fs/ext2/super.c

   1 /*
   2  *  linux/fs/ext2/super.c
   3  *
   4  * Copyright (C) 1992, 1993, 1994, 1995
   5  * Remy Card (card@masi.ibp.fr)
   6  * Laboratoire MASI - Institut Blaise Pascal
   7  * Universite Pierre et Marie Curie (Paris VI)
   8  *
   9  *  from
  10  *
  11  *  linux/fs/minix/inode.c
  12  *
  13  *  Copyright (C) 1991, 1992  Linus Torvalds
  14  *
  15  *  Big-endian to little-endian byte-swapping/bitmaps by
  16  *        David S. Miller (davem@caip.rutgers.edu), 1995
  17  */
  18
  19 #include <linux/config.h>
  20 #include <linux/module.h>
  21 #include <linux/string.h>
  22 #include <linux/slab.h>
  23 #include <linux/init.h>
  24 #include <linux/blkdev.h>
  25 #include <linux/parser.h>
  26 #include <linux/random.h>
  27 #include <linux/buffer_head.h>
  28 #include <linux/smp_lock.h>
  29 #include <linux/vfs.h>
  30 #include <asm/uaccess.h>
  31 #include "ext2.h"
  32 #include "xattr.h"
  33 #include "acl.h"
  34
  35 static void ext2_sync_super(struct super_block *sb,
  36                             struct ext2_super_block *es);
  37 static int ext2_remount (struct super_block * sb, int * flags, char * data);
  38 static int ext2_statfs (struct super_block * sb, struct kstatfs * buf);
  39
  40 static char error_buf[1024];
  41
  42 void ext2_error (struct super_block * sb, const char * function,
  43                  const char * fmt, ...)
  44 {
  45         va_list args;
  46         struct ext2_sb_info *sbi = EXT2_SB(sb);
  47         struct ext2_super_block *es = sbi->s_es;
  48
  49         if (!(sb->s_flags & MS_RDONLY)) {
  50                 sbi->s_mount_state |= EXT2_ERROR_FS;
  51                 es->s_state =
  52                         cpu_to_le16(le16_to_cpu(es->s_state) | EXT2_ERROR_FS);
  53                 ext2_sync_super(sb, es);
  54         }
  55         va_start (args, fmt);
  56         vsprintf (error_buf, fmt, args);
  57         va_end (args);
  58         if (test_opt (sb, ERRORS_PANIC))
  59                 panic ("EXT2-fs panic (device %s): %s: %s\n",
  60                        sb->s_id, function, error_buf);
  61         printk (KERN_CRIT "EXT2-fs error (device %s): %s: %s\n",
  62                 sb->s_id, function, error_buf);
  63         if (test_opt (sb, ERRORS_RO)) {
  64                 printk ("Remounting filesystem read-only\n");
  65                 sb->s_flags |= MS_RDONLY;
  66         }
  67 }
  68
  69 NORET_TYPE void ext2_panic (struct super_block * sb, const char * function,
  70                             const char * fmt, ...)
  71 {
  72         va_list args;
  73         struct ext2_sb_info *sbi = EXT2_SB(sb);
  74
  75         if (!(sb->s_flags & MS_RDONLY)) {
  76                 sbi->s_mount_state |= EXT2_ERROR_FS;
  77                 sbi->s_es->s_state =
  78                         cpu_to_le16(le16_to_cpu(sbi->s_es->s_state) | EXT2_ERROR_FS);
  79                 mark_buffer_dirty(sbi->s_sbh);
  80                 sb->s_dirt = 1;
  81         }
  82         va_start (args, fmt);
  83         vsprintf (error_buf, fmt, args);
  84         va_end (args);
  85         sb->s_flags |= MS_RDONLY;
  86         panic ("EXT2-fs panic (device %s): %s: %s\n",
  87                sb->s_id, function, error_buf);
  88 }
  89
  90 void ext2_warning (struct super_block * sb, const char * function,
  91                    const char * fmt, ...)
  92 {
  93         va_list args;
  94
  95         va_start (args, fmt);
  96         vsprintf (error_buf, fmt, args);
  97         va_end (args);
  98         printk (KERN_WARNING "EXT2-fs warning (device %s): %s: %s\n",
  99                 sb->s_id, function, error_buf);
 100 }
 101
 102 void ext2_update_dynamic_rev(struct super_block *sb)
 103 {
 104         struct ext2_super_block *es = EXT2_SB(sb)->s_es;
 105
 106         if (le32_to_cpu(es->s_rev_level) > EXT2_GOOD_OLD_REV)
 107                 return;
 108
 109         ext2_warning(sb, __FUNCTION__,
 110                      "updating to rev %d because of new feature flag, "
 111                      "running e2fsck is recommended",
 112                      EXT2_DYNAMIC_REV);
 113
 114         es->s_first_ino = cpu_to_le32(EXT2_GOOD_OLD_FIRST_INO);
 115         es->s_inode_size = cpu_to_le16(EXT2_GOOD_OLD_INODE_SIZE);
 116         es->s_rev_level = cpu_to_le32(EXT2_DYNAMIC_REV);
 117         /* leave es->s_feature_*compat flags alone */
 118         /* es->s_uuid will be set by e2fsck if empty */
 119
 120         /*
 121          * The rest of the superblock fields should be zero, and if not it
 122          * means they are likely already in use, so leave them alone.  We
 123          * can leave it up to e2fsck to clean up any inconsistencies there.
