git.oblomov.eu Git - 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/smp_lock.h>
  30 #include <linux/vfs.h>
  31 #include <linux/seq_file.h>
  32 #include <linux/mount.h>
  33 #include <linux/log2.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);
  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
  45 void ext2_error (struct super_block * sb, const char * function,
  46                  const char * fmt, ...)
  47 {
  48         va_list args;
  49         struct ext2_sb_info *sbi = EXT2_SB(sb);
  50         struct ext2_super_block *es = sbi->s_es;
  51
  52         if (!(sb->s_flags & MS_RDONLY)) {
  53                 sbi->s_mount_state |= EXT2_ERROR_FS;
  54                 es->s_state =
  55                         cpu_to_le16(le16_to_cpu(es->s_state) | EXT2_ERROR_FS);
  56                 ext2_sync_super(sb, es);
  57         }
  58
  59         va_start(args, fmt);
  60         printk(KERN_CRIT "EXT2-fs error (device %s): %s: ",sb->s_id, function);
  61         vprintk(fmt, args);
  62         printk("\n");
  63         va_end(args);
  64
  65         if (test_opt(sb, ERRORS_PANIC))
  66                 panic("EXT2-fs panic from previous error\n");
  67         if (test_opt(sb, ERRORS_RO)) {
  68                 printk("Remounting filesystem read-only\n");
  69                 sb->s_flags |= MS_RDONLY;
  70         }
  71 }
  72
  73 void ext2_warning (struct super_block * sb, const char * function,
  74                    const char * fmt, ...)
  75 {
  76         va_list args;
  77
  78         va_start(args, fmt);
  79         printk(KERN_WARNING "EXT2-fs warning (device %s): %s: ",
  80                sb->s_id, function);
  81         vprintk(fmt, args);
  82         printk("\n");
  83         va_end(args);
  84 }
  85
  86 void ext2_update_dynamic_rev(struct super_block *sb)
  87 {
  88         struct ext2_super_block *es = EXT2_SB(sb)->s_es;
  89
  90         if (le32_to_cpu(es->s_rev_level) > EXT2_GOOD_OLD_REV)
  91                 return;
  92
  93         ext2_warning(sb, __FUNCTION__,
  94                      "updating to rev %d because of new feature flag, "
  95                      "running e2fsck is recommended",
  96                      EXT2_DYNAMIC_REV);
  97
  98         es->s_first_ino = cpu_to_le32(EXT2_GOOD_OLD_FIRST_INO);
  99         es->s_inode_size = cpu_to_le16(EXT2_GOOD_OLD_INODE_SIZE);
 100         es->s_rev_level = cpu_to_le32(EXT2_DYNAMIC_REV);
 101         /* leave es->s_feature_*compat flags alone */
 102         /* es->s_uuid will be set by e2fsck if empty */
 103
 104         /*
 105          * The rest of the superblock fields should be zero, and if not it
 106          * means they are likely already in use, so leave them alone.  We
 107          * can leave it up to e2fsck to clean up any inconsistencies there.
 108          */
 109 }
 110
 111 static void ext2_put_super (struct super_block * sb)
 112 {
 113         int db_count;
 114         int i;
 115         struct ext2_sb_info *sbi = EXT2_SB(sb);
 116
 117         ext2_xattr_put_super(sb);
 118         if (!(sb->s_flags & MS_RDONLY)) {
 119                 struct ext2_super_block *es = sbi->s_es;
 120
 121                 es->s_state = cpu_to_le16(sbi->s_mount_state);
 122                 ext2_sync_super(sb, es);
 123         }
 124         db_count = sbi->s_gdb_count;
 125         for (i = 0; i < db_count; i++)
 126                 if (sbi->s_group_desc[i])
 127                         brelse (sbi->s_group_desc[i]);
 128         kfree(sbi->s_group_desc);
 129         kfree(sbi->s_debts);
 130         percpu_counter_destroy(&sbi->s_freeblocks_counter);
 131         percpu_counter_destroy(&sbi->s_freeinodes_counter);
 132         percpu_counter_destroy(&sbi->s_dirs_counter);
 133         brelse (sbi->s_sbh);
 134         sb->s_fs_info = NULL;
 135         kfree(sbi);
 136
 137         return;
 138 }
 139
 140 static struct kmem_cache * ext2_inode_cachep;
 141
 142 static struct inode *ext2_alloc_inode(struct super_block *sb)
 143 {
 144         struct ext2_inode_info *ei;
 145         ei = (struct ext2_inode_info *)kmem_cache_alloc(ext2_inode_cachep, GFP_KERNEL);
 146         if (!ei)
 147                 return NULL;
 148 #ifdef CONFIG_EXT2_FS_POSIX_ACL
 149         ei->i_acl = EXT2_ACL_NOT_CACHED;
 150         ei->i_default_acl = EXT2_ACL_NOT_CACHED;
 151 #endif
 152         ei->i_block_alloc_info = NULL;
 153         ei->vfs_inode.i_version = 1;
 154         return &ei->vfs_inode;
 155 }
 156
 157 static void ext2_destroy_inode(struct inode *inode)
 158 {
 159         kmem_cache_free(ext2_inode_cachep, EXT2_I(inode));
 160 }
 161
 162 static void init_once(struct kmem_cache * cachep, void *foo)
 163 {
 164         struct ext2_inode_info *ei = (struct ext2_inode_info *) foo;
 165
 166         rwlock_init(&ei->i_meta_lock);
 167 #ifdef CONFIG_EXT2_FS_XATTR
 168         init_rwsem(&ei->xattr_sem);
 169 #endif
 170         mutex_init(&ei->truncate_mutex);
 171         inode_init_once(&ei->vfs_inode);
 172 }
 173
 174 static int init_inodecache(void)
 175 {
 176         ext2_inode_cachep = kmem_cache_create("ext2_inode_cache",
 177                                              sizeof(struct ext2_inode_info),
 178                                              0, (SLAB_RECLAIM_ACCOUNT|
 179                                                 SLAB_MEM_SPREAD),
 180                                              init_once);
 181         if (ext2_inode_cachep == NULL)
 182                 return -ENOMEM;
 183         return 0;
 184 }
 185
 186 static void destroy_inodecache(void)
 187 {
 188         kmem_cache_destroy(ext2_inode_cachep);
 189 }
 190
 191 static void ext2_clear_inode(struct inode *inode)
 192 {
 193         struct ext2_block_alloc_info *rsv = EXT2_I(inode)->i_block_alloc_info;
 194 #ifdef CONFIG_EXT2_FS_POSIX_ACL
 195         struct ext2_inode_info *ei = EXT2_I(inode);
 196
 197         if (ei->i_acl && ei->i_acl != EXT2_ACL_NOT_CACHED) {
 198                 posix_acl_release(ei->i_acl);
 199                 ei->i_acl = EXT2_ACL_NOT_CACHED;
 200         }
 201         if (ei->i_default_acl && ei->i_default_acl != EXT2_ACL_NOT_CACHED) {
 202                 posix_acl_release(ei->i_default_acl);
 203                 ei->i_default_acl = EXT2_ACL_NOT_CACHED;
 204         }
 205 #endif
 206         ext2_discard_reservation(inode);
 207         EXT2_I(inode)->i_block_alloc_info = NULL;
 208         if (unlikely(rsv))
 209                 kfree(rsv);
 210 }
 211
 212 static int ext2_show_options(struct seq_file *seq, struct vfsmount *vfs)
 213 {
 214         struct super_block *sb = vfs->mnt_sb;
 215         struct ext2_sb_info *sbi = EXT2_SB(sb);
 216         struct ext2_super_block *es = sbi->s_es;
 217         unsigned long def_mount_opts;
 218
