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 | __GFP_NOWARN);
167 ret = __vmalloc(size, flags, PAGE_KERNEL);
171 void *ext4_kvzalloc(size_t size, gfp_t flags)
175 ret = kzalloc(size, flags | __GFP_NOWARN);
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 #define ext4_error_ratelimit(sb) \
415 ___ratelimit(&(EXT4_SB(sb)->s_err_ratelimit_state), \
418 void __ext4_error(struct super_block *sb, const char *function,
419 unsigned int line, const char *fmt, ...)
421 struct va_format vaf;
424 if (ext4_error_ratelimit(sb)) {
429 "EXT4-fs error (device %s): %s:%d: comm %s: %pV\n",
430 sb->s_id, function, line, current->comm, &vaf);
433 save_error_info(sb, function, line);
434 ext4_handle_error(sb);
437 void __ext4_error_inode(struct inode *inode, const char *function,
438 unsigned int line, ext4_fsblk_t block,
439 const char *fmt, ...)
442 struct va_format vaf;
443 struct ext4_super_block *es = EXT4_SB(inode->i_sb)->s_es;
445 es->s_last_error_ino = cpu_to_le32(inode->i_ino);
446 es->s_last_error_block = cpu_to_le64(block);
447 if (ext4_error_ratelimit(inode->i_sb)) {
452 printk(KERN_CRIT "EXT4-fs error (device %s): %s:%d: "
453 "inode #%lu: block %llu: comm %s: %pV\n",
454 inode->i_sb->s_id, function, line, inode->i_ino,
455 block, current->comm, &vaf);
457 printk(KERN_CRIT "EXT4-fs error (device %s): %s:%d: "
458 "inode #%lu: comm %s: %pV\n",
459 inode->i_sb->s_id, function, line, inode->i_ino,
460 current->comm, &vaf);
463 save_error_info(inode->i_sb, function, line);
464 ext4_handle_error(inode->i_sb);
467 void __ext4_error_file(struct file *file, const char *function,
468 unsigned int line, ext4_fsblk_t block,
469 const char *fmt, ...)
472 struct va_format vaf;
473 struct ext4_super_block *es;
474 struct inode *inode = file_inode(file);
475 char pathname[80], *path;
477 es = EXT4_SB(inode->i_sb)->s_es;
478 es->s_last_error_ino = cpu_to_le32(inode->i_ino);
479 if (ext4_error_ratelimit(inode->i_sb)) {
480 path = d_path(&(file->f_path), pathname, sizeof(pathname));
488 "EXT4-fs error (device %s): %s:%d: inode #%lu: "
489 "block %llu: comm %s: path %s: %pV\n",
490 inode->i_sb->s_id, function, line, inode->i_ino,
491 block, current->comm, path, &vaf);
494 "EXT4-fs error (device %s): %s:%d: inode #%lu: "
495 "comm %s: path %s: %pV\n",
496 inode->i_sb->s_id, function, line, inode->i_ino,
497 current->comm, path, &vaf);
500 save_error_info(inode->i_sb, function, line);
501 ext4_handle_error(inode->i_sb);
504 const char *ext4_decode_error(struct super_block *sb, int errno,
511 errstr = "IO failure";
514 errstr = "Out of memory";
517 if (!sb || (EXT4_SB(sb)->s_journal &&
518 EXT4_SB(sb)->s_journal->j_flags & JBD2_ABORT))
519 errstr = "Journal has aborted";
521 errstr = "Readonly filesystem";
524 /* If the caller passed in an extra buffer for unknown
525 * errors, textualise them now. Else we just return
528 /* Check for truncated error codes... */
529 if (snprintf(nbuf, 16, "error %d", -errno) >= 0)
538 /* __ext4_std_error decodes expected errors from journaling functions
539 * automatically and invokes the appropriate error response. */
541 void __ext4_std_error(struct super_block *sb, const char *function,
542 unsigned int line, int errno)
547 /* Special case: if the error is EROFS, and we're not already
548 * inside a transaction, then there's really no point in logging
550 if (errno == -EROFS && journal_current_handle() == NULL &&
551 (sb->s_flags & MS_RDONLY))
554 if (ext4_error_ratelimit(sb)) {
555 errstr = ext4_decode_error(sb, errno, nbuf);
556 printk(KERN_CRIT "EXT4-fs error (device %s) in %s:%d: %s\n",
557 sb->s_id, function, line, errstr);
560 save_error_info(sb, function, line);
561 ext4_handle_error(sb);
565 * ext4_abort is a much stronger failure handler than ext4_error. The
566 * abort function may be used to deal with unrecoverable failures such
567 * as journal IO errors or ENOMEM at a critical moment in log management.
569 * We unconditionally force the filesystem into an ABORT|READONLY state,
570 * unless the error response on the fs has been set to panic in which
571 * case we take the easy way out and panic immediately.
574 void __ext4_abort(struct super_block *sb, const char *function,
575 unsigned int line, const char *fmt, ...)
579 save_error_info(sb, function, line);
581 printk(KERN_CRIT "EXT4-fs error (device %s): %s:%d: ", sb->s_id,
587 if ((sb->s_flags & MS_RDONLY) == 0) {
588 ext4_msg(sb, KERN_CRIT, "Remounting filesystem read-only");
589 EXT4_SB(sb)->s_mount_flags |= EXT4_MF_FS_ABORTED;
591 * Make sure updated value of ->s_mount_flags will be visible
592 * before ->s_flags update
595 sb->s_flags |= MS_RDONLY;
596 if (EXT4_SB(sb)->s_journal)
597 jbd2_journal_abort(EXT4_SB(sb)->s_journal, -EIO);
598 save_error_info(sb, function, line);
600 if (test_opt(sb, ERRORS_PANIC))
601 panic("EXT4-fs panic from previous error\n");
604 void __ext4_msg(struct super_block *sb,
605 const char *prefix, const char *fmt, ...)
607 struct va_format vaf;
610 if (!___ratelimit(&(EXT4_SB(sb)->s_msg_ratelimit_state), "EXT4-fs"))
616 printk("%sEXT4-fs (%s): %pV\n", prefix, sb->s_id, &vaf);
620 void __ext4_warning(struct super_block *sb, const char *function,
621 unsigned int line, const char *fmt, ...)
623 struct va_format vaf;
626 if (!___ratelimit(&(EXT4_SB(sb)->s_warning_ratelimit_state),
633 printk(KERN_WARNING "EXT4-fs warning (device %s): %s:%d: %pV\n",
634 sb->s_id, function, line, &vaf);
638 void __ext4_grp_locked_error(const char *function, unsigned int line,
639 struct super_block *sb, ext4_group_t grp,
640 unsigned long ino, ext4_fsblk_t block,
641 const char *fmt, ...)
645 struct va_format vaf;
647 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
649 es->s_last_error_ino = cpu_to_le32(ino);
650 es->s_last_error_block = cpu_to_le64(block);
651 __save_error_info(sb, function, line);
653 if (ext4_error_ratelimit(sb)) {
657 printk(KERN_CRIT "EXT4-fs error (device %s): %s:%d: group %u, ",
658 sb->s_id, function, line, grp);
660 printk(KERN_CONT "inode %lu: ", ino);
662 printk(KERN_CONT "block %llu:",
663 (unsigned long long) block);
664 printk(KERN_CONT "%pV\n", &vaf);
668 if (test_opt(sb, ERRORS_CONT)) {
669 ext4_commit_super(sb, 0);
673 ext4_unlock_group(sb, grp);
674 ext4_handle_error(sb);
676 * We only get here in the ERRORS_RO case; relocking the group
677 * may be dangerous, but nothing bad will happen since the
678 * filesystem will have already been marked read/only and the
679 * journal has been aborted. We return 1 as a hint to callers
680 * who might what to use the return value from
681 * ext4_grp_locked_error() to distinguish between the
682 * ERRORS_CONT and ERRORS_RO case, and perhaps return more
683 * aggressively from the ext4 function in question, with a
684 * more appropriate error code.
686 ext4_lock_group(sb, grp);
690 void ext4_update_dynamic_rev(struct super_block *sb)
692 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
694 if (le32_to_cpu(es->s_rev_level) > EXT4_GOOD_OLD_REV)
698 "updating to rev %d because of new feature flag, "
699 "running e2fsck is recommended",
702 es->s_first_ino = cpu_to_le32(EXT4_GOOD_OLD_FIRST_INO);
703 es->s_inode_size = cpu_to_le16(EXT4_GOOD_OLD_INODE_SIZE);
704 es->s_rev_level = cpu_to_le32(EXT4_DYNAMIC_REV);
705 /* leave es->s_feature_*compat flags alone */
706 /* es->s_uuid will be set by e2fsck if empty */
709 * The rest of the superblock fields should be zero, and if not it
710 * means they are likely already in use, so leave them alone. We
711 * can leave it up to e2fsck to clean up any inconsistencies there.
716 * Open the external journal device
718 static struct block_device *ext4_blkdev_get(dev_t dev, struct super_block *sb)
720 struct block_device *bdev;
721 char b[BDEVNAME_SIZE];
723 bdev = blkdev_get_by_dev(dev, FMODE_READ|FMODE_WRITE|FMODE_EXCL, sb);
729 ext4_msg(sb, KERN_ERR, "failed to open journal device %s: %ld",
730 __bdevname(dev, b), PTR_ERR(bdev));
735 * Release the journal device
737 static void ext4_blkdev_put(struct block_device *bdev)
739 blkdev_put(bdev, FMODE_READ|FMODE_WRITE|FMODE_EXCL);
742 static void ext4_blkdev_remove(struct ext4_sb_info *sbi)
744 struct block_device *bdev;
745 bdev = sbi->journal_bdev;
747 ext4_blkdev_put(bdev);
748 sbi->journal_bdev = NULL;
752 static inline struct inode *orphan_list_entry(struct list_head *l)
754 return &list_entry(l, struct ext4_inode_info, i_orphan)->vfs_inode;
757 static void dump_orphan_list(struct super_block *sb, struct ext4_sb_info *sbi)
761 ext4_msg(sb, KERN_ERR, "sb orphan head is %d",
762 le32_to_cpu(sbi->s_es->s_last_orphan));
764 printk(KERN_ERR "sb_info orphan list:\n");
765 list_for_each(l, &sbi->s_orphan) {
766 struct inode *inode = orphan_list_entry(l);
768 "inode %s:%lu at %p: mode %o, nlink %d, next %d\n",
769 inode->i_sb->s_id, inode->i_ino, inode,
770 inode->i_mode, inode->i_nlink,
775 static void ext4_put_super(struct super_block *sb)
777 struct ext4_sb_info *sbi = EXT4_SB(sb);
778 struct ext4_super_block *es = sbi->s_es;
781 ext4_unregister_li_request(sb);
782 dquot_disable(sb, -1, DQUOT_USAGE_ENABLED | DQUOT_LIMITS_ENABLED);
784 flush_workqueue(sbi->rsv_conversion_wq);
785 destroy_workqueue(sbi->rsv_conversion_wq);
787 if (sbi->s_journal) {
788 err = jbd2_journal_destroy(sbi->s_journal);
789 sbi->s_journal = NULL;
791 ext4_abort(sb, "Couldn't clean up the journal");
794 ext4_es_unregister_shrinker(sbi);
795 del_timer_sync(&sbi->s_err_report);
796 ext4_release_system_zone(sb);
798 ext4_ext_release(sb);
799 ext4_xattr_put_super(sb);
801 if (!(sb->s_flags & MS_RDONLY)) {
802 EXT4_CLEAR_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
803 es->s_state = cpu_to_le16(sbi->s_mount_state);
805 if (!(sb->s_flags & MS_RDONLY))
806 ext4_commit_super(sb, 1);
809 remove_proc_entry("options", sbi->s_proc);
810 remove_proc_entry(sb->s_id, ext4_proc_root);
812 kobject_del(&sbi->s_kobj);
814 for (i = 0; i < sbi->s_gdb_count; i++)
815 brelse(sbi->s_group_desc[i]);
816 ext4_kvfree(sbi->s_group_desc);
817 ext4_kvfree(sbi->s_flex_groups);
818 percpu_counter_destroy(&sbi->s_freeclusters_counter);
819 percpu_counter_destroy(&sbi->s_freeinodes_counter);
820 percpu_counter_destroy(&sbi->s_dirs_counter);
821 percpu_counter_destroy(&sbi->s_dirtyclusters_counter);
822 percpu_counter_destroy(&sbi->s_extent_cache_cnt);
825 for (i = 0; i < MAXQUOTAS; i++)
826 kfree(sbi->s_qf_names[i]);
829 /* Debugging code just in case the in-memory inode orphan list
830 * isn't empty. The on-disk one can be non-empty if we've
831 * detected an error and taken the fs readonly, but the
832 * in-memory list had better be clean by this point. */
833 if (!list_empty(&sbi->s_orphan))
834 dump_orphan_list(sb, sbi);
835 J_ASSERT(list_empty(&sbi->s_orphan));
837 invalidate_bdev(sb->s_bdev);
838 if (sbi->journal_bdev && sbi->journal_bdev != sb->s_bdev) {
840 * Invalidate the journal device's buffers. We don't want them
841 * floating about in memory - the physical journal device may
842 * hotswapped, and it breaks the `ro-after' testing code.
844 sync_blockdev(sbi->journal_bdev);
845 invalidate_bdev(sbi->journal_bdev);
846 ext4_blkdev_remove(sbi);
849 kthread_stop(sbi->s_mmp_tsk);
850 sb->s_fs_info = NULL;
852 * Now that we are completely done shutting down the
853 * superblock, we need to actually destroy the kobject.
855 kobject_put(&sbi->s_kobj);
856 wait_for_completion(&sbi->s_kobj_unregister);
857 if (sbi->s_chksum_driver)
858 crypto_free_shash(sbi->s_chksum_driver);
859 kfree(sbi->s_blockgroup_lock);
863 static struct kmem_cache *ext4_inode_cachep;
866 * Called inside transaction, so use GFP_NOFS
868 static struct inode *ext4_alloc_inode(struct super_block *sb)
870 struct ext4_inode_info *ei;
872 ei = kmem_cache_alloc(ext4_inode_cachep, GFP_NOFS);
876 ei->vfs_inode.i_version = 1;
877 INIT_LIST_HEAD(&ei->i_prealloc_list);
878 spin_lock_init(&ei->i_prealloc_lock);
879 ext4_es_init_tree(&ei->i_es_tree);
880 rwlock_init(&ei->i_es_lock);
881 INIT_LIST_HEAD(&ei->i_es_lru);
883 ei->i_touch_when = 0;
884 ei->i_reserved_data_blocks = 0;
885 ei->i_reserved_meta_blocks = 0;
886 ei->i_allocated_meta_blocks = 0;
887 ei->i_da_metadata_calc_len = 0;
888 ei->i_da_metadata_calc_last_lblock = 0;
889 spin_lock_init(&(ei->i_block_reservation_lock));
891 ei->i_reserved_quota = 0;
894 INIT_LIST_HEAD(&ei->i_rsv_conversion_list);
895 spin_lock_init(&ei->i_completed_io_lock);
897 ei->i_datasync_tid = 0;
898 atomic_set(&ei->i_ioend_count, 0);
899 atomic_set(&ei->i_unwritten, 0);
900 INIT_WORK(&ei->i_rsv_conversion_work, ext4_end_io_rsv_work);
902 return &ei->vfs_inode;
905 static int ext4_drop_inode(struct inode *inode)
907 int drop = generic_drop_inode(inode);
909 trace_ext4_drop_inode(inode, drop);
913 static void ext4_i_callback(struct rcu_head *head)
915 struct inode *inode = container_of(head, struct inode, i_rcu);
916 kmem_cache_free(ext4_inode_cachep, EXT4_I(inode));
919 static void ext4_destroy_inode(struct inode *inode)
921 if (!list_empty(&(EXT4_I(inode)->i_orphan))) {
922 ext4_msg(inode->i_sb, KERN_ERR,
923 "Inode %lu (%p): orphan list check failed!",
924 inode->i_ino, EXT4_I(inode));
925 print_hex_dump(KERN_INFO, "", DUMP_PREFIX_ADDRESS, 16, 4,
926 EXT4_I(inode), sizeof(struct ext4_inode_info),
930 call_rcu(&inode->i_rcu, ext4_i_callback);
933 static void init_once(void *foo)
935 struct ext4_inode_info *ei = (struct ext4_inode_info *) foo;
937 INIT_LIST_HEAD(&ei->i_orphan);
938 init_rwsem(&ei->xattr_sem);
939 init_rwsem(&ei->i_data_sem);
940 inode_init_once(&ei->vfs_inode);
943 static int init_inodecache(void)
945 ext4_inode_cachep = kmem_cache_create("ext4_inode_cache",
946 sizeof(struct ext4_inode_info),
947 0, (SLAB_RECLAIM_ACCOUNT|
950 if (ext4_inode_cachep == NULL)
955 static void destroy_inodecache(void)
958 * Make sure all delayed rcu free inodes are flushed before we
962 kmem_cache_destroy(ext4_inode_cachep);
965 void ext4_clear_inode(struct inode *inode)
967 invalidate_inode_buffers(inode);
970 ext4_discard_preallocations(inode);
971 ext4_es_remove_extent(inode, 0, EXT_MAX_BLOCKS);
972 ext4_es_lru_del(inode);
973 if (EXT4_I(inode)->jinode) {
974 jbd2_journal_release_jbd_inode(EXT4_JOURNAL(inode),
975 EXT4_I(inode)->jinode);
976 jbd2_free_inode(EXT4_I(inode)->jinode);
977 EXT4_I(inode)->jinode = NULL;
981 static struct inode *ext4_nfs_get_inode(struct super_block *sb,
982 u64 ino, u32 generation)
986 if (ino < EXT4_FIRST_INO(sb) && ino != EXT4_ROOT_INO)
987 return ERR_PTR(-ESTALE);
988 if (ino > le32_to_cpu(EXT4_SB(sb)->s_es->s_inodes_count))
989 return ERR_PTR(-ESTALE);
991 /* iget isn't really right if the inode is currently unallocated!!
993 * ext4_read_inode will return a bad_inode if the inode had been
994 * deleted, so we should be safe.
996 * Currently we don't know the generation for parent directory, so
997 * a generation of 0 means "accept any"
999 inode = ext4_iget(sb, ino);
1001 return ERR_CAST(inode);
1002 if (generation && inode->i_generation != generation) {
1004 return ERR_PTR(-ESTALE);
1010 static struct dentry *ext4_fh_to_dentry(struct super_block *sb, struct fid *fid,
1011 int fh_len, int fh_type)
1013 return generic_fh_to_dentry(sb, fid, fh_len, fh_type,
1014 ext4_nfs_get_inode);
1017 static struct dentry *ext4_fh_to_parent(struct super_block *sb, struct fid *fid,
1018 int fh_len, int fh_type)
1020 return generic_fh_to_parent(sb, fid, fh_len, fh_type,
1021 ext4_nfs_get_inode);
1025 * Try to release metadata pages (indirect blocks, directories) which are
1026 * mapped via the block device. Since these pages could have journal heads
1027 * which would prevent try_to_free_buffers() from freeing them, we must use
1028 * jbd2 layer's try_to_free_buffers() function to release them.
