fs/ext2/super.c

   1 /*
   2  *  linux/fs/ext2/super.c
   3  *
   4  * Copyright (C) 1992, 1993, 1994, 1995
   5  * Remy Card (card@masi.ibp.fr)
   6  * Laboratoire MASI - Institut Blaise Pascal
   7  * Universite Pierre et Marie Curie (Paris VI)
   8  *
   9  *  from
  10  *
  11  *  linux/fs/minix/inode.c
  12  *
  13  *  Copyright (C) 1991, 1992  Linus Torvalds
  14  *
  15  *  Big-endian to little-endian byte-swapping/bitmaps by
  16  *        David S. Miller (davem@caip.rutgers.edu), 1995
  17  */
  18
  19 #include <linux/module.h>
  20 #include <linux/string.h>
  21 #include <linux/fs.h>
  22 #include <linux/slab.h>
  23 #include <linux/init.h>
  24 #include <linux/blkdev.h>
  25 #include <linux/parser.h>
  26 #include <linux/random.h>
  27 #include <linux/buffer_head.h>
  28 #include <linux/exportfs.h>
  29 #include <linux/smp_lock.h>
  30 #include <linux/vfs.h>
  31 #include <linux/seq_file.h>
  32 #include <linux/mount.h>
  33 #include <linux/log2.h>
  34 #include <linux/quotaops.h>
  35 #include <asm/uaccess.h>
  36 #include "ext2.h"
  37 #include "xattr.h"
  38 #include "acl.h"
  39 #include "xip.h"
  40
  41 static void ext2_sync_super(struct super_block *sb,
  42                             struct ext2_super_block *es);
  43 static int ext2_remount (struct super_block * sb, int * flags, char * data);
  44 static int ext2_statfs (struct dentry * dentry, struct kstatfs * buf);
  45 static int ext2_sync_fs(struct super_block *sb, int wait);
  46
  47 void ext2_error (struct super_block * sb, const char * function,
  48                  const char * fmt, ...)
  49 {
  50         va_list args;
  51         struct ext2_sb_info *sbi = EXT2_SB(sb);
  52         struct ext2_super_block *es = sbi->s_es;
  53
  54         if (!(sb->s_flags & MS_RDONLY)) {
  55                 sbi->s_mount_state |= EXT2_ERROR_FS;
  56                 es->s_state |= cpu_to_le16(EXT2_ERROR_FS);
  57                 ext2_sync_super(sb, es);
  58         }
  59
  60         va_start(args, fmt);
  61         printk(KERN_CRIT "EXT2-fs error (device %s): %s: ",sb->s_id, function);
  62         vprintk(fmt, args);
  63         printk("\n");
  64         va_end(args);
  65
  66         if (test_opt(sb, ERRORS_PANIC))
  67                 panic("EXT2-fs panic from previous error\n");
  68         if (test_opt(sb, ERRORS_RO)) {
  69                 printk("Remounting filesystem read-only\n");
  70                 sb->s_flags |= MS_RDONLY;
  71         }
  72 }
  73
  74 void ext2_warning (struct super_block * sb, const char * function,
  75                    const char * fmt, ...)
  76 {
  77         va_list args;
  78
  79         va_start(args, fmt);
  80         printk(KERN_WARNING "EXT2-fs warning (device %s): %s: ",
  81                sb->s_id, function);
  82         vprintk(fmt, args);
  83         printk("\n");
  84         va_end(args);
  85 }
  86
  87 void ext2_update_dynamic_rev(struct super_block *sb)
  88 {
  89         struct ext2_super_block *es = EXT2_SB(sb)->s_es;
  90
  91         if (le32_to_cpu(es->s_rev_level) > EXT2_GOOD_OLD_REV)
  92                 return;
  93
  94         ext2_warning(sb, __func__,
  95                      "updating to rev %d because of new feature flag, "
  96                      "running e2fsck is recommended",
  97                      EXT2_DYNAMIC_REV);
  98
  99         es->s_first_ino = cpu_to_le32(EXT2_GOOD_OLD_FIRST_INO);
 100         es->s_inode_size = cpu_to_le16(EXT2_GOOD_OLD_INODE_SIZE);
 101         es->s_rev_level = cpu_to_le32(EXT2_DYNAMIC_REV);
 102         /* leave es->s_feature_*compat flags alone */
 103         /* es->s_uuid will be set by e2fsck if empty */
 104
 105         /*
 106          * The rest of the superblock fields should be zero, and if not it
 107          * means they are likely already in use, so leave them alone.  We
 108          * can leave it up to e2fsck to clean up any inconsistencies there.
 109          */
 110 }
 111
 112 static void ext2_put_super (struct super_block * sb)
 113 {
 114         int db_count;
 115         int i;
 116         struct ext2_sb_info *sbi = EXT2_SB(sb);
 117
 118         lock_kernel();
 119
 120         if (sb->s_dirt)
 121                 ext2_write_super(sb);
 122
 123         ext2_xattr_put_super(sb);
 124         if (!(sb->s_flags & MS_RDONLY)) {
 125                 struct ext2_super_block *es = sbi->s_es;
 126
 127                 es->s_state = cpu_to_le16(sbi->s_mount_state);
 128                 ext2_sync_super(sb, es);
 129         }
 130         db_count = sbi->s_gdb_count;
 131         for (i = 0; i < db_count; i++)
 132                 if (sbi->s_group_desc[i])
 133                         brelse (sbi->s_group_desc[i]);
 134         kfree(sbi->s_group_desc);
 135         kfree(sbi->s_debts);
 136         percpu_counter_destroy(&sbi->s_freeblocks_counter);
 137         percpu_counter_destroy(&sbi->s_freeinodes_counter);
 138         percpu_counter_destroy(&sbi->s_dirs_counter);
 139         brelse (sbi->s_sbh);
 140         sb->s_fs_info = NULL;
 141         kfree(sbi->s_blockgroup_lock);
 142         kfree(sbi);
 143
 144         unlock_kernel();
 145 }
 146
 147 static struct kmem_cache * ext2_inode_cachep;
 148
 149 static struct inode *ext2_alloc_inode(struct super_block *sb)
 150 {
 151         struct ext2_inode_info *ei;
 152         ei = (struct ext2_inode_info *)kmem_cache_alloc(ext2_inode_cachep, GFP_KERNEL);
 153         if (!ei)
 154                 return NULL;
 155 #ifdef CONFIG_EXT2_FS_POSIX_ACL
 156         ei->i_acl = EXT2_ACL_NOT_CACHED;
 157         ei->i_default_acl = EXT2_ACL_NOT_CACHED;
 158 #endif
 159         ei->i_block_alloc_info = NULL;
 160         ei->vfs_inode.i_version = 1;
 161         return &ei->vfs_inode;
 162 }
 163
 164 static void ext2_destroy_inode(struct inode *inode)
 165 {
 166         kmem_cache_free(ext2_inode_cachep, EXT2_I(inode));
 167 }
 168
 169 static void init_once(void *foo)
 170 {
 171         struct ext2_inode_info *ei = (struct ext2_inode_info *) foo;
 172
 173         rwlock_init(&ei->i_meta_lock);
 174 #ifdef CONFIG_EXT2_FS_XATTR
 175         init_rwsem(&ei->xattr_sem);
 176 #endif
 177         mutex_init(&ei->truncate_mutex);
 178         inode_init_once(&ei->vfs_inode);
 179 }
 180
 181 static int init_inodecache(void)
 182 {
 183         ext2_inode_cachep = kmem_cache_create("ext2_inode_cache",
 184                                              sizeof(struct ext2_inode_info),
 185                                              0, (SLAB_RECLAIM_ACCOUNT|
 186                                                 SLAB_MEM_SPREAD),
 187                                              init_once);
 188         if (ext2_inode_cachep == NULL)
 189                 return -ENOMEM;
 190         return 0;
 191 }
 192
 193 static void destroy_inodecache(void)
 194 {
 195         kmem_cache_destroy(ext2_inode_cachep);