 219         def_mount_opts = le32_to_cpu(es->s_default_mount_opts);
 220
 221         if (sbi->s_sb_block != 1)
 222                 seq_printf(seq, ",sb=%lu", sbi->s_sb_block);
 223         if (test_opt(sb, MINIX_DF))
 224                 seq_puts(seq, ",minixdf");
 225         if (test_opt(sb, GRPID))
 226                 seq_puts(seq, ",grpid");
 227         if (!test_opt(sb, GRPID) && (def_mount_opts & EXT2_DEFM_BSDGROUPS))
 228                 seq_puts(seq, ",nogrpid");
 229         if (sbi->s_resuid != EXT2_DEF_RESUID ||
 230             le16_to_cpu(es->s_def_resuid) != EXT2_DEF_RESUID) {
 231                 seq_printf(seq, ",resuid=%u", sbi->s_resuid);
 232         }
 233         if (sbi->s_resgid != EXT2_DEF_RESGID ||
 234             le16_to_cpu(es->s_def_resgid) != EXT2_DEF_RESGID) {
 235                 seq_printf(seq, ",resgid=%u", sbi->s_resgid);
 236         }
 237         if (test_opt(sb, ERRORS_CONT)) {
 238                 int def_errors = le16_to_cpu(es->s_errors);
 239
 240                 if (def_errors == EXT2_ERRORS_PANIC ||
 241                     def_errors == EXT2_ERRORS_RO) {
 242                         seq_puts(seq, ",errors=continue");
 243                 }
 244         }
 245         if (test_opt(sb, ERRORS_RO))
 246                 seq_puts(seq, ",errors=remount-ro");
 247         if (test_opt(sb, ERRORS_PANIC))
 248                 seq_puts(seq, ",errors=panic");
 249         if (test_opt(sb, NO_UID32))
 250                 seq_puts(seq, ",nouid32");
 251         if (test_opt(sb, DEBUG))
 252                 seq_puts(seq, ",debug");
 253         if (test_opt(sb, OLDALLOC))
 254                 seq_puts(seq, ",oldalloc");
 255
 256 #ifdef CONFIG_EXT2_FS_XATTR
 257         if (test_opt(sb, XATTR_USER))
 258                 seq_puts(seq, ",user_xattr");
 259         if (!test_opt(sb, XATTR_USER) &&
 260             (def_mount_opts & EXT2_DEFM_XATTR_USER)) {
 261                 seq_puts(seq, ",nouser_xattr");
 262         }
 263 #endif
 264
 265 #ifdef CONFIG_EXT2_FS_POSIX_ACL
 266         if (test_opt(sb, POSIX_ACL))
 267                 seq_puts(seq, ",acl");
 268         if (!test_opt(sb, POSIX_ACL) && (def_mount_opts & EXT2_DEFM_ACL))
 269                 seq_puts(seq, ",noacl");
 270 #endif
 271
 272         if (test_opt(sb, NOBH))
 273                 seq_puts(seq, ",nobh");
 274
 275 #if defined(CONFIG_QUOTA)
 276         if (sbi->s_mount_opt & EXT2_MOUNT_USRQUOTA)
 277                 seq_puts(seq, ",usrquota");
 278
 279         if (sbi->s_mount_opt & EXT2_MOUNT_GRPQUOTA)
 280                 seq_puts(seq, ",grpquota");
 281 #endif
 282
 283 #if defined(CONFIG_EXT2_FS_XIP)
 284         if (sbi->s_mount_opt & EXT2_MOUNT_XIP)
 285                 seq_puts(seq, ",xip");
 286 #endif
 287
 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         .read_inode     = ext2_read_inode,
 300         .write_inode    = ext2_write_inode,
 301         .delete_inode   = ext2_delete_inode,
 302         .put_super      = ext2_put_super,
 303         .write_super    = ext2_write_super,
 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 = iget(sb, ino);
 330         if (inode == NULL)
 331                 return ERR_PTR(-ENOMEM);
 332         if (is_bad_inode(inode) ||
 333             (generation && inode->i_generation != generation)) {
 334                 /* we didn't find the right inode.. */
 335                 iput(inode);
 336                 return ERR_PTR(-ESTALE);
 337         }
 338         return inode;
 339 }
 340
 341 static struct dentry *ext2_fh_to_dentry(struct super_block *sb, struct fid *fid,
 342                 int fh_len, int fh_type)
 343 {
 344         return generic_fh_to_dentry(sb, fid, fh_len, fh_type,
 345                                     ext2_nfs_get_inode);
 346 }
 347
 348 static struct dentry *ext2_fh_to_parent(struct super_block *sb, struct fid *fid,
 349                 int fh_len, int fh_type)
 350 {
 351         return generic_fh_to_parent(sb, fid, fh_len, fh_type,
 352                                     ext2_nfs_get_inode);
 353 }
 354
 355 /* Yes, most of these are left as NULL!!
 356  * A NULL value implies the default, which works with ext2-like file
 357  * systems, but can be improved upon.
 358  * Currently only get_parent is required.
 359  */
 360 static const struct export_operations ext2_export_ops = {
 361         .fh_to_dentry = ext2_fh_to_dentry,
 362         .fh_to_parent = ext2_fh_to_parent,
 363         .get_parent = ext2_get_parent,
 364 };
 365
 366 static unsigned long get_sb_block(void **data)
 367 {
 368         unsigned long   sb_block;
 369         char            *options = (char *) *data;
 370
 371         if (!options || strncmp(options, "sb=", 3) != 0)
 372                 return 1;       /* Default location */
 373         options += 3;
 374         sb_block = simple_strtoul(options, &options, 0);
 375         if (*options && *options != ',') {
 376                 printk("EXT2-fs: Invalid sb specification: %s\n",
 377                        (char *) *data);
 378                 return 1;
 379         }
 380         if (*options == ',')
 381                 options++;
 382         *data = (void *) options;
 383         return sb_block;
 384 }
 385
 386 enum {
 387         Opt_bsd_df, Opt_minix_df, Opt_grpid, Opt_nogrpid,
 388         Opt_resgid, Opt_resuid, Opt_sb, Opt_err_cont, Opt_err_panic,
 389         Opt_err_ro, Opt_nouid32, Opt_nocheck, Opt_debug,
 390         Opt_oldalloc, Opt_orlov, Opt_nobh, Opt_user_xattr, Opt_nouser_xattr,
 391         Opt_acl, Opt_noacl, Opt_xip, Opt_ignore, Opt_err, Opt_quota,
 392         Opt_usrquota, Opt_grpquota, Opt_reservation, Opt_noreservation
 393 };
 394
 395 static match_table_t tokens = {
 396         {Opt_bsd_df, "bsddf"},
 397         {Opt_minix_df, "minixdf"},
 398         {Opt_grpid, "grpid"},
 399         {Opt_grpid, "bsdgroups"},
 400         {Opt_nogrpid, "nogrpid"},
 401         {Opt_nogrpid, "sysvgroups"},
 402         {Opt_resgid, "resgid=%u"},
 403         {Opt_resuid, "resuid=%u"},
 404         {Opt_sb, "sb=%u"},
 405         {Opt_err_cont, "errors=continue"},
 406         {Opt_err_panic, "errors=panic"},
 407         {Opt_err_ro, "errors=remount-ro"},
 408         {Opt_nouid32, "nouid32"},
 409         {Opt_nocheck, "check=none"},
 410         {Opt_nocheck, "nocheck"},
 411         {Opt_debug, "debug"},
 412         {Opt_oldalloc, "oldalloc"},
 413         {Opt_orlov, "orlov"},
 414         {Opt_nobh, "nobh"},
 415         {Opt_user_xattr, "user_xattr"},
 416         {Opt_nouser_xattr, "nouser_xattr"},
 417         {Opt_acl, "acl"},
 418         {Opt_noacl, "noacl"},
 419         {Opt_xip, "xip"},
 420         {Opt_grpquota, "grpquota"},
 421         {Opt_ignore, "noquota"},
 422         {Opt_quota, "quota"},
