2 * linux/fs/ext4/super.c
4 * Copyright (C) 1992, 1993, 1994, 1995
5 * Remy Card (card@masi.ibp.fr)
6 * Laboratoire MASI - Institut Blaise Pascal
7 * Universite Pierre et Marie Curie (Paris VI)
11 * linux/fs/minix/inode.c
13 * Copyright (C) 1991, 1992 Linus Torvalds
15 * Big-endian to little-endian byte-swapping/bitmaps by
16 * David S. Miller (davem@caip.rutgers.edu), 1995
19 #include <linux/module.h>
20 #include <linux/string.h>
22 #include <linux/time.h>
23 #include <linux/vmalloc.h>
24 #include <linux/jbd2.h>
25 #include <linux/slab.h>
26 #include <linux/init.h>
27 #include <linux/blkdev.h>
28 #include <linux/parser.h>
29 #include <linux/buffer_head.h>
30 #include <linux/exportfs.h>
31 #include <linux/vfs.h>
32 #include <linux/random.h>
33 #include <linux/mount.h>
34 #include <linux/namei.h>
35 #include <linux/quotaops.h>
36 #include <linux/seq_file.h>
37 #include <linux/proc_fs.h>
38 #include <linux/ctype.h>
39 #include <linux/log2.h>
40 #include <linux/crc16.h>
41 #include <linux/cleancache.h>
42 #include <asm/uaccess.h>
44 #include <linux/kthread.h>
45 #include <linux/freezer.h>
48 #include "ext4_extents.h" /* Needed for trace points definition */
49 #include "ext4_jbd2.h"
54 #define CREATE_TRACE_POINTS
55 #include <trace/events/ext4.h>
57 static struct proc_dir_entry *ext4_proc_root;
58 static struct kset *ext4_kset;
59 static struct ext4_lazy_init *ext4_li_info;
60 static struct mutex ext4_li_mtx;
61 static struct ext4_features *ext4_feat;
62 static int ext4_mballoc_ready;
64 static int ext4_load_journal(struct super_block *, struct ext4_super_block *,
65 unsigned long journal_devnum);
66 static int ext4_show_options(struct seq_file *seq, struct dentry *root);
67 static int ext4_commit_super(struct super_block *sb, int sync);
68 static void ext4_mark_recovery_complete(struct super_block *sb,
69 struct ext4_super_block *es);
70 static void ext4_clear_journal_err(struct super_block *sb,
71 struct ext4_super_block *es);
72 static int ext4_sync_fs(struct super_block *sb, int wait);
73 static int ext4_remount(struct super_block *sb, int *flags, char *data);
74 static int ext4_statfs(struct dentry *dentry, struct kstatfs *buf);
75 static int ext4_unfreeze(struct super_block *sb);
76 static int ext4_freeze(struct super_block *sb);
77 static struct dentry *ext4_mount(struct file_system_type *fs_type, int flags,
78 const char *dev_name, void *data);
79 static inline int ext2_feature_set_ok(struct super_block *sb);
80 static inline int ext3_feature_set_ok(struct super_block *sb);
81 static int ext4_feature_set_ok(struct super_block *sb, int readonly);
82 static void ext4_destroy_lazyinit_thread(void);
83 static void ext4_unregister_li_request(struct super_block *sb);
84 static void ext4_clear_request_list(void);
85 static int ext4_reserve_clusters(struct ext4_sb_info *, ext4_fsblk_t);
87 #if !defined(CONFIG_EXT2_FS) && !defined(CONFIG_EXT2_FS_MODULE) && defined(CONFIG_EXT4_USE_FOR_EXT23)
88 static struct file_system_type ext2_fs_type = {
92 .kill_sb = kill_block_super,
93 .fs_flags = FS_REQUIRES_DEV,
95 MODULE_ALIAS_FS("ext2");
97 #define IS_EXT2_SB(sb) ((sb)->s_bdev->bd_holder == &ext2_fs_type)
99 #define IS_EXT2_SB(sb) (0)
103 #if !defined(CONFIG_EXT3_FS) && !defined(CONFIG_EXT3_FS_MODULE) && defined(CONFIG_EXT4_USE_FOR_EXT23)
104 static struct file_system_type ext3_fs_type = {
105 .owner = THIS_MODULE,
108 .kill_sb = kill_block_super,
109 .fs_flags = FS_REQUIRES_DEV,
111 MODULE_ALIAS_FS("ext3");
112 MODULE_ALIAS("ext3");
113 #define IS_EXT3_SB(sb) ((sb)->s_bdev->bd_holder == &ext3_fs_type)
115 #define IS_EXT3_SB(sb) (0)
118 static int ext4_verify_csum_type(struct super_block *sb,
119 struct ext4_super_block *es)
121 if (!EXT4_HAS_RO_COMPAT_FEATURE(sb,
122 EXT4_FEATURE_RO_COMPAT_METADATA_CSUM))
125 return es->s_checksum_type == EXT4_CRC32C_CHKSUM;
128 static __le32 ext4_superblock_csum(struct super_block *sb,
129 struct ext4_super_block *es)
131 struct ext4_sb_info *sbi = EXT4_SB(sb);
132 int offset = offsetof(struct ext4_super_block, s_checksum);
135 csum = ext4_chksum(sbi, ~0, (char *)es, offset);
137 return cpu_to_le32(csum);
140 static int ext4_superblock_csum_verify(struct super_block *sb,
141 struct ext4_super_block *es)
143 if (!ext4_has_metadata_csum(sb))
146 return es->s_checksum == ext4_superblock_csum(sb, es);
149 void ext4_superblock_csum_set(struct super_block *sb)
151 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
153 if (!ext4_has_metadata_csum(sb))
156 es->s_checksum = ext4_superblock_csum(sb, es);
159 void *ext4_kvmalloc(size_t size, gfp_t flags)
163 ret = kmalloc(size, flags | __GFP_NOWARN);
165 ret = __vmalloc(size, flags, PAGE_KERNEL);
169 void *ext4_kvzalloc(size_t size, gfp_t flags)
173 ret = kzalloc(size, flags | __GFP_NOWARN);
175 ret = __vmalloc(size, flags | __GFP_ZERO, PAGE_KERNEL);
179 ext4_fsblk_t ext4_block_bitmap(struct super_block *sb,
180 struct ext4_group_desc *bg)
182 return le32_to_cpu(bg->bg_block_bitmap_lo) |
183 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
184 (ext4_fsblk_t)le32_to_cpu(bg->bg_block_bitmap_hi) << 32 : 0);
187 ext4_fsblk_t ext4_inode_bitmap(struct super_block *sb,
188 struct ext4_group_desc *bg)
190 return le32_to_cpu(bg->bg_inode_bitmap_lo) |
191 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
192 (ext4_fsblk_t)le32_to_cpu(bg->bg_inode_bitmap_hi) << 32 : 0);
195 ext4_fsblk_t ext4_inode_table(struct super_block *sb,
196 struct ext4_group_desc *bg)
198 return le32_to_cpu(bg->bg_inode_table_lo) |
199 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
200 (ext4_fsblk_t)le32_to_cpu(bg->bg_inode_table_hi) << 32 : 0);
203 __u32 ext4_free_group_clusters(struct super_block *sb,
204 struct ext4_group_desc *bg)
206 return le16_to_cpu(bg->bg_free_blocks_count_lo) |
207 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
208 (__u32)le16_to_cpu(bg->bg_free_blocks_count_hi) << 16 : 0);
211 __u32 ext4_free_inodes_count(struct super_block *sb,
212 struct ext4_group_desc *bg)
214 return le16_to_cpu(bg->bg_free_inodes_count_lo) |
215 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
216 (__u32)le16_to_cpu(bg->bg_free_inodes_count_hi) << 16 : 0);
219 __u32 ext4_used_dirs_count(struct super_block *sb,
220 struct ext4_group_desc *bg)
222 return le16_to_cpu(bg->bg_used_dirs_count_lo) |
223 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
224 (__u32)le16_to_cpu(bg->bg_used_dirs_count_hi) << 16 : 0);
227 __u32 ext4_itable_unused_count(struct super_block *sb,
228 struct ext4_group_desc *bg)
230 return le16_to_cpu(bg->bg_itable_unused_lo) |
231 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
232 (__u32)le16_to_cpu(bg->bg_itable_unused_hi) << 16 : 0);
235 void ext4_block_bitmap_set(struct super_block *sb,
236 struct ext4_group_desc *bg, ext4_fsblk_t blk)
238 bg->bg_block_bitmap_lo = cpu_to_le32((u32)blk);
239 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
240 bg->bg_block_bitmap_hi = cpu_to_le32(blk >> 32);
243 void ext4_inode_bitmap_set(struct super_block *sb,
244 struct ext4_group_desc *bg, ext4_fsblk_t blk)
246 bg->bg_inode_bitmap_lo = cpu_to_le32((u32)blk);
247 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
248 bg->bg_inode_bitmap_hi = cpu_to_le32(blk >> 32);
251 void ext4_inode_table_set(struct super_block *sb,
252 struct ext4_group_desc *bg, ext4_fsblk_t blk)
254 bg->bg_inode_table_lo = cpu_to_le32((u32)blk);
255 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
256 bg->bg_inode_table_hi = cpu_to_le32(blk >> 32);
259 void ext4_free_group_clusters_set(struct super_block *sb,
260 struct ext4_group_desc *bg, __u32 count)
262 bg->bg_free_blocks_count_lo = cpu_to_le16((__u16)count);
263 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
264 bg->bg_free_blocks_count_hi = cpu_to_le16(count >> 16);
267 void ext4_free_inodes_set(struct super_block *sb,
268 struct ext4_group_desc *bg, __u32 count)
270 bg->bg_free_inodes_count_lo = cpu_to_le16((__u16)count);
271 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
272 bg->bg_free_inodes_count_hi = cpu_to_le16(count >> 16);
275 void ext4_used_dirs_set(struct super_block *sb,
276 struct ext4_group_desc *bg, __u32 count)
278 bg->bg_used_dirs_count_lo = cpu_to_le16((__u16)count);
279 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
280 bg->bg_used_dirs_count_hi = cpu_to_le16(count >> 16);
283 void ext4_itable_unused_set(struct super_block *sb,
284 struct ext4_group_desc *bg, __u32 count)
286 bg->bg_itable_unused_lo = cpu_to_le16((__u16)count);
287 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
288 bg->bg_itable_unused_hi = cpu_to_le16(count >> 16);
292 static void __save_error_info(struct super_block *sb, const char *func,
295 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
297 EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS;
298 es->s_state |= cpu_to_le16(EXT4_ERROR_FS);
299 es->s_last_error_time = cpu_to_le32(get_seconds());
300 strncpy(es->s_last_error_func, func, sizeof(es->s_last_error_func));
301 es->s_last_error_line = cpu_to_le32(line);
302 if (!es->s_first_error_time) {
303 es->s_first_error_time = es->s_last_error_time;
304 strncpy(es->s_first_error_func, func,
305 sizeof(es->s_first_error_func));
306 es->s_first_error_line = cpu_to_le32(line);
307 es->s_first_error_ino = es->s_last_error_ino;
308 es->s_first_error_block = es->s_last_error_block;
311 * Start the daily error reporting function if it hasn't been
314 if (!es->s_error_count)
315 mod_timer(&EXT4_SB(sb)->s_err_report, jiffies + 24*60*60*HZ);
316 le32_add_cpu(&es->s_error_count, 1);
319 static void save_error_info(struct super_block *sb, const char *func,
322 __save_error_info(sb, func, line);
323 ext4_commit_super(sb, 1);
327 * The del_gendisk() function uninitializes the disk-specific data
328 * structures, including the bdi structure, without telling anyone
329 * else. Once this happens, any attempt to call mark_buffer_dirty()
330 * (for example, by ext4_commit_super), will cause a kernel OOPS.
331 * This is a kludge to prevent these oops until we can put in a proper
332 * hook in del_gendisk() to inform the VFS and file system layers.
334 static int block_device_ejected(struct super_block *sb)
336 struct inode *bd_inode = sb->s_bdev->bd_inode;
337 struct backing_dev_info *bdi = inode_to_bdi(bd_inode);
339 return bdi->dev == NULL;
342 static void ext4_journal_commit_callback(journal_t *journal, transaction_t *txn)
344 struct super_block *sb = journal->j_private;
345 struct ext4_sb_info *sbi = EXT4_SB(sb);
346 int error = is_journal_aborted(journal);
347 struct ext4_journal_cb_entry *jce;
349 BUG_ON(txn->t_state == T_FINISHED);
350 spin_lock(&sbi->s_md_lock);
351 while (!list_empty(&txn->t_private_list)) {
352 jce = list_entry(txn->t_private_list.next,
353 struct ext4_journal_cb_entry, jce_list);
354 list_del_init(&jce->jce_list);
355 spin_unlock(&sbi->s_md_lock);
356 jce->jce_func(sb, jce, error);
357 spin_lock(&sbi->s_md_lock);
359 spin_unlock(&sbi->s_md_lock);
362 /* Deal with the reporting of failure conditions on a filesystem such as
363 * inconsistencies detected or read IO failures.
365 * On ext2, we can store the error state of the filesystem in the
366 * superblock. That is not possible on ext4, because we may have other
367 * write ordering constraints on the superblock which prevent us from
368 * writing it out straight away; and given that the journal is about to
369 * be aborted, we can't rely on the current, or future, transactions to
370 * write out the superblock safely.
372 * We'll just use the jbd2_journal_abort() error code to record an error in
373 * the journal instead. On recovery, the journal will complain about
374 * that error until we've noted it down and cleared it.
377 static void ext4_handle_error(struct super_block *sb)
379 if (sb->s_flags & MS_RDONLY)
382 if (!test_opt(sb, ERRORS_CONT)) {
383 journal_t *journal = EXT4_SB(sb)->s_journal;
385 EXT4_SB(sb)->s_mount_flags |= EXT4_MF_FS_ABORTED;
387 jbd2_journal_abort(journal, -EIO);
389 if (test_opt(sb, ERRORS_RO)) {
390 ext4_msg(sb, KERN_CRIT, "Remounting filesystem read-only");
392 * Make sure updated value of ->s_mount_flags will be visible
393 * before ->s_flags update
396 sb->s_flags |= MS_RDONLY;
398 if (test_opt(sb, ERRORS_PANIC))
399 panic("EXT4-fs (device %s): panic forced after error\n",
403 #define ext4_error_ratelimit(sb) \
404 ___ratelimit(&(EXT4_SB(sb)->s_err_ratelimit_state), \
407 void __ext4_error(struct super_block *sb, const char *function,
408 unsigned int line, const char *fmt, ...)
410 struct va_format vaf;
413 if (ext4_error_ratelimit(sb)) {
418 "EXT4-fs error (device %s): %s:%d: comm %s: %pV\n",
419 sb->s_id, function, line, current->comm, &vaf);
422 save_error_info(sb, function, line);
423 ext4_handle_error(sb);
426 void __ext4_error_inode(struct inode *inode, const char *function,
427 unsigned int line, ext4_fsblk_t block,
428 const char *fmt, ...)
431 struct va_format vaf;
432 struct ext4_super_block *es = EXT4_SB(inode->i_sb)->s_es;
434 es->s_last_error_ino = cpu_to_le32(inode->i_ino);
435 es->s_last_error_block = cpu_to_le64(block);
436 if (ext4_error_ratelimit(inode->i_sb)) {
441 printk(KERN_CRIT "EXT4-fs error (device %s): %s:%d: "
442 "inode #%lu: block %llu: comm %s: %pV\n",
443 inode->i_sb->s_id, function, line, inode->i_ino,
444 block, current->comm, &vaf);
446 printk(KERN_CRIT "EXT4-fs error (device %s): %s:%d: "
447 "inode #%lu: comm %s: %pV\n",
448 inode->i_sb->s_id, function, line, inode->i_ino,
449 current->comm, &vaf);
452 save_error_info(inode->i_sb, function, line);
453 ext4_handle_error(inode->i_sb);
456 void __ext4_error_file(struct file *file, const char *function,
457 unsigned int line, ext4_fsblk_t block,
458 const char *fmt, ...)
461 struct va_format vaf;
462 struct ext4_super_block *es;
463 struct inode *inode = file_inode(file);
464 char pathname[80], *path;
466 es = EXT4_SB(inode->i_sb)->s_es;
467 es->s_last_error_ino = cpu_to_le32(inode->i_ino);
468 if (ext4_error_ratelimit(inode->i_sb)) {
469 path = d_path(&(file->f_path), pathname, sizeof(pathname));
477 "EXT4-fs error (device %s): %s:%d: inode #%lu: "
478 "block %llu: comm %s: path %s: %pV\n",
479 inode->i_sb->s_id, function, line, inode->i_ino,
480 block, current->comm, path, &vaf);
483 "EXT4-fs error (device %s): %s:%d: inode #%lu: "
484 "comm %s: path %s: %pV\n",
485 inode->i_sb->s_id, function, line, inode->i_ino,
486 current->comm, path, &vaf);
489 save_error_info(inode->i_sb, function, line);
490 ext4_handle_error(inode->i_sb);
493 const char *ext4_decode_error(struct super_block *sb, int errno,
500 errstr = "IO failure";
503 errstr = "Out of memory";
506 if (!sb || (EXT4_SB(sb)->s_journal &&
507 EXT4_SB(sb)->s_journal->j_flags & JBD2_ABORT))
508 errstr = "Journal has aborted";
510 errstr = "Readonly filesystem";
513 /* If the caller passed in an extra buffer for unknown
514 * errors, textualise them now. Else we just return
517 /* Check for truncated error codes... */
518 if (snprintf(nbuf, 16, "error %d", -errno) >= 0)
527 /* __ext4_std_error decodes expected errors from journaling functions
528 * automatically and invokes the appropriate error response. */
530 void __ext4_std_error(struct super_block *sb, const char *function,
531 unsigned int line, int errno)
536 /* Special case: if the error is EROFS, and we're not already
537 * inside a transaction, then there's really no point in logging
539 if (errno == -EROFS && journal_current_handle() == NULL &&
540 (sb->s_flags & MS_RDONLY))
543 if (ext4_error_ratelimit(sb)) {
544 errstr = ext4_decode_error(sb, errno, nbuf);
545 printk(KERN_CRIT "EXT4-fs error (device %s) in %s:%d: %s\n",
546 sb->s_id, function, line, errstr);
549 save_error_info(sb, function, line);
550 ext4_handle_error(sb);
554 * ext4_abort is a much stronger failure handler than ext4_error. The
555 * abort function may be used to deal with unrecoverable failures such
556 * as journal IO errors or ENOMEM at a critical moment in log management.
558 * We unconditionally force the filesystem into an ABORT|READONLY state,
559 * unless the error response on the fs has been set to panic in which
560 * case we take the easy way out and panic immediately.
563 void __ext4_abort(struct super_block *sb, const char *function,
564 unsigned int line, const char *fmt, ...)
568 save_error_info(sb, function, line);
570 printk(KERN_CRIT "EXT4-fs error (device %s): %s:%d: ", sb->s_id,
576 if ((sb->s_flags & MS_RDONLY) == 0) {
577 ext4_msg(sb, KERN_CRIT, "Remounting filesystem read-only");
578 EXT4_SB(sb)->s_mount_flags |= EXT4_MF_FS_ABORTED;
580 * Make sure updated value of ->s_mount_flags will be visible
581 * before ->s_flags update
584 sb->s_flags |= MS_RDONLY;
585 if (EXT4_SB(sb)->s_journal)
586 jbd2_journal_abort(EXT4_SB(sb)->s_journal, -EIO);
587 save_error_info(sb, function, line);
589 if (test_opt(sb, ERRORS_PANIC))
590 panic("EXT4-fs panic from previous error\n");
593 void __ext4_msg(struct super_block *sb,
594 const char *prefix, const char *fmt, ...)
596 struct va_format vaf;
599 if (!___ratelimit(&(EXT4_SB(sb)->s_msg_ratelimit_state), "EXT4-fs"))
605 printk("%sEXT4-fs (%s): %pV\n", prefix, sb->s_id, &vaf);
609 void __ext4_warning(struct super_block *sb, const char *function,
610 unsigned int line, const char *fmt, ...)
612 struct va_format vaf;
615 if (!___ratelimit(&(EXT4_SB(sb)->s_warning_ratelimit_state),
622 printk(KERN_WARNING "EXT4-fs warning (device %s): %s:%d: %pV\n",
623 sb->s_id, function, line, &vaf);
627 void __ext4_grp_locked_error(const char *function, unsigned int line,
628 struct super_block *sb, ext4_group_t grp,
629 unsigned long ino, ext4_fsblk_t block,
630 const char *fmt, ...)
634 struct va_format vaf;
636 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
638 es->s_last_error_ino = cpu_to_le32(ino);
639 es->s_last_error_block = cpu_to_le64(block);
640 __save_error_info(sb, function, line);
642 if (ext4_error_ratelimit(sb)) {
646 printk(KERN_CRIT "EXT4-fs error (device %s): %s:%d: group %u, ",
647 sb->s_id, function, line, grp);
649 printk(KERN_CONT "inode %lu: ", ino);
651 printk(KERN_CONT "block %llu:",
652 (unsigned long long) block);
653 printk(KERN_CONT "%pV\n", &vaf);
657 if (test_opt(sb, ERRORS_CONT)) {
658 ext4_commit_super(sb, 0);
662 ext4_unlock_group(sb, grp);
663 ext4_handle_error(sb);
665 * We only get here in the ERRORS_RO case; relocking the group
666 * may be dangerous, but nothing bad will happen since the
667 * filesystem will have already been marked read/only and the
668 * journal has been aborted. We return 1 as a hint to callers
669 * who might what to use the return value from
670 * ext4_grp_locked_error() to distinguish between the
671 * ERRORS_CONT and ERRORS_RO case, and perhaps return more
672 * aggressively from the ext4 function in question, with a
673 * more appropriate error code.
675 ext4_lock_group(sb, grp);
679 void ext4_update_dynamic_rev(struct super_block *sb)
681 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
683 if (le32_to_cpu(es->s_rev_level) > EXT4_GOOD_OLD_REV)
687 "updating to rev %d because of new feature flag, "
688 "running e2fsck is recommended",
691 es->s_first_ino = cpu_to_le32(EXT4_GOOD_OLD_FIRST_INO);
692 es->s_inode_size = cpu_to_le16(EXT4_GOOD_OLD_INODE_SIZE);
693 es->s_rev_level = cpu_to_le32(EXT4_DYNAMIC_REV);
694 /* leave es->s_feature_*compat flags alone */
695 /* es->s_uuid will be set by e2fsck if empty */
698 * The rest of the superblock fields should be zero, and if not it
699 * means they are likely already in use, so leave them alone. We
700 * can leave it up to e2fsck to clean up any inconsistencies there.