 196 }
 197
 198 static void ext2_clear_inode(struct inode *inode)
 199 {
 200         struct ext2_block_alloc_info *rsv = EXT2_I(inode)->i_block_alloc_info;
 201 #ifdef CONFIG_EXT2_FS_POSIX_ACL
 202         struct ext2_inode_info *ei = EXT2_I(inode);
 203
 204         if (ei->i_acl && ei->i_acl != EXT2_ACL_NOT_CACHED) {
 205                 posix_acl_release(ei->i_acl);
 206                 ei->i_acl = EXT2_ACL_NOT_CACHED;
 207         }
 208         if (ei->i_default_acl && ei->i_default_acl != EXT2_ACL_NOT_CACHED) {
 209                 posix_acl_release(ei->i_default_acl);
 210                 ei->i_default_acl = EXT2_ACL_NOT_CACHED;
 211         }
 212 #endif
 213         ext2_discard_reservation(inode);
 214         EXT2_I(inode)->i_block_alloc_info = NULL;
 215         if (unlikely(rsv))
 216                 kfree(rsv);
 217 }
 218
 219 static int ext2_show_options(struct seq_file *seq, struct vfsmount *vfs)
 220 {
 221         struct super_block *sb = vfs->mnt_sb;
 222         struct ext2_sb_info *sbi = EXT2_SB(sb);
 223         struct ext2_super_block *es = sbi->s_es;
 224         unsigned long def_mount_opts;
 225
 226         def_mount_opts = le32_to_cpu(es->s_default_mount_opts);
 227
 228         if (sbi->s_sb_block != 1)
 229                 seq_printf(seq, ",sb=%lu", sbi->s_sb_block);
 230         if (test_opt(sb, MINIX_DF))
 231                 seq_puts(seq, ",minixdf");
 232         if (test_opt(sb, GRPID))
 233                 seq_puts(seq, ",grpid");
 234         if (!test_opt(sb, GRPID) && (def_mount_opts & EXT2_DEFM_BSDGROUPS))
 235                 seq_puts(seq, ",nogrpid");
 236         if (sbi->s_resuid != EXT2_DEF_RESUID ||
 237             le16_to_cpu(es->s_def_resuid) != EXT2_DEF_RESUID) {
 238                 seq_printf(seq, ",resuid=%u", sbi->s_resuid);
 239         }
 240         if (sbi->s_resgid != EXT2_DEF_RESGID ||
 241             le16_to_cpu(es->s_def_resgid) != EXT2_DEF_RESGID) {
 242                 seq_printf(seq, ",resgid=%u", sbi->s_resgid);
 243         }
 244         if (test_opt(sb, ERRORS_RO)) {
 245                 int def_errors = le16_to_cpu(es->s_errors);
 246
 247                 if (def_errors == EXT2_ERRORS_PANIC ||
 248                     def_errors == EXT2_ERRORS_CONTINUE) {
 249                         seq_puts(seq, ",errors=remount-ro");
 250                 }
 251         }
 252         if (test_opt(sb, ERRORS_CONT))
 253                 seq_puts(seq, ",errors=continue");
 254         if (test_opt(sb, ERRORS_PANIC))
 255                 seq_puts(seq, ",errors=panic");
 256         if (test_opt(sb, NO_UID32))
 257                 seq_puts(seq, ",nouid32");
 258         if (test_opt(sb, DEBUG))
 259                 seq_puts(seq, ",debug");
 260         if (test_opt(sb, OLDALLOC))
 261                 seq_puts(seq, ",oldalloc");
 262
 263 #ifdef CONFIG_EXT2_FS_XATTR
 264         if (test_opt(sb, XATTR_USER))
 265                 seq_puts(seq, ",user_xattr");
 266         if (!test_opt(sb, XATTR_USER) &&
 267             (def_mount_opts & EXT2_DEFM_XATTR_USER)) {
 268                 seq_puts(seq, ",nouser_xattr");
 269         }
 270 #endif
 271
 272 #ifdef CONFIG_EXT2_FS_POSIX_ACL
 273         if (test_opt(sb, POSIX_ACL))
 274                 seq_puts(seq, ",acl");
 275         if (!test_opt(sb, POSIX_ACL) && (def_mount_opts & EXT2_DEFM_ACL))
 276                 seq_puts(seq, ",noacl");
 277 #endif
 278
 279         if (test_opt(sb, NOBH))
 280                 seq_puts(seq, ",nobh");
 281
 282 #if defined(CONFIG_QUOTA)
 283         if (sbi->s_mount_opt & EXT2_MOUNT_USRQUOTA)
 284                 seq_puts(seq, ",usrquota");
 285
 286         if (sbi->s_mount_opt & EXT2_MOUNT_GRPQUOTA)
 287                 seq_puts(seq, ",grpquota");
 288 #endif
 289
 290 #if defined(CONFIG_EXT2_FS_XIP)
 291         if (sbi->s_mount_opt & EXT2_MOUNT_XIP)
 292                 seq_puts(seq, ",xip");
 293 #endif
 294
 295         if (!test_opt(sb, RESERVATION))
 296                 seq_puts(seq, ",noreservation");
 297
 298         return 0;
 299 }
 300
 301 #ifdef CONFIG_QUOTA
 302 static ssize_t ext2_quota_read(struct super_block *sb, int type, char *data, size_t len, loff_t off);
 303 static ssize_t ext2_quota_write(struct super_block *sb, int type, const char *data, size_t len, loff_t off);
 304 #endif
 305
 306 static const struct super_operations ext2_sops = {
 307         .alloc_inode    = ext2_alloc_inode,
 308         .destroy_inode  = ext2_destroy_inode,
 309         .write_inode    = ext2_write_inode,
 310         .delete_inode   = ext2_delete_inode,
 311         .put_super      = ext2_put_super,
 312         .write_super    = ext2_write_super,
 313         .sync_fs        = ext2_sync_fs,
 314         .statfs         = ext2_statfs,
 315         .remount_fs     = ext2_remount,
 316         .clear_inode    = ext2_clear_inode,
 317         .show_options   = ext2_show_options,
 318 #ifdef CONFIG_QUOTA
 319         .quota_read     = ext2_quota_read,
 320         .quota_write    = ext2_quota_write,
 321 #endif
 322 };
 323
 324 static struct inode *ext2_nfs_get_inode(struct super_block *sb,
 325                 u64 ino, u32 generation)
 326 {
 327         struct inode *inode;
 328
 329         if (ino < EXT2_FIRST_INO(sb) && ino != EXT2_ROOT_INO)
 330                 return ERR_PTR(-ESTALE);
 331         if (ino > le32_to_cpu(EXT2_SB(sb)->s_es->s_inodes_count))
 332                 return ERR_PTR(-ESTALE);
 333
 334         /* iget isn't really right if the inode is currently unallocated!!
 335          * ext2_read_inode currently does appropriate checks, but
 336          * it might be "neater" to call ext2_get_inode first and check
 337          * if the inode is valid.....
 338          */
 339         inode = ext2_iget(sb, ino);
 340         if (IS_ERR(inode))
 341                 return ERR_CAST(inode);
 342         if (generation && inode->i_generation != generation) {
 343                 /* we didn't find the right inode.. */
 344                 iput(inode);
 345                 return ERR_PTR(-ESTALE);
 346         }
 347         return inode;
 348 }
 349
 350 static struct dentry *ext2_fh_to_dentry(struct super_block *sb, struct fid *fid,
 351                 int fh_len, int fh_type)
 352 {
 353         return generic_fh_to_dentry(sb, fid, fh_len, fh_type,
 354                                     ext2_nfs_get_inode);
 355 }
 356
 357 static struct dentry *ext2_fh_to_parent(struct super_block *sb, struct fid *fid,
 358                 int fh_len, int fh_type)
 359 {
 360         return generic_fh_to_parent(sb, fid, fh_len, fh_type,
 361                                     ext2_nfs_get_inode);
 362 }
 363
 364 /* Yes, most of these are left as NULL!!
 365  * A NULL value implies the default, which works with ext2-like file
 366  * systems, but can be improved upon.
 367  * Currently only get_parent is required.
