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/slab.h>
25 #include <linux/init.h>
26 #include <linux/blkdev.h>
27 #include <linux/parser.h>
28 #include <linux/buffer_head.h>
29 #include <linux/exportfs.h>
30 #include <linux/vfs.h>
31 #include <linux/random.h>
32 #include <linux/mount.h>
33 #include <linux/namei.h>
34 #include <linux/quotaops.h>
35 #include <linux/seq_file.h>
36 #include <linux/proc_fs.h>
37 #include <linux/ctype.h>
38 #include <linux/log2.h>
39 #include <linux/crc16.h>
40 #include <linux/cleancache.h>
41 #include <asm/uaccess.h>
43 #include <linux/kthread.h>
44 #include <linux/freezer.h>
47 #include "ext4_extents.h" /* Needed for trace points definition */
48 #include "ext4_jbd2.h"
53 #define CREATE_TRACE_POINTS
54 #include <trace/events/ext4.h>
56 static struct proc_dir_entry *ext4_proc_root;
57 static struct kset *ext4_kset;
58 static struct ext4_lazy_init *ext4_li_info;
59 static struct mutex ext4_li_mtx;
60 static struct ext4_features *ext4_feat;
61 static int ext4_mballoc_ready;
63 static int ext4_load_journal(struct super_block *, struct ext4_super_block *,
64 unsigned long journal_devnum);
65 static int ext4_show_options(struct seq_file *seq, struct dentry *root);
66 static int ext4_commit_super(struct super_block *sb, int sync);
67 static void ext4_mark_recovery_complete(struct super_block *sb,
68 struct ext4_super_block *es);
69 static void ext4_clear_journal_err(struct super_block *sb,
70 struct ext4_super_block *es);
71 static int ext4_sync_fs(struct super_block *sb, int wait);
72 static int ext4_remount(struct super_block *sb, int *flags, char *data);
73 static int ext4_statfs(struct dentry *dentry, struct kstatfs *buf);
74 static int ext4_unfreeze(struct super_block *sb);
75 static int ext4_freeze(struct super_block *sb);
76 static struct dentry *ext4_mount(struct file_system_type *fs_type, int flags,
77 const char *dev_name, void *data);
78 static inline int ext2_feature_set_ok(struct super_block *sb);
79 static inline int ext3_feature_set_ok(struct super_block *sb);
80 static int ext4_feature_set_ok(struct super_block *sb, int readonly);
81 static void ext4_destroy_lazyinit_thread(void);
82 static void ext4_unregister_li_request(struct super_block *sb);
83 static void ext4_clear_request_list(void);
84 static int ext4_reserve_clusters(struct ext4_sb_info *, ext4_fsblk_t);
86 #if !defined(CONFIG_EXT2_FS) && !defined(CONFIG_EXT2_FS_MODULE) && defined(CONFIG_EXT4_USE_FOR_EXT23)
87 static struct file_system_type ext2_fs_type = {
91 .kill_sb = kill_block_super,
92 .fs_flags = FS_REQUIRES_DEV,
94 MODULE_ALIAS_FS("ext2");
96 #define IS_EXT2_SB(sb) ((sb)->s_bdev->bd_holder == &ext2_fs_type)
98 #define IS_EXT2_SB(sb) (0)
102 #if !defined(CONFIG_EXT3_FS) && !defined(CONFIG_EXT3_FS_MODULE) && defined(CONFIG_EXT4_USE_FOR_EXT23)
103 static struct file_system_type ext3_fs_type = {
104 .owner = THIS_MODULE,
107 .kill_sb = kill_block_super,
108 .fs_flags = FS_REQUIRES_DEV,
110 MODULE_ALIAS_FS("ext3");
111 MODULE_ALIAS("ext3");
112 #define IS_EXT3_SB(sb) ((sb)->s_bdev->bd_holder == &ext3_fs_type)
114 #define IS_EXT3_SB(sb) (0)
117 static int ext4_verify_csum_type(struct super_block *sb,
118 struct ext4_super_block *es)
120 if (!EXT4_HAS_RO_COMPAT_FEATURE(sb,
121 EXT4_FEATURE_RO_COMPAT_METADATA_CSUM))
124 return es->s_checksum_type == EXT4_CRC32C_CHKSUM;
127 static __le32 ext4_superblock_csum(struct super_block *sb,
128 struct ext4_super_block *es)
130 struct ext4_sb_info *sbi = EXT4_SB(sb);
131 int offset = offsetof(struct ext4_super_block, s_checksum);
134 csum = ext4_chksum(sbi, ~0, (char *)es, offset);
136 return cpu_to_le32(csum);
139 static int ext4_superblock_csum_verify(struct super_block *sb,
140 struct ext4_super_block *es)
142 if (!ext4_has_metadata_csum(sb))
145 return es->s_checksum == ext4_superblock_csum(sb, es);
148 void ext4_superblock_csum_set(struct super_block *sb)
150 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
152 if (!ext4_has_metadata_csum(sb))
155 es->s_checksum = ext4_superblock_csum(sb, es);
158 void *ext4_kvmalloc(size_t size, gfp_t flags)
162 ret = kmalloc(size, flags | __GFP_NOWARN);
164 ret = __vmalloc(size, flags, PAGE_KERNEL);
168 void *ext4_kvzalloc(size_t size, gfp_t flags)
172 ret = kzalloc(size, flags | __GFP_NOWARN);
174 ret = __vmalloc(size, flags | __GFP_ZERO, PAGE_KERNEL);
178 ext4_fsblk_t ext4_block_bitmap(struct super_block *sb,
179 struct ext4_group_desc *bg)
181 return le32_to_cpu(bg->bg_block_bitmap_lo) |
182 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
183 (ext4_fsblk_t)le32_to_cpu(bg->bg_block_bitmap_hi) << 32 : 0);
186 ext4_fsblk_t ext4_inode_bitmap(struct super_block *sb,
187 struct ext4_group_desc *bg)
189 return le32_to_cpu(bg->bg_inode_bitmap_lo) |
190 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
191 (ext4_fsblk_t)le32_to_cpu(bg->bg_inode_bitmap_hi) << 32 : 0);
194 ext4_fsblk_t ext4_inode_table(struct super_block *sb,
195 struct ext4_group_desc *bg)
197 return le32_to_cpu(bg->bg_inode_table_lo) |
198 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
199 (ext4_fsblk_t)le32_to_cpu(bg->bg_inode_table_hi) << 32 : 0);
202 __u32 ext4_free_group_clusters(struct super_block *sb,
203 struct ext4_group_desc *bg)
205 return le16_to_cpu(bg->bg_free_blocks_count_lo) |
206 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
207 (__u32)le16_to_cpu(bg->bg_free_blocks_count_hi) << 16 : 0);
210 __u32 ext4_free_inodes_count(struct super_block *sb,
211 struct ext4_group_desc *bg)
213 return le16_to_cpu(bg->bg_free_inodes_count_lo) |
214 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
215 (__u32)le16_to_cpu(bg->bg_free_inodes_count_hi) << 16 : 0);
218 __u32 ext4_used_dirs_count(struct super_block *sb,
219 struct ext4_group_desc *bg)
221 return le16_to_cpu(bg->bg_used_dirs_count_lo) |
222 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
223 (__u32)le16_to_cpu(bg->bg_used_dirs_count_hi) << 16 : 0);
226 __u32 ext4_itable_unused_count(struct super_block *sb,
227 struct ext4_group_desc *bg)
229 return le16_to_cpu(bg->bg_itable_unused_lo) |
230 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
231 (__u32)le16_to_cpu(bg->bg_itable_unused_hi) << 16 : 0);
234 void ext4_block_bitmap_set(struct super_block *sb,
235 struct ext4_group_desc *bg, ext4_fsblk_t blk)
237 bg->bg_block_bitmap_lo = cpu_to_le32((u32)blk);
238 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
239 bg->bg_block_bitmap_hi = cpu_to_le32(blk >> 32);
242 void ext4_inode_bitmap_set(struct super_block *sb,
243 struct ext4_group_desc *bg, ext4_fsblk_t blk)
245 bg->bg_inode_bitmap_lo = cpu_to_le32((u32)blk);
246 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
247 bg->bg_inode_bitmap_hi = cpu_to_le32(blk >> 32);
250 void ext4_inode_table_set(struct super_block *sb,
251 struct ext4_group_desc *bg, ext4_fsblk_t blk)
253 bg->bg_inode_table_lo = cpu_to_le32((u32)blk);
254 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
255 bg->bg_inode_table_hi = cpu_to_le32(blk >> 32);
258 void ext4_free_group_clusters_set(struct super_block *sb,
259 struct ext4_group_desc *bg, __u32 count)
261 bg->bg_free_blocks_count_lo = cpu_to_le16((__u16)count);
262 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
263 bg->bg_free_blocks_count_hi = cpu_to_le16(count >> 16);
266 void ext4_free_inodes_set(struct super_block *sb,
267 struct ext4_group_desc *bg, __u32 count)
269 bg->bg_free_inodes_count_lo = cpu_to_le16((__u16)count);
270 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
271 bg->bg_free_inodes_count_hi = cpu_to_le16(count >> 16);
274 void ext4_used_dirs_set(struct super_block *sb,
275 struct ext4_group_desc *bg, __u32 count)
277 bg->bg_used_dirs_count_lo = cpu_to_le16((__u16)count);
278 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
279 bg->bg_used_dirs_count_hi = cpu_to_le16(count >> 16);
282 void ext4_itable_unused_set(struct super_block *sb,
283 struct ext4_group_desc *bg, __u32 count)
285 bg->bg_itable_unused_lo = cpu_to_le16((__u16)count);
286 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
287 bg->bg_itable_unused_hi = cpu_to_le16(count >> 16);
291 static void __save_error_info(struct super_block *sb, const char *func,
294 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
296 EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS;
297 if (bdev_read_only(sb->s_bdev))
299 es->s_state |= cpu_to_le16(EXT4_ERROR_FS);
300 es->s_last_error_time = cpu_to_le32(get_seconds());
301 strncpy(es->s_last_error_func, func, sizeof(es->s_last_error_func));
302 es->s_last_error_line = cpu_to_le32(line);
303 if (!es->s_first_error_time) {
304 es->s_first_error_time = es->s_last_error_time;
305 strncpy(es->s_first_error_func, func,
306 sizeof(es->s_first_error_func));
307 es->s_first_error_line = cpu_to_le32(line);
308 es->s_first_error_ino = es->s_last_error_ino;
309 es->s_first_error_block = es->s_last_error_block;
312 * Start the daily error reporting function if it hasn't been
315 if (!es->s_error_count)
316 mod_timer(&EXT4_SB(sb)->s_err_report, jiffies + 24*60*60*HZ);
317 le32_add_cpu(&es->s_error_count, 1);
320 static void save_error_info(struct super_block *sb, const char *func,
323 __save_error_info(sb, func, line);
324 ext4_commit_super(sb, 1);
327 static void ext4_journal_commit_callback(journal_t *journal, transaction_t *txn)
329 struct super_block *sb = journal->j_private;
330 struct ext4_sb_info *sbi = EXT4_SB(sb);
331 int error = is_journal_aborted(journal);
332 struct ext4_journal_cb_entry *jce;
334 BUG_ON(txn->t_state == T_FINISHED);
335 spin_lock(&sbi->s_md_lock);
336 while (!list_empty(&txn->t_private_list)) {
337 jce = list_entry(txn->t_private_list.next,
338 struct ext4_journal_cb_entry, jce_list);
339 list_del_init(&jce->jce_list);
340 spin_unlock(&sbi->s_md_lock);
341 jce->jce_func(sb, jce, error);
342 spin_lock(&sbi->s_md_lock);
344 spin_unlock(&sbi->s_md_lock);
347 /* Deal with the reporting of failure conditions on a filesystem such as
348 * inconsistencies detected or read IO failures.
350 * On ext2, we can store the error state of the filesystem in the
351 * superblock. That is not possible on ext4, because we may have other
352 * write ordering constraints on the superblock which prevent us from
353 * writing it out straight away; and given that the journal is about to
354 * be aborted, we can't rely on the current, or future, transactions to
355 * write out the superblock safely.
357 * We'll just use the jbd2_journal_abort() error code to record an error in
358 * the journal instead. On recovery, the journal will complain about
359 * that error until we've noted it down and cleared it.
362 static void ext4_handle_error(struct super_block *sb)
364 if (sb->s_flags & MS_RDONLY)
367 if (!test_opt(sb, ERRORS_CONT)) {
368 journal_t *journal = EXT4_SB(sb)->s_journal;
370 EXT4_SB(sb)->s_mount_flags |= EXT4_MF_FS_ABORTED;
372 jbd2_journal_abort(journal, -EIO);
374 if (test_opt(sb, ERRORS_RO)) {
375 ext4_msg(sb, KERN_CRIT, "Remounting filesystem read-only");
377 * Make sure updated value of ->s_mount_flags will be visible
378 * before ->s_flags update
381 sb->s_flags |= MS_RDONLY;
383 if (test_opt(sb, ERRORS_PANIC))
384 panic("EXT4-fs (device %s): panic forced after error\n",
388 #define ext4_error_ratelimit(sb) \
389 ___ratelimit(&(EXT4_SB(sb)->s_err_ratelimit_state), \
392 void __ext4_error(struct super_block *sb, const char *function,
393 unsigned int line, const char *fmt, ...)
395 struct va_format vaf;
398 if (ext4_error_ratelimit(sb)) {
403 "EXT4-fs error (device %s): %s:%d: comm %s: %pV\n",
404 sb->s_id, function, line, current->comm, &vaf);
407 save_error_info(sb, function, line);
408 ext4_handle_error(sb);
411 void __ext4_error_inode(struct inode *inode, const char *function,
412 unsigned int line, ext4_fsblk_t block,
413 const char *fmt, ...)
416 struct va_format vaf;
417 struct ext4_super_block *es = EXT4_SB(inode->i_sb)->s_es;
419 es->s_last_error_ino = cpu_to_le32(inode->i_ino);
420 es->s_last_error_block = cpu_to_le64(block);
421 if (ext4_error_ratelimit(inode->i_sb)) {
426 printk(KERN_CRIT "EXT4-fs error (device %s): %s:%d: "
427 "inode #%lu: block %llu: comm %s: %pV\n",
428 inode->i_sb->s_id, function, line, inode->i_ino,
429 block, current->comm, &vaf);
431 printk(KERN_CRIT "EXT4-fs error (device %s): %s:%d: "
432 "inode #%lu: comm %s: %pV\n",
433 inode->i_sb->s_id, function, line, inode->i_ino,
434 current->comm, &vaf);
437 save_error_info(inode->i_sb, function, line);
438 ext4_handle_error(inode->i_sb);
441 void __ext4_error_file(struct file *file, const char *function,
442 unsigned int line, ext4_fsblk_t block,
443 const char *fmt, ...)
446 struct va_format vaf;
447 struct ext4_super_block *es;
448 struct inode *inode = file_inode(file);
449 char pathname[80], *path;
451 es = EXT4_SB(inode->i_sb)->s_es;
452 es->s_last_error_ino = cpu_to_le32(inode->i_ino);
453 if (ext4_error_ratelimit(inode->i_sb)) {
454 path = d_path(&(file->f_path), pathname, sizeof(pathname));
462 "EXT4-fs error (device %s): %s:%d: inode #%lu: "
463 "block %llu: comm %s: path %s: %pV\n",
464 inode->i_sb->s_id, function, line, inode->i_ino,
465 block, current->comm, path, &vaf);
468 "EXT4-fs error (device %s): %s:%d: inode #%lu: "
469 "comm %s: path %s: %pV\n",
470 inode->i_sb->s_id, function, line, inode->i_ino,
471 current->comm, path, &vaf);
474 save_error_info(inode->i_sb, function, line);
475 ext4_handle_error(inode->i_sb);
478 const char *ext4_decode_error(struct super_block *sb, int errno,
485 errstr = "IO failure";
488 errstr = "Out of memory";
491 if (!sb || (EXT4_SB(sb)->s_journal &&
492 EXT4_SB(sb)->s_journal->j_flags & JBD2_ABORT))
493 errstr = "Journal has aborted";
495 errstr = "Readonly filesystem";
498 /* If the caller passed in an extra buffer for unknown
499 * errors, textualise them now. Else we just return
502 /* Check for truncated error codes... */
503 if (snprintf(nbuf, 16, "error %d", -errno) >= 0)
512 /* __ext4_std_error decodes expected errors from journaling functions
513 * automatically and invokes the appropriate error response. */
515 void __ext4_std_error(struct super_block *sb, const char *function,
516 unsigned int line, int errno)
521 /* Special case: if the error is EROFS, and we're not already
522 * inside a transaction, then there's really no point in logging
524 if (errno == -EROFS && journal_current_handle() == NULL &&
525 (sb->s_flags & MS_RDONLY))
528 if (ext4_error_ratelimit(sb)) {
529 errstr = ext4_decode_error(sb, errno, nbuf);
530 printk(KERN_CRIT "EXT4-fs error (device %s) in %s:%d: %s\n",
531 sb->s_id, function, line, errstr);
534 save_error_info(sb, function, line);
535 ext4_handle_error(sb);
539 * ext4_abort is a much stronger failure handler than ext4_error. The
540 * abort function may be used to deal with unrecoverable failures such
541 * as journal IO errors or ENOMEM at a critical moment in log management.
543 * We unconditionally force the filesystem into an ABORT|READONLY state,
544 * unless the error response on the fs has been set to panic in which
545 * case we take the easy way out and panic immediately.
548 void __ext4_abort(struct super_block *sb, const char *function,
549 unsigned int line, const char *fmt, ...)
553 save_error_info(sb, function, line);
555 printk(KERN_CRIT "EXT4-fs error (device %s): %s:%d: ", sb->s_id,
561 if ((sb->s_flags & MS_RDONLY) == 0) {
562 ext4_msg(sb, KERN_CRIT, "Remounting filesystem read-only");
563 EXT4_SB(sb)->s_mount_flags |= EXT4_MF_FS_ABORTED;
565 * Make sure updated value of ->s_mount_flags will be visible
566 * before ->s_flags update
569 sb->s_flags |= MS_RDONLY;
570 if (EXT4_SB(sb)->s_journal)
571 jbd2_journal_abort(EXT4_SB(sb)->s_journal, -EIO);
572 save_error_info(sb, function, line);
574 if (test_opt(sb, ERRORS_PANIC))
575 panic("EXT4-fs panic from previous error\n");
578 void __ext4_msg(struct super_block *sb,
579 const char *prefix, const char *fmt, ...)
581 struct va_format vaf;
584 if (!___ratelimit(&(EXT4_SB(sb)->s_msg_ratelimit_state), "EXT4-fs"))
590 printk("%sEXT4-fs (%s): %pV\n", prefix, sb->s_id, &vaf);
594 void __ext4_warning(struct super_block *sb, const char *function,
595 unsigned int line, const char *fmt, ...)
597 struct va_format vaf;
600 if (!___ratelimit(&(EXT4_SB(sb)->s_warning_ratelimit_state),
607 printk(KERN_WARNING "EXT4-fs warning (device %s): %s:%d: %pV\n",
608 sb->s_id, function, line, &vaf);
612 void __ext4_grp_locked_error(const char *function, unsigned int line,
613 struct super_block *sb, ext4_group_t grp,
614 unsigned long ino, ext4_fsblk_t block,
615 const char *fmt, ...)
619 struct va_format vaf;
621 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
623 es->s_last_error_ino = cpu_to_le32(ino);
624 es->s_last_error_block = cpu_to_le64(block);
625 __save_error_info(sb, function, line);
627 if (ext4_error_ratelimit(sb)) {
631 printk(KERN_CRIT "EXT4-fs error (device %s): %s:%d: group %u, ",
632 sb->s_id, function, line, grp);
634 printk(KERN_CONT "inode %lu: ", ino);
636 printk(KERN_CONT "block %llu:",
637 (unsigned long long) block);
638 printk(KERN_CONT "%pV\n", &vaf);
642 if (test_opt(sb, ERRORS_CONT)) {
643 ext4_commit_super(sb, 0);
647 ext4_unlock_group(sb, grp);
648 ext4_handle_error(sb);
650 * We only get here in the ERRORS_RO case; relocking the group
651 * may be dangerous, but nothing bad will happen since the
652 * filesystem will have already been marked read/only and the
653 * journal has been aborted. We return 1 as a hint to callers
654 * who might what to use the return value from
655 * ext4_grp_locked_error() to distinguish between the
656 * ERRORS_CONT and ERRORS_RO case, and perhaps return more
657 * aggressively from the ext4 function in question, with a
658 * more appropriate error code.
660 ext4_lock_group(sb, grp);
664 void ext4_update_dynamic_rev(struct super_block *sb)
666 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
668 if (le32_to_cpu(es->s_rev_level) > EXT4_GOOD_OLD_REV)
672 "updating to rev %d because of new feature flag, "
673 "running e2fsck is recommended",
676 es->s_first_ino = cpu_to_le32(EXT4_GOOD_OLD_FIRST_INO);
677 es->s_inode_size = cpu_to_le16(EXT4_GOOD_OLD_INODE_SIZE);
678 es->s_rev_level = cpu_to_le32(EXT4_DYNAMIC_REV);
679 /* leave es->s_feature_*compat flags alone */
680 /* es->s_uuid will be set by e2fsck if empty */
683 * The rest of the superblock fields should be zero, and if not it
684 * means they are likely already in use, so leave them alone. We
685 * can leave it up to e2fsck to clean up any inconsistencies there.
690 * Open the external journal device
692 static struct block_device *ext4_blkdev_get(dev_t dev, struct super_block *sb)
694 struct block_device *bdev;
695 char b[BDEVNAME_SIZE];
697 bdev = blkdev_get_by_dev(dev, FMODE_READ|FMODE_WRITE|FMODE_EXCL, sb);
703 ext4_msg(sb, KERN_ERR, "failed to open journal device %s: %ld",
704 __bdevname(dev, b), PTR_ERR(bdev));
709 * Release the journal device
711 static void ext4_blkdev_put(struct block_device *bdev)
713 blkdev_put(bdev, FMODE_READ|FMODE_WRITE|FMODE_EXCL);
716 static void ext4_blkdev_remove(struct ext4_sb_info *sbi)
718 struct block_device *bdev;
719 bdev = sbi->journal_bdev;
721 ext4_blkdev_put(bdev);
722 sbi->journal_bdev = NULL;
726 static inline struct inode *orphan_list_entry(struct list_head *l)
728 return &list_entry(l, struct ext4_inode_info, i_orphan)->vfs_inode;
731 static void dump_orphan_list(struct super_block *sb, struct ext4_sb_info *sbi)
735 ext4_msg(sb, KERN_ERR, "sb orphan head is %d",
736 le32_to_cpu(sbi->s_es->s_last_orphan));
738 printk(KERN_ERR "sb_info orphan list:\n");
739 list_for_each(l, &sbi->s_orphan) {
740 struct inode *inode = orphan_list_entry(l);
742 "inode %s:%lu at %p: mode %o, nlink %d, next %d\n",
743 inode->i_sb->s_id, inode->i_ino, inode,
744 inode->i_mode, inode->i_nlink,
749 static void ext4_put_super(struct super_block *sb)
751 struct ext4_sb_info *sbi = EXT4_SB(sb);
752 struct ext4_super_block *es = sbi->s_es;
755 ext4_unregister_li_request(sb);
756 dquot_disable(sb, -1, DQUOT_USAGE_ENABLED | DQUOT_LIMITS_ENABLED);
758 flush_workqueue(sbi->rsv_conversion_wq);
759 destroy_workqueue(sbi->rsv_conversion_wq);
761 if (sbi->s_journal) {
762 err = jbd2_journal_destroy(sbi->s_journal);
763 sbi->s_journal = NULL;
765 ext4_abort(sb, "Couldn't clean up the journal");
768 ext4_es_unregister_shrinker(sbi);
769 del_timer_sync(&sbi->s_err_report);
770 ext4_release_system_zone(sb);
772 ext4_ext_release(sb);
773 ext4_xattr_put_super(sb);
775 if (!(sb->s_flags & MS_RDONLY)) {
776 EXT4_CLEAR_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
777 es->s_state = cpu_to_le16(sbi->s_mount_state);
779 if (!(sb->s_flags & MS_RDONLY))
780 ext4_commit_super(sb, 1);
783 remove_proc_entry("options", sbi->s_proc);
784 remove_proc_entry(sb->s_id, ext4_proc_root);
786 kobject_del(&sbi->s_kobj);
788 for (i = 0; i < sbi->s_gdb_count; i++)
789 brelse(sbi->s_group_desc[i]);
790 kvfree(sbi->s_group_desc);
791 kvfree(sbi->s_flex_groups);
792 percpu_counter_destroy(&sbi->s_freeclusters_counter);
793 percpu_counter_destroy(&sbi->s_freeinodes_counter);
794 percpu_counter_destroy(&sbi->s_dirs_counter);
795 percpu_counter_destroy(&sbi->s_dirtyclusters_counter);
798 for (i = 0; i < EXT4_MAXQUOTAS; i++)
799 kfree(sbi->s_qf_names[i]);
802 /* Debugging code just in case the in-memory inode orphan list
803 * isn't empty. The on-disk one can be non-empty if we've
804 * detected an error and taken the fs readonly, but the
805 * in-memory list had better be clean by this point. */
806 if (!list_empty(&sbi->s_orphan))
807 dump_orphan_list(sb, sbi);
808 J_ASSERT(list_empty(&sbi->s_orphan));
810 sync_blockdev(sb->s_bdev);
811 invalidate_bdev(sb->s_bdev);
812 if (sbi->journal_bdev && sbi->journal_bdev != sb->s_bdev) {
814 * Invalidate the journal device's buffers. We don't want them
815 * floating about in memory - the physical journal device may
816 * hotswapped, and it breaks the `ro-after' testing code.
