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/module.h>
  20 #include <linux/string.h>
  21 #include <linux/fs.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/exportfs.h>
  29 #include <linux/smp_lock.h>
  30 #include <linux/vfs.h>
  31 #include <linux/seq_file.h>
  32 #include <linux/mount.h>
  33 #include <linux/log2.h>
  34 #include <linux/quotaops.h>
  35 #include <asm/uaccess.h>
  36 #include "ext2.h"
  37 #include "xattr.h"
  38 #include "acl.h"
  39 #include "xip.h"
  40
  41 static void ext2_sync_super(struct super_block *sb,
  42                             struct ext2_super_block *es);
  43 static int ext2_remount (struct super_block * sb, int * flags, char * data);
  44 static int ext2_statfs (struct dentry * dentry, struct kstatfs * buf);
  45 static int ext2_sync_fs(struct super_block *sb, int wait);
  46
  47 void ext2_error (struct super_block * sb, const char * function,
  48                  const char * fmt, ...)
  49 {
  50         va_list args;
  51         struct ext2_sb_info *sbi = EXT2_SB(sb);
  52         struct ext2_super_block *es = sbi->s_es;
  53
  54         if (!(sb->s_flags & MS_RDONLY)) {
  55                 sbi->s_mount_state |= EXT2_ERROR_FS;
  56                 es->s_state |= cpu_to_le16(EXT2_ERROR_FS);
  57                 ext2_sync_super(sb, es);
  58         }
  59
  60         va_start(args, fmt);
  61         printk(KERN_CRIT "EXT2-fs (%s): error: %s: ", sb->s_id, function);
  62         vprintk(fmt, args);
  63         printk("\n");
  64         va_end(args);
  65
  66         if (test_opt(sb, ERRORS_PANIC))
  67                 panic("EXT2-fs: panic from previous error\n");
  68         if (test_opt(sb, ERRORS_RO)) {
  69                 ext2_msg(sb, KERN_CRIT,
  70                              "error: remounting filesystem read-only");
  71                 sb->s_flags |= MS_RDONLY;
  72         }
  73 }
  74
  75 void ext2_msg(struct super_block *sb, const char *prefix,
  76                 const char *fmt, ...)
  77 {
  78         va_list args;
  79
  80         va_start(args, fmt);
  81         printk("%sEXT2-fs (%s): ", prefix, sb->s_id);
  82         vprintk(fmt, args);
  83         printk("\n");
  84         va_end(args);
  85 }
  86
  87 void ext2_update_dynamic_rev(struct super_block *sb)
  88 {
  89         struct ext2_super_block *es = EXT2_SB(sb)->s_es;
  90
  91         if (le32_to_cpu(es->s_rev_level) > EXT2_GOOD_OLD_REV)
  92                 return;
  93
  94         ext2_msg(sb, KERN_WARNING,
  95                      "warning: updating to rev %d because of "
  96                      "new feature flag, running e2fsck is recommended",
  97                      EXT2_DYNAMIC_REV);
  98
  99         es->s_first_ino = cpu_to_le32(EXT2_GOOD_OLD_FIRST_INO);
 100         es->s_inode_size = cpu_to_le16(EXT2_GOOD_OLD_INODE_SIZE);
 101         es->s_rev_level = cpu_to_le32(EXT2_DYNAMIC_REV);
 102         /* leave es->s_feature_*compat flags alone */
 103         /* es->s_uuid will be set by e2fsck if empty */
 104
 105         /*
 106          * The rest of the superblock fields should be zero, and if not it
 107          * means they are likely already in use, so leave them alone.  We
 108          * can leave it up to e2fsck to clean up any inconsistencies there.
 109          */
 110 }
 111
 112 static void ext2_put_super (struct super_block * sb)
 113 {
 114         int db_count;
 115         int i;
 116         struct ext2_sb_info *sbi = EXT2_SB(sb);
 117
 118         lock_kernel();
 119
 120         if (sb->s_dirt)
 121                 ext2_write_super(sb);
 122
 123         ext2_xattr_put_super(sb);
 124         if (!(sb->s_flags & MS_RDONLY)) {
 125                 struct ext2_super_block *es = sbi->s_es;
 126
 127                 es->s_state = cpu_to_le16(sbi->s_mount_state);
 128                 ext2_sync_super(sb, es);
 129         }
 130         db_count = sbi->s_gdb_count;
 131         for (i = 0; i < db_count; i++)
 132                 if (sbi->s_group_desc[i])
 133                         brelse (sbi->s_group_desc[i]);
 134         kfree(sbi->s_group_desc);
 135         kfree(sbi->s_debts);
 136         percpu_counter_destroy(&sbi->s_freeblocks_counter);
 137         percpu_counter_destroy(&sbi->s_freeinodes_counter);
 138         percpu_counter_destroy(&sbi->s_dirs_counter);
 139         brelse (sbi->s_sbh);
 140         sb->s_fs_info = NULL;
 141         kfree(sbi->s_blockgroup_lock);
 142         kfree(sbi);
 143
 144         unlock_kernel();
 145 }
 146
 147 static struct kmem_cache * ext2_inode_cachep;
 148
 149 static struct inode *ext2_alloc_inode(struct super_block *sb)
 150 {
 151         struct ext2_inode_info *ei;
 152         ei = (struct ext2_inode_info *)kmem_cache_alloc(ext2_inode_cachep, GFP_KERNEL);
 153         if (!ei)
 154                 return NULL;
 155         ei->i_block_alloc_info = NULL;
 156         ei->vfs_inode.i_version = 1;
 157         return &ei->vfs_inode;
 158 }
 159
 160 static void ext2_destroy_inode(struct inode *inode)
 161 {
 162         kmem_cache_free(ext2_inode_cachep, EXT2_I(inode));
 163 }
 164
 165 static void init_once(void *foo)
 166 {
 167         struct ext2_inode_info *ei = (struct ext2_inode_info *) foo;
 168
 169         rwlock_init(&ei->i_meta_lock);
 170 #ifdef CONFIG_EXT2_FS_XATTR
 171         init_rwsem(&ei->xattr_sem);
 172 #endif
 173         mutex_init(&ei->truncate_mutex);
 174         inode_init_once(&ei->vfs_inode);
 175 }
 176
 177 static int init_inodecache(void)
 178 {
 179         ext2_inode_cachep = kmem_cache_create("ext2_inode_cache",
 180                                              sizeof(struct ext2_inode_info),
 181                                              0, (SLAB_RECLAIM_ACCOUNT|
 182                                                 SLAB_MEM_SPREAD),
 183                                              init_once);
 184         if (ext2_inode_cachep == NULL)
 185                 return -ENOMEM;
 186         return 0;
 187 }
 188
 189 static void destroy_inodecache(void)
 190 {
 191         kmem_cache_destroy(ext2_inode_cachep);
 192 }
 193
 194 static void ext2_clear_inode(struct inode *inode)
 195 {
 196         struct ext2_block_alloc_info *rsv = EXT2_I(inode)->i_block_alloc_info;
