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, int wait);
  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                 spin_lock(&sbi->s_lock);
  56                 sbi->s_mount_state |= EXT2_ERROR_FS;
  57                 es->s_state |= cpu_to_le16(EXT2_ERROR_FS);
  58                 spin_unlock(&sbi->s_lock);
  59                 ext2_sync_super(sb, es, 1);
  60         }
  61
  62         va_start(args, fmt);
  63         printk(KERN_CRIT "EXT2-fs (%s): error: %s: ", sb->s_id, function);
  64         vprintk(fmt, args);
  65         printk("\n");
  66         va_end(args);
  67
  68         if (test_opt(sb, ERRORS_PANIC))
  69                 panic("EXT2-fs: panic from previous error\n");
  70         if (test_opt(sb, ERRORS_RO)) {
  71                 ext2_msg(sb, KERN_CRIT,
  72                              "error: remounting filesystem read-only");
  73                 sb->s_flags |= MS_RDONLY;
  74         }
  75 }
  76
  77 void ext2_msg(struct super_block *sb, const char *prefix,
  78                 const char *fmt, ...)
  79 {
  80         va_list args;
  81
  82         va_start(args, fmt);
  83         printk("%sEXT2-fs (%s): ", prefix, sb->s_id);
  84         vprintk(fmt, args);
  85         printk("\n");
  86         va_end(args);
  87 }
  88
  89 /*
  90  * This must be called with sbi->s_lock held.
  91  */
  92 void ext2_update_dynamic_rev(struct super_block *sb)
  93 {
  94         struct ext2_super_block *es = EXT2_SB(sb)->s_es;
  95
  96         if (le32_to_cpu(es->s_rev_level) > EXT2_GOOD_OLD_REV)
  97                 return;
  98
  99         ext2_msg(sb, KERN_WARNING,
 100                      "warning: updating to rev %d because of "
 101                      "new feature flag, running e2fsck is recommended",
 102                      EXT2_DYNAMIC_REV);
 103
 104         es->s_first_ino = cpu_to_le32(EXT2_GOOD_OLD_FIRST_INO);
 105         es->s_inode_size = cpu_to_le16(EXT2_GOOD_OLD_INODE_SIZE);
 106         es->s_rev_level = cpu_to_le32(EXT2_DYNAMIC_REV);
 107         /* leave es->s_feature_*compat flags alone */
 108         /* es->s_uuid will be set by e2fsck if empty */
 109
 110         /*
 111          * The rest of the superblock fields should be zero, and if not it
 112          * means they are likely already in use, so leave them alone.  We
 113          * can leave it up to e2fsck to clean up any inconsistencies there.
 114          */
 115 }
 116
 117 static void ext2_put_super (struct super_block * sb)
 118 {
 119         int db_count;
 120         int i;
 121         struct ext2_sb_info *sbi = EXT2_SB(sb);
 122
 123         lock_kernel();
 124
 125         if (sb->s_dirt)
 126                 ext2_write_super(sb);
 127
 128         ext2_xattr_put_super(sb);
 129         if (!(sb->s_flags & MS_RDONLY)) {
 130                 struct ext2_super_block *es = sbi->s_es;
 131
 132                 spin_lock(&sbi->s_lock);
 133                 es->s_state = cpu_to_le16(sbi->s_mount_state);
 134                 spin_unlock(&sbi->s_lock);
 135                 ext2_sync_super(sb, es, 1);
 136         }
 137         db_count = sbi->s_gdb_count;
 138         for (i = 0; i < db_count; i++)
 139                 if (sbi->s_group_desc[i])
 140                         brelse (sbi->s_group_desc[i]);
 141         kfree(sbi->s_group_desc);
 142         kfree(sbi->s_debts);
 143         percpu_counter_destroy(&sbi->s_freeblocks_counter);
 144         percpu_counter_destroy(&sbi->s_freeinodes_counter);
 145         percpu_counter_destroy(&sbi->s_dirs_counter);
 146         brelse (sbi->s_sbh);
 147         sb->s_fs_info = NULL;
 148         kfree(sbi->s_blockgroup_lock);
 149         kfree(sbi);
 150
 151         unlock_kernel();
 152 }
 153
 154 static struct kmem_cache * ext2_inode_cachep;
 155
 156 static struct inode *ext2_alloc_inode(struct super_block *sb)
 157 {
 158         struct ext2_inode_info *ei;
 159         ei = (struct ext2_inode_info *)kmem_cache_alloc(ext2_inode_cachep, GFP_KERNEL);
 160         if (!ei)
 161                 return NULL;
 162         ei->i_block_alloc_info = NULL;
 163         ei->vfs_inode.i_version = 1;
 164         return &ei->vfs_inode;
 165 }
 166
 167 static void ext2_destroy_inode(struct inode *inode)
 168 {
 169         kmem_cache_free(ext2_inode_cachep, EXT2_I(inode));
 170 }
 171
 172 static void init_once(void *foo)
 173 {
 174         struct ext2_inode_info *ei = (struct ext2_inode_info *) foo;
 175
 176         rwlock_init(&ei->i_meta_lock);
 177 #ifdef CONFIG_EXT2_FS_XATTR
 178         init_rwsem(&ei->xattr_sem);
 179 #endif
 180         mutex_init(&ei->truncate_mutex);
 181         inode_init_once(&ei->vfs_inode);
 182 }
 183
 184 static int init_inodecache(void)
 185 {
 186         ext2_inode_cachep = kmem_cache_create("ext2_inode_cache",
 187                                              sizeof(struct ext2_inode_info),
 188                                              0, (SLAB_RECLAIM_ACCOUNT|
 189                                                 SLAB_MEM_SPREAD),
 190                                              init_once);
 191         if (ext2_inode_cachep == NULL)
 192                 return -ENOMEM;
 193         return 0;
 194 }
 195
 196 static void destroy_inodecache(void)
 197 {
 198         kmem_cache_destroy(ext2_inode_cachep);
 199 }
 200
 201 static void ext2_clear_inode(struct inode *inode)
 202 {
 203         struct ext2_block_alloc_info *rsv = EXT2_I(inode)->i_block_alloc_info;
 204
 205         dquot_drop(inode);
 206         ext2_discard_reservation(inode);
 207         EXT2_I(inode)->i_block_alloc_info = NULL;
 208         if (unlikely(rsv))
 209                 kfree(rsv);
 210 }
 211
 212 static int ext2_show_options(struct seq_file *seq, struct vfsmount *vfs)
 213 {
 214         struct super_block *sb = vfs->mnt_sb;
 215         struct ext2_sb_info *sbi = EXT2_SB(sb);
 216         struct ext2_super_block *es = sbi->s_es;
 217         unsigned long def_mount_opts;
 218
 219         spin_lock(&sbi->s_lock);
 220         def_mount_opts = le32_to_cpu(es->s_default_mount_opts);
 221
 222         if (sbi->s_sb_block != 1)
 223                 seq_printf(seq, ",sb=%lu", sbi->s_sb_block);
 224         if (test_opt(sb, MINIX_DF))
 225                 seq_puts(seq, ",minixdf");
 226         if (test_opt(sb, GRPID))
 227                 seq_puts(seq, ",grpid");
 228         if (!test_opt(sb, GRPID) && (def_mount_opts & EXT2_DEFM_BSDGROUPS))
 229                 seq_puts(seq, ",nogrpid");
 230         if (sbi->s_resuid != EXT2_DEF_RESUID ||
 231             le16_to_cpu(es->s_def_resuid) != EXT2_DEF_RESUID) {
 232                 seq_printf(seq, ",resuid=%u", sbi->s_resuid);
 233         }
 234         if (sbi->s_resgid != EXT2_DEF_RESGID ||
 235             le16_to_cpu(es->s_def_resgid) != EXT2_DEF_RESGID) {
 236                 seq_printf(seq, ",resgid=%u", sbi->s_resgid);
 237         }
 238         if (test_opt(sb, ERRORS_RO)) {
 239                 int def_errors = le16_to_cpu(es->s_errors);
 240
 241                 if (def_errors == EXT2_ERRORS_PANIC ||
 242                     def_errors == EXT2_ERRORS_CONTINUE) {
 243                         seq_puts(seq, ",errors=remount-ro");
 244                 }
 245         }
 246         if (test_opt(sb, ERRORS_CONT))
 247                 seq_puts(seq, ",errors=continue");
 248         if (test_opt(sb, ERRORS_PANIC))
 249                 seq_puts(seq, ",errors=panic");
 250         if (test_opt(sb, NO_UID32))
 251                 seq_puts(seq, ",nouid32");
