2 * linux/fs/ext4/super.c
4 * Copyright (C) 1992, 1993, 1994, 1995
5 * Remy Card (card@masi.ibp.fr)
6 * Laboratoire MASI - Institut Blaise Pascal
7 * Universite Pierre et Marie Curie (Paris VI)
11 * linux/fs/minix/inode.c
13 * Copyright (C) 1991, 1992 Linus Torvalds
15 * Big-endian to little-endian byte-swapping/bitmaps by
16 * David S. Miller (davem@caip.rutgers.edu), 1995
19 #include <linux/module.h>
20 #include <linux/string.h>
22 #include <linux/time.h>
23 #include <linux/vmalloc.h>
24 #include <linux/slab.h>
25 #include <linux/init.h>
26 #include <linux/blkdev.h>
27 #include <linux/parser.h>
28 #include <linux/buffer_head.h>
29 #include <linux/exportfs.h>
30 #include <linux/vfs.h>
31 #include <linux/random.h>
32 #include <linux/mount.h>
33 #include <linux/namei.h>
34 #include <linux/quotaops.h>
35 #include <linux/seq_file.h>
36 #include <linux/proc_fs.h>
37 #include <linux/ctype.h>
38 #include <linux/log2.h>
39 #include <linux/crc16.h>
40 #include <linux/cleancache.h>
41 #include <asm/uaccess.h>
43 #include <linux/kthread.h>
44 #include <linux/freezer.h>
47 #include "ext4_extents.h" /* Needed for trace points definition */
48 #include "ext4_jbd2.h"
53 #define CREATE_TRACE_POINTS
54 #include <trace/events/ext4.h>
56 static struct proc_dir_entry *ext4_proc_root;
57 static struct kset *ext4_kset;
58 static struct ext4_lazy_init *ext4_li_info;
59 static struct mutex ext4_li_mtx;
60 static struct ext4_features *ext4_feat;
61 static int ext4_mballoc_ready;
63 static int ext4_load_journal(struct super_block *, struct ext4_super_block *,
64 unsigned long journal_devnum);
65 static int ext4_show_options(struct seq_file *seq, struct dentry *root);
66 static int ext4_commit_super(struct super_block *sb, int sync);
67 static void ext4_mark_recovery_complete(struct super_block *sb,
68 struct ext4_super_block *es);
69 static void ext4_clear_journal_err(struct super_block *sb,
70 struct ext4_super_block *es);
71 static int ext4_sync_fs(struct super_block *sb, int wait);
72 static int ext4_remount(struct super_block *sb, int *flags, char *data);
73 static int ext4_statfs(struct dentry *dentry, struct kstatfs *buf);
74 static int ext4_unfreeze(struct super_block *sb);
75 static int ext4_freeze(struct super_block *sb);
76 static struct dentry *ext4_mount(struct file_system_type *fs_type, int flags,
77 const char *dev_name, void *data);
78 static inline int ext2_feature_set_ok(struct super_block *sb);
79 static inline int ext3_feature_set_ok(struct super_block *sb);
80 static int ext4_feature_set_ok(struct super_block *sb, int readonly);
81 static void ext4_destroy_lazyinit_thread(void);
82 static void ext4_unregister_li_request(struct super_block *sb);
83 static void ext4_clear_request_list(void);
84 static int ext4_reserve_clusters(struct ext4_sb_info *, ext4_fsblk_t);
86 #if !defined(CONFIG_EXT2_FS) && !defined(CONFIG_EXT2_FS_MODULE) && defined(CONFIG_EXT4_USE_FOR_EXT23)
87 static struct file_system_type ext2_fs_type = {
91 .kill_sb = kill_block_super,
92 .fs_flags = FS_REQUIRES_DEV,
94 MODULE_ALIAS_FS("ext2");
96 #define IS_EXT2_SB(sb) ((sb)->s_bdev->bd_holder == &ext2_fs_type)
98 #define IS_EXT2_SB(sb) (0)
102 #if !defined(CONFIG_EXT3_FS) && !defined(CONFIG_EXT3_FS_MODULE) && defined(CONFIG_EXT4_USE_FOR_EXT23)
103 static struct file_system_type ext3_fs_type = {
104 .owner = THIS_MODULE,
107 .kill_sb = kill_block_super,
108 .fs_flags = FS_REQUIRES_DEV,
110 MODULE_ALIAS_FS("ext3");
111 MODULE_ALIAS("ext3");
112 #define IS_EXT3_SB(sb) ((sb)->s_bdev->bd_holder == &ext3_fs_type)
114 #define IS_EXT3_SB(sb) (0)
117 static int ext4_verify_csum_type(struct super_block *sb,
118 struct ext4_super_block *es)
120 if (!EXT4_HAS_RO_COMPAT_FEATURE(sb,
121 EXT4_FEATURE_RO_COMPAT_METADATA_CSUM))
124 return es->s_checksum_type == EXT4_CRC32C_CHKSUM;
127 static __le32 ext4_superblock_csum(struct super_block *sb,
128 struct ext4_super_block *es)
130 struct ext4_sb_info *sbi = EXT4_SB(sb);
131 int offset = offsetof(struct ext4_super_block, s_checksum);
134 csum = ext4_chksum(sbi, ~0, (char *)es, offset);
136 return cpu_to_le32(csum);
139 static int ext4_superblock_csum_verify(struct super_block *sb,
140 struct ext4_super_block *es)
142 if (!ext4_has_metadata_csum(sb))
145 return es->s_checksum == ext4_superblock_csum(sb, es);
148 void ext4_superblock_csum_set(struct super_block *sb)
150 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
152 if (!ext4_has_metadata_csum(sb))
155 es->s_checksum = ext4_superblock_csum(sb, es);
158 void *ext4_kvmalloc(size_t size, gfp_t flags)
162 ret = kmalloc(size, flags | __GFP_NOWARN);
164 ret = __vmalloc(size, flags, PAGE_KERNEL);
168 void *ext4_kvzalloc(size_t size, gfp_t flags)
172 ret = kzalloc(size, flags | __GFP_NOWARN);
174 ret = __vmalloc(size, flags | __GFP_ZERO, PAGE_KERNEL);
178 ext4_fsblk_t ext4_block_bitmap(struct super_block *sb,
179 struct ext4_group_desc *bg)
181 return le32_to_cpu(bg->bg_block_bitmap_lo) |
182 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
183 (ext4_fsblk_t)le32_to_cpu(bg->bg_block_bitmap_hi) << 32 : 0);
186 ext4_fsblk_t ext4_inode_bitmap(struct super_block *sb,
187 struct ext4_group_desc *bg)
189 return le32_to_cpu(bg->bg_inode_bitmap_lo) |
190 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
191 (ext4_fsblk_t)le32_to_cpu(bg->bg_inode_bitmap_hi) << 32 : 0);
194 ext4_fsblk_t ext4_inode_table(struct super_block *sb,
195 struct ext4_group_desc *bg)
197 return le32_to_cpu(bg->bg_inode_table_lo) |
198 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
199 (ext4_fsblk_t)le32_to_cpu(bg->bg_inode_table_hi) << 32 : 0);
202 __u32 ext4_free_group_clusters(struct super_block *sb,
203 struct ext4_group_desc *bg)
205 return le16_to_cpu(bg->bg_free_blocks_count_lo) |
206 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
207 (__u32)le16_to_cpu(bg->bg_free_blocks_count_hi) << 16 : 0);
210 __u32 ext4_free_inodes_count(struct super_block *sb,
211 struct ext4_group_desc *bg)
213 return le16_to_cpu(bg->bg_free_inodes_count_lo) |
214 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
215 (__u32)le16_to_cpu(bg->bg_free_inodes_count_hi) << 16 : 0);
218 __u32 ext4_used_dirs_count(struct super_block *sb,
219 struct ext4_group_desc *bg)
221 return le16_to_cpu(bg->bg_used_dirs_count_lo) |
222 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
223 (__u32)le16_to_cpu(bg->bg_used_dirs_count_hi) << 16 : 0);
226 __u32 ext4_itable_unused_count(struct super_block *sb,
227 struct ext4_group_desc *bg)
229 return le16_to_cpu(bg->bg_itable_unused_lo) |
230 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
231 (__u32)le16_to_cpu(bg->bg_itable_unused_hi) << 16 : 0);
234 void ext4_block_bitmap_set(struct super_block *sb,
235 struct ext4_group_desc *bg, ext4_fsblk_t blk)
237 bg->bg_block_bitmap_lo = cpu_to_le32((u32)blk);
238 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
239 bg->bg_block_bitmap_hi = cpu_to_le32(blk >> 32);
242 void ext4_inode_bitmap_set(struct super_block *sb,
243 struct ext4_group_desc *bg, ext4_fsblk_t blk)
245 bg->bg_inode_bitmap_lo = cpu_to_le32((u32)blk);
246 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
247 bg->bg_inode_bitmap_hi = cpu_to_le32(blk >> 32);
250 void ext4_inode_table_set(struct super_block *sb,
251 struct ext4_group_desc *bg, ext4_fsblk_t blk)
253 bg->bg_inode_table_lo = cpu_to_le32((u32)blk);
254 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
255 bg->bg_inode_table_hi = cpu_to_le32(blk >> 32);
258 void ext4_free_group_clusters_set(struct super_block *sb,
259 struct ext4_group_desc *bg, __u32 count)
261 bg->bg_free_blocks_count_lo = cpu_to_le16((__u16)count);
262 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
263 bg->bg_free_blocks_count_hi = cpu_to_le16(count >> 16);
266 void ext4_free_inodes_set(struct super_block *sb,
267 struct ext4_group_desc *bg, __u32 count)
269 bg->bg_free_inodes_count_lo = cpu_to_le16((__u16)count);
270 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
271 bg->bg_free_inodes_count_hi = cpu_to_le16(count >> 16);
274 void ext4_used_dirs_set(struct super_block *sb,
275 struct ext4_group_desc *bg, __u32 count)
277 bg->bg_used_dirs_count_lo = cpu_to_le16((__u16)count);
278 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
279 bg->bg_used_dirs_count_hi = cpu_to_le16(count >> 16);
282 void ext4_itable_unused_set(struct super_block *sb,
283 struct ext4_group_desc *bg, __u32 count)
285 bg->bg_itable_unused_lo = cpu_to_le16((__u16)count);
286 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
287 bg->bg_itable_unused_hi = cpu_to_le16(count >> 16);
291 static void __save_error_info(struct super_block *sb, const char *func,
294 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
296 EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS;
297 if (bdev_read_only(sb->s_bdev))
299 es->s_state |= cpu_to_le16(EXT4_ERROR_FS);
300 es->s_last_error_time = cpu_to_le32(get_seconds());
301 strncpy(es->s_last_error_func, func, sizeof(es->s_last_error_func));
302 es->s_last_error_line = cpu_to_le32(line);
303 if (!es->s_first_error_time) {
304 es->s_first_error_time = es->s_last_error_time;
305 strncpy(es->s_first_error_func, func,
306 sizeof(es->s_first_error_func));
307 es->s_first_error_line = cpu_to_le32(line);
308 es->s_first_error_ino = es->s_last_error_ino;
309 es->s_first_error_block = es->s_last_error_block;
312 * Start the daily error reporting function if it hasn't been
315 if (!es->s_error_count)
316 mod_timer(&EXT4_SB(sb)->s_err_report, jiffies + 24*60*60*HZ);
317 le32_add_cpu(&es->s_error_count, 1);
320 static void save_error_info(struct super_block *sb, const char *func,
323 __save_error_info(sb, func, line);
324 ext4_commit_super(sb, 1);
327 static void ext4_journal_commit_callback(journal_t *journal, transaction_t *txn)
329 struct super_block *sb = journal->j_private;
330 struct ext4_sb_info *sbi = EXT4_SB(sb);
331 int error = is_journal_aborted(journal);
332 struct ext4_journal_cb_entry *jce;
334 BUG_ON(txn->t_state == T_FINISHED);
335 spin_lock(&sbi->s_md_lock);
336 while (!list_empty(&txn->t_private_list)) {
337 jce = list_entry(txn->t_private_list.next,
338 struct ext4_journal_cb_entry, jce_list);
339 list_del_init(&jce->jce_list);
340 spin_unlock(&sbi->s_md_lock);
341 jce->jce_func(sb, jce, error);
342 spin_lock(&sbi->s_md_lock);
344 spin_unlock(&sbi->s_md_lock);
347 /* Deal with the reporting of failure conditions on a filesystem such as
348 * inconsistencies detected or read IO failures.
350 * On ext2, we can store the error state of the filesystem in the
351 * superblock. That is not possible on ext4, because we may have other
352 * write ordering constraints on the superblock which prevent us from
353 * writing it out straight away; and given that the journal is about to
354 * be aborted, we can't rely on the current, or future, transactions to
355 * write out the superblock safely.
357 * We'll just use the jbd2_journal_abort() error code to record an error in
358 * the journal instead. On recovery, the journal will complain about
359 * that error until we've noted it down and cleared it.
362 static void ext4_handle_error(struct super_block *sb)
364 if (sb->s_flags & MS_RDONLY)
367 if (!test_opt(sb, ERRORS_CONT)) {
368 journal_t *journal = EXT4_SB(sb)->s_journal;
370 EXT4_SB(sb)->s_mount_flags |= EXT4_MF_FS_ABORTED;
372 jbd2_journal_abort(journal, -EIO);
374 if (test_opt(sb, ERRORS_RO)) {
375 ext4_msg(sb, KERN_CRIT, "Remounting filesystem read-only");
377 * Make sure updated value of ->s_mount_flags will be visible
378 * before ->s_flags update
381 sb->s_flags |= MS_RDONLY;
383 if (test_opt(sb, ERRORS_PANIC))
384 panic("EXT4-fs (device %s): panic forced after error\n",
388 #define ext4_error_ratelimit(sb) \
389 ___ratelimit(&(EXT4_SB(sb)->s_err_ratelimit_state), \
392 void __ext4_error(struct super_block *sb, const char *function,
393 unsigned int line, const char *fmt, ...)
395 struct va_format vaf;
398 if (ext4_error_ratelimit(sb)) {
403 "EXT4-fs error (device %s): %s:%d: comm %s: %pV\n",
404 sb->s_id, function, line, current->comm, &vaf);
407 save_error_info(sb, function, line);
408 ext4_handle_error(sb);
411 void __ext4_error_inode(struct inode *inode, const char *function,
412 unsigned int line, ext4_fsblk_t block,
413 const char *fmt, ...)
416 struct va_format vaf;
417 struct ext4_super_block *es = EXT4_SB(inode->i_sb)->s_es;
419 es->s_last_error_ino = cpu_to_le32(inode->i_ino);
420 es->s_last_error_block = cpu_to_le64(block);
421 if (ext4_error_ratelimit(inode->i_sb)) {
426 printk(KERN_CRIT "EXT4-fs error (device %s): %s:%d: "
427 "inode #%lu: block %llu: comm %s: %pV\n",
428 inode->i_sb->s_id, function, line, inode->i_ino,
429 block, current->comm, &vaf);
431 printk(KERN_CRIT "EXT4-fs error (device %s): %s:%d: "
432 "inode #%lu: comm %s: %pV\n",
433 inode->i_sb->s_id, function, line, inode->i_ino,
434 current->comm, &vaf);
437 save_error_info(inode->i_sb, function, line);
438 ext4_handle_error(inode->i_sb);
441 void __ext4_error_file(struct file *file, const char *function,
442 unsigned int line, ext4_fsblk_t block,
443 const char *fmt, ...)
446 struct va_format vaf;
447 struct ext4_super_block *es;
448 struct inode *inode = file_inode(file);
449 char pathname[80], *path;
451 es = EXT4_SB(inode->i_sb)->s_es;
452 es->s_last_error_ino = cpu_to_le32(inode->i_ino);
453 if (ext4_error_ratelimit(inode->i_sb)) {
454 path = d_path(&(file->f_path), pathname, sizeof(pathname));
462 "EXT4-fs error (device %s): %s:%d: inode #%lu: "
463 "block %llu: comm %s: path %s: %pV\n",
464 inode->i_sb->s_id, function, line, inode->i_ino,
465 block, current->comm, path, &vaf);
468 "EXT4-fs error (device %s): %s:%d: inode #%lu: "
469 "comm %s: path %s: %pV\n",
470 inode->i_sb->s_id, function, line, inode->i_ino,
471 current->comm, path, &vaf);
474 save_error_info(inode->i_sb, function, line);
475 ext4_handle_error(inode->i_sb);
478 const char *ext4_decode_error(struct super_block *sb, int errno,
485 errstr = "IO failure";
488 errstr = "Out of memory";
491 if (!sb || (EXT4_SB(sb)->s_journal &&
492 EXT4_SB(sb)->s_journal->j_flags & JBD2_ABORT))
493 errstr = "Journal has aborted";
495 errstr = "Readonly filesystem";
498 /* If the caller passed in an extra buffer for unknown
499 * errors, textualise them now. Else we just return
502 /* Check for truncated error codes... */
503 if (snprintf(nbuf, 16, "error %d", -errno) >= 0)
512 /* __ext4_std_error decodes expected errors from journaling functions
513 * automatically and invokes the appropriate error response. */
515 void __ext4_std_error(struct super_block *sb, const char *function,
516 unsigned int line, int errno)
521 /* Special case: if the error is EROFS, and we're not already
522 * inside a transaction, then there's really no point in logging
524 if (errno == -EROFS && journal_current_handle() == NULL &&
525 (sb->s_flags & MS_RDONLY))
528 if (ext4_error_ratelimit(sb)) {
529 errstr = ext4_decode_error(sb, errno, nbuf);
530 printk(KERN_CRIT "EXT4-fs error (device %s) in %s:%d: %s\n",
531 sb->s_id, function, line, errstr);
534 save_error_info(sb, function, line);
535 ext4_handle_error(sb);
539 * ext4_abort is a much stronger failure handler than ext4_error. The
540 * abort function may be used to deal with unrecoverable failures such
541 * as journal IO errors or ENOMEM at a critical moment in log management.
543 * We unconditionally force the filesystem into an ABORT|READONLY state,
544 * unless the error response on the fs has been set to panic in which
545 * case we take the easy way out and panic immediately.
548 void __ext4_abort(struct super_block *sb, const char *function,
549 unsigned int line, const char *fmt, ...)
553 save_error_info(sb, function, line);
555 printk(KERN_CRIT "EXT4-fs error (device %s): %s:%d: ", sb->s_id,
561 if ((sb->s_flags & MS_RDONLY) == 0) {
562 ext4_msg(sb, KERN_CRIT, "Remounting filesystem read-only");
563 EXT4_SB(sb)->s_mount_flags |= EXT4_MF_FS_ABORTED;
565 * Make sure updated value of ->s_mount_flags will be visible
566 * before ->s_flags update
569 sb->s_flags |= MS_RDONLY;
570 if (EXT4_SB(sb)->s_journal)
571 jbd2_journal_abort(EXT4_SB(sb)->s_journal, -EIO);
572 save_error_info(sb, function, line);
574 if (test_opt(sb, ERRORS_PANIC))
575 panic("EXT4-fs panic from previous error\n");
578 void __ext4_msg(struct super_block *sb,
579 const char *prefix, const char *fmt, ...)
581 struct va_format vaf;
584 if (!___ratelimit(&(EXT4_SB(sb)->s_msg_ratelimit_state), "EXT4-fs"))
590 printk("%sEXT4-fs (%s): %pV\n", prefix, sb->s_id, &vaf);
594 #define ext4_warning_ratelimit(sb) \
595 ___ratelimit(&(EXT4_SB(sb)->s_warning_ratelimit_state), \
598 void __ext4_warning(struct super_block *sb, const char *function,
599 unsigned int line, const char *fmt, ...)
601 struct va_format vaf;
604 if (!ext4_warning_ratelimit(sb))
610 printk(KERN_WARNING "EXT4-fs warning (device %s): %s:%d: %pV\n",
611 sb->s_id, function, line, &vaf);
615 void __ext4_warning_inode(const struct inode *inode, const char *function,
616 unsigned int line, const char *fmt, ...)
618 struct va_format vaf;
621 if (!ext4_warning_ratelimit(inode->i_sb))
627 printk(KERN_WARNING "EXT4-fs warning (device %s): %s:%d: "
628 "inode #%lu: comm %s: %pV\n", inode->i_sb->s_id,
629 function, line, inode->i_ino, current->comm, &vaf);
633 void __ext4_grp_locked_error(const char *function, unsigned int line,
634 struct super_block *sb, ext4_group_t grp,
635 unsigned long ino, ext4_fsblk_t block,
636 const char *fmt, ...)
640 struct va_format vaf;
642 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
644 es->s_last_error_ino = cpu_to_le32(ino);
645 es->s_last_error_block = cpu_to_le64(block);
646 __save_error_info(sb, function, line);
648 if (ext4_error_ratelimit(sb)) {
652 printk(KERN_CRIT "EXT4-fs error (device %s): %s:%d: group %u, ",
653 sb->s_id, function, line, grp);
655 printk(KERN_CONT "inode %lu: ", ino);
657 printk(KERN_CONT "block %llu:",
658 (unsigned long long) block);
659 printk(KERN_CONT "%pV\n", &vaf);
663 if (test_opt(sb, ERRORS_CONT)) {
664 ext4_commit_super(sb, 0);
668 ext4_unlock_group(sb, grp);
669 ext4_handle_error(sb);
671 * We only get here in the ERRORS_RO case; relocking the group
672 * may be dangerous, but nothing bad will happen since the
673 * filesystem will have already been marked read/only and the
674 * journal has been aborted. We return 1 as a hint to callers
675 * who might what to use the return value from
676 * ext4_grp_locked_error() to distinguish between the
677 * ERRORS_CONT and ERRORS_RO case, and perhaps return more
678 * aggressively from the ext4 function in question, with a
679 * more appropriate error code.
681 ext4_lock_group(sb, grp);
685 void ext4_update_dynamic_rev(struct super_block *sb)
687 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
689 if (le32_to_cpu(es->s_rev_level) > EXT4_GOOD_OLD_REV)
693 "updating to rev %d because of new feature flag, "
694 "running e2fsck is recommended",
697 es->s_first_ino = cpu_to_le32(EXT4_GOOD_OLD_FIRST_INO);
698 es->s_inode_size = cpu_to_le16(EXT4_GOOD_OLD_INODE_SIZE);
699 es->s_rev_level = cpu_to_le32(EXT4_DYNAMIC_REV);
700 /* leave es->s_feature_*compat flags alone */
701 /* es->s_uuid will be set by e2fsck if empty */
704 * The rest of the superblock fields should be zero, and if not it
705 * means they are likely already in use, so leave them alone. We
706 * can leave it up to e2fsck to clean up any inconsistencies there.
711 * Open the external journal device
713 static struct block_device *ext4_blkdev_get(dev_t dev, struct super_block *sb)
715 struct block_device *bdev;
716 char b[BDEVNAME_SIZE];
718 bdev = blkdev_get_by_dev(dev, FMODE_READ|FMODE_WRITE|FMODE_EXCL, sb);
724 ext4_msg(sb, KERN_ERR, "failed to open journal device %s: %ld",
725 __bdevname(dev, b), PTR_ERR(bdev));
730 * Release the journal device
732 static void ext4_blkdev_put(struct block_device *bdev)
734 blkdev_put(bdev, FMODE_READ|FMODE_WRITE|FMODE_EXCL);
737 static void ext4_blkdev_remove(struct ext4_sb_info *sbi)
739 struct block_device *bdev;
740 bdev = sbi->journal_bdev;
742 ext4_blkdev_put(bdev);
743 sbi->journal_bdev = NULL;
747 static inline struct inode *orphan_list_entry(struct list_head *l)
749 return &list_entry(l, struct ext4_inode_info, i_orphan)->vfs_inode;
752 static void dump_orphan_list(struct super_block *sb, struct ext4_sb_info *sbi)
756 ext4_msg(sb, KERN_ERR, "sb orphan head is %d",
757 le32_to_cpu(sbi->s_es->s_last_orphan));
759 printk(KERN_ERR "sb_info orphan list:\n");
760 list_for_each(l, &sbi->s_orphan) {
761 struct inode *inode = orphan_list_entry(l);
763 "inode %s:%lu at %p: mode %o, nlink %d, next %d\n",
764 inode->i_sb->s_id, inode->i_ino, inode,
765 inode->i_mode, inode->i_nlink,
770 static void ext4_put_super(struct super_block *sb)
772 struct ext4_sb_info *sbi = EXT4_SB(sb);
773 struct ext4_super_block *es = sbi->s_es;
776 ext4_unregister_li_request(sb);
777 dquot_disable(sb, -1, DQUOT_USAGE_ENABLED | DQUOT_LIMITS_ENABLED);
779 flush_workqueue(sbi->rsv_conversion_wq);
780 destroy_workqueue(sbi->rsv_conversion_wq);
782 if (sbi->s_journal) {
783 err = jbd2_journal_destroy(sbi->s_journal);
784 sbi->s_journal = NULL;
786 ext4_abort(sb, "Couldn't clean up the journal");
789 ext4_es_unregister_shrinker(sbi);
790 del_timer_sync(&sbi->s_err_report);
791 ext4_release_system_zone(sb);
793 ext4_ext_release(sb);
794 ext4_xattr_put_super(sb);
796 if (!(sb->s_flags & MS_RDONLY)) {
797 EXT4_CLEAR_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
798 es->s_state = cpu_to_le16(sbi->s_mount_state);
800 if (!(sb->s_flags & MS_RDONLY))
801 ext4_commit_super(sb, 1);
804 remove_proc_entry("options", sbi->s_proc);
805 remove_proc_entry(sb->s_id, ext4_proc_root);
807 kobject_del(&sbi->s_kobj);
809 for (i = 0; i < sbi->s_gdb_count; i++)
810 brelse(sbi->s_group_desc[i]);
811 kvfree(sbi->s_group_desc);
812 kvfree(sbi->s_flex_groups);
813 percpu_counter_destroy(&sbi->s_freeclusters_counter);
814 percpu_counter_destroy(&sbi->s_freeinodes_counter);
815 percpu_counter_destroy(&sbi->s_dirs_counter);
816 percpu_counter_destroy(&sbi->s_dirtyclusters_counter);
819 for (i = 0; i < EXT4_MAXQUOTAS; i++)
820 kfree(sbi->s_qf_names[i]);
823 /* Debugging code just in case the in-memory inode orphan list
824 * isn't empty. The on-disk one can be non-empty if we've
825 * detected an error and taken the fs readonly, but the
826 * in-memory list had better be clean by this point. */
827 if (!list_empty(&sbi->s_orphan))
828 dump_orphan_list(sb, sbi);
829 J_ASSERT(list_empty(&sbi->s_orphan));
831 sync_blockdev(sb->s_bdev);
832 invalidate_bdev(sb->s_bdev);
833 if (sbi->journal_bdev && sbi->journal_bdev != sb->s_bdev) {
835 * Invalidate the journal device's buffers. We don't want them
836 * floating about in memory - the physical journal device may
837 * hotswapped, and it breaks the `ro-after' testing code.