 197
 198         dquot_drop(inode);
 199         ext2_discard_reservation(inode);
 200         EXT2_I(inode)->i_block_alloc_info = NULL;
 201         if (unlikely(rsv))
 202                 kfree(rsv);
 203 }
 204
 205 static int ext2_show_options(struct seq_file *seq, struct vfsmount *vfs)
 206 {
 207         struct super_block *sb = vfs->mnt_sb;
 208         struct ext2_sb_info *sbi = EXT2_SB(sb);
 209         struct ext2_super_block *es = sbi->s_es;
 210         unsigned long def_mount_opts;
 211
 212         def_mount_opts = le32_to_cpu(es->s_default_mount_opts);
 213
 214         if (sbi->s_sb_block != 1)
 215                 seq_printf(seq, ",sb=%lu", sbi->s_sb_block);
 216         if (test_opt(sb, MINIX_DF))
 217                 seq_puts(seq, ",minixdf");
 218         if (test_opt(sb, GRPID))
 219                 seq_puts(seq, ",grpid");
 220         if (!test_opt(sb, GRPID) && (def_mount_opts & EXT2_DEFM_BSDGROUPS))
 221                 seq_puts(seq, ",nogrpid");
 222         if (sbi->s_resuid != EXT2_DEF_RESUID ||
 223             le16_to_cpu(es->s_def_resuid) != EXT2_DEF_RESUID) {
 224                 seq_printf(seq, ",resuid=%u", sbi->s_resuid);
 225         }
 226         if (sbi->s_resgid != EXT2_DEF_RESGID ||
 227             le16_to_cpu(es->s_def_resgid) != EXT2_DEF_RESGID) {
 228                 seq_printf(seq, ",resgid=%u", sbi->s_resgid);
 229         }
 230         if (test_opt(sb, ERRORS_RO)) {
 231                 int def_errors = le16_to_cpu(es->s_errors);
 232
 233                 if (def_errors == EXT2_ERRORS_PANIC ||
 234                     def_errors == EXT2_ERRORS_CONTINUE) {
 235                         seq_puts(seq, ",errors=remount-ro");
 236                 }
 237         }
 238         if (test_opt(sb, ERRORS_CONT))
 239                 seq_puts(seq, ",errors=continue");
 240         if (test_opt(sb, ERRORS_PANIC))
 241                 seq_puts(seq, ",errors=panic");
 242         if (test_opt(sb, NO_UID32))
 243                 seq_puts(seq, ",nouid32");
 244         if (test_opt(sb, DEBUG))
 245                 seq_puts(seq, ",debug");
 246         if (test_opt(sb, OLDALLOC))
 247                 seq_puts(seq, ",oldalloc");
 248
 249 #ifdef CONFIG_EXT2_FS_XATTR
 250         if (test_opt(sb, XATTR_USER))
 251                 seq_puts(seq, ",user_xattr");
 252         if (!test_opt(sb, XATTR_USER) &&
 253             (def_mount_opts & EXT2_DEFM_XATTR_USER)) {
 254                 seq_puts(seq, ",nouser_xattr");
 255         }
 256 #endif
 257
 258 #ifdef CONFIG_EXT2_FS_POSIX_ACL
 259         if (test_opt(sb, POSIX_ACL))
 260                 seq_puts(seq, ",acl");
 261         if (!test_opt(sb, POSIX_ACL) && (def_mount_opts & EXT2_DEFM_ACL))
 262                 seq_puts(seq, ",noacl");
 263 #endif
 264
 265         if (test_opt(sb, NOBH))
 266                 seq_puts(seq, ",nobh");
 267
 268 #if defined(CONFIG_QUOTA)
 269         if (sbi->s_mount_opt & EXT2_MOUNT_USRQUOTA)
 270                 seq_puts(seq, ",usrquota");
 271
 272         if (sbi->s_mount_opt & EXT2_MOUNT_GRPQUOTA)
 273                 seq_puts(seq, ",grpquota");
 274 #endif
 275
 276 #if defined(CONFIG_EXT2_FS_XIP)
 277         if (sbi->s_mount_opt & EXT2_MOUNT_XIP)
 278                 seq_puts(seq, ",xip");
 279 #endif
 280
 281         if (!test_opt(sb, RESERVATION))
 282                 seq_puts(seq, ",noreservation");
 283
 284         return 0;
 285 }
 286
 287 #ifdef CONFIG_QUOTA
 288 static ssize_t ext2_quota_read(struct super_block *sb, int type, char *data, size_t len, loff_t off);
 289 static ssize_t ext2_quota_write(struct super_block *sb, int type, const char *data, size_t len, loff_t off);
 290 #endif
 291
 292 static const struct super_operations ext2_sops = {
 293         .alloc_inode    = ext2_alloc_inode,
 294         .destroy_inode  = ext2_destroy_inode,
 295         .write_inode    = ext2_write_inode,
 296         .delete_inode   = ext2_delete_inode,
 297         .put_super      = ext2_put_super,
 298         .write_super    = ext2_write_super,
 299         .sync_fs        = ext2_sync_fs,
 300         .statfs         = ext2_statfs,
 301         .remount_fs     = ext2_remount,
 302         .clear_inode    = ext2_clear_inode,
 303         .show_options   = ext2_show_options,
 304 #ifdef CONFIG_QUOTA
 305         .quota_read     = ext2_quota_read,
 306         .quota_write    = ext2_quota_write,
 307 #endif
 308 };
 309
 310 static struct inode *ext2_nfs_get_inode(struct super_block *sb,
 311                 u64 ino, u32 generation)
 312 {
 313         struct inode *inode;
 314
 315         if (ino < EXT2_FIRST_INO(sb) && ino != EXT2_ROOT_INO)
 316                 return ERR_PTR(-ESTALE);
 317         if (ino > le32_to_cpu(EXT2_SB(sb)->s_es->s_inodes_count))
 318                 return ERR_PTR(-ESTALE);
 319
 320         /* iget isn't really right if the inode is currently unallocated!!
 321          * ext2_read_inode currently does appropriate checks, but
 322          * it might be "neater" to call ext2_get_inode first and check
 323          * if the inode is valid.....
 324          */
 325         inode = ext2_iget(sb, ino);
 326         if (IS_ERR(inode))
 327                 return ERR_CAST(inode);
 328         if (generation && inode->i_generation != generation) {
 329                 /* we didn't find the right inode.. */
 330                 iput(inode);
 331                 return ERR_PTR(-ESTALE);
 332         }
 333         return inode;
 334 }
 335
 336 static struct dentry *ext2_fh_to_dentry(struct super_block *sb, struct fid *fid,
 337                 int fh_len, int fh_type)
 338 {
 339         return generic_fh_to_dentry(sb, fid, fh_len, fh_type,
 340                                     ext2_nfs_get_inode);
 341 }
 342
 343 static struct dentry *ext2_fh_to_parent(struct super_block *sb, struct fid *fid,
 344                 int fh_len, int fh_type)
 345 {
 346         return generic_fh_to_parent(sb, fid, fh_len, fh_type,
 347                                     ext2_nfs_get_inode);
 348 }
 349
 350 /* Yes, most of these are left as NULL!!
 351  * A NULL value implies the default, which works with ext2-like file
 352  * systems, but can be improved upon.
 353  * Currently only get_parent is required.
