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
  39 static void ext2_sync_super(struct super_block *sb,
  40                             struct ext2_super_block *es, int wait);
  41 static int ext2_remount (struct super_block * sb, int * flags, char * data);
  42 static int ext2_statfs (struct dentry * dentry, struct kstatfs * buf);
  43 static int ext2_sync_fs(struct super_block *sb, int wait);
  44 static int ext2_freeze(struct super_block *sb);
  45 static int ext2_unfreeze(struct super_block *sb);
  46
  47 void ext2_error(struct super_block *sb, const char *function,
  48                 const char *fmt, ...)
  49 {
  50         struct va_format vaf;
  51         va_list args;
  52         struct ext2_sb_info *sbi = EXT2_SB(sb);
  53         struct ext2_super_block *es = sbi->s_es;
  54
  55         if (!(sb->s_flags & MS_RDONLY)) {
  56                 spin_lock(&sbi->s_lock);
  57                 sbi->s_mount_state |= EXT2_ERROR_FS;
  58                 es->s_state |= cpu_to_le16(EXT2_ERROR_FS);
  59                 spin_unlock(&sbi->s_lock);
  60                 ext2_sync_super(sb, es, 1);
  61         }
  62
  63         va_start(args, fmt);
  64
  65         vaf.fmt = fmt;
  66         vaf.va = &args;
  67
  68         printk(KERN_CRIT "EXT2-fs (%s): error: %s: %pV\n",
  69                sb->s_id, function, &vaf);
  70
  71         va_end(args);
  72
  73         if (test_opt(sb, ERRORS_PANIC))
  74                 panic("EXT2-fs: panic from previous error\n");
  75         if (test_opt(sb, ERRORS_RO)) {
  76                 ext2_msg(sb, KERN_CRIT,
  77                              "error: remounting filesystem read-only");
  78                 sb->s_flags |= MS_RDONLY;
  79         }
  80 }
  81
  82 void ext2_msg(struct super_block *sb, const char *prefix,
  83                 const char *fmt, ...)
  84 {
  85         struct va_format vaf;
  86         va_list args;
  87
  88         va_start(args, fmt);
  89
  90         vaf.fmt = fmt;
  91         vaf.va = &args;
  92
  93         printk("%sEXT2-fs (%s): %pV\n", prefix, sb->s_id, &vaf);
  94
  95         va_end(args);
  96 }
  97
  98 /*
  99  * This must be called with sbi->s_lock held.
 100  */
 101 void ext2_update_dynamic_rev(struct super_block *sb)
 102 {
 103         struct ext2_super_block *es = EXT2_SB(sb)->s_es;
 104
 105         if (le32_to_cpu(es->s_rev_level) > EXT2_GOOD_OLD_REV)
 106                 return;
 107
 108         ext2_msg(sb, KERN_WARNING,
 109                      "warning: updating to rev %d because of "
 110                      "new feature flag, running e2fsck is recommended",
 111                      EXT2_DYNAMIC_REV);
 112
 113         es->s_first_ino = cpu_to_le32(EXT2_GOOD_OLD_FIRST_INO);
 114         es->s_inode_size = cpu_to_le16(EXT2_GOOD_OLD_INODE_SIZE);
 115         es->s_rev_level = cpu_to_le32(EXT2_DYNAMIC_REV);
 116         /* leave es->s_feature_*compat flags alone */
 117         /* es->s_uuid will be set by e2fsck if empty */
 118
 119         /*
 120          * The rest of the superblock fields should be zero, and if not it
 121          * means they are likely already in use, so leave them alone.  We
 122          * can leave it up to e2fsck to clean up any inconsistencies there.
 123          */
 124 }
 125
 126 static void ext2_put_super (struct super_block * sb)
 127 {
 128         int db_count;
 129         int i;
 130         struct ext2_sb_info *sbi = EXT2_SB(sb);
 131
 132         dquot_disable(sb, -1, DQUOT_USAGE_ENABLED | DQUOT_LIMITS_ENABLED);
 133
 134         ext2_xattr_put_super(sb);
 135         if (!(sb->s_flags & MS_RDONLY)) {
 136                 struct ext2_super_block *es = sbi->s_es;
 137
 138                 spin_lock(&sbi->s_lock);
 139                 es->s_state = cpu_to_le16(sbi->s_mount_state);
 140                 spin_unlock(&sbi->s_lock);
 141                 ext2_sync_super(sb, es, 1);
 142         }
 143         db_count = sbi->s_gdb_count;
 144         for (i = 0; i < db_count; i++)
 145                 if (sbi->s_group_desc[i])
 146                         brelse (sbi->s_group_desc[i]);
 147         kfree(sbi->s_group_desc);
 148         kfree(sbi->s_debts);
 149         percpu_counter_destroy(&sbi->s_freeblocks_counter);
 150         percpu_counter_destroy(&sbi->s_freeinodes_counter);
 151         percpu_counter_destroy(&sbi->s_dirs_counter);
 152         brelse (sbi->s_sbh);
 153         sb->s_fs_info = NULL;
 154         kfree(sbi->s_blockgroup_lock);
 155         kfree(sbi);
 156 }
 157
 158 static struct kmem_cache * ext2_inode_cachep;
 159
 160 static struct inode *ext2_alloc_inode(struct super_block *sb)
 161 {
 162         struct ext2_inode_info *ei;
 163         ei = kmem_cache_alloc(ext2_inode_cachep, GFP_KERNEL);
 164         if (!ei)
 165                 return NULL;
 166         ei->i_block_alloc_info = NULL;
 167         ei->vfs_inode.i_version = 1;
 168 #ifdef CONFIG_QUOTA
 169         memset(&ei->i_dquot, 0, sizeof(ei->i_dquot));
 170 #endif
 171
 172         return &ei->vfs_inode;
 173 }
 174
 175 static void ext2_i_callback(struct rcu_head *head)
 176 {
 177         struct inode *inode = container_of(head, struct inode, i_rcu);
 178         kmem_cache_free(ext2_inode_cachep, EXT2_I(inode));
 179 }
 180
 181 static void ext2_destroy_inode(struct inode *inode)
 182 {
 183         call_rcu(&inode->i_rcu, ext2_i_callback);
 184 }
 185
 186 static void init_once(void *foo)
 187 {
 188         struct ext2_inode_info *ei = (struct ext2_inode_info *) foo;
 189
 190         rwlock_init(&ei->i_meta_lock);
 191 #ifdef CONFIG_EXT2_FS_XATTR
 192         init_rwsem(&ei->xattr_sem);
 193 #endif
 194         mutex_init(&ei->truncate_mutex);
 195         inode_init_once(&ei->vfs_inode);
 196 }
 197
 198 static int __init init_inodecache(void)
 199 {
 200         ext2_inode_cachep = kmem_cache_create("ext2_inode_cache",
 201                                              sizeof(struct ext2_inode_info),
 202                                              0, (SLAB_RECLAIM_ACCOUNT|
 203                                                 SLAB_MEM_SPREAD),
 204                                              init_once);
 205         if (ext2_inode_cachep == NULL)
 206                 return -ENOMEM;
 207         return 0;
 208 }
 209
 210 static void destroy_inodecache(void)
 211 {
 212         /*
 213          * Make sure all delayed rcu free inodes are flushed before we
 214          * destroy cache.
