fs/jfs/jfs_mount.c

   1 /*
   2  *   Copyright (c) International Business Machines Corp., 2000-2003
   3  *
   4  *   This program is free software;  you can redistribute it and/or modify
   5  *   it under the terms of the GNU General Public License as published by
   6  *   the Free Software Foundation; either version 2 of the License, or
   7  *   (at your option) any later version.
   8  *
   9  *   This program is distributed in the hope that it will be useful,
  10  *   but WITHOUT ANY WARRANTY;  without even the implied warranty of
  11  *   MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See
  12  *   the GNU General Public License for more details.
  13  *
  14  *   You should have received a copy of the GNU General Public License
  15  *   along with this program;  if not, write to the Free Software
  16  *   Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA 02111-1307 USA
  17  */
  18
  19 /*
  20  * Module: jfs_mount.c
  21  *
  22  * note: file system in transition to aggregate/fileset:
  23  *
  24  * file system mount is interpreted as the mount of aggregate,
  25  * if not already mounted, and mount of the single/only fileset in
  26  * the aggregate;
  27  *
  28  * a file system/aggregate is represented by an internal inode
  29  * (aka mount inode) initialized with aggregate superblock;
  30  * each vfs represents a fileset, and points to its "fileset inode
  31  * allocation map inode" (aka fileset inode):
  32  * (an aggregate itself is structured recursively as a filset:
  33  * an internal vfs is constructed and points to its "fileset inode
  34  * allocation map inode" (aka aggregate inode) where each inode
  35  * represents a fileset inode) so that inode number is mapped to
  36  * on-disk inode in uniform way at both aggregate and fileset level;
  37  *
  38  * each vnode/inode of a fileset is linked to its vfs (to facilitate
  39  * per fileset inode operations, e.g., unmount of a fileset, etc.);
  40  * each inode points to the mount inode (to facilitate access to
  41  * per aggregate information, e.g., block size, etc.) as well as
  42  * its file set inode.
  43  *
  44  *   aggregate
  45  *   ipmnt
  46  *   mntvfs -> fileset ipimap+ -> aggregate ipbmap -> aggregate ipaimap;
  47  *             fileset vfs     -> vp(1) <-> ... <-> vp(n) <->vproot;
  48  */
  49
  50 #include <linux/fs.h>
  51 #include <linux/buffer_head.h>
  52
  53 #include "jfs_incore.h"
  54 #include "jfs_filsys.h"
  55 #include "jfs_superblock.h"
  56 #include "jfs_dmap.h"
  57 #include "jfs_imap.h"
  58 #include "jfs_metapage.h"
  59 #include "jfs_debug.h"
  60
  61
  62 /*
  63  * forward references
  64  */
  65 static int chkSuper(struct super_block *);
  66 static int logMOUNT(struct super_block *sb);
  67
  68 /*
  69  * NAME:        jfs_mount(sb)
  70  *
  71  * FUNCTION:    vfs_mount()
  72  *
  73  * PARAMETER:   sb      - super block
  74  *
  75  * RETURN:      -EBUSY  - device already mounted or open for write
  76  *              -EBUSY  - cvrdvp already mounted;
  77  *              -EBUSY  - mount table full
  78  *              -ENOTDIR- cvrdvp not directory on a device mount
  79  *              -ENXIO  - device open failure
  80  */
  81 int jfs_mount(struct super_block *sb)
  82 {
  83         int rc = 0;             /* Return code          */
  84         struct jfs_sb_info *sbi = JFS_SBI(sb);
  85         struct inode *ipaimap = NULL;
  86         struct inode *ipaimap2 = NULL;
  87         struct inode *ipimap = NULL;
  88         struct inode *ipbmap = NULL;
  89
  90         /*
  91          * read/validate superblock
