fs/ext4/super.c

   1 /*
   2  *  linux/fs/ext4/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/time.h>
  23 #include <linux/vmalloc.h>
  24 #include <linux/jbd2.h>
  25 #include <linux/slab.h>
  26 #include <linux/init.h>
  27 #include <linux/blkdev.h>
  28 #include <linux/parser.h>
  29 #include <linux/smp_lock.h>
  30 #include <linux/buffer_head.h>
  31 #include <linux/exportfs.h>
  32 #include <linux/vfs.h>
  33 #include <linux/random.h>
  34 #include <linux/mount.h>
  35 #include <linux/namei.h>
  36 #include <linux/quotaops.h>
  37 #include <linux/seq_file.h>
  38 #include <linux/proc_fs.h>
  39 #include <linux/ctype.h>
  40 #include <linux/log2.h>
  41 #include <linux/crc16.h>
  42 #include <asm/uaccess.h>
  43
  44 #include "ext4.h"
  45 #include "ext4_jbd2.h"
  46 #include "xattr.h"
  47 #include "acl.h"
  48
  49 #define CREATE_TRACE_POINTS
  50 #include <trace/events/ext4.h>
  51
  52 static int default_mb_history_length = 1000;
  53
  54 module_param_named(default_mb_history_length, default_mb_history_length,
  55                    int, 0644);
  56 MODULE_PARM_DESC(default_mb_history_length,
  57                  "Default number of entries saved for mb_history");
  58
  59 struct proc_dir_entry *ext4_proc_root;
  60 static struct kset *ext4_kset;
  61
  62 static int ext4_load_journal(struct super_block *, struct ext4_super_block *,
  63                              unsigned long journal_devnum);
  64 static int ext4_commit_super(struct super_block *sb, int sync);
  65 static void ext4_mark_recovery_complete(struct super_block *sb,
  66                                         struct ext4_super_block *es);
  67 static void ext4_clear_journal_err(struct super_block *sb,
  68                                    struct ext4_super_block *es);
  69 static int ext4_sync_fs(struct super_block *sb, int wait);
  70 static const char *ext4_decode_error(struct super_block *sb, int errno,
  71                                      char nbuf[16]);
  72 static int ext4_remount(struct super_block *sb, int *flags, char *data);
  73 static int ext4_statfs(struct dentry *dentry, struct kstatfs *buf);
  74 static int ext4_unfreeze(struct super_block *sb);
  75 static void ext4_write_super(struct super_block *sb);
  76 static int ext4_freeze(struct super_block *sb);
  77
  78
  79 ext4_fsblk_t ext4_block_bitmap(struct super_block *sb,
  80                                struct ext4_group_desc *bg)
  81 {
  82         return le32_to_cpu(bg->bg_block_bitmap_lo) |
  83                 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
  84                  (ext4_fsblk_t)le32_to_cpu(bg->bg_block_bitmap_hi) << 32 : 0);
  85 }
  86
  87 ext4_fsblk_t ext4_inode_bitmap(struct super_block *sb,
  88                                struct ext4_group_desc *bg)
  89 {
  90         return le32_to_cpu(bg->bg_inode_bitmap_lo) |
  91                 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
  92                  (ext4_fsblk_t)le32_to_cpu(bg->bg_inode_bitmap_hi) << 32 : 0);
  93 }
  94
  95 ext4_fsblk_t ext4_inode_table(struct super_block *sb,
  96                               struct ext4_group_desc *bg)
  97 {
  98         return le32_to_cpu(bg->bg_inode_table_lo) |
  99                 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
 100                  (ext4_fsblk_t)le32_to_cpu(bg->bg_inode_table_hi) << 32 : 0);
 101 }
 102
 103 __u32 ext4_free_blks_count(struct super_block *sb,
 104                               struct ext4_group_desc *bg)
 105 {
 106         return le16_to_cpu(bg->bg_free_blocks_count_lo) |
 107                 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
 108                  (__u32)le16_to_cpu(bg->bg_free_blocks_count_hi) << 16 : 0);
 109 }
 110
 111 __u32 ext4_free_inodes_count(struct super_block *sb,
 112                               struct ext4_group_desc *bg)
 113 {
 114         return le16_to_cpu(bg->bg_free_inodes_count_lo) |
 115                 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
 116                  (__u32)le16_to_cpu(bg->bg_free_inodes_count_hi) << 16 : 0);
 117 }
 118
 119 __u32 ext4_used_dirs_count(struct super_block *sb,
 120                               struct ext4_group_desc *bg)
 121 {
 122         return le16_to_cpu(bg->bg_used_dirs_count_lo) |
 123                 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
 124                  (__u32)le16_to_cpu(bg->bg_used_dirs_count_hi) << 16 : 0);
 125 }
 126
 127 __u32 ext4_itable_unused_count(struct super_block *sb,
 128                               struct ext4_group_desc *bg)
 129 {
 130         return le16_to_cpu(bg->bg_itable_unused_lo) |
 131                 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
 132                  (__u32)le16_to_cpu(bg->bg_itable_unused_hi) << 16 : 0);
 133 }
 134
 135 void ext4_block_bitmap_set(struct super_block *sb,
 136                            struct ext4_group_desc *bg, ext4_fsblk_t blk)
 137 {
 138         bg->bg_block_bitmap_lo = cpu_to_le32((u32)blk);
 139         if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
 140                 bg->bg_block_bitmap_hi = cpu_to_le32(blk >> 32);
 141 }
 142
 143 void ext4_inode_bitmap_set(struct super_block *sb,
 144                            struct ext4_group_desc *bg, ext4_fsblk_t blk)
 145 {
 146         bg->bg_inode_bitmap_lo  = cpu_to_le32((u32)blk);
 147         if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
 148                 bg->bg_inode_bitmap_hi = cpu_to_le32(blk >> 32);
 149 }
 150
 151 void ext4_inode_table_set(struct super_block *sb,
 152                           struct ext4_group_desc *bg, ext4_fsblk_t blk)
 153 {
 154         bg->bg_inode_table_lo = cpu_to_le32((u32)blk);
 155         if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
 156                 bg->bg_inode_table_hi = cpu_to_le32(blk >> 32);
 157 }
 158
 159 void ext4_free_blks_set(struct super_block *sb,
 160                           struct ext4_group_desc *bg, __u32 count)
 161 {
 162         bg->bg_free_blocks_count_lo = cpu_to_le16((__u16)count);
 163         if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
 164                 bg->bg_free_blocks_count_hi = cpu_to_le16(count >> 16);
 165 }
 166
 167 void ext4_free_inodes_set(struct super_block *sb,
 168                           struct ext4_group_desc *bg, __u32 count)
 169 {
 170         bg->bg_free_inodes_count_lo = cpu_to_le16((__u16)count);
 171         if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
 172                 bg->bg_free_inodes_count_hi = cpu_to_le16(count >> 16);
 173 }
 174
 175 void ext4_used_dirs_set(struct super_block *sb,
 176                           struct ext4_group_desc *bg, __u32 count)
 177 {
 178         bg->bg_used_dirs_count_lo = cpu_to_le16((__u16)count);
 179         if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
 180                 bg->bg_used_dirs_count_hi = cpu_to_le16(count >> 16);
 181 }
 182
 183 void ext4_itable_unused_set(struct super_block *sb,
 184                           struct ext4_group_desc *bg, __u32 count)
 185 {
 186         bg->bg_itable_unused_lo = cpu_to_le16((__u16)count);
 187         if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
 188                 bg->bg_itable_unused_hi = cpu_to_le16(count >> 16);
 189 }
 190
 191 /*
 192  * Wrappers for jbd2_journal_start/end.
 193  *
 194  * The only special thing we need to do here is to make sure that all
 195  * journal_end calls result in the superblock being marked dirty, so
 196  * that sync() will call the filesystem's write_super callback if
 197  * appropriate.
 198  */
 199 handle_t *ext4_journal_start_sb(struct super_block *sb, int nblocks)
 200 {
 201         journal_t *journal;
 202
 203         if (sb->s_flags & MS_RDONLY)
 204                 return ERR_PTR(-EROFS);
 205
 206         /* Special case here: if the journal has aborted behind our
 207          * backs (eg. EIO in the commit thread), then we still need to
 208          * take the FS itself readonly cleanly. */
 209         journal = EXT4_SB(sb)->s_journal;
 210         if (journal) {
 211                 if (is_journal_aborted(journal)) {
 212                         ext4_abort(sb, __func__, "Detected aborted journal");
 213                         return ERR_PTR(-EROFS);
 214                 }
 215                 return jbd2_journal_start(journal, nblocks);
 216         }
 217         /*
 218          * We're not journaling, return the appropriate indication.
 219          */
 220         current->journal_info = EXT4_NOJOURNAL_HANDLE;
 221         return current->journal_info;
 222 }
 223
 224 /*
 225  * The only special thing we need to do here is to make sure that all
 226  * jbd2_journal_stop calls result in the superblock being marked dirty, so
 227  * that sync() will call the filesystem's write_super callback if
 228  * appropriate.
 229  */
 230 int __ext4_journal_stop(const char *where, handle_t *handle)
 231 {
 232         struct super_block *sb;
 233         int err;
 234         int rc;
 235
 236         if (!ext4_handle_valid(handle)) {
 237                 /*
 238                  * Do this here since we don't call jbd2_journal_stop() in
 239                  * no-journal mode.
 240                  */
 241                 current->journal_info = NULL;
 242                 return 0;
 243         }
 244         sb = handle->h_transaction->t_journal->j_private;
 245         err = handle->h_err;
 246         rc = jbd2_journal_stop(handle);
 247
 248         if (!err)
 249                 err = rc;
 250         if (err)
 251                 __ext4_std_error(sb, where, err);
 252         return err;
 253 }
 254
 255 void ext4_journal_abort_handle(const char *caller, const char *err_fn,
 256                 struct buffer_head *bh, handle_t *handle, int err)
 257 {
 258         char nbuf[16];
 259         const char *errstr = ext4_decode_error(NULL, err, nbuf);
 260
 261         BUG_ON(!ext4_handle_valid(handle));
 262
 263         if (bh)
 264                 BUFFER_TRACE(bh, "abort");
 265
 266         if (!handle->h_err)
 267                 handle->h_err = err;
 268
 269         if (is_handle_aborted(handle))
 270                 return;
 271
 272         printk(KERN_ERR "%s: aborting transaction: %s in %s\n",
 273                caller, errstr, err_fn);
 274
 275         jbd2_journal_abort_handle(handle);
 276 }
 277
 278 /* Deal with the reporting of failure conditions on a filesystem such as
 279  * inconsistencies detected or read IO failures.
 280  *
 281  * On ext2, we can store the error state of the filesystem in the
 282  * superblock.  That is not possible on ext4, because we may have other
 283  * write ordering constraints on the superblock which prevent us from
 284  * writing it out straight away; and given that the journal is about to
 285  * be aborted, we can't rely on the current, or future, transactions to
 286  * write out the superblock safely.
 287  *
 288  * We'll just use the jbd2_journal_abort() error code to record an error in
 289  * the journal instead.  On recovery, the journal will compain about
 290  * that error until we've noted it down and cleared it.
 291  */
 292
 293 static void ext4_handle_error(struct super_block *sb)
 294 {
 295         struct ext4_super_block *es = EXT4_SB(sb)->s_es;
 296
 297         EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS;
 298         es->s_state |= cpu_to_le16(EXT4_ERROR_FS);
 299
 300         if (sb->s_flags & MS_RDONLY)
 301                 return;
 302
 303         if (!test_opt(sb, ERRORS_CONT)) {
 304                 journal_t *journal = EXT4_SB(sb)->s_journal;
 305
 306                 EXT4_SB(sb)->s_mount_flags |= EXT4_MF_FS_ABORTED;
 307                 if (journal)
 308                         jbd2_journal_abort(journal, -EIO);
 309         }
 310         if (test_opt(sb, ERRORS_RO)) {
 311                 ext4_msg(sb, KERN_CRIT, "Remounting filesystem read-only");
 312                 sb->s_flags |= MS_RDONLY;
 313         }
 314         ext4_commit_super(sb, 1);
 315         if (test_opt(sb, ERRORS_PANIC))
 316                 panic("EXT4-fs (device %s): panic forced after error\n",
 317                         sb->s_id);
 318 }
 319
 320 void ext4_error(struct super_block *sb, const char *function,
 321                 const char *fmt, ...)
 322 {
 323         va_list args;
 324
 325         va_start(args, fmt);
 326         printk(KERN_CRIT "EXT4-fs error (device %s): %s: ", sb->s_id, function);
 327         vprintk(fmt, args);
 328         printk("\n");
 329         va_end(args);
 330
 331         ext4_handle_error(sb);
 332 }
 333
 334 static const char *ext4_decode_error(struct super_block *sb, int errno,
 335                                      char nbuf[16])
 336 {
 337         char *errstr = NULL;
 338
 339         switch (errno) {
 340         case -EIO:
 341                 errstr = "IO failure";
 342                 break;
 343         case -ENOMEM:
 344                 errstr = "Out of memory";
 345                 break;
 346         case -EROFS:
 347                 if (!sb || EXT4_SB(sb)->s_journal->j_flags & JBD2_ABORT)
 348                         errstr = "Journal has aborted";
 349                 else
 350                         errstr = "Readonly filesystem";
 351                 break;
 352         default:
 353                 /* If the caller passed in an extra buffer for unknown
 354                  * errors, textualise them now.  Else we just return
 355                  * NULL. */
 356                 if (nbuf) {
 357                         /* Check for truncated error codes... */
 358                         if (snprintf(nbuf, 16, "error %d", -errno) >= 0)
 359                                 errstr = nbuf;
 360                 }
 361                 break;
 362         }
 363
 364         return errstr;
 365 }
 366
 367 /* __ext4_std_error decodes expected errors from journaling functions
 368  * automatically and invokes the appropriate error response.  */
 369
 370 void __ext4_std_error(struct super_block *sb, const char *function, int errno)
 371 {
 372         char nbuf[16];
 373         const char *errstr;
 374
 375         /* Special case: if the error is EROFS, and we're not already
 376          * inside a transaction, then there's really no point in logging
 377          * an error. */
 378         if (errno == -EROFS && journal_current_handle() == NULL &&
 379             (sb->s_flags & MS_RDONLY))
 380                 return;
 381
 382         errstr = ext4_decode_error(sb, errno, nbuf);
 383         printk(KERN_CRIT "EXT4-fs error (device %s) in %s: %s\n",
 384                sb->s_id, function, errstr);
 385
 386         ext4_handle_error(sb);
 387 }
 388
 389 /*
 390  * ext4_abort is a much stronger failure handler than ext4_error.  The
 391  * abort function may be used to deal with unrecoverable failures such
 392  * as journal IO errors or ENOMEM at a critical moment in log management.
 393  *
 394  * We unconditionally force the filesystem into an ABORT|READONLY state,
 395  * unless the error response on the fs has been set to panic in which
 396  * case we take the easy way out and panic immediately.
 397  */
 398
 399 void ext4_abort(struct super_block *sb, const char *function,
 400                 const char *fmt, ...)
 401 {
 402         va_list args;
 403
 404         va_start(args, fmt);
 405         printk(KERN_CRIT "EXT4-fs error (device %s): %s: ", sb->s_id, function);
 406         vprintk(fmt, args);
 407         printk("\n");
 408         va_end(args);
 409
 410         if (test_opt(sb, ERRORS_PANIC))
 411                 panic("EXT4-fs panic from previous error\n");
 412
 413         if (sb->s_flags & MS_RDONLY)
 414                 return;
 415
 416         ext4_msg(sb, KERN_CRIT, "Remounting filesystem read-only");
 417         EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS;
 418         sb->s_flags |= MS_RDONLY;
 419         EXT4_SB(sb)->s_mount_flags |= EXT4_MF_FS_ABORTED;
 420         if (EXT4_SB(sb)->s_journal)
 421                 jbd2_journal_abort(EXT4_SB(sb)->s_journal, -EIO);
 422 }
 423
 424 void ext4_msg (struct super_block * sb, const char *prefix,
 425                    const char *fmt, ...)
 426 {
 427         va_list args;
 428
 429         va_start(args, fmt);
 430         printk("%sEXT4-fs (%s): ", prefix, sb->s_id);
 431         vprintk(fmt, args);
 432         printk("\n");
 433         va_end(args);
 434 }
 435
 436 void ext4_warning(struct super_block *sb, const char *function,
 437                   const char *fmt, ...)
 438 {
 439         va_list args;
 440
 441         va_start(args, fmt);
 442         printk(KERN_WARNING "EXT4-fs warning (device %s): %s: ",
 443                sb->s_id, function);
 444         vprintk(fmt, args);
 445         printk("\n");
 446         va_end(args);
 447 }
 448
 449 void ext4_grp_locked_error(struct super_block *sb, ext4_group_t grp,
 450                            const char *function, const char *fmt, ...)
 451 __releases(bitlock)
 452 __acquires(bitlock)
 453 {
 454         va_list args;
 455         struct ext4_super_block *es = EXT4_SB(sb)->s_es;
 456
 457         va_start(args, fmt);
 458         printk(KERN_CRIT "EXT4-fs error (device %s): %s: ", sb->s_id, function);
 459         vprintk(fmt, args);
 460         printk("\n");
 461         va_end(args);
 462
 463         if (test_opt(sb, ERRORS_CONT)) {
 464                 EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS;
 465                 es->s_state |= cpu_to_le16(EXT4_ERROR_FS);
 466                 ext4_commit_super(sb, 0);
 467                 return;
 468         }
 469         ext4_unlock_group(sb, grp);
 470         ext4_handle_error(sb);
 471         /*
 472          * We only get here in the ERRORS_RO case; relocking the group
 473          * may be dangerous, but nothing bad will happen since the
 474          * filesystem will have already been marked read/only and the
 475          * journal has been aborted.  We return 1 as a hint to callers
 476          * who might what to use the return value from
 477          * ext4_grp_locked_error() to distinguish beween the
 478          * ERRORS_CONT and ERRORS_RO case, and perhaps return more
 479          * aggressively from the ext4 function in question, with a
 480          * more appropriate error code.
 481          */
 482         ext4_lock_group(sb, grp);
 483         return;
 484 }
 485
 486 void ext4_update_dynamic_rev(struct super_block *sb)
 487 {
 488         struct ext4_super_block *es = EXT4_SB(sb)->s_es;
 489
 490         if (le32_to_cpu(es->s_rev_level) > EXT4_GOOD_OLD_REV)
 491                 return;
 492
 493         ext4_warning(sb, __func__,
 494                      "updating to rev %d because of new feature flag, "
 495                      "running e2fsck is recommended",
 496                      EXT4_DYNAMIC_REV);
 497
 498         es->s_first_ino = cpu_to_le32(EXT4_GOOD_OLD_FIRST_INO);
 499         es->s_inode_size = cpu_to_le16(EXT4_GOOD_OLD_INODE_SIZE);
 500         es->s_rev_level = cpu_to_le32(EXT4_DYNAMIC_REV);
 501         /* leave es->s_feature_*compat flags alone */
 502         /* es->s_uuid will be set by e2fsck if empty */
 503
 504         /*
 505          * The rest of the superblock fields should be zero, and if not it
 506          * means they are likely already in use, so leave them alone.  We
 507          * can leave it up to e2fsck to clean up any inconsistencies there.
