linux/fs/ext4/super.c
<<
>>
Prefs
   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"
  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
  57static struct proc_dir_entry *ext4_proc_root;
  58static struct kset *ext4_kset;
  59static struct ext4_lazy_init *ext4_li_info;
  60static struct mutex ext4_li_mtx;
  61static struct ext4_features *ext4_feat;
  62
  63static int ext4_load_journal(struct super_block *, struct ext4_super_block *,
  64                             unsigned long journal_devnum);
  65static int ext4_show_options(struct seq_file *seq, struct dentry *root);
  66static int ext4_commit_super(struct super_block *sb, int sync);
  67static void ext4_mark_recovery_complete(struct super_block *sb,
  68                                        struct ext4_super_block *es);
  69static void ext4_clear_journal_err(struct super_block *sb,
  70                                   struct ext4_super_block *es);
  71static int ext4_sync_fs(struct super_block *sb, int wait);
  72static const char *ext4_decode_error(struct super_block *sb, int errno,
  73                                     char nbuf[16]);
  74static int ext4_remount(struct super_block *sb, int *flags, char *data);
  75static int ext4_statfs(struct dentry *dentry, struct kstatfs *buf);
  76static int ext4_unfreeze(struct super_block *sb);
  77static int ext4_freeze(struct super_block *sb);
  78static struct dentry *ext4_mount(struct file_system_type *fs_type, int flags,
  79                       const char *dev_name, void *data);
  80static inline int ext2_feature_set_ok(struct super_block *sb);
  81static inline int ext3_feature_set_ok(struct super_block *sb);
  82static int ext4_feature_set_ok(struct super_block *sb, int readonly);
  83static void ext4_destroy_lazyinit_thread(void);
  84static void ext4_unregister_li_request(struct super_block *sb);
  85static void ext4_clear_request_list(void);
  86
  87#if !defined(CONFIG_EXT2_FS) && !defined(CONFIG_EXT2_FS_MODULE) && defined(CONFIG_EXT4_USE_FOR_EXT23)
  88static struct file_system_type ext2_fs_type = {
  89        .owner          = THIS_MODULE,
  90        .name           = "ext2",
  91        .mount          = ext4_mount,
  92        .kill_sb        = kill_block_super,
  93        .fs_flags       = FS_REQUIRES_DEV,
  94};
  95#define IS_EXT2_SB(sb) ((sb)->s_bdev->bd_holder == &ext2_fs_type)
  96#else
  97#define IS_EXT2_SB(sb) (0)
  98#endif
  99
 100
 101#if !defined(CONFIG_EXT3_FS) && !defined(CONFIG_EXT3_FS_MODULE) && defined(CONFIG_EXT4_USE_FOR_EXT23)
 102static struct file_system_type ext3_fs_type = {
 103        .owner          = THIS_MODULE,
 104        .name           = "ext3",
 105        .mount          = ext4_mount,
 106        .kill_sb        = kill_block_super,
 107        .fs_flags       = FS_REQUIRES_DEV,
 108};
 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
 114static 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
 124static __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
 136int 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
 146void 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
 157void *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
 167void *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
 177void ext4_kvfree(void *ptr)
 178{
 179        if (is_vmalloc_addr(ptr))
 180                vfree(ptr);
 181        else
 182                kfree(ptr);
 183
 184}
 185
 186ext4_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
 194ext4_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
 202ext4_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
 242void 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
 250void 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
 258void 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
 266void 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
 274void 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
 282void 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
 290void 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/* Just increment the non-pointer handle value */
 300static handle_t *ext4_get_nojournal(void)
 301{
 302        handle_t *handle = current->journal_info;
 303        unsigned long ref_cnt = (unsigned long)handle;
 304
 305        BUG_ON(ref_cnt >= EXT4_NOJOURNAL_MAX_REF_COUNT);
 306
 307        ref_cnt++;
 308        handle = (handle_t *)ref_cnt;
 309
 310        current->journal_info = handle;
 311        return handle;
 312}
 313
 314
 315/* Decrement the non-pointer handle value */
 316static void ext4_put_nojournal(handle_t *handle)
 317{
 318        unsigned long ref_cnt = (unsigned long)handle;
 319
 320        BUG_ON(ref_cnt == 0);
 321
 322        ref_cnt--;
 323        handle = (handle_t *)ref_cnt;
 324
 325        current->journal_info = handle;
 326}
 327
 328/*
 329 * Wrappers for jbd2_journal_start/end.
 330 */
 331handle_t *ext4_journal_start_sb(struct super_block *sb, int nblocks)
 332{
 333        journal_t *journal;
 334
 335        trace_ext4_journal_start(sb, nblocks, _RET_IP_);
 336        if (sb->s_flags & MS_RDONLY)
 337                return ERR_PTR(-EROFS);
 338
 339        WARN_ON(sb->s_writers.frozen == SB_FREEZE_COMPLETE);
 340        journal = EXT4_SB(sb)->s_journal;
 341        if (!journal)
 342                return ext4_get_nojournal();
 343        /*
 344         * Special case here: if the journal has aborted behind our
 345         * backs (eg. EIO in the commit thread), then we still need to
 346         * take the FS itself readonly cleanly.
 347         */
 348        if (is_journal_aborted(journal)) {
 349                ext4_abort(sb, "Detected aborted journal");
 350                return ERR_PTR(-EROFS);
 351        }
 352        return jbd2_journal_start(journal, nblocks);
 353}
 354
 355int __ext4_journal_stop(const char *where, unsigned int line, handle_t *handle)
 356{
 357        struct super_block *sb;
 358        int err;
 359        int rc;
 360
 361        if (!ext4_handle_valid(handle)) {
 362                ext4_put_nojournal(handle);
 363                return 0;
 364        }
 365        sb = handle->h_transaction->t_journal->j_private;
 366        err = handle->h_err;
 367        rc = jbd2_journal_stop(handle);
 368
 369        if (!err)
 370                err = rc;
 371        if (err)
 372                __ext4_std_error(sb, where, line, err);
 373        return err;
 374}
 375
 376void ext4_journal_abort_handle(const char *caller, unsigned int line,
 377                               const char *err_fn, struct buffer_head *bh,
 378                               handle_t *handle, int err)
 379{
 380        char nbuf[16];
 381        const char *errstr = ext4_decode_error(NULL, err, nbuf);
 382
 383        BUG_ON(!ext4_handle_valid(handle));
 384
 385        if (bh)
 386                BUFFER_TRACE(bh, "abort");
 387
 388        if (!handle->h_err)
 389                handle->h_err = err;
 390
 391        if (is_handle_aborted(handle))
 392                return;
 393
 394        printk(KERN_ERR "EXT4-fs: %s:%d: aborting transaction: %s in %s\n",
 395               caller, line, errstr, err_fn);
 396
 397        jbd2_journal_abort_handle(handle);
 398}
 399
 400static void __save_error_info(struct super_block *sb, const char *func,
 401                            unsigned int line)
 402{
 403        struct ext4_super_block *es = EXT4_SB(sb)->s_es;
 404
 405        EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS;
 406        es->s_state |= cpu_to_le16(EXT4_ERROR_FS);
 407        es->s_last_error_time = cpu_to_le32(get_seconds());
 408        strncpy(es->s_last_error_func, func, sizeof(es->s_last_error_func));
 409        es->s_last_error_line = cpu_to_le32(line);
 410        if (!es->s_first_error_time) {
 411                es->s_first_error_time = es->s_last_error_time;
 412                strncpy(es->s_first_error_func, func,
 413                        sizeof(es->s_first_error_func));
 414                es->s_first_error_line = cpu_to_le32(line);
 415                es->s_first_error_ino = es->s_last_error_ino;
 416                es->s_first_error_block = es->s_last_error_block;
 417        }
 418        /*
 419         * Start the daily error reporting function if it hasn't been
 420         * started already
 421         */
 422        if (!es->s_error_count)
 423                mod_timer(&EXT4_SB(sb)->s_err_report, jiffies + 24*60*60*HZ);
 424        le32_add_cpu(&es->s_error_count, 1);
 425}
 426
 427static void save_error_info(struct super_block *sb, const char *func,
 428                            unsigned int line)
 429{
 430        __save_error_info(sb, func, line);
 431        ext4_commit_super(sb, 1);
 432}
 433
 434/*
 435 * The del_gendisk() function uninitializes the disk-specific data
 436 * structures, including the bdi structure, without telling anyone
 437 * else.  Once this happens, any attempt to call mark_buffer_dirty()
 438 * (for example, by ext4_commit_super), will cause a kernel OOPS.
 439 * This is a kludge to prevent these oops until we can put in a proper
 440 * hook in del_gendisk() to inform the VFS and file system layers.
 441 */
 442static int block_device_ejected(struct super_block *sb)
 443{
 444        struct inode *bd_inode = sb->s_bdev->bd_inode;
 445        struct backing_dev_info *bdi = bd_inode->i_mapping->backing_dev_info;
 446
 447        return bdi->dev == NULL;
 448}
 449
 450static void ext4_journal_commit_callback(journal_t *journal, transaction_t *txn)
 451{
 452        struct super_block              *sb = journal->j_private;
 453        struct ext4_sb_info             *sbi = EXT4_SB(sb);
 454        int                             error = is_journal_aborted(journal);
 455        struct ext4_journal_cb_entry    *jce, *tmp;
 456
 457        spin_lock(&sbi->s_md_lock);
 458        list_for_each_entry_safe(jce, tmp, &txn->t_private_list, jce_list) {
 459                list_del_init(&jce->jce_list);
 460                spin_unlock(&sbi->s_md_lock);
 461                jce->jce_func(sb, jce, error);
 462                spin_lock(&sbi->s_md_lock);
 463        }
 464        spin_unlock(&sbi->s_md_lock);
 465}
 466
 467/* Deal with the reporting of failure conditions on a filesystem such as
 468 * inconsistencies detected or read IO failures.
 469 *
 470 * On ext2, we can store the error state of the filesystem in the
 471 * superblock.  That is not possible on ext4, because we may have other
 472 * write ordering constraints on the superblock which prevent us from
 473 * writing it out straight away; and given that the journal is about to
 474 * be aborted, we can't rely on the current, or future, transactions to
 475 * write out the superblock safely.
 476 *
 477 * We'll just use the jbd2_journal_abort() error code to record an error in
 478 * the journal instead.  On recovery, the journal will complain about
 479 * that error until we've noted it down and cleared it.
 480 */
 481
 482static void ext4_handle_error(struct super_block *sb)
 483{
 484        if (sb->s_flags & MS_RDONLY)
 485                return;
 486
 487        if (!test_opt(sb, ERRORS_CONT)) {
 488                journal_t *journal = EXT4_SB(sb)->s_journal;
 489
 490                EXT4_SB(sb)->s_mount_flags |= EXT4_MF_FS_ABORTED;
 491                if (journal)
 492                        jbd2_journal_abort(journal, -EIO);
 493        }
 494        if (test_opt(sb, ERRORS_RO)) {
 495                ext4_msg(sb, KERN_CRIT, "Remounting filesystem read-only");
 496                sb->s_flags |= MS_RDONLY;
 497        }
 498        if (test_opt(sb, ERRORS_PANIC))
 499                panic("EXT4-fs (device %s): panic forced after error\n",
 500                        sb->s_id);
 501}
 502
 503void __ext4_error(struct super_block *sb, const char *function,
 504                  unsigned int line, const char *fmt, ...)
 505{
 506        struct va_format vaf;
 507        va_list args;
 508
 509        va_start(args, fmt);
 510        vaf.fmt = fmt;
 511        vaf.va = &args;
 512        printk(KERN_CRIT "EXT4-fs error (device %s): %s:%d: comm %s: %pV\n",
 513               sb->s_id, function, line, current->comm, &vaf);
 514        va_end(args);
 515        save_error_info(sb, function, line);
 516
 517        ext4_handle_error(sb);
 518}
 519
 520void ext4_error_inode(struct inode *inode, const char *function,
 521                      unsigned int line, ext4_fsblk_t block,
 522                      const char *fmt, ...)
 523{
 524        va_list args;
 525        struct va_format vaf;
 526        struct ext4_super_block *es = EXT4_SB(inode->i_sb)->s_es;
 527
 528        es->s_last_error_ino = cpu_to_le32(inode->i_ino);
 529        es->s_last_error_block = cpu_to_le64(block);
 530        save_error_info(inode->i_sb, function, line);
 531        va_start(args, fmt);
 532        vaf.fmt = fmt;
 533        vaf.va = &args;
 534        if (block)
 535                printk(KERN_CRIT "EXT4-fs error (device %s): %s:%d: "
 536                       "inode #%lu: block %llu: comm %s: %pV\n",
 537                       inode->i_sb->s_id, function, line, inode->i_ino,
 538                       block, current->comm, &vaf);
 539        else
 540                printk(KERN_CRIT "EXT4-fs error (device %s): %s:%d: "
 541                       "inode #%lu: comm %s: %pV\n",
 542                       inode->i_sb->s_id, function, line, inode->i_ino,
 543                       current->comm, &vaf);
 544        va_end(args);
 545
 546        ext4_handle_error(inode->i_sb);
 547}
 548
 549void ext4_error_file(struct file *file, const char *function,
 550                     unsigned int line, ext4_fsblk_t block,
 551                     const char *fmt, ...)
 552{
 553        va_list args;
 554        struct va_format vaf;
 555        struct ext4_super_block *es;
 556        struct inode *inode = file->f_dentry->d_inode;
 557        char pathname[80], *path;
 558
 559        es = EXT4_SB(inode->i_sb)->s_es;
 560        es->s_last_error_ino = cpu_to_le32(inode->i_ino);
 561        save_error_info(inode->i_sb, function, line);
 562        path = d_path(&(file->f_path), pathname, sizeof(pathname));
 563        if (IS_ERR(path))
 564                path = "(unknown)";
 565        va_start(args, fmt);
 566        vaf.fmt = fmt;
 567        vaf.va = &args;
 568        if (block)
 569                printk(KERN_CRIT
 570                       "EXT4-fs error (device %s): %s:%d: inode #%lu: "
 571                       "block %llu: comm %s: path %s: %pV\n",
 572                       inode->i_sb->s_id, function, line, inode->i_ino,
 573                       block, current->comm, path, &vaf);
 574        else
 575                printk(KERN_CRIT
 576                       "EXT4-fs error (device %s): %s:%d: inode #%lu: "
 577                       "comm %s: path %s: %pV\n",
 578                       inode->i_sb->s_id, function, line, inode->i_ino,
 579                       current->comm, path, &vaf);
 580        va_end(args);
 581
 582        ext4_handle_error(inode->i_sb);
 583}
 584
 585static const char *ext4_decode_error(struct super_block *sb, int errno,
 586                                     char nbuf[16])
 587{
 588        char *errstr = NULL;
 589
 590        switch (errno) {
 591        case -EIO:
 592                errstr = "IO failure";
 593                break;
 594        case -ENOMEM:
 595                errstr = "Out of memory";
 596                break;
 597        case -EROFS:
 598                if (!sb || (EXT4_SB(sb)->s_journal &&
 599                            EXT4_SB(sb)->s_journal->j_flags & JBD2_ABORT))
 600                        errstr = "Journal has aborted";
 601                else
 602                        errstr = "Readonly filesystem";
 603                break;
 604        default:
 605                /* If the caller passed in an extra buffer for unknown
 606                 * errors, textualise them now.  Else we just return
 607                 * NULL. */
 608                if (nbuf) {
 609                        /* Check for truncated error codes... */
 610                        if (snprintf(nbuf, 16, "error %d", -errno) >= 0)
 611                                errstr = nbuf;
 612                }
 613                break;
 614        }
 615
 616        return errstr;
 617}
 618
 619/* __ext4_std_error decodes expected errors from journaling functions
 620 * automatically and invokes the appropriate error response.  */
 621
 622void __ext4_std_error(struct super_block *sb, const char *function,
 623                      unsigned int line, int errno)
 624{
 625        char nbuf[16];
 626        const char *errstr;
 627
 628        /* Special case: if the error is EROFS, and we're not already
 629         * inside a transaction, then there's really no point in logging
 630         * an error. */
 631        if (errno == -EROFS && journal_current_handle() == NULL &&
 632            (sb->s_flags & MS_RDONLY))
 633                return;
 634
 635        errstr = ext4_decode_error(sb, errno, nbuf);
 636        printk(KERN_CRIT "EXT4-fs error (device %s) in %s:%d: %s\n",
 637               sb->s_id, function, line, errstr);
 638        save_error_info(sb, function, line);
 639
 640        ext4_handle_error(sb);
 641}
 642
 643/*
 644 * ext4_abort is a much stronger failure handler than ext4_error.  The
 645 * abort function may be used to deal with unrecoverable failures such
 646 * as journal IO errors or ENOMEM at a critical moment in log management.
 647 *
 648 * We unconditionally force the filesystem into an ABORT|READONLY state,
 649 * unless the error response on the fs has been set to panic in which
 650 * case we take the easy way out and panic immediately.
 651 */
 652
 653void __ext4_abort(struct super_block *sb, const char *function,
 654                unsigned int line, const char *fmt, ...)
 655{
 656        va_list args;
 657
 658        save_error_info(sb, function, line);
 659        va_start(args, fmt);
 660        printk(KERN_CRIT "EXT4-fs error (device %s): %s:%d: ", sb->s_id,
 661               function, line);
 662        vprintk(fmt, args);
 663        printk("\n");
 664        va_end(args);
 665
 666        if ((sb->s_flags & MS_RDONLY) == 0) {
 667                ext4_msg(sb, KERN_CRIT, "Remounting filesystem read-only");
 668                sb->s_flags |= MS_RDONLY;
 669                EXT4_SB(sb)->s_mount_flags |= EXT4_MF_FS_ABORTED;
 670                if (EXT4_SB(sb)->s_journal)
 671                        jbd2_journal_abort(EXT4_SB(sb)->s_journal, -EIO);
 672                save_error_info(sb, function, line);
 673        }
 674        if (test_opt(sb, ERRORS_PANIC))
 675                panic("EXT4-fs panic from previous error\n");
 676}
 677
 678void ext4_msg(struct super_block *sb, const char *prefix, const char *fmt, ...)
 679{
 680        struct va_format vaf;
 681        va_list args;
 682
 683        va_start(args, fmt);
 684        vaf.fmt = fmt;
 685        vaf.va = &args;
 686        printk("%sEXT4-fs (%s): %pV\n", prefix, sb->s_id, &vaf);
 687        va_end(args);
 688}
 689
 690void __ext4_warning(struct super_block *sb, const char *function,
 691                    unsigned int line, const char *fmt, ...)
 692{
 693        struct va_format vaf;
 694        va_list args;
 695
 696        va_start(args, fmt);
 697        vaf.fmt = fmt;
 698        vaf.va = &args;
 699        printk(KERN_WARNING "EXT4-fs warning (device %s): %s:%d: %pV\n",
 700               sb->s_id, function, line, &vaf);
 701        va_end(args);
 702}
 703
 704void __ext4_grp_locked_error(const char *function, unsigned int line,
 705                             struct super_block *sb, ext4_group_t grp,
 706                             unsigned long ino, ext4_fsblk_t block,
 707                             const char *fmt, ...)
 708__releases(bitlock)
 709__acquires(bitlock)
 710{
 711        struct va_format vaf;
 712        va_list args;
 713        struct ext4_super_block *es = EXT4_SB(sb)->s_es;
 714
 715        es->s_last_error_ino = cpu_to_le32(ino);
 716        es->s_last_error_block = cpu_to_le64(block);
 717        __save_error_info(sb, function, line);
 718
 719        va_start(args, fmt);
 720
 721        vaf.fmt = fmt;
 722        vaf.va = &args;
 723        printk(KERN_CRIT "EXT4-fs error (device %s): %s:%d: group %u, ",
 724               sb->s_id, function, line, grp);
 725        if (ino)
 726                printk(KERN_CONT "inode %lu: ", ino);
 727        if (block)
 728                printk(KERN_CONT "block %llu:", (unsigned long long) block);
 729        printk(KERN_CONT "%pV\n", &vaf);
 730        va_end(args);
 731
 732        if (test_opt(sb, ERRORS_CONT)) {
 733                ext4_commit_super(sb, 0);
 734                return;
 735        }
 736
 737        ext4_unlock_group(sb, grp);
 738        ext4_handle_error(sb);
 739        /*
 740         * We only get here in the ERRORS_RO case; relocking the group
 741         * may be dangerous, but nothing bad will happen since the
 742         * filesystem will have already been marked read/only and the
 743         * journal has been aborted.  We return 1 as a hint to callers
 744         * who might what to use the return value from
 745         * ext4_grp_locked_error() to distinguish between the
 746         * ERRORS_CONT and ERRORS_RO case, and perhaps return more
 747         * aggressively from the ext4 function in question, with a
 748         * more appropriate error code.
 749         */
 750        ext4_lock_group(sb, grp);
 751        return;
 752}
 753
 754void ext4_update_dynamic_rev(struct super_block *sb)
 755{
 756        struct ext4_super_block *es = EXT4_SB(sb)->s_es;
 757
 758        if (le32_to_cpu(es->s_rev_level) > EXT4_GOOD_OLD_REV)
 759                return;
 760
 761        ext4_warning(sb,
 762                     "updating to rev %d because of new feature flag, "
 763                     "running e2fsck is recommended",
 764                     EXT4_DYNAMIC_REV);
 765
 766        es->s_first_ino = cpu_to_le32(EXT4_GOOD_OLD_FIRST_INO);
 767        es->s_inode_size = cpu_to_le16(EXT4_GOOD_OLD_INODE_SIZE);
 768        es->s_rev_level = cpu_to_le32(EXT4_DYNAMIC_REV);
 769        /* leave es->s_feature_*compat flags alone */
 770        /* es->s_uuid will be set by e2fsck if empty */
 771
 772        /*
 773         * The rest of the superblock fields should be zero, and if not it
 774         * means they are likely already in use, so leave them alone.  We
 775         * can leave it up to e2fsck to clean up any inconsistencies there.
 776         */
 777}
 778
 779/*
 780 * Open the external journal device
 781 */
 782static struct block_device *ext4_blkdev_get(dev_t dev, struct super_block *sb)
 783{
 784        struct block_device *bdev;
 785        char b[BDEVNAME_SIZE];
 786
 787        bdev = blkdev_get_by_dev(dev, FMODE_READ|FMODE_WRITE|FMODE_EXCL, sb);
 788        if (IS_ERR(bdev))
 789                goto fail;
 790        return bdev;
 791
 792fail:
 793        ext4_msg(sb, KERN_ERR, "failed to open journal device %s: %ld",
 794                        __bdevname(dev, b), PTR_ERR(bdev));
 795        return NULL;
 796}
 797
 798/*
 799 * Release the journal device
 800 */
 801static int ext4_blkdev_put(struct block_device *bdev)
 802{
 803        return blkdev_put(bdev, FMODE_READ|FMODE_WRITE|FMODE_EXCL);
 804}
 805
 806static int ext4_blkdev_remove(struct ext4_sb_info *sbi)
 807{
 808        struct block_device *bdev;
 809        int ret = -ENODEV;
 810
 811        bdev = sbi->journal_bdev;
 812        if (bdev) {
 813                ret = ext4_blkdev_put(bdev);
 814                sbi->journal_bdev = NULL;
 815        }
 816        return ret;
 817}
 818
 819static inline struct inode *orphan_list_entry(struct list_head *l)
 820{
 821        return &list_entry(l, struct ext4_inode_info, i_orphan)->vfs_inode;
 822}
 823
 824static void dump_orphan_list(struct super_block *sb, struct ext4_sb_info *sbi)
 825{
 826        struct list_head *l;
 827
 828        ext4_msg(sb, KERN_ERR, "sb orphan head is %d",
 829                 le32_to_cpu(sbi->s_es->s_last_orphan));
 830
 831        printk(KERN_ERR "sb_info orphan list:\n");
 832        list_for_each(l, &sbi->s_orphan) {
 833                struct inode *inode = orphan_list_entry(l);
 834                printk(KERN_ERR "  "
 835                       "inode %s:%lu at %p: mode %o, nlink %d, next %d\n",
 836                       inode->i_sb->s_id, inode->i_ino, inode,
 837                       inode->i_mode, inode->i_nlink,
 838                       NEXT_ORPHAN(inode));
 839        }
 840}
 841
 842static void ext4_put_super(struct super_block *sb)
 843{
 844        struct ext4_sb_info *sbi = EXT4_SB(sb);
 845        struct ext4_super_block *es = sbi->s_es;
 846        int i, err;
 847
 848        ext4_unregister_li_request(sb);
 849        dquot_disable(sb, -1, DQUOT_USAGE_ENABLED | DQUOT_LIMITS_ENABLED);
 850
 851        flush_workqueue(sbi->dio_unwritten_wq);
 852        destroy_workqueue(sbi->dio_unwritten_wq);
 853
 854        if (sbi->s_journal) {
 855                err = jbd2_journal_destroy(sbi->s_journal);
 856                sbi->s_journal = NULL;
 857                if (err < 0)
 858                        ext4_abort(sb, "Couldn't clean up the journal");
 859        }
 860
 861        del_timer(&sbi->s_err_report);
 862        ext4_release_system_zone(sb);
 863        ext4_mb_release(sb);
 864        ext4_ext_release(sb);
 865        ext4_xattr_put_super(sb);
 866
 867        if (!(sb->s_flags & MS_RDONLY)) {
 868                EXT4_CLEAR_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
 869                es->s_state = cpu_to_le16(sbi->s_mount_state);
 870        }
 871        if (!(sb->s_flags & MS_RDONLY))
 872                ext4_commit_super(sb, 1);
 873
 874        if (sbi->s_proc) {
 875                remove_proc_entry("options", sbi->s_proc);
 876                remove_proc_entry(sb->s_id, ext4_proc_root);
 877        }
 878        kobject_del(&sbi->s_kobj);
 879
 880        for (i = 0; i < sbi->s_gdb_count; i++)
 881                brelse(sbi->s_group_desc[i]);
 882        ext4_kvfree(sbi->s_group_desc);
 883        ext4_kvfree(sbi->s_flex_groups);
 884        percpu_counter_destroy(&sbi->s_freeclusters_counter);
 885        percpu_counter_destroy(&sbi->s_freeinodes_counter);
 886        percpu_counter_destroy(&sbi->s_dirs_counter);
 887        percpu_counter_destroy(&sbi->s_dirtyclusters_counter);
 888        brelse(sbi->s_sbh);
 889#ifdef CONFIG_QUOTA
 890        for (i = 0; i < MAXQUOTAS; i++)
 891                kfree(sbi->s_qf_names[i]);
 892#endif
 893
 894        /* Debugging code just in case the in-memory inode orphan list
 895         * isn't empty.  The on-disk one can be non-empty if we've
 896         * detected an error and taken the fs readonly, but the
 897         * in-memory list had better be clean by this point. */
 898        if (!list_empty(&sbi->s_orphan))
 899                dump_orphan_list(sb, sbi);
 900        J_ASSERT(list_empty(&sbi->s_orphan));
 901
 902        invalidate_bdev(sb->s_bdev);
 903        if (sbi->journal_bdev && sbi->journal_bdev != sb->s_bdev) {
 904                /*
 905                 * Invalidate the journal device's buffers.  We don't want them
 906                 * floating about in memory - the physical journal device may
 907                 * hotswapped, and it breaks the `ro-after' testing code.