818 sync_blockdev(sbi->journal_bdev);
819 invalidate_bdev(sbi->journal_bdev);
820 ext4_blkdev_remove(sbi);
822 if (sbi->s_mb_cache) {
823 ext4_xattr_destroy_cache(sbi->s_mb_cache);
824 sbi->s_mb_cache = NULL;
827 kthread_stop(sbi->s_mmp_tsk);
828 sb->s_fs_info = NULL;
830 * Now that we are completely done shutting down the
831 * superblock, we need to actually destroy the kobject.
833 kobject_put(&sbi->s_kobj);
834 wait_for_completion(&sbi->s_kobj_unregister);
835 if (sbi->s_chksum_driver)
836 crypto_free_shash(sbi->s_chksum_driver);
837 kfree(sbi->s_blockgroup_lock);
841 static struct kmem_cache *ext4_inode_cachep;
844 * Called inside transaction, so use GFP_NOFS
846 static struct inode *ext4_alloc_inode(struct super_block *sb)
848 struct ext4_inode_info *ei;
850 ei = kmem_cache_alloc(ext4_inode_cachep, GFP_NOFS);
854 ei->vfs_inode.i_version = 1;
855 spin_lock_init(&ei->i_raw_lock);
856 INIT_LIST_HEAD(&ei->i_prealloc_list);
857 spin_lock_init(&ei->i_prealloc_lock);
858 ext4_es_init_tree(&ei->i_es_tree);
859 rwlock_init(&ei->i_es_lock);
860 INIT_LIST_HEAD(&ei->i_es_list);
863 ei->i_es_shrink_lblk = 0;
864 ei->i_reserved_data_blocks = 0;
865 ei->i_reserved_meta_blocks = 0;
866 ei->i_allocated_meta_blocks = 0;
867 ei->i_da_metadata_calc_len = 0;
868 ei->i_da_metadata_calc_last_lblock = 0;
869 spin_lock_init(&(ei->i_block_reservation_lock));
871 ei->i_reserved_quota = 0;
872 memset(&ei->i_dquot, 0, sizeof(ei->i_dquot));
875 INIT_LIST_HEAD(&ei->i_rsv_conversion_list);
876 spin_lock_init(&ei->i_completed_io_lock);
878 ei->i_datasync_tid = 0;
879 atomic_set(&ei->i_ioend_count, 0);
880 atomic_set(&ei->i_unwritten, 0);
881 INIT_WORK(&ei->i_rsv_conversion_work, ext4_end_io_rsv_work);
882 #ifdef CONFIG_EXT4_FS_ENCRYPTION
883 ei->i_encryption_key.mode = EXT4_ENCRYPTION_MODE_INVALID;
886 return &ei->vfs_inode;
889 static int ext4_drop_inode(struct inode *inode)
891 int drop = generic_drop_inode(inode);
893 trace_ext4_drop_inode(inode, drop);
897 static void ext4_i_callback(struct rcu_head *head)
899 struct inode *inode = container_of(head, struct inode, i_rcu);
900 kmem_cache_free(ext4_inode_cachep, EXT4_I(inode));
903 static void ext4_destroy_inode(struct inode *inode)
905 if (!list_empty(&(EXT4_I(inode)->i_orphan))) {
906 ext4_msg(inode->i_sb, KERN_ERR,
907 "Inode %lu (%p): orphan list check failed!",
908 inode->i_ino, EXT4_I(inode));
909 print_hex_dump(KERN_INFO, "", DUMP_PREFIX_ADDRESS, 16, 4,
910 EXT4_I(inode), sizeof(struct ext4_inode_info),
914 call_rcu(&inode->i_rcu, ext4_i_callback);
917 static void init_once(void *foo)
919 struct ext4_inode_info *ei = (struct ext4_inode_info *) foo;
921 INIT_LIST_HEAD(&ei->i_orphan);
922 init_rwsem(&ei->xattr_sem);
923 init_rwsem(&ei->i_data_sem);
924 inode_init_once(&ei->vfs_inode);
927 static int __init init_inodecache(void)
929 ext4_inode_cachep = kmem_cache_create("ext4_inode_cache",
930 sizeof(struct ext4_inode_info),
931 0, (SLAB_RECLAIM_ACCOUNT|
934 if (ext4_inode_cachep == NULL)
939 static void destroy_inodecache(void)
942 * Make sure all delayed rcu free inodes are flushed before we
946 kmem_cache_destroy(ext4_inode_cachep);
949 void ext4_clear_inode(struct inode *inode)
951 invalidate_inode_buffers(inode);
954 ext4_discard_preallocations(inode);
955 ext4_es_remove_extent(inode, 0, EXT_MAX_BLOCKS);
956 if (EXT4_I(inode)->jinode) {
957 jbd2_journal_release_jbd_inode(EXT4_JOURNAL(inode),
958 EXT4_I(inode)->jinode);
959 jbd2_free_inode(EXT4_I(inode)->jinode);
960 EXT4_I(inode)->jinode = NULL;
964 static struct inode *ext4_nfs_get_inode(struct super_block *sb,
965 u64 ino, u32 generation)
969 if (ino < EXT4_FIRST_INO(sb) && ino != EXT4_ROOT_INO)
970 return ERR_PTR(-ESTALE);
971 if (ino > le32_to_cpu(EXT4_SB(sb)->s_es->s_inodes_count))
972 return ERR_PTR(-ESTALE);
974 /* iget isn't really right if the inode is currently unallocated!!
976 * ext4_read_inode will return a bad_inode if the inode had been
977 * deleted, so we should be safe.
979 * Currently we don't know the generation for parent directory, so
980 * a generation of 0 means "accept any"
982 inode = ext4_iget_normal(sb, ino);
984 return ERR_CAST(inode);
985 if (generation && inode->i_generation != generation) {
987 return ERR_PTR(-ESTALE);
993 static struct dentry *ext4_fh_to_dentry(struct super_block *sb, struct fid *fid,
994 int fh_len, int fh_type)
996 return generic_fh_to_dentry(sb, fid, fh_len, fh_type,
1000 static struct dentry *ext4_fh_to_parent(struct super_block *sb, struct fid *fid,
1001 int fh_len, int fh_type)
1003 return generic_fh_to_parent(sb, fid, fh_len, fh_type,
1004 ext4_nfs_get_inode);
1008 * Try to release metadata pages (indirect blocks, directories) which are
1009 * mapped via the block device. Since these pages could have journal heads
1010 * which would prevent try_to_free_buffers() from freeing them, we must use
1011 * jbd2 layer's try_to_free_buffers() function to release them.
1013 static int bdev_try_to_free_page(struct super_block *sb, struct page *page,
1016 journal_t *journal = EXT4_SB(sb)->s_journal;
1018 WARN_ON(PageChecked(page));
1019 if (!page_has_buffers(page))
1022 return jbd2_journal_try_to_free_buffers(journal, page,
1023 wait & ~__GFP_WAIT);
1024 return try_to_free_buffers(page);
1028 #define QTYPE2NAME(t) ((t) == USRQUOTA ? "user" : "group")
1029 #define QTYPE2MOPT(on, t) ((t) == USRQUOTA?((on)##USRJQUOTA):((on)##GRPJQUOTA))
1031 static int ext4_write_dquot(struct dquot *dquot);
1032 static int ext4_acquire_dquot(struct dquot *dquot);
1033 static int ext4_release_dquot(struct dquot *dquot);
1034 static int ext4_mark_dquot_dirty(struct dquot *dquot);
1035 static int ext4_write_info(struct super_block *sb, int type);
1036 static int ext4_quota_on(struct super_block *sb, int type, int format_id,
1038 static int ext4_quota_off(struct super_block *sb, int type);
1039 static int ext4_quota_on_mount(struct super_block *sb, int type);
1040 static ssize_t ext4_quota_read(struct super_block *sb, int type, char *data,
1041 size_t len, loff_t off);
1042 static ssize_t ext4_quota_write(struct super_block *sb, int type,
1043 const char *data, size_t len, loff_t off);
1044 static int ext4_quota_enable(struct super_block *sb, int type, int format_id,
1045 unsigned int flags);
1046 static int ext4_enable_quotas(struct super_block *sb);
1048 static struct dquot **ext4_get_dquots(struct inode *inode)
1050 return EXT4_I(inode)->i_dquot;
1053 static const struct dquot_operations ext4_quota_operations = {
1054 .get_reserved_space = ext4_get_reserved_space,
1055 .write_dquot = ext4_write_dquot,
1056 .acquire_dquot = ext4_acquire_dquot,
1057 .release_dquot = ext4_release_dquot,
1058 .mark_dirty = ext4_mark_dquot_dirty,
1059 .write_info = ext4_write_info,
1060 .alloc_dquot = dquot_alloc,
1061 .destroy_dquot = dquot_destroy,
1064 static const struct quotactl_ops ext4_qctl_operations = {
1065 .quota_on = ext4_quota_on,
1066 .quota_off = ext4_quota_off,
1067 .quota_sync = dquot_quota_sync,
1068 .get_state = dquot_get_state,
1069 .set_info = dquot_set_dqinfo,
1070 .get_dqblk = dquot_get_dqblk,
1071 .set_dqblk = dquot_set_dqblk
1075 static const struct super_operations ext4_sops = {
1076 .alloc_inode = ext4_alloc_inode,
1077 .destroy_inode = ext4_destroy_inode,
1078 .write_inode = ext4_write_inode,
1079 .dirty_inode = ext4_dirty_inode,
1080 .drop_inode = ext4_drop_inode,
1081 .evict_inode = ext4_evict_inode,
1082 .put_super = ext4_put_super,
1083 .sync_fs = ext4_sync_fs,
1084 .freeze_fs = ext4_freeze,
1085 .unfreeze_fs = ext4_unfreeze,
1086 .statfs = ext4_statfs,
1087 .remount_fs = ext4_remount,
1088 .show_options = ext4_show_options,
1090 .quota_read = ext4_quota_read,
1091 .quota_write = ext4_quota_write,
1092 .get_dquots = ext4_get_dquots,
1094 .bdev_try_to_free_page = bdev_try_to_free_page,
1097 static const struct export_operations ext4_export_ops = {
1098 .fh_to_dentry = ext4_fh_to_dentry,
1099 .fh_to_parent = ext4_fh_to_parent,
1100 .get_parent = ext4_get_parent,
1104 Opt_bsd_df, Opt_minix_df, Opt_grpid, Opt_nogrpid,
1105 Opt_resgid, Opt_resuid, Opt_sb, Opt_err_cont, Opt_err_panic, Opt_err_ro,
1106 Opt_nouid32, Opt_debug, Opt_removed,
1107 Opt_user_xattr, Opt_nouser_xattr, Opt_acl, Opt_noacl,
1108 Opt_auto_da_alloc, Opt_noauto_da_alloc, Opt_noload,
1109 Opt_commit, Opt_min_batch_time, Opt_max_batch_time, Opt_journal_dev,
1110 Opt_journal_path, Opt_journal_checksum, Opt_journal_async_commit,
1111 Opt_abort, Opt_data_journal, Opt_data_ordered, Opt_data_writeback,
1112 Opt_data_err_abort, Opt_data_err_ignore, Opt_test_dummy_encryption,
1113 Opt_usrjquota, Opt_grpjquota, Opt_offusrjquota, Opt_offgrpjquota,
1114 Opt_jqfmt_vfsold, Opt_jqfmt_vfsv0, Opt_jqfmt_vfsv1, Opt_quota,
1115 Opt_noquota, Opt_barrier, Opt_nobarrier, Opt_err,
1116 Opt_usrquota, Opt_grpquota, Opt_i_version, Opt_dax,
1117 Opt_stripe, Opt_delalloc, Opt_nodelalloc, Opt_mblk_io_submit,
1118 Opt_lazytime, Opt_nolazytime,
1119 Opt_nomblk_io_submit, Opt_block_validity, Opt_noblock_validity,
1120 Opt_inode_readahead_blks, Opt_journal_ioprio,
1121 Opt_dioread_nolock, Opt_dioread_lock,
1122 Opt_discard, Opt_nodiscard, Opt_init_itable, Opt_noinit_itable,
1123 Opt_max_dir_size_kb, Opt_nojournal_checksum,
1126 static const match_table_t tokens = {
1127 {Opt_bsd_df, "bsddf"},
1128 {Opt_minix_df, "minixdf"},
1129 {Opt_grpid, "grpid"},
1130 {Opt_grpid, "bsdgroups"},
1131 {Opt_nogrpid, "nogrpid"},
1132 {Opt_nogrpid, "sysvgroups"},
1133 {Opt_resgid, "resgid=%u"},
1134 {Opt_resuid, "resuid=%u"},
1136 {Opt_err_cont, "errors=continue"},
1137 {Opt_err_panic, "errors=panic"},
1138 {Opt_err_ro, "errors=remount-ro"},
1139 {Opt_nouid32, "nouid32"},
1140 {Opt_debug, "debug"},
1141 {Opt_removed, "oldalloc"},
1142 {Opt_removed, "orlov"},
1143 {Opt_user_xattr, "user_xattr"},
1144 {Opt_nouser_xattr, "nouser_xattr"},
1146 {Opt_noacl, "noacl"},
1147 {Opt_noload, "norecovery"},
1148 {Opt_noload, "noload"},
1149 {Opt_removed, "nobh"},
1150 {Opt_removed, "bh"},
1151 {Opt_commit, "commit=%u"},
1152 {Opt_min_batch_time, "min_batch_time=%u"},
1153 {Opt_max_batch_time, "max_batch_time=%u"},
1154 {Opt_journal_dev, "journal_dev=%u"},
1155 {Opt_journal_path, "journal_path=%s"},
1156 {Opt_journal_checksum, "journal_checksum"},
1157 {Opt_nojournal_checksum, "nojournal_checksum"},
1158 {Opt_journal_async_commit, "journal_async_commit"},
1159 {Opt_abort, "abort"},
1160 {Opt_data_journal, "data=journal"},
1161 {Opt_data_ordered, "data=ordered"},
1162 {Opt_data_writeback, "data=writeback"},
1163 {Opt_data_err_abort, "data_err=abort"},
1164 {Opt_data_err_ignore, "data_err=ignore"},
1165 {Opt_offusrjquota, "usrjquota="},
1166 {Opt_usrjquota, "usrjquota=%s"},
1167 {Opt_offgrpjquota, "grpjquota="},
1168 {Opt_grpjquota, "grpjquota=%s"},
1169 {Opt_jqfmt_vfsold, "jqfmt=vfsold"},
1170 {Opt_jqfmt_vfsv0, "jqfmt=vfsv0"},
1171 {Opt_jqfmt_vfsv1, "jqfmt=vfsv1"},
1172 {Opt_grpquota, "grpquota"},
1173 {Opt_noquota, "noquota"},
1174 {Opt_quota, "quota"},
1175 {Opt_usrquota, "usrquota"},
1176 {Opt_barrier, "barrier=%u"},
1177 {Opt_barrier, "barrier"},
1178 {Opt_nobarrier, "nobarrier"},
1179 {Opt_i_version, "i_version"},
1181 {Opt_stripe, "stripe=%u"},
1182 {Opt_delalloc, "delalloc"},
1183 {Opt_lazytime, "lazytime"},
1184 {Opt_nolazytime, "nolazytime"},
1185 {Opt_nodelalloc, "nodelalloc"},
1186 {Opt_removed, "mblk_io_submit"},
1187 {Opt_removed, "nomblk_io_submit"},
1188 {Opt_block_validity, "block_validity"},
1189 {Opt_noblock_validity, "noblock_validity"},
1190 {Opt_inode_readahead_blks, "inode_readahead_blks=%u"},
1191 {Opt_journal_ioprio, "journal_ioprio=%u"},
1192 {Opt_auto_da_alloc, "auto_da_alloc=%u"},
1193 {Opt_auto_da_alloc, "auto_da_alloc"},
1194 {Opt_noauto_da_alloc, "noauto_da_alloc"},
1195 {Opt_dioread_nolock, "dioread_nolock"},
1196 {Opt_dioread_lock, "dioread_lock"},
1197 {Opt_discard, "discard"},
1198 {Opt_nodiscard, "nodiscard"},
1199 {Opt_init_itable, "init_itable=%u"},
1200 {Opt_init_itable, "init_itable"},
1201 {Opt_noinit_itable, "noinit_itable"},
1202 {Opt_max_dir_size_kb, "max_dir_size_kb=%u"},
1203 {Opt_test_dummy_encryption, "test_dummy_encryption"},
1204 {Opt_removed, "check=none"}, /* mount option from ext2/3 */
1205 {Opt_removed, "nocheck"}, /* mount option from ext2/3 */
1206 {Opt_removed, "reservation"}, /* mount option from ext2/3 */
1207 {Opt_removed, "noreservation"}, /* mount option from ext2/3 */
1208 {Opt_removed, "journal=%u"}, /* mount option from ext2/3 */
1212 static ext4_fsblk_t get_sb_block(void **data)
1214 ext4_fsblk_t sb_block;
1215 char *options = (char *) *data;
1217 if (!options || strncmp(options, "sb=", 3) != 0)
1218 return 1; /* Default location */
1221 /* TODO: use simple_strtoll with >32bit ext4 */
1222 sb_block = simple_strtoul(options, &options, 0);
1223 if (*options && *options != ',') {
1224 printk(KERN_ERR "EXT4-fs: Invalid sb specification: %s\n",
1228 if (*options == ',')
1230 *data = (void *) options;
1235 #define DEFAULT_JOURNAL_IOPRIO (IOPRIO_PRIO_VALUE(IOPRIO_CLASS_BE, 3))
1236 static char deprecated_msg[] = "Mount option \"%s\" will be removed by %s\n"
1237 "Contact linux-ext4@vger.kernel.org if you think we should keep it.\n";
1240 static int set_qf_name(struct super_block *sb, int qtype, substring_t *args)
1242 struct ext4_sb_info *sbi = EXT4_SB(sb);
1246 if (sb_any_quota_loaded(sb) &&
1247 !sbi->s_qf_names[qtype]) {
1248 ext4_msg(sb, KERN_ERR,
1249 "Cannot change journaled "
1250 "quota options when quota turned on");
1253 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_QUOTA)) {
1254 ext4_msg(sb, KERN_ERR, "Cannot set journaled quota options "
1255 "when QUOTA feature is enabled");
1258 qname = match_strdup(args);
1260 ext4_msg(sb, KERN_ERR,
1261 "Not enough memory for storing quotafile name");
1264 if (sbi->s_qf_names[qtype]) {
1265 if (strcmp(sbi->s_qf_names[qtype], qname) == 0)
1268 ext4_msg(sb, KERN_ERR,
1269 "%s quota file already specified",
1273 if (strchr(qname, '/')) {
1274 ext4_msg(sb, KERN_ERR,
1275 "quotafile must be on filesystem root");
1278 sbi->s_qf_names[qtype] = qname;
1286 static int clear_qf_name(struct super_block *sb, int qtype)
1289 struct ext4_sb_info *sbi = EXT4_SB(sb);
1291 if (sb_any_quota_loaded(sb) &&
1292 sbi->s_qf_names[qtype]) {
1293 ext4_msg(sb, KERN_ERR, "Cannot change journaled quota options"
1294 " when quota turned on");
1297 kfree(sbi->s_qf_names[qtype]);
1298 sbi->s_qf_names[qtype] = NULL;
1303 #define MOPT_SET 0x0001
1304 #define MOPT_CLEAR 0x0002
1305 #define MOPT_NOSUPPORT 0x0004
1306 #define MOPT_EXPLICIT 0x0008
1307 #define MOPT_CLEAR_ERR 0x0010
1308 #define MOPT_GTE0 0x0020
1311 #define MOPT_QFMT 0x0040
1313 #define MOPT_Q MOPT_NOSUPPORT
1314 #define MOPT_QFMT MOPT_NOSUPPORT
1316 #define MOPT_DATAJ 0x0080
1317 #define MOPT_NO_EXT2 0x0100
1318 #define MOPT_NO_EXT3 0x0200
1319 #define MOPT_EXT4_ONLY (MOPT_NO_EXT2 | MOPT_NO_EXT3)
1320 #define MOPT_STRING 0x0400
1322 static const struct mount_opts {
1326 } ext4_mount_opts[] = {
1327 {Opt_minix_df, EXT4_MOUNT_MINIX_DF, MOPT_SET},
1328 {Opt_bsd_df, EXT4_MOUNT_MINIX_DF, MOPT_CLEAR},
1329 {Opt_grpid, EXT4_MOUNT_GRPID, MOPT_SET},
1330 {Opt_nogrpid, EXT4_MOUNT_GRPID, MOPT_CLEAR},
1331 {Opt_block_validity, EXT4_MOUNT_BLOCK_VALIDITY, MOPT_SET},
1332 {Opt_noblock_validity, EXT4_MOUNT_BLOCK_VALIDITY, MOPT_CLEAR},
1333 {Opt_dioread_nolock, EXT4_MOUNT_DIOREAD_NOLOCK,
1334 MOPT_EXT4_ONLY | MOPT_SET},
1335 {Opt_dioread_lock, EXT4_MOUNT_DIOREAD_NOLOCK,
1336 MOPT_EXT4_ONLY | MOPT_CLEAR},
1337 {Opt_discard, EXT4_MOUNT_DISCARD, MOPT_SET},
1338 {Opt_nodiscard, EXT4_MOUNT_DISCARD, MOPT_CLEAR},
1339 {Opt_delalloc, EXT4_MOUNT_DELALLOC,
1340 MOPT_EXT4_ONLY | MOPT_SET | MOPT_EXPLICIT},
1341 {Opt_nodelalloc, EXT4_MOUNT_DELALLOC,
1342 MOPT_EXT4_ONLY | MOPT_CLEAR},
1343 {Opt_nojournal_checksum, EXT4_MOUNT_JOURNAL_CHECKSUM,
1344 MOPT_EXT4_ONLY | MOPT_CLEAR},
1345 {Opt_journal_checksum, EXT4_MOUNT_JOURNAL_CHECKSUM,
1346 MOPT_EXT4_ONLY | MOPT_SET},
1347 {Opt_journal_async_commit, (EXT4_MOUNT_JOURNAL_ASYNC_COMMIT |
1348 EXT4_MOUNT_JOURNAL_CHECKSUM),
1349 MOPT_EXT4_ONLY | MOPT_SET},
1350 {Opt_noload, EXT4_MOUNT_NOLOAD, MOPT_NO_EXT2 | MOPT_SET},
1351 {Opt_err_panic, EXT4_MOUNT_ERRORS_PANIC, MOPT_SET | MOPT_CLEAR_ERR},
1352 {Opt_err_ro, EXT4_MOUNT_ERRORS_RO, MOPT_SET | MOPT_CLEAR_ERR},
1353 {Opt_err_cont, EXT4_MOUNT_ERRORS_CONT, MOPT_SET | MOPT_CLEAR_ERR},
1354 {Opt_data_err_abort, EXT4_MOUNT_DATA_ERR_ABORT,
1355 MOPT_NO_EXT2 | MOPT_SET},
1356 {Opt_data_err_ignore, EXT4_MOUNT_DATA_ERR_ABORT,
1357 MOPT_NO_EXT2 | MOPT_CLEAR},
1358 {Opt_barrier, EXT4_MOUNT_BARRIER, MOPT_SET},
1359 {Opt_nobarrier, EXT4_MOUNT_BARRIER, MOPT_CLEAR},
1360 {Opt_noauto_da_alloc, EXT4_MOUNT_NO_AUTO_DA_ALLOC, MOPT_SET},
1361 {Opt_auto_da_alloc, EXT4_MOUNT_NO_AUTO_DA_ALLOC, MOPT_CLEAR},
1362 {Opt_noinit_itable, EXT4_MOUNT_INIT_INODE_TABLE, MOPT_CLEAR},
1363 {Opt_commit, 0, MOPT_GTE0},
1364 {Opt_max_batch_time, 0, MOPT_GTE0},
1365 {Opt_min_batch_time, 0, MOPT_GTE0},
1366 {Opt_inode_readahead_blks, 0, MOPT_GTE0},
1367 {Opt_init_itable, 0, MOPT_GTE0},
1368 {Opt_dax, EXT4_MOUNT_DAX, MOPT_SET},
1369 {Opt_stripe, 0, MOPT_GTE0},
1370 {Opt_resuid, 0, MOPT_GTE0},
1371 {Opt_resgid, 0, MOPT_GTE0},
1372 {Opt_journal_dev, 0, MOPT_GTE0},
1373 {Opt_journal_path, 0, MOPT_STRING},
1374 {Opt_journal_ioprio, 0, MOPT_GTE0},
1375 {Opt_data_journal, EXT4_MOUNT_JOURNAL_DATA, MOPT_NO_EXT2 | MOPT_DATAJ},
1376 {Opt_data_ordered, EXT4_MOUNT_ORDERED_DATA, MOPT_NO_EXT2 | MOPT_DATAJ},
1377 {Opt_data_writeback, EXT4_MOUNT_WRITEBACK_DATA,
1378 MOPT_NO_EXT2 | MOPT_DATAJ},
1379 {Opt_user_xattr, EXT4_MOUNT_XATTR_USER, MOPT_SET},
1380 {Opt_nouser_xattr, EXT4_MOUNT_XATTR_USER, MOPT_CLEAR},
1381 #ifdef CONFIG_EXT4_FS_POSIX_ACL
1382 {Opt_acl, EXT4_MOUNT_POSIX_ACL, MOPT_SET},
1383 {Opt_noacl, EXT4_MOUNT_POSIX_ACL, MOPT_CLEAR},
1385 {Opt_acl, 0, MOPT_NOSUPPORT},
1386 {Opt_noacl, 0, MOPT_NOSUPPORT},
1388 {Opt_nouid32, EXT4_MOUNT_NO_UID32, MOPT_SET},
1389 {Opt_debug, EXT4_MOUNT_DEBUG, MOPT_SET},
1390 {Opt_quota, EXT4_MOUNT_QUOTA | EXT4_MOUNT_USRQUOTA, MOPT_SET | MOPT_Q},
1391 {Opt_usrquota, EXT4_MOUNT_QUOTA | EXT4_MOUNT_USRQUOTA,
1393 {Opt_grpquota, EXT4_MOUNT_QUOTA | EXT4_MOUNT_GRPQUOTA,
1395 {Opt_noquota, (EXT4_MOUNT_QUOTA | EXT4_MOUNT_USRQUOTA |
1396 EXT4_MOUNT_GRPQUOTA), MOPT_CLEAR | MOPT_Q},
1397 {Opt_usrjquota, 0, MOPT_Q},
1398 {Opt_grpjquota, 0, MOPT_Q},
1399 {Opt_offusrjquota, 0, MOPT_Q},
1400 {Opt_offgrpjquota, 0, MOPT_Q},
1401 {Opt_jqfmt_vfsold, QFMT_VFS_OLD, MOPT_QFMT},
1402 {Opt_jqfmt_vfsv0, QFMT_VFS_V0, MOPT_QFMT},
1403 {Opt_jqfmt_vfsv1, QFMT_VFS_V1, MOPT_QFMT},
1404 {Opt_max_dir_size_kb, 0, MOPT_GTE0},
1405 {Opt_test_dummy_encryption, 0, MOPT_GTE0},
1409 static int handle_mount_opt(struct super_block *sb, char *opt, int token,
1410 substring_t *args, unsigned long *journal_devnum,
1411 unsigned int *journal_ioprio, int is_remount)
1413 struct ext4_sb_info *sbi = EXT4_SB(sb);
1414 const struct mount_opts *m;
1420 if (token == Opt_usrjquota)
1421 return set_qf_name(sb, USRQUOTA, &args[0]);
1422 else if (token == Opt_grpjquota)
1423 return set_qf_name(sb, GRPQUOTA, &args[0]);
1424 else if (token == Opt_offusrjquota)
1425 return clear_qf_name(sb, USRQUOTA);
1426 else if (token == Opt_offgrpjquota)
1427 return clear_qf_name(sb, GRPQUOTA);
1431 case Opt_nouser_xattr:
1432 ext4_msg(sb, KERN_WARNING, deprecated_msg, opt, "3.5");
1435 return 1; /* handled by get_sb_block() */
1437 ext4_msg(sb, KERN_WARNING, "Ignoring removed %s option", opt);
