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|SLAB_ACCOUNT),
 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         /* Compute how many metadata blocks are needed */
 747         meta_blocks = 1;
 748         meta_blocks += 1 + ppb;
 749         meta_blocks += 1 + ppb + ppb * ppb;
 750         /* Does block tree limit file size? */
 751         if (res + meta_blocks <= upper_limit)
 752                 goto check_lfs;
 753
 754         res = upper_limit;
 755         /* How many metadata blocks are needed for addressing upper_limit? */
 756         upper_limit -= EXT2_NDIR_BLOCKS;
 757         /* indirect blocks */
 758         meta_blocks = 1;
 759         upper_limit -= ppb;
 760         /* double indirect blocks */
 761         if (upper_limit < ppb * ppb) {
 762                 meta_blocks += 1 + DIV_ROUND_UP(upper_limit, ppb);
 763                 res -= meta_blocks;
 764                 goto check_lfs;
 765         }
 766         meta_blocks += 1 + ppb;
 767         upper_limit -= ppb * ppb;
 768         /* tripple indirect blocks for the rest */
 769         meta_blocks += 1 + DIV_ROUND_UP(upper_limit, ppb) +
 770                 DIV_ROUND_UP(upper_limit, ppb*ppb);
 771         res -= meta_blocks;
 772 check_lfs:
 773         res <<= bits;
 774         if (res > MAX_LFS_FILESIZE)
 775                 res = MAX_LFS_FILESIZE;
 776
 777         return res;
 778 }
 779
 780 static unsigned long descriptor_loc(struct super_block *sb,
 781                                     unsigned long logic_sb_block,
 782                                     int nr)
 783 {
 784         struct ext2_sb_info *sbi = EXT2_SB(sb);
 785         unsigned long bg, first_meta_bg;
 786         int has_super = 0;
 787
 788         first_meta_bg = le32_to_cpu(sbi->s_es->s_first_meta_bg);
 789
 790         if (!EXT2_HAS_INCOMPAT_FEATURE(sb, EXT2_FEATURE_INCOMPAT_META_BG) ||
 791             nr < first_meta_bg)
 792                 return (logic_sb_block + nr + 1);
 793         bg = sbi->s_desc_per_block * nr;
 794         if (ext2_bg_has_super(sb, bg))
 795                 has_super = 1;
 796
 797         return ext2_group_first_block_no(sb, bg) + has_super;
 798 }
 799
 800 static int ext2_fill_super(struct super_block *sb, void *data, int silent)
 801 {
 802         struct buffer_head * bh;
 803         struct ext2_sb_info * sbi;
 804         struct ext2_super_block * es;
 805         struct inode *root;
 806         unsigned long block;
 807         unsigned long sb_block = get_sb_block(&data);
 808         unsigned long logic_sb_block;
 809         unsigned long offset = 0;
 810         unsigned long def_mount_opts;
 811         long ret = -EINVAL;
 812         int blocksize = BLOCK_SIZE;
 813         int db_count;
 814         int i, j;
 815         __le32 features;
 816         int err;
 817
 818         err = -ENOMEM;
 819         sbi = kzalloc(sizeof(*sbi), GFP_KERNEL);
 820         if (!sbi)
 821                 goto failed;
 822
 823         sbi->s_blockgroup_lock =
 824                 kzalloc(sizeof(struct blockgroup_lock), GFP_KERNEL);
 825         if (!sbi->s_blockgroup_lock) {
 826                 kfree(sbi);
 827                 goto failed;
 828         }
 829         sb->s_fs_info = sbi;
 830         sbi->s_sb_block = sb_block;
 831
 832         spin_lock_init(&sbi->s_lock);
 833
 834         /*
 835          * See what the current blocksize for the device is, and
 836          * use that as the blocksize.  Otherwise (or if the blocksize
 837          * is smaller than the default) use the default.
 838          * This is important for devices that have a hardware
 839          * sectorsize that is larger than the default.
 840          */
 841         blocksize = sb_min_blocksize(sb, BLOCK_SIZE);
 842         if (!blocksize) {
 843                 ext2_msg(sb, KERN_ERR, "error: unable to set blocksize");
 844                 goto failed_sbi;
 845         }
 846
 847         /*
 848          * If the superblock doesn't start on a hardware sector boundary,
 849          * calculate the offset.