 908                 */
 909                sync_blockdev(sbi->journal_bdev);
 910                invalidate_bdev(sbi->journal_bdev);
 911                ext4_blkdev_remove(sbi);
 912        }
 913        if (sbi->s_mmp_tsk)
 914                kthread_stop(sbi->s_mmp_tsk);
 915        sb->s_fs_info = NULL;
 916        /*
 917         * Now that we are completely done shutting down the
 918         * superblock, we need to actually destroy the kobject.
 919         */
 920        kobject_put(&sbi->s_kobj);
 921        wait_for_completion(&sbi->s_kobj_unregister);
 922        if (sbi->s_chksum_driver)
 923                crypto_free_shash(sbi->s_chksum_driver);
 924        kfree(sbi->s_blockgroup_lock);
 925        kfree(sbi);
 926}
 927
 928static struct kmem_cache *ext4_inode_cachep;
 929
 930/*
 931 * Called inside transaction, so use GFP_NOFS
 932 */
 933static struct inode *ext4_alloc_inode(struct super_block *sb)
 934{
 935        struct ext4_inode_info *ei;
 936
 937        ei = kmem_cache_alloc(ext4_inode_cachep, GFP_NOFS);
 938        if (!ei)
 939                return NULL;
 940
 941        ei->vfs_inode.i_version = 1;
 942        ei->vfs_inode.i_data.writeback_index = 0;
 943        memset(&ei->i_cached_extent, 0, sizeof(struct ext4_ext_cache));
 944        INIT_LIST_HEAD(&ei->i_prealloc_list);
 945        spin_lock_init(&ei->i_prealloc_lock);
 946        ei->i_reserved_data_blocks = 0;
 947        ei->i_reserved_meta_blocks = 0;
 948        ei->i_allocated_meta_blocks = 0;
 949        ei->i_da_metadata_calc_len = 0;
 950        ei->i_da_metadata_calc_last_lblock = 0;
 951        spin_lock_init(&(ei->i_block_reservation_lock));
 952#ifdef CONFIG_QUOTA
 953        ei->i_reserved_quota = 0;
 954#endif
 955        ei->jinode = NULL;
 956        INIT_LIST_HEAD(&ei->i_completed_io_list);
 957        spin_lock_init(&ei->i_completed_io_lock);
 958        ei->i_sync_tid = 0;
 959        ei->i_datasync_tid = 0;
 960        atomic_set(&ei->i_ioend_count, 0);
 961        atomic_set(&ei->i_unwritten, 0);
 962
 963        return &ei->vfs_inode;
 964}
 965
 966static int ext4_drop_inode(struct inode *inode)
 967{
 968        int drop = generic_drop_inode(inode);
 969
 970        trace_ext4_drop_inode(inode, drop);
 971        return drop;
 972}
 973
 974static void ext4_i_callback(struct rcu_head *head)
 975{
 976        struct inode *inode = container_of(head, struct inode, i_rcu);
 977        kmem_cache_free(ext4_inode_cachep, EXT4_I(inode));
 978}
 979
 980static void ext4_destroy_inode(struct inode *inode)
 981{
 982        if (!list_empty(&(EXT4_I(inode)->i_orphan))) {
 983                ext4_msg(inode->i_sb, KERN_ERR,
 984                         "Inode %lu (%p): orphan list check failed!",
 985                         inode->i_ino, EXT4_I(inode));
 986                print_hex_dump(KERN_INFO, "", DUMP_PREFIX_ADDRESS, 16, 4,
 987                                EXT4_I(inode), sizeof(struct ext4_inode_info),
 988                                true);
 989                dump_stack();
 990        }
 991        call_rcu(&inode->i_rcu, ext4_i_callback);
 992}
 993
 994static void init_once(void *foo)
 995{
 996        struct ext4_inode_info *ei = (struct ext4_inode_info *) foo;
 997
 998        INIT_LIST_HEAD(&ei->i_orphan);
 999#ifdef CONFIG_EXT4_FS_XATTR
1000        init_rwsem(&ei->xattr_sem);
1001#endif
1002        init_rwsem(&ei->i_data_sem);
1003        inode_init_once(&ei->vfs_inode);
1004}
1005
1006static int init_inodecache(void)
1007{
1008        ext4_inode_cachep = kmem_cache_create("ext4_inode_cache",
1009                                             sizeof(struct ext4_inode_info),
1010                                             0, (SLAB_RECLAIM_ACCOUNT|
1011                                                SLAB_MEM_SPREAD),
1012                                             init_once);
1013        if (ext4_inode_cachep == NULL)
1014                return -ENOMEM;
1015        return 0;
1016}
1017
1018static void destroy_inodecache(void)
1019{
1020        /*
1021         * Make sure all delayed rcu free inodes are flushed before we
1022         * destroy cache.
1023         */
1024        rcu_barrier();
1025        kmem_cache_destroy(ext4_inode_cachep);
1026}
1027
1028void ext4_clear_inode(struct inode *inode)
1029{
1030        invalidate_inode_buffers(inode);
1031        clear_inode(inode);
1032        dquot_drop(inode);
1033        ext4_discard_preallocations(inode);
1034        if (EXT4_I(inode)->jinode) {
1035                jbd2_journal_release_jbd_inode(EXT4_JOURNAL(inode),
1036                                               EXT4_I(inode)->jinode);
1037                jbd2_free_inode(EXT4_I(inode)->jinode);
1038                EXT4_I(inode)->jinode = NULL;
1039        }
1040}
1041
1042static struct inode *ext4_nfs_get_inode(struct super_block *sb,
1043                                        u64 ino, u32 generation)
1044{
1045        struct inode *inode;
1046
1047        if (ino < EXT4_FIRST_INO(sb) && ino != EXT4_ROOT_INO)
1048                return ERR_PTR(-ESTALE);
1049        if (ino > le32_to_cpu(EXT4_SB(sb)->s_es->s_inodes_count))
1050                return ERR_PTR(-ESTALE);
1051
1052        /* iget isn't really right if the inode is currently unallocated!!
1053         *
1054         * ext4_read_inode will return a bad_inode if the inode had been
1055         * deleted, so we should be safe.
1056         *
1057         * Currently we don't know the generation for parent directory, so
1058         * a generation of 0 means "accept any"
1059         */
1060        inode = ext4_iget(sb, ino);
1061        if (IS_ERR(inode))
1062                return ERR_CAST(inode);
1063        if (generation && inode->i_generation != generation) {
1064                iput(inode);
1065                return ERR_PTR(-ESTALE);
1066        }
1067
1068        return inode;
1069}
1070
1071static struct dentry *ext4_fh_to_dentry(struct super_block *sb, struct fid *fid,
1072                                        int fh_len, int fh_type)
1073{
1074        return generic_fh_to_dentry(sb, fid, fh_len, fh_type,
1075                                    ext4_nfs_get_inode);
1076}
1077
1078static struct dentry *ext4_fh_to_parent(struct super_block *sb, struct fid *fid,
1079                                        int fh_len, int fh_type)
1080{
1081        return generic_fh_to_parent(sb, fid, fh_len, fh_type,
1082                                    ext4_nfs_get_inode);
1083}
1084
1085/*
1086 * Try to release metadata pages (indirect blocks, directories) which are
1087 * mapped via the block device.  Since these pages could have journal heads
1088 * which would prevent try_to_free_buffers() from freeing them, we must use
1089 * jbd2 layer's try_to_free_buffers() function to release them.
1090 */
1091static int bdev_try_to_free_page(struct super_block *sb, struct page *page,
1092                                 gfp_t wait)
1093{
1094        journal_t *journal = EXT4_SB(sb)->s_journal;
1095
1096        WARN_ON(PageChecked(page));
1097        if (!page_has_buffers(page))
1098                return 0;
1099        if (journal)
1100                return jbd2_journal_try_to_free_buffers(journal, page,
1101                                                        wait & ~__GFP_WAIT);
1102        return try_to_free_buffers(page);
1103}
1104
1105#ifdef CONFIG_QUOTA
1106#define QTYPE2NAME(t) ((t) == USRQUOTA ? "user" : "group")
1107#define QTYPE2MOPT(on, t) ((t) == USRQUOTA?((on)##USRJQUOTA):((on)##GRPJQUOTA))
1108
1109static int ext4_write_dquot(struct dquot *dquot);
1110static int ext4_acquire_dquot(struct dquot *dquot);
1111static int ext4_release_dquot(struct dquot *dquot);
1112static int ext4_mark_dquot_dirty(struct dquot *dquot);
1113static int ext4_write_info(struct super_block *sb, int type);
1114static int ext4_quota_on(struct super_block *sb, int type, int format_id,
1115                         struct path *path);
1116static int ext4_quota_on_sysfile(struct super_block *sb, int type,
1117                                 int format_id);
1118static int ext4_quota_off(struct super_block *sb, int type);
1119static int ext4_quota_off_sysfile(struct super_block *sb, int type);
1120static int ext4_quota_on_mount(struct super_block *sb, int type);
1121static ssize_t ext4_quota_read(struct super_block *sb, int type, char *data,
1122                               size_t len, loff_t off);
1123static ssize_t ext4_quota_write(struct super_block *sb, int type,
1124                                const char *data, size_t len, loff_t off);
1125static int ext4_quota_enable(struct super_block *sb, int type, int format_id,
1126                             unsigned int flags);
1127static int ext4_enable_quotas(struct super_block *sb);
1128
1129static const struct dquot_operations ext4_quota_operations = {
1130        .get_reserved_space = ext4_get_reserved_space,
1131        .write_dquot    = ext4_write_dquot,
1132        .acquire_dquot  = ext4_acquire_dquot,
1133        .release_dquot  = ext4_release_dquot,
1134        .mark_dirty     = ext4_mark_dquot_dirty,
1135        .write_info     = ext4_write_info,
1136        .alloc_dquot    = dquot_alloc,
1137        .destroy_dquot  = dquot_destroy,
1138};
1139
1140static const struct quotactl_ops ext4_qctl_operations = {
1141        .quota_on       = ext4_quota_on,
1142        .quota_off      = ext4_quota_off,
1143        .quota_sync     = dquot_quota_sync,
1144        .get_info       = dquot_get_dqinfo,
1145        .set_info       = dquot_set_dqinfo,
1146        .get_dqblk      = dquot_get_dqblk,
1147        .set_dqblk      = dquot_set_dqblk
1148};
1149
1150static const struct quotactl_ops ext4_qctl_sysfile_operations = {
1151        .quota_on_meta  = ext4_quota_on_sysfile,
1152        .quota_off      = ext4_quota_off_sysfile,
1153        .quota_sync     = dquot_quota_sync,
1154        .get_info       = dquot_get_dqinfo,
1155        .set_info       = dquot_set_dqinfo,
1156        .get_dqblk      = dquot_get_dqblk,
1157        .set_dqblk      = dquot_set_dqblk
1158};
1159#endif
1160
1161static const struct super_operations ext4_sops = {
1162        .alloc_inode    = ext4_alloc_inode,
1163        .destroy_inode  = ext4_destroy_inode,
1164        .write_inode    = ext4_write_inode,
1165        .dirty_inode    = ext4_dirty_inode,
1166        .drop_inode     = ext4_drop_inode,
1167        .evict_inode    = ext4_evict_inode,
1168        .put_super      = ext4_put_super,
1169        .sync_fs        = ext4_sync_fs,
1170        .freeze_fs      = ext4_freeze,
1171        .unfreeze_fs    = ext4_unfreeze,
1172        .statfs         = ext4_statfs,
1173        .remount_fs     = ext4_remount,
1174        .show_options   = ext4_show_options,
1175#ifdef CONFIG_QUOTA
1176        .quota_read     = ext4_quota_read,
1177        .quota_write    = ext4_quota_write,
1178#endif
1179        .bdev_try_to_free_page = bdev_try_to_free_page,
1180};
1181
1182static const struct super_operations ext4_nojournal_sops = {
1183        .alloc_inode    = ext4_alloc_inode,
1184        .destroy_inode  = ext4_destroy_inode,
1185        .write_inode    = ext4_write_inode,
1186        .dirty_inode    = ext4_dirty_inode,
1187        .drop_inode     = ext4_drop_inode,
1188        .evict_inode    = ext4_evict_inode,
1189        .put_super      = ext4_put_super,
1190        .statfs         = ext4_statfs,
1191        .remount_fs     = ext4_remount,
1192        .show_options   = ext4_show_options,
1193#ifdef CONFIG_QUOTA
1194        .quota_read     = ext4_quota_read,
1195        .quota_write    = ext4_quota_write,
1196#endif
1197        .bdev_try_to_free_page = bdev_try_to_free_page,
1198};
1199
1200static const struct export_operations ext4_export_ops = {
1201        .fh_to_dentry = ext4_fh_to_dentry,
1202        .fh_to_parent = ext4_fh_to_parent,
1203        .get_parent = ext4_get_parent,
1204};
1205
1206enum {
1207        Opt_bsd_df, Opt_minix_df, Opt_grpid, Opt_nogrpid,
1208        Opt_resgid, Opt_resuid, Opt_sb, Opt_err_cont, Opt_err_panic, Opt_err_ro,
1209        Opt_nouid32, Opt_debug, Opt_removed,
1210        Opt_user_xattr, Opt_nouser_xattr, Opt_acl, Opt_noacl,
1211        Opt_auto_da_alloc, Opt_noauto_da_alloc, Opt_noload,
1212        Opt_commit, Opt_min_batch_time, Opt_max_batch_time,
1213        Opt_journal_dev, Opt_journal_checksum, Opt_journal_async_commit,
1214        Opt_abort, Opt_data_journal, Opt_data_ordered, Opt_data_writeback,
1215        Opt_data_err_abort, Opt_data_err_ignore,
1216        Opt_usrjquota, Opt_grpjquota, Opt_offusrjquota, Opt_offgrpjquota,
1217        Opt_jqfmt_vfsold, Opt_jqfmt_vfsv0, Opt_jqfmt_vfsv1, Opt_quota,
1218        Opt_noquota, Opt_barrier, Opt_nobarrier, Opt_err,
1219        Opt_usrquota, Opt_grpquota, Opt_i_version,
1220        Opt_stripe, Opt_delalloc, Opt_nodelalloc, Opt_mblk_io_submit,
1221        Opt_nomblk_io_submit, Opt_block_validity, Opt_noblock_validity,
1222        Opt_inode_readahead_blks, Opt_journal_ioprio,
1223        Opt_dioread_nolock, Opt_dioread_lock,
1224        Opt_discard, Opt_nodiscard, Opt_init_itable, Opt_noinit_itable,
1225        Opt_max_dir_size_kb,
1226};
1227
1228static const match_table_t tokens = {
1229        {Opt_bsd_df, "bsddf"},
1230        {Opt_minix_df, "minixdf"},
1231        {Opt_grpid, "grpid"},
1232        {Opt_grpid, "bsdgroups"},
1233        {Opt_nogrpid, "nogrpid"},
1234        {Opt_nogrpid, "sysvgroups"},
1235        {Opt_resgid, "resgid=%u"},
1236        {Opt_resuid, "resuid=%u"},
1237        {Opt_sb, "sb=%u"},
1238        {Opt_err_cont, "errors=continue"},
1239        {Opt_err_panic, "errors=panic"},
1240        {Opt_err_ro, "errors=remount-ro"},
1241        {Opt_nouid32, "nouid32"},
1242        {Opt_debug, "debug"},
1243        {Opt_removed, "oldalloc"},
1244        {Opt_removed, "orlov"},
1245        {Opt_user_xattr, "user_xattr"},
1246        {Opt_nouser_xattr, "nouser_xattr"},
1247        {Opt_acl, "acl"},
1248        {Opt_noacl, "noacl"},
1249        {Opt_noload, "norecovery"},
1250        {Opt_noload, "noload"},
1251        {Opt_removed, "nobh"},
1252        {Opt_removed, "bh"},
1253        {Opt_commit, "commit=%u"},
1254        {Opt_min_batch_time, "min_batch_time=%u"},
1255        {Opt_max_batch_time, "max_batch_time=%u"},
1256        {Opt_journal_dev, "journal_dev=%u"},
1257        {Opt_journal_checksum, "journal_checksum"},
1258        {Opt_journal_async_commit, "journal_async_commit"},
1259        {Opt_abort, "abort"},
1260        {Opt_data_journal, "data=journal"},
1261        {Opt_data_ordered, "data=ordered"},
1262        {Opt_data_writeback, "data=writeback"},
1263        {Opt_data_err_abort, "data_err=abort"},
1264        {Opt_data_err_ignore, "data_err=ignore"},
1265        {Opt_offusrjquota, "usrjquota="},
1266        {Opt_usrjquota, "usrjquota=%s"},
1267        {Opt_offgrpjquota, "grpjquota="},
1268        {Opt_grpjquota, "grpjquota=%s"},
1269        {Opt_jqfmt_vfsold, "jqfmt=vfsold"},
1270        {Opt_jqfmt_vfsv0, "jqfmt=vfsv0"},
1271        {Opt_jqfmt_vfsv1, "jqfmt=vfsv1"},
1272        {Opt_grpquota, "grpquota"},
1273        {Opt_noquota, "noquota"},
1274        {Opt_quota, "quota"},
1275        {Opt_usrquota, "usrquota"},
1276        {Opt_barrier, "barrier=%u"},
1277        {Opt_barrier, "barrier"},
1278        {Opt_nobarrier, "nobarrier"},
1279        {Opt_i_version, "i_version"},
1280        {Opt_stripe, "stripe=%u"},
1281        {Opt_delalloc, "delalloc"},
1282        {Opt_nodelalloc, "nodelalloc"},
1283        {Opt_mblk_io_submit, "mblk_io_submit"},
1284        {Opt_nomblk_io_submit, "nomblk_io_submit"},
1285        {Opt_block_validity, "block_validity"},
1286        {Opt_noblock_validity, "noblock_validity"},
1287        {Opt_inode_readahead_blks, "inode_readahead_blks=%u"},
1288        {Opt_journal_ioprio, "journal_ioprio=%u"},
1289        {Opt_auto_da_alloc, "auto_da_alloc=%u"},
1290        {Opt_auto_da_alloc, "auto_da_alloc"},
1291        {Opt_noauto_da_alloc, "noauto_da_alloc"},
1292        {Opt_dioread_nolock, "dioread_nolock"},
1293        {Opt_dioread_lock, "dioread_lock"},
1294        {Opt_discard, "discard"},
1295        {Opt_nodiscard, "nodiscard"},
1296        {Opt_init_itable, "init_itable=%u"},
1297        {Opt_init_itable, "init_itable"},
1298        {Opt_noinit_itable, "noinit_itable"},
1299        {Opt_max_dir_size_kb, "max_dir_size_kb=%u"},
1300        {Opt_removed, "check=none"},    /* mount option from ext2/3 */
1301        {Opt_removed, "nocheck"},       /* mount option from ext2/3 */
1302        {Opt_removed, "reservation"},   /* mount option from ext2/3 */
1303        {Opt_removed, "noreservation"}, /* mount option from ext2/3 */
1304        {Opt_removed, "journal=%u"},    /* mount option from ext2/3 */
1305        {Opt_err, NULL},
1306};
1307
1308static ext4_fsblk_t get_sb_block(void **data)
1309{
1310        ext4_fsblk_t    sb_block;
1311        char            *options = (char *) *data;
1312
1313        if (!options || strncmp(options, "sb=", 3) != 0)
1314                return 1;       /* Default location */
1315
1316        options += 3;
1317        /* TODO: use simple_strtoll with >32bit ext4 */
1318        sb_block = simple_strtoul(options, &options, 0);
1319        if (*options && *options != ',') {
1320                printk(KERN_ERR "EXT4-fs: Invalid sb specification: %s\n",
1321                       (char *) *data);
1322                return 1;
1323        }
1324        if (*options == ',')
1325                options++;
1326        *data = (void *) options;
1327
1328        return sb_block;
1329}
1330
1331#define DEFAULT_JOURNAL_IOPRIO (IOPRIO_PRIO_VALUE(IOPRIO_CLASS_BE, 3))
1332static char deprecated_msg[] = "Mount option \"%s\" will be removed by %s\n"
1333        "Contact linux-ext4@vger.kernel.org if you think we should keep it.\n";
1334
1335#ifdef CONFIG_QUOTA
1336static int set_qf_name(struct super_block *sb, int qtype, substring_t *args)
1337{
1338        struct ext4_sb_info *sbi = EXT4_SB(sb);
1339        char *qname;
1340
1341        if (sb_any_quota_loaded(sb) &&
1342                !sbi->s_qf_names[qtype]) {
1343                ext4_msg(sb, KERN_ERR,
1344                        "Cannot change journaled "
1345                        "quota options when quota turned on");
1346                return -1;
1347        }
1348        qname = match_strdup(args);
1349        if (!qname) {
1350                ext4_msg(sb, KERN_ERR,
1351                        "Not enough memory for storing quotafile name");
1352                return -1;
1353        }
1354        if (sbi->s_qf_names[qtype] &&
1355                strcmp(sbi->s_qf_names[qtype], qname)) {
1356                ext4_msg(sb, KERN_ERR,
1357                        "%s quota file already specified", QTYPE2NAME(qtype));
1358                kfree(qname);
1359                return -1;
1360        }
1361        sbi->s_qf_names[qtype] = qname;
1362        if (strchr(sbi->s_qf_names[qtype], '/')) {
1363                ext4_msg(sb, KERN_ERR,
1364                        "quotafile must be on filesystem root");
1365                kfree(sbi->s_qf_names[qtype]);
1366                sbi->s_qf_names[qtype] = NULL;
1367                return -1;
1368        }
1369        set_opt(sb, QUOTA);
1370        return 1;
1371}
1372
1373static int clear_qf_name(struct super_block *sb, int qtype)
1374{
1375
1376        struct ext4_sb_info *sbi = EXT4_SB(sb);
1377
1378        if (sb_any_quota_loaded(sb) &&
1379                sbi->s_qf_names[qtype]) {
1380                ext4_msg(sb, KERN_ERR, "Cannot change journaled quota options"
1381                        " when quota turned on");
1382                return -1;
1383        }
1384        /*
1385         * The space will be released later when all options are confirmed
1386         * to be correct
1387         */
1388        sbi->s_qf_names[qtype] = NULL;
1389        return 1;
1390}
1391#endif
1392
1393#define MOPT_SET        0x0001
1394#define MOPT_CLEAR      0x0002
1395#define MOPT_NOSUPPORT  0x0004
1396#define MOPT_EXPLICIT   0x0008
1397#define MOPT_CLEAR_ERR  0x0010
1398#define MOPT_GTE0       0x0020
1399#ifdef CONFIG_QUOTA
1400#define MOPT_Q          0
1401#define MOPT_QFMT       0x0040
1402#else
1403#define MOPT_Q          MOPT_NOSUPPORT
1404#define MOPT_QFMT       MOPT_NOSUPPORT
1405#endif
1406#define MOPT_DATAJ      0x0080
1407
1408static const struct mount_opts {
1409        int     token;
1410        int     mount_opt;
1411        int     flags;
1412} ext4_mount_opts[] = {
1413        {Opt_minix_df, EXT4_MOUNT_MINIX_DF, MOPT_SET},
1414        {Opt_bsd_df, EXT4_MOUNT_MINIX_DF, MOPT_CLEAR},
1415        {Opt_grpid, EXT4_MOUNT_GRPID, MOPT_SET},
1416        {Opt_nogrpid, EXT4_MOUNT_GRPID, MOPT_CLEAR},
1417        {Opt_mblk_io_submit, EXT4_MOUNT_MBLK_IO_SUBMIT, MOPT_SET},
1418        {Opt_nomblk_io_submit, EXT4_MOUNT_MBLK_IO_SUBMIT, MOPT_CLEAR},
1419        {Opt_block_validity, EXT4_MOUNT_BLOCK_VALIDITY, MOPT_SET},
1420        {Opt_noblock_validity, EXT4_MOUNT_BLOCK_VALIDITY, MOPT_CLEAR},
1421        {Opt_dioread_nolock, EXT4_MOUNT_DIOREAD_NOLOCK, MOPT_SET},
1422        {Opt_dioread_lock, EXT4_MOUNT_DIOREAD_NOLOCK, MOPT_CLEAR},
1423        {Opt_discard, EXT4_MOUNT_DISCARD, MOPT_SET},
1424        {Opt_nodiscard, EXT4_MOUNT_DISCARD, MOPT_CLEAR},
1425        {Opt_delalloc, EXT4_MOUNT_DELALLOC, MOPT_SET | MOPT_EXPLICIT},
1426        {Opt_nodelalloc, EXT4_MOUNT_DELALLOC, MOPT_CLEAR | MOPT_EXPLICIT},
1427        {Opt_journal_checksum, EXT4_MOUNT_JOURNAL_CHECKSUM, MOPT_SET},
1428        {Opt_journal_async_commit, (EXT4_MOUNT_JOURNAL_ASYNC_COMMIT |
1429                                    EXT4_MOUNT_JOURNAL_CHECKSUM), MOPT_SET},
1430        {Opt_noload, EXT4_MOUNT_NOLOAD, MOPT_SET},
1431        {Opt_err_panic, EXT4_MOUNT_ERRORS_PANIC, MOPT_SET | MOPT_CLEAR_ERR},
1432        {Opt_err_ro, EXT4_MOUNT_ERRORS_RO, MOPT_SET | MOPT_CLEAR_ERR},
1433        {Opt_err_cont, EXT4_MOUNT_ERRORS_CONT, MOPT_SET | MOPT_CLEAR_ERR},
1434        {Opt_data_err_abort, EXT4_MOUNT_DATA_ERR_ABORT, MOPT_SET},
1435        {Opt_data_err_ignore, EXT4_MOUNT_DATA_ERR_ABORT, MOPT_CLEAR},
1436        {Opt_barrier, EXT4_MOUNT_BARRIER, MOPT_SET},
1437        {Opt_nobarrier, EXT4_MOUNT_BARRIER, MOPT_CLEAR},
1438        {Opt_noauto_da_alloc, EXT4_MOUNT_NO_AUTO_DA_ALLOC, MOPT_SET},
1439        {Opt_auto_da_alloc, EXT4_MOUNT_NO_AUTO_DA_ALLOC, MOPT_CLEAR},
1440        {Opt_noinit_itable, EXT4_MOUNT_INIT_INODE_TABLE, MOPT_CLEAR},
1441        {Opt_commit, 0, MOPT_GTE0},
1442        {Opt_max_batch_time, 0, MOPT_GTE0},
1443        {Opt_min_batch_time, 0, MOPT_GTE0},
1444        {Opt_inode_readahead_blks, 0, MOPT_GTE0},
1445        {Opt_init_itable, 0, MOPT_GTE0},
1446        {Opt_stripe, 0, MOPT_GTE0},
1447        {Opt_data_journal, EXT4_MOUNT_JOURNAL_DATA, MOPT_DATAJ},
1448        {Opt_data_ordered, EXT4_MOUNT_ORDERED_DATA, MOPT_DATAJ},
1449        {Opt_data_writeback, EXT4_MOUNT_WRITEBACK_DATA, MOPT_DATAJ},
1450#ifdef CONFIG_EXT4_FS_XATTR
1451        {Opt_user_xattr, EXT4_MOUNT_XATTR_USER, MOPT_SET},
1452        {Opt_nouser_xattr, EXT4_MOUNT_XATTR_USER, MOPT_CLEAR},
1453#else
1454        {Opt_user_xattr, 0, MOPT_NOSUPPORT},
1455        {Opt_nouser_xattr, 0, MOPT_NOSUPPORT},
1456#endif
1457#ifdef CONFIG_EXT4_FS_POSIX_ACL
1458        {Opt_acl, EXT4_MOUNT_POSIX_ACL, MOPT_SET},
1459        {Opt_noacl, EXT4_MOUNT_POSIX_ACL, MOPT_CLEAR},
1460#else
1461        {Opt_acl, 0, MOPT_NOSUPPORT},
1462        {Opt_noacl, 0, MOPT_NOSUPPORT},
1463#endif
1464        {Opt_nouid32, EXT4_MOUNT_NO_UID32, MOPT_SET},
1465        {Opt_debug, EXT4_MOUNT_DEBUG, MOPT_SET},
1466        {Opt_quota, EXT4_MOUNT_QUOTA | EXT4_MOUNT_USRQUOTA, MOPT_SET | MOPT_Q},
1467        {Opt_usrquota, EXT4_MOUNT_QUOTA | EXT4_MOUNT_USRQUOTA,
1468                                                        MOPT_SET | MOPT_Q},
1469        {Opt_grpquota, EXT4_MOUNT_QUOTA | EXT4_MOUNT_GRPQUOTA,
1470                                                        MOPT_SET | MOPT_Q},
1471        {Opt_noquota, (EXT4_MOUNT_QUOTA | EXT4_MOUNT_USRQUOTA |
1472                       EXT4_MOUNT_GRPQUOTA), MOPT_CLEAR | MOPT_Q},
1473        {Opt_usrjquota, 0, MOPT_Q},
1474        {Opt_grpjquota, 0, MOPT_Q},
1475        {Opt_offusrjquota, 0, MOPT_Q},
1476        {Opt_offgrpjquota, 0, MOPT_Q},
1477        {Opt_jqfmt_vfsold, QFMT_VFS_OLD, MOPT_QFMT},
1478        {Opt_jqfmt_vfsv0, QFMT_VFS_V0, MOPT_QFMT},
1479        {Opt_jqfmt_vfsv1, QFMT_VFS_V1, MOPT_QFMT},
1480        {Opt_max_dir_size_kb, 0, MOPT_GTE0},
1481        {Opt_err, 0, 0}
1482};
1483
1484static int handle_mount_opt(struct super_block *sb, char *opt, int token,
1485                            substring_t *args, unsigned long *journal_devnum,
1486                            unsigned int *journal_ioprio, int is_remount)
1487{
1488        struct ext4_sb_info *sbi = EXT4_SB(sb);
1489        const struct mount_opts *m;
1490        kuid_t uid;
1491        kgid_t gid;
1492        int arg = 0;
1493
1494#ifdef CONFIG_QUOTA
1495        if (token == Opt_usrjquota)
1496                return set_qf_name(sb, USRQUOTA, &args[0]);
1497        else if (token == Opt_grpjquota)
1498                return set_qf_name(sb, GRPQUOTA, &args[0]);
1499        else if (token == Opt_offusrjquota)
1500                return clear_qf_name(sb, USRQUOTA);
1501        else if (token == Opt_offgrpjquota)
1502                return clear_qf_name(sb, GRPQUOTA);
1503#endif
1504        if (args->from && match_int(args, &arg))
1505                return -1;
1506        switch (token) {
1507        case Opt_noacl:
1508        case Opt_nouser_xattr:
1509                ext4_msg(sb, KERN_WARNING, deprecated_msg, opt, "3.5");
1510                break;
1511        case Opt_sb:
1512                return 1;       /* handled by get_sb_block() */
1513        case Opt_removed:
1514                ext4_msg(sb, KERN_WARNING,
1515                         "Ignoring removed %s option", opt);
1516                return 1;
1517        case Opt_resuid:
1518                uid = make_kuid(current_user_ns(), arg);
1519                if (!uid_valid(uid)) {
1520                        ext4_msg(sb, KERN_ERR, "Invalid uid value %d", arg);
1521                        return -1;
1522                }
1523                sbi->s_resuid = uid;
1524                return 1;
1525        case Opt_resgid:
1526                gid = make_kgid(current_user_ns(), arg);
1527                if (!gid_valid(gid)) {
1528                        ext4_msg(sb, KERN_ERR, "Invalid gid value %d", arg);
1529                        return -1;
1530                }
1531                sbi->s_resgid = gid;
1532                return 1;
1533        case Opt_abort:
1534                sbi->s_mount_flags |= EXT4_MF_FS_ABORTED;
1535                return 1;
1536        case Opt_i_version:
1537                sb->s_flags |= MS_I_VERSION;
1538                return 1;
1539        case Opt_journal_dev:
1540                if (is_remount) {
1541                        ext4_msg(sb, KERN_ERR,
1542                                 "Cannot specify journal on remount");
1543                        return -1;
1544                }
1545                *journal_devnum = arg;
1546                return 1;
1547        case Opt_journal_ioprio:
1548                if (arg < 0 || arg > 7)