705 * Open the external journal device
707 static struct block_device *ext4_blkdev_get(dev_t dev, struct super_block *sb)
709 struct block_device *bdev;
710 char b[BDEVNAME_SIZE];
712 bdev = blkdev_get_by_dev(dev, FMODE_READ|FMODE_WRITE|FMODE_EXCL, sb);
718 ext4_msg(sb, KERN_ERR, "failed to open journal device %s: %ld",
719 __bdevname(dev, b), PTR_ERR(bdev));
724 * Release the journal device
726 static void ext4_blkdev_put(struct block_device *bdev)
728 blkdev_put(bdev, FMODE_READ|FMODE_WRITE|FMODE_EXCL);
731 static void ext4_blkdev_remove(struct ext4_sb_info *sbi)
733 struct block_device *bdev;
734 bdev = sbi->journal_bdev;
736 ext4_blkdev_put(bdev);
737 sbi->journal_bdev = NULL;
741 static inline struct inode *orphan_list_entry(struct list_head *l)
743 return &list_entry(l, struct ext4_inode_info, i_orphan)->vfs_inode;
746 static void dump_orphan_list(struct super_block *sb, struct ext4_sb_info *sbi)
750 ext4_msg(sb, KERN_ERR, "sb orphan head is %d",
751 le32_to_cpu(sbi->s_es->s_last_orphan));
753 printk(KERN_ERR "sb_info orphan list:\n");
754 list_for_each(l, &sbi->s_orphan) {
755 struct inode *inode = orphan_list_entry(l);
757 "inode %s:%lu at %p: mode %o, nlink %d, next %d\n",
758 inode->i_sb->s_id, inode->i_ino, inode,
759 inode->i_mode, inode->i_nlink,
764 static void ext4_put_super(struct super_block *sb)
766 struct ext4_sb_info *sbi = EXT4_SB(sb);
767 struct ext4_super_block *es = sbi->s_es;
770 ext4_unregister_li_request(sb);
771 dquot_disable(sb, -1, DQUOT_USAGE_ENABLED | DQUOT_LIMITS_ENABLED);
773 flush_workqueue(sbi->rsv_conversion_wq);
774 destroy_workqueue(sbi->rsv_conversion_wq);
776 if (sbi->s_journal) {
777 err = jbd2_journal_destroy(sbi->s_journal);
778 sbi->s_journal = NULL;
780 ext4_abort(sb, "Couldn't clean up the journal");
783 ext4_es_unregister_shrinker(sbi);
784 del_timer_sync(&sbi->s_err_report);
785 ext4_release_system_zone(sb);
787 ext4_ext_release(sb);
788 ext4_xattr_put_super(sb);
790 if (!(sb->s_flags & MS_RDONLY)) {
791 EXT4_CLEAR_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
792 es->s_state = cpu_to_le16(sbi->s_mount_state);
794 if (!(sb->s_flags & MS_RDONLY))
795 ext4_commit_super(sb, 1);
798 remove_proc_entry("options", sbi->s_proc);
799 remove_proc_entry(sb->s_id, ext4_proc_root);
801 kobject_del(&sbi->s_kobj);
803 for (i = 0; i < sbi->s_gdb_count; i++)
804 brelse(sbi->s_group_desc[i]);
805 kvfree(sbi->s_group_desc);
806 kvfree(sbi->s_flex_groups);
807 percpu_counter_destroy(&sbi->s_freeclusters_counter);
808 percpu_counter_destroy(&sbi->s_freeinodes_counter);
809 percpu_counter_destroy(&sbi->s_dirs_counter);
810 percpu_counter_destroy(&sbi->s_dirtyclusters_counter);
813 for (i = 0; i < EXT4_MAXQUOTAS; i++)
814 kfree(sbi->s_qf_names[i]);
817 /* Debugging code just in case the in-memory inode orphan list
818 * isn't empty. The on-disk one can be non-empty if we've
819 * detected an error and taken the fs readonly, but the
820 * in-memory list had better be clean by this point. */
821 if (!list_empty(&sbi->s_orphan))
822 dump_orphan_list(sb, sbi);
823 J_ASSERT(list_empty(&sbi->s_orphan));
825 invalidate_bdev(sb->s_bdev);
826 if (sbi->journal_bdev && sbi->journal_bdev != sb->s_bdev) {
828 * Invalidate the journal device's buffers. We don't want them
829 * floating about in memory - the physical journal device may
830 * hotswapped, and it breaks the `ro-after' testing code.
832 sync_blockdev(sbi->journal_bdev);
833 invalidate_bdev(sbi->journal_bdev);
834 ext4_blkdev_remove(sbi);
836 if (sbi->s_mb_cache) {
837 ext4_xattr_destroy_cache(sbi->s_mb_cache);
838 sbi->s_mb_cache = NULL;
841 kthread_stop(sbi->s_mmp_tsk);
842 sb->s_fs_info = NULL;
844 * Now that we are completely done shutting down the
845 * superblock, we need to actually destroy the kobject.
847 kobject_put(&sbi->s_kobj);
848 wait_for_completion(&sbi->s_kobj_unregister);
849 if (sbi->s_chksum_driver)
850 crypto_free_shash(sbi->s_chksum_driver);
851 kfree(sbi->s_blockgroup_lock);
855 static struct kmem_cache *ext4_inode_cachep;
858 * Called inside transaction, so use GFP_NOFS
860 static struct inode *ext4_alloc_inode(struct super_block *sb)
862 struct ext4_inode_info *ei;
864 ei = kmem_cache_alloc(ext4_inode_cachep, GFP_NOFS);
868 ei->vfs_inode.i_version = 1;
869 spin_lock_init(&ei->i_raw_lock);
870 INIT_LIST_HEAD(&ei->i_prealloc_list);
871 spin_lock_init(&ei->i_prealloc_lock);
872 ext4_es_init_tree(&ei->i_es_tree);
873 rwlock_init(&ei->i_es_lock);
874 INIT_LIST_HEAD(&ei->i_es_list);
877 ei->i_es_shrink_lblk = 0;
878 ei->i_reserved_data_blocks = 0;
879 ei->i_reserved_meta_blocks = 0;
880 ei->i_allocated_meta_blocks = 0;
881 ei->i_da_metadata_calc_len = 0;
882 ei->i_da_metadata_calc_last_lblock = 0;
883 spin_lock_init(&(ei->i_block_reservation_lock));
885 ei->i_reserved_quota = 0;
886 memset(&ei->i_dquot, 0, sizeof(ei->i_dquot));
889 INIT_LIST_HEAD(&ei->i_rsv_conversion_list);
890 spin_lock_init(&ei->i_completed_io_lock);
892 ei->i_datasync_tid = 0;
893 atomic_set(&ei->i_ioend_count, 0);
894 atomic_set(&ei->i_unwritten, 0);
895 INIT_WORK(&ei->i_rsv_conversion_work, ext4_end_io_rsv_work);
897 return &ei->vfs_inode;
900 static int ext4_drop_inode(struct inode *inode)
902 int drop = generic_drop_inode(inode);
904 trace_ext4_drop_inode(inode, drop);
908 static void ext4_i_callback(struct rcu_head *head)
910 struct inode *inode = container_of(head, struct inode, i_rcu);
911 kmem_cache_free(ext4_inode_cachep, EXT4_I(inode));
914 static void ext4_destroy_inode(struct inode *inode)
916 if (!list_empty(&(EXT4_I(inode)->i_orphan))) {
917 ext4_msg(inode->i_sb, KERN_ERR,
918 "Inode %lu (%p): orphan list check failed!",
919 inode->i_ino, EXT4_I(inode));
920 print_hex_dump(KERN_INFO, "", DUMP_PREFIX_ADDRESS, 16, 4,
921 EXT4_I(inode), sizeof(struct ext4_inode_info),
925 call_rcu(&inode->i_rcu, ext4_i_callback);
928 static void init_once(void *foo)
930 struct ext4_inode_info *ei = (struct ext4_inode_info *) foo;
932 INIT_LIST_HEAD(&ei->i_orphan);
933 init_rwsem(&ei->xattr_sem);
934 init_rwsem(&ei->i_data_sem);
935 inode_init_once(&ei->vfs_inode);
938 static int __init init_inodecache(void)
940 ext4_inode_cachep = kmem_cache_create("ext4_inode_cache",
941 sizeof(struct ext4_inode_info),
942 0, (SLAB_RECLAIM_ACCOUNT|
945 if (ext4_inode_cachep == NULL)
950 static void destroy_inodecache(void)
953 * Make sure all delayed rcu free inodes are flushed before we
957 kmem_cache_destroy(ext4_inode_cachep);
960 void ext4_clear_inode(struct inode *inode)
962 invalidate_inode_buffers(inode);
965 ext4_discard_preallocations(inode);
966 ext4_es_remove_extent(inode, 0, EXT_MAX_BLOCKS);
967 if (EXT4_I(inode)->jinode) {
968 jbd2_journal_release_jbd_inode(EXT4_JOURNAL(inode),
969 EXT4_I(inode)->jinode);
970 jbd2_free_inode(EXT4_I(inode)->jinode);
971 EXT4_I(inode)->jinode = NULL;
975 static struct inode *ext4_nfs_get_inode(struct super_block *sb,
976 u64 ino, u32 generation)
980 if (ino < EXT4_FIRST_INO(sb) && ino != EXT4_ROOT_INO)
981 return ERR_PTR(-ESTALE);
982 if (ino > le32_to_cpu(EXT4_SB(sb)->s_es->s_inodes_count))
983 return ERR_PTR(-ESTALE);
985 /* iget isn't really right if the inode is currently unallocated!!
987 * ext4_read_inode will return a bad_inode if the inode had been
988 * deleted, so we should be safe.
990 * Currently we don't know the generation for parent directory, so
991 * a generation of 0 means "accept any"
993 inode = ext4_iget_normal(sb, ino);
995 return ERR_CAST(inode);
996 if (generation && inode->i_generation != generation) {
998 return ERR_PTR(-ESTALE);
1004 static struct dentry *ext4_fh_to_dentry(struct super_block *sb, struct fid *fid,
1005 int fh_len, int fh_type)
1007 return generic_fh_to_dentry(sb, fid, fh_len, fh_type,
1008 ext4_nfs_get_inode);
1011 static struct dentry *ext4_fh_to_parent(struct super_block *sb, struct fid *fid,
1012 int fh_len, int fh_type)
1014 return generic_fh_to_parent(sb, fid, fh_len, fh_type,
1015 ext4_nfs_get_inode);
1019 * Try to release metadata pages (indirect blocks, directories) which are
1020 * mapped via the block device. Since these pages could have journal heads
1021 * which would prevent try_to_free_buffers() from freeing them, we must use
1022 * jbd2 layer's try_to_free_buffers() function to release them.
1024 static int bdev_try_to_free_page(struct super_block *sb, struct page *page,
1027 journal_t *journal = EXT4_SB(sb)->s_journal;
1029 WARN_ON(PageChecked(page));
1030 if (!page_has_buffers(page))
1033 return jbd2_journal_try_to_free_buffers(journal, page,
1034 wait & ~__GFP_WAIT);
1035 return try_to_free_buffers(page);
1039 #define QTYPE2NAME(t) ((t) == USRQUOTA ? "user" : "group")
1040 #define QTYPE2MOPT(on, t) ((t) == USRQUOTA?((on)##USRJQUOTA):((on)##GRPJQUOTA))
1042 static int ext4_write_dquot(struct dquot *dquot);
1043 static int ext4_acquire_dquot(struct dquot *dquot);
1044 static int ext4_release_dquot(struct dquot *dquot);
1045 static int ext4_mark_dquot_dirty(struct dquot *dquot);
1046 static int ext4_write_info(struct super_block *sb, int type);
1047 static int ext4_quota_on(struct super_block *sb, int type, int format_id,
1049 static int ext4_quota_off(struct super_block *sb, int type);
1050 static int ext4_quota_on_mount(struct super_block *sb, int type);
1051 static ssize_t ext4_quota_read(struct super_block *sb, int type, char *data,
1052 size_t len, loff_t off);
1053 static ssize_t ext4_quota_write(struct super_block *sb, int type,
1054 const char *data, size_t len, loff_t off);
1055 static int ext4_quota_enable(struct super_block *sb, int type, int format_id,
1056 unsigned int flags);
1057 static int ext4_enable_quotas(struct super_block *sb);
1059 static struct dquot **ext4_get_dquots(struct inode *inode)
1061 return EXT4_I(inode)->i_dquot;
1064 static const struct dquot_operations ext4_quota_operations = {
1065 .get_reserved_space = ext4_get_reserved_space,
1066 .write_dquot = ext4_write_dquot,
1067 .acquire_dquot = ext4_acquire_dquot,
1068 .release_dquot = ext4_release_dquot,
1069 .mark_dirty = ext4_mark_dquot_dirty,
1070 .write_info = ext4_write_info,
1071 .alloc_dquot = dquot_alloc,
1072 .destroy_dquot = dquot_destroy,
1075 static const struct quotactl_ops ext4_qctl_operations = {
1076 .quota_on = ext4_quota_on,
1077 .quota_off = ext4_quota_off,
1078 .quota_sync = dquot_quota_sync,
1079 .get_info = dquot_get_dqinfo,
1080 .set_info = dquot_set_dqinfo,
1081 .get_dqblk = dquot_get_dqblk,
1082 .set_dqblk = dquot_set_dqblk
1086 static const struct super_operations ext4_sops = {
1087 .alloc_inode = ext4_alloc_inode,
1088 .destroy_inode = ext4_destroy_inode,
1089 .write_inode = ext4_write_inode,
1090 .dirty_inode = ext4_dirty_inode,
1091 .drop_inode = ext4_drop_inode,
1092 .evict_inode = ext4_evict_inode,
1093 .put_super = ext4_put_super,
1094 .sync_fs = ext4_sync_fs,
1095 .freeze_fs = ext4_freeze,
1096 .unfreeze_fs = ext4_unfreeze,
1097 .statfs = ext4_statfs,
1098 .remount_fs = ext4_remount,
1099 .show_options = ext4_show_options,
1101 .quota_read = ext4_quota_read,
1102 .quota_write = ext4_quota_write,
1103 .get_dquots = ext4_get_dquots,
1105 .bdev_try_to_free_page = bdev_try_to_free_page,
1108 static const struct export_operations ext4_export_ops = {
1109 .fh_to_dentry = ext4_fh_to_dentry,
1110 .fh_to_parent = ext4_fh_to_parent,
1111 .get_parent = ext4_get_parent,
1115 Opt_bsd_df, Opt_minix_df, Opt_grpid, Opt_nogrpid,
1116 Opt_resgid, Opt_resuid, Opt_sb, Opt_err_cont, Opt_err_panic, Opt_err_ro,
1117 Opt_nouid32, Opt_debug, Opt_removed,
1118 Opt_user_xattr, Opt_nouser_xattr, Opt_acl, Opt_noacl,
1119 Opt_auto_da_alloc, Opt_noauto_da_alloc, Opt_noload,
1120 Opt_commit, Opt_min_batch_time, Opt_max_batch_time, Opt_journal_dev,
1121 Opt_journal_path, Opt_journal_checksum, Opt_journal_async_commit,
1122 Opt_abort, Opt_data_journal, Opt_data_ordered, Opt_data_writeback,
1123 Opt_data_err_abort, Opt_data_err_ignore,
1124 Opt_usrjquota, Opt_grpjquota, Opt_offusrjquota, Opt_offgrpjquota,
1125 Opt_jqfmt_vfsold, Opt_jqfmt_vfsv0, Opt_jqfmt_vfsv1, Opt_quota,
1126 Opt_noquota, Opt_barrier, Opt_nobarrier, Opt_err,
1127 Opt_usrquota, Opt_grpquota, Opt_i_version,
1128 Opt_stripe, Opt_delalloc, Opt_nodelalloc, Opt_mblk_io_submit,
1129 Opt_nomblk_io_submit, Opt_block_validity, Opt_noblock_validity,
1130 Opt_inode_readahead_blks, Opt_journal_ioprio,
1131 Opt_dioread_nolock, Opt_dioread_lock,
1132 Opt_discard, Opt_nodiscard, Opt_init_itable, Opt_noinit_itable,
1133 Opt_max_dir_size_kb, Opt_nojournal_checksum,
1136 static const match_table_t tokens = {
1137 {Opt_bsd_df, "bsddf"},
1138 {Opt_minix_df, "minixdf"},
1139 {Opt_grpid, "grpid"},
1140 {Opt_grpid, "bsdgroups"},
1141 {Opt_nogrpid, "nogrpid"},
1142 {Opt_nogrpid, "sysvgroups"},
1143 {Opt_resgid, "resgid=%u"},
1144 {Opt_resuid, "resuid=%u"},
1146 {Opt_err_cont, "errors=continue"},
1147 {Opt_err_panic, "errors=panic"},
1148 {Opt_err_ro, "errors=remount-ro"},
1149 {Opt_nouid32, "nouid32"},
1150 {Opt_debug, "debug"},
1151 {Opt_removed, "oldalloc"},
1152 {Opt_removed, "orlov"},
1153 {Opt_user_xattr, "user_xattr"},
1154 {Opt_nouser_xattr, "nouser_xattr"},
1156 {Opt_noacl, "noacl"},
1157 {Opt_noload, "norecovery"},
1158 {Opt_noload, "noload"},
1159 {Opt_removed, "nobh"},
1160 {Opt_removed, "bh"},
1161 {Opt_commit, "commit=%u"},
1162 {Opt_min_batch_time, "min_batch_time=%u"},
1163 {Opt_max_batch_time, "max_batch_time=%u"},
1164 {Opt_journal_dev, "journal_dev=%u"},
1165 {Opt_journal_path, "journal_path=%s"},
1166 {Opt_journal_checksum, "journal_checksum"},
1167 {Opt_nojournal_checksum, "nojournal_checksum"},
1168 {Opt_journal_async_commit, "journal_async_commit"},
1169 {Opt_abort, "abort"},
1170 {Opt_data_journal, "data=journal"},
1171 {Opt_data_ordered, "data=ordered"},
1172 {Opt_data_writeback, "data=writeback"},
1173 {Opt_data_err_abort, "data_err=abort"},
1174 {Opt_data_err_ignore, "data_err=ignore"},
1175 {Opt_offusrjquota, "usrjquota="},
1176 {Opt_usrjquota, "usrjquota=%s"},
1177 {Opt_offgrpjquota, "grpjquota="},
1178 {Opt_grpjquota, "grpjquota=%s"},
1179 {Opt_jqfmt_vfsold, "jqfmt=vfsold"},
1180 {Opt_jqfmt_vfsv0, "jqfmt=vfsv0"},
1181 {Opt_jqfmt_vfsv1, "jqfmt=vfsv1"},
1182 {Opt_grpquota, "grpquota"},
1183 {Opt_noquota, "noquota"},
1184 {Opt_quota, "quota"},
1185 {Opt_usrquota, "usrquota"},
1186 {Opt_barrier, "barrier=%u"},
1187 {Opt_barrier, "barrier"},
1188 {Opt_nobarrier, "nobarrier"},
1189 {Opt_i_version, "i_version"},
1190 {Opt_stripe, "stripe=%u"},
1191 {Opt_delalloc, "delalloc"},
1192 {Opt_nodelalloc, "nodelalloc"},
1193 {Opt_removed, "mblk_io_submit"},
1194 {Opt_removed, "nomblk_io_submit"},
1195 {Opt_block_validity, "block_validity"},
1196 {Opt_noblock_validity, "noblock_validity"},
1197 {Opt_inode_readahead_blks, "inode_readahead_blks=%u"},
1198 {Opt_journal_ioprio, "journal_ioprio=%u"},
1199 {Opt_auto_da_alloc, "auto_da_alloc=%u"},
1200 {Opt_auto_da_alloc, "auto_da_alloc"},
1201 {Opt_noauto_da_alloc, "noauto_da_alloc"},
1202 {Opt_dioread_nolock, "dioread_nolock"},
1203 {Opt_dioread_lock, "dioread_lock"},
1204 {Opt_discard, "discard"},
1205 {Opt_nodiscard, "nodiscard"},
1206 {Opt_init_itable, "init_itable=%u"},
1207 {Opt_init_itable, "init_itable"},
1208 {Opt_noinit_itable, "noinit_itable"},
1209 {Opt_max_dir_size_kb, "max_dir_size_kb=%u"},
1210 {Opt_removed, "check=none"}, /* mount option from ext2/3 */
1211 {Opt_removed, "nocheck"}, /* mount option from ext2/3 */
1212 {Opt_removed, "reservation"}, /* mount option from ext2/3 */
1213 {Opt_removed, "noreservation"}, /* mount option from ext2/3 */
1214 {Opt_removed, "journal=%u"}, /* mount option from ext2/3 */
1218 static ext4_fsblk_t get_sb_block(void **data)
1220 ext4_fsblk_t sb_block;
1221 char *options = (char *) *data;
1223 if (!options || strncmp(options, "sb=", 3) != 0)
1224 return 1; /* Default location */
1227 /* TODO: use simple_strtoll with >32bit ext4 */
1228 sb_block = simple_strtoul(options, &options, 0);
1229 if (*options && *options != ',') {
1230 printk(KERN_ERR "EXT4-fs: Invalid sb specification: %s\n",
1234 if (*options == ',')
1236 *data = (void *) options;
1241 #define DEFAULT_JOURNAL_IOPRIO (IOPRIO_PRIO_VALUE(IOPRIO_CLASS_BE, 3))
1242 static char deprecated_msg[] = "Mount option \"%s\" will be removed by %s\n"
1243 "Contact linux-ext4@vger.kernel.org if you think we should keep it.\n";
1246 static int set_qf_name(struct super_block *sb, int qtype, substring_t *args)
1248 struct ext4_sb_info *sbi = EXT4_SB(sb);
1252 if (sb_any_quota_loaded(sb) &&
1253 !sbi->s_qf_names[qtype]) {
1254 ext4_msg(sb, KERN_ERR,
1255 "Cannot change journaled "
1256 "quota options when quota turned on");
1259 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_QUOTA)) {
1260 ext4_msg(sb, KERN_ERR, "Cannot set journaled quota options "
1261 "when QUOTA feature is enabled");
1264 qname = match_strdup(args);
1266 ext4_msg(sb, KERN_ERR,
1267 "Not enough memory for storing quotafile name");
1270 if (sbi->s_qf_names[qtype]) {
1271 if (strcmp(sbi->s_qf_names[qtype], qname) == 0)
1274 ext4_msg(sb, KERN_ERR,
1275 "%s quota file already specified",
1279 if (strchr(qname, '/')) {
1280 ext4_msg(sb, KERN_ERR,
1281 "quotafile must be on filesystem root");
1284 sbi->s_qf_names[qtype] = qname;
1292 static int clear_qf_name(struct super_block *sb, int qtype)
1295 struct ext4_sb_info *sbi = EXT4_SB(sb);
1297 if (sb_any_quota_loaded(sb) &&
1298 sbi->s_qf_names[qtype]) {
1299 ext4_msg(sb, KERN_ERR, "Cannot change journaled quota options"
1300 " when quota turned on");
1303 kfree(sbi->s_qf_names[qtype]);
1304 sbi->s_qf_names[qtype] = NULL;
1309 #define MOPT_SET 0x0001
1310 #define MOPT_CLEAR 0x0002
1311 #define MOPT_NOSUPPORT 0x0004
1312 #define MOPT_EXPLICIT 0x0008
1313 #define MOPT_CLEAR_ERR 0x0010
1314 #define MOPT_GTE0 0x0020
1317 #define MOPT_QFMT 0x0040
1319 #define MOPT_Q MOPT_NOSUPPORT
1320 #define MOPT_QFMT MOPT_NOSUPPORT
1322 #define MOPT_DATAJ 0x0080
1323 #define MOPT_NO_EXT2 0x0100
1324 #define MOPT_NO_EXT3 0x0200
1325 #define MOPT_EXT4_ONLY (MOPT_NO_EXT2 | MOPT_NO_EXT3)
1326 #define MOPT_STRING 0x0400
1328 static const struct mount_opts {
1332 } ext4_mount_opts[] = {
1333 {Opt_minix_df, EXT4_MOUNT_MINIX_DF, MOPT_SET},
1334 {Opt_bsd_df, EXT4_MOUNT_MINIX_DF, MOPT_CLEAR},
1335 {Opt_grpid, EXT4_MOUNT_GRPID, MOPT_SET},
1336 {Opt_nogrpid, EXT4_MOUNT_GRPID, MOPT_CLEAR},
1337 {Opt_block_validity, EXT4_MOUNT_BLOCK_VALIDITY, MOPT_SET},
1338 {Opt_noblock_validity, EXT4_MOUNT_BLOCK_VALIDITY, MOPT_CLEAR},
1339 {Opt_dioread_nolock, EXT4_MOUNT_DIOREAD_NOLOCK,
1340 MOPT_EXT4_ONLY | MOPT_SET},
1341 {Opt_dioread_lock, EXT4_MOUNT_DIOREAD_NOLOCK,
1342 MOPT_EXT4_ONLY | MOPT_CLEAR},
1343 {Opt_discard, EXT4_MOUNT_DISCARD, MOPT_SET},
1344 {Opt_nodiscard, EXT4_MOUNT_DISCARD, MOPT_CLEAR},
1345 {Opt_delalloc, EXT4_MOUNT_DELALLOC,
1346 MOPT_EXT4_ONLY | MOPT_SET | MOPT_EXPLICIT},
1347 {Opt_nodelalloc, EXT4_MOUNT_DELALLOC,
1348 MOPT_EXT4_ONLY | MOPT_CLEAR},
1349 {Opt_nojournal_checksum, EXT4_MOUNT_JOURNAL_CHECKSUM,
1350 MOPT_EXT4_ONLY | MOPT_CLEAR},
1351 {Opt_journal_checksum, EXT4_MOUNT_JOURNAL_CHECKSUM,
1352 MOPT_EXT4_ONLY | MOPT_SET},
1353 {Opt_journal_async_commit, (EXT4_MOUNT_JOURNAL_ASYNC_COMMIT |
1354 EXT4_MOUNT_JOURNAL_CHECKSUM),
1355 MOPT_EXT4_ONLY | MOPT_SET},
1356 {Opt_noload, EXT4_MOUNT_NOLOAD, MOPT_NO_EXT2 | MOPT_SET},
1357 {Opt_err_panic, EXT4_MOUNT_ERRORS_PANIC, MOPT_SET | MOPT_CLEAR_ERR},
1358 {Opt_err_ro, EXT4_MOUNT_ERRORS_RO, MOPT_SET | MOPT_CLEAR_ERR},
1359 {Opt_err_cont, EXT4_MOUNT_ERRORS_CONT, MOPT_SET | MOPT_CLEAR_ERR},
1360 {Opt_data_err_abort, EXT4_MOUNT_DATA_ERR_ABORT,
1361 MOPT_NO_EXT2 | MOPT_SET},
1362 {Opt_data_err_ignore, EXT4_MOUNT_DATA_ERR_ABORT,
1363 MOPT_NO_EXT2 | MOPT_CLEAR},
1364 {Opt_barrier, EXT4_MOUNT_BARRIER, MOPT_SET},
1365 {Opt_nobarrier, EXT4_MOUNT_BARRIER, MOPT_CLEAR},
1366 {Opt_noauto_da_alloc, EXT4_MOUNT_NO_AUTO_DA_ALLOC, MOPT_SET},
1367 {Opt_auto_da_alloc, EXT4_MOUNT_NO_AUTO_DA_ALLOC, MOPT_CLEAR},
1368 {Opt_noinit_itable, EXT4_MOUNT_INIT_INODE_TABLE, MOPT_CLEAR},
1369 {Opt_commit, 0, MOPT_GTE0},
1370 {Opt_max_batch_time, 0, MOPT_GTE0},
1371 {Opt_min_batch_time, 0, MOPT_GTE0},
1372 {Opt_inode_readahead_blks, 0, MOPT_GTE0},
1373 {Opt_init_itable, 0, MOPT_GTE0},
1374 {Opt_stripe, 0, MOPT_GTE0},
1375 {Opt_resuid, 0, MOPT_GTE0},
1376 {Opt_resgid, 0, MOPT_GTE0},
1377 {Opt_journal_dev, 0, MOPT_GTE0},
1378 {Opt_journal_path, 0, MOPT_STRING},
1379 {Opt_journal_ioprio, 0, MOPT_GTE0},
1380 {Opt_data_journal, EXT4_MOUNT_JOURNAL_DATA, MOPT_NO_EXT2 | MOPT_DATAJ},
1381 {Opt_data_ordered, EXT4_MOUNT_ORDERED_DATA, MOPT_NO_EXT2 | MOPT_DATAJ},
1382 {Opt_data_writeback, EXT4_MOUNT_WRITEBACK_DATA,
1383 MOPT_NO_EXT2 | MOPT_DATAJ},
1384 {Opt_user_xattr, EXT4_MOUNT_XATTR_USER, MOPT_SET},
