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         va_list args;
  50         struct ext2_sb_info *sbi = EXT2_SB(sb);
  51         struct ext2_super_block *es = sbi->s_es;
  52
  53         if (!(sb->s_flags & MS_RDONLY)) {
  54                 spin_lock(&sbi->s_lock);
  55                 sbi->s_mount_state |= EXT2_ERROR_FS;
  56                 es->s_state |= cpu_to_le16(EXT2_ERROR_FS);
  57                 spin_unlock(&sbi->s_lock);
  58                 ext2_sync_super(sb, es, 1);
  59         }
  60
  61         va_start(args, fmt);
  62         printk(KERN_CRIT "EXT2-fs (%s): error: %s: ", sb->s_id, function);
  63         vprintk(fmt, args);
  64         printk("\n");
  65         va_end(args);
  66
  67         if (test_opt(sb, ERRORS_PANIC))
  68                 panic("EXT2-fs: panic from previous error\n");
  69         if (test_opt(sb, ERRORS_RO)) {
  70                 ext2_msg(sb, KERN_CRIT,
  71                              "error: remounting filesystem read-only");
  72                 sb->s_flags |= MS_RDONLY;
  73         }
  74 }
  75
  76 void ext2_msg(struct super_block *sb, const char *prefix,
  77                 const char *fmt, ...)
  78 {
  79         va_list args;
  80
  81         va_start(args, fmt);
  82         printk("%sEXT2-fs (%s): ", prefix, sb->s_id);
  83         vprintk(fmt, args);
  84         printk("\n");
  85         va_end(args);
  86 }
  87
  88 /*
  89  * This must be called with sbi->s_lock held.
  90  */
  91 void ext2_update_dynamic_rev(struct super_block *sb)
  92 {
  93         struct ext2_super_block *es = EXT2_SB(sb)->s_es;
  94
  95         if (le32_to_cpu(es->s_rev_level) > EXT2_GOOD_OLD_REV)
  96                 return;
  97
  98         ext2_msg(sb, KERN_WARNING,
  99                      "warning: updating to rev %d because of "
 100                      "new feature flag, running e2fsck is recommended",
 101                      EXT2_DYNAMIC_REV);
 102
 103         es->s_first_ino = cpu_to_le32(EXT2_GOOD_OLD_FIRST_INO);
 104         es->s_inode_size = cpu_to_le16(EXT2_GOOD_OLD_INODE_SIZE);
 105         es->s_rev_level = cpu_to_le32(EXT2_DYNAMIC_REV);
 106         /* leave es->s_feature_*compat flags alone */
 107         /* es->s_uuid will be set by e2fsck if empty */
 108
 109         /*
 110          * The rest of the superblock fields should be zero, and if not it
 111          * means they are likely already in use, so leave them alone.  We
 112          * can leave it up to e2fsck to clean up any inconsistencies there.
 113          */
 114 }
 115
 116 static void ext2_put_super (struct super_block * sb)
 117 {
 118         int db_count;
 119         int i;
 120         struct ext2_sb_info *sbi = EXT2_SB(sb);
 121
 122         dquot_disable(sb, -1, DQUOT_USAGE_ENABLED | DQUOT_LIMITS_ENABLED);
 123
 124         if (sb->s_dirt)
 125                 ext2_write_super(sb);
 126
 127         ext2_xattr_put_super(sb);
 128         if (!(sb->s_flags & MS_RDONLY)) {
 129                 struct ext2_super_block *es = sbi->s_es;
 130
 131                 spin_lock(&sbi->s_lock);
 132                 es->s_state = cpu_to_le16(sbi->s_mount_state);
 133                 spin_unlock(&sbi->s_lock);
 134                 ext2_sync_super(sb, es, 1);
 135         }
 136         db_count = sbi->s_gdb_count;
 137         for (i = 0; i < db_count; i++)
 138                 if (sbi->s_group_desc[i])
 139                         brelse (sbi->s_group_desc[i]);
 140         kfree(sbi->s_group_desc);
 141         kfree(sbi->s_debts);
 142         percpu_counter_destroy(&sbi->s_freeblocks_counter);
 143         percpu_counter_destroy(&sbi->s_freeinodes_counter);
 144         percpu_counter_destroy(&sbi->s_dirs_counter);
 145         brelse (sbi->s_sbh);
 146         sb->s_fs_info = NULL;
 147         kfree(sbi->s_blockgroup_lock);
 148         kfree(sbi);
 149 }
 150
 151 static struct kmem_cache * ext2_inode_cachep;
 152
 153 static struct inode *ext2_alloc_inode(struct super_block *sb)
 154 {
 155         struct ext2_inode_info *ei;
 156         ei = (struct ext2_inode_info *)kmem_cache_alloc(ext2_inode_cachep, GFP_KERNEL);
 157         if (!ei)
 158                 return NULL;
 159         ei->i_block_alloc_info = NULL;
 160         ei->vfs_inode.i_version = 1;
 161         return &ei->vfs_inode;
 162 }
 163
 164 static void ext2_i_callback(struct rcu_head *head)
 165 {
 166         struct inode *inode = container_of(head, struct inode, i_rcu);
 167         INIT_LIST_HEAD(&inode->i_dentry);
 168         kmem_cache_free(ext2_inode_cachep, EXT2_I(inode));
 169 }
 170
 171 static void ext2_destroy_inode(struct inode *inode)
 172 {
 173         call_rcu(&inode->i_rcu, ext2_i_callback);
 174 }
 175
 176 static void init_once(void *foo)
 177 {
 178         struct ext2_inode_info *ei = (struct ext2_inode_info *) foo;
 179
 180         rwlock_init(&ei->i_meta_lock);
 181 #ifdef CONFIG_EXT2_FS_XATTR
 182         init_rwsem(&ei->xattr_sem);
 183 #endif
 184         mutex_init(&ei->truncate_mutex);
 185         inode_init_once(&ei->vfs_inode);
 186 }
 187
 188 static int init_inodecache(void)
 189 {
 190         ext2_inode_cachep = kmem_cache_create("ext2_inode_cache",
 191                                              sizeof(struct ext2_inode_info),
 192                                              0, (SLAB_RECLAIM_ACCOUNT|
 193                                                 SLAB_MEM_SPREAD),
 194                                              init_once);
 195         if (ext2_inode_cachep == NULL)
 196                 return -ENOMEM;
 197         return 0;
 198 }
 199
 200 static void destroy_inodecache(void)
 201 {
 202         kmem_cache_destroy(ext2_inode_cachep);
 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         spin_lock(&sbi->s_lock);
