fs/ext2/super.c

   1 /*
   2  *  linux/fs/ext2/super.c
   3  *
   4  * Copyright (C) 1992, 1993, 1994, 1995
   5  * Remy Card (card@masi.ibp.fr)
   6  * Laboratoire MASI - Institut Blaise Pascal
   7  * Universite Pierre et Marie Curie (Paris VI)
   8  *
   9  *  from
  10  *
  11  *  linux/fs/minix/inode.c
  12  *
  13  *  Copyright (C) 1991, 1992  Linus Torvalds
  14  *
  15  *  Big-endian to little-endian byte-swapping/bitmaps by
  16  *        David S. Miller (davem@caip.rutgers.edu), 1995
  17  */
  18
  19 #include <linux/module.h>
  20 #include <linux/string.h>
  21 #include <linux/fs.h>
  22 #include <linux/slab.h>
  23 #include <linux/init.h>
  24 #include <linux/blkdev.h>
  25 #include <linux/parser.h>
  26 #include <linux/random.h>
  27 #include <linux/buffer_head.h>
  28 #include <linux/exportfs.h>
  29 #include <linux/vfs.h>
  30 #include <linux/seq_file.h>
  31 #include <linux/mount.h>
  32 #include <linux/log2.h>
  33 #include <linux/quotaops.h>
  34 #include <asm/uaccess.h>
  35 #include "ext2.h"
  36 #include "xattr.h"
  37 #include "acl.h"
  38
  39 static void ext2_sync_super(struct super_block *sb,
  40                             struct ext2_super_block *es, int wait);
  41 static int ext2_remount (struct super_block * sb, int * flags, char * data);
  42 static int ext2_statfs (struct dentry * dentry, struct kstatfs * buf);
  43 static int ext2_sync_fs(struct super_block *sb, int wait);
  44 static int ext2_freeze(struct super_block *sb);
  45 static int ext2_unfreeze(struct super_block *sb);
  46
  47 void ext2_error(struct super_block *sb, const char *function,
  48                 const char *fmt, ...)
  49 {
  50         struct va_format vaf;
  51         va_list args;
  52         struct ext2_sb_info *sbi = EXT2_SB(sb);
  53         struct ext2_super_block *es = sbi->s_es;
  54
  55         if (!(sb->s_flags & MS_RDONLY)) {
  56                 spin_lock(&sbi->s_lock);
  57                 sbi->s_mount_state |= EXT2_ERROR_FS;
  58                 es->s_state |= cpu_to_le16(EXT2_ERROR_FS);
  59                 spin_unlock(&sbi->s_lock);
  60                 ext2_sync_super(sb, es, 1);
  61         }
  62
  63         va_start(args, fmt);
  64
  65         vaf.fmt = fmt;
  66         vaf.va = &args;
  67
  68         printk(KERN_CRIT "EXT2-fs (%s): error: %s: %pV\n",
  69                sb->s_id, function, &vaf);
  70
  71         va_end(args);
  72
  73         if (test_opt(sb, ERRORS_PANIC))
  74                 panic("EXT2-fs: panic from previous error\n");
  75         if (test_opt(sb, ERRORS_RO)) {
  76                 ext2_msg(sb, KERN_CRIT,
  77                              "error: remounting filesystem read-only");
  78                 sb->s_flags |= MS_RDONLY;
  79         }
  80 }
  81
  82 void ext2_msg(struct super_block *sb, const char *prefix,
  83                 const char *fmt, ...)
  84 {
  85         struct va_format vaf;
  86         va_list args;
  87
  88         va_start(args, fmt);
  89
  90         vaf.fmt = fmt;
  91         vaf.va = &args;
  92
  93         printk("%sEXT2-fs (%s): %pV\n", prefix, sb->s_id, &vaf);
  94
  95         va_end(args);
  96 }
  97
  98 /*
  99  * This must be called with sbi->s_lock held.
 100  */
 101 void ext2_update_dynamic_rev(struct super_block *sb)
 102 {
 103         struct ext2_super_block *es = EXT2_SB(sb)->s_es;
 104
 105         if (le32_to_cpu(es->s_rev_level) > EXT2_GOOD_OLD_REV)
 106                 return;
 107
 108         ext2_msg(sb, KERN_WARNING,
 109                      "warning: updating to rev %d because of "
 110                      "new feature flag, running e2fsck is recommended",
 111                      EXT2_DYNAMIC_REV);
 112
 113         es->s_first_ino = cpu_to_le32(EXT2_GOOD_OLD_FIRST_INO);
 114         es->s_inode_size = cpu_to_le16(EXT2_GOOD_OLD_INODE_SIZE);
 115         es->s_rev_level = cpu_to_le32(EXT2_DYNAMIC_REV);
 116         /* leave es->s_feature_*compat flags alone */
 117         /* es->s_uuid will be set by e2fsck if empty */
 118
 119         /*
 120          * The rest of the superblock fields should be zero, and if not it
 121          * means they are likely already in use, so leave them alone.  We
 122          * can leave it up to e2fsck to clean up any inconsistencies there.
 123          */
 124 }
 125
 126 static void ext2_put_super (struct super_block * sb)
 127 {
 128         int db_count;
 129         int i;
 130         struct ext2_sb_info *sbi = EXT2_SB(sb);
 131
 132         dquot_disable(sb, -1, DQUOT_USAGE_ENABLED | DQUOT_LIMITS_ENABLED);
 133
 134         ext2_xattr_put_super(sb);
 135         if (!(sb->s_flags & MS_RDONLY)) {
 136                 struct ext2_super_block *es = sbi->s_es;
 137
 138                 spin_lock(&sbi->s_lock);
 139                 es->s_state = cpu_to_le16(sbi->s_mount_state);
 140                 spin_unlock(&sbi->s_lock);
 141                 ext2_sync_super(sb, es, 1);
 142         }
 143         db_count = sbi->s_gdb_count;
 144         for (i = 0; i < db_count; i++)
 145                 if (sbi->s_group_desc[i])
 146                         brelse (sbi->s_group_desc[i]);
 147         kfree(sbi->s_group_desc);
 148         kfree(sbi->s_debts);
 149         percpu_counter_destroy(&sbi->s_freeblocks_counter);
 150         percpu_counter_destroy(&sbi->s_freeinodes_counter);
 151         percpu_counter_destroy(&sbi->s_dirs_counter);
 152         brelse (sbi->s_sbh);
 153         sb->s_fs_info = NULL;
 154         kfree(sbi->s_blockgroup_lock);
 155         kfree(sbi);
 156 }
 157
 158 static struct kmem_cache * ext2_inode_cachep;
 159
 160 static struct inode *ext2_alloc_inode(struct super_block *sb)
 161 {
 162         struct ext2_inode_info *ei;
 163         ei = kmem_cache_alloc(ext2_inode_cachep, GFP_KERNEL);
 164         if (!ei)
 165                 return NULL;
 166         ei->i_block_alloc_info = NULL;
 167         ei->vfs_inode.i_version = 1;
 168 #ifdef CONFIG_QUOTA
 169         memset(&ei->i_dquot, 0, sizeof(ei->i_dquot));
 170 #endif
 171
 172         return &ei->vfs_inode;
 173 }
 174
 175 static void ext2_i_callback(struct rcu_head *head)
 176 {
 177         struct inode *inode = container_of(head, struct inode, i_rcu);
 178         kmem_cache_free(ext2_inode_cachep, EXT2_I(inode));
 179 }
 180
 181 static void ext2_destroy_inode(struct inode *inode)
 182 {
 183         call_rcu(&inode->i_rcu, ext2_i_callback);
 184 }
 185
 186 static void init_once(void *foo)
 187 {
 188         struct ext2_inode_info *ei = (struct ext2_inode_info *) foo;
 189
 190         rwlock_init(&ei->i_meta_lock);
 191 #ifdef CONFIG_EXT2_FS_XATTR
 192         init_rwsem(&ei->xattr_sem);
 193 #endif
 194         mutex_init(&ei->truncate_mutex);
 195 #ifdef CONFIG_FS_DAX
 196         init_rwsem(&ei->dax_sem);
 197 #endif
 198         inode_init_once(&ei->vfs_inode);
 199 }
 200
 201 static int __init init_inodecache(void)
 202 {
 203         ext2_inode_cachep = kmem_cache_create("ext2_inode_cache",
 204                                              sizeof(struct ext2_inode_info),
 205                                              0, (SLAB_RECLAIM_ACCOUNT|
 206                                                 SLAB_MEM_SPREAD),
 207                                              init_once);
 208         if (ext2_inode_cachep == NULL)
 209                 return -ENOMEM;
 210         return 0;
 211 }
 212
 213 static void destroy_inodecache(void)
 214 {
 215         /*
 216          * Make sure all delayed rcu free inodes are flushed before we
 217          * destroy cache.