 368  */
 369 static const struct export_operations ext2_export_ops = {
 370         .fh_to_dentry = ext2_fh_to_dentry,
 371         .fh_to_parent = ext2_fh_to_parent,
 372         .get_parent = ext2_get_parent,
 373 };
 374
 375 static unsigned long get_sb_block(void **data)
 376 {
 377         unsigned long   sb_block;
 378         char            *options = (char *) *data;
 379
 380         if (!options || strncmp(options, "sb=", 3) != 0)
 381                 return 1;       /* Default location */
 382         options += 3;
 383         sb_block = simple_strtoul(options, &options, 0);
 384         if (*options && *options != ',') {
 385                 printk("EXT2-fs: Invalid sb specification: %s\n",
 386                        (char *) *data);
 387                 return 1;
 388         }
 389         if (*options == ',')
 390                 options++;
 391         *data = (void *) options;
 392         return sb_block;
 393 }
 394
 395 enum {
 396         Opt_bsd_df, Opt_minix_df, Opt_grpid, Opt_nogrpid,
 397         Opt_resgid, Opt_resuid, Opt_sb, Opt_err_cont, Opt_err_panic,
 398         Opt_err_ro, Opt_nouid32, Opt_nocheck, Opt_debug,
 399         Opt_oldalloc, Opt_orlov, Opt_nobh, Opt_user_xattr, Opt_nouser_xattr,
 400         Opt_acl, Opt_noacl, Opt_xip, Opt_ignore, Opt_err, Opt_quota,
 401         Opt_usrquota, Opt_grpquota, Opt_reservation, Opt_noreservation
 402 };
 403
 404 static const match_table_t tokens = {
 405         {Opt_bsd_df, "bsddf"},
 406         {Opt_minix_df, "minixdf"},
 407         {Opt_grpid, "grpid"},
 408         {Opt_grpid, "bsdgroups"},
 409         {Opt_nogrpid, "nogrpid"},
 410         {Opt_nogrpid, "sysvgroups"},
 411         {Opt_resgid, "resgid=%u"},
 412         {Opt_resuid, "resuid=%u"},
 413         {Opt_sb, "sb=%u"},
 414         {Opt_err_cont, "errors=continue"},
 415         {Opt_err_panic, "errors=panic"},
 416         {Opt_err_ro, "errors=remount-ro"},
 417         {Opt_nouid32, "nouid32"},
 418         {Opt_nocheck, "check=none"},
 419         {Opt_nocheck, "nocheck"},
 420         {Opt_debug, "debug"},
 421         {Opt_oldalloc, "oldalloc"},
 422         {Opt_orlov, "orlov"},
 423         {Opt_nobh, "nobh"},
 424         {Opt_user_xattr, "user_xattr"},
 425         {Opt_nouser_xattr, "nouser_xattr"},
 426         {Opt_acl, "acl"},
 427         {Opt_noacl, "noacl"},
 428         {Opt_xip, "xip"},
 429         {Opt_grpquota, "grpquota"},
 430         {Opt_ignore, "noquota"},
 431         {Opt_quota, "quota"},
 432         {Opt_usrquota, "usrquota"},
 433         {Opt_reservation, "reservation"},
 434         {Opt_noreservation, "noreservation"},
 435         {Opt_err, NULL}
 436 };
 437
 438 static int parse_options (char * options,
 439                           struct ext2_sb_info *sbi)
 440 {
 441         char * p;
 442         substring_t args[MAX_OPT_ARGS];
 443         int option;
 444
 445         if (!options)
 446                 return 1;
 447
 448         while ((p = strsep (&options, ",")) != NULL) {
 449                 int token;
 450                 if (!*p)
 451                         continue;
 452
 453                 token = match_token(p, tokens, args);
 454                 switch (token) {
 455                 case Opt_bsd_df:
 456                         clear_opt (sbi->s_mount_opt, MINIX_DF);
 457                         break;
 458                 case Opt_minix_df:
 459                         set_opt (sbi->s_mount_opt, MINIX_DF);
 460                         break;
 461                 case Opt_grpid:
 462                         set_opt (sbi->s_mount_opt, GRPID);
 463                         break;
 464                 case Opt_nogrpid:
 465                         clear_opt (sbi->s_mount_opt, GRPID);
 466                         break;
 467                 case Opt_resuid:
 468                         if (match_int(&args[0], &option))
 469                                 return 0;
 470                         sbi->s_resuid = option;
 471                         break;
 472                 case Opt_resgid:
 473                         if (match_int(&args[0], &option))
 474                                 return 0;
 475                         sbi->s_resgid = option;
 476                         break;
 477                 case Opt_sb:
 478                         /* handled by get_sb_block() instead of here */
 479                         /* *sb_block = match_int(&args[0]); */
 480                         break;
 481                 case Opt_err_panic:
 482                         clear_opt (sbi->s_mount_opt, ERRORS_CONT);
 483                         clear_opt (sbi->s_mount_opt, ERRORS_RO);
 484                         set_opt (sbi->s_mount_opt, ERRORS_PANIC);
 485                         break;
 486                 case Opt_err_ro:
 487                         clear_opt (sbi->s_mount_opt, ERRORS_CONT);
 488                         clear_opt (sbi->s_mount_opt, ERRORS_PANIC);
 489                         set_opt (sbi->s_mount_opt, ERRORS_RO);
 490                         break;
 491                 case Opt_err_cont:
 492                         clear_opt (sbi->s_mount_opt, ERRORS_RO);
 493                         clear_opt (sbi->s_mount_opt, ERRORS_PANIC);
 494                         set_opt (sbi->s_mount_opt, ERRORS_CONT);
 495                         break;
 496                 case Opt_nouid32:
 497                         set_opt (sbi->s_mount_opt, NO_UID32);
 498                         break;
 499                 case Opt_nocheck:
 500                         clear_opt (sbi->s_mount_opt, CHECK);
 501                         break;
 502                 case Opt_debug:
 503                         set_opt (sbi->s_mount_opt, DEBUG);
 504                         break;
 505                 case Opt_oldalloc:
 506                         set_opt (sbi->s_mount_opt, OLDALLOC);
 507                         break;
 508                 case Opt_orlov:
 509                         clear_opt (sbi->s_mount_opt, OLDALLOC);
 510                         break;
 511                 case Opt_nobh:
 512                         set_opt (sbi->s_mount_opt, NOBH);
 513                         break;
 514 #ifdef CONFIG_EXT2_FS_XATTR
 515                 case Opt_user_xattr:
 516                         set_opt (sbi->s_mount_opt, XATTR_USER);
 517                         break;
 518                 case Opt_nouser_xattr:
 519                         clear_opt (sbi->s_mount_opt, XATTR_USER);
 520                         break;
 521 #else
 522                 case Opt_user_xattr:
 523                 case Opt_nouser_xattr:
 524                         printk("EXT2 (no)user_xattr options not supported\n");
 525                         break;
 526 #endif
 527 #ifdef CONFIG_EXT2_FS_POSIX_ACL
 528                 case Opt_acl:
 529                         set_opt(sbi->s_mount_opt, POSIX_ACL);
 530                         break;
 531                 case Opt_noacl:
 532                         clear_opt(sbi->s_mount_opt, POSIX_ACL);
 533                         break;
 534 #else
 535                 case Opt_acl:
 536                 case Opt_noacl:
 537                         printk("EXT2 (no)acl options not supported\n");
 538                         break;
 539 #endif
 540                 case Opt_xip:
 541 #ifdef CONFIG_EXT2_FS_XIP
 542                         set_opt (sbi->s_mount_opt, XIP);
 543 #else
 544                         printk("EXT2 xip option not supported\n");
 545 #endif
 546                         break;
 547
 548 #if defined(CONFIG_QUOTA)
 549                 case Opt_quota:
 550                 case Opt_usrquota:
 551                         set_opt(sbi->s_mount_opt, USRQUOTA);
 552                         break;
 553
 554                 case Opt_grpquota:
 555                         set_opt(sbi->s_mount_opt, GRPQUOTA);
 556                         break;
 557 #else
 558                 case Opt_quota:
 559                 case Opt_usrquota:
 560                 case Opt_grpquota:
 561                         printk(KERN_ERR
 562                                 "EXT2-fs: quota operations not supported.\n");
 563
 564                         break;
 565 #endif
 566
 567                 case Opt_reservation:
 568                         set_opt(sbi->s_mount_opt, RESERVATION);
 569                         printk("reservations ON\n");
 570                         break;
 571                 case Opt_noreservation:
 572                         clear_opt(sbi->s_mount_opt, RESERVATION);
 573                         printk("reservations OFF\n");
 574                         break;
 575                 case Opt_ignore:
 576                         break;
 577                 default:
 578                         return 0;
 579                 }
 580         }
 581         return 1;
 582 }
 583
 584 static int ext2_setup_super (struct super_block * sb,
 585                               struct ext2_super_block * es,
 586                               int read_only)
 587 {
 588         int res = 0;
 589         struct ext2_sb_info *sbi = EXT2_SB(sb);
 590
 591         if (le32_to_cpu(es->s_rev_level) > EXT2_MAX_SUPP_REV) {
 592                 printk ("EXT2-fs warning: revision level too high, "
 593                         "forcing read-only mode\n");
 594                 res = MS_RDONLY;
 595         }
 596         if (read_only)
 597                 return res;
 598         if (!(sbi->s_mount_state & EXT2_VALID_FS))
 599                 printk ("EXT2-fs warning: mounting unchecked fs, "
 600                         "running e2fsck is recommended\n");
 601         else if ((sbi->s_mount_state & EXT2_ERROR_FS))
 602                 printk ("EXT2-fs warning: mounting fs with errors, "
 603                         "running e2fsck is recommended\n");
 604         else if ((__s16) le16_to_cpu(es->s_max_mnt_count) >= 0 &&
 605                  le16_to_cpu(es->s_mnt_count) >=
 606                  (unsigned short) (__s16) le16_to_cpu(es->s_max_mnt_count))
 607                 printk ("EXT2-fs warning: maximal mount count reached, "
 608                         "running e2fsck is recommended\n");
 609         else if (le32_to_cpu(es->s_checkinterval) &&
 610                 (le32_to_cpu(es->s_lastcheck) + le32_to_cpu(es->s_checkinterval) <= get_seconds()))
 611                 printk ("EXT2-fs warning: checktime reached, "
 612                         "running e2fsck is recommended\n");
 613         if (!le16_to_cpu(es->s_max_mnt_count))
 614                 es->s_max_mnt_count = cpu_to_le16(EXT2_DFL_MAX_MNT_COUNT);
 615         le16_add_cpu(&es->s_mnt_count, 1);
 616         ext2_write_super(sb);
 617         if (test_opt (sb, DEBUG))
 618                 printk ("[EXT II FS %s, %s, bs=%lu, fs=%lu, gc=%lu, "
 619                         "bpg=%lu, ipg=%lu, mo=%04lx]\n",
 620                         EXT2FS_VERSION, EXT2FS_DATE, sb->s_blocksize,
 621                         sbi->s_frag_size,
 622                         sbi->s_groups_count,
 623                         EXT2_BLOCKS_PER_GROUP(sb),
 624                         EXT2_INODES_PER_GROUP(sb),
 625                         sbi->s_mount_opt);
 626         return res;
 627 }
 628
 629 static int ext2_check_descriptors(struct super_block *sb)
 630 {
 631         int i;
 632         struct ext2_sb_info *sbi = EXT2_SB(sb);
 633
 634         ext2_debug ("Checking group descriptors");
 635
 636         for (i = 0; i < sbi->s_groups_count; i++) {
 637                 struct ext2_group_desc *gdp = ext2_get_group_desc(sb, i, NULL);
 638                 ext2_fsblk_t first_block = ext2_group_first_block_no(sb, i);
 639                 ext2_fsblk_t last_block;
 640
 641                 if (i == sbi->s_groups_count - 1)
 642                         last_block = le32_to_cpu(sbi->s_es->s_blocks_count) - 1;
 643                 else
 644                         last_block = first_block +
 645                                 (EXT2_BLOCKS_PER_GROUP(sb) - 1);
 646
 647                 if (le32_to_cpu(gdp->bg_block_bitmap) < first_block ||
 648                     le32_to_cpu(gdp->bg_block_bitmap) > last_block)
 649                 {
 650                         ext2_error (sb, "ext2_check_descriptors",
 651                                     "Block bitmap for group %d"
 652                                     " not in group (block %lu)!",
 653                                     i, (unsigned long) le32_to_cpu(gdp->bg_block_bitmap));
 654                         return 0;
 655                 }
 656                 if (le32_to_cpu(gdp->bg_inode_bitmap) < first_block ||
 657                     le32_to_cpu(gdp->bg_inode_bitmap) > last_block)
 658                 {
 659                         ext2_error (sb, "ext2_check_descriptors",
 660                                     "Inode bitmap for group %d"
 661                                     " not in group (block %lu)!",
 662                                     i, (unsigned long) le32_to_cpu(gdp->bg_inode_bitmap));
 663                         return 0;
 664                 }
 665                 if (le32_to_cpu(gdp->bg_inode_table) < first_block ||
 666                     le32_to_cpu(gdp->bg_inode_table) + sbi->s_itb_per_group - 1 >
 667                     last_block)
 668                 {
 669                         ext2_error (sb, "ext2_check_descriptors",
 670                                     "Inode table for group %d"
 671                                     " not in group (block %lu)!",
 672                                     i, (unsigned long) le32_to_cpu(gdp->bg_inode_table));
 673                         return 0;
 674                 }
 675         }
 676         return 1;
 677 }
 678
 679 /*
 680  * Maximal file size.  There is a direct, and {,double-,triple-}indirect
 681  * block limit, and also a limit of (2^32 - 1) 512-byte sectors in i_blocks.
 682  * We need to be 1 filesystem block less than the 2^32 sector limit.
 683  */
 684 static loff_t ext2_max_size(int bits)
 685 {
 686         loff_t res = EXT2_NDIR_BLOCKS;
 687         int meta_blocks;
 688         loff_t upper_limit;
 689
 690         /* This is calculated to be the largest file size for a
 691          * dense, file such that the total number of
 692          * sectors in the file, including data and all indirect blocks,
 693          * does not exceed 2^32 -1
 694          * __u32 i_blocks representing the total number of
 695          * 512 bytes blocks of the file
 696          */
 697         upper_limit = (1LL << 32) - 1;
 698
 699         /* total blocks in file system block size */
 700         upper_limit >>= (bits - 9);
 701
 702
 703         /* indirect blocks */
 704         meta_blocks = 1;
 705         /* double indirect blocks */
 706         meta_blocks += 1 + (1LL << (bits-2));
 707         /* tripple indirect blocks */
 708         meta_blocks += 1 + (1LL << (bits-2)) + (1LL << (2*(bits-2)));
 709
 710         upper_limit -= meta_blocks;
 711         upper_limit <<= bits;
 712
 713         res += 1LL << (bits-2);
 714         res += 1LL << (2*(bits-2));
 715         res += 1LL << (3*(bits-2));
 716         res <<= bits;
 717         if (res > upper_limit)
 718                 res = upper_limit;
 719
 720         if (res > MAX_LFS_FILESIZE)
 721                 res = MAX_LFS_FILESIZE;
 722
 723         return res;
 724 }
 725
 726 static unsigned long descriptor_loc(struct super_block *sb,
 727                                     unsigned long logic_sb_block,
 728                                     int nr)
 729 {
 730         struct ext2_sb_info *sbi = EXT2_SB(sb);
 731         unsigned long bg, first_meta_bg;
 732         int has_super = 0;
 733
 734         first_meta_bg = le32_to_cpu(sbi->s_es->s_first_meta_bg);
 735
 736         if (!EXT2_HAS_INCOMPAT_FEATURE(sb, EXT2_FEATURE_INCOMPAT_META_BG) ||
 737             nr < first_meta_bg)
 738                 return (logic_sb_block + nr + 1);
 739         bg = sbi->s_desc_per_block * nr;
 740         if (ext2_bg_has_super(sb, bg))
 741                 has_super = 1;
 742
 743         return ext2_group_first_block_no(sb, bg) + has_super;
 744 }
 745
 746 static int ext2_fill_super(struct super_block *sb, void *data, int silent)
 747 {
 748         struct buffer_head * bh;
 749         struct ext2_sb_info * sbi;
 750         struct ext2_super_block * es;
 751         struct inode *root;
 752         unsigned long block;
 753         unsigned long sb_block = get_sb_block(&data);
 754         unsigned long logic_sb_block;
 755         unsigned long offset = 0;
 756         unsigned long def_mount_opts;
 757         long ret = -EINVAL;
 758         int blocksize = BLOCK_SIZE;
 759         int db_count;
 760         int i, j;
 761         __le32 features;
 762         int err;
 763
 764         sbi = kzalloc(sizeof(*sbi), GFP_KERNEL);
 765         if (!sbi)
 766                 return -ENOMEM;
 767
 768         sbi->s_blockgroup_lock =
 769                 kzalloc(sizeof(struct blockgroup_lock), GFP_KERNEL);
 770         if (!sbi->s_blockgroup_lock) {
 771                 kfree(sbi);
 772                 return -ENOMEM;
 773         }
 774         sb->s_fs_info = sbi;
 775         sbi->s_sb_block = sb_block;
 776
 777         /*
 778          * See what the current blocksize for the device is, and
 779          * use that as the blocksize.  Otherwise (or if the blocksize
 780          * is smaller than the default) use the default.
