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/backing-dev.h>
28 #include <linux/parser.h>
29 #include <linux/buffer_head.h>
30 #include <linux/exportfs.h>
31 #include <linux/vfs.h>
32 #include <linux/random.h>
33 #include <linux/mount.h>
34 #include <linux/namei.h>
35 #include <linux/quotaops.h>
36 #include <linux/seq_file.h>
37 #include <linux/proc_fs.h>
38 #include <linux/ctype.h>
39 #include <linux/log2.h>
40 #include <linux/crc16.h>
41 #include <linux/cleancache.h>
42 #include <asm/uaccess.h>
44 #include <linux/kthread.h>
45 #include <linux/freezer.h>
48 #include "ext4_extents.h" /* Needed for trace points definition */
49 #include "ext4_jbd2.h"
54 #define CREATE_TRACE_POINTS
55 #include <trace/events/ext4.h>
57 static struct proc_dir_entry *ext4_proc_root;
58 static struct kset *ext4_kset;
59 static struct ext4_lazy_init *ext4_li_info;
60 static struct mutex ext4_li_mtx;
61 static struct ext4_features *ext4_feat;
62 static int ext4_mballoc_ready;
64 static int ext4_load_journal(struct super_block *, struct ext4_super_block *,
65 unsigned long journal_devnum);
66 static int ext4_show_options(struct seq_file *seq, struct dentry *root);
67 static int ext4_commit_super(struct super_block *sb, int sync);
68 static void ext4_mark_recovery_complete(struct super_block *sb,
69 struct ext4_super_block *es);
70 static void ext4_clear_journal_err(struct super_block *sb,
71 struct ext4_super_block *es);
72 static int ext4_sync_fs(struct super_block *sb, int wait);
73 static int ext4_remount(struct super_block *sb, int *flags, char *data);
74 static int ext4_statfs(struct dentry *dentry, struct kstatfs *buf);
75 static int ext4_unfreeze(struct super_block *sb);
76 static int ext4_freeze(struct super_block *sb);
77 static struct dentry *ext4_mount(struct file_system_type *fs_type, int flags,
78 const char *dev_name, void *data);
79 static inline int ext2_feature_set_ok(struct super_block *sb);
80 static inline int ext3_feature_set_ok(struct super_block *sb);
81 static int ext4_feature_set_ok(struct super_block *sb, int readonly);
82 static void ext4_destroy_lazyinit_thread(void);
83 static void ext4_unregister_li_request(struct super_block *sb);
84 static void ext4_clear_request_list(void);
85 static int ext4_reserve_clusters(struct ext4_sb_info *, ext4_fsblk_t);
87 #if !defined(CONFIG_EXT2_FS) && !defined(CONFIG_EXT2_FS_MODULE) && defined(CONFIG_EXT4_USE_FOR_EXT23)
88 static struct file_system_type ext2_fs_type = {
92 .kill_sb = kill_block_super,
93 .fs_flags = FS_REQUIRES_DEV,
95 MODULE_ALIAS_FS("ext2");
97 #define IS_EXT2_SB(sb) ((sb)->s_bdev->bd_holder == &ext2_fs_type)
99 #define IS_EXT2_SB(sb) (0)
103 #if !defined(CONFIG_EXT3_FS) && !defined(CONFIG_EXT3_FS_MODULE) && defined(CONFIG_EXT4_USE_FOR_EXT23)
104 static struct file_system_type ext3_fs_type = {
105 .owner = THIS_MODULE,
108 .kill_sb = kill_block_super,
109 .fs_flags = FS_REQUIRES_DEV,
111 MODULE_ALIAS_FS("ext3");
112 MODULE_ALIAS("ext3");
113 #define IS_EXT3_SB(sb) ((sb)->s_bdev->bd_holder == &ext3_fs_type)
115 #define IS_EXT3_SB(sb) (0)
118 static int ext4_verify_csum_type(struct super_block *sb,
119 struct ext4_super_block *es)
121 if (!EXT4_HAS_RO_COMPAT_FEATURE(sb,
122 EXT4_FEATURE_RO_COMPAT_METADATA_CSUM))
125 return es->s_checksum_type == EXT4_CRC32C_CHKSUM;
128 static __le32 ext4_superblock_csum(struct super_block *sb,
129 struct ext4_super_block *es)
131 struct ext4_sb_info *sbi = EXT4_SB(sb);
132 int offset = offsetof(struct ext4_super_block, s_checksum);
135 csum = ext4_chksum(sbi, ~0, (char *)es, offset);
137 return cpu_to_le32(csum);
140 static int ext4_superblock_csum_verify(struct super_block *sb,
141 struct ext4_super_block *es)
143 if (!ext4_has_metadata_csum(sb))
146 return es->s_checksum == ext4_superblock_csum(sb, es);
149 void ext4_superblock_csum_set(struct super_block *sb)
151 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
153 if (!ext4_has_metadata_csum(sb))
156 es->s_checksum = ext4_superblock_csum(sb, es);
159 void *ext4_kvmalloc(size_t size, gfp_t flags)
163 ret = kmalloc(size, flags | __GFP_NOWARN);
165 ret = __vmalloc(size, flags, PAGE_KERNEL);
169 void *ext4_kvzalloc(size_t size, gfp_t flags)
173 ret = kzalloc(size, flags | __GFP_NOWARN);
175 ret = __vmalloc(size, flags | __GFP_ZERO, PAGE_KERNEL);
179 ext4_fsblk_t ext4_block_bitmap(struct super_block *sb,
180 struct ext4_group_desc *bg)
182 return le32_to_cpu(bg->bg_block_bitmap_lo) |
183 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
184 (ext4_fsblk_t)le32_to_cpu(bg->bg_block_bitmap_hi) << 32 : 0);
187 ext4_fsblk_t ext4_inode_bitmap(struct super_block *sb,
188 struct ext4_group_desc *bg)
190 return le32_to_cpu(bg->bg_inode_bitmap_lo) |
191 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
192 (ext4_fsblk_t)le32_to_cpu(bg->bg_inode_bitmap_hi) << 32 : 0);
195 ext4_fsblk_t ext4_inode_table(struct super_block *sb,
196 struct ext4_group_desc *bg)
198 return le32_to_cpu(bg->bg_inode_table_lo) |
199 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
200 (ext4_fsblk_t)le32_to_cpu(bg->bg_inode_table_hi) << 32 : 0);
203 __u32 ext4_free_group_clusters(struct super_block *sb,
204 struct ext4_group_desc *bg)
206 return le16_to_cpu(bg->bg_free_blocks_count_lo) |
207 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
208 (__u32)le16_to_cpu(bg->bg_free_blocks_count_hi) << 16 : 0);
211 __u32 ext4_free_inodes_count(struct super_block *sb,
212 struct ext4_group_desc *bg)
214 return le16_to_cpu(bg->bg_free_inodes_count_lo) |
215 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
216 (__u32)le16_to_cpu(bg->bg_free_inodes_count_hi) << 16 : 0);
219 __u32 ext4_used_dirs_count(struct super_block *sb,
220 struct ext4_group_desc *bg)
222 return le16_to_cpu(bg->bg_used_dirs_count_lo) |
223 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
224 (__u32)le16_to_cpu(bg->bg_used_dirs_count_hi) << 16 : 0);
227 __u32 ext4_itable_unused_count(struct super_block *sb,
228 struct ext4_group_desc *bg)
230 return le16_to_cpu(bg->bg_itable_unused_lo) |
231 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
232 (__u32)le16_to_cpu(bg->bg_itable_unused_hi) << 16 : 0);
235 void ext4_block_bitmap_set(struct super_block *sb,
236 struct ext4_group_desc *bg, ext4_fsblk_t blk)
238 bg->bg_block_bitmap_lo = cpu_to_le32((u32)blk);
239 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
240 bg->bg_block_bitmap_hi = cpu_to_le32(blk >> 32);
243 void ext4_inode_bitmap_set(struct super_block *sb,
244 struct ext4_group_desc *bg, ext4_fsblk_t blk)
246 bg->bg_inode_bitmap_lo = cpu_to_le32((u32)blk);
247 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
248 bg->bg_inode_bitmap_hi = cpu_to_le32(blk >> 32);
251 void ext4_inode_table_set(struct super_block *sb,
252 struct ext4_group_desc *bg, ext4_fsblk_t blk)
254 bg->bg_inode_table_lo = cpu_to_le32((u32)blk);
255 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
256 bg->bg_inode_table_hi = cpu_to_le32(blk >> 32);
259 void ext4_free_group_clusters_set(struct super_block *sb,
260 struct ext4_group_desc *bg, __u32 count)
262 bg->bg_free_blocks_count_lo = cpu_to_le16((__u16)count);
263 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
264 bg->bg_free_blocks_count_hi = cpu_to_le16(count >> 16);
267 void ext4_free_inodes_set(struct super_block *sb,
268 struct ext4_group_desc *bg, __u32 count)
270 bg->bg_free_inodes_count_lo = cpu_to_le16((__u16)count);
271 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
272 bg->bg_free_inodes_count_hi = cpu_to_le16(count >> 16);
275 void ext4_used_dirs_set(struct super_block *sb,
276 struct ext4_group_desc *bg, __u32 count)
278 bg->bg_used_dirs_count_lo = cpu_to_le16((__u16)count);
279 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
280 bg->bg_used_dirs_count_hi = cpu_to_le16(count >> 16);
283 void ext4_itable_unused_set(struct super_block *sb,
284 struct ext4_group_desc *bg, __u32 count)
286 bg->bg_itable_unused_lo = cpu_to_le16((__u16)count);
287 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
288 bg->bg_itable_unused_hi = cpu_to_le16(count >> 16);
292 static void __save_error_info(struct super_block *sb, const char *func,
295 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
297 EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS;
298 if (bdev_read_only(sb->s_bdev))
300 es->s_state |= cpu_to_le16(EXT4_ERROR_FS);
301 es->s_last_error_time = cpu_to_le32(get_seconds());
302 strncpy(es->s_last_error_func, func, sizeof(es->s_last_error_func));
303 es->s_last_error_line = cpu_to_le32(line);
304 if (!es->s_first_error_time) {
305 es->s_first_error_time = es->s_last_error_time;
306 strncpy(es->s_first_error_func, func,
307 sizeof(es->s_first_error_func));
308 es->s_first_error_line = cpu_to_le32(line);
309 es->s_first_error_ino = es->s_last_error_ino;
310 es->s_first_error_block = es->s_last_error_block;
313 * Start the daily error reporting function if it hasn't been
316 if (!es->s_error_count)
317 mod_timer(&EXT4_SB(sb)->s_err_report, jiffies + 24*60*60*HZ);
318 le32_add_cpu(&es->s_error_count, 1);
321 static void save_error_info(struct super_block *sb, const char *func,
324 __save_error_info(sb, func, line);
325 ext4_commit_super(sb, 1);
328 static void ext4_journal_commit_callback(journal_t *journal, transaction_t *txn)
330 struct super_block *sb = journal->j_private;
331 struct ext4_sb_info *sbi = EXT4_SB(sb);
332 int error = is_journal_aborted(journal);
333 struct ext4_journal_cb_entry *jce;
335 BUG_ON(txn->t_state == T_FINISHED);
336 spin_lock(&sbi->s_md_lock);
337 while (!list_empty(&txn->t_private_list)) {
338 jce = list_entry(txn->t_private_list.next,
339 struct ext4_journal_cb_entry, jce_list);
340 list_del_init(&jce->jce_list);
341 spin_unlock(&sbi->s_md_lock);
342 jce->jce_func(sb, jce, error);
343 spin_lock(&sbi->s_md_lock);
345 spin_unlock(&sbi->s_md_lock);
348 /* Deal with the reporting of failure conditions on a filesystem such as
349 * inconsistencies detected or read IO failures.
351 * On ext2, we can store the error state of the filesystem in the
352 * superblock. That is not possible on ext4, because we may have other
353 * write ordering constraints on the superblock which prevent us from
354 * writing it out straight away; and given that the journal is about to
355 * be aborted, we can't rely on the current, or future, transactions to
356 * write out the superblock safely.
358 * We'll just use the jbd2_journal_abort() error code to record an error in
359 * the journal instead. On recovery, the journal will complain about
360 * that error until we've noted it down and cleared it.
363 static void ext4_handle_error(struct super_block *sb)
365 if (sb->s_flags & MS_RDONLY)
368 if (!test_opt(sb, ERRORS_CONT)) {
369 journal_t *journal = EXT4_SB(sb)->s_journal;
371 EXT4_SB(sb)->s_mount_flags |= EXT4_MF_FS_ABORTED;
373 jbd2_journal_abort(journal, -EIO);
375 if (test_opt(sb, ERRORS_RO)) {
376 ext4_msg(sb, KERN_CRIT, "Remounting filesystem read-only");
378 * Make sure updated value of ->s_mount_flags will be visible
379 * before ->s_flags update
382 sb->s_flags |= MS_RDONLY;
384 if (test_opt(sb, ERRORS_PANIC))
385 panic("EXT4-fs (device %s): panic forced after error\n",
389 #define ext4_error_ratelimit(sb) \
390 ___ratelimit(&(EXT4_SB(sb)->s_err_ratelimit_state), \
393 void __ext4_error(struct super_block *sb, const char *function,
394 unsigned int line, const char *fmt, ...)
396 struct va_format vaf;
399 if (ext4_error_ratelimit(sb)) {
404 "EXT4-fs error (device %s): %s:%d: comm %s: %pV\n",
405 sb->s_id, function, line, current->comm, &vaf);
408 save_error_info(sb, function, line);
409 ext4_handle_error(sb);
412 void __ext4_error_inode(struct inode *inode, const char *function,
413 unsigned int line, ext4_fsblk_t block,
414 const char *fmt, ...)
417 struct va_format vaf;
418 struct ext4_super_block *es = EXT4_SB(inode->i_sb)->s_es;
420 es->s_last_error_ino = cpu_to_le32(inode->i_ino);
421 es->s_last_error_block = cpu_to_le64(block);
422 if (ext4_error_ratelimit(inode->i_sb)) {
427 printk(KERN_CRIT "EXT4-fs error (device %s): %s:%d: "
428 "inode #%lu: block %llu: comm %s: %pV\n",
429 inode->i_sb->s_id, function, line, inode->i_ino,
430 block, current->comm, &vaf);
432 printk(KERN_CRIT "EXT4-fs error (device %s): %s:%d: "
433 "inode #%lu: comm %s: %pV\n",
434 inode->i_sb->s_id, function, line, inode->i_ino,
435 current->comm, &vaf);
438 save_error_info(inode->i_sb, function, line);
439 ext4_handle_error(inode->i_sb);
442 void __ext4_error_file(struct file *file, const char *function,
443 unsigned int line, ext4_fsblk_t block,
444 const char *fmt, ...)
447 struct va_format vaf;
448 struct ext4_super_block *es;
449 struct inode *inode = file_inode(file);
450 char pathname[80], *path;
452 es = EXT4_SB(inode->i_sb)->s_es;
453 es->s_last_error_ino = cpu_to_le32(inode->i_ino);
454 if (ext4_error_ratelimit(inode->i_sb)) {
455 path = file_path(file, pathname, sizeof(pathname));
463 "EXT4-fs error (device %s): %s:%d: inode #%lu: "
464 "block %llu: comm %s: path %s: %pV\n",
465 inode->i_sb->s_id, function, line, inode->i_ino,
466 block, current->comm, path, &vaf);
469 "EXT4-fs error (device %s): %s:%d: inode #%lu: "
470 "comm %s: path %s: %pV\n",
471 inode->i_sb->s_id, function, line, inode->i_ino,
472 current->comm, path, &vaf);
475 save_error_info(inode->i_sb, function, line);
476 ext4_handle_error(inode->i_sb);
479 const char *ext4_decode_error(struct super_block *sb, int errno,
486 errstr = "IO failure";
489 errstr = "Out of memory";
492 if (!sb || (EXT4_SB(sb)->s_journal &&
493 EXT4_SB(sb)->s_journal->j_flags & JBD2_ABORT))
494 errstr = "Journal has aborted";
496 errstr = "Readonly filesystem";
499 /* If the caller passed in an extra buffer for unknown
500 * errors, textualise them now. Else we just return
503 /* Check for truncated error codes... */
504 if (snprintf(nbuf, 16, "error %d", -errno) >= 0)
513 /* __ext4_std_error decodes expected errors from journaling functions
514 * automatically and invokes the appropriate error response. */
516 void __ext4_std_error(struct super_block *sb, const char *function,
517 unsigned int line, int errno)
522 /* Special case: if the error is EROFS, and we're not already
523 * inside a transaction, then there's really no point in logging
525 if (errno == -EROFS && journal_current_handle() == NULL &&
526 (sb->s_flags & MS_RDONLY))
529 if (ext4_error_ratelimit(sb)) {
530 errstr = ext4_decode_error(sb, errno, nbuf);
531 printk(KERN_CRIT "EXT4-fs error (device %s) in %s:%d: %s\n",
532 sb->s_id, function, line, errstr);
535 save_error_info(sb, function, line);
536 ext4_handle_error(sb);
540 * ext4_abort is a much stronger failure handler than ext4_error. The
541 * abort function may be used to deal with unrecoverable failures such
542 * as journal IO errors or ENOMEM at a critical moment in log management.
544 * We unconditionally force the filesystem into an ABORT|READONLY state,
545 * unless the error response on the fs has been set to panic in which
546 * case we take the easy way out and panic immediately.
549 void __ext4_abort(struct super_block *sb, const char *function,
550 unsigned int line, const char *fmt, ...)
554 save_error_info(sb, function, line);
556 printk(KERN_CRIT "EXT4-fs error (device %s): %s:%d: ", sb->s_id,
562 if ((sb->s_flags & MS_RDONLY) == 0) {
563 ext4_msg(sb, KERN_CRIT, "Remounting filesystem read-only");
564 EXT4_SB(sb)->s_mount_flags |= EXT4_MF_FS_ABORTED;
566 * Make sure updated value of ->s_mount_flags will be visible
567 * before ->s_flags update
570 sb->s_flags |= MS_RDONLY;
571 if (EXT4_SB(sb)->s_journal)
572 jbd2_journal_abort(EXT4_SB(sb)->s_journal, -EIO);
573 save_error_info(sb, function, line);
575 if (test_opt(sb, ERRORS_PANIC))
576 panic("EXT4-fs panic from previous error\n");
579 void __ext4_msg(struct super_block *sb,
580 const char *prefix, const char *fmt, ...)
582 struct va_format vaf;
585 if (!___ratelimit(&(EXT4_SB(sb)->s_msg_ratelimit_state), "EXT4-fs"))
591 printk("%sEXT4-fs (%s): %pV\n", prefix, sb->s_id, &vaf);
595 #define ext4_warning_ratelimit(sb) \
596 ___ratelimit(&(EXT4_SB(sb)->s_warning_ratelimit_state), \
599 void __ext4_warning(struct super_block *sb, const char *function,
600 unsigned int line, const char *fmt, ...)
602 struct va_format vaf;
605 if (!ext4_warning_ratelimit(sb))
611 printk(KERN_WARNING "EXT4-fs warning (device %s): %s:%d: %pV\n",
612 sb->s_id, function, line, &vaf);
616 void __ext4_warning_inode(const struct inode *inode, const char *function,
617 unsigned int line, const char *fmt, ...)
619 struct va_format vaf;
622 if (!ext4_warning_ratelimit(inode->i_sb))
628 printk(KERN_WARNING "EXT4-fs warning (device %s): %s:%d: "
629 "inode #%lu: comm %s: %pV\n", inode->i_sb->s_id,
630 function, line, inode->i_ino, current->comm, &vaf);
634 void __ext4_grp_locked_error(const char *function, unsigned int line,
635 struct super_block *sb, ext4_group_t grp,
636 unsigned long ino, ext4_fsblk_t block,
637 const char *fmt, ...)
641 struct va_format vaf;
643 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
645 es->s_last_error_ino = cpu_to_le32(ino);
646 es->s_last_error_block = cpu_to_le64(block);
647 __save_error_info(sb, function, line);
649 if (ext4_error_ratelimit(sb)) {
653 printk(KERN_CRIT "EXT4-fs error (device %s): %s:%d: group %u, ",
654 sb->s_id, function, line, grp);
656 printk(KERN_CONT "inode %lu: ", ino);
658 printk(KERN_CONT "block %llu:",
659 (unsigned long long) block);
660 printk(KERN_CONT "%pV\n", &vaf);
664 if (test_opt(sb, ERRORS_CONT)) {
665 ext4_commit_super(sb, 0);
669 ext4_unlock_group(sb, grp);
670 ext4_handle_error(sb);
672 * We only get here in the ERRORS_RO case; relocking the group
673 * may be dangerous, but nothing bad will happen since the
674 * filesystem will have already been marked read/only and the
675 * journal has been aborted. We return 1 as a hint to callers
676 * who might what to use the return value from
677 * ext4_grp_locked_error() to distinguish between the
678 * ERRORS_CONT and ERRORS_RO case, and perhaps return more
679 * aggressively from the ext4 function in question, with a
680 * more appropriate error code.
682 ext4_lock_group(sb, grp);
686 void ext4_update_dynamic_rev(struct super_block *sb)
688 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
690 if (le32_to_cpu(es->s_rev_level) > EXT4_GOOD_OLD_REV)
694 "updating to rev %d because of new feature flag, "
695 "running e2fsck is recommended",
698 es->s_first_ino = cpu_to_le32(EXT4_GOOD_OLD_FIRST_INO);
699 es->s_inode_size = cpu_to_le16(EXT4_GOOD_OLD_INODE_SIZE);
700 es->s_rev_level = cpu_to_le32(EXT4_DYNAMIC_REV);
701 /* leave es->s_feature_*compat flags alone */
702 /* es->s_uuid will be set by e2fsck if empty */
705 * The rest of the superblock fields should be zero, and if not it
706 * means they are likely already in use, so leave them alone. We
707 * can leave it up to e2fsck to clean up any inconsistencies there.
712 * Open the external journal device
714 static struct block_device *ext4_blkdev_get(dev_t dev, struct super_block *sb)
716 struct block_device *bdev;
717 char b[BDEVNAME_SIZE];
719 bdev = blkdev_get_by_dev(dev, FMODE_READ|FMODE_WRITE|FMODE_EXCL, sb);
725 ext4_msg(sb, KERN_ERR, "failed to open journal device %s: %ld",
726 __bdevname(dev, b), PTR_ERR(bdev));
731 * Release the journal device
733 static void ext4_blkdev_put(struct block_device *bdev)
735 blkdev_put(bdev, FMODE_READ|FMODE_WRITE|FMODE_EXCL);
738 static void ext4_blkdev_remove(struct ext4_sb_info *sbi)
740 struct block_device *bdev;
741 bdev = sbi->journal_bdev;
743 ext4_blkdev_put(bdev);
744 sbi->journal_bdev = NULL;
748 static inline struct inode *orphan_list_entry(struct list_head *l)
750 return &list_entry(l, struct ext4_inode_info, i_orphan)->vfs_inode;
753 static void dump_orphan_list(struct super_block *sb, struct ext4_sb_info *sbi)
757 ext4_msg(sb, KERN_ERR, "sb orphan head is %d",
758 le32_to_cpu(sbi->s_es->s_last_orphan));
760 printk(KERN_ERR "sb_info orphan list:\n");
761 list_for_each(l, &sbi->s_orphan) {
762 struct inode *inode = orphan_list_entry(l);
764 "inode %s:%lu at %p: mode %o, nlink %d, next %d\n",
765 inode->i_sb->s_id, inode->i_ino, inode,
766 inode->i_mode, inode->i_nlink,
771 static void ext4_put_super(struct super_block *sb)
773 struct ext4_sb_info *sbi = EXT4_SB(sb);
774 struct ext4_super_block *es = sbi->s_es;
777 ext4_unregister_li_request(sb);
778 dquot_disable(sb, -1, DQUOT_USAGE_ENABLED | DQUOT_LIMITS_ENABLED);
780 flush_workqueue(sbi->rsv_conversion_wq);
781 destroy_workqueue(sbi->rsv_conversion_wq);
783 if (sbi->s_journal) {
784 err = jbd2_journal_destroy(sbi->s_journal);
785 sbi->s_journal = NULL;
787 ext4_abort(sb, "Couldn't clean up the journal");
790 ext4_es_unregister_shrinker(sbi);
791 del_timer_sync(&sbi->s_err_report);
792 ext4_release_system_zone(sb);
794 ext4_ext_release(sb);
795 ext4_xattr_put_super(sb);
797 if (!(sb->s_flags & MS_RDONLY)) {
798 EXT4_CLEAR_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
799 es->s_state = cpu_to_le16(sbi->s_mount_state);
801 if (!(sb->s_flags & MS_RDONLY))
802 ext4_commit_super(sb, 1);
805 remove_proc_entry("options", sbi->s_proc);
806 remove_proc_entry(sb->s_id, ext4_proc_root);
808 kobject_del(&sbi->s_kobj);
810 for (i = 0; i < sbi->s_gdb_count; i++)
811 brelse(sbi->s_group_desc[i]);
812 kvfree(sbi->s_group_desc);
813 kvfree(sbi->s_flex_groups);
814 percpu_counter_destroy(&sbi->s_freeclusters_counter);
815 percpu_counter_destroy(&sbi->s_freeinodes_counter);
816 percpu_counter_destroy(&sbi->s_dirs_counter);
817 percpu_counter_destroy(&sbi->s_dirtyclusters_counter);
820 for (i = 0; i < EXT4_MAXQUOTAS; i++)
821 kfree(sbi->s_qf_names[i]);
824 /* Debugging code just in case the in-memory inode orphan list
825 * isn't empty. The on-disk one can be non-empty if we've
826 * detected an error and taken the fs readonly, but the
827 * in-memory list had better be clean by this point. */
828 if (!list_empty(&sbi->s_orphan))
829 dump_orphan_list(sb, sbi);
830 J_ASSERT(list_empty(&sbi->s_orphan));
832 sync_blockdev(sb->s_bdev);
833 invalidate_bdev(sb->s_bdev);
834 if (sbi->journal_bdev && sbi->journal_bdev != sb->s_bdev) {
836 * Invalidate the journal device's buffers. We don't want them
837 * floating about in memory - the physical journal device may
838 * hotswapped, and it breaks the `ro-after' testing code.
