fs/ext2/super.c

   1 /*
   2  *  linux/fs/ext2/super.c
   3  *
   4  * Copyright (C) 1992, 1993, 1994, 1995
   5  * Remy Card (card@masi.ibp.fr)
   6  * Laboratoire MASI - Institut Blaise Pascal
   7  * Universite Pierre et Marie Curie (Paris VI)
   8  *
   9  *  from
  10  *
  11  *  linux/fs/minix/inode.c
  12  *
  13  *  Copyright (C) 1991, 1992  Linus Torvalds
  14  *
  15  *  Big-endian to little-endian byte-swapping/bitmaps by
  16  *        David S. Miller (davem@caip.rutgers.edu), 1995
  17  */
  18
  19 #include <linux/module.h>
  20 #include <linux/string.h>
  21 #include <linux/fs.h>
  22 #include <linux/slab.h>
  23 #include <linux/init.h>
  24 #include <linux/blkdev.h>
  25 #include <linux/parser.h>
  26 #include <linux/random.h>
  27 #include <linux/buffer_head.h>
  28 #include <linux/exportfs.h>
  29 #include <linux/smp_lock.h>
  30 #include <linux/vfs.h>
  31 #include <linux/seq_file.h>
  32 #include <linux/mount.h>
  33 #include <linux/log2.h>
  34 #include <asm/uaccess.h>
  35 #include "ext2.h"
  36 #include "xattr.h"
  37 #include "acl.h"
  38 #include "xip.h"
  39
  40 static void ext2_sync_super(struct super_block *sb,
  41                             struct ext2_super_block *es);
  42 static int ext2_remount (struct super_block * sb, int * flags, char * data);
  43 static int ext2_statfs (struct dentry * dentry, struct kstatfs * buf);
  44
  45 void ext2_error (struct super_block * sb, const char * function,
  46                  const char * fmt, ...)
  47 {
  48         va_list args;
  49         struct ext2_sb_info *sbi = EXT2_SB(sb);
  50         struct ext2_super_block *es = sbi->s_es;
  51
  52         if (!(sb->s_flags & MS_RDONLY)) {
  53                 sbi->s_mount_state |= EXT2_ERROR_FS;
  54                 es->s_state |= cpu_to_le16(EXT2_ERROR_FS);
  55                 ext2_sync_super(sb, es);
  56         }
  57
  58         va_start(args, fmt);
  59         printk(KERN_CRIT "EXT2-fs error (device %s): %s: ",sb->s_id, function);
  60         vprintk(fmt, args);
  61         printk("\n");
  62         va_end(args);
  63
  64         if (test_opt(sb, ERRORS_PANIC))
  65                 panic("EXT2-fs panic from previous error\n");
  66         if (test_opt(sb, ERRORS_RO)) {
  67                 printk("Remounting filesystem read-only\n");
  68                 sb->s_flags |= MS_RDONLY;
  69         }
  70 }
  71
  72 void ext2_warning (struct super_block * sb, const char * function,
  73                    const char * fmt, ...)
  74 {
  75         va_list args;
  76
  77         va_start(args, fmt);
  78         printk(KERN_WARNING "EXT2-fs warning (device %s): %s: ",
  79                sb->s_id, function);
  80         vprintk(fmt, args);
  81         printk("\n");
  82         va_end(args);
  83 }
  84
  85 void ext2_update_dynamic_rev(struct super_block *sb)
  86 {
  87         struct ext2_super_block *es = EXT2_SB(sb)->s_es;
  88
  89         if (le32_to_cpu(es->s_rev_level) > EXT2_GOOD_OLD_REV)
  90                 return;
  91
  92         ext2_warning(sb, __func__,
  93                      "updating to rev %d because of new feature flag, "
  94                      "running e2fsck is recommended",
  95                      EXT2_DYNAMIC_REV);
  96
  97         es->s_first_ino = cpu_to_le32(EXT2_GOOD_OLD_FIRST_INO);
  98         es->s_inode_size = cpu_to_le16(EXT2_GOOD_OLD_INODE_SIZE);
  99         es->s_rev_level = cpu_to_le32(EXT2_DYNAMIC_REV);
 100         /* leave es->s_feature_*compat flags alone */
 101         /* es->s_uuid will be set by e2fsck if empty */
 102
 103         /*
 104          * The rest of the superblock fields should be zero, and if not it
 105          * means they are likely already in use, so leave them alone.  We
 106          * can leave it up to e2fsck to clean up any inconsistencies there.
 107          */
 108 }
 109
 110 static void ext2_put_super (struct super_block * sb)
 111 {
 112         int db_count;
 113         int i;
 114         struct ext2_sb_info *sbi = EXT2_SB(sb);
 115
 116         ext2_xattr_put_super(sb);
 117         if (!(sb->s_flags & MS_RDONLY)) {
 118                 struct ext2_super_block *es = sbi->s_es;
 119
 120                 es->s_state = cpu_to_le16(sbi->s_mount_state);
 121                 ext2_sync_super(sb, es);
 122         }
 123         db_count = sbi->s_gdb_count;
 124         for (i = 0; i < db_count; i++)
 125                 if (sbi->s_group_desc[i])
 126                         brelse (sbi->s_group_desc[i]);
 127         kfree(sbi->s_group_desc);
 128         kfree(sbi->s_debts);
 129         percpu_counter_destroy(&sbi->s_freeblocks_counter);
 130         percpu_counter_destroy(&sbi->s_freeinodes_counter);
 131         percpu_counter_destroy(&sbi->s_dirs_counter);
 132         brelse (sbi->s_sbh);
 133         sb->s_fs_info = NULL;
 134         kfree(sbi);
 135
 136         return;
 137 }
 138
 139 static struct kmem_cache * ext2_inode_cachep;
 140
 141 static struct inode *ext2_alloc_inode(struct super_block *sb)
 142 {
 143         struct ext2_inode_info *ei;
 144         ei = (struct ext2_inode_info *)kmem_cache_alloc(ext2_inode_cachep, GFP_KERNEL);
 145         if (!ei)
 146                 return NULL;
 147 #ifdef CONFIG_EXT2_FS_POSIX_ACL
 148         ei->i_acl = EXT2_ACL_NOT_CACHED;
 149         ei->i_default_acl = EXT2_ACL_NOT_CACHED;
 150 #endif
 151         ei->i_block_alloc_info = NULL;
 152         ei->vfs_inode.i_version = 1;
 153         return &ei->vfs_inode;
 154 }
 155
 156 static void ext2_destroy_inode(struct inode *inode)
 157 {
 158         kmem_cache_free(ext2_inode_cachep, EXT2_I(inode));
 159 }
 160
 161 static void init_once(struct kmem_cache * cachep, void *foo)
 162 {
 163         struct ext2_inode_info *ei = (struct ext2_inode_info *) foo;
 164
 165         rwlock_init(&ei->i_meta_lock);
 166 #ifdef CONFIG_EXT2_FS_XATTR
 167         init_rwsem(&ei->xattr_sem);
 168 #endif
 169         mutex_init(&ei->truncate_mutex);
 170         inode_init_once(&ei->vfs_inode);
 171 }
 172
 173 static int init_inodecache(void)
 174 {
 175         ext2_inode_cachep = kmem_cache_create("ext2_inode_cache",
 176                                              sizeof(struct ext2_inode_info),
 177                                              0, (SLAB_RECLAIM_ACCOUNT|
 178                                                 SLAB_MEM_SPREAD),
 179                                              init_once);
 180         if (ext2_inode_cachep == NULL)
 181                 return -ENOMEM;
 182         return 0;
 183 }
 184
 185 static void destroy_inodecache(void)
 186 {
 187         kmem_cache_destroy(ext2_inode_cachep);
 188 }
 189
 190 static void ext2_clear_inode(struct inode *inode)
 191 {
