fs/ext2/super.c

   1 /*
   2  *  linux/fs/ext2/super.c
   3  *
   4  * Copyright (C) 1992, 1993, 1994, 1995
   5  * Remy Card (card@masi.ibp.fr)
   6  * Laboratoire MASI - Institut Blaise Pascal
   7  * Universite Pierre et Marie Curie (Paris VI)
   8  *
   9  *  from
  10  *
  11  *  linux/fs/minix/inode.c
  12  *
  13  *  Copyright (C) 1991, 1992  Linus Torvalds
  14  *
  15  *  Big-endian to little-endian byte-swapping/bitmaps by
  16  *        David S. Miller (davem@caip.rutgers.edu), 1995
  17  */
  18
  19 #include <linux/module.h>
  20 #include <linux/string.h>
  21 #include <linux/fs.h>
  22 #include <linux/slab.h>
  23 #include <linux/init.h>
  24 #include <linux/blkdev.h>
  25 #include <linux/parser.h>
  26 #include <linux/random.h>
  27 #include <linux/buffer_head.h>
  28 #include <linux/exportfs.h>
  29 #include <linux/smp_lock.h>
  30 #include <linux/vfs.h>
  31 #include <linux/seq_file.h>
  32 #include <linux/mount.h>
  33 #include <linux/log2.h>
  34 #include <linux/quotaops.h>
  35 #include <asm/uaccess.h>
  36 #include "ext2.h"
  37 #include "xattr.h"
  38 #include "acl.h"
  39 #include "xip.h"
  40
  41 static void ext2_sync_super(struct super_block *sb,
  42                             struct ext2_super_block *es, int wait);
  43 static int ext2_remount (struct super_block * sb, int * flags, char * data);
  44 static int ext2_statfs (struct dentry * dentry, struct kstatfs * buf);
  45 static int ext2_sync_fs(struct super_block *sb, int wait);
  46
  47 void ext2_error (struct super_block * sb, const char * function,
  48                  const char * fmt, ...)
  49 {
  50         va_list args;
  51         struct ext2_sb_info *sbi = EXT2_SB(sb);
  52         struct ext2_super_block *es = sbi->s_es;
  53
  54         if (!(sb->s_flags & MS_RDONLY)) {
  55                 sbi->s_mount_state |= EXT2_ERROR_FS;
  56                 es->s_state |= cpu_to_le16(EXT2_ERROR_FS);
  57                 ext2_sync_super(sb, es, 1);
  58         }
  59
  60         va_start(args, fmt);
  61         printk(KERN_CRIT "EXT2-fs (%s): error: %s: ", sb->s_id, function);
  62         vprintk(fmt, args);
  63         printk("\n");
  64         va_end(args);
  65
  66         if (test_opt(sb, ERRORS_PANIC))
  67                 panic("EXT2-fs: panic from previous error\n");
  68         if (test_opt(sb, ERRORS_RO)) {
  69                 ext2_msg(sb, KERN_CRIT,
  70                              "error: remounting filesystem read-only");
  71                 sb->s_flags |= MS_RDONLY;
  72         }
  73 }
  74
  75 void ext2_msg(struct super_block *sb, const char *prefix,
  76                 const char *fmt, ...)
  77 {
  78         va_list args;
  79
  80         va_start(args, fmt);
  81         printk("%sEXT2-fs (%s): ", prefix, sb->s_id);
  82         vprintk(fmt, args);
  83         printk("\n");
  84         va_end(args);
  85 }
  86
  87 void ext2_update_dynamic_rev(struct super_block *sb)
  88 {
  89         struct ext2_super_block *es = EXT2_SB(sb)->s_es;
  90
  91         if (le32_to_cpu(es->s_rev_level) > EXT2_GOOD_OLD_REV)
  92                 return;
  93
  94         ext2_msg(sb, KERN_WARNING,
  95                      "warning: updating to rev %d because of "
  96                      "new feature flag, running e2fsck is recommended",
  97                      EXT2_DYNAMIC_REV);
  98
  99         es->s_first_ino = cpu_to_le32(EXT2_GOOD_OLD_FIRST_INO);
 100         es->s_inode_size = cpu_to_le16(EXT2_GOOD_OLD_INODE_SIZE);
 101         es->s_rev_level = cpu_to_le32(EXT2_DYNAMIC_REV);
 102         /* leave es->s_feature_*compat flags alone */
 103         /* es->s_uuid will be set by e2fsck if empty */
 104
 105         /*
 106          * The rest of the superblock fields should be zero, and if not it
 107          * means they are likely already in use, so leave them alone.  We
 108          * can leave it up to e2fsck to clean up any inconsistencies there.
 109          */
 110 }
 111
 112 static void ext2_put_super (struct super_block * sb)
 113 {
 114         int db_count;
 115         int i;
 116         struct ext2_sb_info *sbi = EXT2_SB(sb);
 117
 118         lock_kernel();
 119
 120         if (sb->s_dirt)
 121                 ext2_write_super(sb);
 122
 123         ext2_xattr_put_super(sb);
 124         if (!(sb->s_flags & MS_RDONLY)) {
 125                 struct ext2_super_block *es = sbi->s_es;
 126
 127                 es->s_state = cpu_to_le16(sbi->s_mount_state);
 128                 ext2_sync_super(sb, es, 1);
 129         }
 130         db_count = sbi->s_gdb_count;
 131         for (i = 0; i < db_count; i++)
 132                 if (sbi->s_group_desc[i])
 133                         brelse (sbi->s_group_desc[i]);
 134         kfree(sbi->s_group_desc);
 135         kfree(sbi->s_debts);
 136         percpu_counter_destroy(&sbi->s_freeblocks_counter);
 137         percpu_counter_destroy(&sbi->s_freeinodes_counter);
 138         percpu_counter_destroy(&sbi->s_dirs_counter);
 139         brelse (sbi->s_sbh);
 140         sb->s_fs_info = NULL;
 141         kfree(sbi->s_blockgroup_lock);
 142         kfree(sbi);
 143
 144         unlock_kernel();
 145 }
 146
 147 static struct kmem_cache * ext2_inode_cachep;
 148
 149 static struct inode *ext2_alloc_inode(struct super_block *sb)
 150 {
 151         struct ext2_inode_info *ei;
 152         ei = (struct ext2_inode_info *)kmem_cache_alloc(ext2_inode_cachep, GFP_KERNEL);
 153         if (!ei)
 154                 return NULL;
 155         ei->i_block_alloc_info = NULL;
 156         ei->vfs_inode.i_version = 1;
 157         return &ei->vfs_inode;
 158 }
 159
 160 static void ext2_destroy_inode(struct inode *inode)
 161 {
 162         kmem_cache_free(ext2_inode_cachep, EXT2_I(inode));
 163 }
 164
 165 static void init_once(void *foo)
 166 {
 167         struct ext2_inode_info *ei = (struct ext2_inode_info *) foo;
 168
 169         rwlock_init(&ei->i_meta_lock);
 170 #ifdef CONFIG_EXT2_FS_XATTR
 171         init_rwsem(&ei->xattr_sem);
 172 #endif
 173         mutex_init(&ei->truncate_mutex);
 174         inode_init_once(&ei->vfs_inode);
 175 }
 176
 177 static int init_inodecache(void)
 178 {
 179         ext2_inode_cachep = kmem_cache_create("ext2_inode_cache",
 180                                              sizeof(struct ext2_inode_info),
 181                                              0, (SLAB_RECLAIM_ACCOUNT|
 182                                                 SLAB_MEM_SPREAD),
 183                                              init_once);
 184         if (ext2_inode_cachep == NULL)