 508          */
 509 }
 510
 511 /*
 512  * Open the external journal device
 513  */
 514 static struct block_device *ext4_blkdev_get(dev_t dev, struct super_block *sb)
 515 {
 516         struct block_device *bdev;
 517         char b[BDEVNAME_SIZE];
 518
 519         bdev = open_by_devnum(dev, FMODE_READ|FMODE_WRITE);
 520         if (IS_ERR(bdev))
 521                 goto fail;
 522         return bdev;
 523
 524 fail:
 525         ext4_msg(sb, KERN_ERR, "failed to open journal device %s: %ld",
 526                         __bdevname(dev, b), PTR_ERR(bdev));
 527         return NULL;
 528 }
 529
 530 /*
 531  * Release the journal device
 532  */
 533 static int ext4_blkdev_put(struct block_device *bdev)
 534 {
 535         bd_release(bdev);
 536         return blkdev_put(bdev, FMODE_READ|FMODE_WRITE);
 537 }
 538
 539 static int ext4_blkdev_remove(struct ext4_sb_info *sbi)
 540 {
 541         struct block_device *bdev;
 542         int ret = -ENODEV;
 543
 544         bdev = sbi->journal_bdev;
 545         if (bdev) {
 546                 ret = ext4_blkdev_put(bdev);
 547                 sbi->journal_bdev = NULL;
 548         }
 549         return ret;
 550 }
 551
 552 static inline struct inode *orphan_list_entry(struct list_head *l)
 553 {
 554         return &list_entry(l, struct ext4_inode_info, i_orphan)->vfs_inode;
 555 }
 556
 557 static void dump_orphan_list(struct super_block *sb, struct ext4_sb_info *sbi)
 558 {
 559         struct list_head *l;
 560
 561         ext4_msg(sb, KERN_ERR, "sb orphan head is %d",
 562                  le32_to_cpu(sbi->s_es->s_last_orphan));
 563
 564         printk(KERN_ERR "sb_info orphan list:\n");
 565         list_for_each(l, &sbi->s_orphan) {
 566                 struct inode *inode = orphan_list_entry(l);
 567                 printk(KERN_ERR "  "
 568                        "inode %s:%lu at %p: mode %o, nlink %d, next %d\n",
 569                        inode->i_sb->s_id, inode->i_ino, inode,
 570                        inode->i_mode, inode->i_nlink,
 571                        NEXT_ORPHAN(inode));
 572         }
 573 }
 574
 575 static void ext4_put_super(struct super_block *sb)
 576 {
 577         struct ext4_sb_info *sbi = EXT4_SB(sb);
 578         struct ext4_super_block *es = sbi->s_es;
 579         int i, err;
 580
 581         lock_super(sb);
 582         lock_kernel();
 583         if (sb->s_dirt)
 584                 ext4_commit_super(sb, 1);
 585
 586         ext4_release_system_zone(sb);
 587         ext4_mb_release(sb);
 588         ext4_ext_release(sb);
 589         ext4_xattr_put_super(sb);
 590         if (sbi->s_journal) {
 591                 err = jbd2_journal_destroy(sbi->s_journal);
 592                 sbi->s_journal = NULL;
 593                 if (err < 0)
 594                         ext4_abort(sb, __func__,
 595                                    "Couldn't clean up the journal");
 596         }
 597         if (!(sb->s_flags & MS_RDONLY)) {
 598                 EXT4_CLEAR_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
 599                 es->s_state = cpu_to_le16(sbi->s_mount_state);
 600                 ext4_commit_super(sb, 1);
 601         }
 602         if (sbi->s_proc) {
 603                 remove_proc_entry(sb->s_id, ext4_proc_root);
 604         }
 605         kobject_del(&sbi->s_kobj);
 606
 607         for (i = 0; i < sbi->s_gdb_count; i++)
 608                 brelse(sbi->s_group_desc[i]);
 609         kfree(sbi->s_group_desc);
 610         if (is_vmalloc_addr(sbi->s_flex_groups))
 611                 vfree(sbi->s_flex_groups);
 612         else
 613                 kfree(sbi->s_flex_groups);
 614         percpu_counter_destroy(&sbi->s_freeblocks_counter);
 615         percpu_counter_destroy(&sbi->s_freeinodes_counter);
 616         percpu_counter_destroy(&sbi->s_dirs_counter);
 617         percpu_counter_destroy(&sbi->s_dirtyblocks_counter);
 618         brelse(sbi->s_sbh);
 619 #ifdef CONFIG_QUOTA
 620         for (i = 0; i < MAXQUOTAS; i++)
 621                 kfree(sbi->s_qf_names[i]);
 622 #endif
 623
 624         /* Debugging code just in case the in-memory inode orphan list
 625          * isn't empty.  The on-disk one can be non-empty if we've
 626          * detected an error and taken the fs readonly, but the
 627          * in-memory list had better be clean by this point. */
 628         if (!list_empty(&sbi->s_orphan))
 629                 dump_orphan_list(sb, sbi);
 630         J_ASSERT(list_empty(&sbi->s_orphan));
 631
 632         invalidate_bdev(sb->s_bdev);
 633         if (sbi->journal_bdev && sbi->journal_bdev != sb->s_bdev) {
 634                 /*
 635                  * Invalidate the journal device's buffers.  We don't want them
 636                  * floating about in memory - the physical journal device may
 637                  * hotswapped, and it breaks the `ro-after' testing code.
 638                  */
 639                 sync_blockdev(sbi->journal_bdev);
 640                 invalidate_bdev(sbi->journal_bdev);
 641                 ext4_blkdev_remove(sbi);
 642         }
 643         sb->s_fs_info = NULL;
 644         /*
 645          * Now that we are completely done shutting down the
 646          * superblock, we need to actually destroy the kobject.
 647          */
 648         unlock_kernel();
 649         unlock_super(sb);
 650         kobject_put(&sbi->s_kobj);
 651         wait_for_completion(&sbi->s_kobj_unregister);
 652         kfree(sbi->s_blockgroup_lock);
 653         kfree(sbi);
 654 }
 655
 656 static struct kmem_cache *ext4_inode_cachep;
 657
 658 /*
 659  * Called inside transaction, so use GFP_NOFS
 660  */
 661 static struct inode *ext4_alloc_inode(struct super_block *sb)
 662 {
 663         struct ext4_inode_info *ei;
 664
 665         ei = kmem_cache_alloc(ext4_inode_cachep, GFP_NOFS);
 666         if (!ei)
 667                 return NULL;
 668
 669         ei->vfs_inode.i_version = 1;
 670         ei->vfs_inode.i_data.writeback_index = 0;
 671         memset(&ei->i_cached_extent, 0, sizeof(struct ext4_ext_cache));
 672         INIT_LIST_HEAD(&ei->i_prealloc_list);
 673         spin_lock_init(&ei->i_prealloc_lock);
 674         /*
 675          * Note:  We can be called before EXT4_SB(sb)->s_journal is set,
 676          * therefore it can be null here.  Don't check it, just initialize
 677          * jinode.
 678          */
 679         jbd2_journal_init_jbd_inode(&ei->jinode, &ei->vfs_inode);
 680         ei->i_reserved_data_blocks = 0;
 681         ei->i_reserved_meta_blocks = 0;
 682         ei->i_allocated_meta_blocks = 0;
 683         ei->i_delalloc_reserved_flag = 0;
 684         spin_lock_init(&(ei->i_block_reservation_lock));
 685
 686         return &ei->vfs_inode;
 687 }
 688
 689 static void ext4_destroy_inode(struct inode *inode)
 690 {
 691         if (!list_empty(&(EXT4_I(inode)->i_orphan))) {
 692                 ext4_msg(inode->i_sb, KERN_ERR,
 693                          "Inode %lu (%p): orphan list check failed!",
 694                          inode->i_ino, EXT4_I(inode));
 695                 print_hex_dump(KERN_INFO, "", DUMP_PREFIX_ADDRESS, 16, 4,
 696                                 EXT4_I(inode), sizeof(struct ext4_inode_info),
 697                                 true);
 698                 dump_stack();
 699         }
 700         kmem_cache_free(ext4_inode_cachep, EXT4_I(inode));
 701 }
 702
 703 static void init_once(void *foo)
 704 {
 705         struct ext4_inode_info *ei = (struct ext4_inode_info *) foo;
 706
 707         INIT_LIST_HEAD(&ei->i_orphan);
 708 #ifdef CONFIG_EXT4_FS_XATTR
 709         init_rwsem(&ei->xattr_sem);
 710 #endif
 711         init_rwsem(&ei->i_data_sem);
 712         inode_init_once(&ei->vfs_inode);
 713 }
 714
 715 static int init_inodecache(void)
 716 {
 717         ext4_inode_cachep = kmem_cache_create("ext4_inode_cache",
 718                                              sizeof(struct ext4_inode_info),
 719                                              0, (SLAB_RECLAIM_ACCOUNT|
 720                                                 SLAB_MEM_SPREAD),
 721                                              init_once);
 722         if (ext4_inode_cachep == NULL)
 723                 return -ENOMEM;
 724         return 0;
 725 }
 726
 727 static void destroy_inodecache(void)
 728 {
 729         kmem_cache_destroy(ext4_inode_cachep);
 730 }
 731
 732 static void ext4_clear_inode(struct inode *inode)
 733 {
 734         ext4_discard_preallocations(inode);
 735         if (EXT4_JOURNAL(inode))
 736                 jbd2_journal_release_jbd_inode(EXT4_SB(inode->i_sb)->s_journal,
 737                                        &EXT4_I(inode)->jinode);
 738 }
 739
 740 static inline void ext4_show_quota_options(struct seq_file *seq,
 741                                            struct super_block *sb)
 742 {
 743 #if defined(CONFIG_QUOTA)
 744         struct ext4_sb_info *sbi = EXT4_SB(sb);
 745
 746         if (sbi->s_jquota_fmt)
 747                 seq_printf(seq, ",jqfmt=%s",
 748                 (sbi->s_jquota_fmt == QFMT_VFS_OLD) ? "vfsold" : "vfsv0");
 749
 750         if (sbi->s_qf_names[USRQUOTA])
 751                 seq_printf(seq, ",usrjquota=%s", sbi->s_qf_names[USRQUOTA]);
 752
 753         if (sbi->s_qf_names[GRPQUOTA])
 754                 seq_printf(seq, ",grpjquota=%s", sbi->s_qf_names[GRPQUOTA]);
 755
 756         if (sbi->s_mount_opt & EXT4_MOUNT_USRQUOTA)
 757                 seq_puts(seq, ",usrquota");
 758
 759         if (sbi->s_mount_opt & EXT4_MOUNT_GRPQUOTA)
 760                 seq_puts(seq, ",grpquota");
 761 #endif
 762 }
 763
 764 /*
 765  * Show an option if
 766  *  - it's set to a non-default value OR
 767  *  - if the per-sb default is different from the global default
 768  */
 769 static int ext4_show_options(struct seq_file *seq, struct vfsmount *vfs)
 770 {
 771         int def_errors;
 772         unsigned long def_mount_opts;
 773         struct super_block *sb = vfs->mnt_sb;
 774         struct ext4_sb_info *sbi = EXT4_SB(sb);
 775         struct ext4_super_block *es = sbi->s_es;
 776
 777         def_mount_opts = le32_to_cpu(es->s_default_mount_opts);
 778         def_errors     = le16_to_cpu(es->s_errors);
 779
 780         if (sbi->s_sb_block != 1)
 781                 seq_printf(seq, ",sb=%llu", sbi->s_sb_block);
 782         if (test_opt(sb, MINIX_DF))
 783                 seq_puts(seq, ",minixdf");
 784         if (test_opt(sb, GRPID) && !(def_mount_opts & EXT4_DEFM_BSDGROUPS))
 785                 seq_puts(seq, ",grpid");
 786         if (!test_opt(sb, GRPID) && (def_mount_opts & EXT4_DEFM_BSDGROUPS))
 787                 seq_puts(seq, ",nogrpid");
 788         if (sbi->s_resuid != EXT4_DEF_RESUID ||
 789             le16_to_cpu(es->s_def_resuid) != EXT4_DEF_RESUID) {
 790                 seq_printf(seq, ",resuid=%u", sbi->s_resuid);
 791         }
 792         if (sbi->s_resgid != EXT4_DEF_RESGID ||
 793             le16_to_cpu(es->s_def_resgid) != EXT4_DEF_RESGID) {
 794                 seq_printf(seq, ",resgid=%u", sbi->s_resgid);
 795         }
 796         if (test_opt(sb, ERRORS_RO)) {
 797                 if (def_errors == EXT4_ERRORS_PANIC ||
 798                     def_errors == EXT4_ERRORS_CONTINUE) {
 799                         seq_puts(seq, ",errors=remount-ro");
 800                 }
 801         }
 802         if (test_opt(sb, ERRORS_CONT) && def_errors != EXT4_ERRORS_CONTINUE)
 803                 seq_puts(seq, ",errors=continue");
 804         if (test_opt(sb, ERRORS_PANIC) && def_errors != EXT4_ERRORS_PANIC)
 805                 seq_puts(seq, ",errors=panic");
 806         if (test_opt(sb, NO_UID32) && !(def_mount_opts & EXT4_DEFM_UID16))
 807                 seq_puts(seq, ",nouid32");
 808         if (test_opt(sb, DEBUG) && !(def_mount_opts & EXT4_DEFM_DEBUG))
 809                 seq_puts(seq, ",debug");
 810         if (test_opt(sb, OLDALLOC))
 811                 seq_puts(seq, ",oldalloc");
 812 #ifdef CONFIG_EXT4_FS_XATTR
 813         if (test_opt(sb, XATTR_USER) &&
 814                 !(def_mount_opts & EXT4_DEFM_XATTR_USER))
 815                 seq_puts(seq, ",user_xattr");
 816         if (!test_opt(sb, XATTR_USER) &&
 817             (def_mount_opts & EXT4_DEFM_XATTR_USER)) {
 818                 seq_puts(seq, ",nouser_xattr");
 819         }
 820 #endif
 821 #ifdef CONFIG_EXT4_FS_POSIX_ACL
 822         if (test_opt(sb, POSIX_ACL) && !(def_mount_opts & EXT4_DEFM_ACL))
 823                 seq_puts(seq, ",acl");
 824         if (!test_opt(sb, POSIX_ACL) && (def_mount_opts & EXT4_DEFM_ACL))
 825                 seq_puts(seq, ",noacl");
 826 #endif
 827         if (sbi->s_commit_interval != JBD2_DEFAULT_MAX_COMMIT_AGE*HZ) {
 828                 seq_printf(seq, ",commit=%u",
 829                            (unsigned) (sbi->s_commit_interval / HZ));
 830         }
 831         if (sbi->s_min_batch_time != EXT4_DEF_MIN_BATCH_TIME) {
 832                 seq_printf(seq, ",min_batch_time=%u",
 833                            (unsigned) sbi->s_min_batch_time);
 834         }
 835         if (sbi->s_max_batch_time != EXT4_DEF_MAX_BATCH_TIME) {
 836                 seq_printf(seq, ",max_batch_time=%u",
 837                            (unsigned) sbi->s_min_batch_time);
 838         }
 839
 840         /*
 841          * We're changing the default of barrier mount option, so
 842          * let's always display its mount state so it's clear what its
 843          * status is.
 844          */
 845         seq_puts(seq, ",barrier=");
 846         seq_puts(seq, test_opt(sb, BARRIER) ? "1" : "0");
 847         if (test_opt(sb, JOURNAL_ASYNC_COMMIT))
 848                 seq_puts(seq, ",journal_async_commit");
 849         if (test_opt(sb, NOBH))
 850                 seq_puts(seq, ",nobh");
 851         if (test_opt(sb, I_VERSION))
 852                 seq_puts(seq, ",i_version");
 853         if (!test_opt(sb, DELALLOC))
 854                 seq_puts(seq, ",nodelalloc");
 855
 856
 857         if (sbi->s_stripe)
 858                 seq_printf(seq, ",stripe=%lu", sbi->s_stripe);
 859         /*
 860          * journal mode get enabled in different ways
 861          * So just print the value even if we didn't specify it
 862          */
 863         if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA)
 864                 seq_puts(seq, ",data=journal");
 865         else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA)
 866                 seq_puts(seq, ",data=ordered");
 867         else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_WRITEBACK_DATA)
 868                 seq_puts(seq, ",data=writeback");
 869
 870         if (sbi->s_inode_readahead_blks != EXT4_DEF_INODE_READAHEAD_BLKS)
 871                 seq_printf(seq, ",inode_readahead_blks=%u",
 872                            sbi->s_inode_readahead_blks);
 873
 874         if (test_opt(sb, DATA_ERR_ABORT))
 875                 seq_puts(seq, ",data_err=abort");
 876
 877         if (test_opt(sb, NO_AUTO_DA_ALLOC))
 878                 seq_puts(seq, ",noauto_da_alloc");
 879
 880         ext4_show_quota_options(seq, sb);
 881
 882         return 0;
 883 }
 884
 885 static struct inode *ext4_nfs_get_inode(struct super_block *sb,
 886                                         u64 ino, u32 generation)
 887 {
 888         struct inode *inode;
 889
 890         if (ino < EXT4_FIRST_INO(sb) && ino != EXT4_ROOT_INO)
 891                 return ERR_PTR(-ESTALE);
 892         if (ino > le32_to_cpu(EXT4_SB(sb)->s_es->s_inodes_count))
 893                 return ERR_PTR(-ESTALE);
 894
 895         /* iget isn't really right if the inode is currently unallocated!!
 896          *
 897          * ext4_read_inode will return a bad_inode if the inode had been
 898          * deleted, so we should be safe.
 899          *
 900          * Currently we don't know the generation for parent directory, so
 901          * a generation of 0 means "accept any"
 902          */
 903         inode = ext4_iget(sb, ino);
 904         if (IS_ERR(inode))
 905                 return ERR_CAST(inode);
 906         if (generation && inode->i_generation != generation) {
 907                 iput(inode);
 908                 return ERR_PTR(-ESTALE);
 909         }
 910
 911         return inode;
 912 }
 913
 914 static struct dentry *ext4_fh_to_dentry(struct super_block *sb, struct fid *fid,
 915                                         int fh_len, int fh_type)
 916 {
 917         return generic_fh_to_dentry(sb, fid, fh_len, fh_type,
 918                                     ext4_nfs_get_inode);
 919 }
 920
 921 static struct dentry *ext4_fh_to_parent(struct super_block *sb, struct fid *fid,
 922                                         int fh_len, int fh_type)
 923 {
 924         return generic_fh_to_parent(sb, fid, fh_len, fh_type,
 925                                     ext4_nfs_get_inode);
 926 }
 927
 928 /*
 929  * Try to release metadata pages (indirect blocks, directories) which are
 930  * mapped via the block device.  Since these pages could have journal heads
 931  * which would prevent try_to_free_buffers() from freeing them, we must use
 932  * jbd2 layer's try_to_free_buffers() function to release them.
