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/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 <linux/quotaops.h>
  35 #include <asm/uaccess.h>
  36 #include "ext2.h"
  37 #include "xattr.h"
  38 #include "acl.h"
  39 #include "xip.h"
  40
  41 static void ext2_sync_super(struct super_block *sb,
  42                             struct ext2_super_block *es);
  43 static int ext2_remount (struct super_block * sb, int * flags, char * data);
  44 static int ext2_statfs (struct dentry * dentry, struct kstatfs * buf);
  45 static int ext2_sync_fs(struct super_block *sb, int wait);
  46
  47 void ext2_error (struct super_block * sb, const char * function,
  48                  const char * fmt, ...)
  49 {
  50         va_list args;
  51         struct ext2_sb_info *sbi = EXT2_SB(sb);
  52         struct ext2_super_block *es = sbi->s_es;
  53
  54         if (!(sb->s_flags & MS_RDONLY)) {
  55                 sbi->s_mount_state |= EXT2_ERROR_FS;
  56                 es->s_state |= cpu_to_le16(EXT2_ERROR_FS);
  57                 ext2_sync_super(sb, es);
  58         }
  59
  60         va_start(args, fmt);
  61         printk(KERN_CRIT "EXT2-fs error (device %s): %s: ",sb->s_id, function);
  62         vprintk(fmt, args);
  63         printk("\n");
  64         va_end(args);
  65
  66         if (test_opt(sb, ERRORS_PANIC))
  67                 panic("EXT2-fs panic from previous error\n");
  68         if (test_opt(sb, ERRORS_RO)) {
  69                 printk("Remounting filesystem read-only\n");
  70                 sb->s_flags |= MS_RDONLY;
  71         }
  72 }
  73
  74 void ext2_warning (struct super_block * sb, const char * function,
  75                    const char * fmt, ...)
  76 {
  77         va_list args;
  78
  79         va_start(args, fmt);
  80         printk(KERN_WARNING "EXT2-fs warning (device %s): %s: ",
  81                sb->s_id, function);
  82         vprintk(fmt, args);
  83         printk("\n");
  84         va_end(args);
  85 }
  86
  87 void ext2_update_dynamic_rev(struct super_block *sb)
  88 {
  89         struct ext2_super_block *es = EXT2_SB(sb)->s_es;
  90
  91         if (le32_to_cpu(es->s_rev_level) > EXT2_GOOD_OLD_REV)
  92                 return;
  93
  94         ext2_warning(sb, __func__,
  95                      "updating to rev %d because of new feature flag, "
  96                      "running e2fsck is recommended",
  97                      EXT2_DYNAMIC_REV);
  98
  99         es->s_first_ino = cpu_to_le32(EXT2_GOOD_OLD_FIRST_INO);
 100         es->s_inode_size = cpu_to_le16(EXT2_GOOD_OLD_INODE_SIZE);
 101         es->s_rev_level = cpu_to_le32(EXT2_DYNAMIC_REV);
 102         /* leave es->s_feature_*compat flags alone */
 103         /* es->s_uuid will be set by e2fsck if empty */
 104
 105         /*
 106          * The rest of the superblock fields should be zero, and if not it
 107          * means they are likely already in use, so leave them alone.  We
 108          * can leave it up to e2fsck to clean up any inconsistencies there.
 109          */
 110 }
 111
 112 static void ext2_put_super (struct super_block * sb)
 113 {
 114         int db_count;
 115         int i;
 116         struct ext2_sb_info *sbi = EXT2_SB(sb);
 117
 118         lock_kernel();
 119
 120         if (sb->s_dirt)
 121                 ext2_write_super(sb);
 122
 123         ext2_xattr_put_super(sb);
 124         if (!(sb->s_flags & MS_RDONLY)) {
 125                 struct ext2_super_block *es = sbi->s_es;
 126
 127                 es->s_state = cpu_to_le16(sbi->s_mount_state);
 128                 ext2_sync_super(sb, es);
 129         }
 130         db_count = sbi->s_gdb_count;
 131         for (i = 0; i < db_count; i++)
 132                 if (sbi->s_group_desc[i])
 133                         brelse (sbi->s_group_desc[i]);
 134         kfree(sbi->s_group_desc);
 135         kfree(sbi->s_debts);
 136         percpu_counter_destroy(&sbi->s_freeblocks_counter);
 137         percpu_counter_destroy(&sbi->s_freeinodes_counter);
 138         percpu_counter_destroy(&sbi->s_dirs_counter);
 139         brelse (sbi->s_sbh);
 140         sb->s_fs_info = NULL;
 141         kfree(sbi->s_blockgroup_lock);
 142         kfree(sbi);
 143
 144         unlock_kernel();
 145 }
 146
 147 static struct kmem_cache * ext2_inode_cachep;
 148
 149 static struct inode *ext2_alloc_inode(struct super_block *sb)
 150 {
 151         struct ext2_inode_info *ei;
 152         ei = (struct ext2_inode_info *)kmem_cache_alloc(ext2_inode_cachep, GFP_KERNEL);
 153         if (!ei)
 154                 return NULL;
 155         ei->i_block_alloc_info = NULL;
 156         ei->vfs_inode.i_version = 1;
 157         return &ei->vfs_inode;
 158 }
 159
 160 static void ext2_destroy_inode(struct inode *inode)
 161 {
 162         kmem_cache_free(ext2_inode_cachep, EXT2_I(inode));
 163 }
 164
 165 static void init_once(void *foo)
 166 {
 167         struct ext2_inode_info *ei = (struct ext2_inode_info *) foo;
 168
 169         rwlock_init(&ei->i_meta_lock);
 170 #ifdef CONFIG_EXT2_FS_XATTR
 171         init_rwsem(&ei->xattr_sem);
 172 #endif
 173         mutex_init(&ei->truncate_mutex);
 174         inode_init_once(&ei->vfs_inode);
 175 }
 176
 177 static int init_inodecache(void)
 178 {
 179         ext2_inode_cachep = kmem_cache_create("ext2_inode_cache",
 180                                              sizeof(struct ext2_inode_info),
 181                                              0, (SLAB_RECLAIM_ACCOUNT|
 182                                                 SLAB_MEM_SPREAD),
 183                                              init_once);
 184         if (ext2_inode_cachep == NULL)
 185                 return -ENOMEM;
 186         return 0;
 187 }
 188
 189 static void destroy_inodecache(void)
 190 {
 191         kmem_cache_destroy(ext2_inode_cachep);
 192 }
 193
 194 static void ext2_clear_inode(struct inode *inode)
 195 {
 196         struct ext2_block_alloc_info *rsv = EXT2_I(inode)->i_block_alloc_info;
 197         ext2_discard_reservation(inode);
 198         EXT2_I(inode)->i_block_alloc_info = NULL;
