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/buffer_head.h>
  30 #include <linux/exportfs.h>
  31 #include <linux/vfs.h>
  32 #include <linux/random.h>
  33 #include <linux/mount.h>
  34 #include <linux/namei.h>
  35 #include <linux/quotaops.h>
  36 #include <linux/seq_file.h>
  37 #include <linux/proc_fs.h>
  38 #include <linux/ctype.h>
  39 #include <linux/log2.h>
  40 #include <linux/crc16.h>
  41 #include <linux/cleancache.h>
  42 #include <asm/uaccess.h>
  43
  44 #include <linux/kthread.h>
  45 #include <linux/freezer.h>
  46
  47 #include "ext4.h"
  48 #include "ext4_extents.h"       /* Needed for trace points definition */
  49 #include "ext4_jbd2.h"
  50 #include "xattr.h"
  51 #include "acl.h"
  52 #include "mballoc.h"
  53
  54 #define CREATE_TRACE_POINTS
  55 #include <trace/events/ext4.h>
  56
  57 static struct proc_dir_entry *ext4_proc_root;
  58 static struct kset *ext4_kset;
  59 static struct ext4_lazy_init *ext4_li_info;
  60 static struct mutex ext4_li_mtx;
  61 static struct ext4_features *ext4_feat;
  62
  63 static int ext4_load_journal(struct super_block *, struct ext4_super_block *,
  64                              unsigned long journal_devnum);
  65 static int ext4_show_options(struct seq_file *seq, struct dentry *root);
  66 static int ext4_commit_super(struct super_block *sb, int sync);
  67 static void ext4_mark_recovery_complete(struct super_block *sb,
  68                                         struct ext4_super_block *es);
  69 static void ext4_clear_journal_err(struct super_block *sb,
  70                                    struct ext4_super_block *es);
  71 static int ext4_sync_fs(struct super_block *sb, int wait);
  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 int ext4_freeze(struct super_block *sb);
  76 static struct dentry *ext4_mount(struct file_system_type *fs_type, int flags,
  77                        const char *dev_name, void *data);
  78 static inline int ext2_feature_set_ok(struct super_block *sb);
  79 static inline int ext3_feature_set_ok(struct super_block *sb);
  80 static int ext4_feature_set_ok(struct super_block *sb, int readonly);
  81 static void ext4_destroy_lazyinit_thread(void);
  82 static void ext4_unregister_li_request(struct super_block *sb);
  83 static void ext4_clear_request_list(void);
  84
  85 #if !defined(CONFIG_EXT2_FS) && !defined(CONFIG_EXT2_FS_MODULE) && defined(CONFIG_EXT4_USE_FOR_EXT23)
  86 static struct file_system_type ext2_fs_type = {
  87         .owner          = THIS_MODULE,
  88         .name           = "ext2",
  89         .mount          = ext4_mount,
  90         .kill_sb        = kill_block_super,
  91         .fs_flags       = FS_REQUIRES_DEV,
  92 };
  93 MODULE_ALIAS_FS("ext2");
  94 #define IS_EXT2_SB(sb) ((sb)->s_bdev->bd_holder == &ext2_fs_type)
  95 #else
  96 #define IS_EXT2_SB(sb) (0)
  97 #endif
  98
  99
 100 #if !defined(CONFIG_EXT3_FS) && !defined(CONFIG_EXT3_FS_MODULE) && defined(CONFIG_EXT4_USE_FOR_EXT23)
 101 static struct file_system_type ext3_fs_type = {
 102         .owner          = THIS_MODULE,
 103         .name           = "ext3",
 104         .mount          = ext4_mount,
 105         .kill_sb        = kill_block_super,
 106         .fs_flags       = FS_REQUIRES_DEV,
 107 };
 108 MODULE_ALIAS_FS("ext3");
 109 #define IS_EXT3_SB(sb) ((sb)->s_bdev->bd_holder == &ext3_fs_type)
 110 #else
 111 #define IS_EXT3_SB(sb) (0)
 112 #endif
 113
 114 static int ext4_verify_csum_type(struct super_block *sb,
 115                                  struct ext4_super_block *es)
 116 {
 117         if (!EXT4_HAS_RO_COMPAT_FEATURE(sb,
 118                                         EXT4_FEATURE_RO_COMPAT_METADATA_CSUM))
 119                 return 1;
 120
 121         return es->s_checksum_type == EXT4_CRC32C_CHKSUM;
 122 }
 123
 124 static __le32 ext4_superblock_csum(struct super_block *sb,
 125                                    struct ext4_super_block *es)
 126 {
 127         struct ext4_sb_info *sbi = EXT4_SB(sb);
 128         int offset = offsetof(struct ext4_super_block, s_checksum);
 129         __u32 csum;
 130
 131         csum = ext4_chksum(sbi, ~0, (char *)es, offset);
 132
 133         return cpu_to_le32(csum);
 134 }
 135
 136 int ext4_superblock_csum_verify(struct super_block *sb,
 137                                 struct ext4_super_block *es)
 138 {
 139         if (!EXT4_HAS_RO_COMPAT_FEATURE(sb,
 140                                        EXT4_FEATURE_RO_COMPAT_METADATA_CSUM))
 141                 return 1;
 142
 143         return es->s_checksum == ext4_superblock_csum(sb, es);
 144 }
 145
 146 void ext4_superblock_csum_set(struct super_block *sb)
 147 {
 148         struct ext4_super_block *es = EXT4_SB(sb)->s_es;
 149
 150         if (!EXT4_HAS_RO_COMPAT_FEATURE(sb,
 151                 EXT4_FEATURE_RO_COMPAT_METADATA_CSUM))
 152                 return;
 153
 154         es->s_checksum = ext4_superblock_csum(sb, es);
 155 }
 156
 157 void *ext4_kvmalloc(size_t size, gfp_t flags)
 158 {
 159         void *ret;
 160
 161         ret = kmalloc(size, flags);
 162         if (!ret)
 163                 ret = __vmalloc(size, flags, PAGE_KERNEL);
 164         return ret;
 165 }
 166
 167 void *ext4_kvzalloc(size_t size, gfp_t flags)
 168 {
 169         void *ret;
 170
 171         ret = kzalloc(size, flags);
 172         if (!ret)
 173                 ret = __vmalloc(size, flags | __GFP_ZERO, PAGE_KERNEL);
 174         return ret;
 175 }
 176
 177 void ext4_kvfree(void *ptr)
 178 {
 179         if (is_vmalloc_addr(ptr))
 180                 vfree(ptr);
 181         else
 182                 kfree(ptr);
 183
 184 }
 185
 186 ext4_fsblk_t ext4_block_bitmap(struct super_block *sb,
 187                                struct ext4_group_desc *bg)
 188 {
 189         return le32_to_cpu(bg->bg_block_bitmap_lo) |
 190                 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
 191                  (ext4_fsblk_t)le32_to_cpu(bg->bg_block_bitmap_hi) << 32 : 0);
 192 }
 193
 194 ext4_fsblk_t ext4_inode_bitmap(struct super_block *sb,
 195                                struct ext4_group_desc *bg)
 196 {
 197         return le32_to_cpu(bg->bg_inode_bitmap_lo) |
 198                 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
 199                  (ext4_fsblk_t)le32_to_cpu(bg->bg_inode_bitmap_hi) << 32 : 0);
 200 }
 201
 202 ext4_fsblk_t ext4_inode_table(struct super_block *sb,
 203                               struct ext4_group_desc *bg)
 204 {
 205         return le32_to_cpu(bg->bg_inode_table_lo) |
 206                 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
 207                  (ext4_fsblk_t)le32_to_cpu(bg->bg_inode_table_hi) << 32 : 0);
 208 }
 209
 210 __u32 ext4_free_group_clusters(struct super_block *sb,
 211                                struct ext4_group_desc *bg)
 212 {
 213         return le16_to_cpu(bg->bg_free_blocks_count_lo) |
 214                 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
 215                  (__u32)le16_to_cpu(bg->bg_free_blocks_count_hi) << 16 : 0);
 216 }
 217
 218 __u32 ext4_free_inodes_count(struct super_block *sb,
 219                               struct ext4_group_desc *bg)
 220 {
 221         return le16_to_cpu(bg->bg_free_inodes_count_lo) |
 222                 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
 223                  (__u32)le16_to_cpu(bg->bg_free_inodes_count_hi) << 16 : 0);
 224 }
 225
 226 __u32 ext4_used_dirs_count(struct super_block *sb,
 227                               struct ext4_group_desc *bg)
 228 {
 229         return le16_to_cpu(bg->bg_used_dirs_count_lo) |
 230                 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
 231                  (__u32)le16_to_cpu(bg->bg_used_dirs_count_hi) << 16 : 0);
 232 }
 233
 234 __u32 ext4_itable_unused_count(struct super_block *sb,
 235                               struct ext4_group_desc *bg)
 236 {
 237         return le16_to_cpu(bg->bg_itable_unused_lo) |
 238                 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
 239                  (__u32)le16_to_cpu(bg->bg_itable_unused_hi) << 16 : 0);
 240 }
 241
 242 void ext4_block_bitmap_set(struct super_block *sb,
 243                            struct ext4_group_desc *bg, ext4_fsblk_t blk)
 244 {
 245         bg->bg_block_bitmap_lo = cpu_to_le32((u32)blk);
 246         if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
 247                 bg->bg_block_bitmap_hi = cpu_to_le32(blk >> 32);
 248 }
 249
 250 void ext4_inode_bitmap_set(struct super_block *sb,
 251                            struct ext4_group_desc *bg, ext4_fsblk_t blk)
 252 {
 253         bg->bg_inode_bitmap_lo  = cpu_to_le32((u32)blk);
 254         if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
 255                 bg->bg_inode_bitmap_hi = cpu_to_le32(blk >> 32);
 256 }
 257
 258 void ext4_inode_table_set(struct super_block *sb,
 259                           struct ext4_group_desc *bg, ext4_fsblk_t blk)
 260 {
 261         bg->bg_inode_table_lo = cpu_to_le32((u32)blk);
 262         if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
 263                 bg->bg_inode_table_hi = cpu_to_le32(blk >> 32);
 264 }
 265
 266 void ext4_free_group_clusters_set(struct super_block *sb,
 267                                   struct ext4_group_desc *bg, __u32 count)
 268 {
 269         bg->bg_free_blocks_count_lo = cpu_to_le16((__u16)count);
 270         if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
 271                 bg->bg_free_blocks_count_hi = cpu_to_le16(count >> 16);
 272 }
 273
 274 void ext4_free_inodes_set(struct super_block *sb,
 275                           struct ext4_group_desc *bg, __u32 count)
 276 {
 277         bg->bg_free_inodes_count_lo = cpu_to_le16((__u16)count);
 278         if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
 279                 bg->bg_free_inodes_count_hi = cpu_to_le16(count >> 16);
 280 }
 281
 282 void ext4_used_dirs_set(struct super_block *sb,
 283                           struct ext4_group_desc *bg, __u32 count)
 284 {
 285         bg->bg_used_dirs_count_lo = cpu_to_le16((__u16)count);
 286         if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
 287                 bg->bg_used_dirs_count_hi = cpu_to_le16(count >> 16);
 288 }
 289
 290 void ext4_itable_unused_set(struct super_block *sb,
 291                           struct ext4_group_desc *bg, __u32 count)
 292 {
 293         bg->bg_itable_unused_lo = cpu_to_le16((__u16)count);
 294         if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
 295                 bg->bg_itable_unused_hi = cpu_to_le16(count >> 16);
 296 }
 297
 298
 299 static void __save_error_info(struct super_block *sb, const char *func,
 300                             unsigned int line)
 301 {
 302         struct ext4_super_block *es = EXT4_SB(sb)->s_es;
 303
 304         EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS;
 305         es->s_state |= cpu_to_le16(EXT4_ERROR_FS);
 306         es->s_last_error_time = cpu_to_le32(get_seconds());
 307         strncpy(es->s_last_error_func, func, sizeof(es->s_last_error_func));
 308         es->s_last_error_line = cpu_to_le32(line);
 309         if (!es->s_first_error_time) {
 310                 es->s_first_error_time = es->s_last_error_time;
 311                 strncpy(es->s_first_error_func, func,
 312                         sizeof(es->s_first_error_func));
 313                 es->s_first_error_line = cpu_to_le32(line);
 314                 es->s_first_error_ino = es->s_last_error_ino;
 315                 es->s_first_error_block = es->s_last_error_block;
 316         }
 317         /*
 318          * Start the daily error reporting function if it hasn't been
 319          * started already
 320          */
 321         if (!es->s_error_count)
 322                 mod_timer(&EXT4_SB(sb)->s_err_report, jiffies + 24*60*60*HZ);
 323         le32_add_cpu(&es->s_error_count, 1);
 324 }
 325
 326 static void save_error_info(struct super_block *sb, const char *func,
 327                             unsigned int line)
 328 {
 329         __save_error_info(sb, func, line);
 330         ext4_commit_super(sb, 1);
 331 }
 332
 333 /*
 334  * The del_gendisk() function uninitializes the disk-specific data
 335  * structures, including the bdi structure, without telling anyone
 336  * else.  Once this happens, any attempt to call mark_buffer_dirty()
 337  * (for example, by ext4_commit_super), will cause a kernel OOPS.
 338  * This is a kludge to prevent these oops until we can put in a proper
 339  * hook in del_gendisk() to inform the VFS and file system layers.
 340  */
 341 static int block_device_ejected(struct super_block *sb)
 342 {
 343         struct inode *bd_inode = sb->s_bdev->bd_inode;
 344         struct backing_dev_info *bdi = bd_inode->i_mapping->backing_dev_info;
 345
 346         return bdi->dev == NULL;
 347 }
 348
 349 static void ext4_journal_commit_callback(journal_t *journal, transaction_t *txn)
 350 {
 351         struct super_block              *sb = journal->j_private;
 352         struct ext4_sb_info             *sbi = EXT4_SB(sb);
 353         int                             error = is_journal_aborted(journal);
 354         struct ext4_journal_cb_entry    *jce, *tmp;
 355
 356         spin_lock(&sbi->s_md_lock);
 357         list_for_each_entry_safe(jce, tmp, &txn->t_private_list, jce_list) {
 358                 list_del_init(&jce->jce_list);
 359                 spin_unlock(&sbi->s_md_lock);
 360                 jce->jce_func(sb, jce, error);
 361                 spin_lock(&sbi->s_md_lock);
 362         }
 363         spin_unlock(&sbi->s_md_lock);
 364 }
 365
 366 /* Deal with the reporting of failure conditions on a filesystem such as
 367  * inconsistencies detected or read IO failures.
 368  *
 369  * On ext2, we can store the error state of the filesystem in the
 370  * superblock.  That is not possible on ext4, because we may have other
 371  * write ordering constraints on the superblock which prevent us from
 372  * writing it out straight away; and given that the journal is about to
 373  * be aborted, we can't rely on the current, or future, transactions to
 374  * write out the superblock safely.
 375  *
 376  * We'll just use the jbd2_journal_abort() error code to record an error in
 377  * the journal instead.  On recovery, the journal will complain about
 378  * that error until we've noted it down and cleared it.
 379  */
 380
 381 static void ext4_handle_error(struct super_block *sb)
 382 {
 383         if (sb->s_flags & MS_RDONLY)
 384                 return;
 385
 386         if (!test_opt(sb, ERRORS_CONT)) {
 387                 journal_t *journal = EXT4_SB(sb)->s_journal;
 388
 389                 EXT4_SB(sb)->s_mount_flags |= EXT4_MF_FS_ABORTED;
 390                 if (journal)
 391                         jbd2_journal_abort(journal, -EIO);
 392         }
 393         if (test_opt(sb, ERRORS_RO)) {
 394                 ext4_msg(sb, KERN_CRIT, "Remounting filesystem read-only");
 395                 sb->s_flags |= MS_RDONLY;
 396         }
 397         if (test_opt(sb, ERRORS_PANIC))
 398                 panic("EXT4-fs (device %s): panic forced after error\n",
 399                         sb->s_id);
 400 }
 401
 402 void __ext4_error(struct super_block *sb, const char *function,
 403                   unsigned int line, const char *fmt, ...)
 404 {
 405         struct va_format vaf;
 406         va_list args;
 407
 408         va_start(args, fmt);
 409         vaf.fmt = fmt;
 410         vaf.va = &args;
 411         printk(KERN_CRIT "EXT4-fs error (device %s): %s:%d: comm %s: %pV\n",
 412                sb->s_id, function, line, current->comm, &vaf);
 413         va_end(args);
 414         save_error_info(sb, function, line);
 415
 416         ext4_handle_error(sb);
 417 }
 418
 419 void ext4_error_inode(struct inode *inode, const char *function,
 420                       unsigned int line, ext4_fsblk_t block,
 421                       const char *fmt, ...)
 422 {
 423         va_list args;
 424         struct va_format vaf;
 425         struct ext4_super_block *es = EXT4_SB(inode->i_sb)->s_es;
 426
 427         es->s_last_error_ino = cpu_to_le32(inode->i_ino);
 428         es->s_last_error_block = cpu_to_le64(block);
 429         save_error_info(inode->i_sb, function, line);
 430         va_start(args, fmt);
 431         vaf.fmt = fmt;
 432         vaf.va = &args;
 433         if (block)
 434                 printk(KERN_CRIT "EXT4-fs error (device %s): %s:%d: "
 435                        "inode #%lu: block %llu: comm %s: %pV\n",
 436                        inode->i_sb->s_id, function, line, inode->i_ino,
 437                        block, current->comm, &vaf);
 438         else
 439                 printk(KERN_CRIT "EXT4-fs error (device %s): %s:%d: "
 440                        "inode #%lu: comm %s: %pV\n",
 441                        inode->i_sb->s_id, function, line, inode->i_ino,
 442                        current->comm, &vaf);
 443         va_end(args);
 444
 445         ext4_handle_error(inode->i_sb);
 446 }
 447
 448 void ext4_error_file(struct file *file, const char *function,
 449                      unsigned int line, ext4_fsblk_t block,
 450                      const char *fmt, ...)
 451 {
 452         va_list args;
 453         struct va_format vaf;
 454         struct ext4_super_block *es;
 455         struct inode *inode = file_inode(file);
 456         char pathname[80], *path;
 457
 458         es = EXT4_SB(inode->i_sb)->s_es;
 459         es->s_last_error_ino = cpu_to_le32(inode->i_ino);
 460         save_error_info(inode->i_sb, function, line);
 461         path = d_path(&(file->f_path), pathname, sizeof(pathname));
 462         if (IS_ERR(path))
 463                 path = "(unknown)";
 464         va_start(args, fmt);
 465         vaf.fmt = fmt;
 466         vaf.va = &args;
 467         if (block)
 468                 printk(KERN_CRIT
 469                        "EXT4-fs error (device %s): %s:%d: inode #%lu: "
 470                        "block %llu: comm %s: path %s: %pV\n",
 471                        inode->i_sb->s_id, function, line, inode->i_ino,
 472                        block, current->comm, path, &vaf);
 473         else
 474                 printk(KERN_CRIT
 475                        "EXT4-fs error (device %s): %s:%d: inode #%lu: "
 476                        "comm %s: path %s: %pV\n",
 477                        inode->i_sb->s_id, function, line, inode->i_ino,
 478                        current->comm, path, &vaf);
 479         va_end(args);
 480
 481         ext4_handle_error(inode->i_sb);
 482 }
 483
 484 const char *ext4_decode_error(struct super_block *sb, int errno,
 485                               char nbuf[16])
 486 {
 487         char *errstr = NULL;
 488
 489         switch (errno) {
 490         case -EIO:
 491                 errstr = "IO failure";
 492                 break;
 493         case -ENOMEM:
 494                 errstr = "Out of memory";
 495                 break;
 496         case -EROFS:
 497                 if (!sb || (EXT4_SB(sb)->s_journal &&
 498                             EXT4_SB(sb)->s_journal->j_flags & JBD2_ABORT))
 499                         errstr = "Journal has aborted";
 500                 else
 501                         errstr = "Readonly filesystem";
 502                 break;
 503         default:
 504                 /* If the caller passed in an extra buffer for unknown
 505                  * errors, textualise them now.  Else we just return
 506                  * NULL. */
 507                 if (nbuf) {
 508                         /* Check for truncated error codes... */
 509                         if (snprintf(nbuf, 16, "error %d", -errno) >= 0)
 510                                 errstr = nbuf;
 511                 }
 512                 break;
 513         }
 514
 515         return errstr;
 516 }
 517
 518 /* __ext4_std_error decodes expected errors from journaling functions
 519  * automatically and invokes the appropriate error response.  */
 520
 521 void __ext4_std_error(struct super_block *sb, const char *function,
 522                       unsigned int line, int errno)
 523 {
 524         char nbuf[16];
 525         const char *errstr;
 526
 527         /* Special case: if the error is EROFS, and we're not already
 528          * inside a transaction, then there's really no point in logging
 529          * an error. */
 530         if (errno == -EROFS && journal_current_handle() == NULL &&
 531             (sb->s_flags & MS_RDONLY))
 532                 return;
 533
 534         errstr = ext4_decode_error(sb, errno, nbuf);
 535         printk(KERN_CRIT "EXT4-fs error (device %s) in %s:%d: %s\n",
 536                sb->s_id, function, line, errstr);
 537         save_error_info(sb, function, line);
 538
 539         ext4_handle_error(sb);
 540 }
 541
 542 /*
 543  * ext4_abort is a much stronger failure handler than ext4_error.  The
 544  * abort function may be used to deal with unrecoverable failures such
 545  * as journal IO errors or ENOMEM at a critical moment in log management.
 546  *
 547  * We unconditionally force the filesystem into an ABORT|READONLY state,
 548  * unless the error response on the fs has been set to panic in which
 549  * case we take the easy way out and panic immediately.
 550  */
 551
 552 void __ext4_abort(struct super_block *sb, const char *function,
 553                 unsigned int line, const char *fmt, ...)
 554 {
 555         va_list args;
 556
 557         save_error_info(sb, function, line);
 558         va_start(args, fmt);
 559         printk(KERN_CRIT "EXT4-fs error (device %s): %s:%d: ", sb->s_id,
 560                function, line);
 561         vprintk(fmt, args);
 562         printk("\n");
 563         va_end(args);
 564
 565         if ((sb->s_flags & MS_RDONLY) == 0) {
 566                 ext4_msg(sb, KERN_CRIT, "Remounting filesystem read-only");
 567                 sb->s_flags |= MS_RDONLY;
 568                 EXT4_SB(sb)->s_mount_flags |= EXT4_MF_FS_ABORTED;
 569                 if (EXT4_SB(sb)->s_journal)
 570                         jbd2_journal_abort(EXT4_SB(sb)->s_journal, -EIO);
 571                 save_error_info(sb, function, line);
 572         }
 573         if (test_opt(sb, ERRORS_PANIC))
 574                 panic("EXT4-fs panic from previous error\n");
 575 }
 576
 577 void ext4_msg(struct super_block *sb, const char *prefix, const char *fmt, ...)
 578 {
 579         struct va_format vaf;
 580         va_list args;
 581
 582         va_start(args, fmt);
 583         vaf.fmt = fmt;
 584         vaf.va = &args;
 585         printk("%sEXT4-fs (%s): %pV\n", prefix, sb->s_id, &vaf);
 586         va_end(args);
 587 }
 588
 589 void __ext4_warning(struct super_block *sb, const char *function,
 590                     unsigned int line, const char *fmt, ...)
 591 {
 592         struct va_format vaf;
 593         va_list args;
 594
 595         va_start(args, fmt);
 596         vaf.fmt = fmt;
 597         vaf.va = &args;
 598         printk(KERN_WARNING "EXT4-fs warning (device %s): %s:%d: %pV\n",
 599                sb->s_id, function, line, &vaf);
 600         va_end(args);
 601 }
 602
 603 void __ext4_grp_locked_error(const char *function, unsigned int line,
 604                              struct super_block *sb, ext4_group_t grp,
 605                              unsigned long ino, ext4_fsblk_t block,
 606                              const char *fmt, ...)
 607 __releases(bitlock)
 608 __acquires(bitlock)
 609 {
 610         struct va_format vaf;
 611         va_list args;
 612         struct ext4_super_block *es = EXT4_SB(sb)->s_es;
 613
 614         es->s_last_error_ino = cpu_to_le32(ino);
 615         es->s_last_error_block = cpu_to_le64(block);
 616         __save_error_info(sb, function, line);
 617
 618         va_start(args, fmt);
 619
 620         vaf.fmt = fmt;
 621         vaf.va = &args;
 622         printk(KERN_CRIT "EXT4-fs error (device %s): %s:%d: group %u, ",
 623                sb->s_id, function, line, grp);
 624         if (ino)
 625                 printk(KERN_CONT "inode %lu: ", ino);
 626         if (block)
 627                 printk(KERN_CONT "block %llu:", (unsigned long long) block);
 628         printk(KERN_CONT "%pV\n", &vaf);
 629         va_end(args);
 630
 631         if (test_opt(sb, ERRORS_CONT)) {
 632                 ext4_commit_super(sb, 0);
 633                 return;
 634         }
 635
 636         ext4_unlock_group(sb, grp);
 637         ext4_handle_error(sb);
 638         /*
 639          * We only get here in the ERRORS_RO case; relocking the group
 640          * may be dangerous, but nothing bad will happen since the
 641          * filesystem will have already been marked read/only and the
 642          * journal has been aborted.  We return 1 as a hint to callers
 643          * who might what to use the return value from
 644          * ext4_grp_locked_error() to distinguish between the
 645          * ERRORS_CONT and ERRORS_RO case, and perhaps return more
 646          * aggressively from the ext4 function in question, with a
 647          * more appropriate error code.
 648          */
 649         ext4_lock_group(sb, grp);
 650         return;
 651 }
 652
 653 void ext4_update_dynamic_rev(struct super_block *sb)
 654 {
 655         struct ext4_super_block *es = EXT4_SB(sb)->s_es;
 656
 657         if (le32_to_cpu(es->s_rev_level) > EXT4_GOOD_OLD_REV)
 658                 return;
 659
 660         ext4_warning(sb,
 661                      "updating to rev %d because of new feature flag, "
 662                      "running e2fsck is recommended",
 663                      EXT4_DYNAMIC_REV);
 664
 665         es->s_first_ino = cpu_to_le32(EXT4_GOOD_OLD_FIRST_INO);
 666         es->s_inode_size = cpu_to_le16(EXT4_GOOD_OLD_INODE_SIZE);
 667         es->s_rev_level = cpu_to_le32(EXT4_DYNAMIC_REV);
 668         /* leave es->s_feature_*compat flags alone */
 669         /* es->s_uuid will be set by e2fsck if empty */
 670
 671         /*
 672          * The rest of the superblock fields should be zero, and if not it
 673          * means they are likely already in use, so leave them alone.  We
 674          * can leave it up to e2fsck to clean up any inconsistencies there.
