fs/ext2/super.c

   1 /*
   2  *  linux/fs/ext2/super.c
   3  *
   4  * Copyright (C) 1992, 1993, 1994, 1995
   5  * Remy Card ([email protected])
   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 ([email protected]), 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 (%s): error: %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                 ext2_msg(sb, KERN_CRIT,
  70                              "error: remounting filesystem read-only");
  71                 sb->s_flags |= MS_RDONLY;
  72         }
  73 }
  74
  75 void ext2_msg(struct super_block *sb, const char *prefix,
  76                 const char *fmt, ...)
  77 {
  78         va_list args;
  79
  80         va_start(args, fmt);
  81         printk("%sEXT2-fs (%s): ", prefix, sb->s_id);
  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_msg(sb, KERN_WARNING,
  95                      "warning: updating to rev %d because of "
  96                      "new feature flag, 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, struct super_block *sb)
 423 {
 424         char *p;
 425         struct ext2_sb_info *sbi = EXT2_SB(sb);
 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                         ext2_msg(sb, KERN_INFO, "(no)user_xattr options"
 509                                 "not supported");
 510                         break;
 511 #endif
 512 #ifdef CONFIG_EXT2_FS_POSIX_ACL
 513                 case Opt_acl:
 514                         set_opt(sbi->s_mount_opt, POSIX_ACL);
 515                         break;
 516                 case Opt_noacl:
 517                         clear_opt(sbi->s_mount_opt, POSIX_ACL);
 518                         break;
 519 #else
 520                 case Opt_acl:
 521                 case Opt_noacl:
 522                         ext2_msg(sb, KERN_INFO,
 523                                 "(no)acl options not supported");
 524                         break;
 525 #endif
 526                 case Opt_xip:
 527 #ifdef CONFIG_EXT2_FS_XIP
 528                         set_opt (sbi->s_mount_opt, XIP);
 529 #else
 530                         ext2_msg(sb, KERN_INFO, "xip option not supported");
 531 #endif
 532                         break;
 533
 534 #if defined(CONFIG_QUOTA)
 535                 case Opt_quota:
 536                 case Opt_usrquota:
 537                         set_opt(sbi->s_mount_opt, USRQUOTA);
 538                         break;
 539
 540                 case Opt_grpquota:
 541                         set_opt(sbi->s_mount_opt, GRPQUOTA);
 542                         break;
 543 #else
 544                 case Opt_quota:
 545                 case Opt_usrquota:
 546                 case Opt_grpquota:
 547                         ext2_msg(sb, KERN_INFO,
 548                                 "quota operations not supported");
 549                         break;
 550 #endif
 551
 552                 case Opt_reservation:
 553                         set_opt(sbi->s_mount_opt, RESERVATION);
 554                         ext2_msg(sb, KERN_INFO, "reservations ON");
 555                         break;
 556                 case Opt_noreservation:
 557                         clear_opt(sbi->s_mount_opt, RESERVATION);
 558                         ext2_msg(sb, KERN_INFO, "reservations OFF");
 559                         break;
 560                 case Opt_ignore:
 561                         break;
 562                 default:
 563                         return 0;
 564                 }
 565         }
 566         return 1;
 567 }
 568
 569 static int ext2_setup_super (struct super_block * sb,
 570                               struct ext2_super_block * es,
 571                               int read_only)
 572 {
 573         int res = 0;
 574         struct ext2_sb_info *sbi = EXT2_SB(sb);
 575
 576         if (le32_to_cpu(es->s_rev_level) > EXT2_MAX_SUPP_REV) {
 577                 ext2_msg(sb, KERN_ERR,
 578                         "error: revision level too high, "
 579                         "forcing read-only mode");
 580                 res = MS_RDONLY;
 581         }
 582         if (read_only)
 583                 return res;
 584         if (!(sbi->s_mount_state & EXT2_VALID_FS))
 585                 ext2_msg(sb, KERN_WARNING,
 586                         "warning: mounting unchecked fs, "
 587                         "running e2fsck is recommended");
 588         else if ((sbi->s_mount_state & EXT2_ERROR_FS))
 589                 ext2_msg(sb, KERN_WARNING,
 590                         "warning: mounting fs with errors, "
 591                         "running e2fsck is recommended");
 592         else if ((__s16) le16_to_cpu(es->s_max_mnt_count) >= 0 &&
 593                  le16_to_cpu(es->s_mnt_count) >=
 594                  (unsigned short) (__s16) le16_to_cpu(es->s_max_mnt_count))
 595                 ext2_msg(sb, KERN_WARNING,
 596                         "warning: maximal mount count reached, "
 597                         "running e2fsck is recommended");
 598         else if (le32_to_cpu(es->s_checkinterval) &&
 599                 (le32_to_cpu(es->s_lastcheck) +
 600                         le32_to_cpu(es->s_checkinterval) <= get_seconds()))
 601                 ext2_msg(sb, KERN_WARNING,
 602                         "warning: checktime reached, "
 603                         "running e2fsck is recommended");
 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         le16_add_cpu(&es->s_mnt_count, 1);
 607         ext2_write_super(sb);
 608         if (test_opt (sb, DEBUG))
 609                 ext2_msg(sb, KERN_INFO, "%s, %s, bs=%lu, fs=%lu, gc=%lu, "
 610                         "bpg=%lu, ipg=%lu, mo=%04lx]",
 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         struct ext2_sb_info *sbi = EXT2_SB(sb);
 624
 625         ext2_debug ("Checking group descriptors");
 626
 627         for (i = 0; i < sbi->s_groups_count; i++) {
 628                 struct ext2_group_desc *gdp = ext2_get_group_desc(sb, i, NULL);