1440 sbi->s_mount_flags |= EXT4_MF_FS_ABORTED;
1443 sb->s_flags |= MS_I_VERSION;
1446 sb->s_flags |= MS_LAZYTIME;
1448 case Opt_nolazytime:
1449 sb->s_flags &= ~MS_LAZYTIME;
1453 for (m = ext4_mount_opts; m->token != Opt_err; m++)
1454 if (token == m->token)
1457 if (m->token == Opt_err) {
1458 ext4_msg(sb, KERN_ERR, "Unrecognized mount option \"%s\" "
1459 "or missing value", opt);
1463 if ((m->flags & MOPT_NO_EXT2) && IS_EXT2_SB(sb)) {
1464 ext4_msg(sb, KERN_ERR,
1465 "Mount option \"%s\" incompatible with ext2", opt);
1468 if ((m->flags & MOPT_NO_EXT3) && IS_EXT3_SB(sb)) {
1469 ext4_msg(sb, KERN_ERR,
1470 "Mount option \"%s\" incompatible with ext3", opt);
1474 if (args->from && !(m->flags & MOPT_STRING) && match_int(args, &arg))
1476 if (args->from && (m->flags & MOPT_GTE0) && (arg < 0))
1478 if (m->flags & MOPT_EXPLICIT)
1479 set_opt2(sb, EXPLICIT_DELALLOC);
1480 if (m->flags & MOPT_CLEAR_ERR)
1481 clear_opt(sb, ERRORS_MASK);
1482 if (token == Opt_noquota && sb_any_quota_loaded(sb)) {
1483 ext4_msg(sb, KERN_ERR, "Cannot change quota "
1484 "options when quota turned on");
1488 if (m->flags & MOPT_NOSUPPORT) {
1489 ext4_msg(sb, KERN_ERR, "%s option not supported", opt);
1490 } else if (token == Opt_commit) {
1492 arg = JBD2_DEFAULT_MAX_COMMIT_AGE;
1493 sbi->s_commit_interval = HZ * arg;
1494 } else if (token == Opt_max_batch_time) {
1495 sbi->s_max_batch_time = arg;
1496 } else if (token == Opt_min_batch_time) {
1497 sbi->s_min_batch_time = arg;
1498 } else if (token == Opt_inode_readahead_blks) {
1499 if (arg && (arg > (1 << 30) || !is_power_of_2(arg))) {
1500 ext4_msg(sb, KERN_ERR,
1501 "EXT4-fs: inode_readahead_blks must be "
1502 "0 or a power of 2 smaller than 2^31");
1505 sbi->s_inode_readahead_blks = arg;
1506 } else if (token == Opt_init_itable) {
1507 set_opt(sb, INIT_INODE_TABLE);
1509 arg = EXT4_DEF_LI_WAIT_MULT;
1510 sbi->s_li_wait_mult = arg;
1511 } else if (token == Opt_max_dir_size_kb) {
1512 sbi->s_max_dir_size_kb = arg;
1513 } else if (token == Opt_stripe) {
1514 sbi->s_stripe = arg;
1515 } else if (token == Opt_resuid) {
1516 uid = make_kuid(current_user_ns(), arg);
1517 if (!uid_valid(uid)) {
1518 ext4_msg(sb, KERN_ERR, "Invalid uid value %d", arg);
1521 sbi->s_resuid = uid;
1522 } else if (token == Opt_resgid) {
1523 gid = make_kgid(current_user_ns(), arg);
1524 if (!gid_valid(gid)) {
1525 ext4_msg(sb, KERN_ERR, "Invalid gid value %d", arg);
1528 sbi->s_resgid = gid;
1529 } else if (token == Opt_journal_dev) {
1531 ext4_msg(sb, KERN_ERR,
1532 "Cannot specify journal on remount");
1535 *journal_devnum = arg;
1536 } else if (token == Opt_journal_path) {
1538 struct inode *journal_inode;
1543 ext4_msg(sb, KERN_ERR,
1544 "Cannot specify journal on remount");
1547 journal_path = match_strdup(&args[0]);
1548 if (!journal_path) {
1549 ext4_msg(sb, KERN_ERR, "error: could not dup "
1550 "journal device string");
1554 error = kern_path(journal_path, LOOKUP_FOLLOW, &path);
1556 ext4_msg(sb, KERN_ERR, "error: could not find "
1557 "journal device path: error %d", error);
1558 kfree(journal_path);
1562 journal_inode = d_inode(path.dentry);
1563 if (!S_ISBLK(journal_inode->i_mode)) {
1564 ext4_msg(sb, KERN_ERR, "error: journal path %s "
1565 "is not a block device", journal_path);
1567 kfree(journal_path);
1571 *journal_devnum = new_encode_dev(journal_inode->i_rdev);
1573 kfree(journal_path);
1574 } else if (token == Opt_journal_ioprio) {
1576 ext4_msg(sb, KERN_ERR, "Invalid journal IO priority"
1581 IOPRIO_PRIO_VALUE(IOPRIO_CLASS_BE, arg);
1582 } else if (token == Opt_test_dummy_encryption) {
1583 #ifdef CONFIG_EXT4_FS_ENCRYPTION
1584 sbi->s_mount_flags |= EXT4_MF_TEST_DUMMY_ENCRYPTION;
1585 ext4_msg(sb, KERN_WARNING,
1586 "Test dummy encryption mode enabled");
1588 ext4_msg(sb, KERN_WARNING,
1589 "Test dummy encryption mount option ignored");
1591 } else if (m->flags & MOPT_DATAJ) {
1593 if (!sbi->s_journal)
1594 ext4_msg(sb, KERN_WARNING, "Remounting file system with no journal so ignoring journalled data option");
1595 else if (test_opt(sb, DATA_FLAGS) != m->mount_opt) {
1596 ext4_msg(sb, KERN_ERR,
1597 "Cannot change data mode on remount");
1601 clear_opt(sb, DATA_FLAGS);
1602 sbi->s_mount_opt |= m->mount_opt;
1605 } else if (m->flags & MOPT_QFMT) {
1606 if (sb_any_quota_loaded(sb) &&
1607 sbi->s_jquota_fmt != m->mount_opt) {
1608 ext4_msg(sb, KERN_ERR, "Cannot change journaled "
1609 "quota options when quota turned on");
1612 if (EXT4_HAS_RO_COMPAT_FEATURE(sb,
1613 EXT4_FEATURE_RO_COMPAT_QUOTA)) {
1614 ext4_msg(sb, KERN_ERR,
1615 "Cannot set journaled quota options "
1616 "when QUOTA feature is enabled");
1619 sbi->s_jquota_fmt = m->mount_opt;
1621 #ifndef CONFIG_FS_DAX
1622 } else if (token == Opt_dax) {
1623 ext4_msg(sb, KERN_INFO, "dax option not supported");
1629 if (m->flags & MOPT_CLEAR)
1631 else if (unlikely(!(m->flags & MOPT_SET))) {
1632 ext4_msg(sb, KERN_WARNING,
1633 "buggy handling of option %s", opt);
1638 sbi->s_mount_opt |= m->mount_opt;
1640 sbi->s_mount_opt &= ~m->mount_opt;
1645 static int parse_options(char *options, struct super_block *sb,
1646 unsigned long *journal_devnum,
1647 unsigned int *journal_ioprio,
1650 struct ext4_sb_info *sbi = EXT4_SB(sb);
1652 substring_t args[MAX_OPT_ARGS];
1658 while ((p = strsep(&options, ",")) != NULL) {
1662 * Initialize args struct so we know whether arg was
1663 * found; some options take optional arguments.
1665 args[0].to = args[0].from = NULL;
1666 token = match_token(p, tokens, args);
1667 if (handle_mount_opt(sb, p, token, args, journal_devnum,
1668 journal_ioprio, is_remount) < 0)
1672 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_QUOTA) &&
1673 (test_opt(sb, USRQUOTA) || test_opt(sb, GRPQUOTA))) {
1674 ext4_msg(sb, KERN_ERR, "Cannot set quota options when QUOTA "
1675 "feature is enabled");
1678 if (sbi->s_qf_names[USRQUOTA] || sbi->s_qf_names[GRPQUOTA]) {
1679 if (test_opt(sb, USRQUOTA) && sbi->s_qf_names[USRQUOTA])
1680 clear_opt(sb, USRQUOTA);
1682 if (test_opt(sb, GRPQUOTA) && sbi->s_qf_names[GRPQUOTA])
1683 clear_opt(sb, GRPQUOTA);
1685 if (test_opt(sb, GRPQUOTA) || test_opt(sb, USRQUOTA)) {
1686 ext4_msg(sb, KERN_ERR, "old and new quota "
1691 if (!sbi->s_jquota_fmt) {
1692 ext4_msg(sb, KERN_ERR, "journaled quota format "
1698 if (test_opt(sb, DIOREAD_NOLOCK)) {
1700 BLOCK_SIZE << le32_to_cpu(sbi->s_es->s_log_block_size);
1702 if (blocksize < PAGE_CACHE_SIZE) {
1703 ext4_msg(sb, KERN_ERR, "can't mount with "
1704 "dioread_nolock if block size != PAGE_SIZE");
1708 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA &&
1709 test_opt(sb, JOURNAL_ASYNC_COMMIT)) {
1710 ext4_msg(sb, KERN_ERR, "can't mount with journal_async_commit "
1711 "in data=ordered mode");
1717 static inline void ext4_show_quota_options(struct seq_file *seq,
1718 struct super_block *sb)
1720 #if defined(CONFIG_QUOTA)
1721 struct ext4_sb_info *sbi = EXT4_SB(sb);
1723 if (sbi->s_jquota_fmt) {
1726 switch (sbi->s_jquota_fmt) {
1737 seq_printf(seq, ",jqfmt=%s", fmtname);
1740 if (sbi->s_qf_names[USRQUOTA])
1741 seq_printf(seq, ",usrjquota=%s", sbi->s_qf_names[USRQUOTA]);
1743 if (sbi->s_qf_names[GRPQUOTA])
1744 seq_printf(seq, ",grpjquota=%s", sbi->s_qf_names[GRPQUOTA]);
1748 static const char *token2str(int token)
1750 const struct match_token *t;
1752 for (t = tokens; t->token != Opt_err; t++)
1753 if (t->token == token && !strchr(t->pattern, '='))
1760 * - it's set to a non-default value OR
1761 * - if the per-sb default is different from the global default
1763 static int _ext4_show_options(struct seq_file *seq, struct super_block *sb,
1766 struct ext4_sb_info *sbi = EXT4_SB(sb);
1767 struct ext4_super_block *es = sbi->s_es;
1768 int def_errors, def_mount_opt = nodefs ? 0 : sbi->s_def_mount_opt;
1769 const struct mount_opts *m;
1770 char sep = nodefs ? '\n' : ',';
1772 #define SEQ_OPTS_PUTS(str) seq_printf(seq, "%c" str, sep)
1773 #define SEQ_OPTS_PRINT(str, arg) seq_printf(seq, "%c" str, sep, arg)
1775 if (sbi->s_sb_block != 1)
1776 SEQ_OPTS_PRINT("sb=%llu", sbi->s_sb_block);
1778 for (m = ext4_mount_opts; m->token != Opt_err; m++) {
1779 int want_set = m->flags & MOPT_SET;
1780 if (((m->flags & (MOPT_SET|MOPT_CLEAR)) == 0) ||
1781 (m->flags & MOPT_CLEAR_ERR))
1783 if (!(m->mount_opt & (sbi->s_mount_opt ^ def_mount_opt)))
1784 continue; /* skip if same as the default */
1786 (sbi->s_mount_opt & m->mount_opt) != m->mount_opt) ||
1787 (!want_set && (sbi->s_mount_opt & m->mount_opt)))
1788 continue; /* select Opt_noFoo vs Opt_Foo */
1789 SEQ_OPTS_PRINT("%s", token2str(m->token));
1792 if (nodefs || !uid_eq(sbi->s_resuid, make_kuid(&init_user_ns, EXT4_DEF_RESUID)) ||
1793 le16_to_cpu(es->s_def_resuid) != EXT4_DEF_RESUID)
1794 SEQ_OPTS_PRINT("resuid=%u",
1795 from_kuid_munged(&init_user_ns, sbi->s_resuid));
1796 if (nodefs || !gid_eq(sbi->s_resgid, make_kgid(&init_user_ns, EXT4_DEF_RESGID)) ||
1797 le16_to_cpu(es->s_def_resgid) != EXT4_DEF_RESGID)
1798 SEQ_OPTS_PRINT("resgid=%u",
1799 from_kgid_munged(&init_user_ns, sbi->s_resgid));
1800 def_errors = nodefs ? -1 : le16_to_cpu(es->s_errors);
1801 if (test_opt(sb, ERRORS_RO) && def_errors != EXT4_ERRORS_RO)
1802 SEQ_OPTS_PUTS("errors=remount-ro");
1803 if (test_opt(sb, ERRORS_CONT) && def_errors != EXT4_ERRORS_CONTINUE)
1804 SEQ_OPTS_PUTS("errors=continue");
1805 if (test_opt(sb, ERRORS_PANIC) && def_errors != EXT4_ERRORS_PANIC)
1806 SEQ_OPTS_PUTS("errors=panic");
1807 if (nodefs || sbi->s_commit_interval != JBD2_DEFAULT_MAX_COMMIT_AGE*HZ)
1808 SEQ_OPTS_PRINT("commit=%lu", sbi->s_commit_interval / HZ);
1809 if (nodefs || sbi->s_min_batch_time != EXT4_DEF_MIN_BATCH_TIME)
1810 SEQ_OPTS_PRINT("min_batch_time=%u", sbi->s_min_batch_time);
1811 if (nodefs || sbi->s_max_batch_time != EXT4_DEF_MAX_BATCH_TIME)
1812 SEQ_OPTS_PRINT("max_batch_time=%u", sbi->s_max_batch_time);
1813 if (sb->s_flags & MS_I_VERSION)
1814 SEQ_OPTS_PUTS("i_version");
1815 if (nodefs || sbi->s_stripe)
1816 SEQ_OPTS_PRINT("stripe=%lu", sbi->s_stripe);
1817 if (EXT4_MOUNT_DATA_FLAGS & (sbi->s_mount_opt ^ def_mount_opt)) {
1818 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA)
1819 SEQ_OPTS_PUTS("data=journal");
1820 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA)
1821 SEQ_OPTS_PUTS("data=ordered");
1822 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_WRITEBACK_DATA)
1823 SEQ_OPTS_PUTS("data=writeback");
1826 sbi->s_inode_readahead_blks != EXT4_DEF_INODE_READAHEAD_BLKS)
1827 SEQ_OPTS_PRINT("inode_readahead_blks=%u",
1828 sbi->s_inode_readahead_blks);
1830 if (nodefs || (test_opt(sb, INIT_INODE_TABLE) &&
1831 (sbi->s_li_wait_mult != EXT4_DEF_LI_WAIT_MULT)))
1832 SEQ_OPTS_PRINT("init_itable=%u", sbi->s_li_wait_mult);
1833 if (nodefs || sbi->s_max_dir_size_kb)
1834 SEQ_OPTS_PRINT("max_dir_size_kb=%u", sbi->s_max_dir_size_kb);
1836 ext4_show_quota_options(seq, sb);
1840 static int ext4_show_options(struct seq_file *seq, struct dentry *root)
1842 return _ext4_show_options(seq, root->d_sb, 0);
1845 static int options_seq_show(struct seq_file *seq, void *offset)
1847 struct super_block *sb = seq->private;
1850 seq_puts(seq, (sb->s_flags & MS_RDONLY) ? "ro" : "rw");
1851 rc = _ext4_show_options(seq, sb, 1);
1852 seq_puts(seq, "\n");
1856 static int options_open_fs(struct inode *inode, struct file *file)
1858 return single_open(file, options_seq_show, PDE_DATA(inode));
1861 static const struct file_operations ext4_seq_options_fops = {
1862 .owner = THIS_MODULE,
1863 .open = options_open_fs,
1865 .llseek = seq_lseek,
1866 .release = single_release,
1869 static int ext4_setup_super(struct super_block *sb, struct ext4_super_block *es,
1872 struct ext4_sb_info *sbi = EXT4_SB(sb);
1875 if (le32_to_cpu(es->s_rev_level) > EXT4_MAX_SUPP_REV) {
1876 ext4_msg(sb, KERN_ERR, "revision level too high, "
1877 "forcing read-only mode");
1882 if (!(sbi->s_mount_state & EXT4_VALID_FS))
1883 ext4_msg(sb, KERN_WARNING, "warning: mounting unchecked fs, "
1884 "running e2fsck is recommended");
1885 else if (sbi->s_mount_state & EXT4_ERROR_FS)
1886 ext4_msg(sb, KERN_WARNING,
1887 "warning: mounting fs with errors, "
1888 "running e2fsck is recommended");
1889 else if ((__s16) le16_to_cpu(es->s_max_mnt_count) > 0 &&
1890 le16_to_cpu(es->s_mnt_count) >=
1891 (unsigned short) (__s16) le16_to_cpu(es->s_max_mnt_count))
1892 ext4_msg(sb, KERN_WARNING,
1893 "warning: maximal mount count reached, "
1894 "running e2fsck is recommended");
1895 else if (le32_to_cpu(es->s_checkinterval) &&
1896 (le32_to_cpu(es->s_lastcheck) +
1897 le32_to_cpu(es->s_checkinterval) <= get_seconds()))
1898 ext4_msg(sb, KERN_WARNING,
1899 "warning: checktime reached, "
1900 "running e2fsck is recommended");
1901 if (!sbi->s_journal)
1902 es->s_state &= cpu_to_le16(~EXT4_VALID_FS);
1903 if (!(__s16) le16_to_cpu(es->s_max_mnt_count))
1904 es->s_max_mnt_count = cpu_to_le16(EXT4_DFL_MAX_MNT_COUNT);
1905 le16_add_cpu(&es->s_mnt_count, 1);
1906 es->s_mtime = cpu_to_le32(get_seconds());
1907 ext4_update_dynamic_rev(sb);
1909 EXT4_SET_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
1911 ext4_commit_super(sb, 1);
1913 if (test_opt(sb, DEBUG))
1914 printk(KERN_INFO "[EXT4 FS bs=%lu, gc=%u, "
1915 "bpg=%lu, ipg=%lu, mo=%04x, mo2=%04x]\n",
1917 sbi->s_groups_count,
1918 EXT4_BLOCKS_PER_GROUP(sb),
1919 EXT4_INODES_PER_GROUP(sb),
1920 sbi->s_mount_opt, sbi->s_mount_opt2);
1922 cleancache_init_fs(sb);
1926 int ext4_alloc_flex_bg_array(struct super_block *sb, ext4_group_t ngroup)
1928 struct ext4_sb_info *sbi = EXT4_SB(sb);
1929 struct flex_groups *new_groups;
1932 if (!sbi->s_log_groups_per_flex)
1935 size = ext4_flex_group(sbi, ngroup - 1) + 1;
1936 if (size <= sbi->s_flex_groups_allocated)
1939 size = roundup_pow_of_two(size * sizeof(struct flex_groups));
1940 new_groups = ext4_kvzalloc(size, GFP_KERNEL);
1942 ext4_msg(sb, KERN_ERR, "not enough memory for %d flex groups",
1943 size / (int) sizeof(struct flex_groups));
1947 if (sbi->s_flex_groups) {
1948 memcpy(new_groups, sbi->s_flex_groups,
1949 (sbi->s_flex_groups_allocated *
1950 sizeof(struct flex_groups)));
1951 kvfree(sbi->s_flex_groups);
1953 sbi->s_flex_groups = new_groups;
1954 sbi->s_flex_groups_allocated = size / sizeof(struct flex_groups);
1958 static int ext4_fill_flex_info(struct super_block *sb)
1960 struct ext4_sb_info *sbi = EXT4_SB(sb);
1961 struct ext4_group_desc *gdp = NULL;
1962 ext4_group_t flex_group;
1965 sbi->s_log_groups_per_flex = sbi->s_es->s_log_groups_per_flex;
1966 if (sbi->s_log_groups_per_flex < 1 || sbi->s_log_groups_per_flex > 31) {
1967 sbi->s_log_groups_per_flex = 0;
1971 err = ext4_alloc_flex_bg_array(sb, sbi->s_groups_count);
1975 for (i = 0; i < sbi->s_groups_count; i++) {
1976 gdp = ext4_get_group_desc(sb, i, NULL);
1978 flex_group = ext4_flex_group(sbi, i);
1979 atomic_add(ext4_free_inodes_count(sb, gdp),
1980 &sbi->s_flex_groups[flex_group].free_inodes);
1981 atomic64_add(ext4_free_group_clusters(sb, gdp),
1982 &sbi->s_flex_groups[flex_group].free_clusters);
1983 atomic_add(ext4_used_dirs_count(sb, gdp),
1984 &sbi->s_flex_groups[flex_group].used_dirs);
1992 static __le16 ext4_group_desc_csum(struct ext4_sb_info *sbi, __u32 block_group,
1993 struct ext4_group_desc *gdp)
1997 __le32 le_group = cpu_to_le32(block_group);
1999 if (ext4_has_metadata_csum(sbi->s_sb)) {
2000 /* Use new metadata_csum algorithm */
2004 save_csum = gdp->bg_checksum;
2005 gdp->bg_checksum = 0;
2006 csum32 = ext4_chksum(sbi, sbi->s_csum_seed, (__u8 *)&le_group,
2008 csum32 = ext4_chksum(sbi, csum32, (__u8 *)gdp,
2010 gdp->bg_checksum = save_csum;
2012 crc = csum32 & 0xFFFF;
2016 /* old crc16 code */
2017 if (!(sbi->s_es->s_feature_ro_compat &
2018 cpu_to_le32(EXT4_FEATURE_RO_COMPAT_GDT_CSUM)))
2021 offset = offsetof(struct ext4_group_desc, bg_checksum);
2023 crc = crc16(~0, sbi->s_es->s_uuid, sizeof(sbi->s_es->s_uuid));
2024 crc = crc16(crc, (__u8 *)&le_group, sizeof(le_group));
2025 crc = crc16(crc, (__u8 *)gdp, offset);
2026 offset += sizeof(gdp->bg_checksum); /* skip checksum */
2027 /* for checksum of struct ext4_group_desc do the rest...*/
2028 if ((sbi->s_es->s_feature_incompat &
2029 cpu_to_le32(EXT4_FEATURE_INCOMPAT_64BIT)) &&
2030 offset < le16_to_cpu(sbi->s_es->s_desc_size))
2031 crc = crc16(crc, (__u8 *)gdp + offset,
2032 le16_to_cpu(sbi->s_es->s_desc_size) -
2036 return cpu_to_le16(crc);
2039 int ext4_group_desc_csum_verify(struct super_block *sb, __u32 block_group,
2040 struct ext4_group_desc *gdp)
2042 if (ext4_has_group_desc_csum(sb) &&
2043 (gdp->bg_checksum != ext4_group_desc_csum(EXT4_SB(sb),
2050 void ext4_group_desc_csum_set(struct super_block *sb, __u32 block_group,
2051 struct ext4_group_desc *gdp)
2053 if (!ext4_has_group_desc_csum(sb))
2055 gdp->bg_checksum = ext4_group_desc_csum(EXT4_SB(sb), block_group, gdp);
2058 /* Called at mount-time, super-block is locked */
2059 static int ext4_check_descriptors(struct super_block *sb,
2060 ext4_group_t *first_not_zeroed)
2062 struct ext4_sb_info *sbi = EXT4_SB(sb);
2063 ext4_fsblk_t first_block = le32_to_cpu(sbi->s_es->s_first_data_block);
2064 ext4_fsblk_t last_block;
2065 ext4_fsblk_t block_bitmap;
2066 ext4_fsblk_t inode_bitmap;
2067 ext4_fsblk_t inode_table;
2068 int flexbg_flag = 0;
2069 ext4_group_t i, grp = sbi->s_groups_count;
2071 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_FLEX_BG))
2074 ext4_debug("Checking group descriptors");
2076 for (i = 0; i < sbi->s_groups_count; i++) {
2077 struct ext4_group_desc *gdp = ext4_get_group_desc(sb, i, NULL);
2079 if (i == sbi->s_groups_count - 1 || flexbg_flag)
2080 last_block = ext4_blocks_count(sbi->s_es) - 1;
2082 last_block = first_block +
2083 (EXT4_BLOCKS_PER_GROUP(sb) - 1);
2085 if ((grp == sbi->s_groups_count) &&
2086 !(gdp->bg_flags & cpu_to_le16(EXT4_BG_INODE_ZEROED)))
2089 block_bitmap = ext4_block_bitmap(sb, gdp);
2090 if (block_bitmap < first_block || block_bitmap > last_block) {
2091 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2092 "Block bitmap for group %u not in group "
2093 "(block %llu)!", i, block_bitmap);
2096 inode_bitmap = ext4_inode_bitmap(sb, gdp);
2097 if (inode_bitmap < first_block || inode_bitmap > last_block) {
2098 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2099 "Inode bitmap for group %u not in group "
2100 "(block %llu)!", i, inode_bitmap);
2103 inode_table = ext4_inode_table(sb, gdp);
2104 if (inode_table < first_block ||
2105 inode_table + sbi->s_itb_per_group - 1 > last_block) {
2106 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2107 "Inode table for group %u not in group "
2108 "(block %llu)!", i, inode_table);
2111 ext4_lock_group(sb, i);
2112 if (!ext4_group_desc_csum_verify(sb, i, gdp)) {
2113 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2114 "Checksum for group %u failed (%u!=%u)",
2115 i, le16_to_cpu(ext4_group_desc_csum(sbi, i,
2116 gdp)), le16_to_cpu(gdp->bg_checksum));
2117 if (!(sb->s_flags & MS_RDONLY)) {
2118 ext4_unlock_group(sb, i);
2122 ext4_unlock_group(sb, i);
2124 first_block += EXT4_BLOCKS_PER_GROUP(sb);
2126 if (NULL != first_not_zeroed)
2127 *first_not_zeroed = grp;
2131 /* ext4_orphan_cleanup() walks a singly-linked list of inodes (starting at
2132 * the superblock) which were deleted from all directories, but held open by
2133 * a process at the time of a crash. We walk the list and try to delete these
2134 * inodes at recovery time (only with a read-write filesystem).