1385 {Opt_nouser_xattr, EXT4_MOUNT_XATTR_USER, MOPT_CLEAR},
1386 #ifdef CONFIG_EXT4_FS_POSIX_ACL
1387 {Opt_acl, EXT4_MOUNT_POSIX_ACL, MOPT_SET},
1388 {Opt_noacl, EXT4_MOUNT_POSIX_ACL, MOPT_CLEAR},
1390 {Opt_acl, 0, MOPT_NOSUPPORT},
1391 {Opt_noacl, 0, MOPT_NOSUPPORT},
1393 {Opt_nouid32, EXT4_MOUNT_NO_UID32, MOPT_SET},
1394 {Opt_debug, EXT4_MOUNT_DEBUG, MOPT_SET},
1395 {Opt_quota, EXT4_MOUNT_QUOTA | EXT4_MOUNT_USRQUOTA, MOPT_SET | MOPT_Q},
1396 {Opt_usrquota, EXT4_MOUNT_QUOTA | EXT4_MOUNT_USRQUOTA,
1398 {Opt_grpquota, EXT4_MOUNT_QUOTA | EXT4_MOUNT_GRPQUOTA,
1400 {Opt_noquota, (EXT4_MOUNT_QUOTA | EXT4_MOUNT_USRQUOTA |
1401 EXT4_MOUNT_GRPQUOTA), MOPT_CLEAR | MOPT_Q},
1402 {Opt_usrjquota, 0, MOPT_Q},
1403 {Opt_grpjquota, 0, MOPT_Q},
1404 {Opt_offusrjquota, 0, MOPT_Q},
1405 {Opt_offgrpjquota, 0, MOPT_Q},
1406 {Opt_jqfmt_vfsold, QFMT_VFS_OLD, MOPT_QFMT},
1407 {Opt_jqfmt_vfsv0, QFMT_VFS_V0, MOPT_QFMT},
1408 {Opt_jqfmt_vfsv1, QFMT_VFS_V1, MOPT_QFMT},
1409 {Opt_max_dir_size_kb, 0, MOPT_GTE0},
1413 static int handle_mount_opt(struct super_block *sb, char *opt, int token,
1414 substring_t *args, unsigned long *journal_devnum,
1415 unsigned int *journal_ioprio, int is_remount)
1417 struct ext4_sb_info *sbi = EXT4_SB(sb);
1418 const struct mount_opts *m;
1424 if (token == Opt_usrjquota)
1425 return set_qf_name(sb, USRQUOTA, &args[0]);
1426 else if (token == Opt_grpjquota)
1427 return set_qf_name(sb, GRPQUOTA, &args[0]);
1428 else if (token == Opt_offusrjquota)
1429 return clear_qf_name(sb, USRQUOTA);
1430 else if (token == Opt_offgrpjquota)
1431 return clear_qf_name(sb, GRPQUOTA);
1435 case Opt_nouser_xattr:
1436 ext4_msg(sb, KERN_WARNING, deprecated_msg, opt, "3.5");
1439 return 1; /* handled by get_sb_block() */
1441 ext4_msg(sb, KERN_WARNING, "Ignoring removed %s option", opt);
1444 sbi->s_mount_flags |= EXT4_MF_FS_ABORTED;
1447 sb->s_flags |= MS_I_VERSION;
1451 for (m = ext4_mount_opts; m->token != Opt_err; m++)
1452 if (token == m->token)
1455 if (m->token == Opt_err) {
1456 ext4_msg(sb, KERN_ERR, "Unrecognized mount option \"%s\" "
1457 "or missing value", opt);
1461 if ((m->flags & MOPT_NO_EXT2) && IS_EXT2_SB(sb)) {
1462 ext4_msg(sb, KERN_ERR,
1463 "Mount option \"%s\" incompatible with ext2", opt);
1466 if ((m->flags & MOPT_NO_EXT3) && IS_EXT3_SB(sb)) {
1467 ext4_msg(sb, KERN_ERR,
1468 "Mount option \"%s\" incompatible with ext3", opt);
1472 if (args->from && !(m->flags & MOPT_STRING) && match_int(args, &arg))
1474 if (args->from && (m->flags & MOPT_GTE0) && (arg < 0))
1476 if (m->flags & MOPT_EXPLICIT)
1477 set_opt2(sb, EXPLICIT_DELALLOC);
1478 if (m->flags & MOPT_CLEAR_ERR)
1479 clear_opt(sb, ERRORS_MASK);
1480 if (token == Opt_noquota && sb_any_quota_loaded(sb)) {
1481 ext4_msg(sb, KERN_ERR, "Cannot change quota "
1482 "options when quota turned on");
1486 if (m->flags & MOPT_NOSUPPORT) {
1487 ext4_msg(sb, KERN_ERR, "%s option not supported", opt);
1488 } else if (token == Opt_commit) {
1490 arg = JBD2_DEFAULT_MAX_COMMIT_AGE;
1491 sbi->s_commit_interval = HZ * arg;
1492 } else if (token == Opt_max_batch_time) {
1493 sbi->s_max_batch_time = arg;
1494 } else if (token == Opt_min_batch_time) {
1495 sbi->s_min_batch_time = arg;
1496 } else if (token == Opt_inode_readahead_blks) {
1497 if (arg && (arg > (1 << 30) || !is_power_of_2(arg))) {
1498 ext4_msg(sb, KERN_ERR,
1499 "EXT4-fs: inode_readahead_blks must be "
1500 "0 or a power of 2 smaller than 2^31");
1503 sbi->s_inode_readahead_blks = arg;
1504 } else if (token == Opt_init_itable) {
1505 set_opt(sb, INIT_INODE_TABLE);
1507 arg = EXT4_DEF_LI_WAIT_MULT;
1508 sbi->s_li_wait_mult = arg;
1509 } else if (token == Opt_max_dir_size_kb) {
1510 sbi->s_max_dir_size_kb = arg;
1511 } else if (token == Opt_stripe) {
1512 sbi->s_stripe = arg;
1513 } else if (token == Opt_resuid) {
1514 uid = make_kuid(current_user_ns(), arg);
1515 if (!uid_valid(uid)) {
1516 ext4_msg(sb, KERN_ERR, "Invalid uid value %d", arg);
1519 sbi->s_resuid = uid;
1520 } else if (token == Opt_resgid) {
1521 gid = make_kgid(current_user_ns(), arg);
1522 if (!gid_valid(gid)) {
1523 ext4_msg(sb, KERN_ERR, "Invalid gid value %d", arg);
1526 sbi->s_resgid = gid;
1527 } else if (token == Opt_journal_dev) {
1529 ext4_msg(sb, KERN_ERR,
1530 "Cannot specify journal on remount");
1533 *journal_devnum = arg;
1534 } else if (token == Opt_journal_path) {
1536 struct inode *journal_inode;
1541 ext4_msg(sb, KERN_ERR,
1542 "Cannot specify journal on remount");
1545 journal_path = match_strdup(&args[0]);
1546 if (!journal_path) {
1547 ext4_msg(sb, KERN_ERR, "error: could not dup "
1548 "journal device string");
1552 error = kern_path(journal_path, LOOKUP_FOLLOW, &path);
1554 ext4_msg(sb, KERN_ERR, "error: could not find "
1555 "journal device path: error %d", error);
1556 kfree(journal_path);
1560 journal_inode = path.dentry->d_inode;
1561 if (!S_ISBLK(journal_inode->i_mode)) {
1562 ext4_msg(sb, KERN_ERR, "error: journal path %s "
1563 "is not a block device", journal_path);
1565 kfree(journal_path);
1569 *journal_devnum = new_encode_dev(journal_inode->i_rdev);
1571 kfree(journal_path);
1572 } else if (token == Opt_journal_ioprio) {
1574 ext4_msg(sb, KERN_ERR, "Invalid journal IO priority"
1579 IOPRIO_PRIO_VALUE(IOPRIO_CLASS_BE, arg);
1580 } else if (m->flags & MOPT_DATAJ) {
1582 if (!sbi->s_journal)
1583 ext4_msg(sb, KERN_WARNING, "Remounting file system with no journal so ignoring journalled data option");
1584 else if (test_opt(sb, DATA_FLAGS) != m->mount_opt) {
1585 ext4_msg(sb, KERN_ERR,
1586 "Cannot change data mode on remount");
1590 clear_opt(sb, DATA_FLAGS);
1591 sbi->s_mount_opt |= m->mount_opt;
1594 } else if (m->flags & MOPT_QFMT) {
1595 if (sb_any_quota_loaded(sb) &&
1596 sbi->s_jquota_fmt != m->mount_opt) {
1597 ext4_msg(sb, KERN_ERR, "Cannot change journaled "
1598 "quota options when quota turned on");
1601 if (EXT4_HAS_RO_COMPAT_FEATURE(sb,
1602 EXT4_FEATURE_RO_COMPAT_QUOTA)) {
1603 ext4_msg(sb, KERN_ERR,
1604 "Cannot set journaled quota options "
1605 "when QUOTA feature is enabled");
1608 sbi->s_jquota_fmt = m->mount_opt;
1613 if (m->flags & MOPT_CLEAR)
1615 else if (unlikely(!(m->flags & MOPT_SET))) {
1616 ext4_msg(sb, KERN_WARNING,
1617 "buggy handling of option %s", opt);
1622 sbi->s_mount_opt |= m->mount_opt;
1624 sbi->s_mount_opt &= ~m->mount_opt;
1629 static int parse_options(char *options, struct super_block *sb,
1630 unsigned long *journal_devnum,
1631 unsigned int *journal_ioprio,
1634 struct ext4_sb_info *sbi = EXT4_SB(sb);
1636 substring_t args[MAX_OPT_ARGS];
1642 while ((p = strsep(&options, ",")) != NULL) {
1646 * Initialize args struct so we know whether arg was
1647 * found; some options take optional arguments.
1649 args[0].to = args[0].from = NULL;
1650 token = match_token(p, tokens, args);
1651 if (handle_mount_opt(sb, p, token, args, journal_devnum,
1652 journal_ioprio, is_remount) < 0)
1656 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_QUOTA) &&
1657 (test_opt(sb, USRQUOTA) || test_opt(sb, GRPQUOTA))) {
1658 ext4_msg(sb, KERN_ERR, "Cannot set quota options when QUOTA "
1659 "feature is enabled");
1662 if (sbi->s_qf_names[USRQUOTA] || sbi->s_qf_names[GRPQUOTA]) {
1663 if (test_opt(sb, USRQUOTA) && sbi->s_qf_names[USRQUOTA])
1664 clear_opt(sb, USRQUOTA);
1666 if (test_opt(sb, GRPQUOTA) && sbi->s_qf_names[GRPQUOTA])
1667 clear_opt(sb, GRPQUOTA);
1669 if (test_opt(sb, GRPQUOTA) || test_opt(sb, USRQUOTA)) {
1670 ext4_msg(sb, KERN_ERR, "old and new quota "
1675 if (!sbi->s_jquota_fmt) {
1676 ext4_msg(sb, KERN_ERR, "journaled quota format "
1682 if (test_opt(sb, DIOREAD_NOLOCK)) {
1684 BLOCK_SIZE << le32_to_cpu(sbi->s_es->s_log_block_size);
1686 if (blocksize < PAGE_CACHE_SIZE) {
1687 ext4_msg(sb, KERN_ERR, "can't mount with "
1688 "dioread_nolock if block size != PAGE_SIZE");
1692 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA &&
1693 test_opt(sb, JOURNAL_ASYNC_COMMIT)) {
1694 ext4_msg(sb, KERN_ERR, "can't mount with journal_async_commit "
1695 "in data=ordered mode");
1701 static inline void ext4_show_quota_options(struct seq_file *seq,
1702 struct super_block *sb)
1704 #if defined(CONFIG_QUOTA)
1705 struct ext4_sb_info *sbi = EXT4_SB(sb);
1707 if (sbi->s_jquota_fmt) {
1710 switch (sbi->s_jquota_fmt) {
1721 seq_printf(seq, ",jqfmt=%s", fmtname);
1724 if (sbi->s_qf_names[USRQUOTA])
1725 seq_printf(seq, ",usrjquota=%s", sbi->s_qf_names[USRQUOTA]);
1727 if (sbi->s_qf_names[GRPQUOTA])
1728 seq_printf(seq, ",grpjquota=%s", sbi->s_qf_names[GRPQUOTA]);
1732 static const char *token2str(int token)
1734 const struct match_token *t;
1736 for (t = tokens; t->token != Opt_err; t++)
1737 if (t->token == token && !strchr(t->pattern, '='))
1744 * - it's set to a non-default value OR
1745 * - if the per-sb default is different from the global default
1747 static int _ext4_show_options(struct seq_file *seq, struct super_block *sb,
1750 struct ext4_sb_info *sbi = EXT4_SB(sb);
1751 struct ext4_super_block *es = sbi->s_es;
1752 int def_errors, def_mount_opt = nodefs ? 0 : sbi->s_def_mount_opt;
1753 const struct mount_opts *m;
1754 char sep = nodefs ? '\n' : ',';
1756 #define SEQ_OPTS_PUTS(str) seq_printf(seq, "%c" str, sep)
1757 #define SEQ_OPTS_PRINT(str, arg) seq_printf(seq, "%c" str, sep, arg)
1759 if (sbi->s_sb_block != 1)
1760 SEQ_OPTS_PRINT("sb=%llu", sbi->s_sb_block);
1762 for (m = ext4_mount_opts; m->token != Opt_err; m++) {
1763 int want_set = m->flags & MOPT_SET;
1764 if (((m->flags & (MOPT_SET|MOPT_CLEAR)) == 0) ||
1765 (m->flags & MOPT_CLEAR_ERR))
1767 if (!(m->mount_opt & (sbi->s_mount_opt ^ def_mount_opt)))
1768 continue; /* skip if same as the default */
1770 (sbi->s_mount_opt & m->mount_opt) != m->mount_opt) ||
1771 (!want_set && (sbi->s_mount_opt & m->mount_opt)))
1772 continue; /* select Opt_noFoo vs Opt_Foo */
1773 SEQ_OPTS_PRINT("%s", token2str(m->token));
1776 if (nodefs || !uid_eq(sbi->s_resuid, make_kuid(&init_user_ns, EXT4_DEF_RESUID)) ||
1777 le16_to_cpu(es->s_def_resuid) != EXT4_DEF_RESUID)
1778 SEQ_OPTS_PRINT("resuid=%u",
1779 from_kuid_munged(&init_user_ns, sbi->s_resuid));
1780 if (nodefs || !gid_eq(sbi->s_resgid, make_kgid(&init_user_ns, EXT4_DEF_RESGID)) ||
1781 le16_to_cpu(es->s_def_resgid) != EXT4_DEF_RESGID)
1782 SEQ_OPTS_PRINT("resgid=%u",
1783 from_kgid_munged(&init_user_ns, sbi->s_resgid));
1784 def_errors = nodefs ? -1 : le16_to_cpu(es->s_errors);
1785 if (test_opt(sb, ERRORS_RO) && def_errors != EXT4_ERRORS_RO)
1786 SEQ_OPTS_PUTS("errors=remount-ro");
1787 if (test_opt(sb, ERRORS_CONT) && def_errors != EXT4_ERRORS_CONTINUE)
1788 SEQ_OPTS_PUTS("errors=continue");
1789 if (test_opt(sb, ERRORS_PANIC) && def_errors != EXT4_ERRORS_PANIC)
1790 SEQ_OPTS_PUTS("errors=panic");
1791 if (nodefs || sbi->s_commit_interval != JBD2_DEFAULT_MAX_COMMIT_AGE*HZ)
1792 SEQ_OPTS_PRINT("commit=%lu", sbi->s_commit_interval / HZ);
1793 if (nodefs || sbi->s_min_batch_time != EXT4_DEF_MIN_BATCH_TIME)
1794 SEQ_OPTS_PRINT("min_batch_time=%u", sbi->s_min_batch_time);
1795 if (nodefs || sbi->s_max_batch_time != EXT4_DEF_MAX_BATCH_TIME)
1796 SEQ_OPTS_PRINT("max_batch_time=%u", sbi->s_max_batch_time);
1797 if (sb->s_flags & MS_I_VERSION)
1798 SEQ_OPTS_PUTS("i_version");
1799 if (nodefs || sbi->s_stripe)
1800 SEQ_OPTS_PRINT("stripe=%lu", sbi->s_stripe);
1801 if (EXT4_MOUNT_DATA_FLAGS & (sbi->s_mount_opt ^ def_mount_opt)) {
1802 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA)
1803 SEQ_OPTS_PUTS("data=journal");
1804 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA)
1805 SEQ_OPTS_PUTS("data=ordered");
1806 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_WRITEBACK_DATA)
1807 SEQ_OPTS_PUTS("data=writeback");
1810 sbi->s_inode_readahead_blks != EXT4_DEF_INODE_READAHEAD_BLKS)
1811 SEQ_OPTS_PRINT("inode_readahead_blks=%u",
1812 sbi->s_inode_readahead_blks);
1814 if (nodefs || (test_opt(sb, INIT_INODE_TABLE) &&
1815 (sbi->s_li_wait_mult != EXT4_DEF_LI_WAIT_MULT)))
1816 SEQ_OPTS_PRINT("init_itable=%u", sbi->s_li_wait_mult);
1817 if (nodefs || sbi->s_max_dir_size_kb)
1818 SEQ_OPTS_PRINT("max_dir_size_kb=%u", sbi->s_max_dir_size_kb);
1820 ext4_show_quota_options(seq, sb);
1824 static int ext4_show_options(struct seq_file *seq, struct dentry *root)
1826 return _ext4_show_options(seq, root->d_sb, 0);
1829 static int options_seq_show(struct seq_file *seq, void *offset)
1831 struct super_block *sb = seq->private;
1834 seq_puts(seq, (sb->s_flags & MS_RDONLY) ? "ro" : "rw");
1835 rc = _ext4_show_options(seq, sb, 1);
1836 seq_puts(seq, "\n");
1840 static int options_open_fs(struct inode *inode, struct file *file)
1842 return single_open(file, options_seq_show, PDE_DATA(inode));
1845 static const struct file_operations ext4_seq_options_fops = {
1846 .owner = THIS_MODULE,
1847 .open = options_open_fs,
1849 .llseek = seq_lseek,
1850 .release = single_release,
1853 static int ext4_setup_super(struct super_block *sb, struct ext4_super_block *es,
1856 struct ext4_sb_info *sbi = EXT4_SB(sb);
1859 if (le32_to_cpu(es->s_rev_level) > EXT4_MAX_SUPP_REV) {
1860 ext4_msg(sb, KERN_ERR, "revision level too high, "
1861 "forcing read-only mode");
1866 if (!(sbi->s_mount_state & EXT4_VALID_FS))
1867 ext4_msg(sb, KERN_WARNING, "warning: mounting unchecked fs, "
1868 "running e2fsck is recommended");
1869 else if (sbi->s_mount_state & EXT4_ERROR_FS)
1870 ext4_msg(sb, KERN_WARNING,
1871 "warning: mounting fs with errors, "
1872 "running e2fsck is recommended");
1873 else if ((__s16) le16_to_cpu(es->s_max_mnt_count) > 0 &&
1874 le16_to_cpu(es->s_mnt_count) >=
1875 (unsigned short) (__s16) le16_to_cpu(es->s_max_mnt_count))
1876 ext4_msg(sb, KERN_WARNING,
1877 "warning: maximal mount count reached, "
1878 "running e2fsck is recommended");
1879 else if (le32_to_cpu(es->s_checkinterval) &&
1880 (le32_to_cpu(es->s_lastcheck) +
1881 le32_to_cpu(es->s_checkinterval) <= get_seconds()))
1882 ext4_msg(sb, KERN_WARNING,
1883 "warning: checktime reached, "
1884 "running e2fsck is recommended");
1885 if (!sbi->s_journal)
1886 es->s_state &= cpu_to_le16(~EXT4_VALID_FS);
1887 if (!(__s16) le16_to_cpu(es->s_max_mnt_count))
1888 es->s_max_mnt_count = cpu_to_le16(EXT4_DFL_MAX_MNT_COUNT);
1889 le16_add_cpu(&es->s_mnt_count, 1);
1890 es->s_mtime = cpu_to_le32(get_seconds());
1891 ext4_update_dynamic_rev(sb);
1893 EXT4_SET_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
1895 ext4_commit_super(sb, 1);
1897 if (test_opt(sb, DEBUG))
1898 printk(KERN_INFO "[EXT4 FS bs=%lu, gc=%u, "
1899 "bpg=%lu, ipg=%lu, mo=%04x, mo2=%04x]\n",
1901 sbi->s_groups_count,
1902 EXT4_BLOCKS_PER_GROUP(sb),
1903 EXT4_INODES_PER_GROUP(sb),
1904 sbi->s_mount_opt, sbi->s_mount_opt2);
1906 cleancache_init_fs(sb);
1910 int ext4_alloc_flex_bg_array(struct super_block *sb, ext4_group_t ngroup)
1912 struct ext4_sb_info *sbi = EXT4_SB(sb);
1913 struct flex_groups *new_groups;
1916 if (!sbi->s_log_groups_per_flex)
1919 size = ext4_flex_group(sbi, ngroup - 1) + 1;
1920 if (size <= sbi->s_flex_groups_allocated)
1923 size = roundup_pow_of_two(size * sizeof(struct flex_groups));
1924 new_groups = ext4_kvzalloc(size, GFP_KERNEL);
1926 ext4_msg(sb, KERN_ERR, "not enough memory for %d flex groups",
1927 size / (int) sizeof(struct flex_groups));
1931 if (sbi->s_flex_groups) {
1932 memcpy(new_groups, sbi->s_flex_groups,
1933 (sbi->s_flex_groups_allocated *
1934 sizeof(struct flex_groups)));
1935 kvfree(sbi->s_flex_groups);
1937 sbi->s_flex_groups = new_groups;
1938 sbi->s_flex_groups_allocated = size / sizeof(struct flex_groups);
1942 static int ext4_fill_flex_info(struct super_block *sb)
1944 struct ext4_sb_info *sbi = EXT4_SB(sb);
1945 struct ext4_group_desc *gdp = NULL;
1946 ext4_group_t flex_group;
1949 sbi->s_log_groups_per_flex = sbi->s_es->s_log_groups_per_flex;
1950 if (sbi->s_log_groups_per_flex < 1 || sbi->s_log_groups_per_flex > 31) {
1951 sbi->s_log_groups_per_flex = 0;
1955 err = ext4_alloc_flex_bg_array(sb, sbi->s_groups_count);
1959 for (i = 0; i < sbi->s_groups_count; i++) {
1960 gdp = ext4_get_group_desc(sb, i, NULL);
1962 flex_group = ext4_flex_group(sbi, i);
1963 atomic_add(ext4_free_inodes_count(sb, gdp),
1964 &sbi->s_flex_groups[flex_group].free_inodes);
1965 atomic64_add(ext4_free_group_clusters(sb, gdp),
1966 &sbi->s_flex_groups[flex_group].free_clusters);
1967 atomic_add(ext4_used_dirs_count(sb, gdp),
1968 &sbi->s_flex_groups[flex_group].used_dirs);
1976 static __le16 ext4_group_desc_csum(struct ext4_sb_info *sbi, __u32 block_group,
1977 struct ext4_group_desc *gdp)
1981 __le32 le_group = cpu_to_le32(block_group);
1983 if (ext4_has_metadata_csum(sbi->s_sb)) {
1984 /* Use new metadata_csum algorithm */
1988 save_csum = gdp->bg_checksum;
1989 gdp->bg_checksum = 0;
1990 csum32 = ext4_chksum(sbi, sbi->s_csum_seed, (__u8 *)&le_group,
1992 csum32 = ext4_chksum(sbi, csum32, (__u8 *)gdp,
1994 gdp->bg_checksum = save_csum;
1996 crc = csum32 & 0xFFFF;
2000 /* old crc16 code */
2001 if (!(sbi->s_es->s_feature_ro_compat &
2002 cpu_to_le32(EXT4_FEATURE_RO_COMPAT_GDT_CSUM)))
2005 offset = offsetof(struct ext4_group_desc, bg_checksum);
2007 crc = crc16(~0, sbi->s_es->s_uuid, sizeof(sbi->s_es->s_uuid));
2008 crc = crc16(crc, (__u8 *)&le_group, sizeof(le_group));
2009 crc = crc16(crc, (__u8 *)gdp, offset);
2010 offset += sizeof(gdp->bg_checksum); /* skip checksum */
2011 /* for checksum of struct ext4_group_desc do the rest...*/
2012 if ((sbi->s_es->s_feature_incompat &
2013 cpu_to_le32(EXT4_FEATURE_INCOMPAT_64BIT)) &&
2014 offset < le16_to_cpu(sbi->s_es->s_desc_size))
2015 crc = crc16(crc, (__u8 *)gdp + offset,
2016 le16_to_cpu(sbi->s_es->s_desc_size) -
2020 return cpu_to_le16(crc);
2023 int ext4_group_desc_csum_verify(struct super_block *sb, __u32 block_group,
2024 struct ext4_group_desc *gdp)
2026 if (ext4_has_group_desc_csum(sb) &&
2027 (gdp->bg_checksum != ext4_group_desc_csum(EXT4_SB(sb),
2034 void ext4_group_desc_csum_set(struct super_block *sb, __u32 block_group,
2035 struct ext4_group_desc *gdp)
2037 if (!ext4_has_group_desc_csum(sb))
2039 gdp->bg_checksum = ext4_group_desc_csum(EXT4_SB(sb), block_group, gdp);
2042 /* Called at mount-time, super-block is locked */
2043 static int ext4_check_descriptors(struct super_block *sb,
2044 ext4_group_t *first_not_zeroed)
2046 struct ext4_sb_info *sbi = EXT4_SB(sb);
2047 ext4_fsblk_t first_block = le32_to_cpu(sbi->s_es->s_first_data_block);
2048 ext4_fsblk_t last_block;
2049 ext4_fsblk_t block_bitmap;
2050 ext4_fsblk_t inode_bitmap;
2051 ext4_fsblk_t inode_table;
2052 int flexbg_flag = 0;
2053 ext4_group_t i, grp = sbi->s_groups_count;
2055 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_FLEX_BG))
2058 ext4_debug("Checking group descriptors");
2060 for (i = 0; i < sbi->s_groups_count; i++) {
2061 struct ext4_group_desc *gdp = ext4_get_group_desc(sb, i, NULL);
2063 if (i == sbi->s_groups_count - 1 || flexbg_flag)
2064 last_block = ext4_blocks_count(sbi->s_es) - 1;
2066 last_block = first_block +
2067 (EXT4_BLOCKS_PER_GROUP(sb) - 1);
2069 if ((grp == sbi->s_groups_count) &&
2070 !(gdp->bg_flags & cpu_to_le16(EXT4_BG_INODE_ZEROED)))
2073 block_bitmap = ext4_block_bitmap(sb, gdp);
2074 if (block_bitmap < first_block || block_bitmap > last_block) {
2075 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2076 "Block bitmap for group %u not in group "
2077 "(block %llu)!", i, block_bitmap);
2080 inode_bitmap = ext4_inode_bitmap(sb, gdp);
2081 if (inode_bitmap < first_block || inode_bitmap > last_block) {
2082 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2083 "Inode bitmap for group %u not in group "
2084 "(block %llu)!", i, inode_bitmap);
2087 inode_table = ext4_inode_table(sb, gdp);
2088 if (inode_table < first_block ||
2089 inode_table + sbi->s_itb_per_group - 1 > last_block) {
2090 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2091 "Inode table for group %u not in group "
2092 "(block %llu)!", i, inode_table);
2095 ext4_lock_group(sb, i);
2096 if (!ext4_group_desc_csum_verify(sb, i, gdp)) {
2097 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2098 "Checksum for group %u failed (%u!=%u)",
2099 i, le16_to_cpu(ext4_group_desc_csum(sbi, i,
2100 gdp)), le16_to_cpu(gdp->bg_checksum));
2101 if (!(sb->s_flags & MS_RDONLY)) {
2102 ext4_unlock_group(sb, i);
2106 ext4_unlock_group(sb, i);
2108 first_block += EXT4_BLOCKS_PER_GROUP(sb);
2110 if (NULL != first_not_zeroed)
2111 *first_not_zeroed = grp;
2115 /* ext4_orphan_cleanup() walks a singly-linked list of inodes (starting at
2116 * the superblock) which were deleted from all directories, but held open by
2117 * a process at the time of a crash. We walk the list and try to delete these
2118 * inodes at recovery time (only with a read-write filesystem).