 213         def_mount_opts = le32_to_cpu(es->s_default_mount_opts);
 214
 215         if (sbi->s_sb_block != 1)
 216                 seq_printf(seq, ",sb=%lu", sbi->s_sb_block);
 217         if (test_opt(sb, MINIX_DF))
 218                 seq_puts(seq, ",minixdf");
 219         if (test_opt(sb, GRPID))
 220                 seq_puts(seq, ",grpid");
 221         if (!test_opt(sb, GRPID) && (def_mount_opts & EXT2_DEFM_BSDGROUPS))
 222                 seq_puts(seq, ",nogrpid");
 223         if (sbi->s_resuid != EXT2_DEF_RESUID ||
 224             le16_to_cpu(es->s_def_resuid) != EXT2_DEF_RESUID) {
 225                 seq_printf(seq, ",resuid=%u", sbi->s_resuid);
 226         }
 227         if (sbi->s_resgid != EXT2_DEF_RESGID ||
 228             le16_to_cpu(es->s_def_resgid) != EXT2_DEF_RESGID) {
 229                 seq_printf(seq, ",resgid=%u", sbi->s_resgid);
 230         }
 231         if (test_opt(sb, ERRORS_RO)) {
 232                 int def_errors = le16_to_cpu(es->s_errors);
 233
 234                 if (def_errors == EXT2_ERRORS_PANIC ||
 235                     def_errors == EXT2_ERRORS_CONTINUE) {
 236                         seq_puts(seq, ",errors=remount-ro");
 237                 }
 238         }
 239         if (test_opt(sb, ERRORS_CONT))
 240                 seq_puts(seq, ",errors=continue");
 241         if (test_opt(sb, ERRORS_PANIC))
 242                 seq_puts(seq, ",errors=panic");
 243         if (test_opt(sb, NO_UID32))
 244                 seq_puts(seq, ",nouid32");
 245         if (test_opt(sb, DEBUG))
 246                 seq_puts(seq, ",debug");
 247         if (test_opt(sb, OLDALLOC))
 248                 seq_puts(seq, ",oldalloc");
 249
 250 #ifdef CONFIG_EXT2_FS_XATTR
 251         if (test_opt(sb, XATTR_USER))
 252                 seq_puts(seq, ",user_xattr");
 253         if (!test_opt(sb, XATTR_USER) &&
 254             (def_mount_opts & EXT2_DEFM_XATTR_USER)) {
 255                 seq_puts(seq, ",nouser_xattr");
 256         }
 257 #endif
 258
 259 #ifdef CONFIG_EXT2_FS_POSIX_ACL
 260         if (test_opt(sb, POSIX_ACL))
 261                 seq_puts(seq, ",acl");
 262         if (!test_opt(sb, POSIX_ACL) && (def_mount_opts & EXT2_DEFM_ACL))
 263                 seq_puts(seq, ",noacl");
 264 #endif
 265
 266         if (test_opt(sb, NOBH))
 267                 seq_puts(seq, ",nobh");
 268
 269 #if defined(CONFIG_QUOTA)
 270         if (sbi->s_mount_opt & EXT2_MOUNT_USRQUOTA)
 271                 seq_puts(seq, ",usrquota");
 272
 273         if (sbi->s_mount_opt & EXT2_MOUNT_GRPQUOTA)
 274                 seq_puts(seq, ",grpquota");
 275 #endif
 276
 277 #if defined(CONFIG_EXT2_FS_XIP)
 278         if (sbi->s_mount_opt & EXT2_MOUNT_XIP)
 279                 seq_puts(seq, ",xip");
 280 #endif
 281
 282         if (!test_opt(sb, RESERVATION))
 283                 seq_puts(seq, ",noreservation");
 284
 285         spin_unlock(&sbi->s_lock);
 286         return 0;
 287 }
 288
 289 #ifdef CONFIG_QUOTA
 290 static ssize_t ext2_quota_read(struct super_block *sb, int type, char *data, size_t len, loff_t off);
 291 static ssize_t ext2_quota_write(struct super_block *sb, int type, const char *data, size_t len, loff_t off);
 292 #endif
 293
 294 static const struct super_operations ext2_sops = {
 295         .alloc_inode    = ext2_alloc_inode,
 296         .destroy_inode  = ext2_destroy_inode,
 297         .write_inode    = ext2_write_inode,
 298         .evict_inode    = ext2_evict_inode,
 299         .put_super      = ext2_put_super,
 300         .write_super    = ext2_write_super,
 301         .sync_fs        = ext2_sync_fs,
 302         .statfs         = ext2_statfs,
 303         .remount_fs     = ext2_remount,
 304         .show_options   = ext2_show_options,
 305 #ifdef CONFIG_QUOTA
 306         .quota_read     = ext2_quota_read,
 307         .quota_write    = ext2_quota_write,
 308 #endif
 309 };
 310
 311 static struct inode *ext2_nfs_get_inode(struct super_block *sb,
 312                 u64 ino, u32 generation)
 313 {
 314         struct inode *inode;
 315
 316         if (ino < EXT2_FIRST_INO(sb) && ino != EXT2_ROOT_INO)
 317                 return ERR_PTR(-ESTALE);
 318         if (ino > le32_to_cpu(EXT2_SB(sb)->s_es->s_inodes_count))
 319                 return ERR_PTR(-ESTALE);
 320
 321         /* iget isn't really right if the inode is currently unallocated!!
 322          * ext2_read_inode currently does appropriate checks, but
 323          * it might be "neater" to call ext2_get_inode first and check
 324          * if the inode is valid.....
 325          */
 326         inode = ext2_iget(sb, ino);
 327         if (IS_ERR(inode))
 328                 return ERR_CAST(inode);
 329         if (generation && inode->i_generation != generation) {
 330                 /* we didn't find the right inode.. */
 331                 iput(inode);
 332                 return ERR_PTR(-ESTALE);
 333         }
 334         return inode;
 335 }
 336
 337 static struct dentry *ext2_fh_to_dentry(struct super_block *sb, struct fid *fid,
 338                 int fh_len, int fh_type)
 339 {
 340         return generic_fh_to_dentry(sb, fid, fh_len, fh_type,
 341                                     ext2_nfs_get_inode);
 342 }
 343
 344 static struct dentry *ext2_fh_to_parent(struct super_block *sb, struct fid *fid,
 345                 int fh_len, int fh_type)
 346 {
 347         return generic_fh_to_parent(sb, fid, fh_len, fh_type,
 348                                     ext2_nfs_get_inode);
 349 }
 350
 351 /* Yes, most of these are left as NULL!!
 352  * A NULL value implies the default, which works with ext2-like file
 353  * systems, but can be improved upon.
 354  * Currently only get_parent is required.