 781          * This is important for devices that have a hardware
 782          * sectorsize that is larger than the default.
 783          */
 784         blocksize = sb_min_blocksize(sb, BLOCK_SIZE);
 785         if (!blocksize) {
 786                 printk ("EXT2-fs: unable to set blocksize\n");
 787                 goto failed_sbi;
 788         }
 789
 790         /*
 791          * If the superblock doesn't start on a hardware sector boundary,
 792          * calculate the offset.
 793          */
 794         if (blocksize != BLOCK_SIZE) {
 795                 logic_sb_block = (sb_block*BLOCK_SIZE) / blocksize;
 796                 offset = (sb_block*BLOCK_SIZE) % blocksize;
 797         } else {
 798                 logic_sb_block = sb_block;
 799         }
 800
 801         if (!(bh = sb_bread(sb, logic_sb_block))) {
 802                 printk ("EXT2-fs: unable to read superblock\n");
 803                 goto failed_sbi;
 804         }
 805         /*
 806          * Note: s_es must be initialized as soon as possible because
 807          *       some ext2 macro-instructions depend on its value
 808          */
 809         es = (struct ext2_super_block *) (((char *)bh->b_data) + offset);
 810         sbi->s_es = es;
 811         sb->s_magic = le16_to_cpu(es->s_magic);
 812
 813         if (sb->s_magic != EXT2_SUPER_MAGIC)
 814                 goto cantfind_ext2;
 815
 816         /* Set defaults before we parse the mount options */
 817         def_mount_opts = le32_to_cpu(es->s_default_mount_opts);
 818         if (def_mount_opts & EXT2_DEFM_DEBUG)
 819                 set_opt(sbi->s_mount_opt, DEBUG);
 820         if (def_mount_opts & EXT2_DEFM_BSDGROUPS)
 821                 set_opt(sbi->s_mount_opt, GRPID);
 822         if (def_mount_opts & EXT2_DEFM_UID16)
 823                 set_opt(sbi->s_mount_opt, NO_UID32);
 824 #ifdef CONFIG_EXT2_FS_XATTR
 825         if (def_mount_opts & EXT2_DEFM_XATTR_USER)
 826                 set_opt(sbi->s_mount_opt, XATTR_USER);
 827 #endif
 828 #ifdef CONFIG_EXT2_FS_POSIX_ACL
 829         if (def_mount_opts & EXT2_DEFM_ACL)
 830                 set_opt(sbi->s_mount_opt, POSIX_ACL);
 831 #endif
 832
 833         if (le16_to_cpu(sbi->s_es->s_errors) == EXT2_ERRORS_PANIC)
 834                 set_opt(sbi->s_mount_opt, ERRORS_PANIC);
 835         else if (le16_to_cpu(sbi->s_es->s_errors) == EXT2_ERRORS_CONTINUE)
 836                 set_opt(sbi->s_mount_opt, ERRORS_CONT);
 837         else
 838                 set_opt(sbi->s_mount_opt, ERRORS_RO);
 839
 840         sbi->s_resuid = le16_to_cpu(es->s_def_resuid);
 841         sbi->s_resgid = le16_to_cpu(es->s_def_resgid);
 842
 843         set_opt(sbi->s_mount_opt, RESERVATION);
 844
 845         if (!parse_options ((char *) data, sbi))
 846                 goto failed_mount;
 847
 848         sb->s_flags = (sb->s_flags & ~MS_POSIXACL) |
 849                 ((EXT2_SB(sb)->s_mount_opt & EXT2_MOUNT_POSIX_ACL) ?
 850                  MS_POSIXACL : 0);
 851
 852         ext2_xip_verify_sb(sb); /* see if bdev supports xip, unset
 853                                     EXT2_MOUNT_XIP if not */
 854
 855         if (le32_to_cpu(es->s_rev_level) == EXT2_GOOD_OLD_REV &&
 856             (EXT2_HAS_COMPAT_FEATURE(sb, ~0U) ||
 857              EXT2_HAS_RO_COMPAT_FEATURE(sb, ~0U) ||
 858              EXT2_HAS_INCOMPAT_FEATURE(sb, ~0U)))
 859                 printk("EXT2-fs warning: feature flags set on rev 0 fs, "
 860                        "running e2fsck is recommended\n");
 861         /*
 862          * Check feature flags regardless of the revision level, since we
 863          * previously didn't change the revision level when setting the flags,
 864          * so there is a chance incompat flags are set on a rev 0 filesystem.
 865          */
 866         features = EXT2_HAS_INCOMPAT_FEATURE(sb, ~EXT2_FEATURE_INCOMPAT_SUPP);
 867         if (features) {
 868                 printk("EXT2-fs: %s: couldn't mount because of "
 869                        "unsupported optional features (%x).\n",
 870                        sb->s_id, le32_to_cpu(features));
 871                 goto failed_mount;
 872         }
 873         if (!(sb->s_flags & MS_RDONLY) &&
 874             (features = EXT2_HAS_RO_COMPAT_FEATURE(sb, ~EXT2_FEATURE_RO_COMPAT_SUPP))){
 875                 printk("EXT2-fs: %s: couldn't mount RDWR because of "
 876                        "unsupported optional features (%x).\n",
 877                        sb->s_id, le32_to_cpu(features));
 878                 goto failed_mount;
 879         }
 880
 881         blocksize = BLOCK_SIZE << le32_to_cpu(sbi->s_es->s_log_block_size);
 882
 883         if (ext2_use_xip(sb) && blocksize != PAGE_SIZE) {
 884                 if (!silent)
 885                         printk("XIP: Unsupported blocksize\n");
 886                 goto failed_mount;
 887         }
 888
 889         /* If the blocksize doesn't match, re-read the thing.. */
 890         if (sb->s_blocksize != blocksize) {
 891                 brelse(bh);
 892
 893                 if (!sb_set_blocksize(sb, blocksize)) {
 894                         printk(KERN_ERR "EXT2-fs: blocksize too small for device.\n");
 895                         goto failed_sbi;
 896                 }
 897
 898                 logic_sb_block = (sb_block*BLOCK_SIZE) / blocksize;
 899                 offset = (sb_block*BLOCK_SIZE) % blocksize;
 900                 bh = sb_bread(sb, logic_sb_block);
 901                 if(!bh) {
 902                         printk("EXT2-fs: Couldn't read superblock on "
 903                                "2nd try.\n");
 904                         goto failed_sbi;
 905                 }
 906                 es = (struct ext2_super_block *) (((char *)bh->b_data) + offset);
 907                 sbi->s_es = es;
 908                 if (es->s_magic != cpu_to_le16(EXT2_SUPER_MAGIC)) {
 909                         printk ("EXT2-fs: Magic mismatch, very weird !\n");
 910                         goto failed_mount;
 911                 }
 912         }
 913
 914         sb->s_maxbytes = ext2_max_size(sb->s_blocksize_bits);
 915
 916         if (le32_to_cpu(es->s_rev_level) == EXT2_GOOD_OLD_REV) {
 917                 sbi->s_inode_size = EXT2_GOOD_OLD_INODE_SIZE;
 918                 sbi->s_first_ino = EXT2_GOOD_OLD_FIRST_INO;
 919         } else {
 920                 sbi->s_inode_size = le16_to_cpu(es->s_inode_size);
 921                 sbi->s_first_ino = le32_to_cpu(es->s_first_ino);
 922                 if ((sbi->s_inode_size < EXT2_GOOD_OLD_INODE_SIZE) ||
 923                     !is_power_of_2(sbi->s_inode_size) ||
 924                     (sbi->s_inode_size > blocksize)) {
 925                         printk ("EXT2-fs: unsupported inode size: %d\n",
 926                                 sbi->s_inode_size);
 927                         goto failed_mount;
 928                 }
 929         }
 930
 931         sbi->s_frag_size = EXT2_MIN_FRAG_SIZE <<
 932                                    le32_to_cpu(es->s_log_frag_size);
 933         if (sbi->s_frag_size == 0)
 934                 goto cantfind_ext2;
 935         sbi->s_frags_per_block = sb->s_blocksize / sbi->s_frag_size;
 936
 937         sbi->s_blocks_per_group = le32_to_cpu(es->s_blocks_per_group);
 938         sbi->s_frags_per_group = le32_to_cpu(es->s_frags_per_group);
 939         sbi->s_inodes_per_group = le32_to_cpu(es->s_inodes_per_group);
 940
 941         if (EXT2_INODE_SIZE(sb) == 0)
 942                 goto cantfind_ext2;
 943         sbi->s_inodes_per_block = sb->s_blocksize / EXT2_INODE_SIZE(sb);