839 sync_blockdev(sbi->journal_bdev);
840 invalidate_bdev(sbi->journal_bdev);
841 ext4_blkdev_remove(sbi);
843 if (sbi->s_mb_cache) {
844 ext4_xattr_destroy_cache(sbi->s_mb_cache);
845 sbi->s_mb_cache = NULL;
848 kthread_stop(sbi->s_mmp_tsk);
849 sb->s_fs_info = NULL;
851 * Now that we are completely done shutting down the
852 * superblock, we need to actually destroy the kobject.
854 kobject_put(&sbi->s_kobj);
855 wait_for_completion(&sbi->s_kobj_unregister);
856 if (sbi->s_chksum_driver)
857 crypto_free_shash(sbi->s_chksum_driver);
858 kfree(sbi->s_blockgroup_lock);
862 static struct kmem_cache *ext4_inode_cachep;
865 * Called inside transaction, so use GFP_NOFS
867 static struct inode *ext4_alloc_inode(struct super_block *sb)
869 struct ext4_inode_info *ei;
871 ei = kmem_cache_alloc(ext4_inode_cachep, GFP_NOFS);
875 ei->vfs_inode.i_version = 1;
876 spin_lock_init(&ei->i_raw_lock);
877 INIT_LIST_HEAD(&ei->i_prealloc_list);
878 spin_lock_init(&ei->i_prealloc_lock);
879 ext4_es_init_tree(&ei->i_es_tree);
880 rwlock_init(&ei->i_es_lock);
881 INIT_LIST_HEAD(&ei->i_es_list);
884 ei->i_es_shrink_lblk = 0;
885 ei->i_reserved_data_blocks = 0;
886 ei->i_reserved_meta_blocks = 0;
887 ei->i_allocated_meta_blocks = 0;
888 ei->i_da_metadata_calc_len = 0;
889 ei->i_da_metadata_calc_last_lblock = 0;
890 spin_lock_init(&(ei->i_block_reservation_lock));
892 ei->i_reserved_quota = 0;
893 memset(&ei->i_dquot, 0, sizeof(ei->i_dquot));
896 INIT_LIST_HEAD(&ei->i_rsv_conversion_list);
897 spin_lock_init(&ei->i_completed_io_lock);
899 ei->i_datasync_tid = 0;
900 atomic_set(&ei->i_ioend_count, 0);
901 atomic_set(&ei->i_unwritten, 0);
902 INIT_WORK(&ei->i_rsv_conversion_work, ext4_end_io_rsv_work);
903 #ifdef CONFIG_EXT4_FS_ENCRYPTION
904 ei->i_crypt_info = NULL;
906 return &ei->vfs_inode;
909 static int ext4_drop_inode(struct inode *inode)
911 int drop = generic_drop_inode(inode);
913 trace_ext4_drop_inode(inode, drop);
917 static void ext4_i_callback(struct rcu_head *head)
919 struct inode *inode = container_of(head, struct inode, i_rcu);
920 kmem_cache_free(ext4_inode_cachep, EXT4_I(inode));
923 static void ext4_destroy_inode(struct inode *inode)
925 if (!list_empty(&(EXT4_I(inode)->i_orphan))) {
926 ext4_msg(inode->i_sb, KERN_ERR,
927 "Inode %lu (%p): orphan list check failed!",
928 inode->i_ino, EXT4_I(inode));
929 print_hex_dump(KERN_INFO, "", DUMP_PREFIX_ADDRESS, 16, 4,
930 EXT4_I(inode), sizeof(struct ext4_inode_info),
934 call_rcu(&inode->i_rcu, ext4_i_callback);
937 static void init_once(void *foo)
939 struct ext4_inode_info *ei = (struct ext4_inode_info *) foo;
941 INIT_LIST_HEAD(&ei->i_orphan);
942 init_rwsem(&ei->xattr_sem);
943 init_rwsem(&ei->i_data_sem);
944 inode_init_once(&ei->vfs_inode);
947 static int __init init_inodecache(void)
949 ext4_inode_cachep = kmem_cache_create("ext4_inode_cache",
950 sizeof(struct ext4_inode_info),
951 0, (SLAB_RECLAIM_ACCOUNT|
954 if (ext4_inode_cachep == NULL)
959 static void destroy_inodecache(void)
962 * Make sure all delayed rcu free inodes are flushed before we
966 kmem_cache_destroy(ext4_inode_cachep);
969 void ext4_clear_inode(struct inode *inode)
971 invalidate_inode_buffers(inode);
974 ext4_discard_preallocations(inode);
975 ext4_es_remove_extent(inode, 0, EXT_MAX_BLOCKS);
976 if (EXT4_I(inode)->jinode) {
977 jbd2_journal_release_jbd_inode(EXT4_JOURNAL(inode),
978 EXT4_I(inode)->jinode);
979 jbd2_free_inode(EXT4_I(inode)->jinode);
980 EXT4_I(inode)->jinode = NULL;
982 #ifdef CONFIG_EXT4_FS_ENCRYPTION
983 if (EXT4_I(inode)->i_crypt_info)
984 ext4_free_encryption_info(inode, EXT4_I(inode)->i_crypt_info);
988 static struct inode *ext4_nfs_get_inode(struct super_block *sb,
989 u64 ino, u32 generation)
993 if (ino < EXT4_FIRST_INO(sb) && ino != EXT4_ROOT_INO)
994 return ERR_PTR(-ESTALE);
995 if (ino > le32_to_cpu(EXT4_SB(sb)->s_es->s_inodes_count))
996 return ERR_PTR(-ESTALE);
998 /* iget isn't really right if the inode is currently unallocated!!
1000 * ext4_read_inode will return a bad_inode if the inode had been
1001 * deleted, so we should be safe.
1003 * Currently we don't know the generation for parent directory, so
1004 * a generation of 0 means "accept any"
1006 inode = ext4_iget_normal(sb, ino);
1008 return ERR_CAST(inode);
1009 if (generation && inode->i_generation != generation) {
1011 return ERR_PTR(-ESTALE);
1017 static struct dentry *ext4_fh_to_dentry(struct super_block *sb, struct fid *fid,
1018 int fh_len, int fh_type)
1020 return generic_fh_to_dentry(sb, fid, fh_len, fh_type,
1021 ext4_nfs_get_inode);
1024 static struct dentry *ext4_fh_to_parent(struct super_block *sb, struct fid *fid,
1025 int fh_len, int fh_type)
1027 return generic_fh_to_parent(sb, fid, fh_len, fh_type,
1028 ext4_nfs_get_inode);
1032 * Try to release metadata pages (indirect blocks, directories) which are
1033 * mapped via the block device. Since these pages could have journal heads
1034 * which would prevent try_to_free_buffers() from freeing them, we must use
1035 * jbd2 layer's try_to_free_buffers() function to release them.
1037 static int bdev_try_to_free_page(struct super_block *sb, struct page *page,
1040 journal_t *journal = EXT4_SB(sb)->s_journal;
1042 WARN_ON(PageChecked(page));
1043 if (!page_has_buffers(page))
1046 return jbd2_journal_try_to_free_buffers(journal, page,
1047 wait & ~__GFP_WAIT);
1048 return try_to_free_buffers(page);
1052 #define QTYPE2NAME(t) ((t) == USRQUOTA ? "user" : "group")
1053 #define QTYPE2MOPT(on, t) ((t) == USRQUOTA?((on)##USRJQUOTA):((on)##GRPJQUOTA))
1055 static int ext4_write_dquot(struct dquot *dquot);
1056 static int ext4_acquire_dquot(struct dquot *dquot);
1057 static int ext4_release_dquot(struct dquot *dquot);
1058 static int ext4_mark_dquot_dirty(struct dquot *dquot);
1059 static int ext4_write_info(struct super_block *sb, int type);
1060 static int ext4_quota_on(struct super_block *sb, int type, int format_id,
1062 static int ext4_quota_off(struct super_block *sb, int type);
1063 static int ext4_quota_on_mount(struct super_block *sb, int type);
1064 static ssize_t ext4_quota_read(struct super_block *sb, int type, char *data,
1065 size_t len, loff_t off);
1066 static ssize_t ext4_quota_write(struct super_block *sb, int type,
1067 const char *data, size_t len, loff_t off);
1068 static int ext4_quota_enable(struct super_block *sb, int type, int format_id,
1069 unsigned int flags);
1070 static int ext4_enable_quotas(struct super_block *sb);
1072 static struct dquot **ext4_get_dquots(struct inode *inode)
1074 return EXT4_I(inode)->i_dquot;
1077 static const struct dquot_operations ext4_quota_operations = {
1078 .get_reserved_space = ext4_get_reserved_space,
1079 .write_dquot = ext4_write_dquot,
1080 .acquire_dquot = ext4_acquire_dquot,
1081 .release_dquot = ext4_release_dquot,
1082 .mark_dirty = ext4_mark_dquot_dirty,
1083 .write_info = ext4_write_info,
1084 .alloc_dquot = dquot_alloc,
1085 .destroy_dquot = dquot_destroy,
1088 static const struct quotactl_ops ext4_qctl_operations = {
1089 .quota_on = ext4_quota_on,
1090 .quota_off = ext4_quota_off,
1091 .quota_sync = dquot_quota_sync,
1092 .get_state = dquot_get_state,
1093 .set_info = dquot_set_dqinfo,
1094 .get_dqblk = dquot_get_dqblk,
1095 .set_dqblk = dquot_set_dqblk
1099 static const struct super_operations ext4_sops = {
1100 .alloc_inode = ext4_alloc_inode,
1101 .destroy_inode = ext4_destroy_inode,
1102 .write_inode = ext4_write_inode,
1103 .dirty_inode = ext4_dirty_inode,
1104 .drop_inode = ext4_drop_inode,
1105 .evict_inode = ext4_evict_inode,
1106 .put_super = ext4_put_super,
1107 .sync_fs = ext4_sync_fs,
1108 .freeze_fs = ext4_freeze,
1109 .unfreeze_fs = ext4_unfreeze,
1110 .statfs = ext4_statfs,
1111 .remount_fs = ext4_remount,
1112 .show_options = ext4_show_options,
1114 .quota_read = ext4_quota_read,
1115 .quota_write = ext4_quota_write,
1116 .get_dquots = ext4_get_dquots,
1118 .bdev_try_to_free_page = bdev_try_to_free_page,
1121 static const struct export_operations ext4_export_ops = {
1122 .fh_to_dentry = ext4_fh_to_dentry,
1123 .fh_to_parent = ext4_fh_to_parent,
1124 .get_parent = ext4_get_parent,
1128 Opt_bsd_df, Opt_minix_df, Opt_grpid, Opt_nogrpid,
1129 Opt_resgid, Opt_resuid, Opt_sb, Opt_err_cont, Opt_err_panic, Opt_err_ro,
1130 Opt_nouid32, Opt_debug, Opt_removed,
1131 Opt_user_xattr, Opt_nouser_xattr, Opt_acl, Opt_noacl,
1132 Opt_auto_da_alloc, Opt_noauto_da_alloc, Opt_noload,
1133 Opt_commit, Opt_min_batch_time, Opt_max_batch_time, Opt_journal_dev,
1134 Opt_journal_path, Opt_journal_checksum, Opt_journal_async_commit,
1135 Opt_abort, Opt_data_journal, Opt_data_ordered, Opt_data_writeback,
1136 Opt_data_err_abort, Opt_data_err_ignore, Opt_test_dummy_encryption,
1137 Opt_usrjquota, Opt_grpjquota, Opt_offusrjquota, Opt_offgrpjquota,
1138 Opt_jqfmt_vfsold, Opt_jqfmt_vfsv0, Opt_jqfmt_vfsv1, Opt_quota,
1139 Opt_noquota, Opt_barrier, Opt_nobarrier, Opt_err,
1140 Opt_usrquota, Opt_grpquota, Opt_i_version, Opt_dax,
1141 Opt_stripe, Opt_delalloc, Opt_nodelalloc, Opt_mblk_io_submit,
1142 Opt_lazytime, Opt_nolazytime,
1143 Opt_nomblk_io_submit, Opt_block_validity, Opt_noblock_validity,
1144 Opt_inode_readahead_blks, Opt_journal_ioprio,
1145 Opt_dioread_nolock, Opt_dioread_lock,
1146 Opt_discard, Opt_nodiscard, Opt_init_itable, Opt_noinit_itable,
1147 Opt_max_dir_size_kb, Opt_nojournal_checksum,
1150 static const match_table_t tokens = {
1151 {Opt_bsd_df, "bsddf"},
1152 {Opt_minix_df, "minixdf"},
1153 {Opt_grpid, "grpid"},
1154 {Opt_grpid, "bsdgroups"},
1155 {Opt_nogrpid, "nogrpid"},
1156 {Opt_nogrpid, "sysvgroups"},
1157 {Opt_resgid, "resgid=%u"},
1158 {Opt_resuid, "resuid=%u"},
1160 {Opt_err_cont, "errors=continue"},
1161 {Opt_err_panic, "errors=panic"},
1162 {Opt_err_ro, "errors=remount-ro"},
1163 {Opt_nouid32, "nouid32"},
1164 {Opt_debug, "debug"},
1165 {Opt_removed, "oldalloc"},
1166 {Opt_removed, "orlov"},
1167 {Opt_user_xattr, "user_xattr"},
1168 {Opt_nouser_xattr, "nouser_xattr"},
1170 {Opt_noacl, "noacl"},
1171 {Opt_noload, "norecovery"},
1172 {Opt_noload, "noload"},
1173 {Opt_removed, "nobh"},
1174 {Opt_removed, "bh"},
1175 {Opt_commit, "commit=%u"},
1176 {Opt_min_batch_time, "min_batch_time=%u"},
1177 {Opt_max_batch_time, "max_batch_time=%u"},
1178 {Opt_journal_dev, "journal_dev=%u"},
1179 {Opt_journal_path, "journal_path=%s"},
1180 {Opt_journal_checksum, "journal_checksum"},
1181 {Opt_nojournal_checksum, "nojournal_checksum"},
1182 {Opt_journal_async_commit, "journal_async_commit"},
1183 {Opt_abort, "abort"},
1184 {Opt_data_journal, "data=journal"},
1185 {Opt_data_ordered, "data=ordered"},
1186 {Opt_data_writeback, "data=writeback"},
1187 {Opt_data_err_abort, "data_err=abort"},
1188 {Opt_data_err_ignore, "data_err=ignore"},
1189 {Opt_offusrjquota, "usrjquota="},
1190 {Opt_usrjquota, "usrjquota=%s"},
1191 {Opt_offgrpjquota, "grpjquota="},
1192 {Opt_grpjquota, "grpjquota=%s"},
1193 {Opt_jqfmt_vfsold, "jqfmt=vfsold"},
1194 {Opt_jqfmt_vfsv0, "jqfmt=vfsv0"},
1195 {Opt_jqfmt_vfsv1, "jqfmt=vfsv1"},
1196 {Opt_grpquota, "grpquota"},
1197 {Opt_noquota, "noquota"},
1198 {Opt_quota, "quota"},
1199 {Opt_usrquota, "usrquota"},
1200 {Opt_barrier, "barrier=%u"},
1201 {Opt_barrier, "barrier"},
1202 {Opt_nobarrier, "nobarrier"},
1203 {Opt_i_version, "i_version"},
1205 {Opt_stripe, "stripe=%u"},
1206 {Opt_delalloc, "delalloc"},
1207 {Opt_lazytime, "lazytime"},
1208 {Opt_nolazytime, "nolazytime"},
1209 {Opt_nodelalloc, "nodelalloc"},
1210 {Opt_removed, "mblk_io_submit"},
1211 {Opt_removed, "nomblk_io_submit"},
1212 {Opt_block_validity, "block_validity"},
1213 {Opt_noblock_validity, "noblock_validity"},
1214 {Opt_inode_readahead_blks, "inode_readahead_blks=%u"},
1215 {Opt_journal_ioprio, "journal_ioprio=%u"},
1216 {Opt_auto_da_alloc, "auto_da_alloc=%u"},
1217 {Opt_auto_da_alloc, "auto_da_alloc"},
1218 {Opt_noauto_da_alloc, "noauto_da_alloc"},
1219 {Opt_dioread_nolock, "dioread_nolock"},
1220 {Opt_dioread_lock, "dioread_lock"},
1221 {Opt_discard, "discard"},
1222 {Opt_nodiscard, "nodiscard"},
1223 {Opt_init_itable, "init_itable=%u"},
1224 {Opt_init_itable, "init_itable"},
1225 {Opt_noinit_itable, "noinit_itable"},
1226 {Opt_max_dir_size_kb, "max_dir_size_kb=%u"},
1227 {Opt_test_dummy_encryption, "test_dummy_encryption"},
1228 {Opt_removed, "check=none"}, /* mount option from ext2/3 */
1229 {Opt_removed, "nocheck"}, /* mount option from ext2/3 */
1230 {Opt_removed, "reservation"}, /* mount option from ext2/3 */
1231 {Opt_removed, "noreservation"}, /* mount option from ext2/3 */
1232 {Opt_removed, "journal=%u"}, /* mount option from ext2/3 */
1236 static ext4_fsblk_t get_sb_block(void **data)
1238 ext4_fsblk_t sb_block;
1239 char *options = (char *) *data;
1241 if (!options || strncmp(options, "sb=", 3) != 0)
1242 return 1; /* Default location */
1245 /* TODO: use simple_strtoll with >32bit ext4 */
1246 sb_block = simple_strtoul(options, &options, 0);
1247 if (*options && *options != ',') {
1248 printk(KERN_ERR "EXT4-fs: Invalid sb specification: %s\n",
1252 if (*options == ',')
1254 *data = (void *) options;
1259 #define DEFAULT_JOURNAL_IOPRIO (IOPRIO_PRIO_VALUE(IOPRIO_CLASS_BE, 3))
1260 static char deprecated_msg[] = "Mount option \"%s\" will be removed by %s\n"
1261 "Contact linux-ext4@vger.kernel.org if you think we should keep it.\n";
1264 static int set_qf_name(struct super_block *sb, int qtype, substring_t *args)
1266 struct ext4_sb_info *sbi = EXT4_SB(sb);
1270 if (sb_any_quota_loaded(sb) &&
1271 !sbi->s_qf_names[qtype]) {
1272 ext4_msg(sb, KERN_ERR,
1273 "Cannot change journaled "
1274 "quota options when quota turned on");
1277 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_QUOTA)) {
1278 ext4_msg(sb, KERN_ERR, "Cannot set journaled quota options "
1279 "when QUOTA feature is enabled");
1282 qname = match_strdup(args);
1284 ext4_msg(sb, KERN_ERR,
1285 "Not enough memory for storing quotafile name");
1288 if (sbi->s_qf_names[qtype]) {
1289 if (strcmp(sbi->s_qf_names[qtype], qname) == 0)
1292 ext4_msg(sb, KERN_ERR,
1293 "%s quota file already specified",
1297 if (strchr(qname, '/')) {
1298 ext4_msg(sb, KERN_ERR,
1299 "quotafile must be on filesystem root");
1302 sbi->s_qf_names[qtype] = qname;
1310 static int clear_qf_name(struct super_block *sb, int qtype)
1313 struct ext4_sb_info *sbi = EXT4_SB(sb);
1315 if (sb_any_quota_loaded(sb) &&
1316 sbi->s_qf_names[qtype]) {
1317 ext4_msg(sb, KERN_ERR, "Cannot change journaled quota options"
1318 " when quota turned on");
1321 kfree(sbi->s_qf_names[qtype]);
1322 sbi->s_qf_names[qtype] = NULL;
1327 #define MOPT_SET 0x0001
1328 #define MOPT_CLEAR 0x0002
1329 #define MOPT_NOSUPPORT 0x0004
1330 #define MOPT_EXPLICIT 0x0008
1331 #define MOPT_CLEAR_ERR 0x0010
1332 #define MOPT_GTE0 0x0020
1335 #define MOPT_QFMT 0x0040
1337 #define MOPT_Q MOPT_NOSUPPORT
1338 #define MOPT_QFMT MOPT_NOSUPPORT
1340 #define MOPT_DATAJ 0x0080
1341 #define MOPT_NO_EXT2 0x0100
1342 #define MOPT_NO_EXT3 0x0200
1343 #define MOPT_EXT4_ONLY (MOPT_NO_EXT2 | MOPT_NO_EXT3)
1344 #define MOPT_STRING 0x0400
1346 static const struct mount_opts {
1350 } ext4_mount_opts[] = {
1351 {Opt_minix_df, EXT4_MOUNT_MINIX_DF, MOPT_SET},
1352 {Opt_bsd_df, EXT4_MOUNT_MINIX_DF, MOPT_CLEAR},
1353 {Opt_grpid, EXT4_MOUNT_GRPID, MOPT_SET},
1354 {Opt_nogrpid, EXT4_MOUNT_GRPID, MOPT_CLEAR},
1355 {Opt_block_validity, EXT4_MOUNT_BLOCK_VALIDITY, MOPT_SET},
1356 {Opt_noblock_validity, EXT4_MOUNT_BLOCK_VALIDITY, MOPT_CLEAR},
1357 {Opt_dioread_nolock, EXT4_MOUNT_DIOREAD_NOLOCK,
1358 MOPT_EXT4_ONLY | MOPT_SET},
1359 {Opt_dioread_lock, EXT4_MOUNT_DIOREAD_NOLOCK,
1360 MOPT_EXT4_ONLY | MOPT_CLEAR},
1361 {Opt_discard, EXT4_MOUNT_DISCARD, MOPT_SET},
1362 {Opt_nodiscard, EXT4_MOUNT_DISCARD, MOPT_CLEAR},
1363 {Opt_delalloc, EXT4_MOUNT_DELALLOC,
1364 MOPT_EXT4_ONLY | MOPT_SET | MOPT_EXPLICIT},
1365 {Opt_nodelalloc, EXT4_MOUNT_DELALLOC,
1366 MOPT_EXT4_ONLY | MOPT_CLEAR},
1367 {Opt_nojournal_checksum, EXT4_MOUNT_JOURNAL_CHECKSUM,
1368 MOPT_EXT4_ONLY | MOPT_CLEAR},
1369 {Opt_journal_checksum, EXT4_MOUNT_JOURNAL_CHECKSUM,
1370 MOPT_EXT4_ONLY | MOPT_SET},
1371 {Opt_journal_async_commit, (EXT4_MOUNT_JOURNAL_ASYNC_COMMIT |
1372 EXT4_MOUNT_JOURNAL_CHECKSUM),
1373 MOPT_EXT4_ONLY | MOPT_SET},
1374 {Opt_noload, EXT4_MOUNT_NOLOAD, MOPT_NO_EXT2 | MOPT_SET},
1375 {Opt_err_panic, EXT4_MOUNT_ERRORS_PANIC, MOPT_SET | MOPT_CLEAR_ERR},
1376 {Opt_err_ro, EXT4_MOUNT_ERRORS_RO, MOPT_SET | MOPT_CLEAR_ERR},
1377 {Opt_err_cont, EXT4_MOUNT_ERRORS_CONT, MOPT_SET | MOPT_CLEAR_ERR},
1378 {Opt_data_err_abort, EXT4_MOUNT_DATA_ERR_ABORT,
1379 MOPT_NO_EXT2 | MOPT_SET},
1380 {Opt_data_err_ignore, EXT4_MOUNT_DATA_ERR_ABORT,
1381 MOPT_NO_EXT2 | MOPT_CLEAR},
1382 {Opt_barrier, EXT4_MOUNT_BARRIER, MOPT_SET},
1383 {Opt_nobarrier, EXT4_MOUNT_BARRIER, MOPT_CLEAR},
1384 {Opt_noauto_da_alloc, EXT4_MOUNT_NO_AUTO_DA_ALLOC, MOPT_SET},
1385 {Opt_auto_da_alloc, EXT4_MOUNT_NO_AUTO_DA_ALLOC, MOPT_CLEAR},
1386 {Opt_noinit_itable, EXT4_MOUNT_INIT_INODE_TABLE, MOPT_CLEAR},
1387 {Opt_commit, 0, MOPT_GTE0},
1388 {Opt_max_batch_time, 0, MOPT_GTE0},
1389 {Opt_min_batch_time, 0, MOPT_GTE0},
1390 {Opt_inode_readahead_blks, 0, MOPT_GTE0},
1391 {Opt_init_itable, 0, MOPT_GTE0},
1392 {Opt_dax, EXT4_MOUNT_DAX, MOPT_SET},
1393 {Opt_stripe, 0, MOPT_GTE0},
1394 {Opt_resuid, 0, MOPT_GTE0},
1395 {Opt_resgid, 0, MOPT_GTE0},
1396 {Opt_journal_dev, 0, MOPT_GTE0},
1397 {Opt_journal_path, 0, MOPT_STRING},
1398 {Opt_journal_ioprio, 0, MOPT_GTE0},
1399 {Opt_data_journal, EXT4_MOUNT_JOURNAL_DATA, MOPT_NO_EXT2 | MOPT_DATAJ},
1400 {Opt_data_ordered, EXT4_MOUNT_ORDERED_DATA, MOPT_NO_EXT2 | MOPT_DATAJ},
1401 {Opt_data_writeback, EXT4_MOUNT_WRITEBACK_DATA,
1402 MOPT_NO_EXT2 | MOPT_DATAJ},
1403 {Opt_user_xattr, EXT4_MOUNT_XATTR_USER, MOPT_SET},
1404 {Opt_nouser_xattr, EXT4_MOUNT_XATTR_USER, MOPT_CLEAR},
1405 #ifdef CONFIG_EXT4_FS_POSIX_ACL
1406 {Opt_acl, EXT4_MOUNT_POSIX_ACL, MOPT_SET},
1407 {Opt_noacl, EXT4_MOUNT_POSIX_ACL, MOPT_CLEAR},
1409 {Opt_acl, 0, MOPT_NOSUPPORT},
1410 {Opt_noacl, 0, MOPT_NOSUPPORT},
1412 {Opt_nouid32, EXT4_MOUNT_NO_UID32, MOPT_SET},
1413 {Opt_debug, EXT4_MOUNT_DEBUG, MOPT_SET},
1414 {Opt_quota, EXT4_MOUNT_QUOTA | EXT4_MOUNT_USRQUOTA, MOPT_SET | MOPT_Q},
1415 {Opt_usrquota, EXT4_MOUNT_QUOTA | EXT4_MOUNT_USRQUOTA,
1417 {Opt_grpquota, EXT4_MOUNT_QUOTA | EXT4_MOUNT_GRPQUOTA,
1419 {Opt_noquota, (EXT4_MOUNT_QUOTA | EXT4_MOUNT_USRQUOTA |
1420 EXT4_MOUNT_GRPQUOTA), MOPT_CLEAR | MOPT_Q},
1421 {Opt_usrjquota, 0, MOPT_Q},
1422 {Opt_grpjquota, 0, MOPT_Q},
1423 {Opt_offusrjquota, 0, MOPT_Q},
1424 {Opt_offgrpjquota, 0, MOPT_Q},
1425 {Opt_jqfmt_vfsold, QFMT_VFS_OLD, MOPT_QFMT},
1426 {Opt_jqfmt_vfsv0, QFMT_VFS_V0, MOPT_QFMT},
1427 {Opt_jqfmt_vfsv1, QFMT_VFS_V1, MOPT_QFMT},
1428 {Opt_max_dir_size_kb, 0, MOPT_GTE0},
1429 {Opt_test_dummy_encryption, 0, MOPT_GTE0},
1433 static int handle_mount_opt(struct super_block *sb, char *opt, int token,
1434 substring_t *args, unsigned long *journal_devnum,
1435 unsigned int *journal_ioprio, int is_remount)
1437 struct ext4_sb_info *sbi = EXT4_SB(sb);
1438 const struct mount_opts *m;
1444 if (token == Opt_usrjquota)
1445 return set_qf_name(sb, USRQUOTA, &args[0]);
1446 else if (token == Opt_grpjquota)
1447 return set_qf_name(sb, GRPQUOTA, &args[0]);
1448 else if (token == Opt_offusrjquota)
1449 return clear_qf_name(sb, USRQUOTA);
1450 else if (token == Opt_offgrpjquota)
1451 return clear_qf_name(sb, GRPQUOTA);
1455 case Opt_nouser_xattr:
1456 ext4_msg(sb, KERN_WARNING, deprecated_msg, opt, "3.5");