 850          */
 851         if (blocksize != BLOCK_SIZE) {
 852                 logic_sb_block = (sb_block*BLOCK_SIZE) / blocksize;
 853                 offset = (sb_block*BLOCK_SIZE) % blocksize;
 854         } else {
 855                 logic_sb_block = sb_block;
 856         }
 857
 858         if (!(bh = sb_bread(sb, logic_sb_block))) {
 859                 ext2_msg(sb, KERN_ERR, "error: unable to read superblock");
 860                 goto failed_sbi;
 861         }
 862         /*
 863          * Note: s_es must be initialized as soon as possible because
 864          *       some ext2 macro-instructions depend on its value
 865          */
 866         es = (struct ext2_super_block *) (((char *)bh->b_data) + offset);
 867         sbi->s_es = es;
 868         sb->s_magic = le16_to_cpu(es->s_magic);
 869
 870         if (sb->s_magic != EXT2_SUPER_MAGIC)
 871                 goto cantfind_ext2;
 872
 873         /* Set defaults before we parse the mount options */
 874         def_mount_opts = le32_to_cpu(es->s_default_mount_opts);
 875         if (def_mount_opts & EXT2_DEFM_DEBUG)
 876                 set_opt(sbi->s_mount_opt, DEBUG);
 877         if (def_mount_opts & EXT2_DEFM_BSDGROUPS)
 878                 set_opt(sbi->s_mount_opt, GRPID);
 879         if (def_mount_opts & EXT2_DEFM_UID16)
 880                 set_opt(sbi->s_mount_opt, NO_UID32);
 881 #ifdef CONFIG_EXT2_FS_XATTR
 882         if (def_mount_opts & EXT2_DEFM_XATTR_USER)
 883                 set_opt(sbi->s_mount_opt, XATTR_USER);
 884 #endif
 885 #ifdef CONFIG_EXT2_FS_POSIX_ACL
 886         if (def_mount_opts & EXT2_DEFM_ACL)
 887                 set_opt(sbi->s_mount_opt, POSIX_ACL);
 888 #endif
 889
 890         if (le16_to_cpu(sbi->s_es->s_errors) == EXT2_ERRORS_PANIC)
 891                 set_opt(sbi->s_mount_opt, ERRORS_PANIC);
 892         else if (le16_to_cpu(sbi->s_es->s_errors) == EXT2_ERRORS_CONTINUE)
 893                 set_opt(sbi->s_mount_opt, ERRORS_CONT);
 894         else
 895                 set_opt(sbi->s_mount_opt, ERRORS_RO);
 896
 897         sbi->s_resuid = make_kuid(&init_user_ns, le16_to_cpu(es->s_def_resuid));
 898         sbi->s_resgid = make_kgid(&init_user_ns, le16_to_cpu(es->s_def_resgid));
 899
 900         set_opt(sbi->s_mount_opt, RESERVATION);
 901
 902         if (!parse_options((char *) data, sb))
 903                 goto failed_mount;
 904
 905         sb->s_flags = (sb->s_flags & ~MS_POSIXACL) |
 906                 ((EXT2_SB(sb)->s_mount_opt & EXT2_MOUNT_POSIX_ACL) ?
 907                  MS_POSIXACL : 0);
 908         sb->s_iflags |= SB_I_CGROUPWB;
 909
 910         if (le32_to_cpu(es->s_rev_level) == EXT2_GOOD_OLD_REV &&
 911             (EXT2_HAS_COMPAT_FEATURE(sb, ~0U) ||
 912              EXT2_HAS_RO_COMPAT_FEATURE(sb, ~0U) ||
 913              EXT2_HAS_INCOMPAT_FEATURE(sb, ~0U)))
 914                 ext2_msg(sb, KERN_WARNING,
 915                         "warning: feature flags set on rev 0 fs, "
 916                         "running e2fsck is recommended");
 917         /*
 918          * Check feature flags regardless of the revision level, since we
 919          * previously didn't change the revision level when setting the flags,
 920          * so there is a chance incompat flags are set on a rev 0 filesystem.
 921          */
 922         features = EXT2_HAS_INCOMPAT_FEATURE(sb, ~EXT2_FEATURE_INCOMPAT_SUPP);
 923         if (features) {
 924                 ext2_msg(sb, KERN_ERR,  "error: couldn't mount because of "
 925                        "unsupported optional features (%x)",
 926                         le32_to_cpu(features));
 927                 goto failed_mount;
 928         }
 929         if (!(sb->s_flags & MS_RDONLY) &&
 930             (features = EXT2_HAS_RO_COMPAT_FEATURE(sb, ~EXT2_FEATURE_RO_COMPAT_SUPP))){
 931                 ext2_msg(sb, KERN_ERR, "error: couldn't mount RDWR because of "
 932                        "unsupported optional features (%x)",
 933                        le32_to_cpu(features));
 934                 goto failed_mount;
 935         }
 936
 937         blocksize = BLOCK_SIZE << le32_to_cpu(sbi->s_es->s_log_block_size);
 938
 939         if (sbi->s_mount_opt & EXT2_MOUNT_DAX) {
 940                 err = bdev_dax_supported(sb, blocksize);
 941                 if (err)
 942                         goto failed_mount;
 943         }
 944