 185                 return -ENOMEM;
 186         return 0;
 187 }
 188
 189 static void destroy_inodecache(void)
 190 {
 191         kmem_cache_destroy(ext2_inode_cachep);
 192 }
 193
 194 static void ext2_clear_inode(struct inode *inode)
 195 {
 196         struct ext2_block_alloc_info *rsv = EXT2_I(inode)->i_block_alloc_info;
 197
 198         dquot_drop(inode);
 199         ext2_discard_reservation(inode);
 200         EXT2_I(inode)->i_block_alloc_info = NULL;
 201         if (unlikely(rsv))
 202                 kfree(rsv);
 203 }
 204
 205 static int ext2_show_options(struct seq_file *seq, struct vfsmount *vfs)
 206 {
 207         struct super_block *sb = vfs->mnt_sb;
 208         struct ext2_sb_info *sbi = EXT2_SB(sb);
 209         struct ext2_super_block *es = sbi->s_es;
 210         unsigned long def_mount_opts;
 211
 212         def_mount_opts = le32_to_cpu(es->s_default_mount_opts);
 213
 214         if (sbi->s_sb_block != 1)
 215                 seq_printf(seq, ",sb=%lu", sbi->s_sb_block);
 216         if (test_opt(sb, MINIX_DF))
 217                 seq_puts(seq, ",minixdf");
 218         if (test_opt(sb, GRPID))
 219                 seq_puts(seq, ",grpid");
 220         if (!test_opt(sb, GRPID) && (def_mount_opts & EXT2_DEFM_BSDGROUPS))
 221                 seq_puts(seq, ",nogrpid");
 222         if (sbi->s_resuid != EXT2_DEF_RESUID ||
 223             le16_to_cpu(es->s_def_resuid) != EXT2_DEF_RESUID) {
 224                 seq_printf(seq, ",resuid=%u", sbi->s_resuid);
 225         }
 226         if (sbi->s_resgid != EXT2_DEF_RESGID ||
 227             le16_to_cpu(es->s_def_resgid) != EXT2_DEF_RESGID) {
 228                 seq_printf(seq, ",resgid=%u", sbi->s_resgid);
 229         }
 230         if (test_opt(sb, ERRORS_RO)) {
 231                 int def_errors = le16_to_cpu(es->s_errors);
 232
 233                 if (def_errors == EXT2_ERRORS_PANIC ||
 234                     def_errors == EXT2_ERRORS_CONTINUE) {
 235                         seq_puts(seq, ",errors=remount-ro");
 236                 }
 237         }
 238         if (test_opt(sb, ERRORS_CONT))
 239                 seq_puts(seq, ",errors=continue");
 240         if (test_opt(sb, ERRORS_PANIC))
 241                 seq_puts(seq, ",errors=panic");
 242         if (test_opt(sb, NO_UID32))
 243                 seq_puts(seq, ",nouid32");
 244         if (test_opt(sb, DEBUG))
 245                 seq_puts(seq, ",debug");
 246         if (test_opt(sb, OLDALLOC))
 247                 seq_puts(seq, ",oldalloc");
 248
 249 #ifdef CONFIG_EXT2_FS_XATTR
 250         if (test_opt(sb, XATTR_USER))
 251                 seq_puts(seq, ",user_xattr");
 252         if (!test_opt(sb, XATTR_USER) &&
 253             (def_mount_opts & EXT2_DEFM_XATTR_USER)) {
 254                 seq_puts(seq, ",nouser_xattr");
 255         }
 256 #endif
 257
 258 #ifdef CONFIG_EXT2_FS_POSIX_ACL
 259         if (test_opt(sb, POSIX_ACL))
 260                 seq_puts(seq, ",acl");
 261         if (!test_opt(sb, POSIX_ACL) && (def_mount_opts & EXT2_DEFM_ACL))
 262                 seq_puts(seq, ",noacl");
 263 #endif
 264
 265         if (test_opt(sb, NOBH))
 266                 seq_puts(seq, ",nobh");
 267
 268 #if defined(CONFIG_QUOTA)
 269         if (sbi->s_mount_opt & EXT2_MOUNT_USRQUOTA)
 270                 seq_puts(seq, ",usrquota");
 271
 272         if (sbi->s_mount_opt & EXT2_MOUNT_GRPQUOTA)
 273                 seq_puts(seq, ",grpquota");
 274 #endif
 275
 276 #if defined(CONFIG_EXT2_FS_XIP)
 277         if (sbi->s_mount_opt & EXT2_MOUNT_XIP)
 278                 seq_puts(seq, ",xip");
 279 #endif
 280
 281         if (!test_opt(sb, RESERVATION))
 282                 seq_puts(seq, ",noreservation");
 283
 284         return 0;
 285 }
 286
 287 #ifdef CONFIG_QUOTA
 288 static ssize_t ext2_quota_read(struct super_block *sb, int type, char *data, size_t len, loff_t off);
 289 static ssize_t ext2_quota_write(struct super_block *sb, int type, const char *data, size_t len, loff_t off);
 290 #endif
 291
 292 static const struct super_operations ext2_sops = {
 293         .alloc_inode    = ext2_alloc_inode,
 294         .destroy_inode  = ext2_destroy_inode,
 295         .write_inode    = ext2_write_inode,
 296         .delete_inode   = ext2_delete_inode,
 297         .put_super      = ext2_put_super,
 298         .write_super    = ext2_write_super,
 299         .sync_fs        = ext2_sync_fs,
 300         .statfs         = ext2_statfs,
 301         .remount_fs     = ext2_remount,
 302         .clear_inode    = ext2_clear_inode,
 303         .show_options   = ext2_show_options,
 304 #ifdef CONFIG_QUOTA
 305         .quota_read     = ext2_quota_read,
 306         .quota_write    = ext2_quota_write,
 307 #endif
 308 };
 309
 310 static struct inode *ext2_nfs_get_inode(struct super_block *sb,
 311                 u64 ino, u32 generation)
 312 {
 313         struct inode *inode;
 314
 315         if (ino < EXT2_FIRST_INO(sb) && ino != EXT2_ROOT_INO)
 316                 return ERR_PTR(-ESTALE);
 317         if (ino > le32_to_cpu(EXT2_SB(sb)->s_es->s_inodes_count))
 318                 return ERR_PTR(-ESTALE);
 319
 320         /* iget isn't really right if the inode is currently unallocated!!
 321          * ext2_read_inode currently does appropriate checks, but
 322          * it might be "neater" to call ext2_get_inode first and check
 323          * if the inode is valid.....
 324          */
 325         inode = ext2_iget(sb, ino);
 326         if (IS_ERR(inode))
 327                 return ERR_CAST(inode);
 328         if (generation && inode->i_generation != generation) {
 329                 /* we didn't find the right inode.. */
 330                 iput(inode);
 331                 return ERR_PTR(-ESTALE);
 332         }
 333         return inode;
 334 }
 335
 336 static struct dentry *ext2_fh_to_dentry(struct super_block *sb, struct fid *fid,
 337                 int fh_len, int fh_type)
 338 {
 339         return generic_fh_to_dentry(sb, fid, fh_len, fh_type,
 340                                     ext2_nfs_get_inode);
 341 }
 342
 343 static struct dentry *ext2_fh_to_parent(struct super_block *sb, struct fid *fid,
 344                 int fh_len, int fh_type)
 345 {
 346         return generic_fh_to_parent(sb, fid, fh_len, fh_type,
 347                                     ext2_nfs_get_inode);
 348 }
 349
 350 /* Yes, most of these are left as NULL!!
 351  * A NULL value implies the default, which works with ext2-like file
 352  * systems, but can be improved upon.
 353  * Currently only get_parent is required.
