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 if (!(sbi->s_es->s_feature_ro_compat &
2009 cpu_to_le32(EXT4_FEATURE_RO_COMPAT_GDT_CSUM)))
2012 offset = offsetof(struct ext4_group_desc, bg_checksum);
2014 crc = crc16(~0, sbi->s_es->s_uuid, sizeof(sbi->s_es->s_uuid));
2015 crc = crc16(crc, (__u8 *)&le_group, sizeof(le_group));
2016 crc = crc16(crc, (__u8 *)gdp, offset);
2017 offset += sizeof(gdp->bg_checksum); /* skip checksum */
2018 /* for checksum of struct ext4_group_desc do the rest...*/
2019 if ((sbi->s_es->s_feature_incompat &
2020 cpu_to_le32(EXT4_FEATURE_INCOMPAT_64BIT)) &&
2021 offset < le16_to_cpu(sbi->s_es->s_desc_size))
2022 crc = crc16(crc, (__u8 *)gdp + offset,
2023 le16_to_cpu(sbi->s_es->s_desc_size) -
2027 return cpu_to_le16(crc);
2030 int ext4_group_desc_csum_verify(struct super_block *sb, __u32 block_group,
2031 struct ext4_group_desc *gdp)
2033 if (ext4_has_group_desc_csum(sb) &&
2034 (gdp->bg_checksum != ext4_group_desc_csum(EXT4_SB(sb),
2041 void ext4_group_desc_csum_set(struct super_block *sb, __u32 block_group,
2042 struct ext4_group_desc *gdp)
2044 if (!ext4_has_group_desc_csum(sb))
2046 gdp->bg_checksum = ext4_group_desc_csum(EXT4_SB(sb), block_group, gdp);
2049 /* Called at mount-time, super-block is locked */
2050 static int ext4_check_descriptors(struct super_block *sb,
2051 ext4_group_t *first_not_zeroed)
2053 struct ext4_sb_info *sbi = EXT4_SB(sb);
2054 ext4_fsblk_t first_block = le32_to_cpu(sbi->s_es->s_first_data_block);
2055 ext4_fsblk_t last_block;
2056 ext4_fsblk_t block_bitmap;
2057 ext4_fsblk_t inode_bitmap;
2058 ext4_fsblk_t inode_table;
2059 int flexbg_flag = 0;
2060 ext4_group_t i, grp = sbi->s_groups_count;
2062 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_FLEX_BG))
2065 ext4_debug("Checking group descriptors");
2067 for (i = 0; i < sbi->s_groups_count; i++) {
2068 struct ext4_group_desc *gdp = ext4_get_group_desc(sb, i, NULL);
2070 if (i == sbi->s_groups_count - 1 || flexbg_flag)
2071 last_block = ext4_blocks_count(sbi->s_es) - 1;
2073 last_block = first_block +
2074 (EXT4_BLOCKS_PER_GROUP(sb) - 1);
2076 if ((grp == sbi->s_groups_count) &&
2077 !(gdp->bg_flags & cpu_to_le16(EXT4_BG_INODE_ZEROED)))
2080 block_bitmap = ext4_block_bitmap(sb, gdp);
2081 if (block_bitmap < first_block || block_bitmap > last_block) {
2082 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2083 "Block bitmap for group %u not in group "
2084 "(block %llu)!", i, block_bitmap);
2087 inode_bitmap = ext4_inode_bitmap(sb, gdp);
2088 if (inode_bitmap < first_block || inode_bitmap > last_block) {
2089 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2090 "Inode bitmap for group %u not in group "
2091 "(block %llu)!", i, inode_bitmap);
2094 inode_table = ext4_inode_table(sb, gdp);
2095 if (inode_table < first_block ||
2096 inode_table + sbi->s_itb_per_group - 1 > last_block) {
2097 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2098 "Inode table for group %u not in group "
2099 "(block %llu)!", i, inode_table);
2102 ext4_lock_group(sb, i);
2103 if (!ext4_group_desc_csum_verify(sb, i, gdp)) {
2104 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2105 "Checksum for group %u failed (%u!=%u)",
2106 i, le16_to_cpu(ext4_group_desc_csum(sbi, i,
2107 gdp)), le16_to_cpu(gdp->bg_checksum));
2108 if (!(sb->s_flags & MS_RDONLY)) {
2109 ext4_unlock_group(sb, i);
2113 ext4_unlock_group(sb, i);
2115 first_block += EXT4_BLOCKS_PER_GROUP(sb);
2117 if (NULL != first_not_zeroed)
2118 *first_not_zeroed = grp;
2122 /* ext4_orphan_cleanup() walks a singly-linked list of inodes (starting at
2123 * the superblock) which were deleted from all directories, but held open by
2124 * a process at the time of a crash. We walk the list and try to delete these
2125 * inodes at recovery time (only with a read-write filesystem).
2127 * In order to keep the orphan inode chain consistent during traversal (in
2128 * case of crash during recovery), we link each inode into the superblock
2129 * orphan list_head and handle it the same way as an inode deletion during
2130 * normal operation (which journals the operations for us).
2132 * We only do an iget() and an iput() on each inode, which is very safe if we
2133 * accidentally point at an in-use or already deleted inode. The worst that
2134 * can happen in this case is that we get a "bit already cleared" message from
2135 * ext4_free_inode(). The only reason we would point at a wrong inode is if
2136 * e2fsck was run on this filesystem, and it must have already done the orphan
2137 * inode cleanup for us, so we can safely abort without any further action.
2139 static void ext4_orphan_cleanup(struct super_block *sb,
2140 struct ext4_super_block *es)
2142 unsigned int s_flags = sb->s_flags;
2143 int nr_orphans = 0, nr_truncates = 0;
2147 if (!es->s_last_orphan) {
2148 jbd_debug(4, "no orphan inodes to clean up\n");
2152 if (bdev_read_only(sb->s_bdev)) {
2153 ext4_msg(sb, KERN_ERR, "write access "
2154 "unavailable, skipping orphan cleanup");
2158 /* Check if feature set would not allow a r/w mount */
2159 if (!ext4_feature_set_ok(sb, 0)) {
2160 ext4_msg(sb, KERN_INFO, "Skipping orphan cleanup due to "
2161 "unknown ROCOMPAT features");
2165 if (EXT4_SB(sb)->s_mount_state & EXT4_ERROR_FS) {
2166 /* don't clear list on RO mount w/ errors */
2167 if (es->s_last_orphan && !(s_flags & MS_RDONLY)) {
2168 ext4_msg(sb, KERN_INFO, "Errors on filesystem, "
2169 "clearing orphan list.\n");
2170 es->s_last_orphan = 0;
2172 jbd_debug(1, "Skipping orphan recovery on fs with errors.\n");
2176 if (s_flags & MS_RDONLY) {
2177 ext4_msg(sb, KERN_INFO, "orphan cleanup on readonly fs");
2178 sb->s_flags &= ~MS_RDONLY;
2181 /* Needed for iput() to work correctly and not trash data */
2182 sb->s_flags |= MS_ACTIVE;
2183 /* Turn on quotas so that they are updated correctly */
2184 for (i = 0; i < EXT4_MAXQUOTAS; i++) {
2185 if (EXT4_SB(sb)->s_qf_names[i]) {
2186 int ret = ext4_quota_on_mount(sb, i);
2188 ext4_msg(sb, KERN_ERR,
2189 "Cannot turn on journaled "
2190 "quota: error %d", ret);
2195 while (es->s_last_orphan) {
2196 struct inode *inode;
2198 inode = ext4_orphan_get(sb, le32_to_cpu(es->s_last_orphan));
2199 if (IS_ERR(inode)) {
2200 es->s_last_orphan = 0;
2204 list_add(&EXT4_I(inode)->i_orphan, &EXT4_SB(sb)->s_orphan);
2205 dquot_initialize(inode);
2206 if (inode->i_nlink) {
2207 if (test_opt(sb, DEBUG))
2208 ext4_msg(sb, KERN_DEBUG,
2209 "%s: truncating inode %lu to %lld bytes",
2210 __func__, inode->i_ino, inode->i_size);
2211 jbd_debug(2, "truncating inode %lu to %lld bytes\n",
2212 inode->i_ino, inode->i_size);
2213 mutex_lock(&inode->i_mutex);
2214 truncate_inode_pages(inode->i_mapping, inode->i_size);
2215 ext4_truncate(inode);
2216 mutex_unlock(&inode->i_mutex);
2219 if (test_opt(sb, DEBUG))
2220 ext4_msg(sb, KERN_DEBUG,
2221 "%s: deleting unreferenced inode %lu",
2222 __func__, inode->i_ino);
2223 jbd_debug(2, "deleting unreferenced inode %lu\n",
2227 iput(inode); /* The delete magic happens here! */
2230 #define PLURAL(x) (x), ((x) == 1) ? "" : "s"
2233 ext4_msg(sb, KERN_INFO, "%d orphan inode%s deleted",
2234 PLURAL(nr_orphans));
2236 ext4_msg(sb, KERN_INFO, "%d truncate%s cleaned up",
2237 PLURAL(nr_truncates));
2239 /* Turn quotas off */
2240 for (i = 0; i < EXT4_MAXQUOTAS; i++) {
2241 if (sb_dqopt(sb)->files[i])
2242 dquot_quota_off(sb, i);
2245 sb->s_flags = s_flags; /* Restore MS_RDONLY status */
2249 * Maximal extent format file size.
2250 * Resulting logical blkno at s_maxbytes must fit in our on-disk
2251 * extent format containers, within a sector_t, and within i_blocks
2252 * in the vfs. ext4 inode has 48 bits of i_block in fsblock units,
2253 * so that won't be a limiting factor.
2255 * However there is other limiting factor. We do store extents in the form
2256 * of starting block and length, hence the resulting length of the extent
2257 * covering maximum file size must fit into on-disk format containers as
2258 * well. Given that length is always by 1 unit bigger than max unit (because
2259 * we count 0 as well) we have to lower the s_maxbytes by one fs block.
2261 * Note, this does *not* consider any metadata overhead for vfs i_blocks.
2263 static loff_t ext4_max_size(int blkbits, int has_huge_files)
2266 loff_t upper_limit = MAX_LFS_FILESIZE;
2268 /* small i_blocks in vfs inode? */
2269 if (!has_huge_files || sizeof(blkcnt_t) < sizeof(u64)) {
2271 * CONFIG_LBDAF is not enabled implies the inode
2272 * i_block represent total blocks in 512 bytes
2273 * 32 == size of vfs inode i_blocks * 8
2275 upper_limit = (1LL << 32) - 1;
2277 /* total blocks in file system block size */
2278 upper_limit >>= (blkbits - 9);
2279 upper_limit <<= blkbits;
2283 * 32-bit extent-start container, ee_block. We lower the maxbytes
2284 * by one fs block, so ee_len can cover the extent of maximum file
2287 res = (1LL << 32) - 1;
2290 /* Sanity check against vm- & vfs- imposed limits */
2291 if (res > upper_limit)
2298 * Maximal bitmap file size. There is a direct, and {,double-,triple-}indirect
2299 * block limit, and also a limit of (2^48 - 1) 512-byte sectors in i_blocks.
2300 * We need to be 1 filesystem block less than the 2^48 sector limit.
2302 static loff_t ext4_max_bitmap_size(int bits, int has_huge_files)
2304 loff_t res = EXT4_NDIR_BLOCKS;
2307 /* This is calculated to be the largest file size for a dense, block
2308 * mapped file such that the file's total number of 512-byte sectors,
2309 * including data and all indirect blocks, does not exceed (2^48 - 1).
2311 * __u32 i_blocks_lo and _u16 i_blocks_high represent the total
2312 * number of 512-byte sectors of the file.
2315 if (!has_huge_files || sizeof(blkcnt_t) < sizeof(u64)) {
2317 * !has_huge_files or CONFIG_LBDAF not enabled implies that
2318 * the inode i_block field represents total file blocks in
2319 * 2^32 512-byte sectors == size of vfs inode i_blocks * 8
2321 upper_limit = (1LL << 32) - 1;
2323 /* total blocks in file system block size */
2324 upper_limit >>= (bits - 9);
2328 * We use 48 bit ext4_inode i_blocks
2329 * With EXT4_HUGE_FILE_FL set the i_blocks
2330 * represent total number of blocks in
2331 * file system block size
2333 upper_limit = (1LL << 48) - 1;
2337 /* indirect blocks */
2339 /* double indirect blocks */
2340 meta_blocks += 1 + (1LL << (bits-2));
2341 /* tripple indirect blocks */
2342 meta_blocks += 1 + (1LL << (bits-2)) + (1LL << (2*(bits-2)));
2344 upper_limit -= meta_blocks;
2345 upper_limit <<= bits;
2347 res += 1LL << (bits-2);
2348 res += 1LL << (2*(bits-2));
2349 res += 1LL << (3*(bits-2));
2351 if (res > upper_limit)
2354 if (res > MAX_LFS_FILESIZE)
2355 res = MAX_LFS_FILESIZE;
2360 static ext4_fsblk_t descriptor_loc(struct super_block *sb,
2361 ext4_fsblk_t logical_sb_block, int nr)
2363 struct ext4_sb_info *sbi = EXT4_SB(sb);
2364 ext4_group_t bg, first_meta_bg;
2367 first_meta_bg = le32_to_cpu(sbi->s_es->s_first_meta_bg);
2369 if (!EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_META_BG) ||
2371 return logical_sb_block + nr + 1;
2372 bg = sbi->s_desc_per_block * nr;
2373 if (ext4_bg_has_super(sb, bg))
2377 * If we have a meta_bg fs with 1k blocks, group 0's GDT is at
2378 * block 2, not 1. If s_first_data_block == 0 (bigalloc is enabled
2379 * on modern mke2fs or blksize > 1k on older mke2fs) then we must
2382 if (sb->s_blocksize == 1024 && nr == 0 &&
2383 le32_to_cpu(EXT4_SB(sb)->s_es->s_first_data_block) == 0)
2386 return (has_super + ext4_group_first_block_no(sb, bg));
2390 * ext4_get_stripe_size: Get the stripe size.
2391 * @sbi: In memory super block info
2393 * If we have specified it via mount option, then
2394 * use the mount option value. If the value specified at mount time is
2395 * greater than the blocks per group use the super block value.
2396 * If the super block value is greater than blocks per group return 0.
2397 * Allocator needs it be less than blocks per group.
2400 static unsigned long ext4_get_stripe_size(struct ext4_sb_info *sbi)
2402 unsigned long stride = le16_to_cpu(sbi->s_es->s_raid_stride);
2403 unsigned long stripe_width =
2404 le32_to_cpu(sbi->s_es->s_raid_stripe_width);
2407 if (sbi->s_stripe && sbi->s_stripe <= sbi->s_blocks_per_group)
2408 ret = sbi->s_stripe;
2409 else if (stripe_width <= sbi->s_blocks_per_group)
2411 else if (stride <= sbi->s_blocks_per_group)
2417 * If the stripe width is 1, this makes no sense and
2418 * we set it to 0 to turn off stripe handling code.