 933  */
 934 static int bdev_try_to_free_page(struct super_block *sb, struct page *page,
 935                                  gfp_t wait)
 936 {
 937         journal_t *journal = EXT4_SB(sb)->s_journal;
 938
 939         WARN_ON(PageChecked(page));
 940         if (!page_has_buffers(page))
 941                 return 0;
 942         if (journal)
 943                 return jbd2_journal_try_to_free_buffers(journal, page,
 944                                                         wait & ~__GFP_WAIT);
 945         return try_to_free_buffers(page);
 946 }
 947
 948 #ifdef CONFIG_QUOTA
 949 #define QTYPE2NAME(t) ((t) == USRQUOTA ? "user" : "group")
 950 #define QTYPE2MOPT(on, t) ((t) == USRQUOTA?((on)##USRJQUOTA):((on)##GRPJQUOTA))
 951
 952 static int ext4_write_dquot(struct dquot *dquot);
 953 static int ext4_acquire_dquot(struct dquot *dquot);
 954 static int ext4_release_dquot(struct dquot *dquot);
 955 static int ext4_mark_dquot_dirty(struct dquot *dquot);
 956 static int ext4_write_info(struct super_block *sb, int type);
 957 static int ext4_quota_on(struct super_block *sb, int type, int format_id,
 958                                 char *path, int remount);
 959 static int ext4_quota_on_mount(struct super_block *sb, int type);
 960 static ssize_t ext4_quota_read(struct super_block *sb, int type, char *data,
 961                                size_t len, loff_t off);
 962 static ssize_t ext4_quota_write(struct super_block *sb, int type,
 963                                 const char *data, size_t len, loff_t off);
 964
 965 static struct dquot_operations ext4_quota_operations = {
 966         .initialize     = dquot_initialize,
 967         .drop           = dquot_drop,
 968         .alloc_space    = dquot_alloc_space,
 969         .reserve_space  = dquot_reserve_space,
 970         .claim_space    = dquot_claim_space,
 971         .release_rsv    = dquot_release_reserved_space,
 972         .get_reserved_space = ext4_get_reserved_space,
 973         .alloc_inode    = dquot_alloc_inode,
 974         .free_space     = dquot_free_space,
 975         .free_inode     = dquot_free_inode,
 976         .transfer       = dquot_transfer,
 977         .write_dquot    = ext4_write_dquot,
 978         .acquire_dquot  = ext4_acquire_dquot,
 979         .release_dquot  = ext4_release_dquot,
 980         .mark_dirty     = ext4_mark_dquot_dirty,
 981         .write_info     = ext4_write_info,
 982         .alloc_dquot    = dquot_alloc,
 983         .destroy_dquot  = dquot_destroy,
 984 };
 985
 986 static struct quotactl_ops ext4_qctl_operations = {
 987         .quota_on       = ext4_quota_on,
 988         .quota_off      = vfs_quota_off,
 989         .quota_sync     = vfs_quota_sync,
 990         .get_info       = vfs_get_dqinfo,
 991         .set_info       = vfs_set_dqinfo,
 992         .get_dqblk      = vfs_get_dqblk,
 993         .set_dqblk      = vfs_set_dqblk
 994 };
 995 #endif
 996
 997 static const struct super_operations ext4_sops = {
 998         .alloc_inode    = ext4_alloc_inode,
 999         .destroy_inode  = ext4_destroy_inode,
1000         .write_inode    = ext4_write_inode,
1001         .dirty_inode    = ext4_dirty_inode,
1002         .delete_inode   = ext4_delete_inode,
1003         .put_super      = ext4_put_super,
1004         .sync_fs        = ext4_sync_fs,
1005         .freeze_fs      = ext4_freeze,
1006         .unfreeze_fs    = ext4_unfreeze,
1007         .statfs         = ext4_statfs,
1008         .remount_fs     = ext4_remount,
1009         .clear_inode    = ext4_clear_inode,
1010         .show_options   = ext4_show_options,
1011 #ifdef CONFIG_QUOTA
1012         .quota_read     = ext4_quota_read,
1013         .quota_write    = ext4_quota_write,
1014 #endif
1015         .bdev_try_to_free_page = bdev_try_to_free_page,
1016 };
1017
1018 static const struct super_operations ext4_nojournal_sops = {
1019         .alloc_inode    = ext4_alloc_inode,
1020         .destroy_inode  = ext4_destroy_inode,
1021         .write_inode    = ext4_write_inode,
1022         .dirty_inode    = ext4_dirty_inode,
1023         .delete_inode   = ext4_delete_inode,
1024         .write_super    = ext4_write_super,
1025         .put_super      = ext4_put_super,
1026         .statfs         = ext4_statfs,
1027         .remount_fs     = ext4_remount,
1028         .clear_inode    = ext4_clear_inode,
1029         .show_options   = ext4_show_options,
1030 #ifdef CONFIG_QUOTA
1031         .quota_read     = ext4_quota_read,
1032         .quota_write    = ext4_quota_write,
1033 #endif
1034         .bdev_try_to_free_page = bdev_try_to_free_page,
1035 };
1036
1037 static const struct export_operations ext4_export_ops = {
1038         .fh_to_dentry = ext4_fh_to_dentry,
1039         .fh_to_parent = ext4_fh_to_parent,
1040         .get_parent = ext4_get_parent,
1041 };
1042
1043 enum {
1044         Opt_bsd_df, Opt_minix_df, Opt_grpid, Opt_nogrpid,
1045         Opt_resgid, Opt_resuid, Opt_sb, Opt_err_cont, Opt_err_panic, Opt_err_ro,
1046         Opt_nouid32, Opt_debug, Opt_oldalloc, Opt_orlov,
1047         Opt_user_xattr, Opt_nouser_xattr, Opt_acl, Opt_noacl,
1048         Opt_auto_da_alloc, Opt_noauto_da_alloc, Opt_noload, Opt_nobh, Opt_bh,
1049         Opt_commit, Opt_min_batch_time, Opt_max_batch_time,
1050         Opt_journal_update, Opt_journal_dev,
1051         Opt_journal_checksum, Opt_journal_async_commit,
1052         Opt_abort, Opt_data_journal, Opt_data_ordered, Opt_data_writeback,
1053         Opt_data_err_abort, Opt_data_err_ignore, Opt_mb_history_length,
1054         Opt_usrjquota, Opt_grpjquota, Opt_offusrjquota, Opt_offgrpjquota,
1055         Opt_jqfmt_vfsold, Opt_jqfmt_vfsv0, Opt_quota, Opt_noquota,
1056         Opt_ignore, Opt_barrier, Opt_nobarrier, Opt_err, Opt_resize,
1057         Opt_usrquota, Opt_grpquota, Opt_i_version,
1058         Opt_stripe, Opt_delalloc, Opt_nodelalloc,
1059         Opt_block_validity, Opt_noblock_validity,
1060         Opt_inode_readahead_blks, Opt_journal_ioprio
1061 };
1062
1063 static const match_table_t tokens = {
1064         {Opt_bsd_df, "bsddf"},
1065         {Opt_minix_df, "minixdf"},
1066         {Opt_grpid, "grpid"},
1067         {Opt_grpid, "bsdgroups"},
1068         {Opt_nogrpid, "nogrpid"},
1069         {Opt_nogrpid, "sysvgroups"},
1070         {Opt_resgid, "resgid=%u"},
1071         {Opt_resuid, "resuid=%u"},
1072         {Opt_sb, "sb=%u"},
1073         {Opt_err_cont, "errors=continue"},
1074         {Opt_err_panic, "errors=panic"},
1075         {Opt_err_ro, "errors=remount-ro"},
1076         {Opt_nouid32, "nouid32"},
1077         {Opt_debug, "debug"},
1078         {Opt_oldalloc, "oldalloc"},
1079         {Opt_orlov, "orlov"},
1080         {Opt_user_xattr, "user_xattr"},
1081         {Opt_nouser_xattr, "nouser_xattr"},
1082         {Opt_acl, "acl"},
1083         {Opt_noacl, "noacl"},
1084         {Opt_noload, "noload"},
1085         {Opt_nobh, "nobh"},
1086         {Opt_bh, "bh"},
1087         {Opt_commit, "commit=%u"},
1088         {Opt_min_batch_time, "min_batch_time=%u"},
1089         {Opt_max_batch_time, "max_batch_time=%u"},
1090         {Opt_journal_update, "journal=update"},
1091         {Opt_journal_dev, "journal_dev=%u"},
1092         {Opt_journal_checksum, "journal_checksum"},
1093         {Opt_journal_async_commit, "journal_async_commit"},
1094         {Opt_abort, "abort"},
1095         {Opt_data_journal, "data=journal"},
1096         {Opt_data_ordered, "data=ordered"},
1097         {Opt_data_writeback, "data=writeback"},
1098         {Opt_data_err_abort, "data_err=abort"},
1099         {Opt_data_err_ignore, "data_err=ignore"},
1100         {Opt_mb_history_length, "mb_history_length=%u"},
1101         {Opt_offusrjquota, "usrjquota="},
1102         {Opt_usrjquota, "usrjquota=%s"},
1103         {Opt_offgrpjquota, "grpjquota="},
1104         {Opt_grpjquota, "grpjquota=%s"},
1105         {Opt_jqfmt_vfsold, "jqfmt=vfsold"},
1106         {Opt_jqfmt_vfsv0, "jqfmt=vfsv0"},
1107         {Opt_grpquota, "grpquota"},
1108         {Opt_noquota, "noquota"},
1109         {Opt_quota, "quota"},
1110         {Opt_usrquota, "usrquota"},
1111         {Opt_barrier, "barrier=%u"},
1112         {Opt_barrier, "barrier"},
1113         {Opt_nobarrier, "nobarrier"},
1114         {Opt_i_version, "i_version"},
1115         {Opt_stripe, "stripe=%u"},
1116         {Opt_resize, "resize"},
1117         {Opt_delalloc, "delalloc"},
1118         {Opt_nodelalloc, "nodelalloc"},
1119         {Opt_block_validity, "block_validity"},
1120         {Opt_noblock_validity, "noblock_validity"},
1121         {Opt_inode_readahead_blks, "inode_readahead_blks=%u"},
1122         {Opt_journal_ioprio, "journal_ioprio=%u"},
1123         {Opt_auto_da_alloc, "auto_da_alloc=%u"},
1124         {Opt_auto_da_alloc, "auto_da_alloc"},
1125         {Opt_noauto_da_alloc, "noauto_da_alloc"},
1126         {Opt_err, NULL},
1127 };
1128
1129 static ext4_fsblk_t get_sb_block(void **data)
1130 {
1131         ext4_fsblk_t    sb_block;
1132         char            *options = (char *) *data;
1133
1134         if (!options || strncmp(options, "sb=", 3) != 0)
1135                 return 1;       /* Default location */
1136
1137         options += 3;
1138         /* TODO: use simple_strtoll with >32bit ext4 */
1139         sb_block = simple_strtoul(options, &options, 0);
1140         if (*options && *options != ',') {
1141                 printk(KERN_ERR "EXT4-fs: Invalid sb specification: %s\n",
1142                        (char *) *data);
1143                 return 1;
1144         }
1145         if (*options == ',')
1146                 options++;
1147         *data = (void *) options;
1148
1149         return sb_block;
1150 }
1151
1152 #define DEFAULT_JOURNAL_IOPRIO (IOPRIO_PRIO_VALUE(IOPRIO_CLASS_BE, 3))
1153
1154 static int parse_options(char *options, struct super_block *sb,
1155                          unsigned long *journal_devnum,
1156                          unsigned int *journal_ioprio,
1157                          ext4_fsblk_t *n_blocks_count, int is_remount)
1158 {
1159         struct ext4_sb_info *sbi = EXT4_SB(sb);
1160         char *p;
1161         substring_t args[MAX_OPT_ARGS];
1162         int data_opt = 0;
1163         int option;
1164 #ifdef CONFIG_QUOTA
1165         int qtype, qfmt;
1166         char *qname;
1167 #endif
1168
1169         if (!options)
1170                 return 1;
1171
1172         while ((p = strsep(&options, ",")) != NULL) {
1173                 int token;
1174                 if (!*p)
1175                         continue;
1176
1177                 token = match_token(p, tokens, args);
1178                 switch (token) {
1179                 case Opt_bsd_df:
1180                         clear_opt(sbi->s_mount_opt, MINIX_DF);
1181                         break;
1182                 case Opt_minix_df:
1183                         set_opt(sbi->s_mount_opt, MINIX_DF);
1184                         break;
1185                 case Opt_grpid:
1186                         set_opt(sbi->s_mount_opt, GRPID);
1187                         break;
1188                 case Opt_nogrpid:
1189                         clear_opt(sbi->s_mount_opt, GRPID);
1190                         break;
1191                 case Opt_resuid:
1192                         if (match_int(&args[0], &option))
1193                                 return 0;
1194                         sbi->s_resuid = option;
1195                         break;
1196                 case Opt_resgid:
1197                         if (match_int(&args[0], &option))
1198                                 return 0;
1199                         sbi->s_resgid = option;
1200                         break;
1201                 case Opt_sb:
1202                         /* handled by get_sb_block() instead of here */
1203                         /* *sb_block = match_int(&args[0]); */
1204                         break;
1205                 case Opt_err_panic:
1206                         clear_opt(sbi->s_mount_opt, ERRORS_CONT);
1207                         clear_opt(sbi->s_mount_opt, ERRORS_RO);
1208                         set_opt(sbi->s_mount_opt, ERRORS_PANIC);
1209                         break;
1210                 case Opt_err_ro:
1211                         clear_opt(sbi->s_mount_opt, ERRORS_CONT);
1212                         clear_opt(sbi->s_mount_opt, ERRORS_PANIC);
1213                         set_opt(sbi->s_mount_opt, ERRORS_RO);
1214                         break;
1215                 case Opt_err_cont:
1216                         clear_opt(sbi->s_mount_opt, ERRORS_RO);
1217                         clear_opt(sbi->s_mount_opt, ERRORS_PANIC);
1218                         set_opt(sbi->s_mount_opt, ERRORS_CONT);
1219                         break;
1220                 case Opt_nouid32:
1221                         set_opt(sbi->s_mount_opt, NO_UID32);
1222                         break;
1223                 case Opt_debug:
1224                         set_opt(sbi->s_mount_opt, DEBUG);
1225                         break;
1226                 case Opt_oldalloc:
1227                         set_opt(sbi->s_mount_opt, OLDALLOC);
1228                         break;
1229                 case Opt_orlov:
1230                         clear_opt(sbi->s_mount_opt, OLDALLOC);
1231                         break;
1232 #ifdef CONFIG_EXT4_FS_XATTR
1233                 case Opt_user_xattr:
1234                         set_opt(sbi->s_mount_opt, XATTR_USER);
1235                         break;
1236                 case Opt_nouser_xattr:
1237                         clear_opt(sbi->s_mount_opt, XATTR_USER);
1238                         break;
1239 #else
1240                 case Opt_user_xattr:
1241                 case Opt_nouser_xattr:
1242                         ext4_msg(sb, KERN_ERR, "(no)user_xattr options not supported");
1243                         break;
1244 #endif
1245 #ifdef CONFIG_EXT4_FS_POSIX_ACL
1246                 case Opt_acl:
1247                         set_opt(sbi->s_mount_opt, POSIX_ACL);
1248                         break;
1249                 case Opt_noacl:
1250                         clear_opt(sbi->s_mount_opt, POSIX_ACL);
1251                         break;
1252 #else
1253                 case Opt_acl:
1254                 case Opt_noacl:
1255                         ext4_msg(sb, KERN_ERR, "(no)acl options not supported");
1256                         break;
1257 #endif
1258                 case Opt_journal_update:
1259                         /* @@@ FIXME */
1260                         /* Eventually we will want to be able to create
1261                            a journal file here.  For now, only allow the
1262                            user to specify an existing inode to be the
1263                            journal file. */
1264                         if (is_remount) {
1265                                 ext4_msg(sb, KERN_ERR,
1266                                          "Cannot specify journal on remount");
1267                                 return 0;
1268                         }
1269                         set_opt(sbi->s_mount_opt, UPDATE_JOURNAL);
1270                         break;
1271                 case Opt_journal_dev:
1272                         if (is_remount) {
1273                                 ext4_msg(sb, KERN_ERR,
1274                                         "Cannot specify journal on remount");
1275                                 return 0;
1276                         }
1277                         if (match_int(&args[0], &option))
1278                                 return 0;
1279                         *journal_devnum = option;
1280                         break;
1281                 case Opt_journal_checksum:
1282                         set_opt(sbi->s_mount_opt, JOURNAL_CHECKSUM);
1283                         break;
1284                 case Opt_journal_async_commit:
1285                         set_opt(sbi->s_mount_opt, JOURNAL_ASYNC_COMMIT);
1286                         set_opt(sbi->s_mount_opt, JOURNAL_CHECKSUM);
1287                         break;
1288                 case Opt_noload:
1289                         set_opt(sbi->s_mount_opt, NOLOAD);
1290                         break;
1291                 case Opt_commit:
1292                         if (match_int(&args[0], &option))
1293                                 return 0;
1294                         if (option < 0)
1295                                 return 0;
1296                         if (option == 0)
1297                                 option = JBD2_DEFAULT_MAX_COMMIT_AGE;
1298                         sbi->s_commit_interval = HZ * option;
1299                         break;
1300                 case Opt_max_batch_time:
1301                         if (match_int(&args[0], &option))
1302                                 return 0;
1303                         if (option < 0)
1304                                 return 0;
1305                         if (option == 0)
1306                                 option = EXT4_DEF_MAX_BATCH_TIME;
1307                         sbi->s_max_batch_time = option;
1308                         break;
1309                 case Opt_min_batch_time:
1310                         if (match_int(&args[0], &option))
1311                                 return 0;
1312                         if (option < 0)
1313                                 return 0;
1314                         sbi->s_min_batch_time = option;
1315                         break;
1316                 case Opt_data_journal:
1317                         data_opt = EXT4_MOUNT_JOURNAL_DATA;
1318                         goto datacheck;
1319                 case Opt_data_ordered:
1320                         data_opt = EXT4_MOUNT_ORDERED_DATA;
1321                         goto datacheck;
1322                 case Opt_data_writeback:
1323                         data_opt = EXT4_MOUNT_WRITEBACK_DATA;
1324                 datacheck:
1325                         if (is_remount) {
1326                                 if ((sbi->s_mount_opt & EXT4_MOUNT_DATA_FLAGS)
1327                                                 != data_opt) {
1328                                         ext4_msg(sb, KERN_ERR,
1329                                                 "Cannot change data mode on remount");
1330                                         return 0;
1331                                 }
1332                         } else {
1333                                 sbi->s_mount_opt &= ~EXT4_MOUNT_DATA_FLAGS;
1334                                 sbi->s_mount_opt |= data_opt;
1335                         }
1336                         break;
1337                 case Opt_data_err_abort:
1338                         set_opt(sbi->s_mount_opt, DATA_ERR_ABORT);
1339                         break;
1340                 case Opt_data_err_ignore:
1341                         clear_opt(sbi->s_mount_opt, DATA_ERR_ABORT);
1342                         break;
1343                 case Opt_mb_history_length:
1344                         if (match_int(&args[0], &option))
1345                                 return 0;
1346                         if (option < 0)
1347                                 return 0;
1348                         sbi->s_mb_history_max = option;
1349                         break;
1350 #ifdef CONFIG_QUOTA
1351                 case Opt_usrjquota:
1352                         qtype = USRQUOTA;
1353                         goto set_qf_name;
1354                 case Opt_grpjquota:
1355                         qtype = GRPQUOTA;
1356 set_qf_name:
1357                         if (sb_any_quota_loaded(sb) &&
1358                             !sbi->s_qf_names[qtype]) {
1359                                 ext4_msg(sb, KERN_ERR,
1360                                        "Cannot change journaled "
1361                                        "quota options when quota turned on");
1362                                 return 0;
1363                         }
1364                         qname = match_strdup(&args[0]);
1365                         if (!qname) {
1366                                 ext4_msg(sb, KERN_ERR,
1367                                         "Not enough memory for "
1368                                         "storing quotafile name");
1369                                 return 0;
1370                         }
1371                         if (sbi->s_qf_names[qtype] &&
1372                             strcmp(sbi->s_qf_names[qtype], qname)) {
1373                                 ext4_msg(sb, KERN_ERR,
1374                                         "%s quota file already "
1375                                         "specified", QTYPE2NAME(qtype));
1376                                 kfree(qname);
1377                                 return 0;
1378                         }
1379                         sbi->s_qf_names[qtype] = qname;
1380                         if (strchr(sbi->s_qf_names[qtype], '/')) {
1381                                 ext4_msg(sb, KERN_ERR,
1382                                         "quotafile must be on "
1383                                         "filesystem root");
1384                                 kfree(sbi->s_qf_names[qtype]);
1385                                 sbi->s_qf_names[qtype] = NULL;
1386                                 return 0;
1387                         }
1388                         set_opt(sbi->s_mount_opt, QUOTA);
1389                         break;
1390                 case Opt_offusrjquota:
1391                         qtype = USRQUOTA;
1392                         goto clear_qf_name;
1393                 case Opt_offgrpjquota:
1394                         qtype = GRPQUOTA;
1395 clear_qf_name:
1396                         if (sb_any_quota_loaded(sb) &&
1397                             sbi->s_qf_names[qtype]) {
1398                                 ext4_msg(sb, KERN_ERR, "Cannot change "
1399                                         "journaled quota options when "
1400                                         "quota turned on");
1401                                 return 0;
1402                         }
1403                         /*
1404                          * The space will be released later when all options
1405                          * are confirmed to be correct
1406                          */
1407                         sbi->s_qf_names[qtype] = NULL;
1408                         break;
1409                 case Opt_jqfmt_vfsold:
1410                         qfmt = QFMT_VFS_OLD;
1411                         goto set_qf_format;
1412                 case Opt_jqfmt_vfsv0:
1413                         qfmt = QFMT_VFS_V0;
1414 set_qf_format:
1415                         if (sb_any_quota_loaded(sb) &&
1416                             sbi->s_jquota_fmt != qfmt) {
1417                                 ext4_msg(sb, KERN_ERR, "Cannot change "
1418                                         "journaled quota options when "
1419                                         "quota turned on");
1420                                 return 0;
1421                         }
1422                         sbi->s_jquota_fmt = qfmt;
1423                         break;
1424                 case Opt_quota:
1425                 case Opt_usrquota:
1426                         set_opt(sbi->s_mount_opt, QUOTA);
1427                         set_opt(sbi->s_mount_opt, USRQUOTA);
1428                         break;
1429                 case Opt_grpquota:
1430                         set_opt(sbi->s_mount_opt, QUOTA);
1431                         set_opt(sbi->s_mount_opt, GRPQUOTA);
1432                         break;
1433                 case Opt_noquota:
1434                         if (sb_any_quota_loaded(sb)) {
1435                                 ext4_msg(sb, KERN_ERR, "Cannot change quota "
1436                                         "options when quota turned on");
1437                                 return 0;
1438                         }
1439                         clear_opt(sbi->s_mount_opt, QUOTA);
1440                         clear_opt(sbi->s_mount_opt, USRQUOTA);
1441                         clear_opt(sbi->s_mount_opt, GRPQUOTA);
1442                         break;
1443 #else
1444                 case Opt_quota:
1445                 case Opt_usrquota:
1446                 case Opt_grpquota:
1447                         ext4_msg(sb, KERN_ERR,
1448                                 "quota options not supported");
1449                         break;
1450                 case Opt_usrjquota:
1451                 case Opt_grpjquota:
1452                 case Opt_offusrjquota:
1453                 case Opt_offgrpjquota:
1454                 case Opt_jqfmt_vfsold:
1455                 case Opt_jqfmt_vfsv0:
1456                         ext4_msg(sb, KERN_ERR,
1457                                 "journaled quota options not supported");
1458                         break;