 354  */
 355 static const struct export_operations ext2_export_ops = {
 356         .fh_to_dentry = ext2_fh_to_dentry,
 357         .fh_to_parent = ext2_fh_to_parent,
 358         .get_parent = ext2_get_parent,
 359 };
 360
 361 static unsigned long get_sb_block(void **data)
 362 {
 363         unsigned long   sb_block;
 364         char            *options = (char *) *data;
 365
 366         if (!options || strncmp(options, "sb=", 3) != 0)
 367                 return 1;       /* Default location */
 368         options += 3;
 369         sb_block = simple_strtoul(options, &options, 0);
 370         if (*options && *options != ',') {
 371                 printk("EXT2-fs: Invalid sb specification: %s\n",
 372                        (char *) *data);
 373                 return 1;
 374         }
 375         if (*options == ',')
 376                 options++;
 377         *data = (void *) options;
 378         return sb_block;
 379 }
 380
 381 enum {
 382         Opt_bsd_df, Opt_minix_df, Opt_grpid, Opt_nogrpid,
 383         Opt_resgid, Opt_resuid, Opt_sb, Opt_err_cont, Opt_err_panic,
 384         Opt_err_ro, Opt_nouid32, Opt_nocheck, Opt_debug,
 385         Opt_oldalloc, Opt_orlov, Opt_nobh, Opt_user_xattr, Opt_nouser_xattr,
 386         Opt_acl, Opt_noacl, Opt_xip, Opt_ignore, Opt_err, Opt_quota,
 387         Opt_usrquota, Opt_grpquota, Opt_reservation, Opt_noreservation
 388 };
 389
 390 static const match_table_t tokens = {
 391         {Opt_bsd_df, "bsddf"},
 392         {Opt_minix_df, "minixdf"},
 393         {Opt_grpid, "grpid"},
 394         {Opt_grpid, "bsdgroups"},
 395         {Opt_nogrpid, "nogrpid"},
 396         {Opt_nogrpid, "sysvgroups"},
 397         {Opt_resgid, "resgid=%u"},
 398         {Opt_resuid, "resuid=%u"},
 399         {Opt_sb, "sb=%u"},
 400         {Opt_err_cont, "errors=continue"},
 401         {Opt_err_panic, "errors=panic"},
 402         {Opt_err_ro, "errors=remount-ro"},
 403         {Opt_nouid32, "nouid32"},
 404         {Opt_nocheck, "check=none"},
 405         {Opt_nocheck, "nocheck"},
 406         {Opt_debug, "debug"},
 407         {Opt_oldalloc, "oldalloc"},
 408         {Opt_orlov, "orlov"},
 409         {Opt_nobh, "nobh"},
 410         {Opt_user_xattr, "user_xattr"},
 411         {Opt_nouser_xattr, "nouser_xattr"},
 412         {Opt_acl, "acl"},
 413         {Opt_noacl, "noacl"},
 414         {Opt_xip, "xip"},
 415         {Opt_grpquota, "grpquota"},
 416         {Opt_ignore, "noquota"},
 417         {Opt_quota, "quota"},
 418         {Opt_usrquota, "usrquota"},
 419         {Opt_reservation, "reservation"},
 420         {Opt_noreservation, "noreservation"},
 421         {Opt_err, NULL}
 422 };
 423
 424 static int parse_options(char *options, struct super_block *sb)
 425 {
 426         char *p;
 427         struct ext2_sb_info *sbi = EXT2_SB(sb);
 428         substring_t args[MAX_OPT_ARGS];
 429         int option;
 430
 431         if (!options)
 432                 return 1;
 433
 434         while ((p = strsep (&options, ",")) != NULL) {
 435                 int token;
 436                 if (!*p)
 437                         continue;
 438
 439                 token = match_token(p, tokens, args);
 440                 switch (token) {
 441                 case Opt_bsd_df:
 442                         clear_opt (sbi->s_mount_opt, MINIX_DF);
 443                         break;
 444                 case Opt_minix_df:
 445                         set_opt (sbi->s_mount_opt, MINIX_DF);
 446                         break;
 447                 case Opt_grpid:
 448                         set_opt (sbi->s_mount_opt, GRPID);
 449                         break;
 450                 case Opt_nogrpid:
 451                         clear_opt (sbi->s_mount_opt, GRPID);
 452                         break;
 453                 case Opt_resuid:
 454                         if (match_int(&args[0], &option))
 455                                 return 0;
 456                         sbi->s_resuid = option;
 457                         break;
 458                 case Opt_resgid:
 459                         if (match_int(&args[0], &option))
 460                                 return 0;
 461                         sbi->s_resgid = option;
 462                         break;
 463                 case Opt_sb:
 464                         /* handled by get_sb_block() instead of here */
 465                         /* *sb_block = match_int(&args[0]); */
 466                         break;
 467                 case Opt_err_panic:
 468                         clear_opt (sbi->s_mount_opt, ERRORS_CONT);
 469                         clear_opt (sbi->s_mount_opt, ERRORS_RO);
 470                         set_opt (sbi->s_mount_opt, ERRORS_PANIC);
 471                         break;
 472                 case Opt_err_ro:
 473                         clear_opt (sbi->s_mount_opt, ERRORS_CONT);
 474                         clear_opt (sbi->s_mount_opt, ERRORS_PANIC);
 475                         set_opt (sbi->s_mount_opt, ERRORS_RO);
 476                         break;
 477                 case Opt_err_cont:
 478                         clear_opt (sbi->s_mount_opt, ERRORS_RO);
 479                         clear_opt (sbi->s_mount_opt, ERRORS_PANIC);
 480                         set_opt (sbi->s_mount_opt, ERRORS_CONT);
 481                         break;
 482                 case Opt_nouid32:
 483                         set_opt (sbi->s_mount_opt, NO_UID32);
 484                         break;
 485                 case Opt_nocheck:
 486                         clear_opt (sbi->s_mount_opt, CHECK);
 487                         break;
 488                 case Opt_debug:
 489                         set_opt (sbi->s_mount_opt, DEBUG);
 490                         break;
 491                 case Opt_oldalloc:
 492                         set_opt (sbi->s_mount_opt, OLDALLOC);
 493                         break;
 494                 case Opt_orlov:
 495                         clear_opt (sbi->s_mount_opt, OLDALLOC);
 496                         break;
 497                 case Opt_nobh:
 498                         set_opt (sbi->s_mount_opt, NOBH);
 499                         break;
 500 #ifdef CONFIG_EXT2_FS_XATTR
 501                 case Opt_user_xattr:
 502                         set_opt (sbi->s_mount_opt, XATTR_USER);
 503                         break;
 504                 case Opt_nouser_xattr:
 505                         clear_opt (sbi->s_mount_opt, XATTR_USER);
 506                         break;
 507 #else
 508                 case Opt_user_xattr:
 509                 case Opt_nouser_xattr:
 510                         ext2_msg(sb, KERN_INFO, "(no)user_xattr options"
 511                                 "not supported");
 512                         break;
 513 #endif
 514 #ifdef CONFIG_EXT2_FS_POSIX_ACL
 515                 case Opt_acl:
 516                         set_opt(sbi->s_mount_opt, POSIX_ACL);
 517                         break;
 518                 case Opt_noacl:
 519                         clear_opt(sbi->s_mount_opt, POSIX_ACL);
 520                         break;
 521 #else
 522                 case Opt_acl:
 523                 case Opt_noacl:
 524                         ext2_msg(sb, KERN_INFO,
 525                                 "(no)acl options not supported");
 526                         break;
 527 #endif
 528                 case Opt_xip:
 529 #ifdef CONFIG_EXT2_FS_XIP
 530                         set_opt (sbi->s_mount_opt, XIP);
 531 #else
 532                         ext2_msg(sb, KERN_INFO, "xip option not supported");
 533 #endif
 534                         break;
 535
 536 #if defined(CONFIG_QUOTA)
 537                 case Opt_quota:
 538                 case Opt_usrquota:
 539                         set_opt(sbi->s_mount_opt, USRQUOTA);
 540                         break;
 541
 542                 case Opt_grpquota:
 543                         set_opt(sbi->s_mount_opt, GRPQUOTA);
 544                         break;
 545 #else
 546                 case Opt_quota:
 547                 case Opt_usrquota:
 548                 case Opt_grpquota:
 549                         ext2_msg(sb, KERN_INFO,
 550                                 "quota operations not supported");
 551                         break;
 552 #endif
 553
 554                 case Opt_reservation:
 555                         set_opt(sbi->s_mount_opt, RESERVATION);
 556                         ext2_msg(sb, KERN_INFO, "reservations ON");
 557                         break;
 558                 case Opt_noreservation:
 559                         clear_opt(sbi->s_mount_opt, RESERVATION);
 560                         ext2_msg(sb, KERN_INFO, "reservations OFF");
 561                         break;
 562                 case Opt_ignore:
 563                         break;
 564                 default:
 565                         return 0;