 675          */
 676 }
 677
 678 /*
 679  * Open the external journal device
 680  */
 681 static struct block_device *ext4_blkdev_get(dev_t dev, struct super_block *sb)
 682 {
 683         struct block_device *bdev;
 684         char b[BDEVNAME_SIZE];
 685
 686         bdev = blkdev_get_by_dev(dev, FMODE_READ|FMODE_WRITE|FMODE_EXCL, sb);
 687         if (IS_ERR(bdev))
 688                 goto fail;
 689         return bdev;
 690
 691 fail:
 692         ext4_msg(sb, KERN_ERR, "failed to open journal device %s: %ld",
 693                         __bdevname(dev, b), PTR_ERR(bdev));
 694         return NULL;
 695 }
 696
 697 /*
 698  * Release the journal device
 699  */
 700 static int ext4_blkdev_put(struct block_device *bdev)
 701 {
 702         return blkdev_put(bdev, FMODE_READ|FMODE_WRITE|FMODE_EXCL);
 703 }
 704
 705 static int ext4_blkdev_remove(struct ext4_sb_info *sbi)
 706 {
 707         struct block_device *bdev;
 708         int ret = -ENODEV;
 709
 710         bdev = sbi->journal_bdev;
 711         if (bdev) {
 712                 ret = ext4_blkdev_put(bdev);
 713                 sbi->journal_bdev = NULL;
 714         }
 715         return ret;
 716 }
 717
 718 static inline struct inode *orphan_list_entry(struct list_head *l)
 719 {
 720         return &list_entry(l, struct ext4_inode_info, i_orphan)->vfs_inode;
 721 }
 722
 723 static void dump_orphan_list(struct super_block *sb, struct ext4_sb_info *sbi)
 724 {
 725         struct list_head *l;
 726
 727         ext4_msg(sb, KERN_ERR, "sb orphan head is %d",
 728                  le32_to_cpu(sbi->s_es->s_last_orphan));
 729
 730         printk(KERN_ERR "sb_info orphan list:\n");
 731         list_for_each(l, &sbi->s_orphan) {
 732                 struct inode *inode = orphan_list_entry(l);
 733                 printk(KERN_ERR "  "
 734                        "inode %s:%lu at %p: mode %o, nlink %d, next %d\n",
 735                        inode->i_sb->s_id, inode->i_ino, inode,
 736                        inode->i_mode, inode->i_nlink,
 737                        NEXT_ORPHAN(inode));
 738         }
 739 }
 740
 741 static void ext4_put_super(struct super_block *sb)
 742 {
 743         struct ext4_sb_info *sbi = EXT4_SB(sb);
 744         struct ext4_super_block *es = sbi->s_es;
 745         int i, err;
 746
 747         ext4_unregister_li_request(sb);
 748         dquot_disable(sb, -1, DQUOT_USAGE_ENABLED | DQUOT_LIMITS_ENABLED);
 749
 750         flush_workqueue(sbi->dio_unwritten_wq);
 751         destroy_workqueue(sbi->dio_unwritten_wq);
 752
 753         if (sbi->s_journal) {
 754                 err = jbd2_journal_destroy(sbi->s_journal);
 755                 sbi->s_journal = NULL;
 756                 if (err < 0)
 757                         ext4_abort(sb, "Couldn't clean up the journal");
 758         }
 759
 760         ext4_es_unregister_shrinker(sb);
 761         del_timer(&sbi->s_err_report);
 762         ext4_release_system_zone(sb);
 763         ext4_mb_release(sb);
 764         ext4_ext_release(sb);
 765         ext4_xattr_put_super(sb);
 766
 767         if (!(sb->s_flags & MS_RDONLY)) {
 768                 EXT4_CLEAR_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
 769                 es->s_state = cpu_to_le16(sbi->s_mount_state);
 770         }
 771         if (!(sb->s_flags & MS_RDONLY))
 772                 ext4_commit_super(sb, 1);
 773
 774         if (sbi->s_proc) {
 775                 remove_proc_entry("options", sbi->s_proc);
 776                 remove_proc_entry(sb->s_id, ext4_proc_root);
 777         }
 778         kobject_del(&sbi->s_kobj);
 779
 780         for (i = 0; i < sbi->s_gdb_count; i++)
 781                 brelse(sbi->s_group_desc[i]);
 782         ext4_kvfree(sbi->s_group_desc);
 783         ext4_kvfree(sbi->s_flex_groups);
 784         percpu_counter_destroy(&sbi->s_freeclusters_counter);
 785         percpu_counter_destroy(&sbi->s_freeinodes_counter);
 786         percpu_counter_destroy(&sbi->s_dirs_counter);
 787         percpu_counter_destroy(&sbi->s_dirtyclusters_counter);
 788         percpu_counter_destroy(&sbi->s_extent_cache_cnt);
 789         brelse(sbi->s_sbh);
 790 #ifdef CONFIG_QUOTA
 791         for (i = 0; i < MAXQUOTAS; i++)
 792                 kfree(sbi->s_qf_names[i]);
 793 #endif
 794
 795         /* Debugging code just in case the in-memory inode orphan list
 796          * isn't empty.  The on-disk one can be non-empty if we've
 797          * detected an error and taken the fs readonly, but the
 798          * in-memory list had better be clean by this point. */
 799         if (!list_empty(&sbi->s_orphan))
 800                 dump_orphan_list(sb, sbi);
 801         J_ASSERT(list_empty(&sbi->s_orphan));
 802
 803         invalidate_bdev(sb->s_bdev);
 804         if (sbi->journal_bdev && sbi->journal_bdev != sb->s_bdev) {
 805                 /*
 806                  * Invalidate the journal device's buffers.  We don't want them
 807                  * floating about in memory - the physical journal device may
 808                  * hotswapped, and it breaks the `ro-after' testing code.
 809                  */
 810                 sync_blockdev(sbi->journal_bdev);
 811                 invalidate_bdev(sbi->journal_bdev);
 812                 ext4_blkdev_remove(sbi);
 813         }
 814         if (sbi->s_mmp_tsk)
 815                 kthread_stop(sbi->s_mmp_tsk);
 816         sb->s_fs_info = NULL;
 817         /*
 818          * Now that we are completely done shutting down the
 819          * superblock, we need to actually destroy the kobject.
 820          */
 821         kobject_put(&sbi->s_kobj);
 822         wait_for_completion(&sbi->s_kobj_unregister);
 823         if (sbi->s_chksum_driver)
 824                 crypto_free_shash(sbi->s_chksum_driver);
 825         kfree(sbi->s_blockgroup_lock);
 826         kfree(sbi);
 827 }
 828
 829 static struct kmem_cache *ext4_inode_cachep;
 830
 831 /*
 832  * Called inside transaction, so use GFP_NOFS
 833  */
 834 static struct inode *ext4_alloc_inode(struct super_block *sb)
 835 {
 836         struct ext4_inode_info *ei;
 837
 838         ei = kmem_cache_alloc(ext4_inode_cachep, GFP_NOFS);
 839         if (!ei)
 840                 return NULL;
 841
 842         ei->vfs_inode.i_version = 1;
 843         INIT_LIST_HEAD(&ei->i_prealloc_list);
 844         spin_lock_init(&ei->i_prealloc_lock);
 845         ext4_es_init_tree(&ei->i_es_tree);
 846         rwlock_init(&ei->i_es_lock);
 847         INIT_LIST_HEAD(&ei->i_es_lru);
 848         ei->i_es_lru_nr = 0;
 849         ei->i_reserved_data_blocks = 0;
 850         ei->i_reserved_meta_blocks = 0;
 851         ei->i_allocated_meta_blocks = 0;
 852         ei->i_da_metadata_calc_len = 0;
 853         ei->i_da_metadata_calc_last_lblock = 0;
 854         spin_lock_init(&(ei->i_block_reservation_lock));
 855 #ifdef CONFIG_QUOTA
 856         ei->i_reserved_quota = 0;
 857 #endif
 858         ei->jinode = NULL;
 859         INIT_LIST_HEAD(&ei->i_completed_io_list);
 860         spin_lock_init(&ei->i_completed_io_lock);
 861         ei->i_sync_tid = 0;
 862         ei->i_datasync_tid = 0;
 863         atomic_set(&ei->i_ioend_count, 0);
 864         atomic_set(&ei->i_unwritten, 0);
 865         INIT_WORK(&ei->i_unwritten_work, ext4_end_io_work);
 866
 867         return &ei->vfs_inode;
 868 }
 869
 870 static int ext4_drop_inode(struct inode *inode)
 871 {
 872         int drop = generic_drop_inode(inode);
 873
 874         trace_ext4_drop_inode(inode, drop);
 875         return drop;
 876 }
 877
 878 static void ext4_i_callback(struct rcu_head *head)
 879 {
 880         struct inode *inode = container_of(head, struct inode, i_rcu);
 881         kmem_cache_free(ext4_inode_cachep, EXT4_I(inode));
 882 }
 883
 884 static void ext4_destroy_inode(struct inode *inode)
 885 {
 886         if (!list_empty(&(EXT4_I(inode)->i_orphan))) {
 887                 ext4_msg(inode->i_sb, KERN_ERR,
 888                          "Inode %lu (%p): orphan list check failed!",
 889                          inode->i_ino, EXT4_I(inode));
 890                 print_hex_dump(KERN_INFO, "", DUMP_PREFIX_ADDRESS, 16, 4,
 891                                 EXT4_I(inode), sizeof(struct ext4_inode_info),
 892                                 true);
 893                 dump_stack();
 894         }
 895         call_rcu(&inode->i_rcu, ext4_i_callback);
 896 }
 897
 898 static void init_once(void *foo)
 899 {
 900         struct ext4_inode_info *ei = (struct ext4_inode_info *) foo;
 901
 902         INIT_LIST_HEAD(&ei->i_orphan);
 903         init_rwsem(&ei->xattr_sem);
 904         init_rwsem(&ei->i_data_sem);
 905         inode_init_once(&ei->vfs_inode);
 906 }
 907
 908 static int init_inodecache(void)
 909 {
 910         ext4_inode_cachep = kmem_cache_create("ext4_inode_cache",
 911                                              sizeof(struct ext4_inode_info),
 912                                              0, (SLAB_RECLAIM_ACCOUNT|
 913                                                 SLAB_MEM_SPREAD),
 914                                              init_once);
 915         if (ext4_inode_cachep == NULL)
 916                 return -ENOMEM;
 917         return 0;
 918 }
 919
 920 static void destroy_inodecache(void)
 921 {
 922         /*
 923          * Make sure all delayed rcu free inodes are flushed before we
 924          * destroy cache.
 925          */
 926         rcu_barrier();
 927         kmem_cache_destroy(ext4_inode_cachep);
 928 }
 929
 930 void ext4_clear_inode(struct inode *inode)
 931 {
 932         invalidate_inode_buffers(inode);
 933         clear_inode(inode);
 934         dquot_drop(inode);
 935         ext4_discard_preallocations(inode);
 936         ext4_es_remove_extent(inode, 0, EXT_MAX_BLOCKS);
 937         ext4_es_lru_del(inode);
 938         if (EXT4_I(inode)->jinode) {
 939                 jbd2_journal_release_jbd_inode(EXT4_JOURNAL(inode),
 940                                                EXT4_I(inode)->jinode);
 941                 jbd2_free_inode(EXT4_I(inode)->jinode);
 942                 EXT4_I(inode)->jinode = NULL;
 943         }
 944 }
 945
 946 static struct inode *ext4_nfs_get_inode(struct super_block *sb,
 947                                         u64 ino, u32 generation)
 948 {
 949         struct inode *inode;
 950
 951         if (ino < EXT4_FIRST_INO(sb) && ino != EXT4_ROOT_INO)
 952                 return ERR_PTR(-ESTALE);
 953         if (ino > le32_to_cpu(EXT4_SB(sb)->s_es->s_inodes_count))
 954                 return ERR_PTR(-ESTALE);
 955
 956         /* iget isn't really right if the inode is currently unallocated!!
 957          *
 958          * ext4_read_inode will return a bad_inode if the inode had been
 959          * deleted, so we should be safe.
 960          *
 961          * Currently we don't know the generation for parent directory, so
 962          * a generation of 0 means "accept any"
 963          */
 964         inode = ext4_iget(sb, ino);
 965         if (IS_ERR(inode))
 966                 return ERR_CAST(inode);
 967         if (generation && inode->i_generation != generation) {
 968                 iput(inode);
 969                 return ERR_PTR(-ESTALE);
 970         }
 971
 972         return inode;
 973 }
 974
 975 static struct dentry *ext4_fh_to_dentry(struct super_block *sb, struct fid *fid,
 976                                         int fh_len, int fh_type)
 977 {
 978         return generic_fh_to_dentry(sb, fid, fh_len, fh_type,
 979                                     ext4_nfs_get_inode);
 980 }
 981
 982 static struct dentry *ext4_fh_to_parent(struct super_block *sb, struct fid *fid,
 983                                         int fh_len, int fh_type)
 984 {
 985         return generic_fh_to_parent(sb, fid, fh_len, fh_type,
 986                                     ext4_nfs_get_inode);
 987 }
 988
 989 /*
 990  * Try to release metadata pages (indirect blocks, directories) which are
 991  * mapped via the block device.  Since these pages could have journal heads
 992  * which would prevent try_to_free_buffers() from freeing them, we must use
 993  * jbd2 layer's try_to_free_buffers() function to release them.
 994  */
 995 static int bdev_try_to_free_page(struct super_block *sb, struct page *page,
 996                                  gfp_t wait)
 997 {
 998         journal_t *journal = EXT4_SB(sb)->s_journal;
 999
1000         WARN_ON(PageChecked(page));
1001         if (!page_has_buffers(page))
1002                 return 0;
1003         if (journal)
1004                 return jbd2_journal_try_to_free_buffers(journal, page,
1005                                                         wait & ~__GFP_WAIT);
1006         return try_to_free_buffers(page);
1007 }
1008
1009 #ifdef CONFIG_QUOTA
1010 #define QTYPE2NAME(t) ((t) == USRQUOTA ? "user" : "group")
1011 #define QTYPE2MOPT(on, t) ((t) == USRQUOTA?((on)##USRJQUOTA):((on)##GRPJQUOTA))
1012
1013 static int ext4_write_dquot(struct dquot *dquot);
1014 static int ext4_acquire_dquot(struct dquot *dquot);
1015 static int ext4_release_dquot(struct dquot *dquot);
1016 static int ext4_mark_dquot_dirty(struct dquot *dquot);
1017 static int ext4_write_info(struct super_block *sb, int type);
1018 static int ext4_quota_on(struct super_block *sb, int type, int format_id,
1019                          struct path *path);
1020 static int ext4_quota_on_sysfile(struct super_block *sb, int type,
1021                                  int format_id);
1022 static int ext4_quota_off(struct super_block *sb, int type);
1023 static int ext4_quota_off_sysfile(struct super_block *sb, int type);
1024 static int ext4_quota_on_mount(struct super_block *sb, int type);
1025 static ssize_t ext4_quota_read(struct super_block *sb, int type, char *data,
1026                                size_t len, loff_t off);
1027 static ssize_t ext4_quota_write(struct super_block *sb, int type,
1028                                 const char *data, size_t len, loff_t off);
1029 static int ext4_quota_enable(struct super_block *sb, int type, int format_id,
1030                              unsigned int flags);
1031 static int ext4_enable_quotas(struct super_block *sb);
1032
1033 static const struct dquot_operations ext4_quota_operations = {
1034         .get_reserved_space = ext4_get_reserved_space,
1035         .write_dquot    = ext4_write_dquot,
1036         .acquire_dquot  = ext4_acquire_dquot,
1037         .release_dquot  = ext4_release_dquot,
1038         .mark_dirty     = ext4_mark_dquot_dirty,
1039         .write_info     = ext4_write_info,
1040         .alloc_dquot    = dquot_alloc,
1041         .destroy_dquot  = dquot_destroy,
1042 };
1043
1044 static const struct quotactl_ops ext4_qctl_operations = {
1045         .quota_on       = ext4_quota_on,
1046         .quota_off      = ext4_quota_off,
1047         .quota_sync     = dquot_quota_sync,
1048         .get_info       = dquot_get_dqinfo,
1049         .set_info       = dquot_set_dqinfo,
1050         .get_dqblk      = dquot_get_dqblk,
1051         .set_dqblk      = dquot_set_dqblk
1052 };
1053
1054 static const struct quotactl_ops ext4_qctl_sysfile_operations = {
1055         .quota_on_meta  = ext4_quota_on_sysfile,
1056         .quota_off      = ext4_quota_off_sysfile,
1057         .quota_sync     = dquot_quota_sync,
1058         .get_info       = dquot_get_dqinfo,
1059         .set_info       = dquot_set_dqinfo,
1060         .get_dqblk      = dquot_get_dqblk,
1061         .set_dqblk      = dquot_set_dqblk
1062 };
1063 #endif
1064
1065 static const struct super_operations ext4_sops = {
1066         .alloc_inode    = ext4_alloc_inode,
1067         .destroy_inode  = ext4_destroy_inode,
1068         .write_inode    = ext4_write_inode,
1069         .dirty_inode    = ext4_dirty_inode,
1070         .drop_inode     = ext4_drop_inode,
1071         .evict_inode    = ext4_evict_inode,
1072         .put_super      = ext4_put_super,
1073         .sync_fs        = ext4_sync_fs,
1074         .freeze_fs      = ext4_freeze,
1075         .unfreeze_fs    = ext4_unfreeze,
1076         .statfs         = ext4_statfs,
1077         .remount_fs     = ext4_remount,
1078         .show_options   = ext4_show_options,
1079 #ifdef CONFIG_QUOTA
1080         .quota_read     = ext4_quota_read,
1081         .quota_write    = ext4_quota_write,
1082 #endif
1083         .bdev_try_to_free_page = bdev_try_to_free_page,
1084 };
1085
1086 static const struct super_operations ext4_nojournal_sops = {
1087         .alloc_inode    = ext4_alloc_inode,
1088         .destroy_inode  = ext4_destroy_inode,
1089         .write_inode    = ext4_write_inode,
1090         .dirty_inode    = ext4_dirty_inode,
1091         .drop_inode     = ext4_drop_inode,
1092         .evict_inode    = ext4_evict_inode,
1093         .put_super      = ext4_put_super,
1094         .statfs         = ext4_statfs,
1095         .remount_fs     = ext4_remount,
1096         .show_options   = ext4_show_options,
1097 #ifdef CONFIG_QUOTA
1098         .quota_read     = ext4_quota_read,
1099         .quota_write    = ext4_quota_write,
1100 #endif
1101         .bdev_try_to_free_page = bdev_try_to_free_page,
1102 };
1103
1104 static const struct export_operations ext4_export_ops = {
1105         .fh_to_dentry = ext4_fh_to_dentry,
1106         .fh_to_parent = ext4_fh_to_parent,
1107         .get_parent = ext4_get_parent,
1108 };
1109
1110 enum {
1111         Opt_bsd_df, Opt_minix_df, Opt_grpid, Opt_nogrpid,
1112         Opt_resgid, Opt_resuid, Opt_sb, Opt_err_cont, Opt_err_panic, Opt_err_ro,
1113         Opt_nouid32, Opt_debug, Opt_removed,
1114         Opt_user_xattr, Opt_nouser_xattr, Opt_acl, Opt_noacl,
1115         Opt_auto_da_alloc, Opt_noauto_da_alloc, Opt_noload,
1116         Opt_commit, Opt_min_batch_time, Opt_max_batch_time,
1117         Opt_journal_dev, Opt_journal_checksum, Opt_journal_async_commit,
1118         Opt_abort, Opt_data_journal, Opt_data_ordered, Opt_data_writeback,
1119         Opt_data_err_abort, Opt_data_err_ignore,
1120         Opt_usrjquota, Opt_grpjquota, Opt_offusrjquota, Opt_offgrpjquota,
1121         Opt_jqfmt_vfsold, Opt_jqfmt_vfsv0, Opt_jqfmt_vfsv1, Opt_quota,
1122         Opt_noquota, Opt_barrier, Opt_nobarrier, Opt_err,
1123         Opt_usrquota, Opt_grpquota, Opt_i_version,
1124         Opt_stripe, Opt_delalloc, Opt_nodelalloc, Opt_mblk_io_submit,
1125         Opt_nomblk_io_submit, Opt_block_validity, Opt_noblock_validity,
1126         Opt_inode_readahead_blks, Opt_journal_ioprio,
1127         Opt_dioread_nolock, Opt_dioread_lock,
1128         Opt_discard, Opt_nodiscard, Opt_init_itable, Opt_noinit_itable,
1129         Opt_max_dir_size_kb,
1130 };
1131
1132 static const match_table_t tokens = {
1133         {Opt_bsd_df, "bsddf"},
1134         {Opt_minix_df, "minixdf"},
1135         {Opt_grpid, "grpid"},
1136         {Opt_grpid, "bsdgroups"},
1137         {Opt_nogrpid, "nogrpid"},
1138         {Opt_nogrpid, "sysvgroups"},
1139         {Opt_resgid, "resgid=%u"},
1140         {Opt_resuid, "resuid=%u"},
1141         {Opt_sb, "sb=%u"},
1142         {Opt_err_cont, "errors=continue"},
1143         {Opt_err_panic, "errors=panic"},
1144         {Opt_err_ro, "errors=remount-ro"},
1145         {Opt_nouid32, "nouid32"},
1146         {Opt_debug, "debug"},
1147         {Opt_removed, "oldalloc"},
1148         {Opt_removed, "orlov"},
1149         {Opt_user_xattr, "user_xattr"},
1150         {Opt_nouser_xattr, "nouser_xattr"},
1151         {Opt_acl, "acl"},
1152         {Opt_noacl, "noacl"},
1153         {Opt_noload, "norecovery"},
1154         {Opt_noload, "noload"},
1155         {Opt_removed, "nobh"},
1156         {Opt_removed, "bh"},
1157         {Opt_commit, "commit=%u"},
1158         {Opt_min_batch_time, "min_batch_time=%u"},
1159         {Opt_max_batch_time, "max_batch_time=%u"},
1160         {Opt_journal_dev, "journal_dev=%u"},
1161         {Opt_journal_checksum, "journal_checksum"},
1162         {Opt_journal_async_commit, "journal_async_commit"},
1163         {Opt_abort, "abort"},
1164         {Opt_data_journal, "data=journal"},
1165         {Opt_data_ordered, "data=ordered"},
1166         {Opt_data_writeback, "data=writeback"},
1167         {Opt_data_err_abort, "data_err=abort"},
1168         {Opt_data_err_ignore, "data_err=ignore"},
1169         {Opt_offusrjquota, "usrjquota="},
1170         {Opt_usrjquota, "usrjquota=%s"},
1171         {Opt_offgrpjquota, "grpjquota="},
1172         {Opt_grpjquota, "grpjquota=%s"},
1173         {Opt_jqfmt_vfsold, "jqfmt=vfsold"},
1174         {Opt_jqfmt_vfsv0, "jqfmt=vfsv0"},
1175         {Opt_jqfmt_vfsv1, "jqfmt=vfsv1"},
1176         {Opt_grpquota, "grpquota"},
1177         {Opt_noquota, "noquota"},
1178         {Opt_quota, "quota"},
1179         {Opt_usrquota, "usrquota"},
1180         {Opt_barrier, "barrier=%u"},
1181         {Opt_barrier, "barrier"},
1182         {Opt_nobarrier, "nobarrier"},
1183         {Opt_i_version, "i_version"},
1184         {Opt_stripe, "stripe=%u"},
1185         {Opt_delalloc, "delalloc"},
1186         {Opt_nodelalloc, "nodelalloc"},
1187         {Opt_removed, "mblk_io_submit"},
1188         {Opt_removed, "nomblk_io_submit"},
1189         {Opt_block_validity, "block_validity"},
1190         {Opt_noblock_validity, "noblock_validity"},
1191         {Opt_inode_readahead_blks, "inode_readahead_blks=%u"},
1192         {Opt_journal_ioprio, "journal_ioprio=%u"},
1193         {Opt_auto_da_alloc, "auto_da_alloc=%u"},
1194         {Opt_auto_da_alloc, "auto_da_alloc"},
1195         {Opt_noauto_da_alloc, "noauto_da_alloc"},
1196         {Opt_dioread_nolock, "dioread_nolock"},
1197         {Opt_dioread_lock, "dioread_lock"},
1198         {Opt_discard, "discard"},
1199         {Opt_nodiscard, "nodiscard"},
1200         {Opt_init_itable, "init_itable=%u"},
1201         {Opt_init_itable, "init_itable"},
1202         {Opt_noinit_itable, "noinit_itable"},
1203         {Opt_max_dir_size_kb, "max_dir_size_kb=%u"},
1204         {Opt_removed, "check=none"},    /* mount option from ext2/3 */
1205         {Opt_removed, "nocheck"},       /* mount option from ext2/3 */
1206         {Opt_removed, "reservation"},   /* mount option from ext2/3 */
1207         {Opt_removed, "noreservation"}, /* mount option from ext2/3 */
1208         {Opt_removed, "journal=%u"},    /* mount option from ext2/3 */
1209         {Opt_err, NULL},
1210 };
1211
1212 static ext4_fsblk_t get_sb_block(void **data)
1213 {
1214         ext4_fsblk_t    sb_block;
1215         char            *options = (char *) *data;
1216
1217         if (!options || strncmp(options, "sb=", 3) != 0)
1218                 return 1;       /* Default location */
1219
1220         options += 3;
1221         /* TODO: use simple_strtoll with >32bit ext4 */
1222         sb_block = simple_strtoul(options, &options, 0);
1223         if (*options && *options != ',') {
1224                 printk(KERN_ERR "EXT4-fs: Invalid sb specification: %s\n",
1225                        (char *) *data);
1226                 return 1;
1227         }
1228         if (*options == ',')
1229                 options++;
1230         *data = (void *) options;
1231
1232         return sb_block;
1233 }
1234
1235 #define DEFAULT_JOURNAL_IOPRIO (IOPRIO_PRIO_VALUE(IOPRIO_CLASS_BE, 3))
1236 static char deprecated_msg[] = "Mount option \"%s\" will be removed by %s\n"
1237         "Contact linux-ext4@vger.kernel.org if you think we should keep it.\n";
1238
1239 #ifdef CONFIG_QUOTA
1240 static int set_qf_name(struct super_block *sb, int qtype, substring_t *args)
1241 {
1242         struct ext4_sb_info *sbi = EXT4_SB(sb);
1243         char *qname;
1244         int ret = -1;
1245
1246         if (sb_any_quota_loaded(sb) &&
1247                 !sbi->s_qf_names[qtype]) {
1248                 ext4_msg(sb, KERN_ERR,
1249                         "Cannot change journaled "
1250                         "quota options when quota turned on");
1251                 return -1;
1252         }
1253         if (EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_QUOTA)) {
1254                 ext4_msg(sb, KERN_ERR, "Cannot set journaled quota options "
1255                          "when QUOTA feature is enabled");
1256                 return -1;
1257         }
1258         qname = match_strdup(args);
1259         if (!qname) {
1260                 ext4_msg(sb, KERN_ERR,
1261                         "Not enough memory for storing quotafile name");
1262                 return -1;
1263         }
1264         if (sbi->s_qf_names[qtype]) {
1265                 if (strcmp(sbi->s_qf_names[qtype], qname) == 0)
1266                         ret = 1;
1267                 else
1268                         ext4_msg(sb, KERN_ERR,
1269                                  "%s quota file already specified",
1270                                  QTYPE2NAME(qtype));
1271                 goto errout;
1272         }
1273         if (strchr(qname, '/')) {
1274                 ext4_msg(sb, KERN_ERR,
1275                         "quotafile must be on filesystem root");
1276                 goto errout;
1277         }
1278         sbi->s_qf_names[qtype] = qname;
1279         set_opt(sb, QUOTA);
1280         return 1;
1281 errout:
1282         kfree(qname);
1283         return ret;
1284 }
1285
1286 static int clear_qf_name(struct super_block *sb, int qtype)
1287 {
1288
1289         struct ext4_sb_info *sbi = EXT4_SB(sb);
1290
1291         if (sb_any_quota_loaded(sb) &&
1292                 sbi->s_qf_names[qtype]) {
1293                 ext4_msg(sb, KERN_ERR, "Cannot change journaled quota options"
1294                         " when quota turned on");
1295                 return -1;
1296         }
1297         kfree(sbi->s_qf_names[qtype]);
1298         sbi->s_qf_names[qtype] = NULL;
1299         return 1;
1300 }
1301 #endif
1302
1303 #define MOPT_SET        0x0001
1304 #define MOPT_CLEAR      0x0002
1305 #define MOPT_NOSUPPORT  0x0004
1306 #define MOPT_EXPLICIT   0x0008
1307 #define MOPT_CLEAR_ERR  0x0010
1308 #define MOPT_GTE0       0x0020
1309 #ifdef CONFIG_QUOTA
1310 #define MOPT_Q          0
1311 #define MOPT_QFMT       0x0040
1312 #else
1313 #define MOPT_Q          MOPT_NOSUPPORT
1314 #define MOPT_QFMT       MOPT_NOSUPPORT
1315 #endif
1316 #define MOPT_DATAJ      0x0080
1317 #define MOPT_NO_EXT2    0x0100
1318 #define MOPT_NO_EXT3    0x0200
1319 #define MOPT_EXT4_ONLY  (MOPT_NO_EXT2 | MOPT_NO_EXT3)
1320
1321 static const struct mount_opts {
1322         int     token;
1323         int     mount_opt;
1324         int     flags;
1325 } ext4_mount_opts[] = {
1326         {Opt_minix_df, EXT4_MOUNT_MINIX_DF, MOPT_SET},
1327         {Opt_bsd_df, EXT4_MOUNT_MINIX_DF, MOPT_CLEAR},
1328         {Opt_grpid, EXT4_MOUNT_GRPID, MOPT_SET},
1329         {Opt_nogrpid, EXT4_MOUNT_GRPID, MOPT_CLEAR},
1330         {Opt_block_validity, EXT4_MOUNT_BLOCK_VALIDITY, MOPT_SET},
1331         {Opt_noblock_validity, EXT4_MOUNT_BLOCK_VALIDITY, MOPT_CLEAR},
1332         {Opt_dioread_nolock, EXT4_MOUNT_DIOREAD_NOLOCK,
1333          MOPT_EXT4_ONLY | MOPT_SET},
1334         {Opt_dioread_lock, EXT4_MOUNT_DIOREAD_NOLOCK,
1335          MOPT_EXT4_ONLY | MOPT_CLEAR},
1336         {Opt_discard, EXT4_MOUNT_DISCARD, MOPT_SET},
1337         {Opt_nodiscard, EXT4_MOUNT_DISCARD, MOPT_CLEAR},
1338         {Opt_delalloc, EXT4_MOUNT_DELALLOC,
1339          MOPT_EXT4_ONLY | MOPT_SET | MOPT_EXPLICIT},
1340         {Opt_nodelalloc, EXT4_MOUNT_DELALLOC,
1341          MOPT_EXT4_ONLY | MOPT_CLEAR | MOPT_EXPLICIT},
1342         {Opt_journal_checksum, EXT4_MOUNT_JOURNAL_CHECKSUM,
1343          MOPT_EXT4_ONLY | MOPT_SET},
1344         {Opt_journal_async_commit, (EXT4_MOUNT_JOURNAL_ASYNC_COMMIT |
1345                                     EXT4_MOUNT_JOURNAL_CHECKSUM),
1346          MOPT_EXT4_ONLY | MOPT_SET},
1347         {Opt_noload, EXT4_MOUNT_NOLOAD, MOPT_NO_EXT2 | MOPT_SET},
1348         {Opt_err_panic, EXT4_MOUNT_ERRORS_PANIC, MOPT_SET | MOPT_CLEAR_ERR},
1349         {Opt_err_ro, EXT4_MOUNT_ERRORS_RO, MOPT_SET | MOPT_CLEAR_ERR},
1350         {Opt_err_cont, EXT4_MOUNT_ERRORS_CONT, MOPT_SET | MOPT_CLEAR_ERR},
1351         {Opt_data_err_abort, EXT4_MOUNT_DATA_ERR_ABORT,
1352          MOPT_NO_EXT2 | MOPT_SET},
1353         {Opt_data_err_ignore, EXT4_MOUNT_DATA_ERR_ABORT,
1354          MOPT_NO_EXT2 | MOPT_CLEAR},
1355         {Opt_barrier, EXT4_MOUNT_BARRIER, MOPT_SET},
1356         {Opt_nobarrier, EXT4_MOUNT_BARRIER, MOPT_CLEAR},
1357         {Opt_noauto_da_alloc, EXT4_MOUNT_NO_AUTO_DA_ALLOC, MOPT_SET},
1358         {Opt_auto_da_alloc, EXT4_MOUNT_NO_AUTO_DA_ALLOC, MOPT_CLEAR},
1359         {Opt_noinit_itable, EXT4_MOUNT_INIT_INODE_TABLE, MOPT_CLEAR},
1360         {Opt_commit, 0, MOPT_GTE0},
1361         {Opt_max_batch_time, 0, MOPT_GTE0},
1362         {Opt_min_batch_time, 0, MOPT_GTE0},
1363         {Opt_inode_readahead_blks, 0, MOPT_GTE0},
1364         {Opt_init_itable, 0, MOPT_GTE0},
1365         {Opt_stripe, 0, MOPT_GTE0},
1366         {Opt_resuid, 0, MOPT_GTE0},
1367         {Opt_resgid, 0, MOPT_GTE0},
1368         {Opt_journal_dev, 0, MOPT_GTE0},
1369         {Opt_journal_ioprio, 0, MOPT_GTE0},
1370         {Opt_data_journal, EXT4_MOUNT_JOURNAL_DATA, MOPT_NO_EXT2 | MOPT_DATAJ},
1371         {Opt_data_ordered, EXT4_MOUNT_ORDERED_DATA, MOPT_NO_EXT2 | MOPT_DATAJ},
1372         {Opt_data_writeback, EXT4_MOUNT_WRITEBACK_DATA,
1373          MOPT_NO_EXT2 | MOPT_DATAJ},
1374         {Opt_user_xattr, EXT4_MOUNT_XATTR_USER, MOPT_SET},
1375         {Opt_nouser_xattr, EXT4_MOUNT_XATTR_USER, MOPT_CLEAR},
1376 #ifdef CONFIG_EXT4_FS_POSIX_ACL
1377         {Opt_acl, EXT4_MOUNT_POSIX_ACL, MOPT_SET},
1378         {Opt_noacl, EXT4_MOUNT_POSIX_ACL, MOPT_CLEAR},
1379 #else
1380         {Opt_acl, 0, MOPT_NOSUPPORT},
1381         {Opt_noacl, 0, MOPT_NOSUPPORT},
1382 #endif
1383         {Opt_nouid32, EXT4_MOUNT_NO_UID32, MOPT_SET},
1384         {Opt_debug, EXT4_MOUNT_DEBUG, MOPT_SET},
1385         {Opt_quota, EXT4_MOUNT_QUOTA | EXT4_MOUNT_USRQUOTA, MOPT_SET | MOPT_Q},
1386         {Opt_usrquota, EXT4_MOUNT_QUOTA | EXT4_MOUNT_USRQUOTA,
1387                                                         MOPT_SET | MOPT_Q},
1388         {Opt_grpquota, EXT4_MOUNT_QUOTA | EXT4_MOUNT_GRPQUOTA,
1389                                                         MOPT_SET | MOPT_Q},
1390         {Opt_noquota, (EXT4_MOUNT_QUOTA | EXT4_MOUNT_USRQUOTA |
1391                        EXT4_MOUNT_GRPQUOTA), MOPT_CLEAR | MOPT_Q},
1392         {Opt_usrjquota, 0, MOPT_Q},
1393         {Opt_grpjquota, 0, MOPT_Q},
1394         {Opt_offusrjquota, 0, MOPT_Q},
1395         {Opt_offgrpjquota, 0, MOPT_Q},
1396         {Opt_jqfmt_vfsold, QFMT_VFS_OLD, MOPT_QFMT},
1397         {Opt_jqfmt_vfsv0, QFMT_VFS_V0, MOPT_QFMT},
1398         {Opt_jqfmt_vfsv1, QFMT_VFS_V1, MOPT_QFMT},
1399         {Opt_max_dir_size_kb, 0, MOPT_GTE0},
1400         {Opt_err, 0, 0}
1401 };
1402
1403 static int handle_mount_opt(struct super_block *sb, char *opt, int token,
1404                             substring_t *args, unsigned long *journal_devnum,