 354  */
 355 static const struct export_operations ext2_export_ops = {
 356         .fh_to_dentry = ext2_fh_to_dentry,
 357         .fh_to_parent = ext2_fh_to_parent,
 358         .get_parent = ext2_get_parent,
 359 };
 360
 361 static unsigned long get_sb_block(void **data)
 362 {
 363         unsigned long   sb_block;
 364         char            *options = (char *) *data;
 365
 366         if (!options || strncmp(options, "sb=", 3) != 0)
 367                 return 1;       /* Default location */
 368         options += 3;
 369         sb_block = simple_strtoul(options, &options, 0);
 370         if (*options && *options != ',') {
 371                 printk("EXT2-fs: Invalid sb specification: %s\n",
 372                        (char *) *data);
 373                 return 1;
 374         }
 375         if (*options == ',')
 376                 options++;
 377         *data = (void *) options;
 378         return sb_block;
 379 }
 380
 381 enum {
 382         Opt_bsd_df, Opt_minix_df, Opt_grpid, Opt_nogrpid,
 383         Opt_resgid, Opt_resuid, Opt_sb, Opt_err_cont, Opt_err_panic,
 384         Opt_err_ro, Opt_nouid32, Opt_nocheck, Opt_debug,
 385         Opt_oldalloc, Opt_orlov, Opt_nobh, Opt_user_xattr, Opt_nouser_xattr,
 386         Opt_acl, Opt_noacl, Opt_xip, Opt_ignore, Opt_err, Opt_quota,
 387         Opt_usrquota, Opt_grpquota, Opt_reservation, Opt_noreservation
 388 };
 389
 390 static const match_table_t tokens = {
 391         {Opt_bsd_df, "bsddf"},
 392         {Opt_minix_df, "minixdf"},
 393         {Opt_grpid, "grpid"},
 394         {Opt_grpid, "bsdgroups"},
 395         {Opt_nogrpid, "nogrpid"},
 396         {Opt_nogrpid, "sysvgroups"},
 397         {Opt_resgid, "resgid=%u"},
 398         {Opt_resuid, "resuid=%u"},
 399         {Opt_sb, "sb=%u"},
 400         {Opt_err_cont, "errors=continue"},
 401         {Opt_err_panic, "errors=panic"},
 402         {Opt_err_ro, "errors=remount-ro"},
 403         {Opt_nouid32, "nouid32"},
 404         {Opt_nocheck, "check=none"},
 405         {Opt_nocheck, "nocheck"},
 406         {Opt_debug, "debug"},
 407         {Opt_oldalloc, "oldalloc"},
 408         {Opt_orlov, "orlov"},
 409         {Opt_nobh, "nobh"},
 410         {Opt_user_xattr, "user_xattr"},
 411         {Opt_nouser_xattr, "nouser_xattr"},
 412         {Opt_acl, "acl"},
 413         {Opt_noacl, "noacl"},
 414         {Opt_xip, "xip"},
 415         {Opt_grpquota, "grpquota"},
 416         {Opt_ignore, "noquota"},
 417         {Opt_quota, "quota"},
 418         {Opt_usrquota, "usrquota"},
 419         {Opt_reservation, "reservation"},
 420         {Opt_noreservation, "noreservation"},
 421         {Opt_err, NULL}
 422 };
 423
 424 static int parse_options(char *options, struct super_block *sb)
 425 {
 426         char *p;
 427         struct ext2_sb_info *sbi = EXT2_SB(sb);
 428         substring_t args[MAX_OPT_ARGS];
 429         int option;
 430
 431         if (!options)
 432                 return 1;
 433
 434         while ((p = strsep (&options, ",")) != NULL) {
 435                 int token;
 436                 if (!*p)
 437                         continue;
 438
 439                 token = match_token(p, tokens, args);
 440                 switch (token) {
 441                 case Opt_bsd_df:
 442                         clear_opt (sbi->s_mount_opt, MINIX_DF);
 443                         break;
 444                 case Opt_minix_df:
 445                         set_opt (sbi->s_mount_opt, MINIX_DF);
 446                         break;
 447                 case Opt_grpid:
 448                         set_opt (sbi->s_mount_opt, GRPID);
 449                         break;
 450                 case Opt_nogrpid:
 451                         clear_opt (sbi->s_mount_opt, GRPID);
 452                         break;
 453                 case Opt_resuid:
 454                         if (match_int(&args[0], &option))
 455                                 return 0;
 456                         sbi->s_resuid = option;
 457                         break;
 458                 case Opt_resgid:
 459                         if (match_int(&args[0], &option))
 460                                 return 0;
 461                         sbi->s_resgid = option;
 462                         break;
 463                 case Opt_sb:
 464                         /* handled by get_sb_block() instead of here */
 465                         /* *sb_block = match_int(&args[0]); */
 466                         break;
 467                 case Opt_err_panic:
 468                         clear_opt (sbi->s_mount_opt, ERRORS_CONT);
 469                         clear_opt (sbi->s_mount_opt, ERRORS_RO);
 470                         set_opt (sbi->s_mount_opt, ERRORS_PANIC);
 471                         break;
 472                 case Opt_err_ro:
 473                         clear_opt (sbi->s_mount_opt, ERRORS_CONT);
 474                         clear_opt (sbi->s_mount_opt, ERRORS_PANIC);
 475                         set_opt (sbi->s_mount_opt, ERRORS_RO);
 476                         break;
 477                 case Opt_err_cont:
 478                         clear_opt (sbi->s_mount_opt, ERRORS_RO);
 479                         clear_opt (sbi->s_mount_opt, ERRORS_PANIC);
 480                         set_opt (sbi->s_mount_opt, ERRORS_CONT);
 481                         break;
 482                 case Opt_nouid32:
 483                         set_opt (sbi->s_mount_opt, NO_UID32);
 484                         break;
 485                 case Opt_nocheck:
 486                         clear_opt (sbi->s_mount_opt, CHECK);
 487                         break;
 488                 case Opt_debug:
 489                         set_opt (sbi->s_mount_opt, DEBUG);
 490                         break;
 491                 case Opt_oldalloc:
 492                         set_opt (sbi->s_mount_opt, OLDALLOC);
 493                         break;
 494                 case Opt_orlov:
 495                         clear_opt (sbi->s_mount_opt, OLDALLOC);
 496                         break;
 497                 case Opt_nobh:
 498                         set_opt (sbi->s_mount_opt, NOBH);
 499                         break;
 500 #ifdef CONFIG_EXT2_FS_XATTR
 501                 case Opt_user_xattr:
 502                         set_opt (sbi->s_mount_opt, XATTR_USER);
 503                         break;
 504                 case Opt_nouser_xattr:
 505                         clear_opt (sbi->s_mount_opt, XATTR_USER);
 506                         break;
 507 #else
 508                 case Opt_user_xattr:
 509                 case Opt_nouser_xattr:
 510                         ext2_msg(sb, KERN_INFO, "(no)user_xattr options"
 511                                 "not supported");
 512                         break;
 513 #endif
 514 #ifdef CONFIG_EXT2_FS_POSIX_ACL
 515                 case Opt_acl:
 516                         set_opt(sbi->s_mount_opt, POSIX_ACL);
 517                         break;
 518                 case Opt_noacl:
 519                         clear_opt(sbi->s_mount_opt, POSIX_ACL);
 520                         break;
 521 #else
 522                 case Opt_acl:
 523                 case Opt_noacl:
 524                         ext2_msg(sb, KERN_INFO,
 525                                 "(no)acl options not supported");
 526                         break;
 527 #endif
 528                 case Opt_xip:
 529 #ifdef CONFIG_EXT2_FS_XIP
 530                         set_opt (sbi->s_mount_opt, XIP);
 531 #else
 532                         ext2_msg(sb, KERN_INFO, "xip option not supported");
 533 #endif
 534                         break;
 535
 536 #if defined(CONFIG_QUOTA)
 537                 case Opt_quota:
 538                 case Opt_usrquota:
 539                         set_opt(sbi->s_mount_opt, USRQUOTA);
 540                         break;
 541
 542                 case Opt_grpquota:
 543                         set_opt(sbi->s_mount_opt, GRPQUOTA);
 544                         break;
 545 #else
 546                 case Opt_quota:
 547                 case Opt_usrquota:
 548                 case Opt_grpquota:
 549                         ext2_msg(sb, KERN_INFO,
 550                                 "quota operations not supported");
 551                         break;
 552 #endif
 553
 554                 case Opt_reservation:
 555                         set_opt(sbi->s_mount_opt, RESERVATION);
 556                         ext2_msg(sb, KERN_INFO, "reservations ON");
 557                         break;
 558                 case Opt_noreservation:
 559                         clear_opt(sbi->s_mount_opt, RESERVATION);
 560                         ext2_msg(sb, KERN_INFO, "reservations OFF");
 561                         break;
 562                 case Opt_ignore:
 563                         break;
 564                 default:
 565                         return 0;
 566                 }
 567         }
 568         return 1;
 569 }
 570
 571 static int ext2_setup_super (struct super_block * sb,
 572                               struct ext2_super_block * es,
 573                               int read_only)
 574 {