2136 * In order to keep the orphan inode chain consistent during traversal (in
2137 * case of crash during recovery), we link each inode into the superblock
2138 * orphan list_head and handle it the same way as an inode deletion during
2139 * normal operation (which journals the operations for us).
2141 * We only do an iget() and an iput() on each inode, which is very safe if we
2142 * accidentally point at an in-use or already deleted inode. The worst that
2143 * can happen in this case is that we get a "bit already cleared" message from
2144 * ext4_free_inode(). The only reason we would point at a wrong inode is if
2145 * e2fsck was run on this filesystem, and it must have already done the orphan
2146 * inode cleanup for us, so we can safely abort without any further action.
2148 static void ext4_orphan_cleanup(struct super_block *sb,
2149 struct ext4_super_block *es)
2151 unsigned int s_flags = sb->s_flags;
2152 int nr_orphans = 0, nr_truncates = 0;
2156 if (!es->s_last_orphan) {
2157 jbd_debug(4, "no orphan inodes to clean up\n");
2161 if (bdev_read_only(sb->s_bdev)) {
2162 ext4_msg(sb, KERN_ERR, "write access "
2163 "unavailable, skipping orphan cleanup");
2167 /* Check if feature set would not allow a r/w mount */
2168 if (!ext4_feature_set_ok(sb, 0)) {
2169 ext4_msg(sb, KERN_INFO, "Skipping orphan cleanup due to "
2170 "unknown ROCOMPAT features");
2174 if (EXT4_SB(sb)->s_mount_state & EXT4_ERROR_FS) {
2175 /* don't clear list on RO mount w/ errors */
2176 if (es->s_last_orphan && !(s_flags & MS_RDONLY)) {
2177 ext4_msg(sb, KERN_INFO, "Errors on filesystem, "
2178 "clearing orphan list.\n");
2179 es->s_last_orphan = 0;
2181 jbd_debug(1, "Skipping orphan recovery on fs with errors.\n");
2185 if (s_flags & MS_RDONLY) {
2186 ext4_msg(sb, KERN_INFO, "orphan cleanup on readonly fs");
2187 sb->s_flags &= ~MS_RDONLY;
2190 /* Needed for iput() to work correctly and not trash data */
2191 sb->s_flags |= MS_ACTIVE;
2192 /* Turn on quotas so that they are updated correctly */
2193 for (i = 0; i < EXT4_MAXQUOTAS; i++) {
2194 if (EXT4_SB(sb)->s_qf_names[i]) {
2195 int ret = ext4_quota_on_mount(sb, i);
2197 ext4_msg(sb, KERN_ERR,
2198 "Cannot turn on journaled "
2199 "quota: error %d", ret);
2204 while (es->s_last_orphan) {
2205 struct inode *inode;
2207 inode = ext4_orphan_get(sb, le32_to_cpu(es->s_last_orphan));
2208 if (IS_ERR(inode)) {
2209 es->s_last_orphan = 0;
2213 list_add(&EXT4_I(inode)->i_orphan, &EXT4_SB(sb)->s_orphan);
2214 dquot_initialize(inode);
2215 if (inode->i_nlink) {
2216 if (test_opt(sb, DEBUG))
2217 ext4_msg(sb, KERN_DEBUG,
2218 "%s: truncating inode %lu to %lld bytes",
2219 __func__, inode->i_ino, inode->i_size);
2220 jbd_debug(2, "truncating inode %lu to %lld bytes\n",
2221 inode->i_ino, inode->i_size);
2222 mutex_lock(&inode->i_mutex);
2223 truncate_inode_pages(inode->i_mapping, inode->i_size);
2224 ext4_truncate(inode);
2225 mutex_unlock(&inode->i_mutex);
2228 if (test_opt(sb, DEBUG))
2229 ext4_msg(sb, KERN_DEBUG,
2230 "%s: deleting unreferenced inode %lu",
2231 __func__, inode->i_ino);
2232 jbd_debug(2, "deleting unreferenced inode %lu\n",
2236 iput(inode); /* The delete magic happens here! */
2239 #define PLURAL(x) (x), ((x) == 1) ? "" : "s"
2242 ext4_msg(sb, KERN_INFO, "%d orphan inode%s deleted",
2243 PLURAL(nr_orphans));
2245 ext4_msg(sb, KERN_INFO, "%d truncate%s cleaned up",
2246 PLURAL(nr_truncates));
2248 /* Turn quotas off */
2249 for (i = 0; i < EXT4_MAXQUOTAS; i++) {
2250 if (sb_dqopt(sb)->files[i])
2251 dquot_quota_off(sb, i);
2254 sb->s_flags = s_flags; /* Restore MS_RDONLY status */
2258 * Maximal extent format file size.
2259 * Resulting logical blkno at s_maxbytes must fit in our on-disk
2260 * extent format containers, within a sector_t, and within i_blocks
2261 * in the vfs. ext4 inode has 48 bits of i_block in fsblock units,
2262 * so that won't be a limiting factor.
2264 * However there is other limiting factor. We do store extents in the form
2265 * of starting block and length, hence the resulting length of the extent
2266 * covering maximum file size must fit into on-disk format containers as
2267 * well. Given that length is always by 1 unit bigger than max unit (because
2268 * we count 0 as well) we have to lower the s_maxbytes by one fs block.
2270 * Note, this does *not* consider any metadata overhead for vfs i_blocks.
2272 static loff_t ext4_max_size(int blkbits, int has_huge_files)
2275 loff_t upper_limit = MAX_LFS_FILESIZE;
2277 /* small i_blocks in vfs inode? */
2278 if (!has_huge_files || sizeof(blkcnt_t) < sizeof(u64)) {
2280 * CONFIG_LBDAF is not enabled implies the inode
2281 * i_block represent total blocks in 512 bytes
2282 * 32 == size of vfs inode i_blocks * 8
2284 upper_limit = (1LL << 32) - 1;
2286 /* total blocks in file system block size */
2287 upper_limit >>= (blkbits - 9);
2288 upper_limit <<= blkbits;
2292 * 32-bit extent-start container, ee_block. We lower the maxbytes
2293 * by one fs block, so ee_len can cover the extent of maximum file
2296 res = (1LL << 32) - 1;
2299 /* Sanity check against vm- & vfs- imposed limits */
2300 if (res > upper_limit)
2307 * Maximal bitmap file size. There is a direct, and {,double-,triple-}indirect
2308 * block limit, and also a limit of (2^48 - 1) 512-byte sectors in i_blocks.
2309 * We need to be 1 filesystem block less than the 2^48 sector limit.
2311 static loff_t ext4_max_bitmap_size(int bits, int has_huge_files)
2313 loff_t res = EXT4_NDIR_BLOCKS;
2316 /* This is calculated to be the largest file size for a dense, block
2317 * mapped file such that the file's total number of 512-byte sectors,
2318 * including data and all indirect blocks, does not exceed (2^48 - 1).
2320 * __u32 i_blocks_lo and _u16 i_blocks_high represent the total
2321 * number of 512-byte sectors of the file.
2324 if (!has_huge_files || sizeof(blkcnt_t) < sizeof(u64)) {
2326 * !has_huge_files or CONFIG_LBDAF not enabled implies that
2327 * the inode i_block field represents total file blocks in
2328 * 2^32 512-byte sectors == size of vfs inode i_blocks * 8
2330 upper_limit = (1LL << 32) - 1;
2332 /* total blocks in file system block size */
2333 upper_limit >>= (bits - 9);
2337 * We use 48 bit ext4_inode i_blocks
2338 * With EXT4_HUGE_FILE_FL set the i_blocks
2339 * represent total number of blocks in
2340 * file system block size
2342 upper_limit = (1LL << 48) - 1;
2346 /* indirect blocks */
2348 /* double indirect blocks */
2349 meta_blocks += 1 + (1LL << (bits-2));
2350 /* tripple indirect blocks */
2351 meta_blocks += 1 + (1LL << (bits-2)) + (1LL << (2*(bits-2)));
2353 upper_limit -= meta_blocks;
2354 upper_limit <<= bits;
2356 res += 1LL << (bits-2);
2357 res += 1LL << (2*(bits-2));
2358 res += 1LL << (3*(bits-2));
2360 if (res > upper_limit)
2363 if (res > MAX_LFS_FILESIZE)
2364 res = MAX_LFS_FILESIZE;
2369 static ext4_fsblk_t descriptor_loc(struct super_block *sb,
2370 ext4_fsblk_t logical_sb_block, int nr)
2372 struct ext4_sb_info *sbi = EXT4_SB(sb);
2373 ext4_group_t bg, first_meta_bg;
2376 first_meta_bg = le32_to_cpu(sbi->s_es->s_first_meta_bg);
2378 if (!EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_META_BG) ||
2380 return logical_sb_block + nr + 1;
2381 bg = sbi->s_desc_per_block * nr;
2382 if (ext4_bg_has_super(sb, bg))
2386 * If we have a meta_bg fs with 1k blocks, group 0's GDT is at
2387 * block 2, not 1. If s_first_data_block == 0 (bigalloc is enabled
2388 * on modern mke2fs or blksize > 1k on older mke2fs) then we must
2391 if (sb->s_blocksize == 1024 && nr == 0 &&
2392 le32_to_cpu(EXT4_SB(sb)->s_es->s_first_data_block) == 0)
2395 return (has_super + ext4_group_first_block_no(sb, bg));
2399 * ext4_get_stripe_size: Get the stripe size.
2400 * @sbi: In memory super block info
2402 * If we have specified it via mount option, then
2403 * use the mount option value. If the value specified at mount time is
2404 * greater than the blocks per group use the super block value.
2405 * If the super block value is greater than blocks per group return 0.
2406 * Allocator needs it be less than blocks per group.
2409 static unsigned long ext4_get_stripe_size(struct ext4_sb_info *sbi)
2411 unsigned long stride = le16_to_cpu(sbi->s_es->s_raid_stride);
2412 unsigned long stripe_width =
2413 le32_to_cpu(sbi->s_es->s_raid_stripe_width);
2416 if (sbi->s_stripe && sbi->s_stripe <= sbi->s_blocks_per_group)
2417 ret = sbi->s_stripe;
2418 else if (stripe_width <= sbi->s_blocks_per_group)
2420 else if (stride <= sbi->s_blocks_per_group)
2426 * If the stripe width is 1, this makes no sense and
2427 * we set it to 0 to turn off stripe handling code.
2438 struct attribute attr;
2439 ssize_t (*show)(struct ext4_attr *, struct ext4_sb_info *, char *);
2440 ssize_t (*store)(struct ext4_attr *, struct ext4_sb_info *,
2441 const char *, size_t);
2448 static int parse_strtoull(const char *buf,
2449 unsigned long long max, unsigned long long *value)
2453 ret = kstrtoull(skip_spaces(buf), 0, value);
2454 if (!ret && *value > max)
2459 static ssize_t delayed_allocation_blocks_show(struct ext4_attr *a,
2460 struct ext4_sb_info *sbi,
2463 return snprintf(buf, PAGE_SIZE, "%llu\n",
2465 percpu_counter_sum(&sbi->s_dirtyclusters_counter)));
2468 static ssize_t session_write_kbytes_show(struct ext4_attr *a,
2469 struct ext4_sb_info *sbi, char *buf)
2471 struct super_block *sb = sbi->s_buddy_cache->i_sb;
2473 if (!sb->s_bdev->bd_part)
2474 return snprintf(buf, PAGE_SIZE, "0\n");
2475 return snprintf(buf, PAGE_SIZE, "%lu\n",
2476 (part_stat_read(sb->s_bdev->bd_part, sectors[1]) -
2477 sbi->s_sectors_written_start) >> 1);
2480 static ssize_t lifetime_write_kbytes_show(struct ext4_attr *a,
2481 struct ext4_sb_info *sbi, char *buf)
2483 struct super_block *sb = sbi->s_buddy_cache->i_sb;
2485 if (!sb->s_bdev->bd_part)
2486 return snprintf(buf, PAGE_SIZE, "0\n");
2487 return snprintf(buf, PAGE_SIZE, "%llu\n",
2488 (unsigned long long)(sbi->s_kbytes_written +
2489 ((part_stat_read(sb->s_bdev->bd_part, sectors[1]) -
2490 EXT4_SB(sb)->s_sectors_written_start) >> 1)));
2493 static ssize_t inode_readahead_blks_store(struct ext4_attr *a,
2494 struct ext4_sb_info *sbi,
2495 const char *buf, size_t count)
2500 ret = kstrtoul(skip_spaces(buf), 0, &t);
2504 if (t && (!is_power_of_2(t) || t > 0x40000000))
2507 sbi->s_inode_readahead_blks = t;
2511 static ssize_t sbi_ui_show(struct ext4_attr *a,
2512 struct ext4_sb_info *sbi, char *buf)
2514 unsigned int *ui = (unsigned int *) (((char *) sbi) + a->u.offset);
2516 return snprintf(buf, PAGE_SIZE, "%u\n", *ui);
2519 static ssize_t sbi_ui_store(struct ext4_attr *a,
2520 struct ext4_sb_info *sbi,
2521 const char *buf, size_t count)
2523 unsigned int *ui = (unsigned int *) (((char *) sbi) + a->u.offset);
2527 ret = kstrtoul(skip_spaces(buf), 0, &t);
2534 static ssize_t es_ui_show(struct ext4_attr *a,
2535 struct ext4_sb_info *sbi, char *buf)
2538 unsigned int *ui = (unsigned int *) (((char *) sbi->s_es) +
2541 return snprintf(buf, PAGE_SIZE, "%u\n", *ui);
2544 static ssize_t reserved_clusters_show(struct ext4_attr *a,
2545 struct ext4_sb_info *sbi, char *buf)
2547 return snprintf(buf, PAGE_SIZE, "%llu\n",
2548 (unsigned long long) atomic64_read(&sbi->s_resv_clusters));
2551 static ssize_t reserved_clusters_store(struct ext4_attr *a,
2552 struct ext4_sb_info *sbi,
2553 const char *buf, size_t count)
2555 unsigned long long val;
2558 if (parse_strtoull(buf, -1ULL, &val))
2560 ret = ext4_reserve_clusters(sbi, val);
2562 return ret ? ret : count;
2565 static ssize_t trigger_test_error(struct ext4_attr *a,
2566 struct ext4_sb_info *sbi,
2567 const char *buf, size_t count)
2571 if (!capable(CAP_SYS_ADMIN))
2574 if (len && buf[len-1] == '\n')
2578 ext4_error(sbi->s_sb, "%.*s", len, buf);
2582 static ssize_t sbi_deprecated_show(struct ext4_attr *a,
2583 struct ext4_sb_info *sbi, char *buf)
2585 return snprintf(buf, PAGE_SIZE, "%d\n", a->u.deprecated_val);
2588 #define EXT4_ATTR_OFFSET(_name,_mode,_show,_store,_elname) \
2589 static struct ext4_attr ext4_attr_##_name = { \
2590 .attr = {.name = __stringify(_name), .mode = _mode }, \
2594 .offset = offsetof(struct ext4_sb_info, _elname),\
2598 #define EXT4_ATTR_OFFSET_ES(_name,_mode,_show,_store,_elname) \
2599 static struct ext4_attr ext4_attr_##_name = { \
2600 .attr = {.name = __stringify(_name), .mode = _mode }, \
2604 .offset = offsetof(struct ext4_super_block, _elname), \
2608 #define EXT4_ATTR(name, mode, show, store) \
2609 static struct ext4_attr ext4_attr_##name = __ATTR(name, mode, show, store)
2611 #define EXT4_INFO_ATTR(name) EXT4_ATTR(name, 0444, NULL, NULL)
2612 #define EXT4_RO_ATTR(name) EXT4_ATTR(name, 0444, name##_show, NULL)
2613 #define EXT4_RW_ATTR(name) EXT4_ATTR(name, 0644, name##_show, name##_store)
2615 #define EXT4_RO_ATTR_ES_UI(name, elname) \
2616 EXT4_ATTR_OFFSET_ES(name, 0444, es_ui_show, NULL, elname)
2617 #define EXT4_RW_ATTR_SBI_UI(name, elname) \
2618 EXT4_ATTR_OFFSET(name, 0644, sbi_ui_show, sbi_ui_store, elname)
2620 #define ATTR_LIST(name) &ext4_attr_##name.attr
2621 #define EXT4_DEPRECATED_ATTR(_name, _val) \
2622 static struct ext4_attr ext4_attr_##_name = { \
2623 .attr = {.name = __stringify(_name), .mode = 0444 }, \
2624 .show = sbi_deprecated_show, \
2626 .deprecated_val = _val, \
2630 EXT4_RO_ATTR(delayed_allocation_blocks);
2631 EXT4_RO_ATTR(session_write_kbytes);
2632 EXT4_RO_ATTR(lifetime_write_kbytes);
2633 EXT4_RW_ATTR(reserved_clusters);
2634 EXT4_ATTR_OFFSET(inode_readahead_blks, 0644, sbi_ui_show,
2635 inode_readahead_blks_store, s_inode_readahead_blks);
2636 EXT4_RW_ATTR_SBI_UI(inode_goal, s_inode_goal);
2637 EXT4_RW_ATTR_SBI_UI(mb_stats, s_mb_stats);
2638 EXT4_RW_ATTR_SBI_UI(mb_max_to_scan, s_mb_max_to_scan);
2639 EXT4_RW_ATTR_SBI_UI(mb_min_to_scan, s_mb_min_to_scan);
2640 EXT4_RW_ATTR_SBI_UI(mb_order2_req, s_mb_order2_reqs);
2641 EXT4_RW_ATTR_SBI_UI(mb_stream_req, s_mb_stream_request);
2642 EXT4_RW_ATTR_SBI_UI(mb_group_prealloc, s_mb_group_prealloc);
2643 EXT4_DEPRECATED_ATTR(max_writeback_mb_bump, 128);
2644 EXT4_RW_ATTR_SBI_UI(extent_max_zeroout_kb, s_extent_max_zeroout_kb);
2645 EXT4_ATTR(trigger_fs_error, 0200, NULL, trigger_test_error);
2646 EXT4_RW_ATTR_SBI_UI(err_ratelimit_interval_ms, s_err_ratelimit_state.interval);
2647 EXT4_RW_ATTR_SBI_UI(err_ratelimit_burst, s_err_ratelimit_state.burst);
2648 EXT4_RW_ATTR_SBI_UI(warning_ratelimit_interval_ms, s_warning_ratelimit_state.interval);
2649 EXT4_RW_ATTR_SBI_UI(warning_ratelimit_burst, s_warning_ratelimit_state.burst);
2650 EXT4_RW_ATTR_SBI_UI(msg_ratelimit_interval_ms, s_msg_ratelimit_state.interval);
2651 EXT4_RW_ATTR_SBI_UI(msg_ratelimit_burst, s_msg_ratelimit_state.burst);
2652 EXT4_RO_ATTR_ES_UI(errors_count, s_error_count);
2653 EXT4_RO_ATTR_ES_UI(first_error_time, s_first_error_time);
2654 EXT4_RO_ATTR_ES_UI(last_error_time, s_last_error_time);
2656 static struct attribute *ext4_attrs[] = {
2657 ATTR_LIST(delayed_allocation_blocks),
2658 ATTR_LIST(session_write_kbytes),
2659 ATTR_LIST(lifetime_write_kbytes),
2660 ATTR_LIST(reserved_clusters),
2661 ATTR_LIST(inode_readahead_blks),
2662 ATTR_LIST(inode_goal),
2663 ATTR_LIST(mb_stats),
2664 ATTR_LIST(mb_max_to_scan),
2665 ATTR_LIST(mb_min_to_scan),
2666 ATTR_LIST(mb_order2_req),
2667 ATTR_LIST(mb_stream_req),
2668 ATTR_LIST(mb_group_prealloc),
2669 ATTR_LIST(max_writeback_mb_bump),
2670 ATTR_LIST(extent_max_zeroout_kb),
2671 ATTR_LIST(trigger_fs_error),
2672 ATTR_LIST(err_ratelimit_interval_ms),
2673 ATTR_LIST(err_ratelimit_burst),
2674 ATTR_LIST(warning_ratelimit_interval_ms),
2675 ATTR_LIST(warning_ratelimit_burst),
2676 ATTR_LIST(msg_ratelimit_interval_ms),
2677 ATTR_LIST(msg_ratelimit_burst),
2678 ATTR_LIST(errors_count),
2679 ATTR_LIST(first_error_time),
2680 ATTR_LIST(last_error_time),
2684 /* Features this copy of ext4 supports */
2685 EXT4_INFO_ATTR(lazy_itable_init);
2686 EXT4_INFO_ATTR(batched_discard);
2687 EXT4_INFO_ATTR(meta_bg_resize);
2688 EXT4_INFO_ATTR(encryption);
2690 static struct attribute *ext4_feat_attrs[] = {
2691 ATTR_LIST(lazy_itable_init),
2692 ATTR_LIST(batched_discard),
2693 ATTR_LIST(meta_bg_resize),
2694 ATTR_LIST(encryption),
2698 static ssize_t ext4_attr_show(struct kobject *kobj,
2699 struct attribute *attr, char *buf)
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->show ? a->show(a, sbi, buf) : 0;
2708 static ssize_t ext4_attr_store(struct kobject *kobj,
2709 struct attribute *attr,
2710 const char *buf, size_t len)
2712 struct ext4_sb_info *sbi = container_of(kobj, struct ext4_sb_info,
2714 struct ext4_attr *a = container_of(attr, struct ext4_attr, attr);
2716 return a->store ? a->store(a, sbi, buf, len) : 0;
2719 static void ext4_sb_release(struct kobject *kobj)
2721 struct ext4_sb_info *sbi = container_of(kobj, struct ext4_sb_info,
2723 complete(&sbi->s_kobj_unregister);
2726 static const struct sysfs_ops ext4_attr_ops = {
2727 .show = ext4_attr_show,
2728 .store = ext4_attr_store,
2731 static struct kobj_type ext4_ktype = {
2732 .default_attrs = ext4_attrs,
2733 .sysfs_ops = &ext4_attr_ops,
2734 .release = ext4_sb_release,
2737 static void ext4_feat_release(struct kobject *kobj)
2739 complete(&ext4_feat->f_kobj_unregister);
2742 static ssize_t ext4_feat_show(struct kobject *kobj,
2743 struct attribute *attr, char *buf)
2745 return snprintf(buf, PAGE_SIZE, "supported\n");
2749 * We can not use ext4_attr_show/store because it relies on the kobject
2750 * being embedded in the ext4_sb_info structure which is definitely not
2751 * true in this case.