1030 static int bdev_try_to_free_page(struct super_block *sb, struct page *page,
1033 journal_t *journal = EXT4_SB(sb)->s_journal;
1035 WARN_ON(PageChecked(page));
1036 if (!page_has_buffers(page))
1039 return jbd2_journal_try_to_free_buffers(journal, page,
1040 wait & ~__GFP_WAIT);
1041 return try_to_free_buffers(page);
1045 #define QTYPE2NAME(t) ((t) == USRQUOTA ? "user" : "group")
1046 #define QTYPE2MOPT(on, t) ((t) == USRQUOTA?((on)##USRJQUOTA):((on)##GRPJQUOTA))
1048 static int ext4_write_dquot(struct dquot *dquot);
1049 static int ext4_acquire_dquot(struct dquot *dquot);
1050 static int ext4_release_dquot(struct dquot *dquot);
1051 static int ext4_mark_dquot_dirty(struct dquot *dquot);
1052 static int ext4_write_info(struct super_block *sb, int type);
1053 static int ext4_quota_on(struct super_block *sb, int type, int format_id,
1055 static int ext4_quota_on_sysfile(struct super_block *sb, int type,
1057 static int ext4_quota_off(struct super_block *sb, int type);
1058 static int ext4_quota_off_sysfile(struct super_block *sb, int type);
1059 static int ext4_quota_on_mount(struct super_block *sb, int type);
1060 static ssize_t ext4_quota_read(struct super_block *sb, int type, char *data,
1061 size_t len, loff_t off);
1062 static ssize_t ext4_quota_write(struct super_block *sb, int type,
1063 const char *data, size_t len, loff_t off);
1064 static int ext4_quota_enable(struct super_block *sb, int type, int format_id,
1065 unsigned int flags);
1066 static int ext4_enable_quotas(struct super_block *sb);
1068 static const struct dquot_operations ext4_quota_operations = {
1069 .get_reserved_space = ext4_get_reserved_space,
1070 .write_dquot = ext4_write_dquot,
1071 .acquire_dquot = ext4_acquire_dquot,
1072 .release_dquot = ext4_release_dquot,
1073 .mark_dirty = ext4_mark_dquot_dirty,
1074 .write_info = ext4_write_info,
1075 .alloc_dquot = dquot_alloc,
1076 .destroy_dquot = dquot_destroy,
1079 static const struct quotactl_ops ext4_qctl_operations = {
1080 .quota_on = ext4_quota_on,
1081 .quota_off = ext4_quota_off,
1082 .quota_sync = dquot_quota_sync,
1083 .get_info = dquot_get_dqinfo,
1084 .set_info = dquot_set_dqinfo,
1085 .get_dqblk = dquot_get_dqblk,
1086 .set_dqblk = dquot_set_dqblk
1089 static const struct quotactl_ops ext4_qctl_sysfile_operations = {
1090 .quota_on_meta = ext4_quota_on_sysfile,
1091 .quota_off = ext4_quota_off_sysfile,
1092 .quota_sync = dquot_quota_sync,
1093 .get_info = dquot_get_dqinfo,
1094 .set_info = dquot_set_dqinfo,
1095 .get_dqblk = dquot_get_dqblk,
1096 .set_dqblk = dquot_set_dqblk
1100 static const struct super_operations ext4_sops = {
1101 .alloc_inode = ext4_alloc_inode,
1102 .destroy_inode = ext4_destroy_inode,
1103 .write_inode = ext4_write_inode,
1104 .dirty_inode = ext4_dirty_inode,
1105 .drop_inode = ext4_drop_inode,
1106 .evict_inode = ext4_evict_inode,
1107 .put_super = ext4_put_super,
1108 .sync_fs = ext4_sync_fs,
1109 .freeze_fs = ext4_freeze,
1110 .unfreeze_fs = ext4_unfreeze,
1111 .statfs = ext4_statfs,
1112 .remount_fs = ext4_remount,
1113 .show_options = ext4_show_options,
1115 .quota_read = ext4_quota_read,
1116 .quota_write = ext4_quota_write,
1118 .bdev_try_to_free_page = bdev_try_to_free_page,
1121 static const struct super_operations ext4_nojournal_sops = {
1122 .alloc_inode = ext4_alloc_inode,
1123 .destroy_inode = ext4_destroy_inode,
1124 .write_inode = ext4_write_inode,
1125 .dirty_inode = ext4_dirty_inode,
1126 .drop_inode = ext4_drop_inode,
1127 .evict_inode = ext4_evict_inode,
1128 .sync_fs = ext4_sync_fs_nojournal,
1129 .put_super = ext4_put_super,
1130 .statfs = ext4_statfs,
1131 .remount_fs = ext4_remount,
1132 .show_options = ext4_show_options,
1134 .quota_read = ext4_quota_read,
1135 .quota_write = ext4_quota_write,
1137 .bdev_try_to_free_page = bdev_try_to_free_page,
1140 static const struct export_operations ext4_export_ops = {
1141 .fh_to_dentry = ext4_fh_to_dentry,
1142 .fh_to_parent = ext4_fh_to_parent,
1143 .get_parent = ext4_get_parent,
1147 Opt_bsd_df, Opt_minix_df, Opt_grpid, Opt_nogrpid,
1148 Opt_resgid, Opt_resuid, Opt_sb, Opt_err_cont, Opt_err_panic, Opt_err_ro,
1149 Opt_nouid32, Opt_debug, Opt_removed,
1150 Opt_user_xattr, Opt_nouser_xattr, Opt_acl, Opt_noacl,
1151 Opt_auto_da_alloc, Opt_noauto_da_alloc, Opt_noload,
1152 Opt_commit, Opt_min_batch_time, Opt_max_batch_time, Opt_journal_dev,
1153 Opt_journal_path, Opt_journal_checksum, Opt_journal_async_commit,
1154 Opt_abort, Opt_data_journal, Opt_data_ordered, Opt_data_writeback,
1155 Opt_data_err_abort, Opt_data_err_ignore,
1156 Opt_usrjquota, Opt_grpjquota, Opt_offusrjquota, Opt_offgrpjquota,
1157 Opt_jqfmt_vfsold, Opt_jqfmt_vfsv0, Opt_jqfmt_vfsv1, Opt_quota,
1158 Opt_noquota, Opt_barrier, Opt_nobarrier, Opt_err,
1159 Opt_usrquota, Opt_grpquota, Opt_i_version,
1160 Opt_stripe, Opt_delalloc, Opt_nodelalloc, Opt_mblk_io_submit,
1161 Opt_nomblk_io_submit, Opt_block_validity, Opt_noblock_validity,
1162 Opt_inode_readahead_blks, Opt_journal_ioprio,
1163 Opt_dioread_nolock, Opt_dioread_lock,
1164 Opt_discard, Opt_nodiscard, Opt_init_itable, Opt_noinit_itable,
1165 Opt_max_dir_size_kb,
1168 static const match_table_t tokens = {
1169 {Opt_bsd_df, "bsddf"},
1170 {Opt_minix_df, "minixdf"},
1171 {Opt_grpid, "grpid"},
1172 {Opt_grpid, "bsdgroups"},
1173 {Opt_nogrpid, "nogrpid"},
1174 {Opt_nogrpid, "sysvgroups"},
1175 {Opt_resgid, "resgid=%u"},
1176 {Opt_resuid, "resuid=%u"},
1178 {Opt_err_cont, "errors=continue"},
1179 {Opt_err_panic, "errors=panic"},
1180 {Opt_err_ro, "errors=remount-ro"},
1181 {Opt_nouid32, "nouid32"},
1182 {Opt_debug, "debug"},
1183 {Opt_removed, "oldalloc"},
1184 {Opt_removed, "orlov"},
1185 {Opt_user_xattr, "user_xattr"},
1186 {Opt_nouser_xattr, "nouser_xattr"},
1188 {Opt_noacl, "noacl"},
1189 {Opt_noload, "norecovery"},
1190 {Opt_noload, "noload"},
1191 {Opt_removed, "nobh"},
1192 {Opt_removed, "bh"},
1193 {Opt_commit, "commit=%u"},
1194 {Opt_min_batch_time, "min_batch_time=%u"},
1195 {Opt_max_batch_time, "max_batch_time=%u"},
1196 {Opt_journal_dev, "journal_dev=%u"},
1197 {Opt_journal_path, "journal_path=%s"},
1198 {Opt_journal_checksum, "journal_checksum"},
1199 {Opt_journal_async_commit, "journal_async_commit"},
1200 {Opt_abort, "abort"},
1201 {Opt_data_journal, "data=journal"},
1202 {Opt_data_ordered, "data=ordered"},
1203 {Opt_data_writeback, "data=writeback"},
1204 {Opt_data_err_abort, "data_err=abort"},
1205 {Opt_data_err_ignore, "data_err=ignore"},
1206 {Opt_offusrjquota, "usrjquota="},
1207 {Opt_usrjquota, "usrjquota=%s"},
1208 {Opt_offgrpjquota, "grpjquota="},
1209 {Opt_grpjquota, "grpjquota=%s"},
1210 {Opt_jqfmt_vfsold, "jqfmt=vfsold"},
1211 {Opt_jqfmt_vfsv0, "jqfmt=vfsv0"},
1212 {Opt_jqfmt_vfsv1, "jqfmt=vfsv1"},
1213 {Opt_grpquota, "grpquota"},
1214 {Opt_noquota, "noquota"},
1215 {Opt_quota, "quota"},
1216 {Opt_usrquota, "usrquota"},
1217 {Opt_barrier, "barrier=%u"},
1218 {Opt_barrier, "barrier"},
1219 {Opt_nobarrier, "nobarrier"},
1220 {Opt_i_version, "i_version"},
1221 {Opt_stripe, "stripe=%u"},
1222 {Opt_delalloc, "delalloc"},
1223 {Opt_nodelalloc, "nodelalloc"},
1224 {Opt_removed, "mblk_io_submit"},
1225 {Opt_removed, "nomblk_io_submit"},
1226 {Opt_block_validity, "block_validity"},
1227 {Opt_noblock_validity, "noblock_validity"},
1228 {Opt_inode_readahead_blks, "inode_readahead_blks=%u"},
1229 {Opt_journal_ioprio, "journal_ioprio=%u"},
1230 {Opt_auto_da_alloc, "auto_da_alloc=%u"},
1231 {Opt_auto_da_alloc, "auto_da_alloc"},
1232 {Opt_noauto_da_alloc, "noauto_da_alloc"},
1233 {Opt_dioread_nolock, "dioread_nolock"},
1234 {Opt_dioread_lock, "dioread_lock"},
1235 {Opt_discard, "discard"},
1236 {Opt_nodiscard, "nodiscard"},
1237 {Opt_init_itable, "init_itable=%u"},
1238 {Opt_init_itable, "init_itable"},
1239 {Opt_noinit_itable, "noinit_itable"},
1240 {Opt_max_dir_size_kb, "max_dir_size_kb=%u"},
1241 {Opt_removed, "check=none"}, /* mount option from ext2/3 */
1242 {Opt_removed, "nocheck"}, /* mount option from ext2/3 */
1243 {Opt_removed, "reservation"}, /* mount option from ext2/3 */
1244 {Opt_removed, "noreservation"}, /* mount option from ext2/3 */
1245 {Opt_removed, "journal=%u"}, /* mount option from ext2/3 */
1249 static ext4_fsblk_t get_sb_block(void **data)
1251 ext4_fsblk_t sb_block;
1252 char *options = (char *) *data;
1254 if (!options || strncmp(options, "sb=", 3) != 0)
1255 return 1; /* Default location */
1258 /* TODO: use simple_strtoll with >32bit ext4 */
1259 sb_block = simple_strtoul(options, &options, 0);
1260 if (*options && *options != ',') {
1261 printk(KERN_ERR "EXT4-fs: Invalid sb specification: %s\n",
1265 if (*options == ',')
1267 *data = (void *) options;
1272 #define DEFAULT_JOURNAL_IOPRIO (IOPRIO_PRIO_VALUE(IOPRIO_CLASS_BE, 3))
1273 static char deprecated_msg[] = "Mount option \"%s\" will be removed by %s\n"
1274 "Contact linux-ext4@vger.kernel.org if you think we should keep it.\n";
1277 static int set_qf_name(struct super_block *sb, int qtype, substring_t *args)
1279 struct ext4_sb_info *sbi = EXT4_SB(sb);
1283 if (sb_any_quota_loaded(sb) &&
1284 !sbi->s_qf_names[qtype]) {
1285 ext4_msg(sb, KERN_ERR,
1286 "Cannot change journaled "
1287 "quota options when quota turned on");
1290 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_QUOTA)) {
1291 ext4_msg(sb, KERN_ERR, "Cannot set journaled quota options "
1292 "when QUOTA feature is enabled");
1295 qname = match_strdup(args);
1297 ext4_msg(sb, KERN_ERR,
1298 "Not enough memory for storing quotafile name");
1301 if (sbi->s_qf_names[qtype]) {
1302 if (strcmp(sbi->s_qf_names[qtype], qname) == 0)
1305 ext4_msg(sb, KERN_ERR,
1306 "%s quota file already specified",
1310 if (strchr(qname, '/')) {
1311 ext4_msg(sb, KERN_ERR,
1312 "quotafile must be on filesystem root");
1315 sbi->s_qf_names[qtype] = qname;
1323 static int clear_qf_name(struct super_block *sb, int qtype)
1326 struct ext4_sb_info *sbi = EXT4_SB(sb);
1328 if (sb_any_quota_loaded(sb) &&
1329 sbi->s_qf_names[qtype]) {
1330 ext4_msg(sb, KERN_ERR, "Cannot change journaled quota options"
1331 " when quota turned on");
1334 kfree(sbi->s_qf_names[qtype]);
1335 sbi->s_qf_names[qtype] = NULL;
1340 #define MOPT_SET 0x0001
1341 #define MOPT_CLEAR 0x0002
1342 #define MOPT_NOSUPPORT 0x0004
1343 #define MOPT_EXPLICIT 0x0008
1344 #define MOPT_CLEAR_ERR 0x0010
1345 #define MOPT_GTE0 0x0020
1348 #define MOPT_QFMT 0x0040
1350 #define MOPT_Q MOPT_NOSUPPORT
1351 #define MOPT_QFMT MOPT_NOSUPPORT
1353 #define MOPT_DATAJ 0x0080
1354 #define MOPT_NO_EXT2 0x0100
1355 #define MOPT_NO_EXT3 0x0200
1356 #define MOPT_EXT4_ONLY (MOPT_NO_EXT2 | MOPT_NO_EXT3)
1357 #define MOPT_STRING 0x0400
1359 static const struct mount_opts {
1363 } ext4_mount_opts[] = {
1364 {Opt_minix_df, EXT4_MOUNT_MINIX_DF, MOPT_SET},
1365 {Opt_bsd_df, EXT4_MOUNT_MINIX_DF, MOPT_CLEAR},
1366 {Opt_grpid, EXT4_MOUNT_GRPID, MOPT_SET},
1367 {Opt_nogrpid, EXT4_MOUNT_GRPID, MOPT_CLEAR},
1368 {Opt_block_validity, EXT4_MOUNT_BLOCK_VALIDITY, MOPT_SET},
1369 {Opt_noblock_validity, EXT4_MOUNT_BLOCK_VALIDITY, MOPT_CLEAR},
1370 {Opt_dioread_nolock, EXT4_MOUNT_DIOREAD_NOLOCK,
1371 MOPT_EXT4_ONLY | MOPT_SET},
1372 {Opt_dioread_lock, EXT4_MOUNT_DIOREAD_NOLOCK,
1373 MOPT_EXT4_ONLY | MOPT_CLEAR},
1374 {Opt_discard, EXT4_MOUNT_DISCARD, MOPT_SET},
1375 {Opt_nodiscard, EXT4_MOUNT_DISCARD, MOPT_CLEAR},
1376 {Opt_delalloc, EXT4_MOUNT_DELALLOC,
1377 MOPT_EXT4_ONLY | MOPT_SET | MOPT_EXPLICIT},
1378 {Opt_nodelalloc, EXT4_MOUNT_DELALLOC,
1379 MOPT_EXT4_ONLY | MOPT_CLEAR},
1380 {Opt_journal_checksum, EXT4_MOUNT_JOURNAL_CHECKSUM,
1381 MOPT_EXT4_ONLY | MOPT_SET},
1382 {Opt_journal_async_commit, (EXT4_MOUNT_JOURNAL_ASYNC_COMMIT |
1383 EXT4_MOUNT_JOURNAL_CHECKSUM),
1384 MOPT_EXT4_ONLY | MOPT_SET},
1385 {Opt_noload, EXT4_MOUNT_NOLOAD, MOPT_NO_EXT2 | MOPT_SET},
1386 {Opt_err_panic, EXT4_MOUNT_ERRORS_PANIC, MOPT_SET | MOPT_CLEAR_ERR},
1387 {Opt_err_ro, EXT4_MOUNT_ERRORS_RO, MOPT_SET | MOPT_CLEAR_ERR},
1388 {Opt_err_cont, EXT4_MOUNT_ERRORS_CONT, MOPT_SET | MOPT_CLEAR_ERR},
1389 {Opt_data_err_abort, EXT4_MOUNT_DATA_ERR_ABORT,
1390 MOPT_NO_EXT2 | MOPT_SET},
1391 {Opt_data_err_ignore, EXT4_MOUNT_DATA_ERR_ABORT,
1392 MOPT_NO_EXT2 | MOPT_CLEAR},
1393 {Opt_barrier, EXT4_MOUNT_BARRIER, MOPT_SET},
1394 {Opt_nobarrier, EXT4_MOUNT_BARRIER, MOPT_CLEAR},
1395 {Opt_noauto_da_alloc, EXT4_MOUNT_NO_AUTO_DA_ALLOC, MOPT_SET},
1396 {Opt_auto_da_alloc, EXT4_MOUNT_NO_AUTO_DA_ALLOC, MOPT_CLEAR},
1397 {Opt_noinit_itable, EXT4_MOUNT_INIT_INODE_TABLE, MOPT_CLEAR},
1398 {Opt_commit, 0, MOPT_GTE0},
1399 {Opt_max_batch_time, 0, MOPT_GTE0},
1400 {Opt_min_batch_time, 0, MOPT_GTE0},
1401 {Opt_inode_readahead_blks, 0, MOPT_GTE0},
1402 {Opt_init_itable, 0, MOPT_GTE0},
1403 {Opt_stripe, 0, MOPT_GTE0},
1404 {Opt_resuid, 0, MOPT_GTE0},
1405 {Opt_resgid, 0, MOPT_GTE0},
1406 {Opt_journal_dev, 0, MOPT_GTE0},
1407 {Opt_journal_path, 0, MOPT_STRING},
1408 {Opt_journal_ioprio, 0, MOPT_GTE0},
1409 {Opt_data_journal, EXT4_MOUNT_JOURNAL_DATA, MOPT_NO_EXT2 | MOPT_DATAJ},
1410 {Opt_data_ordered, EXT4_MOUNT_ORDERED_DATA, MOPT_NO_EXT2 | MOPT_DATAJ},
1411 {Opt_data_writeback, EXT4_MOUNT_WRITEBACK_DATA,
1412 MOPT_NO_EXT2 | MOPT_DATAJ},
1413 {Opt_user_xattr, EXT4_MOUNT_XATTR_USER, MOPT_SET},
1414 {Opt_nouser_xattr, EXT4_MOUNT_XATTR_USER, MOPT_CLEAR},
1415 #ifdef CONFIG_EXT4_FS_POSIX_ACL
1416 {Opt_acl, EXT4_MOUNT_POSIX_ACL, MOPT_SET},
1417 {Opt_noacl, EXT4_MOUNT_POSIX_ACL, MOPT_CLEAR},
1419 {Opt_acl, 0, MOPT_NOSUPPORT},
1420 {Opt_noacl, 0, MOPT_NOSUPPORT},
1422 {Opt_nouid32, EXT4_MOUNT_NO_UID32, MOPT_SET},
1423 {Opt_debug, EXT4_MOUNT_DEBUG, MOPT_SET},
1424 {Opt_quota, EXT4_MOUNT_QUOTA | EXT4_MOUNT_USRQUOTA, MOPT_SET | MOPT_Q},
1425 {Opt_usrquota, EXT4_MOUNT_QUOTA | EXT4_MOUNT_USRQUOTA,
1427 {Opt_grpquota, EXT4_MOUNT_QUOTA | EXT4_MOUNT_GRPQUOTA,
1429 {Opt_noquota, (EXT4_MOUNT_QUOTA | EXT4_MOUNT_USRQUOTA |
1430 EXT4_MOUNT_GRPQUOTA), MOPT_CLEAR | MOPT_Q},
1431 {Opt_usrjquota, 0, MOPT_Q},
1432 {Opt_grpjquota, 0, MOPT_Q},
1433 {Opt_offusrjquota, 0, MOPT_Q},
1434 {Opt_offgrpjquota, 0, MOPT_Q},
1435 {Opt_jqfmt_vfsold, QFMT_VFS_OLD, MOPT_QFMT},
1436 {Opt_jqfmt_vfsv0, QFMT_VFS_V0, MOPT_QFMT},
1437 {Opt_jqfmt_vfsv1, QFMT_VFS_V1, MOPT_QFMT},
1438 {Opt_max_dir_size_kb, 0, MOPT_GTE0},
1442 static int handle_mount_opt(struct super_block *sb, char *opt, int token,
1443 substring_t *args, unsigned long *journal_devnum,
1444 unsigned int *journal_ioprio, int is_remount)
1446 struct ext4_sb_info *sbi = EXT4_SB(sb);
1447 const struct mount_opts *m;
1453 if (token == Opt_usrjquota)
1454 return set_qf_name(sb, USRQUOTA, &args[0]);
1455 else if (token == Opt_grpjquota)
1456 return set_qf_name(sb, GRPQUOTA, &args[0]);
1457 else if (token == Opt_offusrjquota)
1458 return clear_qf_name(sb, USRQUOTA);
1459 else if (token == Opt_offgrpjquota)
1460 return clear_qf_name(sb, GRPQUOTA);
1464 case Opt_nouser_xattr:
1465 ext4_msg(sb, KERN_WARNING, deprecated_msg, opt, "3.5");
1468 return 1; /* handled by get_sb_block() */
1470 ext4_msg(sb, KERN_WARNING, "Ignoring removed %s option", opt);
1473 sbi->s_mount_flags |= EXT4_MF_FS_ABORTED;
1476 sb->s_flags |= MS_I_VERSION;
1480 for (m = ext4_mount_opts; m->token != Opt_err; m++)
1481 if (token == m->token)
1484 if (m->token == Opt_err) {
1485 ext4_msg(sb, KERN_ERR, "Unrecognized mount option \"%s\" "
1486 "or missing value", opt);
1490 if ((m->flags & MOPT_NO_EXT2) && IS_EXT2_SB(sb)) {
1491 ext4_msg(sb, KERN_ERR,
1492 "Mount option \"%s\" incompatible with ext2", opt);
1495 if ((m->flags & MOPT_NO_EXT3) && IS_EXT3_SB(sb)) {
1496 ext4_msg(sb, KERN_ERR,
1497 "Mount option \"%s\" incompatible with ext3", opt);
1501 if (args->from && !(m->flags & MOPT_STRING) && match_int(args, &arg))
1503 if (args->from && (m->flags & MOPT_GTE0) && (arg < 0))
1505 if (m->flags & MOPT_EXPLICIT)
1506 set_opt2(sb, EXPLICIT_DELALLOC);
1507 if (m->flags & MOPT_CLEAR_ERR)
1508 clear_opt(sb, ERRORS_MASK);
1509 if (token == Opt_noquota && sb_any_quota_loaded(sb)) {
1510 ext4_msg(sb, KERN_ERR, "Cannot change quota "
1511 "options when quota turned on");
1515 if (m->flags & MOPT_NOSUPPORT) {
1516 ext4_msg(sb, KERN_ERR, "%s option not supported", opt);
1517 } else if (token == Opt_commit) {
1519 arg = JBD2_DEFAULT_MAX_COMMIT_AGE;
1520 sbi->s_commit_interval = HZ * arg;
1521 } else if (token == Opt_max_batch_time) {
1522 sbi->s_max_batch_time = arg;
1523 } else if (token == Opt_min_batch_time) {
1524 sbi->s_min_batch_time = arg;
1525 } else if (token == Opt_inode_readahead_blks) {
1526 if (arg && (arg > (1 << 30) || !is_power_of_2(arg))) {
1527 ext4_msg(sb, KERN_ERR,
1528 "EXT4-fs: inode_readahead_blks must be "
1529 "0 or a power of 2 smaller than 2^31");
1532 sbi->s_inode_readahead_blks = arg;
1533 } else if (token == Opt_init_itable) {
1534 set_opt(sb, INIT_INODE_TABLE);
1536 arg = EXT4_DEF_LI_WAIT_MULT;
1537 sbi->s_li_wait_mult = arg;
1538 } else if (token == Opt_max_dir_size_kb) {
1539 sbi->s_max_dir_size_kb = arg;
1540 } else if (token == Opt_stripe) {
1541 sbi->s_stripe = arg;
1542 } else if (token == Opt_resuid) {
1543 uid = make_kuid(current_user_ns(), arg);
1544 if (!uid_valid(uid)) {
1545 ext4_msg(sb, KERN_ERR, "Invalid uid value %d", arg);
1548 sbi->s_resuid = uid;
1549 } else if (token == Opt_resgid) {
1550 gid = make_kgid(current_user_ns(), arg);
1551 if (!gid_valid(gid)) {
1552 ext4_msg(sb, KERN_ERR, "Invalid gid value %d", arg);
1555 sbi->s_resgid = gid;
1556 } else if (token == Opt_journal_dev) {
1558 ext4_msg(sb, KERN_ERR,
1559 "Cannot specify journal on remount");
1562 *journal_devnum = arg;
1563 } else if (token == Opt_journal_path) {
1565 struct inode *journal_inode;
1570 ext4_msg(sb, KERN_ERR,
1571 "Cannot specify journal on remount");
1574 journal_path = match_strdup(&args[0]);
1575 if (!journal_path) {
1576 ext4_msg(sb, KERN_ERR, "error: could not dup "
1577 "journal device string");
1581 error = kern_path(journal_path, LOOKUP_FOLLOW, &path);
1583 ext4_msg(sb, KERN_ERR, "error: could not find "
1584 "journal device path: error %d", error);
1585 kfree(journal_path);
1589 journal_inode = path.dentry->d_inode;
1590 if (!S_ISBLK(journal_inode->i_mode)) {
1591 ext4_msg(sb, KERN_ERR, "error: journal path %s "
1592 "is not a block device", journal_path);
1594 kfree(journal_path);
1598 *journal_devnum = new_encode_dev(journal_inode->i_rdev);
1600 kfree(journal_path);
1601 } else if (token == Opt_journal_ioprio) {
1603 ext4_msg(sb, KERN_ERR, "Invalid journal IO priority"
1608 IOPRIO_PRIO_VALUE(IOPRIO_CLASS_BE, arg);
1609 } else if (m->flags & MOPT_DATAJ) {
1611 if (!sbi->s_journal)
1612 ext4_msg(sb, KERN_WARNING, "Remounting file system with no journal so ignoring journalled data option");
1613 else if (test_opt(sb, DATA_FLAGS) != m->mount_opt) {
1614 ext4_msg(sb, KERN_ERR,
1615 "Cannot change data mode on remount");
1619 clear_opt(sb, DATA_FLAGS);
1620 sbi->s_mount_opt |= m->mount_opt;
1623 } else if (m->flags & MOPT_QFMT) {
1624 if (sb_any_quota_loaded(sb) &&
1625 sbi->s_jquota_fmt != m->mount_opt) {
1626 ext4_msg(sb, KERN_ERR, "Cannot change journaled "
1627 "quota options when quota turned on");
1630 if (EXT4_HAS_RO_COMPAT_FEATURE(sb,
1631 EXT4_FEATURE_RO_COMPAT_QUOTA)) {
1632 ext4_msg(sb, KERN_ERR,
1633 "Cannot set journaled quota options "
1634 "when QUOTA feature is enabled");
1637 sbi->s_jquota_fmt = m->mount_opt;
1642 if (m->flags & MOPT_CLEAR)
1644 else if (unlikely(!(m->flags & MOPT_SET))) {
1645 ext4_msg(sb, KERN_WARNING,
1646 "buggy handling of option %s", opt);
1651 sbi->s_mount_opt |= m->mount_opt;
1653 sbi->s_mount_opt &= ~m->mount_opt;
1658 static int parse_options(char *options, struct super_block *sb,
1659 unsigned long *journal_devnum,
1660 unsigned int *journal_ioprio,
1663 struct ext4_sb_info *sbi = EXT4_SB(sb);
1665 substring_t args[MAX_OPT_ARGS];
1671 while ((p = strsep(&options, ",")) != NULL) {
1675 * Initialize args struct so we know whether arg was
1676 * found; some options take optional arguments.
