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/ctype.h>
38 #include <linux/log2.h>
39 #include <linux/crc16.h>
40 #include <linux/cleancache.h>
41 #include <asm/uaccess.h>
43 #include <linux/kthread.h>
44 #include <linux/freezer.h>
47 #include "ext4_extents.h" /* Needed for trace points definition */
48 #include "ext4_jbd2.h"
53 #define CREATE_TRACE_POINTS
54 #include <trace/events/ext4.h>
56 static struct ext4_lazy_init *ext4_li_info;
57 static struct mutex ext4_li_mtx;
58 static int ext4_mballoc_ready;
59 static struct ratelimit_state ext4_mount_msg_ratelimit;
61 static int ext4_load_journal(struct super_block *, struct ext4_super_block *,
62 unsigned long journal_devnum);
63 static int ext4_show_options(struct seq_file *seq, struct dentry *root);
64 static int ext4_commit_super(struct super_block *sb, int sync);
65 static void ext4_mark_recovery_complete(struct super_block *sb,
66 struct ext4_super_block *es);
67 static void ext4_clear_journal_err(struct super_block *sb,
68 struct ext4_super_block *es);
69 static int ext4_sync_fs(struct super_block *sb, int wait);
70 static int ext4_remount(struct super_block *sb, int *flags, char *data);
71 static int ext4_statfs(struct dentry *dentry, struct kstatfs *buf);
72 static int ext4_unfreeze(struct super_block *sb);
73 static int ext4_freeze(struct super_block *sb);
74 static struct dentry *ext4_mount(struct file_system_type *fs_type, int flags,
75 const char *dev_name, void *data);
76 static inline int ext2_feature_set_ok(struct super_block *sb);
77 static inline int ext3_feature_set_ok(struct super_block *sb);
78 static int ext4_feature_set_ok(struct super_block *sb, int readonly);
79 static void ext4_destroy_lazyinit_thread(void);
80 static void ext4_unregister_li_request(struct super_block *sb);
81 static void ext4_clear_request_list(void);
83 #if !defined(CONFIG_EXT2_FS) && !defined(CONFIG_EXT2_FS_MODULE) && defined(CONFIG_EXT4_USE_FOR_EXT2)
84 static struct file_system_type ext2_fs_type = {
88 .kill_sb = kill_block_super,
89 .fs_flags = FS_REQUIRES_DEV,
91 MODULE_ALIAS_FS("ext2");
93 #define IS_EXT2_SB(sb) ((sb)->s_bdev->bd_holder == &ext2_fs_type)
95 #define IS_EXT2_SB(sb) (0)
99 static struct file_system_type ext3_fs_type = {
100 .owner = THIS_MODULE,
103 .kill_sb = kill_block_super,
104 .fs_flags = FS_REQUIRES_DEV,
106 MODULE_ALIAS_FS("ext3");
107 MODULE_ALIAS("ext3");
108 #define IS_EXT3_SB(sb) ((sb)->s_bdev->bd_holder == &ext3_fs_type)
110 static int ext4_verify_csum_type(struct super_block *sb,
111 struct ext4_super_block *es)
113 if (!EXT4_HAS_RO_COMPAT_FEATURE(sb,
114 EXT4_FEATURE_RO_COMPAT_METADATA_CSUM))
117 return es->s_checksum_type == EXT4_CRC32C_CHKSUM;
120 static __le32 ext4_superblock_csum(struct super_block *sb,
121 struct ext4_super_block *es)
123 struct ext4_sb_info *sbi = EXT4_SB(sb);
124 int offset = offsetof(struct ext4_super_block, s_checksum);
127 csum = ext4_chksum(sbi, ~0, (char *)es, offset);
129 return cpu_to_le32(csum);
132 static int ext4_superblock_csum_verify(struct super_block *sb,
133 struct ext4_super_block *es)
135 if (!ext4_has_metadata_csum(sb))
138 return es->s_checksum == ext4_superblock_csum(sb, es);
141 void ext4_superblock_csum_set(struct super_block *sb)
143 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
145 if (!ext4_has_metadata_csum(sb))
148 es->s_checksum = ext4_superblock_csum(sb, es);
151 void *ext4_kvmalloc(size_t size, gfp_t flags)
155 ret = kmalloc(size, flags | __GFP_NOWARN);
157 ret = __vmalloc(size, flags, PAGE_KERNEL);
161 void *ext4_kvzalloc(size_t size, gfp_t flags)
165 ret = kzalloc(size, flags | __GFP_NOWARN);
167 ret = __vmalloc(size, flags | __GFP_ZERO, PAGE_KERNEL);
171 ext4_fsblk_t ext4_block_bitmap(struct super_block *sb,
172 struct ext4_group_desc *bg)
174 return le32_to_cpu(bg->bg_block_bitmap_lo) |
175 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
176 (ext4_fsblk_t)le32_to_cpu(bg->bg_block_bitmap_hi) << 32 : 0);
179 ext4_fsblk_t ext4_inode_bitmap(struct super_block *sb,
180 struct ext4_group_desc *bg)
182 return le32_to_cpu(bg->bg_inode_bitmap_lo) |
183 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
184 (ext4_fsblk_t)le32_to_cpu(bg->bg_inode_bitmap_hi) << 32 : 0);
187 ext4_fsblk_t ext4_inode_table(struct super_block *sb,
188 struct ext4_group_desc *bg)
190 return le32_to_cpu(bg->bg_inode_table_lo) |
191 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
192 (ext4_fsblk_t)le32_to_cpu(bg->bg_inode_table_hi) << 32 : 0);
195 __u32 ext4_free_group_clusters(struct super_block *sb,
196 struct ext4_group_desc *bg)
198 return le16_to_cpu(bg->bg_free_blocks_count_lo) |
199 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
200 (__u32)le16_to_cpu(bg->bg_free_blocks_count_hi) << 16 : 0);
203 __u32 ext4_free_inodes_count(struct super_block *sb,
204 struct ext4_group_desc *bg)
206 return le16_to_cpu(bg->bg_free_inodes_count_lo) |
207 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
208 (__u32)le16_to_cpu(bg->bg_free_inodes_count_hi) << 16 : 0);
211 __u32 ext4_used_dirs_count(struct super_block *sb,
212 struct ext4_group_desc *bg)
214 return le16_to_cpu(bg->bg_used_dirs_count_lo) |
215 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
216 (__u32)le16_to_cpu(bg->bg_used_dirs_count_hi) << 16 : 0);
219 __u32 ext4_itable_unused_count(struct super_block *sb,
220 struct ext4_group_desc *bg)
222 return le16_to_cpu(bg->bg_itable_unused_lo) |
223 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
224 (__u32)le16_to_cpu(bg->bg_itable_unused_hi) << 16 : 0);
227 void ext4_block_bitmap_set(struct super_block *sb,
228 struct ext4_group_desc *bg, ext4_fsblk_t blk)
230 bg->bg_block_bitmap_lo = cpu_to_le32((u32)blk);
