2 * linux/fs/ext4/super.c
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)
11 * linux/fs/minix/inode.c
13 * Copyright (C) 1991, 1992 Linus Torvalds
15 * Big-endian to little-endian byte-swapping/bitmaps by
16 * David S. Miller (davem@caip.rutgers.edu), 1995
19 #include <linux/module.h>
20 #include <linux/string.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>
44 #include <linux/kthread.h>
45 #include <linux/freezer.h>
48 #include "ext4_extents.h" /* Needed for trace points definition */
49 #include "ext4_jbd2.h"
54 #define CREATE_TRACE_POINTS
55 #include <trace/events/ext4.h>
57 static struct proc_dir_entry *ext4_proc_root;
58 static struct kset *ext4_kset;
59 static struct ext4_lazy_init *ext4_li_info;
60 static struct mutex ext4_li_mtx;
61 static struct ext4_features *ext4_feat;
62 static int ext4_mballoc_ready;
64 static int ext4_load_journal(struct super_block *, struct ext4_super_block *,
65 unsigned long journal_devnum);
66 static int ext4_show_options(struct seq_file *seq, struct dentry *root);
67 static int ext4_commit_super(struct super_block *sb, int sync);
68 static void ext4_mark_recovery_complete(struct super_block *sb,
69 struct ext4_super_block *es);
70 static void ext4_clear_journal_err(struct super_block *sb,
71 struct ext4_super_block *es);
72 static int ext4_sync_fs(struct super_block *sb, int wait);
73 static int ext4_remount(struct super_block *sb, int *flags, char *data);
74 static int ext4_statfs(struct dentry *dentry, struct kstatfs *buf);
75 static int ext4_unfreeze(struct super_block *sb);
76 static int ext4_freeze(struct super_block *sb);
77 static struct dentry *ext4_mount(struct file_system_type *fs_type, int flags,
78 const char *dev_name, void *data);
79 static inline int ext2_feature_set_ok(struct super_block *sb);
80 static inline int ext3_feature_set_ok(struct super_block *sb);
81 static int ext4_feature_set_ok(struct super_block *sb, int readonly);
82 static void ext4_destroy_lazyinit_thread(void);
83 static void ext4_unregister_li_request(struct super_block *sb);
84 static void ext4_clear_request_list(void);
85 static int ext4_reserve_clusters(struct ext4_sb_info *, ext4_fsblk_t);
87 #if !defined(CONFIG_EXT2_FS) && !defined(CONFIG_EXT2_FS_MODULE) && defined(CONFIG_EXT4_USE_FOR_EXT23)
88 static struct file_system_type ext2_fs_type = {
92 .kill_sb = kill_block_super,
93 .fs_flags = FS_REQUIRES_DEV,
95 MODULE_ALIAS_FS("ext2");
97 #define IS_EXT2_SB(sb) ((sb)->s_bdev->bd_holder == &ext2_fs_type)
99 #define IS_EXT2_SB(sb) (0)
103 #if !defined(CONFIG_EXT3_FS) && !defined(CONFIG_EXT3_FS_MODULE) && defined(CONFIG_EXT4_USE_FOR_EXT23)
104 static struct file_system_type ext3_fs_type = {
105 .owner = THIS_MODULE,
108 .kill_sb = kill_block_super,
109 .fs_flags = FS_REQUIRES_DEV,
111 MODULE_ALIAS_FS("ext3");
112 MODULE_ALIAS("ext3");
113 #define IS_EXT3_SB(sb) ((sb)->s_bdev->bd_holder == &ext3_fs_type)
115 #define IS_EXT3_SB(sb) (0)
118 static int ext4_verify_csum_type(struct super_block *sb,
119 struct ext4_super_block *es)
121 if (!EXT4_HAS_RO_COMPAT_FEATURE(sb,
122 EXT4_FEATURE_RO_COMPAT_METADATA_CSUM))
125 return es->s_checksum_type == EXT4_CRC32C_CHKSUM;
128 static __le32 ext4_superblock_csum(struct super_block *sb,
129 struct ext4_super_block *es)
131 struct ext4_sb_info *sbi = EXT4_SB(sb);
132 int offset = offsetof(struct ext4_super_block, s_checksum);
135 csum = ext4_chksum(sbi, ~0, (char *)es, offset);
137 return cpu_to_le32(csum);
140 static int ext4_superblock_csum_verify(struct super_block *sb,
141 struct ext4_super_block *es)
143 if (!ext4_has_metadata_csum(sb))
146 return es->s_checksum == ext4_superblock_csum(sb, es);
149 void ext4_superblock_csum_set(struct super_block *sb)
151 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
153 if (!ext4_has_metadata_csum(sb))
156 es->s_checksum = ext4_superblock_csum(sb, es);
159 void *ext4_kvmalloc(size_t size, gfp_t flags)
163 ret = kmalloc(size, flags | __GFP_NOWARN);
165 ret = __vmalloc(size, flags, PAGE_KERNEL);
169 void *ext4_kvzalloc(size_t size, gfp_t flags)
173 ret = kzalloc(size, flags | __GFP_NOWARN);
175 ret = __vmalloc(size, flags | __GFP_ZERO, PAGE_KERNEL);
179 void ext4_kvfree(void *ptr)
181 if (is_vmalloc_addr(ptr))
188 ext4_fsblk_t ext4_block_bitmap(struct super_block *sb,
189 struct ext4_group_desc *bg)
191 return le32_to_cpu(bg->bg_block_bitmap_lo) |
192 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
193 (ext4_fsblk_t)le32_to_cpu(bg->bg_block_bitmap_hi) << 32 : 0);
196 ext4_fsblk_t ext4_inode_bitmap(struct super_block *sb,
197 struct ext4_group_desc *bg)
199 return le32_to_cpu(bg->bg_inode_bitmap_lo) |
200 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
201 (ext4_fsblk_t)le32_to_cpu(bg->bg_inode_bitmap_hi) << 32 : 0);
204 ext4_fsblk_t ext4_inode_table(struct super_block *sb,
205 struct ext4_group_desc *bg)
207 return le32_to_cpu(bg->bg_inode_table_lo) |
208 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
209 (ext4_fsblk_t)le32_to_cpu(bg->bg_inode_table_hi) << 32 : 0);
212 __u32 ext4_free_group_clusters(struct super_block *sb,
213 struct ext4_group_desc *bg)
215 return le16_to_cpu(bg->bg_free_blocks_count_lo) |
216 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
217 (__u32)le16_to_cpu(bg->bg_free_blocks_count_hi) << 16 : 0);
220 __u32 ext4_free_inodes_count(struct super_block *sb,
221 struct ext4_group_desc *bg)
223 return le16_to_cpu(bg->bg_free_inodes_count_lo) |
224 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
225 (__u32)le16_to_cpu(bg->bg_free_inodes_count_hi) << 16 : 0);
228 __u32 ext4_used_dirs_count(struct super_block *sb,
229 struct ext4_group_desc *bg)
231 return le16_to_cpu(bg->bg_used_dirs_count_lo) |
232 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
233 (__u32)le16_to_cpu(bg->bg_used_dirs_count_hi) << 16 : 0);
236 __u32 ext4_itable_unused_count(struct super_block *sb,
237 struct ext4_group_desc *bg)
239 return le16_to_cpu(bg->bg_itable_unused_lo) |
240 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
241 (__u32)le16_to_cpu(bg->bg_itable_unused_hi) << 16 : 0);
244 void ext4_block_bitmap_set(struct super_block *sb,
245 struct ext4_group_desc *bg, ext4_fsblk_t blk)
247 bg->bg_block_bitmap_lo = cpu_to_le32((u32)blk);
248 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
249 bg->bg_block_bitmap_hi = cpu_to_le32(blk >> 32);
252 void ext4_inode_bitmap_set(struct super_block *sb,
253 struct ext4_group_desc *bg, ext4_fsblk_t blk)
255 bg->bg_inode_bitmap_lo = cpu_to_le32((u32)blk);
256 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
257 bg->bg_inode_bitmap_hi = cpu_to_le32(blk >> 32);
260 void ext4_inode_table_set(struct super_block *sb,
261 struct ext4_group_desc *bg, ext4_fsblk_t blk)
263 bg->bg_inode_table_lo = cpu_to_le32((u32)blk);
264 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
265 bg->bg_inode_table_hi = cpu_to_le32(blk >> 32);
268 void ext4_free_group_clusters_set(struct super_block *sb,
269 struct ext4_group_desc *bg, __u32 count)
271 bg->bg_free_blocks_count_lo = cpu_to_le16((__u16)count);
272 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
273 bg->bg_free_blocks_count_hi = cpu_to_le16(count >> 16);
276 void ext4_free_inodes_set(struct super_block *sb,
277 struct ext4_group_desc *bg, __u32 count)
279 bg->bg_free_inodes_count_lo = cpu_to_le16((__u16)count);
280 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
281 bg->bg_free_inodes_count_hi = cpu_to_le16(count >> 16);
284 void ext4_used_dirs_set(struct super_block *sb,
285 struct ext4_group_desc *bg, __u32 count)
287 bg->bg_used_dirs_count_lo = cpu_to_le16((__u16)count);
288 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
289 bg->bg_used_dirs_count_hi = cpu_to_le16(count >> 16);
292 void ext4_itable_unused_set(struct super_block *sb,
293 struct ext4_group_desc *bg, __u32 count)
295 bg->bg_itable_unused_lo = cpu_to_le16((__u16)count);
296 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
297 bg->bg_itable_unused_hi = cpu_to_le16(count >> 16);
301 static void __save_error_info(struct super_block *sb, const char *func,
304 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
306 EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS;
307 es->s_state |= cpu_to_le16(EXT4_ERROR_FS);
308 es->s_last_error_time = cpu_to_le32(get_seconds());
309 strncpy(es->s_last_error_func, func, sizeof(es->s_last_error_func));
310 es->s_last_error_line = cpu_to_le32(line);
311 if (!es->s_first_error_time) {
312 es->s_first_error_time = es->s_last_error_time;
313 strncpy(es->s_first_error_func, func,
314 sizeof(es->s_first_error_func));
315 es->s_first_error_line = cpu_to_le32(line);
316 es->s_first_error_ino = es->s_last_error_ino;
317 es->s_first_error_block = es->s_last_error_block;
320 * Start the daily error reporting function if it hasn't been
323 if (!es->s_error_count)
324 mod_timer(&EXT4_SB(sb)->s_err_report, jiffies + 24*60*60*HZ);
325 le32_add_cpu(&es->s_error_count, 1);
328 static void save_error_info(struct super_block *sb, const char *func,
331 __save_error_info(sb, func, line);
332 ext4_commit_super(sb, 1);
336 * The del_gendisk() function uninitializes the disk-specific data
337 * structures, including the bdi structure, without telling anyone
338 * else. Once this happens, any attempt to call mark_buffer_dirty()
339 * (for example, by ext4_commit_super), will cause a kernel OOPS.
340 * This is a kludge to prevent these oops until we can put in a proper
341 * hook in del_gendisk() to inform the VFS and file system layers.
343 static int block_device_ejected(struct super_block *sb)
345 struct inode *bd_inode = sb->s_bdev->bd_inode;
346 struct backing_dev_info *bdi = bd_inode->i_mapping->backing_dev_info;
348 return bdi->dev == NULL;
351 static void ext4_journal_commit_callback(journal_t *journal, transaction_t *txn)
353 struct super_block *sb = journal->j_private;
354 struct ext4_sb_info *sbi = EXT4_SB(sb);
355 int error = is_journal_aborted(journal);
356 struct ext4_journal_cb_entry *jce;
358 BUG_ON(txn->t_state == T_FINISHED);
359 spin_lock(&sbi->s_md_lock);
360 while (!list_empty(&txn->t_private_list)) {
361 jce = list_entry(txn->t_private_list.next,
362 struct ext4_journal_cb_entry, jce_list);
363 list_del_init(&jce->jce_list);
364 spin_unlock(&sbi->s_md_lock);
365 jce->jce_func(sb, jce, error);
366 spin_lock(&sbi->s_md_lock);
368 spin_unlock(&sbi->s_md_lock);
371 /* Deal with the reporting of failure conditions on a filesystem such as
372 * inconsistencies detected or read IO failures.
374 * On ext2, we can store the error state of the filesystem in the
375 * superblock. That is not possible on ext4, because we may have other
376 * write ordering constraints on the superblock which prevent us from
377 * writing it out straight away; and given that the journal is about to
378 * be aborted, we can't rely on the current, or future, transactions to
379 * write out the superblock safely.
381 * We'll just use the jbd2_journal_abort() error code to record an error in
382 * the journal instead. On recovery, the journal will complain about
383 * that error until we've noted it down and cleared it.
386 static void ext4_handle_error(struct super_block *sb)
388 if (sb->s_flags & MS_RDONLY)
391 if (!test_opt(sb, ERRORS_CONT)) {
392 journal_t *journal = EXT4_SB(sb)->s_journal;
394 EXT4_SB(sb)->s_mount_flags |= EXT4_MF_FS_ABORTED;
396 jbd2_journal_abort(journal, -EIO);
398 if (test_opt(sb, ERRORS_RO)) {
399 ext4_msg(sb, KERN_CRIT, "Remounting filesystem read-only");
401 * Make sure updated value of ->s_mount_flags will be visible
402 * before ->s_flags update
405 sb->s_flags |= MS_RDONLY;
407 if (test_opt(sb, ERRORS_PANIC))
408 panic("EXT4-fs (device %s): panic forced after error\n",
412 #define ext4_error_ratelimit(sb) \
413 ___ratelimit(&(EXT4_SB(sb)->s_err_ratelimit_state), \
416 void __ext4_error(struct super_block *sb, const char *function,
417 unsigned int line, const char *fmt, ...)
419 struct va_format vaf;
422 if (ext4_error_ratelimit(sb)) {
427 "EXT4-fs error (device %s): %s:%d: comm %s: %pV\n",
428 sb->s_id, function, line, current->comm, &vaf);
431 save_error_info(sb, function, line);
432 ext4_handle_error(sb);
435 void __ext4_error_inode(struct inode *inode, const char *function,
436 unsigned int line, ext4_fsblk_t block,
437 const char *fmt, ...)
440 struct va_format vaf;
441 struct ext4_super_block *es = EXT4_SB(inode->i_sb)->s_es;
443 es->s_last_error_ino = cpu_to_le32(inode->i_ino);
444 es->s_last_error_block = cpu_to_le64(block);
445 if (ext4_error_ratelimit(inode->i_sb)) {
450 printk(KERN_CRIT "EXT4-fs error (device %s): %s:%d: "
451 "inode #%lu: block %llu: comm %s: %pV\n",
452 inode->i_sb->s_id, function, line, inode->i_ino,
453 block, current->comm, &vaf);
455 printk(KERN_CRIT "EXT4-fs error (device %s): %s:%d: "
456 "inode #%lu: comm %s: %pV\n",
457 inode->i_sb->s_id, function, line, inode->i_ino,
458 current->comm, &vaf);
461 save_error_info(inode->i_sb, function, line);
462 ext4_handle_error(inode->i_sb);
465 void __ext4_error_file(struct file *file, const char *function,
466 unsigned int line, ext4_fsblk_t block,
467 const char *fmt, ...)
470 struct va_format vaf;
471 struct ext4_super_block *es;
472 struct inode *inode = file_inode(file);
473 char pathname[80], *path;
475 es = EXT4_SB(inode->i_sb)->s_es;
476 es->s_last_error_ino = cpu_to_le32(inode->i_ino);
477 if (ext4_error_ratelimit(inode->i_sb)) {
478 path = d_path(&(file->f_path), pathname, sizeof(pathname));
486 "EXT4-fs error (device %s): %s:%d: inode #%lu: "
487 "block %llu: comm %s: path %s: %pV\n",
488 inode->i_sb->s_id, function, line, inode->i_ino,
489 block, current->comm, path, &vaf);
492 "EXT4-fs error (device %s): %s:%d: inode #%lu: "
493 "comm %s: path %s: %pV\n",
494 inode->i_sb->s_id, function, line, inode->i_ino,
495 current->comm, path, &vaf);
498 save_error_info(inode->i_sb, function, line);
499 ext4_handle_error(inode->i_sb);
502 const char *ext4_decode_error(struct super_block *sb, int errno,
509 errstr = "IO failure";
512 errstr = "Out of memory";
515 if (!sb || (EXT4_SB(sb)->s_journal &&
516 EXT4_SB(sb)->s_journal->j_flags & JBD2_ABORT))
517 errstr = "Journal has aborted";
519 errstr = "Readonly filesystem";
522 /* If the caller passed in an extra buffer for unknown
523 * errors, textualise them now. Else we just return
526 /* Check for truncated error codes... */
527 if (snprintf(nbuf, 16, "error %d", -errno) >= 0)
536 /* __ext4_std_error decodes expected errors from journaling functions
537 * automatically and invokes the appropriate error response. */
539 void __ext4_std_error(struct super_block *sb, const char *function,
540 unsigned int line, int errno)
545 /* Special case: if the error is EROFS, and we're not already
546 * inside a transaction, then there's really no point in logging
548 if (errno == -EROFS && journal_current_handle() == NULL &&
549 (sb->s_flags & MS_RDONLY))
552 if (ext4_error_ratelimit(sb)) {
553 errstr = ext4_decode_error(sb, errno, nbuf);
554 printk(KERN_CRIT "EXT4-fs error (device %s) in %s:%d: %s\n",
555 sb->s_id, function, line, errstr);
558 save_error_info(sb, function, line);
559 ext4_handle_error(sb);
563 * ext4_abort is a much stronger failure handler than ext4_error. The
564 * abort function may be used to deal with unrecoverable failures such
565 * as journal IO errors or ENOMEM at a critical moment in log management.
567 * We unconditionally force the filesystem into an ABORT|READONLY state,
568 * unless the error response on the fs has been set to panic in which
569 * case we take the easy way out and panic immediately.
572 void __ext4_abort(struct super_block *sb, const char *function,
573 unsigned int line, const char *fmt, ...)
577 save_error_info(sb, function, line);
579 printk(KERN_CRIT "EXT4-fs error (device %s): %s:%d: ", sb->s_id,
585 if ((sb->s_flags & MS_RDONLY) == 0) {
586 ext4_msg(sb, KERN_CRIT, "Remounting filesystem read-only");
587 EXT4_SB(sb)->s_mount_flags |= EXT4_MF_FS_ABORTED;
589 * Make sure updated value of ->s_mount_flags will be visible
590 * before ->s_flags update
593 sb->s_flags |= MS_RDONLY;
594 if (EXT4_SB(sb)->s_journal)
595 jbd2_journal_abort(EXT4_SB(sb)->s_journal, -EIO);
596 save_error_info(sb, function, line);
598 if (test_opt(sb, ERRORS_PANIC))
599 panic("EXT4-fs panic from previous error\n");
602 void __ext4_msg(struct super_block *sb,
603 const char *prefix, const char *fmt, ...)
605 struct va_format vaf;
608 if (!___ratelimit(&(EXT4_SB(sb)->s_msg_ratelimit_state), "EXT4-fs"))
614 printk("%sEXT4-fs (%s): %pV\n", prefix, sb->s_id, &vaf);
618 void __ext4_warning(struct super_block *sb, const char *function,
619 unsigned int line, const char *fmt, ...)
621 struct va_format vaf;
624 if (!___ratelimit(&(EXT4_SB(sb)->s_warning_ratelimit_state),
631 printk(KERN_WARNING "EXT4-fs warning (device %s): %s:%d: %pV\n",
632 sb->s_id, function, line, &vaf);
636 void __ext4_grp_locked_error(const char *function, unsigned int line,
637 struct super_block *sb, ext4_group_t grp,
638 unsigned long ino, ext4_fsblk_t block,
639 const char *fmt, ...)
643 struct va_format vaf;
645 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
647 es->s_last_error_ino = cpu_to_le32(ino);
648 es->s_last_error_block = cpu_to_le64(block);
649 __save_error_info(sb, function, line);
651 if (ext4_error_ratelimit(sb)) {
655 printk(KERN_CRIT "EXT4-fs error (device %s): %s:%d: group %u, ",
656 sb->s_id, function, line, grp);
658 printk(KERN_CONT "inode %lu: ", ino);
660 printk(KERN_CONT "block %llu:",
661 (unsigned long long) block);
662 printk(KERN_CONT "%pV\n", &vaf);
666 if (test_opt(sb, ERRORS_CONT)) {
667 ext4_commit_super(sb, 0);
671 ext4_unlock_group(sb, grp);
672 ext4_handle_error(sb);
674 * We only get here in the ERRORS_RO case; relocking the group
675 * may be dangerous, but nothing bad will happen since the
676 * filesystem will have already been marked read/only and the
677 * journal has been aborted. We return 1 as a hint to callers
678 * who might what to use the return value from
679 * ext4_grp_locked_error() to distinguish between the
680 * ERRORS_CONT and ERRORS_RO case, and perhaps return more
681 * aggressively from the ext4 function in question, with a
682 * more appropriate error code.
684 ext4_lock_group(sb, grp);
688 void ext4_update_dynamic_rev(struct super_block *sb)
690 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
692 if (le32_to_cpu(es->s_rev_level) > EXT4_GOOD_OLD_REV)
696 "updating to rev %d because of new feature flag, "
697 "running e2fsck is recommended",
700 es->s_first_ino = cpu_to_le32(EXT4_GOOD_OLD_FIRST_INO);
701 es->s_inode_size = cpu_to_le16(EXT4_GOOD_OLD_INODE_SIZE);
702 es->s_rev_level = cpu_to_le32(EXT4_DYNAMIC_REV);
703 /* leave es->s_feature_*compat flags alone */
704 /* es->s_uuid will be set by e2fsck if empty */
707 * The rest of the superblock fields should be zero, and if not it
708 * means they are likely already in use, so leave them alone. We
709 * can leave it up to e2fsck to clean up any inconsistencies there.
714 * Open the external journal device
716 static struct block_device *ext4_blkdev_get(dev_t dev, struct super_block *sb)
718 struct block_device *bdev;
719 char b[BDEVNAME_SIZE];
721 bdev = blkdev_get_by_dev(dev, FMODE_READ|FMODE_WRITE|FMODE_EXCL, sb);
727 ext4_msg(sb, KERN_ERR, "failed to open journal device %s: %ld",
728 __bdevname(dev, b), PTR_ERR(bdev));
733 * Release the journal device
735 static void ext4_blkdev_put(struct block_device *bdev)
737 blkdev_put(bdev, FMODE_READ|FMODE_WRITE|FMODE_EXCL);
740 static void ext4_blkdev_remove(struct ext4_sb_info *sbi)
742 struct block_device *bdev;
743 bdev = sbi->journal_bdev;
745 ext4_blkdev_put(bdev);
746 sbi->journal_bdev = NULL;
750 static inline struct inode *orphan_list_entry(struct list_head *l)
752 return &list_entry(l, struct ext4_inode_info, i_orphan)->vfs_inode;
755 static void dump_orphan_list(struct super_block *sb, struct ext4_sb_info *sbi)
759 ext4_msg(sb, KERN_ERR, "sb orphan head is %d",
760 le32_to_cpu(sbi->s_es->s_last_orphan));
762 printk(KERN_ERR "sb_info orphan list:\n");
763 list_for_each(l, &sbi->s_orphan) {
764 struct inode *inode = orphan_list_entry(l);
766 "inode %s:%lu at %p: mode %o, nlink %d, next %d\n",
767 inode->i_sb->s_id, inode->i_ino, inode,
768 inode->i_mode, inode->i_nlink,
773 static void ext4_put_super(struct super_block *sb)
775 struct ext4_sb_info *sbi = EXT4_SB(sb);
776 struct ext4_super_block *es = sbi->s_es;
779 ext4_unregister_li_request(sb);
780 dquot_disable(sb, -1, DQUOT_USAGE_ENABLED | DQUOT_LIMITS_ENABLED);
782 flush_workqueue(sbi->rsv_conversion_wq);
783 destroy_workqueue(sbi->rsv_conversion_wq);
785 if (sbi->s_journal) {
786 err = jbd2_journal_destroy(sbi->s_journal);
787 sbi->s_journal = NULL;
789 ext4_abort(sb, "Couldn't clean up the journal");
792 ext4_es_unregister_shrinker(sbi);
793 del_timer_sync(&sbi->s_err_report);
794 ext4_release_system_zone(sb);
796 ext4_ext_release(sb);
797 ext4_xattr_put_super(sb);
799 if (!(sb->s_flags & MS_RDONLY)) {
800 EXT4_CLEAR_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
801 es->s_state = cpu_to_le16(sbi->s_mount_state);
803 if (!(sb->s_flags & MS_RDONLY))
804 ext4_commit_super(sb, 1);
807 remove_proc_entry("options", sbi->s_proc);
808 remove_proc_entry(sb->s_id, ext4_proc_root);
810 kobject_del(&sbi->s_kobj);
812 for (i = 0; i < sbi->s_gdb_count; i++)
813 brelse(sbi->s_group_desc[i]);
814 ext4_kvfree(sbi->s_group_desc);
815 ext4_kvfree(sbi->s_flex_groups);
816 percpu_counter_destroy(&sbi->s_freeclusters_counter);
817 percpu_counter_destroy(&sbi->s_freeinodes_counter);
818 percpu_counter_destroy(&sbi->s_dirs_counter);
819 percpu_counter_destroy(&sbi->s_dirtyclusters_counter);
822 for (i = 0; i < EXT4_MAXQUOTAS; i++)
823 kfree(sbi->s_qf_names[i]);
826 /* Debugging code just in case the in-memory inode orphan list
827 * isn't empty. The on-disk one can be non-empty if we've
828 * detected an error and taken the fs readonly, but the
829 * in-memory list had better be clean by this point. */
830 if (!list_empty(&sbi->s_orphan))
831 dump_orphan_list(sb, sbi);
832 J_ASSERT(list_empty(&sbi->s_orphan));
834 invalidate_bdev(sb->s_bdev);
835 if (sbi->journal_bdev && sbi->journal_bdev != sb->s_bdev) {
837 * Invalidate the journal device's buffers. We don't want them
838 * floating about in memory - the physical journal device may
839 * hotswapped, and it breaks the `ro-after' testing code.