2429 struct attribute attr;
2430 ssize_t (*show)(struct ext4_attr *, struct ext4_sb_info *, char *);
2431 ssize_t (*store)(struct ext4_attr *, struct ext4_sb_info *,
2432 const char *, size_t);
2439 static int parse_strtoull(const char *buf,
2440 unsigned long long max, unsigned long long *value)
2444 ret = kstrtoull(skip_spaces(buf), 0, value);
2445 if (!ret && *value > max)
2450 static ssize_t delayed_allocation_blocks_show(struct ext4_attr *a,
2451 struct ext4_sb_info *sbi,
2454 return snprintf(buf, PAGE_SIZE, "%llu\n",
2456 percpu_counter_sum(&sbi->s_dirtyclusters_counter)));
2459 static ssize_t session_write_kbytes_show(struct ext4_attr *a,
2460 struct ext4_sb_info *sbi, char *buf)
2462 struct super_block *sb = sbi->s_buddy_cache->i_sb;
2464 if (!sb->s_bdev->bd_part)
2465 return snprintf(buf, PAGE_SIZE, "0\n");
2466 return snprintf(buf, PAGE_SIZE, "%lu\n",
2467 (part_stat_read(sb->s_bdev->bd_part, sectors[1]) -
2468 sbi->s_sectors_written_start) >> 1);
2471 static ssize_t lifetime_write_kbytes_show(struct ext4_attr *a,
2472 struct ext4_sb_info *sbi, char *buf)
2474 struct super_block *sb = sbi->s_buddy_cache->i_sb;
2476 if (!sb->s_bdev->bd_part)
2477 return snprintf(buf, PAGE_SIZE, "0\n");
2478 return snprintf(buf, PAGE_SIZE, "%llu\n",
2479 (unsigned long long)(sbi->s_kbytes_written +
2480 ((part_stat_read(sb->s_bdev->bd_part, sectors[1]) -
2481 EXT4_SB(sb)->s_sectors_written_start) >> 1)));
2484 static ssize_t inode_readahead_blks_store(struct ext4_attr *a,
2485 struct ext4_sb_info *sbi,
2486 const char *buf, size_t count)
2491 ret = kstrtoul(skip_spaces(buf), 0, &t);
2495 if (t && (!is_power_of_2(t) || t > 0x40000000))
2498 sbi->s_inode_readahead_blks = t;
2502 static ssize_t sbi_ui_show(struct ext4_attr *a,
2503 struct ext4_sb_info *sbi, char *buf)
2505 unsigned int *ui = (unsigned int *) (((char *) sbi) + a->u.offset);
2507 return snprintf(buf, PAGE_SIZE, "%u\n", *ui);
2510 static ssize_t sbi_ui_store(struct ext4_attr *a,
2511 struct ext4_sb_info *sbi,
2512 const char *buf, size_t count)
2514 unsigned int *ui = (unsigned int *) (((char *) sbi) + a->u.offset);
2518 ret = kstrtoul(skip_spaces(buf), 0, &t);
2525 static ssize_t es_ui_show(struct ext4_attr *a,
2526 struct ext4_sb_info *sbi, char *buf)
2529 unsigned int *ui = (unsigned int *) (((char *) sbi->s_es) +
2532 return snprintf(buf, PAGE_SIZE, "%u\n", *ui);
2535 static ssize_t reserved_clusters_show(struct ext4_attr *a,
2536 struct ext4_sb_info *sbi, char *buf)
2538 return snprintf(buf, PAGE_SIZE, "%llu\n",
2539 (unsigned long long) atomic64_read(&sbi->s_resv_clusters));
2542 static ssize_t reserved_clusters_store(struct ext4_attr *a,
2543 struct ext4_sb_info *sbi,
2544 const char *buf, size_t count)
2546 unsigned long long val;
2549 if (parse_strtoull(buf, -1ULL, &val))
2551 ret = ext4_reserve_clusters(sbi, val);
2553 return ret ? ret : count;
2556 static ssize_t trigger_test_error(struct ext4_attr *a,
2557 struct ext4_sb_info *sbi,
2558 const char *buf, size_t count)
2562 if (!capable(CAP_SYS_ADMIN))
2565 if (len && buf[len-1] == '\n')
2569 ext4_error(sbi->s_sb, "%.*s", len, buf);
2573 static ssize_t sbi_deprecated_show(struct ext4_attr *a,
2574 struct ext4_sb_info *sbi, char *buf)
2576 return snprintf(buf, PAGE_SIZE, "%d\n", a->u.deprecated_val);
2579 #define EXT4_ATTR_OFFSET(_name,_mode,_show,_store,_elname) \
2580 static struct ext4_attr ext4_attr_##_name = { \
2581 .attr = {.name = __stringify(_name), .mode = _mode }, \
2585 .offset = offsetof(struct ext4_sb_info, _elname),\
2589 #define EXT4_ATTR_OFFSET_ES(_name,_mode,_show,_store,_elname) \
2590 static struct ext4_attr ext4_attr_##_name = { \
2591 .attr = {.name = __stringify(_name), .mode = _mode }, \
2595 .offset = offsetof(struct ext4_super_block, _elname), \
2599 #define EXT4_ATTR(name, mode, show, store) \
2600 static struct ext4_attr ext4_attr_##name = __ATTR(name, mode, show, store)
2602 #define EXT4_INFO_ATTR(name) EXT4_ATTR(name, 0444, NULL, NULL)
2603 #define EXT4_RO_ATTR(name) EXT4_ATTR(name, 0444, name##_show, NULL)
2604 #define EXT4_RW_ATTR(name) EXT4_ATTR(name, 0644, name##_show, name##_store)
2606 #define EXT4_RO_ATTR_ES_UI(name, elname) \
2607 EXT4_ATTR_OFFSET_ES(name, 0444, es_ui_show, NULL, elname)
2608 #define EXT4_RW_ATTR_SBI_UI(name, elname) \
2609 EXT4_ATTR_OFFSET(name, 0644, sbi_ui_show, sbi_ui_store, elname)
2611 #define ATTR_LIST(name) &ext4_attr_##name.attr
2612 #define EXT4_DEPRECATED_ATTR(_name, _val) \
2613 static struct ext4_attr ext4_attr_##_name = { \
2614 .attr = {.name = __stringify(_name), .mode = 0444 }, \
2615 .show = sbi_deprecated_show, \
2617 .deprecated_val = _val, \
2621 EXT4_RO_ATTR(delayed_allocation_blocks);
2622 EXT4_RO_ATTR(session_write_kbytes);
2623 EXT4_RO_ATTR(lifetime_write_kbytes);
2624 EXT4_RW_ATTR(reserved_clusters);
2625 EXT4_ATTR_OFFSET(inode_readahead_blks, 0644, sbi_ui_show,
2626 inode_readahead_blks_store, s_inode_readahead_blks);
2627 EXT4_RW_ATTR_SBI_UI(inode_goal, s_inode_goal);
2628 EXT4_RW_ATTR_SBI_UI(mb_stats, s_mb_stats);
2629 EXT4_RW_ATTR_SBI_UI(mb_max_to_scan, s_mb_max_to_scan);
2630 EXT4_RW_ATTR_SBI_UI(mb_min_to_scan, s_mb_min_to_scan);
2631 EXT4_RW_ATTR_SBI_UI(mb_order2_req, s_mb_order2_reqs);
2632 EXT4_RW_ATTR_SBI_UI(mb_stream_req, s_mb_stream_request);
2633 EXT4_RW_ATTR_SBI_UI(mb_group_prealloc, s_mb_group_prealloc);
2634 EXT4_DEPRECATED_ATTR(max_writeback_mb_bump, 128);
2635 EXT4_RW_ATTR_SBI_UI(extent_max_zeroout_kb, s_extent_max_zeroout_kb);
2636 EXT4_ATTR(trigger_fs_error, 0200, NULL, trigger_test_error);
2637 EXT4_RW_ATTR_SBI_UI(err_ratelimit_interval_ms, s_err_ratelimit_state.interval);
2638 EXT4_RW_ATTR_SBI_UI(err_ratelimit_burst, s_err_ratelimit_state.burst);
2639 EXT4_RW_ATTR_SBI_UI(warning_ratelimit_interval_ms, s_warning_ratelimit_state.interval);
2640 EXT4_RW_ATTR_SBI_UI(warning_ratelimit_burst, s_warning_ratelimit_state.burst);
2641 EXT4_RW_ATTR_SBI_UI(msg_ratelimit_interval_ms, s_msg_ratelimit_state.interval);
2642 EXT4_RW_ATTR_SBI_UI(msg_ratelimit_burst, s_msg_ratelimit_state.burst);
2643 EXT4_RO_ATTR_ES_UI(errors_count, s_error_count);
2644 EXT4_RO_ATTR_ES_UI(first_error_time, s_first_error_time);
2645 EXT4_RO_ATTR_ES_UI(last_error_time, s_last_error_time);
2647 static struct attribute *ext4_attrs[] = {
2648 ATTR_LIST(delayed_allocation_blocks),
2649 ATTR_LIST(session_write_kbytes),
2650 ATTR_LIST(lifetime_write_kbytes),
2651 ATTR_LIST(reserved_clusters),
2652 ATTR_LIST(inode_readahead_blks),
2653 ATTR_LIST(inode_goal),
2654 ATTR_LIST(mb_stats),
2655 ATTR_LIST(mb_max_to_scan),
2656 ATTR_LIST(mb_min_to_scan),
2657 ATTR_LIST(mb_order2_req),
2658 ATTR_LIST(mb_stream_req),
2659 ATTR_LIST(mb_group_prealloc),
2660 ATTR_LIST(max_writeback_mb_bump),
2661 ATTR_LIST(extent_max_zeroout_kb),
2662 ATTR_LIST(trigger_fs_error),
2663 ATTR_LIST(err_ratelimit_interval_ms),
2664 ATTR_LIST(err_ratelimit_burst),
2665 ATTR_LIST(warning_ratelimit_interval_ms),
2666 ATTR_LIST(warning_ratelimit_burst),
2667 ATTR_LIST(msg_ratelimit_interval_ms),
2668 ATTR_LIST(msg_ratelimit_burst),
2669 ATTR_LIST(errors_count),
2670 ATTR_LIST(first_error_time),
2671 ATTR_LIST(last_error_time),
2675 /* Features this copy of ext4 supports */
2676 EXT4_INFO_ATTR(lazy_itable_init);
2677 EXT4_INFO_ATTR(batched_discard);
2678 EXT4_INFO_ATTR(meta_bg_resize);
2680 static struct attribute *ext4_feat_attrs[] = {
2681 ATTR_LIST(lazy_itable_init),
2682 ATTR_LIST(batched_discard),
2683 ATTR_LIST(meta_bg_resize),
2687 static ssize_t ext4_attr_show(struct kobject *kobj,
2688 struct attribute *attr, char *buf)
2690 struct ext4_sb_info *sbi = container_of(kobj, struct ext4_sb_info,
2692 struct ext4_attr *a = container_of(attr, struct ext4_attr, attr);
2694 return a->show ? a->show(a, sbi, buf) : 0;
2697 static ssize_t ext4_attr_store(struct kobject *kobj,
2698 struct attribute *attr,
2699 const char *buf, size_t len)
2701 struct ext4_sb_info *sbi = container_of(kobj, struct ext4_sb_info,
2703 struct ext4_attr *a = container_of(attr, struct ext4_attr, attr);
2705 return a->store ? a->store(a, sbi, buf, len) : 0;
2708 static void ext4_sb_release(struct kobject *kobj)
2710 struct ext4_sb_info *sbi = container_of(kobj, struct ext4_sb_info,
2712 complete(&sbi->s_kobj_unregister);
2715 static const struct sysfs_ops ext4_attr_ops = {
2716 .show = ext4_attr_show,
2717 .store = ext4_attr_store,
2720 static struct kobj_type ext4_ktype = {
2721 .default_attrs = ext4_attrs,
2722 .sysfs_ops = &ext4_attr_ops,
2723 .release = ext4_sb_release,
2726 static void ext4_feat_release(struct kobject *kobj)
2728 complete(&ext4_feat->f_kobj_unregister);
2731 static ssize_t ext4_feat_show(struct kobject *kobj,
2732 struct attribute *attr, char *buf)
2734 return snprintf(buf, PAGE_SIZE, "supported\n");
2738 * We can not use ext4_attr_show/store because it relies on the kobject
2739 * being embedded in the ext4_sb_info structure which is definitely not
2740 * true in this case.
2742 static const struct sysfs_ops ext4_feat_ops = {
2743 .show = ext4_feat_show,
2747 static struct kobj_type ext4_feat_ktype = {
2748 .default_attrs = ext4_feat_attrs,
2749 .sysfs_ops = &ext4_feat_ops,
2750 .release = ext4_feat_release,
2754 * Check whether this filesystem can be mounted based on
2755 * the features present and the RDONLY/RDWR mount requested.
2756 * Returns 1 if this filesystem can be mounted as requested,
2757 * 0 if it cannot be.
2759 static int ext4_feature_set_ok(struct super_block *sb, int readonly)
2761 if (EXT4_HAS_INCOMPAT_FEATURE(sb, ~EXT4_FEATURE_INCOMPAT_SUPP)) {
2762 ext4_msg(sb, KERN_ERR,
2763 "Couldn't mount because of "
2764 "unsupported optional features (%x)",
2765 (le32_to_cpu(EXT4_SB(sb)->s_es->s_feature_incompat) &
2766 ~EXT4_FEATURE_INCOMPAT_SUPP));
2773 /* Check that feature set is OK for a read-write mount */
2774 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, ~EXT4_FEATURE_RO_COMPAT_SUPP)) {
2775 ext4_msg(sb, KERN_ERR, "couldn't mount RDWR because of "
2776 "unsupported optional features (%x)",
2777 (le32_to_cpu(EXT4_SB(sb)->s_es->s_feature_ro_compat) &
2778 ~EXT4_FEATURE_RO_COMPAT_SUPP));
2782 * Large file size enabled file system can only be mounted
2783 * read-write on 32-bit systems if kernel is built with CONFIG_LBDAF
2785 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_HUGE_FILE)) {
2786 if (sizeof(blkcnt_t) < sizeof(u64)) {
2787 ext4_msg(sb, KERN_ERR, "Filesystem with huge files "
2788 "cannot be mounted RDWR without "
2793 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_BIGALLOC) &&
2794 !EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_EXTENTS)) {
2795 ext4_msg(sb, KERN_ERR,
2796 "Can't support bigalloc feature without "
2797 "extents feature\n");
2801 #ifndef CONFIG_QUOTA
2802 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_QUOTA) &&
2804 ext4_msg(sb, KERN_ERR,
2805 "Filesystem with quota feature cannot be mounted RDWR "
2806 "without CONFIG_QUOTA");
2809 #endif /* CONFIG_QUOTA */
2814 * This function is called once a day if we have errors logged
2815 * on the file system
2817 static void print_daily_error_info(unsigned long arg)
2819 struct super_block *sb = (struct super_block *) arg;
2820 struct ext4_sb_info *sbi;
2821 struct ext4_super_block *es;
2826 if (es->s_error_count)
2827 /* fsck newer than v1.41.13 is needed to clean this condition. */
2828 ext4_msg(sb, KERN_NOTICE, "error count since last fsck: %u",
2829 le32_to_cpu(es->s_error_count));
2830 if (es->s_first_error_time) {
2831 printk(KERN_NOTICE "EXT4-fs (%s): initial error at time %u: %.*s:%d",
2832 sb->s_id, le32_to_cpu(es->s_first_error_time),
2833 (int) sizeof(es->s_first_error_func),
2834 es->s_first_error_func,
2835 le32_to_cpu(es->s_first_error_line));
2836 if (es->s_first_error_ino)
2837 printk(": inode %u",
2838 le32_to_cpu(es->s_first_error_ino));
2839 if (es->s_first_error_block)
2840 printk(": block %llu", (unsigned long long)
2841 le64_to_cpu(es->s_first_error_block));
2844 if (es->s_last_error_time) {
2845 printk(KERN_NOTICE "EXT4-fs (%s): last error at time %u: %.*s:%d",
2846 sb->s_id, le32_to_cpu(es->s_last_error_time),
2847 (int) sizeof(es->s_last_error_func),
2848 es->s_last_error_func,
2849 le32_to_cpu(es->s_last_error_line));
2850 if (es->s_last_error_ino)
2851 printk(": inode %u",
2852 le32_to_cpu(es->s_last_error_ino));
2853 if (es->s_last_error_block)
2854 printk(": block %llu", (unsigned long long)
2855 le64_to_cpu(es->s_last_error_block));
2858 mod_timer(&sbi->s_err_report, jiffies + 24*60*60*HZ); /* Once a day */
2861 /* Find next suitable group and run ext4_init_inode_table */
2862 static int ext4_run_li_request(struct ext4_li_request *elr)
2864 struct ext4_group_desc *gdp = NULL;
2865 ext4_group_t group, ngroups;
2866 struct super_block *sb;
2867 unsigned long timeout = 0;
2871 ngroups = EXT4_SB(sb)->s_groups_count;
2874 for (group = elr->lr_next_group; group < ngroups; group++) {
2875 gdp = ext4_get_group_desc(sb, group, NULL);
2881 if (!(gdp->bg_flags & cpu_to_le16(EXT4_BG_INODE_ZEROED)))
2885 if (group >= ngroups)
2890 ret = ext4_init_inode_table(sb, group,
2891 elr->lr_timeout ? 0 : 1);
2892 if (elr->lr_timeout == 0) {
2893 timeout = (jiffies - timeout) *
2894 elr->lr_sbi->s_li_wait_mult;
2895 elr->lr_timeout = timeout;
2897 elr->lr_next_sched = jiffies + elr->lr_timeout;
2898 elr->lr_next_group = group + 1;
2906 * Remove lr_request from the list_request and free the
2907 * request structure. Should be called with li_list_mtx held
2909 static void ext4_remove_li_request(struct ext4_li_request *elr)
2911 struct ext4_sb_info *sbi;
2918 list_del(&elr->lr_request);
2919 sbi->s_li_request = NULL;
2923 static void ext4_unregister_li_request(struct super_block *sb)
2925 mutex_lock(&ext4_li_mtx);
2926 if (!ext4_li_info) {
2927 mutex_unlock(&ext4_li_mtx);
2931 mutex_lock(&ext4_li_info->li_list_mtx);
2932 ext4_remove_li_request(EXT4_SB(sb)->s_li_request);
2933 mutex_unlock(&ext4_li_info->li_list_mtx);
2934 mutex_unlock(&ext4_li_mtx);
2937 static struct task_struct *ext4_lazyinit_task;
2940 * This is the function where ext4lazyinit thread lives. It walks
2941 * through the request list searching for next scheduled filesystem.