231 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
232 bg->bg_block_bitmap_hi = cpu_to_le32(blk >> 32);
235 void ext4_inode_bitmap_set(struct super_block *sb,
236 struct ext4_group_desc *bg, ext4_fsblk_t blk)
238 bg->bg_inode_bitmap_lo = cpu_to_le32((u32)blk);
239 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
240 bg->bg_inode_bitmap_hi = cpu_to_le32(blk >> 32);
243 void ext4_inode_table_set(struct super_block *sb,
244 struct ext4_group_desc *bg, ext4_fsblk_t blk)
246 bg->bg_inode_table_lo = cpu_to_le32((u32)blk);
247 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
248 bg->bg_inode_table_hi = cpu_to_le32(blk >> 32);
251 void ext4_free_group_clusters_set(struct super_block *sb,
252 struct ext4_group_desc *bg, __u32 count)
254 bg->bg_free_blocks_count_lo = cpu_to_le16((__u16)count);
255 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
256 bg->bg_free_blocks_count_hi = cpu_to_le16(count >> 16);
259 void ext4_free_inodes_set(struct super_block *sb,
260 struct ext4_group_desc *bg, __u32 count)
262 bg->bg_free_inodes_count_lo = cpu_to_le16((__u16)count);
263 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
264 bg->bg_free_inodes_count_hi = cpu_to_le16(count >> 16);
267 void ext4_used_dirs_set(struct super_block *sb,
268 struct ext4_group_desc *bg, __u32 count)
270 bg->bg_used_dirs_count_lo = cpu_to_le16((__u16)count);
271 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
272 bg->bg_used_dirs_count_hi = cpu_to_le16(count >> 16);
275 void ext4_itable_unused_set(struct super_block *sb,
276 struct ext4_group_desc *bg, __u32 count)
278 bg->bg_itable_unused_lo = cpu_to_le16((__u16)count);
279 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
280 bg->bg_itable_unused_hi = cpu_to_le16(count >> 16);
284 static void __save_error_info(struct super_block *sb, const char *func,
287 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
289 EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS;
290 if (bdev_read_only(sb->s_bdev))
292 es->s_state |= cpu_to_le16(EXT4_ERROR_FS);
293 es->s_last_error_time = cpu_to_le32(get_seconds());
294 strncpy(es->s_last_error_func, func, sizeof(es->s_last_error_func));
295 es->s_last_error_line = cpu_to_le32(line);
296 if (!es->s_first_error_time) {
297 es->s_first_error_time = es->s_last_error_time;
298 strncpy(es->s_first_error_func, func,
299 sizeof(es->s_first_error_func));
300 es->s_first_error_line = cpu_to_le32(line);
301 es->s_first_error_ino = es->s_last_error_ino;
302 es->s_first_error_block = es->s_last_error_block;
305 * Start the daily error reporting function if it hasn't been
308 if (!es->s_error_count)
309 mod_timer(&EXT4_SB(sb)->s_err_report, jiffies + 24*60*60*HZ);
310 le32_add_cpu(&es->s_error_count, 1);
313 static void save_error_info(struct super_block *sb, const char *func,
316 __save_error_info(sb, func, line);
317 ext4_commit_super(sb, 1);
321 * The del_gendisk() function uninitializes the disk-specific data
322 * structures, including the bdi structure, without telling anyone
323 * else. Once this happens, any attempt to call mark_buffer_dirty()
324 * (for example, by ext4_commit_super), will cause a kernel OOPS.
325 * This is a kludge to prevent these oops until we can put in a proper
326 * hook in del_gendisk() to inform the VFS and file system layers.
328 static int block_device_ejected(struct super_block *sb)
330 struct inode *bd_inode = sb->s_bdev->bd_inode;
331 struct backing_dev_info *bdi = inode_to_bdi(bd_inode);
333 return bdi->dev == NULL;
336 static void ext4_journal_commit_callback(journal_t *journal, transaction_t *txn)
338 struct super_block *sb = journal->j_private;
339 struct ext4_sb_info *sbi = EXT4_SB(sb);
340 int error = is_journal_aborted(journal);
341 struct ext4_journal_cb_entry *jce;
343 BUG_ON(txn->t_state == T_FINISHED);
344 spin_lock(&sbi->s_md_lock);
345 while (!list_empty(&txn->t_private_list)) {
346 jce = list_entry(txn->t_private_list.next,
347 struct ext4_journal_cb_entry, jce_list);
348 list_del_init(&jce->jce_list);
349 spin_unlock(&sbi->s_md_lock);
350 jce->jce_func(sb, jce, error);
351 spin_lock(&sbi->s_md_lock);
353 spin_unlock(&sbi->s_md_lock);
356 /* Deal with the reporting of failure conditions on a filesystem such as
357 * inconsistencies detected or read IO failures.
359 * On ext2, we can store the error state of the filesystem in the
360 * superblock. That is not possible on ext4, because we may have other
361 * write ordering constraints on the superblock which prevent us from
362 * writing it out straight away; and given that the journal is about to
363 * be aborted, we can't rely on the current, or future, transactions to
364 * write out the superblock safely.
366 * We'll just use the jbd2_journal_abort() error code to record an error in
367 * the journal instead. On recovery, the journal will complain about
368 * that error until we've noted it down and cleared it.
371 static void ext4_handle_error(struct super_block *sb)
373 if (sb->s_flags & MS_RDONLY)
376 if (!test_opt(sb, ERRORS_CONT)) {
377 journal_t *journal = EXT4_SB(sb)->s_journal;
379 EXT4_SB(sb)->s_mount_flags |= EXT4_MF_FS_ABORTED;
381 jbd2_journal_abort(journal, -EIO);
383 if (test_opt(sb, ERRORS_RO)) {
384 ext4_msg(sb, KERN_CRIT, "Remounting filesystem read-only");
386 * Make sure updated value of ->s_mount_flags will be visible
387 * before ->s_flags update
390 sb->s_flags |= MS_RDONLY;
392 if (test_opt(sb, ERRORS_PANIC))
393 panic("EXT4-fs (device %s): panic forced after error\n",
397 #define ext4_error_ratelimit(sb) \
398 ___ratelimit(&(EXT4_SB(sb)->s_err_ratelimit_state), \
401 void __ext4_error(struct super_block *sb, const char *function,
402 unsigned int line, const char *fmt, ...)