2120 * In order to keep the orphan inode chain consistent during traversal (in
2121 * case of crash during recovery), we link each inode into the superblock
2122 * orphan list_head and handle it the same way as an inode deletion during
2123 * normal operation (which journals the operations for us).
2125 * We only do an iget() and an iput() on each inode, which is very safe if we
2126 * accidentally point at an in-use or already deleted inode. The worst that
2127 * can happen in this case is that we get a "bit already cleared" message from
2128 * ext4_free_inode(). The only reason we would point at a wrong inode is if
2129 * e2fsck was run on this filesystem, and it must have already done the orphan
2130 * inode cleanup for us, so we can safely abort without any further action.
2132 static void ext4_orphan_cleanup(struct super_block *sb,
2133 struct ext4_super_block *es)
2135 unsigned int s_flags = sb->s_flags;
2136 int nr_orphans = 0, nr_truncates = 0;
2140 if (!es->s_last_orphan) {
2141 jbd_debug(4, "no orphan inodes to clean up\n");
2145 if (bdev_read_only(sb->s_bdev)) {
2146 ext4_msg(sb, KERN_ERR, "write access "
2147 "unavailable, skipping orphan cleanup");
2151 /* Check if feature set would not allow a r/w mount */
2152 if (!ext4_feature_set_ok(sb, 0)) {
2153 ext4_msg(sb, KERN_INFO, "Skipping orphan cleanup due to "
2154 "unknown ROCOMPAT features");
2158 if (EXT4_SB(sb)->s_mount_state & EXT4_ERROR_FS) {
2159 /* don't clear list on RO mount w/ errors */
2160 if (es->s_last_orphan && !(s_flags & MS_RDONLY)) {
2161 ext4_msg(sb, KERN_INFO, "Errors on filesystem, "
2162 "clearing orphan list.\n");
2163 es->s_last_orphan = 0;
2165 jbd_debug(1, "Skipping orphan recovery on fs with errors.\n");
2169 if (s_flags & MS_RDONLY) {
2170 ext4_msg(sb, KERN_INFO, "orphan cleanup on readonly fs");
2171 sb->s_flags &= ~MS_RDONLY;
2174 /* Needed for iput() to work correctly and not trash data */
2175 sb->s_flags |= MS_ACTIVE;
2176 /* Turn on quotas so that they are updated correctly */
2177 for (i = 0; i < EXT4_MAXQUOTAS; i++) {
2178 if (EXT4_SB(sb)->s_qf_names[i]) {
2179 int ret = ext4_quota_on_mount(sb, i);
2181 ext4_msg(sb, KERN_ERR,
2182 "Cannot turn on journaled "
2183 "quota: error %d", ret);
2188 while (es->s_last_orphan) {
2189 struct inode *inode;
2191 inode = ext4_orphan_get(sb, le32_to_cpu(es->s_last_orphan));
2192 if (IS_ERR(inode)) {
2193 es->s_last_orphan = 0;
2197 list_add(&EXT4_I(inode)->i_orphan, &EXT4_SB(sb)->s_orphan);
2198 dquot_initialize(inode);
2199 if (inode->i_nlink) {
2200 if (test_opt(sb, DEBUG))
2201 ext4_msg(sb, KERN_DEBUG,
2202 "%s: truncating inode %lu to %lld bytes",
2203 __func__, inode->i_ino, inode->i_size);
2204 jbd_debug(2, "truncating inode %lu to %lld bytes\n",
2205 inode->i_ino, inode->i_size);
2206 mutex_lock(&inode->i_mutex);
2207 truncate_inode_pages(inode->i_mapping, inode->i_size);
2208 ext4_truncate(inode);
2209 mutex_unlock(&inode->i_mutex);
2212 if (test_opt(sb, DEBUG))
2213 ext4_msg(sb, KERN_DEBUG,
2214 "%s: deleting unreferenced inode %lu",
2215 __func__, inode->i_ino);
2216 jbd_debug(2, "deleting unreferenced inode %lu\n",
2220 iput(inode); /* The delete magic happens here! */
2223 #define PLURAL(x) (x), ((x) == 1) ? "" : "s"
2226 ext4_msg(sb, KERN_INFO, "%d orphan inode%s deleted",
2227 PLURAL(nr_orphans));
2229 ext4_msg(sb, KERN_INFO, "%d truncate%s cleaned up",
2230 PLURAL(nr_truncates));
2232 /* Turn quotas off */
2233 for (i = 0; i < EXT4_MAXQUOTAS; i++) {
2234 if (sb_dqopt(sb)->files[i])
2235 dquot_quota_off(sb, i);
2238 sb->s_flags = s_flags; /* Restore MS_RDONLY status */
2242 * Maximal extent format file size.
2243 * Resulting logical blkno at s_maxbytes must fit in our on-disk
2244 * extent format containers, within a sector_t, and within i_blocks
2245 * in the vfs. ext4 inode has 48 bits of i_block in fsblock units,
2246 * so that won't be a limiting factor.
2248 * However there is other limiting factor. We do store extents in the form
2249 * of starting block and length, hence the resulting length of the extent
2250 * covering maximum file size must fit into on-disk format containers as
2251 * well. Given that length is always by 1 unit bigger than max unit (because
2252 * we count 0 as well) we have to lower the s_maxbytes by one fs block.
2254 * Note, this does *not* consider any metadata overhead for vfs i_blocks.
2256 static loff_t ext4_max_size(int blkbits, int has_huge_files)
2259 loff_t upper_limit = MAX_LFS_FILESIZE;
2261 /* small i_blocks in vfs inode? */
2262 if (!has_huge_files || sizeof(blkcnt_t) < sizeof(u64)) {
2264 * CONFIG_LBDAF is not enabled implies the inode
2265 * i_block represent total blocks in 512 bytes
2266 * 32 == size of vfs inode i_blocks * 8
2268 upper_limit = (1LL << 32) - 1;
2270 /* total blocks in file system block size */
2271 upper_limit >>= (blkbits - 9);
2272 upper_limit <<= blkbits;
2276 * 32-bit extent-start container, ee_block. We lower the maxbytes
2277 * by one fs block, so ee_len can cover the extent of maximum file
2280 res = (1LL << 32) - 1;
2283 /* Sanity check against vm- & vfs- imposed limits */
2284 if (res > upper_limit)
2291 * Maximal bitmap file size. There is a direct, and {,double-,triple-}indirect
2292 * block limit, and also a limit of (2^48 - 1) 512-byte sectors in i_blocks.
2293 * We need to be 1 filesystem block less than the 2^48 sector limit.
2295 static loff_t ext4_max_bitmap_size(int bits, int has_huge_files)
2297 loff_t res = EXT4_NDIR_BLOCKS;
2300 /* This is calculated to be the largest file size for a dense, block
2301 * mapped file such that the file's total number of 512-byte sectors,
2302 * including data and all indirect blocks, does not exceed (2^48 - 1).
2304 * __u32 i_blocks_lo and _u16 i_blocks_high represent the total
2305 * number of 512-byte sectors of the file.
2308 if (!has_huge_files || sizeof(blkcnt_t) < sizeof(u64)) {
2310 * !has_huge_files or CONFIG_LBDAF not enabled implies that
2311 * the inode i_block field represents total file blocks in
2312 * 2^32 512-byte sectors == size of vfs inode i_blocks * 8
2314 upper_limit = (1LL << 32) - 1;
2316 /* total blocks in file system block size */
2317 upper_limit >>= (bits - 9);
2321 * We use 48 bit ext4_inode i_blocks
2322 * With EXT4_HUGE_FILE_FL set the i_blocks
2323 * represent total number of blocks in
2324 * file system block size
2326 upper_limit = (1LL << 48) - 1;
2330 /* indirect blocks */
2332 /* double indirect blocks */
2333 meta_blocks += 1 + (1LL << (bits-2));
2334 /* tripple indirect blocks */
2335 meta_blocks += 1 + (1LL << (bits-2)) + (1LL << (2*(bits-2)));
2337 upper_limit -= meta_blocks;
2338 upper_limit <<= bits;
2340 res += 1LL << (bits-2);
2341 res += 1LL << (2*(bits-2));
2342 res += 1LL << (3*(bits-2));
2344 if (res > upper_limit)
2347 if (res > MAX_LFS_FILESIZE)
2348 res = MAX_LFS_FILESIZE;
2353 static ext4_fsblk_t descriptor_loc(struct super_block *sb,
2354 ext4_fsblk_t logical_sb_block, int nr)
2356 struct ext4_sb_info *sbi = EXT4_SB(sb);
2357 ext4_group_t bg, first_meta_bg;
2360 first_meta_bg = le32_to_cpu(sbi->s_es->s_first_meta_bg);
2362 if (!EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_META_BG) ||
2364 return logical_sb_block + nr + 1;
2365 bg = sbi->s_desc_per_block * nr;
2366 if (ext4_bg_has_super(sb, bg))
2370 * If we have a meta_bg fs with 1k blocks, group 0's GDT is at
2371 * block 2, not 1. If s_first_data_block == 0 (bigalloc is enabled
2372 * on modern mke2fs or blksize > 1k on older mke2fs) then we must
2375 if (sb->s_blocksize == 1024 && nr == 0 &&
2376 le32_to_cpu(EXT4_SB(sb)->s_es->s_first_data_block) == 0)
2379 return (has_super + ext4_group_first_block_no(sb, bg));
2383 * ext4_get_stripe_size: Get the stripe size.
2384 * @sbi: In memory super block info
2386 * If we have specified it via mount option, then
2387 * use the mount option value. If the value specified at mount time is
2388 * greater than the blocks per group use the super block value.
2389 * If the super block value is greater than blocks per group return 0.
2390 * Allocator needs it be less than blocks per group.
2393 static unsigned long ext4_get_stripe_size(struct ext4_sb_info *sbi)
2395 unsigned long stride = le16_to_cpu(sbi->s_es->s_raid_stride);
2396 unsigned long stripe_width =
2397 le32_to_cpu(sbi->s_es->s_raid_stripe_width);
2400 if (sbi->s_stripe && sbi->s_stripe <= sbi->s_blocks_per_group)
2401 ret = sbi->s_stripe;
2402 else if (stripe_width <= sbi->s_blocks_per_group)
2404 else if (stride <= sbi->s_blocks_per_group)
2410 * If the stripe width is 1, this makes no sense and
2411 * we set it to 0 to turn off stripe handling code.
2422 struct attribute attr;
2423 ssize_t (*show)(struct ext4_attr *, struct ext4_sb_info *, char *);
2424 ssize_t (*store)(struct ext4_attr *, struct ext4_sb_info *,
2425 const char *, size_t);
2432 static int parse_strtoull(const char *buf,
2433 unsigned long long max, unsigned long long *value)
2437 ret = kstrtoull(skip_spaces(buf), 0, value);
2438 if (!ret && *value > max)
2443 static ssize_t delayed_allocation_blocks_show(struct ext4_attr *a,
2444 struct ext4_sb_info *sbi,
2447 return snprintf(buf, PAGE_SIZE, "%llu\n",
2449 percpu_counter_sum(&sbi->s_dirtyclusters_counter)));
2452 static ssize_t session_write_kbytes_show(struct ext4_attr *a,
2453 struct ext4_sb_info *sbi, char *buf)
2455 struct super_block *sb = sbi->s_buddy_cache->i_sb;
2457 if (!sb->s_bdev->bd_part)
2458 return snprintf(buf, PAGE_SIZE, "0\n");
2459 return snprintf(buf, PAGE_SIZE, "%lu\n",
2460 (part_stat_read(sb->s_bdev->bd_part, sectors[1]) -
2461 sbi->s_sectors_written_start) >> 1);
2464 static ssize_t lifetime_write_kbytes_show(struct ext4_attr *a,
2465 struct ext4_sb_info *sbi, char *buf)
2467 struct super_block *sb = sbi->s_buddy_cache->i_sb;
2469 if (!sb->s_bdev->bd_part)
2470 return snprintf(buf, PAGE_SIZE, "0\n");
2471 return snprintf(buf, PAGE_SIZE, "%llu\n",
2472 (unsigned long long)(sbi->s_kbytes_written +
2473 ((part_stat_read(sb->s_bdev->bd_part, sectors[1]) -
2474 EXT4_SB(sb)->s_sectors_written_start) >> 1)));
2477 static ssize_t inode_readahead_blks_store(struct ext4_attr *a,
2478 struct ext4_sb_info *sbi,
2479 const char *buf, size_t count)
2484 ret = kstrtoul(skip_spaces(buf), 0, &t);
2488 if (t && (!is_power_of_2(t) || t > 0x40000000))
2491 sbi->s_inode_readahead_blks = t;
2495 static ssize_t sbi_ui_show(struct ext4_attr *a,
2496 struct ext4_sb_info *sbi, char *buf)
2498 unsigned int *ui = (unsigned int *) (((char *) sbi) + a->u.offset);
2500 return snprintf(buf, PAGE_SIZE, "%u\n", *ui);
2503 static ssize_t sbi_ui_store(struct ext4_attr *a,
2504 struct ext4_sb_info *sbi,
2505 const char *buf, size_t count)
2507 unsigned int *ui = (unsigned int *) (((char *) sbi) + a->u.offset);
2511 ret = kstrtoul(skip_spaces(buf), 0, &t);
2518 static ssize_t es_ui_show(struct ext4_attr *a,
2519 struct ext4_sb_info *sbi, char *buf)
2522 unsigned int *ui = (unsigned int *) (((char *) sbi->s_es) +
2525 return snprintf(buf, PAGE_SIZE, "%u\n", *ui);
2528 static ssize_t reserved_clusters_show(struct ext4_attr *a,
2529 struct ext4_sb_info *sbi, char *buf)
2531 return snprintf(buf, PAGE_SIZE, "%llu\n",
2532 (unsigned long long) atomic64_read(&sbi->s_resv_clusters));
2535 static ssize_t reserved_clusters_store(struct ext4_attr *a,
2536 struct ext4_sb_info *sbi,
2537 const char *buf, size_t count)
2539 unsigned long long val;
2542 if (parse_strtoull(buf, -1ULL, &val))
2544 ret = ext4_reserve_clusters(sbi, val);
2546 return ret ? ret : count;
2549 static ssize_t trigger_test_error(struct ext4_attr *a,
2550 struct ext4_sb_info *sbi,
2551 const char *buf, size_t count)
2555 if (!capable(CAP_SYS_ADMIN))
2558 if (len && buf[len-1] == '\n')
2562 ext4_error(sbi->s_sb, "%.*s", len, buf);
2566 static ssize_t sbi_deprecated_show(struct ext4_attr *a,
2567 struct ext4_sb_info *sbi, char *buf)
2569 return snprintf(buf, PAGE_SIZE, "%d\n", a->u.deprecated_val);
2572 #define EXT4_ATTR_OFFSET(_name,_mode,_show,_store,_elname) \
2573 static struct ext4_attr ext4_attr_##_name = { \
2574 .attr = {.name = __stringify(_name), .mode = _mode }, \
2578 .offset = offsetof(struct ext4_sb_info, _elname),\
2582 #define EXT4_ATTR_OFFSET_ES(_name,_mode,_show,_store,_elname) \
2583 static struct ext4_attr ext4_attr_##_name = { \
2584 .attr = {.name = __stringify(_name), .mode = _mode }, \
2588 .offset = offsetof(struct ext4_super_block, _elname), \
2592 #define EXT4_ATTR(name, mode, show, store) \
2593 static struct ext4_attr ext4_attr_##name = __ATTR(name, mode, show, store)
2595 #define EXT4_INFO_ATTR(name) EXT4_ATTR(name, 0444, NULL, NULL)
2596 #define EXT4_RO_ATTR(name) EXT4_ATTR(name, 0444, name##_show, NULL)
2597 #define EXT4_RW_ATTR(name) EXT4_ATTR(name, 0644, name##_show, name##_store)
2599 #define EXT4_RO_ATTR_ES_UI(name, elname) \
2600 EXT4_ATTR_OFFSET_ES(name, 0444, es_ui_show, NULL, elname)
2601 #define EXT4_RW_ATTR_SBI_UI(name, elname) \
2602 EXT4_ATTR_OFFSET(name, 0644, sbi_ui_show, sbi_ui_store, elname)
2604 #define ATTR_LIST(name) &ext4_attr_##name.attr
2605 #define EXT4_DEPRECATED_ATTR(_name, _val) \
2606 static struct ext4_attr ext4_attr_##_name = { \
2607 .attr = {.name = __stringify(_name), .mode = 0444 }, \
2608 .show = sbi_deprecated_show, \
2610 .deprecated_val = _val, \
2614 EXT4_RO_ATTR(delayed_allocation_blocks);
2615 EXT4_RO_ATTR(session_write_kbytes);
2616 EXT4_RO_ATTR(lifetime_write_kbytes);
2617 EXT4_RW_ATTR(reserved_clusters);
2618 EXT4_ATTR_OFFSET(inode_readahead_blks, 0644, sbi_ui_show,
2619 inode_readahead_blks_store, s_inode_readahead_blks);
2620 EXT4_RW_ATTR_SBI_UI(inode_goal, s_inode_goal);
2621 EXT4_RW_ATTR_SBI_UI(mb_stats, s_mb_stats);
2622 EXT4_RW_ATTR_SBI_UI(mb_max_to_scan, s_mb_max_to_scan);
2623 EXT4_RW_ATTR_SBI_UI(mb_min_to_scan, s_mb_min_to_scan);
2624 EXT4_RW_ATTR_SBI_UI(mb_order2_req, s_mb_order2_reqs);
2625 EXT4_RW_ATTR_SBI_UI(mb_stream_req, s_mb_stream_request);
2626 EXT4_RW_ATTR_SBI_UI(mb_group_prealloc, s_mb_group_prealloc);
2627 EXT4_DEPRECATED_ATTR(max_writeback_mb_bump, 128);
2628 EXT4_RW_ATTR_SBI_UI(extent_max_zeroout_kb, s_extent_max_zeroout_kb);
2629 EXT4_ATTR(trigger_fs_error, 0200, NULL, trigger_test_error);
2630 EXT4_RW_ATTR_SBI_UI(err_ratelimit_interval_ms, s_err_ratelimit_state.interval);
2631 EXT4_RW_ATTR_SBI_UI(err_ratelimit_burst, s_err_ratelimit_state.burst);
2632 EXT4_RW_ATTR_SBI_UI(warning_ratelimit_interval_ms, s_warning_ratelimit_state.interval);
2633 EXT4_RW_ATTR_SBI_UI(warning_ratelimit_burst, s_warning_ratelimit_state.burst);
2634 EXT4_RW_ATTR_SBI_UI(msg_ratelimit_interval_ms, s_msg_ratelimit_state.interval);
2635 EXT4_RW_ATTR_SBI_UI(msg_ratelimit_burst, s_msg_ratelimit_state.burst);
2636 EXT4_RO_ATTR_ES_UI(errors_count, s_error_count);
2637 EXT4_RO_ATTR_ES_UI(first_error_time, s_first_error_time);
2638 EXT4_RO_ATTR_ES_UI(last_error_time, s_last_error_time);
2640 static struct attribute *ext4_attrs[] = {
2641 ATTR_LIST(delayed_allocation_blocks),
2642 ATTR_LIST(session_write_kbytes),
2643 ATTR_LIST(lifetime_write_kbytes),
2644 ATTR_LIST(reserved_clusters),
2645 ATTR_LIST(inode_readahead_blks),
2646 ATTR_LIST(inode_goal),
2647 ATTR_LIST(mb_stats),
2648 ATTR_LIST(mb_max_to_scan),
2649 ATTR_LIST(mb_min_to_scan),
2650 ATTR_LIST(mb_order2_req),
2651 ATTR_LIST(mb_stream_req),
2652 ATTR_LIST(mb_group_prealloc),
2653 ATTR_LIST(max_writeback_mb_bump),
2654 ATTR_LIST(extent_max_zeroout_kb),
2655 ATTR_LIST(trigger_fs_error),
2656 ATTR_LIST(err_ratelimit_interval_ms),
2657 ATTR_LIST(err_ratelimit_burst),
2658 ATTR_LIST(warning_ratelimit_interval_ms),
2659 ATTR_LIST(warning_ratelimit_burst),
2660 ATTR_LIST(msg_ratelimit_interval_ms),
2661 ATTR_LIST(msg_ratelimit_burst),
2662 ATTR_LIST(errors_count),
2663 ATTR_LIST(first_error_time),
2664 ATTR_LIST(last_error_time),
2668 /* Features this copy of ext4 supports */
2669 EXT4_INFO_ATTR(lazy_itable_init);
2670 EXT4_INFO_ATTR(batched_discard);
2671 EXT4_INFO_ATTR(meta_bg_resize);
2673 static struct attribute *ext4_feat_attrs[] = {
2674 ATTR_LIST(lazy_itable_init),
2675 ATTR_LIST(batched_discard),
2676 ATTR_LIST(meta_bg_resize),
2680 static ssize_t ext4_attr_show(struct kobject *kobj,
2681 struct attribute *attr, char *buf)
2683 struct ext4_sb_info *sbi = container_of(kobj, struct ext4_sb_info,
2685 struct ext4_attr *a = container_of(attr, struct ext4_attr, attr);
2687 return a->show ? a->show(a, sbi, buf) : 0;
2690 static ssize_t ext4_attr_store(struct kobject *kobj,
2691 struct attribute *attr,
2692 const char *buf, size_t len)
2694 struct ext4_sb_info *sbi = container_of(kobj, struct ext4_sb_info,
2696 struct ext4_attr *a = container_of(attr, struct ext4_attr, attr);
2698 return a->store ? a->store(a, sbi, buf, len) : 0;
2701 static void ext4_sb_release(struct kobject *kobj)
2703 struct ext4_sb_info *sbi = container_of(kobj, struct ext4_sb_info,
2705 complete(&sbi->s_kobj_unregister);
2708 static const struct sysfs_ops ext4_attr_ops = {
2709 .show = ext4_attr_show,
2710 .store = ext4_attr_store,
2713 static struct kobj_type ext4_ktype = {
2714 .default_attrs = ext4_attrs,
2715 .sysfs_ops = &ext4_attr_ops,
2716 .release = ext4_sb_release,
2719 static void ext4_feat_release(struct kobject *kobj)
2721 complete(&ext4_feat->f_kobj_unregister);
2724 static ssize_t ext4_feat_show(struct kobject *kobj,
2725 struct attribute *attr, char *buf)
2727 return snprintf(buf, PAGE_SIZE, "supported\n");
2731 * We can not use ext4_attr_show/store because it relies on the kobject
2732 * being embedded in the ext4_sb_info structure which is definitely not
2733 * true in this case.
