SAFE public projects git trees. - safe/jmp/linux-2.6/blob - fs/ext2/super.c

   1 /*
   2  *  linux/fs/ext2/super.c
   3  *
   4  * Copyright (C) 1992, 1993, 1994, 1995
   5  * Remy Card (card@masi.ibp.fr)
   6  * Laboratoire MASI - Institut Blaise Pascal
   7  * Universite Pierre et Marie Curie (Paris VI)
   8  *
   9  *  from
  10  *
  11  *  linux/fs/minix/inode.c
  12  *
  13  *  Copyright (C) 1991, 1992  Linus Torvalds
  14  *
  15  *  Big-endian to little-endian byte-swapping/bitmaps by
  16  *        David S. Miller (davem@caip.rutgers.edu), 1995
  17  */
  18
  19 #include <linux/module.h>
  20 #include <linux/string.h>
  21 #include <linux/fs.h>
  22 #include <linux/slab.h>
  23 #include <linux/init.h>
  24 #include <linux/blkdev.h>
  25 #include <linux/parser.h>
  26 #include <linux/random.h>
  27 #include <linux/buffer_head.h>
  28 #include <linux/exportfs.h>
  29 #include <linux/vfs.h>
  30 #include <linux/seq_file.h>
  31 #include <linux/mount.h>
  32 #include <linux/log2.h>
  33 #include <linux/quotaops.h>
  34 #include <asm/uaccess.h>
  35 #include "ext2.h"
  36 #include "xattr.h"
  37 #include "acl.h"
  38 #include "xip.h"
  39
  40 static void ext2_sync_super(struct super_block *sb,
  41                             struct ext2_super_block *es, int wait);
  42 static int ext2_remount (struct super_block * sb, int * flags, char * data);
  43 static int ext2_statfs (struct dentry * dentry, struct kstatfs * buf);
  44 static int ext2_sync_fs(struct super_block *sb, int wait);
  45
  46 void ext2_error (struct super_block * sb, const char * function,
  47                  const char * fmt, ...)
  48 {
  49         va_list args;
  50         struct ext2_sb_info *sbi = EXT2_SB(sb);
  51         struct ext2_super_block *es = sbi->s_es;
  52
  53         if (!(sb->s_flags & MS_RDONLY)) {
  54                 spin_lock(&sbi->s_lock);
  55                 sbi->s_mount_state |= EXT2_ERROR_FS;
  56                 es->s_state |= cpu_to_le16(EXT2_ERROR_FS);
  57                 spin_unlock(&sbi->s_lock);
  58                 ext2_sync_super(sb, es, 1);
  59         }
  60
  61         va_start(args, fmt);
  62         printk(KERN_CRIT "EXT2-fs (%s): error: %s: ", sb->s_id, function);
  63         vprintk(fmt, args);
  64         printk("\n");
  65         va_end(args);
  66
  67         if (test_opt(sb, ERRORS_PANIC))
  68                 panic("EXT2-fs: panic from previous error\n");
  69         if (test_opt(sb, ERRORS_RO)) {
  70                 ext2_msg(sb, KERN_CRIT,
  71                              "error: remounting filesystem read-only");
  72                 sb->s_flags |= MS_RDONLY;
  73         }
  74 }
  75
  76 void ext2_msg(struct super_block *sb, const char *prefix,
  77                 const char *fmt, ...)
  78 {
  79         va_list args;
  80
  81         va_start(args, fmt);
  82         printk("%sEXT2-fs (%s): ", prefix, sb->s_id);
  83         vprintk(fmt, args);
  84         printk("\n");
  85         va_end(args);
  86 }
  87
  88 /*
  89  * This must be called with sbi->s_lock held.
  90  */
  91 void ext2_update_dynamic_rev(struct super_block *sb)
  92 {
  93         struct ext2_super_block *es = EXT2_SB(sb)->s_es;
  94
  95         if (le32_to_cpu(es->s_rev_level) > EXT2_GOOD_OLD_REV)
  96                 return;
  97
  98         ext2_msg(sb, KERN_WARNING,
  99                      "warning: updating to rev %d because of "
 100                      "new feature flag, running e2fsck is recommended",
 101                      EXT2_DYNAMIC_REV);
 102
 103         es->s_first_ino = cpu_to_le32(EXT2_GOOD_OLD_FIRST_INO);
 104         es->s_inode_size = cpu_to_le16(EXT2_GOOD_OLD_INODE_SIZE);
 105         es->s_rev_level = cpu_to_le32(EXT2_DYNAMIC_REV);
 106         /* leave es->s_feature_*compat flags alone */
 107         /* es->s_uuid will be set by e2fsck if empty */
 108
 109         /*
 110          * The rest of the superblock fields should be zero, and if not it
 111          * means they are likely already in use, so leave them alone.  We
 112          * can leave it up to e2fsck to clean up any inconsistencies there.
 113          */
 114 }
 115
 116 static void ext2_put_super (struct super_block * sb)
 117 {
 118         int db_count;
 119         int i;
 120         struct ext2_sb_info *sbi = EXT2_SB(sb);
 121
 122         if (sb->s_dirt)
 123                 ext2_write_super(sb);
 124
 125         ext2_xattr_put_super(sb);
 126         if (!(sb->s_flags & MS_RDONLY)) {
 127                 struct ext2_super_block *es = sbi->s_es;
 128
 129                 spin_lock(&sbi->s_lock);
 130                 es->s_state = cpu_to_le16(sbi->s_mount_state);
 131                 spin_unlock(&sbi->s_lock);
 132                 ext2_sync_super(sb, es, 1);
 133         }
 134         db_count = sbi->s_gdb_count;
 135         for (i = 0; i < db_count; i++)
 136                 if (sbi->s_group_desc[i])
 137                         brelse (sbi->s_group_desc[i]);
 138         kfree(sbi->s_group_desc);
 139         kfree(sbi->s_debts);
 140         percpu_counter_destroy(&sbi->s_freeblocks_counter);
 141         percpu_counter_destroy(&sbi->s_freeinodes_counter);
 142         percpu_counter_destroy(&sbi->s_dirs_counter);
 143         brelse (sbi->s_sbh);
 144         sb->s_fs_info = NULL;
 145         kfree(sbi->s_blockgroup_lock);
 146         kfree(sbi);
 147 }
 148
 149 static struct kmem_cache * ext2_inode_cachep;
 150
 151 static struct inode *ext2_alloc_inode(struct super_block *sb)
 152 {
 153         struct ext2_inode_info *ei;
 154         ei = (struct ext2_inode_info *)kmem_cache_alloc(ext2_inode_cachep, GFP_KERNEL);
 155         if (!ei)
 156                 return NULL;
 157         ei->i_block_alloc_info = NULL;
 158         ei->vfs_inode.i_version = 1;
 159         return &ei->vfs_inode;
 160 }
 161
 162 static void ext2_destroy_inode(struct inode *inode)
 163 {
 164         kmem_cache_free(ext2_inode_cachep, EXT2_I(inode));
 165 }
 166
 167 static void init_once(void *foo)
 168 {
 169         struct ext2_inode_info *ei = (struct ext2_inode_info *) foo;
 170
 171         rwlock_init(&ei->i_meta_lock);
 172 #ifdef CONFIG_EXT2_FS_XATTR
 173         init_rwsem(&ei->xattr_sem);
 174 #endif
 175         mutex_init(&ei->truncate_mutex);
 176         inode_init_once(&ei->vfs_inode);
 177 }
 178
 179 static int init_inodecache(void)
 180 {
 181         ext2_inode_cachep = kmem_cache_create("ext2_inode_cache",
 182                                              sizeof(struct ext2_inode_info),
 183                                              0, (SLAB_RECLAIM_ACCOUNT|
 184                                                 SLAB_MEM_SPREAD),
 185                                              init_once);