 199         if (unlikely(rsv))
 200                 kfree(rsv);
 201 }
 202
 203 static int ext2_show_options(struct seq_file *seq, struct vfsmount *vfs)
 204 {
 205         struct super_block *sb = vfs->mnt_sb;
 206         struct ext2_sb_info *sbi = EXT2_SB(sb);
 207         struct ext2_super_block *es = sbi->s_es;
 208         unsigned long def_mount_opts;
 209
 210         def_mount_opts = le32_to_cpu(es->s_default_mount_opts);
 211
 212         if (sbi->s_sb_block != 1)
 213                 seq_printf(seq, ",sb=%lu", sbi->s_sb_block);
 214         if (test_opt(sb, MINIX_DF))
 215                 seq_puts(seq, ",minixdf");
 216         if (test_opt(sb, GRPID))
 217                 seq_puts(seq, ",grpid");
 218         if (!test_opt(sb, GRPID) && (def_mount_opts & EXT2_DEFM_BSDGROUPS))
 219                 seq_puts(seq, ",nogrpid");
 220         if (sbi->s_resuid != EXT2_DEF_RESUID ||
 221             le16_to_cpu(es->s_def_resuid) != EXT2_DEF_RESUID) {
 222                 seq_printf(seq, ",resuid=%u", sbi->s_resuid);
 223         }
 224         if (sbi->s_resgid != EXT2_DEF_RESGID ||
 225             le16_to_cpu(es->s_def_resgid) != EXT2_DEF_RESGID) {
 226                 seq_printf(seq, ",resgid=%u", sbi->s_resgid);
 227         }
 228         if (test_opt(sb, ERRORS_RO)) {
 229                 int def_errors = le16_to_cpu(es->s_errors);
 230
 231                 if (def_errors == EXT2_ERRORS_PANIC ||
 232                     def_errors == EXT2_ERRORS_CONTINUE) {
 233                         seq_puts(seq, ",errors=remount-ro");
 234                 }
 235         }
 236         if (test_opt(sb, ERRORS_CONT))
 237                 seq_puts(seq, ",errors=continue");
 238         if (test_opt(sb, ERRORS_PANIC))
 239                 seq_puts(seq, ",errors=panic");
 240         if (test_opt(sb, NO_UID32))
 241                 seq_puts(seq, ",nouid32");
 242         if (test_opt(sb, DEBUG))
 243                 seq_puts(seq, ",debug");
 244         if (test_opt(sb, OLDALLOC))
 245                 seq_puts(seq, ",oldalloc");
 246
 247 #ifdef CONFIG_EXT2_FS_XATTR
 248         if (test_opt(sb, XATTR_USER))
 249                 seq_puts(seq, ",user_xattr");
 250         if (!test_opt(sb, XATTR_USER) &&
 251             (def_mount_opts & EXT2_DEFM_XATTR_USER)) {
 252                 seq_puts(seq, ",nouser_xattr");
 253         }
 254 #endif
 255
 256 #ifdef CONFIG_EXT2_FS_POSIX_ACL
 257         if (test_opt(sb, POSIX_ACL))
 258                 seq_puts(seq, ",acl");
 259         if (!test_opt(sb, POSIX_ACL) && (def_mount_opts & EXT2_DEFM_ACL))
 260                 seq_puts(seq, ",noacl");
 261 #endif
 262
 263         if (test_opt(sb, NOBH))
 264                 seq_puts(seq, ",nobh");
 265
 266 #if defined(CONFIG_QUOTA)
 267         if (sbi->s_mount_opt & EXT2_MOUNT_USRQUOTA)
 268                 seq_puts(seq, ",usrquota");
 269
 270         if (sbi->s_mount_opt & EXT2_MOUNT_GRPQUOTA)
 271                 seq_puts(seq, ",grpquota");
 272 #endif
 273
 274 #if defined(CONFIG_EXT2_FS_XIP)
 275         if (sbi->s_mount_opt & EXT2_MOUNT_XIP)
 276                 seq_puts(seq, ",xip");
 277 #endif
 278
 279         if (!test_opt(sb, RESERVATION))
 280                 seq_puts(seq, ",noreservation");
 281
 282         return 0;
 283 }
 284
 285 #ifdef CONFIG_QUOTA
 286 static ssize_t ext2_quota_read(struct super_block *sb, int type, char *data, size_t len, loff_t off);
 287 static ssize_t ext2_quota_write(struct super_block *sb, int type, const char *data, size_t len, loff_t off);
 288 #endif
 289
 290 static const struct super_operations ext2_sops = {
 291         .alloc_inode    = ext2_alloc_inode,
 292         .destroy_inode  = ext2_destroy_inode,
 293         .write_inode    = ext2_write_inode,
 294         .delete_inode   = ext2_delete_inode,
 295         .put_super      = ext2_put_super,
 296         .write_super    = ext2_write_super,
 297         .sync_fs        = ext2_sync_fs,
 298         .statfs         = ext2_statfs,
 299         .remount_fs     = ext2_remount,
 300         .clear_inode    = ext2_clear_inode,
 301         .show_options   = ext2_show_options,
 302 #ifdef CONFIG_QUOTA
 303         .quota_read     = ext2_quota_read,
 304         .quota_write    = ext2_quota_write,
 305 #endif
 306 };
 307
 308 static struct inode *ext2_nfs_get_inode(struct super_block *sb,
 309                 u64 ino, u32 generation)
 310 {
 311         struct inode *inode;
 312
 313         if (ino < EXT2_FIRST_INO(sb) && ino != EXT2_ROOT_INO)
 314                 return ERR_PTR(-ESTALE);
 315         if (ino > le32_to_cpu(EXT2_SB(sb)->s_es->s_inodes_count))
 316                 return ERR_PTR(-ESTALE);
 317
 318         /* iget isn't really right if the inode is currently unallocated!!
 319          * ext2_read_inode currently does appropriate checks, but
 320          * it might be "neater" to call ext2_get_inode first and check
 321          * if the inode is valid.....
 322          */
 323         inode = ext2_iget(sb, ino);
 324         if (IS_ERR(inode))
 325                 return ERR_CAST(inode);
 326         if (generation && inode->i_generation != generation) {
 327                 /* we didn't find the right inode.. */
 328                 iput(inode);
 329                 return ERR_PTR(-ESTALE);
 330         }
 331         return inode;
 332 }
 333
 334 static struct dentry *ext2_fh_to_dentry(struct super_block *sb, struct fid *fid,
 335                 int fh_len, int fh_type)
 336 {
 337         return generic_fh_to_dentry(sb, fid, fh_len, fh_type,
 338                                     ext2_nfs_get_inode);
 339 }
 340
 341 static struct dentry *ext2_fh_to_parent(struct super_block *sb, struct fid *fid,
 342                 int fh_len, int fh_type)
 343 {
 344         return generic_fh_to_parent(sb, fid, fh_len, fh_type,
 345                                     ext2_nfs_get_inode);
 346 }
 347
 348 /* Yes, most of these are left as NULL!!
 349  * A NULL value implies the default, which works with ext2-like file
 350  * systems, but can be improved upon.
 351  * Currently only get_parent is required.
