SAFE public projects git trees. - safe/jmp/linux-2.6/blob - fs/ext2/super.c

   1 /*
   2  *  linux/fs/ext2/super.c
   3  *
   4  * Copyright (C) 1992, 1993, 1994, 1995
   5  * Remy Card (card@masi.ibp.fr)
   6  * Laboratoire MASI - Institut Blaise Pascal
   7  * Universite Pierre et Marie Curie (Paris VI)
   8  *
   9  *  from
  10  *
  11  *  linux/fs/minix/inode.c
  12  *
  13  *  Copyright (C) 1991, 1992  Linus Torvalds
  14  *
  15  *  Big-endian to little-endian byte-swapping/bitmaps by
  16  *        David S. Miller (davem@caip.rutgers.edu), 1995
  17  */
  18
  19 #include <linux/module.h>
  20 #include <linux/string.h>
  21 #include <linux/fs.h>
  22 #include <linux/slab.h>
  23 #include <linux/init.h>
  24 #include <linux/blkdev.h>
  25 #include <linux/parser.h>
  26 #include <linux/random.h>
  27 #include <linux/buffer_head.h>
  28 #include <linux/exportfs.h>
  29 #include <linux/smp_lock.h>
  30 #include <linux/vfs.h>
  31 #include <linux/seq_file.h>
  32 #include <linux/mount.h>
  33 #include <linux/log2.h>
  34 #include <linux/quotaops.h>
  35 #include <asm/uaccess.h>
  36 #include "ext2.h"
  37 #include "xattr.h"
  38 #include "acl.h"
  39 #include "xip.h"
  40
  41 static void ext2_sync_super(struct super_block *sb,
  42                             struct ext2_super_block *es);
  43 static int ext2_remount (struct super_block * sb, int * flags, char * data);
  44 static int ext2_statfs (struct dentry * dentry, struct kstatfs * buf);
  45
  46 void ext2_error (struct super_block * sb, const char * function,
  47                  const char * fmt, ...)
  48 {
  49         va_list args;
  50         struct ext2_sb_info *sbi = EXT2_SB(sb);
  51         struct ext2_super_block *es = sbi->s_es;
  52
  53         if (!(sb->s_flags & MS_RDONLY)) {
  54                 sbi->s_mount_state |= EXT2_ERROR_FS;
  55                 es->s_state |= cpu_to_le16(EXT2_ERROR_FS);
  56                 ext2_sync_super(sb, es);
  57         }
  58
  59         va_start(args, fmt);
  60         printk(KERN_CRIT "EXT2-fs error (device %s): %s: ",sb->s_id, function);
  61         vprintk(fmt, args);
  62         printk("\n");
  63         va_end(args);
  64
  65         if (test_opt(sb, ERRORS_PANIC))
  66                 panic("EXT2-fs panic from previous error\n");
  67         if (test_opt(sb, ERRORS_RO)) {
  68                 printk("Remounting filesystem read-only\n");
  69                 sb->s_flags |= MS_RDONLY;
  70         }
  71 }
  72
  73 void ext2_warning (struct super_block * sb, const char * function,
  74                    const char * fmt, ...)
  75 {
  76         va_list args;
  77
  78         va_start(args, fmt);
  79         printk(KERN_WARNING "EXT2-fs warning (device %s): %s: ",
  80                sb->s_id, function);
  81         vprintk(fmt, args);
  82         printk("\n");
  83         va_end(args);
  84 }
  85
  86 void ext2_update_dynamic_rev(struct super_block *sb)
  87 {
  88         struct ext2_super_block *es = EXT2_SB(sb)->s_es;
  89
  90         if (le32_to_cpu(es->s_rev_level) > EXT2_GOOD_OLD_REV)
  91                 return;
  92
  93         ext2_warning(sb, __func__,
  94                      "updating to rev %d because of new feature flag, "
  95                      "running e2fsck is recommended",
  96                      EXT2_DYNAMIC_REV);
  97
  98         es->s_first_ino = cpu_to_le32(EXT2_GOOD_OLD_FIRST_INO);
  99         es->s_inode_size = cpu_to_le16(EXT2_GOOD_OLD_INODE_SIZE);
 100         es->s_rev_level = cpu_to_le32(EXT2_DYNAMIC_REV);
 101         /* leave es->s_feature_*compat flags alone */
 102         /* es->s_uuid will be set by e2fsck if empty */
 103
 104         /*
 105          * The rest of the superblock fields should be zero, and if not it
 106          * means they are likely already in use, so leave them alone.  We
 107          * can leave it up to e2fsck to clean up any inconsistencies there.
 108          */
 109 }
 110
 111 static void ext2_put_super (struct super_block * sb)
 112 {
 113         int db_count;
 114         int i;
 115         struct ext2_sb_info *sbi = EXT2_SB(sb);
 116
 117         lock_kernel();
 118
 119         if (sb->s_dirt)
 120                 ext2_write_super(sb);
 121
 122         ext2_xattr_put_super(sb);
 123         if (!(sb->s_flags & MS_RDONLY)) {
 124                 struct ext2_super_block *es = sbi->s_es;
 125
 126                 es->s_state = cpu_to_le16(sbi->s_mount_state);
 127                 ext2_sync_super(sb, es);
 128         }
 129         db_count = sbi->s_gdb_count;
 130         for (i = 0; i < db_count; i++)
 131                 if (sbi->s_group_desc[i])
 132                         brelse (sbi->s_group_desc[i]);
 133         kfree(sbi->s_group_desc);
 134         kfree(sbi->s_debts);
 135         percpu_counter_destroy(&sbi->s_freeblocks_counter);
 136         percpu_counter_destroy(&sbi->s_freeinodes_counter);
 137         percpu_counter_destroy(&sbi->s_dirs_counter);
 138         brelse (sbi->s_sbh);
 139         sb->s_fs_info = NULL;
 140         kfree(sbi->s_blockgroup_lock);
 141         kfree(sbi);
 142
 143         unlock_kernel();
 144 }
 145
 146 static struct kmem_cache * ext2_inode_cachep;
 147
 148 static struct inode *ext2_alloc_inode(struct super_block *sb)
 149 {
 150         struct ext2_inode_info *ei;
 151         ei = (struct ext2_inode_info *)kmem_cache_alloc(ext2_inode_cachep, GFP_KERNEL);
 152         if (!ei)
 153                 return NULL;
 154 #ifdef CONFIG_EXT2_FS_POSIX_ACL
 155         ei->i_acl = EXT2_ACL_NOT_CACHED;
 156         ei->i_default_acl = EXT2_ACL_NOT_CACHED;
 157 #endif
 158         ei->i_block_alloc_info = NULL;
 159         ei->vfs_inode.i_version = 1;
 160         return &ei->vfs_inode;
 161 }
 162
 163 static void ext2_destroy_inode(struct inode *inode)
 164 {
 165         kmem_cache_free(ext2_inode_cachep, EXT2_I(inode));
 166 }
 167
 168 static void init_once(void *foo)
 169 {
 170         struct ext2_inode_info *ei = (struct ext2_inode_info *) foo;
 171
 172         rwlock_init(&ei->i_meta_lock);
 173 #ifdef CONFIG_EXT2_FS_XATTR
 174         init_rwsem(&ei->xattr_sem);
 175 #endif
 176         mutex_init(&ei->truncate_mutex);
 177         inode_init_once(&ei->vfs_inode);
 178 }
 179
 180 static int init_inodecache(void)
 181 {
 182         ext2_inode_cachep = kmem_cache_create("ext2_inode_cache",
 183                                              sizeof(struct ext2_inode_info),
 184                                              0, (SLAB_RECLAIM_ACCOUNT|
 185                                                 SLAB_MEM_SPREAD),
 186                                              init_once);
 187         if (ext2_inode_cachep == NULL)
 188                 return -ENOMEM;
 189         return 0;
 190 }
 191
 192 static void destroy_inodecache(void)
 193 {
 194         kmem_cache_destroy(ext2_inode_cachep);