 186         if (ext2_inode_cachep == NULL)
 187                 return -ENOMEM;
 188         return 0;
 189 }
 190
 191 static void destroy_inodecache(void)
 192 {
 193         kmem_cache_destroy(ext2_inode_cachep);
 194 }
 195
 196 static void ext2_clear_inode(struct inode *inode)
 197 {
 198         struct ext2_block_alloc_info *rsv = EXT2_I(inode)->i_block_alloc_info;
 199
 200         dquot_drop(inode);
 201         ext2_discard_reservation(inode);
 202         EXT2_I(inode)->i_block_alloc_info = NULL;
 203         if (unlikely(rsv))
 204                 kfree(rsv);
 205 }
 206
 207 static int ext2_show_options(struct seq_file *seq, struct vfsmount *vfs)
 208 {
 209         struct super_block *sb = vfs->mnt_sb;
 210         struct ext2_sb_info *sbi = EXT2_SB(sb);
 211         struct ext2_super_block *es = sbi->s_es;
 212         unsigned long def_mount_opts;
 213
 214         spin_lock(&sbi->s_lock);
 215         def_mount_opts = le32_to_cpu(es->s_default_mount_opts);
 216
 217         if (sbi->s_sb_block != 1)
 218                 seq_printf(seq, ",sb=%lu", sbi->s_sb_block);
 219         if (test_opt(sb, MINIX_DF))
 220                 seq_puts(seq, ",minixdf");
 221         if (test_opt(sb, GRPID))
 222                 seq_puts(seq, ",grpid");
 223         if (!test_opt(sb, GRPID) && (def_mount_opts & EXT2_DEFM_BSDGROUPS))
 224                 seq_puts(seq, ",nogrpid");
 225         if (sbi->s_resuid != EXT2_DEF_RESUID ||
 226             le16_to_cpu(es->s_def_resuid) != EXT2_DEF_RESUID) {
 227                 seq_printf(seq, ",resuid=%u", sbi->s_resuid);
 228         }
 229         if (sbi->s_resgid != EXT2_DEF_RESGID ||
 230             le16_to_cpu(es->s_def_resgid) != EXT2_DEF_RESGID) {
 231                 seq_printf(seq, ",resgid=%u", sbi->s_resgid);
 232         }
 233         if (test_opt(sb, ERRORS_RO)) {
 234                 int def_errors = le16_to_cpu(es->s_errors);
 235
 236                 if (def_errors == EXT2_ERRORS_PANIC ||
 237                     def_errors == EXT2_ERRORS_CONTINUE) {
 238                         seq_puts(seq, ",errors=remount-ro");
 239                 }
 240         }
 241         if (test_opt(sb, ERRORS_CONT))
 242                 seq_puts(seq, ",errors=continue");
 243         if (test_opt(sb, ERRORS_PANIC))
 244                 seq_puts(seq, ",errors=panic");
 245         if (test_opt(sb, NO_UID32))
 246                 seq_puts(seq, ",nouid32");
 247         if (test_opt(sb, DEBUG))
 248                 seq_puts(seq, ",debug");
 249         if (test_opt(sb, OLDALLOC))
 250                 seq_puts(seq, ",oldalloc");
 251
 252 #ifdef CONFIG_EXT2_FS_XATTR
 253         if (test_opt(sb, XATTR_USER))
 254                 seq_puts(seq, ",user_xattr");
 255         if (!test_opt(sb, XATTR_USER) &&
 256             (def_mount_opts & EXT2_DEFM_XATTR_USER)) {
 257                 seq_puts(seq, ",nouser_xattr");
 258         }
 259 #endif
 260
 261 #ifdef CONFIG_EXT2_FS_POSIX_ACL
 262         if (test_opt(sb, POSIX_ACL))
 263                 seq_puts(seq, ",acl");
 264         if (!test_opt(sb, POSIX_ACL) && (def_mount_opts & EXT2_DEFM_ACL))
 265                 seq_puts(seq, ",noacl");
 266 #endif
 267
 268         if (test_opt(sb, NOBH))
 269                 seq_puts(seq, ",nobh");
 270
 271 #if defined(CONFIG_QUOTA)
 272         if (sbi->s_mount_opt & EXT2_MOUNT_USRQUOTA)
 273                 seq_puts(seq, ",usrquota");
 274
 275         if (sbi->s_mount_opt & EXT2_MOUNT_GRPQUOTA)
 276                 seq_puts(seq, ",grpquota");
 277 #endif
 278
 279 #if defined(CONFIG_EXT2_FS_XIP)
 280         if (sbi->s_mount_opt & EXT2_MOUNT_XIP)
 281                 seq_puts(seq, ",xip");
 282 #endif
 283
 284         if (!test_opt(sb, RESERVATION))
 285                 seq_puts(seq, ",noreservation");
 286
 287         spin_unlock(&sbi->s_lock);
 288         return 0;
 289 }
 290
 291 #ifdef CONFIG_QUOTA
 292 static ssize_t ext2_quota_read(struct super_block *sb, int type, char *data, size_t len, loff_t off);
 293 static ssize_t ext2_quota_write(struct super_block *sb, int type, const char *data, size_t len, loff_t off);
 294 #endif
 295
 296 static const struct super_operations ext2_sops = {
 297         .alloc_inode    = ext2_alloc_inode,
 298         .destroy_inode  = ext2_destroy_inode,
 299         .write_inode    = ext2_write_inode,
 300         .delete_inode   = ext2_delete_inode,
 301         .put_super      = ext2_put_super,
 302         .write_super    = ext2_write_super,
 303         .sync_fs        = ext2_sync_fs,
 304         .statfs         = ext2_statfs,
 305         .remount_fs     = ext2_remount,
 306         .clear_inode    = ext2_clear_inode,
 307         .show_options   = ext2_show_options,
 308 #ifdef CONFIG_QUOTA
 309         .quota_read     = ext2_quota_read,
 310         .quota_write    = ext2_quota_write,
 311 #endif
 312 };
 313
 314 static struct inode *ext2_nfs_get_inode(struct super_block *sb,
 315                 u64 ino, u32 generation)
 316 {
 317         struct inode *inode;
 318
 319         if (ino < EXT2_FIRST_INO(sb) && ino != EXT2_ROOT_INO)
 320                 return ERR_PTR(-ESTALE);
 321         if (ino > le32_to_cpu(EXT2_SB(sb)->s_es->s_inodes_count))
 322                 return ERR_PTR(-ESTALE);
 323
 324         /* iget isn't really right if the inode is currently unallocated!!
 325          * ext2_read_inode currently does appropriate checks, but
 326          * it might be "neater" to call ext2_get_inode first and check
 327          * if the inode is valid.....
 328          */
 329         inode = ext2_iget(sb, ino);
 330         if (IS_ERR(inode))
 331                 return ERR_CAST(inode);
 332         if (generation && inode->i_generation != generation) {
 333                 /* we didn't find the right inode.. */
 334                 iput(inode);
 335                 return ERR_PTR(-ESTALE);
 336         }
 337         return inode;
 338 }
 339
 340 static struct dentry *ext2_fh_to_dentry(struct super_block *sb, struct fid *fid,
 341                 int fh_len, int fh_type)
 342 {
 343         return generic_fh_to_dentry(sb, fid, fh_len, fh_type,
 344                                     ext2_nfs_get_inode);
 345 }
 346
 347 static struct dentry *ext2_fh_to_parent(struct super_block *sb, struct fid *fid,
 348                 int fh_len, int fh_type)
 349 {
 350         return generic_fh_to_parent(sb, fid, fh_len, fh_type,
 351                                     ext2_nfs_get_inode);
 352 }
 353
 354 /* Yes, most of these are left as NULL!!
 355  * A NULL value implies the default, which works with ext2-like file
 356  * systems, but can be improved upon.
 357  * Currently only get_parent is required.