1405                             unsigned int *journal_ioprio, int is_remount)
1406 {
1407         struct ext4_sb_info *sbi = EXT4_SB(sb);
1408         const struct mount_opts *m;
1409         kuid_t uid;
1410         kgid_t gid;
1411         int arg = 0;
1412
1413 #ifdef CONFIG_QUOTA
1414         if (token == Opt_usrjquota)
1415                 return set_qf_name(sb, USRQUOTA, &args[0]);
1416         else if (token == Opt_grpjquota)
1417                 return set_qf_name(sb, GRPQUOTA, &args[0]);
1418         else if (token == Opt_offusrjquota)
1419                 return clear_qf_name(sb, USRQUOTA);
1420         else if (token == Opt_offgrpjquota)
1421                 return clear_qf_name(sb, GRPQUOTA);
1422 #endif
1423         switch (token) {
1424         case Opt_noacl:
1425         case Opt_nouser_xattr:
1426                 ext4_msg(sb, KERN_WARNING, deprecated_msg, opt, "3.5");
1427                 break;
1428         case Opt_sb:
1429                 return 1;       /* handled by get_sb_block() */
1430         case Opt_removed:
1431                 ext4_msg(sb, KERN_WARNING, "Ignoring removed %s option", opt);
1432                 return 1;
1433         case Opt_abort:
1434                 sbi->s_mount_flags |= EXT4_MF_FS_ABORTED;
1435                 return 1;
1436         case Opt_i_version:
1437                 sb->s_flags |= MS_I_VERSION;
1438                 return 1;
1439         }
1440
1441         for (m = ext4_mount_opts; m->token != Opt_err; m++)
1442                 if (token == m->token)
1443                         break;
1444
1445         if (m->token == Opt_err) {
1446                 ext4_msg(sb, KERN_ERR, "Unrecognized mount option \"%s\" "
1447                          "or missing value", opt);
1448                 return -1;
1449         }
1450
1451         if ((m->flags & MOPT_NO_EXT2) && IS_EXT2_SB(sb)) {
1452                 ext4_msg(sb, KERN_ERR,
1453                          "Mount option \"%s\" incompatible with ext2", opt);
1454                 return -1;
1455         }
1456         if ((m->flags & MOPT_NO_EXT3) && IS_EXT3_SB(sb)) {
1457                 ext4_msg(sb, KERN_ERR,
1458                          "Mount option \"%s\" incompatible with ext3", opt);
1459                 return -1;
1460         }
1461
1462         if (args->from && match_int(args, &arg))
1463                 return -1;
1464         if (args->from && (m->flags & MOPT_GTE0) && (arg < 0))
1465                 return -1;
1466         if (m->flags & MOPT_EXPLICIT)
1467                 set_opt2(sb, EXPLICIT_DELALLOC);
1468         if (m->flags & MOPT_CLEAR_ERR)
1469                 clear_opt(sb, ERRORS_MASK);
1470         if (token == Opt_noquota && sb_any_quota_loaded(sb)) {
1471                 ext4_msg(sb, KERN_ERR, "Cannot change quota "
1472                          "options when quota turned on");
1473                 return -1;
1474         }
1475
1476         if (m->flags & MOPT_NOSUPPORT) {
1477                 ext4_msg(sb, KERN_ERR, "%s option not supported", opt);
1478         } else if (token == Opt_commit) {
1479                 if (arg == 0)
1480                         arg = JBD2_DEFAULT_MAX_COMMIT_AGE;
1481                 sbi->s_commit_interval = HZ * arg;
1482         } else if (token == Opt_max_batch_time) {
1483                 if (arg == 0)
1484                         arg = EXT4_DEF_MAX_BATCH_TIME;
1485                 sbi->s_max_batch_time = arg;
1486         } else if (token == Opt_min_batch_time) {
1487                 sbi->s_min_batch_time = arg;
1488         } else if (token == Opt_inode_readahead_blks) {
1489                 if (arg && (arg > (1 << 30) || !is_power_of_2(arg))) {
1490                         ext4_msg(sb, KERN_ERR,
1491                                  "EXT4-fs: inode_readahead_blks must be "
1492                                  "0 or a power of 2 smaller than 2^31");
1493                         return -1;
1494                 }
1495                 sbi->s_inode_readahead_blks = arg;
1496         } else if (token == Opt_init_itable) {
1497                 set_opt(sb, INIT_INODE_TABLE);
1498                 if (!args->from)
1499                         arg = EXT4_DEF_LI_WAIT_MULT;
1500                 sbi->s_li_wait_mult = arg;
1501         } else if (token == Opt_max_dir_size_kb) {
1502                 sbi->s_max_dir_size_kb = arg;
1503         } else if (token == Opt_stripe) {
1504                 sbi->s_stripe = arg;
1505         } else if (token == Opt_resuid) {
1506                 uid = make_kuid(current_user_ns(), arg);
1507                 if (!uid_valid(uid)) {
1508                         ext4_msg(sb, KERN_ERR, "Invalid uid value %d", arg);
1509                         return -1;
1510                 }
1511                 sbi->s_resuid = uid;
1512         } else if (token == Opt_resgid) {
1513                 gid = make_kgid(current_user_ns(), arg);
1514                 if (!gid_valid(gid)) {
1515                         ext4_msg(sb, KERN_ERR, "Invalid gid value %d", arg);
1516                         return -1;
1517                 }
1518                 sbi->s_resgid = gid;
1519         } else if (token == Opt_journal_dev) {
1520                 if (is_remount) {
1521                         ext4_msg(sb, KERN_ERR,
1522                                  "Cannot specify journal on remount");
1523                         return -1;
1524                 }
1525                 *journal_devnum = arg;
1526         } else if (token == Opt_journal_ioprio) {
1527                 if (arg > 7) {
1528                         ext4_msg(sb, KERN_ERR, "Invalid journal IO priority"
1529                                  " (must be 0-7)");
1530                         return -1;
1531                 }
1532                 *journal_ioprio =
1533                         IOPRIO_PRIO_VALUE(IOPRIO_CLASS_BE, arg);
1534         } else if (m->flags & MOPT_DATAJ) {
1535                 if (is_remount) {
1536                         if (!sbi->s_journal)
1537                                 ext4_msg(sb, KERN_WARNING, "Remounting file system with no journal so ignoring journalled data option");
1538                         else if (test_opt(sb, DATA_FLAGS) != m->mount_opt) {
1539                                 ext4_msg(sb, KERN_ERR,
1540                                          "Cannot change data mode on remount");
1541                                 return -1;
1542                         }
1543                 } else {
1544                         clear_opt(sb, DATA_FLAGS);
1545                         sbi->s_mount_opt |= m->mount_opt;
1546                 }
1547 #ifdef CONFIG_QUOTA
1548         } else if (m->flags & MOPT_QFMT) {
1549                 if (sb_any_quota_loaded(sb) &&
1550                     sbi->s_jquota_fmt != m->mount_opt) {
1551                         ext4_msg(sb, KERN_ERR, "Cannot change journaled "
1552                                  "quota options when quota turned on");
1553                         return -1;
1554                 }
1555                 if (EXT4_HAS_RO_COMPAT_FEATURE(sb,
1556                                                EXT4_FEATURE_RO_COMPAT_QUOTA)) {
1557                         ext4_msg(sb, KERN_ERR,
1558                                  "Cannot set journaled quota options "
1559                                  "when QUOTA feature is enabled");
1560                         return -1;
1561                 }
1562                 sbi->s_jquota_fmt = m->mount_opt;
1563 #endif
1564         } else {
1565                 if (!args->from)
1566                         arg = 1;
1567                 if (m->flags & MOPT_CLEAR)
1568                         arg = !arg;
1569                 else if (unlikely(!(m->flags & MOPT_SET))) {
1570                         ext4_msg(sb, KERN_WARNING,
1571                                  "buggy handling of option %s", opt);
1572                         WARN_ON(1);
1573                         return -1;
1574                 }
1575                 if (arg != 0)
1576                         sbi->s_mount_opt |= m->mount_opt;
1577                 else
1578                         sbi->s_mount_opt &= ~m->mount_opt;
1579         }
1580         return 1;
1581 }
1582
1583 static int parse_options(char *options, struct super_block *sb,
1584                          unsigned long *journal_devnum,
1585                          unsigned int *journal_ioprio,
1586                          int is_remount)
1587 {
1588         struct ext4_sb_info *sbi = EXT4_SB(sb);
1589         char *p;
1590         substring_t args[MAX_OPT_ARGS];
1591         int token;
1592
1593         if (!options)
1594                 return 1;
1595
1596         while ((p = strsep(&options, ",")) != NULL) {
1597                 if (!*p)
1598                         continue;
1599                 /*
1600                  * Initialize args struct so we know whether arg was
1601                  * found; some options take optional arguments.
1602                  */
1603                 args[0].to = args[0].from = NULL;
1604                 token = match_token(p, tokens, args);
1605                 if (handle_mount_opt(sb, p, token, args, journal_devnum,
1606                                      journal_ioprio, is_remount) < 0)
1607                         return 0;
1608         }
1609 #ifdef CONFIG_QUOTA
1610         if (EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_QUOTA) &&
1611             (test_opt(sb, USRQUOTA) || test_opt(sb, GRPQUOTA))) {
1612                 ext4_msg(sb, KERN_ERR, "Cannot set quota options when QUOTA "
1613                          "feature is enabled");
1614                 return 0;
1615         }
1616         if (sbi->s_qf_names[USRQUOTA] || sbi->s_qf_names[GRPQUOTA]) {
1617                 if (test_opt(sb, USRQUOTA) && sbi->s_qf_names[USRQUOTA])
1618                         clear_opt(sb, USRQUOTA);
1619
1620                 if (test_opt(sb, GRPQUOTA) && sbi->s_qf_names[GRPQUOTA])
1621                         clear_opt(sb, GRPQUOTA);
1622
1623                 if (test_opt(sb, GRPQUOTA) || test_opt(sb, USRQUOTA)) {
1624                         ext4_msg(sb, KERN_ERR, "old and new quota "
1625                                         "format mixing");
1626                         return 0;
1627                 }
1628
1629                 if (!sbi->s_jquota_fmt) {
1630                         ext4_msg(sb, KERN_ERR, "journaled quota format "
1631                                         "not specified");
1632                         return 0;
1633                 }
1634         } else {
1635                 if (sbi->s_jquota_fmt) {
1636                         ext4_msg(sb, KERN_ERR, "journaled quota format "
1637                                         "specified with no journaling "
1638                                         "enabled");
1639                         return 0;
1640                 }
1641         }
1642 #endif
1643         if (test_opt(sb, DIOREAD_NOLOCK)) {
1644                 int blocksize =
1645                         BLOCK_SIZE << le32_to_cpu(sbi->s_es->s_log_block_size);
1646
1647                 if (blocksize < PAGE_CACHE_SIZE) {
1648                         ext4_msg(sb, KERN_ERR, "can't mount with "
1649                                  "dioread_nolock if block size != PAGE_SIZE");
1650                         return 0;
1651                 }
1652         }
1653         return 1;
1654 }
1655
1656 static inline void ext4_show_quota_options(struct seq_file *seq,
1657                                            struct super_block *sb)
1658 {
1659 #if defined(CONFIG_QUOTA)
1660         struct ext4_sb_info *sbi = EXT4_SB(sb);
1661
1662         if (sbi->s_jquota_fmt) {
1663                 char *fmtname = "";
1664
1665                 switch (sbi->s_jquota_fmt) {
1666                 case QFMT_VFS_OLD:
1667                         fmtname = "vfsold";
1668                         break;
1669                 case QFMT_VFS_V0:
1670                         fmtname = "vfsv0";
1671                         break;
1672                 case QFMT_VFS_V1:
1673                         fmtname = "vfsv1";
1674                         break;
1675                 }
1676                 seq_printf(seq, ",jqfmt=%s", fmtname);
1677         }
1678
1679         if (sbi->s_qf_names[USRQUOTA])
1680                 seq_printf(seq, ",usrjquota=%s", sbi->s_qf_names[USRQUOTA]);
1681
1682         if (sbi->s_qf_names[GRPQUOTA])
1683                 seq_printf(seq, ",grpjquota=%s", sbi->s_qf_names[GRPQUOTA]);
1684
1685         if (test_opt(sb, USRQUOTA))
1686                 seq_puts(seq, ",usrquota");
1687
1688         if (test_opt(sb, GRPQUOTA))
1689                 seq_puts(seq, ",grpquota");
1690 #endif
1691 }
1692
1693 static const char *token2str(int token)
1694 {
1695         const struct match_token *t;
1696
1697         for (t = tokens; t->token != Opt_err; t++)
1698                 if (t->token == token && !strchr(t->pattern, '='))
1699                         break;
1700         return t->pattern;
1701 }
1702
1703 /*
1704  * Show an option if
1705  *  - it's set to a non-default value OR
1706  *  - if the per-sb default is different from the global default
1707  */
1708 static int _ext4_show_options(struct seq_file *seq, struct super_block *sb,
1709                               int nodefs)
1710 {
1711         struct ext4_sb_info *sbi = EXT4_SB(sb);
1712         struct ext4_super_block *es = sbi->s_es;
1713         int def_errors, def_mount_opt = nodefs ? 0 : sbi->s_def_mount_opt;
1714         const struct mount_opts *m;
1715         char sep = nodefs ? '\n' : ',';
1716
1717 #define SEQ_OPTS_PUTS(str) seq_printf(seq, "%c" str, sep)
1718 #define SEQ_OPTS_PRINT(str, arg) seq_printf(seq, "%c" str, sep, arg)
1719
1720         if (sbi->s_sb_block != 1)
1721                 SEQ_OPTS_PRINT("sb=%llu", sbi->s_sb_block);
1722
1723         for (m = ext4_mount_opts; m->token != Opt_err; m++) {
1724                 int want_set = m->flags & MOPT_SET;
1725                 if (((m->flags & (MOPT_SET|MOPT_CLEAR)) == 0) ||
1726                     (m->flags & MOPT_CLEAR_ERR))
1727                         continue;
1728                 if (!(m->mount_opt & (sbi->s_mount_opt ^ def_mount_opt)))
1729                         continue; /* skip if same as the default */
1730                 if ((want_set &&
1731                      (sbi->s_mount_opt & m->mount_opt) != m->mount_opt) ||
1732                     (!want_set && (sbi->s_mount_opt & m->mount_opt)))
1733                         continue; /* select Opt_noFoo vs Opt_Foo */
1734                 SEQ_OPTS_PRINT("%s", token2str(m->token));
1735         }
1736
1737         if (nodefs || !uid_eq(sbi->s_resuid, make_kuid(&init_user_ns, EXT4_DEF_RESUID)) ||
1738             le16_to_cpu(es->s_def_resuid) != EXT4_DEF_RESUID)
1739                 SEQ_OPTS_PRINT("resuid=%u",
1740                                 from_kuid_munged(&init_user_ns, sbi->s_resuid));
1741         if (nodefs || !gid_eq(sbi->s_resgid, make_kgid(&init_user_ns, EXT4_DEF_RESGID)) ||
1742             le16_to_cpu(es->s_def_resgid) != EXT4_DEF_RESGID)
1743                 SEQ_OPTS_PRINT("resgid=%u",
1744                                 from_kgid_munged(&init_user_ns, sbi->s_resgid));
1745         def_errors = nodefs ? -1 : le16_to_cpu(es->s_errors);
1746         if (test_opt(sb, ERRORS_RO) && def_errors != EXT4_ERRORS_RO)
1747                 SEQ_OPTS_PUTS("errors=remount-ro");
1748         if (test_opt(sb, ERRORS_CONT) && def_errors != EXT4_ERRORS_CONTINUE)
1749                 SEQ_OPTS_PUTS("errors=continue");
1750         if (test_opt(sb, ERRORS_PANIC) && def_errors != EXT4_ERRORS_PANIC)
1751                 SEQ_OPTS_PUTS("errors=panic");
1752         if (nodefs || sbi->s_commit_interval != JBD2_DEFAULT_MAX_COMMIT_AGE*HZ)
1753                 SEQ_OPTS_PRINT("commit=%lu", sbi->s_commit_interval / HZ);
1754         if (nodefs || sbi->s_min_batch_time != EXT4_DEF_MIN_BATCH_TIME)
1755                 SEQ_OPTS_PRINT("min_batch_time=%u", sbi->s_min_batch_time);
1756         if (nodefs || sbi->s_max_batch_time != EXT4_DEF_MAX_BATCH_TIME)
1757                 SEQ_OPTS_PRINT("max_batch_time=%u", sbi->s_max_batch_time);
1758         if (sb->s_flags & MS_I_VERSION)
1759                 SEQ_OPTS_PUTS("i_version");
1760         if (nodefs || sbi->s_stripe)
1761                 SEQ_OPTS_PRINT("stripe=%lu", sbi->s_stripe);
1762         if (EXT4_MOUNT_DATA_FLAGS & (sbi->s_mount_opt ^ def_mount_opt)) {
1763                 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA)
1764                         SEQ_OPTS_PUTS("data=journal");
1765                 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA)
1766                         SEQ_OPTS_PUTS("data=ordered");
1767                 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_WRITEBACK_DATA)
1768                         SEQ_OPTS_PUTS("data=writeback");
1769         }
1770         if (nodefs ||
1771             sbi->s_inode_readahead_blks != EXT4_DEF_INODE_READAHEAD_BLKS)
1772                 SEQ_OPTS_PRINT("inode_readahead_blks=%u",
1773                                sbi->s_inode_readahead_blks);
1774
1775         if (nodefs || (test_opt(sb, INIT_INODE_TABLE) &&
1776                        (sbi->s_li_wait_mult != EXT4_DEF_LI_WAIT_MULT)))
1777                 SEQ_OPTS_PRINT("init_itable=%u", sbi->s_li_wait_mult);
1778         if (nodefs || sbi->s_max_dir_size_kb)
1779                 SEQ_OPTS_PRINT("max_dir_size_kb=%u", sbi->s_max_dir_size_kb);
1780
1781         ext4_show_quota_options(seq, sb);
1782         return 0;
1783 }
1784
1785 static int ext4_show_options(struct seq_file *seq, struct dentry *root)
1786 {
1787         return _ext4_show_options(seq, root->d_sb, 0);
1788 }
1789
1790 static int options_seq_show(struct seq_file *seq, void *offset)
1791 {
1792         struct super_block *sb = seq->private;
1793         int rc;
1794
1795         seq_puts(seq, (sb->s_flags & MS_RDONLY) ? "ro" : "rw");
1796         rc = _ext4_show_options(seq, sb, 1);
1797         seq_puts(seq, "\n");
1798         return rc;
1799 }
1800
1801 static int options_open_fs(struct inode *inode, struct file *file)
1802 {
1803         return single_open(file, options_seq_show, PDE(inode)->data);
1804 }
1805
1806 static const struct file_operations ext4_seq_options_fops = {
1807         .owner = THIS_MODULE,
1808         .open = options_open_fs,
1809         .read = seq_read,
1810         .llseek = seq_lseek,
1811         .release = single_release,
1812 };
1813
1814 static int ext4_setup_super(struct super_block *sb, struct ext4_super_block *es,
1815                             int read_only)
1816 {
1817         struct ext4_sb_info *sbi = EXT4_SB(sb);
1818         int res = 0;
1819
1820         if (le32_to_cpu(es->s_rev_level) > EXT4_MAX_SUPP_REV) {
1821                 ext4_msg(sb, KERN_ERR, "revision level too high, "
1822                          "forcing read-only mode");
1823                 res = MS_RDONLY;
1824         }
1825         if (read_only)
1826                 goto done;
1827         if (!(sbi->s_mount_state & EXT4_VALID_FS))
1828                 ext4_msg(sb, KERN_WARNING, "warning: mounting unchecked fs, "
1829                          "running e2fsck is recommended");
1830         else if ((sbi->s_mount_state & EXT4_ERROR_FS))
1831                 ext4_msg(sb, KERN_WARNING,
1832                          "warning: mounting fs with errors, "
1833                          "running e2fsck is recommended");
1834         else if ((__s16) le16_to_cpu(es->s_max_mnt_count) > 0 &&
1835                  le16_to_cpu(es->s_mnt_count) >=
1836                  (unsigned short) (__s16) le16_to_cpu(es->s_max_mnt_count))
1837                 ext4_msg(sb, KERN_WARNING,
1838                          "warning: maximal mount count reached, "
1839                          "running e2fsck is recommended");
1840         else if (le32_to_cpu(es->s_checkinterval) &&
1841                 (le32_to_cpu(es->s_lastcheck) +
1842                         le32_to_cpu(es->s_checkinterval) <= get_seconds()))
1843                 ext4_msg(sb, KERN_WARNING,
1844                          "warning: checktime reached, "
1845                          "running e2fsck is recommended");
1846         if (!sbi->s_journal)
1847                 es->s_state &= cpu_to_le16(~EXT4_VALID_FS);
1848         if (!(__s16) le16_to_cpu(es->s_max_mnt_count))
1849                 es->s_max_mnt_count = cpu_to_le16(EXT4_DFL_MAX_MNT_COUNT);
1850         le16_add_cpu(&es->s_mnt_count, 1);
1851         es->s_mtime = cpu_to_le32(get_seconds());
1852         ext4_update_dynamic_rev(sb);
1853         if (sbi->s_journal)
1854                 EXT4_SET_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
1855
1856         ext4_commit_super(sb, 1);
1857 done:
1858         if (test_opt(sb, DEBUG))
1859                 printk(KERN_INFO "[EXT4 FS bs=%lu, gc=%u, "
1860                                 "bpg=%lu, ipg=%lu, mo=%04x, mo2=%04x]\n",
1861                         sb->s_blocksize,
1862                         sbi->s_groups_count,
1863                         EXT4_BLOCKS_PER_GROUP(sb),
1864                         EXT4_INODES_PER_GROUP(sb),
1865                         sbi->s_mount_opt, sbi->s_mount_opt2);
1866
1867         cleancache_init_fs(sb);
1868         return res;
1869 }
1870
1871 int ext4_alloc_flex_bg_array(struct super_block *sb, ext4_group_t ngroup)
1872 {
1873         struct ext4_sb_info *sbi = EXT4_SB(sb);
1874         struct flex_groups *new_groups;
1875         int size;
1876
1877         if (!sbi->s_log_groups_per_flex)
1878                 return 0;
1879
1880         size = ext4_flex_group(sbi, ngroup - 1) + 1;
1881         if (size <= sbi->s_flex_groups_allocated)
1882                 return 0;
1883
1884         size = roundup_pow_of_two(size * sizeof(struct flex_groups));
1885         new_groups = ext4_kvzalloc(size, GFP_KERNEL);
1886         if (!new_groups) {
1887                 ext4_msg(sb, KERN_ERR, "not enough memory for %d flex groups",
1888                          size / (int) sizeof(struct flex_groups));
1889                 return -ENOMEM;
1890         }
1891
1892         if (sbi->s_flex_groups) {
1893                 memcpy(new_groups, sbi->s_flex_groups,
1894                        (sbi->s_flex_groups_allocated *
1895                         sizeof(struct flex_groups)));
1896                 ext4_kvfree(sbi->s_flex_groups);
1897         }
1898         sbi->s_flex_groups = new_groups;
1899         sbi->s_flex_groups_allocated = size / sizeof(struct flex_groups);
1900         return 0;
1901 }
1902
1903 static int ext4_fill_flex_info(struct super_block *sb)
1904 {
1905         struct ext4_sb_info *sbi = EXT4_SB(sb);
1906         struct ext4_group_desc *gdp = NULL;
1907         ext4_group_t flex_group;
1908         unsigned int groups_per_flex = 0;
1909         int i, err;
1910
1911         sbi->s_log_groups_per_flex = sbi->s_es->s_log_groups_per_flex;
1912         if (sbi->s_log_groups_per_flex < 1 || sbi->s_log_groups_per_flex > 31) {
1913                 sbi->s_log_groups_per_flex = 0;
1914                 return 1;
1915         }
1916         groups_per_flex = 1U << sbi->s_log_groups_per_flex;
1917
1918         err = ext4_alloc_flex_bg_array(sb, sbi->s_groups_count);
1919         if (err)
1920                 goto failed;
1921
1922         for (i = 0; i < sbi->s_groups_count; i++) {
1923                 gdp = ext4_get_group_desc(sb, i, NULL);
1924
1925                 flex_group = ext4_flex_group(sbi, i);
1926                 atomic_add(ext4_free_inodes_count(sb, gdp),
1927                            &sbi->s_flex_groups[flex_group].free_inodes);
1928                 atomic_add(ext4_free_group_clusters(sb, gdp),
1929                            &sbi->s_flex_groups[flex_group].free_clusters);
1930                 atomic_add(ext4_used_dirs_count(sb, gdp),
1931                            &sbi->s_flex_groups[flex_group].used_dirs);
1932         }
1933
1934         return 1;
1935 failed:
1936         return 0;
1937 }
1938
1939 static __le16 ext4_group_desc_csum(struct ext4_sb_info *sbi, __u32 block_group,
1940                                    struct ext4_group_desc *gdp)
1941 {
1942         int offset;
1943         __u16 crc = 0;
1944         __le32 le_group = cpu_to_le32(block_group);
1945
1946         if ((sbi->s_es->s_feature_ro_compat &
1947              cpu_to_le32(EXT4_FEATURE_RO_COMPAT_METADATA_CSUM))) {
1948                 /* Use new metadata_csum algorithm */
1949                 __u16 old_csum;
1950                 __u32 csum32;
1951
1952                 old_csum = gdp->bg_checksum;
1953                 gdp->bg_checksum = 0;
1954                 csum32 = ext4_chksum(sbi, sbi->s_csum_seed, (__u8 *)&le_group,
1955                                      sizeof(le_group));
1956                 csum32 = ext4_chksum(sbi, csum32, (__u8 *)gdp,
1957                                      sbi->s_desc_size);
1958                 gdp->bg_checksum = old_csum;
1959
1960                 crc = csum32 & 0xFFFF;
1961                 goto out;
1962         }
1963
1964         /* old crc16 code */
1965         offset = offsetof(struct ext4_group_desc, bg_checksum);
1966
1967         crc = crc16(~0, sbi->s_es->s_uuid, sizeof(sbi->s_es->s_uuid));
1968         crc = crc16(crc, (__u8 *)&le_group, sizeof(le_group));
1969         crc = crc16(crc, (__u8 *)gdp, offset);
1970         offset += sizeof(gdp->bg_checksum); /* skip checksum */
1971         /* for checksum of struct ext4_group_desc do the rest...*/
1972         if ((sbi->s_es->s_feature_incompat &
1973              cpu_to_le32(EXT4_FEATURE_INCOMPAT_64BIT)) &&
1974             offset < le16_to_cpu(sbi->s_es->s_desc_size))
1975                 crc = crc16(crc, (__u8 *)gdp + offset,
1976                             le16_to_cpu(sbi->s_es->s_desc_size) -
1977                                 offset);
1978
1979 out:
1980         return cpu_to_le16(crc);
1981 }
1982
1983 int ext4_group_desc_csum_verify(struct super_block *sb, __u32 block_group,
1984                                 struct ext4_group_desc *gdp)
1985 {
1986         if (ext4_has_group_desc_csum(sb) &&
1987             (gdp->bg_checksum != ext4_group_desc_csum(EXT4_SB(sb),
1988                                                       block_group, gdp)))
1989                 return 0;
1990
1991         return 1;
1992 }
1993
1994 void ext4_group_desc_csum_set(struct super_block *sb, __u32 block_group,
1995                               struct ext4_group_desc *gdp)
1996 {
1997         if (!ext4_has_group_desc_csum(sb))
1998                 return;
1999         gdp->bg_checksum = ext4_group_desc_csum(EXT4_SB(sb), block_group, gdp);
2000 }
2001
2002 /* Called at mount-time, super-block is locked */
2003 static int ext4_check_descriptors(struct super_block *sb,
2004                                   ext4_group_t *first_not_zeroed)
2005 {
2006         struct ext4_sb_info *sbi = EXT4_SB(sb);
2007         ext4_fsblk_t first_block = le32_to_cpu(sbi->s_es->s_first_data_block);
2008         ext4_fsblk_t last_block;
2009         ext4_fsblk_t block_bitmap;
2010         ext4_fsblk_t inode_bitmap;
2011         ext4_fsblk_t inode_table;
2012         int flexbg_flag = 0;
2013         ext4_group_t i, grp = sbi->s_groups_count;
2014
2015         if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_FLEX_BG))
2016                 flexbg_flag = 1;
2017
2018         ext4_debug("Checking group descriptors");
2019
2020         for (i = 0; i < sbi->s_groups_count; i++) {
2021                 struct ext4_group_desc *gdp = ext4_get_group_desc(sb, i, NULL);
2022
2023                 if (i == sbi->s_groups_count - 1 || flexbg_flag)
2024                         last_block = ext4_blocks_count(sbi->s_es) - 1;
2025                 else
2026                         last_block = first_block +
2027                                 (EXT4_BLOCKS_PER_GROUP(sb) - 1);
2028
2029                 if ((grp == sbi->s_groups_count) &&
2030                    !(gdp->bg_flags & cpu_to_le16(EXT4_BG_INODE_ZEROED)))
2031                         grp = i;
2032
2033                 block_bitmap = ext4_block_bitmap(sb, gdp);
2034                 if (block_bitmap < first_block || block_bitmap > last_block) {
2035                         ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2036                                "Block bitmap for group %u not in group "
2037                                "(block %llu)!", i, block_bitmap);
2038                         return 0;
2039                 }
2040                 inode_bitmap = ext4_inode_bitmap(sb, gdp);
2041                 if (inode_bitmap < first_block || inode_bitmap > last_block) {
2042                         ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2043                                "Inode bitmap for group %u not in group "
2044                                "(block %llu)!", i, inode_bitmap);
2045                         return 0;
2046                 }
2047                 inode_table = ext4_inode_table(sb, gdp);
2048                 if (inode_table < first_block ||
2049                     inode_table + sbi->s_itb_per_group - 1 > last_block) {
2050                         ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2051                                "Inode table for group %u not in group "
2052                                "(block %llu)!", i, inode_table);
2053                         return 0;
2054                 }
2055                 ext4_lock_group(sb, i);
2056                 if (!ext4_group_desc_csum_verify(sb, i, gdp)) {
2057                         ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2058                                  "Checksum for group %u failed (%u!=%u)",
2059                                  i, le16_to_cpu(ext4_group_desc_csum(sbi, i,
2060                                      gdp)), le16_to_cpu(gdp->bg_checksum));
2061                         if (!(sb->s_flags & MS_RDONLY)) {
2062                                 ext4_unlock_group(sb, i);
2063                                 return 0;
2064                         }
2065                 }
2066                 ext4_unlock_group(sb, i);
2067                 if (!flexbg_flag)
2068                         first_block += EXT4_BLOCKS_PER_GROUP(sb);
2069         }
2070         if (NULL != first_not_zeroed)
2071                 *first_not_zeroed = grp;
2072
2073         ext4_free_blocks_count_set(sbi->s_es,
2074                                    EXT4_C2B(sbi, ext4_count_free_clusters(sb)));
2075         sbi->s_es->s_free_inodes_count =cpu_to_le32(ext4_count_free_inodes(sb));
2076         return 1;
2077 }
2078
2079 /* ext4_orphan_cleanup() walks a singly-linked list of inodes (starting at
2080  * the superblock) which were deleted from all directories, but held open by
2081  * a process at the time of a crash.  We walk the list and try to delete these
2082  * inodes at recovery time (only with a read-write filesystem).