2942 * When such a fs is found, run the lazy initialization request
2943 * (ext4_rn_li_request) and keep track of the time spend in this
2944 * function. Based on that time we compute next schedule time of
2945 * the request. When walking through the list is complete, compute
2946 * next waking time and put itself into sleep.
2948 static int ext4_lazyinit_thread(void *arg)
2950 struct ext4_lazy_init *eli = (struct ext4_lazy_init *)arg;
2951 struct list_head *pos, *n;
2952 struct ext4_li_request *elr;
2953 unsigned long next_wakeup, cur;
2955 BUG_ON(NULL == eli);
2959 next_wakeup = MAX_JIFFY_OFFSET;
2961 mutex_lock(&eli->li_list_mtx);
2962 if (list_empty(&eli->li_request_list)) {
2963 mutex_unlock(&eli->li_list_mtx);
2967 list_for_each_safe(pos, n, &eli->li_request_list) {
2968 elr = list_entry(pos, struct ext4_li_request,
2971 if (time_after_eq(jiffies, elr->lr_next_sched)) {
2972 if (ext4_run_li_request(elr) != 0) {
2973 /* error, remove the lazy_init job */
2974 ext4_remove_li_request(elr);
2979 if (time_before(elr->lr_next_sched, next_wakeup))
2980 next_wakeup = elr->lr_next_sched;
2982 mutex_unlock(&eli->li_list_mtx);
2987 if ((time_after_eq(cur, next_wakeup)) ||
2988 (MAX_JIFFY_OFFSET == next_wakeup)) {
2993 schedule_timeout_interruptible(next_wakeup - cur);
2995 if (kthread_should_stop()) {
2996 ext4_clear_request_list();
3003 * It looks like the request list is empty, but we need
3004 * to check it under the li_list_mtx lock, to prevent any
3005 * additions into it, and of course we should lock ext4_li_mtx
3006 * to atomically free the list and ext4_li_info, because at
3007 * this point another ext4 filesystem could be registering
3010 mutex_lock(&ext4_li_mtx);
3011 mutex_lock(&eli->li_list_mtx);
3012 if (!list_empty(&eli->li_request_list)) {
3013 mutex_unlock(&eli->li_list_mtx);
3014 mutex_unlock(&ext4_li_mtx);
3017 mutex_unlock(&eli->li_list_mtx);
3018 kfree(ext4_li_info);
3019 ext4_li_info = NULL;
3020 mutex_unlock(&ext4_li_mtx);
3025 static void ext4_clear_request_list(void)
3027 struct list_head *pos, *n;
3028 struct ext4_li_request *elr;
3030 mutex_lock(&ext4_li_info->li_list_mtx);
3031 list_for_each_safe(pos, n, &ext4_li_info->li_request_list) {
3032 elr = list_entry(pos, struct ext4_li_request,
3034 ext4_remove_li_request(elr);
3036 mutex_unlock(&ext4_li_info->li_list_mtx);
3039 static int ext4_run_lazyinit_thread(void)
3041 ext4_lazyinit_task = kthread_run(ext4_lazyinit_thread,
3042 ext4_li_info, "ext4lazyinit");
3043 if (IS_ERR(ext4_lazyinit_task)) {
3044 int err = PTR_ERR(ext4_lazyinit_task);
3045 ext4_clear_request_list();
3046 kfree(ext4_li_info);
3047 ext4_li_info = NULL;
3048 printk(KERN_CRIT "EXT4-fs: error %d creating inode table "
3049 "initialization thread\n",
3053 ext4_li_info->li_state |= EXT4_LAZYINIT_RUNNING;
3058 * Check whether it make sense to run itable init. thread or not.
3059 * If there is at least one uninitialized inode table, return
3060 * corresponding group number, else the loop goes through all
3061 * groups and return total number of groups.
3063 static ext4_group_t ext4_has_uninit_itable(struct super_block *sb)
3065 ext4_group_t group, ngroups = EXT4_SB(sb)->s_groups_count;
3066 struct ext4_group_desc *gdp = NULL;
3068 for (group = 0; group < ngroups; group++) {
3069 gdp = ext4_get_group_desc(sb, group, NULL);
3073 if (!(gdp->bg_flags & cpu_to_le16(EXT4_BG_INODE_ZEROED)))
3080 static int ext4_li_info_new(void)
3082 struct ext4_lazy_init *eli = NULL;
3084 eli = kzalloc(sizeof(*eli), GFP_KERNEL);
3088 INIT_LIST_HEAD(&eli->li_request_list);
3089 mutex_init(&eli->li_list_mtx);
3091 eli->li_state |= EXT4_LAZYINIT_QUIT;
3098 static struct ext4_li_request *ext4_li_request_new(struct super_block *sb,
3101 struct ext4_sb_info *sbi = EXT4_SB(sb);
3102 struct ext4_li_request *elr;
3104 elr = kzalloc(sizeof(*elr), GFP_KERNEL);
3110 elr->lr_next_group = start;
3113 * Randomize first schedule time of the request to
3114 * spread the inode table initialization requests
3117 elr->lr_next_sched = jiffies + (prandom_u32() %
3118 (EXT4_DEF_LI_MAX_START_DELAY * HZ));
3122 int ext4_register_li_request(struct super_block *sb,
3123 ext4_group_t first_not_zeroed)
3125 struct ext4_sb_info *sbi = EXT4_SB(sb);
3126 struct ext4_li_request *elr = NULL;
3127 ext4_group_t ngroups = EXT4_SB(sb)->s_groups_count;
3130 mutex_lock(&ext4_li_mtx);
3131 if (sbi->s_li_request != NULL) {
3133 * Reset timeout so it can be computed again, because
3134 * s_li_wait_mult might have changed.
3136 sbi->s_li_request->lr_timeout = 0;
3140 if (first_not_zeroed == ngroups ||
3141 (sb->s_flags & MS_RDONLY) ||
3142 !test_opt(sb, INIT_INODE_TABLE))
3145 elr = ext4_li_request_new(sb, first_not_zeroed);
3151 if (NULL == ext4_li_info) {
3152 ret = ext4_li_info_new();
3157 mutex_lock(&ext4_li_info->li_list_mtx);
3158 list_add(&elr->lr_request, &ext4_li_info->li_request_list);
3159 mutex_unlock(&ext4_li_info->li_list_mtx);
3161 sbi->s_li_request = elr;
3163 * set elr to NULL here since it has been inserted to
3164 * the request_list and the removal and free of it is
3165 * handled by ext4_clear_request_list from now on.
3169 if (!(ext4_li_info->li_state & EXT4_LAZYINIT_RUNNING)) {
3170 ret = ext4_run_lazyinit_thread();
3175 mutex_unlock(&ext4_li_mtx);
3182 * We do not need to lock anything since this is called on
3185 static void ext4_destroy_lazyinit_thread(void)
3188 * If thread exited earlier
3189 * there's nothing to be done.
3191 if (!ext4_li_info || !ext4_lazyinit_task)
3194 kthread_stop(ext4_lazyinit_task);
3197 static int set_journal_csum_feature_set(struct super_block *sb)
3200 int compat, incompat;
3201 struct ext4_sb_info *sbi = EXT4_SB(sb);
3203 if (ext4_has_metadata_csum(sb)) {
3204 /* journal checksum v3 */
3206 incompat = JBD2_FEATURE_INCOMPAT_CSUM_V3;
3208 /* journal checksum v1 */
3209 compat = JBD2_FEATURE_COMPAT_CHECKSUM;
3213 jbd2_journal_clear_features(sbi->s_journal,
3214 JBD2_FEATURE_COMPAT_CHECKSUM, 0,
3215 JBD2_FEATURE_INCOMPAT_CSUM_V3 |
3216 JBD2_FEATURE_INCOMPAT_CSUM_V2);
3217 if (test_opt(sb, JOURNAL_ASYNC_COMMIT)) {
3218 ret = jbd2_journal_set_features(sbi->s_journal,
3220 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT |
3222 } else if (test_opt(sb, JOURNAL_CHECKSUM)) {
3223 ret = jbd2_journal_set_features(sbi->s_journal,
3226 jbd2_journal_clear_features(sbi->s_journal, 0, 0,
3227 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT);
3229 jbd2_journal_clear_features(sbi->s_journal, 0, 0,
3230 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT);
3237 * Note: calculating the overhead so we can be compatible with
3238 * historical BSD practice is quite difficult in the face of
3239 * clusters/bigalloc. This is because multiple metadata blocks from
3240 * different block group can end up in the same allocation cluster.
3241 * Calculating the exact overhead in the face of clustered allocation
3242 * requires either O(all block bitmaps) in memory or O(number of block
3243 * groups**2) in time. We will still calculate the superblock for
3244 * older file systems --- and if we come across with a bigalloc file
3245 * system with zero in s_overhead_clusters the estimate will be close to
3246 * correct especially for very large cluster sizes --- but for newer
3247 * file systems, it's better to calculate this figure once at mkfs
3248 * time, and store it in the superblock. If the superblock value is
3249 * present (even for non-bigalloc file systems), we will use it.
3251 static int count_overhead(struct super_block *sb, ext4_group_t grp,
3254 struct ext4_sb_info *sbi = EXT4_SB(sb);
3255 struct ext4_group_desc *gdp;
3256 ext4_fsblk_t first_block, last_block, b;
3257 ext4_group_t i, ngroups = ext4_get_groups_count(sb);
3258 int s, j, count = 0;
3260 if (!EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_BIGALLOC))
3261 return (ext4_bg_has_super(sb, grp) + ext4_bg_num_gdb(sb, grp) +
3262 sbi->s_itb_per_group + 2);
3264 first_block = le32_to_cpu(sbi->s_es->s_first_data_block) +
3265 (grp * EXT4_BLOCKS_PER_GROUP(sb));
3266 last_block = first_block + EXT4_BLOCKS_PER_GROUP(sb) - 1;
3267 for (i = 0; i < ngroups; i++) {
3268 gdp = ext4_get_group_desc(sb, i, NULL);
3269 b = ext4_block_bitmap(sb, gdp);
3270 if (b >= first_block && b <= last_block) {
3271 ext4_set_bit(EXT4_B2C(sbi, b - first_block), buf);
3274 b = ext4_inode_bitmap(sb, gdp);
3275 if (b >= first_block && b <= last_block) {
3276 ext4_set_bit(EXT4_B2C(sbi, b - first_block), buf);
3279 b = ext4_inode_table(sb, gdp);
3280 if (b >= first_block && b + sbi->s_itb_per_group <= last_block)
3281 for (j = 0; j < sbi->s_itb_per_group; j++, b++) {
3282 int c = EXT4_B2C(sbi, b - first_block);
3283 ext4_set_bit(c, buf);
3289 if (ext4_bg_has_super(sb, grp)) {
3290 ext4_set_bit(s++, buf);
3293 for (j = ext4_bg_num_gdb(sb, grp); j > 0; j--) {
3294 ext4_set_bit(EXT4_B2C(sbi, s++), buf);
3300 return EXT4_CLUSTERS_PER_GROUP(sb) -
3301 ext4_count_free(buf, EXT4_CLUSTERS_PER_GROUP(sb) / 8);
3305 * Compute the overhead and stash it in sbi->s_overhead
3307 int ext4_calculate_overhead(struct super_block *sb)
3309 struct ext4_sb_info *sbi = EXT4_SB(sb);
3310 struct ext4_super_block *es = sbi->s_es;
3311 ext4_group_t i, ngroups = ext4_get_groups_count(sb);
3312 ext4_fsblk_t overhead = 0;
3313 char *buf = (char *) get_zeroed_page(GFP_KERNEL);
3319 * Compute the overhead (FS structures). This is constant
3320 * for a given filesystem unless the number of block groups
3321 * changes so we cache the previous value until it does.
3325 * All of the blocks before first_data_block are overhead
3327 overhead = EXT4_B2C(sbi, le32_to_cpu(es->s_first_data_block));
3330 * Add the overhead found in each block group
3332 for (i = 0; i < ngroups; i++) {
3335 blks = count_overhead(sb, i, buf);
3338 memset(buf, 0, PAGE_SIZE);
3341 /* Add the journal blocks as well */
3343 overhead += EXT4_NUM_B2C(sbi, sbi->s_journal->j_maxlen);
3345 sbi->s_overhead = overhead;
3347 free_page((unsigned long) buf);
3352 static ext4_fsblk_t ext4_calculate_resv_clusters(struct super_block *sb)
3354 ext4_fsblk_t resv_clusters;
3357 * There's no need to reserve anything when we aren't using extents.
3358 * The space estimates are exact, there are no unwritten extents,
3359 * hole punching doesn't need new metadata... This is needed especially
3360 * to keep ext2/3 backward compatibility.