1678 args[0].to = args[0].from = NULL;
1679 token = match_token(p, tokens, args);
1680 if (handle_mount_opt(sb, p, token, args, journal_devnum,
1681 journal_ioprio, is_remount) < 0)
1685 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_QUOTA) &&
1686 (test_opt(sb, USRQUOTA) || test_opt(sb, GRPQUOTA))) {
1687 ext4_msg(sb, KERN_ERR, "Cannot set quota options when QUOTA "
1688 "feature is enabled");
1691 if (sbi->s_qf_names[USRQUOTA] || sbi->s_qf_names[GRPQUOTA]) {
1692 if (test_opt(sb, USRQUOTA) && sbi->s_qf_names[USRQUOTA])
1693 clear_opt(sb, USRQUOTA);
1695 if (test_opt(sb, GRPQUOTA) && sbi->s_qf_names[GRPQUOTA])
1696 clear_opt(sb, GRPQUOTA);
1698 if (test_opt(sb, GRPQUOTA) || test_opt(sb, USRQUOTA)) {
1699 ext4_msg(sb, KERN_ERR, "old and new quota "
1704 if (!sbi->s_jquota_fmt) {
1705 ext4_msg(sb, KERN_ERR, "journaled quota format "
1711 if (test_opt(sb, DIOREAD_NOLOCK)) {
1713 BLOCK_SIZE << le32_to_cpu(sbi->s_es->s_log_block_size);
1715 if (blocksize < PAGE_CACHE_SIZE) {
1716 ext4_msg(sb, KERN_ERR, "can't mount with "
1717 "dioread_nolock if block size != PAGE_SIZE");
1724 static inline void ext4_show_quota_options(struct seq_file *seq,
1725 struct super_block *sb)
1727 #if defined(CONFIG_QUOTA)
1728 struct ext4_sb_info *sbi = EXT4_SB(sb);
1730 if (sbi->s_jquota_fmt) {
1733 switch (sbi->s_jquota_fmt) {
1744 seq_printf(seq, ",jqfmt=%s", fmtname);
1747 if (sbi->s_qf_names[USRQUOTA])
1748 seq_printf(seq, ",usrjquota=%s", sbi->s_qf_names[USRQUOTA]);
1750 if (sbi->s_qf_names[GRPQUOTA])
1751 seq_printf(seq, ",grpjquota=%s", sbi->s_qf_names[GRPQUOTA]);
1755 static const char *token2str(int token)
1757 const struct match_token *t;
1759 for (t = tokens; t->token != Opt_err; t++)
1760 if (t->token == token && !strchr(t->pattern, '='))
1767 * - it's set to a non-default value OR
1768 * - if the per-sb default is different from the global default
1770 static int _ext4_show_options(struct seq_file *seq, struct super_block *sb,
1773 struct ext4_sb_info *sbi = EXT4_SB(sb);
1774 struct ext4_super_block *es = sbi->s_es;
1775 int def_errors, def_mount_opt = nodefs ? 0 : sbi->s_def_mount_opt;
1776 const struct mount_opts *m;
1777 char sep = nodefs ? '\n' : ',';
1779 #define SEQ_OPTS_PUTS(str) seq_printf(seq, "%c" str, sep)
1780 #define SEQ_OPTS_PRINT(str, arg) seq_printf(seq, "%c" str, sep, arg)
1782 if (sbi->s_sb_block != 1)
1783 SEQ_OPTS_PRINT("sb=%llu", sbi->s_sb_block);
1785 for (m = ext4_mount_opts; m->token != Opt_err; m++) {
1786 int want_set = m->flags & MOPT_SET;
1787 if (((m->flags & (MOPT_SET|MOPT_CLEAR)) == 0) ||
1788 (m->flags & MOPT_CLEAR_ERR))
1790 if (!(m->mount_opt & (sbi->s_mount_opt ^ def_mount_opt)))
1791 continue; /* skip if same as the default */
1793 (sbi->s_mount_opt & m->mount_opt) != m->mount_opt) ||
1794 (!want_set && (sbi->s_mount_opt & m->mount_opt)))
1795 continue; /* select Opt_noFoo vs Opt_Foo */
1796 SEQ_OPTS_PRINT("%s", token2str(m->token));
1799 if (nodefs || !uid_eq(sbi->s_resuid, make_kuid(&init_user_ns, EXT4_DEF_RESUID)) ||
1800 le16_to_cpu(es->s_def_resuid) != EXT4_DEF_RESUID)
1801 SEQ_OPTS_PRINT("resuid=%u",
1802 from_kuid_munged(&init_user_ns, sbi->s_resuid));
1803 if (nodefs || !gid_eq(sbi->s_resgid, make_kgid(&init_user_ns, EXT4_DEF_RESGID)) ||
1804 le16_to_cpu(es->s_def_resgid) != EXT4_DEF_RESGID)
1805 SEQ_OPTS_PRINT("resgid=%u",
1806 from_kgid_munged(&init_user_ns, sbi->s_resgid));
1807 def_errors = nodefs ? -1 : le16_to_cpu(es->s_errors);
1808 if (test_opt(sb, ERRORS_RO) && def_errors != EXT4_ERRORS_RO)
1809 SEQ_OPTS_PUTS("errors=remount-ro");
1810 if (test_opt(sb, ERRORS_CONT) && def_errors != EXT4_ERRORS_CONTINUE)
1811 SEQ_OPTS_PUTS("errors=continue");
1812 if (test_opt(sb, ERRORS_PANIC) && def_errors != EXT4_ERRORS_PANIC)
1813 SEQ_OPTS_PUTS("errors=panic");
1814 if (nodefs || sbi->s_commit_interval != JBD2_DEFAULT_MAX_COMMIT_AGE*HZ)
1815 SEQ_OPTS_PRINT("commit=%lu", sbi->s_commit_interval / HZ);
1816 if (nodefs || sbi->s_min_batch_time != EXT4_DEF_MIN_BATCH_TIME)
1817 SEQ_OPTS_PRINT("min_batch_time=%u", sbi->s_min_batch_time);
1818 if (nodefs || sbi->s_max_batch_time != EXT4_DEF_MAX_BATCH_TIME)
1819 SEQ_OPTS_PRINT("max_batch_time=%u", sbi->s_max_batch_time);
1820 if (sb->s_flags & MS_I_VERSION)
1821 SEQ_OPTS_PUTS("i_version");
1822 if (nodefs || sbi->s_stripe)
1823 SEQ_OPTS_PRINT("stripe=%lu", sbi->s_stripe);
1824 if (EXT4_MOUNT_DATA_FLAGS & (sbi->s_mount_opt ^ def_mount_opt)) {
1825 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA)
1826 SEQ_OPTS_PUTS("data=journal");
1827 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA)
1828 SEQ_OPTS_PUTS("data=ordered");
1829 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_WRITEBACK_DATA)
1830 SEQ_OPTS_PUTS("data=writeback");
1833 sbi->s_inode_readahead_blks != EXT4_DEF_INODE_READAHEAD_BLKS)
1834 SEQ_OPTS_PRINT("inode_readahead_blks=%u",
1835 sbi->s_inode_readahead_blks);
1837 if (nodefs || (test_opt(sb, INIT_INODE_TABLE) &&
1838 (sbi->s_li_wait_mult != EXT4_DEF_LI_WAIT_MULT)))
1839 SEQ_OPTS_PRINT("init_itable=%u", sbi->s_li_wait_mult);
1840 if (nodefs || sbi->s_max_dir_size_kb)
1841 SEQ_OPTS_PRINT("max_dir_size_kb=%u", sbi->s_max_dir_size_kb);
1843 ext4_show_quota_options(seq, sb);
1847 static int ext4_show_options(struct seq_file *seq, struct dentry *root)
1849 return _ext4_show_options(seq, root->d_sb, 0);
1852 static int options_seq_show(struct seq_file *seq, void *offset)
1854 struct super_block *sb = seq->private;
1857 seq_puts(seq, (sb->s_flags & MS_RDONLY) ? "ro" : "rw");
1858 rc = _ext4_show_options(seq, sb, 1);
1859 seq_puts(seq, "\n");
1863 static int options_open_fs(struct inode *inode, struct file *file)
1865 return single_open(file, options_seq_show, PDE_DATA(inode));
1868 static const struct file_operations ext4_seq_options_fops = {
1869 .owner = THIS_MODULE,
1870 .open = options_open_fs,
1872 .llseek = seq_lseek,
1873 .release = single_release,
1876 static int ext4_setup_super(struct super_block *sb, struct ext4_super_block *es,
1879 struct ext4_sb_info *sbi = EXT4_SB(sb);
1882 if (le32_to_cpu(es->s_rev_level) > EXT4_MAX_SUPP_REV) {
1883 ext4_msg(sb, KERN_ERR, "revision level too high, "
1884 "forcing read-only mode");
1889 if (!(sbi->s_mount_state & EXT4_VALID_FS))
1890 ext4_msg(sb, KERN_WARNING, "warning: mounting unchecked fs, "
1891 "running e2fsck is recommended");
1892 else if ((sbi->s_mount_state & EXT4_ERROR_FS))
1893 ext4_msg(sb, KERN_WARNING,
1894 "warning: mounting fs with errors, "
1895 "running e2fsck is recommended");
1896 else if ((__s16) le16_to_cpu(es->s_max_mnt_count) > 0 &&
1897 le16_to_cpu(es->s_mnt_count) >=
1898 (unsigned short) (__s16) le16_to_cpu(es->s_max_mnt_count))
1899 ext4_msg(sb, KERN_WARNING,
1900 "warning: maximal mount count reached, "
1901 "running e2fsck is recommended");
1902 else if (le32_to_cpu(es->s_checkinterval) &&
1903 (le32_to_cpu(es->s_lastcheck) +
1904 le32_to_cpu(es->s_checkinterval) <= get_seconds()))
1905 ext4_msg(sb, KERN_WARNING,
1906 "warning: checktime reached, "
1907 "running e2fsck is recommended");
1908 if (!sbi->s_journal)
1909 es->s_state &= cpu_to_le16(~EXT4_VALID_FS);
1910 if (!(__s16) le16_to_cpu(es->s_max_mnt_count))
1911 es->s_max_mnt_count = cpu_to_le16(EXT4_DFL_MAX_MNT_COUNT);
1912 le16_add_cpu(&es->s_mnt_count, 1);
1913 es->s_mtime = cpu_to_le32(get_seconds());
1914 ext4_update_dynamic_rev(sb);
1916 EXT4_SET_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
1918 ext4_commit_super(sb, 1);
1920 if (test_opt(sb, DEBUG))
1921 printk(KERN_INFO "[EXT4 FS bs=%lu, gc=%u, "
1922 "bpg=%lu, ipg=%lu, mo=%04x, mo2=%04x]\n",
1924 sbi->s_groups_count,
1925 EXT4_BLOCKS_PER_GROUP(sb),
1926 EXT4_INODES_PER_GROUP(sb),
1927 sbi->s_mount_opt, sbi->s_mount_opt2);
1929 cleancache_init_fs(sb);
1933 int ext4_alloc_flex_bg_array(struct super_block *sb, ext4_group_t ngroup)
1935 struct ext4_sb_info *sbi = EXT4_SB(sb);
1936 struct flex_groups *new_groups;
1939 if (!sbi->s_log_groups_per_flex)
1942 size = ext4_flex_group(sbi, ngroup - 1) + 1;
1943 if (size <= sbi->s_flex_groups_allocated)
1946 size = roundup_pow_of_two(size * sizeof(struct flex_groups));
1947 new_groups = ext4_kvzalloc(size, GFP_KERNEL);
1949 ext4_msg(sb, KERN_ERR, "not enough memory for %d flex groups",
1950 size / (int) sizeof(struct flex_groups));
1954 if (sbi->s_flex_groups) {
1955 memcpy(new_groups, sbi->s_flex_groups,
1956 (sbi->s_flex_groups_allocated *
1957 sizeof(struct flex_groups)));
1958 ext4_kvfree(sbi->s_flex_groups);
1960 sbi->s_flex_groups = new_groups;
1961 sbi->s_flex_groups_allocated = size / sizeof(struct flex_groups);
1965 static int ext4_fill_flex_info(struct super_block *sb)
1967 struct ext4_sb_info *sbi = EXT4_SB(sb);
1968 struct ext4_group_desc *gdp = NULL;
1969 ext4_group_t flex_group;
1972 sbi->s_log_groups_per_flex = sbi->s_es->s_log_groups_per_flex;
1973 if (sbi->s_log_groups_per_flex < 1 || sbi->s_log_groups_per_flex > 31) {
1974 sbi->s_log_groups_per_flex = 0;
1978 err = ext4_alloc_flex_bg_array(sb, sbi->s_groups_count);
1982 for (i = 0; i < sbi->s_groups_count; i++) {
1983 gdp = ext4_get_group_desc(sb, i, NULL);
1985 flex_group = ext4_flex_group(sbi, i);
1986 atomic_add(ext4_free_inodes_count(sb, gdp),
1987 &sbi->s_flex_groups[flex_group].free_inodes);
1988 atomic64_add(ext4_free_group_clusters(sb, gdp),
1989 &sbi->s_flex_groups[flex_group].free_clusters);
1990 atomic_add(ext4_used_dirs_count(sb, gdp),
1991 &sbi->s_flex_groups[flex_group].used_dirs);
1999 static __le16 ext4_group_desc_csum(struct ext4_sb_info *sbi, __u32 block_group,
2000 struct ext4_group_desc *gdp)
2004 __le32 le_group = cpu_to_le32(block_group);
2006 if ((sbi->s_es->s_feature_ro_compat &
2007 cpu_to_le32(EXT4_FEATURE_RO_COMPAT_METADATA_CSUM))) {
2008 /* Use new metadata_csum algorithm */
2012 save_csum = gdp->bg_checksum;
2013 gdp->bg_checksum = 0;
2014 csum32 = ext4_chksum(sbi, sbi->s_csum_seed, (__u8 *)&le_group,
2016 csum32 = ext4_chksum(sbi, csum32, (__u8 *)gdp,
2018 gdp->bg_checksum = save_csum;
2020 crc = csum32 & 0xFFFF;
2024 /* old crc16 code */
2025 offset = offsetof(struct ext4_group_desc, bg_checksum);
2027 crc = crc16(~0, sbi->s_es->s_uuid, sizeof(sbi->s_es->s_uuid));
2028 crc = crc16(crc, (__u8 *)&le_group, sizeof(le_group));
2029 crc = crc16(crc, (__u8 *)gdp, offset);
2030 offset += sizeof(gdp->bg_checksum); /* skip checksum */
2031 /* for checksum of struct ext4_group_desc do the rest...*/
2032 if ((sbi->s_es->s_feature_incompat &
2033 cpu_to_le32(EXT4_FEATURE_INCOMPAT_64BIT)) &&
2034 offset < le16_to_cpu(sbi->s_es->s_desc_size))
2035 crc = crc16(crc, (__u8 *)gdp + offset,
2036 le16_to_cpu(sbi->s_es->s_desc_size) -
2040 return cpu_to_le16(crc);
2043 int ext4_group_desc_csum_verify(struct super_block *sb, __u32 block_group,
2044 struct ext4_group_desc *gdp)
2046 if (ext4_has_group_desc_csum(sb) &&
2047 (gdp->bg_checksum != ext4_group_desc_csum(EXT4_SB(sb),
2054 void ext4_group_desc_csum_set(struct super_block *sb, __u32 block_group,
2055 struct ext4_group_desc *gdp)
2057 if (!ext4_has_group_desc_csum(sb))
2059 gdp->bg_checksum = ext4_group_desc_csum(EXT4_SB(sb), block_group, gdp);
2062 /* Called at mount-time, super-block is locked */
2063 static int ext4_check_descriptors(struct super_block *sb,
2064 ext4_group_t *first_not_zeroed)
2066 struct ext4_sb_info *sbi = EXT4_SB(sb);
2067 ext4_fsblk_t first_block = le32_to_cpu(sbi->s_es->s_first_data_block);
2068 ext4_fsblk_t last_block;
2069 ext4_fsblk_t block_bitmap;
2070 ext4_fsblk_t inode_bitmap;
2071 ext4_fsblk_t inode_table;
2072 int flexbg_flag = 0;
2073 ext4_group_t i, grp = sbi->s_groups_count;
2075 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_FLEX_BG))
2078 ext4_debug("Checking group descriptors");
2080 for (i = 0; i < sbi->s_groups_count; i++) {
2081 struct ext4_group_desc *gdp = ext4_get_group_desc(sb, i, NULL);
2083 if (i == sbi->s_groups_count - 1 || flexbg_flag)
2084 last_block = ext4_blocks_count(sbi->s_es) - 1;
2086 last_block = first_block +
2087 (EXT4_BLOCKS_PER_GROUP(sb) - 1);
2089 if ((grp == sbi->s_groups_count) &&
2090 !(gdp->bg_flags & cpu_to_le16(EXT4_BG_INODE_ZEROED)))
2093 block_bitmap = ext4_block_bitmap(sb, gdp);
2094 if (block_bitmap < first_block || block_bitmap > last_block) {
2095 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2096 "Block bitmap for group %u not in group "
2097 "(block %llu)!", i, block_bitmap);
2100 inode_bitmap = ext4_inode_bitmap(sb, gdp);
2101 if (inode_bitmap < first_block || inode_bitmap > last_block) {
2102 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2103 "Inode bitmap for group %u not in group "
2104 "(block %llu)!", i, inode_bitmap);
2107 inode_table = ext4_inode_table(sb, gdp);
2108 if (inode_table < first_block ||
2109 inode_table + sbi->s_itb_per_group - 1 > last_block) {
2110 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2111 "Inode table for group %u not in group "
2112 "(block %llu)!", i, inode_table);
2115 ext4_lock_group(sb, i);
2116 if (!ext4_group_desc_csum_verify(sb, i, gdp)) {
2117 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2118 "Checksum for group %u failed (%u!=%u)",
2119 i, le16_to_cpu(ext4_group_desc_csum(sbi, i,
2120 gdp)), le16_to_cpu(gdp->bg_checksum));
2121 if (!(sb->s_flags & MS_RDONLY)) {
2122 ext4_unlock_group(sb, i);
2126 ext4_unlock_group(sb, i);
2128 first_block += EXT4_BLOCKS_PER_GROUP(sb);
2130 if (NULL != first_not_zeroed)
2131 *first_not_zeroed = grp;
2133 ext4_free_blocks_count_set(sbi->s_es,
2134 EXT4_C2B(sbi, ext4_count_free_clusters(sb)));
2135 sbi->s_es->s_free_inodes_count =cpu_to_le32(ext4_count_free_inodes(sb));
2139 /* ext4_orphan_cleanup() walks a singly-linked list of inodes (starting at
2140 * the superblock) which were deleted from all directories, but held open by
2141 * a process at the time of a crash. We walk the list and try to delete these
2142 * inodes at recovery time (only with a read-write filesystem).
2144 * In order to keep the orphan inode chain consistent during traversal (in
2145 * case of crash during recovery), we link each inode into the superblock
2146 * orphan list_head and handle it the same way as an inode deletion during
2147 * normal operation (which journals the operations for us).
2149 * We only do an iget() and an iput() on each inode, which is very safe if we
2150 * accidentally point at an in-use or already deleted inode. The worst that
2151 * can happen in this case is that we get a "bit already cleared" message from
2152 * ext4_free_inode(). The only reason we would point at a wrong inode is if
2153 * e2fsck was run on this filesystem, and it must have already done the orphan
2154 * inode cleanup for us, so we can safely abort without any further action.
