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         if (sbi->s_mb_cache) {
 135                 ext2_xattr_destroy_cache(sbi->s_mb_cache);
 136                 sbi->s_mb_cache = NULL;
 137         }
 138         if (!(sb->s_flags & MS_RDONLY)) {
 139                 struct ext2_super_block *es = sbi->s_es;
 140
 141                 spin_lock(&sbi->s_lock);
 142                 es->s_state = cpu_to_le16(sbi->s_mount_state);
 143                 spin_unlock(&sbi->s_lock);
 144                 ext2_sync_super(sb, es, 1);
 145         }
 146         db_count = sbi->s_gdb_count;
 147         for (i = 0; i < db_count; i++)
 148                 if (sbi->s_group_desc[i])
 149                         brelse (sbi->s_group_desc[i]);
 150         kfree(sbi->s_group_desc);
 151         kfree(sbi->s_debts);
 152         percpu_counter_destroy(&sbi->s_freeblocks_counter);
 153         percpu_counter_destroy(&sbi->s_freeinodes_counter);
 154         percpu_counter_destroy(&sbi->s_dirs_counter);
 155         brelse (sbi->s_sbh);
 156         sb->s_fs_info = NULL;
 157         kfree(sbi->s_blockgroup_lock);
 158         kfree(sbi);
 159 }
 160
 161 static struct kmem_cache * ext2_inode_cachep;
 162
 163 static struct inode *ext2_alloc_inode(struct super_block *sb)
 164 {
 165         struct ext2_inode_info *ei;
 166         ei = kmem_cache_alloc(ext2_inode_cachep, GFP_KERNEL);
 167         if (!ei)
 168                 return NULL;
 169         ei->i_block_alloc_info = NULL;
 170         ei->vfs_inode.i_version = 1;
 171 #ifdef CONFIG_QUOTA
 172         memset(&ei->i_dquot, 0, sizeof(ei->i_dquot));
 173 #endif
 174
 175         return &ei->vfs_inode;
 176 }
 177
 178 static void ext2_i_callback(struct rcu_head *head)
 179 {
 180         struct inode *inode = container_of(head, struct inode, i_rcu);
 181         kmem_cache_free(ext2_inode_cachep, EXT2_I(inode));
 182 }
 183
 184 static void ext2_destroy_inode(struct inode *inode)
 185 {
 186         call_rcu(&inode->i_rcu, ext2_i_callback);
 187 }
 188
 189 static void init_once(void *foo)
 190 {
 191         struct ext2_inode_info *ei = (struct ext2_inode_info *) foo;
 192
 193         rwlock_init(&ei->i_meta_lock);
 194 #ifdef CONFIG_EXT2_FS_XATTR
 195         init_rwsem(&ei->xattr_sem);
 196 #endif
 197         mutex_init(&ei->truncate_mutex);
 198 #ifdef CONFIG_FS_DAX
 199         init_rwsem(&ei->dax_sem);
 200 #endif
 201         inode_init_once(&ei->vfs_inode);
 202 }
 203
 204 static int __init init_inodecache(void)
 205 {
 206         ext2_inode_cachep = kmem_cache_create("ext2_inode_cache",
 207                                              sizeof(struct ext2_inode_info),
 208                                              0, (SLAB_RECLAIM_ACCOUNT|
 209                                                 SLAB_MEM_SPREAD),
 210                                              init_once);
 211         if (ext2_inode_cachep == NULL)
 212                 return -ENOMEM;
 213         return 0;
 214 }
 215
 216 static void destroy_inodecache(void)
 217 {
 218         /*
 219          * Make sure all delayed rcu free inodes are flushed before we
 220          * destroy cache.
 221          */
 222         rcu_barrier();
 223         kmem_cache_destroy(ext2_inode_cachep);
 224 }
 225
 226 static int ext2_show_options(struct seq_file *seq, struct dentry *root)
 227 {
 228         struct super_block *sb = root->d_sb;
 229         struct ext2_sb_info *sbi = EXT2_SB(sb);
 230         struct ext2_super_block *es = sbi->s_es;
 231         unsigned long def_mount_opts;
 232
 233         spin_lock(&sbi->s_lock);
 234         def_mount_opts = le32_to_cpu(es->s_default_mount_opts);
 235
 236         if (sbi->s_sb_block != 1)
 237                 seq_printf(seq, ",sb=%lu", sbi->s_sb_block);
 238         if (test_opt(sb, MINIX_DF))
 239                 seq_puts(seq, ",minixdf");
 240         if (test_opt(sb, GRPID))
 241                 seq_puts(seq, ",grpid");
 242         if (!test_opt(sb, GRPID) && (def_mount_opts & EXT2_DEFM_BSDGROUPS))
 243                 seq_puts(seq, ",nogrpid");
 244         if (!uid_eq(sbi->s_resuid, make_kuid(&init_user_ns, EXT2_DEF_RESUID)) ||
 245             le16_to_cpu(es->s_def_resuid) != EXT2_DEF_RESUID) {
 246                 seq_printf(seq, ",resuid=%u",
 247                                 from_kuid_munged(&init_user_ns, sbi->s_resuid));
 248         }
 249         if (!gid_eq(sbi->s_resgid, make_kgid(&init_user_ns, EXT2_DEF_RESGID)) ||
 250             le16_to_cpu(es->s_def_resgid) != EXT2_DEF_RESGID) {
 251                 seq_printf(seq, ",resgid=%u",
 252                                 from_kgid_munged(&init_user_ns, sbi->s_resgid));
 253         }
 254         if (test_opt(sb, ERRORS_RO)) {
 255                 int def_errors = le16_to_cpu(es->s_errors);
 256
 257                 if (def_errors == EXT2_ERRORS_PANIC ||
 258                     def_errors == EXT2_ERRORS_CONTINUE) {
 259                         seq_puts(seq, ",errors=remount-ro");
 260                 }
 261         }
 262         if (test_opt(sb, ERRORS_CONT))
 263                 seq_puts(seq, ",errors=continue");
 264         if (test_opt(sb, ERRORS_PANIC))
 265                 seq_puts(seq, ",errors=panic");
 266         if (test_opt(sb, NO_UID32))
 267                 seq_puts(seq, ",nouid32");
 268         if (test_opt(sb, DEBUG))
 269                 seq_puts(seq, ",debug");
 270         if (test_opt(sb, OLDALLOC))
 271                 seq_puts(seq, ",oldalloc");
 272
 273 #ifdef CONFIG_EXT2_FS_XATTR
 274         if (test_opt(sb, XATTR_USER))
 275                 seq_puts(seq, ",user_xattr");
 276         if (!test_opt(sb, XATTR_USER) &&
 277             (def_mount_opts & EXT2_DEFM_XATTR_USER)) {
 278                 seq_puts(seq, ",nouser_xattr");
 279         }
 280 #endif
 281
 282 #ifdef CONFIG_EXT2_FS_POSIX_ACL
 283         if (test_opt(sb, POSIX_ACL))
 284                 seq_puts(seq, ",acl");
 285         if (!test_opt(sb, POSIX_ACL) && (def_mount_opts & EXT2_DEFM_ACL))
 286                 seq_puts(seq, ",noacl");
 287 #endif
 288
 289         if (test_opt(sb, NOBH))
 290                 seq_puts(seq, ",nobh");
 291
 292 #if defined(CONFIG_QUOTA)
 293         if (sbi->s_mount_opt & EXT2_MOUNT_USRQUOTA)
 294                 seq_puts(seq, ",usrquota");
 295
 296         if (sbi->s_mount_opt & EXT2_MOUNT_GRPQUOTA)
 297                 seq_puts(seq, ",grpquota");
 298 #endif
 299
 300 #ifdef CONFIG_FS_DAX
 301         if (sbi->s_mount_opt & EXT2_MOUNT_XIP)
 302                 seq_puts(seq, ",xip");
 303         if (sbi->s_mount_opt & EXT2_MOUNT_DAX)
 304                 seq_puts(seq, ",dax");
 305 #endif
 306
 307         if (!test_opt(sb, RESERVATION))
 308                 seq_puts(seq, ",noreservation");
 309
 310         spin_unlock(&sbi->s_lock);
 311         return 0;
 312 }
 313
 314 #ifdef CONFIG_QUOTA
 315 static ssize_t ext2_quota_read(struct super_block *sb, int type, char *data, size_t len, loff_t off);
 316 static ssize_t ext2_quota_write(struct super_block *sb, int type, const char *data, size_t len, loff_t off);
 317 static struct dquot **ext2_get_dquots(struct inode *inode)
 318 {
 319         return EXT2_I(inode)->i_dquot;
 320 }
 321 #endif
 322
 323 static const struct super_operations ext2_sops = {
 324         .alloc_inode    = ext2_alloc_inode,
 325         .destroy_inode  = ext2_destroy_inode,
 326         .write_inode    = ext2_write_inode,
 327         .evict_inode    = ext2_evict_inode,
 328         .put_super      = ext2_put_super,
 329         .sync_fs        = ext2_sync_fs,
 330         .freeze_fs      = ext2_freeze,
 331         .unfreeze_fs    = ext2_unfreeze,
 332         .statfs         = ext2_statfs,
 333         .remount_fs     = ext2_remount,
 334         .show_options   = ext2_show_options,
 335 #ifdef CONFIG_QUOTA
 336         .quota_read     = ext2_quota_read,
 337         .quota_write    = ext2_quota_write,
 338         .get_dquots     = ext2_get_dquots,
 339 #endif
 340 };
 341
 342 static struct inode *ext2_nfs_get_inode(struct super_block *sb,
 343                 u64 ino, u32 generation)
 344 {
 345         struct inode *inode;
 346
 347         if (ino < EXT2_FIRST_INO(sb) && ino != EXT2_ROOT_INO)
 348                 return ERR_PTR(-ESTALE);
 349         if (ino > le32_to_cpu(EXT2_SB(sb)->s_es->s_inodes_count))
 350                 return ERR_PTR(-ESTALE);
 351
 352         /*
 353          * ext2_iget isn't quite right if the inode is currently unallocated!