 124          */
 125 }
 126
 127 static void ext2_put_super (struct super_block * sb)
 128 {
 129         int db_count;
 130         int i;
 131         struct ext2_sb_info *sbi = EXT2_SB(sb);
 132
 133         ext2_xattr_put_super(sb);
 134         if (!(sb->s_flags & MS_RDONLY)) {
 135                 struct ext2_super_block *es = sbi->s_es;
 136
 137                 es->s_state = le16_to_cpu(sbi->s_mount_state);
 138                 ext2_sync_super(sb, es);
 139         }
 140         db_count = sbi->s_gdb_count;
 141         for (i = 0; i < db_count; i++)
 142                 if (sbi->s_group_desc[i])
 143                         brelse (sbi->s_group_desc[i]);
 144         kfree(sbi->s_group_desc);
 145         kfree(sbi->s_debts);
 146         brelse (sbi->s_sbh);
 147         sb->s_fs_info = NULL;
 148         kfree(sbi);
 149
 150         return;
 151 }
 152
 153 static kmem_cache_t * ext2_inode_cachep;
 154
 155 static struct inode *ext2_alloc_inode(struct super_block *sb)
 156 {
 157         struct ext2_inode_info *ei;
 158         ei = (struct ext2_inode_info *)kmem_cache_alloc(ext2_inode_cachep, SLAB_KERNEL);
 159         if (!ei)
 160                 return NULL;
 161 #ifdef CONFIG_EXT2_FS_POSIX_ACL
 162         ei->i_acl = EXT2_ACL_NOT_CACHED;
 163         ei->i_default_acl = EXT2_ACL_NOT_CACHED;
 164 #endif
 165         ei->vfs_inode.i_version = 1;
 166         return &ei->vfs_inode;
 167 }
 168
 169 static void ext2_destroy_inode(struct inode *inode)
 170 {
 171         kmem_cache_free(ext2_inode_cachep, EXT2_I(inode));
 172 }
 173
 174 static void init_once(void * foo, kmem_cache_t * cachep, unsigned long flags)
 175 {
 176         struct ext2_inode_info *ei = (struct ext2_inode_info *) foo;
 177
 178         if ((flags & (SLAB_CTOR_VERIFY|SLAB_CTOR_CONSTRUCTOR)) ==
 179             SLAB_CTOR_CONSTRUCTOR) {
 180                 rwlock_init(&ei->i_meta_lock);
 181 #ifdef CONFIG_EXT2_FS_XATTR
 182                 init_rwsem(&ei->xattr_sem);
 183 #endif
 184                 inode_init_once(&ei->vfs_inode);
 185         }
 186 }
 187
 188 static int init_inodecache(void)
 189 {
 190         ext2_inode_cachep = kmem_cache_create("ext2_inode_cache",
 191                                              sizeof(struct ext2_inode_info),
 192                                              0, SLAB_HWCACHE_ALIGN|SLAB_RECLAIM_ACCOUNT,
 193                                              init_once, NULL);
 194         if (ext2_inode_cachep == NULL)
 195                 return -ENOMEM;
 196         return 0;
 197 }
 198
 199 static void destroy_inodecache(void)
 200 {
 201         if (kmem_cache_destroy(ext2_inode_cachep))
 202                 printk(KERN_INFO "ext2_inode_cache: not all structures were freed\n");
 203 }
 204
 205 #ifdef CONFIG_EXT2_FS_POSIX_ACL
 206
 207 static void ext2_clear_inode(struct inode *inode)
 208 {
 209         struct ext2_inode_info *ei = EXT2_I(inode);
 210
 211         if (ei->i_acl && ei->i_acl != EXT2_ACL_NOT_CACHED) {
 212                 posix_acl_release(ei->i_acl);
 213                 ei->i_acl = EXT2_ACL_NOT_CACHED;
 214         }
 215         if (ei->i_default_acl && ei->i_default_acl != EXT2_ACL_NOT_CACHED) {
 216                 posix_acl_release(ei->i_default_acl);
 217                 ei->i_default_acl = EXT2_ACL_NOT_CACHED;
 218         }
 219 }
 220
 221 #else
 222 # define ext2_clear_inode NULL
 223 #endif
 224
 225 static struct super_operations ext2_sops = {
 226         .alloc_inode    = ext2_alloc_inode,
 227         .destroy_inode  = ext2_destroy_inode,
 228         .read_inode     = ext2_read_inode,
 229         .write_inode    = ext2_write_inode,
 230         .put_inode      = ext2_put_inode,
 231         .delete_inode   = ext2_delete_inode,
 232         .put_super      = ext2_put_super,
 233         .write_super    = ext2_write_super,
 234         .statfs         = ext2_statfs,
 235         .remount_fs     = ext2_remount,
 236         .clear_inode    = ext2_clear_inode,
 237 };
 238
 239 /* Yes, most of these are left as NULL!!
 240  * A NULL value implies the default, which works with ext2-like file
 241  * systems, but can be improved upon.
 242  * Currently only get_parent is required.