2156 static void ext4_orphan_cleanup(struct super_block *sb,
2157 struct ext4_super_block *es)
2159 unsigned int s_flags = sb->s_flags;
2160 int nr_orphans = 0, nr_truncates = 0;
2164 if (!es->s_last_orphan) {
2165 jbd_debug(4, "no orphan inodes to clean up\n");
2169 if (bdev_read_only(sb->s_bdev)) {
2170 ext4_msg(sb, KERN_ERR, "write access "
2171 "unavailable, skipping orphan cleanup");
2175 /* Check if feature set would not allow a r/w mount */
2176 if (!ext4_feature_set_ok(sb, 0)) {
2177 ext4_msg(sb, KERN_INFO, "Skipping orphan cleanup due to "
2178 "unknown ROCOMPAT features");
2182 if (EXT4_SB(sb)->s_mount_state & EXT4_ERROR_FS) {
2183 /* don't clear list on RO mount w/ errors */
2184 if (es->s_last_orphan && !(s_flags & MS_RDONLY)) {
2185 jbd_debug(1, "Errors on filesystem, "
2186 "clearing orphan list.\n");
2187 es->s_last_orphan = 0;
2189 jbd_debug(1, "Skipping orphan recovery on fs with errors.\n");
2193 if (s_flags & MS_RDONLY) {
2194 ext4_msg(sb, KERN_INFO, "orphan cleanup on readonly fs");
2195 sb->s_flags &= ~MS_RDONLY;
2198 /* Needed for iput() to work correctly and not trash data */
2199 sb->s_flags |= MS_ACTIVE;
2200 /* Turn on quotas so that they are updated correctly */
2201 for (i = 0; i < MAXQUOTAS; i++) {
2202 if (EXT4_SB(sb)->s_qf_names[i]) {
2203 int ret = ext4_quota_on_mount(sb, i);
2205 ext4_msg(sb, KERN_ERR,
2206 "Cannot turn on journaled "
2207 "quota: error %d", ret);
2212 while (es->s_last_orphan) {
2213 struct inode *inode;
2215 inode = ext4_orphan_get(sb, le32_to_cpu(es->s_last_orphan));
2216 if (IS_ERR(inode)) {
2217 es->s_last_orphan = 0;
2221 list_add(&EXT4_I(inode)->i_orphan, &EXT4_SB(sb)->s_orphan);
2222 dquot_initialize(inode);
2223 if (inode->i_nlink) {
2224 if (test_opt(sb, DEBUG))
2225 ext4_msg(sb, KERN_DEBUG,
2226 "%s: truncating inode %lu to %lld bytes",
2227 __func__, inode->i_ino, inode->i_size);
2228 jbd_debug(2, "truncating inode %lu to %lld bytes\n",
2229 inode->i_ino, inode->i_size);
2230 mutex_lock(&inode->i_mutex);
2231 truncate_inode_pages(inode->i_mapping, inode->i_size);
2232 ext4_truncate(inode);
2233 mutex_unlock(&inode->i_mutex);
2236 if (test_opt(sb, DEBUG))
2237 ext4_msg(sb, KERN_DEBUG,
2238 "%s: deleting unreferenced inode %lu",
2239 __func__, inode->i_ino);
2240 jbd_debug(2, "deleting unreferenced inode %lu\n",
2244 iput(inode); /* The delete magic happens here! */
2247 #define PLURAL(x) (x), ((x) == 1) ? "" : "s"
2250 ext4_msg(sb, KERN_INFO, "%d orphan inode%s deleted",
2251 PLURAL(nr_orphans));
2253 ext4_msg(sb, KERN_INFO, "%d truncate%s cleaned up",
2254 PLURAL(nr_truncates));
2256 /* Turn quotas off */
2257 for (i = 0; i < MAXQUOTAS; i++) {
2258 if (sb_dqopt(sb)->files[i])
2259 dquot_quota_off(sb, i);
2262 sb->s_flags = s_flags; /* Restore MS_RDONLY status */
2266 * Maximal extent format file size.
2267 * Resulting logical blkno at s_maxbytes must fit in our on-disk
2268 * extent format containers, within a sector_t, and within i_blocks
2269 * in the vfs. ext4 inode has 48 bits of i_block in fsblock units,
2270 * so that won't be a limiting factor.
2272 * However there is other limiting factor. We do store extents in the form
2273 * of starting block and length, hence the resulting length of the extent
2274 * covering maximum file size must fit into on-disk format containers as
2275 * well. Given that length is always by 1 unit bigger than max unit (because
2276 * we count 0 as well) we have to lower the s_maxbytes by one fs block.
2278 * Note, this does *not* consider any metadata overhead for vfs i_blocks.
2280 static loff_t ext4_max_size(int blkbits, int has_huge_files)
2283 loff_t upper_limit = MAX_LFS_FILESIZE;
2285 /* small i_blocks in vfs inode? */
2286 if (!has_huge_files || sizeof(blkcnt_t) < sizeof(u64)) {
2288 * CONFIG_LBDAF is not enabled implies the inode
2289 * i_block represent total blocks in 512 bytes
2290 * 32 == size of vfs inode i_blocks * 8
2292 upper_limit = (1LL << 32) - 1;
2294 /* total blocks in file system block size */
2295 upper_limit >>= (blkbits - 9);
2296 upper_limit <<= blkbits;
2300 * 32-bit extent-start container, ee_block. We lower the maxbytes
2301 * by one fs block, so ee_len can cover the extent of maximum file
2304 res = (1LL << 32) - 1;
2307 /* Sanity check against vm- & vfs- imposed limits */
2308 if (res > upper_limit)
2315 * Maximal bitmap file size. There is a direct, and {,double-,triple-}indirect
2316 * block limit, and also a limit of (2^48 - 1) 512-byte sectors in i_blocks.
2317 * We need to be 1 filesystem block less than the 2^48 sector limit.
2319 static loff_t ext4_max_bitmap_size(int bits, int has_huge_files)
2321 loff_t res = EXT4_NDIR_BLOCKS;
2324 /* This is calculated to be the largest file size for a dense, block
2325 * mapped file such that the file's total number of 512-byte sectors,
2326 * including data and all indirect blocks, does not exceed (2^48 - 1).
2328 * __u32 i_blocks_lo and _u16 i_blocks_high represent the total
2329 * number of 512-byte sectors of the file.
2332 if (!has_huge_files || sizeof(blkcnt_t) < sizeof(u64)) {
2334 * !has_huge_files or CONFIG_LBDAF not enabled implies that
2335 * the inode i_block field represents total file blocks in
2336 * 2^32 512-byte sectors == size of vfs inode i_blocks * 8
2338 upper_limit = (1LL << 32) - 1;
2340 /* total blocks in file system block size */
2341 upper_limit >>= (bits - 9);
2345 * We use 48 bit ext4_inode i_blocks
2346 * With EXT4_HUGE_FILE_FL set the i_blocks
2347 * represent total number of blocks in
2348 * file system block size
2350 upper_limit = (1LL << 48) - 1;
2354 /* indirect blocks */
2356 /* double indirect blocks */
2357 meta_blocks += 1 + (1LL << (bits-2));
2358 /* tripple indirect blocks */
2359 meta_blocks += 1 + (1LL << (bits-2)) + (1LL << (2*(bits-2)));
2361 upper_limit -= meta_blocks;
2362 upper_limit <<= bits;
2364 res += 1LL << (bits-2);
2365 res += 1LL << (2*(bits-2));
2366 res += 1LL << (3*(bits-2));
2368 if (res > upper_limit)
2371 if (res > MAX_LFS_FILESIZE)
2372 res = MAX_LFS_FILESIZE;
2377 static ext4_fsblk_t descriptor_loc(struct super_block *sb,
2378 ext4_fsblk_t logical_sb_block, int nr)
2380 struct ext4_sb_info *sbi = EXT4_SB(sb);
2381 ext4_group_t bg, first_meta_bg;
2384 first_meta_bg = le32_to_cpu(sbi->s_es->s_first_meta_bg);
2386 if (!EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_META_BG) ||
2388 return logical_sb_block + nr + 1;
2389 bg = sbi->s_desc_per_block * nr;
2390 if (ext4_bg_has_super(sb, bg))
2393 return (has_super + ext4_group_first_block_no(sb, bg));
2397 * ext4_get_stripe_size: Get the stripe size.
2398 * @sbi: In memory super block info
2400 * If we have specified it via mount option, then
2401 * use the mount option value. If the value specified at mount time is
2402 * greater than the blocks per group use the super block value.
2403 * If the super block value is greater than blocks per group return 0.
2404 * Allocator needs it be less than blocks per group.
2407 static unsigned long ext4_get_stripe_size(struct ext4_sb_info *sbi)
2409 unsigned long stride = le16_to_cpu(sbi->s_es->s_raid_stride);
2410 unsigned long stripe_width =
2411 le32_to_cpu(sbi->s_es->s_raid_stripe_width);
2414 if (sbi->s_stripe && sbi->s_stripe <= sbi->s_blocks_per_group)
2415 ret = sbi->s_stripe;
2416 else if (stripe_width <= sbi->s_blocks_per_group)
2418 else if (stride <= sbi->s_blocks_per_group)
2424 * If the stripe width is 1, this makes no sense and
2425 * we set it to 0 to turn off stripe handling code.
2436 struct attribute attr;
2437 ssize_t (*show)(struct ext4_attr *, struct ext4_sb_info *, char *);
2438 ssize_t (*store)(struct ext4_attr *, struct ext4_sb_info *,
2439 const char *, size_t);
2446 static int parse_strtoull(const char *buf,
2447 unsigned long long max, unsigned long long *value)
2451 ret = kstrtoull(skip_spaces(buf), 0, value);
2452 if (!ret && *value > max)
2457 static ssize_t delayed_allocation_blocks_show(struct ext4_attr *a,
2458 struct ext4_sb_info *sbi,
2461 return snprintf(buf, PAGE_SIZE, "%llu\n",
2463 percpu_counter_sum(&sbi->s_dirtyclusters_counter)));
2466 static ssize_t session_write_kbytes_show(struct ext4_attr *a,
2467 struct ext4_sb_info *sbi, char *buf)
2469 struct super_block *sb = sbi->s_buddy_cache->i_sb;
2471 if (!sb->s_bdev->bd_part)
2472 return snprintf(buf, PAGE_SIZE, "0\n");
2473 return snprintf(buf, PAGE_SIZE, "%lu\n",
2474 (part_stat_read(sb->s_bdev->bd_part, sectors[1]) -
2475 sbi->s_sectors_written_start) >> 1);
2478 static ssize_t lifetime_write_kbytes_show(struct ext4_attr *a,
2479 struct ext4_sb_info *sbi, char *buf)
2481 struct super_block *sb = sbi->s_buddy_cache->i_sb;
2483 if (!sb->s_bdev->bd_part)
2484 return snprintf(buf, PAGE_SIZE, "0\n");
2485 return snprintf(buf, PAGE_SIZE, "%llu\n",
2486 (unsigned long long)(sbi->s_kbytes_written +
2487 ((part_stat_read(sb->s_bdev->bd_part, sectors[1]) -
2488 EXT4_SB(sb)->s_sectors_written_start) >> 1)));
2491 static ssize_t inode_readahead_blks_store(struct ext4_attr *a,
2492 struct ext4_sb_info *sbi,
2493 const char *buf, size_t count)
2498 ret = kstrtoul(skip_spaces(buf), 0, &t);
2502 if (t && (!is_power_of_2(t) || t > 0x40000000))
2505 sbi->s_inode_readahead_blks = t;
2509 static ssize_t sbi_ui_show(struct ext4_attr *a,
2510 struct ext4_sb_info *sbi, char *buf)
2512 unsigned int *ui = (unsigned int *) (((char *) sbi) + a->u.offset);
2514 return snprintf(buf, PAGE_SIZE, "%u\n", *ui);
2517 static ssize_t sbi_ui_store(struct ext4_attr *a,
2518 struct ext4_sb_info *sbi,
2519 const char *buf, size_t count)
2521 unsigned int *ui = (unsigned int *) (((char *) sbi) + a->u.offset);
2525 ret = kstrtoul(skip_spaces(buf), 0, &t);
2532 static ssize_t reserved_clusters_show(struct ext4_attr *a,
2533 struct ext4_sb_info *sbi, char *buf)
2535 return snprintf(buf, PAGE_SIZE, "%llu\n",
2536 (unsigned long long) atomic64_read(&sbi->s_resv_clusters));
2539 static ssize_t reserved_clusters_store(struct ext4_attr *a,
2540 struct ext4_sb_info *sbi,
2541 const char *buf, size_t count)
2543 unsigned long long val;
2546 if (parse_strtoull(buf, -1ULL, &val))
2548 ret = ext4_reserve_clusters(sbi, val);
2550 return ret ? ret : count;
2553 static ssize_t trigger_test_error(struct ext4_attr *a,
2554 struct ext4_sb_info *sbi,
2555 const char *buf, size_t count)
2559 if (!capable(CAP_SYS_ADMIN))
2562 if (len && buf[len-1] == '\n')
2566 ext4_error(sbi->s_sb, "%.*s", len, buf);
2570 static ssize_t sbi_deprecated_show(struct ext4_attr *a,
2571 struct ext4_sb_info *sbi, char *buf)
2573 return snprintf(buf, PAGE_SIZE, "%d\n", a->u.deprecated_val);
2576 #define EXT4_ATTR_OFFSET(_name,_mode,_show,_store,_elname) \
2577 static struct ext4_attr ext4_attr_##_name = { \
2578 .attr = {.name = __stringify(_name), .mode = _mode }, \
2582 .offset = offsetof(struct ext4_sb_info, _elname),\
2585 #define EXT4_ATTR(name, mode, show, store) \
2586 static struct ext4_attr ext4_attr_##name = __ATTR(name, mode, show, store)
2588 #define EXT4_INFO_ATTR(name) EXT4_ATTR(name, 0444, NULL, NULL)
2589 #define EXT4_RO_ATTR(name) EXT4_ATTR(name, 0444, name##_show, NULL)
2590 #define EXT4_RW_ATTR(name) EXT4_ATTR(name, 0644, name##_show, name##_store)
2591 #define EXT4_RW_ATTR_SBI_UI(name, elname) \
2592 EXT4_ATTR_OFFSET(name, 0644, sbi_ui_show, sbi_ui_store, elname)
2593 #define ATTR_LIST(name) &ext4_attr_##name.attr
2594 #define EXT4_DEPRECATED_ATTR(_name, _val) \
2595 static struct ext4_attr ext4_attr_##_name = { \
2596 .attr = {.name = __stringify(_name), .mode = 0444 }, \
2597 .show = sbi_deprecated_show, \
2599 .deprecated_val = _val, \
2603 EXT4_RO_ATTR(delayed_allocation_blocks);
2604 EXT4_RO_ATTR(session_write_kbytes);
2605 EXT4_RO_ATTR(lifetime_write_kbytes);
2606 EXT4_RW_ATTR(reserved_clusters);
2607 EXT4_ATTR_OFFSET(inode_readahead_blks, 0644, sbi_ui_show,
2608 inode_readahead_blks_store, s_inode_readahead_blks);
2609 EXT4_RW_ATTR_SBI_UI(inode_goal, s_inode_goal);
2610 EXT4_RW_ATTR_SBI_UI(mb_stats, s_mb_stats);
2611 EXT4_RW_ATTR_SBI_UI(mb_max_to_scan, s_mb_max_to_scan);
2612 EXT4_RW_ATTR_SBI_UI(mb_min_to_scan, s_mb_min_to_scan);
2613 EXT4_RW_ATTR_SBI_UI(mb_order2_req, s_mb_order2_reqs);
2614 EXT4_RW_ATTR_SBI_UI(mb_stream_req, s_mb_stream_request);
2615 EXT4_RW_ATTR_SBI_UI(mb_group_prealloc, s_mb_group_prealloc);
2616 EXT4_DEPRECATED_ATTR(max_writeback_mb_bump, 128);
2617 EXT4_RW_ATTR_SBI_UI(extent_max_zeroout_kb, s_extent_max_zeroout_kb);
2618 EXT4_ATTR(trigger_fs_error, 0200, NULL, trigger_test_error);
2619 EXT4_RW_ATTR_SBI_UI(err_ratelimit_interval_ms, s_err_ratelimit_state.interval);
2620 EXT4_RW_ATTR_SBI_UI(err_ratelimit_burst, s_err_ratelimit_state.burst);
2621 EXT4_RW_ATTR_SBI_UI(warning_ratelimit_interval_ms, s_warning_ratelimit_state.interval);
2622 EXT4_RW_ATTR_SBI_UI(warning_ratelimit_burst, s_warning_ratelimit_state.burst);
2623 EXT4_RW_ATTR_SBI_UI(msg_ratelimit_interval_ms, s_msg_ratelimit_state.interval);
2624 EXT4_RW_ATTR_SBI_UI(msg_ratelimit_burst, s_msg_ratelimit_state.burst);
2626 static struct attribute *ext4_attrs[] = {
2627 ATTR_LIST(delayed_allocation_blocks),
2628 ATTR_LIST(session_write_kbytes),
2629 ATTR_LIST(lifetime_write_kbytes),
2630 ATTR_LIST(reserved_clusters),
2631 ATTR_LIST(inode_readahead_blks),
2632 ATTR_LIST(inode_goal),
2633 ATTR_LIST(mb_stats),
2634 ATTR_LIST(mb_max_to_scan),
2635 ATTR_LIST(mb_min_to_scan),
2636 ATTR_LIST(mb_order2_req),
2637 ATTR_LIST(mb_stream_req),
2638 ATTR_LIST(mb_group_prealloc),
2639 ATTR_LIST(max_writeback_mb_bump),
2640 ATTR_LIST(extent_max_zeroout_kb),
2641 ATTR_LIST(trigger_fs_error),
2642 ATTR_LIST(err_ratelimit_interval_ms),
2643 ATTR_LIST(err_ratelimit_burst),
2644 ATTR_LIST(warning_ratelimit_interval_ms),
2645 ATTR_LIST(warning_ratelimit_burst),
2646 ATTR_LIST(msg_ratelimit_interval_ms),
2647 ATTR_LIST(msg_ratelimit_burst),
2651 /* Features this copy of ext4 supports */
2652 EXT4_INFO_ATTR(lazy_itable_init);
2653 EXT4_INFO_ATTR(batched_discard);
2654 EXT4_INFO_ATTR(meta_bg_resize);
2656 static struct attribute *ext4_feat_attrs[] = {
2657 ATTR_LIST(lazy_itable_init),
2658 ATTR_LIST(batched_discard),
2659 ATTR_LIST(meta_bg_resize),
2663 static ssize_t ext4_attr_show(struct kobject *kobj,
2664 struct attribute *attr, char *buf)
2666 struct ext4_sb_info *sbi = container_of(kobj, struct ext4_sb_info,
2668 struct ext4_attr *a = container_of(attr, struct ext4_attr, attr);
2670 return a->show ? a->show(a, sbi, buf) : 0;
2673 static ssize_t ext4_attr_store(struct kobject *kobj,
2674 struct attribute *attr,
2675 const char *buf, size_t len)
2677 struct ext4_sb_info *sbi = container_of(kobj, struct ext4_sb_info,
2679 struct ext4_attr *a = container_of(attr, struct ext4_attr, attr);
2681 return a->store ? a->store(a, sbi, buf, len) : 0;
2684 static void ext4_sb_release(struct kobject *kobj)
2686 struct ext4_sb_info *sbi = container_of(kobj, struct ext4_sb_info,
2688 complete(&sbi->s_kobj_unregister);
2691 static const struct sysfs_ops ext4_attr_ops = {
2692 .show = ext4_attr_show,
2693 .store = ext4_attr_store,
2696 static struct kobj_type ext4_ktype = {
2697 .default_attrs = ext4_attrs,
2698 .sysfs_ops = &ext4_attr_ops,
2699 .release = ext4_sb_release,
2702 static void ext4_feat_release(struct kobject *kobj)
2704 complete(&ext4_feat->f_kobj_unregister);
2707 static struct kobj_type ext4_feat_ktype = {
2708 .default_attrs = ext4_feat_attrs,
2709 .sysfs_ops = &ext4_attr_ops,
2710 .release = ext4_feat_release,
2714 * Check whether this filesystem can be mounted based on
2715 * the features present and the RDONLY/RDWR mount requested.
2716 * Returns 1 if this filesystem can be mounted as requested,
2717 * 0 if it cannot be.
2719 static int ext4_feature_set_ok(struct super_block *sb, int readonly)
2721 if (EXT4_HAS_INCOMPAT_FEATURE(sb, ~EXT4_FEATURE_INCOMPAT_SUPP)) {
2722 ext4_msg(sb, KERN_ERR,
2723 "Couldn't mount because of "
2724 "unsupported optional features (%x)",
2725 (le32_to_cpu(EXT4_SB(sb)->s_es->s_feature_incompat) &
2726 ~EXT4_FEATURE_INCOMPAT_SUPP));
2733 /* Check that feature set is OK for a read-write mount */
2734 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, ~EXT4_FEATURE_RO_COMPAT_SUPP)) {
2735 ext4_msg(sb, KERN_ERR, "couldn't mount RDWR because of "
2736 "unsupported optional features (%x)",
2737 (le32_to_cpu(EXT4_SB(sb)->s_es->s_feature_ro_compat) &
2738 ~EXT4_FEATURE_RO_COMPAT_SUPP));
2742 * Large file size enabled file system can only be mounted
2743 * read-write on 32-bit systems if kernel is built with CONFIG_LBDAF
2745 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_HUGE_FILE)) {
2746 if (sizeof(blkcnt_t) < sizeof(u64)) {
2747 ext4_msg(sb, KERN_ERR, "Filesystem with huge files "
2748 "cannot be mounted RDWR without "
2753 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_BIGALLOC) &&
2754 !EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_EXTENTS)) {
2755 ext4_msg(sb, KERN_ERR,
2756 "Can't support bigalloc feature without "
2757 "extents feature\n");
2761 #ifndef CONFIG_QUOTA
2762 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_QUOTA) &&
2764 ext4_msg(sb, KERN_ERR,
2765 "Filesystem with quota feature cannot be mounted RDWR "
2766 "without CONFIG_QUOTA");
2769 #endif /* CONFIG_QUOTA */
2774 * This function is called once a day if we have errors logged
2775 * on the file system
2777 static void print_daily_error_info(unsigned long arg)
2779 struct super_block *sb = (struct super_block *) arg;
2780 struct ext4_sb_info *sbi;
2781 struct ext4_super_block *es;
2786 if (es->s_error_count)
2787 /* fsck newer than v1.41.13 is needed to clean this condition. */
2788 ext4_msg(sb, KERN_NOTICE, "error count since last fsck: %u",
2789 le32_to_cpu(es->s_error_count));
2790 if (es->s_first_error_time) {
2791 printk(KERN_NOTICE "EXT4-fs (%s): initial error at time %u: %.*s:%d",
2792 sb->s_id, le32_to_cpu(es->s_first_error_time),
2793 (int) sizeof(es->s_first_error_func),
2794 es->s_first_error_func,
2795 le32_to_cpu(es->s_first_error_line));
2796 if (es->s_first_error_ino)
2797 printk(": inode %u",
2798 le32_to_cpu(es->s_first_error_ino));
2799 if (es->s_first_error_block)
2800 printk(": block %llu", (unsigned long long)
2801 le64_to_cpu(es->s_first_error_block));
2804 if (es->s_last_error_time) {
2805 printk(KERN_NOTICE "EXT4-fs (%s): last error at time %u: %.*s:%d",
2806 sb->s_id, le32_to_cpu(es->s_last_error_time),
2807 (int) sizeof(es->s_last_error_func),
2808 es->s_last_error_func,
2809 le32_to_cpu(es->s_last_error_line));
2810 if (es->s_last_error_ino)
2811 printk(": inode %u",
2812 le32_to_cpu(es->s_last_error_ino));
2813 if (es->s_last_error_block)
2814 printk(": block %llu", (unsigned long long)
2815 le64_to_cpu(es->s_last_error_block));
2818 mod_timer(&sbi->s_err_report, jiffies + 24*60*60*HZ); /* Once a day */
2821 /* Find next suitable group and run ext4_init_inode_table */
2822 static int ext4_run_li_request(struct ext4_li_request *elr)
2824 struct ext4_group_desc *gdp = NULL;
2825 ext4_group_t group, ngroups;
2826 struct super_block *sb;
2827 unsigned long timeout = 0;
2831 ngroups = EXT4_SB(sb)->s_groups_count;
2834 for (group = elr->lr_next_group; group < ngroups; group++) {
2835 gdp = ext4_get_group_desc(sb, group, NULL);
2841 if (!(gdp->bg_flags & cpu_to_le16(EXT4_BG_INODE_ZEROED)))
2845 if (group >= ngroups)
2850 ret = ext4_init_inode_table(sb, group,
2851 elr->lr_timeout ? 0 : 1);
2852 if (elr->lr_timeout == 0) {
2853 timeout = (jiffies - timeout) *
2854 elr->lr_sbi->s_li_wait_mult;
2855 elr->lr_timeout = timeout;
2857 elr->lr_next_sched = jiffies + elr->lr_timeout;
2858 elr->lr_next_group = group + 1;
2866 * Remove lr_request from the list_request and free the
2867 * request structure. Should be called with li_list_mtx held
2869 static void ext4_remove_li_request(struct ext4_li_request *elr)
2871 struct ext4_sb_info *sbi;
2878 list_del(&elr->lr_request);
2879 sbi->s_li_request = NULL;
2883 static void ext4_unregister_li_request(struct super_block *sb)
2885 mutex_lock(&ext4_li_mtx);
2886 if (!ext4_li_info) {
2887 mutex_unlock(&ext4_li_mtx);
2891 mutex_lock(&ext4_li_info->li_list_mtx);
2892 ext4_remove_li_request(EXT4_SB(sb)->s_li_request);
2893 mutex_unlock(&ext4_li_info->li_list_mtx);
2894 mutex_unlock(&ext4_li_mtx);
2897 static struct task_struct *ext4_lazyinit_task;
2900 * This is the function where ext4lazyinit thread lives. It walks
2901 * through the request list searching for next scheduled filesystem.