 355  */
 356 static const struct export_operations ext2_export_ops = {
 357         .fh_to_dentry = ext2_fh_to_dentry,
 358         .fh_to_parent = ext2_fh_to_parent,
 359         .get_parent = ext2_get_parent,
 360 };
 361
 362 static unsigned long get_sb_block(void **data)
 363 {
 364         unsigned long   sb_block;
 365         char            *options = (char *) *data;
 366
 367         if (!options || strncmp(options, "sb=", 3) != 0)
 368                 return 1;       /* Default location */
 369         options += 3;
 370         sb_block = simple_strtoul(options, &options, 0);
 371         if (*options && *options != ',') {
 372                 printk("EXT2-fs: Invalid sb specification: %s\n",
 373                        (char *) *data);
 374                 return 1;
 375         }
 376         if (*options == ',')
 377                 options++;
 378         *data = (void *) options;
 379         return sb_block;
 380 }
 381
 382 enum {
 383         Opt_bsd_df, Opt_minix_df, Opt_grpid, Opt_nogrpid,
 384         Opt_resgid, Opt_resuid, Opt_sb, Opt_err_cont, Opt_err_panic,
 385         Opt_err_ro, Opt_nouid32, Opt_nocheck, Opt_debug,
 386         Opt_oldalloc, Opt_orlov, Opt_nobh, Opt_user_xattr, Opt_nouser_xattr,
 387         Opt_acl, Opt_noacl, Opt_xip, Opt_ignore, Opt_err, Opt_quota,
 388         Opt_usrquota, Opt_grpquota, Opt_reservation, Opt_noreservation
 389 };
 390
 391 static const match_table_t tokens = {
 392         {Opt_bsd_df, "bsddf"},
 393         {Opt_minix_df, "minixdf"},
 394         {Opt_grpid, "grpid"},
 395         {Opt_grpid, "bsdgroups"},
 396         {Opt_nogrpid, "nogrpid"},
 397         {Opt_nogrpid, "sysvgroups"},
 398         {Opt_resgid, "resgid=%u"},
 399         {Opt_resuid, "resuid=%u"},
 400         {Opt_sb, "sb=%u"},
 401         {Opt_err_cont, "errors=continue"},
 402         {Opt_err_panic, "errors=panic"},
 403         {Opt_err_ro, "errors=remount-ro"},
 404         {Opt_nouid32, "nouid32"},
 405         {Opt_nocheck, "check=none"},
 406         {Opt_nocheck, "nocheck"},
 407         {Opt_debug, "debug"},
 408         {Opt_oldalloc, "oldalloc"},
 409         {Opt_orlov, "orlov"},
 410         {Opt_nobh, "nobh"},
 411         {Opt_user_xattr, "user_xattr"},
 412         {Opt_nouser_xattr, "nouser_xattr"},
 413         {Opt_acl, "acl"},
 414         {Opt_noacl, "noacl"},
 415         {Opt_xip, "xip"},
 416         {Opt_grpquota, "grpquota"},
 417         {Opt_ignore, "noquota"},
 418         {Opt_quota, "quota"},
 419         {Opt_usrquota, "usrquota"},
 420         {Opt_reservation, "reservation"},
 421         {Opt_noreservation, "noreservation"},
 422         {Opt_err, NULL}
 423 };
 424
 425 static int parse_options(char *options, struct super_block *sb)
 426 {
 427         char *p;
 428         struct ext2_sb_info *sbi = EXT2_SB(sb);
 429         substring_t args[MAX_OPT_ARGS];
 430         int option;
 431
 432         if (!options)
 433                 return 1;
 434
 435         while ((p = strsep (&options, ",")) != NULL) {
 436                 int token;
 437                 if (!*p)
 438                         continue;
 439
 440                 token = match_token(p, tokens, args);
 441                 switch (token) {
 442                 case Opt_bsd_df:
 443                         clear_opt (sbi->s_mount_opt, MINIX_DF);
 444                         break;
 445                 case Opt_minix_df:
 446                         set_opt (sbi->s_mount_opt, MINIX_DF);
 447                         break;
 448                 case Opt_grpid:
 449                         set_opt (sbi->s_mount_opt, GRPID);
 450                         break;
 451                 case Opt_nogrpid:
 452                         clear_opt (sbi->s_mount_opt, GRPID);
 453                         break;
 454                 case Opt_resuid:
 455                         if (match_int(&args[0], &option))
 456                                 return 0;
 457                         sbi->s_resuid = option;
 458                         break;
 459                 case Opt_resgid:
 460                         if (match_int(&args[0], &option))
 461                                 return 0;
 462                         sbi->s_resgid = option;
 463                         break;
 464                 case Opt_sb:
 465                         /* handled by get_sb_block() instead of here */
 466                         /* *sb_block = match_int(&args[0]); */
 467                         break;
 468                 case Opt_err_panic:
 469                         clear_opt (sbi->s_mount_opt, ERRORS_CONT);
 470                         clear_opt (sbi->s_mount_opt, ERRORS_RO);
 471                         set_opt (sbi->s_mount_opt, ERRORS_PANIC);
 472                         break;
 473                 case Opt_err_ro:
 474                         clear_opt (sbi->s_mount_opt, ERRORS_CONT);
 475                         clear_opt (sbi->s_mount_opt, ERRORS_PANIC);
 476                         set_opt (sbi->s_mount_opt, ERRORS_RO);
 477                         break;
 478                 case Opt_err_cont:
 479                         clear_opt (sbi->s_mount_opt, ERRORS_RO);
 480                         clear_opt (sbi->s_mount_opt, ERRORS_PANIC);
 481                         set_opt (sbi->s_mount_opt, ERRORS_CONT);
 482                         break;
 483                 case Opt_nouid32:
 484                         set_opt (sbi->s_mount_opt, NO_UID32);
 485                         break;
 486                 case Opt_nocheck:
 487                         clear_opt (sbi->s_mount_opt, CHECK);
 488                         break;
 489                 case Opt_debug:
 490                         set_opt (sbi->s_mount_opt, DEBUG);
 491                         break;
 492                 case Opt_oldalloc:
 493                         set_opt (sbi->s_mount_opt, OLDALLOC);
 494                         break;
 495                 case Opt_orlov:
 496                         clear_opt (sbi->s_mount_opt, OLDALLOC);
 497                         break;
 498                 case Opt_nobh:
 499                         set_opt (sbi->s_mount_opt, NOBH);
 500                         break;
 501 #ifdef CONFIG_EXT2_FS_XATTR
 502                 case Opt_user_xattr:
 503                         set_opt (sbi->s_mount_opt, XATTR_USER);
 504                         break;
 505                 case Opt_nouser_xattr:
 506                         clear_opt (sbi->s_mount_opt, XATTR_USER);
 507                         break;
 508 #else
 509                 case Opt_user_xattr:
 510                 case Opt_nouser_xattr:
 511                         ext2_msg(sb, KERN_INFO, "(no)user_xattr options"
 512                                 "not supported");
 513                         break;
 514 #endif
 515 #ifdef CONFIG_EXT2_FS_POSIX_ACL
 516                 case Opt_acl:
 517                         set_opt(sbi->s_mount_opt, POSIX_ACL);
 518                         break;
 519                 case Opt_noacl:
 520                         clear_opt(sbi->s_mount_opt, POSIX_ACL);
 521                         break;
 522 #else
 523                 case Opt_acl:
 524                 case Opt_noacl:
 525                         ext2_msg(sb, KERN_INFO,
 526                                 "(no)acl options not supported");
 527                         break;
 528 #endif
 529                 case Opt_xip:
 530 #ifdef CONFIG_EXT2_FS_XIP
 531                         set_opt (sbi->s_mount_opt, XIP);
 532 #else
 533                         ext2_msg(sb, KERN_INFO, "xip option not supported");
 534 #endif
 535                         break;
 536
 537 #if defined(CONFIG_QUOTA)
 538                 case Opt_quota:
 539                 case Opt_usrquota:
 540                         set_opt(sbi->s_mount_opt, USRQUOTA);