3362 if (!EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_EXTENTS))
3365 * By default we reserve 2% or 4096 clusters, whichever is smaller.
3366 * This should cover the situations where we can not afford to run
3367 * out of space like for example punch hole, or converting
3368 * unwritten extents in delalloc path. In most cases such
3369 * allocation would require 1, or 2 blocks, higher numbers are
3372 resv_clusters = ext4_blocks_count(EXT4_SB(sb)->s_es) >>
3373 EXT4_SB(sb)->s_cluster_bits;
3375 do_div(resv_clusters, 50);
3376 resv_clusters = min_t(ext4_fsblk_t, resv_clusters, 4096);
3378 return resv_clusters;
3382 static int ext4_reserve_clusters(struct ext4_sb_info *sbi, ext4_fsblk_t count)
3384 ext4_fsblk_t clusters = ext4_blocks_count(sbi->s_es) >>
3385 sbi->s_cluster_bits;
3387 if (count >= clusters)
3390 atomic64_set(&sbi->s_resv_clusters, count);
3394 static int ext4_fill_super(struct super_block *sb, void *data, int silent)
3396 char *orig_data = kstrdup(data, GFP_KERNEL);
3397 struct buffer_head *bh;
3398 struct ext4_super_block *es = NULL;
3399 struct ext4_sb_info *sbi;
3401 ext4_fsblk_t sb_block = get_sb_block(&data);
3402 ext4_fsblk_t logical_sb_block;
3403 unsigned long offset = 0;
3404 unsigned long journal_devnum = 0;
3405 unsigned long def_mount_opts;
3410 int blocksize, clustersize;
3411 unsigned int db_count;
3413 int needs_recovery, has_huge_files, has_bigalloc;
3416 unsigned int journal_ioprio = DEFAULT_JOURNAL_IOPRIO;
3417 ext4_group_t first_not_zeroed;
3419 sbi = kzalloc(sizeof(*sbi), GFP_KERNEL);
3423 sbi->s_blockgroup_lock =
3424 kzalloc(sizeof(struct blockgroup_lock), GFP_KERNEL);
3425 if (!sbi->s_blockgroup_lock) {
3429 sb->s_fs_info = sbi;
3431 sbi->s_inode_readahead_blks = EXT4_DEF_INODE_READAHEAD_BLKS;
3432 sbi->s_sb_block = sb_block;
3433 if (sb->s_bdev->bd_part)
3434 sbi->s_sectors_written_start =
3435 part_stat_read(sb->s_bdev->bd_part, sectors[1]);
3437 /* Cleanup superblock name */
3438 for (cp = sb->s_id; (cp = strchr(cp, '/'));)
3441 /* -EINVAL is default */
3443 blocksize = sb_min_blocksize(sb, EXT4_MIN_BLOCK_SIZE);
3445 ext4_msg(sb, KERN_ERR, "unable to set blocksize");
3450 * The ext4 superblock will not be buffer aligned for other than 1kB
3451 * block sizes. We need to calculate the offset from buffer start.
3453 if (blocksize != EXT4_MIN_BLOCK_SIZE) {
3454 logical_sb_block = sb_block * EXT4_MIN_BLOCK_SIZE;
3455 offset = do_div(logical_sb_block, blocksize);
3457 logical_sb_block = sb_block;
3460 if (!(bh = sb_bread_unmovable(sb, logical_sb_block))) {
3461 ext4_msg(sb, KERN_ERR, "unable to read superblock");
3465 * Note: s_es must be initialized as soon as possible because
3466 * some ext4 macro-instructions depend on its value
3468 es = (struct ext4_super_block *) (bh->b_data + offset);
3470 sb->s_magic = le16_to_cpu(es->s_magic);
3471 if (sb->s_magic != EXT4_SUPER_MAGIC)
3473 sbi->s_kbytes_written = le64_to_cpu(es->s_kbytes_written);
3475 /* Warn if metadata_csum and gdt_csum are both set. */
3476 if (EXT4_HAS_RO_COMPAT_FEATURE(sb,
3477 EXT4_FEATURE_RO_COMPAT_METADATA_CSUM) &&
3478 EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_GDT_CSUM))
3479 ext4_warning(sb, KERN_INFO "metadata_csum and uninit_bg are "
3480 "redundant flags; please run fsck.");
3482 /* Check for a known checksum algorithm */
3483 if (!ext4_verify_csum_type(sb, es)) {
3484 ext4_msg(sb, KERN_ERR, "VFS: Found ext4 filesystem with "
3485 "unknown checksum algorithm.");
3490 /* Load the checksum driver */
3491 if (EXT4_HAS_RO_COMPAT_FEATURE(sb,
3492 EXT4_FEATURE_RO_COMPAT_METADATA_CSUM)) {
3493 sbi->s_chksum_driver = crypto_alloc_shash("crc32c", 0, 0);
3494 if (IS_ERR(sbi->s_chksum_driver)) {
3495 ext4_msg(sb, KERN_ERR, "Cannot load crc32c driver.");
3496 ret = PTR_ERR(sbi->s_chksum_driver);
3497 sbi->s_chksum_driver = NULL;
3502 /* Check superblock checksum */
3503 if (!ext4_superblock_csum_verify(sb, es)) {
3504 ext4_msg(sb, KERN_ERR, "VFS: Found ext4 filesystem with "
3505 "invalid superblock checksum. Run e2fsck?");
3510 /* Precompute checksum seed for all metadata */
3511 if (ext4_has_metadata_csum(sb))
3512 sbi->s_csum_seed = ext4_chksum(sbi, ~0, es->s_uuid,
3513 sizeof(es->s_uuid));
3515 /* Set defaults before we parse the mount options */
3516 def_mount_opts = le32_to_cpu(es->s_default_mount_opts);
3517 set_opt(sb, INIT_INODE_TABLE);
3518 if (def_mount_opts & EXT4_DEFM_DEBUG)
3520 if (def_mount_opts & EXT4_DEFM_BSDGROUPS)
3522 if (def_mount_opts & EXT4_DEFM_UID16)
3523 set_opt(sb, NO_UID32);
3524 /* xattr user namespace & acls are now defaulted on */
3525 set_opt(sb, XATTR_USER);
3526 #ifdef CONFIG_EXT4_FS_POSIX_ACL
3527 set_opt(sb, POSIX_ACL);
3529 /* don't forget to enable journal_csum when metadata_csum is enabled. */
3530 if (ext4_has_metadata_csum(sb))
3531 set_opt(sb, JOURNAL_CHECKSUM);
3533 if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_DATA)
3534 set_opt(sb, JOURNAL_DATA);
3535 else if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_ORDERED)
3536 set_opt(sb, ORDERED_DATA);
3537 else if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_WBACK)
3538 set_opt(sb, WRITEBACK_DATA);
3540 if (le16_to_cpu(sbi->s_es->s_errors) == EXT4_ERRORS_PANIC)
3541 set_opt(sb, ERRORS_PANIC);
3542 else if (le16_to_cpu(sbi->s_es->s_errors) == EXT4_ERRORS_CONTINUE)
3543 set_opt(sb, ERRORS_CONT);
3545 set_opt(sb, ERRORS_RO);
3546 /* block_validity enabled by default; disable with noblock_validity */
3547 set_opt(sb, BLOCK_VALIDITY);
3548 if (def_mount_opts & EXT4_DEFM_DISCARD)
3549 set_opt(sb, DISCARD);
3551 sbi->s_resuid = make_kuid(&init_user_ns, le16_to_cpu(es->s_def_resuid));
3552 sbi->s_resgid = make_kgid(&init_user_ns, le16_to_cpu(es->s_def_resgid));
3553 sbi->s_commit_interval = JBD2_DEFAULT_MAX_COMMIT_AGE * HZ;
3554 sbi->s_min_batch_time = EXT4_DEF_MIN_BATCH_TIME;
3555 sbi->s_max_batch_time = EXT4_DEF_MAX_BATCH_TIME;
3557 if ((def_mount_opts & EXT4_DEFM_NOBARRIER) == 0)
3558 set_opt(sb, BARRIER);
3561 * enable delayed allocation by default
3562 * Use -o nodelalloc to turn it off
3564 if (!IS_EXT3_SB(sb) && !IS_EXT2_SB(sb) &&
3565 ((def_mount_opts & EXT4_DEFM_NODELALLOC) == 0))
3566 set_opt(sb, DELALLOC);
3569 * set default s_li_wait_mult for lazyinit, for the case there is
3570 * no mount option specified.
3572 sbi->s_li_wait_mult = EXT4_DEF_LI_WAIT_MULT;
3574 if (!parse_options((char *) sbi->s_es->s_mount_opts, sb,
3575 &journal_devnum, &journal_ioprio, 0)) {
3576 ext4_msg(sb, KERN_WARNING,
3577 "failed to parse options in superblock: %s",
3578 sbi->s_es->s_mount_opts);
3580 sbi->s_def_mount_opt = sbi->s_mount_opt;
3581 if (!parse_options((char *) data, sb, &journal_devnum,
3582 &journal_ioprio, 0))
3585 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA) {
3586 printk_once(KERN_WARNING "EXT4-fs: Warning: mounting "
3587 "with data=journal disables delayed "
3588 "allocation and O_DIRECT support!\n");
3589 if (test_opt2(sb, EXPLICIT_DELALLOC)) {
3590 ext4_msg(sb, KERN_ERR, "can't mount with "
3591 "both data=journal and delalloc");
3594 if (test_opt(sb, DIOREAD_NOLOCK)) {
3595 ext4_msg(sb, KERN_ERR, "can't mount with "
3596 "both data=journal and dioread_nolock");
3599 if (test_opt(sb, DELALLOC))
3600 clear_opt(sb, DELALLOC);
3603 sb->s_flags = (sb->s_flags & ~MS_POSIXACL) |
3604 (test_opt(sb, POSIX_ACL) ? MS_POSIXACL : 0);
3606 if (le32_to_cpu(es->s_rev_level) == EXT4_GOOD_OLD_REV &&
3607 (EXT4_HAS_COMPAT_FEATURE(sb, ~0U) ||
3608 EXT4_HAS_RO_COMPAT_FEATURE(sb, ~0U) ||
3609 EXT4_HAS_INCOMPAT_FEATURE(sb, ~0U)))
3610 ext4_msg(sb, KERN_WARNING,
3611 "feature flags set on rev 0 fs, "
3612 "running e2fsck is recommended");
3614 if (es->s_creator_os == cpu_to_le32(EXT4_OS_HURD)) {
3615 set_opt2(sb, HURD_COMPAT);
3616 if (EXT4_HAS_INCOMPAT_FEATURE(sb,
3617 EXT4_FEATURE_INCOMPAT_64BIT)) {
3618 ext4_msg(sb, KERN_ERR,
3619 "The Hurd can't support 64-bit file systems");
3624 if (IS_EXT2_SB(sb)) {
3625 if (ext2_feature_set_ok(sb))
3626 ext4_msg(sb, KERN_INFO, "mounting ext2 file system "
3627 "using the ext4 subsystem");
3629 ext4_msg(sb, KERN_ERR, "couldn't mount as ext2 due "
3630 "to feature incompatibilities");
3635 if (IS_EXT3_SB(sb)) {
3636 if (ext3_feature_set_ok(sb))
3637 ext4_msg(sb, KERN_INFO, "mounting ext3 file system "
3638 "using the ext4 subsystem");
3640 ext4_msg(sb, KERN_ERR, "couldn't mount as ext3 due "
3641 "to feature incompatibilities");
3647 * Check feature flags regardless of the revision level, since we
3648 * previously didn't change the revision level when setting the flags,
3649 * so there is a chance incompat flags are set on a rev 0 filesystem.
3651 if (!ext4_feature_set_ok(sb, (sb->s_flags & MS_RDONLY)))
3654 blocksize = BLOCK_SIZE << le32_to_cpu(es->s_log_block_size);
3655 if (blocksize < EXT4_MIN_BLOCK_SIZE ||
3656 blocksize > EXT4_MAX_BLOCK_SIZE) {
3657 ext4_msg(sb, KERN_ERR,
3658 "Unsupported filesystem blocksize %d", blocksize);
3662 if (sb->s_blocksize != blocksize) {
3663 /* Validate the filesystem blocksize */
3664 if (!sb_set_blocksize(sb, blocksize)) {
3665 ext4_msg(sb, KERN_ERR, "bad block size %d",
3671 logical_sb_block = sb_block * EXT4_MIN_BLOCK_SIZE;
3672 offset = do_div(logical_sb_block, blocksize);
3673 bh = sb_bread_unmovable(sb, logical_sb_block);
3675 ext4_msg(sb, KERN_ERR,
3676 "Can't read superblock on 2nd try");
3679 es = (struct ext4_super_block *)(bh->b_data + offset);
3681 if (es->s_magic != cpu_to_le16(EXT4_SUPER_MAGIC)) {
3682 ext4_msg(sb, KERN_ERR,
3683 "Magic mismatch, very weird!");
3688 has_huge_files = EXT4_HAS_RO_COMPAT_FEATURE(sb,
3689 EXT4_FEATURE_RO_COMPAT_HUGE_FILE);
3690 sbi->s_bitmap_maxbytes = ext4_max_bitmap_size(sb->s_blocksize_bits,
3692 sb->s_maxbytes = ext4_max_size(sb->s_blocksize_bits, has_huge_files);
3694 if (le32_to_cpu(es->s_rev_level) == EXT4_GOOD_OLD_REV) {
3695 sbi->s_inode_size = EXT4_GOOD_OLD_INODE_SIZE;
3696 sbi->s_first_ino = EXT4_GOOD_OLD_FIRST_INO;
3698 sbi->s_inode_size = le16_to_cpu(es->s_inode_size);
3699 sbi->s_first_ino = le32_to_cpu(es->s_first_ino);
3700 if ((sbi->s_inode_size < EXT4_GOOD_OLD_INODE_SIZE) ||
3701 (!is_power_of_2(sbi->s_inode_size)) ||
3702 (sbi->s_inode_size > blocksize)) {
3703 ext4_msg(sb, KERN_ERR,
3704 "unsupported inode size: %d",
3708 if (sbi->s_inode_size > EXT4_GOOD_OLD_INODE_SIZE)
3709 sb->s_time_gran = 1 << (EXT4_EPOCH_BITS - 2);
3712 sbi->s_desc_size = le16_to_cpu(es->s_desc_size);
3713 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_64BIT)) {
3714 if (sbi->s_desc_size < EXT4_MIN_DESC_SIZE_64BIT ||
3715 sbi->s_desc_size > EXT4_MAX_DESC_SIZE ||
3716 !is_power_of_2(sbi->s_desc_size)) {
3717 ext4_msg(sb, KERN_ERR,
3718 "unsupported descriptor size %lu",
3723 sbi->s_desc_size = EXT4_MIN_DESC_SIZE;
3725 sbi->s_blocks_per_group = le32_to_cpu(es->s_blocks_per_group);
3726 sbi->s_inodes_per_group = le32_to_cpu(es->s_inodes_per_group);
3727 if (EXT4_INODE_SIZE(sb) == 0 || EXT4_INODES_PER_GROUP(sb) == 0)
3730 sbi->s_inodes_per_block = blocksize / EXT4_INODE_SIZE(sb);
3731 if (sbi->s_inodes_per_block == 0)
3733 sbi->s_itb_per_group = sbi->s_inodes_per_group /
3734 sbi->s_inodes_per_block;
3735 sbi->s_desc_per_block = blocksize / EXT4_DESC_SIZE(sb);
3737 sbi->s_mount_state = le16_to_cpu(es->s_state);
3738 sbi->s_addr_per_block_bits = ilog2(EXT4_ADDR_PER_BLOCK(sb));
3739 sbi->s_desc_per_block_bits = ilog2(EXT4_DESC_PER_BLOCK(sb));
3741 for (i = 0; i < 4; i++)
3742 sbi->s_hash_seed[i] = le32_to_cpu(es->s_hash_seed[i]);
3743 sbi->s_def_hash_version = es->s_def_hash_version;
3744 if (EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_DIR_INDEX)) {
3745 i = le32_to_cpu(es->s_flags);
3746 if (i & EXT2_FLAGS_UNSIGNED_HASH)
3747 sbi->s_hash_unsigned = 3;
3748 else if ((i & EXT2_FLAGS_SIGNED_HASH) == 0) {
3749 #ifdef __CHAR_UNSIGNED__
3750 if (!(sb->s_flags & MS_RDONLY))
3752 cpu_to_le32(EXT2_FLAGS_UNSIGNED_HASH);
3753 sbi->s_hash_unsigned = 3;
3755 if (!(sb->s_flags & MS_RDONLY))
3757 cpu_to_le32(EXT2_FLAGS_SIGNED_HASH);
3762 /* Handle clustersize */
3763 clustersize = BLOCK_SIZE << le32_to_cpu(es->s_log_cluster_size);
3764 has_bigalloc = EXT4_HAS_RO_COMPAT_FEATURE(sb,
3765 EXT4_FEATURE_RO_COMPAT_BIGALLOC);
3767 if (clustersize < blocksize) {
3768 ext4_msg(sb, KERN_ERR,
3769 "cluster size (%d) smaller than "
3770 "block size (%d)", clustersize, blocksize);
3773 sbi->s_cluster_bits = le32_to_cpu(es->s_log_cluster_size) -
3774 le32_to_cpu(es->s_log_block_size);
3775 sbi->s_clusters_per_group =
3776 le32_to_cpu(es->s_clusters_per_group);
3777 if (sbi->s_clusters_per_group > blocksize * 8) {
3778 ext4_msg(sb, KERN_ERR,
3779 "#clusters per group too big: %lu",
3780 sbi->s_clusters_per_group);
3783 if (sbi->s_blocks_per_group !=
3784 (sbi->s_clusters_per_group * (clustersize / blocksize))) {
3785 ext4_msg(sb, KERN_ERR, "blocks per group (%lu) and "
3786 "clusters per group (%lu) inconsistent",
3787 sbi->s_blocks_per_group,
3788 sbi->s_clusters_per_group);
3792 if (clustersize != blocksize) {
3793 ext4_warning(sb, "fragment/cluster size (%d) != "
3794 "block size (%d)", clustersize,
3796 clustersize = blocksize;
3798 if (sbi->s_blocks_per_group > blocksize * 8) {
3799 ext4_msg(sb, KERN_ERR,
3800 "#blocks per group too big: %lu",
3801 sbi->s_blocks_per_group);
3804 sbi->s_clusters_per_group = sbi->s_blocks_per_group;
3805 sbi->s_cluster_bits = 0;
3807 sbi->s_cluster_ratio = clustersize / blocksize;
3809 if (sbi->s_inodes_per_group > blocksize * 8) {
3810 ext4_msg(sb, KERN_ERR,
3811 "#inodes per group too big: %lu",
3812 sbi->s_inodes_per_group);
3816 /* Do we have standard group size of clustersize * 8 blocks ? */
3817 if (sbi->s_blocks_per_group == clustersize << 3)
3818 set_opt2(sb, STD_GROUP_SIZE);
3821 * Test whether we have more sectors than will fit in sector_t,
3822 * and whether the max offset is addressable by the page cache.