 945         /* If the blocksize doesn't match, re-read the thing.. */
 946         if (sb->s_blocksize != blocksize) {
 947                 brelse(bh);
 948
 949                 if (!sb_set_blocksize(sb, blocksize)) {
 950                         ext2_msg(sb, KERN_ERR,
 951                                 "error: bad blocksize %d", blocksize);
 952                         goto failed_sbi;
 953                 }
 954
 955                 logic_sb_block = (sb_block*BLOCK_SIZE) / blocksize;
 956                 offset = (sb_block*BLOCK_SIZE) % blocksize;
 957                 bh = sb_bread(sb, logic_sb_block);
 958                 if(!bh) {
 959                         ext2_msg(sb, KERN_ERR, "error: couldn't read"
 960                                 "superblock on 2nd try");
 961                         goto failed_sbi;
 962                 }
 963                 es = (struct ext2_super_block *) (((char *)bh->b_data) + offset);
 964                 sbi->s_es = es;
 965                 if (es->s_magic != cpu_to_le16(EXT2_SUPER_MAGIC)) {
 966                         ext2_msg(sb, KERN_ERR, "error: magic mismatch");
 967                         goto failed_mount;
 968                 }
 969         }
 970
 971         sb->s_maxbytes = ext2_max_size(sb->s_blocksize_bits);
 972         sb->s_max_links = EXT2_LINK_MAX;
 973
 974         if (le32_to_cpu(es->s_rev_level) == EXT2_GOOD_OLD_REV) {
 975                 sbi->s_inode_size = EXT2_GOOD_OLD_INODE_SIZE;
 976                 sbi->s_first_ino = EXT2_GOOD_OLD_FIRST_INO;
 977         } else {
 978                 sbi->s_inode_size = le16_to_cpu(es->s_inode_size);
 979                 sbi->s_first_ino = le32_to_cpu(es->s_first_ino);
 980                 if ((sbi->s_inode_size < EXT2_GOOD_OLD_INODE_SIZE) ||
 981                     !is_power_of_2(sbi->s_inode_size) ||
 982                     (sbi->s_inode_size > blocksize)) {
 983                         ext2_msg(sb, KERN_ERR,
 984                                 "error: unsupported inode size: %d",
 985                                 sbi->s_inode_size);
 986                         goto failed_mount;
 987                 }
 988         }
 989
 990         sbi->s_frag_size = EXT2_MIN_FRAG_SIZE <<
 991                                    le32_to_cpu(es->s_log_frag_size);
 992         if (sbi->s_frag_size == 0)
 993                 goto cantfind_ext2;
 994         sbi->s_frags_per_block = sb->s_blocksize / sbi->s_frag_size;
 995
 996         sbi->s_blocks_per_group = le32_to_cpu(es->s_blocks_per_group);
 997         sbi->s_frags_per_group = le32_to_cpu(es->s_frags_per_group);
 998         sbi->s_inodes_per_group = le32_to_cpu(es->s_inodes_per_group);
 999
1000         if (EXT2_INODE_SIZE(sb) == 0)
1001                 goto cantfind_ext2;
1002         sbi->s_inodes_per_block = sb->s_blocksize / EXT2_INODE_SIZE(sb);
1003         if (sbi->s_inodes_per_block == 0 || sbi->s_inodes_per_group == 0)
1004                 goto cantfind_ext2;
1005         sbi->s_itb_per_group = sbi->s_inodes_per_group /
1006                                         sbi->s_inodes_per_block;
1007         sbi->s_desc_per_block = sb->s_blocksize /
1008                                         sizeof (struct ext2_group_desc);
1009         sbi->s_sbh = bh;
1010         sbi->s_mount_state = le16_to_cpu(es->s_state);
1011         sbi->s_addr_per_block_bits =
1012                 ilog2 (EXT2_ADDR_PER_BLOCK(sb));
1013         sbi->s_desc_per_block_bits =
1014                 ilog2 (EXT2_DESC_PER_BLOCK(sb));
1015
1016         if (sb->s_magic != EXT2_SUPER_MAGIC)
1017                 goto cantfind_ext2;
1018
1019         if (sb->s_blocksize != bh->b_size) {
1020                 if (!silent)
1021                         ext2_msg(sb, KERN_ERR, "error: unsupported blocksize");
1022                 goto failed_mount;
1023         }
1024
1025         if (sb->s_blocksize != sbi->s_frag_size) {
1026                 ext2_msg(sb, KERN_ERR,
1027                         "error: fragsize %lu != blocksize %lu"
1028                         "(not supported yet)",
1029                         sbi->s_frag_size, sb->s_blocksize);
1030                 goto failed_mount;
1031         }
1032
1033         if (sbi->s_blocks_per_group > sb->s_blocksize * 8) {
1034                 ext2_msg(sb, KERN_ERR,
1035                         "error: #blocks per group too big: %lu",
1036                         sbi->s_blocks_per_group);