1459 return 1; /* handled by get_sb_block() */
1461 ext4_msg(sb, KERN_WARNING, "Ignoring removed %s option", opt);
1464 sbi->s_mount_flags |= EXT4_MF_FS_ABORTED;
1467 sb->s_flags |= MS_I_VERSION;
1470 sb->s_flags |= MS_LAZYTIME;
1472 case Opt_nolazytime:
1473 sb->s_flags &= ~MS_LAZYTIME;
1477 for (m = ext4_mount_opts; m->token != Opt_err; m++)
1478 if (token == m->token)
1481 if (m->token == Opt_err) {
1482 ext4_msg(sb, KERN_ERR, "Unrecognized mount option \"%s\" "
1483 "or missing value", opt);
1487 if ((m->flags & MOPT_NO_EXT2) && IS_EXT2_SB(sb)) {
1488 ext4_msg(sb, KERN_ERR,
1489 "Mount option \"%s\" incompatible with ext2", opt);
1492 if ((m->flags & MOPT_NO_EXT3) && IS_EXT3_SB(sb)) {
1493 ext4_msg(sb, KERN_ERR,
1494 "Mount option \"%s\" incompatible with ext3", opt);
1498 if (args->from && !(m->flags & MOPT_STRING) && match_int(args, &arg))
1500 if (args->from && (m->flags & MOPT_GTE0) && (arg < 0))
1502 if (m->flags & MOPT_EXPLICIT)
1503 set_opt2(sb, EXPLICIT_DELALLOC);
1504 if (m->flags & MOPT_CLEAR_ERR)
1505 clear_opt(sb, ERRORS_MASK);
1506 if (token == Opt_noquota && sb_any_quota_loaded(sb)) {
1507 ext4_msg(sb, KERN_ERR, "Cannot change quota "
1508 "options when quota turned on");
1512 if (m->flags & MOPT_NOSUPPORT) {
1513 ext4_msg(sb, KERN_ERR, "%s option not supported", opt);
1514 } else if (token == Opt_commit) {
1516 arg = JBD2_DEFAULT_MAX_COMMIT_AGE;
1517 sbi->s_commit_interval = HZ * arg;
1518 } else if (token == Opt_max_batch_time) {
1519 sbi->s_max_batch_time = arg;
1520 } else if (token == Opt_min_batch_time) {
1521 sbi->s_min_batch_time = arg;
1522 } else if (token == Opt_inode_readahead_blks) {
1523 if (arg && (arg > (1 << 30) || !is_power_of_2(arg))) {
1524 ext4_msg(sb, KERN_ERR,
1525 "EXT4-fs: inode_readahead_blks must be "
1526 "0 or a power of 2 smaller than 2^31");
1529 sbi->s_inode_readahead_blks = arg;
1530 } else if (token == Opt_init_itable) {
1531 set_opt(sb, INIT_INODE_TABLE);
1533 arg = EXT4_DEF_LI_WAIT_MULT;
1534 sbi->s_li_wait_mult = arg;
1535 } else if (token == Opt_max_dir_size_kb) {
1536 sbi->s_max_dir_size_kb = arg;
1537 } else if (token == Opt_stripe) {
1538 sbi->s_stripe = arg;
1539 } else if (token == Opt_resuid) {
1540 uid = make_kuid(current_user_ns(), arg);
1541 if (!uid_valid(uid)) {
1542 ext4_msg(sb, KERN_ERR, "Invalid uid value %d", arg);
1545 sbi->s_resuid = uid;
1546 } else if (token == Opt_resgid) {
1547 gid = make_kgid(current_user_ns(), arg);
1548 if (!gid_valid(gid)) {
1549 ext4_msg(sb, KERN_ERR, "Invalid gid value %d", arg);
1552 sbi->s_resgid = gid;
1553 } else if (token == Opt_journal_dev) {
1555 ext4_msg(sb, KERN_ERR,
1556 "Cannot specify journal on remount");
1559 *journal_devnum = arg;
1560 } else if (token == Opt_journal_path) {
1562 struct inode *journal_inode;
1567 ext4_msg(sb, KERN_ERR,
1568 "Cannot specify journal on remount");
1571 journal_path = match_strdup(&args[0]);
1572 if (!journal_path) {
1573 ext4_msg(sb, KERN_ERR, "error: could not dup "
1574 "journal device string");
1578 error = kern_path(journal_path, LOOKUP_FOLLOW, &path);
1580 ext4_msg(sb, KERN_ERR, "error: could not find "
1581 "journal device path: error %d", error);
1582 kfree(journal_path);
1586 journal_inode = d_inode(path.dentry);
1587 if (!S_ISBLK(journal_inode->i_mode)) {
1588 ext4_msg(sb, KERN_ERR, "error: journal path %s "
1589 "is not a block device", journal_path);
1591 kfree(journal_path);
1595 *journal_devnum = new_encode_dev(journal_inode->i_rdev);
1597 kfree(journal_path);
1598 } else if (token == Opt_journal_ioprio) {
1600 ext4_msg(sb, KERN_ERR, "Invalid journal IO priority"
1605 IOPRIO_PRIO_VALUE(IOPRIO_CLASS_BE, arg);
1606 } else if (token == Opt_test_dummy_encryption) {
1607 #ifdef CONFIG_EXT4_FS_ENCRYPTION
1608 sbi->s_mount_flags |= EXT4_MF_TEST_DUMMY_ENCRYPTION;
1609 ext4_msg(sb, KERN_WARNING,
1610 "Test dummy encryption mode enabled");
1612 ext4_msg(sb, KERN_WARNING,
1613 "Test dummy encryption mount option ignored");
1615 } else if (m->flags & MOPT_DATAJ) {
1617 if (!sbi->s_journal)
1618 ext4_msg(sb, KERN_WARNING, "Remounting file system with no journal so ignoring journalled data option");
1619 else if (test_opt(sb, DATA_FLAGS) != m->mount_opt) {
1620 ext4_msg(sb, KERN_ERR,
1621 "Cannot change data mode on remount");
1625 clear_opt(sb, DATA_FLAGS);
1626 sbi->s_mount_opt |= m->mount_opt;
1629 } else if (m->flags & MOPT_QFMT) {
1630 if (sb_any_quota_loaded(sb) &&
1631 sbi->s_jquota_fmt != m->mount_opt) {
1632 ext4_msg(sb, KERN_ERR, "Cannot change journaled "
1633 "quota options when quota turned on");
1636 if (EXT4_HAS_RO_COMPAT_FEATURE(sb,
1637 EXT4_FEATURE_RO_COMPAT_QUOTA)) {
1638 ext4_msg(sb, KERN_ERR,
1639 "Cannot set journaled quota options "
1640 "when QUOTA feature is enabled");
1643 sbi->s_jquota_fmt = m->mount_opt;
1645 #ifndef CONFIG_FS_DAX
1646 } else if (token == Opt_dax) {
1647 ext4_msg(sb, KERN_INFO, "dax option not supported");
1653 if (m->flags & MOPT_CLEAR)
1655 else if (unlikely(!(m->flags & MOPT_SET))) {
1656 ext4_msg(sb, KERN_WARNING,
1657 "buggy handling of option %s", opt);
1662 sbi->s_mount_opt |= m->mount_opt;
1664 sbi->s_mount_opt &= ~m->mount_opt;
1669 static int parse_options(char *options, struct super_block *sb,
1670 unsigned long *journal_devnum,
1671 unsigned int *journal_ioprio,
1674 struct ext4_sb_info *sbi = EXT4_SB(sb);
1676 substring_t args[MAX_OPT_ARGS];
1682 while ((p = strsep(&options, ",")) != NULL) {
1686 * Initialize args struct so we know whether arg was
1687 * found; some options take optional arguments.
1689 args[0].to = args[0].from = NULL;
1690 token = match_token(p, tokens, args);
1691 if (handle_mount_opt(sb, p, token, args, journal_devnum,
1692 journal_ioprio, is_remount) < 0)
1696 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_QUOTA) &&
1697 (test_opt(sb, USRQUOTA) || test_opt(sb, GRPQUOTA))) {
1698 ext4_msg(sb, KERN_ERR, "Cannot set quota options when QUOTA "
1699 "feature is enabled");
1702 if (sbi->s_qf_names[USRQUOTA] || sbi->s_qf_names[GRPQUOTA]) {
1703 if (test_opt(sb, USRQUOTA) && sbi->s_qf_names[USRQUOTA])
1704 clear_opt(sb, USRQUOTA);
1706 if (test_opt(sb, GRPQUOTA) && sbi->s_qf_names[GRPQUOTA])
1707 clear_opt(sb, GRPQUOTA);
1709 if (test_opt(sb, GRPQUOTA) || test_opt(sb, USRQUOTA)) {
1710 ext4_msg(sb, KERN_ERR, "old and new quota "
1715 if (!sbi->s_jquota_fmt) {
1716 ext4_msg(sb, KERN_ERR, "journaled quota format "
1722 if (test_opt(sb, DIOREAD_NOLOCK)) {
1724 BLOCK_SIZE << le32_to_cpu(sbi->s_es->s_log_block_size);
1726 if (blocksize < PAGE_CACHE_SIZE) {
1727 ext4_msg(sb, KERN_ERR, "can't mount with "
1728 "dioread_nolock if block size != PAGE_SIZE");
1732 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA &&
1733 test_opt(sb, JOURNAL_ASYNC_COMMIT)) {
1734 ext4_msg(sb, KERN_ERR, "can't mount with journal_async_commit "
1735 "in data=ordered mode");
1741 static inline void ext4_show_quota_options(struct seq_file *seq,
1742 struct super_block *sb)
1744 #if defined(CONFIG_QUOTA)
1745 struct ext4_sb_info *sbi = EXT4_SB(sb);
1747 if (sbi->s_jquota_fmt) {
1750 switch (sbi->s_jquota_fmt) {
1761 seq_printf(seq, ",jqfmt=%s", fmtname);
1764 if (sbi->s_qf_names[USRQUOTA])
1765 seq_printf(seq, ",usrjquota=%s", sbi->s_qf_names[USRQUOTA]);
1767 if (sbi->s_qf_names[GRPQUOTA])
1768 seq_printf(seq, ",grpjquota=%s", sbi->s_qf_names[GRPQUOTA]);
1772 static const char *token2str(int token)
1774 const struct match_token *t;
1776 for (t = tokens; t->token != Opt_err; t++)
1777 if (t->token == token && !strchr(t->pattern, '='))
1784 * - it's set to a non-default value OR
1785 * - if the per-sb default is different from the global default
1787 static int _ext4_show_options(struct seq_file *seq, struct super_block *sb,
1790 struct ext4_sb_info *sbi = EXT4_SB(sb);
1791 struct ext4_super_block *es = sbi->s_es;
1792 int def_errors, def_mount_opt = nodefs ? 0 : sbi->s_def_mount_opt;
1793 const struct mount_opts *m;
1794 char sep = nodefs ? '\n' : ',';
1796 #define SEQ_OPTS_PUTS(str) seq_printf(seq, "%c" str, sep)
1797 #define SEQ_OPTS_PRINT(str, arg) seq_printf(seq, "%c" str, sep, arg)
1799 if (sbi->s_sb_block != 1)
1800 SEQ_OPTS_PRINT("sb=%llu", sbi->s_sb_block);
1802 for (m = ext4_mount_opts; m->token != Opt_err; m++) {
1803 int want_set = m->flags & MOPT_SET;
1804 if (((m->flags & (MOPT_SET|MOPT_CLEAR)) == 0) ||
1805 (m->flags & MOPT_CLEAR_ERR))
1807 if (!(m->mount_opt & (sbi->s_mount_opt ^ def_mount_opt)))
1808 continue; /* skip if same as the default */
1810 (sbi->s_mount_opt & m->mount_opt) != m->mount_opt) ||
1811 (!want_set && (sbi->s_mount_opt & m->mount_opt)))
1812 continue; /* select Opt_noFoo vs Opt_Foo */
1813 SEQ_OPTS_PRINT("%s", token2str(m->token));
1816 if (nodefs || !uid_eq(sbi->s_resuid, make_kuid(&init_user_ns, EXT4_DEF_RESUID)) ||
1817 le16_to_cpu(es->s_def_resuid) != EXT4_DEF_RESUID)
1818 SEQ_OPTS_PRINT("resuid=%u",
1819 from_kuid_munged(&init_user_ns, sbi->s_resuid));
1820 if (nodefs || !gid_eq(sbi->s_resgid, make_kgid(&init_user_ns, EXT4_DEF_RESGID)) ||
1821 le16_to_cpu(es->s_def_resgid) != EXT4_DEF_RESGID)
1822 SEQ_OPTS_PRINT("resgid=%u",
1823 from_kgid_munged(&init_user_ns, sbi->s_resgid));
1824 def_errors = nodefs ? -1 : le16_to_cpu(es->s_errors);
1825 if (test_opt(sb, ERRORS_RO) && def_errors != EXT4_ERRORS_RO)
1826 SEQ_OPTS_PUTS("errors=remount-ro");
1827 if (test_opt(sb, ERRORS_CONT) && def_errors != EXT4_ERRORS_CONTINUE)
1828 SEQ_OPTS_PUTS("errors=continue");
1829 if (test_opt(sb, ERRORS_PANIC) && def_errors != EXT4_ERRORS_PANIC)
1830 SEQ_OPTS_PUTS("errors=panic");
1831 if (nodefs || sbi->s_commit_interval != JBD2_DEFAULT_MAX_COMMIT_AGE*HZ)
1832 SEQ_OPTS_PRINT("commit=%lu", sbi->s_commit_interval / HZ);
1833 if (nodefs || sbi->s_min_batch_time != EXT4_DEF_MIN_BATCH_TIME)
1834 SEQ_OPTS_PRINT("min_batch_time=%u", sbi->s_min_batch_time);
1835 if (nodefs || sbi->s_max_batch_time != EXT4_DEF_MAX_BATCH_TIME)
1836 SEQ_OPTS_PRINT("max_batch_time=%u", sbi->s_max_batch_time);
1837 if (sb->s_flags & MS_I_VERSION)
1838 SEQ_OPTS_PUTS("i_version");
1839 if (nodefs || sbi->s_stripe)
1840 SEQ_OPTS_PRINT("stripe=%lu", sbi->s_stripe);
1841 if (EXT4_MOUNT_DATA_FLAGS & (sbi->s_mount_opt ^ def_mount_opt)) {
1842 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA)
1843 SEQ_OPTS_PUTS("data=journal");
1844 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA)
1845 SEQ_OPTS_PUTS("data=ordered");
1846 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_WRITEBACK_DATA)
1847 SEQ_OPTS_PUTS("data=writeback");
1850 sbi->s_inode_readahead_blks != EXT4_DEF_INODE_READAHEAD_BLKS)
1851 SEQ_OPTS_PRINT("inode_readahead_blks=%u",
1852 sbi->s_inode_readahead_blks);
1854 if (nodefs || (test_opt(sb, INIT_INODE_TABLE) &&
1855 (sbi->s_li_wait_mult != EXT4_DEF_LI_WAIT_MULT)))
1856 SEQ_OPTS_PRINT("init_itable=%u", sbi->s_li_wait_mult);
1857 if (nodefs || sbi->s_max_dir_size_kb)
1858 SEQ_OPTS_PRINT("max_dir_size_kb=%u", sbi->s_max_dir_size_kb);
1860 ext4_show_quota_options(seq, sb);
1864 static int ext4_show_options(struct seq_file *seq, struct dentry *root)
1866 return _ext4_show_options(seq, root->d_sb, 0);
1869 static int options_seq_show(struct seq_file *seq, void *offset)
1871 struct super_block *sb = seq->private;
1874 seq_puts(seq, (sb->s_flags & MS_RDONLY) ? "ro" : "rw");
1875 rc = _ext4_show_options(seq, sb, 1);
1876 seq_puts(seq, "\n");
1880 static int options_open_fs(struct inode *inode, struct file *file)
1882 return single_open(file, options_seq_show, PDE_DATA(inode));
1885 static const struct file_operations ext4_seq_options_fops = {
1886 .owner = THIS_MODULE,
1887 .open = options_open_fs,
1889 .llseek = seq_lseek,
1890 .release = single_release,
1893 static int ext4_setup_super(struct super_block *sb, struct ext4_super_block *es,
1896 struct ext4_sb_info *sbi = EXT4_SB(sb);
1899 if (le32_to_cpu(es->s_rev_level) > EXT4_MAX_SUPP_REV) {
1900 ext4_msg(sb, KERN_ERR, "revision level too high, "
1901 "forcing read-only mode");
1906 if (!(sbi->s_mount_state & EXT4_VALID_FS))
1907 ext4_msg(sb, KERN_WARNING, "warning: mounting unchecked fs, "
1908 "running e2fsck is recommended");
1909 else if (sbi->s_mount_state & EXT4_ERROR_FS)
1910 ext4_msg(sb, KERN_WARNING,
1911 "warning: mounting fs with errors, "
1912 "running e2fsck is recommended");
1913 else if ((__s16) le16_to_cpu(es->s_max_mnt_count) > 0 &&
1914 le16_to_cpu(es->s_mnt_count) >=
1915 (unsigned short) (__s16) le16_to_cpu(es->s_max_mnt_count))
1916 ext4_msg(sb, KERN_WARNING,
1917 "warning: maximal mount count reached, "
1918 "running e2fsck is recommended");
1919 else if (le32_to_cpu(es->s_checkinterval) &&
1920 (le32_to_cpu(es->s_lastcheck) +
1921 le32_to_cpu(es->s_checkinterval) <= get_seconds()))
1922 ext4_msg(sb, KERN_WARNING,
1923 "warning: checktime reached, "
1924 "running e2fsck is recommended");
1925 if (!sbi->s_journal)
1926 es->s_state &= cpu_to_le16(~EXT4_VALID_FS);
1927 if (!(__s16) le16_to_cpu(es->s_max_mnt_count))
1928 es->s_max_mnt_count = cpu_to_le16(EXT4_DFL_MAX_MNT_COUNT);
1929 le16_add_cpu(&es->s_mnt_count, 1);
1930 es->s_mtime = cpu_to_le32(get_seconds());
1931 ext4_update_dynamic_rev(sb);
1933 EXT4_SET_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
1935 ext4_commit_super(sb, 1);
1937 if (test_opt(sb, DEBUG))
1938 printk(KERN_INFO "[EXT4 FS bs=%lu, gc=%u, "
1939 "bpg=%lu, ipg=%lu, mo=%04x, mo2=%04x]\n",
1941 sbi->s_groups_count,
1942 EXT4_BLOCKS_PER_GROUP(sb),
1943 EXT4_INODES_PER_GROUP(sb),
1944 sbi->s_mount_opt, sbi->s_mount_opt2);
1946 cleancache_init_fs(sb);
1950 int ext4_alloc_flex_bg_array(struct super_block *sb, ext4_group_t ngroup)
1952 struct ext4_sb_info *sbi = EXT4_SB(sb);
1953 struct flex_groups *new_groups;
1956 if (!sbi->s_log_groups_per_flex)
1959 size = ext4_flex_group(sbi, ngroup - 1) + 1;
1960 if (size <= sbi->s_flex_groups_allocated)
1963 size = roundup_pow_of_two(size * sizeof(struct flex_groups));
1964 new_groups = ext4_kvzalloc(size, GFP_KERNEL);
1966 ext4_msg(sb, KERN_ERR, "not enough memory for %d flex groups",
1967 size / (int) sizeof(struct flex_groups));
1971 if (sbi->s_flex_groups) {
1972 memcpy(new_groups, sbi->s_flex_groups,
1973 (sbi->s_flex_groups_allocated *
1974 sizeof(struct flex_groups)));
1975 kvfree(sbi->s_flex_groups);
1977 sbi->s_flex_groups = new_groups;
1978 sbi->s_flex_groups_allocated = size / sizeof(struct flex_groups);
1982 static int ext4_fill_flex_info(struct super_block *sb)
1984 struct ext4_sb_info *sbi = EXT4_SB(sb);
1985 struct ext4_group_desc *gdp = NULL;
1986 ext4_group_t flex_group;
1989 sbi->s_log_groups_per_flex = sbi->s_es->s_log_groups_per_flex;
1990 if (sbi->s_log_groups_per_flex < 1 || sbi->s_log_groups_per_flex > 31) {
1991 sbi->s_log_groups_per_flex = 0;
1995 err = ext4_alloc_flex_bg_array(sb, sbi->s_groups_count);
1999 for (i = 0; i < sbi->s_groups_count; i++) {
2000 gdp = ext4_get_group_desc(sb, i, NULL);
2002 flex_group = ext4_flex_group(sbi, i);
2003 atomic_add(ext4_free_inodes_count(sb, gdp),
2004 &sbi->s_flex_groups[flex_group].free_inodes);
2005 atomic64_add(ext4_free_group_clusters(sb, gdp),
2006 &sbi->s_flex_groups[flex_group].free_clusters);
2007 atomic_add(ext4_used_dirs_count(sb, gdp),
2008 &sbi->s_flex_groups[flex_group].used_dirs);
2016 static __le16 ext4_group_desc_csum(struct ext4_sb_info *sbi, __u32 block_group,
2017 struct ext4_group_desc *gdp)
2021 __le32 le_group = cpu_to_le32(block_group);
2023 if (ext4_has_metadata_csum(sbi->s_sb)) {
2024 /* Use new metadata_csum algorithm */
2028 save_csum = gdp->bg_checksum;
2029 gdp->bg_checksum = 0;
2030 csum32 = ext4_chksum(sbi, sbi->s_csum_seed, (__u8 *)&le_group,
2032 csum32 = ext4_chksum(sbi, csum32, (__u8 *)gdp,
2034 gdp->bg_checksum = save_csum;
2036 crc = csum32 & 0xFFFF;
2040 /* old crc16 code */
2041 if (!(sbi->s_es->s_feature_ro_compat &
2042 cpu_to_le32(EXT4_FEATURE_RO_COMPAT_GDT_CSUM)))
2045 offset = offsetof(struct ext4_group_desc, bg_checksum);
2047 crc = crc16(~0, sbi->s_es->s_uuid, sizeof(sbi->s_es->s_uuid));
2048 crc = crc16(crc, (__u8 *)&le_group, sizeof(le_group));
2049 crc = crc16(crc, (__u8 *)gdp, offset);
2050 offset += sizeof(gdp->bg_checksum); /* skip checksum */
2051 /* for checksum of struct ext4_group_desc do the rest...*/
2052 if ((sbi->s_es->s_feature_incompat &
2053 cpu_to_le32(EXT4_FEATURE_INCOMPAT_64BIT)) &&
2054 offset < le16_to_cpu(sbi->s_es->s_desc_size))
2055 crc = crc16(crc, (__u8 *)gdp + offset,
2056 le16_to_cpu(sbi->s_es->s_desc_size) -
2060 return cpu_to_le16(crc);
2063 int ext4_group_desc_csum_verify(struct super_block *sb, __u32 block_group,
2064 struct ext4_group_desc *gdp)
2066 if (ext4_has_group_desc_csum(sb) &&
2067 (gdp->bg_checksum != ext4_group_desc_csum(EXT4_SB(sb),
2074 void ext4_group_desc_csum_set(struct super_block *sb, __u32 block_group,
2075 struct ext4_group_desc *gdp)
2077 if (!ext4_has_group_desc_csum(sb))
2079 gdp->bg_checksum = ext4_group_desc_csum(EXT4_SB(sb), block_group, gdp);
2082 /* Called at mount-time, super-block is locked */
2083 static int ext4_check_descriptors(struct super_block *sb,
2084 ext4_group_t *first_not_zeroed)
2086 struct ext4_sb_info *sbi = EXT4_SB(sb);
2087 ext4_fsblk_t first_block = le32_to_cpu(sbi->s_es->s_first_data_block);
2088 ext4_fsblk_t last_block;
2089 ext4_fsblk_t block_bitmap;
2090 ext4_fsblk_t inode_bitmap;
2091 ext4_fsblk_t inode_table;
2092 int flexbg_flag = 0;
2093 ext4_group_t i, grp = sbi->s_groups_count;
2095 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_FLEX_BG))
2098 ext4_debug("Checking group descriptors");
2100 for (i = 0; i < sbi->s_groups_count; i++) {
2101 struct ext4_group_desc *gdp = ext4_get_group_desc(sb, i, NULL);
2103 if (i == sbi->s_groups_count - 1 || flexbg_flag)
2104 last_block = ext4_blocks_count(sbi->s_es) - 1;
2106 last_block = first_block +
2107 (EXT4_BLOCKS_PER_GROUP(sb) - 1);
2109 if ((grp == sbi->s_groups_count) &&
2110 !(gdp->bg_flags & cpu_to_le16(EXT4_BG_INODE_ZEROED)))
2113 block_bitmap = ext4_block_bitmap(sb, gdp);
2114 if (block_bitmap < first_block || block_bitmap > last_block) {
2115 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2116 "Block bitmap for group %u not in group "
2117 "(block %llu)!", i, block_bitmap);
2120 inode_bitmap = ext4_inode_bitmap(sb, gdp);
2121 if (inode_bitmap < first_block || inode_bitmap > last_block) {
2122 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2123 "Inode bitmap for group %u not in group "
2124 "(block %llu)!", i, inode_bitmap);
2127 inode_table = ext4_inode_table(sb, gdp);
2128 if (inode_table < first_block ||
2129 inode_table + sbi->s_itb_per_group - 1 > last_block) {
2130 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2131 "Inode table for group %u not in group "
2132 "(block %llu)!", i, inode_table);
2135 ext4_lock_group(sb, i);
2136 if (!ext4_group_desc_csum_verify(sb, i, gdp)) {
2137 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2138 "Checksum for group %u failed (%u!=%u)",
2139 i, le16_to_cpu(ext4_group_desc_csum(sbi, i,
2140 gdp)), le16_to_cpu(gdp->bg_checksum));
2141 if (!(sb->s_flags & MS_RDONLY)) {
2142 ext4_unlock_group(sb, i);
2146 ext4_unlock_group(sb, i);
2148 first_block += EXT4_BLOCKS_PER_GROUP(sb);
2150 if (NULL != first_not_zeroed)
2151 *first_not_zeroed = grp;
2155 /* ext4_orphan_cleanup() walks a singly-linked list of inodes (starting at
2156 * the superblock) which were deleted from all directories, but held open by
2157 * a process at the time of a crash. We walk the list and try to delete these
2158 * inodes at recovery time (only with a read-write filesystem).