 629                 ext2_fsblk_t first_block = ext2_group_first_block_no(sb, i);
 630                 ext2_fsblk_t last_block;
 631
 632                 if (i == sbi->s_groups_count - 1)
 633                         last_block = le32_to_cpu(sbi->s_es->s_blocks_count) - 1;
 634                 else
 635                         last_block = first_block +
 636                                 (EXT2_BLOCKS_PER_GROUP(sb) - 1);
 637
 638                 if (le32_to_cpu(gdp->bg_block_bitmap) < first_block ||
 639                     le32_to_cpu(gdp->bg_block_bitmap) > last_block)
 640                 {
 641                         ext2_error (sb, "ext2_check_descriptors",
 642                                     "Block bitmap for group %d"
 643                                     " not in group (block %lu)!",
 644                                     i, (unsigned long) le32_to_cpu(gdp->bg_block_bitmap));
 645                         return 0;
 646                 }
 647                 if (le32_to_cpu(gdp->bg_inode_bitmap) < first_block ||
 648                     le32_to_cpu(gdp->bg_inode_bitmap) > last_block)
 649                 {
 650                         ext2_error (sb, "ext2_check_descriptors",
 651                                     "Inode bitmap for group %d"
 652                                     " not in group (block %lu)!",
 653                                     i, (unsigned long) le32_to_cpu(gdp->bg_inode_bitmap));
 654                         return 0;
 655                 }
 656                 if (le32_to_cpu(gdp->bg_inode_table) < first_block ||
 657                     le32_to_cpu(gdp->bg_inode_table) + sbi->s_itb_per_group - 1 >
 658                     last_block)
 659                 {
 660                         ext2_error (sb, "ext2_check_descriptors",
 661                                     "Inode table for group %d"
 662                                     " not in group (block %lu)!",
 663                                     i, (unsigned long) le32_to_cpu(gdp->bg_inode_table));
 664                         return 0;
 665                 }
 666         }
 667         return 1;
 668 }
 669
 670 /*
 671  * Maximal file size.  There is a direct, and {,double-,triple-}indirect
 672  * block limit, and also a limit of (2^32 - 1) 512-byte sectors in i_blocks.
 673  * We need to be 1 filesystem block less than the 2^32 sector limit.
 674  */
 675 static loff_t ext2_max_size(int bits)
 676 {
 677         loff_t res = EXT2_NDIR_BLOCKS;
 678         int meta_blocks;
 679         loff_t upper_limit;
 680
 681         /* This is calculated to be the largest file size for a
 682          * dense, file such that the total number of
 683          * sectors in the file, including data and all indirect blocks,
 684          * does not exceed 2^32 -1
 685          * __u32 i_blocks representing the total number of
 686          * 512 bytes blocks of the file
 687          */
 688         upper_limit = (1LL << 32) - 1;
 689
 690         /* total blocks in file system block size */
 691         upper_limit >>= (bits - 9);
 692
 693
 694         /* indirect blocks */
 695         meta_blocks = 1;
 696         /* double indirect blocks */
 697         meta_blocks += 1 + (1LL << (bits-2));
 698         /* tripple indirect blocks */
 699         meta_blocks += 1 + (1LL << (bits-2)) + (1LL << (2*(bits-2)));
 700
 701         upper_limit -= meta_blocks;
 702         upper_limit <<= bits;
 703
 704         res += 1LL << (bits-2);
 705         res += 1LL << (2*(bits-2));
 706         res += 1LL << (3*(bits-2));
 707         res <<= bits;
 708         if (res > upper_limit)
 709                 res = upper_limit;
 710
 711         if (res > MAX_LFS_FILESIZE)
 712                 res = MAX_LFS_FILESIZE;
 713
 714         return res;
 715 }
 716
 717 static unsigned long descriptor_loc(struct super_block *sb,
 718                                     unsigned long logic_sb_block,
 719                                     int nr)
 720 {
 721         struct ext2_sb_info *sbi = EXT2_SB(sb);
 722         unsigned long bg, first_meta_bg;
 723         int has_super = 0;
 724
 725         first_meta_bg = le32_to_cpu(sbi->s_es->s_first_meta_bg);
 726
 727         if (!EXT2_HAS_INCOMPAT_FEATURE(sb, EXT2_FEATURE_INCOMPAT_META_BG) ||
 728             nr < first_meta_bg)
 729                 return (logic_sb_block + nr + 1);
 730         bg = sbi->s_desc_per_block * nr;
 731         if (ext2_bg_has_super(sb, bg))
 732                 has_super = 1;
 733
 734         return ext2_group_first_block_no(sb, bg) + has_super;
 735 }
 736
 737 static int ext2_fill_super(struct super_block *sb, void *data, int silent)
 738 {
 739         struct buffer_head * bh;
 740         struct ext2_sb_info * sbi;
 741         struct ext2_super_block * es;
 742         struct inode *root;
 743         unsigned long block;
 744         unsigned long sb_block = get_sb_block(&data);
 745         unsigned long logic_sb_block;
 746         unsigned long offset = 0;
 747         unsigned long def_mount_opts;
 748         long ret = -EINVAL;
 749         int blocksize = BLOCK_SIZE;
 750         int db_count;
 751         int i, j;
 752         __le32 features;
 753         int err;
 754
 755         sbi = kzalloc(sizeof(*sbi), GFP_KERNEL);
 756         if (!sbi)
 757                 return -ENOMEM;
 758
 759         sbi->s_blockgroup_lock =
 760                 kzalloc(sizeof(struct blockgroup_lock), GFP_KERNEL);
 761         if (!sbi->s_blockgroup_lock) {
 762                 kfree(sbi);
 763                 return -ENOMEM;
 764         }
 765         sb->s_fs_info = sbi;
 766         sbi->s_sb_block = sb_block;
 767
 768         /*
 769          * See what the current blocksize for the device is, and
 770          * use that as the blocksize.  Otherwise (or if the blocksize
 771          * is smaller than the default) use the default.