840 sync_blockdev(sbi->journal_bdev);
841 invalidate_bdev(sbi->journal_bdev);
842 ext4_blkdev_remove(sbi);
844 if (sbi->s_mb_cache) {
845 ext4_xattr_destroy_cache(sbi->s_mb_cache);
846 sbi->s_mb_cache = NULL;
849 kthread_stop(sbi->s_mmp_tsk);
850 sb->s_fs_info = NULL;
852 * Now that we are completely done shutting down the
853 * superblock, we need to actually destroy the kobject.
855 kobject_put(&sbi->s_kobj);
856 wait_for_completion(&sbi->s_kobj_unregister);
857 if (sbi->s_chksum_driver)
858 crypto_free_shash(sbi->s_chksum_driver);
859 kfree(sbi->s_blockgroup_lock);
863 static struct kmem_cache *ext4_inode_cachep;
866 * Called inside transaction, so use GFP_NOFS
868 static struct inode *ext4_alloc_inode(struct super_block *sb)
870 struct ext4_inode_info *ei;
872 ei = kmem_cache_alloc(ext4_inode_cachep, GFP_NOFS);
876 ei->vfs_inode.i_version = 1;
877 spin_lock_init(&ei->i_raw_lock);
878 INIT_LIST_HEAD(&ei->i_prealloc_list);
879 spin_lock_init(&ei->i_prealloc_lock);
880 ext4_es_init_tree(&ei->i_es_tree);
881 rwlock_init(&ei->i_es_lock);
882 INIT_LIST_HEAD(&ei->i_es_list);
885 ei->i_es_shrink_lblk = 0;
886 ei->i_reserved_data_blocks = 0;
887 ei->i_reserved_meta_blocks = 0;
888 ei->i_allocated_meta_blocks = 0;
889 ei->i_da_metadata_calc_len = 0;
890 ei->i_da_metadata_calc_last_lblock = 0;
891 spin_lock_init(&(ei->i_block_reservation_lock));
893 ei->i_reserved_quota = 0;
894 memset(&ei->i_dquot, 0, sizeof(ei->i_dquot));
897 INIT_LIST_HEAD(&ei->i_rsv_conversion_list);
898 spin_lock_init(&ei->i_completed_io_lock);
900 ei->i_datasync_tid = 0;
901 atomic_set(&ei->i_ioend_count, 0);
902 atomic_set(&ei->i_unwritten, 0);
903 INIT_WORK(&ei->i_rsv_conversion_work, ext4_end_io_rsv_work);
904 #ifdef CONFIG_EXT4_FS_ENCRYPTION
905 ei->i_crypt_info = NULL;
907 return &ei->vfs_inode;
910 static int ext4_drop_inode(struct inode *inode)
912 int drop = generic_drop_inode(inode);
914 trace_ext4_drop_inode(inode, drop);
918 static void ext4_i_callback(struct rcu_head *head)
920 struct inode *inode = container_of(head, struct inode, i_rcu);
921 kmem_cache_free(ext4_inode_cachep, EXT4_I(inode));
924 static void ext4_destroy_inode(struct inode *inode)
926 if (!list_empty(&(EXT4_I(inode)->i_orphan))) {
927 ext4_msg(inode->i_sb, KERN_ERR,
928 "Inode %lu (%p): orphan list check failed!",
929 inode->i_ino, EXT4_I(inode));
930 print_hex_dump(KERN_INFO, "", DUMP_PREFIX_ADDRESS, 16, 4,
931 EXT4_I(inode), sizeof(struct ext4_inode_info),
935 call_rcu(&inode->i_rcu, ext4_i_callback);
938 static void init_once(void *foo)
940 struct ext4_inode_info *ei = (struct ext4_inode_info *) foo;
942 INIT_LIST_HEAD(&ei->i_orphan);
943 init_rwsem(&ei->xattr_sem);
944 init_rwsem(&ei->i_data_sem);
945 inode_init_once(&ei->vfs_inode);
948 static int __init init_inodecache(void)
950 ext4_inode_cachep = kmem_cache_create("ext4_inode_cache",
951 sizeof(struct ext4_inode_info),
952 0, (SLAB_RECLAIM_ACCOUNT|
955 if (ext4_inode_cachep == NULL)
960 static void destroy_inodecache(void)
963 * Make sure all delayed rcu free inodes are flushed before we
967 kmem_cache_destroy(ext4_inode_cachep);
970 void ext4_clear_inode(struct inode *inode)
972 invalidate_inode_buffers(inode);
975 ext4_discard_preallocations(inode);
976 ext4_es_remove_extent(inode, 0, EXT_MAX_BLOCKS);
977 if (EXT4_I(inode)->jinode) {
978 jbd2_journal_release_jbd_inode(EXT4_JOURNAL(inode),
979 EXT4_I(inode)->jinode);
980 jbd2_free_inode(EXT4_I(inode)->jinode);
981 EXT4_I(inode)->jinode = NULL;
983 #ifdef CONFIG_EXT4_FS_ENCRYPTION
984 if (EXT4_I(inode)->i_crypt_info)
985 ext4_free_encryption_info(inode, EXT4_I(inode)->i_crypt_info);
989 static struct inode *ext4_nfs_get_inode(struct super_block *sb,
990 u64 ino, u32 generation)
994 if (ino < EXT4_FIRST_INO(sb) && ino != EXT4_ROOT_INO)
995 return ERR_PTR(-ESTALE);
996 if (ino > le32_to_cpu(EXT4_SB(sb)->s_es->s_inodes_count))
997 return ERR_PTR(-ESTALE);
999 /* iget isn't really right if the inode is currently unallocated!!
1001 * ext4_read_inode will return a bad_inode if the inode had been
1002 * deleted, so we should be safe.
1004 * Currently we don't know the generation for parent directory, so
1005 * a generation of 0 means "accept any"
1007 inode = ext4_iget_normal(sb, ino);
1009 return ERR_CAST(inode);
1010 if (generation && inode->i_generation != generation) {
1012 return ERR_PTR(-ESTALE);
1018 static struct dentry *ext4_fh_to_dentry(struct super_block *sb, struct fid *fid,
1019 int fh_len, int fh_type)
1021 return generic_fh_to_dentry(sb, fid, fh_len, fh_type,
1022 ext4_nfs_get_inode);
1025 static struct dentry *ext4_fh_to_parent(struct super_block *sb, struct fid *fid,
1026 int fh_len, int fh_type)
1028 return generic_fh_to_parent(sb, fid, fh_len, fh_type,
1029 ext4_nfs_get_inode);
1033 * Try to release metadata pages (indirect blocks, directories) which are
1034 * mapped via the block device. Since these pages could have journal heads
1035 * which would prevent try_to_free_buffers() from freeing them, we must use
1036 * jbd2 layer's try_to_free_buffers() function to release them.
1038 static int bdev_try_to_free_page(struct super_block *sb, struct page *page,
1041 journal_t *journal = EXT4_SB(sb)->s_journal;
1043 WARN_ON(PageChecked(page));
1044 if (!page_has_buffers(page))
1047 return jbd2_journal_try_to_free_buffers(journal, page,
1048 wait & ~__GFP_WAIT);
1049 return try_to_free_buffers(page);
1053 #define QTYPE2NAME(t) ((t) == USRQUOTA ? "user" : "group")
1054 #define QTYPE2MOPT(on, t) ((t) == USRQUOTA?((on)##USRJQUOTA):((on)##GRPJQUOTA))
1056 static int ext4_write_dquot(struct dquot *dquot);
1057 static int ext4_acquire_dquot(struct dquot *dquot);
1058 static int ext4_release_dquot(struct dquot *dquot);
1059 static int ext4_mark_dquot_dirty(struct dquot *dquot);
1060 static int ext4_write_info(struct super_block *sb, int type);
1061 static int ext4_quota_on(struct super_block *sb, int type, int format_id,
1063 static int ext4_quota_off(struct super_block *sb, int type);
1064 static int ext4_quota_on_mount(struct super_block *sb, int type);
1065 static ssize_t ext4_quota_read(struct super_block *sb, int type, char *data,
1066 size_t len, loff_t off);
1067 static ssize_t ext4_quota_write(struct super_block *sb, int type,
1068 const char *data, size_t len, loff_t off);
1069 static int ext4_quota_enable(struct super_block *sb, int type, int format_id,
1070 unsigned int flags);
1071 static int ext4_enable_quotas(struct super_block *sb);
1073 static struct dquot **ext4_get_dquots(struct inode *inode)
1075 return EXT4_I(inode)->i_dquot;
1078 static const struct dquot_operations ext4_quota_operations = {
1079 .get_reserved_space = ext4_get_reserved_space,
1080 .write_dquot = ext4_write_dquot,
1081 .acquire_dquot = ext4_acquire_dquot,
1082 .release_dquot = ext4_release_dquot,
1083 .mark_dirty = ext4_mark_dquot_dirty,
1084 .write_info = ext4_write_info,
1085 .alloc_dquot = dquot_alloc,
1086 .destroy_dquot = dquot_destroy,
1089 static const struct quotactl_ops ext4_qctl_operations = {
1090 .quota_on = ext4_quota_on,
1091 .quota_off = ext4_quota_off,
1092 .quota_sync = dquot_quota_sync,
1093 .get_state = dquot_get_state,
1094 .set_info = dquot_set_dqinfo,
1095 .get_dqblk = dquot_get_dqblk,
1096 .set_dqblk = dquot_set_dqblk
1100 static const struct super_operations ext4_sops = {
1101 .alloc_inode = ext4_alloc_inode,
1102 .destroy_inode = ext4_destroy_inode,
1103 .write_inode = ext4_write_inode,
1104 .dirty_inode = ext4_dirty_inode,
1105 .drop_inode = ext4_drop_inode,
1106 .evict_inode = ext4_evict_inode,
1107 .put_super = ext4_put_super,
1108 .sync_fs = ext4_sync_fs,
1109 .freeze_fs = ext4_freeze,
1110 .unfreeze_fs = ext4_unfreeze,
1111 .statfs = ext4_statfs,
1112 .remount_fs = ext4_remount,
1113 .show_options = ext4_show_options,
1115 .quota_read = ext4_quota_read,
1116 .quota_write = ext4_quota_write,
1117 .get_dquots = ext4_get_dquots,
1119 .bdev_try_to_free_page = bdev_try_to_free_page,
1122 static const struct export_operations ext4_export_ops = {
1123 .fh_to_dentry = ext4_fh_to_dentry,
1124 .fh_to_parent = ext4_fh_to_parent,
1125 .get_parent = ext4_get_parent,
1129 Opt_bsd_df, Opt_minix_df, Opt_grpid, Opt_nogrpid,
1130 Opt_resgid, Opt_resuid, Opt_sb, Opt_err_cont, Opt_err_panic, Opt_err_ro,
1131 Opt_nouid32, Opt_debug, Opt_removed,
1132 Opt_user_xattr, Opt_nouser_xattr, Opt_acl, Opt_noacl,
1133 Opt_auto_da_alloc, Opt_noauto_da_alloc, Opt_noload,
1134 Opt_commit, Opt_min_batch_time, Opt_max_batch_time, Opt_journal_dev,
1135 Opt_journal_path, Opt_journal_checksum, Opt_journal_async_commit,
1136 Opt_abort, Opt_data_journal, Opt_data_ordered, Opt_data_writeback,
1137 Opt_data_err_abort, Opt_data_err_ignore, Opt_test_dummy_encryption,
1138 Opt_usrjquota, Opt_grpjquota, Opt_offusrjquota, Opt_offgrpjquota,
1139 Opt_jqfmt_vfsold, Opt_jqfmt_vfsv0, Opt_jqfmt_vfsv1, Opt_quota,
1140 Opt_noquota, Opt_barrier, Opt_nobarrier, Opt_err,
1141 Opt_usrquota, Opt_grpquota, Opt_i_version, Opt_dax,
1142 Opt_stripe, Opt_delalloc, Opt_nodelalloc, Opt_mblk_io_submit,
1143 Opt_lazytime, Opt_nolazytime,
1144 Opt_nomblk_io_submit, Opt_block_validity, Opt_noblock_validity,
1145 Opt_inode_readahead_blks, Opt_journal_ioprio,
1146 Opt_dioread_nolock, Opt_dioread_lock,
1147 Opt_discard, Opt_nodiscard, Opt_init_itable, Opt_noinit_itable,
1148 Opt_max_dir_size_kb, Opt_nojournal_checksum,
1151 static const match_table_t tokens = {
1152 {Opt_bsd_df, "bsddf"},
1153 {Opt_minix_df, "minixdf"},
1154 {Opt_grpid, "grpid"},
1155 {Opt_grpid, "bsdgroups"},
1156 {Opt_nogrpid, "nogrpid"},
1157 {Opt_nogrpid, "sysvgroups"},
1158 {Opt_resgid, "resgid=%u"},
1159 {Opt_resuid, "resuid=%u"},
1161 {Opt_err_cont, "errors=continue"},
1162 {Opt_err_panic, "errors=panic"},
1163 {Opt_err_ro, "errors=remount-ro"},
1164 {Opt_nouid32, "nouid32"},
1165 {Opt_debug, "debug"},
1166 {Opt_removed, "oldalloc"},
1167 {Opt_removed, "orlov"},
1168 {Opt_user_xattr, "user_xattr"},
1169 {Opt_nouser_xattr, "nouser_xattr"},
1171 {Opt_noacl, "noacl"},
1172 {Opt_noload, "norecovery"},
1173 {Opt_noload, "noload"},
1174 {Opt_removed, "nobh"},
1175 {Opt_removed, "bh"},
1176 {Opt_commit, "commit=%u"},
1177 {Opt_min_batch_time, "min_batch_time=%u"},
1178 {Opt_max_batch_time, "max_batch_time=%u"},
1179 {Opt_journal_dev, "journal_dev=%u"},
1180 {Opt_journal_path, "journal_path=%s"},
1181 {Opt_journal_checksum, "journal_checksum"},
1182 {Opt_nojournal_checksum, "nojournal_checksum"},
1183 {Opt_journal_async_commit, "journal_async_commit"},
1184 {Opt_abort, "abort"},
1185 {Opt_data_journal, "data=journal"},
1186 {Opt_data_ordered, "data=ordered"},
1187 {Opt_data_writeback, "data=writeback"},
1188 {Opt_data_err_abort, "data_err=abort"},
1189 {Opt_data_err_ignore, "data_err=ignore"},
1190 {Opt_offusrjquota, "usrjquota="},
1191 {Opt_usrjquota, "usrjquota=%s"},
1192 {Opt_offgrpjquota, "grpjquota="},
1193 {Opt_grpjquota, "grpjquota=%s"},
1194 {Opt_jqfmt_vfsold, "jqfmt=vfsold"},
1195 {Opt_jqfmt_vfsv0, "jqfmt=vfsv0"},
1196 {Opt_jqfmt_vfsv1, "jqfmt=vfsv1"},
1197 {Opt_grpquota, "grpquota"},
1198 {Opt_noquota, "noquota"},
1199 {Opt_quota, "quota"},
1200 {Opt_usrquota, "usrquota"},
1201 {Opt_barrier, "barrier=%u"},
1202 {Opt_barrier, "barrier"},
1203 {Opt_nobarrier, "nobarrier"},
1204 {Opt_i_version, "i_version"},
1206 {Opt_stripe, "stripe=%u"},
1207 {Opt_delalloc, "delalloc"},
1208 {Opt_lazytime, "lazytime"},
1209 {Opt_nolazytime, "nolazytime"},
1210 {Opt_nodelalloc, "nodelalloc"},
1211 {Opt_removed, "mblk_io_submit"},
1212 {Opt_removed, "nomblk_io_submit"},
1213 {Opt_block_validity, "block_validity"},
1214 {Opt_noblock_validity, "noblock_validity"},
1215 {Opt_inode_readahead_blks, "inode_readahead_blks=%u"},
1216 {Opt_journal_ioprio, "journal_ioprio=%u"},
1217 {Opt_auto_da_alloc, "auto_da_alloc=%u"},
1218 {Opt_auto_da_alloc, "auto_da_alloc"},
1219 {Opt_noauto_da_alloc, "noauto_da_alloc"},
1220 {Opt_dioread_nolock, "dioread_nolock"},
1221 {Opt_dioread_lock, "dioread_lock"},
1222 {Opt_discard, "discard"},
1223 {Opt_nodiscard, "nodiscard"},
1224 {Opt_init_itable, "init_itable=%u"},
1225 {Opt_init_itable, "init_itable"},
1226 {Opt_noinit_itable, "noinit_itable"},
1227 {Opt_max_dir_size_kb, "max_dir_size_kb=%u"},
1228 {Opt_test_dummy_encryption, "test_dummy_encryption"},
1229 {Opt_removed, "check=none"}, /* mount option from ext2/3 */
1230 {Opt_removed, "nocheck"}, /* mount option from ext2/3 */
1231 {Opt_removed, "reservation"}, /* mount option from ext2/3 */
1232 {Opt_removed, "noreservation"}, /* mount option from ext2/3 */
1233 {Opt_removed, "journal=%u"}, /* mount option from ext2/3 */
1237 static ext4_fsblk_t get_sb_block(void **data)
1239 ext4_fsblk_t sb_block;
1240 char *options = (char *) *data;
1242 if (!options || strncmp(options, "sb=", 3) != 0)
1243 return 1; /* Default location */
1246 /* TODO: use simple_strtoll with >32bit ext4 */
1247 sb_block = simple_strtoul(options, &options, 0);
1248 if (*options && *options != ',') {
1249 printk(KERN_ERR "EXT4-fs: Invalid sb specification: %s\n",
1253 if (*options == ',')
1255 *data = (void *) options;
1260 #define DEFAULT_JOURNAL_IOPRIO (IOPRIO_PRIO_VALUE(IOPRIO_CLASS_BE, 3))
1261 static char deprecated_msg[] = "Mount option \"%s\" will be removed by %s\n"
1262 "Contact linux-ext4@vger.kernel.org if you think we should keep it.\n";
1265 static int set_qf_name(struct super_block *sb, int qtype, substring_t *args)
1267 struct ext4_sb_info *sbi = EXT4_SB(sb);
1271 if (sb_any_quota_loaded(sb) &&
1272 !sbi->s_qf_names[qtype]) {
1273 ext4_msg(sb, KERN_ERR,
1274 "Cannot change journaled "
1275 "quota options when quota turned on");
1278 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_QUOTA)) {
1279 ext4_msg(sb, KERN_ERR, "Cannot set journaled quota options "
1280 "when QUOTA feature is enabled");
1283 qname = match_strdup(args);
1285 ext4_msg(sb, KERN_ERR,
1286 "Not enough memory for storing quotafile name");
1289 if (sbi->s_qf_names[qtype]) {
1290 if (strcmp(sbi->s_qf_names[qtype], qname) == 0)
1293 ext4_msg(sb, KERN_ERR,
1294 "%s quota file already specified",
1298 if (strchr(qname, '/')) {
1299 ext4_msg(sb, KERN_ERR,
1300 "quotafile must be on filesystem root");
1303 sbi->s_qf_names[qtype] = qname;
1311 static int clear_qf_name(struct super_block *sb, int qtype)
1314 struct ext4_sb_info *sbi = EXT4_SB(sb);
1316 if (sb_any_quota_loaded(sb) &&
1317 sbi->s_qf_names[qtype]) {
1318 ext4_msg(sb, KERN_ERR, "Cannot change journaled quota options"
1319 " when quota turned on");
1322 kfree(sbi->s_qf_names[qtype]);
1323 sbi->s_qf_names[qtype] = NULL;
1328 #define MOPT_SET 0x0001
1329 #define MOPT_CLEAR 0x0002
1330 #define MOPT_NOSUPPORT 0x0004
1331 #define MOPT_EXPLICIT 0x0008
1332 #define MOPT_CLEAR_ERR 0x0010
1333 #define MOPT_GTE0 0x0020
1336 #define MOPT_QFMT 0x0040
1338 #define MOPT_Q MOPT_NOSUPPORT
1339 #define MOPT_QFMT MOPT_NOSUPPORT
1341 #define MOPT_DATAJ 0x0080
1342 #define MOPT_NO_EXT2 0x0100
1343 #define MOPT_NO_EXT3 0x0200
1344 #define MOPT_EXT4_ONLY (MOPT_NO_EXT2 | MOPT_NO_EXT3)
1345 #define MOPT_STRING 0x0400
1347 static const struct mount_opts {
1351 } ext4_mount_opts[] = {
1352 {Opt_minix_df, EXT4_MOUNT_MINIX_DF, MOPT_SET},
1353 {Opt_bsd_df, EXT4_MOUNT_MINIX_DF, MOPT_CLEAR},
1354 {Opt_grpid, EXT4_MOUNT_GRPID, MOPT_SET},
1355 {Opt_nogrpid, EXT4_MOUNT_GRPID, MOPT_CLEAR},
1356 {Opt_block_validity, EXT4_MOUNT_BLOCK_VALIDITY, MOPT_SET},
1357 {Opt_noblock_validity, EXT4_MOUNT_BLOCK_VALIDITY, MOPT_CLEAR},
1358 {Opt_dioread_nolock, EXT4_MOUNT_DIOREAD_NOLOCK,
1359 MOPT_EXT4_ONLY | MOPT_SET},
1360 {Opt_dioread_lock, EXT4_MOUNT_DIOREAD_NOLOCK,
1361 MOPT_EXT4_ONLY | MOPT_CLEAR},
1362 {Opt_discard, EXT4_MOUNT_DISCARD, MOPT_SET},
1363 {Opt_nodiscard, EXT4_MOUNT_DISCARD, MOPT_CLEAR},
1364 {Opt_delalloc, EXT4_MOUNT_DELALLOC,
1365 MOPT_EXT4_ONLY | MOPT_SET | MOPT_EXPLICIT},
1366 {Opt_nodelalloc, EXT4_MOUNT_DELALLOC,
1367 MOPT_EXT4_ONLY | MOPT_CLEAR},
1368 {Opt_nojournal_checksum, EXT4_MOUNT_JOURNAL_CHECKSUM,
1369 MOPT_EXT4_ONLY | MOPT_CLEAR},
1370 {Opt_journal_checksum, EXT4_MOUNT_JOURNAL_CHECKSUM,
1371 MOPT_EXT4_ONLY | MOPT_SET},
1372 {Opt_journal_async_commit, (EXT4_MOUNT_JOURNAL_ASYNC_COMMIT |
1373 EXT4_MOUNT_JOURNAL_CHECKSUM),
1374 MOPT_EXT4_ONLY | MOPT_SET},
1375 {Opt_noload, EXT4_MOUNT_NOLOAD, MOPT_NO_EXT2 | MOPT_SET},
1376 {Opt_err_panic, EXT4_MOUNT_ERRORS_PANIC, MOPT_SET | MOPT_CLEAR_ERR},
1377 {Opt_err_ro, EXT4_MOUNT_ERRORS_RO, MOPT_SET | MOPT_CLEAR_ERR},
1378 {Opt_err_cont, EXT4_MOUNT_ERRORS_CONT, MOPT_SET | MOPT_CLEAR_ERR},
1379 {Opt_data_err_abort, EXT4_MOUNT_DATA_ERR_ABORT,
1380 MOPT_NO_EXT2 | MOPT_SET},
1381 {Opt_data_err_ignore, EXT4_MOUNT_DATA_ERR_ABORT,
1382 MOPT_NO_EXT2 | MOPT_CLEAR},
1383 {Opt_barrier, EXT4_MOUNT_BARRIER, MOPT_SET},
1384 {Opt_nobarrier, EXT4_MOUNT_BARRIER, MOPT_CLEAR},
1385 {Opt_noauto_da_alloc, EXT4_MOUNT_NO_AUTO_DA_ALLOC, MOPT_SET},
1386 {Opt_auto_da_alloc, EXT4_MOUNT_NO_AUTO_DA_ALLOC, MOPT_CLEAR},
1387 {Opt_noinit_itable, EXT4_MOUNT_INIT_INODE_TABLE, MOPT_CLEAR},
1388 {Opt_commit, 0, MOPT_GTE0},
1389 {Opt_max_batch_time, 0, MOPT_GTE0},
1390 {Opt_min_batch_time, 0, MOPT_GTE0},
1391 {Opt_inode_readahead_blks, 0, MOPT_GTE0},
1392 {Opt_init_itable, 0, MOPT_GTE0},
1393 {Opt_dax, EXT4_MOUNT_DAX, MOPT_SET},
1394 {Opt_stripe, 0, MOPT_GTE0},
1395 {Opt_resuid, 0, MOPT_GTE0},
1396 {Opt_resgid, 0, MOPT_GTE0},
1397 {Opt_journal_dev, 0, MOPT_GTE0},
1398 {Opt_journal_path, 0, MOPT_STRING},
1399 {Opt_journal_ioprio, 0, MOPT_GTE0},
1400 {Opt_data_journal, EXT4_MOUNT_JOURNAL_DATA, MOPT_NO_EXT2 | MOPT_DATAJ},
1401 {Opt_data_ordered, EXT4_MOUNT_ORDERED_DATA, MOPT_NO_EXT2 | MOPT_DATAJ},
1402 {Opt_data_writeback, EXT4_MOUNT_WRITEBACK_DATA,
1403 MOPT_NO_EXT2 | MOPT_DATAJ},
1404 {Opt_user_xattr, EXT4_MOUNT_XATTR_USER, MOPT_SET},
1405 {Opt_nouser_xattr, EXT4_MOUNT_XATTR_USER, MOPT_CLEAR},
1406 #ifdef CONFIG_EXT4_FS_POSIX_ACL
1407 {Opt_acl, EXT4_MOUNT_POSIX_ACL, MOPT_SET},
1408 {Opt_noacl, EXT4_MOUNT_POSIX_ACL, MOPT_CLEAR},
1410 {Opt_acl, 0, MOPT_NOSUPPORT},
1411 {Opt_noacl, 0, MOPT_NOSUPPORT},
1413 {Opt_nouid32, EXT4_MOUNT_NO_UID32, MOPT_SET},
1414 {Opt_debug, EXT4_MOUNT_DEBUG, MOPT_SET},
1415 {Opt_quota, EXT4_MOUNT_QUOTA | EXT4_MOUNT_USRQUOTA, MOPT_SET | MOPT_Q},
1416 {Opt_usrquota, EXT4_MOUNT_QUOTA | EXT4_MOUNT_USRQUOTA,
1418 {Opt_grpquota, EXT4_MOUNT_QUOTA | EXT4_MOUNT_GRPQUOTA,
1420 {Opt_noquota, (EXT4_MOUNT_QUOTA | EXT4_MOUNT_USRQUOTA |
1421 EXT4_MOUNT_GRPQUOTA), MOPT_CLEAR | MOPT_Q},
1422 {Opt_usrjquota, 0, MOPT_Q},
1423 {Opt_grpjquota, 0, MOPT_Q},
1424 {Opt_offusrjquota, 0, MOPT_Q},
1425 {Opt_offgrpjquota, 0, MOPT_Q},
1426 {Opt_jqfmt_vfsold, QFMT_VFS_OLD, MOPT_QFMT},
1427 {Opt_jqfmt_vfsv0, QFMT_VFS_V0, MOPT_QFMT},
1428 {Opt_jqfmt_vfsv1, QFMT_VFS_V1, MOPT_QFMT},
1429 {Opt_max_dir_size_kb, 0, MOPT_GTE0},
1430 {Opt_test_dummy_encryption, 0, MOPT_GTE0},
1434 static int handle_mount_opt(struct super_block *sb, char *opt, int token,
1435 substring_t *args, unsigned long *journal_devnum,
1436 unsigned int *journal_ioprio, int is_remount)
1438 struct ext4_sb_info *sbi = EXT4_SB(sb);
1439 const struct mount_opts *m;
1445 if (token == Opt_usrjquota)
1446 return set_qf_name(sb, USRQUOTA, &args[0]);
1447 else if (token == Opt_grpjquota)
1448 return set_qf_name(sb, GRPQUOTA, &args[0]);
1449 else if (token == Opt_offusrjquota)
1450 return clear_qf_name(sb, USRQUOTA);
1451 else if (token == Opt_offgrpjquota)
1452 return clear_qf_name(sb, GRPQUOTA);
1456 case Opt_nouser_xattr:
1457 ext4_msg(sb, KERN_WARNING, deprecated_msg, opt, "3.5");