 354          * However ext2_iget currently does appropriate checks to handle stale
 355          * inodes so everything is OK.
 356          */
 357         inode = ext2_iget(sb, ino);
 358         if (IS_ERR(inode))
 359                 return ERR_CAST(inode);
 360         if (generation && inode->i_generation != generation) {
 361                 /* we didn't find the right inode.. */
 362                 iput(inode);
 363                 return ERR_PTR(-ESTALE);
 364         }
 365         return inode;
 366 }
 367
 368 static struct dentry *ext2_fh_to_dentry(struct super_block *sb, struct fid *fid,
 369                 int fh_len, int fh_type)
 370 {
 371         return generic_fh_to_dentry(sb, fid, fh_len, fh_type,
 372                                     ext2_nfs_get_inode);
 373 }
 374
 375 static struct dentry *ext2_fh_to_parent(struct super_block *sb, struct fid *fid,
 376                 int fh_len, int fh_type)
 377 {
 378         return generic_fh_to_parent(sb, fid, fh_len, fh_type,
 379                                     ext2_nfs_get_inode);
 380 }
 381
 382 static const struct export_operations ext2_export_ops = {
 383         .fh_to_dentry = ext2_fh_to_dentry,
 384         .fh_to_parent = ext2_fh_to_parent,
 385         .get_parent = ext2_get_parent,
 386 };
 387
 388 static unsigned long get_sb_block(void **data)
 389 {
 390         unsigned long   sb_block;
 391         char            *options = (char *) *data;
 392
 393         if (!options || strncmp(options, "sb=", 3) != 0)
 394                 return 1;       /* Default location */
 395         options += 3;
 396         sb_block = simple_strtoul(options, &options, 0);
 397         if (*options && *options != ',') {
 398                 printk("EXT2-fs: Invalid sb specification: %s\n",
 399                        (char *) *data);
 400                 return 1;
 401         }
 402         if (*options == ',')
 403                 options++;
 404         *data = (void *) options;
 405         return sb_block;
 406 }
 407
 408 enum {
 409         Opt_bsd_df, Opt_minix_df, Opt_grpid, Opt_nogrpid,
 410         Opt_resgid, Opt_resuid, Opt_sb, Opt_err_cont, Opt_err_panic,
 411         Opt_err_ro, Opt_nouid32, Opt_nocheck, Opt_debug,
 412         Opt_oldalloc, Opt_orlov, Opt_nobh, Opt_user_xattr, Opt_nouser_xattr,
 413         Opt_acl, Opt_noacl, Opt_xip, Opt_dax, Opt_ignore, Opt_err, Opt_quota,
 414         Opt_usrquota, Opt_grpquota, Opt_reservation, Opt_noreservation
 415 };
 416
 417 static const match_table_t tokens = {
 418         {Opt_bsd_df, "bsddf"},
 419         {Opt_minix_df, "minixdf"},
 420         {Opt_grpid, "grpid"},
 421         {Opt_grpid, "bsdgroups"},
 422         {Opt_nogrpid, "nogrpid"},
 423         {Opt_nogrpid, "sysvgroups"},
 424         {Opt_resgid, "resgid=%u"},
 425         {Opt_resuid, "resuid=%u"},
 426         {Opt_sb, "sb=%u"},
 427         {Opt_err_cont, "errors=continue"},
 428         {Opt_err_panic, "errors=panic"},
 429         {Opt_err_ro, "errors=remount-ro"},
 430         {Opt_nouid32, "nouid32"},
 431         {Opt_nocheck, "check=none"},
 432         {Opt_nocheck, "nocheck"},
 433         {Opt_debug, "debug"},
 434         {Opt_oldalloc, "oldalloc"},
 435         {Opt_orlov, "orlov"},
 436         {Opt_nobh, "nobh"},
 437         {Opt_user_xattr, "user_xattr"},
 438         {Opt_nouser_xattr, "nouser_xattr"},
 439         {Opt_acl, "acl"},
 440         {Opt_noacl, "noacl"},
 441         {Opt_xip, "xip"},
 442         {Opt_dax, "dax"},
 443         {Opt_grpquota, "grpquota"},
 444         {Opt_ignore, "noquota"},
 445         {Opt_quota, "quota"},
 446         {Opt_usrquota, "usrquota"},
 447         {Opt_reservation, "reservation"},
 448         {Opt_noreservation, "noreservation"},
 449         {Opt_err, NULL}
 450 };
 451
 452 static int parse_options(char *options, struct super_block *sb)
 453 {
 454         char *p;
 455         struct ext2_sb_info *sbi = EXT2_SB(sb);
 456         substring_t args[MAX_OPT_ARGS];
 457         int option;
 458         kuid_t uid;
 459         kgid_t gid;
 460
 461         if (!options)
 462                 return 1;
 463
 464         while ((p = strsep (&options, ",")) != NULL) {
 465                 int token;
 466                 if (!*p)
 467                         continue;
 468
 469                 token = match_token(p, tokens, args);
 470                 switch (token) {
 471                 case Opt_bsd_df:
 472                         clear_opt (sbi->s_mount_opt, MINIX_DF);
 473                         break;
 474                 case Opt_minix_df:
 475                         set_opt (sbi->s_mount_opt, MINIX_DF);
 476                         break;
 477                 case Opt_grpid:
 478                         set_opt (sbi->s_mount_opt, GRPID);
 479                         break;
 480                 case Opt_nogrpid:
 481                         clear_opt (sbi->s_mount_opt, GRPID);
 482                         break;
 483                 case Opt_resuid:
 484                         if (match_int(&args[0], &option))
 485                                 return 0;
 486                         uid = make_kuid(current_user_ns(), option);
 487                         if (!uid_valid(uid)) {
 488                                 ext2_msg(sb, KERN_ERR, "Invalid uid value %d", option);
 489                                 return 0;
 490
 491                         }
 492                         sbi->s_resuid = uid;
 493                         break;
 494                 case Opt_resgid:
 495                         if (match_int(&args[0], &option))
 496                                 return 0;
 497                         gid = make_kgid(current_user_ns(), option);
 498                         if (!gid_valid(gid)) {
 499                                 ext2_msg(sb, KERN_ERR, "Invalid gid value %d", option);
 500                                 return 0;
 501                         }
 502                         sbi->s_resgid = gid;
 503                         break;
 504                 case Opt_sb:
 505                         /* handled by get_sb_block() instead of here */
 506                         /* *sb_block = match_int(&args[0]); */
 507                         break;
 508                 case Opt_err_panic:
 509                         clear_opt (sbi->s_mount_opt, ERRORS_CONT);
 510                         clear_opt (sbi->s_mount_opt, ERRORS_RO);
 511                         set_opt (sbi->s_mount_opt, ERRORS_PANIC);
 512                         break;
 513                 case Opt_err_ro:
 514                         clear_opt (sbi->s_mount_opt, ERRORS_CONT);
 515                         clear_opt (sbi->s_mount_opt, ERRORS_PANIC);
 516                         set_opt (sbi->s_mount_opt, ERRORS_RO);
 517                         break;
 518                 case Opt_err_cont:
 519                         clear_opt (sbi->s_mount_opt, ERRORS_RO);
 520                         clear_opt (sbi->s_mount_opt, ERRORS_PANIC);
 521                         set_opt (sbi->s_mount_opt, ERRORS_CONT);
 522                         break;
 523                 case Opt_nouid32:
 524                         set_opt (sbi->s_mount_opt, NO_UID32);
 525                         break;
 526                 case Opt_nocheck:
 527                         clear_opt (sbi->s_mount_opt, CHECK);
 528                         break;
 529                 case Opt_debug:
 530                         set_opt (sbi->s_mount_opt, DEBUG);
 531                         break;
 532                 case Opt_oldalloc:
 533                         set_opt (sbi->s_mount_opt, OLDALLOC);
 534                         break;
 535                 case Opt_orlov:
 536                         clear_opt (sbi->s_mount_opt, OLDALLOC);
 537                         break;
 538                 case Opt_nobh:
 539                         set_opt (sbi->s_mount_opt, NOBH);
 540                         break;
 541 #ifdef CONFIG_EXT2_FS_XATTR
 542                 case Opt_user_xattr:
 543                         set_opt (sbi->s_mount_opt, XATTR_USER);
 544                         break;
 545                 case Opt_nouser_xattr:
 546                         clear_opt (sbi->s_mount_opt, XATTR_USER);
 547                         break;
 548 #else
 549                 case Opt_user_xattr:
 550                 case Opt_nouser_xattr:
 551                         ext2_msg(sb, KERN_INFO, "(no)user_xattr options"
 552                                 "not supported");
 553                         break;
 554 #endif
 555 #ifdef CONFIG_EXT2_FS_POSIX_ACL
 556                 case Opt_acl:
 557                         set_opt(sbi->s_mount_opt, POSIX_ACL);
 558                         break;
 559                 case Opt_noacl:
 560                         clear_opt(sbi->s_mount_opt, POSIX_ACL);
 561                         break;
 562 #else
 563                 case Opt_acl:
 564                 case Opt_noacl:
 565                         ext2_msg(sb, KERN_INFO,
 566                                 "(no)acl options not supported");
 567                         break;
 568 #endif
 569                 case Opt_xip:
 570                         ext2_msg(sb, KERN_INFO, "use dax instead of xip");
 571                         set_opt(sbi->s_mount_opt, XIP);
 572                         /* Fall through */
 573                 case Opt_dax:
 574 #ifdef CONFIG_FS_DAX
 575                         ext2_msg(sb, KERN_WARNING,
 576                 "DAX enabled. Warning: EXPERIMENTAL, use at your own risk");
 577                         set_opt(sbi->s_mount_opt, DAX);
 578 #else
 579                         ext2_msg(sb, KERN_INFO, "dax option not supported");
 580 #endif
 581                         break;
 582
 583 #if defined(CONFIG_QUOTA)
 584                 case Opt_quota:
 585                 case Opt_usrquota:
 586                         set_opt(sbi->s_mount_opt, USRQUOTA);
 587                         break;
 588
 589                 case Opt_grpquota:
 590                         set_opt(sbi->s_mount_opt, GRPQUOTA);
 591                         break;
 592 #else
 593                 case Opt_quota:
 594                 case Opt_usrquota:
 595                 case Opt_grpquota:
 596                         ext2_msg(sb, KERN_INFO,
 597                                 "quota operations not supported");
 598                         break;
 599 #endif
 600
 601                 case Opt_reservation:
 602                         set_opt(sbi->s_mount_opt, RESERVATION);
 603                         ext2_msg(sb, KERN_INFO, "reservations ON");
 604                         break;
 605                 case Opt_noreservation:
 606                         clear_opt(sbi->s_mount_opt, RESERVATION);
 607                         ext2_msg(sb, KERN_INFO, "reservations OFF");
 608                         break;
 609                 case Opt_ignore:
 610                         break;
 611                 default:
 612                         return 0;
 613                 }
 614         }
 615         return 1;
 616 }
 617
 618 static int ext2_setup_super (struct super_block * sb,
 619                               struct ext2_super_block * es,
 620                               int read_only)
 621 {
 622         int res = 0;
 623         struct ext2_sb_info *sbi = EXT2_SB(sb);
 624
 625         if (le32_to_cpu(es->s_rev_level) > EXT2_MAX_SUPP_REV) {
 626                 ext2_msg(sb, KERN_ERR,
 627                         "error: revision level too high, "
 628                         "forcing read-only mode");
 629                 res = MS_RDONLY;
 630         }
 631         if (read_only)
 632                 return res;
 633         if (!(sbi->s_mount_state & EXT2_VALID_FS))
 634                 ext2_msg(sb, KERN_WARNING,
 635                         "warning: mounting unchecked fs, "
 636                         "running e2fsck is recommended");
 637         else if ((sbi->s_mount_state & EXT2_ERROR_FS))
 638                 ext2_msg(sb, KERN_WARNING,
 639                         "warning: mounting fs with errors, "
 640                         "running e2fsck is recommended");
 641         else if ((__s16) le16_to_cpu(es->s_max_mnt_count) >= 0 &&
 642                  le16_to_cpu(es->s_mnt_count) >=
 643                  (unsigned short) (__s16) le16_to_cpu(es->s_max_mnt_count))
 644                 ext2_msg(sb, KERN_WARNING,
 645                         "warning: maximal mount count reached, "
 646                         "running e2fsck is recommended");
 647         else if (le32_to_cpu(es->s_checkinterval) &&
 648                 (le32_to_cpu(es->s_lastcheck) +
 649                         le32_to_cpu(es->s_checkinterval) <= get_seconds()))
 650                 ext2_msg(sb, KERN_WARNING,
 651                         "warning: checktime reached, "
 652                         "running e2fsck is recommended");
 653         if (!le16_to_cpu(es->s_max_mnt_count))
 654                 es->s_max_mnt_count = cpu_to_le16(EXT2_DFL_MAX_MNT_COUNT);
 655         le16_add_cpu(&es->s_mnt_count, 1);
 656         if (test_opt (sb, DEBUG))
 657                 ext2_msg(sb, KERN_INFO, "%s, %s, bs=%lu, fs=%lu, gc=%lu, "
 658                         "bpg=%lu, ipg=%lu, mo=%04lx]",
 659                         EXT2FS_VERSION, EXT2FS_DATE, sb->s_blocksize,
 660                         sbi->s_frag_size,
 661                         sbi->s_groups_count,
 662                         EXT2_BLOCKS_PER_GROUP(sb),
 663                         EXT2_INODES_PER_GROUP(sb),
 664                         sbi->s_mount_opt);
 665         return res;
 666 }
 667
 668 static int ext2_check_descriptors(struct super_block *sb)
 669 {
 670         int i;
 671         struct ext2_sb_info *sbi = EXT2_SB(sb);
 672
 673         ext2_debug ("Checking group descriptors");
 674
 675         for (i = 0; i < sbi->s_groups_count; i++) {
 676                 struct ext2_group_desc *gdp = ext2_get_group_desc(sb, i, NULL);
 677                 ext2_fsblk_t first_block = ext2_group_first_block_no(sb, i);
 678                 ext2_fsblk_t last_block;
 679
 680                 if (i == sbi->s_groups_count - 1)
 681                         last_block = le32_to_cpu(sbi->s_es->s_blocks_count) - 1;
 682                 else
 683                         last_block = first_block +
 684                                 (EXT2_BLOCKS_PER_GROUP(sb) - 1);
 685
 686                 if (le32_to_cpu(gdp->bg_block_bitmap) < first_block ||
 687                     le32_to_cpu(gdp->bg_block_bitmap) > last_block)
 688                 {
 689                         ext2_error (sb, "ext2_check_descriptors",
 690                                     "Block bitmap for group %d"
 691                                     " not in group (block %lu)!",
 692                                     i, (unsigned long) le32_to_cpu(gdp->bg_block_bitmap));
 693                         return 0;
 694                 }
 695                 if (le32_to_cpu(gdp->bg_inode_bitmap) < first_block ||
 696                     le32_to_cpu(gdp->bg_inode_bitmap) > last_block)
 697                 {
 698                         ext2_error (sb, "ext2_check_descriptors",
 699                                     "Inode bitmap for group %d"
 700                                     " not in group (block %lu)!",
 701                                     i, (unsigned long) le32_to_cpu(gdp->bg_inode_bitmap));
 702                         return 0;
 703                 }
 704                 if (le32_to_cpu(gdp->bg_inode_table) < first_block ||
 705                     le32_to_cpu(gdp->bg_inode_table) + sbi->s_itb_per_group - 1 >
 706                     last_block)
 707                 {
 708                         ext2_error (sb, "ext2_check_descriptors",
 709                                     "Inode table for group %d"
 710                                     " not in group (block %lu)!",
 711                                     i, (unsigned long) le32_to_cpu(gdp->bg_inode_table));
 712                         return 0;
 713                 }
 714         }
 715         return 1;
 716 }
 717
 718 /*
 719  * Maximal file size.  There is a direct, and {,double-,triple-}indirect
 720  * block limit, and also a limit of (2^32 - 1) 512-byte sectors in i_blocks.