 541                         break;
 542
 543                 case Opt_grpquota:
 544                         set_opt(sbi->s_mount_opt, GRPQUOTA);
 545                         break;
 546 #else
 547                 case Opt_quota:
 548                 case Opt_usrquota:
 549                 case Opt_grpquota:
 550                         ext2_msg(sb, KERN_INFO,
 551                                 "quota operations not supported");
 552                         break;
 553 #endif
 554
 555                 case Opt_reservation:
 556                         set_opt(sbi->s_mount_opt, RESERVATION);
 557                         ext2_msg(sb, KERN_INFO, "reservations ON");
 558                         break;
 559                 case Opt_noreservation:
 560                         clear_opt(sbi->s_mount_opt, RESERVATION);
 561                         ext2_msg(sb, KERN_INFO, "reservations OFF");
 562                         break;
 563                 case Opt_ignore:
 564                         break;
 565                 default:
 566                         return 0;
 567                 }
 568         }
 569         return 1;
 570 }
 571
 572 static int ext2_setup_super (struct super_block * sb,
 573                               struct ext2_super_block * es,
 574                               int read_only)
 575 {
 576         int res = 0;
 577         struct ext2_sb_info *sbi = EXT2_SB(sb);
 578
 579         if (le32_to_cpu(es->s_rev_level) > EXT2_MAX_SUPP_REV) {
 580                 ext2_msg(sb, KERN_ERR,
 581                         "error: revision level too high, "
 582                         "forcing read-only mode");
 583                 res = MS_RDONLY;
 584         }
 585         if (read_only)
 586                 return res;
 587         if (!(sbi->s_mount_state & EXT2_VALID_FS))
 588                 ext2_msg(sb, KERN_WARNING,
 589                         "warning: mounting unchecked fs, "
 590                         "running e2fsck is recommended");
 591         else if ((sbi->s_mount_state & EXT2_ERROR_FS))
 592                 ext2_msg(sb, KERN_WARNING,
 593                         "warning: mounting fs with errors, "
 594                         "running e2fsck is recommended");
 595         else if ((__s16) le16_to_cpu(es->s_max_mnt_count) >= 0 &&
 596                  le16_to_cpu(es->s_mnt_count) >=
 597                  (unsigned short) (__s16) le16_to_cpu(es->s_max_mnt_count))
 598                 ext2_msg(sb, KERN_WARNING,
 599                         "warning: maximal mount count reached, "
 600                         "running e2fsck is recommended");
 601         else if (le32_to_cpu(es->s_checkinterval) &&
 602                 (le32_to_cpu(es->s_lastcheck) +
 603                         le32_to_cpu(es->s_checkinterval) <= get_seconds()))
 604                 ext2_msg(sb, KERN_WARNING,
 605                         "warning: checktime reached, "
 606                         "running e2fsck is recommended");
 607         if (!le16_to_cpu(es->s_max_mnt_count))
 608                 es->s_max_mnt_count = cpu_to_le16(EXT2_DFL_MAX_MNT_COUNT);
 609         le16_add_cpu(&es->s_mnt_count, 1);
 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         err = -ENOMEM;
 758         sbi = kzalloc(sizeof(*sbi), GFP_KERNEL);
 759         if (!sbi)
 760                 goto failed_unlock;
 761
 762         sbi->s_blockgroup_lock =
 763                 kzalloc(sizeof(struct blockgroup_lock), GFP_KERNEL);
 764         if (!sbi->s_blockgroup_lock) {
 765                 kfree(sbi);
 766                 goto failed_unlock;
 767         }
 768         sb->s_fs_info = sbi;
 769         sbi->s_sb_block = sb_block;
 770
 771         spin_lock_init(&sbi->s_lock);
 772
 773         /*
 774          * See what the current blocksize for the device is, and
 775          * use that as the blocksize.  Otherwise (or if the blocksize
 776          * is smaller than the default) use the default.
 777          * This is important for devices that have a hardware
 778          * sectorsize that is larger than the default.
 779          */
 780         blocksize = sb_min_blocksize(sb, BLOCK_SIZE);
 781         if (!blocksize) {
 782                 ext2_msg(sb, KERN_ERR, "error: unable to set blocksize");
 783                 goto failed_sbi;
 784         }
 785
 786         /*
 787          * If the superblock doesn't start on a hardware sector boundary,
 788          * calculate the offset.
 789          */
 790         if (blocksize != BLOCK_SIZE) {
 791                 logic_sb_block = (sb_block*BLOCK_SIZE) / blocksize;
 792                 offset = (sb_block*BLOCK_SIZE) % blocksize;
 793         } else {
 794                 logic_sb_block = sb_block;
 795         }
 796
 797         if (!(bh = sb_bread(sb, logic_sb_block))) {
 798                 ext2_msg(sb, KERN_ERR, "error: unable to read superblock");
 799                 goto failed_sbi;
 800         }
 801         /*
 802          * Note: s_es must be initialized as soon as possible because
 803          *       some ext2 macro-instructions depend on its value
 804          */
 805         es = (struct ext2_super_block *) (((char *)bh->b_data) + offset);
 806         sbi->s_es = es;
 807         sb->s_magic = le16_to_cpu(es->s_magic);
 808
 809         if (sb->s_magic != EXT2_SUPER_MAGIC)
 810                 goto cantfind_ext2;
 811
 812         /* Set defaults before we parse the mount options */
 813         def_mount_opts = le32_to_cpu(es->s_default_mount_opts);
 814         if (def_mount_opts & EXT2_DEFM_DEBUG)
 815                 set_opt(sbi->s_mount_opt, DEBUG);
 816         if (def_mount_opts & EXT2_DEFM_BSDGROUPS)
 817                 set_opt(sbi->s_mount_opt, GRPID);
 818         if (def_mount_opts & EXT2_DEFM_UID16)
 819                 set_opt(sbi->s_mount_opt, NO_UID32);
 820 #ifdef CONFIG_EXT2_FS_XATTR
 821         if (def_mount_opts & EXT2_DEFM_XATTR_USER)
 822                 set_opt(sbi->s_mount_opt, XATTR_USER);
 823 #endif
 824 #ifdef CONFIG_EXT2_FS_POSIX_ACL
 825         if (def_mount_opts & EXT2_DEFM_ACL)
 826                 set_opt(sbi->s_mount_opt, POSIX_ACL);
 827 #endif
 828
 829         if (le16_to_cpu(sbi->s_es->s_errors) == EXT2_ERRORS_PANIC)
 830                 set_opt(sbi->s_mount_opt, ERRORS_PANIC);
 831         else if (le16_to_cpu(sbi->s_es->s_errors) == EXT2_ERRORS_CONTINUE)
 832                 set_opt(sbi->s_mount_opt, ERRORS_CONT);
 833         else
 834                 set_opt(sbi->s_mount_opt, ERRORS_RO);
 835
 836         sbi->s_resuid = le16_to_cpu(es->s_def_resuid);
 837         sbi->s_resgid = le16_to_cpu(es->s_def_resgid);
 838
 839         set_opt(sbi->s_mount_opt, RESERVATION);
 840
 841         if (!parse_options((char *) data, sb))
 842                 goto failed_mount;
 843
 844         sb->s_flags = (sb->s_flags & ~MS_POSIXACL) |
 845                 ((EXT2_SB(sb)->s_mount_opt & EXT2_MOUNT_POSIX_ACL) ?
 846                  MS_POSIXACL : 0);
 847
 848         ext2_xip_verify_sb(sb); /* see if bdev supports xip, unset
 849                                     EXT2_MOUNT_XIP if not */
 850
 851         if (le32_to_cpu(es->s_rev_level) == EXT2_GOOD_OLD_REV &&
 852             (EXT2_HAS_COMPAT_FEATURE(sb, ~0U) ||
 853              EXT2_HAS_RO_COMPAT_FEATURE(sb, ~0U) ||
 854              EXT2_HAS_INCOMPAT_FEATURE(sb, ~0U)))
 855                 ext2_msg(sb, KERN_WARNING,
 856                         "warning: feature flags set on rev 0 fs, "
 857                         "running e2fsck is recommended");
 858         /*
 859          * Check feature flags regardless of the revision level, since we
 860          * previously didn't change the revision level when setting the flags,
 861          * so there is a chance incompat flags are set on a rev 0 filesystem.