1037                 goto failed_mount;
1038         }
1039         if (sbi->s_frags_per_group > sb->s_blocksize * 8) {
1040                 ext2_msg(sb, KERN_ERR,
1041                         "error: #fragments per group too big: %lu",
1042                         sbi->s_frags_per_group);
1043                 goto failed_mount;
1044         }
1045         if (sbi->s_inodes_per_group > sb->s_blocksize * 8) {
1046                 ext2_msg(sb, KERN_ERR,
1047                         "error: #inodes per group too big: %lu",
1048                         sbi->s_inodes_per_group);
1049                 goto failed_mount;
1050         }
1051
1052         if (EXT2_BLOCKS_PER_GROUP(sb) == 0)
1053                 goto cantfind_ext2;
1054         sbi->s_groups_count = ((le32_to_cpu(es->s_blocks_count) -
1055                                 le32_to_cpu(es->s_first_data_block) - 1)
1056                                         / EXT2_BLOCKS_PER_GROUP(sb)) + 1;
1057         db_count = (sbi->s_groups_count + EXT2_DESC_PER_BLOCK(sb) - 1) /
1058                    EXT2_DESC_PER_BLOCK(sb);
1059         sbi->s_group_desc = kmalloc (db_count * sizeof (struct buffer_head *), GFP_KERNEL);
1060         if (sbi->s_group_desc == NULL) {
1061                 ext2_msg(sb, KERN_ERR, "error: not enough memory");
1062                 goto failed_mount;
1063         }
1064         bgl_lock_init(sbi->s_blockgroup_lock);
1065         sbi->s_debts = kcalloc(sbi->s_groups_count, sizeof(*sbi->s_debts), GFP_KERNEL);
1066         if (!sbi->s_debts) {
1067                 ext2_msg(sb, KERN_ERR, "error: not enough memory");
1068                 goto failed_mount_group_desc;
1069         }
1070         for (i = 0; i < db_count; i++) {
1071                 block = descriptor_loc(sb, logic_sb_block, i);
1072                 sbi->s_group_desc[i] = sb_bread(sb, block);
1073                 if (!sbi->s_group_desc[i]) {
1074                         for (j = 0; j < i; j++)
1075                                 brelse (sbi->s_group_desc[j]);
1076                         ext2_msg(sb, KERN_ERR,
1077                                 "error: unable to read group descriptors");
1078                         goto failed_mount_group_desc;
1079                 }
1080         }
1081         if (!ext2_check_descriptors (sb)) {
1082                 ext2_msg(sb, KERN_ERR, "group descriptors corrupted");
1083                 goto failed_mount2;
1084         }
1085         sbi->s_gdb_count = db_count;
1086         get_random_bytes(&sbi->s_next_generation, sizeof(u32));
1087         spin_lock_init(&sbi->s_next_gen_lock);
1088
1089         /* per fileystem reservation list head & lock */
1090         spin_lock_init(&sbi->s_rsv_window_lock);
1091         sbi->s_rsv_window_root = RB_ROOT;
1092         /*
1093          * Add a single, static dummy reservation to the start of the
1094          * reservation window list --- it gives us a placeholder for
1095          * append-at-start-of-list which makes the allocation logic
1096          * _much_ simpler.
1097          */
1098         sbi->s_rsv_window_head.rsv_start = EXT2_RESERVE_WINDOW_NOT_ALLOCATED;
1099         sbi->s_rsv_window_head.rsv_end = EXT2_RESERVE_WINDOW_NOT_ALLOCATED;
1100         sbi->s_rsv_window_head.rsv_alloc_hit = 0;
1101         sbi->s_rsv_window_head.rsv_goal_size = 0;
1102         ext2_rsv_window_add(sb, &sbi->s_rsv_window_head);
1103
1104         err = percpu_counter_init(&sbi->s_freeblocks_counter,
1105                                 ext2_count_free_blocks(sb), GFP_KERNEL);
1106         if (!err) {
1107                 err = percpu_counter_init(&sbi->s_freeinodes_counter,
1108                                 ext2_count_free_inodes(sb), GFP_KERNEL);
1109         }
1110         if (!err) {
1111                 err = percpu_counter_init(&sbi->s_dirs_counter,
1112                                 ext2_count_dirs(sb), GFP_KERNEL);
1113         }
1114         if (err) {
1115                 ext2_msg(sb, KERN_ERR, "error: insufficient memory");
1116                 goto failed_mount3;
1117         }
1118
1119 #ifdef CONFIG_EXT2_FS_XATTR
1120         sbi->s_mb_cache = ext2_xattr_create_cache();
1121         if (!sbi->s_mb_cache) {
1122                 ext2_msg(sb, KERN_ERR, "Failed to create an mb_cache");
1123                 goto failed_mount3;
1124         }
1125 #endif
1126         /*
1127          * set up enough so that it can read an inode