 352  */
 353 static const struct export_operations ext2_export_ops = {
 354         .fh_to_dentry = ext2_fh_to_dentry,
 355         .fh_to_parent = ext2_fh_to_parent,
 356         .get_parent = ext2_get_parent,
 357 };
 358
 359 static unsigned long get_sb_block(void **data)
 360 {
 361         unsigned long   sb_block;
 362         char            *options = (char *) *data;
 363
 364         if (!options || strncmp(options, "sb=", 3) != 0)
 365                 return 1;       /* Default location */
 366         options += 3;
 367         sb_block = simple_strtoul(options, &options, 0);
 368         if (*options && *options != ',') {
 369                 printk("EXT2-fs: Invalid sb specification: %s\n",
 370                        (char *) *data);
 371                 return 1;
 372         }
 373         if (*options == ',')
 374                 options++;
 375         *data = (void *) options;
 376         return sb_block;
 377 }
 378
 379 enum {
 380         Opt_bsd_df, Opt_minix_df, Opt_grpid, Opt_nogrpid,
 381         Opt_resgid, Opt_resuid, Opt_sb, Opt_err_cont, Opt_err_panic,
 382         Opt_err_ro, Opt_nouid32, Opt_nocheck, Opt_debug,
 383         Opt_oldalloc, Opt_orlov, Opt_nobh, Opt_user_xattr, Opt_nouser_xattr,
 384         Opt_acl, Opt_noacl, Opt_xip, Opt_ignore, Opt_err, Opt_quota,
 385         Opt_usrquota, Opt_grpquota, Opt_reservation, Opt_noreservation
 386 };
 387
 388 static const match_table_t tokens = {
 389         {Opt_bsd_df, "bsddf"},
 390         {Opt_minix_df, "minixdf"},
 391         {Opt_grpid, "grpid"},
 392         {Opt_grpid, "bsdgroups"},
 393         {Opt_nogrpid, "nogrpid"},
 394         {Opt_nogrpid, "sysvgroups"},
 395         {Opt_resgid, "resgid=%u"},
 396         {Opt_resuid, "resuid=%u"},
 397         {Opt_sb, "sb=%u"},
 398         {Opt_err_cont, "errors=continue"},
 399         {Opt_err_panic, "errors=panic"},
 400         {Opt_err_ro, "errors=remount-ro"},
 401         {Opt_nouid32, "nouid32"},
 402         {Opt_nocheck, "check=none"},
 403         {Opt_nocheck, "nocheck"},
 404         {Opt_debug, "debug"},
 405         {Opt_oldalloc, "oldalloc"},
 406         {Opt_orlov, "orlov"},
 407         {Opt_nobh, "nobh"},
 408         {Opt_user_xattr, "user_xattr"},
 409         {Opt_nouser_xattr, "nouser_xattr"},
 410         {Opt_acl, "acl"},
 411         {Opt_noacl, "noacl"},
 412         {Opt_xip, "xip"},
 413         {Opt_grpquota, "grpquota"},
 414         {Opt_ignore, "noquota"},
 415         {Opt_quota, "quota"},
 416         {Opt_usrquota, "usrquota"},
 417         {Opt_reservation, "reservation"},
 418         {Opt_noreservation, "noreservation"},
 419         {Opt_err, NULL}
 420 };
 421
 422 static int parse_options (char * options,
 423                           struct ext2_sb_info *sbi)
 424 {
 425         char * p;
 426         substring_t args[MAX_OPT_ARGS];
 427         int option;
 428
 429         if (!options)
 430                 return 1;
 431
 432         while ((p = strsep (&options, ",")) != NULL) {
 433                 int token;
 434                 if (!*p)
 435                         continue;
 436
 437                 token = match_token(p, tokens, args);
 438                 switch (token) {
 439                 case Opt_bsd_df:
 440                         clear_opt (sbi->s_mount_opt, MINIX_DF);
 441                         break;
 442                 case Opt_minix_df:
 443                         set_opt (sbi->s_mount_opt, MINIX_DF);
 444                         break;
 445                 case Opt_grpid:
 446                         set_opt (sbi->s_mount_opt, GRPID);
 447                         break;
 448                 case Opt_nogrpid:
 449                         clear_opt (sbi->s_mount_opt, GRPID);
 450                         break;
 451                 case Opt_resuid:
 452                         if (match_int(&args[0], &option))
 453                                 return 0;
 454                         sbi->s_resuid = option;
 455                         break;
 456                 case Opt_resgid:
 457                         if (match_int(&args[0], &option))
 458                                 return 0;
 459                         sbi->s_resgid = option;
 460                         break;
 461                 case Opt_sb:
 462                         /* handled by get_sb_block() instead of here */
 463                         /* *sb_block = match_int(&args[0]); */
 464                         break;
 465                 case Opt_err_panic:
 466                         clear_opt (sbi->s_mount_opt, ERRORS_CONT);
 467                         clear_opt (sbi->s_mount_opt, ERRORS_RO);
 468                         set_opt (sbi->s_mount_opt, ERRORS_PANIC);
 469                         break;
 470                 case Opt_err_ro:
 471                         clear_opt (sbi->s_mount_opt, ERRORS_CONT);
 472                         clear_opt (sbi->s_mount_opt, ERRORS_PANIC);
 473                         set_opt (sbi->s_mount_opt, ERRORS_RO);
 474                         break;
 475                 case Opt_err_cont:
 476                         clear_opt (sbi->s_mount_opt, ERRORS_RO);
 477                         clear_opt (sbi->s_mount_opt, ERRORS_PANIC);
 478                         set_opt (sbi->s_mount_opt, ERRORS_CONT);
 479                         break;
 480                 case Opt_nouid32:
 481                         set_opt (sbi->s_mount_opt, NO_UID32);
 482                         break;
 483                 case Opt_nocheck:
 484                         clear_opt (sbi->s_mount_opt, CHECK);
 485                         break;
 486                 case Opt_debug:
 487                         set_opt (sbi->s_mount_opt, DEBUG);
 488                         break;
 489                 case Opt_oldalloc:
 490                         set_opt (sbi->s_mount_opt, OLDALLOC);
 491                         break;
 492                 case Opt_orlov:
 493                         clear_opt (sbi->s_mount_opt, OLDALLOC);
 494                         break;
 495                 case Opt_nobh:
 496                         set_opt (sbi->s_mount_opt, NOBH);
 497                         break;
 498 #ifdef CONFIG_EXT2_FS_XATTR
 499                 case Opt_user_xattr:
 500                         set_opt (sbi->s_mount_opt, XATTR_USER);
 501                         break;
 502                 case Opt_nouser_xattr:
 503                         clear_opt (sbi->s_mount_opt, XATTR_USER);
 504                         break;
 505 #else
 506                 case Opt_user_xattr:
 507                 case Opt_nouser_xattr:
 508                         printk("EXT2 (no)user_xattr options not supported\n");
 509                         break;
 510 #endif
 511 #ifdef CONFIG_EXT2_FS_POSIX_ACL
 512                 case Opt_acl:
 513                         set_opt(sbi->s_mount_opt, POSIX_ACL);
 514                         break;
 515                 case Opt_noacl:
 516                         clear_opt(sbi->s_mount_opt, POSIX_ACL);
 517                         break;
 518 #else
 519                 case Opt_acl:
 520                 case Opt_noacl:
 521                         printk("EXT2 (no)acl options not supported\n");
 522                         break;
 523 #endif
 524                 case Opt_xip:
 525 #ifdef CONFIG_EXT2_FS_XIP
 526                         set_opt (sbi->s_mount_opt, XIP);
 527 #else
 528                         printk("EXT2 xip option not supported\n");
 529 #endif
 530                         break;
 531
 532 #if defined(CONFIG_QUOTA)
 533                 case Opt_quota:
 534                 case Opt_usrquota:
 535                         set_opt(sbi->s_mount_opt, USRQUOTA);
 536                         break;
 537
 538                 case Opt_grpquota:
 539                         set_opt(sbi->s_mount_opt, GRPQUOTA);
 540                         break;
 541 #else
 542                 case Opt_quota:
 543                 case Opt_usrquota:
 544                 case Opt_grpquota:
 545                         printk(KERN_ERR
 546                                 "EXT2-fs: quota operations not supported.\n");
 547
 548                         break;
 549 #endif
 550
 551                 case Opt_reservation:
 552                         set_opt(sbi->s_mount_opt, RESERVATION);
 553                         printk("reservations ON\n");
 554                         break;
 555                 case Opt_noreservation:
 556                         clear_opt(sbi->s_mount_opt, RESERVATION);
 557                         printk("reservations OFF\n");
 558                         break;
 559                 case Opt_ignore:
 560                         break;
 561                 default:
 562                         return 0;
 563                 }
 564         }
 565         return 1;
 566 }
 567
 568 static int ext2_setup_super (struct super_block * sb,
 569                               struct ext2_super_block * es,