 195 }
 196
 197 static void ext2_clear_inode(struct inode *inode)
 198 {
 199         struct ext2_block_alloc_info *rsv = EXT2_I(inode)->i_block_alloc_info;
 200 #ifdef CONFIG_EXT2_FS_POSIX_ACL
 201         struct ext2_inode_info *ei = EXT2_I(inode);
 202
 203         if (ei->i_acl && ei->i_acl != EXT2_ACL_NOT_CACHED) {
 204                 posix_acl_release(ei->i_acl);
 205                 ei->i_acl = EXT2_ACL_NOT_CACHED;
 206         }
 207         if (ei->i_default_acl && ei->i_default_acl != EXT2_ACL_NOT_CACHED) {
 208                 posix_acl_release(ei->i_default_acl);
 209                 ei->i_default_acl = EXT2_ACL_NOT_CACHED;
 210         }
 211 #endif
 212         ext2_discard_reservation(inode);
 213         EXT2_I(inode)->i_block_alloc_info = NULL;
 214         if (unlikely(rsv))
 215                 kfree(rsv);
 216 }
 217
 218 static int ext2_show_options(struct seq_file *seq, struct vfsmount *vfs)
 219 {
 220         struct super_block *sb = vfs->mnt_sb;
 221         struct ext2_sb_info *sbi = EXT2_SB(sb);
 222         struct ext2_super_block *es = sbi->s_es;
 223         unsigned long def_mount_opts;
 224
 225         def_mount_opts = le32_to_cpu(es->s_default_mount_opts);
 226
 227         if (sbi->s_sb_block != 1)
 228                 seq_printf(seq, ",sb=%lu", sbi->s_sb_block);
 229         if (test_opt(sb, MINIX_DF))
 230                 seq_puts(seq, ",minixdf");
 231         if (test_opt(sb, GRPID))
 232                 seq_puts(seq, ",grpid");
 233         if (!test_opt(sb, GRPID) && (def_mount_opts & EXT2_DEFM_BSDGROUPS))
 234                 seq_puts(seq, ",nogrpid");
 235         if (sbi->s_resuid != EXT2_DEF_RESUID ||
 236             le16_to_cpu(es->s_def_resuid) != EXT2_DEF_RESUID) {
 237                 seq_printf(seq, ",resuid=%u", sbi->s_resuid);
 238         }
 239         if (sbi->s_resgid != EXT2_DEF_RESGID ||
 240             le16_to_cpu(es->s_def_resgid) != EXT2_DEF_RESGID) {
 241                 seq_printf(seq, ",resgid=%u", sbi->s_resgid);
 242         }
 243         if (test_opt(sb, ERRORS_RO)) {
 244                 int def_errors = le16_to_cpu(es->s_errors);
 245
 246                 if (def_errors == EXT2_ERRORS_PANIC ||
 247                     def_errors == EXT2_ERRORS_CONTINUE) {
 248                         seq_puts(seq, ",errors=remount-ro");
 249                 }
 250         }
 251         if (test_opt(sb, ERRORS_CONT))
 252                 seq_puts(seq, ",errors=continue");
 253         if (test_opt(sb, ERRORS_PANIC))
 254                 seq_puts(seq, ",errors=panic");
 255         if (test_opt(sb, NO_UID32))
 256                 seq_puts(seq, ",nouid32");
 257         if (test_opt(sb, DEBUG))
 258                 seq_puts(seq, ",debug");
 259         if (test_opt(sb, OLDALLOC))
 260                 seq_puts(seq, ",oldalloc");
 261
 262 #ifdef CONFIG_EXT2_FS_XATTR
 263         if (test_opt(sb, XATTR_USER))
 264                 seq_puts(seq, ",user_xattr");
 265         if (!test_opt(sb, XATTR_USER) &&
 266             (def_mount_opts & EXT2_DEFM_XATTR_USER)) {
 267                 seq_puts(seq, ",nouser_xattr");
 268         }
 269 #endif
 270
 271 #ifdef CONFIG_EXT2_FS_POSIX_ACL
 272         if (test_opt(sb, POSIX_ACL))
 273                 seq_puts(seq, ",acl");
 274         if (!test_opt(sb, POSIX_ACL) && (def_mount_opts & EXT2_DEFM_ACL))
 275                 seq_puts(seq, ",noacl");
 276 #endif
 277
 278         if (test_opt(sb, NOBH))
 279                 seq_puts(seq, ",nobh");
 280
 281 #if defined(CONFIG_QUOTA)
 282         if (sbi->s_mount_opt & EXT2_MOUNT_USRQUOTA)
 283                 seq_puts(seq, ",usrquota");
 284
 285         if (sbi->s_mount_opt & EXT2_MOUNT_GRPQUOTA)
 286                 seq_puts(seq, ",grpquota");
 287 #endif
 288
 289 #if defined(CONFIG_EXT2_FS_XIP)
 290         if (sbi->s_mount_opt & EXT2_MOUNT_XIP)
 291                 seq_puts(seq, ",xip");
 292 #endif
 293
 294         if (!test_opt(sb, RESERVATION))
 295                 seq_puts(seq, ",noreservation");
 296
 297         return 0;
 298 }
 299
 300 #ifdef CONFIG_QUOTA
 301 static ssize_t ext2_quota_read(struct super_block *sb, int type, char *data, size_t len, loff_t off);
 302 static ssize_t ext2_quota_write(struct super_block *sb, int type, const char *data, size_t len, loff_t off);
 303 #endif
 304
 305 static const struct super_operations ext2_sops = {
 306         .alloc_inode    = ext2_alloc_inode,
 307         .destroy_inode  = ext2_destroy_inode,
 308         .write_inode    = ext2_write_inode,
 309         .delete_inode   = ext2_delete_inode,
 310         .put_super      = ext2_put_super,
 311         .write_super    = ext2_write_super,
 312         .statfs         = ext2_statfs,
 313         .remount_fs     = ext2_remount,
 314         .clear_inode    = ext2_clear_inode,
 315         .show_options   = ext2_show_options,
 316 #ifdef CONFIG_QUOTA
 317         .quota_read     = ext2_quota_read,
 318         .quota_write    = ext2_quota_write,
 319 #endif
 320 };
 321
 322 static struct inode *ext2_nfs_get_inode(struct super_block *sb,
 323                 u64 ino, u32 generation)
 324 {
 325         struct inode *inode;
 326
 327         if (ino < EXT2_FIRST_INO(sb) && ino != EXT2_ROOT_INO)
 328                 return ERR_PTR(-ESTALE);
 329         if (ino > le32_to_cpu(EXT2_SB(sb)->s_es->s_inodes_count))
 330                 return ERR_PTR(-ESTALE);
 331
 332         /* iget isn't really right if the inode is currently unallocated!!
 333          * ext2_read_inode currently does appropriate checks, but
 334          * it might be "neater" to call ext2_get_inode first and check
 335          * if the inode is valid.....
 336          */
 337         inode = ext2_iget(sb, ino);
 338         if (IS_ERR(inode))
 339                 return ERR_CAST(inode);
 340         if (generation && inode->i_generation != generation) {
 341                 /* we didn't find the right inode.. */
 342                 iput(inode);
 343                 return ERR_PTR(-ESTALE);
 344         }
 345         return inode;
 346 }
 347
 348 static struct dentry *ext2_fh_to_dentry(struct super_block *sb, struct fid *fid,
 349                 int fh_len, int fh_type)
 350 {
 351         return generic_fh_to_dentry(sb, fid, fh_len, fh_type,
 352                                     ext2_nfs_get_inode);
 353 }
 354
 355 static struct dentry *ext2_fh_to_parent(struct super_block *sb, struct fid *fid,
 356                 int fh_len, int fh_type)
 357 {
 358         return generic_fh_to_parent(sb, fid, fh_len, fh_type,
 359                                     ext2_nfs_get_inode);
 360 }
 361
 362 /* Yes, most of these are left as NULL!!
 363  * A NULL value implies the default, which works with ext2-like file
 364  * systems, but can be improved upon.
 365  * Currently only get_parent is required.