  92          * (initialize mount inode from the superblock)
  93          */
  94         if ((rc = chkSuper(sb))) {
  95                 goto errout20;
  96         }
  97
  98         ipaimap = diReadSpecial(sb, AGGREGATE_I, 0);
  99         if (ipaimap == NULL) {
 100                 jfs_err("jfs_mount: Faild to read AGGREGATE_I");
 101                 rc = -EIO;
 102                 goto errout20;
 103         }
 104         sbi->ipaimap = ipaimap;
 105
 106         jfs_info("jfs_mount: ipaimap:0x%p", ipaimap);
 107
 108         /*
 109          * initialize aggregate inode allocation map
 110          */
 111         if ((rc = diMount(ipaimap))) {
 112                 jfs_err("jfs_mount: diMount(ipaimap) failed w/rc = %d", rc);
 113                 goto errout21;
 114         }
 115
 116         /*
 117          * open aggregate block allocation map
 118          */
 119         ipbmap = diReadSpecial(sb, BMAP_I, 0);
 120         if (ipbmap == NULL) {
 121                 rc = -EIO;
 122                 goto errout22;
 123         }
 124
 125         jfs_info("jfs_mount: ipbmap:0x%p", ipbmap);
 126
 127         sbi->ipbmap = ipbmap;
 128
 129         /*
 130          * initialize aggregate block allocation map
 131          */
 132         if ((rc = dbMount(ipbmap))) {
 133                 jfs_err("jfs_mount: dbMount failed w/rc = %d", rc);
 134                 goto errout22;
 135         }
 136
 137         /*
 138          * open the secondary aggregate inode allocation map
 139          *
 140          * This is a duplicate of the aggregate inode allocation map.
 141          *
 142          * hand craft a vfs in the same fashion as we did to read ipaimap.
 143          * By adding INOSPEREXT (32) to the inode number, we are telling
 144          * diReadSpecial that we are reading from the secondary aggregate
 145          * inode table.  This also creates a unique entry in the inode hash
 146          * table.
 147          */
 148         if ((sbi->mntflag & JFS_BAD_SAIT) == 0) {
 149                 ipaimap2 = diReadSpecial(sb, AGGREGATE_I, 1);
 150                 if (ipaimap2 == 0) {
 151                         jfs_err("jfs_mount: Faild to read AGGREGATE_I");
 152                         rc = -EIO;
 153                         goto errout35;
 154                 }
 155                 sbi->ipaimap2 = ipaimap2;
 156
 157                 jfs_info("jfs_mount: ipaimap2:0x%p", ipaimap2);
 158
 159                 /*
 160                  * initialize secondary aggregate inode allocation map
 161                  */
 162                 if ((rc = diMount(ipaimap2))) {
 163                         jfs_err("jfs_mount: diMount(ipaimap2) failed, rc = %d",
 164                                 rc);
 165                         goto errout35;
 166                 }
 167         } else
 168                 /* Secondary aggregate inode table is not valid */
 169                 sbi->ipaimap2 = 0;
 170
 171         /*
 172          *      mount (the only/single) fileset
 173          */
 174         /*
 175          * open fileset inode allocation map (aka fileset inode)
 176          */
 177         ipimap = diReadSpecial(sb, FILESYSTEM_I, 0);
 178         if (ipimap == NULL) {
 179                 jfs_err("jfs_mount: Failed to read FILESYSTEM_I");
 180                 /* open fileset secondary inode allocation map */
 181                 rc = -EIO;
 182                 goto errout40;
 183         }
 184         jfs_info("jfs_mount: ipimap:0x%p", ipimap);
 185
 186         /* map further access of per fileset inodes by the fileset inode */