 252         if (test_opt(sb, DEBUG))
 253                 seq_puts(seq, ",debug");
 254         if (test_opt(sb, OLDALLOC))
 255                 seq_puts(seq, ",oldalloc");
 256
 257 #ifdef CONFIG_EXT2_FS_XATTR
 258         if (test_opt(sb, XATTR_USER))
 259                 seq_puts(seq, ",user_xattr");
 260         if (!test_opt(sb, XATTR_USER) &&
 261             (def_mount_opts & EXT2_DEFM_XATTR_USER)) {
 262                 seq_puts(seq, ",nouser_xattr");
 263         }
 264 #endif
 265
 266 #ifdef CONFIG_EXT2_FS_POSIX_ACL
 267         if (test_opt(sb, POSIX_ACL))
 268                 seq_puts(seq, ",acl");
 269         if (!test_opt(sb, POSIX_ACL) && (def_mount_opts & EXT2_DEFM_ACL))
 270                 seq_puts(seq, ",noacl");
 271 #endif
 272
 273         if (test_opt(sb, NOBH))
 274                 seq_puts(seq, ",nobh");
 275
 276 #if defined(CONFIG_QUOTA)
 277         if (sbi->s_mount_opt & EXT2_MOUNT_USRQUOTA)
 278                 seq_puts(seq, ",usrquota");
 279
 280         if (sbi->s_mount_opt & EXT2_MOUNT_GRPQUOTA)
 281                 seq_puts(seq, ",grpquota");
 282 #endif
 283
 284 #if defined(CONFIG_EXT2_FS_XIP)
 285         if (sbi->s_mount_opt & EXT2_MOUNT_XIP)
 286                 seq_puts(seq, ",xip");
 287 #endif
 288
 289         if (!test_opt(sb, RESERVATION))
 290                 seq_puts(seq, ",noreservation");
 291
 292         spin_unlock(&sbi->s_lock);
 293         return 0;
 294 }
 295
 296 #ifdef CONFIG_QUOTA
 297 static ssize_t ext2_quota_read(struct super_block *sb, int type, char *data, size_t len, loff_t off);
 298 static ssize_t ext2_quota_write(struct super_block *sb, int type, const char *data, size_t len, loff_t off);
 299 #endif
 300
 301 static const struct super_operations ext2_sops = {
 302         .alloc_inode    = ext2_alloc_inode,
 303         .destroy_inode  = ext2_destroy_inode,
 304         .write_inode    = ext2_write_inode,
 305         .delete_inode   = ext2_delete_inode,
 306         .put_super      = ext2_put_super,
 307         .write_super    = ext2_write_super,
 308         .sync_fs        = ext2_sync_fs,
 309         .statfs         = ext2_statfs,
 310         .remount_fs     = ext2_remount,
 311         .clear_inode    = ext2_clear_inode,
 312         .show_options   = ext2_show_options,
 313 #ifdef CONFIG_QUOTA
 314         .quota_read     = ext2_quota_read,
 315         .quota_write    = ext2_quota_write,
 316 #endif
 317 };
 318
 319 static struct inode *ext2_nfs_get_inode(struct super_block *sb,
 320                 u64 ino, u32 generation)
 321 {
 322         struct inode *inode;
 323
 324         if (ino < EXT2_FIRST_INO(sb) && ino != EXT2_ROOT_INO)
 325                 return ERR_PTR(-ESTALE);
 326         if (ino > le32_to_cpu(EXT2_SB(sb)->s_es->s_inodes_count))
 327                 return ERR_PTR(-ESTALE);
 328
 329         /* iget isn't really right if the inode is currently unallocated!!
 330          * ext2_read_inode currently does appropriate checks, but
 331          * it might be "neater" to call ext2_get_inode first and check
 332          * if the inode is valid.....
 333          */
 334         inode = ext2_iget(sb, ino);
 335         if (IS_ERR(inode))
 336                 return ERR_CAST(inode);
 337         if (generation && inode->i_generation != generation) {
 338                 /* we didn't find the right inode.. */
 339                 iput(inode);
 340                 return ERR_PTR(-ESTALE);
 341         }
 342         return inode;
 343 }
 344
 345 static struct dentry *ext2_fh_to_dentry(struct super_block *sb, struct fid *fid,
 346                 int fh_len, int fh_type)
 347 {
 348         return generic_fh_to_dentry(sb, fid, fh_len, fh_type,
 349                                     ext2_nfs_get_inode);
 350 }
 351
 352 static struct dentry *ext2_fh_to_parent(struct super_block *sb, struct fid *fid,
 353                 int fh_len, int fh_type)
 354 {
 355         return generic_fh_to_parent(sb, fid, fh_len, fh_type,
 356                                     ext2_nfs_get_inode);
 357 }
 358
 359 /* Yes, most of these are left as NULL!!
 360  * A NULL value implies the default, which works with ext2-like file
 361  * systems, but can be improved upon.
 362  * Currently only get_parent is required.
 363  */
 364 static const struct export_operations ext2_export_ops = {
 365         .fh_to_dentry = ext2_fh_to_dentry,
 366         .fh_to_parent = ext2_fh_to_parent,
 367         .get_parent = ext2_get_parent,
 368 };
 369
 370 static unsigned long get_sb_block(void **data)
 371 {
 372         unsigned long   sb_block;
 373         char            *options = (char *) *data;
 374
 375         if (!options || strncmp(options, "sb=", 3) != 0)
 376                 return 1;       /* Default location */
 377         options += 3;
 378         sb_block = simple_strtoul(options, &options, 0);
 379         if (*options && *options != ',') {
 380                 printk("EXT2-fs: Invalid sb specification: %s\n",
 381                        (char *) *data);
 382                 return 1;
 383         }
 384         if (*options == ',')
 385                 options++;
 386         *data = (void *) options;
 387         return sb_block;
 388 }
 389
 390 enum {
 391         Opt_bsd_df, Opt_minix_df, Opt_grpid, Opt_nogrpid,
 392         Opt_resgid, Opt_resuid, Opt_sb, Opt_err_cont, Opt_err_panic,
 393         Opt_err_ro, Opt_nouid32, Opt_nocheck, Opt_debug,
 394         Opt_oldalloc, Opt_orlov, Opt_nobh, Opt_user_xattr, Opt_nouser_xattr,
 395         Opt_acl, Opt_noacl, Opt_xip, Opt_ignore, Opt_err, Opt_quota,
 396         Opt_usrquota, Opt_grpquota, Opt_reservation, Opt_noreservation
 397 };
 398
 399 static const match_table_t tokens = {
 400         {Opt_bsd_df, "bsddf"},
 401         {Opt_minix_df, "minixdf"},
 402         {Opt_grpid, "grpid"},
 403         {Opt_grpid, "bsdgroups"},
 404         {Opt_nogrpid, "nogrpid"},
 405         {Opt_nogrpid, "sysvgroups"},
 406         {Opt_resgid, "resgid=%u"},
 407         {Opt_resuid, "resuid=%u"},
 408         {Opt_sb, "sb=%u"},
 409         {Opt_err_cont, "errors=continue"},
 410         {Opt_err_panic, "errors=panic"},
 411         {Opt_err_ro, "errors=remount-ro"},
 412         {Opt_nouid32, "nouid32"},
 413         {Opt_nocheck, "check=none"},
 414         {Opt_nocheck, "nocheck"},
 415         {Opt_debug, "debug"},
 416         {Opt_oldalloc, "oldalloc"},
 417         {Opt_orlov, "orlov"},
 418         {Opt_nobh, "nobh"},
 419         {Opt_user_xattr, "user_xattr"},
 420         {Opt_nouser_xattr, "nouser_xattr"},
 421         {Opt_acl, "acl"},
 422         {Opt_noacl, "noacl"},
 423         {Opt_xip, "xip"},
 424         {Opt_grpquota, "grpquota"},
 425         {Opt_ignore, "noquota"},
 426         {Opt_quota, "quota"},
 427         {Opt_usrquota, "usrquota"},
 428         {Opt_reservation, "reservation"},
 429         {Opt_noreservation, "noreservation"},
 430         {Opt_err, NULL}
 431 };
 432
 433 static int parse_options(char *options, struct super_block *sb)
 434 {
 435         char *p;
 436         struct ext2_sb_info *sbi = EXT2_SB(sb);
 437         substring_t args[MAX_OPT_ARGS];
 438         int option;
 439
 440         if (!options)
 441                 return 1;
 442
 443         while ((p = strsep (&options, ",")) != NULL) {
 444                 int token;
 445                 if (!*p)
 446                         continue;
 447
 448                 token = match_token(p, tokens, args);
 449                 switch (token) {
 450                 case Opt_bsd_df:
 451                         clear_opt (sbi->s_mount_opt, MINIX_DF);
 452                         break;
 453                 case Opt_minix_df:
 454                         set_opt (sbi->s_mount_opt, MINIX_DF);
 455                         break;
 456                 case Opt_grpid:
 457                         set_opt (sbi->s_mount_opt, GRPID);
 458                         break;
 459                 case Opt_nogrpid:
 460                         clear_opt (sbi->s_mount_opt, GRPID);
 461                         break;
 462                 case Opt_resuid:
 463                         if (match_int(&args[0], &option))
 464                                 return 0;
 465                         sbi->s_resuid = option;
 466                         break;
 467                 case Opt_resgid:
 468                         if (match_int(&args[0], &option))
 469                                 return 0;
 470                         sbi->s_resgid = option;
 471                         break;
 472                 case Opt_sb:
 473                         /* handled by get_sb_block() instead of here */
 474                         /* *sb_block = match_int(&args[0]); */
 475                         break;
 476                 case Opt_err_panic:
 477                         clear_opt (sbi->s_mount_opt, ERRORS_CONT);
 478                         clear_opt (sbi->s_mount_opt, ERRORS_RO);
 479                         set_opt (sbi->s_mount_opt, ERRORS_PANIC);
 480                         break;
 481                 case Opt_err_ro:
 482                         clear_opt (sbi->s_mount_opt, ERRORS_CONT);
 483                         clear_opt (sbi->s_mount_opt, ERRORS_PANIC);
 484                         set_opt (sbi->s_mount_opt, ERRORS_RO);
 485                         break;
 486                 case Opt_err_cont:
 487                         clear_opt (sbi->s_mount_opt, ERRORS_RO);
 488                         clear_opt (sbi->s_mount_opt, ERRORS_PANIC);
 489                         set_opt (sbi->s_mount_opt, ERRORS_CONT);
 490                         break;
 491                 case Opt_nouid32:
 492                         set_opt (sbi->s_mount_opt, NO_UID32);
 493                         break;
 494                 case Opt_nocheck:
 495                         clear_opt (sbi->s_mount_opt, CHECK);
 496                         break;
 497                 case Opt_debug:
 498                         set_opt (sbi->s_mount_opt, DEBUG);
 499                         break;
 500                 case Opt_oldalloc:
 501                         set_opt (sbi->s_mount_opt, OLDALLOC);
 502                         break;
 503                 case Opt_orlov:
 504                         clear_opt (sbi->s_mount_opt, OLDALLOC);
 505                         break;
 506                 case Opt_nobh:
 507                         set_opt (sbi->s_mount_opt, NOBH);
 508                         break;
 509 #ifdef CONFIG_EXT2_FS_XATTR
 510                 case Opt_user_xattr:
 511                         set_opt (sbi->s_mount_opt, XATTR_USER);
 512                         break;
 513                 case Opt_nouser_xattr:
 514                         clear_opt (sbi->s_mount_opt, XATTR_USER);
 515                         break;
 516 #else
 517                 case Opt_user_xattr:
 518                 case Opt_nouser_xattr:
 519                         ext2_msg(sb, KERN_INFO, "(no)user_xattr options"
 520                                 "not supported");
 521                         break;
 522 #endif
 523 #ifdef CONFIG_EXT2_FS_POSIX_ACL
 524                 case Opt_acl:
 525                         set_opt(sbi->s_mount_opt, POSIX_ACL);
 526                         break;
 527                 case Opt_noacl:
 528                         clear_opt(sbi->s_mount_opt, POSIX_ACL);
 529                         break;
 530 #else
 531                 case Opt_acl:
 532                 case Opt_noacl:
 533                         ext2_msg(sb, KERN_INFO,
 534                                 "(no)acl options not supported");
 535                         break;
 536 #endif
 537                 case Opt_xip:
 538 #ifdef CONFIG_EXT2_FS_XIP
 539                         set_opt (sbi->s_mount_opt, XIP);
 540 #else
 541                         ext2_msg(sb, KERN_INFO, "xip option not supported");
 542 #endif
 543                         break;
 544
 545 #if defined(CONFIG_QUOTA)
 546                 case Opt_quota:
 547                 case Opt_usrquota:
 548                         set_opt(sbi->s_mount_opt, USRQUOTA);
 549                         break;
 550
 551                 case Opt_grpquota:
 552                         set_opt(sbi->s_mount_opt, GRPQUOTA);
 553                         break;
 554 #else
 555                 case Opt_quota:
 556                 case Opt_usrquota:
 557                 case Opt_grpquota:
 558                         ext2_msg(sb, KERN_INFO,
 559                                 "quota operations not supported");
 560                         break;
 561 #endif
 562
 563                 case Opt_reservation:
 564                         set_opt(sbi->s_mount_opt, RESERVATION);
 565                         ext2_msg(sb, KERN_INFO, "reservations ON");
 566                         break;
 567                 case Opt_noreservation:
 568                         clear_opt(sbi->s_mount_opt, RESERVATION);
 569                         ext2_msg(sb, KERN_INFO, "reservations OFF");
 570                         break;
 571                 case Opt_ignore:
 572                         break;
 573                 default:
 574                         return 0;
 575                 }
 576         }
 577         return 1;
 578 }
 579
 580 static int ext2_setup_super (struct super_block * sb,
 581                               struct ext2_super_block * es,
 582                               int read_only)
 583 {
 584         int res = 0;
 585         struct ext2_sb_info *sbi = EXT2_SB(sb);
 586
 587         if (le32_to_cpu(es->s_rev_level) > EXT2_MAX_SUPP_REV) {
 588                 ext2_msg(sb, KERN_ERR,
 589                         "error: revision level too high, "
 590                         "forcing read-only mode");
 591                 res = MS_RDONLY;
 592         }
 593         if (read_only)
 594                 return res;
 595         if (!(sbi->s_mount_state & EXT2_VALID_FS))
 596                 ext2_msg(sb, KERN_WARNING,
 597                         "warning: mounting unchecked fs, "
 598                         "running e2fsck is recommended");
 599         else if ((sbi->s_mount_state & EXT2_ERROR_FS))
 600                 ext2_msg(sb, KERN_WARNING,
 601                         "warning: mounting fs with errors, "
 602                         "running e2fsck is recommended");
 603         else if ((__s16) le16_to_cpu(es->s_max_mnt_count) >= 0 &&
 604                  le16_to_cpu(es->s_mnt_count) >=
 605                  (unsigned short) (__s16) le16_to_cpu(es->s_max_mnt_count))
 606                 ext2_msg(sb, KERN_WARNING,
 607                         "warning: maximal mount count reached, "
 608                         "running e2fsck is recommended");
 609         else if (le32_to_cpu(es->s_checkinterval) &&
 610                 (le32_to_cpu(es->s_lastcheck) +
 611                         le32_to_cpu(es->s_checkinterval) <= get_seconds()))
 612                 ext2_msg(sb, KERN_WARNING,
 613                         "warning: checktime reached, "
 614                         "running e2fsck is recommended");