 215          */
 216         rcu_barrier();
 217         kmem_cache_destroy(ext2_inode_cachep);
 218 }
 219
 220 static int ext2_show_options(struct seq_file *seq, struct dentry *root)
 221 {
 222         struct super_block *sb = root->d_sb;
 223         struct ext2_sb_info *sbi = EXT2_SB(sb);
 224         struct ext2_super_block *es = sbi->s_es;
 225         unsigned long def_mount_opts;
 226
 227         spin_lock(&sbi->s_lock);
 228         def_mount_opts = le32_to_cpu(es->s_default_mount_opts);
 229
 230         if (sbi->s_sb_block != 1)
 231                 seq_printf(seq, ",sb=%lu", sbi->s_sb_block);
 232         if (test_opt(sb, MINIX_DF))
 233                 seq_puts(seq, ",minixdf");
 234         if (test_opt(sb, GRPID))
 235                 seq_puts(seq, ",grpid");
 236         if (!test_opt(sb, GRPID) && (def_mount_opts & EXT2_DEFM_BSDGROUPS))
 237                 seq_puts(seq, ",nogrpid");
 238         if (!uid_eq(sbi->s_resuid, make_kuid(&init_user_ns, EXT2_DEF_RESUID)) ||
 239             le16_to_cpu(es->s_def_resuid) != EXT2_DEF_RESUID) {
 240                 seq_printf(seq, ",resuid=%u",
 241                                 from_kuid_munged(&init_user_ns, sbi->s_resuid));
 242         }
 243         if (!gid_eq(sbi->s_resgid, make_kgid(&init_user_ns, EXT2_DEF_RESGID)) ||
 244             le16_to_cpu(es->s_def_resgid) != EXT2_DEF_RESGID) {
 245                 seq_printf(seq, ",resgid=%u",
 246                                 from_kgid_munged(&init_user_ns, sbi->s_resgid));
 247         }
 248         if (test_opt(sb, ERRORS_RO)) {
 249                 int def_errors = le16_to_cpu(es->s_errors);
 250
 251                 if (def_errors == EXT2_ERRORS_PANIC ||
 252                     def_errors == EXT2_ERRORS_CONTINUE) {
 253                         seq_puts(seq, ",errors=remount-ro");
 254                 }
 255         }
 256         if (test_opt(sb, ERRORS_CONT))
 257                 seq_puts(seq, ",errors=continue");
 258         if (test_opt(sb, ERRORS_PANIC))
 259                 seq_puts(seq, ",errors=panic");
 260         if (test_opt(sb, NO_UID32))
 261                 seq_puts(seq, ",nouid32");
 262         if (test_opt(sb, DEBUG))
 263                 seq_puts(seq, ",debug");
 264         if (test_opt(sb, OLDALLOC))
 265                 seq_puts(seq, ",oldalloc");
 266
 267 #ifdef CONFIG_EXT2_FS_XATTR
 268         if (test_opt(sb, XATTR_USER))
 269                 seq_puts(seq, ",user_xattr");
 270         if (!test_opt(sb, XATTR_USER) &&
 271             (def_mount_opts & EXT2_DEFM_XATTR_USER)) {
 272                 seq_puts(seq, ",nouser_xattr");
 273         }
 274 #endif
 275
 276 #ifdef CONFIG_EXT2_FS_POSIX_ACL
 277         if (test_opt(sb, POSIX_ACL))
 278                 seq_puts(seq, ",acl");
 279         if (!test_opt(sb, POSIX_ACL) && (def_mount_opts & EXT2_DEFM_ACL))
 280                 seq_puts(seq, ",noacl");
 281 #endif
 282
 283         if (test_opt(sb, NOBH))
 284                 seq_puts(seq, ",nobh");
 285
 286 #if defined(CONFIG_QUOTA)
 287         if (sbi->s_mount_opt & EXT2_MOUNT_USRQUOTA)
 288                 seq_puts(seq, ",usrquota");
 289
 290         if (sbi->s_mount_opt & EXT2_MOUNT_GRPQUOTA)
 291                 seq_puts(seq, ",grpquota");
 292 #endif
 293
 294 #if defined(CONFIG_EXT2_FS_XIP)
 295         if (sbi->s_mount_opt & EXT2_MOUNT_XIP)
 296                 seq_puts(seq, ",xip");
 297 #endif
 298
 299         if (!test_opt(sb, RESERVATION))
 300                 seq_puts(seq, ",noreservation");
 301
 302         spin_unlock(&sbi->s_lock);
 303         return 0;
 304 }
 305
 306 #ifdef CONFIG_QUOTA
 307 static ssize_t ext2_quota_read(struct super_block *sb, int type, char *data, size_t len, loff_t off);
 308 static ssize_t ext2_quota_write(struct super_block *sb, int type, const char *data, size_t len, loff_t off);
 309 static struct dquot **ext2_get_dquots(struct inode *inode)
 310 {
 311         return EXT2_I(inode)->i_dquot;
 312 }
 313 #endif
 314
 315 static const struct super_operations ext2_sops = {
 316         .alloc_inode    = ext2_alloc_inode,
 317         .destroy_inode  = ext2_destroy_inode,
 318         .write_inode    = ext2_write_inode,
 319         .evict_inode    = ext2_evict_inode,
 320         .put_super      = ext2_put_super,
 321         .sync_fs        = ext2_sync_fs,
 322         .freeze_fs      = ext2_freeze,
 323         .unfreeze_fs    = ext2_unfreeze,
 324         .statfs         = ext2_statfs,
 325         .remount_fs     = ext2_remount,
 326         .show_options   = ext2_show_options,
 327 #ifdef CONFIG_QUOTA
 328         .quota_read     = ext2_quota_read,
 329         .quota_write    = ext2_quota_write,
 330         .get_dquots     = ext2_get_dquots,
 331 #endif
 332 };
 333
 334 static struct inode *ext2_nfs_get_inode(struct super_block *sb,
 335                 u64 ino, u32 generation)
 336 {
 337         struct inode *inode;
 338
 339         if (ino < EXT2_FIRST_INO(sb) && ino != EXT2_ROOT_INO)
 340                 return ERR_PTR(-ESTALE);
 341         if (ino > le32_to_cpu(EXT2_SB(sb)->s_es->s_inodes_count))
 342                 return ERR_PTR(-ESTALE);
 343
 344         /*
 345          * ext2_iget isn't quite right if the inode is currently unallocated!
 346          * However ext2_iget currently does appropriate checks to handle stale
 347          * inodes so everything is OK.
 348          */
 349         inode = ext2_iget(sb, ino);
 350         if (IS_ERR(inode))
 351                 return ERR_CAST(inode);
 352         if (generation && inode->i_generation != generation) {
 353                 /* we didn't find the right inode.. */
 354                 iput(inode);
 355                 return ERR_PTR(-ESTALE);
 356         }
 357         return inode;
 358 }
 359
 360 static struct dentry *ext2_fh_to_dentry(struct super_block *sb, struct fid *fid,
 361                 int fh_len, int fh_type)
 362 {
 363         return generic_fh_to_dentry(sb, fid, fh_len, fh_type,
 364                                     ext2_nfs_get_inode);
 365 }
 366
 367 static struct dentry *ext2_fh_to_parent(struct super_block *sb, struct fid *fid,
 368                 int fh_len, int fh_type)
 369 {
 370         return generic_fh_to_parent(sb, fid, fh_len, fh_type,
 371                                     ext2_nfs_get_inode);
 372 }
 373
 374 static const struct export_operations ext2_export_ops = {
 375         .fh_to_dentry = ext2_fh_to_dentry,
 376         .fh_to_parent = ext2_fh_to_parent,
 377         .get_parent = ext2_get_parent,
 378 };
 379
 380 static unsigned long get_sb_block(void **data)
 381 {
 382         unsigned long   sb_block;
 383         char            *options = (char *) *data;
 384
 385         if (!options || strncmp(options, "sb=", 3) != 0)
 386                 return 1;       /* Default location */
 387         options += 3;
 388         sb_block = simple_strtoul(options, &options, 0);
 389         if (*options && *options != ',') {
 390                 printk("EXT2-fs: Invalid sb specification: %s\n",
 391                        (char *) *data);
 392                 return 1;
 393         }
 394         if (*options == ',')
 395                 options++;
 396         *data = (void *) options;
 397         return sb_block;
 398 }
 399
 400 enum {
 401         Opt_bsd_df, Opt_minix_df, Opt_grpid, Opt_nogrpid,
 402         Opt_resgid, Opt_resuid, Opt_sb, Opt_err_cont, Opt_err_panic,
 403         Opt_err_ro, Opt_nouid32, Opt_nocheck, Opt_debug,
 404         Opt_oldalloc, Opt_orlov, Opt_nobh, Opt_user_xattr, Opt_nouser_xattr,
 405         Opt_acl, Opt_noacl, Opt_xip, Opt_ignore, Opt_err, Opt_quota,
 406         Opt_usrquota, Opt_grpquota, Opt_reservation, Opt_noreservation
 407 };
 408
 409 static const match_table_t tokens = {
 410         {Opt_bsd_df, "bsddf"},
 411         {Opt_minix_df, "minixdf"},
 412         {Opt_grpid, "grpid"},
 413         {Opt_grpid, "bsdgroups"},
 414         {Opt_nogrpid, "nogrpid"},
 415         {Opt_nogrpid, "sysvgroups"},
 416         {Opt_resgid, "resgid=%u"},
 417         {Opt_resuid, "resuid=%u"},
 418         {Opt_sb, "sb=%u"},
 419         {Opt_err_cont, "errors=continue"},
 420         {Opt_err_panic, "errors=panic"},
 421         {Opt_err_ro, "errors=remount-ro"},
 422         {Opt_nouid32, "nouid32"},
 423         {Opt_nocheck, "check=none"},
 424         {Opt_nocheck, "nocheck"},
 425         {Opt_debug, "debug"},
 426         {Opt_oldalloc, "oldalloc"},