841 sync_blockdev(sbi->journal_bdev);
842 invalidate_bdev(sbi->journal_bdev);
843 ext4_blkdev_remove(sbi);
845 if (sbi->s_mb_cache) {
846 ext4_xattr_destroy_cache(sbi->s_mb_cache);
847 sbi->s_mb_cache = NULL;
850 kthread_stop(sbi->s_mmp_tsk);
851 sb->s_fs_info = NULL;
853 * Now that we are completely done shutting down the
854 * superblock, we need to actually destroy the kobject.
856 kobject_put(&sbi->s_kobj);
857 wait_for_completion(&sbi->s_kobj_unregister);
858 if (sbi->s_chksum_driver)
859 crypto_free_shash(sbi->s_chksum_driver);
860 kfree(sbi->s_blockgroup_lock);
864 static struct kmem_cache *ext4_inode_cachep;
867 * Called inside transaction, so use GFP_NOFS
869 static struct inode *ext4_alloc_inode(struct super_block *sb)
871 struct ext4_inode_info *ei;
873 ei = kmem_cache_alloc(ext4_inode_cachep, GFP_NOFS);
877 ei->vfs_inode.i_version = 1;
878 spin_lock_init(&ei->i_raw_lock);
879 INIT_LIST_HEAD(&ei->i_prealloc_list);
880 spin_lock_init(&ei->i_prealloc_lock);
881 ext4_es_init_tree(&ei->i_es_tree);
882 rwlock_init(&ei->i_es_lock);
883 INIT_LIST_HEAD(&ei->i_es_lru);
886 ei->i_touch_when = 0;
887 ei->i_reserved_data_blocks = 0;
888 ei->i_reserved_meta_blocks = 0;
889 ei->i_allocated_meta_blocks = 0;
890 ei->i_da_metadata_calc_len = 0;
891 ei->i_da_metadata_calc_last_lblock = 0;
892 spin_lock_init(&(ei->i_block_reservation_lock));
894 ei->i_reserved_quota = 0;
897 INIT_LIST_HEAD(&ei->i_rsv_conversion_list);
898 spin_lock_init(&ei->i_completed_io_lock);
900 ei->i_datasync_tid = 0;
901 atomic_set(&ei->i_ioend_count, 0);
902 atomic_set(&ei->i_unwritten, 0);
903 INIT_WORK(&ei->i_rsv_conversion_work, ext4_end_io_rsv_work);
905 return &ei->vfs_inode;
908 static int ext4_drop_inode(struct inode *inode)
910 int drop = generic_drop_inode(inode);
912 trace_ext4_drop_inode(inode, drop);
916 static void ext4_i_callback(struct rcu_head *head)
918 struct inode *inode = container_of(head, struct inode, i_rcu);
919 kmem_cache_free(ext4_inode_cachep, EXT4_I(inode));
922 static void ext4_destroy_inode(struct inode *inode)
924 if (!list_empty(&(EXT4_I(inode)->i_orphan))) {
925 ext4_msg(inode->i_sb, KERN_ERR,
926 "Inode %lu (%p): orphan list check failed!",
927 inode->i_ino, EXT4_I(inode));
928 print_hex_dump(KERN_INFO, "", DUMP_PREFIX_ADDRESS, 16, 4,
929 EXT4_I(inode), sizeof(struct ext4_inode_info),
933 call_rcu(&inode->i_rcu, ext4_i_callback);
936 static void init_once(void *foo)
938 struct ext4_inode_info *ei = (struct ext4_inode_info *) foo;
940 INIT_LIST_HEAD(&ei->i_orphan);
941 init_rwsem(&ei->xattr_sem);
942 init_rwsem(&ei->i_data_sem);
943 inode_init_once(&ei->vfs_inode);
946 static int __init init_inodecache(void)
948 ext4_inode_cachep = kmem_cache_create("ext4_inode_cache",
949 sizeof(struct ext4_inode_info),
950 0, (SLAB_RECLAIM_ACCOUNT|
953 if (ext4_inode_cachep == NULL)
958 static void destroy_inodecache(void)
961 * Make sure all delayed rcu free inodes are flushed before we
965 kmem_cache_destroy(ext4_inode_cachep);
968 void ext4_clear_inode(struct inode *inode)
970 invalidate_inode_buffers(inode);
973 ext4_discard_preallocations(inode);
974 ext4_es_remove_extent(inode, 0, EXT_MAX_BLOCKS);
975 ext4_es_lru_del(inode);
976 if (EXT4_I(inode)->jinode) {
977 jbd2_journal_release_jbd_inode(EXT4_JOURNAL(inode),
978 EXT4_I(inode)->jinode);
979 jbd2_free_inode(EXT4_I(inode)->jinode);
980 EXT4_I(inode)->jinode = NULL;
984 static struct inode *ext4_nfs_get_inode(struct super_block *sb,
985 u64 ino, u32 generation)
989 if (ino < EXT4_FIRST_INO(sb) && ino != EXT4_ROOT_INO)
990 return ERR_PTR(-ESTALE);
991 if (ino > le32_to_cpu(EXT4_SB(sb)->s_es->s_inodes_count))
992 return ERR_PTR(-ESTALE);
994 /* iget isn't really right if the inode is currently unallocated!!
996 * ext4_read_inode will return a bad_inode if the inode had been
997 * deleted, so we should be safe.
999 * Currently we don't know the generation for parent directory, so
1000 * a generation of 0 means "accept any"
1002 inode = ext4_iget_normal(sb, ino);
1004 return ERR_CAST(inode);
1005 if (generation && inode->i_generation != generation) {
1007 return ERR_PTR(-ESTALE);
1013 static struct dentry *ext4_fh_to_dentry(struct super_block *sb, struct fid *fid,
1014 int fh_len, int fh_type)
1016 return generic_fh_to_dentry(sb, fid, fh_len, fh_type,
1017 ext4_nfs_get_inode);
1020 static struct dentry *ext4_fh_to_parent(struct super_block *sb, struct fid *fid,
1021 int fh_len, int fh_type)
1023 return generic_fh_to_parent(sb, fid, fh_len, fh_type,
1024 ext4_nfs_get_inode);
1028 * Try to release metadata pages (indirect blocks, directories) which are
1029 * mapped via the block device. Since these pages could have journal heads
1030 * which would prevent try_to_free_buffers() from freeing them, we must use
1031 * jbd2 layer's try_to_free_buffers() function to release them.
1033 static int bdev_try_to_free_page(struct super_block *sb, struct page *page,
1036 journal_t *journal = EXT4_SB(sb)->s_journal;
1038 WARN_ON(PageChecked(page));
1039 if (!page_has_buffers(page))
1042 return jbd2_journal_try_to_free_buffers(journal, page,
1043 wait & ~__GFP_WAIT);
1044 return try_to_free_buffers(page);
1048 #define QTYPE2NAME(t) ((t) == USRQUOTA ? "user" : "group")
1049 #define QTYPE2MOPT(on, t) ((t) == USRQUOTA?((on)##USRJQUOTA):((on)##GRPJQUOTA))
1051 static int ext4_write_dquot(struct dquot *dquot);
1052 static int ext4_acquire_dquot(struct dquot *dquot);
1053 static int ext4_release_dquot(struct dquot *dquot);
1054 static int ext4_mark_dquot_dirty(struct dquot *dquot);
1055 static int ext4_write_info(struct super_block *sb, int type);
1056 static int ext4_quota_on(struct super_block *sb, int type, int format_id,
1058 static int ext4_quota_on_sysfile(struct super_block *sb, int type,
1060 static int ext4_quota_off(struct super_block *sb, int type);
1061 static int ext4_quota_off_sysfile(struct super_block *sb, int type);
1062 static int ext4_quota_on_mount(struct super_block *sb, int type);
1063 static ssize_t ext4_quota_read(struct super_block *sb, int type, char *data,
1064 size_t len, loff_t off);
1065 static ssize_t ext4_quota_write(struct super_block *sb, int type,
1066 const char *data, size_t len, loff_t off);
1067 static int ext4_quota_enable(struct super_block *sb, int type, int format_id,
1068 unsigned int flags);
1069 static int ext4_enable_quotas(struct super_block *sb);
1071 static const struct dquot_operations ext4_quota_operations = {
1072 .get_reserved_space = ext4_get_reserved_space,
1073 .write_dquot = ext4_write_dquot,
1074 .acquire_dquot = ext4_acquire_dquot,
1075 .release_dquot = ext4_release_dquot,
1076 .mark_dirty = ext4_mark_dquot_dirty,
1077 .write_info = ext4_write_info,
1078 .alloc_dquot = dquot_alloc,
1079 .destroy_dquot = dquot_destroy,
1082 static const struct quotactl_ops ext4_qctl_operations = {
1083 .quota_on = ext4_quota_on,
1084 .quota_off = ext4_quota_off,
1085 .quota_sync = dquot_quota_sync,
1086 .get_info = dquot_get_dqinfo,
1087 .set_info = dquot_set_dqinfo,
1088 .get_dqblk = dquot_get_dqblk,
1089 .set_dqblk = dquot_set_dqblk
1092 static const struct quotactl_ops ext4_qctl_sysfile_operations = {
1093 .quota_on_meta = ext4_quota_on_sysfile,
1094 .quota_off = ext4_quota_off_sysfile,
1095 .quota_sync = dquot_quota_sync,
1096 .get_info = dquot_get_dqinfo,
1097 .set_info = dquot_set_dqinfo,
1098 .get_dqblk = dquot_get_dqblk,
1099 .set_dqblk = dquot_set_dqblk
1103 static const struct super_operations ext4_sops = {
1104 .alloc_inode = ext4_alloc_inode,
1105 .destroy_inode = ext4_destroy_inode,
1106 .write_inode = ext4_write_inode,
1107 .dirty_inode = ext4_dirty_inode,
1108 .drop_inode = ext4_drop_inode,
1109 .evict_inode = ext4_evict_inode,
1110 .put_super = ext4_put_super,
1111 .sync_fs = ext4_sync_fs,
1112 .freeze_fs = ext4_freeze,
1113 .unfreeze_fs = ext4_unfreeze,
1114 .statfs = ext4_statfs,
1115 .remount_fs = ext4_remount,
1116 .show_options = ext4_show_options,
1118 .quota_read = ext4_quota_read,
1119 .quota_write = ext4_quota_write,
1121 .bdev_try_to_free_page = bdev_try_to_free_page,
1124 static const struct export_operations ext4_export_ops = {
1125 .fh_to_dentry = ext4_fh_to_dentry,
1126 .fh_to_parent = ext4_fh_to_parent,
1127 .get_parent = ext4_get_parent,
1131 Opt_bsd_df, Opt_minix_df, Opt_grpid, Opt_nogrpid,
1132 Opt_resgid, Opt_resuid, Opt_sb, Opt_err_cont, Opt_err_panic, Opt_err_ro,
1133 Opt_nouid32, Opt_debug, Opt_removed,
1134 Opt_user_xattr, Opt_nouser_xattr, Opt_acl, Opt_noacl,
1135 Opt_auto_da_alloc, Opt_noauto_da_alloc, Opt_noload,
1136 Opt_commit, Opt_min_batch_time, Opt_max_batch_time, Opt_journal_dev,
1137 Opt_journal_path, Opt_journal_checksum, Opt_journal_async_commit,
1138 Opt_abort, Opt_data_journal, Opt_data_ordered, Opt_data_writeback,
1139 Opt_data_err_abort, Opt_data_err_ignore,
1140 Opt_usrjquota, Opt_grpjquota, Opt_offusrjquota, Opt_offgrpjquota,
1141 Opt_jqfmt_vfsold, Opt_jqfmt_vfsv0, Opt_jqfmt_vfsv1, Opt_quota,
1142 Opt_noquota, Opt_barrier, Opt_nobarrier, Opt_err,
1143 Opt_usrquota, Opt_grpquota, Opt_i_version,
1144 Opt_stripe, Opt_delalloc, Opt_nodelalloc, Opt_mblk_io_submit,
1145 Opt_nomblk_io_submit, Opt_block_validity, Opt_noblock_validity,
1146 Opt_inode_readahead_blks, Opt_journal_ioprio,
1147 Opt_dioread_nolock, Opt_dioread_lock,
1148 Opt_discard, Opt_nodiscard, Opt_init_itable, Opt_noinit_itable,
1149 Opt_max_dir_size_kb,
1152 static const match_table_t tokens = {
1153 {Opt_bsd_df, "bsddf"},
1154 {Opt_minix_df, "minixdf"},
1155 {Opt_grpid, "grpid"},
1156 {Opt_grpid, "bsdgroups"},
1157 {Opt_nogrpid, "nogrpid"},
1158 {Opt_nogrpid, "sysvgroups"},
1159 {Opt_resgid, "resgid=%u"},
1160 {Opt_resuid, "resuid=%u"},
1162 {Opt_err_cont, "errors=continue"},
1163 {Opt_err_panic, "errors=panic"},
1164 {Opt_err_ro, "errors=remount-ro"},
1165 {Opt_nouid32, "nouid32"},
1166 {Opt_debug, "debug"},
1167 {Opt_removed, "oldalloc"},
1168 {Opt_removed, "orlov"},
1169 {Opt_user_xattr, "user_xattr"},
1170 {Opt_nouser_xattr, "nouser_xattr"},
1172 {Opt_noacl, "noacl"},
1173 {Opt_noload, "norecovery"},
1174 {Opt_noload, "noload"},
1175 {Opt_removed, "nobh"},
1176 {Opt_removed, "bh"},
1177 {Opt_commit, "commit=%u"},
1178 {Opt_min_batch_time, "min_batch_time=%u"},
1179 {Opt_max_batch_time, "max_batch_time=%u"},
1180 {Opt_journal_dev, "journal_dev=%u"},
1181 {Opt_journal_path, "journal_path=%s"},
1182 {Opt_journal_checksum, "journal_checksum"},
1183 {Opt_journal_async_commit, "journal_async_commit"},
1184 {Opt_abort, "abort"},
1185 {Opt_data_journal, "data=journal"},
1186 {Opt_data_ordered, "data=ordered"},
1187 {Opt_data_writeback, "data=writeback"},
1188 {Opt_data_err_abort, "data_err=abort"},
1189 {Opt_data_err_ignore, "data_err=ignore"},
1190 {Opt_offusrjquota, "usrjquota="},
1191 {Opt_usrjquota, "usrjquota=%s"},
1192 {Opt_offgrpjquota, "grpjquota="},
1193 {Opt_grpjquota, "grpjquota=%s"},
1194 {Opt_jqfmt_vfsold, "jqfmt=vfsold"},
1195 {Opt_jqfmt_vfsv0, "jqfmt=vfsv0"},
1196 {Opt_jqfmt_vfsv1, "jqfmt=vfsv1"},
1197 {Opt_grpquota, "grpquota"},
1198 {Opt_noquota, "noquota"},
1199 {Opt_quota, "quota"},
1200 {Opt_usrquota, "usrquota"},
1201 {Opt_barrier, "barrier=%u"},
1202 {Opt_barrier, "barrier"},
1203 {Opt_nobarrier, "nobarrier"},
1204 {Opt_i_version, "i_version"},
1205 {Opt_stripe, "stripe=%u"},
1206 {Opt_delalloc, "delalloc"},
1207 {Opt_nodelalloc, "nodelalloc"},
1208 {Opt_removed, "mblk_io_submit"},
1209 {Opt_removed, "nomblk_io_submit"},
1210 {Opt_block_validity, "block_validity"},
1211 {Opt_noblock_validity, "noblock_validity"},
1212 {Opt_inode_readahead_blks, "inode_readahead_blks=%u"},
1213 {Opt_journal_ioprio, "journal_ioprio=%u"},
1214 {Opt_auto_da_alloc, "auto_da_alloc=%u"},
1215 {Opt_auto_da_alloc, "auto_da_alloc"},
1216 {Opt_noauto_da_alloc, "noauto_da_alloc"},
1217 {Opt_dioread_nolock, "dioread_nolock"},
1218 {Opt_dioread_lock, "dioread_lock"},
1219 {Opt_discard, "discard"},
1220 {Opt_nodiscard, "nodiscard"},
1221 {Opt_init_itable, "init_itable=%u"},
1222 {Opt_init_itable, "init_itable"},
1223 {Opt_noinit_itable, "noinit_itable"},
1224 {Opt_max_dir_size_kb, "max_dir_size_kb=%u"},
1225 {Opt_removed, "check=none"}, /* mount option from ext2/3 */
1226 {Opt_removed, "nocheck"}, /* mount option from ext2/3 */
1227 {Opt_removed, "reservation"}, /* mount option from ext2/3 */
1228 {Opt_removed, "noreservation"}, /* mount option from ext2/3 */
1229 {Opt_removed, "journal=%u"}, /* mount option from ext2/3 */
1233 static ext4_fsblk_t get_sb_block(void **data)
1235 ext4_fsblk_t sb_block;
1236 char *options = (char *) *data;
1238 if (!options || strncmp(options, "sb=", 3) != 0)
1239 return 1; /* Default location */
1242 /* TODO: use simple_strtoll with >32bit ext4 */
1243 sb_block = simple_strtoul(options, &options, 0);
1244 if (*options && *options != ',') {
1245 printk(KERN_ERR "EXT4-fs: Invalid sb specification: %s\n",
1249 if (*options == ',')
1251 *data = (void *) options;
1256 #define DEFAULT_JOURNAL_IOPRIO (IOPRIO_PRIO_VALUE(IOPRIO_CLASS_BE, 3))
1257 static char deprecated_msg[] = "Mount option \"%s\" will be removed by %s\n"
1258 "Contact linux-ext4@vger.kernel.org if you think we should keep it.\n";
1261 static int set_qf_name(struct super_block *sb, int qtype, substring_t *args)
1263 struct ext4_sb_info *sbi = EXT4_SB(sb);
1267 if (sb_any_quota_loaded(sb) &&
1268 !sbi->s_qf_names[qtype]) {
1269 ext4_msg(sb, KERN_ERR,
1270 "Cannot change journaled "
1271 "quota options when quota turned on");
1274 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_QUOTA)) {
1275 ext4_msg(sb, KERN_ERR, "Cannot set journaled quota options "
1276 "when QUOTA feature is enabled");
1279 qname = match_strdup(args);
1281 ext4_msg(sb, KERN_ERR,
1282 "Not enough memory for storing quotafile name");
1285 if (sbi->s_qf_names[qtype]) {
1286 if (strcmp(sbi->s_qf_names[qtype], qname) == 0)
1289 ext4_msg(sb, KERN_ERR,
1290 "%s quota file already specified",
1294 if (strchr(qname, '/')) {
1295 ext4_msg(sb, KERN_ERR,
1296 "quotafile must be on filesystem root");
1299 sbi->s_qf_names[qtype] = qname;
1307 static int clear_qf_name(struct super_block *sb, int qtype)
1310 struct ext4_sb_info *sbi = EXT4_SB(sb);
1312 if (sb_any_quota_loaded(sb) &&
1313 sbi->s_qf_names[qtype]) {
1314 ext4_msg(sb, KERN_ERR, "Cannot change journaled quota options"
1315 " when quota turned on");
1318 kfree(sbi->s_qf_names[qtype]);
1319 sbi->s_qf_names[qtype] = NULL;
1324 #define MOPT_SET 0x0001
1325 #define MOPT_CLEAR 0x0002
1326 #define MOPT_NOSUPPORT 0x0004
1327 #define MOPT_EXPLICIT 0x0008
1328 #define MOPT_CLEAR_ERR 0x0010
1329 #define MOPT_GTE0 0x0020
1332 #define MOPT_QFMT 0x0040
1334 #define MOPT_Q MOPT_NOSUPPORT
1335 #define MOPT_QFMT MOPT_NOSUPPORT
1337 #define MOPT_DATAJ 0x0080
1338 #define MOPT_NO_EXT2 0x0100
1339 #define MOPT_NO_EXT3 0x0200
1340 #define MOPT_EXT4_ONLY (MOPT_NO_EXT2 | MOPT_NO_EXT3)
1341 #define MOPT_STRING 0x0400
1343 static const struct mount_opts {
1347 } ext4_mount_opts[] = {
1348 {Opt_minix_df, EXT4_MOUNT_MINIX_DF, MOPT_SET},
1349 {Opt_bsd_df, EXT4_MOUNT_MINIX_DF, MOPT_CLEAR},
1350 {Opt_grpid, EXT4_MOUNT_GRPID, MOPT_SET},
1351 {Opt_nogrpid, EXT4_MOUNT_GRPID, MOPT_CLEAR},
1352 {Opt_block_validity, EXT4_MOUNT_BLOCK_VALIDITY, MOPT_SET},
1353 {Opt_noblock_validity, EXT4_MOUNT_BLOCK_VALIDITY, MOPT_CLEAR},
1354 {Opt_dioread_nolock, EXT4_MOUNT_DIOREAD_NOLOCK,
1355 MOPT_EXT4_ONLY | MOPT_SET},
1356 {Opt_dioread_lock, EXT4_MOUNT_DIOREAD_NOLOCK,
1357 MOPT_EXT4_ONLY | MOPT_CLEAR},
1358 {Opt_discard, EXT4_MOUNT_DISCARD, MOPT_SET},
1359 {Opt_nodiscard, EXT4_MOUNT_DISCARD, MOPT_CLEAR},
1360 {Opt_delalloc, EXT4_MOUNT_DELALLOC,
1361 MOPT_EXT4_ONLY | MOPT_SET | MOPT_EXPLICIT},
1362 {Opt_nodelalloc, EXT4_MOUNT_DELALLOC,
1363 MOPT_EXT4_ONLY | MOPT_CLEAR},
1364 {Opt_journal_checksum, EXT4_MOUNT_JOURNAL_CHECKSUM,
1365 MOPT_EXT4_ONLY | MOPT_SET},
1366 {Opt_journal_async_commit, (EXT4_MOUNT_JOURNAL_ASYNC_COMMIT |
1367 EXT4_MOUNT_JOURNAL_CHECKSUM),
1368 MOPT_EXT4_ONLY | MOPT_SET},
1369 {Opt_noload, EXT4_MOUNT_NOLOAD, MOPT_NO_EXT2 | MOPT_SET},
1370 {Opt_err_panic, EXT4_MOUNT_ERRORS_PANIC, MOPT_SET | MOPT_CLEAR_ERR},
1371 {Opt_err_ro, EXT4_MOUNT_ERRORS_RO, MOPT_SET | MOPT_CLEAR_ERR},
1372 {Opt_err_cont, EXT4_MOUNT_ERRORS_CONT, MOPT_SET | MOPT_CLEAR_ERR},
1373 {Opt_data_err_abort, EXT4_MOUNT_DATA_ERR_ABORT,
1374 MOPT_NO_EXT2 | MOPT_SET},
1375 {Opt_data_err_ignore, EXT4_MOUNT_DATA_ERR_ABORT,
1376 MOPT_NO_EXT2 | MOPT_CLEAR},
1377 {Opt_barrier, EXT4_MOUNT_BARRIER, MOPT_SET},
1378 {Opt_nobarrier, EXT4_MOUNT_BARRIER, MOPT_CLEAR},
1379 {Opt_noauto_da_alloc, EXT4_MOUNT_NO_AUTO_DA_ALLOC, MOPT_SET},
1380 {Opt_auto_da_alloc, EXT4_MOUNT_NO_AUTO_DA_ALLOC, MOPT_CLEAR},
1381 {Opt_noinit_itable, EXT4_MOUNT_INIT_INODE_TABLE, MOPT_CLEAR},
1382 {Opt_commit, 0, MOPT_GTE0},
1383 {Opt_max_batch_time, 0, MOPT_GTE0},
1384 {Opt_min_batch_time, 0, MOPT_GTE0},
1385 {Opt_inode_readahead_blks, 0, MOPT_GTE0},
1386 {Opt_init_itable, 0, MOPT_GTE0},
1387 {Opt_stripe, 0, MOPT_GTE0},
1388 {Opt_resuid, 0, MOPT_GTE0},
1389 {Opt_resgid, 0, MOPT_GTE0},
1390 {Opt_journal_dev, 0, MOPT_GTE0},
1391 {Opt_journal_path, 0, MOPT_STRING},
1392 {Opt_journal_ioprio, 0, MOPT_GTE0},
1393 {Opt_data_journal, EXT4_MOUNT_JOURNAL_DATA, MOPT_NO_EXT2 | MOPT_DATAJ},
1394 {Opt_data_ordered, EXT4_MOUNT_ORDERED_DATA, MOPT_NO_EXT2 | MOPT_DATAJ},
1395 {Opt_data_writeback, EXT4_MOUNT_WRITEBACK_DATA,
1396 MOPT_NO_EXT2 | MOPT_DATAJ},
1397 {Opt_user_xattr, EXT4_MOUNT_XATTR_USER, MOPT_SET},
1398 {Opt_nouser_xattr, EXT4_MOUNT_XATTR_USER, MOPT_CLEAR},
1399 #ifdef CONFIG_EXT4_FS_POSIX_ACL
1400 {Opt_acl, EXT4_MOUNT_POSIX_ACL, MOPT_SET},
1401 {Opt_noacl, EXT4_MOUNT_POSIX_ACL, MOPT_CLEAR},
1403 {Opt_acl, 0, MOPT_NOSUPPORT},
1404 {Opt_noacl, 0, MOPT_NOSUPPORT},
1406 {Opt_nouid32, EXT4_MOUNT_NO_UID32, MOPT_SET},
1407 {Opt_debug, EXT4_MOUNT_DEBUG, MOPT_SET},
1408 {Opt_quota, EXT4_MOUNT_QUOTA | EXT4_MOUNT_USRQUOTA, MOPT_SET | MOPT_Q},
1409 {Opt_usrquota, EXT4_MOUNT_QUOTA | EXT4_MOUNT_USRQUOTA,
1411 {Opt_grpquota, EXT4_MOUNT_QUOTA | EXT4_MOUNT_GRPQUOTA,
1413 {Opt_noquota, (EXT4_MOUNT_QUOTA | EXT4_MOUNT_USRQUOTA |
1414 EXT4_MOUNT_GRPQUOTA), MOPT_CLEAR | MOPT_Q},
1415 {Opt_usrjquota, 0, MOPT_Q},
1416 {Opt_grpjquota, 0, MOPT_Q},
1417 {Opt_offusrjquota, 0, MOPT_Q},
1418 {Opt_offgrpjquota, 0, MOPT_Q},
1419 {Opt_jqfmt_vfsold, QFMT_VFS_OLD, MOPT_QFMT},
1420 {Opt_jqfmt_vfsv0, QFMT_VFS_V0, MOPT_QFMT},
1421 {Opt_jqfmt_vfsv1, QFMT_VFS_V1, MOPT_QFMT},
1422 {Opt_max_dir_size_kb, 0, MOPT_GTE0},
1426 static int handle_mount_opt(struct super_block *sb, char *opt, int token,
1427 substring_t *args, unsigned long *journal_devnum,
1428 unsigned int *journal_ioprio, int is_remount)
1430 struct ext4_sb_info *sbi = EXT4_SB(sb);
1431 const struct mount_opts *m;
1437 if (token == Opt_usrjquota)
1438 return set_qf_name(sb, USRQUOTA, &args[0]);
1439 else if (token == Opt_grpjquota)
1440 return set_qf_name(sb, GRPQUOTA, &args[0]);