 358  */
 359 static const struct export_operations ext2_export_ops = {
 360         .fh_to_dentry = ext2_fh_to_dentry,
 361         .fh_to_parent = ext2_fh_to_parent,
 362         .get_parent = ext2_get_parent,
 363 };
 364
 365 static unsigned long get_sb_block(void **data)
 366 {
 367         unsigned long   sb_block;
 368         char            *options = (char *) *data;
 369
 370         if (!options || strncmp(options, "sb=", 3) != 0)
 371                 return 1;       /* Default location */
 372         options += 3;
 373         sb_block = simple_strtoul(options, &options, 0);
 374         if (*options && *options != ',') {
 375                 printk("EXT2-fs: Invalid sb specification: %s\n",
 376                        (char *) *data);
 377                 return 1;
 378         }
 379         if (*options == ',')
 380                 options++;
 381         *data = (void *) options;
 382         return sb_block;
 383 }
 384
 385 enum {
 386         Opt_bsd_df, Opt_minix_df, Opt_grpid, Opt_nogrpid,
 387         Opt_resgid, Opt_resuid, Opt_sb, Opt_err_cont, Opt_err_panic,
 388         Opt_err_ro, Opt_nouid32, Opt_nocheck, Opt_debug,
 389         Opt_oldalloc, Opt_orlov, Opt_nobh, Opt_user_xattr, Opt_nouser_xattr,
 390         Opt_acl, Opt_noacl, Opt_xip, Opt_ignore, Opt_err, Opt_quota,
 391         Opt_usrquota, Opt_grpquota, Opt_reservation, Opt_noreservation
 392 };
 393
 394 static const match_table_t tokens = {
 395         {Opt_bsd_df, "bsddf"},
 396         {Opt_minix_df, "minixdf"},
 397         {Opt_grpid, "grpid"},
 398         {Opt_grpid, "bsdgroups"},
 399         {Opt_nogrpid, "nogrpid"},
 400         {Opt_nogrpid, "sysvgroups"},
 401         {Opt_resgid, "resgid=%u"},
 402         {Opt_resuid, "resuid=%u"},
 403         {Opt_sb, "sb=%u"},
 404         {Opt_err_cont, "errors=continue"},
 405         {Opt_err_panic, "errors=panic"},
 406         {Opt_err_ro, "errors=remount-ro"},
 407         {Opt_nouid32, "nouid32"},
 408         {Opt_nocheck, "check=none"},
 409         {Opt_nocheck, "nocheck"},
 410         {Opt_debug, "debug"},
 411         {Opt_oldalloc, "oldalloc"},
 412         {Opt_orlov, "orlov"},
 413         {Opt_nobh, "nobh"},
 414         {Opt_user_xattr, "user_xattr"},
 415         {Opt_nouser_xattr, "nouser_xattr"},
 416         {Opt_acl, "acl"},
 417         {Opt_noacl, "noacl"},
 418         {Opt_xip, "xip"},
 419         {Opt_grpquota, "grpquota"},
 420         {Opt_ignore, "noquota"},
 421         {Opt_quota, "quota"},
 422         {Opt_usrquota, "usrquota"},
 423         {Opt_reservation, "reservation"},
 424         {Opt_noreservation, "noreservation"},
 425         {Opt_err, NULL}
 426 };
 427
 428 static int parse_options(char *options, struct super_block *sb)
 429 {
 430         char *p;
 431         struct ext2_sb_info *sbi = EXT2_SB(sb);
 432         substring_t args[MAX_OPT_ARGS];
 433         int option;
 434
 435         if (!options)
 436                 return 1;
 437
 438         while ((p = strsep (&options, ",")) != NULL) {
 439                 int token;
 440                 if (!*p)
 441                         continue;
 442
 443                 token = match_token(p, tokens, args);
 444                 switch (token) {
 445                 case Opt_bsd_df:
 446                         clear_opt (sbi->s_mount_opt, MINIX_DF);
 447                         break;
 448                 case Opt_minix_df:
 449                         set_opt (sbi->s_mount_opt, MINIX_DF);
 450                         break;
 451                 case Opt_grpid:
 452                         set_opt (sbi->s_mount_opt, GRPID);
 453                         break;
 454                 case Opt_nogrpid:
 455                         clear_opt (sbi->s_mount_opt, GRPID);
 456                         break;
 457                 case Opt_resuid:
 458                         if (match_int(&args[0], &option))
 459                                 return 0;
 460                         sbi->s_resuid = option;
 461                         break;
 462                 case Opt_resgid:
 463                         if (match_int(&args[0], &option))
 464                                 return 0;
 465                         sbi->s_resgid = option;
 466                         break;
 467                 case Opt_sb:
 468                         /* handled by get_sb_block() instead of here */
 469                         /* *sb_block = match_int(&args[0]); */
 470                         break;
 471                 case Opt_err_panic:
 472                         clear_opt (sbi->s_mount_opt, ERRORS_CONT);
 473                         clear_opt (sbi->s_mount_opt, ERRORS_RO);
 474                         set_opt (sbi->s_mount_opt, ERRORS_PANIC);
 475                         break;
 476                 case Opt_err_ro:
 477                         clear_opt (sbi->s_mount_opt, ERRORS_CONT);
 478                         clear_opt (sbi->s_mount_opt, ERRORS_PANIC);
 479                         set_opt (sbi->s_mount_opt, ERRORS_RO);
 480                         break;
 481                 case Opt_err_cont:
 482                         clear_opt (sbi->s_mount_opt, ERRORS_RO);
 483                         clear_opt (sbi->s_mount_opt, ERRORS_PANIC);
 484                         set_opt (sbi->s_mount_opt, ERRORS_CONT);
 485                         break;
 486                 case Opt_nouid32:
 487                         set_opt (sbi->s_mount_opt, NO_UID32);
 488                         break;
 489                 case Opt_nocheck:
 490                         clear_opt (sbi->s_mount_opt, CHECK);
 491                         break;
 492                 case Opt_debug:
 493                         set_opt (sbi->s_mount_opt, DEBUG);
 494                         break;
 495                 case Opt_oldalloc:
 496                         set_opt (sbi->s_mount_opt, OLDALLOC);
 497                         break;
 498                 case Opt_orlov:
 499                         clear_opt (sbi->s_mount_opt, OLDALLOC);
 500                         break;
 501                 case Opt_nobh:
 502                         set_opt (sbi->s_mount_opt, NOBH);
 503                         break;
 504 #ifdef CONFIG_EXT2_FS_XATTR
 505                 case Opt_user_xattr:
 506                         set_opt (sbi->s_mount_opt, XATTR_USER);
 507                         break;
 508                 case Opt_nouser_xattr:
 509                         clear_opt (sbi->s_mount_opt, XATTR_USER);
 510                         break;
 511 #else
 512                 case Opt_user_xattr:
 513                 case Opt_nouser_xattr:
 514                         ext2_msg(sb, KERN_INFO, "(no)user_xattr options"
 515                                 "not supported");
 516                         break;
 517 #endif
 518 #ifdef CONFIG_EXT2_FS_POSIX_ACL
 519                 case Opt_acl:
 520                         set_opt(sbi->s_mount_opt, POSIX_ACL);
 521                         break;
 522                 case Opt_noacl:
 523                         clear_opt(sbi->s_mount_opt, POSIX_ACL);
 524                         break;
 525 #else
 526                 case Opt_acl:
 527                 case Opt_noacl:
 528                         ext2_msg(sb, KERN_INFO,
 529                                 "(no)acl options not supported");
 530                         break;
 531 #endif
 532                 case Opt_xip:
 533 #ifdef CONFIG_EXT2_FS_XIP
 534                         set_opt (sbi->s_mount_opt, XIP);
 535 #else
 536                         ext2_msg(sb, KERN_INFO, "xip option not supported");
 537 #endif
 538                         break;
 539
 540 #if defined(CONFIG_QUOTA)
 541                 case Opt_quota:
 542                 case Opt_usrquota:
 543                         set_opt(sbi->s_mount_opt, USRQUOTA);
 544                         break;
 545
 546                 case Opt_grpquota:
 547                         set_opt(sbi->s_mount_opt, GRPQUOTA);
 548                         break;
 549 #else
 550                 case Opt_quota:
 551                 case Opt_usrquota:
 552                 case Opt_grpquota:
 553                         ext2_msg(sb, KERN_INFO,
 554                                 "quota operations not supported");
 555                         break;
 556 #endif
 557
 558                 case Opt_reservation:
 559                         set_opt(sbi->s_mount_opt, RESERVATION);
 560                         ext2_msg(sb, KERN_INFO, "reservations ON");
 561                         break;
 562                 case Opt_noreservation:
 563                         clear_opt(sbi->s_mount_opt, RESERVATION);
 564                         ext2_msg(sb, KERN_INFO, "reservations OFF");
 565                         break;
 566                 case Opt_ignore:
 567                         break;
 568                 default:
 569                         return 0;
 570                 }