 243  */
 244 struct dentry *ext2_get_parent(struct dentry *child);
 245 static struct export_operations ext2_export_ops = {
 246         .get_parent = ext2_get_parent,
 247 };
 248
 249 static unsigned long get_sb_block(void **data)
 250 {
 251         unsigned long   sb_block;
 252         char            *options = (char *) *data;
 253
 254         if (!options || strncmp(options, "sb=", 3) != 0)
 255                 return 1;       /* Default location */
 256         options += 3;
 257         sb_block = simple_strtoul(options, &options, 0);
 258         if (*options && *options != ',') {
 259                 printk("EXT2-fs: Invalid sb specification: %s\n",
 260                        (char *) *data);
 261                 return 1;
 262         }
 263         if (*options == ',')
 264                 options++;
 265         *data = (void *) options;
 266         return sb_block;
 267 }
 268
 269 enum {
 270         Opt_bsd_df, Opt_minix_df, Opt_grpid, Opt_nogrpid,
 271         Opt_resgid, Opt_resuid, Opt_sb, Opt_err_cont, Opt_err_panic, Opt_err_ro,
 272         Opt_nouid32, Opt_check, Opt_nocheck, Opt_debug, Opt_oldalloc, Opt_orlov, Opt_nobh,
 273         Opt_user_xattr, Opt_nouser_xattr, Opt_acl, Opt_noacl,
 274         Opt_ignore, Opt_err,
 275 };
 276
 277 static match_table_t tokens = {
 278         {Opt_bsd_df, "bsddf"},
 279         {Opt_minix_df, "minixdf"},
 280         {Opt_grpid, "grpid"},
 281         {Opt_grpid, "bsdgroups"},
 282         {Opt_nogrpid, "nogrpid"},
 283         {Opt_nogrpid, "sysvgroups"},
 284         {Opt_resgid, "resgid=%u"},
 285         {Opt_resuid, "resuid=%u"},
 286         {Opt_sb, "sb=%u"},
 287         {Opt_err_cont, "errors=continue"},
 288         {Opt_err_panic, "errors=panic"},
 289         {Opt_err_ro, "errors=remount-ro"},
 290         {Opt_nouid32, "nouid32"},
 291         {Opt_nocheck, "check=none"},
 292         {Opt_nocheck, "nocheck"},
 293         {Opt_check, "check"},
 294         {Opt_debug, "debug"},
 295         {Opt_oldalloc, "oldalloc"},
 296         {Opt_orlov, "orlov"},
 297         {Opt_nobh, "nobh"},
 298         {Opt_user_xattr, "user_xattr"},
 299         {Opt_nouser_xattr, "nouser_xattr"},
 300         {Opt_acl, "acl"},
 301         {Opt_noacl, "noacl"},
 302         {Opt_ignore, "grpquota"},
 303         {Opt_ignore, "noquota"},
 304         {Opt_ignore, "quota"},
 305         {Opt_ignore, "usrquota"},
 306         {Opt_err, NULL}
 307 };
 308
 309 static int parse_options (char * options,
 310                           struct ext2_sb_info *sbi)
 311 {
 312         char * p;
 313         substring_t args[MAX_OPT_ARGS];
 314         unsigned long kind = EXT2_MOUNT_ERRORS_CONT;
 315         int option;
 316
 317         if (!options)
 318                 return 1;
 319
 320         while ((p = strsep (&options, ",")) != NULL) {
 321                 int token;
 322                 if (!*p)
 323                         continue;
 324
 325                 token = match_token(p, tokens, args);
 326                 switch (token) {
 327                 case Opt_bsd_df:
 328                         clear_opt (sbi->s_mount_opt, MINIX_DF);
 329                         break;
 330                 case Opt_minix_df:
 331                         set_opt (sbi->s_mount_opt, MINIX_DF);
 332                         break;
 333                 case Opt_grpid:
 334                         set_opt (sbi->s_mount_opt, GRPID);
 335                         break;
 336                 case Opt_nogrpid:
 337                         clear_opt (sbi->s_mount_opt, GRPID);
 338                         break;
 339                 case Opt_resuid:
 340                         if (match_int(&args[0], &option))
 341                                 return 0;
 342                         sbi->s_resuid = option;
 343                         break;
 344                 case Opt_resgid:
 345                         if (match_int(&args[0], &option))
 346                                 return 0;
 347                         sbi->s_resgid = option;
 348                         break;
 349                 case Opt_sb:
 350                         /* handled by get_sb_block() instead of here */
 351                         /* *sb_block = match_int(&args[0]); */
 352                         break;
 353                 case Opt_err_panic:
 354                         kind = EXT2_MOUNT_ERRORS_PANIC;
 355                         break;
 356                 case Opt_err_ro:
 357                         kind = EXT2_MOUNT_ERRORS_RO;
 358                         break;
 359                 case Opt_err_cont:
 360                         kind = EXT2_MOUNT_ERRORS_CONT;
 361                         break;
 362                 case Opt_nouid32:
 363                         set_opt (sbi->s_mount_opt, NO_UID32);
 364                         break;
 365                 case Opt_check:
 366 #ifdef CONFIG_EXT2_CHECK
 367                         set_opt (sbi->s_mount_opt, CHECK);
 368 #else
 369                         printk("EXT2 Check option not supported\n");
 370 #endif
 371                         break;
 372                 case Opt_nocheck:
 373                         clear_opt (sbi->s_mount_opt, CHECK);
 374                         break;
 375                 case Opt_debug:
 376                         set_opt (sbi->s_mount_opt, DEBUG);
 377                         break;
 378                 case Opt_oldalloc:
 379                         set_opt (sbi->s_mount_opt, OLDALLOC);
 380                         break;
 381                 case Opt_orlov:
 382                         clear_opt (sbi->s_mount_opt, OLDALLOC);
 383                         break;
 384                 case Opt_nobh:
 385                         set_opt (sbi->s_mount_opt, NOBH);
 386                         break;
 387 #ifdef CONFIG_EXT2_FS_XATTR
 388                 case Opt_user_xattr:
 389                         set_opt (sbi->s_mount_opt, XATTR_USER);
 390                         break;
 391                 case Opt_nouser_xattr:
 392                         clear_opt (sbi->s_mount_opt, XATTR_USER);
 393                         break;
 394 #else
 395                 case Opt_user_xattr:
 396                 case Opt_nouser_xattr:
 397                         printk("EXT2 (no)user_xattr options not supported\n");
 398                         break;
 399 #endif
 400 #ifdef CONFIG_EXT2_FS_POSIX_ACL
 401                 case Opt_acl:
 402                         set_opt(sbi->s_mount_opt, POSIX_ACL);
 403                         break;
 404                 case Opt_noacl:
 405                         clear_opt(sbi->s_mount_opt, POSIX_ACL);
 406                         break;
 407 #else
 408                 case Opt_acl:
 409                 case Opt_noacl:
 410                         printk("EXT2 (no)acl options not supported\n");
 411                         break;
 412 #endif
 413                 case Opt_ignore:
 414                         break;
 415                 default:
 416                         return 0;
 417                 }
 418         }
 419         sbi->s_mount_opt |= kind;
 420         return 1;
 421 }
 422
 423 static int ext2_setup_super (struct super_block * sb,
 424                               struct ext2_super_block * es,
 425                               int read_only)
 426 {
 427         int res = 0;
 428         struct ext2_sb_info *sbi = EXT2_SB(sb);
 429
 430         if (le32_to_cpu(es->s_rev_level) > EXT2_MAX_SUPP_REV) {
 431                 printk ("EXT2-fs warning: revision level too high, "
 432                         "forcing read-only mode\n");
 433                 res = MS_RDONLY;
 434         }
 435         if (read_only)
 436                 return res;
 437         if (!(sbi->s_mount_state & EXT2_VALID_FS))
 438                 printk ("EXT2-fs warning: mounting unchecked fs, "
 439                         "running e2fsck is recommended\n");
 440         else if ((sbi->s_mount_state & EXT2_ERROR_FS))
 441                 printk ("EXT2-fs warning: mounting fs with errors, "
 442                         "running e2fsck is recommended\n");
 443         else if ((__s16) le16_to_cpu(es->s_max_mnt_count) >= 0 &&
 444                  le16_to_cpu(es->s_mnt_count) >=
 445                  (unsigned short) (__s16) le16_to_cpu(es->s_max_mnt_count))
 446                 printk ("EXT2-fs warning: maximal mount count reached, "
 447                         "running e2fsck is recommended\n");
 448         else if (le32_to_cpu(es->s_checkinterval) &&
 449                 (le32_to_cpu(es->s_lastcheck) + le32_to_cpu(es->s_checkinterval) <= get_seconds()))
 450                 printk ("EXT2-fs warning: checktime reached, "
 451                         "running e2fsck is recommended\n");
 452         if (!(__s16) le16_to_cpu(es->s_max_mnt_count))
 453                 es->s_max_mnt_count = (__s16) cpu_to_le16(EXT2_DFL_MAX_MNT_COUNT);
 454         es->s_mnt_count=cpu_to_le16(le16_to_cpu(es->s_mnt_count) + 1);
 455         ext2_write_super(sb);
 456         if (test_opt (sb, DEBUG))
 457                 printk ("[EXT II FS %s, %s, bs=%lu, fs=%lu, gc=%lu, "
 458                         "bpg=%lu, ipg=%lu, mo=%04lx]\n",
 459                         EXT2FS_VERSION, EXT2FS_DATE, sb->s_blocksize,
 460                         sbi->s_frag_size,
 461                         sbi->s_groups_count,
 462                         EXT2_BLOCKS_PER_GROUP(sb),
 463                         EXT2_INODES_PER_GROUP(sb),
 464                         sbi->s_mount_opt);
 465 #ifdef CONFIG_EXT2_CHECK
 466         if (test_opt (sb, CHECK)) {
 467                 ext2_check_blocks_bitmap (sb);
 468                 ext2_check_inodes_bitmap (sb);
 469         }
 470 #endif
 471         return res;
 472 }
 473
 474 static int ext2_check_descriptors (struct super_block * sb)
 475 {
 476         int i;
 477         int desc_block = 0;
 478         struct ext2_sb_info *sbi = EXT2_SB(sb);
 479         unsigned long block = le32_to_cpu(sbi->s_es->s_first_data_block);
 480         struct ext2_group_desc * gdp = NULL;
 481
 482         ext2_debug ("Checking group descriptors");
 483
 484         for (i = 0; i < sbi->s_groups_count; i++)
 485         {
 486                 if ((i % EXT2_DESC_PER_BLOCK(sb)) == 0)
 487                         gdp = (struct ext2_group_desc *) sbi->s_group_desc[desc_block++]->b_data;
 488                 if (le32_to_cpu(gdp->bg_block_bitmap) < block ||
 489                     le32_to_cpu(gdp->bg_block_bitmap) >= block + EXT2_BLOCKS_PER_GROUP(sb))
 490                 {
 491                         ext2_error (sb, "ext2_check_descriptors",
 492                                     "Block bitmap for group %d"
 493                                     " not in group (block %lu)!",
 494                                     i, (unsigned long) le32_to_cpu(gdp->bg_block_bitmap));
 495                         return 0;
 496                 }
 497                 if (le32_to_cpu(gdp->bg_inode_bitmap) < block ||
 498                     le32_to_cpu(gdp->bg_inode_bitmap) >= block + EXT2_BLOCKS_PER_GROUP(sb))
 499                 {
 500                         ext2_error (sb, "ext2_check_descriptors",
 501                                     "Inode bitmap for group %d"
 502                                     " not in group (block %lu)!",
 503                                     i, (unsigned long) le32_to_cpu(gdp->bg_inode_bitmap));
 504                         return 0;
 505                 }
 506                 if (le32_to_cpu(gdp->bg_inode_table) < block ||
 507                     le32_to_cpu(gdp->bg_inode_table) + sbi->s_itb_per_group >=
 508                     block + EXT2_BLOCKS_PER_GROUP(sb))
 509                 {
 510                         ext2_error (sb, "ext2_check_descriptors",
 511                                     "Inode table for group %d"
 512                                     " not in group (block %lu)!",
 513                                     i, (unsigned long) le32_to_cpu(gdp->bg_inode_table));
 514                         return 0;
 515                 }
 516                 block += EXT2_BLOCKS_PER_GROUP(sb);
 517                 gdp++;
 518         }
 519         return 1;
 520 }
 521
 522 #define log2(n) ffz(~(n))
 523
 524 /*
 525  * Maximal file size.  There is a direct, and {,double-,triple-}indirect
 526  * block limit, and also a limit of (2^32 - 1) 512-byte sectors in i_blocks.
 527  * We need to be 1 filesystem block less than the 2^32 sector limit.
