2 * linux/fs/ext4/super.c
4 * Copyright (C) 1992, 1993, 1994, 1995
5 * Remy Card (card@masi.ibp.fr)
6 * Laboratoire MASI - Institut Blaise Pascal
7 * Universite Pierre et Marie Curie (Paris VI)
11 * linux/fs/minix/inode.c
13 * Copyright (C) 1991, 1992 Linus Torvalds
15 * Big-endian to little-endian byte-swapping/bitmaps by
16 * David S. Miller (davem@caip.rutgers.edu), 1995
19 #include <linux/module.h>
20 #include <linux/string.h>
22 #include <linux/time.h>
23 #include <linux/vmalloc.h>
24 #include <linux/jbd2.h>
25 #include <linux/slab.h>
26 #include <linux/init.h>
27 #include <linux/blkdev.h>
28 #include <linux/parser.h>
29 #include <linux/buffer_head.h>
30 #include <linux/exportfs.h>
31 #include <linux/vfs.h>
32 #include <linux/random.h>
33 #include <linux/mount.h>
34 #include <linux/namei.h>
35 #include <linux/quotaops.h>
36 #include <linux/seq_file.h>
37 #include <linux/proc_fs.h>
38 #include <linux/ctype.h>
39 #include <linux/log2.h>
40 #include <linux/crc16.h>
41 #include <linux/cleancache.h>
42 #include <asm/uaccess.h>
44 #include <linux/kthread.h>
45 #include <linux/freezer.h>
48 #include "ext4_extents.h" /* Needed for trace points definition */
49 #include "ext4_jbd2.h"
54 #define CREATE_TRACE_POINTS
55 #include <trace/events/ext4.h>
57 static struct proc_dir_entry *ext4_proc_root;
58 static struct kset *ext4_kset;
59 static struct ext4_lazy_init *ext4_li_info;
60 static struct mutex ext4_li_mtx;
61 static struct ext4_features *ext4_feat;
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_sync_fs_nojournal(struct super_block *sb, int wait);
73 static int ext4_remount(struct super_block *sb, int *flags, char *data);
74 static int ext4_statfs(struct dentry *dentry, struct kstatfs *buf);
75 static int ext4_unfreeze(struct super_block *sb);
76 static int ext4_freeze(struct super_block *sb);
77 static struct dentry *ext4_mount(struct file_system_type *fs_type, int flags,
78 const char *dev_name, void *data);
79 static inline int ext2_feature_set_ok(struct super_block *sb);
80 static inline int ext3_feature_set_ok(struct super_block *sb);
81 static int ext4_feature_set_ok(struct super_block *sb, int readonly);
82 static void ext4_destroy_lazyinit_thread(void);
83 static void ext4_unregister_li_request(struct super_block *sb);
84 static void ext4_clear_request_list(void);
85 static int ext4_reserve_clusters(struct ext4_sb_info *, ext4_fsblk_t);
87 #if !defined(CONFIG_EXT2_FS) && !defined(CONFIG_EXT2_FS_MODULE) && defined(CONFIG_EXT4_USE_FOR_EXT23)
88 static struct file_system_type ext2_fs_type = {
92 .kill_sb = kill_block_super,
93 .fs_flags = FS_REQUIRES_DEV,
95 MODULE_ALIAS_FS("ext2");
97 #define IS_EXT2_SB(sb) ((sb)->s_bdev->bd_holder == &ext2_fs_type)
99 #define IS_EXT2_SB(sb) (0)
103 #if !defined(CONFIG_EXT3_FS) && !defined(CONFIG_EXT3_FS_MODULE) && defined(CONFIG_EXT4_USE_FOR_EXT23)
104 static struct file_system_type ext3_fs_type = {
105 .owner = THIS_MODULE,
108 .kill_sb = kill_block_super,
109 .fs_flags = FS_REQUIRES_DEV,
111 MODULE_ALIAS_FS("ext3");
112 MODULE_ALIAS("ext3");
113 #define IS_EXT3_SB(sb) ((sb)->s_bdev->bd_holder == &ext3_fs_type)
115 #define IS_EXT3_SB(sb) (0)
118 static int ext4_verify_csum_type(struct super_block *sb,
119 struct ext4_super_block *es)
121 if (!EXT4_HAS_RO_COMPAT_FEATURE(sb,
122 EXT4_FEATURE_RO_COMPAT_METADATA_CSUM))
125 return es->s_checksum_type == EXT4_CRC32C_CHKSUM;
128 static __le32 ext4_superblock_csum(struct super_block *sb,
129 struct ext4_super_block *es)
131 struct ext4_sb_info *sbi = EXT4_SB(sb);
132 int offset = offsetof(struct ext4_super_block, s_checksum);
135 csum = ext4_chksum(sbi, ~0, (char *)es, offset);
137 return cpu_to_le32(csum);
140 int ext4_superblock_csum_verify(struct super_block *sb,
141 struct ext4_super_block *es)
143 if (!EXT4_HAS_RO_COMPAT_FEATURE(sb,
144 EXT4_FEATURE_RO_COMPAT_METADATA_CSUM))
147 return es->s_checksum == ext4_superblock_csum(sb, es);
150 void ext4_superblock_csum_set(struct super_block *sb)
152 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
154 if (!EXT4_HAS_RO_COMPAT_FEATURE(sb,
155 EXT4_FEATURE_RO_COMPAT_METADATA_CSUM))
158 es->s_checksum = ext4_superblock_csum(sb, es);
161 void *ext4_kvmalloc(size_t size, gfp_t flags)
165 ret = kmalloc(size, flags);
167 ret = __vmalloc(size, flags, PAGE_KERNEL);
171 void *ext4_kvzalloc(size_t size, gfp_t flags)
175 ret = kzalloc(size, flags);
177 ret = __vmalloc(size, flags | __GFP_ZERO, PAGE_KERNEL);
181 void ext4_kvfree(void *ptr)
183 if (is_vmalloc_addr(ptr))
190 ext4_fsblk_t ext4_block_bitmap(struct super_block *sb,
191 struct ext4_group_desc *bg)
193 return le32_to_cpu(bg->bg_block_bitmap_lo) |
194 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
195 (ext4_fsblk_t)le32_to_cpu(bg->bg_block_bitmap_hi) << 32 : 0);
198 ext4_fsblk_t ext4_inode_bitmap(struct super_block *sb,
199 struct ext4_group_desc *bg)
201 return le32_to_cpu(bg->bg_inode_bitmap_lo) |
202 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
203 (ext4_fsblk_t)le32_to_cpu(bg->bg_inode_bitmap_hi) << 32 : 0);
206 ext4_fsblk_t ext4_inode_table(struct super_block *sb,
207 struct ext4_group_desc *bg)
209 return le32_to_cpu(bg->bg_inode_table_lo) |
210 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
211 (ext4_fsblk_t)le32_to_cpu(bg->bg_inode_table_hi) << 32 : 0);
214 __u32 ext4_free_group_clusters(struct super_block *sb,
215 struct ext4_group_desc *bg)
217 return le16_to_cpu(bg->bg_free_blocks_count_lo) |
218 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
219 (__u32)le16_to_cpu(bg->bg_free_blocks_count_hi) << 16 : 0);
222 __u32 ext4_free_inodes_count(struct super_block *sb,
223 struct ext4_group_desc *bg)
225 return le16_to_cpu(bg->bg_free_inodes_count_lo) |
226 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
227 (__u32)le16_to_cpu(bg->bg_free_inodes_count_hi) << 16 : 0);
230 __u32 ext4_used_dirs_count(struct super_block *sb,
231 struct ext4_group_desc *bg)
233 return le16_to_cpu(bg->bg_used_dirs_count_lo) |
234 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
235 (__u32)le16_to_cpu(bg->bg_used_dirs_count_hi) << 16 : 0);
238 __u32 ext4_itable_unused_count(struct super_block *sb,
239 struct ext4_group_desc *bg)
241 return le16_to_cpu(bg->bg_itable_unused_lo) |
242 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
243 (__u32)le16_to_cpu(bg->bg_itable_unused_hi) << 16 : 0);
246 void ext4_block_bitmap_set(struct super_block *sb,
247 struct ext4_group_desc *bg, ext4_fsblk_t blk)
249 bg->bg_block_bitmap_lo = cpu_to_le32((u32)blk);
250 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
251 bg->bg_block_bitmap_hi = cpu_to_le32(blk >> 32);
254 void ext4_inode_bitmap_set(struct super_block *sb,
255 struct ext4_group_desc *bg, ext4_fsblk_t blk)
257 bg->bg_inode_bitmap_lo = cpu_to_le32((u32)blk);
258 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
259 bg->bg_inode_bitmap_hi = cpu_to_le32(blk >> 32);
262 void ext4_inode_table_set(struct super_block *sb,
263 struct ext4_group_desc *bg, ext4_fsblk_t blk)
265 bg->bg_inode_table_lo = cpu_to_le32((u32)blk);
266 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
267 bg->bg_inode_table_hi = cpu_to_le32(blk >> 32);
270 void ext4_free_group_clusters_set(struct super_block *sb,
271 struct ext4_group_desc *bg, __u32 count)
273 bg->bg_free_blocks_count_lo = cpu_to_le16((__u16)count);
274 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
275 bg->bg_free_blocks_count_hi = cpu_to_le16(count >> 16);
278 void ext4_free_inodes_set(struct super_block *sb,
279 struct ext4_group_desc *bg, __u32 count)
281 bg->bg_free_inodes_count_lo = cpu_to_le16((__u16)count);
282 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
283 bg->bg_free_inodes_count_hi = cpu_to_le16(count >> 16);
286 void ext4_used_dirs_set(struct super_block *sb,
287 struct ext4_group_desc *bg, __u32 count)
289 bg->bg_used_dirs_count_lo = cpu_to_le16((__u16)count);
290 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
291 bg->bg_used_dirs_count_hi = cpu_to_le16(count >> 16);
294 void ext4_itable_unused_set(struct super_block *sb,
295 struct ext4_group_desc *bg, __u32 count)
297 bg->bg_itable_unused_lo = cpu_to_le16((__u16)count);
298 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
299 bg->bg_itable_unused_hi = cpu_to_le16(count >> 16);
303 static void __save_error_info(struct super_block *sb, const char *func,
306 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
308 EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS;
309 es->s_state |= cpu_to_le16(EXT4_ERROR_FS);
310 es->s_last_error_time = cpu_to_le32(get_seconds());
311 strncpy(es->s_last_error_func, func, sizeof(es->s_last_error_func));
312 es->s_last_error_line = cpu_to_le32(line);
313 if (!es->s_first_error_time) {
314 es->s_first_error_time = es->s_last_error_time;
315 strncpy(es->s_first_error_func, func,
316 sizeof(es->s_first_error_func));
317 es->s_first_error_line = cpu_to_le32(line);
318 es->s_first_error_ino = es->s_last_error_ino;
319 es->s_first_error_block = es->s_last_error_block;
322 * Start the daily error reporting function if it hasn't been
325 if (!es->s_error_count)
326 mod_timer(&EXT4_SB(sb)->s_err_report, jiffies + 24*60*60*HZ);
327 le32_add_cpu(&es->s_error_count, 1);
330 static void save_error_info(struct super_block *sb, const char *func,
333 __save_error_info(sb, func, line);
334 ext4_commit_super(sb, 1);
338 * The del_gendisk() function uninitializes the disk-specific data
339 * structures, including the bdi structure, without telling anyone
340 * else. Once this happens, any attempt to call mark_buffer_dirty()
341 * (for example, by ext4_commit_super), will cause a kernel OOPS.
342 * This is a kludge to prevent these oops until we can put in a proper
343 * hook in del_gendisk() to inform the VFS and file system layers.
345 static int block_device_ejected(struct super_block *sb)
347 struct inode *bd_inode = sb->s_bdev->bd_inode;
348 struct backing_dev_info *bdi = bd_inode->i_mapping->backing_dev_info;
350 return bdi->dev == NULL;
353 static void ext4_journal_commit_callback(journal_t *journal, transaction_t *txn)
355 struct super_block *sb = journal->j_private;
356 struct ext4_sb_info *sbi = EXT4_SB(sb);
357 int error = is_journal_aborted(journal);
358 struct ext4_journal_cb_entry *jce;
360 BUG_ON(txn->t_state == T_FINISHED);
361 spin_lock(&sbi->s_md_lock);
362 while (!list_empty(&txn->t_private_list)) {
363 jce = list_entry(txn->t_private_list.next,
364 struct ext4_journal_cb_entry, jce_list);
365 list_del_init(&jce->jce_list);
366 spin_unlock(&sbi->s_md_lock);
367 jce->jce_func(sb, jce, error);
368 spin_lock(&sbi->s_md_lock);
370 spin_unlock(&sbi->s_md_lock);
373 /* Deal with the reporting of failure conditions on a filesystem such as
374 * inconsistencies detected or read IO failures.
376 * On ext2, we can store the error state of the filesystem in the
377 * superblock. That is not possible on ext4, because we may have other
378 * write ordering constraints on the superblock which prevent us from
379 * writing it out straight away; and given that the journal is about to
380 * be aborted, we can't rely on the current, or future, transactions to
381 * write out the superblock safely.
383 * We'll just use the jbd2_journal_abort() error code to record an error in
384 * the journal instead. On recovery, the journal will complain about
385 * that error until we've noted it down and cleared it.
388 static void ext4_handle_error(struct super_block *sb)
390 if (sb->s_flags & MS_RDONLY)
393 if (!test_opt(sb, ERRORS_CONT)) {
394 journal_t *journal = EXT4_SB(sb)->s_journal;
396 EXT4_SB(sb)->s_mount_flags |= EXT4_MF_FS_ABORTED;
398 jbd2_journal_abort(journal, -EIO);
400 if (test_opt(sb, ERRORS_RO)) {
401 ext4_msg(sb, KERN_CRIT, "Remounting filesystem read-only");
403 * Make sure updated value of ->s_mount_flags will be visible
404 * before ->s_flags update
407 sb->s_flags |= MS_RDONLY;
409 if (test_opt(sb, ERRORS_PANIC))
410 panic("EXT4-fs (device %s): panic forced after error\n",
414 void __ext4_error(struct super_block *sb, const char *function,
415 unsigned int line, const char *fmt, ...)
417 struct va_format vaf;
423 printk(KERN_CRIT "EXT4-fs error (device %s): %s:%d: comm %s: %pV\n",
424 sb->s_id, function, line, current->comm, &vaf);
426 save_error_info(sb, function, line);
428 ext4_handle_error(sb);
431 void __ext4_error_inode(struct inode *inode, const char *function,
432 unsigned int line, ext4_fsblk_t block,
433 const char *fmt, ...)
436 struct va_format vaf;
437 struct ext4_super_block *es = EXT4_SB(inode->i_sb)->s_es;
439 es->s_last_error_ino = cpu_to_le32(inode->i_ino);
440 es->s_last_error_block = cpu_to_le64(block);
441 save_error_info(inode->i_sb, function, line);
446 printk(KERN_CRIT "EXT4-fs error (device %s): %s:%d: "
447 "inode #%lu: block %llu: comm %s: %pV\n",
448 inode->i_sb->s_id, function, line, inode->i_ino,
449 block, current->comm, &vaf);
451 printk(KERN_CRIT "EXT4-fs error (device %s): %s:%d: "
452 "inode #%lu: comm %s: %pV\n",
453 inode->i_sb->s_id, function, line, inode->i_ino,
454 current->comm, &vaf);
457 ext4_handle_error(inode->i_sb);
460 void __ext4_error_file(struct file *file, const char *function,
461 unsigned int line, ext4_fsblk_t block,
462 const char *fmt, ...)
465 struct va_format vaf;
466 struct ext4_super_block *es;
467 struct inode *inode = file_inode(file);
468 char pathname[80], *path;
470 es = EXT4_SB(inode->i_sb)->s_es;
471 es->s_last_error_ino = cpu_to_le32(inode->i_ino);
472 save_error_info(inode->i_sb, function, line);
473 path = d_path(&(file->f_path), pathname, sizeof(pathname));
481 "EXT4-fs error (device %s): %s:%d: inode #%lu: "
482 "block %llu: comm %s: path %s: %pV\n",
483 inode->i_sb->s_id, function, line, inode->i_ino,
484 block, current->comm, path, &vaf);
487 "EXT4-fs error (device %s): %s:%d: inode #%lu: "
488 "comm %s: path %s: %pV\n",
489 inode->i_sb->s_id, function, line, inode->i_ino,
490 current->comm, path, &vaf);
493 ext4_handle_error(inode->i_sb);
496 const char *ext4_decode_error(struct super_block *sb, int errno,
503 errstr = "IO failure";
506 errstr = "Out of memory";
509 if (!sb || (EXT4_SB(sb)->s_journal &&
510 EXT4_SB(sb)->s_journal->j_flags & JBD2_ABORT))
511 errstr = "Journal has aborted";
513 errstr = "Readonly filesystem";
516 /* If the caller passed in an extra buffer for unknown
517 * errors, textualise them now. Else we just return
520 /* Check for truncated error codes... */
521 if (snprintf(nbuf, 16, "error %d", -errno) >= 0)
530 /* __ext4_std_error decodes expected errors from journaling functions
531 * automatically and invokes the appropriate error response. */
533 void __ext4_std_error(struct super_block *sb, const char *function,
534 unsigned int line, int errno)
539 /* Special case: if the error is EROFS, and we're not already
540 * inside a transaction, then there's really no point in logging
542 if (errno == -EROFS && journal_current_handle() == NULL &&
543 (sb->s_flags & MS_RDONLY))
546 errstr = ext4_decode_error(sb, errno, nbuf);
547 printk(KERN_CRIT "EXT4-fs error (device %s) in %s:%d: %s\n",
548 sb->s_id, function, line, errstr);
549 save_error_info(sb, function, line);
551 ext4_handle_error(sb);
555 * ext4_abort is a much stronger failure handler than ext4_error. The
556 * abort function may be used to deal with unrecoverable failures such
557 * as journal IO errors or ENOMEM at a critical moment in log management.
559 * We unconditionally force the filesystem into an ABORT|READONLY state,
560 * unless the error response on the fs has been set to panic in which
561 * case we take the easy way out and panic immediately.
564 void __ext4_abort(struct super_block *sb, const char *function,
565 unsigned int line, const char *fmt, ...)
569 save_error_info(sb, function, line);
571 printk(KERN_CRIT "EXT4-fs error (device %s): %s:%d: ", sb->s_id,
577 if ((sb->s_flags & MS_RDONLY) == 0) {
578 ext4_msg(sb, KERN_CRIT, "Remounting filesystem read-only");
579 EXT4_SB(sb)->s_mount_flags |= EXT4_MF_FS_ABORTED;
581 * Make sure updated value of ->s_mount_flags will be visible
582 * before ->s_flags update
585 sb->s_flags |= MS_RDONLY;
586 if (EXT4_SB(sb)->s_journal)
587 jbd2_journal_abort(EXT4_SB(sb)->s_journal, -EIO);
588 save_error_info(sb, function, line);
590 if (test_opt(sb, ERRORS_PANIC))
591 panic("EXT4-fs panic from previous error\n");
594 void __ext4_msg(struct super_block *sb,
595 const char *prefix, const char *fmt, ...)
597 struct va_format vaf;
603 printk("%sEXT4-fs (%s): %pV\n", prefix, sb->s_id, &vaf);
607 void __ext4_warning(struct super_block *sb, const char *function,
608 unsigned int line, const char *fmt, ...)
610 struct va_format vaf;
616 printk(KERN_WARNING "EXT4-fs warning (device %s): %s:%d: %pV\n",
617 sb->s_id, function, line, &vaf);
621 void __ext4_grp_locked_error(const char *function, unsigned int line,
622 struct super_block *sb, ext4_group_t grp,
623 unsigned long ino, ext4_fsblk_t block,
624 const char *fmt, ...)
628 struct va_format vaf;
630 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
632 es->s_last_error_ino = cpu_to_le32(ino);
633 es->s_last_error_block = cpu_to_le64(block);
634 __save_error_info(sb, function, line);
640 printk(KERN_CRIT "EXT4-fs error (device %s): %s:%d: group %u, ",
641 sb->s_id, function, line, grp);
643 printk(KERN_CONT "inode %lu: ", ino);
645 printk(KERN_CONT "block %llu:", (unsigned long long) block);
646 printk(KERN_CONT "%pV\n", &vaf);
649 if (test_opt(sb, ERRORS_CONT)) {
650 ext4_commit_super(sb, 0);
654 ext4_unlock_group(sb, grp);
655 ext4_handle_error(sb);
657 * We only get here in the ERRORS_RO case; relocking the group
658 * may be dangerous, but nothing bad will happen since the
659 * filesystem will have already been marked read/only and the
660 * journal has been aborted. We return 1 as a hint to callers
661 * who might what to use the return value from
662 * ext4_grp_locked_error() to distinguish between the
663 * ERRORS_CONT and ERRORS_RO case, and perhaps return more
664 * aggressively from the ext4 function in question, with a
665 * more appropriate error code.
667 ext4_lock_group(sb, grp);
671 void ext4_update_dynamic_rev(struct super_block *sb)
673 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
675 if (le32_to_cpu(es->s_rev_level) > EXT4_GOOD_OLD_REV)
679 "updating to rev %d because of new feature flag, "
680 "running e2fsck is recommended",
683 es->s_first_ino = cpu_to_le32(EXT4_GOOD_OLD_FIRST_INO);
684 es->s_inode_size = cpu_to_le16(EXT4_GOOD_OLD_INODE_SIZE);
685 es->s_rev_level = cpu_to_le32(EXT4_DYNAMIC_REV);
686 /* leave es->s_feature_*compat flags alone */
687 /* es->s_uuid will be set by e2fsck if empty */
690 * The rest of the superblock fields should be zero, and if not it
691 * means they are likely already in use, so leave them alone. We
692 * can leave it up to e2fsck to clean up any inconsistencies there.