2160 * In order to keep the orphan inode chain consistent during traversal (in
2161 * case of crash during recovery), we link each inode into the superblock
2162 * orphan list_head and handle it the same way as an inode deletion during
2163 * normal operation (which journals the operations for us).
2165 * We only do an iget() and an iput() on each inode, which is very safe if we
2166 * accidentally point at an in-use or already deleted inode. The worst that
2167 * can happen in this case is that we get a "bit already cleared" message from
2168 * ext4_free_inode(). The only reason we would point at a wrong inode is if
2169 * e2fsck was run on this filesystem, and it must have already done the orphan
2170 * inode cleanup for us, so we can safely abort without any further action.
2172 static void ext4_orphan_cleanup(struct super_block *sb,
2173 struct ext4_super_block *es)
2175 unsigned int s_flags = sb->s_flags;
2176 int nr_orphans = 0, nr_truncates = 0;
2180 if (!es->s_last_orphan) {
2181 jbd_debug(4, "no orphan inodes to clean up\n");
2185 if (bdev_read_only(sb->s_bdev)) {
2186 ext4_msg(sb, KERN_ERR, "write access "
2187 "unavailable, skipping orphan cleanup");
2191 /* Check if feature set would not allow a r/w mount */
2192 if (!ext4_feature_set_ok(sb, 0)) {
2193 ext4_msg(sb, KERN_INFO, "Skipping orphan cleanup due to "
2194 "unknown ROCOMPAT features");
2198 if (EXT4_SB(sb)->s_mount_state & EXT4_ERROR_FS) {
2199 /* don't clear list on RO mount w/ errors */
2200 if (es->s_last_orphan && !(s_flags & MS_RDONLY)) {
2201 ext4_msg(sb, KERN_INFO, "Errors on filesystem, "
2202 "clearing orphan list.\n");
2203 es->s_last_orphan = 0;
2205 jbd_debug(1, "Skipping orphan recovery on fs with errors.\n");
2209 if (s_flags & MS_RDONLY) {
2210 ext4_msg(sb, KERN_INFO, "orphan cleanup on readonly fs");
2211 sb->s_flags &= ~MS_RDONLY;
2214 /* Needed for iput() to work correctly and not trash data */
2215 sb->s_flags |= MS_ACTIVE;
2216 /* Turn on quotas so that they are updated correctly */
2217 for (i = 0; i < EXT4_MAXQUOTAS; i++) {
2218 if (EXT4_SB(sb)->s_qf_names[i]) {
2219 int ret = ext4_quota_on_mount(sb, i);
2221 ext4_msg(sb, KERN_ERR,
2222 "Cannot turn on journaled "
2223 "quota: error %d", ret);
2228 while (es->s_last_orphan) {
2229 struct inode *inode;
2231 inode = ext4_orphan_get(sb, le32_to_cpu(es->s_last_orphan));
2232 if (IS_ERR(inode)) {
2233 es->s_last_orphan = 0;
2237 list_add(&EXT4_I(inode)->i_orphan, &EXT4_SB(sb)->s_orphan);
2238 dquot_initialize(inode);
2239 if (inode->i_nlink) {
2240 if (test_opt(sb, DEBUG))
2241 ext4_msg(sb, KERN_DEBUG,
2242 "%s: truncating inode %lu to %lld bytes",
2243 __func__, inode->i_ino, inode->i_size);
2244 jbd_debug(2, "truncating inode %lu to %lld bytes\n",
2245 inode->i_ino, inode->i_size);
2246 mutex_lock(&inode->i_mutex);
2247 truncate_inode_pages(inode->i_mapping, inode->i_size);
2248 ext4_truncate(inode);
2249 mutex_unlock(&inode->i_mutex);
2252 if (test_opt(sb, DEBUG))
2253 ext4_msg(sb, KERN_DEBUG,
2254 "%s: deleting unreferenced inode %lu",
2255 __func__, inode->i_ino);
2256 jbd_debug(2, "deleting unreferenced inode %lu\n",
2260 iput(inode); /* The delete magic happens here! */
2263 #define PLURAL(x) (x), ((x) == 1) ? "" : "s"
2266 ext4_msg(sb, KERN_INFO, "%d orphan inode%s deleted",
2267 PLURAL(nr_orphans));
2269 ext4_msg(sb, KERN_INFO, "%d truncate%s cleaned up",
2270 PLURAL(nr_truncates));
2272 /* Turn quotas off */
2273 for (i = 0; i < EXT4_MAXQUOTAS; i++) {
2274 if (sb_dqopt(sb)->files[i])
2275 dquot_quota_off(sb, i);
2278 sb->s_flags = s_flags; /* Restore MS_RDONLY status */
2282 * Maximal extent format file size.
2283 * Resulting logical blkno at s_maxbytes must fit in our on-disk
2284 * extent format containers, within a sector_t, and within i_blocks
2285 * in the vfs. ext4 inode has 48 bits of i_block in fsblock units,
2286 * so that won't be a limiting factor.
2288 * However there is other limiting factor. We do store extents in the form
2289 * of starting block and length, hence the resulting length of the extent
2290 * covering maximum file size must fit into on-disk format containers as
2291 * well. Given that length is always by 1 unit bigger than max unit (because
2292 * we count 0 as well) we have to lower the s_maxbytes by one fs block.
2294 * Note, this does *not* consider any metadata overhead for vfs i_blocks.
2296 static loff_t ext4_max_size(int blkbits, int has_huge_files)
2299 loff_t upper_limit = MAX_LFS_FILESIZE;
2301 /* small i_blocks in vfs inode? */
2302 if (!has_huge_files || sizeof(blkcnt_t) < sizeof(u64)) {
2304 * CONFIG_LBDAF is not enabled implies the inode
2305 * i_block represent total blocks in 512 bytes
2306 * 32 == size of vfs inode i_blocks * 8
2308 upper_limit = (1LL << 32) - 1;
2310 /* total blocks in file system block size */
2311 upper_limit >>= (blkbits - 9);
2312 upper_limit <<= blkbits;
2316 * 32-bit extent-start container, ee_block. We lower the maxbytes
2317 * by one fs block, so ee_len can cover the extent of maximum file
2320 res = (1LL << 32) - 1;
2323 /* Sanity check against vm- & vfs- imposed limits */
2324 if (res > upper_limit)
2331 * Maximal bitmap file size. There is a direct, and {,double-,triple-}indirect
2332 * block limit, and also a limit of (2^48 - 1) 512-byte sectors in i_blocks.
2333 * We need to be 1 filesystem block less than the 2^48 sector limit.
2335 static loff_t ext4_max_bitmap_size(int bits, int has_huge_files)
2337 loff_t res = EXT4_NDIR_BLOCKS;
2340 /* This is calculated to be the largest file size for a dense, block
2341 * mapped file such that the file's total number of 512-byte sectors,
2342 * including data and all indirect blocks, does not exceed (2^48 - 1).
2344 * __u32 i_blocks_lo and _u16 i_blocks_high represent the total
2345 * number of 512-byte sectors of the file.
2348 if (!has_huge_files || sizeof(blkcnt_t) < sizeof(u64)) {
2350 * !has_huge_files or CONFIG_LBDAF not enabled implies that
2351 * the inode i_block field represents total file blocks in
2352 * 2^32 512-byte sectors == size of vfs inode i_blocks * 8
2354 upper_limit = (1LL << 32) - 1;
2356 /* total blocks in file system block size */
2357 upper_limit >>= (bits - 9);
2361 * We use 48 bit ext4_inode i_blocks
2362 * With EXT4_HUGE_FILE_FL set the i_blocks
2363 * represent total number of blocks in
2364 * file system block size
2366 upper_limit = (1LL << 48) - 1;
2370 /* indirect blocks */
2372 /* double indirect blocks */
2373 meta_blocks += 1 + (1LL << (bits-2));
2374 /* tripple indirect blocks */
2375 meta_blocks += 1 + (1LL << (bits-2)) + (1LL << (2*(bits-2)));
2377 upper_limit -= meta_blocks;
2378 upper_limit <<= bits;
2380 res += 1LL << (bits-2);
2381 res += 1LL << (2*(bits-2));
2382 res += 1LL << (3*(bits-2));
2384 if (res > upper_limit)
2387 if (res > MAX_LFS_FILESIZE)
2388 res = MAX_LFS_FILESIZE;
2393 static ext4_fsblk_t descriptor_loc(struct super_block *sb,
2394 ext4_fsblk_t logical_sb_block, int nr)
2396 struct ext4_sb_info *sbi = EXT4_SB(sb);
2397 ext4_group_t bg, first_meta_bg;
2400 first_meta_bg = le32_to_cpu(sbi->s_es->s_first_meta_bg);
2402 if (!EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_META_BG) ||
2404 return logical_sb_block + nr + 1;
2405 bg = sbi->s_desc_per_block * nr;
2406 if (ext4_bg_has_super(sb, bg))
2410 * If we have a meta_bg fs with 1k blocks, group 0's GDT is at
2411 * block 2, not 1. If s_first_data_block == 0 (bigalloc is enabled
2412 * on modern mke2fs or blksize > 1k on older mke2fs) then we must
2415 if (sb->s_blocksize == 1024 && nr == 0 &&
2416 le32_to_cpu(EXT4_SB(sb)->s_es->s_first_data_block) == 0)
2419 return (has_super + ext4_group_first_block_no(sb, bg));
2423 * ext4_get_stripe_size: Get the stripe size.
2424 * @sbi: In memory super block info
2426 * If we have specified it via mount option, then
2427 * use the mount option value. If the value specified at mount time is
2428 * greater than the blocks per group use the super block value.
2429 * If the super block value is greater than blocks per group return 0.
2430 * Allocator needs it be less than blocks per group.
2433 static unsigned long ext4_get_stripe_size(struct ext4_sb_info *sbi)
2435 unsigned long stride = le16_to_cpu(sbi->s_es->s_raid_stride);
2436 unsigned long stripe_width =
2437 le32_to_cpu(sbi->s_es->s_raid_stripe_width);
2440 if (sbi->s_stripe && sbi->s_stripe <= sbi->s_blocks_per_group)
2441 ret = sbi->s_stripe;
2442 else if (stripe_width <= sbi->s_blocks_per_group)
2444 else if (stride <= sbi->s_blocks_per_group)
2450 * If the stripe width is 1, this makes no sense and
2451 * we set it to 0 to turn off stripe handling code.
2462 struct attribute attr;
2463 ssize_t (*show)(struct ext4_attr *, struct ext4_sb_info *, char *);
2464 ssize_t (*store)(struct ext4_attr *, struct ext4_sb_info *,
2465 const char *, size_t);
2472 static int parse_strtoull(const char *buf,
2473 unsigned long long max, unsigned long long *value)
2477 ret = kstrtoull(skip_spaces(buf), 0, value);
2478 if (!ret && *value > max)
2483 static ssize_t delayed_allocation_blocks_show(struct ext4_attr *a,
2484 struct ext4_sb_info *sbi,
2487 return snprintf(buf, PAGE_SIZE, "%llu\n",
2489 percpu_counter_sum(&sbi->s_dirtyclusters_counter)));
2492 static ssize_t session_write_kbytes_show(struct ext4_attr *a,
2493 struct ext4_sb_info *sbi, char *buf)
2495 struct super_block *sb = sbi->s_buddy_cache->i_sb;
2497 if (!sb->s_bdev->bd_part)
2498 return snprintf(buf, PAGE_SIZE, "0\n");
2499 return snprintf(buf, PAGE_SIZE, "%lu\n",
2500 (part_stat_read(sb->s_bdev->bd_part, sectors[1]) -
2501 sbi->s_sectors_written_start) >> 1);
2504 static ssize_t lifetime_write_kbytes_show(struct ext4_attr *a,
2505 struct ext4_sb_info *sbi, char *buf)
2507 struct super_block *sb = sbi->s_buddy_cache->i_sb;
2509 if (!sb->s_bdev->bd_part)
2510 return snprintf(buf, PAGE_SIZE, "0\n");
2511 return snprintf(buf, PAGE_SIZE, "%llu\n",
2512 (unsigned long long)(sbi->s_kbytes_written +
2513 ((part_stat_read(sb->s_bdev->bd_part, sectors[1]) -
2514 EXT4_SB(sb)->s_sectors_written_start) >> 1)));
2517 static ssize_t inode_readahead_blks_store(struct ext4_attr *a,
2518 struct ext4_sb_info *sbi,
2519 const char *buf, size_t count)
2524 ret = kstrtoul(skip_spaces(buf), 0, &t);
2528 if (t && (!is_power_of_2(t) || t > 0x40000000))
2531 sbi->s_inode_readahead_blks = t;
2535 static ssize_t sbi_ui_show(struct ext4_attr *a,
2536 struct ext4_sb_info *sbi, char *buf)
2538 unsigned int *ui = (unsigned int *) (((char *) sbi) + a->u.offset);
2540 return snprintf(buf, PAGE_SIZE, "%u\n", *ui);
2543 static ssize_t sbi_ui_store(struct ext4_attr *a,
2544 struct ext4_sb_info *sbi,
2545 const char *buf, size_t count)
2547 unsigned int *ui = (unsigned int *) (((char *) sbi) + a->u.offset);
2551 ret = kstrtoul(skip_spaces(buf), 0, &t);
2558 static ssize_t es_ui_show(struct ext4_attr *a,
2559 struct ext4_sb_info *sbi, char *buf)
2562 unsigned int *ui = (unsigned int *) (((char *) sbi->s_es) +
2565 return snprintf(buf, PAGE_SIZE, "%u\n", *ui);
2568 static ssize_t reserved_clusters_show(struct ext4_attr *a,
2569 struct ext4_sb_info *sbi, char *buf)
2571 return snprintf(buf, PAGE_SIZE, "%llu\n",
2572 (unsigned long long) atomic64_read(&sbi->s_resv_clusters));
2575 static ssize_t reserved_clusters_store(struct ext4_attr *a,
2576 struct ext4_sb_info *sbi,
2577 const char *buf, size_t count)
2579 unsigned long long val;
2582 if (parse_strtoull(buf, -1ULL, &val))
2584 ret = ext4_reserve_clusters(sbi, val);
2586 return ret ? ret : count;
2589 static ssize_t trigger_test_error(struct ext4_attr *a,
2590 struct ext4_sb_info *sbi,
2591 const char *buf, size_t count)
2595 if (!capable(CAP_SYS_ADMIN))
2598 if (len && buf[len-1] == '\n')
2602 ext4_error(sbi->s_sb, "%.*s", len, buf);
2606 static ssize_t sbi_deprecated_show(struct ext4_attr *a,
2607 struct ext4_sb_info *sbi, char *buf)
2609 return snprintf(buf, PAGE_SIZE, "%d\n", a->u.deprecated_val);
2612 #define EXT4_ATTR_OFFSET(_name,_mode,_show,_store,_elname) \
2613 static struct ext4_attr ext4_attr_##_name = { \
2614 .attr = {.name = __stringify(_name), .mode = _mode }, \
2618 .offset = offsetof(struct ext4_sb_info, _elname),\
2622 #define EXT4_ATTR_OFFSET_ES(_name,_mode,_show,_store,_elname) \
2623 static struct ext4_attr ext4_attr_##_name = { \
2624 .attr = {.name = __stringify(_name), .mode = _mode }, \
2628 .offset = offsetof(struct ext4_super_block, _elname), \
2632 #define EXT4_ATTR(name, mode, show, store) \
2633 static struct ext4_attr ext4_attr_##name = __ATTR(name, mode, show, store)
2635 #define EXT4_INFO_ATTR(name) EXT4_ATTR(name, 0444, NULL, NULL)
2636 #define EXT4_RO_ATTR(name) EXT4_ATTR(name, 0444, name##_show, NULL)
2637 #define EXT4_RW_ATTR(name) EXT4_ATTR(name, 0644, name##_show, name##_store)
2639 #define EXT4_RO_ATTR_ES_UI(name, elname) \
2640 EXT4_ATTR_OFFSET_ES(name, 0444, es_ui_show, NULL, elname)
2641 #define EXT4_RW_ATTR_SBI_UI(name, elname) \
2642 EXT4_ATTR_OFFSET(name, 0644, sbi_ui_show, sbi_ui_store, elname)
2644 #define ATTR_LIST(name) &ext4_attr_##name.attr
2645 #define EXT4_DEPRECATED_ATTR(_name, _val) \
2646 static struct ext4_attr ext4_attr_##_name = { \
2647 .attr = {.name = __stringify(_name), .mode = 0444 }, \
2648 .show = sbi_deprecated_show, \
2650 .deprecated_val = _val, \
2654 EXT4_RO_ATTR(delayed_allocation_blocks);
2655 EXT4_RO_ATTR(session_write_kbytes);
2656 EXT4_RO_ATTR(lifetime_write_kbytes);
2657 EXT4_RW_ATTR(reserved_clusters);
2658 EXT4_ATTR_OFFSET(inode_readahead_blks, 0644, sbi_ui_show,
2659 inode_readahead_blks_store, s_inode_readahead_blks);
2660 EXT4_RW_ATTR_SBI_UI(inode_goal, s_inode_goal);
2661 EXT4_RW_ATTR_SBI_UI(mb_stats, s_mb_stats);
2662 EXT4_RW_ATTR_SBI_UI(mb_max_to_scan, s_mb_max_to_scan);
2663 EXT4_RW_ATTR_SBI_UI(mb_min_to_scan, s_mb_min_to_scan);
2664 EXT4_RW_ATTR_SBI_UI(mb_order2_req, s_mb_order2_reqs);
2665 EXT4_RW_ATTR_SBI_UI(mb_stream_req, s_mb_stream_request);
2666 EXT4_RW_ATTR_SBI_UI(mb_group_prealloc, s_mb_group_prealloc);
2667 EXT4_DEPRECATED_ATTR(max_writeback_mb_bump, 128);
2668 EXT4_RW_ATTR_SBI_UI(extent_max_zeroout_kb, s_extent_max_zeroout_kb);
2669 EXT4_ATTR(trigger_fs_error, 0200, NULL, trigger_test_error);
2670 EXT4_RW_ATTR_SBI_UI(err_ratelimit_interval_ms, s_err_ratelimit_state.interval);
2671 EXT4_RW_ATTR_SBI_UI(err_ratelimit_burst, s_err_ratelimit_state.burst);
2672 EXT4_RW_ATTR_SBI_UI(warning_ratelimit_interval_ms, s_warning_ratelimit_state.interval);
2673 EXT4_RW_ATTR_SBI_UI(warning_ratelimit_burst, s_warning_ratelimit_state.burst);
2674 EXT4_RW_ATTR_SBI_UI(msg_ratelimit_interval_ms, s_msg_ratelimit_state.interval);
2675 EXT4_RW_ATTR_SBI_UI(msg_ratelimit_burst, s_msg_ratelimit_state.burst);
2676 EXT4_RO_ATTR_ES_UI(errors_count, s_error_count);
2677 EXT4_RO_ATTR_ES_UI(first_error_time, s_first_error_time);
2678 EXT4_RO_ATTR_ES_UI(last_error_time, s_last_error_time);
2680 static struct attribute *ext4_attrs[] = {
2681 ATTR_LIST(delayed_allocation_blocks),
2682 ATTR_LIST(session_write_kbytes),
2683 ATTR_LIST(lifetime_write_kbytes),
2684 ATTR_LIST(reserved_clusters),
2685 ATTR_LIST(inode_readahead_blks),
2686 ATTR_LIST(inode_goal),
2687 ATTR_LIST(mb_stats),
2688 ATTR_LIST(mb_max_to_scan),
2689 ATTR_LIST(mb_min_to_scan),
2690 ATTR_LIST(mb_order2_req),
2691 ATTR_LIST(mb_stream_req),
2692 ATTR_LIST(mb_group_prealloc),
2693 ATTR_LIST(max_writeback_mb_bump),
2694 ATTR_LIST(extent_max_zeroout_kb),
2695 ATTR_LIST(trigger_fs_error),
2696 ATTR_LIST(err_ratelimit_interval_ms),
2697 ATTR_LIST(err_ratelimit_burst),
2698 ATTR_LIST(warning_ratelimit_interval_ms),
2699 ATTR_LIST(warning_ratelimit_burst),
2700 ATTR_LIST(msg_ratelimit_interval_ms),
2701 ATTR_LIST(msg_ratelimit_burst),
2702 ATTR_LIST(errors_count),
2703 ATTR_LIST(first_error_time),
2704 ATTR_LIST(last_error_time),
2708 /* Features this copy of ext4 supports */
2709 EXT4_INFO_ATTR(lazy_itable_init);
2710 EXT4_INFO_ATTR(batched_discard);
2711 EXT4_INFO_ATTR(meta_bg_resize);
2712 EXT4_INFO_ATTR(encryption);
2714 static struct attribute *ext4_feat_attrs[] = {
2715 ATTR_LIST(lazy_itable_init),
2716 ATTR_LIST(batched_discard),
2717 ATTR_LIST(meta_bg_resize),
2718 ATTR_LIST(encryption),
2722 static ssize_t ext4_attr_show(struct kobject *kobj,
2723 struct attribute *attr, char *buf)
2725 struct ext4_sb_info *sbi = container_of(kobj, struct ext4_sb_info,
2727 struct ext4_attr *a = container_of(attr, struct ext4_attr, attr);
2729 return a->show ? a->show(a, sbi, buf) : 0;
2732 static ssize_t ext4_attr_store(struct kobject *kobj,
2733 struct attribute *attr,
2734 const char *buf, size_t len)
2736 struct ext4_sb_info *sbi = container_of(kobj, struct ext4_sb_info,
2738 struct ext4_attr *a = container_of(attr, struct ext4_attr, attr);
2740 return a->store ? a->store(a, sbi, buf, len) : 0;
2743 static void ext4_sb_release(struct kobject *kobj)
2745 struct ext4_sb_info *sbi = container_of(kobj, struct ext4_sb_info,
2747 complete(&sbi->s_kobj_unregister);
2750 static const struct sysfs_ops ext4_attr_ops = {
2751 .show = ext4_attr_show,
2752 .store = ext4_attr_store,
2755 static struct kobj_type ext4_ktype = {
2756 .default_attrs = ext4_attrs,
2757 .sysfs_ops = &ext4_attr_ops,
2758 .release = ext4_sb_release,
2761 static void ext4_feat_release(struct kobject *kobj)
2763 complete(&ext4_feat->f_kobj_unregister);
2766 static ssize_t ext4_feat_show(struct kobject *kobj,
2767 struct attribute *attr, char *buf)
2769 return snprintf(buf, PAGE_SIZE, "supported\n");
2773 * We can not use ext4_attr_show/store because it relies on the kobject
2774 * being embedded in the ext4_sb_info structure which is definitely not
2775 * true in this case.
