;
 566         cdev_t dev;
 567 };
 568
 569
 570 cdev_t
 571 kgetdiskbyname(const char *name)
 572 {
 573         char *cp;
 574         cdev_t rdev;
 575
 576         /*
 577          * Get the base name of the device
 578          */
 579         if (strncmp(name, __SYS_PATH_DEV, sizeof(__SYS_PATH_DEV) - 1) == 0)
 580                 name += sizeof(__SYS_PATH_DEV) - 1;
 581         cp = __DECONST(char *, name);
 582
 583         /*
 584          * Locate the device
 585          */
 586         kprintf("tryroot %s\n", name);
 587         rdev = devfs_find_device_by_name(name);
 588         if (rdev == NULL) {
 589                 kprintf("no disk named '%s'\n", name);
 590         }
 591         /*
 592          * FOUND DEVICE
 593          */
 594         return(rdev);
 595 }
 596
 597 /*
 598  * Set rootdev to match (name), given that we expect it to
 599  * refer to a disk-like device.
 600  */
 601 static int
 602 setrootbyname(char *name)
 603 {
 604         cdev_t diskdev;
 605
 606         diskdev = kgetdiskbyname(name);
 607         if (diskdev != NULL) {
 608                 rootdev = diskdev;
 609                 return (0);
 610         }
 611         /* set to NULL if kgetdiskbyname() fails so that if the first rootdev is
 612          * found by fails to mount and the second one isn't found, mountroot_try
 613          * doesn't try again with the first one
 614          */
 615         rootdev = NULL;
 616         return (1);
 617 }
 618
 619 #ifdef DDB
 620 DB_SHOW_COMMAND(disk, db_getdiskbyname)
 621 {
 622         cdev_t dev;
 623
 624         if (modif[0] == '\0') {
 625                 db_error("usage: show disk/devicename");
 626                 return;
 627         }
 628         dev = kgetdiskbyname(modif);
 629         if (dev != NULL)
 630                 db_printf("cdev_t = %p\n", dev);
 631         else
 632                 db_printf("No disk device matched.\n");
 633 }
 634 #endif