2753 static const struct sysfs_ops ext4_feat_ops = {
2754 .show = ext4_feat_show,
2758 static struct kobj_type ext4_feat_ktype = {
2759 .default_attrs = ext4_feat_attrs,
2760 .sysfs_ops = &ext4_feat_ops,
2761 .release = ext4_feat_release,
2765 * Check whether this filesystem can be mounted based on
2766 * the features present and the RDONLY/RDWR mount requested.
2767 * Returns 1 if this filesystem can be mounted as requested,
2768 * 0 if it cannot be.
2770 static int ext4_feature_set_ok(struct super_block *sb, int readonly)
2772 if (EXT4_HAS_INCOMPAT_FEATURE(sb, ~EXT4_FEATURE_INCOMPAT_SUPP)) {
2773 ext4_msg(sb, KERN_ERR,
2774 "Couldn't mount because of "
2775 "unsupported optional features (%x)",
2776 (le32_to_cpu(EXT4_SB(sb)->s_es->s_feature_incompat) &
2777 ~EXT4_FEATURE_INCOMPAT_SUPP));
2784 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_READONLY)) {
2785 ext4_msg(sb, KERN_INFO, "filesystem is read-only");
2786 sb->s_flags |= MS_RDONLY;
2790 /* Check that feature set is OK for a read-write mount */
2791 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, ~EXT4_FEATURE_RO_COMPAT_SUPP)) {
2792 ext4_msg(sb, KERN_ERR, "couldn't mount RDWR because of "
2793 "unsupported optional features (%x)",
2794 (le32_to_cpu(EXT4_SB(sb)->s_es->s_feature_ro_compat) &
2795 ~EXT4_FEATURE_RO_COMPAT_SUPP));
2799 * Large file size enabled file system can only be mounted
2800 * read-write on 32-bit systems if kernel is built with CONFIG_LBDAF
2802 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_HUGE_FILE)) {
2803 if (sizeof(blkcnt_t) < sizeof(u64)) {
2804 ext4_msg(sb, KERN_ERR, "Filesystem with huge files "
2805 "cannot be mounted RDWR without "
2810 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_BIGALLOC) &&
2811 !EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_EXTENTS)) {
2812 ext4_msg(sb, KERN_ERR,
2813 "Can't support bigalloc feature without "
2814 "extents feature\n");
2818 #ifndef CONFIG_QUOTA
2819 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_QUOTA) &&
2821 ext4_msg(sb, KERN_ERR,
2822 "Filesystem with quota feature cannot be mounted RDWR "
2823 "without CONFIG_QUOTA");
2826 #endif /* CONFIG_QUOTA */
2831 * This function is called once a day if we have errors logged
2832 * on the file system
2834 static void print_daily_error_info(unsigned long arg)
2836 struct super_block *sb = (struct super_block *) arg;
2837 struct ext4_sb_info *sbi;
2838 struct ext4_super_block *es;
2843 if (es->s_error_count)
2844 /* fsck newer than v1.41.13 is needed to clean this condition. */
2845 ext4_msg(sb, KERN_NOTICE, "error count since last fsck: %u",
2846 le32_to_cpu(es->s_error_count));
2847 if (es->s_first_error_time) {
2848 printk(KERN_NOTICE "EXT4-fs (%s): initial error at time %u: %.*s:%d",
2849 sb->s_id, le32_to_cpu(es->s_first_error_time),
2850 (int) sizeof(es->s_first_error_func),
2851 es->s_first_error_func,
2852 le32_to_cpu(es->s_first_error_line));
2853 if (es->s_first_error_ino)
2854 printk(": inode %u",
2855 le32_to_cpu(es->s_first_error_ino));
2856 if (es->s_first_error_block)
2857 printk(": block %llu", (unsigned long long)
2858 le64_to_cpu(es->s_first_error_block));
2861 if (es->s_last_error_time) {
2862 printk(KERN_NOTICE "EXT4-fs (%s): last error at time %u: %.*s:%d",
2863 sb->s_id, le32_to_cpu(es->s_last_error_time),
2864 (int) sizeof(es->s_last_error_func),
2865 es->s_last_error_func,
2866 le32_to_cpu(es->s_last_error_line));
2867 if (es->s_last_error_ino)
2868 printk(": inode %u",
2869 le32_to_cpu(es->s_last_error_ino));
2870 if (es->s_last_error_block)
2871 printk(": block %llu", (unsigned long long)
2872 le64_to_cpu(es->s_last_error_block));
2875 mod_timer(&sbi->s_err_report, jiffies + 24*60*60*HZ); /* Once a day */
2878 /* Find next suitable group and run ext4_init_inode_table */
2879 static int ext4_run_li_request(struct ext4_li_request *elr)
2881 struct ext4_group_desc *gdp = NULL;
2882 ext4_group_t group, ngroups;
2883 struct super_block *sb;
2884 unsigned long timeout = 0;
2888 ngroups = EXT4_SB(sb)->s_groups_count;
2891 for (group = elr->lr_next_group; group < ngroups; group++) {
2892 gdp = ext4_get_group_desc(sb, group, NULL);
2898 if (!(gdp->bg_flags & cpu_to_le16(EXT4_BG_INODE_ZEROED)))
2902 if (group >= ngroups)
2907 ret = ext4_init_inode_table(sb, group,
2908 elr->lr_timeout ? 0 : 1);
2909 if (elr->lr_timeout == 0) {
2910 timeout = (jiffies - timeout) *
2911 elr->lr_sbi->s_li_wait_mult;
2912 elr->lr_timeout = timeout;
2914 elr->lr_next_sched = jiffies + elr->lr_timeout;
2915 elr->lr_next_group = group + 1;
2923 * Remove lr_request from the list_request and free the
2924 * request structure. Should be called with li_list_mtx held
2926 static void ext4_remove_li_request(struct ext4_li_request *elr)
2928 struct ext4_sb_info *sbi;
2935 list_del(&elr->lr_request);
2936 sbi->s_li_request = NULL;
2940 static void ext4_unregister_li_request(struct super_block *sb)
2942 mutex_lock(&ext4_li_mtx);
2943 if (!ext4_li_info) {
2944 mutex_unlock(&ext4_li_mtx);
2948 mutex_lock(&ext4_li_info->li_list_mtx);
2949 ext4_remove_li_request(EXT4_SB(sb)->s_li_request);
2950 mutex_unlock(&ext4_li_info->li_list_mtx);
2951 mutex_unlock(&ext4_li_mtx);
2954 static struct task_struct *ext4_lazyinit_task;
2957 * This is the function where ext4lazyinit thread lives. It walks
2958 * through the request list searching for next scheduled filesystem.
2959 * When such a fs is found, run the lazy initialization request
2960 * (ext4_rn_li_request) and keep track of the time spend in this
2961 * function. Based on that time we compute next schedule time of
2962 * the request. When walking through the list is complete, compute
2963 * next waking time and put itself into sleep.
2965 static int ext4_lazyinit_thread(void *arg)
2967 struct ext4_lazy_init *eli = (struct ext4_lazy_init *)arg;
2968 struct list_head *pos, *n;
2969 struct ext4_li_request *elr;
2970 unsigned long next_wakeup, cur;
2972 BUG_ON(NULL == eli);
2976 next_wakeup = MAX_JIFFY_OFFSET;
2978 mutex_lock(&eli->li_list_mtx);
2979 if (list_empty(&eli->li_request_list)) {
2980 mutex_unlock(&eli->li_list_mtx);
2984 list_for_each_safe(pos, n, &eli->li_request_list) {
2985 elr = list_entry(pos, struct ext4_li_request,
2988 if (time_after_eq(jiffies, elr->lr_next_sched)) {
2989 if (ext4_run_li_request(elr) != 0) {
2990 /* error, remove the lazy_init job */
2991 ext4_remove_li_request(elr);
2996 if (time_before(elr->lr_next_sched, next_wakeup))
2997 next_wakeup = elr->lr_next_sched;
2999 mutex_unlock(&eli->li_list_mtx);
3004 if ((time_after_eq(cur, next_wakeup)) ||
3005 (MAX_JIFFY_OFFSET == next_wakeup)) {
3010 schedule_timeout_interruptible(next_wakeup - cur);
3012 if (kthread_should_stop()) {
3013 ext4_clear_request_list();
3020 * It looks like the request list is empty, but we need
3021 * to check it under the li_list_mtx lock, to prevent any
3022 * additions into it, and of course we should lock ext4_li_mtx
3023 * to atomically free the list and ext4_li_info, because at
3024 * this point another ext4 filesystem could be registering
3027 mutex_lock(&ext4_li_mtx);
3028 mutex_lock(&eli->li_list_mtx);
3029 if (!list_empty(&eli->li_request_list)) {
3030 mutex_unlock(&eli->li_list_mtx);
3031 mutex_unlock(&ext4_li_mtx);
3034 mutex_unlock(&eli->li_list_mtx);
3035 kfree(ext4_li_info);
3036 ext4_li_info = NULL;
3037 mutex_unlock(&ext4_li_mtx);
3042 static void ext4_clear_request_list(void)
3044 struct list_head *pos, *n;
3045 struct ext4_li_request *elr;
3047 mutex_lock(&ext4_li_info->li_list_mtx);
3048 list_for_each_safe(pos, n, &ext4_li_info->li_request_list) {
3049 elr = list_entry(pos, struct ext4_li_request,
3051 ext4_remove_li_request(elr);
3053 mutex_unlock(&ext4_li_info->li_list_mtx);
3056 static int ext4_run_lazyinit_thread(void)
3058 ext4_lazyinit_task = kthread_run(ext4_lazyinit_thread,
3059 ext4_li_info, "ext4lazyinit");
3060 if (IS_ERR(ext4_lazyinit_task)) {
3061 int err = PTR_ERR(ext4_lazyinit_task);
3062 ext4_clear_request_list();
3063 kfree(ext4_li_info);
3064 ext4_li_info = NULL;
3065 printk(KERN_CRIT "EXT4-fs: error %d creating inode table "
3066 "initialization thread\n",
3070 ext4_li_info->li_state |= EXT4_LAZYINIT_RUNNING;
3075 * Check whether it make sense to run itable init. thread or not.
3076 * If there is at least one uninitialized inode table, return
3077 * corresponding group number, else the loop goes through all
3078 * groups and return total number of groups.
3080 static ext4_group_t ext4_has_uninit_itable(struct super_block *sb)
3082 ext4_group_t group, ngroups = EXT4_SB(sb)->s_groups_count;
3083 struct ext4_group_desc *gdp = NULL;
3085 for (group = 0; group < ngroups; group++) {
3086 gdp = ext4_get_group_desc(sb, group, NULL);
3090 if (!(gdp->bg_flags & cpu_to_le16(EXT4_BG_INODE_ZEROED)))
3097 static int ext4_li_info_new(void)
3099 struct ext4_lazy_init *eli = NULL;
3101 eli = kzalloc(sizeof(*eli), GFP_KERNEL);
3105 INIT_LIST_HEAD(&eli->li_request_list);
3106 mutex_init(&eli->li_list_mtx);
3108 eli->li_state |= EXT4_LAZYINIT_QUIT;
3115 static struct ext4_li_request *ext4_li_request_new(struct super_block *sb,
3118 struct ext4_sb_info *sbi = EXT4_SB(sb);
3119 struct ext4_li_request *elr;
3121 elr = kzalloc(sizeof(*elr), GFP_KERNEL);
3127 elr->lr_next_group = start;
3130 * Randomize first schedule time of the request to
3131 * spread the inode table initialization requests
3134 elr->lr_next_sched = jiffies + (prandom_u32() %
3135 (EXT4_DEF_LI_MAX_START_DELAY * HZ));
3139 int ext4_register_li_request(struct super_block *sb,
3140 ext4_group_t first_not_zeroed)
3142 struct ext4_sb_info *sbi = EXT4_SB(sb);
3143 struct ext4_li_request *elr = NULL;
3144 ext4_group_t ngroups = EXT4_SB(sb)->s_groups_count;
3147 mutex_lock(&ext4_li_mtx);
3148 if (sbi->s_li_request != NULL) {
3150 * Reset timeout so it can be computed again, because
3151 * s_li_wait_mult might have changed.
3153 sbi->s_li_request->lr_timeout = 0;
3157 if (first_not_zeroed == ngroups ||
3158 (sb->s_flags & MS_RDONLY) ||
3159 !test_opt(sb, INIT_INODE_TABLE))
3162 elr = ext4_li_request_new(sb, first_not_zeroed);
3168 if (NULL == ext4_li_info) {
3169 ret = ext4_li_info_new();
3174 mutex_lock(&ext4_li_info->li_list_mtx);
3175 list_add(&elr->lr_request, &ext4_li_info->li_request_list);
3176 mutex_unlock(&ext4_li_info->li_list_mtx);
3178 sbi->s_li_request = elr;
3180 * set elr to NULL here since it has been inserted to
3181 * the request_list and the removal and free of it is
3182 * handled by ext4_clear_request_list from now on.
3186 if (!(ext4_li_info->li_state & EXT4_LAZYINIT_RUNNING)) {
3187 ret = ext4_run_lazyinit_thread();
3192 mutex_unlock(&ext4_li_mtx);
3199 * We do not need to lock anything since this is called on
3202 static void ext4_destroy_lazyinit_thread(void)
3205 * If thread exited earlier
3206 * there's nothing to be done.
3208 if (!ext4_li_info || !ext4_lazyinit_task)
3211 kthread_stop(ext4_lazyinit_task);
3214 static int set_journal_csum_feature_set(struct super_block *sb)
3217 int compat, incompat;
3218 struct ext4_sb_info *sbi = EXT4_SB(sb);
3220 if (ext4_has_metadata_csum(sb)) {
3221 /* journal checksum v3 */
3223 incompat = JBD2_FEATURE_INCOMPAT_CSUM_V3;
3225 /* journal checksum v1 */
3226 compat = JBD2_FEATURE_COMPAT_CHECKSUM;
3230 jbd2_journal_clear_features(sbi->s_journal,
3231 JBD2_FEATURE_COMPAT_CHECKSUM, 0,
3232 JBD2_FEATURE_INCOMPAT_CSUM_V3 |
3233 JBD2_FEATURE_INCOMPAT_CSUM_V2);
3234 if (test_opt(sb, JOURNAL_ASYNC_COMMIT)) {
3235 ret = jbd2_journal_set_features(sbi->s_journal,
3237 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT |
3239 } else if (test_opt(sb, JOURNAL_CHECKSUM)) {
3240 ret = jbd2_journal_set_features(sbi->s_journal,
3243 jbd2_journal_clear_features(sbi->s_journal, 0, 0,
3244 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT);
3246 jbd2_journal_clear_features(sbi->s_journal, 0, 0,
3247 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT);
3254 * Note: calculating the overhead so we can be compatible with
3255 * historical BSD practice is quite difficult in the face of
3256 * clusters/bigalloc. This is because multiple metadata blocks from
3257 * different block group can end up in the same allocation cluster.
3258 * Calculating the exact overhead in the face of clustered allocation
3259 * requires either O(all block bitmaps) in memory or O(number of block
3260 * groups**2) in time. We will still calculate the superblock for
3261 * older file systems --- and if we come across with a bigalloc file
3262 * system with zero in s_overhead_clusters the estimate will be close to
3263 * correct especially for very large cluster sizes --- but for newer
3264 * file systems, it's better to calculate this figure once at mkfs
3265 * time, and store it in the superblock. If the superblock value is
3266 * present (even for non-bigalloc file systems), we will use it.
3268 static int count_overhead(struct super_block *sb, ext4_group_t grp,
3271 struct ext4_sb_info *sbi = EXT4_SB(sb);
3272 struct ext4_group_desc *gdp;
3273 ext4_fsblk_t first_block, last_block, b;
3274 ext4_group_t i, ngroups = ext4_get_groups_count(sb);
3275 int s, j, count = 0;
3277 if (!EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_BIGALLOC))
3278 return (ext4_bg_has_super(sb, grp) + ext4_bg_num_gdb(sb, grp) +
3279 sbi->s_itb_per_group + 2);
3281 first_block = le32_to_cpu(sbi->s_es->s_first_data_block) +
3282 (grp * EXT4_BLOCKS_PER_GROUP(sb));
3283 last_block = first_block + EXT4_BLOCKS_PER_GROUP(sb) - 1;
3284 for (i = 0; i < ngroups; i++) {
3285 gdp = ext4_get_group_desc(sb, i, NULL);
3286 b = ext4_block_bitmap(sb, gdp);
3287 if (b >= first_block && b <= last_block) {
3288 ext4_set_bit(EXT4_B2C(sbi, b - first_block), buf);
3291 b = ext4_inode_bitmap(sb, gdp);
3292 if (b >= first_block && b <= last_block) {
3293 ext4_set_bit(EXT4_B2C(sbi, b - first_block), buf);
3296 b = ext4_inode_table(sb, gdp);
3297 if (b >= first_block && b + sbi->s_itb_per_group <= last_block)
3298 for (j = 0; j < sbi->s_itb_per_group; j++, b++) {
3299 int c = EXT4_B2C(sbi, b - first_block);
3300 ext4_set_bit(c, buf);
3306 if (ext4_bg_has_super(sb, grp)) {
3307 ext4_set_bit(s++, buf);
3310 for (j = ext4_bg_num_gdb(sb, grp); j > 0; j--) {
3311 ext4_set_bit(EXT4_B2C(sbi, s++), buf);
3317 return EXT4_CLUSTERS_PER_GROUP(sb) -
3318 ext4_count_free(buf, EXT4_CLUSTERS_PER_GROUP(sb) / 8);
3322 * Compute the overhead and stash it in sbi->s_overhead
3324 int ext4_calculate_overhead(struct super_block *sb)
3326 struct ext4_sb_info *sbi = EXT4_SB(sb);
3327 struct ext4_super_block *es = sbi->s_es;
3328 ext4_group_t i, ngroups = ext4_get_groups_count(sb);
3329 ext4_fsblk_t overhead = 0;
3330 char *buf = (char *) get_zeroed_page(GFP_NOFS);
3336 * Compute the overhead (FS structures). This is constant
3337 * for a given filesystem unless the number of block groups
3338 * changes so we cache the previous value until it does.
3342 * All of the blocks before first_data_block are overhead
3344 overhead = EXT4_B2C(sbi, le32_to_cpu(es->s_first_data_block));
3347 * Add the overhead found in each block group
3349 for (i = 0; i < ngroups; i++) {
3352 blks = count_overhead(sb, i, buf);
3355 memset(buf, 0, PAGE_SIZE);
3358 /* Add the internal journal blocks as well */
3359 if (sbi->s_journal && !sbi->journal_bdev)
3360 overhead += EXT4_NUM_B2C(sbi, sbi->s_journal->j_maxlen);
3362 sbi->s_overhead = overhead;
3364 free_page((unsigned long) buf);
3369 static ext4_fsblk_t ext4_calculate_resv_clusters(struct super_block *sb)
3371 ext4_fsblk_t resv_clusters;
3374 * There's no need to reserve anything when we aren't using extents.
3375 * The space estimates are exact, there are no unwritten extents,
3376 * hole punching doesn't need new metadata... This is needed especially
3377 * to keep ext2/3 backward compatibility.
3379 if (!EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_EXTENTS))
3382 * By default we reserve 2% or 4096 clusters, whichever is smaller.
3383 * This should cover the situations where we can not afford to run
3384 * out of space like for example punch hole, or converting
3385 * unwritten extents in delalloc path. In most cases such
3386 * allocation would require 1, or 2 blocks, higher numbers are
3389 resv_clusters = ext4_blocks_count(EXT4_SB(sb)->s_es) >>
3390 EXT4_SB(sb)->s_cluster_bits;
3392 do_div(resv_clusters, 50);
3393 resv_clusters = min_t(ext4_fsblk_t, resv_clusters, 4096);
3395 return resv_clusters;
3399 static int ext4_reserve_clusters(struct ext4_sb_info *sbi, ext4_fsblk_t count)
3401 ext4_fsblk_t clusters = ext4_blocks_count(sbi->s_es) >>
3402 sbi->s_cluster_bits;
3404 if (count >= clusters)
3407 atomic64_set(&sbi->s_resv_clusters, count);
3411 static int ext4_fill_super(struct super_block *sb, void *data, int silent)
3413 char *orig_data = kstrdup(data, GFP_KERNEL);
3414 struct buffer_head *bh;
3415 struct ext4_super_block *es = NULL;
3416 struct ext4_sb_info *sbi;
3418 ext4_fsblk_t sb_block = get_sb_block(&data);
3419 ext4_fsblk_t logical_sb_block;
3420 unsigned long offset = 0;
3421 unsigned long journal_devnum = 0;
3422 unsigned long def_mount_opts;
3427 int blocksize, clustersize;
3428 unsigned int db_count;
3430 int needs_recovery, has_huge_files, has_bigalloc;
3433 unsigned int journal_ioprio = DEFAULT_JOURNAL_IOPRIO;
3434 ext4_group_t first_not_zeroed;
3436 sbi = kzalloc(sizeof(*sbi), GFP_KERNEL);
3440 sbi->s_blockgroup_lock =
3441 kzalloc(sizeof(struct blockgroup_lock), GFP_KERNEL);
3442 if (!sbi->s_blockgroup_lock) {
3446 sb->s_fs_info = sbi;
3448 sbi->s_inode_readahead_blks = EXT4_DEF_INODE_READAHEAD_BLKS;
3449 sbi->s_sb_block = sb_block;
3450 if (sb->s_bdev->bd_part)
3451 sbi->s_sectors_written_start =
3452 part_stat_read(sb->s_bdev->bd_part, sectors[1]);
3453 #ifdef CONFIG_EXT4_FS_ENCRYPTION
3454 /* Modes of operations for file and directory encryption. */
3455 sbi->s_file_encryption_mode = EXT4_ENCRYPTION_MODE_AES_256_XTS;
3456 sbi->s_dir_encryption_mode = EXT4_ENCRYPTION_MODE_INVALID;
3459 /* Cleanup superblock name */
3460 for (cp = sb->s_id; (cp = strchr(cp, '/'));)
3463 /* -EINVAL is default */
3465 blocksize = sb_min_blocksize(sb, EXT4_MIN_BLOCK_SIZE);
3467 ext4_msg(sb, KERN_ERR, "unable to set blocksize");
3472 * The ext4 superblock will not be buffer aligned for other than 1kB
3473 * block sizes. We need to calculate the offset from buffer start.