 187         sbi->ipimap = ipimap;
 188
 189         /* initialize fileset inode allocation map */
 190         if ((rc = diMount(ipimap))) {
 191                 jfs_err("jfs_mount: diMount failed w/rc = %d", rc);
 192                 goto errout41;
 193         }
 194
 195         goto out;
 196
 197         /*
 198          *      unwind on error
 199          */
 200 //errout42: /* close fileset inode allocation map */
 201         diUnmount(ipimap, 1);
 202
 203       errout41:         /* close fileset inode allocation map inode */
 204         diFreeSpecial(ipimap);
 205
 206       errout40:         /* fileset closed */
 207
 208         /* close secondary aggregate inode allocation map */
 209         if (ipaimap2) {
 210                 diUnmount(ipaimap2, 1);
 211                 diFreeSpecial(ipaimap2);
 212         }
 213
 214       errout35:
 215
 216         /* close aggregate block allocation map */
 217         dbUnmount(ipbmap, 1);
 218         diFreeSpecial(ipbmap);
 219
 220       errout22:         /* close aggregate inode allocation map */
 221
 222         diUnmount(ipaimap, 1);
 223
 224       errout21:         /* close aggregate inodes */
 225         diFreeSpecial(ipaimap);
 226       errout20:         /* aggregate closed */
 227
 228       out:
 229
 230         if (rc)
 231                 jfs_err("Mount JFS Failure: %d", rc);
 232
 233         return rc;
 234 }
 235
 236 /*
 237  * NAME:        jfs_mount_rw(sb, remount)
 238  *
 239  * FUNCTION:    Completes read-write mount, or remounts read-only volume
 240  *              as read-write
 241  */
 242 int jfs_mount_rw(struct super_block *sb, int remount)
 243 {
 244         struct jfs_sb_info *sbi = JFS_SBI(sb);
 245         struct jfs_log *log;
 246         int rc;
 247
 248         /*
 249          * If we are re-mounting a previously read-only volume, we want to
 250          * re-read the inode and block maps, since fsck.jfs may have updated
 251          * them.
 252          */
 253         if (remount) {
 254                 if (chkSuper(sb) || (sbi->state != FM_CLEAN))
 255                         return -EINVAL;
 256
 257                 truncate_inode_pages(sbi->ipimap->i_mapping, 0);
 258                 truncate_inode_pages(sbi->ipbmap->i_mapping, 0);
 259                 diUnmount(sbi->ipimap, 1);
 260                 if ((rc = diMount(sbi->ipimap))) {
 261                         jfs_err("jfs_mount_rw: diMount failed!");
 262                         return rc;
 263                 }
 264
 265                 dbUnmount(sbi->ipbmap, 1);
 266                 if ((rc = dbMount(sbi->ipbmap))) {
 267                         jfs_err("jfs_mount_rw: dbMount failed!");
 268                         return rc;
 269                 }
 270         }
 271
 272         /*
 273          * open/initialize log
 274          */
 275         if ((rc = lmLogOpen(sb, &log)))
 276                 return rc;
 277
 278         JFS_SBI(sb)->log = log;
 279
 280         /*
 281          * update file system superblock;
 282          */
 283         if ((rc = updateSuper(sb, FM_MOUNT))) {
 284                 jfs_err("jfs_mount: updateSuper failed w/rc = %d", rc);
 285                 lmLogClose(sb, log);
 286                 JFS_SBI(sb)->log = 0;
 287                 return rc;
 288         }
 289
 290         /*
 291          * write MOUNT log record of the file system
 292          */
 293         logMOUNT(sb);
 294
 295         return rc;
 296 }
 297
 298 /*
 299  *      chkSuper()
 300  *
 301  * validate the superblock of the file system to be mounted and
 302  * get the file system parameters.