 427         {Opt_orlov, "orlov"},
 428         {Opt_nobh, "nobh"},
 429         {Opt_user_xattr, "user_xattr"},
 430         {Opt_nouser_xattr, "nouser_xattr"},
 431         {Opt_acl, "acl"},
 432         {Opt_noacl, "noacl"},
 433         {Opt_xip, "xip"},
 434         {Opt_grpquota, "grpquota"},
 435         {Opt_ignore, "noquota"},
 436         {Opt_quota, "quota"},
 437         {Opt_usrquota, "usrquota"},
 438         {Opt_reservation, "reservation"},
 439         {Opt_noreservation, "noreservation"},
 440         {Opt_err, NULL}
 441 };
 442
 443 static int parse_options(char *options, struct super_block *sb)
 444 {
 445         char *p;
 446         struct ext2_sb_info *sbi = EXT2_SB(sb);
 447         substring_t args[MAX_OPT_ARGS];
 448         int option;
 449         kuid_t uid;
 450         kgid_t gid;
 451
 452         if (!options)
 453                 return 1;
 454
 455         while ((p = strsep (&options, ",")) != NULL) {
 456                 int token;
 457                 if (!*p)
 458                         continue;
 459
 460                 token = match_token(p, tokens, args);
 461                 switch (token) {
 462                 case Opt_bsd_df:
 463                         clear_opt (sbi->s_mount_opt, MINIX_DF);
 464                         break;
 465                 case Opt_minix_df:
 466                         set_opt (sbi->s_mount_opt, MINIX_DF);
 467                         break;
 468                 case Opt_grpid:
 469                         set_opt (sbi->s_mount_opt, GRPID);
 470                         break;
 471                 case Opt_nogrpid:
 472                         clear_opt (sbi->s_mount_opt, GRPID);
 473                         break;
 474                 case Opt_resuid:
 475                         if (match_int(&args[0], &option))
 476                                 return 0;
 477                         uid = make_kuid(current_user_ns(), option);
 478                         if (!uid_valid(uid)) {
 479                                 ext2_msg(sb, KERN_ERR, "Invalid uid value %d", option);
 480                                 return 0;
 481
 482                         }
 483                         sbi->s_resuid = uid;
 484                         break;
 485                 case Opt_resgid:
 486                         if (match_int(&args[0], &option))
 487                                 return 0;
 488                         gid = make_kgid(current_user_ns(), option);
 489                         if (!gid_valid(gid)) {
 490                                 ext2_msg(sb, KERN_ERR, "Invalid gid value %d", option);
 491                                 return 0;
 492                         }
 493                         sbi->s_resgid = gid;
 494                         break;
 495                 case Opt_sb:
 496                         /* handled by get_sb_block() instead of here */
 497                         /* *sb_block = match_int(&args[0]); */
 498                         break;
 499                 case Opt_err_panic:
 500                         clear_opt (sbi->s_mount_opt, ERRORS_CONT);
 501                         clear_opt (sbi->s_mount_opt, ERRORS_RO);
 502                         set_opt (sbi->s_mount_opt, ERRORS_PANIC);
 503                         break;
 504                 case Opt_err_ro:
 505                         clear_opt (sbi->s_mount_opt, ERRORS_CONT);
 506                         clear_opt (sbi->s_mount_opt, ERRORS_PANIC);
 507                         set_opt (sbi->s_mount_opt, ERRORS_RO);
 508                         break;
 509                 case Opt_err_cont:
 510                         clear_opt (sbi->s_mount_opt, ERRORS_RO);
 511                         clear_opt (sbi->s_mount_opt, ERRORS_PANIC);
 512                         set_opt (sbi->s_mount_opt, ERRORS_CONT);
 513                         break;
 514                 case Opt_nouid32:
 515                         set_opt (sbi->s_mount_opt, NO_UID32);
 516                         break;
 517                 case Opt_nocheck:
 518                         clear_opt (sbi->s_mount_opt, CHECK);
 519                         break;
 520                 case Opt_debug:
 521                         set_opt (sbi->s_mount_opt, DEBUG);
 522                         break;
 523                 case Opt_oldalloc:
 524                         set_opt (sbi->s_mount_opt, OLDALLOC);
 525                         break;
 526                 case Opt_orlov:
 527                         clear_opt (sbi->s_mount_opt, OLDALLOC);
 528                         break;
 529                 case Opt_nobh:
 530                         set_opt (sbi->s_mount_opt, NOBH);
 531                         break;
 532 #ifdef CONFIG_EXT2_FS_XATTR
 533                 case Opt_user_xattr:
 534                         set_opt (sbi->s_mount_opt, XATTR_USER);
 535                         break;
 536                 case Opt_nouser_xattr:
 537                         clear_opt (sbi->s_mount_opt, XATTR_USER);
 538                         break;
 539 #else
 540                 case Opt_user_xattr:
 541                 case Opt_nouser_xattr:
 542                         ext2_msg(sb, KERN_INFO, "(no)user_xattr options"
 543                                 "not supported");
 544                         break;
 545 #endif
 546 #ifdef CONFIG_EXT2_FS_POSIX_ACL
 547                 case Opt_acl:
 548                         set_opt(sbi->s_mount_opt, POSIX_ACL);
 549                         break;
 550                 case Opt_noacl:
 551                         clear_opt(sbi->s_mount_opt, POSIX_ACL);
 552                         break;
 553 #else
 554                 case Opt_acl:
 555                 case Opt_noacl:
 556                         ext2_msg(sb, KERN_INFO,
 557                                 "(no)acl options not supported");
 558                         break;
 559 #endif
 560                 case Opt_xip:
 561 #ifdef CONFIG_EXT2_FS_XIP
 562                         set_opt (sbi->s_mount_opt, XIP);
 563 #else
 564                         ext2_msg(sb, KERN_INFO, "xip option not supported");
 565 #endif
 566                         break;
 567
 568 #if defined(CONFIG_QUOTA)
 569                 case Opt_quota:
 570                 case Opt_usrquota:
 571                         set_opt(sbi->s_mount_opt, USRQUOTA);
 572                         break;
 573
 574                 case Opt_grpquota:
 575                         set_opt(sbi->s_mount_opt, GRPQUOTA);
 576                         break;
 577 #else
 578                 case Opt_quota:
 579                 case Opt_usrquota:
 580                 case Opt_grpquota:
 581                         ext2_msg(sb, KERN_INFO,
 582                                 "quota operations not supported");
 583                         break;
 584 #endif
 585
 586                 case Opt_reservation:
 587                         set_opt(sbi->s_mount_opt, RESERVATION);
 588                         ext2_msg(sb, KERN_INFO, "reservations ON");
 589                         break;
 590                 case Opt_noreservation:
 591                         clear_opt(sbi->s_mount_opt, RESERVATION);
 592                         ext2_msg(sb, KERN_INFO, "reservations OFF");
 593                         break;
 594                 case Opt_ignore:
 595                         break;
 596                 default:
 597                         return 0;
 598                 }
 599         }
 600         return 1;
 601 }
 602
 603 static int ext2_setup_super (struct super_block * sb,
 604                               struct ext2_super_block * es,
 605                               int read_only)
 606 {
 607         int res = 0;
 608         struct ext2_sb_info *sbi = EXT2_SB(sb);
 609
 610         if (le32_to_cpu(es->s_rev_level) > EXT2_MAX_SUPP_REV) {
 611                 ext2_msg(sb, KERN_ERR,
 612                         "error: revision level too high, "
 613                         "forcing read-only mode");
 614                 res = MS_RDONLY;
 615         }
 616         if (read_only)
 617                 return res;
 618         if (!(sbi->s_mount_state & EXT2_VALID_FS))
 619                 ext2_msg(sb, KERN_WARNING,
 620                         "warning: mounting unchecked fs, "
 621                         "running e2fsck is recommended");
 622         else if ((sbi->s_mount_state & EXT2_ERROR_FS))
 623                 ext2_msg(sb, KERN_WARNING,