1441 else if (token == Opt_offusrjquota)
1442 return clear_qf_name(sb, USRQUOTA);
1443 else if (token == Opt_offgrpjquota)
1444 return clear_qf_name(sb, GRPQUOTA);
1448 case Opt_nouser_xattr:
1449 ext4_msg(sb, KERN_WARNING, deprecated_msg, opt, "3.5");
1452 return 1; /* handled by get_sb_block() */
1454 ext4_msg(sb, KERN_WARNING, "Ignoring removed %s option", opt);
1457 sbi->s_mount_flags |= EXT4_MF_FS_ABORTED;
1460 sb->s_flags |= MS_I_VERSION;
1464 for (m = ext4_mount_opts; m->token != Opt_err; m++)
1465 if (token == m->token)
1468 if (m->token == Opt_err) {
1469 ext4_msg(sb, KERN_ERR, "Unrecognized mount option \"%s\" "
1470 "or missing value", opt);
1474 if ((m->flags & MOPT_NO_EXT2) && IS_EXT2_SB(sb)) {
1475 ext4_msg(sb, KERN_ERR,
1476 "Mount option \"%s\" incompatible with ext2", opt);
1479 if ((m->flags & MOPT_NO_EXT3) && IS_EXT3_SB(sb)) {
1480 ext4_msg(sb, KERN_ERR,
1481 "Mount option \"%s\" incompatible with ext3", opt);
1485 if (args->from && !(m->flags & MOPT_STRING) && match_int(args, &arg))
1487 if (args->from && (m->flags & MOPT_GTE0) && (arg < 0))
1489 if (m->flags & MOPT_EXPLICIT)
1490 set_opt2(sb, EXPLICIT_DELALLOC);
1491 if (m->flags & MOPT_CLEAR_ERR)
1492 clear_opt(sb, ERRORS_MASK);
1493 if (token == Opt_noquota && sb_any_quota_loaded(sb)) {
1494 ext4_msg(sb, KERN_ERR, "Cannot change quota "
1495 "options when quota turned on");
1499 if (m->flags & MOPT_NOSUPPORT) {
1500 ext4_msg(sb, KERN_ERR, "%s option not supported", opt);
1501 } else if (token == Opt_commit) {
1503 arg = JBD2_DEFAULT_MAX_COMMIT_AGE;
1504 sbi->s_commit_interval = HZ * arg;
1505 } else if (token == Opt_max_batch_time) {
1506 sbi->s_max_batch_time = arg;
1507 } else if (token == Opt_min_batch_time) {
1508 sbi->s_min_batch_time = arg;
1509 } else if (token == Opt_inode_readahead_blks) {
1510 if (arg && (arg > (1 << 30) || !is_power_of_2(arg))) {
1511 ext4_msg(sb, KERN_ERR,
1512 "EXT4-fs: inode_readahead_blks must be "
1513 "0 or a power of 2 smaller than 2^31");
1516 sbi->s_inode_readahead_blks = arg;
1517 } else if (token == Opt_init_itable) {
1518 set_opt(sb, INIT_INODE_TABLE);
1520 arg = EXT4_DEF_LI_WAIT_MULT;
1521 sbi->s_li_wait_mult = arg;
1522 } else if (token == Opt_max_dir_size_kb) {
1523 sbi->s_max_dir_size_kb = arg;
1524 } else if (token == Opt_stripe) {
1525 sbi->s_stripe = arg;
1526 } else if (token == Opt_resuid) {
1527 uid = make_kuid(current_user_ns(), arg);
1528 if (!uid_valid(uid)) {
1529 ext4_msg(sb, KERN_ERR, "Invalid uid value %d", arg);
1532 sbi->s_resuid = uid;
1533 } else if (token == Opt_resgid) {
1534 gid = make_kgid(current_user_ns(), arg);
1535 if (!gid_valid(gid)) {
1536 ext4_msg(sb, KERN_ERR, "Invalid gid value %d", arg);
1539 sbi->s_resgid = gid;
1540 } else if (token == Opt_journal_dev) {
1542 ext4_msg(sb, KERN_ERR,
1543 "Cannot specify journal on remount");
1546 *journal_devnum = arg;
1547 } else if (token == Opt_journal_path) {
1549 struct inode *journal_inode;
1554 ext4_msg(sb, KERN_ERR,
1555 "Cannot specify journal on remount");
1558 journal_path = match_strdup(&args[0]);
1559 if (!journal_path) {
1560 ext4_msg(sb, KERN_ERR, "error: could not dup "
1561 "journal device string");
1565 error = kern_path(journal_path, LOOKUP_FOLLOW, &path);
1567 ext4_msg(sb, KERN_ERR, "error: could not find "
1568 "journal device path: error %d", error);
1569 kfree(journal_path);
1573 journal_inode = path.dentry->d_inode;
1574 if (!S_ISBLK(journal_inode->i_mode)) {
1575 ext4_msg(sb, KERN_ERR, "error: journal path %s "
1576 "is not a block device", journal_path);
1578 kfree(journal_path);
1582 *journal_devnum = new_encode_dev(journal_inode->i_rdev);
1584 kfree(journal_path);
1585 } else if (token == Opt_journal_ioprio) {
1587 ext4_msg(sb, KERN_ERR, "Invalid journal IO priority"
1592 IOPRIO_PRIO_VALUE(IOPRIO_CLASS_BE, arg);
1593 } else if (m->flags & MOPT_DATAJ) {
1595 if (!sbi->s_journal)
1596 ext4_msg(sb, KERN_WARNING, "Remounting file system with no journal so ignoring journalled data option");
1597 else if (test_opt(sb, DATA_FLAGS) != m->mount_opt) {
1598 ext4_msg(sb, KERN_ERR,
1599 "Cannot change data mode on remount");
1603 clear_opt(sb, DATA_FLAGS);
1604 sbi->s_mount_opt |= m->mount_opt;
1607 } else if (m->flags & MOPT_QFMT) {
1608 if (sb_any_quota_loaded(sb) &&
1609 sbi->s_jquota_fmt != m->mount_opt) {
1610 ext4_msg(sb, KERN_ERR, "Cannot change journaled "
1611 "quota options when quota turned on");
1614 if (EXT4_HAS_RO_COMPAT_FEATURE(sb,
1615 EXT4_FEATURE_RO_COMPAT_QUOTA)) {
1616 ext4_msg(sb, KERN_ERR,
1617 "Cannot set journaled quota options "
1618 "when QUOTA feature is enabled");
1621 sbi->s_jquota_fmt = m->mount_opt;
1626 if (m->flags & MOPT_CLEAR)
1628 else if (unlikely(!(m->flags & MOPT_SET))) {
1629 ext4_msg(sb, KERN_WARNING,
1630 "buggy handling of option %s", opt);
1635 sbi->s_mount_opt |= m->mount_opt;
1637 sbi->s_mount_opt &= ~m->mount_opt;
1642 static int parse_options(char *options, struct super_block *sb,
1643 unsigned long *journal_devnum,
1644 unsigned int *journal_ioprio,
1647 struct ext4_sb_info *sbi = EXT4_SB(sb);
1649 substring_t args[MAX_OPT_ARGS];
1655 while ((p = strsep(&options, ",")) != NULL) {
1659 * Initialize args struct so we know whether arg was
1660 * found; some options take optional arguments.
1662 args[0].to = args[0].from = NULL;
1663 token = match_token(p, tokens, args);
1664 if (handle_mount_opt(sb, p, token, args, journal_devnum,
1665 journal_ioprio, is_remount) < 0)
1669 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_QUOTA) &&
1670 (test_opt(sb, USRQUOTA) || test_opt(sb, GRPQUOTA))) {
1671 ext4_msg(sb, KERN_ERR, "Cannot set quota options when QUOTA "
1672 "feature is enabled");
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);
1679 if (test_opt(sb, GRPQUOTA) && sbi->s_qf_names[GRPQUOTA])
1680 clear_opt(sb, GRPQUOTA);
1682 if (test_opt(sb, GRPQUOTA) || test_opt(sb, USRQUOTA)) {
1683 ext4_msg(sb, KERN_ERR, "old and new quota "
1688 if (!sbi->s_jquota_fmt) {
1689 ext4_msg(sb, KERN_ERR, "journaled quota format "
1695 if (test_opt(sb, DIOREAD_NOLOCK)) {
1697 BLOCK_SIZE << le32_to_cpu(sbi->s_es->s_log_block_size);
1699 if (blocksize < PAGE_CACHE_SIZE) {
1700 ext4_msg(sb, KERN_ERR, "can't mount with "
1701 "dioread_nolock if block size != PAGE_SIZE");
1708 static inline void ext4_show_quota_options(struct seq_file *seq,
1709 struct super_block *sb)
1711 #if defined(CONFIG_QUOTA)
1712 struct ext4_sb_info *sbi = EXT4_SB(sb);
1714 if (sbi->s_jquota_fmt) {
1717 switch (sbi->s_jquota_fmt) {
1728 seq_printf(seq, ",jqfmt=%s", fmtname);
1731 if (sbi->s_qf_names[USRQUOTA])
1732 seq_printf(seq, ",usrjquota=%s", sbi->s_qf_names[USRQUOTA]);
1734 if (sbi->s_qf_names[GRPQUOTA])
1735 seq_printf(seq, ",grpjquota=%s", sbi->s_qf_names[GRPQUOTA]);
1739 static const char *token2str(int token)
1741 const struct match_token *t;
1743 for (t = tokens; t->token != Opt_err; t++)
1744 if (t->token == token && !strchr(t->pattern, '='))
1751 * - it's set to a non-default value OR
1752 * - if the per-sb default is different from the global default
1754 static int _ext4_show_options(struct seq_file *seq, struct super_block *sb,
1757 struct ext4_sb_info *sbi = EXT4_SB(sb);
1758 struct ext4_super_block *es = sbi->s_es;
1759 int def_errors, def_mount_opt = nodefs ? 0 : sbi->s_def_mount_opt;
1760 const struct mount_opts *m;
1761 char sep = nodefs ? '\n' : ',';
1763 #define SEQ_OPTS_PUTS(str) seq_printf(seq, "%c" str, sep)
1764 #define SEQ_OPTS_PRINT(str, arg) seq_printf(seq, "%c" str, sep, arg)
1766 if (sbi->s_sb_block != 1)
1767 SEQ_OPTS_PRINT("sb=%llu", sbi->s_sb_block);
1769 for (m = ext4_mount_opts; m->token != Opt_err; m++) {
1770 int want_set = m->flags & MOPT_SET;
1771 if (((m->flags & (MOPT_SET|MOPT_CLEAR)) == 0) ||
1772 (m->flags & MOPT_CLEAR_ERR))
1774 if (!(m->mount_opt & (sbi->s_mount_opt ^ def_mount_opt)))
1775 continue; /* skip if same as the default */
1777 (sbi->s_mount_opt & m->mount_opt) != m->mount_opt) ||
1778 (!want_set && (sbi->s_mount_opt & m->mount_opt)))
1779 continue; /* select Opt_noFoo vs Opt_Foo */
1780 SEQ_OPTS_PRINT("%s", token2str(m->token));
1783 if (nodefs || !uid_eq(sbi->s_resuid, make_kuid(&init_user_ns, EXT4_DEF_RESUID)) ||
1784 le16_to_cpu(es->s_def_resuid) != EXT4_DEF_RESUID)
1785 SEQ_OPTS_PRINT("resuid=%u",
1786 from_kuid_munged(&init_user_ns, sbi->s_resuid));
1787 if (nodefs || !gid_eq(sbi->s_resgid, make_kgid(&init_user_ns, EXT4_DEF_RESGID)) ||
1788 le16_to_cpu(es->s_def_resgid) != EXT4_DEF_RESGID)
1789 SEQ_OPTS_PRINT("resgid=%u",
1790 from_kgid_munged(&init_user_ns, sbi->s_resgid));
1791 def_errors = nodefs ? -1 : le16_to_cpu(es->s_errors);
1792 if (test_opt(sb, ERRORS_RO) && def_errors != EXT4_ERRORS_RO)
1793 SEQ_OPTS_PUTS("errors=remount-ro");
1794 if (test_opt(sb, ERRORS_CONT) && def_errors != EXT4_ERRORS_CONTINUE)
1795 SEQ_OPTS_PUTS("errors=continue");
1796 if (test_opt(sb, ERRORS_PANIC) && def_errors != EXT4_ERRORS_PANIC)
1797 SEQ_OPTS_PUTS("errors=panic");
1798 if (nodefs || sbi->s_commit_interval != JBD2_DEFAULT_MAX_COMMIT_AGE*HZ)
1799 SEQ_OPTS_PRINT("commit=%lu", sbi->s_commit_interval / HZ);
1800 if (nodefs || sbi->s_min_batch_time != EXT4_DEF_MIN_BATCH_TIME)
1801 SEQ_OPTS_PRINT("min_batch_time=%u", sbi->s_min_batch_time);
1802 if (nodefs || sbi->s_max_batch_time != EXT4_DEF_MAX_BATCH_TIME)
1803 SEQ_OPTS_PRINT("max_batch_time=%u", sbi->s_max_batch_time);
1804 if (sb->s_flags & MS_I_VERSION)
1805 SEQ_OPTS_PUTS("i_version");
1806 if (nodefs || sbi->s_stripe)
1807 SEQ_OPTS_PRINT("stripe=%lu", sbi->s_stripe);
1808 if (EXT4_MOUNT_DATA_FLAGS & (sbi->s_mount_opt ^ def_mount_opt)) {
1809 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA)
1810 SEQ_OPTS_PUTS("data=journal");
1811 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA)
1812 SEQ_OPTS_PUTS("data=ordered");
1813 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_WRITEBACK_DATA)
1814 SEQ_OPTS_PUTS("data=writeback");
1817 sbi->s_inode_readahead_blks != EXT4_DEF_INODE_READAHEAD_BLKS)
1818 SEQ_OPTS_PRINT("inode_readahead_blks=%u",
1819 sbi->s_inode_readahead_blks);
1821 if (nodefs || (test_opt(sb, INIT_INODE_TABLE) &&
1822 (sbi->s_li_wait_mult != EXT4_DEF_LI_WAIT_MULT)))
1823 SEQ_OPTS_PRINT("init_itable=%u", sbi->s_li_wait_mult);
1824 if (nodefs || sbi->s_max_dir_size_kb)
1825 SEQ_OPTS_PRINT("max_dir_size_kb=%u", sbi->s_max_dir_size_kb);
1827 ext4_show_quota_options(seq, sb);
1831 static int ext4_show_options(struct seq_file *seq, struct dentry *root)
1833 return _ext4_show_options(seq, root->d_sb, 0);
1836 static int options_seq_show(struct seq_file *seq, void *offset)
1838 struct super_block *sb = seq->private;
1841 seq_puts(seq, (sb->s_flags & MS_RDONLY) ? "ro" : "rw");
1842 rc = _ext4_show_options(seq, sb, 1);
1843 seq_puts(seq, "\n");
1847 static int options_open_fs(struct inode *inode, struct file *file)
1849 return single_open(file, options_seq_show, PDE_DATA(inode));
1852 static const struct file_operations ext4_seq_options_fops = {
1853 .owner = THIS_MODULE,
1854 .open = options_open_fs,
1856 .llseek = seq_lseek,
1857 .release = single_release,
1860 static int ext4_setup_super(struct super_block *sb, struct ext4_super_block *es,
1863 struct ext4_sb_info *sbi = EXT4_SB(sb);
1866 if (le32_to_cpu(es->s_rev_level) > EXT4_MAX_SUPP_REV) {
1867 ext4_msg(sb, KERN_ERR, "revision level too high, "
1868 "forcing read-only mode");
1873 if (!(sbi->s_mount_state & EXT4_VALID_FS))
1874 ext4_msg(sb, KERN_WARNING, "warning: mounting unchecked fs, "
1875 "running e2fsck is recommended");
1876 else if (sbi->s_mount_state & EXT4_ERROR_FS)
1877 ext4_msg(sb, KERN_WARNING,
1878 "warning: mounting fs with errors, "
1879 "running e2fsck is recommended");
1880 else if ((__s16) le16_to_cpu(es->s_max_mnt_count) > 0 &&
1881 le16_to_cpu(es->s_mnt_count) >=
1882 (unsigned short) (__s16) le16_to_cpu(es->s_max_mnt_count))
1883 ext4_msg(sb, KERN_WARNING,
1884 "warning: maximal mount count reached, "
1885 "running e2fsck is recommended");
1886 else if (le32_to_cpu(es->s_checkinterval) &&
1887 (le32_to_cpu(es->s_lastcheck) +
1888 le32_to_cpu(es->s_checkinterval) <= get_seconds()))
1889 ext4_msg(sb, KERN_WARNING,
1890 "warning: checktime reached, "
1891 "running e2fsck is recommended");
1892 if (!sbi->s_journal)
1893 es->s_state &= cpu_to_le16(~EXT4_VALID_FS);
1894 if (!(__s16) le16_to_cpu(es->s_max_mnt_count))
1895 es->s_max_mnt_count = cpu_to_le16(EXT4_DFL_MAX_MNT_COUNT);
1896 le16_add_cpu(&es->s_mnt_count, 1);
1897 es->s_mtime = cpu_to_le32(get_seconds());
1898 ext4_update_dynamic_rev(sb);
1900 EXT4_SET_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
1902 ext4_commit_super(sb, 1);
1904 if (test_opt(sb, DEBUG))
1905 printk(KERN_INFO "[EXT4 FS bs=%lu, gc=%u, "
1906 "bpg=%lu, ipg=%lu, mo=%04x, mo2=%04x]\n",
1908 sbi->s_groups_count,
1909 EXT4_BLOCKS_PER_GROUP(sb),
1910 EXT4_INODES_PER_GROUP(sb),
1911 sbi->s_mount_opt, sbi->s_mount_opt2);
1913 cleancache_init_fs(sb);
1917 int ext4_alloc_flex_bg_array(struct super_block *sb, ext4_group_t ngroup)
1919 struct ext4_sb_info *sbi = EXT4_SB(sb);
1920 struct flex_groups *new_groups;
1923 if (!sbi->s_log_groups_per_flex)
1926 size = ext4_flex_group(sbi, ngroup - 1) + 1;
1927 if (size <= sbi->s_flex_groups_allocated)
1930 size = roundup_pow_of_two(size * sizeof(struct flex_groups));
1931 new_groups = ext4_kvzalloc(size, GFP_KERNEL);
1933 ext4_msg(sb, KERN_ERR, "not enough memory for %d flex groups",
1934 size / (int) sizeof(struct flex_groups));
1938 if (sbi->s_flex_groups) {
1939 memcpy(new_groups, sbi->s_flex_groups,
1940 (sbi->s_flex_groups_allocated *
1941 sizeof(struct flex_groups)));
1942 ext4_kvfree(sbi->s_flex_groups);
1944 sbi->s_flex_groups = new_groups;
1945 sbi->s_flex_groups_allocated = size / sizeof(struct flex_groups);
1949 static int ext4_fill_flex_info(struct super_block *sb)
1951 struct ext4_sb_info *sbi = EXT4_SB(sb);
1952 struct ext4_group_desc *gdp = NULL;
1953 ext4_group_t flex_group;
1956 sbi->s_log_groups_per_flex = sbi->s_es->s_log_groups_per_flex;
1957 if (sbi->s_log_groups_per_flex < 1 || sbi->s_log_groups_per_flex > 31) {
1958 sbi->s_log_groups_per_flex = 0;
1962 err = ext4_alloc_flex_bg_array(sb, sbi->s_groups_count);
1966 for (i = 0; i < sbi->s_groups_count; i++) {
1967 gdp = ext4_get_group_desc(sb, i, NULL);
1969 flex_group = ext4_flex_group(sbi, i);
1970 atomic_add(ext4_free_inodes_count(sb, gdp),
1971 &sbi->s_flex_groups[flex_group].free_inodes);
1972 atomic64_add(ext4_free_group_clusters(sb, gdp),
1973 &sbi->s_flex_groups[flex_group].free_clusters);
1974 atomic_add(ext4_used_dirs_count(sb, gdp),
1975 &sbi->s_flex_groups[flex_group].used_dirs);
1983 static __le16 ext4_group_desc_csum(struct ext4_sb_info *sbi, __u32 block_group,
1984 struct ext4_group_desc *gdp)
1988 __le32 le_group = cpu_to_le32(block_group);
1990 if (ext4_has_metadata_csum(sbi->s_sb)) {
1991 /* Use new metadata_csum algorithm */
1995 save_csum = gdp->bg_checksum;
1996 gdp->bg_checksum = 0;
1997 csum32 = ext4_chksum(sbi, sbi->s_csum_seed, (__u8 *)&le_group,
1999 csum32 = ext4_chksum(sbi, csum32, (__u8 *)gdp,
2001 gdp->bg_checksum = save_csum;
2003 crc = csum32 & 0xFFFF;
2007 /* old crc16 code */
2008 offset = offsetof(struct ext4_group_desc, bg_checksum);
2010 crc = crc16(~0, sbi->s_es->s_uuid, sizeof(sbi->s_es->s_uuid));
2011 crc = crc16(crc, (__u8 *)&le_group, sizeof(le_group));
2012 crc = crc16(crc, (__u8 *)gdp, offset);
2013 offset += sizeof(gdp->bg_checksum); /* skip checksum */
2014 /* for checksum of struct ext4_group_desc do the rest...*/
2015 if ((sbi->s_es->s_feature_incompat &
2016 cpu_to_le32(EXT4_FEATURE_INCOMPAT_64BIT)) &&
2017 offset < le16_to_cpu(sbi->s_es->s_desc_size))
2018 crc = crc16(crc, (__u8 *)gdp + offset,
2019 le16_to_cpu(sbi->s_es->s_desc_size) -
2023 return cpu_to_le16(crc);
2026 int ext4_group_desc_csum_verify(struct super_block *sb, __u32 block_group,
2027 struct ext4_group_desc *gdp)
2029 if (ext4_has_group_desc_csum(sb) &&
2030 (gdp->bg_checksum != ext4_group_desc_csum(EXT4_SB(sb),
2037 void ext4_group_desc_csum_set(struct super_block *sb, __u32 block_group,
2038 struct ext4_group_desc *gdp)
2040 if (!ext4_has_group_desc_csum(sb))
2042 gdp->bg_checksum = ext4_group_desc_csum(EXT4_SB(sb), block_group, gdp);
2045 /* Called at mount-time, super-block is locked */
2046 static int ext4_check_descriptors(struct super_block *sb,
2047 ext4_group_t *first_not_zeroed)
2049 struct ext4_sb_info *sbi = EXT4_SB(sb);
2050 ext4_fsblk_t first_block = le32_to_cpu(sbi->s_es->s_first_data_block);
2051 ext4_fsblk_t last_block;
2052 ext4_fsblk_t block_bitmap;
2053 ext4_fsblk_t inode_bitmap;
2054 ext4_fsblk_t inode_table;
2055 int flexbg_flag = 0;
2056 ext4_group_t i, grp = sbi->s_groups_count;
2058 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_FLEX_BG))
2061 ext4_debug("Checking group descriptors");
2063 for (i = 0; i < sbi->s_groups_count; i++) {
2064 struct ext4_group_desc *gdp = ext4_get_group_desc(sb, i, NULL);
2066 if (i == sbi->s_groups_count - 1 || flexbg_flag)
2067 last_block = ext4_blocks_count(sbi->s_es) - 1;
2069 last_block = first_block +
2070 (EXT4_BLOCKS_PER_GROUP(sb) - 1);
2072 if ((grp == sbi->s_groups_count) &&
2073 !(gdp->bg_flags & cpu_to_le16(EXT4_BG_INODE_ZEROED)))
2076 block_bitmap = ext4_block_bitmap(sb, gdp);
2077 if (block_bitmap < first_block || block_bitmap > last_block) {
2078 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2079 "Block bitmap for group %u not in group "
2080 "(block %llu)!", i, block_bitmap);
2083 inode_bitmap = ext4_inode_bitmap(sb, gdp);
2084 if (inode_bitmap < first_block || inode_bitmap > last_block) {
2085 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2086 "Inode bitmap for group %u not in group "
2087 "(block %llu)!", i, inode_bitmap);
2090 inode_table = ext4_inode_table(sb, gdp);
2091 if (inode_table < first_block ||
2092 inode_table + sbi->s_itb_per_group - 1 > last_block) {
2093 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2094 "Inode table for group %u not in group "
2095 "(block %llu)!", i, inode_table);
2098 ext4_lock_group(sb, i);
2099 if (!ext4_group_desc_csum_verify(sb, i, gdp)) {
2100 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2101 "Checksum for group %u failed (%u!=%u)",
2102 i, le16_to_cpu(ext4_group_desc_csum(sbi, i,
2103 gdp)), le16_to_cpu(gdp->bg_checksum));
2104 if (!(sb->s_flags & MS_RDONLY)) {
2105 ext4_unlock_group(sb, i);
2109 ext4_unlock_group(sb, i);
2111 first_block += EXT4_BLOCKS_PER_GROUP(sb);
2113 if (NULL != first_not_zeroed)
2114 *first_not_zeroed = grp;
2118 /* ext4_orphan_cleanup() walks a singly-linked list of inodes (starting at
2119 * the superblock) which were deleted from all directories, but held open by
2120 * a process at the time of a crash. We walk the list and try to delete these
2121 * inodes at recovery time (only with a read-write filesystem).