697 * Open the external journal device
699 static struct block_device *ext4_blkdev_get(dev_t dev, struct super_block *sb)
701 struct block_device *bdev;
702 char b[BDEVNAME_SIZE];
704 bdev = blkdev_get_by_dev(dev, FMODE_READ|FMODE_WRITE|FMODE_EXCL, sb);
710 ext4_msg(sb, KERN_ERR, "failed to open journal device %s: %ld",
711 __bdevname(dev, b), PTR_ERR(bdev));
716 * Release the journal device
718 static void ext4_blkdev_put(struct block_device *bdev)
720 blkdev_put(bdev, FMODE_READ|FMODE_WRITE|FMODE_EXCL);
723 static void ext4_blkdev_remove(struct ext4_sb_info *sbi)
725 struct block_device *bdev;
726 bdev = sbi->journal_bdev;
728 ext4_blkdev_put(bdev);
729 sbi->journal_bdev = NULL;
733 static inline struct inode *orphan_list_entry(struct list_head *l)
735 return &list_entry(l, struct ext4_inode_info, i_orphan)->vfs_inode;
738 static void dump_orphan_list(struct super_block *sb, struct ext4_sb_info *sbi)
742 ext4_msg(sb, KERN_ERR, "sb orphan head is %d",
743 le32_to_cpu(sbi->s_es->s_last_orphan));
745 printk(KERN_ERR "sb_info orphan list:\n");
746 list_for_each(l, &sbi->s_orphan) {
747 struct inode *inode = orphan_list_entry(l);
749 "inode %s:%lu at %p: mode %o, nlink %d, next %d\n",
750 inode->i_sb->s_id, inode->i_ino, inode,
751 inode->i_mode, inode->i_nlink,
756 static void ext4_put_super(struct super_block *sb)
758 struct ext4_sb_info *sbi = EXT4_SB(sb);
759 struct ext4_super_block *es = sbi->s_es;
762 ext4_unregister_li_request(sb);
763 dquot_disable(sb, -1, DQUOT_USAGE_ENABLED | DQUOT_LIMITS_ENABLED);
765 flush_workqueue(sbi->unrsv_conversion_wq);
766 flush_workqueue(sbi->rsv_conversion_wq);
767 destroy_workqueue(sbi->unrsv_conversion_wq);
768 destroy_workqueue(sbi->rsv_conversion_wq);
770 if (sbi->s_journal) {
771 err = jbd2_journal_destroy(sbi->s_journal);
772 sbi->s_journal = NULL;
774 ext4_abort(sb, "Couldn't clean up the journal");
777 ext4_es_unregister_shrinker(sbi);
778 del_timer(&sbi->s_err_report);
779 ext4_release_system_zone(sb);
781 ext4_ext_release(sb);
782 ext4_xattr_put_super(sb);
784 if (!(sb->s_flags & MS_RDONLY)) {
785 EXT4_CLEAR_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
786 es->s_state = cpu_to_le16(sbi->s_mount_state);
788 if (!(sb->s_flags & MS_RDONLY))
789 ext4_commit_super(sb, 1);
792 remove_proc_entry("options", sbi->s_proc);
793 remove_proc_entry(sb->s_id, ext4_proc_root);
795 kobject_del(&sbi->s_kobj);
797 for (i = 0; i < sbi->s_gdb_count; i++)
798 brelse(sbi->s_group_desc[i]);
799 ext4_kvfree(sbi->s_group_desc);
800 ext4_kvfree(sbi->s_flex_groups);
801 percpu_counter_destroy(&sbi->s_freeclusters_counter);
802 percpu_counter_destroy(&sbi->s_freeinodes_counter);
803 percpu_counter_destroy(&sbi->s_dirs_counter);
804 percpu_counter_destroy(&sbi->s_dirtyclusters_counter);
805 percpu_counter_destroy(&sbi->s_extent_cache_cnt);
808 for (i = 0; i < MAXQUOTAS; i++)
809 kfree(sbi->s_qf_names[i]);
812 /* Debugging code just in case the in-memory inode orphan list
813 * isn't empty. The on-disk one can be non-empty if we've
814 * detected an error and taken the fs readonly, but the
815 * in-memory list had better be clean by this point. */
816 if (!list_empty(&sbi->s_orphan))
817 dump_orphan_list(sb, sbi);
818 J_ASSERT(list_empty(&sbi->s_orphan));
820 invalidate_bdev(sb->s_bdev);
821 if (sbi->journal_bdev && sbi->journal_bdev != sb->s_bdev) {
823 * Invalidate the journal device's buffers. We don't want them
824 * floating about in memory - the physical journal device may
825 * hotswapped, and it breaks the `ro-after' testing code.
827 sync_blockdev(sbi->journal_bdev);
828 invalidate_bdev(sbi->journal_bdev);
829 ext4_blkdev_remove(sbi);
832 kthread_stop(sbi->s_mmp_tsk);
833 sb->s_fs_info = NULL;
835 * Now that we are completely done shutting down the
836 * superblock, we need to actually destroy the kobject.
838 kobject_put(&sbi->s_kobj);
839 wait_for_completion(&sbi->s_kobj_unregister);
840 if (sbi->s_chksum_driver)
841 crypto_free_shash(sbi->s_chksum_driver);
842 kfree(sbi->s_blockgroup_lock);
846 static struct kmem_cache *ext4_inode_cachep;
849 * Called inside transaction, so use GFP_NOFS
851 static struct inode *ext4_alloc_inode(struct super_block *sb)
853 struct ext4_inode_info *ei;
855 ei = kmem_cache_alloc(ext4_inode_cachep, GFP_NOFS);
859 ei->vfs_inode.i_version = 1;
860 INIT_LIST_HEAD(&ei->i_prealloc_list);
861 spin_lock_init(&ei->i_prealloc_lock);
862 ext4_es_init_tree(&ei->i_es_tree);
863 rwlock_init(&ei->i_es_lock);
864 INIT_LIST_HEAD(&ei->i_es_lru);
866 ei->i_touch_when = 0;
867 ei->i_reserved_data_blocks = 0;
868 ei->i_reserved_meta_blocks = 0;
869 ei->i_allocated_meta_blocks = 0;
870 ei->i_da_metadata_calc_len = 0;
871 ei->i_da_metadata_calc_last_lblock = 0;
872 spin_lock_init(&(ei->i_block_reservation_lock));
874 ei->i_reserved_quota = 0;
877 INIT_LIST_HEAD(&ei->i_rsv_conversion_list);
878 INIT_LIST_HEAD(&ei->i_unrsv_conversion_list);
879 spin_lock_init(&ei->i_completed_io_lock);
881 ei->i_datasync_tid = 0;
882 atomic_set(&ei->i_ioend_count, 0);
883 atomic_set(&ei->i_unwritten, 0);
884 INIT_WORK(&ei->i_rsv_conversion_work, ext4_end_io_rsv_work);
885 INIT_WORK(&ei->i_unrsv_conversion_work, ext4_end_io_unrsv_work);
887 return &ei->vfs_inode;
890 static int ext4_drop_inode(struct inode *inode)
892 int drop = generic_drop_inode(inode);
894 trace_ext4_drop_inode(inode, drop);
898 static void ext4_i_callback(struct rcu_head *head)
900 struct inode *inode = container_of(head, struct inode, i_rcu);
901 kmem_cache_free(ext4_inode_cachep, EXT4_I(inode));
904 static void ext4_destroy_inode(struct inode *inode)
906 if (!list_empty(&(EXT4_I(inode)->i_orphan))) {
907 ext4_msg(inode->i_sb, KERN_ERR,
908 "Inode %lu (%p): orphan list check failed!",
909 inode->i_ino, EXT4_I(inode));
910 print_hex_dump(KERN_INFO, "", DUMP_PREFIX_ADDRESS, 16, 4,
911 EXT4_I(inode), sizeof(struct ext4_inode_info),
915 call_rcu(&inode->i_rcu, ext4_i_callback);
918 static void init_once(void *foo)
920 struct ext4_inode_info *ei = (struct ext4_inode_info *) foo;
922 INIT_LIST_HEAD(&ei->i_orphan);
923 init_rwsem(&ei->xattr_sem);
924 init_rwsem(&ei->i_data_sem);
925 inode_init_once(&ei->vfs_inode);
928 static int init_inodecache(void)
930 ext4_inode_cachep = kmem_cache_create("ext4_inode_cache",
931 sizeof(struct ext4_inode_info),
932 0, (SLAB_RECLAIM_ACCOUNT|
935 if (ext4_inode_cachep == NULL)
940 static void destroy_inodecache(void)
943 * Make sure all delayed rcu free inodes are flushed before we
947 kmem_cache_destroy(ext4_inode_cachep);
950 void ext4_clear_inode(struct inode *inode)
952 invalidate_inode_buffers(inode);
955 ext4_discard_preallocations(inode);
956 ext4_es_remove_extent(inode, 0, EXT_MAX_BLOCKS);
957 ext4_es_lru_del(inode);
958 if (EXT4_I(inode)->jinode) {
959 jbd2_journal_release_jbd_inode(EXT4_JOURNAL(inode),
960 EXT4_I(inode)->jinode);
961 jbd2_free_inode(EXT4_I(inode)->jinode);
962 EXT4_I(inode)->jinode = NULL;
966 static struct inode *ext4_nfs_get_inode(struct super_block *sb,
967 u64 ino, u32 generation)
971 if (ino < EXT4_FIRST_INO(sb) && ino != EXT4_ROOT_INO)
972 return ERR_PTR(-ESTALE);
973 if (ino > le32_to_cpu(EXT4_SB(sb)->s_es->s_inodes_count))
974 return ERR_PTR(-ESTALE);
976 /* iget isn't really right if the inode is currently unallocated!!
978 * ext4_read_inode will return a bad_inode if the inode had been
979 * deleted, so we should be safe.
981 * Currently we don't know the generation for parent directory, so
982 * a generation of 0 means "accept any"
984 inode = ext4_iget(sb, ino);
986 return ERR_CAST(inode);
987 if (generation && inode->i_generation != generation) {
989 return ERR_PTR(-ESTALE);
995 static struct dentry *ext4_fh_to_dentry(struct super_block *sb, struct fid *fid,
996 int fh_len, int fh_type)
998 return generic_fh_to_dentry(sb, fid, fh_len, fh_type,
1002 static struct dentry *ext4_fh_to_parent(struct super_block *sb, struct fid *fid,
1003 int fh_len, int fh_type)
1005 return generic_fh_to_parent(sb, fid, fh_len, fh_type,
1006 ext4_nfs_get_inode);
1010 * Try to release metadata pages (indirect blocks, directories) which are
1011 * mapped via the block device. Since these pages could have journal heads
1012 * which would prevent try_to_free_buffers() from freeing them, we must use
1013 * jbd2 layer's try_to_free_buffers() function to release them.
1015 static int bdev_try_to_free_page(struct super_block *sb, struct page *page,
1018 journal_t *journal = EXT4_SB(sb)->s_journal;
1020 WARN_ON(PageChecked(page));
1021 if (!page_has_buffers(page))
1024 return jbd2_journal_try_to_free_buffers(journal, page,
1025 wait & ~__GFP_WAIT);
1026 return try_to_free_buffers(page);
1030 #define QTYPE2NAME(t) ((t) == USRQUOTA ? "user" : "group")
1031 #define QTYPE2MOPT(on, t) ((t) == USRQUOTA?((on)##USRJQUOTA):((on)##GRPJQUOTA))
1033 static int ext4_write_dquot(struct dquot *dquot);
1034 static int ext4_acquire_dquot(struct dquot *dquot);
1035 static int ext4_release_dquot(struct dquot *dquot);
1036 static int ext4_mark_dquot_dirty(struct dquot *dquot);
1037 static int ext4_write_info(struct super_block *sb, int type);
1038 static int ext4_quota_on(struct super_block *sb, int type, int format_id,
1040 static int ext4_quota_on_sysfile(struct super_block *sb, int type,
1042 static int ext4_quota_off(struct super_block *sb, int type);
1043 static int ext4_quota_off_sysfile(struct super_block *sb, int type);
1044 static int ext4_quota_on_mount(struct super_block *sb, int type);
1045 static ssize_t ext4_quota_read(struct super_block *sb, int type, char *data,
1046 size_t len, loff_t off);
1047 static ssize_t ext4_quota_write(struct super_block *sb, int type,
1048 const char *data, size_t len, loff_t off);
1049 static int ext4_quota_enable(struct super_block *sb, int type, int format_id,
1050 unsigned int flags);
1051 static int ext4_enable_quotas(struct super_block *sb);
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_info = dquot_get_dqinfo,
1069 .set_info = dquot_set_dqinfo,
1070 .get_dqblk = dquot_get_dqblk,
1071 .set_dqblk = dquot_set_dqblk
1074 static const struct quotactl_ops ext4_qctl_sysfile_operations = {
1075 .quota_on_meta = ext4_quota_on_sysfile,
1076 .quota_off = ext4_quota_off_sysfile,
1077 .quota_sync = dquot_quota_sync,
1078 .get_info = dquot_get_dqinfo,
1079 .set_info = dquot_set_dqinfo,
1080 .get_dqblk = dquot_get_dqblk,
1081 .set_dqblk = dquot_set_dqblk
1085 static const struct super_operations ext4_sops = {
1086 .alloc_inode = ext4_alloc_inode,
1087 .destroy_inode = ext4_destroy_inode,
1088 .write_inode = ext4_write_inode,
1089 .dirty_inode = ext4_dirty_inode,
1090 .drop_inode = ext4_drop_inode,
1091 .evict_inode = ext4_evict_inode,
1092 .put_super = ext4_put_super,
1093 .sync_fs = ext4_sync_fs,
1094 .freeze_fs = ext4_freeze,
1095 .unfreeze_fs = ext4_unfreeze,
1096 .statfs = ext4_statfs,
1097 .remount_fs = ext4_remount,
1098 .show_options = ext4_show_options,
1100 .quota_read = ext4_quota_read,
1101 .quota_write = ext4_quota_write,
1103 .bdev_try_to_free_page = bdev_try_to_free_page,
1106 static const struct super_operations ext4_nojournal_sops = {
1107 .alloc_inode = ext4_alloc_inode,
1108 .destroy_inode = ext4_destroy_inode,
1109 .write_inode = ext4_write_inode,
1110 .dirty_inode = ext4_dirty_inode,
1111 .drop_inode = ext4_drop_inode,
1112 .evict_inode = ext4_evict_inode,
1113 .sync_fs = ext4_sync_fs_nojournal,
1114 .put_super = ext4_put_super,
1115 .statfs = ext4_statfs,
1116 .remount_fs = ext4_remount,
1117 .show_options = ext4_show_options,
1119 .quota_read = ext4_quota_read,
1120 .quota_write = ext4_quota_write,
1122 .bdev_try_to_free_page = bdev_try_to_free_page,
1125 static const struct export_operations ext4_export_ops = {
1126 .fh_to_dentry = ext4_fh_to_dentry,
1127 .fh_to_parent = ext4_fh_to_parent,
1128 .get_parent = ext4_get_parent,
1132 Opt_bsd_df, Opt_minix_df, Opt_grpid, Opt_nogrpid,
1133 Opt_resgid, Opt_resuid, Opt_sb, Opt_err_cont, Opt_err_panic, Opt_err_ro,
1134 Opt_nouid32, Opt_debug, Opt_removed,
1135 Opt_user_xattr, Opt_nouser_xattr, Opt_acl, Opt_noacl,
1136 Opt_auto_da_alloc, Opt_noauto_da_alloc, Opt_noload,
1137 Opt_commit, Opt_min_batch_time, Opt_max_batch_time, Opt_journal_dev,
1138 Opt_journal_path, Opt_journal_checksum, Opt_journal_async_commit,
1139 Opt_abort, Opt_data_journal, Opt_data_ordered, Opt_data_writeback,
1140 Opt_data_err_abort, Opt_data_err_ignore,
1141 Opt_usrjquota, Opt_grpjquota, Opt_offusrjquota, Opt_offgrpjquota,
1142 Opt_jqfmt_vfsold, Opt_jqfmt_vfsv0, Opt_jqfmt_vfsv1, Opt_quota,
1143 Opt_noquota, Opt_barrier, Opt_nobarrier, Opt_err,
1144 Opt_usrquota, Opt_grpquota, Opt_i_version,
1145 Opt_stripe, Opt_delalloc, Opt_nodelalloc, Opt_mblk_io_submit,
1146 Opt_nomblk_io_submit, Opt_block_validity, Opt_noblock_validity,
1147 Opt_inode_readahead_blks, Opt_journal_ioprio,
1148 Opt_dioread_nolock, Opt_dioread_lock,
1149 Opt_discard, Opt_nodiscard, Opt_init_itable, Opt_noinit_itable,
1150 Opt_max_dir_size_kb,
1153 static const match_table_t tokens = {
1154 {Opt_bsd_df, "bsddf"},
1155 {Opt_minix_df, "minixdf"},
1156 {Opt_grpid, "grpid"},
1157 {Opt_grpid, "bsdgroups"},
1158 {Opt_nogrpid, "nogrpid"},
1159 {Opt_nogrpid, "sysvgroups"},
1160 {Opt_resgid, "resgid=%u"},
1161 {Opt_resuid, "resuid=%u"},
1163 {Opt_err_cont, "errors=continue"},
1164 {Opt_err_panic, "errors=panic"},
1165 {Opt_err_ro, "errors=remount-ro"},
1166 {Opt_nouid32, "nouid32"},
1167 {Opt_debug, "debug"},
1168 {Opt_removed, "oldalloc"},
1169 {Opt_removed, "orlov"},
1170 {Opt_user_xattr, "user_xattr"},
1171 {Opt_nouser_xattr, "nouser_xattr"},
1173 {Opt_noacl, "noacl"},
1174 {Opt_noload, "norecovery"},
1175 {Opt_noload, "noload"},
1176 {Opt_removed, "nobh"},
1177 {Opt_removed, "bh"},
1178 {Opt_commit, "commit=%u"},
1179 {Opt_min_batch_time, "min_batch_time=%u"},
1180 {Opt_max_batch_time, "max_batch_time=%u"},
1181 {Opt_journal_dev, "journal_dev=%u"},
1182 {Opt_journal_path, "journal_path=%s"},
1183 {Opt_journal_checksum, "journal_checksum"},
1184 {Opt_journal_async_commit, "journal_async_commit"},
1185 {Opt_abort, "abort"},
1186 {Opt_data_journal, "data=journal"},
1187 {Opt_data_ordered, "data=ordered"},
1188 {Opt_data_writeback, "data=writeback"},
1189 {Opt_data_err_abort, "data_err=abort"},
1190 {Opt_data_err_ignore, "data_err=ignore"},
1191 {Opt_offusrjquota, "usrjquota="},
1192 {Opt_usrjquota, "usrjquota=%s"},
1193 {Opt_offgrpjquota, "grpjquota="},
1194 {Opt_grpjquota, "grpjquota=%s"},
1195 {Opt_jqfmt_vfsold, "jqfmt=vfsold"},
1196 {Opt_jqfmt_vfsv0, "jqfmt=vfsv0"},
1197 {Opt_jqfmt_vfsv1, "jqfmt=vfsv1"},
1198 {Opt_grpquota, "grpquota"},
1199 {Opt_noquota, "noquota"},
1200 {Opt_quota, "quota"},
1201 {Opt_usrquota, "usrquota"},
1202 {Opt_barrier, "barrier=%u"},
1203 {Opt_barrier, "barrier"},
1204 {Opt_nobarrier, "nobarrier"},
1205 {Opt_i_version, "i_version"},
1206 {Opt_stripe, "stripe=%u"},
1207 {Opt_delalloc, "delalloc"},
1208 {Opt_nodelalloc, "nodelalloc"},
1209 {Opt_removed, "mblk_io_submit"},
1210 {Opt_removed, "nomblk_io_submit"},
1211 {Opt_block_validity, "block_validity"},
1212 {Opt_noblock_validity, "noblock_validity"},
1213 {Opt_inode_readahead_blks, "inode_readahead_blks=%u"},
1214 {Opt_journal_ioprio, "journal_ioprio=%u"},
1215 {Opt_auto_da_alloc, "auto_da_alloc=%u"},
1216 {Opt_auto_da_alloc, "auto_da_alloc"},
1217 {Opt_noauto_da_alloc, "noauto_da_alloc"},
1218 {Opt_dioread_nolock, "dioread_nolock"},
1219 {Opt_dioread_lock, "dioread_lock"},
1220 {Opt_discard, "discard"},
1221 {Opt_nodiscard, "nodiscard"},
1222 {Opt_init_itable, "init_itable=%u"},
1223 {Opt_init_itable, "init_itable"},
1224 {Opt_noinit_itable, "noinit_itable"},
1225 {Opt_max_dir_size_kb, "max_dir_size_kb=%u"},
1226 {Opt_removed, "check=none"}, /* mount option from ext2/3 */
1227 {Opt_removed, "nocheck"}, /* mount option from ext2/3 */
1228 {Opt_removed, "reservation"}, /* mount option from ext2/3 */
1229 {Opt_removed, "noreservation"}, /* mount option from ext2/3 */
1230 {Opt_removed, "journal=%u"}, /* mount option from ext2/3 */
1234 static ext4_fsblk_t get_sb_block(void **data)
1236 ext4_fsblk_t sb_block;
1237 char *options = (char *) *data;
1239 if (!options || strncmp(options, "sb=", 3) != 0)
1240 return 1; /* Default location */
1243 /* TODO: use simple_strtoll with >32bit ext4 */
1244 sb_block = simple_strtoul(options, &options, 0);
1245 if (*options && *options != ',') {
1246 printk(KERN_ERR "EXT4-fs: Invalid sb specification: %s\n",
1250 if (*options == ',')
1252 *data = (void *) options;
1257 #define DEFAULT_JOURNAL_IOPRIO (IOPRIO_PRIO_VALUE(IOPRIO_CLASS_BE, 3))
1258 static char deprecated_msg[] = "Mount option \"%s\" will be removed by %s\n"
1259 "Contact linux-ext4@vger.kernel.org if you think we should keep it.\n";
1262 static int set_qf_name(struct super_block *sb, int qtype, substring_t *args)
1264 struct ext4_sb_info *sbi = EXT4_SB(sb);
1268 if (sb_any_quota_loaded(sb) &&
1269 !sbi->s_qf_names[qtype]) {
1270 ext4_msg(sb, KERN_ERR,
1271 "Cannot change journaled "
1272 "quota options when quota turned on");
1275 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_QUOTA)) {
1276 ext4_msg(sb, KERN_ERR, "Cannot set journaled quota options "
1277 "when QUOTA feature is enabled");
1280 qname = match_strdup(args);
1282 ext4_msg(sb, KERN_ERR,
1283 "Not enough memory for storing quotafile name");
1286 if (sbi->s_qf_names[qtype]) {
1287 if (strcmp(sbi->s_qf_names[qtype], qname) == 0)
1290 ext4_msg(sb, KERN_ERR,
1291 "%s quota file already specified",
1295 if (strchr(qname, '/')) {
1296 ext4_msg(sb, KERN_ERR,
1297 "quotafile must be on filesystem root");
1300 sbi->s_qf_names[qtype] = qname;
1308 static int clear_qf_name(struct super_block *sb, int qtype)
1311 struct ext4_sb_info *sbi = EXT4_SB(sb);
1313 if (sb_any_quota_loaded(sb) &&
1314 sbi->s_qf_names[qtype]) {
1315 ext4_msg(sb, KERN_ERR, "Cannot change journaled quota options"
1316 " when quota turned on");
1319 kfree(sbi->s_qf_names[qtype]);
1320 sbi->s_qf_names[qtype] = NULL;
1325 #define MOPT_SET 0x0001
1326 #define MOPT_CLEAR 0x0002
1327 #define MOPT_NOSUPPORT 0x0004
1328 #define MOPT_EXPLICIT 0x0008
1329 #define MOPT_CLEAR_ERR 0x0010
1330 #define MOPT_GTE0 0x0020
1333 #define MOPT_QFMT 0x0040
1335 #define MOPT_Q MOPT_NOSUPPORT
1336 #define MOPT_QFMT MOPT_NOSUPPORT
1338 #define MOPT_DATAJ 0x0080
1339 #define MOPT_NO_EXT2 0x0100
1340 #define MOPT_NO_EXT3 0x0200
1341 #define MOPT_EXT4_ONLY (MOPT_NO_EXT2 | MOPT_NO_EXT3)
1342 #define MOPT_STRING 0x0400
1344 static const struct mount_opts {
1348 } ext4_mount_opts[] = {
1349 {Opt_minix_df, EXT4_MOUNT_MINIX_DF, MOPT_SET},
1350 {Opt_bsd_df, EXT4_MOUNT_MINIX_DF, MOPT_CLEAR},
1351 {Opt_grpid, EXT4_MOUNT_GRPID, MOPT_SET},
1352 {Opt_nogrpid, EXT4_MOUNT_GRPID, MOPT_CLEAR},
1353 {Opt_block_validity, EXT4_MOUNT_BLOCK_VALIDITY, MOPT_SET},
1354 {Opt_noblock_validity, EXT4_MOUNT_BLOCK_VALIDITY, MOPT_CLEAR},
1355 {Opt_dioread_nolock, EXT4_MOUNT_DIOREAD_NOLOCK,
1356 MOPT_EXT4_ONLY | MOPT_SET},
1357 {Opt_dioread_lock, EXT4_MOUNT_DIOREAD_NOLOCK,
1358 MOPT_EXT4_ONLY | MOPT_CLEAR},
1359 {Opt_discard, EXT4_MOUNT_DISCARD, MOPT_SET},
1360 {Opt_nodiscard, EXT4_MOUNT_DISCARD, MOPT_CLEAR},
1361 {Opt_delalloc, EXT4_MOUNT_DELALLOC,
1362 MOPT_EXT4_ONLY | MOPT_SET | MOPT_EXPLICIT},
1363 {Opt_nodelalloc, EXT4_MOUNT_DELALLOC,
1364 MOPT_EXT4_ONLY | MOPT_CLEAR},
1365 {Opt_journal_checksum, EXT4_MOUNT_JOURNAL_CHECKSUM,
1366 MOPT_EXT4_ONLY | MOPT_SET},
1367 {Opt_journal_async_commit, (EXT4_MOUNT_JOURNAL_ASYNC_COMMIT |
1368 EXT4_MOUNT_JOURNAL_CHECKSUM),
1369 MOPT_EXT4_ONLY | MOPT_SET},
1370 {Opt_noload, EXT4_MOUNT_NOLOAD, MOPT_NO_EXT2 | MOPT_SET},
1371 {Opt_err_panic, EXT4_MOUNT_ERRORS_PANIC, MOPT_SET | MOPT_CLEAR_ERR},
1372 {Opt_err_ro, EXT4_MOUNT_ERRORS_RO, MOPT_SET | MOPT_CLEAR_ERR},
1373 {Opt_err_cont, EXT4_MOUNT_ERRORS_CONT, MOPT_SET | MOPT_CLEAR_ERR},
1374 {Opt_data_err_abort, EXT4_MOUNT_DATA_ERR_ABORT,
1375 MOPT_NO_EXT2 | MOPT_SET},
1376 {Opt_data_err_ignore, EXT4_MOUNT_DATA_ERR_ABORT,
1377 MOPT_NO_EXT2 | MOPT_CLEAR},
1378 {Opt_barrier, EXT4_MOUNT_BARRIER, MOPT_SET},
1379 {Opt_nobarrier, EXT4_MOUNT_BARRIER, MOPT_CLEAR},
1380 {Opt_noauto_da_alloc, EXT4_MOUNT_NO_AUTO_DA_ALLOC, MOPT_SET},