 615         if (!le16_to_cpu(es->s_max_mnt_count))
 616                 es->s_max_mnt_count = cpu_to_le16(EXT2_DFL_MAX_MNT_COUNT);
 617         le16_add_cpu(&es->s_mnt_count, 1);
 618         if (test_opt (sb, DEBUG))
 619                 ext2_msg(sb, KERN_INFO, "%s, %s, bs=%lu, fs=%lu, gc=%lu, "
 620                         "bpg=%lu, ipg=%lu, mo=%04lx]",
 621                         EXT2FS_VERSION, EXT2FS_DATE, sb->s_blocksize,
 622                         sbi->s_frag_size,
 623                         sbi->s_groups_count,
 624                         EXT2_BLOCKS_PER_GROUP(sb),
 625                         EXT2_INODES_PER_GROUP(sb),
 626                         sbi->s_mount_opt);
 627         return res;
 628 }
 629
 630 static int ext2_check_descriptors(struct super_block *sb)
 631 {
 632         int i;
 633         struct ext2_sb_info *sbi = EXT2_SB(sb);
 634
 635         ext2_debug ("Checking group descriptors");
 636
 637         for (i = 0; i < sbi->s_groups_count; i++) {
 638                 struct ext2_group_desc *gdp = ext2_get_group_desc(sb, i, NULL);
 639                 ext2_fsblk_t first_block = ext2_group_first_block_no(sb, i);
 640                 ext2_fsblk_t last_block;
 641
 642                 if (i == sbi->s_groups_count - 1)
 643                         last_block = le32_to_cpu(sbi->s_es->s_blocks_count) - 1;
 644                 else
 645                         last_block = first_block +
 646                                 (EXT2_BLOCKS_PER_GROUP(sb) - 1);
 647
 648                 if (le32_to_cpu(gdp->bg_block_bitmap) < first_block ||
 649                     le32_to_cpu(gdp->bg_block_bitmap) > last_block)
 650                 {
 651                         ext2_error (sb, "ext2_check_descriptors",
 652                                     "Block bitmap for group %d"
 653                                     " not in group (block %lu)!",
 654                                     i, (unsigned long) le32_to_cpu(gdp->bg_block_bitmap));
 655                         return 0;
 656                 }
 657                 if (le32_to_cpu(gdp->bg_inode_bitmap) < first_block ||
 658                     le32_to_cpu(gdp->bg_inode_bitmap) > last_block)
 659                 {
 660                         ext2_error (sb, "ext2_check_descriptors",
 661                                     "Inode bitmap for group %d"
 662                                     " not in group (block %lu)!",
 663                                     i, (unsigned long) le32_to_cpu(gdp->bg_inode_bitmap));
 664                         return 0;
 665                 }
 666                 if (le32_to_cpu(gdp->bg_inode_table) < first_block ||
 667                     le32_to_cpu(gdp->bg_inode_table) + sbi->s_itb_per_group - 1 >
 668                     last_block)
 669                 {
 670                         ext2_error (sb, "ext2_check_descriptors",
 671                                     "Inode table for group %d"
 672                                     " not in group (block %lu)!",
 673                                     i, (unsigned long) le32_to_cpu(gdp->bg_inode_table));
 674                         return 0;
 675                 }
 676         }
 677         return 1;
 678 }
 679
 680 /*
 681  * Maximal file size.  There is a direct, and {,double-,triple-}indirect
 682  * block limit, and also a limit of (2^32 - 1) 512-byte sectors in i_blocks.
 683  * We need to be 1 filesystem block less than the 2^32 sector limit.
 684  */
 685 static loff_t ext2_max_size(int bits)
 686 {
 687         loff_t res = EXT2_NDIR_BLOCKS;
 688         int meta_blocks;
 689         loff_t upper_limit;
 690
 691         /* This is calculated to be the largest file size for a
 692          * dense, file such that the total number of
 693          * sectors in the file, including data and all indirect blocks,
 694          * does not exceed 2^32 -1
 695          * __u32 i_blocks representing the total number of
 696          * 512 bytes blocks of the file
 697          */
 698         upper_limit = (1LL << 32) - 1;
 699
 700         /* total blocks in file system block size */
 701         upper_limit >>= (bits - 9);
 702
 703
 704         /* indirect blocks */
 705         meta_blocks = 1;
 706         /* double indirect blocks */
 707         meta_blocks += 1 + (1LL << (bits-2));
 708         /* tripple indirect blocks */
 709         meta_blocks += 1 + (1LL << (bits-2)) + (1LL << (2*(bits-2)));
 710
 711         upper_limit -= meta_blocks;
 712         upper_limit <<= bits;
 713
 714         res += 1LL << (bits-2);
 715         res += 1LL << (2*(bits-2));
 716         res += 1LL << (3*(bits-2));
 717         res <<= bits;
 718         if (res > upper_limit)
 719                 res = upper_limit;
 720
 721         if (res > MAX_LFS_FILESIZE)
 722                 res = MAX_LFS_FILESIZE;
 723
 724         return res;
 725 }
 726
 727 static unsigned long descriptor_loc(struct super_block *sb,
 728                                     unsigned long logic_sb_block,
 729                                     int nr)
 730 {
 731         struct ext2_sb_info *sbi = EXT2_SB(sb);
 732         unsigned long bg, first_meta_bg;
 733         int has_super = 0;
 734
 735         first_meta_bg = le32_to_cpu(sbi->s_es->s_first_meta_bg);
 736
 737         if (!EXT2_HAS_INCOMPAT_FEATURE(sb, EXT2_FEATURE_INCOMPAT_META_BG) ||
 738             nr < first_meta_bg)
 739                 return (logic_sb_block + nr + 1);
 740         bg = sbi->s_desc_per_block * nr;
 741         if (ext2_bg_has_super(sb, bg))
 742                 has_super = 1;
 743
 744         return ext2_group_first_block_no(sb, bg) + has_super;
 745 }
 746
 747 static int ext2_fill_super(struct super_block *sb, void *data, int silent)
 748 {
 749         struct buffer_head * bh;
 750         struct ext2_sb_info * sbi;
 751         struct ext2_super_block * es;
 752         struct inode *root;
 753         unsigned long block;
 754         unsigned long sb_block = get_sb_block(&data);
 755         unsigned long logic_sb_block;
 756         unsigned long offset = 0;
 757         unsigned long def_mount_opts;
 758         long ret = -EINVAL;
 759         int blocksize = BLOCK_SIZE;
 760         int db_count;
 761         int i, j;
 762         __le32 features;
 763         int err;
 764
 765         sbi = kzalloc(sizeof(*sbi), GFP_KERNEL);
 766         if (!sbi)
 767                 return -ENOMEM;
 768
 769         sbi->s_blockgroup_lock =
 770                 kzalloc(sizeof(struct blockgroup_lock), GFP_KERNEL);
 771         if (!sbi->s_blockgroup_lock) {
 772                 kfree(sbi);
 773                 return -ENOMEM;
 774         }
 775         sb->s_fs_info = sbi;
 776         sbi->s_sb_block = sb_block;
 777
 778         spin_lock_init(&sbi->s_lock);
 779
 780         /*
 781          * See what the current blocksize for the device is, and
 782          * use that as the blocksize.  Otherwise (or if the blocksize
 783          * is smaller than the default) use the default.
 784          * This is important for devices that have a hardware
 785          * sectorsize that is larger than the default.
 786          */
 787         blocksize = sb_min_blocksize(sb, BLOCK_SIZE);
 788         if (!blocksize) {
 789                 ext2_msg(sb, KERN_ERR, "error: unable to set blocksize");
 790                 goto failed_sbi;
 791         }
 792
 793         /*
 794          * If the superblock doesn't start on a hardware sector boundary,
 795          * calculate the offset.
