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/vfs.h>
  30 #include <linux/seq_file.h>
  31 #include <linux/mount.h>
  32 #include <linux/log2.h>
  33 #include <linux/quotaops.h>
  34 #include <asm/uaccess.h>
  35 #include "ext2.h"
  36 #include "xattr.h"
  37 #include "acl.h"
  38 #include "xip.h"
  39
  40 static void ext2_sync_super(struct super_block *sb,
  41                             struct ext2_super_block *es, int wait);
  42 static int ext2_remount (struct super_block * sb, int * flags, char * data);
  43 static int ext2_statfs (struct dentry * dentry, struct kstatfs * buf);
  44 static int ext2_sync_fs(struct super_block *sb, int wait);
  45
  46 void ext2_error(struct super_block *sb, const char *function,
  47                 const char *fmt, ...)
  48 {
  49         struct va_format vaf;
  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                 spin_lock(&sbi->s_lock);
  56                 sbi->s_mount_state |= EXT2_ERROR_FS;
  57                 es->s_state |= cpu_to_le16(EXT2_ERROR_FS);
  58                 spin_unlock(&sbi->s_lock);
  59                 ext2_sync_super(sb, es, 1);
  60         }
  61
  62         va_start(args, fmt);
  63
  64         vaf.fmt = fmt;
  65         vaf.va = &args;
  66
  67         printk(KERN_CRIT "EXT2-fs (%s): error: %s: %pV\n",
  68                sb->s_id, function, &vaf);
  69
  70         va_end(args);
  71
  72         if (test_opt(sb, ERRORS_PANIC))
  73                 panic("EXT2-fs: panic from previous error\n");
  74         if (test_opt(sb, ERRORS_RO)) {
  75                 ext2_msg(sb, KERN_CRIT,
  76                              "error: remounting filesystem read-only");
  77                 sb->s_flags |= MS_RDONLY;
  78         }
  79 }
  80
  81 void ext2_msg(struct super_block *sb, const char *prefix,
  82                 const char *fmt, ...)
  83 {
  84         struct va_format vaf;
  85         va_list args;
  86
  87         va_start(args, fmt);
  88
  89         vaf.fmt = fmt;
  90         vaf.va = &args;
  91
  92         printk("%sEXT2-fs (%s): %pV\n", prefix, sb->s_id, &vaf);
  93
  94         va_end(args);
  95 }
  96
  97 /*
  98  * This must be called with sbi->s_lock held.
  99  */
 100 void ext2_update_dynamic_rev(struct super_block *sb)
 101 {
 102         struct ext2_super_block *es = EXT2_SB(sb)->s_es;
 103
 104         if (le32_to_cpu(es->s_rev_level) > EXT2_GOOD_OLD_REV)
 105                 return;
 106
 107         ext2_msg(sb, KERN_WARNING,
 108                      "warning: updating to rev %d because of "
 109                      "new feature flag, running e2fsck is recommended",
 110                      EXT2_DYNAMIC_REV);
 111
 112         es->s_first_ino = cpu_to_le32(EXT2_GOOD_OLD_FIRST_INO);
 113         es->s_inode_size = cpu_to_le16(EXT2_GOOD_OLD_INODE_SIZE);
 114         es->s_rev_level = cpu_to_le32(EXT2_DYNAMIC_REV);
 115         /* leave es->s_feature_*compat flags alone */
 116         /* es->s_uuid will be set by e2fsck if empty */
 117
 118         /*
 119          * The rest of the superblock fields should be zero, and if not it
 120          * means they are likely already in use, so leave them alone.  We
 121          * can leave it up to e2fsck to clean up any inconsistencies there.
 122          */
 123 }
 124
 125 static void ext2_put_super (struct super_block * sb)
 126 {
 127         int db_count;
 128         int i;
 129         struct ext2_sb_info *sbi = EXT2_SB(sb);
 130
 131         dquot_disable(sb, -1, DQUOT_USAGE_ENABLED | DQUOT_LIMITS_ENABLED);
 132
 133         if (sb->s_dirt)
 134                 ext2_write_super(sb);
 135
 136         ext2_xattr_put_super(sb);
 137         if (!(sb->s_flags & MS_RDONLY)) {
 138                 struct ext2_super_block *es = sbi->s_es;
 139
 140                 spin_lock(&sbi->s_lock);
 141                 es->s_state = cpu_to_le16(sbi->s_mount_state);
 142                 spin_unlock(&sbi->s_lock);
 143                 ext2_sync_super(sb, es, 1);
 144         }
 145         db_count = sbi->s_gdb_count;
 146         for (i = 0; i < db_count; i++)
 147                 if (sbi->s_group_desc[i])
 148                         brelse (sbi->s_group_desc[i]);
 149         kfree(sbi->s_group_desc);
 150         kfree(sbi->s_debts);
 151         percpu_counter_destroy(&sbi->s_freeblocks_counter);
 152         percpu_counter_destroy(&sbi->s_freeinodes_counter);
 153         percpu_counter_destroy(&sbi->s_dirs_counter);
 154         brelse (sbi->s_sbh);
 155         sb->s_fs_info = NULL;
 156         kfree(sbi->s_blockgroup_lock);
 157         kfree(sbi);
 158 }
 159
 160 static struct kmem_cache * ext2_inode_cachep;
 161
 162 static struct inode *ext2_alloc_inode(struct super_block *sb)
 163 {
 164         struct ext2_inode_info *ei;
 165         ei = (struct ext2_inode_info *)kmem_cache_alloc(ext2_inode_cachep, GFP_KERNEL);
 166         if (!ei)
 167                 return NULL;
 168         ei->i_block_alloc_info = NULL;
 169         ei->vfs_inode.i_version = 1;
 170         return &ei->vfs_inode;
 171 }
 172
 173 static void ext2_destroy_inode(struct inode *inode)
 174 {
 175         kmem_cache_free(ext2_inode_cachep, EXT2_I(inode));
 176 }
 177
 178 static void init_once(void *foo)
 179 {
 180         struct ext2_inode_info *ei = (struct ext2_inode_info *) foo;
 181
 182         rwlock_init(&ei->i_meta_lock);
 183 #ifdef CONFIG_EXT2_FS_XATTR
 184         init_rwsem(&ei->xattr_sem);
 185 #endif
 186         mutex_init(&ei->truncate_mutex);
 187         inode_init_once(&ei->vfs_inode);
 188 }
 189
 190 static int init_inodecache(void)
 191 {
 192         ext2_inode_cachep = kmem_cache_create("ext2_inode_cache",
 193                                              sizeof(struct ext2_inode_info),
 194                                              0, (SLAB_RECLAIM_ACCOUNT|
 195                                                 SLAB_MEM_SPREAD),
 196                                              init_once);
 197         if (ext2_inode_cachep == NULL)
 198                 return -ENOMEM;
 199         return 0;
 200 }
 201
 202 static void destroy_inodecache(void)