 192         struct ext2_block_alloc_info *rsv = EXT2_I(inode)->i_block_alloc_info;
 193 #ifdef CONFIG_EXT2_FS_POSIX_ACL
 194         struct ext2_inode_info *ei = EXT2_I(inode);
 195
 196         if (ei->i_acl && ei->i_acl != EXT2_ACL_NOT_CACHED) {
 197                 posix_acl_release(ei->i_acl);
 198                 ei->i_acl = EXT2_ACL_NOT_CACHED;
 199         }
 200         if (ei->i_default_acl && ei->i_default_acl != EXT2_ACL_NOT_CACHED) {
 201                 posix_acl_release(ei->i_default_acl);
 202                 ei->i_default_acl = EXT2_ACL_NOT_CACHED;
 203         }
 204 #endif
 205         ext2_discard_reservation(inode);
 206         EXT2_I(inode)->i_block_alloc_info = NULL;
 207         if (unlikely(rsv))
 208                 kfree(rsv);
 209 }
 210
 211 static int ext2_show_options(struct seq_file *seq, struct vfsmount *vfs)
 212 {
 213         struct super_block *sb = vfs->mnt_sb;
 214         struct ext2_sb_info *sbi = EXT2_SB(sb);
 215         struct ext2_super_block *es = sbi->s_es;
 216         unsigned long def_mount_opts;
 217
 218         def_mount_opts = le32_to_cpu(es->s_default_mount_opts);
 219
 220         if (sbi->s_sb_block != 1)
 221                 seq_printf(seq, ",sb=%lu", sbi->s_sb_block);
 222         if (test_opt(sb, MINIX_DF))
 223                 seq_puts(seq, ",minixdf");
 224         if (test_opt(sb, GRPID))
 225                 seq_puts(seq, ",grpid");
 226         if (!test_opt(sb, GRPID) && (def_mount_opts & EXT2_DEFM_BSDGROUPS))
 227                 seq_puts(seq, ",nogrpid");
 228         if (sbi->s_resuid != EXT2_DEF_RESUID ||
 229             le16_to_cpu(es->s_def_resuid) != EXT2_DEF_RESUID) {
 230                 seq_printf(seq, ",resuid=%u", sbi->s_resuid);
 231         }
 232         if (sbi->s_resgid != EXT2_DEF_RESGID ||
 233             le16_to_cpu(es->s_def_resgid) != EXT2_DEF_RESGID) {
 234                 seq_printf(seq, ",resgid=%u", sbi->s_resgid);
 235         }
 236         if (test_opt(sb, ERRORS_RO)) {
 237                 int def_errors = le16_to_cpu(es->s_errors);
 238
 239                 if (def_errors == EXT2_ERRORS_PANIC ||
 240                     def_errors == EXT2_ERRORS_CONTINUE) {
 241                         seq_puts(seq, ",errors=remount-ro");
 242                 }
 243         }
 244         if (test_opt(sb, ERRORS_CONT))
 245                 seq_puts(seq, ",errors=continue");
 246         if (test_opt(sb, ERRORS_PANIC))
 247                 seq_puts(seq, ",errors=panic");
 248         if (test_opt(sb, NO_UID32))
 249                 seq_puts(seq, ",nouid32");
 250         if (test_opt(sb, DEBUG))
 251                 seq_puts(seq, ",debug");
 252         if (test_opt(sb, OLDALLOC))
 253                 seq_puts(seq, ",oldalloc");
 254
 255 #ifdef CONFIG_EXT2_FS_XATTR
 256         if (test_opt(sb, XATTR_USER))
 257                 seq_puts(seq, ",user_xattr");
 258         if (!test_opt(sb, XATTR_USER) &&
 259             (def_mount_opts & EXT2_DEFM_XATTR_USER)) {
 260                 seq_puts(seq, ",nouser_xattr");
 261         }
 262 #endif
 263
 264 #ifdef CONFIG_EXT2_FS_POSIX_ACL
 265         if (test_opt(sb, POSIX_ACL))
 266                 seq_puts(seq, ",acl");
 267         if (!test_opt(sb, POSIX_ACL) && (def_mount_opts & EXT2_DEFM_ACL))
 268                 seq_puts(seq, ",noacl");
 269 #endif
 270
 271         if (test_opt(sb, NOBH))
 272                 seq_puts(seq, ",nobh");
 273
 274 #if defined(CONFIG_QUOTA)
 275         if (sbi->s_mount_opt & EXT2_MOUNT_USRQUOTA)
 276                 seq_puts(seq, ",usrquota");
 277
 278         if (sbi->s_mount_opt & EXT2_MOUNT_GRPQUOTA)
 279                 seq_puts(seq, ",grpquota");
 280 #endif
 281
 282 #if defined(CONFIG_EXT2_FS_XIP)
 283         if (sbi->s_mount_opt & EXT2_MOUNT_XIP)
 284                 seq_puts(seq, ",xip");
 285 #endif
 286
 287         if (!test_opt(sb, RESERVATION))
 288                 seq_puts(seq, ",noreservation");
 289
 290         return 0;
 291 }
 292
 293 #ifdef CONFIG_QUOTA
 294 static ssize_t ext2_quota_read(struct super_block *sb, int type, char *data, size_t len, loff_t off);
 295 static ssize_t ext2_quota_write(struct super_block *sb, int type, const char *data, size_t len, loff_t off);
 296 #endif
 297
 298 static const struct super_operations ext2_sops = {
 299         .alloc_inode    = ext2_alloc_inode,
 300         .destroy_inode  = ext2_destroy_inode,
 301         .write_inode    = ext2_write_inode,
 302         .delete_inode   = ext2_delete_inode,
 303         .put_super      = ext2_put_super,
 304         .write_super    = ext2_write_super,
 305         .statfs         = ext2_statfs,
 306         .remount_fs     = ext2_remount,
 307         .clear_inode    = ext2_clear_inode,
 308         .show_options   = ext2_show_options,
 309 #ifdef CONFIG_QUOTA
 310         .quota_read     = ext2_quota_read,
 311         .quota_write    = ext2_quota_write,
 312 #endif
 313 };
 314
 315 static struct inode *ext2_nfs_get_inode(struct super_block *sb,
 316                 u64 ino, u32 generation)
 317 {
 318         struct inode *inode;
 319
 320         if (ino < EXT2_FIRST_INO(sb) && ino != EXT2_ROOT_INO)
 321                 return ERR_PTR(-ESTALE);
 322         if (ino > le32_to_cpu(EXT2_SB(sb)->s_es->s_inodes_count))
 323                 return ERR_PTR(-ESTALE);
 324
 325         /* iget isn't really right if the inode is currently unallocated!!
 326          * ext2_read_inode currently does appropriate checks, but
 327          * it might be "neater" to call ext2_get_inode first and check
 328          * if the inode is valid.....
 329          */
 330         inode = ext2_iget(sb, ino);
 331         if (IS_ERR(inode))
 332                 return ERR_CAST(inode);
 333         if (generation && inode->i_generation != generation) {
 334                 /* we didn't find the right inode.. */
 335                 iput(inode);
 336                 return ERR_PTR(-ESTALE);
 337         }
 338         return inode;
 339 }
 340
 341 static struct dentry *ext2_fh_to_dentry(struct super_block *sb, struct fid *fid,
 342                 int fh_len, int fh_type)
 343 {
 344         return generic_fh_to_dentry(sb, fid, fh_len, fh_type,
 345                                     ext2_nfs_get_inode);
 346 }
 347
 348 static struct dentry *ext2_fh_to_parent(struct super_block *sb, struct fid *fid,
 349                 int fh_len, int fh_type)
 350 {
 351         return generic_fh_to_parent(sb, fid, fh_len, fh_type,
 352                                     ext2_nfs_get_inode);
 353 }
 354
 355 /* Yes, most of these are left as NULL!!
 356  * A NULL value implies the default, which works with ext2-like file
 357  * systems, but can be improved upon.
 358  * Currently only get_parent is required.