 944         if (sbi->s_inodes_per_block == 0 || sbi->s_inodes_per_group == 0)
 945                 goto cantfind_ext2;
 946         sbi->s_itb_per_group = sbi->s_inodes_per_group /
 947                                         sbi->s_inodes_per_block;
 948         sbi->s_desc_per_block = sb->s_blocksize /
 949                                         sizeof (struct ext2_group_desc);
 950         sbi->s_sbh = bh;
 951         sbi->s_mount_state = le16_to_cpu(es->s_state);
 952         sbi->s_addr_per_block_bits =
 953                 ilog2 (EXT2_ADDR_PER_BLOCK(sb));
 954         sbi->s_desc_per_block_bits =
 955                 ilog2 (EXT2_DESC_PER_BLOCK(sb));
 956
 957         if (sb->s_magic != EXT2_SUPER_MAGIC)
 958                 goto cantfind_ext2;
 959
 960         if (sb->s_blocksize != bh->b_size) {
 961                 if (!silent)
 962                         printk ("VFS: Unsupported blocksize on dev "
 963                                 "%s.\n", sb->s_id);
 964                 goto failed_mount;
 965         }
 966
 967         if (sb->s_blocksize != sbi->s_frag_size) {
 968                 printk ("EXT2-fs: fragsize %lu != blocksize %lu (not supported yet)\n",
 969                         sbi->s_frag_size, sb->s_blocksize);
 970                 goto failed_mount;
 971         }
 972
 973         if (sbi->s_blocks_per_group > sb->s_blocksize * 8) {
 974                 printk ("EXT2-fs: #blocks per group too big: %lu\n",
 975                         sbi->s_blocks_per_group);
 976                 goto failed_mount;
 977         }
 978         if (sbi->s_frags_per_group > sb->s_blocksize * 8) {
 979                 printk ("EXT2-fs: #fragments per group too big: %lu\n",
 980                         sbi->s_frags_per_group);
 981                 goto failed_mount;
 982         }
 983         if (sbi->s_inodes_per_group > sb->s_blocksize * 8) {
 984                 printk ("EXT2-fs: #inodes per group too big: %lu\n",
 985                         sbi->s_inodes_per_group);
 986                 goto failed_mount;
 987         }
 988
 989         if (EXT2_BLOCKS_PER_GROUP(sb) == 0)
 990                 goto cantfind_ext2;
 991         sbi->s_groups_count = ((le32_to_cpu(es->s_blocks_count) -
 992                                 le32_to_cpu(es->s_first_data_block) - 1)
 993                                         / EXT2_BLOCKS_PER_GROUP(sb)) + 1;
 994         db_count = (sbi->s_groups_count + EXT2_DESC_PER_BLOCK(sb) - 1) /
 995                    EXT2_DESC_PER_BLOCK(sb);
 996         sbi->s_group_desc = kmalloc (db_count * sizeof (struct buffer_head *), GFP_KERNEL);
 997         if (sbi->s_group_desc == NULL) {
 998                 printk ("EXT2-fs: not enough memory\n");
 999                 goto failed_mount;
1000         }
1001         bgl_lock_init(sbi->s_blockgroup_lock);
1002         sbi->s_debts = kcalloc(sbi->s_groups_count, sizeof(*sbi->s_debts), GFP_KERNEL);
1003         if (!sbi->s_debts) {
1004                 printk ("EXT2-fs: not enough memory\n");
1005                 goto failed_mount_group_desc;
1006         }
1007         for (i = 0; i < db_count; i++) {
1008                 block = descriptor_loc(sb, logic_sb_block, i);
1009                 sbi->s_group_desc[i] = sb_bread(sb, block);
1010                 if (!sbi->s_group_desc[i]) {
1011                         for (j = 0; j < i; j++)
1012                                 brelse (sbi->s_group_desc[j]);
1013                         printk ("EXT2-fs: unable to read group descriptors\n");
1014                         goto failed_mount_group_desc;
1015                 }
1016         }
1017         if (!ext2_check_descriptors (sb)) {
1018                 printk ("EXT2-fs: group descriptors corrupted!\n");
1019                 goto failed_mount2;
1020         }
1021         sbi->s_gdb_count = db_count;
1022         get_random_bytes(&sbi->s_next_generation, sizeof(u32));
1023         spin_lock_init(&sbi->s_next_gen_lock);
1024
1025         /* per fileystem reservation list head & lock */
1026         spin_lock_init(&sbi->s_rsv_window_lock);
1027         sbi->s_rsv_window_root = RB_ROOT;
1028         /*
1029          * Add a single, static dummy reservation to the start of the
1030          * reservation window list --- it gives us a placeholder for
1031          * append-at-start-of-list which makes the allocation logic
1032          * _much_ simpler.
1033          */
1034         sbi->s_rsv_window_head.rsv_start = EXT2_RESERVE_WINDOW_NOT_ALLOCATED;
1035         sbi->s_rsv_window_head.rsv_end = EXT2_RESERVE_WINDOW_NOT_ALLOCATED;
1036         sbi->s_rsv_window_head.rsv_alloc_hit = 0;
1037         sbi->s_rsv_window_head.rsv_goal_size = 0;
1038         ext2_rsv_window_add(sb, &sbi->s_rsv_window_head);
1039
1040         err = percpu_counter_init(&sbi->s_freeblocks_counter,
1041                                 ext2_count_free_blocks(sb));
1042         if (!err) {
1043                 err = percpu_counter_init(&sbi->s_freeinodes_counter,
1044                                 ext2_count_free_inodes(sb));
1045         }
1046         if (!err) {
1047                 err = percpu_counter_init(&sbi->s_dirs_counter,
1048                                 ext2_count_dirs(sb));
1049         }
1050         if (err) {
1051                 printk(KERN_ERR "EXT2-fs: insufficient memory\n");
1052                 goto failed_mount3;
1053         }
1054         /*
1055          * set up enough so that it can read an inode
1056          */
1057         sb->s_op = &ext2_sops;
1058         sb->s_export_op = &ext2_export_ops;
1059         sb->s_xattr = ext2_xattr_handlers;
1060         root = ext2_iget(sb, EXT2_ROOT_INO);
1061         if (IS_ERR(root)) {
1062                 ret = PTR_ERR(root);
1063                 goto failed_mount3;
1064         }
1065         if (!S_ISDIR(root->i_mode) || !root->i_blocks || !root->i_size) {
1066                 iput(root);
1067                 printk(KERN_ERR "EXT2-fs: corrupt root inode, run e2fsck\n");
1068                 goto failed_mount3;
1069         }
1070
1071         sb->s_root = d_alloc_root(root);
1072         if (!sb->s_root) {
1073                 iput(root);
1074                 printk(KERN_ERR "EXT2-fs: get root inode failed\n");
1075                 ret = -ENOMEM;
1076                 goto failed_mount3;
1077         }
1078         if (EXT2_HAS_COMPAT_FEATURE(sb, EXT3_FEATURE_COMPAT_HAS_JOURNAL))
1079                 ext2_warning(sb, __func__,
1080                         "mounting ext3 filesystem as ext2");
1081         ext2_setup_super (sb, es, sb->s_flags & MS_RDONLY);
1082         return 0;
1083
1084 cantfind_ext2:
1085         if (!silent)
1086                 printk("VFS: Can't find an ext2 filesystem on dev %s.\n",
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_commit_super (struct super_block * sb,
1109                                struct ext2_super_block * es)
1110 {
1111         es->s_wtime = cpu_to_le32(get_seconds());
1112         mark_buffer_dirty(EXT2_SB(sb)->s_sbh);
1113         sb->s_dirt = 0;
1114 }
1115
1116 static void ext2_sync_super(struct super_block *sb, struct ext2_super_block *es)
1117 {
1118         es->s_free_blocks_count = cpu_to_le32(ext2_count_free_blocks(sb));
1119         es->s_free_inodes_count = cpu_to_le32(ext2_count_free_inodes(sb));
1120         es->s_wtime = cpu_to_le32(get_seconds());
1121         mark_buffer_dirty(EXT2_SB(sb)->s_sbh);
1122         sync_dirty_buffer(EXT2_SB(sb)->s_sbh);
1123         sb->s_dirt = 0;
1124 }
1125
1126 /*
1127  * In the second extended file system, it is not necessary to
1128  * write the super block since we use a mapping of the
1129  * disk super block in a buffer.