 566                 }
 567         }
 568         return 1;
 569 }
 570
 571 static int ext2_setup_super (struct super_block * sb,
 572                               struct ext2_super_block * es,
 573                               int read_only)
 574 {
 575         int res = 0;
 576         struct ext2_sb_info *sbi = EXT2_SB(sb);
 577
 578         if (le32_to_cpu(es->s_rev_level) > EXT2_MAX_SUPP_REV) {
 579                 ext2_msg(sb, KERN_ERR,
 580                         "error: revision level too high, "
 581                         "forcing read-only mode");
 582                 res = MS_RDONLY;
 583         }
 584         if (read_only)
 585                 return res;
 586         if (!(sbi->s_mount_state & EXT2_VALID_FS))
 587                 ext2_msg(sb, KERN_WARNING,
 588                         "warning: mounting unchecked fs, "
 589                         "running e2fsck is recommended");
 590         else if ((sbi->s_mount_state & EXT2_ERROR_FS))
 591                 ext2_msg(sb, KERN_WARNING,
 592                         "warning: mounting fs with errors, "
 593                         "running e2fsck is recommended");
 594         else if ((__s16) le16_to_cpu(es->s_max_mnt_count) >= 0 &&
 595                  le16_to_cpu(es->s_mnt_count) >=
 596                  (unsigned short) (__s16) le16_to_cpu(es->s_max_mnt_count))
 597                 ext2_msg(sb, KERN_WARNING,
 598                         "warning: maximal mount count reached, "
 599                         "running e2fsck is recommended");
 600         else if (le32_to_cpu(es->s_checkinterval) &&
 601                 (le32_to_cpu(es->s_lastcheck) +
 602                         le32_to_cpu(es->s_checkinterval) <= get_seconds()))
 603                 ext2_msg(sb, KERN_WARNING,
 604                         "warning: checktime reached, "
 605                         "running e2fsck is recommended");
 606         if (!le16_to_cpu(es->s_max_mnt_count))
 607                 es->s_max_mnt_count = cpu_to_le16(EXT2_DFL_MAX_MNT_COUNT);
 608         le16_add_cpu(&es->s_mnt_count, 1);
 609         if (test_opt (sb, DEBUG))
 610                 ext2_msg(sb, KERN_INFO, "%s, %s, bs=%lu, fs=%lu, gc=%lu, "
 611                         "bpg=%lu, ipg=%lu, mo=%04lx]",
 612                         EXT2FS_VERSION, EXT2FS_DATE, sb->s_blocksize,
 613                         sbi->s_frag_size,
 614                         sbi->s_groups_count,
 615                         EXT2_BLOCKS_PER_GROUP(sb),
 616                         EXT2_INODES_PER_GROUP(sb),
 617                         sbi->s_mount_opt);
 618         return res;
 619 }
 620
 621 static int ext2_check_descriptors(struct super_block *sb)
 622 {
 623         int i;
 624         struct ext2_sb_info *sbi = EXT2_SB(sb);
 625
 626         ext2_debug ("Checking group descriptors");
 627
 628         for (i = 0; i < sbi->s_groups_count; i++) {
 629                 struct ext2_group_desc *gdp = ext2_get_group_desc(sb, i, NULL);
 630                 ext2_fsblk_t first_block = ext2_group_first_block_no(sb, i);
 631                 ext2_fsblk_t last_block;
 632
 633                 if (i == sbi->s_groups_count - 1)
 634                         last_block = le32_to_cpu(sbi->s_es->s_blocks_count) - 1;
 635                 else
 636                         last_block = first_block +
 637                                 (EXT2_BLOCKS_PER_GROUP(sb) - 1);
 638
 639                 if (le32_to_cpu(gdp->bg_block_bitmap) < first_block ||
 640                     le32_to_cpu(gdp->bg_block_bitmap) > last_block)
 641                 {
 642                         ext2_error (sb, "ext2_check_descriptors",
 643                                     "Block bitmap for group %d"
 644                                     " not in group (block %lu)!",
 645                                     i, (unsigned long) le32_to_cpu(gdp->bg_block_bitmap));
 646                         return 0;
 647                 }
 648                 if (le32_to_cpu(gdp->bg_inode_bitmap) < first_block ||
 649                     le32_to_cpu(gdp->bg_inode_bitmap) > last_block)
 650                 {
 651                         ext2_error (sb, "ext2_check_descriptors",
 652                                     "Inode bitmap for group %d"
 653                                     " not in group (block %lu)!",
 654                                     i, (unsigned long) le32_to_cpu(gdp->bg_inode_bitmap));
 655                         return 0;
 656                 }
 657                 if (le32_to_cpu(gdp->bg_inode_table) < first_block ||
 658                     le32_to_cpu(gdp->bg_inode_table) + sbi->s_itb_per_group - 1 >
 659                     last_block)
 660                 {
 661                         ext2_error (sb, "ext2_check_descriptors",
 662                                     "Inode table for group %d"
 663                                     " not in group (block %lu)!",
 664                                     i, (unsigned long) le32_to_cpu(gdp->bg_inode_table));
 665                         return 0;
 666                 }
 667         }
 668         return 1;
 669 }
 670
 671 /*
 672  * Maximal file size.  There is a direct, and {,double-,triple-}indirect
 673  * block limit, and also a limit of (2^32 - 1) 512-byte sectors in i_blocks.
 674  * We need to be 1 filesystem block less than the 2^32 sector limit.
 675  */
 676 static loff_t ext2_max_size(int bits)
 677 {
 678         loff_t res = EXT2_NDIR_BLOCKS;
 679         int meta_blocks;
 680         loff_t upper_limit;
 681
 682         /* This is calculated to be the largest file size for a
 683          * dense, file such that the total number of
 684          * sectors in the file, including data and all indirect blocks,
 685          * does not exceed 2^32 -1
 686          * __u32 i_blocks representing the total number of
 687          * 512 bytes blocks of the file
 688          */
 689         upper_limit = (1LL << 32) - 1;
 690
 691         /* total blocks in file system block size */
 692         upper_limit >>= (bits - 9);
 693
 694
 695         /* indirect blocks */
 696         meta_blocks = 1;
 697         /* double indirect blocks */
 698         meta_blocks += 1 + (1LL << (bits-2));
 699         /* tripple indirect blocks */
 700         meta_blocks += 1 + (1LL << (bits-2)) + (1LL << (2*(bits-2)));
 701
 702         upper_limit -= meta_blocks;
 703         upper_limit <<= bits;
 704
 705         res += 1LL << (bits-2);
 706         res += 1LL << (2*(bits-2));
 707         res += 1LL << (3*(bits-2));
 708         res <<= bits;
 709         if (res > upper_limit)
 710                 res = upper_limit;
 711
 712         if (res > MAX_LFS_FILESIZE)
 713                 res = MAX_LFS_FILESIZE;
 714
 715         return res;
 716 }
 717
 718 static unsigned long descriptor_loc(struct super_block *sb,
 719                                     unsigned long logic_sb_block,
 720                                     int nr)
 721 {
 722         struct ext2_sb_info *sbi = EXT2_SB(sb);
 723         unsigned long bg, first_meta_bg;
 724         int has_super = 0;
 725
 726         first_meta_bg = le32_to_cpu(sbi->s_es->s_first_meta_bg);
 727
 728         if (!EXT2_HAS_INCOMPAT_FEATURE(sb, EXT2_FEATURE_INCOMPAT_META_BG) ||
 729             nr < first_meta_bg)
 730                 return (logic_sb_block + nr + 1);
 731         bg = sbi->s_desc_per_block * nr;
 732         if (ext2_bg_has_super(sb, bg))
 733                 has_super = 1;
 734
 735         return ext2_group_first_block_no(sb, bg) + has_super;
 736 }
 737
 738 static int ext2_fill_super(struct super_block *sb, void *data, int silent)
 739 {
 740         struct buffer_head * bh;
 741         struct ext2_sb_info * sbi;
 742         struct ext2_super_block * es;
 743         struct inode *root;
 744         unsigned long block;
 745         unsigned long sb_block = get_sb_block(&data);
 746         unsigned long logic_sb_block;
 747         unsigned long offset = 0;
 748         unsigned long def_mount_opts;
 749         long ret = -EINVAL;
 750         int blocksize = BLOCK_SIZE;
 751         int db_count;
 752         int i, j;
 753         __le32 features;
 754         int err;
 755
 756         sbi = kzalloc(sizeof(*sbi), GFP_KERNEL);
 757         if (!sbi)
 758                 return -ENOMEM;
 759
 760         sbi->s_blockgroup_lock =
 761                 kzalloc(sizeof(struct blockgroup_lock), GFP_KERNEL);
 762         if (!sbi->s_blockgroup_lock) {
 763                 kfree(sbi);
 764                 return -ENOMEM;
 765         }
 766         sb->s_fs_info = sbi;
 767         sbi->s_sb_block = sb_block;
 768
 769         /*
 770          * See what the current blocksize for the device is, and
 771          * use that as the blocksize.  Otherwise (or if the blocksize
 772          * is smaller than the default) use the default.