1459                 case Opt_noquota:
1460                         break;
1461 #endif
1462                 case Opt_abort:
1463                         sbi->s_mount_flags |= EXT4_MF_FS_ABORTED;
1464                         break;
1465                 case Opt_nobarrier:
1466                         clear_opt(sbi->s_mount_opt, BARRIER);
1467                         break;
1468                 case Opt_barrier:
1469                         if (match_int(&args[0], &option)) {
1470                                 set_opt(sbi->s_mount_opt, BARRIER);
1471                                 break;
1472                         }
1473                         if (option)
1474                                 set_opt(sbi->s_mount_opt, BARRIER);
1475                         else
1476                                 clear_opt(sbi->s_mount_opt, BARRIER);
1477                         break;
1478                 case Opt_ignore:
1479                         break;
1480                 case Opt_resize:
1481                         if (!is_remount) {
1482                                 ext4_msg(sb, KERN_ERR,
1483                                         "resize option only available "
1484                                         "for remount");
1485                                 return 0;
1486                         }
1487                         if (match_int(&args[0], &option) != 0)
1488                                 return 0;
1489                         *n_blocks_count = option;
1490                         break;
1491                 case Opt_nobh:
1492                         set_opt(sbi->s_mount_opt, NOBH);
1493                         break;
1494                 case Opt_bh:
1495                         clear_opt(sbi->s_mount_opt, NOBH);
1496                         break;
1497                 case Opt_i_version:
1498                         set_opt(sbi->s_mount_opt, I_VERSION);
1499                         sb->s_flags |= MS_I_VERSION;
1500                         break;
1501                 case Opt_nodelalloc:
1502                         clear_opt(sbi->s_mount_opt, DELALLOC);
1503                         break;
1504                 case Opt_stripe:
1505                         if (match_int(&args[0], &option))
1506                                 return 0;
1507                         if (option < 0)
1508                                 return 0;
1509                         sbi->s_stripe = option;
1510                         break;
1511                 case Opt_delalloc:
1512                         set_opt(sbi->s_mount_opt, DELALLOC);
1513                         break;
1514                 case Opt_block_validity:
1515                         set_opt(sbi->s_mount_opt, BLOCK_VALIDITY);
1516                         break;
1517                 case Opt_noblock_validity:
1518                         clear_opt(sbi->s_mount_opt, BLOCK_VALIDITY);
1519                         break;
1520                 case Opt_inode_readahead_blks:
1521                         if (match_int(&args[0], &option))
1522                                 return 0;
1523                         if (option < 0 || option > (1 << 30))
1524                                 return 0;
1525                         if (!is_power_of_2(option)) {
1526                                 ext4_msg(sb, KERN_ERR,
1527                                          "EXT4-fs: inode_readahead_blks"
1528                                          " must be a power of 2");
1529                                 return 0;
1530                         }
1531                         sbi->s_inode_readahead_blks = option;
1532                         break;
1533                 case Opt_journal_ioprio:
1534                         if (match_int(&args[0], &option))
1535                                 return 0;
1536                         if (option < 0 || option > 7)
1537                                 break;
1538                         *journal_ioprio = IOPRIO_PRIO_VALUE(IOPRIO_CLASS_BE,
1539                                                             option);
1540                         break;
1541                 case Opt_noauto_da_alloc:
1542                         set_opt(sbi->s_mount_opt,NO_AUTO_DA_ALLOC);
1543                         break;
1544                 case Opt_auto_da_alloc:
1545                         if (match_int(&args[0], &option)) {
1546                                 clear_opt(sbi->s_mount_opt, NO_AUTO_DA_ALLOC);
1547                                 break;
1548                         }
1549                         if (option)
1550                                 clear_opt(sbi->s_mount_opt, NO_AUTO_DA_ALLOC);
1551                         else
1552                                 set_opt(sbi->s_mount_opt,NO_AUTO_DA_ALLOC);
1553                         break;
1554                 default:
1555                         ext4_msg(sb, KERN_ERR,
1556                                "Unrecognized mount option \"%s\" "
1557                                "or missing value", p);
1558                         return 0;
1559                 }
1560         }
1561 #ifdef CONFIG_QUOTA
1562         if (sbi->s_qf_names[USRQUOTA] || sbi->s_qf_names[GRPQUOTA]) {
1563                 if ((sbi->s_mount_opt & EXT4_MOUNT_USRQUOTA) &&
1564                      sbi->s_qf_names[USRQUOTA])
1565                         clear_opt(sbi->s_mount_opt, USRQUOTA);
1566
1567                 if ((sbi->s_mount_opt & EXT4_MOUNT_GRPQUOTA) &&
1568                      sbi->s_qf_names[GRPQUOTA])
1569                         clear_opt(sbi->s_mount_opt, GRPQUOTA);
1570
1571                 if ((sbi->s_qf_names[USRQUOTA] &&
1572                                 (sbi->s_mount_opt & EXT4_MOUNT_GRPQUOTA)) ||
1573                     (sbi->s_qf_names[GRPQUOTA] &&
1574                                 (sbi->s_mount_opt & EXT4_MOUNT_USRQUOTA))) {
1575                         ext4_msg(sb, KERN_ERR, "old and new quota "
1576                                         "format mixing");
1577                         return 0;
1578                 }
1579
1580                 if (!sbi->s_jquota_fmt) {
1581                         ext4_msg(sb, KERN_ERR, "journaled quota format "
1582                                         "not specified");
1583                         return 0;
1584                 }
1585         } else {
1586                 if (sbi->s_jquota_fmt) {
1587                         ext4_msg(sb, KERN_ERR, "journaled quota format "
1588                                         "specified with no journaling "
1589                                         "enabled");
1590                         return 0;
1591                 }
1592         }
1593 #endif
1594         return 1;
1595 }
1596
1597 static int ext4_setup_super(struct super_block *sb, struct ext4_super_block *es,
1598                             int read_only)
1599 {
1600         struct ext4_sb_info *sbi = EXT4_SB(sb);
1601         int res = 0;
1602
1603         if (le32_to_cpu(es->s_rev_level) > EXT4_MAX_SUPP_REV) {
1604                 ext4_msg(sb, KERN_ERR, "revision level too high, "
1605                          "forcing read-only mode");
1606                 res = MS_RDONLY;
1607         }
1608         if (read_only)
1609                 return res;
1610         if (!(sbi->s_mount_state & EXT4_VALID_FS))
1611                 ext4_msg(sb, KERN_WARNING, "warning: mounting unchecked fs, "
1612                          "running e2fsck is recommended");
1613         else if ((sbi->s_mount_state & EXT4_ERROR_FS))
1614                 ext4_msg(sb, KERN_WARNING,
1615                          "warning: mounting fs with errors, "
1616                          "running e2fsck is recommended");
1617         else if ((__s16) le16_to_cpu(es->s_max_mnt_count) >= 0 &&
1618                  le16_to_cpu(es->s_mnt_count) >=
1619                  (unsigned short) (__s16) le16_to_cpu(es->s_max_mnt_count))
1620                 ext4_msg(sb, KERN_WARNING,
1621                          "warning: maximal mount count reached, "
1622                          "running e2fsck is recommended");
1623         else if (le32_to_cpu(es->s_checkinterval) &&
1624                 (le32_to_cpu(es->s_lastcheck) +
1625                         le32_to_cpu(es->s_checkinterval) <= get_seconds()))
1626                 ext4_msg(sb, KERN_WARNING,
1627                          "warning: checktime reached, "
1628                          "running e2fsck is recommended");
1629         if (!sbi->s_journal)
1630                 es->s_state &= cpu_to_le16(~EXT4_VALID_FS);
1631         if (!(__s16) le16_to_cpu(es->s_max_mnt_count))
1632                 es->s_max_mnt_count = cpu_to_le16(EXT4_DFL_MAX_MNT_COUNT);
1633         le16_add_cpu(&es->s_mnt_count, 1);
1634         es->s_mtime = cpu_to_le32(get_seconds());
1635         ext4_update_dynamic_rev(sb);
1636         if (sbi->s_journal)
1637                 EXT4_SET_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
1638
1639         ext4_commit_super(sb, 1);
1640         if (test_opt(sb, DEBUG))
1641                 printk(KERN_INFO "[EXT4 FS bs=%lu, gc=%u, "
1642                                 "bpg=%lu, ipg=%lu, mo=%04x]\n",
1643                         sb->s_blocksize,
1644                         sbi->s_groups_count,
1645                         EXT4_BLOCKS_PER_GROUP(sb),
1646                         EXT4_INODES_PER_GROUP(sb),
1647                         sbi->s_mount_opt);
1648
1649         if (EXT4_SB(sb)->s_journal) {
1650                 ext4_msg(sb, KERN_INFO, "%s journal on %s",
1651                        EXT4_SB(sb)->s_journal->j_inode ? "internal" :
1652                        "external", EXT4_SB(sb)->s_journal->j_devname);
1653         } else {
1654                 ext4_msg(sb, KERN_INFO, "no journal");
1655         }
1656         return res;
1657 }
1658
1659 static int ext4_fill_flex_info(struct super_block *sb)
1660 {
1661         struct ext4_sb_info *sbi = EXT4_SB(sb);
1662         struct ext4_group_desc *gdp = NULL;
1663         ext4_group_t flex_group_count;
1664         ext4_group_t flex_group;
1665         int groups_per_flex = 0;
1666         size_t size;
1667         int i;
1668
1669         if (!sbi->s_es->s_log_groups_per_flex) {
1670                 sbi->s_log_groups_per_flex = 0;
1671                 return 1;
1672         }
1673
1674         sbi->s_log_groups_per_flex = sbi->s_es->s_log_groups_per_flex;
1675         groups_per_flex = 1 << sbi->s_log_groups_per_flex;
1676
1677         /* We allocate both existing and potentially added groups */
1678         flex_group_count = ((sbi->s_groups_count + groups_per_flex - 1) +
1679                         ((le16_to_cpu(sbi->s_es->s_reserved_gdt_blocks) + 1) <<
1680                               EXT4_DESC_PER_BLOCK_BITS(sb))) / groups_per_flex;
1681         size = flex_group_count * sizeof(struct flex_groups);
1682         sbi->s_flex_groups = kzalloc(size, GFP_KERNEL);
1683         if (sbi->s_flex_groups == NULL) {
1684                 sbi->s_flex_groups = vmalloc(size);
1685                 if (sbi->s_flex_groups)
1686                         memset(sbi->s_flex_groups, 0, size);
1687         }
1688         if (sbi->s_flex_groups == NULL) {
1689                 ext4_msg(sb, KERN_ERR, "not enough memory for "
1690                                 "%u flex groups", flex_group_count);
1691                 goto failed;
1692         }
1693
1694         for (i = 0; i < sbi->s_groups_count; i++) {
1695                 gdp = ext4_get_group_desc(sb, i, NULL);
1696
1697                 flex_group = ext4_flex_group(sbi, i);
1698                 atomic_set(&sbi->s_flex_groups[flex_group].free_inodes,
1699                            ext4_free_inodes_count(sb, gdp));
1700                 atomic_set(&sbi->s_flex_groups[flex_group].free_blocks,
1701                            ext4_free_blks_count(sb, gdp));
1702                 atomic_set(&sbi->s_flex_groups[flex_group].used_dirs,
1703                            ext4_used_dirs_count(sb, gdp));
1704         }
1705
1706         return 1;
1707 failed:
1708         return 0;
1709 }
1710
1711 __le16 ext4_group_desc_csum(struct ext4_sb_info *sbi, __u32 block_group,
1712                             struct ext4_group_desc *gdp)
1713 {
1714         __u16 crc = 0;
1715
1716         if (sbi->s_es->s_feature_ro_compat &
1717             cpu_to_le32(EXT4_FEATURE_RO_COMPAT_GDT_CSUM)) {
1718                 int offset = offsetof(struct ext4_group_desc, bg_checksum);
1719                 __le32 le_group = cpu_to_le32(block_group);
1720
1721                 crc = crc16(~0, sbi->s_es->s_uuid, sizeof(sbi->s_es->s_uuid));
1722                 crc = crc16(crc, (__u8 *)&le_group, sizeof(le_group));
1723                 crc = crc16(crc, (__u8 *)gdp, offset);
1724                 offset += sizeof(gdp->bg_checksum); /* skip checksum */
1725                 /* for checksum of struct ext4_group_desc do the rest...*/
1726                 if ((sbi->s_es->s_feature_incompat &
1727                      cpu_to_le32(EXT4_FEATURE_INCOMPAT_64BIT)) &&
1728                     offset < le16_to_cpu(sbi->s_es->s_desc_size))
1729                         crc = crc16(crc, (__u8 *)gdp + offset,
1730                                     le16_to_cpu(sbi->s_es->s_desc_size) -
1731                                         offset);
1732         }
1733
1734         return cpu_to_le16(crc);
1735 }
1736
1737 int ext4_group_desc_csum_verify(struct ext4_sb_info *sbi, __u32 block_group,
1738                                 struct ext4_group_desc *gdp)
1739 {
1740         if ((sbi->s_es->s_feature_ro_compat &
1741              cpu_to_le32(EXT4_FEATURE_RO_COMPAT_GDT_CSUM)) &&
1742             (gdp->bg_checksum != ext4_group_desc_csum(sbi, block_group, gdp)))
1743                 return 0;
1744
1745         return 1;
1746 }
1747
1748 /* Called at mount-time, super-block is locked */
1749 static int ext4_check_descriptors(struct super_block *sb)
1750 {
1751         struct ext4_sb_info *sbi = EXT4_SB(sb);
1752         ext4_fsblk_t first_block = le32_to_cpu(sbi->s_es->s_first_data_block);
1753         ext4_fsblk_t last_block;
1754         ext4_fsblk_t block_bitmap;
1755         ext4_fsblk_t inode_bitmap;
1756         ext4_fsblk_t inode_table;
1757         int flexbg_flag = 0;
1758         ext4_group_t i;
1759
1760         if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_FLEX_BG))
1761                 flexbg_flag = 1;
1762
1763         ext4_debug("Checking group descriptors");
1764
1765         for (i = 0; i < sbi->s_groups_count; i++) {
1766                 struct ext4_group_desc *gdp = ext4_get_group_desc(sb, i, NULL);
1767
1768                 if (i == sbi->s_groups_count - 1 || flexbg_flag)
1769                         last_block = ext4_blocks_count(sbi->s_es) - 1;
1770                 else
1771                         last_block = first_block +
1772                                 (EXT4_BLOCKS_PER_GROUP(sb) - 1);
1773
1774                 block_bitmap = ext4_block_bitmap(sb, gdp);
1775                 if (block_bitmap < first_block || block_bitmap > last_block) {
1776                         ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
1777                                "Block bitmap for group %u not in group "
1778                                "(block %llu)!", i, block_bitmap);
1779                         return 0;
1780                 }
1781                 inode_bitmap = ext4_inode_bitmap(sb, gdp);
1782                 if (inode_bitmap < first_block || inode_bitmap > last_block) {
1783                         ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
1784                                "Inode bitmap for group %u not in group "
1785                                "(block %llu)!", i, inode_bitmap);
1786                         return 0;
1787                 }
1788                 inode_table = ext4_inode_table(sb, gdp);
1789                 if (inode_table < first_block ||
1790                     inode_table + sbi->s_itb_per_group - 1 > last_block) {
1791                         ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
1792                                "Inode table for group %u not in group "
1793                                "(block %llu)!", i, inode_table);
1794                         return 0;
1795                 }
1796                 ext4_lock_group(sb, i);
1797                 if (!ext4_group_desc_csum_verify(sbi, i, gdp)) {
1798                         ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
1799                                  "Checksum for group %u failed (%u!=%u)",
1800                                  i, le16_to_cpu(ext4_group_desc_csum(sbi, i,
1801                                      gdp)), le16_to_cpu(gdp->bg_checksum));
1802                         if (!(sb->s_flags & MS_RDONLY)) {
1803                                 ext4_unlock_group(sb, i);
1804                                 return 0;
1805                         }
1806                 }
1807                 ext4_unlock_group(sb, i);
1808                 if (!flexbg_flag)
1809                         first_block += EXT4_BLOCKS_PER_GROUP(sb);
1810         }
1811
1812         ext4_free_blocks_count_set(sbi->s_es, ext4_count_free_blocks(sb));
1813         sbi->s_es->s_free_inodes_count =cpu_to_le32(ext4_count_free_inodes(sb));
1814         return 1;
1815 }
1816
1817 /* ext4_orphan_cleanup() walks a singly-linked list of inodes (starting at
1818  * the superblock) which were deleted from all directories, but held open by
1819  * a process at the time of a crash.  We walk the list and try to delete these
1820  * inodes at recovery time (only with a read-write filesystem).
1821  *
1822  * In order to keep the orphan inode chain consistent during traversal (in
1823  * case of crash during recovery), we link each inode into the superblock
1824  * orphan list_head and handle it the same way as an inode deletion during
1825  * normal operation (which journals the operations for us).
1826  *
1827  * We only do an iget() and an iput() on each inode, which is very safe if we
1828  * accidentally point at an in-use or already deleted inode.  The worst that
1829  * can happen in this case is that we get a "bit already cleared" message from
1830  * ext4_free_inode().  The only reason we would point at a wrong inode is if
1831  * e2fsck was run on this filesystem, and it must have already done the orphan
1832  * inode cleanup for us, so we can safely abort without any further action.
1833  */
1834 static void ext4_orphan_cleanup(struct super_block *sb,
1835                                 struct ext4_super_block *es)
1836 {
1837         unsigned int s_flags = sb->s_flags;
1838         int nr_orphans = 0, nr_truncates = 0;
1839 #ifdef CONFIG_QUOTA
1840         int i;
1841 #endif
1842         if (!es->s_last_orphan) {
1843                 jbd_debug(4, "no orphan inodes to clean up\n");
1844                 return;
1845         }
1846
1847         if (bdev_read_only(sb->s_bdev)) {
1848                 ext4_msg(sb, KERN_ERR, "write access "
1849                         "unavailable, skipping orphan cleanup");
1850                 return;
1851         }
1852
1853         if (EXT4_SB(sb)->s_mount_state & EXT4_ERROR_FS) {
1854                 if (es->s_last_orphan)
1855                         jbd_debug(1, "Errors on filesystem, "
1856                                   "clearing orphan list.\n");
1857                 es->s_last_orphan = 0;
1858                 jbd_debug(1, "Skipping orphan recovery on fs with errors.\n");
1859                 return;
1860         }
1861
1862         if (s_flags & MS_RDONLY) {
1863                 ext4_msg(sb, KERN_INFO, "orphan cleanup on readonly fs");
1864                 sb->s_flags &= ~MS_RDONLY;
1865         }
1866 #ifdef CONFIG_QUOTA
1867         /* Needed for iput() to work correctly and not trash data */
1868         sb->s_flags |= MS_ACTIVE;
1869         /* Turn on quotas so that they are updated correctly */
1870         for (i = 0; i < MAXQUOTAS; i++) {
1871                 if (EXT4_SB(sb)->s_qf_names[i]) {
1872                         int ret = ext4_quota_on_mount(sb, i);
1873                         if (ret < 0)
1874                                 ext4_msg(sb, KERN_ERR,
1875                                         "Cannot turn on journaled "
1876                                         "quota: error %d", ret);
1877                 }
1878         }
1879 #endif
1880
1881         while (es->s_last_orphan) {
1882                 struct inode *inode;
1883
1884                 inode = ext4_orphan_get(sb, le32_to_cpu(es->s_last_orphan));
1885                 if (IS_ERR(inode)) {
1886                         es->s_last_orphan = 0;
1887                         break;
1888                 }
1889
1890                 list_add(&EXT4_I(inode)->i_orphan, &EXT4_SB(sb)->s_orphan);
1891                 vfs_dq_init(inode);
1892                 if (inode->i_nlink) {
1893                         ext4_msg(sb, KERN_DEBUG,
1894                                 "%s: truncating inode %lu to %lld bytes",
1895                                 __func__, inode->i_ino, inode->i_size);
1896                         jbd_debug(2, "truncating inode %lu to %lld bytes\n",
1897                                   inode->i_ino, inode->i_size);
1898                         ext4_truncate(inode);
1899                         nr_truncates++;
1900                 } else {
1901                         ext4_msg(sb, KERN_DEBUG,
1902                                 "%s: deleting unreferenced inode %lu",
1903                                 __func__, inode->i_ino);
1904                         jbd_debug(2, "deleting unreferenced inode %lu\n",
1905                                   inode->i_ino);
1906                         nr_orphans++;
1907                 }
1908                 iput(inode);  /* The delete magic happens here! */
1909         }
1910
1911 #define PLURAL(x) (x), ((x) == 1) ? "" : "s"
1912
1913         if (nr_orphans)
1914                 ext4_msg(sb, KERN_INFO, "%d orphan inode%s deleted",
1915                        PLURAL(nr_orphans));
1916         if (nr_truncates)
1917                 ext4_msg(sb, KERN_INFO, "%d truncate%s cleaned up",
1918                        PLURAL(nr_truncates));
1919 #ifdef CONFIG_QUOTA
1920         /* Turn quotas off */
1921         for (i = 0; i < MAXQUOTAS; i++) {
1922                 if (sb_dqopt(sb)->files[i])
1923                         vfs_quota_off(sb, i, 0);
1924         }
1925 #endif
1926         sb->s_flags = s_flags; /* Restore MS_RDONLY status */
1927 }
1928
1929 /*
1930  * Maximal extent format file size.
1931  * Resulting logical blkno at s_maxbytes must fit in our on-disk
1932  * extent format containers, within a sector_t, and within i_blocks
1933  * in the vfs.  ext4 inode has 48 bits of i_block in fsblock units,
1934  * so that won't be a limiting factor.
1935  *
1936  * Note, this does *not* consider any metadata overhead for vfs i_blocks.
1937  */
1938 static loff_t ext4_max_size(int blkbits, int has_huge_files)
1939 {
1940         loff_t res;
1941         loff_t upper_limit = MAX_LFS_FILESIZE;
1942
1943         /* small i_blocks in vfs inode? */
1944         if (!has_huge_files || sizeof(blkcnt_t) < sizeof(u64)) {
1945                 /*
1946                  * CONFIG_LBDAF is not enabled implies the inode
1947                  * i_block represent total blocks in 512 bytes
1948                  * 32 == size of vfs inode i_blocks * 8
1949                  */
1950                 upper_limit = (1LL << 32) - 1;
1951
1952                 /* total blocks in file system block size */
1953                 upper_limit >>= (blkbits - 9);
1954                 upper_limit <<= blkbits;
1955         }
1956
1957         /* 32-bit extent-start container, ee_block */
1958         res = 1LL << 32;
1959         res <<= blkbits;
1960         res -= 1;
1961
1962         /* Sanity check against vm- & vfs- imposed limits */
1963         if (res > upper_limit)
1964                 res = upper_limit;
1965
1966         return res;
1967 }
1968
1969 /*
1970  * Maximal bitmap file size.  There is a direct, and {,double-,triple-}indirect
1971  * block limit, and also a limit of (2^48 - 1) 512-byte sectors in i_blocks.
1972  * We need to be 1 filesystem block less than the 2^48 sector limit.
1973  */
1974 static loff_t ext4_max_bitmap_size(int bits, int has_huge_files)
1975 {
1976         loff_t res = EXT4_NDIR_BLOCKS;
1977         int meta_blocks;
1978         loff_t upper_limit;
1979         /* This is calculated to be the largest file size for a dense, block
1980          * mapped file such that the file's total number of 512-byte sectors,
1981          * including data and all indirect blocks, does not exceed (2^48 - 1).
1982          *
1983          * __u32 i_blocks_lo and _u16 i_blocks_high represent the total
1984          * number of 512-byte sectors of the file.