404 struct va_format vaf;
407 if (ext4_error_ratelimit(sb)) {
412 "EXT4-fs error (device %s): %s:%d: comm %s: %pV\n",
413 sb->s_id, function, line, current->comm, &vaf);
416 save_error_info(sb, function, line);
417 ext4_handle_error(sb);
420 void __ext4_error_inode(struct inode *inode, const char *function,
421 unsigned int line, ext4_fsblk_t block,
422 const char *fmt, ...)
425 struct va_format vaf;
426 struct ext4_super_block *es = EXT4_SB(inode->i_sb)->s_es;
428 es->s_last_error_ino = cpu_to_le32(inode->i_ino);
429 es->s_last_error_block = cpu_to_le64(block);
430 if (ext4_error_ratelimit(inode->i_sb)) {
435 printk(KERN_CRIT "EXT4-fs error (device %s): %s:%d: "
436 "inode #%lu: block %llu: comm %s: %pV\n",
437 inode->i_sb->s_id, function, line, inode->i_ino,
438 block, current->comm, &vaf);
440 printk(KERN_CRIT "EXT4-fs error (device %s): %s:%d: "
441 "inode #%lu: comm %s: %pV\n",
442 inode->i_sb->s_id, function, line, inode->i_ino,
443 current->comm, &vaf);
446 save_error_info(inode->i_sb, function, line);
447 ext4_handle_error(inode->i_sb);
450 void __ext4_error_file(struct file *file, const char *function,
451 unsigned int line, ext4_fsblk_t block,
452 const char *fmt, ...)
455 struct va_format vaf;
456 struct ext4_super_block *es;
457 struct inode *inode = file_inode(file);
458 char pathname[80], *path;
460 es = EXT4_SB(inode->i_sb)->s_es;
461 es->s_last_error_ino = cpu_to_le32(inode->i_ino);
462 if (ext4_error_ratelimit(inode->i_sb)) {
463 path = file_path(file, pathname, sizeof(pathname));
471 "EXT4-fs error (device %s): %s:%d: inode #%lu: "
472 "block %llu: comm %s: path %s: %pV\n",
473 inode->i_sb->s_id, function, line, inode->i_ino,
474 block, current->comm, path, &vaf);
477 "EXT4-fs error (device %s): %s:%d: inode #%lu: "
478 "comm %s: path %s: %pV\n",
479 inode->i_sb->s_id, function, line, inode->i_ino,
480 current->comm, path, &vaf);
483 save_error_info(inode->i_sb, function, line);
484 ext4_handle_error(inode->i_sb);
487 const char *ext4_decode_error(struct super_block *sb, int errno,
494 errstr = "IO failure";
497 errstr = "Out of memory";
500 if (!sb || (EXT4_SB(sb)->s_journal &&
501 EXT4_SB(sb)->s_journal->j_flags & JBD2_ABORT))
502 errstr = "Journal has aborted";
504 errstr = "Readonly filesystem";
507 /* If the caller passed in an extra buffer for unknown
508 * errors, textualise them now. Else we just return
511 /* Check for truncated error codes... */
512 if (snprintf(nbuf, 16, "error %d", -errno) >= 0)
521 /* __ext4_std_error decodes expected errors from journaling functions
522 * automatically and invokes the appropriate error response. */
524 void __ext4_std_error(struct super_block *sb, const char *function,
525 unsigned int line, int errno)
530 /* Special case: if the error is EROFS, and we're not already
531 * inside a transaction, then there's really no point in logging
533 if (errno == -EROFS && journal_current_handle() == NULL &&
534 (sb->s_flags & MS_RDONLY))
537 if (ext4_error_ratelimit(sb)) {
538 errstr = ext4_decode_error(sb, errno, nbuf);
539 printk(KERN_CRIT "EXT4-fs error (device %s) in %s:%d: %s\n",
540 sb->s_id, function, line, errstr);
543 save_error_info(sb, function, line);
544 ext4_handle_error(sb);
548 * ext4_abort is a much stronger failure handler than ext4_error. The
549 * abort function may be used to deal with unrecoverable failures such
550 * as journal IO errors or ENOMEM at a critical moment in log management.
552 * We unconditionally force the filesystem into an ABORT|READONLY state,
553 * unless the error response on the fs has been set to panic in which
554 * case we take the easy way out and panic immediately.
557 void __ext4_abort(struct super_block *sb, const char *function,
558 unsigned int line, const char *fmt, ...)
562 save_error_info(sb, function, line);
564 printk(KERN_CRIT "EXT4-fs error (device %s): %s:%d: ", sb->s_id,
570 if ((sb->s_flags & MS_RDONLY) == 0) {
571 ext4_msg(sb, KERN_CRIT, "Remounting filesystem read-only");
572 EXT4_SB(sb)->s_mount_flags |= EXT4_MF_FS_ABORTED;
574 * Make sure updated value of ->s_mount_flags will be visible
575 * before ->s_flags update
578 sb->s_flags |= MS_RDONLY;
579 if (EXT4_SB(sb)->s_journal)
580 jbd2_journal_abort(EXT4_SB(sb)->s_journal, -EIO);
581 save_error_info(sb, function, line);
583 if (test_opt(sb, ERRORS_PANIC))
584 panic("EXT4-fs panic from previous error\n");
587 void __ext4_msg(struct super_block *sb,
588 const char *prefix, const char *fmt, ...)
590 struct va_format vaf;
593 if (!___ratelimit(&(EXT4_SB(sb)->s_msg_ratelimit_state), "EXT4-fs"))
599 printk("%sEXT4-fs (%s): %pV\n", prefix, sb->s_id, &vaf);
603 #define ext4_warning_ratelimit(sb) \
604 ___ratelimit(&(EXT4_SB(sb)->s_warning_ratelimit_state), \
607 void __ext4_warning(struct super_block *sb, const char *function,
608 unsigned int line, const char *fmt, ...)