 575         int res = 0;
 576         struct ext2_sb_info *sbi = EXT2_SB(sb);
 577
 578         if (le32_to_cpu(es->s_rev_level) > EXT2_MAX_SUPP_REV) {
 579                 ext2_msg(sb, KERN_ERR,
 580                         "error: revision level too high, "
 581                         "forcing read-only mode");
 582                 res = MS_RDONLY;
 583         }
 584         if (read_only)
 585                 return res;
 586         if (!(sbi->s_mount_state & EXT2_VALID_FS))
 587                 ext2_msg(sb, KERN_WARNING,
 588                         "warning: mounting unchecked fs, "
 589                         "running e2fsck is recommended");
 590         else if ((sbi->s_mount_state & EXT2_ERROR_FS))
 591                 ext2_msg(sb, KERN_WARNING,
 592                         "warning: mounting fs with errors, "
 593                         "running e2fsck is recommended");
 594         else if ((__s16) le16_to_cpu(es->s_max_mnt_count) >= 0 &&
 595                  le16_to_cpu(es->s_mnt_count) >=
 596                  (unsigned short) (__s16) le16_to_cpu(es->s_max_mnt_count))
 597                 ext2_msg(sb, KERN_WARNING,
 598                         "warning: maximal mount count reached, "
 599                         "running e2fsck is recommended");
 600         else if (le32_to_cpu(es->s_checkinterval) &&
 601                 (le32_to_cpu(es->s_lastcheck) +
 602                         le32_to_cpu(es->s_checkinterval) <= get_seconds()))
 603                 ext2_msg(sb, KERN_WARNING,
 604                         "warning: checktime reached, "
 605                         "running e2fsck is recommended");
 606         if (!le16_to_cpu(es->s_max_mnt_count))
 607                 es->s_max_mnt_count = cpu_to_le16(EXT2_DFL_MAX_MNT_COUNT);
 608         le16_add_cpu(&es->s_mnt_count, 1);
 609         ext2_write_super(sb);
 610         if (test_opt (sb, DEBUG))
 611                 ext2_msg(sb, KERN_INFO, "%s, %s, bs=%lu, fs=%lu, gc=%lu, "
 612                         "bpg=%lu, ipg=%lu, mo=%04lx]",
 613                         EXT2FS_VERSION, EXT2FS_DATE, sb->s_blocksize,
 614                         sbi->s_frag_size,
 615                         sbi->s_groups_count,
 616                         EXT2_BLOCKS_PER_GROUP(sb),
 617                         EXT2_INODES_PER_GROUP(sb),
 618                         sbi->s_mount_opt);
 619         return res;
 620 }
 621
 622 static int ext2_check_descriptors(struct super_block *sb)
 623 {
 624         int i;
 625         struct ext2_sb_info *sbi = EXT2_SB(sb);
 626
 627         ext2_debug ("Checking group descriptors");
 628
 629         for (i = 0; i < sbi->s_groups_count; i++) {
 630                 struct ext2_group_desc *gdp = ext2_get_group_desc(sb, i, NULL);
 631                 ext2_fsblk_t first_block = ext2_group_first_block_no(sb, i);
 632                 ext2_fsblk_t last_block;
 633
 634                 if (i == sbi->s_groups_count - 1)
 635                         last_block = le32_to_cpu(sbi->s_es->s_blocks_count) - 1;
 636                 else
 637                         last_block = first_block +
 638                                 (EXT2_BLOCKS_PER_GROUP(sb) - 1);
 639
 640                 if (le32_to_cpu(gdp->bg_block_bitmap) < first_block ||
 641                     le32_to_cpu(gdp->bg_block_bitmap) > last_block)
 642                 {
 643                         ext2_error (sb, "ext2_check_descriptors",
 644                                     "Block bitmap for group %d"
 645                                     " not in group (block %lu)!",
 646                                     i, (unsigned long) le32_to_cpu(gdp->bg_block_bitmap));
 647                         return 0;
 648                 }
 649                 if (le32_to_cpu(gdp->bg_inode_bitmap) < first_block ||
 650                     le32_to_cpu(gdp->bg_inode_bitmap) > last_block)
 651                 {
 652                         ext2_error (sb, "ext2_check_descriptors",
 653                                     "Inode bitmap for group %d"
 654                                     " not in group (block %lu)!",
 655                                     i, (unsigned long) le32_to_cpu(gdp->bg_inode_bitmap));
 656                         return 0;
 657                 }
 658                 if (le32_to_cpu(gdp->bg_inode_table) < first_block ||
 659                     le32_to_cpu(gdp->bg_inode_table) + sbi->s_itb_per_group - 1 >
 660                     last_block)
 661                 {
 662                         ext2_error (sb, "ext2_check_descriptors",
 663                                     "Inode table for group %d"
 664                                     " not in group (block %lu)!",
 665                                     i, (unsigned long) le32_to_cpu(gdp->bg_inode_table));
 666                         return 0;
 667                 }
 668         }
 669         return 1;
 670 }
 671
 672 /*
 673  * Maximal file size.  There is a direct, and {,double-,triple-}indirect
 674  * block limit, and also a limit of (2^32 - 1) 512-byte sectors in i_blocks.
 675  * We need to be 1 filesystem block less than the 2^32 sector limit.
 676  */
 677 static loff_t ext2_max_size(int bits)
 678 {
 679         loff_t res = EXT2_NDIR_BLOCKS;
 680         int meta_blocks;
 681         loff_t upper_limit;
 682
 683         /* This is calculated to be the largest file size for a
 684          * dense, file such that the total number of
 685          * sectors in the file, including data and all indirect blocks,
 686          * does not exceed 2^32 -1
 687          * __u32 i_blocks representing the total number of
 688          * 512 bytes blocks of the file
 689          */
 690         upper_limit = (1LL << 32) - 1;
 691
 692         /* total blocks in file system block size */
 693         upper_limit >>= (bits - 9);
 694
 695
 696         /* indirect blocks */
 697         meta_blocks = 1;
 698         /* double indirect blocks */
 699         meta_blocks += 1 + (1LL << (bits-2));
 700         /* tripple indirect blocks */
 701         meta_blocks += 1 + (1LL << (bits-2)) + (1LL << (2*(bits-2)));
 702
 703         upper_limit -= meta_blocks;
 704         upper_limit <<= bits;
 705
 706         res += 1LL << (bits-2);
 707         res += 1LL << (2*(bits-2));
 708         res += 1LL << (3*(bits-2));
 709         res <<= bits;
 710         if (res > upper_limit)
 711                 res = upper_limit;
 712
 713         if (res > MAX_LFS_FILESIZE)
 714                 res = MAX_LFS_FILESIZE;
 715
 716         return res;
 717 }
 718
 719 static unsigned long descriptor_loc(struct super_block *sb,
 720                                     unsigned long logic_sb_block,
 721                                     int nr)
 722 {
 723         struct ext2_sb_info *sbi = EXT2_SB(sb);
 724         unsigned long bg, first_meta_bg;
 725         int has_super = 0;
 726
 727         first_meta_bg = le32_to_cpu(sbi->s_es->s_first_meta_bg);
 728
 729         if (!EXT2_HAS_INCOMPAT_FEATURE(sb, EXT2_FEATURE_INCOMPAT_META_BG) ||
 730             nr < first_meta_bg)
 731                 return (logic_sb_block + nr + 1);
 732         bg = sbi->s_desc_per_block * nr;
 733         if (ext2_bg_has_super(sb, bg))
 734                 has_super = 1;
 735
 736         return ext2_group_first_block_no(sb, bg) + has_super;
 737 }
 738
 739 static int ext2_fill_super(struct super_block *sb, void *data, int silent)
 740 {
 741         struct buffer_head * bh;
 742         struct ext2_sb_info * sbi;
 743         struct ext2_super_block * es;
 744         struct inode *root;
 745         unsigned long block;
 746         unsigned long sb_block = get_sb_block(&data);
 747         unsigned long logic_sb_block;
 748         unsigned long offset = 0;
 749         unsigned long def_mount_opts;
 750         long ret = -EINVAL;
 751         int blocksize = BLOCK_SIZE;
 752         int db_count;
 753         int i, j;
 754         __le32 features;
 755         int err;
 756
 757         sbi = kzalloc(sizeof(*sbi), GFP_KERNEL);
 758         if (!sbi)
 759                 return -ENOMEM;
 760
 761         sbi->s_blockgroup_lock =
 762                 kzalloc(sizeof(struct blockgroup_lock), GFP_KERNEL);
 763         if (!sbi->s_blockgroup_lock) {
 764                 kfree(sbi);
 765                 return -ENOMEM;
 766         }
 767         sb->s_fs_info = sbi;
 768         sbi->s_sb_block = sb_block;
 769
 770         /*
 771          * See what the current blocksize for the device is, and
 772          * use that as the blocksize.  Otherwise (or if the blocksize
 773          * is smaller than the default) use the default.
 774          * This is important for devices that have a hardware
 775          * sectorsize that is larger than the default.
 776          */
 777         blocksize = sb_min_blocksize(sb, BLOCK_SIZE);
 778         if (!blocksize) {
 779                 ext2_msg(sb, KERN_ERR, "error: unable to set blocksize");
 780                 goto failed_sbi;
 781         }
 782
 783         /*
 784          * If the superblock doesn't start on a hardware sector boundary,
 785          * calculate the offset.