2777 static const struct sysfs_ops ext4_feat_ops = {
2778 .show = ext4_feat_show,
2782 static struct kobj_type ext4_feat_ktype = {
2783 .default_attrs = ext4_feat_attrs,
2784 .sysfs_ops = &ext4_feat_ops,
2785 .release = ext4_feat_release,
2789 * Check whether this filesystem can be mounted based on
2790 * the features present and the RDONLY/RDWR mount requested.
2791 * Returns 1 if this filesystem can be mounted as requested,
2792 * 0 if it cannot be.
2794 static int ext4_feature_set_ok(struct super_block *sb, int readonly)
2796 if (EXT4_HAS_INCOMPAT_FEATURE(sb, ~EXT4_FEATURE_INCOMPAT_SUPP)) {
2797 ext4_msg(sb, KERN_ERR,
2798 "Couldn't mount because of "
2799 "unsupported optional features (%x)",
2800 (le32_to_cpu(EXT4_SB(sb)->s_es->s_feature_incompat) &
2801 ~EXT4_FEATURE_INCOMPAT_SUPP));
2808 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_READONLY)) {
2809 ext4_msg(sb, KERN_INFO, "filesystem is read-only");
2810 sb->s_flags |= MS_RDONLY;
2814 /* Check that feature set is OK for a read-write mount */
2815 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, ~EXT4_FEATURE_RO_COMPAT_SUPP)) {
2816 ext4_msg(sb, KERN_ERR, "couldn't mount RDWR because of "
2817 "unsupported optional features (%x)",
2818 (le32_to_cpu(EXT4_SB(sb)->s_es->s_feature_ro_compat) &
2819 ~EXT4_FEATURE_RO_COMPAT_SUPP));
2823 * Large file size enabled file system can only be mounted
2824 * read-write on 32-bit systems if kernel is built with CONFIG_LBDAF
2826 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_HUGE_FILE)) {
2827 if (sizeof(blkcnt_t) < sizeof(u64)) {
2828 ext4_msg(sb, KERN_ERR, "Filesystem with huge files "
2829 "cannot be mounted RDWR without "
2834 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_BIGALLOC) &&
2835 !EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_EXTENTS)) {
2836 ext4_msg(sb, KERN_ERR,
2837 "Can't support bigalloc feature without "
2838 "extents feature\n");
2842 #ifndef CONFIG_QUOTA
2843 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_QUOTA) &&
2845 ext4_msg(sb, KERN_ERR,
2846 "Filesystem with quota feature cannot be mounted RDWR "
2847 "without CONFIG_QUOTA");
2850 #endif /* CONFIG_QUOTA */
2855 * This function is called once a day if we have errors logged
2856 * on the file system
2858 static void print_daily_error_info(unsigned long arg)
2860 struct super_block *sb = (struct super_block *) arg;
2861 struct ext4_sb_info *sbi;
2862 struct ext4_super_block *es;
2867 if (es->s_error_count)
2868 /* fsck newer than v1.41.13 is needed to clean this condition. */
2869 ext4_msg(sb, KERN_NOTICE, "error count since last fsck: %u",
2870 le32_to_cpu(es->s_error_count));
2871 if (es->s_first_error_time) {
2872 printk(KERN_NOTICE "EXT4-fs (%s): initial error at time %u: %.*s:%d",
2873 sb->s_id, le32_to_cpu(es->s_first_error_time),
2874 (int) sizeof(es->s_first_error_func),
2875 es->s_first_error_func,
2876 le32_to_cpu(es->s_first_error_line));
2877 if (es->s_first_error_ino)
2878 printk(": inode %u",
2879 le32_to_cpu(es->s_first_error_ino));
2880 if (es->s_first_error_block)
2881 printk(": block %llu", (unsigned long long)
2882 le64_to_cpu(es->s_first_error_block));
2885 if (es->s_last_error_time) {
2886 printk(KERN_NOTICE "EXT4-fs (%s): last error at time %u: %.*s:%d",
2887 sb->s_id, le32_to_cpu(es->s_last_error_time),
2888 (int) sizeof(es->s_last_error_func),
2889 es->s_last_error_func,
2890 le32_to_cpu(es->s_last_error_line));
2891 if (es->s_last_error_ino)
2892 printk(": inode %u",
2893 le32_to_cpu(es->s_last_error_ino));
2894 if (es->s_last_error_block)
2895 printk(": block %llu", (unsigned long long)
2896 le64_to_cpu(es->s_last_error_block));
2899 mod_timer(&sbi->s_err_report, jiffies + 24*60*60*HZ); /* Once a day */
2902 /* Find next suitable group and run ext4_init_inode_table */
2903 static int ext4_run_li_request(struct ext4_li_request *elr)
2905 struct ext4_group_desc *gdp = NULL;
2906 ext4_group_t group, ngroups;
2907 struct super_block *sb;
2908 unsigned long timeout = 0;
2912 ngroups = EXT4_SB(sb)->s_groups_count;
2915 for (group = elr->lr_next_group; group < ngroups; group++) {
2916 gdp = ext4_get_group_desc(sb, group, NULL);
2922 if (!(gdp->bg_flags & cpu_to_le16(EXT4_BG_INODE_ZEROED)))
2926 if (group >= ngroups)
2931 ret = ext4_init_inode_table(sb, group,
2932 elr->lr_timeout ? 0 : 1);
2933 if (elr->lr_timeout == 0) {
2934 timeout = (jiffies - timeout) *
2935 elr->lr_sbi->s_li_wait_mult;
2936 elr->lr_timeout = timeout;
2938 elr->lr_next_sched = jiffies + elr->lr_timeout;
2939 elr->lr_next_group = group + 1;
2947 * Remove lr_request from the list_request and free the
2948 * request structure. Should be called with li_list_mtx held
2950 static void ext4_remove_li_request(struct ext4_li_request *elr)
2952 struct ext4_sb_info *sbi;
2959 list_del(&elr->lr_request);
2960 sbi->s_li_request = NULL;
2964 static void ext4_unregister_li_request(struct super_block *sb)
2966 mutex_lock(&ext4_li_mtx);
2967 if (!ext4_li_info) {
2968 mutex_unlock(&ext4_li_mtx);
2972 mutex_lock(&ext4_li_info->li_list_mtx);
2973 ext4_remove_li_request(EXT4_SB(sb)->s_li_request);
2974 mutex_unlock(&ext4_li_info->li_list_mtx);
2975 mutex_unlock(&ext4_li_mtx);
2978 static struct task_struct *ext4_lazyinit_task;
2981 * This is the function where ext4lazyinit thread lives. It walks
2982 * through the request list searching for next scheduled filesystem.
2983 * When such a fs is found, run the lazy initialization request
2984 * (ext4_rn_li_request) and keep track of the time spend in this
2985 * function. Based on that time we compute next schedule time of
2986 * the request. When walking through the list is complete, compute
2987 * next waking time and put itself into sleep.
2989 static int ext4_lazyinit_thread(void *arg)
2991 struct ext4_lazy_init *eli = (struct ext4_lazy_init *)arg;
2992 struct list_head *pos, *n;
2993 struct ext4_li_request *elr;
2994 unsigned long next_wakeup, cur;
2996 BUG_ON(NULL == eli);
3000 next_wakeup = MAX_JIFFY_OFFSET;
3002 mutex_lock(&eli->li_list_mtx);
3003 if (list_empty(&eli->li_request_list)) {
3004 mutex_unlock(&eli->li_list_mtx);
3008 list_for_each_safe(pos, n, &eli->li_request_list) {
3009 elr = list_entry(pos, struct ext4_li_request,
3012 if (time_after_eq(jiffies, elr->lr_next_sched)) {
3013 if (ext4_run_li_request(elr) != 0) {
3014 /* error, remove the lazy_init job */
3015 ext4_remove_li_request(elr);
3020 if (time_before(elr->lr_next_sched, next_wakeup))
3021 next_wakeup = elr->lr_next_sched;
3023 mutex_unlock(&eli->li_list_mtx);
3028 if ((time_after_eq(cur, next_wakeup)) ||
3029 (MAX_JIFFY_OFFSET == next_wakeup)) {
3034 schedule_timeout_interruptible(next_wakeup - cur);
3036 if (kthread_should_stop()) {
3037 ext4_clear_request_list();
3044 * It looks like the request list is empty, but we need
3045 * to check it under the li_list_mtx lock, to prevent any
3046 * additions into it, and of course we should lock ext4_li_mtx
3047 * to atomically free the list and ext4_li_info, because at
3048 * this point another ext4 filesystem could be registering
3051 mutex_lock(&ext4_li_mtx);
3052 mutex_lock(&eli->li_list_mtx);
3053 if (!list_empty(&eli->li_request_list)) {
3054 mutex_unlock(&eli->li_list_mtx);
3055 mutex_unlock(&ext4_li_mtx);
3058 mutex_unlock(&eli->li_list_mtx);
3059 kfree(ext4_li_info);
3060 ext4_li_info = NULL;
3061 mutex_unlock(&ext4_li_mtx);
3066 static void ext4_clear_request_list(void)
3068 struct list_head *pos, *n;
3069 struct ext4_li_request *elr;
3071 mutex_lock(&ext4_li_info->li_list_mtx);
3072 list_for_each_safe(pos, n, &ext4_li_info->li_request_list) {
3073 elr = list_entry(pos, struct ext4_li_request,
3075 ext4_remove_li_request(elr);
3077 mutex_unlock(&ext4_li_info->li_list_mtx);
3080 static int ext4_run_lazyinit_thread(void)
3082 ext4_lazyinit_task = kthread_run(ext4_lazyinit_thread,
3083 ext4_li_info, "ext4lazyinit");
3084 if (IS_ERR(ext4_lazyinit_task)) {
3085 int err = PTR_ERR(ext4_lazyinit_task);
3086 ext4_clear_request_list();
3087 kfree(ext4_li_info);
3088 ext4_li_info = NULL;
3089 printk(KERN_CRIT "EXT4-fs: error %d creating inode table "
3090 "initialization thread\n",
3094 ext4_li_info->li_state |= EXT4_LAZYINIT_RUNNING;
3099 * Check whether it make sense to run itable init. thread or not.
3100 * If there is at least one uninitialized inode table, return
3101 * corresponding group number, else the loop goes through all
3102 * groups and return total number of groups.
3104 static ext4_group_t ext4_has_uninit_itable(struct super_block *sb)
3106 ext4_group_t group, ngroups = EXT4_SB(sb)->s_groups_count;
3107 struct ext4_group_desc *gdp = NULL;
3109 for (group = 0; group < ngroups; group++) {
3110 gdp = ext4_get_group_desc(sb, group, NULL);
3114 if (!(gdp->bg_flags & cpu_to_le16(EXT4_BG_INODE_ZEROED)))
3121 static int ext4_li_info_new(void)
3123 struct ext4_lazy_init *eli = NULL;
3125 eli = kzalloc(sizeof(*eli), GFP_KERNEL);
3129 INIT_LIST_HEAD(&eli->li_request_list);
3130 mutex_init(&eli->li_list_mtx);
3132 eli->li_state |= EXT4_LAZYINIT_QUIT;
3139 static struct ext4_li_request *ext4_li_request_new(struct super_block *sb,
3142 struct ext4_sb_info *sbi = EXT4_SB(sb);
3143 struct ext4_li_request *elr;
3145 elr = kzalloc(sizeof(*elr), GFP_KERNEL);
3151 elr->lr_next_group = start;
3154 * Randomize first schedule time of the request to
3155 * spread the inode table initialization requests
3158 elr->lr_next_sched = jiffies + (prandom_u32() %
3159 (EXT4_DEF_LI_MAX_START_DELAY * HZ));
3163 int ext4_register_li_request(struct super_block *sb,
3164 ext4_group_t first_not_zeroed)
3166 struct ext4_sb_info *sbi = EXT4_SB(sb);
3167 struct ext4_li_request *elr = NULL;
3168 ext4_group_t ngroups = EXT4_SB(sb)->s_groups_count;
3171 mutex_lock(&ext4_li_mtx);
3172 if (sbi->s_li_request != NULL) {
3174 * Reset timeout so it can be computed again, because
3175 * s_li_wait_mult might have changed.
3177 sbi->s_li_request->lr_timeout = 0;
3181 if (first_not_zeroed == ngroups ||
3182 (sb->s_flags & MS_RDONLY) ||
3183 !test_opt(sb, INIT_INODE_TABLE))
3186 elr = ext4_li_request_new(sb, first_not_zeroed);
3192 if (NULL == ext4_li_info) {
3193 ret = ext4_li_info_new();
3198 mutex_lock(&ext4_li_info->li_list_mtx);
3199 list_add(&elr->lr_request, &ext4_li_info->li_request_list);
3200 mutex_unlock(&ext4_li_info->li_list_mtx);
3202 sbi->s_li_request = elr;
3204 * set elr to NULL here since it has been inserted to
3205 * the request_list and the removal and free of it is
3206 * handled by ext4_clear_request_list from now on.
3210 if (!(ext4_li_info->li_state & EXT4_LAZYINIT_RUNNING)) {
3211 ret = ext4_run_lazyinit_thread();
3216 mutex_unlock(&ext4_li_mtx);
3223 * We do not need to lock anything since this is called on
3226 static void ext4_destroy_lazyinit_thread(void)
3229 * If thread exited earlier
3230 * there's nothing to be done.
3232 if (!ext4_li_info || !ext4_lazyinit_task)
3235 kthread_stop(ext4_lazyinit_task);
3238 static int set_journal_csum_feature_set(struct super_block *sb)
3241 int compat, incompat;
3242 struct ext4_sb_info *sbi = EXT4_SB(sb);
3244 if (ext4_has_metadata_csum(sb)) {
3245 /* journal checksum v3 */
3247 incompat = JBD2_FEATURE_INCOMPAT_CSUM_V3;
3249 /* journal checksum v1 */
3250 compat = JBD2_FEATURE_COMPAT_CHECKSUM;
3254 jbd2_journal_clear_features(sbi->s_journal,
3255 JBD2_FEATURE_COMPAT_CHECKSUM, 0,
3256 JBD2_FEATURE_INCOMPAT_CSUM_V3 |
3257 JBD2_FEATURE_INCOMPAT_CSUM_V2);
3258 if (test_opt(sb, JOURNAL_ASYNC_COMMIT)) {
3259 ret = jbd2_journal_set_features(sbi->s_journal,
3261 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT |
3263 } else if (test_opt(sb, JOURNAL_CHECKSUM)) {
3264 ret = jbd2_journal_set_features(sbi->s_journal,
3267 jbd2_journal_clear_features(sbi->s_journal, 0, 0,
3268 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT);
3270 jbd2_journal_clear_features(sbi->s_journal, 0, 0,
3271 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT);
3278 * Note: calculating the overhead so we can be compatible with
3279 * historical BSD practice is quite difficult in the face of
3280 * clusters/bigalloc. This is because multiple metadata blocks from
3281 * different block group can end up in the same allocation cluster.
3282 * Calculating the exact overhead in the face of clustered allocation
3283 * requires either O(all block bitmaps) in memory or O(number of block
3284 * groups**2) in time. We will still calculate the superblock for
3285 * older file systems --- and if we come across with a bigalloc file
3286 * system with zero in s_overhead_clusters the estimate will be close to
3287 * correct especially for very large cluster sizes --- but for newer
3288 * file systems, it's better to calculate this figure once at mkfs
3289 * time, and store it in the superblock. If the superblock value is
3290 * present (even for non-bigalloc file systems), we will use it.
3292 static int count_overhead(struct super_block *sb, ext4_group_t grp,
3295 struct ext4_sb_info *sbi = EXT4_SB(sb);
3296 struct ext4_group_desc *gdp;
3297 ext4_fsblk_t first_block, last_block, b;
3298 ext4_group_t i, ngroups = ext4_get_groups_count(sb);
3299 int s, j, count = 0;
3301 if (!EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_BIGALLOC))
3302 return (ext4_bg_has_super(sb, grp) + ext4_bg_num_gdb(sb, grp) +
3303 sbi->s_itb_per_group + 2);
3305 first_block = le32_to_cpu(sbi->s_es->s_first_data_block) +
3306 (grp * EXT4_BLOCKS_PER_GROUP(sb));
3307 last_block = first_block + EXT4_BLOCKS_PER_GROUP(sb) - 1;
3308 for (i = 0; i < ngroups; i++) {
3309 gdp = ext4_get_group_desc(sb, i, NULL);
3310 b = ext4_block_bitmap(sb, gdp);
3311 if (b >= first_block && b <= last_block) {
3312 ext4_set_bit(EXT4_B2C(sbi, b - first_block), buf);
3315 b = ext4_inode_bitmap(sb, gdp);
3316 if (b >= first_block && b <= last_block) {
3317 ext4_set_bit(EXT4_B2C(sbi, b - first_block), buf);
3320 b = ext4_inode_table(sb, gdp);
3321 if (b >= first_block && b + sbi->s_itb_per_group <= last_block)
3322 for (j = 0; j < sbi->s_itb_per_group; j++, b++) {
3323 int c = EXT4_B2C(sbi, b - first_block);
3324 ext4_set_bit(c, buf);
3330 if (ext4_bg_has_super(sb, grp)) {
3331 ext4_set_bit(s++, buf);
3334 for (j = ext4_bg_num_gdb(sb, grp); j > 0; j--) {
3335 ext4_set_bit(EXT4_B2C(sbi, s++), buf);
3341 return EXT4_CLUSTERS_PER_GROUP(sb) -
3342 ext4_count_free(buf, EXT4_CLUSTERS_PER_GROUP(sb) / 8);
3346 * Compute the overhead and stash it in sbi->s_overhead
3348 int ext4_calculate_overhead(struct super_block *sb)
3350 struct ext4_sb_info *sbi = EXT4_SB(sb);
3351 struct ext4_super_block *es = sbi->s_es;
3352 ext4_group_t i, ngroups = ext4_get_groups_count(sb);
3353 ext4_fsblk_t overhead = 0;
3354 char *buf = (char *) get_zeroed_page(GFP_NOFS);
3360 * Compute the overhead (FS structures). This is constant
3361 * for a given filesystem unless the number of block groups
3362 * changes so we cache the previous value until it does.
3366 * All of the blocks before first_data_block are overhead
3368 overhead = EXT4_B2C(sbi, le32_to_cpu(es->s_first_data_block));
3371 * Add the overhead found in each block group
3373 for (i = 0; i < ngroups; i++) {
3376 blks = count_overhead(sb, i, buf);
3379 memset(buf, 0, PAGE_SIZE);
3382 /* Add the internal journal blocks as well */
3383 if (sbi->s_journal && !sbi->journal_bdev)
3384 overhead += EXT4_NUM_B2C(sbi, sbi->s_journal->j_maxlen);
3386 sbi->s_overhead = overhead;
3388 free_page((unsigned long) buf);
3393 static ext4_fsblk_t ext4_calculate_resv_clusters(struct super_block *sb)
3395 ext4_fsblk_t resv_clusters;
3398 * There's no need to reserve anything when we aren't using extents.
3399 * The space estimates are exact, there are no unwritten extents,
3400 * hole punching doesn't need new metadata... This is needed especially
3401 * to keep ext2/3 backward compatibility.
3403 if (!EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_EXTENTS))
3406 * By default we reserve 2% or 4096 clusters, whichever is smaller.
3407 * This should cover the situations where we can not afford to run
3408 * out of space like for example punch hole, or converting
3409 * unwritten extents in delalloc path. In most cases such
3410 * allocation would require 1, or 2 blocks, higher numbers are
3413 resv_clusters = ext4_blocks_count(EXT4_SB(sb)->s_es) >>
3414 EXT4_SB(sb)->s_cluster_bits;
3416 do_div(resv_clusters, 50);
3417 resv_clusters = min_t(ext4_fsblk_t, resv_clusters, 4096);
3419 return resv_clusters;
3423 static int ext4_reserve_clusters(struct ext4_sb_info *sbi, ext4_fsblk_t count)
3425 ext4_fsblk_t clusters = ext4_blocks_count(sbi->s_es) >>
3426 sbi->s_cluster_bits;
3428 if (count >= clusters)
3431 atomic64_set(&sbi->s_resv_clusters, count);
3435 static int ext4_fill_super(struct super_block *sb, void *data, int silent)
3437 char *orig_data = kstrdup(data, GFP_KERNEL);
3438 struct buffer_head *bh;
3439 struct ext4_super_block *es = NULL;
3440 struct ext4_sb_info *sbi;
3442 ext4_fsblk_t sb_block = get_sb_block(&data);
3443 ext4_fsblk_t logical_sb_block;
3444 unsigned long offset = 0;
3445 unsigned long journal_devnum = 0;
3446 unsigned long def_mount_opts;
3451 int blocksize, clustersize;
3452 unsigned int db_count;
3454 int needs_recovery, has_huge_files, has_bigalloc;
3457 unsigned int journal_ioprio = DEFAULT_JOURNAL_IOPRIO;
3458 ext4_group_t first_not_zeroed;
3460 sbi = kzalloc(sizeof(*sbi), GFP_KERNEL);
3464 sbi->s_blockgroup_lock =
3465 kzalloc(sizeof(struct blockgroup_lock), GFP_KERNEL);
3466 if (!sbi->s_blockgroup_lock) {
3470 sb->s_fs_info = sbi;
3472 sbi->s_inode_readahead_blks = EXT4_DEF_INODE_READAHEAD_BLKS;
3473 sbi->s_sb_block = sb_block;
3474 if (sb->s_bdev->bd_part)
3475 sbi->s_sectors_written_start =
3476 part_stat_read(sb->s_bdev->bd_part, sectors[1]);
3478 /* Cleanup superblock name */
3479 for (cp = sb->s_id; (cp = strchr(cp, '/'));)
3482 /* -EINVAL is default */
3484 blocksize = sb_min_blocksize(sb, EXT4_MIN_BLOCK_SIZE);
3486 ext4_msg(sb, KERN_ERR, "unable to set blocksize");
3491 * The ext4 superblock will not be buffer aligned for other than 1kB
3492 * block sizes. We need to calculate the offset from buffer start.