 218          */
 219         rcu_barrier();
 220         kmem_cache_destroy(ext2_inode_cachep);
 221 }
 222
 223 static int ext2_show_options(struct seq_file *seq, struct dentry *root)
 224 {
 225         struct super_block *sb = root->d_sb;
 226         struct ext2_sb_info *sbi = EXT2_SB(sb);
 227         struct ext2_super_block *es = sbi->s_es;
 228         unsigned long def_mount_opts;
 229
 230         spin_lock(&sbi->s_lock);
 231         def_mount_opts = le32_to_cpu(es->s_default_mount_opts);
 232
 233         if (sbi->s_sb_block != 1)
 234                 seq_printf(seq, ",sb=%lu", sbi->s_sb_block);
 235         if (test_opt(sb, MINIX_DF))
 236                 seq_puts(seq, ",minixdf");
 237         if (test_opt(sb, GRPID))
 238                 seq_puts(seq, ",grpid");
 239         if (!test_opt(sb, GRPID) && (def_mount_opts & EXT2_DEFM_BSDGROUPS))
 240                 seq_puts(seq, ",nogrpid");
 241         if (!uid_eq(sbi->s_resuid, make_kuid(&init_user_ns, EXT2_DEF_RESUID)) ||
 242             le16_to_cpu(es->s_def_resuid) != EXT2_DEF_RESUID) {
 243                 seq_printf(seq, ",resuid=%u",
 244                                 from_kuid_munged(&init_user_ns, sbi->s_resuid));
 245         }
 246         if (!gid_eq(sbi->s_resgid, make_kgid(&init_user_ns, EXT2_DEF_RESGID)) ||
 247             le16_to_cpu(es->s_def_resgid) != EXT2_DEF_RESGID) {
 248                 seq_printf(seq, ",resgid=%u",
 249                                 from_kgid_munged(&init_user_ns, sbi->s_resgid));
 250         }
 251         if (test_opt(sb, ERRORS_RO)) {
 252                 int def_errors = le16_to_cpu(es->s_errors);
 253
 254                 if (def_errors == EXT2_ERRORS_PANIC ||
 255                     def_errors == EXT2_ERRORS_CONTINUE) {
 256                         seq_puts(seq, ",errors=remount-ro");
 257                 }
 258         }
 259         if (test_opt(sb, ERRORS_CONT))
 260                 seq_puts(seq, ",errors=continue");
 261         if (test_opt(sb, ERRORS_PANIC))
 262                 seq_puts(seq, ",errors=panic");
 263         if (test_opt(sb, NO_UID32))
 264                 seq_puts(seq, ",nouid32");
 265         if (test_opt(sb, DEBUG))
 266                 seq_puts(seq, ",debug");
 267         if (test_opt(sb, OLDALLOC))
 268                 seq_puts(seq, ",oldalloc");
 269
 270 #ifdef CONFIG_EXT2_FS_XATTR
 271         if (test_opt(sb, XATTR_USER))
 272                 seq_puts(seq, ",user_xattr");
 273         if (!test_opt(sb, XATTR_USER) &&
 274             (def_mount_opts & EXT2_DEFM_XATTR_USER)) {
 275                 seq_puts(seq, ",nouser_xattr");
 276         }
 277 #endif
 278
 279 #ifdef CONFIG_EXT2_FS_POSIX_ACL
 280         if (test_opt(sb, POSIX_ACL))
 281                 seq_puts(seq, ",acl");
 282         if (!test_opt(sb, POSIX_ACL) && (def_mount_opts & EXT2_DEFM_ACL))
 283                 seq_puts(seq, ",noacl");
 284 #endif
 285
 286         if (test_opt(sb, NOBH))
 287                 seq_puts(seq, ",nobh");
 288
 289 #if defined(CONFIG_QUOTA)
 290         if (sbi->s_mount_opt & EXT2_MOUNT_USRQUOTA)
 291                 seq_puts(seq, ",usrquota");
 292
 293         if (sbi->s_mount_opt & EXT2_MOUNT_GRPQUOTA)
 294                 seq_puts(seq, ",grpquota");
 295 #endif
 296
 297 #ifdef CONFIG_FS_DAX
 298         if (sbi->s_mount_opt & EXT2_MOUNT_XIP)
 299                 seq_puts(seq, ",xip");
 300         if (sbi->s_mount_opt & EXT2_MOUNT_DAX)
 301                 seq_puts(seq, ",dax");
 302 #endif
 303
 304         if (!test_opt(sb, RESERVATION))
 305                 seq_puts(seq, ",noreservation");
 306
 307         spin_unlock(&sbi->s_lock);
 308         return 0;
 309 }
 310
 311 #ifdef CONFIG_QUOTA
 312 static ssize_t ext2_quota_read(struct super_block *sb, int type, char *data, size_t len, loff_t off);
 313 static ssize_t ext2_quota_write(struct super_block *sb, int type, const char *data, size_t len, loff_t off);
 314 static struct dquot **ext2_get_dquots(struct inode *inode)
 315 {
 316         return EXT2_I(inode)->i_dquot;
 317 }
 318 #endif
 319
 320 static const struct super_operations ext2_sops = {
 321         .alloc_inode    = ext2_alloc_inode,
 322         .destroy_inode  = ext2_destroy_inode,
 323         .write_inode    = ext2_write_inode,
 324         .evict_inode    = ext2_evict_inode,
 325         .put_super      = ext2_put_super,
 326         .sync_fs        = ext2_sync_fs,
 327         .freeze_fs      = ext2_freeze,
 328         .unfreeze_fs    = ext2_unfreeze,
 329         .statfs         = ext2_statfs,
 330         .remount_fs     = ext2_remount,
 331         .show_options   = ext2_show_options,
 332 #ifdef CONFIG_QUOTA
 333         .quota_read     = ext2_quota_read,
 334         .quota_write    = ext2_quota_write,
 335         .get_dquots     = ext2_get_dquots,
 336 #endif
 337 };
 338
 339 static struct inode *ext2_nfs_get_inode(struct super_block *sb,
 340                 u64 ino, u32 generation)
 341 {
 342         struct inode *inode;
 343
 344         if (ino < EXT2_FIRST_INO(sb) && ino != EXT2_ROOT_INO)
 345                 return ERR_PTR(-ESTALE);
 346         if (ino > le32_to_cpu(EXT2_SB(sb)->s_es->s_inodes_count))
 347                 return ERR_PTR(-ESTALE);
 348
 349         /*
 350          * ext2_iget isn't quite right if the inode is currently unallocated!
 351          * However ext2_iget currently does appropriate checks to handle stale
 352          * inodes so everything is OK.