 786          */
 787         if (blocksize != BLOCK_SIZE) {
 788                 logic_sb_block = (sb_block*BLOCK_SIZE) / blocksize;
 789                 offset = (sb_block*BLOCK_SIZE) % blocksize;
 790         } else {
 791                 logic_sb_block = sb_block;
 792         }
 793
 794         if (!(bh = sb_bread(sb, logic_sb_block))) {
 795                 ext2_msg(sb, KERN_ERR, "error: unable to read superblock");
 796                 goto failed_sbi;
 797         }
 798         /*
 799          * Note: s_es must be initialized as soon as possible because
 800          *       some ext2 macro-instructions depend on its value
 801          */
 802         es = (struct ext2_super_block *) (((char *)bh->b_data) + offset);
 803         sbi->s_es = es;
 804         sb->s_magic = le16_to_cpu(es->s_magic);
 805
 806         if (sb->s_magic != EXT2_SUPER_MAGIC)
 807                 goto cantfind_ext2;
 808
 809         /* Set defaults before we parse the mount options */
 810         def_mount_opts = le32_to_cpu(es->s_default_mount_opts);
 811         if (def_mount_opts & EXT2_DEFM_DEBUG)
 812                 set_opt(sbi->s_mount_opt, DEBUG);
 813         if (def_mount_opts & EXT2_DEFM_BSDGROUPS)
 814                 set_opt(sbi->s_mount_opt, GRPID);
 815         if (def_mount_opts & EXT2_DEFM_UID16)
 816                 set_opt(sbi->s_mount_opt, NO_UID32);
 817 #ifdef CONFIG_EXT2_FS_XATTR
 818         if (def_mount_opts & EXT2_DEFM_XATTR_USER)
 819                 set_opt(sbi->s_mount_opt, XATTR_USER);
 820 #endif
 821 #ifdef CONFIG_EXT2_FS_POSIX_ACL
 822         if (def_mount_opts & EXT2_DEFM_ACL)
 823                 set_opt(sbi->s_mount_opt, POSIX_ACL);
 824 #endif
 825
 826         if (le16_to_cpu(sbi->s_es->s_errors) == EXT2_ERRORS_PANIC)
 827                 set_opt(sbi->s_mount_opt, ERRORS_PANIC);
 828         else if (le16_to_cpu(sbi->s_es->s_errors) == EXT2_ERRORS_CONTINUE)
 829                 set_opt(sbi->s_mount_opt, ERRORS_CONT);
 830         else
 831                 set_opt(sbi->s_mount_opt, ERRORS_RO);
 832
 833         sbi->s_resuid = le16_to_cpu(es->s_def_resuid);
 834         sbi->s_resgid = le16_to_cpu(es->s_def_resgid);
 835
 836         set_opt(sbi->s_mount_opt, RESERVATION);
 837
 838         if (!parse_options((char *) data, sb))
 839                 goto failed_mount;
 840
 841         sb->s_flags = (sb->s_flags & ~MS_POSIXACL) |
 842                 ((EXT2_SB(sb)->s_mount_opt & EXT2_MOUNT_POSIX_ACL) ?
 843                  MS_POSIXACL : 0);
 844
 845         ext2_xip_verify_sb(sb); /* see if bdev supports xip, unset
 846                                     EXT2_MOUNT_XIP if not */
 847
 848         if (le32_to_cpu(es->s_rev_level) == EXT2_GOOD_OLD_REV &&
 849             (EXT2_HAS_COMPAT_FEATURE(sb, ~0U) ||
 850              EXT2_HAS_RO_COMPAT_FEATURE(sb, ~0U) ||
 851              EXT2_HAS_INCOMPAT_FEATURE(sb, ~0U)))
 852                 ext2_msg(sb, KERN_WARNING,
 853                         "warning: feature flags set on rev 0 fs, "
 854                         "running e2fsck is recommended");
 855         /*
 856          * Check feature flags regardless of the revision level, since we
 857          * previously didn't change the revision level when setting the flags,
 858          * so there is a chance incompat flags are set on a rev 0 filesystem.
 859          */
 860         features = EXT2_HAS_INCOMPAT_FEATURE(sb, ~EXT2_FEATURE_INCOMPAT_SUPP);
 861         if (features) {
 862                 ext2_msg(sb, KERN_ERR,  "error: couldn't mount because of "
 863                        "unsupported optional features (%x)",
 864                         le32_to_cpu(features));
 865                 goto failed_mount;
 866         }
 867         if (!(sb->s_flags & MS_RDONLY) &&
 868             (features = EXT2_HAS_RO_COMPAT_FEATURE(sb, ~EXT2_FEATURE_RO_COMPAT_SUPP))){
 869                 ext2_msg(sb, KERN_ERR, "error: couldn't mount RDWR because of "
 870                        "unsupported optional features (%x)",
 871                        le32_to_cpu(features));
 872                 goto failed_mount;
 873         }
 874
 875         blocksize = BLOCK_SIZE << le32_to_cpu(sbi->s_es->s_log_block_size);
 876
 877         if (ext2_use_xip(sb) && blocksize != PAGE_SIZE) {
 878                 if (!silent)
 879                         ext2_msg(sb, KERN_ERR,
 880                                 "error: unsupported blocksize for xip");
 881                 goto failed_mount;
 882         }
 883
 884         /* If the blocksize doesn't match, re-read the thing.. */
 885         if (sb->s_blocksize != blocksize) {
 886                 brelse(bh);
 887
 888                 if (!sb_set_blocksize(sb, blocksize)) {
 889                         ext2_msg(sb, KERN_ERR, "error: blocksize is too small");
 890                         goto failed_sbi;
 891                 }
 892
 893                 logic_sb_block = (sb_block*BLOCK_SIZE) / blocksize;
 894                 offset = (sb_block*BLOCK_SIZE) % blocksize;
 895                 bh = sb_bread(sb, logic_sb_block);
 896                 if(!bh) {
 897                         ext2_msg(sb, KERN_ERR, "error: couldn't read"
 898                                 "superblock on 2nd try");
 899                         goto failed_sbi;
 900                 }
 901                 es = (struct ext2_super_block *) (((char *)bh->b_data) + offset);
 902                 sbi->s_es = es;
 903                 if (es->s_magic != cpu_to_le16(EXT2_SUPER_MAGIC)) {
 904                         ext2_msg(sb, KERN_ERR, "error: magic mismatch");
 905                         goto failed_mount;
 906                 }
 907         }
 908
 909         sb->s_maxbytes = ext2_max_size(sb->s_blocksize_bits);
 910
 911         if (le32_to_cpu(es->s_rev_level) == EXT2_GOOD_OLD_REV) {
 912                 sbi->s_inode_size = EXT2_GOOD_OLD_INODE_SIZE;
 913                 sbi->s_first_ino = EXT2_GOOD_OLD_FIRST_INO;
 914         } else {
 915                 sbi->s_inode_size = le16_to_cpu(es->s_inode_size);
 916                 sbi->s_first_ino = le32_to_cpu(es->s_first_ino);
 917                 if ((sbi->s_inode_size < EXT2_GOOD_OLD_INODE_SIZE) ||
 918                     !is_power_of_2(sbi->s_inode_size) ||
 919                     (sbi->s_inode_size > blocksize)) {
 920                         ext2_msg(sb, KERN_ERR,
 921                                 "error: unsupported inode size: %d",
 922                                 sbi->s_inode_size);
 923                         goto failed_mount;
 924                 }
 925         }
 926
 927         sbi->s_frag_size = EXT2_MIN_FRAG_SIZE <<
 928                                    le32_to_cpu(es->s_log_frag_size);
 929         if (sbi->s_frag_size == 0)
 930                 goto cantfind_ext2;
 931         sbi->s_frags_per_block = sb->s_blocksize / sbi->s_frag_size;
 932
 933         sbi->s_blocks_per_group = le32_to_cpu(es->s_blocks_per_group);
 934         sbi->s_frags_per_group = le32_to_cpu(es->s_frags_per_group);
 935         sbi->s_inodes_per_group = le32_to_cpu(es->s_inodes_per_group);
 936
 937         if (EXT2_INODE_SIZE(sb) == 0)
 938                 goto cantfind_ext2;
 939         sbi->s_inodes_per_block = sb->s_blocksize / EXT2_INODE_SIZE(sb);
 940         if (sbi->s_inodes_per_block == 0 || sbi->s_inodes_per_group == 0)
 941                 goto cantfind_ext2;
 942         sbi->s_itb_per_group = sbi->s_inodes_per_group /
 943                                         sbi->s_inodes_per_block;
 944         sbi->s_desc_per_block = sb->s_blocksize /
 945                                         sizeof (struct ext2_group_desc);
 946         sbi->s_sbh = bh;
 947         sbi->s_mount_state = le16_to_cpu(es->s_state);
 948         sbi->s_addr_per_block_bits =