 203 {
 204         kmem_cache_destroy(ext2_inode_cachep);
 205 }
 206
 207 static int ext2_show_options(struct seq_file *seq, struct vfsmount *vfs)
 208 {
 209         struct super_block *sb = vfs->mnt_sb;
 210         struct ext2_sb_info *sbi = EXT2_SB(sb);
 211         struct ext2_super_block *es = sbi->s_es;
 212         unsigned long def_mount_opts;
 213
 214         spin_lock(&sbi->s_lock);
 215         def_mount_opts = le32_to_cpu(es->s_default_mount_opts);
 216
 217         if (sbi->s_sb_block != 1)
 218                 seq_printf(seq, ",sb=%lu", sbi->s_sb_block);
 219         if (test_opt(sb, MINIX_DF))
 220                 seq_puts(seq, ",minixdf");
 221         if (test_opt(sb, GRPID))
 222                 seq_puts(seq, ",grpid");
 223         if (!test_opt(sb, GRPID) && (def_mount_opts & EXT2_DEFM_BSDGROUPS))
 224                 seq_puts(seq, ",nogrpid");
 225         if (sbi->s_resuid != EXT2_DEF_RESUID ||
 226             le16_to_cpu(es->s_def_resuid) != EXT2_DEF_RESUID) {
 227                 seq_printf(seq, ",resuid=%u", sbi->s_resuid);
 228         }
 229         if (sbi->s_resgid != EXT2_DEF_RESGID ||
 230             le16_to_cpu(es->s_def_resgid) != EXT2_DEF_RESGID) {
 231                 seq_printf(seq, ",resgid=%u", sbi->s_resgid);
 232         }
 233         if (test_opt(sb, ERRORS_RO)) {
 234                 int def_errors = le16_to_cpu(es->s_errors);
 235
 236                 if (def_errors == EXT2_ERRORS_PANIC ||
 237                     def_errors == EXT2_ERRORS_CONTINUE) {
 238                         seq_puts(seq, ",errors=remount-ro");
 239                 }
 240         }
 241         if (test_opt(sb, ERRORS_CONT))
 242                 seq_puts(seq, ",errors=continue");
 243         if (test_opt(sb, ERRORS_PANIC))
 244                 seq_puts(seq, ",errors=panic");
 245         if (test_opt(sb, NO_UID32))
 246                 seq_puts(seq, ",nouid32");
 247         if (test_opt(sb, DEBUG))
 248                 seq_puts(seq, ",debug");
 249         if (test_opt(sb, OLDALLOC))
 250                 seq_puts(seq, ",oldalloc");
 251
 252 #ifdef CONFIG_EXT2_FS_XATTR
 253         if (test_opt(sb, XATTR_USER))
 254                 seq_puts(seq, ",user_xattr");
 255         if (!test_opt(sb, XATTR_USER) &&
 256             (def_mount_opts & EXT2_DEFM_XATTR_USER)) {
 257                 seq_puts(seq, ",nouser_xattr");
 258         }
 259 #endif
 260
 261 #ifdef CONFIG_EXT2_FS_POSIX_ACL
 262         if (test_opt(sb, POSIX_ACL))
 263                 seq_puts(seq, ",acl");
 264         if (!test_opt(sb, POSIX_ACL) && (def_mount_opts & EXT2_DEFM_ACL))
 265                 seq_puts(seq, ",noacl");
 266 #endif
 267
 268         if (test_opt(sb, NOBH))
 269                 seq_puts(seq, ",nobh");
 270
 271 #if defined(CONFIG_QUOTA)
 272         if (sbi->s_mount_opt & EXT2_MOUNT_USRQUOTA)
 273                 seq_puts(seq, ",usrquota");
 274
 275         if (sbi->s_mount_opt & EXT2_MOUNT_GRPQUOTA)
 276                 seq_puts(seq, ",grpquota");
 277 #endif
 278
 279 #if defined(CONFIG_EXT2_FS_XIP)
 280         if (sbi->s_mount_opt & EXT2_MOUNT_XIP)
 281                 seq_puts(seq, ",xip");
 282 #endif
 283
 284         if (!test_opt(sb, RESERVATION))
 285                 seq_puts(seq, ",noreservation");
 286
 287         spin_unlock(&sbi->s_lock);
 288         return 0;
 289 }
 290
 291 #ifdef CONFIG_QUOTA
 292 static ssize_t ext2_quota_read(struct super_block *sb, int type, char *data, size_t len, loff_t off);
 293 static ssize_t ext2_quota_write(struct super_block *sb, int type, const char *data, size_t len, loff_t off);
 294 #endif
 295
 296 static const struct super_operations ext2_sops = {
 297         .alloc_inode    = ext2_alloc_inode,
 298         .destroy_inode  = ext2_destroy_inode,
 299         .write_inode    = ext2_write_inode,
 300         .evict_inode    = ext2_evict_inode,
 301         .put_super      = ext2_put_super,
 302         .write_super    = ext2_write_super,
 303         .sync_fs        = ext2_sync_fs,
 304         .statfs         = ext2_statfs,
 305         .remount_fs     = ext2_remount,
 306         .show_options   = ext2_show_options,
 307 #ifdef CONFIG_QUOTA
 308         .quota_read     = ext2_quota_read,
 309         .quota_write    = ext2_quota_write,
 310 #endif
 311 };
 312
 313 static struct inode *ext2_nfs_get_inode(struct super_block *sb,
 314                 u64 ino, u32 generation)
 315 {
 316         struct inode *inode;
 317
 318         if (ino < EXT2_FIRST_INO(sb) && ino != EXT2_ROOT_INO)
 319                 return ERR_PTR(-ESTALE);
 320         if (ino > le32_to_cpu(EXT2_SB(sb)->s_es->s_inodes_count))
 321                 return ERR_PTR(-ESTALE);
 322
 323         /* iget isn't really right if the inode is currently unallocated!!
 324          * ext2_read_inode currently does appropriate checks, but
 325          * it might be "neater" to call ext2_get_inode first and check
 326          * if the inode is valid.....
 327          */
 328         inode = ext2_iget(sb, ino);
 329         if (IS_ERR(inode))
 330                 return ERR_CAST(inode);
 331         if (generation && inode->i_generation != generation) {
 332                 /* we didn't find the right inode.. */
 333                 iput(inode);
 334                 return ERR_PTR(-ESTALE);
 335         }
 336         return inode;
 337 }
 338
 339 static struct dentry *ext2_fh_to_dentry(struct super_block *sb, struct fid *fid,
 340                 int fh_len, int fh_type)
 341 {
 342         return generic_fh_to_dentry(sb, fid, fh_len, fh_type,
 343                                     ext2_nfs_get_inode);
 344 }
 345
 346 static struct dentry *ext2_fh_to_parent(struct super_block *sb, struct fid *fid,
 347                 int fh_len, int fh_type)
 348 {
 349         return generic_fh_to_parent(sb, fid, fh_len, fh_type,
 350                                     ext2_nfs_get_inode);
 351 }
 352
 353 /* Yes, most of these are left as NULL!!
 354  * A NULL value implies the default, which works with ext2-like file
 355  * systems, but can be improved upon.
 356  * Currently only get_parent is required.