 353          */
 354         inode = ext2_iget(sb, ino);
 355         if (IS_ERR(inode))
 356                 return ERR_CAST(inode);
 357         if (generation && inode->i_generation != generation) {
 358                 /* we didn't find the right inode.. */
 359                 iput(inode);
 360                 return ERR_PTR(-ESTALE);
 361         }
 362         return inode;
 363 }
 364
 365 static struct dentry *ext2_fh_to_dentry(struct super_block *sb, struct fid *fid,
 366                 int fh_len, int fh_type)
 367 {
 368         return generic_fh_to_dentry(sb, fid, fh_len, fh_type,
 369                                     ext2_nfs_get_inode);
 370 }
 371
 372 static struct dentry *ext2_fh_to_parent(struct super_block *sb, struct fid *fid,
 373                 int fh_len, int fh_type)
 374 {
 375         return generic_fh_to_parent(sb, fid, fh_len, fh_type,
 376                                     ext2_nfs_get_inode);
 377 }
 378
 379 static const struct export_operations ext2_export_ops = {
 380         .fh_to_dentry = ext2_fh_to_dentry,
 381         .fh_to_parent = ext2_fh_to_parent,
 382         .get_parent = ext2_get_parent,
 383 };
 384
 385 static unsigned long get_sb_block(void **data)
 386 {
 387         unsigned long   sb_block;
 388         char            *options = (char *) *data;
 389
 390         if (!options || strncmp(options, "sb=", 3) != 0)
 391                 return 1;       /* Default location */
 392         options += 3;
 393         sb_block = simple_strtoul(options, &options, 0);
 394         if (*options && *options != ',') {
 395                 printk("EXT2-fs: Invalid sb specification: %s\n",
 396                        (char *) *data);
 397                 return 1;
 398         }
 399         if (*options == ',')
 400                 options++;
 401         *data = (void *) options;
 402         return sb_block;
 403 }
 404
 405 enum {
 406         Opt_bsd_df, Opt_minix_df, Opt_grpid, Opt_nogrpid,
 407         Opt_resgid, Opt_resuid, Opt_sb, Opt_err_cont, Opt_err_panic,
 408         Opt_err_ro, Opt_nouid32, Opt_nocheck, Opt_debug,
 409         Opt_oldalloc, Opt_orlov, Opt_nobh, Opt_user_xattr, Opt_nouser_xattr,
 410         Opt_acl, Opt_noacl, Opt_xip, Opt_dax, Opt_ignore, Opt_err, Opt_quota,
 411         Opt_usrquota, Opt_grpquota, Opt_reservation, Opt_noreservation
 412 };
 413
 414 static const match_table_t tokens = {
 415         {Opt_bsd_df, "bsddf"},
 416         {Opt_minix_df, "minixdf"},
 417         {Opt_grpid, "grpid"},
 418         {Opt_grpid, "bsdgroups"},
 419         {Opt_nogrpid, "nogrpid"},
 420         {Opt_nogrpid, "sysvgroups"},
 421         {Opt_resgid, "resgid=%u"},
 422         {Opt_resuid, "resuid=%u"},
 423         {Opt_sb, "sb=%u"},
 424         {Opt_err_cont, "errors=continue"},
 425         {Opt_err_panic, "errors=panic"},
 426         {Opt_err_ro, "errors=remount-ro"},
 427         {Opt_nouid32, "nouid32"},
 428         {Opt_nocheck, "check=none"},
 429         {Opt_nocheck, "nocheck"},
 430         {Opt_debug, "debug"},
 431         {Opt_oldalloc, "oldalloc"},
 432         {Opt_orlov, "orlov"},
 433         {Opt_nobh, "nobh"},
 434         {Opt_user_xattr, "user_xattr"},
 435         {Opt_nouser_xattr, "nouser_xattr"},
 436         {Opt_acl, "acl"},
 437         {Opt_noacl, "noacl"},
 438         {Opt_xip, "xip"},
 439         {Opt_dax, "dax"},
 440         {Opt_grpquota, "grpquota"},
 441         {Opt_ignore, "noquota"},
 442         {Opt_quota, "quota"},
 443         {Opt_usrquota, "usrquota"},
 444         {Opt_reservation, "reservation"},
 445         {Opt_noreservation, "noreservation"},
 446         {Opt_err, NULL}
 447 };
 448
 449 static int parse_options(char *options, struct super_block *sb)
 450 {
 451         char *p;
 452         struct ext2_sb_info *sbi = EXT2_SB(sb);
 453         substring_t args[MAX_OPT_ARGS];
 454         int option;
 455         kuid_t uid;
 456         kgid_t gid;
 457
 458         if (!options)
 459                 return 1;
 460
 461         while ((p = strsep (&options, ",")) != NULL) {
 462                 int token;
 463                 if (!*p)
 464                         continue;
 465
 466                 token = match_token(p, tokens, args);
 467                 switch (token) {
 468                 case Opt_bsd_df:
 469                         clear_opt (sbi->s_mount_opt, MINIX_DF);
 470                         break;
 471                 case Opt_minix_df:
 472                         set_opt (sbi->s_mount_opt, MINIX_DF);
 473                         break;
 474                 case Opt_grpid:
 475                         set_opt (sbi->s_mount_opt, GRPID);
 476                         break;
 477                 case Opt_nogrpid:
 478                         clear_opt (sbi->s_mount_opt, GRPID);
 479                         break;
 480                 case Opt_resuid:
 481                         if (match_int(&args[0], &option))
 482                                 return 0;
 483                         uid = make_kuid(current_user_ns(), option);
 484                         if (!uid_valid(uid)) {
 485                                 ext2_msg(sb, KERN_ERR, "Invalid uid value %d", option);
 486                                 return 0;
 487
 488                         }
 489                         sbi->s_resuid = uid;
 490                         break;
 491                 case Opt_resgid:
 492                         if (match_int(&args[0], &option))
 493                                 return 0;
 494                         gid = make_kgid(current_user_ns(), option);
 495                         if (!gid_valid(gid)) {
 496                                 ext2_msg(sb, KERN_ERR, "Invalid gid value %d", option);
 497                                 return 0;
 498                         }
 499                         sbi->s_resgid = gid;
 500                         break;
 501                 case Opt_sb:
 502                         /* handled by get_sb_block() instead of here */
 503                         /* *sb_block = match_int(&args[0]); */
 504                         break;
 505                 case Opt_err_panic:
 506                         clear_opt (sbi->s_mount_opt, ERRORS_CONT);
 507                         clear_opt (sbi->s_mount_opt, ERRORS_RO);
 508                         set_opt (sbi->s_mount_opt, ERRORS_PANIC);
 509                         break;
 510                 case Opt_err_ro:
 511                         clear_opt (sbi->s_mount_opt, ERRORS_CONT);
 512                         clear_opt (sbi->s_mount_opt, ERRORS_PANIC);
 513                         set_opt (sbi->s_mount_opt, ERRORS_RO);
 514                         break;
 515                 case Opt_err_cont:
 516                         clear_opt (sbi->s_mount_opt, ERRORS_RO);
 517                         clear_opt (sbi->s_mount_opt, ERRORS_PANIC);
 518                         set_opt (sbi->s_mount_opt, ERRORS_CONT);
 519                         break;
 520                 case Opt_nouid32:
 521                         set_opt (sbi->s_mount_opt, NO_UID32);
 522                         break;
 523                 case Opt_nocheck:
 524                         clear_opt (sbi->s_mount_opt, CHECK);
 525                         break;
 526                 case Opt_debug:
 527                         set_opt (sbi->s_mount_opt, DEBUG);
 528                         break;
 529                 case Opt_oldalloc:
 530                         set_opt (sbi->s_mount_opt, OLDALLOC);
 531                         break;
 532                 case Opt_orlov:
 533                         clear_opt (sbi->s_mount_opt, OLDALLOC);
 534                         break;
 535                 case Opt_nobh:
 536                         set_opt (sbi->s_mount_opt, NOBH);
 537                         break;
 538 #ifdef CONFIG_EXT2_FS_XATTR
 539                 case Opt_user_xattr:
 540                         set_opt (sbi->s_mount_opt, XATTR_USER);
 541                         break;
 542                 case Opt_nouser_xattr:
 543                         clear_opt (sbi->s_mount_opt, XATTR_USER);