2902 * When such a fs is found, run the lazy initialization request
2903 * (ext4_rn_li_request) and keep track of the time spend in this
2904 * function. Based on that time we compute next schedule time of
2905 * the request. When walking through the list is complete, compute
2906 * next waking time and put itself into sleep.
2908 static int ext4_lazyinit_thread(void *arg)
2910 struct ext4_lazy_init *eli = (struct ext4_lazy_init *)arg;
2911 struct list_head *pos, *n;
2912 struct ext4_li_request *elr;
2913 unsigned long next_wakeup, cur;
2915 BUG_ON(NULL == eli);
2919 next_wakeup = MAX_JIFFY_OFFSET;
2921 mutex_lock(&eli->li_list_mtx);
2922 if (list_empty(&eli->li_request_list)) {
2923 mutex_unlock(&eli->li_list_mtx);
2927 list_for_each_safe(pos, n, &eli->li_request_list) {
2928 elr = list_entry(pos, struct ext4_li_request,
2931 if (time_after_eq(jiffies, elr->lr_next_sched)) {
2932 if (ext4_run_li_request(elr) != 0) {
2933 /* error, remove the lazy_init job */
2934 ext4_remove_li_request(elr);
2939 if (time_before(elr->lr_next_sched, next_wakeup))
2940 next_wakeup = elr->lr_next_sched;
2942 mutex_unlock(&eli->li_list_mtx);
2947 if ((time_after_eq(cur, next_wakeup)) ||
2948 (MAX_JIFFY_OFFSET == next_wakeup)) {
2953 schedule_timeout_interruptible(next_wakeup - cur);
2955 if (kthread_should_stop()) {
2956 ext4_clear_request_list();
2963 * It looks like the request list is empty, but we need
2964 * to check it under the li_list_mtx lock, to prevent any
2965 * additions into it, and of course we should lock ext4_li_mtx
2966 * to atomically free the list and ext4_li_info, because at
2967 * this point another ext4 filesystem could be registering
2970 mutex_lock(&ext4_li_mtx);
2971 mutex_lock(&eli->li_list_mtx);
2972 if (!list_empty(&eli->li_request_list)) {
2973 mutex_unlock(&eli->li_list_mtx);
2974 mutex_unlock(&ext4_li_mtx);
2977 mutex_unlock(&eli->li_list_mtx);
2978 kfree(ext4_li_info);
2979 ext4_li_info = NULL;
2980 mutex_unlock(&ext4_li_mtx);
2985 static void ext4_clear_request_list(void)
2987 struct list_head *pos, *n;
2988 struct ext4_li_request *elr;
2990 mutex_lock(&ext4_li_info->li_list_mtx);
2991 list_for_each_safe(pos, n, &ext4_li_info->li_request_list) {
2992 elr = list_entry(pos, struct ext4_li_request,
2994 ext4_remove_li_request(elr);
2996 mutex_unlock(&ext4_li_info->li_list_mtx);
2999 static int ext4_run_lazyinit_thread(void)
3001 ext4_lazyinit_task = kthread_run(ext4_lazyinit_thread,
3002 ext4_li_info, "ext4lazyinit");
3003 if (IS_ERR(ext4_lazyinit_task)) {
3004 int err = PTR_ERR(ext4_lazyinit_task);
3005 ext4_clear_request_list();
3006 kfree(ext4_li_info);
3007 ext4_li_info = NULL;
3008 printk(KERN_CRIT "EXT4-fs: error %d creating inode table "
3009 "initialization thread\n",
3013 ext4_li_info->li_state |= EXT4_LAZYINIT_RUNNING;
3018 * Check whether it make sense to run itable init. thread or not.
3019 * If there is at least one uninitialized inode table, return
3020 * corresponding group number, else the loop goes through all
3021 * groups and return total number of groups.
3023 static ext4_group_t ext4_has_uninit_itable(struct super_block *sb)
3025 ext4_group_t group, ngroups = EXT4_SB(sb)->s_groups_count;
3026 struct ext4_group_desc *gdp = NULL;
3028 for (group = 0; group < ngroups; group++) {
3029 gdp = ext4_get_group_desc(sb, group, NULL);
3033 if (!(gdp->bg_flags & cpu_to_le16(EXT4_BG_INODE_ZEROED)))
3040 static int ext4_li_info_new(void)
3042 struct ext4_lazy_init *eli = NULL;
3044 eli = kzalloc(sizeof(*eli), GFP_KERNEL);
3048 INIT_LIST_HEAD(&eli->li_request_list);
3049 mutex_init(&eli->li_list_mtx);
3051 eli->li_state |= EXT4_LAZYINIT_QUIT;
3058 static struct ext4_li_request *ext4_li_request_new(struct super_block *sb,
3061 struct ext4_sb_info *sbi = EXT4_SB(sb);
3062 struct ext4_li_request *elr;
3064 elr = kzalloc(sizeof(*elr), GFP_KERNEL);
3070 elr->lr_next_group = start;
3073 * Randomize first schedule time of the request to
3074 * spread the inode table initialization requests
3077 elr->lr_next_sched = jiffies + (prandom_u32() %
3078 (EXT4_DEF_LI_MAX_START_DELAY * HZ));
3082 int ext4_register_li_request(struct super_block *sb,
3083 ext4_group_t first_not_zeroed)
3085 struct ext4_sb_info *sbi = EXT4_SB(sb);
3086 struct ext4_li_request *elr = NULL;
3087 ext4_group_t ngroups = EXT4_SB(sb)->s_groups_count;
3090 mutex_lock(&ext4_li_mtx);
3091 if (sbi->s_li_request != NULL) {
3093 * Reset timeout so it can be computed again, because
3094 * s_li_wait_mult might have changed.
3096 sbi->s_li_request->lr_timeout = 0;
3100 if (first_not_zeroed == ngroups ||
3101 (sb->s_flags & MS_RDONLY) ||
3102 !test_opt(sb, INIT_INODE_TABLE))
3105 elr = ext4_li_request_new(sb, first_not_zeroed);
3111 if (NULL == ext4_li_info) {
3112 ret = ext4_li_info_new();
3117 mutex_lock(&ext4_li_info->li_list_mtx);
3118 list_add(&elr->lr_request, &ext4_li_info->li_request_list);
3119 mutex_unlock(&ext4_li_info->li_list_mtx);
3121 sbi->s_li_request = elr;
3123 * set elr to NULL here since it has been inserted to
3124 * the request_list and the removal and free of it is
3125 * handled by ext4_clear_request_list from now on.
3129 if (!(ext4_li_info->li_state & EXT4_LAZYINIT_RUNNING)) {
3130 ret = ext4_run_lazyinit_thread();
3135 mutex_unlock(&ext4_li_mtx);
3142 * We do not need to lock anything since this is called on
3145 static void ext4_destroy_lazyinit_thread(void)
3148 * If thread exited earlier
3149 * there's nothing to be done.
3151 if (!ext4_li_info || !ext4_lazyinit_task)
3154 kthread_stop(ext4_lazyinit_task);
3157 static int set_journal_csum_feature_set(struct super_block *sb)
3160 int compat, incompat;
3161 struct ext4_sb_info *sbi = EXT4_SB(sb);
3163 if (EXT4_HAS_RO_COMPAT_FEATURE(sb,
3164 EXT4_FEATURE_RO_COMPAT_METADATA_CSUM)) {
3165 /* journal checksum v3 */
3167 incompat = JBD2_FEATURE_INCOMPAT_CSUM_V3;
3169 /* journal checksum v1 */
3170 compat = JBD2_FEATURE_COMPAT_CHECKSUM;
3174 if (test_opt(sb, JOURNAL_ASYNC_COMMIT)) {
3175 ret = jbd2_journal_set_features(sbi->s_journal,
3177 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT |
3179 } else if (test_opt(sb, JOURNAL_CHECKSUM)) {
3180 ret = jbd2_journal_set_features(sbi->s_journal,
3183 jbd2_journal_clear_features(sbi->s_journal, 0, 0,
3184 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT);
3186 jbd2_journal_clear_features(sbi->s_journal,
3187 JBD2_FEATURE_COMPAT_CHECKSUM, 0,
3188 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT |
3189 JBD2_FEATURE_INCOMPAT_CSUM_V3 |
3190 JBD2_FEATURE_INCOMPAT_CSUM_V2);
3197 * Note: calculating the overhead so we can be compatible with
3198 * historical BSD practice is quite difficult in the face of
3199 * clusters/bigalloc. This is because multiple metadata blocks from
3200 * different block group can end up in the same allocation cluster.
3201 * Calculating the exact overhead in the face of clustered allocation
3202 * requires either O(all block bitmaps) in memory or O(number of block
3203 * groups**2) in time. We will still calculate the superblock for
3204 * older file systems --- and if we come across with a bigalloc file
3205 * system with zero in s_overhead_clusters the estimate will be close to
3206 * correct especially for very large cluster sizes --- but for newer
3207 * file systems, it's better to calculate this figure once at mkfs
3208 * time, and store it in the superblock. If the superblock value is
3209 * present (even for non-bigalloc file systems), we will use it.
3211 static int count_overhead(struct super_block *sb, ext4_group_t grp,
3214 struct ext4_sb_info *sbi = EXT4_SB(sb);
3215 struct ext4_group_desc *gdp;
3216 ext4_fsblk_t first_block, last_block, b;
3217 ext4_group_t i, ngroups = ext4_get_groups_count(sb);
3218 int s, j, count = 0;
3220 if (!EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_BIGALLOC))
3221 return (ext4_bg_has_super(sb, grp) + ext4_bg_num_gdb(sb, grp) +
3222 sbi->s_itb_per_group + 2);
3224 first_block = le32_to_cpu(sbi->s_es->s_first_data_block) +
3225 (grp * EXT4_BLOCKS_PER_GROUP(sb));
3226 last_block = first_block + EXT4_BLOCKS_PER_GROUP(sb) - 1;
3227 for (i = 0; i < ngroups; i++) {
3228 gdp = ext4_get_group_desc(sb, i, NULL);
3229 b = ext4_block_bitmap(sb, gdp);
3230 if (b >= first_block && b <= last_block) {
3231 ext4_set_bit(EXT4_B2C(sbi, b - first_block), buf);
3234 b = ext4_inode_bitmap(sb, gdp);
3235 if (b >= first_block && b <= last_block) {
3236 ext4_set_bit(EXT4_B2C(sbi, b - first_block), buf);
3239 b = ext4_inode_table(sb, gdp);
3240 if (b >= first_block && b + sbi->s_itb_per_group <= last_block)
3241 for (j = 0; j < sbi->s_itb_per_group; j++, b++) {
3242 int c = EXT4_B2C(sbi, b - first_block);
3243 ext4_set_bit(c, buf);
3249 if (ext4_bg_has_super(sb, grp)) {
3250 ext4_set_bit(s++, buf);
3253 for (j = ext4_bg_num_gdb(sb, grp); j > 0; j--) {
3254 ext4_set_bit(EXT4_B2C(sbi, s++), buf);
3260 return EXT4_CLUSTERS_PER_GROUP(sb) -
3261 ext4_count_free(buf, EXT4_CLUSTERS_PER_GROUP(sb) / 8);
3265 * Compute the overhead and stash it in sbi->s_overhead
3267 int ext4_calculate_overhead(struct super_block *sb)
3269 struct ext4_sb_info *sbi = EXT4_SB(sb);
3270 struct ext4_super_block *es = sbi->s_es;
3271 ext4_group_t i, ngroups = ext4_get_groups_count(sb);
3272 ext4_fsblk_t overhead = 0;
3273 char *buf = (char *) get_zeroed_page(GFP_KERNEL);
3279 * Compute the overhead (FS structures). This is constant
3280 * for a given filesystem unless the number of block groups
3281 * changes so we cache the previous value until it does.
3285 * All of the blocks before first_data_block are overhead
3287 overhead = EXT4_B2C(sbi, le32_to_cpu(es->s_first_data_block));
3290 * Add the overhead found in each block group
3292 for (i = 0; i < ngroups; i++) {
3295 blks = count_overhead(sb, i, buf);
3298 memset(buf, 0, PAGE_SIZE);
3301 /* Add the journal blocks as well */
3303 overhead += EXT4_NUM_B2C(sbi, sbi->s_journal->j_maxlen);
3305 sbi->s_overhead = overhead;
3307 free_page((unsigned long) buf);
3312 static ext4_fsblk_t ext4_calculate_resv_clusters(struct super_block *sb)
3314 ext4_fsblk_t resv_clusters;
3317 * There's no need to reserve anything when we aren't using extents.
3318 * The space estimates are exact, there are no unwritten extents,
3319 * hole punching doesn't need new metadata... This is needed especially
3320 * to keep ext2/3 backward compatibility.
3322 if (!EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_EXTENTS))
3325 * By default we reserve 2% or 4096 clusters, whichever is smaller.
3326 * This should cover the situations where we can not afford to run
3327 * out of space like for example punch hole, or converting
3328 * uninitialized extents in delalloc path. In most cases such
3329 * allocation would require 1, or 2 blocks, higher numbers are
3332 resv_clusters = ext4_blocks_count(EXT4_SB(sb)->s_es) >>
3333 EXT4_SB(sb)->s_cluster_bits;
3335 do_div(resv_clusters, 50);
3336 resv_clusters = min_t(ext4_fsblk_t, resv_clusters, 4096);
3338 return resv_clusters;
3342 static int ext4_reserve_clusters(struct ext4_sb_info *sbi, ext4_fsblk_t count)
3344 ext4_fsblk_t clusters = ext4_blocks_count(sbi->s_es) >>
3345 sbi->s_cluster_bits;
3347 if (count >= clusters)
3350 atomic64_set(&sbi->s_resv_clusters, count);
3354 static int ext4_fill_super(struct super_block *sb, void *data, int silent)
3356 char *orig_data = kstrdup(data, GFP_KERNEL);
3357 struct buffer_head *bh;
3358 struct ext4_super_block *es = NULL;
3359 struct ext4_sb_info *sbi;
3361 ext4_fsblk_t sb_block = get_sb_block(&data);
3362 ext4_fsblk_t logical_sb_block;
3363 unsigned long offset = 0;
3364 unsigned long journal_devnum = 0;
3365 unsigned long def_mount_opts;
3370 int blocksize, clustersize;
3371 unsigned int db_count;
3373 int needs_recovery, has_huge_files, has_bigalloc;
3376 unsigned int journal_ioprio = DEFAULT_JOURNAL_IOPRIO;
3377 ext4_group_t first_not_zeroed;
3379 sbi = kzalloc(sizeof(*sbi), GFP_KERNEL);
3383 sbi->s_blockgroup_lock =
3384 kzalloc(sizeof(struct blockgroup_lock), GFP_KERNEL);
3385 if (!sbi->s_blockgroup_lock) {
3389 sb->s_fs_info = sbi;
3391 sbi->s_inode_readahead_blks = EXT4_DEF_INODE_READAHEAD_BLKS;
3392 sbi->s_sb_block = sb_block;
3393 if (sb->s_bdev->bd_part)
3394 sbi->s_sectors_written_start =
3395 part_stat_read(sb->s_bdev->bd_part, sectors[1]);
3397 /* Cleanup superblock name */
3398 for (cp = sb->s_id; (cp = strchr(cp, '/'));)
3401 /* -EINVAL is default */
3403 blocksize = sb_min_blocksize(sb, EXT4_MIN_BLOCK_SIZE);
3405 ext4_msg(sb, KERN_ERR, "unable to set blocksize");
3410 * The ext4 superblock will not be buffer aligned for other than 1kB
3411 * block sizes. We need to calculate the offset from buffer start.
3413 if (blocksize != EXT4_MIN_BLOCK_SIZE) {
3414 logical_sb_block = sb_block * EXT4_MIN_BLOCK_SIZE;
3415 offset = do_div(logical_sb_block, blocksize);
3417 logical_sb_block = sb_block;
3420 if (!(bh = sb_bread(sb, logical_sb_block))) {
3421 ext4_msg(sb, KERN_ERR, "unable to read superblock");
3425 * Note: s_es must be initialized as soon as possible because
3426 * some ext4 macro-instructions depend on its value
3428 es = (struct ext4_super_block *) (bh->b_data + offset);
3430 sb->s_magic = le16_to_cpu(es->s_magic);
3431 if (sb->s_magic != EXT4_SUPER_MAGIC)
3433 sbi->s_kbytes_written = le64_to_cpu(es->s_kbytes_written);
3435 /* Warn if metadata_csum and gdt_csum are both set. */
3436 if (EXT4_HAS_RO_COMPAT_FEATURE(sb,
3437 EXT4_FEATURE_RO_COMPAT_METADATA_CSUM) &&
3438 EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_GDT_CSUM))
3439 ext4_warning(sb, KERN_INFO "metadata_csum and uninit_bg are "
3440 "redundant flags; please run fsck.");
3442 /* Check for a known checksum algorithm */
3443 if (!ext4_verify_csum_type(sb, es)) {
3444 ext4_msg(sb, KERN_ERR, "VFS: Found ext4 filesystem with "
3445 "unknown checksum algorithm.");
3450 /* Load the checksum driver */
3451 if (EXT4_HAS_RO_COMPAT_FEATURE(sb,
3452 EXT4_FEATURE_RO_COMPAT_METADATA_CSUM)) {
3453 sbi->s_chksum_driver = crypto_alloc_shash("crc32c", 0, 0);
3454 if (IS_ERR(sbi->s_chksum_driver)) {
3455 ext4_msg(sb, KERN_ERR, "Cannot load crc32c driver.");
3456 ret = PTR_ERR(sbi->s_chksum_driver);
3457 sbi->s_chksum_driver = NULL;
3462 /* Check superblock checksum */
3463 if (!ext4_superblock_csum_verify(sb, es)) {
3464 ext4_msg(sb, KERN_ERR, "VFS: Found ext4 filesystem with "
3465 "invalid superblock checksum. Run e2fsck?");
3470 /* Precompute checksum seed for all metadata */
3471 if (EXT4_HAS_RO_COMPAT_FEATURE(sb,
3472 EXT4_FEATURE_RO_COMPAT_METADATA_CSUM))
3473 sbi->s_csum_seed = ext4_chksum(sbi, ~0, es->s_uuid,
3474 sizeof(es->s_uuid));
3476 /* Set defaults before we parse the mount options */
3477 def_mount_opts = le32_to_cpu(es->s_default_mount_opts);
3478 set_opt(sb, INIT_INODE_TABLE);
3479 if (def_mount_opts & EXT4_DEFM_DEBUG)
3481 if (def_mount_opts & EXT4_DEFM_BSDGROUPS)
3483 if (def_mount_opts & EXT4_DEFM_UID16)
3484 set_opt(sb, NO_UID32);
3485 /* xattr user namespace & acls are now defaulted on */
3486 set_opt(sb, XATTR_USER);
3487 #ifdef CONFIG_EXT4_FS_POSIX_ACL
3488 set_opt(sb, POSIX_ACL);
3490 if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_DATA)
3491 set_opt(sb, JOURNAL_DATA);
3492 else if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_ORDERED)
3493 set_opt(sb, ORDERED_DATA);
3494 else if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_WBACK)
3495 set_opt(sb, WRITEBACK_DATA);
3497 if (le16_to_cpu(sbi->s_es->s_errors) == EXT4_ERRORS_PANIC)
3498 set_opt(sb, ERRORS_PANIC);
3499 else if (le16_to_cpu(sbi->s_es->s_errors) == EXT4_ERRORS_CONTINUE)
3500 set_opt(sb, ERRORS_CONT);
3502 set_opt(sb, ERRORS_RO);
3503 if (def_mount_opts & EXT4_DEFM_BLOCK_VALIDITY)
3504 set_opt(sb, BLOCK_VALIDITY);
3505 if (def_mount_opts & EXT4_DEFM_DISCARD)
3506 set_opt(sb, DISCARD);
3508 sbi->s_resuid = make_kuid(&init_user_ns, le16_to_cpu(es->s_def_resuid));
3509 sbi->s_resgid = make_kgid(&init_user_ns, le16_to_cpu(es->s_def_resgid));
3510 sbi->s_commit_interval = JBD2_DEFAULT_MAX_COMMIT_AGE * HZ;
3511 sbi->s_min_batch_time = EXT4_DEF_MIN_BATCH_TIME;
3512 sbi->s_max_batch_time = EXT4_DEF_MAX_BATCH_TIME;
3514 if ((def_mount_opts & EXT4_DEFM_NOBARRIER) == 0)
3515 set_opt(sb, BARRIER);
3518 * enable delayed allocation by default
3519 * Use -o nodelalloc to turn it off
3521 if (!IS_EXT3_SB(sb) && !IS_EXT2_SB(sb) &&
3522 ((def_mount_opts & EXT4_DEFM_NODELALLOC) == 0))
3523 set_opt(sb, DELALLOC);
3526 * set default s_li_wait_mult for lazyinit, for the case there is
3527 * no mount option specified.