1549                        return -1;
1550                *journal_ioprio = IOPRIO_PRIO_VALUE(IOPRIO_CLASS_BE, arg);
1551                return 1;
1552        }
1553
1554        for (m = ext4_mount_opts; m->token != Opt_err; m++) {
1555                if (token != m->token)
1556                        continue;
1557                if (args->from && (m->flags & MOPT_GTE0) && (arg < 0))
1558                        return -1;
1559                if (m->flags & MOPT_EXPLICIT)
1560                        set_opt2(sb, EXPLICIT_DELALLOC);
1561                if (m->flags & MOPT_CLEAR_ERR)
1562                        clear_opt(sb, ERRORS_MASK);
1563                if (token == Opt_noquota && sb_any_quota_loaded(sb)) {
1564                        ext4_msg(sb, KERN_ERR, "Cannot change quota "
1565                                 "options when quota turned on");
1566                        return -1;
1567                }
1568
1569                if (m->flags & MOPT_NOSUPPORT) {
1570                        ext4_msg(sb, KERN_ERR, "%s option not supported", opt);
1571                } else if (token == Opt_commit) {
1572                        if (arg == 0)
1573                                arg = JBD2_DEFAULT_MAX_COMMIT_AGE;
1574                        sbi->s_commit_interval = HZ * arg;
1575                } else if (token == Opt_max_batch_time) {
1576                        if (arg == 0)
1577                                arg = EXT4_DEF_MAX_BATCH_TIME;
1578                        sbi->s_max_batch_time = arg;
1579                } else if (token == Opt_min_batch_time) {
1580                        sbi->s_min_batch_time = arg;
1581                } else if (token == Opt_inode_readahead_blks) {
1582                        if (arg > (1 << 30))
1583                                return -1;
1584                        if (arg && !is_power_of_2(arg)) {
1585                                ext4_msg(sb, KERN_ERR,
1586                                         "EXT4-fs: inode_readahead_blks"
1587                                         " must be a power of 2");
1588                                return -1;
1589                        }
1590                        sbi->s_inode_readahead_blks = arg;
1591                } else if (token == Opt_init_itable) {
1592                        set_opt(sb, INIT_INODE_TABLE);
1593                        if (!args->from)
1594                                arg = EXT4_DEF_LI_WAIT_MULT;
1595                        sbi->s_li_wait_mult = arg;
1596                } else if (token == Opt_max_dir_size_kb) {
1597                        sbi->s_max_dir_size_kb = arg;
1598                } else if (token == Opt_stripe) {
1599                        sbi->s_stripe = arg;
1600                } else if (m->flags & MOPT_DATAJ) {
1601                        if (is_remount) {
1602                                if (!sbi->s_journal)
1603                                        ext4_msg(sb, KERN_WARNING, "Remounting file system with no journal so ignoring journalled data option");
1604                                else if (test_opt(sb, DATA_FLAGS) !=
1605                                         m->mount_opt) {
1606                                        ext4_msg(sb, KERN_ERR,
1607                                         "Cannot change data mode on remount");
1608                                        return -1;
1609                                }
1610                        } else {
1611                                clear_opt(sb, DATA_FLAGS);
1612                                sbi->s_mount_opt |= m->mount_opt;
1613                        }
1614#ifdef CONFIG_QUOTA
1615                } else if (m->flags & MOPT_QFMT) {
1616                        if (sb_any_quota_loaded(sb) &&
1617                            sbi->s_jquota_fmt != m->mount_opt) {
1618                                ext4_msg(sb, KERN_ERR, "Cannot "
1619                                         "change journaled quota options "
1620                                         "when quota turned on");
1621                                return -1;
1622                        }
1623                        sbi->s_jquota_fmt = m->mount_opt;
1624#endif
1625                } else {
1626                        if (!args->from)
1627                                arg = 1;
1628                        if (m->flags & MOPT_CLEAR)
1629                                arg = !arg;
1630                        else if (unlikely(!(m->flags & MOPT_SET))) {
1631                                ext4_msg(sb, KERN_WARNING,
1632                                         "buggy handling of option %s", opt);
1633                                WARN_ON(1);
1634                                return -1;
1635                        }
1636                        if (arg != 0)
1637                                sbi->s_mount_opt |= m->mount_opt;
1638                        else
1639                                sbi->s_mount_opt &= ~m->mount_opt;
1640                }
1641                return 1;
1642        }
1643        ext4_msg(sb, KERN_ERR, "Unrecognized mount option \"%s\" "
1644                 "or missing value", opt);
1645        return -1;
1646}
1647
1648static int parse_options(char *options, struct super_block *sb,
1649                         unsigned long *journal_devnum,
1650                         unsigned int *journal_ioprio,
1651                         int is_remount)
1652{
1653        struct ext4_sb_info *sbi = EXT4_SB(sb);
1654        char *p;
1655        substring_t args[MAX_OPT_ARGS];
1656        int token;
1657
1658        if (!options)
1659                return 1;
1660
1661        while ((p = strsep(&options, ",")) != NULL) {
1662                if (!*p)
1663                        continue;
1664                /*
1665                 * Initialize args struct so we know whether arg was
1666                 * found; some options take optional arguments.
1667                 */
1668                args[0].to = args[0].from = NULL;
1669                token = match_token(p, tokens, args);
1670                if (handle_mount_opt(sb, p, token, args, journal_devnum,
1671                                     journal_ioprio, is_remount) < 0)
1672                        return 0;
1673        }
1674#ifdef CONFIG_QUOTA
1675        if (sbi->s_qf_names[USRQUOTA] || sbi->s_qf_names[GRPQUOTA]) {
1676                if (test_opt(sb, USRQUOTA) && sbi->s_qf_names[USRQUOTA])
1677                        clear_opt(sb, USRQUOTA);
1678
1679                if (test_opt(sb, GRPQUOTA) && sbi->s_qf_names[GRPQUOTA])
1680                        clear_opt(sb, GRPQUOTA);
1681
1682                if (test_opt(sb, GRPQUOTA) || test_opt(sb, USRQUOTA)) {
1683                        ext4_msg(sb, KERN_ERR, "old and new quota "
1684                                        "format mixing");
1685                        return 0;
1686                }
1687
1688                if (!sbi->s_jquota_fmt) {
1689                        ext4_msg(sb, KERN_ERR, "journaled quota format "
1690                                        "not specified");
1691                        return 0;
1692                }
1693        } else {
1694                if (sbi->s_jquota_fmt) {
1695                        ext4_msg(sb, KERN_ERR, "journaled quota format "
1696                                        "specified with no journaling "
1697                                        "enabled");
1698                        return 0;
1699                }
1700        }
1701#endif
1702        if (test_opt(sb, DIOREAD_NOLOCK)) {
1703                int blocksize =
1704                        BLOCK_SIZE << le32_to_cpu(sbi->s_es->s_log_block_size);
1705
1706                if (blocksize < PAGE_CACHE_SIZE) {
1707                        ext4_msg(sb, KERN_ERR, "can't mount with "
1708                                 "dioread_nolock if block size != PAGE_SIZE");
1709                        return 0;
1710                }
1711        }
1712        return 1;
1713}
1714
1715static inline void ext4_show_quota_options(struct seq_file *seq,
1716                                           struct super_block *sb)
1717{
1718#if defined(CONFIG_QUOTA)
1719        struct ext4_sb_info *sbi = EXT4_SB(sb);
1720
1721        if (sbi->s_jquota_fmt) {
1722                char *fmtname = "";
1723
1724                switch (sbi->s_jquota_fmt) {
1725                case QFMT_VFS_OLD:
1726                        fmtname = "vfsold";
1727                        break;
1728                case QFMT_VFS_V0:
1729                        fmtname = "vfsv0";
1730                        break;
1731                case QFMT_VFS_V1:
1732                        fmtname = "vfsv1";
1733                        break;
1734                }
1735                seq_printf(seq, ",jqfmt=%s", fmtname);
1736        }
1737
1738        if (sbi->s_qf_names[USRQUOTA])
1739                seq_printf(seq, ",usrjquota=%s", sbi->s_qf_names[USRQUOTA]);
1740
1741        if (sbi->s_qf_names[GRPQUOTA])
1742                seq_printf(seq, ",grpjquota=%s", sbi->s_qf_names[GRPQUOTA]);
1743
1744        if (test_opt(sb, USRQUOTA))
1745                seq_puts(seq, ",usrquota");
1746
1747        if (test_opt(sb, GRPQUOTA))
1748                seq_puts(seq, ",grpquota");
1749#endif
1750}
1751
1752static const char *token2str(int token)
1753{
1754        const struct match_token *t;
1755
1756        for (t = tokens; t->token != Opt_err; t++)
1757                if (t->token == token && !strchr(t->pattern, '='))
1758                        break;
1759        return t->pattern;
1760}
1761
1762/*
1763 * Show an option if
1764 *  - it's set to a non-default value OR
1765 *  - if the per-sb default is different from the global default
1766 */
1767static int _ext4_show_options(struct seq_file *seq, struct super_block *sb,
1768                              int nodefs)
1769{
1770        struct ext4_sb_info *sbi = EXT4_SB(sb);
1771        struct ext4_super_block *es = sbi->s_es;
1772        int def_errors, def_mount_opt = nodefs ? 0 : sbi->s_def_mount_opt;
1773        const struct mount_opts *m;
1774        char sep = nodefs ? '\n' : ',';
1775
1776#define SEQ_OPTS_PUTS(str) seq_printf(seq, "%c" str, sep)
1777#define SEQ_OPTS_PRINT(str, arg) seq_printf(seq, "%c" str, sep, arg)
1778
1779        if (sbi->s_sb_block != 1)
1780                SEQ_OPTS_PRINT("sb=%llu", sbi->s_sb_block);
1781
1782        for (m = ext4_mount_opts; m->token != Opt_err; m++) {
1783                int want_set = m->flags & MOPT_SET;
1784                if (((m->flags & (MOPT_SET|MOPT_CLEAR)) == 0) ||
1785                    (m->flags & MOPT_CLEAR_ERR))
1786                        continue;
1787                if (!(m->mount_opt & (sbi->s_mount_opt ^ def_mount_opt)))
1788                        continue; /* skip if same as the default */
1789                if ((want_set &&
1790                     (sbi->s_mount_opt & m->mount_opt) != m->mount_opt) ||
1791                    (!want_set && (sbi->s_mount_opt & m->mount_opt)))
1792                        continue; /* select Opt_noFoo vs Opt_Foo */
1793                SEQ_OPTS_PRINT("%s", token2str(m->token));
1794        }
1795
1796        if (nodefs || !uid_eq(sbi->s_resuid, make_kuid(&init_user_ns, EXT4_DEF_RESUID)) ||
1797            le16_to_cpu(es->s_def_resuid) != EXT4_DEF_RESUID)
1798                SEQ_OPTS_PRINT("resuid=%u",
1799                                from_kuid_munged(&init_user_ns, sbi->s_resuid));
1800        if (nodefs || !gid_eq(sbi->s_resgid, make_kgid(&init_user_ns, EXT4_DEF_RESGID)) ||
1801            le16_to_cpu(es->s_def_resgid) != EXT4_DEF_RESGID)
1802                SEQ_OPTS_PRINT("resgid=%u",
1803                                from_kgid_munged(&init_user_ns, sbi->s_resgid));
1804        def_errors = nodefs ? -1 : le16_to_cpu(es->s_errors);
1805        if (test_opt(sb, ERRORS_RO) && def_errors != EXT4_ERRORS_RO)
1806                SEQ_OPTS_PUTS("errors=remount-ro");
1807        if (test_opt(sb, ERRORS_CONT) && def_errors != EXT4_ERRORS_CONTINUE)
1808                SEQ_OPTS_PUTS("errors=continue");
1809        if (test_opt(sb, ERRORS_PANIC) && def_errors != EXT4_ERRORS_PANIC)
1810                SEQ_OPTS_PUTS("errors=panic");
1811        if (nodefs || sbi->s_commit_interval != JBD2_DEFAULT_MAX_COMMIT_AGE*HZ)
1812                SEQ_OPTS_PRINT("commit=%lu", sbi->s_commit_interval / HZ);
1813        if (nodefs || sbi->s_min_batch_time != EXT4_DEF_MIN_BATCH_TIME)
1814                SEQ_OPTS_PRINT("min_batch_time=%u", sbi->s_min_batch_time);
1815        if (nodefs || sbi->s_max_batch_time != EXT4_DEF_MAX_BATCH_TIME)
1816                SEQ_OPTS_PRINT("max_batch_time=%u", sbi->s_max_batch_time);
1817        if (sb->s_flags & MS_I_VERSION)
1818                SEQ_OPTS_PUTS("i_version");
1819        if (nodefs || sbi->s_stripe)
1820                SEQ_OPTS_PRINT("stripe=%lu", sbi->s_stripe);
1821        if (EXT4_MOUNT_DATA_FLAGS & (sbi->s_mount_opt ^ def_mount_opt)) {
1822                if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA)
1823                        SEQ_OPTS_PUTS("data=journal");
1824                else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA)
1825                        SEQ_OPTS_PUTS("data=ordered");
1826                else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_WRITEBACK_DATA)
1827                        SEQ_OPTS_PUTS("data=writeback");
1828        }
1829        if (nodefs ||
1830            sbi->s_inode_readahead_blks != EXT4_DEF_INODE_READAHEAD_BLKS)
1831                SEQ_OPTS_PRINT("inode_readahead_blks=%u",
1832                               sbi->s_inode_readahead_blks);
1833
1834        if (nodefs || (test_opt(sb, INIT_INODE_TABLE) &&
1835                       (sbi->s_li_wait_mult != EXT4_DEF_LI_WAIT_MULT)))
1836                SEQ_OPTS_PRINT("init_itable=%u", sbi->s_li_wait_mult);
1837        if (nodefs || sbi->s_max_dir_size_kb)
1838                SEQ_OPTS_PRINT("max_dir_size_kb=%u", sbi->s_max_dir_size_kb);
1839
1840        ext4_show_quota_options(seq, sb);
1841        return 0;
1842}
1843
1844static int ext4_show_options(struct seq_file *seq, struct dentry *root)
1845{
1846        return _ext4_show_options(seq, root->d_sb, 0);
1847}
1848
1849static int options_seq_show(struct seq_file *seq, void *offset)
1850{
1851        struct super_block *sb = seq->private;
1852        int rc;
1853
1854        seq_puts(seq, (sb->s_flags & MS_RDONLY) ? "ro" : "rw");
1855        rc = _ext4_show_options(seq, sb, 1);
1856        seq_puts(seq, "\n");
1857        return rc;
1858}
1859
1860static int options_open_fs(struct inode *inode, struct file *file)
1861{
1862        return single_open(file, options_seq_show, PDE(inode)->data);
1863}
1864
1865static const struct file_operations ext4_seq_options_fops = {
1866        .owner = THIS_MODULE,
1867        .open = options_open_fs,
1868        .read = seq_read,
1869        .llseek = seq_lseek,
1870        .release = single_release,
1871};
1872
1873static int ext4_setup_super(struct super_block *sb, struct ext4_super_block *es,
1874                            int read_only)
1875{
1876        struct ext4_sb_info *sbi = EXT4_SB(sb);
1877        int res = 0;
1878
1879        if (le32_to_cpu(es->s_rev_level) > EXT4_MAX_SUPP_REV) {
1880                ext4_msg(sb, KERN_ERR, "revision level too high, "
1881                         "forcing read-only mode");
1882                res = MS_RDONLY;
1883        }
1884        if (read_only)
1885                goto done;
1886        if (!(sbi->s_mount_state & EXT4_VALID_FS))
1887                ext4_msg(sb, KERN_WARNING, "warning: mounting unchecked fs, "
1888                         "running e2fsck is recommended");
1889        else if ((sbi->s_mount_state & EXT4_ERROR_FS))
1890                ext4_msg(sb, KERN_WARNING,
1891                         "warning: mounting fs with errors, "
1892                         "running e2fsck is recommended");
1893        else if ((__s16) le16_to_cpu(es->s_max_mnt_count) > 0 &&
1894                 le16_to_cpu(es->s_mnt_count) >=
1895                 (unsigned short) (__s16) le16_to_cpu(es->s_max_mnt_count))
1896                ext4_msg(sb, KERN_WARNING,
1897                         "warning: maximal mount count reached, "
1898                         "running e2fsck is recommended");
1899        else if (le32_to_cpu(es->s_checkinterval) &&
1900                (le32_to_cpu(es->s_lastcheck) +
1901                        le32_to_cpu(es->s_checkinterval) <= get_seconds()))
1902                ext4_msg(sb, KERN_WARNING,
1903                         "warning: checktime reached, "
1904                         "running e2fsck is recommended");
1905        if (!sbi->s_journal)
1906                es->s_state &= cpu_to_le16(~EXT4_VALID_FS);
1907        if (!(__s16) le16_to_cpu(es->s_max_mnt_count))
1908                es->s_max_mnt_count = cpu_to_le16(EXT4_DFL_MAX_MNT_COUNT);
1909        le16_add_cpu(&es->s_mnt_count, 1);
1910        es->s_mtime = cpu_to_le32(get_seconds());
1911        ext4_update_dynamic_rev(sb);
1912        if (sbi->s_journal)
1913                EXT4_SET_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
1914
1915        ext4_commit_super(sb, 1);
1916done:
1917        if (test_opt(sb, DEBUG))
1918                printk(KERN_INFO "[EXT4 FS bs=%lu, gc=%u, "
1919                                "bpg=%lu, ipg=%lu, mo=%04x, mo2=%04x]\n",
1920                        sb->s_blocksize,
1921                        sbi->s_groups_count,
1922                        EXT4_BLOCKS_PER_GROUP(sb),
1923                        EXT4_INODES_PER_GROUP(sb),
1924                        sbi->s_mount_opt, sbi->s_mount_opt2);
1925
1926        cleancache_init_fs(sb);
1927        return res;
1928}
1929
1930int ext4_alloc_flex_bg_array(struct super_block *sb, ext4_group_t ngroup)
1931{
1932        struct ext4_sb_info *sbi = EXT4_SB(sb);
1933        struct flex_groups *new_groups;
1934        int size;
1935
1936        if (!sbi->s_log_groups_per_flex)
1937                return 0;
1938
1939        size = ext4_flex_group(sbi, ngroup - 1) + 1;
1940        if (size <= sbi->s_flex_groups_allocated)
1941                return 0;
1942
1943        size = roundup_pow_of_two(size * sizeof(struct flex_groups));
1944        new_groups = ext4_kvzalloc(size, GFP_KERNEL);
1945        if (!new_groups) {
1946                ext4_msg(sb, KERN_ERR, "not enough memory for %d flex groups",
1947                         size / (int) sizeof(struct flex_groups));
1948                return -ENOMEM;
1949        }
1950
1951        if (sbi->s_flex_groups) {
1952                memcpy(new_groups, sbi->s_flex_groups,
1953                       (sbi->s_flex_groups_allocated *
1954                        sizeof(struct flex_groups)));
1955                ext4_kvfree(sbi->s_flex_groups);
1956        }
1957        sbi->s_flex_groups = new_groups;
1958        sbi->s_flex_groups_allocated = size / sizeof(struct flex_groups);
1959        return 0;
1960}
1961
1962static int ext4_fill_flex_info(struct super_block *sb)
1963{
1964        struct ext4_sb_info *sbi = EXT4_SB(sb);
1965        struct ext4_group_desc *gdp = NULL;
1966        ext4_group_t flex_group;
1967        unsigned int groups_per_flex = 0;
1968        int i, err;
1969
1970        sbi->s_log_groups_per_flex = sbi->s_es->s_log_groups_per_flex;
1971        if (sbi->s_log_groups_per_flex < 1 || sbi->s_log_groups_per_flex > 31) {
1972                sbi->s_log_groups_per_flex = 0;
1973                return 1;
1974        }
1975        groups_per_flex = 1U << sbi->s_log_groups_per_flex;
1976
1977        err = ext4_alloc_flex_bg_array(sb, sbi->s_groups_count);
1978        if (err)
1979                goto failed;
1980
1981        for (i = 0; i < sbi->s_groups_count; i++) {
1982                gdp = ext4_get_group_desc(sb, i, NULL);
1983
1984                flex_group = ext4_flex_group(sbi, i);
1985                atomic_add(ext4_free_inodes_count(sb, gdp),
1986                           &sbi->s_flex_groups[flex_group].free_inodes);
1987                atomic_add(ext4_free_group_clusters(sb, gdp),
1988                           &sbi->s_flex_groups[flex_group].free_clusters);
1989                atomic_add(ext4_used_dirs_count(sb, gdp),
1990                           &sbi->s_flex_groups[flex_group].used_dirs);
1991        }
1992
1993        return 1;
1994failed:
1995        return 0;
1996}
1997
1998static __le16 ext4_group_desc_csum(struct ext4_sb_info *sbi, __u32 block_group,
1999                                   struct ext4_group_desc *gdp)
2000{
2001        int offset;
2002        __u16 crc = 0;
2003        __le32 le_group = cpu_to_le32(block_group);
2004
2005        if ((sbi->s_es->s_feature_ro_compat &
2006             cpu_to_le32(EXT4_FEATURE_RO_COMPAT_METADATA_CSUM))) {
2007                /* Use new metadata_csum algorithm */
2008                __u16 old_csum;
2009                __u32 csum32;
2010
2011                old_csum = gdp->bg_checksum;
2012                gdp->bg_checksum = 0;
2013                csum32 = ext4_chksum(sbi, sbi->s_csum_seed, (__u8 *)&le_group,
2014                                     sizeof(le_group));
2015                csum32 = ext4_chksum(sbi, csum32, (__u8 *)gdp,
2016                                     sbi->s_desc_size);
2017                gdp->bg_checksum = old_csum;
2018
2019                crc = csum32 & 0xFFFF;
2020                goto out;
2021        }
2022
2023        /* old crc16 code */
2024        offset = offsetof(struct ext4_group_desc, bg_checksum);
2025
2026        crc = crc16(~0, sbi->s_es->s_uuid, sizeof(sbi->s_es->s_uuid));
2027        crc = crc16(crc, (__u8 *)&le_group, sizeof(le_group));
2028        crc = crc16(crc, (__u8 *)gdp, offset);
2029        offset += sizeof(gdp->bg_checksum); /* skip checksum */
2030        /* for checksum of struct ext4_group_desc do the rest...*/
2031        if ((sbi->s_es->s_feature_incompat &
2032             cpu_to_le32(EXT4_FEATURE_INCOMPAT_64BIT)) &&
2033            offset < le16_to_cpu(sbi->s_es->s_desc_size))
2034                crc = crc16(crc, (__u8 *)gdp + offset,
2035                            le16_to_cpu(sbi->s_es->s_desc_size) -
2036                                offset);
2037
2038out:
2039        return cpu_to_le16(crc);
2040}
2041
2042int ext4_group_desc_csum_verify(struct super_block *sb, __u32 block_group,
2043                                struct ext4_group_desc *gdp)
2044{
2045        if (ext4_has_group_desc_csum(sb) &&
2046            (gdp->bg_checksum != ext4_group_desc_csum(EXT4_SB(sb),
2047                                                      block_group, gdp)))
2048                return 0;
2049
2050        return 1;
2051}
2052
2053void ext4_group_desc_csum_set(struct super_block *sb, __u32 block_group,
2054                              struct ext4_group_desc *gdp)
2055{
2056        if (!ext4_has_group_desc_csum(sb))
2057                return;
2058        gdp->bg_checksum = ext4_group_desc_csum(EXT4_SB(sb), block_group, gdp);
2059}
2060
2061/* Called at mount-time, super-block is locked */
2062static int ext4_check_descriptors(struct super_block *sb,
2063                                  ext4_group_t *first_not_zeroed)
2064{
2065        struct ext4_sb_info *sbi = EXT4_SB(sb);
2066        ext4_fsblk_t first_block = le32_to_cpu(sbi->s_es->s_first_data_block);
2067        ext4_fsblk_t last_block;
2068        ext4_fsblk_t block_bitmap;
2069        ext4_fsblk_t inode_bitmap;
2070        ext4_fsblk_t inode_table;
2071        int flexbg_flag = 0;
2072        ext4_group_t i, grp = sbi->s_groups_count;
2073
2074        if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_FLEX_BG))
2075                flexbg_flag = 1;
2076
2077        ext4_debug("Checking group descriptors");
2078
2079        for (i = 0; i < sbi->s_groups_count; i++) {
2080                struct ext4_group_desc *gdp = ext4_get_group_desc(sb, i, NULL);
2081
2082                if (i == sbi->s_groups_count - 1 || flexbg_flag)
2083                        last_block = ext4_blocks_count(sbi->s_es) - 1;
2084                else
2085                        last_block = first_block +
2086                                (EXT4_BLOCKS_PER_GROUP(sb) - 1);
2087
2088                if ((grp == sbi->s_groups_count) &&
2089                   !(gdp->bg_flags & cpu_to_le16(EXT4_BG_INODE_ZEROED)))
2090                        grp = i;
2091
2092                block_bitmap = ext4_block_bitmap(sb, gdp);
2093                if (block_bitmap < first_block || block_bitmap > last_block) {
2094                        ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2095                               "Block bitmap for group %u not in group "
2096                               "(block %llu)!", i, block_bitmap);
2097                        return 0;
2098                }
2099                inode_bitmap = ext4_inode_bitmap(sb, gdp);
2100                if (inode_bitmap < first_block || inode_bitmap > last_block) {
2101                        ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2102                               "Inode bitmap for group %u not in group "
2103                               "(block %llu)!", i, inode_bitmap);
2104                        return 0;
2105                }
2106                inode_table = ext4_inode_table(sb, gdp);
2107                if (inode_table < first_block ||
2108                    inode_table + sbi->s_itb_per_group - 1 > last_block) {
2109                        ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2110                               "Inode table for group %u not in group "
2111                               "(block %llu)!", i, inode_table);
2112                        return 0;
2113                }
2114                ext4_lock_group(sb, i);
2115                if (!ext4_group_desc_csum_verify(sb, i, gdp)) {
2116                        ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2117                                 "Checksum for group %u failed (%u!=%u)",
2118                                 i, le16_to_cpu(ext4_group_desc_csum(sbi, i,
2119                                     gdp)), le16_to_cpu(gdp->bg_checksum));
2120                        if (!(sb->s_flags & MS_RDONLY)) {
2121                                ext4_unlock_group(sb, i);
2122                                return 0;
2123                        }
2124                }
2125                ext4_unlock_group(sb, i);
2126                if (!flexbg_flag)
2127                        first_block += EXT4_BLOCKS_PER_GROUP(sb);
2128        }
2129        if (NULL != first_not_zeroed)
2130                *first_not_zeroed = grp;
2131
2132        ext4_free_blocks_count_set(sbi->s_es,
2133                                   EXT4_C2B(sbi, ext4_count_free_clusters(sb)));
2134        sbi->s_es->s_free_inodes_count =cpu_to_le32(ext4_count_free_inodes(sb));
2135        return 1;
2136}
2137
2138/* ext4_orphan_cleanup() walks a singly-linked list of inodes (starting at
2139 * the superblock) which were deleted from all directories, but held open by
2140 * a process at the time of a crash.  We walk the list and try to delete these
2141 * inodes at recovery time (only with a read-write filesystem).