 571         }
 572         return 1;
 573 }
 574
 575 static int ext2_setup_super (struct super_block * sb,
 576                               struct ext2_super_block * es,
 577                               int read_only)
 578 {
 579         int res = 0;
 580         struct ext2_sb_info *sbi = EXT2_SB(sb);
 581
 582         if (le32_to_cpu(es->s_rev_level) > EXT2_MAX_SUPP_REV) {
 583                 ext2_msg(sb, KERN_ERR,
 584                         "error: revision level too high, "
 585                         "forcing read-only mode");
 586                 res = MS_RDONLY;
 587         }
 588         if (read_only)
 589                 return res;
 590         if (!(sbi->s_mount_state & EXT2_VALID_FS))
 591                 ext2_msg(sb, KERN_WARNING,
 592                         "warning: mounting unchecked fs, "
 593                         "running e2fsck is recommended");
 594         else if ((sbi->s_mount_state & EXT2_ERROR_FS))
 595                 ext2_msg(sb, KERN_WARNING,
 596                         "warning: mounting fs with errors, "
 597                         "running e2fsck is recommended");
 598         else if ((__s16) le16_to_cpu(es->s_max_mnt_count) >= 0 &&
 599                  le16_to_cpu(es->s_mnt_count) >=
 600                  (unsigned short) (__s16) le16_to_cpu(es->s_max_mnt_count))
 601                 ext2_msg(sb, KERN_WARNING,
 602                         "warning: maximal mount count reached, "
 603                         "running e2fsck is recommended");
 604         else if (le32_to_cpu(es->s_checkinterval) &&
 605                 (le32_to_cpu(es->s_lastcheck) +
 606                         le32_to_cpu(es->s_checkinterval) <= get_seconds()))
 607                 ext2_msg(sb, KERN_WARNING,
 608                         "warning: checktime reached, "
 609                         "running e2fsck is recommended");
 610         if (!le16_to_cpu(es->s_max_mnt_count))
 611                 es->s_max_mnt_count = cpu_to_le16(EXT2_DFL_MAX_MNT_COUNT);
 612         le16_add_cpu(&es->s_mnt_count, 1);
 613         if (test_opt (sb, DEBUG))
 614                 ext2_msg(sb, KERN_INFO, "%s, %s, bs=%lu, fs=%lu, gc=%lu, "
 615                         "bpg=%lu, ipg=%lu, mo=%04lx]",
 616                         EXT2FS_VERSION, EXT2FS_DATE, sb->s_blocksize,
 617                         sbi->s_frag_size,
 618                         sbi->s_groups_count,
 619                         EXT2_BLOCKS_PER_GROUP(sb),
 620                         EXT2_INODES_PER_GROUP(sb),
 621                         sbi->s_mount_opt);
 622         return res;
 623 }
 624
 625 static int ext2_check_descriptors(struct super_block *sb)
 626 {
 627         int i;
 628         struct ext2_sb_info *sbi = EXT2_SB(sb);
 629
 630         ext2_debug ("Checking group descriptors");
 631
 632         for (i = 0; i < sbi->s_groups_count; i++) {
 633                 struct ext2_group_desc *gdp = ext2_get_group_desc(sb, i, NULL);
 634                 ext2_fsblk_t first_block = ext2_group_first_block_no(sb, i);
 635                 ext2_fsblk_t last_block;
 636
 637                 if (i == sbi->s_groups_count - 1)
 638                         last_block = le32_to_cpu(sbi->s_es->s_blocks_count) - 1;
 639                 else
 640                         last_block = first_block +
 641                                 (EXT2_BLOCKS_PER_GROUP(sb) - 1);
 642
 643                 if (le32_to_cpu(gdp->bg_block_bitmap) < first_block ||
 644                     le32_to_cpu(gdp->bg_block_bitmap) > last_block)
 645                 {
 646                         ext2_error (sb, "ext2_check_descriptors",
 647                                     "Block bitmap for group %d"
 648                                     " not in group (block %lu)!",
 649                                     i, (unsigned long) le32_to_cpu(gdp->bg_block_bitmap));
 650                         return 0;
 651                 }
 652                 if (le32_to_cpu(gdp->bg_inode_bitmap) < first_block ||
 653                     le32_to_cpu(gdp->bg_inode_bitmap) > last_block)
 654                 {
 655                         ext2_error (sb, "ext2_check_descriptors",
 656                                     "Inode bitmap for group %d"
 657                                     " not in group (block %lu)!",
 658                                     i, (unsigned long) le32_to_cpu(gdp->bg_inode_bitmap));
 659                         return 0;
 660                 }
 661                 if (le32_to_cpu(gdp->bg_inode_table) < first_block ||
 662                     le32_to_cpu(gdp->bg_inode_table) + sbi->s_itb_per_group - 1 >
 663                     last_block)
 664                 {
 665                         ext2_error (sb, "ext2_check_descriptors",
 666                                     "Inode table for group %d"
 667                                     " not in group (block %lu)!",
 668                                     i, (unsigned long) le32_to_cpu(gdp->bg_inode_table));
 669                         return 0;
 670                 }
 671         }
 672         return 1;
 673 }
 674
 675 /*
 676  * Maximal file size.  There is a direct, and {,double-,triple-}indirect
 677  * block limit, and also a limit of (2^32 - 1) 512-byte sectors in i_blocks.
 678  * We need to be 1 filesystem block less than the 2^32 sector limit.
 679  */
 680 static loff_t ext2_max_size(int bits)
 681 {
 682         loff_t res = EXT2_NDIR_BLOCKS;
 683         int meta_blocks;
 684         loff_t upper_limit;
 685
 686         /* This is calculated to be the largest file size for a
 687          * dense, file such that the total number of
 688          * sectors in the file, including data and all indirect blocks,
 689          * does not exceed 2^32 -1
 690          * __u32 i_blocks representing the total number of
 691          * 512 bytes blocks of the file
 692          */
 693         upper_limit = (1LL << 32) - 1;
 694
 695         /* total blocks in file system block size */
 696         upper_limit >>= (bits - 9);
 697
 698
 699         /* indirect blocks */
 700         meta_blocks = 1;
 701         /* double indirect blocks */
 702         meta_blocks += 1 + (1LL << (bits-2));
 703         /* tripple indirect blocks */
 704         meta_blocks += 1 + (1LL << (bits-2)) + (1LL << (2*(bits-2)));
 705
 706         upper_limit -= meta_blocks;
 707         upper_limit <<= bits;
 708
 709         res += 1LL << (bits-2);
 710         res += 1LL << (2*(bits-2));
 711         res += 1LL << (3*(bits-2));
 712         res <<= bits;
 713         if (res > upper_limit)
 714                 res = upper_limit;
 715
 716         if (res > MAX_LFS_FILESIZE)
 717                 res = MAX_LFS_FILESIZE;
 718
 719         return res;
 720 }
 721
 722 static unsigned long descriptor_loc(struct super_block *sb,
 723                                     unsigned long logic_sb_block,
 724                                     int nr)
 725 {
 726         struct ext2_sb_info *sbi = EXT2_SB(sb);
 727         unsigned long bg, first_meta_bg;
 728         int has_super = 0;
 729
 730         first_meta_bg = le32_to_cpu(sbi->s_es->s_first_meta_bg);
 731
 732         if (!EXT2_HAS_INCOMPAT_FEATURE(sb, EXT2_FEATURE_INCOMPAT_META_BG) ||
 733             nr < first_meta_bg)
 734                 return (logic_sb_block + nr + 1);
 735         bg = sbi->s_desc_per_block * nr;
 736         if (ext2_bg_has_super(sb, bg))
 737                 has_super = 1;
 738
 739         return ext2_group_first_block_no(sb, bg) + has_super;
 740 }
 741
 742 static int ext2_fill_super(struct super_block *sb, void *data, int silent)
 743 {
 744         struct buffer_head * bh;
 745         struct ext2_sb_info * sbi;
 746         struct ext2_super_block * es;
 747         struct inode *root;
 748         unsigned long block;
 749         unsigned long sb_block = get_sb_block(&data);
 750         unsigned long logic_sb_block;
 751         unsigned long offset = 0;
 752         unsigned long def_mount_opts;
 753         long ret = -EINVAL;
 754         int blocksize = BLOCK_SIZE;
 755         int db_count;
 756         int i, j;
 757         __le32 features;
 758         int err;
 759
 760         sbi = kzalloc(sizeof(*sbi), GFP_KERNEL);
 761         if (!sbi)
 762                 return -ENOMEM;
 763
 764         sbi->s_blockgroup_lock =
 765                 kzalloc(sizeof(struct blockgroup_lock), GFP_KERNEL);
 766         if (!sbi->s_blockgroup_lock) {
 767                 kfree(sbi);
 768                 return -ENOMEM;
 769         }
 770         sb->s_fs_info = sbi;
 771         sbi->s_sb_block = sb_block;
 772
 773         spin_lock_init(&sbi->s_lock);
 774
 775         /*
 776          * See what the current blocksize for the device is, and
 777          * use that as the blocksize.  Otherwise (or if the blocksize
 778          * is smaller than the default) use the default.