 544                         break;
 545 #else
 546                 case Opt_user_xattr:
 547                 case Opt_nouser_xattr:
 548                         ext2_msg(sb, KERN_INFO, "(no)user_xattr options"
 549                                 "not supported");
 550                         break;
 551 #endif
 552 #ifdef CONFIG_EXT2_FS_POSIX_ACL
 553                 case Opt_acl:
 554                         set_opt(sbi->s_mount_opt, POSIX_ACL);
 555                         break;
 556                 case Opt_noacl:
 557                         clear_opt(sbi->s_mount_opt, POSIX_ACL);
 558                         break;
 559 #else
 560                 case Opt_acl:
 561                 case Opt_noacl:
 562                         ext2_msg(sb, KERN_INFO,
 563                                 "(no)acl options not supported");
 564                         break;
 565 #endif
 566                 case Opt_xip:
 567                         ext2_msg(sb, KERN_INFO, "use dax instead of xip");
 568                         set_opt(sbi->s_mount_opt, XIP);
 569                         /* Fall through */
 570                 case Opt_dax:
 571 #ifdef CONFIG_FS_DAX
 572                         ext2_msg(sb, KERN_WARNING,
 573                 "DAX enabled. Warning: EXPERIMENTAL, use at your own risk");
 574                         set_opt(sbi->s_mount_opt, DAX);
 575 #else
 576                         ext2_msg(sb, KERN_INFO, "dax option not supported");
 577 #endif
 578                         break;
 579
 580 #if defined(CONFIG_QUOTA)
 581                 case Opt_quota:
 582                 case Opt_usrquota:
 583                         set_opt(sbi->s_mount_opt, USRQUOTA);
 584                         break;
 585
 586                 case Opt_grpquota:
 587                         set_opt(sbi->s_mount_opt, GRPQUOTA);
 588                         break;
 589 #else
 590                 case Opt_quota:
 591                 case Opt_usrquota:
 592                 case Opt_grpquota:
 593                         ext2_msg(sb, KERN_INFO,
 594                                 "quota operations not supported");
 595                         break;
 596 #endif
 597
 598                 case Opt_reservation:
 599                         set_opt(sbi->s_mount_opt, RESERVATION);
 600                         ext2_msg(sb, KERN_INFO, "reservations ON");
 601                         break;
 602                 case Opt_noreservation:
 603                         clear_opt(sbi->s_mount_opt, RESERVATION);
 604                         ext2_msg(sb, KERN_INFO, "reservations OFF");
 605                         break;
 606                 case Opt_ignore:
 607                         break;
 608                 default:
 609                         return 0;
 610                 }
 611         }
 612         return 1;
 613 }
 614
 615 static int ext2_setup_super (struct super_block * sb,
 616                               struct ext2_super_block * es,
 617                               int read_only)
 618 {
 619         int res = 0;
 620         struct ext2_sb_info *sbi = EXT2_SB(sb);
 621
 622         if (le32_to_cpu(es->s_rev_level) > EXT2_MAX_SUPP_REV) {
 623                 ext2_msg(sb, KERN_ERR,
 624                         "error: revision level too high, "
 625                         "forcing read-only mode");
 626                 res = MS_RDONLY;
 627         }
 628         if (read_only)
 629                 return res;
 630         if (!(sbi->s_mount_state & EXT2_VALID_FS))
 631                 ext2_msg(sb, KERN_WARNING,
 632                         "warning: mounting unchecked fs, "
 633                         "running e2fsck is recommended");
 634         else if ((sbi->s_mount_state & EXT2_ERROR_FS))
 635                 ext2_msg(sb, KERN_WARNING,
 636                         "warning: mounting fs with errors, "
 637                         "running e2fsck is recommended");
 638         else if ((__s16) le16_to_cpu(es->s_max_mnt_count) >= 0 &&
 639                  le16_to_cpu(es->s_mnt_count) >=
 640                  (unsigned short) (__s16) le16_to_cpu(es->s_max_mnt_count))
 641                 ext2_msg(sb, KERN_WARNING,
 642                         "warning: maximal mount count reached, "
 643                         "running e2fsck is recommended");
 644         else if (le32_to_cpu(es->s_checkinterval) &&
 645                 (le32_to_cpu(es->s_lastcheck) +
 646                         le32_to_cpu(es->s_checkinterval) <= get_seconds()))
 647                 ext2_msg(sb, KERN_WARNING,
 648                         "warning: checktime reached, "
 649                         "running e2fsck is recommended");
 650         if (!le16_to_cpu(es->s_max_mnt_count))
 651                 es->s_max_mnt_count = cpu_to_le16(EXT2_DFL_MAX_MNT_COUNT);
 652         le16_add_cpu(&es->s_mnt_count, 1);
 653         if (test_opt (sb, DEBUG))
 654                 ext2_msg(sb, KERN_INFO, "%s, %s, bs=%lu, fs=%lu, gc=%lu, "
 655                         "bpg=%lu, ipg=%lu, mo=%04lx]",
 656                         EXT2FS_VERSION, EXT2FS_DATE, sb->s_blocksize,
 657                         sbi->s_frag_size,
 658                         sbi->s_groups_count,
 659                         EXT2_BLOCKS_PER_GROUP(sb),
 660                         EXT2_INODES_PER_GROUP(sb),
 661                         sbi->s_mount_opt);
 662         return res;
 663 }
 664
 665 static int ext2_check_descriptors(struct super_block *sb)
 666 {
 667         int i;
 668         struct ext2_sb_info *sbi = EXT2_SB(sb);
 669
 670         ext2_debug ("Checking group descriptors");
 671
 672         for (i = 0; i < sbi->s_groups_count; i++) {
 673                 struct ext2_group_desc *gdp = ext2_get_group_desc(sb, i, NULL);
 674                 ext2_fsblk_t first_block = ext2_group_first_block_no(sb, i);
 675                 ext2_fsblk_t last_block;
 676
 677                 if (i == sbi->s_groups_count - 1)
 678                         last_block = le32_to_cpu(sbi->s_es->s_blocks_count) - 1;
 679                 else
 680                         last_block = first_block +
 681                                 (EXT2_BLOCKS_PER_GROUP(sb) - 1);
 682
 683                 if (le32_to_cpu(gdp->bg_block_bitmap) < first_block ||
 684                     le32_to_cpu(gdp->bg_block_bitmap) > last_block)
 685                 {
 686                         ext2_error (sb, "ext2_check_descriptors",
 687                                     "Block bitmap for group %d"
 688                                     " not in group (block %lu)!",
 689                                     i, (unsigned long) le32_to_cpu(gdp->bg_block_bitmap));
 690                         return 0;
 691                 }
 692                 if (le32_to_cpu(gdp->bg_inode_bitmap) < first_block ||
 693                     le32_to_cpu(gdp->bg_inode_bitmap) > last_block)
 694                 {
 695                         ext2_error (sb, "ext2_check_descriptors",
 696                                     "Inode bitmap for group %d"
 697                                     " not in group (block %lu)!",
 698                                     i, (unsigned long) le32_to_cpu(gdp->bg_inode_bitmap));
 699                         return 0;
 700                 }
 701                 if (le32_to_cpu(gdp->bg_inode_table) < first_block ||
 702                     le32_to_cpu(gdp->bg_inode_table) + sbi->s_itb_per_group - 1 >
 703                     last_block)
 704                 {
 705                         ext2_error (sb, "ext2_check_descriptors",
 706                                     "Inode table for group %d"
 707                                     " not in group (block %lu)!",
 708                                     i, (unsigned long) le32_to_cpu(gdp->bg_inode_table));
 709                         return 0;
 710                 }
 711         }
 712         return 1;
 713 }
 714
 715 /*
 716  * Maximal file size.  There is a direct, and {,double-,triple-}indirect
 717  * block limit, and also a limit of (2^32 - 1) 512-byte sectors in i_blocks.