 721  * We need to be 1 filesystem block less than the 2^32 sector limit.
 722  */
 723 static loff_t ext2_max_size(int bits)
 724 {
 725         loff_t res = EXT2_NDIR_BLOCKS;
 726         int meta_blocks;
 727         unsigned int upper_limit;
 728         unsigned int ppb = 1 << (bits-2);
 729
 730         /* This is calculated to be the largest file size for a
 731          * dense, file such that the total number of
 732          * sectors in the file, including data and all indirect blocks,
 733          * does not exceed 2^32 -1
 734          * __u32 i_blocks representing the total number of
 735          * 512 bytes blocks of the file
 736          */
 737         upper_limit = (1LL << 32) - 1;
 738
 739         /* total blocks in file system block size */
 740         upper_limit >>= (bits - 9);
 741
 742         /* Compute how many blocks we can address by block tree */
 743         res += 1LL << (bits-2);
 744         res += 1LL << (2*(bits-2));
 745         res += 1LL << (3*(bits-2));
 746         /* Does block tree limit file size? */
 747         if (res < upper_limit)
 748                 goto check_lfs;
 749
 750         res = upper_limit;
 751         /* How many metadata blocks are needed for addressing upper_limit? */
 752         upper_limit -= EXT2_NDIR_BLOCKS;
 753         /* indirect blocks */
 754         meta_blocks = 1;
 755         upper_limit -= ppb;
 756         /* double indirect blocks */
 757         if (upper_limit < ppb * ppb) {
 758                 meta_blocks += 1 + DIV_ROUND_UP(upper_limit, ppb);
 759                 res -= meta_blocks;
 760                 goto check_lfs;
 761         }
 762         meta_blocks += 1 + ppb;
 763         upper_limit -= ppb * ppb;
 764         /* tripple indirect blocks for the rest */
 765         meta_blocks += 1 + DIV_ROUND_UP(upper_limit, ppb) +
 766                 DIV_ROUND_UP(upper_limit, ppb*ppb);
 767         res -= meta_blocks;
 768 check_lfs:
 769         res <<= bits;
 770         if (res > MAX_LFS_FILESIZE)
 771                 res = MAX_LFS_FILESIZE;
 772
 773         return res;
 774 }
 775
 776 static unsigned long descriptor_loc(struct super_block *sb,
 777                                     unsigned long logic_sb_block,
 778                                     int nr)
 779 {
 780         struct ext2_sb_info *sbi = EXT2_SB(sb);
 781         unsigned long bg, first_meta_bg;
 782         int has_super = 0;
 783
 784         first_meta_bg = le32_to_cpu(sbi->s_es->s_first_meta_bg);
 785
 786         if (!EXT2_HAS_INCOMPAT_FEATURE(sb, EXT2_FEATURE_INCOMPAT_META_BG) ||
 787             nr < first_meta_bg)
 788                 return (logic_sb_block + nr + 1);
 789         bg = sbi->s_desc_per_block * nr;
 790         if (ext2_bg_has_super(sb, bg))
 791                 has_super = 1;
 792
 793         return ext2_group_first_block_no(sb, bg) + has_super;
 794 }
 795
 796 static int ext2_fill_super(struct super_block *sb, void *data, int silent)
 797 {
 798         struct buffer_head * bh;
 799         struct ext2_sb_info * sbi;
 800         struct ext2_super_block * es;
 801         struct inode *root;
 802         unsigned long block;
 803         unsigned long sb_block = get_sb_block(&data);
 804         unsigned long logic_sb_block;
 805         unsigned long offset = 0;
 806         unsigned long def_mount_opts;
 807         long ret = -EINVAL;
 808         int blocksize = BLOCK_SIZE;
 809         int db_count;
 810         int i, j;
 811         __le32 features;
 812         int err;
 813
 814         err = -ENOMEM;
 815         sbi = kzalloc(sizeof(*sbi), GFP_KERNEL);
 816         if (!sbi)
 817                 goto failed;
 818
 819         sbi->s_blockgroup_lock =
 820                 kzalloc(sizeof(struct blockgroup_lock), GFP_KERNEL);
 821         if (!sbi->s_blockgroup_lock) {
 822                 kfree(sbi);
 823                 goto failed;
 824         }
 825         sb->s_fs_info = sbi;
 826         sbi->s_sb_block = sb_block;
 827
 828         spin_lock_init(&sbi->s_lock);
 829
 830         /*
 831          * See what the current blocksize for the device is, and
 832          * use that as the blocksize.  Otherwise (or if the blocksize
 833          * is smaller than the default) use the default.
 834          * This is important for devices that have a hardware
 835          * sectorsize that is larger than the default.
 836          */
 837         blocksize = sb_min_blocksize(sb, BLOCK_SIZE);
 838         if (!blocksize) {
 839                 ext2_msg(sb, KERN_ERR, "error: unable to set blocksize");
 840                 goto failed_sbi;
 841         }
 842
 843         /*
 844          * If the superblock doesn't start on a hardware sector boundary,
 845          * calculate the offset.
 846          */
 847         if (blocksize != BLOCK_SIZE) {
 848                 logic_sb_block = (sb_block*BLOCK_SIZE) / blocksize;
 849                 offset = (sb_block*BLOCK_SIZE) % blocksize;
 850         } else {
 851                 logic_sb_block = sb_block;
 852         }
 853
 854         if (!(bh = sb_bread(sb, logic_sb_block))) {
 855                 ext2_msg(sb, KERN_ERR, "error: unable to read superblock");
 856                 goto failed_sbi;
 857         }
 858         /*
 859          * Note: s_es must be initialized as soon as possible because
 860          *       some ext2 macro-instructions depend on its value
 861          */
 862         es = (struct ext2_super_block *) (((char *)bh->b_data) + offset);
 863         sbi->s_es = es;
 864         sb->s_magic = le16_to_cpu(es->s_magic);
 865
 866         if (sb->s_magic != EXT2_SUPER_MAGIC)
 867                 goto cantfind_ext2;
 868
 869         /* Set defaults before we parse the mount options */
 870         def_mount_opts = le32_to_cpu(es->s_default_mount_opts);
 871         if (def_mount_opts & EXT2_DEFM_DEBUG)
 872                 set_opt(sbi->s_mount_opt, DEBUG);
 873         if (def_mount_opts & EXT2_DEFM_BSDGROUPS)
 874                 set_opt(sbi->s_mount_opt, GRPID);
 875         if (def_mount_opts & EXT2_DEFM_UID16)
 876                 set_opt(sbi->s_mount_opt, NO_UID32);
 877 #ifdef CONFIG_EXT2_FS_XATTR
 878         if (def_mount_opts & EXT2_DEFM_XATTR_USER)
 879                 set_opt(sbi->s_mount_opt, XATTR_USER);
 880 #endif
 881 #ifdef CONFIG_EXT2_FS_POSIX_ACL
 882         if (def_mount_opts & EXT2_DEFM_ACL)
 883                 set_opt(sbi->s_mount_opt, POSIX_ACL);
 884 #endif
 885
 886         if (le16_to_cpu(sbi->s_es->s_errors) == EXT2_ERRORS_PANIC)
 887                 set_opt(sbi->s_mount_opt, ERRORS_PANIC);
 888         else if (le16_to_cpu(sbi->s_es->s_errors) == EXT2_ERRORS_CONTINUE)
 889                 set_opt(sbi->s_mount_opt, ERRORS_CONT);
 890         else
 891                 set_opt(sbi->s_mount_opt, ERRORS_RO);
 892
 893         sbi->s_resuid = make_kuid(&init_user_ns, le16_to_cpu(es->s_def_resuid));
 894         sbi->s_resgid = make_kgid(&init_user_ns, le16_to_cpu(es->s_def_resgid));
 895
 896         set_opt(sbi->s_mount_opt, RESERVATION);
 897
 898         if (!parse_options((char *) data, sb))
 899                 goto failed_mount;
 900
 901         sb->s_flags = (sb->s_flags & ~MS_POSIXACL) |
 902                 ((EXT2_SB(sb)->s_mount_opt & EXT2_MOUNT_POSIX_ACL) ?