2142 *
2143 * In order to keep the orphan inode chain consistent during traversal (in
2144 * case of crash during recovery), we link each inode into the superblock
2145 * orphan list_head and handle it the same way as an inode deletion during
2146 * normal operation (which journals the operations for us).
2147 *
2148 * We only do an iget() and an iput() on each inode, which is very safe if we
2149 * accidentally point at an in-use or already deleted inode.  The worst that
2150 * can happen in this case is that we get a "bit already cleared" message from
2151 * ext4_free_inode().  The only reason we would point at a wrong inode is if
2152 * e2fsck was run on this filesystem, and it must have already done the orphan
2153 * inode cleanup for us, so we can safely abort without any further action.
2154 */
2155static void ext4_orphan_cleanup(struct super_block *sb,
2156                                struct ext4_super_block *es)
2157{
2158        unsigned int s_flags = sb->s_flags;
2159        int nr_orphans = 0, nr_truncates = 0;
2160#ifdef CONFIG_QUOTA
2161        int i;
2162#endif
2163        if (!es->s_last_orphan) {
2164                jbd_debug(4, "no orphan inodes to clean up\n");
2165                return;
2166        }
2167
2168        if (bdev_read_only(sb->s_bdev)) {
2169                ext4_msg(sb, KERN_ERR, "write access "
2170                        "unavailable, skipping orphan cleanup");
2171                return;
2172        }
2173
2174        /* Check if feature set would not allow a r/w mount */
2175        if (!ext4_feature_set_ok(sb, 0)) {
2176                ext4_msg(sb, KERN_INFO, "Skipping orphan cleanup due to "
2177                         "unknown ROCOMPAT features");
2178                return;
2179        }
2180
2181        if (EXT4_SB(sb)->s_mount_state & EXT4_ERROR_FS) {
2182                /* don't clear list on RO mount w/ errors */
2183                if (es->s_last_orphan && !(s_flags & MS_RDONLY)) {
2184                        jbd_debug(1, "Errors on filesystem, "
2185                                  "clearing orphan list.\n");
2186                        es->s_last_orphan = 0;
2187                }
2188                jbd_debug(1, "Skipping orphan recovery on fs with errors.\n");
2189                return;
2190        }
2191
2192        if (s_flags & MS_RDONLY) {
2193                ext4_msg(sb, KERN_INFO, "orphan cleanup on readonly fs");
2194                sb->s_flags &= ~MS_RDONLY;
2195        }
2196#ifdef CONFIG_QUOTA
2197        /* Needed for iput() to work correctly and not trash data */
2198        sb->s_flags |= MS_ACTIVE;
2199        /* Turn on quotas so that they are updated correctly */
2200        for (i = 0; i < MAXQUOTAS; i++) {
2201                if (EXT4_SB(sb)->s_qf_names[i]) {
2202                        int ret = ext4_quota_on_mount(sb, i);
2203                        if (ret < 0)
2204                                ext4_msg(sb, KERN_ERR,
2205                                        "Cannot turn on journaled "
2206                                        "quota: error %d", ret);
2207                }
2208        }
2209#endif
2210
2211        while (es->s_last_orphan) {
2212                struct inode *inode;
2213
2214                inode = ext4_orphan_get(sb, le32_to_cpu(es->s_last_orphan));
2215                if (IS_ERR(inode)) {
2216                        es->s_last_orphan = 0;
2217                        break;
2218                }
2219
2220                list_add(&EXT4_I(inode)->i_orphan, &EXT4_SB(sb)->s_orphan);
2221                dquot_initialize(inode);
2222                if (inode->i_nlink) {
2223                        ext4_msg(sb, KERN_DEBUG,
2224                                "%s: truncating inode %lu to %lld bytes",
2225                                __func__, inode->i_ino, inode->i_size);
2226                        jbd_debug(2, "truncating inode %lu to %lld bytes\n",
2227                                  inode->i_ino, inode->i_size);
2228                        mutex_lock(&inode->i_mutex);
2229                        ext4_truncate(inode);
2230                        mutex_unlock(&inode->i_mutex);
2231                        nr_truncates++;
2232                } else {
2233                        ext4_msg(sb, KERN_DEBUG,
2234                                "%s: deleting unreferenced inode %lu",
2235                                __func__, inode->i_ino);
2236                        jbd_debug(2, "deleting unreferenced inode %lu\n",
2237                                  inode->i_ino);
2238                        nr_orphans++;
2239                }
2240                iput(inode);  /* The delete magic happens here! */
2241        }
2242
2243#define PLURAL(x) (x), ((x) == 1) ? "" : "s"
2244
2245        if (nr_orphans)
2246                ext4_msg(sb, KERN_INFO, "%d orphan inode%s deleted",
2247                       PLURAL(nr_orphans));
2248        if (nr_truncates)
2249                ext4_msg(sb, KERN_INFO, "%d truncate%s cleaned up",
2250                       PLURAL(nr_truncates));
2251#ifdef CONFIG_QUOTA
2252        /* Turn quotas off */
2253        for (i = 0; i < MAXQUOTAS; i++) {
2254                if (sb_dqopt(sb)->files[i])
2255                        dquot_quota_off(sb, i);
2256        }
2257#endif
2258        sb->s_flags = s_flags; /* Restore MS_RDONLY status */
2259}
2260
2261/*
2262 * Maximal extent format file size.
2263 * Resulting logical blkno at s_maxbytes must fit in our on-disk
2264 * extent format containers, within a sector_t, and within i_blocks
2265 * in the vfs.  ext4 inode has 48 bits of i_block in fsblock units,
2266 * so that won't be a limiting factor.
2267 *
2268 * However there is other limiting factor. We do store extents in the form
2269 * of starting block and length, hence the resulting length of the extent
2270 * covering maximum file size must fit into on-disk format containers as
2271 * well. Given that length is always by 1 unit bigger than max unit (because
2272 * we count 0 as well) we have to lower the s_maxbytes by one fs block.
2273 *
2274 * Note, this does *not* consider any metadata overhead for vfs i_blocks.
2275 */
2276static loff_t ext4_max_size(int blkbits, int has_huge_files)
2277{
2278        loff_t res;
2279        loff_t upper_limit = MAX_LFS_FILESIZE;
2280
2281        /* small i_blocks in vfs inode? */
2282        if (!has_huge_files || sizeof(blkcnt_t) < sizeof(u64)) {
2283                /*
2284                 * CONFIG_LBDAF is not enabled implies the inode
2285                 * i_block represent total blocks in 512 bytes
2286                 * 32 == size of vfs inode i_blocks * 8
2287                 */
2288                upper_limit = (1LL << 32) - 1;
2289
2290                /* total blocks in file system block size */
2291                upper_limit >>= (blkbits - 9);
2292                upper_limit <<= blkbits;
2293        }
2294
2295        /*
2296         * 32-bit extent-start container, ee_block. We lower the maxbytes
2297         * by one fs block, so ee_len can cover the extent of maximum file
2298         * size
2299         */
2300        res = (1LL << 32) - 1;
2301        res <<= blkbits;
2302
2303        /* Sanity check against vm- & vfs- imposed limits */
2304        if (res > upper_limit)
2305                res = upper_limit;
2306
2307        return res;
2308}
2309
2310/*
2311 * Maximal bitmap file size.  There is a direct, and {,double-,triple-}indirect
2312 * block limit, and also a limit of (2^48 - 1) 512-byte sectors in i_blocks.
2313 * We need to be 1 filesystem block less than the 2^48 sector limit.
2314 */
2315static loff_t ext4_max_bitmap_size(int bits, int has_huge_files)
2316{
2317        loff_t res = EXT4_NDIR_BLOCKS;
2318        int meta_blocks;
2319        loff_t upper_limit;
2320        /* This is calculated to be the largest file size for a dense, block
2321         * mapped file such that the file's total number of 512-byte sectors,
2322         * including data and all indirect blocks, does not exceed (2^48 - 1).
2323         *
2324         * __u32 i_blocks_lo and _u16 i_blocks_high represent the total
2325         * number of 512-byte sectors of the file.
2326         */
2327
2328        if (!has_huge_files || sizeof(blkcnt_t) < sizeof(u64)) {
2329                /*
2330                 * !has_huge_files or CONFIG_LBDAF not enabled implies that
2331                 * the inode i_block field represents total file blocks in
2332                 * 2^32 512-byte sectors == size of vfs inode i_blocks * 8
2333                 */
2334                upper_limit = (1LL << 32) - 1;
2335
2336                /* total blocks in file system block size */
2337                upper_limit >>= (bits - 9);
2338
2339        } else {
2340                /*
2341                 * We use 48 bit ext4_inode i_blocks
2342                 * With EXT4_HUGE_FILE_FL set the i_blocks
2343                 * represent total number of blocks in
2344                 * file system block size
2345                 */
2346                upper_limit = (1LL << 48) - 1;
2347
2348        }
2349
2350        /* indirect blocks */
2351        meta_blocks = 1;
2352        /* double indirect blocks */
2353        meta_blocks += 1 + (1LL << (bits-2));
2354        /* tripple indirect blocks */
2355        meta_blocks += 1 + (1LL << (bits-2)) + (1LL << (2*(bits-2)));
2356
2357        upper_limit -= meta_blocks;
2358        upper_limit <<= bits;
2359
2360        res += 1LL << (bits-2);
2361        res += 1LL << (2*(bits-2));
2362        res += 1LL << (3*(bits-2));
2363        res <<= bits;
2364        if (res > upper_limit)
2365                res = upper_limit;
2366
2367        if (res > MAX_LFS_FILESIZE)
2368                res = MAX_LFS_FILESIZE;
2369
2370        return res;
2371}
2372
2373static ext4_fsblk_t descriptor_loc(struct super_block *sb,
2374                                   ext4_fsblk_t logical_sb_block, int nr)
2375{
2376        struct ext4_sb_info *sbi = EXT4_SB(sb);
2377        ext4_group_t bg, first_meta_bg;
2378        int has_super = 0;
2379
2380        first_meta_bg = le32_to_cpu(sbi->s_es->s_first_meta_bg);
2381
2382        if (!EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_META_BG) ||
2383            nr < first_meta_bg)
2384                return logical_sb_block + nr + 1;
2385        bg = sbi->s_desc_per_block * nr;
2386        if (ext4_bg_has_super(sb, bg))
2387                has_super = 1;
2388
2389        return (has_super + ext4_group_first_block_no(sb, bg));
2390}
2391
2392/**
2393 * ext4_get_stripe_size: Get the stripe size.
2394 * @sbi: In memory super block info
2395 *
2396 * If we have specified it via mount option, then
2397 * use the mount option value. If the value specified at mount time is
2398 * greater than the blocks per group use the super block value.
2399 * If the super block value is greater than blocks per group return 0.
2400 * Allocator needs it be less than blocks per group.
2401 *
2402 */
2403static unsigned long ext4_get_stripe_size(struct ext4_sb_info *sbi)
2404{
2405        unsigned long stride = le16_to_cpu(sbi->s_es->s_raid_stride);
2406        unsigned long stripe_width =
2407                        le32_to_cpu(sbi->s_es->s_raid_stripe_width);
2408        int ret;
2409
2410        if (sbi->s_stripe && sbi->s_stripe <= sbi->s_blocks_per_group)
2411                ret = sbi->s_stripe;
2412        else if (stripe_width <= sbi->s_blocks_per_group)
2413                ret = stripe_width;
2414        else if (stride <= sbi->s_blocks_per_group)
2415                ret = stride;
2416        else
2417                ret = 0;
2418
2419        /*
2420         * If the stripe width is 1, this makes no sense and
2421         * we set it to 0 to turn off stripe handling code.
2422         */
2423        if (ret <= 1)
2424                ret = 0;
2425
2426        return ret;
2427}
2428
2429/* sysfs supprt */
2430
2431struct ext4_attr {
2432        struct attribute attr;
2433        ssize_t (*show)(struct ext4_attr *, struct ext4_sb_info *, char *);
2434        ssize_t (*store)(struct ext4_attr *, struct ext4_sb_info *,
2435                         const char *, size_t);
2436        int offset;
2437};
2438
2439static int parse_strtoul(const char *buf,
2440                unsigned long max, unsigned long *value)
2441{
2442        char *endp;
2443
2444        *value = simple_strtoul(skip_spaces(buf), &endp, 0);
2445        endp = skip_spaces(endp);
2446        if (*endp || *value > max)
2447                return -EINVAL;
2448
2449        return 0;
2450}
2451
2452static ssize_t delayed_allocation_blocks_show(struct ext4_attr *a,
2453                                              struct ext4_sb_info *sbi,
2454                                              char *buf)
2455{
2456        return snprintf(buf, PAGE_SIZE, "%llu\n",
2457                (s64) EXT4_C2B(sbi,
2458                        percpu_counter_sum(&sbi->s_dirtyclusters_counter)));
2459}
2460
2461static ssize_t session_write_kbytes_show(struct ext4_attr *a,
2462                                         struct ext4_sb_info *sbi, char *buf)
2463{
2464        struct super_block *sb = sbi->s_buddy_cache->i_sb;
2465
2466        if (!sb->s_bdev->bd_part)
2467                return snprintf(buf, PAGE_SIZE, "0\n");
2468        return snprintf(buf, PAGE_SIZE, "%lu\n",
2469                        (part_stat_read(sb->s_bdev->bd_part, sectors[1]) -
2470                         sbi->s_sectors_written_start) >> 1);
2471}
2472
2473static ssize_t lifetime_write_kbytes_show(struct ext4_attr *a,
2474                                          struct ext4_sb_info *sbi, char *buf)
2475{
2476        struct super_block *sb = sbi->s_buddy_cache->i_sb;
2477
2478        if (!sb->s_bdev->bd_part)
2479                return snprintf(buf, PAGE_SIZE, "0\n");
2480        return snprintf(buf, PAGE_SIZE, "%llu\n",
2481                        (unsigned long long)(sbi->s_kbytes_written +
2482                        ((part_stat_read(sb->s_bdev->bd_part, sectors[1]) -
2483                          EXT4_SB(sb)->s_sectors_written_start) >> 1)));
2484}
2485
2486static ssize_t inode_readahead_blks_store(struct ext4_attr *a,
2487                                          struct ext4_sb_info *sbi,
2488                                          const char *buf, size_t count)
2489{
2490        unsigned long t;
2491
2492        if (parse_strtoul(buf, 0x40000000, &t))
2493                return -EINVAL;
2494
2495        if (t && !is_power_of_2(t))
2496                return -EINVAL;
2497
2498        sbi->s_inode_readahead_blks = t;
2499        return count;
2500}
2501
2502static ssize_t sbi_ui_show(struct ext4_attr *a,
2503                           struct ext4_sb_info *sbi, char *buf)
2504{
2505        unsigned int *ui = (unsigned int *) (((char *) sbi) + a->offset);
2506
2507        return snprintf(buf, PAGE_SIZE, "%u\n", *ui);
2508}
2509
2510static ssize_t sbi_ui_store(struct ext4_attr *a,
2511                            struct ext4_sb_info *sbi,
2512                            const char *buf, size_t count)
2513{
2514        unsigned int *ui = (unsigned int *) (((char *) sbi) + a->offset);
2515        unsigned long t;
2516
2517        if (parse_strtoul(buf, 0xffffffff, &t))
2518                return -EINVAL;
2519        *ui = t;
2520        return count;
2521}
2522
2523static ssize_t trigger_test_error(struct ext4_attr *a,
2524                                  struct ext4_sb_info *sbi,
2525                                  const char *buf, size_t count)
2526{
2527        int len = count;
2528
2529        if (!capable(CAP_SYS_ADMIN))
2530                return -EPERM;
2531
2532        if (len && buf[len-1] == '\n')
2533                len--;
2534
2535        if (len)
2536                ext4_error(sbi->s_sb, "%.*s", len, buf);
2537        return count;
2538}
2539
2540#define EXT4_ATTR_OFFSET(_name,_mode,_show,_store,_elname) \
2541static struct ext4_attr ext4_attr_##_name = {                   \
2542        .attr = {.name = __stringify(_name), .mode = _mode },   \
2543        .show   = _show,                                        \
2544        .store  = _store,                                       \
2545        .offset = offsetof(struct ext4_sb_info, _elname),       \
2546}
2547#define EXT4_ATTR(name, mode, show, store) \
2548static struct ext4_attr ext4_attr_##name = __ATTR(name, mode, show, store)
2549
2550#define EXT4_INFO_ATTR(name) EXT4_ATTR(name, 0444, NULL, NULL)
2551#define EXT4_RO_ATTR(name) EXT4_ATTR(name, 0444, name##_show, NULL)
2552#define EXT4_RW_ATTR(name) EXT4_ATTR(name, 0644, name##_show, name##_store)
2553#define EXT4_RW_ATTR_SBI_UI(name, elname)       \
2554        EXT4_ATTR_OFFSET(name, 0644, sbi_ui_show, sbi_ui_store, elname)
2555#define ATTR_LIST(name) &ext4_attr_##name.attr
2556
2557EXT4_RO_ATTR(delayed_allocation_blocks);
2558EXT4_RO_ATTR(session_write_kbytes);
2559EXT4_RO_ATTR(lifetime_write_kbytes);
2560EXT4_ATTR_OFFSET(inode_readahead_blks, 0644, sbi_ui_show,
2561                 inode_readahead_blks_store, s_inode_readahead_blks);
2562EXT4_RW_ATTR_SBI_UI(inode_goal, s_inode_goal);
2563EXT4_RW_ATTR_SBI_UI(mb_stats, s_mb_stats);
2564EXT4_RW_ATTR_SBI_UI(mb_max_to_scan, s_mb_max_to_scan);
2565EXT4_RW_ATTR_SBI_UI(mb_min_to_scan, s_mb_min_to_scan);
2566EXT4_RW_ATTR_SBI_UI(mb_order2_req, s_mb_order2_reqs);
2567EXT4_RW_ATTR_SBI_UI(mb_stream_req, s_mb_stream_request);
2568EXT4_RW_ATTR_SBI_UI(mb_group_prealloc, s_mb_group_prealloc);
2569EXT4_RW_ATTR_SBI_UI(max_writeback_mb_bump, s_max_writeback_mb_bump);
2570EXT4_RW_ATTR_SBI_UI(extent_max_zeroout_kb, s_extent_max_zeroout_kb);
2571EXT4_ATTR(trigger_fs_error, 0200, NULL, trigger_test_error);
2572
2573static struct attribute *ext4_attrs[] = {
2574        ATTR_LIST(delayed_allocation_blocks),
2575        ATTR_LIST(session_write_kbytes),
2576        ATTR_LIST(lifetime_write_kbytes),
2577        ATTR_LIST(inode_readahead_blks),
2578        ATTR_LIST(inode_goal),
2579        ATTR_LIST(mb_stats),
2580        ATTR_LIST(mb_max_to_scan),
2581        ATTR_LIST(mb_min_to_scan),
2582        ATTR_LIST(mb_order2_req),
2583        ATTR_LIST(mb_stream_req),
2584        ATTR_LIST(mb_group_prealloc),
2585        ATTR_LIST(max_writeback_mb_bump),
2586        ATTR_LIST(extent_max_zeroout_kb),
2587        ATTR_LIST(trigger_fs_error),
2588        NULL,
2589};
2590
2591/* Features this copy of ext4 supports */
2592EXT4_INFO_ATTR(lazy_itable_init);
2593EXT4_INFO_ATTR(batched_discard);
2594EXT4_INFO_ATTR(meta_bg_resize);
2595
2596static struct attribute *ext4_feat_attrs[] = {
2597        ATTR_LIST(lazy_itable_init),
2598        ATTR_LIST(batched_discard),
2599        ATTR_LIST(meta_bg_resize),
2600        NULL,
2601};
2602
2603static ssize_t ext4_attr_show(struct kobject *kobj,
2604                              struct attribute *attr, char *buf)
2605{
2606        struct ext4_sb_info *sbi = container_of(kobj, struct ext4_sb_info,
2607                                                s_kobj);
2608        struct ext4_attr *a = container_of(attr, struct ext4_attr, attr);
2609
2610        return a->show ? a->show(a, sbi, buf) : 0;
2611}
2612
2613static ssize_t ext4_attr_store(struct kobject *kobj,
2614                               struct attribute *attr,
2615                               const char *buf, size_t len)
2616{
2617        struct ext4_sb_info *sbi = container_of(kobj, struct ext4_sb_info,
2618                                                s_kobj);
2619        struct ext4_attr *a = container_of(attr, struct ext4_attr, attr);
2620
2621        return a->store ? a->store(a, sbi, buf, len) : 0;
2622}
2623
2624static void ext4_sb_release(struct kobject *kobj)
2625{
2626        struct ext4_sb_info *sbi = container_of(kobj, struct ext4_sb_info,
2627                                                s_kobj);
2628        complete(&sbi->s_kobj_unregister);
2629}
2630
2631static const struct sysfs_ops ext4_attr_ops = {
2632        .show   = ext4_attr_show,
2633        .store  = ext4_attr_store,
2634};
2635
2636static struct kobj_type ext4_ktype = {
2637        .default_attrs  = ext4_attrs,
2638        .sysfs_ops      = &ext4_attr_ops,
2639        .release        = ext4_sb_release,
2640};
2641
2642static void ext4_feat_release(struct kobject *kobj)
2643{
2644        complete(&ext4_feat->f_kobj_unregister);
2645}
2646
2647static struct kobj_type ext4_feat_ktype = {
2648        .default_attrs  = ext4_feat_attrs,
2649        .sysfs_ops      = &ext4_attr_ops,
2650        .release        = ext4_feat_release,
2651};
2652
2653/*
2654 * Check whether this filesystem can be mounted based on
2655 * the features present and the RDONLY/RDWR mount requested.