 423         {Opt_usrquota, "usrquota"},
 424         {Opt_reservation, "reservation"},
 425         {Opt_noreservation, "noreservation"},
 426         {Opt_err, NULL}
 427 };
 428
 429 static int parse_options (char * options,
 430                           struct ext2_sb_info *sbi)
 431 {
 432         char * p;
 433         substring_t args[MAX_OPT_ARGS];
 434         int option;
 435
 436         if (!options)
 437                 return 1;
 438
 439         while ((p = strsep (&options, ",")) != NULL) {
 440                 int token;
 441                 if (!*p)
 442                         continue;
 443
 444                 token = match_token(p, tokens, args);
 445                 switch (token) {
 446                 case Opt_bsd_df:
 447                         clear_opt (sbi->s_mount_opt, MINIX_DF);
 448                         break;
 449                 case Opt_minix_df:
 450                         set_opt (sbi->s_mount_opt, MINIX_DF);
 451                         break;
 452                 case Opt_grpid:
 453                         set_opt (sbi->s_mount_opt, GRPID);
 454                         break;
 455                 case Opt_nogrpid:
 456                         clear_opt (sbi->s_mount_opt, GRPID);
 457                         break;
 458                 case Opt_resuid:
 459                         if (match_int(&args[0], &option))
 460                                 return 0;
 461                         sbi->s_resuid = option;
 462                         break;
 463                 case Opt_resgid:
 464                         if (match_int(&args[0], &option))
 465                                 return 0;
 466                         sbi->s_resgid = option;
 467                         break;
 468                 case Opt_sb:
 469                         /* handled by get_sb_block() instead of here */
 470                         /* *sb_block = match_int(&args[0]); */
 471                         break;
 472                 case Opt_err_panic:
 473                         clear_opt (sbi->s_mount_opt, ERRORS_CONT);
 474                         clear_opt (sbi->s_mount_opt, ERRORS_RO);
 475                         set_opt (sbi->s_mount_opt, ERRORS_PANIC);
 476                         break;
 477                 case Opt_err_ro:
 478                         clear_opt (sbi->s_mount_opt, ERRORS_CONT);
 479                         clear_opt (sbi->s_mount_opt, ERRORS_PANIC);
 480                         set_opt (sbi->s_mount_opt, ERRORS_RO);
 481                         break;
 482                 case Opt_err_cont:
 483                         clear_opt (sbi->s_mount_opt, ERRORS_RO);
 484                         clear_opt (sbi->s_mount_opt, ERRORS_PANIC);
 485                         set_opt (sbi->s_mount_opt, ERRORS_CONT);
 486                         break;
 487                 case Opt_nouid32:
 488                         set_opt (sbi->s_mount_opt, NO_UID32);
 489                         break;
 490                 case Opt_nocheck:
 491                         clear_opt (sbi->s_mount_opt, CHECK);
 492                         break;
 493                 case Opt_debug:
 494                         set_opt (sbi->s_mount_opt, DEBUG);
 495                         break;
 496                 case Opt_oldalloc:
 497                         set_opt (sbi->s_mount_opt, OLDALLOC);
 498                         break;
 499                 case Opt_orlov:
 500                         clear_opt (sbi->s_mount_opt, OLDALLOC);
 501                         break;
 502                 case Opt_nobh:
 503                         set_opt (sbi->s_mount_opt, NOBH);
 504                         break;
 505 #ifdef CONFIG_EXT2_FS_XATTR
 506                 case Opt_user_xattr:
 507                         set_opt (sbi->s_mount_opt, XATTR_USER);
 508                         break;
 509                 case Opt_nouser_xattr:
 510                         clear_opt (sbi->s_mount_opt, XATTR_USER);
 511                         break;
 512 #else
 513                 case Opt_user_xattr:
 514                 case Opt_nouser_xattr:
 515                         printk("EXT2 (no)user_xattr options not supported\n");
 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                         printk("EXT2 (no)acl options not supported\n");
 529                         break;
 530 #endif
 531                 case Opt_xip:
 532 #ifdef CONFIG_EXT2_FS_XIP
 533                         set_opt (sbi->s_mount_opt, XIP);
 534 #else
 535                         printk("EXT2 xip option not supported\n");
 536 #endif
 537                         break;
 538
 539 #if defined(CONFIG_QUOTA)
 540                 case Opt_quota:
 541                 case Opt_usrquota:
 542                         set_opt(sbi->s_mount_opt, USRQUOTA);
 543                         break;
 544
 545                 case Opt_grpquota:
 546                         set_opt(sbi->s_mount_opt, GRPQUOTA);
 547                         break;
 548 #else
 549                 case Opt_quota:
 550                 case Opt_usrquota:
 551                 case Opt_grpquota:
 552                         printk(KERN_ERR
 553                                 "EXT2-fs: quota operations not supported.\n");
 554
 555                         break;
 556 #endif
 557
 558                 case Opt_reservation:
 559                         set_opt(sbi->s_mount_opt, RESERVATION);
 560                         printk("reservations ON\n");
 561                         break;
 562                 case Opt_noreservation:
 563                         clear_opt(sbi->s_mount_opt, RESERVATION);
 564                         printk("reservations OFF\n");
 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                 printk ("EXT2-fs warning: revision level too high, "
 584                         "forcing read-only mode\n");
 585                 res = MS_RDONLY;
 586         }
 587         if (read_only)
 588                 return res;
 589         if (!(sbi->s_mount_state & EXT2_VALID_FS))
 590                 printk ("EXT2-fs warning: mounting unchecked fs, "
 591                         "running e2fsck is recommended\n");
 592         else if ((sbi->s_mount_state & EXT2_ERROR_FS))
 593                 printk ("EXT2-fs warning: mounting fs with errors, "
 594                         "running e2fsck is recommended\n");
 595         else if ((__s16) le16_to_cpu(es->s_max_mnt_count) >= 0 &&
 596                  le16_to_cpu(es->s_mnt_count) >=
 597                  (unsigned short) (__s16) le16_to_cpu(es->s_max_mnt_count))
 598                 printk ("EXT2-fs warning: maximal mount count reached, "
 599                         "running e2fsck is recommended\n");
 600         else if (le32_to_cpu(es->s_checkinterval) &&