2083  *
2084  * In order to keep the orphan inode chain consistent during traversal (in
2085  * case of crash during recovery), we link each inode into the superblock
2086  * orphan list_head and handle it the same way as an inode deletion during
2087  * normal operation (which journals the operations for us).
2088  *
2089  * We only do an iget() and an iput() on each inode, which is very safe if we
2090  * accidentally point at an in-use or already deleted inode.  The worst that
2091  * can happen in this case is that we get a "bit already cleared" message from
2092  * ext4_free_inode().  The only reason we would point at a wrong inode is if
2093  * e2fsck was run on this filesystem, and it must have already done the orphan
2094  * inode cleanup for us, so we can safely abort without any further action.
2095  */
2096 static void ext4_orphan_cleanup(struct super_block *sb,
2097                                 struct ext4_super_block *es)
2098 {
2099         unsigned int s_flags = sb->s_flags;
2100         int nr_orphans = 0, nr_truncates = 0;
2101 #ifdef CONFIG_QUOTA
2102         int i;
2103 #endif
2104         if (!es->s_last_orphan) {
2105                 jbd_debug(4, "no orphan inodes to clean up\n");
2106                 return;
2107         }
2108
2109         if (bdev_read_only(sb->s_bdev)) {
2110                 ext4_msg(sb, KERN_ERR, "write access "
2111                         "unavailable, skipping orphan cleanup");
2112                 return;
2113         }
2114
2115         /* Check if feature set would not allow a r/w mount */
2116         if (!ext4_feature_set_ok(sb, 0)) {
2117                 ext4_msg(sb, KERN_INFO, "Skipping orphan cleanup due to "
2118                          "unknown ROCOMPAT features");
2119                 return;
2120         }
2121
2122         if (EXT4_SB(sb)->s_mount_state & EXT4_ERROR_FS) {
2123                 /* don't clear list on RO mount w/ errors */
2124                 if (es->s_last_orphan && !(s_flags & MS_RDONLY)) {
2125                         jbd_debug(1, "Errors on filesystem, "
2126                                   "clearing orphan list.\n");
2127                         es->s_last_orphan = 0;
2128                 }
2129                 jbd_debug(1, "Skipping orphan recovery on fs with errors.\n");
2130                 return;
2131         }
2132
2133         if (s_flags & MS_RDONLY) {
2134                 ext4_msg(sb, KERN_INFO, "orphan cleanup on readonly fs");
2135                 sb->s_flags &= ~MS_RDONLY;
2136         }
2137 #ifdef CONFIG_QUOTA
2138         /* Needed for iput() to work correctly and not trash data */
2139         sb->s_flags |= MS_ACTIVE;
2140         /* Turn on quotas so that they are updated correctly */
2141         for (i = 0; i < MAXQUOTAS; i++) {
2142                 if (EXT4_SB(sb)->s_qf_names[i]) {
2143                         int ret = ext4_quota_on_mount(sb, i);
2144                         if (ret < 0)
2145                                 ext4_msg(sb, KERN_ERR,
2146                                         "Cannot turn on journaled "
2147                                         "quota: error %d", ret);
2148                 }
2149         }
2150 #endif
2151
2152         while (es->s_last_orphan) {
2153                 struct inode *inode;
2154
2155                 inode = ext4_orphan_get(sb, le32_to_cpu(es->s_last_orphan));
2156                 if (IS_ERR(inode)) {
2157                         es->s_last_orphan = 0;
2158                         break;
2159                 }
2160
2161                 list_add(&EXT4_I(inode)->i_orphan, &EXT4_SB(sb)->s_orphan);
2162                 dquot_initialize(inode);
2163                 if (inode->i_nlink) {
2164                         ext4_msg(sb, KERN_DEBUG,
2165                                 "%s: truncating inode %lu to %lld bytes",
2166                                 __func__, inode->i_ino, inode->i_size);
2167                         jbd_debug(2, "truncating inode %lu to %lld bytes\n",
2168                                   inode->i_ino, inode->i_size);
2169                         mutex_lock(&inode->i_mutex);
2170                         ext4_truncate(inode);
2171                         mutex_unlock(&inode->i_mutex);
2172                         nr_truncates++;
2173                 } else {
2174                         ext4_msg(sb, KERN_DEBUG,
2175                                 "%s: deleting unreferenced inode %lu",
2176                                 __func__, inode->i_ino);
2177                         jbd_debug(2, "deleting unreferenced inode %lu\n",
2178                                   inode->i_ino);
2179                         nr_orphans++;
2180                 }
2181                 iput(inode);  /* The delete magic happens here! */
2182         }
2183
2184 #define PLURAL(x) (x), ((x) == 1) ? "" : "s"
2185
2186         if (nr_orphans)
2187                 ext4_msg(sb, KERN_INFO, "%d orphan inode%s deleted",
2188                        PLURAL(nr_orphans));
2189         if (nr_truncates)
2190                 ext4_msg(sb, KERN_INFO, "%d truncate%s cleaned up",
2191                        PLURAL(nr_truncates));
2192 #ifdef CONFIG_QUOTA
2193         /* Turn quotas off */
2194         for (i = 0; i < MAXQUOTAS; i++) {
2195                 if (sb_dqopt(sb)->files[i])
2196                         dquot_quota_off(sb, i);
2197         }
2198 #endif
2199         sb->s_flags = s_flags; /* Restore MS_RDONLY status */
2200 }
2201
2202 /*
2203  * Maximal extent format file size.
2204  * Resulting logical blkno at s_maxbytes must fit in our on-disk
2205  * extent format containers, within a sector_t, and within i_blocks
2206  * in the vfs.  ext4 inode has 48 bits of i_block in fsblock units,
2207  * so that won't be a limiting factor.
2208  *
2209  * However there is other limiting factor. We do store extents in the form
2210  * of starting block and length, hence the resulting length of the extent
2211  * covering maximum file size must fit into on-disk format containers as
2212  * well. Given that length is always by 1 unit bigger than max unit (because
2213  * we count 0 as well) we have to lower the s_maxbytes by one fs block.
2214  *
2215  * Note, this does *not* consider any metadata overhead for vfs i_blocks.
2216  */
2217 static loff_t ext4_max_size(int blkbits, int has_huge_files)
2218 {
2219         loff_t res;
2220         loff_t upper_limit = MAX_LFS_FILESIZE;
2221
2222         /* small i_blocks in vfs inode? */
2223         if (!has_huge_files || sizeof(blkcnt_t) < sizeof(u64)) {
2224                 /*
2225                  * CONFIG_LBDAF is not enabled implies the inode
2226                  * i_block represent total blocks in 512 bytes
2227                  * 32 == size of vfs inode i_blocks * 8
2228                  */
2229                 upper_limit = (1LL << 32) - 1;
2230
2231                 /* total blocks in file system block size */
2232                 upper_limit >>= (blkbits - 9);
2233                 upper_limit <<= blkbits;
2234         }
2235
2236         /*
2237          * 32-bit extent-start container, ee_block. We lower the maxbytes
2238          * by one fs block, so ee_len can cover the extent of maximum file
2239          * size
2240          */
2241         res = (1LL << 32) - 1;
2242         res <<= blkbits;
2243
2244         /* Sanity check against vm- & vfs- imposed limits */
2245         if (res > upper_limit)
2246                 res = upper_limit;
2247
2248         return res;
2249 }
2250
2251 /*
2252  * Maximal bitmap file size.  There is a direct, and {,double-,triple-}indirect
2253  * block limit, and also a limit of (2^48 - 1) 512-byte sectors in i_blocks.
2254  * We need to be 1 filesystem block less than the 2^48 sector limit.
2255  */
2256 static loff_t ext4_max_bitmap_size(int bits, int has_huge_files)
2257 {
2258         loff_t res = EXT4_NDIR_BLOCKS;
2259         int meta_blocks;
2260         loff_t upper_limit;
2261         /* This is calculated to be the largest file size for a dense, block
2262          * mapped file such that the file's total number of 512-byte sectors,
2263          * including data and all indirect blocks, does not exceed (2^48 - 1).
2264          *
2265          * __u32 i_blocks_lo and _u16 i_blocks_high represent the total
2266          * number of 512-byte sectors of the file.
2267          */
2268
2269         if (!has_huge_files || sizeof(blkcnt_t) < sizeof(u64)) {
2270                 /*
2271                  * !has_huge_files or CONFIG_LBDAF not enabled implies that
2272                  * the inode i_block field represents total file blocks in
2273                  * 2^32 512-byte sectors == size of vfs inode i_blocks * 8
2274                  */
2275                 upper_limit = (1LL << 32) - 1;
2276
2277                 /* total blocks in file system block size */
2278                 upper_limit >>= (bits - 9);
2279
2280         } else {
2281                 /*
2282                  * We use 48 bit ext4_inode i_blocks
2283                  * With EXT4_HUGE_FILE_FL set the i_blocks
2284                  * represent total number of blocks in
2285                  * file system block size
2286                  */
2287                 upper_limit = (1LL << 48) - 1;
2288
2289         }
2290
2291         /* indirect blocks */
2292         meta_blocks = 1;
2293         /* double indirect blocks */
2294         meta_blocks += 1 + (1LL << (bits-2));
2295         /* tripple indirect blocks */
2296         meta_blocks += 1 + (1LL << (bits-2)) + (1LL << (2*(bits-2)));
2297
2298         upper_limit -= meta_blocks;
2299         upper_limit <<= bits;
2300
2301         res += 1LL << (bits-2);
2302         res += 1LL << (2*(bits-2));
2303         res += 1LL << (3*(bits-2));
2304         res <<= bits;
2305         if (res > upper_limit)
2306                 res = upper_limit;
2307
2308         if (res > MAX_LFS_FILESIZE)
2309                 res = MAX_LFS_FILESIZE;
2310
2311         return res;
2312 }
2313
2314 static ext4_fsblk_t descriptor_loc(struct super_block *sb,
2315                                    ext4_fsblk_t logical_sb_block, int nr)
2316 {
2317         struct ext4_sb_info *sbi = EXT4_SB(sb);
2318         ext4_group_t bg, first_meta_bg;
2319         int has_super = 0;
2320
2321         first_meta_bg = le32_to_cpu(sbi->s_es->s_first_meta_bg);
2322
2323         if (!EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_META_BG) ||
2324             nr < first_meta_bg)
2325                 return logical_sb_block + nr + 1;
2326         bg = sbi->s_desc_per_block * nr;
2327         if (ext4_bg_has_super(sb, bg))
2328                 has_super = 1;
2329
2330         return (has_super + ext4_group_first_block_no(sb, bg));
2331 }
2332
2333 /**
2334  * ext4_get_stripe_size: Get the stripe size.
2335  * @sbi: In memory super block info
2336  *
2337  * If we have specified it via mount option, then
2338  * use the mount option value. If the value specified at mount time is
2339  * greater than the blocks per group use the super block value.
2340  * If the super block value is greater than blocks per group return 0.
2341  * Allocator needs it be less than blocks per group.
2342  *
2343  */
2344 static unsigned long ext4_get_stripe_size(struct ext4_sb_info *sbi)
2345 {
2346         unsigned long stride = le16_to_cpu(sbi->s_es->s_raid_stride);
2347         unsigned long stripe_width =
2348                         le32_to_cpu(sbi->s_es->s_raid_stripe_width);
2349         int ret;
2350
2351         if (sbi->s_stripe && sbi->s_stripe <= sbi->s_blocks_per_group)
2352                 ret = sbi->s_stripe;
2353         else if (stripe_width <= sbi->s_blocks_per_group)
2354                 ret = stripe_width;
2355         else if (stride <= sbi->s_blocks_per_group)
2356                 ret = stride;
2357         else
2358                 ret = 0;
2359
2360         /*
2361          * If the stripe width is 1, this makes no sense and
2362          * we set it to 0 to turn off stripe handling code.
2363          */
2364         if (ret <= 1)
2365                 ret = 0;
2366
2367         return ret;
2368 }
2369
2370 /* sysfs supprt */
2371
2372 struct ext4_attr {
2373         struct attribute attr;
2374         ssize_t (*show)(struct ext4_attr *, struct ext4_sb_info *, char *);
2375         ssize_t (*store)(struct ext4_attr *, struct ext4_sb_info *,
2376                          const char *, size_t);
2377         int offset;
2378 };
2379
2380 static int parse_strtoul(const char *buf,
2381                 unsigned long max, unsigned long *value)
2382 {
2383         char *endp;
2384
2385         *value = simple_strtoul(skip_spaces(buf), &endp, 0);
2386         endp = skip_spaces(endp);
2387         if (*endp || *value > max)
2388                 return -EINVAL;
2389
2390         return 0;
2391 }
2392
2393 static ssize_t delayed_allocation_blocks_show(struct ext4_attr *a,
2394                                               struct ext4_sb_info *sbi,
2395                                               char *buf)
2396 {
2397         return snprintf(buf, PAGE_SIZE, "%llu\n",
2398                 (s64) EXT4_C2B(sbi,
2399                         percpu_counter_sum(&sbi->s_dirtyclusters_counter)));
2400 }
2401
2402 static ssize_t session_write_kbytes_show(struct ext4_attr *a,
2403                                          struct ext4_sb_info *sbi, char *buf)
2404 {
2405         struct super_block *sb = sbi->s_buddy_cache->i_sb;
2406
2407         if (!sb->s_bdev->bd_part)
2408                 return snprintf(buf, PAGE_SIZE, "0\n");
2409         return snprintf(buf, PAGE_SIZE, "%lu\n",
2410                         (part_stat_read(sb->s_bdev->bd_part, sectors[1]) -
2411                          sbi->s_sectors_written_start) >> 1);
2412 }
2413
2414 static ssize_t lifetime_write_kbytes_show(struct ext4_attr *a,
2415                                           struct ext4_sb_info *sbi, char *buf)
2416 {
2417         struct super_block *sb = sbi->s_buddy_cache->i_sb;
2418
2419         if (!sb->s_bdev->bd_part)
2420                 return snprintf(buf, PAGE_SIZE, "0\n");
2421         return snprintf(buf, PAGE_SIZE, "%llu\n",
2422                         (unsigned long long)(sbi->s_kbytes_written +
2423                         ((part_stat_read(sb->s_bdev->bd_part, sectors[1]) -
2424                           EXT4_SB(sb)->s_sectors_written_start) >> 1)));
2425 }
2426
2427 static ssize_t inode_readahead_blks_store(struct ext4_attr *a,
2428                                           struct ext4_sb_info *sbi,
2429                                           const char *buf, size_t count)
2430 {
2431         unsigned long t;
2432
2433         if (parse_strtoul(buf, 0x40000000, &t))
2434                 return -EINVAL;
2435
2436         if (t && !is_power_of_2(t))
2437                 return -EINVAL;
2438
2439         sbi->s_inode_readahead_blks = t;
2440         return count;
2441 }
2442
2443 static ssize_t sbi_ui_show(struct ext4_attr *a,
2444                            struct ext4_sb_info *sbi, char *buf)
2445 {
2446         unsigned int *ui = (unsigned int *) (((char *) sbi) + a->offset);
2447
2448         return snprintf(buf, PAGE_SIZE, "%u\n", *ui);
2449 }
2450
2451 static ssize_t sbi_ui_store(struct ext4_attr *a,
2452                             struct ext4_sb_info *sbi,
2453                             const char *buf, size_t count)
2454 {
2455         unsigned int *ui = (unsigned int *) (((char *) sbi) + a->offset);
2456         unsigned long t;
2457
2458         if (parse_strtoul(buf, 0xffffffff, &t))
2459                 return -EINVAL;
2460         *ui = t;
2461         return count;
2462 }
2463
2464 static ssize_t trigger_test_error(struct ext4_attr *a,
2465                                   struct ext4_sb_info *sbi,
2466                                   const char *buf, size_t count)
2467 {
2468         int len = count;
2469
2470         if (!capable(CAP_SYS_ADMIN))
2471                 return -EPERM;
2472
2473         if (len && buf[len-1] == '\n')
2474                 len--;
2475
2476         if (len)
2477                 ext4_error(sbi->s_sb, "%.*s", len, buf);
2478         return count;
2479 }
2480
2481 #define EXT4_ATTR_OFFSET(_name,_mode,_show,_store,_elname) \
2482 static struct ext4_attr ext4_attr_##_name = {                   \
2483         .attr = {.name = __stringify(_name), .mode = _mode },   \
2484         .show   = _show,                                        \
2485         .store  = _store,                                       \
2486         .offset = offsetof(struct ext4_sb_info, _elname),       \
2487 }
2488 #define EXT4_ATTR(name, mode, show, store) \
2489 static struct ext4_attr ext4_attr_##name = __ATTR(name, mode, show, store)
2490
2491 #define EXT4_INFO_ATTR(name) EXT4_ATTR(name, 0444, NULL, NULL)
2492 #define EXT4_RO_ATTR(name) EXT4_ATTR(name, 0444, name##_show, NULL)
2493 #define EXT4_RW_ATTR(name) EXT4_ATTR(name, 0644, name##_show, name##_store)
2494 #define EXT4_RW_ATTR_SBI_UI(name, elname)       \
2495         EXT4_ATTR_OFFSET(name, 0644, sbi_ui_show, sbi_ui_store, elname)
2496 #define ATTR_LIST(name) &ext4_attr_##name.attr
2497
2498 EXT4_RO_ATTR(delayed_allocation_blocks);
2499 EXT4_RO_ATTR(session_write_kbytes);
2500 EXT4_RO_ATTR(lifetime_write_kbytes);
2501 EXT4_ATTR_OFFSET(inode_readahead_blks, 0644, sbi_ui_show,
2502                  inode_readahead_blks_store, s_inode_readahead_blks);
2503 EXT4_RW_ATTR_SBI_UI(inode_goal, s_inode_goal);
2504 EXT4_RW_ATTR_SBI_UI(mb_stats, s_mb_stats);
2505 EXT4_RW_ATTR_SBI_UI(mb_max_to_scan, s_mb_max_to_scan);
2506 EXT4_RW_ATTR_SBI_UI(mb_min_to_scan, s_mb_min_to_scan);
2507 EXT4_RW_ATTR_SBI_UI(mb_order2_req, s_mb_order2_reqs);
2508 EXT4_RW_ATTR_SBI_UI(mb_stream_req, s_mb_stream_request);
2509 EXT4_RW_ATTR_SBI_UI(mb_group_prealloc, s_mb_group_prealloc);
2510 EXT4_RW_ATTR_SBI_UI(max_writeback_mb_bump, s_max_writeback_mb_bump);
2511 EXT4_RW_ATTR_SBI_UI(extent_max_zeroout_kb, s_extent_max_zeroout_kb);
2512 EXT4_ATTR(trigger_fs_error, 0200, NULL, trigger_test_error);
2513
2514 static struct attribute *ext4_attrs[] = {
2515         ATTR_LIST(delayed_allocation_blocks),
2516         ATTR_LIST(session_write_kbytes),
2517         ATTR_LIST(lifetime_write_kbytes),
2518         ATTR_LIST(inode_readahead_blks),
2519         ATTR_LIST(inode_goal),
2520         ATTR_LIST(mb_stats),
2521         ATTR_LIST(mb_max_to_scan),
2522         ATTR_LIST(mb_min_to_scan),
2523         ATTR_LIST(mb_order2_req),
2524         ATTR_LIST(mb_stream_req),
2525         ATTR_LIST(mb_group_prealloc),
2526         ATTR_LIST(max_writeback_mb_bump),
2527         ATTR_LIST(extent_max_zeroout_kb),
2528         ATTR_LIST(trigger_fs_error),
2529         NULL,
2530 };
2531
2532 /* Features this copy of ext4 supports */
2533 EXT4_INFO_ATTR(lazy_itable_init);
2534 EXT4_INFO_ATTR(batched_discard);
2535 EXT4_INFO_ATTR(meta_bg_resize);
2536
2537 static struct attribute *ext4_feat_attrs[] = {
2538         ATTR_LIST(lazy_itable_init),
2539         ATTR_LIST(batched_discard),
2540         ATTR_LIST(meta_bg_resize),
2541         NULL,
2542 };
2543
2544 static ssize_t ext4_attr_show(struct kobject *kobj,
2545                               struct attribute *attr, char *buf)
2546 {
2547         struct ext4_sb_info *sbi = container_of(kobj, struct ext4_sb_info,
2548                                                 s_kobj);
2549         struct ext4_attr *a = container_of(attr, struct ext4_attr, attr);
2550
2551         return a->show ? a->show(a, sbi, buf) : 0;
2552 }
2553
2554 static ssize_t ext4_attr_store(struct kobject *kobj,
2555                                struct attribute *attr,
2556                                const char *buf, size_t len)
2557 {
2558         struct ext4_sb_info *sbi = container_of(kobj, struct ext4_sb_info,
2559                                                 s_kobj);
2560         struct ext4_attr *a = container_of(attr, struct ext4_attr, attr);
2561
2562         return a->store ? a->store(a, sbi, buf, len) : 0;
2563 }
2564
2565 static void ext4_sb_release(struct kobject *kobj)
2566 {
2567         struct ext4_sb_info *sbi = container_of(kobj, struct ext4_sb_info,
2568                                                 s_kobj);
2569         complete(&sbi->s_kobj_unregister);
2570 }
2571
2572 static const struct sysfs_ops ext4_attr_ops = {
2573         .show   = ext4_attr_show,
2574         .store  = ext4_attr_store,
2575 };
2576
2577 static struct kobj_type ext4_ktype = {
2578         .default_attrs  = ext4_attrs,
2579         .sysfs_ops      = &ext4_attr_ops,
2580         .release        = ext4_sb_release,
2581 };
2582
2583 static void ext4_feat_release(struct kobject *kobj)
2584 {
2585         complete(&ext4_feat->f_kobj_unregister);
2586 }
2587
2588 static struct kobj_type ext4_feat_ktype = {
2589         .default_attrs  = ext4_feat_attrs,
2590         .sysfs_ops      = &ext4_attr_ops,
2591         .release        = ext4_feat_release,
2592 };
2593
2594 /*
2595  * Check whether this filesystem can be mounted based on
2596  * the features present and the RDONLY/RDWR mount requested.