 718  * We need to be 1 filesystem block less than the 2^32 sector limit.
 719  */
 720 static loff_t ext2_max_size(int bits)
 721 {
 722         loff_t res = EXT2_NDIR_BLOCKS;
 723         int meta_blocks;
 724         unsigned int upper_limit;
 725         unsigned int ppb = 1 << (bits-2);
 726
 727         /* This is calculated to be the largest file size for a
 728          * dense, file such that the total number of
 729          * sectors in the file, including data and all indirect blocks,
 730          * does not exceed 2^32 -1
 731          * __u32 i_blocks representing the total number of
 732          * 512 bytes blocks of the file
 733          */
 734         upper_limit = (1LL << 32) - 1;
 735
 736         /* total blocks in file system block size */
 737         upper_limit >>= (bits - 9);
 738
 739         /* Compute how many blocks we can address by block tree */
 740         res += 1LL << (bits-2);
 741         res += 1LL << (2*(bits-2));
 742         res += 1LL << (3*(bits-2));
 743         /* Does block tree limit file size? */
 744         if (res < upper_limit)
 745                 goto check_lfs;
 746
 747         res = upper_limit;
 748         /* How many metadata blocks are needed for addressing upper_limit? */
 749         upper_limit -= EXT2_NDIR_BLOCKS;
 750         /* indirect blocks */
 751         meta_blocks = 1;
 752         upper_limit -= ppb;
 753         /* double indirect blocks */
 754         if (upper_limit < ppb * ppb) {
 755                 meta_blocks += 1 + DIV_ROUND_UP(upper_limit, ppb);
 756                 res -= meta_blocks;
 757                 goto check_lfs;
 758         }
 759         meta_blocks += 1 + ppb;
 760         upper_limit -= ppb * ppb;
 761         /* tripple indirect blocks for the rest */
 762         meta_blocks += 1 + DIV_ROUND_UP(upper_limit, ppb) +
 763                 DIV_ROUND_UP(upper_limit, ppb*ppb);
 764         res -= meta_blocks;
 765 check_lfs:
 766         res <<= bits;
 767         if (res > MAX_LFS_FILESIZE)
 768                 res = MAX_LFS_FILESIZE;
 769
 770         return res;
 771 }
 772
 773 static unsigned long descriptor_loc(struct super_block *sb,
 774                                     unsigned long logic_sb_block,
 775                                     int nr)
 776 {
 777         struct ext2_sb_info *sbi = EXT2_SB(sb);
 778         unsigned long bg, first_meta_bg;
 779         int has_super = 0;
 780
 781         first_meta_bg = le32_to_cpu(sbi->s_es->s_first_meta_bg);
 782
 783         if (!EXT2_HAS_INCOMPAT_FEATURE(sb, EXT2_FEATURE_INCOMPAT_META_BG) ||
 784             nr < first_meta_bg)
 785                 return (logic_sb_block + nr + 1);
 786         bg = sbi->s_desc_per_block * nr;
 787         if (ext2_bg_has_super(sb, bg))
 788                 has_super = 1;
 789
 790         return ext2_group_first_block_no(sb, bg) + has_super;
 791 }
 792
 793 static int ext2_fill_super(struct super_block *sb, void *data, int silent)
 794 {
 795         struct buffer_head * bh;
 796         struct ext2_sb_info * sbi;
 797         struct ext2_super_block * es;
 798         struct inode *root;
 799         unsigned long block;
 800         unsigned long sb_block = get_sb_block(&data);
 801         unsigned long logic_sb_block;
 802         unsigned long offset = 0;
 803         unsigned long def_mount_opts;
 804         long ret = -EINVAL;
 805         int blocksize = BLOCK_SIZE;
 806         int db_count;
 807         int i, j;
 808         __le32 features;
 809         int err;
 810
 811         err = -ENOMEM;
 812         sbi = kzalloc(sizeof(*sbi), GFP_KERNEL);
 813         if (!sbi)
 814                 goto failed;
 815
 816         sbi->s_blockgroup_lock =
 817                 kzalloc(sizeof(struct blockgroup_lock), GFP_KERNEL);
 818         if (!sbi->s_blockgroup_lock) {
 819                 kfree(sbi);
 820                 goto failed;
 821         }
 822         sb->s_fs_info = sbi;
 823         sbi->s_sb_block = sb_block;
 824
 825         spin_lock_init(&sbi->s_lock);
 826
 827         /*
 828          * See what the current blocksize for the device is, and
 829          * use that as the blocksize.  Otherwise (or if the blocksize
 830          * is smaller than the default) use the default.
 831          * This is important for devices that have a hardware
 832          * sectorsize that is larger than the default.
 833          */
 834         blocksize = sb_min_blocksize(sb, BLOCK_SIZE);
 835         if (!blocksize) {
 836                 ext2_msg(sb, KERN_ERR, "error: unable to set blocksize");
 837                 goto failed_sbi;
 838         }
 839
 840         /*
 841          * If the superblock doesn't start on a hardware sector boundary,
 842          * calculate the offset.
 843          */
 844         if (blocksize != BLOCK_SIZE) {
 845                 logic_sb_block = (sb_block*BLOCK_SIZE) / blocksize;
 846                 offset = (sb_block*BLOCK_SIZE) % blocksize;
 847         } else {
 848                 logic_sb_block = sb_block;
 849         }
 850
 851         if (!(bh = sb_bread(sb, logic_sb_block))) {
 852                 ext2_msg(sb, KERN_ERR, "error: unable to read superblock");
 853                 goto failed_sbi;
 854         }
 855         /*
 856          * Note: s_es must be initialized as soon as possible because
 857          *       some ext2 macro-instructions depend on its value
 858          */
 859         es = (struct ext2_super_block *) (((char *)bh->b_data) + offset);
 860         sbi->s_es = es;
 861         sb->s_magic = le16_to_cpu(es->s_magic);
 862
 863         if (sb->s_magic != EXT2_SUPER_MAGIC)
 864                 goto cantfind_ext2;
 865
 866         /* Set defaults before we parse the mount options */
 867         def_mount_opts = le32_to_cpu(es->s_default_mount_opts);
 868         if (def_mount_opts & EXT2_DEFM_DEBUG)
 869                 set_opt(sbi->s_mount_opt, DEBUG);
 870         if (def_mount_opts & EXT2_DEFM_BSDGROUPS)
 871                 set_opt(sbi->s_mount_opt, GRPID);
 872         if (def_mount_opts & EXT2_DEFM_UID16)
 873                 set_opt(sbi->s_mount_opt, NO_UID32);
 874 #ifdef CONFIG_EXT2_FS_XATTR
 875         if (def_mount_opts & EXT2_DEFM_XATTR_USER)
 876                 set_opt(sbi->s_mount_opt, XATTR_USER);
 877 #endif
 878 #ifdef CONFIG_EXT2_FS_POSIX_ACL
 879         if (def_mount_opts & EXT2_DEFM_ACL)
 880                 set_opt(sbi->s_mount_opt, POSIX_ACL);
 881 #endif
 882
 883         if (le16_to_cpu(sbi->s_es->s_errors) == EXT2_ERRORS_PANIC)
 884                 set_opt(sbi->s_mount_opt, ERRORS_PANIC);
 885         else if (le16_to_cpu(sbi->s_es->s_errors) == EXT2_ERRORS_CONTINUE)
 886                 set_opt(sbi->s_mount_opt, ERRORS_CONT);
 887         else
 888                 set_opt(sbi->s_mount_opt, ERRORS_RO);
 889
 890         sbi->s_resuid = make_kuid(&init_user_ns, le16_to_cpu(es->s_def_resuid));
 891         sbi->s_resgid = make_kgid(&init_user_ns, le16_to_cpu(es->s_def_resgid));
 892
 893         set_opt(sbi->s_mount_opt, RESERVATION);
 894
 895         if (!parse_options((char *) data, sb))
 896                 goto failed_mount;
 897
 898         sb->s_flags = (sb->s_flags & ~MS_POSIXACL) |
 899                 ((EXT2_SB(sb)->s_mount_opt & EXT2_MOUNT_POSIX_ACL) ?