 772          * This is important for devices that have a hardware
 773          * sectorsize that is larger than the default.
 774          */
 775         blocksize = sb_min_blocksize(sb, BLOCK_SIZE);
 776         if (!blocksize) {
 777                 ext2_msg(sb, KERN_ERR, "error: unable to set blocksize");
 778                 goto failed_sbi;
 779         }
 780
 781         /*
 782          * If the superblock doesn't start on a hardware sector boundary,
 783          * calculate the offset.
 784          */
 785         if (blocksize != BLOCK_SIZE) {
 786                 logic_sb_block = (sb_block*BLOCK_SIZE) / blocksize;
 787                 offset = (sb_block*BLOCK_SIZE) % blocksize;
 788         } else {
 789                 logic_sb_block = sb_block;
 790         }
 791
 792         if (!(bh = sb_bread(sb, logic_sb_block))) {
 793                 ext2_msg(sb, KERN_ERR, "error: unable to read superblock");
 794                 goto failed_sbi;
 795         }
 796         /*
 797          * Note: s_es must be initialized as soon as possible because
 798          *       some ext2 macro-instructions depend on its value
 799          */
 800         es = (struct ext2_super_block *) (((char *)bh->b_data) + offset);
 801         sbi->s_es = es;
 802         sb->s_magic = le16_to_cpu(es->s_magic);
 803
 804         if (sb->s_magic != EXT2_SUPER_MAGIC)
 805                 goto cantfind_ext2;
 806
 807         /* Set defaults before we parse the mount options */
 808         def_mount_opts = le32_to_cpu(es->s_default_mount_opts);
 809         if (def_mount_opts & EXT2_DEFM_DEBUG)
 810                 set_opt(sbi->s_mount_opt, DEBUG);
 811         if (def_mount_opts & EXT2_DEFM_BSDGROUPS)
 812                 set_opt(sbi->s_mount_opt, GRPID);
 813         if (def_mount_opts & EXT2_DEFM_UID16)
 814                 set_opt(sbi->s_mount_opt, NO_UID32);
 815 #ifdef CONFIG_EXT2_FS_XATTR
 816         if (def_mount_opts & EXT2_DEFM_XATTR_USER)
 817                 set_opt(sbi->s_mount_opt, XATTR_USER);
 818 #endif
 819 #ifdef CONFIG_EXT2_FS_POSIX_ACL
 820         if (def_mount_opts & EXT2_DEFM_ACL)
 821                 set_opt(sbi->s_mount_opt, POSIX_ACL);
 822 #endif
 823
 824         if (le16_to_cpu(sbi->s_es->s_errors) == EXT2_ERRORS_PANIC)
 825                 set_opt(sbi->s_mount_opt, ERRORS_PANIC);
 826         else if (le16_to_cpu(sbi->s_es->s_errors) == EXT2_ERRORS_CONTINUE)
 827                 set_opt(sbi->s_mount_opt, ERRORS_CONT);
 828         else
 829                 set_opt(sbi->s_mount_opt, ERRORS_RO);
 830
 831         sbi->s_resuid = le16_to_cpu(es->s_def_resuid);
 832         sbi->s_resgid = le16_to_cpu(es->s_def_resgid);
 833
 834         set_opt(sbi->s_mount_opt, RESERVATION);
 835
 836         if (!parse_options((char *) data, sb))
 837                 goto failed_mount;
 838
 839         sb->s_flags = (sb->s_flags & ~MS_POSIXACL) |
 840                 ((EXT2_SB(sb)->s_mount_opt & EXT2_MOUNT_POSIX_ACL) ?
 841                  MS_POSIXACL : 0);
 842
 843         ext2_xip_verify_sb(sb); /* see if bdev supports xip, unset
 844                                     EXT2_MOUNT_XIP if not */
 845
 846         if (le32_to_cpu(es->s_rev_level) == EXT2_GOOD_OLD_REV &&
 847             (EXT2_HAS_COMPAT_FEATURE(sb, ~0U) ||
 848              EXT2_HAS_RO_COMPAT_FEATURE(sb, ~0U) ||
 849              EXT2_HAS_INCOMPAT_FEATURE(sb, ~0U)))
 850                 ext2_msg(sb, KERN_WARNING,
 851                         "warning: feature flags set on rev 0 fs, "
 852                         "running e2fsck is recommended");
 853         /*
 854          * Check feature flags regardless of the revision level, since we
 855          * previously didn't change the revision level when setting the flags,
 856          * so there is a chance incompat flags are set on a rev 0 filesystem.