 601                 (le32_to_cpu(es->s_lastcheck) + le32_to_cpu(es->s_checkinterval) <= get_seconds()))
 602                 printk ("EXT2-fs warning: checktime reached, "
 603                         "running e2fsck is recommended\n");
 604         if (!le16_to_cpu(es->s_max_mnt_count))
 605                 es->s_max_mnt_count = cpu_to_le16(EXT2_DFL_MAX_MNT_COUNT);
 606         es->s_mnt_count=cpu_to_le16(le16_to_cpu(es->s_mnt_count) + 1);
 607         ext2_write_super(sb);
 608         if (test_opt (sb, DEBUG))
 609                 printk ("[EXT II FS %s, %s, bs=%lu, fs=%lu, gc=%lu, "
 610                         "bpg=%lu, ipg=%lu, mo=%04lx]\n",
 611                         EXT2FS_VERSION, EXT2FS_DATE, sb->s_blocksize,
 612                         sbi->s_frag_size,
 613                         sbi->s_groups_count,
 614                         EXT2_BLOCKS_PER_GROUP(sb),
 615                         EXT2_INODES_PER_GROUP(sb),
 616                         sbi->s_mount_opt);
 617         return res;
 618 }
 619
 620 static int ext2_check_descriptors (struct super_block * sb)
 621 {
 622         int i;
 623         int desc_block = 0;
 624         struct ext2_sb_info *sbi = EXT2_SB(sb);
 625         unsigned long first_block = le32_to_cpu(sbi->s_es->s_first_data_block);
 626         unsigned long last_block;
 627         struct ext2_group_desc * gdp = NULL;
 628
 629         ext2_debug ("Checking group descriptors");
 630
 631         for (i = 0; i < sbi->s_groups_count; i++)
 632         {
 633                 if (i == sbi->s_groups_count - 1)
 634                         last_block = le32_to_cpu(sbi->s_es->s_blocks_count) - 1;
 635                 else
 636                         last_block = first_block +
 637                                 (EXT2_BLOCKS_PER_GROUP(sb) - 1);
 638
 639                 if ((i % EXT2_DESC_PER_BLOCK(sb)) == 0)
 640                         gdp = (struct ext2_group_desc *) sbi->s_group_desc[desc_block++]->b_data;
 641                 if (le32_to_cpu(gdp->bg_block_bitmap) < first_block ||
 642                     le32_to_cpu(gdp->bg_block_bitmap) > last_block)
 643                 {
 644                         ext2_error (sb, "ext2_check_descriptors",
 645                                     "Block bitmap for group %d"
 646                                     " not in group (block %lu)!",
 647                                     i, (unsigned long) le32_to_cpu(gdp->bg_block_bitmap));
 648                         return 0;
 649                 }
 650                 if (le32_to_cpu(gdp->bg_inode_bitmap) < first_block ||
 651                     le32_to_cpu(gdp->bg_inode_bitmap) > last_block)
 652                 {
 653                         ext2_error (sb, "ext2_check_descriptors",
 654                                     "Inode bitmap for group %d"
 655                                     " not in group (block %lu)!",
 656                                     i, (unsigned long) le32_to_cpu(gdp->bg_inode_bitmap));
 657                         return 0;
 658                 }
 659                 if (le32_to_cpu(gdp->bg_inode_table) < first_block ||
 660                     le32_to_cpu(gdp->bg_inode_table) + sbi->s_itb_per_group - 1 >
 661                     last_block)
 662                 {
 663                         ext2_error (sb, "ext2_check_descriptors",
 664                                     "Inode table for group %d"
 665                                     " not in group (block %lu)!",
 666                                     i, (unsigned long) le32_to_cpu(gdp->bg_inode_table));
 667                         return 0;
 668                 }
 669                 first_block += EXT2_BLOCKS_PER_GROUP(sb);
 670                 gdp++;
 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         /* This constant is calculated to be the largest file size for a
 684          * dense, 4k-blocksize file such that the total number of
 685          * sectors in the file, including data and all indirect blocks,
 686          * does not exceed 2^32. */
 687         const loff_t upper_limit = 0x1ff7fffd000LL;
 688
 689         res += 1LL << (bits-2);
 690         res += 1LL << (2*(bits-2));
 691         res += 1LL << (3*(bits-2));
 692         res <<= bits;
 693         if (res > upper_limit)
 694                 res = upper_limit;
 695         return res;
 696 }
 697
 698 static unsigned long descriptor_loc(struct super_block *sb,
 699                                     unsigned long logic_sb_block,
 700                                     int nr)
 701 {
 702         struct ext2_sb_info *sbi = EXT2_SB(sb);
 703         unsigned long bg, first_data_block, first_meta_bg;
 704         int has_super = 0;
 705
 706         first_data_block = le32_to_cpu(sbi->s_es->s_first_data_block);
 707         first_meta_bg = le32_to_cpu(sbi->s_es->s_first_meta_bg);
 708
 709         if (!EXT2_HAS_INCOMPAT_FEATURE(sb, EXT2_FEATURE_INCOMPAT_META_BG) ||
 710             nr < first_meta_bg)
 711                 return (logic_sb_block + nr + 1);
 712         bg = sbi->s_desc_per_block * nr;
 713         if (ext2_bg_has_super(sb, bg))
 714                 has_super = 1;
 715         return (first_data_block + has_super + (bg * sbi->s_blocks_per_group));
 716 }
 717
 718 static int ext2_fill_super(struct super_block *sb, void *data, int silent)
 719 {
 720         struct buffer_head * bh;
 721         struct ext2_sb_info * sbi;
 722         struct ext2_super_block * es;
 723         struct inode *root;
 724         unsigned long block;
 725         unsigned long sb_block = get_sb_block(&data);
 726         unsigned long logic_sb_block;
 727         unsigned long offset = 0;
 728         unsigned long def_mount_opts;
 729         int blocksize = BLOCK_SIZE;
 730         int db_count;
 731         int i, j;
 732         __le32 features;
 733         int err;
 734
 735         sbi = kzalloc(sizeof(*sbi), GFP_KERNEL);
 736         if (!sbi)
 737                 return -ENOMEM;
 738         sb->s_fs_info = sbi;
 739         sbi->s_sb_block = sb_block;
 740
 741         /*
 742          * See what the current blocksize for the device is, and
 743          * use that as the blocksize.  Otherwise (or if the blocksize
 744          * is smaller than the default) use the default.
 745          * This is important for devices that have a hardware
 746          * sectorsize that is larger than the default.
 747          */
 748         blocksize = sb_min_blocksize(sb, BLOCK_SIZE);
 749         if (!blocksize) {
 750                 printk ("EXT2-fs: unable to set blocksize\n");
 751                 goto failed_sbi;
 752         }
 753
 754         /*
 755          * If the superblock doesn't start on a hardware sector boundary,
 756          * calculate the offset.