 303  *
 304  * returns
 305  *      0 with fragsize set if check successful
 306  *      error code if not successful
 307  */
 308 static int chkSuper(struct super_block *sb)
 309 {
 310         int rc = 0;
 311         struct jfs_sb_info *sbi = JFS_SBI(sb);
 312         struct jfs_superblock *j_sb;
 313         struct buffer_head *bh;
 314         int AIM_bytesize, AIT_bytesize;
 315         int expected_AIM_bytesize, expected_AIT_bytesize;
 316         s64 AIM_byte_addr, AIT_byte_addr, fsckwsp_addr;
 317         s64 byte_addr_diff0, byte_addr_diff1;
 318         s32 bsize;
 319
 320         if ((rc = readSuper(sb, &bh)))
 321                 return rc;
 322         j_sb = (struct jfs_superblock *)bh->b_data;
 323
 324         /*
 325          * validate superblock
 326          */
 327         /* validate fs signature */
 328         if (strncmp(j_sb->s_magic, JFS_MAGIC, 4) ||
 329             j_sb->s_version > cpu_to_le32(JFS_VERSION)) {
 330                 rc = -EINVAL;
 331                 goto out;
 332         }
 333
 334         bsize = le32_to_cpu(j_sb->s_bsize);
 335 #ifdef _JFS_4K
 336         if (bsize != PSIZE) {
 337                 jfs_err("Currently only 4K block size supported!");
 338                 rc = -EINVAL;
 339                 goto out;
 340         }
 341 #endif                          /* _JFS_4K */
 342
 343         jfs_info("superblock: flag:0x%08x state:0x%08x size:0x%Lx",
 344                  le32_to_cpu(j_sb->s_flag), le32_to_cpu(j_sb->s_state),
 345                  (unsigned long long) le64_to_cpu(j_sb->s_size));
 346
 347         /* validate the descriptors for Secondary AIM and AIT */
 348         if ((j_sb->s_flag & cpu_to_le32(JFS_BAD_SAIT)) !=
 349             cpu_to_le32(JFS_BAD_SAIT)) {
 350                 expected_AIM_bytesize = 2 * PSIZE;
 351                 AIM_bytesize = lengthPXD(&(j_sb->s_aim2)) * bsize;
 352                 expected_AIT_bytesize = 4 * PSIZE;
 353                 AIT_bytesize = lengthPXD(&(j_sb->s_ait2)) * bsize;
 354                 AIM_byte_addr = addressPXD(&(j_sb->s_aim2)) * bsize;
 355                 AIT_byte_addr = addressPXD(&(j_sb->s_ait2)) * bsize;
 356                 byte_addr_diff0 = AIT_byte_addr - AIM_byte_addr;
 357                 fsckwsp_addr = addressPXD(&(j_sb->s_fsckpxd)) * bsize;
 358                 byte_addr_diff1 = fsckwsp_addr - AIT_byte_addr;
 359                 if ((AIM_bytesize != expected_AIM_bytesize) ||
 360                     (AIT_bytesize != expected_AIT_bytesize) ||
 361                     (byte_addr_diff0 != AIM_bytesize) ||
 362                     (byte_addr_diff1 <= AIT_bytesize))
 363                         j_sb->s_flag |= cpu_to_le32(JFS_BAD_SAIT);
 364         }
 365
 366         if ((j_sb->s_flag & cpu_to_le32(JFS_GROUPCOMMIT)) !=
 367             cpu_to_le32(JFS_GROUPCOMMIT))
 368                 j_sb->s_flag |= cpu_to_le32(JFS_GROUPCOMMIT);
 369
 370         /* validate fs state */
 371         if (j_sb->s_state != cpu_to_le32(FM_CLEAN) &&
 372             !(sb->s_flags & MS_RDONLY)) {
 373                 jfs_err("jfs_mount: Mount Failure: File System Dirty.");
 374                 rc = -EINVAL;
 375                 goto out;
 376         }
 377
 378         sbi->state = le32_to_cpu(j_sb->s_state);
 379         sbi->mntflag = le32_to_cpu(j_sb->s_flag);
 380
 381         /*
 382          * JFS always does I/O by 4K pages.  Don't tell the buffer cache
 383          * that we use anything else (leave s_blocksize alone).
 384          */
 385         sbi->bsize = bsize;
 386         sbi->l2bsize = le16_to_cpu(j_sb->s_l2bsize);
 387
 388         /*
 389          * For now, ignore s_pbsize, l2bfactor.  All I/O going through buffer
 390          * cache.
 391          */
 392         sbi->nbperpage = PSIZE >> sbi->l2bsize;
 393         sbi->l2nbperpage = L2PSIZE - sbi->l2bsize;
 394         sbi->l2niperblk = sbi->l2bsize - L2DISIZE;
 395         if (sbi->mntflag & JFS_INLINELOG)
 396                 sbi->logpxd = j_sb->s_logpxd;
 397         else {
 398                 sbi->logdev = new_decode_dev(le32_to_cpu(j_sb->s_logdev));
 399                 memcpy(sbi->uuid, j_sb->s_uuid, sizeof(sbi->uuid));
 400                 memcpy(sbi->loguuid, j_sb->s_loguuid, sizeof(sbi->uuid));
 401         }
 402         sbi->fsckpxd = j_sb->s_fsckpxd;
 403         sbi->ait2 = j_sb->s_ait2;
 404
 405       out:
 406         brelse(bh);
 407         return rc;
 408 }
 409
 410
 411 /*
 412  *      updateSuper()
 413  *
 414  * update synchronously superblock if it is mounted read-write.