 357  */
 358 static const struct export_operations ext2_export_ops = {
 359         .fh_to_dentry = ext2_fh_to_dentry,
 360         .fh_to_parent = ext2_fh_to_parent,
 361         .get_parent = ext2_get_parent,
 362 };
 363
 364 static unsigned long get_sb_block(void **data)
 365 {
 366         unsigned long   sb_block;
 367         char            *options = (char *) *data;
 368
 369         if (!options || strncmp(options, "sb=", 3) != 0)
 370                 return 1;       /* Default location */
 371         options += 3;
 372         sb_block = simple_strtoul(options, &options, 0);
 373         if (*options && *options != ',') {
 374                 printk("EXT2-fs: Invalid sb specification: %s\n",
 375                        (char *) *data);
 376                 return 1;
 377         }
 378         if (*options == ',')
 379                 options++;
 380         *data = (void *) options;
 381         return sb_block;
 382 }
 383
 384 enum {
 385         Opt_bsd_df, Opt_minix_df, Opt_grpid, Opt_nogrpid,
 386         Opt_resgid, Opt_resuid, Opt_sb, Opt_err_cont, Opt_err_panic,
 387         Opt_err_ro, Opt_nouid32, Opt_nocheck, Opt_debug,
 388         Opt_oldalloc, Opt_orlov, Opt_nobh, Opt_user_xattr, Opt_nouser_xattr,
 389         Opt_acl, Opt_noacl, Opt_xip, Opt_ignore, Opt_err, Opt_quota,
 390         Opt_usrquota, Opt_grpquota, Opt_reservation, Opt_noreservation
 391 };
 392
 393 static const match_table_t tokens = {
 394         {Opt_bsd_df, "bsddf"},
 395         {Opt_minix_df, "minixdf"},
 396         {Opt_grpid, "grpid"},
 397         {Opt_grpid, "bsdgroups"},
 398         {Opt_nogrpid, "nogrpid"},
 399         {Opt_nogrpid, "sysvgroups"},
 400         {Opt_resgid, "resgid=%u"},
 401         {Opt_resuid, "resuid=%u"},
 402         {Opt_sb, "sb=%u"},
 403         {Opt_err_cont, "errors=continue"},
 404         {Opt_err_panic, "errors=panic"},
 405         {Opt_err_ro, "errors=remount-ro"},
 406         {Opt_nouid32, "nouid32"},
 407         {Opt_nocheck, "check=none"},
 408         {Opt_nocheck, "nocheck"},
 409         {Opt_debug, "debug"},
 410         {Opt_oldalloc, "oldalloc"},
 411         {Opt_orlov, "orlov"},
 412         {Opt_nobh, "nobh"},
 413         {Opt_user_xattr, "user_xattr"},
 414         {Opt_nouser_xattr, "nouser_xattr"},
 415         {Opt_acl, "acl"},
 416         {Opt_noacl, "noacl"},
 417         {Opt_xip, "xip"},
 418         {Opt_grpquota, "grpquota"},
 419         {Opt_ignore, "noquota"},
 420         {Opt_quota, "quota"},
 421         {Opt_usrquota, "usrquota"},
 422         {Opt_reservation, "reservation"},
 423         {Opt_noreservation, "noreservation"},
 424         {Opt_err, NULL}
 425 };
 426
 427 static int parse_options(char *options, struct super_block *sb)
 428 {
 429         char *p;
 430         struct ext2_sb_info *sbi = EXT2_SB(sb);
 431         substring_t args[MAX_OPT_ARGS];
 432         int option;
 433
 434         if (!options)
 435                 return 1;
 436
 437         while ((p = strsep (&options, ",")) != NULL) {
 438                 int token;
 439                 if (!*p)
 440                         continue;
 441
 442                 token = match_token(p, tokens, args);
 443                 switch (token) {
 444                 case Opt_bsd_df:
 445                         clear_opt (sbi->s_mount_opt, MINIX_DF);
 446                         break;
 447                 case Opt_minix_df:
 448                         set_opt (sbi->s_mount_opt, MINIX_DF);
 449                         break;
 450                 case Opt_grpid:
 451                         set_opt (sbi->s_mount_opt, GRPID);
 452                         break;
 453                 case Opt_nogrpid:
 454                         clear_opt (sbi->s_mount_opt, GRPID);
 455                         break;
 456                 case Opt_resuid:
 457                         if (match_int(&args[0], &option))
 458                                 return 0;
 459                         sbi->s_resuid = option;
 460                         break;
 461                 case Opt_resgid:
 462                         if (match_int(&args[0], &option))
 463                                 return 0;
 464                         sbi->s_resgid = option;
 465                         break;
 466                 case Opt_sb:
 467                         /* handled by get_sb_block() instead of here */
 468                         /* *sb_block = match_int(&args[0]); */
 469                         break;
 470                 case Opt_err_panic:
 471                         clear_opt (sbi->s_mount_opt, ERRORS_CONT);
 472                         clear_opt (sbi->s_mount_opt, ERRORS_RO);
 473                         set_opt (sbi->s_mount_opt, ERRORS_PANIC);
 474                         break;
 475                 case Opt_err_ro:
 476                         clear_opt (sbi->s_mount_opt, ERRORS_CONT);
 477                         clear_opt (sbi->s_mount_opt, ERRORS_PANIC);
 478                         set_opt (sbi->s_mount_opt, ERRORS_RO);
 479                         break;
 480                 case Opt_err_cont:
 481                         clear_opt (sbi->s_mount_opt, ERRORS_RO);
 482                         clear_opt (sbi->s_mount_opt, ERRORS_PANIC);
 483                         set_opt (sbi->s_mount_opt, ERRORS_CONT);
 484                         break;
 485                 case Opt_nouid32:
 486                         set_opt (sbi->s_mount_opt, NO_UID32);
 487                         break;
 488                 case Opt_nocheck:
 489                         clear_opt (sbi->s_mount_opt, CHECK);
 490                         break;
 491                 case Opt_debug:
 492                         set_opt (sbi->s_mount_opt, DEBUG);
 493                         break;
 494                 case Opt_oldalloc:
 495                         set_opt (sbi->s_mount_opt, OLDALLOC);
 496                         break;
 497                 case Opt_orlov:
 498                         clear_opt (sbi->s_mount_opt, OLDALLOC);
 499                         break;
 500                 case Opt_nobh:
 501                         set_opt (sbi->s_mount_opt, NOBH);
 502                         break;
 503 #ifdef CONFIG_EXT2_FS_XATTR
 504                 case Opt_user_xattr:
 505                         set_opt (sbi->s_mount_opt, XATTR_USER);
 506                         break;
 507                 case Opt_nouser_xattr:
 508                         clear_opt (sbi->s_mount_opt, XATTR_USER);
 509                         break;
 510 #else
 511                 case Opt_user_xattr:
 512                 case Opt_nouser_xattr:
 513                         ext2_msg(sb, KERN_INFO, "(no)user_xattr options"
 514                                 "not supported");
 515                         break;
 516 #endif
 517 #ifdef CONFIG_EXT2_FS_POSIX_ACL
 518                 case Opt_acl:
 519                         set_opt(sbi->s_mount_opt, POSIX_ACL);
 520                         break;
 521                 case Opt_noacl:
 522                         clear_opt(sbi->s_mount_opt, POSIX_ACL);
 523                         break;
 524 #else
 525                 case Opt_acl:
 526                 case Opt_noacl:
 527                         ext2_msg(sb, KERN_INFO,
 528                                 "(no)acl options not supported");
 529                         break;
 530 #endif
 531                 case Opt_xip:
 532 #ifdef CONFIG_EXT2_FS_XIP
 533                         set_opt (sbi->s_mount_opt, XIP);
 534 #else
 535                         ext2_msg(sb, KERN_INFO, "xip option not supported");
 536 #endif
 537                         break;
 538
 539 #if defined(CONFIG_QUOTA)
 540                 case Opt_quota:
 541                 case Opt_usrquota:
 542                         set_opt(sbi->s_mount_opt, USRQUOTA);
 543                         break;
 544
 545                 case Opt_grpquota:
 546                         set_opt(sbi->s_mount_opt, GRPQUOTA);
 547                         break;
 548 #else
 549                 case Opt_quota:
 550                 case Opt_usrquota:
 551                 case Opt_grpquota:
 552                         ext2_msg(sb, KERN_INFO,
 553                                 "quota operations not supported");
 554                         break;
 555 #endif
 556
 557                 case Opt_reservation:
 558                         set_opt(sbi->s_mount_opt, RESERVATION);
 559                         ext2_msg(sb, KERN_INFO, "reservations ON");
 560                         break;
 561                 case Opt_noreservation:
 562                         clear_opt(sbi->s_mount_opt, RESERVATION);
 563                         ext2_msg(sb, KERN_INFO, "reservations OFF");
 564                         break;
 565                 case Opt_ignore:
 566                         break;
 567                 default:
 568                         return 0;
 569                 }