 366  */
 367 static const struct export_operations ext2_export_ops = {
 368         .fh_to_dentry = ext2_fh_to_dentry,
 369         .fh_to_parent = ext2_fh_to_parent,
 370         .get_parent = ext2_get_parent,
 371 };
 372
 373 static unsigned long get_sb_block(void **data)
 374 {
 375         unsigned long   sb_block;
 376         char            *options = (char *) *data;
 377
 378         if (!options || strncmp(options, "sb=", 3) != 0)
 379                 return 1;       /* Default location */
 380         options += 3;
 381         sb_block = simple_strtoul(options, &options, 0);
 382         if (*options && *options != ',') {
 383                 printk("EXT2-fs: Invalid sb specification: %s\n",
 384                        (char *) *data);
 385                 return 1;
 386         }
 387         if (*options == ',')
 388                 options++;
 389         *data = (void *) options;
 390         return sb_block;
 391 }
 392
 393 enum {
 394         Opt_bsd_df, Opt_minix_df, Opt_grpid, Opt_nogrpid,
 395         Opt_resgid, Opt_resuid, Opt_sb, Opt_err_cont, Opt_err_panic,
 396         Opt_err_ro, Opt_nouid32, Opt_nocheck, Opt_debug,
 397         Opt_oldalloc, Opt_orlov, Opt_nobh, Opt_user_xattr, Opt_nouser_xattr,
 398         Opt_acl, Opt_noacl, Opt_xip, Opt_ignore, Opt_err, Opt_quota,
 399         Opt_usrquota, Opt_grpquota, Opt_reservation, Opt_noreservation
 400 };
 401
 402 static const match_table_t tokens = {
 403         {Opt_bsd_df, "bsddf"},
 404         {Opt_minix_df, "minixdf"},
 405         {Opt_grpid, "grpid"},
 406         {Opt_grpid, "bsdgroups"},
 407         {Opt_nogrpid, "nogrpid"},
 408         {Opt_nogrpid, "sysvgroups"},
 409         {Opt_resgid, "resgid=%u"},
 410         {Opt_resuid, "resuid=%u"},
 411         {Opt_sb, "sb=%u"},
 412         {Opt_err_cont, "errors=continue"},
 413         {Opt_err_panic, "errors=panic"},
 414         {Opt_err_ro, "errors=remount-ro"},
 415         {Opt_nouid32, "nouid32"},
 416         {Opt_nocheck, "check=none"},
 417         {Opt_nocheck, "nocheck"},
 418         {Opt_debug, "debug"},
 419         {Opt_oldalloc, "oldalloc"},
 420         {Opt_orlov, "orlov"},
 421         {Opt_nobh, "nobh"},
 422         {Opt_user_xattr, "user_xattr"},
 423         {Opt_nouser_xattr, "nouser_xattr"},
 424         {Opt_acl, "acl"},
 425         {Opt_noacl, "noacl"},
 426         {Opt_xip, "xip"},
 427         {Opt_grpquota, "grpquota"},
 428         {Opt_ignore, "noquota"},
 429         {Opt_quota, "quota"},
 430         {Opt_usrquota, "usrquota"},
 431         {Opt_reservation, "reservation"},
 432         {Opt_noreservation, "noreservation"},
 433         {Opt_err, NULL}
 434 };
 435
 436 static int parse_options (char * options,
 437                           struct ext2_sb_info *sbi)
 438 {
 439         char * p;
 440         substring_t args[MAX_OPT_ARGS];
 441         int option;
 442
 443         if (!options)
 444                 return 1;
 445
 446         while ((p = strsep (&options, ",")) != NULL) {
 447                 int token;
 448                 if (!*p)
 449                         continue;
 450
 451                 token = match_token(p, tokens, args);
 452                 switch (token) {
 453                 case Opt_bsd_df:
 454                         clear_opt (sbi->s_mount_opt, MINIX_DF);
 455                         break;
 456                 case Opt_minix_df:
 457                         set_opt (sbi->s_mount_opt, MINIX_DF);
 458                         break;
 459                 case Opt_grpid:
 460                         set_opt (sbi->s_mount_opt, GRPID);
 461                         break;
 462                 case Opt_nogrpid:
 463                         clear_opt (sbi->s_mount_opt, GRPID);
 464                         break;
 465                 case Opt_resuid:
 466                         if (match_int(&args[0], &option))
 467                                 return 0;
 468                         sbi->s_resuid = option;
 469                         break;
 470                 case Opt_resgid:
 471                         if (match_int(&args[0], &option))
 472                                 return 0;
 473                         sbi->s_resgid = option;
 474                         break;
 475                 case Opt_sb:
 476                         /* handled by get_sb_block() instead of here */
 477                         /* *sb_block = match_int(&args[0]); */
 478                         break;
 479                 case Opt_err_panic:
 480                         clear_opt (sbi->s_mount_opt, ERRORS_CONT);
 481                         clear_opt (sbi->s_mount_opt, ERRORS_RO);
 482                         set_opt (sbi->s_mount_opt, ERRORS_PANIC);
 483                         break;
 484                 case Opt_err_ro:
 485                         clear_opt (sbi->s_mount_opt, ERRORS_CONT);
 486                         clear_opt (sbi->s_mount_opt, ERRORS_PANIC);
 487                         set_opt (sbi->s_mount_opt, ERRORS_RO);
 488                         break;
 489                 case Opt_err_cont:
 490                         clear_opt (sbi->s_mount_opt, ERRORS_RO);
 491                         clear_opt (sbi->s_mount_opt, ERRORS_PANIC);
 492                         set_opt (sbi->s_mount_opt, ERRORS_CONT);
 493                         break;
 494                 case Opt_nouid32:
 495                         set_opt (sbi->s_mount_opt, NO_UID32);
 496                         break;
 497                 case Opt_nocheck:
 498                         clear_opt (sbi->s_mount_opt, CHECK);
 499                         break;
 500                 case Opt_debug:
 501                         set_opt (sbi->s_mount_opt, DEBUG);
 502                         break;
 503                 case Opt_oldalloc:
 504                         set_opt (sbi->s_mount_opt, OLDALLOC);
 505                         break;
 506                 case Opt_orlov:
 507                         clear_opt (sbi->s_mount_opt, OLDALLOC);
 508                         break;
 509                 case Opt_nobh:
 510                         set_opt (sbi->s_mount_opt, NOBH);
 511                         break;
 512 #ifdef CONFIG_EXT2_FS_XATTR
 513                 case Opt_user_xattr:
 514                         set_opt (sbi->s_mount_opt, XATTR_USER);
 515                         break;
 516                 case Opt_nouser_xattr:
 517                         clear_opt (sbi->s_mount_opt, XATTR_USER);
 518                         break;
 519 #else
 520                 case Opt_user_xattr:
 521                 case Opt_nouser_xattr:
 522                         printk("EXT2 (no)user_xattr options not supported\n");
 523                         break;
 524 #endif
 525 #ifdef CONFIG_EXT2_FS_POSIX_ACL
 526                 case Opt_acl:
 527                         set_opt(sbi->s_mount_opt, POSIX_ACL);
 528                         break;
 529                 case Opt_noacl:
 530                         clear_opt(sbi->s_mount_opt, POSIX_ACL);
 531                         break;
 532 #else
 533                 case Opt_acl:
 534                 case Opt_noacl:
 535                         printk("EXT2 (no)acl options not supported\n");
 536                         break;
 537 #endif
 538                 case Opt_xip:
 539 #ifdef CONFIG_EXT2_FS_XIP
 540                         set_opt (sbi->s_mount_opt, XIP);
 541 #else
 542                         printk("EXT2 xip option not supported\n");
 543 #endif
 544                         break;
 545
 546 #if defined(CONFIG_QUOTA)
 547                 case Opt_quota:
 548                 case Opt_usrquota:
 549                         set_opt(sbi->s_mount_opt, USRQUOTA);
 550                         break;
 551
 552                 case Opt_grpquota:
 553                         set_opt(sbi->s_mount_opt, GRPQUOTA);
 554                         break;
 555 #else
 556                 case Opt_quota:
 557                 case Opt_usrquota:
 558                 case Opt_grpquota:
 559                         printk(KERN_ERR
 560                                 "EXT2-fs: quota operations not supported.\n");
 561
 562                         break;
 563 #endif
 564
 565                 case Opt_reservation:
 566                         set_opt(sbi->s_mount_opt, RESERVATION);
 567                         printk("reservations ON\n");
 568                         break;
 569                 case Opt_noreservation:
 570                         clear_opt(sbi->s_mount_opt, RESERVATION);
 571                         printk("reservations OFF\n");
 572                         break;
 573                 case Opt_ignore:
 574                         break;
 575                 default:
 576                         return 0;
 577                 }
 578         }
 579         return 1;
 580 }
 581
 582 static int ext2_setup_super (struct super_block * sb,
 583                               struct ext2_super_block * es,
 584                               int read_only)
 585 {
 586         int res = 0;
 587         struct ext2_sb_info *sbi = EXT2_SB(sb);
 588
 589         if (le32_to_cpu(es->s_rev_level) > EXT2_MAX_SUPP_REV) {
 590                 printk ("EXT2-fs warning: revision level too high, "
 591                         "forcing read-only mode\n");
 592                 res = MS_RDONLY;
 593         }
 594         if (read_only)
 595                 return res;
 596         if (!(sbi->s_mount_state & EXT2_VALID_FS))
 597                 printk ("EXT2-fs warning: mounting unchecked fs, "
 598                         "running e2fsck is recommended\n");
 599         else if ((sbi->s_mount_state & EXT2_ERROR_FS))
 600                 printk ("EXT2-fs warning: mounting fs with errors, "
 601                         "running e2fsck is recommended\n");
 602         else if ((__s16) le16_to_cpu(es->s_max_mnt_count) >= 0 &&
 603                  le16_to_cpu(es->s_mnt_count) >=
 604                  (unsigned short) (__s16) le16_to_cpu(es->s_max_mnt_count))
 605                 printk ("EXT2-fs warning: maximal mount count reached, "
 606                         "running e2fsck is recommended\n");
 607         else if (le32_to_cpu(es->s_checkinterval) &&
 608                 (le32_to_cpu(es->s_lastcheck) + le32_to_cpu(es->s_checkinterval) <= get_seconds()))
 609                 printk ("EXT2-fs warning: checktime reached, "
 610                         "running e2fsck is recommended\n");
 611         if (!le16_to_cpu(es->s_max_mnt_count))
 612                 es->s_max_mnt_count = cpu_to_le16(EXT2_DFL_MAX_MNT_COUNT);
 613         le16_add_cpu(&es->s_mnt_count, 1);
 614         ext2_write_super(sb);
 615         if (test_opt (sb, DEBUG))
 616                 printk ("[EXT II FS %s, %s, bs=%lu, fs=%lu, gc=%lu, "
 617                         "bpg=%lu, ipg=%lu, mo=%04lx]\n",
 618                         EXT2FS_VERSION, EXT2FS_DATE, sb->s_blocksize,
 619                         sbi->s_frag_size,
 620                         sbi->s_groups_count,
 621                         EXT2_BLOCKS_PER_GROUP(sb),
 622                         EXT2_INODES_PER_GROUP(sb),
 623                         sbi->s_mount_opt);
 624         return res;
 625 }
 626
 627 static int ext2_check_descriptors(struct super_block *sb)
 628 {
 629         int i;
 630         struct ext2_sb_info *sbi = EXT2_SB(sb);
 631
 632         ext2_debug ("Checking group descriptors");
 633
 634         for (i = 0; i < sbi->s_groups_count; i++) {
 635                 struct ext2_group_desc *gdp = ext2_get_group_desc(sb, i, NULL);
 636                 ext2_fsblk_t first_block = ext2_group_first_block_no(sb, i);
 637                 ext2_fsblk_t last_block;
 638
 639                 if (i == sbi->s_groups_count - 1)
 640                         last_block = le32_to_cpu(sbi->s_es->s_blocks_count) - 1;
 641                 else
 642                         last_block = first_block +
 643                                 (EXT2_BLOCKS_PER_GROUP(sb) - 1);
 644
 645                 if (le32_to_cpu(gdp->bg_block_bitmap) < first_block ||
 646                     le32_to_cpu(gdp->bg_block_bitmap) > last_block)
 647                 {
 648                         ext2_error (sb, "ext2_check_descriptors",
 649                                     "Block bitmap for group %d"
 650                                     " not in group (block %lu)!",
 651                                     i, (unsigned long) le32_to_cpu(gdp->bg_block_bitmap));
 652                         return 0;
 653                 }
 654                 if (le32_to_cpu(gdp->bg_inode_bitmap) < first_block ||
 655                     le32_to_cpu(gdp->bg_inode_bitmap) > last_block)
 656                 {
 657                         ext2_error (sb, "ext2_check_descriptors",
 658                                     "Inode bitmap for group %d"
 659                                     " not in group (block %lu)!",
 660                                     i, (unsigned long) le32_to_cpu(gdp->bg_inode_bitmap));
 661                         return 0;
 662                 }
 663                 if (le32_to_cpu(gdp->bg_inode_table) < first_block ||
 664                     le32_to_cpu(gdp->bg_inode_table) + sbi->s_itb_per_group - 1 >
 665                     last_block)
 666                 {
 667                         ext2_error (sb, "ext2_check_descriptors",
 668                                     "Inode table for group %d"
 669                                     " not in group (block %lu)!",
 670                                     i, (unsigned long) le32_to_cpu(gdp->bg_inode_table));
 671                         return 0;
 672                 }
 673         }
 674         return 1;
 675 }
 676
 677 /*
 678  * Maximal file size.  There is a direct, and {,double-,triple-}indirect
 679  * block limit, and also a limit of (2^32 - 1) 512-byte sectors in i_blocks.
 680  * We need to be 1 filesystem block less than the 2^32 sector limit.
 681  */
 682 static loff_t ext2_max_size(int bits)
 683 {
 684         loff_t res = EXT2_NDIR_BLOCKS;
 685         int meta_blocks;
 686         loff_t upper_limit;
 687
 688         /* This is calculated to be the largest file size for a
 689          * dense, file such that the total number of
 690          * sectors in the file, including data and all indirect blocks,
 691          * does not exceed 2^32 -1
 692          * __u32 i_blocks representing the total number of
 693          * 512 bytes blocks of the file
 694          */
 695         upper_limit = (1LL << 32) - 1;
 696
 697         /* total blocks in file system block size */
 698         upper_limit >>= (bits - 9);
 699
 700
 701         /* indirect blocks */
 702         meta_blocks = 1;
 703         /* double indirect blocks */
 704         meta_blocks += 1 + (1LL << (bits-2));
 705         /* tripple indirect blocks */
 706         meta_blocks += 1 + (1LL << (bits-2)) + (1LL << (2*(bits-2)));
 707
 708         upper_limit -= meta_blocks;
 709         upper_limit <<= bits;
 710
 711         res += 1LL << (bits-2);
 712         res += 1LL << (2*(bits-2));
 713         res += 1LL << (3*(bits-2));
 714         res <<= bits;
 715         if (res > upper_limit)
 716                 res = upper_limit;
 717
 718         if (res > MAX_LFS_FILESIZE)
 719                 res = MAX_LFS_FILESIZE;
 720
 721         return res;
 722 }
 723
 724 static unsigned long descriptor_loc(struct super_block *sb,
 725                                     unsigned long logic_sb_block,
 726                                     int nr)
 727 {
 728         struct ext2_sb_info *sbi = EXT2_SB(sb);
 729         unsigned long bg, first_meta_bg;
 730         int has_super = 0;
 731
 732         first_meta_bg = le32_to_cpu(sbi->s_es->s_first_meta_bg);
 733
 734         if (!EXT2_HAS_INCOMPAT_FEATURE(sb, EXT2_FEATURE_INCOMPAT_META_BG) ||
 735             nr < first_meta_bg)
 736                 return (logic_sb_block + nr + 1);
 737         bg = sbi->s_desc_per_block * nr;
 738         if (ext2_bg_has_super(sb, bg))
 739                 has_super = 1;
 740
 741         return ext2_group_first_block_no(sb, bg) + has_super;
 742 }
 743
 744 static int ext2_fill_super(struct super_block *sb, void *data, int silent)
 745 {
 746         struct buffer_head * bh;
 747         struct ext2_sb_info * sbi;
 748         struct ext2_super_block * es;
 749         struct inode *root;
 750         unsigned long block;
 751         unsigned long sb_block = get_sb_block(&data);
 752         unsigned long logic_sb_block;
 753         unsigned long offset = 0;
 754         unsigned long def_mount_opts;
 755         long ret = -EINVAL;
 756         int blocksize = BLOCK_SIZE;
 757         int db_count;
 758         int i, j;
 759         __le32 features;
 760         int err;
 761
 762         sbi = kzalloc(sizeof(*sbi), GFP_KERNEL);
 763         if (!sbi)
 764                 return -ENOMEM;
 765
 766         sbi->s_blockgroup_lock =
 767                 kzalloc(sizeof(struct blockgroup_lock), GFP_KERNEL);
 768         if (!sbi->s_blockgroup_lock) {
 769                 kfree(sbi);
 770                 return -ENOMEM;
 771         }
 772         sb->s_fs_info = sbi;
 773         sbi->s_sb_block = sb_block;
 774
 775         /*
 776          * See what the current blocksize for the device is, and
 777          * use that as the blocksize.  Otherwise (or if the blocksize
 778          * is smaller than the default) use the default.