2656 * Returns 1 if this filesystem can be mounted as requested,
2657 * 0 if it cannot be.
2658 */
2659static int ext4_feature_set_ok(struct super_block *sb, int readonly)
2660{
2661        if (EXT4_HAS_INCOMPAT_FEATURE(sb, ~EXT4_FEATURE_INCOMPAT_SUPP)) {
2662                ext4_msg(sb, KERN_ERR,
2663                        "Couldn't mount because of "
2664                        "unsupported optional features (%x)",
2665                        (le32_to_cpu(EXT4_SB(sb)->s_es->s_feature_incompat) &
2666                        ~EXT4_FEATURE_INCOMPAT_SUPP));
2667                return 0;
2668        }
2669
2670        if (readonly)
2671                return 1;
2672
2673        /* Check that feature set is OK for a read-write mount */
2674        if (EXT4_HAS_RO_COMPAT_FEATURE(sb, ~EXT4_FEATURE_RO_COMPAT_SUPP)) {
2675                ext4_msg(sb, KERN_ERR, "couldn't mount RDWR because of "
2676                         "unsupported optional features (%x)",
2677                         (le32_to_cpu(EXT4_SB(sb)->s_es->s_feature_ro_compat) &
2678                                ~EXT4_FEATURE_RO_COMPAT_SUPP));
2679                return 0;
2680        }
2681        /*
2682         * Large file size enabled file system can only be mounted
2683         * read-write on 32-bit systems if kernel is built with CONFIG_LBDAF
2684         */
2685        if (EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_HUGE_FILE)) {
2686                if (sizeof(blkcnt_t) < sizeof(u64)) {
2687                        ext4_msg(sb, KERN_ERR, "Filesystem with huge files "
2688                                 "cannot be mounted RDWR without "
2689                                 "CONFIG_LBDAF");
2690                        return 0;
2691                }
2692        }
2693        if (EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_BIGALLOC) &&
2694            !EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_EXTENTS)) {
2695                ext4_msg(sb, KERN_ERR,
2696                         "Can't support bigalloc feature without "
2697                         "extents feature\n");
2698                return 0;
2699        }
2700
2701#ifndef CONFIG_QUOTA
2702        if (EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_QUOTA) &&
2703            !readonly) {
2704                ext4_msg(sb, KERN_ERR,
2705                         "Filesystem with quota feature cannot be mounted RDWR "
2706                         "without CONFIG_QUOTA");
2707                return 0;
2708        }
2709#endif  /* CONFIG_QUOTA */
2710        return 1;
2711}
2712
2713/*
2714 * This function is called once a day if we have errors logged
2715 * on the file system
2716 */
2717static void print_daily_error_info(unsigned long arg)
2718{
2719        struct super_block *sb = (struct super_block *) arg;
2720        struct ext4_sb_info *sbi;
2721        struct ext4_super_block *es;
2722
2723        sbi = EXT4_SB(sb);
2724        es = sbi->s_es;
2725
2726        if (es->s_error_count)
2727                ext4_msg(sb, KERN_NOTICE, "error count: %u",
2728                         le32_to_cpu(es->s_error_count));
2729        if (es->s_first_error_time) {
2730                printk(KERN_NOTICE "EXT4-fs (%s): initial error at %u: %.*s:%d",
2731                       sb->s_id, le32_to_cpu(es->s_first_error_time),
2732                       (int) sizeof(es->s_first_error_func),
2733                       es->s_first_error_func,
2734                       le32_to_cpu(es->s_first_error_line));
2735                if (es->s_first_error_ino)
2736                        printk(": inode %u",
2737                               le32_to_cpu(es->s_first_error_ino));
2738                if (es->s_first_error_block)
2739                        printk(": block %llu", (unsigned long long)
2740                               le64_to_cpu(es->s_first_error_block));
2741                printk("\n");
2742        }
2743        if (es->s_last_error_time) {
2744                printk(KERN_NOTICE "EXT4-fs (%s): last error at %u: %.*s:%d",
2745                       sb->s_id, le32_to_cpu(es->s_last_error_time),
2746                       (int) sizeof(es->s_last_error_func),
2747                       es->s_last_error_func,
2748                       le32_to_cpu(es->s_last_error_line));
2749                if (es->s_last_error_ino)
2750                        printk(": inode %u",
2751                               le32_to_cpu(es->s_last_error_ino));
2752                if (es->s_last_error_block)
2753                        printk(": block %llu", (unsigned long long)
2754                               le64_to_cpu(es->s_last_error_block));
2755                printk("\n");
2756        }
2757        mod_timer(&sbi->s_err_report, jiffies + 24*60*60*HZ);  /* Once a day */
2758}
2759
2760/* Find next suitable group and run ext4_init_inode_table */
2761static int ext4_run_li_request(struct ext4_li_request *elr)
2762{
2763        struct ext4_group_desc *gdp = NULL;
2764        ext4_group_t group, ngroups;
2765        struct super_block *sb;
2766        unsigned long timeout = 0;
2767        int ret = 0;
2768
2769        sb = elr->lr_super;
2770        ngroups = EXT4_SB(sb)->s_groups_count;
2771
2772        sb_start_write(sb);
2773        for (group = elr->lr_next_group; group < ngroups; group++) {
2774                gdp = ext4_get_group_desc(sb, group, NULL);
2775                if (!gdp) {
2776                        ret = 1;
2777                        break;
2778                }
2779
2780                if (!(gdp->bg_flags & cpu_to_le16(EXT4_BG_INODE_ZEROED)))
2781                        break;
2782        }
2783
2784        if (group == ngroups)
2785                ret = 1;
2786
2787        if (!ret) {
2788                timeout = jiffies;
2789                ret = ext4_init_inode_table(sb, group,
2790                                            elr->lr_timeout ? 0 : 1);
2791                if (elr->lr_timeout == 0) {
2792                        timeout = (jiffies - timeout) *
2793                                  elr->lr_sbi->s_li_wait_mult;
2794                        elr->lr_timeout = timeout;
2795                }
2796                elr->lr_next_sched = jiffies + elr->lr_timeout;
2797                elr->lr_next_group = group + 1;
2798        }
2799        sb_end_write(sb);
2800
2801        return ret;
2802}
2803
2804/*
2805 * Remove lr_request from the list_request and free the
2806 * request structure. Should be called with li_list_mtx held
2807 */
2808static void ext4_remove_li_request(struct ext4_li_request *elr)
2809{
2810        struct ext4_sb_info *sbi;
2811
2812        if (!elr)
2813                return;
2814
2815        sbi = elr->lr_sbi;
2816
2817        list_del(&elr->lr_request);
2818        sbi->s_li_request = NULL;
2819        kfree(elr);
2820}
2821
2822static void ext4_unregister_li_request(struct super_block *sb)
2823{
2824        mutex_lock(&ext4_li_mtx);
2825        if (!ext4_li_info) {
2826                mutex_unlock(&ext4_li_mtx);
2827                return;
2828        }
2829
2830        mutex_lock(&ext4_li_info->li_list_mtx);
2831        ext4_remove_li_request(EXT4_SB(sb)->s_li_request);
2832        mutex_unlock(&ext4_li_info->li_list_mtx);
2833        mutex_unlock(&ext4_li_mtx);
2834}
2835
2836static struct task_struct *ext4_lazyinit_task;
2837
2838/*
2839 * This is the function where ext4lazyinit thread lives. It walks
2840 * through the request list searching for next scheduled filesystem.
2841 * When such a fs is found, run the lazy initialization request
2842 * (ext4_rn_li_request) and keep track of the time spend in this
2843 * function. Based on that time we compute next schedule time of
2844 * the request. When walking through the list is complete, compute
2845 * next waking time and put itself into sleep.
2846 */
2847static int ext4_lazyinit_thread(void *arg)
2848{
2849        struct ext4_lazy_init *eli = (struct ext4_lazy_init *)arg;
2850        struct list_head *pos, *n;
2851        struct ext4_li_request *elr;
2852        unsigned long next_wakeup, cur;
2853
2854        BUG_ON(NULL == eli);
2855
2856cont_thread:
2857        while (true) {
2858                next_wakeup = MAX_JIFFY_OFFSET;
2859
2860                mutex_lock(&eli->li_list_mtx);
2861                if (list_empty(&eli->li_request_list)) {
2862                        mutex_unlock(&eli->li_list_mtx);
2863                        goto exit_thread;
2864                }
2865
2866                list_for_each_safe(pos, n, &eli->li_request_list) {
2867                        elr = list_entry(pos, struct ext4_li_request,
2868                                         lr_request);
2869
2870                        if (time_after_eq(jiffies, elr->lr_next_sched)) {
2871                                if (ext4_run_li_request(elr) != 0) {
2872                                        /* error, remove the lazy_init job */
2873                                        ext4_remove_li_request(elr);
2874                                        continue;
2875                                }
2876                        }
2877
2878                        if (time_before(elr->lr_next_sched, next_wakeup))
2879                                next_wakeup = elr->lr_next_sched;
2880                }
2881                mutex_unlock(&eli->li_list_mtx);
2882
2883                try_to_freeze();
2884
2885                cur = jiffies;
2886                if ((time_after_eq(cur, next_wakeup)) ||
2887                    (MAX_JIFFY_OFFSET == next_wakeup)) {
2888                        cond_resched();
2889                        continue;
2890                }
2891
2892                schedule_timeout_interruptible(next_wakeup - cur);
2893
2894                if (kthread_should_stop()) {
2895                        ext4_clear_request_list();
2896                        goto exit_thread;
2897                }
2898        }
2899
2900exit_thread:
2901        /*
2902         * It looks like the request list is empty, but we need
2903         * to check it under the li_list_mtx lock, to prevent any
2904         * additions into it, and of course we should lock ext4_li_mtx
2905         * to atomically free the list and ext4_li_info, because at
2906         * this point another ext4 filesystem could be registering
2907         * new one.
2908         */
2909        mutex_lock(&ext4_li_mtx);
2910        mutex_lock(&eli->li_list_mtx);
2911        if (!list_empty(&eli->li_request_list)) {
2912                mutex_unlock(&eli->li_list_mtx);
2913                mutex_unlock(&ext4_li_mtx);
2914                goto cont_thread;
2915        }
2916        mutex_unlock(&eli->li_list_mtx);
2917        kfree(ext4_li_info);
2918        ext4_li_info = NULL;
2919        mutex_unlock(&ext4_li_mtx);
2920
2921        return 0;
2922}
2923
2924static void ext4_clear_request_list(void)
2925{
2926        struct list_head *pos, *n;
2927        struct ext4_li_request *elr;
2928
2929        mutex_lock(&ext4_li_info->li_list_mtx);
2930        list_for_each_safe(pos, n, &ext4_li_info->li_request_list) {
2931                elr = list_entry(pos, struct ext4_li_request,
2932                                 lr_request);
2933                ext4_remove_li_request(elr);
2934        }
2935        mutex_unlock(&ext4_li_info->li_list_mtx);
2936}
2937
2938static int ext4_run_lazyinit_thread(void)
2939{
2940        ext4_lazyinit_task = kthread_run(ext4_lazyinit_thread,
2941                                         ext4_li_info, "ext4lazyinit");
2942        if (IS_ERR(ext4_lazyinit_task)) {
2943                int err = PTR_ERR(ext4_lazyinit_task);
2944                ext4_clear_request_list();
2945                kfree(ext4_li_info);
2946                ext4_li_info = NULL;
2947                printk(KERN_CRIT "EXT4-fs: error %d creating inode table "
2948                                 "initialization thread\n",
2949                                 err);
2950                return err;
2951        }
2952        ext4_li_info->li_state |= EXT4_LAZYINIT_RUNNING;
2953        return 0;
2954}
2955
2956/*
2957 * Check whether it make sense to run itable init. thread or not.
2958 * If there is at least one uninitialized inode table, return
2959 * corresponding group number, else the loop goes through all
2960 * groups and return total number of groups.
2961 */
2962static ext4_group_t ext4_has_uninit_itable(struct super_block *sb)
2963{
2964        ext4_group_t group, ngroups = EXT4_SB(sb)->s_groups_count;
2965        struct ext4_group_desc *gdp = NULL;
2966
2967        for (group = 0; group < ngroups; group++) {
2968                gdp = ext4_get_group_desc(sb, group, NULL);
2969                if (!gdp)
2970                        continue;
2971
2972                if (!(gdp->bg_flags & cpu_to_le16(EXT4_BG_INODE_ZEROED)))
2973                        break;
2974        }
2975
2976        return group;
2977}
2978
2979static int ext4_li_info_new(void)
2980{
2981        struct ext4_lazy_init *eli = NULL;
2982
2983        eli = kzalloc(sizeof(*eli), GFP_KERNEL);
2984        if (!eli)
2985                return -ENOMEM;
2986
2987        INIT_LIST_HEAD(&eli->li_request_list);
2988        mutex_init(&eli->li_list_mtx);
2989
2990        eli->li_state |= EXT4_LAZYINIT_QUIT;
2991
2992        ext4_li_info = eli;
2993
2994        return 0;
2995}
2996
2997static struct ext4_li_request *ext4_li_request_new(struct super_block *sb,
2998                                            ext4_group_t start)
2999{
3000        struct ext4_sb_info *sbi = EXT4_SB(sb);
3001        struct ext4_li_request *elr;
3002        unsigned long rnd;
3003
3004        elr = kzalloc(sizeof(*elr), GFP_KERNEL);
3005        if (!elr)
3006                return NULL;
3007
3008        elr->lr_super = sb;
3009        elr->lr_sbi = sbi;
3010        elr->lr_next_group = start;
3011
3012        /*
3013         * Randomize first schedule time of the request to
3014         * spread the inode table initialization requests
3015         * better.
3016         */
3017        get_random_bytes(&rnd, sizeof(rnd));
3018        elr->lr_next_sched = jiffies + (unsigned long)rnd %
3019                             (EXT4_DEF_LI_MAX_START_DELAY * HZ);
3020
3021        return elr;
3022}
3023
3024static int ext4_register_li_request(struct super_block *sb,
3025                                    ext4_group_t first_not_zeroed)
3026{
3027        struct ext4_sb_info *sbi = EXT4_SB(sb);
3028        struct ext4_li_request *elr;
3029        ext4_group_t ngroups = EXT4_SB(sb)->s_groups_count;
3030        int ret = 0;
3031
3032        if (sbi->s_li_request != NULL) {
3033                /*
3034                 * Reset timeout so it can be computed again, because
3035                 * s_li_wait_mult might have changed.
3036                 */
3037                sbi->s_li_request->lr_timeout = 0;
3038                return 0;
3039        }
3040
3041        if (first_not_zeroed == ngroups ||
3042            (sb->s_flags & MS_RDONLY) ||
3043            !test_opt(sb, INIT_INODE_TABLE))
3044                return 0;
3045
3046        elr = ext4_li_request_new(sb, first_not_zeroed);
3047        if (!elr)
3048                return -ENOMEM;
3049
3050        mutex_lock(&ext4_li_mtx);
3051
3052        if (NULL == ext4_li_info) {
3053                ret = ext4_li_info_new();
3054                if (ret)
3055                        goto out;
3056        }
3057
3058        mutex_lock(&ext4_li_info->li_list_mtx);
3059        list_add(&elr->lr_request, &ext4_li_info->li_request_list);
3060        mutex_unlock(&ext4_li_info->li_list_mtx);
3061
3062        sbi->s_li_request = elr;
3063        /*
3064         * set elr to NULL here since it has been inserted to
3065         * the request_list and the removal and free of it is
3066         * handled by ext4_clear_request_list from now on.
3067         */
3068        elr = NULL;
3069
3070        if (!(ext4_li_info->li_state & EXT4_LAZYINIT_RUNNING)) {
3071                ret = ext4_run_lazyinit_thread();
3072                if (ret)
3073                        goto out;
3074        }
3075out:
3076        mutex_unlock(&ext4_li_mtx);
3077        if (ret)
3078                kfree(elr);
3079        return ret;
3080}
3081
3082/*
3083 * We do not need to lock anything since this is called on
3084 * module unload.
3085 */
3086static void ext4_destroy_lazyinit_thread(void)
3087{
3088        /*
3089         * If thread exited earlier
3090         * there's nothing to be done.
3091         */
3092        if (!ext4_li_info || !ext4_lazyinit_task)
3093                return;
3094
3095        kthread_stop(ext4_lazyinit_task);
3096}
3097
3098static int set_journal_csum_feature_set(struct super_block *sb)
3099{
3100        int ret = 1;
3101        int compat, incompat;
3102        struct ext4_sb_info *sbi = EXT4_SB(sb);
3103
3104        if (EXT4_HAS_RO_COMPAT_FEATURE(sb,
3105                                       EXT4_FEATURE_RO_COMPAT_METADATA_CSUM)) {
3106                /* journal checksum v2 */
3107                compat = 0;
3108                incompat = JBD2_FEATURE_INCOMPAT_CSUM_V2;
3109        } else {
3110                /* journal checksum v1 */
3111                compat = JBD2_FEATURE_COMPAT_CHECKSUM;
3112                incompat = 0;
3113        }
3114
3115        if (test_opt(sb, JOURNAL_ASYNC_COMMIT)) {
3116                ret = jbd2_journal_set_features(sbi->s_journal,
3117                                compat, 0,
3118                                JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT |
3119                                incompat);
3120        } else if (test_opt(sb, JOURNAL_CHECKSUM)) {
3121                ret = jbd2_journal_set_features(sbi->s_journal,
3122                                compat, 0,
3123                                incompat);
3124                jbd2_journal_clear_features(sbi->s_journal, 0, 0,
3125                                JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT);
3126        } else {
3127                jbd2_journal_clear_features(sbi->s_journal,
3128                                JBD2_FEATURE_COMPAT_CHECKSUM, 0,
3129                                JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT |
3130                                JBD2_FEATURE_INCOMPAT_CSUM_V2);
3131        }
3132
3133        return ret;
3134}
3135
3136/*
3137 * Note: calculating the overhead so we can be compatible with
3138 * historical BSD practice is quite difficult in the face of
3139 * clusters/bigalloc.  This is because multiple metadata blocks from
3140 * different block group can end up in the same allocation cluster.
3141 * Calculating the exact overhead in the face of clustered allocation
3142 * requires either O(all block bitmaps) in memory or O(number of block
3143 * groups**2) in time.  We will still calculate the superblock for
3144 * older file systems --- and if we come across with a bigalloc file
3145 * system with zero in s_overhead_clusters the estimate will be close to
3146 * correct especially for very large cluster sizes --- but for newer
3147 * file systems, it's better to calculate this figure once at mkfs
3148 * time, and store it in the superblock.  If the superblock value is
3149 * present (even for non-bigalloc file systems), we will use it.
3150 */
3151static int count_overhead(struct super_block *sb, ext4_group_t grp,
3152                          char *buf)
3153{
3154        struct ext4_sb_info     *sbi = EXT4_SB(sb);
3155        struct ext4_group_desc  *gdp;
3156        ext4_fsblk_t            first_block, last_block, b;
3157        ext4_group_t            i, ngroups = ext4_get_groups_count(sb);
3158        int                     s, j, count = 0;
3159
3160        if (!EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_BIGALLOC))
3161                return (ext4_bg_has_super(sb, grp) + ext4_bg_num_gdb(sb, grp) +
3162                        sbi->s_itb_per_group + 2);
3163
3164        first_block = le32_to_cpu(sbi->s_es->s_first_data_block) +
3165                (grp * EXT4_BLOCKS_PER_GROUP(sb));
3166        last_block = first_block + EXT4_BLOCKS_PER_GROUP(sb) - 1;
3167        for (i = 0; i < ngroups; i++) {
3168                gdp = ext4_get_group_desc(sb, i, NULL);
3169                b = ext4_block_bitmap(sb, gdp);
3170                if (b >= first_block && b <= last_block) {
3171                        ext4_set_bit(EXT4_B2C(sbi, b - first_block), buf);
3172                        count++;
3173                }
3174                b = ext4_inode_bitmap(sb, gdp);
3175                if (b >= first_block && b <= last_block) {
3176                        ext4_set_bit(EXT4_B2C(sbi, b - first_block), buf);
3177                        count++;
3178                }
3179                b = ext4_inode_table(sb, gdp);
3180                if (b >= first_block && b + sbi->s_itb_per_group <= last_block)
3181                        for (j = 0; j < sbi->s_itb_per_group; j++, b++) {
3182                                int c = EXT4_B2C(sbi, b - first_block);
3183                                ext4_set_bit(c, buf);
3184                                count++;
3185                        }
3186                if (i != grp)
3187                        continue;
3188                s = 0;
3189                if (ext4_bg_has_super(sb, grp)) {
3190                        ext4_set_bit(s++, buf);
3191                        count++;
3192                }
3193                for (j = ext4_bg_num_gdb(sb, grp); j > 0; j--) {
3194                        ext4_set_bit(EXT4_B2C(sbi, s++), buf);
3195                        count++;
3196                }
3197        }
3198        if (!count)
3199                return 0;
3200        return EXT4_CLUSTERS_PER_GROUP(sb) -
3201                ext4_count_free(buf, EXT4_CLUSTERS_PER_GROUP(sb) / 8);
3202}
3203
3204/*
3205 * Compute the overhead and stash it in sbi->s_overhead
3206 */
3207int ext4_calculate_overhead(struct super_block *sb)
3208{
3209        struct ext4_sb_info *sbi = EXT4_SB(sb);
3210        struct ext4_super_block *es = sbi->s_es;
3211        ext4_group_t i, ngroups = ext4_get_groups_count(sb);
3212        ext4_fsblk_t overhead = 0;
3213        char *buf = (char *) get_zeroed_page(GFP_KERNEL);
3214
3215        memset(buf, 0, PAGE_SIZE);
3216        if (!buf)
3217                return -ENOMEM;
3218
3219        /*
3220         * Compute the overhead (FS structures).  This is constant
3221         * for a given filesystem unless the number of block groups
3222         * changes so we cache the previous value until it does.