3824 err = generic_check_addressable(sb->s_blocksize_bits,
3825 ext4_blocks_count(es));
3827 ext4_msg(sb, KERN_ERR, "filesystem"
3828 " too large to mount safely on this system");
3829 if (sizeof(sector_t) < 8)
3830 ext4_msg(sb, KERN_WARNING, "CONFIG_LBDAF not enabled");
3834 if (EXT4_BLOCKS_PER_GROUP(sb) == 0)
3837 /* check blocks count against device size */
3838 blocks_count = sb->s_bdev->bd_inode->i_size >> sb->s_blocksize_bits;
3839 if (blocks_count && ext4_blocks_count(es) > blocks_count) {
3840 ext4_msg(sb, KERN_WARNING, "bad geometry: block count %llu "
3841 "exceeds size of device (%llu blocks)",
3842 ext4_blocks_count(es), blocks_count);
3847 * It makes no sense for the first data block to be beyond the end
3848 * of the filesystem.
3850 if (le32_to_cpu(es->s_first_data_block) >= ext4_blocks_count(es)) {
3851 ext4_msg(sb, KERN_WARNING, "bad geometry: first data "
3852 "block %u is beyond end of filesystem (%llu)",
3853 le32_to_cpu(es->s_first_data_block),
3854 ext4_blocks_count(es));
3857 blocks_count = (ext4_blocks_count(es) -
3858 le32_to_cpu(es->s_first_data_block) +
3859 EXT4_BLOCKS_PER_GROUP(sb) - 1);
3860 do_div(blocks_count, EXT4_BLOCKS_PER_GROUP(sb));
3861 if (blocks_count > ((uint64_t)1<<32) - EXT4_DESC_PER_BLOCK(sb)) {
3862 ext4_msg(sb, KERN_WARNING, "groups count too large: %u "
3863 "(block count %llu, first data block %u, "
3864 "blocks per group %lu)", sbi->s_groups_count,
3865 ext4_blocks_count(es),
3866 le32_to_cpu(es->s_first_data_block),
3867 EXT4_BLOCKS_PER_GROUP(sb));
3870 sbi->s_groups_count = blocks_count;
3871 sbi->s_blockfile_groups = min_t(ext4_group_t, sbi->s_groups_count,
3872 (EXT4_MAX_BLOCK_FILE_PHYS / EXT4_BLOCKS_PER_GROUP(sb)));
3873 db_count = (sbi->s_groups_count + EXT4_DESC_PER_BLOCK(sb) - 1) /
3874 EXT4_DESC_PER_BLOCK(sb);
3875 sbi->s_group_desc = ext4_kvmalloc(db_count *
3876 sizeof(struct buffer_head *),
3878 if (sbi->s_group_desc == NULL) {
3879 ext4_msg(sb, KERN_ERR, "not enough memory");
3885 sbi->s_proc = proc_mkdir(sb->s_id, ext4_proc_root);
3888 proc_create_data("options", S_IRUGO, sbi->s_proc,
3889 &ext4_seq_options_fops, sb);
3891 bgl_lock_init(sbi->s_blockgroup_lock);
3893 for (i = 0; i < db_count; i++) {
3894 block = descriptor_loc(sb, logical_sb_block, i);
3895 sbi->s_group_desc[i] = sb_bread_unmovable(sb, block);
3896 if (!sbi->s_group_desc[i]) {
3897 ext4_msg(sb, KERN_ERR,
3898 "can't read group descriptor %d", i);
3903 if (!ext4_check_descriptors(sb, &first_not_zeroed)) {
3904 ext4_msg(sb, KERN_ERR, "group descriptors corrupted!");
3908 sbi->s_gdb_count = db_count;
3909 get_random_bytes(&sbi->s_next_generation, sizeof(u32));
3910 spin_lock_init(&sbi->s_next_gen_lock);
3912 init_timer(&sbi->s_err_report);
3913 sbi->s_err_report.function = print_daily_error_info;
3914 sbi->s_err_report.data = (unsigned long) sb;
3916 /* Register extent status tree shrinker */
3917 if (ext4_es_register_shrinker(sbi))
3920 sbi->s_stripe = ext4_get_stripe_size(sbi);
3921 sbi->s_extent_max_zeroout_kb = 32;
3924 * set up enough so that it can read an inode
3926 sb->s_op = &ext4_sops;
3927 sb->s_export_op = &ext4_export_ops;
3928 sb->s_xattr = ext4_xattr_handlers;
3930 sb->dq_op = &ext4_quota_operations;
3931 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_QUOTA))
3932 sb->s_qcop = &ext4_qctl_sysfile_operations;
3934 sb->s_qcop = &ext4_qctl_operations;
3936 memcpy(sb->s_uuid, es->s_uuid, sizeof(es->s_uuid));
3938 INIT_LIST_HEAD(&sbi->s_orphan); /* unlinked but open files */
3939 mutex_init(&sbi->s_orphan_lock);
3943 needs_recovery = (es->s_last_orphan != 0 ||
3944 EXT4_HAS_INCOMPAT_FEATURE(sb,
3945 EXT4_FEATURE_INCOMPAT_RECOVER));
3947 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_MMP) &&
3948 !(sb->s_flags & MS_RDONLY))
3949 if (ext4_multi_mount_protect(sb, le64_to_cpu(es->s_mmp_block)))
3950 goto failed_mount3a;
3953 * The first inode we look at is the journal inode. Don't try
3954 * root first: it may be modified in the journal!
3956 if (!test_opt(sb, NOLOAD) &&
3957 EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL)) {
3958 if (ext4_load_journal(sb, es, journal_devnum))
3959 goto failed_mount3a;
3960 } else if (test_opt(sb, NOLOAD) && !(sb->s_flags & MS_RDONLY) &&
3961 EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER)) {
3962 ext4_msg(sb, KERN_ERR, "required journal recovery "
3963 "suppressed and not mounted read-only");
3964 goto failed_mount_wq;
3966 clear_opt(sb, DATA_FLAGS);
3967 sbi->s_journal = NULL;
3972 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_64BIT) &&
3973 !jbd2_journal_set_features(EXT4_SB(sb)->s_journal, 0, 0,
3974 JBD2_FEATURE_INCOMPAT_64BIT)) {
3975 ext4_msg(sb, KERN_ERR, "Failed to set 64-bit journal feature");
3976 goto failed_mount_wq;
3979 if (!set_journal_csum_feature_set(sb)) {
3980 ext4_msg(sb, KERN_ERR, "Failed to set journal checksum "
3982 goto failed_mount_wq;
3985 /* We have now updated the journal if required, so we can
3986 * validate the data journaling mode. */
3987 switch (test_opt(sb, DATA_FLAGS)) {
3989 /* No mode set, assume a default based on the journal
3990 * capabilities: ORDERED_DATA if the journal can
3991 * cope, else JOURNAL_DATA
3993 if (jbd2_journal_check_available_features
3994 (sbi->s_journal, 0, 0, JBD2_FEATURE_INCOMPAT_REVOKE))
3995 set_opt(sb, ORDERED_DATA);
3997 set_opt(sb, JOURNAL_DATA);
4000 case EXT4_MOUNT_ORDERED_DATA:
4001 case EXT4_MOUNT_WRITEBACK_DATA:
4002 if (!jbd2_journal_check_available_features
4003 (sbi->s_journal, 0, 0, JBD2_FEATURE_INCOMPAT_REVOKE)) {
4004 ext4_msg(sb, KERN_ERR, "Journal does not support "
4005 "requested data journaling mode");
4006 goto failed_mount_wq;
4011 set_task_ioprio(sbi->s_journal->j_task, journal_ioprio);
4013 sbi->s_journal->j_commit_callback = ext4_journal_commit_callback;
4016 if (ext4_mballoc_ready) {
4017 sbi->s_mb_cache = ext4_xattr_create_cache(sb->s_id);
4018 if (!sbi->s_mb_cache) {
4019 ext4_msg(sb, KERN_ERR, "Failed to create an mb_cache");
4020 goto failed_mount_wq;
4025 * Get the # of file system overhead blocks from the
4026 * superblock if present.
4028 if (es->s_overhead_clusters)
4029 sbi->s_overhead = le32_to_cpu(es->s_overhead_clusters);
4031 err = ext4_calculate_overhead(sb);
4033 goto failed_mount_wq;
4037 * The maximum number of concurrent works can be high and
4038 * concurrency isn't really necessary. Limit it to 1.
4040 EXT4_SB(sb)->rsv_conversion_wq =
4041 alloc_workqueue("ext4-rsv-conversion", WQ_MEM_RECLAIM | WQ_UNBOUND, 1);
4042 if (!EXT4_SB(sb)->rsv_conversion_wq) {
4043 printk(KERN_ERR "EXT4-fs: failed to create workqueue\n");
4049 * The jbd2_journal_load will have done any necessary log recovery,
4050 * so we can safely mount the rest of the filesystem now.