1130  *
1131  * However, this function is still used to set the fs valid
1132  * flags to 0.  We need to set this flag to 0 since the fs
1133  * may have been checked while mounted and e2fsck may have
1134  * set s_state to EXT2_VALID_FS after some corrections.
1135  */
1136
1137 static int ext2_sync_fs(struct super_block *sb, int wait)
1138 {
1139         struct ext2_super_block *es = EXT2_SB(sb)->s_es;
1140
1141         lock_kernel();
1142         if (es->s_state & cpu_to_le16(EXT2_VALID_FS)) {
1143                 ext2_debug("setting valid to 0\n");
1144                 es->s_state &= cpu_to_le16(~EXT2_VALID_FS);
1145                 es->s_free_blocks_count =
1146                         cpu_to_le32(ext2_count_free_blocks(sb));
1147                 es->s_free_inodes_count =
1148                         cpu_to_le32(ext2_count_free_inodes(sb));
1149                 es->s_mtime = cpu_to_le32(get_seconds());
1150                 ext2_sync_super(sb, es);
1151         } else {
1152                 ext2_commit_super(sb, es);
1153         }
1154         sb->s_dirt = 0;
1155         unlock_kernel();
1156
1157         return 0;
1158 }
1159
1160
1161 void ext2_write_super(struct super_block *sb)
1162 {
1163         if (!(sb->s_flags & MS_RDONLY))
1164                 ext2_sync_fs(sb, 1);
1165         else
1166                 sb->s_dirt = 0;
1167 }
1168
1169 static int ext2_remount (struct super_block * sb, int * flags, char * data)
1170 {
1171         struct ext2_sb_info * sbi = EXT2_SB(sb);
1172         struct ext2_super_block * es;
1173         unsigned long old_mount_opt = sbi->s_mount_opt;
1174         struct ext2_mount_options old_opts;
1175         unsigned long old_sb_flags;
1176         int err;
1177
1178         lock_kernel();
1179
1180         /* Store the old options */
1181         old_sb_flags = sb->s_flags;
1182         old_opts.s_mount_opt = sbi->s_mount_opt;
1183         old_opts.s_resuid = sbi->s_resuid;
1184         old_opts.s_resgid = sbi->s_resgid;
1185
1186         /*
1187          * Allow the "check" option to be passed as a remount option.
1188          */
1189         if (!parse_options (data, sbi)) {
1190                 err = -EINVAL;
1191                 goto restore_opts;
1192         }
1193
1194         sb->s_flags = (sb->s_flags & ~MS_POSIXACL) |
1195                 ((sbi->s_mount_opt & EXT2_MOUNT_POSIX_ACL) ? MS_POSIXACL : 0);
1196
1197         ext2_xip_verify_sb(sb); /* see if bdev supports xip, unset
1198                                     EXT2_MOUNT_XIP if not */
1199
1200         if ((ext2_use_xip(sb)) && (sb->s_blocksize != PAGE_SIZE)) {
1201                 printk("XIP: Unsupported blocksize\n");
1202                 err = -EINVAL;
1203                 goto restore_opts;
1204         }
1205
1206         es = sbi->s_es;
1207         if (((sbi->s_mount_opt & EXT2_MOUNT_XIP) !=
1208             (old_mount_opt & EXT2_MOUNT_XIP)) &&
1209             invalidate_inodes(sb)) {
1210                 ext2_warning(sb, __func__, "refusing change of xip flag "
1211                              "with busy inodes while remounting");
1212                 sbi->s_mount_opt &= ~EXT2_MOUNT_XIP;
1213                 sbi->s_mount_opt |= old_mount_opt & EXT2_MOUNT_XIP;
1214         }
1215         if ((*flags & MS_RDONLY) == (sb->s_flags & MS_RDONLY)) {
1216                 unlock_kernel();
1217                 return 0;
1218         }
1219         if (*flags & MS_RDONLY) {
1220                 if (le16_to_cpu(es->s_state) & EXT2_VALID_FS ||
1221                     !(sbi->s_mount_state & EXT2_VALID_FS)) {
1222                         unlock_kernel();
1223                         return 0;
1224                 }
1225                 /*
1226                  * OK, we are remounting a valid rw partition rdonly, so set
1227                  * the rdonly flag and then mark the partition as valid again.
1228                  */
1229                 es->s_state = cpu_to_le16(sbi->s_mount_state);
1230                 es->s_mtime = cpu_to_le32(get_seconds());
1231         } else {
1232                 __le32 ret = EXT2_HAS_RO_COMPAT_FEATURE(sb,
1233                                                ~EXT2_FEATURE_RO_COMPAT_SUPP);
1234                 if (ret) {
1235                         printk("EXT2-fs: %s: couldn't remount RDWR because of "
1236                                "unsupported optional features (%x).\n",
1237                                sb->s_id, le32_to_cpu(ret));
1238                         err = -EROFS;
1239                         goto restore_opts;
1240                 }
1241                 /*
1242                  * Mounting a RDONLY partition read-write, so reread and
1243                  * store the current valid flag.  (It may have been changed
1244                  * by e2fsck since we originally mounted the partition.)
1245                  */
1246                 sbi->s_mount_state = le16_to_cpu(es->s_state);
1247                 if (!ext2_setup_super (sb, es, 0))
1248                         sb->s_flags &= ~MS_RDONLY;
1249         }
1250         ext2_sync_super(sb, es);
1251         unlock_kernel();
1252         return 0;
1253 restore_opts:
1254         sbi->s_mount_opt = old_opts.s_mount_opt;
1255         sbi->s_resuid = old_opts.s_resuid;
1256         sbi->s_resgid = old_opts.s_resgid;
1257         sb->s_flags = old_sb_flags;
1258         unlock_kernel();
1259         return err;
1260 }
1261
1262 static int ext2_statfs (struct dentry * dentry, struct kstatfs * buf)
1263 {
1264         struct super_block *sb = dentry->d_sb;
1265         struct ext2_sb_info *sbi = EXT2_SB(sb);
1266         struct ext2_super_block *es = sbi->s_es;
1267         u64 fsid;
1268
1269         if (test_opt (sb, MINIX_DF))
1270                 sbi->s_overhead_last = 0;
1271         else if (sbi->s_blocks_last != le32_to_cpu(es->s_blocks_count)) {
1272                 unsigned long i, overhead = 0;
1273                 smp_rmb();
1274
1275                 /*
1276                  * Compute the overhead (FS structures). This is constant
1277                  * for a given filesystem unless the number of block groups
1278                  * changes so we cache the previous value until it does.