1128          */
1129         sb->s_op = &ext2_sops;
1130         sb->s_export_op = &ext2_export_ops;
1131         sb->s_xattr = ext2_xattr_handlers;
1132
1133 #ifdef CONFIG_QUOTA
1134         sb->dq_op = &dquot_operations;
1135         sb->s_qcop = &dquot_quotactl_ops;
1136         sb->s_quota_types = QTYPE_MASK_USR | QTYPE_MASK_GRP;
1137 #endif
1138
1139         root = ext2_iget(sb, EXT2_ROOT_INO);
1140         if (IS_ERR(root)) {
1141                 ret = PTR_ERR(root);
1142                 goto failed_mount3;
1143         }
1144         if (!S_ISDIR(root->i_mode) || !root->i_blocks || !root->i_size) {
1145                 iput(root);
1146                 ext2_msg(sb, KERN_ERR, "error: corrupt root inode, run e2fsck");
1147                 goto failed_mount3;
1148         }
1149
1150         sb->s_root = d_make_root(root);
1151         if (!sb->s_root) {
1152                 ext2_msg(sb, KERN_ERR, "error: get root inode failed");
1153                 ret = -ENOMEM;
1154                 goto failed_mount3;
1155         }
1156         if (EXT2_HAS_COMPAT_FEATURE(sb, EXT3_FEATURE_COMPAT_HAS_JOURNAL))
1157                 ext2_msg(sb, KERN_WARNING,
1158                         "warning: mounting ext3 filesystem as ext2");
1159         if (ext2_setup_super (sb, es, sb->s_flags & MS_RDONLY))
1160                 sb->s_flags |= MS_RDONLY;
1161         ext2_write_super(sb);
1162         return 0;
1163
1164 cantfind_ext2:
1165         if (!silent)
1166                 ext2_msg(sb, KERN_ERR,
1167                         "error: can't find an ext2 filesystem on dev %s.",
1168                         sb->s_id);
1169         goto failed_mount;
1170 failed_mount3:
1171         if (sbi->s_mb_cache)
1172                 ext2_xattr_destroy_cache(sbi->s_mb_cache);
1173         percpu_counter_destroy(&sbi->s_freeblocks_counter);
1174         percpu_counter_destroy(&sbi->s_freeinodes_counter);
1175         percpu_counter_destroy(&sbi->s_dirs_counter);
1176 failed_mount2:
1177         for (i = 0; i < db_count; i++)
1178                 brelse(sbi->s_group_desc[i]);
1179 failed_mount_group_desc:
1180         kfree(sbi->s_group_desc);
1181         kfree(sbi->s_debts);
1182 failed_mount:
1183         brelse(bh);
1184 failed_sbi:
1185         sb->s_fs_info = NULL;
1186         kfree(sbi->s_blockgroup_lock);
1187         kfree(sbi);
1188 failed:
1189         return ret;
1190 }
1191
1192 static void ext2_clear_super_error(struct super_block *sb)
1193 {
1194         struct buffer_head *sbh = EXT2_SB(sb)->s_sbh;
1195
1196         if (buffer_write_io_error(sbh)) {
1197                 /*
1198                  * Oh, dear.  A previous attempt to write the
1199                  * superblock failed.  This could happen because the
1200                  * USB device was yanked out.  Or it could happen to
1201                  * be a transient write error and maybe the block will
1202                  * be remapped.  Nothing we can do but to retry the
1203                  * write and hope for the best.
1204                  */
1205                 ext2_msg(sb, KERN_ERR,
1206                        "previous I/O error to superblock detected\n");
1207                 clear_buffer_write_io_error(sbh);
1208                 set_buffer_uptodate(sbh);
1209         }
1210 }
1211
1212 static void ext2_sync_super(struct super_block *sb, struct ext2_super_block *es,
1213                             int wait)
1214 {
1215         ext2_clear_super_error(sb);
1216         spin_lock(&EXT2_SB(sb)->s_lock);
1217         es->s_free_blocks_count = cpu_to_le32(ext2_count_free_blocks(sb));
1218         es->s_free_inodes_count = cpu_to_le32(ext2_count_free_inodes(sb));
1219         es->s_wtime = cpu_to_le32(get_seconds());
1220         /* unlock before we do IO */
1221         spin_unlock(&EXT2_SB(sb)->s_lock);
1222         mark_buffer_dirty(EXT2_SB(sb)->s_sbh);
1223         if (wait)
1224                 sync_dirty_buffer(EXT2_SB(sb)->s_sbh);
1225 }
1226
1227 /*
1228  * In the second extended file system, it is not necessary to
1229  * write the super block since we use a mapping of the
1230  * disk super block in a buffer.
1231  *
1232  * However, this function is still used to set the fs valid
1233  * flags to 0.  We need to set this flag to 0 since the fs
1234  * may have been checked while mounted and e2fsck may have
1235  * set s_state to EXT2_VALID_FS after some corrections.