 796          */
 797         if (blocksize != BLOCK_SIZE) {
 798                 logic_sb_block = (sb_block*BLOCK_SIZE) / blocksize;
 799                 offset = (sb_block*BLOCK_SIZE) % blocksize;
 800         } else {
 801                 logic_sb_block = sb_block;
 802         }
 803
 804         if (!(bh = sb_bread(sb, logic_sb_block))) {
 805                 ext2_msg(sb, KERN_ERR, "error: unable to read superblock");
 806                 goto failed_sbi;
 807         }
 808         /*
 809          * Note: s_es must be initialized as soon as possible because
 810          *       some ext2 macro-instructions depend on its value
 811          */
 812         es = (struct ext2_super_block *) (((char *)bh->b_data) + offset);
 813         sbi->s_es = es;
 814         sb->s_magic = le16_to_cpu(es->s_magic);
 815
 816         if (sb->s_magic != EXT2_SUPER_MAGIC)
 817                 goto cantfind_ext2;
 818
 819         /* Set defaults before we parse the mount options */
 820         def_mount_opts = le32_to_cpu(es->s_default_mount_opts);
 821         if (def_mount_opts & EXT2_DEFM_DEBUG)
 822                 set_opt(sbi->s_mount_opt, DEBUG);
 823         if (def_mount_opts & EXT2_DEFM_BSDGROUPS)
 824                 set_opt(sbi->s_mount_opt, GRPID);
 825         if (def_mount_opts & EXT2_DEFM_UID16)
 826                 set_opt(sbi->s_mount_opt, NO_UID32);
 827 #ifdef CONFIG_EXT2_FS_XATTR
 828         if (def_mount_opts & EXT2_DEFM_XATTR_USER)
 829                 set_opt(sbi->s_mount_opt, XATTR_USER);
 830 #endif
 831 #ifdef CONFIG_EXT2_FS_POSIX_ACL
 832         if (def_mount_opts & EXT2_DEFM_ACL)
 833                 set_opt(sbi->s_mount_opt, POSIX_ACL);
 834 #endif
 835
 836         if (le16_to_cpu(sbi->s_es->s_errors) == EXT2_ERRORS_PANIC)
 837                 set_opt(sbi->s_mount_opt, ERRORS_PANIC);
 838         else if (le16_to_cpu(sbi->s_es->s_errors) == EXT2_ERRORS_CONTINUE)
 839                 set_opt(sbi->s_mount_opt, ERRORS_CONT);
 840         else
 841                 set_opt(sbi->s_mount_opt, ERRORS_RO);
 842
 843         sbi->s_resuid = le16_to_cpu(es->s_def_resuid);
 844         sbi->s_resgid = le16_to_cpu(es->s_def_resgid);
 845
 846         set_opt(sbi->s_mount_opt, RESERVATION);
 847
 848         if (!parse_options((char *) data, sb))
 849                 goto failed_mount;
 850
 851         sb->s_flags = (sb->s_flags & ~MS_POSIXACL) |
 852                 ((EXT2_SB(sb)->s_mount_opt & EXT2_MOUNT_POSIX_ACL) ?
 853                  MS_POSIXACL : 0);
 854
 855         ext2_xip_verify_sb(sb); /* see if bdev supports xip, unset
 856                                     EXT2_MOUNT_XIP if not */
 857
 858         if (le32_to_cpu(es->s_rev_level) == EXT2_GOOD_OLD_REV &&
 859             (EXT2_HAS_COMPAT_FEATURE(sb, ~0U) ||
 860              EXT2_HAS_RO_COMPAT_FEATURE(sb, ~0U) ||
 861              EXT2_HAS_INCOMPAT_FEATURE(sb, ~0U)))
 862                 ext2_msg(sb, KERN_WARNING,
 863                         "warning: feature flags set on rev 0 fs, "
 864                         "running e2fsck is recommended");
 865         /*
 866          * Check feature flags regardless of the revision level, since we
 867          * previously didn't change the revision level when setting the flags,
 868          * so there is a chance incompat flags are set on a rev 0 filesystem.
 869          */
 870         features = EXT2_HAS_INCOMPAT_FEATURE(sb, ~EXT2_FEATURE_INCOMPAT_SUPP);
 871         if (features) {
 872                 ext2_msg(sb, KERN_ERR,  "error: couldn't mount because of "
 873                        "unsupported optional features (%x)",
 874                         le32_to_cpu(features));
 875                 goto failed_mount;
 876         }
 877         if (!(sb->s_flags & MS_RDONLY) &&
 878             (features = EXT2_HAS_RO_COMPAT_FEATURE(sb, ~EXT2_FEATURE_RO_COMPAT_SUPP))){
 879                 ext2_msg(sb, KERN_ERR, "error: couldn't mount RDWR because of "
 880                        "unsupported optional features (%x)",
 881                        le32_to_cpu(features));
 882                 goto failed_mount;
 883         }
 884
 885         blocksize = BLOCK_SIZE << le32_to_cpu(sbi->s_es->s_log_block_size);
 886
 887         if (ext2_use_xip(sb) && blocksize != PAGE_SIZE) {
 888                 if (!silent)
 889                         ext2_msg(sb, KERN_ERR,
 890                                 "error: unsupported blocksize for xip");
 891                 goto failed_mount;
 892         }
 893
 894         /* If the blocksize doesn't match, re-read the thing.. */
 895         if (sb->s_blocksize != blocksize) {
 896                 brelse(bh);
 897
 898                 if (!sb_set_blocksize(sb, blocksize)) {
 899                         ext2_msg(sb, KERN_ERR, "error: blocksize is too small");
 900                         goto failed_sbi;
 901                 }
 902
 903                 logic_sb_block = (sb_block*BLOCK_SIZE) / blocksize;
 904                 offset = (sb_block*BLOCK_SIZE) % blocksize;
 905                 bh = sb_bread(sb, logic_sb_block);
 906                 if(!bh) {
 907                         ext2_msg(sb, KERN_ERR, "error: couldn't read"
 908                                 "superblock on 2nd try");
 909                         goto failed_sbi;
 910                 }
 911                 es = (struct ext2_super_block *) (((char *)bh->b_data) + offset);
 912                 sbi->s_es = es;
 913                 if (es->s_magic != cpu_to_le16(EXT2_SUPER_MAGIC)) {
 914                         ext2_msg(sb, KERN_ERR, "error: magic mismatch");
 915                         goto failed_mount;
 916                 }
 917         }
 918
 919         sb->s_maxbytes = ext2_max_size(sb->s_blocksize_bits);
 920
 921         if (le32_to_cpu(es->s_rev_level) == EXT2_GOOD_OLD_REV) {
 922                 sbi->s_inode_size = EXT2_GOOD_OLD_INODE_SIZE;
 923                 sbi->s_first_ino = EXT2_GOOD_OLD_FIRST_INO;
 924         } else {
 925                 sbi->s_inode_size = le16_to_cpu(es->s_inode_size);
 926                 sbi->s_first_ino = le32_to_cpu(es->s_first_ino);
 927                 if ((sbi->s_inode_size < EXT2_GOOD_OLD_INODE_SIZE) ||
 928                     !is_power_of_2(sbi->s_inode_size) ||
 929                     (sbi->s_inode_size > blocksize)) {
 930                         ext2_msg(sb, KERN_ERR,
 931                                 "error: unsupported inode size: %d",
 932                                 sbi->s_inode_size);
 933                         goto failed_mount;
 934                 }
 935         }
 936
 937         sbi->s_frag_size = EXT2_MIN_FRAG_SIZE <<
 938                                    le32_to_cpu(es->s_log_frag_size);
 939         if (sbi->s_frag_size == 0)
 940                 goto cantfind_ext2;
 941         sbi->s_frags_per_block = sb->s_blocksize / sbi->s_frag_size;
 942
 943         sbi->s_blocks_per_group = le32_to_cpu(es->s_blocks_per_group);
 944         sbi->s_frags_per_group = le32_to_cpu(es->s_frags_per_group);
 945         sbi->s_inodes_per_group = le32_to_cpu(es->s_inodes_per_group);
 946
 947         if (EXT2_INODE_SIZE(sb) == 0)
 948                 goto cantfind_ext2;
 949         sbi->s_inodes_per_block = sb->s_blocksize / EXT2_INODE_SIZE(sb);
 950         if (sbi->s_inodes_per_block == 0 || sbi->s_inodes_per_group == 0)
 951                 goto cantfind_ext2;
 952         sbi->s_itb_per_group = sbi->s_inodes_per_group /
 953                                         sbi->s_inodes_per_block;
 954         sbi->s_desc_per_block = sb->s_blocksize /
 955                                         sizeof (struct ext2_group_desc);
 956         sbi->s_sbh = bh;
 957         sbi->s_mount_state = le16_to_cpu(es->s_state);
 958         sbi->s_addr_per_block_bits =
 959                 ilog2 (EXT2_ADDR_PER_BLOCK(sb));