2123 * In order to keep the orphan inode chain consistent during traversal (in
2124 * case of crash during recovery), we link each inode into the superblock
2125 * orphan list_head and handle it the same way as an inode deletion during
2126 * normal operation (which journals the operations for us).
2128 * We only do an iget() and an iput() on each inode, which is very safe if we
2129 * accidentally point at an in-use or already deleted inode. The worst that
2130 * can happen in this case is that we get a "bit already cleared" message from
2131 * ext4_free_inode(). The only reason we would point at a wrong inode is if
2132 * e2fsck was run on this filesystem, and it must have already done the orphan
2133 * inode cleanup for us, so we can safely abort without any further action.
2135 static void ext4_orphan_cleanup(struct super_block *sb,
2136 struct ext4_super_block *es)
2138 unsigned int s_flags = sb->s_flags;
2139 int nr_orphans = 0, nr_truncates = 0;
2143 if (!es->s_last_orphan) {
2144 jbd_debug(4, "no orphan inodes to clean up\n");
2148 if (bdev_read_only(sb->s_bdev)) {
2149 ext4_msg(sb, KERN_ERR, "write access "
2150 "unavailable, skipping orphan cleanup");
2154 /* Check if feature set would not allow a r/w mount */
2155 if (!ext4_feature_set_ok(sb, 0)) {
2156 ext4_msg(sb, KERN_INFO, "Skipping orphan cleanup due to "
2157 "unknown ROCOMPAT features");
2161 if (EXT4_SB(sb)->s_mount_state & EXT4_ERROR_FS) {
2162 /* don't clear list on RO mount w/ errors */
2163 if (es->s_last_orphan && !(s_flags & MS_RDONLY)) {
2164 ext4_msg(sb, KERN_INFO, "Errors on filesystem, "
2165 "clearing orphan list.\n");
2166 es->s_last_orphan = 0;
2168 jbd_debug(1, "Skipping orphan recovery on fs with errors.\n");
2172 if (s_flags & MS_RDONLY) {
2173 ext4_msg(sb, KERN_INFO, "orphan cleanup on readonly fs");
2174 sb->s_flags &= ~MS_RDONLY;
2177 /* Needed for iput() to work correctly and not trash data */
2178 sb->s_flags |= MS_ACTIVE;
2179 /* Turn on quotas so that they are updated correctly */
2180 for (i = 0; i < EXT4_MAXQUOTAS; i++) {
2181 if (EXT4_SB(sb)->s_qf_names[i]) {
2182 int ret = ext4_quota_on_mount(sb, i);
2184 ext4_msg(sb, KERN_ERR,
2185 "Cannot turn on journaled "
2186 "quota: error %d", ret);
2191 while (es->s_last_orphan) {
2192 struct inode *inode;
2194 inode = ext4_orphan_get(sb, le32_to_cpu(es->s_last_orphan));
2195 if (IS_ERR(inode)) {
2196 es->s_last_orphan = 0;
2200 list_add(&EXT4_I(inode)->i_orphan, &EXT4_SB(sb)->s_orphan);
2201 dquot_initialize(inode);
2202 if (inode->i_nlink) {
2203 if (test_opt(sb, DEBUG))
2204 ext4_msg(sb, KERN_DEBUG,
2205 "%s: truncating inode %lu to %lld bytes",
2206 __func__, inode->i_ino, inode->i_size);
2207 jbd_debug(2, "truncating inode %lu to %lld bytes\n",
2208 inode->i_ino, inode->i_size);
2209 mutex_lock(&inode->i_mutex);
2210 truncate_inode_pages(inode->i_mapping, inode->i_size);
2211 ext4_truncate(inode);
2212 mutex_unlock(&inode->i_mutex);
2215 if (test_opt(sb, DEBUG))
2216 ext4_msg(sb, KERN_DEBUG,
2217 "%s: deleting unreferenced inode %lu",
2218 __func__, inode->i_ino);
2219 jbd_debug(2, "deleting unreferenced inode %lu\n",
2223 iput(inode); /* The delete magic happens here! */
2226 #define PLURAL(x) (x), ((x) == 1) ? "" : "s"
2229 ext4_msg(sb, KERN_INFO, "%d orphan inode%s deleted",
2230 PLURAL(nr_orphans));
2232 ext4_msg(sb, KERN_INFO, "%d truncate%s cleaned up",
2233 PLURAL(nr_truncates));
2235 /* Turn quotas off */
2236 for (i = 0; i < EXT4_MAXQUOTAS; i++) {
2237 if (sb_dqopt(sb)->files[i])
2238 dquot_quota_off(sb, i);
2241 sb->s_flags = s_flags; /* Restore MS_RDONLY status */
2245 * Maximal extent format file size.
2246 * Resulting logical blkno at s_maxbytes must fit in our on-disk
2247 * extent format containers, within a sector_t, and within i_blocks
2248 * in the vfs. ext4 inode has 48 bits of i_block in fsblock units,
2249 * so that won't be a limiting factor.
2251 * However there is other limiting factor. We do store extents in the form
2252 * of starting block and length, hence the resulting length of the extent
2253 * covering maximum file size must fit into on-disk format containers as
2254 * well. Given that length is always by 1 unit bigger than max unit (because
2255 * we count 0 as well) we have to lower the s_maxbytes by one fs block.
2257 * Note, this does *not* consider any metadata overhead for vfs i_blocks.
2259 static loff_t ext4_max_size(int blkbits, int has_huge_files)
2262 loff_t upper_limit = MAX_LFS_FILESIZE;
2264 /* small i_blocks in vfs inode? */
2265 if (!has_huge_files || sizeof(blkcnt_t) < sizeof(u64)) {
2267 * CONFIG_LBDAF is not enabled implies the inode
2268 * i_block represent total blocks in 512 bytes
2269 * 32 == size of vfs inode i_blocks * 8
2271 upper_limit = (1LL << 32) - 1;
2273 /* total blocks in file system block size */
2274 upper_limit >>= (blkbits - 9);
2275 upper_limit <<= blkbits;
2279 * 32-bit extent-start container, ee_block. We lower the maxbytes
2280 * by one fs block, so ee_len can cover the extent of maximum file
2283 res = (1LL << 32) - 1;
2286 /* Sanity check against vm- & vfs- imposed limits */
2287 if (res > upper_limit)
2294 * Maximal bitmap file size. There is a direct, and {,double-,triple-}indirect
2295 * block limit, and also a limit of (2^48 - 1) 512-byte sectors in i_blocks.
2296 * We need to be 1 filesystem block less than the 2^48 sector limit.
2298 static loff_t ext4_max_bitmap_size(int bits, int has_huge_files)
2300 loff_t res = EXT4_NDIR_BLOCKS;
2303 /* This is calculated to be the largest file size for a dense, block
2304 * mapped file such that the file's total number of 512-byte sectors,
2305 * including data and all indirect blocks, does not exceed (2^48 - 1).
2307 * __u32 i_blocks_lo and _u16 i_blocks_high represent the total
2308 * number of 512-byte sectors of the file.
2311 if (!has_huge_files || sizeof(blkcnt_t) < sizeof(u64)) {
2313 * !has_huge_files or CONFIG_LBDAF not enabled implies that
2314 * the inode i_block field represents total file blocks in
2315 * 2^32 512-byte sectors == size of vfs inode i_blocks * 8
2317 upper_limit = (1LL << 32) - 1;
2319 /* total blocks in file system block size */
2320 upper_limit >>= (bits - 9);
2324 * We use 48 bit ext4_inode i_blocks
2325 * With EXT4_HUGE_FILE_FL set the i_blocks
2326 * represent total number of blocks in
2327 * file system block size
2329 upper_limit = (1LL << 48) - 1;
2333 /* indirect blocks */
2335 /* double indirect blocks */
2336 meta_blocks += 1 + (1LL << (bits-2));
2337 /* tripple indirect blocks */
2338 meta_blocks += 1 + (1LL << (bits-2)) + (1LL << (2*(bits-2)));
2340 upper_limit -= meta_blocks;
2341 upper_limit <<= bits;
2343 res += 1LL << (bits-2);
2344 res += 1LL << (2*(bits-2));
2345 res += 1LL << (3*(bits-2));
2347 if (res > upper_limit)
2350 if (res > MAX_LFS_FILESIZE)
2351 res = MAX_LFS_FILESIZE;
2356 static ext4_fsblk_t descriptor_loc(struct super_block *sb,
2357 ext4_fsblk_t logical_sb_block, int nr)
2359 struct ext4_sb_info *sbi = EXT4_SB(sb);
2360 ext4_group_t bg, first_meta_bg;
2363 first_meta_bg = le32_to_cpu(sbi->s_es->s_first_meta_bg);
2365 if (!EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_META_BG) ||
2367 return logical_sb_block + nr + 1;
2368 bg = sbi->s_desc_per_block * nr;
2369 if (ext4_bg_has_super(sb, bg))
2373 * If we have a meta_bg fs with 1k blocks, group 0's GDT is at
2374 * block 2, not 1. If s_first_data_block == 0 (bigalloc is enabled
2375 * on modern mke2fs or blksize > 1k on older mke2fs) then we must
2378 if (sb->s_blocksize == 1024 && nr == 0 &&
2379 le32_to_cpu(EXT4_SB(sb)->s_es->s_first_data_block) == 0)
2382 return (has_super + ext4_group_first_block_no(sb, bg));
2386 * ext4_get_stripe_size: Get the stripe size.
2387 * @sbi: In memory super block info
2389 * If we have specified it via mount option, then
2390 * use the mount option value. If the value specified at mount time is
2391 * greater than the blocks per group use the super block value.
2392 * If the super block value is greater than blocks per group return 0.
2393 * Allocator needs it be less than blocks per group.
2396 static unsigned long ext4_get_stripe_size(struct ext4_sb_info *sbi)
2398 unsigned long stride = le16_to_cpu(sbi->s_es->s_raid_stride);
2399 unsigned long stripe_width =
2400 le32_to_cpu(sbi->s_es->s_raid_stripe_width);
2403 if (sbi->s_stripe && sbi->s_stripe <= sbi->s_blocks_per_group)
2404 ret = sbi->s_stripe;
2405 else if (stripe_width <= sbi->s_blocks_per_group)
2407 else if (stride <= sbi->s_blocks_per_group)
2413 * If the stripe width is 1, this makes no sense and
2414 * we set it to 0 to turn off stripe handling code.
2425 struct attribute attr;
2426 ssize_t (*show)(struct ext4_attr *, struct ext4_sb_info *, char *);
2427 ssize_t (*store)(struct ext4_attr *, struct ext4_sb_info *,
2428 const char *, size_t);
2435 static int parse_strtoull(const char *buf,
2436 unsigned long long max, unsigned long long *value)
2440 ret = kstrtoull(skip_spaces(buf), 0, value);
2441 if (!ret && *value > max)
2446 static ssize_t delayed_allocation_blocks_show(struct ext4_attr *a,
2447 struct ext4_sb_info *sbi,
2450 return snprintf(buf, PAGE_SIZE, "%llu\n",
2452 percpu_counter_sum(&sbi->s_dirtyclusters_counter)));
2455 static ssize_t session_write_kbytes_show(struct ext4_attr *a,
2456 struct ext4_sb_info *sbi, char *buf)
2458 struct super_block *sb = sbi->s_buddy_cache->i_sb;
2460 if (!sb->s_bdev->bd_part)
2461 return snprintf(buf, PAGE_SIZE, "0\n");
2462 return snprintf(buf, PAGE_SIZE, "%lu\n",
2463 (part_stat_read(sb->s_bdev->bd_part, sectors[1]) -
2464 sbi->s_sectors_written_start) >> 1);
2467 static ssize_t lifetime_write_kbytes_show(struct ext4_attr *a,
2468 struct ext4_sb_info *sbi, char *buf)
2470 struct super_block *sb = sbi->s_buddy_cache->i_sb;
2472 if (!sb->s_bdev->bd_part)
2473 return snprintf(buf, PAGE_SIZE, "0\n");
2474 return snprintf(buf, PAGE_SIZE, "%llu\n",
2475 (unsigned long long)(sbi->s_kbytes_written +
2476 ((part_stat_read(sb->s_bdev->bd_part, sectors[1]) -
2477 EXT4_SB(sb)->s_sectors_written_start) >> 1)));
2480 static ssize_t inode_readahead_blks_store(struct ext4_attr *a,
2481 struct ext4_sb_info *sbi,
2482 const char *buf, size_t count)
2487 ret = kstrtoul(skip_spaces(buf), 0, &t);
2491 if (t && (!is_power_of_2(t) || t > 0x40000000))
2494 sbi->s_inode_readahead_blks = t;
2498 static ssize_t sbi_ui_show(struct ext4_attr *a,
2499 struct ext4_sb_info *sbi, char *buf)
2501 unsigned int *ui = (unsigned int *) (((char *) sbi) + a->u.offset);
2503 return snprintf(buf, PAGE_SIZE, "%u\n", *ui);
2506 static ssize_t sbi_ui_store(struct ext4_attr *a,
2507 struct ext4_sb_info *sbi,
2508 const char *buf, size_t count)
2510 unsigned int *ui = (unsigned int *) (((char *) sbi) + a->u.offset);
2514 ret = kstrtoul(skip_spaces(buf), 0, &t);
2521 static ssize_t es_ui_show(struct ext4_attr *a,
2522 struct ext4_sb_info *sbi, char *buf)
2525 unsigned int *ui = (unsigned int *) (((char *) sbi->s_es) +
2528 return snprintf(buf, PAGE_SIZE, "%u\n", *ui);
2531 static ssize_t reserved_clusters_show(struct ext4_attr *a,
2532 struct ext4_sb_info *sbi, char *buf)
2534 return snprintf(buf, PAGE_SIZE, "%llu\n",
2535 (unsigned long long) atomic64_read(&sbi->s_resv_clusters));
2538 static ssize_t reserved_clusters_store(struct ext4_attr *a,
2539 struct ext4_sb_info *sbi,
2540 const char *buf, size_t count)
2542 unsigned long long val;
2545 if (parse_strtoull(buf, -1ULL, &val))
2547 ret = ext4_reserve_clusters(sbi, val);
2549 return ret ? ret : count;
2552 static ssize_t trigger_test_error(struct ext4_attr *a,
2553 struct ext4_sb_info *sbi,
2554 const char *buf, size_t count)
2558 if (!capable(CAP_SYS_ADMIN))
2561 if (len && buf[len-1] == '\n')
2565 ext4_error(sbi->s_sb, "%.*s", len, buf);
2569 static ssize_t sbi_deprecated_show(struct ext4_attr *a,
2570 struct ext4_sb_info *sbi, char *buf)
2572 return snprintf(buf, PAGE_SIZE, "%d\n", a->u.deprecated_val);
2575 #define EXT4_ATTR_OFFSET(_name,_mode,_show,_store,_elname) \
2576 static struct ext4_attr ext4_attr_##_name = { \
2577 .attr = {.name = __stringify(_name), .mode = _mode }, \
2581 .offset = offsetof(struct ext4_sb_info, _elname),\
2585 #define EXT4_ATTR_OFFSET_ES(_name,_mode,_show,_store,_elname) \
2586 static struct ext4_attr ext4_attr_##_name = { \
2587 .attr = {.name = __stringify(_name), .mode = _mode }, \
2591 .offset = offsetof(struct ext4_super_block, _elname), \
2595 #define EXT4_ATTR(name, mode, show, store) \
2596 static struct ext4_attr ext4_attr_##name = __ATTR(name, mode, show, store)
2598 #define EXT4_INFO_ATTR(name) EXT4_ATTR(name, 0444, NULL, NULL)
2599 #define EXT4_RO_ATTR(name) EXT4_ATTR(name, 0444, name##_show, NULL)
2600 #define EXT4_RW_ATTR(name) EXT4_ATTR(name, 0644, name##_show, name##_store)
2602 #define EXT4_RO_ATTR_ES_UI(name, elname) \
2603 EXT4_ATTR_OFFSET_ES(name, 0444, es_ui_show, NULL, elname)
2604 #define EXT4_RW_ATTR_SBI_UI(name, elname) \
2605 EXT4_ATTR_OFFSET(name, 0644, sbi_ui_show, sbi_ui_store, elname)
2607 #define ATTR_LIST(name) &ext4_attr_##name.attr
2608 #define EXT4_DEPRECATED_ATTR(_name, _val) \
2609 static struct ext4_attr ext4_attr_##_name = { \
2610 .attr = {.name = __stringify(_name), .mode = 0444 }, \
2611 .show = sbi_deprecated_show, \
2613 .deprecated_val = _val, \
2617 EXT4_RO_ATTR(delayed_allocation_blocks);
2618 EXT4_RO_ATTR(session_write_kbytes);
2619 EXT4_RO_ATTR(lifetime_write_kbytes);
2620 EXT4_RW_ATTR(reserved_clusters);
2621 EXT4_ATTR_OFFSET(inode_readahead_blks, 0644, sbi_ui_show,
2622 inode_readahead_blks_store, s_inode_readahead_blks);
2623 EXT4_RW_ATTR_SBI_UI(inode_goal, s_inode_goal);
2624 EXT4_RW_ATTR_SBI_UI(mb_stats, s_mb_stats);
2625 EXT4_RW_ATTR_SBI_UI(mb_max_to_scan, s_mb_max_to_scan);
2626 EXT4_RW_ATTR_SBI_UI(mb_min_to_scan, s_mb_min_to_scan);
2627 EXT4_RW_ATTR_SBI_UI(mb_order2_req, s_mb_order2_reqs);
2628 EXT4_RW_ATTR_SBI_UI(mb_stream_req, s_mb_stream_request);
2629 EXT4_RW_ATTR_SBI_UI(mb_group_prealloc, s_mb_group_prealloc);
2630 EXT4_DEPRECATED_ATTR(max_writeback_mb_bump, 128);
2631 EXT4_RW_ATTR_SBI_UI(extent_max_zeroout_kb, s_extent_max_zeroout_kb);
2632 EXT4_ATTR(trigger_fs_error, 0200, NULL, trigger_test_error);
2633 EXT4_RW_ATTR_SBI_UI(err_ratelimit_interval_ms, s_err_ratelimit_state.interval);
2634 EXT4_RW_ATTR_SBI_UI(err_ratelimit_burst, s_err_ratelimit_state.burst);
2635 EXT4_RW_ATTR_SBI_UI(warning_ratelimit_interval_ms, s_warning_ratelimit_state.interval);
2636 EXT4_RW_ATTR_SBI_UI(warning_ratelimit_burst, s_warning_ratelimit_state.burst);
2637 EXT4_RW_ATTR_SBI_UI(msg_ratelimit_interval_ms, s_msg_ratelimit_state.interval);
2638 EXT4_RW_ATTR_SBI_UI(msg_ratelimit_burst, s_msg_ratelimit_state.burst);
2639 EXT4_RO_ATTR_ES_UI(errors_count, s_error_count);
2640 EXT4_RO_ATTR_ES_UI(first_error_time, s_first_error_time);
2641 EXT4_RO_ATTR_ES_UI(last_error_time, s_last_error_time);
2643 static struct attribute *ext4_attrs[] = {
2644 ATTR_LIST(delayed_allocation_blocks),
2645 ATTR_LIST(session_write_kbytes),
2646 ATTR_LIST(lifetime_write_kbytes),
2647 ATTR_LIST(reserved_clusters),
2648 ATTR_LIST(inode_readahead_blks),
2649 ATTR_LIST(inode_goal),
2650 ATTR_LIST(mb_stats),
2651 ATTR_LIST(mb_max_to_scan),
2652 ATTR_LIST(mb_min_to_scan),
2653 ATTR_LIST(mb_order2_req),
2654 ATTR_LIST(mb_stream_req),
2655 ATTR_LIST(mb_group_prealloc),
2656 ATTR_LIST(max_writeback_mb_bump),
2657 ATTR_LIST(extent_max_zeroout_kb),
2658 ATTR_LIST(trigger_fs_error),
2659 ATTR_LIST(err_ratelimit_interval_ms),
2660 ATTR_LIST(err_ratelimit_burst),
2661 ATTR_LIST(warning_ratelimit_interval_ms),
2662 ATTR_LIST(warning_ratelimit_burst),
2663 ATTR_LIST(msg_ratelimit_interval_ms),
2664 ATTR_LIST(msg_ratelimit_burst),
2665 ATTR_LIST(errors_count),
2666 ATTR_LIST(first_error_time),
2667 ATTR_LIST(last_error_time),
2671 /* Features this copy of ext4 supports */
2672 EXT4_INFO_ATTR(lazy_itable_init);
2673 EXT4_INFO_ATTR(batched_discard);
2674 EXT4_INFO_ATTR(meta_bg_resize);
2676 static struct attribute *ext4_feat_attrs[] = {
2677 ATTR_LIST(lazy_itable_init),
2678 ATTR_LIST(batched_discard),
2679 ATTR_LIST(meta_bg_resize),
2683 static ssize_t ext4_attr_show(struct kobject *kobj,
2684 struct attribute *attr, char *buf)
2686 struct ext4_sb_info *sbi = container_of(kobj, struct ext4_sb_info,
2688 struct ext4_attr *a = container_of(attr, struct ext4_attr, attr);
2690 return a->show ? a->show(a, sbi, buf) : 0;
2693 static ssize_t ext4_attr_store(struct kobject *kobj,
2694 struct attribute *attr,
2695 const char *buf, size_t len)
2697 struct ext4_sb_info *sbi = container_of(kobj, struct ext4_sb_info,
2699 struct ext4_attr *a = container_of(attr, struct ext4_attr, attr);
2701 return a->store ? a->store(a, sbi, buf, len) : 0;
2704 static void ext4_sb_release(struct kobject *kobj)
2706 struct ext4_sb_info *sbi = container_of(kobj, struct ext4_sb_info,
2708 complete(&sbi->s_kobj_unregister);
2711 static const struct sysfs_ops ext4_attr_ops = {
2712 .show = ext4_attr_show,
2713 .store = ext4_attr_store,
2716 static struct kobj_type ext4_ktype = {
2717 .default_attrs = ext4_attrs,
2718 .sysfs_ops = &ext4_attr_ops,
2719 .release = ext4_sb_release,
2722 static void ext4_feat_release(struct kobject *kobj)
2724 complete(&ext4_feat->f_kobj_unregister);
2727 static ssize_t ext4_feat_show(struct kobject *kobj,
2728 struct attribute *attr, char *buf)
2730 return snprintf(buf, PAGE_SIZE, "supported\n");
2734 * We can not use ext4_attr_show/store because it relies on the kobject
2735 * being embedded in the ext4_sb_info structure which is definitely not
2736 * true in this case.