 862          */
 863         features = EXT2_HAS_INCOMPAT_FEATURE(sb, ~EXT2_FEATURE_INCOMPAT_SUPP);
 864         if (features) {
 865                 ext2_msg(sb, KERN_ERR,  "error: couldn't mount because of "
 866                        "unsupported optional features (%x)",
 867                         le32_to_cpu(features));
 868                 goto failed_mount;
 869         }
 870         if (!(sb->s_flags & MS_RDONLY) &&
 871             (features = EXT2_HAS_RO_COMPAT_FEATURE(sb, ~EXT2_FEATURE_RO_COMPAT_SUPP))){
 872                 ext2_msg(sb, KERN_ERR, "error: couldn't mount RDWR because of "
 873                        "unsupported optional features (%x)",
 874                        le32_to_cpu(features));
 875                 goto failed_mount;
 876         }
 877
 878         blocksize = BLOCK_SIZE << le32_to_cpu(sbi->s_es->s_log_block_size);
 879
 880         if (ext2_use_xip(sb) && blocksize != PAGE_SIZE) {
 881                 if (!silent)
 882                         ext2_msg(sb, KERN_ERR,
 883                                 "error: unsupported blocksize for xip");
 884                 goto failed_mount;
 885         }
 886
 887         /* If the blocksize doesn't match, re-read the thing.. */
 888         if (sb->s_blocksize != blocksize) {
 889                 brelse(bh);
 890
 891                 if (!sb_set_blocksize(sb, blocksize)) {
 892                         ext2_msg(sb, KERN_ERR, "error: blocksize is too small");
 893                         goto failed_sbi;
 894                 }
 895
 896                 logic_sb_block = (sb_block*BLOCK_SIZE) / blocksize;
 897                 offset = (sb_block*BLOCK_SIZE) % blocksize;
 898                 bh = sb_bread(sb, logic_sb_block);
 899                 if(!bh) {
 900                         ext2_msg(sb, KERN_ERR, "error: couldn't read"
 901                                 "superblock on 2nd try");
 902                         goto failed_sbi;
 903                 }
 904                 es = (struct ext2_super_block *) (((char *)bh->b_data) + offset);
 905                 sbi->s_es = es;
 906                 if (es->s_magic != cpu_to_le16(EXT2_SUPER_MAGIC)) {
 907                         ext2_msg(sb, KERN_ERR, "error: magic mismatch");
 908                         goto failed_mount;
 909                 }
 910         }
 911
 912         sb->s_maxbytes = ext2_max_size(sb->s_blocksize_bits);
 913
 914         if (le32_to_cpu(es->s_rev_level) == EXT2_GOOD_OLD_REV) {
 915                 sbi->s_inode_size = EXT2_GOOD_OLD_INODE_SIZE;
 916                 sbi->s_first_ino = EXT2_GOOD_OLD_FIRST_INO;
 917         } else {
 918                 sbi->s_inode_size = le16_to_cpu(es->s_inode_size);
 919                 sbi->s_first_ino = le32_to_cpu(es->s_first_ino);
 920                 if ((sbi->s_inode_size < EXT2_GOOD_OLD_INODE_SIZE) ||
 921                     !is_power_of_2(sbi->s_inode_size) ||
 922                     (sbi->s_inode_size > blocksize)) {
 923                         ext2_msg(sb, KERN_ERR,
 924                                 "error: unsupported inode size: %d",
 925                                 sbi->s_inode_size);
 926                         goto failed_mount;
 927                 }
 928         }
 929
 930         sbi->s_frag_size = EXT2_MIN_FRAG_SIZE <<
 931                                    le32_to_cpu(es->s_log_frag_size);
 932         if (sbi->s_frag_size == 0)
 933                 goto cantfind_ext2;
 934         sbi->s_frags_per_block = sb->s_blocksize / sbi->s_frag_size;
 935
 936         sbi->s_blocks_per_group = le32_to_cpu(es->s_blocks_per_group);
 937         sbi->s_frags_per_group = le32_to_cpu(es->s_frags_per_group);
 938         sbi->s_inodes_per_group = le32_to_cpu(es->s_inodes_per_group);
 939
 940         if (EXT2_INODE_SIZE(sb) == 0)
 941                 goto cantfind_ext2;
 942         sbi->s_inodes_per_block = sb->s_blocksize / EXT2_INODE_SIZE(sb);
 943         if (sbi->s_inodes_per_block == 0 || sbi->s_inodes_per_group == 0)
 944                 goto cantfind_ext2;
 945         sbi->s_itb_per_group = sbi->s_inodes_per_group /
 946                                         sbi->s_inodes_per_block;
 947         sbi->s_desc_per_block = sb->s_blocksize /
 948                                         sizeof (struct ext2_group_desc);
 949         sbi->s_sbh = bh;
 950         sbi->s_mount_state = le16_to_cpu(es->s_state);
 951         sbi->s_addr_per_block_bits =
 952                 ilog2 (EXT2_ADDR_PER_BLOCK(sb));
 953         sbi->s_desc_per_block_bits =
 954                 ilog2 (EXT2_DESC_PER_BLOCK(sb));
 955
 956         if (sb->s_magic != EXT2_SUPER_MAGIC)
 957                 goto cantfind_ext2;
 958
 959         if (sb->s_blocksize != bh->b_size) {
 960                 if (!silent)
 961                         ext2_msg(sb, KERN_ERR, "error: unsupported blocksize");
 962                 goto failed_mount;
 963         }
 964
 965         if (sb->s_blocksize != sbi->s_frag_size) {
 966                 ext2_msg(sb, KERN_ERR,
 967                         "error: fragsize %lu != blocksize %lu"
 968                         "(not supported yet)",
 969                         sbi->s_frag_size, sb->s_blocksize);
 970                 goto failed_mount;
 971         }
 972
 973         if (sbi->s_blocks_per_group > sb->s_blocksize * 8) {
 974                 ext2_msg(sb, KERN_ERR,
 975                         "error: #blocks per group too big: %lu",
 976                         sbi->s_blocks_per_group);
 977                 goto failed_mount;
 978         }
 979         if (sbi->s_frags_per_group > sb->s_blocksize * 8) {
 980                 ext2_msg(sb, KERN_ERR,
 981                         "error: #fragments per group too big: %lu",
 982                         sbi->s_frags_per_group);
 983                 goto failed_mount;
 984         }
 985         if (sbi->s_inodes_per_group > sb->s_blocksize * 8) {
 986                 ext2_msg(sb, KERN_ERR,
 987                         "error: #inodes per group too big: %lu",
 988                         sbi->s_inodes_per_group);
 989                 goto failed_mount;
 990         }
 991
 992         if (EXT2_BLOCKS_PER_GROUP(sb) == 0)
 993                 goto cantfind_ext2;
 994         sbi->s_groups_count = ((le32_to_cpu(es->s_blocks_count) -
 995                                 le32_to_cpu(es->s_first_data_block) - 1)
 996                                         / EXT2_BLOCKS_PER_GROUP(sb)) + 1;
 997         db_count = (sbi->s_groups_count + EXT2_DESC_PER_BLOCK(sb) - 1) /
 998                    EXT2_DESC_PER_BLOCK(sb);
 999         sbi->s_group_desc = kmalloc (db_count * sizeof (struct buffer_head *), GFP_KERNEL);
1000         if (sbi->s_group_desc == NULL) {
1001                 ext2_msg(sb, KERN_ERR, "error: not enough memory");
1002                 goto failed_mount;
1003         }
1004         bgl_lock_init(sbi->s_blockgroup_lock);
1005         sbi->s_debts = kcalloc(sbi->s_groups_count, sizeof(*sbi->s_debts), GFP_KERNEL);
1006         if (!sbi->s_debts) {
1007                 ext2_msg(sb, KERN_ERR, "error: not enough memory");
1008                 goto failed_mount_group_desc;
1009         }
1010         for (i = 0; i < db_count; i++) {
1011                 block = descriptor_loc(sb, logic_sb_block, i);
1012                 sbi->s_group_desc[i] = sb_bread(sb, block);
1013                 if (!sbi->s_group_desc[i]) {
1014                         for (j = 0; j < i; j++)
1015                                 brelse (sbi->s_group_desc[j]);
1016                         ext2_msg(sb, KERN_ERR,
1017                                 "error: unable to read group descriptors");
1018                         goto failed_mount_group_desc;
1019                 }
1020         }
1021         if (!ext2_check_descriptors (sb)) {
1022                 ext2_msg(sb, KERN_ERR, "group descriptors corrupted");
1023                 goto failed_mount2;
1024         }
1025         sbi->s_gdb_count = db_count;
1026         get_random_bytes(&sbi->s_next_generation, sizeof(u32));
1027         spin_lock_init(&sbi->s_next_gen_lock);
1028
1029         /* per fileystem reservation list head & lock */
1030         spin_lock_init(&sbi->s_rsv_window_lock);
1031         sbi->s_rsv_window_root = RB_ROOT;
1032         /*
1033          * Add a single, static dummy reservation to the start of the
1034          * reservation window list --- it gives us a placeholder for
1035          * append-at-start-of-list which makes the allocation logic
1036          * _much_ simpler.