3529 sbi->s_li_wait_mult = EXT4_DEF_LI_WAIT_MULT;
3531 if (!parse_options((char *) sbi->s_es->s_mount_opts, sb,
3532 &journal_devnum, &journal_ioprio, 0)) {
3533 ext4_msg(sb, KERN_WARNING,
3534 "failed to parse options in superblock: %s",
3535 sbi->s_es->s_mount_opts);
3537 sbi->s_def_mount_opt = sbi->s_mount_opt;
3538 if (!parse_options((char *) data, sb, &journal_devnum,
3539 &journal_ioprio, 0))
3542 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA) {
3543 printk_once(KERN_WARNING "EXT4-fs: Warning: mounting "
3544 "with data=journal disables delayed "
3545 "allocation and O_DIRECT support!\n");
3546 if (test_opt2(sb, EXPLICIT_DELALLOC)) {
3547 ext4_msg(sb, KERN_ERR, "can't mount with "
3548 "both data=journal and delalloc");
3551 if (test_opt(sb, DIOREAD_NOLOCK)) {
3552 ext4_msg(sb, KERN_ERR, "can't mount with "
3553 "both data=journal and dioread_nolock");
3556 if (test_opt(sb, DELALLOC))
3557 clear_opt(sb, DELALLOC);
3560 sb->s_flags = (sb->s_flags & ~MS_POSIXACL) |
3561 (test_opt(sb, POSIX_ACL) ? MS_POSIXACL : 0);
3563 if (le32_to_cpu(es->s_rev_level) == EXT4_GOOD_OLD_REV &&
3564 (EXT4_HAS_COMPAT_FEATURE(sb, ~0U) ||
3565 EXT4_HAS_RO_COMPAT_FEATURE(sb, ~0U) ||
3566 EXT4_HAS_INCOMPAT_FEATURE(sb, ~0U)))
3567 ext4_msg(sb, KERN_WARNING,
3568 "feature flags set on rev 0 fs, "
3569 "running e2fsck is recommended");
3571 if (IS_EXT2_SB(sb)) {
3572 if (ext2_feature_set_ok(sb))
3573 ext4_msg(sb, KERN_INFO, "mounting ext2 file system "
3574 "using the ext4 subsystem");
3576 ext4_msg(sb, KERN_ERR, "couldn't mount as ext2 due "
3577 "to feature incompatibilities");
3582 if (IS_EXT3_SB(sb)) {
3583 if (ext3_feature_set_ok(sb))
3584 ext4_msg(sb, KERN_INFO, "mounting ext3 file system "
3585 "using the ext4 subsystem");
3587 ext4_msg(sb, KERN_ERR, "couldn't mount as ext3 due "
3588 "to feature incompatibilities");
3594 * Check feature flags regardless of the revision level, since we
3595 * previously didn't change the revision level when setting the flags,
3596 * so there is a chance incompat flags are set on a rev 0 filesystem.
3598 if (!ext4_feature_set_ok(sb, (sb->s_flags & MS_RDONLY)))
3601 blocksize = BLOCK_SIZE << le32_to_cpu(es->s_log_block_size);
3602 if (blocksize < EXT4_MIN_BLOCK_SIZE ||
3603 blocksize > EXT4_MAX_BLOCK_SIZE) {
3604 ext4_msg(sb, KERN_ERR,
3605 "Unsupported filesystem blocksize %d", blocksize);
3609 if (sb->s_blocksize != blocksize) {
3610 /* Validate the filesystem blocksize */
3611 if (!sb_set_blocksize(sb, blocksize)) {
3612 ext4_msg(sb, KERN_ERR, "bad block size %d",
3618 logical_sb_block = sb_block * EXT4_MIN_BLOCK_SIZE;
3619 offset = do_div(logical_sb_block, blocksize);
3620 bh = sb_bread(sb, logical_sb_block);
3622 ext4_msg(sb, KERN_ERR,
3623 "Can't read superblock on 2nd try");
3626 es = (struct ext4_super_block *)(bh->b_data + offset);
3628 if (es->s_magic != cpu_to_le16(EXT4_SUPER_MAGIC)) {
3629 ext4_msg(sb, KERN_ERR,
3630 "Magic mismatch, very weird!");
3635 has_huge_files = EXT4_HAS_RO_COMPAT_FEATURE(sb,
3636 EXT4_FEATURE_RO_COMPAT_HUGE_FILE);
3637 sbi->s_bitmap_maxbytes = ext4_max_bitmap_size(sb->s_blocksize_bits,
3639 sb->s_maxbytes = ext4_max_size(sb->s_blocksize_bits, has_huge_files);
3641 if (le32_to_cpu(es->s_rev_level) == EXT4_GOOD_OLD_REV) {
3642 sbi->s_inode_size = EXT4_GOOD_OLD_INODE_SIZE;
3643 sbi->s_first_ino = EXT4_GOOD_OLD_FIRST_INO;
3645 sbi->s_inode_size = le16_to_cpu(es->s_inode_size);
3646 sbi->s_first_ino = le32_to_cpu(es->s_first_ino);
3647 if ((sbi->s_inode_size < EXT4_GOOD_OLD_INODE_SIZE) ||
3648 (!is_power_of_2(sbi->s_inode_size)) ||
3649 (sbi->s_inode_size > blocksize)) {
3650 ext4_msg(sb, KERN_ERR,
3651 "unsupported inode size: %d",
3655 if (sbi->s_inode_size > EXT4_GOOD_OLD_INODE_SIZE)
3656 sb->s_time_gran = 1 << (EXT4_EPOCH_BITS - 2);
3659 sbi->s_desc_size = le16_to_cpu(es->s_desc_size);
3660 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_64BIT)) {
3661 if (sbi->s_desc_size < EXT4_MIN_DESC_SIZE_64BIT ||
3662 sbi->s_desc_size > EXT4_MAX_DESC_SIZE ||
3663 !is_power_of_2(sbi->s_desc_size)) {
3664 ext4_msg(sb, KERN_ERR,
3665 "unsupported descriptor size %lu",
3670 sbi->s_desc_size = EXT4_MIN_DESC_SIZE;
3672 sbi->s_blocks_per_group = le32_to_cpu(es->s_blocks_per_group);
3673 sbi->s_inodes_per_group = le32_to_cpu(es->s_inodes_per_group);
3674 if (EXT4_INODE_SIZE(sb) == 0 || EXT4_INODES_PER_GROUP(sb) == 0)
3677 sbi->s_inodes_per_block = blocksize / EXT4_INODE_SIZE(sb);
3678 if (sbi->s_inodes_per_block == 0)
3680 sbi->s_itb_per_group = sbi->s_inodes_per_group /
3681 sbi->s_inodes_per_block;
3682 sbi->s_desc_per_block = blocksize / EXT4_DESC_SIZE(sb);
3684 sbi->s_mount_state = le16_to_cpu(es->s_state);
3685 sbi->s_addr_per_block_bits = ilog2(EXT4_ADDR_PER_BLOCK(sb));
3686 sbi->s_desc_per_block_bits = ilog2(EXT4_DESC_PER_BLOCK(sb));
3688 for (i = 0; i < 4; i++)
3689 sbi->s_hash_seed[i] = le32_to_cpu(es->s_hash_seed[i]);
3690 sbi->s_def_hash_version = es->s_def_hash_version;
3691 if (EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_DIR_INDEX)) {
3692 i = le32_to_cpu(es->s_flags);
3693 if (i & EXT2_FLAGS_UNSIGNED_HASH)
3694 sbi->s_hash_unsigned = 3;
3695 else if ((i & EXT2_FLAGS_SIGNED_HASH) == 0) {
3696 #ifdef __CHAR_UNSIGNED__
3697 if (!(sb->s_flags & MS_RDONLY))
3699 cpu_to_le32(EXT2_FLAGS_UNSIGNED_HASH);
3700 sbi->s_hash_unsigned = 3;
3702 if (!(sb->s_flags & MS_RDONLY))
3704 cpu_to_le32(EXT2_FLAGS_SIGNED_HASH);
3709 /* Handle clustersize */
3710 clustersize = BLOCK_SIZE << le32_to_cpu(es->s_log_cluster_size);
3711 has_bigalloc = EXT4_HAS_RO_COMPAT_FEATURE(sb,
3712 EXT4_FEATURE_RO_COMPAT_BIGALLOC);
3714 if (clustersize < blocksize) {
3715 ext4_msg(sb, KERN_ERR,
3716 "cluster size (%d) smaller than "
3717 "block size (%d)", clustersize, blocksize);
3720 sbi->s_cluster_bits = le32_to_cpu(es->s_log_cluster_size) -
3721 le32_to_cpu(es->s_log_block_size);
3722 sbi->s_clusters_per_group =
3723 le32_to_cpu(es->s_clusters_per_group);
3724 if (sbi->s_clusters_per_group > blocksize * 8) {
3725 ext4_msg(sb, KERN_ERR,
3726 "#clusters per group too big: %lu",
3727 sbi->s_clusters_per_group);
3730 if (sbi->s_blocks_per_group !=
3731 (sbi->s_clusters_per_group * (clustersize / blocksize))) {
3732 ext4_msg(sb, KERN_ERR, "blocks per group (%lu) and "
3733 "clusters per group (%lu) inconsistent",
3734 sbi->s_blocks_per_group,
3735 sbi->s_clusters_per_group);
3739 if (clustersize != blocksize) {
3740 ext4_warning(sb, "fragment/cluster size (%d) != "
3741 "block size (%d)", clustersize,
3743 clustersize = blocksize;
3745 if (sbi->s_blocks_per_group > blocksize * 8) {
3746 ext4_msg(sb, KERN_ERR,
3747 "#blocks per group too big: %lu",
3748 sbi->s_blocks_per_group);
3751 sbi->s_clusters_per_group = sbi->s_blocks_per_group;
3752 sbi->s_cluster_bits = 0;
3754 sbi->s_cluster_ratio = clustersize / blocksize;
3756 if (sbi->s_inodes_per_group > blocksize * 8) {
3757 ext4_msg(sb, KERN_ERR,
3758 "#inodes per group too big: %lu",
3759 sbi->s_inodes_per_group);
3763 /* Do we have standard group size of clustersize * 8 blocks ? */
3764 if (sbi->s_blocks_per_group == clustersize << 3)
3765 set_opt2(sb, STD_GROUP_SIZE);
3768 * Test whether we have more sectors than will fit in sector_t,
3769 * and whether the max offset is addressable by the page cache.
3771 err = generic_check_addressable(sb->s_blocksize_bits,
3772 ext4_blocks_count(es));
3774 ext4_msg(sb, KERN_ERR, "filesystem"
3775 " too large to mount safely on this system");
3776 if (sizeof(sector_t) < 8)
3777 ext4_msg(sb, KERN_WARNING, "CONFIG_LBDAF not enabled");
3781 if (EXT4_BLOCKS_PER_GROUP(sb) == 0)
3784 /* check blocks count against device size */
3785 blocks_count = sb->s_bdev->bd_inode->i_size >> sb->s_blocksize_bits;
3786 if (blocks_count && ext4_blocks_count(es) > blocks_count) {
3787 ext4_msg(sb, KERN_WARNING, "bad geometry: block count %llu "
3788 "exceeds size of device (%llu blocks)",
3789 ext4_blocks_count(es), blocks_count);
3794 * It makes no sense for the first data block to be beyond the end
3795 * of the filesystem.
3797 if (le32_to_cpu(es->s_first_data_block) >= ext4_blocks_count(es)) {
3798 ext4_msg(sb, KERN_WARNING, "bad geometry: first data "
3799 "block %u is beyond end of filesystem (%llu)",
3800 le32_to_cpu(es->s_first_data_block),
3801 ext4_blocks_count(es));
3804 blocks_count = (ext4_blocks_count(es) -
3805 le32_to_cpu(es->s_first_data_block) +
3806 EXT4_BLOCKS_PER_GROUP(sb) - 1);
3807 do_div(blocks_count, EXT4_BLOCKS_PER_GROUP(sb));
3808 if (blocks_count > ((uint64_t)1<<32) - EXT4_DESC_PER_BLOCK(sb)) {
3809 ext4_msg(sb, KERN_WARNING, "groups count too large: %u "
3810 "(block count %llu, first data block %u, "
3811 "blocks per group %lu)", sbi->s_groups_count,
3812 ext4_blocks_count(es),
3813 le32_to_cpu(es->s_first_data_block),
3814 EXT4_BLOCKS_PER_GROUP(sb));
3817 sbi->s_groups_count = blocks_count;
3818 sbi->s_blockfile_groups = min_t(ext4_group_t, sbi->s_groups_count,
3819 (EXT4_MAX_BLOCK_FILE_PHYS / EXT4_BLOCKS_PER_GROUP(sb)));
3820 db_count = (sbi->s_groups_count + EXT4_DESC_PER_BLOCK(sb) - 1) /
3821 EXT4_DESC_PER_BLOCK(sb);
3822 sbi->s_group_desc = ext4_kvmalloc(db_count *
3823 sizeof(struct buffer_head *),
3825 if (sbi->s_group_desc == NULL) {
3826 ext4_msg(sb, KERN_ERR, "not enough memory");
3832 sbi->s_proc = proc_mkdir(sb->s_id, ext4_proc_root);
3835 proc_create_data("options", S_IRUGO, sbi->s_proc,
3836 &ext4_seq_options_fops, sb);
3838 bgl_lock_init(sbi->s_blockgroup_lock);
3840 for (i = 0; i < db_count; i++) {
3841 block = descriptor_loc(sb, logical_sb_block, i);
3842 sbi->s_group_desc[i] = sb_bread(sb, block);
3843 if (!sbi->s_group_desc[i]) {
3844 ext4_msg(sb, KERN_ERR,
3845 "can't read group descriptor %d", i);
3850 if (!ext4_check_descriptors(sb, &first_not_zeroed)) {
3851 ext4_msg(sb, KERN_ERR, "group descriptors corrupted!");
3854 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_FLEX_BG))
3855 if (!ext4_fill_flex_info(sb)) {
3856 ext4_msg(sb, KERN_ERR,
3857 "unable to initialize "
3858 "flex_bg meta info!");
3862 sbi->s_gdb_count = db_count;
3863 get_random_bytes(&sbi->s_next_generation, sizeof(u32));
3864 spin_lock_init(&sbi->s_next_gen_lock);
3866 init_timer(&sbi->s_err_report);
3867 sbi->s_err_report.function = print_daily_error_info;
3868 sbi->s_err_report.data = (unsigned long) sb;
3870 /* Register extent status tree shrinker */
3871 ext4_es_register_shrinker(sbi);
3873 err = percpu_counter_init(&sbi->s_freeclusters_counter,
3874 ext4_count_free_clusters(sb));
3876 err = percpu_counter_init(&sbi->s_freeinodes_counter,
3877 ext4_count_free_inodes(sb));
3880 err = percpu_counter_init(&sbi->s_dirs_counter,
3881 ext4_count_dirs(sb));
3884 err = percpu_counter_init(&sbi->s_dirtyclusters_counter, 0);
3887 err = percpu_counter_init(&sbi->s_extent_cache_cnt, 0);
3890 ext4_msg(sb, KERN_ERR, "insufficient memory");
3894 sbi->s_stripe = ext4_get_stripe_size(sbi);
3895 sbi->s_extent_max_zeroout_kb = 32;
3898 * set up enough so that it can read an inode
3900 if (!test_opt(sb, NOLOAD) &&
3901 EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL))
3902 sb->s_op = &ext4_sops;
3904 sb->s_op = &ext4_nojournal_sops;
3905 sb->s_export_op = &ext4_export_ops;
3906 sb->s_xattr = ext4_xattr_handlers;
3908 sb->dq_op = &ext4_quota_operations;
3909 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_QUOTA))
3910 sb->s_qcop = &ext4_qctl_sysfile_operations;
3912 sb->s_qcop = &ext4_qctl_operations;
3914 memcpy(sb->s_uuid, es->s_uuid, sizeof(es->s_uuid));
3916 INIT_LIST_HEAD(&sbi->s_orphan); /* unlinked but open files */
3917 mutex_init(&sbi->s_orphan_lock);
3921 needs_recovery = (es->s_last_orphan != 0 ||
3922 EXT4_HAS_INCOMPAT_FEATURE(sb,
3923 EXT4_FEATURE_INCOMPAT_RECOVER));
3925 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_MMP) &&
3926 !(sb->s_flags & MS_RDONLY))
3927 if (ext4_multi_mount_protect(sb, le64_to_cpu(es->s_mmp_block)))
3931 * The first inode we look at is the journal inode. Don't try
3932 * root first: it may be modified in the journal!
3934 if (!test_opt(sb, NOLOAD) &&
3935 EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL)) {
3936 if (ext4_load_journal(sb, es, journal_devnum))
3938 } else if (test_opt(sb, NOLOAD) && !(sb->s_flags & MS_RDONLY) &&
3939 EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER)) {
3940 ext4_msg(sb, KERN_ERR, "required journal recovery "
3941 "suppressed and not mounted read-only");
3942 goto failed_mount_wq;
3944 clear_opt(sb, DATA_FLAGS);
3945 sbi->s_journal = NULL;
3950 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_64BIT) &&
3951 !jbd2_journal_set_features(EXT4_SB(sb)->s_journal, 0, 0,
3952 JBD2_FEATURE_INCOMPAT_64BIT)) {
3953 ext4_msg(sb, KERN_ERR, "Failed to set 64-bit journal feature");
3954 goto failed_mount_wq;
3957 if (!set_journal_csum_feature_set(sb)) {
3958 ext4_msg(sb, KERN_ERR, "Failed to set journal checksum "
3960 goto failed_mount_wq;
3963 /* We have now updated the journal if required, so we can
3964 * validate the data journaling mode. */
3965 switch (test_opt(sb, DATA_FLAGS)) {
3967 /* No mode set, assume a default based on the journal
3968 * capabilities: ORDERED_DATA if the journal can
3969 * cope, else JOURNAL_DATA
3971 if (jbd2_journal_check_available_features
3972 (sbi->s_journal, 0, 0, JBD2_FEATURE_INCOMPAT_REVOKE))
3973 set_opt(sb, ORDERED_DATA);
3975 set_opt(sb, JOURNAL_DATA);
3978 case EXT4_MOUNT_ORDERED_DATA:
3979 case EXT4_MOUNT_WRITEBACK_DATA:
3980 if (!jbd2_journal_check_available_features
3981 (sbi->s_journal, 0, 0, JBD2_FEATURE_INCOMPAT_REVOKE)) {
3982 ext4_msg(sb, KERN_ERR, "Journal does not support "
3983 "requested data journaling mode");
3984 goto failed_mount_wq;
3989 set_task_ioprio(sbi->s_journal->j_task, journal_ioprio);
3991 sbi->s_journal->j_commit_callback = ext4_journal_commit_callback;
3994 * The journal may have updated the bg summary counts, so we
3995 * need to update the global counters.
3997 percpu_counter_set(&sbi->s_freeclusters_counter,
3998 ext4_count_free_clusters(sb));
3999 percpu_counter_set(&sbi->s_freeinodes_counter,
4000 ext4_count_free_inodes(sb));
4001 percpu_counter_set(&sbi->s_dirs_counter,
4002 ext4_count_dirs(sb));
4003 percpu_counter_set(&sbi->s_dirtyclusters_counter, 0);
4007 * Get the # of file system overhead blocks from the
4008 * superblock if present.
4010 if (es->s_overhead_clusters)
4011 sbi->s_overhead = le32_to_cpu(es->s_overhead_clusters);
4013 err = ext4_calculate_overhead(sb);
4015 goto failed_mount_wq;
4019 * The maximum number of concurrent works can be high and
4020 * concurrency isn't really necessary. Limit it to 1.
4022 EXT4_SB(sb)->rsv_conversion_wq =
4023 alloc_workqueue("ext4-rsv-conversion", WQ_MEM_RECLAIM | WQ_UNBOUND, 1);
4024 if (!EXT4_SB(sb)->rsv_conversion_wq) {
4025 printk(KERN_ERR "EXT4-fs: failed to create workqueue\n");
4031 * The jbd2_journal_load will have done any necessary log recovery,
4032 * so we can safely mount the rest of the filesystem now.