 570                               int read_only)
 571 {
 572         int res = 0;
 573         struct ext2_sb_info *sbi = EXT2_SB(sb);
 574
 575         if (le32_to_cpu(es->s_rev_level) > EXT2_MAX_SUPP_REV) {
 576                 printk ("EXT2-fs warning: revision level too high, "
 577                         "forcing read-only mode\n");
 578                 res = MS_RDONLY;
 579         }
 580         if (read_only)
 581                 return res;
 582         if (!(sbi->s_mount_state & EXT2_VALID_FS))
 583                 printk ("EXT2-fs warning: mounting unchecked fs, "
 584                         "running e2fsck is recommended\n");
 585         else if ((sbi->s_mount_state & EXT2_ERROR_FS))
 586                 printk ("EXT2-fs warning: mounting fs with errors, "
 587                         "running e2fsck is recommended\n");
 588         else if ((__s16) le16_to_cpu(es->s_max_mnt_count) >= 0 &&
 589                  le16_to_cpu(es->s_mnt_count) >=
 590                  (unsigned short) (__s16) le16_to_cpu(es->s_max_mnt_count))
 591                 printk ("EXT2-fs warning: maximal mount count reached, "
 592                         "running e2fsck is recommended\n");
 593         else if (le32_to_cpu(es->s_checkinterval) &&
 594                 (le32_to_cpu(es->s_lastcheck) + le32_to_cpu(es->s_checkinterval) <= get_seconds()))
 595                 printk ("EXT2-fs warning: checktime reached, "
 596                         "running e2fsck is recommended\n");
 597         if (!le16_to_cpu(es->s_max_mnt_count))
 598                 es->s_max_mnt_count = cpu_to_le16(EXT2_DFL_MAX_MNT_COUNT);
 599         le16_add_cpu(&es->s_mnt_count, 1);
 600         ext2_write_super(sb);
 601         if (test_opt (sb, DEBUG))
 602                 printk ("[EXT II FS %s, %s, bs=%lu, fs=%lu, gc=%lu, "
 603                         "bpg=%lu, ipg=%lu, mo=%04lx]\n",
 604                         EXT2FS_VERSION, EXT2FS_DATE, sb->s_blocksize,
 605                         sbi->s_frag_size,
 606                         sbi->s_groups_count,
 607                         EXT2_BLOCKS_PER_GROUP(sb),
 608                         EXT2_INODES_PER_GROUP(sb),
 609                         sbi->s_mount_opt);
 610         return res;
 611 }
 612
 613 static int ext2_check_descriptors(struct super_block *sb)
 614 {
 615         int i;
 616         struct ext2_sb_info *sbi = EXT2_SB(sb);
 617
 618         ext2_debug ("Checking group descriptors");
 619
 620         for (i = 0; i < sbi->s_groups_count; i++) {
 621                 struct ext2_group_desc *gdp = ext2_get_group_desc(sb, i, NULL);
 622                 ext2_fsblk_t first_block = ext2_group_first_block_no(sb, i);
 623                 ext2_fsblk_t last_block;
 624
 625                 if (i == sbi->s_groups_count - 1)
 626                         last_block = le32_to_cpu(sbi->s_es->s_blocks_count) - 1;
 627                 else
 628                         last_block = first_block +
 629                                 (EXT2_BLOCKS_PER_GROUP(sb) - 1);
 630
 631                 if (le32_to_cpu(gdp->bg_block_bitmap) < first_block ||
 632                     le32_to_cpu(gdp->bg_block_bitmap) > last_block)
 633                 {
 634                         ext2_error (sb, "ext2_check_descriptors",
 635                                     "Block bitmap for group %d"
 636                                     " not in group (block %lu)!",
 637                                     i, (unsigned long) le32_to_cpu(gdp->bg_block_bitmap));
 638                         return 0;
 639                 }
 640                 if (le32_to_cpu(gdp->bg_inode_bitmap) < first_block ||
 641                     le32_to_cpu(gdp->bg_inode_bitmap) > last_block)
 642                 {
 643                         ext2_error (sb, "ext2_check_descriptors",
 644                                     "Inode bitmap for group %d"
 645                                     " not in group (block %lu)!",
 646                                     i, (unsigned long) le32_to_cpu(gdp->bg_inode_bitmap));
 647                         return 0;
 648                 }
 649                 if (le32_to_cpu(gdp->bg_inode_table) < first_block ||
 650                     le32_to_cpu(gdp->bg_inode_table) + sbi->s_itb_per_group - 1 >
 651                     last_block)
 652                 {
 653                         ext2_error (sb, "ext2_check_descriptors",
 654                                     "Inode table for group %d"
 655                                     " not in group (block %lu)!",
 656                                     i, (unsigned long) le32_to_cpu(gdp->bg_inode_table));
 657                         return 0;
 658                 }
 659         }
 660         return 1;
 661 }
 662
 663 /*
 664  * Maximal file size.  There is a direct, and {,double-,triple-}indirect
 665  * block limit, and also a limit of (2^32 - 1) 512-byte sectors in i_blocks.
 666  * We need to be 1 filesystem block less than the 2^32 sector limit.
 667  */
 668 static loff_t ext2_max_size(int bits)
 669 {
 670         loff_t res = EXT2_NDIR_BLOCKS;
 671         int meta_blocks;
 672         loff_t upper_limit;
 673
 674         /* This is calculated to be the largest file size for a
 675          * dense, file such that the total number of
 676          * sectors in the file, including data and all indirect blocks,
 677          * does not exceed 2^32 -1
 678          * __u32 i_blocks representing the total number of
 679          * 512 bytes blocks of the file
 680          */
 681         upper_limit = (1LL << 32) - 1;
 682
 683         /* total blocks in file system block size */
 684         upper_limit >>= (bits - 9);
 685
 686
 687         /* indirect blocks */
 688         meta_blocks = 1;
 689         /* double indirect blocks */
 690         meta_blocks += 1 + (1LL << (bits-2));
 691         /* tripple indirect blocks */
 692         meta_blocks += 1 + (1LL << (bits-2)) + (1LL << (2*(bits-2)));
 693
 694         upper_limit -= meta_blocks;
 695         upper_limit <<= bits;
 696
 697         res += 1LL << (bits-2);
 698         res += 1LL << (2*(bits-2));
 699         res += 1LL << (3*(bits-2));
 700         res <<= bits;
 701         if (res > upper_limit)
 702                 res = upper_limit;
 703
 704         if (res > MAX_LFS_FILESIZE)
 705                 res = MAX_LFS_FILESIZE;
 706
 707         return res;
 708 }
 709
 710 static unsigned long descriptor_loc(struct super_block *sb,
 711                                     unsigned long logic_sb_block,
 712                                     int nr)
 713 {
 714         struct ext2_sb_info *sbi = EXT2_SB(sb);
 715         unsigned long bg, first_meta_bg;
 716         int has_super = 0;
 717
 718         first_meta_bg = le32_to_cpu(sbi->s_es->s_first_meta_bg);
 719
 720         if (!EXT2_HAS_INCOMPAT_FEATURE(sb, EXT2_FEATURE_INCOMPAT_META_BG) ||
 721             nr < first_meta_bg)
 722                 return (logic_sb_block + nr + 1);
 723         bg = sbi->s_desc_per_block * nr;
 724         if (ext2_bg_has_super(sb, bg))
 725                 has_super = 1;
 726
 727         return ext2_group_first_block_no(sb, bg) + has_super;
 728 }
 729
 730 static int ext2_fill_super(struct super_block *sb, void *data, int silent)
 731 {
 732         struct buffer_head * bh;
 733         struct ext2_sb_info * sbi;
 734         struct ext2_super_block * es;
 735         struct inode *root;
 736         unsigned long block;
 737         unsigned long sb_block = get_sb_block(&data);
 738         unsigned long logic_sb_block;
 739         unsigned long offset = 0;
 740         unsigned long def_mount_opts;
 741         long ret = -EINVAL;
 742         int blocksize = BLOCK_SIZE;
 743         int db_count;
 744         int i, j;
 745         __le32 features;
 746         int err;
 747
 748         sbi = kzalloc(sizeof(*sbi), GFP_KERNEL);
 749         if (!sbi)
 750                 return -ENOMEM;
 751
 752         sbi->s_blockgroup_lock =
 753                 kzalloc(sizeof(struct blockgroup_lock), GFP_KERNEL);
 754         if (!sbi->s_blockgroup_lock) {
 755                 kfree(sbi);
 756                 return -ENOMEM;
 757         }
 758         sb->s_fs_info = sbi;
 759         sbi->s_sb_block = sb_block;
 760
 761         /*
 762          * See what the current blocksize for the device is, and
 763          * use that as the blocksize.  Otherwise (or if the blocksize
 764          * is smaller than the default) use the default.
 765          * This is important for devices that have a hardware
 766          * sectorsize that is larger than the default.
 767          */
 768         blocksize = sb_min_blocksize(sb, BLOCK_SIZE);
 769         if (!blocksize) {
 770                 printk ("EXT2-fs: unable to set blocksize\n");
 771                 goto failed_sbi;
 772         }
 773
 774         /*
 775          * If the superblock doesn't start on a hardware sector boundary,
 776          * calculate the offset.