 779          * This is important for devices that have a hardware
 780          * sectorsize that is larger than the default.
 781          */
 782         blocksize = sb_min_blocksize(sb, BLOCK_SIZE);
 783         if (!blocksize) {
 784                 ext2_msg(sb, KERN_ERR, "error: unable to set blocksize");
 785                 goto failed_sbi;
 786         }
 787
 788         /*
 789          * If the superblock doesn't start on a hardware sector boundary,
 790          * calculate the offset.
 791          */
 792         if (blocksize != BLOCK_SIZE) {
 793                 logic_sb_block = (sb_block*BLOCK_SIZE) / blocksize;
 794                 offset = (sb_block*BLOCK_SIZE) % blocksize;
 795         } else {
 796                 logic_sb_block = sb_block;
 797         }
 798
 799         if (!(bh = sb_bread(sb, logic_sb_block))) {
 800                 ext2_msg(sb, KERN_ERR, "error: unable to read superblock");
 801                 goto failed_sbi;
 802         }
 803         /*
 804          * Note: s_es must be initialized as soon as possible because
 805          *       some ext2 macro-instructions depend on its value
 806          */
 807         es = (struct ext2_super_block *) (((char *)bh->b_data) + offset);
 808         sbi->s_es = es;
 809         sb->s_magic = le16_to_cpu(es->s_magic);
 810
 811         if (sb->s_magic != EXT2_SUPER_MAGIC)
 812                 goto cantfind_ext2;
 813
 814         /* Set defaults before we parse the mount options */
 815         def_mount_opts = le32_to_cpu(es->s_default_mount_opts);
 816         if (def_mount_opts & EXT2_DEFM_DEBUG)
 817                 set_opt(sbi->s_mount_opt, DEBUG);
 818         if (def_mount_opts & EXT2_DEFM_BSDGROUPS)
 819                 set_opt(sbi->s_mount_opt, GRPID);
 820         if (def_mount_opts & EXT2_DEFM_UID16)
 821                 set_opt(sbi->s_mount_opt, NO_UID32);
 822 #ifdef CONFIG_EXT2_FS_XATTR
 823         if (def_mount_opts & EXT2_DEFM_XATTR_USER)
 824                 set_opt(sbi->s_mount_opt, XATTR_USER);
 825 #endif
 826 #ifdef CONFIG_EXT2_FS_POSIX_ACL
 827         if (def_mount_opts & EXT2_DEFM_ACL)
 828                 set_opt(sbi->s_mount_opt, POSIX_ACL);
 829 #endif
 830
 831         if (le16_to_cpu(sbi->s_es->s_errors) == EXT2_ERRORS_PANIC)
 832                 set_opt(sbi->s_mount_opt, ERRORS_PANIC);
 833         else if (le16_to_cpu(sbi->s_es->s_errors) == EXT2_ERRORS_CONTINUE)
 834                 set_opt(sbi->s_mount_opt, ERRORS_CONT);
 835         else
 836                 set_opt(sbi->s_mount_opt, ERRORS_RO);
 837
 838         sbi->s_resuid = le16_to_cpu(es->s_def_resuid);
 839         sbi->s_resgid = le16_to_cpu(es->s_def_resgid);
 840
 841         set_opt(sbi->s_mount_opt, RESERVATION);
 842
 843         if (!parse_options((char *) data, sb))
 844                 goto failed_mount;
 845
 846         sb->s_flags = (sb->s_flags & ~MS_POSIXACL) |
 847                 ((EXT2_SB(sb)->s_mount_opt & EXT2_MOUNT_POSIX_ACL) ?
 848                  MS_POSIXACL : 0);
 849
 850         ext2_xip_verify_sb(sb); /* see if bdev supports xip, unset
 851                                     EXT2_MOUNT_XIP if not */
 852
 853         if (le32_to_cpu(es->s_rev_level) == EXT2_GOOD_OLD_REV &&
 854             (EXT2_HAS_COMPAT_FEATURE(sb, ~0U) ||
 855              EXT2_HAS_RO_COMPAT_FEATURE(sb, ~0U) ||
 856              EXT2_HAS_INCOMPAT_FEATURE(sb, ~0U)))
 857                 ext2_msg(sb, KERN_WARNING,
 858                         "warning: feature flags set on rev 0 fs, "
 859                         "running e2fsck is recommended");
 860         /*
 861          * Check feature flags regardless of the revision level, since we
 862          * previously didn't change the revision level when setting the flags,
 863          * so there is a chance incompat flags are set on a rev 0 filesystem.
 864          */
 865         features = EXT2_HAS_INCOMPAT_FEATURE(sb, ~EXT2_FEATURE_INCOMPAT_SUPP);
 866         if (features) {
 867                 ext2_msg(sb, KERN_ERR,  "error: couldn't mount because of "
 868                        "unsupported optional features (%x)",
 869                         le32_to_cpu(features));
 870                 goto failed_mount;
 871         }
 872         if (!(sb->s_flags & MS_RDONLY) &&
 873             (features = EXT2_HAS_RO_COMPAT_FEATURE(sb, ~EXT2_FEATURE_RO_COMPAT_SUPP))){
 874                 ext2_msg(sb, KERN_ERR, "error: couldn't mount RDWR because of "
 875                        "unsupported optional features (%x)",
 876                        le32_to_cpu(features));
 877                 goto failed_mount;
 878         }
 879
 880         blocksize = BLOCK_SIZE << le32_to_cpu(sbi->s_es->s_log_block_size);
 881
 882         if (ext2_use_xip(sb) && blocksize != PAGE_SIZE) {
 883                 if (!silent)
 884                         ext2_msg(sb, KERN_ERR,
 885                                 "error: unsupported blocksize for xip");
 886                 goto failed_mount;
 887         }
 888
 889         /* If the blocksize doesn't match, re-read the thing.. */
 890         if (sb->s_blocksize != blocksize) {
 891                 brelse(bh);
 892
 893                 if (!sb_set_blocksize(sb, blocksize)) {
 894                         ext2_msg(sb, KERN_ERR, "error: blocksize is too small");
 895                         goto failed_sbi;
 896                 }
 897
 898                 logic_sb_block = (sb_block*BLOCK_SIZE) / blocksize;
 899                 offset = (sb_block*BLOCK_SIZE) % blocksize;
 900                 bh = sb_bread(sb, logic_sb_block);
 901                 if(!bh) {
 902                         ext2_msg(sb, KERN_ERR, "error: couldn't read"
 903                                 "superblock on 2nd try");
 904                         goto failed_sbi;
 905                 }
 906                 es = (struct ext2_super_block *) (((char *)bh->b_data) + offset);
 907                 sbi->s_es = es;
 908                 if (es->s_magic != cpu_to_le16(EXT2_SUPER_MAGIC)) {
 909                         ext2_msg(sb, KERN_ERR, "error: magic mismatch");
 910                         goto failed_mount;
 911                 }
 912         }
 913
 914         sb->s_maxbytes = ext2_max_size(sb->s_blocksize_bits);
 915
 916         if (le32_to_cpu(es->s_rev_level) == EXT2_GOOD_OLD_REV) {
 917                 sbi->s_inode_size = EXT2_GOOD_OLD_INODE_SIZE;
 918                 sbi->s_first_ino = EXT2_GOOD_OLD_FIRST_INO;
 919         } else {
 920                 sbi->s_inode_size = le16_to_cpu(es->s_inode_size);
 921                 sbi->s_first_ino = le32_to_cpu(es->s_first_ino);
 922                 if ((sbi->s_inode_size < EXT2_GOOD_OLD_INODE_SIZE) ||
 923                     !is_power_of_2(sbi->s_inode_size) ||
 924                     (sbi->s_inode_size > blocksize)) {
 925                         ext2_msg(sb, KERN_ERR,
 926                                 "error: unsupported inode size: %d",
 927                                 sbi->s_inode_size);
 928                         goto failed_mount;
 929                 }
 930         }
 931
 932         sbi->s_frag_size = EXT2_MIN_FRAG_SIZE <<
 933                                    le32_to_cpu(es->s_log_frag_size);
 934         if (sbi->s_frag_size == 0)
 935                 goto cantfind_ext2;
 936         sbi->s_frags_per_block = sb->s_blocksize / sbi->s_frag_size;
 937
 938         sbi->s_blocks_per_group = le32_to_cpu(es->s_blocks_per_group);
 939         sbi->s_frags_per_group = le32_to_cpu(es->s_frags_per_group);
 940         sbi->s_inodes_per_group = le32_to_cpu(es->s_inodes_per_group);
 941
 942         if (EXT2_INODE_SIZE(sb) == 0)
 943                 goto cantfind_ext2;
 944         sbi->s_inodes_per_block = sb->s_blocksize / EXT2_INODE_SIZE(sb);
 945         if (sbi->s_inodes_per_block == 0 || sbi->s_inodes_per_group == 0)
 946                 goto cantfind_ext2;
 947         sbi->s_itb_per_group = sbi->s_inodes_per_group /
 948                                         sbi->s_inodes_per_block;
 949         sbi->s_desc_per_block = sb->s_blocksize /