2735 static const struct sysfs_ops ext4_feat_ops = {
2736 .show = ext4_feat_show,
2740 static struct kobj_type ext4_feat_ktype = {
2741 .default_attrs = ext4_feat_attrs,
2742 .sysfs_ops = &ext4_feat_ops,
2743 .release = ext4_feat_release,
2747 * Check whether this filesystem can be mounted based on
2748 * the features present and the RDONLY/RDWR mount requested.
2749 * Returns 1 if this filesystem can be mounted as requested,
2750 * 0 if it cannot be.
2752 static int ext4_feature_set_ok(struct super_block *sb, int readonly)
2754 if (EXT4_HAS_INCOMPAT_FEATURE(sb, ~EXT4_FEATURE_INCOMPAT_SUPP)) {
2755 ext4_msg(sb, KERN_ERR,
2756 "Couldn't mount because of "
2757 "unsupported optional features (%x)",
2758 (le32_to_cpu(EXT4_SB(sb)->s_es->s_feature_incompat) &
2759 ~EXT4_FEATURE_INCOMPAT_SUPP));
2766 /* Check that feature set is OK for a read-write mount */
2767 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, ~EXT4_FEATURE_RO_COMPAT_SUPP)) {
2768 ext4_msg(sb, KERN_ERR, "couldn't mount RDWR because of "
2769 "unsupported optional features (%x)",
2770 (le32_to_cpu(EXT4_SB(sb)->s_es->s_feature_ro_compat) &
2771 ~EXT4_FEATURE_RO_COMPAT_SUPP));
2775 * Large file size enabled file system can only be mounted
2776 * read-write on 32-bit systems if kernel is built with CONFIG_LBDAF
2778 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_HUGE_FILE)) {
2779 if (sizeof(blkcnt_t) < sizeof(u64)) {
2780 ext4_msg(sb, KERN_ERR, "Filesystem with huge files "
2781 "cannot be mounted RDWR without "
2786 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_BIGALLOC) &&
2787 !EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_EXTENTS)) {
2788 ext4_msg(sb, KERN_ERR,
2789 "Can't support bigalloc feature without "
2790 "extents feature\n");
2794 #ifndef CONFIG_QUOTA
2795 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_QUOTA) &&
2797 ext4_msg(sb, KERN_ERR,
2798 "Filesystem with quota feature cannot be mounted RDWR "
2799 "without CONFIG_QUOTA");
2802 #endif /* CONFIG_QUOTA */
2807 * This function is called once a day if we have errors logged
2808 * on the file system
2810 static void print_daily_error_info(unsigned long arg)
2812 struct super_block *sb = (struct super_block *) arg;
2813 struct ext4_sb_info *sbi;
2814 struct ext4_super_block *es;
2819 if (es->s_error_count)
2820 /* fsck newer than v1.41.13 is needed to clean this condition. */
2821 ext4_msg(sb, KERN_NOTICE, "error count since last fsck: %u",
2822 le32_to_cpu(es->s_error_count));
2823 if (es->s_first_error_time) {
2824 printk(KERN_NOTICE "EXT4-fs (%s): initial error at time %u: %.*s:%d",
2825 sb->s_id, le32_to_cpu(es->s_first_error_time),
2826 (int) sizeof(es->s_first_error_func),
2827 es->s_first_error_func,
2828 le32_to_cpu(es->s_first_error_line));
2829 if (es->s_first_error_ino)
2830 printk(": inode %u",
2831 le32_to_cpu(es->s_first_error_ino));
2832 if (es->s_first_error_block)
2833 printk(": block %llu", (unsigned long long)
2834 le64_to_cpu(es->s_first_error_block));
2837 if (es->s_last_error_time) {
2838 printk(KERN_NOTICE "EXT4-fs (%s): last error at time %u: %.*s:%d",
2839 sb->s_id, le32_to_cpu(es->s_last_error_time),
2840 (int) sizeof(es->s_last_error_func),
2841 es->s_last_error_func,
2842 le32_to_cpu(es->s_last_error_line));
2843 if (es->s_last_error_ino)
2844 printk(": inode %u",
2845 le32_to_cpu(es->s_last_error_ino));
2846 if (es->s_last_error_block)
2847 printk(": block %llu", (unsigned long long)
2848 le64_to_cpu(es->s_last_error_block));
2851 mod_timer(&sbi->s_err_report, jiffies + 24*60*60*HZ); /* Once a day */
2854 /* Find next suitable group and run ext4_init_inode_table */
2855 static int ext4_run_li_request(struct ext4_li_request *elr)
2857 struct ext4_group_desc *gdp = NULL;
2858 ext4_group_t group, ngroups;
2859 struct super_block *sb;
2860 unsigned long timeout = 0;
2864 ngroups = EXT4_SB(sb)->s_groups_count;
2867 for (group = elr->lr_next_group; group < ngroups; group++) {
2868 gdp = ext4_get_group_desc(sb, group, NULL);
2874 if (!(gdp->bg_flags & cpu_to_le16(EXT4_BG_INODE_ZEROED)))
2878 if (group >= ngroups)
2883 ret = ext4_init_inode_table(sb, group,
2884 elr->lr_timeout ? 0 : 1);
2885 if (elr->lr_timeout == 0) {
2886 timeout = (jiffies - timeout) *
2887 elr->lr_sbi->s_li_wait_mult;
2888 elr->lr_timeout = timeout;
2890 elr->lr_next_sched = jiffies + elr->lr_timeout;
2891 elr->lr_next_group = group + 1;
2899 * Remove lr_request from the list_request and free the
2900 * request structure. Should be called with li_list_mtx held
2902 static void ext4_remove_li_request(struct ext4_li_request *elr)
2904 struct ext4_sb_info *sbi;
2911 list_del(&elr->lr_request);
2912 sbi->s_li_request = NULL;
2916 static void ext4_unregister_li_request(struct super_block *sb)
2918 mutex_lock(&ext4_li_mtx);
2919 if (!ext4_li_info) {
2920 mutex_unlock(&ext4_li_mtx);
2924 mutex_lock(&ext4_li_info->li_list_mtx);
2925 ext4_remove_li_request(EXT4_SB(sb)->s_li_request);
2926 mutex_unlock(&ext4_li_info->li_list_mtx);
2927 mutex_unlock(&ext4_li_mtx);
2930 static struct task_struct *ext4_lazyinit_task;
2933 * This is the function where ext4lazyinit thread lives. It walks
2934 * through the request list searching for next scheduled filesystem.
2935 * When such a fs is found, run the lazy initialization request
2936 * (ext4_rn_li_request) and keep track of the time spend in this
2937 * function. Based on that time we compute next schedule time of
2938 * the request. When walking through the list is complete, compute
2939 * next waking time and put itself into sleep.
2941 static int ext4_lazyinit_thread(void *arg)
2943 struct ext4_lazy_init *eli = (struct ext4_lazy_init *)arg;
2944 struct list_head *pos, *n;
2945 struct ext4_li_request *elr;
2946 unsigned long next_wakeup, cur;
2948 BUG_ON(NULL == eli);
2952 next_wakeup = MAX_JIFFY_OFFSET;
2954 mutex_lock(&eli->li_list_mtx);
2955 if (list_empty(&eli->li_request_list)) {
2956 mutex_unlock(&eli->li_list_mtx);
2960 list_for_each_safe(pos, n, &eli->li_request_list) {
2961 elr = list_entry(pos, struct ext4_li_request,
2964 if (time_after_eq(jiffies, elr->lr_next_sched)) {
2965 if (ext4_run_li_request(elr) != 0) {
2966 /* error, remove the lazy_init job */
2967 ext4_remove_li_request(elr);
2972 if (time_before(elr->lr_next_sched, next_wakeup))
2973 next_wakeup = elr->lr_next_sched;
2975 mutex_unlock(&eli->li_list_mtx);
2980 if ((time_after_eq(cur, next_wakeup)) ||
2981 (MAX_JIFFY_OFFSET == next_wakeup)) {
2986 schedule_timeout_interruptible(next_wakeup - cur);
2988 if (kthread_should_stop()) {
2989 ext4_clear_request_list();
2996 * It looks like the request list is empty, but we need
2997 * to check it under the li_list_mtx lock, to prevent any
2998 * additions into it, and of course we should lock ext4_li_mtx
2999 * to atomically free the list and ext4_li_info, because at
3000 * this point another ext4 filesystem could be registering
3003 mutex_lock(&ext4_li_mtx);
3004 mutex_lock(&eli->li_list_mtx);
3005 if (!list_empty(&eli->li_request_list)) {
3006 mutex_unlock(&eli->li_list_mtx);
3007 mutex_unlock(&ext4_li_mtx);
3010 mutex_unlock(&eli->li_list_mtx);
3011 kfree(ext4_li_info);
3012 ext4_li_info = NULL;
3013 mutex_unlock(&ext4_li_mtx);
3018 static void ext4_clear_request_list(void)
3020 struct list_head *pos, *n;
3021 struct ext4_li_request *elr;
3023 mutex_lock(&ext4_li_info->li_list_mtx);
3024 list_for_each_safe(pos, n, &ext4_li_info->li_request_list) {
3025 elr = list_entry(pos, struct ext4_li_request,
3027 ext4_remove_li_request(elr);
3029 mutex_unlock(&ext4_li_info->li_list_mtx);
3032 static int ext4_run_lazyinit_thread(void)
3034 ext4_lazyinit_task = kthread_run(ext4_lazyinit_thread,
3035 ext4_li_info, "ext4lazyinit");
3036 if (IS_ERR(ext4_lazyinit_task)) {
3037 int err = PTR_ERR(ext4_lazyinit_task);
3038 ext4_clear_request_list();
3039 kfree(ext4_li_info);
3040 ext4_li_info = NULL;
3041 printk(KERN_CRIT "EXT4-fs: error %d creating inode table "
3042 "initialization thread\n",
3046 ext4_li_info->li_state |= EXT4_LAZYINIT_RUNNING;
3051 * Check whether it make sense to run itable init. thread or not.
3052 * If there is at least one uninitialized inode table, return
3053 * corresponding group number, else the loop goes through all
3054 * groups and return total number of groups.
3056 static ext4_group_t ext4_has_uninit_itable(struct super_block *sb)
3058 ext4_group_t group, ngroups = EXT4_SB(sb)->s_groups_count;
3059 struct ext4_group_desc *gdp = NULL;
3061 for (group = 0; group < ngroups; group++) {
3062 gdp = ext4_get_group_desc(sb, group, NULL);
3066 if (!(gdp->bg_flags & cpu_to_le16(EXT4_BG_INODE_ZEROED)))
3073 static int ext4_li_info_new(void)
3075 struct ext4_lazy_init *eli = NULL;
3077 eli = kzalloc(sizeof(*eli), GFP_KERNEL);
3081 INIT_LIST_HEAD(&eli->li_request_list);
3082 mutex_init(&eli->li_list_mtx);
3084 eli->li_state |= EXT4_LAZYINIT_QUIT;
3091 static struct ext4_li_request *ext4_li_request_new(struct super_block *sb,
3094 struct ext4_sb_info *sbi = EXT4_SB(sb);
3095 struct ext4_li_request *elr;
3097 elr = kzalloc(sizeof(*elr), GFP_KERNEL);
3103 elr->lr_next_group = start;
3106 * Randomize first schedule time of the request to
3107 * spread the inode table initialization requests
3110 elr->lr_next_sched = jiffies + (prandom_u32() %
3111 (EXT4_DEF_LI_MAX_START_DELAY * HZ));
3115 int ext4_register_li_request(struct super_block *sb,
3116 ext4_group_t first_not_zeroed)
3118 struct ext4_sb_info *sbi = EXT4_SB(sb);
3119 struct ext4_li_request *elr = NULL;
3120 ext4_group_t ngroups = EXT4_SB(sb)->s_groups_count;
3123 mutex_lock(&ext4_li_mtx);
3124 if (sbi->s_li_request != NULL) {
3126 * Reset timeout so it can be computed again, because
3127 * s_li_wait_mult might have changed.
3129 sbi->s_li_request->lr_timeout = 0;
3133 if (first_not_zeroed == ngroups ||
3134 (sb->s_flags & MS_RDONLY) ||
3135 !test_opt(sb, INIT_INODE_TABLE))
3138 elr = ext4_li_request_new(sb, first_not_zeroed);
3144 if (NULL == ext4_li_info) {
3145 ret = ext4_li_info_new();
3150 mutex_lock(&ext4_li_info->li_list_mtx);
3151 list_add(&elr->lr_request, &ext4_li_info->li_request_list);
3152 mutex_unlock(&ext4_li_info->li_list_mtx);
3154 sbi->s_li_request = elr;
3156 * set elr to NULL here since it has been inserted to
3157 * the request_list and the removal and free of it is
3158 * handled by ext4_clear_request_list from now on.
3162 if (!(ext4_li_info->li_state & EXT4_LAZYINIT_RUNNING)) {
3163 ret = ext4_run_lazyinit_thread();
3168 mutex_unlock(&ext4_li_mtx);
3175 * We do not need to lock anything since this is called on
3178 static void ext4_destroy_lazyinit_thread(void)
3181 * If thread exited earlier
3182 * there's nothing to be done.
3184 if (!ext4_li_info || !ext4_lazyinit_task)
3187 kthread_stop(ext4_lazyinit_task);
3190 static int set_journal_csum_feature_set(struct super_block *sb)
3193 int compat, incompat;
3194 struct ext4_sb_info *sbi = EXT4_SB(sb);
3196 if (ext4_has_metadata_csum(sb)) {
3197 /* journal checksum v3 */
3199 incompat = JBD2_FEATURE_INCOMPAT_CSUM_V3;
3201 /* journal checksum v1 */
3202 compat = JBD2_FEATURE_COMPAT_CHECKSUM;
3206 jbd2_journal_clear_features(sbi->s_journal,
3207 JBD2_FEATURE_COMPAT_CHECKSUM, 0,
3208 JBD2_FEATURE_INCOMPAT_CSUM_V3 |
3209 JBD2_FEATURE_INCOMPAT_CSUM_V2);
3210 if (test_opt(sb, JOURNAL_ASYNC_COMMIT)) {
3211 ret = jbd2_journal_set_features(sbi->s_journal,
3213 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT |
3215 } else if (test_opt(sb, JOURNAL_CHECKSUM)) {
3216 ret = jbd2_journal_set_features(sbi->s_journal,
3219 jbd2_journal_clear_features(sbi->s_journal, 0, 0,
3220 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT);
3222 jbd2_journal_clear_features(sbi->s_journal, 0, 0,
3223 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT);
3230 * Note: calculating the overhead so we can be compatible with
3231 * historical BSD practice is quite difficult in the face of
3232 * clusters/bigalloc. This is because multiple metadata blocks from
3233 * different block group can end up in the same allocation cluster.
3234 * Calculating the exact overhead in the face of clustered allocation
3235 * requires either O(all block bitmaps) in memory or O(number of block
3236 * groups**2) in time. We will still calculate the superblock for
3237 * older file systems --- and if we come across with a bigalloc file
3238 * system with zero in s_overhead_clusters the estimate will be close to
3239 * correct especially for very large cluster sizes --- but for newer
3240 * file systems, it's better to calculate this figure once at mkfs
3241 * time, and store it in the superblock. If the superblock value is
3242 * present (even for non-bigalloc file systems), we will use it.
3244 static int count_overhead(struct super_block *sb, ext4_group_t grp,
3247 struct ext4_sb_info *sbi = EXT4_SB(sb);
3248 struct ext4_group_desc *gdp;
3249 ext4_fsblk_t first_block, last_block, b;
3250 ext4_group_t i, ngroups = ext4_get_groups_count(sb);
3251 int s, j, count = 0;
3253 if (!EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_BIGALLOC))
3254 return (ext4_bg_has_super(sb, grp) + ext4_bg_num_gdb(sb, grp) +
3255 sbi->s_itb_per_group + 2);
3257 first_block = le32_to_cpu(sbi->s_es->s_first_data_block) +
3258 (grp * EXT4_BLOCKS_PER_GROUP(sb));
3259 last_block = first_block + EXT4_BLOCKS_PER_GROUP(sb) - 1;
3260 for (i = 0; i < ngroups; i++) {
3261 gdp = ext4_get_group_desc(sb, i, NULL);
3262 b = ext4_block_bitmap(sb, gdp);
3263 if (b >= first_block && b <= last_block) {
3264 ext4_set_bit(EXT4_B2C(sbi, b - first_block), buf);
3267 b = ext4_inode_bitmap(sb, gdp);
3268 if (b >= first_block && b <= last_block) {
3269 ext4_set_bit(EXT4_B2C(sbi, b - first_block), buf);
3272 b = ext4_inode_table(sb, gdp);
3273 if (b >= first_block && b + sbi->s_itb_per_group <= last_block)
3274 for (j = 0; j < sbi->s_itb_per_group; j++, b++) {
3275 int c = EXT4_B2C(sbi, b - first_block);
3276 ext4_set_bit(c, buf);
3282 if (ext4_bg_has_super(sb, grp)) {
3283 ext4_set_bit(s++, buf);
3286 for (j = ext4_bg_num_gdb(sb, grp); j > 0; j--) {
3287 ext4_set_bit(EXT4_B2C(sbi, s++), buf);
3293 return EXT4_CLUSTERS_PER_GROUP(sb) -
3294 ext4_count_free(buf, EXT4_CLUSTERS_PER_GROUP(sb) / 8);
3298 * Compute the overhead and stash it in sbi->s_overhead
3300 int ext4_calculate_overhead(struct super_block *sb)
3302 struct ext4_sb_info *sbi = EXT4_SB(sb);
3303 struct ext4_super_block *es = sbi->s_es;
3304 ext4_group_t i, ngroups = ext4_get_groups_count(sb);
3305 ext4_fsblk_t overhead = 0;
3306 char *buf = (char *) get_zeroed_page(GFP_NOFS);
3312 * Compute the overhead (FS structures). This is constant
3313 * for a given filesystem unless the number of block groups
3314 * changes so we cache the previous value until it does.
3318 * All of the blocks before first_data_block are overhead
3320 overhead = EXT4_B2C(sbi, le32_to_cpu(es->s_first_data_block));
3323 * Add the overhead found in each block group
3325 for (i = 0; i < ngroups; i++) {
3328 blks = count_overhead(sb, i, buf);
3331 memset(buf, 0, PAGE_SIZE);
3334 /* Add the internal journal blocks as well */
3335 if (sbi->s_journal && !sbi->journal_bdev)
3336 overhead += EXT4_NUM_B2C(sbi, sbi->s_journal->j_maxlen);
3338 sbi->s_overhead = overhead;
3340 free_page((unsigned long) buf);
3345 static ext4_fsblk_t ext4_calculate_resv_clusters(struct super_block *sb)
3347 ext4_fsblk_t resv_clusters;
3350 * There's no need to reserve anything when we aren't using extents.
3351 * The space estimates are exact, there are no unwritten extents,
3352 * hole punching doesn't need new metadata... This is needed especially
3353 * to keep ext2/3 backward compatibility.
3355 if (!EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_EXTENTS))
3358 * By default we reserve 2% or 4096 clusters, whichever is smaller.
3359 * This should cover the situations where we can not afford to run
3360 * out of space like for example punch hole, or converting
3361 * unwritten extents in delalloc path. In most cases such
3362 * allocation would require 1, or 2 blocks, higher numbers are
3365 resv_clusters = ext4_blocks_count(EXT4_SB(sb)->s_es) >>
3366 EXT4_SB(sb)->s_cluster_bits;
3368 do_div(resv_clusters, 50);
3369 resv_clusters = min_t(ext4_fsblk_t, resv_clusters, 4096);
3371 return resv_clusters;
3375 static int ext4_reserve_clusters(struct ext4_sb_info *sbi, ext4_fsblk_t count)
3377 ext4_fsblk_t clusters = ext4_blocks_count(sbi->s_es) >>
3378 sbi->s_cluster_bits;
3380 if (count >= clusters)
3383 atomic64_set(&sbi->s_resv_clusters, count);
3387 static int ext4_fill_super(struct super_block *sb, void *data, int silent)
3389 char *orig_data = kstrdup(data, GFP_KERNEL);
3390 struct buffer_head *bh;
3391 struct ext4_super_block *es = NULL;
3392 struct ext4_sb_info *sbi;
3394 ext4_fsblk_t sb_block = get_sb_block(&data);
3395 ext4_fsblk_t logical_sb_block;
3396 unsigned long offset = 0;
3397 unsigned long journal_devnum = 0;
3398 unsigned long def_mount_opts;
3403 int blocksize, clustersize;
3404 unsigned int db_count;
3406 int needs_recovery, has_huge_files, has_bigalloc;
3409 unsigned int journal_ioprio = DEFAULT_JOURNAL_IOPRIO;
3410 ext4_group_t first_not_zeroed;
3412 sbi = kzalloc(sizeof(*sbi), GFP_KERNEL);
3416 sbi->s_blockgroup_lock =
3417 kzalloc(sizeof(struct blockgroup_lock), GFP_KERNEL);
3418 if (!sbi->s_blockgroup_lock) {
3422 sb->s_fs_info = sbi;
3424 sbi->s_inode_readahead_blks = EXT4_DEF_INODE_READAHEAD_BLKS;
3425 sbi->s_sb_block = sb_block;
3426 if (sb->s_bdev->bd_part)
3427 sbi->s_sectors_written_start =
3428 part_stat_read(sb->s_bdev->bd_part, sectors[1]);
3430 /* Cleanup superblock name */
3431 for (cp = sb->s_id; (cp = strchr(cp, '/'));)
3434 /* -EINVAL is default */
3436 blocksize = sb_min_blocksize(sb, EXT4_MIN_BLOCK_SIZE);
3438 ext4_msg(sb, KERN_ERR, "unable to set blocksize");
3443 * The ext4 superblock will not be buffer aligned for other than 1kB
3444 * block sizes. We need to calculate the offset from buffer start.