 773          * This is important for devices that have a hardware
 774          * sectorsize that is larger than the default.
 775          */
 776         blocksize = sb_min_blocksize(sb, BLOCK_SIZE);
 777         if (!blocksize) {
 778                 ext2_msg(sb, KERN_ERR, "error: unable to set blocksize");
 779                 goto failed_sbi;
 780         }
 781
 782         /*
 783          * If the superblock doesn't start on a hardware sector boundary,
 784          * calculate the offset.
 785          */
 786         if (blocksize != BLOCK_SIZE) {
 787                 logic_sb_block = (sb_block*BLOCK_SIZE) / blocksize;
 788                 offset = (sb_block*BLOCK_SIZE) % blocksize;
 789         } else {
 790                 logic_sb_block = sb_block;
 791         }
 792
 793         if (!(bh = sb_bread(sb, logic_sb_block))) {
 794                 ext2_msg(sb, KERN_ERR, "error: unable to read superblock");
 795                 goto failed_sbi;
 796         }
 797         /*
 798          * Note: s_es must be initialized as soon as possible because
 799          *       some ext2 macro-instructions depend on its value
 800          */
 801         es = (struct ext2_super_block *) (((char *)bh->b_data) + offset);
 802         sbi->s_es = es;
 803         sb->s_magic = le16_to_cpu(es->s_magic);
 804
 805         if (sb->s_magic != EXT2_SUPER_MAGIC)
 806                 goto cantfind_ext2;
 807
 808         /* Set defaults before we parse the mount options */
 809         def_mount_opts = le32_to_cpu(es->s_default_mount_opts);
 810         if (def_mount_opts & EXT2_DEFM_DEBUG)
 811                 set_opt(sbi->s_mount_opt, DEBUG);
 812         if (def_mount_opts & EXT2_DEFM_BSDGROUPS)
 813                 set_opt(sbi->s_mount_opt, GRPID);
 814         if (def_mount_opts & EXT2_DEFM_UID16)
 815                 set_opt(sbi->s_mount_opt, NO_UID32);
 816 #ifdef CONFIG_EXT2_FS_XATTR
 817         if (def_mount_opts & EXT2_DEFM_XATTR_USER)
 818                 set_opt(sbi->s_mount_opt, XATTR_USER);
 819 #endif
 820 #ifdef CONFIG_EXT2_FS_POSIX_ACL
 821         if (def_mount_opts & EXT2_DEFM_ACL)
 822                 set_opt(sbi->s_mount_opt, POSIX_ACL);
 823 #endif
 824
 825         if (le16_to_cpu(sbi->s_es->s_errors) == EXT2_ERRORS_PANIC)
 826                 set_opt(sbi->s_mount_opt, ERRORS_PANIC);
 827         else if (le16_to_cpu(sbi->s_es->s_errors) == EXT2_ERRORS_CONTINUE)
 828                 set_opt(sbi->s_mount_opt, ERRORS_CONT);
 829         else
 830                 set_opt(sbi->s_mount_opt, ERRORS_RO);
 831
 832         sbi->s_resuid = le16_to_cpu(es->s_def_resuid);
 833         sbi->s_resgid = le16_to_cpu(es->s_def_resgid);
 834
 835         set_opt(sbi->s_mount_opt, RESERVATION);
 836
 837         if (!parse_options((char *) data, sb))
 838                 goto failed_mount;
 839
 840         sb->s_flags = (sb->s_flags & ~MS_POSIXACL) |
 841                 ((EXT2_SB(sb)->s_mount_opt & EXT2_MOUNT_POSIX_ACL) ?
 842                  MS_POSIXACL : 0);
 843
 844         ext2_xip_verify_sb(sb); /* see if bdev supports xip, unset
 845                                     EXT2_MOUNT_XIP if not */
 846
 847         if (le32_to_cpu(es->s_rev_level) == EXT2_GOOD_OLD_REV &&
 848             (EXT2_HAS_COMPAT_FEATURE(sb, ~0U) ||
 849              EXT2_HAS_RO_COMPAT_FEATURE(sb, ~0U) ||
 850              EXT2_HAS_INCOMPAT_FEATURE(sb, ~0U)))
 851                 ext2_msg(sb, KERN_WARNING,
 852                         "warning: feature flags set on rev 0 fs, "
 853                         "running e2fsck is recommended");
 854         /*
 855          * Check feature flags regardless of the revision level, since we
 856          * previously didn't change the revision level when setting the flags,
 857          * so there is a chance incompat flags are set on a rev 0 filesystem.