 903                  MS_POSIXACL : 0);
 904         sb->s_iflags |= SB_I_CGROUPWB;
 905
 906         if (le32_to_cpu(es->s_rev_level) == EXT2_GOOD_OLD_REV &&
 907             (EXT2_HAS_COMPAT_FEATURE(sb, ~0U) ||
 908              EXT2_HAS_RO_COMPAT_FEATURE(sb, ~0U) ||
 909              EXT2_HAS_INCOMPAT_FEATURE(sb, ~0U)))
 910                 ext2_msg(sb, KERN_WARNING,
 911                         "warning: feature flags set on rev 0 fs, "
 912                         "running e2fsck is recommended");
 913         /*
 914          * Check feature flags regardless of the revision level, since we
 915          * previously didn't change the revision level when setting the flags,
 916          * so there is a chance incompat flags are set on a rev 0 filesystem.
 917          */
 918         features = EXT2_HAS_INCOMPAT_FEATURE(sb, ~EXT2_FEATURE_INCOMPAT_SUPP);
 919         if (features) {
 920                 ext2_msg(sb, KERN_ERR,  "error: couldn't mount because of "
 921                        "unsupported optional features (%x)",
 922                         le32_to_cpu(features));
 923                 goto failed_mount;
 924         }
 925         if (!(sb->s_flags & MS_RDONLY) &&
 926             (features = EXT2_HAS_RO_COMPAT_FEATURE(sb, ~EXT2_FEATURE_RO_COMPAT_SUPP))){
 927                 ext2_msg(sb, KERN_ERR, "error: couldn't mount RDWR because of "
 928                        "unsupported optional features (%x)",
 929                        le32_to_cpu(features));
 930                 goto failed_mount;
 931         }
 932
 933         blocksize = BLOCK_SIZE << le32_to_cpu(sbi->s_es->s_log_block_size);
 934
 935         if (sbi->s_mount_opt & EXT2_MOUNT_DAX) {
 936                 if (blocksize != PAGE_SIZE) {
 937                         ext2_msg(sb, KERN_ERR,
 938                                         "error: unsupported blocksize for dax");
 939                         goto failed_mount;
 940                 }
 941                 if (!sb->s_bdev->bd_disk->fops->direct_access) {
 942                         ext2_msg(sb, KERN_ERR,
 943                                         "error: device does not support dax");
 944                         goto failed_mount;
 945                 }
 946         }
 947
 948         /* If the blocksize doesn't match, re-read the thing.. */
 949         if (sb->s_blocksize != blocksize) {
 950                 brelse(bh);
 951
 952                 if (!sb_set_blocksize(sb, blocksize)) {
 953                         ext2_msg(sb, KERN_ERR,
 954                                 "error: bad blocksize %d", blocksize);
 955                         goto failed_sbi;
 956                 }
 957
 958                 logic_sb_block = (sb_block*BLOCK_SIZE) / blocksize;
 959                 offset = (sb_block*BLOCK_SIZE) % blocksize;
 960                 bh = sb_bread(sb, logic_sb_block);
 961                 if(!bh) {
 962                         ext2_msg(sb, KERN_ERR, "error: couldn't read"
 963                                 "superblock on 2nd try");
 964                         goto failed_sbi;
 965                 }
 966                 es = (struct ext2_super_block *) (((char *)bh->b_data) + offset);
 967                 sbi->s_es = es;
 968                 if (es->s_magic != cpu_to_le16(EXT2_SUPER_MAGIC)) {
 969                         ext2_msg(sb, KERN_ERR, "error: magic mismatch");
 970                         goto failed_mount;
 971                 }
 972         }
 973
 974         sb->s_maxbytes = ext2_max_size(sb->s_blocksize_bits);
 975         sb->s_max_links = EXT2_LINK_MAX;
 976
 977         if (le32_to_cpu(es->s_rev_level) == EXT2_GOOD_OLD_REV) {
 978                 sbi->s_inode_size = EXT2_GOOD_OLD_INODE_SIZE;
 979                 sbi->s_first_ino = EXT2_GOOD_OLD_FIRST_INO;
 980         } else {
 981                 sbi->s_inode_size = le16_to_cpu(es->s_inode_size);
 982                 sbi->s_first_ino = le32_to_cpu(es->s_first_ino);
 983                 if ((sbi->s_inode_size < EXT2_GOOD_OLD_INODE_SIZE) ||
 984                     !is_power_of_2(sbi->s_inode_size) ||
 985                     (sbi->s_inode_size > blocksize)) {
 986                         ext2_msg(sb, KERN_ERR,
 987                                 "error: unsupported inode size: %d",
 988                                 sbi->s_inode_size);
 989                         goto failed_mount;
 990                 }
 991         }
 992
 993         sbi->s_frag_size = EXT2_MIN_FRAG_SIZE <<
 994                                    le32_to_cpu(es->s_log_frag_size);
 995         if (sbi->s_frag_size == 0)
 996                 goto cantfind_ext2;
 997         sbi->s_frags_per_block = sb->s_blocksize / sbi->s_frag_size;
 998
 999         sbi->s_blocks_per_group = le32_to_cpu(es->s_blocks_per_group);
1000         sbi->s_frags_per_group = le32_to_cpu(es->s_frags_per_group);
1001         sbi->s_inodes_per_group = le32_to_cpu(es->s_inodes_per_group);
1002
1003         if (EXT2_INODE_SIZE(sb) == 0)
1004                 goto cantfind_ext2;
1005         sbi->s_inodes_per_block = sb->s_blocksize / EXT2_INODE_SIZE(sb);
1006         if (sbi->s_inodes_per_block == 0 || sbi->s_inodes_per_group == 0)
1007                 goto cantfind_ext2;
1008         sbi->s_itb_per_group = sbi->s_inodes_per_group /
1009                                         sbi->s_inodes_per_block;
1010         sbi->s_desc_per_block = sb->s_blocksize /
1011                                         sizeof (struct ext2_group_desc);
1012         sbi->s_sbh = bh;
1013         sbi->s_mount_state = le16_to_cpu(es->s_state);
1014         sbi->s_addr_per_block_bits =
1015                 ilog2 (EXT2_ADDR_PER_BLOCK(sb));
1016         sbi->s_desc_per_block_bits =
1017                 ilog2 (EXT2_DESC_PER_BLOCK(sb));
1018
1019         if (sb->s_magic != EXT2_SUPER_MAGIC)
1020                 goto cantfind_ext2;
1021
1022         if (sb->s_blocksize != bh->b_size) {
1023                 if (!silent)
1024                         ext2_msg(sb, KERN_ERR, "error: unsupported blocksize");
1025                 goto failed_mount;
1026         }
1027
1028         if (sb->s_blocksize != sbi->s_frag_size) {
1029                 ext2_msg(sb, KERN_ERR,
1030                         "error: fragsize %lu != blocksize %lu"
1031                         "(not supported yet)",
1032                         sbi->s_frag_size, sb->s_blocksize);
1033                 goto failed_mount;
1034         }
1035
1036         if (sbi->s_blocks_per_group > sb->s_blocksize * 8) {
1037                 ext2_msg(sb, KERN_ERR,
1038                         "error: #blocks per group too big: %lu",
1039                         sbi->s_blocks_per_group);
1040                 goto failed_mount;
1041         }
1042         if (sbi->s_frags_per_group > sb->s_blocksize * 8) {
1043                 ext2_msg(sb, KERN_ERR,
1044                         "error: #fragments per group too big: %lu",
1045                         sbi->s_frags_per_group);
1046                 goto failed_mount;
1047         }
1048         if (sbi->s_inodes_per_group > sb->s_blocksize * 8) {