610 struct va_format vaf;
613 if (!ext4_warning_ratelimit(sb))
619 printk(KERN_WARNING "EXT4-fs warning (device %s): %s:%d: %pV\n",
620 sb->s_id, function, line, &vaf);
624 void __ext4_warning_inode(const struct inode *inode, const char *function,
625 unsigned int line, const char *fmt, ...)
627 struct va_format vaf;
630 if (!ext4_warning_ratelimit(inode->i_sb))
636 printk(KERN_WARNING "EXT4-fs warning (device %s): %s:%d: "
637 "inode #%lu: comm %s: %pV\n", inode->i_sb->s_id,
638 function, line, inode->i_ino, current->comm, &vaf);
642 void __ext4_grp_locked_error(const char *function, unsigned int line,
643 struct super_block *sb, ext4_group_t grp,
644 unsigned long ino, ext4_fsblk_t block,
645 const char *fmt, ...)
649 struct va_format vaf;
651 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
653 es->s_last_error_ino = cpu_to_le32(ino);
654 es->s_last_error_block = cpu_to_le64(block);
655 __save_error_info(sb, function, line);
657 if (ext4_error_ratelimit(sb)) {
661 printk(KERN_CRIT "EXT4-fs error (device %s): %s:%d: group %u, ",
662 sb->s_id, function, line, grp);
664 printk(KERN_CONT "inode %lu: ", ino);
666 printk(KERN_CONT "block %llu:",
667 (unsigned long long) block);
668 printk(KERN_CONT "%pV\n", &vaf);
672 if (test_opt(sb, ERRORS_CONT)) {
673 ext4_commit_super(sb, 0);
677 ext4_unlock_group(sb, grp);
678 ext4_handle_error(sb);
680 * We only get here in the ERRORS_RO case; relocking the group
681 * may be dangerous, but nothing bad will happen since the
682 * filesystem will have already been marked read/only and the
683 * journal has been aborted. We return 1 as a hint to callers
684 * who might what to use the return value from
685 * ext4_grp_locked_error() to distinguish between the
686 * ERRORS_CONT and ERRORS_RO case, and perhaps return more
687 * aggressively from the ext4 function in question, with a
688 * more appropriate error code.
690 ext4_lock_group(sb, grp);
694 void ext4_update_dynamic_rev(struct super_block *sb)
696 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
698 if (le32_to_cpu(es->s_rev_level) > EXT4_GOOD_OLD_REV)
702 "updating to rev %d because of new feature flag, "
703 "running e2fsck is recommended",
706 es->s_first_ino = cpu_to_le32(EXT4_GOOD_OLD_FIRST_INO);
707 es->s_inode_size = cpu_to_le16(EXT4_GOOD_OLD_INODE_SIZE);
708 es->s_rev_level = cpu_to_le32(EXT4_DYNAMIC_REV);
709 /* leave es->s_feature_*compat flags alone */
710 /* es->s_uuid will be set by e2fsck if empty */
713 * The rest of the superblock fields should be zero, and if not it
714 * means they are likely already in use, so leave them alone. We
715 * can leave it up to e2fsck to clean up any inconsistencies there.
720 * Open the external journal device
722 static struct block_device *ext4_blkdev_get(dev_t dev, struct super_block *sb)
724 struct block_device *bdev;
725 char b[BDEVNAME_SIZE];
727 bdev = blkdev_get_by_dev(dev, FMODE_READ|FMODE_WRITE|FMODE_EXCL, sb);
733 ext4_msg(sb, KERN_ERR, "failed to open journal device %s: %ld",
734 __bdevname(dev, b), PTR_ERR(bdev));
739 * Release the journal device
741 static void ext4_blkdev_put(struct block_device *bdev)
743 blkdev_put(bdev, FMODE_READ|FMODE_WRITE|FMODE_EXCL);
746 static void ext4_blkdev_remove(struct ext4_sb_info *sbi)
748 struct block_device *bdev;
749 bdev = sbi->journal_bdev;
751 ext4_blkdev_put(bdev);
752 sbi->journal_bdev = NULL;
756 static inline struct inode *orphan_list_entry(struct list_head *l)
758 return &list_entry(l, struct ext4_inode_info, i_orphan)->vfs_inode;
761 static void dump_orphan_list(struct super_block *sb, struct ext4_sb_info *sbi)
765 ext4_msg(sb, KERN_ERR, "sb orphan head is %d",
766 le32_to_cpu(sbi->s_es->s_last_orphan));
768 printk(KERN_ERR "sb_info orphan list:\n");
769 list_for_each(l, &sbi->s_orphan) {
770 struct inode *inode = orphan_list_entry(l);
772 "inode %s:%lu at %p: mode %o, nlink %d, next %d\n",
773 inode->i_sb->s_id, inode->i_ino, inode,
774 inode->i_mode, inode->i_nlink,
779 static void ext4_put_super(struct super_block *sb)
781 struct ext4_sb_info *sbi = EXT4_SB(sb);
782 struct ext4_super_block *es = sbi->s_es;
785 ext4_unregister_li_request(sb);
786 dquot_disable(sb, -1, DQUOT_USAGE_ENABLED | DQUOT_LIMITS_ENABLED);
788 flush_workqueue(sbi->rsv_conversion_wq);
789 destroy_workqueue(sbi->rsv_conversion_wq);
791 if (sbi->s_journal) {
792 err = jbd2_journal_destroy(sbi->s_journal);
793 sbi->s_journal = NULL;
795 ext4_abort(sb, "Couldn't clean up the journal");
798 ext4_unregister_sysfs(sb);
799 ext4_es_unregister_shrinker(sbi);
800 del_timer_sync(&sbi->s_err_report);
801 ext4_release_system_zone(sb);
803 ext4_ext_release(sb);
804 ext4_xattr_put_super(sb);
806 if (!(sb->s_flags & MS_RDONLY)) {
807 EXT4_CLEAR_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
808 es->s_state = cpu_to_le16(sbi->s_mount_state);
810 if (!(sb->s_flags & MS_RDONLY))
811 ext4_commit_super(sb, 1);
813 for (i = 0; i < sbi->s_gdb_count; i++)
814 brelse(sbi->s_group_desc[i]);
815 kvfree(sbi->s_group_desc);
816 kvfree(sbi->s_flex_groups);
817 percpu_counter_destroy(&sbi->s_freeclusters_counter);
818 percpu_counter_destroy(&sbi->s_freeinodes_counter);
819 percpu_counter_destroy(&sbi->s_dirs_counter);
820 percpu_counter_destroy(&sbi->s_dirtyclusters_counter);
823 for (i = 0; i < EXT4_MAXQUOTAS; i++)
824 kfree(sbi->s_qf_names[i]);
827 /* Debugging code just in case the in-memory inode orphan list
828 * isn't empty. The on-disk one can be non-empty if we've
829 * detected an error and taken the fs readonly, but the
830 * in-memory list had better be clean by this point. */
831 if (!list_empty(&sbi->s_orphan))
832 dump_orphan_list(sb, sbi);
833 J_ASSERT(list_empty(&sbi->s_orphan));
835 sync_blockdev(sb->s_bdev);
836 invalidate_bdev(sb->s_bdev);
837 if (sbi->journal_bdev && sbi->journal_bdev != sb->s_bdev) {
839 * Invalidate the journal device's buffers. We don't want them
840 * floating about in memory - the physical journal device may
841 * hotswapped, and it breaks the `ro-after' testing code.