3446 if (blocksize != EXT4_MIN_BLOCK_SIZE) {
3447 logical_sb_block = sb_block * EXT4_MIN_BLOCK_SIZE;
3448 offset = do_div(logical_sb_block, blocksize);
3450 logical_sb_block = sb_block;
3453 if (!(bh = sb_bread_unmovable(sb, logical_sb_block))) {
3454 ext4_msg(sb, KERN_ERR, "unable to read superblock");
3458 * Note: s_es must be initialized as soon as possible because
3459 * some ext4 macro-instructions depend on its value
3461 es = (struct ext4_super_block *) (bh->b_data + offset);
3463 sb->s_magic = le16_to_cpu(es->s_magic);
3464 if (sb->s_magic != EXT4_SUPER_MAGIC)
3466 sbi->s_kbytes_written = le64_to_cpu(es->s_kbytes_written);
3468 /* Warn if metadata_csum and gdt_csum are both set. */
3469 if (EXT4_HAS_RO_COMPAT_FEATURE(sb,
3470 EXT4_FEATURE_RO_COMPAT_METADATA_CSUM) &&
3471 EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_GDT_CSUM))
3472 ext4_warning(sb, "metadata_csum and uninit_bg are "
3473 "redundant flags; please run fsck.");
3475 /* Check for a known checksum algorithm */
3476 if (!ext4_verify_csum_type(sb, es)) {
3477 ext4_msg(sb, KERN_ERR, "VFS: Found ext4 filesystem with "
3478 "unknown checksum algorithm.");
3483 /* Load the checksum driver */
3484 if (EXT4_HAS_RO_COMPAT_FEATURE(sb,
3485 EXT4_FEATURE_RO_COMPAT_METADATA_CSUM)) {
3486 sbi->s_chksum_driver = crypto_alloc_shash("crc32c", 0, 0);
3487 if (IS_ERR(sbi->s_chksum_driver)) {
3488 ext4_msg(sb, KERN_ERR, "Cannot load crc32c driver.");
3489 ret = PTR_ERR(sbi->s_chksum_driver);
3490 sbi->s_chksum_driver = NULL;
3495 /* Check superblock checksum */
3496 if (!ext4_superblock_csum_verify(sb, es)) {
3497 ext4_msg(sb, KERN_ERR, "VFS: Found ext4 filesystem with "
3498 "invalid superblock checksum. Run e2fsck?");
3503 /* Precompute checksum seed for all metadata */
3504 if (ext4_has_metadata_csum(sb))
3505 sbi->s_csum_seed = ext4_chksum(sbi, ~0, es->s_uuid,
3506 sizeof(es->s_uuid));
3508 /* Set defaults before we parse the mount options */
3509 def_mount_opts = le32_to_cpu(es->s_default_mount_opts);
3510 set_opt(sb, INIT_INODE_TABLE);
3511 if (def_mount_opts & EXT4_DEFM_DEBUG)
3513 if (def_mount_opts & EXT4_DEFM_BSDGROUPS)
3515 if (def_mount_opts & EXT4_DEFM_UID16)
3516 set_opt(sb, NO_UID32);
3517 /* xattr user namespace & acls are now defaulted on */
3518 set_opt(sb, XATTR_USER);
3519 #ifdef CONFIG_EXT4_FS_POSIX_ACL
3520 set_opt(sb, POSIX_ACL);
3522 /* don't forget to enable journal_csum when metadata_csum is enabled. */
3523 if (ext4_has_metadata_csum(sb))
3524 set_opt(sb, JOURNAL_CHECKSUM);
3526 if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_DATA)
3527 set_opt(sb, JOURNAL_DATA);
3528 else if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_ORDERED)
3529 set_opt(sb, ORDERED_DATA);
3530 else if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_WBACK)
3531 set_opt(sb, WRITEBACK_DATA);
3533 if (le16_to_cpu(sbi->s_es->s_errors) == EXT4_ERRORS_PANIC)
3534 set_opt(sb, ERRORS_PANIC);
3535 else if (le16_to_cpu(sbi->s_es->s_errors) == EXT4_ERRORS_CONTINUE)
3536 set_opt(sb, ERRORS_CONT);
3538 set_opt(sb, ERRORS_RO);
3539 /* block_validity enabled by default; disable with noblock_validity */
3540 set_opt(sb, BLOCK_VALIDITY);
3541 if (def_mount_opts & EXT4_DEFM_DISCARD)
3542 set_opt(sb, DISCARD);
3544 sbi->s_resuid = make_kuid(&init_user_ns, le16_to_cpu(es->s_def_resuid));
3545 sbi->s_resgid = make_kgid(&init_user_ns, le16_to_cpu(es->s_def_resgid));
3546 sbi->s_commit_interval = JBD2_DEFAULT_MAX_COMMIT_AGE * HZ;
3547 sbi->s_min_batch_time = EXT4_DEF_MIN_BATCH_TIME;
3548 sbi->s_max_batch_time = EXT4_DEF_MAX_BATCH_TIME;
3550 if ((def_mount_opts & EXT4_DEFM_NOBARRIER) == 0)
3551 set_opt(sb, BARRIER);
3554 * enable delayed allocation by default
3555 * Use -o nodelalloc to turn it off
3557 if (!IS_EXT3_SB(sb) && !IS_EXT2_SB(sb) &&
3558 ((def_mount_opts & EXT4_DEFM_NODELALLOC) == 0))
3559 set_opt(sb, DELALLOC);
3562 * set default s_li_wait_mult for lazyinit, for the case there is
3563 * no mount option specified.
3565 sbi->s_li_wait_mult = EXT4_DEF_LI_WAIT_MULT;
3567 if (!parse_options((char *) sbi->s_es->s_mount_opts, sb,
3568 &journal_devnum, &journal_ioprio, 0)) {
3569 ext4_msg(sb, KERN_WARNING,
3570 "failed to parse options in superblock: %s",
3571 sbi->s_es->s_mount_opts);
3573 sbi->s_def_mount_opt = sbi->s_mount_opt;
3574 if (!parse_options((char *) data, sb, &journal_devnum,
3575 &journal_ioprio, 0))
3578 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA) {
3579 printk_once(KERN_WARNING "EXT4-fs: Warning: mounting "
3580 "with data=journal disables delayed "
3581 "allocation and O_DIRECT support!\n");
3582 if (test_opt2(sb, EXPLICIT_DELALLOC)) {
3583 ext4_msg(sb, KERN_ERR, "can't mount with "
3584 "both data=journal and delalloc");
3587 if (test_opt(sb, DIOREAD_NOLOCK)) {
3588 ext4_msg(sb, KERN_ERR, "can't mount with "
3589 "both data=journal and dioread_nolock");
3592 if (test_opt(sb, DELALLOC))
3593 clear_opt(sb, DELALLOC);
3596 sb->s_flags = (sb->s_flags & ~MS_POSIXACL) |
3597 (test_opt(sb, POSIX_ACL) ? MS_POSIXACL : 0);
3599 if (le32_to_cpu(es->s_rev_level) == EXT4_GOOD_OLD_REV &&
3600 (EXT4_HAS_COMPAT_FEATURE(sb, ~0U) ||
3601 EXT4_HAS_RO_COMPAT_FEATURE(sb, ~0U) ||
3602 EXT4_HAS_INCOMPAT_FEATURE(sb, ~0U)))
3603 ext4_msg(sb, KERN_WARNING,
3604 "feature flags set on rev 0 fs, "
3605 "running e2fsck is recommended");
3607 if (es->s_creator_os == cpu_to_le32(EXT4_OS_HURD)) {
3608 set_opt2(sb, HURD_COMPAT);
3609 if (EXT4_HAS_INCOMPAT_FEATURE(sb,
3610 EXT4_FEATURE_INCOMPAT_64BIT)) {
3611 ext4_msg(sb, KERN_ERR,
3612 "The Hurd can't support 64-bit file systems");
3617 if (IS_EXT2_SB(sb)) {
3618 if (ext2_feature_set_ok(sb))
3619 ext4_msg(sb, KERN_INFO, "mounting ext2 file system "
3620 "using the ext4 subsystem");
3622 ext4_msg(sb, KERN_ERR, "couldn't mount as ext2 due "
3623 "to feature incompatibilities");
3628 if (IS_EXT3_SB(sb)) {
3629 if (ext3_feature_set_ok(sb))
3630 ext4_msg(sb, KERN_INFO, "mounting ext3 file system "
3631 "using the ext4 subsystem");
3633 ext4_msg(sb, KERN_ERR, "couldn't mount as ext3 due "
3634 "to feature incompatibilities");
3640 * Check feature flags regardless of the revision level, since we
3641 * previously didn't change the revision level when setting the flags,
3642 * so there is a chance incompat flags are set on a rev 0 filesystem.
3644 if (!ext4_feature_set_ok(sb, (sb->s_flags & MS_RDONLY)))
3647 blocksize = BLOCK_SIZE << le32_to_cpu(es->s_log_block_size);
3648 if (blocksize < EXT4_MIN_BLOCK_SIZE ||
3649 blocksize > EXT4_MAX_BLOCK_SIZE) {
3650 ext4_msg(sb, KERN_ERR,
3651 "Unsupported filesystem blocksize %d", blocksize);
3655 if (sb->s_blocksize != blocksize) {
3656 /* Validate the filesystem blocksize */
3657 if (!sb_set_blocksize(sb, blocksize)) {
3658 ext4_msg(sb, KERN_ERR, "bad block size %d",
3664 logical_sb_block = sb_block * EXT4_MIN_BLOCK_SIZE;
3665 offset = do_div(logical_sb_block, blocksize);
3666 bh = sb_bread_unmovable(sb, logical_sb_block);
3668 ext4_msg(sb, KERN_ERR,
3669 "Can't read superblock on 2nd try");
3672 es = (struct ext4_super_block *)(bh->b_data + offset);
3674 if (es->s_magic != cpu_to_le16(EXT4_SUPER_MAGIC)) {
3675 ext4_msg(sb, KERN_ERR,
3676 "Magic mismatch, very weird!");
3681 has_huge_files = EXT4_HAS_RO_COMPAT_FEATURE(sb,
3682 EXT4_FEATURE_RO_COMPAT_HUGE_FILE);
3683 sbi->s_bitmap_maxbytes = ext4_max_bitmap_size(sb->s_blocksize_bits,
3685 sb->s_maxbytes = ext4_max_size(sb->s_blocksize_bits, has_huge_files);
3687 if (le32_to_cpu(es->s_rev_level) == EXT4_GOOD_OLD_REV) {
3688 sbi->s_inode_size = EXT4_GOOD_OLD_INODE_SIZE;
3689 sbi->s_first_ino = EXT4_GOOD_OLD_FIRST_INO;
3691 sbi->s_inode_size = le16_to_cpu(es->s_inode_size);
3692 sbi->s_first_ino = le32_to_cpu(es->s_first_ino);
3693 if ((sbi->s_inode_size < EXT4_GOOD_OLD_INODE_SIZE) ||
3694 (!is_power_of_2(sbi->s_inode_size)) ||
3695 (sbi->s_inode_size > blocksize)) {
3696 ext4_msg(sb, KERN_ERR,
3697 "unsupported inode size: %d",
3701 if (sbi->s_inode_size > EXT4_GOOD_OLD_INODE_SIZE)
3702 sb->s_time_gran = 1 << (EXT4_EPOCH_BITS - 2);
3705 sbi->s_desc_size = le16_to_cpu(es->s_desc_size);
3706 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_64BIT)) {
3707 if (sbi->s_desc_size < EXT4_MIN_DESC_SIZE_64BIT ||
3708 sbi->s_desc_size > EXT4_MAX_DESC_SIZE ||
3709 !is_power_of_2(sbi->s_desc_size)) {
3710 ext4_msg(sb, KERN_ERR,
3711 "unsupported descriptor size %lu",
3716 sbi->s_desc_size = EXT4_MIN_DESC_SIZE;
3718 sbi->s_blocks_per_group = le32_to_cpu(es->s_blocks_per_group);
3719 sbi->s_inodes_per_group = le32_to_cpu(es->s_inodes_per_group);
3720 if (EXT4_INODE_SIZE(sb) == 0 || EXT4_INODES_PER_GROUP(sb) == 0)
3723 sbi->s_inodes_per_block = blocksize / EXT4_INODE_SIZE(sb);
3724 if (sbi->s_inodes_per_block == 0)
3726 sbi->s_itb_per_group = sbi->s_inodes_per_group /
3727 sbi->s_inodes_per_block;
3728 sbi->s_desc_per_block = blocksize / EXT4_DESC_SIZE(sb);
3730 sbi->s_mount_state = le16_to_cpu(es->s_state);
3731 sbi->s_addr_per_block_bits = ilog2(EXT4_ADDR_PER_BLOCK(sb));
3732 sbi->s_desc_per_block_bits = ilog2(EXT4_DESC_PER_BLOCK(sb));
3734 for (i = 0; i < 4; i++)
3735 sbi->s_hash_seed[i] = le32_to_cpu(es->s_hash_seed[i]);
3736 sbi->s_def_hash_version = es->s_def_hash_version;
3737 if (EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_DIR_INDEX)) {
3738 i = le32_to_cpu(es->s_flags);
3739 if (i & EXT2_FLAGS_UNSIGNED_HASH)
3740 sbi->s_hash_unsigned = 3;
3741 else if ((i & EXT2_FLAGS_SIGNED_HASH) == 0) {
3742 #ifdef __CHAR_UNSIGNED__
3743 if (!(sb->s_flags & MS_RDONLY))
3745 cpu_to_le32(EXT2_FLAGS_UNSIGNED_HASH);
3746 sbi->s_hash_unsigned = 3;
3748 if (!(sb->s_flags & MS_RDONLY))
3750 cpu_to_le32(EXT2_FLAGS_SIGNED_HASH);
3755 /* Handle clustersize */
3756 clustersize = BLOCK_SIZE << le32_to_cpu(es->s_log_cluster_size);
3757 has_bigalloc = EXT4_HAS_RO_COMPAT_FEATURE(sb,
3758 EXT4_FEATURE_RO_COMPAT_BIGALLOC);
3760 if (clustersize < blocksize) {
3761 ext4_msg(sb, KERN_ERR,
3762 "cluster size (%d) smaller than "
3763 "block size (%d)", clustersize, blocksize);
3766 sbi->s_cluster_bits = le32_to_cpu(es->s_log_cluster_size) -
3767 le32_to_cpu(es->s_log_block_size);
3768 sbi->s_clusters_per_group =
3769 le32_to_cpu(es->s_clusters_per_group);
3770 if (sbi->s_clusters_per_group > blocksize * 8) {
3771 ext4_msg(sb, KERN_ERR,
3772 "#clusters per group too big: %lu",
3773 sbi->s_clusters_per_group);
3776 if (sbi->s_blocks_per_group !=
3777 (sbi->s_clusters_per_group * (clustersize / blocksize))) {
3778 ext4_msg(sb, KERN_ERR, "blocks per group (%lu) and "
3779 "clusters per group (%lu) inconsistent",
3780 sbi->s_blocks_per_group,
3781 sbi->s_clusters_per_group);
3785 if (clustersize != blocksize) {
3786 ext4_warning(sb, "fragment/cluster size (%d) != "
3787 "block size (%d)", clustersize,
3789 clustersize = blocksize;
3791 if (sbi->s_blocks_per_group > blocksize * 8) {
3792 ext4_msg(sb, KERN_ERR,
3793 "#blocks per group too big: %lu",
3794 sbi->s_blocks_per_group);
3797 sbi->s_clusters_per_group = sbi->s_blocks_per_group;
3798 sbi->s_cluster_bits = 0;
3800 sbi->s_cluster_ratio = clustersize / blocksize;
3802 if (sbi->s_inodes_per_group > blocksize * 8) {
3803 ext4_msg(sb, KERN_ERR,
3804 "#inodes per group too big: %lu",
3805 sbi->s_inodes_per_group);
3809 /* Do we have standard group size of clustersize * 8 blocks ? */
3810 if (sbi->s_blocks_per_group == clustersize << 3)
3811 set_opt2(sb, STD_GROUP_SIZE);
3814 * Test whether we have more sectors than will fit in sector_t,
3815 * and whether the max offset is addressable by the page cache.
3817 err = generic_check_addressable(sb->s_blocksize_bits,
3818 ext4_blocks_count(es));
3820 ext4_msg(sb, KERN_ERR, "filesystem"
3821 " too large to mount safely on this system");
3822 if (sizeof(sector_t) < 8)
3823 ext4_msg(sb, KERN_WARNING, "CONFIG_LBDAF not enabled");
3827 if (EXT4_BLOCKS_PER_GROUP(sb) == 0)
3830 /* check blocks count against device size */
3831 blocks_count = sb->s_bdev->bd_inode->i_size >> sb->s_blocksize_bits;
3832 if (blocks_count && ext4_blocks_count(es) > blocks_count) {
3833 ext4_msg(sb, KERN_WARNING, "bad geometry: block count %llu "
3834 "exceeds size of device (%llu blocks)",
3835 ext4_blocks_count(es), blocks_count);
3840 * It makes no sense for the first data block to be beyond the end
3841 * of the filesystem.
3843 if (le32_to_cpu(es->s_first_data_block) >= ext4_blocks_count(es)) {
3844 ext4_msg(sb, KERN_WARNING, "bad geometry: first data "
3845 "block %u is beyond end of filesystem (%llu)",
3846 le32_to_cpu(es->s_first_data_block),
3847 ext4_blocks_count(es));
3850 blocks_count = (ext4_blocks_count(es) -
3851 le32_to_cpu(es->s_first_data_block) +
3852 EXT4_BLOCKS_PER_GROUP(sb) - 1);
3853 do_div(blocks_count, EXT4_BLOCKS_PER_GROUP(sb));
3854 if (blocks_count > ((uint64_t)1<<32) - EXT4_DESC_PER_BLOCK(sb)) {
3855 ext4_msg(sb, KERN_WARNING, "groups count too large: %u "
3856 "(block count %llu, first data block %u, "
3857 "blocks per group %lu)", sbi->s_groups_count,
3858 ext4_blocks_count(es),
3859 le32_to_cpu(es->s_first_data_block),
3860 EXT4_BLOCKS_PER_GROUP(sb));
3863 sbi->s_groups_count = blocks_count;
3864 sbi->s_blockfile_groups = min_t(ext4_group_t, sbi->s_groups_count,
3865 (EXT4_MAX_BLOCK_FILE_PHYS / EXT4_BLOCKS_PER_GROUP(sb)));
3866 db_count = (sbi->s_groups_count + EXT4_DESC_PER_BLOCK(sb) - 1) /
3867 EXT4_DESC_PER_BLOCK(sb);
3868 sbi->s_group_desc = ext4_kvmalloc(db_count *
3869 sizeof(struct buffer_head *),
3871 if (sbi->s_group_desc == NULL) {
3872 ext4_msg(sb, KERN_ERR, "not enough memory");
3878 sbi->s_proc = proc_mkdir(sb->s_id, ext4_proc_root);
3881 proc_create_data("options", S_IRUGO, sbi->s_proc,
3882 &ext4_seq_options_fops, sb);
3884 bgl_lock_init(sbi->s_blockgroup_lock);
3886 for (i = 0; i < db_count; i++) {
3887 block = descriptor_loc(sb, logical_sb_block, i);
3888 sbi->s_group_desc[i] = sb_bread_unmovable(sb, block);
3889 if (!sbi->s_group_desc[i]) {
3890 ext4_msg(sb, KERN_ERR,
3891 "can't read group descriptor %d", i);
3896 if (!ext4_check_descriptors(sb, &first_not_zeroed)) {
3897 ext4_msg(sb, KERN_ERR, "group descriptors corrupted!");
3901 sbi->s_gdb_count = db_count;
3902 get_random_bytes(&sbi->s_next_generation, sizeof(u32));
3903 spin_lock_init(&sbi->s_next_gen_lock);
3905 init_timer(&sbi->s_err_report);
3906 sbi->s_err_report.function = print_daily_error_info;
3907 sbi->s_err_report.data = (unsigned long) sb;
3909 /* Register extent status tree shrinker */
3910 if (ext4_es_register_shrinker(sbi))
3913 sbi->s_stripe = ext4_get_stripe_size(sbi);
3914 sbi->s_extent_max_zeroout_kb = 32;
3917 * set up enough so that it can read an inode
3919 sb->s_op = &ext4_sops;
3920 sb->s_export_op = &ext4_export_ops;
3921 sb->s_xattr = ext4_xattr_handlers;
3923 sb->dq_op = &ext4_quota_operations;
3924 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_QUOTA))
3925 sb->s_qcop = &dquot_quotactl_sysfile_ops;
3927 sb->s_qcop = &ext4_qctl_operations;
3928 sb->s_quota_types = QTYPE_MASK_USR | QTYPE_MASK_GRP;
3930 memcpy(sb->s_uuid, es->s_uuid, sizeof(es->s_uuid));
3932 INIT_LIST_HEAD(&sbi->s_orphan); /* unlinked but open files */
3933 mutex_init(&sbi->s_orphan_lock);
3937 needs_recovery = (es->s_last_orphan != 0 ||
3938 EXT4_HAS_INCOMPAT_FEATURE(sb,
3939 EXT4_FEATURE_INCOMPAT_RECOVER));
3941 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_MMP) &&
3942 !(sb->s_flags & MS_RDONLY))
3943 if (ext4_multi_mount_protect(sb, le64_to_cpu(es->s_mmp_block)))
3944 goto failed_mount3a;
3947 * The first inode we look at is the journal inode. Don't try
3948 * root first: it may be modified in the journal!
3950 if (!test_opt(sb, NOLOAD) &&
3951 EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL)) {
3952 if (ext4_load_journal(sb, es, journal_devnum))
3953 goto failed_mount3a;
3954 } else if (test_opt(sb, NOLOAD) && !(sb->s_flags & MS_RDONLY) &&
3955 EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER)) {
3956 ext4_msg(sb, KERN_ERR, "required journal recovery "
3957 "suppressed and not mounted read-only");
3958 goto failed_mount_wq;
3960 clear_opt(sb, DATA_FLAGS);
3961 sbi->s_journal = NULL;
3966 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_64BIT) &&
3967 !jbd2_journal_set_features(EXT4_SB(sb)->s_journal, 0, 0,
3968 JBD2_FEATURE_INCOMPAT_64BIT)) {
3969 ext4_msg(sb, KERN_ERR, "Failed to set 64-bit journal feature");
3970 goto failed_mount_wq;
3973 if (!set_journal_csum_feature_set(sb)) {
3974 ext4_msg(sb, KERN_ERR, "Failed to set journal checksum "
3976 goto failed_mount_wq;
3979 /* We have now updated the journal if required, so we can
3980 * validate the data journaling mode. */
3981 switch (test_opt(sb, DATA_FLAGS)) {
3983 /* No mode set, assume a default based on the journal
3984 * capabilities: ORDERED_DATA if the journal can
3985 * cope, else JOURNAL_DATA
3987 if (jbd2_journal_check_available_features
3988 (sbi->s_journal, 0, 0, JBD2_FEATURE_INCOMPAT_REVOKE))
3989 set_opt(sb, ORDERED_DATA);
3991 set_opt(sb, JOURNAL_DATA);
3994 case EXT4_MOUNT_ORDERED_DATA:
3995 case EXT4_MOUNT_WRITEBACK_DATA:
3996 if (!jbd2_journal_check_available_features
3997 (sbi->s_journal, 0, 0, JBD2_FEATURE_INCOMPAT_REVOKE)) {
3998 ext4_msg(sb, KERN_ERR, "Journal does not support "
3999 "requested data journaling mode");
4000 goto failed_mount_wq;
4005 set_task_ioprio(sbi->s_journal->j_task, journal_ioprio);
4007 sbi->s_journal->j_commit_callback = ext4_journal_commit_callback;
4010 if (ext4_mballoc_ready) {
4011 sbi->s_mb_cache = ext4_xattr_create_cache(sb->s_id);
4012 if (!sbi->s_mb_cache) {
4013 ext4_msg(sb, KERN_ERR, "Failed to create an mb_cache");
4014 goto failed_mount_wq;
4019 * Get the # of file system overhead blocks from the
4020 * superblock if present.