 777          */
 778         if (blocksize != BLOCK_SIZE) {
 779                 logic_sb_block = (sb_block*BLOCK_SIZE) / blocksize;
 780                 offset = (sb_block*BLOCK_SIZE) % blocksize;
 781         } else {
 782                 logic_sb_block = sb_block;
 783         }
 784
 785         if (!(bh = sb_bread(sb, logic_sb_block))) {
 786                 printk ("EXT2-fs: unable to read superblock\n");
 787                 goto failed_sbi;
 788         }
 789         /*
 790          * Note: s_es must be initialized as soon as possible because
 791          *       some ext2 macro-instructions depend on its value
 792          */
 793         es = (struct ext2_super_block *) (((char *)bh->b_data) + offset);
 794         sbi->s_es = es;
 795         sb->s_magic = le16_to_cpu(es->s_magic);
 796
 797         if (sb->s_magic != EXT2_SUPER_MAGIC)
 798                 goto cantfind_ext2;
 799
 800         /* Set defaults before we parse the mount options */
 801         def_mount_opts = le32_to_cpu(es->s_default_mount_opts);
 802         if (def_mount_opts & EXT2_DEFM_DEBUG)
 803                 set_opt(sbi->s_mount_opt, DEBUG);
 804         if (def_mount_opts & EXT2_DEFM_BSDGROUPS)
 805                 set_opt(sbi->s_mount_opt, GRPID);
 806         if (def_mount_opts & EXT2_DEFM_UID16)
 807                 set_opt(sbi->s_mount_opt, NO_UID32);
 808 #ifdef CONFIG_EXT2_FS_XATTR
 809         if (def_mount_opts & EXT2_DEFM_XATTR_USER)
 810                 set_opt(sbi->s_mount_opt, XATTR_USER);
 811 #endif
 812 #ifdef CONFIG_EXT2_FS_POSIX_ACL
 813         if (def_mount_opts & EXT2_DEFM_ACL)
 814                 set_opt(sbi->s_mount_opt, POSIX_ACL);
 815 #endif
 816
 817         if (le16_to_cpu(sbi->s_es->s_errors) == EXT2_ERRORS_PANIC)
 818                 set_opt(sbi->s_mount_opt, ERRORS_PANIC);
 819         else if (le16_to_cpu(sbi->s_es->s_errors) == EXT2_ERRORS_CONTINUE)
 820                 set_opt(sbi->s_mount_opt, ERRORS_CONT);
 821         else
 822                 set_opt(sbi->s_mount_opt, ERRORS_RO);
 823
 824         sbi->s_resuid = le16_to_cpu(es->s_def_resuid);
 825         sbi->s_resgid = le16_to_cpu(es->s_def_resgid);
 826
 827         set_opt(sbi->s_mount_opt, RESERVATION);
 828
 829         if (!parse_options ((char *) data, sbi))
 830                 goto failed_mount;
 831
 832         sb->s_flags = (sb->s_flags & ~MS_POSIXACL) |
 833                 ((EXT2_SB(sb)->s_mount_opt & EXT2_MOUNT_POSIX_ACL) ?
 834                  MS_POSIXACL : 0);
 835
 836         ext2_xip_verify_sb(sb); /* see if bdev supports xip, unset
 837                                     EXT2_MOUNT_XIP if not */
 838
 839         if (le32_to_cpu(es->s_rev_level) == EXT2_GOOD_OLD_REV &&
 840             (EXT2_HAS_COMPAT_FEATURE(sb, ~0U) ||
 841              EXT2_HAS_RO_COMPAT_FEATURE(sb, ~0U) ||
 842              EXT2_HAS_INCOMPAT_FEATURE(sb, ~0U)))
 843                 printk("EXT2-fs warning: feature flags set on rev 0 fs, "
 844                        "running e2fsck is recommended\n");
 845         /*
 846          * Check feature flags regardless of the revision level, since we
 847          * previously didn't change the revision level when setting the flags,
 848          * so there is a chance incompat flags are set on a rev 0 filesystem.
 849          */
 850         features = EXT2_HAS_INCOMPAT_FEATURE(sb, ~EXT2_FEATURE_INCOMPAT_SUPP);
 851         if (features) {
 852                 printk("EXT2-fs: %s: couldn't mount because of "
 853                        "unsupported optional features (%x).\n",
 854                        sb->s_id, le32_to_cpu(features));
 855                 goto failed_mount;
 856         }
 857         if (!(sb->s_flags & MS_RDONLY) &&
 858             (features = EXT2_HAS_RO_COMPAT_FEATURE(sb, ~EXT2_FEATURE_RO_COMPAT_SUPP))){
 859                 printk("EXT2-fs: %s: couldn't mount RDWR because of "
 860                        "unsupported optional features (%x).\n",
 861                        sb->s_id, le32_to_cpu(features));
 862                 goto failed_mount;
 863         }
 864
 865         blocksize = BLOCK_SIZE << le32_to_cpu(sbi->s_es->s_log_block_size);
 866
 867         if (ext2_use_xip(sb) && blocksize != PAGE_SIZE) {
 868                 if (!silent)
 869                         printk("XIP: Unsupported blocksize\n");
 870                 goto failed_mount;
 871         }
 872
 873         /* If the blocksize doesn't match, re-read the thing.. */
 874         if (sb->s_blocksize != blocksize) {
 875                 brelse(bh);
 876
 877                 if (!sb_set_blocksize(sb, blocksize)) {
 878                         printk(KERN_ERR "EXT2-fs: blocksize too small for device.\n");
 879                         goto failed_sbi;
 880                 }
 881
 882                 logic_sb_block = (sb_block*BLOCK_SIZE) / blocksize;
 883                 offset = (sb_block*BLOCK_SIZE) % blocksize;
 884                 bh = sb_bread(sb, logic_sb_block);
 885                 if(!bh) {
 886                         printk("EXT2-fs: Couldn't read superblock on "
 887                                "2nd try.\n");
 888                         goto failed_sbi;
 889                 }
 890                 es = (struct ext2_super_block *) (((char *)bh->b_data) + offset);
 891                 sbi->s_es = es;
 892                 if (es->s_magic != cpu_to_le16(EXT2_SUPER_MAGIC)) {
 893                         printk ("EXT2-fs: Magic mismatch, very weird !\n");
 894                         goto failed_mount;
 895                 }
 896         }
 897
 898         sb->s_maxbytes = ext2_max_size(sb->s_blocksize_bits);
 899
 900         if (le32_to_cpu(es->s_rev_level) == EXT2_GOOD_OLD_REV) {
 901                 sbi->s_inode_size = EXT2_GOOD_OLD_INODE_SIZE;
 902                 sbi->s_first_ino = EXT2_GOOD_OLD_FIRST_INO;
 903         } else {
 904                 sbi->s_inode_size = le16_to_cpu(es->s_inode_size);
 905                 sbi->s_first_ino = le32_to_cpu(es->s_first_ino);
 906                 if ((sbi->s_inode_size < EXT2_GOOD_OLD_INODE_SIZE) ||
 907                     !is_power_of_2(sbi->s_inode_size) ||
 908                     (sbi->s_inode_size > blocksize)) {
 909                         printk ("EXT2-fs: unsupported inode size: %d\n",
 910                                 sbi->s_inode_size);
 911                         goto failed_mount;
 912                 }
 913         }
 914
 915         sbi->s_frag_size = EXT2_MIN_FRAG_SIZE <<
 916                                    le32_to_cpu(es->s_log_frag_size);
 917         if (sbi->s_frag_size == 0)
 918                 goto cantfind_ext2;
 919         sbi->s_frags_per_block = sb->s_blocksize / sbi->s_frag_size;
 920
 921         sbi->s_blocks_per_group = le32_to_cpu(es->s_blocks_per_group);
 922         sbi->s_frags_per_group = le32_to_cpu(es->s_frags_per_group);
 923         sbi->s_inodes_per_group = le32_to_cpu(es->s_inodes_per_group);
 924
 925         if (EXT2_INODE_SIZE(sb) == 0)
 926                 goto cantfind_ext2;
 927         sbi->s_inodes_per_block = sb->s_blocksize / EXT2_INODE_SIZE(sb);
 928         if (sbi->s_inodes_per_block == 0 || sbi->s_inodes_per_group == 0)
 929                 goto cantfind_ext2;
 930         sbi->s_itb_per_group = sbi->s_inodes_per_group /
 931                                         sbi->s_inodes_per_block;
 932         sbi->s_desc_per_block = sb->s_blocksize /
 933                                         sizeof (struct ext2_group_desc);
 934         sbi->s_sbh = bh;
 935         sbi->s_mount_state = le16_to_cpu(es->s_state);
 936         sbi->s_addr_per_block_bits =
 937                 ilog2 (EXT2_ADDR_PER_BLOCK(sb));
 938         sbi->s_desc_per_block_bits =