1381 {Opt_auto_da_alloc, EXT4_MOUNT_NO_AUTO_DA_ALLOC, MOPT_CLEAR},
1382 {Opt_noinit_itable, EXT4_MOUNT_INIT_INODE_TABLE, MOPT_CLEAR},
1383 {Opt_commit, 0, MOPT_GTE0},
1384 {Opt_max_batch_time, 0, MOPT_GTE0},
1385 {Opt_min_batch_time, 0, MOPT_GTE0},
1386 {Opt_inode_readahead_blks, 0, MOPT_GTE0},
1387 {Opt_init_itable, 0, MOPT_GTE0},
1388 {Opt_stripe, 0, MOPT_GTE0},
1389 {Opt_resuid, 0, MOPT_GTE0},
1390 {Opt_resgid, 0, MOPT_GTE0},
1391 {Opt_journal_dev, 0, MOPT_GTE0},
1392 {Opt_journal_path, 0, MOPT_STRING},
1393 {Opt_journal_ioprio, 0, MOPT_GTE0},
1394 {Opt_data_journal, EXT4_MOUNT_JOURNAL_DATA, MOPT_NO_EXT2 | MOPT_DATAJ},
1395 {Opt_data_ordered, EXT4_MOUNT_ORDERED_DATA, MOPT_NO_EXT2 | MOPT_DATAJ},
1396 {Opt_data_writeback, EXT4_MOUNT_WRITEBACK_DATA,
1397 MOPT_NO_EXT2 | MOPT_DATAJ},
1398 {Opt_user_xattr, EXT4_MOUNT_XATTR_USER, MOPT_SET},
1399 {Opt_nouser_xattr, EXT4_MOUNT_XATTR_USER, MOPT_CLEAR},
1400 #ifdef CONFIG_EXT4_FS_POSIX_ACL
1401 {Opt_acl, EXT4_MOUNT_POSIX_ACL, MOPT_SET},
1402 {Opt_noacl, EXT4_MOUNT_POSIX_ACL, MOPT_CLEAR},
1404 {Opt_acl, 0, MOPT_NOSUPPORT},
1405 {Opt_noacl, 0, MOPT_NOSUPPORT},
1407 {Opt_nouid32, EXT4_MOUNT_NO_UID32, MOPT_SET},
1408 {Opt_debug, EXT4_MOUNT_DEBUG, MOPT_SET},
1409 {Opt_quota, EXT4_MOUNT_QUOTA | EXT4_MOUNT_USRQUOTA, MOPT_SET | MOPT_Q},
1410 {Opt_usrquota, EXT4_MOUNT_QUOTA | EXT4_MOUNT_USRQUOTA,
1412 {Opt_grpquota, EXT4_MOUNT_QUOTA | EXT4_MOUNT_GRPQUOTA,
1414 {Opt_noquota, (EXT4_MOUNT_QUOTA | EXT4_MOUNT_USRQUOTA |
1415 EXT4_MOUNT_GRPQUOTA), MOPT_CLEAR | MOPT_Q},
1416 {Opt_usrjquota, 0, MOPT_Q},
1417 {Opt_grpjquota, 0, MOPT_Q},
1418 {Opt_offusrjquota, 0, MOPT_Q},
1419 {Opt_offgrpjquota, 0, MOPT_Q},
1420 {Opt_jqfmt_vfsold, QFMT_VFS_OLD, MOPT_QFMT},
1421 {Opt_jqfmt_vfsv0, QFMT_VFS_V0, MOPT_QFMT},
1422 {Opt_jqfmt_vfsv1, QFMT_VFS_V1, MOPT_QFMT},
1423 {Opt_max_dir_size_kb, 0, MOPT_GTE0},
1427 static int handle_mount_opt(struct super_block *sb, char *opt, int token,
1428 substring_t *args, unsigned long *journal_devnum,
1429 unsigned int *journal_ioprio, int is_remount)
1431 struct ext4_sb_info *sbi = EXT4_SB(sb);
1432 const struct mount_opts *m;
1438 if (token == Opt_usrjquota)
1439 return set_qf_name(sb, USRQUOTA, &args[0]);
1440 else if (token == Opt_grpjquota)
1441 return set_qf_name(sb, GRPQUOTA, &args[0]);
1442 else if (token == Opt_offusrjquota)
1443 return clear_qf_name(sb, USRQUOTA);
1444 else if (token == Opt_offgrpjquota)
1445 return clear_qf_name(sb, GRPQUOTA);
1449 case Opt_nouser_xattr:
1450 ext4_msg(sb, KERN_WARNING, deprecated_msg, opt, "3.5");
1453 return 1; /* handled by get_sb_block() */
1455 ext4_msg(sb, KERN_WARNING, "Ignoring removed %s option", opt);
1458 sbi->s_mount_flags |= EXT4_MF_FS_ABORTED;
1461 sb->s_flags |= MS_I_VERSION;
1465 for (m = ext4_mount_opts; m->token != Opt_err; m++)
1466 if (token == m->token)
1469 if (m->token == Opt_err) {
1470 ext4_msg(sb, KERN_ERR, "Unrecognized mount option \"%s\" "
1471 "or missing value", opt);
1475 if ((m->flags & MOPT_NO_EXT2) && IS_EXT2_SB(sb)) {
1476 ext4_msg(sb, KERN_ERR,
1477 "Mount option \"%s\" incompatible with ext2", opt);
1480 if ((m->flags & MOPT_NO_EXT3) && IS_EXT3_SB(sb)) {
1481 ext4_msg(sb, KERN_ERR,
1482 "Mount option \"%s\" incompatible with ext3", opt);
1486 if (args->from && !(m->flags & MOPT_STRING) && match_int(args, &arg))
1488 if (args->from && (m->flags & MOPT_GTE0) && (arg < 0))
1490 if (m->flags & MOPT_EXPLICIT)
1491 set_opt2(sb, EXPLICIT_DELALLOC);
1492 if (m->flags & MOPT_CLEAR_ERR)
1493 clear_opt(sb, ERRORS_MASK);
1494 if (token == Opt_noquota && sb_any_quota_loaded(sb)) {
1495 ext4_msg(sb, KERN_ERR, "Cannot change quota "
1496 "options when quota turned on");
1500 if (m->flags & MOPT_NOSUPPORT) {
1501 ext4_msg(sb, KERN_ERR, "%s option not supported", opt);
1502 } else if (token == Opt_commit) {
1504 arg = JBD2_DEFAULT_MAX_COMMIT_AGE;
1505 sbi->s_commit_interval = HZ * arg;
1506 } else if (token == Opt_max_batch_time) {
1508 arg = EXT4_DEF_MAX_BATCH_TIME;
1509 sbi->s_max_batch_time = arg;
1510 } else if (token == Opt_min_batch_time) {
1511 sbi->s_min_batch_time = arg;
1512 } else if (token == Opt_inode_readahead_blks) {
1513 if (arg && (arg > (1 << 30) || !is_power_of_2(arg))) {
1514 ext4_msg(sb, KERN_ERR,
1515 "EXT4-fs: inode_readahead_blks must be "
1516 "0 or a power of 2 smaller than 2^31");
1519 sbi->s_inode_readahead_blks = arg;
1520 } else if (token == Opt_init_itable) {
1521 set_opt(sb, INIT_INODE_TABLE);
1523 arg = EXT4_DEF_LI_WAIT_MULT;
1524 sbi->s_li_wait_mult = arg;
1525 } else if (token == Opt_max_dir_size_kb) {
1526 sbi->s_max_dir_size_kb = arg;
1527 } else if (token == Opt_stripe) {
1528 sbi->s_stripe = arg;
1529 } else if (token == Opt_resuid) {
1530 uid = make_kuid(current_user_ns(), arg);
1531 if (!uid_valid(uid)) {
1532 ext4_msg(sb, KERN_ERR, "Invalid uid value %d", arg);
1535 sbi->s_resuid = uid;
1536 } else if (token == Opt_resgid) {
1537 gid = make_kgid(current_user_ns(), arg);
1538 if (!gid_valid(gid)) {
1539 ext4_msg(sb, KERN_ERR, "Invalid gid value %d", arg);
1542 sbi->s_resgid = gid;
1543 } else if (token == Opt_journal_dev) {
1545 ext4_msg(sb, KERN_ERR,
1546 "Cannot specify journal on remount");
1549 *journal_devnum = arg;
1550 } else if (token == Opt_journal_path) {
1552 struct inode *journal_inode;
1557 ext4_msg(sb, KERN_ERR,
1558 "Cannot specify journal on remount");
1561 journal_path = match_strdup(&args[0]);
1562 if (!journal_path) {
1563 ext4_msg(sb, KERN_ERR, "error: could not dup "
1564 "journal device string");
1568 error = kern_path(journal_path, LOOKUP_FOLLOW, &path);
1570 ext4_msg(sb, KERN_ERR, "error: could not find "
1571 "journal device path: error %d", error);
1572 kfree(journal_path);
1576 journal_inode = path.dentry->d_inode;
1577 if (!S_ISBLK(journal_inode->i_mode)) {
1578 ext4_msg(sb, KERN_ERR, "error: journal path %s "
1579 "is not a block device", journal_path);
1581 kfree(journal_path);
1585 *journal_devnum = new_encode_dev(journal_inode->i_rdev);
1587 kfree(journal_path);
1588 } else if (token == Opt_journal_ioprio) {
1590 ext4_msg(sb, KERN_ERR, "Invalid journal IO priority"
1595 IOPRIO_PRIO_VALUE(IOPRIO_CLASS_BE, arg);
1596 } else if (m->flags & MOPT_DATAJ) {
1598 if (!sbi->s_journal)
1599 ext4_msg(sb, KERN_WARNING, "Remounting file system with no journal so ignoring journalled data option");
1600 else if (test_opt(sb, DATA_FLAGS) != m->mount_opt) {
1601 ext4_msg(sb, KERN_ERR,
1602 "Cannot change data mode on remount");
1606 clear_opt(sb, DATA_FLAGS);
1607 sbi->s_mount_opt |= m->mount_opt;
1610 } else if (m->flags & MOPT_QFMT) {
1611 if (sb_any_quota_loaded(sb) &&
1612 sbi->s_jquota_fmt != m->mount_opt) {
1613 ext4_msg(sb, KERN_ERR, "Cannot change journaled "
1614 "quota options when quota turned on");
1617 if (EXT4_HAS_RO_COMPAT_FEATURE(sb,
1618 EXT4_FEATURE_RO_COMPAT_QUOTA)) {
1619 ext4_msg(sb, KERN_ERR,
1620 "Cannot set journaled quota options "
1621 "when QUOTA feature is enabled");
1624 sbi->s_jquota_fmt = m->mount_opt;
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 "
1697 if (sbi->s_jquota_fmt) {
1698 ext4_msg(sb, KERN_ERR, "journaled quota format "
1699 "specified with no journaling "
1705 if (test_opt(sb, DIOREAD_NOLOCK)) {
1707 BLOCK_SIZE << le32_to_cpu(sbi->s_es->s_log_block_size);
1709 if (blocksize < PAGE_CACHE_SIZE) {
1710 ext4_msg(sb, KERN_ERR, "can't mount with "
1711 "dioread_nolock if block size != PAGE_SIZE");
1718 static inline void ext4_show_quota_options(struct seq_file *seq,
1719 struct super_block *sb)
1721 #if defined(CONFIG_QUOTA)
1722 struct ext4_sb_info *sbi = EXT4_SB(sb);
1724 if (sbi->s_jquota_fmt) {
1727 switch (sbi->s_jquota_fmt) {
1738 seq_printf(seq, ",jqfmt=%s", fmtname);
1741 if (sbi->s_qf_names[USRQUOTA])
1742 seq_printf(seq, ",usrjquota=%s", sbi->s_qf_names[USRQUOTA]);
1744 if (sbi->s_qf_names[GRPQUOTA])
1745 seq_printf(seq, ",grpjquota=%s", sbi->s_qf_names[GRPQUOTA]);
1749 static const char *token2str(int token)
1751 const struct match_token *t;
1753 for (t = tokens; t->token != Opt_err; t++)
1754 if (t->token == token && !strchr(t->pattern, '='))
1761 * - it's set to a non-default value OR
1762 * - if the per-sb default is different from the global default
1764 static int _ext4_show_options(struct seq_file *seq, struct super_block *sb,
1767 struct ext4_sb_info *sbi = EXT4_SB(sb);
1768 struct ext4_super_block *es = sbi->s_es;
1769 int def_errors, def_mount_opt = nodefs ? 0 : sbi->s_def_mount_opt;
1770 const struct mount_opts *m;
1771 char sep = nodefs ? '\n' : ',';
1773 #define SEQ_OPTS_PUTS(str) seq_printf(seq, "%c" str, sep)
1774 #define SEQ_OPTS_PRINT(str, arg) seq_printf(seq, "%c" str, sep, arg)
1776 if (sbi->s_sb_block != 1)
1777 SEQ_OPTS_PRINT("sb=%llu", sbi->s_sb_block);
1779 for (m = ext4_mount_opts; m->token != Opt_err; m++) {
1780 int want_set = m->flags & MOPT_SET;
1781 if (((m->flags & (MOPT_SET|MOPT_CLEAR)) == 0) ||
1782 (m->flags & MOPT_CLEAR_ERR))
1784 if (!(m->mount_opt & (sbi->s_mount_opt ^ def_mount_opt)))
1785 continue; /* skip if same as the default */
1787 (sbi->s_mount_opt & m->mount_opt) != m->mount_opt) ||
1788 (!want_set && (sbi->s_mount_opt & m->mount_opt)))
1789 continue; /* select Opt_noFoo vs Opt_Foo */
1790 SEQ_OPTS_PRINT("%s", token2str(m->token));
1793 if (nodefs || !uid_eq(sbi->s_resuid, make_kuid(&init_user_ns, EXT4_DEF_RESUID)) ||
1794 le16_to_cpu(es->s_def_resuid) != EXT4_DEF_RESUID)
1795 SEQ_OPTS_PRINT("resuid=%u",
1796 from_kuid_munged(&init_user_ns, sbi->s_resuid));
1797 if (nodefs || !gid_eq(sbi->s_resgid, make_kgid(&init_user_ns, EXT4_DEF_RESGID)) ||
1798 le16_to_cpu(es->s_def_resgid) != EXT4_DEF_RESGID)
1799 SEQ_OPTS_PRINT("resgid=%u",
1800 from_kgid_munged(&init_user_ns, sbi->s_resgid));
1801 def_errors = nodefs ? -1 : le16_to_cpu(es->s_errors);
1802 if (test_opt(sb, ERRORS_RO) && def_errors != EXT4_ERRORS_RO)
1803 SEQ_OPTS_PUTS("errors=remount-ro");
1804 if (test_opt(sb, ERRORS_CONT) && def_errors != EXT4_ERRORS_CONTINUE)
1805 SEQ_OPTS_PUTS("errors=continue");
1806 if (test_opt(sb, ERRORS_PANIC) && def_errors != EXT4_ERRORS_PANIC)
1807 SEQ_OPTS_PUTS("errors=panic");
1808 if (nodefs || sbi->s_commit_interval != JBD2_DEFAULT_MAX_COMMIT_AGE*HZ)
1809 SEQ_OPTS_PRINT("commit=%lu", sbi->s_commit_interval / HZ);
1810 if (nodefs || sbi->s_min_batch_time != EXT4_DEF_MIN_BATCH_TIME)
1811 SEQ_OPTS_PRINT("min_batch_time=%u", sbi->s_min_batch_time);
1812 if (nodefs || sbi->s_max_batch_time != EXT4_DEF_MAX_BATCH_TIME)
1813 SEQ_OPTS_PRINT("max_batch_time=%u", sbi->s_max_batch_time);
1814 if (sb->s_flags & MS_I_VERSION)
1815 SEQ_OPTS_PUTS("i_version");
1816 if (nodefs || sbi->s_stripe)
1817 SEQ_OPTS_PRINT("stripe=%lu", sbi->s_stripe);
1818 if (EXT4_MOUNT_DATA_FLAGS & (sbi->s_mount_opt ^ def_mount_opt)) {
1819 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA)
1820 SEQ_OPTS_PUTS("data=journal");
1821 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA)
1822 SEQ_OPTS_PUTS("data=ordered");
1823 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_WRITEBACK_DATA)
1824 SEQ_OPTS_PUTS("data=writeback");
1827 sbi->s_inode_readahead_blks != EXT4_DEF_INODE_READAHEAD_BLKS)
1828 SEQ_OPTS_PRINT("inode_readahead_blks=%u",
1829 sbi->s_inode_readahead_blks);
1831 if (nodefs || (test_opt(sb, INIT_INODE_TABLE) &&
1832 (sbi->s_li_wait_mult != EXT4_DEF_LI_WAIT_MULT)))
1833 SEQ_OPTS_PRINT("init_itable=%u", sbi->s_li_wait_mult);
1834 if (nodefs || sbi->s_max_dir_size_kb)
1835 SEQ_OPTS_PRINT("max_dir_size_kb=%u", sbi->s_max_dir_size_kb);
1837 ext4_show_quota_options(seq, sb);
1841 static int ext4_show_options(struct seq_file *seq, struct dentry *root)
1843 return _ext4_show_options(seq, root->d_sb, 0);
1846 static int options_seq_show(struct seq_file *seq, void *offset)
1848 struct super_block *sb = seq->private;
1851 seq_puts(seq, (sb->s_flags & MS_RDONLY) ? "ro" : "rw");
1852 rc = _ext4_show_options(seq, sb, 1);
1853 seq_puts(seq, "\n");
1857 static int options_open_fs(struct inode *inode, struct file *file)
1859 return single_open(file, options_seq_show, PDE_DATA(inode));
1862 static const struct file_operations ext4_seq_options_fops = {
1863 .owner = THIS_MODULE,
1864 .open = options_open_fs,
1866 .llseek = seq_lseek,
1867 .release = single_release,
1870 static int ext4_setup_super(struct super_block *sb, struct ext4_super_block *es,
1873 struct ext4_sb_info *sbi = EXT4_SB(sb);
1876 if (le32_to_cpu(es->s_rev_level) > EXT4_MAX_SUPP_REV) {
1877 ext4_msg(sb, KERN_ERR, "revision level too high, "
1878 "forcing read-only mode");
1883 if (!(sbi->s_mount_state & EXT4_VALID_FS))
1884 ext4_msg(sb, KERN_WARNING, "warning: mounting unchecked fs, "
1885 "running e2fsck is recommended");
1886 else if ((sbi->s_mount_state & EXT4_ERROR_FS))
1887 ext4_msg(sb, KERN_WARNING,
1888 "warning: mounting fs with errors, "
1889 "running e2fsck is recommended");
1890 else if ((__s16) le16_to_cpu(es->s_max_mnt_count) > 0 &&
1891 le16_to_cpu(es->s_mnt_count) >=
1892 (unsigned short) (__s16) le16_to_cpu(es->s_max_mnt_count))
1893 ext4_msg(sb, KERN_WARNING,
1894 "warning: maximal mount count reached, "
1895 "running e2fsck is recommended");
1896 else if (le32_to_cpu(es->s_checkinterval) &&
1897 (le32_to_cpu(es->s_lastcheck) +
1898 le32_to_cpu(es->s_checkinterval) <= get_seconds()))
1899 ext4_msg(sb, KERN_WARNING,
1900 "warning: checktime reached, "
1901 "running e2fsck is recommended");
1902 if (!sbi->s_journal)
1903 es->s_state &= cpu_to_le16(~EXT4_VALID_FS);
1904 if (!(__s16) le16_to_cpu(es->s_max_mnt_count))
1905 es->s_max_mnt_count = cpu_to_le16(EXT4_DFL_MAX_MNT_COUNT);
1906 le16_add_cpu(&es->s_mnt_count, 1);
1907 es->s_mtime = cpu_to_le32(get_seconds());
1908 ext4_update_dynamic_rev(sb);
1910 EXT4_SET_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
1912 ext4_commit_super(sb, 1);
1914 if (test_opt(sb, DEBUG))
1915 printk(KERN_INFO "[EXT4 FS bs=%lu, gc=%u, "
1916 "bpg=%lu, ipg=%lu, mo=%04x, mo2=%04x]\n",
1918 sbi->s_groups_count,
1919 EXT4_BLOCKS_PER_GROUP(sb),
1920 EXT4_INODES_PER_GROUP(sb),
1921 sbi->s_mount_opt, sbi->s_mount_opt2);
1923 cleancache_init_fs(sb);
1927 int ext4_alloc_flex_bg_array(struct super_block *sb, ext4_group_t ngroup)
1929 struct ext4_sb_info *sbi = EXT4_SB(sb);
1930 struct flex_groups *new_groups;
1933 if (!sbi->s_log_groups_per_flex)
1936 size = ext4_flex_group(sbi, ngroup - 1) + 1;
1937 if (size <= sbi->s_flex_groups_allocated)
1940 size = roundup_pow_of_two(size * sizeof(struct flex_groups));
1941 new_groups = ext4_kvzalloc(size, GFP_KERNEL);
1943 ext4_msg(sb, KERN_ERR, "not enough memory for %d flex groups",
1944 size / (int) sizeof(struct flex_groups));
1948 if (sbi->s_flex_groups) {
1949 memcpy(new_groups, sbi->s_flex_groups,
1950 (sbi->s_flex_groups_allocated *
1951 sizeof(struct flex_groups)));
1952 ext4_kvfree(sbi->s_flex_groups);
1954 sbi->s_flex_groups = new_groups;
1955 sbi->s_flex_groups_allocated = size / sizeof(struct flex_groups);
1959 static int ext4_fill_flex_info(struct super_block *sb)
1961 struct ext4_sb_info *sbi = EXT4_SB(sb);
1962 struct ext4_group_desc *gdp = NULL;
1963 ext4_group_t flex_group;
1966 sbi->s_log_groups_per_flex = sbi->s_es->s_log_groups_per_flex;
1967 if (sbi->s_log_groups_per_flex < 1 || sbi->s_log_groups_per_flex > 31) {
1968 sbi->s_log_groups_per_flex = 0;
1972 err = ext4_alloc_flex_bg_array(sb, sbi->s_groups_count);
1976 for (i = 0; i < sbi->s_groups_count; i++) {
1977 gdp = ext4_get_group_desc(sb, i, NULL);
1979 flex_group = ext4_flex_group(sbi, i);
1980 atomic_add(ext4_free_inodes_count(sb, gdp),
1981 &sbi->s_flex_groups[flex_group].free_inodes);
1982 atomic64_add(ext4_free_group_clusters(sb, gdp),
1983 &sbi->s_flex_groups[flex_group].free_clusters);
1984 atomic_add(ext4_used_dirs_count(sb, gdp),
1985 &sbi->s_flex_groups[flex_group].used_dirs);
1993 static __le16 ext4_group_desc_csum(struct ext4_sb_info *sbi, __u32 block_group,
1994 struct ext4_group_desc *gdp)
1998 __le32 le_group = cpu_to_le32(block_group);
2000 if ((sbi->s_es->s_feature_ro_compat &
2001 cpu_to_le32(EXT4_FEATURE_RO_COMPAT_METADATA_CSUM))) {
2002 /* Use new metadata_csum algorithm */
2006 save_csum = gdp->bg_checksum;
2007 gdp->bg_checksum = 0;
2008 csum32 = ext4_chksum(sbi, sbi->s_csum_seed, (__u8 *)&le_group,
2010 csum32 = ext4_chksum(sbi, csum32, (__u8 *)gdp,
2012 gdp->bg_checksum = save_csum;
2014 crc = csum32 & 0xFFFF;
2018 /* old crc16 code */
2019 offset = offsetof(struct ext4_group_desc, bg_checksum);
2021 crc = crc16(~0, sbi->s_es->s_uuid, sizeof(sbi->s_es->s_uuid));
2022 crc = crc16(crc, (__u8 *)&le_group, sizeof(le_group));
2023 crc = crc16(crc, (__u8 *)gdp, offset);
2024 offset += sizeof(gdp->bg_checksum); /* skip checksum */
2025 /* for checksum of struct ext4_group_desc do the rest...*/
2026 if ((sbi->s_es->s_feature_incompat &
2027 cpu_to_le32(EXT4_FEATURE_INCOMPAT_64BIT)) &&
2028 offset < le16_to_cpu(sbi->s_es->s_desc_size))
2029 crc = crc16(crc, (__u8 *)gdp + offset,
2030 le16_to_cpu(sbi->s_es->s_desc_size) -
2034 return cpu_to_le16(crc);
2037 int ext4_group_desc_csum_verify(struct super_block *sb, __u32 block_group,
2038 struct ext4_group_desc *gdp)
2040 if (ext4_has_group_desc_csum(sb) &&
2041 (gdp->bg_checksum != ext4_group_desc_csum(EXT4_SB(sb),
2048 void ext4_group_desc_csum_set(struct super_block *sb, __u32 block_group,
2049 struct ext4_group_desc *gdp)
2051 if (!ext4_has_group_desc_csum(sb))
2053 gdp->bg_checksum = ext4_group_desc_csum(EXT4_SB(sb), block_group, gdp);
2056 /* Called at mount-time, super-block is locked */
2057 static int ext4_check_descriptors(struct super_block *sb,
2058 ext4_group_t *first_not_zeroed)
2060 struct ext4_sb_info *sbi = EXT4_SB(sb);
2061 ext4_fsblk_t first_block = le32_to_cpu(sbi->s_es->s_first_data_block);
2062 ext4_fsblk_t last_block;
2063 ext4_fsblk_t block_bitmap;
2064 ext4_fsblk_t inode_bitmap;
2065 ext4_fsblk_t inode_table;
2066 int flexbg_flag = 0;
2067 ext4_group_t i, grp = sbi->s_groups_count;
2069 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_FLEX_BG))
2072 ext4_debug("Checking group descriptors");
2074 for (i = 0; i < sbi->s_groups_count; i++) {
2075 struct ext4_group_desc *gdp = ext4_get_group_desc(sb, i, NULL);
2077 if (i == sbi->s_groups_count - 1 || flexbg_flag)
2078 last_block = ext4_blocks_count(sbi->s_es) - 1;
2080 last_block = first_block +
2081 (EXT4_BLOCKS_PER_GROUP(sb) - 1);
2083 if ((grp == sbi->s_groups_count) &&
2084 !(gdp->bg_flags & cpu_to_le16(EXT4_BG_INODE_ZEROED)))
2087 block_bitmap = ext4_block_bitmap(sb, gdp);
2088 if (block_bitmap < first_block || block_bitmap > last_block) {
2089 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2090 "Block bitmap for group %u not in group "
2091 "(block %llu)!", i, block_bitmap);
2094 inode_bitmap = ext4_inode_bitmap(sb, gdp);
2095 if (inode_bitmap < first_block || inode_bitmap > last_block) {
2096 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2097 "Inode bitmap for group %u not in group "
2098 "(block %llu)!", i, inode_bitmap);
2101 inode_table = ext4_inode_table(sb, gdp);
2102 if (inode_table < first_block ||
2103 inode_table + sbi->s_itb_per_group - 1 > last_block) {
2104 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2105 "Inode table for group %u not in group "
2106 "(block %llu)!", i, inode_table);
2109 ext4_lock_group(sb, i);
2110 if (!ext4_group_desc_csum_verify(sb, i, gdp)) {
2111 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2112 "Checksum for group %u failed (%u!=%u)",
2113 i, le16_to_cpu(ext4_group_desc_csum(sbi, i,
2114 gdp)), le16_to_cpu(gdp->bg_checksum));
2115 if (!(sb->s_flags & MS_RDONLY)) {
2116 ext4_unlock_group(sb, i);
2120 ext4_unlock_group(sb, i);
2122 first_block += EXT4_BLOCKS_PER_GROUP(sb);
2124 if (NULL != first_not_zeroed)
2125 *first_not_zeroed = grp;
2127 ext4_free_blocks_count_set(sbi->s_es,
2128 EXT4_C2B(sbi, ext4_count_free_clusters(sb)));
2129 sbi->s_es->s_free_inodes_count =cpu_to_le32(ext4_count_free_inodes(sb));
2133 /* ext4_orphan_cleanup() walks a singly-linked list of inodes (starting at
2134 * the superblock) which were deleted from all directories, but held open by
2135 * a process at the time of a crash. We walk the list and try to delete these
2136 * inodes at recovery time (only with a read-write filesystem).