4022 if (es->s_overhead_clusters)
4023 sbi->s_overhead = le32_to_cpu(es->s_overhead_clusters);
4025 err = ext4_calculate_overhead(sb);
4027 goto failed_mount_wq;
4031 * The maximum number of concurrent works can be high and
4032 * concurrency isn't really necessary. Limit it to 1.
4034 EXT4_SB(sb)->rsv_conversion_wq =
4035 alloc_workqueue("ext4-rsv-conversion", WQ_MEM_RECLAIM | WQ_UNBOUND, 1);
4036 if (!EXT4_SB(sb)->rsv_conversion_wq) {
4037 printk(KERN_ERR "EXT4-fs: failed to create workqueue\n");
4043 * The jbd2_journal_load will have done any necessary log recovery,
4044 * so we can safely mount the rest of the filesystem now.
4047 root = ext4_iget(sb, EXT4_ROOT_INO);
4049 ext4_msg(sb, KERN_ERR, "get root inode failed");
4050 ret = PTR_ERR(root);
4054 if (!S_ISDIR(root->i_mode) || !root->i_blocks || !root->i_size) {
4055 ext4_msg(sb, KERN_ERR, "corrupt root inode, run e2fsck");
4059 sb->s_root = d_make_root(root);
4061 ext4_msg(sb, KERN_ERR, "get root dentry failed");
4066 if (ext4_setup_super(sb, es, sb->s_flags & MS_RDONLY))
4067 sb->s_flags |= MS_RDONLY;
4069 /* determine the minimum size of new large inodes, if present */
4070 if (sbi->s_inode_size > EXT4_GOOD_OLD_INODE_SIZE) {
4071 sbi->s_want_extra_isize = sizeof(struct ext4_inode) -
4072 EXT4_GOOD_OLD_INODE_SIZE;
4073 if (EXT4_HAS_RO_COMPAT_FEATURE(sb,
4074 EXT4_FEATURE_RO_COMPAT_EXTRA_ISIZE)) {
4075 if (sbi->s_want_extra_isize <
4076 le16_to_cpu(es->s_want_extra_isize))
4077 sbi->s_want_extra_isize =
4078 le16_to_cpu(es->s_want_extra_isize);
4079 if (sbi->s_want_extra_isize <
4080 le16_to_cpu(es->s_min_extra_isize))
4081 sbi->s_want_extra_isize =
4082 le16_to_cpu(es->s_min_extra_isize);
4085 /* Check if enough inode space is available */
4086 if (EXT4_GOOD_OLD_INODE_SIZE + sbi->s_want_extra_isize >
4087 sbi->s_inode_size) {
4088 sbi->s_want_extra_isize = sizeof(struct ext4_inode) -
4089 EXT4_GOOD_OLD_INODE_SIZE;
4090 ext4_msg(sb, KERN_INFO, "required extra inode space not"
4094 err = ext4_reserve_clusters(sbi, ext4_calculate_resv_clusters(sb));
4096 ext4_msg(sb, KERN_ERR, "failed to reserve %llu clusters for "
4097 "reserved pool", ext4_calculate_resv_clusters(sb));
4098 goto failed_mount4a;
4101 err = ext4_setup_system_zone(sb);
4103 ext4_msg(sb, KERN_ERR, "failed to initialize system "
4105 goto failed_mount4a;
4109 err = ext4_mb_init(sb);
4111 ext4_msg(sb, KERN_ERR, "failed to initialize mballoc (%d)",
4116 block = ext4_count_free_clusters(sb);
4117 ext4_free_blocks_count_set(sbi->s_es,
4118 EXT4_C2B(sbi, block));
4119 err = percpu_counter_init(&sbi->s_freeclusters_counter, block,
4122 unsigned long freei = ext4_count_free_inodes(sb);
4123 sbi->s_es->s_free_inodes_count = cpu_to_le32(freei);
4124 err = percpu_counter_init(&sbi->s_freeinodes_counter, freei,
4128 err = percpu_counter_init(&sbi->s_dirs_counter,
4129 ext4_count_dirs(sb), GFP_KERNEL);
4131 err = percpu_counter_init(&sbi->s_dirtyclusters_counter, 0,
4134 ext4_msg(sb, KERN_ERR, "insufficient memory");
4138 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_FLEX_BG))
4139 if (!ext4_fill_flex_info(sb)) {
4140 ext4_msg(sb, KERN_ERR,
4141 "unable to initialize "
4142 "flex_bg meta info!");
4146 err = ext4_register_li_request(sb, first_not_zeroed);
4150 sbi->s_kobj.kset = ext4_kset;
4151 init_completion(&sbi->s_kobj_unregister);
4152 err = kobject_init_and_add(&sbi->s_kobj, &ext4_ktype, NULL,
4158 /* Enable quota usage during mount. */
4159 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_QUOTA) &&
4160 !(sb->s_flags & MS_RDONLY)) {
4161 err = ext4_enable_quotas(sb);
4165 #endif /* CONFIG_QUOTA */
4167 EXT4_SB(sb)->s_mount_state |= EXT4_ORPHAN_FS;
4168 ext4_orphan_cleanup(sb, es);
4169 EXT4_SB(sb)->s_mount_state &= ~EXT4_ORPHAN_FS;
4170 if (needs_recovery) {
4171 ext4_msg(sb, KERN_INFO, "recovery complete");
4172 ext4_mark_recovery_complete(sb, es);
4174 if (EXT4_SB(sb)->s_journal) {
4175 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA)
4176 descr = " journalled data mode";
4177 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA)
4178 descr = " ordered data mode";
4180 descr = " writeback data mode";
4182 descr = "out journal";
4184 if (test_opt(sb, DISCARD)) {
4185 struct request_queue *q = bdev_get_queue(sb->s_bdev);
4186 if (!blk_queue_discard(q))
4187 ext4_msg(sb, KERN_WARNING,
4188 "mounting with \"discard\" option, but "
4189 "the device does not support discard");
4192 ext4_msg(sb, KERN_INFO, "mounted filesystem with%s. "
4193 "Opts: %s%s%s", descr, sbi->s_es->s_mount_opts,
4194 *sbi->s_es->s_mount_opts ? "; " : "", orig_data);
4196 if (es->s_error_count)
4197 mod_timer(&sbi->s_err_report, jiffies + 300*HZ); /* 5 minutes */
4199 /* Enable message ratelimiting. Default is 10 messages per 5 secs. */
4200 ratelimit_state_init(&sbi->s_err_ratelimit_state, 5 * HZ, 10);
4201 ratelimit_state_init(&sbi->s_warning_ratelimit_state, 5 * HZ, 10);
4202 ratelimit_state_init(&sbi->s_msg_ratelimit_state, 5 * HZ, 10);
4209 ext4_msg(sb, KERN_ERR, "VFS: Can't find ext4 filesystem");
4214 kobject_del(&sbi->s_kobj);
4217 ext4_unregister_li_request(sb);
4219 ext4_mb_release(sb);
4220 if (sbi->s_flex_groups)
4221 kvfree(sbi->s_flex_groups);
4222 percpu_counter_destroy(&sbi->s_freeclusters_counter);
4223 percpu_counter_destroy(&sbi->s_freeinodes_counter);
4224 percpu_counter_destroy(&sbi->s_dirs_counter);
4225 percpu_counter_destroy(&sbi->s_dirtyclusters_counter);
4227 ext4_ext_release(sb);
4228 ext4_release_system_zone(sb);
4233 ext4_msg(sb, KERN_ERR, "mount failed");
4234 if (EXT4_SB(sb)->rsv_conversion_wq)
4235 destroy_workqueue(EXT4_SB(sb)->rsv_conversion_wq);
4237 if (sbi->s_journal) {
4238 jbd2_journal_destroy(sbi->s_journal);
4239 sbi->s_journal = NULL;
4242 ext4_es_unregister_shrinker(sbi);
4244 del_timer_sync(&sbi->s_err_report);
4246 kthread_stop(sbi->s_mmp_tsk);
4248 for (i = 0; i < db_count; i++)
4249 brelse(sbi->s_group_desc[i]);
4250 kvfree(sbi->s_group_desc);
4252 if (sbi->s_chksum_driver)
4253 crypto_free_shash(sbi->s_chksum_driver);
4255 remove_proc_entry("options", sbi->s_proc);
4256 remove_proc_entry(sb->s_id, ext4_proc_root);
4259 for (i = 0; i < EXT4_MAXQUOTAS; i++)
4260 kfree(sbi->s_qf_names[i]);
4262 ext4_blkdev_remove(sbi);
4265 sb->s_fs_info = NULL;
4266 kfree(sbi->s_blockgroup_lock);
4270 return err ? err : ret;
4274 * Setup any per-fs journal parameters now. We'll do this both on
4275 * initial mount, once the journal has been initialised but before we've
4276 * done any recovery; and again on any subsequent remount.
4278 static void ext4_init_journal_params(struct super_block *sb, journal_t *journal)
4280 struct ext4_sb_info *sbi = EXT4_SB(sb);
4282 journal->j_commit_interval = sbi->s_commit_interval;
4283 journal->j_min_batch_time = sbi->s_min_batch_time;
4284 journal->j_max_batch_time = sbi->s_max_batch_time;
4286 write_lock(&journal->j_state_lock);
4287 if (test_opt(sb, BARRIER))
4288 journal->j_flags |= JBD2_BARRIER;
4290 journal->j_flags &= ~JBD2_BARRIER;
4291 if (test_opt(sb, DATA_ERR_ABORT))
4292 journal->j_flags |= JBD2_ABORT_ON_SYNCDATA_ERR;
4294 journal->j_flags &= ~JBD2_ABORT_ON_SYNCDATA_ERR;
4295 write_unlock(&journal->j_state_lock);
4298 static journal_t *ext4_get_journal(struct super_block *sb,
4299 unsigned int journal_inum)
4301 struct inode *journal_inode;
4304 BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL));
4306 /* First, test for the existence of a valid inode on disk. Bad
4307 * things happen if we iget() an unused inode, as the subsequent
4308 * iput() will try to delete it. */
4310 journal_inode = ext4_iget(sb, journal_inum);
4311 if (IS_ERR(journal_inode)) {
4312 ext4_msg(sb, KERN_ERR, "no journal found");
4315 if (!journal_inode->i_nlink) {
4316 make_bad_inode(journal_inode);
4317 iput(journal_inode);
4318 ext4_msg(sb, KERN_ERR, "journal inode is deleted");
4322 jbd_debug(2, "Journal inode found at %p: %lld bytes\n",
4323 journal_inode, journal_inode->i_size);
4324 if (!S_ISREG(journal_inode->i_mode)) {
4325 ext4_msg(sb, KERN_ERR, "invalid journal inode");
4326 iput(journal_inode);
4330 journal = jbd2_journal_init_inode(journal_inode);
4332 ext4_msg(sb, KERN_ERR, "Could not load journal inode");
4333 iput(journal_inode);
4336 journal->j_private = sb;
4337 ext4_init_journal_params(sb, journal);
4341 static journal_t *ext4_get_dev_journal(struct super_block *sb,
4344 struct buffer_head *bh;
4348 int hblock, blocksize;
4349 ext4_fsblk_t sb_block;
4350 unsigned long offset;
4351 struct ext4_super_block *es;
4352 struct block_device *bdev;
4354 BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL));
4356 bdev = ext4_blkdev_get(j_dev, sb);
4360 blocksize = sb->s_blocksize;
4361 hblock = bdev_logical_block_size(bdev);
4362 if (blocksize < hblock) {
4363 ext4_msg(sb, KERN_ERR,
4364 "blocksize too small for journal device");
4368 sb_block = EXT4_MIN_BLOCK_SIZE / blocksize;
4369 offset = EXT4_MIN_BLOCK_SIZE % blocksize;
4370 set_blocksize(bdev, blocksize);
4371 if (!(bh = __bread(bdev, sb_block, blocksize))) {
4372 ext4_msg(sb, KERN_ERR, "couldn't read superblock of "
4373 "external journal");
4377 es = (struct ext4_super_block *) (bh->b_data + offset);
4378 if ((le16_to_cpu(es->s_magic) != EXT4_SUPER_MAGIC) ||
4379 !(le32_to_cpu(es->s_feature_incompat) &
4380 EXT4_FEATURE_INCOMPAT_JOURNAL_DEV)) {
4381 ext4_msg(sb, KERN_ERR, "external journal has "
4387 if ((le32_to_cpu(es->s_feature_ro_compat) &
4388 EXT4_FEATURE_RO_COMPAT_METADATA_CSUM) &&
4389 es->s_checksum != ext4_superblock_csum(sb, es)) {
4390 ext4_msg(sb, KERN_ERR, "external journal has "
4391 "corrupt superblock");
4396 if (memcmp(EXT4_SB(sb)->s_es->s_journal_uuid, es->s_uuid, 16)) {
4397 ext4_msg(sb, KERN_ERR, "journal UUID does not match");
4402 len = ext4_blocks_count(es);
4403 start = sb_block + 1;
4404 brelse(bh); /* we're done with the superblock */
4406 journal = jbd2_journal_init_dev(bdev, sb->s_bdev,
4407 start, len, blocksize);
4409 ext4_msg(sb, KERN_ERR, "failed to create device journal");
4412 journal->j_private = sb;
4413 ll_rw_block(READ | REQ_META | REQ_PRIO, 1, &journal->j_sb_buffer);
4414 wait_on_buffer(journal->j_sb_buffer);
4415 if (!buffer_uptodate(journal->j_sb_buffer)) {
4416 ext4_msg(sb, KERN_ERR, "I/O error on journal device");
4419 if (be32_to_cpu(journal->j_superblock->s_nr_users) != 1) {
4420 ext4_msg(sb, KERN_ERR, "External journal has more than one "
4421 "user (unsupported) - %d",
4422 be32_to_cpu(journal->j_superblock->s_nr_users));
4425 EXT4_SB(sb)->journal_bdev = bdev;
4426 ext4_init_journal_params(sb, journal);
4430 jbd2_journal_destroy(journal);
4432 ext4_blkdev_put(bdev);
4436 static int ext4_load_journal(struct super_block *sb,
4437 struct ext4_super_block *es,
4438 unsigned long journal_devnum)
4441 unsigned int journal_inum = le32_to_cpu(es->s_journal_inum);
4444 int really_read_only;
4446 BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL));
4448 if (journal_devnum &&
4449 journal_devnum != le32_to_cpu(es->s_journal_dev)) {
4450 ext4_msg(sb, KERN_INFO, "external journal device major/minor "
4451 "numbers have changed");
4452 journal_dev = new_decode_dev(journal_devnum);
4454 journal_dev = new_decode_dev(le32_to_cpu(es->s_journal_dev));
4456 really_read_only = bdev_read_only(sb->s_bdev);
4459 * Are we loading a blank journal or performing recovery after a
4460 * crash? For recovery, we need to check in advance whether we
4461 * can get read-write access to the device.
4463 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER)) {
4464 if (sb->s_flags & MS_RDONLY) {
4465 ext4_msg(sb, KERN_INFO, "INFO: recovery "
4466 "required on readonly filesystem");
4467 if (really_read_only) {
4468 ext4_msg(sb, KERN_ERR, "write access "
4469 "unavailable, cannot proceed");
4472 ext4_msg(sb, KERN_INFO, "write access will "
4473 "be enabled during recovery");
4477 if (journal_inum && journal_dev) {
4478 ext4_msg(sb, KERN_ERR, "filesystem has both journal "
4479 "and inode journals!");
4484 if (!(journal = ext4_get_journal(sb, journal_inum)))
4487 if (!(journal = ext4_get_dev_journal(sb, journal_dev)))
4491 if (!(journal->j_flags & JBD2_BARRIER))
4492 ext4_msg(sb, KERN_INFO, "barriers disabled");
4494 if (!EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER))
4495 err = jbd2_journal_wipe(journal, !really_read_only);
4497 char *save = kmalloc(EXT4_S_ERR_LEN, GFP_KERNEL);
4499 memcpy(save, ((char *) es) +
4500 EXT4_S_ERR_START, EXT4_S_ERR_LEN);
4501 err = jbd2_journal_load(journal);
4503 memcpy(((char *) es) + EXT4_S_ERR_START,
4504 save, EXT4_S_ERR_LEN);
4509 ext4_msg(sb, KERN_ERR, "error loading journal");
4510 jbd2_journal_destroy(journal);
4514 EXT4_SB(sb)->s_journal = journal;
4515 ext4_clear_journal_err(sb, es);
4517 if (!really_read_only && journal_devnum &&
4518 journal_devnum != le32_to_cpu(es->s_journal_dev)) {
4519 es->s_journal_dev = cpu_to_le32(journal_devnum);
4521 /* Make sure we flush the recovery flag to disk. */
4522 ext4_commit_super(sb, 1);
4528 static int ext4_commit_super(struct super_block *sb, int sync)
4530 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
4531 struct buffer_head *sbh = EXT4_SB(sb)->s_sbh;
4534 if (!sbh || block_device_ejected(sb))
4536 if (buffer_write_io_error(sbh)) {
4538 * Oh, dear. A previous attempt to write the
4539 * superblock failed. This could happen because the
4540 * USB device was yanked out. Or it could happen to
4541 * be a transient write error and maybe the block will
4542 * be remapped. Nothing we can do but to retry the
4543 * write and hope for the best.
4545 ext4_msg(sb, KERN_ERR, "previous I/O error to "
4546 "superblock detected");
4547 clear_buffer_write_io_error(sbh);
4548 set_buffer_uptodate(sbh);
4551 * If the file system is mounted read-only, don't update the
4552 * superblock write time. This avoids updating the superblock
4553 * write time when we are mounting the root file system
4554 * read/only but we need to replay the journal; at that point,
4555 * for people who are east of GMT and who make their clock
4556 * tick in localtime for Windows bug-for-bug compatibility,
4557 * the clock is set in the future, and this will cause e2fsck
4558 * to complain and force a full file system check.
4560 if (!(sb->s_flags & MS_RDONLY))
4561 es->s_wtime = cpu_to_le32(get_seconds());
4562 if (sb->s_bdev->bd_part)
4563 es->s_kbytes_written =
4564 cpu_to_le64(EXT4_SB(sb)->s_kbytes_written +
4565 ((part_stat_read(sb->s_bdev->bd_part, sectors[1]) -
4566 EXT4_SB(sb)->s_sectors_written_start) >> 1));
4568 es->s_kbytes_written =
4569 cpu_to_le64(EXT4_SB(sb)->s_kbytes_written);
4570 if (percpu_counter_initialized(&EXT4_SB(sb)->s_freeclusters_counter))
4571 ext4_free_blocks_count_set(es,
4572 EXT4_C2B(EXT4_SB(sb), percpu_counter_sum_positive(
4573 &EXT4_SB(sb)->s_freeclusters_counter)));
4574 if (percpu_counter_initialized(&EXT4_SB(sb)->s_freeinodes_counter))
4575 es->s_free_inodes_count =
4576 cpu_to_le32(percpu_counter_sum_positive(
4577 &EXT4_SB(sb)->s_freeinodes_counter));
4578 BUFFER_TRACE(sbh, "marking dirty");
4579 ext4_superblock_csum_set(sb);
4580 mark_buffer_dirty(sbh);
4582 error = sync_dirty_buffer(sbh);
4586 error = buffer_write_io_error(sbh);
4588 ext4_msg(sb, KERN_ERR, "I/O error while writing "
4590 clear_buffer_write_io_error(sbh);
4591 set_buffer_uptodate(sbh);
4598 * Have we just finished recovery? If so, and if we are mounting (or
4599 * remounting) the filesystem readonly, then we will end up with a
4600 * consistent fs on disk. Record that fact.
4602 static void ext4_mark_recovery_complete(struct super_block *sb,
4603 struct ext4_super_block *es)
4605 journal_t *journal = EXT4_SB(sb)->s_journal;
4607 if (!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL)) {
4608 BUG_ON(journal != NULL);
4611 jbd2_journal_lock_updates(journal);
4612 if (jbd2_journal_flush(journal) < 0)
4615 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER) &&
4616 sb->s_flags & MS_RDONLY) {
4617 EXT4_CLEAR_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
4618 ext4_commit_super(sb, 1);
4622 jbd2_journal_unlock_updates(journal);
4626 * If we are mounting (or read-write remounting) a filesystem whose journal
4627 * has recorded an error from a previous lifetime, move that error to the
4628 * main filesystem now.
4630 static void ext4_clear_journal_err(struct super_block *sb,
4631 struct ext4_super_block *es)
4637 BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL));
4639 journal = EXT4_SB(sb)->s_journal;
4642 * Now check for any error status which may have been recorded in the
4643 * journal by a prior ext4_error() or ext4_abort()
4646 j_errno = jbd2_journal_errno(journal);
4650 errstr = ext4_decode_error(sb, j_errno, nbuf);
4651 ext4_warning(sb, "Filesystem error recorded "
4652 "from previous mount: %s", errstr);
4653 ext4_warning(sb, "Marking fs in need of filesystem check.");
4655 EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS;
4656 es->s_state |= cpu_to_le16(EXT4_ERROR_FS);
4657 ext4_commit_super(sb, 1);
4659 jbd2_journal_clear_err(journal);
4660 jbd2_journal_update_sb_errno(journal);
4665 * Force the running and committing transactions to commit,
4666 * and wait on the commit.
4668 int ext4_force_commit(struct super_block *sb)
4672 if (sb->s_flags & MS_RDONLY)
4675 journal = EXT4_SB(sb)->s_journal;
4676 return ext4_journal_force_commit(journal);
4679 static int ext4_sync_fs(struct super_block *sb, int wait)
4683 bool needs_barrier = false;
4684 struct ext4_sb_info *sbi = EXT4_SB(sb);
4686 trace_ext4_sync_fs(sb, wait);
4687 flush_workqueue(sbi->rsv_conversion_wq);
4689 * Writeback quota in non-journalled quota case - journalled quota has
4692 dquot_writeback_dquots(sb, -1);
4694 * Data writeback is possible w/o journal transaction, so barrier must
4695 * being sent at the end of the function. But we can skip it if
4696 * transaction_commit will do it for us.
4698 if (sbi->s_journal) {
4699 target = jbd2_get_latest_transaction(sbi->s_journal);
4700 if (wait && sbi->s_journal->j_flags & JBD2_BARRIER &&
4701 !jbd2_trans_will_send_data_barrier(sbi->s_journal, target))
4702 needs_barrier = true;
4704 if (jbd2_journal_start_commit(sbi->s_journal, &target)) {
4706 ret = jbd2_log_wait_commit(sbi->s_journal,
4709 } else if (wait && test_opt(sb, BARRIER))
4710 needs_barrier = true;
4711 if (needs_barrier) {
4713 err = blkdev_issue_flush(sb->s_bdev, GFP_KERNEL, NULL);
4722 * LVM calls this function before a (read-only) snapshot is created. This
4723 * gives us a chance to flush the journal completely and mark the fs clean.
4725 * Note that only this function cannot bring a filesystem to be in a clean
4726 * state independently. It relies on upper layer to stop all data & metadata
4729 static int ext4_freeze(struct super_block *sb)
4734 if (sb->s_flags & MS_RDONLY)
4737 journal = EXT4_SB(sb)->s_journal;
4740 /* Now we set up the journal barrier. */
4741 jbd2_journal_lock_updates(journal);
4744 * Don't clear the needs_recovery flag if we failed to
4745 * flush the journal.
4747 error = jbd2_journal_flush(journal);
4752 /* Journal blocked and flushed, clear needs_recovery flag. */
4753 EXT4_CLEAR_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
4754 error = ext4_commit_super(sb, 1);
4757 /* we rely on upper layer to stop further updates */
4758 jbd2_journal_unlock_updates(journal);
4763 * Called by LVM after the snapshot is done. We need to reset the RECOVER
4764 * flag here, even though the filesystem is not technically dirty yet.