 570         }
 571         return 1;
 572 }
 573
 574 static int ext2_setup_super (struct super_block * sb,
 575                               struct ext2_super_block * es,
 576                               int read_only)
 577 {
 578         int res = 0;
 579         struct ext2_sb_info *sbi = EXT2_SB(sb);
 580
 581         if (le32_to_cpu(es->s_rev_level) > EXT2_MAX_SUPP_REV) {
 582                 ext2_msg(sb, KERN_ERR,
 583                         "error: revision level too high, "
 584                         "forcing read-only mode");
 585                 res = MS_RDONLY;
 586         }
 587         if (read_only)
 588                 return res;
 589         if (!(sbi->s_mount_state & EXT2_VALID_FS))
 590                 ext2_msg(sb, KERN_WARNING,
 591                         "warning: mounting unchecked fs, "
 592                         "running e2fsck is recommended");
 593         else if ((sbi->s_mount_state & EXT2_ERROR_FS))
 594                 ext2_msg(sb, KERN_WARNING,
 595                         "warning: mounting fs with errors, "
 596                         "running e2fsck is recommended");
 597         else if ((__s16) le16_to_cpu(es->s_max_mnt_count) >= 0 &&
 598                  le16_to_cpu(es->s_mnt_count) >=
 599                  (unsigned short) (__s16) le16_to_cpu(es->s_max_mnt_count))
 600                 ext2_msg(sb, KERN_WARNING,
 601                         "warning: maximal mount count reached, "
 602                         "running e2fsck is recommended");
 603         else if (le32_to_cpu(es->s_checkinterval) &&
 604                 (le32_to_cpu(es->s_lastcheck) +
 605                         le32_to_cpu(es->s_checkinterval) <= get_seconds()))
 606                 ext2_msg(sb, KERN_WARNING,
 607                         "warning: checktime reached, "
 608                         "running e2fsck is recommended");
 609         if (!le16_to_cpu(es->s_max_mnt_count))
 610                 es->s_max_mnt_count = cpu_to_le16(EXT2_DFL_MAX_MNT_COUNT);
 611         le16_add_cpu(&es->s_mnt_count, 1);
 612         if (test_opt (sb, DEBUG))
 613                 ext2_msg(sb, KERN_INFO, "%s, %s, bs=%lu, fs=%lu, gc=%lu, "
 614                         "bpg=%lu, ipg=%lu, mo=%04lx]",
 615                         EXT2FS_VERSION, EXT2FS_DATE, sb->s_blocksize,
 616                         sbi->s_frag_size,
 617                         sbi->s_groups_count,
 618                         EXT2_BLOCKS_PER_GROUP(sb),
 619                         EXT2_INODES_PER_GROUP(sb),
 620                         sbi->s_mount_opt);
 621         return res;
 622 }
 623
 624 static int ext2_check_descriptors(struct super_block *sb)
 625 {
 626         int i;
 627         struct ext2_sb_info *sbi = EXT2_SB(sb);
 628
 629         ext2_debug ("Checking group descriptors");
 630
 631         for (i = 0; i < sbi->s_groups_count; i++) {
 632                 struct ext2_group_desc *gdp = ext2_get_group_desc(sb, i, NULL);
 633                 ext2_fsblk_t first_block = ext2_group_first_block_no(sb, i);
 634                 ext2_fsblk_t last_block;
 635
 636                 if (i == sbi->s_groups_count - 1)
 637                         last_block = le32_to_cpu(sbi->s_es->s_blocks_count) - 1;
 638                 else
 639                         last_block = first_block +
 640                                 (EXT2_BLOCKS_PER_GROUP(sb) - 1);
 641
 642                 if (le32_to_cpu(gdp->bg_block_bitmap) < first_block ||
 643                     le32_to_cpu(gdp->bg_block_bitmap) > last_block)
 644                 {
 645                         ext2_error (sb, "ext2_check_descriptors",
 646                                     "Block bitmap for group %d"
 647                                     " not in group (block %lu)!",
 648                                     i, (unsigned long) le32_to_cpu(gdp->bg_block_bitmap));
 649                         return 0;
 650                 }
 651                 if (le32_to_cpu(gdp->bg_inode_bitmap) < first_block ||
 652                     le32_to_cpu(gdp->bg_inode_bitmap) > last_block)
 653                 {
 654                         ext2_error (sb, "ext2_check_descriptors",
 655                                     "Inode bitmap for group %d"
 656                                     " not in group (block %lu)!",
 657                                     i, (unsigned long) le32_to_cpu(gdp->bg_inode_bitmap));
 658                         return 0;
 659                 }
 660                 if (le32_to_cpu(gdp->bg_inode_table) < first_block ||
 661                     le32_to_cpu(gdp->bg_inode_table) + sbi->s_itb_per_group - 1 >
 662                     last_block)
 663                 {
 664                         ext2_error (sb, "ext2_check_descriptors",
 665                                     "Inode table for group %d"
 666                                     " not in group (block %lu)!",
 667                                     i, (unsigned long) le32_to_cpu(gdp->bg_inode_table));
 668                         return 0;
 669                 }
 670         }
 671         return 1;
 672 }
 673
 674 /*
 675  * Maximal file size.  There is a direct, and {,double-,triple-}indirect
 676  * block limit, and also a limit of (2^32 - 1) 512-byte sectors in i_blocks.
 677  * We need to be 1 filesystem block less than the 2^32 sector limit.
 678  */
 679 static loff_t ext2_max_size(int bits)
 680 {
 681         loff_t res = EXT2_NDIR_BLOCKS;
 682         int meta_blocks;
 683         loff_t upper_limit;
 684
 685         /* This is calculated to be the largest file size for a
 686          * dense, file such that the total number of
 687          * sectors in the file, including data and all indirect blocks,
 688          * does not exceed 2^32 -1
 689          * __u32 i_blocks representing the total number of
 690          * 512 bytes blocks of the file
 691          */
 692         upper_limit = (1LL << 32) - 1;
 693
 694         /* total blocks in file system block size */
 695         upper_limit >>= (bits - 9);
 696
 697
 698         /* indirect blocks */
 699         meta_blocks = 1;
 700         /* double indirect blocks */
 701         meta_blocks += 1 + (1LL << (bits-2));
 702         /* tripple indirect blocks */
 703         meta_blocks += 1 + (1LL << (bits-2)) + (1LL << (2*(bits-2)));
 704
 705         upper_limit -= meta_blocks;
 706         upper_limit <<= bits;
 707
 708         res += 1LL << (bits-2);
 709         res += 1LL << (2*(bits-2));
 710         res += 1LL << (3*(bits-2));
 711         res <<= bits;
 712         if (res > upper_limit)
 713                 res = upper_limit;
 714
 715         if (res > MAX_LFS_FILESIZE)
 716                 res = MAX_LFS_FILESIZE;
 717
 718         return res;
 719 }
 720
 721 static unsigned long descriptor_loc(struct super_block *sb,
 722                                     unsigned long logic_sb_block,
 723                                     int nr)
 724 {
 725         struct ext2_sb_info *sbi = EXT2_SB(sb);
 726         unsigned long bg, first_meta_bg;
 727         int has_super = 0;
 728
 729         first_meta_bg = le32_to_cpu(sbi->s_es->s_first_meta_bg);
 730
 731         if (!EXT2_HAS_INCOMPAT_FEATURE(sb, EXT2_FEATURE_INCOMPAT_META_BG) ||
 732             nr < first_meta_bg)
 733                 return (logic_sb_block + nr + 1);
 734         bg = sbi->s_desc_per_block * nr;
 735         if (ext2_bg_has_super(sb, bg))
 736                 has_super = 1;
 737
 738         return ext2_group_first_block_no(sb, bg) + has_super;
 739 }
 740
 741 static int ext2_fill_super(struct super_block *sb, void *data, int silent)
 742 {
 743         struct buffer_head * bh;
 744         struct ext2_sb_info * sbi;
 745         struct ext2_super_block * es;
 746         struct inode *root;
 747         unsigned long block;
 748         unsigned long sb_block = get_sb_block(&data);
 749         unsigned long logic_sb_block;
 750         unsigned long offset = 0;
 751         unsigned long def_mount_opts;
 752         long ret = -EINVAL;
 753         int blocksize = BLOCK_SIZE;
 754         int db_count;
 755         int i, j;
 756         __le32 features;
 757         int err;
 758
 759         err = -ENOMEM;
 760         sbi = kzalloc(sizeof(*sbi), GFP_KERNEL);
 761         if (!sbi)
 762                 goto failed_unlock;
 763
 764         sbi->s_blockgroup_lock =
 765                 kzalloc(sizeof(struct blockgroup_lock), GFP_KERNEL);
 766         if (!sbi->s_blockgroup_lock) {
 767                 kfree(sbi);
 768                 goto failed_unlock;
 769         }
 770         sb->s_fs_info = sbi;
 771         sbi->s_sb_block = sb_block;
 772
 773         spin_lock_init(&sbi->s_lock);
 774
 775         /*
 776          * See what the current blocksize for the device is, and
 777          * use that as the blocksize.  Otherwise (or if the blocksize
 778          * is smaller than the default) use the default.