 624                         "warning: mounting fs with errors, "
 625                         "running e2fsck is recommended");
 626         else if ((__s16) le16_to_cpu(es->s_max_mnt_count) >= 0 &&
 627                  le16_to_cpu(es->s_mnt_count) >=
 628                  (unsigned short) (__s16) le16_to_cpu(es->s_max_mnt_count))
 629                 ext2_msg(sb, KERN_WARNING,
 630                         "warning: maximal mount count reached, "
 631                         "running e2fsck is recommended");
 632         else if (le32_to_cpu(es->s_checkinterval) &&
 633                 (le32_to_cpu(es->s_lastcheck) +
 634                         le32_to_cpu(es->s_checkinterval) <= get_seconds()))
 635                 ext2_msg(sb, KERN_WARNING,
 636                         "warning: checktime reached, "
 637                         "running e2fsck is recommended");
 638         if (!le16_to_cpu(es->s_max_mnt_count))
 639                 es->s_max_mnt_count = cpu_to_le16(EXT2_DFL_MAX_MNT_COUNT);
 640         le16_add_cpu(&es->s_mnt_count, 1);
 641         if (test_opt (sb, DEBUG))
 642                 ext2_msg(sb, KERN_INFO, "%s, %s, bs=%lu, fs=%lu, gc=%lu, "
 643                         "bpg=%lu, ipg=%lu, mo=%04lx]",
 644                         EXT2FS_VERSION, EXT2FS_DATE, sb->s_blocksize,
 645                         sbi->s_frag_size,
 646                         sbi->s_groups_count,
 647                         EXT2_BLOCKS_PER_GROUP(sb),
 648                         EXT2_INODES_PER_GROUP(sb),
 649                         sbi->s_mount_opt);
 650         return res;
 651 }
 652
 653 static int ext2_check_descriptors(struct super_block *sb)
 654 {
 655         int i;
 656         struct ext2_sb_info *sbi = EXT2_SB(sb);
 657
 658         ext2_debug ("Checking group descriptors");
 659
 660         for (i = 0; i < sbi->s_groups_count; i++) {
 661                 struct ext2_group_desc *gdp = ext2_get_group_desc(sb, i, NULL);
 662                 ext2_fsblk_t first_block = ext2_group_first_block_no(sb, i);
 663                 ext2_fsblk_t last_block;
 664
 665                 if (i == sbi->s_groups_count - 1)
 666                         last_block = le32_to_cpu(sbi->s_es->s_blocks_count) - 1;
 667                 else
 668                         last_block = first_block +
 669                                 (EXT2_BLOCKS_PER_GROUP(sb) - 1);
 670
 671                 if (le32_to_cpu(gdp->bg_block_bitmap) < first_block ||
 672                     le32_to_cpu(gdp->bg_block_bitmap) > last_block)
 673                 {
 674                         ext2_error (sb, "ext2_check_descriptors",
 675                                     "Block bitmap for group %d"
 676                                     " not in group (block %lu)!",
 677                                     i, (unsigned long) le32_to_cpu(gdp->bg_block_bitmap));
 678                         return 0;
 679                 }
 680                 if (le32_to_cpu(gdp->bg_inode_bitmap) < first_block ||
 681                     le32_to_cpu(gdp->bg_inode_bitmap) > last_block)
 682                 {
 683                         ext2_error (sb, "ext2_check_descriptors",
 684                                     "Inode bitmap for group %d"
 685                                     " not in group (block %lu)!",
 686                                     i, (unsigned long) le32_to_cpu(gdp->bg_inode_bitmap));
 687                         return 0;
 688                 }
 689                 if (le32_to_cpu(gdp->bg_inode_table) < first_block ||
 690                     le32_to_cpu(gdp->bg_inode_table) + sbi->s_itb_per_group - 1 >
 691                     last_block)
 692                 {
 693                         ext2_error (sb, "ext2_check_descriptors",
 694                                     "Inode table for group %d"
 695                                     " not in group (block %lu)!",
 696                                     i, (unsigned long) le32_to_cpu(gdp->bg_inode_table));
 697                         return 0;
 698                 }
 699         }
 700         return 1;
 701 }
 702
 703 /*
 704  * Maximal file size.  There is a direct, and {,double-,triple-}indirect
 705  * block limit, and also a limit of (2^32 - 1) 512-byte sectors in i_blocks.
 706  * We need to be 1 filesystem block less than the 2^32 sector limit.
 707  */
 708 static loff_t ext2_max_size(int bits)
 709 {
 710         loff_t res = EXT2_NDIR_BLOCKS;
 711         int meta_blocks;
 712         loff_t upper_limit;
 713
 714         /* This is calculated to be the largest file size for a
 715          * dense, file such that the total number of
 716          * sectors in the file, including data and all indirect blocks,
 717          * does not exceed 2^32 -1
 718          * __u32 i_blocks representing the total number of
 719          * 512 bytes blocks of the file
 720          */
 721         upper_limit = (1LL << 32) - 1;
 722
 723         /* total blocks in file system block size */
 724         upper_limit >>= (bits - 9);
 725
 726
 727         /* indirect blocks */
 728         meta_blocks = 1;
 729         /* double indirect blocks */
 730         meta_blocks += 1 + (1LL << (bits-2));
 731         /* tripple indirect blocks */
 732         meta_blocks += 1 + (1LL << (bits-2)) + (1LL << (2*(bits-2)));
 733
 734         upper_limit -= meta_blocks;
 735         upper_limit <<= bits;
 736
 737         res += 1LL << (bits-2);
 738         res += 1LL << (2*(bits-2));
 739         res += 1LL << (3*(bits-2));
 740         res <<= bits;
 741         if (res > upper_limit)
 742                 res = upper_limit;
 743
 744         if (res > MAX_LFS_FILESIZE)
 745                 res = MAX_LFS_FILESIZE;
 746
 747         return res;
 748 }
 749
 750 static unsigned long descriptor_loc(struct super_block *sb,
 751                                     unsigned long logic_sb_block,
 752                                     int nr)
 753 {
 754         struct ext2_sb_info *sbi = EXT2_SB(sb);
 755         unsigned long bg, first_meta_bg;
 756         int has_super = 0;
 757
 758         first_meta_bg = le32_to_cpu(sbi->s_es->s_first_meta_bg);
 759
 760         if (!EXT2_HAS_INCOMPAT_FEATURE(sb, EXT2_FEATURE_INCOMPAT_META_BG) ||
 761             nr < first_meta_bg)
 762                 return (logic_sb_block + nr + 1);
 763         bg = sbi->s_desc_per_block * nr;
 764         if (ext2_bg_has_super(sb, bg))
 765                 has_super = 1;
 766
 767         return ext2_group_first_block_no(sb, bg) + has_super;
 768 }
 769
 770 static int ext2_fill_super(struct super_block *sb, void *data, int silent)
 771 {
 772         struct buffer_head * bh;
 773         struct ext2_sb_info * sbi;
 774         struct ext2_super_block * es;
 775         struct inode *root;
 776         unsigned long block;
 777         unsigned long sb_block = get_sb_block(&data);
 778         unsigned long logic_sb_block;
 779         unsigned long offset = 0;
 780         unsigned long def_mount_opts;
 781         long ret = -EINVAL;
 782         int blocksize = BLOCK_SIZE;
 783         int db_count;
 784         int i, j;
 785         __le32 features;
 786         int err;
 787
 788         err = -ENOMEM;
 789         sbi = kzalloc(sizeof(*sbi), GFP_KERNEL);
 790         if (!sbi)
 791                 goto failed;
 792
 793         sbi->s_blockgroup_lock =
 794                 kzalloc(sizeof(struct blockgroup_lock), GFP_KERNEL);
 795         if (!sbi->s_blockgroup_lock) {
 796                 kfree(sbi);
 797                 goto failed;
 798         }
 799         sb->s_fs_info = sbi;
 800         sbi->s_sb_block = sb_block;
 801
 802         spin_lock_init(&sbi->s_lock);
 803
 804         /*
 805          * See what the current blocksize for the device is, and
 806          * use that as the blocksize.  Otherwise (or if the blocksize
 807          * is smaller than the default) use the default.
 808          * This is important for devices that have a hardware
 809          * sectorsize that is larger than the default.
 810          */
 811         blocksize = sb_min_blocksize(sb, BLOCK_SIZE);
 812         if (!blocksize) {
 813                 ext2_msg(sb, KERN_ERR, "error: unable to set blocksize");
 814                 goto failed_sbi;
 815         }
 816
 817         /*
 818          * If the superblock doesn't start on a hardware sector boundary,
 819          * calculate the offset.