1985          */
1986
1987         if (!has_huge_files || sizeof(blkcnt_t) < sizeof(u64)) {
1988                 /*
1989                  * !has_huge_files or CONFIG_LBDAF not enabled implies that
1990                  * the inode i_block field represents total file blocks in
1991                  * 2^32 512-byte sectors == size of vfs inode i_blocks * 8
1992                  */
1993                 upper_limit = (1LL << 32) - 1;
1994
1995                 /* total blocks in file system block size */
1996                 upper_limit >>= (bits - 9);
1997
1998         } else {
1999                 /*
2000                  * We use 48 bit ext4_inode i_blocks
2001                  * With EXT4_HUGE_FILE_FL set the i_blocks
2002                  * represent total number of blocks in
2003                  * file system block size
2004                  */
2005                 upper_limit = (1LL << 48) - 1;
2006
2007         }
2008
2009         /* indirect blocks */
2010         meta_blocks = 1;
2011         /* double indirect blocks */
2012         meta_blocks += 1 + (1LL << (bits-2));
2013         /* tripple indirect blocks */
2014         meta_blocks += 1 + (1LL << (bits-2)) + (1LL << (2*(bits-2)));
2015
2016         upper_limit -= meta_blocks;
2017         upper_limit <<= bits;
2018
2019         res += 1LL << (bits-2);
2020         res += 1LL << (2*(bits-2));
2021         res += 1LL << (3*(bits-2));
2022         res <<= bits;
2023         if (res > upper_limit)
2024                 res = upper_limit;
2025
2026         if (res > MAX_LFS_FILESIZE)
2027                 res = MAX_LFS_FILESIZE;
2028
2029         return res;
2030 }
2031
2032 static ext4_fsblk_t descriptor_loc(struct super_block *sb,
2033                                    ext4_fsblk_t logical_sb_block, int nr)
2034 {
2035         struct ext4_sb_info *sbi = EXT4_SB(sb);
2036         ext4_group_t bg, first_meta_bg;
2037         int has_super = 0;
2038
2039         first_meta_bg = le32_to_cpu(sbi->s_es->s_first_meta_bg);
2040
2041         if (!EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_META_BG) ||
2042             nr < first_meta_bg)
2043                 return logical_sb_block + nr + 1;
2044         bg = sbi->s_desc_per_block * nr;
2045         if (ext4_bg_has_super(sb, bg))
2046                 has_super = 1;
2047
2048         return (has_super + ext4_group_first_block_no(sb, bg));
2049 }
2050
2051 /**
2052  * ext4_get_stripe_size: Get the stripe size.
2053  * @sbi: In memory super block info
2054  *
2055  * If we have specified it via mount option, then
2056  * use the mount option value. If the value specified at mount time is
2057  * greater than the blocks per group use the super block value.
2058  * If the super block value is greater than blocks per group return 0.
2059  * Allocator needs it be less than blocks per group.
2060  *
2061  */
2062 static unsigned long ext4_get_stripe_size(struct ext4_sb_info *sbi)
2063 {
2064         unsigned long stride = le16_to_cpu(sbi->s_es->s_raid_stride);
2065         unsigned long stripe_width =
2066                         le32_to_cpu(sbi->s_es->s_raid_stripe_width);
2067
2068         if (sbi->s_stripe && sbi->s_stripe <= sbi->s_blocks_per_group)
2069                 return sbi->s_stripe;
2070
2071         if (stripe_width <= sbi->s_blocks_per_group)
2072                 return stripe_width;
2073
2074         if (stride <= sbi->s_blocks_per_group)
2075                 return stride;
2076
2077         return 0;
2078 }
2079
2080 /* sysfs supprt */
2081
2082 struct ext4_attr {
2083         struct attribute attr;
2084         ssize_t (*show)(struct ext4_attr *, struct ext4_sb_info *, char *);
2085         ssize_t (*store)(struct ext4_attr *, struct ext4_sb_info *,
2086                          const char *, size_t);
2087         int offset;
2088 };
2089
2090 static int parse_strtoul(const char *buf,
2091                 unsigned long max, unsigned long *value)
2092 {
2093         char *endp;
2094
2095         while (*buf && isspace(*buf))
2096                 buf++;
2097         *value = simple_strtoul(buf, &endp, 0);
2098         while (*endp && isspace(*endp))
2099                 endp++;
2100         if (*endp || *value > max)
2101                 return -EINVAL;
2102
2103         return 0;
2104 }
2105
2106 static ssize_t delayed_allocation_blocks_show(struct ext4_attr *a,
2107                                               struct ext4_sb_info *sbi,
2108                                               char *buf)
2109 {
2110         return snprintf(buf, PAGE_SIZE, "%llu\n",
2111                         (s64) percpu_counter_sum(&sbi->s_dirtyblocks_counter));
2112 }
2113
2114 static ssize_t session_write_kbytes_show(struct ext4_attr *a,
2115                                          struct ext4_sb_info *sbi, char *buf)
2116 {
2117         struct super_block *sb = sbi->s_buddy_cache->i_sb;
2118
2119         return snprintf(buf, PAGE_SIZE, "%lu\n",
2120                         (part_stat_read(sb->s_bdev->bd_part, sectors[1]) -
2121                          sbi->s_sectors_written_start) >> 1);
2122 }
2123
2124 static ssize_t lifetime_write_kbytes_show(struct ext4_attr *a,
2125                                           struct ext4_sb_info *sbi, char *buf)
2126 {
2127         struct super_block *sb = sbi->s_buddy_cache->i_sb;
2128
2129         return snprintf(buf, PAGE_SIZE, "%llu\n",
2130                         sbi->s_kbytes_written +
2131                         ((part_stat_read(sb->s_bdev->bd_part, sectors[1]) -
2132                           EXT4_SB(sb)->s_sectors_written_start) >> 1));
2133 }
2134
2135 static ssize_t inode_readahead_blks_store(struct ext4_attr *a,
2136                                           struct ext4_sb_info *sbi,
2137                                           const char *buf, size_t count)
2138 {
2139         unsigned long t;
2140
2141         if (parse_strtoul(buf, 0x40000000, &t))
2142                 return -EINVAL;
2143
2144         if (!is_power_of_2(t))
2145                 return -EINVAL;
2146
2147         sbi->s_inode_readahead_blks = t;
2148         return count;
2149 }
2150
2151 static ssize_t sbi_ui_show(struct ext4_attr *a,
2152                            struct ext4_sb_info *sbi, char *buf)
2153 {
2154         unsigned int *ui = (unsigned int *) (((char *) sbi) + a->offset);
2155
2156         return snprintf(buf, PAGE_SIZE, "%u\n", *ui);
2157 }
2158
2159 static ssize_t sbi_ui_store(struct ext4_attr *a,
2160                             struct ext4_sb_info *sbi,
2161                             const char *buf, size_t count)
2162 {
2163         unsigned int *ui = (unsigned int *) (((char *) sbi) + a->offset);
2164         unsigned long t;
2165
2166         if (parse_strtoul(buf, 0xffffffff, &t))
2167                 return -EINVAL;
2168         *ui = t;
2169         return count;
2170 }
2171
2172 #define EXT4_ATTR_OFFSET(_name,_mode,_show,_store,_elname) \
2173 static struct ext4_attr ext4_attr_##_name = {                   \
2174         .attr = {.name = __stringify(_name), .mode = _mode },   \
2175         .show   = _show,                                        \
2176         .store  = _store,                                       \
2177         .offset = offsetof(struct ext4_sb_info, _elname),       \
2178 }
2179 #define EXT4_ATTR(name, mode, show, store) \
2180 static struct ext4_attr ext4_attr_##name = __ATTR(name, mode, show, store)
2181
2182 #define EXT4_RO_ATTR(name) EXT4_ATTR(name, 0444, name##_show, NULL)
2183 #define EXT4_RW_ATTR(name) EXT4_ATTR(name, 0644, name##_show, name##_store)
2184 #define EXT4_RW_ATTR_SBI_UI(name, elname)       \
2185         EXT4_ATTR_OFFSET(name, 0644, sbi_ui_show, sbi_ui_store, elname)
2186 #define ATTR_LIST(name) &ext4_attr_##name.attr
2187
2188 EXT4_RO_ATTR(delayed_allocation_blocks);
2189 EXT4_RO_ATTR(session_write_kbytes);
2190 EXT4_RO_ATTR(lifetime_write_kbytes);
2191 EXT4_ATTR_OFFSET(inode_readahead_blks, 0644, sbi_ui_show,
2192                  inode_readahead_blks_store, s_inode_readahead_blks);
2193 EXT4_RW_ATTR_SBI_UI(inode_goal, s_inode_goal);
2194 EXT4_RW_ATTR_SBI_UI(mb_stats, s_mb_stats);
2195 EXT4_RW_ATTR_SBI_UI(mb_max_to_scan, s_mb_max_to_scan);
2196 EXT4_RW_ATTR_SBI_UI(mb_min_to_scan, s_mb_min_to_scan);
2197 EXT4_RW_ATTR_SBI_UI(mb_order2_req, s_mb_order2_reqs);
2198 EXT4_RW_ATTR_SBI_UI(mb_stream_req, s_mb_stream_request);
2199 EXT4_RW_ATTR_SBI_UI(mb_group_prealloc, s_mb_group_prealloc);
2200
2201 static struct attribute *ext4_attrs[] = {
2202         ATTR_LIST(delayed_allocation_blocks),
2203         ATTR_LIST(session_write_kbytes),
2204         ATTR_LIST(lifetime_write_kbytes),
2205         ATTR_LIST(inode_readahead_blks),
2206         ATTR_LIST(inode_goal),
2207         ATTR_LIST(mb_stats),
2208         ATTR_LIST(mb_max_to_scan),
2209         ATTR_LIST(mb_min_to_scan),
2210         ATTR_LIST(mb_order2_req),
2211         ATTR_LIST(mb_stream_req),
2212         ATTR_LIST(mb_group_prealloc),
2213         NULL,
2214 };
2215
2216 static ssize_t ext4_attr_show(struct kobject *kobj,
2217                               struct attribute *attr, char *buf)
2218 {
2219         struct ext4_sb_info *sbi = container_of(kobj, struct ext4_sb_info,
2220                                                 s_kobj);
2221         struct ext4_attr *a = container_of(attr, struct ext4_attr, attr);
2222
2223         return a->show ? a->show(a, sbi, buf) : 0;
2224 }
2225
2226 static ssize_t ext4_attr_store(struct kobject *kobj,
2227                                struct attribute *attr,
2228                                const char *buf, size_t len)
2229 {
2230         struct ext4_sb_info *sbi = container_of(kobj, struct ext4_sb_info,
2231                                                 s_kobj);
2232         struct ext4_attr *a = container_of(attr, struct ext4_attr, attr);
2233
2234         return a->store ? a->store(a, sbi, buf, len) : 0;
2235 }
2236
2237 static void ext4_sb_release(struct kobject *kobj)
2238 {
2239         struct ext4_sb_info *sbi = container_of(kobj, struct ext4_sb_info,
2240                                                 s_kobj);
2241         complete(&sbi->s_kobj_unregister);
2242 }
2243
2244
2245 static struct sysfs_ops ext4_attr_ops = {
2246         .show   = ext4_attr_show,
2247         .store  = ext4_attr_store,
2248 };
2249
2250 static struct kobj_type ext4_ktype = {
2251         .default_attrs  = ext4_attrs,
2252         .sysfs_ops      = &ext4_attr_ops,
2253         .release        = ext4_sb_release,
2254 };
2255
2256 static int ext4_fill_super(struct super_block *sb, void *data, int silent)
2257                                 __releases(kernel_lock)
2258                                 __acquires(kernel_lock)
2259 {
2260         struct buffer_head *bh;
2261         struct ext4_super_block *es = NULL;
2262         struct ext4_sb_info *sbi;
2263         ext4_fsblk_t block;
2264         ext4_fsblk_t sb_block = get_sb_block(&data);
2265         ext4_fsblk_t logical_sb_block;
2266         unsigned long offset = 0;
2267         unsigned long journal_devnum = 0;
2268         unsigned long def_mount_opts;
2269         struct inode *root;
2270         char *cp;
2271         const char *descr;
2272         int ret = -EINVAL;
2273         int blocksize;
2274         unsigned int db_count;
2275         unsigned int i;
2276         int needs_recovery, has_huge_files;
2277         int features;
2278         __u64 blocks_count;
2279         int err;
2280         unsigned int journal_ioprio = DEFAULT_JOURNAL_IOPRIO;
2281
2282         sbi = kzalloc(sizeof(*sbi), GFP_KERNEL);
2283         if (!sbi)
2284                 return -ENOMEM;
2285
2286         sbi->s_blockgroup_lock =
2287                 kzalloc(sizeof(struct blockgroup_lock), GFP_KERNEL);
2288         if (!sbi->s_blockgroup_lock) {
2289                 kfree(sbi);
2290                 return -ENOMEM;
2291         }
2292         sb->s_fs_info = sbi;
2293         sbi->s_mount_opt = 0;
2294         sbi->s_resuid = EXT4_DEF_RESUID;
2295         sbi->s_resgid = EXT4_DEF_RESGID;
2296         sbi->s_inode_readahead_blks = EXT4_DEF_INODE_READAHEAD_BLKS;
2297         sbi->s_sb_block = sb_block;
2298         sbi->s_sectors_written_start = part_stat_read(sb->s_bdev->bd_part,
2299                                                       sectors[1]);
2300
2301         unlock_kernel();
2302
2303         /* Cleanup superblock name */
2304         for (cp = sb->s_id; (cp = strchr(cp, '/'));)
2305                 *cp = '!';
2306
2307         blocksize = sb_min_blocksize(sb, EXT4_MIN_BLOCK_SIZE);
2308         if (!blocksize) {
2309                 ext4_msg(sb, KERN_ERR, "unable to set blocksize");
2310                 goto out_fail;
2311         }
2312
2313         /*
2314          * The ext4 superblock will not be buffer aligned for other than 1kB
2315          * block sizes.  We need to calculate the offset from buffer start.
2316          */
2317         if (blocksize != EXT4_MIN_BLOCK_SIZE) {
2318                 logical_sb_block = sb_block * EXT4_MIN_BLOCK_SIZE;
2319                 offset = do_div(logical_sb_block, blocksize);
2320         } else {
2321                 logical_sb_block = sb_block;
2322         }
2323
2324         if (!(bh = sb_bread(sb, logical_sb_block))) {
2325                 ext4_msg(sb, KERN_ERR, "unable to read superblock");
2326                 goto out_fail;
2327         }
2328         /*
2329          * Note: s_es must be initialized as soon as possible because
2330          *       some ext4 macro-instructions depend on its value
2331          */
2332         es = (struct ext4_super_block *) (((char *)bh->b_data) + offset);
2333         sbi->s_es = es;
2334         sb->s_magic = le16_to_cpu(es->s_magic);
2335         if (sb->s_magic != EXT4_SUPER_MAGIC)
2336                 goto cantfind_ext4;
2337         sbi->s_kbytes_written = le64_to_cpu(es->s_kbytes_written);
2338
2339         /* Set defaults before we parse the mount options */
2340         def_mount_opts = le32_to_cpu(es->s_default_mount_opts);
2341         if (def_mount_opts & EXT4_DEFM_DEBUG)
2342                 set_opt(sbi->s_mount_opt, DEBUG);
2343         if (def_mount_opts & EXT4_DEFM_BSDGROUPS)
2344                 set_opt(sbi->s_mount_opt, GRPID);
2345         if (def_mount_opts & EXT4_DEFM_UID16)
2346                 set_opt(sbi->s_mount_opt, NO_UID32);
2347 #ifdef CONFIG_EXT4_FS_XATTR
2348         if (def_mount_opts & EXT4_DEFM_XATTR_USER)
2349                 set_opt(sbi->s_mount_opt, XATTR_USER);
2350 #endif
2351 #ifdef CONFIG_EXT4_FS_POSIX_ACL
2352         if (def_mount_opts & EXT4_DEFM_ACL)
2353                 set_opt(sbi->s_mount_opt, POSIX_ACL);
2354 #endif
2355         if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_DATA)
2356                 sbi->s_mount_opt |= EXT4_MOUNT_JOURNAL_DATA;
2357         else if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_ORDERED)
2358                 sbi->s_mount_opt |= EXT4_MOUNT_ORDERED_DATA;
2359         else if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_WBACK)
2360                 sbi->s_mount_opt |= EXT4_MOUNT_WRITEBACK_DATA;
2361
2362         if (le16_to_cpu(sbi->s_es->s_errors) == EXT4_ERRORS_PANIC)
2363                 set_opt(sbi->s_mount_opt, ERRORS_PANIC);
2364         else if (le16_to_cpu(sbi->s_es->s_errors) == EXT4_ERRORS_CONTINUE)
2365                 set_opt(sbi->s_mount_opt, ERRORS_CONT);
2366         else
2367                 set_opt(sbi->s_mount_opt, ERRORS_RO);
2368
2369         sbi->s_resuid = le16_to_cpu(es->s_def_resuid);
2370         sbi->s_resgid = le16_to_cpu(es->s_def_resgid);
2371         sbi->s_commit_interval = JBD2_DEFAULT_MAX_COMMIT_AGE * HZ;
2372         sbi->s_min_batch_time = EXT4_DEF_MIN_BATCH_TIME;
2373         sbi->s_max_batch_time = EXT4_DEF_MAX_BATCH_TIME;
2374         sbi->s_mb_history_max = default_mb_history_length;
2375
2376         set_opt(sbi->s_mount_opt, BARRIER);
2377
2378         /*
2379          * enable delayed allocation by default
2380          * Use -o nodelalloc to turn it off
2381          */
2382         set_opt(sbi->s_mount_opt, DELALLOC);
2383
2384         if (!parse_options((char *) data, sb, &journal_devnum,
2385                            &journal_ioprio, NULL, 0))
2386                 goto failed_mount;
2387
2388         sb->s_flags = (sb->s_flags & ~MS_POSIXACL) |
2389                 ((sbi->s_mount_opt & EXT4_MOUNT_POSIX_ACL) ? MS_POSIXACL : 0);
2390
2391         if (le32_to_cpu(es->s_rev_level) == EXT4_GOOD_OLD_REV &&
2392             (EXT4_HAS_COMPAT_FEATURE(sb, ~0U) ||
2393              EXT4_HAS_RO_COMPAT_FEATURE(sb, ~0U) ||
2394              EXT4_HAS_INCOMPAT_FEATURE(sb, ~0U)))
2395                 ext4_msg(sb, KERN_WARNING,
2396                        "feature flags set on rev 0 fs, "
2397                        "running e2fsck is recommended");
2398
2399         /*
2400          * Check feature flags regardless of the revision level, since we
2401          * previously didn't change the revision level when setting the flags,
2402          * so there is a chance incompat flags are set on a rev 0 filesystem.