2138 * In order to keep the orphan inode chain consistent during traversal (in
2139 * case of crash during recovery), we link each inode into the superblock
2140 * orphan list_head and handle it the same way as an inode deletion during
2141 * normal operation (which journals the operations for us).
2143 * We only do an iget() and an iput() on each inode, which is very safe if we
2144 * accidentally point at an in-use or already deleted inode. The worst that
2145 * can happen in this case is that we get a "bit already cleared" message from
2146 * ext4_free_inode(). The only reason we would point at a wrong inode is if
2147 * e2fsck was run on this filesystem, and it must have already done the orphan
2148 * inode cleanup for us, so we can safely abort without any further action.
2150 static void ext4_orphan_cleanup(struct super_block *sb,
2151 struct ext4_super_block *es)
2153 unsigned int s_flags = sb->s_flags;
2154 int nr_orphans = 0, nr_truncates = 0;
2158 if (!es->s_last_orphan) {
2159 jbd_debug(4, "no orphan inodes to clean up\n");
2163 if (bdev_read_only(sb->s_bdev)) {
2164 ext4_msg(sb, KERN_ERR, "write access "
2165 "unavailable, skipping orphan cleanup");
2169 /* Check if feature set would not allow a r/w mount */
2170 if (!ext4_feature_set_ok(sb, 0)) {
2171 ext4_msg(sb, KERN_INFO, "Skipping orphan cleanup due to "
2172 "unknown ROCOMPAT features");
2176 if (EXT4_SB(sb)->s_mount_state & EXT4_ERROR_FS) {
2177 /* don't clear list on RO mount w/ errors */
2178 if (es->s_last_orphan && !(s_flags & MS_RDONLY)) {
2179 jbd_debug(1, "Errors on filesystem, "
2180 "clearing orphan list.\n");
2181 es->s_last_orphan = 0;
2183 jbd_debug(1, "Skipping orphan recovery on fs with errors.\n");
2187 if (s_flags & MS_RDONLY) {
2188 ext4_msg(sb, KERN_INFO, "orphan cleanup on readonly fs");
2189 sb->s_flags &= ~MS_RDONLY;
2192 /* Needed for iput() to work correctly and not trash data */
2193 sb->s_flags |= MS_ACTIVE;
2194 /* Turn on quotas so that they are updated correctly */
2195 for (i = 0; i < MAXQUOTAS; i++) {
2196 if (EXT4_SB(sb)->s_qf_names[i]) {
2197 int ret = ext4_quota_on_mount(sb, i);
2199 ext4_msg(sb, KERN_ERR,
2200 "Cannot turn on journaled "
2201 "quota: error %d", ret);
2206 while (es->s_last_orphan) {
2207 struct inode *inode;
2209 inode = ext4_orphan_get(sb, le32_to_cpu(es->s_last_orphan));
2210 if (IS_ERR(inode)) {
2211 es->s_last_orphan = 0;
2215 list_add(&EXT4_I(inode)->i_orphan, &EXT4_SB(sb)->s_orphan);
2216 dquot_initialize(inode);
2217 if (inode->i_nlink) {
2218 if (test_opt(sb, DEBUG))
2219 ext4_msg(sb, KERN_DEBUG,
2220 "%s: truncating inode %lu to %lld bytes",
2221 __func__, inode->i_ino, inode->i_size);
2222 jbd_debug(2, "truncating inode %lu to %lld bytes\n",
2223 inode->i_ino, inode->i_size);
2224 mutex_lock(&inode->i_mutex);
2225 truncate_inode_pages(inode->i_mapping, inode->i_size);
2226 ext4_truncate(inode);
2227 mutex_unlock(&inode->i_mutex);
2230 if (test_opt(sb, DEBUG))
2231 ext4_msg(sb, KERN_DEBUG,
2232 "%s: deleting unreferenced inode %lu",
2233 __func__, inode->i_ino);
2234 jbd_debug(2, "deleting unreferenced inode %lu\n",
2238 iput(inode); /* The delete magic happens here! */
2241 #define PLURAL(x) (x), ((x) == 1) ? "" : "s"
2244 ext4_msg(sb, KERN_INFO, "%d orphan inode%s deleted",
2245 PLURAL(nr_orphans));
2247 ext4_msg(sb, KERN_INFO, "%d truncate%s cleaned up",
2248 PLURAL(nr_truncates));
2250 /* Turn quotas off */
2251 for (i = 0; i < MAXQUOTAS; i++) {
2252 if (sb_dqopt(sb)->files[i])
2253 dquot_quota_off(sb, i);
2256 sb->s_flags = s_flags; /* Restore MS_RDONLY status */
2260 * Maximal extent format file size.
2261 * Resulting logical blkno at s_maxbytes must fit in our on-disk
2262 * extent format containers, within a sector_t, and within i_blocks
2263 * in the vfs. ext4 inode has 48 bits of i_block in fsblock units,
2264 * so that won't be a limiting factor.
2266 * However there is other limiting factor. We do store extents in the form
2267 * of starting block and length, hence the resulting length of the extent
2268 * covering maximum file size must fit into on-disk format containers as
2269 * well. Given that length is always by 1 unit bigger than max unit (because
2270 * we count 0 as well) we have to lower the s_maxbytes by one fs block.
2272 * Note, this does *not* consider any metadata overhead for vfs i_blocks.
2274 static loff_t ext4_max_size(int blkbits, int has_huge_files)
2277 loff_t upper_limit = MAX_LFS_FILESIZE;
2279 /* small i_blocks in vfs inode? */
2280 if (!has_huge_files || sizeof(blkcnt_t) < sizeof(u64)) {
2282 * CONFIG_LBDAF is not enabled implies the inode
2283 * i_block represent total blocks in 512 bytes
2284 * 32 == size of vfs inode i_blocks * 8
2286 upper_limit = (1LL << 32) - 1;
2288 /* total blocks in file system block size */
2289 upper_limit >>= (blkbits - 9);
2290 upper_limit <<= blkbits;
2294 * 32-bit extent-start container, ee_block. We lower the maxbytes
2295 * by one fs block, so ee_len can cover the extent of maximum file
2298 res = (1LL << 32) - 1;
2301 /* Sanity check against vm- & vfs- imposed limits */
2302 if (res > upper_limit)
2309 * Maximal bitmap file size. There is a direct, and {,double-,triple-}indirect
2310 * block limit, and also a limit of (2^48 - 1) 512-byte sectors in i_blocks.
2311 * We need to be 1 filesystem block less than the 2^48 sector limit.
2313 static loff_t ext4_max_bitmap_size(int bits, int has_huge_files)
2315 loff_t res = EXT4_NDIR_BLOCKS;
2318 /* This is calculated to be the largest file size for a dense, block
2319 * mapped file such that the file's total number of 512-byte sectors,
2320 * including data and all indirect blocks, does not exceed (2^48 - 1).
2322 * __u32 i_blocks_lo and _u16 i_blocks_high represent the total
2323 * number of 512-byte sectors of the file.
2326 if (!has_huge_files || sizeof(blkcnt_t) < sizeof(u64)) {
2328 * !has_huge_files or CONFIG_LBDAF not enabled implies that
2329 * the inode i_block field represents total file blocks in
2330 * 2^32 512-byte sectors == size of vfs inode i_blocks * 8
2332 upper_limit = (1LL << 32) - 1;
2334 /* total blocks in file system block size */
2335 upper_limit >>= (bits - 9);
2339 * We use 48 bit ext4_inode i_blocks
2340 * With EXT4_HUGE_FILE_FL set the i_blocks
2341 * represent total number of blocks in
2342 * file system block size
2344 upper_limit = (1LL << 48) - 1;
2348 /* indirect blocks */
2350 /* double indirect blocks */
2351 meta_blocks += 1 + (1LL << (bits-2));
2352 /* tripple indirect blocks */
2353 meta_blocks += 1 + (1LL << (bits-2)) + (1LL << (2*(bits-2)));
2355 upper_limit -= meta_blocks;
2356 upper_limit <<= bits;
2358 res += 1LL << (bits-2);
2359 res += 1LL << (2*(bits-2));
2360 res += 1LL << (3*(bits-2));
2362 if (res > upper_limit)
2365 if (res > MAX_LFS_FILESIZE)
2366 res = MAX_LFS_FILESIZE;
2371 static ext4_fsblk_t descriptor_loc(struct super_block *sb,
2372 ext4_fsblk_t logical_sb_block, int nr)
2374 struct ext4_sb_info *sbi = EXT4_SB(sb);
2375 ext4_group_t bg, first_meta_bg;
2378 first_meta_bg = le32_to_cpu(sbi->s_es->s_first_meta_bg);
2380 if (!EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_META_BG) ||
2382 return logical_sb_block + nr + 1;
2383 bg = sbi->s_desc_per_block * nr;
2384 if (ext4_bg_has_super(sb, bg))
2387 return (has_super + ext4_group_first_block_no(sb, bg));
2391 * ext4_get_stripe_size: Get the stripe size.
2392 * @sbi: In memory super block info
2394 * If we have specified it via mount option, then
2395 * use the mount option value. If the value specified at mount time is
2396 * greater than the blocks per group use the super block value.
2397 * If the super block value is greater than blocks per group return 0.
2398 * Allocator needs it be less than blocks per group.
2401 static unsigned long ext4_get_stripe_size(struct ext4_sb_info *sbi)
2403 unsigned long stride = le16_to_cpu(sbi->s_es->s_raid_stride);
2404 unsigned long stripe_width =
2405 le32_to_cpu(sbi->s_es->s_raid_stripe_width);
2408 if (sbi->s_stripe && sbi->s_stripe <= sbi->s_blocks_per_group)
2409 ret = sbi->s_stripe;
2410 else if (stripe_width <= sbi->s_blocks_per_group)
2412 else if (stride <= sbi->s_blocks_per_group)
2418 * If the stripe width is 1, this makes no sense and
2419 * we set it to 0 to turn off stripe handling code.
2430 struct attribute attr;
2431 ssize_t (*show)(struct ext4_attr *, struct ext4_sb_info *, char *);
2432 ssize_t (*store)(struct ext4_attr *, struct ext4_sb_info *,
2433 const char *, size_t);
2440 static int parse_strtoull(const char *buf,
2441 unsigned long long max, unsigned long long *value)
2445 ret = kstrtoull(skip_spaces(buf), 0, value);
2446 if (!ret && *value > max)
2451 static ssize_t delayed_allocation_blocks_show(struct ext4_attr *a,
2452 struct ext4_sb_info *sbi,
2455 return snprintf(buf, PAGE_SIZE, "%llu\n",
2457 percpu_counter_sum(&sbi->s_dirtyclusters_counter)));
2460 static ssize_t session_write_kbytes_show(struct ext4_attr *a,
2461 struct ext4_sb_info *sbi, char *buf)
2463 struct super_block *sb = sbi->s_buddy_cache->i_sb;
2465 if (!sb->s_bdev->bd_part)
2466 return snprintf(buf, PAGE_SIZE, "0\n");
2467 return snprintf(buf, PAGE_SIZE, "%lu\n",
2468 (part_stat_read(sb->s_bdev->bd_part, sectors[1]) -
2469 sbi->s_sectors_written_start) >> 1);
2472 static ssize_t lifetime_write_kbytes_show(struct ext4_attr *a,
2473 struct ext4_sb_info *sbi, char *buf)
2475 struct super_block *sb = sbi->s_buddy_cache->i_sb;
2477 if (!sb->s_bdev->bd_part)
2478 return snprintf(buf, PAGE_SIZE, "0\n");
2479 return snprintf(buf, PAGE_SIZE, "%llu\n",
2480 (unsigned long long)(sbi->s_kbytes_written +
2481 ((part_stat_read(sb->s_bdev->bd_part, sectors[1]) -
2482 EXT4_SB(sb)->s_sectors_written_start) >> 1)));
2485 static ssize_t inode_readahead_blks_store(struct ext4_attr *a,
2486 struct ext4_sb_info *sbi,
2487 const char *buf, size_t count)
2492 ret = kstrtoul(skip_spaces(buf), 0, &t);
2496 if (t && (!is_power_of_2(t) || t > 0x40000000))
2499 sbi->s_inode_readahead_blks = t;
2503 static ssize_t sbi_ui_show(struct ext4_attr *a,
2504 struct ext4_sb_info *sbi, char *buf)
2506 unsigned int *ui = (unsigned int *) (((char *) sbi) + a->u.offset);
2508 return snprintf(buf, PAGE_SIZE, "%u\n", *ui);
2511 static ssize_t sbi_ui_store(struct ext4_attr *a,
2512 struct ext4_sb_info *sbi,
2513 const char *buf, size_t count)
2515 unsigned int *ui = (unsigned int *) (((char *) sbi) + a->u.offset);
2519 ret = kstrtoul(skip_spaces(buf), 0, &t);
2526 static ssize_t reserved_clusters_show(struct ext4_attr *a,
2527 struct ext4_sb_info *sbi, char *buf)
2529 return snprintf(buf, PAGE_SIZE, "%llu\n",
2530 (unsigned long long) atomic64_read(&sbi->s_resv_clusters));
2533 static ssize_t reserved_clusters_store(struct ext4_attr *a,
2534 struct ext4_sb_info *sbi,
2535 const char *buf, size_t count)
2537 unsigned long long val;
2540 if (parse_strtoull(buf, -1ULL, &val))
2542 ret = ext4_reserve_clusters(sbi, val);
2544 return ret ? ret : count;
2547 static ssize_t trigger_test_error(struct ext4_attr *a,
2548 struct ext4_sb_info *sbi,
2549 const char *buf, size_t count)
2553 if (!capable(CAP_SYS_ADMIN))
2556 if (len && buf[len-1] == '\n')
2560 ext4_error(sbi->s_sb, "%.*s", len, buf);
2564 static ssize_t sbi_deprecated_show(struct ext4_attr *a,
2565 struct ext4_sb_info *sbi, char *buf)
2567 return snprintf(buf, PAGE_SIZE, "%d\n", a->u.deprecated_val);
2570 #define EXT4_ATTR_OFFSET(_name,_mode,_show,_store,_elname) \
2571 static struct ext4_attr ext4_attr_##_name = { \
2572 .attr = {.name = __stringify(_name), .mode = _mode }, \
2576 .offset = offsetof(struct ext4_sb_info, _elname),\
2579 #define EXT4_ATTR(name, mode, show, store) \
2580 static struct ext4_attr ext4_attr_##name = __ATTR(name, mode, show, store)
2582 #define EXT4_INFO_ATTR(name) EXT4_ATTR(name, 0444, NULL, NULL)
2583 #define EXT4_RO_ATTR(name) EXT4_ATTR(name, 0444, name##_show, NULL)
2584 #define EXT4_RW_ATTR(name) EXT4_ATTR(name, 0644, name##_show, name##_store)
2585 #define EXT4_RW_ATTR_SBI_UI(name, elname) \
2586 EXT4_ATTR_OFFSET(name, 0644, sbi_ui_show, sbi_ui_store, elname)
2587 #define ATTR_LIST(name) &ext4_attr_##name.attr
2588 #define EXT4_DEPRECATED_ATTR(_name, _val) \
2589 static struct ext4_attr ext4_attr_##_name = { \
2590 .attr = {.name = __stringify(_name), .mode = 0444 }, \
2591 .show = sbi_deprecated_show, \
2593 .deprecated_val = _val, \
2597 EXT4_RO_ATTR(delayed_allocation_blocks);
2598 EXT4_RO_ATTR(session_write_kbytes);
2599 EXT4_RO_ATTR(lifetime_write_kbytes);
2600 EXT4_RW_ATTR(reserved_clusters);
2601 EXT4_ATTR_OFFSET(inode_readahead_blks, 0644, sbi_ui_show,
2602 inode_readahead_blks_store, s_inode_readahead_blks);
2603 EXT4_RW_ATTR_SBI_UI(inode_goal, s_inode_goal);
2604 EXT4_RW_ATTR_SBI_UI(mb_stats, s_mb_stats);
2605 EXT4_RW_ATTR_SBI_UI(mb_max_to_scan, s_mb_max_to_scan);
2606 EXT4_RW_ATTR_SBI_UI(mb_min_to_scan, s_mb_min_to_scan);
2607 EXT4_RW_ATTR_SBI_UI(mb_order2_req, s_mb_order2_reqs);
2608 EXT4_RW_ATTR_SBI_UI(mb_stream_req, s_mb_stream_request);
2609 EXT4_RW_ATTR_SBI_UI(mb_group_prealloc, s_mb_group_prealloc);
2610 EXT4_DEPRECATED_ATTR(max_writeback_mb_bump, 128);
2611 EXT4_RW_ATTR_SBI_UI(extent_max_zeroout_kb, s_extent_max_zeroout_kb);
2612 EXT4_ATTR(trigger_fs_error, 0200, NULL, trigger_test_error);
2614 static struct attribute *ext4_attrs[] = {
2615 ATTR_LIST(delayed_allocation_blocks),
2616 ATTR_LIST(session_write_kbytes),
2617 ATTR_LIST(lifetime_write_kbytes),
2618 ATTR_LIST(reserved_clusters),
2619 ATTR_LIST(inode_readahead_blks),
2620 ATTR_LIST(inode_goal),
2621 ATTR_LIST(mb_stats),
2622 ATTR_LIST(mb_max_to_scan),
2623 ATTR_LIST(mb_min_to_scan),
2624 ATTR_LIST(mb_order2_req),
2625 ATTR_LIST(mb_stream_req),
2626 ATTR_LIST(mb_group_prealloc),
2627 ATTR_LIST(max_writeback_mb_bump),
2628 ATTR_LIST(extent_max_zeroout_kb),
2629 ATTR_LIST(trigger_fs_error),
2633 /* Features this copy of ext4 supports */
2634 EXT4_INFO_ATTR(lazy_itable_init);
2635 EXT4_INFO_ATTR(batched_discard);
2636 EXT4_INFO_ATTR(meta_bg_resize);
2638 static struct attribute *ext4_feat_attrs[] = {
2639 ATTR_LIST(lazy_itable_init),
2640 ATTR_LIST(batched_discard),
2641 ATTR_LIST(meta_bg_resize),
2645 static ssize_t ext4_attr_show(struct kobject *kobj,
2646 struct attribute *attr, char *buf)
2648 struct ext4_sb_info *sbi = container_of(kobj, struct ext4_sb_info,
2650 struct ext4_attr *a = container_of(attr, struct ext4_attr, attr);
2652 return a->show ? a->show(a, sbi, buf) : 0;
2655 static ssize_t ext4_attr_store(struct kobject *kobj,
2656 struct attribute *attr,
2657 const char *buf, size_t len)
2659 struct ext4_sb_info *sbi = container_of(kobj, struct ext4_sb_info,
2661 struct ext4_attr *a = container_of(attr, struct ext4_attr, attr);
2663 return a->store ? a->store(a, sbi, buf, len) : 0;
2666 static void ext4_sb_release(struct kobject *kobj)
2668 struct ext4_sb_info *sbi = container_of(kobj, struct ext4_sb_info,
2670 complete(&sbi->s_kobj_unregister);
2673 static const struct sysfs_ops ext4_attr_ops = {
2674 .show = ext4_attr_show,
2675 .store = ext4_attr_store,
2678 static struct kobj_type ext4_ktype = {
2679 .default_attrs = ext4_attrs,
2680 .sysfs_ops = &ext4_attr_ops,
2681 .release = ext4_sb_release,
2684 static void ext4_feat_release(struct kobject *kobj)
2686 complete(&ext4_feat->f_kobj_unregister);
2689 static struct kobj_type ext4_feat_ktype = {
2690 .default_attrs = ext4_feat_attrs,
2691 .sysfs_ops = &ext4_attr_ops,
2692 .release = ext4_feat_release,
2696 * Check whether this filesystem can be mounted based on
2697 * the features present and the RDONLY/RDWR mount requested.