 757          */
 758         if (blocksize != BLOCK_SIZE) {
 759                 logic_sb_block = (sb_block*BLOCK_SIZE) / blocksize;
 760                 offset = (sb_block*BLOCK_SIZE) % blocksize;
 761         } else {
 762                 logic_sb_block = sb_block;
 763         }
 764
 765         if (!(bh = sb_bread(sb, logic_sb_block))) {
 766                 printk ("EXT2-fs: unable to read superblock\n");
 767                 goto failed_sbi;
 768         }
 769         /*
 770          * Note: s_es must be initialized as soon as possible because
 771          *       some ext2 macro-instructions depend on its value
 772          */
 773         es = (struct ext2_super_block *) (((char *)bh->b_data) + offset);
 774         sbi->s_es = es;
 775         sb->s_magic = le16_to_cpu(es->s_magic);
 776
 777         if (sb->s_magic != EXT2_SUPER_MAGIC)
 778                 goto cantfind_ext2;
 779
 780         /* Set defaults before we parse the mount options */
 781         def_mount_opts = le32_to_cpu(es->s_default_mount_opts);
 782         if (def_mount_opts & EXT2_DEFM_DEBUG)
 783                 set_opt(sbi->s_mount_opt, DEBUG);
 784         if (def_mount_opts & EXT2_DEFM_BSDGROUPS)
 785                 set_opt(sbi->s_mount_opt, GRPID);
 786         if (def_mount_opts & EXT2_DEFM_UID16)
 787                 set_opt(sbi->s_mount_opt, NO_UID32);
 788 #ifdef CONFIG_EXT2_FS_XATTR
 789         if (def_mount_opts & EXT2_DEFM_XATTR_USER)
 790                 set_opt(sbi->s_mount_opt, XATTR_USER);
 791 #endif
 792 #ifdef CONFIG_EXT2_FS_POSIX_ACL
 793         if (def_mount_opts & EXT2_DEFM_ACL)
 794                 set_opt(sbi->s_mount_opt, POSIX_ACL);
 795 #endif
 796
 797         if (le16_to_cpu(sbi->s_es->s_errors) == EXT2_ERRORS_PANIC)
 798                 set_opt(sbi->s_mount_opt, ERRORS_PANIC);
 799         else if (le16_to_cpu(sbi->s_es->s_errors) == EXT2_ERRORS_RO)
 800                 set_opt(sbi->s_mount_opt, ERRORS_RO);
 801         else
 802                 set_opt(sbi->s_mount_opt, ERRORS_CONT);
 803
 804         sbi->s_resuid = le16_to_cpu(es->s_def_resuid);
 805         sbi->s_resgid = le16_to_cpu(es->s_def_resgid);
 806
 807         set_opt(sbi->s_mount_opt, RESERVATION);
 808
 809         if (!parse_options ((char *) data, sbi))
 810                 goto failed_mount;
 811
 812         sb->s_flags = (sb->s_flags & ~MS_POSIXACL) |
 813                 ((EXT2_SB(sb)->s_mount_opt & EXT2_MOUNT_POSIX_ACL) ?
 814                  MS_POSIXACL : 0);
 815
 816         ext2_xip_verify_sb(sb); /* see if bdev supports xip, unset
 817                                     EXT2_MOUNT_XIP if not */
 818
 819         if (le32_to_cpu(es->s_rev_level) == EXT2_GOOD_OLD_REV &&
 820             (EXT2_HAS_COMPAT_FEATURE(sb, ~0U) ||
 821              EXT2_HAS_RO_COMPAT_FEATURE(sb, ~0U) ||
 822              EXT2_HAS_INCOMPAT_FEATURE(sb, ~0U)))
 823                 printk("EXT2-fs warning: feature flags set on rev 0 fs, "
 824                        "running e2fsck is recommended\n");
 825         /*
 826          * Check feature flags regardless of the revision level, since we
 827          * previously didn't change the revision level when setting the flags,
 828          * so there is a chance incompat flags are set on a rev 0 filesystem.
 829          */
 830         features = EXT2_HAS_INCOMPAT_FEATURE(sb, ~EXT2_FEATURE_INCOMPAT_SUPP);
 831         if (features) {
 832                 printk("EXT2-fs: %s: couldn't mount because of "
 833                        "unsupported optional features (%x).\n",
 834                        sb->s_id, le32_to_cpu(features));
 835                 goto failed_mount;
 836         }
 837         if (!(sb->s_flags & MS_RDONLY) &&
 838             (features = EXT2_HAS_RO_COMPAT_FEATURE(sb, ~EXT2_FEATURE_RO_COMPAT_SUPP))){
 839                 printk("EXT2-fs: %s: couldn't mount RDWR because of "
 840                        "unsupported optional features (%x).\n",
 841                        sb->s_id, le32_to_cpu(features));
 842                 goto failed_mount;
 843         }
 844
 845         blocksize = BLOCK_SIZE << le32_to_cpu(sbi->s_es->s_log_block_size);
 846
 847         if ((ext2_use_xip(sb)) && ((blocksize != PAGE_SIZE) ||
 848                                   (sb->s_blocksize != blocksize))) {
 849                 if (!silent)
 850                         printk("XIP: Unsupported blocksize\n");
 851                 goto failed_mount;
 852         }
 853
 854         /* If the blocksize doesn't match, re-read the thing.. */
 855         if (sb->s_blocksize != blocksize) {
 856                 brelse(bh);
 857
 858                 if (!sb_set_blocksize(sb, blocksize)) {
 859                         printk(KERN_ERR "EXT2-fs: blocksize too small for device.\n");
 860                         goto failed_sbi;
 861                 }
 862
 863                 logic_sb_block = (sb_block*BLOCK_SIZE) / blocksize;
 864                 offset = (sb_block*BLOCK_SIZE) % blocksize;
 865                 bh = sb_bread(sb, logic_sb_block);
 866                 if(!bh) {
 867                         printk("EXT2-fs: Couldn't read superblock on "
 868                                "2nd try.\n");
 869                         goto failed_sbi;
 870                 }
 871                 es = (struct ext2_super_block *) (((char *)bh->b_data) + offset);
 872                 sbi->s_es = es;
 873                 if (es->s_magic != cpu_to_le16(EXT2_SUPER_MAGIC)) {
 874                         printk ("EXT2-fs: Magic mismatch, very weird !\n");
 875                         goto failed_mount;
 876                 }
 877         }
 878
 879         sb->s_maxbytes = ext2_max_size(sb->s_blocksize_bits);
 880
 881         if (le32_to_cpu(es->s_rev_level) == EXT2_GOOD_OLD_REV) {
 882                 sbi->s_inode_size = EXT2_GOOD_OLD_INODE_SIZE;
 883                 sbi->s_first_ino = EXT2_GOOD_OLD_FIRST_INO;
 884         } else {
 885                 sbi->s_inode_size = le16_to_cpu(es->s_inode_size);
 886                 sbi->s_first_ino = le32_to_cpu(es->s_first_ino);
 887                 if ((sbi->s_inode_size < EXT2_GOOD_OLD_INODE_SIZE) ||
 888                     !is_power_of_2(sbi->s_inode_size) ||
 889                     (sbi->s_inode_size > blocksize)) {
 890                         printk ("EXT2-fs: unsupported inode size: %d\n",
 891                                 sbi->s_inode_size);
 892                         goto failed_mount;
 893                 }
 894         }
 895
 896         sbi->s_frag_size = EXT2_MIN_FRAG_SIZE <<
 897                                    le32_to_cpu(es->s_log_frag_size);
 898         if (sbi->s_frag_size == 0)
 899                 goto cantfind_ext2;
 900         sbi->s_frags_per_block = sb->s_blocksize / sbi->s_frag_size;
 901
 902         sbi->s_blocks_per_group = le32_to_cpu(es->s_blocks_per_group);
 903         sbi->s_frags_per_group = le32_to_cpu(es->s_frags_per_group);
 904         sbi->s_inodes_per_group = le32_to_cpu(es->s_inodes_per_group);
 905
 906         if (EXT2_INODE_SIZE(sb) == 0)
 907                 goto cantfind_ext2;
 908         sbi->s_inodes_per_block = sb->s_blocksize / EXT2_INODE_SIZE(sb);
 909         if (sbi->s_inodes_per_block == 0 || sbi->s_inodes_per_group == 0)
 910                 goto cantfind_ext2;
 911         sbi->s_itb_per_group = sbi->s_inodes_per_group /
 912                                         sbi->s_inodes_per_block;
 913         sbi->s_desc_per_block = sb->s_blocksize /
 914                                         sizeof (struct ext2_group_desc);
 915         sbi->s_sbh = bh;
 916         sbi->s_mount_state = le16_to_cpu(es->s_state);
 917         sbi->s_addr_per_block_bits =