 949                 ilog2 (EXT2_ADDR_PER_BLOCK(sb));
 950         sbi->s_desc_per_block_bits =
 951                 ilog2 (EXT2_DESC_PER_BLOCK(sb));
 952
 953         if (sb->s_magic != EXT2_SUPER_MAGIC)
 954                 goto cantfind_ext2;
 955
 956         if (sb->s_blocksize != bh->b_size) {
 957                 if (!silent)
 958                         ext2_msg(sb, KERN_ERR, "error: unsupported blocksize");
 959                 goto failed_mount;
 960         }
 961
 962         if (sb->s_blocksize != sbi->s_frag_size) {
 963                 ext2_msg(sb, KERN_ERR,
 964                         "error: fragsize %lu != blocksize %lu"
 965                         "(not supported yet)",
 966                         sbi->s_frag_size, sb->s_blocksize);
 967                 goto failed_mount;
 968         }
 969
 970         if (sbi->s_blocks_per_group > sb->s_blocksize * 8) {
 971                 ext2_msg(sb, KERN_ERR,
 972                         "error: #blocks per group too big: %lu",
 973                         sbi->s_blocks_per_group);
 974                 goto failed_mount;
 975         }
 976         if (sbi->s_frags_per_group > sb->s_blocksize * 8) {
 977                 ext2_msg(sb, KERN_ERR,
 978                         "error: #fragments per group too big: %lu",
 979                         sbi->s_frags_per_group);
 980                 goto failed_mount;
 981         }
 982         if (sbi->s_inodes_per_group > sb->s_blocksize * 8) {
 983                 ext2_msg(sb, KERN_ERR,
 984                         "error: #inodes per group too big: %lu",
 985                         sbi->s_inodes_per_group);
 986                 goto failed_mount;
 987         }
 988
 989         if (EXT2_BLOCKS_PER_GROUP(sb) == 0)
 990                 goto cantfind_ext2;
 991         sbi->s_groups_count = ((le32_to_cpu(es->s_blocks_count) -
 992                                 le32_to_cpu(es->s_first_data_block) - 1)
 993                                         / EXT2_BLOCKS_PER_GROUP(sb)) + 1;
 994         db_count = (sbi->s_groups_count + EXT2_DESC_PER_BLOCK(sb) - 1) /
 995                    EXT2_DESC_PER_BLOCK(sb);
 996         sbi->s_group_desc = kmalloc (db_count * sizeof (struct buffer_head *), GFP_KERNEL);
 997         if (sbi->s_group_desc == NULL) {
 998                 ext2_msg(sb, KERN_ERR, "error: not enough memory");
 999                 goto failed_mount;
1000         }
1001         bgl_lock_init(sbi->s_blockgroup_lock);
1002         sbi->s_debts = kcalloc(sbi->s_groups_count, sizeof(*sbi->s_debts), GFP_KERNEL);
1003         if (!sbi->s_debts) {
1004                 ext2_msg(sb, KERN_ERR, "error: not enough memory");
1005                 goto failed_mount_group_desc;
1006         }
1007         for (i = 0; i < db_count; i++) {
1008                 block = descriptor_loc(sb, logic_sb_block, i);
1009                 sbi->s_group_desc[i] = sb_bread(sb, block);
1010                 if (!sbi->s_group_desc[i]) {
1011                         for (j = 0; j < i; j++)
1012                                 brelse (sbi->s_group_desc[j]);
1013                         ext2_msg(sb, KERN_ERR,
1014                                 "error: unable to read group descriptors");
1015                         goto failed_mount_group_desc;
1016                 }
1017         }
1018         if (!ext2_check_descriptors (sb)) {
1019                 ext2_msg(sb, KERN_ERR, "group descriptors corrupted");
1020                 goto failed_mount2;
1021         }
1022         sbi->s_gdb_count = db_count;
1023         get_random_bytes(&sbi->s_next_generation, sizeof(u32));
1024         spin_lock_init(&sbi->s_next_gen_lock);
1025
1026         /* per fileystem reservation list head & lock */
1027         spin_lock_init(&sbi->s_rsv_window_lock);
1028         sbi->s_rsv_window_root = RB_ROOT;
1029         /*
1030          * Add a single, static dummy reservation to the start of the
1031          * reservation window list --- it gives us a placeholder for
1032          * append-at-start-of-list which makes the allocation logic
1033          * _much_ simpler.
1034          */
1035         sbi->s_rsv_window_head.rsv_start = EXT2_RESERVE_WINDOW_NOT_ALLOCATED;
1036         sbi->s_rsv_window_head.rsv_end = EXT2_RESERVE_WINDOW_NOT_ALLOCATED;
1037         sbi->s_rsv_window_head.rsv_alloc_hit = 0;
1038         sbi->s_rsv_window_head.rsv_goal_size = 0;
1039         ext2_rsv_window_add(sb, &sbi->s_rsv_window_head);
1040
1041         err = percpu_counter_init(&sbi->s_freeblocks_counter,
1042                                 ext2_count_free_blocks(sb));
1043         if (!err) {
1044                 err = percpu_counter_init(&sbi->s_freeinodes_counter,
1045                                 ext2_count_free_inodes(sb));
1046         }
1047         if (!err) {
1048                 err = percpu_counter_init(&sbi->s_dirs_counter,
1049                                 ext2_count_dirs(sb));
1050         }
1051         if (err) {
1052                 ext2_msg(sb, KERN_ERR, "error: insufficient memory");
1053                 goto failed_mount3;
1054         }
1055         /*
1056          * set up enough so that it can read an inode
1057          */
1058         sb->s_op = &ext2_sops;
1059         sb->s_export_op = &ext2_export_ops;
1060         sb->s_xattr = ext2_xattr_handlers;
1061         root = ext2_iget(sb, EXT2_ROOT_INO);
1062         if (IS_ERR(root)) {
1063                 ret = PTR_ERR(root);
1064                 goto failed_mount3;
1065         }
1066         if (!S_ISDIR(root->i_mode) || !root->i_blocks || !root->i_size) {
1067                 iput(root);
1068                 ext2_msg(sb, KERN_ERR, "error: corrupt root inode, run e2fsck");
1069                 goto failed_mount3;
1070         }
1071
1072         sb->s_root = d_alloc_root(root);
1073         if (!sb->s_root) {
1074                 iput(root);
1075                 ext2_msg(sb, KERN_ERR, "error: get root inode failed");
1076                 ret = -ENOMEM;
1077                 goto failed_mount3;
1078         }
1079         if (EXT2_HAS_COMPAT_FEATURE(sb, EXT3_FEATURE_COMPAT_HAS_JOURNAL))
1080                 ext2_msg(sb, KERN_WARNING,
1081                         "warning: mounting ext3 filesystem as ext2");
1082         ext2_setup_super (sb, es, sb->s_flags & MS_RDONLY);
1083         return 0;
1084
1085 cantfind_ext2:
1086         if (!silent)
1087                 ext2_msg(sb, KERN_ERR,
1088                         "error: can't find an ext2 filesystem on dev %s.",
1089                         sb->s_id);
1090         goto failed_mount;
1091 failed_mount3:
1092         percpu_counter_destroy(&sbi->s_freeblocks_counter);
1093         percpu_counter_destroy(&sbi->s_freeinodes_counter);
1094         percpu_counter_destroy(&sbi->s_dirs_counter);
1095 failed_mount2:
1096         for (i = 0; i < db_count; i++)
1097                 brelse(sbi->s_group_desc[i]);
1098 failed_mount_group_desc:
1099         kfree(sbi->s_group_desc);
1100         kfree(sbi->s_debts);
1101 failed_mount:
1102         brelse(bh);
1103 failed_sbi:
1104         sb->s_fs_info = NULL;
1105         kfree(sbi->s_blockgroup_lock);
1106         kfree(sbi);
1107         return ret;
1108 }
1109
1110 static void ext2_clear_super_error(struct super_block *sb)
1111 {
1112         struct buffer_head *sbh = EXT2_SB(sb)->s_sbh;
1113
1114         if (buffer_write_io_error(sbh)) {
1115                 /*
1116                  * Oh, dear.  A previous attempt to write the
1117                  * superblock failed.  This could happen because the
1118                  * USB device was yanked out.  Or it could happen to
1119                  * be a transient write error and maybe the block will
1120                  * be remapped.  Nothing we can do but to retry the
1121                  * write and hope for the best.