2698 * Returns 1 if this filesystem can be mounted as requested,
2699 * 0 if it cannot be.
2701 static int ext4_feature_set_ok(struct super_block *sb, int readonly)
2703 if (EXT4_HAS_INCOMPAT_FEATURE(sb, ~EXT4_FEATURE_INCOMPAT_SUPP)) {
2704 ext4_msg(sb, KERN_ERR,
2705 "Couldn't mount because of "
2706 "unsupported optional features (%x)",
2707 (le32_to_cpu(EXT4_SB(sb)->s_es->s_feature_incompat) &
2708 ~EXT4_FEATURE_INCOMPAT_SUPP));
2715 /* Check that feature set is OK for a read-write mount */
2716 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, ~EXT4_FEATURE_RO_COMPAT_SUPP)) {
2717 ext4_msg(sb, KERN_ERR, "couldn't mount RDWR because of "
2718 "unsupported optional features (%x)",
2719 (le32_to_cpu(EXT4_SB(sb)->s_es->s_feature_ro_compat) &
2720 ~EXT4_FEATURE_RO_COMPAT_SUPP));
2724 * Large file size enabled file system can only be mounted
2725 * read-write on 32-bit systems if kernel is built with CONFIG_LBDAF
2727 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_HUGE_FILE)) {
2728 if (sizeof(blkcnt_t) < sizeof(u64)) {
2729 ext4_msg(sb, KERN_ERR, "Filesystem with huge files "
2730 "cannot be mounted RDWR without "
2735 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_BIGALLOC) &&
2736 !EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_EXTENTS)) {
2737 ext4_msg(sb, KERN_ERR,
2738 "Can't support bigalloc feature without "
2739 "extents feature\n");
2743 #ifndef CONFIG_QUOTA
2744 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_QUOTA) &&
2746 ext4_msg(sb, KERN_ERR,
2747 "Filesystem with quota feature cannot be mounted RDWR "
2748 "without CONFIG_QUOTA");
2751 #endif /* CONFIG_QUOTA */
2756 * This function is called once a day if we have errors logged
2757 * on the file system
2759 static void print_daily_error_info(unsigned long arg)
2761 struct super_block *sb = (struct super_block *) arg;
2762 struct ext4_sb_info *sbi;
2763 struct ext4_super_block *es;
2768 if (es->s_error_count)
2769 ext4_msg(sb, KERN_NOTICE, "error count: %u",
2770 le32_to_cpu(es->s_error_count));
2771 if (es->s_first_error_time) {
2772 printk(KERN_NOTICE "EXT4-fs (%s): initial error at %u: %.*s:%d",
2773 sb->s_id, le32_to_cpu(es->s_first_error_time),
2774 (int) sizeof(es->s_first_error_func),
2775 es->s_first_error_func,
2776 le32_to_cpu(es->s_first_error_line));
2777 if (es->s_first_error_ino)
2778 printk(": inode %u",
2779 le32_to_cpu(es->s_first_error_ino));
2780 if (es->s_first_error_block)
2781 printk(": block %llu", (unsigned long long)
2782 le64_to_cpu(es->s_first_error_block));
2785 if (es->s_last_error_time) {
2786 printk(KERN_NOTICE "EXT4-fs (%s): last error at %u: %.*s:%d",
2787 sb->s_id, le32_to_cpu(es->s_last_error_time),
2788 (int) sizeof(es->s_last_error_func),
2789 es->s_last_error_func,
2790 le32_to_cpu(es->s_last_error_line));
2791 if (es->s_last_error_ino)
2792 printk(": inode %u",
2793 le32_to_cpu(es->s_last_error_ino));
2794 if (es->s_last_error_block)
2795 printk(": block %llu", (unsigned long long)
2796 le64_to_cpu(es->s_last_error_block));
2799 mod_timer(&sbi->s_err_report, jiffies + 24*60*60*HZ); /* Once a day */
2802 /* Find next suitable group and run ext4_init_inode_table */
2803 static int ext4_run_li_request(struct ext4_li_request *elr)
2805 struct ext4_group_desc *gdp = NULL;
2806 ext4_group_t group, ngroups;
2807 struct super_block *sb;
2808 unsigned long timeout = 0;
2812 ngroups = EXT4_SB(sb)->s_groups_count;
2815 for (group = elr->lr_next_group; group < ngroups; group++) {
2816 gdp = ext4_get_group_desc(sb, group, NULL);
2822 if (!(gdp->bg_flags & cpu_to_le16(EXT4_BG_INODE_ZEROED)))
2826 if (group >= ngroups)
2831 ret = ext4_init_inode_table(sb, group,
2832 elr->lr_timeout ? 0 : 1);
2833 if (elr->lr_timeout == 0) {
2834 timeout = (jiffies - timeout) *
2835 elr->lr_sbi->s_li_wait_mult;
2836 elr->lr_timeout = timeout;
2838 elr->lr_next_sched = jiffies + elr->lr_timeout;
2839 elr->lr_next_group = group + 1;
2847 * Remove lr_request from the list_request and free the
2848 * request structure. Should be called with li_list_mtx held
2850 static void ext4_remove_li_request(struct ext4_li_request *elr)
2852 struct ext4_sb_info *sbi;
2859 list_del(&elr->lr_request);
2860 sbi->s_li_request = NULL;
2864 static void ext4_unregister_li_request(struct super_block *sb)
2866 mutex_lock(&ext4_li_mtx);
2867 if (!ext4_li_info) {
2868 mutex_unlock(&ext4_li_mtx);
2872 mutex_lock(&ext4_li_info->li_list_mtx);
2873 ext4_remove_li_request(EXT4_SB(sb)->s_li_request);
2874 mutex_unlock(&ext4_li_info->li_list_mtx);
2875 mutex_unlock(&ext4_li_mtx);
2878 static struct task_struct *ext4_lazyinit_task;
2881 * This is the function where ext4lazyinit thread lives. It walks
2882 * through the request list searching for next scheduled filesystem.
2883 * When such a fs is found, run the lazy initialization request
2884 * (ext4_rn_li_request) and keep track of the time spend in this
2885 * function. Based on that time we compute next schedule time of
2886 * the request. When walking through the list is complete, compute
2887 * next waking time and put itself into sleep.
2889 static int ext4_lazyinit_thread(void *arg)
2891 struct ext4_lazy_init *eli = (struct ext4_lazy_init *)arg;
2892 struct list_head *pos, *n;
2893 struct ext4_li_request *elr;
2894 unsigned long next_wakeup, cur;
2896 BUG_ON(NULL == eli);
2900 next_wakeup = MAX_JIFFY_OFFSET;
2902 mutex_lock(&eli->li_list_mtx);
2903 if (list_empty(&eli->li_request_list)) {
2904 mutex_unlock(&eli->li_list_mtx);
2908 list_for_each_safe(pos, n, &eli->li_request_list) {
2909 elr = list_entry(pos, struct ext4_li_request,
2912 if (time_after_eq(jiffies, elr->lr_next_sched)) {
2913 if (ext4_run_li_request(elr) != 0) {
2914 /* error, remove the lazy_init job */
2915 ext4_remove_li_request(elr);
2920 if (time_before(elr->lr_next_sched, next_wakeup))
2921 next_wakeup = elr->lr_next_sched;
2923 mutex_unlock(&eli->li_list_mtx);
2928 if ((time_after_eq(cur, next_wakeup)) ||
2929 (MAX_JIFFY_OFFSET == next_wakeup)) {
2934 schedule_timeout_interruptible(next_wakeup - cur);
2936 if (kthread_should_stop()) {
2937 ext4_clear_request_list();
2944 * It looks like the request list is empty, but we need
2945 * to check it under the li_list_mtx lock, to prevent any
2946 * additions into it, and of course we should lock ext4_li_mtx
2947 * to atomically free the list and ext4_li_info, because at
2948 * this point another ext4 filesystem could be registering
2951 mutex_lock(&ext4_li_mtx);
2952 mutex_lock(&eli->li_list_mtx);
2953 if (!list_empty(&eli->li_request_list)) {
2954 mutex_unlock(&eli->li_list_mtx);
2955 mutex_unlock(&ext4_li_mtx);
2958 mutex_unlock(&eli->li_list_mtx);
2959 kfree(ext4_li_info);
2960 ext4_li_info = NULL;
2961 mutex_unlock(&ext4_li_mtx);
2966 static void ext4_clear_request_list(void)
2968 struct list_head *pos, *n;
2969 struct ext4_li_request *elr;
2971 mutex_lock(&ext4_li_info->li_list_mtx);
2972 list_for_each_safe(pos, n, &ext4_li_info->li_request_list) {
2973 elr = list_entry(pos, struct ext4_li_request,
2975 ext4_remove_li_request(elr);
2977 mutex_unlock(&ext4_li_info->li_list_mtx);
2980 static int ext4_run_lazyinit_thread(void)
2982 ext4_lazyinit_task = kthread_run(ext4_lazyinit_thread,
2983 ext4_li_info, "ext4lazyinit");
2984 if (IS_ERR(ext4_lazyinit_task)) {
2985 int err = PTR_ERR(ext4_lazyinit_task);
2986 ext4_clear_request_list();
2987 kfree(ext4_li_info);
2988 ext4_li_info = NULL;
2989 printk(KERN_CRIT "EXT4-fs: error %d creating inode table "
2990 "initialization thread\n",
2994 ext4_li_info->li_state |= EXT4_LAZYINIT_RUNNING;
2999 * Check whether it make sense to run itable init. thread or not.
3000 * If there is at least one uninitialized inode table, return
3001 * corresponding group number, else the loop goes through all
3002 * groups and return total number of groups.
3004 static ext4_group_t ext4_has_uninit_itable(struct super_block *sb)
3006 ext4_group_t group, ngroups = EXT4_SB(sb)->s_groups_count;
3007 struct ext4_group_desc *gdp = NULL;
3009 for (group = 0; group < ngroups; group++) {
3010 gdp = ext4_get_group_desc(sb, group, NULL);
3014 if (!(gdp->bg_flags & cpu_to_le16(EXT4_BG_INODE_ZEROED)))
3021 static int ext4_li_info_new(void)
3023 struct ext4_lazy_init *eli = NULL;
3025 eli = kzalloc(sizeof(*eli), GFP_KERNEL);
3029 INIT_LIST_HEAD(&eli->li_request_list);
3030 mutex_init(&eli->li_list_mtx);
3032 eli->li_state |= EXT4_LAZYINIT_QUIT;
3039 static struct ext4_li_request *ext4_li_request_new(struct super_block *sb,
3042 struct ext4_sb_info *sbi = EXT4_SB(sb);
3043 struct ext4_li_request *elr;
3046 elr = kzalloc(sizeof(*elr), GFP_KERNEL);
3052 elr->lr_next_group = start;
3055 * Randomize first schedule time of the request to
3056 * spread the inode table initialization requests
3059 get_random_bytes(&rnd, sizeof(rnd));
3060 elr->lr_next_sched = jiffies + (unsigned long)rnd %
3061 (EXT4_DEF_LI_MAX_START_DELAY * HZ);
3066 int ext4_register_li_request(struct super_block *sb,
3067 ext4_group_t first_not_zeroed)
3069 struct ext4_sb_info *sbi = EXT4_SB(sb);
3070 struct ext4_li_request *elr = NULL;
3071 ext4_group_t ngroups = EXT4_SB(sb)->s_groups_count;
3074 mutex_lock(&ext4_li_mtx);
3075 if (sbi->s_li_request != NULL) {
3077 * Reset timeout so it can be computed again, because
3078 * s_li_wait_mult might have changed.
3080 sbi->s_li_request->lr_timeout = 0;
3084 if (first_not_zeroed == ngroups ||
3085 (sb->s_flags & MS_RDONLY) ||
3086 !test_opt(sb, INIT_INODE_TABLE))
3089 elr = ext4_li_request_new(sb, first_not_zeroed);
3095 if (NULL == ext4_li_info) {
3096 ret = ext4_li_info_new();
3101 mutex_lock(&ext4_li_info->li_list_mtx);
3102 list_add(&elr->lr_request, &ext4_li_info->li_request_list);
3103 mutex_unlock(&ext4_li_info->li_list_mtx);
3105 sbi->s_li_request = elr;
3107 * set elr to NULL here since it has been inserted to
3108 * the request_list and the removal and free of it is
3109 * handled by ext4_clear_request_list from now on.
3113 if (!(ext4_li_info->li_state & EXT4_LAZYINIT_RUNNING)) {
3114 ret = ext4_run_lazyinit_thread();
3119 mutex_unlock(&ext4_li_mtx);
3126 * We do not need to lock anything since this is called on
3129 static void ext4_destroy_lazyinit_thread(void)
3132 * If thread exited earlier
3133 * there's nothing to be done.
3135 if (!ext4_li_info || !ext4_lazyinit_task)
3138 kthread_stop(ext4_lazyinit_task);
3141 static int set_journal_csum_feature_set(struct super_block *sb)
3144 int compat, incompat;
3145 struct ext4_sb_info *sbi = EXT4_SB(sb);
3147 if (EXT4_HAS_RO_COMPAT_FEATURE(sb,
3148 EXT4_FEATURE_RO_COMPAT_METADATA_CSUM)) {
3149 /* journal checksum v2 */
3151 incompat = JBD2_FEATURE_INCOMPAT_CSUM_V2;
3153 /* journal checksum v1 */
3154 compat = JBD2_FEATURE_COMPAT_CHECKSUM;
3158 if (test_opt(sb, JOURNAL_ASYNC_COMMIT)) {
3159 ret = jbd2_journal_set_features(sbi->s_journal,
3161 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT |
3163 } else if (test_opt(sb, JOURNAL_CHECKSUM)) {
3164 ret = jbd2_journal_set_features(sbi->s_journal,
3167 jbd2_journal_clear_features(sbi->s_journal, 0, 0,
3168 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT);
3170 jbd2_journal_clear_features(sbi->s_journal,
3171 JBD2_FEATURE_COMPAT_CHECKSUM, 0,
3172 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT |
3173 JBD2_FEATURE_INCOMPAT_CSUM_V2);
3180 * Note: calculating the overhead so we can be compatible with
3181 * historical BSD practice is quite difficult in the face of
3182 * clusters/bigalloc. This is because multiple metadata blocks from
3183 * different block group can end up in the same allocation cluster.
3184 * Calculating the exact overhead in the face of clustered allocation
3185 * requires either O(all block bitmaps) in memory or O(number of block
3186 * groups**2) in time. We will still calculate the superblock for
3187 * older file systems --- and if we come across with a bigalloc file
3188 * system with zero in s_overhead_clusters the estimate will be close to
3189 * correct especially for very large cluster sizes --- but for newer
3190 * file systems, it's better to calculate this figure once at mkfs
3191 * time, and store it in the superblock. If the superblock value is
3192 * present (even for non-bigalloc file systems), we will use it.
3194 static int count_overhead(struct super_block *sb, ext4_group_t grp,
3197 struct ext4_sb_info *sbi = EXT4_SB(sb);
3198 struct ext4_group_desc *gdp;
3199 ext4_fsblk_t first_block, last_block, b;
3200 ext4_group_t i, ngroups = ext4_get_groups_count(sb);
3201 int s, j, count = 0;
3203 if (!EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_BIGALLOC))
3204 return (ext4_bg_has_super(sb, grp) + ext4_bg_num_gdb(sb, grp) +
3205 sbi->s_itb_per_group + 2);
3207 first_block = le32_to_cpu(sbi->s_es->s_first_data_block) +
3208 (grp * EXT4_BLOCKS_PER_GROUP(sb));
3209 last_block = first_block + EXT4_BLOCKS_PER_GROUP(sb) - 1;
3210 for (i = 0; i < ngroups; i++) {
3211 gdp = ext4_get_group_desc(sb, i, NULL);
3212 b = ext4_block_bitmap(sb, gdp);
3213 if (b >= first_block && b <= last_block) {
3214 ext4_set_bit(EXT4_B2C(sbi, b - first_block), buf);
3217 b = ext4_inode_bitmap(sb, gdp);
3218 if (b >= first_block && b <= last_block) {
3219 ext4_set_bit(EXT4_B2C(sbi, b - first_block), buf);
3222 b = ext4_inode_table(sb, gdp);
3223 if (b >= first_block && b + sbi->s_itb_per_group <= last_block)
3224 for (j = 0; j < sbi->s_itb_per_group; j++, b++) {
3225 int c = EXT4_B2C(sbi, b - first_block);
3226 ext4_set_bit(c, buf);
3232 if (ext4_bg_has_super(sb, grp)) {
3233 ext4_set_bit(s++, buf);
3236 for (j = ext4_bg_num_gdb(sb, grp); j > 0; j--) {
3237 ext4_set_bit(EXT4_B2C(sbi, s++), buf);
3243 return EXT4_CLUSTERS_PER_GROUP(sb) -
3244 ext4_count_free(buf, EXT4_CLUSTERS_PER_GROUP(sb) / 8);
3248 * Compute the overhead and stash it in sbi->s_overhead
3250 int ext4_calculate_overhead(struct super_block *sb)
3252 struct ext4_sb_info *sbi = EXT4_SB(sb);
3253 struct ext4_super_block *es = sbi->s_es;
3254 ext4_group_t i, ngroups = ext4_get_groups_count(sb);
3255 ext4_fsblk_t overhead = 0;
3256 char *buf = (char *) get_zeroed_page(GFP_KERNEL);
3262 * Compute the overhead (FS structures). This is constant
3263 * for a given filesystem unless the number of block groups
3264 * changes so we cache the previous value until it does.
3268 * All of the blocks before first_data_block are overhead
3270 overhead = EXT4_B2C(sbi, le32_to_cpu(es->s_first_data_block));
3273 * Add the overhead found in each block group
3275 for (i = 0; i < ngroups; i++) {
3278 blks = count_overhead(sb, i, buf);
3281 memset(buf, 0, PAGE_SIZE);
3284 /* Add the journal blocks as well */
3286 overhead += EXT4_NUM_B2C(sbi, sbi->s_journal->j_maxlen);
3288 sbi->s_overhead = overhead;
3290 free_page((unsigned long) buf);
3295 static ext4_fsblk_t ext4_calculate_resv_clusters(struct ext4_sb_info *sbi)
3297 ext4_fsblk_t resv_clusters;
3300 * By default we reserve 2% or 4096 clusters, whichever is smaller.
3301 * This should cover the situations where we can not afford to run
3302 * out of space like for example punch hole, or converting
3303 * uninitialized extents in delalloc path. In most cases such
3304 * allocation would require 1, or 2 blocks, higher numbers are
3307 resv_clusters = ext4_blocks_count(sbi->s_es) >> sbi->s_cluster_bits;
3309 do_div(resv_clusters, 50);
3310 resv_clusters = min_t(ext4_fsblk_t, resv_clusters, 4096);
3312 return resv_clusters;
3316 static int ext4_reserve_clusters(struct ext4_sb_info *sbi, ext4_fsblk_t count)
3318 ext4_fsblk_t clusters = ext4_blocks_count(sbi->s_es) >>
3319 sbi->s_cluster_bits;
3321 if (count >= clusters)
3324 atomic64_set(&sbi->s_resv_clusters, count);
3328 static int ext4_fill_super(struct super_block *sb, void *data, int silent)
3330 char *orig_data = kstrdup(data, GFP_KERNEL);
3331 struct buffer_head *bh;
3332 struct ext4_super_block *es = NULL;
3333 struct ext4_sb_info *sbi;
3335 ext4_fsblk_t sb_block = get_sb_block(&data);
3336 ext4_fsblk_t logical_sb_block;
3337 unsigned long offset = 0;
3338 unsigned long journal_devnum = 0;
3339 unsigned long def_mount_opts;
3344 int blocksize, clustersize;
3345 unsigned int db_count;
3347 int needs_recovery, has_huge_files, has_bigalloc;
3350 unsigned int journal_ioprio = DEFAULT_JOURNAL_IOPRIO;
3351 ext4_group_t first_not_zeroed;
3353 sbi = kzalloc(sizeof(*sbi), GFP_KERNEL);
3357 sbi->s_blockgroup_lock =
3358 kzalloc(sizeof(struct blockgroup_lock), GFP_KERNEL);
3359 if (!sbi->s_blockgroup_lock) {
3363 sb->s_fs_info = sbi;
3365 sbi->s_inode_readahead_blks = EXT4_DEF_INODE_READAHEAD_BLKS;
3366 sbi->s_sb_block = sb_block;
3367 if (sb->s_bdev->bd_part)
3368 sbi->s_sectors_written_start =
3369 part_stat_read(sb->s_bdev->bd_part, sectors[1]);
3371 /* Cleanup superblock name */
3372 for (cp = sb->s_id; (cp = strchr(cp, '/'));)
3375 /* -EINVAL is default */
3377 blocksize = sb_min_blocksize(sb, EXT4_MIN_BLOCK_SIZE);
3379 ext4_msg(sb, KERN_ERR, "unable to set blocksize");
3384 * The ext4 superblock will not be buffer aligned for other than 1kB
3385 * block sizes. We need to calculate the offset from buffer start.