 918                 ilog2 (EXT2_ADDR_PER_BLOCK(sb));
 919         sbi->s_desc_per_block_bits =
 920                 ilog2 (EXT2_DESC_PER_BLOCK(sb));
 921
 922         if (sb->s_magic != EXT2_SUPER_MAGIC)
 923                 goto cantfind_ext2;
 924
 925         if (sb->s_blocksize != bh->b_size) {
 926                 if (!silent)
 927                         printk ("VFS: Unsupported blocksize on dev "
 928                                 "%s.\n", sb->s_id);
 929                 goto failed_mount;
 930         }
 931
 932         if (sb->s_blocksize != sbi->s_frag_size) {
 933                 printk ("EXT2-fs: fragsize %lu != blocksize %lu (not supported yet)\n",
 934                         sbi->s_frag_size, sb->s_blocksize);
 935                 goto failed_mount;
 936         }
 937
 938         if (sbi->s_blocks_per_group > sb->s_blocksize * 8) {
 939                 printk ("EXT2-fs: #blocks per group too big: %lu\n",
 940                         sbi->s_blocks_per_group);
 941                 goto failed_mount;
 942         }
 943         if (sbi->s_frags_per_group > sb->s_blocksize * 8) {
 944                 printk ("EXT2-fs: #fragments per group too big: %lu\n",
 945                         sbi->s_frags_per_group);
 946                 goto failed_mount;
 947         }
 948         if (sbi->s_inodes_per_group > sb->s_blocksize * 8) {
 949                 printk ("EXT2-fs: #inodes per group too big: %lu\n",
 950                         sbi->s_inodes_per_group);
 951                 goto failed_mount;
 952         }
 953
 954         if (EXT2_BLOCKS_PER_GROUP(sb) == 0)
 955                 goto cantfind_ext2;
 956         sbi->s_groups_count = ((le32_to_cpu(es->s_blocks_count) -
 957                                 le32_to_cpu(es->s_first_data_block) - 1)
 958                                         / EXT2_BLOCKS_PER_GROUP(sb)) + 1;
 959         db_count = (sbi->s_groups_count + EXT2_DESC_PER_BLOCK(sb) - 1) /
 960                    EXT2_DESC_PER_BLOCK(sb);
 961         sbi->s_group_desc = kmalloc (db_count * sizeof (struct buffer_head *), GFP_KERNEL);
 962         if (sbi->s_group_desc == NULL) {
 963                 printk ("EXT2-fs: not enough memory\n");
 964                 goto failed_mount;
 965         }
 966         bgl_lock_init(&sbi->s_blockgroup_lock);
 967         sbi->s_debts = kcalloc(sbi->s_groups_count, sizeof(*sbi->s_debts), GFP_KERNEL);
 968         if (!sbi->s_debts) {
 969                 printk ("EXT2-fs: not enough memory\n");
 970                 goto failed_mount_group_desc;
 971         }
 972         for (i = 0; i < db_count; i++) {
 973                 block = descriptor_loc(sb, logic_sb_block, i);
 974                 sbi->s_group_desc[i] = sb_bread(sb, block);
 975                 if (!sbi->s_group_desc[i]) {
 976                         for (j = 0; j < i; j++)
 977                                 brelse (sbi->s_group_desc[j]);
 978                         printk ("EXT2-fs: unable to read group descriptors\n");
 979                         goto failed_mount_group_desc;
 980                 }
 981         }
 982         if (!ext2_check_descriptors (sb)) {
 983                 printk ("EXT2-fs: group descriptors corrupted!\n");
 984                 goto failed_mount2;
 985         }
 986         sbi->s_gdb_count = db_count;
 987         get_random_bytes(&sbi->s_next_generation, sizeof(u32));
 988         spin_lock_init(&sbi->s_next_gen_lock);
 989
 990         /* per fileystem reservation list head & lock */
 991         spin_lock_init(&sbi->s_rsv_window_lock);
 992         sbi->s_rsv_window_root = RB_ROOT;
 993         /*
 994          * Add a single, static dummy reservation to the start of the
 995          * reservation window list --- it gives us a placeholder for
 996          * append-at-start-of-list which makes the allocation logic
 997          * _much_ simpler.
 998          */
 999         sbi->s_rsv_window_head.rsv_start = EXT2_RESERVE_WINDOW_NOT_ALLOCATED;
1000         sbi->s_rsv_window_head.rsv_end = EXT2_RESERVE_WINDOW_NOT_ALLOCATED;
1001         sbi->s_rsv_window_head.rsv_alloc_hit = 0;
1002         sbi->s_rsv_window_head.rsv_goal_size = 0;
1003         ext2_rsv_window_add(sb, &sbi->s_rsv_window_head);
1004
1005         err = percpu_counter_init(&sbi->s_freeblocks_counter,
1006                                 ext2_count_free_blocks(sb));
1007         if (!err) {
1008                 err = percpu_counter_init(&sbi->s_freeinodes_counter,
1009                                 ext2_count_free_inodes(sb));
1010         }
1011         if (!err) {
1012                 err = percpu_counter_init(&sbi->s_dirs_counter,
1013                                 ext2_count_dirs(sb));
1014         }
1015         if (err) {
1016                 printk(KERN_ERR "EXT2-fs: insufficient memory\n");
1017                 goto failed_mount3;
1018         }
1019         /*
1020          * set up enough so that it can read an inode
1021          */
1022         sb->s_op = &ext2_sops;
1023         sb->s_export_op = &ext2_export_ops;
1024         sb->s_xattr = ext2_xattr_handlers;
1025         root = iget(sb, EXT2_ROOT_INO);
1026         sb->s_root = d_alloc_root(root);
1027         if (!sb->s_root) {
1028                 iput(root);
1029                 printk(KERN_ERR "EXT2-fs: get root inode failed\n");
1030                 goto failed_mount3;
1031         }
1032         if (!S_ISDIR(root->i_mode) || !root->i_blocks || !root->i_size) {
1033                 dput(sb->s_root);
1034                 sb->s_root = NULL;
1035                 printk(KERN_ERR "EXT2-fs: corrupt root inode, run e2fsck\n");
1036                 goto failed_mount3;
1037         }
1038         if (EXT2_HAS_COMPAT_FEATURE(sb, EXT3_FEATURE_COMPAT_HAS_JOURNAL))
1039                 ext2_warning(sb, __FUNCTION__,
1040                         "mounting ext3 filesystem as ext2");
1041         ext2_setup_super (sb, es, sb->s_flags & MS_RDONLY);
1042         return 0;
1043
1044 cantfind_ext2:
1045         if (!silent)
1046                 printk("VFS: Can't find an ext2 filesystem on dev %s.\n",
1047                        sb->s_id);
1048         goto failed_mount;
1049 failed_mount3:
1050         percpu_counter_destroy(&sbi->s_freeblocks_counter);
1051         percpu_counter_destroy(&sbi->s_freeinodes_counter);
1052         percpu_counter_destroy(&sbi->s_dirs_counter);
1053 failed_mount2:
1054         for (i = 0; i < db_count; i++)
1055                 brelse(sbi->s_group_desc[i]);
1056 failed_mount_group_desc:
1057         kfree(sbi->s_group_desc);
1058         kfree(sbi->s_debts);
1059 failed_mount:
1060         brelse(bh);
1061 failed_sbi:
1062         sb->s_fs_info = NULL;
1063         kfree(sbi);
1064         return -EINVAL;
1065 }
1066
1067 static void ext2_commit_super (struct super_block * sb,
1068                                struct ext2_super_block * es)
1069 {
1070         es->s_wtime = cpu_to_le32(get_seconds());
1071         mark_buffer_dirty(EXT2_SB(sb)->s_sbh);
1072         sb->s_dirt = 0;
1073 }
1074
1075 static void ext2_sync_super(struct super_block *sb, struct ext2_super_block *es)
1076 {
1077         es->s_free_blocks_count = cpu_to_le32(ext2_count_free_blocks(sb));
1078         es->s_free_inodes_count = cpu_to_le32(ext2_count_free_inodes(sb));
1079         es->s_wtime = cpu_to_le32(get_seconds());
1080         mark_buffer_dirty(EXT2_SB(sb)->s_sbh);
1081         sync_dirty_buffer(EXT2_SB(sb)->s_sbh);
1082         sb->s_dirt = 0;
1083 }
1084
1085 /*
1086  * In the second extended file system, it is not necessary to
1087  * write the super block since we use a mapping of the
1088  * disk super block in a buffer.