1037          */
1038         sbi->s_rsv_window_head.rsv_start = EXT2_RESERVE_WINDOW_NOT_ALLOCATED;
1039         sbi->s_rsv_window_head.rsv_end = EXT2_RESERVE_WINDOW_NOT_ALLOCATED;
1040         sbi->s_rsv_window_head.rsv_alloc_hit = 0;
1041         sbi->s_rsv_window_head.rsv_goal_size = 0;
1042         ext2_rsv_window_add(sb, &sbi->s_rsv_window_head);
1043
1044         err = percpu_counter_init(&sbi->s_freeblocks_counter,
1045                                 ext2_count_free_blocks(sb));
1046         if (!err) {
1047                 err = percpu_counter_init(&sbi->s_freeinodes_counter,
1048                                 ext2_count_free_inodes(sb));
1049         }
1050         if (!err) {
1051                 err = percpu_counter_init(&sbi->s_dirs_counter,
1052                                 ext2_count_dirs(sb));
1053         }
1054         if (err) {
1055                 ext2_msg(sb, KERN_ERR, "error: insufficient memory");
1056                 goto failed_mount3;
1057         }
1058         /*
1059          * set up enough so that it can read an inode
1060          */
1061         sb->s_op = &ext2_sops;
1062         sb->s_export_op = &ext2_export_ops;
1063         sb->s_xattr = ext2_xattr_handlers;
1064
1065 #ifdef CONFIG_QUOTA
1066         sb->dq_op = &dquot_operations;
1067         sb->s_qcop = &dquot_quotactl_ops;
1068 #endif
1069
1070         root = ext2_iget(sb, EXT2_ROOT_INO);
1071         if (IS_ERR(root)) {
1072                 ret = PTR_ERR(root);
1073                 goto failed_mount3;
1074         }
1075         if (!S_ISDIR(root->i_mode) || !root->i_blocks || !root->i_size) {
1076                 iput(root);
1077                 ext2_msg(sb, KERN_ERR, "error: corrupt root inode, run e2fsck");
1078                 goto failed_mount3;
1079         }
1080
1081         sb->s_root = d_alloc_root(root);
1082         if (!sb->s_root) {
1083                 iput(root);
1084                 ext2_msg(sb, KERN_ERR, "error: get root inode failed");
1085                 ret = -ENOMEM;
1086                 goto failed_mount3;
1087         }
1088         if (EXT2_HAS_COMPAT_FEATURE(sb, EXT3_FEATURE_COMPAT_HAS_JOURNAL))
1089                 ext2_msg(sb, KERN_WARNING,
1090                         "warning: mounting ext3 filesystem as ext2");
1091         if (ext2_setup_super (sb, es, sb->s_flags & MS_RDONLY))
1092                 sb->s_flags |= MS_RDONLY;
1093         ext2_write_super(sb);
1094         return 0;
1095
1096 cantfind_ext2:
1097         if (!silent)
1098                 ext2_msg(sb, KERN_ERR,
1099                         "error: can't find an ext2 filesystem on dev %s.",
1100                         sb->s_id);
1101         goto failed_mount;
1102 failed_mount3:
1103         percpu_counter_destroy(&sbi->s_freeblocks_counter);
1104         percpu_counter_destroy(&sbi->s_freeinodes_counter);
1105         percpu_counter_destroy(&sbi->s_dirs_counter);
1106 failed_mount2:
1107         for (i = 0; i < db_count; i++)
1108                 brelse(sbi->s_group_desc[i]);
1109 failed_mount_group_desc:
1110         kfree(sbi->s_group_desc);
1111         kfree(sbi->s_debts);
1112 failed_mount:
1113         brelse(bh);
1114 failed_sbi:
1115         sb->s_fs_info = NULL;
1116         kfree(sbi->s_blockgroup_lock);
1117         kfree(sbi);
1118 failed_unlock:
1119         return ret;
1120 }
1121
1122 static void ext2_clear_super_error(struct super_block *sb)
1123 {
1124         struct buffer_head *sbh = EXT2_SB(sb)->s_sbh;
1125
1126         if (buffer_write_io_error(sbh)) {
1127                 /*
1128                  * Oh, dear.  A previous attempt to write the
1129                  * superblock failed.  This could happen because the
1130                  * USB device was yanked out.  Or it could happen to
1131                  * be a transient write error and maybe the block will
1132                  * be remapped.  Nothing we can do but to retry the
1133                  * write and hope for the best.
1134                  */
1135                 ext2_msg(sb, KERN_ERR,
1136                        "previous I/O error to superblock detected\n");
1137                 clear_buffer_write_io_error(sbh);
1138                 set_buffer_uptodate(sbh);
1139         }
1140 }
1141
1142 static void ext2_sync_super(struct super_block *sb, struct ext2_super_block *es,
1143                             int wait)
1144 {
1145         ext2_clear_super_error(sb);
1146         spin_lock(&EXT2_SB(sb)->s_lock);
1147         es->s_free_blocks_count = cpu_to_le32(ext2_count_free_blocks(sb));
1148         es->s_free_inodes_count = cpu_to_le32(ext2_count_free_inodes(sb));
1149         es->s_wtime = cpu_to_le32(get_seconds());
1150         /* unlock before we do IO */
1151         spin_unlock(&EXT2_SB(sb)->s_lock);
1152         mark_buffer_dirty(EXT2_SB(sb)->s_sbh);
1153         if (wait)
1154                 sync_dirty_buffer(EXT2_SB(sb)->s_sbh);
1155         sb->s_dirt = 0;
1156 }
1157
1158 /*
1159  * In the second extended file system, it is not necessary to
1160  * write the super block since we use a mapping of the
1161  * disk super block in a buffer.