843 sync_blockdev(sbi->journal_bdev);
844 invalidate_bdev(sbi->journal_bdev);
845 ext4_blkdev_remove(sbi);
847 if (sbi->s_mb_cache) {
848 ext4_xattr_destroy_cache(sbi->s_mb_cache);
849 sbi->s_mb_cache = NULL;
852 kthread_stop(sbi->s_mmp_tsk);
853 sb->s_fs_info = NULL;
855 * Now that we are completely done shutting down the
856 * superblock, we need to actually destroy the kobject.
858 kobject_put(&sbi->s_kobj);
859 wait_for_completion(&sbi->s_kobj_unregister);
860 if (sbi->s_chksum_driver)
861 crypto_free_shash(sbi->s_chksum_driver);
862 kfree(sbi->s_blockgroup_lock);
866 static struct kmem_cache *ext4_inode_cachep;
869 * Called inside transaction, so use GFP_NOFS
871 static struct inode *ext4_alloc_inode(struct super_block *sb)
873 struct ext4_inode_info *ei;
875 ei = kmem_cache_alloc(ext4_inode_cachep, GFP_NOFS);
879 ei->vfs_inode.i_version = 1;
880 spin_lock_init(&ei->i_raw_lock);
881 INIT_LIST_HEAD(&ei->i_prealloc_list);
882 spin_lock_init(&ei->i_prealloc_lock);
883 ext4_es_init_tree(&ei->i_es_tree);
884 rwlock_init(&ei->i_es_lock);
885 INIT_LIST_HEAD(&ei->i_es_list);
888 ei->i_es_shrink_lblk = 0;
889 ei->i_reserved_data_blocks = 0;
890 ei->i_reserved_meta_blocks = 0;
891 ei->i_allocated_meta_blocks = 0;
892 ei->i_da_metadata_calc_len = 0;
893 ei->i_da_metadata_calc_last_lblock = 0;
894 spin_lock_init(&(ei->i_block_reservation_lock));
896 ei->i_reserved_quota = 0;
897 memset(&ei->i_dquot, 0, sizeof(ei->i_dquot));
900 INIT_LIST_HEAD(&ei->i_rsv_conversion_list);
901 spin_lock_init(&ei->i_completed_io_lock);
903 ei->i_datasync_tid = 0;
904 atomic_set(&ei->i_ioend_count, 0);
905 atomic_set(&ei->i_unwritten, 0);
906 INIT_WORK(&ei->i_rsv_conversion_work, ext4_end_io_rsv_work);
907 #ifdef CONFIG_EXT4_FS_ENCRYPTION
908 ei->i_crypt_info = NULL;
910 return &ei->vfs_inode;
913 static int ext4_drop_inode(struct inode *inode)
915 int drop = generic_drop_inode(inode);
917 trace_ext4_drop_inode(inode, drop);
921 static void ext4_i_callback(struct rcu_head *head)
923 struct inode *inode = container_of(head, struct inode, i_rcu);
924 kmem_cache_free(ext4_inode_cachep, EXT4_I(inode));
927 static void ext4_destroy_inode(struct inode *inode)
929 if (!list_empty(&(EXT4_I(inode)->i_orphan))) {
930 ext4_msg(inode->i_sb, KERN_ERR,
931 "Inode %lu (%p): orphan list check failed!",
932 inode->i_ino, EXT4_I(inode));
933 print_hex_dump(KERN_INFO, "", DUMP_PREFIX_ADDRESS, 16, 4,
934 EXT4_I(inode), sizeof(struct ext4_inode_info),
938 call_rcu(&inode->i_rcu, ext4_i_callback);
941 static void init_once(void *foo)
943 struct ext4_inode_info *ei = (struct ext4_inode_info *) foo;
945 INIT_LIST_HEAD(&ei->i_orphan);
946 init_rwsem(&ei->xattr_sem);
947 init_rwsem(&ei->i_data_sem);
948 inode_init_once(&ei->vfs_inode);
951 static int __init init_inodecache(void)
953 ext4_inode_cachep = kmem_cache_create("ext4_inode_cache",
954 sizeof(struct ext4_inode_info),
955 0, (SLAB_RECLAIM_ACCOUNT|
958 if (ext4_inode_cachep == NULL)
963 static void destroy_inodecache(void)
966 * Make sure all delayed rcu free inodes are flushed before we
970 kmem_cache_destroy(ext4_inode_cachep);
973 void ext4_clear_inode(struct inode *inode)
975 invalidate_inode_buffers(inode);
978 ext4_discard_preallocations(inode);
979 ext4_es_remove_extent(inode, 0, EXT_MAX_BLOCKS);
980 if (EXT4_I(inode)->jinode) {
981 jbd2_journal_release_jbd_inode(EXT4_JOURNAL(inode),
982 EXT4_I(inode)->jinode);
983 jbd2_free_inode(EXT4_I(inode)->jinode);
984 EXT4_I(inode)->jinode = NULL;
986 #ifdef CONFIG_EXT4_FS_ENCRYPTION
987 if (EXT4_I(inode)->i_crypt_info)
988 ext4_free_encryption_info(inode, EXT4_I(inode)->i_crypt_info);
992 static struct inode *ext4_nfs_get_inode(struct super_block *sb,
993 u64 ino, u32 generation)
997 if (ino < EXT4_FIRST_INO(sb) && ino != EXT4_ROOT_INO)
998 return ERR_PTR(-ESTALE);
999 if (ino > le32_to_cpu(EXT4_SB(sb)->s_es->s_inodes_count))
1000 return ERR_PTR(-ESTALE);
1002 /* iget isn't really right if the inode is currently unallocated!!