3223         */
3224
3225        /*
3226         * All of the blocks before first_data_block are overhead
3227         */
3228        overhead = EXT4_B2C(sbi, le32_to_cpu(es->s_first_data_block));
3229
3230        /*
3231         * Add the overhead found in each block group
3232         */
3233        for (i = 0; i < ngroups; i++) {
3234                int blks;
3235
3236                blks = count_overhead(sb, i, buf);
3237                overhead += blks;
3238                if (blks)
3239                        memset(buf, 0, PAGE_SIZE);
3240                cond_resched();
3241        }
3242        sbi->s_overhead = overhead;
3243        smp_wmb();
3244        free_page((unsigned long) buf);
3245        return 0;
3246}
3247
3248static int ext4_fill_super(struct super_block *sb, void *data, int silent)
3249{
3250        char *orig_data = kstrdup(data, GFP_KERNEL);
3251        struct buffer_head *bh;
3252        struct ext4_super_block *es = NULL;
3253        struct ext4_sb_info *sbi;
3254        ext4_fsblk_t block;
3255        ext4_fsblk_t sb_block = get_sb_block(&data);
3256        ext4_fsblk_t logical_sb_block;
3257        unsigned long offset = 0;
3258        unsigned long journal_devnum = 0;
3259        unsigned long def_mount_opts;
3260        struct inode *root;
3261        char *cp;
3262        const char *descr;
3263        int ret = -ENOMEM;
3264        int blocksize, clustersize;
3265        unsigned int db_count;
3266        unsigned int i;
3267        int needs_recovery, has_huge_files, has_bigalloc;
3268        __u64 blocks_count;
3269        int err;
3270        unsigned int journal_ioprio = DEFAULT_JOURNAL_IOPRIO;
3271        ext4_group_t first_not_zeroed;
3272
3273        sbi = kzalloc(sizeof(*sbi), GFP_KERNEL);
3274        if (!sbi)
3275                goto out_free_orig;
3276
3277        sbi->s_blockgroup_lock =
3278                kzalloc(sizeof(struct blockgroup_lock), GFP_KERNEL);
3279        if (!sbi->s_blockgroup_lock) {
3280                kfree(sbi);
3281                goto out_free_orig;
3282        }
3283        sb->s_fs_info = sbi;
3284        sbi->s_sb = sb;
3285        sbi->s_mount_opt = 0;
3286        sbi->s_resuid = make_kuid(&init_user_ns, EXT4_DEF_RESUID);
3287        sbi->s_resgid = make_kgid(&init_user_ns, EXT4_DEF_RESGID);
3288        sbi->s_inode_readahead_blks = EXT4_DEF_INODE_READAHEAD_BLKS;
3289        sbi->s_sb_block = sb_block;
3290        if (sb->s_bdev->bd_part)
3291                sbi->s_sectors_written_start =
3292                        part_stat_read(sb->s_bdev->bd_part, sectors[1]);
3293
3294        /* Cleanup superblock name */
3295        for (cp = sb->s_id; (cp = strchr(cp, '/'));)
3296                *cp = '!';
3297
3298        ret = -EINVAL;
3299        blocksize = sb_min_blocksize(sb, EXT4_MIN_BLOCK_SIZE);
3300        if (!blocksize) {
3301                ext4_msg(sb, KERN_ERR, "unable to set blocksize");
3302                goto out_fail;
3303        }
3304
3305        /*
3306         * The ext4 superblock will not be buffer aligned for other than 1kB
3307         * block sizes.  We need to calculate the offset from buffer start.
3308         */
3309        if (blocksize != EXT4_MIN_BLOCK_SIZE) {
3310                logical_sb_block = sb_block * EXT4_MIN_BLOCK_SIZE;
3311                offset = do_div(logical_sb_block, blocksize);
3312        } else {
3313                logical_sb_block = sb_block;
3314        }
3315
3316        if (!(bh = sb_bread(sb, logical_sb_block))) {
3317                ext4_msg(sb, KERN_ERR, "unable to read superblock");
3318                goto out_fail;
3319        }
3320        /*
3321         * Note: s_es must be initialized as soon as possible because
3322         *       some ext4 macro-instructions depend on its value
3323         */
3324        es = (struct ext4_super_block *) (bh->b_data + offset);
3325        sbi->s_es = es;
3326        sb->s_magic = le16_to_cpu(es->s_magic);
3327        if (sb->s_magic != EXT4_SUPER_MAGIC)
3328                goto cantfind_ext4;
3329        sbi->s_kbytes_written = le64_to_cpu(es->s_kbytes_written);
3330
3331        /* Warn if metadata_csum and gdt_csum are both set. */
3332        if (EXT4_HAS_RO_COMPAT_FEATURE(sb,
3333                                       EXT4_FEATURE_RO_COMPAT_METADATA_CSUM) &&
3334            EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_GDT_CSUM))
3335                ext4_warning(sb, KERN_INFO "metadata_csum and uninit_bg are "
3336                             "redundant flags; please run fsck.");
3337
3338        /* Check for a known checksum algorithm */
3339        if (!ext4_verify_csum_type(sb, es)) {
3340                ext4_msg(sb, KERN_ERR, "VFS: Found ext4 filesystem with "
3341                         "unknown checksum algorithm.");
3342                silent = 1;
3343                goto cantfind_ext4;
3344        }
3345
3346        /* Load the checksum driver */
3347        if (EXT4_HAS_RO_COMPAT_FEATURE(sb,
3348                                       EXT4_FEATURE_RO_COMPAT_METADATA_CSUM)) {
3349                sbi->s_chksum_driver = crypto_alloc_shash("crc32c", 0, 0);
3350                if (IS_ERR(sbi->s_chksum_driver)) {
3351                        ext4_msg(sb, KERN_ERR, "Cannot load crc32c driver.");
3352                        ret = PTR_ERR(sbi->s_chksum_driver);
3353                        sbi->s_chksum_driver = NULL;
3354                        goto failed_mount;
3355                }
3356        }
3357
3358        /* Check superblock checksum */
3359        if (!ext4_superblock_csum_verify(sb, es)) {
3360                ext4_msg(sb, KERN_ERR, "VFS: Found ext4 filesystem with "
3361                         "invalid superblock checksum.  Run e2fsck?");
3362                silent = 1;
3363                goto cantfind_ext4;
3364        }
3365
3366        /* Precompute checksum seed for all metadata */
3367        if (EXT4_HAS_RO_COMPAT_FEATURE(sb,
3368                        EXT4_FEATURE_RO_COMPAT_METADATA_CSUM))
3369                sbi->s_csum_seed = ext4_chksum(sbi, ~0, es->s_uuid,
3370                                               sizeof(es->s_uuid));
3371
3372        /* Set defaults before we parse the mount options */
3373        def_mount_opts = le32_to_cpu(es->s_default_mount_opts);
3374        set_opt(sb, INIT_INODE_TABLE);
3375        if (def_mount_opts & EXT4_DEFM_DEBUG)
3376                set_opt(sb, DEBUG);
3377        if (def_mount_opts & EXT4_DEFM_BSDGROUPS)
3378                set_opt(sb, GRPID);
3379        if (def_mount_opts & EXT4_DEFM_UID16)
3380                set_opt(sb, NO_UID32);
3381        /* xattr user namespace & acls are now defaulted on */
3382#ifdef CONFIG_EXT4_FS_XATTR
3383        set_opt(sb, XATTR_USER);
3384#endif
3385#ifdef CONFIG_EXT4_FS_POSIX_ACL
3386        set_opt(sb, POSIX_ACL);
3387#endif
3388        set_opt(sb, MBLK_IO_SUBMIT);
3389        if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_DATA)
3390                set_opt(sb, JOURNAL_DATA);
3391        else if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_ORDERED)
3392                set_opt(sb, ORDERED_DATA);
3393        else if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_WBACK)
3394                set_opt(sb, WRITEBACK_DATA);
3395
3396        if (le16_to_cpu(sbi->s_es->s_errors) == EXT4_ERRORS_PANIC)
3397                set_opt(sb, ERRORS_PANIC);
3398        else if (le16_to_cpu(sbi->s_es->s_errors) == EXT4_ERRORS_CONTINUE)
3399                set_opt(sb, ERRORS_CONT);
3400        else
3401                set_opt(sb, ERRORS_RO);
3402        if (def_mount_opts & EXT4_DEFM_BLOCK_VALIDITY)
3403                set_opt(sb, BLOCK_VALIDITY);
3404        if (def_mount_opts & EXT4_DEFM_DISCARD)
3405                set_opt(sb, DISCARD);
3406
3407        sbi->s_resuid = make_kuid(&init_user_ns, le16_to_cpu(es->s_def_resuid));
3408        sbi->s_resgid = make_kgid(&init_user_ns, le16_to_cpu(es->s_def_resgid));
3409        sbi->s_commit_interval = JBD2_DEFAULT_MAX_COMMIT_AGE * HZ;
3410        sbi->s_min_batch_time = EXT4_DEF_MIN_BATCH_TIME;
3411        sbi->s_max_batch_time = EXT4_DEF_MAX_BATCH_TIME;
3412
3413        if ((def_mount_opts & EXT4_DEFM_NOBARRIER) == 0)
3414                set_opt(sb, BARRIER);
3415
3416        /*
3417         * enable delayed allocation by default
3418         * Use -o nodelalloc to turn it off
3419         */
3420        if (!IS_EXT3_SB(sb) && !IS_EXT2_SB(sb) &&
3421            ((def_mount_opts & EXT4_DEFM_NODELALLOC) == 0))
3422                set_opt(sb, DELALLOC);
3423
3424        /*
3425         * set default s_li_wait_mult for lazyinit, for the case there is
3426         * no mount option specified.
3427         */
3428        sbi->s_li_wait_mult = EXT4_DEF_LI_WAIT_MULT;
3429
3430        if (!parse_options((char *) sbi->s_es->s_mount_opts, sb,
3431                           &journal_devnum, &journal_ioprio, 0)) {
3432                ext4_msg(sb, KERN_WARNING,
3433                         "failed to parse options in superblock: %s",
3434                         sbi->s_es->s_mount_opts);
3435        }
3436        sbi->s_def_mount_opt = sbi->s_mount_opt;
3437        if (!parse_options((char *) data, sb, &journal_devnum,
3438                           &journal_ioprio, 0))
3439                goto failed_mount;
3440
3441        if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA) {
3442                printk_once(KERN_WARNING "EXT4-fs: Warning: mounting "
3443                            "with data=journal disables delayed "
3444                            "allocation and O_DIRECT support!\n");
3445                if (test_opt2(sb, EXPLICIT_DELALLOC)) {
3446                        ext4_msg(sb, KERN_ERR, "can't mount with "
3447                                 "both data=journal and delalloc");
3448                        goto failed_mount;
3449                }
3450                if (test_opt(sb, DIOREAD_NOLOCK)) {
3451                        ext4_msg(sb, KERN_ERR, "can't mount with "
3452                                 "both data=journal and delalloc");
3453                        goto failed_mount;
3454                }
3455                if (test_opt(sb, DELALLOC))
3456                        clear_opt(sb, DELALLOC);
3457        }
3458
3459        sb->s_flags = (sb->s_flags & ~MS_POSIXACL) |
3460                (test_opt(sb, POSIX_ACL) ? MS_POSIXACL : 0);
3461
3462        if (le32_to_cpu(es->s_rev_level) == EXT4_GOOD_OLD_REV &&
3463            (EXT4_HAS_COMPAT_FEATURE(sb, ~0U) ||
3464             EXT4_HAS_RO_COMPAT_FEATURE(sb, ~0U) ||
3465             EXT4_HAS_INCOMPAT_FEATURE(sb, ~0U)))
3466                ext4_msg(sb, KERN_WARNING,
3467                       "feature flags set on rev 0 fs, "
3468                       "running e2fsck is recommended");
3469
3470        if (IS_EXT2_SB(sb)) {
3471                if (ext2_feature_set_ok(sb))
3472                        ext4_msg(sb, KERN_INFO, "mounting ext2 file system "
3473                                 "using the ext4 subsystem");
3474                else {
3475                        ext4_msg(sb, KERN_ERR, "couldn't mount as ext2 due "
3476                                 "to feature incompatibilities");
3477                        goto failed_mount;
3478                }
3479        }
3480
3481        if (IS_EXT3_SB(sb)) {
3482                if (ext3_feature_set_ok(sb))
3483                        ext4_msg(sb, KERN_INFO, "mounting ext3 file system "
3484                                 "using the ext4 subsystem");
3485                else {
3486                        ext4_msg(sb, KERN_ERR, "couldn't mount as ext3 due "
3487                                 "to feature incompatibilities");
3488                        goto failed_mount;
3489                }
3490        }
3491
3492        /*
3493         * Check feature flags regardless of the revision level, since we
3494         * previously didn't change the revision level when setting the flags,
3495         * so there is a chance incompat flags are set on a rev 0 filesystem.
3496         */
3497        if (!ext4_feature_set_ok(sb, (sb->s_flags & MS_RDONLY)))
3498                goto failed_mount;
3499
3500        blocksize = BLOCK_SIZE << le32_to_cpu(es->s_log_block_size);
3501        if (blocksize < EXT4_MIN_BLOCK_SIZE ||
3502            blocksize > EXT4_MAX_BLOCK_SIZE) {
3503                ext4_msg(sb, KERN_ERR,
3504                       "Unsupported filesystem blocksize %d", blocksize);
3505                goto failed_mount;
3506        }
3507
3508        if (sb->s_blocksize != blocksize) {
3509                /* Validate the filesystem blocksize */
3510                if (!sb_set_blocksize(sb, blocksize)) {
3511                        ext4_msg(sb, KERN_ERR, "bad block size %d",
3512                                        blocksize);
3513                        goto failed_mount;
3514                }
3515
3516                brelse(bh);
3517                logical_sb_block = sb_block * EXT4_MIN_BLOCK_SIZE;
3518                offset = do_div(logical_sb_block, blocksize);
3519                bh = sb_bread(sb, logical_sb_block);
3520                if (!bh) {
3521                        ext4_msg(sb, KERN_ERR,
3522                               "Can't read superblock on 2nd try");
3523                        goto failed_mount;
3524                }
3525                es = (struct ext4_super_block *)(bh->b_data + offset);
3526                sbi->s_es = es;
3527                if (es->s_magic != cpu_to_le16(EXT4_SUPER_MAGIC)) {
3528                        ext4_msg(sb, KERN_ERR,
3529                               "Magic mismatch, very weird!");
3530                        goto failed_mount;
3531                }
3532        }
3533
3534        has_huge_files = EXT4_HAS_RO_COMPAT_FEATURE(sb,
3535                                EXT4_FEATURE_RO_COMPAT_HUGE_FILE);
3536        sbi->s_bitmap_maxbytes = ext4_max_bitmap_size(sb->s_blocksize_bits,
3537                                                      has_huge_files);
3538        sb->s_maxbytes = ext4_max_size(sb->s_blocksize_bits, has_huge_files);
3539
3540        if (le32_to_cpu(es->s_rev_level) == EXT4_GOOD_OLD_REV) {
3541                sbi->s_inode_size = EXT4_GOOD_OLD_INODE_SIZE;
3542                sbi->s_first_ino = EXT4_GOOD_OLD_FIRST_INO;
3543        } else {
3544                sbi->s_inode_size = le16_to_cpu(es->s_inode_size);
3545                sbi->s_first_ino = le32_to_cpu(es->s_first_ino);
3546                if ((sbi->s_inode_size < EXT4_GOOD_OLD_INODE_SIZE) ||
3547                    (!is_power_of_2(sbi->s_inode_size)) ||
3548                    (sbi->s_inode_size > blocksize)) {
3549                        ext4_msg(sb, KERN_ERR,
3550                               "unsupported inode size: %d",
3551                               sbi->s_inode_size);
3552                        goto failed_mount;
3553                }
3554                if (sbi->s_inode_size > EXT4_GOOD_OLD_INODE_SIZE)
3555                        sb->s_time_gran = 1 << (EXT4_EPOCH_BITS - 2);
3556        }
3557
3558        sbi->s_desc_size = le16_to_cpu(es->s_desc_size);
3559        if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_64BIT)) {
3560                if (sbi->s_desc_size < EXT4_MIN_DESC_SIZE_64BIT ||
3561                    sbi->s_desc_size > EXT4_MAX_DESC_SIZE ||
3562                    !is_power_of_2(sbi->s_desc_size)) {
3563                        ext4_msg(sb, KERN_ERR,
3564                               "unsupported descriptor size %lu",
3565                               sbi->s_desc_size);
3566                        goto failed_mount;
3567                }
3568        } else
3569                sbi->s_desc_size = EXT4_MIN_DESC_SIZE;
3570
3571        sbi->s_blocks_per_group = le32_to_cpu(es->s_blocks_per_group);
3572        sbi->s_inodes_per_group = le32_to_cpu(es->s_inodes_per_group);
3573        if (EXT4_INODE_SIZE(sb) == 0 || EXT4_INODES_PER_GROUP(sb) == 0)
3574                goto cantfind_ext4;
3575
3576        sbi->s_inodes_per_block = blocksize / EXT4_INODE_SIZE(sb);
3577        if (sbi->s_inodes_per_block == 0)
3578                goto cantfind_ext4;
3579        sbi->s_itb_per_group = sbi->s_inodes_per_group /
3580                                        sbi->s_inodes_per_block;
3581        sbi->s_desc_per_block = blocksize / EXT4_DESC_SIZE(sb);
3582        sbi->s_sbh = bh;
3583        sbi->s_mount_state = le16_to_cpu(es->s_state);
3584        sbi->s_addr_per_block_bits = ilog2(EXT4_ADDR_PER_BLOCK(sb));
3585        sbi->s_desc_per_block_bits = ilog2(EXT4_DESC_PER_BLOCK(sb));
3586
3587        for (i = 0; i < 4; i++)
3588                sbi->s_hash_seed[i] = le32_to_cpu(es->s_hash_seed[i]);
3589        sbi->s_def_hash_version = es->s_def_hash_version;
3590        i = le32_to_cpu(es->s_flags);
3591        if (i & EXT2_FLAGS_UNSIGNED_HASH)
3592                sbi->s_hash_unsigned = 3;
3593        else if ((i & EXT2_FLAGS_SIGNED_HASH) == 0) {
3594#ifdef __CHAR_UNSIGNED__
3595                es->s_flags |= cpu_to_le32(EXT2_FLAGS_UNSIGNED_HASH);
3596                sbi->s_hash_unsigned = 3;
3597#else
3598                es->s_flags |= cpu_to_le32(EXT2_FLAGS_SIGNED_HASH);
3599#endif
3600        }
3601
3602        /* Handle clustersize */
3603        clustersize = BLOCK_SIZE << le32_to_cpu(es->s_log_cluster_size);
3604        has_bigalloc = EXT4_HAS_RO_COMPAT_FEATURE(sb,
3605                                EXT4_FEATURE_RO_COMPAT_BIGALLOC);
3606        if (has_bigalloc) {
3607                if (clustersize < blocksize) {
3608                        ext4_msg(sb, KERN_ERR,
3609                                 "cluster size (%d) smaller than "
3610                                 "block size (%d)", clustersize, blocksize);
3611                        goto failed_mount;
3612                }
3613                sbi->s_cluster_bits = le32_to_cpu(es->s_log_cluster_size) -
3614                        le32_to_cpu(es->s_log_block_size);
3615                sbi->s_clusters_per_group =
3616                        le32_to_cpu(es->s_clusters_per_group);
3617                if (sbi->s_clusters_per_group > blocksize * 8) {
3618                        ext4_msg(sb, KERN_ERR,
3619                                 "#clusters per group too big: %lu",
3620                                 sbi->s_clusters_per_group);
3621                        goto failed_mount;
3622                }
3623                if (sbi->s_blocks_per_group !=
3624                    (sbi->s_clusters_per_group * (clustersize / blocksize))) {
3625                        ext4_msg(sb, KERN_ERR, "blocks per group (%lu) and "
3626                                 "clusters per group (%lu) inconsistent",
3627                                 sbi->s_blocks_per_group,
3628                                 sbi->s_clusters_per_group);
3629                        goto failed_mount;
3630                }
3631        } else {
3632                if (clustersize != blocksize) {
3633                        ext4_warning(sb, "fragment/cluster size (%d) != "
3634                                     "block size (%d)", clustersize,
3635                                     blocksize);
3636                        clustersize = blocksize;
3637                }
3638                if (sbi->s_blocks_per_group > blocksize * 8) {
3639                        ext4_msg(sb, KERN_ERR,
3640                                 "#blocks per group too big: %lu",
3641                                 sbi->s_blocks_per_group);
3642                        goto failed_mount;
3643                }
3644                sbi->s_clusters_per_group = sbi->s_blocks_per_group;
3645                sbi->s_cluster_bits = 0;
3646        }
3647        sbi->s_cluster_ratio = clustersize / blocksize;
3648
3649        if (sbi->s_inodes_per_group > blocksize * 8) {
3650                ext4_msg(sb, KERN_ERR,
3651                       "#inodes per group too big: %lu",
3652                       sbi->s_inodes_per_group);
3653                goto failed_mount;
3654        }
3655
3656        /*
3657         * Test whether we have more sectors than will fit in sector_t,
3658         * and whether the max offset is addressable by the page cache.
3659         */
3660        err = generic_check_addressable(sb->s_blocksize_bits,
3661                                        ext4_blocks_count(es));
3662        if (err) {
3663                ext4_msg(sb, KERN_ERR, "filesystem"
3664                         " too large to mount safely on this system");
3665                if (sizeof(sector_t) < 8)
3666                        ext4_msg(sb, KERN_WARNING, "CONFIG_LBDAF not enabled");
3667                ret = err;
3668                goto failed_mount;
3669        }
3670
3671        if (EXT4_BLOCKS_PER_GROUP(sb) == 0)
3672                goto cantfind_ext4;
3673
3674        /* check blocks count against device size */
3675        blocks_count = sb->s_bdev->bd_inode->i_size >> sb->s_blocksize_bits;
3676        if (blocks_count && ext4_blocks_count(es) > blocks_count) {
3677                ext4_msg(sb, KERN_WARNING, "bad geometry: block count %llu "
3678                       "exceeds size of device (%llu blocks)",
3679                       ext4_blocks_count(es), blocks_count);
3680                goto failed_mount;
3681        }
3682
3683        /*
3684         * It makes no sense for the first data block to be beyond the end
3685         * of the filesystem.
3686         */
3687        if (le32_to_cpu(es->s_first_data_block) >= ext4_blocks_count(es)) {
3688                ext4_msg(sb, KERN_WARNING, "bad geometry: first data "
3689                         "block %u is beyond end of filesystem (%llu)",
3690                         le32_to_cpu(es->s_first_data_block),
3691                         ext4_blocks_count(es));
3692                goto failed_mount;
3693        }
3694        blocks_count = (ext4_blocks_count(es) -
3695                        le32_to_cpu(es->s_first_data_block) +
3696                        EXT4_BLOCKS_PER_GROUP(sb) - 1);
3697        do_div(blocks_count, EXT4_BLOCKS_PER_GROUP(sb));
3698        if (blocks_count > ((uint64_t)1<<32) - EXT4_DESC_PER_BLOCK(sb)) {
3699                ext4_msg(sb, KERN_WARNING, "groups count too large: %u "
3700                       "(block count %llu, first data block %u, "
3701                       "blocks per group %lu)", sbi->s_groups_count,
3702                       ext4_blocks_count(es),
3703                       le32_to_cpu(es->s_first_data_block),
3704                       EXT4_BLOCKS_PER_GROUP(sb));
3705                goto failed_mount;
3706        }
3707        sbi->s_groups_count = blocks_count;
3708        sbi->s_blockfile_groups = min_t(ext4_group_t, sbi->s_groups_count,
3709                        (EXT4_MAX_BLOCK_FILE_PHYS / EXT4_BLOCKS_PER_GROUP(sb)));
3710        db_count = (sbi->s_groups_count + EXT4_DESC_PER_BLOCK(sb) - 1) /
3711                   EXT4_DESC_PER_BLOCK(sb);
3712        sbi->s_group_desc = ext4_kvmalloc(db_count *
3713                                          sizeof(struct buffer_head *),
3714                                          GFP_KERNEL);
3715        if (sbi->s_group_desc == NULL) {
3716                ext4_msg(sb, KERN_ERR, "not enough memory");
3717                ret = -ENOMEM;
3718                goto failed_mount;
3719        }
3720
3721        if (ext4_proc_root)
3722                sbi->s_proc = proc_mkdir(sb->s_id, ext4_proc_root);
3723
3724        if (sbi->s_proc)
3725                proc_create_data("options", S_IRUGO, sbi->s_proc,
3726                                 &ext4_seq_options_fops, sb);
3727
3728        bgl_lock_init(sbi->s_blockgroup_lock);
3729
3730        for (i = 0; i < db_count; i++) {
3731                block = descriptor_loc(sb, logical_sb_block, i);
3732                sbi->s_group_desc[i] = sb_bread(sb, block);
3733                if (!sbi->s_group_desc[i]) {
3734                        ext4_msg(sb, KERN_ERR,
3735                               "can't read group descriptor %d", i);
3736                        db_count = i;
3737                        goto failed_mount2;
3738                }
3739        }
3740        if (!ext4_check_descriptors(sb, &first_not_zeroed)) {
3741                ext4_msg(sb, KERN_ERR, "group descriptors corrupted!");
3742                goto failed_mount2;
3743        }
3744        if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_FLEX_BG))
3745                if (!ext4_fill_flex_info(sb)) {
3746                        ext4_msg(sb, KERN_ERR,
3747                               "unable to initialize "
3748                               "flex_bg meta info!");
3749                        goto failed_mount2;
3750                }
3751
3752        sbi->s_gdb_count = db_count;
3753        get_random_bytes(&sbi->s_next_generation, sizeof(u32));
3754        spin_lock_init(&sbi->s_next_gen_lock);
3755
3756        init_timer(&sbi->s_err_report);
3757        sbi->s_err_report.function = print_daily_error_info;
3758        sbi->s_err_report.data = (unsigned long) sb;
3759
3760        err = percpu_counter_init(&sbi->s_freeclusters_counter,
3761                        ext4_count_free_clusters(sb));
3762        if (!err) {
3763                err = percpu_counter_init(&sbi->s_freeinodes_counter,
3764                                ext4_count_free_inodes(sb));
3765        }
3766        if (!err) {
3767                err = percpu_counter_init(&sbi->s_dirs_counter,
3768                                ext4_count_dirs(sb));
3769        }
3770        if (!err) {
3771                err = percpu_counter_init(&sbi->s_dirtyclusters_counter, 0);
3772        }
3773        if (err) {
3774                ext4_msg(sb, KERN_ERR, "insufficient memory");
3775                ret = err;
3776                goto failed_mount3;
3777        }
3778
3779        sbi->s_stripe = ext4_get_stripe_size(sbi);
3780        sbi->s_max_writeback_mb_bump = 128;
3781        sbi->s_extent_max_zeroout_kb = 32;
3782
3783        /*
3784         * set up enough so that it can read an inode
3785         */
3786        if (!test_opt(sb, NOLOAD) &&
3787            EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL))
3788                sb->s_op = &ext4_sops;
3789        else
3790                sb->s_op = &ext4_nojournal_sops;
3791        sb->s_export_op = &ext4_export_ops;
3792        sb->s_xattr = ext4_xattr_handlers;
3793#ifdef CONFIG_QUOTA
3794        sb->s_qcop = &ext4_qctl_operations;
3795        sb->dq_op = &ext4_quota_operations;
3796
3797        if (EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_QUOTA)) {
3798                /* Use qctl operations for hidden quota files. */
3799                sb->s_qcop = &ext4_qctl_sysfile_operations;
3800        }
3801#endif
3802        memcpy(sb->s_uuid, es->s_uuid, sizeof(es->s_uuid));
3803
3804        INIT_LIST_HEAD(&sbi->s_orphan); /* unlinked but open files */
3805        mutex_init(&sbi->s_orphan_lock);
3806        sbi->s_resize_flags = 0;
3807
3808        sb->s_root = NULL;
3809
3810        needs_recovery = (es->s_last_orphan != 0 ||
3811                          EXT4_HAS_INCOMPAT_FEATURE(sb,
3812                                    EXT4_FEATURE_INCOMPAT_RECOVER));
3813
3814        if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_MMP) &&
3815            !(sb->s_flags & MS_RDONLY))
3816                if (ext4_multi_mount_protect(sb, le64_to_cpu(es->s_mmp_block)))
3817                        goto failed_mount3;
3818
3819        /*
3820         * The first inode we look at is the journal inode.  Don't try
3821         * root first: it may be modified in the journal!