 900                  MS_POSIXACL : 0);
 901         sb->s_iflags |= SB_I_CGROUPWB;
 902
 903         if (le32_to_cpu(es->s_rev_level) == EXT2_GOOD_OLD_REV &&
 904             (EXT2_HAS_COMPAT_FEATURE(sb, ~0U) ||
 905              EXT2_HAS_RO_COMPAT_FEATURE(sb, ~0U) ||
 906              EXT2_HAS_INCOMPAT_FEATURE(sb, ~0U)))
 907                 ext2_msg(sb, KERN_WARNING,
 908                         "warning: feature flags set on rev 0 fs, "
 909                         "running e2fsck is recommended");
 910         /*
 911          * Check feature flags regardless of the revision level, since we
 912          * previously didn't change the revision level when setting the flags,
 913          * so there is a chance incompat flags are set on a rev 0 filesystem.
 914          */
 915         features = EXT2_HAS_INCOMPAT_FEATURE(sb, ~EXT2_FEATURE_INCOMPAT_SUPP);
 916         if (features) {
 917                 ext2_msg(sb, KERN_ERR,  "error: couldn't mount because of "
 918                        "unsupported optional features (%x)",
 919                         le32_to_cpu(features));
 920                 goto failed_mount;
 921         }
 922         if (!(sb->s_flags & MS_RDONLY) &&
 923             (features = EXT2_HAS_RO_COMPAT_FEATURE(sb, ~EXT2_FEATURE_RO_COMPAT_SUPP))){
 924                 ext2_msg(sb, KERN_ERR, "error: couldn't mount RDWR because of "
 925                        "unsupported optional features (%x)",
 926                        le32_to_cpu(features));
 927                 goto failed_mount;
 928         }
 929
 930         blocksize = BLOCK_SIZE << le32_to_cpu(sbi->s_es->s_log_block_size);
 931
 932         if (sbi->s_mount_opt & EXT2_MOUNT_DAX) {
 933                 if (blocksize != PAGE_SIZE) {
 934                         ext2_msg(sb, KERN_ERR,
 935                                         "error: unsupported blocksize for dax");
 936                         goto failed_mount;
 937                 }
 938                 if (!sb->s_bdev->bd_disk->fops->direct_access) {
 939                         ext2_msg(sb, KERN_ERR,
 940                                         "error: device does not support dax");
 941                         goto failed_mount;
 942                 }
 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          * set up enough so that it can read an inode
1120          */
1121         sb->s_op = &ext2_sops;
1122         sb->s_export_op = &ext2_export_ops;
1123         sb->s_xattr = ext2_xattr_handlers;
1124
1125 #ifdef CONFIG_QUOTA
1126         sb->dq_op = &dquot_operations;
1127         sb->s_qcop = &dquot_quotactl_ops;
1128         sb->s_quota_types = QTYPE_MASK_USR | QTYPE_MASK_GRP;
1129 #endif
1130
1131         root = ext2_iget(sb, EXT2_ROOT_INO);
1132         if (IS_ERR(root)) {
1133                 ret = PTR_ERR(root);
1134                 goto failed_mount3;
1135         }
1136         if (!S_ISDIR(root->i_mode) || !root->i_blocks || !root->i_size) {
1137                 iput(root);
1138                 ext2_msg(sb, KERN_ERR, "error: corrupt root inode, run e2fsck");
1139                 goto failed_mount3;
1140         }
1141
1142         sb->s_root = d_make_root(root);
1143         if (!sb->s_root) {
1144                 ext2_msg(sb, KERN_ERR, "error: get root inode failed");
1145                 ret = -ENOMEM;
1146                 goto failed_mount3;
1147         }
1148         if (EXT2_HAS_COMPAT_FEATURE(sb, EXT3_FEATURE_COMPAT_HAS_JOURNAL))
1149                 ext2_msg(sb, KERN_WARNING,
1150                         "warning: mounting ext3 filesystem as ext2");
1151         if (ext2_setup_super (sb, es, sb->s_flags & MS_RDONLY))
1152                 sb->s_flags |= MS_RDONLY;
1153         ext2_write_super(sb);
1154         return 0;
1155
1156 cantfind_ext2:
1157         if (!silent)
1158                 ext2_msg(sb, KERN_ERR,
1159                         "error: can't find an ext2 filesystem on dev %s.",
1160                         sb->s_id);
1161         goto failed_mount;
1162 failed_mount3:
1163         percpu_counter_destroy(&sbi->s_freeblocks_counter);
1164         percpu_counter_destroy(&sbi->s_freeinodes_counter);
1165         percpu_counter_destroy(&sbi->s_dirs_counter);
1166 failed_mount2:
1167         for (i = 0; i < db_count; i++)
1168                 brelse(sbi->s_group_desc[i]);
1169 failed_mount_group_desc:
1170         kfree(sbi->s_group_desc);
1171         kfree(sbi->s_debts);
1172 failed_mount:
1173         brelse(bh);
1174 failed_sbi:
1175         sb->s_fs_info = NULL;
1176         kfree(sbi->s_blockgroup_lock);
1177         kfree(sbi);
1178 failed:
1179         return ret;
1180 }
1181
1182 static void ext2_clear_super_error(struct super_block *sb)
1183 {
1184         struct buffer_head *sbh = EXT2_SB(sb)->s_sbh;
1185
1186         if (buffer_write_io_error(sbh)) {
1187                 /*
1188                  * Oh, dear.  A previous attempt to write the
1189                  * superblock failed.  This could happen because the
1190                  * USB device was yanked out.  Or it could happen to
1191                  * be a transient write error and maybe the block will
1192                  * be remapped.  Nothing we can do but to retry the
1193                  * write and hope for the best.
1194                  */
1195                 ext2_msg(sb, KERN_ERR,
1196                        "previous I/O error to superblock detected\n");
1197                 clear_buffer_write_io_error(sbh);
1198                 set_buffer_uptodate(sbh);
1199         }
1200 }
1201
1202 static void ext2_sync_super(struct super_block *sb, struct ext2_super_block *es,
1203                             int wait)
1204 {
1205         ext2_clear_super_error(sb);
1206         spin_lock(&EXT2_SB(sb)->s_lock);
1207         es->s_free_blocks_count = cpu_to_le32(ext2_count_free_blocks(sb));
1208         es->s_free_inodes_count = cpu_to_le32(ext2_count_free_inodes(sb));
1209         es->s_wtime = cpu_to_le32(get_seconds());
1210         /* unlock before we do IO */
1211         spin_unlock(&EXT2_SB(sb)->s_lock);
1212         mark_buffer_dirty(EXT2_SB(sb)->s_sbh);
1213         if (wait)
1214                 sync_dirty_buffer(EXT2_SB(sb)->s_sbh);
1215 }
1216
1217 /*
1218  * In the second extended file system, it is not necessary to
1219  * write the super block since we use a mapping of the
1220  * disk super block in a buffer.