1236  */
1237 static int ext2_sync_fs(struct super_block *sb, int wait)
1238 {
1239         struct ext2_sb_info *sbi = EXT2_SB(sb);
1240         struct ext2_super_block *es = EXT2_SB(sb)->s_es;
1241
1242         /*
1243          * Write quota structures to quota file, sync_blockdev() will write
1244          * them to disk later
1245          */
1246         dquot_writeback_dquots(sb, -1);
1247
1248         spin_lock(&sbi->s_lock);
1249         if (es->s_state & cpu_to_le16(EXT2_VALID_FS)) {
1250                 ext2_debug("setting valid to 0\n");
1251                 es->s_state &= cpu_to_le16(~EXT2_VALID_FS);
1252         }
1253         spin_unlock(&sbi->s_lock);
1254         ext2_sync_super(sb, es, wait);
1255         return 0;
1256 }
1257
1258 static int ext2_freeze(struct super_block *sb)
1259 {
1260         struct ext2_sb_info *sbi = EXT2_SB(sb);
1261
1262         /*
1263          * Open but unlinked files present? Keep EXT2_VALID_FS flag cleared
1264          * because we have unattached inodes and thus filesystem is not fully
1265          * consistent.
1266          */
1267         if (atomic_long_read(&sb->s_remove_count)) {
1268                 ext2_sync_fs(sb, 1);
1269                 return 0;
1270         }
1271         /* Set EXT2_FS_VALID flag */
1272         spin_lock(&sbi->s_lock);
1273         sbi->s_es->s_state = cpu_to_le16(sbi->s_mount_state);
1274         spin_unlock(&sbi->s_lock);
1275         ext2_sync_super(sb, sbi->s_es, 1);
1276
1277         return 0;
1278 }
1279
1280 static int ext2_unfreeze(struct super_block *sb)
1281 {
1282         /* Just write sb to clear EXT2_VALID_FS flag */
1283         ext2_write_super(sb);
1284
1285         return 0;
1286 }
1287
1288 void ext2_write_super(struct super_block *sb)
1289 {
1290         if (!(sb->s_flags & MS_RDONLY))
1291                 ext2_sync_fs(sb, 1);
1292 }
1293
1294 static int ext2_remount (struct super_block * sb, int * flags, char * data)
1295 {
1296         struct ext2_sb_info * sbi = EXT2_SB(sb);
1297         struct ext2_super_block * es;
1298         struct ext2_mount_options old_opts;
1299         unsigned long old_sb_flags;
1300         int err;
1301
1302         sync_filesystem(sb);
1303         spin_lock(&sbi->s_lock);
1304
1305         /* Store the old options */
1306         old_sb_flags = sb->s_flags;
1307         old_opts.s_mount_opt = sbi->s_mount_opt;
1308         old_opts.s_resuid = sbi->s_resuid;
1309         old_opts.s_resgid = sbi->s_resgid;
1310
1311         /*
1312          * Allow the "check" option to be passed as a remount option.
1313          */
1314         if (!parse_options(data, sb)) {
1315                 err = -EINVAL;
1316                 goto restore_opts;
1317         }
1318
1319         sb->s_flags = (sb->s_flags & ~MS_POSIXACL) |
1320                 ((sbi->s_mount_opt & EXT2_MOUNT_POSIX_ACL) ? MS_POSIXACL : 0);
1321
1322         es = sbi->s_es;
1323         if ((sbi->s_mount_opt ^ old_opts.s_mount_opt) & EXT2_MOUNT_DAX) {
1324                 ext2_msg(sb, KERN_WARNING, "warning: refusing change of "
1325                          "dax flag with busy inodes while remounting");
1326                 sbi->s_mount_opt ^= EXT2_MOUNT_DAX;
1327         }
1328         if ((*flags & MS_RDONLY) == (sb->s_flags & MS_RDONLY)) {
1329                 spin_unlock(&sbi->s_lock);
1330                 return 0;
1331         }
1332         if (*flags & MS_RDONLY) {
1333                 if (le16_to_cpu(es->s_state) & EXT2_VALID_FS ||
1334                     !(sbi->s_mount_state & EXT2_VALID_FS)) {
1335                         spin_unlock(&sbi->s_lock);
1336                         return 0;
1337                 }
1338
1339                 /*
1340                  * OK, we are remounting a valid rw partition rdonly, so set
1341                  * the rdonly flag and then mark the partition as valid again.
1342                  */
1343                 es->s_state = cpu_to_le16(sbi->s_mount_state);
1344                 es->s_mtime = cpu_to_le32(get_seconds());
1345                 spin_unlock(&sbi->s_lock);
1346
1347                 err = dquot_suspend(sb, -1);
1348                 if (err < 0) {
1349                         spin_lock(&sbi->s_lock);
1350                         goto restore_opts;
1351                 }
1352
1353                 ext2_sync_super(sb, es, 1);
1354         } else {
1355                 __le32 ret = EXT2_HAS_RO_COMPAT_FEATURE(sb,
1356                                                ~EXT2_FEATURE_RO_COMPAT_SUPP);
1357                 if (ret) {
1358                         ext2_msg(sb, KERN_WARNING,
1359                                 "warning: couldn't remount RDWR because of "
1360                                 "unsupported optional features (%x).",
1361                                 le32_to_cpu(ret));
1362                         err = -EROFS;
1363                         goto restore_opts;
1364                 }
1365                 /*
1366                  * Mounting a RDONLY partition read-write, so reread and
1367                  * store the current valid flag.  (It may have been changed
1368                  * by e2fsck since we originally mounted the partition.)