 857          */
 858         features = EXT2_HAS_INCOMPAT_FEATURE(sb, ~EXT2_FEATURE_INCOMPAT_SUPP);
 859         if (features) {
 860                 ext2_msg(sb, KERN_ERR,  "error: couldn't mount because of "
 861                        "unsupported optional features (%x)",
 862                         le32_to_cpu(features));
 863                 goto failed_mount;
 864         }
 865         if (!(sb->s_flags & MS_RDONLY) &&
 866             (features = EXT2_HAS_RO_COMPAT_FEATURE(sb, ~EXT2_FEATURE_RO_COMPAT_SUPP))){
 867                 ext2_msg(sb, KERN_ERR, "error: couldn't mount RDWR because of "
 868                        "unsupported optional features (%x)",
 869                        le32_to_cpu(features));
 870                 goto failed_mount;
 871         }
 872
 873         blocksize = BLOCK_SIZE << le32_to_cpu(sbi->s_es->s_log_block_size);
 874
 875         if (ext2_use_xip(sb) && blocksize != PAGE_SIZE) {
 876                 if (!silent)
 877                         ext2_msg(sb, KERN_ERR,
 878                                 "error: unsupported blocksize for xip");
 879                 goto failed_mount;
 880         }
 881
 882         /* If the blocksize doesn't match, re-read the thing.. */
 883         if (sb->s_blocksize != blocksize) {
 884                 brelse(bh);
 885
 886                 if (!sb_set_blocksize(sb, blocksize)) {
 887                         ext2_msg(sb, KERN_ERR, "error: blocksize is too small");
 888                         goto failed_sbi;
 889                 }
 890
 891                 logic_sb_block = (sb_block*BLOCK_SIZE) / blocksize;
 892                 offset = (sb_block*BLOCK_SIZE) % blocksize;
 893                 bh = sb_bread(sb, logic_sb_block);
 894                 if(!bh) {
 895                         ext2_msg(sb, KERN_ERR, "error: couldn't read"
 896                                 "superblock on 2nd try");
 897                         goto failed_sbi;
 898                 }
 899                 es = (struct ext2_super_block *) (((char *)bh->b_data) + offset);
 900                 sbi->s_es = es;
 901                 if (es->s_magic != cpu_to_le16(EXT2_SUPER_MAGIC)) {
 902                         ext2_msg(sb, KERN_ERR, "error: magic mismatch");
 903                         goto failed_mount;
 904                 }
 905         }
 906
 907         sb->s_maxbytes = ext2_max_size(sb->s_blocksize_bits);
 908
 909         if (le32_to_cpu(es->s_rev_level) == EXT2_GOOD_OLD_REV) {
 910                 sbi->s_inode_size = EXT2_GOOD_OLD_INODE_SIZE;
 911                 sbi->s_first_ino = EXT2_GOOD_OLD_FIRST_INO;
 912         } else {
 913                 sbi->s_inode_size = le16_to_cpu(es->s_inode_size);
 914                 sbi->s_first_ino = le32_to_cpu(es->s_first_ino);
 915                 if ((sbi->s_inode_size < EXT2_GOOD_OLD_INODE_SIZE) ||
 916                     !is_power_of_2(sbi->s_inode_size) ||
 917                     (sbi->s_inode_size > blocksize)) {
 918                         ext2_msg(sb, KERN_ERR,
 919                                 "error: unsupported inode size: %d",
 920                                 sbi->s_inode_size);
 921                         goto failed_mount;
 922                 }
 923         }
 924
 925         sbi->s_frag_size = EXT2_MIN_FRAG_SIZE <<
 926                                    le32_to_cpu(es->s_log_frag_size);
 927         if (sbi->s_frag_size == 0)
 928                 goto cantfind_ext2;
 929         sbi->s_frags_per_block = sb->s_blocksize / sbi->s_frag_size;
 930
 931         sbi->s_blocks_per_group = le32_to_cpu(es->s_blocks_per_group);
 932         sbi->s_frags_per_group = le32_to_cpu(es->s_frags_per_group);
 933         sbi->s_inodes_per_group = le32_to_cpu(es->s_inodes_per_group);
 934
 935         if (EXT2_INODE_SIZE(sb) == 0)
 936                 goto cantfind_ext2;
 937         sbi->s_inodes_per_block = sb->s_blocksize / EXT2_INODE_SIZE(sb);
 938         if (sbi->s_inodes_per_block == 0 || sbi->s_inodes_per_group == 0)
 939                 goto cantfind_ext2;
 940         sbi->s_itb_per_group = sbi->s_inodes_per_group /
 941                                         sbi->s_inodes_per_block;
 942         sbi->s_desc_per_block = sb->s_blocksize /
 943                                         sizeof (struct ext2_group_desc);
 944         sbi->s_sbh = bh;
 945         sbi->s_mount_state = le16_to_cpu(es->s_state);
 946         sbi->s_addr_per_block_bits =
 947                 ilog2 (EXT2_ADDR_PER_BLOCK(sb));
 948         sbi->s_desc_per_block_bits =
 949                 ilog2 (EXT2_DESC_PER_BLOCK(sb));
 950
 951         if (sb->s_magic != EXT2_SUPER_MAGIC)
 952                 goto cantfind_ext2;
 953
 954         if (sb->s_blocksize != bh->b_size) {
 955                 if (!silent)
 956                         ext2_msg(sb, KERN_ERR, "error: unsupported blocksize");
 957                 goto failed_mount;
 958         }
 959
 960         if (sb->s_blocksize != sbi->s_frag_size) {
 961                 ext2_msg(sb, KERN_ERR,
 962                         "error: fragsize %lu != blocksize %lu"
 963                         "(not supported yet)",
 964                         sbi->s_frag_size, sb->s_blocksize);
 965                 goto failed_mount;
 966         }
 967
 968         if (sbi->s_blocks_per_group > sb->s_blocksize * 8) {
 969                 ext2_msg(sb, KERN_ERR,
 970                         "error: #blocks per group too big: %lu",
 971                         sbi->s_blocks_per_group);
 972                 goto failed_mount;
 973         }
 974         if (sbi->s_frags_per_group > sb->s_blocksize * 8) {
 975                 ext2_msg(sb, KERN_ERR,
 976                         "error: #fragments per group too big: %lu",
 977                         sbi->s_frags_per_group);
 978                 goto failed_mount;
 979         }
 980         if (sbi->s_inodes_per_group > sb->s_blocksize * 8) {
 981                 ext2_msg(sb, KERN_ERR,
 982                         "error: #inodes per group too big: %lu",
 983                         sbi->s_inodes_per_group);
 984                 goto failed_mount;
 985         }