3494 if (blocksize != EXT4_MIN_BLOCK_SIZE) {
3495 logical_sb_block = sb_block * EXT4_MIN_BLOCK_SIZE;
3496 offset = do_div(logical_sb_block, blocksize);
3498 logical_sb_block = sb_block;
3501 if (!(bh = sb_bread_unmovable(sb, logical_sb_block))) {
3502 ext4_msg(sb, KERN_ERR, "unable to read superblock");
3506 * Note: s_es must be initialized as soon as possible because
3507 * some ext4 macro-instructions depend on its value
3509 es = (struct ext4_super_block *) (bh->b_data + offset);
3511 sb->s_magic = le16_to_cpu(es->s_magic);
3512 if (sb->s_magic != EXT4_SUPER_MAGIC)
3514 sbi->s_kbytes_written = le64_to_cpu(es->s_kbytes_written);
3516 /* Warn if metadata_csum and gdt_csum are both set. */
3517 if (EXT4_HAS_RO_COMPAT_FEATURE(sb,
3518 EXT4_FEATURE_RO_COMPAT_METADATA_CSUM) &&
3519 EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_GDT_CSUM))
3520 ext4_warning(sb, "metadata_csum and uninit_bg are "
3521 "redundant flags; please run fsck.");
3523 /* Check for a known checksum algorithm */
3524 if (!ext4_verify_csum_type(sb, es)) {
3525 ext4_msg(sb, KERN_ERR, "VFS: Found ext4 filesystem with "
3526 "unknown checksum algorithm.");
3531 /* Load the checksum driver */
3532 if (EXT4_HAS_RO_COMPAT_FEATURE(sb,
3533 EXT4_FEATURE_RO_COMPAT_METADATA_CSUM)) {
3534 sbi->s_chksum_driver = crypto_alloc_shash("crc32c", 0, 0);
3535 if (IS_ERR(sbi->s_chksum_driver)) {
3536 ext4_msg(sb, KERN_ERR, "Cannot load crc32c driver.");
3537 ret = PTR_ERR(sbi->s_chksum_driver);
3538 sbi->s_chksum_driver = NULL;
3543 /* Check superblock checksum */
3544 if (!ext4_superblock_csum_verify(sb, es)) {
3545 ext4_msg(sb, KERN_ERR, "VFS: Found ext4 filesystem with "
3546 "invalid superblock checksum. Run e2fsck?");
3551 /* Precompute checksum seed for all metadata */
3552 if (ext4_has_metadata_csum(sb))
3553 sbi->s_csum_seed = ext4_chksum(sbi, ~0, es->s_uuid,
3554 sizeof(es->s_uuid));
3556 /* Set defaults before we parse the mount options */
3557 def_mount_opts = le32_to_cpu(es->s_default_mount_opts);
3558 set_opt(sb, INIT_INODE_TABLE);
3559 if (def_mount_opts & EXT4_DEFM_DEBUG)
3561 if (def_mount_opts & EXT4_DEFM_BSDGROUPS)
3563 if (def_mount_opts & EXT4_DEFM_UID16)
3564 set_opt(sb, NO_UID32);
3565 /* xattr user namespace & acls are now defaulted on */
3566 set_opt(sb, XATTR_USER);
3567 #ifdef CONFIG_EXT4_FS_POSIX_ACL
3568 set_opt(sb, POSIX_ACL);
3570 /* don't forget to enable journal_csum when metadata_csum is enabled. */
3571 if (ext4_has_metadata_csum(sb))
3572 set_opt(sb, JOURNAL_CHECKSUM);
3574 if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_DATA)
3575 set_opt(sb, JOURNAL_DATA);
3576 else if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_ORDERED)
3577 set_opt(sb, ORDERED_DATA);
3578 else if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_WBACK)
3579 set_opt(sb, WRITEBACK_DATA);
3581 if (le16_to_cpu(sbi->s_es->s_errors) == EXT4_ERRORS_PANIC)
3582 set_opt(sb, ERRORS_PANIC);
3583 else if (le16_to_cpu(sbi->s_es->s_errors) == EXT4_ERRORS_CONTINUE)
3584 set_opt(sb, ERRORS_CONT);
3586 set_opt(sb, ERRORS_RO);
3587 /* block_validity enabled by default; disable with noblock_validity */
3588 set_opt(sb, BLOCK_VALIDITY);
3589 if (def_mount_opts & EXT4_DEFM_DISCARD)
3590 set_opt(sb, DISCARD);
3592 sbi->s_resuid = make_kuid(&init_user_ns, le16_to_cpu(es->s_def_resuid));
3593 sbi->s_resgid = make_kgid(&init_user_ns, le16_to_cpu(es->s_def_resgid));
3594 sbi->s_commit_interval = JBD2_DEFAULT_MAX_COMMIT_AGE * HZ;
3595 sbi->s_min_batch_time = EXT4_DEF_MIN_BATCH_TIME;
3596 sbi->s_max_batch_time = EXT4_DEF_MAX_BATCH_TIME;
3598 if ((def_mount_opts & EXT4_DEFM_NOBARRIER) == 0)
3599 set_opt(sb, BARRIER);
3602 * enable delayed allocation by default
3603 * Use -o nodelalloc to turn it off
3605 if (!IS_EXT3_SB(sb) && !IS_EXT2_SB(sb) &&
3606 ((def_mount_opts & EXT4_DEFM_NODELALLOC) == 0))
3607 set_opt(sb, DELALLOC);
3610 * set default s_li_wait_mult for lazyinit, for the case there is
3611 * no mount option specified.
3613 sbi->s_li_wait_mult = EXT4_DEF_LI_WAIT_MULT;
3615 if (!parse_options((char *) sbi->s_es->s_mount_opts, sb,
3616 &journal_devnum, &journal_ioprio, 0)) {
3617 ext4_msg(sb, KERN_WARNING,
3618 "failed to parse options in superblock: %s",
3619 sbi->s_es->s_mount_opts);
3621 sbi->s_def_mount_opt = sbi->s_mount_opt;
3622 if (!parse_options((char *) data, sb, &journal_devnum,
3623 &journal_ioprio, 0))
3626 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA) {
3627 printk_once(KERN_WARNING "EXT4-fs: Warning: mounting "
3628 "with data=journal disables delayed "
3629 "allocation and O_DIRECT support!\n");
3630 if (test_opt2(sb, EXPLICIT_DELALLOC)) {
3631 ext4_msg(sb, KERN_ERR, "can't mount with "
3632 "both data=journal and delalloc");
3635 if (test_opt(sb, DIOREAD_NOLOCK)) {
3636 ext4_msg(sb, KERN_ERR, "can't mount with "
3637 "both data=journal and dioread_nolock");
3640 if (test_opt(sb, DAX)) {
3641 ext4_msg(sb, KERN_ERR, "can't mount with "
3642 "both data=journal and dax");
3645 if (test_opt(sb, DELALLOC))
3646 clear_opt(sb, DELALLOC);
3649 sb->s_flags = (sb->s_flags & ~MS_POSIXACL) |
3650 (test_opt(sb, POSIX_ACL) ? MS_POSIXACL : 0);
3652 if (le32_to_cpu(es->s_rev_level) == EXT4_GOOD_OLD_REV &&
3653 (EXT4_HAS_COMPAT_FEATURE(sb, ~0U) ||
3654 EXT4_HAS_RO_COMPAT_FEATURE(sb, ~0U) ||
3655 EXT4_HAS_INCOMPAT_FEATURE(sb, ~0U)))
3656 ext4_msg(sb, KERN_WARNING,
3657 "feature flags set on rev 0 fs, "
3658 "running e2fsck is recommended");
3660 if (es->s_creator_os == cpu_to_le32(EXT4_OS_HURD)) {
3661 set_opt2(sb, HURD_COMPAT);
3662 if (EXT4_HAS_INCOMPAT_FEATURE(sb,
3663 EXT4_FEATURE_INCOMPAT_64BIT)) {
3664 ext4_msg(sb, KERN_ERR,
3665 "The Hurd can't support 64-bit file systems");
3670 if (IS_EXT2_SB(sb)) {
3671 if (ext2_feature_set_ok(sb))
3672 ext4_msg(sb, KERN_INFO, "mounting ext2 file system "
3673 "using the ext4 subsystem");
3675 ext4_msg(sb, KERN_ERR, "couldn't mount as ext2 due "
3676 "to feature incompatibilities");
3681 if (IS_EXT3_SB(sb)) {
3682 if (ext3_feature_set_ok(sb))
3683 ext4_msg(sb, KERN_INFO, "mounting ext3 file system "
3684 "using the ext4 subsystem");
3686 ext4_msg(sb, KERN_ERR, "couldn't mount as ext3 due "
3687 "to feature incompatibilities");
3693 * Check feature flags regardless of the revision level, since we
3694 * previously didn't change the revision level when setting the flags,
3695 * so there is a chance incompat flags are set on a rev 0 filesystem.
3697 if (!ext4_feature_set_ok(sb, (sb->s_flags & MS_RDONLY)))
3700 blocksize = BLOCK_SIZE << le32_to_cpu(es->s_log_block_size);
3701 if (blocksize < EXT4_MIN_BLOCK_SIZE ||
3702 blocksize > EXT4_MAX_BLOCK_SIZE) {
3703 ext4_msg(sb, KERN_ERR,
3704 "Unsupported filesystem blocksize %d", blocksize);
3708 if (sbi->s_mount_opt & EXT4_MOUNT_DAX) {
3709 if (blocksize != PAGE_SIZE) {
3710 ext4_msg(sb, KERN_ERR,
3711 "error: unsupported blocksize for dax");
3714 if (!sb->s_bdev->bd_disk->fops->direct_access) {
3715 ext4_msg(sb, KERN_ERR,
3716 "error: device does not support dax");
3721 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_ENCRYPT) &&
3722 es->s_encryption_level) {
3723 ext4_msg(sb, KERN_ERR, "Unsupported encryption level %d",
3724 es->s_encryption_level);
3728 if (sb->s_blocksize != blocksize) {
3729 /* Validate the filesystem blocksize */
3730 if (!sb_set_blocksize(sb, blocksize)) {
3731 ext4_msg(sb, KERN_ERR, "bad block size %d",
3737 logical_sb_block = sb_block * EXT4_MIN_BLOCK_SIZE;
3738 offset = do_div(logical_sb_block, blocksize);
3739 bh = sb_bread_unmovable(sb, logical_sb_block);
3741 ext4_msg(sb, KERN_ERR,
3742 "Can't read superblock on 2nd try");
3745 es = (struct ext4_super_block *)(bh->b_data + offset);
3747 if (es->s_magic != cpu_to_le16(EXT4_SUPER_MAGIC)) {
3748 ext4_msg(sb, KERN_ERR,
3749 "Magic mismatch, very weird!");
3754 has_huge_files = EXT4_HAS_RO_COMPAT_FEATURE(sb,
3755 EXT4_FEATURE_RO_COMPAT_HUGE_FILE);
3756 sbi->s_bitmap_maxbytes = ext4_max_bitmap_size(sb->s_blocksize_bits,
3758 sb->s_maxbytes = ext4_max_size(sb->s_blocksize_bits, has_huge_files);
3760 if (le32_to_cpu(es->s_rev_level) == EXT4_GOOD_OLD_REV) {
3761 sbi->s_inode_size = EXT4_GOOD_OLD_INODE_SIZE;
3762 sbi->s_first_ino = EXT4_GOOD_OLD_FIRST_INO;
3764 sbi->s_inode_size = le16_to_cpu(es->s_inode_size);
3765 sbi->s_first_ino = le32_to_cpu(es->s_first_ino);
3766 if ((sbi->s_inode_size < EXT4_GOOD_OLD_INODE_SIZE) ||
3767 (!is_power_of_2(sbi->s_inode_size)) ||
3768 (sbi->s_inode_size > blocksize)) {
3769 ext4_msg(sb, KERN_ERR,
3770 "unsupported inode size: %d",
3774 if (sbi->s_inode_size > EXT4_GOOD_OLD_INODE_SIZE)
3775 sb->s_time_gran = 1 << (EXT4_EPOCH_BITS - 2);
3778 sbi->s_desc_size = le16_to_cpu(es->s_desc_size);
3779 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_64BIT)) {
3780 if (sbi->s_desc_size < EXT4_MIN_DESC_SIZE_64BIT ||
3781 sbi->s_desc_size > EXT4_MAX_DESC_SIZE ||
3782 !is_power_of_2(sbi->s_desc_size)) {
3783 ext4_msg(sb, KERN_ERR,
3784 "unsupported descriptor size %lu",
3789 sbi->s_desc_size = EXT4_MIN_DESC_SIZE;
3791 sbi->s_blocks_per_group = le32_to_cpu(es->s_blocks_per_group);
3792 sbi->s_inodes_per_group = le32_to_cpu(es->s_inodes_per_group);
3793 if (EXT4_INODE_SIZE(sb) == 0 || EXT4_INODES_PER_GROUP(sb) == 0)
3796 sbi->s_inodes_per_block = blocksize / EXT4_INODE_SIZE(sb);
3797 if (sbi->s_inodes_per_block == 0)
3799 sbi->s_itb_per_group = sbi->s_inodes_per_group /
3800 sbi->s_inodes_per_block;
3801 sbi->s_desc_per_block = blocksize / EXT4_DESC_SIZE(sb);
3803 sbi->s_mount_state = le16_to_cpu(es->s_state);
3804 sbi->s_addr_per_block_bits = ilog2(EXT4_ADDR_PER_BLOCK(sb));
3805 sbi->s_desc_per_block_bits = ilog2(EXT4_DESC_PER_BLOCK(sb));
3807 for (i = 0; i < 4; i++)
3808 sbi->s_hash_seed[i] = le32_to_cpu(es->s_hash_seed[i]);
3809 sbi->s_def_hash_version = es->s_def_hash_version;
3810 if (EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_DIR_INDEX)) {
3811 i = le32_to_cpu(es->s_flags);
3812 if (i & EXT2_FLAGS_UNSIGNED_HASH)
3813 sbi->s_hash_unsigned = 3;
3814 else if ((i & EXT2_FLAGS_SIGNED_HASH) == 0) {
3815 #ifdef __CHAR_UNSIGNED__
3816 if (!(sb->s_flags & MS_RDONLY))
3818 cpu_to_le32(EXT2_FLAGS_UNSIGNED_HASH);
3819 sbi->s_hash_unsigned = 3;
3821 if (!(sb->s_flags & MS_RDONLY))
3823 cpu_to_le32(EXT2_FLAGS_SIGNED_HASH);
3828 /* Handle clustersize */
3829 clustersize = BLOCK_SIZE << le32_to_cpu(es->s_log_cluster_size);
3830 has_bigalloc = EXT4_HAS_RO_COMPAT_FEATURE(sb,
3831 EXT4_FEATURE_RO_COMPAT_BIGALLOC);
3833 if (clustersize < blocksize) {
3834 ext4_msg(sb, KERN_ERR,
3835 "cluster size (%d) smaller than "
3836 "block size (%d)", clustersize, blocksize);
3839 sbi->s_cluster_bits = le32_to_cpu(es->s_log_cluster_size) -
3840 le32_to_cpu(es->s_log_block_size);
3841 sbi->s_clusters_per_group =
3842 le32_to_cpu(es->s_clusters_per_group);
3843 if (sbi->s_clusters_per_group > blocksize * 8) {
3844 ext4_msg(sb, KERN_ERR,
3845 "#clusters per group too big: %lu",
3846 sbi->s_clusters_per_group);
3849 if (sbi->s_blocks_per_group !=
3850 (sbi->s_clusters_per_group * (clustersize / blocksize))) {
3851 ext4_msg(sb, KERN_ERR, "blocks per group (%lu) and "
3852 "clusters per group (%lu) inconsistent",
3853 sbi->s_blocks_per_group,
3854 sbi->s_clusters_per_group);
3858 if (clustersize != blocksize) {
3859 ext4_warning(sb, "fragment/cluster size (%d) != "
3860 "block size (%d)", clustersize,
3862 clustersize = blocksize;
3864 if (sbi->s_blocks_per_group > blocksize * 8) {
3865 ext4_msg(sb, KERN_ERR,
3866 "#blocks per group too big: %lu",
3867 sbi->s_blocks_per_group);
3870 sbi->s_clusters_per_group = sbi->s_blocks_per_group;
3871 sbi->s_cluster_bits = 0;
3873 sbi->s_cluster_ratio = clustersize / blocksize;
3875 if (sbi->s_inodes_per_group > blocksize * 8) {
3876 ext4_msg(sb, KERN_ERR,
3877 "#inodes per group too big: %lu",
3878 sbi->s_inodes_per_group);
3882 /* Do we have standard group size of clustersize * 8 blocks ? */
3883 if (sbi->s_blocks_per_group == clustersize << 3)
3884 set_opt2(sb, STD_GROUP_SIZE);
3887 * Test whether we have more sectors than will fit in sector_t,
3888 * and whether the max offset is addressable by the page cache.
3890 err = generic_check_addressable(sb->s_blocksize_bits,
3891 ext4_blocks_count(es));
3893 ext4_msg(sb, KERN_ERR, "filesystem"
3894 " too large to mount safely on this system");
3895 if (sizeof(sector_t) < 8)
3896 ext4_msg(sb, KERN_WARNING, "CONFIG_LBDAF not enabled");
3900 if (EXT4_BLOCKS_PER_GROUP(sb) == 0)
3903 /* check blocks count against device size */
3904 blocks_count = sb->s_bdev->bd_inode->i_size >> sb->s_blocksize_bits;
3905 if (blocks_count && ext4_blocks_count(es) > blocks_count) {
3906 ext4_msg(sb, KERN_WARNING, "bad geometry: block count %llu "
3907 "exceeds size of device (%llu blocks)",
3908 ext4_blocks_count(es), blocks_count);
3913 * It makes no sense for the first data block to be beyond the end
3914 * of the filesystem.
3916 if (le32_to_cpu(es->s_first_data_block) >= ext4_blocks_count(es)) {
3917 ext4_msg(sb, KERN_WARNING, "bad geometry: first data "
3918 "block %u is beyond end of filesystem (%llu)",
3919 le32_to_cpu(es->s_first_data_block),
3920 ext4_blocks_count(es));
3923 blocks_count = (ext4_blocks_count(es) -
3924 le32_to_cpu(es->s_first_data_block) +
3925 EXT4_BLOCKS_PER_GROUP(sb) - 1);
3926 do_div(blocks_count, EXT4_BLOCKS_PER_GROUP(sb));
3927 if (blocks_count > ((uint64_t)1<<32) - EXT4_DESC_PER_BLOCK(sb)) {
3928 ext4_msg(sb, KERN_WARNING, "groups count too large: %u "
3929 "(block count %llu, first data block %u, "
3930 "blocks per group %lu)", sbi->s_groups_count,
3931 ext4_blocks_count(es),
3932 le32_to_cpu(es->s_first_data_block),
3933 EXT4_BLOCKS_PER_GROUP(sb));
3936 sbi->s_groups_count = blocks_count;
3937 sbi->s_blockfile_groups = min_t(ext4_group_t, sbi->s_groups_count,
3938 (EXT4_MAX_BLOCK_FILE_PHYS / EXT4_BLOCKS_PER_GROUP(sb)));
3939 db_count = (sbi->s_groups_count + EXT4_DESC_PER_BLOCK(sb) - 1) /
3940 EXT4_DESC_PER_BLOCK(sb);
3941 sbi->s_group_desc = ext4_kvmalloc(db_count *
3942 sizeof(struct buffer_head *),
3944 if (sbi->s_group_desc == NULL) {
3945 ext4_msg(sb, KERN_ERR, "not enough memory");
3951 sbi->s_proc = proc_mkdir(sb->s_id, ext4_proc_root);
3954 proc_create_data("options", S_IRUGO, sbi->s_proc,
3955 &ext4_seq_options_fops, sb);
3957 bgl_lock_init(sbi->s_blockgroup_lock);
3959 for (i = 0; i < db_count; i++) {
3960 block = descriptor_loc(sb, logical_sb_block, i);
3961 sbi->s_group_desc[i] = sb_bread_unmovable(sb, block);
3962 if (!sbi->s_group_desc[i]) {
3963 ext4_msg(sb, KERN_ERR,
3964 "can't read group descriptor %d", i);
3969 if (!ext4_check_descriptors(sb, &first_not_zeroed)) {
3970 ext4_msg(sb, KERN_ERR, "group descriptors corrupted!");
3974 sbi->s_gdb_count = db_count;
3975 get_random_bytes(&sbi->s_next_generation, sizeof(u32));
3976 spin_lock_init(&sbi->s_next_gen_lock);
3978 setup_timer(&sbi->s_err_report, print_daily_error_info,
3979 (unsigned long) sb);
3981 /* Register extent status tree shrinker */
3982 if (ext4_es_register_shrinker(sbi))
3985 sbi->s_stripe = ext4_get_stripe_size(sbi);
3986 sbi->s_extent_max_zeroout_kb = 32;
3989 * set up enough so that it can read an inode
3991 sb->s_op = &ext4_sops;
3992 sb->s_export_op = &ext4_export_ops;
3993 sb->s_xattr = ext4_xattr_handlers;
3995 sb->dq_op = &ext4_quota_operations;
3996 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_QUOTA))
3997 sb->s_qcop = &dquot_quotactl_sysfile_ops;
3999 sb->s_qcop = &ext4_qctl_operations;
4000 sb->s_quota_types = QTYPE_MASK_USR | QTYPE_MASK_GRP;
4002 memcpy(sb->s_uuid, es->s_uuid, sizeof(es->s_uuid));
4004 INIT_LIST_HEAD(&sbi->s_orphan); /* unlinked but open files */
4005 mutex_init(&sbi->s_orphan_lock);
4009 needs_recovery = (es->s_last_orphan != 0 ||
4010 EXT4_HAS_INCOMPAT_FEATURE(sb,
4011 EXT4_FEATURE_INCOMPAT_RECOVER));
4013 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_MMP) &&
4014 !(sb->s_flags & MS_RDONLY))
4015 if (ext4_multi_mount_protect(sb, le64_to_cpu(es->s_mmp_block)))
4016 goto failed_mount3a;
4019 * The first inode we look at is the journal inode. Don't try
4020 * root first: it may be modified in the journal!
4022 if (!test_opt(sb, NOLOAD) &&
4023 EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL)) {
4024 if (ext4_load_journal(sb, es, journal_devnum))
4025 goto failed_mount3a;
4026 } else if (test_opt(sb, NOLOAD) && !(sb->s_flags & MS_RDONLY) &&
4027 EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER)) {
4028 ext4_msg(sb, KERN_ERR, "required journal recovery "
4029 "suppressed and not mounted read-only");
4030 goto failed_mount_wq;
4032 clear_opt(sb, DATA_FLAGS);
4033 sbi->s_journal = NULL;
4038 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_64BIT) &&
4039 !jbd2_journal_set_features(EXT4_SB(sb)->s_journal, 0, 0,
4040 JBD2_FEATURE_INCOMPAT_64BIT)) {
4041 ext4_msg(sb, KERN_ERR, "Failed to set 64-bit journal feature");
4042 goto failed_mount_wq;
4045 if (!set_journal_csum_feature_set(sb)) {
4046 ext4_msg(sb, KERN_ERR, "Failed to set journal checksum "
4048 goto failed_mount_wq;
4051 /* We have now updated the journal if required, so we can
4052 * validate the data journaling mode. */
4053 switch (test_opt(sb, DATA_FLAGS)) {
4055 /* No mode set, assume a default based on the journal
4056 * capabilities: ORDERED_DATA if the journal can
4057 * cope, else JOURNAL_DATA
4059 if (jbd2_journal_check_available_features
4060 (sbi->s_journal, 0, 0, JBD2_FEATURE_INCOMPAT_REVOKE))
4061 set_opt(sb, ORDERED_DATA);
4063 set_opt(sb, JOURNAL_DATA);
4066 case EXT4_MOUNT_ORDERED_DATA:
4067 case EXT4_MOUNT_WRITEBACK_DATA:
4068 if (!jbd2_journal_check_available_features
4069 (sbi->s_journal, 0, 0, JBD2_FEATURE_INCOMPAT_REVOKE)) {
4070 ext4_msg(sb, KERN_ERR, "Journal does not support "
4071 "requested data journaling mode");
4072 goto failed_mount_wq;
4077 set_task_ioprio(sbi->s_journal->j_task, journal_ioprio);
4079 sbi->s_journal->j_commit_callback = ext4_journal_commit_callback;
4082 if (ext4_mballoc_ready) {
4083 sbi->s_mb_cache = ext4_xattr_create_cache(sb->s_id);
4084 if (!sbi->s_mb_cache) {
4085 ext4_msg(sb, KERN_ERR, "Failed to create an mb_cache");
4086 goto failed_mount_wq;
4090 if ((DUMMY_ENCRYPTION_ENABLED(sbi) ||
4091 EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_ENCRYPT)) &&
4092 (blocksize != PAGE_CACHE_SIZE)) {
4093 ext4_msg(sb, KERN_ERR,
4094 "Unsupported blocksize for fs encryption");
4095 goto failed_mount_wq;
4098 if (DUMMY_ENCRYPTION_ENABLED(sbi) &&
4099 !(sb->s_flags & MS_RDONLY) &&
4100 !EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_ENCRYPT)) {
4101 EXT4_SET_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_ENCRYPT);
4102 ext4_commit_super(sb, 1);
4106 * Get the # of file system overhead blocks from the
4107 * superblock if present.