3387 if (blocksize != EXT4_MIN_BLOCK_SIZE) {
3388 logical_sb_block = sb_block * EXT4_MIN_BLOCK_SIZE;
3389 offset = do_div(logical_sb_block, blocksize);
3391 logical_sb_block = sb_block;
3394 if (!(bh = sb_bread(sb, logical_sb_block))) {
3395 ext4_msg(sb, KERN_ERR, "unable to read superblock");
3399 * Note: s_es must be initialized as soon as possible because
3400 * some ext4 macro-instructions depend on its value
3402 es = (struct ext4_super_block *) (bh->b_data + offset);
3404 sb->s_magic = le16_to_cpu(es->s_magic);
3405 if (sb->s_magic != EXT4_SUPER_MAGIC)
3407 sbi->s_kbytes_written = le64_to_cpu(es->s_kbytes_written);
3409 /* Warn if metadata_csum and gdt_csum are both set. */
3410 if (EXT4_HAS_RO_COMPAT_FEATURE(sb,
3411 EXT4_FEATURE_RO_COMPAT_METADATA_CSUM) &&
3412 EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_GDT_CSUM))
3413 ext4_warning(sb, KERN_INFO "metadata_csum and uninit_bg are "
3414 "redundant flags; please run fsck.");
3416 /* Check for a known checksum algorithm */
3417 if (!ext4_verify_csum_type(sb, es)) {
3418 ext4_msg(sb, KERN_ERR, "VFS: Found ext4 filesystem with "
3419 "unknown checksum algorithm.");
3424 /* Load the checksum driver */
3425 if (EXT4_HAS_RO_COMPAT_FEATURE(sb,
3426 EXT4_FEATURE_RO_COMPAT_METADATA_CSUM)) {
3427 sbi->s_chksum_driver = crypto_alloc_shash("crc32c", 0, 0);
3428 if (IS_ERR(sbi->s_chksum_driver)) {
3429 ext4_msg(sb, KERN_ERR, "Cannot load crc32c driver.");
3430 ret = PTR_ERR(sbi->s_chksum_driver);
3431 sbi->s_chksum_driver = NULL;
3436 /* Check superblock checksum */
3437 if (!ext4_superblock_csum_verify(sb, es)) {
3438 ext4_msg(sb, KERN_ERR, "VFS: Found ext4 filesystem with "
3439 "invalid superblock checksum. Run e2fsck?");
3444 /* Precompute checksum seed for all metadata */
3445 if (EXT4_HAS_RO_COMPAT_FEATURE(sb,
3446 EXT4_FEATURE_RO_COMPAT_METADATA_CSUM))
3447 sbi->s_csum_seed = ext4_chksum(sbi, ~0, es->s_uuid,
3448 sizeof(es->s_uuid));
3450 /* Set defaults before we parse the mount options */
3451 def_mount_opts = le32_to_cpu(es->s_default_mount_opts);
3452 set_opt(sb, INIT_INODE_TABLE);
3453 if (def_mount_opts & EXT4_DEFM_DEBUG)
3455 if (def_mount_opts & EXT4_DEFM_BSDGROUPS)
3457 if (def_mount_opts & EXT4_DEFM_UID16)
3458 set_opt(sb, NO_UID32);
3459 /* xattr user namespace & acls are now defaulted on */
3460 set_opt(sb, XATTR_USER);
3461 #ifdef CONFIG_EXT4_FS_POSIX_ACL
3462 set_opt(sb, POSIX_ACL);
3464 if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_DATA)
3465 set_opt(sb, JOURNAL_DATA);
3466 else if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_ORDERED)
3467 set_opt(sb, ORDERED_DATA);
3468 else if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_WBACK)
3469 set_opt(sb, WRITEBACK_DATA);
3471 if (le16_to_cpu(sbi->s_es->s_errors) == EXT4_ERRORS_PANIC)
3472 set_opt(sb, ERRORS_PANIC);
3473 else if (le16_to_cpu(sbi->s_es->s_errors) == EXT4_ERRORS_CONTINUE)
3474 set_opt(sb, ERRORS_CONT);
3476 set_opt(sb, ERRORS_RO);
3477 if (def_mount_opts & EXT4_DEFM_BLOCK_VALIDITY)
3478 set_opt(sb, BLOCK_VALIDITY);
3479 if (def_mount_opts & EXT4_DEFM_DISCARD)
3480 set_opt(sb, DISCARD);
3482 sbi->s_resuid = make_kuid(&init_user_ns, le16_to_cpu(es->s_def_resuid));
3483 sbi->s_resgid = make_kgid(&init_user_ns, le16_to_cpu(es->s_def_resgid));
3484 sbi->s_commit_interval = JBD2_DEFAULT_MAX_COMMIT_AGE * HZ;
3485 sbi->s_min_batch_time = EXT4_DEF_MIN_BATCH_TIME;
3486 sbi->s_max_batch_time = EXT4_DEF_MAX_BATCH_TIME;
3488 if ((def_mount_opts & EXT4_DEFM_NOBARRIER) == 0)
3489 set_opt(sb, BARRIER);
3492 * enable delayed allocation by default
3493 * Use -o nodelalloc to turn it off
3495 if (!IS_EXT3_SB(sb) && !IS_EXT2_SB(sb) &&
3496 ((def_mount_opts & EXT4_DEFM_NODELALLOC) == 0))
3497 set_opt(sb, DELALLOC);
3500 * set default s_li_wait_mult for lazyinit, for the case there is
3501 * no mount option specified.
3503 sbi->s_li_wait_mult = EXT4_DEF_LI_WAIT_MULT;
3505 if (!parse_options((char *) sbi->s_es->s_mount_opts, sb,
3506 &journal_devnum, &journal_ioprio, 0)) {
3507 ext4_msg(sb, KERN_WARNING,
3508 "failed to parse options in superblock: %s",
3509 sbi->s_es->s_mount_opts);
3511 sbi->s_def_mount_opt = sbi->s_mount_opt;
3512 if (!parse_options((char *) data, sb, &journal_devnum,
3513 &journal_ioprio, 0))
3516 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA) {
3517 printk_once(KERN_WARNING "EXT4-fs: Warning: mounting "
3518 "with data=journal disables delayed "
3519 "allocation and O_DIRECT support!\n");
3520 if (test_opt2(sb, EXPLICIT_DELALLOC)) {
3521 ext4_msg(sb, KERN_ERR, "can't mount with "
3522 "both data=journal and delalloc");
3525 if (test_opt(sb, DIOREAD_NOLOCK)) {
3526 ext4_msg(sb, KERN_ERR, "can't mount with "
3527 "both data=journal and dioread_nolock");
3530 if (test_opt(sb, DELALLOC))
3531 clear_opt(sb, DELALLOC);
3534 sb->s_flags = (sb->s_flags & ~MS_POSIXACL) |
3535 (test_opt(sb, POSIX_ACL) ? MS_POSIXACL : 0);
3537 if (le32_to_cpu(es->s_rev_level) == EXT4_GOOD_OLD_REV &&
3538 (EXT4_HAS_COMPAT_FEATURE(sb, ~0U) ||
3539 EXT4_HAS_RO_COMPAT_FEATURE(sb, ~0U) ||
3540 EXT4_HAS_INCOMPAT_FEATURE(sb, ~0U)))
3541 ext4_msg(sb, KERN_WARNING,
3542 "feature flags set on rev 0 fs, "
3543 "running e2fsck is recommended");
3545 if (IS_EXT2_SB(sb)) {
3546 if (ext2_feature_set_ok(sb))
3547 ext4_msg(sb, KERN_INFO, "mounting ext2 file system "
3548 "using the ext4 subsystem");
3550 ext4_msg(sb, KERN_ERR, "couldn't mount as ext2 due "
3551 "to feature incompatibilities");
3556 if (IS_EXT3_SB(sb)) {
3557 if (ext3_feature_set_ok(sb))
3558 ext4_msg(sb, KERN_INFO, "mounting ext3 file system "
3559 "using the ext4 subsystem");
3561 ext4_msg(sb, KERN_ERR, "couldn't mount as ext3 due "
3562 "to feature incompatibilities");
3568 * Check feature flags regardless of the revision level, since we
3569 * previously didn't change the revision level when setting the flags,
3570 * so there is a chance incompat flags are set on a rev 0 filesystem.
3572 if (!ext4_feature_set_ok(sb, (sb->s_flags & MS_RDONLY)))
3575 blocksize = BLOCK_SIZE << le32_to_cpu(es->s_log_block_size);
3576 if (blocksize < EXT4_MIN_BLOCK_SIZE ||
3577 blocksize > EXT4_MAX_BLOCK_SIZE) {
3578 ext4_msg(sb, KERN_ERR,
3579 "Unsupported filesystem blocksize %d", blocksize);
3583 if (sb->s_blocksize != blocksize) {
3584 /* Validate the filesystem blocksize */
3585 if (!sb_set_blocksize(sb, blocksize)) {
3586 ext4_msg(sb, KERN_ERR, "bad block size %d",
3592 logical_sb_block = sb_block * EXT4_MIN_BLOCK_SIZE;
3593 offset = do_div(logical_sb_block, blocksize);
3594 bh = sb_bread(sb, logical_sb_block);
3596 ext4_msg(sb, KERN_ERR,
3597 "Can't read superblock on 2nd try");
3600 es = (struct ext4_super_block *)(bh->b_data + offset);
3602 if (es->s_magic != cpu_to_le16(EXT4_SUPER_MAGIC)) {
3603 ext4_msg(sb, KERN_ERR,
3604 "Magic mismatch, very weird!");
3609 has_huge_files = EXT4_HAS_RO_COMPAT_FEATURE(sb,
3610 EXT4_FEATURE_RO_COMPAT_HUGE_FILE);
3611 sbi->s_bitmap_maxbytes = ext4_max_bitmap_size(sb->s_blocksize_bits,
3613 sb->s_maxbytes = ext4_max_size(sb->s_blocksize_bits, has_huge_files);
3615 if (le32_to_cpu(es->s_rev_level) == EXT4_GOOD_OLD_REV) {
3616 sbi->s_inode_size = EXT4_GOOD_OLD_INODE_SIZE;
3617 sbi->s_first_ino = EXT4_GOOD_OLD_FIRST_INO;
3619 sbi->s_inode_size = le16_to_cpu(es->s_inode_size);
3620 sbi->s_first_ino = le32_to_cpu(es->s_first_ino);
3621 if ((sbi->s_inode_size < EXT4_GOOD_OLD_INODE_SIZE) ||
3622 (!is_power_of_2(sbi->s_inode_size)) ||
3623 (sbi->s_inode_size > blocksize)) {
3624 ext4_msg(sb, KERN_ERR,
3625 "unsupported inode size: %d",
3629 if (sbi->s_inode_size > EXT4_GOOD_OLD_INODE_SIZE)
3630 sb->s_time_gran = 1 << (EXT4_EPOCH_BITS - 2);
3633 sbi->s_desc_size = le16_to_cpu(es->s_desc_size);
3634 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_64BIT)) {
3635 if (sbi->s_desc_size < EXT4_MIN_DESC_SIZE_64BIT ||
3636 sbi->s_desc_size > EXT4_MAX_DESC_SIZE ||
3637 !is_power_of_2(sbi->s_desc_size)) {
3638 ext4_msg(sb, KERN_ERR,
3639 "unsupported descriptor size %lu",
3644 sbi->s_desc_size = EXT4_MIN_DESC_SIZE;
3646 sbi->s_blocks_per_group = le32_to_cpu(es->s_blocks_per_group);
3647 sbi->s_inodes_per_group = le32_to_cpu(es->s_inodes_per_group);
3648 if (EXT4_INODE_SIZE(sb) == 0 || EXT4_INODES_PER_GROUP(sb) == 0)
3651 sbi->s_inodes_per_block = blocksize / EXT4_INODE_SIZE(sb);
3652 if (sbi->s_inodes_per_block == 0)
3654 sbi->s_itb_per_group = sbi->s_inodes_per_group /
3655 sbi->s_inodes_per_block;
3656 sbi->s_desc_per_block = blocksize / EXT4_DESC_SIZE(sb);
3658 sbi->s_mount_state = le16_to_cpu(es->s_state);
3659 sbi->s_addr_per_block_bits = ilog2(EXT4_ADDR_PER_BLOCK(sb));
3660 sbi->s_desc_per_block_bits = ilog2(EXT4_DESC_PER_BLOCK(sb));
3662 for (i = 0; i < 4; i++)
3663 sbi->s_hash_seed[i] = le32_to_cpu(es->s_hash_seed[i]);
3664 sbi->s_def_hash_version = es->s_def_hash_version;
3665 i = le32_to_cpu(es->s_flags);
3666 if (i & EXT2_FLAGS_UNSIGNED_HASH)
3667 sbi->s_hash_unsigned = 3;
3668 else if ((i & EXT2_FLAGS_SIGNED_HASH) == 0) {
3669 #ifdef __CHAR_UNSIGNED__
3670 es->s_flags |= cpu_to_le32(EXT2_FLAGS_UNSIGNED_HASH);
3671 sbi->s_hash_unsigned = 3;
3673 es->s_flags |= cpu_to_le32(EXT2_FLAGS_SIGNED_HASH);
3677 /* Handle clustersize */
3678 clustersize = BLOCK_SIZE << le32_to_cpu(es->s_log_cluster_size);
3679 has_bigalloc = EXT4_HAS_RO_COMPAT_FEATURE(sb,
3680 EXT4_FEATURE_RO_COMPAT_BIGALLOC);
3682 if (clustersize < blocksize) {
3683 ext4_msg(sb, KERN_ERR,
3684 "cluster size (%d) smaller than "
3685 "block size (%d)", clustersize, blocksize);
3688 sbi->s_cluster_bits = le32_to_cpu(es->s_log_cluster_size) -
3689 le32_to_cpu(es->s_log_block_size);
3690 sbi->s_clusters_per_group =
3691 le32_to_cpu(es->s_clusters_per_group);
3692 if (sbi->s_clusters_per_group > blocksize * 8) {
3693 ext4_msg(sb, KERN_ERR,
3694 "#clusters per group too big: %lu",
3695 sbi->s_clusters_per_group);
3698 if (sbi->s_blocks_per_group !=
3699 (sbi->s_clusters_per_group * (clustersize / blocksize))) {
3700 ext4_msg(sb, KERN_ERR, "blocks per group (%lu) and "
3701 "clusters per group (%lu) inconsistent",
3702 sbi->s_blocks_per_group,
3703 sbi->s_clusters_per_group);
3707 if (clustersize != blocksize) {
3708 ext4_warning(sb, "fragment/cluster size (%d) != "
3709 "block size (%d)", clustersize,
3711 clustersize = blocksize;
3713 if (sbi->s_blocks_per_group > blocksize * 8) {
3714 ext4_msg(sb, KERN_ERR,
3715 "#blocks per group too big: %lu",
3716 sbi->s_blocks_per_group);
3719 sbi->s_clusters_per_group = sbi->s_blocks_per_group;
3720 sbi->s_cluster_bits = 0;
3722 sbi->s_cluster_ratio = clustersize / blocksize;
3724 if (sbi->s_inodes_per_group > blocksize * 8) {
3725 ext4_msg(sb, KERN_ERR,
3726 "#inodes per group too big: %lu",
3727 sbi->s_inodes_per_group);
3731 /* Do we have standard group size of clustersize * 8 blocks ? */
3732 if (sbi->s_blocks_per_group == clustersize << 3)
3733 set_opt2(sb, STD_GROUP_SIZE);
3736 * Test whether we have more sectors than will fit in sector_t,
3737 * and whether the max offset is addressable by the page cache.
3739 err = generic_check_addressable(sb->s_blocksize_bits,
3740 ext4_blocks_count(es));
3742 ext4_msg(sb, KERN_ERR, "filesystem"
3743 " too large to mount safely on this system");
3744 if (sizeof(sector_t) < 8)
3745 ext4_msg(sb, KERN_WARNING, "CONFIG_LBDAF not enabled");
3749 if (EXT4_BLOCKS_PER_GROUP(sb) == 0)
3752 /* check blocks count against device size */
3753 blocks_count = sb->s_bdev->bd_inode->i_size >> sb->s_blocksize_bits;
3754 if (blocks_count && ext4_blocks_count(es) > blocks_count) {
3755 ext4_msg(sb, KERN_WARNING, "bad geometry: block count %llu "
3756 "exceeds size of device (%llu blocks)",
3757 ext4_blocks_count(es), blocks_count);
3762 * It makes no sense for the first data block to be beyond the end
3763 * of the filesystem.
3765 if (le32_to_cpu(es->s_first_data_block) >= ext4_blocks_count(es)) {
3766 ext4_msg(sb, KERN_WARNING, "bad geometry: first data "
3767 "block %u is beyond end of filesystem (%llu)",
3768 le32_to_cpu(es->s_first_data_block),
3769 ext4_blocks_count(es));
3772 blocks_count = (ext4_blocks_count(es) -
3773 le32_to_cpu(es->s_first_data_block) +
3774 EXT4_BLOCKS_PER_GROUP(sb) - 1);
3775 do_div(blocks_count, EXT4_BLOCKS_PER_GROUP(sb));
3776 if (blocks_count > ((uint64_t)1<<32) - EXT4_DESC_PER_BLOCK(sb)) {
3777 ext4_msg(sb, KERN_WARNING, "groups count too large: %u "
3778 "(block count %llu, first data block %u, "
3779 "blocks per group %lu)", sbi->s_groups_count,
3780 ext4_blocks_count(es),
3781 le32_to_cpu(es->s_first_data_block),
3782 EXT4_BLOCKS_PER_GROUP(sb));
3785 sbi->s_groups_count = blocks_count;
3786 sbi->s_blockfile_groups = min_t(ext4_group_t, sbi->s_groups_count,
3787 (EXT4_MAX_BLOCK_FILE_PHYS / EXT4_BLOCKS_PER_GROUP(sb)));
3788 db_count = (sbi->s_groups_count + EXT4_DESC_PER_BLOCK(sb) - 1) /
3789 EXT4_DESC_PER_BLOCK(sb);
3790 sbi->s_group_desc = ext4_kvmalloc(db_count *
3791 sizeof(struct buffer_head *),
3793 if (sbi->s_group_desc == NULL) {
3794 ext4_msg(sb, KERN_ERR, "not enough memory");
3800 sbi->s_proc = proc_mkdir(sb->s_id, ext4_proc_root);
3803 proc_create_data("options", S_IRUGO, sbi->s_proc,
3804 &ext4_seq_options_fops, sb);
3806 bgl_lock_init(sbi->s_blockgroup_lock);
3808 for (i = 0; i < db_count; i++) {
3809 block = descriptor_loc(sb, logical_sb_block, i);
3810 sbi->s_group_desc[i] = sb_bread(sb, block);
3811 if (!sbi->s_group_desc[i]) {
3812 ext4_msg(sb, KERN_ERR,
3813 "can't read group descriptor %d", i);
3818 if (!ext4_check_descriptors(sb, &first_not_zeroed)) {
3819 ext4_msg(sb, KERN_ERR, "group descriptors corrupted!");
3822 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_FLEX_BG))
3823 if (!ext4_fill_flex_info(sb)) {
3824 ext4_msg(sb, KERN_ERR,
3825 "unable to initialize "
3826 "flex_bg meta info!");
3830 sbi->s_gdb_count = db_count;
3831 get_random_bytes(&sbi->s_next_generation, sizeof(u32));
3832 spin_lock_init(&sbi->s_next_gen_lock);
3834 init_timer(&sbi->s_err_report);
3835 sbi->s_err_report.function = print_daily_error_info;
3836 sbi->s_err_report.data = (unsigned long) sb;
3838 /* Register extent status tree shrinker */
3839 ext4_es_register_shrinker(sbi);
3841 err = percpu_counter_init(&sbi->s_freeclusters_counter,
3842 ext4_count_free_clusters(sb));
3844 err = percpu_counter_init(&sbi->s_freeinodes_counter,
3845 ext4_count_free_inodes(sb));
3848 err = percpu_counter_init(&sbi->s_dirs_counter,
3849 ext4_count_dirs(sb));
3852 err = percpu_counter_init(&sbi->s_dirtyclusters_counter, 0);
3855 err = percpu_counter_init(&sbi->s_extent_cache_cnt, 0);
3858 ext4_msg(sb, KERN_ERR, "insufficient memory");
3862 sbi->s_stripe = ext4_get_stripe_size(sbi);
3863 sbi->s_extent_max_zeroout_kb = 32;
3866 * set up enough so that it can read an inode
3868 if (!test_opt(sb, NOLOAD) &&
3869 EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL))
3870 sb->s_op = &ext4_sops;
3872 sb->s_op = &ext4_nojournal_sops;
3873 sb->s_export_op = &ext4_export_ops;
3874 sb->s_xattr = ext4_xattr_handlers;
3876 sb->dq_op = &ext4_quota_operations;
3877 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_QUOTA))
3878 sb->s_qcop = &ext4_qctl_sysfile_operations;
3880 sb->s_qcop = &ext4_qctl_operations;
3882 memcpy(sb->s_uuid, es->s_uuid, sizeof(es->s_uuid));
3884 INIT_LIST_HEAD(&sbi->s_orphan); /* unlinked but open files */
3885 mutex_init(&sbi->s_orphan_lock);
3889 needs_recovery = (es->s_last_orphan != 0 ||
3890 EXT4_HAS_INCOMPAT_FEATURE(sb,
3891 EXT4_FEATURE_INCOMPAT_RECOVER));
3893 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_MMP) &&
3894 !(sb->s_flags & MS_RDONLY))
3895 if (ext4_multi_mount_protect(sb, le64_to_cpu(es->s_mmp_block)))
3899 * The first inode we look at is the journal inode. Don't try
3900 * root first: it may be modified in the journal!
3902 if (!test_opt(sb, NOLOAD) &&
3903 EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL)) {
3904 if (ext4_load_journal(sb, es, journal_devnum))
3906 } else if (test_opt(sb, NOLOAD) && !(sb->s_flags & MS_RDONLY) &&
3907 EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER)) {
3908 ext4_msg(sb, KERN_ERR, "required journal recovery "
3909 "suppressed and not mounted read-only");
3910 goto failed_mount_wq;
3912 clear_opt(sb, DATA_FLAGS);
3913 sbi->s_journal = NULL;
3918 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_64BIT) &&
3919 !jbd2_journal_set_features(EXT4_SB(sb)->s_journal, 0, 0,
3920 JBD2_FEATURE_INCOMPAT_64BIT)) {
3921 ext4_msg(sb, KERN_ERR, "Failed to set 64-bit journal feature");
3922 goto failed_mount_wq;
3925 if (!set_journal_csum_feature_set(sb)) {
3926 ext4_msg(sb, KERN_ERR, "Failed to set journal checksum "
3928 goto failed_mount_wq;
3931 /* We have now updated the journal if required, so we can
3932 * validate the data journaling mode. */
3933 switch (test_opt(sb, DATA_FLAGS)) {
3935 /* No mode set, assume a default based on the journal
3936 * capabilities: ORDERED_DATA if the journal can
3937 * cope, else JOURNAL_DATA
3939 if (jbd2_journal_check_available_features
3940 (sbi->s_journal, 0, 0, JBD2_FEATURE_INCOMPAT_REVOKE))
3941 set_opt(sb, ORDERED_DATA);
3943 set_opt(sb, JOURNAL_DATA);
3946 case EXT4_MOUNT_ORDERED_DATA:
3947 case EXT4_MOUNT_WRITEBACK_DATA:
3948 if (!jbd2_journal_check_available_features
3949 (sbi->s_journal, 0, 0, JBD2_FEATURE_INCOMPAT_REVOKE)) {
3950 ext4_msg(sb, KERN_ERR, "Journal does not support "
3951 "requested data journaling mode");
3952 goto failed_mount_wq;
3957 set_task_ioprio(sbi->s_journal->j_task, journal_ioprio);
3959 sbi->s_journal->j_commit_callback = ext4_journal_commit_callback;
3962 * The journal may have updated the bg summary counts, so we
3963 * need to update the global counters.
3965 percpu_counter_set(&sbi->s_freeclusters_counter,
3966 ext4_count_free_clusters(sb));
3967 percpu_counter_set(&sbi->s_freeinodes_counter,
3968 ext4_count_free_inodes(sb));
3969 percpu_counter_set(&sbi->s_dirs_counter,
3970 ext4_count_dirs(sb));
3971 percpu_counter_set(&sbi->s_dirtyclusters_counter, 0);
3975 * Get the # of file system overhead blocks from the
3976 * superblock if present.
3978 if (es->s_overhead_clusters)
3979 sbi->s_overhead = le32_to_cpu(es->s_overhead_clusters);
3981 err = ext4_calculate_overhead(sb);
3983 goto failed_mount_wq;
3987 * The maximum number of concurrent works can be high and
3988 * concurrency isn't really necessary. Limit it to 1.