 779          * This is important for devices that have a hardware
 780          * sectorsize that is larger than the default.
 781          */
 782         blocksize = sb_min_blocksize(sb, BLOCK_SIZE);
 783         if (!blocksize) {
 784                 printk ("EXT2-fs: unable to set blocksize\n");
 785                 goto failed_sbi;
 786         }
 787
 788         /*
 789          * If the superblock doesn't start on a hardware sector boundary,
 790          * calculate the offset.
 791          */
 792         if (blocksize != BLOCK_SIZE) {
 793                 logic_sb_block = (sb_block*BLOCK_SIZE) / blocksize;
 794                 offset = (sb_block*BLOCK_SIZE) % blocksize;
 795         } else {
 796                 logic_sb_block = sb_block;
 797         }
 798
 799         if (!(bh = sb_bread(sb, logic_sb_block))) {
 800                 printk ("EXT2-fs: unable to read superblock\n");
 801                 goto failed_sbi;
 802         }
 803         /*
 804          * Note: s_es must be initialized as soon as possible because
 805          *       some ext2 macro-instructions depend on its value
 806          */
 807         es = (struct ext2_super_block *) (((char *)bh->b_data) + offset);
 808         sbi->s_es = es;
 809         sb->s_magic = le16_to_cpu(es->s_magic);
 810
 811         if (sb->s_magic != EXT2_SUPER_MAGIC)
 812                 goto cantfind_ext2;
 813
 814         /* Set defaults before we parse the mount options */
 815         def_mount_opts = le32_to_cpu(es->s_default_mount_opts);
 816         if (def_mount_opts & EXT2_DEFM_DEBUG)
 817                 set_opt(sbi->s_mount_opt, DEBUG);
 818         if (def_mount_opts & EXT2_DEFM_BSDGROUPS)
 819                 set_opt(sbi->s_mount_opt, GRPID);
 820         if (def_mount_opts & EXT2_DEFM_UID16)
 821                 set_opt(sbi->s_mount_opt, NO_UID32);
 822 #ifdef CONFIG_EXT2_FS_XATTR
 823         if (def_mount_opts & EXT2_DEFM_XATTR_USER)
 824                 set_opt(sbi->s_mount_opt, XATTR_USER);
 825 #endif
 826 #ifdef CONFIG_EXT2_FS_POSIX_ACL
 827         if (def_mount_opts & EXT2_DEFM_ACL)
 828                 set_opt(sbi->s_mount_opt, POSIX_ACL);
 829 #endif
 830
 831         if (le16_to_cpu(sbi->s_es->s_errors) == EXT2_ERRORS_PANIC)
 832                 set_opt(sbi->s_mount_opt, ERRORS_PANIC);
 833         else if (le16_to_cpu(sbi->s_es->s_errors) == EXT2_ERRORS_CONTINUE)
 834                 set_opt(sbi->s_mount_opt, ERRORS_CONT);
 835         else
 836                 set_opt(sbi->s_mount_opt, ERRORS_RO);
 837
 838         sbi->s_resuid = le16_to_cpu(es->s_def_resuid);
 839         sbi->s_resgid = le16_to_cpu(es->s_def_resgid);
 840
 841         set_opt(sbi->s_mount_opt, RESERVATION);
 842
 843         if (!parse_options ((char *) data, sbi))
 844                 goto failed_mount;
 845
 846         sb->s_flags = (sb->s_flags & ~MS_POSIXACL) |
 847                 ((EXT2_SB(sb)->s_mount_opt & EXT2_MOUNT_POSIX_ACL) ?
 848                  MS_POSIXACL : 0);
 849
 850         ext2_xip_verify_sb(sb); /* see if bdev supports xip, unset
 851                                     EXT2_MOUNT_XIP if not */
 852
 853         if (le32_to_cpu(es->s_rev_level) == EXT2_GOOD_OLD_REV &&
 854             (EXT2_HAS_COMPAT_FEATURE(sb, ~0U) ||
 855              EXT2_HAS_RO_COMPAT_FEATURE(sb, ~0U) ||
 856              EXT2_HAS_INCOMPAT_FEATURE(sb, ~0U)))
 857                 printk("EXT2-fs warning: feature flags set on rev 0 fs, "
 858                        "running e2fsck is recommended\n");
 859         /*
 860          * Check feature flags regardless of the revision level, since we
 861          * previously didn't change the revision level when setting the flags,
 862          * so there is a chance incompat flags are set on a rev 0 filesystem.
 863          */
 864         features = EXT2_HAS_INCOMPAT_FEATURE(sb, ~EXT2_FEATURE_INCOMPAT_SUPP);
 865         if (features) {
 866                 printk("EXT2-fs: %s: couldn't mount because of "
 867                        "unsupported optional features (%x).\n",
 868                        sb->s_id, le32_to_cpu(features));
 869                 goto failed_mount;
 870         }
 871         if (!(sb->s_flags & MS_RDONLY) &&
 872             (features = EXT2_HAS_RO_COMPAT_FEATURE(sb, ~EXT2_FEATURE_RO_COMPAT_SUPP))){
 873                 printk("EXT2-fs: %s: couldn't mount RDWR because of "
 874                        "unsupported optional features (%x).\n",
 875                        sb->s_id, le32_to_cpu(features));
 876                 goto failed_mount;
 877         }
 878
 879         blocksize = BLOCK_SIZE << le32_to_cpu(sbi->s_es->s_log_block_size);
 880
 881         if (ext2_use_xip(sb) && blocksize != PAGE_SIZE) {
 882                 if (!silent)
 883                         printk("XIP: Unsupported blocksize\n");
 884                 goto failed_mount;
 885         }
 886
 887         /* If the blocksize doesn't match, re-read the thing.. */
 888         if (sb->s_blocksize != blocksize) {
 889                 brelse(bh);
 890
 891                 if (!sb_set_blocksize(sb, blocksize)) {
 892                         printk(KERN_ERR "EXT2-fs: blocksize too small for device.\n");
 893                         goto failed_sbi;
 894                 }
 895
 896                 logic_sb_block = (sb_block*BLOCK_SIZE) / blocksize;
 897                 offset = (sb_block*BLOCK_SIZE) % blocksize;
 898                 bh = sb_bread(sb, logic_sb_block);
 899                 if(!bh) {
 900                         printk("EXT2-fs: Couldn't read superblock on "
 901                                "2nd try.\n");
 902                         goto failed_sbi;
 903                 }
 904                 es = (struct ext2_super_block *) (((char *)bh->b_data) + offset);
 905                 sbi->s_es = es;
 906                 if (es->s_magic != cpu_to_le16(EXT2_SUPER_MAGIC)) {
 907                         printk ("EXT2-fs: Magic mismatch, very weird !\n");
 908                         goto failed_mount;