2597  * Returns 1 if this filesystem can be mounted as requested,
2598  * 0 if it cannot be.
2599  */
2600 static int ext4_feature_set_ok(struct super_block *sb, int readonly)
2601 {
2602         if (EXT4_HAS_INCOMPAT_FEATURE(sb, ~EXT4_FEATURE_INCOMPAT_SUPP)) {
2603                 ext4_msg(sb, KERN_ERR,
2604                         "Couldn't mount because of "
2605                         "unsupported optional features (%x)",
2606                         (le32_to_cpu(EXT4_SB(sb)->s_es->s_feature_incompat) &
2607                         ~EXT4_FEATURE_INCOMPAT_SUPP));
2608                 return 0;
2609         }
2610
2611         if (readonly)
2612                 return 1;
2613
2614         /* Check that feature set is OK for a read-write mount */
2615         if (EXT4_HAS_RO_COMPAT_FEATURE(sb, ~EXT4_FEATURE_RO_COMPAT_SUPP)) {
2616                 ext4_msg(sb, KERN_ERR, "couldn't mount RDWR because of "
2617                          "unsupported optional features (%x)",
2618                          (le32_to_cpu(EXT4_SB(sb)->s_es->s_feature_ro_compat) &
2619                                 ~EXT4_FEATURE_RO_COMPAT_SUPP));
2620                 return 0;
2621         }
2622         /*
2623          * Large file size enabled file system can only be mounted
2624          * read-write on 32-bit systems if kernel is built with CONFIG_LBDAF
2625          */
2626         if (EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_HUGE_FILE)) {
2627                 if (sizeof(blkcnt_t) < sizeof(u64)) {
2628                         ext4_msg(sb, KERN_ERR, "Filesystem with huge files "
2629                                  "cannot be mounted RDWR without "
2630                                  "CONFIG_LBDAF");
2631                         return 0;
2632                 }
2633         }
2634         if (EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_BIGALLOC) &&
2635             !EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_EXTENTS)) {
2636                 ext4_msg(sb, KERN_ERR,
2637                          "Can't support bigalloc feature without "
2638                          "extents feature\n");
2639                 return 0;
2640         }
2641
2642 #ifndef CONFIG_QUOTA
2643         if (EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_QUOTA) &&
2644             !readonly) {
2645                 ext4_msg(sb, KERN_ERR,
2646                          "Filesystem with quota feature cannot be mounted RDWR "
2647                          "without CONFIG_QUOTA");
2648                 return 0;
2649         }
2650 #endif  /* CONFIG_QUOTA */
2651         return 1;
2652 }
2653
2654 /*
2655  * This function is called once a day if we have errors logged
2656  * on the file system
2657  */
2658 static void print_daily_error_info(unsigned long arg)
2659 {
2660         struct super_block *sb = (struct super_block *) arg;
2661         struct ext4_sb_info *sbi;
2662         struct ext4_super_block *es;
2663
2664         sbi = EXT4_SB(sb);
2665         es = sbi->s_es;
2666
2667         if (es->s_error_count)
2668                 ext4_msg(sb, KERN_NOTICE, "error count: %u",
2669                          le32_to_cpu(es->s_error_count));
2670         if (es->s_first_error_time) {
2671                 printk(KERN_NOTICE "EXT4-fs (%s): initial error at %u: %.*s:%d",
2672                        sb->s_id, le32_to_cpu(es->s_first_error_time),
2673                        (int) sizeof(es->s_first_error_func),
2674                        es->s_first_error_func,
2675                        le32_to_cpu(es->s_first_error_line));
2676                 if (es->s_first_error_ino)
2677                         printk(": inode %u",
2678                                le32_to_cpu(es->s_first_error_ino));
2679                 if (es->s_first_error_block)
2680                         printk(": block %llu", (unsigned long long)
2681                                le64_to_cpu(es->s_first_error_block));
2682                 printk("\n");
2683         }
2684         if (es->s_last_error_time) {
2685                 printk(KERN_NOTICE "EXT4-fs (%s): last error at %u: %.*s:%d",
2686                        sb->s_id, le32_to_cpu(es->s_last_error_time),
2687                        (int) sizeof(es->s_last_error_func),
2688                        es->s_last_error_func,
2689                        le32_to_cpu(es->s_last_error_line));
2690                 if (es->s_last_error_ino)
2691                         printk(": inode %u",
2692                                le32_to_cpu(es->s_last_error_ino));
2693                 if (es->s_last_error_block)
2694                         printk(": block %llu", (unsigned long long)
2695                                le64_to_cpu(es->s_last_error_block));
2696                 printk("\n");
2697         }
2698         mod_timer(&sbi->s_err_report, jiffies + 24*60*60*HZ);  /* Once a day */
2699 }
2700
2701 /* Find next suitable group and run ext4_init_inode_table */
2702 static int ext4_run_li_request(struct ext4_li_request *elr)
2703 {
2704         struct ext4_group_desc *gdp = NULL;
2705         ext4_group_t group, ngroups;
2706         struct super_block *sb;
2707         unsigned long timeout = 0;
2708         int ret = 0;
2709
2710         sb = elr->lr_super;
2711         ngroups = EXT4_SB(sb)->s_groups_count;
2712
2713         sb_start_write(sb);
2714         for (group = elr->lr_next_group; group < ngroups; group++) {
2715                 gdp = ext4_get_group_desc(sb, group, NULL);
2716                 if (!gdp) {
2717                         ret = 1;
2718                         break;
2719                 }
2720
2721                 if (!(gdp->bg_flags & cpu_to_le16(EXT4_BG_INODE_ZEROED)))
2722                         break;
2723         }
2724
2725         if (group >= ngroups)
2726                 ret = 1;
2727
2728         if (!ret) {
2729                 timeout = jiffies;
2730                 ret = ext4_init_inode_table(sb, group,
2731                                             elr->lr_timeout ? 0 : 1);
2732                 if (elr->lr_timeout == 0) {
2733                         timeout = (jiffies - timeout) *
2734                                   elr->lr_sbi->s_li_wait_mult;
2735                         elr->lr_timeout = timeout;
2736                 }
2737                 elr->lr_next_sched = jiffies + elr->lr_timeout;
2738                 elr->lr_next_group = group + 1;
2739         }
2740         sb_end_write(sb);
2741
2742         return ret;
2743 }
2744
2745 /*
2746  * Remove lr_request from the list_request and free the
2747  * request structure. Should be called with li_list_mtx held
2748  */
2749 static void ext4_remove_li_request(struct ext4_li_request *elr)
2750 {
2751         struct ext4_sb_info *sbi;
2752
2753         if (!elr)
2754                 return;
2755
2756         sbi = elr->lr_sbi;
2757
2758         list_del(&elr->lr_request);
2759         sbi->s_li_request = NULL;
2760         kfree(elr);
2761 }
2762
2763 static void ext4_unregister_li_request(struct super_block *sb)
2764 {
2765         mutex_lock(&ext4_li_mtx);
2766         if (!ext4_li_info) {
2767                 mutex_unlock(&ext4_li_mtx);
2768                 return;
2769         }
2770
2771         mutex_lock(&ext4_li_info->li_list_mtx);
2772         ext4_remove_li_request(EXT4_SB(sb)->s_li_request);
2773         mutex_unlock(&ext4_li_info->li_list_mtx);
2774         mutex_unlock(&ext4_li_mtx);
2775 }
2776
2777 static struct task_struct *ext4_lazyinit_task;
2778
2779 /*
2780  * This is the function where ext4lazyinit thread lives. It walks
2781  * through the request list searching for next scheduled filesystem.
2782  * When such a fs is found, run the lazy initialization request
2783  * (ext4_rn_li_request) and keep track of the time spend in this
2784  * function. Based on that time we compute next schedule time of
2785  * the request. When walking through the list is complete, compute
2786  * next waking time and put itself into sleep.
2787  */
2788 static int ext4_lazyinit_thread(void *arg)
2789 {
2790         struct ext4_lazy_init *eli = (struct ext4_lazy_init *)arg;
2791         struct list_head *pos, *n;
2792         struct ext4_li_request *elr;
2793         unsigned long next_wakeup, cur;
2794
2795         BUG_ON(NULL == eli);
2796
2797 cont_thread:
2798         while (true) {
2799                 next_wakeup = MAX_JIFFY_OFFSET;
2800
2801                 mutex_lock(&eli->li_list_mtx);
2802                 if (list_empty(&eli->li_request_list)) {
2803                         mutex_unlock(&eli->li_list_mtx);
2804                         goto exit_thread;
2805                 }
2806
2807                 list_for_each_safe(pos, n, &eli->li_request_list) {
2808                         elr = list_entry(pos, struct ext4_li_request,
2809                                          lr_request);
2810
2811                         if (time_after_eq(jiffies, elr->lr_next_sched)) {
2812                                 if (ext4_run_li_request(elr) != 0) {
2813                                         /* error, remove the lazy_init job */
2814                                         ext4_remove_li_request(elr);
2815                                         continue;
2816                                 }
2817                         }
2818
2819                         if (time_before(elr->lr_next_sched, next_wakeup))
2820                                 next_wakeup = elr->lr_next_sched;
2821                 }
2822                 mutex_unlock(&eli->li_list_mtx);
2823
2824                 try_to_freeze();
2825
2826                 cur = jiffies;
2827                 if ((time_after_eq(cur, next_wakeup)) ||
2828                     (MAX_JIFFY_OFFSET == next_wakeup)) {
2829                         cond_resched();
2830                         continue;
2831                 }
2832
2833                 schedule_timeout_interruptible(next_wakeup - cur);
2834
2835                 if (kthread_should_stop()) {
2836                         ext4_clear_request_list();
2837                         goto exit_thread;
2838                 }
2839         }
2840
2841 exit_thread:
2842         /*
2843          * It looks like the request list is empty, but we need
2844          * to check it under the li_list_mtx lock, to prevent any
2845          * additions into it, and of course we should lock ext4_li_mtx
2846          * to atomically free the list and ext4_li_info, because at
2847          * this point another ext4 filesystem could be registering
2848          * new one.
2849          */
2850         mutex_lock(&ext4_li_mtx);
2851         mutex_lock(&eli->li_list_mtx);
2852         if (!list_empty(&eli->li_request_list)) {
2853                 mutex_unlock(&eli->li_list_mtx);
2854                 mutex_unlock(&ext4_li_mtx);
2855                 goto cont_thread;
2856         }
2857         mutex_unlock(&eli->li_list_mtx);
2858         kfree(ext4_li_info);
2859         ext4_li_info = NULL;
2860         mutex_unlock(&ext4_li_mtx);
2861
2862         return 0;
2863 }
2864
2865 static void ext4_clear_request_list(void)
2866 {
2867         struct list_head *pos, *n;
2868         struct ext4_li_request *elr;
2869
2870         mutex_lock(&ext4_li_info->li_list_mtx);
2871         list_for_each_safe(pos, n, &ext4_li_info->li_request_list) {
2872                 elr = list_entry(pos, struct ext4_li_request,
2873                                  lr_request);
2874                 ext4_remove_li_request(elr);
2875         }
2876         mutex_unlock(&ext4_li_info->li_list_mtx);
2877 }
2878
2879 static int ext4_run_lazyinit_thread(void)
2880 {
2881         ext4_lazyinit_task = kthread_run(ext4_lazyinit_thread,
2882                                          ext4_li_info, "ext4lazyinit");
2883         if (IS_ERR(ext4_lazyinit_task)) {
2884                 int err = PTR_ERR(ext4_lazyinit_task);
2885                 ext4_clear_request_list();
2886                 kfree(ext4_li_info);
2887                 ext4_li_info = NULL;
2888                 printk(KERN_CRIT "EXT4-fs: error %d creating inode table "
2889                                  "initialization thread\n",
2890                                  err);
2891                 return err;
2892         }
2893         ext4_li_info->li_state |= EXT4_LAZYINIT_RUNNING;
2894         return 0;
2895 }
2896
2897 /*
2898  * Check whether it make sense to run itable init. thread or not.
2899  * If there is at least one uninitialized inode table, return
2900  * corresponding group number, else the loop goes through all
2901  * groups and return total number of groups.
2902  */
2903 static ext4_group_t ext4_has_uninit_itable(struct super_block *sb)
2904 {
2905         ext4_group_t group, ngroups = EXT4_SB(sb)->s_groups_count;
2906         struct ext4_group_desc *gdp = NULL;
2907
2908         for (group = 0; group < ngroups; group++) {
2909                 gdp = ext4_get_group_desc(sb, group, NULL);
2910                 if (!gdp)
2911                         continue;
2912
2913                 if (!(gdp->bg_flags & cpu_to_le16(EXT4_BG_INODE_ZEROED)))
2914                         break;
2915         }
2916
2917         return group;
2918 }
2919
2920 static int ext4_li_info_new(void)
2921 {
2922         struct ext4_lazy_init *eli = NULL;
2923
2924         eli = kzalloc(sizeof(*eli), GFP_KERNEL);
2925         if (!eli)
2926                 return -ENOMEM;
2927
2928         INIT_LIST_HEAD(&eli->li_request_list);
2929         mutex_init(&eli->li_list_mtx);
2930
2931         eli->li_state |= EXT4_LAZYINIT_QUIT;
2932
2933         ext4_li_info = eli;
2934
2935         return 0;
2936 }
2937
2938 static struct ext4_li_request *ext4_li_request_new(struct super_block *sb,
2939                                             ext4_group_t start)
2940 {
2941         struct ext4_sb_info *sbi = EXT4_SB(sb);
2942         struct ext4_li_request *elr;
2943         unsigned long rnd;
2944
2945         elr = kzalloc(sizeof(*elr), GFP_KERNEL);
2946         if (!elr)
2947                 return NULL;
2948
2949         elr->lr_super = sb;
2950         elr->lr_sbi = sbi;
2951         elr->lr_next_group = start;
2952
2953         /*
2954          * Randomize first schedule time of the request to
2955          * spread the inode table initialization requests
2956          * better.
2957          */
2958         get_random_bytes(&rnd, sizeof(rnd));
2959         elr->lr_next_sched = jiffies + (unsigned long)rnd %
2960                              (EXT4_DEF_LI_MAX_START_DELAY * HZ);
2961
2962         return elr;
2963 }
2964
2965 int ext4_register_li_request(struct super_block *sb,
2966                              ext4_group_t first_not_zeroed)
2967 {
2968         struct ext4_sb_info *sbi = EXT4_SB(sb);
2969         struct ext4_li_request *elr = NULL;
2970         ext4_group_t ngroups = EXT4_SB(sb)->s_groups_count;
2971         int ret = 0;
2972
2973         mutex_lock(&ext4_li_mtx);
2974         if (sbi->s_li_request != NULL) {
2975                 /*
2976                  * Reset timeout so it can be computed again, because
2977                  * s_li_wait_mult might have changed.
2978                  */
2979                 sbi->s_li_request->lr_timeout = 0;
2980                 goto out;
2981         }
2982
2983         if (first_not_zeroed == ngroups ||
2984             (sb->s_flags & MS_RDONLY) ||
2985             !test_opt(sb, INIT_INODE_TABLE))
2986                 goto out;
2987
2988         elr = ext4_li_request_new(sb, first_not_zeroed);
2989         if (!elr) {
2990                 ret = -ENOMEM;
2991                 goto out;
2992         }
2993
2994         if (NULL == ext4_li_info) {
2995                 ret = ext4_li_info_new();
2996                 if (ret)
2997                         goto out;
2998         }
2999
3000         mutex_lock(&ext4_li_info->li_list_mtx);
3001         list_add(&elr->lr_request, &ext4_li_info->li_request_list);
3002         mutex_unlock(&ext4_li_info->li_list_mtx);
3003
3004         sbi->s_li_request = elr;
3005         /*
3006          * set elr to NULL here since it has been inserted to
3007          * the request_list and the removal and free of it is
3008          * handled by ext4_clear_request_list from now on.
3009          */
3010         elr = NULL;
3011
3012         if (!(ext4_li_info->li_state & EXT4_LAZYINIT_RUNNING)) {
3013                 ret = ext4_run_lazyinit_thread();
3014                 if (ret)
3015                         goto out;
3016         }
3017 out:
3018         mutex_unlock(&ext4_li_mtx);
3019         if (ret)
3020                 kfree(elr);
3021         return ret;
3022 }
3023
3024 /*
3025  * We do not need to lock anything since this is called on
3026  * module unload.
3027  */
3028 static void ext4_destroy_lazyinit_thread(void)
3029 {
3030         /*
3031          * If thread exited earlier
3032          * there's nothing to be done.
3033          */
3034         if (!ext4_li_info || !ext4_lazyinit_task)
3035                 return;
3036
3037         kthread_stop(ext4_lazyinit_task);
3038 }
3039
3040 static int set_journal_csum_feature_set(struct super_block *sb)
3041 {
3042         int ret = 1;
3043         int compat, incompat;
3044         struct ext4_sb_info *sbi = EXT4_SB(sb);
3045
3046         if (EXT4_HAS_RO_COMPAT_FEATURE(sb,
3047                                        EXT4_FEATURE_RO_COMPAT_METADATA_CSUM)) {
3048                 /* journal checksum v2 */
3049                 compat = 0;
3050                 incompat = JBD2_FEATURE_INCOMPAT_CSUM_V2;
3051         } else {
3052                 /* journal checksum v1 */
3053                 compat = JBD2_FEATURE_COMPAT_CHECKSUM;
3054                 incompat = 0;
3055         }
3056
3057         if (test_opt(sb, JOURNAL_ASYNC_COMMIT)) {
3058                 ret = jbd2_journal_set_features(sbi->s_journal,
3059                                 compat, 0,
3060                                 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT |
3061                                 incompat);
3062         } else if (test_opt(sb, JOURNAL_CHECKSUM)) {
3063                 ret = jbd2_journal_set_features(sbi->s_journal,
3064                                 compat, 0,
3065                                 incompat);
3066                 jbd2_journal_clear_features(sbi->s_journal, 0, 0,
3067                                 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT);
3068         } else {
3069                 jbd2_journal_clear_features(sbi->s_journal,
3070                                 JBD2_FEATURE_COMPAT_CHECKSUM, 0,
3071                                 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT |
3072                                 JBD2_FEATURE_INCOMPAT_CSUM_V2);
3073         }
3074
3075         return ret;
3076 }
3077
3078 /*
3079  * Note: calculating the overhead so we can be compatible with
3080  * historical BSD practice is quite difficult in the face of
3081  * clusters/bigalloc.  This is because multiple metadata blocks from
3082  * different block group can end up in the same allocation cluster.
3083  * Calculating the exact overhead in the face of clustered allocation
3084  * requires either O(all block bitmaps) in memory or O(number of block
3085  * groups**2) in time.  We will still calculate the superblock for
3086  * older file systems --- and if we come across with a bigalloc file
3087  * system with zero in s_overhead_clusters the estimate will be close to
3088  * correct especially for very large cluster sizes --- but for newer
3089  * file systems, it's better to calculate this figure once at mkfs
3090  * time, and store it in the superblock.  If the superblock value is
3091  * present (even for non-bigalloc file systems), we will use it.
3092  */
3093 static int count_overhead(struct super_block *sb, ext4_group_t grp,
3094                           char *buf)
3095 {
3096         struct ext4_sb_info     *sbi = EXT4_SB(sb);
3097         struct ext4_group_desc  *gdp;
3098         ext4_fsblk_t            first_block, last_block, b;
3099         ext4_group_t            i, ngroups = ext4_get_groups_count(sb);
3100         int                     s, j, count = 0;
3101
3102         if (!EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_BIGALLOC))
3103                 return (ext4_bg_has_super(sb, grp) + ext4_bg_num_gdb(sb, grp) +
3104                         sbi->s_itb_per_group + 2);
3105
3106         first_block = le32_to_cpu(sbi->s_es->s_first_data_block) +
3107                 (grp * EXT4_BLOCKS_PER_GROUP(sb));
3108         last_block = first_block + EXT4_BLOCKS_PER_GROUP(sb) - 1;
3109         for (i = 0; i < ngroups; i++) {
3110                 gdp = ext4_get_group_desc(sb, i, NULL);
3111                 b = ext4_block_bitmap(sb, gdp);
3112                 if (b >= first_block && b <= last_block) {
3113                         ext4_set_bit(EXT4_B2C(sbi, b - first_block), buf);
3114                         count++;
3115                 }
3116                 b = ext4_inode_bitmap(sb, gdp);
3117                 if (b >= first_block && b <= last_block) {
3118                         ext4_set_bit(EXT4_B2C(sbi, b - first_block), buf);
3119                         count++;
3120                 }
3121                 b = ext4_inode_table(sb, gdp);
3122                 if (b >= first_block && b + sbi->s_itb_per_group <= last_block)
3123                         for (j = 0; j < sbi->s_itb_per_group; j++, b++) {
3124                                 int c = EXT4_B2C(sbi, b - first_block);
3125                                 ext4_set_bit(c, buf);
3126                                 count++;
3127                         }
3128                 if (i != grp)
3129                         continue;
3130                 s = 0;
3131                 if (ext4_bg_has_super(sb, grp)) {
3132                         ext4_set_bit(s++, buf);
3133                         count++;
3134                 }
3135                 for (j = ext4_bg_num_gdb(sb, grp); j > 0; j--) {
3136                         ext4_set_bit(EXT4_B2C(sbi, s++), buf);
3137                         count++;
3138                 }
3139         }
3140         if (!count)
3141                 return 0;
3142         return EXT4_CLUSTERS_PER_GROUP(sb) -
3143                 ext4_count_free(buf, EXT4_CLUSTERS_PER_GROUP(sb) / 8);
3144 }
3145
3146 /*
3147  * Compute the overhead and stash it in sbi->s_overhead
3148  */
3149 int ext4_calculate_overhead(struct super_block *sb)
3150 {
3151         struct ext4_sb_info *sbi = EXT4_SB(sb);
3152         struct ext4_super_block *es = sbi->s_es;
3153         ext4_group_t i, ngroups = ext4_get_groups_count(sb);
3154         ext4_fsblk_t overhead = 0;
3155         char *buf = (char *) get_zeroed_page(GFP_KERNEL);
3156
3157         if (!buf)
3158                 return -ENOMEM;
3159
3160         /*
3161          * Compute the overhead (FS structures).  This is constant
3162          * for a given filesystem unless the number of block groups
3163          * changes so we cache the previous value until it does.
3164          */
3165
3166         /*
3167          * All of the blocks before first_data_block are overhead
3168          */
3169         overhead = EXT4_B2C(sbi, le32_to_cpu(es->s_first_data_block));
3170
3171         /*
3172          * Add the overhead found in each block group
3173          */
3174         for (i = 0; i < ngroups; i++) {
3175                 int blks;
3176
3177                 blks = count_overhead(sb, i, buf);
3178                 overhead += blks;
3179                 if (blks)
3180                         memset(buf, 0, PAGE_SIZE);
3181                 cond_resched();
3182         }
3183         /* Add the journal blocks as well */
3184         if (sbi->s_journal)
3185                 overhead += EXT4_NUM_B2C(sbi, sbi->s_journal->j_maxlen);
3186
3187         sbi->s_overhead = overhead;
3188         smp_wmb();
3189         free_page((unsigned long) buf);
3190         return 0;
3191 }
3192
3193 static int ext4_fill_super(struct super_block *sb, void *data, int silent)
3194 {
3195         char *orig_data = kstrdup(data, GFP_KERNEL);
3196         struct buffer_head *bh;
3197         struct ext4_super_block *es = NULL;
3198         struct ext4_sb_info *sbi;
3199         ext4_fsblk_t block;
3200         ext4_fsblk_t sb_block = get_sb_block(&data);
3201         ext4_fsblk_t logical_sb_block;
3202         unsigned long offset = 0;
3203         unsigned long journal_devnum = 0;
3204         unsigned long def_mount_opts;
3205         struct inode *root;
3206         char *cp;
3207         const char *descr;
3208         int ret = -ENOMEM;
3209         int blocksize, clustersize;
3210         unsigned int db_count;
3211         unsigned int i;
3212         int needs_recovery, has_huge_files, has_bigalloc;
3213         __u64 blocks_count;
3214         int err = 0;
3215         unsigned int journal_ioprio = DEFAULT_JOURNAL_IOPRIO;
3216         ext4_group_t first_not_zeroed;
3217
3218         sbi = kzalloc(sizeof(*sbi), GFP_KERNEL);
3219         if (!sbi)
3220                 goto out_free_orig;
3221
3222         sbi->s_blockgroup_lock =
3223                 kzalloc(sizeof(struct blockgroup_lock), GFP_KERNEL);
3224         if (!sbi->s_blockgroup_lock) {
3225                 kfree(sbi);
3226                 goto out_free_orig;
3227         }
3228         sb->s_fs_info = sbi;
3229         sbi->s_sb = sb;
3230         sbi->s_inode_readahead_blks = EXT4_DEF_INODE_READAHEAD_BLKS;
3231         sbi->s_sb_block = sb_block;
3232         if (sb->s_bdev->bd_part)
3233                 sbi->s_sectors_written_start =
3234                         part_stat_read(sb->s_bdev->bd_part, sectors[1]);
3235
3236         /* Cleanup superblock name */
3237         for (cp = sb->s_id; (cp = strchr(cp, '/'));)
3238                 *cp = '!';
3239
3240         /* -EINVAL is default */
3241         ret = -EINVAL;
3242         blocksize = sb_min_blocksize(sb, EXT4_MIN_BLOCK_SIZE);
3243         if (!blocksize) {
3244                 ext4_msg(sb, KERN_ERR, "unable to set blocksize");
3245                 goto out_fail;
3246         }
3247
3248         /*
3249          * The ext4 superblock will not be buffer aligned for other than 1kB
3250          * block sizes.  We need to calculate the offset from buffer start.