4109 if (es->s_overhead_clusters)
4110 sbi->s_overhead = le32_to_cpu(es->s_overhead_clusters);
4112 err = ext4_calculate_overhead(sb);
4114 goto failed_mount_wq;
4118 * The maximum number of concurrent works can be high and
4119 * concurrency isn't really necessary. Limit it to 1.
4121 EXT4_SB(sb)->rsv_conversion_wq =
4122 alloc_workqueue("ext4-rsv-conversion", WQ_MEM_RECLAIM | WQ_UNBOUND, 1);
4123 if (!EXT4_SB(sb)->rsv_conversion_wq) {
4124 printk(KERN_ERR "EXT4-fs: failed to create workqueue\n");
4130 * The jbd2_journal_load will have done any necessary log recovery,
4131 * so we can safely mount the rest of the filesystem now.
4134 root = ext4_iget(sb, EXT4_ROOT_INO);
4136 ext4_msg(sb, KERN_ERR, "get root inode failed");
4137 ret = PTR_ERR(root);
4141 if (!S_ISDIR(root->i_mode) || !root->i_blocks || !root->i_size) {
4142 ext4_msg(sb, KERN_ERR, "corrupt root inode, run e2fsck");
4146 sb->s_root = d_make_root(root);
4148 ext4_msg(sb, KERN_ERR, "get root dentry failed");
4153 if (ext4_setup_super(sb, es, sb->s_flags & MS_RDONLY))
4154 sb->s_flags |= MS_RDONLY;
4156 /* determine the minimum size of new large inodes, if present */
4157 if (sbi->s_inode_size > EXT4_GOOD_OLD_INODE_SIZE) {
4158 sbi->s_want_extra_isize = sizeof(struct ext4_inode) -
4159 EXT4_GOOD_OLD_INODE_SIZE;
4160 if (EXT4_HAS_RO_COMPAT_FEATURE(sb,
4161 EXT4_FEATURE_RO_COMPAT_EXTRA_ISIZE)) {
4162 if (sbi->s_want_extra_isize <
4163 le16_to_cpu(es->s_want_extra_isize))
4164 sbi->s_want_extra_isize =
4165 le16_to_cpu(es->s_want_extra_isize);
4166 if (sbi->s_want_extra_isize <
4167 le16_to_cpu(es->s_min_extra_isize))
4168 sbi->s_want_extra_isize =
4169 le16_to_cpu(es->s_min_extra_isize);
4172 /* Check if enough inode space is available */
4173 if (EXT4_GOOD_OLD_INODE_SIZE + sbi->s_want_extra_isize >
4174 sbi->s_inode_size) {
4175 sbi->s_want_extra_isize = sizeof(struct ext4_inode) -
4176 EXT4_GOOD_OLD_INODE_SIZE;
4177 ext4_msg(sb, KERN_INFO, "required extra inode space not"
4181 err = ext4_reserve_clusters(sbi, ext4_calculate_resv_clusters(sb));
4183 ext4_msg(sb, KERN_ERR, "failed to reserve %llu clusters for "
4184 "reserved pool", ext4_calculate_resv_clusters(sb));
4185 goto failed_mount4a;
4188 err = ext4_setup_system_zone(sb);
4190 ext4_msg(sb, KERN_ERR, "failed to initialize system "
4192 goto failed_mount4a;
4196 err = ext4_mb_init(sb);
4198 ext4_msg(sb, KERN_ERR, "failed to initialize mballoc (%d)",
4203 block = ext4_count_free_clusters(sb);
4204 ext4_free_blocks_count_set(sbi->s_es,
4205 EXT4_C2B(sbi, block));
4206 err = percpu_counter_init(&sbi->s_freeclusters_counter, block,
4209 unsigned long freei = ext4_count_free_inodes(sb);
4210 sbi->s_es->s_free_inodes_count = cpu_to_le32(freei);
4211 err = percpu_counter_init(&sbi->s_freeinodes_counter, freei,
4215 err = percpu_counter_init(&sbi->s_dirs_counter,
4216 ext4_count_dirs(sb), GFP_KERNEL);
4218 err = percpu_counter_init(&sbi->s_dirtyclusters_counter, 0,
4221 ext4_msg(sb, KERN_ERR, "insufficient memory");
4225 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_FLEX_BG))
4226 if (!ext4_fill_flex_info(sb)) {
4227 ext4_msg(sb, KERN_ERR,
4228 "unable to initialize "
4229 "flex_bg meta info!");
4233 err = ext4_register_li_request(sb, first_not_zeroed);
4237 sbi->s_kobj.kset = ext4_kset;
4238 init_completion(&sbi->s_kobj_unregister);
4239 err = kobject_init_and_add(&sbi->s_kobj, &ext4_ktype, NULL,
4245 /* Enable quota usage during mount. */
4246 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_QUOTA) &&
4247 !(sb->s_flags & MS_RDONLY)) {
4248 err = ext4_enable_quotas(sb);
4252 #endif /* CONFIG_QUOTA */
4254 EXT4_SB(sb)->s_mount_state |= EXT4_ORPHAN_FS;
4255 ext4_orphan_cleanup(sb, es);
4256 EXT4_SB(sb)->s_mount_state &= ~EXT4_ORPHAN_FS;
4257 if (needs_recovery) {
4258 ext4_msg(sb, KERN_INFO, "recovery complete");
4259 ext4_mark_recovery_complete(sb, es);
4261 if (EXT4_SB(sb)->s_journal) {
4262 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA)
4263 descr = " journalled data mode";
4264 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA)
4265 descr = " ordered data mode";
4267 descr = " writeback data mode";
4269 descr = "out journal";
4271 if (test_opt(sb, DISCARD)) {
4272 struct request_queue *q = bdev_get_queue(sb->s_bdev);
4273 if (!blk_queue_discard(q))
4274 ext4_msg(sb, KERN_WARNING,
4275 "mounting with \"discard\" option, but "
4276 "the device does not support discard");
4279 ext4_msg(sb, KERN_INFO, "mounted filesystem with%s. "
4280 "Opts: %s%s%s", descr, sbi->s_es->s_mount_opts,
4281 *sbi->s_es->s_mount_opts ? "; " : "", orig_data);
4283 if (es->s_error_count)
4284 mod_timer(&sbi->s_err_report, jiffies + 300*HZ); /* 5 minutes */
4286 /* Enable message ratelimiting. Default is 10 messages per 5 secs. */
4287 ratelimit_state_init(&sbi->s_err_ratelimit_state, 5 * HZ, 10);
4288 ratelimit_state_init(&sbi->s_warning_ratelimit_state, 5 * HZ, 10);
4289 ratelimit_state_init(&sbi->s_msg_ratelimit_state, 5 * HZ, 10);
4296 ext4_msg(sb, KERN_ERR, "VFS: Can't find ext4 filesystem");
4301 kobject_del(&sbi->s_kobj);
4304 ext4_unregister_li_request(sb);
4306 ext4_mb_release(sb);
4307 if (sbi->s_flex_groups)
4308 kvfree(sbi->s_flex_groups);
4309 percpu_counter_destroy(&sbi->s_freeclusters_counter);
4310 percpu_counter_destroy(&sbi->s_freeinodes_counter);
4311 percpu_counter_destroy(&sbi->s_dirs_counter);
4312 percpu_counter_destroy(&sbi->s_dirtyclusters_counter);
4314 ext4_ext_release(sb);
4315 ext4_release_system_zone(sb);
4320 ext4_msg(sb, KERN_ERR, "mount failed");
4321 if (EXT4_SB(sb)->rsv_conversion_wq)
4322 destroy_workqueue(EXT4_SB(sb)->rsv_conversion_wq);
4324 if (sbi->s_journal) {
4325 jbd2_journal_destroy(sbi->s_journal);
4326 sbi->s_journal = NULL;
4329 ext4_es_unregister_shrinker(sbi);
4331 del_timer_sync(&sbi->s_err_report);
4333 kthread_stop(sbi->s_mmp_tsk);
4335 for (i = 0; i < db_count; i++)
4336 brelse(sbi->s_group_desc[i]);
4337 kvfree(sbi->s_group_desc);
4339 if (sbi->s_chksum_driver)
4340 crypto_free_shash(sbi->s_chksum_driver);
4342 remove_proc_entry("options", sbi->s_proc);
4343 remove_proc_entry(sb->s_id, ext4_proc_root);
4346 for (i = 0; i < EXT4_MAXQUOTAS; i++)
4347 kfree(sbi->s_qf_names[i]);
4349 ext4_blkdev_remove(sbi);
4352 sb->s_fs_info = NULL;
4353 kfree(sbi->s_blockgroup_lock);
4357 return err ? err : ret;
4361 * Setup any per-fs journal parameters now. We'll do this both on
4362 * initial mount, once the journal has been initialised but before we've
4363 * done any recovery; and again on any subsequent remount.
4365 static void ext4_init_journal_params(struct super_block *sb, journal_t *journal)
4367 struct ext4_sb_info *sbi = EXT4_SB(sb);
4369 journal->j_commit_interval = sbi->s_commit_interval;
4370 journal->j_min_batch_time = sbi->s_min_batch_time;
4371 journal->j_max_batch_time = sbi->s_max_batch_time;
4373 write_lock(&journal->j_state_lock);
4374 if (test_opt(sb, BARRIER))
4375 journal->j_flags |= JBD2_BARRIER;
4377 journal->j_flags &= ~JBD2_BARRIER;
4378 if (test_opt(sb, DATA_ERR_ABORT))
4379 journal->j_flags |= JBD2_ABORT_ON_SYNCDATA_ERR;
4381 journal->j_flags &= ~JBD2_ABORT_ON_SYNCDATA_ERR;
4382 write_unlock(&journal->j_state_lock);
4385 static journal_t *ext4_get_journal(struct super_block *sb,
4386 unsigned int journal_inum)
4388 struct inode *journal_inode;
4391 BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL));
4393 /* First, test for the existence of a valid inode on disk. Bad
4394 * things happen if we iget() an unused inode, as the subsequent
4395 * iput() will try to delete it. */
4397 journal_inode = ext4_iget(sb, journal_inum);
4398 if (IS_ERR(journal_inode)) {
4399 ext4_msg(sb, KERN_ERR, "no journal found");
4402 if (!journal_inode->i_nlink) {
4403 make_bad_inode(journal_inode);
4404 iput(journal_inode);
4405 ext4_msg(sb, KERN_ERR, "journal inode is deleted");
4409 jbd_debug(2, "Journal inode found at %p: %lld bytes\n",
4410 journal_inode, journal_inode->i_size);
4411 if (!S_ISREG(journal_inode->i_mode)) {
4412 ext4_msg(sb, KERN_ERR, "invalid journal inode");
4413 iput(journal_inode);
4417 journal = jbd2_journal_init_inode(journal_inode);
4419 ext4_msg(sb, KERN_ERR, "Could not load journal inode");
4420 iput(journal_inode);
4423 journal->j_private = sb;
4424 ext4_init_journal_params(sb, journal);
4428 static journal_t *ext4_get_dev_journal(struct super_block *sb,
4431 struct buffer_head *bh;
4435 int hblock, blocksize;
4436 ext4_fsblk_t sb_block;
4437 unsigned long offset;
4438 struct ext4_super_block *es;
4439 struct block_device *bdev;
4441 BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL));
4443 bdev = ext4_blkdev_get(j_dev, sb);
4447 blocksize = sb->s_blocksize;
4448 hblock = bdev_logical_block_size(bdev);
4449 if (blocksize < hblock) {
4450 ext4_msg(sb, KERN_ERR,
4451 "blocksize too small for journal device");
4455 sb_block = EXT4_MIN_BLOCK_SIZE / blocksize;
4456 offset = EXT4_MIN_BLOCK_SIZE % blocksize;
4457 set_blocksize(bdev, blocksize);
4458 if (!(bh = __bread(bdev, sb_block, blocksize))) {
4459 ext4_msg(sb, KERN_ERR, "couldn't read superblock of "
4460 "external journal");
4464 es = (struct ext4_super_block *) (bh->b_data + offset);
4465 if ((le16_to_cpu(es->s_magic) != EXT4_SUPER_MAGIC) ||
4466 !(le32_to_cpu(es->s_feature_incompat) &
4467 EXT4_FEATURE_INCOMPAT_JOURNAL_DEV)) {
4468 ext4_msg(sb, KERN_ERR, "external journal has "
4474 if ((le32_to_cpu(es->s_feature_ro_compat) &
4475 EXT4_FEATURE_RO_COMPAT_METADATA_CSUM) &&
4476 es->s_checksum != ext4_superblock_csum(sb, es)) {
4477 ext4_msg(sb, KERN_ERR, "external journal has "
4478 "corrupt superblock");
4483 if (memcmp(EXT4_SB(sb)->s_es->s_journal_uuid, es->s_uuid, 16)) {
4484 ext4_msg(sb, KERN_ERR, "journal UUID does not match");
4489 len = ext4_blocks_count(es);
4490 start = sb_block + 1;
4491 brelse(bh); /* we're done with the superblock */
4493 journal = jbd2_journal_init_dev(bdev, sb->s_bdev,
4494 start, len, blocksize);
4496 ext4_msg(sb, KERN_ERR, "failed to create device journal");
4499 journal->j_private = sb;
4500 ll_rw_block(READ | REQ_META | REQ_PRIO, 1, &journal->j_sb_buffer);
4501 wait_on_buffer(journal->j_sb_buffer);
4502 if (!buffer_uptodate(journal->j_sb_buffer)) {
4503 ext4_msg(sb, KERN_ERR, "I/O error on journal device");
4506 if (be32_to_cpu(journal->j_superblock->s_nr_users) != 1) {
4507 ext4_msg(sb, KERN_ERR, "External journal has more than one "
4508 "user (unsupported) - %d",
4509 be32_to_cpu(journal->j_superblock->s_nr_users));
4512 EXT4_SB(sb)->journal_bdev = bdev;
4513 ext4_init_journal_params(sb, journal);
4517 jbd2_journal_destroy(journal);
4519 ext4_blkdev_put(bdev);
4523 static int ext4_load_journal(struct super_block *sb,
4524 struct ext4_super_block *es,
4525 unsigned long journal_devnum)
4528 unsigned int journal_inum = le32_to_cpu(es->s_journal_inum);
4531 int really_read_only;
4533 BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL));
4535 if (journal_devnum &&
4536 journal_devnum != le32_to_cpu(es->s_journal_dev)) {
4537 ext4_msg(sb, KERN_INFO, "external journal device major/minor "
4538 "numbers have changed");
4539 journal_dev = new_decode_dev(journal_devnum);
4541 journal_dev = new_decode_dev(le32_to_cpu(es->s_journal_dev));
4543 really_read_only = bdev_read_only(sb->s_bdev);
4546 * Are we loading a blank journal or performing recovery after a
4547 * crash? For recovery, we need to check in advance whether we
4548 * can get read-write access to the device.
4550 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER)) {
4551 if (sb->s_flags & MS_RDONLY) {
4552 ext4_msg(sb, KERN_INFO, "INFO: recovery "
4553 "required on readonly filesystem");
4554 if (really_read_only) {
4555 ext4_msg(sb, KERN_ERR, "write access "
4556 "unavailable, cannot proceed");
4559 ext4_msg(sb, KERN_INFO, "write access will "
4560 "be enabled during recovery");
4564 if (journal_inum && journal_dev) {
4565 ext4_msg(sb, KERN_ERR, "filesystem has both journal "
4566 "and inode journals!");
4571 if (!(journal = ext4_get_journal(sb, journal_inum)))
4574 if (!(journal = ext4_get_dev_journal(sb, journal_dev)))
4578 if (!(journal->j_flags & JBD2_BARRIER))
4579 ext4_msg(sb, KERN_INFO, "barriers disabled");
4581 if (!EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER))
4582 err = jbd2_journal_wipe(journal, !really_read_only);
4584 char *save = kmalloc(EXT4_S_ERR_LEN, GFP_KERNEL);
4586 memcpy(save, ((char *) es) +
4587 EXT4_S_ERR_START, EXT4_S_ERR_LEN);
4588 err = jbd2_journal_load(journal);
4590 memcpy(((char *) es) + EXT4_S_ERR_START,
4591 save, EXT4_S_ERR_LEN);
4596 ext4_msg(sb, KERN_ERR, "error loading journal");
4597 jbd2_journal_destroy(journal);
4601 EXT4_SB(sb)->s_journal = journal;
4602 ext4_clear_journal_err(sb, es);
4604 if (!really_read_only && journal_devnum &&
4605 journal_devnum != le32_to_cpu(es->s_journal_dev)) {
4606 es->s_journal_dev = cpu_to_le32(journal_devnum);
4608 /* Make sure we flush the recovery flag to disk. */
4609 ext4_commit_super(sb, 1);
4615 static int ext4_commit_super(struct super_block *sb, int sync)
4617 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
4618 struct buffer_head *sbh = EXT4_SB(sb)->s_sbh;
4623 if (buffer_write_io_error(sbh)) {
4625 * Oh, dear. A previous attempt to write the
4626 * superblock failed. This could happen because the
4627 * USB device was yanked out. Or it could happen to
4628 * be a transient write error and maybe the block will
4629 * be remapped. Nothing we can do but to retry the
4630 * write and hope for the best.
4632 ext4_msg(sb, KERN_ERR, "previous I/O error to "
4633 "superblock detected");
4634 clear_buffer_write_io_error(sbh);
4635 set_buffer_uptodate(sbh);
4638 * If the file system is mounted read-only, don't update the
4639 * superblock write time. This avoids updating the superblock
4640 * write time when we are mounting the root file system
4641 * read/only but we need to replay the journal; at that point,
4642 * for people who are east of GMT and who make their clock
4643 * tick in localtime for Windows bug-for-bug compatibility,
4644 * the clock is set in the future, and this will cause e2fsck
4645 * to complain and force a full file system check.
4647 if (!(sb->s_flags & MS_RDONLY))
4648 es->s_wtime = cpu_to_le32(get_seconds());
4649 if (sb->s_bdev->bd_part)
4650 es->s_kbytes_written =
4651 cpu_to_le64(EXT4_SB(sb)->s_kbytes_written +
4652 ((part_stat_read(sb->s_bdev->bd_part, sectors[1]) -
4653 EXT4_SB(sb)->s_sectors_written_start) >> 1));
4655 es->s_kbytes_written =
4656 cpu_to_le64(EXT4_SB(sb)->s_kbytes_written);
4657 if (percpu_counter_initialized(&EXT4_SB(sb)->s_freeclusters_counter))
4658 ext4_free_blocks_count_set(es,
4659 EXT4_C2B(EXT4_SB(sb), percpu_counter_sum_positive(
4660 &EXT4_SB(sb)->s_freeclusters_counter)));
4661 if (percpu_counter_initialized(&EXT4_SB(sb)->s_freeinodes_counter))
4662 es->s_free_inodes_count =
4663 cpu_to_le32(percpu_counter_sum_positive(
4664 &EXT4_SB(sb)->s_freeinodes_counter));
4665 BUFFER_TRACE(sbh, "marking dirty");
4666 ext4_superblock_csum_set(sb);
4667 mark_buffer_dirty(sbh);
4669 error = sync_dirty_buffer(sbh);
4673 error = buffer_write_io_error(sbh);
4675 ext4_msg(sb, KERN_ERR, "I/O error while writing "
4677 clear_buffer_write_io_error(sbh);
4678 set_buffer_uptodate(sbh);
4685 * Have we just finished recovery? If so, and if we are mounting (or
4686 * remounting) the filesystem readonly, then we will end up with a
4687 * consistent fs on disk. Record that fact.
4689 static void ext4_mark_recovery_complete(struct super_block *sb,
4690 struct ext4_super_block *es)
4692 journal_t *journal = EXT4_SB(sb)->s_journal;
4694 if (!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL)) {
4695 BUG_ON(journal != NULL);
4698 jbd2_journal_lock_updates(journal);
4699 if (jbd2_journal_flush(journal) < 0)
4702 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER) &&
4703 sb->s_flags & MS_RDONLY) {
4704 EXT4_CLEAR_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
4705 ext4_commit_super(sb, 1);
4709 jbd2_journal_unlock_updates(journal);
4713 * If we are mounting (or read-write remounting) a filesystem whose journal
4714 * has recorded an error from a previous lifetime, move that error to the
4715 * main filesystem now.
4717 static void ext4_clear_journal_err(struct super_block *sb,
4718 struct ext4_super_block *es)
4724 BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL));
4726 journal = EXT4_SB(sb)->s_journal;
4729 * Now check for any error status which may have been recorded in the
4730 * journal by a prior ext4_error() or ext4_abort()
4733 j_errno = jbd2_journal_errno(journal);
4737 errstr = ext4_decode_error(sb, j_errno, nbuf);
4738 ext4_warning(sb, "Filesystem error recorded "
4739 "from previous mount: %s", errstr);
4740 ext4_warning(sb, "Marking fs in need of filesystem check.");
4742 EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS;
4743 es->s_state |= cpu_to_le16(EXT4_ERROR_FS);
4744 ext4_commit_super(sb, 1);
4746 jbd2_journal_clear_err(journal);
4747 jbd2_journal_update_sb_errno(journal);
4752 * Force the running and committing transactions to commit,
4753 * and wait on the commit.
4755 int ext4_force_commit(struct super_block *sb)
4759 if (sb->s_flags & MS_RDONLY)
4762 journal = EXT4_SB(sb)->s_journal;
4763 return ext4_journal_force_commit(journal);
4766 static int ext4_sync_fs(struct super_block *sb, int wait)
4770 bool needs_barrier = false;
4771 struct ext4_sb_info *sbi = EXT4_SB(sb);
4773 trace_ext4_sync_fs(sb, wait);
4774 flush_workqueue(sbi->rsv_conversion_wq);
4776 * Writeback quota in non-journalled quota case - journalled quota has
4779 dquot_writeback_dquots(sb, -1);
4781 * Data writeback is possible w/o journal transaction, so barrier must
4782 * being sent at the end of the function. But we can skip it if
4783 * transaction_commit will do it for us.
4785 if (sbi->s_journal) {
4786 target = jbd2_get_latest_transaction(sbi->s_journal);
4787 if (wait && sbi->s_journal->j_flags & JBD2_BARRIER &&
4788 !jbd2_trans_will_send_data_barrier(sbi->s_journal, target))
4789 needs_barrier = true;
4791 if (jbd2_journal_start_commit(sbi->s_journal, &target)) {
4793 ret = jbd2_log_wait_commit(sbi->s_journal,
4796 } else if (wait && test_opt(sb, BARRIER))
4797 needs_barrier = true;
4798 if (needs_barrier) {
4800 err = blkdev_issue_flush(sb->s_bdev, GFP_KERNEL, NULL);
4809 * LVM calls this function before a (read-only) snapshot is created. This
4810 * gives us a chance to flush the journal completely and mark the fs clean.
4812 * Note that only this function cannot bring a filesystem to be in a clean
4813 * state independently. It relies on upper layer to stop all data & metadata
4816 static int ext4_freeze(struct super_block *sb)
4821 if (sb->s_flags & MS_RDONLY)
4824 journal = EXT4_SB(sb)->s_journal;
4827 /* Now we set up the journal barrier. */
4828 jbd2_journal_lock_updates(journal);
4831 * Don't clear the needs_recovery flag if we failed to
4832 * flush the journal.
4834 error = jbd2_journal_flush(journal);
4839 /* Journal blocked and flushed, clear needs_recovery flag. */
4840 EXT4_CLEAR_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
4841 error = ext4_commit_super(sb, 1);
4844 /* we rely on upper layer to stop further updates */
4845 jbd2_journal_unlock_updates(journal);
4850 * Called by LVM after the snapshot is done. We need to reset the RECOVER
4851 * flag here, even though the filesystem is not technically dirty yet.