1369                  */
1370                 sbi->s_mount_state = le16_to_cpu(es->s_state);
1371                 if (!ext2_setup_super (sb, es, 0))
1372                         sb->s_flags &= ~MS_RDONLY;
1373                 spin_unlock(&sbi->s_lock);
1374
1375                 ext2_write_super(sb);
1376
1377                 dquot_resume(sb, -1);
1378         }
1379
1380         return 0;
1381 restore_opts:
1382         sbi->s_mount_opt = old_opts.s_mount_opt;
1383         sbi->s_resuid = old_opts.s_resuid;
1384         sbi->s_resgid = old_opts.s_resgid;
1385         sb->s_flags = old_sb_flags;
1386         spin_unlock(&sbi->s_lock);
1387         return err;
1388 }
1389
1390 static int ext2_statfs (struct dentry * dentry, struct kstatfs * buf)
1391 {
1392         struct super_block *sb = dentry->d_sb;
1393         struct ext2_sb_info *sbi = EXT2_SB(sb);
1394         struct ext2_super_block *es = sbi->s_es;
1395         u64 fsid;
1396
1397         spin_lock(&sbi->s_lock);
1398
1399         if (test_opt (sb, MINIX_DF))
1400                 sbi->s_overhead_last = 0;
1401         else if (sbi->s_blocks_last != le32_to_cpu(es->s_blocks_count)) {
1402                 unsigned long i, overhead = 0;
1403                 smp_rmb();
1404
1405                 /*
1406                  * Compute the overhead (FS structures). This is constant
1407                  * for a given filesystem unless the number of block groups
1408                  * changes so we cache the previous value until it does.
1409                  */
1410
1411                 /*
1412                  * All of the blocks before first_data_block are
1413                  * overhead
1414                  */
1415                 overhead = le32_to_cpu(es->s_first_data_block);
1416
1417                 /*
1418                  * Add the overhead attributed to the superblock and
1419                  * block group descriptors.  If the sparse superblocks
1420                  * feature is turned on, then not all groups have this.
1421                  */
1422                 for (i = 0; i < sbi->s_groups_count; i++)
1423                         overhead += ext2_bg_has_super(sb, i) +
1424                                 ext2_bg_num_gdb(sb, i);
1425
1426                 /*
1427                  * Every block group has an inode bitmap, a block
1428                  * bitmap, and an inode table.
1429                  */
1430                 overhead += (sbi->s_groups_count *
1431                              (2 + sbi->s_itb_per_group));
1432                 sbi->s_overhead_last = overhead;
1433                 smp_wmb();
1434                 sbi->s_blocks_last = le32_to_cpu(es->s_blocks_count);
1435         }
1436
1437         buf->f_type = EXT2_SUPER_MAGIC;
1438         buf->f_bsize = sb->s_blocksize;
1439         buf->f_blocks = le32_to_cpu(es->s_blocks_count) - sbi->s_overhead_last;
1440         buf->f_bfree = ext2_count_free_blocks(sb);
1441         es->s_free_blocks_count = cpu_to_le32(buf->f_bfree);
1442         buf->f_bavail = buf->f_bfree - le32_to_cpu(es->s_r_blocks_count);
1443         if (buf->f_bfree < le32_to_cpu(es->s_r_blocks_count))
1444                 buf->f_bavail = 0;
1445         buf->f_files = le32_to_cpu(es->s_inodes_count);
1446         buf->f_ffree = ext2_count_free_inodes(sb);
1447         es->s_free_inodes_count = cpu_to_le32(buf->f_ffree);
1448         buf->f_namelen = EXT2_NAME_LEN;
1449         fsid = le64_to_cpup((void *)es->s_uuid) ^
1450                le64_to_cpup((void *)es->s_uuid + sizeof(u64));
1451         buf->f_fsid.val[0] = fsid & 0xFFFFFFFFUL;
1452         buf->f_fsid.val[1] = (fsid >> 32) & 0xFFFFFFFFUL;
1453         spin_unlock(&sbi->s_lock);
1454         return 0;
1455 }
1456
1457 static struct dentry *ext2_mount(struct file_system_type *fs_type,
1458         int flags, const char *dev_name, void *data)
1459 {
1460         return mount_bdev(fs_type, flags, dev_name, data, ext2_fill_super);
1461 }
1462
1463 #ifdef CONFIG_QUOTA
1464
1465 /* Read data from quotafile - avoid pagecache and such because we cannot afford
1466  * acquiring the locks... As quota files are never truncated and quota code
1467  * itself serializes the operations (and no one else should touch the files)
1468  * we don't have to be afraid of races */
1469 static ssize_t ext2_quota_read(struct super_block *sb, int type, char *data,
1470                                size_t len, loff_t off)
1471 {
1472         struct inode *inode = sb_dqopt(sb)->files[type];
1473         sector_t blk = off >> EXT2_BLOCK_SIZE_BITS(sb);
1474         int err = 0;
1475         int offset = off & (sb->s_blocksize - 1);
1476         int tocopy;
1477         size_t toread;
1478         struct buffer_head tmp_bh;
1479         struct buffer_head *bh;
1480         loff_t i_size = i_size_read(inode);
1481
1482         if (off > i_size)
1483                 return 0;
1484         if (off+len > i_size)
1485                 len = i_size-off;
1486         toread = len;
1487         while (toread > 0) {
1488                 tocopy = sb->s_blocksize - offset < toread ?