 779          * This is important for devices that have a hardware
 780          * sectorsize that is larger than the default.
 781          */
 782         blocksize = sb_min_blocksize(sb, BLOCK_SIZE);
 783         if (!blocksize) {
 784                 ext2_msg(sb, KERN_ERR, "error: unable to set blocksize");
 785                 goto failed_sbi;
 786         }
 787
 788         /*
 789          * If the superblock doesn't start on a hardware sector boundary,
 790          * calculate the offset.
 791          */
 792         if (blocksize != BLOCK_SIZE) {
 793                 logic_sb_block = (sb_block*BLOCK_SIZE) / blocksize;
 794                 offset = (sb_block*BLOCK_SIZE) % blocksize;
 795         } else {
 796                 logic_sb_block = sb_block;
 797         }
 798
 799         if (!(bh = sb_bread(sb, logic_sb_block))) {
 800                 ext2_msg(sb, KERN_ERR, "error: unable to read superblock");
 801                 goto failed_sbi;
 802         }
 803         /*
 804          * Note: s_es must be initialized as soon as possible because
 805          *       some ext2 macro-instructions depend on its value
 806          */
 807         es = (struct ext2_super_block *) (((char *)bh->b_data) + offset);
 808         sbi->s_es = es;
 809         sb->s_magic = le16_to_cpu(es->s_magic);
 810
 811         if (sb->s_magic != EXT2_SUPER_MAGIC)
 812                 goto cantfind_ext2;
 813
 814         /* Set defaults before we parse the mount options */
 815         def_mount_opts = le32_to_cpu(es->s_default_mount_opts);
 816         if (def_mount_opts & EXT2_DEFM_DEBUG)
 817                 set_opt(sbi->s_mount_opt, DEBUG);
 818         if (def_mount_opts & EXT2_DEFM_BSDGROUPS)
 819                 set_opt(sbi->s_mount_opt, GRPID);
 820         if (def_mount_opts & EXT2_DEFM_UID16)
 821                 set_opt(sbi->s_mount_opt, NO_UID32);
 822 #ifdef CONFIG_EXT2_FS_XATTR
 823         if (def_mount_opts & EXT2_DEFM_XATTR_USER)
 824                 set_opt(sbi->s_mount_opt, XATTR_USER);
 825 #endif
 826 #ifdef CONFIG_EXT2_FS_POSIX_ACL
 827         if (def_mount_opts & EXT2_DEFM_ACL)
 828                 set_opt(sbi->s_mount_opt, POSIX_ACL);
 829 #endif
 830
 831         if (le16_to_cpu(sbi->s_es->s_errors) == EXT2_ERRORS_PANIC)
 832                 set_opt(sbi->s_mount_opt, ERRORS_PANIC);
 833         else if (le16_to_cpu(sbi->s_es->s_errors) == EXT2_ERRORS_CONTINUE)
 834                 set_opt(sbi->s_mount_opt, ERRORS_CONT);
 835         else
 836                 set_opt(sbi->s_mount_opt, ERRORS_RO);
 837
 838         sbi->s_resuid = le16_to_cpu(es->s_def_resuid);
 839         sbi->s_resgid = le16_to_cpu(es->s_def_resgid);
 840
 841         set_opt(sbi->s_mount_opt, RESERVATION);
 842
 843         if (!parse_options((char *) data, sb))
 844                 goto failed_mount;
 845
 846         sb->s_flags = (sb->s_flags & ~MS_POSIXACL) |
 847                 ((EXT2_SB(sb)->s_mount_opt & EXT2_MOUNT_POSIX_ACL) ?
 848                  MS_POSIXACL : 0);
 849
 850         ext2_xip_verify_sb(sb); /* see if bdev supports xip, unset
 851                                     EXT2_MOUNT_XIP if not */
 852
 853         if (le32_to_cpu(es->s_rev_level) == EXT2_GOOD_OLD_REV &&
 854             (EXT2_HAS_COMPAT_FEATURE(sb, ~0U) ||
 855              EXT2_HAS_RO_COMPAT_FEATURE(sb, ~0U) ||
 856              EXT2_HAS_INCOMPAT_FEATURE(sb, ~0U)))
 857                 ext2_msg(sb, KERN_WARNING,
 858                         "warning: feature flags set on rev 0 fs, "
 859                         "running e2fsck is recommended");
 860         /*
 861          * Check feature flags regardless of the revision level, since we
 862          * previously didn't change the revision level when setting the flags,
 863          * so there is a chance incompat flags are set on a rev 0 filesystem.