 960         sbi->s_desc_per_block_bits =
 961                 ilog2 (EXT2_DESC_PER_BLOCK(sb));
 962
 963         if (sb->s_magic != EXT2_SUPER_MAGIC)
 964                 goto cantfind_ext2;
 965
 966         if (sb->s_blocksize != bh->b_size) {
 967                 if (!silent)
 968                         ext2_msg(sb, KERN_ERR, "error: unsupported blocksize");
 969                 goto failed_mount;
 970         }
 971
 972         if (sb->s_blocksize != sbi->s_frag_size) {
 973                 ext2_msg(sb, KERN_ERR,
 974                         "error: fragsize %lu != blocksize %lu"
 975                         "(not supported yet)",
 976                         sbi->s_frag_size, sb->s_blocksize);
 977                 goto failed_mount;
 978         }
 979
 980         if (sbi->s_blocks_per_group > sb->s_blocksize * 8) {
 981                 ext2_msg(sb, KERN_ERR,
 982                         "error: #blocks per group too big: %lu",
 983                         sbi->s_blocks_per_group);
 984                 goto failed_mount;
 985         }
 986         if (sbi->s_frags_per_group > sb->s_blocksize * 8) {
 987                 ext2_msg(sb, KERN_ERR,
 988                         "error: #fragments per group too big: %lu",
 989                         sbi->s_frags_per_group);
 990                 goto failed_mount;
 991         }
 992         if (sbi->s_inodes_per_group > sb->s_blocksize * 8) {
 993                 ext2_msg(sb, KERN_ERR,
 994                         "error: #inodes per group too big: %lu",
 995                         sbi->s_inodes_per_group);
 996                 goto failed_mount;
 997         }
 998
 999         if (EXT2_BLOCKS_PER_GROUP(sb) == 0)
1000                 goto cantfind_ext2;
1001         sbi->s_groups_count = ((le32_to_cpu(es->s_blocks_count) -
1002                                 le32_to_cpu(es->s_first_data_block) - 1)
1003                                         / EXT2_BLOCKS_PER_GROUP(sb)) + 1;
1004         db_count = (sbi->s_groups_count + EXT2_DESC_PER_BLOCK(sb) - 1) /
1005                    EXT2_DESC_PER_BLOCK(sb);
1006         sbi->s_group_desc = kmalloc (db_count * sizeof (struct buffer_head *), GFP_KERNEL);
1007         if (sbi->s_group_desc == NULL) {
1008                 ext2_msg(sb, KERN_ERR, "error: not enough memory");
1009                 goto failed_mount;
1010         }
1011         bgl_lock_init(sbi->s_blockgroup_lock);
1012         sbi->s_debts = kcalloc(sbi->s_groups_count, sizeof(*sbi->s_debts), GFP_KERNEL);
1013         if (!sbi->s_debts) {
1014                 ext2_msg(sb, KERN_ERR, "error: not enough memory");
1015                 goto failed_mount_group_desc;
1016         }
1017         for (i = 0; i < db_count; i++) {
1018                 block = descriptor_loc(sb, logic_sb_block, i);
1019                 sbi->s_group_desc[i] = sb_bread(sb, block);
1020                 if (!sbi->s_group_desc[i]) {
1021                         for (j = 0; j < i; j++)
1022                                 brelse (sbi->s_group_desc[j]);
1023                         ext2_msg(sb, KERN_ERR,
1024                                 "error: unable to read group descriptors");
1025                         goto failed_mount_group_desc;
1026                 }
1027         }
1028         if (!ext2_check_descriptors (sb)) {
1029                 ext2_msg(sb, KERN_ERR, "group descriptors corrupted");
1030                 goto failed_mount2;
1031         }
1032         sbi->s_gdb_count = db_count;
1033         get_random_bytes(&sbi->s_next_generation, sizeof(u32));
1034         spin_lock_init(&sbi->s_next_gen_lock);
1035
1036         /* per fileystem reservation list head & lock */
1037         spin_lock_init(&sbi->s_rsv_window_lock);
1038         sbi->s_rsv_window_root = RB_ROOT;
1039         /*
1040          * Add a single, static dummy reservation to the start of the
1041          * reservation window list --- it gives us a placeholder for
1042          * append-at-start-of-list which makes the allocation logic
1043          * _much_ simpler.
1044          */
1045         sbi->s_rsv_window_head.rsv_start = EXT2_RESERVE_WINDOW_NOT_ALLOCATED;
1046         sbi->s_rsv_window_head.rsv_end = EXT2_RESERVE_WINDOW_NOT_ALLOCATED;
1047         sbi->s_rsv_window_head.rsv_alloc_hit = 0;
1048         sbi->s_rsv_window_head.rsv_goal_size = 0;
1049         ext2_rsv_window_add(sb, &sbi->s_rsv_window_head);
1050
1051         err = percpu_counter_init(&sbi->s_freeblocks_counter,
1052                                 ext2_count_free_blocks(sb));
1053         if (!err) {
1054                 err = percpu_counter_init(&sbi->s_freeinodes_counter,
1055                                 ext2_count_free_inodes(sb));
1056         }
1057         if (!err) {
1058                 err = percpu_counter_init(&sbi->s_dirs_counter,
1059                                 ext2_count_dirs(sb));
1060         }
1061         if (err) {
1062                 ext2_msg(sb, KERN_ERR, "error: insufficient memory");
1063                 goto failed_mount3;
1064         }
1065         /*
1066          * set up enough so that it can read an inode
1067          */
1068         sb->s_op = &ext2_sops;
1069         sb->s_export_op = &ext2_export_ops;
1070         sb->s_xattr = ext2_xattr_handlers;
1071         root = ext2_iget(sb, EXT2_ROOT_INO);
1072         if (IS_ERR(root)) {
1073                 ret = PTR_ERR(root);
1074                 goto failed_mount3;
1075         }
1076         if (!S_ISDIR(root->i_mode) || !root->i_blocks || !root->i_size) {
1077                 iput(root);
1078                 ext2_msg(sb, KERN_ERR, "error: corrupt root inode, run e2fsck");
1079                 goto failed_mount3;
1080         }
1081
1082         sb->s_root = d_alloc_root(root);
1083         if (!sb->s_root) {
1084                 iput(root);
1085                 ext2_msg(sb, KERN_ERR, "error: get root inode failed");
1086                 ret = -ENOMEM;
1087                 goto failed_mount3;
1088         }
1089         if (EXT2_HAS_COMPAT_FEATURE(sb, EXT3_FEATURE_COMPAT_HAS_JOURNAL))
1090                 ext2_msg(sb, KERN_WARNING,
1091                         "warning: mounting ext3 filesystem as ext2");
1092         if (ext2_setup_super (sb, es, sb->s_flags & MS_RDONLY))
1093                 sb->s_flags |= MS_RDONLY;
1094         ext2_write_super(sb);
1095         return 0;
1096
1097 cantfind_ext2:
1098         if (!silent)
1099                 ext2_msg(sb, KERN_ERR,
1100                         "error: can't find an ext2 filesystem on dev %s.",
1101                         sb->s_id);
1102         goto failed_mount;
1103 failed_mount3:
1104         percpu_counter_destroy(&sbi->s_freeblocks_counter);
1105         percpu_counter_destroy(&sbi->s_freeinodes_counter);
1106         percpu_counter_destroy(&sbi->s_dirs_counter);
1107 failed_mount2:
1108         for (i = 0; i < db_count; i++)
1109                 brelse(sbi->s_group_desc[i]);
1110 failed_mount_group_desc:
1111         kfree(sbi->s_group_desc);
1112         kfree(sbi->s_debts);
1113 failed_mount:
1114         brelse(bh);
1115 failed_sbi:
1116         sb->s_fs_info = NULL;
1117         kfree(sbi->s_blockgroup_lock);
1118         kfree(sbi);
1119         return ret;
1120 }
1121
1122 static void ext2_clear_super_error(struct super_block *sb)
1123 {
1124         struct buffer_head *sbh = EXT2_SB(sb)->s_sbh;
1125
1126         if (buffer_write_io_error(sbh)) {
1127                 /*
1128                  * Oh, dear.  A previous attempt to write the
1129                  * superblock failed.  This could happen because the
1130                  * USB device was yanked out.  Or it could happen to
1131                  * be a transient write error and maybe the block will
1132                  * be remapped.  Nothing we can do but to retry the
1133                  * write and hope for the best.