1489                                 sb->s_blocksize - offset : toread;
1490
1491                 tmp_bh.b_state = 0;
1492                 tmp_bh.b_size = sb->s_blocksize;
1493                 err = ext2_get_block(inode, blk, &tmp_bh, 0);
1494                 if (err < 0)
1495                         return err;
1496                 if (!buffer_mapped(&tmp_bh))    /* A hole? */
1497                         memset(data, 0, tocopy);
1498                 else {
1499                         bh = sb_bread(sb, tmp_bh.b_blocknr);
1500                         if (!bh)
1501                                 return -EIO;
1502                         memcpy(data, bh->b_data+offset, tocopy);
1503                         brelse(bh);
1504                 }
1505                 offset = 0;
1506                 toread -= tocopy;
1507                 data += tocopy;
1508                 blk++;
1509         }
1510         return len;
1511 }
1512
1513 /* Write to quotafile */
1514 static ssize_t ext2_quota_write(struct super_block *sb, int type,
1515                                 const char *data, size_t len, loff_t off)
1516 {
1517         struct inode *inode = sb_dqopt(sb)->files[type];
1518         sector_t blk = off >> EXT2_BLOCK_SIZE_BITS(sb);
1519         int err = 0;
1520         int offset = off & (sb->s_blocksize - 1);
1521         int tocopy;
1522         size_t towrite = len;
1523         struct buffer_head tmp_bh;
1524         struct buffer_head *bh;
1525
1526         while (towrite > 0) {
1527                 tocopy = sb->s_blocksize - offset < towrite ?
1528                                 sb->s_blocksize - offset : towrite;
1529
1530                 tmp_bh.b_state = 0;
1531                 tmp_bh.b_size = sb->s_blocksize;
1532                 err = ext2_get_block(inode, blk, &tmp_bh, 1);
1533                 if (err < 0)
1534                         goto out;
1535                 if (offset || tocopy != EXT2_BLOCK_SIZE(sb))
1536                         bh = sb_bread(sb, tmp_bh.b_blocknr);
1537                 else
1538                         bh = sb_getblk(sb, tmp_bh.b_blocknr);
1539                 if (unlikely(!bh)) {
1540                         err = -EIO;
1541                         goto out;
1542                 }
1543                 lock_buffer(bh);
1544                 memcpy(bh->b_data+offset, data, tocopy);
1545                 flush_dcache_page(bh->b_page);
1546                 set_buffer_uptodate(bh);
1547                 mark_buffer_dirty(bh);
1548                 unlock_buffer(bh);
1549                 brelse(bh);
1550                 offset = 0;
1551                 towrite -= tocopy;
1552                 data += tocopy;
1553                 blk++;
1554         }
1555 out:
1556         if (len == towrite)
1557                 return err;
1558         if (inode->i_size < off+len-towrite)
1559                 i_size_write(inode, off+len-towrite);
1560         inode->i_version++;
1561         inode->i_mtime = inode->i_ctime = current_time(inode);
1562         mark_inode_dirty(inode);
1563         return len - towrite;
1564 }
1565
1566 #endif
1567
1568 static struct file_system_type ext2_fs_type = {
1569         .owner          = THIS_MODULE,
1570         .name           = "ext2",
1571         .mount          = ext2_mount,
1572         .kill_sb        = kill_block_super,
1573         .fs_flags       = FS_REQUIRES_DEV,
1574 };
1575 MODULE_ALIAS_FS("ext2");
1576
1577 static int __init init_ext2_fs(void)
1578 {
1579         int err;
1580
1581         err = init_inodecache();
1582         if (err)
1583                 return err;
1584         err = register_filesystem(&ext2_fs_type);
1585         if (err)
1586                 goto out;
1587         return 0;
1588 out:
1589         destroy_inodecache();
1590         return err;
1591 }
1592
1593 static void __exit exit_ext2_fs(void)
1594 {
1595         unregister_filesystem(&ext2_fs_type);
1596         destroy_inodecache();
1597 }
1598
1599 MODULE_AUTHOR("Remy Card and others");
1600 MODULE_DESCRIPTION("Second Extended Filesystem");
1601 MODULE_LICENSE("GPL");
1602 module_init(init_ext2_fs)
1603 module_exit(exit_ext2_fs)