2738 static const struct sysfs_ops ext4_feat_ops = {
2739 .show = ext4_feat_show,
2743 static struct kobj_type ext4_feat_ktype = {
2744 .default_attrs = ext4_feat_attrs,
2745 .sysfs_ops = &ext4_feat_ops,
2746 .release = ext4_feat_release,
2750 * Check whether this filesystem can be mounted based on
2751 * the features present and the RDONLY/RDWR mount requested.
2752 * Returns 1 if this filesystem can be mounted as requested,
2753 * 0 if it cannot be.
2755 static int ext4_feature_set_ok(struct super_block *sb, int readonly)
2757 if (EXT4_HAS_INCOMPAT_FEATURE(sb, ~EXT4_FEATURE_INCOMPAT_SUPP)) {
2758 ext4_msg(sb, KERN_ERR,
2759 "Couldn't mount because of "
2760 "unsupported optional features (%x)",
2761 (le32_to_cpu(EXT4_SB(sb)->s_es->s_feature_incompat) &
2762 ~EXT4_FEATURE_INCOMPAT_SUPP));
2769 /* Check that feature set is OK for a read-write mount */
2770 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, ~EXT4_FEATURE_RO_COMPAT_SUPP)) {
2771 ext4_msg(sb, KERN_ERR, "couldn't mount RDWR because of "
2772 "unsupported optional features (%x)",
2773 (le32_to_cpu(EXT4_SB(sb)->s_es->s_feature_ro_compat) &
2774 ~EXT4_FEATURE_RO_COMPAT_SUPP));
2778 * Large file size enabled file system can only be mounted
2779 * read-write on 32-bit systems if kernel is built with CONFIG_LBDAF
2781 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_HUGE_FILE)) {
2782 if (sizeof(blkcnt_t) < sizeof(u64)) {
2783 ext4_msg(sb, KERN_ERR, "Filesystem with huge files "
2784 "cannot be mounted RDWR without "
2789 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_BIGALLOC) &&
2790 !EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_EXTENTS)) {
2791 ext4_msg(sb, KERN_ERR,
2792 "Can't support bigalloc feature without "
2793 "extents feature\n");
2797 #ifndef CONFIG_QUOTA
2798 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_QUOTA) &&
2800 ext4_msg(sb, KERN_ERR,
2801 "Filesystem with quota feature cannot be mounted RDWR "
2802 "without CONFIG_QUOTA");
2805 #endif /* CONFIG_QUOTA */
2810 * This function is called once a day if we have errors logged
2811 * on the file system
2813 static void print_daily_error_info(unsigned long arg)
2815 struct super_block *sb = (struct super_block *) arg;
2816 struct ext4_sb_info *sbi;
2817 struct ext4_super_block *es;
2822 if (es->s_error_count)
2823 /* fsck newer than v1.41.13 is needed to clean this condition. */
2824 ext4_msg(sb, KERN_NOTICE, "error count since last fsck: %u",
2825 le32_to_cpu(es->s_error_count));
2826 if (es->s_first_error_time) {
2827 printk(KERN_NOTICE "EXT4-fs (%s): initial error at time %u: %.*s:%d",
2828 sb->s_id, le32_to_cpu(es->s_first_error_time),
2829 (int) sizeof(es->s_first_error_func),
2830 es->s_first_error_func,
2831 le32_to_cpu(es->s_first_error_line));
2832 if (es->s_first_error_ino)
2833 printk(": inode %u",
2834 le32_to_cpu(es->s_first_error_ino));
2835 if (es->s_first_error_block)
2836 printk(": block %llu", (unsigned long long)
2837 le64_to_cpu(es->s_first_error_block));
2840 if (es->s_last_error_time) {
2841 printk(KERN_NOTICE "EXT4-fs (%s): last error at time %u: %.*s:%d",
2842 sb->s_id, le32_to_cpu(es->s_last_error_time),
2843 (int) sizeof(es->s_last_error_func),
2844 es->s_last_error_func,
2845 le32_to_cpu(es->s_last_error_line));
2846 if (es->s_last_error_ino)
2847 printk(": inode %u",
2848 le32_to_cpu(es->s_last_error_ino));
2849 if (es->s_last_error_block)
2850 printk(": block %llu", (unsigned long long)
2851 le64_to_cpu(es->s_last_error_block));
2854 mod_timer(&sbi->s_err_report, jiffies + 24*60*60*HZ); /* Once a day */
2857 /* Find next suitable group and run ext4_init_inode_table */
2858 static int ext4_run_li_request(struct ext4_li_request *elr)
2860 struct ext4_group_desc *gdp = NULL;
2861 ext4_group_t group, ngroups;
2862 struct super_block *sb;
2863 unsigned long timeout = 0;
2867 ngroups = EXT4_SB(sb)->s_groups_count;
2870 for (group = elr->lr_next_group; group < ngroups; group++) {
2871 gdp = ext4_get_group_desc(sb, group, NULL);
2877 if (!(gdp->bg_flags & cpu_to_le16(EXT4_BG_INODE_ZEROED)))
2881 if (group >= ngroups)
2886 ret = ext4_init_inode_table(sb, group,
2887 elr->lr_timeout ? 0 : 1);
2888 if (elr->lr_timeout == 0) {
2889 timeout = (jiffies - timeout) *
2890 elr->lr_sbi->s_li_wait_mult;
2891 elr->lr_timeout = timeout;
2893 elr->lr_next_sched = jiffies + elr->lr_timeout;
2894 elr->lr_next_group = group + 1;
2902 * Remove lr_request from the list_request and free the
2903 * request structure. Should be called with li_list_mtx held
2905 static void ext4_remove_li_request(struct ext4_li_request *elr)
2907 struct ext4_sb_info *sbi;
2914 list_del(&elr->lr_request);
2915 sbi->s_li_request = NULL;
2919 static void ext4_unregister_li_request(struct super_block *sb)
2921 mutex_lock(&ext4_li_mtx);
2922 if (!ext4_li_info) {
2923 mutex_unlock(&ext4_li_mtx);
2927 mutex_lock(&ext4_li_info->li_list_mtx);
2928 ext4_remove_li_request(EXT4_SB(sb)->s_li_request);
2929 mutex_unlock(&ext4_li_info->li_list_mtx);
2930 mutex_unlock(&ext4_li_mtx);
2933 static struct task_struct *ext4_lazyinit_task;
2936 * This is the function where ext4lazyinit thread lives. It walks
2937 * through the request list searching for next scheduled filesystem.
2938 * When such a fs is found, run the lazy initialization request
2939 * (ext4_rn_li_request) and keep track of the time spend in this
2940 * function. Based on that time we compute next schedule time of
2941 * the request. When walking through the list is complete, compute
2942 * next waking time and put itself into sleep.
2944 static int ext4_lazyinit_thread(void *arg)
2946 struct ext4_lazy_init *eli = (struct ext4_lazy_init *)arg;
2947 struct list_head *pos, *n;
2948 struct ext4_li_request *elr;
2949 unsigned long next_wakeup, cur;
2951 BUG_ON(NULL == eli);
2955 next_wakeup = MAX_JIFFY_OFFSET;
2957 mutex_lock(&eli->li_list_mtx);
2958 if (list_empty(&eli->li_request_list)) {
2959 mutex_unlock(&eli->li_list_mtx);
2963 list_for_each_safe(pos, n, &eli->li_request_list) {
2964 elr = list_entry(pos, struct ext4_li_request,
2967 if (time_after_eq(jiffies, elr->lr_next_sched)) {
2968 if (ext4_run_li_request(elr) != 0) {
2969 /* error, remove the lazy_init job */
2970 ext4_remove_li_request(elr);
2975 if (time_before(elr->lr_next_sched, next_wakeup))
2976 next_wakeup = elr->lr_next_sched;
2978 mutex_unlock(&eli->li_list_mtx);
2983 if ((time_after_eq(cur, next_wakeup)) ||
2984 (MAX_JIFFY_OFFSET == next_wakeup)) {
2989 schedule_timeout_interruptible(next_wakeup - cur);
2991 if (kthread_should_stop()) {
2992 ext4_clear_request_list();
2999 * It looks like the request list is empty, but we need
3000 * to check it under the li_list_mtx lock, to prevent any
3001 * additions into it, and of course we should lock ext4_li_mtx
3002 * to atomically free the list and ext4_li_info, because at
3003 * this point another ext4 filesystem could be registering
3006 mutex_lock(&ext4_li_mtx);
3007 mutex_lock(&eli->li_list_mtx);
3008 if (!list_empty(&eli->li_request_list)) {
3009 mutex_unlock(&eli->li_list_mtx);
3010 mutex_unlock(&ext4_li_mtx);
3013 mutex_unlock(&eli->li_list_mtx);
3014 kfree(ext4_li_info);
3015 ext4_li_info = NULL;
3016 mutex_unlock(&ext4_li_mtx);
3021 static void ext4_clear_request_list(void)
3023 struct list_head *pos, *n;
3024 struct ext4_li_request *elr;
3026 mutex_lock(&ext4_li_info->li_list_mtx);
3027 list_for_each_safe(pos, n, &ext4_li_info->li_request_list) {
3028 elr = list_entry(pos, struct ext4_li_request,
3030 ext4_remove_li_request(elr);
3032 mutex_unlock(&ext4_li_info->li_list_mtx);
3035 static int ext4_run_lazyinit_thread(void)
3037 ext4_lazyinit_task = kthread_run(ext4_lazyinit_thread,
3038 ext4_li_info, "ext4lazyinit");
3039 if (IS_ERR(ext4_lazyinit_task)) {
3040 int err = PTR_ERR(ext4_lazyinit_task);
3041 ext4_clear_request_list();
3042 kfree(ext4_li_info);
3043 ext4_li_info = NULL;
3044 printk(KERN_CRIT "EXT4-fs: error %d creating inode table "
3045 "initialization thread\n",
3049 ext4_li_info->li_state |= EXT4_LAZYINIT_RUNNING;
3054 * Check whether it make sense to run itable init. thread or not.
3055 * If there is at least one uninitialized inode table, return
3056 * corresponding group number, else the loop goes through all
3057 * groups and return total number of groups.
3059 static ext4_group_t ext4_has_uninit_itable(struct super_block *sb)
3061 ext4_group_t group, ngroups = EXT4_SB(sb)->s_groups_count;
3062 struct ext4_group_desc *gdp = NULL;
3064 for (group = 0; group < ngroups; group++) {
3065 gdp = ext4_get_group_desc(sb, group, NULL);
3069 if (!(gdp->bg_flags & cpu_to_le16(EXT4_BG_INODE_ZEROED)))
3076 static int ext4_li_info_new(void)
3078 struct ext4_lazy_init *eli = NULL;
3080 eli = kzalloc(sizeof(*eli), GFP_KERNEL);
3084 INIT_LIST_HEAD(&eli->li_request_list);
3085 mutex_init(&eli->li_list_mtx);
3087 eli->li_state |= EXT4_LAZYINIT_QUIT;
3094 static struct ext4_li_request *ext4_li_request_new(struct super_block *sb,
3097 struct ext4_sb_info *sbi = EXT4_SB(sb);
3098 struct ext4_li_request *elr;
3100 elr = kzalloc(sizeof(*elr), GFP_KERNEL);
3106 elr->lr_next_group = start;
3109 * Randomize first schedule time of the request to
3110 * spread the inode table initialization requests
3113 elr->lr_next_sched = jiffies + (prandom_u32() %
3114 (EXT4_DEF_LI_MAX_START_DELAY * HZ));
3118 int ext4_register_li_request(struct super_block *sb,
3119 ext4_group_t first_not_zeroed)
3121 struct ext4_sb_info *sbi = EXT4_SB(sb);
3122 struct ext4_li_request *elr = NULL;
3123 ext4_group_t ngroups = EXT4_SB(sb)->s_groups_count;
3126 mutex_lock(&ext4_li_mtx);
3127 if (sbi->s_li_request != NULL) {
3129 * Reset timeout so it can be computed again, because
3130 * s_li_wait_mult might have changed.
3132 sbi->s_li_request->lr_timeout = 0;
3136 if (first_not_zeroed == ngroups ||
3137 (sb->s_flags & MS_RDONLY) ||
3138 !test_opt(sb, INIT_INODE_TABLE))
3141 elr = ext4_li_request_new(sb, first_not_zeroed);
3147 if (NULL == ext4_li_info) {
3148 ret = ext4_li_info_new();
3153 mutex_lock(&ext4_li_info->li_list_mtx);
3154 list_add(&elr->lr_request, &ext4_li_info->li_request_list);
3155 mutex_unlock(&ext4_li_info->li_list_mtx);
3157 sbi->s_li_request = elr;
3159 * set elr to NULL here since it has been inserted to
3160 * the request_list and the removal and free of it is
3161 * handled by ext4_clear_request_list from now on.
3165 if (!(ext4_li_info->li_state & EXT4_LAZYINIT_RUNNING)) {
3166 ret = ext4_run_lazyinit_thread();
3171 mutex_unlock(&ext4_li_mtx);
3178 * We do not need to lock anything since this is called on
3181 static void ext4_destroy_lazyinit_thread(void)
3184 * If thread exited earlier
3185 * there's nothing to be done.
3187 if (!ext4_li_info || !ext4_lazyinit_task)
3190 kthread_stop(ext4_lazyinit_task);
3193 static int set_journal_csum_feature_set(struct super_block *sb)
3196 int compat, incompat;
3197 struct ext4_sb_info *sbi = EXT4_SB(sb);
3199 if (ext4_has_metadata_csum(sb)) {
3200 /* journal checksum v3 */
3202 incompat = JBD2_FEATURE_INCOMPAT_CSUM_V3;
3204 /* journal checksum v1 */
3205 compat = JBD2_FEATURE_COMPAT_CHECKSUM;
3209 jbd2_journal_clear_features(sbi->s_journal,
3210 JBD2_FEATURE_COMPAT_CHECKSUM, 0,
3211 JBD2_FEATURE_INCOMPAT_CSUM_V3 |
3212 JBD2_FEATURE_INCOMPAT_CSUM_V2);
3213 if (test_opt(sb, JOURNAL_ASYNC_COMMIT)) {
3214 ret = jbd2_journal_set_features(sbi->s_journal,
3216 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT |
3218 } else if (test_opt(sb, JOURNAL_CHECKSUM)) {
3219 ret = jbd2_journal_set_features(sbi->s_journal,
3222 jbd2_journal_clear_features(sbi->s_journal, 0, 0,
3223 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT);
3225 jbd2_journal_clear_features(sbi->s_journal, 0, 0,
3226 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT);
3233 * Note: calculating the overhead so we can be compatible with
3234 * historical BSD practice is quite difficult in the face of
3235 * clusters/bigalloc. This is because multiple metadata blocks from
3236 * different block group can end up in the same allocation cluster.
3237 * Calculating the exact overhead in the face of clustered allocation
3238 * requires either O(all block bitmaps) in memory or O(number of block
3239 * groups**2) in time. We will still calculate the superblock for
3240 * older file systems --- and if we come across with a bigalloc file
3241 * system with zero in s_overhead_clusters the estimate will be close to
3242 * correct especially for very large cluster sizes --- but for newer
3243 * file systems, it's better to calculate this figure once at mkfs
3244 * time, and store it in the superblock. If the superblock value is
3245 * present (even for non-bigalloc file systems), we will use it.
3247 static int count_overhead(struct super_block *sb, ext4_group_t grp,
3250 struct ext4_sb_info *sbi = EXT4_SB(sb);
3251 struct ext4_group_desc *gdp;
3252 ext4_fsblk_t first_block, last_block, b;
3253 ext4_group_t i, ngroups = ext4_get_groups_count(sb);
3254 int s, j, count = 0;
3256 if (!EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_BIGALLOC))
3257 return (ext4_bg_has_super(sb, grp) + ext4_bg_num_gdb(sb, grp) +
3258 sbi->s_itb_per_group + 2);
3260 first_block = le32_to_cpu(sbi->s_es->s_first_data_block) +
3261 (grp * EXT4_BLOCKS_PER_GROUP(sb));
3262 last_block = first_block + EXT4_BLOCKS_PER_GROUP(sb) - 1;
3263 for (i = 0; i < ngroups; i++) {
3264 gdp = ext4_get_group_desc(sb, i, NULL);
3265 b = ext4_block_bitmap(sb, gdp);
3266 if (b >= first_block && b <= last_block) {
3267 ext4_set_bit(EXT4_B2C(sbi, b - first_block), buf);
3270 b = ext4_inode_bitmap(sb, gdp);
3271 if (b >= first_block && b <= last_block) {
3272 ext4_set_bit(EXT4_B2C(sbi, b - first_block), buf);
3275 b = ext4_inode_table(sb, gdp);
3276 if (b >= first_block && b + sbi->s_itb_per_group <= last_block)
3277 for (j = 0; j < sbi->s_itb_per_group; j++, b++) {
3278 int c = EXT4_B2C(sbi, b - first_block);
3279 ext4_set_bit(c, buf);
3285 if (ext4_bg_has_super(sb, grp)) {
3286 ext4_set_bit(s++, buf);
3289 for (j = ext4_bg_num_gdb(sb, grp); j > 0; j--) {
3290 ext4_set_bit(EXT4_B2C(sbi, s++), buf);
3296 return EXT4_CLUSTERS_PER_GROUP(sb) -
3297 ext4_count_free(buf, EXT4_CLUSTERS_PER_GROUP(sb) / 8);
3301 * Compute the overhead and stash it in sbi->s_overhead
3303 int ext4_calculate_overhead(struct super_block *sb)
3305 struct ext4_sb_info *sbi = EXT4_SB(sb);
3306 struct ext4_super_block *es = sbi->s_es;
3307 ext4_group_t i, ngroups = ext4_get_groups_count(sb);
3308 ext4_fsblk_t overhead = 0;
3309 char *buf = (char *) get_zeroed_page(GFP_KERNEL);
3315 * Compute the overhead (FS structures). This is constant
3316 * for a given filesystem unless the number of block groups
3317 * changes so we cache the previous value until it does.
3321 * All of the blocks before first_data_block are overhead
3323 overhead = EXT4_B2C(sbi, le32_to_cpu(es->s_first_data_block));
3326 * Add the overhead found in each block group
3328 for (i = 0; i < ngroups; i++) {
3331 blks = count_overhead(sb, i, buf);
3334 memset(buf, 0, PAGE_SIZE);
3337 /* Add the journal blocks as well */
3339 overhead += EXT4_NUM_B2C(sbi, sbi->s_journal->j_maxlen);
3341 sbi->s_overhead = overhead;
3343 free_page((unsigned long) buf);
3348 static ext4_fsblk_t ext4_calculate_resv_clusters(struct super_block *sb)
3350 ext4_fsblk_t resv_clusters;
3353 * There's no need to reserve anything when we aren't using extents.
3354 * The space estimates are exact, there are no unwritten extents,
3355 * hole punching doesn't need new metadata... This is needed especially
3356 * to keep ext2/3 backward compatibility.
3358 if (!EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_EXTENTS))
3361 * By default we reserve 2% or 4096 clusters, whichever is smaller.
3362 * This should cover the situations where we can not afford to run
3363 * out of space like for example punch hole, or converting
3364 * unwritten extents in delalloc path. In most cases such
3365 * allocation would require 1, or 2 blocks, higher numbers are
3368 resv_clusters = ext4_blocks_count(EXT4_SB(sb)->s_es) >>
3369 EXT4_SB(sb)->s_cluster_bits;
3371 do_div(resv_clusters, 50);
3372 resv_clusters = min_t(ext4_fsblk_t, resv_clusters, 4096);
3374 return resv_clusters;
3378 static int ext4_reserve_clusters(struct ext4_sb_info *sbi, ext4_fsblk_t count)
3380 ext4_fsblk_t clusters = ext4_blocks_count(sbi->s_es) >>
3381 sbi->s_cluster_bits;
3383 if (count >= clusters)
3386 atomic64_set(&sbi->s_resv_clusters, count);
3390 static int ext4_fill_super(struct super_block *sb, void *data, int silent)
3392 char *orig_data = kstrdup(data, GFP_KERNEL);
3393 struct buffer_head *bh;
3394 struct ext4_super_block *es = NULL;
3395 struct ext4_sb_info *sbi;
3397 ext4_fsblk_t sb_block = get_sb_block(&data);
3398 ext4_fsblk_t logical_sb_block;
3399 unsigned long offset = 0;
3400 unsigned long journal_devnum = 0;
3401 unsigned long def_mount_opts;
3406 int blocksize, clustersize;
3407 unsigned int db_count;
3409 int needs_recovery, has_huge_files, has_bigalloc;
3412 unsigned int journal_ioprio = DEFAULT_JOURNAL_IOPRIO;
3413 ext4_group_t first_not_zeroed;
3415 sbi = kzalloc(sizeof(*sbi), GFP_KERNEL);
3419 sbi->s_blockgroup_lock =
3420 kzalloc(sizeof(struct blockgroup_lock), GFP_KERNEL);
3421 if (!sbi->s_blockgroup_lock) {
3425 sb->s_fs_info = sbi;
3427 sbi->s_inode_readahead_blks = EXT4_DEF_INODE_READAHEAD_BLKS;
3428 sbi->s_sb_block = sb_block;
3429 if (sb->s_bdev->bd_part)
3430 sbi->s_sectors_written_start =
3431 part_stat_read(sb->s_bdev->bd_part, sectors[1]);
3433 /* Cleanup superblock name */
3434 for (cp = sb->s_id; (cp = strchr(cp, '/'));)
3437 /* -EINVAL is default */
3439 blocksize = sb_min_blocksize(sb, EXT4_MIN_BLOCK_SIZE);
3441 ext4_msg(sb, KERN_ERR, "unable to set blocksize");
3446 * The ext4 superblock will not be buffer aligned for other than 1kB
3447 * block sizes. We need to calculate the offset from buffer start.