 359  */
 360 static const struct export_operations ext2_export_ops = {
 361         .fh_to_dentry = ext2_fh_to_dentry,
 362         .fh_to_parent = ext2_fh_to_parent,
 363         .get_parent = ext2_get_parent,
 364 };
 365
 366 static unsigned long get_sb_block(void **data)
 367 {
 368         unsigned long   sb_block;
 369         char            *options = (char *) *data;
 370
 371         if (!options || strncmp(options, "sb=", 3) != 0)
 372                 return 1;       /* Default location */
 373         options += 3;
 374         sb_block = simple_strtoul(options, &options, 0);
 375         if (*options && *options != ',') {
 376                 printk("EXT2-fs: Invalid sb specification: %s\n",
 377                        (char *) *data);
 378                 return 1;
 379         }
 380         if (*options == ',')
 381                 options++;
 382         *data = (void *) options;
 383         return sb_block;
 384 }
 385
 386 enum {
 387         Opt_bsd_df, Opt_minix_df, Opt_grpid, Opt_nogrpid,
 388         Opt_resgid, Opt_resuid, Opt_sb, Opt_err_cont, Opt_err_panic,
 389         Opt_err_ro, Opt_nouid32, Opt_nocheck, Opt_debug,
 390         Opt_oldalloc, Opt_orlov, Opt_nobh, Opt_user_xattr, Opt_nouser_xattr,
 391         Opt_acl, Opt_noacl, Opt_xip, Opt_ignore, Opt_err, Opt_quota,
 392         Opt_usrquota, Opt_grpquota, Opt_reservation, Opt_noreservation
 393 };
 394
 395 static match_table_t tokens = {
 396         {Opt_bsd_df, "bsddf"},
 397         {Opt_minix_df, "minixdf"},
 398         {Opt_grpid, "grpid"},
 399         {Opt_grpid, "bsdgroups"},
 400         {Opt_nogrpid, "nogrpid"},
 401         {Opt_nogrpid, "sysvgroups"},
 402         {Opt_resgid, "resgid=%u"},
 403         {Opt_resuid, "resuid=%u"},
 404         {Opt_sb, "sb=%u"},
 405         {Opt_err_cont, "errors=continue"},
 406         {Opt_err_panic, "errors=panic"},
 407         {Opt_err_ro, "errors=remount-ro"},
 408         {Opt_nouid32, "nouid32"},
 409         {Opt_nocheck, "check=none"},
 410         {Opt_nocheck, "nocheck"},
 411         {Opt_debug, "debug"},
 412         {Opt_oldalloc, "oldalloc"},
 413         {Opt_orlov, "orlov"},
 414         {Opt_nobh, "nobh"},
 415         {Opt_user_xattr, "user_xattr"},
 416         {Opt_nouser_xattr, "nouser_xattr"},
 417         {Opt_acl, "acl"},
 418         {Opt_noacl, "noacl"},
 419         {Opt_xip, "xip"},
 420         {Opt_grpquota, "grpquota"},
 421         {Opt_ignore, "noquota"},
 422         {Opt_quota, "quota"},
 423         {Opt_usrquota, "usrquota"},
 424         {Opt_reservation, "reservation"},
 425         {Opt_noreservation, "noreservation"},
 426         {Opt_err, NULL}
 427 };
 428
 429 static int parse_options (char * options,
 430                           struct ext2_sb_info *sbi)
 431 {
 432         char * p;
 433         substring_t args[MAX_OPT_ARGS];
 434         int option;
 435
 436         if (!options)
 437                 return 1;
 438
 439         while ((p = strsep (&options, ",")) != NULL) {
 440                 int token;
 441                 if (!*p)
 442                         continue;
 443
 444                 token = match_token(p, tokens, args);
 445                 switch (token) {
 446                 case Opt_bsd_df:
 447                         clear_opt (sbi->s_mount_opt, MINIX_DF);
 448                         break;
 449                 case Opt_minix_df:
 450                         set_opt (sbi->s_mount_opt, MINIX_DF);
 451                         break;
 452                 case Opt_grpid:
 453                         set_opt (sbi->s_mount_opt, GRPID);
 454                         break;
 455                 case Opt_nogrpid:
 456                         clear_opt (sbi->s_mount_opt, GRPID);
 457                         break;
 458                 case Opt_resuid:
 459                         if (match_int(&args[0], &option))
 460                                 return 0;
 461                         sbi->s_resuid = option;
 462                         break;
 463                 case Opt_resgid:
 464                         if (match_int(&args[0], &option))
 465                                 return 0;
 466                         sbi->s_resgid = option;
 467                         break;
 468                 case Opt_sb:
 469                         /* handled by get_sb_block() instead of here */
 470                         /* *sb_block = match_int(&args[0]); */
 471                         break;
 472                 case Opt_err_panic:
 473                         clear_opt (sbi->s_mount_opt, ERRORS_CONT);
 474                         clear_opt (sbi->s_mount_opt, ERRORS_RO);
 475                         set_opt (sbi->s_mount_opt, ERRORS_PANIC);
 476                         break;
 477                 case Opt_err_ro:
 478                         clear_opt (sbi->s_mount_opt, ERRORS_CONT);
 479                         clear_opt (sbi->s_mount_opt, ERRORS_PANIC);
 480                         set_opt (sbi->s_mount_opt, ERRORS_RO);
 481                         break;
 482                 case Opt_err_cont:
 483                         clear_opt (sbi->s_mount_opt, ERRORS_RO);
 484                         clear_opt (sbi->s_mount_opt, ERRORS_PANIC);
 485                         set_opt (sbi->s_mount_opt, ERRORS_CONT);
 486                         break;
 487                 case Opt_nouid32:
 488                         set_opt (sbi->s_mount_opt, NO_UID32);
 489                         break;
 490                 case Opt_nocheck:
 491                         clear_opt (sbi->s_mount_opt, CHECK);
 492                         break;
 493                 case Opt_debug:
 494                         set_opt (sbi->s_mount_opt, DEBUG);
 495                         break;
 496                 case Opt_oldalloc:
 497                         set_opt (sbi->s_mount_opt, OLDALLOC);
 498                         break;
 499                 case Opt_orlov:
 500                         clear_opt (sbi->s_mount_opt, OLDALLOC);
 501                         break;
 502                 case Opt_nobh:
 503                         set_opt (sbi->s_mount_opt, NOBH);
 504                         break;
 505 #ifdef CONFIG_EXT2_FS_XATTR
 506                 case Opt_user_xattr:
 507                         set_opt (sbi->s_mount_opt, XATTR_USER);
 508                         break;
 509                 case Opt_nouser_xattr:
 510                         clear_opt (sbi->s_mount_opt, XATTR_USER);
 511                         break;
 512 #else
 513                 case Opt_user_xattr:
 514                 case Opt_nouser_xattr:
 515                         printk("EXT2 (no)user_xattr options not supported\n");
 516                         break;
 517 #endif
 518 #ifdef CONFIG_EXT2_FS_POSIX_ACL
 519                 case Opt_acl:
 520                         set_opt(sbi->s_mount_opt, POSIX_ACL);
 521                         break;
 522                 case Opt_noacl:
 523                         clear_opt(sbi->s_mount_opt, POSIX_ACL);
 524                         break;
 525 #else
 526                 case Opt_acl:
 527                 case Opt_noacl:
 528                         printk("EXT2 (no)acl options not supported\n");
 529                         break;
 530 #endif
 531                 case Opt_xip:
 532 #ifdef CONFIG_EXT2_FS_XIP
 533                         set_opt (sbi->s_mount_opt, XIP);
 534 #else
 535                         printk("EXT2 xip option not supported\n");
 536 #endif
 537                         break;
 538
 539 #if defined(CONFIG_QUOTA)
 540                 case Opt_quota:
 541                 case Opt_usrquota:
 542                         set_opt(sbi->s_mount_opt, USRQUOTA);
 543                         break;
 544
 545                 case Opt_grpquota:
 546                         set_opt(sbi->s_mount_opt, GRPQUOTA);
 547                         break;
 548 #else
 549                 case Opt_quota:
 550                 case Opt_usrquota:
 551                 case Opt_grpquota:
 552                         printk(KERN_ERR
 553                                 "EXT2-fs: quota operations not supported.\n");
 554
 555                         break;
 556 #endif
 557
 558                 case Opt_reservation:
 559                         set_opt(sbi->s_mount_opt, RESERVATION);
 560                         printk("reservations ON\n");
 561                         break;
 562                 case Opt_noreservation:
 563                         clear_opt(sbi->s_mount_opt, RESERVATION);
 564                         printk("reservations OFF\n");
 565                         break;