1162  *
1163  * However, this function is still used to set the fs valid
1164  * flags to 0.  We need to set this flag to 0 since the fs
1165  * may have been checked while mounted and e2fsck may have
1166  * set s_state to EXT2_VALID_FS after some corrections.
1167  */
1168 static int ext2_sync_fs(struct super_block *sb, int wait)
1169 {
1170         struct ext2_sb_info *sbi = EXT2_SB(sb);
1171         struct ext2_super_block *es = EXT2_SB(sb)->s_es;
1172
1173         spin_lock(&sbi->s_lock);
1174         if (es->s_state & cpu_to_le16(EXT2_VALID_FS)) {
1175                 ext2_debug("setting valid to 0\n");
1176                 es->s_state &= cpu_to_le16(~EXT2_VALID_FS);
1177         }
1178         spin_unlock(&sbi->s_lock);
1179         ext2_sync_super(sb, es, wait);
1180         return 0;
1181 }
1182
1183
1184 void ext2_write_super(struct super_block *sb)
1185 {
1186         if (!(sb->s_flags & MS_RDONLY))
1187                 ext2_sync_fs(sb, 1);
1188         else
1189                 sb->s_dirt = 0;
1190 }
1191
1192 static int ext2_remount (struct super_block * sb, int * flags, char * data)
1193 {
1194         struct ext2_sb_info * sbi = EXT2_SB(sb);
1195         struct ext2_super_block * es;
1196         unsigned long old_mount_opt = sbi->s_mount_opt;
1197         struct ext2_mount_options old_opts;
1198         unsigned long old_sb_flags;
1199         int err;
1200
1201         spin_lock(&sbi->s_lock);
1202
1203         /* Store the old options */
1204         old_sb_flags = sb->s_flags;
1205         old_opts.s_mount_opt = sbi->s_mount_opt;
1206         old_opts.s_resuid = sbi->s_resuid;
1207         old_opts.s_resgid = sbi->s_resgid;
1208
1209         /*
1210          * Allow the "check" option to be passed as a remount option.
1211          */
1212         if (!parse_options(data, sb)) {
1213                 err = -EINVAL;
1214                 goto restore_opts;
1215         }
1216
1217         sb->s_flags = (sb->s_flags & ~MS_POSIXACL) |
1218                 ((sbi->s_mount_opt & EXT2_MOUNT_POSIX_ACL) ? MS_POSIXACL : 0);
1219
1220         ext2_xip_verify_sb(sb); /* see if bdev supports xip, unset
1221                                     EXT2_MOUNT_XIP if not */
1222
1223         if ((ext2_use_xip(sb)) && (sb->s_blocksize != PAGE_SIZE)) {
1224                 ext2_msg(sb, KERN_WARNING,
1225                         "warning: unsupported blocksize for xip");
1226                 err = -EINVAL;
1227                 goto restore_opts;
1228         }
1229
1230         es = sbi->s_es;
1231         if ((sbi->s_mount_opt ^ old_mount_opt) & EXT2_MOUNT_XIP) {
1232                 ext2_msg(sb, KERN_WARNING, "warning: refusing change of "
1233                          "xip flag with busy inodes while remounting");
1234                 sbi->s_mount_opt &= ~EXT2_MOUNT_XIP;
1235                 sbi->s_mount_opt |= old_mount_opt & EXT2_MOUNT_XIP;
1236         }
1237         if ((*flags & MS_RDONLY) == (sb->s_flags & MS_RDONLY)) {
1238                 spin_unlock(&sbi->s_lock);
1239                 return 0;
1240         }
1241         if (*flags & MS_RDONLY) {
1242                 if (le16_to_cpu(es->s_state) & EXT2_VALID_FS ||
1243                     !(sbi->s_mount_state & EXT2_VALID_FS)) {
1244                         spin_unlock(&sbi->s_lock);
1245                         return 0;
1246                 }
1247
1248                 /*
1249                  * OK, we are remounting a valid rw partition rdonly, so set
1250                  * the rdonly flag and then mark the partition as valid again.
1251                  */
1252                 es->s_state = cpu_to_le16(sbi->s_mount_state);
1253                 es->s_mtime = cpu_to_le32(get_seconds());
1254                 spin_unlock(&sbi->s_lock);
1255
1256                 err = dquot_suspend(sb, -1);
1257                 if (err < 0) {
1258                         spin_lock(&sbi->s_lock);
1259                         goto restore_opts;
1260                 }
1261
1262                 ext2_sync_super(sb, es, 1);
1263         } else {
1264                 __le32 ret = EXT2_HAS_RO_COMPAT_FEATURE(sb,
1265                                                ~EXT2_FEATURE_RO_COMPAT_SUPP);
1266                 if (ret) {
1267                         ext2_msg(sb, KERN_WARNING,
1268                                 "warning: couldn't remount RDWR because of "
1269                                 "unsupported optional features (%x).",
1270                                 le32_to_cpu(ret));
1271                         err = -EROFS;
1272                         goto restore_opts;
1273                 }
1274                 /*
1275                  * Mounting a RDONLY partition read-write, so reread and
1276                  * store the current valid flag.  (It may have been changed
1277                  * by e2fsck since we originally mounted the partition.)
1278                  */
1279                 sbi->s_mount_state = le16_to_cpu(es->s_state);
1280                 if (!ext2_setup_super (sb, es, 0))
1281                         sb->s_flags &= ~MS_RDONLY;
1282                 spin_unlock(&sbi->s_lock);
1283
1284                 ext2_write_super(sb);
1285
1286                 dquot_resume(sb, -1);
1287         }
1288
1289         return 0;
1290 restore_opts:
1291         sbi->s_mount_opt = old_opts.s_mount_opt;
1292         sbi->s_resuid = old_opts.s_resuid;
1293         sbi->s_resgid = old_opts.s_resgid;
1294         sb->s_flags = old_sb_flags;
1295         spin_unlock(&sbi->s_lock);
1296         return err;
1297 }
1298
1299 static int ext2_statfs (struct dentry * dentry, struct kstatfs * buf)
1300 {
1301         struct super_block *sb = dentry->d_sb;
1302         struct ext2_sb_info *sbi = EXT2_SB(sb);
1303         struct ext2_super_block *es = sbi->s_es;
1304         u64 fsid;
1305
1306         spin_lock(&sbi->s_lock);
1307
1308         if (test_opt (sb, MINIX_DF))
1309                 sbi->s_overhead_last = 0;
1310         else if (sbi->s_blocks_last != le32_to_cpu(es->s_blocks_count)) {
1311                 unsigned long i, overhead = 0;
1312                 smp_rmb();
1313
1314                 /*
1315                  * Compute the overhead (FS structures). This is constant
1316                  * for a given filesystem unless the number of block groups
1317                  * changes so we cache the previous value until it does.
1318                  */
1319
1320                 /*
1321                  * All of the blocks before first_data_block are
1322                  * overhead
1323                  */
1324                 overhead = le32_to_cpu(es->s_first_data_block);
1325
1326                 /*
1327                  * Add the overhead attributed to the superblock and
1328                  * block group descriptors.  If the sparse superblocks
1329                  * feature is turned on, then not all groups have this.
1330                  */
1331                 for (i = 0; i < sbi->s_groups_count; i++)
1332                         overhead += ext2_bg_has_super(sb, i) +
1333                                 ext2_bg_num_gdb(sb, i);
1334
1335                 /*
1336                  * Every block group has an inode bitmap, a block
1337                  * bitmap, and an inode table.