1460 return 1; /* handled by get_sb_block() */
1462 ext4_msg(sb, KERN_WARNING, "Ignoring removed %s option", opt);
1465 sbi->s_mount_flags |= EXT4_MF_FS_ABORTED;
1468 sb->s_flags |= MS_I_VERSION;
1471 sb->s_flags |= MS_LAZYTIME;
1473 case Opt_nolazytime:
1474 sb->s_flags &= ~MS_LAZYTIME;
1478 for (m = ext4_mount_opts; m->token != Opt_err; m++)
1479 if (token == m->token)
1482 if (m->token == Opt_err) {
1483 ext4_msg(sb, KERN_ERR, "Unrecognized mount option \"%s\" "
1484 "or missing value", opt);
1488 if ((m->flags & MOPT_NO_EXT2) && IS_EXT2_SB(sb)) {
1489 ext4_msg(sb, KERN_ERR,
1490 "Mount option \"%s\" incompatible with ext2", opt);
1493 if ((m->flags & MOPT_NO_EXT3) && IS_EXT3_SB(sb)) {
1494 ext4_msg(sb, KERN_ERR,
1495 "Mount option \"%s\" incompatible with ext3", opt);
1499 if (args->from && !(m->flags & MOPT_STRING) && match_int(args, &arg))
1501 if (args->from && (m->flags & MOPT_GTE0) && (arg < 0))
1503 if (m->flags & MOPT_EXPLICIT)
1504 set_opt2(sb, EXPLICIT_DELALLOC);
1505 if (m->flags & MOPT_CLEAR_ERR)
1506 clear_opt(sb, ERRORS_MASK);
1507 if (token == Opt_noquota && sb_any_quota_loaded(sb)) {
1508 ext4_msg(sb, KERN_ERR, "Cannot change quota "
1509 "options when quota turned on");
1513 if (m->flags & MOPT_NOSUPPORT) {
1514 ext4_msg(sb, KERN_ERR, "%s option not supported", opt);
1515 } else if (token == Opt_commit) {
1517 arg = JBD2_DEFAULT_MAX_COMMIT_AGE;
1518 sbi->s_commit_interval = HZ * arg;
1519 } else if (token == Opt_max_batch_time) {
1520 sbi->s_max_batch_time = arg;
1521 } else if (token == Opt_min_batch_time) {
1522 sbi->s_min_batch_time = arg;
1523 } else if (token == Opt_inode_readahead_blks) {
1524 if (arg && (arg > (1 << 30) || !is_power_of_2(arg))) {
1525 ext4_msg(sb, KERN_ERR,
1526 "EXT4-fs: inode_readahead_blks must be "
1527 "0 or a power of 2 smaller than 2^31");
1530 sbi->s_inode_readahead_blks = arg;
1531 } else if (token == Opt_init_itable) {
1532 set_opt(sb, INIT_INODE_TABLE);
1534 arg = EXT4_DEF_LI_WAIT_MULT;
1535 sbi->s_li_wait_mult = arg;
1536 } else if (token == Opt_max_dir_size_kb) {
1537 sbi->s_max_dir_size_kb = arg;
1538 } else if (token == Opt_stripe) {
1539 sbi->s_stripe = arg;
1540 } else if (token == Opt_resuid) {
1541 uid = make_kuid(current_user_ns(), arg);
1542 if (!uid_valid(uid)) {
1543 ext4_msg(sb, KERN_ERR, "Invalid uid value %d", arg);
1546 sbi->s_resuid = uid;
1547 } else if (token == Opt_resgid) {
1548 gid = make_kgid(current_user_ns(), arg);
1549 if (!gid_valid(gid)) {
1550 ext4_msg(sb, KERN_ERR, "Invalid gid value %d", arg);
1553 sbi->s_resgid = gid;
1554 } else if (token == Opt_journal_dev) {
1556 ext4_msg(sb, KERN_ERR,
1557 "Cannot specify journal on remount");
1560 *journal_devnum = arg;
1561 } else if (token == Opt_journal_path) {
1563 struct inode *journal_inode;
1568 ext4_msg(sb, KERN_ERR,
1569 "Cannot specify journal on remount");
1572 journal_path = match_strdup(&args[0]);
1573 if (!journal_path) {
1574 ext4_msg(sb, KERN_ERR, "error: could not dup "
1575 "journal device string");
1579 error = kern_path(journal_path, LOOKUP_FOLLOW, &path);
1581 ext4_msg(sb, KERN_ERR, "error: could not find "
1582 "journal device path: error %d", error);
1583 kfree(journal_path);
1587 journal_inode = d_inode(path.dentry);
1588 if (!S_ISBLK(journal_inode->i_mode)) {
1589 ext4_msg(sb, KERN_ERR, "error: journal path %s "
1590 "is not a block device", journal_path);
1592 kfree(journal_path);
1596 *journal_devnum = new_encode_dev(journal_inode->i_rdev);
1598 kfree(journal_path);
1599 } else if (token == Opt_journal_ioprio) {
1601 ext4_msg(sb, KERN_ERR, "Invalid journal IO priority"
1606 IOPRIO_PRIO_VALUE(IOPRIO_CLASS_BE, arg);
1607 } else if (token == Opt_test_dummy_encryption) {
1608 #ifdef CONFIG_EXT4_FS_ENCRYPTION
1609 sbi->s_mount_flags |= EXT4_MF_TEST_DUMMY_ENCRYPTION;
1610 ext4_msg(sb, KERN_WARNING,
1611 "Test dummy encryption mode enabled");
1613 ext4_msg(sb, KERN_WARNING,
1614 "Test dummy encryption mount option ignored");
1616 } else if (m->flags & MOPT_DATAJ) {
1618 if (!sbi->s_journal)
1619 ext4_msg(sb, KERN_WARNING, "Remounting file system with no journal so ignoring journalled data option");
1620 else if (test_opt(sb, DATA_FLAGS) != m->mount_opt) {
1621 ext4_msg(sb, KERN_ERR,
1622 "Cannot change data mode on remount");
1626 clear_opt(sb, DATA_FLAGS);
1627 sbi->s_mount_opt |= m->mount_opt;
1630 } else if (m->flags & MOPT_QFMT) {
1631 if (sb_any_quota_loaded(sb) &&
1632 sbi->s_jquota_fmt != m->mount_opt) {
1633 ext4_msg(sb, KERN_ERR, "Cannot change journaled "
1634 "quota options when quota turned on");
1637 if (EXT4_HAS_RO_COMPAT_FEATURE(sb,
1638 EXT4_FEATURE_RO_COMPAT_QUOTA)) {
1639 ext4_msg(sb, KERN_ERR,
1640 "Cannot set journaled quota options "
1641 "when QUOTA feature is enabled");
1644 sbi->s_jquota_fmt = m->mount_opt;
1646 #ifndef CONFIG_FS_DAX
1647 } else if (token == Opt_dax) {
1648 ext4_msg(sb, KERN_INFO, "dax option not supported");
1654 if (m->flags & MOPT_CLEAR)
1656 else if (unlikely(!(m->flags & MOPT_SET))) {
1657 ext4_msg(sb, KERN_WARNING,
1658 "buggy handling of option %s", opt);
1663 sbi->s_mount_opt |= m->mount_opt;
1665 sbi->s_mount_opt &= ~m->mount_opt;
1670 static int parse_options(char *options, struct super_block *sb,
1671 unsigned long *journal_devnum,
1672 unsigned int *journal_ioprio,
1675 struct ext4_sb_info *sbi = EXT4_SB(sb);
1677 substring_t args[MAX_OPT_ARGS];
1683 while ((p = strsep(&options, ",")) != NULL) {
1687 * Initialize args struct so we know whether arg was
1688 * found; some options take optional arguments.
1690 args[0].to = args[0].from = NULL;
1691 token = match_token(p, tokens, args);
1692 if (handle_mount_opt(sb, p, token, args, journal_devnum,
1693 journal_ioprio, is_remount) < 0)
1697 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_QUOTA) &&
1698 (test_opt(sb, USRQUOTA) || test_opt(sb, GRPQUOTA))) {
1699 ext4_msg(sb, KERN_ERR, "Cannot set quota options when QUOTA "
1700 "feature is enabled");
1703 if (sbi->s_qf_names[USRQUOTA] || sbi->s_qf_names[GRPQUOTA]) {
1704 if (test_opt(sb, USRQUOTA) && sbi->s_qf_names[USRQUOTA])
1705 clear_opt(sb, USRQUOTA);
1707 if (test_opt(sb, GRPQUOTA) && sbi->s_qf_names[GRPQUOTA])
1708 clear_opt(sb, GRPQUOTA);
1710 if (test_opt(sb, GRPQUOTA) || test_opt(sb, USRQUOTA)) {
1711 ext4_msg(sb, KERN_ERR, "old and new quota "
1716 if (!sbi->s_jquota_fmt) {
1717 ext4_msg(sb, KERN_ERR, "journaled quota format "
1723 if (test_opt(sb, DIOREAD_NOLOCK)) {
1725 BLOCK_SIZE << le32_to_cpu(sbi->s_es->s_log_block_size);
1727 if (blocksize < PAGE_CACHE_SIZE) {
1728 ext4_msg(sb, KERN_ERR, "can't mount with "
1729 "dioread_nolock if block size != PAGE_SIZE");
1733 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA &&
1734 test_opt(sb, JOURNAL_ASYNC_COMMIT)) {
1735 ext4_msg(sb, KERN_ERR, "can't mount with journal_async_commit "
1736 "in data=ordered mode");
1742 static inline void ext4_show_quota_options(struct seq_file *seq,
1743 struct super_block *sb)
1745 #if defined(CONFIG_QUOTA)
1746 struct ext4_sb_info *sbi = EXT4_SB(sb);
1748 if (sbi->s_jquota_fmt) {
1751 switch (sbi->s_jquota_fmt) {
1762 seq_printf(seq, ",jqfmt=%s", fmtname);
1765 if (sbi->s_qf_names[USRQUOTA])
1766 seq_printf(seq, ",usrjquota=%s", sbi->s_qf_names[USRQUOTA]);
1768 if (sbi->s_qf_names[GRPQUOTA])
1769 seq_printf(seq, ",grpjquota=%s", sbi->s_qf_names[GRPQUOTA]);
1773 static const char *token2str(int token)
1775 const struct match_token *t;
1777 for (t = tokens; t->token != Opt_err; t++)
1778 if (t->token == token && !strchr(t->pattern, '='))
1785 * - it's set to a non-default value OR
1786 * - if the per-sb default is different from the global default
1788 static int _ext4_show_options(struct seq_file *seq, struct super_block *sb,
1791 struct ext4_sb_info *sbi = EXT4_SB(sb);
1792 struct ext4_super_block *es = sbi->s_es;
1793 int def_errors, def_mount_opt = nodefs ? 0 : sbi->s_def_mount_opt;
1794 const struct mount_opts *m;
1795 char sep = nodefs ? '\n' : ',';
1797 #define SEQ_OPTS_PUTS(str) seq_printf(seq, "%c" str, sep)
1798 #define SEQ_OPTS_PRINT(str, arg) seq_printf(seq, "%c" str, sep, arg)
1800 if (sbi->s_sb_block != 1)
1801 SEQ_OPTS_PRINT("sb=%llu", sbi->s_sb_block);
1803 for (m = ext4_mount_opts; m->token != Opt_err; m++) {
1804 int want_set = m->flags & MOPT_SET;
1805 if (((m->flags & (MOPT_SET|MOPT_CLEAR)) == 0) ||
1806 (m->flags & MOPT_CLEAR_ERR))
1808 if (!(m->mount_opt & (sbi->s_mount_opt ^ def_mount_opt)))
1809 continue; /* skip if same as the default */
1811 (sbi->s_mount_opt & m->mount_opt) != m->mount_opt) ||
1812 (!want_set && (sbi->s_mount_opt & m->mount_opt)))
1813 continue; /* select Opt_noFoo vs Opt_Foo */
1814 SEQ_OPTS_PRINT("%s", token2str(m->token));
1817 if (nodefs || !uid_eq(sbi->s_resuid, make_kuid(&init_user_ns, EXT4_DEF_RESUID)) ||
1818 le16_to_cpu(es->s_def_resuid) != EXT4_DEF_RESUID)
1819 SEQ_OPTS_PRINT("resuid=%u",
1820 from_kuid_munged(&init_user_ns, sbi->s_resuid));
1821 if (nodefs || !gid_eq(sbi->s_resgid, make_kgid(&init_user_ns, EXT4_DEF_RESGID)) ||
1822 le16_to_cpu(es->s_def_resgid) != EXT4_DEF_RESGID)
1823 SEQ_OPTS_PRINT("resgid=%u",
1824 from_kgid_munged(&init_user_ns, sbi->s_resgid));
1825 def_errors = nodefs ? -1 : le16_to_cpu(es->s_errors);
1826 if (test_opt(sb, ERRORS_RO) && def_errors != EXT4_ERRORS_RO)
1827 SEQ_OPTS_PUTS("errors=remount-ro");
1828 if (test_opt(sb, ERRORS_CONT) && def_errors != EXT4_ERRORS_CONTINUE)
1829 SEQ_OPTS_PUTS("errors=continue");
1830 if (test_opt(sb, ERRORS_PANIC) && def_errors != EXT4_ERRORS_PANIC)
1831 SEQ_OPTS_PUTS("errors=panic");
1832 if (nodefs || sbi->s_commit_interval != JBD2_DEFAULT_MAX_COMMIT_AGE*HZ)
1833 SEQ_OPTS_PRINT("commit=%lu", sbi->s_commit_interval / HZ);
1834 if (nodefs || sbi->s_min_batch_time != EXT4_DEF_MIN_BATCH_TIME)
1835 SEQ_OPTS_PRINT("min_batch_time=%u", sbi->s_min_batch_time);
1836 if (nodefs || sbi->s_max_batch_time != EXT4_DEF_MAX_BATCH_TIME)
1837 SEQ_OPTS_PRINT("max_batch_time=%u", sbi->s_max_batch_time);
1838 if (sb->s_flags & MS_I_VERSION)
1839 SEQ_OPTS_PUTS("i_version");
1840 if (nodefs || sbi->s_stripe)
1841 SEQ_OPTS_PRINT("stripe=%lu", sbi->s_stripe);
1842 if (EXT4_MOUNT_DATA_FLAGS & (sbi->s_mount_opt ^ def_mount_opt)) {
1843 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA)
1844 SEQ_OPTS_PUTS("data=journal");
1845 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA)
1846 SEQ_OPTS_PUTS("data=ordered");
1847 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_WRITEBACK_DATA)
1848 SEQ_OPTS_PUTS("data=writeback");
1851 sbi->s_inode_readahead_blks != EXT4_DEF_INODE_READAHEAD_BLKS)
1852 SEQ_OPTS_PRINT("inode_readahead_blks=%u",
1853 sbi->s_inode_readahead_blks);
1855 if (nodefs || (test_opt(sb, INIT_INODE_TABLE) &&
1856 (sbi->s_li_wait_mult != EXT4_DEF_LI_WAIT_MULT)))
1857 SEQ_OPTS_PRINT("init_itable=%u", sbi->s_li_wait_mult);
1858 if (nodefs || sbi->s_max_dir_size_kb)
1859 SEQ_OPTS_PRINT("max_dir_size_kb=%u", sbi->s_max_dir_size_kb);
1861 ext4_show_quota_options(seq, sb);
1865 static int ext4_show_options(struct seq_file *seq, struct dentry *root)
1867 return _ext4_show_options(seq, root->d_sb, 0);
1870 static int options_seq_show(struct seq_file *seq, void *offset)
1872 struct super_block *sb = seq->private;
1875 seq_puts(seq, (sb->s_flags & MS_RDONLY) ? "ro" : "rw");
1876 rc = _ext4_show_options(seq, sb, 1);
1877 seq_puts(seq, "\n");
1881 static int options_open_fs(struct inode *inode, struct file *file)
1883 return single_open(file, options_seq_show, PDE_DATA(inode));
1886 static const struct file_operations ext4_seq_options_fops = {
1887 .owner = THIS_MODULE,
1888 .open = options_open_fs,
1890 .llseek = seq_lseek,
1891 .release = single_release,
1894 static int ext4_setup_super(struct super_block *sb, struct ext4_super_block *es,
1897 struct ext4_sb_info *sbi = EXT4_SB(sb);
1900 if (le32_to_cpu(es->s_rev_level) > EXT4_MAX_SUPP_REV) {
1901 ext4_msg(sb, KERN_ERR, "revision level too high, "
1902 "forcing read-only mode");
1907 if (!(sbi->s_mount_state & EXT4_VALID_FS))
1908 ext4_msg(sb, KERN_WARNING, "warning: mounting unchecked fs, "
1909 "running e2fsck is recommended");
1910 else if (sbi->s_mount_state & EXT4_ERROR_FS)
1911 ext4_msg(sb, KERN_WARNING,
1912 "warning: mounting fs with errors, "
1913 "running e2fsck is recommended");
1914 else if ((__s16) le16_to_cpu(es->s_max_mnt_count) > 0 &&
1915 le16_to_cpu(es->s_mnt_count) >=
1916 (unsigned short) (__s16) le16_to_cpu(es->s_max_mnt_count))
1917 ext4_msg(sb, KERN_WARNING,
1918 "warning: maximal mount count reached, "
1919 "running e2fsck is recommended");
1920 else if (le32_to_cpu(es->s_checkinterval) &&
1921 (le32_to_cpu(es->s_lastcheck) +
1922 le32_to_cpu(es->s_checkinterval) <= get_seconds()))
1923 ext4_msg(sb, KERN_WARNING,
1924 "warning: checktime reached, "
1925 "running e2fsck is recommended");
1926 if (!sbi->s_journal)
1927 es->s_state &= cpu_to_le16(~EXT4_VALID_FS);
1928 if (!(__s16) le16_to_cpu(es->s_max_mnt_count))
1929 es->s_max_mnt_count = cpu_to_le16(EXT4_DFL_MAX_MNT_COUNT);
1930 le16_add_cpu(&es->s_mnt_count, 1);
1931 es->s_mtime = cpu_to_le32(get_seconds());
1932 ext4_update_dynamic_rev(sb);
1934 EXT4_SET_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
1936 ext4_commit_super(sb, 1);
1938 if (test_opt(sb, DEBUG))
1939 printk(KERN_INFO "[EXT4 FS bs=%lu, gc=%u, "
1940 "bpg=%lu, ipg=%lu, mo=%04x, mo2=%04x]\n",
1942 sbi->s_groups_count,
1943 EXT4_BLOCKS_PER_GROUP(sb),
1944 EXT4_INODES_PER_GROUP(sb),
1945 sbi->s_mount_opt, sbi->s_mount_opt2);
1947 cleancache_init_fs(sb);
1951 int ext4_alloc_flex_bg_array(struct super_block *sb, ext4_group_t ngroup)
1953 struct ext4_sb_info *sbi = EXT4_SB(sb);
1954 struct flex_groups *new_groups;
1957 if (!sbi->s_log_groups_per_flex)
1960 size = ext4_flex_group(sbi, ngroup - 1) + 1;
1961 if (size <= sbi->s_flex_groups_allocated)
1964 size = roundup_pow_of_two(size * sizeof(struct flex_groups));
1965 new_groups = ext4_kvzalloc(size, GFP_KERNEL);
1967 ext4_msg(sb, KERN_ERR, "not enough memory for %d flex groups",
1968 size / (int) sizeof(struct flex_groups));
1972 if (sbi->s_flex_groups) {
1973 memcpy(new_groups, sbi->s_flex_groups,
1974 (sbi->s_flex_groups_allocated *
1975 sizeof(struct flex_groups)));
1976 kvfree(sbi->s_flex_groups);
1978 sbi->s_flex_groups = new_groups;
1979 sbi->s_flex_groups_allocated = size / sizeof(struct flex_groups);
1983 static int ext4_fill_flex_info(struct super_block *sb)
1985 struct ext4_sb_info *sbi = EXT4_SB(sb);
1986 struct ext4_group_desc *gdp = NULL;
1987 ext4_group_t flex_group;
1990 sbi->s_log_groups_per_flex = sbi->s_es->s_log_groups_per_flex;
1991 if (sbi->s_log_groups_per_flex < 1 || sbi->s_log_groups_per_flex > 31) {
1992 sbi->s_log_groups_per_flex = 0;
1996 err = ext4_alloc_flex_bg_array(sb, sbi->s_groups_count);
2000 for (i = 0; i < sbi->s_groups_count; i++) {
2001 gdp = ext4_get_group_desc(sb, i, NULL);
2003 flex_group = ext4_flex_group(sbi, i);
2004 atomic_add(ext4_free_inodes_count(sb, gdp),
2005 &sbi->s_flex_groups[flex_group].free_inodes);
2006 atomic64_add(ext4_free_group_clusters(sb, gdp),
2007 &sbi->s_flex_groups[flex_group].free_clusters);
2008 atomic_add(ext4_used_dirs_count(sb, gdp),
2009 &sbi->s_flex_groups[flex_group].used_dirs);
2017 static __le16 ext4_group_desc_csum(struct ext4_sb_info *sbi, __u32 block_group,
2018 struct ext4_group_desc *gdp)
2022 __le32 le_group = cpu_to_le32(block_group);
2024 if (ext4_has_metadata_csum(sbi->s_sb)) {
2025 /* Use new metadata_csum algorithm */
2029 save_csum = gdp->bg_checksum;
2030 gdp->bg_checksum = 0;
2031 csum32 = ext4_chksum(sbi, sbi->s_csum_seed, (__u8 *)&le_group,
2033 csum32 = ext4_chksum(sbi, csum32, (__u8 *)gdp,
2035 gdp->bg_checksum = save_csum;
2037 crc = csum32 & 0xFFFF;
2041 /* old crc16 code */
2042 if (!(sbi->s_es->s_feature_ro_compat &
2043 cpu_to_le32(EXT4_FEATURE_RO_COMPAT_GDT_CSUM)))
2046 offset = offsetof(struct ext4_group_desc, bg_checksum);
2048 crc = crc16(~0, sbi->s_es->s_uuid, sizeof(sbi->s_es->s_uuid));
2049 crc = crc16(crc, (__u8 *)&le_group, sizeof(le_group));
2050 crc = crc16(crc, (__u8 *)gdp, offset);
2051 offset += sizeof(gdp->bg_checksum); /* skip checksum */
2052 /* for checksum of struct ext4_group_desc do the rest...*/
2053 if ((sbi->s_es->s_feature_incompat &
2054 cpu_to_le32(EXT4_FEATURE_INCOMPAT_64BIT)) &&
2055 offset < le16_to_cpu(sbi->s_es->s_desc_size))
2056 crc = crc16(crc, (__u8 *)gdp + offset,
2057 le16_to_cpu(sbi->s_es->s_desc_size) -
2061 return cpu_to_le16(crc);
2064 int ext4_group_desc_csum_verify(struct super_block *sb, __u32 block_group,
2065 struct ext4_group_desc *gdp)
2067 if (ext4_has_group_desc_csum(sb) &&
2068 (gdp->bg_checksum != ext4_group_desc_csum(EXT4_SB(sb),
2075 void ext4_group_desc_csum_set(struct super_block *sb, __u32 block_group,
2076 struct ext4_group_desc *gdp)
2078 if (!ext4_has_group_desc_csum(sb))
2080 gdp->bg_checksum = ext4_group_desc_csum(EXT4_SB(sb), block_group, gdp);
2083 /* Called at mount-time, super-block is locked */
2084 static int ext4_check_descriptors(struct super_block *sb,
2085 ext4_group_t *first_not_zeroed)
2087 struct ext4_sb_info *sbi = EXT4_SB(sb);
2088 ext4_fsblk_t first_block = le32_to_cpu(sbi->s_es->s_first_data_block);
2089 ext4_fsblk_t last_block;
2090 ext4_fsblk_t block_bitmap;
2091 ext4_fsblk_t inode_bitmap;
2092 ext4_fsblk_t inode_table;
2093 int flexbg_flag = 0;
2094 ext4_group_t i, grp = sbi->s_groups_count;
2096 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_FLEX_BG))
2099 ext4_debug("Checking group descriptors");
2101 for (i = 0; i < sbi->s_groups_count; i++) {
2102 struct ext4_group_desc *gdp = ext4_get_group_desc(sb, i, NULL);
2104 if (i == sbi->s_groups_count - 1 || flexbg_flag)
2105 last_block = ext4_blocks_count(sbi->s_es) - 1;
2107 last_block = first_block +
2108 (EXT4_BLOCKS_PER_GROUP(sb) - 1);
2110 if ((grp == sbi->s_groups_count) &&
2111 !(gdp->bg_flags & cpu_to_le16(EXT4_BG_INODE_ZEROED)))
2114 block_bitmap = ext4_block_bitmap(sb, gdp);
2115 if (block_bitmap < first_block || block_bitmap > last_block) {
2116 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2117 "Block bitmap for group %u not in group "
2118 "(block %llu)!", i, block_bitmap);
2121 inode_bitmap = ext4_inode_bitmap(sb, gdp);
2122 if (inode_bitmap < first_block || inode_bitmap > last_block) {
2123 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2124 "Inode bitmap for group %u not in group "
2125 "(block %llu)!", i, inode_bitmap);
2128 inode_table = ext4_inode_table(sb, gdp);
2129 if (inode_table < first_block ||
2130 inode_table + sbi->s_itb_per_group - 1 > last_block) {
2131 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2132 "Inode table for group %u not in group "
2133 "(block %llu)!", i, inode_table);
2136 ext4_lock_group(sb, i);
2137 if (!ext4_group_desc_csum_verify(sb, i, gdp)) {
2138 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2139 "Checksum for group %u failed (%u!=%u)",
2140 i, le16_to_cpu(ext4_group_desc_csum(sbi, i,
2141 gdp)), le16_to_cpu(gdp->bg_checksum));
2142 if (!(sb->s_flags & MS_RDONLY)) {
2143 ext4_unlock_group(sb, i);
2147 ext4_unlock_group(sb, i);
2149 first_block += EXT4_BLOCKS_PER_GROUP(sb);
2151 if (NULL != first_not_zeroed)
2152 *first_not_zeroed = grp;
2156 /* ext4_orphan_cleanup() walks a singly-linked list of inodes (starting at
2157 * the superblock) which were deleted from all directories, but held open by
2158 * a process at the time of a crash. We walk the list and try to delete these
2159 * inodes at recovery time (only with a read-write filesystem).