 939                 ilog2 (EXT2_DESC_PER_BLOCK(sb));
 940
 941         if (sb->s_magic != EXT2_SUPER_MAGIC)
 942                 goto cantfind_ext2;
 943
 944         if (sb->s_blocksize != bh->b_size) {
 945                 if (!silent)
 946                         printk ("VFS: Unsupported blocksize on dev "
 947                                 "%s.\n", sb->s_id);
 948                 goto failed_mount;
 949         }
 950
 951         if (sb->s_blocksize != sbi->s_frag_size) {
 952                 printk ("EXT2-fs: fragsize %lu != blocksize %lu (not supported yet)\n",
 953                         sbi->s_frag_size, sb->s_blocksize);
 954                 goto failed_mount;
 955         }
 956
 957         if (sbi->s_blocks_per_group > sb->s_blocksize * 8) {
 958                 printk ("EXT2-fs: #blocks per group too big: %lu\n",
 959                         sbi->s_blocks_per_group);
 960                 goto failed_mount;
 961         }
 962         if (sbi->s_frags_per_group > sb->s_blocksize * 8) {
 963                 printk ("EXT2-fs: #fragments per group too big: %lu\n",
 964                         sbi->s_frags_per_group);
 965                 goto failed_mount;
 966         }
 967         if (sbi->s_inodes_per_group > sb->s_blocksize * 8) {
 968                 printk ("EXT2-fs: #inodes per group too big: %lu\n",
 969                         sbi->s_inodes_per_group);
 970                 goto failed_mount;
 971         }
 972
 973         if (EXT2_BLOCKS_PER_GROUP(sb) == 0)
 974                 goto cantfind_ext2;
 975         sbi->s_groups_count = ((le32_to_cpu(es->s_blocks_count) -
 976                                 le32_to_cpu(es->s_first_data_block) - 1)
 977                                         / EXT2_BLOCKS_PER_GROUP(sb)) + 1;
 978         db_count = (sbi->s_groups_count + EXT2_DESC_PER_BLOCK(sb) - 1) /
 979                    EXT2_DESC_PER_BLOCK(sb);
 980         sbi->s_group_desc = kmalloc (db_count * sizeof (struct buffer_head *), GFP_KERNEL);
 981         if (sbi->s_group_desc == NULL) {
 982                 printk ("EXT2-fs: not enough memory\n");
 983                 goto failed_mount;
 984         }
 985         bgl_lock_init(sbi->s_blockgroup_lock);
 986         sbi->s_debts = kcalloc(sbi->s_groups_count, sizeof(*sbi->s_debts), GFP_KERNEL);
 987         if (!sbi->s_debts) {
 988                 printk ("EXT2-fs: not enough memory\n");
 989                 goto failed_mount_group_desc;
 990         }
 991         for (i = 0; i < db_count; i++) {
 992                 block = descriptor_loc(sb, logic_sb_block, i);
 993                 sbi->s_group_desc[i] = sb_bread(sb, block);
 994                 if (!sbi->s_group_desc[i]) {
 995                         for (j = 0; j < i; j++)
 996                                 brelse (sbi->s_group_desc[j]);
 997                         printk ("EXT2-fs: unable to read group descriptors\n");
 998                         goto failed_mount_group_desc;
 999                 }
1000         }
1001         if (!ext2_check_descriptors (sb)) {
1002                 printk ("EXT2-fs: group descriptors corrupted!\n");
1003                 goto failed_mount2;
1004         }
1005         sbi->s_gdb_count = db_count;
1006         get_random_bytes(&sbi->s_next_generation, sizeof(u32));
1007         spin_lock_init(&sbi->s_next_gen_lock);
1008
1009         /* per fileystem reservation list head & lock */
1010         spin_lock_init(&sbi->s_rsv_window_lock);
1011         sbi->s_rsv_window_root = RB_ROOT;
1012         /*
1013          * Add a single, static dummy reservation to the start of the
1014          * reservation window list --- it gives us a placeholder for
1015          * append-at-start-of-list which makes the allocation logic
1016          * _much_ simpler.
1017          */
1018         sbi->s_rsv_window_head.rsv_start = EXT2_RESERVE_WINDOW_NOT_ALLOCATED;
1019         sbi->s_rsv_window_head.rsv_end = EXT2_RESERVE_WINDOW_NOT_ALLOCATED;
1020         sbi->s_rsv_window_head.rsv_alloc_hit = 0;
1021         sbi->s_rsv_window_head.rsv_goal_size = 0;
1022         ext2_rsv_window_add(sb, &sbi->s_rsv_window_head);
1023
1024         err = percpu_counter_init(&sbi->s_freeblocks_counter,
1025                                 ext2_count_free_blocks(sb));
1026         if (!err) {
1027                 err = percpu_counter_init(&sbi->s_freeinodes_counter,
1028                                 ext2_count_free_inodes(sb));
1029         }
1030         if (!err) {
1031                 err = percpu_counter_init(&sbi->s_dirs_counter,
1032                                 ext2_count_dirs(sb));
1033         }
1034         if (err) {
1035                 printk(KERN_ERR "EXT2-fs: insufficient memory\n");
1036                 goto failed_mount3;
1037         }
1038         /*
1039          * set up enough so that it can read an inode
1040          */
1041         sb->s_op = &ext2_sops;
1042         sb->s_export_op = &ext2_export_ops;
1043         sb->s_xattr = ext2_xattr_handlers;
1044         root = ext2_iget(sb, EXT2_ROOT_INO);
1045         if (IS_ERR(root)) {
1046                 ret = PTR_ERR(root);
1047                 goto failed_mount3;
1048         }
1049         if (!S_ISDIR(root->i_mode) || !root->i_blocks || !root->i_size) {
1050                 iput(root);
1051                 printk(KERN_ERR "EXT2-fs: corrupt root inode, run e2fsck\n");
1052                 goto failed_mount3;
1053         }
1054
1055         sb->s_root = d_alloc_root(root);
1056         if (!sb->s_root) {
1057                 iput(root);
1058                 printk(KERN_ERR "EXT2-fs: get root inode failed\n");
1059                 ret = -ENOMEM;
1060                 goto failed_mount3;
1061         }
1062         if (EXT2_HAS_COMPAT_FEATURE(sb, EXT3_FEATURE_COMPAT_HAS_JOURNAL))
1063                 ext2_warning(sb, __func__,
1064                         "mounting ext3 filesystem as ext2");
1065         ext2_setup_super (sb, es, sb->s_flags & MS_RDONLY);
1066         return 0;
1067
1068 cantfind_ext2:
1069         if (!silent)
1070                 printk("VFS: Can't find an ext2 filesystem on dev %s.\n",
1071                        sb->s_id);
1072         goto failed_mount;
1073 failed_mount3:
1074         percpu_counter_destroy(&sbi->s_freeblocks_counter);
1075         percpu_counter_destroy(&sbi->s_freeinodes_counter);
1076         percpu_counter_destroy(&sbi->s_dirs_counter);
1077 failed_mount2:
1078         for (i = 0; i < db_count; i++)
1079                 brelse(sbi->s_group_desc[i]);
1080 failed_mount_group_desc:
1081         kfree(sbi->s_group_desc);
1082         kfree(sbi->s_debts);
1083 failed_mount:
1084         brelse(bh);
1085 failed_sbi:
1086         sb->s_fs_info = NULL;
1087         kfree(sbi->s_blockgroup_lock);
1088         kfree(sbi);
1089         return ret;
1090 }
1091
1092 static void ext2_commit_super (struct super_block * sb,
1093                                struct ext2_super_block * es)
1094 {
1095         es->s_wtime = cpu_to_le32(get_seconds());
1096         mark_buffer_dirty(EXT2_SB(sb)->s_sbh);
1097         sb->s_dirt = 0;
1098 }
1099
1100 static void ext2_sync_super(struct super_block *sb, struct ext2_super_block *es)
1101 {
1102         es->s_free_blocks_count = cpu_to_le32(ext2_count_free_blocks(sb));
1103         es->s_free_inodes_count = cpu_to_le32(ext2_count_free_inodes(sb));
1104         es->s_wtime = cpu_to_le32(get_seconds());
1105         mark_buffer_dirty(EXT2_SB(sb)->s_sbh);
1106         sync_dirty_buffer(EXT2_SB(sb)->s_sbh);
1107         sb->s_dirt = 0;
1108 }
1109
1110 /*
1111  * In the second extended file system, it is not necessary to
1112  * write the super block since we use a mapping of the
1113  * disk super block in a buffer.