3449 if (blocksize != EXT4_MIN_BLOCK_SIZE) {
3450 logical_sb_block = sb_block * EXT4_MIN_BLOCK_SIZE;
3451 offset = do_div(logical_sb_block, blocksize);
3453 logical_sb_block = sb_block;
3456 if (!(bh = sb_bread_unmovable(sb, logical_sb_block))) {
3457 ext4_msg(sb, KERN_ERR, "unable to read superblock");
3461 * Note: s_es must be initialized as soon as possible because
3462 * some ext4 macro-instructions depend on its value
3464 es = (struct ext4_super_block *) (bh->b_data + offset);
3466 sb->s_magic = le16_to_cpu(es->s_magic);
3467 if (sb->s_magic != EXT4_SUPER_MAGIC)
3469 sbi->s_kbytes_written = le64_to_cpu(es->s_kbytes_written);
3471 /* Warn if metadata_csum and gdt_csum are both set. */
3472 if (EXT4_HAS_RO_COMPAT_FEATURE(sb,
3473 EXT4_FEATURE_RO_COMPAT_METADATA_CSUM) &&
3474 EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_GDT_CSUM))
3475 ext4_warning(sb, KERN_INFO "metadata_csum and uninit_bg are "
3476 "redundant flags; please run fsck.");
3478 /* Check for a known checksum algorithm */
3479 if (!ext4_verify_csum_type(sb, es)) {
3480 ext4_msg(sb, KERN_ERR, "VFS: Found ext4 filesystem with "
3481 "unknown checksum algorithm.");
3486 /* Load the checksum driver */
3487 if (EXT4_HAS_RO_COMPAT_FEATURE(sb,
3488 EXT4_FEATURE_RO_COMPAT_METADATA_CSUM)) {
3489 sbi->s_chksum_driver = crypto_alloc_shash("crc32c", 0, 0);
3490 if (IS_ERR(sbi->s_chksum_driver)) {
3491 ext4_msg(sb, KERN_ERR, "Cannot load crc32c driver.");
3492 ret = PTR_ERR(sbi->s_chksum_driver);
3493 sbi->s_chksum_driver = NULL;
3498 /* Check superblock checksum */
3499 if (!ext4_superblock_csum_verify(sb, es)) {
3500 ext4_msg(sb, KERN_ERR, "VFS: Found ext4 filesystem with "
3501 "invalid superblock checksum. Run e2fsck?");
3506 /* Precompute checksum seed for all metadata */
3507 if (ext4_has_metadata_csum(sb))
3508 sbi->s_csum_seed = ext4_chksum(sbi, ~0, es->s_uuid,
3509 sizeof(es->s_uuid));
3511 /* Set defaults before we parse the mount options */
3512 def_mount_opts = le32_to_cpu(es->s_default_mount_opts);
3513 set_opt(sb, INIT_INODE_TABLE);
3514 if (def_mount_opts & EXT4_DEFM_DEBUG)
3516 if (def_mount_opts & EXT4_DEFM_BSDGROUPS)
3518 if (def_mount_opts & EXT4_DEFM_UID16)
3519 set_opt(sb, NO_UID32);
3520 /* xattr user namespace & acls are now defaulted on */
3521 set_opt(sb, XATTR_USER);
3522 #ifdef CONFIG_EXT4_FS_POSIX_ACL
3523 set_opt(sb, POSIX_ACL);
3525 if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_DATA)
3526 set_opt(sb, JOURNAL_DATA);
3527 else if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_ORDERED)
3528 set_opt(sb, ORDERED_DATA);
3529 else if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_WBACK)
3530 set_opt(sb, WRITEBACK_DATA);
3532 if (le16_to_cpu(sbi->s_es->s_errors) == EXT4_ERRORS_PANIC)
3533 set_opt(sb, ERRORS_PANIC);
3534 else if (le16_to_cpu(sbi->s_es->s_errors) == EXT4_ERRORS_CONTINUE)
3535 set_opt(sb, ERRORS_CONT);
3537 set_opt(sb, ERRORS_RO);
3538 /* block_validity enabled by default; disable with noblock_validity */
3539 set_opt(sb, BLOCK_VALIDITY);
3540 if (def_mount_opts & EXT4_DEFM_DISCARD)
3541 set_opt(sb, DISCARD);
3543 sbi->s_resuid = make_kuid(&init_user_ns, le16_to_cpu(es->s_def_resuid));
3544 sbi->s_resgid = make_kgid(&init_user_ns, le16_to_cpu(es->s_def_resgid));
3545 sbi->s_commit_interval = JBD2_DEFAULT_MAX_COMMIT_AGE * HZ;
3546 sbi->s_min_batch_time = EXT4_DEF_MIN_BATCH_TIME;
3547 sbi->s_max_batch_time = EXT4_DEF_MAX_BATCH_TIME;
3549 if ((def_mount_opts & EXT4_DEFM_NOBARRIER) == 0)
3550 set_opt(sb, BARRIER);
3553 * enable delayed allocation by default
3554 * Use -o nodelalloc to turn it off
3556 if (!IS_EXT3_SB(sb) && !IS_EXT2_SB(sb) &&
3557 ((def_mount_opts & EXT4_DEFM_NODELALLOC) == 0))
3558 set_opt(sb, DELALLOC);
3561 * set default s_li_wait_mult for lazyinit, for the case there is
3562 * no mount option specified.
3564 sbi->s_li_wait_mult = EXT4_DEF_LI_WAIT_MULT;
3566 if (!parse_options((char *) sbi->s_es->s_mount_opts, sb,
3567 &journal_devnum, &journal_ioprio, 0)) {
3568 ext4_msg(sb, KERN_WARNING,
3569 "failed to parse options in superblock: %s",
3570 sbi->s_es->s_mount_opts);
3572 sbi->s_def_mount_opt = sbi->s_mount_opt;
3573 if (!parse_options((char *) data, sb, &journal_devnum,
3574 &journal_ioprio, 0))
3577 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA) {
3578 printk_once(KERN_WARNING "EXT4-fs: Warning: mounting "
3579 "with data=journal disables delayed "
3580 "allocation and O_DIRECT support!\n");
3581 if (test_opt2(sb, EXPLICIT_DELALLOC)) {
3582 ext4_msg(sb, KERN_ERR, "can't mount with "
3583 "both data=journal and delalloc");
3586 if (test_opt(sb, DIOREAD_NOLOCK)) {
3587 ext4_msg(sb, KERN_ERR, "can't mount with "
3588 "both data=journal and dioread_nolock");
3591 if (test_opt(sb, DELALLOC))
3592 clear_opt(sb, DELALLOC);
3595 sb->s_flags = (sb->s_flags & ~MS_POSIXACL) |
3596 (test_opt(sb, POSIX_ACL) ? MS_POSIXACL : 0);
3598 if (le32_to_cpu(es->s_rev_level) == EXT4_GOOD_OLD_REV &&
3599 (EXT4_HAS_COMPAT_FEATURE(sb, ~0U) ||
3600 EXT4_HAS_RO_COMPAT_FEATURE(sb, ~0U) ||
3601 EXT4_HAS_INCOMPAT_FEATURE(sb, ~0U)))
3602 ext4_msg(sb, KERN_WARNING,
3603 "feature flags set on rev 0 fs, "
3604 "running e2fsck is recommended");
3606 if (es->s_creator_os == cpu_to_le32(EXT4_OS_HURD)) {
3607 set_opt2(sb, HURD_COMPAT);
3608 if (EXT4_HAS_INCOMPAT_FEATURE(sb,
3609 EXT4_FEATURE_INCOMPAT_64BIT)) {
3610 ext4_msg(sb, KERN_ERR,
3611 "The Hurd can't support 64-bit file systems");
3616 if (IS_EXT2_SB(sb)) {
3617 if (ext2_feature_set_ok(sb))
3618 ext4_msg(sb, KERN_INFO, "mounting ext2 file system "
3619 "using the ext4 subsystem");
3621 ext4_msg(sb, KERN_ERR, "couldn't mount as ext2 due "
3622 "to feature incompatibilities");
3627 if (IS_EXT3_SB(sb)) {
3628 if (ext3_feature_set_ok(sb))
3629 ext4_msg(sb, KERN_INFO, "mounting ext3 file system "
3630 "using the ext4 subsystem");
3632 ext4_msg(sb, KERN_ERR, "couldn't mount as ext3 due "
3633 "to feature incompatibilities");
3639 * Check feature flags regardless of the revision level, since we
3640 * previously didn't change the revision level when setting the flags,
3641 * so there is a chance incompat flags are set on a rev 0 filesystem.
3643 if (!ext4_feature_set_ok(sb, (sb->s_flags & MS_RDONLY)))
3646 blocksize = BLOCK_SIZE << le32_to_cpu(es->s_log_block_size);
3647 if (blocksize < EXT4_MIN_BLOCK_SIZE ||
3648 blocksize > EXT4_MAX_BLOCK_SIZE) {
3649 ext4_msg(sb, KERN_ERR,
3650 "Unsupported filesystem blocksize %d", blocksize);
3654 if (sb->s_blocksize != blocksize) {
3655 /* Validate the filesystem blocksize */
3656 if (!sb_set_blocksize(sb, blocksize)) {
3657 ext4_msg(sb, KERN_ERR, "bad block size %d",
3663 logical_sb_block = sb_block * EXT4_MIN_BLOCK_SIZE;
3664 offset = do_div(logical_sb_block, blocksize);
3665 bh = sb_bread_unmovable(sb, logical_sb_block);
3667 ext4_msg(sb, KERN_ERR,
3668 "Can't read superblock on 2nd try");
3671 es = (struct ext4_super_block *)(bh->b_data + offset);
3673 if (es->s_magic != cpu_to_le16(EXT4_SUPER_MAGIC)) {
3674 ext4_msg(sb, KERN_ERR,
3675 "Magic mismatch, very weird!");
3680 has_huge_files = EXT4_HAS_RO_COMPAT_FEATURE(sb,
3681 EXT4_FEATURE_RO_COMPAT_HUGE_FILE);
3682 sbi->s_bitmap_maxbytes = ext4_max_bitmap_size(sb->s_blocksize_bits,
3684 sb->s_maxbytes = ext4_max_size(sb->s_blocksize_bits, has_huge_files);
3686 if (le32_to_cpu(es->s_rev_level) == EXT4_GOOD_OLD_REV) {
3687 sbi->s_inode_size = EXT4_GOOD_OLD_INODE_SIZE;
3688 sbi->s_first_ino = EXT4_GOOD_OLD_FIRST_INO;
3690 sbi->s_inode_size = le16_to_cpu(es->s_inode_size);
3691 sbi->s_first_ino = le32_to_cpu(es->s_first_ino);
3692 if ((sbi->s_inode_size < EXT4_GOOD_OLD_INODE_SIZE) ||
3693 (!is_power_of_2(sbi->s_inode_size)) ||
3694 (sbi->s_inode_size > blocksize)) {
3695 ext4_msg(sb, KERN_ERR,
3696 "unsupported inode size: %d",
3700 if (sbi->s_inode_size > EXT4_GOOD_OLD_INODE_SIZE)
3701 sb->s_time_gran = 1 << (EXT4_EPOCH_BITS - 2);
3704 sbi->s_desc_size = le16_to_cpu(es->s_desc_size);
3705 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_64BIT)) {
3706 if (sbi->s_desc_size < EXT4_MIN_DESC_SIZE_64BIT ||
3707 sbi->s_desc_size > EXT4_MAX_DESC_SIZE ||
3708 !is_power_of_2(sbi->s_desc_size)) {
3709 ext4_msg(sb, KERN_ERR,
3710 "unsupported descriptor size %lu",
3715 sbi->s_desc_size = EXT4_MIN_DESC_SIZE;
3717 sbi->s_blocks_per_group = le32_to_cpu(es->s_blocks_per_group);
3718 sbi->s_inodes_per_group = le32_to_cpu(es->s_inodes_per_group);
3719 if (EXT4_INODE_SIZE(sb) == 0 || EXT4_INODES_PER_GROUP(sb) == 0)
3722 sbi->s_inodes_per_block = blocksize / EXT4_INODE_SIZE(sb);
3723 if (sbi->s_inodes_per_block == 0)
3725 sbi->s_itb_per_group = sbi->s_inodes_per_group /
3726 sbi->s_inodes_per_block;
3727 sbi->s_desc_per_block = blocksize / EXT4_DESC_SIZE(sb);
3729 sbi->s_mount_state = le16_to_cpu(es->s_state);
3730 sbi->s_addr_per_block_bits = ilog2(EXT4_ADDR_PER_BLOCK(sb));
3731 sbi->s_desc_per_block_bits = ilog2(EXT4_DESC_PER_BLOCK(sb));
3733 for (i = 0; i < 4; i++)
3734 sbi->s_hash_seed[i] = le32_to_cpu(es->s_hash_seed[i]);
3735 sbi->s_def_hash_version = es->s_def_hash_version;
3736 if (EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_DIR_INDEX)) {
3737 i = le32_to_cpu(es->s_flags);
3738 if (i & EXT2_FLAGS_UNSIGNED_HASH)
3739 sbi->s_hash_unsigned = 3;
3740 else if ((i & EXT2_FLAGS_SIGNED_HASH) == 0) {
3741 #ifdef __CHAR_UNSIGNED__
3742 if (!(sb->s_flags & MS_RDONLY))
3744 cpu_to_le32(EXT2_FLAGS_UNSIGNED_HASH);
3745 sbi->s_hash_unsigned = 3;
3747 if (!(sb->s_flags & MS_RDONLY))
3749 cpu_to_le32(EXT2_FLAGS_SIGNED_HASH);
3754 /* Handle clustersize */
3755 clustersize = BLOCK_SIZE << le32_to_cpu(es->s_log_cluster_size);
3756 has_bigalloc = EXT4_HAS_RO_COMPAT_FEATURE(sb,
3757 EXT4_FEATURE_RO_COMPAT_BIGALLOC);
3759 if (clustersize < blocksize) {
3760 ext4_msg(sb, KERN_ERR,
3761 "cluster size (%d) smaller than "
3762 "block size (%d)", clustersize, blocksize);
3765 sbi->s_cluster_bits = le32_to_cpu(es->s_log_cluster_size) -
3766 le32_to_cpu(es->s_log_block_size);
3767 sbi->s_clusters_per_group =
3768 le32_to_cpu(es->s_clusters_per_group);
3769 if (sbi->s_clusters_per_group > blocksize * 8) {
3770 ext4_msg(sb, KERN_ERR,
3771 "#clusters per group too big: %lu",
3772 sbi->s_clusters_per_group);
3775 if (sbi->s_blocks_per_group !=
3776 (sbi->s_clusters_per_group * (clustersize / blocksize))) {
3777 ext4_msg(sb, KERN_ERR, "blocks per group (%lu) and "
3778 "clusters per group (%lu) inconsistent",
3779 sbi->s_blocks_per_group,
3780 sbi->s_clusters_per_group);
3784 if (clustersize != blocksize) {
3785 ext4_warning(sb, "fragment/cluster size (%d) != "
3786 "block size (%d)", clustersize,
3788 clustersize = blocksize;
3790 if (sbi->s_blocks_per_group > blocksize * 8) {
3791 ext4_msg(sb, KERN_ERR,
3792 "#blocks per group too big: %lu",
3793 sbi->s_blocks_per_group);
3796 sbi->s_clusters_per_group = sbi->s_blocks_per_group;
3797 sbi->s_cluster_bits = 0;
3799 sbi->s_cluster_ratio = clustersize / blocksize;
3801 if (sbi->s_inodes_per_group > blocksize * 8) {
3802 ext4_msg(sb, KERN_ERR,
3803 "#inodes per group too big: %lu",
3804 sbi->s_inodes_per_group);
3808 /* Do we have standard group size of clustersize * 8 blocks ? */
3809 if (sbi->s_blocks_per_group == clustersize << 3)
3810 set_opt2(sb, STD_GROUP_SIZE);
3813 * Test whether we have more sectors than will fit in sector_t,
3814 * and whether the max offset is addressable by the page cache.
3816 err = generic_check_addressable(sb->s_blocksize_bits,
3817 ext4_blocks_count(es));
3819 ext4_msg(sb, KERN_ERR, "filesystem"
3820 " too large to mount safely on this system");
3821 if (sizeof(sector_t) < 8)
3822 ext4_msg(sb, KERN_WARNING, "CONFIG_LBDAF not enabled");
3826 if (EXT4_BLOCKS_PER_GROUP(sb) == 0)
3829 /* check blocks count against device size */
3830 blocks_count = sb->s_bdev->bd_inode->i_size >> sb->s_blocksize_bits;
3831 if (blocks_count && ext4_blocks_count(es) > blocks_count) {
3832 ext4_msg(sb, KERN_WARNING, "bad geometry: block count %llu "
3833 "exceeds size of device (%llu blocks)",
3834 ext4_blocks_count(es), blocks_count);
3839 * It makes no sense for the first data block to be beyond the end
3840 * of the filesystem.
3842 if (le32_to_cpu(es->s_first_data_block) >= ext4_blocks_count(es)) {
3843 ext4_msg(sb, KERN_WARNING, "bad geometry: first data "
3844 "block %u is beyond end of filesystem (%llu)",
3845 le32_to_cpu(es->s_first_data_block),
3846 ext4_blocks_count(es));
3849 blocks_count = (ext4_blocks_count(es) -
3850 le32_to_cpu(es->s_first_data_block) +
3851 EXT4_BLOCKS_PER_GROUP(sb) - 1);
3852 do_div(blocks_count, EXT4_BLOCKS_PER_GROUP(sb));
3853 if (blocks_count > ((uint64_t)1<<32) - EXT4_DESC_PER_BLOCK(sb)) {
3854 ext4_msg(sb, KERN_WARNING, "groups count too large: %u "
3855 "(block count %llu, first data block %u, "
3856 "blocks per group %lu)", sbi->s_groups_count,
3857 ext4_blocks_count(es),
3858 le32_to_cpu(es->s_first_data_block),
3859 EXT4_BLOCKS_PER_GROUP(sb));
3862 sbi->s_groups_count = blocks_count;
3863 sbi->s_blockfile_groups = min_t(ext4_group_t, sbi->s_groups_count,
3864 (EXT4_MAX_BLOCK_FILE_PHYS / EXT4_BLOCKS_PER_GROUP(sb)));
3865 db_count = (sbi->s_groups_count + EXT4_DESC_PER_BLOCK(sb) - 1) /
3866 EXT4_DESC_PER_BLOCK(sb);
3867 sbi->s_group_desc = ext4_kvmalloc(db_count *
3868 sizeof(struct buffer_head *),
3870 if (sbi->s_group_desc == NULL) {
3871 ext4_msg(sb, KERN_ERR, "not enough memory");
3877 sbi->s_proc = proc_mkdir(sb->s_id, ext4_proc_root);
3880 proc_create_data("options", S_IRUGO, sbi->s_proc,
3881 &ext4_seq_options_fops, sb);
3883 bgl_lock_init(sbi->s_blockgroup_lock);
3885 for (i = 0; i < db_count; i++) {
3886 block = descriptor_loc(sb, logical_sb_block, i);
3887 sbi->s_group_desc[i] = sb_bread_unmovable(sb, block);
3888 if (!sbi->s_group_desc[i]) {
3889 ext4_msg(sb, KERN_ERR,
3890 "can't read group descriptor %d", i);
3895 if (!ext4_check_descriptors(sb, &first_not_zeroed)) {
3896 ext4_msg(sb, KERN_ERR, "group descriptors corrupted!");
3900 sbi->s_gdb_count = db_count;
3901 get_random_bytes(&sbi->s_next_generation, sizeof(u32));
3902 spin_lock_init(&sbi->s_next_gen_lock);
3904 init_timer(&sbi->s_err_report);
3905 sbi->s_err_report.function = print_daily_error_info;
3906 sbi->s_err_report.data = (unsigned long) sb;
3908 /* Register extent status tree shrinker */
3909 if (ext4_es_register_shrinker(sbi))
3912 sbi->s_stripe = ext4_get_stripe_size(sbi);
3913 sbi->s_extent_max_zeroout_kb = 32;
3916 * set up enough so that it can read an inode
3918 sb->s_op = &ext4_sops;
3919 sb->s_export_op = &ext4_export_ops;
3920 sb->s_xattr = ext4_xattr_handlers;
3922 sb->dq_op = &ext4_quota_operations;
3923 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_QUOTA))
3924 sb->s_qcop = &ext4_qctl_sysfile_operations;
3926 sb->s_qcop = &ext4_qctl_operations;
3928 memcpy(sb->s_uuid, es->s_uuid, sizeof(es->s_uuid));
3930 INIT_LIST_HEAD(&sbi->s_orphan); /* unlinked but open files */
3931 mutex_init(&sbi->s_orphan_lock);
3935 needs_recovery = (es->s_last_orphan != 0 ||
3936 EXT4_HAS_INCOMPAT_FEATURE(sb,
3937 EXT4_FEATURE_INCOMPAT_RECOVER));
3939 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_MMP) &&
3940 !(sb->s_flags & MS_RDONLY))
3941 if (ext4_multi_mount_protect(sb, le64_to_cpu(es->s_mmp_block)))
3945 * The first inode we look at is the journal inode. Don't try
3946 * root first: it may be modified in the journal!
3948 if (!test_opt(sb, NOLOAD) &&
3949 EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL)) {
3950 if (ext4_load_journal(sb, es, journal_devnum))
3952 } else if (test_opt(sb, NOLOAD) && !(sb->s_flags & MS_RDONLY) &&
3953 EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER)) {
3954 ext4_msg(sb, KERN_ERR, "required journal recovery "
3955 "suppressed and not mounted read-only");
3956 goto failed_mount_wq;
3958 clear_opt(sb, DATA_FLAGS);
3959 sbi->s_journal = NULL;
3964 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_64BIT) &&
3965 !jbd2_journal_set_features(EXT4_SB(sb)->s_journal, 0, 0,
3966 JBD2_FEATURE_INCOMPAT_64BIT)) {
3967 ext4_msg(sb, KERN_ERR, "Failed to set 64-bit journal feature");
3968 goto failed_mount_wq;
3971 if (!set_journal_csum_feature_set(sb)) {
3972 ext4_msg(sb, KERN_ERR, "Failed to set journal checksum "
3974 goto failed_mount_wq;
3977 /* We have now updated the journal if required, so we can
3978 * validate the data journaling mode. */
3979 switch (test_opt(sb, DATA_FLAGS)) {
3981 /* No mode set, assume a default based on the journal
3982 * capabilities: ORDERED_DATA if the journal can
3983 * cope, else JOURNAL_DATA
3985 if (jbd2_journal_check_available_features
3986 (sbi->s_journal, 0, 0, JBD2_FEATURE_INCOMPAT_REVOKE))
3987 set_opt(sb, ORDERED_DATA);
3989 set_opt(sb, JOURNAL_DATA);
3992 case EXT4_MOUNT_ORDERED_DATA:
3993 case EXT4_MOUNT_WRITEBACK_DATA:
3994 if (!jbd2_journal_check_available_features
3995 (sbi->s_journal, 0, 0, JBD2_FEATURE_INCOMPAT_REVOKE)) {
3996 ext4_msg(sb, KERN_ERR, "Journal does not support "
3997 "requested data journaling mode");
3998 goto failed_mount_wq;
4003 set_task_ioprio(sbi->s_journal->j_task, journal_ioprio);
4005 sbi->s_journal->j_commit_callback = ext4_journal_commit_callback;
4008 if (ext4_mballoc_ready) {
4009 sbi->s_mb_cache = ext4_xattr_create_cache(sb->s_id);
4010 if (!sbi->s_mb_cache) {
4011 ext4_msg(sb, KERN_ERR, "Failed to create an mb_cache");
4012 goto failed_mount_wq;
4017 * Get the # of file system overhead blocks from the
4018 * superblock if present.
4020 if (es->s_overhead_clusters)
4021 sbi->s_overhead = le32_to_cpu(es->s_overhead_clusters);
4023 err = ext4_calculate_overhead(sb);
4025 goto failed_mount_wq;
4029 * The maximum number of concurrent works can be high and
4030 * concurrency isn't really necessary. Limit it to 1.