1279                  */
1280
1281                 /*
1282                  * All of the blocks before first_data_block are
1283                  * overhead
1284                  */
1285                 overhead = le32_to_cpu(es->s_first_data_block);
1286
1287                 /*
1288                  * Add the overhead attributed to the superblock and
1289                  * block group descriptors.  If the sparse superblocks
1290                  * feature is turned on, then not all groups have this.
1291                  */
1292                 for (i = 0; i < sbi->s_groups_count; i++)
1293                         overhead += ext2_bg_has_super(sb, i) +
1294                                 ext2_bg_num_gdb(sb, i);
1295
1296                 /*
1297                  * Every block group has an inode bitmap, a block
1298                  * bitmap, and an inode table.
1299                  */
1300                 overhead += (sbi->s_groups_count *
1301                              (2 + sbi->s_itb_per_group));
1302                 sbi->s_overhead_last = overhead;
1303                 smp_wmb();
1304                 sbi->s_blocks_last = le32_to_cpu(es->s_blocks_count);
1305         }
1306
1307         buf->f_type = EXT2_SUPER_MAGIC;
1308         buf->f_bsize = sb->s_blocksize;
1309         buf->f_blocks = le32_to_cpu(es->s_blocks_count) - sbi->s_overhead_last;
1310         buf->f_bfree = ext2_count_free_blocks(sb);
1311         es->s_free_blocks_count = cpu_to_le32(buf->f_bfree);
1312         buf->f_bavail = buf->f_bfree - le32_to_cpu(es->s_r_blocks_count);
1313         if (buf->f_bfree < le32_to_cpu(es->s_r_blocks_count))
1314                 buf->f_bavail = 0;
1315         buf->f_files = le32_to_cpu(es->s_inodes_count);
1316         buf->f_ffree = ext2_count_free_inodes(sb);
1317         es->s_free_inodes_count = cpu_to_le32(buf->f_ffree);
1318         buf->f_namelen = EXT2_NAME_LEN;
1319         fsid = le64_to_cpup((void *)es->s_uuid) ^
1320                le64_to_cpup((void *)es->s_uuid + sizeof(u64));
1321         buf->f_fsid.val[0] = fsid & 0xFFFFFFFFUL;
1322         buf->f_fsid.val[1] = (fsid >> 32) & 0xFFFFFFFFUL;
1323         return 0;
1324 }
1325
1326 static int ext2_get_sb(struct file_system_type *fs_type,
1327         int flags, const char *dev_name, void *data, struct vfsmount *mnt)
1328 {
1329         return get_sb_bdev(fs_type, flags, dev_name, data, ext2_fill_super, mnt);
1330 }
1331
1332 #ifdef CONFIG_QUOTA
1333
1334 /* Read data from quotafile - avoid pagecache and such because we cannot afford
1335  * acquiring the locks... As quota files are never truncated and quota code
1336  * itself serializes the operations (and noone else should touch the files)
1337  * we don't have to be afraid of races */
1338 static ssize_t ext2_quota_read(struct super_block *sb, int type, char *data,
1339                                size_t len, loff_t off)
1340 {
1341         struct inode *inode = sb_dqopt(sb)->files[type];
1342         sector_t blk = off >> EXT2_BLOCK_SIZE_BITS(sb);
1343         int err = 0;
1344         int offset = off & (sb->s_blocksize - 1);
1345         int tocopy;
1346         size_t toread;
1347         struct buffer_head tmp_bh;
1348         struct buffer_head *bh;
1349         loff_t i_size = i_size_read(inode);
1350
1351         if (off > i_size)
1352                 return 0;
1353         if (off+len > i_size)
1354                 len = i_size-off;
1355         toread = len;
1356         while (toread > 0) {
1357                 tocopy = sb->s_blocksize - offset < toread ?
1358                                 sb->s_blocksize - offset : toread;
1359
1360                 tmp_bh.b_state = 0;
1361                 tmp_bh.b_size = sb->s_blocksize;
1362                 err = ext2_get_block(inode, blk, &tmp_bh, 0);
1363                 if (err < 0)
1364                         return err;
1365                 if (!buffer_mapped(&tmp_bh))    /* A hole? */
1366                         memset(data, 0, tocopy);
1367                 else {
1368                         bh = sb_bread(sb, tmp_bh.b_blocknr);
1369                         if (!bh)
1370                                 return -EIO;
1371                         memcpy(data, bh->b_data+offset, tocopy);
1372                         brelse(bh);
1373                 }
1374                 offset = 0;
1375                 toread -= tocopy;
1376                 data += tocopy;
1377                 blk++;
1378         }
1379         return len;
1380 }
1381
1382 /* Write to quotafile */
1383 static ssize_t ext2_quota_write(struct super_block *sb, int type,
1384                                 const char *data, size_t len, loff_t off)
1385 {
1386         struct inode *inode = sb_dqopt(sb)->files[type];
1387         sector_t blk = off >> EXT2_BLOCK_SIZE_BITS(sb);
1388         int err = 0;
1389         int offset = off & (sb->s_blocksize - 1);
1390         int tocopy;
1391         size_t towrite = len;
1392         struct buffer_head tmp_bh;
1393         struct buffer_head *bh;
1394
1395         mutex_lock_nested(&inode->i_mutex, I_MUTEX_QUOTA);
1396         while (towrite > 0) {
1397                 tocopy = sb->s_blocksize - offset < towrite ?
1398                                 sb->s_blocksize - offset : towrite;
1399
1400                 tmp_bh.b_state = 0;
1401                 err = ext2_get_block(inode, blk, &tmp_bh, 1);
1402                 if (err < 0)
1403                         goto out;
1404                 if (offset || tocopy != EXT2_BLOCK_SIZE(sb))
1405                         bh = sb_bread(sb, tmp_bh.b_blocknr);
1406                 else
1407                         bh = sb_getblk(sb, tmp_bh.b_blocknr);
1408                 if (!bh) {
1409                         err = -EIO;
1410                         goto out;
1411                 }
1412                 lock_buffer(bh);
1413                 memcpy(bh->b_data+offset, data, tocopy);
1414                 flush_dcache_page(bh->b_page);
1415                 set_buffer_uptodate(bh);
1416                 mark_buffer_dirty(bh);
1417                 unlock_buffer(bh);
1418                 brelse(bh);
1419                 offset = 0;
1420                 towrite -= tocopy;
1421                 data += tocopy;
1422                 blk++;
1423         }
1424 out:
1425         if (len == towrite) {
1426                 mutex_unlock(&inode->i_mutex);
1427                 return err;
1428         }
1429         if (inode->i_size < off+len-towrite)
1430                 i_size_write(inode, off+len-towrite);
1431         inode->i_version++;
1432         inode->i_mtime = inode->i_ctime = CURRENT_TIME;
1433         mark_inode_dirty(inode);
1434         mutex_unlock(&inode->i_mutex);
1435         return len - towrite;
1436 }
1437
1438 #endif
1439
1440 static struct file_system_type ext2_fs_type = {
1441         .owner          = THIS_MODULE,
1442         .name           = "ext2",
1443         .get_sb         = ext2_get_sb,
1444         .kill_sb        = kill_block_super,
1445         .fs_flags       = FS_REQUIRES_DEV,
1446 };
1447
1448 static int __init init_ext2_fs(void)
1449 {
1450         int err = init_ext2_xattr();
1451         if (err)
1452                 return err;
1453         err = init_inodecache();
1454         if (err)
1455                 goto out1;
1456         err = register_filesystem(&ext2_fs_type);
1457         if (err)
1458                 goto out;
1459         return 0;
1460 out:
1461         destroy_inodecache();
1462 out1:
1463         exit_ext2_xattr();
1464         return err;
1465 }
1466
1467 static void __exit exit_ext2_fs(void)
1468 {
1469         unregister_filesystem(&ext2_fs_type);
1470         destroy_inodecache();
1471         exit_ext2_xattr();
1472 }
1473
1474 module_init(init_ext2_fs)
1475 module_exit(exit_ext2_fs)