 864          */
 865         features = EXT2_HAS_INCOMPAT_FEATURE(sb, ~EXT2_FEATURE_INCOMPAT_SUPP);
 866         if (features) {
 867                 ext2_msg(sb, KERN_ERR,  "error: couldn't mount because of "
 868                        "unsupported optional features (%x)",
 869                         le32_to_cpu(features));
 870                 goto failed_mount;
 871         }
 872         if (!(sb->s_flags & MS_RDONLY) &&
 873             (features = EXT2_HAS_RO_COMPAT_FEATURE(sb, ~EXT2_FEATURE_RO_COMPAT_SUPP))){
 874                 ext2_msg(sb, KERN_ERR, "error: couldn't mount RDWR because of "
 875                        "unsupported optional features (%x)",
 876                        le32_to_cpu(features));
 877                 goto failed_mount;
 878         }
 879
 880         blocksize = BLOCK_SIZE << le32_to_cpu(sbi->s_es->s_log_block_size);
 881
 882         if (ext2_use_xip(sb) && blocksize != PAGE_SIZE) {
 883                 if (!silent)
 884                         ext2_msg(sb, KERN_ERR,
 885                                 "error: unsupported blocksize for xip");
 886                 goto failed_mount;
 887         }
 888
 889         /* If the blocksize doesn't match, re-read the thing.. */
 890         if (sb->s_blocksize != blocksize) {
 891                 brelse(bh);
 892
 893                 if (!sb_set_blocksize(sb, blocksize)) {
 894                         ext2_msg(sb, KERN_ERR, "error: blocksize is too small");
 895                         goto failed_sbi;
 896                 }
 897
 898                 logic_sb_block = (sb_block*BLOCK_SIZE) / blocksize;
 899                 offset = (sb_block*BLOCK_SIZE) % blocksize;
 900                 bh = sb_bread(sb, logic_sb_block);
 901                 if(!bh) {
 902                         ext2_msg(sb, KERN_ERR, "error: couldn't read"
 903                                 "superblock on 2nd try");
 904                         goto failed_sbi;
 905                 }
 906                 es = (struct ext2_super_block *) (((char *)bh->b_data) + offset);
 907                 sbi->s_es = es;
 908                 if (es->s_magic != cpu_to_le16(EXT2_SUPER_MAGIC)) {
 909                         ext2_msg(sb, KERN_ERR, "error: magic mismatch");
 910                         goto failed_mount;
 911                 }
 912         }
 913
 914         sb->s_maxbytes = ext2_max_size(sb->s_blocksize_bits);
 915
 916         if (le32_to_cpu(es->s_rev_level) == EXT2_GOOD_OLD_REV) {
 917                 sbi->s_inode_size = EXT2_GOOD_OLD_INODE_SIZE;
 918                 sbi->s_first_ino = EXT2_GOOD_OLD_FIRST_INO;
 919         } else {
 920                 sbi->s_inode_size = le16_to_cpu(es->s_inode_size);
 921                 sbi->s_first_ino = le32_to_cpu(es->s_first_ino);
 922                 if ((sbi->s_inode_size < EXT2_GOOD_OLD_INODE_SIZE) ||
 923                     !is_power_of_2(sbi->s_inode_size) ||
 924                     (sbi->s_inode_size > blocksize)) {
 925                         ext2_msg(sb, KERN_ERR,
 926                                 "error: unsupported inode size: %d",
 927                                 sbi->s_inode_size);
 928                         goto failed_mount;
 929                 }
 930         }
 931
 932         sbi->s_frag_size = EXT2_MIN_FRAG_SIZE <<
 933                                    le32_to_cpu(es->s_log_frag_size);
 934         if (sbi->s_frag_size == 0)
 935                 goto cantfind_ext2;
 936         sbi->s_frags_per_block = sb->s_blocksize / sbi->s_frag_size;
 937
 938         sbi->s_blocks_per_group = le32_to_cpu(es->s_blocks_per_group);
 939         sbi->s_frags_per_group = le32_to_cpu(es->s_frags_per_group);
 940         sbi->s_inodes_per_group = le32_to_cpu(es->s_inodes_per_group);
 941
 942         if (EXT2_INODE_SIZE(sb) == 0)
 943                 goto cantfind_ext2;
 944         sbi->s_inodes_per_block = sb->s_blocksize / EXT2_INODE_SIZE(sb);
 945         if (sbi->s_inodes_per_block == 0 || sbi->s_inodes_per_group == 0)
 946                 goto cantfind_ext2;
 947         sbi->s_itb_per_group = sbi->s_inodes_per_group /
 948                                         sbi->s_inodes_per_block;
 949         sbi->s_desc_per_block = sb->s_blocksize /
 950                                         sizeof (struct ext2_group_desc);
 951         sbi->s_sbh = bh;
 952         sbi->s_mount_state = le16_to_cpu(es->s_state);
 953         sbi->s_addr_per_block_bits =
 954                 ilog2 (EXT2_ADDR_PER_BLOCK(sb));
 955         sbi->s_desc_per_block_bits =
 956                 ilog2 (EXT2_DESC_PER_BLOCK(sb));
 957
 958         if (sb->s_magic != EXT2_SUPER_MAGIC)
 959                 goto cantfind_ext2;
 960
 961         if (sb->s_blocksize != bh->b_size) {
 962                 if (!silent)
 963                         ext2_msg(sb, KERN_ERR, "error: unsupported blocksize");
 964                 goto failed_mount;
 965         }
 966
 967         if (sb->s_blocksize != sbi->s_frag_size) {
 968                 ext2_msg(sb, KERN_ERR,
 969                         "error: fragsize %lu != blocksize %lu"
 970                         "(not supported yet)",
 971                         sbi->s_frag_size, sb->s_blocksize);
 972                 goto failed_mount;
 973         }
 974
 975         if (sbi->s_blocks_per_group > sb->s_blocksize * 8) {
 976                 ext2_msg(sb, KERN_ERR,
 977                         "error: #blocks per group too big: %lu",
 978                         sbi->s_blocks_per_group);
 979                 goto failed_mount;
 980         }
 981         if (sbi->s_frags_per_group > sb->s_blocksize * 8) {
 982                 ext2_msg(sb, KERN_ERR,
 983                         "error: #fragments per group too big: %lu",
 984                         sbi->s_frags_per_group);
 985                 goto failed_mount;
 986         }
 987         if (sbi->s_inodes_per_group > sb->s_blocksize * 8) {
 988                 ext2_msg(sb, KERN_ERR,
 989                         "error: #inodes per group too big: %lu",
 990                         sbi->s_inodes_per_group);
 991                 goto failed_mount;
 992         }
 993
 994         if (EXT2_BLOCKS_PER_GROUP(sb) == 0)
 995                 goto cantfind_ext2;
 996         sbi->s_groups_count = ((le32_to_cpu(es->s_blocks_count) -
 997                                 le32_to_cpu(es->s_first_data_block) - 1)
 998                                         / EXT2_BLOCKS_PER_GROUP(sb)) + 1;
 999         db_count = (sbi->s_groups_count + EXT2_DESC_PER_BLOCK(sb) - 1) /
1000                    EXT2_DESC_PER_BLOCK(sb);
1001         sbi->s_group_desc = kmalloc (db_count * sizeof (struct buffer_head *), GFP_KERNEL);
1002         if (sbi->s_group_desc == NULL) {
1003                 ext2_msg(sb, KERN_ERR, "error: not enough memory");
1004                 goto failed_mount;
1005         }
1006         bgl_lock_init(sbi->s_blockgroup_lock);
1007         sbi->s_debts = kcalloc(sbi->s_groups_count, sizeof(*sbi->s_debts), GFP_KERNEL);
1008         if (!sbi->s_debts) {
1009                 ext2_msg(sb, KERN_ERR, "error: not enough memory");
1010                 goto failed_mount_group_desc;
1011         }
1012         for (i = 0; i < db_count; i++) {
1013                 block = descriptor_loc(sb, logic_sb_block, i);
1014                 sbi->s_group_desc[i] = sb_bread(sb, block);
1015                 if (!sbi->s_group_desc[i]) {
1016                         for (j = 0; j < i; j++)
1017                                 brelse (sbi->s_group_desc[j]);
1018                         ext2_msg(sb, KERN_ERR,
1019                                 "error: unable to read group descriptors");
1020                         goto failed_mount_group_desc;
1021                 }
1022         }
1023         if (!ext2_check_descriptors (sb)) {
1024                 ext2_msg(sb, KERN_ERR, "group descriptors corrupted");
1025                 goto failed_mount2;
1026         }
1027         sbi->s_gdb_count = db_count;
1028         get_random_bytes(&sbi->s_next_generation, sizeof(u32));
1029         spin_lock_init(&sbi->s_next_gen_lock);
1030
1031         /* per fileystem reservation list head & lock */
1032         spin_lock_init(&sbi->s_rsv_window_lock);
1033         sbi->s_rsv_window_root = RB_ROOT;
1034         /*
1035          * Add a single, static dummy reservation to the start of the
1036          * reservation window list --- it gives us a placeholder for
1037          * append-at-start-of-list which makes the allocation logic
1038          * _much_ simpler.