 566                 case Opt_ignore:
 567                         break;
 568                 default:
 569                         return 0;
 570                 }
 571         }
 572         return 1;
 573 }
 574
 575 static int ext2_setup_super (struct super_block * sb,
 576                               struct ext2_super_block * es,
 577                               int read_only)
 578 {
 579         int res = 0;
 580         struct ext2_sb_info *sbi = EXT2_SB(sb);
 581
 582         if (le32_to_cpu(es->s_rev_level) > EXT2_MAX_SUPP_REV) {
 583                 printk ("EXT2-fs warning: revision level too high, "
 584                         "forcing read-only mode\n");
 585                 res = MS_RDONLY;
 586         }
 587         if (read_only)
 588                 return res;
 589         if (!(sbi->s_mount_state & EXT2_VALID_FS))
 590                 printk ("EXT2-fs warning: mounting unchecked fs, "
 591                         "running e2fsck is recommended\n");
 592         else if ((sbi->s_mount_state & EXT2_ERROR_FS))
 593                 printk ("EXT2-fs warning: mounting fs with errors, "
 594                         "running e2fsck is recommended\n");
 595         else if ((__s16) le16_to_cpu(es->s_max_mnt_count) >= 0 &&
 596                  le16_to_cpu(es->s_mnt_count) >=
 597                  (unsigned short) (__s16) le16_to_cpu(es->s_max_mnt_count))
 598                 printk ("EXT2-fs warning: maximal mount count reached, "
 599                         "running e2fsck is recommended\n");
 600         else if (le32_to_cpu(es->s_checkinterval) &&
 601                 (le32_to_cpu(es->s_lastcheck) + le32_to_cpu(es->s_checkinterval) <= get_seconds()))
 602                 printk ("EXT2-fs warning: checktime reached, "
 603                         "running e2fsck is recommended\n");
 604         if (!le16_to_cpu(es->s_max_mnt_count))
 605                 es->s_max_mnt_count = cpu_to_le16(EXT2_DFL_MAX_MNT_COUNT);
 606         le16_add_cpu(&es->s_mnt_count, 1);
 607         ext2_write_super(sb);
 608         if (test_opt (sb, DEBUG))
 609                 printk ("[EXT II FS %s, %s, bs=%lu, fs=%lu, gc=%lu, "
 610                         "bpg=%lu, ipg=%lu, mo=%04lx]\n",
 611                         EXT2FS_VERSION, EXT2FS_DATE, sb->s_blocksize,
 612                         sbi->s_frag_size,
 613                         sbi->s_groups_count,
 614                         EXT2_BLOCKS_PER_GROUP(sb),
 615                         EXT2_INODES_PER_GROUP(sb),
 616                         sbi->s_mount_opt);
 617         return res;
 618 }
 619
 620 static int ext2_check_descriptors(struct super_block *sb)
 621 {
 622         int i;
 623         struct ext2_sb_info *sbi = EXT2_SB(sb);
 624
 625         ext2_debug ("Checking group descriptors");
 626
 627         for (i = 0; i < sbi->s_groups_count; i++) {
 628                 struct ext2_group_desc *gdp = ext2_get_group_desc(sb, i, NULL);
 629                 ext2_fsblk_t first_block = ext2_group_first_block_no(sb, i);
 630                 ext2_fsblk_t last_block;
 631
 632                 if (i == sbi->s_groups_count - 1)
 633                         last_block = le32_to_cpu(sbi->s_es->s_blocks_count) - 1;
 634                 else
 635                         last_block = first_block +
 636                                 (EXT2_BLOCKS_PER_GROUP(sb) - 1);
 637
 638                 if (le32_to_cpu(gdp->bg_block_bitmap) < first_block ||
 639                     le32_to_cpu(gdp->bg_block_bitmap) > last_block)
 640                 {
 641                         ext2_error (sb, "ext2_check_descriptors",
 642                                     "Block bitmap for group %d"
 643                                     " not in group (block %lu)!",
 644                                     i, (unsigned long) le32_to_cpu(gdp->bg_block_bitmap));
 645                         return 0;
 646                 }
 647                 if (le32_to_cpu(gdp->bg_inode_bitmap) < first_block ||
 648                     le32_to_cpu(gdp->bg_inode_bitmap) > last_block)
 649                 {
 650                         ext2_error (sb, "ext2_check_descriptors",
 651                                     "Inode bitmap for group %d"
 652                                     " not in group (block %lu)!",
 653                                     i, (unsigned long) le32_to_cpu(gdp->bg_inode_bitmap));
 654                         return 0;
 655                 }
 656                 if (le32_to_cpu(gdp->bg_inode_table) < first_block ||
 657                     le32_to_cpu(gdp->bg_inode_table) + sbi->s_itb_per_group - 1 >
 658                     last_block)
 659                 {
 660                         ext2_error (sb, "ext2_check_descriptors",
 661                                     "Inode table for group %d"
 662                                     " not in group (block %lu)!",
 663                                     i, (unsigned long) le32_to_cpu(gdp->bg_inode_table));
 664                         return 0;
 665                 }
 666         }
 667         return 1;
 668 }
 669
 670 /*
 671  * Maximal file size.  There is a direct, and {,double-,triple-}indirect
 672  * block limit, and also a limit of (2^32 - 1) 512-byte sectors in i_blocks.
 673  * We need to be 1 filesystem block less than the 2^32 sector limit.
 674  */
 675 static loff_t ext2_max_size(int bits)
 676 {
 677         loff_t res = EXT2_NDIR_BLOCKS;
 678         int meta_blocks;
 679         loff_t upper_limit;
 680
 681         /* This is calculated to be the largest file size for a
 682          * dense, file such that the total number of
 683          * sectors in the file, including data and all indirect blocks,
 684          * does not exceed 2^32 -1
 685          * __u32 i_blocks representing the total number of
 686          * 512 bytes blocks of the file
 687          */
 688         upper_limit = (1LL << 32) - 1;
 689
 690         /* total blocks in file system block size */
 691         upper_limit >>= (bits - 9);
 692
 693
 694         /* indirect blocks */
 695         meta_blocks = 1;
 696         /* double indirect blocks */
 697         meta_blocks += 1 + (1LL << (bits-2));
 698         /* tripple indirect blocks */
 699         meta_blocks += 1 + (1LL << (bits-2)) + (1LL << (2*(bits-2)));
 700
 701         upper_limit -= meta_blocks;
 702         upper_limit <<= bits;
 703
 704         res += 1LL << (bits-2);
 705         res += 1LL << (2*(bits-2));
 706         res += 1LL << (3*(bits-2));
 707         res <<= bits;
 708         if (res > upper_limit)
 709                 res = upper_limit;
 710
 711         if (res > MAX_LFS_FILESIZE)
 712                 res = MAX_LFS_FILESIZE;
 713
 714         return res;
 715 }
 716
 717 static unsigned long descriptor_loc(struct super_block *sb,
 718                                     unsigned long logic_sb_block,
 719                                     int nr)
 720 {
 721         struct ext2_sb_info *sbi = EXT2_SB(sb);
 722         unsigned long bg, first_meta_bg;
 723         int has_super = 0;
 724
 725         first_meta_bg = le32_to_cpu(sbi->s_es->s_first_meta_bg);
 726
 727         if (!EXT2_HAS_INCOMPAT_FEATURE(sb, EXT2_FEATURE_INCOMPAT_META_BG) ||
 728             nr < first_meta_bg)
 729                 return (logic_sb_block + nr + 1);
 730         bg = sbi->s_desc_per_block * nr;
 731         if (ext2_bg_has_super(sb, bg))
 732                 has_super = 1;
 733
 734         return ext2_group_first_block_no(sb, bg) + has_super;
 735 }
 736
 737 static int ext2_fill_super(struct super_block *sb, void *data, int silent)
 738 {
 739         struct buffer_head * bh;
 740         struct ext2_sb_info * sbi;
 741         struct ext2_super_block * es;
 742         struct inode *root;
 743         unsigned long block;
 744         unsigned long sb_block = get_sb_block(&data);
 745         unsigned long logic_sb_block;
 746         unsigned long offset = 0;
 747         unsigned long def_mount_opts;
 748         long ret = -EINVAL;
 749         int blocksize = BLOCK_SIZE;
 750         int db_count;
 751         int i, j;
 752         __le32 features;
 753         int err;
 754
 755         sbi = kzalloc(sizeof(*sbi), GFP_KERNEL);
 756         if (!sbi)
 757                 return -ENOMEM;
 758         sb->s_fs_info = sbi;
 759         sbi->s_sb_block = sb_block;
 760
 761         /*
 762          * See what the current blocksize for the device is, and
 763          * use that as the blocksize.  Otherwise (or if the blocksize
 764          * is smaller than the default) use the default.