4035 root = ext4_iget(sb, EXT4_ROOT_INO);
4037 ext4_msg(sb, KERN_ERR, "get root inode failed");
4038 ret = PTR_ERR(root);
4042 if (!S_ISDIR(root->i_mode) || !root->i_blocks || !root->i_size) {
4043 ext4_msg(sb, KERN_ERR, "corrupt root inode, run e2fsck");
4047 sb->s_root = d_make_root(root);
4049 ext4_msg(sb, KERN_ERR, "get root dentry failed");
4054 if (ext4_setup_super(sb, es, sb->s_flags & MS_RDONLY))
4055 sb->s_flags |= MS_RDONLY;
4057 /* determine the minimum size of new large inodes, if present */
4058 if (sbi->s_inode_size > EXT4_GOOD_OLD_INODE_SIZE) {
4059 sbi->s_want_extra_isize = sizeof(struct ext4_inode) -
4060 EXT4_GOOD_OLD_INODE_SIZE;
4061 if (EXT4_HAS_RO_COMPAT_FEATURE(sb,
4062 EXT4_FEATURE_RO_COMPAT_EXTRA_ISIZE)) {
4063 if (sbi->s_want_extra_isize <
4064 le16_to_cpu(es->s_want_extra_isize))
4065 sbi->s_want_extra_isize =
4066 le16_to_cpu(es->s_want_extra_isize);
4067 if (sbi->s_want_extra_isize <
4068 le16_to_cpu(es->s_min_extra_isize))
4069 sbi->s_want_extra_isize =
4070 le16_to_cpu(es->s_min_extra_isize);
4073 /* Check if enough inode space is available */
4074 if (EXT4_GOOD_OLD_INODE_SIZE + sbi->s_want_extra_isize >
4075 sbi->s_inode_size) {
4076 sbi->s_want_extra_isize = sizeof(struct ext4_inode) -
4077 EXT4_GOOD_OLD_INODE_SIZE;
4078 ext4_msg(sb, KERN_INFO, "required extra inode space not"
4082 err = ext4_reserve_clusters(sbi, ext4_calculate_resv_clusters(sb));
4084 ext4_msg(sb, KERN_ERR, "failed to reserve %llu clusters for "
4085 "reserved pool", ext4_calculate_resv_clusters(sb));
4086 goto failed_mount4a;
4089 err = ext4_setup_system_zone(sb);
4091 ext4_msg(sb, KERN_ERR, "failed to initialize system "
4093 goto failed_mount4a;
4097 err = ext4_mb_init(sb);
4099 ext4_msg(sb, KERN_ERR, "failed to initialize mballoc (%d)",
4104 err = ext4_register_li_request(sb, first_not_zeroed);
4108 sbi->s_kobj.kset = ext4_kset;
4109 init_completion(&sbi->s_kobj_unregister);
4110 err = kobject_init_and_add(&sbi->s_kobj, &ext4_ktype, NULL,
4116 /* Enable quota usage during mount. */
4117 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_QUOTA) &&
4118 !(sb->s_flags & MS_RDONLY)) {
4119 err = ext4_enable_quotas(sb);
4123 #endif /* CONFIG_QUOTA */
4125 EXT4_SB(sb)->s_mount_state |= EXT4_ORPHAN_FS;
4126 ext4_orphan_cleanup(sb, es);
4127 EXT4_SB(sb)->s_mount_state &= ~EXT4_ORPHAN_FS;
4128 if (needs_recovery) {
4129 ext4_msg(sb, KERN_INFO, "recovery complete");
4130 ext4_mark_recovery_complete(sb, es);
4132 if (EXT4_SB(sb)->s_journal) {
4133 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA)
4134 descr = " journalled data mode";
4135 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA)
4136 descr = " ordered data mode";
4138 descr = " writeback data mode";
4140 descr = "out journal";
4142 if (test_opt(sb, DISCARD)) {
4143 struct request_queue *q = bdev_get_queue(sb->s_bdev);
4144 if (!blk_queue_discard(q))
4145 ext4_msg(sb, KERN_WARNING,
4146 "mounting with \"discard\" option, but "
4147 "the device does not support discard");
4150 ext4_msg(sb, KERN_INFO, "mounted filesystem with%s. "
4151 "Opts: %s%s%s", descr, sbi->s_es->s_mount_opts,
4152 *sbi->s_es->s_mount_opts ? "; " : "", orig_data);
4154 if (es->s_error_count)
4155 mod_timer(&sbi->s_err_report, jiffies + 300*HZ); /* 5 minutes */
4157 /* Enable message ratelimiting. Default is 10 messages per 5 secs. */
4158 ratelimit_state_init(&sbi->s_err_ratelimit_state, 5 * HZ, 10);
4159 ratelimit_state_init(&sbi->s_warning_ratelimit_state, 5 * HZ, 10);
4160 ratelimit_state_init(&sbi->s_msg_ratelimit_state, 5 * HZ, 10);
4167 ext4_msg(sb, KERN_ERR, "VFS: Can't find ext4 filesystem");
4172 kobject_del(&sbi->s_kobj);
4175 ext4_unregister_li_request(sb);
4177 ext4_mb_release(sb);
4179 ext4_ext_release(sb);
4180 ext4_release_system_zone(sb);
4185 ext4_msg(sb, KERN_ERR, "mount failed");
4186 if (EXT4_SB(sb)->rsv_conversion_wq)
4187 destroy_workqueue(EXT4_SB(sb)->rsv_conversion_wq);
4189 if (sbi->s_journal) {
4190 jbd2_journal_destroy(sbi->s_journal);
4191 sbi->s_journal = NULL;
4194 ext4_es_unregister_shrinker(sbi);
4195 del_timer_sync(&sbi->s_err_report);
4196 if (sbi->s_flex_groups)
4197 ext4_kvfree(sbi->s_flex_groups);
4198 percpu_counter_destroy(&sbi->s_freeclusters_counter);
4199 percpu_counter_destroy(&sbi->s_freeinodes_counter);
4200 percpu_counter_destroy(&sbi->s_dirs_counter);
4201 percpu_counter_destroy(&sbi->s_dirtyclusters_counter);
4202 percpu_counter_destroy(&sbi->s_extent_cache_cnt);
4204 kthread_stop(sbi->s_mmp_tsk);
4206 for (i = 0; i < db_count; i++)
4207 brelse(sbi->s_group_desc[i]);
4208 ext4_kvfree(sbi->s_group_desc);
4210 if (sbi->s_chksum_driver)
4211 crypto_free_shash(sbi->s_chksum_driver);
4213 remove_proc_entry("options", sbi->s_proc);
4214 remove_proc_entry(sb->s_id, ext4_proc_root);
4217 for (i = 0; i < MAXQUOTAS; i++)
4218 kfree(sbi->s_qf_names[i]);
4220 ext4_blkdev_remove(sbi);
4223 sb->s_fs_info = NULL;
4224 kfree(sbi->s_blockgroup_lock);
4228 return err ? err : ret;
4232 * Setup any per-fs journal parameters now. We'll do this both on
4233 * initial mount, once the journal has been initialised but before we've
4234 * done any recovery; and again on any subsequent remount.
4236 static void ext4_init_journal_params(struct super_block *sb, journal_t *journal)
4238 struct ext4_sb_info *sbi = EXT4_SB(sb);
4240 journal->j_commit_interval = sbi->s_commit_interval;
4241 journal->j_min_batch_time = sbi->s_min_batch_time;
4242 journal->j_max_batch_time = sbi->s_max_batch_time;
4244 write_lock(&journal->j_state_lock);
4245 if (test_opt(sb, BARRIER))
4246 journal->j_flags |= JBD2_BARRIER;
4248 journal->j_flags &= ~JBD2_BARRIER;
4249 if (test_opt(sb, DATA_ERR_ABORT))
4250 journal->j_flags |= JBD2_ABORT_ON_SYNCDATA_ERR;
4252 journal->j_flags &= ~JBD2_ABORT_ON_SYNCDATA_ERR;
4253 write_unlock(&journal->j_state_lock);
4256 static journal_t *ext4_get_journal(struct super_block *sb,
4257 unsigned int journal_inum)
4259 struct inode *journal_inode;
4262 BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL));
4264 /* First, test for the existence of a valid inode on disk. Bad
4265 * things happen if we iget() an unused inode, as the subsequent
4266 * iput() will try to delete it. */
4268 journal_inode = ext4_iget(sb, journal_inum);
4269 if (IS_ERR(journal_inode)) {
4270 ext4_msg(sb, KERN_ERR, "no journal found");
4273 if (!journal_inode->i_nlink) {
4274 make_bad_inode(journal_inode);
4275 iput(journal_inode);
4276 ext4_msg(sb, KERN_ERR, "journal inode is deleted");
4280 jbd_debug(2, "Journal inode found at %p: %lld bytes\n",
4281 journal_inode, journal_inode->i_size);
4282 if (!S_ISREG(journal_inode->i_mode)) {
4283 ext4_msg(sb, KERN_ERR, "invalid journal inode");
4284 iput(journal_inode);
4288 journal = jbd2_journal_init_inode(journal_inode);
4290 ext4_msg(sb, KERN_ERR, "Could not load journal inode");
4291 iput(journal_inode);
4294 journal->j_private = sb;
4295 ext4_init_journal_params(sb, journal);
4299 static journal_t *ext4_get_dev_journal(struct super_block *sb,
4302 struct buffer_head *bh;
4306 int hblock, blocksize;
4307 ext4_fsblk_t sb_block;
4308 unsigned long offset;
4309 struct ext4_super_block *es;
4310 struct block_device *bdev;
4312 BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL));
4314 bdev = ext4_blkdev_get(j_dev, sb);
4318 blocksize = sb->s_blocksize;
4319 hblock = bdev_logical_block_size(bdev);
4320 if (blocksize < hblock) {
4321 ext4_msg(sb, KERN_ERR,
4322 "blocksize too small for journal device");
4326 sb_block = EXT4_MIN_BLOCK_SIZE / blocksize;
4327 offset = EXT4_MIN_BLOCK_SIZE % blocksize;
4328 set_blocksize(bdev, blocksize);
4329 if (!(bh = __bread(bdev, sb_block, blocksize))) {
4330 ext4_msg(sb, KERN_ERR, "couldn't read superblock of "
4331 "external journal");
4335 es = (struct ext4_super_block *) (bh->b_data + offset);
4336 if ((le16_to_cpu(es->s_magic) != EXT4_SUPER_MAGIC) ||
4337 !(le32_to_cpu(es->s_feature_incompat) &
4338 EXT4_FEATURE_INCOMPAT_JOURNAL_DEV)) {
4339 ext4_msg(sb, KERN_ERR, "external journal has "
4345 if (memcmp(EXT4_SB(sb)->s_es->s_journal_uuid, es->s_uuid, 16)) {
4346 ext4_msg(sb, KERN_ERR, "journal UUID does not match");
4351 len = ext4_blocks_count(es);
4352 start = sb_block + 1;
4353 brelse(bh); /* we're done with the superblock */
4355 journal = jbd2_journal_init_dev(bdev, sb->s_bdev,
4356 start, len, blocksize);
4358 ext4_msg(sb, KERN_ERR, "failed to create device journal");
4361 journal->j_private = sb;
4362 ll_rw_block(READ | REQ_META | REQ_PRIO, 1, &journal->j_sb_buffer);
4363 wait_on_buffer(journal->j_sb_buffer);
4364 if (!buffer_uptodate(journal->j_sb_buffer)) {
4365 ext4_msg(sb, KERN_ERR, "I/O error on journal device");
4368 if (be32_to_cpu(journal->j_superblock->s_nr_users) != 1) {
4369 ext4_msg(sb, KERN_ERR, "External journal has more than one "
4370 "user (unsupported) - %d",
4371 be32_to_cpu(journal->j_superblock->s_nr_users));
4374 EXT4_SB(sb)->journal_bdev = bdev;
4375 ext4_init_journal_params(sb, journal);
4379 jbd2_journal_destroy(journal);
4381 ext4_blkdev_put(bdev);
4385 static int ext4_load_journal(struct super_block *sb,
4386 struct ext4_super_block *es,
4387 unsigned long journal_devnum)
4390 unsigned int journal_inum = le32_to_cpu(es->s_journal_inum);
4393 int really_read_only;
4395 BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL));
4397 if (journal_devnum &&
4398 journal_devnum != le32_to_cpu(es->s_journal_dev)) {
4399 ext4_msg(sb, KERN_INFO, "external journal device major/minor "
4400 "numbers have changed");
4401 journal_dev = new_decode_dev(journal_devnum);
4403 journal_dev = new_decode_dev(le32_to_cpu(es->s_journal_dev));
4405 really_read_only = bdev_read_only(sb->s_bdev);
4408 * Are we loading a blank journal or performing recovery after a
4409 * crash? For recovery, we need to check in advance whether we
4410 * can get read-write access to the device.
4412 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER)) {
4413 if (sb->s_flags & MS_RDONLY) {
4414 ext4_msg(sb, KERN_INFO, "INFO: recovery "
4415 "required on readonly filesystem");
4416 if (really_read_only) {
4417 ext4_msg(sb, KERN_ERR, "write access "
4418 "unavailable, cannot proceed");
4421 ext4_msg(sb, KERN_INFO, "write access will "
4422 "be enabled during recovery");
4426 if (journal_inum && journal_dev) {
4427 ext4_msg(sb, KERN_ERR, "filesystem has both journal "
4428 "and inode journals!");
4433 if (!(journal = ext4_get_journal(sb, journal_inum)))
4436 if (!(journal = ext4_get_dev_journal(sb, journal_dev)))
4440 if (!(journal->j_flags & JBD2_BARRIER))
4441 ext4_msg(sb, KERN_INFO, "barriers disabled");
4443 if (!EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER))
4444 err = jbd2_journal_wipe(journal, !really_read_only);
4446 char *save = kmalloc(EXT4_S_ERR_LEN, GFP_KERNEL);
4448 memcpy(save, ((char *) es) +
4449 EXT4_S_ERR_START, EXT4_S_ERR_LEN);
4450 err = jbd2_journal_load(journal);
4452 memcpy(((char *) es) + EXT4_S_ERR_START,
4453 save, EXT4_S_ERR_LEN);
4458 ext4_msg(sb, KERN_ERR, "error loading journal");
4459 jbd2_journal_destroy(journal);
4463 EXT4_SB(sb)->s_journal = journal;
4464 ext4_clear_journal_err(sb, es);
4466 if (!really_read_only && journal_devnum &&
4467 journal_devnum != le32_to_cpu(es->s_journal_dev)) {
4468 es->s_journal_dev = cpu_to_le32(journal_devnum);
4470 /* Make sure we flush the recovery flag to disk. */
4471 ext4_commit_super(sb, 1);
4477 static int ext4_commit_super(struct super_block *sb, int sync)
4479 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
4480 struct buffer_head *sbh = EXT4_SB(sb)->s_sbh;
4483 if (!sbh || block_device_ejected(sb))
4485 if (buffer_write_io_error(sbh)) {
4487 * Oh, dear. A previous attempt to write the
4488 * superblock failed. This could happen because the
4489 * USB device was yanked out. Or it could happen to
4490 * be a transient write error and maybe the block will
4491 * be remapped. Nothing we can do but to retry the
4492 * write and hope for the best.
4494 ext4_msg(sb, KERN_ERR, "previous I/O error to "
4495 "superblock detected");
4496 clear_buffer_write_io_error(sbh);
4497 set_buffer_uptodate(sbh);
4500 * If the file system is mounted read-only, don't update the
4501 * superblock write time. This avoids updating the superblock
4502 * write time when we are mounting the root file system
4503 * read/only but we need to replay the journal; at that point,
4504 * for people who are east of GMT and who make their clock
4505 * tick in localtime for Windows bug-for-bug compatibility,
4506 * the clock is set in the future, and this will cause e2fsck
4507 * to complain and force a full file system check.
4509 if (!(sb->s_flags & MS_RDONLY))
4510 es->s_wtime = cpu_to_le32(get_seconds());
4511 if (sb->s_bdev->bd_part)
4512 es->s_kbytes_written =
4513 cpu_to_le64(EXT4_SB(sb)->s_kbytes_written +
4514 ((part_stat_read(sb->s_bdev->bd_part, sectors[1]) -
4515 EXT4_SB(sb)->s_sectors_written_start) >> 1));
4517 es->s_kbytes_written =
4518 cpu_to_le64(EXT4_SB(sb)->s_kbytes_written);
4519 ext4_free_blocks_count_set(es,
4520 EXT4_C2B(EXT4_SB(sb), percpu_counter_sum_positive(
4521 &EXT4_SB(sb)->s_freeclusters_counter)));
4522 es->s_free_inodes_count =
4523 cpu_to_le32(percpu_counter_sum_positive(
4524 &EXT4_SB(sb)->s_freeinodes_counter));
4525 BUFFER_TRACE(sbh, "marking dirty");
4526 ext4_superblock_csum_set(sb);
4527 mark_buffer_dirty(sbh);
4529 error = sync_dirty_buffer(sbh);
4533 error = buffer_write_io_error(sbh);
4535 ext4_msg(sb, KERN_ERR, "I/O error while writing "
4537 clear_buffer_write_io_error(sbh);
4538 set_buffer_uptodate(sbh);
4545 * Have we just finished recovery? If so, and if we are mounting (or
4546 * remounting) the filesystem readonly, then we will end up with a
4547 * consistent fs on disk. Record that fact.
4549 static void ext4_mark_recovery_complete(struct super_block *sb,
4550 struct ext4_super_block *es)
4552 journal_t *journal = EXT4_SB(sb)->s_journal;
4554 if (!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL)) {
4555 BUG_ON(journal != NULL);
4558 jbd2_journal_lock_updates(journal);
4559 if (jbd2_journal_flush(journal) < 0)
4562 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER) &&
4563 sb->s_flags & MS_RDONLY) {
4564 EXT4_CLEAR_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
4565 ext4_commit_super(sb, 1);
4569 jbd2_journal_unlock_updates(journal);
4573 * If we are mounting (or read-write remounting) a filesystem whose journal
4574 * has recorded an error from a previous lifetime, move that error to the
4575 * main filesystem now.
4577 static void ext4_clear_journal_err(struct super_block *sb,
4578 struct ext4_super_block *es)
4584 BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL));
4586 journal = EXT4_SB(sb)->s_journal;
4589 * Now check for any error status which may have been recorded in the
4590 * journal by a prior ext4_error() or ext4_abort()
4593 j_errno = jbd2_journal_errno(journal);
4597 errstr = ext4_decode_error(sb, j_errno, nbuf);
4598 ext4_warning(sb, "Filesystem error recorded "
4599 "from previous mount: %s", errstr);
4600 ext4_warning(sb, "Marking fs in need of filesystem check.");
4602 EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS;
4603 es->s_state |= cpu_to_le16(EXT4_ERROR_FS);
4604 ext4_commit_super(sb, 1);
4606 jbd2_journal_clear_err(journal);
4607 jbd2_journal_update_sb_errno(journal);
4612 * Force the running and committing transactions to commit,
4613 * and wait on the commit.
4615 int ext4_force_commit(struct super_block *sb)
4619 if (sb->s_flags & MS_RDONLY)
4622 journal = EXT4_SB(sb)->s_journal;
4623 return ext4_journal_force_commit(journal);
4626 static int ext4_sync_fs(struct super_block *sb, int wait)
4630 bool needs_barrier = false;
4631 struct ext4_sb_info *sbi = EXT4_SB(sb);
4633 trace_ext4_sync_fs(sb, wait);
4634 flush_workqueue(sbi->rsv_conversion_wq);
4636 * Writeback quota in non-journalled quota case - journalled quota has
4639 dquot_writeback_dquots(sb, -1);
4641 * Data writeback is possible w/o journal transaction, so barrier must
4642 * being sent at the end of the function. But we can skip it if
4643 * transaction_commit will do it for us.
4645 target = jbd2_get_latest_transaction(sbi->s_journal);
4646 if (wait && sbi->s_journal->j_flags & JBD2_BARRIER &&
4647 !jbd2_trans_will_send_data_barrier(sbi->s_journal, target))
4648 needs_barrier = true;
4650 if (jbd2_journal_start_commit(sbi->s_journal, &target)) {
4652 ret = jbd2_log_wait_commit(sbi->s_journal, target);
4654 if (needs_barrier) {
4656 err = blkdev_issue_flush(sb->s_bdev, GFP_KERNEL, NULL);
4664 static int ext4_sync_fs_nojournal(struct super_block *sb, int wait)
4668 trace_ext4_sync_fs(sb, wait);
4669 flush_workqueue(EXT4_SB(sb)->rsv_conversion_wq);
4670 dquot_writeback_dquots(sb, -1);
4671 if (wait && test_opt(sb, BARRIER))
4672 ret = blkdev_issue_flush(sb->s_bdev, GFP_KERNEL, NULL);
4678 * LVM calls this function before a (read-only) snapshot is created. This
4679 * gives us a chance to flush the journal completely and mark the fs clean.
4681 * Note that only this function cannot bring a filesystem to be in a clean
4682 * state independently. It relies on upper layer to stop all data & metadata
4685 static int ext4_freeze(struct super_block *sb)
4690 if (sb->s_flags & MS_RDONLY)
4693 journal = EXT4_SB(sb)->s_journal;
4695 /* Now we set up the journal barrier. */
4696 jbd2_journal_lock_updates(journal);
4699 * Don't clear the needs_recovery flag if we failed to flush
4702 error = jbd2_journal_flush(journal);
4706 /* Journal blocked and flushed, clear needs_recovery flag. */
4707 EXT4_CLEAR_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
4708 error = ext4_commit_super(sb, 1);
4710 /* we rely on upper layer to stop further updates */
4711 jbd2_journal_unlock_updates(EXT4_SB(sb)->s_journal);
4716 * Called by LVM after the snapshot is done. We need to reset the RECOVER
4717 * flag here, even though the filesystem is not technically dirty yet.
4719 static int ext4_unfreeze(struct super_block *sb)
4721 if (sb->s_flags & MS_RDONLY)
4724 /* Reset the needs_recovery flag before the fs is unlocked. */
4725 EXT4_SET_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
4726 ext4_commit_super(sb, 1);
4731 * Structure to save mount options for ext4_remount's benefit
4733 struct ext4_mount_options {
4734 unsigned long s_mount_opt;
4735 unsigned long s_mount_opt2;
4738 unsigned long s_commit_interval;
4739 u32 s_min_batch_time, s_max_batch_time;
4742 char *s_qf_names[MAXQUOTAS];
4746 static int ext4_remount(struct super_block *sb, int *flags, char *data)
4748 struct ext4_super_block *es;
4749 struct ext4_sb_info *sbi = EXT4_SB(sb);
4750 unsigned long old_sb_flags;
4751 struct ext4_mount_options old_opts;
4752 int enable_quota = 0;
4754 unsigned int journal_ioprio = DEFAULT_JOURNAL_IOPRIO;
4759 char *orig_data = kstrdup(data, GFP_KERNEL);
4761 /* Store the original options */
4762 old_sb_flags = sb->s_flags;
4763 old_opts.s_mount_opt = sbi->s_mount_opt;
4764 old_opts.s_mount_opt2 = sbi->s_mount_opt2;
4765 old_opts.s_resuid = sbi->s_resuid;
4766 old_opts.s_resgid = sbi->s_resgid;
4767 old_opts.s_commit_interval = sbi->s_commit_interval;
4768 old_opts.s_min_batch_time = sbi->s_min_batch_time;
4769 old_opts.s_max_batch_time = sbi->s_max_batch_time;
4771 old_opts.s_jquota_fmt = sbi->s_jquota_fmt;
4772 for (i = 0; i < MAXQUOTAS; i++)
4773 if (sbi->s_qf_names[i]) {
4774 old_opts.s_qf_names[i] = kstrdup(sbi->s_qf_names[i],
4776 if (!old_opts.s_qf_names[i]) {
4777 for (j = 0; j < i; j++)
4778 kfree(old_opts.s_qf_names[j]);
4783 old_opts.s_qf_names[i] = NULL;
4785 if (sbi->s_journal && sbi->s_journal->j_task->io_context)
4786 journal_ioprio = sbi->s_journal->j_task->io_context->ioprio;
4789 * Allow the "check" option to be passed as a remount option.