 415  */
 416 int updateSuper(struct super_block *sb, uint state)
 417 {
 418         struct jfs_superblock *j_sb;
 419         struct jfs_sb_info *sbi = JFS_SBI(sb);
 420         struct buffer_head *bh;
 421         int rc;
 422
 423         if (sbi->flag & JFS_NOINTEGRITY) {
 424                 if (state == FM_DIRTY) {
 425                         sbi->p_state = state;
 426                         return 0;
 427                 } else if (state == FM_MOUNT) {
 428                         sbi->p_state = sbi->state;
 429                         state = FM_DIRTY;
 430                 } else if (state == FM_CLEAN) {
 431                         state = sbi->p_state;
 432                 } else
 433                         jfs_err("updateSuper: bad state");
 434         } else if (sbi->state == FM_DIRTY)
 435                 return 0;
 436
 437         if ((rc = readSuper(sb, &bh)))
 438                 return rc;
 439
 440         j_sb = (struct jfs_superblock *)bh->b_data;
 441
 442         j_sb->s_state = cpu_to_le32(state);
 443         sbi->state = state;
 444
 445         if (state == FM_MOUNT) {
 446                 /* record log's dev_t and mount serial number */
 447                 j_sb->s_logdev = cpu_to_le32(new_encode_dev(sbi->log->bdev->bd_dev));
 448                 j_sb->s_logserial = cpu_to_le32(sbi->log->serial);
 449         } else if (state == FM_CLEAN) {
 450                 /*
 451                  * If this volume is shared with OS/2, OS/2 will need to
 452                  * recalculate DASD usage, since we don't deal with it.
 453                  */
 454                 if (j_sb->s_flag & cpu_to_le32(JFS_DASD_ENABLED))
 455                         j_sb->s_flag |= cpu_to_le32(JFS_DASD_PRIME);
 456         }
 457
 458         mark_buffer_dirty(bh);
 459         sync_dirty_buffer(bh);
 460         brelse(bh);
 461
 462         return 0;
 463 }
 464
 465
 466 /*
 467  *      readSuper()
 468  *
 469  * read superblock by raw sector address
 470  */
 471 int readSuper(struct super_block *sb, struct buffer_head **bpp)
 472 {
 473         /* read in primary superblock */
 474         *bpp = sb_bread(sb, SUPER1_OFF >> sb->s_blocksize_bits);
 475         if (*bpp)
 476                 return 0;
 477
 478         /* read in secondary/replicated superblock */
 479         *bpp = sb_bread(sb, SUPER2_OFF >> sb->s_blocksize_bits);
 480         if (*bpp)
 481                 return 0;
 482
 483         return -EIO;
 484 }
 485
 486
 487 /*
 488  *      logMOUNT()
 489  *
 490  * function: write a MOUNT log record for file system.
 491  *
 492  * MOUNT record keeps logredo() from processing log records
 493  * for this file system past this point in log.
 494  * it is harmless if mount fails.
 495  *
 496  * note: MOUNT record is at aggregate level, not at fileset level,
 497  * since log records of previous mounts of a fileset
 498  * (e.g., AFTER record of extent allocation) have to be processed
 499  * to update block allocation map at aggregate level.
 500  */
 501 static int logMOUNT(struct super_block *sb)
 502 {
 503         struct jfs_log *log = JFS_SBI(sb)->log;
 504         struct lrd lrd;
 505
 506         lrd.logtid = 0;
 507         lrd.backchain = 0;
 508         lrd.type = cpu_to_le16(LOG_MOUNT);
 509         lrd.length = 0;
 510         lrd.aggregate = cpu_to_le32(new_encode_dev(sb->s_bdev->bd_dev));
 511         lmLog(log, NULL, &lrd, NULL);
 512
 513         return 0;
 514 }