1004 * ext4_read_inode will return a bad_inode if the inode had been
1005 * deleted, so we should be safe.
1007 * Currently we don't know the generation for parent directory, so
1008 * a generation of 0 means "accept any"
1010 inode = ext4_iget_normal(sb, ino);
1012 return ERR_CAST(inode);
1013 if (generation && inode->i_generation != generation) {
1015 return ERR_PTR(-ESTALE);
1021 static struct dentry *ext4_fh_to_dentry(struct super_block *sb, struct fid *fid,
1022 int fh_len, int fh_type)
1024 return generic_fh_to_dentry(sb, fid, fh_len, fh_type,
1025 ext4_nfs_get_inode);
1028 static struct dentry *ext4_fh_to_parent(struct super_block *sb, struct fid *fid,
1029 int fh_len, int fh_type)
1031 return generic_fh_to_parent(sb, fid, fh_len, fh_type,
1032 ext4_nfs_get_inode);
1036 * Try to release metadata pages (indirect blocks, directories) which are
1037 * mapped via the block device. Since these pages could have journal heads
1038 * which would prevent try_to_free_buffers() from freeing them, we must use
1039 * jbd2 layer's try_to_free_buffers() function to release them.
1041 static int bdev_try_to_free_page(struct super_block *sb, struct page *page,
1044 journal_t *journal = EXT4_SB(sb)->s_journal;
1046 WARN_ON(PageChecked(page));
1047 if (!page_has_buffers(page))
1050 return jbd2_journal_try_to_free_buffers(journal, page,
1051 wait & ~__GFP_WAIT);
1052 return try_to_free_buffers(page);
1056 #define QTYPE2NAME(t) ((t) == USRQUOTA ? "user" : "group")
1057 #define QTYPE2MOPT(on, t) ((t) == USRQUOTA?((on)##USRJQUOTA):((on)##GRPJQUOTA))
1059 static int ext4_write_dquot(struct dquot *dquot);
1060 static int ext4_acquire_dquot(struct dquot *dquot);
1061 static int ext4_release_dquot(struct dquot *dquot);
1062 static int ext4_mark_dquot_dirty(struct dquot *dquot);
1063 static int ext4_write_info(struct super_block *sb, int type);
1064 static int ext4_quota_on(struct super_block *sb, int type, int format_id,
1066 static int ext4_quota_off(struct super_block *sb, int type);
1067 static int ext4_quota_on_mount(struct super_block *sb, int type);
1068 static ssize_t ext4_quota_read(struct super_block *sb, int type, char *data,
1069 size_t len, loff_t off);
1070 static ssize_t ext4_quota_write(struct super_block *sb, int type,
1071 const char *data, size_t len, loff_t off);
1072 static int ext4_quota_enable(struct super_block *sb, int type, int format_id,
1073 unsigned int flags);
1074 static int ext4_enable_quotas(struct super_block *sb);
1076 static struct dquot **ext4_get_dquots(struct inode *inode)
1078 return EXT4_I(inode)->i_dquot;
1081 static const struct dquot_operations ext4_quota_operations = {
1082 .get_reserved_space = ext4_get_reserved_space,
1083 .write_dquot = ext4_write_dquot,
1084 .acquire_dquot = ext4_acquire_dquot,
1085 .release_dquot = ext4_release_dquot,
1086 .mark_dirty = ext4_mark_dquot_dirty,
1087 .write_info = ext4_write_info,
1088 .alloc_dquot = dquot_alloc,
1089 .destroy_dquot = dquot_destroy,
1092 static const struct quotactl_ops ext4_qctl_operations = {
1093 .quota_on = ext4_quota_on,
1094 .quota_off = ext4_quota_off,
1095 .quota_sync = dquot_quota_sync,
1096 .get_state = dquot_get_state,
1097 .set_info = dquot_set_dqinfo,
1098 .get_dqblk = dquot_get_dqblk,
1099 .set_dqblk = dquot_set_dqblk
1103 static const struct super_operations ext4_sops = {
1104 .alloc_inode = ext4_alloc_inode,
1105 .destroy_inode = ext4_destroy_inode,
1106 .write_inode = ext4_write_inode,
1107 .dirty_inode = ext4_dirty_inode,
1108 .drop_inode = ext4_drop_inode,
1109 .evict_inode = ext4_evict_inode,
1110 .put_super = ext4_put_super,
1111 .sync_fs = ext4_sync_fs,
1112 .freeze_fs = ext4_freeze,
1113 .unfreeze_fs = ext4_unfreeze,
1114 .statfs = ext4_statfs,
1115 .remount_fs = ext4_remount,
1116 .show_options = ext4_show_options,
1118 .quota_read = ext4_quota_read,
1119 .quota_write = ext4_quota_write,
1120 .get_dquots = ext4_get_dquots,
1122 .bdev_try_to_free_page = bdev_try_to_free_page,
1125 static const struct export_operations ext4_export_ops = {
1126 .fh_to_dentry = ext4_fh_to_dentry,
1127 .fh_to_parent = ext4_fh_to_parent,
1128 .get_parent = ext4_get_parent,
1132 Opt_bsd_df, Opt_minix_df, Opt_grpid, Opt_nogrpid,
1133 Opt_resgid, Opt_resuid, Opt_sb, Opt_err_cont, Opt_err_panic, Opt_err_ro,
1134 Opt_nouid32, Opt_debug, Opt_removed,
1135 Opt_user_xattr, Opt_nouser_xattr, Opt_acl, Opt_noacl,