4766 static int ext4_unfreeze(struct super_block *sb)
4768 if (sb->s_flags & MS_RDONLY)
4771 /* Reset the needs_recovery flag before the fs is unlocked. */
4772 EXT4_SET_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
4773 ext4_commit_super(sb, 1);
4778 * Structure to save mount options for ext4_remount's benefit
4780 struct ext4_mount_options {
4781 unsigned long s_mount_opt;
4782 unsigned long s_mount_opt2;
4785 unsigned long s_commit_interval;
4786 u32 s_min_batch_time, s_max_batch_time;
4789 char *s_qf_names[EXT4_MAXQUOTAS];
4793 static int ext4_remount(struct super_block *sb, int *flags, char *data)
4795 struct ext4_super_block *es;
4796 struct ext4_sb_info *sbi = EXT4_SB(sb);
4797 unsigned long old_sb_flags;
4798 struct ext4_mount_options old_opts;
4799 int enable_quota = 0;
4801 unsigned int journal_ioprio = DEFAULT_JOURNAL_IOPRIO;
4806 char *orig_data = kstrdup(data, GFP_KERNEL);
4808 /* Store the original options */
4809 old_sb_flags = sb->s_flags;
4810 old_opts.s_mount_opt = sbi->s_mount_opt;
4811 old_opts.s_mount_opt2 = sbi->s_mount_opt2;
4812 old_opts.s_resuid = sbi->s_resuid;
4813 old_opts.s_resgid = sbi->s_resgid;
4814 old_opts.s_commit_interval = sbi->s_commit_interval;
4815 old_opts.s_min_batch_time = sbi->s_min_batch_time;
4816 old_opts.s_max_batch_time = sbi->s_max_batch_time;
4818 old_opts.s_jquota_fmt = sbi->s_jquota_fmt;
4819 for (i = 0; i < EXT4_MAXQUOTAS; i++)
4820 if (sbi->s_qf_names[i]) {
4821 old_opts.s_qf_names[i] = kstrdup(sbi->s_qf_names[i],
4823 if (!old_opts.s_qf_names[i]) {
4824 for (j = 0; j < i; j++)
4825 kfree(old_opts.s_qf_names[j]);
4830 old_opts.s_qf_names[i] = NULL;
4832 if (sbi->s_journal && sbi->s_journal->j_task->io_context)
4833 journal_ioprio = sbi->s_journal->j_task->io_context->ioprio;
4836 * Allow the "check" option to be passed as a remount option.
4838 if (!parse_options(data, sb, NULL, &journal_ioprio, 1)) {
4843 if ((old_opts.s_mount_opt & EXT4_MOUNT_JOURNAL_CHECKSUM) ^
4844 test_opt(sb, JOURNAL_CHECKSUM)) {
4845 ext4_msg(sb, KERN_ERR, "changing journal_checksum "
4846 "during remount not supported");
4851 if ((old_opts.s_mount_opt & EXT4_MOUNT_JOURNAL_CHECKSUM) ^
4852 test_opt(sb, JOURNAL_CHECKSUM)) {
4853 ext4_msg(sb, KERN_ERR, "changing journal_checksum "
4854 "during remount not supported");
4859 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA) {
4860 if (test_opt2(sb, EXPLICIT_DELALLOC)) {
4861 ext4_msg(sb, KERN_ERR, "can't mount with "
4862 "both data=journal and delalloc");
4866 if (test_opt(sb, DIOREAD_NOLOCK)) {
4867 ext4_msg(sb, KERN_ERR, "can't mount with "
4868 "both data=journal and dioread_nolock");
4874 if (sbi->s_mount_flags & EXT4_MF_FS_ABORTED)
4875 ext4_abort(sb, "Abort forced by user");
4877 sb->s_flags = (sb->s_flags & ~MS_POSIXACL) |
4878 (test_opt(sb, POSIX_ACL) ? MS_POSIXACL : 0);
4882 if (sbi->s_journal) {
4883 ext4_init_journal_params(sb, sbi->s_journal);
4884 set_task_ioprio(sbi->s_journal->j_task, journal_ioprio);
4887 if ((*flags & MS_RDONLY) != (sb->s_flags & MS_RDONLY)) {
4888 if (sbi->s_mount_flags & EXT4_MF_FS_ABORTED) {
4893 if (*flags & MS_RDONLY) {
4894 err = sync_filesystem(sb);
4897 err = dquot_suspend(sb, -1);
4902 * First of all, the unconditional stuff we have to do
4903 * to disable replay of the journal when we next remount
4905 sb->s_flags |= MS_RDONLY;
4908 * OK, test if we are remounting a valid rw partition
4909 * readonly, and if so set the rdonly flag and then
4910 * mark the partition as valid again.
4912 if (!(es->s_state & cpu_to_le16(EXT4_VALID_FS)) &&
4913 (sbi->s_mount_state & EXT4_VALID_FS))
4914 es->s_state = cpu_to_le16(sbi->s_mount_state);
4917 ext4_mark_recovery_complete(sb, es);
4919 /* Make sure we can mount this feature set readwrite */
4920 if (!ext4_feature_set_ok(sb, 0)) {
4925 * Make sure the group descriptor checksums
4926 * are sane. If they aren't, refuse to remount r/w.
4928 for (g = 0; g < sbi->s_groups_count; g++) {
4929 struct ext4_group_desc *gdp =
4930 ext4_get_group_desc(sb, g, NULL);
4932 if (!ext4_group_desc_csum_verify(sb, g, gdp)) {
4933 ext4_msg(sb, KERN_ERR,
4934 "ext4_remount: Checksum for group %u failed (%u!=%u)",
4935 g, le16_to_cpu(ext4_group_desc_csum(sbi, g, gdp)),
4936 le16_to_cpu(gdp->bg_checksum));
4943 * If we have an unprocessed orphan list hanging
4944 * around from a previously readonly bdev mount,
4945 * require a full umount/remount for now.
4947 if (es->s_last_orphan) {
4948 ext4_msg(sb, KERN_WARNING, "Couldn't "
4949 "remount RDWR because of unprocessed "
4950 "orphan inode list. Please "
4951 "umount/remount instead");
4957 * Mounting a RDONLY partition read-write, so reread
4958 * and store the current valid flag. (It may have
4959 * been changed by e2fsck since we originally mounted
4963 ext4_clear_journal_err(sb, es);
4964 sbi->s_mount_state = le16_to_cpu(es->s_state);
4965 if (!ext4_setup_super(sb, es, 0))
4966 sb->s_flags &= ~MS_RDONLY;
4967 if (EXT4_HAS_INCOMPAT_FEATURE(sb,
4968 EXT4_FEATURE_INCOMPAT_MMP))
4969 if (ext4_multi_mount_protect(sb,
4970 le64_to_cpu(es->s_mmp_block))) {
4979 * Reinitialize lazy itable initialization thread based on
4982 if ((sb->s_flags & MS_RDONLY) || !test_opt(sb, INIT_INODE_TABLE))
4983 ext4_unregister_li_request(sb);
4985 ext4_group_t first_not_zeroed;
4986 first_not_zeroed = ext4_has_uninit_itable(sb);
4987 ext4_register_li_request(sb, first_not_zeroed);
4990 ext4_setup_system_zone(sb);
4991 if (sbi->s_journal == NULL && !(old_sb_flags & MS_RDONLY))
4992 ext4_commit_super(sb, 1);
4995 /* Release old quota file names */
4996 for (i = 0; i < EXT4_MAXQUOTAS; i++)
4997 kfree(old_opts.s_qf_names[i]);
4999 if (sb_any_quota_suspended(sb))
5000 dquot_resume(sb, -1);
5001 else if (EXT4_HAS_RO_COMPAT_FEATURE(sb,
5002 EXT4_FEATURE_RO_COMPAT_QUOTA)) {
5003 err = ext4_enable_quotas(sb);
5010 ext4_msg(sb, KERN_INFO, "re-mounted. Opts: %s", orig_data);
5015 sb->s_flags = old_sb_flags;
5016 sbi->s_mount_opt = old_opts.s_mount_opt;
5017 sbi->s_mount_opt2 = old_opts.s_mount_opt2;
5018 sbi->s_resuid = old_opts.s_resuid;
5019 sbi->s_resgid = old_opts.s_resgid;
5020 sbi->s_commit_interval = old_opts.s_commit_interval;
5021 sbi->s_min_batch_time = old_opts.s_min_batch_time;
5022 sbi->s_max_batch_time = old_opts.s_max_batch_time;
5024 sbi->s_jquota_fmt = old_opts.s_jquota_fmt;
5025 for (i = 0; i < EXT4_MAXQUOTAS; i++) {
5026 kfree(sbi->s_qf_names[i]);
5027 sbi->s_qf_names[i] = old_opts.s_qf_names[i];
5034 static int ext4_statfs(struct dentry *dentry, struct kstatfs *buf)
5036 struct super_block *sb = dentry->d_sb;
5037 struct ext4_sb_info *sbi = EXT4_SB(sb);
5038 struct ext4_super_block *es = sbi->s_es;
5039 ext4_fsblk_t overhead = 0, resv_blocks;
5042 resv_blocks = EXT4_C2B(sbi, atomic64_read(&sbi->s_resv_clusters));
5044 if (!test_opt(sb, MINIX_DF))
5045 overhead = sbi->s_overhead;
5047 buf->f_type = EXT4_SUPER_MAGIC;
5048 buf->f_bsize = sb->s_blocksize;
5049 buf->f_blocks = ext4_blocks_count(es) - EXT4_C2B(sbi, overhead);
5050 bfree = percpu_counter_sum_positive(&sbi->s_freeclusters_counter) -
5051 percpu_counter_sum_positive(&sbi->s_dirtyclusters_counter);
5052 /* prevent underflow in case that few free space is available */
5053 buf->f_bfree = EXT4_C2B(sbi, max_t(s64, bfree, 0));
5054 buf->f_bavail = buf->f_bfree -
5055 (ext4_r_blocks_count(es) + resv_blocks);
5056 if (buf->f_bfree < (ext4_r_blocks_count(es) + resv_blocks))
5058 buf->f_files = le32_to_cpu(es->s_inodes_count);
5059 buf->f_ffree = percpu_counter_sum_positive(&sbi->s_freeinodes_counter);
5060 buf->f_namelen = EXT4_NAME_LEN;
5061 fsid = le64_to_cpup((void *)es->s_uuid) ^
5062 le64_to_cpup((void *)es->s_uuid + sizeof(u64));
5063 buf->f_fsid.val[0] = fsid & 0xFFFFFFFFUL;
5064 buf->f_fsid.val[1] = (fsid >> 32) & 0xFFFFFFFFUL;
5069 /* Helper function for writing quotas on sync - we need to start transaction
5070 * before quota file is locked for write. Otherwise the are possible deadlocks:
5071 * Process 1 Process 2
5072 * ext4_create() quota_sync()
5073 * jbd2_journal_start() write_dquot()
5074 * dquot_initialize() down(dqio_mutex)
5075 * down(dqio_mutex) jbd2_journal_start()
5081 static inline struct inode *dquot_to_inode(struct dquot *dquot)
5083 return sb_dqopt(dquot->dq_sb)->files[dquot->dq_id.type];
5086 static int ext4_write_dquot(struct dquot *dquot)
5090 struct inode *inode;
5092 inode = dquot_to_inode(dquot);
5093 handle = ext4_journal_start(inode, EXT4_HT_QUOTA,
5094 EXT4_QUOTA_TRANS_BLOCKS(dquot->dq_sb));
5096 return PTR_ERR(handle);
5097 ret = dquot_commit(dquot);
5098 err = ext4_journal_stop(handle);
5104 static int ext4_acquire_dquot(struct dquot *dquot)
5109 handle = ext4_journal_start(dquot_to_inode(dquot), EXT4_HT_QUOTA,
5110 EXT4_QUOTA_INIT_BLOCKS(dquot->dq_sb));
5112 return PTR_ERR(handle);
5113 ret = dquot_acquire(dquot);
5114 err = ext4_journal_stop(handle);
5120 static int ext4_release_dquot(struct dquot *dquot)
5125 handle = ext4_journal_start(dquot_to_inode(dquot), EXT4_HT_QUOTA,
5126 EXT4_QUOTA_DEL_BLOCKS(dquot->dq_sb));
5127 if (IS_ERR(handle)) {
5128 /* Release dquot anyway to avoid endless cycle in dqput() */
5129 dquot_release(dquot);
5130 return PTR_ERR(handle);
5132 ret = dquot_release(dquot);
5133 err = ext4_journal_stop(handle);
5139 static int ext4_mark_dquot_dirty(struct dquot *dquot)
5141 struct super_block *sb = dquot->dq_sb;
5142 struct ext4_sb_info *sbi = EXT4_SB(sb);
5144 /* Are we journaling quotas? */
5145 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_QUOTA) ||
5146 sbi->s_qf_names[USRQUOTA] || sbi->s_qf_names[GRPQUOTA]) {
5147 dquot_mark_dquot_dirty(dquot);
5148 return ext4_write_dquot(dquot);
5150 return dquot_mark_dquot_dirty(dquot);
5154 static int ext4_write_info(struct super_block *sb, int type)
5159 /* Data block + inode block */
5160 handle = ext4_journal_start(sb->s_root->d_inode, EXT4_HT_QUOTA, 2);
5162 return PTR_ERR(handle);
5163 ret = dquot_commit_info(sb, type);
5164 err = ext4_journal_stop(handle);
5171 * Turn on quotas during mount time - we need to find
5172 * the quota file and such...
5174 static int ext4_quota_on_mount(struct super_block *sb, int type)
5176 return dquot_quota_on_mount(sb, EXT4_SB(sb)->s_qf_names[type],
5177 EXT4_SB(sb)->s_jquota_fmt, type);
5181 * Standard function to be called on quota_on
5183 static int ext4_quota_on(struct super_block *sb, int type, int format_id,
5188 if (!test_opt(sb, QUOTA))
5191 /* Quotafile not on the same filesystem? */
5192 if (path->dentry->d_sb != sb)
5194 /* Journaling quota? */
5195 if (EXT4_SB(sb)->s_qf_names[type]) {
5196 /* Quotafile not in fs root? */
5197 if (path->dentry->d_parent != sb->s_root)
5198 ext4_msg(sb, KERN_WARNING,
5199 "Quota file not on filesystem root. "
5200 "Journaled quota will not work");
5204 * When we journal data on quota file, we have to flush journal to see
5205 * all updates to the file when we bypass pagecache...
5207 if (EXT4_SB(sb)->s_journal &&
5208 ext4_should_journal_data(path->dentry->d_inode)) {
5210 * We don't need to lock updates but journal_flush() could
5211 * otherwise be livelocked...
5213 jbd2_journal_lock_updates(EXT4_SB(sb)->s_journal);
5214 err = jbd2_journal_flush(EXT4_SB(sb)->s_journal);
5215 jbd2_journal_unlock_updates(EXT4_SB(sb)->s_journal);
5220 return dquot_quota_on(sb, type, format_id, path);
5223 static int ext4_quota_enable(struct super_block *sb, int type, int format_id,
5227 struct inode *qf_inode;
5228 unsigned long qf_inums[EXT4_MAXQUOTAS] = {
5229 le32_to_cpu(EXT4_SB(sb)->s_es->s_usr_quota_inum),
5230 le32_to_cpu(EXT4_SB(sb)->s_es->s_grp_quota_inum)
5233 BUG_ON(!EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_QUOTA));
5235 if (!qf_inums[type])
5238 qf_inode = ext4_iget(sb, qf_inums[type]);
5239 if (IS_ERR(qf_inode)) {
5240 ext4_error(sb, "Bad quota inode # %lu", qf_inums[type]);
5241 return PTR_ERR(qf_inode);
5244 /* Don't account quota for quota files to avoid recursion */
5245 qf_inode->i_flags |= S_NOQUOTA;
5246 err = dquot_enable(qf_inode, type, format_id, flags);
5252 /* Enable usage tracking for all quota types. */
5253 static int ext4_enable_quotas(struct super_block *sb)
5256 unsigned long qf_inums[EXT4_MAXQUOTAS] = {
5257 le32_to_cpu(EXT4_SB(sb)->s_es->s_usr_quota_inum),
5258 le32_to_cpu(EXT4_SB(sb)->s_es->s_grp_quota_inum)
5261 sb_dqopt(sb)->flags |= DQUOT_QUOTA_SYS_FILE;
5262 for (type = 0; type < EXT4_MAXQUOTAS; type++) {
5263 if (qf_inums[type]) {
5264 err = ext4_quota_enable(sb, type, QFMT_VFS_V1,
5265 DQUOT_USAGE_ENABLED);
5268 "Failed to enable quota tracking "
5269 "(type=%d, err=%d). Please run "
5270 "e2fsck to fix.", type, err);
5278 static int ext4_quota_off(struct super_block *sb, int type)
5280 struct inode *inode = sb_dqopt(sb)->files[type];
5283 /* Force all delayed allocation blocks to be allocated.
5284 * Caller already holds s_umount sem */
5285 if (test_opt(sb, DELALLOC))
5286 sync_filesystem(sb);
5291 /* Update modification times of quota files when userspace can
5292 * start looking at them */
5293 handle = ext4_journal_start(inode, EXT4_HT_QUOTA, 1);
5296 inode->i_mtime = inode->i_ctime = CURRENT_TIME;
5297 ext4_mark_inode_dirty(handle, inode);
5298 ext4_journal_stop(handle);
5301 return dquot_quota_off(sb, type);
5304 /* Read data from quotafile - avoid pagecache and such because we cannot afford
5305 * acquiring the locks... As quota files are never truncated and quota code
5306 * itself serializes the operations (and no one else should touch the files)
5307 * we don't have to be afraid of races */
5308 static ssize_t ext4_quota_read(struct super_block *sb, int type, char *data,
5309 size_t len, loff_t off)
5311 struct inode *inode = sb_dqopt(sb)->files[type];
5312 ext4_lblk_t blk = off >> EXT4_BLOCK_SIZE_BITS(sb);
5313 int offset = off & (sb->s_blocksize - 1);
5316 struct buffer_head *bh;
5317 loff_t i_size = i_size_read(inode);
5321 if (off+len > i_size)
5324 while (toread > 0) {
5325 tocopy = sb->s_blocksize - offset < toread ?
5326 sb->s_blocksize - offset : toread;
5327 bh = ext4_bread(NULL, inode, blk, 0);
5330 if (!bh) /* A hole? */
5331 memset(data, 0, tocopy);
5333 memcpy(data, bh->b_data+offset, tocopy);
5343 /* Write to quotafile (we know the transaction is already started and has
5344 * enough credits) */
5345 static ssize_t ext4_quota_write(struct super_block *sb, int type,
5346 const char *data, size_t len, loff_t off)
5348 struct inode *inode = sb_dqopt(sb)->files[type];
5349 ext4_lblk_t blk = off >> EXT4_BLOCK_SIZE_BITS(sb);
5350 int err, offset = off & (sb->s_blocksize - 1);
5351 struct buffer_head *bh;
5352 handle_t *handle = journal_current_handle();
5354 if (EXT4_SB(sb)->s_journal && !handle) {
5355 ext4_msg(sb, KERN_WARNING, "Quota write (off=%llu, len=%llu)"
5356 " cancelled because transaction is not started",
5357 (unsigned long long)off, (unsigned long long)len);
5361 * Since we account only one data block in transaction credits,
5362 * then it is impossible to cross a block boundary.
5364 if (sb->s_blocksize - offset < len) {
5365 ext4_msg(sb, KERN_WARNING, "Quota write (off=%llu, len=%llu)"
5366 " cancelled because not block aligned",
5367 (unsigned long long)off, (unsigned long long)len);
5371 bh = ext4_bread(handle, inode, blk, 1);
5376 BUFFER_TRACE(bh, "get write access");
5377 err = ext4_journal_get_write_access(handle, bh);
5383 memcpy(bh->b_data+offset, data, len);
5384 flush_dcache_page(bh->b_page);
5386 err = ext4_handle_dirty_metadata(handle, NULL, bh);
5389 if (inode->i_size < off + len) {
5390 i_size_write(inode, off + len);
5391 EXT4_I(inode)->i_disksize = inode->i_size;
5392 ext4_mark_inode_dirty(handle, inode);
5399 static struct dentry *ext4_mount(struct file_system_type *fs_type, int flags,
5400 const char *dev_name, void *data)
5402 return mount_bdev(fs_type, flags, dev_name, data, ext4_fill_super);
5405 #if !defined(CONFIG_EXT2_FS) && !defined(CONFIG_EXT2_FS_MODULE) && defined(CONFIG_EXT4_USE_FOR_EXT23)
5406 static inline void register_as_ext2(void)
5408 int err = register_filesystem(&ext2_fs_type);
5411 "EXT4-fs: Unable to register as ext2 (%d)\n", err);
5414 static inline void unregister_as_ext2(void)
5416 unregister_filesystem(&ext2_fs_type);
5419 static inline int ext2_feature_set_ok(struct super_block *sb)
5421 if (EXT4_HAS_INCOMPAT_FEATURE(sb, ~EXT2_FEATURE_INCOMPAT_SUPP))
5423 if (sb->s_flags & MS_RDONLY)
5425 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, ~EXT2_FEATURE_RO_COMPAT_SUPP))
5430 static inline void register_as_ext2(void) { }
5431 static inline void unregister_as_ext2(void) { }
5432 static inline int ext2_feature_set_ok(struct super_block *sb) { return 0; }
5435 #if !defined(CONFIG_EXT3_FS) && !defined(CONFIG_EXT3_FS_MODULE) && defined(CONFIG_EXT4_USE_FOR_EXT23)
5436 static inline void register_as_ext3(void)
5438 int err = register_filesystem(&ext3_fs_type);
5441 "EXT4-fs: Unable to register as ext3 (%d)\n", err);
5444 static inline void unregister_as_ext3(void)
5446 unregister_filesystem(&ext3_fs_type);
5449 static inline int ext3_feature_set_ok(struct super_block *sb)
5451 if (EXT4_HAS_INCOMPAT_FEATURE(sb, ~EXT3_FEATURE_INCOMPAT_SUPP))
5453 if (!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL))
5455 if (sb->s_flags & MS_RDONLY)
5457 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, ~EXT3_FEATURE_RO_COMPAT_SUPP))
5462 static inline void register_as_ext3(void) { }
5463 static inline void unregister_as_ext3(void) { }
5464 static inline int ext3_feature_set_ok(struct super_block *sb) { return 0; }
5467 static struct file_system_type ext4_fs_type = {
5468 .owner = THIS_MODULE,
5470 .mount = ext4_mount,
5471 .kill_sb = kill_block_super,
5472 .fs_flags = FS_REQUIRES_DEV,
5474 MODULE_ALIAS_FS("ext4");
5476 static int __init ext4_init_feat_adverts(void)
5478 struct ext4_features *ef;
5481 ef = kzalloc(sizeof(struct ext4_features), GFP_KERNEL);
5485 ef->f_kobj.kset = ext4_kset;
5486 init_completion(&ef->f_kobj_unregister);
5487 ret = kobject_init_and_add(&ef->f_kobj, &ext4_feat_ktype, NULL,
5500 static void ext4_exit_feat_adverts(void)
5502 kobject_put(&ext4_feat->f_kobj);
5503 wait_for_completion(&ext4_feat->f_kobj_unregister);
5507 /* Shared across all ext4 file systems */
5508 wait_queue_head_t ext4__ioend_wq[EXT4_WQ_HASH_SZ];
5509 struct mutex ext4__aio_mutex[EXT4_WQ_HASH_SZ];
5511 static int __init ext4_init_fs(void)
5515 ext4_li_info = NULL;
5516 mutex_init(&ext4_li_mtx);
5518 /* Build-time check for flags consistency */
5519 ext4_check_flag_values();
5521 for (i = 0; i < EXT4_WQ_HASH_SZ; i++) {
5522 mutex_init(&ext4__aio_mutex[i]);
5523 init_waitqueue_head(&ext4__ioend_wq[i]);
5526 err = ext4_init_es();
5530 err = ext4_init_pageio();
5534 err = ext4_init_system_zone();
5537 ext4_kset = kset_create_and_add("ext4", NULL, fs_kobj);
5542 ext4_proc_root = proc_mkdir("fs/ext4", NULL);
5544 err = ext4_init_feat_adverts();
5548 err = ext4_init_mballoc();
5552 ext4_mballoc_ready = 1;
5553 err = init_inodecache();
5558 err = register_filesystem(&ext4_fs_type);
5564 unregister_as_ext2();
5565 unregister_as_ext3();
5566 destroy_inodecache();
5568 ext4_mballoc_ready = 0;
5569 ext4_exit_mballoc();
5571 ext4_exit_feat_adverts();
5574 remove_proc_entry("fs/ext4", NULL);
5575 kset_unregister(ext4_kset);
5577 ext4_exit_system_zone();
5586 static void __exit ext4_exit_fs(void)
5588 ext4_destroy_lazyinit_thread();
5589 unregister_as_ext2();
5590 unregister_as_ext3();
5591 unregister_filesystem(&ext4_fs_type);
5592 destroy_inodecache();
5593 ext4_exit_mballoc();
5594 ext4_exit_feat_adverts();
5595 remove_proc_entry("fs/ext4", NULL);
5596 kset_unregister(ext4_kset);
5597 ext4_exit_system_zone();
5602 MODULE_AUTHOR("Remy Card, Stephen Tweedie, Andrew Morton, Andreas Dilger, Theodore Ts'o and others");
5603 MODULE_DESCRIPTION("Fourth Extended Filesystem");
5604 MODULE_LICENSE("GPL");
5605 module_init(ext4_init_fs)
5606 module_exit(ext4_exit_fs)