2161 * In order to keep the orphan inode chain consistent during traversal (in
2162 * case of crash during recovery), we link each inode into the superblock
2163 * orphan list_head and handle it the same way as an inode deletion during
2164 * normal operation (which journals the operations for us).
2166 * We only do an iget() and an iput() on each inode, which is very safe if we
2167 * accidentally point at an in-use or already deleted inode. The worst that
2168 * can happen in this case is that we get a "bit already cleared" message from
2169 * ext4_free_inode(). The only reason we would point at a wrong inode is if
2170 * e2fsck was run on this filesystem, and it must have already done the orphan
2171 * inode cleanup for us, so we can safely abort without any further action.
2173 static void ext4_orphan_cleanup(struct super_block *sb,
2174 struct ext4_super_block *es)
2176 unsigned int s_flags = sb->s_flags;
2177 int nr_orphans = 0, nr_truncates = 0;
2181 if (!es->s_last_orphan) {
2182 jbd_debug(4, "no orphan inodes to clean up\n");
2186 if (bdev_read_only(sb->s_bdev)) {
2187 ext4_msg(sb, KERN_ERR, "write access "
2188 "unavailable, skipping orphan cleanup");
2192 /* Check if feature set would not allow a r/w mount */
2193 if (!ext4_feature_set_ok(sb, 0)) {
2194 ext4_msg(sb, KERN_INFO, "Skipping orphan cleanup due to "
2195 "unknown ROCOMPAT features");
2199 if (EXT4_SB(sb)->s_mount_state & EXT4_ERROR_FS) {
2200 /* don't clear list on RO mount w/ errors */
2201 if (es->s_last_orphan && !(s_flags & MS_RDONLY)) {
2202 ext4_msg(sb, KERN_INFO, "Errors on filesystem, "
2203 "clearing orphan list.\n");
2204 es->s_last_orphan = 0;
2206 jbd_debug(1, "Skipping orphan recovery on fs with errors.\n");
2210 if (s_flags & MS_RDONLY) {
2211 ext4_msg(sb, KERN_INFO, "orphan cleanup on readonly fs");
2212 sb->s_flags &= ~MS_RDONLY;
2215 /* Needed for iput() to work correctly and not trash data */
2216 sb->s_flags |= MS_ACTIVE;
2217 /* Turn on quotas so that they are updated correctly */
2218 for (i = 0; i < EXT4_MAXQUOTAS; i++) {
2219 if (EXT4_SB(sb)->s_qf_names[i]) {
2220 int ret = ext4_quota_on_mount(sb, i);
2222 ext4_msg(sb, KERN_ERR,
2223 "Cannot turn on journaled "
2224 "quota: error %d", ret);
2229 while (es->s_last_orphan) {
2230 struct inode *inode;
2232 inode = ext4_orphan_get(sb, le32_to_cpu(es->s_last_orphan));
2233 if (IS_ERR(inode)) {
2234 es->s_last_orphan = 0;
2238 list_add(&EXT4_I(inode)->i_orphan, &EXT4_SB(sb)->s_orphan);
2239 dquot_initialize(inode);
2240 if (inode->i_nlink) {
2241 if (test_opt(sb, DEBUG))
2242 ext4_msg(sb, KERN_DEBUG,
2243 "%s: truncating inode %lu to %lld bytes",
2244 __func__, inode->i_ino, inode->i_size);
2245 jbd_debug(2, "truncating inode %lu to %lld bytes\n",
2246 inode->i_ino, inode->i_size);
2247 mutex_lock(&inode->i_mutex);
2248 truncate_inode_pages(inode->i_mapping, inode->i_size);
2249 ext4_truncate(inode);
2250 mutex_unlock(&inode->i_mutex);
2253 if (test_opt(sb, DEBUG))
2254 ext4_msg(sb, KERN_DEBUG,
2255 "%s: deleting unreferenced inode %lu",
2256 __func__, inode->i_ino);
2257 jbd_debug(2, "deleting unreferenced inode %lu\n",
2261 iput(inode); /* The delete magic happens here! */
2264 #define PLURAL(x) (x), ((x) == 1) ? "" : "s"
2267 ext4_msg(sb, KERN_INFO, "%d orphan inode%s deleted",
2268 PLURAL(nr_orphans));
2270 ext4_msg(sb, KERN_INFO, "%d truncate%s cleaned up",
2271 PLURAL(nr_truncates));
2273 /* Turn quotas off */
2274 for (i = 0; i < EXT4_MAXQUOTAS; i++) {
2275 if (sb_dqopt(sb)->files[i])
2276 dquot_quota_off(sb, i);
2279 sb->s_flags = s_flags; /* Restore MS_RDONLY status */
2283 * Maximal extent format file size.
2284 * Resulting logical blkno at s_maxbytes must fit in our on-disk
2285 * extent format containers, within a sector_t, and within i_blocks
2286 * in the vfs. ext4 inode has 48 bits of i_block in fsblock units,
2287 * so that won't be a limiting factor.
2289 * However there is other limiting factor. We do store extents in the form
2290 * of starting block and length, hence the resulting length of the extent
2291 * covering maximum file size must fit into on-disk format containers as
2292 * well. Given that length is always by 1 unit bigger than max unit (because
2293 * we count 0 as well) we have to lower the s_maxbytes by one fs block.
2295 * Note, this does *not* consider any metadata overhead for vfs i_blocks.
2297 static loff_t ext4_max_size(int blkbits, int has_huge_files)
2300 loff_t upper_limit = MAX_LFS_FILESIZE;
2302 /* small i_blocks in vfs inode? */
2303 if (!has_huge_files || sizeof(blkcnt_t) < sizeof(u64)) {
2305 * CONFIG_LBDAF is not enabled implies the inode
2306 * i_block represent total blocks in 512 bytes
2307 * 32 == size of vfs inode i_blocks * 8
2309 upper_limit = (1LL << 32) - 1;
2311 /* total blocks in file system block size */
2312 upper_limit >>= (blkbits - 9);
2313 upper_limit <<= blkbits;
2317 * 32-bit extent-start container, ee_block. We lower the maxbytes
2318 * by one fs block, so ee_len can cover the extent of maximum file
2321 res = (1LL << 32) - 1;
2324 /* Sanity check against vm- & vfs- imposed limits */
2325 if (res > upper_limit)
2332 * Maximal bitmap file size. There is a direct, and {,double-,triple-}indirect
2333 * block limit, and also a limit of (2^48 - 1) 512-byte sectors in i_blocks.
2334 * We need to be 1 filesystem block less than the 2^48 sector limit.
2336 static loff_t ext4_max_bitmap_size(int bits, int has_huge_files)
2338 loff_t res = EXT4_NDIR_BLOCKS;
2341 /* This is calculated to be the largest file size for a dense, block
2342 * mapped file such that the file's total number of 512-byte sectors,
2343 * including data and all indirect blocks, does not exceed (2^48 - 1).
2345 * __u32 i_blocks_lo and _u16 i_blocks_high represent the total
2346 * number of 512-byte sectors of the file.
2349 if (!has_huge_files || sizeof(blkcnt_t) < sizeof(u64)) {
2351 * !has_huge_files or CONFIG_LBDAF not enabled implies that
2352 * the inode i_block field represents total file blocks in
2353 * 2^32 512-byte sectors == size of vfs inode i_blocks * 8
2355 upper_limit = (1LL << 32) - 1;
2357 /* total blocks in file system block size */
2358 upper_limit >>= (bits - 9);
2362 * We use 48 bit ext4_inode i_blocks
2363 * With EXT4_HUGE_FILE_FL set the i_blocks
2364 * represent total number of blocks in
2365 * file system block size
2367 upper_limit = (1LL << 48) - 1;
2371 /* indirect blocks */
2373 /* double indirect blocks */
2374 meta_blocks += 1 + (1LL << (bits-2));
2375 /* tripple indirect blocks */
2376 meta_blocks += 1 + (1LL << (bits-2)) + (1LL << (2*(bits-2)));
2378 upper_limit -= meta_blocks;
2379 upper_limit <<= bits;
2381 res += 1LL << (bits-2);
2382 res += 1LL << (2*(bits-2));
2383 res += 1LL << (3*(bits-2));
2385 if (res > upper_limit)
2388 if (res > MAX_LFS_FILESIZE)
2389 res = MAX_LFS_FILESIZE;
2394 static ext4_fsblk_t descriptor_loc(struct super_block *sb,
2395 ext4_fsblk_t logical_sb_block, int nr)
2397 struct ext4_sb_info *sbi = EXT4_SB(sb);
2398 ext4_group_t bg, first_meta_bg;
2401 first_meta_bg = le32_to_cpu(sbi->s_es->s_first_meta_bg);
2403 if (!EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_META_BG) ||
2405 return logical_sb_block + nr + 1;
2406 bg = sbi->s_desc_per_block * nr;
2407 if (ext4_bg_has_super(sb, bg))
2411 * If we have a meta_bg fs with 1k blocks, group 0's GDT is at
2412 * block 2, not 1. If s_first_data_block == 0 (bigalloc is enabled
2413 * on modern mke2fs or blksize > 1k on older mke2fs) then we must
2416 if (sb->s_blocksize == 1024 && nr == 0 &&
2417 le32_to_cpu(EXT4_SB(sb)->s_es->s_first_data_block) == 0)
2420 return (has_super + ext4_group_first_block_no(sb, bg));
2424 * ext4_get_stripe_size: Get the stripe size.
2425 * @sbi: In memory super block info
2427 * If we have specified it via mount option, then
2428 * use the mount option value. If the value specified at mount time is
2429 * greater than the blocks per group use the super block value.
2430 * If the super block value is greater than blocks per group return 0.
2431 * Allocator needs it be less than blocks per group.
2434 static unsigned long ext4_get_stripe_size(struct ext4_sb_info *sbi)
2436 unsigned long stride = le16_to_cpu(sbi->s_es->s_raid_stride);
2437 unsigned long stripe_width =
2438 le32_to_cpu(sbi->s_es->s_raid_stripe_width);
2441 if (sbi->s_stripe && sbi->s_stripe <= sbi->s_blocks_per_group)
2442 ret = sbi->s_stripe;
2443 else if (stripe_width <= sbi->s_blocks_per_group)
2445 else if (stride <= sbi->s_blocks_per_group)
2451 * If the stripe width is 1, this makes no sense and
2452 * we set it to 0 to turn off stripe handling code.
2463 struct attribute attr;
2464 ssize_t (*show)(struct ext4_attr *, struct ext4_sb_info *, char *);
2465 ssize_t (*store)(struct ext4_attr *, struct ext4_sb_info *,
2466 const char *, size_t);
2473 static int parse_strtoull(const char *buf,
2474 unsigned long long max, unsigned long long *value)
2478 ret = kstrtoull(skip_spaces(buf), 0, value);
2479 if (!ret && *value > max)
2484 static ssize_t delayed_allocation_blocks_show(struct ext4_attr *a,
2485 struct ext4_sb_info *sbi,
2488 return snprintf(buf, PAGE_SIZE, "%llu\n",
2490 percpu_counter_sum(&sbi->s_dirtyclusters_counter)));
2493 static ssize_t session_write_kbytes_show(struct ext4_attr *a,
2494 struct ext4_sb_info *sbi, char *buf)
2496 struct super_block *sb = sbi->s_buddy_cache->i_sb;
2498 if (!sb->s_bdev->bd_part)
2499 return snprintf(buf, PAGE_SIZE, "0\n");
2500 return snprintf(buf, PAGE_SIZE, "%lu\n",
2501 (part_stat_read(sb->s_bdev->bd_part, sectors[1]) -
2502 sbi->s_sectors_written_start) >> 1);
2505 static ssize_t lifetime_write_kbytes_show(struct ext4_attr *a,
2506 struct ext4_sb_info *sbi, char *buf)
2508 struct super_block *sb = sbi->s_buddy_cache->i_sb;
2510 if (!sb->s_bdev->bd_part)
2511 return snprintf(buf, PAGE_SIZE, "0\n");
2512 return snprintf(buf, PAGE_SIZE, "%llu\n",
2513 (unsigned long long)(sbi->s_kbytes_written +
2514 ((part_stat_read(sb->s_bdev->bd_part, sectors[1]) -
2515 EXT4_SB(sb)->s_sectors_written_start) >> 1)));
2518 static ssize_t inode_readahead_blks_store(struct ext4_attr *a,
2519 struct ext4_sb_info *sbi,
2520 const char *buf, size_t count)
2525 ret = kstrtoul(skip_spaces(buf), 0, &t);
2529 if (t && (!is_power_of_2(t) || t > 0x40000000))
2532 sbi->s_inode_readahead_blks = t;
2536 static ssize_t sbi_ui_show(struct ext4_attr *a,
2537 struct ext4_sb_info *sbi, char *buf)
2539 unsigned int *ui = (unsigned int *) (((char *) sbi) + a->u.offset);
2541 return snprintf(buf, PAGE_SIZE, "%u\n", *ui);
2544 static ssize_t sbi_ui_store(struct ext4_attr *a,
2545 struct ext4_sb_info *sbi,
2546 const char *buf, size_t count)
2548 unsigned int *ui = (unsigned int *) (((char *) sbi) + a->u.offset);
2552 ret = kstrtoul(skip_spaces(buf), 0, &t);
2559 static ssize_t es_ui_show(struct ext4_attr *a,
2560 struct ext4_sb_info *sbi, char *buf)
2563 unsigned int *ui = (unsigned int *) (((char *) sbi->s_es) +
2566 return snprintf(buf, PAGE_SIZE, "%u\n", *ui);
2569 static ssize_t reserved_clusters_show(struct ext4_attr *a,
2570 struct ext4_sb_info *sbi, char *buf)
2572 return snprintf(buf, PAGE_SIZE, "%llu\n",
2573 (unsigned long long) atomic64_read(&sbi->s_resv_clusters));
2576 static ssize_t reserved_clusters_store(struct ext4_attr *a,
2577 struct ext4_sb_info *sbi,
2578 const char *buf, size_t count)
2580 unsigned long long val;
2583 if (parse_strtoull(buf, -1ULL, &val))
2585 ret = ext4_reserve_clusters(sbi, val);
2587 return ret ? ret : count;
2590 static ssize_t trigger_test_error(struct ext4_attr *a,
2591 struct ext4_sb_info *sbi,
2592 const char *buf, size_t count)
2596 if (!capable(CAP_SYS_ADMIN))
2599 if (len && buf[len-1] == '\n')
2603 ext4_error(sbi->s_sb, "%.*s", len, buf);
2607 static ssize_t sbi_deprecated_show(struct ext4_attr *a,
2608 struct ext4_sb_info *sbi, char *buf)
2610 return snprintf(buf, PAGE_SIZE, "%d\n", a->u.deprecated_val);
2613 #define EXT4_ATTR_OFFSET(_name,_mode,_show,_store,_elname) \
2614 static struct ext4_attr ext4_attr_##_name = { \
2615 .attr = {.name = __stringify(_name), .mode = _mode }, \
2619 .offset = offsetof(struct ext4_sb_info, _elname),\
2623 #define EXT4_ATTR_OFFSET_ES(_name,_mode,_show,_store,_elname) \
2624 static struct ext4_attr ext4_attr_##_name = { \
2625 .attr = {.name = __stringify(_name), .mode = _mode }, \
2629 .offset = offsetof(struct ext4_super_block, _elname), \
2633 #define EXT4_ATTR(name, mode, show, store) \
2634 static struct ext4_attr ext4_attr_##name = __ATTR(name, mode, show, store)
2636 #define EXT4_INFO_ATTR(name) EXT4_ATTR(name, 0444, NULL, NULL)
2637 #define EXT4_RO_ATTR(name) EXT4_ATTR(name, 0444, name##_show, NULL)
2638 #define EXT4_RW_ATTR(name) EXT4_ATTR(name, 0644, name##_show, name##_store)
2640 #define EXT4_RO_ATTR_ES_UI(name, elname) \
2641 EXT4_ATTR_OFFSET_ES(name, 0444, es_ui_show, NULL, elname)
2642 #define EXT4_RW_ATTR_SBI_UI(name, elname) \
2643 EXT4_ATTR_OFFSET(name, 0644, sbi_ui_show, sbi_ui_store, elname)
2645 #define ATTR_LIST(name) &ext4_attr_##name.attr
2646 #define EXT4_DEPRECATED_ATTR(_name, _val) \
2647 static struct ext4_attr ext4_attr_##_name = { \
2648 .attr = {.name = __stringify(_name), .mode = 0444 }, \
2649 .show = sbi_deprecated_show, \
2651 .deprecated_val = _val, \
2655 EXT4_RO_ATTR(delayed_allocation_blocks);
2656 EXT4_RO_ATTR(session_write_kbytes);
2657 EXT4_RO_ATTR(lifetime_write_kbytes);
2658 EXT4_RW_ATTR(reserved_clusters);
2659 EXT4_ATTR_OFFSET(inode_readahead_blks, 0644, sbi_ui_show,
2660 inode_readahead_blks_store, s_inode_readahead_blks);
2661 EXT4_RW_ATTR_SBI_UI(inode_goal, s_inode_goal);
2662 EXT4_RW_ATTR_SBI_UI(mb_stats, s_mb_stats);
2663 EXT4_RW_ATTR_SBI_UI(mb_max_to_scan, s_mb_max_to_scan);
2664 EXT4_RW_ATTR_SBI_UI(mb_min_to_scan, s_mb_min_to_scan);
2665 EXT4_RW_ATTR_SBI_UI(mb_order2_req, s_mb_order2_reqs);
2666 EXT4_RW_ATTR_SBI_UI(mb_stream_req, s_mb_stream_request);
2667 EXT4_RW_ATTR_SBI_UI(mb_group_prealloc, s_mb_group_prealloc);
2668 EXT4_DEPRECATED_ATTR(max_writeback_mb_bump, 128);
2669 EXT4_RW_ATTR_SBI_UI(extent_max_zeroout_kb, s_extent_max_zeroout_kb);
2670 EXT4_ATTR(trigger_fs_error, 0200, NULL, trigger_test_error);
2671 EXT4_RW_ATTR_SBI_UI(err_ratelimit_interval_ms, s_err_ratelimit_state.interval);
2672 EXT4_RW_ATTR_SBI_UI(err_ratelimit_burst, s_err_ratelimit_state.burst);
2673 EXT4_RW_ATTR_SBI_UI(warning_ratelimit_interval_ms, s_warning_ratelimit_state.interval);
2674 EXT4_RW_ATTR_SBI_UI(warning_ratelimit_burst, s_warning_ratelimit_state.burst);
2675 EXT4_RW_ATTR_SBI_UI(msg_ratelimit_interval_ms, s_msg_ratelimit_state.interval);
2676 EXT4_RW_ATTR_SBI_UI(msg_ratelimit_burst, s_msg_ratelimit_state.burst);
2677 EXT4_RO_ATTR_ES_UI(errors_count, s_error_count);
2678 EXT4_RO_ATTR_ES_UI(first_error_time, s_first_error_time);
2679 EXT4_RO_ATTR_ES_UI(last_error_time, s_last_error_time);
2681 static struct attribute *ext4_attrs[] = {
2682 ATTR_LIST(delayed_allocation_blocks),
2683 ATTR_LIST(session_write_kbytes),
2684 ATTR_LIST(lifetime_write_kbytes),
2685 ATTR_LIST(reserved_clusters),
2686 ATTR_LIST(inode_readahead_blks),
2687 ATTR_LIST(inode_goal),
2688 ATTR_LIST(mb_stats),
2689 ATTR_LIST(mb_max_to_scan),
2690 ATTR_LIST(mb_min_to_scan),
2691 ATTR_LIST(mb_order2_req),
2692 ATTR_LIST(mb_stream_req),
2693 ATTR_LIST(mb_group_prealloc),
2694 ATTR_LIST(max_writeback_mb_bump),
2695 ATTR_LIST(extent_max_zeroout_kb),
2696 ATTR_LIST(trigger_fs_error),
2697 ATTR_LIST(err_ratelimit_interval_ms),
2698 ATTR_LIST(err_ratelimit_burst),
2699 ATTR_LIST(warning_ratelimit_interval_ms),
2700 ATTR_LIST(warning_ratelimit_burst),
2701 ATTR_LIST(msg_ratelimit_interval_ms),
2702 ATTR_LIST(msg_ratelimit_burst),
2703 ATTR_LIST(errors_count),
2704 ATTR_LIST(first_error_time),
2705 ATTR_LIST(last_error_time),
2709 /* Features this copy of ext4 supports */
2710 EXT4_INFO_ATTR(lazy_itable_init);
2711 EXT4_INFO_ATTR(batched_discard);
2712 EXT4_INFO_ATTR(meta_bg_resize);
2713 EXT4_INFO_ATTR(encryption);
2715 static struct attribute *ext4_feat_attrs[] = {
2716 ATTR_LIST(lazy_itable_init),
2717 ATTR_LIST(batched_discard),
2718 ATTR_LIST(meta_bg_resize),
2719 ATTR_LIST(encryption),
2723 static ssize_t ext4_attr_show(struct kobject *kobj,
2724 struct attribute *attr, char *buf)
2726 struct ext4_sb_info *sbi = container_of(kobj, struct ext4_sb_info,
2728 struct ext4_attr *a = container_of(attr, struct ext4_attr, attr);
2730 return a->show ? a->show(a, sbi, buf) : 0;
2733 static ssize_t ext4_attr_store(struct kobject *kobj,
2734 struct attribute *attr,
2735 const char *buf, size_t len)
2737 struct ext4_sb_info *sbi = container_of(kobj, struct ext4_sb_info,
2739 struct ext4_attr *a = container_of(attr, struct ext4_attr, attr);
2741 return a->store ? a->store(a, sbi, buf, len) : 0;
2744 static void ext4_sb_release(struct kobject *kobj)
2746 struct ext4_sb_info *sbi = container_of(kobj, struct ext4_sb_info,
2748 complete(&sbi->s_kobj_unregister);
2751 static const struct sysfs_ops ext4_attr_ops = {
2752 .show = ext4_attr_show,
2753 .store = ext4_attr_store,
2756 static struct kobj_type ext4_ktype = {
2757 .default_attrs = ext4_attrs,
2758 .sysfs_ops = &ext4_attr_ops,
2759 .release = ext4_sb_release,
2762 static void ext4_feat_release(struct kobject *kobj)
2764 complete(&ext4_feat->f_kobj_unregister);
2767 static ssize_t ext4_feat_show(struct kobject *kobj,
2768 struct attribute *attr, char *buf)
2770 return snprintf(buf, PAGE_SIZE, "supported\n");
2774 * We can not use ext4_attr_show/store because it relies on the kobject
2775 * being embedded in the ext4_sb_info structure which is definitely not
2776 * true in this case.