4053 root = ext4_iget(sb, EXT4_ROOT_INO);
4055 ext4_msg(sb, KERN_ERR, "get root inode failed");
4056 ret = PTR_ERR(root);
4060 if (!S_ISDIR(root->i_mode) || !root->i_blocks || !root->i_size) {
4061 ext4_msg(sb, KERN_ERR, "corrupt root inode, run e2fsck");
4065 sb->s_root = d_make_root(root);
4067 ext4_msg(sb, KERN_ERR, "get root dentry failed");
4072 if (ext4_setup_super(sb, es, sb->s_flags & MS_RDONLY))
4073 sb->s_flags |= MS_RDONLY;
4075 /* determine the minimum size of new large inodes, if present */
4076 if (sbi->s_inode_size > EXT4_GOOD_OLD_INODE_SIZE) {
4077 sbi->s_want_extra_isize = sizeof(struct ext4_inode) -
4078 EXT4_GOOD_OLD_INODE_SIZE;
4079 if (EXT4_HAS_RO_COMPAT_FEATURE(sb,
4080 EXT4_FEATURE_RO_COMPAT_EXTRA_ISIZE)) {
4081 if (sbi->s_want_extra_isize <
4082 le16_to_cpu(es->s_want_extra_isize))
4083 sbi->s_want_extra_isize =
4084 le16_to_cpu(es->s_want_extra_isize);
4085 if (sbi->s_want_extra_isize <
4086 le16_to_cpu(es->s_min_extra_isize))
4087 sbi->s_want_extra_isize =
4088 le16_to_cpu(es->s_min_extra_isize);
4091 /* Check if enough inode space is available */
4092 if (EXT4_GOOD_OLD_INODE_SIZE + sbi->s_want_extra_isize >
4093 sbi->s_inode_size) {
4094 sbi->s_want_extra_isize = sizeof(struct ext4_inode) -
4095 EXT4_GOOD_OLD_INODE_SIZE;
4096 ext4_msg(sb, KERN_INFO, "required extra inode space not"
4100 err = ext4_reserve_clusters(sbi, ext4_calculate_resv_clusters(sb));
4102 ext4_msg(sb, KERN_ERR, "failed to reserve %llu clusters for "
4103 "reserved pool", ext4_calculate_resv_clusters(sb));
4104 goto failed_mount4a;
4107 err = ext4_setup_system_zone(sb);
4109 ext4_msg(sb, KERN_ERR, "failed to initialize system "
4111 goto failed_mount4a;
4115 err = ext4_mb_init(sb);
4117 ext4_msg(sb, KERN_ERR, "failed to initialize mballoc (%d)",
4122 block = ext4_count_free_clusters(sb);
4123 ext4_free_blocks_count_set(sbi->s_es,
4124 EXT4_C2B(sbi, block));
4125 err = percpu_counter_init(&sbi->s_freeclusters_counter, block,
4128 unsigned long freei = ext4_count_free_inodes(sb);
4129 sbi->s_es->s_free_inodes_count = cpu_to_le32(freei);
4130 err = percpu_counter_init(&sbi->s_freeinodes_counter, freei,
4134 err = percpu_counter_init(&sbi->s_dirs_counter,
4135 ext4_count_dirs(sb), GFP_KERNEL);
4137 err = percpu_counter_init(&sbi->s_dirtyclusters_counter, 0,
4140 ext4_msg(sb, KERN_ERR, "insufficient memory");
4144 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_FLEX_BG))
4145 if (!ext4_fill_flex_info(sb)) {
4146 ext4_msg(sb, KERN_ERR,
4147 "unable to initialize "
4148 "flex_bg meta info!");
4152 err = ext4_register_li_request(sb, first_not_zeroed);
4156 sbi->s_kobj.kset = ext4_kset;
4157 init_completion(&sbi->s_kobj_unregister);
4158 err = kobject_init_and_add(&sbi->s_kobj, &ext4_ktype, NULL,
4164 /* Enable quota usage during mount. */
4165 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_QUOTA) &&
4166 !(sb->s_flags & MS_RDONLY)) {
4167 err = ext4_enable_quotas(sb);
4171 #endif /* CONFIG_QUOTA */
4173 EXT4_SB(sb)->s_mount_state |= EXT4_ORPHAN_FS;
4174 ext4_orphan_cleanup(sb, es);
4175 EXT4_SB(sb)->s_mount_state &= ~EXT4_ORPHAN_FS;
4176 if (needs_recovery) {
4177 ext4_msg(sb, KERN_INFO, "recovery complete");
4178 ext4_mark_recovery_complete(sb, es);
4180 if (EXT4_SB(sb)->s_journal) {
4181 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA)
4182 descr = " journalled data mode";
4183 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA)
4184 descr = " ordered data mode";
4186 descr = " writeback data mode";
4188 descr = "out journal";
4190 if (test_opt(sb, DISCARD)) {
4191 struct request_queue *q = bdev_get_queue(sb->s_bdev);
4192 if (!blk_queue_discard(q))
4193 ext4_msg(sb, KERN_WARNING,
4194 "mounting with \"discard\" option, but "
4195 "the device does not support discard");
4198 ext4_msg(sb, KERN_INFO, "mounted filesystem with%s. "
4199 "Opts: %s%s%s", descr, sbi->s_es->s_mount_opts,
4200 *sbi->s_es->s_mount_opts ? "; " : "", orig_data);
4202 if (es->s_error_count)
4203 mod_timer(&sbi->s_err_report, jiffies + 300*HZ); /* 5 minutes */
4205 /* Enable message ratelimiting. Default is 10 messages per 5 secs. */
4206 ratelimit_state_init(&sbi->s_err_ratelimit_state, 5 * HZ, 10);
4207 ratelimit_state_init(&sbi->s_warning_ratelimit_state, 5 * HZ, 10);
4208 ratelimit_state_init(&sbi->s_msg_ratelimit_state, 5 * HZ, 10);
4215 ext4_msg(sb, KERN_ERR, "VFS: Can't find ext4 filesystem");
4220 kobject_del(&sbi->s_kobj);
4223 ext4_unregister_li_request(sb);
4225 ext4_mb_release(sb);
4226 if (sbi->s_flex_groups)
4227 ext4_kvfree(sbi->s_flex_groups);
4228 percpu_counter_destroy(&sbi->s_freeclusters_counter);
4229 percpu_counter_destroy(&sbi->s_freeinodes_counter);
4230 percpu_counter_destroy(&sbi->s_dirs_counter);
4231 percpu_counter_destroy(&sbi->s_dirtyclusters_counter);
4233 ext4_ext_release(sb);
4234 ext4_release_system_zone(sb);
4239 ext4_msg(sb, KERN_ERR, "mount failed");
4240 if (EXT4_SB(sb)->rsv_conversion_wq)
4241 destroy_workqueue(EXT4_SB(sb)->rsv_conversion_wq);
4243 if (sbi->s_journal) {
4244 jbd2_journal_destroy(sbi->s_journal);
4245 sbi->s_journal = NULL;
4248 ext4_es_unregister_shrinker(sbi);
4250 del_timer_sync(&sbi->s_err_report);
4252 kthread_stop(sbi->s_mmp_tsk);
4254 for (i = 0; i < db_count; i++)
4255 brelse(sbi->s_group_desc[i]);
4256 ext4_kvfree(sbi->s_group_desc);
4258 if (sbi->s_chksum_driver)
4259 crypto_free_shash(sbi->s_chksum_driver);
4261 remove_proc_entry("options", sbi->s_proc);
4262 remove_proc_entry(sb->s_id, ext4_proc_root);
4265 for (i = 0; i < EXT4_MAXQUOTAS; i++)
4266 kfree(sbi->s_qf_names[i]);
4268 ext4_blkdev_remove(sbi);
4271 sb->s_fs_info = NULL;
4272 kfree(sbi->s_blockgroup_lock);
4276 return err ? err : ret;
4280 * Setup any per-fs journal parameters now. We'll do this both on
4281 * initial mount, once the journal has been initialised but before we've
4282 * done any recovery; and again on any subsequent remount.
4284 static void ext4_init_journal_params(struct super_block *sb, journal_t *journal)
4286 struct ext4_sb_info *sbi = EXT4_SB(sb);
4288 journal->j_commit_interval = sbi->s_commit_interval;
4289 journal->j_min_batch_time = sbi->s_min_batch_time;
4290 journal->j_max_batch_time = sbi->s_max_batch_time;
4292 write_lock(&journal->j_state_lock);
4293 if (test_opt(sb, BARRIER))
4294 journal->j_flags |= JBD2_BARRIER;
4296 journal->j_flags &= ~JBD2_BARRIER;
4297 if (test_opt(sb, DATA_ERR_ABORT))
4298 journal->j_flags |= JBD2_ABORT_ON_SYNCDATA_ERR;
4300 journal->j_flags &= ~JBD2_ABORT_ON_SYNCDATA_ERR;
4301 write_unlock(&journal->j_state_lock);
4304 static journal_t *ext4_get_journal(struct super_block *sb,
4305 unsigned int journal_inum)
4307 struct inode *journal_inode;
4310 BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL));
4312 /* First, test for the existence of a valid inode on disk. Bad
4313 * things happen if we iget() an unused inode, as the subsequent
4314 * iput() will try to delete it. */
4316 journal_inode = ext4_iget(sb, journal_inum);
4317 if (IS_ERR(journal_inode)) {
4318 ext4_msg(sb, KERN_ERR, "no journal found");
4321 if (!journal_inode->i_nlink) {
4322 make_bad_inode(journal_inode);
4323 iput(journal_inode);
4324 ext4_msg(sb, KERN_ERR, "journal inode is deleted");
4328 jbd_debug(2, "Journal inode found at %p: %lld bytes\n",
4329 journal_inode, journal_inode->i_size);
4330 if (!S_ISREG(journal_inode->i_mode)) {
4331 ext4_msg(sb, KERN_ERR, "invalid journal inode");
4332 iput(journal_inode);
4336 journal = jbd2_journal_init_inode(journal_inode);
4338 ext4_msg(sb, KERN_ERR, "Could not load journal inode");
4339 iput(journal_inode);
4342 journal->j_private = sb;
4343 ext4_init_journal_params(sb, journal);
4347 static journal_t *ext4_get_dev_journal(struct super_block *sb,
4350 struct buffer_head *bh;
4354 int hblock, blocksize;
4355 ext4_fsblk_t sb_block;
4356 unsigned long offset;
4357 struct ext4_super_block *es;
4358 struct block_device *bdev;
4360 BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL));
4362 bdev = ext4_blkdev_get(j_dev, sb);
4366 blocksize = sb->s_blocksize;
4367 hblock = bdev_logical_block_size(bdev);
4368 if (blocksize < hblock) {
4369 ext4_msg(sb, KERN_ERR,
4370 "blocksize too small for journal device");
4374 sb_block = EXT4_MIN_BLOCK_SIZE / blocksize;
4375 offset = EXT4_MIN_BLOCK_SIZE % blocksize;
4376 set_blocksize(bdev, blocksize);
4377 if (!(bh = __bread(bdev, sb_block, blocksize))) {
4378 ext4_msg(sb, KERN_ERR, "couldn't read superblock of "
4379 "external journal");
4383 es = (struct ext4_super_block *) (bh->b_data + offset);
4384 if ((le16_to_cpu(es->s_magic) != EXT4_SUPER_MAGIC) ||
4385 !(le32_to_cpu(es->s_feature_incompat) &
4386 EXT4_FEATURE_INCOMPAT_JOURNAL_DEV)) {
4387 ext4_msg(sb, KERN_ERR, "external journal has "
4393 if ((le32_to_cpu(es->s_feature_ro_compat) &
4394 EXT4_FEATURE_RO_COMPAT_METADATA_CSUM) &&
4395 es->s_checksum != ext4_superblock_csum(sb, es)) {
4396 ext4_msg(sb, KERN_ERR, "external journal has "
4397 "corrupt superblock");
4402 if (memcmp(EXT4_SB(sb)->s_es->s_journal_uuid, es->s_uuid, 16)) {
4403 ext4_msg(sb, KERN_ERR, "journal UUID does not match");
4408 len = ext4_blocks_count(es);
4409 start = sb_block + 1;
4410 brelse(bh); /* we're done with the superblock */
4412 journal = jbd2_journal_init_dev(bdev, sb->s_bdev,
4413 start, len, blocksize);
4415 ext4_msg(sb, KERN_ERR, "failed to create device journal");
4418 journal->j_private = sb;
4419 ll_rw_block(READ | REQ_META | REQ_PRIO, 1, &journal->j_sb_buffer);
4420 wait_on_buffer(journal->j_sb_buffer);
4421 if (!buffer_uptodate(journal->j_sb_buffer)) {
4422 ext4_msg(sb, KERN_ERR, "I/O error on journal device");
4425 if (be32_to_cpu(journal->j_superblock->s_nr_users) != 1) {
4426 ext4_msg(sb, KERN_ERR, "External journal has more than one "
4427 "user (unsupported) - %d",
4428 be32_to_cpu(journal->j_superblock->s_nr_users));
4431 EXT4_SB(sb)->journal_bdev = bdev;
4432 ext4_init_journal_params(sb, journal);
4436 jbd2_journal_destroy(journal);
4438 ext4_blkdev_put(bdev);
4442 static int ext4_load_journal(struct super_block *sb,
4443 struct ext4_super_block *es,
4444 unsigned long journal_devnum)
4447 unsigned int journal_inum = le32_to_cpu(es->s_journal_inum);
4450 int really_read_only;
4452 BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL));
4454 if (journal_devnum &&
4455 journal_devnum != le32_to_cpu(es->s_journal_dev)) {
4456 ext4_msg(sb, KERN_INFO, "external journal device major/minor "
4457 "numbers have changed");
4458 journal_dev = new_decode_dev(journal_devnum);
4460 journal_dev = new_decode_dev(le32_to_cpu(es->s_journal_dev));
4462 really_read_only = bdev_read_only(sb->s_bdev);
4465 * Are we loading a blank journal or performing recovery after a
4466 * crash? For recovery, we need to check in advance whether we
4467 * can get read-write access to the device.
4469 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER)) {
4470 if (sb->s_flags & MS_RDONLY) {
4471 ext4_msg(sb, KERN_INFO, "INFO: recovery "
4472 "required on readonly filesystem");
4473 if (really_read_only) {
4474 ext4_msg(sb, KERN_ERR, "write access "
4475 "unavailable, cannot proceed");
4478 ext4_msg(sb, KERN_INFO, "write access will "
4479 "be enabled during recovery");
4483 if (journal_inum && journal_dev) {
4484 ext4_msg(sb, KERN_ERR, "filesystem has both journal "
4485 "and inode journals!");
4490 if (!(journal = ext4_get_journal(sb, journal_inum)))
4493 if (!(journal = ext4_get_dev_journal(sb, journal_dev)))
4497 if (!(journal->j_flags & JBD2_BARRIER))
4498 ext4_msg(sb, KERN_INFO, "barriers disabled");
4500 if (!EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER))
4501 err = jbd2_journal_wipe(journal, !really_read_only);
4503 char *save = kmalloc(EXT4_S_ERR_LEN, GFP_KERNEL);
4505 memcpy(save, ((char *) es) +
4506 EXT4_S_ERR_START, EXT4_S_ERR_LEN);
4507 err = jbd2_journal_load(journal);
4509 memcpy(((char *) es) + EXT4_S_ERR_START,
4510 save, EXT4_S_ERR_LEN);
4515 ext4_msg(sb, KERN_ERR, "error loading journal");
4516 jbd2_journal_destroy(journal);
4520 EXT4_SB(sb)->s_journal = journal;
4521 ext4_clear_journal_err(sb, es);
4523 if (!really_read_only && journal_devnum &&
4524 journal_devnum != le32_to_cpu(es->s_journal_dev)) {
4525 es->s_journal_dev = cpu_to_le32(journal_devnum);
4527 /* Make sure we flush the recovery flag to disk. */
4528 ext4_commit_super(sb, 1);
4534 static int ext4_commit_super(struct super_block *sb, int sync)
4536 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
4537 struct buffer_head *sbh = EXT4_SB(sb)->s_sbh;
4540 if (!sbh || block_device_ejected(sb))
4542 if (buffer_write_io_error(sbh)) {
4544 * Oh, dear. A previous attempt to write the
4545 * superblock failed. This could happen because the
4546 * USB device was yanked out. Or it could happen to
4547 * be a transient write error and maybe the block will
4548 * be remapped. Nothing we can do but to retry the
4549 * write and hope for the best.
4551 ext4_msg(sb, KERN_ERR, "previous I/O error to "
4552 "superblock detected");
4553 clear_buffer_write_io_error(sbh);
4554 set_buffer_uptodate(sbh);
4557 * If the file system is mounted read-only, don't update the
4558 * superblock write time. This avoids updating the superblock
4559 * write time when we are mounting the root file system
4560 * read/only but we need to replay the journal; at that point,
4561 * for people who are east of GMT and who make their clock
4562 * tick in localtime for Windows bug-for-bug compatibility,
4563 * the clock is set in the future, and this will cause e2fsck
4564 * to complain and force a full file system check.
4566 if (!(sb->s_flags & MS_RDONLY))
4567 es->s_wtime = cpu_to_le32(get_seconds());
4568 if (sb->s_bdev->bd_part)
4569 es->s_kbytes_written =
4570 cpu_to_le64(EXT4_SB(sb)->s_kbytes_written +
4571 ((part_stat_read(sb->s_bdev->bd_part, sectors[1]) -
4572 EXT4_SB(sb)->s_sectors_written_start) >> 1));
4574 es->s_kbytes_written =
4575 cpu_to_le64(EXT4_SB(sb)->s_kbytes_written);
4576 if (percpu_counter_initialized(&EXT4_SB(sb)->s_freeclusters_counter))
4577 ext4_free_blocks_count_set(es,
4578 EXT4_C2B(EXT4_SB(sb), percpu_counter_sum_positive(
4579 &EXT4_SB(sb)->s_freeclusters_counter)));
4580 if (percpu_counter_initialized(&EXT4_SB(sb)->s_freeinodes_counter))
4581 es->s_free_inodes_count =
4582 cpu_to_le32(percpu_counter_sum_positive(
4583 &EXT4_SB(sb)->s_freeinodes_counter));
4584 BUFFER_TRACE(sbh, "marking dirty");
4585 ext4_superblock_csum_set(sb);
4586 mark_buffer_dirty(sbh);
4588 error = sync_dirty_buffer(sbh);
4592 error = buffer_write_io_error(sbh);
4594 ext4_msg(sb, KERN_ERR, "I/O error while writing "
4596 clear_buffer_write_io_error(sbh);
4597 set_buffer_uptodate(sbh);
4604 * Have we just finished recovery? If so, and if we are mounting (or
4605 * remounting) the filesystem readonly, then we will end up with a
4606 * consistent fs on disk. Record that fact.