 858          */
 859         features = EXT2_HAS_INCOMPAT_FEATURE(sb, ~EXT2_FEATURE_INCOMPAT_SUPP);
 860         if (features) {
 861                 ext2_msg(sb, KERN_ERR,  "error: couldn't mount because of "
 862                        "unsupported optional features (%x)",
 863                         le32_to_cpu(features));
 864                 goto failed_mount;
 865         }
 866         if (!(sb->s_flags & MS_RDONLY) &&
 867             (features = EXT2_HAS_RO_COMPAT_FEATURE(sb, ~EXT2_FEATURE_RO_COMPAT_SUPP))){
 868                 ext2_msg(sb, KERN_ERR, "error: couldn't mount RDWR because of "
 869                        "unsupported optional features (%x)",
 870                        le32_to_cpu(features));
 871                 goto failed_mount;
 872         }
 873
 874         blocksize = BLOCK_SIZE << le32_to_cpu(sbi->s_es->s_log_block_size);
 875
 876         if (ext2_use_xip(sb) && blocksize != PAGE_SIZE) {
 877                 if (!silent)
 878                         ext2_msg(sb, KERN_ERR,
 879                                 "error: unsupported blocksize for xip");
 880                 goto failed_mount;
 881         }
 882
 883         /* If the blocksize doesn't match, re-read the thing.. */
 884         if (sb->s_blocksize != blocksize) {
 885                 brelse(bh);
 886
 887                 if (!sb_set_blocksize(sb, blocksize)) {
 888                         ext2_msg(sb, KERN_ERR, "error: blocksize is too small");
 889                         goto failed_sbi;
 890                 }
 891
 892                 logic_sb_block = (sb_block*BLOCK_SIZE) / blocksize;
 893                 offset = (sb_block*BLOCK_SIZE) % blocksize;
 894                 bh = sb_bread(sb, logic_sb_block);
 895                 if(!bh) {
 896                         ext2_msg(sb, KERN_ERR, "error: couldn't read"
 897                                 "superblock on 2nd try");
 898                         goto failed_sbi;
 899                 }
 900                 es = (struct ext2_super_block *) (((char *)bh->b_data) + offset);
 901                 sbi->s_es = es;
 902                 if (es->s_magic != cpu_to_le16(EXT2_SUPER_MAGIC)) {
 903                         ext2_msg(sb, KERN_ERR, "error: magic mismatch");
 904                         goto failed_mount;
 905                 }
 906         }
 907
 908         sb->s_maxbytes = ext2_max_size(sb->s_blocksize_bits);
 909
 910         if (le32_to_cpu(es->s_rev_level) == EXT2_GOOD_OLD_REV) {
 911                 sbi->s_inode_size = EXT2_GOOD_OLD_INODE_SIZE;
 912                 sbi->s_first_ino = EXT2_GOOD_OLD_FIRST_INO;
 913         } else {
 914                 sbi->s_inode_size = le16_to_cpu(es->s_inode_size);
 915                 sbi->s_first_ino = le32_to_cpu(es->s_first_ino);
 916                 if ((sbi->s_inode_size < EXT2_GOOD_OLD_INODE_SIZE) ||
 917                     !is_power_of_2(sbi->s_inode_size) ||
 918                     (sbi->s_inode_size > blocksize)) {
 919                         ext2_msg(sb, KERN_ERR,
 920                                 "error: unsupported inode size: %d",
 921                                 sbi->s_inode_size);
 922                         goto failed_mount;
 923                 }
 924         }
 925
 926         sbi->s_frag_size = EXT2_MIN_FRAG_SIZE <<
 927                                    le32_to_cpu(es->s_log_frag_size);
 928         if (sbi->s_frag_size == 0)
 929                 goto cantfind_ext2;
 930         sbi->s_frags_per_block = sb->s_blocksize / sbi->s_frag_size;
 931
 932         sbi->s_blocks_per_group = le32_to_cpu(es->s_blocks_per_group);
 933         sbi->s_frags_per_group = le32_to_cpu(es->s_frags_per_group);
 934         sbi->s_inodes_per_group = le32_to_cpu(es->s_inodes_per_group);
 935
 936         if (EXT2_INODE_SIZE(sb) == 0)
 937                 goto cantfind_ext2;
 938         sbi->s_inodes_per_block = sb->s_blocksize / EXT2_INODE_SIZE(sb);
 939         if (sbi->s_inodes_per_block == 0 || sbi->s_inodes_per_group == 0)
 940                 goto cantfind_ext2;
 941         sbi->s_itb_per_group = sbi->s_inodes_per_group /
 942                                         sbi->s_inodes_per_block;
 943         sbi->s_desc_per_block = sb->s_blocksize /
 944                                         sizeof (struct ext2_group_desc);
 945         sbi->s_sbh = bh;
 946         sbi->s_mount_state = le16_to_cpu(es->s_state);
 947         sbi->s_addr_per_block_bits =
 948                 ilog2 (EXT2_ADDR_PER_BLOCK(sb));
 949         sbi->s_desc_per_block_bits =
 950                 ilog2 (EXT2_DESC_PER_BLOCK(sb));
 951
 952         if (sb->s_magic != EXT2_SUPER_MAGIC)
 953                 goto cantfind_ext2;
 954
 955         if (sb->s_blocksize != bh->b_size) {
 956                 if (!silent)
 957                         ext2_msg(sb, KERN_ERR, "error: unsupported blocksize");
 958                 goto failed_mount;
 959         }
 960
 961         if (sb->s_blocksize != sbi->s_frag_size) {
 962                 ext2_msg(sb, KERN_ERR,
 963                         "error: fragsize %lu != blocksize %lu"
 964                         "(not supported yet)",
 965                         sbi->s_frag_size, sb->s_blocksize);
 966                 goto failed_mount;
 967         }
 968
 969         if (sbi->s_blocks_per_group > sb->s_blocksize * 8) {
 970                 ext2_msg(sb, KERN_ERR,
 971                         "error: #blocks per group too big: %lu",
 972                         sbi->s_blocks_per_group);
 973                 goto failed_mount;
 974         }
 975         if (sbi->s_frags_per_group > sb->s_blocksize * 8) {
 976                 ext2_msg(sb, KERN_ERR,
 977                         "error: #fragments per group too big: %lu",
 978                         sbi->s_frags_per_group);
 979                 goto failed_mount;
 980         }
 981         if (sbi->s_inodes_per_group > sb->s_blocksize * 8) {
 982                 ext2_msg(sb, KERN_ERR,
 983                         "error: #inodes per group too big: %lu",
 984                         sbi->s_inodes_per_group);
 985                 goto failed_mount;
 986         }
 987
 988         if (EXT2_BLOCKS_PER_GROUP(sb) == 0)
 989                 goto cantfind_ext2;
 990         sbi->s_groups_count = ((le32_to_cpu(es->s_blocks_count) -
 991                                 le32_to_cpu(es->s_first_data_block) - 1)
 992                                         / EXT2_BLOCKS_PER_GROUP(sb)) + 1;
 993         db_count = (sbi->s_groups_count + EXT2_DESC_PER_BLOCK(sb) - 1) /
 994                    EXT2_DESC_PER_BLOCK(sb);
 995         sbi->s_group_desc = kmalloc (db_count * sizeof (struct buffer_head *), GFP_KERNEL);
 996         if (sbi->s_group_desc == NULL) {
 997                 ext2_msg(sb, KERN_ERR, "error: not enough memory");
 998                 goto failed_mount;
 999         }
1000         bgl_lock_init(sbi->s_blockgroup_lock);
1001         sbi->s_debts = kcalloc(sbi->s_groups_count, sizeof(*sbi->s_debts), GFP_KERNEL);
1002         if (!sbi->s_debts) {
1003                 ext2_msg(sb, KERN_ERR, "error: not enough memory");
1004                 goto failed_mount_group_desc;
1005         }
1006         for (i = 0; i < db_count; i++) {
1007                 block = descriptor_loc(sb, logic_sb_block, i);
1008                 sbi->s_group_desc[i] = sb_bread(sb, block);
1009                 if (!sbi->s_group_desc[i]) {
1010                         for (j = 0; j < i; j++)
1011                                 brelse (sbi->s_group_desc[j]);
1012                         ext2_msg(sb, KERN_ERR,
1013                                 "error: unable to read group descriptors");
1014                         goto failed_mount_group_desc;
1015                 }
1016         }
1017         if (!ext2_check_descriptors (sb)) {
1018                 ext2_msg(sb, KERN_ERR, "group descriptors corrupted");
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                 ext2_msg(sb, KERN_ERR, "error: insufficient memory");
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                 ext2_msg(sb, KERN_ERR, "error: corrupt root inode, run e2fsck");
1068                 goto failed_mount3;
1069         }
1070
1071         sb->s_root = d_alloc_root(root);
1072         if (!sb->s_root) {
1073                 iput(root);
1074                 ext2_msg(sb, KERN_ERR, "error: get root inode failed");
1075                 ret = -ENOMEM;
1076                 goto failed_mount3;
1077         }
1078         if (EXT2_HAS_COMPAT_FEATURE(sb, EXT3_FEATURE_COMPAT_HAS_JOURNAL))
1079                 ext2_msg(sb, KERN_WARNING,
1080                         "warning: mounting ext3 filesystem as ext2");
1081         if (ext2_setup_super (sb, es, sb->s_flags & MS_RDONLY))
1082                 sb->s_flags |= MS_RDONLY;
1083         ext2_write_super(sb);
1084         return 0;
1085
1086 cantfind_ext2:
1087         if (!silent)
1088                 ext2_msg(sb, KERN_ERR,
1089                         "error: can't find an ext2 filesystem on dev %s.",
1090                         sb->s_id);
1091         goto failed_mount;
1092 failed_mount3:
1093         percpu_counter_destroy(&sbi->s_freeblocks_counter);
1094         percpu_counter_destroy(&sbi->s_freeinodes_counter);
1095         percpu_counter_destroy(&sbi->s_dirs_counter);
1096 failed_mount2:
1097         for (i = 0; i < db_count; i++)
1098                 brelse(sbi->s_group_desc[i]);
1099 failed_mount_group_desc:
1100         kfree(sbi->s_group_desc);
1101         kfree(sbi->s_debts);
1102 failed_mount:
1103         brelse(bh);
1104 failed_sbi:
1105         sb->s_fs_info = NULL;
1106         kfree(sbi->s_blockgroup_lock);
1107         kfree(sbi);
1108         return ret;
1109 }
1110
1111 static void ext2_clear_super_error(struct super_block *sb)
1112 {
1113         struct buffer_head *sbh = EXT2_SB(sb)->s_sbh;
1114
1115         if (buffer_write_io_error(sbh)) {
1116                 /*
1117                  * Oh, dear.  A previous attempt to write the
1118                  * superblock failed.  This could happen because the
1119                  * USB device was yanked out.  Or it could happen to
1120                  * be a transient write error and maybe the block will
1121                  * be remapped.  Nothing we can do but to retry the
1122                  * write and hope for the best.