4791 if (!parse_options(data, sb, NULL, &journal_ioprio, 1)) {
4796 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA) {
4797 if (test_opt2(sb, EXPLICIT_DELALLOC)) {
4798 ext4_msg(sb, KERN_ERR, "can't mount with "
4799 "both data=journal and delalloc");
4803 if (test_opt(sb, DIOREAD_NOLOCK)) {
4804 ext4_msg(sb, KERN_ERR, "can't mount with "
4805 "both data=journal and dioread_nolock");
4811 if (sbi->s_mount_flags & EXT4_MF_FS_ABORTED)
4812 ext4_abort(sb, "Abort forced by user");
4814 sb->s_flags = (sb->s_flags & ~MS_POSIXACL) |
4815 (test_opt(sb, POSIX_ACL) ? MS_POSIXACL : 0);
4819 if (sbi->s_journal) {
4820 ext4_init_journal_params(sb, sbi->s_journal);
4821 set_task_ioprio(sbi->s_journal->j_task, journal_ioprio);
4824 if ((*flags & MS_RDONLY) != (sb->s_flags & MS_RDONLY)) {
4825 if (sbi->s_mount_flags & EXT4_MF_FS_ABORTED) {
4830 if (*flags & MS_RDONLY) {
4831 err = dquot_suspend(sb, -1);
4836 * First of all, the unconditional stuff we have to do
4837 * to disable replay of the journal when we next remount
4839 sb->s_flags |= MS_RDONLY;
4842 * OK, test if we are remounting a valid rw partition
4843 * readonly, and if so set the rdonly flag and then
4844 * mark the partition as valid again.
4846 if (!(es->s_state & cpu_to_le16(EXT4_VALID_FS)) &&
4847 (sbi->s_mount_state & EXT4_VALID_FS))
4848 es->s_state = cpu_to_le16(sbi->s_mount_state);
4851 ext4_mark_recovery_complete(sb, es);
4853 /* Make sure we can mount this feature set readwrite */
4854 if (!ext4_feature_set_ok(sb, 0)) {
4859 * Make sure the group descriptor checksums
4860 * are sane. If they aren't, refuse to remount r/w.
4862 for (g = 0; g < sbi->s_groups_count; g++) {
4863 struct ext4_group_desc *gdp =
4864 ext4_get_group_desc(sb, g, NULL);
4866 if (!ext4_group_desc_csum_verify(sb, g, gdp)) {
4867 ext4_msg(sb, KERN_ERR,
4868 "ext4_remount: Checksum for group %u failed (%u!=%u)",
4869 g, le16_to_cpu(ext4_group_desc_csum(sbi, g, gdp)),
4870 le16_to_cpu(gdp->bg_checksum));
4877 * If we have an unprocessed orphan list hanging
4878 * around from a previously readonly bdev mount,
4879 * require a full umount/remount for now.
4881 if (es->s_last_orphan) {
4882 ext4_msg(sb, KERN_WARNING, "Couldn't "
4883 "remount RDWR because of unprocessed "
4884 "orphan inode list. Please "
4885 "umount/remount instead");
4891 * Mounting a RDONLY partition read-write, so reread
4892 * and store the current valid flag. (It may have
4893 * been changed by e2fsck since we originally mounted
4897 ext4_clear_journal_err(sb, es);
4898 sbi->s_mount_state = le16_to_cpu(es->s_state);
4899 if (!ext4_setup_super(sb, es, 0))
4900 sb->s_flags &= ~MS_RDONLY;
4901 if (EXT4_HAS_INCOMPAT_FEATURE(sb,
4902 EXT4_FEATURE_INCOMPAT_MMP))
4903 if (ext4_multi_mount_protect(sb,
4904 le64_to_cpu(es->s_mmp_block))) {
4913 * Reinitialize lazy itable initialization thread based on
4916 if ((sb->s_flags & MS_RDONLY) || !test_opt(sb, INIT_INODE_TABLE))
4917 ext4_unregister_li_request(sb);
4919 ext4_group_t first_not_zeroed;
4920 first_not_zeroed = ext4_has_uninit_itable(sb);
4921 ext4_register_li_request(sb, first_not_zeroed);
4924 ext4_setup_system_zone(sb);
4925 if (sbi->s_journal == NULL && !(old_sb_flags & MS_RDONLY))
4926 ext4_commit_super(sb, 1);
4929 /* Release old quota file names */
4930 for (i = 0; i < MAXQUOTAS; i++)
4931 kfree(old_opts.s_qf_names[i]);
4933 if (sb_any_quota_suspended(sb))
4934 dquot_resume(sb, -1);
4935 else if (EXT4_HAS_RO_COMPAT_FEATURE(sb,
4936 EXT4_FEATURE_RO_COMPAT_QUOTA)) {
4937 err = ext4_enable_quotas(sb);
4944 ext4_msg(sb, KERN_INFO, "re-mounted. Opts: %s", orig_data);
4949 sb->s_flags = old_sb_flags;
4950 sbi->s_mount_opt = old_opts.s_mount_opt;
4951 sbi->s_mount_opt2 = old_opts.s_mount_opt2;
4952 sbi->s_resuid = old_opts.s_resuid;
4953 sbi->s_resgid = old_opts.s_resgid;
4954 sbi->s_commit_interval = old_opts.s_commit_interval;
4955 sbi->s_min_batch_time = old_opts.s_min_batch_time;
4956 sbi->s_max_batch_time = old_opts.s_max_batch_time;
4958 sbi->s_jquota_fmt = old_opts.s_jquota_fmt;
4959 for (i = 0; i < MAXQUOTAS; i++) {
4960 kfree(sbi->s_qf_names[i]);
4961 sbi->s_qf_names[i] = old_opts.s_qf_names[i];
4968 static int ext4_statfs(struct dentry *dentry, struct kstatfs *buf)
4970 struct super_block *sb = dentry->d_sb;
4971 struct ext4_sb_info *sbi = EXT4_SB(sb);
4972 struct ext4_super_block *es = sbi->s_es;
4973 ext4_fsblk_t overhead = 0, resv_blocks;
4976 resv_blocks = EXT4_C2B(sbi, atomic64_read(&sbi->s_resv_clusters));
4978 if (!test_opt(sb, MINIX_DF))
4979 overhead = sbi->s_overhead;
4981 buf->f_type = EXT4_SUPER_MAGIC;
4982 buf->f_bsize = sb->s_blocksize;
4983 buf->f_blocks = ext4_blocks_count(es) - EXT4_C2B(sbi, overhead);
4984 bfree = percpu_counter_sum_positive(&sbi->s_freeclusters_counter) -
4985 percpu_counter_sum_positive(&sbi->s_dirtyclusters_counter);
4986 /* prevent underflow in case that few free space is available */
4987 buf->f_bfree = EXT4_C2B(sbi, max_t(s64, bfree, 0));
4988 buf->f_bavail = buf->f_bfree -
4989 (ext4_r_blocks_count(es) + resv_blocks);
4990 if (buf->f_bfree < (ext4_r_blocks_count(es) + resv_blocks))
4992 buf->f_files = le32_to_cpu(es->s_inodes_count);
4993 buf->f_ffree = percpu_counter_sum_positive(&sbi->s_freeinodes_counter);
4994 buf->f_namelen = EXT4_NAME_LEN;
4995 fsid = le64_to_cpup((void *)es->s_uuid) ^
4996 le64_to_cpup((void *)es->s_uuid + sizeof(u64));
4997 buf->f_fsid.val[0] = fsid & 0xFFFFFFFFUL;
4998 buf->f_fsid.val[1] = (fsid >> 32) & 0xFFFFFFFFUL;
5003 /* Helper function for writing quotas on sync - we need to start transaction
5004 * before quota file is locked for write. Otherwise the are possible deadlocks:
5005 * Process 1 Process 2
5006 * ext4_create() quota_sync()
5007 * jbd2_journal_start() write_dquot()
5008 * dquot_initialize() down(dqio_mutex)
5009 * down(dqio_mutex) jbd2_journal_start()
5015 static inline struct inode *dquot_to_inode(struct dquot *dquot)
5017 return sb_dqopt(dquot->dq_sb)->files[dquot->dq_id.type];
5020 static int ext4_write_dquot(struct dquot *dquot)
5024 struct inode *inode;
5026 inode = dquot_to_inode(dquot);
5027 handle = ext4_journal_start(inode, EXT4_HT_QUOTA,
5028 EXT4_QUOTA_TRANS_BLOCKS(dquot->dq_sb));
5030 return PTR_ERR(handle);
5031 ret = dquot_commit(dquot);
5032 err = ext4_journal_stop(handle);
5038 static int ext4_acquire_dquot(struct dquot *dquot)
5043 handle = ext4_journal_start(dquot_to_inode(dquot), EXT4_HT_QUOTA,
5044 EXT4_QUOTA_INIT_BLOCKS(dquot->dq_sb));
5046 return PTR_ERR(handle);
5047 ret = dquot_acquire(dquot);
5048 err = ext4_journal_stop(handle);
5054 static int ext4_release_dquot(struct dquot *dquot)
5059 handle = ext4_journal_start(dquot_to_inode(dquot), EXT4_HT_QUOTA,
5060 EXT4_QUOTA_DEL_BLOCKS(dquot->dq_sb));
5061 if (IS_ERR(handle)) {
5062 /* Release dquot anyway to avoid endless cycle in dqput() */
5063 dquot_release(dquot);
5064 return PTR_ERR(handle);
5066 ret = dquot_release(dquot);
5067 err = ext4_journal_stop(handle);
5073 static int ext4_mark_dquot_dirty(struct dquot *dquot)
5075 struct super_block *sb = dquot->dq_sb;
5076 struct ext4_sb_info *sbi = EXT4_SB(sb);
5078 /* Are we journaling quotas? */
5079 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_QUOTA) ||
5080 sbi->s_qf_names[USRQUOTA] || sbi->s_qf_names[GRPQUOTA]) {
5081 dquot_mark_dquot_dirty(dquot);
5082 return ext4_write_dquot(dquot);
5084 return dquot_mark_dquot_dirty(dquot);
5088 static int ext4_write_info(struct super_block *sb, int type)
5093 /* Data block + inode block */
5094 handle = ext4_journal_start(sb->s_root->d_inode, EXT4_HT_QUOTA, 2);
5096 return PTR_ERR(handle);
5097 ret = dquot_commit_info(sb, type);
5098 err = ext4_journal_stop(handle);
5105 * Turn on quotas during mount time - we need to find
5106 * the quota file and such...
5108 static int ext4_quota_on_mount(struct super_block *sb, int type)
5110 return dquot_quota_on_mount(sb, EXT4_SB(sb)->s_qf_names[type],
5111 EXT4_SB(sb)->s_jquota_fmt, type);
5115 * Standard function to be called on quota_on
5117 static int ext4_quota_on(struct super_block *sb, int type, int format_id,
5122 if (!test_opt(sb, QUOTA))
5125 /* Quotafile not on the same filesystem? */
5126 if (path->dentry->d_sb != sb)
5128 /* Journaling quota? */
5129 if (EXT4_SB(sb)->s_qf_names[type]) {
5130 /* Quotafile not in fs root? */
5131 if (path->dentry->d_parent != sb->s_root)
5132 ext4_msg(sb, KERN_WARNING,
5133 "Quota file not on filesystem root. "
5134 "Journaled quota will not work");
5138 * When we journal data on quota file, we have to flush journal to see
5139 * all updates to the file when we bypass pagecache...
5141 if (EXT4_SB(sb)->s_journal &&
5142 ext4_should_journal_data(path->dentry->d_inode)) {
5144 * We don't need to lock updates but journal_flush() could
5145 * otherwise be livelocked...
5147 jbd2_journal_lock_updates(EXT4_SB(sb)->s_journal);
5148 err = jbd2_journal_flush(EXT4_SB(sb)->s_journal);
5149 jbd2_journal_unlock_updates(EXT4_SB(sb)->s_journal);
5154 return dquot_quota_on(sb, type, format_id, path);
5157 static int ext4_quota_enable(struct super_block *sb, int type, int format_id,
5161 struct inode *qf_inode;
5162 unsigned long qf_inums[MAXQUOTAS] = {
5163 le32_to_cpu(EXT4_SB(sb)->s_es->s_usr_quota_inum),
5164 le32_to_cpu(EXT4_SB(sb)->s_es->s_grp_quota_inum)
5167 BUG_ON(!EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_QUOTA));
5169 if (!qf_inums[type])
5172 qf_inode = ext4_iget(sb, qf_inums[type]);
5173 if (IS_ERR(qf_inode)) {
5174 ext4_error(sb, "Bad quota inode # %lu", qf_inums[type]);
5175 return PTR_ERR(qf_inode);
5178 /* Don't account quota for quota files to avoid recursion */
5179 qf_inode->i_flags |= S_NOQUOTA;
5180 err = dquot_enable(qf_inode, type, format_id, flags);
5186 /* Enable usage tracking for all quota types. */
5187 static int ext4_enable_quotas(struct super_block *sb)
5190 unsigned long qf_inums[MAXQUOTAS] = {
5191 le32_to_cpu(EXT4_SB(sb)->s_es->s_usr_quota_inum),
5192 le32_to_cpu(EXT4_SB(sb)->s_es->s_grp_quota_inum)
5195 sb_dqopt(sb)->flags |= DQUOT_QUOTA_SYS_FILE;
5196 for (type = 0; type < MAXQUOTAS; type++) {
5197 if (qf_inums[type]) {
5198 err = ext4_quota_enable(sb, type, QFMT_VFS_V1,
5199 DQUOT_USAGE_ENABLED);
5202 "Failed to enable quota tracking "
5203 "(type=%d, err=%d). Please run "
5204 "e2fsck to fix.", type, err);
5213 * quota_on function that is used when QUOTA feature is set.
5215 static int ext4_quota_on_sysfile(struct super_block *sb, int type,
5218 if (!EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_QUOTA))
5222 * USAGE was enabled at mount time. Only need to enable LIMITS now.
5224 return ext4_quota_enable(sb, type, format_id, DQUOT_LIMITS_ENABLED);
5227 static int ext4_quota_off(struct super_block *sb, int type)
5229 struct inode *inode = sb_dqopt(sb)->files[type];
5232 /* Force all delayed allocation blocks to be allocated.
5233 * Caller already holds s_umount sem */
5234 if (test_opt(sb, DELALLOC))
5235 sync_filesystem(sb);
5240 /* Update modification times of quota files when userspace can
5241 * start looking at them */
5242 handle = ext4_journal_start(inode, EXT4_HT_QUOTA, 1);
5245 inode->i_mtime = inode->i_ctime = CURRENT_TIME;
5246 ext4_mark_inode_dirty(handle, inode);
5247 ext4_journal_stop(handle);
5250 return dquot_quota_off(sb, type);
5254 * quota_off function that is used when QUOTA feature is set.
5256 static int ext4_quota_off_sysfile(struct super_block *sb, int type)
5258 if (!EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_QUOTA))
5261 /* Disable only the limits. */
5262 return dquot_disable(sb, type, DQUOT_LIMITS_ENABLED);
5265 /* Read data from quotafile - avoid pagecache and such because we cannot afford
5266 * acquiring the locks... As quota files are never truncated and quota code
5267 * itself serializes the operations (and no one else should touch the files)
5268 * we don't have to be afraid of races */
5269 static ssize_t ext4_quota_read(struct super_block *sb, int type, char *data,
5270 size_t len, loff_t off)
5272 struct inode *inode = sb_dqopt(sb)->files[type];
5273 ext4_lblk_t blk = off >> EXT4_BLOCK_SIZE_BITS(sb);
5275 int offset = off & (sb->s_blocksize - 1);
5278 struct buffer_head *bh;
5279 loff_t i_size = i_size_read(inode);
5283 if (off+len > i_size)
5286 while (toread > 0) {
5287 tocopy = sb->s_blocksize - offset < toread ?
5288 sb->s_blocksize - offset : toread;
5289 bh = ext4_bread(NULL, inode, blk, 0, &err);
5292 if (!bh) /* A hole? */
5293 memset(data, 0, tocopy);
5295 memcpy(data, bh->b_data+offset, tocopy);
5305 /* Write to quotafile (we know the transaction is already started and has
5306 * enough credits) */
5307 static ssize_t ext4_quota_write(struct super_block *sb, int type,
5308 const char *data, size_t len, loff_t off)
5310 struct inode *inode = sb_dqopt(sb)->files[type];
5311 ext4_lblk_t blk = off >> EXT4_BLOCK_SIZE_BITS(sb);
5313 int offset = off & (sb->s_blocksize - 1);
5314 struct buffer_head *bh;
5315 handle_t *handle = journal_current_handle();
5317 if (EXT4_SB(sb)->s_journal && !handle) {
5318 ext4_msg(sb, KERN_WARNING, "Quota write (off=%llu, len=%llu)"
5319 " cancelled because transaction is not started",
5320 (unsigned long long)off, (unsigned long long)len);
5324 * Since we account only one data block in transaction credits,
5325 * then it is impossible to cross a block boundary.
5327 if (sb->s_blocksize - offset < len) {
5328 ext4_msg(sb, KERN_WARNING, "Quota write (off=%llu, len=%llu)"
5329 " cancelled because not block aligned",
5330 (unsigned long long)off, (unsigned long long)len);
5334 bh = ext4_bread(handle, inode, blk, 1, &err);
5337 err = ext4_journal_get_write_access(handle, bh);
5343 memcpy(bh->b_data+offset, data, len);
5344 flush_dcache_page(bh->b_page);
5346 err = ext4_handle_dirty_metadata(handle, NULL, bh);
5351 if (inode->i_size < off + len) {
5352 i_size_write(inode, off + len);
5353 EXT4_I(inode)->i_disksize = inode->i_size;
5354 ext4_mark_inode_dirty(handle, inode);
5361 static struct dentry *ext4_mount(struct file_system_type *fs_type, int flags,
5362 const char *dev_name, void *data)
5364 return mount_bdev(fs_type, flags, dev_name, data, ext4_fill_super);
5367 #if !defined(CONFIG_EXT2_FS) && !defined(CONFIG_EXT2_FS_MODULE) && defined(CONFIG_EXT4_USE_FOR_EXT23)
5368 static inline void register_as_ext2(void)
5370 int err = register_filesystem(&ext2_fs_type);
5373 "EXT4-fs: Unable to register as ext2 (%d)\n", err);
5376 static inline void unregister_as_ext2(void)
5378 unregister_filesystem(&ext2_fs_type);
5381 static inline int ext2_feature_set_ok(struct super_block *sb)
5383 if (EXT4_HAS_INCOMPAT_FEATURE(sb, ~EXT2_FEATURE_INCOMPAT_SUPP))
5385 if (sb->s_flags & MS_RDONLY)
5387 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, ~EXT2_FEATURE_RO_COMPAT_SUPP))
5392 static inline void register_as_ext2(void) { }
5393 static inline void unregister_as_ext2(void) { }
5394 static inline int ext2_feature_set_ok(struct super_block *sb) { return 0; }
5397 #if !defined(CONFIG_EXT3_FS) && !defined(CONFIG_EXT3_FS_MODULE) && defined(CONFIG_EXT4_USE_FOR_EXT23)
5398 static inline void register_as_ext3(void)
5400 int err = register_filesystem(&ext3_fs_type);
5403 "EXT4-fs: Unable to register as ext3 (%d)\n", err);
5406 static inline void unregister_as_ext3(void)
5408 unregister_filesystem(&ext3_fs_type);
5411 static inline int ext3_feature_set_ok(struct super_block *sb)
5413 if (EXT4_HAS_INCOMPAT_FEATURE(sb, ~EXT3_FEATURE_INCOMPAT_SUPP))
5415 if (!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL))
5417 if (sb->s_flags & MS_RDONLY)
5419 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, ~EXT3_FEATURE_RO_COMPAT_SUPP))
5424 static inline void register_as_ext3(void) { }
5425 static inline void unregister_as_ext3(void) { }
5426 static inline int ext3_feature_set_ok(struct super_block *sb) { return 0; }
5429 static struct file_system_type ext4_fs_type = {
5430 .owner = THIS_MODULE,
5432 .mount = ext4_mount,
5433 .kill_sb = kill_block_super,
5434 .fs_flags = FS_REQUIRES_DEV,
5436 MODULE_ALIAS_FS("ext4");
5438 static int __init ext4_init_feat_adverts(void)
5440 struct ext4_features *ef;
5443 ef = kzalloc(sizeof(struct ext4_features), GFP_KERNEL);
5447 ef->f_kobj.kset = ext4_kset;
5448 init_completion(&ef->f_kobj_unregister);
5449 ret = kobject_init_and_add(&ef->f_kobj, &ext4_feat_ktype, NULL,
5462 static void ext4_exit_feat_adverts(void)
5464 kobject_put(&ext4_feat->f_kobj);
5465 wait_for_completion(&ext4_feat->f_kobj_unregister);
5469 /* Shared across all ext4 file systems */
5470 wait_queue_head_t ext4__ioend_wq[EXT4_WQ_HASH_SZ];
5471 struct mutex ext4__aio_mutex[EXT4_WQ_HASH_SZ];
5473 static int __init ext4_init_fs(void)
5477 ext4_li_info = NULL;
5478 mutex_init(&ext4_li_mtx);
5480 /* Build-time check for flags consistency */
5481 ext4_check_flag_values();
5483 for (i = 0; i < EXT4_WQ_HASH_SZ; i++) {
5484 mutex_init(&ext4__aio_mutex[i]);
5485 init_waitqueue_head(&ext4__ioend_wq[i]);
5488 err = ext4_init_es();
5492 err = ext4_init_pageio();
5496 err = ext4_init_system_zone();
5499 ext4_kset = kset_create_and_add("ext4", NULL, fs_kobj);
5504 ext4_proc_root = proc_mkdir("fs/ext4", NULL);
5506 err = ext4_init_feat_adverts();
5510 err = ext4_init_mballoc();
5514 err = ext4_init_xattr();
5517 err = init_inodecache();
5522 err = register_filesystem(&ext4_fs_type);
5528 unregister_as_ext2();
5529 unregister_as_ext3();
5530 destroy_inodecache();
5534 ext4_exit_mballoc();
5536 ext4_exit_feat_adverts();
5539 remove_proc_entry("fs/ext4", NULL);
5540 kset_unregister(ext4_kset);
5542 ext4_exit_system_zone();
5551 static void __exit ext4_exit_fs(void)
5553 ext4_destroy_lazyinit_thread();
5554 unregister_as_ext2();
5555 unregister_as_ext3();
5556 unregister_filesystem(&ext4_fs_type);
5557 destroy_inodecache();
5559 ext4_exit_mballoc();
5560 ext4_exit_feat_adverts();
5561 remove_proc_entry("fs/ext4", NULL);
5562 kset_unregister(ext4_kset);
5563 ext4_exit_system_zone();
5568 MODULE_AUTHOR("Remy Card, Stephen Tweedie, Andrew Morton, Andreas Dilger, Theodore Ts'o and others");
5569 MODULE_DESCRIPTION("Fourth Extended Filesystem");
5570 MODULE_LICENSE("GPL");
5571 module_init(ext4_init_fs)
5572 module_exit(ext4_exit_fs)