3990 EXT4_SB(sb)->rsv_conversion_wq =
3991 alloc_workqueue("ext4-rsv-conversion", WQ_MEM_RECLAIM | WQ_UNBOUND, 1);
3992 if (!EXT4_SB(sb)->rsv_conversion_wq) {
3993 printk(KERN_ERR "EXT4-fs: failed to create workqueue\n");
3998 EXT4_SB(sb)->unrsv_conversion_wq =
3999 alloc_workqueue("ext4-unrsv-conversion", WQ_MEM_RECLAIM | WQ_UNBOUND, 1);
4000 if (!EXT4_SB(sb)->unrsv_conversion_wq) {
4001 printk(KERN_ERR "EXT4-fs: failed to create workqueue\n");
4007 * The jbd2_journal_load will have done any necessary log recovery,
4008 * so we can safely mount the rest of the filesystem now.
4011 root = ext4_iget(sb, EXT4_ROOT_INO);
4013 ext4_msg(sb, KERN_ERR, "get root inode failed");
4014 ret = PTR_ERR(root);
4018 if (!S_ISDIR(root->i_mode) || !root->i_blocks || !root->i_size) {
4019 ext4_msg(sb, KERN_ERR, "corrupt root inode, run e2fsck");
4023 sb->s_root = d_make_root(root);
4025 ext4_msg(sb, KERN_ERR, "get root dentry failed");
4030 if (ext4_setup_super(sb, es, sb->s_flags & MS_RDONLY))
4031 sb->s_flags |= MS_RDONLY;
4033 /* determine the minimum size of new large inodes, if present */
4034 if (sbi->s_inode_size > EXT4_GOOD_OLD_INODE_SIZE) {
4035 sbi->s_want_extra_isize = sizeof(struct ext4_inode) -
4036 EXT4_GOOD_OLD_INODE_SIZE;
4037 if (EXT4_HAS_RO_COMPAT_FEATURE(sb,
4038 EXT4_FEATURE_RO_COMPAT_EXTRA_ISIZE)) {
4039 if (sbi->s_want_extra_isize <
4040 le16_to_cpu(es->s_want_extra_isize))
4041 sbi->s_want_extra_isize =
4042 le16_to_cpu(es->s_want_extra_isize);
4043 if (sbi->s_want_extra_isize <
4044 le16_to_cpu(es->s_min_extra_isize))
4045 sbi->s_want_extra_isize =
4046 le16_to_cpu(es->s_min_extra_isize);
4049 /* Check if enough inode space is available */
4050 if (EXT4_GOOD_OLD_INODE_SIZE + sbi->s_want_extra_isize >
4051 sbi->s_inode_size) {
4052 sbi->s_want_extra_isize = sizeof(struct ext4_inode) -
4053 EXT4_GOOD_OLD_INODE_SIZE;
4054 ext4_msg(sb, KERN_INFO, "required extra inode space not"
4058 err = ext4_reserve_clusters(sbi, ext4_calculate_resv_clusters(sbi));
4060 ext4_msg(sb, KERN_ERR, "failed to reserve %llu clusters for "
4061 "reserved pool", ext4_calculate_resv_clusters(sbi));
4062 goto failed_mount4a;
4065 err = ext4_setup_system_zone(sb);
4067 ext4_msg(sb, KERN_ERR, "failed to initialize system "
4069 goto failed_mount4a;
4073 err = ext4_mb_init(sb);
4075 ext4_msg(sb, KERN_ERR, "failed to initialize mballoc (%d)",
4080 err = ext4_register_li_request(sb, first_not_zeroed);
4084 sbi->s_kobj.kset = ext4_kset;
4085 init_completion(&sbi->s_kobj_unregister);
4086 err = kobject_init_and_add(&sbi->s_kobj, &ext4_ktype, NULL,
4092 /* Enable quota usage during mount. */
4093 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_QUOTA) &&
4094 !(sb->s_flags & MS_RDONLY)) {
4095 err = ext4_enable_quotas(sb);
4099 #endif /* CONFIG_QUOTA */
4101 EXT4_SB(sb)->s_mount_state |= EXT4_ORPHAN_FS;
4102 ext4_orphan_cleanup(sb, es);
4103 EXT4_SB(sb)->s_mount_state &= ~EXT4_ORPHAN_FS;
4104 if (needs_recovery) {
4105 ext4_msg(sb, KERN_INFO, "recovery complete");
4106 ext4_mark_recovery_complete(sb, es);
4108 if (EXT4_SB(sb)->s_journal) {
4109 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA)
4110 descr = " journalled data mode";
4111 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA)
4112 descr = " ordered data mode";
4114 descr = " writeback data mode";
4116 descr = "out journal";
4118 if (test_opt(sb, DISCARD)) {
4119 struct request_queue *q = bdev_get_queue(sb->s_bdev);
4120 if (!blk_queue_discard(q))
4121 ext4_msg(sb, KERN_WARNING,
4122 "mounting with \"discard\" option, but "
4123 "the device does not support discard");
4126 ext4_msg(sb, KERN_INFO, "mounted filesystem with%s. "
4127 "Opts: %s%s%s", descr, sbi->s_es->s_mount_opts,
4128 *sbi->s_es->s_mount_opts ? "; " : "", orig_data);
4130 if (es->s_error_count)
4131 mod_timer(&sbi->s_err_report, jiffies + 300*HZ); /* 5 minutes */
4138 ext4_msg(sb, KERN_ERR, "VFS: Can't find ext4 filesystem");
4143 kobject_del(&sbi->s_kobj);
4146 ext4_unregister_li_request(sb);
4148 ext4_mb_release(sb);
4150 ext4_ext_release(sb);
4151 ext4_release_system_zone(sb);
4156 ext4_msg(sb, KERN_ERR, "mount failed");
4157 if (EXT4_SB(sb)->rsv_conversion_wq)
4158 destroy_workqueue(EXT4_SB(sb)->rsv_conversion_wq);
4159 if (EXT4_SB(sb)->unrsv_conversion_wq)
4160 destroy_workqueue(EXT4_SB(sb)->unrsv_conversion_wq);
4162 if (sbi->s_journal) {
4163 jbd2_journal_destroy(sbi->s_journal);
4164 sbi->s_journal = NULL;
4167 ext4_es_unregister_shrinker(sbi);
4168 del_timer(&sbi->s_err_report);
4169 if (sbi->s_flex_groups)
4170 ext4_kvfree(sbi->s_flex_groups);
4171 percpu_counter_destroy(&sbi->s_freeclusters_counter);
4172 percpu_counter_destroy(&sbi->s_freeinodes_counter);
4173 percpu_counter_destroy(&sbi->s_dirs_counter);
4174 percpu_counter_destroy(&sbi->s_dirtyclusters_counter);
4175 percpu_counter_destroy(&sbi->s_extent_cache_cnt);
4177 kthread_stop(sbi->s_mmp_tsk);
4179 for (i = 0; i < db_count; i++)
4180 brelse(sbi->s_group_desc[i]);
4181 ext4_kvfree(sbi->s_group_desc);
4183 if (sbi->s_chksum_driver)
4184 crypto_free_shash(sbi->s_chksum_driver);
4186 remove_proc_entry("options", sbi->s_proc);
4187 remove_proc_entry(sb->s_id, ext4_proc_root);
4190 for (i = 0; i < MAXQUOTAS; i++)
4191 kfree(sbi->s_qf_names[i]);
4193 ext4_blkdev_remove(sbi);
4196 sb->s_fs_info = NULL;
4197 kfree(sbi->s_blockgroup_lock);
4201 return err ? err : ret;
4205 * Setup any per-fs journal parameters now. We'll do this both on
4206 * initial mount, once the journal has been initialised but before we've
4207 * done any recovery; and again on any subsequent remount.
4209 static void ext4_init_journal_params(struct super_block *sb, journal_t *journal)
4211 struct ext4_sb_info *sbi = EXT4_SB(sb);
4213 journal->j_commit_interval = sbi->s_commit_interval;
4214 journal->j_min_batch_time = sbi->s_min_batch_time;
4215 journal->j_max_batch_time = sbi->s_max_batch_time;
4217 write_lock(&journal->j_state_lock);
4218 if (test_opt(sb, BARRIER))
4219 journal->j_flags |= JBD2_BARRIER;
4221 journal->j_flags &= ~JBD2_BARRIER;
4222 if (test_opt(sb, DATA_ERR_ABORT))
4223 journal->j_flags |= JBD2_ABORT_ON_SYNCDATA_ERR;
4225 journal->j_flags &= ~JBD2_ABORT_ON_SYNCDATA_ERR;
4226 write_unlock(&journal->j_state_lock);
4229 static journal_t *ext4_get_journal(struct super_block *sb,
4230 unsigned int journal_inum)
4232 struct inode *journal_inode;
4235 BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL));
4237 /* First, test for the existence of a valid inode on disk. Bad
4238 * things happen if we iget() an unused inode, as the subsequent
4239 * iput() will try to delete it. */
4241 journal_inode = ext4_iget(sb, journal_inum);
4242 if (IS_ERR(journal_inode)) {
4243 ext4_msg(sb, KERN_ERR, "no journal found");
4246 if (!journal_inode->i_nlink) {
4247 make_bad_inode(journal_inode);
4248 iput(journal_inode);
4249 ext4_msg(sb, KERN_ERR, "journal inode is deleted");
4253 jbd_debug(2, "Journal inode found at %p: %lld bytes\n",
4254 journal_inode, journal_inode->i_size);
4255 if (!S_ISREG(journal_inode->i_mode)) {
4256 ext4_msg(sb, KERN_ERR, "invalid journal inode");
4257 iput(journal_inode);
4261 journal = jbd2_journal_init_inode(journal_inode);
4263 ext4_msg(sb, KERN_ERR, "Could not load journal inode");
4264 iput(journal_inode);
4267 journal->j_private = sb;
4268 ext4_init_journal_params(sb, journal);
4272 static journal_t *ext4_get_dev_journal(struct super_block *sb,
4275 struct buffer_head *bh;
4279 int hblock, blocksize;
4280 ext4_fsblk_t sb_block;
4281 unsigned long offset;
4282 struct ext4_super_block *es;
4283 struct block_device *bdev;
4285 BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL));
4287 bdev = ext4_blkdev_get(j_dev, sb);
4291 blocksize = sb->s_blocksize;
4292 hblock = bdev_logical_block_size(bdev);
4293 if (blocksize < hblock) {
4294 ext4_msg(sb, KERN_ERR,
4295 "blocksize too small for journal device");
4299 sb_block = EXT4_MIN_BLOCK_SIZE / blocksize;
4300 offset = EXT4_MIN_BLOCK_SIZE % blocksize;
4301 set_blocksize(bdev, blocksize);
4302 if (!(bh = __bread(bdev, sb_block, blocksize))) {
4303 ext4_msg(sb, KERN_ERR, "couldn't read superblock of "
4304 "external journal");
4308 es = (struct ext4_super_block *) (bh->b_data + offset);
4309 if ((le16_to_cpu(es->s_magic) != EXT4_SUPER_MAGIC) ||
4310 !(le32_to_cpu(es->s_feature_incompat) &
4311 EXT4_FEATURE_INCOMPAT_JOURNAL_DEV)) {
4312 ext4_msg(sb, KERN_ERR, "external journal has "
4318 if (memcmp(EXT4_SB(sb)->s_es->s_journal_uuid, es->s_uuid, 16)) {
4319 ext4_msg(sb, KERN_ERR, "journal UUID does not match");
4324 len = ext4_blocks_count(es);
4325 start = sb_block + 1;
4326 brelse(bh); /* we're done with the superblock */
4328 journal = jbd2_journal_init_dev(bdev, sb->s_bdev,
4329 start, len, blocksize);
4331 ext4_msg(sb, KERN_ERR, "failed to create device journal");
4334 journal->j_private = sb;
4335 ll_rw_block(READ | REQ_META | REQ_PRIO, 1, &journal->j_sb_buffer);
4336 wait_on_buffer(journal->j_sb_buffer);
4337 if (!buffer_uptodate(journal->j_sb_buffer)) {
4338 ext4_msg(sb, KERN_ERR, "I/O error on journal device");
4341 if (be32_to_cpu(journal->j_superblock->s_nr_users) != 1) {
4342 ext4_msg(sb, KERN_ERR, "External journal has more than one "
4343 "user (unsupported) - %d",
4344 be32_to_cpu(journal->j_superblock->s_nr_users));
4347 EXT4_SB(sb)->journal_bdev = bdev;
4348 ext4_init_journal_params(sb, journal);
4352 jbd2_journal_destroy(journal);
4354 ext4_blkdev_put(bdev);
4358 static int ext4_load_journal(struct super_block *sb,
4359 struct ext4_super_block *es,
4360 unsigned long journal_devnum)
4363 unsigned int journal_inum = le32_to_cpu(es->s_journal_inum);
4366 int really_read_only;
4368 BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL));
4370 if (journal_devnum &&
4371 journal_devnum != le32_to_cpu(es->s_journal_dev)) {
4372 ext4_msg(sb, KERN_INFO, "external journal device major/minor "
4373 "numbers have changed");
4374 journal_dev = new_decode_dev(journal_devnum);
4376 journal_dev = new_decode_dev(le32_to_cpu(es->s_journal_dev));
4378 really_read_only = bdev_read_only(sb->s_bdev);
4381 * Are we loading a blank journal or performing recovery after a
4382 * crash? For recovery, we need to check in advance whether we
4383 * can get read-write access to the device.
4385 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER)) {
4386 if (sb->s_flags & MS_RDONLY) {
4387 ext4_msg(sb, KERN_INFO, "INFO: recovery "
4388 "required on readonly filesystem");
4389 if (really_read_only) {
4390 ext4_msg(sb, KERN_ERR, "write access "
4391 "unavailable, cannot proceed");
4394 ext4_msg(sb, KERN_INFO, "write access will "
4395 "be enabled during recovery");
4399 if (journal_inum && journal_dev) {
4400 ext4_msg(sb, KERN_ERR, "filesystem has both journal "
4401 "and inode journals!");
4406 if (!(journal = ext4_get_journal(sb, journal_inum)))
4409 if (!(journal = ext4_get_dev_journal(sb, journal_dev)))
4413 if (!(journal->j_flags & JBD2_BARRIER))
4414 ext4_msg(sb, KERN_INFO, "barriers disabled");
4416 if (!EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER))
4417 err = jbd2_journal_wipe(journal, !really_read_only);
4419 char *save = kmalloc(EXT4_S_ERR_LEN, GFP_KERNEL);
4421 memcpy(save, ((char *) es) +
4422 EXT4_S_ERR_START, EXT4_S_ERR_LEN);
4423 err = jbd2_journal_load(journal);
4425 memcpy(((char *) es) + EXT4_S_ERR_START,
4426 save, EXT4_S_ERR_LEN);
4431 ext4_msg(sb, KERN_ERR, "error loading journal");
4432 jbd2_journal_destroy(journal);
4436 EXT4_SB(sb)->s_journal = journal;
4437 ext4_clear_journal_err(sb, es);
4439 if (!really_read_only && journal_devnum &&
4440 journal_devnum != le32_to_cpu(es->s_journal_dev)) {
4441 es->s_journal_dev = cpu_to_le32(journal_devnum);
4443 /* Make sure we flush the recovery flag to disk. */
4444 ext4_commit_super(sb, 1);
4450 static int ext4_commit_super(struct super_block *sb, int sync)
4452 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
4453 struct buffer_head *sbh = EXT4_SB(sb)->s_sbh;
4456 if (!sbh || block_device_ejected(sb))
4458 if (buffer_write_io_error(sbh)) {
4460 * Oh, dear. A previous attempt to write the
4461 * superblock failed. This could happen because the
4462 * USB device was yanked out. Or it could happen to
4463 * be a transient write error and maybe the block will
4464 * be remapped. Nothing we can do but to retry the
4465 * write and hope for the best.
4467 ext4_msg(sb, KERN_ERR, "previous I/O error to "
4468 "superblock detected");
4469 clear_buffer_write_io_error(sbh);
4470 set_buffer_uptodate(sbh);
4473 * If the file system is mounted read-only, don't update the
4474 * superblock write time. This avoids updating the superblock
4475 * write time when we are mounting the root file system
4476 * read/only but we need to replay the journal; at that point,
4477 * for people who are east of GMT and who make their clock
4478 * tick in localtime for Windows bug-for-bug compatibility,
4479 * the clock is set in the future, and this will cause e2fsck
4480 * to complain and force a full file system check.
4482 if (!(sb->s_flags & MS_RDONLY))
4483 es->s_wtime = cpu_to_le32(get_seconds());
4484 if (sb->s_bdev->bd_part)
4485 es->s_kbytes_written =
4486 cpu_to_le64(EXT4_SB(sb)->s_kbytes_written +
4487 ((part_stat_read(sb->s_bdev->bd_part, sectors[1]) -
4488 EXT4_SB(sb)->s_sectors_written_start) >> 1));
4490 es->s_kbytes_written =
4491 cpu_to_le64(EXT4_SB(sb)->s_kbytes_written);
4492 ext4_free_blocks_count_set(es,
4493 EXT4_C2B(EXT4_SB(sb), percpu_counter_sum_positive(
4494 &EXT4_SB(sb)->s_freeclusters_counter)));
4495 es->s_free_inodes_count =
4496 cpu_to_le32(percpu_counter_sum_positive(
4497 &EXT4_SB(sb)->s_freeinodes_counter));
4498 BUFFER_TRACE(sbh, "marking dirty");
4499 ext4_superblock_csum_set(sb);
4500 mark_buffer_dirty(sbh);
4502 error = sync_dirty_buffer(sbh);
4506 error = buffer_write_io_error(sbh);
4508 ext4_msg(sb, KERN_ERR, "I/O error while writing "
4510 clear_buffer_write_io_error(sbh);
4511 set_buffer_uptodate(sbh);
4518 * Have we just finished recovery? If so, and if we are mounting (or
4519 * remounting) the filesystem readonly, then we will end up with a
4520 * consistent fs on disk. Record that fact.
4522 static void ext4_mark_recovery_complete(struct super_block *sb,
4523 struct ext4_super_block *es)
4525 journal_t *journal = EXT4_SB(sb)->s_journal;
4527 if (!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL)) {
4528 BUG_ON(journal != NULL);
4531 jbd2_journal_lock_updates(journal);
4532 if (jbd2_journal_flush(journal) < 0)
4535 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER) &&
4536 sb->s_flags & MS_RDONLY) {
4537 EXT4_CLEAR_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
4538 ext4_commit_super(sb, 1);
4542 jbd2_journal_unlock_updates(journal);
4546 * If we are mounting (or read-write remounting) a filesystem whose journal
4547 * has recorded an error from a previous lifetime, move that error to the
4548 * main filesystem now.
4550 static void ext4_clear_journal_err(struct super_block *sb,
4551 struct ext4_super_block *es)
4557 BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL));
4559 journal = EXT4_SB(sb)->s_journal;
4562 * Now check for any error status which may have been recorded in the
4563 * journal by a prior ext4_error() or ext4_abort()
4566 j_errno = jbd2_journal_errno(journal);
4570 errstr = ext4_decode_error(sb, j_errno, nbuf);
4571 ext4_warning(sb, "Filesystem error recorded "
4572 "from previous mount: %s", errstr);
4573 ext4_warning(sb, "Marking fs in need of filesystem check.");
4575 EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS;
4576 es->s_state |= cpu_to_le16(EXT4_ERROR_FS);
4577 ext4_commit_super(sb, 1);
4579 jbd2_journal_clear_err(journal);
4580 jbd2_journal_update_sb_errno(journal);
4585 * Force the running and committing transactions to commit,
4586 * and wait on the commit.
4588 int ext4_force_commit(struct super_block *sb)
4592 if (sb->s_flags & MS_RDONLY)
4595 journal = EXT4_SB(sb)->s_journal;
4596 return ext4_journal_force_commit(journal);
4599 static int ext4_sync_fs(struct super_block *sb, int wait)
4603 bool needs_barrier = false;
4604 struct ext4_sb_info *sbi = EXT4_SB(sb);
4606 trace_ext4_sync_fs(sb, wait);
4607 flush_workqueue(sbi->rsv_conversion_wq);
4608 flush_workqueue(sbi->unrsv_conversion_wq);
4610 * Writeback quota in non-journalled quota case - journalled quota has
4613 dquot_writeback_dquots(sb, -1);
4615 * Data writeback is possible w/o journal transaction, so barrier must
4616 * being sent at the end of the function. But we can skip it if
4617 * transaction_commit will do it for us.
4619 target = jbd2_get_latest_transaction(sbi->s_journal);
4620 if (wait && sbi->s_journal->j_flags & JBD2_BARRIER &&
4621 !jbd2_trans_will_send_data_barrier(sbi->s_journal, target))
4622 needs_barrier = true;
4624 if (jbd2_journal_start_commit(sbi->s_journal, &target)) {
4626 ret = jbd2_log_wait_commit(sbi->s_journal, target);
4628 if (needs_barrier) {
4630 err = blkdev_issue_flush(sb->s_bdev, GFP_KERNEL, NULL);
4638 static int ext4_sync_fs_nojournal(struct super_block *sb, int wait)
4642 trace_ext4_sync_fs(sb, wait);
4643 flush_workqueue(EXT4_SB(sb)->rsv_conversion_wq);
4644 flush_workqueue(EXT4_SB(sb)->unrsv_conversion_wq);
4645 dquot_writeback_dquots(sb, -1);
4646 if (wait && test_opt(sb, BARRIER))
4647 ret = blkdev_issue_flush(sb->s_bdev, GFP_KERNEL, NULL);
4653 * LVM calls this function before a (read-only) snapshot is created. This
4654 * gives us a chance to flush the journal completely and mark the fs clean.
4656 * Note that only this function cannot bring a filesystem to be in a clean
4657 * state independently. It relies on upper layer to stop all data & metadata
4660 static int ext4_freeze(struct super_block *sb)
4665 if (sb->s_flags & MS_RDONLY)
4668 journal = EXT4_SB(sb)->s_journal;
4670 /* Now we set up the journal barrier. */
4671 jbd2_journal_lock_updates(journal);
4674 * Don't clear the needs_recovery flag if we failed to flush
4677 error = jbd2_journal_flush(journal);
4681 /* Journal blocked and flushed, clear needs_recovery flag. */
4682 EXT4_CLEAR_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
4683 error = ext4_commit_super(sb, 1);
4685 /* we rely on upper layer to stop further updates */
4686 jbd2_journal_unlock_updates(EXT4_SB(sb)->s_journal);
4691 * Called by LVM after the snapshot is done. We need to reset the RECOVER
4692 * flag here, even though the filesystem is not technically dirty yet.
4694 static int ext4_unfreeze(struct super_block *sb)
4696 if (sb->s_flags & MS_RDONLY)
4699 /* Reset the needs_recovery flag before the fs is unlocked. */
4700 EXT4_SET_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
4701 ext4_commit_super(sb, 1);
4706 * Structure to save mount options for ext4_remount's benefit
4708 struct ext4_mount_options {
4709 unsigned long s_mount_opt;
4710 unsigned long s_mount_opt2;
4713 unsigned long s_commit_interval;
4714 u32 s_min_batch_time, s_max_batch_time;
4717 char *s_qf_names[MAXQUOTAS];
4721 static int ext4_remount(struct super_block *sb, int *flags, char *data)
4723 struct ext4_super_block *es;
4724 struct ext4_sb_info *sbi = EXT4_SB(sb);
4725 unsigned long old_sb_flags;
4726 struct ext4_mount_options old_opts;
4727 int enable_quota = 0;
4729 unsigned int journal_ioprio = DEFAULT_JOURNAL_IOPRIO;
4734 char *orig_data = kstrdup(data, GFP_KERNEL);
4736 /* Store the original options */
4737 old_sb_flags = sb->s_flags;
4738 old_opts.s_mount_opt = sbi->s_mount_opt;
4739 old_opts.s_mount_opt2 = sbi->s_mount_opt2;
4740 old_opts.s_resuid = sbi->s_resuid;
4741 old_opts.s_resgid = sbi->s_resgid;
4742 old_opts.s_commit_interval = sbi->s_commit_interval;
4743 old_opts.s_min_batch_time = sbi->s_min_batch_time;
4744 old_opts.s_max_batch_time = sbi->s_max_batch_time;
4746 old_opts.s_jquota_fmt = sbi->s_jquota_fmt;
4747 for (i = 0; i < MAXQUOTAS; i++)
4748 if (sbi->s_qf_names[i]) {
4749 old_opts.s_qf_names[i] = kstrdup(sbi->s_qf_names[i],
4751 if (!old_opts.s_qf_names[i]) {
4752 for (j = 0; j < i; j++)
4753 kfree(old_opts.s_qf_names[j]);
4758 old_opts.s_qf_names[i] = NULL;
4760 if (sbi->s_journal && sbi->s_journal->j_task->io_context)
4761 journal_ioprio = sbi->s_journal->j_task->io_context->ioprio;
4764 * Allow the "check" option to be passed as a remount option.