 986
 987         if (EXT2_BLOCKS_PER_GROUP(sb) == 0)
 988                 goto cantfind_ext2;
 989         sbi->s_groups_count = ((le32_to_cpu(es->s_blocks_count) -
 990                                 le32_to_cpu(es->s_first_data_block) - 1)
 991                                         / EXT2_BLOCKS_PER_GROUP(sb)) + 1;
 992         db_count = (sbi->s_groups_count + EXT2_DESC_PER_BLOCK(sb) - 1) /
 993                    EXT2_DESC_PER_BLOCK(sb);
 994         sbi->s_group_desc = kmalloc (db_count * sizeof (struct buffer_head *), GFP_KERNEL);
 995         if (sbi->s_group_desc == NULL) {
 996                 ext2_msg(sb, KERN_ERR, "error: not enough memory");
 997                 goto failed_mount;
 998         }
 999         bgl_lock_init(sbi->s_blockgroup_lock);
1000         sbi->s_debts = kcalloc(sbi->s_groups_count, sizeof(*sbi->s_debts), GFP_KERNEL);
1001         if (!sbi->s_debts) {
1002                 ext2_msg(sb, KERN_ERR, "error: not enough memory");
1003                 goto failed_mount_group_desc;
1004         }
1005         for (i = 0; i < db_count; i++) {
1006                 block = descriptor_loc(sb, logic_sb_block, i);
1007                 sbi->s_group_desc[i] = sb_bread(sb, block);
1008                 if (!sbi->s_group_desc[i]) {
1009                         for (j = 0; j < i; j++)
1010                                 brelse (sbi->s_group_desc[j]);
1011                         ext2_msg(sb, KERN_ERR,
1012                                 "error: unable to read group descriptors");
1013                         goto failed_mount_group_desc;
1014                 }
1015         }
1016         if (!ext2_check_descriptors (sb)) {
1017                 ext2_msg(sb, KERN_ERR, "group descriptors corrupted");
1018                 goto failed_mount2;
1019         }
1020         sbi->s_gdb_count = db_count;
1021         get_random_bytes(&sbi->s_next_generation, sizeof(u32));
1022         spin_lock_init(&sbi->s_next_gen_lock);
1023
1024         /* per fileystem reservation list head & lock */
1025         spin_lock_init(&sbi->s_rsv_window_lock);
1026         sbi->s_rsv_window_root = RB_ROOT;
1027         /*
1028          * Add a single, static dummy reservation to the start of the
1029          * reservation window list --- it gives us a placeholder for
1030          * append-at-start-of-list which makes the allocation logic
1031          * _much_ simpler.
1032          */
1033         sbi->s_rsv_window_head.rsv_start = EXT2_RESERVE_WINDOW_NOT_ALLOCATED;
1034         sbi->s_rsv_window_head.rsv_end = EXT2_RESERVE_WINDOW_NOT_ALLOCATED;
1035         sbi->s_rsv_window_head.rsv_alloc_hit = 0;
1036         sbi->s_rsv_window_head.rsv_goal_size = 0;
1037         ext2_rsv_window_add(sb, &sbi->s_rsv_window_head);
1038
1039         err = percpu_counter_init(&sbi->s_freeblocks_counter,
1040                                 ext2_count_free_blocks(sb));
1041         if (!err) {
1042                 err = percpu_counter_init(&sbi->s_freeinodes_counter,
1043                                 ext2_count_free_inodes(sb));
1044         }
1045         if (!err) {
1046                 err = percpu_counter_init(&sbi->s_dirs_counter,
1047                                 ext2_count_dirs(sb));
1048         }
1049         if (err) {
1050                 ext2_msg(sb, KERN_ERR, "error: insufficient memory");
1051                 goto failed_mount3;
1052         }
1053         /*
1054          * set up enough so that it can read an inode
1055          */
1056         sb->s_op = &ext2_sops;
1057         sb->s_export_op = &ext2_export_ops;
1058         sb->s_xattr = ext2_xattr_handlers;
1059         root = ext2_iget(sb, EXT2_ROOT_INO);
1060         if (IS_ERR(root)) {
1061                 ret = PTR_ERR(root);
1062                 goto failed_mount3;
1063         }
1064         if (!S_ISDIR(root->i_mode) || !root->i_blocks || !root->i_size) {
1065                 iput(root);
1066                 ext2_msg(sb, KERN_ERR, "error: corrupt root inode, run e2fsck");
1067                 goto failed_mount3;
1068         }
1069
1070         sb->s_root = d_alloc_root(root);
1071         if (!sb->s_root) {
1072                 iput(root);
1073                 ext2_msg(sb, KERN_ERR, "error: get root inode failed");
1074                 ret = -ENOMEM;
1075                 goto failed_mount3;
1076         }
1077         if (EXT2_HAS_COMPAT_FEATURE(sb, EXT3_FEATURE_COMPAT_HAS_JOURNAL))
1078                 ext2_msg(sb, KERN_WARNING,
1079                         "warning: mounting ext3 filesystem as ext2");
1080         ext2_setup_super (sb, es, sb->s_flags & MS_RDONLY);
1081         return 0;
1082
1083 cantfind_ext2:
1084         if (!silent)
1085                 ext2_msg(sb, KERN_ERR,
1086                         "error: can't find an ext2 filesystem on dev %s.",
1087                         sb->s_id);
1088         goto failed_mount;
1089 failed_mount3:
1090         percpu_counter_destroy(&sbi->s_freeblocks_counter);
1091         percpu_counter_destroy(&sbi->s_freeinodes_counter);
1092         percpu_counter_destroy(&sbi->s_dirs_counter);
1093 failed_mount2:
1094         for (i = 0; i < db_count; i++)
1095                 brelse(sbi->s_group_desc[i]);
1096 failed_mount_group_desc:
1097         kfree(sbi->s_group_desc);
1098         kfree(sbi->s_debts);
1099 failed_mount:
1100         brelse(bh);
1101 failed_sbi:
1102         sb->s_fs_info = NULL;
1103         kfree(sbi->s_blockgroup_lock);
1104         kfree(sbi);
1105         return ret;
1106 }
1107
1108 static void ext2_clear_super_error(struct super_block *sb)
1109 {
1110         struct buffer_head *sbh = EXT2_SB(sb)->s_sbh;
1111
1112         if (buffer_write_io_error(sbh)) {
1113                 /*
1114                  * Oh, dear.  A previous attempt to write the
1115                  * superblock failed.  This could happen because the
1116                  * USB device was yanked out.  Or it could happen to
1117                  * be a transient write error and maybe the block will
1118                  * be remapped.  Nothing we can do but to retry the
1119                  * write and hope for the best.