1049                 ext2_msg(sb, KERN_ERR,
1050                         "error: #inodes per group too big: %lu",
1051                         sbi->s_inodes_per_group);
1052                 goto failed_mount;
1053         }
1054
1055         if (EXT2_BLOCKS_PER_GROUP(sb) == 0)
1056                 goto cantfind_ext2;
1057         sbi->s_groups_count = ((le32_to_cpu(es->s_blocks_count) -
1058                                 le32_to_cpu(es->s_first_data_block) - 1)
1059                                         / EXT2_BLOCKS_PER_GROUP(sb)) + 1;
1060         db_count = (sbi->s_groups_count + EXT2_DESC_PER_BLOCK(sb) - 1) /
1061                    EXT2_DESC_PER_BLOCK(sb);
1062         sbi->s_group_desc = kmalloc (db_count * sizeof (struct buffer_head *), GFP_KERNEL);
1063         if (sbi->s_group_desc == NULL) {
1064                 ext2_msg(sb, KERN_ERR, "error: not enough memory");
1065                 goto failed_mount;
1066         }
1067         bgl_lock_init(sbi->s_blockgroup_lock);
1068         sbi->s_debts = kcalloc(sbi->s_groups_count, sizeof(*sbi->s_debts), GFP_KERNEL);
1069         if (!sbi->s_debts) {
1070                 ext2_msg(sb, KERN_ERR, "error: not enough memory");
1071                 goto failed_mount_group_desc;
1072         }
1073         for (i = 0; i < db_count; i++) {
1074                 block = descriptor_loc(sb, logic_sb_block, i);
1075                 sbi->s_group_desc[i] = sb_bread(sb, block);
1076                 if (!sbi->s_group_desc[i]) {
1077                         for (j = 0; j < i; j++)
1078                                 brelse (sbi->s_group_desc[j]);
1079                         ext2_msg(sb, KERN_ERR,
1080                                 "error: unable to read group descriptors");
1081                         goto failed_mount_group_desc;
1082                 }
1083         }
1084         if (!ext2_check_descriptors (sb)) {
1085                 ext2_msg(sb, KERN_ERR, "group descriptors corrupted");
1086                 goto failed_mount2;
1087         }
1088         sbi->s_gdb_count = db_count;
1089         get_random_bytes(&sbi->s_next_generation, sizeof(u32));
1090         spin_lock_init(&sbi->s_next_gen_lock);
1091
1092         /* per fileystem reservation list head & lock */
1093         spin_lock_init(&sbi->s_rsv_window_lock);
1094         sbi->s_rsv_window_root = RB_ROOT;
1095         /*
1096          * Add a single, static dummy reservation to the start of the
1097          * reservation window list --- it gives us a placeholder for
1098          * append-at-start-of-list which makes the allocation logic
1099          * _much_ simpler.
1100          */
1101         sbi->s_rsv_window_head.rsv_start = EXT2_RESERVE_WINDOW_NOT_ALLOCATED;
1102         sbi->s_rsv_window_head.rsv_end = EXT2_RESERVE_WINDOW_NOT_ALLOCATED;
1103         sbi->s_rsv_window_head.rsv_alloc_hit = 0;
1104         sbi->s_rsv_window_head.rsv_goal_size = 0;
1105         ext2_rsv_window_add(sb, &sbi->s_rsv_window_head);
1106
1107         err = percpu_counter_init(&sbi->s_freeblocks_counter,
1108                                 ext2_count_free_blocks(sb), GFP_KERNEL);
1109         if (!err) {
1110                 err = percpu_counter_init(&sbi->s_freeinodes_counter,
1111                                 ext2_count_free_inodes(sb), GFP_KERNEL);
1112         }
1113         if (!err) {
1114                 err = percpu_counter_init(&sbi->s_dirs_counter,
1115                                 ext2_count_dirs(sb), GFP_KERNEL);
1116         }
1117         if (err) {
1118                 ext2_msg(sb, KERN_ERR, "error: insufficient memory");
1119                 goto failed_mount3;
1120         }
1121
1122 #ifdef CONFIG_EXT2_FS_XATTR
1123         sbi->s_mb_cache = ext2_xattr_create_cache();
1124         if (!sbi->s_mb_cache) {
1125                 ext2_msg(sb, KERN_ERR, "Failed to create an mb_cache");
1126                 goto failed_mount3;
1127         }
1128 #endif
1129         /*
1130          * set up enough so that it can read an inode
1131          */
1132         sb->s_op = &ext2_sops;
1133         sb->s_export_op = &ext2_export_ops;
1134         sb->s_xattr = ext2_xattr_handlers;
1135
1136 #ifdef CONFIG_QUOTA
1137         sb->dq_op = &dquot_operations;
1138         sb->s_qcop = &dquot_quotactl_ops;
1139         sb->s_quota_types = QTYPE_MASK_USR | QTYPE_MASK_GRP;
1140 #endif
1141
1142         root = ext2_iget(sb, EXT2_ROOT_INO);
1143         if (IS_ERR(root)) {
1144                 ret = PTR_ERR(root);
1145                 goto failed_mount3;
1146         }
1147         if (!S_ISDIR(root->i_mode) || !root->i_blocks || !root->i_size) {
1148                 iput(root);
1149                 ext2_msg(sb, KERN_ERR, "error: corrupt root inode, run e2fsck");
1150                 goto failed_mount3;
1151         }
1152
1153         sb->s_root = d_make_root(root);
1154         if (!sb->s_root) {
1155                 ext2_msg(sb, KERN_ERR, "error: get root inode failed");
1156                 ret = -ENOMEM;
1157                 goto failed_mount3;
1158         }
1159         if (EXT2_HAS_COMPAT_FEATURE(sb, EXT3_FEATURE_COMPAT_HAS_JOURNAL))
1160                 ext2_msg(sb, KERN_WARNING,
1161                         "warning: mounting ext3 filesystem as ext2");
1162         if (ext2_setup_super (sb, es, sb->s_flags & MS_RDONLY))
1163                 sb->s_flags |= MS_RDONLY;
1164         ext2_write_super(sb);
1165         return 0;
1166
1167 cantfind_ext2:
1168         if (!silent)
1169                 ext2_msg(sb, KERN_ERR,
1170                         "error: can't find an ext2 filesystem on dev %s.",
1171                         sb->s_id);
1172         goto failed_mount;
1173 failed_mount3:
1174         if (sbi->s_mb_cache)
1175                 ext2_xattr_destroy_cache(sbi->s_mb_cache);
1176         percpu_counter_destroy(&sbi->s_freeblocks_counter);
1177         percpu_counter_destroy(&sbi->s_freeinodes_counter);
1178         percpu_counter_destroy(&sbi->s_dirs_counter);
1179 failed_mount2:
1180         for (i = 0; i < db_count; i++)
1181                 brelse(sbi->s_group_desc[i]);
1182 failed_mount_group_desc:
1183         kfree(sbi->s_group_desc);
1184         kfree(sbi->s_debts);
1185 failed_mount:
1186         brelse(bh);
1187 failed_sbi:
1188         sb->s_fs_info = NULL;
1189         kfree(sbi->s_blockgroup_lock);
1190         kfree(sbi);
1191 failed:
1192         return ret;
1193 }
1194
1195 static void ext2_clear_super_error(struct super_block *sb)
1196 {
1197         struct buffer_head *sbh = EXT2_SB(sb)->s_sbh;
1198
1199         if (buffer_write_io_error(sbh)) {
1200                 /*
1201                  * Oh, dear.  A previous attempt to write the
1202                  * superblock failed.  This could happen because the
1203                  * USB device was yanked out.  Or it could happen to
1204                  * be a transient write error and maybe the block will
1205                  * be remapped.  Nothing we can do but to retry the
1206                  * write and hope for the best.