2778 static const struct sysfs_ops ext4_feat_ops = {
2779 .show = ext4_feat_show,
2783 static struct kobj_type ext4_feat_ktype = {
2784 .default_attrs = ext4_feat_attrs,
2785 .sysfs_ops = &ext4_feat_ops,
2786 .release = ext4_feat_release,
2790 * Check whether this filesystem can be mounted based on
2791 * the features present and the RDONLY/RDWR mount requested.
2792 * Returns 1 if this filesystem can be mounted as requested,
2793 * 0 if it cannot be.
2795 static int ext4_feature_set_ok(struct super_block *sb, int readonly)
2797 if (EXT4_HAS_INCOMPAT_FEATURE(sb, ~EXT4_FEATURE_INCOMPAT_SUPP)) {
2798 ext4_msg(sb, KERN_ERR,
2799 "Couldn't mount because of "
2800 "unsupported optional features (%x)",
2801 (le32_to_cpu(EXT4_SB(sb)->s_es->s_feature_incompat) &
2802 ~EXT4_FEATURE_INCOMPAT_SUPP));
2809 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_READONLY)) {
2810 ext4_msg(sb, KERN_INFO, "filesystem is read-only");
2811 sb->s_flags |= MS_RDONLY;
2815 /* Check that feature set is OK for a read-write mount */
2816 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, ~EXT4_FEATURE_RO_COMPAT_SUPP)) {
2817 ext4_msg(sb, KERN_ERR, "couldn't mount RDWR because of "
2818 "unsupported optional features (%x)",
2819 (le32_to_cpu(EXT4_SB(sb)->s_es->s_feature_ro_compat) &
2820 ~EXT4_FEATURE_RO_COMPAT_SUPP));
2824 * Large file size enabled file system can only be mounted
2825 * read-write on 32-bit systems if kernel is built with CONFIG_LBDAF
2827 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_HUGE_FILE)) {
2828 if (sizeof(blkcnt_t) < sizeof(u64)) {
2829 ext4_msg(sb, KERN_ERR, "Filesystem with huge files "
2830 "cannot be mounted RDWR without "
2835 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_BIGALLOC) &&
2836 !EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_EXTENTS)) {
2837 ext4_msg(sb, KERN_ERR,
2838 "Can't support bigalloc feature without "
2839 "extents feature\n");
2843 #ifndef CONFIG_QUOTA
2844 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_QUOTA) &&
2846 ext4_msg(sb, KERN_ERR,
2847 "Filesystem with quota feature cannot be mounted RDWR "
2848 "without CONFIG_QUOTA");
2851 #endif /* CONFIG_QUOTA */
2856 * This function is called once a day if we have errors logged
2857 * on the file system
2859 static void print_daily_error_info(unsigned long arg)
2861 struct super_block *sb = (struct super_block *) arg;
2862 struct ext4_sb_info *sbi;
2863 struct ext4_super_block *es;
2868 if (es->s_error_count)
2869 /* fsck newer than v1.41.13 is needed to clean this condition. */
2870 ext4_msg(sb, KERN_NOTICE, "error count since last fsck: %u",
2871 le32_to_cpu(es->s_error_count));
2872 if (es->s_first_error_time) {
2873 printk(KERN_NOTICE "EXT4-fs (%s): initial error at time %u: %.*s:%d",
2874 sb->s_id, le32_to_cpu(es->s_first_error_time),
2875 (int) sizeof(es->s_first_error_func),
2876 es->s_first_error_func,
2877 le32_to_cpu(es->s_first_error_line));
2878 if (es->s_first_error_ino)
2879 printk(": inode %u",
2880 le32_to_cpu(es->s_first_error_ino));
2881 if (es->s_first_error_block)
2882 printk(": block %llu", (unsigned long long)
2883 le64_to_cpu(es->s_first_error_block));
2886 if (es->s_last_error_time) {
2887 printk(KERN_NOTICE "EXT4-fs (%s): last error at time %u: %.*s:%d",
2888 sb->s_id, le32_to_cpu(es->s_last_error_time),
2889 (int) sizeof(es->s_last_error_func),
2890 es->s_last_error_func,
2891 le32_to_cpu(es->s_last_error_line));
2892 if (es->s_last_error_ino)
2893 printk(": inode %u",
2894 le32_to_cpu(es->s_last_error_ino));
2895 if (es->s_last_error_block)
2896 printk(": block %llu", (unsigned long long)
2897 le64_to_cpu(es->s_last_error_block));
2900 mod_timer(&sbi->s_err_report, jiffies + 24*60*60*HZ); /* Once a day */
2903 /* Find next suitable group and run ext4_init_inode_table */
2904 static int ext4_run_li_request(struct ext4_li_request *elr)
2906 struct ext4_group_desc *gdp = NULL;
2907 ext4_group_t group, ngroups;
2908 struct super_block *sb;
2909 unsigned long timeout = 0;
2913 ngroups = EXT4_SB(sb)->s_groups_count;
2916 for (group = elr->lr_next_group; group < ngroups; group++) {
2917 gdp = ext4_get_group_desc(sb, group, NULL);
2923 if (!(gdp->bg_flags & cpu_to_le16(EXT4_BG_INODE_ZEROED)))
2927 if (group >= ngroups)
2932 ret = ext4_init_inode_table(sb, group,
2933 elr->lr_timeout ? 0 : 1);
2934 if (elr->lr_timeout == 0) {
2935 timeout = (jiffies - timeout) *
2936 elr->lr_sbi->s_li_wait_mult;
2937 elr->lr_timeout = timeout;
2939 elr->lr_next_sched = jiffies + elr->lr_timeout;
2940 elr->lr_next_group = group + 1;
2948 * Remove lr_request from the list_request and free the
2949 * request structure. Should be called with li_list_mtx held
2951 static void ext4_remove_li_request(struct ext4_li_request *elr)
2953 struct ext4_sb_info *sbi;
2960 list_del(&elr->lr_request);
2961 sbi->s_li_request = NULL;
2965 static void ext4_unregister_li_request(struct super_block *sb)
2967 mutex_lock(&ext4_li_mtx);
2968 if (!ext4_li_info) {
2969 mutex_unlock(&ext4_li_mtx);
2973 mutex_lock(&ext4_li_info->li_list_mtx);
2974 ext4_remove_li_request(EXT4_SB(sb)->s_li_request);
2975 mutex_unlock(&ext4_li_info->li_list_mtx);
2976 mutex_unlock(&ext4_li_mtx);
2979 static struct task_struct *ext4_lazyinit_task;
2982 * This is the function where ext4lazyinit thread lives. It walks
2983 * through the request list searching for next scheduled filesystem.
2984 * When such a fs is found, run the lazy initialization request
2985 * (ext4_rn_li_request) and keep track of the time spend in this
2986 * function. Based on that time we compute next schedule time of
2987 * the request. When walking through the list is complete, compute
2988 * next waking time and put itself into sleep.
2990 static int ext4_lazyinit_thread(void *arg)
2992 struct ext4_lazy_init *eli = (struct ext4_lazy_init *)arg;
2993 struct list_head *pos, *n;
2994 struct ext4_li_request *elr;
2995 unsigned long next_wakeup, cur;
2997 BUG_ON(NULL == eli);
3001 next_wakeup = MAX_JIFFY_OFFSET;
3003 mutex_lock(&eli->li_list_mtx);
3004 if (list_empty(&eli->li_request_list)) {
3005 mutex_unlock(&eli->li_list_mtx);
3009 list_for_each_safe(pos, n, &eli->li_request_list) {
3010 elr = list_entry(pos, struct ext4_li_request,
3013 if (time_after_eq(jiffies, elr->lr_next_sched)) {
3014 if (ext4_run_li_request(elr) != 0) {
3015 /* error, remove the lazy_init job */
3016 ext4_remove_li_request(elr);
3021 if (time_before(elr->lr_next_sched, next_wakeup))
3022 next_wakeup = elr->lr_next_sched;
3024 mutex_unlock(&eli->li_list_mtx);
3029 if ((time_after_eq(cur, next_wakeup)) ||
3030 (MAX_JIFFY_OFFSET == next_wakeup)) {
3035 schedule_timeout_interruptible(next_wakeup - cur);
3037 if (kthread_should_stop()) {
3038 ext4_clear_request_list();
3045 * It looks like the request list is empty, but we need
3046 * to check it under the li_list_mtx lock, to prevent any
3047 * additions into it, and of course we should lock ext4_li_mtx
3048 * to atomically free the list and ext4_li_info, because at
3049 * this point another ext4 filesystem could be registering
3052 mutex_lock(&ext4_li_mtx);
3053 mutex_lock(&eli->li_list_mtx);
3054 if (!list_empty(&eli->li_request_list)) {
3055 mutex_unlock(&eli->li_list_mtx);
3056 mutex_unlock(&ext4_li_mtx);
3059 mutex_unlock(&eli->li_list_mtx);
3060 kfree(ext4_li_info);
3061 ext4_li_info = NULL;
3062 mutex_unlock(&ext4_li_mtx);
3067 static void ext4_clear_request_list(void)
3069 struct list_head *pos, *n;
3070 struct ext4_li_request *elr;
3072 mutex_lock(&ext4_li_info->li_list_mtx);
3073 list_for_each_safe(pos, n, &ext4_li_info->li_request_list) {
3074 elr = list_entry(pos, struct ext4_li_request,
3076 ext4_remove_li_request(elr);
3078 mutex_unlock(&ext4_li_info->li_list_mtx);
3081 static int ext4_run_lazyinit_thread(void)
3083 ext4_lazyinit_task = kthread_run(ext4_lazyinit_thread,
3084 ext4_li_info, "ext4lazyinit");
3085 if (IS_ERR(ext4_lazyinit_task)) {
3086 int err = PTR_ERR(ext4_lazyinit_task);
3087 ext4_clear_request_list();
3088 kfree(ext4_li_info);
3089 ext4_li_info = NULL;
3090 printk(KERN_CRIT "EXT4-fs: error %d creating inode table "
3091 "initialization thread\n",
3095 ext4_li_info->li_state |= EXT4_LAZYINIT_RUNNING;
3100 * Check whether it make sense to run itable init. thread or not.
3101 * If there is at least one uninitialized inode table, return
3102 * corresponding group number, else the loop goes through all
3103 * groups and return total number of groups.
3105 static ext4_group_t ext4_has_uninit_itable(struct super_block *sb)
3107 ext4_group_t group, ngroups = EXT4_SB(sb)->s_groups_count;
3108 struct ext4_group_desc *gdp = NULL;
3110 for (group = 0; group < ngroups; group++) {
3111 gdp = ext4_get_group_desc(sb, group, NULL);
3115 if (!(gdp->bg_flags & cpu_to_le16(EXT4_BG_INODE_ZEROED)))
3122 static int ext4_li_info_new(void)
3124 struct ext4_lazy_init *eli = NULL;
3126 eli = kzalloc(sizeof(*eli), GFP_KERNEL);
3130 INIT_LIST_HEAD(&eli->li_request_list);
3131 mutex_init(&eli->li_list_mtx);
3133 eli->li_state |= EXT4_LAZYINIT_QUIT;
3140 static struct ext4_li_request *ext4_li_request_new(struct super_block *sb,
3143 struct ext4_sb_info *sbi = EXT4_SB(sb);
3144 struct ext4_li_request *elr;
3146 elr = kzalloc(sizeof(*elr), GFP_KERNEL);
3152 elr->lr_next_group = start;
3155 * Randomize first schedule time of the request to
3156 * spread the inode table initialization requests
3159 elr->lr_next_sched = jiffies + (prandom_u32() %
3160 (EXT4_DEF_LI_MAX_START_DELAY * HZ));
3164 int ext4_register_li_request(struct super_block *sb,
3165 ext4_group_t first_not_zeroed)
3167 struct ext4_sb_info *sbi = EXT4_SB(sb);
3168 struct ext4_li_request *elr = NULL;
3169 ext4_group_t ngroups = EXT4_SB(sb)->s_groups_count;
3172 mutex_lock(&ext4_li_mtx);
3173 if (sbi->s_li_request != NULL) {
3175 * Reset timeout so it can be computed again, because
3176 * s_li_wait_mult might have changed.
3178 sbi->s_li_request->lr_timeout = 0;
3182 if (first_not_zeroed == ngroups ||
3183 (sb->s_flags & MS_RDONLY) ||
3184 !test_opt(sb, INIT_INODE_TABLE))
3187 elr = ext4_li_request_new(sb, first_not_zeroed);
3193 if (NULL == ext4_li_info) {
3194 ret = ext4_li_info_new();
3199 mutex_lock(&ext4_li_info->li_list_mtx);
3200 list_add(&elr->lr_request, &ext4_li_info->li_request_list);
3201 mutex_unlock(&ext4_li_info->li_list_mtx);
3203 sbi->s_li_request = elr;
3205 * set elr to NULL here since it has been inserted to
3206 * the request_list and the removal and free of it is
3207 * handled by ext4_clear_request_list from now on.
3211 if (!(ext4_li_info->li_state & EXT4_LAZYINIT_RUNNING)) {
3212 ret = ext4_run_lazyinit_thread();
3217 mutex_unlock(&ext4_li_mtx);
3224 * We do not need to lock anything since this is called on
3227 static void ext4_destroy_lazyinit_thread(void)
3230 * If thread exited earlier
3231 * there's nothing to be done.
3233 if (!ext4_li_info || !ext4_lazyinit_task)
3236 kthread_stop(ext4_lazyinit_task);
3239 static int set_journal_csum_feature_set(struct super_block *sb)
3242 int compat, incompat;
3243 struct ext4_sb_info *sbi = EXT4_SB(sb);
3245 if (ext4_has_metadata_csum(sb)) {
3246 /* journal checksum v3 */
3248 incompat = JBD2_FEATURE_INCOMPAT_CSUM_V3;
3250 /* journal checksum v1 */
3251 compat = JBD2_FEATURE_COMPAT_CHECKSUM;
3255 jbd2_journal_clear_features(sbi->s_journal,
3256 JBD2_FEATURE_COMPAT_CHECKSUM, 0,
3257 JBD2_FEATURE_INCOMPAT_CSUM_V3 |
3258 JBD2_FEATURE_INCOMPAT_CSUM_V2);
3259 if (test_opt(sb, JOURNAL_ASYNC_COMMIT)) {
3260 ret = jbd2_journal_set_features(sbi->s_journal,
3262 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT |
3264 } else if (test_opt(sb, JOURNAL_CHECKSUM)) {
3265 ret = jbd2_journal_set_features(sbi->s_journal,
3268 jbd2_journal_clear_features(sbi->s_journal, 0, 0,
3269 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT);
3271 jbd2_journal_clear_features(sbi->s_journal, 0, 0,
3272 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT);
3279 * Note: calculating the overhead so we can be compatible with
3280 * historical BSD practice is quite difficult in the face of
3281 * clusters/bigalloc. This is because multiple metadata blocks from
3282 * different block group can end up in the same allocation cluster.
3283 * Calculating the exact overhead in the face of clustered allocation
3284 * requires either O(all block bitmaps) in memory or O(number of block
3285 * groups**2) in time. We will still calculate the superblock for
3286 * older file systems --- and if we come across with a bigalloc file
3287 * system with zero in s_overhead_clusters the estimate will be close to
3288 * correct especially for very large cluster sizes --- but for newer
3289 * file systems, it's better to calculate this figure once at mkfs
3290 * time, and store it in the superblock. If the superblock value is
3291 * present (even for non-bigalloc file systems), we will use it.
3293 static int count_overhead(struct super_block *sb, ext4_group_t grp,
3296 struct ext4_sb_info *sbi = EXT4_SB(sb);
3297 struct ext4_group_desc *gdp;
3298 ext4_fsblk_t first_block, last_block, b;
3299 ext4_group_t i, ngroups = ext4_get_groups_count(sb);
3300 int s, j, count = 0;
3302 if (!EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_BIGALLOC))
3303 return (ext4_bg_has_super(sb, grp) + ext4_bg_num_gdb(sb, grp) +
3304 sbi->s_itb_per_group + 2);
3306 first_block = le32_to_cpu(sbi->s_es->s_first_data_block) +
3307 (grp * EXT4_BLOCKS_PER_GROUP(sb));
3308 last_block = first_block + EXT4_BLOCKS_PER_GROUP(sb) - 1;
3309 for (i = 0; i < ngroups; i++) {
3310 gdp = ext4_get_group_desc(sb, i, NULL);
3311 b = ext4_block_bitmap(sb, gdp);
3312 if (b >= first_block && b <= last_block) {
3313 ext4_set_bit(EXT4_B2C(sbi, b - first_block), buf);
3316 b = ext4_inode_bitmap(sb, gdp);
3317 if (b >= first_block && b <= last_block) {
3318 ext4_set_bit(EXT4_B2C(sbi, b - first_block), buf);
3321 b = ext4_inode_table(sb, gdp);
3322 if (b >= first_block && b + sbi->s_itb_per_group <= last_block)
3323 for (j = 0; j < sbi->s_itb_per_group; j++, b++) {
3324 int c = EXT4_B2C(sbi, b - first_block);
3325 ext4_set_bit(c, buf);
3331 if (ext4_bg_has_super(sb, grp)) {
3332 ext4_set_bit(s++, buf);
3335 for (j = ext4_bg_num_gdb(sb, grp); j > 0; j--) {
3336 ext4_set_bit(EXT4_B2C(sbi, s++), buf);
3342 return EXT4_CLUSTERS_PER_GROUP(sb) -
3343 ext4_count_free(buf, EXT4_CLUSTERS_PER_GROUP(sb) / 8);
3347 * Compute the overhead and stash it in sbi->s_overhead
3349 int ext4_calculate_overhead(struct super_block *sb)
3351 struct ext4_sb_info *sbi = EXT4_SB(sb);
3352 struct ext4_super_block *es = sbi->s_es;
3353 ext4_group_t i, ngroups = ext4_get_groups_count(sb);
3354 ext4_fsblk_t overhead = 0;
3355 char *buf = (char *) get_zeroed_page(GFP_NOFS);
3361 * Compute the overhead (FS structures). This is constant
3362 * for a given filesystem unless the number of block groups
3363 * changes so we cache the previous value until it does.
3367 * All of the blocks before first_data_block are overhead
3369 overhead = EXT4_B2C(sbi, le32_to_cpu(es->s_first_data_block));
3372 * Add the overhead found in each block group
3374 for (i = 0; i < ngroups; i++) {
3377 blks = count_overhead(sb, i, buf);
3380 memset(buf, 0, PAGE_SIZE);
3383 /* Add the internal journal blocks as well */
3384 if (sbi->s_journal && !sbi->journal_bdev)
3385 overhead += EXT4_NUM_B2C(sbi, sbi->s_journal->j_maxlen);
3387 sbi->s_overhead = overhead;
3389 free_page((unsigned long) buf);
3394 static ext4_fsblk_t ext4_calculate_resv_clusters(struct super_block *sb)
3396 ext4_fsblk_t resv_clusters;
3399 * There's no need to reserve anything when we aren't using extents.
3400 * The space estimates are exact, there are no unwritten extents,
3401 * hole punching doesn't need new metadata... This is needed especially
3402 * to keep ext2/3 backward compatibility.
3404 if (!EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_EXTENTS))
3407 * By default we reserve 2% or 4096 clusters, whichever is smaller.
3408 * This should cover the situations where we can not afford to run
3409 * out of space like for example punch hole, or converting
3410 * unwritten extents in delalloc path. In most cases such
3411 * allocation would require 1, or 2 blocks, higher numbers are
3414 resv_clusters = ext4_blocks_count(EXT4_SB(sb)->s_es) >>
3415 EXT4_SB(sb)->s_cluster_bits;
3417 do_div(resv_clusters, 50);
3418 resv_clusters = min_t(ext4_fsblk_t, resv_clusters, 4096);
3420 return resv_clusters;
3424 static int ext4_reserve_clusters(struct ext4_sb_info *sbi, ext4_fsblk_t count)
3426 ext4_fsblk_t clusters = ext4_blocks_count(sbi->s_es) >>
3427 sbi->s_cluster_bits;
3429 if (count >= clusters)
3432 atomic64_set(&sbi->s_resv_clusters, count);
3436 static int ext4_fill_super(struct super_block *sb, void *data, int silent)
3438 char *orig_data = kstrdup(data, GFP_KERNEL);
3439 struct buffer_head *bh;
3440 struct ext4_super_block *es = NULL;
3441 struct ext4_sb_info *sbi;
3443 ext4_fsblk_t sb_block = get_sb_block(&data);
3444 ext4_fsblk_t logical_sb_block;
3445 unsigned long offset = 0;
3446 unsigned long journal_devnum = 0;
3447 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 strreplace(sb->s_id, '/', '!');
3481 /* -EINVAL is default */
3483 blocksize = sb_min_blocksize(sb, EXT4_MIN_BLOCK_SIZE);
3485 ext4_msg(sb, KERN_ERR, "unable to set blocksize");
3490 * The ext4 superblock will not be buffer aligned for other than 1kB
3491 * block sizes. We need to calculate the offset from buffer start.