1120                  */
1121                 printk(KERN_ERR "EXT2-fs: %s previous I/O error to "
1122                        "superblock detected", sb->s_id);
1123                 clear_buffer_write_io_error(sbh);
1124                 set_buffer_uptodate(sbh);
1125         }
1126 }
1127
1128 static void ext2_commit_super (struct super_block * sb,
1129                                struct ext2_super_block * es)
1130 {
1131         ext2_clear_super_error(sb);
1132         es->s_wtime = cpu_to_le32(get_seconds());
1133         mark_buffer_dirty(EXT2_SB(sb)->s_sbh);
1134         sb->s_dirt = 0;
1135 }
1136
1137 static void ext2_sync_super(struct super_block *sb, struct ext2_super_block *es)
1138 {
1139         ext2_clear_super_error(sb);
1140         es->s_free_blocks_count = cpu_to_le32(ext2_count_free_blocks(sb));
1141         es->s_free_inodes_count = cpu_to_le32(ext2_count_free_inodes(sb));
1142         es->s_wtime = cpu_to_le32(get_seconds());
1143         mark_buffer_dirty(EXT2_SB(sb)->s_sbh);
1144         sync_dirty_buffer(EXT2_SB(sb)->s_sbh);
1145         sb->s_dirt = 0;
1146 }
1147
1148 /*
1149  * In the second extended file system, it is not necessary to
1150  * write the super block since we use a mapping of the
1151  * disk super block in a buffer.
1152  *
1153  * However, this function is still used to set the fs valid
1154  * flags to 0.  We need to set this flag to 0 since the fs
1155  * may have been checked while mounted and e2fsck may have
1156  * set s_state to EXT2_VALID_FS after some corrections.
1157  */
1158
1159 static int ext2_sync_fs(struct super_block *sb, int wait)
1160 {
1161         struct ext2_super_block *es = EXT2_SB(sb)->s_es;
1162         struct buffer_head *sbh = EXT2_SB(sb)->s_sbh;
1163
1164         lock_kernel();
1165         if (buffer_write_io_error(sbh)) {
1166                 /*
1167                  * Oh, dear.  A previous attempt to write the
1168                  * superblock failed.  This could happen because the
1169                  * USB device was yanked out.  Or it could happen to
1170                  * be a transient write error and maybe the block will
1171                  * be remapped.  Nothing we can do but to retry the
1172                  * write and hope for the best.
1173                  */
1174                 ext2_msg(sb, KERN_ERR,
1175                        "previous I/O error to superblock detected\n");
1176                 clear_buffer_write_io_error(sbh);
1177                 set_buffer_uptodate(sbh);
1178         }
1179
1180         if (es->s_state & cpu_to_le16(EXT2_VALID_FS)) {
1181                 ext2_debug("setting valid to 0\n");
1182                 es->s_state &= cpu_to_le16(~EXT2_VALID_FS);
1183                 es->s_free_blocks_count =
1184                         cpu_to_le32(ext2_count_free_blocks(sb));
1185                 es->s_free_inodes_count =
1186                         cpu_to_le32(ext2_count_free_inodes(sb));
1187                 es->s_mtime = cpu_to_le32(get_seconds());
1188                 ext2_sync_super(sb, es);
1189         } else {
1190                 ext2_commit_super(sb, es);
1191         }
1192         sb->s_dirt = 0;
1193         unlock_kernel();
1194
1195         return 0;
1196 }
1197
1198
1199 void ext2_write_super(struct super_block *sb)
1200 {
1201         if (!(sb->s_flags & MS_RDONLY))
1202                 ext2_sync_fs(sb, 1);
1203         else
1204                 sb->s_dirt = 0;
1205 }
1206
1207 static int ext2_remount (struct super_block * sb, int * flags, char * data)
1208 {
1209         struct ext2_sb_info * sbi = EXT2_SB(sb);
1210         struct ext2_super_block * es;
1211         unsigned long old_mount_opt = sbi->s_mount_opt;
1212         struct ext2_mount_options old_opts;
1213         unsigned long old_sb_flags;
1214         int err;
1215
1216         lock_kernel();
1217
1218         /* Store the old options */
1219         old_sb_flags = sb->s_flags;
1220         old_opts.s_mount_opt = sbi->s_mount_opt;
1221         old_opts.s_resuid = sbi->s_resuid;
1222         old_opts.s_resgid = sbi->s_resgid;
1223
1224         /*
1225          * Allow the "check" option to be passed as a remount option.