1207                  */
1208                 ext2_msg(sb, KERN_ERR,
1209                        "previous I/O error to superblock detected\n");
1210                 clear_buffer_write_io_error(sbh);
1211                 set_buffer_uptodate(sbh);
1212         }
1213 }
1214
1215 static void ext2_sync_super(struct super_block *sb, struct ext2_super_block *es,
1216                             int wait)
1217 {
1218         ext2_clear_super_error(sb);
1219         spin_lock(&EXT2_SB(sb)->s_lock);
1220         es->s_free_blocks_count = cpu_to_le32(ext2_count_free_blocks(sb));
1221         es->s_free_inodes_count = cpu_to_le32(ext2_count_free_inodes(sb));
1222         es->s_wtime = cpu_to_le32(get_seconds());
1223         /* unlock before we do IO */
1224         spin_unlock(&EXT2_SB(sb)->s_lock);
1225         mark_buffer_dirty(EXT2_SB(sb)->s_sbh);
1226         if (wait)
1227                 sync_dirty_buffer(EXT2_SB(sb)->s_sbh);
1228 }
1229
1230 /*
1231  * In the second extended file system, it is not necessary to
1232  * write the super block since we use a mapping of the
1233  * disk super block in a buffer.
1234  *
1235  * However, this function is still used to set the fs valid
1236  * flags to 0.  We need to set this flag to 0 since the fs
1237  * may have been checked while mounted and e2fsck may have
1238  * set s_state to EXT2_VALID_FS after some corrections.
1239  */
1240 static int ext2_sync_fs(struct super_block *sb, int wait)
1241 {
1242         struct ext2_sb_info *sbi = EXT2_SB(sb);
1243         struct ext2_super_block *es = EXT2_SB(sb)->s_es;
1244
1245         /*
1246          * Write quota structures to quota file, sync_blockdev() will write
1247          * them to disk later
1248          */
1249         dquot_writeback_dquots(sb, -1);
1250
1251         spin_lock(&sbi->s_lock);
1252         if (es->s_state & cpu_to_le16(EXT2_VALID_FS)) {
1253                 ext2_debug("setting valid to 0\n");
1254                 es->s_state &= cpu_to_le16(~EXT2_VALID_FS);
1255         }
1256         spin_unlock(&sbi->s_lock);
1257         ext2_sync_super(sb, es, wait);
1258         return 0;
1259 }
1260
1261 static int ext2_freeze(struct super_block *sb)
1262 {
1263         struct ext2_sb_info *sbi = EXT2_SB(sb);
1264
1265         /*
1266          * Open but unlinked files present? Keep EXT2_VALID_FS flag cleared
1267          * because we have unattached inodes and thus filesystem is not fully
1268          * consistent.
1269          */
1270         if (atomic_long_read(&sb->s_remove_count)) {
1271                 ext2_sync_fs(sb, 1);
1272                 return 0;
1273         }
1274         /* Set EXT2_FS_VALID flag */
1275         spin_lock(&sbi->s_lock);
1276         sbi->s_es->s_state = cpu_to_le16(sbi->s_mount_state);
1277         spin_unlock(&sbi->s_lock);
1278         ext2_sync_super(sb, sbi->s_es, 1);
1279
1280         return 0;
1281 }
1282
1283 static int ext2_unfreeze(struct super_block *sb)
1284 {
1285         /* Just write sb to clear EXT2_VALID_FS flag */
1286         ext2_write_super(sb);
1287
1288         return 0;
1289 }
1290
1291 void ext2_write_super(struct super_block *sb)
1292 {
1293         if (!(sb->s_flags & MS_RDONLY))
1294                 ext2_sync_fs(sb, 1);
1295 }
1296
1297 static int ext2_remount (struct super_block * sb, int * flags, char * data)
1298 {
1299         struct ext2_sb_info * sbi = EXT2_SB(sb);
1300         struct ext2_super_block * es;
1301         struct ext2_mount_options old_opts;
1302         unsigned long old_sb_flags;
1303         int err;
1304
1305         sync_filesystem(sb);
1306         spin_lock(&sbi->s_lock);
1307
1308         /* Store the old options */
1309         old_sb_flags = sb->s_flags;
1310         old_opts.s_mount_opt = sbi->s_mount_opt;
1311         old_opts.s_resuid = sbi->s_resuid;
1312         old_opts.s_resgid = sbi->s_resgid;
1313
1314         /*
1315          * Allow the "check" option to be passed as a remount option.
1316          */
1317         if (!parse_options(data, sb)) {
1318                 err = -EINVAL;
1319                 goto restore_opts;
1320         }
1321
1322         sb->s_flags = (sb->s_flags & ~MS_POSIXACL) |
1323                 ((sbi->s_mount_opt & EXT2_MOUNT_POSIX_ACL) ? MS_POSIXACL : 0);
1324
1325         es = sbi->s_es;
1326         if ((sbi->s_mount_opt ^ old_opts.s_mount_opt) & EXT2_MOUNT_DAX) {
1327                 ext2_msg(sb, KERN_WARNING, "warning: refusing change of "
1328                          "dax flag with busy inodes while remounting");
1329                 sbi->s_mount_opt ^= EXT2_MOUNT_DAX;
1330         }
1331         if ((*flags & MS_RDONLY) == (sb->s_flags & MS_RDONLY)) {
1332                 spin_unlock(&sbi->s_lock);
1333                 return 0;
1334         }
1335         if (*flags & MS_RDONLY) {
1336                 if (le16_to_cpu(es->s_state) & EXT2_VALID_FS ||
1337                     !(sbi->s_mount_state & EXT2_VALID_FS)) {
1338                         spin_unlock(&sbi->s_lock);
1339                         return 0;
1340                 }
1341
1342                 /*
1343                  * OK, we are remounting a valid rw partition rdonly, so set
1344                  * the rdonly flag and then mark the partition as valid again.
1345                  */
1346                 es->s_state = cpu_to_le16(sbi->s_mount_state);
1347                 es->s_mtime = cpu_to_le32(get_seconds());
1348                 spin_unlock(&sbi->s_lock);
1349
1350                 err = dquot_suspend(sb, -1);
1351                 if (err < 0) {
1352                         spin_lock(&sbi->s_lock);
1353                         goto restore_opts;
1354                 }
1355
1356                 ext2_sync_super(sb, es, 1);
1357         } else {
1358                 __le32 ret = EXT2_HAS_RO_COMPAT_FEATURE(sb,
1359                                                ~EXT2_FEATURE_RO_COMPAT_SUPP);
1360                 if (ret) {
1361                         ext2_msg(sb, KERN_WARNING,
1362                                 "warning: couldn't remount RDWR because of "
1363                                 "unsupported optional features (%x).",
1364                                 le32_to_cpu(ret));
1365                         err = -EROFS;
1366                         goto restore_opts;
1367                 }
1368                 /*
1369                  * Mounting a RDONLY partition read-write, so reread and
1370                  * store the current valid flag.  (It may have been changed
1371                  * by e2fsck since we originally mounted the partition.)
1372                  */
1373                 sbi->s_mount_state = le16_to_cpu(es->s_state);
1374                 if (!ext2_setup_super (sb, es, 0))
1375                         sb->s_flags &= ~MS_RDONLY;
1376                 spin_unlock(&sbi->s_lock);
1377
1378                 ext2_write_super(sb);
1379
1380                 dquot_resume(sb, -1);
1381         }
1382
1383         return 0;
1384 restore_opts:
1385         sbi->s_mount_opt = old_opts.s_mount_opt;
1386         sbi->s_resuid = old_opts.s_resuid;
1387         sbi->s_resgid = old_opts.s_resgid;
1388         sb->s_flags = old_sb_flags;
1389         spin_unlock(&sbi->s_lock);
1390         return err;
1391 }
1392
1393 static int ext2_statfs (struct dentry * dentry, struct kstatfs * buf)
1394 {
1395         struct super_block *sb = dentry->d_sb;
1396         struct ext2_sb_info *sbi = EXT2_SB(sb);
1397         struct ext2_super_block *es = sbi->s_es;
1398         u64 fsid;
1399
1400         spin_lock(&sbi->s_lock);
1401
1402         if (test_opt (sb, MINIX_DF))
1403                 sbi->s_overhead_last = 0;
1404         else if (sbi->s_blocks_last != le32_to_cpu(es->s_blocks_count)) {
1405                 unsigned long i, overhead = 0;
1406                 smp_rmb();
1407
1408                 /*
1409                  * Compute the overhead (FS structures). This is constant
1410                  * for a given filesystem unless the number of block groups
1411                  * changes so we cache the previous value until it does.