1338                  */
1339                 overhead += (sbi->s_groups_count *
1340                              (2 + sbi->s_itb_per_group));
1341                 sbi->s_overhead_last = overhead;
1342                 smp_wmb();
1343                 sbi->s_blocks_last = le32_to_cpu(es->s_blocks_count);
1344         }
1345
1346         buf->f_type = EXT2_SUPER_MAGIC;
1347         buf->f_bsize = sb->s_blocksize;
1348         buf->f_blocks = le32_to_cpu(es->s_blocks_count) - sbi->s_overhead_last;
1349         buf->f_bfree = ext2_count_free_blocks(sb);
1350         es->s_free_blocks_count = cpu_to_le32(buf->f_bfree);
1351         buf->f_bavail = buf->f_bfree - le32_to_cpu(es->s_r_blocks_count);
1352         if (buf->f_bfree < le32_to_cpu(es->s_r_blocks_count))
1353                 buf->f_bavail = 0;
1354         buf->f_files = le32_to_cpu(es->s_inodes_count);
1355         buf->f_ffree = ext2_count_free_inodes(sb);
1356         es->s_free_inodes_count = cpu_to_le32(buf->f_ffree);
1357         buf->f_namelen = EXT2_NAME_LEN;
1358         fsid = le64_to_cpup((void *)es->s_uuid) ^
1359                le64_to_cpup((void *)es->s_uuid + sizeof(u64));
1360         buf->f_fsid.val[0] = fsid & 0xFFFFFFFFUL;
1361         buf->f_fsid.val[1] = (fsid >> 32) & 0xFFFFFFFFUL;
1362         spin_unlock(&sbi->s_lock);
1363         return 0;
1364 }
1365
1366 static struct dentry *ext2_mount(struct file_system_type *fs_type,
1367         int flags, const char *dev_name, void *data)
1368 {
1369         return mount_bdev(fs_type, flags, dev_name, data, ext2_fill_super);
1370 }
1371
1372 #ifdef CONFIG_QUOTA
1373
1374 /* Read data from quotafile - avoid pagecache and such because we cannot afford
1375  * acquiring the locks... As quota files are never truncated and quota code
1376  * itself serializes the operations (and noone else should touch the files)
1377  * we don't have to be afraid of races */
1378 static ssize_t ext2_quota_read(struct super_block *sb, int type, char *data,
1379                                size_t len, loff_t off)
1380 {
1381         struct inode *inode = sb_dqopt(sb)->files[type];
1382         sector_t blk = off >> EXT2_BLOCK_SIZE_BITS(sb);
1383         int err = 0;
1384         int offset = off & (sb->s_blocksize - 1);
1385         int tocopy;
1386         size_t toread;
1387         struct buffer_head tmp_bh;
1388         struct buffer_head *bh;
1389         loff_t i_size = i_size_read(inode);
1390
1391         if (off > i_size)
1392                 return 0;
1393         if (off+len > i_size)
1394                 len = i_size-off;
1395         toread = len;
1396         while (toread > 0) {
1397                 tocopy = sb->s_blocksize - offset < toread ?
1398                                 sb->s_blocksize - offset : toread;
1399
1400                 tmp_bh.b_state = 0;
1401                 tmp_bh.b_size = sb->s_blocksize;
1402                 err = ext2_get_block(inode, blk, &tmp_bh, 0);
1403                 if (err < 0)
1404                         return err;
1405                 if (!buffer_mapped(&tmp_bh))    /* A hole? */
1406                         memset(data, 0, tocopy);
1407                 else {
1408                         bh = sb_bread(sb, tmp_bh.b_blocknr);
1409                         if (!bh)
1410                                 return -EIO;
1411                         memcpy(data, bh->b_data+offset, tocopy);
1412                         brelse(bh);
1413                 }
1414                 offset = 0;
1415                 toread -= tocopy;
1416                 data += tocopy;
1417                 blk++;
1418         }
1419         return len;
1420 }
1421
1422 /* Write to quotafile */
1423 static ssize_t ext2_quota_write(struct super_block *sb, int type,
1424                                 const char *data, size_t len, loff_t off)
1425 {
1426         struct inode *inode = sb_dqopt(sb)->files[type];
1427         sector_t blk = off >> EXT2_BLOCK_SIZE_BITS(sb);
1428         int err = 0;
1429         int offset = off & (sb->s_blocksize - 1);
1430         int tocopy;
1431         size_t towrite = len;
1432         struct buffer_head tmp_bh;
1433         struct buffer_head *bh;
1434
1435         mutex_lock_nested(&inode->i_mutex, I_MUTEX_QUOTA);
1436         while (towrite > 0) {
1437                 tocopy = sb->s_blocksize - offset < towrite ?
1438                                 sb->s_blocksize - offset : towrite;
1439
1440                 tmp_bh.b_state = 0;
1441                 err = ext2_get_block(inode, blk, &tmp_bh, 1);
1442                 if (err < 0)
1443                         goto out;
1444                 if (offset || tocopy != EXT2_BLOCK_SIZE(sb))
1445                         bh = sb_bread(sb, tmp_bh.b_blocknr);
1446                 else
1447                         bh = sb_getblk(sb, tmp_bh.b_blocknr);
1448                 if (!bh) {
1449                         err = -EIO;
1450                         goto out;
1451                 }
1452                 lock_buffer(bh);
1453                 memcpy(bh->b_data+offset, data, tocopy);
1454                 flush_dcache_page(bh->b_page);
1455                 set_buffer_uptodate(bh);
1456                 mark_buffer_dirty(bh);
1457                 unlock_buffer(bh);
1458                 brelse(bh);
1459                 offset = 0;
1460                 towrite -= tocopy;
1461                 data += tocopy;
1462                 blk++;
1463         }
1464 out:
1465         if (len == towrite) {
1466                 mutex_unlock(&inode->i_mutex);
1467                 return err;
1468         }
1469         if (inode->i_size < off+len-towrite)
1470                 i_size_write(inode, off+len-towrite);
1471         inode->i_version++;
1472         inode->i_mtime = inode->i_ctime = CURRENT_TIME;
1473         mark_inode_dirty(inode);
1474         mutex_unlock(&inode->i_mutex);
1475         return len - towrite;
1476 }
1477
1478 #endif
1479
1480 static struct file_system_type ext2_fs_type = {
1481         .owner          = THIS_MODULE,
1482         .name           = "ext2",
1483         .mount          = ext2_mount,
1484         .kill_sb        = kill_block_super,
1485         .fs_flags       = FS_REQUIRES_DEV,
1486 };
1487
1488 static int __init init_ext2_fs(void)
1489 {
1490         int err = init_ext2_xattr();
1491         if (err)
1492                 return err;
1493         err = init_inodecache();
1494         if (err)
1495                 goto out1;
1496         err = register_filesystem(&ext2_fs_type);
1497         if (err)
1498                 goto out;
1499         return 0;
1500 out:
1501         destroy_inodecache();
1502 out1:
1503         exit_ext2_xattr();
1504         return err;
1505 }
1506
1507 static void __exit exit_ext2_fs(void)
1508 {
1509         unregister_filesystem(&ext2_fs_type);
1510         destroy_inodecache();
1511         exit_ext2_xattr();
1512 }
1513
1514 module_init(init_ext2_fs)
1515 module_exit(exit_ext2_fs)