 950                                         sizeof (struct ext2_group_desc);
 951         sbi->s_sbh = bh;
 952         sbi->s_mount_state = le16_to_cpu(es->s_state);
 953         sbi->s_addr_per_block_bits =
 954                 ilog2 (EXT2_ADDR_PER_BLOCK(sb));
 955         sbi->s_desc_per_block_bits =
 956                 ilog2 (EXT2_DESC_PER_BLOCK(sb));
 957
 958         if (sb->s_magic != EXT2_SUPER_MAGIC)
 959                 goto cantfind_ext2;
 960
 961         if (sb->s_blocksize != bh->b_size) {
 962                 if (!silent)
 963                         ext2_msg(sb, KERN_ERR, "error: unsupported blocksize");
 964                 goto failed_mount;
 965         }
 966
 967         if (sb->s_blocksize != sbi->s_frag_size) {
 968                 ext2_msg(sb, KERN_ERR,
 969                         "error: fragsize %lu != blocksize %lu"
 970                         "(not supported yet)",
 971                         sbi->s_frag_size, sb->s_blocksize);
 972                 goto failed_mount;
 973         }
 974
 975         if (sbi->s_blocks_per_group > sb->s_blocksize * 8) {
 976                 ext2_msg(sb, KERN_ERR,
 977                         "error: #blocks per group too big: %lu",
 978                         sbi->s_blocks_per_group);
 979                 goto failed_mount;
 980         }
 981         if (sbi->s_frags_per_group > sb->s_blocksize * 8) {
 982                 ext2_msg(sb, KERN_ERR,
 983                         "error: #fragments per group too big: %lu",
 984                         sbi->s_frags_per_group);
 985                 goto failed_mount;
 986         }
 987         if (sbi->s_inodes_per_group > sb->s_blocksize * 8) {
 988                 ext2_msg(sb, KERN_ERR,
 989                         "error: #inodes per group too big: %lu",
 990                         sbi->s_inodes_per_group);
 991                 goto failed_mount;
 992         }
 993
 994         if (EXT2_BLOCKS_PER_GROUP(sb) == 0)
 995                 goto cantfind_ext2;
 996         sbi->s_groups_count = ((le32_to_cpu(es->s_blocks_count) -
 997                                 le32_to_cpu(es->s_first_data_block) - 1)
 998                                         / EXT2_BLOCKS_PER_GROUP(sb)) + 1;
 999         db_count = (sbi->s_groups_count + EXT2_DESC_PER_BLOCK(sb) - 1) /
1000                    EXT2_DESC_PER_BLOCK(sb);
1001         sbi->s_group_desc = kmalloc (db_count * sizeof (struct buffer_head *), GFP_KERNEL);
1002         if (sbi->s_group_desc == NULL) {
1003                 ext2_msg(sb, KERN_ERR, "error: not enough memory");
1004                 goto failed_mount;
1005         }
1006         bgl_lock_init(sbi->s_blockgroup_lock);
1007         sbi->s_debts = kcalloc(sbi->s_groups_count, sizeof(*sbi->s_debts), GFP_KERNEL);
1008         if (!sbi->s_debts) {
1009                 ext2_msg(sb, KERN_ERR, "error: not enough memory");
1010                 goto failed_mount_group_desc;
1011         }
1012         for (i = 0; i < db_count; i++) {
1013                 block = descriptor_loc(sb, logic_sb_block, i);
1014                 sbi->s_group_desc[i] = sb_bread(sb, block);
1015                 if (!sbi->s_group_desc[i]) {
1016                         for (j = 0; j < i; j++)
1017                                 brelse (sbi->s_group_desc[j]);
1018                         ext2_msg(sb, KERN_ERR,
1019                                 "error: unable to read group descriptors");
1020                         goto failed_mount_group_desc;
1021                 }
1022         }
1023         if (!ext2_check_descriptors (sb)) {
1024                 ext2_msg(sb, KERN_ERR, "group descriptors corrupted");
1025                 goto failed_mount2;
1026         }
1027         sbi->s_gdb_count = db_count;
1028         get_random_bytes(&sbi->s_next_generation, sizeof(u32));
1029         spin_lock_init(&sbi->s_next_gen_lock);
1030
1031         /* per fileystem reservation list head & lock */
1032         spin_lock_init(&sbi->s_rsv_window_lock);
1033         sbi->s_rsv_window_root = RB_ROOT;
1034         /*
1035          * Add a single, static dummy reservation to the start of the
1036          * reservation window list --- it gives us a placeholder for
1037          * append-at-start-of-list which makes the allocation logic
1038          * _much_ simpler.
1039          */
1040         sbi->s_rsv_window_head.rsv_start = EXT2_RESERVE_WINDOW_NOT_ALLOCATED;
1041         sbi->s_rsv_window_head.rsv_end = EXT2_RESERVE_WINDOW_NOT_ALLOCATED;
1042         sbi->s_rsv_window_head.rsv_alloc_hit = 0;
1043         sbi->s_rsv_window_head.rsv_goal_size = 0;
1044         ext2_rsv_window_add(sb, &sbi->s_rsv_window_head);
1045
1046         err = percpu_counter_init(&sbi->s_freeblocks_counter,
1047                                 ext2_count_free_blocks(sb));
1048         if (!err) {
1049                 err = percpu_counter_init(&sbi->s_freeinodes_counter,
1050                                 ext2_count_free_inodes(sb));
1051         }
1052         if (!err) {
1053                 err = percpu_counter_init(&sbi->s_dirs_counter,
1054                                 ext2_count_dirs(sb));
1055         }
1056         if (err) {
1057                 ext2_msg(sb, KERN_ERR, "error: insufficient memory");
1058                 goto failed_mount3;
1059         }
1060         /*
1061          * set up enough so that it can read an inode
1062          */
1063         sb->s_op = &ext2_sops;
1064         sb->s_export_op = &ext2_export_ops;
1065         sb->s_xattr = ext2_xattr_handlers;
1066         root = ext2_iget(sb, EXT2_ROOT_INO);
1067         if (IS_ERR(root)) {
1068                 ret = PTR_ERR(root);
1069                 goto failed_mount3;
1070         }
1071         if (!S_ISDIR(root->i_mode) || !root->i_blocks || !root->i_size) {
1072                 iput(root);
1073                 ext2_msg(sb, KERN_ERR, "error: corrupt root inode, run e2fsck");
1074                 goto failed_mount3;
1075         }
1076
1077         sb->s_root = d_alloc_root(root);
1078         if (!sb->s_root) {
1079                 iput(root);
1080                 ext2_msg(sb, KERN_ERR, "error: get root inode failed");
1081                 ret = -ENOMEM;
1082                 goto failed_mount3;
1083         }
1084         if (EXT2_HAS_COMPAT_FEATURE(sb, EXT3_FEATURE_COMPAT_HAS_JOURNAL))
1085                 ext2_msg(sb, KERN_WARNING,
1086                         "warning: mounting ext3 filesystem as ext2");
1087         if (ext2_setup_super (sb, es, sb->s_flags & MS_RDONLY))
1088                 sb->s_flags |= MS_RDONLY;
1089         ext2_write_super(sb);
1090         return 0;
1091
1092 cantfind_ext2:
1093         if (!silent)
1094                 ext2_msg(sb, KERN_ERR,
1095                         "error: can't find an ext2 filesystem on dev %s.",
1096                         sb->s_id);
1097         goto failed_mount;
1098 failed_mount3:
1099         percpu_counter_destroy(&sbi->s_freeblocks_counter);
1100         percpu_counter_destroy(&sbi->s_freeinodes_counter);
1101         percpu_counter_destroy(&sbi->s_dirs_counter);
1102 failed_mount2:
1103         for (i = 0; i < db_count; i++)
1104                 brelse(sbi->s_group_desc[i]);
1105 failed_mount_group_desc:
1106         kfree(sbi->s_group_desc);
1107         kfree(sbi->s_debts);
1108 failed_mount:
1109         brelse(bh);
1110 failed_sbi:
1111         sb->s_fs_info = NULL;
1112         kfree(sbi->s_blockgroup_lock);
1113         kfree(sbi);
1114         return ret;
1115 }
1116
1117 static void ext2_clear_super_error(struct super_block *sb)
1118 {
1119         struct buffer_head *sbh = EXT2_SB(sb)->s_sbh;
1120
1121         if (buffer_write_io_error(sbh)) {
1122                 /*
1123                  * Oh, dear.  A previous attempt to write the
1124                  * superblock failed.  This could happen because the
1125                  * USB device was yanked out.  Or it could happen to
1126                  * be a transient write error and maybe the block will
1127                  * be remapped.  Nothing we can do but to retry the
1128                  * write and hope for the best.