1412                  */
1413
1414                 /*
1415                  * All of the blocks before first_data_block are
1416                  * overhead
1417                  */
1418                 overhead = le32_to_cpu(es->s_first_data_block);
1419
1420                 /*
1421                  * Add the overhead attributed to the superblock and
1422                  * block group descriptors.  If the sparse superblocks
1423                  * feature is turned on, then not all groups have this.
1424                  */
1425                 for (i = 0; i < sbi->s_groups_count; i++)
1426                         overhead += ext2_bg_has_super(sb, i) +
1427                                 ext2_bg_num_gdb(sb, i);
1428
1429                 /*
1430                  * Every block group has an inode bitmap, a block
1431                  * bitmap, and an inode table.
1432                  */
1433                 overhead += (sbi->s_groups_count *
1434                              (2 + sbi->s_itb_per_group));
1435                 sbi->s_overhead_last = overhead;
1436                 smp_wmb();
1437                 sbi->s_blocks_last = le32_to_cpu(es->s_blocks_count);
1438         }
1439
1440         buf->f_type = EXT2_SUPER_MAGIC;
1441         buf->f_bsize = sb->s_blocksize;
1442         buf->f_blocks = le32_to_cpu(es->s_blocks_count) - sbi->s_overhead_last;
1443         buf->f_bfree = ext2_count_free_blocks(sb);
1444         es->s_free_blocks_count = cpu_to_le32(buf->f_bfree);
1445         buf->f_bavail = buf->f_bfree - le32_to_cpu(es->s_r_blocks_count);
1446         if (buf->f_bfree < le32_to_cpu(es->s_r_blocks_count))
1447                 buf->f_bavail = 0;
1448         buf->f_files = le32_to_cpu(es->s_inodes_count);
1449         buf->f_ffree = ext2_count_free_inodes(sb);
1450         es->s_free_inodes_count = cpu_to_le32(buf->f_ffree);
1451         buf->f_namelen = EXT2_NAME_LEN;
1452         fsid = le64_to_cpup((void *)es->s_uuid) ^
1453                le64_to_cpup((void *)es->s_uuid + sizeof(u64));
1454         buf->f_fsid.val[0] = fsid & 0xFFFFFFFFUL;
1455         buf->f_fsid.val[1] = (fsid >> 32) & 0xFFFFFFFFUL;
1456         spin_unlock(&sbi->s_lock);
1457         return 0;
1458 }
1459
1460 static struct dentry *ext2_mount(struct file_system_type *fs_type,
1461         int flags, const char *dev_name, void *data)
1462 {
1463         return mount_bdev(fs_type, flags, dev_name, data, ext2_fill_super);
1464 }
1465
1466 #ifdef CONFIG_QUOTA
1467
1468 /* Read data from quotafile - avoid pagecache and such because we cannot afford
1469  * acquiring the locks... As quota files are never truncated and quota code
1470  * itself serializes the operations (and no one else should touch the files)
1471  * we don't have to be afraid of races */
1472 static ssize_t ext2_quota_read(struct super_block *sb, int type, char *data,
1473                                size_t len, loff_t off)
1474 {
1475         struct inode *inode = sb_dqopt(sb)->files[type];
1476         sector_t blk = off >> EXT2_BLOCK_SIZE_BITS(sb);
1477         int err = 0;
1478         int offset = off & (sb->s_blocksize - 1);
1479         int tocopy;
1480         size_t toread;
1481         struct buffer_head tmp_bh;
1482         struct buffer_head *bh;
1483         loff_t i_size = i_size_read(inode);
1484
1485         if (off > i_size)
1486                 return 0;
1487         if (off+len > i_size)
1488                 len = i_size-off;
1489         toread = len;
1490         while (toread > 0) {
1491                 tocopy = sb->s_blocksize - offset < toread ?
1492                                 sb->s_blocksize - offset : toread;
1493
1494                 tmp_bh.b_state = 0;
1495                 tmp_bh.b_size = sb->s_blocksize;
1496                 err = ext2_get_block(inode, blk, &tmp_bh, 0);
1497                 if (err < 0)
1498                         return err;
1499                 if (!buffer_mapped(&tmp_bh))    /* A hole? */
1500                         memset(data, 0, tocopy);
1501                 else {
1502                         bh = sb_bread(sb, tmp_bh.b_blocknr);
1503                         if (!bh)
1504                                 return -EIO;
1505                         memcpy(data, bh->b_data+offset, tocopy);
1506                         brelse(bh);
1507                 }
1508                 offset = 0;
1509                 toread -= tocopy;
1510                 data += tocopy;
1511                 blk++;
1512         }
1513         return len;
1514 }
1515
1516 /* Write to quotafile */
1517 static ssize_t ext2_quota_write(struct super_block *sb, int type,
1518                                 const char *data, size_t len, loff_t off)
1519 {
1520         struct inode *inode = sb_dqopt(sb)->files[type];
1521         sector_t blk = off >> EXT2_BLOCK_SIZE_BITS(sb);
1522         int err = 0;
1523         int offset = off & (sb->s_blocksize - 1);
1524         int tocopy;
1525         size_t towrite = len;
1526         struct buffer_head tmp_bh;
1527         struct buffer_head *bh;
1528
1529         while (towrite > 0) {
1530                 tocopy = sb->s_blocksize - offset < towrite ?
1531                                 sb->s_blocksize - offset : towrite;
1532
1533                 tmp_bh.b_state = 0;
1534                 tmp_bh.b_size = sb->s_blocksize;
1535                 err = ext2_get_block(inode, blk, &tmp_bh, 1);
1536                 if (err < 0)
1537                         goto out;
1538                 if (offset || tocopy != EXT2_BLOCK_SIZE(sb))
1539                         bh = sb_bread(sb, tmp_bh.b_blocknr);
1540                 else
1541                         bh = sb_getblk(sb, tmp_bh.b_blocknr);
1542                 if (unlikely(!bh)) {
1543                         err = -EIO;
1544                         goto out;
1545                 }
1546                 lock_buffer(bh);
1547                 memcpy(bh->b_data+offset, data, tocopy);
1548                 flush_dcache_page(bh->b_page);
1549                 set_buffer_uptodate(bh);
1550                 mark_buffer_dirty(bh);
1551                 unlock_buffer(bh);
1552                 brelse(bh);
1553                 offset = 0;
1554                 towrite -= tocopy;
1555                 data += tocopy;
1556                 blk++;
1557         }
1558 out:
1559         if (len == towrite)
1560                 return err;
1561         if (inode->i_size < off+len-towrite)
1562                 i_size_write(inode, off+len-towrite);
1563         inode->i_version++;
1564         inode->i_mtime = inode->i_ctime = CURRENT_TIME;
1565         mark_inode_dirty(inode);
1566         return len - towrite;
1567 }
1568
1569 #endif
1570
1571 static struct file_system_type ext2_fs_type = {
1572         .owner          = THIS_MODULE,
1573         .name           = "ext2",
1574         .mount          = ext2_mount,
1575         .kill_sb        = kill_block_super,
1576         .fs_flags       = FS_REQUIRES_DEV,
1577 };
1578 MODULE_ALIAS_FS("ext2");
1579
1580 static int __init init_ext2_fs(void)
1581 {
1582         int err;
1583
1584         err = init_inodecache();
1585         if (err)
1586                 return err;
1587         err = register_filesystem(&ext2_fs_type);
1588         if (err)
1589                 goto out;
1590         return 0;
1591 out:
1592         destroy_inodecache();
1593         return err;
1594 }
1595
1596 static void __exit exit_ext2_fs(void)
1597 {
1598         unregister_filesystem(&ext2_fs_type);
1599         destroy_inodecache();
1600 }
1601
1602 MODULE_AUTHOR("Remy Card and others");
1603 MODULE_DESCRIPTION("Second Extended Filesystem");
1604 MODULE_LICENSE("GPL");
1605 module_init(init_ext2_fs)
1606 module_exit(exit_ext2_fs)