1221  *
1222  * However, this function is still used to set the fs valid
1223  * flags to 0.  We need to set this flag to 0 since the fs
1224  * may have been checked while mounted and e2fsck may have
1225  * set s_state to EXT2_VALID_FS after some corrections.
1226  */
1227 static int ext2_sync_fs(struct super_block *sb, int wait)
1228 {
1229         struct ext2_sb_info *sbi = EXT2_SB(sb);
1230         struct ext2_super_block *es = EXT2_SB(sb)->s_es;
1231
1232         /*
1233          * Write quota structures to quota file, sync_blockdev() will write
1234          * them to disk later
1235          */
1236         dquot_writeback_dquots(sb, -1);
1237
1238         spin_lock(&sbi->s_lock);
1239         if (es->s_state & cpu_to_le16(EXT2_VALID_FS)) {
1240                 ext2_debug("setting valid to 0\n");
1241                 es->s_state &= cpu_to_le16(~EXT2_VALID_FS);
1242         }
1243         spin_unlock(&sbi->s_lock);
1244         ext2_sync_super(sb, es, wait);
1245         return 0;
1246 }
1247
1248 static int ext2_freeze(struct super_block *sb)
1249 {
1250         struct ext2_sb_info *sbi = EXT2_SB(sb);
1251
1252         /*
1253          * Open but unlinked files present? Keep EXT2_VALID_FS flag cleared
1254          * because we have unattached inodes and thus filesystem is not fully
1255          * consistent.
1256          */
1257         if (atomic_long_read(&sb->s_remove_count)) {
1258                 ext2_sync_fs(sb, 1);
1259                 return 0;
1260         }
1261         /* Set EXT2_FS_VALID flag */
1262         spin_lock(&sbi->s_lock);
1263         sbi->s_es->s_state = cpu_to_le16(sbi->s_mount_state);
1264         spin_unlock(&sbi->s_lock);
1265         ext2_sync_super(sb, sbi->s_es, 1);
1266
1267         return 0;
1268 }
1269
1270 static int ext2_unfreeze(struct super_block *sb)
1271 {
1272         /* Just write sb to clear EXT2_VALID_FS flag */
1273         ext2_write_super(sb);
1274
1275         return 0;
1276 }
1277
1278 void ext2_write_super(struct super_block *sb)
1279 {
1280         if (!(sb->s_flags & MS_RDONLY))
1281                 ext2_sync_fs(sb, 1);
1282 }
1283
1284 static int ext2_remount (struct super_block * sb, int * flags, char * data)
1285 {
1286         struct ext2_sb_info * sbi = EXT2_SB(sb);
1287         struct ext2_super_block * es;
1288         struct ext2_mount_options old_opts;
1289         unsigned long old_sb_flags;
1290         int err;
1291
1292         sync_filesystem(sb);
1293         spin_lock(&sbi->s_lock);
1294
1295         /* Store the old options */
1296         old_sb_flags = sb->s_flags;
1297         old_opts.s_mount_opt = sbi->s_mount_opt;
1298         old_opts.s_resuid = sbi->s_resuid;
1299         old_opts.s_resgid = sbi->s_resgid;
1300
1301         /*
1302          * Allow the "check" option to be passed as a remount option.
1303          */
1304         if (!parse_options(data, sb)) {
1305                 err = -EINVAL;
1306                 goto restore_opts;
1307         }
1308
1309         sb->s_flags = (sb->s_flags & ~MS_POSIXACL) |
1310                 ((sbi->s_mount_opt & EXT2_MOUNT_POSIX_ACL) ? MS_POSIXACL : 0);
1311
1312         es = sbi->s_es;
1313         if ((sbi->s_mount_opt ^ old_opts.s_mount_opt) & EXT2_MOUNT_DAX) {
1314                 ext2_msg(sb, KERN_WARNING, "warning: refusing change of "
1315                          "dax flag with busy inodes while remounting");
1316                 sbi->s_mount_opt ^= EXT2_MOUNT_DAX;
1317         }
1318         if ((*flags & MS_RDONLY) == (sb->s_flags & MS_RDONLY)) {
1319                 spin_unlock(&sbi->s_lock);
1320                 return 0;
1321         }
1322         if (*flags & MS_RDONLY) {
1323                 if (le16_to_cpu(es->s_state) & EXT2_VALID_FS ||
1324                     !(sbi->s_mount_state & EXT2_VALID_FS)) {
1325                         spin_unlock(&sbi->s_lock);
1326                         return 0;
1327                 }
1328
1329                 /*
1330                  * OK, we are remounting a valid rw partition rdonly, so set
1331                  * the rdonly flag and then mark the partition as valid again.
1332                  */
1333                 es->s_state = cpu_to_le16(sbi->s_mount_state);
1334                 es->s_mtime = cpu_to_le32(get_seconds());
1335                 spin_unlock(&sbi->s_lock);
1336
1337                 err = dquot_suspend(sb, -1);
1338                 if (err < 0) {
1339                         spin_lock(&sbi->s_lock);
1340                         goto restore_opts;
1341                 }
1342
1343                 ext2_sync_super(sb, es, 1);
1344         } else {
1345                 __le32 ret = EXT2_HAS_RO_COMPAT_FEATURE(sb,
1346                                                ~EXT2_FEATURE_RO_COMPAT_SUPP);
1347                 if (ret) {
1348                         ext2_msg(sb, KERN_WARNING,
1349                                 "warning: couldn't remount RDWR because of "
1350                                 "unsupported optional features (%x).",
1351                                 le32_to_cpu(ret));
1352                         err = -EROFS;
1353                         goto restore_opts;
1354                 }
1355                 /*
1356                  * Mounting a RDONLY partition read-write, so reread and
1357                  * store the current valid flag.  (It may have been changed
1358                  * by e2fsck since we originally mounted the partition.)
1359                  */
1360                 sbi->s_mount_state = le16_to_cpu(es->s_state);
1361                 if (!ext2_setup_super (sb, es, 0))
1362                         sb->s_flags &= ~MS_RDONLY;
1363                 spin_unlock(&sbi->s_lock);
1364
1365                 ext2_write_super(sb);
1366
1367                 dquot_resume(sb, -1);
1368         }
1369
1370         return 0;
1371 restore_opts:
1372         sbi->s_mount_opt = old_opts.s_mount_opt;
1373         sbi->s_resuid = old_opts.s_resuid;
1374         sbi->s_resgid = old_opts.s_resgid;
1375         sb->s_flags = old_sb_flags;
1376         spin_unlock(&sbi->s_lock);
1377         return err;
1378 }
1379
1380 static int ext2_statfs (struct dentry * dentry, struct kstatfs * buf)
1381 {
1382         struct super_block *sb = dentry->d_sb;
1383         struct ext2_sb_info *sbi = EXT2_SB(sb);
1384         struct ext2_super_block *es = sbi->s_es;
1385         u64 fsid;
1386
1387         spin_lock(&sbi->s_lock);
1388
1389         if (test_opt (sb, MINIX_DF))
1390                 sbi->s_overhead_last = 0;
1391         else if (sbi->s_blocks_last != le32_to_cpu(es->s_blocks_count)) {
1392                 unsigned long i, overhead = 0;
1393                 smp_rmb();
1394
1395                 /*
1396                  * Compute the overhead (FS structures). This is constant
1397                  * for a given filesystem unless the number of block groups
1398                  * changes so we cache the previous value until it does.
1399                  */
1400
1401                 /*
1402                  * All of the blocks before first_data_block are
1403                  * overhead
1404                  */
1405                 overhead = le32_to_cpu(es->s_first_data_block);
1406
1407                 /*
1408                  * Add the overhead attributed to the superblock and
1409                  * block group descriptors.  If the sparse superblocks
1410                  * feature is turned on, then not all groups have this.