 909                 }
 910         }
 911
 912         sb->s_maxbytes = ext2_max_size(sb->s_blocksize_bits);
 913
 914         if (le32_to_cpu(es->s_rev_level) == EXT2_GOOD_OLD_REV) {
 915                 sbi->s_inode_size = EXT2_GOOD_OLD_INODE_SIZE;
 916                 sbi->s_first_ino = EXT2_GOOD_OLD_FIRST_INO;
 917         } else {
 918                 sbi->s_inode_size = le16_to_cpu(es->s_inode_size);
 919                 sbi->s_first_ino = le32_to_cpu(es->s_first_ino);
 920                 if ((sbi->s_inode_size < EXT2_GOOD_OLD_INODE_SIZE) ||
 921                     !is_power_of_2(sbi->s_inode_size) ||
 922                     (sbi->s_inode_size > blocksize)) {
 923                         printk ("EXT2-fs: unsupported inode size: %d\n",
 924                                 sbi->s_inode_size);
 925                         goto failed_mount;
 926                 }
 927         }
 928
 929         sbi->s_frag_size = EXT2_MIN_FRAG_SIZE <<
 930                                    le32_to_cpu(es->s_log_frag_size);
 931         if (sbi->s_frag_size == 0)
 932                 goto cantfind_ext2;
 933         sbi->s_frags_per_block = sb->s_blocksize / sbi->s_frag_size;
 934
 935         sbi->s_blocks_per_group = le32_to_cpu(es->s_blocks_per_group);
 936         sbi->s_frags_per_group = le32_to_cpu(es->s_frags_per_group);
 937         sbi->s_inodes_per_group = le32_to_cpu(es->s_inodes_per_group);
 938
 939         if (EXT2_INODE_SIZE(sb) == 0)
 940                 goto cantfind_ext2;
 941         sbi->s_inodes_per_block = sb->s_blocksize / EXT2_INODE_SIZE(sb);
 942         if (sbi->s_inodes_per_block == 0 || sbi->s_inodes_per_group == 0)
 943                 goto cantfind_ext2;
 944         sbi->s_itb_per_group = sbi->s_inodes_per_group /
 945                                         sbi->s_inodes_per_block;
 946         sbi->s_desc_per_block = sb->s_blocksize /
 947                                         sizeof (struct ext2_group_desc);
 948         sbi->s_sbh = bh;
 949         sbi->s_mount_state = le16_to_cpu(es->s_state);
 950         sbi->s_addr_per_block_bits =
 951                 ilog2 (EXT2_ADDR_PER_BLOCK(sb));
 952         sbi->s_desc_per_block_bits =
 953                 ilog2 (EXT2_DESC_PER_BLOCK(sb));
 954
 955         if (sb->s_magic != EXT2_SUPER_MAGIC)
 956                 goto cantfind_ext2;
 957
 958         if (sb->s_blocksize != bh->b_size) {
 959                 if (!silent)
 960                         printk ("VFS: Unsupported blocksize on dev "
 961                                 "%s.\n", sb->s_id);
 962                 goto failed_mount;
 963         }
 964
 965         if (sb->s_blocksize != sbi->s_frag_size) {
 966                 printk ("EXT2-fs: fragsize %lu != blocksize %lu (not supported yet)\n",
 967                         sbi->s_frag_size, sb->s_blocksize);
 968                 goto failed_mount;
 969         }
 970
 971         if (sbi->s_blocks_per_group > sb->s_blocksize * 8) {
 972                 printk ("EXT2-fs: #blocks per group too big: %lu\n",
 973                         sbi->s_blocks_per_group);
 974                 goto failed_mount;
 975         }
 976         if (sbi->s_frags_per_group > sb->s_blocksize * 8) {
 977                 printk ("EXT2-fs: #fragments per group too big: %lu\n",
 978                         sbi->s_frags_per_group);
 979                 goto failed_mount;
 980         }
 981         if (sbi->s_inodes_per_group > sb->s_blocksize * 8) {
 982                 printk ("EXT2-fs: #inodes per group too big: %lu\n",
 983                         sbi->s_inodes_per_group);
 984                 goto failed_mount;
 985         }
 986
 987         if (EXT2_BLOCKS_PER_GROUP(sb) == 0)
 988                 goto cantfind_ext2;
 989         sbi->s_groups_count = ((le32_to_cpu(es->s_blocks_count) -
 990                                 le32_to_cpu(es->s_first_data_block) - 1)
 991                                         / EXT2_BLOCKS_PER_GROUP(sb)) + 1;
 992         db_count = (sbi->s_groups_count + EXT2_DESC_PER_BLOCK(sb) - 1) /
 993                    EXT2_DESC_PER_BLOCK(sb);
 994         sbi->s_group_desc = kmalloc (db_count * sizeof (struct buffer_head *), GFP_KERNEL);
 995         if (sbi->s_group_desc == NULL) {
 996                 printk ("EXT2-fs: not enough memory\n");
 997                 goto failed_mount;
 998         }
 999         bgl_lock_init(sbi->s_blockgroup_lock);
1000         sbi->s_debts = kcalloc(sbi->s_groups_count, sizeof(*sbi->s_debts), GFP_KERNEL);
1001         if (!sbi->s_debts) {
1002                 printk ("EXT2-fs: not enough memory\n");
1003                 goto failed_mount_group_desc;
1004         }
1005         for (i = 0; i < db_count; i++) {
1006                 block = descriptor_loc(sb, logic_sb_block, i);
1007                 sbi->s_group_desc[i] = sb_bread(sb, block);
1008                 if (!sbi->s_group_desc[i]) {
1009                         for (j = 0; j < i; j++)
1010                                 brelse (sbi->s_group_desc[j]);
1011                         printk ("EXT2-fs: unable to read group descriptors\n");
1012                         goto failed_mount_group_desc;
1013                 }
1014         }
1015         if (!ext2_check_descriptors (sb)) {
1016                 printk ("EXT2-fs: group descriptors corrupted!\n");
1017                 goto failed_mount2;
1018         }
1019         sbi->s_gdb_count = db_count;
1020         get_random_bytes(&sbi->s_next_generation, sizeof(u32));
1021         spin_lock_init(&sbi->s_next_gen_lock);
1022
1023         /* per fileystem reservation list head & lock */
1024         spin_lock_init(&sbi->s_rsv_window_lock);
1025         sbi->s_rsv_window_root = RB_ROOT;
1026         /*
1027          * Add a single, static dummy reservation to the start of the
1028          * reservation window list --- it gives us a placeholder for
1029          * append-at-start-of-list which makes the allocation logic
1030          * _much_ simpler.