2403          */
2404         features = EXT4_HAS_INCOMPAT_FEATURE(sb, ~EXT4_FEATURE_INCOMPAT_SUPP);
2405         if (features) {
2406                 ext4_msg(sb, KERN_ERR,
2407                         "Couldn't mount because of "
2408                         "unsupported optional features (%x)",
2409                         (le32_to_cpu(EXT4_SB(sb)->s_es->s_feature_incompat) &
2410                         ~EXT4_FEATURE_INCOMPAT_SUPP));
2411                 goto failed_mount;
2412         }
2413         features = EXT4_HAS_RO_COMPAT_FEATURE(sb, ~EXT4_FEATURE_RO_COMPAT_SUPP);
2414         if (!(sb->s_flags & MS_RDONLY) && features) {
2415                 ext4_msg(sb, KERN_ERR,
2416                         "Couldn't mount RDWR because of "
2417                         "unsupported optional features (%x)",
2418                         (le32_to_cpu(EXT4_SB(sb)->s_es->s_feature_ro_compat) &
2419                         ~EXT4_FEATURE_RO_COMPAT_SUPP));
2420                 goto failed_mount;
2421         }
2422         has_huge_files = EXT4_HAS_RO_COMPAT_FEATURE(sb,
2423                                     EXT4_FEATURE_RO_COMPAT_HUGE_FILE);
2424         if (has_huge_files) {
2425                 /*
2426                  * Large file size enabled file system can only be
2427                  * mount if kernel is build with CONFIG_LBDAF
2428                  */
2429                 if (sizeof(root->i_blocks) < sizeof(u64) &&
2430                                 !(sb->s_flags & MS_RDONLY)) {
2431                         ext4_msg(sb, KERN_ERR, "Filesystem with huge "
2432                                         "files cannot be mounted read-write "
2433                                         "without CONFIG_LBDAF");
2434                         goto failed_mount;
2435                 }
2436         }
2437         blocksize = BLOCK_SIZE << le32_to_cpu(es->s_log_block_size);
2438
2439         if (blocksize < EXT4_MIN_BLOCK_SIZE ||
2440             blocksize > EXT4_MAX_BLOCK_SIZE) {
2441                 ext4_msg(sb, KERN_ERR,
2442                        "Unsupported filesystem blocksize %d", blocksize);
2443                 goto failed_mount;
2444         }
2445
2446         if (sb->s_blocksize != blocksize) {
2447                 /* Validate the filesystem blocksize */
2448                 if (!sb_set_blocksize(sb, blocksize)) {
2449                         ext4_msg(sb, KERN_ERR, "bad block size %d",
2450                                         blocksize);
2451                         goto failed_mount;
2452                 }
2453
2454                 brelse(bh);
2455                 logical_sb_block = sb_block * EXT4_MIN_BLOCK_SIZE;
2456                 offset = do_div(logical_sb_block, blocksize);
2457                 bh = sb_bread(sb, logical_sb_block);
2458                 if (!bh) {
2459                         ext4_msg(sb, KERN_ERR,
2460                                "Can't read superblock on 2nd try");
2461                         goto failed_mount;
2462                 }
2463                 es = (struct ext4_super_block *)(((char *)bh->b_data) + offset);
2464                 sbi->s_es = es;
2465                 if (es->s_magic != cpu_to_le16(EXT4_SUPER_MAGIC)) {
2466                         ext4_msg(sb, KERN_ERR,
2467                                "Magic mismatch, very weird!");
2468                         goto failed_mount;
2469                 }
2470         }
2471
2472         sbi->s_bitmap_maxbytes = ext4_max_bitmap_size(sb->s_blocksize_bits,
2473                                                       has_huge_files);
2474         sb->s_maxbytes = ext4_max_size(sb->s_blocksize_bits, has_huge_files);
2475
2476         if (le32_to_cpu(es->s_rev_level) == EXT4_GOOD_OLD_REV) {
2477                 sbi->s_inode_size = EXT4_GOOD_OLD_INODE_SIZE;
2478                 sbi->s_first_ino = EXT4_GOOD_OLD_FIRST_INO;
2479         } else {
2480                 sbi->s_inode_size = le16_to_cpu(es->s_inode_size);
2481                 sbi->s_first_ino = le32_to_cpu(es->s_first_ino);
2482                 if ((sbi->s_inode_size < EXT4_GOOD_OLD_INODE_SIZE) ||
2483                     (!is_power_of_2(sbi->s_inode_size)) ||
2484                     (sbi->s_inode_size > blocksize)) {
2485                         ext4_msg(sb, KERN_ERR,
2486                                "unsupported inode size: %d",
2487                                sbi->s_inode_size);
2488                         goto failed_mount;
2489                 }
2490                 if (sbi->s_inode_size > EXT4_GOOD_OLD_INODE_SIZE)
2491                         sb->s_time_gran = 1 << (EXT4_EPOCH_BITS - 2);
2492         }
2493
2494         sbi->s_desc_size = le16_to_cpu(es->s_desc_size);
2495         if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_64BIT)) {
2496                 if (sbi->s_desc_size < EXT4_MIN_DESC_SIZE_64BIT ||
2497                     sbi->s_desc_size > EXT4_MAX_DESC_SIZE ||
2498                     !is_power_of_2(sbi->s_desc_size)) {
2499                         ext4_msg(sb, KERN_ERR,
2500                                "unsupported descriptor size %lu",
2501                                sbi->s_desc_size);
2502                         goto failed_mount;
2503                 }
2504         } else
2505                 sbi->s_desc_size = EXT4_MIN_DESC_SIZE;
2506
2507         sbi->s_blocks_per_group = le32_to_cpu(es->s_blocks_per_group);
2508         sbi->s_inodes_per_group = le32_to_cpu(es->s_inodes_per_group);
2509         if (EXT4_INODE_SIZE(sb) == 0 || EXT4_INODES_PER_GROUP(sb) == 0)
2510                 goto cantfind_ext4;
2511
2512         sbi->s_inodes_per_block = blocksize / EXT4_INODE_SIZE(sb);
2513         if (sbi->s_inodes_per_block == 0)
2514                 goto cantfind_ext4;
2515         sbi->s_itb_per_group = sbi->s_inodes_per_group /
2516                                         sbi->s_inodes_per_block;
2517         sbi->s_desc_per_block = blocksize / EXT4_DESC_SIZE(sb);
2518         sbi->s_sbh = bh;
2519         sbi->s_mount_state = le16_to_cpu(es->s_state);
2520         sbi->s_addr_per_block_bits = ilog2(EXT4_ADDR_PER_BLOCK(sb));
2521         sbi->s_desc_per_block_bits = ilog2(EXT4_DESC_PER_BLOCK(sb));
2522
2523         for (i = 0; i < 4; i++)
2524                 sbi->s_hash_seed[i] = le32_to_cpu(es->s_hash_seed[i]);
2525         sbi->s_def_hash_version = es->s_def_hash_version;
2526         i = le32_to_cpu(es->s_flags);
2527         if (i & EXT2_FLAGS_UNSIGNED_HASH)
2528                 sbi->s_hash_unsigned = 3;
2529         else if ((i & EXT2_FLAGS_SIGNED_HASH) == 0) {
2530 #ifdef __CHAR_UNSIGNED__
2531                 es->s_flags |= cpu_to_le32(EXT2_FLAGS_UNSIGNED_HASH);
2532                 sbi->s_hash_unsigned = 3;
2533 #else
2534                 es->s_flags |= cpu_to_le32(EXT2_FLAGS_SIGNED_HASH);
2535 #endif
2536                 sb->s_dirt = 1;
2537         }
2538
2539         if (sbi->s_blocks_per_group > blocksize * 8) {
2540                 ext4_msg(sb, KERN_ERR,
2541                        "#blocks per group too big: %lu",
2542                        sbi->s_blocks_per_group);
2543                 goto failed_mount;
2544         }
2545         if (sbi->s_inodes_per_group > blocksize * 8) {
2546                 ext4_msg(sb, KERN_ERR,
2547                        "#inodes per group too big: %lu",
2548                        sbi->s_inodes_per_group);
2549                 goto failed_mount;
2550         }
2551
2552         if (ext4_blocks_count(es) >
2553                     (sector_t)(~0ULL) >> (sb->s_blocksize_bits - 9)) {
2554                 ext4_msg(sb, KERN_ERR, "filesystem"
2555                         " too large to mount safely");
2556                 if (sizeof(sector_t) < 8)
2557                         ext4_msg(sb, KERN_WARNING, "CONFIG_LBDAF not enabled");
2558                 goto failed_mount;
2559         }
2560
2561         if (EXT4_BLOCKS_PER_GROUP(sb) == 0)
2562                 goto cantfind_ext4;
2563
2564         /* check blocks count against device size */
2565         blocks_count = sb->s_bdev->bd_inode->i_size >> sb->s_blocksize_bits;
2566         if (blocks_count && ext4_blocks_count(es) > blocks_count) {
2567                 ext4_msg(sb, KERN_WARNING, "bad geometry: block count %llu "
2568                        "exceeds size of device (%llu blocks)",
2569                        ext4_blocks_count(es), blocks_count);
2570                 goto failed_mount;
2571         }
2572
2573         /*
2574          * It makes no sense for the first data block to be beyond the end
2575          * of the filesystem.
2576          */
2577         if (le32_to_cpu(es->s_first_data_block) >= ext4_blocks_count(es)) {
2578                 ext4_msg(sb, KERN_WARNING, "bad geometry: first data"
2579                          "block %u is beyond end of filesystem (%llu)",
2580                          le32_to_cpu(es->s_first_data_block),
2581                          ext4_blocks_count(es));
2582                 goto failed_mount;
2583         }
2584         blocks_count = (ext4_blocks_count(es) -
2585                         le32_to_cpu(es->s_first_data_block) +
2586                         EXT4_BLOCKS_PER_GROUP(sb) - 1);
2587         do_div(blocks_count, EXT4_BLOCKS_PER_GROUP(sb));
2588         if (blocks_count > ((uint64_t)1<<32) - EXT4_DESC_PER_BLOCK(sb)) {
2589                 ext4_msg(sb, KERN_WARNING, "groups count too large: %u "
2590                        "(block count %llu, first data block %u, "
2591                        "blocks per group %lu)", sbi->s_groups_count,
2592                        ext4_blocks_count(es),
2593                        le32_to_cpu(es->s_first_data_block),
2594                        EXT4_BLOCKS_PER_GROUP(sb));
2595                 goto failed_mount;
2596         }
2597         sbi->s_groups_count = blocks_count;
2598         db_count = (sbi->s_groups_count + EXT4_DESC_PER_BLOCK(sb) - 1) /
2599                    EXT4_DESC_PER_BLOCK(sb);
2600         sbi->s_group_desc = kmalloc(db_count * sizeof(struct buffer_head *),
2601                                     GFP_KERNEL);
2602         if (sbi->s_group_desc == NULL) {
2603                 ext4_msg(sb, KERN_ERR, "not enough memory");
2604                 goto failed_mount;
2605         }
2606
2607 #ifdef CONFIG_PROC_FS
2608         if (ext4_proc_root)
2609                 sbi->s_proc = proc_mkdir(sb->s_id, ext4_proc_root);
2610 #endif
2611
2612         bgl_lock_init(sbi->s_blockgroup_lock);
2613
2614         for (i = 0; i < db_count; i++) {
2615                 block = descriptor_loc(sb, logical_sb_block, i);
2616                 sbi->s_group_desc[i] = sb_bread(sb, block);
2617                 if (!sbi->s_group_desc[i]) {
2618                         ext4_msg(sb, KERN_ERR,
2619                                "can't read group descriptor %d", i);
2620                         db_count = i;
2621                         goto failed_mount2;
2622                 }
2623         }
2624         if (!ext4_check_descriptors(sb)) {
2625                 ext4_msg(sb, KERN_ERR, "group descriptors corrupted!");
2626                 goto failed_mount2;
2627         }
2628         if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_FLEX_BG))
2629                 if (!ext4_fill_flex_info(sb)) {
2630                         ext4_msg(sb, KERN_ERR,
2631                                "unable to initialize "
2632                                "flex_bg meta info!");
2633                         goto failed_mount2;
2634                 }
2635
2636         sbi->s_gdb_count = db_count;
2637         get_random_bytes(&sbi->s_next_generation, sizeof(u32));
2638         spin_lock_init(&sbi->s_next_gen_lock);
2639
2640         err = percpu_counter_init(&sbi->s_freeblocks_counter,
2641                         ext4_count_free_blocks(sb));
2642         if (!err) {
2643                 err = percpu_counter_init(&sbi->s_freeinodes_counter,
2644                                 ext4_count_free_inodes(sb));
2645         }
2646         if (!err) {
2647                 err = percpu_counter_init(&sbi->s_dirs_counter,
2648                                 ext4_count_dirs(sb));
2649         }
2650         if (!err) {
2651                 err = percpu_counter_init(&sbi->s_dirtyblocks_counter, 0);
2652         }
2653         if (err) {
2654                 ext4_msg(sb, KERN_ERR, "insufficient memory");
2655                 goto failed_mount3;
2656         }
2657
2658         sbi->s_stripe = ext4_get_stripe_size(sbi);
2659
2660         /*
2661          * set up enough so that it can read an inode
2662          */
2663         if (!test_opt(sb, NOLOAD) &&
2664             EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL))
2665                 sb->s_op = &ext4_sops;
2666         else
2667                 sb->s_op = &ext4_nojournal_sops;
2668         sb->s_export_op = &ext4_export_ops;
2669         sb->s_xattr = ext4_xattr_handlers;
2670 #ifdef CONFIG_QUOTA
2671         sb->s_qcop = &ext4_qctl_operations;
2672         sb->dq_op = &ext4_quota_operations;
2673 #endif
2674         INIT_LIST_HEAD(&sbi->s_orphan); /* unlinked but open files */
2675         mutex_init(&sbi->s_orphan_lock);
2676         mutex_init(&sbi->s_resize_lock);
2677
2678         sb->s_root = NULL;
2679
2680         needs_recovery = (es->s_last_orphan != 0 ||
2681                           EXT4_HAS_INCOMPAT_FEATURE(sb,
2682                                     EXT4_FEATURE_INCOMPAT_RECOVER));
2683
2684         /*
2685          * The first inode we look at is the journal inode.  Don't try
2686          * root first: it may be modified in the journal!
2687          */
2688         if (!test_opt(sb, NOLOAD) &&
2689             EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL)) {
2690                 if (ext4_load_journal(sb, es, journal_devnum))
2691                         goto failed_mount3;
2692                 if (!(sb->s_flags & MS_RDONLY) &&
2693                     EXT4_SB(sb)->s_journal->j_failed_commit) {
2694                         ext4_msg(sb, KERN_CRIT, "error: "
2695                                "ext4_fill_super: Journal transaction "
2696                                "%u is corrupt",
2697                                EXT4_SB(sb)->s_journal->j_failed_commit);
2698                         if (test_opt(sb, ERRORS_RO)) {
2699                                 ext4_msg(sb, KERN_CRIT,
2700                                        "Mounting filesystem read-only");
2701                                 sb->s_flags |= MS_RDONLY;
2702                                 EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS;
2703                                 es->s_state |= cpu_to_le16(EXT4_ERROR_FS);
2704                         }
2705                         if (test_opt(sb, ERRORS_PANIC)) {
2706                                 EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS;
2707                                 es->s_state |= cpu_to_le16(EXT4_ERROR_FS);
2708                                 ext4_commit_super(sb, 1);
2709                                 goto failed_mount4;
2710                         }
2711                 }
2712         } else if (test_opt(sb, NOLOAD) && !(sb->s_flags & MS_RDONLY) &&
2713               EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER)) {
2714                 ext4_msg(sb, KERN_ERR, "required journal recovery "
2715                        "suppressed and not mounted read-only");
2716                 goto failed_mount4;
2717         } else {
2718                 clear_opt(sbi->s_mount_opt, DATA_FLAGS);
2719                 set_opt(sbi->s_mount_opt, WRITEBACK_DATA);
2720                 sbi->s_journal = NULL;
2721                 needs_recovery = 0;
2722                 goto no_journal;
2723         }
2724
2725         if (ext4_blocks_count(es) > 0xffffffffULL &&
2726             !jbd2_journal_set_features(EXT4_SB(sb)->s_journal, 0, 0,
2727                                        JBD2_FEATURE_INCOMPAT_64BIT)) {
2728                 ext4_msg(sb, KERN_ERR, "Failed to set 64-bit journal feature");
2729                 goto failed_mount4;
2730         }
2731
2732         if (test_opt(sb, JOURNAL_ASYNC_COMMIT)) {
2733                 jbd2_journal_set_features(sbi->s_journal,
2734                                 JBD2_FEATURE_COMPAT_CHECKSUM, 0,
2735                                 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT);
2736         } else if (test_opt(sb, JOURNAL_CHECKSUM)) {
2737                 jbd2_journal_set_features(sbi->s_journal,
2738                                 JBD2_FEATURE_COMPAT_CHECKSUM, 0, 0);
2739                 jbd2_journal_clear_features(sbi->s_journal, 0, 0,
2740                                 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT);
2741         } else {
2742                 jbd2_journal_clear_features(sbi->s_journal,
2743                                 JBD2_FEATURE_COMPAT_CHECKSUM, 0,
2744                                 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT);
2745         }
2746
2747         /* We have now updated the journal if required, so we can
2748          * validate the data journaling mode. */
2749         switch (test_opt(sb, DATA_FLAGS)) {
2750         case 0:
2751                 /* No mode set, assume a default based on the journal
2752                  * capabilities: ORDERED_DATA if the journal can
2753                  * cope, else JOURNAL_DATA
2754                  */
2755                 if (jbd2_journal_check_available_features
2756                     (sbi->s_journal, 0, 0, JBD2_FEATURE_INCOMPAT_REVOKE))
2757                         set_opt(sbi->s_mount_opt, ORDERED_DATA);
2758                 else
2759                         set_opt(sbi->s_mount_opt, JOURNAL_DATA);
2760                 break;
2761
2762         case EXT4_MOUNT_ORDERED_DATA:
2763         case EXT4_MOUNT_WRITEBACK_DATA:
2764                 if (!jbd2_journal_check_available_features
2765                     (sbi->s_journal, 0, 0, JBD2_FEATURE_INCOMPAT_REVOKE)) {
2766                         ext4_msg(sb, KERN_ERR, "Journal does not support "
2767                                "requested data journaling mode");
2768                         goto failed_mount4;
2769                 }
2770         default:
2771                 break;
2772         }
2773         set_task_ioprio(sbi->s_journal->j_task, journal_ioprio);
2774
2775 no_journal:
2776
2777         if (test_opt(sb, NOBH)) {
2778                 if (!(test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_WRITEBACK_DATA)) {
2779                         ext4_msg(sb, KERN_WARNING, "Ignoring nobh option - "
2780                                 "its supported only with writeback mode");
2781                         clear_opt(sbi->s_mount_opt, NOBH);
2782                 }
2783         }
2784         /*
2785          * The jbd2_journal_load will have done any necessary log recovery,
2786          * so we can safely mount the rest of the filesystem now.
2787          */
2788
2789         root = ext4_iget(sb, EXT4_ROOT_INO);
2790         if (IS_ERR(root)) {
2791                 ext4_msg(sb, KERN_ERR, "get root inode failed");
2792                 ret = PTR_ERR(root);
2793                 goto failed_mount4;
2794         }
2795         if (!S_ISDIR(root->i_mode) || !root->i_blocks || !root->i_size) {
2796                 iput(root);
2797                 ext4_msg(sb, KERN_ERR, "corrupt root inode, run e2fsck");
2798                 goto failed_mount4;
2799         }
2800         sb->s_root = d_alloc_root(root);
2801         if (!sb->s_root) {
2802                 ext4_msg(sb, KERN_ERR, "get root dentry failed");
2803                 iput(root);
2804                 ret = -ENOMEM;
2805                 goto failed_mount4;
2806         }
2807
2808         ext4_setup_super(sb, es, sb->s_flags & MS_RDONLY);
2809
2810         /* determine the minimum size of new large inodes, if present */
2811         if (sbi->s_inode_size > EXT4_GOOD_OLD_INODE_SIZE) {
2812                 sbi->s_want_extra_isize = sizeof(struct ext4_inode) -
2813                                                      EXT4_GOOD_OLD_INODE_SIZE;
2814                 if (EXT4_HAS_RO_COMPAT_FEATURE(sb,
2815                                        EXT4_FEATURE_RO_COMPAT_EXTRA_ISIZE)) {
2816                         if (sbi->s_want_extra_isize <
2817                             le16_to_cpu(es->s_want_extra_isize))
2818                                 sbi->s_want_extra_isize =
2819                                         le16_to_cpu(es->s_want_extra_isize);
2820                         if (sbi->s_want_extra_isize <
2821                             le16_to_cpu(es->s_min_extra_isize))
2822                                 sbi->s_want_extra_isize =
2823                                         le16_to_cpu(es->s_min_extra_isize);
2824                 }
2825         }
2826         /* Check if enough inode space is available */
2827         if (EXT4_GOOD_OLD_INODE_SIZE + sbi->s_want_extra_isize >
2828                                                         sbi->s_inode_size) {
2829                 sbi->s_want_extra_isize = sizeof(struct ext4_inode) -
2830                                                        EXT4_GOOD_OLD_INODE_SIZE;
2831                 ext4_msg(sb, KERN_INFO, "required extra inode space not"
2832                          "available");
2833         }
2834
2835         if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA) {
2836                 ext4_msg(sb, KERN_WARNING, "Ignoring delalloc option - "
2837                          "requested data journaling mode");
2838                 clear_opt(sbi->s_mount_opt, DELALLOC);
2839         } else if (test_opt(sb, DELALLOC))
2840                 ext4_msg(sb, KERN_INFO, "delayed allocation enabled");
2841
2842         err = ext4_setup_system_zone(sb);
2843         if (err) {
2844                 ext4_msg(sb, KERN_ERR, "failed to initialize system "
2845                          "zone (%d)\n", err);
2846                 goto failed_mount4;
2847         }
2848
2849         ext4_ext_init(sb);
2850         err = ext4_mb_init(sb, needs_recovery);
2851         if (err) {
2852                 ext4_msg(sb, KERN_ERR, "failed to initalize mballoc (%d)",
2853                          err);
2854                 goto failed_mount4;
2855         }
2856
2857         sbi->s_kobj.kset = ext4_kset;
2858         init_completion(&sbi->s_kobj_unregister);
2859         err = kobject_init_and_add(&sbi->s_kobj, &ext4_ktype, NULL,
2860                                    "%s", sb->s_id);
2861         if (err) {
2862                 ext4_mb_release(sb);
2863                 ext4_ext_release(sb);
2864                 goto failed_mount4;
2865         };
2866
2867         EXT4_SB(sb)->s_mount_state |= EXT4_ORPHAN_FS;
2868         ext4_orphan_cleanup(sb, es);
2869         EXT4_SB(sb)->s_mount_state &= ~EXT4_ORPHAN_FS;
2870         if (needs_recovery) {
2871                 ext4_msg(sb, KERN_INFO, "recovery complete");
2872                 ext4_mark_recovery_complete(sb, es);
2873         }
2874         if (EXT4_SB(sb)->s_journal) {
2875                 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA)
2876                         descr = " journalled data mode";
2877                 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA)
2878                         descr = " ordered data mode";
2879                 else
2880                         descr = " writeback data mode";
2881         } else
2882                 descr = "out journal";
2883
2884         ext4_msg(sb, KERN_INFO, "mounted filesystem with%s", descr);
2885
2886         lock_kernel();
2887         return 0;
2888
2889 cantfind_ext4:
2890         if (!silent)
2891                 ext4_msg(sb, KERN_ERR, "VFS: Can't find ext4 filesystem");
2892         goto failed_mount;
2893
2894 failed_mount4:
2895         ext4_msg(sb, KERN_ERR, "mount failed");
2896         ext4_release_system_zone(sb);
2897         if (sbi->s_journal) {
2898                 jbd2_journal_destroy(sbi->s_journal);
2899                 sbi->s_journal = NULL;
2900         }
2901 failed_mount3:
2902         if (sbi->s_flex_groups) {
2903                 if (is_vmalloc_addr(sbi->s_flex_groups))
2904                         vfree(sbi->s_flex_groups);
2905                 else
2906                         kfree(sbi->s_flex_groups);
2907         }
2908         percpu_counter_destroy(&sbi->s_freeblocks_counter);
2909         percpu_counter_destroy(&sbi->s_freeinodes_counter);
2910         percpu_counter_destroy(&sbi->s_dirs_counter);
2911         percpu_counter_destroy(&sbi->s_dirtyblocks_counter);
2912 failed_mount2:
2913         for (i = 0; i < db_count; i++)
2914                 brelse(sbi->s_group_desc[i]);
2915         kfree(sbi->s_group_desc);
2916 failed_mount:
2917         if (sbi->s_proc) {
2918                 remove_proc_entry(sb->s_id, ext4_proc_root);
2919         }
2920 #ifdef CONFIG_QUOTA
2921         for (i = 0; i < MAXQUOTAS; i++)
2922                 kfree(sbi->s_qf_names[i]);
2923 #endif
2924         ext4_blkdev_remove(sbi);
2925         brelse(bh);
2926 out_fail:
2927         sb->s_fs_info = NULL;
2928         kfree(sbi->s_blockgroup_lock);
2929         kfree(sbi);
2930         lock_kernel();
2931         return ret;
2932 }
2933
2934 /*
2935  * Setup any per-fs journal parameters now.  We'll do this both on
2936  * initial mount, once the journal has been initialised but before we've
2937  * done any recovery; and again on any subsequent remount.