3475 if (blocksize != EXT4_MIN_BLOCK_SIZE) {
3476 logical_sb_block = sb_block * EXT4_MIN_BLOCK_SIZE;
3477 offset = do_div(logical_sb_block, blocksize);
3479 logical_sb_block = sb_block;
3482 if (!(bh = sb_bread_unmovable(sb, logical_sb_block))) {
3483 ext4_msg(sb, KERN_ERR, "unable to read superblock");
3487 * Note: s_es must be initialized as soon as possible because
3488 * some ext4 macro-instructions depend on its value
3490 es = (struct ext4_super_block *) (bh->b_data + offset);
3492 sb->s_magic = le16_to_cpu(es->s_magic);
3493 if (sb->s_magic != EXT4_SUPER_MAGIC)
3495 sbi->s_kbytes_written = le64_to_cpu(es->s_kbytes_written);
3497 /* Warn if metadata_csum and gdt_csum are both set. */
3498 if (EXT4_HAS_RO_COMPAT_FEATURE(sb,
3499 EXT4_FEATURE_RO_COMPAT_METADATA_CSUM) &&
3500 EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_GDT_CSUM))
3501 ext4_warning(sb, "metadata_csum and uninit_bg are "
3502 "redundant flags; please run fsck.");
3504 /* Check for a known checksum algorithm */
3505 if (!ext4_verify_csum_type(sb, es)) {
3506 ext4_msg(sb, KERN_ERR, "VFS: Found ext4 filesystem with "
3507 "unknown checksum algorithm.");
3512 /* Load the checksum driver */
3513 if (EXT4_HAS_RO_COMPAT_FEATURE(sb,
3514 EXT4_FEATURE_RO_COMPAT_METADATA_CSUM)) {
3515 sbi->s_chksum_driver = crypto_alloc_shash("crc32c", 0, 0);
3516 if (IS_ERR(sbi->s_chksum_driver)) {
3517 ext4_msg(sb, KERN_ERR, "Cannot load crc32c driver.");
3518 ret = PTR_ERR(sbi->s_chksum_driver);
3519 sbi->s_chksum_driver = NULL;
3524 /* Check superblock checksum */
3525 if (!ext4_superblock_csum_verify(sb, es)) {
3526 ext4_msg(sb, KERN_ERR, "VFS: Found ext4 filesystem with "
3527 "invalid superblock checksum. Run e2fsck?");
3532 /* Precompute checksum seed for all metadata */
3533 if (ext4_has_metadata_csum(sb))
3534 sbi->s_csum_seed = ext4_chksum(sbi, ~0, es->s_uuid,
3535 sizeof(es->s_uuid));
3537 /* Set defaults before we parse the mount options */
3538 def_mount_opts = le32_to_cpu(es->s_default_mount_opts);
3539 set_opt(sb, INIT_INODE_TABLE);
3540 if (def_mount_opts & EXT4_DEFM_DEBUG)
3542 if (def_mount_opts & EXT4_DEFM_BSDGROUPS)
3544 if (def_mount_opts & EXT4_DEFM_UID16)
3545 set_opt(sb, NO_UID32);
3546 /* xattr user namespace & acls are now defaulted on */
3547 set_opt(sb, XATTR_USER);
3548 #ifdef CONFIG_EXT4_FS_POSIX_ACL
3549 set_opt(sb, POSIX_ACL);
3551 /* don't forget to enable journal_csum when metadata_csum is enabled. */
3552 if (ext4_has_metadata_csum(sb))
3553 set_opt(sb, JOURNAL_CHECKSUM);
3555 if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_DATA)
3556 set_opt(sb, JOURNAL_DATA);
3557 else if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_ORDERED)
3558 set_opt(sb, ORDERED_DATA);
3559 else if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_WBACK)
3560 set_opt(sb, WRITEBACK_DATA);
3562 if (le16_to_cpu(sbi->s_es->s_errors) == EXT4_ERRORS_PANIC)
3563 set_opt(sb, ERRORS_PANIC);
3564 else if (le16_to_cpu(sbi->s_es->s_errors) == EXT4_ERRORS_CONTINUE)
3565 set_opt(sb, ERRORS_CONT);
3567 set_opt(sb, ERRORS_RO);
3568 /* block_validity enabled by default; disable with noblock_validity */
3569 set_opt(sb, BLOCK_VALIDITY);
3570 if (def_mount_opts & EXT4_DEFM_DISCARD)
3571 set_opt(sb, DISCARD);
3573 sbi->s_resuid = make_kuid(&init_user_ns, le16_to_cpu(es->s_def_resuid));
3574 sbi->s_resgid = make_kgid(&init_user_ns, le16_to_cpu(es->s_def_resgid));
3575 sbi->s_commit_interval = JBD2_DEFAULT_MAX_COMMIT_AGE * HZ;
3576 sbi->s_min_batch_time = EXT4_DEF_MIN_BATCH_TIME;
3577 sbi->s_max_batch_time = EXT4_DEF_MAX_BATCH_TIME;
3579 if ((def_mount_opts & EXT4_DEFM_NOBARRIER) == 0)
3580 set_opt(sb, BARRIER);
3583 * enable delayed allocation by default
3584 * Use -o nodelalloc to turn it off
3586 if (!IS_EXT3_SB(sb) && !IS_EXT2_SB(sb) &&
3587 ((def_mount_opts & EXT4_DEFM_NODELALLOC) == 0))
3588 set_opt(sb, DELALLOC);
3591 * set default s_li_wait_mult for lazyinit, for the case there is
3592 * no mount option specified.
3594 sbi->s_li_wait_mult = EXT4_DEF_LI_WAIT_MULT;
3596 if (!parse_options((char *) sbi->s_es->s_mount_opts, sb,
3597 &journal_devnum, &journal_ioprio, 0)) {
3598 ext4_msg(sb, KERN_WARNING,
3599 "failed to parse options in superblock: %s",
3600 sbi->s_es->s_mount_opts);
3602 sbi->s_def_mount_opt = sbi->s_mount_opt;
3603 if (!parse_options((char *) data, sb, &journal_devnum,
3604 &journal_ioprio, 0))
3607 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA) {
3608 printk_once(KERN_WARNING "EXT4-fs: Warning: mounting "
3609 "with data=journal disables delayed "
3610 "allocation and O_DIRECT support!\n");
3611 if (test_opt2(sb, EXPLICIT_DELALLOC)) {
3612 ext4_msg(sb, KERN_ERR, "can't mount with "
3613 "both data=journal and delalloc");
3616 if (test_opt(sb, DIOREAD_NOLOCK)) {
3617 ext4_msg(sb, KERN_ERR, "can't mount with "
3618 "both data=journal and dioread_nolock");
3621 if (test_opt(sb, DAX)) {
3622 ext4_msg(sb, KERN_ERR, "can't mount with "
3623 "both data=journal and dax");
3626 if (test_opt(sb, DELALLOC))
3627 clear_opt(sb, DELALLOC);
3630 sb->s_flags = (sb->s_flags & ~MS_POSIXACL) |
3631 (test_opt(sb, POSIX_ACL) ? MS_POSIXACL : 0);
3633 if (le32_to_cpu(es->s_rev_level) == EXT4_GOOD_OLD_REV &&
3634 (EXT4_HAS_COMPAT_FEATURE(sb, ~0U) ||
3635 EXT4_HAS_RO_COMPAT_FEATURE(sb, ~0U) ||
3636 EXT4_HAS_INCOMPAT_FEATURE(sb, ~0U)))
3637 ext4_msg(sb, KERN_WARNING,
3638 "feature flags set on rev 0 fs, "
3639 "running e2fsck is recommended");
3641 if (es->s_creator_os == cpu_to_le32(EXT4_OS_HURD)) {
3642 set_opt2(sb, HURD_COMPAT);
3643 if (EXT4_HAS_INCOMPAT_FEATURE(sb,
3644 EXT4_FEATURE_INCOMPAT_64BIT)) {
3645 ext4_msg(sb, KERN_ERR,
3646 "The Hurd can't support 64-bit file systems");
3651 if (IS_EXT2_SB(sb)) {
3652 if (ext2_feature_set_ok(sb))
3653 ext4_msg(sb, KERN_INFO, "mounting ext2 file system "
3654 "using the ext4 subsystem");
3656 ext4_msg(sb, KERN_ERR, "couldn't mount as ext2 due "
3657 "to feature incompatibilities");
3662 if (IS_EXT3_SB(sb)) {
3663 if (ext3_feature_set_ok(sb))
3664 ext4_msg(sb, KERN_INFO, "mounting ext3 file system "
3665 "using the ext4 subsystem");
3667 ext4_msg(sb, KERN_ERR, "couldn't mount as ext3 due "
3668 "to feature incompatibilities");
3674 * Check feature flags regardless of the revision level, since we
3675 * previously didn't change the revision level when setting the flags,
3676 * so there is a chance incompat flags are set on a rev 0 filesystem.
3678 if (!ext4_feature_set_ok(sb, (sb->s_flags & MS_RDONLY)))
3681 blocksize = BLOCK_SIZE << le32_to_cpu(es->s_log_block_size);
3682 if (blocksize < EXT4_MIN_BLOCK_SIZE ||
3683 blocksize > EXT4_MAX_BLOCK_SIZE) {
3684 ext4_msg(sb, KERN_ERR,
3685 "Unsupported filesystem blocksize %d", blocksize);
3689 if (sbi->s_mount_opt & EXT4_MOUNT_DAX) {
3690 if (blocksize != PAGE_SIZE) {
3691 ext4_msg(sb, KERN_ERR,
3692 "error: unsupported blocksize for dax");
3695 if (!sb->s_bdev->bd_disk->fops->direct_access) {
3696 ext4_msg(sb, KERN_ERR,
3697 "error: device does not support dax");
3702 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_ENCRYPT) &&
3703 es->s_encryption_level) {
3704 ext4_msg(sb, KERN_ERR, "Unsupported encryption level %d",
3705 es->s_encryption_level);
3709 if (sb->s_blocksize != blocksize) {
3710 /* Validate the filesystem blocksize */
3711 if (!sb_set_blocksize(sb, blocksize)) {
3712 ext4_msg(sb, KERN_ERR, "bad block size %d",
3718 logical_sb_block = sb_block * EXT4_MIN_BLOCK_SIZE;
3719 offset = do_div(logical_sb_block, blocksize);
3720 bh = sb_bread_unmovable(sb, logical_sb_block);
3722 ext4_msg(sb, KERN_ERR,
3723 "Can't read superblock on 2nd try");
3726 es = (struct ext4_super_block *)(bh->b_data + offset);
3728 if (es->s_magic != cpu_to_le16(EXT4_SUPER_MAGIC)) {
3729 ext4_msg(sb, KERN_ERR,
3730 "Magic mismatch, very weird!");
3735 has_huge_files = EXT4_HAS_RO_COMPAT_FEATURE(sb,
3736 EXT4_FEATURE_RO_COMPAT_HUGE_FILE);
3737 sbi->s_bitmap_maxbytes = ext4_max_bitmap_size(sb->s_blocksize_bits,
3739 sb->s_maxbytes = ext4_max_size(sb->s_blocksize_bits, has_huge_files);
3741 if (le32_to_cpu(es->s_rev_level) == EXT4_GOOD_OLD_REV) {
3742 sbi->s_inode_size = EXT4_GOOD_OLD_INODE_SIZE;
3743 sbi->s_first_ino = EXT4_GOOD_OLD_FIRST_INO;
3745 sbi->s_inode_size = le16_to_cpu(es->s_inode_size);
3746 sbi->s_first_ino = le32_to_cpu(es->s_first_ino);
3747 if ((sbi->s_inode_size < EXT4_GOOD_OLD_INODE_SIZE) ||
3748 (!is_power_of_2(sbi->s_inode_size)) ||
3749 (sbi->s_inode_size > blocksize)) {
3750 ext4_msg(sb, KERN_ERR,
3751 "unsupported inode size: %d",
3755 if (sbi->s_inode_size > EXT4_GOOD_OLD_INODE_SIZE)
3756 sb->s_time_gran = 1 << (EXT4_EPOCH_BITS - 2);
3759 sbi->s_desc_size = le16_to_cpu(es->s_desc_size);
3760 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_64BIT)) {
3761 if (sbi->s_desc_size < EXT4_MIN_DESC_SIZE_64BIT ||
3762 sbi->s_desc_size > EXT4_MAX_DESC_SIZE ||
3763 !is_power_of_2(sbi->s_desc_size)) {
3764 ext4_msg(sb, KERN_ERR,
3765 "unsupported descriptor size %lu",
3770 sbi->s_desc_size = EXT4_MIN_DESC_SIZE;
3772 sbi->s_blocks_per_group = le32_to_cpu(es->s_blocks_per_group);
3773 sbi->s_inodes_per_group = le32_to_cpu(es->s_inodes_per_group);
3774 if (EXT4_INODE_SIZE(sb) == 0 || EXT4_INODES_PER_GROUP(sb) == 0)
3777 sbi->s_inodes_per_block = blocksize / EXT4_INODE_SIZE(sb);
3778 if (sbi->s_inodes_per_block == 0)
3780 sbi->s_itb_per_group = sbi->s_inodes_per_group /
3781 sbi->s_inodes_per_block;
3782 sbi->s_desc_per_block = blocksize / EXT4_DESC_SIZE(sb);
3784 sbi->s_mount_state = le16_to_cpu(es->s_state);
3785 sbi->s_addr_per_block_bits = ilog2(EXT4_ADDR_PER_BLOCK(sb));
3786 sbi->s_desc_per_block_bits = ilog2(EXT4_DESC_PER_BLOCK(sb));
3788 for (i = 0; i < 4; i++)
3789 sbi->s_hash_seed[i] = le32_to_cpu(es->s_hash_seed[i]);
3790 sbi->s_def_hash_version = es->s_def_hash_version;
3791 if (EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_DIR_INDEX)) {
3792 i = le32_to_cpu(es->s_flags);
3793 if (i & EXT2_FLAGS_UNSIGNED_HASH)
3794 sbi->s_hash_unsigned = 3;
3795 else if ((i & EXT2_FLAGS_SIGNED_HASH) == 0) {
3796 #ifdef __CHAR_UNSIGNED__
3797 if (!(sb->s_flags & MS_RDONLY))
3799 cpu_to_le32(EXT2_FLAGS_UNSIGNED_HASH);
3800 sbi->s_hash_unsigned = 3;
3802 if (!(sb->s_flags & MS_RDONLY))
3804 cpu_to_le32(EXT2_FLAGS_SIGNED_HASH);
3809 /* Handle clustersize */
3810 clustersize = BLOCK_SIZE << le32_to_cpu(es->s_log_cluster_size);
3811 has_bigalloc = EXT4_HAS_RO_COMPAT_FEATURE(sb,
3812 EXT4_FEATURE_RO_COMPAT_BIGALLOC);
3814 if (clustersize < blocksize) {
3815 ext4_msg(sb, KERN_ERR,
3816 "cluster size (%d) smaller than "
3817 "block size (%d)", clustersize, blocksize);
3820 sbi->s_cluster_bits = le32_to_cpu(es->s_log_cluster_size) -
3821 le32_to_cpu(es->s_log_block_size);
3822 sbi->s_clusters_per_group =
3823 le32_to_cpu(es->s_clusters_per_group);
3824 if (sbi->s_clusters_per_group > blocksize * 8) {
3825 ext4_msg(sb, KERN_ERR,
3826 "#clusters per group too big: %lu",
3827 sbi->s_clusters_per_group);
3830 if (sbi->s_blocks_per_group !=
3831 (sbi->s_clusters_per_group * (clustersize / blocksize))) {
3832 ext4_msg(sb, KERN_ERR, "blocks per group (%lu) and "
3833 "clusters per group (%lu) inconsistent",
3834 sbi->s_blocks_per_group,
3835 sbi->s_clusters_per_group);
3839 if (clustersize != blocksize) {
3840 ext4_warning(sb, "fragment/cluster size (%d) != "
3841 "block size (%d)", clustersize,
3843 clustersize = blocksize;
3845 if (sbi->s_blocks_per_group > blocksize * 8) {
3846 ext4_msg(sb, KERN_ERR,
3847 "#blocks per group too big: %lu",
3848 sbi->s_blocks_per_group);
3851 sbi->s_clusters_per_group = sbi->s_blocks_per_group;
3852 sbi->s_cluster_bits = 0;
3854 sbi->s_cluster_ratio = clustersize / blocksize;
3856 if (sbi->s_inodes_per_group > blocksize * 8) {
3857 ext4_msg(sb, KERN_ERR,
3858 "#inodes per group too big: %lu",
3859 sbi->s_inodes_per_group);
3863 /* Do we have standard group size of clustersize * 8 blocks ? */
3864 if (sbi->s_blocks_per_group == clustersize << 3)
3865 set_opt2(sb, STD_GROUP_SIZE);
3868 * Test whether we have more sectors than will fit in sector_t,
3869 * and whether the max offset is addressable by the page cache.
3871 err = generic_check_addressable(sb->s_blocksize_bits,
3872 ext4_blocks_count(es));
3874 ext4_msg(sb, KERN_ERR, "filesystem"
3875 " too large to mount safely on this system");
3876 if (sizeof(sector_t) < 8)
3877 ext4_msg(sb, KERN_WARNING, "CONFIG_LBDAF not enabled");
3881 if (EXT4_BLOCKS_PER_GROUP(sb) == 0)
3884 /* check blocks count against device size */
3885 blocks_count = sb->s_bdev->bd_inode->i_size >> sb->s_blocksize_bits;
3886 if (blocks_count && ext4_blocks_count(es) > blocks_count) {
3887 ext4_msg(sb, KERN_WARNING, "bad geometry: block count %llu "
3888 "exceeds size of device (%llu blocks)",
3889 ext4_blocks_count(es), blocks_count);
3894 * It makes no sense for the first data block to be beyond the end
3895 * of the filesystem.
3897 if (le32_to_cpu(es->s_first_data_block) >= ext4_blocks_count(es)) {
3898 ext4_msg(sb, KERN_WARNING, "bad geometry: first data "
3899 "block %u is beyond end of filesystem (%llu)",
3900 le32_to_cpu(es->s_first_data_block),
3901 ext4_blocks_count(es));
3904 blocks_count = (ext4_blocks_count(es) -
3905 le32_to_cpu(es->s_first_data_block) +
3906 EXT4_BLOCKS_PER_GROUP(sb) - 1);
3907 do_div(blocks_count, EXT4_BLOCKS_PER_GROUP(sb));
3908 if (blocks_count > ((uint64_t)1<<32) - EXT4_DESC_PER_BLOCK(sb)) {
3909 ext4_msg(sb, KERN_WARNING, "groups count too large: %u "
3910 "(block count %llu, first data block %u, "
3911 "blocks per group %lu)", sbi->s_groups_count,
3912 ext4_blocks_count(es),
3913 le32_to_cpu(es->s_first_data_block),
3914 EXT4_BLOCKS_PER_GROUP(sb));
3917 sbi->s_groups_count = blocks_count;
3918 sbi->s_blockfile_groups = min_t(ext4_group_t, sbi->s_groups_count,
3919 (EXT4_MAX_BLOCK_FILE_PHYS / EXT4_BLOCKS_PER_GROUP(sb)));
3920 db_count = (sbi->s_groups_count + EXT4_DESC_PER_BLOCK(sb) - 1) /
3921 EXT4_DESC_PER_BLOCK(sb);
3922 sbi->s_group_desc = ext4_kvmalloc(db_count *
3923 sizeof(struct buffer_head *),
3925 if (sbi->s_group_desc == NULL) {
3926 ext4_msg(sb, KERN_ERR, "not enough memory");
3932 sbi->s_proc = proc_mkdir(sb->s_id, ext4_proc_root);
3935 proc_create_data("options", S_IRUGO, sbi->s_proc,
3936 &ext4_seq_options_fops, sb);
3938 bgl_lock_init(sbi->s_blockgroup_lock);
3940 for (i = 0; i < db_count; i++) {
3941 block = descriptor_loc(sb, logical_sb_block, i);
3942 sbi->s_group_desc[i] = sb_bread_unmovable(sb, block);
3943 if (!sbi->s_group_desc[i]) {
3944 ext4_msg(sb, KERN_ERR,
3945 "can't read group descriptor %d", i);
3950 if (!ext4_check_descriptors(sb, &first_not_zeroed)) {
3951 ext4_msg(sb, KERN_ERR, "group descriptors corrupted!");
3955 sbi->s_gdb_count = db_count;
3956 get_random_bytes(&sbi->s_next_generation, sizeof(u32));
3957 spin_lock_init(&sbi->s_next_gen_lock);
3959 setup_timer(&sbi->s_err_report, print_daily_error_info,
3960 (unsigned long) sb);
3962 /* Register extent status tree shrinker */
3963 if (ext4_es_register_shrinker(sbi))
3966 sbi->s_stripe = ext4_get_stripe_size(sbi);
3967 sbi->s_extent_max_zeroout_kb = 32;
3970 * set up enough so that it can read an inode
3972 sb->s_op = &ext4_sops;
3973 sb->s_export_op = &ext4_export_ops;
3974 sb->s_xattr = ext4_xattr_handlers;
3976 sb->dq_op = &ext4_quota_operations;
3977 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_QUOTA))
3978 sb->s_qcop = &dquot_quotactl_sysfile_ops;
3980 sb->s_qcop = &ext4_qctl_operations;
3981 sb->s_quota_types = QTYPE_MASK_USR | QTYPE_MASK_GRP;
3983 memcpy(sb->s_uuid, es->s_uuid, sizeof(es->s_uuid));
3985 INIT_LIST_HEAD(&sbi->s_orphan); /* unlinked but open files */
3986 mutex_init(&sbi->s_orphan_lock);
3990 needs_recovery = (es->s_last_orphan != 0 ||
3991 EXT4_HAS_INCOMPAT_FEATURE(sb,
3992 EXT4_FEATURE_INCOMPAT_RECOVER));
3994 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_MMP) &&
3995 !(sb->s_flags & MS_RDONLY))
3996 if (ext4_multi_mount_protect(sb, le64_to_cpu(es->s_mmp_block)))
3997 goto failed_mount3a;
4000 * The first inode we look at is the journal inode. Don't try
4001 * root first: it may be modified in the journal!
4003 if (!test_opt(sb, NOLOAD) &&
4004 EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL)) {
4005 if (ext4_load_journal(sb, es, journal_devnum))
4006 goto failed_mount3a;
4007 } else if (test_opt(sb, NOLOAD) && !(sb->s_flags & MS_RDONLY) &&
4008 EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER)) {
4009 ext4_msg(sb, KERN_ERR, "required journal recovery "
4010 "suppressed and not mounted read-only");
4011 goto failed_mount_wq;
4013 clear_opt(sb, DATA_FLAGS);
4014 sbi->s_journal = NULL;
4019 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_64BIT) &&
4020 !jbd2_journal_set_features(EXT4_SB(sb)->s_journal, 0, 0,
4021 JBD2_FEATURE_INCOMPAT_64BIT)) {
4022 ext4_msg(sb, KERN_ERR, "Failed to set 64-bit journal feature");
4023 goto failed_mount_wq;
4026 if (!set_journal_csum_feature_set(sb)) {
4027 ext4_msg(sb, KERN_ERR, "Failed to set journal checksum "
4029 goto failed_mount_wq;
4032 /* We have now updated the journal if required, so we can
4033 * validate the data journaling mode. */
4034 switch (test_opt(sb, DATA_FLAGS)) {
4036 /* No mode set, assume a default based on the journal
4037 * capabilities: ORDERED_DATA if the journal can
4038 * cope, else JOURNAL_DATA
4040 if (jbd2_journal_check_available_features
4041 (sbi->s_journal, 0, 0, JBD2_FEATURE_INCOMPAT_REVOKE))
4042 set_opt(sb, ORDERED_DATA);
4044 set_opt(sb, JOURNAL_DATA);
4047 case EXT4_MOUNT_ORDERED_DATA:
4048 case EXT4_MOUNT_WRITEBACK_DATA:
4049 if (!jbd2_journal_check_available_features
4050 (sbi->s_journal, 0, 0, JBD2_FEATURE_INCOMPAT_REVOKE)) {
4051 ext4_msg(sb, KERN_ERR, "Journal does not support "
4052 "requested data journaling mode");
4053 goto failed_mount_wq;
4058 set_task_ioprio(sbi->s_journal->j_task, journal_ioprio);
4060 sbi->s_journal->j_commit_callback = ext4_journal_commit_callback;
4063 if (ext4_mballoc_ready) {
4064 sbi->s_mb_cache = ext4_xattr_create_cache(sb->s_id);
4065 if (!sbi->s_mb_cache) {
4066 ext4_msg(sb, KERN_ERR, "Failed to create an mb_cache");
4067 goto failed_mount_wq;
4071 if (unlikely(sbi->s_mount_flags & EXT4_MF_TEST_DUMMY_ENCRYPTION) &&
4072 !(sb->s_flags & MS_RDONLY) &&
4073 !EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_ENCRYPT)) {
4074 EXT4_SET_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_ENCRYPT);
4075 ext4_commit_super(sb, 1);
4079 * Get the # of file system overhead blocks from the
4080 * superblock if present.