4608 static void ext4_mark_recovery_complete(struct super_block *sb,
4609 struct ext4_super_block *es)
4611 journal_t *journal = EXT4_SB(sb)->s_journal;
4613 if (!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL)) {
4614 BUG_ON(journal != NULL);
4617 jbd2_journal_lock_updates(journal);
4618 if (jbd2_journal_flush(journal) < 0)
4621 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER) &&
4622 sb->s_flags & MS_RDONLY) {
4623 EXT4_CLEAR_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
4624 ext4_commit_super(sb, 1);
4628 jbd2_journal_unlock_updates(journal);
4632 * If we are mounting (or read-write remounting) a filesystem whose journal
4633 * has recorded an error from a previous lifetime, move that error to the
4634 * main filesystem now.
4636 static void ext4_clear_journal_err(struct super_block *sb,
4637 struct ext4_super_block *es)
4643 BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL));
4645 journal = EXT4_SB(sb)->s_journal;
4648 * Now check for any error status which may have been recorded in the
4649 * journal by a prior ext4_error() or ext4_abort()
4652 j_errno = jbd2_journal_errno(journal);
4656 errstr = ext4_decode_error(sb, j_errno, nbuf);
4657 ext4_warning(sb, "Filesystem error recorded "
4658 "from previous mount: %s", errstr);
4659 ext4_warning(sb, "Marking fs in need of filesystem check.");
4661 EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS;
4662 es->s_state |= cpu_to_le16(EXT4_ERROR_FS);
4663 ext4_commit_super(sb, 1);
4665 jbd2_journal_clear_err(journal);
4666 jbd2_journal_update_sb_errno(journal);
4671 * Force the running and committing transactions to commit,
4672 * and wait on the commit.
4674 int ext4_force_commit(struct super_block *sb)
4678 if (sb->s_flags & MS_RDONLY)
4681 journal = EXT4_SB(sb)->s_journal;
4682 return ext4_journal_force_commit(journal);
4685 static int ext4_sync_fs(struct super_block *sb, int wait)
4689 bool needs_barrier = false;
4690 struct ext4_sb_info *sbi = EXT4_SB(sb);
4692 trace_ext4_sync_fs(sb, wait);
4693 flush_workqueue(sbi->rsv_conversion_wq);
4695 * Writeback quota in non-journalled quota case - journalled quota has
4698 dquot_writeback_dquots(sb, -1);
4700 * Data writeback is possible w/o journal transaction, so barrier must
4701 * being sent at the end of the function. But we can skip it if
4702 * transaction_commit will do it for us.
4704 if (sbi->s_journal) {
4705 target = jbd2_get_latest_transaction(sbi->s_journal);
4706 if (wait && sbi->s_journal->j_flags & JBD2_BARRIER &&
4707 !jbd2_trans_will_send_data_barrier(sbi->s_journal, target))
4708 needs_barrier = true;
4710 if (jbd2_journal_start_commit(sbi->s_journal, &target)) {
4712 ret = jbd2_log_wait_commit(sbi->s_journal,
4715 } else if (wait && test_opt(sb, BARRIER))
4716 needs_barrier = true;
4717 if (needs_barrier) {
4719 err = blkdev_issue_flush(sb->s_bdev, GFP_KERNEL, NULL);
4728 * LVM calls this function before a (read-only) snapshot is created. This
4729 * gives us a chance to flush the journal completely and mark the fs clean.
4731 * Note that only this function cannot bring a filesystem to be in a clean
4732 * state independently. It relies on upper layer to stop all data & metadata
4735 static int ext4_freeze(struct super_block *sb)
4740 if (sb->s_flags & MS_RDONLY)
4743 journal = EXT4_SB(sb)->s_journal;
4746 /* Now we set up the journal barrier. */
4747 jbd2_journal_lock_updates(journal);
4750 * Don't clear the needs_recovery flag if we failed to
4751 * flush the journal.
4753 error = jbd2_journal_flush(journal);
4758 /* Journal blocked and flushed, clear needs_recovery flag. */
4759 EXT4_CLEAR_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
4760 error = ext4_commit_super(sb, 1);
4763 /* we rely on upper layer to stop further updates */
4764 jbd2_journal_unlock_updates(journal);
4769 * Called by LVM after the snapshot is done. We need to reset the RECOVER
4770 * flag here, even though the filesystem is not technically dirty yet.
4772 static int ext4_unfreeze(struct super_block *sb)
4774 if (sb->s_flags & MS_RDONLY)
4777 /* Reset the needs_recovery flag before the fs is unlocked. */
4778 EXT4_SET_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
4779 ext4_commit_super(sb, 1);
4784 * Structure to save mount options for ext4_remount's benefit
4786 struct ext4_mount_options {
4787 unsigned long s_mount_opt;
4788 unsigned long s_mount_opt2;
4791 unsigned long s_commit_interval;
4792 u32 s_min_batch_time, s_max_batch_time;
4795 char *s_qf_names[EXT4_MAXQUOTAS];
4799 static int ext4_remount(struct super_block *sb, int *flags, char *data)
4801 struct ext4_super_block *es;
4802 struct ext4_sb_info *sbi = EXT4_SB(sb);
4803 unsigned long old_sb_flags;
4804 struct ext4_mount_options old_opts;
4805 int enable_quota = 0;
4807 unsigned int journal_ioprio = DEFAULT_JOURNAL_IOPRIO;
4812 char *orig_data = kstrdup(data, GFP_KERNEL);
4814 /* Store the original options */
4815 old_sb_flags = sb->s_flags;
4816 old_opts.s_mount_opt = sbi->s_mount_opt;
4817 old_opts.s_mount_opt2 = sbi->s_mount_opt2;
4818 old_opts.s_resuid = sbi->s_resuid;
4819 old_opts.s_resgid = sbi->s_resgid;
4820 old_opts.s_commit_interval = sbi->s_commit_interval;
4821 old_opts.s_min_batch_time = sbi->s_min_batch_time;
4822 old_opts.s_max_batch_time = sbi->s_max_batch_time;
4824 old_opts.s_jquota_fmt = sbi->s_jquota_fmt;
4825 for (i = 0; i < EXT4_MAXQUOTAS; i++)
4826 if (sbi->s_qf_names[i]) {
4827 old_opts.s_qf_names[i] = kstrdup(sbi->s_qf_names[i],
4829 if (!old_opts.s_qf_names[i]) {
4830 for (j = 0; j < i; j++)
4831 kfree(old_opts.s_qf_names[j]);
4836 old_opts.s_qf_names[i] = NULL;
4838 if (sbi->s_journal && sbi->s_journal->j_task->io_context)
4839 journal_ioprio = sbi->s_journal->j_task->io_context->ioprio;
4842 * Allow the "check" option to be passed as a remount option.
4844 if (!parse_options(data, sb, NULL, &journal_ioprio, 1)) {
4849 if ((old_opts.s_mount_opt & EXT4_MOUNT_JOURNAL_CHECKSUM) ^
4850 test_opt(sb, JOURNAL_CHECKSUM)) {
4851 ext4_msg(sb, KERN_ERR, "changing journal_checksum "
4852 "during remount not supported");
4857 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA) {
4858 if (test_opt2(sb, EXPLICIT_DELALLOC)) {
4859 ext4_msg(sb, KERN_ERR, "can't mount with "
4860 "both data=journal and delalloc");
4864 if (test_opt(sb, DIOREAD_NOLOCK)) {
4865 ext4_msg(sb, KERN_ERR, "can't mount with "
4866 "both data=journal and dioread_nolock");
4872 if (sbi->s_mount_flags & EXT4_MF_FS_ABORTED)
4873 ext4_abort(sb, "Abort forced by user");
4875 sb->s_flags = (sb->s_flags & ~MS_POSIXACL) |
4876 (test_opt(sb, POSIX_ACL) ? MS_POSIXACL : 0);
4880 if (sbi->s_journal) {
4881 ext4_init_journal_params(sb, sbi->s_journal);
4882 set_task_ioprio(sbi->s_journal->j_task, journal_ioprio);
4885 if ((*flags & MS_RDONLY) != (sb->s_flags & MS_RDONLY)) {
4886 if (sbi->s_mount_flags & EXT4_MF_FS_ABORTED) {
4891 if (*flags & MS_RDONLY) {
4892 err = sync_filesystem(sb);
4895 err = dquot_suspend(sb, -1);
4900 * First of all, the unconditional stuff we have to do
4901 * to disable replay of the journal when we next remount
4903 sb->s_flags |= MS_RDONLY;
4906 * OK, test if we are remounting a valid rw partition
4907 * readonly, and if so set the rdonly flag and then
4908 * mark the partition as valid again.
4910 if (!(es->s_state & cpu_to_le16(EXT4_VALID_FS)) &&
4911 (sbi->s_mount_state & EXT4_VALID_FS))
4912 es->s_state = cpu_to_le16(sbi->s_mount_state);
4915 ext4_mark_recovery_complete(sb, es);
4917 /* Make sure we can mount this feature set readwrite */
4918 if (!ext4_feature_set_ok(sb, 0)) {
4923 * Make sure the group descriptor checksums
4924 * are sane. If they aren't, refuse to remount r/w.
4926 for (g = 0; g < sbi->s_groups_count; g++) {
4927 struct ext4_group_desc *gdp =
4928 ext4_get_group_desc(sb, g, NULL);
4930 if (!ext4_group_desc_csum_verify(sb, g, gdp)) {
4931 ext4_msg(sb, KERN_ERR,
4932 "ext4_remount: Checksum for group %u failed (%u!=%u)",
4933 g, le16_to_cpu(ext4_group_desc_csum(sbi, g, gdp)),
4934 le16_to_cpu(gdp->bg_checksum));
4941 * If we have an unprocessed orphan list hanging
4942 * around from a previously readonly bdev mount,
4943 * require a full umount/remount for now.
4945 if (es->s_last_orphan) {
4946 ext4_msg(sb, KERN_WARNING, "Couldn't "
4947 "remount RDWR because of unprocessed "
4948 "orphan inode list. Please "
4949 "umount/remount instead");
4955 * Mounting a RDONLY partition read-write, so reread
4956 * and store the current valid flag. (It may have
4957 * been changed by e2fsck since we originally mounted
4961 ext4_clear_journal_err(sb, es);
4962 sbi->s_mount_state = le16_to_cpu(es->s_state);
4963 if (!ext4_setup_super(sb, es, 0))
4964 sb->s_flags &= ~MS_RDONLY;
4965 if (EXT4_HAS_INCOMPAT_FEATURE(sb,
4966 EXT4_FEATURE_INCOMPAT_MMP))
4967 if (ext4_multi_mount_protect(sb,
4968 le64_to_cpu(es->s_mmp_block))) {
4977 * Reinitialize lazy itable initialization thread based on
4980 if ((sb->s_flags & MS_RDONLY) || !test_opt(sb, INIT_INODE_TABLE))
4981 ext4_unregister_li_request(sb);
4983 ext4_group_t first_not_zeroed;
4984 first_not_zeroed = ext4_has_uninit_itable(sb);
4985 ext4_register_li_request(sb, first_not_zeroed);
4988 ext4_setup_system_zone(sb);
4989 if (sbi->s_journal == NULL && !(old_sb_flags & MS_RDONLY))
4990 ext4_commit_super(sb, 1);
4993 /* Release old quota file names */
4994 for (i = 0; i < EXT4_MAXQUOTAS; i++)
4995 kfree(old_opts.s_qf_names[i]);
4997 if (sb_any_quota_suspended(sb))
4998 dquot_resume(sb, -1);
4999 else if (EXT4_HAS_RO_COMPAT_FEATURE(sb,
5000 EXT4_FEATURE_RO_COMPAT_QUOTA)) {
5001 err = ext4_enable_quotas(sb);
5008 ext4_msg(sb, KERN_INFO, "re-mounted. Opts: %s", orig_data);
5013 sb->s_flags = old_sb_flags;
5014 sbi->s_mount_opt = old_opts.s_mount_opt;
5015 sbi->s_mount_opt2 = old_opts.s_mount_opt2;
5016 sbi->s_resuid = old_opts.s_resuid;
5017 sbi->s_resgid = old_opts.s_resgid;
5018 sbi->s_commit_interval = old_opts.s_commit_interval;
5019 sbi->s_min_batch_time = old_opts.s_min_batch_time;
5020 sbi->s_max_batch_time = old_opts.s_max_batch_time;
5022 sbi->s_jquota_fmt = old_opts.s_jquota_fmt;
5023 for (i = 0; i < EXT4_MAXQUOTAS; i++) {
5024 kfree(sbi->s_qf_names[i]);
5025 sbi->s_qf_names[i] = old_opts.s_qf_names[i];
5032 static int ext4_statfs(struct dentry *dentry, struct kstatfs *buf)
5034 struct super_block *sb = dentry->d_sb;
5035 struct ext4_sb_info *sbi = EXT4_SB(sb);
5036 struct ext4_super_block *es = sbi->s_es;
5037 ext4_fsblk_t overhead = 0, resv_blocks;
5040 resv_blocks = EXT4_C2B(sbi, atomic64_read(&sbi->s_resv_clusters));
5042 if (!test_opt(sb, MINIX_DF))
5043 overhead = sbi->s_overhead;
5045 buf->f_type = EXT4_SUPER_MAGIC;
5046 buf->f_bsize = sb->s_blocksize;
5047 buf->f_blocks = ext4_blocks_count(es) - EXT4_C2B(sbi, overhead);
5048 bfree = percpu_counter_sum_positive(&sbi->s_freeclusters_counter) -
5049 percpu_counter_sum_positive(&sbi->s_dirtyclusters_counter);
5050 /* prevent underflow in case that few free space is available */
5051 buf->f_bfree = EXT4_C2B(sbi, max_t(s64, bfree, 0));
5052 buf->f_bavail = buf->f_bfree -
5053 (ext4_r_blocks_count(es) + resv_blocks);
5054 if (buf->f_bfree < (ext4_r_blocks_count(es) + resv_blocks))
5056 buf->f_files = le32_to_cpu(es->s_inodes_count);
5057 buf->f_ffree = percpu_counter_sum_positive(&sbi->s_freeinodes_counter);
5058 buf->f_namelen = EXT4_NAME_LEN;
5059 fsid = le64_to_cpup((void *)es->s_uuid) ^
5060 le64_to_cpup((void *)es->s_uuid + sizeof(u64));
5061 buf->f_fsid.val[0] = fsid & 0xFFFFFFFFUL;
5062 buf->f_fsid.val[1] = (fsid >> 32) & 0xFFFFFFFFUL;
5067 /* Helper function for writing quotas on sync - we need to start transaction
5068 * before quota file is locked for write. Otherwise the are possible deadlocks:
5069 * Process 1 Process 2
5070 * ext4_create() quota_sync()
5071 * jbd2_journal_start() write_dquot()
5072 * dquot_initialize() down(dqio_mutex)
5073 * down(dqio_mutex) jbd2_journal_start()
5079 static inline struct inode *dquot_to_inode(struct dquot *dquot)
5081 return sb_dqopt(dquot->dq_sb)->files[dquot->dq_id.type];
5084 static int ext4_write_dquot(struct dquot *dquot)
5088 struct inode *inode;
5090 inode = dquot_to_inode(dquot);
5091 handle = ext4_journal_start(inode, EXT4_HT_QUOTA,
5092 EXT4_QUOTA_TRANS_BLOCKS(dquot->dq_sb));
5094 return PTR_ERR(handle);
5095 ret = dquot_commit(dquot);
5096 err = ext4_journal_stop(handle);
5102 static int ext4_acquire_dquot(struct dquot *dquot)
5107 handle = ext4_journal_start(dquot_to_inode(dquot), EXT4_HT_QUOTA,
5108 EXT4_QUOTA_INIT_BLOCKS(dquot->dq_sb));
5110 return PTR_ERR(handle);
5111 ret = dquot_acquire(dquot);
5112 err = ext4_journal_stop(handle);
5118 static int ext4_release_dquot(struct dquot *dquot)
5123 handle = ext4_journal_start(dquot_to_inode(dquot), EXT4_HT_QUOTA,
5124 EXT4_QUOTA_DEL_BLOCKS(dquot->dq_sb));
5125 if (IS_ERR(handle)) {
5126 /* Release dquot anyway to avoid endless cycle in dqput() */
5127 dquot_release(dquot);
5128 return PTR_ERR(handle);
5130 ret = dquot_release(dquot);
5131 err = ext4_journal_stop(handle);
5137 static int ext4_mark_dquot_dirty(struct dquot *dquot)
5139 struct super_block *sb = dquot->dq_sb;
5140 struct ext4_sb_info *sbi = EXT4_SB(sb);
5142 /* Are we journaling quotas? */
5143 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_QUOTA) ||
5144 sbi->s_qf_names[USRQUOTA] || sbi->s_qf_names[GRPQUOTA]) {
5145 dquot_mark_dquot_dirty(dquot);
5146 return ext4_write_dquot(dquot);
5148 return dquot_mark_dquot_dirty(dquot);
5152 static int ext4_write_info(struct super_block *sb, int type)
5157 /* Data block + inode block */
5158 handle = ext4_journal_start(sb->s_root->d_inode, EXT4_HT_QUOTA, 2);
5160 return PTR_ERR(handle);
5161 ret = dquot_commit_info(sb, type);
5162 err = ext4_journal_stop(handle);
5169 * Turn on quotas during mount time - we need to find
5170 * the quota file and such...