3493 if (blocksize != EXT4_MIN_BLOCK_SIZE) {
3494 logical_sb_block = sb_block * EXT4_MIN_BLOCK_SIZE;
3495 offset = do_div(logical_sb_block, blocksize);
3497 logical_sb_block = sb_block;
3500 if (!(bh = sb_bread_unmovable(sb, logical_sb_block))) {
3501 ext4_msg(sb, KERN_ERR, "unable to read superblock");
3505 * Note: s_es must be initialized as soon as possible because
3506 * some ext4 macro-instructions depend on its value
3508 es = (struct ext4_super_block *) (bh->b_data + offset);
3510 sb->s_magic = le16_to_cpu(es->s_magic);
3511 if (sb->s_magic != EXT4_SUPER_MAGIC)
3513 sbi->s_kbytes_written = le64_to_cpu(es->s_kbytes_written);
3515 /* Warn if metadata_csum and gdt_csum are both set. */
3516 if (EXT4_HAS_RO_COMPAT_FEATURE(sb,
3517 EXT4_FEATURE_RO_COMPAT_METADATA_CSUM) &&
3518 EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_GDT_CSUM))
3519 ext4_warning(sb, "metadata_csum and uninit_bg are "
3520 "redundant flags; please run fsck.");
3522 /* Check for a known checksum algorithm */
3523 if (!ext4_verify_csum_type(sb, es)) {
3524 ext4_msg(sb, KERN_ERR, "VFS: Found ext4 filesystem with "
3525 "unknown checksum algorithm.");
3530 /* Load the checksum driver */
3531 if (EXT4_HAS_RO_COMPAT_FEATURE(sb,
3532 EXT4_FEATURE_RO_COMPAT_METADATA_CSUM)) {
3533 sbi->s_chksum_driver = crypto_alloc_shash("crc32c", 0, 0);
3534 if (IS_ERR(sbi->s_chksum_driver)) {
3535 ext4_msg(sb, KERN_ERR, "Cannot load crc32c driver.");
3536 ret = PTR_ERR(sbi->s_chksum_driver);
3537 sbi->s_chksum_driver = NULL;
3542 /* Check superblock checksum */
3543 if (!ext4_superblock_csum_verify(sb, es)) {
3544 ext4_msg(sb, KERN_ERR, "VFS: Found ext4 filesystem with "
3545 "invalid superblock checksum. Run e2fsck?");
3550 /* Precompute checksum seed for all metadata */
3551 if (ext4_has_metadata_csum(sb))
3552 sbi->s_csum_seed = ext4_chksum(sbi, ~0, es->s_uuid,
3553 sizeof(es->s_uuid));
3555 /* Set defaults before we parse the mount options */
3556 def_mount_opts = le32_to_cpu(es->s_default_mount_opts);
3557 set_opt(sb, INIT_INODE_TABLE);
3558 if (def_mount_opts & EXT4_DEFM_DEBUG)
3560 if (def_mount_opts & EXT4_DEFM_BSDGROUPS)
3562 if (def_mount_opts & EXT4_DEFM_UID16)
3563 set_opt(sb, NO_UID32);
3564 /* xattr user namespace & acls are now defaulted on */
3565 set_opt(sb, XATTR_USER);
3566 #ifdef CONFIG_EXT4_FS_POSIX_ACL
3567 set_opt(sb, POSIX_ACL);
3569 /* don't forget to enable journal_csum when metadata_csum is enabled. */
3570 if (ext4_has_metadata_csum(sb))
3571 set_opt(sb, JOURNAL_CHECKSUM);
3573 if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_DATA)
3574 set_opt(sb, JOURNAL_DATA);
3575 else if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_ORDERED)
3576 set_opt(sb, ORDERED_DATA);
3577 else if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_WBACK)
3578 set_opt(sb, WRITEBACK_DATA);
3580 if (le16_to_cpu(sbi->s_es->s_errors) == EXT4_ERRORS_PANIC)
3581 set_opt(sb, ERRORS_PANIC);
3582 else if (le16_to_cpu(sbi->s_es->s_errors) == EXT4_ERRORS_CONTINUE)
3583 set_opt(sb, ERRORS_CONT);
3585 set_opt(sb, ERRORS_RO);
3586 /* block_validity enabled by default; disable with noblock_validity */
3587 set_opt(sb, BLOCK_VALIDITY);
3588 if (def_mount_opts & EXT4_DEFM_DISCARD)
3589 set_opt(sb, DISCARD);
3591 sbi->s_resuid = make_kuid(&init_user_ns, le16_to_cpu(es->s_def_resuid));
3592 sbi->s_resgid = make_kgid(&init_user_ns, le16_to_cpu(es->s_def_resgid));
3593 sbi->s_commit_interval = JBD2_DEFAULT_MAX_COMMIT_AGE * HZ;
3594 sbi->s_min_batch_time = EXT4_DEF_MIN_BATCH_TIME;
3595 sbi->s_max_batch_time = EXT4_DEF_MAX_BATCH_TIME;
3597 if ((def_mount_opts & EXT4_DEFM_NOBARRIER) == 0)
3598 set_opt(sb, BARRIER);
3601 * enable delayed allocation by default
3602 * Use -o nodelalloc to turn it off
3604 if (!IS_EXT3_SB(sb) && !IS_EXT2_SB(sb) &&
3605 ((def_mount_opts & EXT4_DEFM_NODELALLOC) == 0))
3606 set_opt(sb, DELALLOC);
3609 * set default s_li_wait_mult for lazyinit, for the case there is
3610 * no mount option specified.
3612 sbi->s_li_wait_mult = EXT4_DEF_LI_WAIT_MULT;
3614 if (!parse_options((char *) sbi->s_es->s_mount_opts, sb,
3615 &journal_devnum, &journal_ioprio, 0)) {
3616 ext4_msg(sb, KERN_WARNING,
3617 "failed to parse options in superblock: %s",
3618 sbi->s_es->s_mount_opts);
3620 sbi->s_def_mount_opt = sbi->s_mount_opt;
3621 if (!parse_options((char *) data, sb, &journal_devnum,
3622 &journal_ioprio, 0))
3625 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA) {
3626 printk_once(KERN_WARNING "EXT4-fs: Warning: mounting "
3627 "with data=journal disables delayed "
3628 "allocation and O_DIRECT support!\n");
3629 if (test_opt2(sb, EXPLICIT_DELALLOC)) {
3630 ext4_msg(sb, KERN_ERR, "can't mount with "
3631 "both data=journal and delalloc");
3634 if (test_opt(sb, DIOREAD_NOLOCK)) {
3635 ext4_msg(sb, KERN_ERR, "can't mount with "
3636 "both data=journal and dioread_nolock");
3639 if (test_opt(sb, DAX)) {
3640 ext4_msg(sb, KERN_ERR, "can't mount with "
3641 "both data=journal and dax");
3644 if (test_opt(sb, DELALLOC))
3645 clear_opt(sb, DELALLOC);
3647 sb->s_iflags |= SB_I_CGROUPWB;
3650 sb->s_flags = (sb->s_flags & ~MS_POSIXACL) |
3651 (test_opt(sb, POSIX_ACL) ? MS_POSIXACL : 0);
3653 if (le32_to_cpu(es->s_rev_level) == EXT4_GOOD_OLD_REV &&
3654 (EXT4_HAS_COMPAT_FEATURE(sb, ~0U) ||
3655 EXT4_HAS_RO_COMPAT_FEATURE(sb, ~0U) ||
3656 EXT4_HAS_INCOMPAT_FEATURE(sb, ~0U)))
3657 ext4_msg(sb, KERN_WARNING,
3658 "feature flags set on rev 0 fs, "
3659 "running e2fsck is recommended");
3661 if (es->s_creator_os == cpu_to_le32(EXT4_OS_HURD)) {
3662 set_opt2(sb, HURD_COMPAT);
3663 if (EXT4_HAS_INCOMPAT_FEATURE(sb,
3664 EXT4_FEATURE_INCOMPAT_64BIT)) {
3665 ext4_msg(sb, KERN_ERR,
3666 "The Hurd can't support 64-bit file systems");
3671 if (IS_EXT2_SB(sb)) {
3672 if (ext2_feature_set_ok(sb))
3673 ext4_msg(sb, KERN_INFO, "mounting ext2 file system "
3674 "using the ext4 subsystem");
3676 ext4_msg(sb, KERN_ERR, "couldn't mount as ext2 due "
3677 "to feature incompatibilities");
3682 if (IS_EXT3_SB(sb)) {
3683 if (ext3_feature_set_ok(sb))
3684 ext4_msg(sb, KERN_INFO, "mounting ext3 file system "
3685 "using the ext4 subsystem");
3687 ext4_msg(sb, KERN_ERR, "couldn't mount as ext3 due "
3688 "to feature incompatibilities");
3694 * Check feature flags regardless of the revision level, since we
3695 * previously didn't change the revision level when setting the flags,
3696 * so there is a chance incompat flags are set on a rev 0 filesystem.
3698 if (!ext4_feature_set_ok(sb, (sb->s_flags & MS_RDONLY)))
3701 blocksize = BLOCK_SIZE << le32_to_cpu(es->s_log_block_size);
3702 if (blocksize < EXT4_MIN_BLOCK_SIZE ||
3703 blocksize > EXT4_MAX_BLOCK_SIZE) {
3704 ext4_msg(sb, KERN_ERR,
3705 "Unsupported filesystem blocksize %d", blocksize);
3709 if (sbi->s_mount_opt & EXT4_MOUNT_DAX) {
3710 if (blocksize != PAGE_SIZE) {
3711 ext4_msg(sb, KERN_ERR,
3712 "error: unsupported blocksize for dax");
3715 if (!sb->s_bdev->bd_disk->fops->direct_access) {
3716 ext4_msg(sb, KERN_ERR,
3717 "error: device does not support dax");
3722 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_ENCRYPT) &&
3723 es->s_encryption_level) {
3724 ext4_msg(sb, KERN_ERR, "Unsupported encryption level %d",
3725 es->s_encryption_level);
3729 if (sb->s_blocksize != blocksize) {
3730 /* Validate the filesystem blocksize */
3731 if (!sb_set_blocksize(sb, blocksize)) {
3732 ext4_msg(sb, KERN_ERR, "bad block size %d",
3738 logical_sb_block = sb_block * EXT4_MIN_BLOCK_SIZE;
3739 offset = do_div(logical_sb_block, blocksize);
3740 bh = sb_bread_unmovable(sb, logical_sb_block);
3742 ext4_msg(sb, KERN_ERR,
3743 "Can't read superblock on 2nd try");
3746 es = (struct ext4_super_block *)(bh->b_data + offset);
3748 if (es->s_magic != cpu_to_le16(EXT4_SUPER_MAGIC)) {
3749 ext4_msg(sb, KERN_ERR,
3750 "Magic mismatch, very weird!");
3755 has_huge_files = EXT4_HAS_RO_COMPAT_FEATURE(sb,
3756 EXT4_FEATURE_RO_COMPAT_HUGE_FILE);
3757 sbi->s_bitmap_maxbytes = ext4_max_bitmap_size(sb->s_blocksize_bits,
3759 sb->s_maxbytes = ext4_max_size(sb->s_blocksize_bits, has_huge_files);
3761 if (le32_to_cpu(es->s_rev_level) == EXT4_GOOD_OLD_REV) {
3762 sbi->s_inode_size = EXT4_GOOD_OLD_INODE_SIZE;
3763 sbi->s_first_ino = EXT4_GOOD_OLD_FIRST_INO;
3765 sbi->s_inode_size = le16_to_cpu(es->s_inode_size);
3766 sbi->s_first_ino = le32_to_cpu(es->s_first_ino);
3767 if ((sbi->s_inode_size < EXT4_GOOD_OLD_INODE_SIZE) ||
3768 (!is_power_of_2(sbi->s_inode_size)) ||
3769 (sbi->s_inode_size > blocksize)) {
3770 ext4_msg(sb, KERN_ERR,
3771 "unsupported inode size: %d",
3775 if (sbi->s_inode_size > EXT4_GOOD_OLD_INODE_SIZE)
3776 sb->s_time_gran = 1 << (EXT4_EPOCH_BITS - 2);
3779 sbi->s_desc_size = le16_to_cpu(es->s_desc_size);
3780 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_64BIT)) {
3781 if (sbi->s_desc_size < EXT4_MIN_DESC_SIZE_64BIT ||
3782 sbi->s_desc_size > EXT4_MAX_DESC_SIZE ||
3783 !is_power_of_2(sbi->s_desc_size)) {
3784 ext4_msg(sb, KERN_ERR,
3785 "unsupported descriptor size %lu",
3790 sbi->s_desc_size = EXT4_MIN_DESC_SIZE;
3792 sbi->s_blocks_per_group = le32_to_cpu(es->s_blocks_per_group);
3793 sbi->s_inodes_per_group = le32_to_cpu(es->s_inodes_per_group);
3794 if (EXT4_INODE_SIZE(sb) == 0 || EXT4_INODES_PER_GROUP(sb) == 0)
3797 sbi->s_inodes_per_block = blocksize / EXT4_INODE_SIZE(sb);
3798 if (sbi->s_inodes_per_block == 0)
3800 sbi->s_itb_per_group = sbi->s_inodes_per_group /
3801 sbi->s_inodes_per_block;
3802 sbi->s_desc_per_block = blocksize / EXT4_DESC_SIZE(sb);
3804 sbi->s_mount_state = le16_to_cpu(es->s_state);
3805 sbi->s_addr_per_block_bits = ilog2(EXT4_ADDR_PER_BLOCK(sb));
3806 sbi->s_desc_per_block_bits = ilog2(EXT4_DESC_PER_BLOCK(sb));
3808 for (i = 0; i < 4; i++)
3809 sbi->s_hash_seed[i] = le32_to_cpu(es->s_hash_seed[i]);
3810 sbi->s_def_hash_version = es->s_def_hash_version;
3811 if (EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_DIR_INDEX)) {
3812 i = le32_to_cpu(es->s_flags);
3813 if (i & EXT2_FLAGS_UNSIGNED_HASH)
3814 sbi->s_hash_unsigned = 3;
3815 else if ((i & EXT2_FLAGS_SIGNED_HASH) == 0) {
3816 #ifdef __CHAR_UNSIGNED__
3817 if (!(sb->s_flags & MS_RDONLY))
3819 cpu_to_le32(EXT2_FLAGS_UNSIGNED_HASH);
3820 sbi->s_hash_unsigned = 3;
3822 if (!(sb->s_flags & MS_RDONLY))
3824 cpu_to_le32(EXT2_FLAGS_SIGNED_HASH);
3829 /* Handle clustersize */
3830 clustersize = BLOCK_SIZE << le32_to_cpu(es->s_log_cluster_size);
3831 has_bigalloc = EXT4_HAS_RO_COMPAT_FEATURE(sb,
3832 EXT4_FEATURE_RO_COMPAT_BIGALLOC);
3834 if (clustersize < blocksize) {
3835 ext4_msg(sb, KERN_ERR,
3836 "cluster size (%d) smaller than "
3837 "block size (%d)", clustersize, blocksize);
3840 sbi->s_cluster_bits = le32_to_cpu(es->s_log_cluster_size) -
3841 le32_to_cpu(es->s_log_block_size);
3842 sbi->s_clusters_per_group =
3843 le32_to_cpu(es->s_clusters_per_group);
3844 if (sbi->s_clusters_per_group > blocksize * 8) {
3845 ext4_msg(sb, KERN_ERR,
3846 "#clusters per group too big: %lu",
3847 sbi->s_clusters_per_group);
3850 if (sbi->s_blocks_per_group !=
3851 (sbi->s_clusters_per_group * (clustersize / blocksize))) {
3852 ext4_msg(sb, KERN_ERR, "blocks per group (%lu) and "
3853 "clusters per group (%lu) inconsistent",
3854 sbi->s_blocks_per_group,
3855 sbi->s_clusters_per_group);
3859 if (clustersize != blocksize) {
3860 ext4_warning(sb, "fragment/cluster size (%d) != "
3861 "block size (%d)", clustersize,
3863 clustersize = blocksize;
3865 if (sbi->s_blocks_per_group > blocksize * 8) {
3866 ext4_msg(sb, KERN_ERR,
3867 "#blocks per group too big: %lu",
3868 sbi->s_blocks_per_group);
3871 sbi->s_clusters_per_group = sbi->s_blocks_per_group;
3872 sbi->s_cluster_bits = 0;
3874 sbi->s_cluster_ratio = clustersize / blocksize;
3876 if (sbi->s_inodes_per_group > blocksize * 8) {
3877 ext4_msg(sb, KERN_ERR,
3878 "#inodes per group too big: %lu",
3879 sbi->s_inodes_per_group);
3883 /* Do we have standard group size of clustersize * 8 blocks ? */
3884 if (sbi->s_blocks_per_group == clustersize << 3)
3885 set_opt2(sb, STD_GROUP_SIZE);
3888 * Test whether we have more sectors than will fit in sector_t,
3889 * and whether the max offset is addressable by the page cache.
3891 err = generic_check_addressable(sb->s_blocksize_bits,
3892 ext4_blocks_count(es));
3894 ext4_msg(sb, KERN_ERR, "filesystem"
3895 " too large to mount safely on this system");
3896 if (sizeof(sector_t) < 8)
3897 ext4_msg(sb, KERN_WARNING, "CONFIG_LBDAF not enabled");
3901 if (EXT4_BLOCKS_PER_GROUP(sb) == 0)
3904 /* check blocks count against device size */
3905 blocks_count = sb->s_bdev->bd_inode->i_size >> sb->s_blocksize_bits;
3906 if (blocks_count && ext4_blocks_count(es) > blocks_count) {
3907 ext4_msg(sb, KERN_WARNING, "bad geometry: block count %llu "
3908 "exceeds size of device (%llu blocks)",
3909 ext4_blocks_count(es), blocks_count);
3914 * It makes no sense for the first data block to be beyond the end
3915 * of the filesystem.
3917 if (le32_to_cpu(es->s_first_data_block) >= ext4_blocks_count(es)) {
3918 ext4_msg(sb, KERN_WARNING, "bad geometry: first data "
3919 "block %u is beyond end of filesystem (%llu)",
3920 le32_to_cpu(es->s_first_data_block),
3921 ext4_blocks_count(es));
3924 blocks_count = (ext4_blocks_count(es) -
3925 le32_to_cpu(es->s_first_data_block) +
3926 EXT4_BLOCKS_PER_GROUP(sb) - 1);
3927 do_div(blocks_count, EXT4_BLOCKS_PER_GROUP(sb));
3928 if (blocks_count > ((uint64_t)1<<32) - EXT4_DESC_PER_BLOCK(sb)) {
3929 ext4_msg(sb, KERN_WARNING, "groups count too large: %u "
3930 "(block count %llu, first data block %u, "
3931 "blocks per group %lu)", sbi->s_groups_count,
3932 ext4_blocks_count(es),
3933 le32_to_cpu(es->s_first_data_block),
3934 EXT4_BLOCKS_PER_GROUP(sb));
3937 sbi->s_groups_count = blocks_count;
3938 sbi->s_blockfile_groups = min_t(ext4_group_t, sbi->s_groups_count,
3939 (EXT4_MAX_BLOCK_FILE_PHYS / EXT4_BLOCKS_PER_GROUP(sb)));
3940 db_count = (sbi->s_groups_count + EXT4_DESC_PER_BLOCK(sb) - 1) /
3941 EXT4_DESC_PER_BLOCK(sb);
3942 sbi->s_group_desc = ext4_kvmalloc(db_count *
3943 sizeof(struct buffer_head *),
3945 if (sbi->s_group_desc == NULL) {
3946 ext4_msg(sb, KERN_ERR, "not enough memory");
3952 sbi->s_proc = proc_mkdir(sb->s_id, ext4_proc_root);
3955 proc_create_data("options", S_IRUGO, sbi->s_proc,
3956 &ext4_seq_options_fops, sb);
3958 bgl_lock_init(sbi->s_blockgroup_lock);
3960 for (i = 0; i < db_count; i++) {
3961 block = descriptor_loc(sb, logical_sb_block, i);
3962 sbi->s_group_desc[i] = sb_bread_unmovable(sb, block);
3963 if (!sbi->s_group_desc[i]) {
3964 ext4_msg(sb, KERN_ERR,
3965 "can't read group descriptor %d", i);
3970 if (!ext4_check_descriptors(sb, &first_not_zeroed)) {
3971 ext4_msg(sb, KERN_ERR, "group descriptors corrupted!");
3975 sbi->s_gdb_count = db_count;
3976 get_random_bytes(&sbi->s_next_generation, sizeof(u32));
3977 spin_lock_init(&sbi->s_next_gen_lock);
3979 setup_timer(&sbi->s_err_report, print_daily_error_info,
3980 (unsigned long) sb);
3982 /* Register extent status tree shrinker */
3983 if (ext4_es_register_shrinker(sbi))
3986 sbi->s_stripe = ext4_get_stripe_size(sbi);
3987 sbi->s_extent_max_zeroout_kb = 32;
3990 * set up enough so that it can read an inode
3992 sb->s_op = &ext4_sops;
3993 sb->s_export_op = &ext4_export_ops;
3994 sb->s_xattr = ext4_xattr_handlers;
3996 sb->dq_op = &ext4_quota_operations;
3997 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_QUOTA))
3998 sb->s_qcop = &dquot_quotactl_sysfile_ops;
4000 sb->s_qcop = &ext4_qctl_operations;
4001 sb->s_quota_types = QTYPE_MASK_USR | QTYPE_MASK_GRP;
4003 memcpy(sb->s_uuid, es->s_uuid, sizeof(es->s_uuid));
4005 INIT_LIST_HEAD(&sbi->s_orphan); /* unlinked but open files */
4006 mutex_init(&sbi->s_orphan_lock);
4010 needs_recovery = (es->s_last_orphan != 0 ||
4011 EXT4_HAS_INCOMPAT_FEATURE(sb,
4012 EXT4_FEATURE_INCOMPAT_RECOVER));
4014 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_MMP) &&
4015 !(sb->s_flags & MS_RDONLY))
4016 if (ext4_multi_mount_protect(sb, le64_to_cpu(es->s_mmp_block)))
4017 goto failed_mount3a;
4020 * The first inode we look at is the journal inode. Don't try
4021 * root first: it may be modified in the journal!
4023 if (!test_opt(sb, NOLOAD) &&
4024 EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL)) {
4025 if (ext4_load_journal(sb, es, journal_devnum))
4026 goto failed_mount3a;
4027 } else if (test_opt(sb, NOLOAD) && !(sb->s_flags & MS_RDONLY) &&
4028 EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER)) {
4029 ext4_msg(sb, KERN_ERR, "required journal recovery "
4030 "suppressed and not mounted read-only");
4031 goto failed_mount_wq;
4033 clear_opt(sb, DATA_FLAGS);
4034 sbi->s_journal = NULL;
4039 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_64BIT) &&
4040 !jbd2_journal_set_features(EXT4_SB(sb)->s_journal, 0, 0,
4041 JBD2_FEATURE_INCOMPAT_64BIT)) {
4042 ext4_msg(sb, KERN_ERR, "Failed to set 64-bit journal feature");
4043 goto failed_mount_wq;
4046 if (!set_journal_csum_feature_set(sb)) {
4047 ext4_msg(sb, KERN_ERR, "Failed to set journal checksum "
4049 goto failed_mount_wq;
4052 /* We have now updated the journal if required, so we can
4053 * validate the data journaling mode. */
4054 switch (test_opt(sb, DATA_FLAGS)) {
4056 /* No mode set, assume a default based on the journal
4057 * capabilities: ORDERED_DATA if the journal can
4058 * cope, else JOURNAL_DATA
4060 if (jbd2_journal_check_available_features
4061 (sbi->s_journal, 0, 0, JBD2_FEATURE_INCOMPAT_REVOKE))
4062 set_opt(sb, ORDERED_DATA);
4064 set_opt(sb, JOURNAL_DATA);
4067 case EXT4_MOUNT_ORDERED_DATA:
4068 case EXT4_MOUNT_WRITEBACK_DATA:
4069 if (!jbd2_journal_check_available_features
4070 (sbi->s_journal, 0, 0, JBD2_FEATURE_INCOMPAT_REVOKE)) {
4071 ext4_msg(sb, KERN_ERR, "Journal does not support "
4072 "requested data journaling mode");
4073 goto failed_mount_wq;
4078 set_task_ioprio(sbi->s_journal->j_task, journal_ioprio);
4080 sbi->s_journal->j_commit_callback = ext4_journal_commit_callback;
4083 if (ext4_mballoc_ready) {
4084 sbi->s_mb_cache = ext4_xattr_create_cache(sb->s_id);
4085 if (!sbi->s_mb_cache) {
4086 ext4_msg(sb, KERN_ERR, "Failed to create an mb_cache");
4087 goto failed_mount_wq;
4091 if ((DUMMY_ENCRYPTION_ENABLED(sbi) ||
4092 EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_ENCRYPT)) &&
4093 (blocksize != PAGE_CACHE_SIZE)) {
4094 ext4_msg(sb, KERN_ERR,
4095 "Unsupported blocksize for fs encryption");
4096 goto failed_mount_wq;
4099 if (DUMMY_ENCRYPTION_ENABLED(sbi) &&
4100 !(sb->s_flags & MS_RDONLY) &&
4101 !EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_ENCRYPT)) {
4102 EXT4_SET_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_ENCRYPT);
4103 ext4_commit_super(sb, 1);
4107 * Get the # of file system overhead blocks from the
4108 * superblock if present.
4110 if (es->s_overhead_clusters)
4111 sbi->s_overhead = le32_to_cpu(es->s_overhead_clusters);
4113 err = ext4_calculate_overhead(sb);
4115 goto failed_mount_wq;
4119 * The maximum number of concurrent works can be high and
4120 * concurrency isn't really necessary. Limit it to 1.