3251          */
3252         if (blocksize != EXT4_MIN_BLOCK_SIZE) {
3253                 logical_sb_block = sb_block * EXT4_MIN_BLOCK_SIZE;
3254                 offset = do_div(logical_sb_block, blocksize);
3255         } else {
3256                 logical_sb_block = sb_block;
3257         }
3258
3259         if (!(bh = sb_bread(sb, logical_sb_block))) {
3260                 ext4_msg(sb, KERN_ERR, "unable to read superblock");
3261                 goto out_fail;
3262         }
3263         /*
3264          * Note: s_es must be initialized as soon as possible because
3265          *       some ext4 macro-instructions depend on its value
3266          */
3267         es = (struct ext4_super_block *) (bh->b_data + offset);
3268         sbi->s_es = es;
3269         sb->s_magic = le16_to_cpu(es->s_magic);
3270         if (sb->s_magic != EXT4_SUPER_MAGIC)
3271                 goto cantfind_ext4;
3272         sbi->s_kbytes_written = le64_to_cpu(es->s_kbytes_written);
3273
3274         /* Warn if metadata_csum and gdt_csum are both set. */
3275         if (EXT4_HAS_RO_COMPAT_FEATURE(sb,
3276                                        EXT4_FEATURE_RO_COMPAT_METADATA_CSUM) &&
3277             EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_GDT_CSUM))
3278                 ext4_warning(sb, KERN_INFO "metadata_csum and uninit_bg are "
3279                              "redundant flags; please run fsck.");
3280
3281         /* Check for a known checksum algorithm */
3282         if (!ext4_verify_csum_type(sb, es)) {
3283                 ext4_msg(sb, KERN_ERR, "VFS: Found ext4 filesystem with "
3284                          "unknown checksum algorithm.");
3285                 silent = 1;
3286                 goto cantfind_ext4;
3287         }
3288
3289         /* Load the checksum driver */
3290         if (EXT4_HAS_RO_COMPAT_FEATURE(sb,
3291                                        EXT4_FEATURE_RO_COMPAT_METADATA_CSUM)) {
3292                 sbi->s_chksum_driver = crypto_alloc_shash("crc32c", 0, 0);
3293                 if (IS_ERR(sbi->s_chksum_driver)) {
3294                         ext4_msg(sb, KERN_ERR, "Cannot load crc32c driver.");
3295                         ret = PTR_ERR(sbi->s_chksum_driver);
3296                         sbi->s_chksum_driver = NULL;
3297                         goto failed_mount;
3298                 }
3299         }
3300
3301         /* Check superblock checksum */
3302         if (!ext4_superblock_csum_verify(sb, es)) {
3303                 ext4_msg(sb, KERN_ERR, "VFS: Found ext4 filesystem with "
3304                          "invalid superblock checksum.  Run e2fsck?");
3305                 silent = 1;
3306                 goto cantfind_ext4;
3307         }
3308
3309         /* Precompute checksum seed for all metadata */
3310         if (EXT4_HAS_RO_COMPAT_FEATURE(sb,
3311                         EXT4_FEATURE_RO_COMPAT_METADATA_CSUM))
3312                 sbi->s_csum_seed = ext4_chksum(sbi, ~0, es->s_uuid,
3313                                                sizeof(es->s_uuid));
3314
3315         /* Set defaults before we parse the mount options */
3316         def_mount_opts = le32_to_cpu(es->s_default_mount_opts);
3317         set_opt(sb, INIT_INODE_TABLE);
3318         if (def_mount_opts & EXT4_DEFM_DEBUG)
3319                 set_opt(sb, DEBUG);
3320         if (def_mount_opts & EXT4_DEFM_BSDGROUPS)
3321                 set_opt(sb, GRPID);
3322         if (def_mount_opts & EXT4_DEFM_UID16)
3323                 set_opt(sb, NO_UID32);
3324         /* xattr user namespace & acls are now defaulted on */
3325         set_opt(sb, XATTR_USER);
3326 #ifdef CONFIG_EXT4_FS_POSIX_ACL
3327         set_opt(sb, POSIX_ACL);
3328 #endif
3329         if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_DATA)
3330                 set_opt(sb, JOURNAL_DATA);
3331         else if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_ORDERED)
3332                 set_opt(sb, ORDERED_DATA);
3333         else if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_WBACK)
3334                 set_opt(sb, WRITEBACK_DATA);
3335
3336         if (le16_to_cpu(sbi->s_es->s_errors) == EXT4_ERRORS_PANIC)
3337                 set_opt(sb, ERRORS_PANIC);
3338         else if (le16_to_cpu(sbi->s_es->s_errors) == EXT4_ERRORS_CONTINUE)
3339                 set_opt(sb, ERRORS_CONT);
3340         else
3341                 set_opt(sb, ERRORS_RO);
3342         if (def_mount_opts & EXT4_DEFM_BLOCK_VALIDITY)
3343                 set_opt(sb, BLOCK_VALIDITY);
3344         if (def_mount_opts & EXT4_DEFM_DISCARD)
3345                 set_opt(sb, DISCARD);
3346
3347         sbi->s_resuid = make_kuid(&init_user_ns, le16_to_cpu(es->s_def_resuid));
3348         sbi->s_resgid = make_kgid(&init_user_ns, le16_to_cpu(es->s_def_resgid));
3349         sbi->s_commit_interval = JBD2_DEFAULT_MAX_COMMIT_AGE * HZ;
3350         sbi->s_min_batch_time = EXT4_DEF_MIN_BATCH_TIME;
3351         sbi->s_max_batch_time = EXT4_DEF_MAX_BATCH_TIME;
3352
3353         if ((def_mount_opts & EXT4_DEFM_NOBARRIER) == 0)
3354                 set_opt(sb, BARRIER);
3355
3356         /*
3357          * enable delayed allocation by default
3358          * Use -o nodelalloc to turn it off
3359          */
3360         if (!IS_EXT3_SB(sb) && !IS_EXT2_SB(sb) &&
3361             ((def_mount_opts & EXT4_DEFM_NODELALLOC) == 0))
3362                 set_opt(sb, DELALLOC);
3363
3364         /*
3365          * set default s_li_wait_mult for lazyinit, for the case there is
3366          * no mount option specified.
3367          */
3368         sbi->s_li_wait_mult = EXT4_DEF_LI_WAIT_MULT;
3369
3370         if (!parse_options((char *) sbi->s_es->s_mount_opts, sb,
3371                            &journal_devnum, &journal_ioprio, 0)) {
3372                 ext4_msg(sb, KERN_WARNING,
3373                          "failed to parse options in superblock: %s",
3374                          sbi->s_es->s_mount_opts);
3375         }
3376         sbi->s_def_mount_opt = sbi->s_mount_opt;
3377         if (!parse_options((char *) data, sb, &journal_devnum,
3378                            &journal_ioprio, 0))
3379                 goto failed_mount;
3380
3381         if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA) {
3382                 printk_once(KERN_WARNING "EXT4-fs: Warning: mounting "
3383                             "with data=journal disables delayed "
3384                             "allocation and O_DIRECT support!\n");
3385                 if (test_opt2(sb, EXPLICIT_DELALLOC)) {
3386                         ext4_msg(sb, KERN_ERR, "can't mount with "
3387                                  "both data=journal and delalloc");
3388                         goto failed_mount;
3389                 }
3390                 if (test_opt(sb, DIOREAD_NOLOCK)) {
3391                         ext4_msg(sb, KERN_ERR, "can't mount with "
3392                                  "both data=journal and delalloc");
3393                         goto failed_mount;
3394                 }
3395                 if (test_opt(sb, DELALLOC))
3396                         clear_opt(sb, DELALLOC);
3397         }
3398
3399         sb->s_flags = (sb->s_flags & ~MS_POSIXACL) |
3400                 (test_opt(sb, POSIX_ACL) ? MS_POSIXACL : 0);
3401
3402         if (le32_to_cpu(es->s_rev_level) == EXT4_GOOD_OLD_REV &&
3403             (EXT4_HAS_COMPAT_FEATURE(sb, ~0U) ||
3404              EXT4_HAS_RO_COMPAT_FEATURE(sb, ~0U) ||
3405              EXT4_HAS_INCOMPAT_FEATURE(sb, ~0U)))
3406                 ext4_msg(sb, KERN_WARNING,
3407                        "feature flags set on rev 0 fs, "
3408                        "running e2fsck is recommended");
3409
3410         if (IS_EXT2_SB(sb)) {
3411                 if (ext2_feature_set_ok(sb))
3412                         ext4_msg(sb, KERN_INFO, "mounting ext2 file system "
3413                                  "using the ext4 subsystem");
3414                 else {
3415                         ext4_msg(sb, KERN_ERR, "couldn't mount as ext2 due "
3416                                  "to feature incompatibilities");
3417                         goto failed_mount;
3418                 }
3419         }
3420
3421         if (IS_EXT3_SB(sb)) {
3422                 if (ext3_feature_set_ok(sb))
3423                         ext4_msg(sb, KERN_INFO, "mounting ext3 file system "
3424                                  "using the ext4 subsystem");
3425                 else {
3426                         ext4_msg(sb, KERN_ERR, "couldn't mount as ext3 due "
3427                                  "to feature incompatibilities");
3428                         goto failed_mount;
3429                 }
3430         }
3431
3432         /*
3433          * Check feature flags regardless of the revision level, since we
3434          * previously didn't change the revision level when setting the flags,
3435          * so there is a chance incompat flags are set on a rev 0 filesystem.
3436          */
3437         if (!ext4_feature_set_ok(sb, (sb->s_flags & MS_RDONLY)))
3438                 goto failed_mount;
3439
3440         blocksize = BLOCK_SIZE << le32_to_cpu(es->s_log_block_size);
3441         if (blocksize < EXT4_MIN_BLOCK_SIZE ||
3442             blocksize > EXT4_MAX_BLOCK_SIZE) {
3443                 ext4_msg(sb, KERN_ERR,
3444                        "Unsupported filesystem blocksize %d", blocksize);
3445                 goto failed_mount;
3446         }
3447
3448         if (sb->s_blocksize != blocksize) {
3449                 /* Validate the filesystem blocksize */
3450                 if (!sb_set_blocksize(sb, blocksize)) {
3451                         ext4_msg(sb, KERN_ERR, "bad block size %d",
3452                                         blocksize);
3453                         goto failed_mount;
3454                 }
3455
3456                 brelse(bh);
3457                 logical_sb_block = sb_block * EXT4_MIN_BLOCK_SIZE;
3458                 offset = do_div(logical_sb_block, blocksize);
3459                 bh = sb_bread(sb, logical_sb_block);
3460                 if (!bh) {
3461                         ext4_msg(sb, KERN_ERR,
3462                                "Can't read superblock on 2nd try");
3463                         goto failed_mount;
3464                 }
3465                 es = (struct ext4_super_block *)(bh->b_data + offset);
3466                 sbi->s_es = es;
3467                 if (es->s_magic != cpu_to_le16(EXT4_SUPER_MAGIC)) {
3468                         ext4_msg(sb, KERN_ERR,
3469                                "Magic mismatch, very weird!");
3470                         goto failed_mount;
3471                 }
3472         }
3473
3474         has_huge_files = EXT4_HAS_RO_COMPAT_FEATURE(sb,
3475                                 EXT4_FEATURE_RO_COMPAT_HUGE_FILE);
3476         sbi->s_bitmap_maxbytes = ext4_max_bitmap_size(sb->s_blocksize_bits,
3477                                                       has_huge_files);
3478         sb->s_maxbytes = ext4_max_size(sb->s_blocksize_bits, has_huge_files);
3479
3480         if (le32_to_cpu(es->s_rev_level) == EXT4_GOOD_OLD_REV) {
3481                 sbi->s_inode_size = EXT4_GOOD_OLD_INODE_SIZE;
3482                 sbi->s_first_ino = EXT4_GOOD_OLD_FIRST_INO;
3483         } else {
3484                 sbi->s_inode_size = le16_to_cpu(es->s_inode_size);
3485                 sbi->s_first_ino = le32_to_cpu(es->s_first_ino);
3486                 if ((sbi->s_inode_size < EXT4_GOOD_OLD_INODE_SIZE) ||
3487                     (!is_power_of_2(sbi->s_inode_size)) ||
3488                     (sbi->s_inode_size > blocksize)) {
3489                         ext4_msg(sb, KERN_ERR,
3490                                "unsupported inode size: %d",
3491                                sbi->s_inode_size);
3492                         goto failed_mount;
3493                 }
3494                 if (sbi->s_inode_size > EXT4_GOOD_OLD_INODE_SIZE)
3495                         sb->s_time_gran = 1 << (EXT4_EPOCH_BITS - 2);
3496         }
3497
3498         sbi->s_desc_size = le16_to_cpu(es->s_desc_size);
3499         if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_64BIT)) {
3500                 if (sbi->s_desc_size < EXT4_MIN_DESC_SIZE_64BIT ||
3501                     sbi->s_desc_size > EXT4_MAX_DESC_SIZE ||
3502                     !is_power_of_2(sbi->s_desc_size)) {
3503                         ext4_msg(sb, KERN_ERR,
3504                                "unsupported descriptor size %lu",
3505                                sbi->s_desc_size);
3506                         goto failed_mount;
3507                 }
3508         } else
3509                 sbi->s_desc_size = EXT4_MIN_DESC_SIZE;
3510
3511         sbi->s_blocks_per_group = le32_to_cpu(es->s_blocks_per_group);
3512         sbi->s_inodes_per_group = le32_to_cpu(es->s_inodes_per_group);
3513         if (EXT4_INODE_SIZE(sb) == 0 || EXT4_INODES_PER_GROUP(sb) == 0)
3514                 goto cantfind_ext4;
3515
3516         sbi->s_inodes_per_block = blocksize / EXT4_INODE_SIZE(sb);
3517         if (sbi->s_inodes_per_block == 0)
3518                 goto cantfind_ext4;
3519         sbi->s_itb_per_group = sbi->s_inodes_per_group /
3520                                         sbi->s_inodes_per_block;
3521         sbi->s_desc_per_block = blocksize / EXT4_DESC_SIZE(sb);
3522         sbi->s_sbh = bh;
3523         sbi->s_mount_state = le16_to_cpu(es->s_state);
3524         sbi->s_addr_per_block_bits = ilog2(EXT4_ADDR_PER_BLOCK(sb));
3525         sbi->s_desc_per_block_bits = ilog2(EXT4_DESC_PER_BLOCK(sb));
3526
3527         for (i = 0; i < 4; i++)
3528                 sbi->s_hash_seed[i] = le32_to_cpu(es->s_hash_seed[i]);
3529         sbi->s_def_hash_version = es->s_def_hash_version;
3530         i = le32_to_cpu(es->s_flags);
3531         if (i & EXT2_FLAGS_UNSIGNED_HASH)
3532                 sbi->s_hash_unsigned = 3;
3533         else if ((i & EXT2_FLAGS_SIGNED_HASH) == 0) {
3534 #ifdef __CHAR_UNSIGNED__
3535                 es->s_flags |= cpu_to_le32(EXT2_FLAGS_UNSIGNED_HASH);
3536                 sbi->s_hash_unsigned = 3;
3537 #else
3538                 es->s_flags |= cpu_to_le32(EXT2_FLAGS_SIGNED_HASH);
3539 #endif
3540         }
3541
3542         /* Handle clustersize */
3543         clustersize = BLOCK_SIZE << le32_to_cpu(es->s_log_cluster_size);
3544         has_bigalloc = EXT4_HAS_RO_COMPAT_FEATURE(sb,
3545                                 EXT4_FEATURE_RO_COMPAT_BIGALLOC);
3546         if (has_bigalloc) {
3547                 if (clustersize < blocksize) {
3548                         ext4_msg(sb, KERN_ERR,
3549                                  "cluster size (%d) smaller than "
3550                                  "block size (%d)", clustersize, blocksize);
3551                         goto failed_mount;
3552                 }
3553                 sbi->s_cluster_bits = le32_to_cpu(es->s_log_cluster_size) -
3554                         le32_to_cpu(es->s_log_block_size);
3555                 sbi->s_clusters_per_group =
3556                         le32_to_cpu(es->s_clusters_per_group);
3557                 if (sbi->s_clusters_per_group > blocksize * 8) {
3558                         ext4_msg(sb, KERN_ERR,
3559                                  "#clusters per group too big: %lu",
3560                                  sbi->s_clusters_per_group);
3561                         goto failed_mount;
3562                 }
3563                 if (sbi->s_blocks_per_group !=
3564                     (sbi->s_clusters_per_group * (clustersize / blocksize))) {
3565                         ext4_msg(sb, KERN_ERR, "blocks per group (%lu) and "
3566                                  "clusters per group (%lu) inconsistent",
3567                                  sbi->s_blocks_per_group,
3568                                  sbi->s_clusters_per_group);
3569                         goto failed_mount;
3570                 }
3571         } else {
3572                 if (clustersize != blocksize) {
3573                         ext4_warning(sb, "fragment/cluster size (%d) != "
3574                                      "block size (%d)", clustersize,
3575                                      blocksize);
3576                         clustersize = blocksize;
3577                 }
3578                 if (sbi->s_blocks_per_group > blocksize * 8) {
3579                         ext4_msg(sb, KERN_ERR,
3580                                  "#blocks per group too big: %lu",
3581                                  sbi->s_blocks_per_group);
3582                         goto failed_mount;
3583                 }
3584                 sbi->s_clusters_per_group = sbi->s_blocks_per_group;
3585                 sbi->s_cluster_bits = 0;
3586         }
3587         sbi->s_cluster_ratio = clustersize / blocksize;
3588
3589         if (sbi->s_inodes_per_group > blocksize * 8) {
3590                 ext4_msg(sb, KERN_ERR,
3591                        "#inodes per group too big: %lu",
3592                        sbi->s_inodes_per_group);
3593                 goto failed_mount;
3594         }
3595
3596         /*
3597          * Test whether we have more sectors than will fit in sector_t,
3598          * and whether the max offset is addressable by the page cache.
3599          */
3600         err = generic_check_addressable(sb->s_blocksize_bits,
3601                                         ext4_blocks_count(es));
3602         if (err) {
3603                 ext4_msg(sb, KERN_ERR, "filesystem"
3604                          " too large to mount safely on this system");
3605                 if (sizeof(sector_t) < 8)
3606                         ext4_msg(sb, KERN_WARNING, "CONFIG_LBDAF not enabled");
3607                 goto failed_mount;
3608         }
3609
3610         if (EXT4_BLOCKS_PER_GROUP(sb) == 0)
3611                 goto cantfind_ext4;
3612
3613         /* check blocks count against device size */
3614         blocks_count = sb->s_bdev->bd_inode->i_size >> sb->s_blocksize_bits;
3615         if (blocks_count && ext4_blocks_count(es) > blocks_count) {
3616                 ext4_msg(sb, KERN_WARNING, "bad geometry: block count %llu "
3617                        "exceeds size of device (%llu blocks)",
3618                        ext4_blocks_count(es), blocks_count);
3619                 goto failed_mount;
3620         }
3621
3622         /*
3623          * It makes no sense for the first data block to be beyond the end
3624          * of the filesystem.
3625          */
3626         if (le32_to_cpu(es->s_first_data_block) >= ext4_blocks_count(es)) {
3627                 ext4_msg(sb, KERN_WARNING, "bad geometry: first data "
3628                          "block %u is beyond end of filesystem (%llu)",
3629                          le32_to_cpu(es->s_first_data_block),
3630                          ext4_blocks_count(es));
3631                 goto failed_mount;
3632         }
3633         blocks_count = (ext4_blocks_count(es) -
3634                         le32_to_cpu(es->s_first_data_block) +
3635                         EXT4_BLOCKS_PER_GROUP(sb) - 1);
3636         do_div(blocks_count, EXT4_BLOCKS_PER_GROUP(sb));
3637         if (blocks_count > ((uint64_t)1<<32) - EXT4_DESC_PER_BLOCK(sb)) {
3638                 ext4_msg(sb, KERN_WARNING, "groups count too large: %u "
3639                        "(block count %llu, first data block %u, "
3640                        "blocks per group %lu)", sbi->s_groups_count,
3641                        ext4_blocks_count(es),
3642                        le32_to_cpu(es->s_first_data_block),
3643                        EXT4_BLOCKS_PER_GROUP(sb));
3644                 goto failed_mount;
3645         }
3646         sbi->s_groups_count = blocks_count;
3647         sbi->s_blockfile_groups = min_t(ext4_group_t, sbi->s_groups_count,
3648                         (EXT4_MAX_BLOCK_FILE_PHYS / EXT4_BLOCKS_PER_GROUP(sb)));
3649         db_count = (sbi->s_groups_count + EXT4_DESC_PER_BLOCK(sb) - 1) /
3650                    EXT4_DESC_PER_BLOCK(sb);
3651         sbi->s_group_desc = ext4_kvmalloc(db_count *
3652                                           sizeof(struct buffer_head *),
3653                                           GFP_KERNEL);
3654         if (sbi->s_group_desc == NULL) {
3655                 ext4_msg(sb, KERN_ERR, "not enough memory");
3656                 ret = -ENOMEM;
3657                 goto failed_mount;
3658         }
3659
3660         if (ext4_proc_root)
3661                 sbi->s_proc = proc_mkdir(sb->s_id, ext4_proc_root);
3662
3663         if (sbi->s_proc)
3664                 proc_create_data("options", S_IRUGO, sbi->s_proc,
3665                                  &ext4_seq_options_fops, sb);
3666
3667         bgl_lock_init(sbi->s_blockgroup_lock);
3668
3669         for (i = 0; i < db_count; i++) {
3670                 block = descriptor_loc(sb, logical_sb_block, i);
3671                 sbi->s_group_desc[i] = sb_bread(sb, block);
3672                 if (!sbi->s_group_desc[i]) {
3673                         ext4_msg(sb, KERN_ERR,
3674                                "can't read group descriptor %d", i);
3675                         db_count = i;
3676                         goto failed_mount2;
3677                 }
3678         }
3679         if (!ext4_check_descriptors(sb, &first_not_zeroed)) {
3680                 ext4_msg(sb, KERN_ERR, "group descriptors corrupted!");
3681                 goto failed_mount2;
3682         }
3683         if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_FLEX_BG))
3684                 if (!ext4_fill_flex_info(sb)) {
3685                         ext4_msg(sb, KERN_ERR,
3686                                "unable to initialize "
3687                                "flex_bg meta info!");
3688                         goto failed_mount2;
3689                 }
3690
3691         sbi->s_gdb_count = db_count;
3692         get_random_bytes(&sbi->s_next_generation, sizeof(u32));
3693         spin_lock_init(&sbi->s_next_gen_lock);
3694
3695         init_timer(&sbi->s_err_report);
3696         sbi->s_err_report.function = print_daily_error_info;
3697         sbi->s_err_report.data = (unsigned long) sb;
3698
3699         err = percpu_counter_init(&sbi->s_freeclusters_counter,
3700                         ext4_count_free_clusters(sb));
3701         if (!err) {
3702                 err = percpu_counter_init(&sbi->s_freeinodes_counter,
3703                                 ext4_count_free_inodes(sb));
3704         }
3705         if (!err) {
3706                 err = percpu_counter_init(&sbi->s_dirs_counter,
3707                                 ext4_count_dirs(sb));
3708         }
3709         if (!err) {
3710                 err = percpu_counter_init(&sbi->s_dirtyclusters_counter, 0);
3711         }
3712         if (!err) {
3713                 err = percpu_counter_init(&sbi->s_extent_cache_cnt, 0);
3714         }
3715         if (err) {
3716                 ext4_msg(sb, KERN_ERR, "insufficient memory");
3717                 goto failed_mount3;
3718         }
3719
3720         sbi->s_stripe = ext4_get_stripe_size(sbi);
3721         sbi->s_max_writeback_mb_bump = 128;
3722         sbi->s_extent_max_zeroout_kb = 32;
3723
3724         /* Register extent status tree shrinker */
3725         ext4_es_register_shrinker(sb);
3726
3727         /*
3728          * set up enough so that it can read an inode
3729          */
3730         if (!test_opt(sb, NOLOAD) &&
3731             EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL))
3732                 sb->s_op = &ext4_sops;
3733         else
3734                 sb->s_op = &ext4_nojournal_sops;
3735         sb->s_export_op = &ext4_export_ops;
3736         sb->s_xattr = ext4_xattr_handlers;
3737 #ifdef CONFIG_QUOTA
3738         sb->dq_op = &ext4_quota_operations;
3739         if (EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_QUOTA))
3740                 sb->s_qcop = &ext4_qctl_sysfile_operations;
3741         else
3742                 sb->s_qcop = &ext4_qctl_operations;
3743 #endif
3744         memcpy(sb->s_uuid, es->s_uuid, sizeof(es->s_uuid));
3745
3746         INIT_LIST_HEAD(&sbi->s_orphan); /* unlinked but open files */
3747         mutex_init(&sbi->s_orphan_lock);
3748
3749         sb->s_root = NULL;
3750
3751         needs_recovery = (es->s_last_orphan != 0 ||
3752                           EXT4_HAS_INCOMPAT_FEATURE(sb,
3753                                     EXT4_FEATURE_INCOMPAT_RECOVER));
3754
3755         if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_MMP) &&
3756             !(sb->s_flags & MS_RDONLY))
3757                 if (ext4_multi_mount_protect(sb, le64_to_cpu(es->s_mmp_block)))
3758                         goto failed_mount3;
3759
3760         /*
3761          * The first inode we look at is the journal inode.  Don't try
3762          * root first: it may be modified in the journal!
3763          */
3764         if (!test_opt(sb, NOLOAD) &&
3765             EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL)) {
3766                 if (ext4_load_journal(sb, es, journal_devnum))
3767                         goto failed_mount3;
3768         } else if (test_opt(sb, NOLOAD) && !(sb->s_flags & MS_RDONLY) &&
3769               EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER)) {
3770                 ext4_msg(sb, KERN_ERR, "required journal recovery "
3771                        "suppressed and not mounted read-only");
3772                 goto failed_mount_wq;
3773         } else {
3774                 clear_opt(sb, DATA_FLAGS);
3775                 sbi->s_journal = NULL;
3776                 needs_recovery = 0;
3777                 goto no_journal;
3778         }
3779
3780         if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_64BIT) &&
3781             !jbd2_journal_set_features(EXT4_SB(sb)->s_journal, 0, 0,
3782                                        JBD2_FEATURE_INCOMPAT_64BIT)) {
3783                 ext4_msg(sb, KERN_ERR, "Failed to set 64-bit journal feature");
3784                 goto failed_mount_wq;
3785         }
3786
3787         if (!set_journal_csum_feature_set(sb)) {
3788                 ext4_msg(sb, KERN_ERR, "Failed to set journal checksum "
3789                          "feature set");
3790                 goto failed_mount_wq;
3791         }
3792
3793         /* We have now updated the journal if required, so we can
3794          * validate the data journaling mode. */
3795         switch (test_opt(sb, DATA_FLAGS)) {
3796         case 0:
3797                 /* No mode set, assume a default based on the journal
3798                  * capabilities: ORDERED_DATA if the journal can
3799                  * cope, else JOURNAL_DATA
3800                  */
3801                 if (jbd2_journal_check_available_features
3802                     (sbi->s_journal, 0, 0, JBD2_FEATURE_INCOMPAT_REVOKE))
3803                         set_opt(sb, ORDERED_DATA);
3804                 else
3805                         set_opt(sb, JOURNAL_DATA);
3806                 break;
3807
3808         case EXT4_MOUNT_ORDERED_DATA:
3809         case EXT4_MOUNT_WRITEBACK_DATA:
3810                 if (!jbd2_journal_check_available_features
3811                     (sbi->s_journal, 0, 0, JBD2_FEATURE_INCOMPAT_REVOKE)) {
3812                         ext4_msg(sb, KERN_ERR, "Journal does not support "
3813                                "requested data journaling mode");
3814                         goto failed_mount_wq;
3815                 }
3816         default:
3817                 break;
3818         }
3819         set_task_ioprio(sbi->s_journal->j_task, journal_ioprio);
3820
3821         sbi->s_journal->j_commit_callback = ext4_journal_commit_callback;
3822
3823         /*
3824          * The journal may have updated the bg summary counts, so we
3825          * need to update the global counters.
3826          */
3827         percpu_counter_set(&sbi->s_freeclusters_counter,
3828                            ext4_count_free_clusters(sb));
3829         percpu_counter_set(&sbi->s_freeinodes_counter,
3830                            ext4_count_free_inodes(sb));
3831         percpu_counter_set(&sbi->s_dirs_counter,
3832                            ext4_count_dirs(sb));
3833         percpu_counter_set(&sbi->s_dirtyclusters_counter, 0);
3834
3835 no_journal:
3836         /*
3837          * Get the # of file system overhead blocks from the
3838          * superblock if present.
3839          */
3840         if (es->s_overhead_clusters)
3841                 sbi->s_overhead = le32_to_cpu(es->s_overhead_clusters);
3842         else {
3843                 err = ext4_calculate_overhead(sb);
3844                 if (err)
3845                         goto failed_mount_wq;
3846         }
3847
3848         /*
3849          * The maximum number of concurrent works can be high and
3850          * concurrency isn't really necessary.  Limit it to 1.
3851          */
3852         EXT4_SB(sb)->dio_unwritten_wq =
3853                 alloc_workqueue("ext4-dio-unwritten", WQ_MEM_RECLAIM | WQ_UNBOUND, 1);
3854         if (!EXT4_SB(sb)->dio_unwritten_wq) {
3855                 printk(KERN_ERR "EXT4-fs: failed to create DIO workqueue\n");
3856                 ret = -ENOMEM;
3857                 goto failed_mount_wq;
3858         }
3859
3860         /*
3861          * The jbd2_journal_load will have done any necessary log recovery,
3862          * so we can safely mount the rest of the filesystem now.