1114  *
1115  * However, this function is still used to set the fs valid
1116  * flags to 0.  We need to set this flag to 0 since the fs
1117  * may have been checked while mounted and e2fsck may have
1118  * set s_state to EXT2_VALID_FS after some corrections.
1119  */
1120
1121 static int ext2_sync_fs(struct super_block *sb, int wait)
1122 {
1123         struct ext2_super_block *es = EXT2_SB(sb)->s_es;
1124
1125         lock_kernel();
1126         if (es->s_state & cpu_to_le16(EXT2_VALID_FS)) {
1127                 ext2_debug("setting valid to 0\n");
1128                 es->s_state &= cpu_to_le16(~EXT2_VALID_FS);
1129                 es->s_free_blocks_count =
1130                         cpu_to_le32(ext2_count_free_blocks(sb));
1131                 es->s_free_inodes_count =
1132                         cpu_to_le32(ext2_count_free_inodes(sb));
1133                 es->s_mtime = cpu_to_le32(get_seconds());
1134                 ext2_sync_super(sb, es);
1135         } else {
1136                 ext2_commit_super(sb, es);
1137         }
1138         sb->s_dirt = 0;
1139         unlock_kernel();
1140
1141         return 0;
1142 }
1143
1144
1145 void ext2_write_super(struct super_block *sb)
1146 {
1147         if (!(sb->s_flags & MS_RDONLY))
1148                 ext2_sync_fs(sb, 1);
1149         else
1150                 sb->s_dirt = 0;
1151 }
1152
1153 static int ext2_remount (struct super_block * sb, int * flags, char * data)
1154 {
1155         struct ext2_sb_info * sbi = EXT2_SB(sb);
1156         struct ext2_super_block * es;
1157         unsigned long old_mount_opt = sbi->s_mount_opt;
1158         struct ext2_mount_options old_opts;
1159         unsigned long old_sb_flags;
1160         int err;
1161
1162         lock_kernel();
1163
1164         /* Store the old options */
1165         old_sb_flags = sb->s_flags;
1166         old_opts.s_mount_opt = sbi->s_mount_opt;
1167         old_opts.s_resuid = sbi->s_resuid;
1168         old_opts.s_resgid = sbi->s_resgid;
1169
1170         /*
1171          * Allow the "check" option to be passed as a remount option.
1172          */
1173         if (!parse_options (data, sbi)) {
1174                 err = -EINVAL;
1175                 goto restore_opts;
1176         }
1177
1178         sb->s_flags = (sb->s_flags & ~MS_POSIXACL) |
1179                 ((sbi->s_mount_opt & EXT2_MOUNT_POSIX_ACL) ? MS_POSIXACL : 0);
1180
1181         ext2_xip_verify_sb(sb); /* see if bdev supports xip, unset
1182                                     EXT2_MOUNT_XIP if not */
1183
1184         if ((ext2_use_xip(sb)) && (sb->s_blocksize != PAGE_SIZE)) {
1185                 printk("XIP: Unsupported blocksize\n");
1186                 err = -EINVAL;
1187                 goto restore_opts;
1188         }
1189
1190         es = sbi->s_es;
1191         if (((sbi->s_mount_opt & EXT2_MOUNT_XIP) !=
1192             (old_mount_opt & EXT2_MOUNT_XIP)) &&
1193             invalidate_inodes(sb)) {
1194                 ext2_warning(sb, __func__, "refusing change of xip flag "
1195                              "with busy inodes while remounting");
1196                 sbi->s_mount_opt &= ~EXT2_MOUNT_XIP;
1197                 sbi->s_mount_opt |= old_mount_opt & EXT2_MOUNT_XIP;
1198         }
1199         if ((*flags & MS_RDONLY) == (sb->s_flags & MS_RDONLY)) {
1200                 unlock_kernel();
1201                 return 0;
1202         }
1203         if (*flags & MS_RDONLY) {
1204                 if (le16_to_cpu(es->s_state) & EXT2_VALID_FS ||
1205                     !(sbi->s_mount_state & EXT2_VALID_FS)) {
1206                         unlock_kernel();
1207                         return 0;
1208                 }
1209                 /*
1210                  * OK, we are remounting a valid rw partition rdonly, so set
1211                  * the rdonly flag and then mark the partition as valid again.
1212                  */
1213                 es->s_state = cpu_to_le16(sbi->s_mount_state);
1214                 es->s_mtime = cpu_to_le32(get_seconds());
1215         } else {
1216                 __le32 ret = EXT2_HAS_RO_COMPAT_FEATURE(sb,
1217                                                ~EXT2_FEATURE_RO_COMPAT_SUPP);
1218                 if (ret) {
1219                         printk("EXT2-fs: %s: couldn't remount RDWR because of "
1220                                "unsupported optional features (%x).\n",
1221                                sb->s_id, le32_to_cpu(ret));
1222                         err = -EROFS;
1223                         goto restore_opts;
1224                 }
1225                 /*
1226                  * Mounting a RDONLY partition read-write, so reread and
1227                  * store the current valid flag.  (It may have been changed
1228                  * by e2fsck since we originally mounted the partition.)
1229                  */
1230                 sbi->s_mount_state = le16_to_cpu(es->s_state);
1231                 if (!ext2_setup_super (sb, es, 0))
1232                         sb->s_flags &= ~MS_RDONLY;
1233         }
1234         ext2_sync_super(sb, es);
1235         unlock_kernel();
1236         return 0;
1237 restore_opts:
1238         sbi->s_mount_opt = old_opts.s_mount_opt;
1239         sbi->s_resuid = old_opts.s_resuid;
1240         sbi->s_resgid = old_opts.s_resgid;
1241         sb->s_flags = old_sb_flags;
1242         unlock_kernel();
1243         return err;
1244 }
1245
1246 static int ext2_statfs (struct dentry * dentry, struct kstatfs * buf)
1247 {
1248         struct super_block *sb = dentry->d_sb;
1249         struct ext2_sb_info *sbi = EXT2_SB(sb);
1250         struct ext2_super_block *es = sbi->s_es;
1251         u64 fsid;
1252
1253         if (test_opt (sb, MINIX_DF))
1254                 sbi->s_overhead_last = 0;
1255         else if (sbi->s_blocks_last != le32_to_cpu(es->s_blocks_count)) {
1256                 unsigned long i, overhead = 0;
1257                 smp_rmb();
1258
1259                 /*
1260                  * Compute the overhead (FS structures). This is constant
1261                  * for a given filesystem unless the number of block groups
1262                  * changes so we cache the previous value until it does.