 528  */
 529 static loff_t ext2_max_size(int bits)
 530 {
 531         loff_t res = EXT2_NDIR_BLOCKS;
 532         res += 1LL << (bits-2);
 533         res += 1LL << (2*(bits-2));
 534         res += 1LL << (3*(bits-2));
 535         res <<= bits;
 536         if (res > (512LL << 32) - (1 << bits))
 537                 res = (512LL << 32) - (1 << bits);
 538         return res;
 539 }
 540
 541 static unsigned long descriptor_loc(struct super_block *sb,
 542                                     unsigned long logic_sb_block,
 543                                     int nr)
 544 {
 545         struct ext2_sb_info *sbi = EXT2_SB(sb);
 546         unsigned long bg, first_data_block, first_meta_bg;
 547         int has_super = 0;
 548
 549         first_data_block = le32_to_cpu(sbi->s_es->s_first_data_block);
 550         first_meta_bg = le32_to_cpu(sbi->s_es->s_first_meta_bg);
 551
 552         if (!EXT2_HAS_INCOMPAT_FEATURE(sb, EXT2_FEATURE_INCOMPAT_META_BG) ||
 553             nr < first_meta_bg)
 554                 return (logic_sb_block + nr + 1);
 555         bg = sbi->s_desc_per_block * nr;
 556         if (ext2_bg_has_super(sb, bg))
 557                 has_super = 1;
 558         return (first_data_block + has_super + (bg * sbi->s_blocks_per_group));
 559 }
 560
 561 static int ext2_fill_super(struct super_block *sb, void *data, int silent)
 562 {
 563         struct buffer_head * bh;
 564         struct ext2_sb_info * sbi;
 565         struct ext2_super_block * es;
 566         struct inode *root;
 567         unsigned long block;
 568         unsigned long sb_block = get_sb_block(&data);
 569         unsigned long logic_sb_block;
 570         unsigned long offset = 0;
 571         unsigned long def_mount_opts;
 572         int blocksize = BLOCK_SIZE;
 573         int db_count;
 574         int i, j;
 575
 576         sbi = kmalloc(sizeof(*sbi), GFP_KERNEL);
 577         if (!sbi)
 578                 return -ENOMEM;
 579         sb->s_fs_info = sbi;
 580         memset(sbi, 0, sizeof(*sbi));
 581
 582         /*
 583          * See what the current blocksize for the device is, and
 584          * use that as the blocksize.  Otherwise (or if the blocksize
 585          * is smaller than the default) use the default.
 586          * This is important for devices that have a hardware
 587          * sectorsize that is larger than the default.
 588          */
 589         blocksize = sb_min_blocksize(sb, BLOCK_SIZE);
 590         if (!blocksize) {
 591                 printk ("EXT2-fs: unable to set blocksize\n");
 592                 goto failed_sbi;
 593         }
 594
 595         /*
 596          * If the superblock doesn't start on a hardware sector boundary,
 597          * calculate the offset.
 598          */
 599         if (blocksize != BLOCK_SIZE) {
 600                 logic_sb_block = (sb_block*BLOCK_SIZE) / blocksize;
 601                 offset = (sb_block*BLOCK_SIZE) % blocksize;
 602         } else {
 603                 logic_sb_block = sb_block;
 604         }
 605
 606         if (!(bh = sb_bread(sb, logic_sb_block))) {
 607                 printk ("EXT2-fs: unable to read superblock\n");
 608                 goto failed_sbi;
 609         }
 610         /*
 611          * Note: s_es must be initialized as soon as possible because
 612          *       some ext2 macro-instructions depend on its value
 613          */
 614         es = (struct ext2_super_block *) (((char *)bh->b_data) + offset);
 615         sbi->s_es = es;
 616         sb->s_magic = le16_to_cpu(es->s_magic);
 617         sb->s_flags |= MS_ONE_SECOND;
 618         if (sb->s_magic != EXT2_SUPER_MAGIC) {
 619                 if (!silent)
 620                         printk ("VFS: Can't find ext2 filesystem on dev %s.\n",
 621                                 sb->s_id);
 622                 goto failed_mount;
 623         }
 624
 625         /* Set defaults before we parse the mount options */
 626         def_mount_opts = le32_to_cpu(es->s_default_mount_opts);
 627         if (def_mount_opts & EXT2_DEFM_DEBUG)
 628                 set_opt(sbi->s_mount_opt, DEBUG);
 629         if (def_mount_opts & EXT2_DEFM_BSDGROUPS)
 630                 set_opt(sbi->s_mount_opt, GRPID);
 631         if (def_mount_opts & EXT2_DEFM_UID16)
 632                 set_opt(sbi->s_mount_opt, NO_UID32);
 633         if (def_mount_opts & EXT2_DEFM_XATTR_USER)
 634                 set_opt(sbi->s_mount_opt, XATTR_USER);
 635         if (def_mount_opts & EXT2_DEFM_ACL)
 636                 set_opt(sbi->s_mount_opt, POSIX_ACL);
 637
 638         if (le16_to_cpu(sbi->s_es->s_errors) == EXT2_ERRORS_PANIC)
 639                 set_opt(sbi->s_mount_opt, ERRORS_PANIC);
 640         else if (le16_to_cpu(sbi->s_es->s_errors) == EXT2_ERRORS_RO)
 641                 set_opt(sbi->s_mount_opt, ERRORS_RO);
 642
 643         sbi->s_resuid = le16_to_cpu(es->s_def_resuid);
 644         sbi->s_resgid = le16_to_cpu(es->s_def_resgid);
 645
 646         if (!parse_options ((char *) data, sbi))
 647                 goto failed_mount;
 648
 649         sb->s_flags = (sb->s_flags & ~MS_POSIXACL) |
 650                 ((EXT2_SB(sb)->s_mount_opt & EXT2_MOUNT_POSIX_ACL) ?
 651                  MS_POSIXACL : 0);
 652
 653         if (le32_to_cpu(es->s_rev_level) == EXT2_GOOD_OLD_REV &&
 654             (EXT2_HAS_COMPAT_FEATURE(sb, ~0U) ||
 655              EXT2_HAS_RO_COMPAT_FEATURE(sb, ~0U) ||
 656              EXT2_HAS_INCOMPAT_FEATURE(sb, ~0U)))
 657                 printk("EXT2-fs warning: feature flags set on rev 0 fs, "
 658                        "running e2fsck is recommended\n");
 659         /*
 660          * Check feature flags regardless of the revision level, since we
 661          * previously didn't change the revision level when setting the flags,
 662          * so there is a chance incompat flags are set on a rev 0 filesystem.