3863          */
3864
3865         root = ext4_iget(sb, EXT4_ROOT_INO);
3866         if (IS_ERR(root)) {
3867                 ext4_msg(sb, KERN_ERR, "get root inode failed");
3868                 ret = PTR_ERR(root);
3869                 root = NULL;
3870                 goto failed_mount4;
3871         }
3872         if (!S_ISDIR(root->i_mode) || !root->i_blocks || !root->i_size) {
3873                 ext4_msg(sb, KERN_ERR, "corrupt root inode, run e2fsck");
3874                 iput(root);
3875                 goto failed_mount4;
3876         }
3877         sb->s_root = d_make_root(root);
3878         if (!sb->s_root) {
3879                 ext4_msg(sb, KERN_ERR, "get root dentry failed");
3880                 ret = -ENOMEM;
3881                 goto failed_mount4;
3882         }
3883
3884         if (ext4_setup_super(sb, es, sb->s_flags & MS_RDONLY))
3885                 sb->s_flags |= MS_RDONLY;
3886
3887         /* determine the minimum size of new large inodes, if present */
3888         if (sbi->s_inode_size > EXT4_GOOD_OLD_INODE_SIZE) {
3889                 sbi->s_want_extra_isize = sizeof(struct ext4_inode) -
3890                                                      EXT4_GOOD_OLD_INODE_SIZE;
3891                 if (EXT4_HAS_RO_COMPAT_FEATURE(sb,
3892                                        EXT4_FEATURE_RO_COMPAT_EXTRA_ISIZE)) {
3893                         if (sbi->s_want_extra_isize <
3894                             le16_to_cpu(es->s_want_extra_isize))
3895                                 sbi->s_want_extra_isize =
3896                                         le16_to_cpu(es->s_want_extra_isize);
3897                         if (sbi->s_want_extra_isize <
3898                             le16_to_cpu(es->s_min_extra_isize))
3899                                 sbi->s_want_extra_isize =
3900                                         le16_to_cpu(es->s_min_extra_isize);
3901                 }
3902         }
3903         /* Check if enough inode space is available */
3904         if (EXT4_GOOD_OLD_INODE_SIZE + sbi->s_want_extra_isize >
3905                                                         sbi->s_inode_size) {
3906                 sbi->s_want_extra_isize = sizeof(struct ext4_inode) -
3907                                                        EXT4_GOOD_OLD_INODE_SIZE;
3908                 ext4_msg(sb, KERN_INFO, "required extra inode space not"
3909                          "available");
3910         }
3911
3912         err = ext4_setup_system_zone(sb);
3913         if (err) {
3914                 ext4_msg(sb, KERN_ERR, "failed to initialize system "
3915                          "zone (%d)", err);
3916                 goto failed_mount4a;
3917         }
3918
3919         ext4_ext_init(sb);
3920         err = ext4_mb_init(sb);
3921         if (err) {
3922                 ext4_msg(sb, KERN_ERR, "failed to initialize mballoc (%d)",
3923                          err);
3924                 goto failed_mount5;
3925         }
3926
3927         err = ext4_register_li_request(sb, first_not_zeroed);
3928         if (err)
3929                 goto failed_mount6;
3930
3931         sbi->s_kobj.kset = ext4_kset;
3932         init_completion(&sbi->s_kobj_unregister);
3933         err = kobject_init_and_add(&sbi->s_kobj, &ext4_ktype, NULL,
3934                                    "%s", sb->s_id);
3935         if (err)
3936                 goto failed_mount7;
3937
3938 #ifdef CONFIG_QUOTA
3939         /* Enable quota usage during mount. */
3940         if (EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_QUOTA) &&
3941             !(sb->s_flags & MS_RDONLY)) {
3942                 err = ext4_enable_quotas(sb);
3943                 if (err)
3944                         goto failed_mount8;
3945         }
3946 #endif  /* CONFIG_QUOTA */
3947
3948         EXT4_SB(sb)->s_mount_state |= EXT4_ORPHAN_FS;
3949         ext4_orphan_cleanup(sb, es);
3950         EXT4_SB(sb)->s_mount_state &= ~EXT4_ORPHAN_FS;
3951         if (needs_recovery) {
3952                 ext4_msg(sb, KERN_INFO, "recovery complete");
3953                 ext4_mark_recovery_complete(sb, es);
3954         }
3955         if (EXT4_SB(sb)->s_journal) {
3956                 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA)
3957                         descr = " journalled data mode";
3958                 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA)
3959                         descr = " ordered data mode";
3960                 else
3961                         descr = " writeback data mode";
3962         } else
3963                 descr = "out journal";
3964
3965         if (test_opt(sb, DISCARD)) {
3966                 struct request_queue *q = bdev_get_queue(sb->s_bdev);
3967                 if (!blk_queue_discard(q))
3968                         ext4_msg(sb, KERN_WARNING,
3969                                  "mounting with \"discard\" option, but "
3970                                  "the device does not support discard");
3971         }
3972
3973         ext4_msg(sb, KERN_INFO, "mounted filesystem with%s. "
3974                  "Opts: %s%s%s", descr, sbi->s_es->s_mount_opts,
3975                  *sbi->s_es->s_mount_opts ? "; " : "", orig_data);
3976
3977         if (es->s_error_count)
3978                 mod_timer(&sbi->s_err_report, jiffies + 300*HZ); /* 5 minutes */
3979
3980         kfree(orig_data);
3981         return 0;
3982
3983 cantfind_ext4:
3984         if (!silent)
3985                 ext4_msg(sb, KERN_ERR, "VFS: Can't find ext4 filesystem");
3986         goto failed_mount;
3987
3988 #ifdef CONFIG_QUOTA
3989 failed_mount8:
3990         kobject_del(&sbi->s_kobj);
3991 #endif
3992 failed_mount7:
3993         ext4_unregister_li_request(sb);
3994 failed_mount6:
3995         ext4_mb_release(sb);
3996 failed_mount5:
3997         ext4_ext_release(sb);
3998         ext4_release_system_zone(sb);
3999 failed_mount4a:
4000         dput(sb->s_root);
4001         sb->s_root = NULL;
4002 failed_mount4:
4003         ext4_msg(sb, KERN_ERR, "mount failed");
4004         destroy_workqueue(EXT4_SB(sb)->dio_unwritten_wq);
4005 failed_mount_wq:
4006         if (sbi->s_journal) {
4007                 jbd2_journal_destroy(sbi->s_journal);
4008                 sbi->s_journal = NULL;
4009         }
4010 failed_mount3:
4011         del_timer(&sbi->s_err_report);
4012         if (sbi->s_flex_groups)
4013                 ext4_kvfree(sbi->s_flex_groups);
4014         percpu_counter_destroy(&sbi->s_freeclusters_counter);
4015         percpu_counter_destroy(&sbi->s_freeinodes_counter);
4016         percpu_counter_destroy(&sbi->s_dirs_counter);
4017         percpu_counter_destroy(&sbi->s_dirtyclusters_counter);
4018         percpu_counter_destroy(&sbi->s_extent_cache_cnt);
4019         if (sbi->s_mmp_tsk)
4020                 kthread_stop(sbi->s_mmp_tsk);
4021 failed_mount2:
4022         for (i = 0; i < db_count; i++)
4023                 brelse(sbi->s_group_desc[i]);
4024         ext4_kvfree(sbi->s_group_desc);
4025 failed_mount:
4026         if (sbi->s_chksum_driver)
4027                 crypto_free_shash(sbi->s_chksum_driver);
4028         if (sbi->s_proc) {
4029                 remove_proc_entry("options", sbi->s_proc);
4030                 remove_proc_entry(sb->s_id, ext4_proc_root);
4031         }
4032 #ifdef CONFIG_QUOTA
4033         for (i = 0; i < MAXQUOTAS; i++)
4034                 kfree(sbi->s_qf_names[i]);
4035 #endif
4036         ext4_blkdev_remove(sbi);
4037         brelse(bh);
4038 out_fail:
4039         sb->s_fs_info = NULL;
4040         kfree(sbi->s_blockgroup_lock);
4041         kfree(sbi);
4042 out_free_orig:
4043         kfree(orig_data);
4044         return err ? err : ret;
4045 }
4046
4047 /*
4048  * Setup any per-fs journal parameters now.  We'll do this both on
4049  * initial mount, once the journal has been initialised but before we've
4050  * done any recovery; and again on any subsequent remount.
4051  */
4052 static void ext4_init_journal_params(struct super_block *sb, journal_t *journal)
4053 {
4054         struct ext4_sb_info *sbi = EXT4_SB(sb);
4055
4056         journal->j_commit_interval = sbi->s_commit_interval;
4057         journal->j_min_batch_time = sbi->s_min_batch_time;
4058         journal->j_max_batch_time = sbi->s_max_batch_time;
4059
4060         write_lock(&journal->j_state_lock);
4061         if (test_opt(sb, BARRIER))
4062                 journal->j_flags |= JBD2_BARRIER;
4063         else
4064                 journal->j_flags &= ~JBD2_BARRIER;
4065         if (test_opt(sb, DATA_ERR_ABORT))
4066                 journal->j_flags |= JBD2_ABORT_ON_SYNCDATA_ERR;
4067         else
4068                 journal->j_flags &= ~JBD2_ABORT_ON_SYNCDATA_ERR;
4069         write_unlock(&journal->j_state_lock);
4070 }
4071
4072 static journal_t *ext4_get_journal(struct super_block *sb,
4073                                    unsigned int journal_inum)
4074 {
4075         struct inode *journal_inode;
4076         journal_t *journal;
4077
4078         BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL));
4079
4080         /* First, test for the existence of a valid inode on disk.  Bad
4081          * things happen if we iget() an unused inode, as the subsequent
4082          * iput() will try to delete it. */
4083
4084         journal_inode = ext4_iget(sb, journal_inum);
4085         if (IS_ERR(journal_inode)) {
4086                 ext4_msg(sb, KERN_ERR, "no journal found");
4087                 return NULL;
4088         }
4089         if (!journal_inode->i_nlink) {
4090                 make_bad_inode(journal_inode);
4091                 iput(journal_inode);
4092                 ext4_msg(sb, KERN_ERR, "journal inode is deleted");
4093                 return NULL;
4094         }
4095
4096         jbd_debug(2, "Journal inode found at %p: %lld bytes\n",
4097                   journal_inode, journal_inode->i_size);
4098         if (!S_ISREG(journal_inode->i_mode)) {
4099                 ext4_msg(sb, KERN_ERR, "invalid journal inode");
4100                 iput(journal_inode);
4101                 return NULL;
4102         }
4103
4104         journal = jbd2_journal_init_inode(journal_inode);
4105         if (!journal) {
4106                 ext4_msg(sb, KERN_ERR, "Could not load journal inode");
4107                 iput(journal_inode);
4108                 return NULL;
4109         }
4110         journal->j_private = sb;
4111         ext4_init_journal_params(sb, journal);
4112         return journal;
4113 }
4114
4115 static journal_t *ext4_get_dev_journal(struct super_block *sb,
4116                                        dev_t j_dev)
4117 {
4118         struct buffer_head *bh;
4119         journal_t *journal;
4120         ext4_fsblk_t start;
4121         ext4_fsblk_t len;
4122         int hblock, blocksize;
4123         ext4_fsblk_t sb_block;
4124         unsigned long offset;
4125         struct ext4_super_block *es;
4126         struct block_device *bdev;
4127
4128         BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL));
4129
4130         bdev = ext4_blkdev_get(j_dev, sb);
4131         if (bdev == NULL)
4132                 return NULL;
4133
4134         blocksize = sb->s_blocksize;
4135         hblock = bdev_logical_block_size(bdev);
4136         if (blocksize < hblock) {
4137                 ext4_msg(sb, KERN_ERR,
4138                         "blocksize too small for journal device");
4139                 goto out_bdev;
4140         }
4141
4142         sb_block = EXT4_MIN_BLOCK_SIZE / blocksize;
4143         offset = EXT4_MIN_BLOCK_SIZE % blocksize;
4144         set_blocksize(bdev, blocksize);
4145         if (!(bh = __bread(bdev, sb_block, blocksize))) {
4146                 ext4_msg(sb, KERN_ERR, "couldn't read superblock of "
4147                        "external journal");
4148                 goto out_bdev;
4149         }
4150
4151         es = (struct ext4_super_block *) (bh->b_data + offset);
4152         if ((le16_to_cpu(es->s_magic) != EXT4_SUPER_MAGIC) ||
4153             !(le32_to_cpu(es->s_feature_incompat) &
4154               EXT4_FEATURE_INCOMPAT_JOURNAL_DEV)) {
4155                 ext4_msg(sb, KERN_ERR, "external journal has "
4156                                         "bad superblock");
4157                 brelse(bh);
4158                 goto out_bdev;
4159         }
4160
4161         if (memcmp(EXT4_SB(sb)->s_es->s_journal_uuid, es->s_uuid, 16)) {
4162                 ext4_msg(sb, KERN_ERR, "journal UUID does not match");
4163                 brelse(bh);
4164                 goto out_bdev;
4165         }
4166
4167         len = ext4_blocks_count(es);
4168         start = sb_block + 1;
4169         brelse(bh);     /* we're done with the superblock */
4170
4171         journal = jbd2_journal_init_dev(bdev, sb->s_bdev,
4172                                         start, len, blocksize);
4173         if (!journal) {
4174                 ext4_msg(sb, KERN_ERR, "failed to create device journal");
4175                 goto out_bdev;
4176         }
4177         journal->j_private = sb;
4178         ll_rw_block(READ, 1, &journal->j_sb_buffer);
4179         wait_on_buffer(journal->j_sb_buffer);
4180         if (!buffer_uptodate(journal->j_sb_buffer)) {
4181                 ext4_msg(sb, KERN_ERR, "I/O error on journal device");
4182                 goto out_journal;
4183         }
4184         if (be32_to_cpu(journal->j_superblock->s_nr_users) != 1) {
4185                 ext4_msg(sb, KERN_ERR, "External journal has more than one "
4186                                         "user (unsupported) - %d",
4187                         be32_to_cpu(journal->j_superblock->s_nr_users));
4188                 goto out_journal;
4189         }
4190         EXT4_SB(sb)->journal_bdev = bdev;
4191         ext4_init_journal_params(sb, journal);
4192         return journal;
4193
4194 out_journal:
4195         jbd2_journal_destroy(journal);
4196 out_bdev:
4197         ext4_blkdev_put(bdev);
4198         return NULL;
4199 }
4200
4201 static int ext4_load_journal(struct super_block *sb,
4202                              struct ext4_super_block *es,
4203                              unsigned long journal_devnum)
4204 {
4205         journal_t *journal;
4206         unsigned int journal_inum = le32_to_cpu(es->s_journal_inum);
4207         dev_t journal_dev;
4208         int err = 0;
4209         int really_read_only;
4210
4211         BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL));
4212
4213         if (journal_devnum &&
4214             journal_devnum != le32_to_cpu(es->s_journal_dev)) {
4215                 ext4_msg(sb, KERN_INFO, "external journal device major/minor "
4216                         "numbers have changed");
4217                 journal_dev = new_decode_dev(journal_devnum);
4218         } else
4219                 journal_dev = new_decode_dev(le32_to_cpu(es->s_journal_dev));
4220
4221         really_read_only = bdev_read_only(sb->s_bdev);
4222
4223         /*
4224          * Are we loading a blank journal or performing recovery after a
4225          * crash?  For recovery, we need to check in advance whether we
4226          * can get read-write access to the device.
4227          */
4228         if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER)) {
4229                 if (sb->s_flags & MS_RDONLY) {
4230                         ext4_msg(sb, KERN_INFO, "INFO: recovery "
4231                                         "required on readonly filesystem");
4232                         if (really_read_only) {
4233                                 ext4_msg(sb, KERN_ERR, "write access "
4234                                         "unavailable, cannot proceed");
4235                                 return -EROFS;
4236                         }
4237                         ext4_msg(sb, KERN_INFO, "write access will "
4238                                "be enabled during recovery");
4239                 }
4240         }
4241
4242         if (journal_inum && journal_dev) {
4243                 ext4_msg(sb, KERN_ERR, "filesystem has both journal "
4244                        "and inode journals!");
4245                 return -EINVAL;
4246         }
4247
4248         if (journal_inum) {
4249                 if (!(journal = ext4_get_journal(sb, journal_inum)))
4250                         return -EINVAL;
4251         } else {
4252                 if (!(journal = ext4_get_dev_journal(sb, journal_dev)))
4253                         return -EINVAL;
4254         }
4255
4256         if (!(journal->j_flags & JBD2_BARRIER))
4257                 ext4_msg(sb, KERN_INFO, "barriers disabled");
4258
4259         if (!EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER))
4260                 err = jbd2_journal_wipe(journal, !really_read_only);
4261         if (!err) {
4262                 char *save = kmalloc(EXT4_S_ERR_LEN, GFP_KERNEL);
4263                 if (save)
4264                         memcpy(save, ((char *) es) +
4265                                EXT4_S_ERR_START, EXT4_S_ERR_LEN);
4266                 err = jbd2_journal_load(journal);
4267                 if (save)
4268                         memcpy(((char *) es) + EXT4_S_ERR_START,
4269                                save, EXT4_S_ERR_LEN);
4270                 kfree(save);
4271         }
4272
4273         if (err) {
4274                 ext4_msg(sb, KERN_ERR, "error loading journal");
4275                 jbd2_journal_destroy(journal);
4276                 return err;
4277         }
4278
4279         EXT4_SB(sb)->s_journal = journal;
4280         ext4_clear_journal_err(sb, es);
4281
4282         if (!really_read_only && journal_devnum &&
4283             journal_devnum != le32_to_cpu(es->s_journal_dev)) {
4284                 es->s_journal_dev = cpu_to_le32(journal_devnum);
4285
4286                 /* Make sure we flush the recovery flag to disk. */
4287                 ext4_commit_super(sb, 1);
4288         }
4289
4290         return 0;
4291 }
4292
4293 static int ext4_commit_super(struct super_block *sb, int sync)
4294 {
4295         struct ext4_super_block *es = EXT4_SB(sb)->s_es;
4296         struct buffer_head *sbh = EXT4_SB(sb)->s_sbh;
4297         int error = 0;
4298
4299         if (!sbh || block_device_ejected(sb))
4300                 return error;
4301         if (buffer_write_io_error(sbh)) {
4302                 /*
4303                  * Oh, dear.  A previous attempt to write the
4304                  * superblock failed.  This could happen because the
4305                  * USB device was yanked out.  Or it could happen to
4306                  * be a transient write error and maybe the block will
4307                  * be remapped.  Nothing we can do but to retry the
4308                  * write and hope for the best.
4309                  */
4310                 ext4_msg(sb, KERN_ERR, "previous I/O error to "
4311                        "superblock detected");
4312                 clear_buffer_write_io_error(sbh);
4313                 set_buffer_uptodate(sbh);
4314         }
4315         /*
4316          * If the file system is mounted read-only, don't update the
4317          * superblock write time.  This avoids updating the superblock
4318          * write time when we are mounting the root file system
4319          * read/only but we need to replay the journal; at that point,
4320          * for people who are east of GMT and who make their clock
4321          * tick in localtime for Windows bug-for-bug compatibility,
4322          * the clock is set in the future, and this will cause e2fsck
4323          * to complain and force a full file system check.
4324          */
4325         if (!(sb->s_flags & MS_RDONLY))
4326                 es->s_wtime = cpu_to_le32(get_seconds());
4327         if (sb->s_bdev->bd_part)
4328                 es->s_kbytes_written =
4329                         cpu_to_le64(EXT4_SB(sb)->s_kbytes_written +
4330                             ((part_stat_read(sb->s_bdev->bd_part, sectors[1]) -
4331                               EXT4_SB(sb)->s_sectors_written_start) >> 1));
4332         else
4333                 es->s_kbytes_written =
4334                         cpu_to_le64(EXT4_SB(sb)->s_kbytes_written);
4335         ext4_free_blocks_count_set(es,
4336                         EXT4_C2B(EXT4_SB(sb), percpu_counter_sum_positive(
4337                                 &EXT4_SB(sb)->s_freeclusters_counter)));
4338         es->s_free_inodes_count =
4339                 cpu_to_le32(percpu_counter_sum_positive(
4340                                 &EXT4_SB(sb)->s_freeinodes_counter));
4341         BUFFER_TRACE(sbh, "marking dirty");
4342         ext4_superblock_csum_set(sb);
4343         mark_buffer_dirty(sbh);
4344         if (sync) {
4345                 error = sync_dirty_buffer(sbh);
4346                 if (error)
4347                         return error;
4348
4349                 error = buffer_write_io_error(sbh);
4350                 if (error) {
4351                         ext4_msg(sb, KERN_ERR, "I/O error while writing "
4352                                "superblock");
4353                         clear_buffer_write_io_error(sbh);
4354                         set_buffer_uptodate(sbh);
4355                 }
4356         }
4357         return error;
4358 }
4359
4360 /*
4361  * Have we just finished recovery?  If so, and if we are mounting (or
4362  * remounting) the filesystem readonly, then we will end up with a
4363  * consistent fs on disk.  Record that fact.
4364  */
4365 static void ext4_mark_recovery_complete(struct super_block *sb,
4366                                         struct ext4_super_block *es)
4367 {
4368         journal_t *journal = EXT4_SB(sb)->s_journal;
4369
4370         if (!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL)) {
4371                 BUG_ON(journal != NULL);
4372                 return;
4373         }
4374         jbd2_journal_lock_updates(journal);
4375         if (jbd2_journal_flush(journal) < 0)
4376                 goto out;
4377
4378         if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER) &&
4379             sb->s_flags & MS_RDONLY) {
4380                 EXT4_CLEAR_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
4381                 ext4_commit_super(sb, 1);
4382         }
4383
4384 out:
4385         jbd2_journal_unlock_updates(journal);
4386 }
4387
4388 /*
4389  * If we are mounting (or read-write remounting) a filesystem whose journal
4390  * has recorded an error from a previous lifetime, move that error to the
4391  * main filesystem now.
4392  */
4393 static void ext4_clear_journal_err(struct super_block *sb,
4394                                    struct ext4_super_block *es)
4395 {
4396         journal_t *journal;
4397         int j_errno;
4398         const char *errstr;
4399
4400         BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL));
4401
4402         journal = EXT4_SB(sb)->s_journal;
4403
4404         /*
4405          * Now check for any error status which may have been recorded in the
4406          * journal by a prior ext4_error() or ext4_abort()
4407          */
4408
4409         j_errno = jbd2_journal_errno(journal);
4410         if (j_errno) {
4411                 char nbuf[16];
4412
4413                 errstr = ext4_decode_error(sb, j_errno, nbuf);
4414                 ext4_warning(sb, "Filesystem error recorded "
4415                              "from previous mount: %s", errstr);
4416                 ext4_warning(sb, "Marking fs in need of filesystem check.");
4417
4418                 EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS;
4419                 es->s_state |= cpu_to_le16(EXT4_ERROR_FS);
4420                 ext4_commit_super(sb, 1);
4421
4422                 jbd2_journal_clear_err(journal);
4423                 jbd2_journal_update_sb_errno(journal);
4424         }
4425 }
4426
4427 /*
4428  * Force the running and committing transactions to commit,
4429  * and wait on the commit.
4430  */
4431 int ext4_force_commit(struct super_block *sb)
4432 {
4433         journal_t *journal;
4434
4435         if (sb->s_flags & MS_RDONLY)
4436                 return 0;
4437
4438         journal = EXT4_SB(sb)->s_journal;
4439         return ext4_journal_force_commit(journal);
4440 }
4441
4442 static int ext4_sync_fs(struct super_block *sb, int wait)
4443 {
4444         int ret = 0;
4445         tid_t target;
4446         struct ext4_sb_info *sbi = EXT4_SB(sb);
4447
4448         trace_ext4_sync_fs(sb, wait);
4449         flush_workqueue(sbi->dio_unwritten_wq);
4450         /*
4451          * Writeback quota in non-journalled quota case - journalled quota has
4452          * no dirty dquots
4453          */
4454         dquot_writeback_dquots(sb, -1);
4455         if (jbd2_journal_start_commit(sbi->s_journal, &target)) {
4456                 if (wait)
4457                         jbd2_log_wait_commit(sbi->s_journal, target);
4458         }
4459         return ret;
4460 }
4461
4462 /*
4463  * LVM calls this function before a (read-only) snapshot is created.  This
4464  * gives us a chance to flush the journal completely and mark the fs clean.
4465  *
4466  * Note that only this function cannot bring a filesystem to be in a clean
4467  * state independently. It relies on upper layer to stop all data & metadata
4468  * modifications.
4469  */
4470 static int ext4_freeze(struct super_block *sb)
4471 {
4472         int error = 0;
4473         journal_t *journal;
4474
4475         if (sb->s_flags & MS_RDONLY)
4476                 return 0;
4477
4478         journal = EXT4_SB(sb)->s_journal;
4479
4480         /* Now we set up the journal barrier. */
4481         jbd2_journal_lock_updates(journal);
4482
4483         /*
4484          * Don't clear the needs_recovery flag if we failed to flush
4485          * the journal.
4486          */
4487         error = jbd2_journal_flush(journal);
4488         if (error < 0)
4489                 goto out;
4490
4491         /* Journal blocked and flushed, clear needs_recovery flag. */
4492         EXT4_CLEAR_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
4493         error = ext4_commit_super(sb, 1);
4494 out:
4495         /* we rely on upper layer to stop further updates */
4496         jbd2_journal_unlock_updates(EXT4_SB(sb)->s_journal);
4497         return error;
4498 }
4499
4500 /*
4501  * Called by LVM after the snapshot is done.  We need to reset the RECOVER
4502  * flag here, even though the filesystem is not technically dirty yet.
4503  */
4504 static int ext4_unfreeze(struct super_block *sb)
4505 {
4506         if (sb->s_flags & MS_RDONLY)
4507                 return 0;
4508
4509         /* Reset the needs_recovery flag before the fs is unlocked. */
4510         EXT4_SET_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
4511         ext4_commit_super(sb, 1);
4512         return 0;
4513 }
4514
4515 /*
4516  * Structure to save mount options for ext4_remount's benefit
4517  */
4518 struct ext4_mount_options {
4519         unsigned long s_mount_opt;
4520         unsigned long s_mount_opt2;
4521         kuid_t s_resuid;
4522         kgid_t s_resgid;
4523         unsigned long s_commit_interval;
4524         u32 s_min_batch_time, s_max_batch_time;
4525 #ifdef CONFIG_QUOTA
4526         int s_jquota_fmt;
4527         char *s_qf_names[MAXQUOTAS];
4528 #endif
4529 };
4530
4531 static int ext4_remount(struct super_block *sb, int *flags, char *data)
4532 {
4533         struct ext4_super_block *es;
4534         struct ext4_sb_info *sbi = EXT4_SB(sb);
4535         unsigned long old_sb_flags;
4536         struct ext4_mount_options old_opts;
4537         int enable_quota = 0;
4538         ext4_group_t g;
4539         unsigned int journal_ioprio = DEFAULT_JOURNAL_IOPRIO;
4540         int err = 0;
4541 #ifdef CONFIG_QUOTA
4542         int i, j;
4543 #endif
4544         char *orig_data = kstrdup(data, GFP_KERNEL);
4545
4546         /* Store the original options */
4547         old_sb_flags = sb->s_flags;
4548         old_opts.s_mount_opt = sbi->s_mount_opt;
4549         old_opts.s_mount_opt2 = sbi->s_mount_opt2;
4550         old_opts.s_resuid = sbi->s_resuid;
4551         old_opts.s_resgid = sbi->s_resgid;
4552         old_opts.s_commit_interval = sbi->s_commit_interval;
4553         old_opts.s_min_batch_time = sbi->s_min_batch_time;
4554         old_opts.s_max_batch_time = sbi->s_max_batch_time;
4555 #ifdef CONFIG_QUOTA
4556         old_opts.s_jquota_fmt = sbi->s_jquota_fmt;
4557         for (i = 0; i < MAXQUOTAS; i++)
4558                 if (sbi->s_qf_names[i]) {
4559                         old_opts.s_qf_names[i] = kstrdup(sbi->s_qf_names[i],
4560                                                          GFP_KERNEL);
4561                         if (!old_opts.s_qf_names[i]) {
4562                                 for (j = 0; j < i; j++)
4563                                         kfree(old_opts.s_qf_names[j]);
4564                                 kfree(orig_data);
4565                                 return -ENOMEM;
4566                         }
4567                 } else
4568                         old_opts.s_qf_names[i] = NULL;
4569 #endif
4570         if (sbi->s_journal && sbi->s_journal->j_task->io_context)
4571                 journal_ioprio = sbi->s_journal->j_task->io_context->ioprio;
4572
4573         /*
4574          * Allow the "check" option to be passed as a remount option.
4575          */
4576         if (!parse_options(data, sb, NULL, &journal_ioprio, 1)) {
4577                 err = -EINVAL;
4578                 goto restore_opts;
4579         }
4580
4581         if (sbi->s_mount_flags & EXT4_MF_FS_ABORTED)
4582                 ext4_abort(sb, "Abort forced by user");
4583
4584         sb->s_flags = (sb->s_flags & ~MS_POSIXACL) |
4585                 (test_opt(sb, POSIX_ACL) ? MS_POSIXACL : 0);
4586
4587         es = sbi->s_es;
4588
4589         if (sbi->s_journal) {
4590                 ext4_init_journal_params(sb, sbi->s_journal);
4591                 set_task_ioprio(sbi->s_journal->j_task, journal_ioprio);
4592         }
4593
4594         if ((*flags & MS_RDONLY) != (sb->s_flags & MS_RDONLY)) {
4595                 if (sbi->s_mount_flags & EXT4_MF_FS_ABORTED) {
4596                         err = -EROFS;
4597                         goto restore_opts;
4598                 }
4599
4600                 if (*flags & MS_RDONLY) {
4601                         err = dquot_suspend(sb, -1);
4602                         if (err < 0)
4603                                 goto restore_opts;
4604
4605                         /*
4606                          * First of all, the unconditional stuff we have to do
4607                          * to disable replay of the journal when we next remount
4608                          */
4609                         sb->s_flags |= MS_RDONLY;
4610
4611                         /*
4612                          * OK, test if we are remounting a valid rw partition
4613                          * readonly, and if so set the rdonly flag and then
4614                          * mark the partition as valid again.