1031          */
1032         sbi->s_rsv_window_head.rsv_start = EXT2_RESERVE_WINDOW_NOT_ALLOCATED;
1033         sbi->s_rsv_window_head.rsv_end = EXT2_RESERVE_WINDOW_NOT_ALLOCATED;
1034         sbi->s_rsv_window_head.rsv_alloc_hit = 0;
1035         sbi->s_rsv_window_head.rsv_goal_size = 0;
1036         ext2_rsv_window_add(sb, &sbi->s_rsv_window_head);
1037
1038         err = percpu_counter_init(&sbi->s_freeblocks_counter,
1039                                 ext2_count_free_blocks(sb));
1040         if (!err) {
1041                 err = percpu_counter_init(&sbi->s_freeinodes_counter,
1042                                 ext2_count_free_inodes(sb));
1043         }
1044         if (!err) {
1045                 err = percpu_counter_init(&sbi->s_dirs_counter,
1046                                 ext2_count_dirs(sb));
1047         }
1048         if (err) {
1049                 printk(KERN_ERR "EXT2-fs: insufficient memory\n");
1050                 goto failed_mount3;
1051         }
1052         /*
1053          * set up enough so that it can read an inode
1054          */
1055         sb->s_op = &ext2_sops;
1056         sb->s_export_op = &ext2_export_ops;
1057         sb->s_xattr = ext2_xattr_handlers;
1058         root = ext2_iget(sb, EXT2_ROOT_INO);
1059         if (IS_ERR(root)) {
1060                 ret = PTR_ERR(root);
1061                 goto failed_mount3;
1062         }
1063         if (!S_ISDIR(root->i_mode) || !root->i_blocks || !root->i_size) {
1064                 iput(root);
1065                 printk(KERN_ERR "EXT2-fs: corrupt root inode, run e2fsck\n");
1066                 goto failed_mount3;
1067         }
1068
1069         sb->s_root = d_alloc_root(root);
1070         if (!sb->s_root) {
1071                 iput(root);
1072                 printk(KERN_ERR "EXT2-fs: get root inode failed\n");
1073                 ret = -ENOMEM;
1074                 goto failed_mount3;
1075         }
1076         if (EXT2_HAS_COMPAT_FEATURE(sb, EXT3_FEATURE_COMPAT_HAS_JOURNAL))
1077                 ext2_warning(sb, __func__,
1078                         "mounting ext3 filesystem as ext2");
1079         ext2_setup_super (sb, es, sb->s_flags & MS_RDONLY);
1080         return 0;
1081
1082 cantfind_ext2:
1083         if (!silent)
1084                 printk("VFS: Can't find an ext2 filesystem on dev %s.\n",
1085                        sb->s_id);
1086         goto failed_mount;
1087 failed_mount3:
1088         percpu_counter_destroy(&sbi->s_freeblocks_counter);
1089         percpu_counter_destroy(&sbi->s_freeinodes_counter);
1090         percpu_counter_destroy(&sbi->s_dirs_counter);
1091 failed_mount2:
1092         for (i = 0; i < db_count; i++)
1093                 brelse(sbi->s_group_desc[i]);
1094 failed_mount_group_desc:
1095         kfree(sbi->s_group_desc);
1096         kfree(sbi->s_debts);
1097 failed_mount:
1098         brelse(bh);
1099 failed_sbi:
1100         sb->s_fs_info = NULL;
1101         kfree(sbi->s_blockgroup_lock);
1102         kfree(sbi);
1103         return ret;
1104 }
1105
1106 static void ext2_commit_super (struct super_block * sb,
1107                                struct ext2_super_block * es)
1108 {
1109         es->s_wtime = cpu_to_le32(get_seconds());
1110         mark_buffer_dirty(EXT2_SB(sb)->s_sbh);
1111         sb->s_dirt = 0;
1112 }
1113
1114 static void ext2_sync_super(struct super_block *sb, struct ext2_super_block *es)
1115 {
1116         es->s_free_blocks_count = cpu_to_le32(ext2_count_free_blocks(sb));
1117         es->s_free_inodes_count = cpu_to_le32(ext2_count_free_inodes(sb));
1118         es->s_wtime = cpu_to_le32(get_seconds());
1119         mark_buffer_dirty(EXT2_SB(sb)->s_sbh);
1120         sync_dirty_buffer(EXT2_SB(sb)->s_sbh);
1121         sb->s_dirt = 0;
1122 }
1123
1124 /*
1125  * In the second extended file system, it is not necessary to
1126  * write the super block since we use a mapping of the
1127  * disk super block in a buffer.
1128  *
1129  * However, this function is still used to set the fs valid
1130  * flags to 0.  We need to set this flag to 0 since the fs
1131  * may have been checked while mounted and e2fsck may have
1132  * set s_state to EXT2_VALID_FS after some corrections.
1133  */
1134
1135 void ext2_write_super (struct super_block * sb)
1136 {
1137         struct ext2_super_block * es;
1138         lock_kernel();
1139         if (!(sb->s_flags & MS_RDONLY)) {
1140                 es = EXT2_SB(sb)->s_es;
1141
1142                 if (es->s_state & cpu_to_le16(EXT2_VALID_FS)) {
1143                         ext2_debug ("setting valid to 0\n");
1144                         es->s_state &= cpu_to_le16(~EXT2_VALID_FS);
1145                         es->s_free_blocks_count = cpu_to_le32(ext2_count_free_blocks(sb));
1146                         es->s_free_inodes_count = cpu_to_le32(ext2_count_free_inodes(sb));
1147                         es->s_mtime = cpu_to_le32(get_seconds());
1148                         ext2_sync_super(sb, es);
1149                 } else
1150                         ext2_commit_super (sb, es);
1151         }
1152         sb->s_dirt = 0;
1153         unlock_kernel();
1154 }
1155
1156 static int ext2_remount (struct super_block * sb, int * flags, char * data)
1157 {
1158         struct ext2_sb_info * sbi = EXT2_SB(sb);
1159         struct ext2_super_block * es;
1160         unsigned long old_mount_opt = sbi->s_mount_opt;
1161         struct ext2_mount_options old_opts;
1162         unsigned long old_sb_flags;
1163         int err;
1164
1165         /* Store the old options */
1166         old_sb_flags = sb->s_flags;
1167         old_opts.s_mount_opt = sbi->s_mount_opt;
1168         old_opts.s_resuid = sbi->s_resuid;
1169         old_opts.s_resgid = sbi->s_resgid;
1170
1171         /*
1172          * Allow the "check" option to be passed as a remount option.
1173          */
1174         if (!parse_options (data, sbi)) {
1175                 err = -EINVAL;
1176                 goto restore_opts;
1177         }
1178
1179         sb->s_flags = (sb->s_flags & ~MS_POSIXACL) |
1180                 ((sbi->s_mount_opt & EXT2_MOUNT_POSIX_ACL) ? MS_POSIXACL : 0);
1181
1182         ext2_xip_verify_sb(sb); /* see if bdev supports xip, unset
1183                                     EXT2_MOUNT_XIP if not */
1184
1185         if ((ext2_use_xip(sb)) && (sb->s_blocksize != PAGE_SIZE)) {
1186                 printk("XIP: Unsupported blocksize\n");
1187                 err = -EINVAL;
1188                 goto restore_opts;
1189         }
1190
1191         es = sbi->s_es;
1192         if (((sbi->s_mount_opt & EXT2_MOUNT_XIP) !=
1193             (old_mount_opt & EXT2_MOUNT_XIP)) &&
1194             invalidate_inodes(sb)) {
1195                 ext2_warning(sb, __func__, "refusing change of xip flag "
1196                              "with busy inodes while remounting");
1197                 sbi->s_mount_opt &= ~EXT2_MOUNT_XIP;
1198                 sbi->s_mount_opt |= old_mount_opt & EXT2_MOUNT_XIP;
1199         }
1200         if ((*flags & MS_RDONLY) == (sb->s_flags & MS_RDONLY))
1201                 return 0;
1202         if (*flags & MS_RDONLY) {
1203                 if (le16_to_cpu(es->s_state) & EXT2_VALID_FS ||
1204                     !(sbi->s_mount_state & EXT2_VALID_FS))
1205                         return 0;
1206                 /*
1207                  * OK, we are remounting a valid rw partition rdonly, so set
1208                  * the rdonly flag and then mark the partition as valid again.
1209                  */
1210                 es->s_state = cpu_to_le16(sbi->s_mount_state);
1211                 es->s_mtime = cpu_to_le32(get_seconds());
1212         } else {
1213                 __le32 ret = EXT2_HAS_RO_COMPAT_FEATURE(sb,
1214                                                ~EXT2_FEATURE_RO_COMPAT_SUPP);
1215                 if (ret) {
1216                         printk("EXT2-fs: %s: couldn't remount RDWR because of "
1217                                "unsupported optional features (%x).\n",
1218                                sb->s_id, le32_to_cpu(ret));
1219                         err = -EROFS;
1220                         goto restore_opts;
1221                 }
1222                 /*
1223                  * Mounting a RDONLY partition read-write, so reread and
1224                  * store the current valid flag.  (It may have been changed
1225                  * by e2fsck since we originally mounted the partition.)
1226                  */
1227                 sbi->s_mount_state = le16_to_cpu(es->s_state);
1228                 if (!ext2_setup_super (sb, es, 0))
1229                         sb->s_flags &= ~MS_RDONLY;
1230         }
1231         ext2_sync_super(sb, es);
1232         return 0;
1233 restore_opts:
1234         sbi->s_mount_opt = old_opts.s_mount_opt;
1235         sbi->s_resuid = old_opts.s_resuid;
1236         sbi->s_resgid = old_opts.s_resgid;
1237         sb->s_flags = old_sb_flags;
1238         return err;
1239 }
1240
1241 static int ext2_statfs (struct dentry * dentry, struct kstatfs * buf)
1242 {
1243         struct super_block *sb = dentry->d_sb;
1244         struct ext2_sb_info *sbi = EXT2_SB(sb);
1245         struct ext2_super_block *es = sbi->s_es;
1246         u64 fsid;
1247
1248         if (test_opt (sb, MINIX_DF))
1249                 sbi->s_overhead_last = 0;
1250         else if (sbi->s_blocks_last != le32_to_cpu(es->s_blocks_count)) {
1251                 unsigned long i, overhead = 0;
1252                 smp_rmb();
1253
1254                 /*
1255                  * Compute the overhead (FS structures). This is constant
1256                  * for a given filesystem unless the number of block groups
1257                  * changes so we cache the previous value until it does.