 663          */
 664         if ((i = EXT2_HAS_INCOMPAT_FEATURE(sb, ~EXT2_FEATURE_INCOMPAT_SUPP))) {
 665                 printk("EXT2-fs: %s: couldn't mount because of "
 666                        "unsupported optional features (%x).\n",
 667                        sb->s_id, i);
 668                 goto failed_mount;
 669         }
 670         if (!(sb->s_flags & MS_RDONLY) &&
 671             (i = EXT2_HAS_RO_COMPAT_FEATURE(sb, ~EXT2_FEATURE_RO_COMPAT_SUPP))){
 672                 printk("EXT2-fs: %s: couldn't mount RDWR because of "
 673                        "unsupported optional features (%x).\n",
 674                        sb->s_id, i);
 675                 goto failed_mount;
 676         }
 677         blocksize = BLOCK_SIZE << le32_to_cpu(sbi->s_es->s_log_block_size);
 678         /* If the blocksize doesn't match, re-read the thing.. */
 679         if (sb->s_blocksize != blocksize) {
 680                 brelse(bh);
 681
 682                 if (!sb_set_blocksize(sb, blocksize)) {
 683                         printk(KERN_ERR "EXT2-fs: blocksize too small for device.\n");
 684                         goto failed_sbi;
 685                 }
 686
 687                 logic_sb_block = (sb_block*BLOCK_SIZE) / blocksize;
 688                 offset = (sb_block*BLOCK_SIZE) % blocksize;
 689                 bh = sb_bread(sb, logic_sb_block);
 690                 if(!bh) {
 691                         printk("EXT2-fs: Couldn't read superblock on "
 692                                "2nd try.\n");
 693                         goto failed_sbi;
 694                 }
 695                 es = (struct ext2_super_block *) (((char *)bh->b_data) + offset);
 696                 sbi->s_es = es;
 697                 if (es->s_magic != le16_to_cpu(EXT2_SUPER_MAGIC)) {
 698                         printk ("EXT2-fs: Magic mismatch, very weird !\n");
 699                         goto failed_mount;
 700                 }
 701         }
 702
 703         sb->s_maxbytes = ext2_max_size(sb->s_blocksize_bits);
 704
 705         if (le32_to_cpu(es->s_rev_level) == EXT2_GOOD_OLD_REV) {
 706                 sbi->s_inode_size = EXT2_GOOD_OLD_INODE_SIZE;
 707                 sbi->s_first_ino = EXT2_GOOD_OLD_FIRST_INO;
 708         } else {
 709                 sbi->s_inode_size = le16_to_cpu(es->s_inode_size);
 710                 sbi->s_first_ino = le32_to_cpu(es->s_first_ino);
 711                 if ((sbi->s_inode_size < EXT2_GOOD_OLD_INODE_SIZE) ||
 712                     (sbi->s_inode_size & (sbi->s_inode_size - 1)) ||
 713                     (sbi->s_inode_size > blocksize)) {
 714                         printk ("EXT2-fs: unsupported inode size: %d\n",
 715                                 sbi->s_inode_size);
 716                         goto failed_mount;
 717                 }
 718         }
 719         sbi->s_frag_size = EXT2_MIN_FRAG_SIZE <<
 720                                    le32_to_cpu(es->s_log_frag_size);
 721         if (sbi->s_frag_size)
 722                 sbi->s_frags_per_block = sb->s_blocksize /
 723                                                   sbi->s_frag_size;
 724         else
 725                 sb->s_magic = 0;
 726         sbi->s_blocks_per_group = le32_to_cpu(es->s_blocks_per_group);
 727         sbi->s_frags_per_group = le32_to_cpu(es->s_frags_per_group);
 728         sbi->s_inodes_per_group = le32_to_cpu(es->s_inodes_per_group);
 729         sbi->s_inodes_per_block = sb->s_blocksize /
 730                                            EXT2_INODE_SIZE(sb);
 731         sbi->s_itb_per_group = sbi->s_inodes_per_group /
 732                                         sbi->s_inodes_per_block;
 733         sbi->s_desc_per_block = sb->s_blocksize /
 734                                          sizeof (struct ext2_group_desc);
 735         sbi->s_sbh = bh;
 736         sbi->s_mount_state = le16_to_cpu(es->s_state);
 737         sbi->s_addr_per_block_bits =
 738                 log2 (EXT2_ADDR_PER_BLOCK(sb));
 739         sbi->s_desc_per_block_bits =
 740                 log2 (EXT2_DESC_PER_BLOCK(sb));
 741         if (sb->s_magic != EXT2_SUPER_MAGIC) {
 742                 if (!silent)
 743                         printk ("VFS: Can't find an ext2 filesystem on dev "
 744                                 "%s.\n",
 745                                 sb->s_id);
 746                 goto failed_mount;
 747         }
 748         if (sb->s_blocksize != bh->b_size) {
 749                 if (!silent)
 750                         printk ("VFS: Unsupported blocksize on dev "
 751                                 "%s.\n", sb->s_id);
 752                 goto failed_mount;
 753         }
 754
 755         if (sb->s_blocksize != sbi->s_frag_size) {
 756                 printk ("EXT2-fs: fragsize %lu != blocksize %lu (not supported yet)\n",
 757                         sbi->s_frag_size, sb->s_blocksize);
 758                 goto failed_mount;
 759         }
 760
 761         if (sbi->s_blocks_per_group > sb->s_blocksize * 8) {
 762                 printk ("EXT2-fs: #blocks per group too big: %lu\n",
 763                         sbi->s_blocks_per_group);
 764                 goto failed_mount;
 765         }
 766         if (sbi->s_frags_per_group > sb->s_blocksize * 8) {
 767                 printk ("EXT2-fs: #fragments per group too big: %lu\n",
 768                         sbi->s_frags_per_group);
 769                 goto failed_mount;
 770         }
 771         if (sbi->s_inodes_per_group > sb->s_blocksize * 8) {
 772                 printk ("EXT2-fs: #inodes per group too big: %lu\n",
 773                         sbi->s_inodes_per_group);
 774                 goto failed_mount;
 775         }
 776
 777         sbi->s_groups_count = (le32_to_cpu(es->s_blocks_count) -
 778                                         le32_to_cpu(es->s_first_data_block) +
 779                                        EXT2_BLOCKS_PER_GROUP(sb) - 1) /
 780                                        EXT2_BLOCKS_PER_GROUP(sb);
 781         db_count = (sbi->s_groups_count + EXT2_DESC_PER_BLOCK(sb) - 1) /