4766 if (!parse_options(data, sb, NULL, &journal_ioprio, 1)) {
4771 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA) {
4772 if (test_opt2(sb, EXPLICIT_DELALLOC)) {
4773 ext4_msg(sb, KERN_ERR, "can't mount with "
4774 "both data=journal and delalloc");
4778 if (test_opt(sb, DIOREAD_NOLOCK)) {
4779 ext4_msg(sb, KERN_ERR, "can't mount with "
4780 "both data=journal and dioread_nolock");
4786 if (sbi->s_mount_flags & EXT4_MF_FS_ABORTED)
4787 ext4_abort(sb, "Abort forced by user");
4789 sb->s_flags = (sb->s_flags & ~MS_POSIXACL) |
4790 (test_opt(sb, POSIX_ACL) ? MS_POSIXACL : 0);
4794 if (sbi->s_journal) {
4795 ext4_init_journal_params(sb, sbi->s_journal);
4796 set_task_ioprio(sbi->s_journal->j_task, journal_ioprio);
4799 if ((*flags & MS_RDONLY) != (sb->s_flags & MS_RDONLY)) {
4800 if (sbi->s_mount_flags & EXT4_MF_FS_ABORTED) {
4805 if (*flags & MS_RDONLY) {
4806 err = dquot_suspend(sb, -1);
4811 * First of all, the unconditional stuff we have to do
4812 * to disable replay of the journal when we next remount
4814 sb->s_flags |= MS_RDONLY;
4817 * OK, test if we are remounting a valid rw partition
4818 * readonly, and if so set the rdonly flag and then
4819 * mark the partition as valid again.
4821 if (!(es->s_state & cpu_to_le16(EXT4_VALID_FS)) &&
4822 (sbi->s_mount_state & EXT4_VALID_FS))
4823 es->s_state = cpu_to_le16(sbi->s_mount_state);
4826 ext4_mark_recovery_complete(sb, es);
4828 /* Make sure we can mount this feature set readwrite */
4829 if (!ext4_feature_set_ok(sb, 0)) {
4834 * Make sure the group descriptor checksums
4835 * are sane. If they aren't, refuse to remount r/w.
4837 for (g = 0; g < sbi->s_groups_count; g++) {
4838 struct ext4_group_desc *gdp =
4839 ext4_get_group_desc(sb, g, NULL);
4841 if (!ext4_group_desc_csum_verify(sb, g, gdp)) {
4842 ext4_msg(sb, KERN_ERR,
4843 "ext4_remount: Checksum for group %u failed (%u!=%u)",
4844 g, le16_to_cpu(ext4_group_desc_csum(sbi, g, gdp)),
4845 le16_to_cpu(gdp->bg_checksum));
4852 * If we have an unprocessed orphan list hanging
4853 * around from a previously readonly bdev mount,
4854 * require a full umount/remount for now.
4856 if (es->s_last_orphan) {
4857 ext4_msg(sb, KERN_WARNING, "Couldn't "
4858 "remount RDWR because of unprocessed "
4859 "orphan inode list. Please "
4860 "umount/remount instead");
4866 * Mounting a RDONLY partition read-write, so reread
4867 * and store the current valid flag. (It may have
4868 * been changed by e2fsck since we originally mounted
4872 ext4_clear_journal_err(sb, es);
4873 sbi->s_mount_state = le16_to_cpu(es->s_state);
4874 if (!ext4_setup_super(sb, es, 0))
4875 sb->s_flags &= ~MS_RDONLY;
4876 if (EXT4_HAS_INCOMPAT_FEATURE(sb,
4877 EXT4_FEATURE_INCOMPAT_MMP))
4878 if (ext4_multi_mount_protect(sb,
4879 le64_to_cpu(es->s_mmp_block))) {
4888 * Reinitialize lazy itable initialization thread based on
4891 if ((sb->s_flags & MS_RDONLY) || !test_opt(sb, INIT_INODE_TABLE))
4892 ext4_unregister_li_request(sb);
4894 ext4_group_t first_not_zeroed;
4895 first_not_zeroed = ext4_has_uninit_itable(sb);
4896 ext4_register_li_request(sb, first_not_zeroed);
4899 ext4_setup_system_zone(sb);
4900 if (sbi->s_journal == NULL && !(old_sb_flags & MS_RDONLY))
4901 ext4_commit_super(sb, 1);
4904 /* Release old quota file names */
4905 for (i = 0; i < MAXQUOTAS; i++)
4906 kfree(old_opts.s_qf_names[i]);
4908 if (sb_any_quota_suspended(sb))
4909 dquot_resume(sb, -1);
4910 else if (EXT4_HAS_RO_COMPAT_FEATURE(sb,
4911 EXT4_FEATURE_RO_COMPAT_QUOTA)) {
4912 err = ext4_enable_quotas(sb);
4919 ext4_msg(sb, KERN_INFO, "re-mounted. Opts: %s", orig_data);
4924 sb->s_flags = old_sb_flags;
4925 sbi->s_mount_opt = old_opts.s_mount_opt;
4926 sbi->s_mount_opt2 = old_opts.s_mount_opt2;
4927 sbi->s_resuid = old_opts.s_resuid;
4928 sbi->s_resgid = old_opts.s_resgid;
4929 sbi->s_commit_interval = old_opts.s_commit_interval;
4930 sbi->s_min_batch_time = old_opts.s_min_batch_time;
4931 sbi->s_max_batch_time = old_opts.s_max_batch_time;
4933 sbi->s_jquota_fmt = old_opts.s_jquota_fmt;
4934 for (i = 0; i < MAXQUOTAS; i++) {
4935 kfree(sbi->s_qf_names[i]);
4936 sbi->s_qf_names[i] = old_opts.s_qf_names[i];
4943 static int ext4_statfs(struct dentry *dentry, struct kstatfs *buf)
4945 struct super_block *sb = dentry->d_sb;
4946 struct ext4_sb_info *sbi = EXT4_SB(sb);
4947 struct ext4_super_block *es = sbi->s_es;
4948 ext4_fsblk_t overhead = 0, resv_blocks;
4951 resv_blocks = EXT4_C2B(sbi, atomic64_read(&sbi->s_resv_clusters));
4953 if (!test_opt(sb, MINIX_DF))
4954 overhead = sbi->s_overhead;
4956 buf->f_type = EXT4_SUPER_MAGIC;
4957 buf->f_bsize = sb->s_blocksize;
4958 buf->f_blocks = ext4_blocks_count(es) - EXT4_C2B(sbi, overhead);
4959 bfree = percpu_counter_sum_positive(&sbi->s_freeclusters_counter) -
4960 percpu_counter_sum_positive(&sbi->s_dirtyclusters_counter);
4961 /* prevent underflow in case that few free space is available */
4962 buf->f_bfree = EXT4_C2B(sbi, max_t(s64, bfree, 0));
4963 buf->f_bavail = buf->f_bfree -
4964 (ext4_r_blocks_count(es) + resv_blocks);
4965 if (buf->f_bfree < (ext4_r_blocks_count(es) + resv_blocks))
4967 buf->f_files = le32_to_cpu(es->s_inodes_count);
4968 buf->f_ffree = percpu_counter_sum_positive(&sbi->s_freeinodes_counter);
4969 buf->f_namelen = EXT4_NAME_LEN;
4970 fsid = le64_to_cpup((void *)es->s_uuid) ^
4971 le64_to_cpup((void *)es->s_uuid + sizeof(u64));
4972 buf->f_fsid.val[0] = fsid & 0xFFFFFFFFUL;
4973 buf->f_fsid.val[1] = (fsid >> 32) & 0xFFFFFFFFUL;
4978 /* Helper function for writing quotas on sync - we need to start transaction
4979 * before quota file is locked for write. Otherwise the are possible deadlocks:
4980 * Process 1 Process 2
4981 * ext4_create() quota_sync()
4982 * jbd2_journal_start() write_dquot()
4983 * dquot_initialize() down(dqio_mutex)
4984 * down(dqio_mutex) jbd2_journal_start()
4990 static inline struct inode *dquot_to_inode(struct dquot *dquot)
4992 return sb_dqopt(dquot->dq_sb)->files[dquot->dq_id.type];
4995 static int ext4_write_dquot(struct dquot *dquot)
4999 struct inode *inode;
5001 inode = dquot_to_inode(dquot);
5002 handle = ext4_journal_start(inode, EXT4_HT_QUOTA,
5003 EXT4_QUOTA_TRANS_BLOCKS(dquot->dq_sb));
5005 return PTR_ERR(handle);
5006 ret = dquot_commit(dquot);
5007 err = ext4_journal_stop(handle);
5013 static int ext4_acquire_dquot(struct dquot *dquot)
5018 handle = ext4_journal_start(dquot_to_inode(dquot), EXT4_HT_QUOTA,
5019 EXT4_QUOTA_INIT_BLOCKS(dquot->dq_sb));
5021 return PTR_ERR(handle);
5022 ret = dquot_acquire(dquot);
5023 err = ext4_journal_stop(handle);
5029 static int ext4_release_dquot(struct dquot *dquot)
5034 handle = ext4_journal_start(dquot_to_inode(dquot), EXT4_HT_QUOTA,
5035 EXT4_QUOTA_DEL_BLOCKS(dquot->dq_sb));
5036 if (IS_ERR(handle)) {
5037 /* Release dquot anyway to avoid endless cycle in dqput() */
5038 dquot_release(dquot);
5039 return PTR_ERR(handle);
5041 ret = dquot_release(dquot);
5042 err = ext4_journal_stop(handle);
5048 static int ext4_mark_dquot_dirty(struct dquot *dquot)
5050 struct super_block *sb = dquot->dq_sb;
5051 struct ext4_sb_info *sbi = EXT4_SB(sb);
5053 /* Are we journaling quotas? */
5054 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_QUOTA) ||
5055 sbi->s_qf_names[USRQUOTA] || sbi->s_qf_names[GRPQUOTA]) {
5056 dquot_mark_dquot_dirty(dquot);
5057 return ext4_write_dquot(dquot);
5059 return dquot_mark_dquot_dirty(dquot);
5063 static int ext4_write_info(struct super_block *sb, int type)
5068 /* Data block + inode block */
5069 handle = ext4_journal_start(sb->s_root->d_inode, EXT4_HT_QUOTA, 2);
5071 return PTR_ERR(handle);
5072 ret = dquot_commit_info(sb, type);
5073 err = ext4_journal_stop(handle);
5080 * Turn on quotas during mount time - we need to find
5081 * the quota file and such...
5083 static int ext4_quota_on_mount(struct super_block *sb, int type)
5085 return dquot_quota_on_mount(sb, EXT4_SB(sb)->s_qf_names[type],
5086 EXT4_SB(sb)->s_jquota_fmt, type);
5090 * Standard function to be called on quota_on
5092 static int ext4_quota_on(struct super_block *sb, int type, int format_id,
5097 if (!test_opt(sb, QUOTA))
5100 /* Quotafile not on the same filesystem? */
5101 if (path->dentry->d_sb != sb)
5103 /* Journaling quota? */
5104 if (EXT4_SB(sb)->s_qf_names[type]) {
5105 /* Quotafile not in fs root? */
5106 if (path->dentry->d_parent != sb->s_root)
5107 ext4_msg(sb, KERN_WARNING,
5108 "Quota file not on filesystem root. "
5109 "Journaled quota will not work");
5113 * When we journal data on quota file, we have to flush journal to see
5114 * all updates to the file when we bypass pagecache...
5116 if (EXT4_SB(sb)->s_journal &&
5117 ext4_should_journal_data(path->dentry->d_inode)) {
5119 * We don't need to lock updates but journal_flush() could
5120 * otherwise be livelocked...
5122 jbd2_journal_lock_updates(EXT4_SB(sb)->s_journal);
5123 err = jbd2_journal_flush(EXT4_SB(sb)->s_journal);
5124 jbd2_journal_unlock_updates(EXT4_SB(sb)->s_journal);
5129 return dquot_quota_on(sb, type, format_id, path);
5132 static int ext4_quota_enable(struct super_block *sb, int type, int format_id,
5136 struct inode *qf_inode;
5137 unsigned long qf_inums[MAXQUOTAS] = {
5138 le32_to_cpu(EXT4_SB(sb)->s_es->s_usr_quota_inum),
5139 le32_to_cpu(EXT4_SB(sb)->s_es->s_grp_quota_inum)
5142 BUG_ON(!EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_QUOTA));
5144 if (!qf_inums[type])
5147 qf_inode = ext4_iget(sb, qf_inums[type]);
5148 if (IS_ERR(qf_inode)) {
5149 ext4_error(sb, "Bad quota inode # %lu", qf_inums[type]);
5150 return PTR_ERR(qf_inode);
5153 /* Don't account quota for quota files to avoid recursion */
5154 qf_inode->i_flags |= S_NOQUOTA;
5155 err = dquot_enable(qf_inode, type, format_id, flags);
5161 /* Enable usage tracking for all quota types. */
5162 static int ext4_enable_quotas(struct super_block *sb)
5165 unsigned long qf_inums[MAXQUOTAS] = {
5166 le32_to_cpu(EXT4_SB(sb)->s_es->s_usr_quota_inum),
5167 le32_to_cpu(EXT4_SB(sb)->s_es->s_grp_quota_inum)
5170 sb_dqopt(sb)->flags |= DQUOT_QUOTA_SYS_FILE;
5171 for (type = 0; type < MAXQUOTAS; type++) {
5172 if (qf_inums[type]) {
5173 err = ext4_quota_enable(sb, type, QFMT_VFS_V1,
5174 DQUOT_USAGE_ENABLED);
5177 "Failed to enable quota tracking "
5178 "(type=%d, err=%d). Please run "
5179 "e2fsck to fix.", type, err);
5188 * quota_on function that is used when QUOTA feature is set.
5190 static int ext4_quota_on_sysfile(struct super_block *sb, int type,
5193 if (!EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_QUOTA))
5197 * USAGE was enabled at mount time. Only need to enable LIMITS now.
5199 return ext4_quota_enable(sb, type, format_id, DQUOT_LIMITS_ENABLED);
5202 static int ext4_quota_off(struct super_block *sb, int type)
5204 struct inode *inode = sb_dqopt(sb)->files[type];
5207 /* Force all delayed allocation blocks to be allocated.
5208 * Caller already holds s_umount sem */
5209 if (test_opt(sb, DELALLOC))
5210 sync_filesystem(sb);
5215 /* Update modification times of quota files when userspace can
5216 * start looking at them */
5217 handle = ext4_journal_start(inode, EXT4_HT_QUOTA, 1);
5220 inode->i_mtime = inode->i_ctime = CURRENT_TIME;
5221 ext4_mark_inode_dirty(handle, inode);
5222 ext4_journal_stop(handle);
5225 return dquot_quota_off(sb, type);
5229 * quota_off function that is used when QUOTA feature is set.
5231 static int ext4_quota_off_sysfile(struct super_block *sb, int type)
5233 if (!EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_QUOTA))
5236 /* Disable only the limits. */
5237 return dquot_disable(sb, type, DQUOT_LIMITS_ENABLED);
5240 /* Read data from quotafile - avoid pagecache and such because we cannot afford
5241 * acquiring the locks... As quota files are never truncated and quota code
5242 * itself serializes the operations (and no one else should touch the files)
5243 * we don't have to be afraid of races */
5244 static ssize_t ext4_quota_read(struct super_block *sb, int type, char *data,
5245 size_t len, loff_t off)
5247 struct inode *inode = sb_dqopt(sb)->files[type];
5248 ext4_lblk_t blk = off >> EXT4_BLOCK_SIZE_BITS(sb);
5250 int offset = off & (sb->s_blocksize - 1);
5253 struct buffer_head *bh;
5254 loff_t i_size = i_size_read(inode);
5258 if (off+len > i_size)
5261 while (toread > 0) {
5262 tocopy = sb->s_blocksize - offset < toread ?
5263 sb->s_blocksize - offset : toread;
5264 bh = ext4_bread(NULL, inode, blk, 0, &err);
5267 if (!bh) /* A hole? */
5268 memset(data, 0, tocopy);
5270 memcpy(data, bh->b_data+offset, tocopy);
5280 /* Write to quotafile (we know the transaction is already started and has
5281 * enough credits) */
5282 static ssize_t ext4_quota_write(struct super_block *sb, int type,
5283 const char *data, size_t len, loff_t off)
5285 struct inode *inode = sb_dqopt(sb)->files[type];
5286 ext4_lblk_t blk = off >> EXT4_BLOCK_SIZE_BITS(sb);
5288 int offset = off & (sb->s_blocksize - 1);
5289 struct buffer_head *bh;
5290 handle_t *handle = journal_current_handle();
5292 if (EXT4_SB(sb)->s_journal && !handle) {
5293 ext4_msg(sb, KERN_WARNING, "Quota write (off=%llu, len=%llu)"
5294 " cancelled because transaction is not started",
5295 (unsigned long long)off, (unsigned long long)len);
5299 * Since we account only one data block in transaction credits,
5300 * then it is impossible to cross a block boundary.
5302 if (sb->s_blocksize - offset < len) {
5303 ext4_msg(sb, KERN_WARNING, "Quota write (off=%llu, len=%llu)"
5304 " cancelled because not block aligned",
5305 (unsigned long long)off, (unsigned long long)len);
5309 bh = ext4_bread(handle, inode, blk, 1, &err);
5312 err = ext4_journal_get_write_access(handle, bh);
5318 memcpy(bh->b_data+offset, data, len);
5319 flush_dcache_page(bh->b_page);
5321 err = ext4_handle_dirty_metadata(handle, NULL, bh);
5326 if (inode->i_size < off + len) {
5327 i_size_write(inode, off + len);
5328 EXT4_I(inode)->i_disksize = inode->i_size;
5329 ext4_mark_inode_dirty(handle, inode);
5336 static struct dentry *ext4_mount(struct file_system_type *fs_type, int flags,
5337 const char *dev_name, void *data)
5339 return mount_bdev(fs_type, flags, dev_name, data, ext4_fill_super);
5342 #if !defined(CONFIG_EXT2_FS) && !defined(CONFIG_EXT2_FS_MODULE) && defined(CONFIG_EXT4_USE_FOR_EXT23)
5343 static inline void register_as_ext2(void)
5345 int err = register_filesystem(&ext2_fs_type);
5348 "EXT4-fs: Unable to register as ext2 (%d)\n", err);
5351 static inline void unregister_as_ext2(void)
5353 unregister_filesystem(&ext2_fs_type);
5356 static inline int ext2_feature_set_ok(struct super_block *sb)
5358 if (EXT4_HAS_INCOMPAT_FEATURE(sb, ~EXT2_FEATURE_INCOMPAT_SUPP))
5360 if (sb->s_flags & MS_RDONLY)
5362 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, ~EXT2_FEATURE_RO_COMPAT_SUPP))
5367 static inline void register_as_ext2(void) { }
5368 static inline void unregister_as_ext2(void) { }
5369 static inline int ext2_feature_set_ok(struct super_block *sb) { return 0; }
5372 #if !defined(CONFIG_EXT3_FS) && !defined(CONFIG_EXT3_FS_MODULE) && defined(CONFIG_EXT4_USE_FOR_EXT23)
5373 static inline void register_as_ext3(void)
5375 int err = register_filesystem(&ext3_fs_type);
5378 "EXT4-fs: Unable to register as ext3 (%d)\n", err);
5381 static inline void unregister_as_ext3(void)
5383 unregister_filesystem(&ext3_fs_type);
5386 static inline int ext3_feature_set_ok(struct super_block *sb)
5388 if (EXT4_HAS_INCOMPAT_FEATURE(sb, ~EXT3_FEATURE_INCOMPAT_SUPP))
5390 if (!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL))
5392 if (sb->s_flags & MS_RDONLY)
5394 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, ~EXT3_FEATURE_RO_COMPAT_SUPP))
5399 static inline void register_as_ext3(void) { }
5400 static inline void unregister_as_ext3(void) { }
5401 static inline int ext3_feature_set_ok(struct super_block *sb) { return 0; }
5404 static struct file_system_type ext4_fs_type = {
5405 .owner = THIS_MODULE,
5407 .mount = ext4_mount,
5408 .kill_sb = kill_block_super,
5409 .fs_flags = FS_REQUIRES_DEV,
5411 MODULE_ALIAS_FS("ext4");
5413 static int __init ext4_init_feat_adverts(void)
5415 struct ext4_features *ef;
5418 ef = kzalloc(sizeof(struct ext4_features), GFP_KERNEL);
5422 ef->f_kobj.kset = ext4_kset;
5423 init_completion(&ef->f_kobj_unregister);
5424 ret = kobject_init_and_add(&ef->f_kobj, &ext4_feat_ktype, NULL,
5437 static void ext4_exit_feat_adverts(void)
5439 kobject_put(&ext4_feat->f_kobj);
5440 wait_for_completion(&ext4_feat->f_kobj_unregister);
5444 /* Shared across all ext4 file systems */
5445 wait_queue_head_t ext4__ioend_wq[EXT4_WQ_HASH_SZ];
5446 struct mutex ext4__aio_mutex[EXT4_WQ_HASH_SZ];
5448 static int __init ext4_init_fs(void)
5452 ext4_li_info = NULL;
5453 mutex_init(&ext4_li_mtx);
5455 /* Build-time check for flags consistency */
5456 ext4_check_flag_values();
5458 for (i = 0; i < EXT4_WQ_HASH_SZ; i++) {
5459 mutex_init(&ext4__aio_mutex[i]);
5460 init_waitqueue_head(&ext4__ioend_wq[i]);
5463 err = ext4_init_es();
5467 err = ext4_init_pageio();
5471 err = ext4_init_system_zone();
5474 ext4_kset = kset_create_and_add("ext4", NULL, fs_kobj);
5479 ext4_proc_root = proc_mkdir("fs/ext4", NULL);
5481 err = ext4_init_feat_adverts();
5485 err = ext4_init_mballoc();
5489 err = ext4_init_xattr();
5492 err = init_inodecache();
5497 err = register_filesystem(&ext4_fs_type);
5503 unregister_as_ext2();
5504 unregister_as_ext3();
5505 destroy_inodecache();
5509 ext4_exit_mballoc();
5511 ext4_exit_feat_adverts();
5514 remove_proc_entry("fs/ext4", NULL);
5515 kset_unregister(ext4_kset);
5517 ext4_exit_system_zone();
5526 static void __exit ext4_exit_fs(void)
5528 ext4_destroy_lazyinit_thread();
5529 unregister_as_ext2();
5530 unregister_as_ext3();
5531 unregister_filesystem(&ext4_fs_type);
5532 destroy_inodecache();
5534 ext4_exit_mballoc();
5535 ext4_exit_feat_adverts();
5536 remove_proc_entry("fs/ext4", NULL);
5537 kset_unregister(ext4_kset);
5538 ext4_exit_system_zone();
5543 MODULE_AUTHOR("Remy Card, Stephen Tweedie, Andrew Morton, Andreas Dilger, Theodore Ts'o and others");
5544 MODULE_DESCRIPTION("Fourth Extended Filesystem");
5545 MODULE_LICENSE("GPL");
5546 module_init(ext4_init_fs)
5547 module_exit(ext4_exit_fs)