 765          * This is important for devices that have a hardware
 766          * sectorsize that is larger than the default.
 767          */
 768         blocksize = sb_min_blocksize(sb, BLOCK_SIZE);
 769         if (!blocksize) {
 770                 printk ("EXT2-fs: unable to set blocksize\n");
 771                 goto failed_sbi;
 772         }
 773
 774         /*
 775          * If the superblock doesn't start on a hardware sector boundary,
 776          * calculate the offset.
 777          */
 778         if (blocksize != BLOCK_SIZE) {
 779                 logic_sb_block = (sb_block*BLOCK_SIZE) / blocksize;
 780                 offset = (sb_block*BLOCK_SIZE) % blocksize;
 781         } else {
 782                 logic_sb_block = sb_block;
 783         }
 784
 785         if (!(bh = sb_bread(sb, logic_sb_block))) {
 786                 printk ("EXT2-fs: unable to read superblock\n");
 787                 goto failed_sbi;
 788         }
 789         /*
 790          * Note: s_es must be initialized as soon as possible because
 791          *       some ext2 macro-instructions depend on its value
 792          */
 793         es = (struct ext2_super_block *) (((char *)bh->b_data) + offset);
 794         sbi->s_es = es;
 795         sb->s_magic = le16_to_cpu(es->s_magic);
 796
 797         if (sb->s_magic != EXT2_SUPER_MAGIC)
 798                 goto cantfind_ext2;
 799
 800         /* Set defaults before we parse the mount options */
 801         def_mount_opts = le32_to_cpu(es->s_default_mount_opts);
 802         if (def_mount_opts & EXT2_DEFM_DEBUG)
 803                 set_opt(sbi->s_mount_opt, DEBUG);
 804         if (def_mount_opts & EXT2_DEFM_BSDGROUPS)
 805                 set_opt(sbi->s_mount_opt, GRPID);
 806         if (def_mount_opts & EXT2_DEFM_UID16)
 807                 set_opt(sbi->s_mount_opt, NO_UID32);
 808 #ifdef CONFIG_EXT2_FS_XATTR
 809         if (def_mount_opts & EXT2_DEFM_XATTR_USER)
 810                 set_opt(sbi->s_mount_opt, XATTR_USER);
 811 #endif
 812 #ifdef CONFIG_EXT2_FS_POSIX_ACL
 813         if (def_mount_opts & EXT2_DEFM_ACL)
 814                 set_opt(sbi->s_mount_opt, POSIX_ACL);
 815 #endif
 816
 817         if (le16_to_cpu(sbi->s_es->s_errors) == EXT2_ERRORS_PANIC)
 818                 set_opt(sbi->s_mount_opt, ERRORS_PANIC);
 819         else if (le16_to_cpu(sbi->s_es->s_errors) == EXT2_ERRORS_CONTINUE)
 820                 set_opt(sbi->s_mount_opt, ERRORS_CONT);
 821         else
 822                 set_opt(sbi->s_mount_opt, ERRORS_RO);
 823
 824         sbi->s_resuid = le16_to_cpu(es->s_def_resuid);
 825         sbi->s_resgid = le16_to_cpu(es->s_def_resgid);
 826
 827         set_opt(sbi->s_mount_opt, RESERVATION);
 828
 829         if (!parse_options ((char *) data, sbi))
 830                 goto failed_mount;
 831
 832         sb->s_flags = (sb->s_flags & ~MS_POSIXACL) |
 833                 ((EXT2_SB(sb)->s_mount_opt & EXT2_MOUNT_POSIX_ACL) ?
 834                  MS_POSIXACL : 0);
 835
 836         ext2_xip_verify_sb(sb); /* see if bdev supports xip, unset
 837                                     EXT2_MOUNT_XIP if not */
 838
 839         if (le32_to_cpu(es->s_rev_level) == EXT2_GOOD_OLD_REV &&
 840             (EXT2_HAS_COMPAT_FEATURE(sb, ~0U) ||
 841              EXT2_HAS_RO_COMPAT_FEATURE(sb, ~0U) ||
 842              EXT2_HAS_INCOMPAT_FEATURE(sb, ~0U)))
 843                 printk("EXT2-fs warning: feature flags set on rev 0 fs, "
 844                        "running e2fsck is recommended\n");
 845         /*
 846          * Check feature flags regardless of the revision level, since we
 847          * previously didn't change the revision level when setting the flags,
 848          * so there is a chance incompat flags are set on a rev 0 filesystem.
 849          */
 850         features = EXT2_HAS_INCOMPAT_FEATURE(sb, ~EXT2_FEATURE_INCOMPAT_SUPP);
 851         if (features) {
 852                 printk("EXT2-fs: %s: couldn't mount because of "
 853                        "unsupported optional features (%x).\n",
 854                        sb->s_id, le32_to_cpu(features));
 855                 goto failed_mount;
 856         }
 857         if (!(sb->s_flags & MS_RDONLY) &&
 858             (features = EXT2_HAS_RO_COMPAT_FEATURE(sb, ~EXT2_FEATURE_RO_COMPAT_SUPP))){
 859                 printk("EXT2-fs: %s: couldn't mount RDWR because of "
 860                        "unsupported optional features (%x).\n",
 861                        sb->s_id, le32_to_cpu(features));
 862                 goto failed_mount;
 863         }
 864
 865         blocksize = BLOCK_SIZE << le32_to_cpu(sbi->s_es->s_log_block_size);
 866
 867         if (ext2_use_xip(sb) && blocksize != PAGE_SIZE) {
 868                 if (!silent)
 869                         printk("XIP: Unsupported blocksize\n");
 870                 goto failed_mount;
 871         }
 872
 873         /* If the blocksize doesn't match, re-read the thing.. */
 874         if (sb->s_blocksize != blocksize) {
 875                 brelse(bh);
 876
 877                 if (!sb_set_blocksize(sb, blocksize)) {
 878                         printk(KERN_ERR "EXT2-fs: blocksize too small for device.\n");
 879                         goto failed_sbi;
 880                 }
 881
 882                 logic_sb_block = (sb_block*BLOCK_SIZE) / blocksize;
 883                 offset = (sb_block*BLOCK_SIZE) % blocksize;
 884                 bh = sb_bread(sb, logic_sb_block);
 885                 if(!bh) {
 886                         printk("EXT2-fs: Couldn't read superblock on "
 887                                "2nd try.\n");
 888                         goto failed_sbi;
 889                 }
 890                 es = (struct ext2_super_block *) (((char *)bh->b_data) + offset);
 891                 sbi->s_es = es;
 892                 if (es->s_magic != cpu_to_le16(EXT2_SUPER_MAGIC)) {
 893                         printk ("EXT2-fs: Magic mismatch, very weird !\n");
 894                         goto failed_mount;