3822         */
3823        if (!test_opt(sb, NOLOAD) &&
3824            EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL)) {
3825                if (ext4_load_journal(sb, es, journal_devnum))
3826                        goto failed_mount3;
3827        } else if (test_opt(sb, NOLOAD) && !(sb->s_flags & MS_RDONLY) &&
3828              EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER)) {
3829                ext4_msg(sb, KERN_ERR, "required journal recovery "
3830                       "suppressed and not mounted read-only");
3831                goto failed_mount_wq;
3832        } else {
3833                clear_opt(sb, DATA_FLAGS);
3834                sbi->s_journal = NULL;
3835                needs_recovery = 0;
3836                goto no_journal;
3837        }
3838
3839        if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_64BIT) &&
3840            !jbd2_journal_set_features(EXT4_SB(sb)->s_journal, 0, 0,
3841                                       JBD2_FEATURE_INCOMPAT_64BIT)) {
3842                ext4_msg(sb, KERN_ERR, "Failed to set 64-bit journal feature");
3843                goto failed_mount_wq;
3844        }
3845
3846        if (!set_journal_csum_feature_set(sb)) {
3847                ext4_msg(sb, KERN_ERR, "Failed to set journal checksum "
3848                         "feature set");
3849                goto failed_mount_wq;
3850        }
3851
3852        /* We have now updated the journal if required, so we can
3853         * validate the data journaling mode. */
3854        switch (test_opt(sb, DATA_FLAGS)) {
3855        case 0:
3856                /* No mode set, assume a default based on the journal
3857                 * capabilities: ORDERED_DATA if the journal can
3858                 * cope, else JOURNAL_DATA
3859                 */
3860                if (jbd2_journal_check_available_features
3861                    (sbi->s_journal, 0, 0, JBD2_FEATURE_INCOMPAT_REVOKE))
3862                        set_opt(sb, ORDERED_DATA);
3863                else
3864                        set_opt(sb, JOURNAL_DATA);
3865                break;
3866
3867        case EXT4_MOUNT_ORDERED_DATA:
3868        case EXT4_MOUNT_WRITEBACK_DATA:
3869                if (!jbd2_journal_check_available_features
3870                    (sbi->s_journal, 0, 0, JBD2_FEATURE_INCOMPAT_REVOKE)) {
3871                        ext4_msg(sb, KERN_ERR, "Journal does not support "
3872                               "requested data journaling mode");
3873                        goto failed_mount_wq;
3874                }
3875        default:
3876                break;
3877        }
3878        set_task_ioprio(sbi->s_journal->j_task, journal_ioprio);
3879
3880        sbi->s_journal->j_commit_callback = ext4_journal_commit_callback;
3881
3882        /*
3883         * The journal may have updated the bg summary counts, so we
3884         * need to update the global counters.
3885         */
3886        percpu_counter_set(&sbi->s_freeclusters_counter,
3887                           ext4_count_free_clusters(sb));
3888        percpu_counter_set(&sbi->s_freeinodes_counter,
3889                           ext4_count_free_inodes(sb));
3890        percpu_counter_set(&sbi->s_dirs_counter,
3891                           ext4_count_dirs(sb));
3892        percpu_counter_set(&sbi->s_dirtyclusters_counter, 0);
3893
3894no_journal:
3895        /*
3896         * Get the # of file system overhead blocks from the
3897         * superblock if present.
3898         */
3899        if (es->s_overhead_clusters)
3900                sbi->s_overhead = le32_to_cpu(es->s_overhead_clusters);
3901        else {
3902                ret = ext4_calculate_overhead(sb);
3903                if (ret)
3904                        goto failed_mount_wq;
3905        }
3906
3907        /*
3908         * The maximum number of concurrent works can be high and
3909         * concurrency isn't really necessary.  Limit it to 1.
3910         */
3911        EXT4_SB(sb)->dio_unwritten_wq =
3912                alloc_workqueue("ext4-dio-unwritten", WQ_MEM_RECLAIM | WQ_UNBOUND, 1);
3913        if (!EXT4_SB(sb)->dio_unwritten_wq) {
3914                printk(KERN_ERR "EXT4-fs: failed to create DIO workqueue\n");
3915                goto failed_mount_wq;
3916        }
3917
3918        /*
3919         * The jbd2_journal_load will have done any necessary log recovery,
3920         * so we can safely mount the rest of the filesystem now.
3921         */
3922
3923        root = ext4_iget(sb, EXT4_ROOT_INO);
3924        if (IS_ERR(root)) {
3925                ext4_msg(sb, KERN_ERR, "get root inode failed");
3926                ret = PTR_ERR(root);
3927                root = NULL;
3928                goto failed_mount4;
3929        }
3930        if (!S_ISDIR(root->i_mode) || !root->i_blocks || !root->i_size) {
3931                ext4_msg(sb, KERN_ERR, "corrupt root inode, run e2fsck");
3932                iput(root);
3933                goto failed_mount4;
3934        }
3935        sb->s_root = d_make_root(root);
3936        if (!sb->s_root) {
3937                ext4_msg(sb, KERN_ERR, "get root dentry failed");
3938                ret = -ENOMEM;
3939                goto failed_mount4;
3940        }
3941
3942        if (ext4_setup_super(sb, es, sb->s_flags & MS_RDONLY))
3943                sb->s_flags |= MS_RDONLY;
3944
3945        /* determine the minimum size of new large inodes, if present */
3946        if (sbi->s_inode_size > EXT4_GOOD_OLD_INODE_SIZE) {
3947                sbi->s_want_extra_isize = sizeof(struct ext4_inode) -
3948                                                     EXT4_GOOD_OLD_INODE_SIZE;
3949                if (EXT4_HAS_RO_COMPAT_FEATURE(sb,
3950                                       EXT4_FEATURE_RO_COMPAT_EXTRA_ISIZE)) {
3951                        if (sbi->s_want_extra_isize <
3952                            le16_to_cpu(es->s_want_extra_isize))
3953                                sbi->s_want_extra_isize =
3954                                        le16_to_cpu(es->s_want_extra_isize);
3955                        if (sbi->s_want_extra_isize <
3956                            le16_to_cpu(es->s_min_extra_isize))
3957                                sbi->s_want_extra_isize =
3958                                        le16_to_cpu(es->s_min_extra_isize);
3959                }
3960        }
3961        /* Check if enough inode space is available */
3962        if (EXT4_GOOD_OLD_INODE_SIZE + sbi->s_want_extra_isize >
3963                                                        sbi->s_inode_size) {
3964                sbi->s_want_extra_isize = sizeof(struct ext4_inode) -
3965                                                       EXT4_GOOD_OLD_INODE_SIZE;
3966                ext4_msg(sb, KERN_INFO, "required extra inode space not"
3967                         "available");
3968        }
3969
3970        err = ext4_setup_system_zone(sb);
3971        if (err) {
3972                ext4_msg(sb, KERN_ERR, "failed to initialize system "
3973                         "zone (%d)", err);
3974                goto failed_mount4a;
3975        }
3976
3977        ext4_ext_init(sb);
3978        err = ext4_mb_init(sb);
3979        if (err) {
3980                ext4_msg(sb, KERN_ERR, "failed to initialize mballoc (%d)",
3981                         err);
3982                goto failed_mount5;
3983        }
3984
3985        err = ext4_register_li_request(sb, first_not_zeroed);
3986        if (err)
3987                goto failed_mount6;
3988
3989        sbi->s_kobj.kset = ext4_kset;
3990        init_completion(&sbi->s_kobj_unregister);
3991        err = kobject_init_and_add(&sbi->s_kobj, &ext4_ktype, NULL,
3992                                   "%s", sb->s_id);
3993        if (err)
3994                goto failed_mount7;
3995
3996        EXT4_SB(sb)->s_mount_state |= EXT4_ORPHAN_FS;
3997        ext4_orphan_cleanup(sb, es);
3998        EXT4_SB(sb)->s_mount_state &= ~EXT4_ORPHAN_FS;
3999        if (needs_recovery) {
4000                ext4_msg(sb, KERN_INFO, "recovery complete");
4001                ext4_mark_recovery_complete(sb, es);
4002        }
4003        if (EXT4_SB(sb)->s_journal) {
4004                if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA)
4005                        descr = " journalled data mode";
4006                else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA)
4007                        descr = " ordered data mode";
4008                else
4009                        descr = " writeback data mode";
4010        } else
4011                descr = "out journal";
4012
4013#ifdef CONFIG_QUOTA
4014        /* Enable quota usage during mount. */
4015        if (EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_QUOTA) &&
4016            !(sb->s_flags & MS_RDONLY)) {
4017                ret = ext4_enable_quotas(sb);
4018                if (ret)
4019                        goto failed_mount7;
4020        }
4021#endif  /* CONFIG_QUOTA */
4022
4023        ext4_msg(sb, KERN_INFO, "mounted filesystem with%s. "
4024                 "Opts: %s%s%s", descr, sbi->s_es->s_mount_opts,
4025                 *sbi->s_es->s_mount_opts ? "; " : "", orig_data);
4026
4027        if (es->s_error_count)
4028                mod_timer(&sbi->s_err_report, jiffies + 300*HZ); /* 5 minutes */
4029
4030        kfree(orig_data);
4031        return 0;
4032
4033cantfind_ext4:
4034        if (!silent)
4035                ext4_msg(sb, KERN_ERR, "VFS: Can't find ext4 filesystem");
4036        goto failed_mount;
4037
4038failed_mount7:
4039        ext4_unregister_li_request(sb);
4040failed_mount6:
4041        ext4_mb_release(sb);
4042failed_mount5:
4043        ext4_ext_release(sb);
4044        ext4_release_system_zone(sb);
4045failed_mount4a:
4046        dput(sb->s_root);
4047        sb->s_root = NULL;
4048failed_mount4:
4049        ext4_msg(sb, KERN_ERR, "mount failed");
4050        destroy_workqueue(EXT4_SB(sb)->dio_unwritten_wq);
4051failed_mount_wq:
4052        if (sbi->s_journal) {
4053                jbd2_journal_destroy(sbi->s_journal);
4054                sbi->s_journal = NULL;
4055        }
4056failed_mount3:
4057        del_timer(&sbi->s_err_report);
4058        if (sbi->s_flex_groups)
4059                ext4_kvfree(sbi->s_flex_groups);
4060        percpu_counter_destroy(&sbi->s_freeclusters_counter);
4061        percpu_counter_destroy(&sbi->s_freeinodes_counter);
4062        percpu_counter_destroy(&sbi->s_dirs_counter);
4063        percpu_counter_destroy(&sbi->s_dirtyclusters_counter);
4064        if (sbi->s_mmp_tsk)
4065                kthread_stop(sbi->s_mmp_tsk);
4066failed_mount2:
4067        for (i = 0; i < db_count; i++)
4068                brelse(sbi->s_group_desc[i]);
4069        ext4_kvfree(sbi->s_group_desc);
4070failed_mount:
4071        if (sbi->s_chksum_driver)
4072                crypto_free_shash(sbi->s_chksum_driver);
4073        if (sbi->s_proc) {
4074                remove_proc_entry("options", sbi->s_proc);
4075                remove_proc_entry(sb->s_id, ext4_proc_root);
4076        }
4077#ifdef CONFIG_QUOTA
4078        for (i = 0; i < MAXQUOTAS; i++)
4079                kfree(sbi->s_qf_names[i]);
4080#endif
4081        ext4_blkdev_remove(sbi);
4082        brelse(bh);
4083out_fail:
4084        sb->s_fs_info = NULL;
4085        kfree(sbi->s_blockgroup_lock);
4086        kfree(sbi);
4087out_free_orig:
4088        kfree(orig_data);
4089        return ret;
4090}
4091
4092/*
4093 * Setup any per-fs journal parameters now.  We'll do this both on
4094 * initial mount, once the journal has been initialised but before we've
4095 * done any recovery; and again on any subsequent remount.
4096 */
4097static void ext4_init_journal_params(struct super_block *sb, journal_t *journal)
4098{
4099        struct ext4_sb_info *sbi = EXT4_SB(sb);
4100
4101        journal->j_commit_interval = sbi->s_commit_interval;
4102        journal->j_min_batch_time = sbi->s_min_batch_time;
4103        journal->j_max_batch_time = sbi->s_max_batch_time;
4104
4105        write_lock(&journal->j_state_lock);
4106        if (test_opt(sb, BARRIER))
4107                journal->j_flags |= JBD2_BARRIER;
4108        else
4109                journal->j_flags &= ~JBD2_BARRIER;
4110        if (test_opt(sb, DATA_ERR_ABORT))
4111                journal->j_flags |= JBD2_ABORT_ON_SYNCDATA_ERR;
4112        else
4113                journal->j_flags &= ~JBD2_ABORT_ON_SYNCDATA_ERR;
4114        write_unlock(&journal->j_state_lock);
4115}
4116
4117static journal_t *ext4_get_journal(struct super_block *sb,
4118                                   unsigned int journal_inum)
4119{
4120        struct inode *journal_inode;
4121        journal_t *journal;
4122
4123        BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL));
4124
4125        /* First, test for the existence of a valid inode on disk.  Bad
4126         * things happen if we iget() an unused inode, as the subsequent
4127         * iput() will try to delete it. */
4128
4129        journal_inode = ext4_iget(sb, journal_inum);
4130        if (IS_ERR(journal_inode)) {
4131                ext4_msg(sb, KERN_ERR, "no journal found");
4132                return NULL;
4133        }
4134        if (!journal_inode->i_nlink) {
4135                make_bad_inode(journal_inode);
4136                iput(journal_inode);
4137                ext4_msg(sb, KERN_ERR, "journal inode is deleted");
4138                return NULL;
4139        }
4140
4141        jbd_debug(2, "Journal inode found at %p: %lld bytes\n",
4142                  journal_inode, journal_inode->i_size);
4143        if (!S_ISREG(journal_inode->i_mode)) {
4144                ext4_msg(sb, KERN_ERR, "invalid journal inode");
4145                iput(journal_inode);
4146                return NULL;
4147        }
4148
4149        journal = jbd2_journal_init_inode(journal_inode);
4150        if (!journal) {
4151                ext4_msg(sb, KERN_ERR, "Could not load journal inode");
4152                iput(journal_inode);
4153                return NULL;
4154        }
4155        journal->j_private = sb;
4156        ext4_init_journal_params(sb, journal);
4157        return journal;
4158}
4159
4160static journal_t *ext4_get_dev_journal(struct super_block *sb,
4161                                       dev_t j_dev)
4162{
4163        struct buffer_head *bh;
4164        journal_t *journal;
4165        ext4_fsblk_t start;
4166        ext4_fsblk_t len;
4167        int hblock, blocksize;
4168        ext4_fsblk_t sb_block;
4169        unsigned long offset;
4170        struct ext4_super_block *es;
4171        struct block_device *bdev;
4172
4173        BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL));
4174
4175        bdev = ext4_blkdev_get(j_dev, sb);
4176        if (bdev == NULL)
4177                return NULL;
4178
4179        blocksize = sb->s_blocksize;
4180        hblock = bdev_logical_block_size(bdev);
4181        if (blocksize < hblock) {
4182                ext4_msg(sb, KERN_ERR,
4183                        "blocksize too small for journal device");
4184                goto out_bdev;
4185        }
4186
4187        sb_block = EXT4_MIN_BLOCK_SIZE / blocksize;
4188        offset = EXT4_MIN_BLOCK_SIZE % blocksize;
4189        set_blocksize(bdev, blocksize);
4190        if (!(bh = __bread(bdev, sb_block, blocksize))) {
4191                ext4_msg(sb, KERN_ERR, "couldn't read superblock of "
4192                       "external journal");
4193                goto out_bdev;
4194        }
4195
4196        es = (struct ext4_super_block *) (bh->b_data + offset);
4197        if ((le16_to_cpu(es->s_magic) != EXT4_SUPER_MAGIC) ||
4198            !(le32_to_cpu(es->s_feature_incompat) &
4199              EXT4_FEATURE_INCOMPAT_JOURNAL_DEV)) {
4200                ext4_msg(sb, KERN_ERR, "external journal has "
4201                                        "bad superblock");
4202                brelse(bh);
4203                goto out_bdev;
4204        }
4205
4206        if (memcmp(EXT4_SB(sb)->s_es->s_journal_uuid, es->s_uuid, 16)) {
4207                ext4_msg(sb, KERN_ERR, "journal UUID does not match");
4208                brelse(bh);
4209                goto out_bdev;
4210        }
4211
4212        len = ext4_blocks_count(es);
4213        start = sb_block + 1;
4214        brelse(bh);     /* we're done with the superblock */
4215
4216        journal = jbd2_journal_init_dev(bdev, sb->s_bdev,
4217                                        start, len, blocksize);
4218        if (!journal) {
4219                ext4_msg(sb, KERN_ERR, "failed to create device journal");
4220                goto out_bdev;
4221        }
4222        journal->j_private = sb;
4223        ll_rw_block(READ, 1, &journal->j_sb_buffer);
4224        wait_on_buffer(journal->j_sb_buffer);
4225        if (!buffer_uptodate(journal->j_sb_buffer)) {
4226                ext4_msg(sb, KERN_ERR, "I/O error on journal device");
4227                goto out_journal;
4228        }
4229        if (be32_to_cpu(journal->j_superblock->s_nr_users) != 1) {
4230                ext4_msg(sb, KERN_ERR, "External journal has more than one "
4231                                        "user (unsupported) - %d",
4232                        be32_to_cpu(journal->j_superblock->s_nr_users));
4233                goto out_journal;
4234        }
4235        EXT4_SB(sb)->journal_bdev = bdev;
4236        ext4_init_journal_params(sb, journal);
4237        return journal;
4238
4239out_journal:
4240        jbd2_journal_destroy(journal);
4241out_bdev:
4242        ext4_blkdev_put(bdev);
4243        return NULL;
4244}
4245
4246static int ext4_load_journal(struct super_block *sb,
4247                             struct ext4_super_block *es,
4248                             unsigned long journal_devnum)
4249{
4250        journal_t *journal;
4251        unsigned int journal_inum = le32_to_cpu(es->s_journal_inum);
4252        dev_t journal_dev;
4253        int err = 0;
4254        int really_read_only;
4255
4256        BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL));
4257
4258        if (journal_devnum &&
4259            journal_devnum != le32_to_cpu(es->s_journal_dev)) {
4260                ext4_msg(sb, KERN_INFO, "external journal device major/minor "
4261                        "numbers have changed");
4262                journal_dev = new_decode_dev(journal_devnum);
4263        } else
4264                journal_dev = new_decode_dev(le32_to_cpu(es->s_journal_dev));
4265
4266        really_read_only = bdev_read_only(sb->s_bdev);
4267
4268        /*
4269         * Are we loading a blank journal or performing recovery after a
4270         * crash?  For recovery, we need to check in advance whether we
4271         * can get read-write access to the device.
4272         */
4273        if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER)) {
4274                if (sb->s_flags & MS_RDONLY) {
4275                        ext4_msg(sb, KERN_INFO, "INFO: recovery "
4276                                        "required on readonly filesystem");
4277                        if (really_read_only) {
4278                                ext4_msg(sb, KERN_ERR, "write access "
4279                                        "unavailable, cannot proceed");
4280                                return -EROFS;
4281                        }
4282                        ext4_msg(sb, KERN_INFO, "write access will "
4283                               "be enabled during recovery");
4284                }
4285        }
4286
4287        if (journal_inum && journal_dev) {
4288                ext4_msg(sb, KERN_ERR, "filesystem has both journal "
4289                       "and inode journals!");
4290                return -EINVAL;
4291        }
4292
4293        if (journal_inum) {
4294                if (!(journal = ext4_get_journal(sb, journal_inum)))
4295                        return -EINVAL;
4296        } else {
4297                if (!(journal = ext4_get_dev_journal(sb, journal_dev)))
4298                        return -EINVAL;
4299        }
4300
4301        if (!(journal->j_flags & JBD2_BARRIER))
4302                ext4_msg(sb, KERN_INFO, "barriers disabled");
4303
4304        if (!EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER))
4305                err = jbd2_journal_wipe(journal, !really_read_only);
4306        if (!err) {
4307                char *save = kmalloc(EXT4_S_ERR_LEN, GFP_KERNEL);
4308                if (save)
4309                        memcpy(save, ((char *) es) +
4310                               EXT4_S_ERR_START, EXT4_S_ERR_LEN);
4311                err = jbd2_journal_load(journal);
4312                if (save)
4313                        memcpy(((char *) es) + EXT4_S_ERR_START,
4314                               save, EXT4_S_ERR_LEN);
4315                kfree(save);
4316        }
4317
4318        if (err) {
4319                ext4_msg(sb, KERN_ERR, "error loading journal");
4320                jbd2_journal_destroy(journal);
4321                return err;
4322        }
4323
4324        EXT4_SB(sb)->s_journal = journal;
4325        ext4_clear_journal_err(sb, es);
4326
4327        if (!really_read_only && journal_devnum &&
4328            journal_devnum != le32_to_cpu(es->s_journal_dev)) {
4329                es->s_journal_dev = cpu_to_le32(journal_devnum);
4330
4331                /* Make sure we flush the recovery flag to disk. */
4332                ext4_commit_super(sb, 1);
4333        }
4334
4335        return 0;
4336}
4337
4338static int ext4_commit_super(struct super_block *sb, int sync)
4339{
4340        struct ext4_super_block *es = EXT4_SB(sb)->s_es;
4341        struct buffer_head *sbh = EXT4_SB(sb)->s_sbh;
4342        int error = 0;
4343
4344        if (!sbh || block_device_ejected(sb))
4345                return error;
4346        if (buffer_write_io_error(sbh)) {
4347                /*
4348                 * Oh, dear.  A previous attempt to write the
4349                 * superblock failed.  This could happen because the
4350                 * USB device was yanked out.  Or it could happen to
4351                 * be a transient write error and maybe the block will
4352                 * be remapped.  Nothing we can do but to retry the
4353                 * write and hope for the best.
4354                 */
4355                ext4_msg(sb, KERN_ERR, "previous I/O error to "
4356                       "superblock detected");
4357                clear_buffer_write_io_error(sbh);
4358                set_buffer_uptodate(sbh);
4359        }
4360        /*
4361         * If the file system is mounted read-only, don't update the
4362         * superblock write time.  This avoids updating the superblock
4363         * write time when we are mounting the root file system
4364         * read/only but we need to replay the journal; at that point,
4365         * for people who are east of GMT and who make their clock
4366         * tick in localtime for Windows bug-for-bug compatibility,
4367         * the clock is set in the future, and this will cause e2fsck
4368         * to complain and force a full file system check.
4369         */
4370        if (!(sb->s_flags & MS_RDONLY))
4371                es->s_wtime = cpu_to_le32(get_seconds());
4372        if (sb->s_bdev->bd_part)
4373                es->s_kbytes_written =
4374                        cpu_to_le64(EXT4_SB(sb)->s_kbytes_written +
4375                            ((part_stat_read(sb->s_bdev->bd_part, sectors[1]) -
4376                              EXT4_SB(sb)->s_sectors_written_start) >> 1));
4377        else
4378                es->s_kbytes_written =
4379                        cpu_to_le64(EXT4_SB(sb)->s_kbytes_written);
4380        ext4_free_blocks_count_set(es,
4381                        EXT4_C2B(EXT4_SB(sb), percpu_counter_sum_positive(
4382                                &EXT4_SB(sb)->s_freeclusters_counter)));
4383        es->s_free_inodes_count =
4384                cpu_to_le32(percpu_counter_sum_positive(
4385                                &EXT4_SB(sb)->s_freeinodes_counter));
4386        BUFFER_TRACE(sbh, "marking dirty");
4387        ext4_superblock_csum_set(sb);
4388        mark_buffer_dirty(sbh);
4389        if (sync) {
4390                error = sync_dirty_buffer(sbh);
4391                if (error)
4392                        return error;
4393
4394                error = buffer_write_io_error(sbh);
4395                if (error) {
4396                        ext4_msg(sb, KERN_ERR, "I/O error while writing "
4397                               "superblock");
4398                        clear_buffer_write_io_error(sbh);
4399                        set_buffer_uptodate(sbh);
4400                }
4401        }
4402        return error;
4403}
4404
4405/*
4406 * Have we just finished recovery?  If so, and if we are mounting (or
4407 * remounting) the filesystem readonly, then we will end up with a
4408 * consistent fs on disk.  Record that fact.
4409 */
4410static void ext4_mark_recovery_complete(struct super_block *sb,
4411                                        struct ext4_super_block *es)
4412{
4413        journal_t *journal = EXT4_SB(sb)->s_journal;
4414
4415        if (!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL)) {
4416                BUG_ON(journal != NULL);
4417                return;
4418        }
4419        jbd2_journal_lock_updates(journal);
4420        if (jbd2_journal_flush(journal) < 0)
4421                goto out;
4422
4423        if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER) &&
4424            sb->s_flags & MS_RDONLY) {
4425                EXT4_CLEAR_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
4426                ext4_commit_super(sb, 1);
4427        }
4428
4429out:
4430        jbd2_journal_unlock_updates(journal);
4431}
4432
4433/*
4434 * If we are mounting (or read-write remounting) a filesystem whose journal
4435 * has recorded an error from a previous lifetime, move that error to the
4436 * main filesystem now.