1122                  */
1123                 printk(KERN_ERR "EXT2-fs: %s previous I/O error to "
1124                        "superblock detected", sb->s_id);
1125                 clear_buffer_write_io_error(sbh);
1126                 set_buffer_uptodate(sbh);
1127         }
1128 }
1129
1130 static void ext2_commit_super (struct super_block * sb,
1131                                struct ext2_super_block * es)
1132 {
1133         ext2_clear_super_error(sb);
1134         es->s_wtime = cpu_to_le32(get_seconds());
1135         mark_buffer_dirty(EXT2_SB(sb)->s_sbh);
1136         sb->s_dirt = 0;
1137 }
1138
1139 static void ext2_sync_super(struct super_block *sb, struct ext2_super_block *es)
1140 {
1141         ext2_clear_super_error(sb);
1142         es->s_free_blocks_count = cpu_to_le32(ext2_count_free_blocks(sb));
1143         es->s_free_inodes_count = cpu_to_le32(ext2_count_free_inodes(sb));
1144         es->s_wtime = cpu_to_le32(get_seconds());
1145         mark_buffer_dirty(EXT2_SB(sb)->s_sbh);
1146         sync_dirty_buffer(EXT2_SB(sb)->s_sbh);
1147         sb->s_dirt = 0;
1148 }
1149
1150 /*
1151  * In the second extended file system, it is not necessary to
1152  * write the super block since we use a mapping of the
1153  * disk super block in a buffer.
1154  *
1155  * However, this function is still used to set the fs valid
1156  * flags to 0.  We need to set this flag to 0 since the fs
1157  * may have been checked while mounted and e2fsck may have
1158  * set s_state to EXT2_VALID_FS after some corrections.
1159  */
1160
1161 static int ext2_sync_fs(struct super_block *sb, int wait)
1162 {
1163         struct ext2_super_block *es = EXT2_SB(sb)->s_es;
1164         struct buffer_head *sbh = EXT2_SB(sb)->s_sbh;
1165
1166         lock_kernel();
1167         if (buffer_write_io_error(sbh)) {
1168                 /*
1169                  * Oh, dear.  A previous attempt to write the
1170                  * superblock failed.  This could happen because the
1171                  * USB device was yanked out.  Or it could happen to
1172                  * be a transient write error and maybe the block will
1173                  * be remapped.  Nothing we can do but to retry the
1174                  * write and hope for the best.
1175                  */
1176                 ext2_msg(sb, KERN_ERR,
1177                        "previous I/O error to superblock detected\n");
1178                 clear_buffer_write_io_error(sbh);
1179                 set_buffer_uptodate(sbh);
1180         }
1181
1182         if (es->s_state & cpu_to_le16(EXT2_VALID_FS)) {
1183                 ext2_debug("setting valid to 0\n");
1184                 es->s_state &= cpu_to_le16(~EXT2_VALID_FS);
1185                 es->s_free_blocks_count =
1186                         cpu_to_le32(ext2_count_free_blocks(sb));
1187                 es->s_free_inodes_count =
1188                         cpu_to_le32(ext2_count_free_inodes(sb));
1189                 es->s_mtime = cpu_to_le32(get_seconds());
1190                 ext2_sync_super(sb, es);
1191         } else {
1192                 ext2_commit_super(sb, es);
1193         }
1194         sb->s_dirt = 0;
1195         unlock_kernel();
1196
1197         return 0;
1198 }
1199
1200
1201 void ext2_write_super(struct super_block *sb)
1202 {
1203         if (!(sb->s_flags & MS_RDONLY))
1204                 ext2_sync_fs(sb, 1);
1205         else
1206                 sb->s_dirt = 0;
1207 }
1208
1209 static int ext2_remount (struct super_block * sb, int * flags, char * data)
1210 {
1211         struct ext2_sb_info * sbi = EXT2_SB(sb);
1212         struct ext2_super_block * es;
1213         unsigned long old_mount_opt = sbi->s_mount_opt;
1214         struct ext2_mount_options old_opts;
1215         unsigned long old_sb_flags;
1216         int err;
1217
1218         lock_kernel();
1219
1220         /* Store the old options */
1221         old_sb_flags = sb->s_flags;
1222         old_opts.s_mount_opt = sbi->s_mount_opt;
1223         old_opts.s_resuid = sbi->s_resuid;
1224         old_opts.s_resgid = sbi->s_resgid;
1225
1226         /*
1227          * Allow the "check" option to be passed as a remount option.
1228          */
1229         if (!parse_options(data, sb)) {
1230                 err = -EINVAL;
1231                 goto restore_opts;
1232         }
1233
1234         sb->s_flags = (sb->s_flags & ~MS_POSIXACL) |
1235                 ((sbi->s_mount_opt & EXT2_MOUNT_POSIX_ACL) ? MS_POSIXACL : 0);
1236
1237         ext2_xip_verify_sb(sb); /* see if bdev supports xip, unset
1238                                     EXT2_MOUNT_XIP if not */
1239
1240         if ((ext2_use_xip(sb)) && (sb->s_blocksize != PAGE_SIZE)) {
1241                 ext2_msg(sb, KERN_WARNING,
1242                         "warning: unsupported blocksize for xip");
1243                 err = -EINVAL;
1244                 goto restore_opts;
1245         }
1246
1247         es = sbi->s_es;
1248         if (((sbi->s_mount_opt & EXT2_MOUNT_XIP) !=
1249             (old_mount_opt & EXT2_MOUNT_XIP)) &&
1250             invalidate_inodes(sb)) {
1251                 ext2_msg(sb, KERN_WARNING, "warning: refusing change of "
1252                          "xip flag with busy inodes while remounting");
1253                 sbi->s_mount_opt &= ~EXT2_MOUNT_XIP;
1254                 sbi->s_mount_opt |= old_mount_opt & EXT2_MOUNT_XIP;
1255         }
1256         if ((*flags & MS_RDONLY) == (sb->s_flags & MS_RDONLY)) {
1257                 unlock_kernel();
1258                 return 0;
1259         }
1260         if (*flags & MS_RDONLY) {
1261                 if (le16_to_cpu(es->s_state) & EXT2_VALID_FS ||
1262                     !(sbi->s_mount_state & EXT2_VALID_FS)) {
1263                         unlock_kernel();
1264                         return 0;
1265                 }
1266                 /*
1267                  * OK, we are remounting a valid rw partition rdonly, so set
1268                  * the rdonly flag and then mark the partition as valid again.
1269                  */
1270                 es->s_state = cpu_to_le16(sbi->s_mount_state);
1271                 es->s_mtime = cpu_to_le32(get_seconds());
1272         } else {
1273                 __le32 ret = EXT2_HAS_RO_COMPAT_FEATURE(sb,
1274                                                ~EXT2_FEATURE_RO_COMPAT_SUPP);
1275                 if (ret) {
1276                         ext2_msg(sb, KERN_WARNING,
1277                                 "warning: couldn't remount RDWR because of "
1278                                 "unsupported optional features (%x).",
1279                                 le32_to_cpu(ret));
1280                         err = -EROFS;
1281                         goto restore_opts;
1282                 }
1283                 /*
1284                  * Mounting a RDONLY partition read-write, so reread and
1285                  * store the current valid flag.  (It may have been changed
1286                  * by e2fsck since we originally mounted the partition.)
1287                  */
1288                 sbi->s_mount_state = le16_to_cpu(es->s_state);
1289                 if (!ext2_setup_super (sb, es, 0))
1290                         sb->s_flags &= ~MS_RDONLY;
1291         }
1292         ext2_sync_super(sb, es);
1293         unlock_kernel();
1294         return 0;
1295 restore_opts:
1296         sbi->s_mount_opt = old_opts.s_mount_opt;
1297         sbi->s_resuid = old_opts.s_resuid;
1298         sbi->s_resgid = old_opts.s_resgid;
1299         sb->s_flags = old_sb_flags;
1300         unlock_kernel();
1301         return err;
1302 }
1303
1304 static int ext2_statfs (struct dentry * dentry, struct kstatfs * buf)
1305 {
1306         struct super_block *sb = dentry->d_sb;
1307         struct ext2_sb_info *sbi = EXT2_SB(sb);
1308         struct ext2_super_block *es = sbi->s_es;
1309         u64 fsid;
1310
1311         if (test_opt (sb, MINIX_DF))
1312                 sbi->s_overhead_last = 0;
1313         else if (sbi->s_blocks_last != le32_to_cpu(es->s_blocks_count)) {
1314                 unsigned long i, overhead = 0;
1315                 smp_rmb();
1316
1317                 /*
1318                  * Compute the overhead (FS structures). This is constant
1319                  * for a given filesystem unless the number of block groups
1320                  * changes so we cache the previous value until it does.