4082 if (es->s_overhead_clusters)
4083 sbi->s_overhead = le32_to_cpu(es->s_overhead_clusters);
4085 err = ext4_calculate_overhead(sb);
4087 goto failed_mount_wq;
4091 * The maximum number of concurrent works can be high and
4092 * concurrency isn't really necessary. Limit it to 1.
4094 EXT4_SB(sb)->rsv_conversion_wq =
4095 alloc_workqueue("ext4-rsv-conversion", WQ_MEM_RECLAIM | WQ_UNBOUND, 1);
4096 if (!EXT4_SB(sb)->rsv_conversion_wq) {
4097 printk(KERN_ERR "EXT4-fs: failed to create workqueue\n");
4103 * The jbd2_journal_load will have done any necessary log recovery,
4104 * so we can safely mount the rest of the filesystem now.
4107 root = ext4_iget(sb, EXT4_ROOT_INO);
4109 ext4_msg(sb, KERN_ERR, "get root inode failed");
4110 ret = PTR_ERR(root);
4114 if (!S_ISDIR(root->i_mode) || !root->i_blocks || !root->i_size) {
4115 ext4_msg(sb, KERN_ERR, "corrupt root inode, run e2fsck");
4119 sb->s_root = d_make_root(root);
4121 ext4_msg(sb, KERN_ERR, "get root dentry failed");
4126 if (ext4_setup_super(sb, es, sb->s_flags & MS_RDONLY))
4127 sb->s_flags |= MS_RDONLY;
4129 /* determine the minimum size of new large inodes, if present */
4130 if (sbi->s_inode_size > EXT4_GOOD_OLD_INODE_SIZE) {
4131 sbi->s_want_extra_isize = sizeof(struct ext4_inode) -
4132 EXT4_GOOD_OLD_INODE_SIZE;
4133 if (EXT4_HAS_RO_COMPAT_FEATURE(sb,
4134 EXT4_FEATURE_RO_COMPAT_EXTRA_ISIZE)) {
4135 if (sbi->s_want_extra_isize <
4136 le16_to_cpu(es->s_want_extra_isize))
4137 sbi->s_want_extra_isize =
4138 le16_to_cpu(es->s_want_extra_isize);
4139 if (sbi->s_want_extra_isize <
4140 le16_to_cpu(es->s_min_extra_isize))
4141 sbi->s_want_extra_isize =
4142 le16_to_cpu(es->s_min_extra_isize);
4145 /* Check if enough inode space is available */
4146 if (EXT4_GOOD_OLD_INODE_SIZE + sbi->s_want_extra_isize >
4147 sbi->s_inode_size) {
4148 sbi->s_want_extra_isize = sizeof(struct ext4_inode) -
4149 EXT4_GOOD_OLD_INODE_SIZE;
4150 ext4_msg(sb, KERN_INFO, "required extra inode space not"
4154 err = ext4_reserve_clusters(sbi, ext4_calculate_resv_clusters(sb));
4156 ext4_msg(sb, KERN_ERR, "failed to reserve %llu clusters for "
4157 "reserved pool", ext4_calculate_resv_clusters(sb));
4158 goto failed_mount4a;
4161 err = ext4_setup_system_zone(sb);
4163 ext4_msg(sb, KERN_ERR, "failed to initialize system "
4165 goto failed_mount4a;
4169 err = ext4_mb_init(sb);
4171 ext4_msg(sb, KERN_ERR, "failed to initialize mballoc (%d)",
4176 block = ext4_count_free_clusters(sb);
4177 ext4_free_blocks_count_set(sbi->s_es,
4178 EXT4_C2B(sbi, block));
4179 err = percpu_counter_init(&sbi->s_freeclusters_counter, block,
4182 unsigned long freei = ext4_count_free_inodes(sb);
4183 sbi->s_es->s_free_inodes_count = cpu_to_le32(freei);
4184 err = percpu_counter_init(&sbi->s_freeinodes_counter, freei,
4188 err = percpu_counter_init(&sbi->s_dirs_counter,
4189 ext4_count_dirs(sb), GFP_KERNEL);
4191 err = percpu_counter_init(&sbi->s_dirtyclusters_counter, 0,
4194 ext4_msg(sb, KERN_ERR, "insufficient memory");
4198 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_FLEX_BG))
4199 if (!ext4_fill_flex_info(sb)) {
4200 ext4_msg(sb, KERN_ERR,
4201 "unable to initialize "
4202 "flex_bg meta info!");
4206 err = ext4_register_li_request(sb, first_not_zeroed);
4210 sbi->s_kobj.kset = ext4_kset;
4211 init_completion(&sbi->s_kobj_unregister);
4212 err = kobject_init_and_add(&sbi->s_kobj, &ext4_ktype, NULL,
4218 /* Enable quota usage during mount. */
4219 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_QUOTA) &&
4220 !(sb->s_flags & MS_RDONLY)) {
4221 err = ext4_enable_quotas(sb);
4225 #endif /* CONFIG_QUOTA */
4227 EXT4_SB(sb)->s_mount_state |= EXT4_ORPHAN_FS;
4228 ext4_orphan_cleanup(sb, es);
4229 EXT4_SB(sb)->s_mount_state &= ~EXT4_ORPHAN_FS;
4230 if (needs_recovery) {
4231 ext4_msg(sb, KERN_INFO, "recovery complete");
4232 ext4_mark_recovery_complete(sb, es);
4234 if (EXT4_SB(sb)->s_journal) {
4235 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA)
4236 descr = " journalled data mode";
4237 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA)
4238 descr = " ordered data mode";
4240 descr = " writeback data mode";
4242 descr = "out journal";
4244 if (test_opt(sb, DISCARD)) {
4245 struct request_queue *q = bdev_get_queue(sb->s_bdev);
4246 if (!blk_queue_discard(q))
4247 ext4_msg(sb, KERN_WARNING,
4248 "mounting with \"discard\" option, but "
4249 "the device does not support discard");
4252 ext4_msg(sb, KERN_INFO, "mounted filesystem with%s. "
4253 "Opts: %s%s%s", descr, sbi->s_es->s_mount_opts,
4254 *sbi->s_es->s_mount_opts ? "; " : "", orig_data);
4256 if (es->s_error_count)
4257 mod_timer(&sbi->s_err_report, jiffies + 300*HZ); /* 5 minutes */
4259 /* Enable message ratelimiting. Default is 10 messages per 5 secs. */
4260 ratelimit_state_init(&sbi->s_err_ratelimit_state, 5 * HZ, 10);
4261 ratelimit_state_init(&sbi->s_warning_ratelimit_state, 5 * HZ, 10);
4262 ratelimit_state_init(&sbi->s_msg_ratelimit_state, 5 * HZ, 10);
4269 ext4_msg(sb, KERN_ERR, "VFS: Can't find ext4 filesystem");
4274 kobject_del(&sbi->s_kobj);
4277 ext4_unregister_li_request(sb);
4279 ext4_mb_release(sb);
4280 if (sbi->s_flex_groups)
4281 kvfree(sbi->s_flex_groups);
4282 percpu_counter_destroy(&sbi->s_freeclusters_counter);
4283 percpu_counter_destroy(&sbi->s_freeinodes_counter);
4284 percpu_counter_destroy(&sbi->s_dirs_counter);
4285 percpu_counter_destroy(&sbi->s_dirtyclusters_counter);
4287 ext4_ext_release(sb);
4288 ext4_release_system_zone(sb);
4293 ext4_msg(sb, KERN_ERR, "mount failed");
4294 if (EXT4_SB(sb)->rsv_conversion_wq)
4295 destroy_workqueue(EXT4_SB(sb)->rsv_conversion_wq);
4297 if (sbi->s_journal) {
4298 jbd2_journal_destroy(sbi->s_journal);
4299 sbi->s_journal = NULL;
4302 ext4_es_unregister_shrinker(sbi);
4304 del_timer_sync(&sbi->s_err_report);
4306 kthread_stop(sbi->s_mmp_tsk);
4308 for (i = 0; i < db_count; i++)
4309 brelse(sbi->s_group_desc[i]);
4310 kvfree(sbi->s_group_desc);
4312 if (sbi->s_chksum_driver)
4313 crypto_free_shash(sbi->s_chksum_driver);
4315 remove_proc_entry("options", sbi->s_proc);
4316 remove_proc_entry(sb->s_id, ext4_proc_root);
4319 for (i = 0; i < EXT4_MAXQUOTAS; i++)
4320 kfree(sbi->s_qf_names[i]);
4322 ext4_blkdev_remove(sbi);
4325 sb->s_fs_info = NULL;
4326 kfree(sbi->s_blockgroup_lock);
4330 return err ? err : ret;
4334 * Setup any per-fs journal parameters now. We'll do this both on
4335 * initial mount, once the journal has been initialised but before we've
4336 * done any recovery; and again on any subsequent remount.
4338 static void ext4_init_journal_params(struct super_block *sb, journal_t *journal)
4340 struct ext4_sb_info *sbi = EXT4_SB(sb);
4342 journal->j_commit_interval = sbi->s_commit_interval;
4343 journal->j_min_batch_time = sbi->s_min_batch_time;
4344 journal->j_max_batch_time = sbi->s_max_batch_time;
4346 write_lock(&journal->j_state_lock);
4347 if (test_opt(sb, BARRIER))
4348 journal->j_flags |= JBD2_BARRIER;
4350 journal->j_flags &= ~JBD2_BARRIER;
4351 if (test_opt(sb, DATA_ERR_ABORT))
4352 journal->j_flags |= JBD2_ABORT_ON_SYNCDATA_ERR;
4354 journal->j_flags &= ~JBD2_ABORT_ON_SYNCDATA_ERR;
4355 write_unlock(&journal->j_state_lock);
4358 static journal_t *ext4_get_journal(struct super_block *sb,
4359 unsigned int journal_inum)
4361 struct inode *journal_inode;
4364 BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL));
4366 /* First, test for the existence of a valid inode on disk. Bad
4367 * things happen if we iget() an unused inode, as the subsequent
4368 * iput() will try to delete it. */
4370 journal_inode = ext4_iget(sb, journal_inum);
4371 if (IS_ERR(journal_inode)) {
4372 ext4_msg(sb, KERN_ERR, "no journal found");
4375 if (!journal_inode->i_nlink) {
4376 make_bad_inode(journal_inode);
4377 iput(journal_inode);
4378 ext4_msg(sb, KERN_ERR, "journal inode is deleted");
4382 jbd_debug(2, "Journal inode found at %p: %lld bytes\n",
4383 journal_inode, journal_inode->i_size);
4384 if (!S_ISREG(journal_inode->i_mode)) {
4385 ext4_msg(sb, KERN_ERR, "invalid journal inode");
4386 iput(journal_inode);
4390 journal = jbd2_journal_init_inode(journal_inode);
4392 ext4_msg(sb, KERN_ERR, "Could not load journal inode");
4393 iput(journal_inode);
4396 journal->j_private = sb;
4397 ext4_init_journal_params(sb, journal);
4401 static journal_t *ext4_get_dev_journal(struct super_block *sb,
4404 struct buffer_head *bh;
4408 int hblock, blocksize;
4409 ext4_fsblk_t sb_block;
4410 unsigned long offset;
4411 struct ext4_super_block *es;
4412 struct block_device *bdev;
4414 BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL));
4416 bdev = ext4_blkdev_get(j_dev, sb);
4420 blocksize = sb->s_blocksize;
4421 hblock = bdev_logical_block_size(bdev);
4422 if (blocksize < hblock) {
4423 ext4_msg(sb, KERN_ERR,
4424 "blocksize too small for journal device");
4428 sb_block = EXT4_MIN_BLOCK_SIZE / blocksize;
4429 offset = EXT4_MIN_BLOCK_SIZE % blocksize;
4430 set_blocksize(bdev, blocksize);
4431 if (!(bh = __bread(bdev, sb_block, blocksize))) {
4432 ext4_msg(sb, KERN_ERR, "couldn't read superblock of "
4433 "external journal");
4437 es = (struct ext4_super_block *) (bh->b_data + offset);
4438 if ((le16_to_cpu(es->s_magic) != EXT4_SUPER_MAGIC) ||
4439 !(le32_to_cpu(es->s_feature_incompat) &
4440 EXT4_FEATURE_INCOMPAT_JOURNAL_DEV)) {
4441 ext4_msg(sb, KERN_ERR, "external journal has "
4447 if ((le32_to_cpu(es->s_feature_ro_compat) &
4448 EXT4_FEATURE_RO_COMPAT_METADATA_CSUM) &&
4449 es->s_checksum != ext4_superblock_csum(sb, es)) {
4450 ext4_msg(sb, KERN_ERR, "external journal has "
4451 "corrupt superblock");
4456 if (memcmp(EXT4_SB(sb)->s_es->s_journal_uuid, es->s_uuid, 16)) {
4457 ext4_msg(sb, KERN_ERR, "journal UUID does not match");
4462 len = ext4_blocks_count(es);
4463 start = sb_block + 1;
4464 brelse(bh); /* we're done with the superblock */
4466 journal = jbd2_journal_init_dev(bdev, sb->s_bdev,
4467 start, len, blocksize);
4469 ext4_msg(sb, KERN_ERR, "failed to create device journal");
4472 journal->j_private = sb;
4473 ll_rw_block(READ | REQ_META | REQ_PRIO, 1, &journal->j_sb_buffer);
4474 wait_on_buffer(journal->j_sb_buffer);
4475 if (!buffer_uptodate(journal->j_sb_buffer)) {
4476 ext4_msg(sb, KERN_ERR, "I/O error on journal device");
4479 if (be32_to_cpu(journal->j_superblock->s_nr_users) != 1) {
4480 ext4_msg(sb, KERN_ERR, "External journal has more than one "
4481 "user (unsupported) - %d",
4482 be32_to_cpu(journal->j_superblock->s_nr_users));
4485 EXT4_SB(sb)->journal_bdev = bdev;
4486 ext4_init_journal_params(sb, journal);
4490 jbd2_journal_destroy(journal);
4492 ext4_blkdev_put(bdev);
4496 static int ext4_load_journal(struct super_block *sb,
4497 struct ext4_super_block *es,
4498 unsigned long journal_devnum)
4501 unsigned int journal_inum = le32_to_cpu(es->s_journal_inum);
4504 int really_read_only;
4506 BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL));
4508 if (journal_devnum &&
4509 journal_devnum != le32_to_cpu(es->s_journal_dev)) {
4510 ext4_msg(sb, KERN_INFO, "external journal device major/minor "
4511 "numbers have changed");
4512 journal_dev = new_decode_dev(journal_devnum);
4514 journal_dev = new_decode_dev(le32_to_cpu(es->s_journal_dev));
4516 really_read_only = bdev_read_only(sb->s_bdev);
4519 * Are we loading a blank journal or performing recovery after a
4520 * crash? For recovery, we need to check in advance whether we
4521 * can get read-write access to the device.
4523 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER)) {
4524 if (sb->s_flags & MS_RDONLY) {
4525 ext4_msg(sb, KERN_INFO, "INFO: recovery "
4526 "required on readonly filesystem");
4527 if (really_read_only) {
4528 ext4_msg(sb, KERN_ERR, "write access "
4529 "unavailable, cannot proceed");
4532 ext4_msg(sb, KERN_INFO, "write access will "
4533 "be enabled during recovery");
4537 if (journal_inum && journal_dev) {
4538 ext4_msg(sb, KERN_ERR, "filesystem has both journal "
4539 "and inode journals!");
4544 if (!(journal = ext4_get_journal(sb, journal_inum)))
4547 if (!(journal = ext4_get_dev_journal(sb, journal_dev)))
4551 if (!(journal->j_flags & JBD2_BARRIER))
4552 ext4_msg(sb, KERN_INFO, "barriers disabled");
4554 if (!EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER))
4555 err = jbd2_journal_wipe(journal, !really_read_only);
4557 char *save = kmalloc(EXT4_S_ERR_LEN, GFP_KERNEL);
4559 memcpy(save, ((char *) es) +
4560 EXT4_S_ERR_START, EXT4_S_ERR_LEN);
4561 err = jbd2_journal_load(journal);
4563 memcpy(((char *) es) + EXT4_S_ERR_START,
4564 save, EXT4_S_ERR_LEN);
4569 ext4_msg(sb, KERN_ERR, "error loading journal");
4570 jbd2_journal_destroy(journal);
4574 EXT4_SB(sb)->s_journal = journal;
4575 ext4_clear_journal_err(sb, es);
4577 if (!really_read_only && journal_devnum &&
4578 journal_devnum != le32_to_cpu(es->s_journal_dev)) {
4579 es->s_journal_dev = cpu_to_le32(journal_devnum);
4581 /* Make sure we flush the recovery flag to disk. */
4582 ext4_commit_super(sb, 1);
4588 static int ext4_commit_super(struct super_block *sb, int sync)
4590 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
4591 struct buffer_head *sbh = EXT4_SB(sb)->s_sbh;
4596 if (buffer_write_io_error(sbh)) {
4598 * Oh, dear. A previous attempt to write the
4599 * superblock failed. This could happen because the
4600 * USB device was yanked out. Or it could happen to
4601 * be a transient write error and maybe the block will
4602 * be remapped. Nothing we can do but to retry the
4603 * write and hope for the best.
4605 ext4_msg(sb, KERN_ERR, "previous I/O error to "
4606 "superblock detected");
4607 clear_buffer_write_io_error(sbh);
4608 set_buffer_uptodate(sbh);
4611 * If the file system is mounted read-only, don't update the
4612 * superblock write time. This avoids updating the superblock
4613 * write time when we are mounting the root file system
4614 * read/only but we need to replay the journal; at that point,
4615 * for people who are east of GMT and who make their clock
4616 * tick in localtime for Windows bug-for-bug compatibility,
4617 * the clock is set in the future, and this will cause e2fsck
4618 * to complain and force a full file system check.
4620 if (!(sb->s_flags & MS_RDONLY))
4621 es->s_wtime = cpu_to_le32(get_seconds());
4622 if (sb->s_bdev->bd_part)
4623 es->s_kbytes_written =
4624 cpu_to_le64(EXT4_SB(sb)->s_kbytes_written +
4625 ((part_stat_read(sb->s_bdev->bd_part, sectors[1]) -
4626 EXT4_SB(sb)->s_sectors_written_start) >> 1));
4628 es->s_kbytes_written =
4629 cpu_to_le64(EXT4_SB(sb)->s_kbytes_written);
4630 if (percpu_counter_initialized(&EXT4_SB(sb)->s_freeclusters_counter))
4631 ext4_free_blocks_count_set(es,
4632 EXT4_C2B(EXT4_SB(sb), percpu_counter_sum_positive(
4633 &EXT4_SB(sb)->s_freeclusters_counter)));
4634 if (percpu_counter_initialized(&EXT4_SB(sb)->s_freeinodes_counter))
4635 es->s_free_inodes_count =
4636 cpu_to_le32(percpu_counter_sum_positive(
4637 &EXT4_SB(sb)->s_freeinodes_counter));
4638 BUFFER_TRACE(sbh, "marking dirty");
4639 ext4_superblock_csum_set(sb);
4640 mark_buffer_dirty(sbh);
4642 error = sync_dirty_buffer(sbh);
4646 error = buffer_write_io_error(sbh);
4648 ext4_msg(sb, KERN_ERR, "I/O error while writing "
4650 clear_buffer_write_io_error(sbh);
4651 set_buffer_uptodate(sbh);
4658 * Have we just finished recovery? If so, and if we are mounting (or
4659 * remounting) the filesystem readonly, then we will end up with a
4660 * consistent fs on disk. Record that fact.
4662 static void ext4_mark_recovery_complete(struct super_block *sb,
4663 struct ext4_super_block *es)
4665 journal_t *journal = EXT4_SB(sb)->s_journal;
4667 if (!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL)) {
4668 BUG_ON(journal != NULL);
4671 jbd2_journal_lock_updates(journal);
4672 if (jbd2_journal_flush(journal) < 0)
4675 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER) &&
4676 sb->s_flags & MS_RDONLY) {
4677 EXT4_CLEAR_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
4678 ext4_commit_super(sb, 1);
4682 jbd2_journal_unlock_updates(journal);
4686 * If we are mounting (or read-write remounting) a filesystem whose journal
4687 * has recorded an error from a previous lifetime, move that error to the
4688 * main filesystem now.
4690 static void ext4_clear_journal_err(struct super_block *sb,
4691 struct ext4_super_block *es)
4697 BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL));
4699 journal = EXT4_SB(sb)->s_journal;
4702 * Now check for any error status which may have been recorded in the
4703 * journal by a prior ext4_error() or ext4_abort()
4706 j_errno = jbd2_journal_errno(journal);
4710 errstr = ext4_decode_error(sb, j_errno, nbuf);
4711 ext4_warning(sb, "Filesystem error recorded "
4712 "from previous mount: %s", errstr);
4713 ext4_warning(sb, "Marking fs in need of filesystem check.");
4715 EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS;
4716 es->s_state |= cpu_to_le16(EXT4_ERROR_FS);
4717 ext4_commit_super(sb, 1);
4719 jbd2_journal_clear_err(journal);
4720 jbd2_journal_update_sb_errno(journal);
4725 * Force the running and committing transactions to commit,
4726 * and wait on the commit.
4728 int ext4_force_commit(struct super_block *sb)
4732 if (sb->s_flags & MS_RDONLY)
4735 journal = EXT4_SB(sb)->s_journal;
4736 return ext4_journal_force_commit(journal);
4739 static int ext4_sync_fs(struct super_block *sb, int wait)
4743 bool needs_barrier = false;
4744 struct ext4_sb_info *sbi = EXT4_SB(sb);
4746 trace_ext4_sync_fs(sb, wait);
4747 flush_workqueue(sbi->rsv_conversion_wq);
4749 * Writeback quota in non-journalled quota case - journalled quota has
4752 dquot_writeback_dquots(sb, -1);
4754 * Data writeback is possible w/o journal transaction, so barrier must
4755 * being sent at the end of the function. But we can skip it if
4756 * transaction_commit will do it for us.
4758 if (sbi->s_journal) {
4759 target = jbd2_get_latest_transaction(sbi->s_journal);
4760 if (wait && sbi->s_journal->j_flags & JBD2_BARRIER &&
4761 !jbd2_trans_will_send_data_barrier(sbi->s_journal, target))
4762 needs_barrier = true;
4764 if (jbd2_journal_start_commit(sbi->s_journal, &target)) {
4766 ret = jbd2_log_wait_commit(sbi->s_journal,
4769 } else if (wait && test_opt(sb, BARRIER))
4770 needs_barrier = true;
4771 if (needs_barrier) {
4773 err = blkdev_issue_flush(sb->s_bdev, GFP_KERNEL, NULL);
4782 * LVM calls this function before a (read-only) snapshot is created. This
4783 * gives us a chance to flush the journal completely and mark the fs clean.
4785 * Note that only this function cannot bring a filesystem to be in a clean
4786 * state independently. It relies on upper layer to stop all data & metadata
4789 static int ext4_freeze(struct super_block *sb)
4794 if (sb->s_flags & MS_RDONLY)
4797 journal = EXT4_SB(sb)->s_journal;
4800 /* Now we set up the journal barrier. */
4801 jbd2_journal_lock_updates(journal);
4804 * Don't clear the needs_recovery flag if we failed to
4805 * flush the journal.
4807 error = jbd2_journal_flush(journal);
4812 /* Journal blocked and flushed, clear needs_recovery flag. */
4813 EXT4_CLEAR_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
4814 error = ext4_commit_super(sb, 1);
4817 /* we rely on upper layer to stop further updates */
4818 jbd2_journal_unlock_updates(journal);
4823 * Called by LVM after the snapshot is done. We need to reset the RECOVER
4824 * flag here, even though the filesystem is not technically dirty yet.