1134                  */
1135                 ext2_msg(sb, KERN_ERR,
1136                        "previous I/O error to superblock detected\n");
1137                 clear_buffer_write_io_error(sbh);
1138                 set_buffer_uptodate(sbh);
1139         }
1140 }
1141
1142 static void ext2_sync_super(struct super_block *sb, struct ext2_super_block *es,
1143                             int wait)
1144 {
1145         ext2_clear_super_error(sb);
1146         spin_lock(&EXT2_SB(sb)->s_lock);
1147         es->s_free_blocks_count = cpu_to_le32(ext2_count_free_blocks(sb));
1148         es->s_free_inodes_count = cpu_to_le32(ext2_count_free_inodes(sb));
1149         es->s_wtime = cpu_to_le32(get_seconds());
1150         /* unlock before we do IO */
1151         spin_unlock(&EXT2_SB(sb)->s_lock);
1152         mark_buffer_dirty(EXT2_SB(sb)->s_sbh);
1153         if (wait)
1154                 sync_dirty_buffer(EXT2_SB(sb)->s_sbh);
1155         sb->s_dirt = 0;
1156 }
1157
1158 /*
1159  * In the second extended file system, it is not necessary to
1160  * write the super block since we use a mapping of the
1161  * disk super block in a buffer.
1162  *
1163  * However, this function is still used to set the fs valid
1164  * flags to 0.  We need to set this flag to 0 since the fs
1165  * may have been checked while mounted and e2fsck may have
1166  * set s_state to EXT2_VALID_FS after some corrections.
1167  */
1168 static int ext2_sync_fs(struct super_block *sb, int wait)
1169 {
1170         struct ext2_sb_info *sbi = EXT2_SB(sb);
1171         struct ext2_super_block *es = EXT2_SB(sb)->s_es;
1172
1173         lock_kernel();
1174         spin_lock(&sbi->s_lock);
1175         if (es->s_state & cpu_to_le16(EXT2_VALID_FS)) {
1176                 ext2_debug("setting valid to 0\n");
1177                 es->s_state &= cpu_to_le16(~EXT2_VALID_FS);
1178         }
1179         spin_unlock(&sbi->s_lock);
1180         ext2_sync_super(sb, es, wait);
1181         unlock_kernel();
1182
1183         return 0;
1184 }
1185
1186
1187 void ext2_write_super(struct super_block *sb)
1188 {
1189         if (!(sb->s_flags & MS_RDONLY))
1190                 ext2_sync_fs(sb, 1);
1191         else
1192                 sb->s_dirt = 0;
1193 }
1194
1195 static int ext2_remount (struct super_block * sb, int * flags, char * data)
1196 {
1197         struct ext2_sb_info * sbi = EXT2_SB(sb);
1198         struct ext2_super_block * es;
1199         unsigned long old_mount_opt = sbi->s_mount_opt;
1200         struct ext2_mount_options old_opts;
1201         unsigned long old_sb_flags;
1202         int err;
1203
1204         lock_kernel();
1205         spin_lock(&sbi->s_lock);
1206
1207         /* Store the old options */
1208         old_sb_flags = sb->s_flags;
1209         old_opts.s_mount_opt = sbi->s_mount_opt;
1210         old_opts.s_resuid = sbi->s_resuid;
1211         old_opts.s_resgid = sbi->s_resgid;
1212
1213         /*
1214          * Allow the "check" option to be passed as a remount option.
1215          */
1216         if (!parse_options(data, sb)) {
1217                 err = -EINVAL;
1218                 goto restore_opts;
1219         }
1220
1221         sb->s_flags = (sb->s_flags & ~MS_POSIXACL) |
1222                 ((sbi->s_mount_opt & EXT2_MOUNT_POSIX_ACL) ? MS_POSIXACL : 0);
1223
1224         ext2_xip_verify_sb(sb); /* see if bdev supports xip, unset
1225                                     EXT2_MOUNT_XIP if not */
1226
1227         if ((ext2_use_xip(sb)) && (sb->s_blocksize != PAGE_SIZE)) {
1228                 ext2_msg(sb, KERN_WARNING,
1229                         "warning: unsupported blocksize for xip");
1230                 err = -EINVAL;
1231                 goto restore_opts;
1232         }
1233
1234         es = sbi->s_es;
1235         if (((sbi->s_mount_opt & EXT2_MOUNT_XIP) !=
1236             (old_mount_opt & EXT2_MOUNT_XIP)) &&
1237             invalidate_inodes(sb)) {
1238                 ext2_msg(sb, KERN_WARNING, "warning: refusing change of "
1239                          "xip flag with busy inodes while remounting");
1240                 sbi->s_mount_opt &= ~EXT2_MOUNT_XIP;
1241                 sbi->s_mount_opt |= old_mount_opt & EXT2_MOUNT_XIP;
1242         }
1243         if ((*flags & MS_RDONLY) == (sb->s_flags & MS_RDONLY)) {
1244                 spin_unlock(&sbi->s_lock);
1245                 unlock_kernel();
1246                 return 0;
1247         }
1248         if (*flags & MS_RDONLY) {
1249                 if (le16_to_cpu(es->s_state) & EXT2_VALID_FS ||
1250                     !(sbi->s_mount_state & EXT2_VALID_FS)) {
1251                         spin_unlock(&sbi->s_lock);
1252                         unlock_kernel();
1253                         return 0;
1254                 }
1255                 /*
1256                  * OK, we are remounting a valid rw partition rdonly, so set
1257                  * the rdonly flag and then mark the partition as valid again.
1258                  */
1259                 es->s_state = cpu_to_le16(sbi->s_mount_state);
1260                 es->s_mtime = cpu_to_le32(get_seconds());
1261                 spin_unlock(&sbi->s_lock);
1262                 ext2_sync_super(sb, es, 1);
1263         } else {
1264                 __le32 ret = EXT2_HAS_RO_COMPAT_FEATURE(sb,
1265                                                ~EXT2_FEATURE_RO_COMPAT_SUPP);
1266                 if (ret) {
1267                         ext2_msg(sb, KERN_WARNING,
1268                                 "warning: couldn't remount RDWR because of "
1269                                 "unsupported optional features (%x).",
1270                                 le32_to_cpu(ret));
1271                         err = -EROFS;
1272                         goto restore_opts;
1273                 }
1274                 /*
1275                  * Mounting a RDONLY partition read-write, so reread and
1276                  * store the current valid flag.  (It may have been changed
1277                  * by e2fsck since we originally mounted the partition.)
1278                  */
1279                 sbi->s_mount_state = le16_to_cpu(es->s_state);
1280                 if (!ext2_setup_super (sb, es, 0))
1281                         sb->s_flags &= ~MS_RDONLY;
1282                 spin_unlock(&sbi->s_lock);
1283                 ext2_write_super(sb);
1284         }
1285         unlock_kernel();
1286         return 0;
1287 restore_opts:
1288         sbi->s_mount_opt = old_opts.s_mount_opt;
1289         sbi->s_resuid = old_opts.s_resuid;
1290         sbi->s_resgid = old_opts.s_resgid;
1291         sb->s_flags = old_sb_flags;
1292         spin_unlock(&sbi->s_lock);
1293         unlock_kernel();
1294         return err;
1295 }
1296
1297 static int ext2_statfs (struct dentry * dentry, struct kstatfs * buf)
1298 {
1299         struct super_block *sb = dentry->d_sb;
1300         struct ext2_sb_info *sbi = EXT2_SB(sb);
1301         struct ext2_super_block *es = sbi->s_es;
1302         u64 fsid;
1303
1304         spin_lock(&sbi->s_lock);
1305
1306         if (test_opt (sb, MINIX_DF))
1307                 sbi->s_overhead_last = 0;
1308         else if (sbi->s_blocks_last != le32_to_cpu(es->s_blocks_count)) {
1309                 unsigned long i, overhead = 0;
1310                 smp_rmb();
1311
1312                 /*
1313                  * Compute the overhead (FS structures). This is constant
1314                  * for a given filesystem unless the number of block groups
1315                  * changes so we cache the previous value until it does.