1136 Opt_auto_da_alloc, Opt_noauto_da_alloc, Opt_noload,
1137 Opt_commit, Opt_min_batch_time, Opt_max_batch_time, Opt_journal_dev,
1138 Opt_journal_path, Opt_journal_checksum, Opt_journal_async_commit,
1139 Opt_abort, Opt_data_journal, Opt_data_ordered, Opt_data_writeback,
1140 Opt_data_err_abort, Opt_data_err_ignore, Opt_test_dummy_encryption,
1141 Opt_usrjquota, Opt_grpjquota, Opt_offusrjquota, Opt_offgrpjquota,
1142 Opt_jqfmt_vfsold, Opt_jqfmt_vfsv0, Opt_jqfmt_vfsv1, Opt_quota,
1143 Opt_noquota, Opt_barrier, Opt_nobarrier, Opt_err,
1144 Opt_usrquota, Opt_grpquota, Opt_i_version, Opt_dax,
1145 Opt_stripe, Opt_delalloc, Opt_nodelalloc, Opt_mblk_io_submit,
1146 Opt_lazytime, Opt_nolazytime,
1147 Opt_nomblk_io_submit, Opt_block_validity, Opt_noblock_validity,
1148 Opt_inode_readahead_blks, Opt_journal_ioprio,
1149 Opt_dioread_nolock, Opt_dioread_lock,
1150 Opt_discard, Opt_nodiscard, Opt_init_itable, Opt_noinit_itable,
1151 Opt_max_dir_size_kb, Opt_nojournal_checksum,
1154 static const match_table_t tokens = {
1155 {Opt_bsd_df, "bsddf"},
1156 {Opt_minix_df, "minixdf"},
1157 {Opt_grpid, "grpid"},
1158 {Opt_grpid, "bsdgroups"},
1159 {Opt_nogrpid, "nogrpid"},
1160 {Opt_nogrpid, "sysvgroups"},
1161 {Opt_resgid, "resgid=%u"},
1162 {Opt_resuid, "resuid=%u"},
1164 {Opt_err_cont, "errors=continue"},
1165 {Opt_err_panic, "errors=panic"},
1166 {Opt_err_ro, "errors=remount-ro"},
1167 {Opt_nouid32, "nouid32"},
1168 {Opt_debug, "debug"},
1169 {Opt_removed, "oldalloc"},
1170 {Opt_removed, "orlov"},
1171 {Opt_user_xattr, "user_xattr"},
1172 {Opt_nouser_xattr, "nouser_xattr"},
1174 {Opt_noacl, "noacl"},
1175 {Opt_noload, "norecovery"},
1176 {Opt_noload, "noload"},
1177 {Opt_removed, "nobh"},
1178 {Opt_removed, "bh"},
1179 {Opt_commit, "commit=%u"},
1180 {Opt_min_batch_time, "min_batch_time=%u"},
1181 {Opt_max_batch_time, "max_batch_time=%u"},
1182 {Opt_journal_dev, "journal_dev=%u"},
1183 {Opt_journal_path, "journal_path=%s"},
1184 {Opt_journal_checksum, "journal_checksum"},
1185 {Opt_nojournal_checksum, "nojournal_checksum"},
1186 {Opt_journal_async_commit, "journal_async_commit"},
1187 {Opt_abort, "abort"},
1188 {Opt_data_journal, "data=journal"},
1189 {Opt_data_ordered, "data=ordered"},
1190 {Opt_data_writeback, "data=writeback"},
1191 {Opt_data_err_abort, "data_err=abort"},
1192 {Opt_data_err_ignore, "data_err=ignore"},
1193 {Opt_offusrjquota, "usrjquota="},
1194 {Opt_usrjquota, "usrjquota=%s"},
1195 {Opt_offgrpjquota, "grpjquota="},
1196 {Opt_grpjquota, "grpjquota=%s"},
1197 {Opt_jqfmt_vfsold, "jqfmt=vfsold"},
1198 {Opt_jqfmt_vfsv0, "jqfmt=vfsv0"},
1199 {Opt_jqfmt_vfsv1, "jqfmt=vfsv1"},
1200 {Opt_grpquota, "grpquota"},
1201 {Opt_noquota, "noquota"},
1202 {Opt_quota, "quota"},
1203 {Opt_usrquota, "usrquota"},
1204 {Opt_barrier, "barrier=%u"},
1205 {Opt_barrier, "barrier"},
1206 {Opt_nobarrier, "nobarrier"},
1207 {Opt_i_version, "i_version"},
1209 {Opt_stripe, "stripe=%u"},
1210 {Opt_delalloc, "delalloc"},
1211 {Opt_lazytime, "lazytime"},
1212 {Opt_nolazytime, "nolazytime"},
1213 {Opt_nodelalloc, "nodelalloc"},
1214 {Opt_removed, "mblk_io_submit"},
1215 {Opt_removed, "nomblk_io_submit"},
1216 {Opt_block_validity, "block_validity"},
1217 {Opt_noblock_validity, "noblock_validity"},
1218 {Opt_inode_readahead_blks, "inode_readahead_blks=%u"},
1219 {Opt_journal_ioprio, "journal_ioprio=%u"},
1220 {Opt_auto_da_alloc, "auto_da_alloc=%u"},
1221 {Opt_auto_da_alloc, "auto_da_alloc"},
1222 {Opt_noauto_da_alloc, "noauto_da_alloc"},
1223 {Opt_dioread_nolock, "dioread_nolock"},
1224 {Opt_dioread_lock, "dioread_lock"},
1225 {Opt_discard, "discard"},
1226 {Opt_nodiscard, "nodiscard"},
1227 {Opt_init_itable, "init_itable=%u"},
1228 {Opt_init_itable, "init_itable"},
1229 {Opt_noinit_itable, "noinit_itable"},
1230 {Opt_max_dir_size_kb, "max_dir_size_kb=%u"},
1231 {Opt_test_dummy_encryption, "test_dummy_encryption"},
1232 {Opt_removed, "check=none"}, /* mount option from ext2/3 */
1233 {Opt_removed, "nocheck"}, /* mount option from ext2/3 */
1234 {Opt_removed, "reservation"}, /* mount option from ext2/3 */
1235 {Opt_removed, "noreservation"}, /* mount option from ext2/3 */
1236 {Opt_removed, "journal=%u"}, /* mount option from ext2/3 */
1240 static ext4_fsblk_t get_sb_block(void **data)
1242 ext4_fsblk_t sb_block;
1243 char *options = (char *) *data;
1245 if (!options || strncmp(options, "sb=", 3) != 0)