 820          */
 821         if (blocksize != BLOCK_SIZE) {
 822                 logic_sb_block = (sb_block*BLOCK_SIZE) / blocksize;
 823                 offset = (sb_block*BLOCK_SIZE) % blocksize;
 824         } else {
 825                 logic_sb_block = sb_block;
 826         }
 827
 828         if (!(bh = sb_bread(sb, logic_sb_block))) {
 829                 ext2_msg(sb, KERN_ERR, "error: unable to read superblock");
 830                 goto failed_sbi;
 831         }
 832         /*
 833          * Note: s_es must be initialized as soon as possible because
 834          *       some ext2 macro-instructions depend on its value
 835          */
 836         es = (struct ext2_super_block *) (((char *)bh->b_data) + offset);
 837         sbi->s_es = es;
 838         sb->s_magic = le16_to_cpu(es->s_magic);
 839
 840         if (sb->s_magic != EXT2_SUPER_MAGIC)
 841                 goto cantfind_ext2;
 842
 843         /* Set defaults before we parse the mount options */
 844         def_mount_opts = le32_to_cpu(es->s_default_mount_opts);
 845         if (def_mount_opts & EXT2_DEFM_DEBUG)
 846                 set_opt(sbi->s_mount_opt, DEBUG);
 847         if (def_mount_opts & EXT2_DEFM_BSDGROUPS)
 848                 set_opt(sbi->s_mount_opt, GRPID);
 849         if (def_mount_opts & EXT2_DEFM_UID16)
 850                 set_opt(sbi->s_mount_opt, NO_UID32);
 851 #ifdef CONFIG_EXT2_FS_XATTR
 852         if (def_mount_opts & EXT2_DEFM_XATTR_USER)
 853                 set_opt(sbi->s_mount_opt, XATTR_USER);
 854 #endif
 855 #ifdef CONFIG_EXT2_FS_POSIX_ACL
 856         if (def_mount_opts & EXT2_DEFM_ACL)
 857                 set_opt(sbi->s_mount_opt, POSIX_ACL);
 858 #endif
 859
 860         if (le16_to_cpu(sbi->s_es->s_errors) == EXT2_ERRORS_PANIC)
 861                 set_opt(sbi->s_mount_opt, ERRORS_PANIC);
 862         else if (le16_to_cpu(sbi->s_es->s_errors) == EXT2_ERRORS_CONTINUE)
 863                 set_opt(sbi->s_mount_opt, ERRORS_CONT);
 864         else
 865                 set_opt(sbi->s_mount_opt, ERRORS_RO);
 866
 867         sbi->s_resuid = make_kuid(&init_user_ns, le16_to_cpu(es->s_def_resuid));
 868         sbi->s_resgid = make_kgid(&init_user_ns, le16_to_cpu(es->s_def_resgid));
 869
 870         set_opt(sbi->s_mount_opt, RESERVATION);
 871
 872         if (!parse_options((char *) data, sb))
 873                 goto failed_mount;
 874
 875         sb->s_flags = (sb->s_flags & ~MS_POSIXACL) |
 876                 ((EXT2_SB(sb)->s_mount_opt & EXT2_MOUNT_POSIX_ACL) ?
 877                  MS_POSIXACL : 0);
 878
 879         if (le32_to_cpu(es->s_rev_level) == EXT2_GOOD_OLD_REV &&
 880             (EXT2_HAS_COMPAT_FEATURE(sb, ~0U) ||
 881              EXT2_HAS_RO_COMPAT_FEATURE(sb, ~0U) ||
 882              EXT2_HAS_INCOMPAT_FEATURE(sb, ~0U)))
 883                 ext2_msg(sb, KERN_WARNING,
 884                         "warning: feature flags set on rev 0 fs, "
 885                         "running e2fsck is recommended");
 886         /*
 887          * Check feature flags regardless of the revision level, since we
 888          * previously didn't change the revision level when setting the flags,
 889          * so there is a chance incompat flags are set on a rev 0 filesystem.
 890          */
 891         features = EXT2_HAS_INCOMPAT_FEATURE(sb, ~EXT2_FEATURE_INCOMPAT_SUPP);
 892         if (features) {
 893                 ext2_msg(sb, KERN_ERR,  "error: couldn't mount because of "
 894                        "unsupported optional features (%x)",
 895                         le32_to_cpu(features));
 896                 goto failed_mount;
 897         }
 898         if (!(sb->s_flags & MS_RDONLY) &&
 899             (features = EXT2_HAS_RO_COMPAT_FEATURE(sb, ~EXT2_FEATURE_RO_COMPAT_SUPP))){
 900                 ext2_msg(sb, KERN_ERR, "error: couldn't mount RDWR because of "
 901                        "unsupported optional features (%x)",
 902                        le32_to_cpu(features));
 903                 goto failed_mount;
 904         }
 905
 906         blocksize = BLOCK_SIZE << le32_to_cpu(sbi->s_es->s_log_block_size);
 907
 908         if (sbi->s_mount_opt & EXT2_MOUNT_XIP) {
 909                 if (blocksize != PAGE_SIZE) {
 910                         ext2_msg(sb, KERN_ERR,
 911                                         "error: unsupported blocksize for xip");
 912                         goto failed_mount;
 913                 }
 914                 if (!sb->s_bdev->bd_disk->fops->direct_access) {
 915                         ext2_msg(sb, KERN_ERR,
 916                                         "error: device does not support xip");
 917                         goto failed_mount;
 918                 }
 919         }
 920
 921         /* If the blocksize doesn't match, re-read the thing.. */
 922         if (sb->s_blocksize != blocksize) {
 923                 brelse(bh);
 924
 925                 if (!sb_set_blocksize(sb, blocksize)) {
 926                         ext2_msg(sb, KERN_ERR,
 927                                 "error: bad blocksize %d", blocksize);
 928                         goto failed_sbi;
 929                 }
 930
 931                 logic_sb_block = (sb_block*BLOCK_SIZE) / blocksize;
 932                 offset = (sb_block*BLOCK_SIZE) % blocksize;
 933                 bh = sb_bread(sb, logic_sb_block);
 934                 if(!bh) {
 935                         ext2_msg(sb, KERN_ERR, "error: couldn't read"
 936                                 "superblock on 2nd try");
 937                         goto failed_sbi;
 938                 }
 939                 es = (struct ext2_super_block *) (((char *)bh->b_data) + offset);
 940                 sbi->s_es = es;
 941                 if (es->s_magic != cpu_to_le16(EXT2_SUPER_MAGIC)) {
 942                         ext2_msg(sb, KERN_ERR, "error: magic mismatch");
 943                         goto failed_mount;
 944                 }
 945         }
 946
 947         sb->s_maxbytes = ext2_max_size(sb->s_blocksize_bits);
 948         sb->s_max_links = EXT2_LINK_MAX;
 949
 950         if (le32_to_cpu(es->s_rev_level) == EXT2_GOOD_OLD_REV) {
 951                 sbi->s_inode_size = EXT2_GOOD_OLD_INODE_SIZE;
 952                 sbi->s_first_ino = EXT2_GOOD_OLD_FIRST_INO;
 953         } else {
 954                 sbi->s_inode_size = le16_to_cpu(es->s_inode_size);
 955                 sbi->s_first_ino = le32_to_cpu(es->s_first_ino);
 956                 if ((sbi->s_inode_size < EXT2_GOOD_OLD_INODE_SIZE) ||
 957                     !is_power_of_2(sbi->s_inode_size) ||
 958                     (sbi->s_inode_size > blocksize)) {
 959                         ext2_msg(sb, KERN_ERR,
 960                                 "error: unsupported inode size: %d",
 961                                 sbi->s_inode_size);
 962                         goto failed_mount;
 963                 }
 964         }
 965
 966         sbi->s_frag_size = EXT2_MIN_FRAG_SIZE <<
 967                                    le32_to_cpu(es->s_log_frag_size);
 968         if (sbi->s_frag_size == 0)
 969                 goto cantfind_ext2;
 970         sbi->s_frags_per_block = sb->s_blocksize / sbi->s_frag_size;
 971
 972         sbi->s_blocks_per_group = le32_to_cpu(es->s_blocks_per_group);
 973         sbi->s_frags_per_group = le32_to_cpu(es->s_frags_per_group);
 974         sbi->s_inodes_per_group = le32_to_cpu(es->s_inodes_per_group);
 975
 976         if (EXT2_INODE_SIZE(sb) == 0)
 977                 goto cantfind_ext2;
 978         sbi->s_inodes_per_block = sb->s_blocksize / EXT2_INODE_SIZE(sb);
 979         if (sbi->s_inodes_per_block == 0 || sbi->s_inodes_per_group == 0)
 980                 goto cantfind_ext2;
 981         sbi->s_itb_per_group = sbi->s_inodes_per_group /
 982                                         sbi->s_inodes_per_block;
 983         sbi->s_desc_per_block = sb->s_blocksize /
 984                                         sizeof (struct ext2_group_desc);
 985         sbi->s_sbh = bh;
 986         sbi->s_mount_state = le16_to_cpu(es->s_state);
 987         sbi->s_addr_per_block_bits =