1123                  */
1124                 ext2_msg(sb, KERN_ERR,
1125                        "previous I/O error to superblock detected\n");
1126                 clear_buffer_write_io_error(sbh);
1127                 set_buffer_uptodate(sbh);
1128         }
1129 }
1130
1131 static void ext2_sync_super(struct super_block *sb, struct ext2_super_block *es,
1132                             int wait)
1133 {
1134         ext2_clear_super_error(sb);
1135         es->s_free_blocks_count = cpu_to_le32(ext2_count_free_blocks(sb));
1136         es->s_free_inodes_count = cpu_to_le32(ext2_count_free_inodes(sb));
1137         es->s_wtime = cpu_to_le32(get_seconds());
1138         mark_buffer_dirty(EXT2_SB(sb)->s_sbh);
1139         if (wait)
1140                 sync_dirty_buffer(EXT2_SB(sb)->s_sbh);
1141         sb->s_dirt = 0;
1142 }
1143
1144 /*
1145  * In the second extended file system, it is not necessary to
1146  * write the super block since we use a mapping of the
1147  * disk super block in a buffer.
1148  *
1149  * However, this function is still used to set the fs valid
1150  * flags to 0.  We need to set this flag to 0 since the fs
1151  * may have been checked while mounted and e2fsck may have
1152  * set s_state to EXT2_VALID_FS after some corrections.
1153  */
1154
1155 static int ext2_sync_fs(struct super_block *sb, int wait)
1156 {
1157         struct ext2_super_block *es = EXT2_SB(sb)->s_es;
1158
1159         lock_kernel();
1160         if (es->s_state & cpu_to_le16(EXT2_VALID_FS)) {
1161                 ext2_debug("setting valid to 0\n");
1162                 es->s_state &= cpu_to_le16(~EXT2_VALID_FS);
1163         }
1164         ext2_sync_super(sb, es, wait);
1165         unlock_kernel();
1166
1167         return 0;
1168 }
1169
1170
1171 void ext2_write_super(struct super_block *sb)
1172 {
1173         if (!(sb->s_flags & MS_RDONLY))
1174                 ext2_sync_fs(sb, 1);
1175         else
1176                 sb->s_dirt = 0;
1177 }
1178
1179 static int ext2_remount (struct super_block * sb, int * flags, char * data)
1180 {
1181         struct ext2_sb_info * sbi = EXT2_SB(sb);
1182         struct ext2_super_block * es;
1183         unsigned long old_mount_opt = sbi->s_mount_opt;
1184         struct ext2_mount_options old_opts;
1185         unsigned long old_sb_flags;
1186         int err;
1187
1188         lock_kernel();
1189
1190         /* Store the old options */
1191         old_sb_flags = sb->s_flags;
1192         old_opts.s_mount_opt = sbi->s_mount_opt;
1193         old_opts.s_resuid = sbi->s_resuid;
1194         old_opts.s_resgid = sbi->s_resgid;
1195
1196         /*
1197          * Allow the "check" option to be passed as a remount option.
1198          */
1199         if (!parse_options(data, sb)) {
1200                 err = -EINVAL;
1201                 goto restore_opts;
1202         }
1203
1204         sb->s_flags = (sb->s_flags & ~MS_POSIXACL) |
1205                 ((sbi->s_mount_opt & EXT2_MOUNT_POSIX_ACL) ? MS_POSIXACL : 0);
1206
1207         ext2_xip_verify_sb(sb); /* see if bdev supports xip, unset
1208                                     EXT2_MOUNT_XIP if not */
1209
1210         if ((ext2_use_xip(sb)) && (sb->s_blocksize != PAGE_SIZE)) {
1211                 ext2_msg(sb, KERN_WARNING,
1212                         "warning: unsupported blocksize for xip");
1213                 err = -EINVAL;
1214                 goto restore_opts;
1215         }
1216
1217         es = sbi->s_es;
1218         if (((sbi->s_mount_opt & EXT2_MOUNT_XIP) !=
1219             (old_mount_opt & EXT2_MOUNT_XIP)) &&
1220             invalidate_inodes(sb)) {
1221                 ext2_msg(sb, KERN_WARNING, "warning: refusing change of "
1222                          "xip flag with busy inodes while remounting");
1223                 sbi->s_mount_opt &= ~EXT2_MOUNT_XIP;
1224                 sbi->s_mount_opt |= old_mount_opt & EXT2_MOUNT_XIP;
1225         }
1226         if ((*flags & MS_RDONLY) == (sb->s_flags & MS_RDONLY)) {
1227                 unlock_kernel();
1228                 return 0;
1229         }
1230         if (*flags & MS_RDONLY) {
1231                 if (le16_to_cpu(es->s_state) & EXT2_VALID_FS ||
1232                     !(sbi->s_mount_state & EXT2_VALID_FS)) {
1233                         unlock_kernel();
1234                         return 0;
1235                 }
1236                 /*
1237                  * OK, we are remounting a valid rw partition rdonly, so set
1238                  * the rdonly flag and then mark the partition as valid again.
1239                  */
1240                 es->s_state = cpu_to_le16(sbi->s_mount_state);
1241                 es->s_mtime = cpu_to_le32(get_seconds());
1242                 ext2_sync_super(sb, es, 1);
1243         } else {
1244                 __le32 ret = EXT2_HAS_RO_COMPAT_FEATURE(sb,
1245                                                ~EXT2_FEATURE_RO_COMPAT_SUPP);
1246                 if (ret) {
1247                         ext2_msg(sb, KERN_WARNING,
1248                                 "warning: couldn't remount RDWR because of "
1249                                 "unsupported optional features (%x).",
1250                                 le32_to_cpu(ret));
1251                         err = -EROFS;
1252                         goto restore_opts;
1253                 }
1254                 /*
1255                  * Mounting a RDONLY partition read-write, so reread and
1256                  * store the current valid flag.  (It may have been changed
1257                  * by e2fsck since we originally mounted the partition.)
1258                  */
1259                 sbi->s_mount_state = le16_to_cpu(es->s_state);
1260                 if (!ext2_setup_super (sb, es, 0))
1261                         sb->s_flags &= ~MS_RDONLY;
1262                 ext2_write_super(sb);
1263         }
1264         unlock_kernel();
1265         return 0;
1266 restore_opts:
1267         sbi->s_mount_opt = old_opts.s_mount_opt;
1268         sbi->s_resuid = old_opts.s_resuid;
1269         sbi->s_resgid = old_opts.s_resgid;
1270         sb->s_flags = old_sb_flags;
1271         unlock_kernel();
1272         return err;
1273 }
1274
1275 static int ext2_statfs (struct dentry * dentry, struct kstatfs * buf)
1276 {
1277         struct super_block *sb = dentry->d_sb;
1278         struct ext2_sb_info *sbi = EXT2_SB(sb);
1279         struct ext2_super_block *es = sbi->s_es;
1280         u64 fsid;
1281
1282         if (test_opt (sb, MINIX_DF))
1283                 sbi->s_overhead_last = 0;
1284         else if (sbi->s_blocks_last != le32_to_cpu(es->s_blocks_count)) {
1285                 unsigned long i, overhead = 0;
1286                 smp_rmb();
1287
1288                 /*
1289                  * Compute the overhead (FS structures). This is constant
1290                  * for a given filesystem unless the number of block groups
1291                  * changes so we cache the previous value until it does.