4615                          */
4616                         if (!(es->s_state & cpu_to_le16(EXT4_VALID_FS)) &&
4617                             (sbi->s_mount_state & EXT4_VALID_FS))
4618                                 es->s_state = cpu_to_le16(sbi->s_mount_state);
4619
4620                         if (sbi->s_journal)
4621                                 ext4_mark_recovery_complete(sb, es);
4622                 } else {
4623                         /* Make sure we can mount this feature set readwrite */
4624                         if (!ext4_feature_set_ok(sb, 0)) {
4625                                 err = -EROFS;
4626                                 goto restore_opts;
4627                         }
4628                         /*
4629                          * Make sure the group descriptor checksums
4630                          * are sane.  If they aren't, refuse to remount r/w.
4631                          */
4632                         for (g = 0; g < sbi->s_groups_count; g++) {
4633                                 struct ext4_group_desc *gdp =
4634                                         ext4_get_group_desc(sb, g, NULL);
4635
4636                                 if (!ext4_group_desc_csum_verify(sb, g, gdp)) {
4637                                         ext4_msg(sb, KERN_ERR,
4638                "ext4_remount: Checksum for group %u failed (%u!=%u)",
4639                 g, le16_to_cpu(ext4_group_desc_csum(sbi, g, gdp)),
4640                                                le16_to_cpu(gdp->bg_checksum));
4641                                         err = -EINVAL;
4642                                         goto restore_opts;
4643                                 }
4644                         }
4645
4646                         /*
4647                          * If we have an unprocessed orphan list hanging
4648                          * around from a previously readonly bdev mount,
4649                          * require a full umount/remount for now.
4650                          */
4651                         if (es->s_last_orphan) {
4652                                 ext4_msg(sb, KERN_WARNING, "Couldn't "
4653                                        "remount RDWR because of unprocessed "
4654                                        "orphan inode list.  Please "
4655                                        "umount/remount instead");
4656                                 err = -EINVAL;
4657                                 goto restore_opts;
4658                         }
4659
4660                         /*
4661                          * Mounting a RDONLY partition read-write, so reread
4662                          * and store the current valid flag.  (It may have
4663                          * been changed by e2fsck since we originally mounted
4664                          * the partition.)
4665                          */
4666                         if (sbi->s_journal)
4667                                 ext4_clear_journal_err(sb, es);
4668                         sbi->s_mount_state = le16_to_cpu(es->s_state);
4669                         if (!ext4_setup_super(sb, es, 0))
4670                                 sb->s_flags &= ~MS_RDONLY;
4671                         if (EXT4_HAS_INCOMPAT_FEATURE(sb,
4672                                                      EXT4_FEATURE_INCOMPAT_MMP))
4673                                 if (ext4_multi_mount_protect(sb,
4674                                                 le64_to_cpu(es->s_mmp_block))) {
4675                                         err = -EROFS;
4676                                         goto restore_opts;
4677                                 }
4678                         enable_quota = 1;
4679                 }
4680         }
4681
4682         /*
4683          * Reinitialize lazy itable initialization thread based on
4684          * current settings
4685          */
4686         if ((sb->s_flags & MS_RDONLY) || !test_opt(sb, INIT_INODE_TABLE))
4687                 ext4_unregister_li_request(sb);
4688         else {
4689                 ext4_group_t first_not_zeroed;
4690                 first_not_zeroed = ext4_has_uninit_itable(sb);
4691                 ext4_register_li_request(sb, first_not_zeroed);
4692         }
4693
4694         ext4_setup_system_zone(sb);
4695         if (sbi->s_journal == NULL && !(old_sb_flags & MS_RDONLY))
4696                 ext4_commit_super(sb, 1);
4697
4698 #ifdef CONFIG_QUOTA
4699         /* Release old quota file names */
4700         for (i = 0; i < MAXQUOTAS; i++)
4701                 kfree(old_opts.s_qf_names[i]);
4702         if (enable_quota) {
4703                 if (sb_any_quota_suspended(sb))
4704                         dquot_resume(sb, -1);
4705                 else if (EXT4_HAS_RO_COMPAT_FEATURE(sb,
4706                                         EXT4_FEATURE_RO_COMPAT_QUOTA)) {
4707                         err = ext4_enable_quotas(sb);
4708                         if (err)
4709                                 goto restore_opts;
4710                 }
4711         }
4712 #endif
4713
4714         ext4_msg(sb, KERN_INFO, "re-mounted. Opts: %s", orig_data);
4715         kfree(orig_data);
4716         return 0;
4717
4718 restore_opts:
4719         sb->s_flags = old_sb_flags;
4720         sbi->s_mount_opt = old_opts.s_mount_opt;
4721         sbi->s_mount_opt2 = old_opts.s_mount_opt2;
4722         sbi->s_resuid = old_opts.s_resuid;
4723         sbi->s_resgid = old_opts.s_resgid;
4724         sbi->s_commit_interval = old_opts.s_commit_interval;
4725         sbi->s_min_batch_time = old_opts.s_min_batch_time;
4726         sbi->s_max_batch_time = old_opts.s_max_batch_time;
4727 #ifdef CONFIG_QUOTA
4728         sbi->s_jquota_fmt = old_opts.s_jquota_fmt;
4729         for (i = 0; i < MAXQUOTAS; i++) {
4730                 kfree(sbi->s_qf_names[i]);
4731                 sbi->s_qf_names[i] = old_opts.s_qf_names[i];
4732         }
4733 #endif
4734         kfree(orig_data);
4735         return err;
4736 }
4737
4738 static int ext4_statfs(struct dentry *dentry, struct kstatfs *buf)
4739 {
4740         struct super_block *sb = dentry->d_sb;
4741         struct ext4_sb_info *sbi = EXT4_SB(sb);
4742         struct ext4_super_block *es = sbi->s_es;
4743         ext4_fsblk_t overhead = 0;
4744         u64 fsid;
4745         s64 bfree;
4746
4747         if (!test_opt(sb, MINIX_DF))
4748                 overhead = sbi->s_overhead;
4749
4750         buf->f_type = EXT4_SUPER_MAGIC;
4751         buf->f_bsize = sb->s_blocksize;
4752         buf->f_blocks = ext4_blocks_count(es) - EXT4_C2B(sbi, overhead);
4753         bfree = percpu_counter_sum_positive(&sbi->s_freeclusters_counter) -
4754                 percpu_counter_sum_positive(&sbi->s_dirtyclusters_counter);
4755         /* prevent underflow in case that few free space is available */
4756         buf->f_bfree = EXT4_C2B(sbi, max_t(s64, bfree, 0));
4757         buf->f_bavail = buf->f_bfree - ext4_r_blocks_count(es);
4758         if (buf->f_bfree < ext4_r_blocks_count(es))
4759                 buf->f_bavail = 0;
4760         buf->f_files = le32_to_cpu(es->s_inodes_count);
4761         buf->f_ffree = percpu_counter_sum_positive(&sbi->s_freeinodes_counter);
4762         buf->f_namelen = EXT4_NAME_LEN;
4763         fsid = le64_to_cpup((void *)es->s_uuid) ^
4764                le64_to_cpup((void *)es->s_uuid + sizeof(u64));
4765         buf->f_fsid.val[0] = fsid & 0xFFFFFFFFUL;
4766         buf->f_fsid.val[1] = (fsid >> 32) & 0xFFFFFFFFUL;
4767
4768         return 0;
4769 }
4770
4771 /* Helper function for writing quotas on sync - we need to start transaction
4772  * before quota file is locked for write. Otherwise the are possible deadlocks:
4773  * Process 1                         Process 2
4774  * ext4_create()                     quota_sync()
4775  *   jbd2_journal_start()                  write_dquot()
4776  *   dquot_initialize()                         down(dqio_mutex)
4777  *     down(dqio_mutex)                    jbd2_journal_start()
4778  *
4779  */
4780
4781 #ifdef CONFIG_QUOTA
4782
4783 static inline struct inode *dquot_to_inode(struct dquot *dquot)
4784 {
4785         return sb_dqopt(dquot->dq_sb)->files[dquot->dq_id.type];
4786 }
4787
4788 static int ext4_write_dquot(struct dquot *dquot)
4789 {
4790         int ret, err;
4791         handle_t *handle;
4792         struct inode *inode;
4793
4794         inode = dquot_to_inode(dquot);
4795         handle = ext4_journal_start(inode, EXT4_HT_QUOTA,
4796                                     EXT4_QUOTA_TRANS_BLOCKS(dquot->dq_sb));
4797         if (IS_ERR(handle))
4798                 return PTR_ERR(handle);
4799         ret = dquot_commit(dquot);
4800         err = ext4_journal_stop(handle);
4801         if (!ret)
4802                 ret = err;
4803         return ret;
4804 }
4805
4806 static int ext4_acquire_dquot(struct dquot *dquot)
4807 {
4808         int ret, err;
4809         handle_t *handle;
4810
4811         handle = ext4_journal_start(dquot_to_inode(dquot), EXT4_HT_QUOTA,
4812                                     EXT4_QUOTA_INIT_BLOCKS(dquot->dq_sb));
4813         if (IS_ERR(handle))
4814                 return PTR_ERR(handle);
4815         ret = dquot_acquire(dquot);
4816         err = ext4_journal_stop(handle);
4817         if (!ret)
4818                 ret = err;
4819         return ret;
4820 }
4821
4822 static int ext4_release_dquot(struct dquot *dquot)
4823 {
4824         int ret, err;
4825         handle_t *handle;
4826
4827         handle = ext4_journal_start(dquot_to_inode(dquot), EXT4_HT_QUOTA,
4828                                     EXT4_QUOTA_DEL_BLOCKS(dquot->dq_sb));
4829         if (IS_ERR(handle)) {
4830                 /* Release dquot anyway to avoid endless cycle in dqput() */
4831                 dquot_release(dquot);
4832                 return PTR_ERR(handle);
4833         }
4834         ret = dquot_release(dquot);
4835         err = ext4_journal_stop(handle);
4836         if (!ret)
4837                 ret = err;
4838         return ret;
4839 }
4840
4841 static int ext4_mark_dquot_dirty(struct dquot *dquot)
4842 {
4843         struct super_block *sb = dquot->dq_sb;
4844         struct ext4_sb_info *sbi = EXT4_SB(sb);
4845
4846         /* Are we journaling quotas? */
4847         if (EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_QUOTA) ||
4848             sbi->s_qf_names[USRQUOTA] || sbi->s_qf_names[GRPQUOTA]) {
4849                 dquot_mark_dquot_dirty(dquot);
4850                 return ext4_write_dquot(dquot);
4851         } else {
4852                 return dquot_mark_dquot_dirty(dquot);
4853         }
4854 }
4855
4856 static int ext4_write_info(struct super_block *sb, int type)
4857 {
4858         int ret, err;
4859         handle_t *handle;
4860
4861         /* Data block + inode block */
4862         handle = ext4_journal_start(sb->s_root->d_inode, EXT4_HT_QUOTA, 2);
4863         if (IS_ERR(handle))
4864                 return PTR_ERR(handle);
4865         ret = dquot_commit_info(sb, type);
4866         err = ext4_journal_stop(handle);
4867         if (!ret)
4868                 ret = err;
4869         return ret;
4870 }
4871
4872 /*
4873  * Turn on quotas during mount time - we need to find
4874  * the quota file and such...
4875  */
4876 static int ext4_quota_on_mount(struct super_block *sb, int type)
4877 {
4878         return dquot_quota_on_mount(sb, EXT4_SB(sb)->s_qf_names[type],
4879                                         EXT4_SB(sb)->s_jquota_fmt, type);
4880 }
4881
4882 /*
4883  * Standard function to be called on quota_on
4884  */
4885 static int ext4_quota_on(struct super_block *sb, int type, int format_id,
4886                          struct path *path)
4887 {
4888         int err;
4889
4890         if (!test_opt(sb, QUOTA))
4891                 return -EINVAL;
4892
4893         /* Quotafile not on the same filesystem? */
4894         if (path->dentry->d_sb != sb)
4895                 return -EXDEV;
4896         /* Journaling quota? */
4897         if (EXT4_SB(sb)->s_qf_names[type]) {
4898                 /* Quotafile not in fs root? */
4899                 if (path->dentry->d_parent != sb->s_root)
4900                         ext4_msg(sb, KERN_WARNING,
4901                                 "Quota file not on filesystem root. "
4902                                 "Journaled quota will not work");
4903         }
4904
4905         /*
4906          * When we journal data on quota file, we have to flush journal to see
4907          * all updates to the file when we bypass pagecache...
4908          */
4909         if (EXT4_SB(sb)->s_journal &&
4910             ext4_should_journal_data(path->dentry->d_inode)) {
4911                 /*
4912                  * We don't need to lock updates but journal_flush() could
4913                  * otherwise be livelocked...
4914                  */
4915                 jbd2_journal_lock_updates(EXT4_SB(sb)->s_journal);
4916                 err = jbd2_journal_flush(EXT4_SB(sb)->s_journal);
4917                 jbd2_journal_unlock_updates(EXT4_SB(sb)->s_journal);
4918                 if (err)
4919                         return err;
4920         }
4921
4922         return dquot_quota_on(sb, type, format_id, path);
4923 }
4924
4925 static int ext4_quota_enable(struct super_block *sb, int type, int format_id,
4926                              unsigned int flags)
4927 {
4928         int err;
4929         struct inode *qf_inode;
4930         unsigned long qf_inums[MAXQUOTAS] = {
4931                 le32_to_cpu(EXT4_SB(sb)->s_es->s_usr_quota_inum),
4932                 le32_to_cpu(EXT4_SB(sb)->s_es->s_grp_quota_inum)
4933         };
4934
4935         BUG_ON(!EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_QUOTA));
4936
4937         if (!qf_inums[type])
4938                 return -EPERM;
4939
4940         qf_inode = ext4_iget(sb, qf_inums[type]);
4941         if (IS_ERR(qf_inode)) {
4942                 ext4_error(sb, "Bad quota inode # %lu", qf_inums[type]);
4943                 return PTR_ERR(qf_inode);
4944         }
4945
4946         err = dquot_enable(qf_inode, type, format_id, flags);
4947         iput(qf_inode);
4948
4949         return err;
4950 }
4951
4952 /* Enable usage tracking for all quota types. */
4953 static int ext4_enable_quotas(struct super_block *sb)
4954 {
4955         int type, err = 0;
4956         unsigned long qf_inums[MAXQUOTAS] = {
4957                 le32_to_cpu(EXT4_SB(sb)->s_es->s_usr_quota_inum),
4958                 le32_to_cpu(EXT4_SB(sb)->s_es->s_grp_quota_inum)
4959         };
4960
4961         sb_dqopt(sb)->flags |= DQUOT_QUOTA_SYS_FILE;
4962         for (type = 0; type < MAXQUOTAS; type++) {
4963                 if (qf_inums[type]) {
4964                         err = ext4_quota_enable(sb, type, QFMT_VFS_V1,
4965                                                 DQUOT_USAGE_ENABLED);
4966                         if (err) {
4967                                 ext4_warning(sb,
4968                                         "Failed to enable quota tracking "
4969                                         "(type=%d, err=%d). Please run "
4970                                         "e2fsck to fix.", type, err);
4971                                 return err;
4972                         }
4973                 }
4974         }
4975         return 0;
4976 }
4977
4978 /*
4979  * quota_on function that is used when QUOTA feature is set.
4980  */
4981 static int ext4_quota_on_sysfile(struct super_block *sb, int type,
4982                                  int format_id)
4983 {
4984         if (!EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_QUOTA))
4985                 return -EINVAL;
4986
4987         /*
4988          * USAGE was enabled at mount time. Only need to enable LIMITS now.
4989          */
4990         return ext4_quota_enable(sb, type, format_id, DQUOT_LIMITS_ENABLED);
4991 }
4992
4993 static int ext4_quota_off(struct super_block *sb, int type)
4994 {
4995         struct inode *inode = sb_dqopt(sb)->files[type];
4996         handle_t *handle;
4997
4998         /* Force all delayed allocation blocks to be allocated.
4999          * Caller already holds s_umount sem */
5000         if (test_opt(sb, DELALLOC))
5001                 sync_filesystem(sb);
5002
5003         if (!inode)
5004                 goto out;
5005
5006         /* Update modification times of quota files when userspace can
5007          * start looking at them */
5008         handle = ext4_journal_start(inode, EXT4_HT_QUOTA, 1);
5009         if (IS_ERR(handle))
5010                 goto out;
5011         inode->i_mtime = inode->i_ctime = CURRENT_TIME;
5012         ext4_mark_inode_dirty(handle, inode);
5013         ext4_journal_stop(handle);
5014
5015 out:
5016         return dquot_quota_off(sb, type);
5017 }
5018
5019 /*
5020  * quota_off function that is used when QUOTA feature is set.
5021  */
5022 static int ext4_quota_off_sysfile(struct super_block *sb, int type)
5023 {
5024         if (!EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_QUOTA))
5025                 return -EINVAL;
5026
5027         /* Disable only the limits. */
5028         return dquot_disable(sb, type, DQUOT_LIMITS_ENABLED);
5029 }
5030
5031 /* Read data from quotafile - avoid pagecache and such because we cannot afford
5032  * acquiring the locks... As quota files are never truncated and quota code
5033  * itself serializes the operations (and no one else should touch the files)
5034  * we don't have to be afraid of races */
5035 static ssize_t ext4_quota_read(struct super_block *sb, int type, char *data,
5036                                size_t len, loff_t off)
5037 {
5038         struct inode *inode = sb_dqopt(sb)->files[type];
5039         ext4_lblk_t blk = off >> EXT4_BLOCK_SIZE_BITS(sb);
5040         int err = 0;
5041         int offset = off & (sb->s_blocksize - 1);
5042         int tocopy;
5043         size_t toread;
5044         struct buffer_head *bh;
5045         loff_t i_size = i_size_read(inode);
5046
5047         if (off > i_size)
5048                 return 0;
5049         if (off+len > i_size)
5050                 len = i_size-off;
5051         toread = len;
5052         while (toread > 0) {
5053                 tocopy = sb->s_blocksize - offset < toread ?
5054                                 sb->s_blocksize - offset : toread;
5055                 bh = ext4_bread(NULL, inode, blk, 0, &err);
5056                 if (err)
5057                         return err;
5058                 if (!bh)        /* A hole? */
5059                         memset(data, 0, tocopy);
5060                 else
5061                         memcpy(data, bh->b_data+offset, tocopy);
5062                 brelse(bh);
5063                 offset = 0;
5064                 toread -= tocopy;
5065                 data += tocopy;
5066                 blk++;
5067         }
5068         return len;
5069 }
5070
5071 /* Write to quotafile (we know the transaction is already started and has
5072  * enough credits) */
5073 static ssize_t ext4_quota_write(struct super_block *sb, int type,
5074                                 const char *data, size_t len, loff_t off)
5075 {
5076         struct inode *inode = sb_dqopt(sb)->files[type];
5077         ext4_lblk_t blk = off >> EXT4_BLOCK_SIZE_BITS(sb);
5078         int err = 0;
5079         int offset = off & (sb->s_blocksize - 1);
5080         struct buffer_head *bh;
5081         handle_t *handle = journal_current_handle();
5082
5083         if (EXT4_SB(sb)->s_journal && !handle) {
5084                 ext4_msg(sb, KERN_WARNING, "Quota write (off=%llu, len=%llu)"
5085                         " cancelled because transaction is not started",
5086                         (unsigned long long)off, (unsigned long long)len);
5087                 return -EIO;
5088         }
5089         /*
5090          * Since we account only one data block in transaction credits,
5091          * then it is impossible to cross a block boundary.
5092          */
5093         if (sb->s_blocksize - offset < len) {
5094                 ext4_msg(sb, KERN_WARNING, "Quota write (off=%llu, len=%llu)"
5095                         " cancelled because not block aligned",
5096                         (unsigned long long)off, (unsigned long long)len);
5097                 return -EIO;
5098         }
5099
5100         bh = ext4_bread(handle, inode, blk, 1, &err);
5101         if (!bh)
5102                 goto out;
5103         err = ext4_journal_get_write_access(handle, bh);
5104         if (err) {
5105                 brelse(bh);
5106                 goto out;
5107         }
5108         lock_buffer(bh);
5109         memcpy(bh->b_data+offset, data, len);
5110         flush_dcache_page(bh->b_page);
5111         unlock_buffer(bh);
5112         err = ext4_handle_dirty_metadata(handle, NULL, bh);
5113         brelse(bh);
5114 out:
5115         if (err)
5116                 return err;
5117         if (inode->i_size < off + len) {
5118                 i_size_write(inode, off + len);
5119                 EXT4_I(inode)->i_disksize = inode->i_size;
5120                 ext4_mark_inode_dirty(handle, inode);
5121         }
5122         return len;
5123 }
5124
5125 #endif
5126
5127 static struct dentry *ext4_mount(struct file_system_type *fs_type, int flags,
5128                        const char *dev_name, void *data)
5129 {
5130         return mount_bdev(fs_type, flags, dev_name, data, ext4_fill_super);
5131 }
5132
5133 #if !defined(CONFIG_EXT2_FS) && !defined(CONFIG_EXT2_FS_MODULE) && defined(CONFIG_EXT4_USE_FOR_EXT23)
5134 static inline void register_as_ext2(void)
5135 {
5136         int err = register_filesystem(&ext2_fs_type);
5137         if (err)
5138                 printk(KERN_WARNING
5139                        "EXT4-fs: Unable to register as ext2 (%d)\n", err);
5140 }
5141
5142 static inline void unregister_as_ext2(void)
5143 {
5144         unregister_filesystem(&ext2_fs_type);
5145 }
5146
5147 static inline int ext2_feature_set_ok(struct super_block *sb)
5148 {
5149         if (EXT4_HAS_INCOMPAT_FEATURE(sb, ~EXT2_FEATURE_INCOMPAT_SUPP))
5150                 return 0;
5151         if (sb->s_flags & MS_RDONLY)
5152                 return 1;
5153         if (EXT4_HAS_RO_COMPAT_FEATURE(sb, ~EXT2_FEATURE_RO_COMPAT_SUPP))
5154                 return 0;
5155         return 1;
5156 }
5157 #else
5158 static inline void register_as_ext2(void) { }
5159 static inline void unregister_as_ext2(void) { }
5160 static inline int ext2_feature_set_ok(struct super_block *sb) { return 0; }
5161 #endif
5162
5163 #if !defined(CONFIG_EXT3_FS) && !defined(CONFIG_EXT3_FS_MODULE) && defined(CONFIG_EXT4_USE_FOR_EXT23)
5164 static inline void register_as_ext3(void)
5165 {
5166         int err = register_filesystem(&ext3_fs_type);
5167         if (err)
5168                 printk(KERN_WARNING
5169                        "EXT4-fs: Unable to register as ext3 (%d)\n", err);
5170 }
5171
5172 static inline void unregister_as_ext3(void)
5173 {
5174         unregister_filesystem(&ext3_fs_type);
5175 }
5176
5177 static inline int ext3_feature_set_ok(struct super_block *sb)
5178 {
5179         if (EXT4_HAS_INCOMPAT_FEATURE(sb, ~EXT3_FEATURE_INCOMPAT_SUPP))
5180                 return 0;
5181         if (!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL))
5182                 return 0;
5183         if (sb->s_flags & MS_RDONLY)
5184                 return 1;
5185         if (EXT4_HAS_RO_COMPAT_FEATURE(sb, ~EXT3_FEATURE_RO_COMPAT_SUPP))
5186                 return 0;
5187         return 1;
5188 }
5189 #else
5190 static inline void register_as_ext3(void) { }
5191 static inline void unregister_as_ext3(void) { }
5192 static inline int ext3_feature_set_ok(struct super_block *sb) { return 0; }
5193 #endif
5194
5195 static struct file_system_type ext4_fs_type = {
5196         .owner          = THIS_MODULE,
5197         .name           = "ext4",
5198         .mount          = ext4_mount,
5199         .kill_sb        = kill_block_super,
5200         .fs_flags       = FS_REQUIRES_DEV,
5201 };
5202 MODULE_ALIAS_FS("ext4");
5203
5204 static int __init ext4_init_feat_adverts(void)
5205 {
5206         struct ext4_features *ef;
5207         int ret = -ENOMEM;
5208
5209         ef = kzalloc(sizeof(struct ext4_features), GFP_KERNEL);
5210         if (!ef)
5211                 goto out;
5212
5213         ef->f_kobj.kset = ext4_kset;
5214         init_completion(&ef->f_kobj_unregister);
5215         ret = kobject_init_and_add(&ef->f_kobj, &ext4_feat_ktype, NULL,
5216                                    "features");
5217         if (ret) {
5218                 kfree(ef);
5219                 goto out;
5220         }
5221
5222         ext4_feat = ef;
5223         ret = 0;
5224 out:
5225         return ret;
5226 }
5227
5228 static void ext4_exit_feat_adverts(void)
5229 {
5230         kobject_put(&ext4_feat->f_kobj);
5231         wait_for_completion(&ext4_feat->f_kobj_unregister);
5232         kfree(ext4_feat);
5233 }
5234
5235 /* Shared across all ext4 file systems */
5236 wait_queue_head_t ext4__ioend_wq[EXT4_WQ_HASH_SZ];
5237 struct mutex ext4__aio_mutex[EXT4_WQ_HASH_SZ];
5238
5239 static int __init ext4_init_fs(void)
5240 {
5241         int i, err;
5242
5243         ext4_li_info = NULL;
5244         mutex_init(&ext4_li_mtx);
5245
5246         /* Build-time check for flags consistency */
5247         ext4_check_flag_values();
5248
5249         for (i = 0; i < EXT4_WQ_HASH_SZ; i++) {
5250                 mutex_init(&ext4__aio_mutex[i]);
5251                 init_waitqueue_head(&ext4__ioend_wq[i]);
5252         }
5253
5254         err = ext4_init_es();
5255         if (err)
5256                 return err;
5257
5258         err = ext4_init_pageio();
5259         if (err)
5260                 goto out7;
5261
5262         err = ext4_init_system_zone();
5263         if (err)
5264                 goto out6;
5265         ext4_kset = kset_create_and_add("ext4", NULL, fs_kobj);
5266         if (!ext4_kset) {
5267                 err = -ENOMEM;
5268                 goto out5;
5269         }
5270         ext4_proc_root = proc_mkdir("fs/ext4", NULL);
5271
5272         err = ext4_init_feat_adverts();
5273         if (err)
5274                 goto out4;
5275
5276         err = ext4_init_mballoc();
5277         if (err)
5278                 goto out3;
5279
5280         err = ext4_init_xattr();
5281         if (err)
5282                 goto out2;
5283         err = init_inodecache();
5284         if (err)
5285                 goto out1;
5286         register_as_ext3();
5287         register_as_ext2();
5288         err = register_filesystem(&ext4_fs_type);
5289         if (err)
5290                 goto out;
5291
5292         return 0;
5293 out:
5294         unregister_as_ext2();
5295         unregister_as_ext3();
5296         destroy_inodecache();
5297 out1:
5298         ext4_exit_xattr();
5299 out2:
5300         ext4_exit_mballoc();
5301 out3:
5302         ext4_exit_feat_adverts();
5303 out4:
5304         if (ext4_proc_root)
5305                 remove_proc_entry("fs/ext4", NULL);
5306         kset_unregister(ext4_kset);
5307 out5:
5308         ext4_exit_system_zone();
5309 out6:
5310         ext4_exit_pageio();
5311 out7:
5312         ext4_exit_es();
5313
5314         return err;
5315 }
5316
5317 static void __exit ext4_exit_fs(void)
5318 {
5319         ext4_destroy_lazyinit_thread();
5320         unregister_as_ext2();
5321         unregister_as_ext3();
5322         unregister_filesystem(&ext4_fs_type);
5323         destroy_inodecache();
5324         ext4_exit_xattr();
5325         ext4_exit_mballoc();
5326         ext4_exit_feat_adverts();
5327         remove_proc_entry("fs/ext4", NULL);
5328         kset_unregister(ext4_kset);
5329         ext4_exit_system_zone();
5330         ext4_exit_pageio();
5331 }
5332
5333 MODULE_AUTHOR("Remy Card, Stephen Tweedie, Andrew Morton, Andreas Dilger, Theodore Ts'o and others");
5334 MODULE_DESCRIPTION("Fourth Extended Filesystem");
5335 MODULE_LICENSE("GPL");
5336 module_init(ext4_init_fs)
5337 module_exit(ext4_exit_fs)