 988                 ilog2 (EXT2_ADDR_PER_BLOCK(sb));
 989         sbi->s_desc_per_block_bits =
 990                 ilog2 (EXT2_DESC_PER_BLOCK(sb));
 991
 992         if (sb->s_magic != EXT2_SUPER_MAGIC)
 993                 goto cantfind_ext2;
 994
 995         if (sb->s_blocksize != bh->b_size) {
 996                 if (!silent)
 997                         ext2_msg(sb, KERN_ERR, "error: unsupported blocksize");
 998                 goto failed_mount;
 999         }
1000
1001         if (sb->s_blocksize != sbi->s_frag_size) {
1002                 ext2_msg(sb, KERN_ERR,
1003                         "error: fragsize %lu != blocksize %lu"
1004                         "(not supported yet)",
1005                         sbi->s_frag_size, sb->s_blocksize);
1006                 goto failed_mount;
1007         }
1008
1009         if (sbi->s_blocks_per_group > sb->s_blocksize * 8) {
1010                 ext2_msg(sb, KERN_ERR,
1011                         "error: #blocks per group too big: %lu",
1012                         sbi->s_blocks_per_group);
1013                 goto failed_mount;
1014         }
1015         if (sbi->s_frags_per_group > sb->s_blocksize * 8) {
1016                 ext2_msg(sb, KERN_ERR,
1017                         "error: #fragments per group too big: %lu",
1018                         sbi->s_frags_per_group);
1019                 goto failed_mount;
1020         }
1021         if (sbi->s_inodes_per_group > sb->s_blocksize * 8) {
1022                 ext2_msg(sb, KERN_ERR,
1023                         "error: #inodes per group too big: %lu",
1024                         sbi->s_inodes_per_group);
1025                 goto failed_mount;
1026         }
1027
1028         if (EXT2_BLOCKS_PER_GROUP(sb) == 0)
1029                 goto cantfind_ext2;
1030         sbi->s_groups_count = ((le32_to_cpu(es->s_blocks_count) -
1031                                 le32_to_cpu(es->s_first_data_block) - 1)
1032                                         / EXT2_BLOCKS_PER_GROUP(sb)) + 1;
1033         db_count = (sbi->s_groups_count + EXT2_DESC_PER_BLOCK(sb) - 1) /
1034                    EXT2_DESC_PER_BLOCK(sb);
1035         sbi->s_group_desc = kmalloc (db_count * sizeof (struct buffer_head *), GFP_KERNEL);
1036         if (sbi->s_group_desc == NULL) {
1037                 ext2_msg(sb, KERN_ERR, "error: not enough memory");
1038                 goto failed_mount;
1039         }
1040         bgl_lock_init(sbi->s_blockgroup_lock);
1041         sbi->s_debts = kcalloc(sbi->s_groups_count, sizeof(*sbi->s_debts), GFP_KERNEL);
1042         if (!sbi->s_debts) {
1043                 ext2_msg(sb, KERN_ERR, "error: not enough memory");
1044                 goto failed_mount_group_desc;
1045         }
1046         for (i = 0; i < db_count; i++) {
1047                 block = descriptor_loc(sb, logic_sb_block, i);
1048                 sbi->s_group_desc[i] = sb_bread(sb, block);
1049                 if (!sbi->s_group_desc[i]) {
1050                         for (j = 0; j < i; j++)
1051                                 brelse (sbi->s_group_desc[j]);
1052                         ext2_msg(sb, KERN_ERR,
1053                                 "error: unable to read group descriptors");
1054                         goto failed_mount_group_desc;
1055                 }
1056         }
1057         if (!ext2_check_descriptors (sb)) {
1058                 ext2_msg(sb, KERN_ERR, "group descriptors corrupted");
1059                 goto failed_mount2;
1060         }
1061         sbi->s_gdb_count = db_count;
1062         get_random_bytes(&sbi->s_next_generation, sizeof(u32));
1063         spin_lock_init(&sbi->s_next_gen_lock);
1064
1065         /* per fileystem reservation list head & lock */
1066         spin_lock_init(&sbi->s_rsv_window_lock);
1067         sbi->s_rsv_window_root = RB_ROOT;
1068         /*
1069          * Add a single, static dummy reservation to the start of the
1070          * reservation window list --- it gives us a placeholder for
1071          * append-at-start-of-list which makes the allocation logic
1072          * _much_ simpler.
1073          */
1074         sbi->s_rsv_window_head.rsv_start = EXT2_RESERVE_WINDOW_NOT_ALLOCATED;
1075         sbi->s_rsv_window_head.rsv_end = EXT2_RESERVE_WINDOW_NOT_ALLOCATED;
1076         sbi->s_rsv_window_head.rsv_alloc_hit = 0;
1077         sbi->s_rsv_window_head.rsv_goal_size = 0;
1078         ext2_rsv_window_add(sb, &sbi->s_rsv_window_head);
1079
1080         err = percpu_counter_init(&sbi->s_freeblocks_counter,
1081                                 ext2_count_free_blocks(sb), GFP_KERNEL);
1082         if (!err) {
1083                 err = percpu_counter_init(&sbi->s_freeinodes_counter,
1084                                 ext2_count_free_inodes(sb), GFP_KERNEL);
1085         }
1086         if (!err) {
1087                 err = percpu_counter_init(&sbi->s_dirs_counter,
1088                                 ext2_count_dirs(sb), GFP_KERNEL);
1089         }
1090         if (err) {
1091                 ext2_msg(sb, KERN_ERR, "error: insufficient memory");
1092                 goto failed_mount3;
1093         }
1094         /*
1095          * set up enough so that it can read an inode
1096          */
1097         sb->s_op = &ext2_sops;
1098         sb->s_export_op = &ext2_export_ops;
1099         sb->s_xattr = ext2_xattr_handlers;
1100
1101 #ifdef CONFIG_QUOTA
1102         sb->dq_op = &dquot_operations;
1103         sb->s_qcop = &dquot_quotactl_ops;
1104         sb->s_quota_types = QTYPE_MASK_USR | QTYPE_MASK_GRP;
1105 #endif
1106
1107         root = ext2_iget(sb, EXT2_ROOT_INO);
1108         if (IS_ERR(root)) {
1109                 ret = PTR_ERR(root);
1110                 goto failed_mount3;
1111         }
1112         if (!S_ISDIR(root->i_mode) || !root->i_blocks || !root->i_size) {
1113                 iput(root);
1114                 ext2_msg(sb, KERN_ERR, "error: corrupt root inode, run e2fsck");
1115                 goto failed_mount3;
1116         }
1117
1118         sb->s_root = d_make_root(root);
1119         if (!sb->s_root) {
1120                 ext2_msg(sb, KERN_ERR, "error: get root inode failed");
1121                 ret = -ENOMEM;
1122                 goto failed_mount3;
1123         }
1124         if (EXT2_HAS_COMPAT_FEATURE(sb, EXT3_FEATURE_COMPAT_HAS_JOURNAL))
1125                 ext2_msg(sb, KERN_WARNING,
1126                         "warning: mounting ext3 filesystem as ext2");
1127         if (ext2_setup_super (sb, es, sb->s_flags & MS_RDONLY))
1128                 sb->s_flags |= MS_RDONLY;
1129         ext2_write_super(sb);
1130         return 0;
1131
1132 cantfind_ext2:
1133         if (!silent)
1134                 ext2_msg(sb, KERN_ERR,
1135                         "error: can't find an ext2 filesystem on dev %s.",
1136                         sb->s_id);
1137         goto failed_mount;
1138 failed_mount3:
1139         percpu_counter_destroy(&sbi->s_freeblocks_counter);
1140         percpu_counter_destroy(&sbi->s_freeinodes_counter);
1141         percpu_counter_destroy(&sbi->s_dirs_counter);
1142 failed_mount2:
1143         for (i = 0; i < db_count; i++)
1144                 brelse(sbi->s_group_desc[i]);
1145 failed_mount_group_desc:
1146         kfree(sbi->s_group_desc);
1147         kfree(sbi->s_debts);
1148 failed_mount:
1149         brelse(bh);
1150 failed_sbi:
1151         sb->s_fs_info = NULL;
1152         kfree(sbi->s_blockgroup_lock);
1153         kfree(sbi);
1154 failed:
1155         return ret;
1156 }
1157
1158 static void ext2_clear_super_error(struct super_block *sb)
1159 {
1160         struct buffer_head *sbh = EXT2_SB(sb)->s_sbh;
1161
1162         if (buffer_write_io_error(sbh)) {
1163                 /*
1164                  * Oh, dear.  A previous attempt to write the
1165                  * superblock failed.  This could happen because the
1166                  * USB device was yanked out.  Or it could happen to
1167                  * be a transient write error and maybe the block will
1168                  * be remapped.  Nothing we can do but to retry the
1169                  * write and hope for the best.