 895                 }
 896         }
 897
 898         sb->s_maxbytes = ext2_max_size(sb->s_blocksize_bits);
 899
 900         if (le32_to_cpu(es->s_rev_level) == EXT2_GOOD_OLD_REV) {
 901                 sbi->s_inode_size = EXT2_GOOD_OLD_INODE_SIZE;
 902                 sbi->s_first_ino = EXT2_GOOD_OLD_FIRST_INO;
 903         } else {
 904                 sbi->s_inode_size = le16_to_cpu(es->s_inode_size);
 905                 sbi->s_first_ino = le32_to_cpu(es->s_first_ino);
 906                 if ((sbi->s_inode_size < EXT2_GOOD_OLD_INODE_SIZE) ||
 907                     !is_power_of_2(sbi->s_inode_size) ||
 908                     (sbi->s_inode_size > blocksize)) {
 909                         printk ("EXT2-fs: unsupported inode size: %d\n",
 910                                 sbi->s_inode_size);
 911                         goto failed_mount;
 912                 }
 913         }
 914
 915         sbi->s_frag_size = EXT2_MIN_FRAG_SIZE <<
 916                                    le32_to_cpu(es->s_log_frag_size);
 917         if (sbi->s_frag_size == 0)
 918                 goto cantfind_ext2;
 919         sbi->s_frags_per_block = sb->s_blocksize / sbi->s_frag_size;
 920
 921         sbi->s_blocks_per_group = le32_to_cpu(es->s_blocks_per_group);
 922         sbi->s_frags_per_group = le32_to_cpu(es->s_frags_per_group);
 923         sbi->s_inodes_per_group = le32_to_cpu(es->s_inodes_per_group);
 924
 925         if (EXT2_INODE_SIZE(sb) == 0)
 926                 goto cantfind_ext2;
 927         sbi->s_inodes_per_block = sb->s_blocksize / EXT2_INODE_SIZE(sb);
 928         if (sbi->s_inodes_per_block == 0 || sbi->s_inodes_per_group == 0)
 929                 goto cantfind_ext2;
 930         sbi->s_itb_per_group = sbi->s_inodes_per_group /
 931                                         sbi->s_inodes_per_block;
 932         sbi->s_desc_per_block = sb->s_blocksize /
 933                                         sizeof (struct ext2_group_desc);
 934         sbi->s_sbh = bh;
 935         sbi->s_mount_state = le16_to_cpu(es->s_state);
 936         sbi->s_addr_per_block_bits =
 937                 ilog2 (EXT2_ADDR_PER_BLOCK(sb));
 938         sbi->s_desc_per_block_bits =
 939                 ilog2 (EXT2_DESC_PER_BLOCK(sb));
 940
 941         if (sb->s_magic != EXT2_SUPER_MAGIC)
 942                 goto cantfind_ext2;
 943
 944         if (sb->s_blocksize != bh->b_size) {
 945                 if (!silent)
 946                         printk ("VFS: Unsupported blocksize on dev "
 947                                 "%s.\n", sb->s_id);
 948                 goto failed_mount;
 949         }
 950
 951         if (sb->s_blocksize != sbi->s_frag_size) {
 952                 printk ("EXT2-fs: fragsize %lu != blocksize %lu (not supported yet)\n",
 953                         sbi->s_frag_size, sb->s_blocksize);
 954                 goto failed_mount;
 955         }
 956
 957         if (sbi->s_blocks_per_group > sb->s_blocksize * 8) {
 958                 printk ("EXT2-fs: #blocks per group too big: %lu\n",
 959                         sbi->s_blocks_per_group);
 960                 goto failed_mount;
 961         }
 962         if (sbi->s_frags_per_group > sb->s_blocksize * 8) {
 963                 printk ("EXT2-fs: #fragments per group too big: %lu\n",
 964                         sbi->s_frags_per_group);
 965                 goto failed_mount;
 966         }
 967         if (sbi->s_inodes_per_group > sb->s_blocksize * 8) {
 968                 printk ("EXT2-fs: #inodes per group too big: %lu\n",
 969                         sbi->s_inodes_per_group);
 970                 goto failed_mount;
 971         }
 972
 973         if (EXT2_BLOCKS_PER_GROUP(sb) == 0)
 974                 goto cantfind_ext2;
 975         sbi->s_groups_count = ((le32_to_cpu(es->s_blocks_count) -
 976                                 le32_to_cpu(es->s_first_data_block) - 1)
 977                                         / EXT2_BLOCKS_PER_GROUP(sb)) + 1;
 978         db_count = (sbi->s_groups_count + EXT2_DESC_PER_BLOCK(sb) - 1) /
 979                    EXT2_DESC_PER_BLOCK(sb);
 980         sbi->s_group_desc = kmalloc (db_count * sizeof (struct buffer_head *), GFP_KERNEL);
 981         if (sbi->s_group_desc == NULL) {
 982                 printk ("EXT2-fs: not enough memory\n");
 983                 goto failed_mount;
 984         }
 985         bgl_lock_init(&sbi->s_blockgroup_lock);
 986         sbi->s_debts = kcalloc(sbi->s_groups_count, sizeof(*sbi->s_debts), GFP_KERNEL);
 987         if (!sbi->s_debts) {
 988                 printk ("EXT2-fs: not enough memory\n");
 989                 goto failed_mount_group_desc;
 990         }
 991         for (i = 0; i < db_count; i++) {
 992                 block = descriptor_loc(sb, logic_sb_block, i);
 993                 sbi->s_group_desc[i] = sb_bread(sb, block);
 994                 if (!sbi->s_group_desc[i]) {
 995                         for (j = 0; j < i; j++)
 996                                 brelse (sbi->s_group_desc[j]);
 997                         printk ("EXT2-fs: unable to read group descriptors\n");
 998                         goto failed_mount_group_desc;
 999                 }
1000         }
1001         if (!ext2_check_descriptors (sb)) {
1002                 printk ("EXT2-fs: group descriptors corrupted!\n");
1003                 goto failed_mount2;
1004         }
1005         sbi->s_gdb_count = db_count;
1006         get_random_bytes(&sbi->s_next_generation, sizeof(u32));
1007         spin_lock_init(&sbi->s_next_gen_lock);
1008
1009         /* per fileystem reservation list head & lock */
1010         spin_lock_init(&sbi->s_rsv_window_lock);
1011         sbi->s_rsv_window_root = RB_ROOT;
1012         /*
1013          * Add a single, static dummy reservation to the start of the
1014          * reservation window list --- it gives us a placeholder for
1015          * append-at-start-of-list which makes the allocation logic
1016          * _much_ simpler.