1039          */
1040         sbi->s_rsv_window_head.rsv_start = EXT2_RESERVE_WINDOW_NOT_ALLOCATED;
1041         sbi->s_rsv_window_head.rsv_end = EXT2_RESERVE_WINDOW_NOT_ALLOCATED;
1042         sbi->s_rsv_window_head.rsv_alloc_hit = 0;
1043         sbi->s_rsv_window_head.rsv_goal_size = 0;
1044         ext2_rsv_window_add(sb, &sbi->s_rsv_window_head);
1045
1046         err = percpu_counter_init(&sbi->s_freeblocks_counter,
1047                                 ext2_count_free_blocks(sb));
1048         if (!err) {
1049                 err = percpu_counter_init(&sbi->s_freeinodes_counter,
1050                                 ext2_count_free_inodes(sb));
1051         }
1052         if (!err) {
1053                 err = percpu_counter_init(&sbi->s_dirs_counter,
1054                                 ext2_count_dirs(sb));
1055         }
1056         if (err) {
1057                 ext2_msg(sb, KERN_ERR, "error: insufficient memory");
1058                 goto failed_mount3;
1059         }
1060         /*
1061          * set up enough so that it can read an inode
1062          */
1063         sb->s_op = &ext2_sops;
1064         sb->s_export_op = &ext2_export_ops;
1065         sb->s_xattr = ext2_xattr_handlers;
1066
1067 #ifdef CONFIG_QUOTA
1068         sb->dq_op = &dquot_operations;
1069         sb->s_qcop = &dquot_quotactl_ops;
1070 #endif
1071
1072         root = ext2_iget(sb, EXT2_ROOT_INO);
1073         if (IS_ERR(root)) {
1074                 ret = PTR_ERR(root);
1075                 goto failed_mount3;
1076         }
1077         if (!S_ISDIR(root->i_mode) || !root->i_blocks || !root->i_size) {
1078                 iput(root);
1079                 ext2_msg(sb, KERN_ERR, "error: corrupt root inode, run e2fsck");
1080                 goto failed_mount3;
1081         }
1082
1083         sb->s_root = d_alloc_root(root);
1084         if (!sb->s_root) {
1085                 iput(root);
1086                 ext2_msg(sb, KERN_ERR, "error: get root inode failed");
1087                 ret = -ENOMEM;
1088                 goto failed_mount3;
1089         }
1090         if (EXT2_HAS_COMPAT_FEATURE(sb, EXT3_FEATURE_COMPAT_HAS_JOURNAL))
1091                 ext2_msg(sb, KERN_WARNING,
1092                         "warning: mounting ext3 filesystem as ext2");
1093         if (ext2_setup_super (sb, es, sb->s_flags & MS_RDONLY))
1094                 sb->s_flags |= MS_RDONLY;
1095         ext2_write_super(sb);
1096         return 0;
1097
1098 cantfind_ext2:
1099         if (!silent)
1100                 ext2_msg(sb, KERN_ERR,
1101                         "error: can't find an ext2 filesystem on dev %s.",
1102                         sb->s_id);
1103         goto failed_mount;
1104 failed_mount3:
1105         percpu_counter_destroy(&sbi->s_freeblocks_counter);
1106         percpu_counter_destroy(&sbi->s_freeinodes_counter);
1107         percpu_counter_destroy(&sbi->s_dirs_counter);
1108 failed_mount2:
1109         for (i = 0; i < db_count; i++)
1110                 brelse(sbi->s_group_desc[i]);
1111 failed_mount_group_desc:
1112         kfree(sbi->s_group_desc);
1113         kfree(sbi->s_debts);
1114 failed_mount:
1115         brelse(bh);
1116 failed_sbi:
1117         sb->s_fs_info = NULL;
1118         kfree(sbi->s_blockgroup_lock);
1119         kfree(sbi);
1120 failed_unlock:
1121         return ret;
1122 }
1123
1124 static void ext2_clear_super_error(struct super_block *sb)
1125 {
1126         struct buffer_head *sbh = EXT2_SB(sb)->s_sbh;
1127
1128         if (buffer_write_io_error(sbh)) {
1129                 /*
1130                  * Oh, dear.  A previous attempt to write the
1131                  * superblock failed.  This could happen because the
1132                  * USB device was yanked out.  Or it could happen to
1133                  * be a transient write error and maybe the block will
1134                  * be remapped.  Nothing we can do but to retry the
1135                  * write and hope for the best.
1136                  */
1137                 ext2_msg(sb, KERN_ERR,
1138                        "previous I/O error to superblock detected\n");
1139                 clear_buffer_write_io_error(sbh);
1140                 set_buffer_uptodate(sbh);
1141         }
1142 }
1143
1144 static void ext2_sync_super(struct super_block *sb, struct ext2_super_block *es,
1145                             int wait)
1146 {
1147         ext2_clear_super_error(sb);
1148         spin_lock(&EXT2_SB(sb)->s_lock);
1149         es->s_free_blocks_count = cpu_to_le32(ext2_count_free_blocks(sb));
1150         es->s_free_inodes_count = cpu_to_le32(ext2_count_free_inodes(sb));
1151         es->s_wtime = cpu_to_le32(get_seconds());
1152         /* unlock before we do IO */
1153         spin_unlock(&EXT2_SB(sb)->s_lock);
1154         mark_buffer_dirty(EXT2_SB(sb)->s_sbh);
1155         if (wait)
1156                 sync_dirty_buffer(EXT2_SB(sb)->s_sbh);
1157         sb->s_dirt = 0;
1158 }
1159
1160 /*
1161  * In the second extended file system, it is not necessary to
1162  * write the super block since we use a mapping of the
1163  * disk super block in a buffer.
1164  *
1165  * However, this function is still used to set the fs valid
1166  * flags to 0.  We need to set this flag to 0 since the fs
1167  * may have been checked while mounted and e2fsck may have
1168  * set s_state to EXT2_VALID_FS after some corrections.
1169  */
1170 static int ext2_sync_fs(struct super_block *sb, int wait)
1171 {
1172         struct ext2_sb_info *sbi = EXT2_SB(sb);
1173         struct ext2_super_block *es = EXT2_SB(sb)->s_es;
1174
1175         spin_lock(&sbi->s_lock);
1176         if (es->s_state & cpu_to_le16(EXT2_VALID_FS)) {
1177                 ext2_debug("setting valid to 0\n");
1178                 es->s_state &= cpu_to_le16(~EXT2_VALID_FS);
1179         }
1180         spin_unlock(&sbi->s_lock);
1181         ext2_sync_super(sb, es, wait);
1182         return 0;
1183 }
1184
1185
1186 void ext2_write_super(struct super_block *sb)
1187 {
1188         if (!(sb->s_flags & MS_RDONLY))
1189                 ext2_sync_fs(sb, 1);
1190         else
1191                 sb->s_dirt = 0;
1192 }
1193
1194 static int ext2_remount (struct super_block * sb, int * flags, char * data)
1195 {
1196         struct ext2_sb_info * sbi = EXT2_SB(sb);
1197         struct ext2_super_block * es;
1198         unsigned long old_mount_opt = sbi->s_mount_opt;
1199         struct ext2_mount_options old_opts;
1200         unsigned long old_sb_flags;
1201         int err;
1202
1203         spin_lock(&sbi->s_lock);
1204
1205         /* Store the old options */
1206         old_sb_flags = sb->s_flags;
1207         old_opts.s_mount_opt = sbi->s_mount_opt;
1208         old_opts.s_resuid = sbi->s_resuid;
1209         old_opts.s_resgid = sbi->s_resgid;
1210
1211         /*
1212          * Allow the "check" option to be passed as a remount option.