1170                  */
1171                 ext2_msg(sb, KERN_ERR,
1172                        "previous I/O error to superblock detected\n");
1173                 clear_buffer_write_io_error(sbh);
1174                 set_buffer_uptodate(sbh);
1175         }
1176 }
1177
1178 static void ext2_sync_super(struct super_block *sb, struct ext2_super_block *es,
1179                             int wait)
1180 {
1181         ext2_clear_super_error(sb);
1182         spin_lock(&EXT2_SB(sb)->s_lock);
1183         es->s_free_blocks_count = cpu_to_le32(ext2_count_free_blocks(sb));
1184         es->s_free_inodes_count = cpu_to_le32(ext2_count_free_inodes(sb));
1185         es->s_wtime = cpu_to_le32(get_seconds());
1186         /* unlock before we do IO */
1187         spin_unlock(&EXT2_SB(sb)->s_lock);
1188         mark_buffer_dirty(EXT2_SB(sb)->s_sbh);
1189         if (wait)
1190                 sync_dirty_buffer(EXT2_SB(sb)->s_sbh);
1191 }
1192
1193 /*
1194  * In the second extended file system, it is not necessary to
1195  * write the super block since we use a mapping of the
1196  * disk super block in a buffer.
1197  *
1198  * However, this function is still used to set the fs valid
1199  * flags to 0.  We need to set this flag to 0 since the fs
1200  * may have been checked while mounted and e2fsck may have
1201  * set s_state to EXT2_VALID_FS after some corrections.
1202  */
1203 static int ext2_sync_fs(struct super_block *sb, int wait)
1204 {
1205         struct ext2_sb_info *sbi = EXT2_SB(sb);
1206         struct ext2_super_block *es = EXT2_SB(sb)->s_es;
1207
1208         /*
1209          * Write quota structures to quota file, sync_blockdev() will write
1210          * them to disk later
1211          */
1212         dquot_writeback_dquots(sb, -1);
1213
1214         spin_lock(&sbi->s_lock);
1215         if (es->s_state & cpu_to_le16(EXT2_VALID_FS)) {
1216                 ext2_debug("setting valid to 0\n");
1217                 es->s_state &= cpu_to_le16(~EXT2_VALID_FS);
1218         }
1219         spin_unlock(&sbi->s_lock);
1220         ext2_sync_super(sb, es, wait);
1221         return 0;
1222 }
1223
1224 static int ext2_freeze(struct super_block *sb)
1225 {
1226         struct ext2_sb_info *sbi = EXT2_SB(sb);
1227
1228         /*
1229          * Open but unlinked files present? Keep EXT2_VALID_FS flag cleared
1230          * because we have unattached inodes and thus filesystem is not fully
1231          * consistent.
1232          */
1233         if (atomic_long_read(&sb->s_remove_count)) {
1234                 ext2_sync_fs(sb, 1);
1235                 return 0;
1236         }
1237         /* Set EXT2_FS_VALID flag */
1238         spin_lock(&sbi->s_lock);
1239         sbi->s_es->s_state = cpu_to_le16(sbi->s_mount_state);
1240         spin_unlock(&sbi->s_lock);
1241         ext2_sync_super(sb, sbi->s_es, 1);
1242
1243         return 0;
1244 }
1245
1246 static int ext2_unfreeze(struct super_block *sb)
1247 {
1248         /* Just write sb to clear EXT2_VALID_FS flag */
1249         ext2_write_super(sb);
1250
1251         return 0;
1252 }
1253
1254 void ext2_write_super(struct super_block *sb)
1255 {
1256         if (!(sb->s_flags & MS_RDONLY))
1257                 ext2_sync_fs(sb, 1);
1258 }
1259
1260 static int ext2_remount (struct super_block * sb, int * flags, char * data)
1261 {
1262         struct ext2_sb_info * sbi = EXT2_SB(sb);
1263         struct ext2_super_block * es;
1264         struct ext2_mount_options old_opts;
1265         unsigned long old_sb_flags;
1266         int err;
1267
1268         sync_filesystem(sb);
1269         spin_lock(&sbi->s_lock);
1270
1271         /* Store the old options */
1272         old_sb_flags = sb->s_flags;
1273         old_opts.s_mount_opt = sbi->s_mount_opt;
1274         old_opts.s_resuid = sbi->s_resuid;
1275         old_opts.s_resgid = sbi->s_resgid;
1276
1277         /*
1278          * Allow the "check" option to be passed as a remount option.
1279          */
1280         if (!parse_options(data, sb)) {
1281                 err = -EINVAL;
1282                 goto restore_opts;
1283         }
1284
1285         sb->s_flags = (sb->s_flags & ~MS_POSIXACL) |
1286                 ((sbi->s_mount_opt & EXT2_MOUNT_POSIX_ACL) ? MS_POSIXACL : 0);
1287
1288         es = sbi->s_es;
1289         if ((sbi->s_mount_opt ^ old_opts.s_mount_opt) & EXT2_MOUNT_XIP) {
1290                 ext2_msg(sb, KERN_WARNING, "warning: refusing change of "
1291                          "xip flag with busy inodes while remounting");
1292                 sbi->s_mount_opt ^= EXT2_MOUNT_XIP;
1293         }
1294         if ((*flags & MS_RDONLY) == (sb->s_flags & MS_RDONLY)) {
1295                 spin_unlock(&sbi->s_lock);
1296                 return 0;
1297         }
1298         if (*flags & MS_RDONLY) {
1299                 if (le16_to_cpu(es->s_state) & EXT2_VALID_FS ||
1300                     !(sbi->s_mount_state & EXT2_VALID_FS)) {
1301                         spin_unlock(&sbi->s_lock);
1302                         return 0;
1303                 }
1304
1305                 /*
1306                  * OK, we are remounting a valid rw partition rdonly, so set
1307                  * the rdonly flag and then mark the partition as valid again.
1308                  */
1309                 es->s_state = cpu_to_le16(sbi->s_mount_state);
1310                 es->s_mtime = cpu_to_le32(get_seconds());
1311                 spin_unlock(&sbi->s_lock);
1312
1313                 err = dquot_suspend(sb, -1);
1314                 if (err < 0) {
1315                         spin_lock(&sbi->s_lock);
1316                         goto restore_opts;
1317                 }
1318
1319                 ext2_sync_super(sb, es, 1);
1320         } else {
1321                 __le32 ret = EXT2_HAS_RO_COMPAT_FEATURE(sb,
1322                                                ~EXT2_FEATURE_RO_COMPAT_SUPP);
1323                 if (ret) {
1324                         ext2_msg(sb, KERN_WARNING,
1325                                 "warning: couldn't remount RDWR because of "
1326                                 "unsupported optional features (%x).",
1327                                 le32_to_cpu(ret));
1328                         err = -EROFS;
1329                         goto restore_opts;
1330                 }
1331                 /*
1332                  * Mounting a RDONLY partition read-write, so reread and
1333                  * store the current valid flag.  (It may have been changed
1334                  * by e2fsck since we originally mounted the partition.)
1335                  */
1336                 sbi->s_mount_state = le16_to_cpu(es->s_state);
1337                 if (!ext2_setup_super (sb, es, 0))
1338                         sb->s_flags &= ~MS_RDONLY;
1339                 spin_unlock(&sbi->s_lock);
1340
1341                 ext2_write_super(sb);
1342
1343                 dquot_resume(sb, -1);
1344         }
1345
1346         return 0;
1347 restore_opts:
1348         sbi->s_mount_opt = old_opts.s_mount_opt;
1349         sbi->s_resuid = old_opts.s_resuid;
1350         sbi->s_resgid = old_opts.s_resgid;
1351         sb->s_flags = old_sb_flags;
1352         spin_unlock(&sbi->s_lock);
1353         return err;
1354 }
1355
1356 static int ext2_statfs (struct dentry * dentry, struct kstatfs * buf)
1357 {
1358         struct super_block *sb = dentry->d_sb;
1359         struct ext2_sb_info *sbi = EXT2_SB(sb);
1360         struct ext2_super_block *es = sbi->s_es;
1361         u64 fsid;
1362
1363         spin_lock(&sbi->s_lock);
1364
1365         if (test_opt (sb, MINIX_DF))
1366                 sbi->s_overhead_last = 0;
1367         else if (sbi->s_blocks_last != le32_to_cpu(es->s_blocks_count)) {
1368                 unsigned long i, overhead = 0;
1369                 smp_rmb();
1370
1371                 /*
1372                  * Compute the overhead (FS structures). This is constant
1373                  * for a given filesystem unless the number of block groups
1374                  * changes so we cache the previous value until it does.