1017          */
1018         sbi->s_rsv_window_head.rsv_start = EXT2_RESERVE_WINDOW_NOT_ALLOCATED;
1019         sbi->s_rsv_window_head.rsv_end = EXT2_RESERVE_WINDOW_NOT_ALLOCATED;
1020         sbi->s_rsv_window_head.rsv_alloc_hit = 0;
1021         sbi->s_rsv_window_head.rsv_goal_size = 0;
1022         ext2_rsv_window_add(sb, &sbi->s_rsv_window_head);
1023
1024         err = percpu_counter_init(&sbi->s_freeblocks_counter,
1025                                 ext2_count_free_blocks(sb));
1026         if (!err) {
1027                 err = percpu_counter_init(&sbi->s_freeinodes_counter,
1028                                 ext2_count_free_inodes(sb));
1029         }
1030         if (!err) {
1031                 err = percpu_counter_init(&sbi->s_dirs_counter,
1032                                 ext2_count_dirs(sb));
1033         }
1034         if (err) {
1035                 printk(KERN_ERR "EXT2-fs: insufficient memory\n");
1036                 goto failed_mount3;
1037         }
1038         /*
1039          * set up enough so that it can read an inode
1040          */
1041         sb->s_op = &ext2_sops;
1042         sb->s_export_op = &ext2_export_ops;
1043         sb->s_xattr = ext2_xattr_handlers;
1044         root = ext2_iget(sb, EXT2_ROOT_INO);
1045         if (IS_ERR(root)) {
1046                 ret = PTR_ERR(root);
1047                 goto failed_mount3;
1048         }
1049         if (!S_ISDIR(root->i_mode) || !root->i_blocks || !root->i_size) {
1050                 iput(root);
1051                 printk(KERN_ERR "EXT2-fs: corrupt root inode, run e2fsck\n");
1052                 goto failed_mount3;
1053         }
1054
1055         sb->s_root = d_alloc_root(root);
1056         if (!sb->s_root) {
1057                 iput(root);
1058                 printk(KERN_ERR "EXT2-fs: get root inode failed\n");
1059                 ret = -ENOMEM;
1060                 goto failed_mount3;
1061         }
1062         if (EXT2_HAS_COMPAT_FEATURE(sb, EXT3_FEATURE_COMPAT_HAS_JOURNAL))
1063                 ext2_warning(sb, __func__,
1064                         "mounting ext3 filesystem as ext2");
1065         ext2_setup_super (sb, es, sb->s_flags & MS_RDONLY);
1066         return 0;
1067
1068 cantfind_ext2:
1069         if (!silent)
1070                 printk("VFS: Can't find an ext2 filesystem on dev %s.\n",
1071                        sb->s_id);
1072         goto failed_mount;
1073 failed_mount3:
1074         percpu_counter_destroy(&sbi->s_freeblocks_counter);
1075         percpu_counter_destroy(&sbi->s_freeinodes_counter);
1076         percpu_counter_destroy(&sbi->s_dirs_counter);
1077 failed_mount2:
1078         for (i = 0; i < db_count; i++)
1079                 brelse(sbi->s_group_desc[i]);
1080 failed_mount_group_desc:
1081         kfree(sbi->s_group_desc);
1082         kfree(sbi->s_debts);
1083 failed_mount:
1084         brelse(bh);
1085 failed_sbi:
1086         sb->s_fs_info = NULL;
1087         kfree(sbi);
1088         return ret;
1089 }
1090
1091 static void ext2_commit_super (struct super_block * sb,
1092                                struct ext2_super_block * es)
1093 {
1094         es->s_wtime = cpu_to_le32(get_seconds());
1095         mark_buffer_dirty(EXT2_SB(sb)->s_sbh);
1096         sb->s_dirt = 0;
1097 }
1098
1099 static void ext2_sync_super(struct super_block *sb, struct ext2_super_block *es)
1100 {
1101         es->s_free_blocks_count = cpu_to_le32(ext2_count_free_blocks(sb));
1102         es->s_free_inodes_count = cpu_to_le32(ext2_count_free_inodes(sb));
1103         es->s_wtime = cpu_to_le32(get_seconds());
1104         mark_buffer_dirty(EXT2_SB(sb)->s_sbh);
1105         sync_dirty_buffer(EXT2_SB(sb)->s_sbh);
1106         sb->s_dirt = 0;
1107 }
1108
1109 /*
1110  * In the second extended file system, it is not necessary to
1111  * write the super block since we use a mapping of the
1112  * disk super block in a buffer.
1113  *
1114  * However, this function is still used to set the fs valid
1115  * flags to 0.  We need to set this flag to 0 since the fs
1116  * may have been checked while mounted and e2fsck may have
1117  * set s_state to EXT2_VALID_FS after some corrections.
1118  */
1119
1120 void ext2_write_super (struct super_block * sb)
1121 {
1122         struct ext2_super_block * es;
1123         lock_kernel();
1124         if (!(sb->s_flags & MS_RDONLY)) {
1125                 es = EXT2_SB(sb)->s_es;
1126
1127                 if (es->s_state & cpu_to_le16(EXT2_VALID_FS)) {
1128                         ext2_debug ("setting valid to 0\n");
1129                         es->s_state &= cpu_to_le16(~EXT2_VALID_FS);
1130                         es->s_free_blocks_count = cpu_to_le32(ext2_count_free_blocks(sb));
1131                         es->s_free_inodes_count = cpu_to_le32(ext2_count_free_inodes(sb));
1132                         es->s_mtime = cpu_to_le32(get_seconds());
1133                         ext2_sync_super(sb, es);
1134                 } else
1135                         ext2_commit_super (sb, es);
1136         }
1137         sb->s_dirt = 0;
1138         unlock_kernel();
1139 }
1140
1141 static int ext2_remount (struct super_block * sb, int * flags, char * data)
1142 {
1143         struct ext2_sb_info * sbi = EXT2_SB(sb);
1144         struct ext2_super_block * es;
1145         unsigned long old_mount_opt = sbi->s_mount_opt;
1146         struct ext2_mount_options old_opts;
1147         unsigned long old_sb_flags;
1148         int err;
1149
1150         /* Store the old options */
1151         old_sb_flags = sb->s_flags;
1152         old_opts.s_mount_opt = sbi->s_mount_opt;
1153         old_opts.s_resuid = sbi->s_resuid;
1154         old_opts.s_resgid = sbi->s_resgid;
1155
1156         /*
1157          * Allow the "check" option to be passed as a remount option.
1158          */
1159         if (!parse_options (data, sbi)) {
1160                 err = -EINVAL;
1161                 goto restore_opts;
1162         }
1163
1164         sb->s_flags = (sb->s_flags & ~MS_POSIXACL) |
1165                 ((sbi->s_mount_opt & EXT2_MOUNT_POSIX_ACL) ? MS_POSIXACL : 0);
1166
1167         ext2_xip_verify_sb(sb); /* see if bdev supports xip, unset
1168                                     EXT2_MOUNT_XIP if not */
1169
1170         if ((ext2_use_xip(sb)) && (sb->s_blocksize != PAGE_SIZE)) {
1171                 printk("XIP: Unsupported blocksize\n");
1172                 err = -EINVAL;
1173                 goto restore_opts;
1174         }
1175
1176         es = sbi->s_es;
1177         if (((sbi->s_mount_opt & EXT2_MOUNT_XIP) !=
1178             (old_mount_opt & EXT2_MOUNT_XIP)) &&
1179             invalidate_inodes(sb))
1180                 ext2_warning(sb, __func__, "busy inodes while remounting "\
1181                              "xip remain in cache (no functional problem)");
1182         if ((*flags & MS_RDONLY) == (sb->s_flags & MS_RDONLY))
1183                 return 0;
1184         if (*flags & MS_RDONLY) {
1185                 if (le16_to_cpu(es->s_state) & EXT2_VALID_FS ||
1186                     !(sbi->s_mount_state & EXT2_VALID_FS))
1187                         return 0;
1188                 /*
1189                  * OK, we are remounting a valid rw partition rdonly, so set
1190                  * the rdonly flag and then mark the partition as valid again.
1191                  */
1192                 es->s_state = cpu_to_le16(sbi->s_mount_state);
1193                 es->s_mtime = cpu_to_le32(get_seconds());
1194         } else {
1195                 __le32 ret = EXT2_HAS_RO_COMPAT_FEATURE(sb,
1196                                                ~EXT2_FEATURE_RO_COMPAT_SUPP);
1197                 if (ret) {
1198                         printk("EXT2-fs: %s: couldn't remount RDWR because of "
1199                                "unsupported optional features (%x).\n",
1200                                sb->s_id, le32_to_cpu(ret));
1201                         err = -EROFS;
1202                         goto restore_opts;
1203                 }
1204                 /*
1205                  * Mounting a RDONLY partition read-write, so reread and
1206                  * store the current valid flag.  (It may have been changed
1207                  * by e2fsck since we originally mounted the partition.)
1208                  */
1209                 sbi->s_mount_state = le16_to_cpu(es->s_state);
1210                 if (!ext2_setup_super (sb, es, 0))
1211                         sb->s_flags &= ~MS_RDONLY;
1212         }
1213         ext2_sync_super(sb, es);
1214         return 0;
1215 restore_opts:
1216         sbi->s_mount_opt = old_opts.s_mount_opt;
1217         sbi->s_resuid = old_opts.s_resuid;
1218         sbi->s_resgid = old_opts.s_resgid;
1219         sb->s_flags = old_sb_flags;
1220         return err;
1221 }
1222
1223 static int ext2_statfs (struct dentry * dentry, struct kstatfs * buf)
1224 {
1225         struct super_block *sb = dentry->d_sb;
1226         struct ext2_sb_info *sbi = EXT2_SB(sb);
1227         struct ext2_super_block *es = sbi->s_es;
1228         u64 fsid;
1229
1230         if (test_opt (sb, MINIX_DF))
1231                 sbi->s_overhead_last = 0;
1232         else if (sbi->s_blocks_last != le32_to_cpu(es->s_blocks_count)) {
1233                 unsigned long i, overhead = 0;
1234                 smp_rmb();
1235
1236                 /*
1237                  * Compute the overhead (FS structures). This is constant
1238                  * for a given filesystem unless the number of block groups
1239                  * changes so we cache the previous value until it does.