1411                  */
1412                 for (i = 0; i < sbi->s_groups_count; i++)
1413                         overhead += ext2_bg_has_super(sb, i) +
1414                                 ext2_bg_num_gdb(sb, i);
1415
1416                 /*
1417                  * Every block group has an inode bitmap, a block
1418                  * bitmap, and an inode table.
1419                  */
1420                 overhead += (sbi->s_groups_count *
1421                              (2 + sbi->s_itb_per_group));
1422                 sbi->s_overhead_last = overhead;
1423                 smp_wmb();
1424                 sbi->s_blocks_last = le32_to_cpu(es->s_blocks_count);
1425         }
1426
1427         buf->f_type = EXT2_SUPER_MAGIC;
1428         buf->f_bsize = sb->s_blocksize;
1429         buf->f_blocks = le32_to_cpu(es->s_blocks_count) - sbi->s_overhead_last;
1430         buf->f_bfree = ext2_count_free_blocks(sb);
1431         es->s_free_blocks_count = cpu_to_le32(buf->f_bfree);
1432         buf->f_bavail = buf->f_bfree - le32_to_cpu(es->s_r_blocks_count);
1433         if (buf->f_bfree < le32_to_cpu(es->s_r_blocks_count))
1434                 buf->f_bavail = 0;
1435         buf->f_files = le32_to_cpu(es->s_inodes_count);
1436         buf->f_ffree = ext2_count_free_inodes(sb);
1437         es->s_free_inodes_count = cpu_to_le32(buf->f_ffree);
1438         buf->f_namelen = EXT2_NAME_LEN;
1439         fsid = le64_to_cpup((void *)es->s_uuid) ^
1440                le64_to_cpup((void *)es->s_uuid + sizeof(u64));
1441         buf->f_fsid.val[0] = fsid & 0xFFFFFFFFUL;
1442         buf->f_fsid.val[1] = (fsid >> 32) & 0xFFFFFFFFUL;
1443         spin_unlock(&sbi->s_lock);
1444         return 0;
1445 }
1446
1447 static struct dentry *ext2_mount(struct file_system_type *fs_type,
1448         int flags, const char *dev_name, void *data)
1449 {
1450         return mount_bdev(fs_type, flags, dev_name, data, ext2_fill_super);
1451 }
1452
1453 #ifdef CONFIG_QUOTA
1454
1455 /* Read data from quotafile - avoid pagecache and such because we cannot afford
1456  * acquiring the locks... As quota files are never truncated and quota code
1457  * itself serializes the operations (and no one else should touch the files)
1458  * we don't have to be afraid of races */
1459 static ssize_t ext2_quota_read(struct super_block *sb, int type, char *data,
1460                                size_t len, loff_t off)
1461 {
1462         struct inode *inode = sb_dqopt(sb)->files[type];
1463         sector_t blk = off >> EXT2_BLOCK_SIZE_BITS(sb);
1464         int err = 0;
1465         int offset = off & (sb->s_blocksize - 1);
1466         int tocopy;
1467         size_t toread;
1468         struct buffer_head tmp_bh;
1469         struct buffer_head *bh;
1470         loff_t i_size = i_size_read(inode);
1471
1472         if (off > i_size)
1473                 return 0;
1474         if (off+len > i_size)
1475                 len = i_size-off;
1476         toread = len;
1477         while (toread > 0) {
1478                 tocopy = sb->s_blocksize - offset < toread ?
1479                                 sb->s_blocksize - offset : toread;
1480
1481                 tmp_bh.b_state = 0;
1482                 tmp_bh.b_size = sb->s_blocksize;
1483                 err = ext2_get_block(inode, blk, &tmp_bh, 0);
1484                 if (err < 0)
1485                         return err;
1486                 if (!buffer_mapped(&tmp_bh))    /* A hole? */
1487                         memset(data, 0, tocopy);
1488                 else {
1489                         bh = sb_bread(sb, tmp_bh.b_blocknr);
1490                         if (!bh)
1491                                 return -EIO;
1492                         memcpy(data, bh->b_data+offset, tocopy);
1493                         brelse(bh);
1494                 }
1495                 offset = 0;
1496                 toread -= tocopy;
1497                 data += tocopy;
1498                 blk++;
1499         }
1500         return len;
1501 }
1502
1503 /* Write to quotafile */
1504 static ssize_t ext2_quota_write(struct super_block *sb, int type,
1505                                 const char *data, size_t len, loff_t off)
1506 {
1507         struct inode *inode = sb_dqopt(sb)->files[type];
1508         sector_t blk = off >> EXT2_BLOCK_SIZE_BITS(sb);
1509         int err = 0;
1510         int offset = off & (sb->s_blocksize - 1);
1511         int tocopy;
1512         size_t towrite = len;
1513         struct buffer_head tmp_bh;
1514         struct buffer_head *bh;
1515
1516         while (towrite > 0) {
1517                 tocopy = sb->s_blocksize - offset < towrite ?
1518                                 sb->s_blocksize - offset : towrite;
1519
1520                 tmp_bh.b_state = 0;
1521                 tmp_bh.b_size = sb->s_blocksize;
1522                 err = ext2_get_block(inode, blk, &tmp_bh, 1);
1523                 if (err < 0)
1524                         goto out;
1525                 if (offset || tocopy != EXT2_BLOCK_SIZE(sb))
1526                         bh = sb_bread(sb, tmp_bh.b_blocknr);
1527                 else
1528                         bh = sb_getblk(sb, tmp_bh.b_blocknr);
1529                 if (unlikely(!bh)) {
1530                         err = -EIO;
1531                         goto out;
1532                 }
1533                 lock_buffer(bh);
1534                 memcpy(bh->b_data+offset, data, tocopy);
1535                 flush_dcache_page(bh->b_page);
1536                 set_buffer_uptodate(bh);
1537                 mark_buffer_dirty(bh);
1538                 unlock_buffer(bh);
1539                 brelse(bh);
1540                 offset = 0;
1541                 towrite -= tocopy;
1542                 data += tocopy;
1543                 blk++;
1544         }
1545 out:
1546         if (len == towrite)
1547                 return err;
1548         if (inode->i_size < off+len-towrite)
1549                 i_size_write(inode, off+len-towrite);
1550         inode->i_version++;
1551         inode->i_mtime = inode->i_ctime = CURRENT_TIME;
1552         mark_inode_dirty(inode);
1553         return len - towrite;
1554 }
1555
1556 #endif
1557
1558 static struct file_system_type ext2_fs_type = {
1559         .owner          = THIS_MODULE,
1560         .name           = "ext2",
1561         .mount          = ext2_mount,
1562         .kill_sb        = kill_block_super,
1563         .fs_flags       = FS_REQUIRES_DEV,
1564 };
1565 MODULE_ALIAS_FS("ext2");
1566
1567 static int __init init_ext2_fs(void)
1568 {
1569         int err = init_ext2_xattr();
1570         if (err)
1571                 return err;
1572         err = init_inodecache();
1573         if (err)
1574                 goto out1;
1575         err = register_filesystem(&ext2_fs_type);
1576         if (err)
1577                 goto out;
1578         return 0;
1579 out:
1580         destroy_inodecache();
1581 out1:
1582         exit_ext2_xattr();
1583         return err;
1584 }
1585
1586 static void __exit exit_ext2_fs(void)
1587 {
1588         unregister_filesystem(&ext2_fs_type);
1589         destroy_inodecache();
1590         exit_ext2_xattr();
1591 }
1592
1593 MODULE_AUTHOR("Remy Card and others");
1594 MODULE_DESCRIPTION("Second Extended Filesystem");
1595 MODULE_LICENSE("GPL");
1596 module_init(init_ext2_fs)
1597 module_exit(exit_ext2_fs)