4122 EXT4_SB(sb)->rsv_conversion_wq =
4123 alloc_workqueue("ext4-rsv-conversion", WQ_MEM_RECLAIM | WQ_UNBOUND, 1);
4124 if (!EXT4_SB(sb)->rsv_conversion_wq) {
4125 printk(KERN_ERR "EXT4-fs: failed to create workqueue\n");
4131 * The jbd2_journal_load will have done any necessary log recovery,
4132 * so we can safely mount the rest of the filesystem now.
4135 root = ext4_iget(sb, EXT4_ROOT_INO);
4137 ext4_msg(sb, KERN_ERR, "get root inode failed");
4138 ret = PTR_ERR(root);
4142 if (!S_ISDIR(root->i_mode) || !root->i_blocks || !root->i_size) {
4143 ext4_msg(sb, KERN_ERR, "corrupt root inode, run e2fsck");
4147 sb->s_root = d_make_root(root);
4149 ext4_msg(sb, KERN_ERR, "get root dentry failed");
4154 if (ext4_setup_super(sb, es, sb->s_flags & MS_RDONLY))
4155 sb->s_flags |= MS_RDONLY;
4157 /* determine the minimum size of new large inodes, if present */
4158 if (sbi->s_inode_size > EXT4_GOOD_OLD_INODE_SIZE) {
4159 sbi->s_want_extra_isize = sizeof(struct ext4_inode) -
4160 EXT4_GOOD_OLD_INODE_SIZE;
4161 if (EXT4_HAS_RO_COMPAT_FEATURE(sb,
4162 EXT4_FEATURE_RO_COMPAT_EXTRA_ISIZE)) {
4163 if (sbi->s_want_extra_isize <
4164 le16_to_cpu(es->s_want_extra_isize))
4165 sbi->s_want_extra_isize =
4166 le16_to_cpu(es->s_want_extra_isize);
4167 if (sbi->s_want_extra_isize <
4168 le16_to_cpu(es->s_min_extra_isize))
4169 sbi->s_want_extra_isize =
4170 le16_to_cpu(es->s_min_extra_isize);
4173 /* Check if enough inode space is available */
4174 if (EXT4_GOOD_OLD_INODE_SIZE + sbi->s_want_extra_isize >
4175 sbi->s_inode_size) {
4176 sbi->s_want_extra_isize = sizeof(struct ext4_inode) -
4177 EXT4_GOOD_OLD_INODE_SIZE;
4178 ext4_msg(sb, KERN_INFO, "required extra inode space not"
4182 err = ext4_reserve_clusters(sbi, ext4_calculate_resv_clusters(sb));
4184 ext4_msg(sb, KERN_ERR, "failed to reserve %llu clusters for "
4185 "reserved pool", ext4_calculate_resv_clusters(sb));
4186 goto failed_mount4a;
4189 err = ext4_setup_system_zone(sb);
4191 ext4_msg(sb, KERN_ERR, "failed to initialize system "
4193 goto failed_mount4a;
4197 err = ext4_mb_init(sb);
4199 ext4_msg(sb, KERN_ERR, "failed to initialize mballoc (%d)",
4204 block = ext4_count_free_clusters(sb);
4205 ext4_free_blocks_count_set(sbi->s_es,
4206 EXT4_C2B(sbi, block));
4207 err = percpu_counter_init(&sbi->s_freeclusters_counter, block,
4210 unsigned long freei = ext4_count_free_inodes(sb);
4211 sbi->s_es->s_free_inodes_count = cpu_to_le32(freei);
4212 err = percpu_counter_init(&sbi->s_freeinodes_counter, freei,
4216 err = percpu_counter_init(&sbi->s_dirs_counter,
4217 ext4_count_dirs(sb), GFP_KERNEL);
4219 err = percpu_counter_init(&sbi->s_dirtyclusters_counter, 0,
4222 ext4_msg(sb, KERN_ERR, "insufficient memory");
4226 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_FLEX_BG))
4227 if (!ext4_fill_flex_info(sb)) {
4228 ext4_msg(sb, KERN_ERR,
4229 "unable to initialize "
4230 "flex_bg meta info!");
4234 err = ext4_register_li_request(sb, first_not_zeroed);
4238 sbi->s_kobj.kset = ext4_kset;
4239 init_completion(&sbi->s_kobj_unregister);
4240 err = kobject_init_and_add(&sbi->s_kobj, &ext4_ktype, NULL,
4246 /* Enable quota usage during mount. */
4247 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_QUOTA) &&
4248 !(sb->s_flags & MS_RDONLY)) {
4249 err = ext4_enable_quotas(sb);
4253 #endif /* CONFIG_QUOTA */
4255 EXT4_SB(sb)->s_mount_state |= EXT4_ORPHAN_FS;
4256 ext4_orphan_cleanup(sb, es);
4257 EXT4_SB(sb)->s_mount_state &= ~EXT4_ORPHAN_FS;
4258 if (needs_recovery) {
4259 ext4_msg(sb, KERN_INFO, "recovery complete");
4260 ext4_mark_recovery_complete(sb, es);
4262 if (EXT4_SB(sb)->s_journal) {
4263 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA)
4264 descr = " journalled data mode";
4265 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA)
4266 descr = " ordered data mode";
4268 descr = " writeback data mode";
4270 descr = "out journal";
4272 if (test_opt(sb, DISCARD)) {
4273 struct request_queue *q = bdev_get_queue(sb->s_bdev);
4274 if (!blk_queue_discard(q))
4275 ext4_msg(sb, KERN_WARNING,
4276 "mounting with \"discard\" option, but "
4277 "the device does not support discard");
4280 ext4_msg(sb, KERN_INFO, "mounted filesystem with%s. "
4281 "Opts: %s%s%s", descr, sbi->s_es->s_mount_opts,
4282 *sbi->s_es->s_mount_opts ? "; " : "", orig_data);
4284 if (es->s_error_count)
4285 mod_timer(&sbi->s_err_report, jiffies + 300*HZ); /* 5 minutes */
4287 /* Enable message ratelimiting. Default is 10 messages per 5 secs. */
4288 ratelimit_state_init(&sbi->s_err_ratelimit_state, 5 * HZ, 10);
4289 ratelimit_state_init(&sbi->s_warning_ratelimit_state, 5 * HZ, 10);
4290 ratelimit_state_init(&sbi->s_msg_ratelimit_state, 5 * HZ, 10);
4297 ext4_msg(sb, KERN_ERR, "VFS: Can't find ext4 filesystem");
4302 kobject_del(&sbi->s_kobj);
4305 ext4_unregister_li_request(sb);
4307 ext4_mb_release(sb);
4308 if (sbi->s_flex_groups)
4309 kvfree(sbi->s_flex_groups);
4310 percpu_counter_destroy(&sbi->s_freeclusters_counter);
4311 percpu_counter_destroy(&sbi->s_freeinodes_counter);
4312 percpu_counter_destroy(&sbi->s_dirs_counter);
4313 percpu_counter_destroy(&sbi->s_dirtyclusters_counter);
4315 ext4_ext_release(sb);
4316 ext4_release_system_zone(sb);
4321 ext4_msg(sb, KERN_ERR, "mount failed");
4322 if (EXT4_SB(sb)->rsv_conversion_wq)
4323 destroy_workqueue(EXT4_SB(sb)->rsv_conversion_wq);
4325 if (sbi->s_journal) {
4326 jbd2_journal_destroy(sbi->s_journal);
4327 sbi->s_journal = NULL;
4330 ext4_es_unregister_shrinker(sbi);
4332 del_timer_sync(&sbi->s_err_report);
4334 kthread_stop(sbi->s_mmp_tsk);
4336 for (i = 0; i < db_count; i++)
4337 brelse(sbi->s_group_desc[i]);
4338 kvfree(sbi->s_group_desc);
4340 if (sbi->s_chksum_driver)
4341 crypto_free_shash(sbi->s_chksum_driver);
4343 remove_proc_entry("options", sbi->s_proc);
4344 remove_proc_entry(sb->s_id, ext4_proc_root);
4347 for (i = 0; i < EXT4_MAXQUOTAS; i++)
4348 kfree(sbi->s_qf_names[i]);
4350 ext4_blkdev_remove(sbi);
4353 sb->s_fs_info = NULL;
4354 kfree(sbi->s_blockgroup_lock);
4358 return err ? err : ret;
4362 * Setup any per-fs journal parameters now. We'll do this both on
4363 * initial mount, once the journal has been initialised but before we've
4364 * done any recovery; and again on any subsequent remount.
4366 static void ext4_init_journal_params(struct super_block *sb, journal_t *journal)
4368 struct ext4_sb_info *sbi = EXT4_SB(sb);
4370 journal->j_commit_interval = sbi->s_commit_interval;
4371 journal->j_min_batch_time = sbi->s_min_batch_time;
4372 journal->j_max_batch_time = sbi->s_max_batch_time;
4374 write_lock(&journal->j_state_lock);
4375 if (test_opt(sb, BARRIER))
4376 journal->j_flags |= JBD2_BARRIER;
4378 journal->j_flags &= ~JBD2_BARRIER;
4379 if (test_opt(sb, DATA_ERR_ABORT))
4380 journal->j_flags |= JBD2_ABORT_ON_SYNCDATA_ERR;
4382 journal->j_flags &= ~JBD2_ABORT_ON_SYNCDATA_ERR;
4383 write_unlock(&journal->j_state_lock);
4386 static journal_t *ext4_get_journal(struct super_block *sb,
4387 unsigned int journal_inum)
4389 struct inode *journal_inode;
4392 BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL));
4394 /* First, test for the existence of a valid inode on disk. Bad
4395 * things happen if we iget() an unused inode, as the subsequent
4396 * iput() will try to delete it. */
4398 journal_inode = ext4_iget(sb, journal_inum);
4399 if (IS_ERR(journal_inode)) {
4400 ext4_msg(sb, KERN_ERR, "no journal found");
4403 if (!journal_inode->i_nlink) {
4404 make_bad_inode(journal_inode);
4405 iput(journal_inode);
4406 ext4_msg(sb, KERN_ERR, "journal inode is deleted");
4410 jbd_debug(2, "Journal inode found at %p: %lld bytes\n",
4411 journal_inode, journal_inode->i_size);
4412 if (!S_ISREG(journal_inode->i_mode)) {
4413 ext4_msg(sb, KERN_ERR, "invalid journal inode");
4414 iput(journal_inode);
4418 journal = jbd2_journal_init_inode(journal_inode);
4420 ext4_msg(sb, KERN_ERR, "Could not load journal inode");
4421 iput(journal_inode);
4424 journal->j_private = sb;
4425 ext4_init_journal_params(sb, journal);
4429 static journal_t *ext4_get_dev_journal(struct super_block *sb,
4432 struct buffer_head *bh;
4436 int hblock, blocksize;
4437 ext4_fsblk_t sb_block;
4438 unsigned long offset;
4439 struct ext4_super_block *es;
4440 struct block_device *bdev;
4442 BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL));
4444 bdev = ext4_blkdev_get(j_dev, sb);
4448 blocksize = sb->s_blocksize;
4449 hblock = bdev_logical_block_size(bdev);
4450 if (blocksize < hblock) {
4451 ext4_msg(sb, KERN_ERR,
4452 "blocksize too small for journal device");
4456 sb_block = EXT4_MIN_BLOCK_SIZE / blocksize;
4457 offset = EXT4_MIN_BLOCK_SIZE % blocksize;
4458 set_blocksize(bdev, blocksize);
4459 if (!(bh = __bread(bdev, sb_block, blocksize))) {
4460 ext4_msg(sb, KERN_ERR, "couldn't read superblock of "
4461 "external journal");
4465 es = (struct ext4_super_block *) (bh->b_data + offset);
4466 if ((le16_to_cpu(es->s_magic) != EXT4_SUPER_MAGIC) ||
4467 !(le32_to_cpu(es->s_feature_incompat) &
4468 EXT4_FEATURE_INCOMPAT_JOURNAL_DEV)) {
4469 ext4_msg(sb, KERN_ERR, "external journal has "
4475 if ((le32_to_cpu(es->s_feature_ro_compat) &
4476 EXT4_FEATURE_RO_COMPAT_METADATA_CSUM) &&
4477 es->s_checksum != ext4_superblock_csum(sb, es)) {
4478 ext4_msg(sb, KERN_ERR, "external journal has "
4479 "corrupt superblock");
4484 if (memcmp(EXT4_SB(sb)->s_es->s_journal_uuid, es->s_uuid, 16)) {
4485 ext4_msg(sb, KERN_ERR, "journal UUID does not match");
4490 len = ext4_blocks_count(es);
4491 start = sb_block + 1;
4492 brelse(bh); /* we're done with the superblock */
4494 journal = jbd2_journal_init_dev(bdev, sb->s_bdev,
4495 start, len, blocksize);
4497 ext4_msg(sb, KERN_ERR, "failed to create device journal");
4500 journal->j_private = sb;
4501 ll_rw_block(READ | REQ_META | REQ_PRIO, 1, &journal->j_sb_buffer);
4502 wait_on_buffer(journal->j_sb_buffer);
4503 if (!buffer_uptodate(journal->j_sb_buffer)) {
4504 ext4_msg(sb, KERN_ERR, "I/O error on journal device");
4507 if (be32_to_cpu(journal->j_superblock->s_nr_users) != 1) {
4508 ext4_msg(sb, KERN_ERR, "External journal has more than one "
4509 "user (unsupported) - %d",
4510 be32_to_cpu(journal->j_superblock->s_nr_users));
4513 EXT4_SB(sb)->journal_bdev = bdev;
4514 ext4_init_journal_params(sb, journal);
4518 jbd2_journal_destroy(journal);
4520 ext4_blkdev_put(bdev);
4524 static int ext4_load_journal(struct super_block *sb,
4525 struct ext4_super_block *es,
4526 unsigned long journal_devnum)
4529 unsigned int journal_inum = le32_to_cpu(es->s_journal_inum);
4532 int really_read_only;
4534 BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL));
4536 if (journal_devnum &&
4537 journal_devnum != le32_to_cpu(es->s_journal_dev)) {
4538 ext4_msg(sb, KERN_INFO, "external journal device major/minor "
4539 "numbers have changed");
4540 journal_dev = new_decode_dev(journal_devnum);
4542 journal_dev = new_decode_dev(le32_to_cpu(es->s_journal_dev));
4544 really_read_only = bdev_read_only(sb->s_bdev);
4547 * Are we loading a blank journal or performing recovery after a
4548 * crash? For recovery, we need to check in advance whether we
4549 * can get read-write access to the device.
4551 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER)) {
4552 if (sb->s_flags & MS_RDONLY) {
4553 ext4_msg(sb, KERN_INFO, "INFO: recovery "
4554 "required on readonly filesystem");
4555 if (really_read_only) {
4556 ext4_msg(sb, KERN_ERR, "write access "
4557 "unavailable, cannot proceed");
4560 ext4_msg(sb, KERN_INFO, "write access will "
4561 "be enabled during recovery");
4565 if (journal_inum && journal_dev) {
4566 ext4_msg(sb, KERN_ERR, "filesystem has both journal "
4567 "and inode journals!");
4572 if (!(journal = ext4_get_journal(sb, journal_inum)))
4575 if (!(journal = ext4_get_dev_journal(sb, journal_dev)))
4579 if (!(journal->j_flags & JBD2_BARRIER))
4580 ext4_msg(sb, KERN_INFO, "barriers disabled");
4582 if (!EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER))
4583 err = jbd2_journal_wipe(journal, !really_read_only);
4585 char *save = kmalloc(EXT4_S_ERR_LEN, GFP_KERNEL);
4587 memcpy(save, ((char *) es) +
4588 EXT4_S_ERR_START, EXT4_S_ERR_LEN);
4589 err = jbd2_journal_load(journal);
4591 memcpy(((char *) es) + EXT4_S_ERR_START,
4592 save, EXT4_S_ERR_LEN);
4597 ext4_msg(sb, KERN_ERR, "error loading journal");
4598 jbd2_journal_destroy(journal);
4602 EXT4_SB(sb)->s_journal = journal;
4603 ext4_clear_journal_err(sb, es);
4605 if (!really_read_only && journal_devnum &&
4606 journal_devnum != le32_to_cpu(es->s_journal_dev)) {
4607 es->s_journal_dev = cpu_to_le32(journal_devnum);
4609 /* Make sure we flush the recovery flag to disk. */
4610 ext4_commit_super(sb, 1);
4616 static int ext4_commit_super(struct super_block *sb, int sync)
4618 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
4619 struct buffer_head *sbh = EXT4_SB(sb)->s_sbh;
4624 if (buffer_write_io_error(sbh)) {
4626 * Oh, dear. A previous attempt to write the
4627 * superblock failed. This could happen because the
4628 * USB device was yanked out. Or it could happen to
4629 * be a transient write error and maybe the block will
4630 * be remapped. Nothing we can do but to retry the
4631 * write and hope for the best.
4633 ext4_msg(sb, KERN_ERR, "previous I/O error to "
4634 "superblock detected");
4635 clear_buffer_write_io_error(sbh);
4636 set_buffer_uptodate(sbh);
4639 * If the file system is mounted read-only, don't update the
4640 * superblock write time. This avoids updating the superblock
4641 * write time when we are mounting the root file system
4642 * read/only but we need to replay the journal; at that point,
4643 * for people who are east of GMT and who make their clock
4644 * tick in localtime for Windows bug-for-bug compatibility,
4645 * the clock is set in the future, and this will cause e2fsck
4646 * to complain and force a full file system check.
4648 if (!(sb->s_flags & MS_RDONLY))
4649 es->s_wtime = cpu_to_le32(get_seconds());
4650 if (sb->s_bdev->bd_part)
4651 es->s_kbytes_written =
4652 cpu_to_le64(EXT4_SB(sb)->s_kbytes_written +
4653 ((part_stat_read(sb->s_bdev->bd_part, sectors[1]) -
4654 EXT4_SB(sb)->s_sectors_written_start) >> 1));
4656 es->s_kbytes_written =
4657 cpu_to_le64(EXT4_SB(sb)->s_kbytes_written);
4658 if (percpu_counter_initialized(&EXT4_SB(sb)->s_freeclusters_counter))
4659 ext4_free_blocks_count_set(es,
4660 EXT4_C2B(EXT4_SB(sb), percpu_counter_sum_positive(
4661 &EXT4_SB(sb)->s_freeclusters_counter)));
4662 if (percpu_counter_initialized(&EXT4_SB(sb)->s_freeinodes_counter))
4663 es->s_free_inodes_count =
4664 cpu_to_le32(percpu_counter_sum_positive(
4665 &EXT4_SB(sb)->s_freeinodes_counter));
4666 BUFFER_TRACE(sbh, "marking dirty");
4667 ext4_superblock_csum_set(sb);
4668 mark_buffer_dirty(sbh);
4670 error = sync_dirty_buffer(sbh);
4674 error = buffer_write_io_error(sbh);
4676 ext4_msg(sb, KERN_ERR, "I/O error while writing "
4678 clear_buffer_write_io_error(sbh);
4679 set_buffer_uptodate(sbh);
4686 * Have we just finished recovery? If so, and if we are mounting (or
4687 * remounting) the filesystem readonly, then we will end up with a
4688 * consistent fs on disk. Record that fact.
4690 static void ext4_mark_recovery_complete(struct super_block *sb,
4691 struct ext4_super_block *es)
4693 journal_t *journal = EXT4_SB(sb)->s_journal;
4695 if (!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL)) {
4696 BUG_ON(journal != NULL);
4699 jbd2_journal_lock_updates(journal);
4700 if (jbd2_journal_flush(journal) < 0)
4703 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER) &&
4704 sb->s_flags & MS_RDONLY) {
4705 EXT4_CLEAR_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
4706 ext4_commit_super(sb, 1);
4710 jbd2_journal_unlock_updates(journal);
4714 * If we are mounting (or read-write remounting) a filesystem whose journal
4715 * has recorded an error from a previous lifetime, move that error to the
4716 * main filesystem now.
4718 static void ext4_clear_journal_err(struct super_block *sb,
4719 struct ext4_super_block *es)
4725 BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL));
4727 journal = EXT4_SB(sb)->s_journal;
4730 * Now check for any error status which may have been recorded in the
4731 * journal by a prior ext4_error() or ext4_abort()
4734 j_errno = jbd2_journal_errno(journal);
4738 errstr = ext4_decode_error(sb, j_errno, nbuf);
4739 ext4_warning(sb, "Filesystem error recorded "
4740 "from previous mount: %s", errstr);
4741 ext4_warning(sb, "Marking fs in need of filesystem check.");
4743 EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS;
4744 es->s_state |= cpu_to_le16(EXT4_ERROR_FS);
4745 ext4_commit_super(sb, 1);
4747 jbd2_journal_clear_err(journal);
4748 jbd2_journal_update_sb_errno(journal);
4753 * Force the running and committing transactions to commit,
4754 * and wait on the commit.
4756 int ext4_force_commit(struct super_block *sb)
4760 if (sb->s_flags & MS_RDONLY)
4763 journal = EXT4_SB(sb)->s_journal;
4764 return ext4_journal_force_commit(journal);
4767 static int ext4_sync_fs(struct super_block *sb, int wait)
4771 bool needs_barrier = false;
4772 struct ext4_sb_info *sbi = EXT4_SB(sb);
4774 trace_ext4_sync_fs(sb, wait);
4775 flush_workqueue(sbi->rsv_conversion_wq);
4777 * Writeback quota in non-journalled quota case - journalled quota has
4780 dquot_writeback_dquots(sb, -1);
4782 * Data writeback is possible w/o journal transaction, so barrier must
4783 * being sent at the end of the function. But we can skip it if
4784 * transaction_commit will do it for us.
4786 if (sbi->s_journal) {
4787 target = jbd2_get_latest_transaction(sbi->s_journal);
4788 if (wait && sbi->s_journal->j_flags & JBD2_BARRIER &&
4789 !jbd2_trans_will_send_data_barrier(sbi->s_journal, target))
4790 needs_barrier = true;
4792 if (jbd2_journal_start_commit(sbi->s_journal, &target)) {
4794 ret = jbd2_log_wait_commit(sbi->s_journal,
4797 } else if (wait && test_opt(sb, BARRIER))
4798 needs_barrier = true;
4799 if (needs_barrier) {
4801 err = blkdev_issue_flush(sb->s_bdev, GFP_KERNEL, NULL);
4810 * LVM calls this function before a (read-only) snapshot is created. This
4811 * gives us a chance to flush the journal completely and mark the fs clean.
4813 * Note that only this function cannot bring a filesystem to be in a clean
4814 * state independently. It relies on upper layer to stop all data & metadata
4817 static int ext4_freeze(struct super_block *sb)
4822 if (sb->s_flags & MS_RDONLY)
4825 journal = EXT4_SB(sb)->s_journal;
4828 /* Now we set up the journal barrier. */
4829 jbd2_journal_lock_updates(journal);
4832 * Don't clear the needs_recovery flag if we failed to
4833 * flush the journal.
4835 error = jbd2_journal_flush(journal);
4840 /* Journal blocked and flushed, clear needs_recovery flag. */
4841 EXT4_CLEAR_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
4842 error = ext4_commit_super(sb, 1);
4845 /* we rely on upper layer to stop further updates */
4846 jbd2_journal_unlock_updates(journal);
4851 * Called by LVM after the snapshot is done. We need to reset the RECOVER
4852 * flag here, even though the filesystem is not technically dirty yet.