4437 */
4438static void ext4_clear_journal_err(struct super_block *sb,
4439                                   struct ext4_super_block *es)
4440{
4441        journal_t *journal;
4442        int j_errno;
4443        const char *errstr;
4444
4445        BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL));
4446
4447        journal = EXT4_SB(sb)->s_journal;
4448
4449        /*
4450         * Now check for any error status which may have been recorded in the
4451         * journal by a prior ext4_error() or ext4_abort()
4452         */
4453
4454        j_errno = jbd2_journal_errno(journal);
4455        if (j_errno) {
4456                char nbuf[16];
4457
4458                errstr = ext4_decode_error(sb, j_errno, nbuf);
4459                ext4_warning(sb, "Filesystem error recorded "
4460                             "from previous mount: %s", errstr);
4461                ext4_warning(sb, "Marking fs in need of filesystem check.");
4462
4463                EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS;
4464                es->s_state |= cpu_to_le16(EXT4_ERROR_FS);
4465                ext4_commit_super(sb, 1);
4466
4467                jbd2_journal_clear_err(journal);
4468                jbd2_journal_update_sb_errno(journal);
4469        }
4470}
4471
4472/*
4473 * Force the running and committing transactions to commit,
4474 * and wait on the commit.
4475 */
4476int ext4_force_commit(struct super_block *sb)
4477{
4478        journal_t *journal;
4479        int ret = 0;
4480
4481        if (sb->s_flags & MS_RDONLY)
4482                return 0;
4483
4484        journal = EXT4_SB(sb)->s_journal;
4485        if (journal)
4486                ret = ext4_journal_force_commit(journal);
4487
4488        return ret;
4489}
4490
4491static int ext4_sync_fs(struct super_block *sb, int wait)
4492{
4493        int ret = 0;
4494        tid_t target;
4495        struct ext4_sb_info *sbi = EXT4_SB(sb);
4496
4497        trace_ext4_sync_fs(sb, wait);
4498        flush_workqueue(sbi->dio_unwritten_wq);
4499        /*
4500         * Writeback quota in non-journalled quota case - journalled quota has
4501         * no dirty dquots
4502         */
4503        dquot_writeback_dquots(sb, -1);
4504        if (jbd2_journal_start_commit(sbi->s_journal, &target)) {
4505                if (wait)
4506                        jbd2_log_wait_commit(sbi->s_journal, target);
4507        }
4508        return ret;
4509}
4510
4511/*
4512 * LVM calls this function before a (read-only) snapshot is created.  This
4513 * gives us a chance to flush the journal completely and mark the fs clean.
4514 *
4515 * Note that only this function cannot bring a filesystem to be in a clean
4516 * state independently. It relies on upper layer to stop all data & metadata
4517 * modifications.
4518 */
4519static int ext4_freeze(struct super_block *sb)
4520{
4521        int error = 0;
4522        journal_t *journal;
4523
4524        if (sb->s_flags & MS_RDONLY)
4525                return 0;
4526
4527        journal = EXT4_SB(sb)->s_journal;
4528
4529        /* Now we set up the journal barrier. */
4530        jbd2_journal_lock_updates(journal);
4531
4532        /*
4533         * Don't clear the needs_recovery flag if we failed to flush
4534         * the journal.
4535         */
4536        error = jbd2_journal_flush(journal);
4537        if (error < 0)
4538                goto out;
4539
4540        /* Journal blocked and flushed, clear needs_recovery flag. */
4541        EXT4_CLEAR_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
4542        error = ext4_commit_super(sb, 1);
4543out:
4544        /* we rely on upper layer to stop further updates */
4545        jbd2_journal_unlock_updates(EXT4_SB(sb)->s_journal);
4546        return error;
4547}
4548
4549/*
4550 * Called by LVM after the snapshot is done.  We need to reset the RECOVER
4551 * flag here, even though the filesystem is not technically dirty yet.
4552 */
4553static int ext4_unfreeze(struct super_block *sb)
4554{
4555        if (sb->s_flags & MS_RDONLY)
4556                return 0;
4557
4558        /* Reset the needs_recovery flag before the fs is unlocked. */
4559        EXT4_SET_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
4560        ext4_commit_super(sb, 1);
4561        return 0;
4562}
4563
4564/*
4565 * Structure to save mount options for ext4_remount's benefit
4566 */
4567struct ext4_mount_options {
4568        unsigned long s_mount_opt;
4569        unsigned long s_mount_opt2;
4570        kuid_t s_resuid;
4571        kgid_t s_resgid;
4572        unsigned long s_commit_interval;
4573        u32 s_min_batch_time, s_max_batch_time;
4574#ifdef CONFIG_QUOTA
4575        int s_jquota_fmt;
4576        char *s_qf_names[MAXQUOTAS];
4577#endif
4578};
4579
4580static int ext4_remount(struct super_block *sb, int *flags, char *data)
4581{
4582        struct ext4_super_block *es;
4583        struct ext4_sb_info *sbi = EXT4_SB(sb);
4584        unsigned long old_sb_flags;
4585        struct ext4_mount_options old_opts;
4586        int enable_quota = 0;
4587        ext4_group_t g;
4588        unsigned int journal_ioprio = DEFAULT_JOURNAL_IOPRIO;
4589        int err = 0;
4590#ifdef CONFIG_QUOTA
4591        int i;
4592#endif
4593        char *orig_data = kstrdup(data, GFP_KERNEL);
4594
4595        /* Store the original options */
4596        old_sb_flags = sb->s_flags;
4597        old_opts.s_mount_opt = sbi->s_mount_opt;
4598        old_opts.s_mount_opt2 = sbi->s_mount_opt2;
4599        old_opts.s_resuid = sbi->s_resuid;
4600        old_opts.s_resgid = sbi->s_resgid;
4601        old_opts.s_commit_interval = sbi->s_commit_interval;
4602        old_opts.s_min_batch_time = sbi->s_min_batch_time;
4603        old_opts.s_max_batch_time = sbi->s_max_batch_time;
4604#ifdef CONFIG_QUOTA
4605        old_opts.s_jquota_fmt = sbi->s_jquota_fmt;
4606        for (i = 0; i < MAXQUOTAS; i++)
4607                old_opts.s_qf_names[i] = sbi->s_qf_names[i];
4608#endif
4609        if (sbi->s_journal && sbi->s_journal->j_task->io_context)
4610                journal_ioprio = sbi->s_journal->j_task->io_context->ioprio;
4611
4612        /*
4613         * Allow the "check" option to be passed as a remount option.
4614         */
4615        if (!parse_options(data, sb, NULL, &journal_ioprio, 1)) {
4616                err = -EINVAL;
4617                goto restore_opts;
4618        }
4619
4620        if (sbi->s_mount_flags & EXT4_MF_FS_ABORTED)
4621                ext4_abort(sb, "Abort forced by user");
4622
4623        sb->s_flags = (sb->s_flags & ~MS_POSIXACL) |
4624                (test_opt(sb, POSIX_ACL) ? MS_POSIXACL : 0);
4625
4626        es = sbi->s_es;
4627
4628        if (sbi->s_journal) {
4629                ext4_init_journal_params(sb, sbi->s_journal);
4630                set_task_ioprio(sbi->s_journal->j_task, journal_ioprio);
4631        }
4632
4633        if ((*flags & MS_RDONLY) != (sb->s_flags & MS_RDONLY)) {
4634                if (sbi->s_mount_flags & EXT4_MF_FS_ABORTED) {
4635                        err = -EROFS;
4636                        goto restore_opts;
4637                }
4638
4639                if (*flags & MS_RDONLY) {
4640                        err = dquot_suspend(sb, -1);
4641                        if (err < 0)
4642                                goto restore_opts;
4643
4644                        /*
4645                         * First of all, the unconditional stuff we have to do
4646                         * to disable replay of the journal when we next remount
4647                         */
4648                        sb->s_flags |= MS_RDONLY;
4649
4650                        /*
4651                         * OK, test if we are remounting a valid rw partition
4652                         * readonly, and if so set the rdonly flag and then
4653                         * mark the partition as valid again.
4654                         */
4655                        if (!(es->s_state & cpu_to_le16(EXT4_VALID_FS)) &&
4656                            (sbi->s_mount_state & EXT4_VALID_FS))
4657                                es->s_state = cpu_to_le16(sbi->s_mount_state);
4658
4659                        if (sbi->s_journal)
4660                                ext4_mark_recovery_complete(sb, es);
4661                } else {
4662                        /* Make sure we can mount this feature set readwrite */
4663                        if (!ext4_feature_set_ok(sb, 0)) {
4664                                err = -EROFS;
4665                                goto restore_opts;
4666                        }
4667                        /*
4668                         * Make sure the group descriptor checksums
4669                         * are sane.  If they aren't, refuse to remount r/w.
4670                         */
4671                        for (g = 0; g < sbi->s_groups_count; g++) {
4672                                struct ext4_group_desc *gdp =
4673                                        ext4_get_group_desc(sb, g, NULL);
4674
4675                                if (!ext4_group_desc_csum_verify(sb, g, gdp)) {
4676                                        ext4_msg(sb, KERN_ERR,
4677               "ext4_remount: Checksum for group %u failed (%u!=%u)",
4678                g, le16_to_cpu(ext4_group_desc_csum(sbi, g, gdp)),
4679                                               le16_to_cpu(gdp->bg_checksum));
4680                                        err = -EINVAL;
4681                                        goto restore_opts;
4682                                }
4683                        }
4684
4685                        /*
4686                         * If we have an unprocessed orphan list hanging
4687                         * around from a previously readonly bdev mount,
4688                         * require a full umount/remount for now.
4689                         */
4690                        if (es->s_last_orphan) {
4691                                ext4_msg(sb, KERN_WARNING, "Couldn't "
4692                                       "remount RDWR because of unprocessed "
4693                                       "orphan inode list.  Please "
4694                                       "umount/remount instead");
4695                                err = -EINVAL;
4696                                goto restore_opts;
4697                        }
4698
4699                        /*
4700                         * Mounting a RDONLY partition read-write, so reread
4701                         * and store the current valid flag.  (It may have
4702                         * been changed by e2fsck since we originally mounted
4703                         * the partition.)
4704                         */
4705                        if (sbi->s_journal)
4706                                ext4_clear_journal_err(sb, es);
4707                        sbi->s_mount_state = le16_to_cpu(es->s_state);
4708                        if (!ext4_setup_super(sb, es, 0))
4709                                sb->s_flags &= ~MS_RDONLY;
4710                        if (EXT4_HAS_INCOMPAT_FEATURE(sb,
4711                                                     EXT4_FEATURE_INCOMPAT_MMP))
4712                                if (ext4_multi_mount_protect(sb,
4713                                                le64_to_cpu(es->s_mmp_block))) {
4714                                        err = -EROFS;
4715                                        goto restore_opts;
4716                                }
4717                        enable_quota = 1;
4718                }
4719        }
4720
4721        /*
4722         * Reinitialize lazy itable initialization thread based on
4723         * current settings
4724         */
4725        if ((sb->s_flags & MS_RDONLY) || !test_opt(sb, INIT_INODE_TABLE))
4726                ext4_unregister_li_request(sb);
4727        else {
4728                ext4_group_t first_not_zeroed;
4729                first_not_zeroed = ext4_has_uninit_itable(sb);
4730                ext4_register_li_request(sb, first_not_zeroed);
4731        }
4732
4733        ext4_setup_system_zone(sb);
4734        if (sbi->s_journal == NULL && !(old_sb_flags & MS_RDONLY))
4735                ext4_commit_super(sb, 1);
4736
4737#ifdef CONFIG_QUOTA
4738        /* Release old quota file names */
4739        for (i = 0; i < MAXQUOTAS; i++)
4740                if (old_opts.s_qf_names[i] &&
4741                    old_opts.s_qf_names[i] != sbi->s_qf_names[i])
4742                        kfree(old_opts.s_qf_names[i]);
4743        if (enable_quota) {
4744                if (sb_any_quota_suspended(sb))
4745                        dquot_resume(sb, -1);
4746                else if (EXT4_HAS_RO_COMPAT_FEATURE(sb,
4747                                        EXT4_FEATURE_RO_COMPAT_QUOTA)) {
4748                        err = ext4_enable_quotas(sb);
4749                        if (err)
4750                                goto restore_opts;
4751                }
4752        }
4753#endif
4754
4755        ext4_msg(sb, KERN_INFO, "re-mounted. Opts: %s", orig_data);
4756        kfree(orig_data);
4757        return 0;
4758
4759restore_opts:
4760        sb->s_flags = old_sb_flags;
4761        sbi->s_mount_opt = old_opts.s_mount_opt;
4762        sbi->s_mount_opt2 = old_opts.s_mount_opt2;
4763        sbi->s_resuid = old_opts.s_resuid;
4764        sbi->s_resgid = old_opts.s_resgid;
4765        sbi->s_commit_interval = old_opts.s_commit_interval;
4766        sbi->s_min_batch_time = old_opts.s_min_batch_time;
4767        sbi->s_max_batch_time = old_opts.s_max_batch_time;
4768#ifdef CONFIG_QUOTA
4769        sbi->s_jquota_fmt = old_opts.s_jquota_fmt;
4770        for (i = 0; i < MAXQUOTAS; i++) {
4771                if (sbi->s_qf_names[i] &&
4772                    old_opts.s_qf_names[i] != sbi->s_qf_names[i])
4773                        kfree(sbi->s_qf_names[i]);
4774                sbi->s_qf_names[i] = old_opts.s_qf_names[i];
4775        }
4776#endif
4777        kfree(orig_data);
4778        return err;
4779}
4780
4781static int ext4_statfs(struct dentry *dentry, struct kstatfs *buf)
4782{
4783        struct super_block *sb = dentry->d_sb;
4784        struct ext4_sb_info *sbi = EXT4_SB(sb);
4785        struct ext4_super_block *es = sbi->s_es;
4786        ext4_fsblk_t overhead = 0;
4787        u64 fsid;
4788        s64 bfree;
4789
4790        if (!test_opt(sb, MINIX_DF))
4791                overhead = sbi->s_overhead;
4792
4793        buf->f_type = EXT4_SUPER_MAGIC;
4794        buf->f_bsize = sb->s_blocksize;
4795        buf->f_blocks = ext4_blocks_count(es) - EXT4_C2B(sbi, sbi->s_overhead);
4796        bfree = percpu_counter_sum_positive(&sbi->s_freeclusters_counter) -
4797                percpu_counter_sum_positive(&sbi->s_dirtyclusters_counter);
4798        /* prevent underflow in case that few free space is available */
4799        buf->f_bfree = EXT4_C2B(sbi, max_t(s64, bfree, 0));
4800        buf->f_bavail = buf->f_bfree - ext4_r_blocks_count(es);
4801        if (buf->f_bfree < ext4_r_blocks_count(es))
4802                buf->f_bavail = 0;
4803        buf->f_files = le32_to_cpu(es->s_inodes_count);
4804        buf->f_ffree = percpu_counter_sum_positive(&sbi->s_freeinodes_counter);
4805        buf->f_namelen = EXT4_NAME_LEN;
4806        fsid = le64_to_cpup((void *)es->s_uuid) ^
4807               le64_to_cpup((void *)es->s_uuid + sizeof(u64));
4808        buf->f_fsid.val[0] = fsid & 0xFFFFFFFFUL;
4809        buf->f_fsid.val[1] = (fsid >> 32) & 0xFFFFFFFFUL;
4810
4811        return 0;
4812}
4813
4814/* Helper function for writing quotas on sync - we need to start transaction
4815 * before quota file is locked for write. Otherwise the are possible deadlocks:
4816 * Process 1                         Process 2
4817 * ext4_create()                     quota_sync()
4818 *   jbd2_journal_start()                  write_dquot()
4819 *   dquot_initialize()                         down(dqio_mutex)
4820 *     down(dqio_mutex)                    jbd2_journal_start()
4821 *
4822 */
4823
4824#ifdef CONFIG_QUOTA
4825
4826static inline struct inode *dquot_to_inode(struct dquot *dquot)
4827{
4828        return sb_dqopt(dquot->dq_sb)->files[dquot->dq_id.type];
4829}
4830
4831static int ext4_write_dquot(struct dquot *dquot)
4832{
4833        int ret, err;
4834        handle_t *handle;
4835        struct inode *inode;
4836
4837        inode = dquot_to_inode(dquot);
4838        handle = ext4_journal_start(inode,
4839                                    EXT4_QUOTA_TRANS_BLOCKS(dquot->dq_sb));
4840        if (IS_ERR(handle))
4841                return PTR_ERR(handle);
4842        ret = dquot_commit(dquot);
4843        err = ext4_journal_stop(handle);
4844        if (!ret)
4845                ret = err;
4846        return ret;
4847}
4848
4849static int ext4_acquire_dquot(struct dquot *dquot)
4850{
4851        int ret, err;
4852        handle_t *handle;
4853
4854        handle = ext4_journal_start(dquot_to_inode(dquot),
4855                                    EXT4_QUOTA_INIT_BLOCKS(dquot->dq_sb));
4856        if (IS_ERR(handle))
4857                return PTR_ERR(handle);
4858        ret = dquot_acquire(dquot);
4859        err = ext4_journal_stop(handle);
4860        if (!ret)
4861                ret = err;
4862        return ret;
4863}
4864
4865static int ext4_release_dquot(struct dquot *dquot)
4866{
4867        int ret, err;
4868        handle_t *handle;
4869
4870        handle = ext4_journal_start(dquot_to_inode(dquot),
4871                                    EXT4_QUOTA_DEL_BLOCKS(dquot->dq_sb));
4872        if (IS_ERR(handle)) {
4873                /* Release dquot anyway to avoid endless cycle in dqput() */
4874                dquot_release(dquot);
4875                return PTR_ERR(handle);
4876        }
4877        ret = dquot_release(dquot);
4878        err = ext4_journal_stop(handle);
4879        if (!ret)
4880                ret = err;
4881        return ret;
4882}
4883
4884static int ext4_mark_dquot_dirty(struct dquot *dquot)
4885{
4886        /* Are we journaling quotas? */
4887        if (EXT4_SB(dquot->dq_sb)->s_qf_names[USRQUOTA] ||
4888            EXT4_SB(dquot->dq_sb)->s_qf_names[GRPQUOTA]) {
4889                dquot_mark_dquot_dirty(dquot);
4890                return ext4_write_dquot(dquot);
4891        } else {
4892                return dquot_mark_dquot_dirty(dquot);
4893        }
4894}
4895
4896static int ext4_write_info(struct super_block *sb, int type)
4897{
4898        int ret, err;
4899        handle_t *handle;
4900
4901        /* Data block + inode block */
4902        handle = ext4_journal_start(sb->s_root->d_inode, 2);
4903        if (IS_ERR(handle))
4904                return PTR_ERR(handle);
4905        ret = dquot_commit_info(sb, type);
4906        err = ext4_journal_stop(handle);
4907        if (!ret)
4908                ret = err;
4909        return ret;
4910}
4911
4912/*
4913 * Turn on quotas during mount time - we need to find
4914 * the quota file and such...
4915 */
4916static int ext4_quota_on_mount(struct super_block *sb, int type)
4917{
4918        return dquot_quota_on_mount(sb, EXT4_SB(sb)->s_qf_names[type],
4919                                        EXT4_SB(sb)->s_jquota_fmt, type);
4920}
4921
4922/*
4923 * Standard function to be called on quota_on
4924 */
4925static int ext4_quota_on(struct super_block *sb, int type, int format_id,
4926                         struct path *path)
4927{
4928        int err;
4929
4930        if (!test_opt(sb, QUOTA))
4931                return -EINVAL;
4932
4933        /* Quotafile not on the same filesystem? */
4934        if (path->dentry->d_sb != sb)
4935                return -EXDEV;
4936        /* Journaling quota? */
4937        if (EXT4_SB(sb)->s_qf_names[type]) {
4938                /* Quotafile not in fs root? */
4939                if (path->dentry->d_parent != sb->s_root)
4940                        ext4_msg(sb, KERN_WARNING,
4941                                "Quota file not on filesystem root. "
4942                                "Journaled quota will not work");
4943        }
4944
4945        /*
4946         * When we journal data on quota file, we have to flush journal to see
4947         * all updates to the file when we bypass pagecache...
4948         */
4949        if (EXT4_SB(sb)->s_journal &&
4950            ext4_should_journal_data(path->dentry->d_inode)) {
4951                /*
4952                 * We don't need to lock updates but journal_flush() could
4953                 * otherwise be livelocked...
4954                 */
4955                jbd2_journal_lock_updates(EXT4_SB(sb)->s_journal);
4956                err = jbd2_journal_flush(EXT4_SB(sb)->s_journal);
4957                jbd2_journal_unlock_updates(EXT4_SB(sb)->s_journal);
4958                if (err)
4959                        return err;
4960        }
4961
4962        return dquot_quota_on(sb, type, format_id, path);
4963}
4964
4965static int ext4_quota_enable(struct super_block *sb, int type, int format_id,
4966                             unsigned int flags)
4967{
4968        int err;
4969        struct inode *qf_inode;
4970        unsigned long qf_inums[MAXQUOTAS] = {
4971                le32_to_cpu(EXT4_SB(sb)->s_es->s_usr_quota_inum),
4972                le32_to_cpu(EXT4_SB(sb)->s_es->s_grp_quota_inum)
4973        };
4974
4975        BUG_ON(!EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_QUOTA));
4976
4977        if (!qf_inums[type])
4978                return -EPERM;
4979
4980        qf_inode = ext4_iget(sb, qf_inums[type]);
4981        if (IS_ERR(qf_inode)) {
4982                ext4_error(sb, "Bad quota inode # %lu", qf_inums[type]);
4983                return PTR_ERR(qf_inode);
4984        }
4985
4986        err = dquot_enable(qf_inode, type, format_id, flags);
4987        iput(qf_inode);
4988
4989        return err;
4990}
4991
4992/* Enable usage tracking for all quota types. */
4993static int ext4_enable_quotas(struct super_block *sb)
4994{
4995        int type, err = 0;
4996        unsigned long qf_inums[MAXQUOTAS] = {
4997                le32_to_cpu(EXT4_SB(sb)->s_es->s_usr_quota_inum),
4998                le32_to_cpu(EXT4_SB(sb)->s_es->s_grp_quota_inum)
4999        };
5000
5001        sb_dqopt(sb)->flags |= DQUOT_QUOTA_SYS_FILE;
5002        for (type = 0; type < MAXQUOTAS; type++) {
5003                if (qf_inums[type]) {
5004                        err = ext4_quota_enable(sb, type, QFMT_VFS_V1,
5005                                                DQUOT_USAGE_ENABLED);
5006                        if (err) {
5007                                ext4_warning(sb,
5008                                        "Failed to enable quota (type=%d) "
5009                                        "tracking. Please run e2fsck to fix.",
5010                                        type);
5011                                return err;
5012                        }
5013                }
5014        }
5015        return 0;
5016}
5017
5018/*
5019 * quota_on function that is used when QUOTA feature is set.
5020 */
5021static int ext4_quota_on_sysfile(struct super_block *sb, int type,
5022                                 int format_id)
5023{
5024        if (!EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_QUOTA))
5025                return -EINVAL;
5026
5027        /*
5028         * USAGE was enabled at mount time. Only need to enable LIMITS now.
5029         */
5030        return ext4_quota_enable(sb, type, format_id, DQUOT_LIMITS_ENABLED);
5031}
5032
5033static int ext4_quota_off(struct super_block *sb, int type)
5034{
5035        struct inode *inode = sb_dqopt(sb)->files[type];
5036        handle_t *handle;
5037
5038        /* Force all delayed allocation blocks to be allocated.
5039         * Caller already holds s_umount sem */
5040        if (test_opt(sb, DELALLOC))
5041                sync_filesystem(sb);
5042
5043        if (!inode)
5044                goto out;
5045
5046        /* Update modification times of quota files when userspace can
5047         * start looking at them */
5048        handle = ext4_journal_start(inode, 1);
5049        if (IS_ERR(handle))
5050                goto out;
5051        inode->i_mtime = inode->i_ctime = CURRENT_TIME;
5052        ext4_mark_inode_dirty(handle, inode);
5053        ext4_journal_stop(handle);
5054
5055out:
5056        return dquot_quota_off(sb, type);
5057}
5058
5059/*
5060 * quota_off function that is used when QUOTA feature is set.
5061 */
5062static int ext4_quota_off_sysfile(struct super_block *sb, int type)
5063{
5064        if (!EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_QUOTA))
5065                return -EINVAL;
5066
5067        /* Disable only the limits. */
5068        return dquot_disable(sb, type, DQUOT_LIMITS_ENABLED);
5069}
5070
5071/* Read data from quotafile - avoid pagecache and such because we cannot afford
5072 * acquiring the locks... As quota files are never truncated and quota code
5073 * itself serializes the operations (and no one else should touch the files)
5074 * we don't have to be afraid of races */
5075static ssize_t ext4_quota_read(struct super_block *sb, int type, char *data,
5076                               size_t len, loff_t off)
5077{
5078        struct inode *inode = sb_dqopt(sb)->files[type];
5079        ext4_lblk_t blk = off >> EXT4_BLOCK_SIZE_BITS(sb);
5080        int err = 0;
5081        int offset = off & (sb->s_blocksize - 1);
5082        int tocopy;
5083        size_t toread;
5084        struct buffer_head *bh;
5085        loff_t i_size = i_size_read(inode);
5086
5087        if (off > i_size)
5088                return 0;
5089        if (off+len > i_size)
5090                len = i_size-off;
5091        toread = len;
5092        while (toread > 0) {
5093                tocopy = sb->s_blocksize - offset < toread ?
5094                                sb->s_blocksize - offset : toread;
5095                bh = ext4_bread(NULL, inode, blk, 0, &err);
5096                if (err)
5097                        return err;
5098                if (!bh)        /* A hole? */
5099                        memset(data, 0, tocopy);
5100                else
5101                        memcpy(data, bh->b_data+offset, tocopy);
5102                brelse(bh);
5103                offset = 0;
5104                toread -= tocopy;
5105                data += tocopy;
5106                blk++;
5107        }
5108        return len;
5109}
5110
5111/* Write to quotafile (we know the transaction is already started and has
5112 * enough credits) */
5113static ssize_t ext4_quota_write(struct super_block *sb, int type,
5114                                const char *data, size_t len, loff_t off)
5115{
5116        struct inode *inode =