1089  *
1090  * However, this function is still used to set the fs valid
1091  * flags to 0.  We need to set this flag to 0 since the fs
1092  * may have been checked while mounted and e2fsck may have
1093  * set s_state to EXT2_VALID_FS after some corrections.
1094  */
1095
1096 void ext2_write_super (struct super_block * sb)
1097 {
1098         struct ext2_super_block * es;
1099         lock_kernel();
1100         if (!(sb->s_flags & MS_RDONLY)) {
1101                 es = EXT2_SB(sb)->s_es;
1102
1103                 if (le16_to_cpu(es->s_state) & EXT2_VALID_FS) {
1104                         ext2_debug ("setting valid to 0\n");
1105                         es->s_state = cpu_to_le16(le16_to_cpu(es->s_state) &
1106                                                   ~EXT2_VALID_FS);
1107                         es->s_free_blocks_count = cpu_to_le32(ext2_count_free_blocks(sb));
1108                         es->s_free_inodes_count = cpu_to_le32(ext2_count_free_inodes(sb));
1109                         es->s_mtime = cpu_to_le32(get_seconds());
1110                         ext2_sync_super(sb, es);
1111                 } else
1112                         ext2_commit_super (sb, es);
1113         }
1114         sb->s_dirt = 0;
1115         unlock_kernel();
1116 }
1117
1118 static int ext2_remount (struct super_block * sb, int * flags, char * data)
1119 {
1120         struct ext2_sb_info * sbi = EXT2_SB(sb);
1121         struct ext2_super_block * es;
1122         unsigned long old_mount_opt = sbi->s_mount_opt;
1123         struct ext2_mount_options old_opts;
1124         unsigned long old_sb_flags;
1125         int err;
1126
1127         /* Store the old options */
1128         old_sb_flags = sb->s_flags;
1129         old_opts.s_mount_opt = sbi->s_mount_opt;
1130         old_opts.s_resuid = sbi->s_resuid;
1131         old_opts.s_resgid = sbi->s_resgid;
1132
1133         /*
1134          * Allow the "check" option to be passed as a remount option.
1135          */
1136         if (!parse_options (data, sbi)) {
1137                 err = -EINVAL;
1138                 goto restore_opts;
1139         }
1140
1141         sb->s_flags = (sb->s_flags & ~MS_POSIXACL) |
1142                 ((sbi->s_mount_opt & EXT2_MOUNT_POSIX_ACL) ? MS_POSIXACL : 0);
1143
1144         ext2_xip_verify_sb(sb); /* see if bdev supports xip, unset
1145                                     EXT2_MOUNT_XIP if not */
1146
1147         if ((ext2_use_xip(sb)) && (sb->s_blocksize != PAGE_SIZE)) {
1148                 printk("XIP: Unsupported blocksize\n");
1149                 err = -EINVAL;
1150                 goto restore_opts;
1151         }
1152
1153         es = sbi->s_es;
1154         if (((sbi->s_mount_opt & EXT2_MOUNT_XIP) !=
1155             (old_mount_opt & EXT2_MOUNT_XIP)) &&
1156             invalidate_inodes(sb))
1157                 ext2_warning(sb, __FUNCTION__, "busy inodes while remounting "\
1158                              "xip remain in cache (no functional problem)");
1159         if ((*flags & MS_RDONLY) == (sb->s_flags & MS_RDONLY))
1160                 return 0;
1161         if (*flags & MS_RDONLY) {
1162                 if (le16_to_cpu(es->s_state) & EXT2_VALID_FS ||
1163                     !(sbi->s_mount_state & EXT2_VALID_FS))
1164                         return 0;
1165                 /*
1166                  * OK, we are remounting a valid rw partition rdonly, so set
1167                  * the rdonly flag and then mark the partition as valid again.
1168                  */
1169                 es->s_state = cpu_to_le16(sbi->s_mount_state);
1170                 es->s_mtime = cpu_to_le32(get_seconds());
1171         } else {
1172                 __le32 ret = EXT2_HAS_RO_COMPAT_FEATURE(sb,
1173                                                ~EXT2_FEATURE_RO_COMPAT_SUPP);
1174                 if (ret) {
1175                         printk("EXT2-fs: %s: couldn't remount RDWR because of "
1176                                "unsupported optional features (%x).\n",
1177                                sb->s_id, le32_to_cpu(ret));
1178                         err = -EROFS;
1179                         goto restore_opts;
1180                 }
1181                 /*
1182                  * Mounting a RDONLY partition read-write, so reread and
1183                  * store the current valid flag.  (It may have been changed
1184                  * by e2fsck since we originally mounted the partition.)
1185                  */
1186                 sbi->s_mount_state = le16_to_cpu(es->s_state);
1187                 if (!ext2_setup_super (sb, es, 0))
1188                         sb->s_flags &= ~MS_RDONLY;
1189         }
1190         ext2_sync_super(sb, es);
1191         return 0;
1192 restore_opts:
1193         sbi->s_mount_opt = old_opts.s_mount_opt;
1194         sbi->s_resuid = old_opts.s_resuid;
1195         sbi->s_resgid = old_opts.s_resgid;
1196         sb->s_flags = old_sb_flags;
1197         return err;
1198 }
1199
1200 static int ext2_statfs (struct dentry * dentry, struct kstatfs * buf)
1201 {
1202         struct super_block *sb = dentry->d_sb;
1203         struct ext2_sb_info *sbi = EXT2_SB(sb);
1204         struct ext2_super_block *es = sbi->s_es;
1205         u64 fsid;
1206
1207         if (test_opt (sb, MINIX_DF))
1208                 sbi->s_overhead_last = 0;
1209         else if (sbi->s_blocks_last != le32_to_cpu(es->s_blocks_count)) {
1210                 unsigned long i, overhead = 0;
1211                 smp_rmb();
1212
1213                 /*
1214                  * Compute the overhead (FS structures). This is constant
1215                  * for a given filesystem unless the number of block groups
1216                  * changes so we cache the previous value until it does.
1217                  */
1218
1219                 /*
1220                  * All of the blocks before first_data_block are
1221                  * overhead
1222                  */
1223                 overhead = le32_to_cpu(es->s_first_data_block);
1224
1225                 /*
1226                  * Add the overhead attributed to the superblock and
1227                  * block group descriptors.  If the sparse superblocks
1228                  * feature is turned on, then not all groups have this.