 782                    EXT2_DESC_PER_BLOCK(sb);
 783         sbi->s_group_desc = kmalloc (db_count * sizeof (struct buffer_head *), GFP_KERNEL);
 784         if (sbi->s_group_desc == NULL) {
 785                 printk ("EXT2-fs: not enough memory\n");
 786                 goto failed_mount;
 787         }
 788         percpu_counter_init(&sbi->s_freeblocks_counter);
 789         percpu_counter_init(&sbi->s_freeinodes_counter);
 790         percpu_counter_init(&sbi->s_dirs_counter);
 791         bgl_lock_init(&sbi->s_blockgroup_lock);
 792         sbi->s_debts = kmalloc(sbi->s_groups_count * sizeof(*sbi->s_debts),
 793                                GFP_KERNEL);
 794         if (!sbi->s_debts) {
 795                 printk ("EXT2-fs: not enough memory\n");
 796                 goto failed_mount_group_desc;
 797         }
 798         memset(sbi->s_debts, 0, sbi->s_groups_count * sizeof(*sbi->s_debts));
 799         for (i = 0; i < db_count; i++) {
 800                 block = descriptor_loc(sb, logic_sb_block, i);
 801                 sbi->s_group_desc[i] = sb_bread(sb, block);
 802                 if (!sbi->s_group_desc[i]) {
 803                         for (j = 0; j < i; j++)
 804                                 brelse (sbi->s_group_desc[j]);
 805                         printk ("EXT2-fs: unable to read group descriptors\n");
 806                         goto failed_mount_group_desc;
 807                 }
 808         }
 809         if (!ext2_check_descriptors (sb)) {
 810                 printk ("EXT2-fs: group descriptors corrupted!\n");
 811                 db_count = i;
 812                 goto failed_mount2;
 813         }
 814         sbi->s_gdb_count = db_count;
 815         get_random_bytes(&sbi->s_next_generation, sizeof(u32));
 816         spin_lock_init(&sbi->s_next_gen_lock);
 817         /*
 818          * set up enough so that it can read an inode
 819          */
 820         sb->s_op = &ext2_sops;
 821         sb->s_export_op = &ext2_export_ops;
 822         root = iget(sb, EXT2_ROOT_INO);
 823         sb->s_root = d_alloc_root(root);
 824         if (!sb->s_root) {
 825                 iput(root);
 826                 printk(KERN_ERR "EXT2-fs: get root inode failed\n");
 827                 goto failed_mount2;
 828         }
 829         if (!S_ISDIR(root->i_mode) || !root->i_blocks || !root->i_size) {
 830                 dput(sb->s_root);
 831                 sb->s_root = NULL;
 832                 printk(KERN_ERR "EXT2-fs: corrupt root inode, run e2fsck\n");
 833                 goto failed_mount2;
 834         }
 835         if (EXT2_HAS_COMPAT_FEATURE(sb, EXT3_FEATURE_COMPAT_HAS_JOURNAL))
 836                 ext2_warning(sb, __FUNCTION__,
 837                         "mounting ext3 filesystem as ext2\n");
 838         ext2_setup_super (sb, es, sb->s_flags & MS_RDONLY);
 839         percpu_counter_mod(&sbi->s_freeblocks_counter,
 840                                 ext2_count_free_blocks(sb));
 841         percpu_counter_mod(&sbi->s_freeinodes_counter,
 842                                 ext2_count_free_inodes(sb));
 843         percpu_counter_mod(&sbi->s_dirs_counter,
 844                                 ext2_count_dirs(sb));
 845         return 0;
 846 failed_mount2:
 847         for (i = 0; i < db_count; i++)
 848                 brelse(sbi->s_group_desc[i]);
 849 failed_mount_group_desc:
 850         kfree(sbi->s_group_desc);
 851         if (sbi->s_debts)
 852                 kfree(sbi->s_debts);
 853 failed_mount:
 854         brelse(bh);
 855 failed_sbi:
 856         sb->s_fs_info = NULL;
 857         kfree(sbi);
 858         return -EINVAL;
 859 }
 860
 861 static void ext2_commit_super (struct super_block * sb,
 862                                struct ext2_super_block * es)
 863 {
 864         es->s_wtime = cpu_to_le32(get_seconds());
 865         mark_buffer_dirty(EXT2_SB(sb)->s_sbh);
 866         sb->s_dirt = 0;
 867 }
 868
 869 static void ext2_sync_super(struct super_block *sb, struct ext2_super_block *es)
 870 {
 871         es->s_free_blocks_count = cpu_to_le32(ext2_count_free_blocks(sb));
 872         es->s_free_inodes_count = cpu_to_le32(ext2_count_free_inodes(sb));
 873         es->s_wtime = cpu_to_le32(get_seconds());
 874         mark_buffer_dirty(EXT2_SB(sb)->s_sbh);
 875         sync_dirty_buffer(EXT2_SB(sb)->s_sbh);
 876         sb->s_dirt = 0;
 877 }
 878
 879 /*
 880  * In the second extended file system, it is not necessary to
 881  * write the super block since we use a mapping of the
 882  * disk super block in a buffer.
 883  *
 884  * However, this function is still used to set the fs valid
 885  * flags to 0.  We need to set this flag to 0 since the fs
 886  * may have been checked while mounted and e2fsck may have
 887  * set s_state to EXT2_VALID_FS after some corrections.
 888  */
 889
 890 void ext2_write_super (struct super_block * sb)
 891 {
 892         struct ext2_super_block * es;
 893         lock_kernel();
 894         if (!(sb->s_flags & MS_RDONLY)) {
 895                 es = EXT2_SB(sb)->s_es;
 896
 897                 if (le16_to_cpu(es->s_state) & EXT2_VALID_FS) {
 898                         ext2_debug ("setting valid to 0\n");
 899                         es->s_state = cpu_to_le16(le16_to_cpu(es->s_state) &
 900                                                   ~EXT2_VALID_FS);
 901                         es->s_free_blocks_count = cpu_to_le32(ext2_count_free_blocks(sb));
 902                         es->s_free_inodes_count = cpu_to_le32(ext2_count_free_inodes(sb));
 903                         es->s_mtime = cpu_to_le32(get_seconds());
 904                         ext2_sync_super(sb, es);
 905                 } else
 906                         ext2_commit_super (sb, es);
 907         }
 908         sb->s_dirt = 0;
 909         unlock_kernel();
 910 }
 911
 912 static int ext2_remount (struct super_block * sb, int * flags, char * data)
 913 {
 914         struct ext2_sb_info * sbi = EXT2_SB(sb);
 915         struct ext2_super_block * es;
 916
 917         /*
 918          * Allow the "check" option to be passed as a remount option.