1226          */
1227         if (!parse_options(data, sb)) {
1228                 err = -EINVAL;
1229                 goto restore_opts;
1230         }
1231
1232         sb->s_flags = (sb->s_flags & ~MS_POSIXACL) |
1233                 ((sbi->s_mount_opt & EXT2_MOUNT_POSIX_ACL) ? MS_POSIXACL : 0);
1234
1235         ext2_xip_verify_sb(sb); /* see if bdev supports xip, unset
1236                                     EXT2_MOUNT_XIP if not */
1237
1238         if ((ext2_use_xip(sb)) && (sb->s_blocksize != PAGE_SIZE)) {
1239                 ext2_msg(sb, KERN_WARNING,
1240                         "warning: unsupported blocksize for xip");
1241                 err = -EINVAL;
1242                 goto restore_opts;
1243         }
1244
1245         es = sbi->s_es;
1246         if (((sbi->s_mount_opt & EXT2_MOUNT_XIP) !=
1247             (old_mount_opt & EXT2_MOUNT_XIP)) &&
1248             invalidate_inodes(sb)) {
1249                 ext2_msg(sb, KERN_WARNING, "warning: refusing change of "
1250                          "xip flag with busy inodes while remounting");
1251                 sbi->s_mount_opt &= ~EXT2_MOUNT_XIP;
1252                 sbi->s_mount_opt |= old_mount_opt & EXT2_MOUNT_XIP;
1253         }
1254         if ((*flags & MS_RDONLY) == (sb->s_flags & MS_RDONLY)) {
1255                 unlock_kernel();
1256                 return 0;
1257         }
1258         if (*flags & MS_RDONLY) {
1259                 if (le16_to_cpu(es->s_state) & EXT2_VALID_FS ||
1260                     !(sbi->s_mount_state & EXT2_VALID_FS)) {
1261                         unlock_kernel();
1262                         return 0;
1263                 }
1264                 /*
1265                  * OK, we are remounting a valid rw partition rdonly, so set
1266                  * the rdonly flag and then mark the partition as valid again.
1267                  */
1268                 es->s_state = cpu_to_le16(sbi->s_mount_state);
1269                 es->s_mtime = cpu_to_le32(get_seconds());
1270         } else {
1271                 __le32 ret = EXT2_HAS_RO_COMPAT_FEATURE(sb,
1272                                                ~EXT2_FEATURE_RO_COMPAT_SUPP);
1273                 if (ret) {
1274                         ext2_msg(sb, KERN_WARNING,
1275                                 "warning: couldn't remount RDWR because of "
1276                                 "unsupported optional features (%x).",
1277                                 le32_to_cpu(ret));
1278                         err = -EROFS;
1279                         goto restore_opts;
1280                 }
1281                 /*
1282                  * Mounting a RDONLY partition read-write, so reread and
1283                  * store the current valid flag.  (It may have been changed
1284                  * by e2fsck since we originally mounted the partition.)
1285                  */
1286                 sbi->s_mount_state = le16_to_cpu(es->s_state);
1287                 if (!ext2_setup_super (sb, es, 0))
1288                         sb->s_flags &= ~MS_RDONLY;
1289         }
1290         ext2_sync_super(sb, es);
1291         unlock_kernel();
1292         return 0;
1293 restore_opts:
1294         sbi->s_mount_opt = old_opts.s_mount_opt;
1295         sbi->s_resuid = old_opts.s_resuid;
1296         sbi->s_resgid = old_opts.s_resgid;
1297         sb->s_flags = old_sb_flags;
1298         unlock_kernel();
1299         return err;
1300 }
1301
1302 static int ext2_statfs (struct dentry * dentry, struct kstatfs * buf)
1303 {
1304         struct super_block *sb = dentry->d_sb;
1305         struct ext2_sb_info *sbi = EXT2_SB(sb);
1306         struct ext2_super_block *es = sbi->s_es;
1307         u64 fsid;
1308
1309         if (test_opt (sb, MINIX_DF))
1310                 sbi->s_overhead_last = 0;
1311         else if (sbi->s_blocks_last != le32_to_cpu(es->s_blocks_count)) {
1312                 unsigned long i, overhead = 0;
1313                 smp_rmb();
1314
1315                 /*
1316                  * Compute the overhead (FS structures). This is constant
1317                  * for a given filesystem unless the number of block groups
1318                  * changes so we cache the previous value until it does.
1319                  */
1320
1321                 /*
1322                  * All of the blocks before first_data_block are
1323                  * overhead
1324                  */
1325                 overhead = le32_to_cpu(es->s_first_data_block);
1326
1327                 /*
1328                  * Add the overhead attributed to the superblock and
1329                  * block group descriptors.  If the sparse superblocks
1330                  * feature is turned on, then not all groups have this.
1331                  */
1332                 for (i = 0; i < sbi->s_groups_count; i++)
1333                         overhead += ext2_bg_has_super(sb, i) +
1334                                 ext2_bg_num_gdb(sb, i);
1335
1336                 /*
1337                  * Every block group has an inode bitmap, a block
1338                  * bitmap, and an inode table.
1339                  */
1340                 overhead += (sbi->s_groups_count *
1341                              (2 + sbi->s_itb_per_group));
1342                 sbi->s_overhead_last = overhead;
1343                 smp_wmb();
1344                 sbi->s_blocks_last = le32_to_cpu(es->s_blocks_count);
1345         }
1346
1347         buf->f_type = EXT2_SUPER_MAGIC;
1348         buf->f_bsize = sb->s_blocksize;
1349         buf->f_blocks = le32_to_cpu(es->s_blocks_count) - sbi->s_overhead_last;
1350         buf->f_bfree = ext2_count_free_blocks(sb);
1351         es->s_free_blocks_count = cpu_to_le32(buf->f_bfree);
1352         buf->f_bavail = buf->f_bfree - le32_to_cpu(es->s_r_blocks_count);
1353         if (buf->f_bfree < le32_to_cpu(es->s_r_blocks_count))
1354                 buf->f_bavail = 0;
1355         buf->f_files = le32_to_cpu(es->s_inodes_count);
1356         buf->f_ffree = ext2_count_free_inodes(sb);
1357         es->s_free_inodes_count = cpu_to_le32(buf->f_ffree);
1358         buf->f_namelen = EXT2_NAME_LEN;
1359         fsid = le64_to_cpup((void *)es->s_uuid) ^
1360                le64_to_cpup((void *)es->s_uuid + sizeof(u64));
1361         buf->f_fsid.val[0] = fsid & 0xFFFFFFFFUL;
1362         buf->f_fsid.val[1] = (fsid >> 32) & 0xFFFFFFFFUL;
1363         return 0;
1364 }
1365
1366 static int ext2_get_sb(struct file_system_type *fs_type,
1367         int flags, const char *dev_name, void *data, struct vfsmount *mnt)
1368 {
1369         return get_sb_bdev(fs_type, flags, dev_name, data, ext2_fill_super, mnt);
1370 }
1371
1372 #ifdef CONFIG_QUOTA
1373
1374 /* Read data from quotafile - avoid pagecache and such because we cannot afford
1375  * acquiring the locks... As quota files are never truncated and quota code
1376  * itself serializes the operations (and noone else should touch the files)
1377  * we don't have to be afraid of races */
1378 static ssize_t ext2_quota_read(struct super_block *sb, int type, char *data,
1379                                size_t len, loff_t off)
1380 {
1381         struct inode *inode = sb_dqopt(sb)->files[type];
1382         sector_t blk = off >> EXT2_BLOCK_SIZE_BITS(sb);
1383         int err = 0;
1384         int offset = off & (sb->s_blocksize - 1);
1385         int tocopy;
1386         size_t toread;
1387         struct buffer_head tmp_bh;
1388         struct buffer_head *bh;
1389         loff_t i_size = i_size_read(inode);
1390
1391         if (off > i_size)
1392                 return 0;
1393         if (off+len > i_size)
1394                 len = i_size-off;
1395         toread = len;
1396         while (toread > 0) {
1397                 tocopy = sb->s_blocksize - offset < toread ?