1229                  */
1230                 for (i = 0; i < sbi->s_groups_count; i++)
1231                         overhead += ext2_bg_has_super(sb, i) +
1232                                 ext2_bg_num_gdb(sb, i);
1233
1234                 /*
1235                  * Every block group has an inode bitmap, a block
1236                  * bitmap, and an inode table.
1237                  */
1238                 overhead += (sbi->s_groups_count *
1239                              (2 + sbi->s_itb_per_group));
1240                 sbi->s_overhead_last = overhead;
1241                 smp_wmb();
1242                 sbi->s_blocks_last = le32_to_cpu(es->s_blocks_count);
1243         }
1244
1245         buf->f_type = EXT2_SUPER_MAGIC;
1246         buf->f_bsize = sb->s_blocksize;
1247         buf->f_blocks = le32_to_cpu(es->s_blocks_count) - sbi->s_overhead_last;
1248         buf->f_bfree = ext2_count_free_blocks(sb);
1249         es->s_free_blocks_count = cpu_to_le32(buf->f_bfree);
1250         buf->f_bavail = buf->f_bfree - le32_to_cpu(es->s_r_blocks_count);
1251         if (buf->f_bfree < le32_to_cpu(es->s_r_blocks_count))
1252                 buf->f_bavail = 0;
1253         buf->f_files = le32_to_cpu(es->s_inodes_count);
1254         buf->f_ffree = ext2_count_free_inodes(sb);
1255         es->s_free_inodes_count = cpu_to_le32(buf->f_ffree);
1256         buf->f_namelen = EXT2_NAME_LEN;
1257         fsid = le64_to_cpup((void *)es->s_uuid) ^
1258                le64_to_cpup((void *)es->s_uuid + sizeof(u64));
1259         buf->f_fsid.val[0] = fsid & 0xFFFFFFFFUL;
1260         buf->f_fsid.val[1] = (fsid >> 32) & 0xFFFFFFFFUL;
1261         return 0;
1262 }
1263
1264 static int ext2_get_sb(struct file_system_type *fs_type,
1265         int flags, const char *dev_name, void *data, struct vfsmount *mnt)
1266 {
1267         return get_sb_bdev(fs_type, flags, dev_name, data, ext2_fill_super, mnt);
1268 }
1269
1270 #ifdef CONFIG_QUOTA
1271
1272 /* Read data from quotafile - avoid pagecache and such because we cannot afford
1273  * acquiring the locks... As quota files are never truncated and quota code
1274  * itself serializes the operations (and noone else should touch the files)
1275  * we don't have to be afraid of races */
1276 static ssize_t ext2_quota_read(struct super_block *sb, int type, char *data,
1277                                size_t len, loff_t off)
1278 {
1279         struct inode *inode = sb_dqopt(sb)->files[type];
1280         sector_t blk = off >> EXT2_BLOCK_SIZE_BITS(sb);
1281         int err = 0;
1282         int offset = off & (sb->s_blocksize - 1);
1283         int tocopy;
1284         size_t toread;
1285         struct buffer_head tmp_bh;
1286         struct buffer_head *bh;
1287         loff_t i_size = i_size_read(inode);
1288
1289         if (off > i_size)
1290                 return 0;
1291         if (off+len > i_size)
1292                 len = i_size-off;
1293         toread = len;
1294         while (toread > 0) {
1295                 tocopy = sb->s_blocksize - offset < toread ?
1296                                 sb->s_blocksize - offset : toread;
1297
1298                 tmp_bh.b_state = 0;
1299                 err = ext2_get_block(inode, blk, &tmp_bh, 0);
1300                 if (err < 0)
1301                         return err;
1302                 if (!buffer_mapped(&tmp_bh))    /* A hole? */
1303                         memset(data, 0, tocopy);
1304                 else {
1305                         bh = sb_bread(sb, tmp_bh.b_blocknr);
1306                         if (!bh)
1307                                 return -EIO;
1308                         memcpy(data, bh->b_data+offset, tocopy);
1309                         brelse(bh);
1310                 }
1311                 offset = 0;
1312                 toread -= tocopy;
1313                 data += tocopy;
1314                 blk++;
1315         }
1316         return len;
1317 }
1318
1319 /* Write to quotafile */
1320 static ssize_t ext2_quota_write(struct super_block *sb, int type,
1321                                 const char *data, size_t len, loff_t off)
1322 {
1323         struct inode *inode = sb_dqopt(sb)->files[type];
1324         sector_t blk = off >> EXT2_BLOCK_SIZE_BITS(sb);
1325         int err = 0;
1326         int offset = off & (sb->s_blocksize - 1);
1327         int tocopy;
1328         size_t towrite = len;
1329         struct buffer_head tmp_bh;
1330         struct buffer_head *bh;
1331
1332         mutex_lock_nested(&inode->i_mutex, I_MUTEX_QUOTA);
1333         while (towrite > 0) {
1334                 tocopy = sb->s_blocksize - offset < towrite ?
1335                                 sb->s_blocksize - offset : towrite;
1336
1337                 tmp_bh.b_state = 0;
1338                 err = ext2_get_block(inode, blk, &tmp_bh, 1);
1339                 if (err < 0)
1340                         goto out;
1341                 if (offset || tocopy != EXT2_BLOCK_SIZE(sb))
1342                         bh = sb_bread(sb, tmp_bh.b_blocknr);
1343                 else
1344                         bh = sb_getblk(sb, tmp_bh.b_blocknr);
1345                 if (!bh) {
1346                         err = -EIO;
1347                         goto out;
1348                 }
1349                 lock_buffer(bh);
1350                 memcpy(bh->b_data+offset, data, tocopy);
1351                 flush_dcache_page(bh->b_page);
1352                 set_buffer_uptodate(bh);
1353                 mark_buffer_dirty(bh);
1354                 unlock_buffer(bh);
1355                 brelse(bh);
1356                 offset = 0;
1357                 towrite -= tocopy;
1358                 data += tocopy;
1359                 blk++;
1360         }
1361 out:
1362         if (len == towrite)
1363                 return err;
1364         if (inode->i_size < off+len-towrite)
1365                 i_size_write(inode, off+len-towrite);
1366         inode->i_version++;
1367         inode->i_mtime = inode->i_ctime = CURRENT_TIME;
1368         mark_inode_dirty(inode);
1369         mutex_unlock(&inode->i_mutex);
1370         return len - towrite;
1371 }
1372
1373 #endif
1374
1375 static struct file_system_type ext2_fs_type = {
1376         .owner          = THIS_MODULE,
1377         .name           = "ext2",
1378         .get_sb         = ext2_get_sb,
1379         .kill_sb        = kill_block_super,
1380         .fs_flags       = FS_REQUIRES_DEV,
1381 };
1382
1383 static int __init init_ext2_fs(void)
1384 {
1385         int err = init_ext2_xattr();
1386         if (err)
1387                 return err;
1388         err = init_inodecache();
1389         if (err)
1390                 goto out1;
1391         err = register_filesystem(&ext2_fs_type);
1392         if (err)
1393                 goto out;
1394         return 0;
1395 out:
1396         destroy_inodecache();
1397 out1:
1398         exit_ext2_xattr();
1399         return err;
1400 }
1401
1402 static void __exit exit_ext2_fs(void)
1403 {
1404         unregister_filesystem(&ext2_fs_type);
1405         destroy_inodecache();
1406         exit_ext2_xattr();
1407 }
1408
1409 module_init(init_ext2_fs)
1410 module_exit(exit_ext2_fs)