1263                  */
1264
1265                 /*
1266                  * All of the blocks before first_data_block are
1267                  * overhead
1268                  */
1269                 overhead = le32_to_cpu(es->s_first_data_block);
1270
1271                 /*
1272                  * Add the overhead attributed to the superblock and
1273                  * block group descriptors.  If the sparse superblocks
1274                  * feature is turned on, then not all groups have this.
1275                  */
1276                 for (i = 0; i < sbi->s_groups_count; i++)
1277                         overhead += ext2_bg_has_super(sb, i) +
1278                                 ext2_bg_num_gdb(sb, i);
1279
1280                 /*
1281                  * Every block group has an inode bitmap, a block
1282                  * bitmap, and an inode table.
1283                  */
1284                 overhead += (sbi->s_groups_count *
1285                              (2 + sbi->s_itb_per_group));
1286                 sbi->s_overhead_last = overhead;
1287                 smp_wmb();
1288                 sbi->s_blocks_last = le32_to_cpu(es->s_blocks_count);
1289         }
1290
1291         buf->f_type = EXT2_SUPER_MAGIC;
1292         buf->f_bsize = sb->s_blocksize;
1293         buf->f_blocks = le32_to_cpu(es->s_blocks_count) - sbi->s_overhead_last;
1294         buf->f_bfree = ext2_count_free_blocks(sb);
1295         es->s_free_blocks_count = cpu_to_le32(buf->f_bfree);
1296         buf->f_bavail = buf->f_bfree - le32_to_cpu(es->s_r_blocks_count);
1297         if (buf->f_bfree < le32_to_cpu(es->s_r_blocks_count))
1298                 buf->f_bavail = 0;
1299         buf->f_files = le32_to_cpu(es->s_inodes_count);
1300         buf->f_ffree = ext2_count_free_inodes(sb);
1301         es->s_free_inodes_count = cpu_to_le32(buf->f_ffree);
1302         buf->f_namelen = EXT2_NAME_LEN;
1303         fsid = le64_to_cpup((void *)es->s_uuid) ^
1304                le64_to_cpup((void *)es->s_uuid + sizeof(u64));
1305         buf->f_fsid.val[0] = fsid & 0xFFFFFFFFUL;
1306         buf->f_fsid.val[1] = (fsid >> 32) & 0xFFFFFFFFUL;
1307         return 0;
1308 }
1309
1310 static int ext2_get_sb(struct file_system_type *fs_type,
1311         int flags, const char *dev_name, void *data, struct vfsmount *mnt)
1312 {
1313         return get_sb_bdev(fs_type, flags, dev_name, data, ext2_fill_super, mnt);
1314 }
1315
1316 #ifdef CONFIG_QUOTA
1317
1318 /* Read data from quotafile - avoid pagecache and such because we cannot afford
1319  * acquiring the locks... As quota files are never truncated and quota code
1320  * itself serializes the operations (and noone else should touch the files)
1321  * we don't have to be afraid of races */
1322 static ssize_t ext2_quota_read(struct super_block *sb, int type, char *data,
1323                                size_t len, loff_t off)
1324 {
1325         struct inode *inode = sb_dqopt(sb)->files[type];
1326         sector_t blk = off >> EXT2_BLOCK_SIZE_BITS(sb);
1327         int err = 0;
1328         int offset = off & (sb->s_blocksize - 1);
1329         int tocopy;
1330         size_t toread;
1331         struct buffer_head tmp_bh;
1332         struct buffer_head *bh;
1333         loff_t i_size = i_size_read(inode);
1334
1335         if (off > i_size)
1336                 return 0;
1337         if (off+len > i_size)
1338                 len = i_size-off;
1339         toread = len;
1340         while (toread > 0) {
1341                 tocopy = sb->s_blocksize - offset < toread ?
1342                                 sb->s_blocksize - offset : toread;
1343
1344                 tmp_bh.b_state = 0;
1345                 tmp_bh.b_size = sb->s_blocksize;
1346                 err = ext2_get_block(inode, blk, &tmp_bh, 0);
1347                 if (err < 0)
1348                         return err;
1349                 if (!buffer_mapped(&tmp_bh))    /* A hole? */
1350                         memset(data, 0, tocopy);
1351                 else {
1352                         bh = sb_bread(sb, tmp_bh.b_blocknr);
1353                         if (!bh)
1354                                 return -EIO;
1355                         memcpy(data, bh->b_data+offset, tocopy);
1356                         brelse(bh);
1357                 }
1358                 offset = 0;
1359                 toread -= tocopy;
1360                 data += tocopy;
1361                 blk++;
1362         }
1363         return len;
1364 }
1365
1366 /* Write to quotafile */
1367 static ssize_t ext2_quota_write(struct super_block *sb, int type,
1368                                 const char *data, size_t len, loff_t off)
1369 {
1370         struct inode *inode = sb_dqopt(sb)->files[type];
1371         sector_t blk = off >> EXT2_BLOCK_SIZE_BITS(sb);
1372         int err = 0;
1373         int offset = off & (sb->s_blocksize - 1);
1374         int tocopy;
1375         size_t towrite = len;
1376         struct buffer_head tmp_bh;
1377         struct buffer_head *bh;
1378
1379         mutex_lock_nested(&inode->i_mutex, I_MUTEX_QUOTA);
1380         while (towrite > 0) {
1381                 tocopy = sb->s_blocksize - offset < towrite ?
1382                                 sb->s_blocksize - offset : towrite;
1383
1384                 tmp_bh.b_state = 0;
1385                 err = ext2_get_block(inode, blk, &tmp_bh, 1);
1386                 if (err < 0)
1387                         goto out;
1388                 if (offset || tocopy != EXT2_BLOCK_SIZE(sb))
1389                         bh = sb_bread(sb, tmp_bh.b_blocknr);
1390                 else
1391                         bh = sb_getblk(sb, tmp_bh.b_blocknr);
1392                 if (!bh) {
1393                         err = -EIO;
1394                         goto out;
1395                 }
1396                 lock_buffer(bh);
1397                 memcpy(bh->b_data+offset, data, tocopy);
1398                 flush_dcache_page(bh->b_page);
1399                 set_buffer_uptodate(bh);
1400                 mark_buffer_dirty(bh);
1401                 unlock_buffer(bh);
1402                 brelse(bh);
1403                 offset = 0;
1404                 towrite -= tocopy;
1405                 data += tocopy;
1406                 blk++;
1407         }
1408 out:
1409         if (len == towrite) {
1410                 mutex_unlock(&inode->i_mutex);
1411                 return err;
1412         }
1413         if (inode->i_size < off+len-towrite)
1414                 i_size_write(inode, off+len-towrite);
1415         inode->i_version++;
1416         inode->i_mtime = inode->i_ctime = CURRENT_TIME;
1417         mark_inode_dirty(inode);
1418         mutex_unlock(&inode->i_mutex);
1419         return len - towrite;
1420 }
1421
1422 #endif
1423
1424 static struct file_system_type ext2_fs_type = {
1425         .owner          = THIS_MODULE,
1426         .name           = "ext2",
1427         .get_sb         = ext2_get_sb,
1428         .kill_sb        = kill_block_super,
1429         .fs_flags       = FS_REQUIRES_DEV,
1430 };
1431
1432 static int __init init_ext2_fs(void)
1433 {
1434         int err = init_ext2_xattr();
1435         if (err)
1436                 return err;
1437         err = init_inodecache();
1438         if (err)
1439                 goto out1;
1440         err = register_filesystem(&ext2_fs_type);
1441         if (err)
1442                 goto out;
1443         return 0;
1444 out:
1445         destroy_inodecache();
1446 out1:
1447         exit_ext2_xattr();
1448         return err;
1449 }
1450
1451 static void __exit exit_ext2_fs(void)
1452 {
1453         unregister_filesystem(&ext2_fs_type);
1454         destroy_inodecache();
1455         exit_ext2_xattr();
1456 }
1457
1458 module_init(init_ext2_fs)
1459 module_exit(exit_ext2_fs)