1213          */
1214         if (!parse_options(data, sb)) {
1215                 err = -EINVAL;
1216                 goto restore_opts;
1217         }
1218
1219         sb->s_flags = (sb->s_flags & ~MS_POSIXACL) |
1220                 ((sbi->s_mount_opt & EXT2_MOUNT_POSIX_ACL) ? MS_POSIXACL : 0);
1221
1222         ext2_xip_verify_sb(sb); /* see if bdev supports xip, unset
1223                                     EXT2_MOUNT_XIP if not */
1224
1225         if ((ext2_use_xip(sb)) && (sb->s_blocksize != PAGE_SIZE)) {
1226                 ext2_msg(sb, KERN_WARNING,
1227                         "warning: unsupported blocksize for xip");
1228                 err = -EINVAL;
1229                 goto restore_opts;
1230         }
1231
1232         es = sbi->s_es;
1233         if ((sbi->s_mount_opt ^ old_mount_opt) & EXT2_MOUNT_XIP) {
1234                 ext2_msg(sb, KERN_WARNING, "warning: refusing change of "
1235                          "xip flag with busy inodes while remounting");
1236                 sbi->s_mount_opt &= ~EXT2_MOUNT_XIP;
1237                 sbi->s_mount_opt |= old_mount_opt & EXT2_MOUNT_XIP;
1238         }
1239         if ((*flags & MS_RDONLY) == (sb->s_flags & MS_RDONLY)) {
1240                 spin_unlock(&sbi->s_lock);
1241                 return 0;
1242         }
1243         if (*flags & MS_RDONLY) {
1244                 if (le16_to_cpu(es->s_state) & EXT2_VALID_FS ||
1245                     !(sbi->s_mount_state & EXT2_VALID_FS)) {
1246                         spin_unlock(&sbi->s_lock);
1247                         return 0;
1248                 }
1249
1250                 /*
1251                  * OK, we are remounting a valid rw partition rdonly, so set
1252                  * the rdonly flag and then mark the partition as valid again.
1253                  */
1254                 es->s_state = cpu_to_le16(sbi->s_mount_state);
1255                 es->s_mtime = cpu_to_le32(get_seconds());
1256                 spin_unlock(&sbi->s_lock);
1257
1258                 err = dquot_suspend(sb, -1);
1259                 if (err < 0) {
1260                         spin_lock(&sbi->s_lock);
1261                         goto restore_opts;
1262                 }
1263
1264                 ext2_sync_super(sb, es, 1);
1265         } else {
1266                 __le32 ret = EXT2_HAS_RO_COMPAT_FEATURE(sb,
1267                                                ~EXT2_FEATURE_RO_COMPAT_SUPP);
1268                 if (ret) {
1269                         ext2_msg(sb, KERN_WARNING,
1270                                 "warning: couldn't remount RDWR because of "
1271                                 "unsupported optional features (%x).",
1272                                 le32_to_cpu(ret));
1273                         err = -EROFS;
1274                         goto restore_opts;
1275                 }
1276                 /*
1277                  * Mounting a RDONLY partition read-write, so reread and
1278                  * store the current valid flag.  (It may have been changed
1279                  * by e2fsck since we originally mounted the partition.)
1280                  */
1281                 sbi->s_mount_state = le16_to_cpu(es->s_state);
1282                 if (!ext2_setup_super (sb, es, 0))
1283                         sb->s_flags &= ~MS_RDONLY;
1284                 spin_unlock(&sbi->s_lock);
1285
1286                 ext2_write_super(sb);
1287
1288                 dquot_resume(sb, -1);
1289         }
1290
1291         return 0;
1292 restore_opts:
1293         sbi->s_mount_opt = old_opts.s_mount_opt;
1294         sbi->s_resuid = old_opts.s_resuid;
1295         sbi->s_resgid = old_opts.s_resgid;
1296         sb->s_flags = old_sb_flags;
1297         spin_unlock(&sbi->s_lock);
1298         return err;
1299 }
1300
1301 static int ext2_statfs (struct dentry * dentry, struct kstatfs * buf)
1302 {
1303         struct super_block *sb = dentry->d_sb;
1304         struct ext2_sb_info *sbi = EXT2_SB(sb);
1305         struct ext2_super_block *es = sbi->s_es;
1306         u64 fsid;
1307
1308         spin_lock(&sbi->s_lock);
1309
1310         if (test_opt (sb, MINIX_DF))
1311                 sbi->s_overhead_last = 0;
1312         else if (sbi->s_blocks_last != le32_to_cpu(es->s_blocks_count)) {
1313                 unsigned long i, overhead = 0;
1314                 smp_rmb();
1315
1316                 /*
1317                  * Compute the overhead (FS structures). This is constant
1318                  * for a given filesystem unless the number of block groups
1319                  * changes so we cache the previous value until it does.
1320                  */
1321
1322                 /*
1323                  * All of the blocks before first_data_block are
1324                  * overhead
1325                  */
1326                 overhead = le32_to_cpu(es->s_first_data_block);
1327
1328                 /*
1329                  * Add the overhead attributed to the superblock and
1330                  * block group descriptors.  If the sparse superblocks
1331                  * feature is turned on, then not all groups have this.
1332                  */
1333                 for (i = 0; i < sbi->s_groups_count; i++)
1334                         overhead += ext2_bg_has_super(sb, i) +
1335                                 ext2_bg_num_gdb(sb, i);
1336
1337                 /*
1338                  * Every block group has an inode bitmap, a block
1339                  * bitmap, and an inode table.
1340                  */
1341                 overhead += (sbi->s_groups_count *
1342                              (2 + sbi->s_itb_per_group));
1343                 sbi->s_overhead_last = overhead;
1344                 smp_wmb();
1345                 sbi->s_blocks_last = le32_to_cpu(es->s_blocks_count);
1346         }
1347
1348         buf->f_type = EXT2_SUPER_MAGIC;
1349         buf->f_bsize = sb->s_blocksize;
1350         buf->f_blocks = le32_to_cpu(es->s_blocks_count) - sbi->s_overhead_last;
1351         buf->f_bfree = ext2_count_free_blocks(sb);
1352         es->s_free_blocks_count = cpu_to_le32(buf->f_bfree);
1353         buf->f_bavail = buf->f_bfree - le32_to_cpu(es->s_r_blocks_count);
1354         if (buf->f_bfree < le32_to_cpu(es->s_r_blocks_count))
1355                 buf->f_bavail = 0;
1356         buf->f_files = le32_to_cpu(es->s_inodes_count);
1357         buf->f_ffree = ext2_count_free_inodes(sb);
1358         es->s_free_inodes_count = cpu_to_le32(buf->f_ffree);
1359         buf->f_namelen = EXT2_NAME_LEN;
1360         fsid = le64_to_cpup((void *)es->s_uuid) ^
1361                le64_to_cpup((void *)es->s_uuid + sizeof(u64));
1362         buf->f_fsid.val[0] = fsid & 0xFFFFFFFFUL;
1363         buf->f_fsid.val[1] = (fsid >> 32) & 0xFFFFFFFFUL;
1364         spin_unlock(&sbi->s_lock);
1365         return 0;
1366 }
1367
1368 static struct dentry *ext2_mount(struct file_system_type *fs_type,
1369         int flags, const char *dev_name, void *data)
1370 {
1371         return mount_bdev(fs_type, flags, dev_name, data, ext2_fill_super);
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         .mount          = ext2_mount,
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)