 919          */
 920         if (!parse_options (data, sbi))
 921                 return -EINVAL;
 922
 923         sb->s_flags = (sb->s_flags & ~MS_POSIXACL) |
 924                 ((sbi->s_mount_opt & EXT2_MOUNT_POSIX_ACL) ? MS_POSIXACL : 0);
 925
 926         es = sbi->s_es;
 927         if ((*flags & MS_RDONLY) == (sb->s_flags & MS_RDONLY))
 928                 return 0;
 929         if (*flags & MS_RDONLY) {
 930                 if (le16_to_cpu(es->s_state) & EXT2_VALID_FS ||
 931                     !(sbi->s_mount_state & EXT2_VALID_FS))
 932                         return 0;
 933                 /*
 934                  * OK, we are remounting a valid rw partition rdonly, so set
 935                  * the rdonly flag and then mark the partition as valid again.
 936                  */
 937                 es->s_state = cpu_to_le16(sbi->s_mount_state);
 938                 es->s_mtime = cpu_to_le32(get_seconds());
 939         } else {
 940                 int ret;
 941                 if ((ret = EXT2_HAS_RO_COMPAT_FEATURE(sb,
 942                                                ~EXT2_FEATURE_RO_COMPAT_SUPP))) {
 943                         printk("EXT2-fs: %s: couldn't remount RDWR because of "
 944                                "unsupported optional features (%x).\n",
 945                                sb->s_id, ret);
 946                         return -EROFS;
 947                 }
 948                 /*
 949                  * Mounting a RDONLY partition read-write, so reread and
 950                  * store the current valid flag.  (It may have been changed
 951                  * by e2fsck since we originally mounted the partition.)
 952                  */
 953                 sbi->s_mount_state = le16_to_cpu(es->s_state);
 954                 if (!ext2_setup_super (sb, es, 0))
 955                         sb->s_flags &= ~MS_RDONLY;
 956         }
 957         ext2_sync_super(sb, es);
 958         return 0;
 959 }
 960
 961 static int ext2_statfs (struct super_block * sb, struct kstatfs * buf)
 962 {
 963         struct ext2_sb_info *sbi = EXT2_SB(sb);
 964         unsigned long overhead;
 965         int i;
 966
 967         if (test_opt (sb, MINIX_DF))
 968                 overhead = 0;
 969         else {
 970                 /*
 971                  * Compute the overhead (FS structures)
 972                  */
 973
 974                 /*
 975                  * All of the blocks before first_data_block are
 976                  * overhead
 977                  */
 978                 overhead = le32_to_cpu(sbi->s_es->s_first_data_block);
 979
 980                 /*
 981                  * Add the overhead attributed to the superblock and
 982                  * block group descriptors.  If the sparse superblocks
 983                  * feature is turned on, then not all groups have this.
 984                  */
 985                 for (i = 0; i < sbi->s_groups_count; i++)
 986                         overhead += ext2_bg_has_super(sb, i) +
 987                                 ext2_bg_num_gdb(sb, i);
 988
 989                 /*
 990                  * Every block group has an inode bitmap, a block
 991                  * bitmap, and an inode table.
 992                  */
 993                 overhead += (sbi->s_groups_count *
 994                              (2 + sbi->s_itb_per_group));
 995         }
 996
 997         buf->f_type = EXT2_SUPER_MAGIC;
 998         buf->f_bsize = sb->s_blocksize;
 999         buf->f_blocks = le32_to_cpu(sbi->s_es->s_blocks_count) - overhead;
1000         buf->f_bfree = ext2_count_free_blocks(sb);
1001         buf->f_bavail = buf->f_bfree - le32_to_cpu(sbi->s_es->s_r_blocks_count);
1002         if (buf->f_bfree < le32_to_cpu(sbi->s_es->s_r_blocks_count))
1003                 buf->f_bavail = 0;
1004         buf->f_files = le32_to_cpu(sbi->s_es->s_inodes_count);
1005         buf->f_ffree = ext2_count_free_inodes (sb);
1006         buf->f_namelen = EXT2_NAME_LEN;
1007         return 0;
1008 }
1009
1010 static struct super_block *ext2_get_sb(struct file_system_type *fs_type,
1011         int flags, const char *dev_name, void *data)
1012 {
1013         return get_sb_bdev(fs_type, flags, dev_name, data, ext2_fill_super);
1014 }
1015
1016 static struct file_system_type ext2_fs_type = {
1017         .owner          = THIS_MODULE,
1018         .name           = "ext2",
1019         .get_sb         = ext2_get_sb,
1020         .kill_sb        = kill_block_super,
1021         .fs_flags       = FS_REQUIRES_DEV,
1022 };
1023
1024 static int __init init_ext2_fs(void)
1025 {
1026         int err = init_ext2_xattr();
1027         if (err)
1028                 return err;
1029         err = init_inodecache();
1030         if (err)
1031                 goto out1;
1032         err = register_filesystem(&ext2_fs_type);
1033         if (err)
1034                 goto out;
1035         return 0;
1036 out:
1037         destroy_inodecache();
1038 out1:
1039         exit_ext2_xattr();
1040         return err;
1041 }
1042
1043 static void __exit exit_ext2_fs(void)
1044 {
1045         unregister_filesystem(&ext2_fs_type);
1046         destroy_inodecache();
1047         exit_ext2_xattr();
1048 }
1049
1050 module_init(init_ext2_fs)
1051 module_exit(exit_ext2_fs)