1258                  */
1259
1260                 /*
1261                  * All of the blocks before first_data_block are
1262                  * overhead
1263                  */
1264                 overhead = le32_to_cpu(es->s_first_data_block);
1265
1266                 /*
1267                  * Add the overhead attributed to the superblock and
1268                  * block group descriptors.  If the sparse superblocks
1269                  * feature is turned on, then not all groups have this.
1270                  */
1271                 for (i = 0; i < sbi->s_groups_count; i++)
1272                         overhead += ext2_bg_has_super(sb, i) +
1273                                 ext2_bg_num_gdb(sb, i);
1274
1275                 /*
1276                  * Every block group has an inode bitmap, a block
1277                  * bitmap, and an inode table.
1278                  */
1279                 overhead += (sbi->s_groups_count *
1280                              (2 + sbi->s_itb_per_group));
1281                 sbi->s_overhead_last = overhead;
1282                 smp_wmb();
1283                 sbi->s_blocks_last = le32_to_cpu(es->s_blocks_count);
1284         }
1285
1286         buf->f_type = EXT2_SUPER_MAGIC;
1287         buf->f_bsize = sb->s_blocksize;
1288         buf->f_blocks = le32_to_cpu(es->s_blocks_count) - sbi->s_overhead_last;
1289         buf->f_bfree = ext2_count_free_blocks(sb);
1290         es->s_free_blocks_count = cpu_to_le32(buf->f_bfree);
1291         buf->f_bavail = buf->f_bfree - le32_to_cpu(es->s_r_blocks_count);
1292         if (buf->f_bfree < le32_to_cpu(es->s_r_blocks_count))
1293                 buf->f_bavail = 0;
1294         buf->f_files = le32_to_cpu(es->s_inodes_count);
1295         buf->f_ffree = ext2_count_free_inodes(sb);
1296         es->s_free_inodes_count = cpu_to_le32(buf->f_ffree);
1297         buf->f_namelen = EXT2_NAME_LEN;
1298         fsid = le64_to_cpup((void *)es->s_uuid) ^
1299                le64_to_cpup((void *)es->s_uuid + sizeof(u64));
1300         buf->f_fsid.val[0] = fsid & 0xFFFFFFFFUL;
1301         buf->f_fsid.val[1] = (fsid >> 32) & 0xFFFFFFFFUL;
1302         return 0;
1303 }
1304
1305 static int ext2_get_sb(struct file_system_type *fs_type,
1306         int flags, const char *dev_name, void *data, struct vfsmount *mnt)
1307 {
1308         return get_sb_bdev(fs_type, flags, dev_name, data, ext2_fill_super, mnt);
1309 }
1310
1311 #ifdef CONFIG_QUOTA
1312
1313 /* Read data from quotafile - avoid pagecache and such because we cannot afford
1314  * acquiring the locks... As quota files are never truncated and quota code
1315  * itself serializes the operations (and noone else should touch the files)
1316  * we don't have to be afraid of races */
1317 static ssize_t ext2_quota_read(struct super_block *sb, int type, char *data,
1318                                size_t len, loff_t off)
1319 {
1320         struct inode *inode = sb_dqopt(sb)->files[type];
1321         sector_t blk = off >> EXT2_BLOCK_SIZE_BITS(sb);
1322         int err = 0;
1323         int offset = off & (sb->s_blocksize - 1);
1324         int tocopy;
1325         size_t toread;
1326         struct buffer_head tmp_bh;
1327         struct buffer_head *bh;
1328         loff_t i_size = i_size_read(inode);
1329
1330         if (off > i_size)
1331                 return 0;
1332         if (off+len > i_size)
1333                 len = i_size-off;
1334         toread = len;
1335         while (toread > 0) {
1336                 tocopy = sb->s_blocksize - offset < toread ?
1337                                 sb->s_blocksize - offset : toread;
1338
1339                 tmp_bh.b_state = 0;
1340                 tmp_bh.b_size = sb->s_blocksize;
1341                 err = ext2_get_block(inode, blk, &tmp_bh, 0);
1342                 if (err < 0)
1343                         return err;
1344                 if (!buffer_mapped(&tmp_bh))    /* A hole? */
1345                         memset(data, 0, tocopy);
1346                 else {
1347                         bh = sb_bread(sb, tmp_bh.b_blocknr);
1348                         if (!bh)
1349                                 return -EIO;
1350                         memcpy(data, bh->b_data+offset, tocopy);
1351                         brelse(bh);
1352                 }
1353                 offset = 0;
1354                 toread -= tocopy;
1355                 data += tocopy;
1356                 blk++;
1357         }
1358         return len;
1359 }
1360
1361 /* Write to quotafile */
1362 static ssize_t ext2_quota_write(struct super_block *sb, int type,
1363                                 const char *data, size_t len, loff_t off)
1364 {
1365         struct inode *inode = sb_dqopt(sb)->files[type];
1366         sector_t blk = off >> EXT2_BLOCK_SIZE_BITS(sb);
1367         int err = 0;
1368         int offset = off & (sb->s_blocksize - 1);
1369         int tocopy;
1370         size_t towrite = len;
1371         struct buffer_head tmp_bh;
1372         struct buffer_head *bh;
1373
1374         mutex_lock_nested(&inode->i_mutex, I_MUTEX_QUOTA);
1375         while (towrite > 0) {
1376                 tocopy = sb->s_blocksize - offset < towrite ?
1377                                 sb->s_blocksize - offset : towrite;
1378
1379                 tmp_bh.b_state = 0;
1380                 err = ext2_get_block(inode, blk, &tmp_bh, 1);
1381                 if (err < 0)
1382                         goto out;
1383                 if (offset || tocopy != EXT2_BLOCK_SIZE(sb))
1384                         bh = sb_bread(sb, tmp_bh.b_blocknr);
1385                 else
1386                         bh = sb_getblk(sb, tmp_bh.b_blocknr);
1387                 if (!bh) {
1388                         err = -EIO;
1389                         goto out;
1390                 }
1391                 lock_buffer(bh);
1392                 memcpy(bh->b_data+offset, data, tocopy);
1393                 flush_dcache_page(bh->b_page);
1394                 set_buffer_uptodate(bh);
1395                 mark_buffer_dirty(bh);
1396                 unlock_buffer(bh);
1397                 brelse(bh);
1398                 offset = 0;
1399                 towrite -= tocopy;
1400                 data += tocopy;
1401                 blk++;
1402         }
1403 out:
1404         if (len == towrite) {
1405                 mutex_unlock(&inode->i_mutex);
1406                 return err;
1407         }
1408         if (inode->i_size < off+len-towrite)
1409                 i_size_write(inode, off+len-towrite);
1410         inode->i_version++;
1411         inode->i_mtime = inode->i_ctime = CURRENT_TIME;
1412         mark_inode_dirty(inode);
1413         mutex_unlock(&inode->i_mutex);
1414         return len - towrite;
1415 }
1416
1417 #endif
1418
1419 static struct file_system_type ext2_fs_type = {
1420         .owner          = THIS_MODULE,
1421         .name           = "ext2",
1422         .get_sb         = ext2_get_sb,
1423         .kill_sb        = kill_block_super,
1424         .fs_flags       = FS_REQUIRES_DEV,
1425 };
1426
1427 static int __init init_ext2_fs(void)
1428 {
1429         int err = init_ext2_xattr();
1430         if (err)
1431                 return err;
1432         err = init_inodecache();
1433         if (err)
1434                 goto out1;
1435         err = register_filesystem(&ext2_fs_type);
1436         if (err)
1437                 goto out;
1438         return 0;
1439 out:
1440         destroy_inodecache();
1441 out1:
1442         exit_ext2_xattr();
1443         return err;
1444 }
1445
1446 static void __exit exit_ext2_fs(void)
1447 {
1448         unregister_filesystem(&ext2_fs_type);
1449         destroy_inodecache();
1450         exit_ext2_xattr();
1451 }
1452
1453 module_init(init_ext2_fs)
1454 module_exit(exit_ext2_fs)