1129                  */
1130                 ext2_msg(sb, KERN_ERR,
1131                        "previous I/O error to superblock detected\n");
1132                 clear_buffer_write_io_error(sbh);
1133                 set_buffer_uptodate(sbh);
1134         }
1135 }
1136
1137 static void ext2_sync_super(struct super_block *sb, struct ext2_super_block *es,
1138                             int wait)
1139 {
1140         ext2_clear_super_error(sb);
1141         spin_lock(&EXT2_SB(sb)->s_lock);
1142         es->s_free_blocks_count = cpu_to_le32(ext2_count_free_blocks(sb));
1143         es->s_free_inodes_count = cpu_to_le32(ext2_count_free_inodes(sb));
1144         es->s_wtime = cpu_to_le32(get_seconds());
1145         /* unlock before we do IO */
1146         spin_unlock(&EXT2_SB(sb)->s_lock);
1147         mark_buffer_dirty(EXT2_SB(sb)->s_sbh);
1148         if (wait)
1149                 sync_dirty_buffer(EXT2_SB(sb)->s_sbh);
1150         sb->s_dirt = 0;
1151 }
1152
1153 /*
1154  * In the second extended file system, it is not necessary to
1155  * write the super block since we use a mapping of the
1156  * disk super block in a buffer.
1157  *
1158  * However, this function is still used to set the fs valid
1159  * flags to 0.  We need to set this flag to 0 since the fs
1160  * may have been checked while mounted and e2fsck may have
1161  * set s_state to EXT2_VALID_FS after some corrections.
1162  */
1163 static int ext2_sync_fs(struct super_block *sb, int wait)
1164 {
1165         struct ext2_sb_info *sbi = EXT2_SB(sb);
1166         struct ext2_super_block *es = EXT2_SB(sb)->s_es;
1167
1168         spin_lock(&sbi->s_lock);
1169         if (es->s_state & cpu_to_le16(EXT2_VALID_FS)) {
1170                 ext2_debug("setting valid to 0\n");
1171                 es->s_state &= cpu_to_le16(~EXT2_VALID_FS);
1172         }
1173         spin_unlock(&sbi->s_lock);
1174         ext2_sync_super(sb, es, wait);
1175         return 0;
1176 }
1177
1178
1179 void ext2_write_super(struct super_block *sb)
1180 {
1181         if (!(sb->s_flags & MS_RDONLY))
1182                 ext2_sync_fs(sb, 1);
1183         else
1184                 sb->s_dirt = 0;
1185 }
1186
1187 static int ext2_remount (struct super_block * sb, int * flags, char * data)
1188 {
1189         struct ext2_sb_info * sbi = EXT2_SB(sb);
1190         struct ext2_super_block * es;
1191         unsigned long old_mount_opt = sbi->s_mount_opt;
1192         struct ext2_mount_options old_opts;
1193         unsigned long old_sb_flags;
1194         int err;
1195
1196         spin_lock(&sbi->s_lock);
1197
1198         /* Store the old options */
1199         old_sb_flags = sb->s_flags;
1200         old_opts.s_mount_opt = sbi->s_mount_opt;
1201         old_opts.s_resuid = sbi->s_resuid;
1202         old_opts.s_resgid = sbi->s_resgid;
1203
1204         /*
1205          * Allow the "check" option to be passed as a remount option.
1206          */
1207         if (!parse_options(data, sb)) {
1208                 err = -EINVAL;
1209                 goto restore_opts;
1210         }
1211
1212         sb->s_flags = (sb->s_flags & ~MS_POSIXACL) |
1213                 ((sbi->s_mount_opt & EXT2_MOUNT_POSIX_ACL) ? MS_POSIXACL : 0);
1214
1215         ext2_xip_verify_sb(sb); /* see if bdev supports xip, unset
1216                                     EXT2_MOUNT_XIP if not */
1217
1218         if ((ext2_use_xip(sb)) && (sb->s_blocksize != PAGE_SIZE)) {
1219                 ext2_msg(sb, KERN_WARNING,
1220                         "warning: unsupported blocksize for xip");
1221                 err = -EINVAL;
1222                 goto restore_opts;
1223         }
1224
1225         es = sbi->s_es;
1226         if (((sbi->s_mount_opt & EXT2_MOUNT_XIP) !=
1227             (old_mount_opt & EXT2_MOUNT_XIP)) &&
1228             invalidate_inodes(sb)) {
1229                 ext2_msg(sb, KERN_WARNING, "warning: refusing change of "
1230                          "xip flag with busy inodes while remounting");
1231                 sbi->s_mount_opt &= ~EXT2_MOUNT_XIP;
1232                 sbi->s_mount_opt |= old_mount_opt & EXT2_MOUNT_XIP;
1233         }
1234         if ((*flags & MS_RDONLY) == (sb->s_flags & MS_RDONLY)) {
1235                 spin_unlock(&sbi->s_lock);
1236                 return 0;
1237         }
1238         if (*flags & MS_RDONLY) {
1239                 if (le16_to_cpu(es->s_state) & EXT2_VALID_FS ||
1240                     !(sbi->s_mount_state & EXT2_VALID_FS)) {
1241                         spin_unlock(&sbi->s_lock);
1242                         return 0;
1243                 }
1244                 /*
1245                  * OK, we are remounting a valid rw partition rdonly, so set
1246                  * the rdonly flag and then mark the partition as valid again.
1247                  */
1248                 es->s_state = cpu_to_le16(sbi->s_mount_state);
1249                 es->s_mtime = cpu_to_le32(get_seconds());
1250                 spin_unlock(&sbi->s_lock);
1251                 ext2_sync_super(sb, es, 1);
1252         } else {
1253                 __le32 ret = EXT2_HAS_RO_COMPAT_FEATURE(sb,
1254                                                ~EXT2_FEATURE_RO_COMPAT_SUPP);
1255                 if (ret) {
1256                         ext2_msg(sb, KERN_WARNING,
1257                                 "warning: couldn't remount RDWR because of "
1258                                 "unsupported optional features (%x).",
1259                                 le32_to_cpu(ret));
1260                         err = -EROFS;
1261                         goto restore_opts;
1262                 }
1263                 /*
1264                  * Mounting a RDONLY partition read-write, so reread and
1265                  * store the current valid flag.  (It may have been changed
1266                  * by e2fsck since we originally mounted the partition.)
1267                  */
1268                 sbi->s_mount_state = le16_to_cpu(es->s_state);
1269                 if (!ext2_setup_super (sb, es, 0))
1270                         sb->s_flags &= ~MS_RDONLY;
1271                 spin_unlock(&sbi->s_lock);
1272                 ext2_write_super(sb);
1273         }
1274         return 0;
1275 restore_opts:
1276         sbi->s_mount_opt = old_opts.s_mount_opt;
1277         sbi->s_resuid = old_opts.s_resuid;
1278         sbi->s_resgid = old_opts.s_resgid;
1279         sb->s_flags = old_sb_flags;
1280         spin_unlock(&sbi->s_lock);
1281         return err;
1282 }
1283
1284 static int ext2_statfs (struct dentry * dentry, struct kstatfs * buf)
1285 {
1286         struct super_block *sb = dentry->d_sb;
1287         struct ext2_sb_info *sbi = EXT2_SB(sb);
1288         struct ext2_super_block *es = sbi->s_es;
1289         u64 fsid;
1290
1291         spin_lock(&sbi->s_lock);
1292
1293         if (test_opt (sb, MINIX_DF))
1294                 sbi->s_overhead_last = 0;
1295         else if (sbi->s_blocks_last != le32_to_cpu(es->s_blocks_count)) {
1296                 unsigned long i, overhead = 0;
1297                 smp_rmb();
1298
1299                 /*
1300                  * Compute the overhead (FS structures). This is constant
1301                  * for a given filesystem unless the number of block groups
1302                  * changes so we cache the previous value until it does.