2938  */
2939 static void ext4_init_journal_params(struct super_block *sb, journal_t *journal)
2940 {
2941         struct ext4_sb_info *sbi = EXT4_SB(sb);
2942
2943         journal->j_commit_interval = sbi->s_commit_interval;
2944         journal->j_min_batch_time = sbi->s_min_batch_time;
2945         journal->j_max_batch_time = sbi->s_max_batch_time;
2946
2947         spin_lock(&journal->j_state_lock);
2948         if (test_opt(sb, BARRIER))
2949                 journal->j_flags |= JBD2_BARRIER;
2950         else
2951                 journal->j_flags &= ~JBD2_BARRIER;
2952         if (test_opt(sb, DATA_ERR_ABORT))
2953                 journal->j_flags |= JBD2_ABORT_ON_SYNCDATA_ERR;
2954         else
2955                 journal->j_flags &= ~JBD2_ABORT_ON_SYNCDATA_ERR;
2956         spin_unlock(&journal->j_state_lock);
2957 }
2958
2959 static journal_t *ext4_get_journal(struct super_block *sb,
2960                                    unsigned int journal_inum)
2961 {
2962         struct inode *journal_inode;
2963         journal_t *journal;
2964
2965         BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL));
2966
2967         /* First, test for the existence of a valid inode on disk.  Bad
2968          * things happen if we iget() an unused inode, as the subsequent
2969          * iput() will try to delete it. */
2970
2971         journal_inode = ext4_iget(sb, journal_inum);
2972         if (IS_ERR(journal_inode)) {
2973                 ext4_msg(sb, KERN_ERR, "no journal found");
2974                 return NULL;
2975         }
2976         if (!journal_inode->i_nlink) {
2977                 make_bad_inode(journal_inode);
2978                 iput(journal_inode);
2979                 ext4_msg(sb, KERN_ERR, "journal inode is deleted");
2980                 return NULL;
2981         }
2982
2983         jbd_debug(2, "Journal inode found at %p: %lld bytes\n",
2984                   journal_inode, journal_inode->i_size);
2985         if (!S_ISREG(journal_inode->i_mode)) {
2986                 ext4_msg(sb, KERN_ERR, "invalid journal inode");
2987                 iput(journal_inode);
2988                 return NULL;
2989         }
2990
2991         journal = jbd2_journal_init_inode(journal_inode);
2992         if (!journal) {
2993                 ext4_msg(sb, KERN_ERR, "Could not load journal inode");
2994                 iput(journal_inode);
2995                 return NULL;
2996         }
2997         journal->j_private = sb;
2998         ext4_init_journal_params(sb, journal);
2999         return journal;
3000 }
3001
3002 static journal_t *ext4_get_dev_journal(struct super_block *sb,
3003                                        dev_t j_dev)
3004 {
3005         struct buffer_head *bh;
3006         journal_t *journal;
3007         ext4_fsblk_t start;
3008         ext4_fsblk_t len;
3009         int hblock, blocksize;
3010         ext4_fsblk_t sb_block;
3011         unsigned long offset;
3012         struct ext4_super_block *es;
3013         struct block_device *bdev;
3014
3015         BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL));
3016
3017         bdev = ext4_blkdev_get(j_dev, sb);
3018         if (bdev == NULL)
3019                 return NULL;
3020
3021         if (bd_claim(bdev, sb)) {
3022                 ext4_msg(sb, KERN_ERR,
3023                         "failed to claim external journal device");
3024                 blkdev_put(bdev, FMODE_READ|FMODE_WRITE);
3025                 return NULL;
3026         }
3027
3028         blocksize = sb->s_blocksize;
3029         hblock = bdev_logical_block_size(bdev);
3030         if (blocksize < hblock) {
3031                 ext4_msg(sb, KERN_ERR,
3032                         "blocksize too small for journal device");
3033                 goto out_bdev;
3034         }
3035
3036         sb_block = EXT4_MIN_BLOCK_SIZE / blocksize;
3037         offset = EXT4_MIN_BLOCK_SIZE % blocksize;
3038         set_blocksize(bdev, blocksize);
3039         if (!(bh = __bread(bdev, sb_block, blocksize))) {
3040                 ext4_msg(sb, KERN_ERR, "couldn't read superblock of "
3041                        "external journal");
3042                 goto out_bdev;
3043         }
3044
3045         es = (struct ext4_super_block *) (((char *)bh->b_data) + offset);
3046         if ((le16_to_cpu(es->s_magic) != EXT4_SUPER_MAGIC) ||
3047             !(le32_to_cpu(es->s_feature_incompat) &
3048               EXT4_FEATURE_INCOMPAT_JOURNAL_DEV)) {
3049                 ext4_msg(sb, KERN_ERR, "external journal has "
3050                                         "bad superblock");
3051                 brelse(bh);
3052                 goto out_bdev;
3053         }
3054
3055         if (memcmp(EXT4_SB(sb)->s_es->s_journal_uuid, es->s_uuid, 16)) {
3056                 ext4_msg(sb, KERN_ERR, "journal UUID does not match");
3057                 brelse(bh);
3058                 goto out_bdev;
3059         }
3060
3061         len = ext4_blocks_count(es);
3062         start = sb_block + 1;
3063         brelse(bh);     /* we're done with the superblock */
3064
3065         journal = jbd2_journal_init_dev(bdev, sb->s_bdev,
3066                                         start, len, blocksize);
3067         if (!journal) {
3068                 ext4_msg(sb, KERN_ERR, "failed to create device journal");
3069                 goto out_bdev;
3070         }
3071         journal->j_private = sb;
3072         ll_rw_block(READ, 1, &journal->j_sb_buffer);
3073         wait_on_buffer(journal->j_sb_buffer);
3074         if (!buffer_uptodate(journal->j_sb_buffer)) {
3075                 ext4_msg(sb, KERN_ERR, "I/O error on journal device");
3076                 goto out_journal;
3077         }
3078         if (be32_to_cpu(journal->j_superblock->s_nr_users) != 1) {
3079                 ext4_msg(sb, KERN_ERR, "External journal has more than one "
3080                                         "user (unsupported) - %d",
3081                         be32_to_cpu(journal->j_superblock->s_nr_users));
3082                 goto out_journal;
3083         }
3084         EXT4_SB(sb)->journal_bdev = bdev;
3085         ext4_init_journal_params(sb, journal);
3086         return journal;
3087
3088 out_journal:
3089         jbd2_journal_destroy(journal);
3090 out_bdev:
3091         ext4_blkdev_put(bdev);
3092         return NULL;
3093 }
3094
3095 static int ext4_load_journal(struct super_block *sb,
3096                              struct ext4_super_block *es,
3097                              unsigned long journal_devnum)
3098 {
3099         journal_t *journal;
3100         unsigned int journal_inum = le32_to_cpu(es->s_journal_inum);
3101         dev_t journal_dev;
3102         int err = 0;
3103         int really_read_only;
3104
3105         BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL));
3106
3107         if (journal_devnum &&
3108             journal_devnum != le32_to_cpu(es->s_journal_dev)) {
3109                 ext4_msg(sb, KERN_INFO, "external journal device major/minor "
3110                         "numbers have changed");
3111                 journal_dev = new_decode_dev(journal_devnum);
3112         } else
3113                 journal_dev = new_decode_dev(le32_to_cpu(es->s_journal_dev));
3114
3115         really_read_only = bdev_read_only(sb->s_bdev);
3116
3117         /*
3118          * Are we loading a blank journal or performing recovery after a
3119          * crash?  For recovery, we need to check in advance whether we
3120          * can get read-write access to the device.
3121          */
3122         if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER)) {
3123                 if (sb->s_flags & MS_RDONLY) {
3124                         ext4_msg(sb, KERN_INFO, "INFO: recovery "
3125                                         "required on readonly filesystem");
3126                         if (really_read_only) {
3127                                 ext4_msg(sb, KERN_ERR, "write access "
3128                                         "unavailable, cannot proceed");
3129                                 return -EROFS;
3130                         }
3131                         ext4_msg(sb, KERN_INFO, "write access will "
3132                                "be enabled during recovery");
3133                 }
3134         }
3135
3136         if (journal_inum && journal_dev) {
3137                 ext4_msg(sb, KERN_ERR, "filesystem has both journal "
3138                        "and inode journals!");
3139                 return -EINVAL;
3140         }
3141
3142         if (journal_inum) {
3143                 if (!(journal = ext4_get_journal(sb, journal_inum)))
3144                         return -EINVAL;
3145         } else {
3146                 if (!(journal = ext4_get_dev_journal(sb, journal_dev)))
3147                         return -EINVAL;
3148         }
3149
3150         if (journal->j_flags & JBD2_BARRIER)
3151                 ext4_msg(sb, KERN_INFO, "barriers enabled");
3152         else
3153                 ext4_msg(sb, KERN_INFO, "barriers disabled");
3154
3155         if (!really_read_only && test_opt(sb, UPDATE_JOURNAL)) {
3156                 err = jbd2_journal_update_format(journal);
3157                 if (err)  {
3158                         ext4_msg(sb, KERN_ERR, "error updating journal");
3159                         jbd2_journal_destroy(journal);
3160                         return err;
3161                 }
3162         }
3163
3164         if (!EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER))
3165                 err = jbd2_journal_wipe(journal, !really_read_only);
3166         if (!err)
3167                 err = jbd2_journal_load(journal);
3168
3169         if (err) {
3170                 ext4_msg(sb, KERN_ERR, "error loading journal");
3171                 jbd2_journal_destroy(journal);
3172                 return err;
3173         }
3174
3175         EXT4_SB(sb)->s_journal = journal;
3176         ext4_clear_journal_err(sb, es);
3177
3178         if (journal_devnum &&
3179             journal_devnum != le32_to_cpu(es->s_journal_dev)) {
3180                 es->s_journal_dev = cpu_to_le32(journal_devnum);
3181
3182                 /* Make sure we flush the recovery flag to disk. */
3183                 ext4_commit_super(sb, 1);
3184         }
3185
3186         return 0;
3187 }
3188
3189 static int ext4_commit_super(struct super_block *sb, int sync)
3190 {
3191         struct ext4_super_block *es = EXT4_SB(sb)->s_es;
3192         struct buffer_head *sbh = EXT4_SB(sb)->s_sbh;
3193         int error = 0;
3194
3195         if (!sbh)
3196                 return error;
3197         if (buffer_write_io_error(sbh)) {
3198                 /*
3199                  * Oh, dear.  A previous attempt to write the
3200                  * superblock failed.  This could happen because the
3201                  * USB device was yanked out.  Or it could happen to
3202                  * be a transient write error and maybe the block will
3203                  * be remapped.  Nothing we can do but to retry the
3204                  * write and hope for the best.
3205                  */
3206                 ext4_msg(sb, KERN_ERR, "previous I/O error to "
3207                        "superblock detected");
3208                 clear_buffer_write_io_error(sbh);
3209                 set_buffer_uptodate(sbh);
3210         }
3211         es->s_wtime = cpu_to_le32(get_seconds());
3212         es->s_kbytes_written =
3213                 cpu_to_le64(EXT4_SB(sb)->s_kbytes_written +
3214                             ((part_stat_read(sb->s_bdev->bd_part, sectors[1]) -
3215                               EXT4_SB(sb)->s_sectors_written_start) >> 1));
3216         ext4_free_blocks_count_set(es, percpu_counter_sum_positive(
3217                                         &EXT4_SB(sb)->s_freeblocks_counter));
3218         es->s_free_inodes_count = cpu_to_le32(percpu_counter_sum_positive(
3219                                         &EXT4_SB(sb)->s_freeinodes_counter));
3220         sb->s_dirt = 0;
3221         BUFFER_TRACE(sbh, "marking dirty");
3222         mark_buffer_dirty(sbh);
3223         if (sync) {
3224                 error = sync_dirty_buffer(sbh);
3225                 if (error)
3226                         return error;
3227
3228                 error = buffer_write_io_error(sbh);
3229                 if (error) {
3230                         ext4_msg(sb, KERN_ERR, "I/O error while writing "
3231                                "superblock");
3232                         clear_buffer_write_io_error(sbh);
3233                         set_buffer_uptodate(sbh);
3234                 }
3235         }
3236         return error;
3237 }
3238
3239 /*
3240  * Have we just finished recovery?  If so, and if we are mounting (or
3241  * remounting) the filesystem readonly, then we will end up with a
3242  * consistent fs on disk.  Record that fact.
3243  */
3244 static void ext4_mark_recovery_complete(struct super_block *sb,
3245                                         struct ext4_super_block *es)
3246 {
3247         journal_t *journal = EXT4_SB(sb)->s_journal;
3248
3249         if (!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL)) {
3250                 BUG_ON(journal != NULL);
3251                 return;
3252         }
3253         jbd2_journal_lock_updates(journal);
3254         if (jbd2_journal_flush(journal) < 0)
3255                 goto out;
3256
3257         if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER) &&
3258             sb->s_flags & MS_RDONLY) {
3259                 EXT4_CLEAR_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
3260                 ext4_commit_super(sb, 1);
3261         }
3262
3263 out:
3264         jbd2_journal_unlock_updates(journal);
3265 }
3266
3267 /*
3268  * If we are mounting (or read-write remounting) a filesystem whose journal
3269  * has recorded an error from a previous lifetime, move that error to the
3270  * main filesystem now.
3271  */
3272 static void ext4_clear_journal_err(struct super_block *sb,
3273                                    struct ext4_super_block *es)
3274 {
3275         journal_t *journal;
3276         int j_errno;
3277         const char *errstr;
3278
3279         BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL));
3280
3281         journal = EXT4_SB(sb)->s_journal;
3282
3283         /*
3284          * Now check for any error status which may have been recorded in the
3285          * journal by a prior ext4_error() or ext4_abort()
3286          */
3287
3288         j_errno = jbd2_journal_errno(journal);
3289         if (j_errno) {
3290                 char nbuf[16];
3291
3292                 errstr = ext4_decode_error(sb, j_errno, nbuf);
3293                 ext4_warning(sb, __func__, "Filesystem error recorded "
3294                              "from previous mount: %s", errstr);
3295                 ext4_warning(sb, __func__, "Marking fs in need of "
3296                              "filesystem check.");
3297
3298                 EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS;
3299                 es->s_state |= cpu_to_le16(EXT4_ERROR_FS);
3300                 ext4_commit_super(sb, 1);
3301
3302                 jbd2_journal_clear_err(journal);
3303         }
3304 }
3305
3306 /*
3307  * Force the running and committing transactions to commit,
3308  * and wait on the commit.
3309  */
3310 int ext4_force_commit(struct super_block *sb)
3311 {
3312         journal_t *journal;
3313         int ret = 0;
3314
3315         if (sb->s_flags & MS_RDONLY)
3316                 return 0;
3317
3318         journal = EXT4_SB(sb)->s_journal;
3319         if (journal)
3320                 ret = ext4_journal_force_commit(journal);
3321
3322         return ret;
3323 }
3324
3325 static void ext4_write_super(struct super_block *sb)
3326 {
3327         lock_super(sb);
3328         ext4_commit_super(sb, 1);
3329         unlock_super(sb);
3330 }
3331
3332 static int ext4_sync_fs(struct super_block *sb, int wait)
3333 {
3334         int ret = 0;
3335         tid_t target;
3336
3337         trace_ext4_sync_fs(sb, wait);
3338         if (jbd2_journal_start_commit(EXT4_SB(sb)->s_journal, &target)) {
3339                 if (wait)
3340                         jbd2_log_wait_commit(EXT4_SB(sb)->s_journal, target);
3341         }
3342         return ret;
3343 }
3344
3345 /*
3346  * LVM calls this function before a (read-only) snapshot is created.  This
3347  * gives us a chance to flush the journal completely and mark the fs clean.
3348  */
3349 static int ext4_freeze(struct super_block *sb)
3350 {
3351         int error = 0;
3352         journal_t *journal;
3353
3354         if (sb->s_flags & MS_RDONLY)
3355                 return 0;
3356
3357         journal = EXT4_SB(sb)->s_journal;
3358
3359         /* Now we set up the journal barrier. */
3360         jbd2_journal_lock_updates(journal);
3361
3362         /*
3363          * Don't clear the needs_recovery flag if we failed to flush
3364          * the journal.
3365          */
3366         error = jbd2_journal_flush(journal);
3367         if (error < 0) {
3368         out:
3369                 jbd2_journal_unlock_updates(journal);
3370                 return error;
3371         }
3372
3373         /* Journal blocked and flushed, clear needs_recovery flag. */
3374         EXT4_CLEAR_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
3375         error = ext4_commit_super(sb, 1);
3376         if (error)
3377                 goto out;
3378         return 0;
3379 }
3380
3381 /*
3382  * Called by LVM after the snapshot is done.  We need to reset the RECOVER
3383  * flag here, even though the filesystem is not technically dirty yet.
3384  */
3385 static int ext4_unfreeze(struct super_block *sb)
3386 {
3387         if (sb->s_flags & MS_RDONLY)
3388                 return 0;
3389
3390         lock_super(sb);
3391         /* Reset the needs_recovery flag before the fs is unlocked. */
3392         EXT4_SET_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
3393         ext4_commit_super(sb, 1);
3394         unlock_super(sb);
3395         jbd2_journal_unlock_updates(EXT4_SB(sb)->s_journal);
3396         return 0;
3397 }
3398
3399 static int ext4_remount(struct super_block *sb, int *flags, char *data)
3400 {
3401         struct ext4_super_block *es;
3402         struct ext4_sb_info *sbi = EXT4_SB(sb);
3403         ext4_fsblk_t n_blocks_count = 0;
3404         unsigned long old_sb_flags;
3405         struct ext4_mount_options old_opts;
3406         ext4_group_t g;
3407         unsigned int journal_ioprio = DEFAULT_JOURNAL_IOPRIO;
3408         int err;
3409 #ifdef CONFIG_QUOTA
3410         int i;
3411 #endif
3412
3413         lock_kernel();
3414
3415         /* Store the original options */
3416         lock_super(sb);
3417         old_sb_flags = sb->s_flags;
3418         old_opts.s_mount_opt = sbi->s_mount_opt;
3419         old_opts.s_resuid = sbi->s_resuid;
3420         old_opts.s_resgid = sbi->s_resgid;
3421         old_opts.s_commit_interval = sbi->s_commit_interval;
3422         old_opts.s_min_batch_time = sbi->s_min_batch_time;
3423         old_opts.s_max_batch_time = sbi->s_max_batch_time;
3424 #ifdef CONFIG_QUOTA
3425         old_opts.s_jquota_fmt = sbi->s_jquota_fmt;
3426         for (i = 0; i < MAXQUOTAS; i++)
3427                 old_opts.s_qf_names[i] = sbi->s_qf_names[i];
3428 #endif
3429         if (sbi->s_journal && sbi->s_journal->j_task->io_context)
3430                 journal_ioprio = sbi->s_journal->j_task->io_context->ioprio;
3431
3432         /*
3433          * Allow the "check" option to be passed as a remount option.
3434          */
3435         if (!parse_options(data, sb, NULL, &journal_ioprio,
3436                            &n_blocks_count, 1)) {
3437                 err = -EINVAL;
3438                 goto restore_opts;
3439         }
3440
3441         if (sbi->s_mount_flags & EXT4_MF_FS_ABORTED)
3442                 ext4_abort(sb, __func__, "Abort forced by user");
3443
3444         sb->s_flags = (sb->s_flags & ~MS_POSIXACL) |
3445                 ((sbi->s_mount_opt & EXT4_MOUNT_POSIX_ACL) ? MS_POSIXACL : 0);
3446
3447         es = sbi->s_es;
3448
3449         if (sbi->s_journal) {
3450                 ext4_init_journal_params(sb, sbi->s_journal);
3451                 set_task_ioprio(sbi->s_journal->j_task, journal_ioprio);
3452         }
3453
3454         if ((*flags & MS_RDONLY) != (sb->s_flags & MS_RDONLY) ||
3455                 n_blocks_count > ext4_blocks_count(es)) {
3456                 if (sbi->s_mount_flags & EXT4_MF_FS_ABORTED) {
3457                         err = -EROFS;
3458                         goto restore_opts;
3459                 }
3460
3461                 if (*flags & MS_RDONLY) {
3462                         /*
3463                          * First of all, the unconditional stuff we have to do
3464                          * to disable replay of the journal when we next remount
3465                          */
3466                         sb->s_flags |= MS_RDONLY;
3467
3468                         /*
3469                          * OK, test if we are remounting a valid rw partition
3470                          * readonly, and if so set the rdonly flag and then
3471                          * mark the partition as valid again.
3472                          */
3473                         if (!(es->s_state & cpu_to_le16(EXT4_VALID_FS)) &&
3474                             (sbi->s_mount_state & EXT4_VALID_FS))
3475                                 es->s_state = cpu_to_le16(sbi->s_mount_state);
3476
3477                         if (sbi->s_journal)
3478                                 ext4_mark_recovery_complete(sb, es);
3479                 } else {
3480                         int ret;
3481                         if ((ret = EXT4_HAS_RO_COMPAT_FEATURE(sb,
3482                                         ~EXT4_FEATURE_RO_COMPAT_SUPP))) {
3483                                 ext4_msg(sb, KERN_WARNING, "couldn't "
3484                                        "remount RDWR because of unsupported "
3485                                        "optional features (%x)",
3486                                 (le32_to_cpu(sbi->s_es->s_feature_ro_compat) &
3487                                         ~EXT4_FEATURE_RO_COMPAT_SUPP));
3488                                 err = -EROFS;
3489                                 goto restore_opts;
3490                         }
3491
3492                         /*
3493                          * Make sure the group descriptor checksums
3494                          * are sane.  If they aren't, refuse to remount r/w.