5172 static int ext4_quota_on_mount(struct super_block *sb, int type)
5174 return dquot_quota_on_mount(sb, EXT4_SB(sb)->s_qf_names[type],
5175 EXT4_SB(sb)->s_jquota_fmt, type);
5179 * Standard function to be called on quota_on
5181 static int ext4_quota_on(struct super_block *sb, int type, int format_id,
5186 if (!test_opt(sb, QUOTA))
5189 /* Quotafile not on the same filesystem? */
5190 if (path->dentry->d_sb != sb)
5192 /* Journaling quota? */
5193 if (EXT4_SB(sb)->s_qf_names[type]) {
5194 /* Quotafile not in fs root? */
5195 if (path->dentry->d_parent != sb->s_root)
5196 ext4_msg(sb, KERN_WARNING,
5197 "Quota file not on filesystem root. "
5198 "Journaled quota will not work");
5202 * When we journal data on quota file, we have to flush journal to see
5203 * all updates to the file when we bypass pagecache...
5205 if (EXT4_SB(sb)->s_journal &&
5206 ext4_should_journal_data(path->dentry->d_inode)) {
5208 * We don't need to lock updates but journal_flush() could
5209 * otherwise be livelocked...
5211 jbd2_journal_lock_updates(EXT4_SB(sb)->s_journal);
5212 err = jbd2_journal_flush(EXT4_SB(sb)->s_journal);
5213 jbd2_journal_unlock_updates(EXT4_SB(sb)->s_journal);
5218 return dquot_quota_on(sb, type, format_id, path);
5221 static int ext4_quota_enable(struct super_block *sb, int type, int format_id,
5225 struct inode *qf_inode;
5226 unsigned long qf_inums[EXT4_MAXQUOTAS] = {
5227 le32_to_cpu(EXT4_SB(sb)->s_es->s_usr_quota_inum),
5228 le32_to_cpu(EXT4_SB(sb)->s_es->s_grp_quota_inum)
5231 BUG_ON(!EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_QUOTA));
5233 if (!qf_inums[type])
5236 qf_inode = ext4_iget(sb, qf_inums[type]);
5237 if (IS_ERR(qf_inode)) {
5238 ext4_error(sb, "Bad quota inode # %lu", qf_inums[type]);
5239 return PTR_ERR(qf_inode);
5242 /* Don't account quota for quota files to avoid recursion */
5243 qf_inode->i_flags |= S_NOQUOTA;
5244 err = dquot_enable(qf_inode, type, format_id, flags);
5250 /* Enable usage tracking for all quota types. */
5251 static int ext4_enable_quotas(struct super_block *sb)
5254 unsigned long qf_inums[EXT4_MAXQUOTAS] = {
5255 le32_to_cpu(EXT4_SB(sb)->s_es->s_usr_quota_inum),
5256 le32_to_cpu(EXT4_SB(sb)->s_es->s_grp_quota_inum)
5259 sb_dqopt(sb)->flags |= DQUOT_QUOTA_SYS_FILE;
5260 for (type = 0; type < EXT4_MAXQUOTAS; type++) {
5261 if (qf_inums[type]) {
5262 err = ext4_quota_enable(sb, type, QFMT_VFS_V1,
5263 DQUOT_USAGE_ENABLED);
5266 "Failed to enable quota tracking "
5267 "(type=%d, err=%d). Please run "
5268 "e2fsck to fix.", type, err);
5277 * quota_on function that is used when QUOTA feature is set.
5279 static int ext4_quota_on_sysfile(struct super_block *sb, int type,
5282 if (!EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_QUOTA))
5286 * USAGE was enabled at mount time. Only need to enable LIMITS now.
5288 return ext4_quota_enable(sb, type, format_id, DQUOT_LIMITS_ENABLED);
5291 static int ext4_quota_off(struct super_block *sb, int type)
5293 struct inode *inode = sb_dqopt(sb)->files[type];
5296 /* Force all delayed allocation blocks to be allocated.
5297 * Caller already holds s_umount sem */
5298 if (test_opt(sb, DELALLOC))
5299 sync_filesystem(sb);
5304 /* Update modification times of quota files when userspace can
5305 * start looking at them */
5306 handle = ext4_journal_start(inode, EXT4_HT_QUOTA, 1);
5309 inode->i_mtime = inode->i_ctime = CURRENT_TIME;
5310 ext4_mark_inode_dirty(handle, inode);
5311 ext4_journal_stop(handle);
5314 return dquot_quota_off(sb, type);
5318 * quota_off function that is used when QUOTA feature is set.
5320 static int ext4_quota_off_sysfile(struct super_block *sb, int type)
5322 if (!EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_QUOTA))
5325 /* Disable only the limits. */
5326 return dquot_disable(sb, type, DQUOT_LIMITS_ENABLED);
5329 /* Read data from quotafile - avoid pagecache and such because we cannot afford
5330 * acquiring the locks... As quota files are never truncated and quota code
5331 * itself serializes the operations (and no one else should touch the files)
5332 * we don't have to be afraid of races */
5333 static ssize_t ext4_quota_read(struct super_block *sb, int type, char *data,
5334 size_t len, loff_t off)
5336 struct inode *inode = sb_dqopt(sb)->files[type];
5337 ext4_lblk_t blk = off >> EXT4_BLOCK_SIZE_BITS(sb);
5338 int offset = off & (sb->s_blocksize - 1);
5341 struct buffer_head *bh;
5342 loff_t i_size = i_size_read(inode);
5346 if (off+len > i_size)
5349 while (toread > 0) {
5350 tocopy = sb->s_blocksize - offset < toread ?
5351 sb->s_blocksize - offset : toread;
5352 bh = ext4_bread(NULL, inode, blk, 0);
5355 if (!bh) /* A hole? */
5356 memset(data, 0, tocopy);
5358 memcpy(data, bh->b_data+offset, tocopy);
5368 /* Write to quotafile (we know the transaction is already started and has
5369 * enough credits) */
5370 static ssize_t ext4_quota_write(struct super_block *sb, int type,
5371 const char *data, size_t len, loff_t off)
5373 struct inode *inode = sb_dqopt(sb)->files[type];
5374 ext4_lblk_t blk = off >> EXT4_BLOCK_SIZE_BITS(sb);
5375 int err, offset = off & (sb->s_blocksize - 1);
5376 struct buffer_head *bh;
5377 handle_t *handle = journal_current_handle();
5379 if (EXT4_SB(sb)->s_journal && !handle) {
5380 ext4_msg(sb, KERN_WARNING, "Quota write (off=%llu, len=%llu)"
5381 " cancelled because transaction is not started",
5382 (unsigned long long)off, (unsigned long long)len);
5386 * Since we account only one data block in transaction credits,
5387 * then it is impossible to cross a block boundary.
5389 if (sb->s_blocksize - offset < len) {
5390 ext4_msg(sb, KERN_WARNING, "Quota write (off=%llu, len=%llu)"
5391 " cancelled because not block aligned",
5392 (unsigned long long)off, (unsigned long long)len);
5396 bh = ext4_bread(handle, inode, blk, 1);
5401 BUFFER_TRACE(bh, "get write access");
5402 err = ext4_journal_get_write_access(handle, bh);
5408 memcpy(bh->b_data+offset, data, len);
5409 flush_dcache_page(bh->b_page);
5411 err = ext4_handle_dirty_metadata(handle, NULL, bh);
5414 if (inode->i_size < off + len) {
5415 i_size_write(inode, off + len);
5416 EXT4_I(inode)->i_disksize = inode->i_size;
5417 ext4_mark_inode_dirty(handle, inode);
5424 static struct dentry *ext4_mount(struct file_system_type *fs_type, int flags,
5425 const char *dev_name, void *data)
5427 return mount_bdev(fs_type, flags, dev_name, data, ext4_fill_super);
5430 #if !defined(CONFIG_EXT2_FS) && !defined(CONFIG_EXT2_FS_MODULE) && defined(CONFIG_EXT4_USE_FOR_EXT23)
5431 static inline void register_as_ext2(void)
5433 int err = register_filesystem(&ext2_fs_type);
5436 "EXT4-fs: Unable to register as ext2 (%d)\n", err);
5439 static inline void unregister_as_ext2(void)
5441 unregister_filesystem(&ext2_fs_type);
5444 static inline int ext2_feature_set_ok(struct super_block *sb)
5446 if (EXT4_HAS_INCOMPAT_FEATURE(sb, ~EXT2_FEATURE_INCOMPAT_SUPP))
5448 if (sb->s_flags & MS_RDONLY)
5450 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, ~EXT2_FEATURE_RO_COMPAT_SUPP))
5455 static inline void register_as_ext2(void) { }
5456 static inline void unregister_as_ext2(void) { }
5457 static inline int ext2_feature_set_ok(struct super_block *sb) { return 0; }
5460 #if !defined(CONFIG_EXT3_FS) && !defined(CONFIG_EXT3_FS_MODULE) && defined(CONFIG_EXT4_USE_FOR_EXT23)
5461 static inline void register_as_ext3(void)
5463 int err = register_filesystem(&ext3_fs_type);
5466 "EXT4-fs: Unable to register as ext3 (%d)\n", err);
5469 static inline void unregister_as_ext3(void)
5471 unregister_filesystem(&ext3_fs_type);
5474 static inline int ext3_feature_set_ok(struct super_block *sb)
5476 if (EXT4_HAS_INCOMPAT_FEATURE(sb, ~EXT3_FEATURE_INCOMPAT_SUPP))
5478 if (!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL))
5480 if (sb->s_flags & MS_RDONLY)
5482 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, ~EXT3_FEATURE_RO_COMPAT_SUPP))
5487 static inline void register_as_ext3(void) { }
5488 static inline void unregister_as_ext3(void) { }
5489 static inline int ext3_feature_set_ok(struct super_block *sb) { return 0; }
5492 static struct file_system_type ext4_fs_type = {
5493 .owner = THIS_MODULE,
5495 .mount = ext4_mount,
5496 .kill_sb = kill_block_super,
5497 .fs_flags = FS_REQUIRES_DEV,
5499 MODULE_ALIAS_FS("ext4");
5501 static int __init ext4_init_feat_adverts(void)
5503 struct ext4_features *ef;
5506 ef = kzalloc(sizeof(struct ext4_features), GFP_KERNEL);
5510 ef->f_kobj.kset = ext4_kset;
5511 init_completion(&ef->f_kobj_unregister);
5512 ret = kobject_init_and_add(&ef->f_kobj, &ext4_feat_ktype, NULL,
5525 static void ext4_exit_feat_adverts(void)
5527 kobject_put(&ext4_feat->f_kobj);
5528 wait_for_completion(&ext4_feat->f_kobj_unregister);
5532 /* Shared across all ext4 file systems */
5533 wait_queue_head_t ext4__ioend_wq[EXT4_WQ_HASH_SZ];
5534 struct mutex ext4__aio_mutex[EXT4_WQ_HASH_SZ];
5536 static int __init ext4_init_fs(void)
5540 ext4_li_info = NULL;
5541 mutex_init(&ext4_li_mtx);
5543 /* Build-time check for flags consistency */
5544 ext4_check_flag_values();
5546 for (i = 0; i < EXT4_WQ_HASH_SZ; i++) {
5547 mutex_init(&ext4__aio_mutex[i]);
5548 init_waitqueue_head(&ext4__ioend_wq[i]);
5551 err = ext4_init_es();
5555 err = ext4_init_pageio();
5559 err = ext4_init_system_zone();
5562 ext4_kset = kset_create_and_add("ext4", NULL, fs_kobj);
5567 ext4_proc_root = proc_mkdir("fs/ext4", NULL);
5569 err = ext4_init_feat_adverts();
5573 err = ext4_init_mballoc();
5577 ext4_mballoc_ready = 1;
5578 err = init_inodecache();
5583 err = register_filesystem(&ext4_fs_type);
5589 unregister_as_ext2();
5590 unregister_as_ext3();
5591 destroy_inodecache();
5593 ext4_mballoc_ready = 0;
5594 ext4_exit_mballoc();
5596 ext4_exit_feat_adverts();
5599 remove_proc_entry("fs/ext4", NULL);
5600 kset_unregister(ext4_kset);
5602 ext4_exit_system_zone();
5611 static void __exit ext4_exit_fs(void)
5613 ext4_destroy_lazyinit_thread();
5614 unregister_as_ext2();
5615 unregister_as_ext3();
5616 unregister_filesystem(&ext4_fs_type);
5617 destroy_inodecache();
5618 ext4_exit_mballoc();
5619 ext4_exit_feat_adverts();
5620 remove_proc_entry("fs/ext4", NULL);
5621 kset_unregister(ext4_kset);
5622 ext4_exit_system_zone();
5627 MODULE_AUTHOR("Remy Card, Stephen Tweedie, Andrew Morton, Andreas Dilger, Theodore Ts'o and others");
5628 MODULE_DESCRIPTION("Fourth Extended Filesystem");
5629 MODULE_LICENSE("GPL");
5630 module_init(ext4_init_fs)
5631 module_exit(ext4_exit_fs)