4853 static int ext4_unfreeze(struct super_block *sb)
4855 if (sb->s_flags & MS_RDONLY)
4858 /* Reset the needs_recovery flag before the fs is unlocked. */
4859 EXT4_SET_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
4860 ext4_commit_super(sb, 1);
4865 * Structure to save mount options for ext4_remount's benefit
4867 struct ext4_mount_options {
4868 unsigned long s_mount_opt;
4869 unsigned long s_mount_opt2;
4872 unsigned long s_commit_interval;
4873 u32 s_min_batch_time, s_max_batch_time;
4876 char *s_qf_names[EXT4_MAXQUOTAS];
4880 static int ext4_remount(struct super_block *sb, int *flags, char *data)
4882 struct ext4_super_block *es;
4883 struct ext4_sb_info *sbi = EXT4_SB(sb);
4884 unsigned long old_sb_flags;
4885 struct ext4_mount_options old_opts;
4886 int enable_quota = 0;
4888 unsigned int journal_ioprio = DEFAULT_JOURNAL_IOPRIO;
4893 char *orig_data = kstrdup(data, GFP_KERNEL);
4895 /* Store the original options */
4896 old_sb_flags = sb->s_flags;
4897 old_opts.s_mount_opt = sbi->s_mount_opt;
4898 old_opts.s_mount_opt2 = sbi->s_mount_opt2;
4899 old_opts.s_resuid = sbi->s_resuid;
4900 old_opts.s_resgid = sbi->s_resgid;
4901 old_opts.s_commit_interval = sbi->s_commit_interval;
4902 old_opts.s_min_batch_time = sbi->s_min_batch_time;
4903 old_opts.s_max_batch_time = sbi->s_max_batch_time;
4905 old_opts.s_jquota_fmt = sbi->s_jquota_fmt;
4906 for (i = 0; i < EXT4_MAXQUOTAS; i++)
4907 if (sbi->s_qf_names[i]) {
4908 old_opts.s_qf_names[i] = kstrdup(sbi->s_qf_names[i],
4910 if (!old_opts.s_qf_names[i]) {
4911 for (j = 0; j < i; j++)
4912 kfree(old_opts.s_qf_names[j]);
4917 old_opts.s_qf_names[i] = NULL;
4919 if (sbi->s_journal && sbi->s_journal->j_task->io_context)
4920 journal_ioprio = sbi->s_journal->j_task->io_context->ioprio;
4922 if (!parse_options(data, sb, NULL, &journal_ioprio, 1)) {
4927 if ((old_opts.s_mount_opt & EXT4_MOUNT_JOURNAL_CHECKSUM) ^
4928 test_opt(sb, JOURNAL_CHECKSUM)) {
4929 ext4_msg(sb, KERN_ERR, "changing journal_checksum "
4930 "during remount not supported; ignoring");
4931 sbi->s_mount_opt ^= EXT4_MOUNT_JOURNAL_CHECKSUM;
4934 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA) {
4935 if (test_opt2(sb, EXPLICIT_DELALLOC)) {
4936 ext4_msg(sb, KERN_ERR, "can't mount with "
4937 "both data=journal and delalloc");
4941 if (test_opt(sb, DIOREAD_NOLOCK)) {
4942 ext4_msg(sb, KERN_ERR, "can't mount with "
4943 "both data=journal and dioread_nolock");
4947 if (test_opt(sb, DAX)) {
4948 ext4_msg(sb, KERN_ERR, "can't mount with "
4949 "both data=journal and dax");
4955 if ((sbi->s_mount_opt ^ old_opts.s_mount_opt) & EXT4_MOUNT_DAX) {
4956 ext4_msg(sb, KERN_WARNING, "warning: refusing change of "
4957 "dax flag with busy inodes while remounting");
4958 sbi->s_mount_opt ^= EXT4_MOUNT_DAX;
4961 if (sbi->s_mount_flags & EXT4_MF_FS_ABORTED)
4962 ext4_abort(sb, "Abort forced by user");
4964 sb->s_flags = (sb->s_flags & ~MS_POSIXACL) |
4965 (test_opt(sb, POSIX_ACL) ? MS_POSIXACL : 0);
4969 if (sbi->s_journal) {
4970 ext4_init_journal_params(sb, sbi->s_journal);
4971 set_task_ioprio(sbi->s_journal->j_task, journal_ioprio);
4974 if ((*flags & MS_RDONLY) != (sb->s_flags & MS_RDONLY)) {
4975 if (sbi->s_mount_flags & EXT4_MF_FS_ABORTED) {
4980 if (*flags & MS_RDONLY) {
4981 err = sync_filesystem(sb);
4984 err = dquot_suspend(sb, -1);
4989 * First of all, the unconditional stuff we have to do
4990 * to disable replay of the journal when we next remount
4992 sb->s_flags |= MS_RDONLY;
4995 * OK, test if we are remounting a valid rw partition
4996 * readonly, and if so set the rdonly flag and then
4997 * mark the partition as valid again.
4999 if (!(es->s_state & cpu_to_le16(EXT4_VALID_FS)) &&
5000 (sbi->s_mount_state & EXT4_VALID_FS))
5001 es->s_state = cpu_to_le16(sbi->s_mount_state);
5004 ext4_mark_recovery_complete(sb, es);
5006 /* Make sure we can mount this feature set readwrite */
5007 if (EXT4_HAS_RO_COMPAT_FEATURE(sb,
5008 EXT4_FEATURE_RO_COMPAT_READONLY) ||
5009 !ext4_feature_set_ok(sb, 0)) {
5014 * Make sure the group descriptor checksums
5015 * are sane. If they aren't, refuse to remount r/w.
5017 for (g = 0; g < sbi->s_groups_count; g++) {
5018 struct ext4_group_desc *gdp =
5019 ext4_get_group_desc(sb, g, NULL);
5021 if (!ext4_group_desc_csum_verify(sb, g, gdp)) {
5022 ext4_msg(sb, KERN_ERR,
5023 "ext4_remount: Checksum for group %u failed (%u!=%u)",
5024 g, le16_to_cpu(ext4_group_desc_csum(sbi, g, gdp)),
5025 le16_to_cpu(gdp->bg_checksum));
5032 * If we have an unprocessed orphan list hanging
5033 * around from a previously readonly bdev mount,
5034 * require a full umount/remount for now.
5036 if (es->s_last_orphan) {
5037 ext4_msg(sb, KERN_WARNING, "Couldn't "
5038 "remount RDWR because of unprocessed "
5039 "orphan inode list. Please "
5040 "umount/remount instead");
5046 * Mounting a RDONLY partition read-write, so reread
5047 * and store the current valid flag. (It may have
5048 * been changed by e2fsck since we originally mounted
5052 ext4_clear_journal_err(sb, es);
5053 sbi->s_mount_state = le16_to_cpu(es->s_state);
5054 if (!ext4_setup_super(sb, es, 0))
5055 sb->s_flags &= ~MS_RDONLY;
5056 if (EXT4_HAS_INCOMPAT_FEATURE(sb,
5057 EXT4_FEATURE_INCOMPAT_MMP))
5058 if (ext4_multi_mount_protect(sb,
5059 le64_to_cpu(es->s_mmp_block))) {
5068 * Reinitialize lazy itable initialization thread based on
5071 if ((sb->s_flags & MS_RDONLY) || !test_opt(sb, INIT_INODE_TABLE))
5072 ext4_unregister_li_request(sb);
5074 ext4_group_t first_not_zeroed;
5075 first_not_zeroed = ext4_has_uninit_itable(sb);
5076 ext4_register_li_request(sb, first_not_zeroed);
5079 ext4_setup_system_zone(sb);
5080 if (sbi->s_journal == NULL && !(old_sb_flags & MS_RDONLY))
5081 ext4_commit_super(sb, 1);
5084 /* Release old quota file names */
5085 for (i = 0; i < EXT4_MAXQUOTAS; i++)
5086 kfree(old_opts.s_qf_names[i]);
5088 if (sb_any_quota_suspended(sb))
5089 dquot_resume(sb, -1);
5090 else if (EXT4_HAS_RO_COMPAT_FEATURE(sb,
5091 EXT4_FEATURE_RO_COMPAT_QUOTA)) {
5092 err = ext4_enable_quotas(sb);
5099 *flags = (*flags & ~MS_LAZYTIME) | (sb->s_flags & MS_LAZYTIME);
5100 ext4_msg(sb, KERN_INFO, "re-mounted. Opts: %s", orig_data);
5105 sb->s_flags = old_sb_flags;
5106 sbi->s_mount_opt = old_opts.s_mount_opt;
5107 sbi->s_mount_opt2 = old_opts.s_mount_opt2;
5108 sbi->s_resuid = old_opts.s_resuid;
5109 sbi->s_resgid = old_opts.s_resgid;
5110 sbi->s_commit_interval = old_opts.s_commit_interval;
5111 sbi->s_min_batch_time = old_opts.s_min_batch_time;
5112 sbi->s_max_batch_time = old_opts.s_max_batch_time;
5114 sbi->s_jquota_fmt = old_opts.s_jquota_fmt;
5115 for (i = 0; i < EXT4_MAXQUOTAS; i++) {
5116 kfree(sbi->s_qf_names[i]);
5117 sbi->s_qf_names[i] = old_opts.s_qf_names[i];
5124 static int ext4_statfs(struct dentry *dentry, struct kstatfs *buf)
5126 struct super_block *sb = dentry->d_sb;
5127 struct ext4_sb_info *sbi = EXT4_SB(sb);
5128 struct ext4_super_block *es = sbi->s_es;
5129 ext4_fsblk_t overhead = 0, resv_blocks;
5132 resv_blocks = EXT4_C2B(sbi, atomic64_read(&sbi->s_resv_clusters));
5134 if (!test_opt(sb, MINIX_DF))
5135 overhead = sbi->s_overhead;
5137 buf->f_type = EXT4_SUPER_MAGIC;
5138 buf->f_bsize = sb->s_blocksize;
5139 buf->f_blocks = ext4_blocks_count(es) - EXT4_C2B(sbi, overhead);
5140 bfree = percpu_counter_sum_positive(&sbi->s_freeclusters_counter) -
5141 percpu_counter_sum_positive(&sbi->s_dirtyclusters_counter);
5142 /* prevent underflow in case that few free space is available */
5143 buf->f_bfree = EXT4_C2B(sbi, max_t(s64, bfree, 0));
5144 buf->f_bavail = buf->f_bfree -
5145 (ext4_r_blocks_count(es) + resv_blocks);
5146 if (buf->f_bfree < (ext4_r_blocks_count(es) + resv_blocks))
5148 buf->f_files = le32_to_cpu(es->s_inodes_count);
5149 buf->f_ffree = percpu_counter_sum_positive(&sbi->s_freeinodes_counter);
5150 buf->f_namelen = EXT4_NAME_LEN;
5151 fsid = le64_to_cpup((void *)es->s_uuid) ^
5152 le64_to_cpup((void *)es->s_uuid + sizeof(u64));
5153 buf->f_fsid.val[0] = fsid & 0xFFFFFFFFUL;
5154 buf->f_fsid.val[1] = (fsid >> 32) & 0xFFFFFFFFUL;
5159 /* Helper function for writing quotas on sync - we need to start transaction
5160 * before quota file is locked for write. Otherwise the are possible deadlocks:
5161 * Process 1 Process 2
5162 * ext4_create() quota_sync()
5163 * jbd2_journal_start() write_dquot()
5164 * dquot_initialize() down(dqio_mutex)
5165 * down(dqio_mutex) jbd2_journal_start()
5171 static inline struct inode *dquot_to_inode(struct dquot *dquot)
5173 return sb_dqopt(dquot->dq_sb)->files[dquot->dq_id.type];
5176 static int ext4_write_dquot(struct dquot *dquot)
5180 struct inode *inode;
5182 inode = dquot_to_inode(dquot);
5183 handle = ext4_journal_start(inode, EXT4_HT_QUOTA,
5184 EXT4_QUOTA_TRANS_BLOCKS(dquot->dq_sb));
5186 return PTR_ERR(handle);
5187 ret = dquot_commit(dquot);
5188 err = ext4_journal_stop(handle);
5194 static int ext4_acquire_dquot(struct dquot *dquot)
5199 handle = ext4_journal_start(dquot_to_inode(dquot), EXT4_HT_QUOTA,
5200 EXT4_QUOTA_INIT_BLOCKS(dquot->dq_sb));
5202 return PTR_ERR(handle);
5203 ret = dquot_acquire(dquot);
5204 err = ext4_journal_stop(handle);
5210 static int ext4_release_dquot(struct dquot *dquot)
5215 handle = ext4_journal_start(dquot_to_inode(dquot), EXT4_HT_QUOTA,
5216 EXT4_QUOTA_DEL_BLOCKS(dquot->dq_sb));
5217 if (IS_ERR(handle)) {
5218 /* Release dquot anyway to avoid endless cycle in dqput() */
5219 dquot_release(dquot);
5220 return PTR_ERR(handle);
5222 ret = dquot_release(dquot);
5223 err = ext4_journal_stop(handle);
5229 static int ext4_mark_dquot_dirty(struct dquot *dquot)
5231 struct super_block *sb = dquot->dq_sb;
5232 struct ext4_sb_info *sbi = EXT4_SB(sb);
5234 /* Are we journaling quotas? */
5235 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_QUOTA) ||
5236 sbi->s_qf_names[USRQUOTA] || sbi->s_qf_names[GRPQUOTA]) {
5237 dquot_mark_dquot_dirty(dquot);
5238 return ext4_write_dquot(dquot);
5240 return dquot_mark_dquot_dirty(dquot);
5244 static int ext4_write_info(struct super_block *sb, int type)
5249 /* Data block + inode block */
5250 handle = ext4_journal_start(d_inode(sb->s_root), EXT4_HT_QUOTA, 2);
5252 return PTR_ERR(handle);
5253 ret = dquot_commit_info(sb, type);
5254 err = ext4_journal_stop(handle);
5261 * Turn on quotas during mount time - we need to find
5262 * the quota file and such...
5264 static int ext4_quota_on_mount(struct super_block *sb, int type)
5266 return dquot_quota_on_mount(sb, EXT4_SB(sb)->s_qf_names[type],
5267 EXT4_SB(sb)->s_jquota_fmt, type);
5271 * Standard function to be called on quota_on
5273 static int ext4_quota_on(struct super_block *sb, int type, int format_id,
5278 if (!test_opt(sb, QUOTA))
5281 /* Quotafile not on the same filesystem? */
5282 if (path->dentry->d_sb != sb)
5284 /* Journaling quota? */
5285 if (EXT4_SB(sb)->s_qf_names[type]) {
5286 /* Quotafile not in fs root? */
5287 if (path->dentry->d_parent != sb->s_root)
5288 ext4_msg(sb, KERN_WARNING,
5289 "Quota file not on filesystem root. "
5290 "Journaled quota will not work");
5294 * When we journal data on quota file, we have to flush journal to see
5295 * all updates to the file when we bypass pagecache...
5297 if (EXT4_SB(sb)->s_journal &&
5298 ext4_should_journal_data(d_inode(path->dentry))) {
5300 * We don't need to lock updates but journal_flush() could
5301 * otherwise be livelocked...
5303 jbd2_journal_lock_updates(EXT4_SB(sb)->s_journal);
5304 err = jbd2_journal_flush(EXT4_SB(sb)->s_journal);
5305 jbd2_journal_unlock_updates(EXT4_SB(sb)->s_journal);
5310 return dquot_quota_on(sb, type, format_id, path);
5313 static int ext4_quota_enable(struct super_block *sb, int type, int format_id,
5317 struct inode *qf_inode;
5318 unsigned long qf_inums[EXT4_MAXQUOTAS] = {
5319 le32_to_cpu(EXT4_SB(sb)->s_es->s_usr_quota_inum),
5320 le32_to_cpu(EXT4_SB(sb)->s_es->s_grp_quota_inum)
5323 BUG_ON(!EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_QUOTA));
5325 if (!qf_inums[type])
5328 qf_inode = ext4_iget(sb, qf_inums[type]);
5329 if (IS_ERR(qf_inode)) {
5330 ext4_error(sb, "Bad quota inode # %lu", qf_inums[type]);
5331 return PTR_ERR(qf_inode);
5334 /* Don't account quota for quota files to avoid recursion */
5335 qf_inode->i_flags |= S_NOQUOTA;
5336 err = dquot_enable(qf_inode, type, format_id, flags);
5342 /* Enable usage tracking for all quota types. */
5343 static int ext4_enable_quotas(struct super_block *sb)
5346 unsigned long qf_inums[EXT4_MAXQUOTAS] = {
5347 le32_to_cpu(EXT4_SB(sb)->s_es->s_usr_quota_inum),
5348 le32_to_cpu(EXT4_SB(sb)->s_es->s_grp_quota_inum)
5351 sb_dqopt(sb)->flags |= DQUOT_QUOTA_SYS_FILE;
5352 for (type = 0; type < EXT4_MAXQUOTAS; type++) {
5353 if (qf_inums[type]) {
5354 err = ext4_quota_enable(sb, type, QFMT_VFS_V1,
5355 DQUOT_USAGE_ENABLED);
5358 "Failed to enable quota tracking "
5359 "(type=%d, err=%d). Please run "
5360 "e2fsck to fix.", type, err);
5368 static int ext4_quota_off(struct super_block *sb, int type)
5370 struct inode *inode = sb_dqopt(sb)->files[type];
5373 /* Force all delayed allocation blocks to be allocated.
5374 * Caller already holds s_umount sem */
5375 if (test_opt(sb, DELALLOC))
5376 sync_filesystem(sb);
5381 /* Update modification times of quota files when userspace can
5382 * start looking at them */
5383 handle = ext4_journal_start(inode, EXT4_HT_QUOTA, 1);
5386 inode->i_mtime = inode->i_ctime = CURRENT_TIME;
5387 ext4_mark_inode_dirty(handle, inode);
5388 ext4_journal_stop(handle);
5391 return dquot_quota_off(sb, type);
5394 /* Read data from quotafile - avoid pagecache and such because we cannot afford
5395 * acquiring the locks... As quota files are never truncated and quota code
5396 * itself serializes the operations (and no one else should touch the files)
5397 * we don't have to be afraid of races */
5398 static ssize_t ext4_quota_read(struct super_block *sb, int type, char *data,
5399 size_t len, loff_t off)
5401 struct inode *inode = sb_dqopt(sb)->files[type];
5402 ext4_lblk_t blk = off >> EXT4_BLOCK_SIZE_BITS(sb);
5403 int offset = off & (sb->s_blocksize - 1);
5406 struct buffer_head *bh;
5407 loff_t i_size = i_size_read(inode);
5411 if (off+len > i_size)
5414 while (toread > 0) {
5415 tocopy = sb->s_blocksize - offset < toread ?
5416 sb->s_blocksize - offset : toread;
5417 bh = ext4_bread(NULL, inode, blk, 0);
5420 if (!bh) /* A hole? */
5421 memset(data, 0, tocopy);
5423 memcpy(data, bh->b_data+offset, tocopy);
5433 /* Write to quotafile (we know the transaction is already started and has
5434 * enough credits) */
5435 static ssize_t ext4_quota_write(struct super_block *sb, int type,
5436 const char *data, size_t len, loff_t off)
5438 struct inode *inode = sb_dqopt(sb)->files[type];
5439 ext4_lblk_t blk = off >> EXT4_BLOCK_SIZE_BITS(sb);
5440 int err, offset = off & (sb->s_blocksize - 1);
5442 struct buffer_head *bh;
5443 handle_t *handle = journal_current_handle();
5445 if (EXT4_SB(sb)->s_journal && !handle) {
5446 ext4_msg(sb, KERN_WARNING, "Quota write (off=%llu, len=%llu)"
5447 " cancelled because transaction is not started",
5448 (unsigned long long)off, (unsigned long long)len);
5452 * Since we account only one data block in transaction credits,
5453 * then it is impossible to cross a block boundary.
5455 if (sb->s_blocksize - offset < len) {
5456 ext4_msg(sb, KERN_WARNING, "Quota write (off=%llu, len=%llu)"
5457 " cancelled because not block aligned",
5458 (unsigned long long)off, (unsigned long long)len);
5463 bh = ext4_bread(handle, inode, blk,
5464 EXT4_GET_BLOCKS_CREATE |
5465 EXT4_GET_BLOCKS_METADATA_NOFAIL);
5466 } while (IS_ERR(bh) && (PTR_ERR(bh) == -ENOSPC) &&
5467 ext4_should_retry_alloc(inode->i_sb, &retries));
5472 BUFFER_TRACE(bh, "get write access");
5473 err = ext4_journal_get_write_access(handle, bh);
5479 memcpy(bh->b_data+offset, data, len);
5480 flush_dcache_page(bh->b_page);
5482 err = ext4_handle_dirty_metadata(handle, NULL, bh);
5485 if (inode->i_size < off + len) {
5486 i_size_write(inode, off + len);
5487 EXT4_I(inode)->i_disksize = inode->i_size;
5488 ext4_mark_inode_dirty(handle, inode);
5495 static struct dentry *ext4_mount(struct file_system_type *fs_type, int flags,
5496 const char *dev_name, void *data)
5498 return mount_bdev(fs_type, flags, dev_name, data, ext4_fill_super);
5501 #if !defined(CONFIG_EXT2_FS) && !defined(CONFIG_EXT2_FS_MODULE) && defined(CONFIG_EXT4_USE_FOR_EXT23)
5502 static inline void register_as_ext2(void)
5504 int err = register_filesystem(&ext2_fs_type);
5507 "EXT4-fs: Unable to register as ext2 (%d)\n", err);
5510 static inline void unregister_as_ext2(void)
5512 unregister_filesystem(&ext2_fs_type);
5515 static inline int ext2_feature_set_ok(struct super_block *sb)
5517 if (EXT4_HAS_INCOMPAT_FEATURE(sb, ~EXT2_FEATURE_INCOMPAT_SUPP))
5519 if (sb->s_flags & MS_RDONLY)
5521 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, ~EXT2_FEATURE_RO_COMPAT_SUPP))
5526 static inline void register_as_ext2(void) { }
5527 static inline void unregister_as_ext2(void) { }
5528 static inline int ext2_feature_set_ok(struct super_block *sb) { return 0; }
5531 #if !defined(CONFIG_EXT3_FS) && !defined(CONFIG_EXT3_FS_MODULE) && defined(CONFIG_EXT4_USE_FOR_EXT23)
5532 static inline void register_as_ext3(void)
5534 int err = register_filesystem(&ext3_fs_type);
5537 "EXT4-fs: Unable to register as ext3 (%d)\n", err);
5540 static inline void unregister_as_ext3(void)
5542 unregister_filesystem(&ext3_fs_type);
5545 static inline int ext3_feature_set_ok(struct super_block *sb)
5547 if (EXT4_HAS_INCOMPAT_FEATURE(sb, ~EXT3_FEATURE_INCOMPAT_SUPP))
5549 if (!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL))
5551 if (sb->s_flags & MS_RDONLY)
5553 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, ~EXT3_FEATURE_RO_COMPAT_SUPP))
5558 static inline void register_as_ext3(void) { }
5559 static inline void unregister_as_ext3(void) { }
5560 static inline int ext3_feature_set_ok(struct super_block *sb) { return 0; }
5563 static struct file_system_type ext4_fs_type = {
5564 .owner = THIS_MODULE,
5566 .mount = ext4_mount,
5567 .kill_sb = kill_block_super,
5568 .fs_flags = FS_REQUIRES_DEV,
5570 MODULE_ALIAS_FS("ext4");
5572 static int __init ext4_init_feat_adverts(void)
5574 struct ext4_features *ef;
5577 ef = kzalloc(sizeof(struct ext4_features), GFP_KERNEL);
5581 ef->f_kobj.kset = ext4_kset;
5582 init_completion(&ef->f_kobj_unregister);
5583 ret = kobject_init_and_add(&ef->f_kobj, &ext4_feat_ktype, NULL,
5596 static void ext4_exit_feat_adverts(void)
5598 kobject_put(&ext4_feat->f_kobj);
5599 wait_for_completion(&ext4_feat->f_kobj_unregister);
5603 /* Shared across all ext4 file systems */
5604 wait_queue_head_t ext4__ioend_wq[EXT4_WQ_HASH_SZ];
5605 struct mutex ext4__aio_mutex[EXT4_WQ_HASH_SZ];
5607 static int __init ext4_init_fs(void)
5611 ext4_li_info = NULL;
5612 mutex_init(&ext4_li_mtx);
5614 /* Build-time check for flags consistency */
5615 ext4_check_flag_values();
5617 for (i = 0; i < EXT4_WQ_HASH_SZ; i++) {
5618 mutex_init(&ext4__aio_mutex[i]);
5619 init_waitqueue_head(&ext4__ioend_wq[i]);
5622 err = ext4_init_es();
5626 err = ext4_init_pageio();
5630 err = ext4_init_system_zone();
5633 ext4_kset = kset_create_and_add("ext4", NULL, fs_kobj);
5638 ext4_proc_root = proc_mkdir("fs/ext4", NULL);
5640 err = ext4_init_feat_adverts();
5644 err = ext4_init_mballoc();
5648 ext4_mballoc_ready = 1;
5649 err = init_inodecache();
5654 err = register_filesystem(&ext4_fs_type);
5660 unregister_as_ext2();
5661 unregister_as_ext3();
5662 destroy_inodecache();
5664 ext4_mballoc_ready = 0;
5665 ext4_exit_mballoc();
5667 ext4_exit_feat_adverts();
5670 remove_proc_entry("fs/ext4", NULL);
5671 kset_unregister(ext4_kset);
5673 ext4_exit_system_zone();
5682 static void __exit ext4_exit_fs(void)
5685 ext4_destroy_lazyinit_thread();
5686 unregister_as_ext2();
5687 unregister_as_ext3();
5688 unregister_filesystem(&ext4_fs_type);
5689 destroy_inodecache();
5690 ext4_exit_mballoc();
5691 ext4_exit_feat_adverts();
5692 remove_proc_entry("fs/ext4", NULL);
5693 kset_unregister(ext4_kset);
5694 ext4_exit_system_zone();
5699 MODULE_AUTHOR("Remy Card, Stephen Tweedie, Andrew Morton, Andreas Dilger, Theodore Ts'o and others");
5700 MODULE_DESCRIPTION("Fourth Extended Filesystem");
5701 MODULE_LICENSE("GPL");
5702 module_init(ext4_init_fs)
5703 module_exit(ext4_exit_fs)