4826 static int ext4_unfreeze(struct super_block *sb)
4828 if (sb->s_flags & MS_RDONLY)
4831 /* Reset the needs_recovery flag before the fs is unlocked. */
4832 EXT4_SET_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
4833 ext4_commit_super(sb, 1);
4838 * Structure to save mount options for ext4_remount's benefit
4840 struct ext4_mount_options {
4841 unsigned long s_mount_opt;
4842 unsigned long s_mount_opt2;
4845 unsigned long s_commit_interval;
4846 u32 s_min_batch_time, s_max_batch_time;
4849 char *s_qf_names[EXT4_MAXQUOTAS];
4853 static int ext4_remount(struct super_block *sb, int *flags, char *data)
4855 struct ext4_super_block *es;
4856 struct ext4_sb_info *sbi = EXT4_SB(sb);
4857 unsigned long old_sb_flags;
4858 struct ext4_mount_options old_opts;
4859 int enable_quota = 0;
4861 unsigned int journal_ioprio = DEFAULT_JOURNAL_IOPRIO;
4866 char *orig_data = kstrdup(data, GFP_KERNEL);
4868 /* Store the original options */
4869 old_sb_flags = sb->s_flags;
4870 old_opts.s_mount_opt = sbi->s_mount_opt;
4871 old_opts.s_mount_opt2 = sbi->s_mount_opt2;
4872 old_opts.s_resuid = sbi->s_resuid;
4873 old_opts.s_resgid = sbi->s_resgid;
4874 old_opts.s_commit_interval = sbi->s_commit_interval;
4875 old_opts.s_min_batch_time = sbi->s_min_batch_time;
4876 old_opts.s_max_batch_time = sbi->s_max_batch_time;
4878 old_opts.s_jquota_fmt = sbi->s_jquota_fmt;
4879 for (i = 0; i < EXT4_MAXQUOTAS; i++)
4880 if (sbi->s_qf_names[i]) {
4881 old_opts.s_qf_names[i] = kstrdup(sbi->s_qf_names[i],
4883 if (!old_opts.s_qf_names[i]) {
4884 for (j = 0; j < i; j++)
4885 kfree(old_opts.s_qf_names[j]);
4890 old_opts.s_qf_names[i] = NULL;
4892 if (sbi->s_journal && sbi->s_journal->j_task->io_context)
4893 journal_ioprio = sbi->s_journal->j_task->io_context->ioprio;
4895 if (!parse_options(data, sb, NULL, &journal_ioprio, 1)) {
4900 if ((old_opts.s_mount_opt & EXT4_MOUNT_JOURNAL_CHECKSUM) ^
4901 test_opt(sb, JOURNAL_CHECKSUM)) {
4902 ext4_msg(sb, KERN_ERR, "changing journal_checksum "
4903 "during remount not supported; ignoring");
4904 sbi->s_mount_opt ^= EXT4_MOUNT_JOURNAL_CHECKSUM;
4907 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA) {
4908 if (test_opt2(sb, EXPLICIT_DELALLOC)) {
4909 ext4_msg(sb, KERN_ERR, "can't mount with "
4910 "both data=journal and delalloc");
4914 if (test_opt(sb, DIOREAD_NOLOCK)) {
4915 ext4_msg(sb, KERN_ERR, "can't mount with "
4916 "both data=journal and dioread_nolock");
4920 if (test_opt(sb, DAX)) {
4921 ext4_msg(sb, KERN_ERR, "can't mount with "
4922 "both data=journal and dax");
4928 if ((sbi->s_mount_opt ^ old_opts.s_mount_opt) & EXT4_MOUNT_DAX) {
4929 ext4_msg(sb, KERN_WARNING, "warning: refusing change of "
4930 "dax flag with busy inodes while remounting");
4931 sbi->s_mount_opt ^= EXT4_MOUNT_DAX;
4934 if (sbi->s_mount_flags & EXT4_MF_FS_ABORTED)
4935 ext4_abort(sb, "Abort forced by user");
4937 sb->s_flags = (sb->s_flags & ~MS_POSIXACL) |
4938 (test_opt(sb, POSIX_ACL) ? MS_POSIXACL : 0);
4942 if (sbi->s_journal) {
4943 ext4_init_journal_params(sb, sbi->s_journal);
4944 set_task_ioprio(sbi->s_journal->j_task, journal_ioprio);
4947 if (*flags & MS_LAZYTIME)
4948 sb->s_flags |= MS_LAZYTIME;
4950 if ((*flags & MS_RDONLY) != (sb->s_flags & MS_RDONLY)) {
4951 if (sbi->s_mount_flags & EXT4_MF_FS_ABORTED) {
4956 if (*flags & MS_RDONLY) {
4957 err = sync_filesystem(sb);
4960 err = dquot_suspend(sb, -1);
4965 * First of all, the unconditional stuff we have to do
4966 * to disable replay of the journal when we next remount
4968 sb->s_flags |= MS_RDONLY;
4971 * OK, test if we are remounting a valid rw partition
4972 * readonly, and if so set the rdonly flag and then
4973 * mark the partition as valid again.
4975 if (!(es->s_state & cpu_to_le16(EXT4_VALID_FS)) &&
4976 (sbi->s_mount_state & EXT4_VALID_FS))
4977 es->s_state = cpu_to_le16(sbi->s_mount_state);
4980 ext4_mark_recovery_complete(sb, es);
4982 /* Make sure we can mount this feature set readwrite */
4983 if (EXT4_HAS_RO_COMPAT_FEATURE(sb,
4984 EXT4_FEATURE_RO_COMPAT_READONLY) ||
4985 !ext4_feature_set_ok(sb, 0)) {
4990 * Make sure the group descriptor checksums
4991 * are sane. If they aren't, refuse to remount r/w.
4993 for (g = 0; g < sbi->s_groups_count; g++) {
4994 struct ext4_group_desc *gdp =
4995 ext4_get_group_desc(sb, g, NULL);
4997 if (!ext4_group_desc_csum_verify(sb, g, gdp)) {
4998 ext4_msg(sb, KERN_ERR,
4999 "ext4_remount: Checksum for group %u failed (%u!=%u)",
5000 g, le16_to_cpu(ext4_group_desc_csum(sbi, g, gdp)),
5001 le16_to_cpu(gdp->bg_checksum));
5008 * If we have an unprocessed orphan list hanging
5009 * around from a previously readonly bdev mount,
5010 * require a full umount/remount for now.
5012 if (es->s_last_orphan) {
5013 ext4_msg(sb, KERN_WARNING, "Couldn't "
5014 "remount RDWR because of unprocessed "
5015 "orphan inode list. Please "
5016 "umount/remount instead");
5022 * Mounting a RDONLY partition read-write, so reread
5023 * and store the current valid flag. (It may have
5024 * been changed by e2fsck since we originally mounted
5028 ext4_clear_journal_err(sb, es);
5029 sbi->s_mount_state = le16_to_cpu(es->s_state);
5030 if (!ext4_setup_super(sb, es, 0))
5031 sb->s_flags &= ~MS_RDONLY;
5032 if (EXT4_HAS_INCOMPAT_FEATURE(sb,
5033 EXT4_FEATURE_INCOMPAT_MMP))
5034 if (ext4_multi_mount_protect(sb,
5035 le64_to_cpu(es->s_mmp_block))) {
5044 * Reinitialize lazy itable initialization thread based on
5047 if ((sb->s_flags & MS_RDONLY) || !test_opt(sb, INIT_INODE_TABLE))
5048 ext4_unregister_li_request(sb);
5050 ext4_group_t first_not_zeroed;
5051 first_not_zeroed = ext4_has_uninit_itable(sb);
5052 ext4_register_li_request(sb, first_not_zeroed);
5055 ext4_setup_system_zone(sb);
5056 if (sbi->s_journal == NULL && !(old_sb_flags & MS_RDONLY))
5057 ext4_commit_super(sb, 1);
5060 /* Release old quota file names */
5061 for (i = 0; i < EXT4_MAXQUOTAS; i++)
5062 kfree(old_opts.s_qf_names[i]);
5064 if (sb_any_quota_suspended(sb))
5065 dquot_resume(sb, -1);
5066 else if (EXT4_HAS_RO_COMPAT_FEATURE(sb,
5067 EXT4_FEATURE_RO_COMPAT_QUOTA)) {
5068 err = ext4_enable_quotas(sb);
5075 *flags = (*flags & ~MS_LAZYTIME) | (sb->s_flags & MS_LAZYTIME);
5076 ext4_msg(sb, KERN_INFO, "re-mounted. Opts: %s", orig_data);
5081 sb->s_flags = old_sb_flags;
5082 sbi->s_mount_opt = old_opts.s_mount_opt;
5083 sbi->s_mount_opt2 = old_opts.s_mount_opt2;
5084 sbi->s_resuid = old_opts.s_resuid;
5085 sbi->s_resgid = old_opts.s_resgid;
5086 sbi->s_commit_interval = old_opts.s_commit_interval;
5087 sbi->s_min_batch_time = old_opts.s_min_batch_time;
5088 sbi->s_max_batch_time = old_opts.s_max_batch_time;
5090 sbi->s_jquota_fmt = old_opts.s_jquota_fmt;
5091 for (i = 0; i < EXT4_MAXQUOTAS; i++) {
5092 kfree(sbi->s_qf_names[i]);
5093 sbi->s_qf_names[i] = old_opts.s_qf_names[i];
5100 static int ext4_statfs(struct dentry *dentry, struct kstatfs *buf)
5102 struct super_block *sb = dentry->d_sb;
5103 struct ext4_sb_info *sbi = EXT4_SB(sb);
5104 struct ext4_super_block *es = sbi->s_es;
5105 ext4_fsblk_t overhead = 0, resv_blocks;
5108 resv_blocks = EXT4_C2B(sbi, atomic64_read(&sbi->s_resv_clusters));
5110 if (!test_opt(sb, MINIX_DF))
5111 overhead = sbi->s_overhead;
5113 buf->f_type = EXT4_SUPER_MAGIC;
5114 buf->f_bsize = sb->s_blocksize;
5115 buf->f_blocks = ext4_blocks_count(es) - EXT4_C2B(sbi, overhead);
5116 bfree = percpu_counter_sum_positive(&sbi->s_freeclusters_counter) -
5117 percpu_counter_sum_positive(&sbi->s_dirtyclusters_counter);
5118 /* prevent underflow in case that few free space is available */
5119 buf->f_bfree = EXT4_C2B(sbi, max_t(s64, bfree, 0));
5120 buf->f_bavail = buf->f_bfree -
5121 (ext4_r_blocks_count(es) + resv_blocks);
5122 if (buf->f_bfree < (ext4_r_blocks_count(es) + resv_blocks))
5124 buf->f_files = le32_to_cpu(es->s_inodes_count);
5125 buf->f_ffree = percpu_counter_sum_positive(&sbi->s_freeinodes_counter);
5126 buf->f_namelen = EXT4_NAME_LEN;
5127 fsid = le64_to_cpup((void *)es->s_uuid) ^
5128 le64_to_cpup((void *)es->s_uuid + sizeof(u64));
5129 buf->f_fsid.val[0] = fsid & 0xFFFFFFFFUL;
5130 buf->f_fsid.val[1] = (fsid >> 32) & 0xFFFFFFFFUL;
5135 /* Helper function for writing quotas on sync - we need to start transaction
5136 * before quota file is locked for write. Otherwise the are possible deadlocks:
5137 * Process 1 Process 2
5138 * ext4_create() quota_sync()
5139 * jbd2_journal_start() write_dquot()
5140 * dquot_initialize() down(dqio_mutex)
5141 * down(dqio_mutex) jbd2_journal_start()
5147 static inline struct inode *dquot_to_inode(struct dquot *dquot)
5149 return sb_dqopt(dquot->dq_sb)->files[dquot->dq_id.type];
5152 static int ext4_write_dquot(struct dquot *dquot)
5156 struct inode *inode;
5158 inode = dquot_to_inode(dquot);
5159 handle = ext4_journal_start(inode, EXT4_HT_QUOTA,
5160 EXT4_QUOTA_TRANS_BLOCKS(dquot->dq_sb));
5162 return PTR_ERR(handle);
5163 ret = dquot_commit(dquot);
5164 err = ext4_journal_stop(handle);
5170 static int ext4_acquire_dquot(struct dquot *dquot)
5175 handle = ext4_journal_start(dquot_to_inode(dquot), EXT4_HT_QUOTA,
5176 EXT4_QUOTA_INIT_BLOCKS(dquot->dq_sb));
5178 return PTR_ERR(handle);
5179 ret = dquot_acquire(dquot);
5180 err = ext4_journal_stop(handle);
5186 static int ext4_release_dquot(struct dquot *dquot)
5191 handle = ext4_journal_start(dquot_to_inode(dquot), EXT4_HT_QUOTA,
5192 EXT4_QUOTA_DEL_BLOCKS(dquot->dq_sb));
5193 if (IS_ERR(handle)) {
5194 /* Release dquot anyway to avoid endless cycle in dqput() */
5195 dquot_release(dquot);
5196 return PTR_ERR(handle);
5198 ret = dquot_release(dquot);
5199 err = ext4_journal_stop(handle);
5205 static int ext4_mark_dquot_dirty(struct dquot *dquot)
5207 struct super_block *sb = dquot->dq_sb;
5208 struct ext4_sb_info *sbi = EXT4_SB(sb);
5210 /* Are we journaling quotas? */
5211 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_QUOTA) ||
5212 sbi->s_qf_names[USRQUOTA] || sbi->s_qf_names[GRPQUOTA]) {
5213 dquot_mark_dquot_dirty(dquot);
5214 return ext4_write_dquot(dquot);
5216 return dquot_mark_dquot_dirty(dquot);
5220 static int ext4_write_info(struct super_block *sb, int type)
5225 /* Data block + inode block */
5226 handle = ext4_journal_start(d_inode(sb->s_root), EXT4_HT_QUOTA, 2);
5228 return PTR_ERR(handle);
5229 ret = dquot_commit_info(sb, type);
5230 err = ext4_journal_stop(handle);
5237 * Turn on quotas during mount time - we need to find
5238 * the quota file and such...
5240 static int ext4_quota_on_mount(struct super_block *sb, int type)
5242 return dquot_quota_on_mount(sb, EXT4_SB(sb)->s_qf_names[type],
5243 EXT4_SB(sb)->s_jquota_fmt, type);
5247 * Standard function to be called on quota_on
5249 static int ext4_quota_on(struct super_block *sb, int type, int format_id,
5254 if (!test_opt(sb, QUOTA))
5257 /* Quotafile not on the same filesystem? */
5258 if (path->dentry->d_sb != sb)
5260 /* Journaling quota? */
5261 if (EXT4_SB(sb)->s_qf_names[type]) {
5262 /* Quotafile not in fs root? */
5263 if (path->dentry->d_parent != sb->s_root)
5264 ext4_msg(sb, KERN_WARNING,
5265 "Quota file not on filesystem root. "
5266 "Journaled quota will not work");
5270 * When we journal data on quota file, we have to flush journal to see
5271 * all updates to the file when we bypass pagecache...
5273 if (EXT4_SB(sb)->s_journal &&
5274 ext4_should_journal_data(d_inode(path->dentry))) {
5276 * We don't need to lock updates but journal_flush() could
5277 * otherwise be livelocked...
5279 jbd2_journal_lock_updates(EXT4_SB(sb)->s_journal);
5280 err = jbd2_journal_flush(EXT4_SB(sb)->s_journal);
5281 jbd2_journal_unlock_updates(EXT4_SB(sb)->s_journal);
5286 return dquot_quota_on(sb, type, format_id, path);
5289 static int ext4_quota_enable(struct super_block *sb, int type, int format_id,
5293 struct inode *qf_inode;
5294 unsigned long qf_inums[EXT4_MAXQUOTAS] = {
5295 le32_to_cpu(EXT4_SB(sb)->s_es->s_usr_quota_inum),
5296 le32_to_cpu(EXT4_SB(sb)->s_es->s_grp_quota_inum)
5299 BUG_ON(!EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_QUOTA));
5301 if (!qf_inums[type])
5304 qf_inode = ext4_iget(sb, qf_inums[type]);
5305 if (IS_ERR(qf_inode)) {
5306 ext4_error(sb, "Bad quota inode # %lu", qf_inums[type]);
5307 return PTR_ERR(qf_inode);
5310 /* Don't account quota for quota files to avoid recursion */
5311 qf_inode->i_flags |= S_NOQUOTA;
5312 err = dquot_enable(qf_inode, type, format_id, flags);
5318 /* Enable usage tracking for all quota types. */
5319 static int ext4_enable_quotas(struct super_block *sb)
5322 unsigned long qf_inums[EXT4_MAXQUOTAS] = {
5323 le32_to_cpu(EXT4_SB(sb)->s_es->s_usr_quota_inum),
5324 le32_to_cpu(EXT4_SB(sb)->s_es->s_grp_quota_inum)
5327 sb_dqopt(sb)->flags |= DQUOT_QUOTA_SYS_FILE;
5328 for (type = 0; type < EXT4_MAXQUOTAS; type++) {
5329 if (qf_inums[type]) {
5330 err = ext4_quota_enable(sb, type, QFMT_VFS_V1,
5331 DQUOT_USAGE_ENABLED);
5334 "Failed to enable quota tracking "
5335 "(type=%d, err=%d). Please run "
5336 "e2fsck to fix.", type, err);
5344 static int ext4_quota_off(struct super_block *sb, int type)
5346 struct inode *inode = sb_dqopt(sb)->files[type];
5349 /* Force all delayed allocation blocks to be allocated.
5350 * Caller already holds s_umount sem */
5351 if (test_opt(sb, DELALLOC))
5352 sync_filesystem(sb);
5357 /* Update modification times of quota files when userspace can
5358 * start looking at them */
5359 handle = ext4_journal_start(inode, EXT4_HT_QUOTA, 1);
5362 inode->i_mtime = inode->i_ctime = CURRENT_TIME;
5363 ext4_mark_inode_dirty(handle, inode);
5364 ext4_journal_stop(handle);
5367 return dquot_quota_off(sb, type);
5370 /* Read data from quotafile - avoid pagecache and such because we cannot afford
5371 * acquiring the locks... As quota files are never truncated and quota code
5372 * itself serializes the operations (and no one else should touch the files)
5373 * we don't have to be afraid of races */
5374 static ssize_t ext4_quota_read(struct super_block *sb, int type, char *data,
5375 size_t len, loff_t off)
5377 struct inode *inode = sb_dqopt(sb)->files[type];
5378 ext4_lblk_t blk = off >> EXT4_BLOCK_SIZE_BITS(sb);
5379 int offset = off & (sb->s_blocksize - 1);
5382 struct buffer_head *bh;
5383 loff_t i_size = i_size_read(inode);
5387 if (off+len > i_size)
5390 while (toread > 0) {
5391 tocopy = sb->s_blocksize - offset < toread ?
5392 sb->s_blocksize - offset : toread;
5393 bh = ext4_bread(NULL, inode, blk, 0);
5396 if (!bh) /* A hole? */
5397 memset(data, 0, tocopy);
5399 memcpy(data, bh->b_data+offset, tocopy);
5409 /* Write to quotafile (we know the transaction is already started and has
5410 * enough credits) */
5411 static ssize_t ext4_quota_write(struct super_block *sb, int type,
5412 const char *data, size_t len, loff_t off)
5414 struct inode *inode = sb_dqopt(sb)->files[type];
5415 ext4_lblk_t blk = off >> EXT4_BLOCK_SIZE_BITS(sb);
5416 int err, offset = off & (sb->s_blocksize - 1);
5417 struct buffer_head *bh;
5418 handle_t *handle = journal_current_handle();
5420 if (EXT4_SB(sb)->s_journal && !handle) {
5421 ext4_msg(sb, KERN_WARNING, "Quota write (off=%llu, len=%llu)"
5422 " cancelled because transaction is not started",
5423 (unsigned long long)off, (unsigned long long)len);
5427 * Since we account only one data block in transaction credits,
5428 * then it is impossible to cross a block boundary.
5430 if (sb->s_blocksize - offset < len) {
5431 ext4_msg(sb, KERN_WARNING, "Quota write (off=%llu, len=%llu)"
5432 " cancelled because not block aligned",
5433 (unsigned long long)off, (unsigned long long)len);
5437 bh = ext4_bread(handle, inode, blk, 1);
5442 BUFFER_TRACE(bh, "get write access");
5443 err = ext4_journal_get_write_access(handle, bh);
5449 memcpy(bh->b_data+offset, data, len);
5450 flush_dcache_page(bh->b_page);
5452 err = ext4_handle_dirty_metadata(handle, NULL, bh);
5455 if (inode->i_size < off + len) {
5456 i_size_write(inode, off + len);
5457 EXT4_I(inode)->i_disksize = inode->i_size;
5458 ext4_mark_inode_dirty(handle, inode);
5465 static struct dentry *ext4_mount(struct file_system_type *fs_type, int flags,
5466 const char *dev_name, void *data)
5468 return mount_bdev(fs_type, flags, dev_name, data, ext4_fill_super);
5471 #if !defined(CONFIG_EXT2_FS) && !defined(CONFIG_EXT2_FS_MODULE) && defined(CONFIG_EXT4_USE_FOR_EXT23)
5472 static inline void register_as_ext2(void)
5474 int err = register_filesystem(&ext2_fs_type);
5477 "EXT4-fs: Unable to register as ext2 (%d)\n", err);
5480 static inline void unregister_as_ext2(void)
5482 unregister_filesystem(&ext2_fs_type);
5485 static inline int ext2_feature_set_ok(struct super_block *sb)
5487 if (EXT4_HAS_INCOMPAT_FEATURE(sb, ~EXT2_FEATURE_INCOMPAT_SUPP))
5489 if (sb->s_flags & MS_RDONLY)
5491 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, ~EXT2_FEATURE_RO_COMPAT_SUPP))
5496 static inline void register_as_ext2(void) { }
5497 static inline void unregister_as_ext2(void) { }
5498 static inline int ext2_feature_set_ok(struct super_block *sb) { return 0; }
5501 #if !defined(CONFIG_EXT3_FS) && !defined(CONFIG_EXT3_FS_MODULE) && defined(CONFIG_EXT4_USE_FOR_EXT23)
5502 static inline void register_as_ext3(void)
5504 int err = register_filesystem(&ext3_fs_type);
5507 "EXT4-fs: Unable to register as ext3 (%d)\n", err);
5510 static inline void unregister_as_ext3(void)
5512 unregister_filesystem(&ext3_fs_type);
5515 static inline int ext3_feature_set_ok(struct super_block *sb)
5517 if (EXT4_HAS_INCOMPAT_FEATURE(sb, ~EXT3_FEATURE_INCOMPAT_SUPP))
5519 if (!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL))
5521 if (sb->s_flags & MS_RDONLY)
5523 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, ~EXT3_FEATURE_RO_COMPAT_SUPP))
5528 static inline void register_as_ext3(void) { }
5529 static inline void unregister_as_ext3(void) { }
5530 static inline int ext3_feature_set_ok(struct super_block *sb) { return 0; }
5533 static struct file_system_type ext4_fs_type = {
5534 .owner = THIS_MODULE,
5536 .mount = ext4_mount,
5537 .kill_sb = kill_block_super,
5538 .fs_flags = FS_REQUIRES_DEV,
5540 MODULE_ALIAS_FS("ext4");
5542 static int __init ext4_init_feat_adverts(void)
5544 struct ext4_features *ef;
5547 ef = kzalloc(sizeof(struct ext4_features), GFP_KERNEL);
5551 ef->f_kobj.kset = ext4_kset;
5552 init_completion(&ef->f_kobj_unregister);
5553 ret = kobject_init_and_add(&ef->f_kobj, &ext4_feat_ktype, NULL,
5566 static void ext4_exit_feat_adverts(void)
5568 kobject_put(&ext4_feat->f_kobj);
5569 wait_for_completion(&ext4_feat->f_kobj_unregister);
5573 /* Shared across all ext4 file systems */
5574 wait_queue_head_t ext4__ioend_wq[EXT4_WQ_HASH_SZ];
5575 struct mutex ext4__aio_mutex[EXT4_WQ_HASH_SZ];
5577 static int __init ext4_init_fs(void)
5581 ext4_li_info = NULL;
5582 mutex_init(&ext4_li_mtx);
5584 /* Build-time check for flags consistency */
5585 ext4_check_flag_values();
5587 for (i = 0; i < EXT4_WQ_HASH_SZ; i++) {
5588 mutex_init(&ext4__aio_mutex[i]);
5589 init_waitqueue_head(&ext4__ioend_wq[i]);
5592 err = ext4_init_es();
5596 err = ext4_init_pageio();
5600 err = ext4_init_system_zone();
5603 ext4_kset = kset_create_and_add("ext4", NULL, fs_kobj);
5608 ext4_proc_root = proc_mkdir("fs/ext4", NULL);
5610 err = ext4_init_feat_adverts();
5614 err = ext4_init_mballoc();
5618 ext4_mballoc_ready = 1;
5619 err = init_inodecache();
5624 err = register_filesystem(&ext4_fs_type);
5630 unregister_as_ext2();
5631 unregister_as_ext3();
5632 destroy_inodecache();
5634 ext4_mballoc_ready = 0;
5635 ext4_exit_mballoc();
5637 ext4_exit_feat_adverts();
5640 remove_proc_entry("fs/ext4", NULL);
5641 kset_unregister(ext4_kset);
5643 ext4_exit_system_zone();
5652 static void __exit ext4_exit_fs(void)
5654 ext4_destroy_lazyinit_thread();
5655 unregister_as_ext2();
5656 unregister_as_ext3();
5657 unregister_filesystem(&ext4_fs_type);
5658 destroy_inodecache();
5659 ext4_exit_mballoc();
5660 ext4_exit_feat_adverts();
5661 remove_proc_entry("fs/ext4", NULL);
5662 kset_unregister(ext4_kset);
5663 ext4_exit_system_zone();
5668 MODULE_AUTHOR("Remy Card, Stephen Tweedie, Andrew Morton, Andreas Dilger, Theodore Ts'o and others");
5669 MODULE_DESCRIPTION("Fourth Extended Filesystem");
5670 MODULE_LICENSE("GPL");
5671 module_init(ext4_init_fs)
5672 module_exit(ext4_exit_fs)