1321                  */
1322
1323                 /*
1324                  * All of the blocks before first_data_block are
1325                  * overhead
1326                  */
1327                 overhead = le32_to_cpu(es->s_first_data_block);
1328
1329                 /*
1330                  * Add the overhead attributed to the superblock and
1331                  * block group descriptors.  If the sparse superblocks
1332                  * feature is turned on, then not all groups have this.
1333                  */
1334                 for (i = 0; i < sbi->s_groups_count; i++)
1335                         overhead += ext2_bg_has_super(sb, i) +
1336                                 ext2_bg_num_gdb(sb, i);
1337
1338                 /*
1339                  * Every block group has an inode bitmap, a block
1340                  * bitmap, and an inode table.
1341                  */
1342                 overhead += (sbi->s_groups_count *
1343                              (2 + sbi->s_itb_per_group));
1344                 sbi->s_overhead_last = overhead;
1345                 smp_wmb();
1346                 sbi->s_blocks_last = le32_to_cpu(es->s_blocks_count);
1347         }
1348
1349         buf->f_type = EXT2_SUPER_MAGIC;
1350         buf->f_bsize = sb->s_blocksize;
1351         buf->f_blocks = le32_to_cpu(es->s_blocks_count) - sbi->s_overhead_last;
1352         buf->f_bfree = ext2_count_free_blocks(sb);
1353         es->s_free_blocks_count = cpu_to_le32(buf->f_bfree);
1354         buf->f_bavail = buf->f_bfree - le32_to_cpu(es->s_r_blocks_count);
1355         if (buf->f_bfree < le32_to_cpu(es->s_r_blocks_count))
1356                 buf->f_bavail = 0;
1357         buf->f_files = le32_to_cpu(es->s_inodes_count);
1358         buf->f_ffree = ext2_count_free_inodes(sb);
1359         es->s_free_inodes_count = cpu_to_le32(buf->f_ffree);
1360         buf->f_namelen = EXT2_NAME_LEN;
1361         fsid = le64_to_cpup((void *)es->s_uuid) ^
1362                le64_to_cpup((void *)es->s_uuid + sizeof(u64));
1363         buf->f_fsid.val[0] = fsid & 0xFFFFFFFFUL;
1364         buf->f_fsid.val[1] = (fsid >> 32) & 0xFFFFFFFFUL;
1365         return 0;
1366 }
1367
1368 static int ext2_get_sb(struct file_system_type *fs_type,
1369         int flags, const char *dev_name, void *data, struct vfsmount *mnt)
1370 {
1371         return get_sb_bdev(fs_type, flags, dev_name, data, ext2_fill_super, mnt);
1372 }
1373
1374 #ifdef CONFIG_QUOTA
1375
1376 /* Read data from quotafile - avoid pagecache and such because we cannot afford
1377  * acquiring the locks... As quota files are never truncated and quota code
1378  * itself serializes the operations (and noone else should touch the files)
1379  * we don't have to be afraid of races */
1380 static ssize_t ext2_quota_read(struct super_block *sb, int type, char *data,
1381                                size_t len, loff_t off)
1382 {
1383         struct inode *inode = sb_dqopt(sb)->files[type];
1384         sector_t blk = off >> EXT2_BLOCK_SIZE_BITS(sb);
1385         int err = 0;
1386         int offset = off & (sb->s_blocksize - 1);
1387         int tocopy;
1388         size_t toread;
1389         struct buffer_head tmp_bh;
1390         struct buffer_head *bh;
1391         loff_t i_size = i_size_read(inode);
1392
1393         if (off > i_size)
1394                 return 0;
1395         if (off+len > i_size)
1396                 len = i_size-off;
1397         toread = len;
1398         while (toread > 0) {
1399                 tocopy = sb->s_blocksize - offset < toread ?
1400                                 sb->s_blocksize - offset : toread;
1401
1402                 tmp_bh.b_state = 0;
1403                 tmp_bh.b_size = sb->s_blocksize;
1404                 err = ext2_get_block(inode, blk, &tmp_bh, 0);
1405                 if (err < 0)
1406                         return err;
1407                 if (!buffer_mapped(&tmp_bh))    /* A hole? */
1408                         memset(data, 0, tocopy);
1409                 else {
1410                         bh = sb_bread(sb, tmp_bh.b_blocknr);
1411                         if (!bh)
1412                                 return -EIO;
1413                         memcpy(data, bh->b_data+offset, tocopy);
1414                         brelse(bh);
1415                 }
1416                 offset = 0;
1417                 toread -= tocopy;
1418                 data += tocopy;
1419                 blk++;
1420         }
1421         return len;
1422 }
1423
1424 /* Write to quotafile */
1425 static ssize_t ext2_quota_write(struct super_block *sb, int type,
1426                                 const char *data, size_t len, loff_t off)
1427 {
1428         struct inode *inode = sb_dqopt(sb)->files[type];
1429         sector_t blk = off >> EXT2_BLOCK_SIZE_BITS(sb);
1430         int err = 0;
1431         int offset = off & (sb->s_blocksize - 1);
1432         int tocopy;
1433         size_t towrite = len;
1434         struct buffer_head tmp_bh;
1435         struct buffer_head *bh;
1436
1437         mutex_lock_nested(&inode->i_mutex, I_MUTEX_QUOTA);
1438         while (towrite > 0) {
1439                 tocopy = sb->s_blocksize - offset < towrite ?
1440                                 sb->s_blocksize - offset : towrite;
1441
1442                 tmp_bh.b_state = 0;
1443                 err = ext2_get_block(inode, blk, &tmp_bh, 1);
1444                 if (err < 0)
1445                         goto out;
1446                 if (offset || tocopy != EXT2_BLOCK_SIZE(sb))
1447                         bh = sb_bread(sb, tmp_bh.b_blocknr);
1448                 else
1449                         bh = sb_getblk(sb, tmp_bh.b_blocknr);
1450                 if (!bh) {
1451                         err = -EIO;
1452                         goto out;
1453                 }
1454                 lock_buffer(bh);
1455                 memcpy(bh->b_data+offset, data, tocopy);
1456                 flush_dcache_page(bh->b_page);
1457                 set_buffer_uptodate(bh);
1458                 mark_buffer_dirty(bh);
1459                 unlock_buffer(bh);
1460                 brelse(bh);
1461                 offset = 0;
1462                 towrite -= tocopy;
1463                 data += tocopy;
1464                 blk++;
1465         }
1466 out:
1467         if (len == towrite) {
1468                 mutex_unlock(&inode->i_mutex);
1469                 return err;
1470         }
1471         if (inode->i_size < off+len-towrite)
1472                 i_size_write(inode, off+len-towrite);
1473         inode->i_version++;
1474         inode->i_mtime = inode->i_ctime = CURRENT_TIME;
1475         mark_inode_dirty(inode);
1476         mutex_unlock(&inode->i_mutex);
1477         return len - towrite;
1478 }
1479
1480 #endif
1481
1482 static struct file_system_type ext2_fs_type = {
1483         .owner          = THIS_MODULE,
1484         .name           = "ext2",
1485         .get_sb         = ext2_get_sb,
1486         .kill_sb        = kill_block_super,
1487         .fs_flags       = FS_REQUIRES_DEV,
1488 };
1489
1490 static int __init init_ext2_fs(void)
1491 {
1492         int err = init_ext2_xattr();
1493         if (err)
1494                 return err;
1495         err = init_inodecache();
1496         if (err)
1497                 goto out1;
1498         err = register_filesystem(&ext2_fs_type);
1499         if (err)
1500                 goto out;
1501         return 0;
1502 out:
1503         destroy_inodecache();
1504 out1:
1505         exit_ext2_xattr();
1506         return err;
1507 }
1508
1509 static void __exit exit_ext2_fs(void)
1510 {
1511         unregister_filesystem(&ext2_fs_type);
1512         destroy_inodecache();
1513         exit_ext2_xattr();
1514 }
1515
1516 module_init(init_ext2_fs)
1517 module_exit(exit_ext2_fs)