1292                  */
1293
1294                 /*
1295                  * All of the blocks before first_data_block are
1296                  * overhead
1297                  */
1298                 overhead = le32_to_cpu(es->s_first_data_block);
1299
1300                 /*
1301                  * Add the overhead attributed to the superblock and
1302                  * block group descriptors.  If the sparse superblocks
1303                  * feature is turned on, then not all groups have this.
1304                  */
1305                 for (i = 0; i < sbi->s_groups_count; i++)
1306                         overhead += ext2_bg_has_super(sb, i) +
1307                                 ext2_bg_num_gdb(sb, i);
1308
1309                 /*
1310                  * Every block group has an inode bitmap, a block
1311                  * bitmap, and an inode table.
1312                  */
1313                 overhead += (sbi->s_groups_count *
1314                              (2 + sbi->s_itb_per_group));
1315                 sbi->s_overhead_last = overhead;
1316                 smp_wmb();
1317                 sbi->s_blocks_last = le32_to_cpu(es->s_blocks_count);
1318         }
1319
1320         buf->f_type = EXT2_SUPER_MAGIC;
1321         buf->f_bsize = sb->s_blocksize;
1322         buf->f_blocks = le32_to_cpu(es->s_blocks_count) - sbi->s_overhead_last;
1323         buf->f_bfree = ext2_count_free_blocks(sb);
1324         es->s_free_blocks_count = cpu_to_le32(buf->f_bfree);
1325         buf->f_bavail = buf->f_bfree - le32_to_cpu(es->s_r_blocks_count);
1326         if (buf->f_bfree < le32_to_cpu(es->s_r_blocks_count))
1327                 buf->f_bavail = 0;
1328         buf->f_files = le32_to_cpu(es->s_inodes_count);
1329         buf->f_ffree = ext2_count_free_inodes(sb);
1330         es->s_free_inodes_count = cpu_to_le32(buf->f_ffree);
1331         buf->f_namelen = EXT2_NAME_LEN;
1332         fsid = le64_to_cpup((void *)es->s_uuid) ^
1333                le64_to_cpup((void *)es->s_uuid + sizeof(u64));
1334         buf->f_fsid.val[0] = fsid & 0xFFFFFFFFUL;
1335         buf->f_fsid.val[1] = (fsid >> 32) & 0xFFFFFFFFUL;
1336         return 0;
1337 }
1338
1339 static int ext2_get_sb(struct file_system_type *fs_type,
1340         int flags, const char *dev_name, void *data, struct vfsmount *mnt)
1341 {
1342         return get_sb_bdev(fs_type, flags, dev_name, data, ext2_fill_super, mnt);
1343 }
1344
1345 #ifdef CONFIG_QUOTA
1346
1347 /* Read data from quotafile - avoid pagecache and such because we cannot afford
1348  * acquiring the locks... As quota files are never truncated and quota code
1349  * itself serializes the operations (and noone else should touch the files)
1350  * we don't have to be afraid of races */
1351 static ssize_t ext2_quota_read(struct super_block *sb, int type, char *data,
1352                                size_t len, loff_t off)
1353 {
1354         struct inode *inode = sb_dqopt(sb)->files[type];
1355         sector_t blk = off >> EXT2_BLOCK_SIZE_BITS(sb);
1356         int err = 0;
1357         int offset = off & (sb->s_blocksize - 1);
1358         int tocopy;
1359         size_t toread;
1360         struct buffer_head tmp_bh;
1361         struct buffer_head *bh;
1362         loff_t i_size = i_size_read(inode);
1363
1364         if (off > i_size)
1365                 return 0;
1366         if (off+len > i_size)
1367                 len = i_size-off;
1368         toread = len;
1369         while (toread > 0) {
1370                 tocopy = sb->s_blocksize - offset < toread ?
1371                                 sb->s_blocksize - offset : toread;
1372
1373                 tmp_bh.b_state = 0;
1374                 tmp_bh.b_size = sb->s_blocksize;
1375                 err = ext2_get_block(inode, blk, &tmp_bh, 0);
1376                 if (err < 0)
1377                         return err;
1378                 if (!buffer_mapped(&tmp_bh))    /* A hole? */
1379                         memset(data, 0, tocopy);
1380                 else {
1381                         bh = sb_bread(sb, tmp_bh.b_blocknr);
1382                         if (!bh)
1383                                 return -EIO;
1384                         memcpy(data, bh->b_data+offset, tocopy);
1385                         brelse(bh);
1386                 }
1387                 offset = 0;
1388                 toread -= tocopy;
1389                 data += tocopy;
1390                 blk++;
1391         }
1392         return len;
1393 }
1394
1395 /* Write to quotafile */
1396 static ssize_t ext2_quota_write(struct super_block *sb, int type,
1397                                 const char *data, size_t len, loff_t off)
1398 {
1399         struct inode *inode = sb_dqopt(sb)->files[type];
1400         sector_t blk = off >> EXT2_BLOCK_SIZE_BITS(sb);
1401         int err = 0;
1402         int offset = off & (sb->s_blocksize - 1);
1403         int tocopy;
1404         size_t towrite = len;
1405         struct buffer_head tmp_bh;
1406         struct buffer_head *bh;
1407
1408         mutex_lock_nested(&inode->i_mutex, I_MUTEX_QUOTA);
1409         while (towrite > 0) {
1410                 tocopy = sb->s_blocksize - offset < towrite ?
1411                                 sb->s_blocksize - offset : towrite;
1412
1413                 tmp_bh.b_state = 0;
1414                 err = ext2_get_block(inode, blk, &tmp_bh, 1);
1415                 if (err < 0)
1416                         goto out;
1417                 if (offset || tocopy != EXT2_BLOCK_SIZE(sb))
1418                         bh = sb_bread(sb, tmp_bh.b_blocknr);
1419                 else
1420                         bh = sb_getblk(sb, tmp_bh.b_blocknr);
1421                 if (!bh) {
1422                         err = -EIO;
1423                         goto out;
1424                 }
1425                 lock_buffer(bh);
1426                 memcpy(bh->b_data+offset, data, tocopy);
1427                 flush_dcache_page(bh->b_page);
1428                 set_buffer_uptodate(bh);
1429                 mark_buffer_dirty(bh);
1430                 unlock_buffer(bh);
1431                 brelse(bh);
1432                 offset = 0;
1433                 towrite -= tocopy;
1434                 data += tocopy;
1435                 blk++;
1436         }
1437 out:
1438         if (len == towrite) {
1439                 mutex_unlock(&inode->i_mutex);
1440                 return err;
1441         }
1442         if (inode->i_size < off+len-towrite)
1443                 i_size_write(inode, off+len-towrite);
1444         inode->i_version++;
1445         inode->i_mtime = inode->i_ctime = CURRENT_TIME;
1446         mark_inode_dirty(inode);
1447         mutex_unlock(&inode->i_mutex);
1448         return len - towrite;
1449 }
1450
1451 #endif
1452
1453 static struct file_system_type ext2_fs_type = {
1454         .owner          = THIS_MODULE,
1455         .name           = "ext2",
1456         .get_sb         = ext2_get_sb,
1457         .kill_sb        = kill_block_super,
1458         .fs_flags       = FS_REQUIRES_DEV,
1459 };
1460
1461 static int __init init_ext2_fs(void)
1462 {
1463         int err = init_ext2_xattr();
1464         if (err)
1465                 return err;
1466         err = init_inodecache();
1467         if (err)
1468                 goto out1;
1469         err = register_filesystem(&ext2_fs_type);
1470         if (err)
1471                 goto out;
1472         return 0;
1473 out:
1474         destroy_inodecache();
1475 out1:
1476         exit_ext2_xattr();
1477         return err;
1478 }
1479
1480 static void __exit exit_ext2_fs(void)
1481 {
1482         unregister_filesystem(&ext2_fs_type);
1483         destroy_inodecache();
1484         exit_ext2_xattr();
1485 }
1486
1487 module_init(init_ext2_fs)
1488 module_exit(exit_ext2_fs)