1240                  */
1241
1242                 /*
1243                  * All of the blocks before first_data_block are
1244                  * overhead
1245                  */
1246                 overhead = le32_to_cpu(es->s_first_data_block);
1247
1248                 /*
1249                  * Add the overhead attributed to the superblock and
1250                  * block group descriptors.  If the sparse superblocks
1251                  * feature is turned on, then not all groups have this.
1252                  */
1253                 for (i = 0; i < sbi->s_groups_count; i++)
1254                         overhead += ext2_bg_has_super(sb, i) +
1255                                 ext2_bg_num_gdb(sb, i);
1256
1257                 /*
1258                  * Every block group has an inode bitmap, a block
1259                  * bitmap, and an inode table.
1260                  */
1261                 overhead += (sbi->s_groups_count *
1262                              (2 + sbi->s_itb_per_group));
1263                 sbi->s_overhead_last = overhead;
1264                 smp_wmb();
1265                 sbi->s_blocks_last = le32_to_cpu(es->s_blocks_count);
1266         }
1267
1268         buf->f_type = EXT2_SUPER_MAGIC;
1269         buf->f_bsize = sb->s_blocksize;
1270         buf->f_blocks = le32_to_cpu(es->s_blocks_count) - sbi->s_overhead_last;
1271         buf->f_bfree = ext2_count_free_blocks(sb);
1272         es->s_free_blocks_count = cpu_to_le32(buf->f_bfree);
1273         buf->f_bavail = buf->f_bfree - le32_to_cpu(es->s_r_blocks_count);
1274         if (buf->f_bfree < le32_to_cpu(es->s_r_blocks_count))
1275                 buf->f_bavail = 0;
1276         buf->f_files = le32_to_cpu(es->s_inodes_count);
1277         buf->f_ffree = ext2_count_free_inodes(sb);
1278         es->s_free_inodes_count = cpu_to_le32(buf->f_ffree);
1279         buf->f_namelen = EXT2_NAME_LEN;
1280         fsid = le64_to_cpup((void *)es->s_uuid) ^
1281                le64_to_cpup((void *)es->s_uuid + sizeof(u64));
1282         buf->f_fsid.val[0] = fsid & 0xFFFFFFFFUL;
1283         buf->f_fsid.val[1] = (fsid >> 32) & 0xFFFFFFFFUL;
1284         return 0;
1285 }
1286
1287 static int ext2_get_sb(struct file_system_type *fs_type,
1288         int flags, const char *dev_name, void *data, struct vfsmount *mnt)
1289 {
1290         return get_sb_bdev(fs_type, flags, dev_name, data, ext2_fill_super, mnt);
1291 }
1292
1293 #ifdef CONFIG_QUOTA
1294
1295 /* Read data from quotafile - avoid pagecache and such because we cannot afford
1296  * acquiring the locks... As quota files are never truncated and quota code
1297  * itself serializes the operations (and noone else should touch the files)
1298  * we don't have to be afraid of races */
1299 static ssize_t ext2_quota_read(struct super_block *sb, int type, char *data,
1300                                size_t len, loff_t off)
1301 {
1302         struct inode *inode = sb_dqopt(sb)->files[type];
1303         sector_t blk = off >> EXT2_BLOCK_SIZE_BITS(sb);
1304         int err = 0;
1305         int offset = off & (sb->s_blocksize - 1);
1306         int tocopy;
1307         size_t toread;
1308         struct buffer_head tmp_bh;
1309         struct buffer_head *bh;
1310         loff_t i_size = i_size_read(inode);
1311
1312         if (off > i_size)
1313                 return 0;
1314         if (off+len > i_size)
1315                 len = i_size-off;
1316         toread = len;
1317         while (toread > 0) {
1318                 tocopy = sb->s_blocksize - offset < toread ?
1319                                 sb->s_blocksize - offset : toread;
1320
1321                 tmp_bh.b_state = 0;
1322                 err = ext2_get_block(inode, blk, &tmp_bh, 0);
1323                 if (err < 0)
1324                         return err;
1325                 if (!buffer_mapped(&tmp_bh))    /* A hole? */
1326                         memset(data, 0, tocopy);
1327                 else {
1328                         bh = sb_bread(sb, tmp_bh.b_blocknr);
1329                         if (!bh)
1330                                 return -EIO;
1331                         memcpy(data, bh->b_data+offset, tocopy);
1332                         brelse(bh);
1333                 }
1334                 offset = 0;
1335                 toread -= tocopy;
1336                 data += tocopy;
1337                 blk++;
1338         }
1339         return len;
1340 }
1341
1342 /* Write to quotafile */
1343 static ssize_t ext2_quota_write(struct super_block *sb, int type,
1344                                 const char *data, size_t len, loff_t off)
1345 {
1346         struct inode *inode = sb_dqopt(sb)->files[type];
1347         sector_t blk = off >> EXT2_BLOCK_SIZE_BITS(sb);
1348         int err = 0;
1349         int offset = off & (sb->s_blocksize - 1);
1350         int tocopy;
1351         size_t towrite = len;
1352         struct buffer_head tmp_bh;
1353         struct buffer_head *bh;
1354
1355         mutex_lock_nested(&inode->i_mutex, I_MUTEX_QUOTA);
1356         while (towrite > 0) {
1357                 tocopy = sb->s_blocksize - offset < towrite ?
1358                                 sb->s_blocksize - offset : towrite;
1359
1360                 tmp_bh.b_state = 0;
1361                 err = ext2_get_block(inode, blk, &tmp_bh, 1);
1362                 if (err < 0)
1363                         goto out;
1364                 if (offset || tocopy != EXT2_BLOCK_SIZE(sb))
1365                         bh = sb_bread(sb, tmp_bh.b_blocknr);
1366                 else
1367                         bh = sb_getblk(sb, tmp_bh.b_blocknr);
1368                 if (!bh) {
1369                         err = -EIO;
1370                         goto out;
1371                 }
1372                 lock_buffer(bh);
1373                 memcpy(bh->b_data+offset, data, tocopy);
1374                 flush_dcache_page(bh->b_page);
1375                 set_buffer_uptodate(bh);
1376                 mark_buffer_dirty(bh);
1377                 unlock_buffer(bh);
1378                 brelse(bh);
1379                 offset = 0;
1380                 towrite -= tocopy;
1381                 data += tocopy;
1382                 blk++;
1383         }
1384 out:
1385         if (len == towrite)
1386                 return err;
1387         if (inode->i_size < off+len-towrite)
1388                 i_size_write(inode, off+len-towrite);
1389         inode->i_version++;
1390         inode->i_mtime = inode->i_ctime = CURRENT_TIME;
1391         mark_inode_dirty(inode);
1392         mutex_unlock(&inode->i_mutex);
1393         return len - towrite;
1394 }
1395
1396 #endif
1397
1398 static struct file_system_type ext2_fs_type = {
1399         .owner          = THIS_MODULE,
1400         .name           = "ext2",
1401         .get_sb         = ext2_get_sb,
1402         .kill_sb        = kill_block_super,
1403         .fs_flags       = FS_REQUIRES_DEV,
1404 };
1405
1406 static int __init init_ext2_fs(void)
1407 {
1408         int err = init_ext2_xattr();
1409         if (err)
1410                 return err;
1411         err = init_inodecache();
1412         if (err)
1413                 goto out1;
1414         err = register_filesystem(&ext2_fs_type);
1415         if (err)
1416                 goto out;
1417         return 0;
1418 out:
1419         destroy_inodecache();
1420 out1:
1421         exit_ext2_xattr();
1422         return err;
1423 }
1424
1425 static void __exit exit_ext2_fs(void)
1426 {
1427         unregister_filesystem(&ext2_fs_type);
1428         destroy_inodecache();
1429         exit_ext2_xattr();
1430 }
1431
1432 module_init(init_ext2_fs)
1433 module_exit(exit_ext2_fs)