3495                          */
3496                         for (g = 0; g < sbi->s_groups_count; g++) {
3497                                 struct ext4_group_desc *gdp =
3498                                         ext4_get_group_desc(sb, g, NULL);
3499
3500                                 if (!ext4_group_desc_csum_verify(sbi, g, gdp)) {
3501                                         ext4_msg(sb, KERN_ERR,
3502                "ext4_remount: Checksum for group %u failed (%u!=%u)",
3503                 g, le16_to_cpu(ext4_group_desc_csum(sbi, g, gdp)),
3504                                                le16_to_cpu(gdp->bg_checksum));
3505                                         err = -EINVAL;
3506                                         goto restore_opts;
3507                                 }
3508                         }
3509
3510                         /*
3511                          * If we have an unprocessed orphan list hanging
3512                          * around from a previously readonly bdev mount,
3513                          * require a full umount/remount for now.
3514                          */
3515                         if (es->s_last_orphan) {
3516                                 ext4_msg(sb, KERN_WARNING, "Couldn't "
3517                                        "remount RDWR because of unprocessed "
3518                                        "orphan inode list.  Please "
3519                                        "umount/remount instead");
3520                                 err = -EINVAL;
3521                                 goto restore_opts;
3522                         }
3523
3524                         /*
3525                          * Mounting a RDONLY partition read-write, so reread
3526                          * and store the current valid flag.  (It may have
3527                          * been changed by e2fsck since we originally mounted
3528                          * the partition.)
3529                          */
3530                         if (sbi->s_journal)
3531                                 ext4_clear_journal_err(sb, es);
3532                         sbi->s_mount_state = le16_to_cpu(es->s_state);
3533                         if ((err = ext4_group_extend(sb, es, n_blocks_count)))
3534                                 goto restore_opts;
3535                         if (!ext4_setup_super(sb, es, 0))
3536                                 sb->s_flags &= ~MS_RDONLY;
3537                 }
3538         }
3539         ext4_setup_system_zone(sb);
3540         if (sbi->s_journal == NULL)
3541                 ext4_commit_super(sb, 1);
3542
3543 #ifdef CONFIG_QUOTA
3544         /* Release old quota file names */
3545         for (i = 0; i < MAXQUOTAS; i++)
3546                 if (old_opts.s_qf_names[i] &&
3547                     old_opts.s_qf_names[i] != sbi->s_qf_names[i])
3548                         kfree(old_opts.s_qf_names[i]);
3549 #endif
3550         unlock_super(sb);
3551         unlock_kernel();
3552         return 0;
3553
3554 restore_opts:
3555         sb->s_flags = old_sb_flags;
3556         sbi->s_mount_opt = old_opts.s_mount_opt;
3557         sbi->s_resuid = old_opts.s_resuid;
3558         sbi->s_resgid = old_opts.s_resgid;
3559         sbi->s_commit_interval = old_opts.s_commit_interval;
3560         sbi->s_min_batch_time = old_opts.s_min_batch_time;
3561         sbi->s_max_batch_time = old_opts.s_max_batch_time;
3562 #ifdef CONFIG_QUOTA
3563         sbi->s_jquota_fmt = old_opts.s_jquota_fmt;
3564         for (i = 0; i < MAXQUOTAS; i++) {
3565                 if (sbi->s_qf_names[i] &&
3566                     old_opts.s_qf_names[i] != sbi->s_qf_names[i])
3567                         kfree(sbi->s_qf_names[i]);
3568                 sbi->s_qf_names[i] = old_opts.s_qf_names[i];
3569         }
3570 #endif
3571         unlock_super(sb);
3572         unlock_kernel();
3573         return err;
3574 }
3575
3576 static int ext4_statfs(struct dentry *dentry, struct kstatfs *buf)
3577 {
3578         struct super_block *sb = dentry->d_sb;
3579         struct ext4_sb_info *sbi = EXT4_SB(sb);
3580         struct ext4_super_block *es = sbi->s_es;
3581         u64 fsid;
3582
3583         if (test_opt(sb, MINIX_DF)) {
3584                 sbi->s_overhead_last = 0;
3585         } else if (sbi->s_blocks_last != ext4_blocks_count(es)) {
3586                 ext4_group_t i, ngroups = ext4_get_groups_count(sb);
3587                 ext4_fsblk_t overhead = 0;
3588
3589                 /*
3590                  * Compute the overhead (FS structures).  This is constant
3591                  * for a given filesystem unless the number of block groups
3592                  * changes so we cache the previous value until it does.
3593                  */
3594
3595                 /*
3596                  * All of the blocks before first_data_block are
3597                  * overhead
3598                  */
3599                 overhead = le32_to_cpu(es->s_first_data_block);
3600
3601                 /*
3602                  * Add the overhead attributed to the superblock and
3603                  * block group descriptors.  If the sparse superblocks
3604                  * feature is turned on, then not all groups have this.
3605                  */
3606                 for (i = 0; i < ngroups; i++) {
3607                         overhead += ext4_bg_has_super(sb, i) +
3608                                 ext4_bg_num_gdb(sb, i);
3609                         cond_resched();
3610                 }
3611
3612                 /*
3613                  * Every block group has an inode bitmap, a block
3614                  * bitmap, and an inode table.
3615                  */
3616                 overhead += ngroups * (2 + sbi->s_itb_per_group);
3617                 sbi->s_overhead_last = overhead;
3618                 smp_wmb();
3619                 sbi->s_blocks_last = ext4_blocks_count(es);
3620         }
3621
3622         buf->f_type = EXT4_SUPER_MAGIC;
3623         buf->f_bsize = sb->s_blocksize;
3624         buf->f_blocks = ext4_blocks_count(es) - sbi->s_overhead_last;
3625         buf->f_bfree = percpu_counter_sum_positive(&sbi->s_freeblocks_counter) -
3626                        percpu_counter_sum_positive(&sbi->s_dirtyblocks_counter);
3627         ext4_free_blocks_count_set(es, buf->f_bfree);
3628         buf->f_bavail = buf->f_bfree - ext4_r_blocks_count(es);
3629         if (buf->f_bfree < ext4_r_blocks_count(es))
3630                 buf->f_bavail = 0;
3631         buf->f_files = le32_to_cpu(es->s_inodes_count);
3632         buf->f_ffree = percpu_counter_sum_positive(&sbi->s_freeinodes_counter);
3633         es->s_free_inodes_count = cpu_to_le32(buf->f_ffree);
3634         buf->f_namelen = EXT4_NAME_LEN;
3635         fsid = le64_to_cpup((void *)es->s_uuid) ^
3636                le64_to_cpup((void *)es->s_uuid + sizeof(u64));
3637         buf->f_fsid.val[0] = fsid & 0xFFFFFFFFUL;
3638         buf->f_fsid.val[1] = (fsid >> 32) & 0xFFFFFFFFUL;
3639
3640         return 0;
3641 }
3642
3643 /* Helper function for writing quotas on sync - we need to start transaction
3644  * before quota file is locked for write. Otherwise the are possible deadlocks:
3645  * Process 1                         Process 2
3646  * ext4_create()                     quota_sync()
3647  *   jbd2_journal_start()                  write_dquot()
3648  *   vfs_dq_init()                         down(dqio_mutex)
3649  *     down(dqio_mutex)                    jbd2_journal_start()
3650  *
3651  */
3652
3653 #ifdef CONFIG_QUOTA
3654
3655 static inline struct inode *dquot_to_inode(struct dquot *dquot)
3656 {
3657         return sb_dqopt(dquot->dq_sb)->files[dquot->dq_type];
3658 }
3659
3660 static int ext4_write_dquot(struct dquot *dquot)
3661 {
3662         int ret, err;
3663         handle_t *handle;
3664         struct inode *inode;
3665
3666         inode = dquot_to_inode(dquot);
3667         handle = ext4_journal_start(inode,
3668                                     EXT4_QUOTA_TRANS_BLOCKS(dquot->dq_sb));
3669         if (IS_ERR(handle))
3670                 return PTR_ERR(handle);
3671         ret = dquot_commit(dquot);
3672         err = ext4_journal_stop(handle);
3673         if (!ret)
3674                 ret = err;
3675         return ret;
3676 }
3677
3678 static int ext4_acquire_dquot(struct dquot *dquot)
3679 {
3680         int ret, err;
3681         handle_t *handle;
3682
3683         handle = ext4_journal_start(dquot_to_inode(dquot),
3684                                     EXT4_QUOTA_INIT_BLOCKS(dquot->dq_sb));
3685         if (IS_ERR(handle))
3686                 return PTR_ERR(handle);
3687         ret = dquot_acquire(dquot);
3688         err = ext4_journal_stop(handle);
3689         if (!ret)
3690                 ret = err;
3691         return ret;
3692 }
3693
3694 static int ext4_release_dquot(struct dquot *dquot)
3695 {
3696         int ret, err;
3697         handle_t *handle;
3698
3699         handle = ext4_journal_start(dquot_to_inode(dquot),
3700                                     EXT4_QUOTA_DEL_BLOCKS(dquot->dq_sb));
3701         if (IS_ERR(handle)) {
3702                 /* Release dquot anyway to avoid endless cycle in dqput() */
3703                 dquot_release(dquot);
3704                 return PTR_ERR(handle);
3705         }
3706         ret = dquot_release(dquot);
3707         err = ext4_journal_stop(handle);
3708         if (!ret)
3709                 ret = err;
3710         return ret;
3711 }
3712
3713 static int ext4_mark_dquot_dirty(struct dquot *dquot)
3714 {
3715         /* Are we journaling quotas? */
3716         if (EXT4_SB(dquot->dq_sb)->s_qf_names[USRQUOTA] ||
3717             EXT4_SB(dquot->dq_sb)->s_qf_names[GRPQUOTA]) {
3718                 dquot_mark_dquot_dirty(dquot);
3719                 return ext4_write_dquot(dquot);
3720         } else {
3721                 return dquot_mark_dquot_dirty(dquot);
3722         }
3723 }
3724
3725 static int ext4_write_info(struct super_block *sb, int type)
3726 {
3727         int ret, err;
3728         handle_t *handle;
3729
3730         /* Data block + inode block */
3731         handle = ext4_journal_start(sb->s_root->d_inode, 2);
3732         if (IS_ERR(handle))
3733                 return PTR_ERR(handle);
3734         ret = dquot_commit_info(sb, type);
3735         err = ext4_journal_stop(handle);
3736         if (!ret)
3737                 ret = err;
3738         return ret;
3739 }
3740
3741 /*
3742  * Turn on quotas during mount time - we need to find
3743  * the quota file and such...
3744  */
3745 static int ext4_quota_on_mount(struct super_block *sb, int type)
3746 {
3747         return vfs_quota_on_mount(sb, EXT4_SB(sb)->s_qf_names[type],
3748                                   EXT4_SB(sb)->s_jquota_fmt, type);
3749 }
3750
3751 /*
3752  * Standard function to be called on quota_on
3753  */
3754 static int ext4_quota_on(struct super_block *sb, int type, int format_id,
3755                          char *name, int remount)
3756 {
3757         int err;
3758         struct path path;
3759
3760         if (!test_opt(sb, QUOTA))
3761                 return -EINVAL;
3762         /* When remounting, no checks are needed and in fact, name is NULL */
3763         if (remount)
3764                 return vfs_quota_on(sb, type, format_id, name, remount);
3765
3766         err = kern_path(name, LOOKUP_FOLLOW, &path);
3767         if (err)
3768                 return err;
3769
3770         /* Quotafile not on the same filesystem? */
3771         if (path.mnt->mnt_sb != sb) {
3772                 path_put(&path);
3773                 return -EXDEV;
3774         }
3775         /* Journaling quota? */
3776         if (EXT4_SB(sb)->s_qf_names[type]) {
3777                 /* Quotafile not in fs root? */
3778                 if (path.dentry->d_parent != sb->s_root)
3779                         ext4_msg(sb, KERN_WARNING,
3780                                 "Quota file not on filesystem root. "
3781                                 "Journaled quota will not work");
3782         }
3783
3784         /*
3785          * When we journal data on quota file, we have to flush journal to see
3786          * all updates to the file when we bypass pagecache...
3787          */
3788         if (EXT4_SB(sb)->s_journal &&
3789             ext4_should_journal_data(path.dentry->d_inode)) {
3790                 /*
3791                  * We don't need to lock updates but journal_flush() could
3792                  * otherwise be livelocked...
3793                  */
3794                 jbd2_journal_lock_updates(EXT4_SB(sb)->s_journal);
3795                 err = jbd2_journal_flush(EXT4_SB(sb)->s_journal);
3796                 jbd2_journal_unlock_updates(EXT4_SB(sb)->s_journal);
3797                 if (err) {
3798                         path_put(&path);
3799                         return err;
3800                 }
3801         }
3802
3803         err = vfs_quota_on_path(sb, type, format_id, &path);
3804         path_put(&path);
3805         return err;
3806 }
3807
3808 /* Read data from quotafile - avoid pagecache and such because we cannot afford
3809  * acquiring the locks... As quota files are never truncated and quota code
3810  * itself serializes the operations (and noone else should touch the files)
3811  * we don't have to be afraid of races */
3812 static ssize_t ext4_quota_read(struct super_block *sb, int type, char *data,
3813                                size_t len, loff_t off)
3814 {
3815         struct inode *inode = sb_dqopt(sb)->files[type];
3816         ext4_lblk_t blk = off >> EXT4_BLOCK_SIZE_BITS(sb);
3817         int err = 0;
3818         int offset = off & (sb->s_blocksize - 1);
3819         int tocopy;
3820         size_t toread;
3821         struct buffer_head *bh;
3822         loff_t i_size = i_size_read(inode);
3823
3824         if (off > i_size)
3825                 return 0;
3826         if (off+len > i_size)
3827                 len = i_size-off;
3828         toread = len;
3829         while (toread > 0) {
3830                 tocopy = sb->s_blocksize - offset < toread ?
3831                                 sb->s_blocksize - offset : toread;
3832                 bh = ext4_bread(NULL, inode, blk, 0, &err);
3833                 if (err)
3834                         return err;
3835                 if (!bh)        /* A hole? */
3836                         memset(data, 0, tocopy);
3837                 else
3838                         memcpy(data, bh->b_data+offset, tocopy);
3839                 brelse(bh);
3840                 offset = 0;
3841                 toread -= tocopy;
3842                 data += tocopy;
3843                 blk++;
3844         }
3845         return len;
3846 }
3847
3848 /* Write to quotafile (we know the transaction is already started and has
3849  * enough credits) */
3850 static ssize_t ext4_quota_write(struct super_block *sb, int type,
3851                                 const char *data, size_t len, loff_t off)
3852 {
3853         struct inode *inode = sb_dqopt(sb)->files[type];
3854         ext4_lblk_t blk = off >> EXT4_BLOCK_SIZE_BITS(sb);
3855         int err = 0;
3856         int offset = off & (sb->s_blocksize - 1);
3857         int tocopy;
3858         int journal_quota = EXT4_SB(sb)->s_qf_names[type] != NULL;
3859         size_t towrite = len;
3860         struct buffer_head *bh;
3861         handle_t *handle = journal_current_handle();
3862
3863         if (EXT4_SB(sb)->s_journal && !handle) {
3864                 ext4_msg(sb, KERN_WARNING, "Quota write (off=%llu, len=%llu)"
3865                         " cancelled because transaction is not started",
3866                         (unsigned long long)off, (unsigned long long)len);
3867                 return -EIO;
3868         }
3869         mutex_lock_nested(&inode->i_mutex, I_MUTEX_QUOTA);
3870         while (towrite > 0) {
3871                 tocopy = sb->s_blocksize - offset < towrite ?
3872                                 sb->s_blocksize - offset : towrite;
3873                 bh = ext4_bread(handle, inode, blk, 1, &err);
3874                 if (!bh)
3875                         goto out;
3876                 if (journal_quota) {
3877                         err = ext4_journal_get_write_access(handle, bh);
3878                         if (err) {
3879                                 brelse(bh);
3880                                 goto out;
3881                         }
3882                 }
3883                 lock_buffer(bh);
3884                 memcpy(bh->b_data+offset, data, tocopy);
3885                 flush_dcache_page(bh->b_page);
3886                 unlock_buffer(bh);
3887                 if (journal_quota)
3888                         err = ext4_handle_dirty_metadata(handle, NULL, bh);
3889                 else {
3890                         /* Always do at least ordered writes for quotas */
3891                         err = ext4_jbd2_file_inode(handle, inode);
3892                         mark_buffer_dirty(bh);
3893                 }
3894                 brelse(bh);
3895                 if (err)
3896                         goto out;
3897                 offset = 0;
3898                 towrite -= tocopy;
3899                 data += tocopy;
3900                 blk++;
3901         }
3902 out:
3903         if (len == towrite) {
3904                 mutex_unlock(&inode->i_mutex);
3905                 return err;
3906         }
3907         if (inode->i_size < off+len-towrite) {
3908                 i_size_write(inode, off+len-towrite);
3909                 EXT4_I(inode)->i_disksize = inode->i_size;
3910         }
3911         inode->i_mtime = inode->i_ctime = CURRENT_TIME;
3912         ext4_mark_inode_dirty(handle, inode);
3913         mutex_unlock(&inode->i_mutex);
3914         return len - towrite;
3915 }
3916
3917 #endif
3918
3919 static int ext4_get_sb(struct file_system_type *fs_type, int flags,
3920                        const char *dev_name, void *data, struct vfsmount *mnt)
3921 {
3922         return get_sb_bdev(fs_type, flags, dev_name, data, ext4_fill_super,mnt);
3923 }
3924
3925 static struct file_system_type ext4_fs_type = {
3926         .owner          = THIS_MODULE,
3927         .name           = "ext4",
3928         .get_sb         = ext4_get_sb,
3929         .kill_sb        = kill_block_super,
3930         .fs_flags       = FS_REQUIRES_DEV,
3931 };
3932
3933 #ifdef CONFIG_EXT4DEV_COMPAT
3934 static int ext4dev_get_sb(struct file_system_type *fs_type, int flags,
3935                           const char *dev_name, void *data,struct vfsmount *mnt)
3936 {
3937         printk(KERN_WARNING "EXT4-fs (%s): Update your userspace programs "
3938                "to mount using ext4\n", dev_name);
3939         printk(KERN_WARNING "EXT4-fs (%s): ext4dev backwards compatibility "
3940                "will go away by 2.6.31\n", dev_name);
3941         return get_sb_bdev(fs_type, flags, dev_name, data, ext4_fill_super,mnt);
3942 }
3943
3944 static struct file_system_type ext4dev_fs_type = {
3945         .owner          = THIS_MODULE,
3946         .name           = "ext4dev",
3947         .get_sb         = ext4dev_get_sb,
3948         .kill_sb        = kill_block_super,
3949         .fs_flags       = FS_REQUIRES_DEV,
3950 };
3951 MODULE_ALIAS("ext4dev");
3952 #endif
3953
3954 static int __init init_ext4_fs(void)
3955 {
3956         int err;
3957
3958         err = init_ext4_system_zone();
3959         if (err)
3960                 return err;
3961         ext4_kset = kset_create_and_add("ext4", NULL, fs_kobj);
3962         if (!ext4_kset)
3963                 goto out4;
3964         ext4_proc_root = proc_mkdir("fs/ext4", NULL);
3965         err = init_ext4_mballoc();
3966         if (err)
3967                 goto out3;
3968
3969         err = init_ext4_xattr();
3970         if (err)
3971                 goto out2;
3972         err = init_inodecache();
3973         if (err)
3974                 goto out1;
3975         err = register_filesystem(&ext4_fs_type);
3976         if (err)
3977                 goto out;
3978 #ifdef CONFIG_EXT4DEV_COMPAT
3979         err = register_filesystem(&ext4dev_fs_type);
3980         if (err) {
3981                 unregister_filesystem(&ext4_fs_type);
3982                 goto out;
3983         }
3984 #endif
3985         return 0;
3986 out:
3987         destroy_inodecache();
3988 out1:
3989         exit_ext4_xattr();
3990 out2:
3991         exit_ext4_mballoc();
3992 out3:
3993         remove_proc_entry("fs/ext4", NULL);
3994         kset_unregister(ext4_kset);
3995 out4:
3996         exit_ext4_system_zone();
3997         return err;
3998 }
3999
4000 static void __exit exit_ext4_fs(void)
4001 {
4002         unregister_filesystem(&ext4_fs_type);
4003 #ifdef CONFIG_EXT4DEV_COMPAT
4004         unregister_filesystem(&ext4dev_fs_type);
4005 #endif
4006         destroy_inodecache();
4007         exit_ext4_xattr();
4008         exit_ext4_mballoc();
4009         remove_proc_entry("fs/ext4", NULL);
4010         kset_unregister(ext4_kset);
4011         exit_ext4_system_zone();
4012 }
4013
4014 MODULE_AUTHOR("Remy Card, Stephen Tweedie, Andrew Morton, Andreas Dilger, Theodore Ts'o and others");
4015 MODULE_DESCRIPTION("Fourth Extended Filesystem");
4016 MODULE_LICENSE("GPL");
4017 module_init(init_ext4_fs)
4018 module_exit(exit_ext4_fs)