1375                  */
1376
1377                 /*
1378                  * All of the blocks before first_data_block are
1379                  * overhead
1380                  */
1381                 overhead = le32_to_cpu(es->s_first_data_block);
1382
1383                 /*
1384                  * Add the overhead attributed to the superblock and
1385                  * block group descriptors.  If the sparse superblocks
1386                  * feature is turned on, then not all groups have this.
1387                  */
1388                 for (i = 0; i < sbi->s_groups_count; i++)
1389                         overhead += ext2_bg_has_super(sb, i) +
1390                                 ext2_bg_num_gdb(sb, i);
1391
1392                 /*
1393                  * Every block group has an inode bitmap, a block
1394                  * bitmap, and an inode table.
1395                  */
1396                 overhead += (sbi->s_groups_count *
1397                              (2 + sbi->s_itb_per_group));
1398                 sbi->s_overhead_last = overhead;
1399                 smp_wmb();
1400                 sbi->s_blocks_last = le32_to_cpu(es->s_blocks_count);
1401         }
1402
1403         buf->f_type = EXT2_SUPER_MAGIC;
1404         buf->f_bsize = sb->s_blocksize;
1405         buf->f_blocks = le32_to_cpu(es->s_blocks_count) - sbi->s_overhead_last;
1406         buf->f_bfree = ext2_count_free_blocks(sb);
1407         es->s_free_blocks_count = cpu_to_le32(buf->f_bfree);
1408         buf->f_bavail = buf->f_bfree - le32_to_cpu(es->s_r_blocks_count);
1409         if (buf->f_bfree < le32_to_cpu(es->s_r_blocks_count))
1410                 buf->f_bavail = 0;
1411         buf->f_files = le32_to_cpu(es->s_inodes_count);
1412         buf->f_ffree = ext2_count_free_inodes(sb);
1413         es->s_free_inodes_count = cpu_to_le32(buf->f_ffree);
1414         buf->f_namelen = EXT2_NAME_LEN;
1415         fsid = le64_to_cpup((void *)es->s_uuid) ^
1416                le64_to_cpup((void *)es->s_uuid + sizeof(u64));
1417         buf->f_fsid.val[0] = fsid & 0xFFFFFFFFUL;
1418         buf->f_fsid.val[1] = (fsid >> 32) & 0xFFFFFFFFUL;
1419         spin_unlock(&sbi->s_lock);
1420         return 0;
1421 }
1422
1423 static struct dentry *ext2_mount(struct file_system_type *fs_type,
1424         int flags, const char *dev_name, void *data)
1425 {
1426         return mount_bdev(fs_type, flags, dev_name, data, ext2_fill_super);
1427 }
1428
1429 #ifdef CONFIG_QUOTA
1430
1431 /* Read data from quotafile - avoid pagecache and such because we cannot afford
1432  * acquiring the locks... As quota files are never truncated and quota code
1433  * itself serializes the operations (and no one else should touch the files)
1434  * we don't have to be afraid of races */
1435 static ssize_t ext2_quota_read(struct super_block *sb, int type, char *data,
1436                                size_t len, loff_t off)
1437 {
1438         struct inode *inode = sb_dqopt(sb)->files[type];
1439         sector_t blk = off >> EXT2_BLOCK_SIZE_BITS(sb);
1440         int err = 0;
1441         int offset = off & (sb->s_blocksize - 1);
1442         int tocopy;
1443         size_t toread;
1444         struct buffer_head tmp_bh;
1445         struct buffer_head *bh;
1446         loff_t i_size = i_size_read(inode);
1447
1448         if (off > i_size)
1449                 return 0;
1450         if (off+len > i_size)
1451                 len = i_size-off;
1452         toread = len;
1453         while (toread > 0) {
1454                 tocopy = sb->s_blocksize - offset < toread ?
1455                                 sb->s_blocksize - offset : toread;
1456
1457                 tmp_bh.b_state = 0;
1458                 tmp_bh.b_size = sb->s_blocksize;
1459                 err = ext2_get_block(inode, blk, &tmp_bh, 0);
1460                 if (err < 0)
1461                         return err;
1462                 if (!buffer_mapped(&tmp_bh))    /* A hole? */
1463                         memset(data, 0, tocopy);
1464                 else {
1465                         bh = sb_bread(sb, tmp_bh.b_blocknr);
1466                         if (!bh)
1467                                 return -EIO;
1468                         memcpy(data, bh->b_data+offset, tocopy);
1469                         brelse(bh);
1470                 }
1471                 offset = 0;
1472                 toread -= tocopy;
1473                 data += tocopy;
1474                 blk++;
1475         }
1476         return len;
1477 }
1478
1479 /* Write to quotafile */
1480 static ssize_t ext2_quota_write(struct super_block *sb, int type,
1481                                 const char *data, size_t len, loff_t off)
1482 {
1483         struct inode *inode = sb_dqopt(sb)->files[type];
1484         sector_t blk = off >> EXT2_BLOCK_SIZE_BITS(sb);
1485         int err = 0;
1486         int offset = off & (sb->s_blocksize - 1);
1487         int tocopy;
1488         size_t towrite = len;
1489         struct buffer_head tmp_bh;
1490         struct buffer_head *bh;
1491
1492         while (towrite > 0) {
1493                 tocopy = sb->s_blocksize - offset < towrite ?
1494                                 sb->s_blocksize - offset : towrite;
1495
1496                 tmp_bh.b_state = 0;
1497                 tmp_bh.b_size = sb->s_blocksize;
1498                 err = ext2_get_block(inode, blk, &tmp_bh, 1);
1499                 if (err < 0)
1500                         goto out;
1501                 if (offset || tocopy != EXT2_BLOCK_SIZE(sb))
1502                         bh = sb_bread(sb, tmp_bh.b_blocknr);
1503                 else
1504                         bh = sb_getblk(sb, tmp_bh.b_blocknr);
1505                 if (unlikely(!bh)) {
1506                         err = -EIO;
1507                         goto out;
1508                 }
1509                 lock_buffer(bh);
1510                 memcpy(bh->b_data+offset, data, tocopy);
1511                 flush_dcache_page(bh->b_page);
1512                 set_buffer_uptodate(bh);
1513                 mark_buffer_dirty(bh);
1514                 unlock_buffer(bh);
1515                 brelse(bh);
1516                 offset = 0;
1517                 towrite -= tocopy;
1518                 data += tocopy;
1519                 blk++;
1520         }
1521 out:
1522         if (len == towrite)
1523                 return err;
1524         if (inode->i_size < off+len-towrite)
1525                 i_size_write(inode, off+len-towrite);
1526         inode->i_version++;
1527         inode->i_mtime = inode->i_ctime = CURRENT_TIME;
1528         mark_inode_dirty(inode);
1529         return len - towrite;
1530 }
1531
1532 #endif
1533
1534 static struct file_system_type ext2_fs_type = {
1535         .owner          = THIS_MODULE,
1536         .name           = "ext2",
1537         .mount          = ext2_mount,
1538         .kill_sb        = kill_block_super,
1539         .fs_flags       = FS_REQUIRES_DEV,
1540 };
1541 MODULE_ALIAS_FS("ext2");
1542
1543 static int __init init_ext2_fs(void)
1544 {
1545         int err = init_ext2_xattr();
1546         if (err)
1547                 return err;
1548         err = init_inodecache();
1549         if (err)
1550                 goto out1;
1551         err = register_filesystem(&ext2_fs_type);
1552         if (err)
1553                 goto out;
1554         return 0;
1555 out:
1556         destroy_inodecache();
1557 out1:
1558         exit_ext2_xattr();
1559         return err;
1560 }
1561
1562 static void __exit exit_ext2_fs(void)
1563 {
1564         unregister_filesystem(&ext2_fs_type);
1565         destroy_inodecache();
1566         exit_ext2_xattr();
1567 }
1568
1569 MODULE_AUTHOR("Remy Card and others");
1570 MODULE_DESCRIPTION("Second Extended Filesystem");
1571 MODULE_LICENSE("GPL");
1572 module_init(init_ext2_fs)
1573 module_exit(exit_ext2_fs)