4854 static int ext4_unfreeze(struct super_block *sb)
4856 if (sb->s_flags & MS_RDONLY)
4859 /* Reset the needs_recovery flag before the fs is unlocked. */
4860 EXT4_SET_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
4861 ext4_commit_super(sb, 1);
4866 * Structure to save mount options for ext4_remount's benefit
4868 struct ext4_mount_options {
4869 unsigned long s_mount_opt;
4870 unsigned long s_mount_opt2;
4873 unsigned long s_commit_interval;
4874 u32 s_min_batch_time, s_max_batch_time;
4877 char *s_qf_names[EXT4_MAXQUOTAS];
4881 static int ext4_remount(struct super_block *sb, int *flags, char *data)
4883 struct ext4_super_block *es;
4884 struct ext4_sb_info *sbi = EXT4_SB(sb);
4885 unsigned long old_sb_flags;
4886 struct ext4_mount_options old_opts;
4887 int enable_quota = 0;
4889 unsigned int journal_ioprio = DEFAULT_JOURNAL_IOPRIO;
4894 char *orig_data = kstrdup(data, GFP_KERNEL);
4896 /* Store the original options */
4897 old_sb_flags = sb->s_flags;
4898 old_opts.s_mount_opt = sbi->s_mount_opt;
4899 old_opts.s_mount_opt2 = sbi->s_mount_opt2;
4900 old_opts.s_resuid = sbi->s_resuid;
4901 old_opts.s_resgid = sbi->s_resgid;
4902 old_opts.s_commit_interval = sbi->s_commit_interval;
4903 old_opts.s_min_batch_time = sbi->s_min_batch_time;
4904 old_opts.s_max_batch_time = sbi->s_max_batch_time;
4906 old_opts.s_jquota_fmt = sbi->s_jquota_fmt;
4907 for (i = 0; i < EXT4_MAXQUOTAS; i++)
4908 if (sbi->s_qf_names[i]) {
4909 old_opts.s_qf_names[i] = kstrdup(sbi->s_qf_names[i],
4911 if (!old_opts.s_qf_names[i]) {
4912 for (j = 0; j < i; j++)
4913 kfree(old_opts.s_qf_names[j]);
4918 old_opts.s_qf_names[i] = NULL;
4920 if (sbi->s_journal && sbi->s_journal->j_task->io_context)
4921 journal_ioprio = sbi->s_journal->j_task->io_context->ioprio;
4923 if (!parse_options(data, sb, NULL, &journal_ioprio, 1)) {
4928 if ((old_opts.s_mount_opt & EXT4_MOUNT_JOURNAL_CHECKSUM) ^
4929 test_opt(sb, JOURNAL_CHECKSUM)) {
4930 ext4_msg(sb, KERN_ERR, "changing journal_checksum "
4931 "during remount not supported; ignoring");
4932 sbi->s_mount_opt ^= EXT4_MOUNT_JOURNAL_CHECKSUM;
4935 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA) {
4936 if (test_opt2(sb, EXPLICIT_DELALLOC)) {
4937 ext4_msg(sb, KERN_ERR, "can't mount with "
4938 "both data=journal and delalloc");
4942 if (test_opt(sb, DIOREAD_NOLOCK)) {
4943 ext4_msg(sb, KERN_ERR, "can't mount with "
4944 "both data=journal and dioread_nolock");
4948 if (test_opt(sb, DAX)) {
4949 ext4_msg(sb, KERN_ERR, "can't mount with "
4950 "both data=journal and dax");
4956 if ((sbi->s_mount_opt ^ old_opts.s_mount_opt) & EXT4_MOUNT_DAX) {
4957 ext4_msg(sb, KERN_WARNING, "warning: refusing change of "
4958 "dax flag with busy inodes while remounting");
4959 sbi->s_mount_opt ^= EXT4_MOUNT_DAX;
4962 if (sbi->s_mount_flags & EXT4_MF_FS_ABORTED)
4963 ext4_abort(sb, "Abort forced by user");
4965 sb->s_flags = (sb->s_flags & ~MS_POSIXACL) |
4966 (test_opt(sb, POSIX_ACL) ? MS_POSIXACL : 0);
4970 if (sbi->s_journal) {
4971 ext4_init_journal_params(sb, sbi->s_journal);
4972 set_task_ioprio(sbi->s_journal->j_task, journal_ioprio);
4975 if (*flags & MS_LAZYTIME)
4976 sb->s_flags |= MS_LAZYTIME;
4978 if ((*flags & MS_RDONLY) != (sb->s_flags & MS_RDONLY)) {
4979 if (sbi->s_mount_flags & EXT4_MF_FS_ABORTED) {
4984 if (*flags & MS_RDONLY) {
4985 err = sync_filesystem(sb);
4988 err = dquot_suspend(sb, -1);
4993 * First of all, the unconditional stuff we have to do
4994 * to disable replay of the journal when we next remount
4996 sb->s_flags |= MS_RDONLY;
4999 * OK, test if we are remounting a valid rw partition
5000 * readonly, and if so set the rdonly flag and then
5001 * mark the partition as valid again.
5003 if (!(es->s_state & cpu_to_le16(EXT4_VALID_FS)) &&
5004 (sbi->s_mount_state & EXT4_VALID_FS))
5005 es->s_state = cpu_to_le16(sbi->s_mount_state);
5008 ext4_mark_recovery_complete(sb, es);
5010 /* Make sure we can mount this feature set readwrite */
5011 if (EXT4_HAS_RO_COMPAT_FEATURE(sb,
5012 EXT4_FEATURE_RO_COMPAT_READONLY) ||
5013 !ext4_feature_set_ok(sb, 0)) {
5018 * Make sure the group descriptor checksums
5019 * are sane. If they aren't, refuse to remount r/w.
5021 for (g = 0; g < sbi->s_groups_count; g++) {
5022 struct ext4_group_desc *gdp =
5023 ext4_get_group_desc(sb, g, NULL);
5025 if (!ext4_group_desc_csum_verify(sb, g, gdp)) {
5026 ext4_msg(sb, KERN_ERR,
5027 "ext4_remount: Checksum for group %u failed (%u!=%u)",
5028 g, le16_to_cpu(ext4_group_desc_csum(sbi, g, gdp)),
5029 le16_to_cpu(gdp->bg_checksum));
5036 * If we have an unprocessed orphan list hanging
5037 * around from a previously readonly bdev mount,
5038 * require a full umount/remount for now.
5040 if (es->s_last_orphan) {
5041 ext4_msg(sb, KERN_WARNING, "Couldn't "
5042 "remount RDWR because of unprocessed "
5043 "orphan inode list. Please "
5044 "umount/remount instead");
5050 * Mounting a RDONLY partition read-write, so reread
5051 * and store the current valid flag. (It may have
5052 * been changed by e2fsck since we originally mounted
5056 ext4_clear_journal_err(sb, es);
5057 sbi->s_mount_state = le16_to_cpu(es->s_state);
5058 if (!ext4_setup_super(sb, es, 0))
5059 sb->s_flags &= ~MS_RDONLY;
5060 if (EXT4_HAS_INCOMPAT_FEATURE(sb,
5061 EXT4_FEATURE_INCOMPAT_MMP))
5062 if (ext4_multi_mount_protect(sb,
5063 le64_to_cpu(es->s_mmp_block))) {
5072 * Reinitialize lazy itable initialization thread based on
5075 if ((sb->s_flags & MS_RDONLY) || !test_opt(sb, INIT_INODE_TABLE))
5076 ext4_unregister_li_request(sb);
5078 ext4_group_t first_not_zeroed;
5079 first_not_zeroed = ext4_has_uninit_itable(sb);
5080 ext4_register_li_request(sb, first_not_zeroed);
5083 ext4_setup_system_zone(sb);
5084 if (sbi->s_journal == NULL && !(old_sb_flags & MS_RDONLY))
5085 ext4_commit_super(sb, 1);
5088 /* Release old quota file names */
5089 for (i = 0; i < EXT4_MAXQUOTAS; i++)
5090 kfree(old_opts.s_qf_names[i]);
5092 if (sb_any_quota_suspended(sb))
5093 dquot_resume(sb, -1);
5094 else if (EXT4_HAS_RO_COMPAT_FEATURE(sb,
5095 EXT4_FEATURE_RO_COMPAT_QUOTA)) {
5096 err = ext4_enable_quotas(sb);
5103 *flags = (*flags & ~MS_LAZYTIME) | (sb->s_flags & MS_LAZYTIME);
5104 ext4_msg(sb, KERN_INFO, "re-mounted. Opts: %s", orig_data);
5109 sb->s_flags = old_sb_flags;
5110 sbi->s_mount_opt = old_opts.s_mount_opt;
5111 sbi->s_mount_opt2 = old_opts.s_mount_opt2;
5112 sbi->s_resuid = old_opts.s_resuid;
5113 sbi->s_resgid = old_opts.s_resgid;
5114 sbi->s_commit_interval = old_opts.s_commit_interval;
5115 sbi->s_min_batch_time = old_opts.s_min_batch_time;
5116 sbi->s_max_batch_time = old_opts.s_max_batch_time;
5118 sbi->s_jquota_fmt = old_opts.s_jquota_fmt;
5119 for (i = 0; i < EXT4_MAXQUOTAS; i++) {
5120 kfree(sbi->s_qf_names[i]);
5121 sbi->s_qf_names[i] = old_opts.s_qf_names[i];
5128 static int ext4_statfs(struct dentry *dentry, struct kstatfs *buf)
5130 struct super_block *sb = dentry->d_sb;
5131 struct ext4_sb_info *sbi = EXT4_SB(sb);
5132 struct ext4_super_block *es = sbi->s_es;
5133 ext4_fsblk_t overhead = 0, resv_blocks;
5136 resv_blocks = EXT4_C2B(sbi, atomic64_read(&sbi->s_resv_clusters));
5138 if (!test_opt(sb, MINIX_DF))
5139 overhead = sbi->s_overhead;
5141 buf->f_type = EXT4_SUPER_MAGIC;
5142 buf->f_bsize = sb->s_blocksize;
5143 buf->f_blocks = ext4_blocks_count(es) - EXT4_C2B(sbi, overhead);
5144 bfree = percpu_counter_sum_positive(&sbi->s_freeclusters_counter) -
5145 percpu_counter_sum_positive(&sbi->s_dirtyclusters_counter);
5146 /* prevent underflow in case that few free space is available */
5147 buf->f_bfree = EXT4_C2B(sbi, max_t(s64, bfree, 0));
5148 buf->f_bavail = buf->f_bfree -
5149 (ext4_r_blocks_count(es) + resv_blocks);
5150 if (buf->f_bfree < (ext4_r_blocks_count(es) + resv_blocks))
5152 buf->f_files = le32_to_cpu(es->s_inodes_count);
5153 buf->f_ffree = percpu_counter_sum_positive(&sbi->s_freeinodes_counter);
5154 buf->f_namelen = EXT4_NAME_LEN;
5155 fsid = le64_to_cpup((void *)es->s_uuid) ^
5156 le64_to_cpup((void *)es->s_uuid + sizeof(u64));
5157 buf->f_fsid.val[0] = fsid & 0xFFFFFFFFUL;
5158 buf->f_fsid.val[1] = (fsid >> 32) & 0xFFFFFFFFUL;
5163 /* Helper function for writing quotas on sync - we need to start transaction
5164 * before quota file is locked for write. Otherwise the are possible deadlocks:
5165 * Process 1 Process 2
5166 * ext4_create() quota_sync()
5167 * jbd2_journal_start() write_dquot()
5168 * dquot_initialize() down(dqio_mutex)
5169 * down(dqio_mutex) jbd2_journal_start()
5175 static inline struct inode *dquot_to_inode(struct dquot *dquot)
5177 return sb_dqopt(dquot->dq_sb)->files[dquot->dq_id.type];
5180 static int ext4_write_dquot(struct dquot *dquot)
5184 struct inode *inode;
5186 inode = dquot_to_inode(dquot);
5187 handle = ext4_journal_start(inode, EXT4_HT_QUOTA,
5188 EXT4_QUOTA_TRANS_BLOCKS(dquot->dq_sb));
5190 return PTR_ERR(handle);
5191 ret = dquot_commit(dquot);
5192 err = ext4_journal_stop(handle);
5198 static int ext4_acquire_dquot(struct dquot *dquot)
5203 handle = ext4_journal_start(dquot_to_inode(dquot), EXT4_HT_QUOTA,
5204 EXT4_QUOTA_INIT_BLOCKS(dquot->dq_sb));
5206 return PTR_ERR(handle);
5207 ret = dquot_acquire(dquot);
5208 err = ext4_journal_stop(handle);
5214 static int ext4_release_dquot(struct dquot *dquot)
5219 handle = ext4_journal_start(dquot_to_inode(dquot), EXT4_HT_QUOTA,
5220 EXT4_QUOTA_DEL_BLOCKS(dquot->dq_sb));
5221 if (IS_ERR(handle)) {
5222 /* Release dquot anyway to avoid endless cycle in dqput() */
5223 dquot_release(dquot);
5224 return PTR_ERR(handle);
5226 ret = dquot_release(dquot);
5227 err = ext4_journal_stop(handle);
5233 static int ext4_mark_dquot_dirty(struct dquot *dquot)
5235 struct super_block *sb = dquot->dq_sb;
5236 struct ext4_sb_info *sbi = EXT4_SB(sb);
5238 /* Are we journaling quotas? */
5239 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_QUOTA) ||
5240 sbi->s_qf_names[USRQUOTA] || sbi->s_qf_names[GRPQUOTA]) {
5241 dquot_mark_dquot_dirty(dquot);
5242 return ext4_write_dquot(dquot);
5244 return dquot_mark_dquot_dirty(dquot);
5248 static int ext4_write_info(struct super_block *sb, int type)
5253 /* Data block + inode block */
5254 handle = ext4_journal_start(d_inode(sb->s_root), EXT4_HT_QUOTA, 2);
5256 return PTR_ERR(handle);
5257 ret = dquot_commit_info(sb, type);
5258 err = ext4_journal_stop(handle);
5265 * Turn on quotas during mount time - we need to find
5266 * the quota file and such...
5268 static int ext4_quota_on_mount(struct super_block *sb, int type)
5270 return dquot_quota_on_mount(sb, EXT4_SB(sb)->s_qf_names[type],
5271 EXT4_SB(sb)->s_jquota_fmt, type);
5275 * Standard function to be called on quota_on
5277 static int ext4_quota_on(struct super_block *sb, int type, int format_id,
5282 if (!test_opt(sb, QUOTA))
5285 /* Quotafile not on the same filesystem? */
5286 if (path->dentry->d_sb != sb)
5288 /* Journaling quota? */
5289 if (EXT4_SB(sb)->s_qf_names[type]) {
5290 /* Quotafile not in fs root? */
5291 if (path->dentry->d_parent != sb->s_root)
5292 ext4_msg(sb, KERN_WARNING,
5293 "Quota file not on filesystem root. "
5294 "Journaled quota will not work");
5298 * When we journal data on quota file, we have to flush journal to see
5299 * all updates to the file when we bypass pagecache...
5301 if (EXT4_SB(sb)->s_journal &&
5302 ext4_should_journal_data(d_inode(path->dentry))) {
5304 * We don't need to lock updates but journal_flush() could
5305 * otherwise be livelocked...
5307 jbd2_journal_lock_updates(EXT4_SB(sb)->s_journal);
5308 err = jbd2_journal_flush(EXT4_SB(sb)->s_journal);
5309 jbd2_journal_unlock_updates(EXT4_SB(sb)->s_journal);
5314 return dquot_quota_on(sb, type, format_id, path);
5317 static int ext4_quota_enable(struct super_block *sb, int type, int format_id,
5321 struct inode *qf_inode;
5322 unsigned long qf_inums[EXT4_MAXQUOTAS] = {
5323 le32_to_cpu(EXT4_SB(sb)->s_es->s_usr_quota_inum),
5324 le32_to_cpu(EXT4_SB(sb)->s_es->s_grp_quota_inum)
5327 BUG_ON(!EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_QUOTA));
5329 if (!qf_inums[type])
5332 qf_inode = ext4_iget(sb, qf_inums[type]);
5333 if (IS_ERR(qf_inode)) {
5334 ext4_error(sb, "Bad quota inode # %lu", qf_inums[type]);
5335 return PTR_ERR(qf_inode);
5338 /* Don't account quota for quota files to avoid recursion */
5339 qf_inode->i_flags |= S_NOQUOTA;
5340 err = dquot_enable(qf_inode, type, format_id, flags);
5346 /* Enable usage tracking for all quota types. */
5347 static int ext4_enable_quotas(struct super_block *sb)
5350 unsigned long qf_inums[EXT4_MAXQUOTAS] = {
5351 le32_to_cpu(EXT4_SB(sb)->s_es->s_usr_quota_inum),
5352 le32_to_cpu(EXT4_SB(sb)->s_es->s_grp_quota_inum)
5355 sb_dqopt(sb)->flags |= DQUOT_QUOTA_SYS_FILE;
5356 for (type = 0; type < EXT4_MAXQUOTAS; type++) {
5357 if (qf_inums[type]) {
5358 err = ext4_quota_enable(sb, type, QFMT_VFS_V1,
5359 DQUOT_USAGE_ENABLED);
5362 "Failed to enable quota tracking "
5363 "(type=%d, err=%d). Please run "
5364 "e2fsck to fix.", type, err);
5372 static int ext4_quota_off(struct super_block *sb, int type)
5374 struct inode *inode = sb_dqopt(sb)->files[type];
5377 /* Force all delayed allocation blocks to be allocated.
5378 * Caller already holds s_umount sem */
5379 if (test_opt(sb, DELALLOC))
5380 sync_filesystem(sb);
5385 /* Update modification times of quota files when userspace can
5386 * start looking at them */
5387 handle = ext4_journal_start(inode, EXT4_HT_QUOTA, 1);
5390 inode->i_mtime = inode->i_ctime = CURRENT_TIME;
5391 ext4_mark_inode_dirty(handle, inode);
5392 ext4_journal_stop(handle);
5395 return dquot_quota_off(sb, type);
5398 /* Read data from quotafile - avoid pagecache and such because we cannot afford
5399 * acquiring the locks... As quota files are never truncated and quota code
5400 * itself serializes the operations (and no one else should touch the files)
5401 * we don't have to be afraid of races */
5402 static ssize_t ext4_quota_read(struct super_block *sb, int type, char *data,
5403 size_t len, loff_t off)
5405 struct inode *inode = sb_dqopt(sb)->files[type];
5406 ext4_lblk_t blk = off >> EXT4_BLOCK_SIZE_BITS(sb);
5407 int offset = off & (sb->s_blocksize - 1);
5410 struct buffer_head *bh;
5411 loff_t i_size = i_size_read(inode);
5415 if (off+len > i_size)
5418 while (toread > 0) {
5419 tocopy = sb->s_blocksize - offset < toread ?
5420 sb->s_blocksize - offset : toread;
5421 bh = ext4_bread(NULL, inode, blk, 0);
5424 if (!bh) /* A hole? */
5425 memset(data, 0, tocopy);
5427 memcpy(data, bh->b_data+offset, tocopy);
5437 /* Write to quotafile (we know the transaction is already started and has
5438 * enough credits) */
5439 static ssize_t ext4_quota_write(struct super_block *sb, int type,
5440 const char *data, size_t len, loff_t off)
5442 struct inode *inode = sb_dqopt(sb)->files[type];
5443 ext4_lblk_t blk = off >> EXT4_BLOCK_SIZE_BITS(sb);
5444 int err, offset = off & (sb->s_blocksize - 1);
5446 struct buffer_head *bh;
5447 handle_t *handle = journal_current_handle();
5449 if (EXT4_SB(sb)->s_journal && !handle) {
5450 ext4_msg(sb, KERN_WARNING, "Quota write (off=%llu, len=%llu)"
5451 " cancelled because transaction is not started",
5452 (unsigned long long)off, (unsigned long long)len);
5456 * Since we account only one data block in transaction credits,
5457 * then it is impossible to cross a block boundary.
5459 if (sb->s_blocksize - offset < len) {
5460 ext4_msg(sb, KERN_WARNING, "Quota write (off=%llu, len=%llu)"
5461 " cancelled because not block aligned",
5462 (unsigned long long)off, (unsigned long long)len);
5467 bh = ext4_bread(handle, inode, blk,
5468 EXT4_GET_BLOCKS_CREATE |
5469 EXT4_GET_BLOCKS_METADATA_NOFAIL);
5470 } while (IS_ERR(bh) && (PTR_ERR(bh) == -ENOSPC) &&
5471 ext4_should_retry_alloc(inode->i_sb, &retries));
5476 BUFFER_TRACE(bh, "get write access");
5477 err = ext4_journal_get_write_access(handle, bh);
5483 memcpy(bh->b_data+offset, data, len);
5484 flush_dcache_page(bh->b_page);
5486 err = ext4_handle_dirty_metadata(handle, NULL, bh);
5489 if (inode->i_size < off + len) {
5490 i_size_write(inode, off + len);
5491 EXT4_I(inode)->i_disksize = inode->i_size;
5492 ext4_mark_inode_dirty(handle, inode);
5499 static struct dentry *ext4_mount(struct file_system_type *fs_type, int flags,
5500 const char *dev_name, void *data)
5502 return mount_bdev(fs_type, flags, dev_name, data, ext4_fill_super);
5505 #if !defined(CONFIG_EXT2_FS) && !defined(CONFIG_EXT2_FS_MODULE) && defined(CONFIG_EXT4_USE_FOR_EXT23)
5506 static inline void register_as_ext2(void)
5508 int err = register_filesystem(&ext2_fs_type);
5511 "EXT4-fs: Unable to register as ext2 (%d)\n", err);
5514 static inline void unregister_as_ext2(void)
5516 unregister_filesystem(&ext2_fs_type);
5519 static inline int ext2_feature_set_ok(struct super_block *sb)
5521 if (EXT4_HAS_INCOMPAT_FEATURE(sb, ~EXT2_FEATURE_INCOMPAT_SUPP))
5523 if (sb->s_flags & MS_RDONLY)
5525 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, ~EXT2_FEATURE_RO_COMPAT_SUPP))
5530 static inline void register_as_ext2(void) { }
5531 static inline void unregister_as_ext2(void) { }
5532 static inline int ext2_feature_set_ok(struct super_block *sb) { return 0; }
5535 #if !defined(CONFIG_EXT3_FS) && !defined(CONFIG_EXT3_FS_MODULE) && defined(CONFIG_EXT4_USE_FOR_EXT23)
5536 static inline void register_as_ext3(void)
5538 int err = register_filesystem(&ext3_fs_type);
5541 "EXT4-fs: Unable to register as ext3 (%d)\n", err);
5544 static inline void unregister_as_ext3(void)
5546 unregister_filesystem(&ext3_fs_type);
5549 static inline int ext3_feature_set_ok(struct super_block *sb)
5551 if (EXT4_HAS_INCOMPAT_FEATURE(sb, ~EXT3_FEATURE_INCOMPAT_SUPP))
5553 if (!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL))
5555 if (sb->s_flags & MS_RDONLY)
5557 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, ~EXT3_FEATURE_RO_COMPAT_SUPP))
5562 static inline void register_as_ext3(void) { }
5563 static inline void unregister_as_ext3(void) { }
5564 static inline int ext3_feature_set_ok(struct super_block *sb) { return 0; }
5567 static struct file_system_type ext4_fs_type = {
5568 .owner = THIS_MODULE,
5570 .mount = ext4_mount,
5571 .kill_sb = kill_block_super,
5572 .fs_flags = FS_REQUIRES_DEV,
5574 MODULE_ALIAS_FS("ext4");
5576 static int __init ext4_init_feat_adverts(void)
5578 struct ext4_features *ef;
5581 ef = kzalloc(sizeof(struct ext4_features), GFP_KERNEL);
5585 ef->f_kobj.kset = ext4_kset;
5586 init_completion(&ef->f_kobj_unregister);
5587 ret = kobject_init_and_add(&ef->f_kobj, &ext4_feat_ktype, NULL,
5600 static void ext4_exit_feat_adverts(void)
5602 kobject_put(&ext4_feat->f_kobj);
5603 wait_for_completion(&ext4_feat->f_kobj_unregister);
5607 /* Shared across all ext4 file systems */
5608 wait_queue_head_t ext4__ioend_wq[EXT4_WQ_HASH_SZ];
5609 struct mutex ext4__aio_mutex[EXT4_WQ_HASH_SZ];
5611 static int __init ext4_init_fs(void)
5615 ext4_li_info = NULL;
5616 mutex_init(&ext4_li_mtx);
5618 /* Build-time check for flags consistency */
5619 ext4_check_flag_values();
5621 for (i = 0; i < EXT4_WQ_HASH_SZ; i++) {
5622 mutex_init(&ext4__aio_mutex[i]);
5623 init_waitqueue_head(&ext4__ioend_wq[i]);
5626 err = ext4_init_es();
5630 err = ext4_init_pageio();
5634 err = ext4_init_system_zone();
5637 ext4_kset = kset_create_and_add("ext4", NULL, fs_kobj);
5642 ext4_proc_root = proc_mkdir("fs/ext4", NULL);
5644 err = ext4_init_feat_adverts();
5648 err = ext4_init_mballoc();
5652 ext4_mballoc_ready = 1;
5653 err = init_inodecache();
5658 err = register_filesystem(&ext4_fs_type);
5664 unregister_as_ext2();
5665 unregister_as_ext3();
5666 destroy_inodecache();
5668 ext4_mballoc_ready = 0;
5669 ext4_exit_mballoc();
5671 ext4_exit_feat_adverts();
5674 remove_proc_entry("fs/ext4", NULL);
5675 kset_unregister(ext4_kset);
5677 ext4_exit_system_zone();
5686 static void __exit ext4_exit_fs(void)
5689 ext4_destroy_lazyinit_thread();
5690 unregister_as_ext2();
5691 unregister_as_ext3();
5692 unregister_filesystem(&ext4_fs_type);
5693 destroy_inodecache();
5694 ext4_exit_mballoc();
5695 ext4_exit_feat_adverts();
5696 remove_proc_entry("fs/ext4", NULL);
5697 kset_unregister(ext4_kset);
5698 ext4_exit_system_zone();
5703 MODULE_AUTHOR("Remy Card, Stephen Tweedie, Andrew Morton, Andreas Dilger, Theodore Ts'o and others");
5704 MODULE_DESCRIPTION("Fourth Extended Filesystem");
5705 MODULE_LICENSE("GPL");
5706 module_init(ext4_init_fs)
5707 module_exit(ext4_exit_fs)