1316                  */
1317
1318                 /*
1319                  * All of the blocks before first_data_block are
1320                  * overhead
1321                  */
1322                 overhead = le32_to_cpu(es->s_first_data_block);
1323
1324                 /*
1325                  * Add the overhead attributed to the superblock and
1326                  * block group descriptors.  If the sparse superblocks
1327                  * feature is turned on, then not all groups have this.
1328                  */
1329                 for (i = 0; i < sbi->s_groups_count; i++)
1330                         overhead += ext2_bg_has_super(sb, i) +
1331                                 ext2_bg_num_gdb(sb, i);
1332
1333                 /*
1334                  * Every block group has an inode bitmap, a block
1335                  * bitmap, and an inode table.
1336                  */
1337                 overhead += (sbi->s_groups_count *
1338                              (2 + sbi->s_itb_per_group));
1339                 sbi->s_overhead_last = overhead;
1340                 smp_wmb();
1341                 sbi->s_blocks_last = le32_to_cpu(es->s_blocks_count);
1342         }
1343
1344         buf->f_type = EXT2_SUPER_MAGIC;
1345         buf->f_bsize = sb->s_blocksize;
1346         buf->f_blocks = le32_to_cpu(es->s_blocks_count) - sbi->s_overhead_last;
1347         buf->f_bfree = ext2_count_free_blocks(sb);
1348         es->s_free_blocks_count = cpu_to_le32(buf->f_bfree);
1349         buf->f_bavail = buf->f_bfree - le32_to_cpu(es->s_r_blocks_count);
1350         if (buf->f_bfree < le32_to_cpu(es->s_r_blocks_count))
1351                 buf->f_bavail = 0;
1352         buf->f_files = le32_to_cpu(es->s_inodes_count);
1353         buf->f_ffree = ext2_count_free_inodes(sb);
1354         es->s_free_inodes_count = cpu_to_le32(buf->f_ffree);
1355         buf->f_namelen = EXT2_NAME_LEN;
1356         fsid = le64_to_cpup((void *)es->s_uuid) ^
1357                le64_to_cpup((void *)es->s_uuid + sizeof(u64));
1358         buf->f_fsid.val[0] = fsid & 0xFFFFFFFFUL;
1359         buf->f_fsid.val[1] = (fsid >> 32) & 0xFFFFFFFFUL;
1360         spin_unlock(&sbi->s_lock);
1361         return 0;
1362 }
1363
1364 static int ext2_get_sb(struct file_system_type *fs_type,
1365         int flags, const char *dev_name, void *data, struct vfsmount *mnt)
1366 {
1367         return get_sb_bdev(fs_type, flags, dev_name, data, ext2_fill_super, mnt);
1368 }
1369
1370 #ifdef CONFIG_QUOTA
1371
1372 /* Read data from quotafile - avoid pagecache and such because we cannot afford
1373  * acquiring the locks... As quota files are never truncated and quota code
1374  * itself serializes the operations (and noone else should touch the files)
1375  * we don't have to be afraid of races */
1376 static ssize_t ext2_quota_read(struct super_block *sb, int type, char *data,
1377                                size_t len, loff_t off)
1378 {
1379         struct inode *inode = sb_dqopt(sb)->files[type];
1380         sector_t blk = off >> EXT2_BLOCK_SIZE_BITS(sb);
1381         int err = 0;
1382         int offset = off & (sb->s_blocksize - 1);
1383         int tocopy;
1384         size_t toread;
1385         struct buffer_head tmp_bh;
1386         struct buffer_head *bh;
1387         loff_t i_size = i_size_read(inode);
1388
1389         if (off > i_size)
1390                 return 0;
1391         if (off+len > i_size)
1392                 len = i_size-off;
1393         toread = len;
1394         while (toread > 0) {
1395                 tocopy = sb->s_blocksize - offset < toread ?
1396                                 sb->s_blocksize - offset : toread;
1397
1398                 tmp_bh.b_state = 0;
1399                 tmp_bh.b_size = sb->s_blocksize;
1400                 err = ext2_get_block(inode, blk, &tmp_bh, 0);
1401                 if (err < 0)
1402                         return err;
1403                 if (!buffer_mapped(&tmp_bh))    /* A hole? */
1404                         memset(data, 0, tocopy);
1405                 else {
1406                         bh = sb_bread(sb, tmp_bh.b_blocknr);
1407                         if (!bh)
1408                                 return -EIO;
1409                         memcpy(data, bh->b_data+offset, tocopy);
1410                         brelse(bh);
1411                 }
1412                 offset = 0;
1413                 toread -= tocopy;
1414                 data += tocopy;
1415                 blk++;
1416         }
1417         return len;
1418 }
1419
1420 /* Write to quotafile */
1421 static ssize_t ext2_quota_write(struct super_block *sb, int type,
1422                                 const char *data, size_t len, loff_t off)
1423 {
1424         struct inode *inode = sb_dqopt(sb)->files[type];
1425         sector_t blk = off >> EXT2_BLOCK_SIZE_BITS(sb);
1426         int err = 0;
1427         int offset = off & (sb->s_blocksize - 1);
1428         int tocopy;
1429         size_t towrite = len;
1430         struct buffer_head tmp_bh;
1431         struct buffer_head *bh;
1432
1433         mutex_lock_nested(&inode->i_mutex, I_MUTEX_QUOTA);
1434         while (towrite > 0) {
1435                 tocopy = sb->s_blocksize - offset < towrite ?
1436                                 sb->s_blocksize - offset : towrite;
1437
1438                 tmp_bh.b_state = 0;
1439                 err = ext2_get_block(inode, blk, &tmp_bh, 1);
1440                 if (err < 0)
1441                         goto out;
1442                 if (offset || tocopy != EXT2_BLOCK_SIZE(sb))
1443                         bh = sb_bread(sb, tmp_bh.b_blocknr);
1444                 else
1445                         bh = sb_getblk(sb, tmp_bh.b_blocknr);
1446                 if (!bh) {
1447                         err = -EIO;
1448                         goto out;
1449                 }
1450                 lock_buffer(bh);
1451                 memcpy(bh->b_data+offset, data, tocopy);
1452                 flush_dcache_page(bh->b_page);
1453                 set_buffer_uptodate(bh);
1454                 mark_buffer_dirty(bh);
1455                 unlock_buffer(bh);
1456                 brelse(bh);
1457                 offset = 0;
1458                 towrite -= tocopy;
1459                 data += tocopy;
1460                 blk++;
1461         }
1462 out:
1463         if (len == towrite) {
1464                 mutex_unlock(&inode->i_mutex);
1465                 return err;
1466         }
1467         if (inode->i_size < off+len-towrite)
1468                 i_size_write(inode, off+len-towrite);
1469         inode->i_version++;
1470         inode->i_mtime = inode->i_ctime = CURRENT_TIME;
1471         mark_inode_dirty(inode);
1472         mutex_unlock(&inode->i_mutex);
1473         return len - towrite;
1474 }
1475
1476 #endif
1477
1478 static struct file_system_type ext2_fs_type = {
1479         .owner          = THIS_MODULE,
1480         .name           = "ext2",
1481         .get_sb         = ext2_get_sb,
1482         .kill_sb        = kill_block_super,
1483         .fs_flags       = FS_REQUIRES_DEV,
1484 };
1485
1486 static int __init init_ext2_fs(void)
1487 {
1488         int err = init_ext2_xattr();
1489         if (err)
1490                 return err;
1491         err = init_inodecache();
1492         if (err)
1493                 goto out1;
1494         err = register_filesystem(&ext2_fs_type);
1495         if (err)
1496                 goto out;
1497         return 0;
1498 out:
1499         destroy_inodecache();
1500 out1:
1501         exit_ext2_xattr();
1502         return err;
1503 }
1504
1505 static void __exit exit_ext2_fs(void)
1506 {
1507         unregister_filesystem(&ext2_fs_type);
1508         destroy_inodecache();
1509         exit_ext2_xattr();
1510 }
1511
1512 module_init(init_ext2_fs)
1513 module_exit(exit_ext2_fs)