4032 EXT4_SB(sb)->rsv_conversion_wq =
4033 alloc_workqueue("ext4-rsv-conversion", WQ_MEM_RECLAIM | WQ_UNBOUND, 1);
4034 if (!EXT4_SB(sb)->rsv_conversion_wq) {
4035 printk(KERN_ERR "EXT4-fs: failed to create workqueue\n");
4041 * The jbd2_journal_load will have done any necessary log recovery,
4042 * so we can safely mount the rest of the filesystem now.
4045 root = ext4_iget(sb, EXT4_ROOT_INO);
4047 ext4_msg(sb, KERN_ERR, "get root inode failed");
4048 ret = PTR_ERR(root);
4052 if (!S_ISDIR(root->i_mode) || !root->i_blocks || !root->i_size) {
4053 ext4_msg(sb, KERN_ERR, "corrupt root inode, run e2fsck");
4057 sb->s_root = d_make_root(root);
4059 ext4_msg(sb, KERN_ERR, "get root dentry failed");
4064 if (ext4_setup_super(sb, es, sb->s_flags & MS_RDONLY))
4065 sb->s_flags |= MS_RDONLY;
4067 /* determine the minimum size of new large inodes, if present */
4068 if (sbi->s_inode_size > EXT4_GOOD_OLD_INODE_SIZE) {
4069 sbi->s_want_extra_isize = sizeof(struct ext4_inode) -
4070 EXT4_GOOD_OLD_INODE_SIZE;
4071 if (EXT4_HAS_RO_COMPAT_FEATURE(sb,
4072 EXT4_FEATURE_RO_COMPAT_EXTRA_ISIZE)) {
4073 if (sbi->s_want_extra_isize <
4074 le16_to_cpu(es->s_want_extra_isize))
4075 sbi->s_want_extra_isize =
4076 le16_to_cpu(es->s_want_extra_isize);
4077 if (sbi->s_want_extra_isize <
4078 le16_to_cpu(es->s_min_extra_isize))
4079 sbi->s_want_extra_isize =
4080 le16_to_cpu(es->s_min_extra_isize);
4083 /* Check if enough inode space is available */
4084 if (EXT4_GOOD_OLD_INODE_SIZE + sbi->s_want_extra_isize >
4085 sbi->s_inode_size) {
4086 sbi->s_want_extra_isize = sizeof(struct ext4_inode) -
4087 EXT4_GOOD_OLD_INODE_SIZE;
4088 ext4_msg(sb, KERN_INFO, "required extra inode space not"
4092 err = ext4_reserve_clusters(sbi, ext4_calculate_resv_clusters(sb));
4094 ext4_msg(sb, KERN_ERR, "failed to reserve %llu clusters for "
4095 "reserved pool", ext4_calculate_resv_clusters(sb));
4096 goto failed_mount4a;
4099 err = ext4_setup_system_zone(sb);
4101 ext4_msg(sb, KERN_ERR, "failed to initialize system "
4103 goto failed_mount4a;
4107 err = ext4_mb_init(sb);
4109 ext4_msg(sb, KERN_ERR, "failed to initialize mballoc (%d)",
4114 block = ext4_count_free_clusters(sb);
4115 ext4_free_blocks_count_set(sbi->s_es,
4116 EXT4_C2B(sbi, block));
4117 err = percpu_counter_init(&sbi->s_freeclusters_counter, block);
4119 unsigned long freei = ext4_count_free_inodes(sb);
4120 sbi->s_es->s_free_inodes_count = cpu_to_le32(freei);
4121 err = percpu_counter_init(&sbi->s_freeinodes_counter, freei);
4124 err = percpu_counter_init(&sbi->s_dirs_counter,
4125 ext4_count_dirs(sb));
4127 err = percpu_counter_init(&sbi->s_dirtyclusters_counter, 0);
4129 ext4_msg(sb, KERN_ERR, "insufficient memory");
4133 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_FLEX_BG))
4134 if (!ext4_fill_flex_info(sb)) {
4135 ext4_msg(sb, KERN_ERR,
4136 "unable to initialize "
4137 "flex_bg meta info!");
4141 err = ext4_register_li_request(sb, first_not_zeroed);
4145 sbi->s_kobj.kset = ext4_kset;
4146 init_completion(&sbi->s_kobj_unregister);
4147 err = kobject_init_and_add(&sbi->s_kobj, &ext4_ktype, NULL,
4153 /* Enable quota usage during mount. */
4154 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_QUOTA) &&
4155 !(sb->s_flags & MS_RDONLY)) {
4156 err = ext4_enable_quotas(sb);
4160 #endif /* CONFIG_QUOTA */
4162 EXT4_SB(sb)->s_mount_state |= EXT4_ORPHAN_FS;
4163 ext4_orphan_cleanup(sb, es);
4164 EXT4_SB(sb)->s_mount_state &= ~EXT4_ORPHAN_FS;
4165 if (needs_recovery) {
4166 ext4_msg(sb, KERN_INFO, "recovery complete");
4167 ext4_mark_recovery_complete(sb, es);
4169 if (EXT4_SB(sb)->s_journal) {
4170 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA)
4171 descr = " journalled data mode";
4172 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA)
4173 descr = " ordered data mode";
4175 descr = " writeback data mode";
4177 descr = "out journal";
4179 if (test_opt(sb, DISCARD)) {
4180 struct request_queue *q = bdev_get_queue(sb->s_bdev);
4181 if (!blk_queue_discard(q))
4182 ext4_msg(sb, KERN_WARNING,
4183 "mounting with \"discard\" option, but "
4184 "the device does not support discard");
4187 ext4_msg(sb, KERN_INFO, "mounted filesystem with%s. "
4188 "Opts: %s%s%s", descr, sbi->s_es->s_mount_opts,
4189 *sbi->s_es->s_mount_opts ? "; " : "", orig_data);
4191 if (es->s_error_count)
4192 mod_timer(&sbi->s_err_report, jiffies + 300*HZ); /* 5 minutes */
4194 /* Enable message ratelimiting. Default is 10 messages per 5 secs. */
4195 ratelimit_state_init(&sbi->s_err_ratelimit_state, 5 * HZ, 10);
4196 ratelimit_state_init(&sbi->s_warning_ratelimit_state, 5 * HZ, 10);
4197 ratelimit_state_init(&sbi->s_msg_ratelimit_state, 5 * HZ, 10);
4204 ext4_msg(sb, KERN_ERR, "VFS: Can't find ext4 filesystem");
4209 kobject_del(&sbi->s_kobj);
4212 ext4_unregister_li_request(sb);
4214 ext4_mb_release(sb);
4215 if (sbi->s_flex_groups)
4216 ext4_kvfree(sbi->s_flex_groups);
4217 percpu_counter_destroy(&sbi->s_freeclusters_counter);
4218 percpu_counter_destroy(&sbi->s_freeinodes_counter);
4219 percpu_counter_destroy(&sbi->s_dirs_counter);
4220 percpu_counter_destroy(&sbi->s_dirtyclusters_counter);
4222 ext4_ext_release(sb);
4223 ext4_release_system_zone(sb);
4228 ext4_msg(sb, KERN_ERR, "mount failed");
4229 if (EXT4_SB(sb)->rsv_conversion_wq)
4230 destroy_workqueue(EXT4_SB(sb)->rsv_conversion_wq);
4232 if (sbi->s_journal) {
4233 jbd2_journal_destroy(sbi->s_journal);
4234 sbi->s_journal = NULL;
4236 ext4_es_unregister_shrinker(sbi);
4238 del_timer_sync(&sbi->s_err_report);
4240 kthread_stop(sbi->s_mmp_tsk);
4242 for (i = 0; i < db_count; i++)
4243 brelse(sbi->s_group_desc[i]);
4244 ext4_kvfree(sbi->s_group_desc);
4246 if (sbi->s_chksum_driver)
4247 crypto_free_shash(sbi->s_chksum_driver);
4249 remove_proc_entry("options", sbi->s_proc);
4250 remove_proc_entry(sb->s_id, ext4_proc_root);
4253 for (i = 0; i < EXT4_MAXQUOTAS; i++)
4254 kfree(sbi->s_qf_names[i]);
4256 ext4_blkdev_remove(sbi);
4259 sb->s_fs_info = NULL;
4260 kfree(sbi->s_blockgroup_lock);
4264 return err ? err : ret;
4268 * Setup any per-fs journal parameters now. We'll do this both on
4269 * initial mount, once the journal has been initialised but before we've
4270 * done any recovery; and again on any subsequent remount.
4272 static void ext4_init_journal_params(struct super_block *sb, journal_t *journal)
4274 struct ext4_sb_info *sbi = EXT4_SB(sb);
4276 journal->j_commit_interval = sbi->s_commit_interval;
4277 journal->j_min_batch_time = sbi->s_min_batch_time;
4278 journal->j_max_batch_time = sbi->s_max_batch_time;
4280 write_lock(&journal->j_state_lock);
4281 if (test_opt(sb, BARRIER))
4282 journal->j_flags |= JBD2_BARRIER;
4284 journal->j_flags &= ~JBD2_BARRIER;
4285 if (test_opt(sb, DATA_ERR_ABORT))
4286 journal->j_flags |= JBD2_ABORT_ON_SYNCDATA_ERR;
4288 journal->j_flags &= ~JBD2_ABORT_ON_SYNCDATA_ERR;
4289 write_unlock(&journal->j_state_lock);
4292 static journal_t *ext4_get_journal(struct super_block *sb,
4293 unsigned int journal_inum)
4295 struct inode *journal_inode;
4298 BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL));
4300 /* First, test for the existence of a valid inode on disk. Bad
4301 * things happen if we iget() an unused inode, as the subsequent
4302 * iput() will try to delete it. */
4304 journal_inode = ext4_iget(sb, journal_inum);
4305 if (IS_ERR(journal_inode)) {
4306 ext4_msg(sb, KERN_ERR, "no journal found");
4309 if (!journal_inode->i_nlink) {
4310 make_bad_inode(journal_inode);
4311 iput(journal_inode);
4312 ext4_msg(sb, KERN_ERR, "journal inode is deleted");
4316 jbd_debug(2, "Journal inode found at %p: %lld bytes\n",
4317 journal_inode, journal_inode->i_size);
4318 if (!S_ISREG(journal_inode->i_mode)) {
4319 ext4_msg(sb, KERN_ERR, "invalid journal inode");
4320 iput(journal_inode);
4324 journal = jbd2_journal_init_inode(journal_inode);
4326 ext4_msg(sb, KERN_ERR, "Could not load journal inode");
4327 iput(journal_inode);
4330 journal->j_private = sb;
4331 ext4_init_journal_params(sb, journal);
4335 static journal_t *ext4_get_dev_journal(struct super_block *sb,
4338 struct buffer_head *bh;
4342 int hblock, blocksize;
4343 ext4_fsblk_t sb_block;
4344 unsigned long offset;
4345 struct ext4_super_block *es;
4346 struct block_device *bdev;
4348 BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL));
4350 bdev = ext4_blkdev_get(j_dev, sb);
4354 blocksize = sb->s_blocksize;
4355 hblock = bdev_logical_block_size(bdev);
4356 if (blocksize < hblock) {
4357 ext4_msg(sb, KERN_ERR,
4358 "blocksize too small for journal device");
4362 sb_block = EXT4_MIN_BLOCK_SIZE / blocksize;
4363 offset = EXT4_MIN_BLOCK_SIZE % blocksize;
4364 set_blocksize(bdev, blocksize);
4365 if (!(bh = __bread(bdev, sb_block, blocksize))) {
4366 ext4_msg(sb, KERN_ERR, "couldn't read superblock of "
4367 "external journal");
4371 es = (struct ext4_super_block *) (bh->b_data + offset);
4372 if ((le16_to_cpu(es->s_magic) != EXT4_SUPER_MAGIC) ||
4373 !(le32_to_cpu(es->s_feature_incompat) &
4374 EXT4_FEATURE_INCOMPAT_JOURNAL_DEV)) {
4375 ext4_msg(sb, KERN_ERR, "external journal has "
4381 if ((le32_to_cpu(es->s_feature_ro_compat) &
4382 EXT4_FEATURE_RO_COMPAT_METADATA_CSUM) &&
4383 es->s_checksum != ext4_superblock_csum(sb, es)) {
4384 ext4_msg(sb, KERN_ERR, "external journal has "
4385 "corrupt superblock");
4390 if (memcmp(EXT4_SB(sb)->s_es->s_journal_uuid, es->s_uuid, 16)) {
4391 ext4_msg(sb, KERN_ERR, "journal UUID does not match");
4396 len = ext4_blocks_count(es);
4397 start = sb_block + 1;
4398 brelse(bh); /* we're done with the superblock */
4400 journal = jbd2_journal_init_dev(bdev, sb->s_bdev,
4401 start, len, blocksize);
4403 ext4_msg(sb, KERN_ERR, "failed to create device journal");
4406 journal->j_private = sb;
4407 ll_rw_block(READ | REQ_META | REQ_PRIO, 1, &journal->j_sb_buffer);
4408 wait_on_buffer(journal->j_sb_buffer);
4409 if (!buffer_uptodate(journal->j_sb_buffer)) {
4410 ext4_msg(sb, KERN_ERR, "I/O error on journal device");
4413 if (be32_to_cpu(journal->j_superblock->s_nr_users) != 1) {
4414 ext4_msg(sb, KERN_ERR, "External journal has more than one "
4415 "user (unsupported) - %d",
4416 be32_to_cpu(journal->j_superblock->s_nr_users));
4419 EXT4_SB(sb)->journal_bdev = bdev;
4420 ext4_init_journal_params(sb, journal);
4424 jbd2_journal_destroy(journal);
4426 ext4_blkdev_put(bdev);
4430 static int ext4_load_journal(struct super_block *sb,
4431 struct ext4_super_block *es,
4432 unsigned long journal_devnum)
4435 unsigned int journal_inum = le32_to_cpu(es->s_journal_inum);
4438 int really_read_only;
4440 BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL));
4442 if (journal_devnum &&
4443 journal_devnum != le32_to_cpu(es->s_journal_dev)) {
4444 ext4_msg(sb, KERN_INFO, "external journal device major/minor "
4445 "numbers have changed");
4446 journal_dev = new_decode_dev(journal_devnum);
4448 journal_dev = new_decode_dev(le32_to_cpu(es->s_journal_dev));
4450 really_read_only = bdev_read_only(sb->s_bdev);
4453 * Are we loading a blank journal or performing recovery after a
4454 * crash? For recovery, we need to check in advance whether we
4455 * can get read-write access to the device.
4457 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER)) {
4458 if (sb->s_flags & MS_RDONLY) {
4459 ext4_msg(sb, KERN_INFO, "INFO: recovery "
4460 "required on readonly filesystem");
4461 if (really_read_only) {
4462 ext4_msg(sb, KERN_ERR, "write access "
4463 "unavailable, cannot proceed");
4466 ext4_msg(sb, KERN_INFO, "write access will "
4467 "be enabled during recovery");
4471 if (journal_inum && journal_dev) {
4472 ext4_msg(sb, KERN_ERR, "filesystem has both journal "
4473 "and inode journals!");
4478 if (!(journal = ext4_get_journal(sb, journal_inum)))
4481 if (!(journal = ext4_get_dev_journal(sb, journal_dev)))
4485 if (!(journal->j_flags & JBD2_BARRIER))
4486 ext4_msg(sb, KERN_INFO, "barriers disabled");
4488 if (!EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER))
4489 err = jbd2_journal_wipe(journal, !really_read_only);
4491 char *save = kmalloc(EXT4_S_ERR_LEN, GFP_KERNEL);
4493 memcpy(save, ((char *) es) +
4494 EXT4_S_ERR_START, EXT4_S_ERR_LEN);
4495 err = jbd2_journal_load(journal);
4497 memcpy(((char *) es) + EXT4_S_ERR_START,
4498 save, EXT4_S_ERR_LEN);
4503 ext4_msg(sb, KERN_ERR, "error loading journal");
4504 jbd2_journal_destroy(journal);
4508 EXT4_SB(sb)->s_journal = journal;
4509 ext4_clear_journal_err(sb, es);
4511 if (!really_read_only && journal_devnum &&
4512 journal_devnum != le32_to_cpu(es->s_journal_dev)) {
4513 es->s_journal_dev = cpu_to_le32(journal_devnum);
4515 /* Make sure we flush the recovery flag to disk. */
4516 ext4_commit_super(sb, 1);
4522 static int ext4_commit_super(struct super_block *sb, int sync)
4524 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
4525 struct buffer_head *sbh = EXT4_SB(sb)->s_sbh;
4528 if (!sbh || block_device_ejected(sb))
4530 if (buffer_write_io_error(sbh)) {
4532 * Oh, dear. A previous attempt to write the
4533 * superblock failed. This could happen because the
4534 * USB device was yanked out. Or it could happen to
4535 * be a transient write error and maybe the block will
4536 * be remapped. Nothing we can do but to retry the
4537 * write and hope for the best.
4539 ext4_msg(sb, KERN_ERR, "previous I/O error to "
4540 "superblock detected");
4541 clear_buffer_write_io_error(sbh);
4542 set_buffer_uptodate(sbh);
4545 * If the file system is mounted read-only, don't update the
4546 * superblock write time. This avoids updating the superblock
4547 * write time when we are mounting the root file system
4548 * read/only but we need to replay the journal; at that point,
4549 * for people who are east of GMT and who make their clock
4550 * tick in localtime for Windows bug-for-bug compatibility,
4551 * the clock is set in the future, and this will cause e2fsck
4552 * to complain and force a full file system check.
4554 if (!(sb->s_flags & MS_RDONLY))
4555 es->s_wtime = cpu_to_le32(get_seconds());
4556 if (sb->s_bdev->bd_part)
4557 es->s_kbytes_written =
4558 cpu_to_le64(EXT4_SB(sb)->s_kbytes_written +
4559 ((part_stat_read(sb->s_bdev->bd_part, sectors[1]) -
4560 EXT4_SB(sb)->s_sectors_written_start) >> 1));
4562 es->s_kbytes_written =
4563 cpu_to_le64(EXT4_SB(sb)->s_kbytes_written);
4564 if (percpu_counter_initialized(&EXT4_SB(sb)->s_freeclusters_counter))
4565 ext4_free_blocks_count_set(es,
4566 EXT4_C2B(EXT4_SB(sb), percpu_counter_sum_positive(
4567 &EXT4_SB(sb)->s_freeclusters_counter)));
4568 if (percpu_counter_initialized(&EXT4_SB(sb)->s_freeinodes_counter))
4569 es->s_free_inodes_count =
4570 cpu_to_le32(percpu_counter_sum_positive(
4571 &EXT4_SB(sb)->s_freeinodes_counter));
4572 BUFFER_TRACE(sbh, "marking dirty");
4573 ext4_superblock_csum_set(sb);
4574 mark_buffer_dirty(sbh);
4576 error = sync_dirty_buffer(sbh);
4580 error = buffer_write_io_error(sbh);
4582 ext4_msg(sb, KERN_ERR, "I/O error while writing "
4584 clear_buffer_write_io_error(sbh);
4585 set_buffer_uptodate(sbh);
4592 * Have we just finished recovery? If so, and if we are mounting (or
4593 * remounting) the filesystem readonly, then we will end up with a
4594 * consistent fs on disk. Record that fact.
4596 static void ext4_mark_recovery_complete(struct super_block *sb,
4597 struct ext4_super_block *es)
4599 journal_t *journal = EXT4_SB(sb)->s_journal;
4601 if (!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL)) {
4602 BUG_ON(journal != NULL);
4605 jbd2_journal_lock_updates(journal);
4606 if (jbd2_journal_flush(journal) < 0)
4609 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER) &&
4610 sb->s_flags & MS_RDONLY) {
4611 EXT4_CLEAR_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
4612 ext4_commit_super(sb, 1);
4616 jbd2_journal_unlock_updates(journal);
4620 * If we are mounting (or read-write remounting) a filesystem whose journal
4621 * has recorded an error from a previous lifetime, move that error to the
4622 * main filesystem now.
4624 static void ext4_clear_journal_err(struct super_block *sb,
4625 struct ext4_super_block *es)
4631 BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL));
4633 journal = EXT4_SB(sb)->s_journal;
4636 * Now check for any error status which may have been recorded in the
4637 * journal by a prior ext4_error() or ext4_abort()
4640 j_errno = jbd2_journal_errno(journal);
4644 errstr = ext4_decode_error(sb, j_errno, nbuf);
4645 ext4_warning(sb, "Filesystem error recorded "
4646 "from previous mount: %s", errstr);
4647 ext4_warning(sb, "Marking fs in need of filesystem check.");
4649 EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS;
4650 es->s_state |= cpu_to_le16(EXT4_ERROR_FS);
4651 ext4_commit_super(sb, 1);
4653 jbd2_journal_clear_err(journal);
4654 jbd2_journal_update_sb_errno(journal);
4659 * Force the running and committing transactions to commit,
4660 * and wait on the commit.
4662 int ext4_force_commit(struct super_block *sb)
4666 if (sb->s_flags & MS_RDONLY)
4669 journal = EXT4_SB(sb)->s_journal;
4670 return ext4_journal_force_commit(journal);
4673 static int ext4_sync_fs(struct super_block *sb, int wait)
4677 bool needs_barrier = false;
4678 struct ext4_sb_info *sbi = EXT4_SB(sb);
4680 trace_ext4_sync_fs(sb, wait);
4681 flush_workqueue(sbi->rsv_conversion_wq);
4683 * Writeback quota in non-journalled quota case - journalled quota has
4686 dquot_writeback_dquots(sb, -1);
4688 * Data writeback is possible w/o journal transaction, so barrier must
4689 * being sent at the end of the function. But we can skip it if
4690 * transaction_commit will do it for us.
4692 if (sbi->s_journal) {
4693 target = jbd2_get_latest_transaction(sbi->s_journal);
4694 if (wait && sbi->s_journal->j_flags & JBD2_BARRIER &&
4695 !jbd2_trans_will_send_data_barrier(sbi->s_journal, target))
4696 needs_barrier = true;
4698 if (jbd2_journal_start_commit(sbi->s_journal, &target)) {
4700 ret = jbd2_log_wait_commit(sbi->s_journal,
4703 } else if (wait && test_opt(sb, BARRIER))
4704 needs_barrier = true;
4705 if (needs_barrier) {
4707 err = blkdev_issue_flush(sb->s_bdev, GFP_KERNEL, NULL);
4716 * LVM calls this function before a (read-only) snapshot is created. This
4717 * gives us a chance to flush the journal completely and mark the fs clean.
4719 * Note that only this function cannot bring a filesystem to be in a clean
4720 * state independently. It relies on upper layer to stop all data & metadata
4723 static int ext4_freeze(struct super_block *sb)
4728 if (sb->s_flags & MS_RDONLY)
4731 journal = EXT4_SB(sb)->s_journal;
4734 /* Now we set up the journal barrier. */
4735 jbd2_journal_lock_updates(journal);
4738 * Don't clear the needs_recovery flag if we failed to
4739 * flush the journal.
4741 error = jbd2_journal_flush(journal);
4746 /* Journal blocked and flushed, clear needs_recovery flag. */
4747 EXT4_CLEAR_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
4748 error = ext4_commit_super(sb, 1);
4751 /* we rely on upper layer to stop further updates */
4752 jbd2_journal_unlock_updates(journal);
4757 * Called by LVM after the snapshot is done. We need to reset the RECOVER
4758 * flag here, even though the filesystem is not technically dirty yet.