1303                  */
1304
1305                 /*
1306                  * All of the blocks before first_data_block are
1307                  * overhead
1308                  */
1309                 overhead = le32_to_cpu(es->s_first_data_block);
1310
1311                 /*
1312                  * Add the overhead attributed to the superblock and
1313                  * block group descriptors.  If the sparse superblocks
1314                  * feature is turned on, then not all groups have this.
1315                  */
1316                 for (i = 0; i < sbi->s_groups_count; i++)
1317                         overhead += ext2_bg_has_super(sb, i) +
1318                                 ext2_bg_num_gdb(sb, i);
1319
1320                 /*
1321                  * Every block group has an inode bitmap, a block
1322                  * bitmap, and an inode table.
1323                  */
1324                 overhead += (sbi->s_groups_count *
1325                              (2 + sbi->s_itb_per_group));
1326                 sbi->s_overhead_last = overhead;
1327                 smp_wmb();
1328                 sbi->s_blocks_last = le32_to_cpu(es->s_blocks_count);
1329         }
1330
1331         buf->f_type = EXT2_SUPER_MAGIC;
1332         buf->f_bsize = sb->s_blocksize;
1333         buf->f_blocks = le32_to_cpu(es->s_blocks_count) - sbi->s_overhead_last;
1334         buf->f_bfree = ext2_count_free_blocks(sb);
1335         es->s_free_blocks_count = cpu_to_le32(buf->f_bfree);
1336         buf->f_bavail = buf->f_bfree - le32_to_cpu(es->s_r_blocks_count);
1337         if (buf->f_bfree < le32_to_cpu(es->s_r_blocks_count))
1338                 buf->f_bavail = 0;
1339         buf->f_files = le32_to_cpu(es->s_inodes_count);
1340         buf->f_ffree = ext2_count_free_inodes(sb);
1341         es->s_free_inodes_count = cpu_to_le32(buf->f_ffree);
1342         buf->f_namelen = EXT2_NAME_LEN;
1343         fsid = le64_to_cpup((void *)es->s_uuid) ^
1344                le64_to_cpup((void *)es->s_uuid + sizeof(u64));
1345         buf->f_fsid.val[0] = fsid & 0xFFFFFFFFUL;
1346         buf->f_fsid.val[1] = (fsid >> 32) & 0xFFFFFFFFUL;
1347         spin_unlock(&sbi->s_lock);
1348         return 0;
1349 }
1350
1351 static int ext2_get_sb(struct file_system_type *fs_type,
1352         int flags, const char *dev_name, void *data, struct vfsmount *mnt)
1353 {
1354         return get_sb_bdev(fs_type, flags, dev_name, data, ext2_fill_super, mnt);
1355 }
1356
1357 #ifdef CONFIG_QUOTA
1358
1359 /* Read data from quotafile - avoid pagecache and such because we cannot afford
1360  * acquiring the locks... As quota files are never truncated and quota code
1361  * itself serializes the operations (and noone else should touch the files)
1362  * we don't have to be afraid of races */
1363 static ssize_t ext2_quota_read(struct super_block *sb, int type, char *data,
1364                                size_t len, loff_t off)
1365 {
1366         struct inode *inode = sb_dqopt(sb)->files[type];
1367         sector_t blk = off >> EXT2_BLOCK_SIZE_BITS(sb);
1368         int err = 0;
1369         int offset = off & (sb->s_blocksize - 1);
1370         int tocopy;
1371         size_t toread;
1372         struct buffer_head tmp_bh;
1373         struct buffer_head *bh;
1374         loff_t i_size = i_size_read(inode);
1375
1376         if (off > i_size)
1377                 return 0;
1378         if (off+len > i_size)
1379                 len = i_size-off;
1380         toread = len;
1381         while (toread > 0) {
1382                 tocopy = sb->s_blocksize - offset < toread ?
1383                                 sb->s_blocksize - offset : toread;
1384
1385                 tmp_bh.b_state = 0;
1386                 tmp_bh.b_size = sb->s_blocksize;
1387                 err = ext2_get_block(inode, blk, &tmp_bh, 0);
1388                 if (err < 0)
1389                         return err;
1390                 if (!buffer_mapped(&tmp_bh))    /* A hole? */
1391                         memset(data, 0, tocopy);
1392                 else {
1393                         bh = sb_bread(sb, tmp_bh.b_blocknr);
1394                         if (!bh)
1395                                 return -EIO;
1396                         memcpy(data, bh->b_data+offset, tocopy);
1397                         brelse(bh);
1398                 }
1399                 offset = 0;
1400                 toread -= tocopy;
1401                 data += tocopy;
1402                 blk++;
1403         }
1404         return len;
1405 }
1406
1407 /* Write to quotafile */
1408 static ssize_t ext2_quota_write(struct super_block *sb, int type,
1409                                 const char *data, size_t len, loff_t off)
1410 {
1411         struct inode *inode = sb_dqopt(sb)->files[type];
1412         sector_t blk = off >> EXT2_BLOCK_SIZE_BITS(sb);
1413         int err = 0;
1414         int offset = off & (sb->s_blocksize - 1);
1415         int tocopy;
1416         size_t towrite = len;
1417         struct buffer_head tmp_bh;
1418         struct buffer_head *bh;
1419
1420         mutex_lock_nested(&inode->i_mutex, I_MUTEX_QUOTA);
1421         while (towrite > 0) {
1422                 tocopy = sb->s_blocksize - offset < towrite ?
1423                                 sb->s_blocksize - offset : towrite;
1424
1425                 tmp_bh.b_state = 0;
1426                 err = ext2_get_block(inode, blk, &tmp_bh, 1);
1427                 if (err < 0)
1428                         goto out;
1429                 if (offset || tocopy != EXT2_BLOCK_SIZE(sb))
1430                         bh = sb_bread(sb, tmp_bh.b_blocknr);
1431                 else
1432                         bh = sb_getblk(sb, tmp_bh.b_blocknr);
1433                 if (!bh) {
1434                         err = -EIO;
1435                         goto out;
1436                 }
1437                 lock_buffer(bh);
1438                 memcpy(bh->b_data+offset, data, tocopy);
1439                 flush_dcache_page(bh->b_page);
1440                 set_buffer_uptodate(bh);
1441                 mark_buffer_dirty(bh);
1442                 unlock_buffer(bh);
1443                 brelse(bh);
1444                 offset = 0;
1445                 towrite -= tocopy;
1446                 data += tocopy;
1447                 blk++;
1448         }
1449 out:
1450         if (len == towrite) {
1451                 mutex_unlock(&inode->i_mutex);
1452                 return err;
1453         }
1454         if (inode->i_size < off+len-towrite)
1455                 i_size_write(inode, off+len-towrite);
1456         inode->i_version++;
1457         inode->i_mtime = inode->i_ctime = CURRENT_TIME;
1458         mark_inode_dirty(inode);
1459         mutex_unlock(&inode->i_mutex);
1460         return len - towrite;
1461 }
1462
1463 #endif
1464
1465 static struct file_system_type ext2_fs_type = {
1466         .owner          = THIS_MODULE,
1467         .name           = "ext2",
1468         .get_sb         = ext2_get_sb,
1469         .kill_sb        = kill_block_super,
1470         .fs_flags       = FS_REQUIRES_DEV,
1471 };
1472
1473 static int __init init_ext2_fs(void)
1474 {
1475         int err = init_ext2_xattr();
1476         if (err)
1477                 return err;
1478         err = init_inodecache();
1479         if (err)
1480                 goto out1;
1481         err = register_filesystem(&ext2_fs_type);
1482         if (err)
1483                 goto out;
1484         return 0;
1485 out:
1486         destroy_inodecache();
1487 out1:
1488         exit_ext2_xattr();
1489         return err;
1490 }
1491
1492 static void __exit exit_ext2_fs(void)
1493 {
1494         unregister_filesystem(&ext2_fs_type);
1495         destroy_inodecache();
1496         exit_ext2_xattr();
1497 }
1498
1499 module_init(init_ext2_fs)
1500 module_exit(exit_ext2_fs)