1398                                 sb->s_blocksize - offset : toread;
1399
1400                 tmp_bh.b_state = 0;
1401                 tmp_bh.b_size = sb->s_blocksize;
1402                 err = ext2_get_block(inode, blk, &tmp_bh, 0);
1403                 if (err < 0)
1404                         return err;
1405                 if (!buffer_mapped(&tmp_bh))    /* A hole? */
1406                         memset(data, 0, tocopy);
1407                 else {
1408                         bh = sb_bread(sb, tmp_bh.b_blocknr);
1409                         if (!bh)
1410                                 return -EIO;
1411                         memcpy(data, bh->b_data+offset, tocopy);
1412                         brelse(bh);
1413                 }
1414                 offset = 0;
1415                 toread -= tocopy;
1416                 data += tocopy;
1417                 blk++;
1418         }
1419         return len;
1420 }
1421
1422 /* Write to quotafile */
1423 static ssize_t ext2_quota_write(struct super_block *sb, int type,
1424                                 const char *data, size_t len, loff_t off)
1425 {
1426         struct inode *inode = sb_dqopt(sb)->files[type];
1427         sector_t blk = off >> EXT2_BLOCK_SIZE_BITS(sb);
1428         int err = 0;
1429         int offset = off & (sb->s_blocksize - 1);
1430         int tocopy;
1431         size_t towrite = len;
1432         struct buffer_head tmp_bh;
1433         struct buffer_head *bh;
1434
1435         mutex_lock_nested(&inode->i_mutex, I_MUTEX_QUOTA);
1436         while (towrite > 0) {
1437                 tocopy = sb->s_blocksize - offset < towrite ?
1438                                 sb->s_blocksize - offset : towrite;
1439
1440                 tmp_bh.b_state = 0;
1441                 err = ext2_get_block(inode, blk, &tmp_bh, 1);
1442                 if (err < 0)
1443                         goto out;
1444                 if (offset || tocopy != EXT2_BLOCK_SIZE(sb))
1445                         bh = sb_bread(sb, tmp_bh.b_blocknr);
1446                 else
1447                         bh = sb_getblk(sb, tmp_bh.b_blocknr);
1448                 if (!bh) {
1449                         err = -EIO;
1450                         goto out;
1451                 }
1452                 lock_buffer(bh);
1453                 memcpy(bh->b_data+offset, data, tocopy);
1454                 flush_dcache_page(bh->b_page);
1455                 set_buffer_uptodate(bh);
1456                 mark_buffer_dirty(bh);
1457                 unlock_buffer(bh);
1458                 brelse(bh);
1459                 offset = 0;
1460                 towrite -= tocopy;
1461                 data += tocopy;
1462                 blk++;
1463         }
1464 out:
1465         if (len == towrite) {
1466                 mutex_unlock(&inode->i_mutex);
1467                 return err;
1468         }
1469         if (inode->i_size < off+len-towrite)
1470                 i_size_write(inode, off+len-towrite);
1471         inode->i_version++;
1472         inode->i_mtime = inode->i_ctime = CURRENT_TIME;
1473         mark_inode_dirty(inode);
1474         mutex_unlock(&inode->i_mutex);
1475         return len - towrite;
1476 }
1477
1478 #endif
1479
1480 static struct file_system_type ext2_fs_type = {
1481         .owner          = THIS_MODULE,
1482         .name           = "ext2",
1483         .get_sb         = ext2_get_sb,
1484         .kill_sb        = kill_block_super,
1485         .fs_flags       = FS_REQUIRES_DEV,
1486 };
1487
1488 static int __init init_ext2_fs(void)
1489 {
1490         int err = init_ext2_xattr();
1491         if (err)
1492                 return err;
1493         err = init_inodecache();
1494         if (err)
1495                 goto out1;
1496         err = register_filesystem(&ext2_fs_type);
1497         if (err)
1498                 goto out;
1499         return 0;
1500 out:
1501         destroy_inodecache();
1502 out1:
1503         exit_ext2_xattr();
1504         return err;
1505 }
1506
1507 static void __exit exit_ext2_fs(void)
1508 {
1509         unregister_filesystem(&ext2_fs_type);
1510         destroy_inodecache();
1511         exit_ext2_xattr();
1512 }
1513
1514 module_init(init_ext2_fs)
1515 module_exit(exit_ext2_fs)