4760 static int ext4_unfreeze(struct super_block *sb)
4762 if (sb->s_flags & MS_RDONLY)
4765 /* Reset the needs_recovery flag before the fs is unlocked. */
4766 EXT4_SET_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
4767 ext4_commit_super(sb, 1);
4772 * Structure to save mount options for ext4_remount's benefit
4774 struct ext4_mount_options {
4775 unsigned long s_mount_opt;
4776 unsigned long s_mount_opt2;
4779 unsigned long s_commit_interval;
4780 u32 s_min_batch_time, s_max_batch_time;
4783 char *s_qf_names[EXT4_MAXQUOTAS];
4787 static int ext4_remount(struct super_block *sb, int *flags, char *data)
4789 struct ext4_super_block *es;
4790 struct ext4_sb_info *sbi = EXT4_SB(sb);
4791 unsigned long old_sb_flags;
4792 struct ext4_mount_options old_opts;
4793 int enable_quota = 0;
4795 unsigned int journal_ioprio = DEFAULT_JOURNAL_IOPRIO;
4800 char *orig_data = kstrdup(data, GFP_KERNEL);
4802 /* Store the original options */
4803 old_sb_flags = sb->s_flags;
4804 old_opts.s_mount_opt = sbi->s_mount_opt;
4805 old_opts.s_mount_opt2 = sbi->s_mount_opt2;
4806 old_opts.s_resuid = sbi->s_resuid;
4807 old_opts.s_resgid = sbi->s_resgid;
4808 old_opts.s_commit_interval = sbi->s_commit_interval;
4809 old_opts.s_min_batch_time = sbi->s_min_batch_time;
4810 old_opts.s_max_batch_time = sbi->s_max_batch_time;
4812 old_opts.s_jquota_fmt = sbi->s_jquota_fmt;
4813 for (i = 0; i < EXT4_MAXQUOTAS; i++)
4814 if (sbi->s_qf_names[i]) {
4815 old_opts.s_qf_names[i] = kstrdup(sbi->s_qf_names[i],
4817 if (!old_opts.s_qf_names[i]) {
4818 for (j = 0; j < i; j++)
4819 kfree(old_opts.s_qf_names[j]);
4824 old_opts.s_qf_names[i] = NULL;
4826 if (sbi->s_journal && sbi->s_journal->j_task->io_context)
4827 journal_ioprio = sbi->s_journal->j_task->io_context->ioprio;
4830 * Allow the "check" option to be passed as a remount option.
4832 if (!parse_options(data, sb, NULL, &journal_ioprio, 1)) {
4837 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA) {
4838 if (test_opt2(sb, EXPLICIT_DELALLOC)) {
4839 ext4_msg(sb, KERN_ERR, "can't mount with "
4840 "both data=journal and delalloc");
4844 if (test_opt(sb, DIOREAD_NOLOCK)) {
4845 ext4_msg(sb, KERN_ERR, "can't mount with "
4846 "both data=journal and dioread_nolock");
4852 if (sbi->s_mount_flags & EXT4_MF_FS_ABORTED)
4853 ext4_abort(sb, "Abort forced by user");
4855 sb->s_flags = (sb->s_flags & ~MS_POSIXACL) |
4856 (test_opt(sb, POSIX_ACL) ? MS_POSIXACL : 0);
4860 if (sbi->s_journal) {
4861 ext4_init_journal_params(sb, sbi->s_journal);
4862 set_task_ioprio(sbi->s_journal->j_task, journal_ioprio);
4865 if ((*flags & MS_RDONLY) != (sb->s_flags & MS_RDONLY)) {
4866 if (sbi->s_mount_flags & EXT4_MF_FS_ABORTED) {
4871 if (*flags & MS_RDONLY) {
4872 err = sync_filesystem(sb);
4875 err = dquot_suspend(sb, -1);
4880 * First of all, the unconditional stuff we have to do
4881 * to disable replay of the journal when we next remount
4883 sb->s_flags |= MS_RDONLY;
4886 * OK, test if we are remounting a valid rw partition
4887 * readonly, and if so set the rdonly flag and then
4888 * mark the partition as valid again.
4890 if (!(es->s_state & cpu_to_le16(EXT4_VALID_FS)) &&
4891 (sbi->s_mount_state & EXT4_VALID_FS))
4892 es->s_state = cpu_to_le16(sbi->s_mount_state);
4895 ext4_mark_recovery_complete(sb, es);
4897 /* Make sure we can mount this feature set readwrite */
4898 if (!ext4_feature_set_ok(sb, 0)) {
4903 * Make sure the group descriptor checksums
4904 * are sane. If they aren't, refuse to remount r/w.
4906 for (g = 0; g < sbi->s_groups_count; g++) {
4907 struct ext4_group_desc *gdp =
4908 ext4_get_group_desc(sb, g, NULL);
4910 if (!ext4_group_desc_csum_verify(sb, g, gdp)) {
4911 ext4_msg(sb, KERN_ERR,
4912 "ext4_remount: Checksum for group %u failed (%u!=%u)",
4913 g, le16_to_cpu(ext4_group_desc_csum(sbi, g, gdp)),
4914 le16_to_cpu(gdp->bg_checksum));
4921 * If we have an unprocessed orphan list hanging
4922 * around from a previously readonly bdev mount,
4923 * require a full umount/remount for now.
4925 if (es->s_last_orphan) {
4926 ext4_msg(sb, KERN_WARNING, "Couldn't "
4927 "remount RDWR because of unprocessed "
4928 "orphan inode list. Please "
4929 "umount/remount instead");
4935 * Mounting a RDONLY partition read-write, so reread
4936 * and store the current valid flag. (It may have
4937 * been changed by e2fsck since we originally mounted
4941 ext4_clear_journal_err(sb, es);
4942 sbi->s_mount_state = le16_to_cpu(es->s_state);
4943 if (!ext4_setup_super(sb, es, 0))
4944 sb->s_flags &= ~MS_RDONLY;
4945 if (EXT4_HAS_INCOMPAT_FEATURE(sb,
4946 EXT4_FEATURE_INCOMPAT_MMP))
4947 if (ext4_multi_mount_protect(sb,
4948 le64_to_cpu(es->s_mmp_block))) {
4957 * Reinitialize lazy itable initialization thread based on
4960 if ((sb->s_flags & MS_RDONLY) || !test_opt(sb, INIT_INODE_TABLE))
4961 ext4_unregister_li_request(sb);
4963 ext4_group_t first_not_zeroed;
4964 first_not_zeroed = ext4_has_uninit_itable(sb);
4965 ext4_register_li_request(sb, first_not_zeroed);
4968 ext4_setup_system_zone(sb);
4969 if (sbi->s_journal == NULL && !(old_sb_flags & MS_RDONLY))
4970 ext4_commit_super(sb, 1);
4973 /* Release old quota file names */
4974 for (i = 0; i < EXT4_MAXQUOTAS; i++)
4975 kfree(old_opts.s_qf_names[i]);
4977 if (sb_any_quota_suspended(sb))
4978 dquot_resume(sb, -1);
4979 else if (EXT4_HAS_RO_COMPAT_FEATURE(sb,
4980 EXT4_FEATURE_RO_COMPAT_QUOTA)) {
4981 err = ext4_enable_quotas(sb);
4988 ext4_msg(sb, KERN_INFO, "re-mounted. Opts: %s", orig_data);
4993 sb->s_flags = old_sb_flags;
4994 sbi->s_mount_opt = old_opts.s_mount_opt;
4995 sbi->s_mount_opt2 = old_opts.s_mount_opt2;
4996 sbi->s_resuid = old_opts.s_resuid;
4997 sbi->s_resgid = old_opts.s_resgid;
4998 sbi->s_commit_interval = old_opts.s_commit_interval;
4999 sbi->s_min_batch_time = old_opts.s_min_batch_time;
5000 sbi->s_max_batch_time = old_opts.s_max_batch_time;
5002 sbi->s_jquota_fmt = old_opts.s_jquota_fmt;
5003 for (i = 0; i < EXT4_MAXQUOTAS; i++) {
5004 kfree(sbi->s_qf_names[i]);
5005 sbi->s_qf_names[i] = old_opts.s_qf_names[i];
5012 static int ext4_statfs(struct dentry *dentry, struct kstatfs *buf)
5014 struct super_block *sb = dentry->d_sb;
5015 struct ext4_sb_info *sbi = EXT4_SB(sb);
5016 struct ext4_super_block *es = sbi->s_es;
5017 ext4_fsblk_t overhead = 0, resv_blocks;
5020 resv_blocks = EXT4_C2B(sbi, atomic64_read(&sbi->s_resv_clusters));
5022 if (!test_opt(sb, MINIX_DF))
5023 overhead = sbi->s_overhead;
5025 buf->f_type = EXT4_SUPER_MAGIC;
5026 buf->f_bsize = sb->s_blocksize;
5027 buf->f_blocks = ext4_blocks_count(es) - EXT4_C2B(sbi, overhead);
5028 bfree = percpu_counter_sum_positive(&sbi->s_freeclusters_counter) -
5029 percpu_counter_sum_positive(&sbi->s_dirtyclusters_counter);
5030 /* prevent underflow in case that few free space is available */
5031 buf->f_bfree = EXT4_C2B(sbi, max_t(s64, bfree, 0));
5032 buf->f_bavail = buf->f_bfree -
5033 (ext4_r_blocks_count(es) + resv_blocks);
5034 if (buf->f_bfree < (ext4_r_blocks_count(es) + resv_blocks))
5036 buf->f_files = le32_to_cpu(es->s_inodes_count);
5037 buf->f_ffree = percpu_counter_sum_positive(&sbi->s_freeinodes_counter);
5038 buf->f_namelen = EXT4_NAME_LEN;
5039 fsid = le64_to_cpup((void *)es->s_uuid) ^
5040 le64_to_cpup((void *)es->s_uuid + sizeof(u64));
5041 buf->f_fsid.val[0] = fsid & 0xFFFFFFFFUL;
5042 buf->f_fsid.val[1] = (fsid >> 32) & 0xFFFFFFFFUL;
5047 /* Helper function for writing quotas on sync - we need to start transaction
5048 * before quota file is locked for write. Otherwise the are possible deadlocks:
5049 * Process 1 Process 2
5050 * ext4_create() quota_sync()
5051 * jbd2_journal_start() write_dquot()
5052 * dquot_initialize() down(dqio_mutex)
5053 * down(dqio_mutex) jbd2_journal_start()
5059 static inline struct inode *dquot_to_inode(struct dquot *dquot)
5061 return sb_dqopt(dquot->dq_sb)->files[dquot->dq_id.type];
5064 static int ext4_write_dquot(struct dquot *dquot)
5068 struct inode *inode;
5070 inode = dquot_to_inode(dquot);
5071 handle = ext4_journal_start(inode, EXT4_HT_QUOTA,
5072 EXT4_QUOTA_TRANS_BLOCKS(dquot->dq_sb));
5074 return PTR_ERR(handle);
5075 ret = dquot_commit(dquot);
5076 err = ext4_journal_stop(handle);
5082 static int ext4_acquire_dquot(struct dquot *dquot)
5087 handle = ext4_journal_start(dquot_to_inode(dquot), EXT4_HT_QUOTA,
5088 EXT4_QUOTA_INIT_BLOCKS(dquot->dq_sb));
5090 return PTR_ERR(handle);
5091 ret = dquot_acquire(dquot);
5092 err = ext4_journal_stop(handle);
5098 static int ext4_release_dquot(struct dquot *dquot)
5103 handle = ext4_journal_start(dquot_to_inode(dquot), EXT4_HT_QUOTA,
5104 EXT4_QUOTA_DEL_BLOCKS(dquot->dq_sb));
5105 if (IS_ERR(handle)) {
5106 /* Release dquot anyway to avoid endless cycle in dqput() */
5107 dquot_release(dquot);
5108 return PTR_ERR(handle);
5110 ret = dquot_release(dquot);
5111 err = ext4_journal_stop(handle);
5117 static int ext4_mark_dquot_dirty(struct dquot *dquot)
5119 struct super_block *sb = dquot->dq_sb;
5120 struct ext4_sb_info *sbi = EXT4_SB(sb);
5122 /* Are we journaling quotas? */
5123 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_QUOTA) ||
5124 sbi->s_qf_names[USRQUOTA] || sbi->s_qf_names[GRPQUOTA]) {
5125 dquot_mark_dquot_dirty(dquot);
5126 return ext4_write_dquot(dquot);
5128 return dquot_mark_dquot_dirty(dquot);
5132 static int ext4_write_info(struct super_block *sb, int type)
5137 /* Data block + inode block */
5138 handle = ext4_journal_start(sb->s_root->d_inode, EXT4_HT_QUOTA, 2);
5140 return PTR_ERR(handle);
5141 ret = dquot_commit_info(sb, type);
5142 err = ext4_journal_stop(handle);
5149 * Turn on quotas during mount time - we need to find
5150 * the quota file and such...
5152 static int ext4_quota_on_mount(struct super_block *sb, int type)
5154 return dquot_quota_on_mount(sb, EXT4_SB(sb)->s_qf_names[type],
5155 EXT4_SB(sb)->s_jquota_fmt, type);
5159 * Standard function to be called on quota_on
5161 static int ext4_quota_on(struct super_block *sb, int type, int format_id,
5166 if (!test_opt(sb, QUOTA))
5169 /* Quotafile not on the same filesystem? */
5170 if (path->dentry->d_sb != sb)
5172 /* Journaling quota? */
5173 if (EXT4_SB(sb)->s_qf_names[type]) {
5174 /* Quotafile not in fs root? */
5175 if (path->dentry->d_parent != sb->s_root)
5176 ext4_msg(sb, KERN_WARNING,
5177 "Quota file not on filesystem root. "
5178 "Journaled quota will not work");
5182 * When we journal data on quota file, we have to flush journal to see
5183 * all updates to the file when we bypass pagecache...
5185 if (EXT4_SB(sb)->s_journal &&
5186 ext4_should_journal_data(path->dentry->d_inode)) {
5188 * We don't need to lock updates but journal_flush() could
5189 * otherwise be livelocked...
5191 jbd2_journal_lock_updates(EXT4_SB(sb)->s_journal);
5192 err = jbd2_journal_flush(EXT4_SB(sb)->s_journal);
5193 jbd2_journal_unlock_updates(EXT4_SB(sb)->s_journal);
5198 return dquot_quota_on(sb, type, format_id, path);
5201 static int ext4_quota_enable(struct super_block *sb, int type, int format_id,
5205 struct inode *qf_inode;
5206 unsigned long qf_inums[EXT4_MAXQUOTAS] = {
5207 le32_to_cpu(EXT4_SB(sb)->s_es->s_usr_quota_inum),
5208 le32_to_cpu(EXT4_SB(sb)->s_es->s_grp_quota_inum)
5211 BUG_ON(!EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_QUOTA));
5213 if (!qf_inums[type])
5216 qf_inode = ext4_iget(sb, qf_inums[type]);
5217 if (IS_ERR(qf_inode)) {
5218 ext4_error(sb, "Bad quota inode # %lu", qf_inums[type]);
5219 return PTR_ERR(qf_inode);
5222 /* Don't account quota for quota files to avoid recursion */
5223 qf_inode->i_flags |= S_NOQUOTA;
5224 err = dquot_enable(qf_inode, type, format_id, flags);
5230 /* Enable usage tracking for all quota types. */
5231 static int ext4_enable_quotas(struct super_block *sb)
5234 unsigned long qf_inums[EXT4_MAXQUOTAS] = {
5235 le32_to_cpu(EXT4_SB(sb)->s_es->s_usr_quota_inum),
5236 le32_to_cpu(EXT4_SB(sb)->s_es->s_grp_quota_inum)
5239 sb_dqopt(sb)->flags |= DQUOT_QUOTA_SYS_FILE;
5240 for (type = 0; type < EXT4_MAXQUOTAS; type++) {
5241 if (qf_inums[type]) {
5242 err = ext4_quota_enable(sb, type, QFMT_VFS_V1,
5243 DQUOT_USAGE_ENABLED);
5246 "Failed to enable quota tracking "
5247 "(type=%d, err=%d). Please run "
5248 "e2fsck to fix.", type, err);
5257 * quota_on function that is used when QUOTA feature is set.
5259 static int ext4_quota_on_sysfile(struct super_block *sb, int type,
5262 if (!EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_QUOTA))
5266 * USAGE was enabled at mount time. Only need to enable LIMITS now.
5268 return ext4_quota_enable(sb, type, format_id, DQUOT_LIMITS_ENABLED);
5271 static int ext4_quota_off(struct super_block *sb, int type)
5273 struct inode *inode = sb_dqopt(sb)->files[type];
5276 /* Force all delayed allocation blocks to be allocated.
5277 * Caller already holds s_umount sem */
5278 if (test_opt(sb, DELALLOC))
5279 sync_filesystem(sb);
5284 /* Update modification times of quota files when userspace can
5285 * start looking at them */
5286 handle = ext4_journal_start(inode, EXT4_HT_QUOTA, 1);
5289 inode->i_mtime = inode->i_ctime = CURRENT_TIME;
5290 ext4_mark_inode_dirty(handle, inode);
5291 ext4_journal_stop(handle);
5294 return dquot_quota_off(sb, type);
5298 * quota_off function that is used when QUOTA feature is set.
5300 static int ext4_quota_off_sysfile(struct super_block *sb, int type)
5302 if (!EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_QUOTA))
5305 /* Disable only the limits. */
5306 return dquot_disable(sb, type, DQUOT_LIMITS_ENABLED);
5309 /* Read data from quotafile - avoid pagecache and such because we cannot afford
5310 * acquiring the locks... As quota files are never truncated and quota code
5311 * itself serializes the operations (and no one else should touch the files)
5312 * we don't have to be afraid of races */
5313 static ssize_t ext4_quota_read(struct super_block *sb, int type, char *data,
5314 size_t len, loff_t off)
5316 struct inode *inode = sb_dqopt(sb)->files[type];
5317 ext4_lblk_t blk = off >> EXT4_BLOCK_SIZE_BITS(sb);
5318 int offset = off & (sb->s_blocksize - 1);
5321 struct buffer_head *bh;
5322 loff_t i_size = i_size_read(inode);
5326 if (off+len > i_size)
5329 while (toread > 0) {
5330 tocopy = sb->s_blocksize - offset < toread ?
5331 sb->s_blocksize - offset : toread;
5332 bh = ext4_bread(NULL, inode, blk, 0);
5335 if (!bh) /* A hole? */
5336 memset(data, 0, tocopy);
5338 memcpy(data, bh->b_data+offset, tocopy);
5348 /* Write to quotafile (we know the transaction is already started and has
5349 * enough credits) */
5350 static ssize_t ext4_quota_write(struct super_block *sb, int type,
5351 const char *data, size_t len, loff_t off)
5353 struct inode *inode = sb_dqopt(sb)->files[type];
5354 ext4_lblk_t blk = off >> EXT4_BLOCK_SIZE_BITS(sb);
5355 int err, offset = off & (sb->s_blocksize - 1);
5356 struct buffer_head *bh;
5357 handle_t *handle = journal_current_handle();
5359 if (EXT4_SB(sb)->s_journal && !handle) {
5360 ext4_msg(sb, KERN_WARNING, "Quota write (off=%llu, len=%llu)"
5361 " cancelled because transaction is not started",
5362 (unsigned long long)off, (unsigned long long)len);
5366 * Since we account only one data block in transaction credits,
5367 * then it is impossible to cross a block boundary.
5369 if (sb->s_blocksize - offset < len) {
5370 ext4_msg(sb, KERN_WARNING, "Quota write (off=%llu, len=%llu)"
5371 " cancelled because not block aligned",
5372 (unsigned long long)off, (unsigned long long)len);
5376 bh = ext4_bread(handle, inode, blk, 1);
5381 BUFFER_TRACE(bh, "get write access");
5382 err = ext4_journal_get_write_access(handle, bh);
5388 memcpy(bh->b_data+offset, data, len);
5389 flush_dcache_page(bh->b_page);
5391 err = ext4_handle_dirty_metadata(handle, NULL, bh);
5394 if (inode->i_size < off + len) {
5395 i_size_write(inode, off + len);
5396 EXT4_I(inode)->i_disksize = inode->i_size;
5397 ext4_mark_inode_dirty(handle, inode);
5404 static struct dentry *ext4_mount(struct file_system_type *fs_type, int flags,
5405 const char *dev_name, void *data)
5407 return mount_bdev(fs_type, flags, dev_name, data, ext4_fill_super);
5410 #if !defined(CONFIG_EXT2_FS) && !defined(CONFIG_EXT2_FS_MODULE) && defined(CONFIG_EXT4_USE_FOR_EXT23)
5411 static inline void register_as_ext2(void)
5413 int err = register_filesystem(&ext2_fs_type);
5416 "EXT4-fs: Unable to register as ext2 (%d)\n", err);
5419 static inline void unregister_as_ext2(void)
5421 unregister_filesystem(&ext2_fs_type);
5424 static inline int ext2_feature_set_ok(struct super_block *sb)
5426 if (EXT4_HAS_INCOMPAT_FEATURE(sb, ~EXT2_FEATURE_INCOMPAT_SUPP))
5428 if (sb->s_flags & MS_RDONLY)
5430 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, ~EXT2_FEATURE_RO_COMPAT_SUPP))
5435 static inline void register_as_ext2(void) { }
5436 static inline void unregister_as_ext2(void) { }
5437 static inline int ext2_feature_set_ok(struct super_block *sb) { return 0; }
5440 #if !defined(CONFIG_EXT3_FS) && !defined(CONFIG_EXT3_FS_MODULE) && defined(CONFIG_EXT4_USE_FOR_EXT23)
5441 static inline void register_as_ext3(void)
5443 int err = register_filesystem(&ext3_fs_type);
5446 "EXT4-fs: Unable to register as ext3 (%d)\n", err);
5449 static inline void unregister_as_ext3(void)
5451 unregister_filesystem(&ext3_fs_type);
5454 static inline int ext3_feature_set_ok(struct super_block *sb)
5456 if (EXT4_HAS_INCOMPAT_FEATURE(sb, ~EXT3_FEATURE_INCOMPAT_SUPP))
5458 if (!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL))
5460 if (sb->s_flags & MS_RDONLY)
5462 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, ~EXT3_FEATURE_RO_COMPAT_SUPP))
5467 static inline void register_as_ext3(void) { }
5468 static inline void unregister_as_ext3(void) { }
5469 static inline int ext3_feature_set_ok(struct super_block *sb) { return 0; }
5472 static struct file_system_type ext4_fs_type = {
5473 .owner = THIS_MODULE,
5475 .mount = ext4_mount,
5476 .kill_sb = kill_block_super,
5477 .fs_flags = FS_REQUIRES_DEV,
5479 MODULE_ALIAS_FS("ext4");
5481 static int __init ext4_init_feat_adverts(void)
5483 struct ext4_features *ef;
5486 ef = kzalloc(sizeof(struct ext4_features), GFP_KERNEL);
5490 ef->f_kobj.kset = ext4_kset;
5491 init_completion(&ef->f_kobj_unregister);
5492 ret = kobject_init_and_add(&ef->f_kobj, &ext4_feat_ktype, NULL,
5505 static void ext4_exit_feat_adverts(void)
5507 kobject_put(&ext4_feat->f_kobj);
5508 wait_for_completion(&ext4_feat->f_kobj_unregister);
5512 /* Shared across all ext4 file systems */
5513 wait_queue_head_t ext4__ioend_wq[EXT4_WQ_HASH_SZ];
5514 struct mutex ext4__aio_mutex[EXT4_WQ_HASH_SZ];
5516 static int __init ext4_init_fs(void)
5520 ext4_li_info = NULL;
5521 mutex_init(&ext4_li_mtx);
5523 /* Build-time check for flags consistency */
5524 ext4_check_flag_values();
5526 for (i = 0; i < EXT4_WQ_HASH_SZ; i++) {
5527 mutex_init(&ext4__aio_mutex[i]);
5528 init_waitqueue_head(&ext4__ioend_wq[i]);
5531 err = ext4_init_es();
5535 err = ext4_init_pageio();
5539 err = ext4_init_system_zone();
5542 ext4_kset = kset_create_and_add("ext4", NULL, fs_kobj);
5547 ext4_proc_root = proc_mkdir("fs/ext4", NULL);
5549 err = ext4_init_feat_adverts();
5553 err = ext4_init_mballoc();
5557 ext4_mballoc_ready = 1;
5558 err = init_inodecache();
5563 err = register_filesystem(&ext4_fs_type);
5569 unregister_as_ext2();
5570 unregister_as_ext3();
5571 destroy_inodecache();
5573 ext4_mballoc_ready = 0;
5574 ext4_exit_mballoc();
5576 ext4_exit_feat_adverts();
5579 remove_proc_entry("fs/ext4", NULL);
5580 kset_unregister(ext4_kset);
5582 ext4_exit_system_zone();
5591 static void __exit ext4_exit_fs(void)
5593 ext4_destroy_lazyinit_thread();
5594 unregister_as_ext2();
5595 unregister_as_ext3();
5596 unregister_filesystem(&ext4_fs_type);
5597 destroy_inodecache();
5598 ext4_exit_mballoc();
5599 ext4_exit_feat_adverts();
5600 remove_proc_entry("fs/ext4", NULL);
5601 kset_unregister(ext4_kset);
5602 ext4_exit_system_zone();
5607 MODULE_AUTHOR("Remy Card, Stephen Tweedie, Andrew Morton, Andreas Dilger, Theodore Ts'o and others");
5608 MODULE_DESCRIPTION("Fourth Extended Filesystem");
5609 MODULE_LICENSE("GPL");
5610 module_init(ext4_init_fs)
5611 module_exit(ext4_exit_fs)