1246 return 1; /* Default location */
1249 /* TODO: use simple_strtoll with >32bit ext4 */
1250 sb_block = simple_strtoul(options, &options, 0);
1251 if (*options && *options != ',') {
1252 printk(KERN_ERR "EXT4-fs: Invalid sb specification: %s\n",
1256 if (*options == ',')
1258 *data = (void *) options;
1263 #define DEFAULT_JOURNAL_IOPRIO (IOPRIO_PRIO_VALUE(IOPRIO_CLASS_BE, 3))
1264 static char deprecated_msg[] = "Mount option \"%s\" will be removed by %s\n"
1265 "Contact linux-ext4@vger.kernel.org if you think we should keep it.\n";
1268 static int set_qf_name(struct super_block *sb, int qtype, substring_t *args)
1270 struct ext4_sb_info *sbi = EXT4_SB(sb);
1274 if (sb_any_quota_loaded(sb) &&
1275 !sbi->s_qf_names[qtype]) {
1276 ext4_msg(sb, KERN_ERR,
1277 "Cannot change journaled "
1278 "quota options when quota turned on");
1281 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_QUOTA)) {
1282 ext4_msg(sb, KERN_ERR, "Cannot set journaled quota options "
1283 "when QUOTA feature is enabled");
1286 qname = match_strdup(args);
1288 ext4_msg(sb, KERN_ERR,
1289 "Not enough memory for storing quotafile name");
1292 if (sbi->s_qf_names[qtype]) {
1293 if (strcmp(sbi->s_qf_names[qtype], qname) == 0)
1296 ext4_msg(sb, KERN_ERR,
1297 "%s quota file already specified",
1301 if (strchr(qname, '/')) {
1302 ext4_msg(sb, KERN_ERR,
1303 "quotafile must be on filesystem root");
1306 sbi->s_qf_names[qtype] = qname;
1314 static int clear_qf_name(struct super_block *sb, int qtype)
1317 struct ext4_sb_info *sbi = EXT4_SB(sb);
1319 if (sb_any_quota_loaded(sb) &&
1320 sbi->s_qf_names[qtype]) {
1321 ext4_msg(sb, KERN_ERR, "Cannot change journaled quota options"
1322 " when quota turned on");
1325 kfree(sbi->s_qf_names[qtype]);
1326 sbi->s_qf_names[qtype] = NULL;
1331 #define MOPT_SET 0x0001
1332 #define MOPT_CLEAR 0x0002
1333 #define MOPT_NOSUPPORT 0x0004
1334 #define MOPT_EXPLICIT 0x0008
1335 #define MOPT_CLEAR_ERR 0x0010
1336 #define MOPT_GTE0 0x0020
1339 #define MOPT_QFMT 0x0040
1341 #define MOPT_Q MOPT_NOSUPPORT
1342 #define MOPT_QFMT MOPT_NOSUPPORT
1344 #define MOPT_DATAJ 0x0080
1345 #define MOPT_NO_EXT2 0x0100
1346 #define MOPT_NO_EXT3 0x0200
1347 #define MOPT_EXT4_ONLY (MOPT_NO_EXT2 | MOPT_NO_EXT3)
1348 #define MOPT_STRING 0x0400
1350 static const struct mount_opts {
1354 } ext4_mount_opts[] = {
1355 {Opt_minix_df, EXT4_MOUNT_MINIX_DF, MOPT_SET},
1356 {Opt_bsd_df, EXT4_MOUNT_MINIX_DF, MOPT_CLEAR},
1357 {Opt_grpid, EXT4_MOUNT_GRPID, MOPT_SET},
1358 {Opt_nogrpid, EXT4_MOUNT_GRPID, MOPT_CLEAR},
1359 {Opt_block_validity, EXT4_MOUNT_BLOCK_VALIDITY, MOPT_SET},
1360 {Opt_noblock_validity, EXT4_MOUNT_BLOCK_VALIDITY, MOPT_CLEAR},
1361 {Opt_dioread_nolock, EXT4_MOUNT_DIOREAD_NOLOCK,
1362 MOPT_EXT4_ONLY | MOPT_SET},
1363 {Opt_dioread_lock, EXT4_MOUNT_DIOREAD_NOLOCK,
1364 MOPT_EXT4_ONLY | MOPT_CLEAR},
1365 {Opt_discard, EXT4_MOUNT_DISCARD, MOPT_SET},
1366 {Opt_nodiscard, EXT4_MOUNT_DISCARD, MOPT_CLEAR},
1367 {Opt_delalloc, EXT4_MOUNT_DELALLOC,
1368 MOPT_EXT4_ONLY | MOPT_SET | MOPT_EXPLICIT},
1369 {Opt_nodelalloc, EXT4_MOUNT_DELALLOC,
1370 MOPT_EXT4_ONLY | MOPT_CLEAR},
1371 {Opt_nojournal_checksum, EXT4_MOUNT_JOURNAL_CHECKSUM,
1372 MOPT_EXT4_ONLY | MOPT_CLEAR},
1373 {Opt_journal_checksum, EXT4_MOUNT_JOURNAL_CHECKSUM,
1374 MOPT_EXT4_ONLY | MOPT_SET},
1375 {Opt_journal_async_commit, (EXT4_MOUNT_JOURNAL_ASYNC_COMMIT |
1376 EXT4_MOUNT_JOURNAL_CHECKSUM),
1377 MOPT_EXT4_ONLY | MOPT_SET},
1378 {Opt_noload, EXT4_MOUNT_NOLOAD, MOPT_NO_EXT2 | MOPT_SET},
1379 {Opt_err_panic, EXT4_MOUNT_ERRORS_PANIC, MOPT_SET | MOPT_CLEAR_ERR},
1380 {Opt_err_ro, EXT4_MOUNT_ERRORS_RO, MOPT_SET | MOPT_CLEAR_ERR},
1381 {Opt_err_cont, EXT4_MOUNT_ERRORS_CONT, MOPT_SET | MOPT_CLEAR_ERR},
1382 {Opt_data_err_abort, EXT4_MOUNT_DATA_ERR_ABORT,
1383 MOPT_NO_EXT2 | MOPT_SET},
1384 {Opt_data_err_ignore, EXT4_MOUNT_DATA_ERR_ABORT,
1385 MOPT_NO_EXT2 | MOPT_CLEAR},
1386 {Opt_barrier, EXT4_MOUNT_BARRIER, MOPT_SET},
1387 {Opt_nobarrier, EXT4_MOUNT_BARRIER, MOPT_CLEAR},
1388 {Opt_noauto_da_alloc, EXT4_MOUNT_NO_AUTO_DA_ALLOC, MOPT_SET},
1389 {Opt_auto_da_alloc, EXT4_MOUNT_NO_AUTO_DA_ALLOC, MOPT_CLEAR},
1390 {Opt_noinit_itable, EXT4_MOUNT_INIT_INODE_TABLE, MOPT_CLEAR},
1391 {Opt_commit, 0, MOPT_GTE0},
1392 {Opt_max_batch_time, 0, MOPT_GTE0},
1393 {Opt_min_batch_time, 0, MOPT_GTE0},
1394 {Opt_inode_readahead_blks, 0, MOPT_GTE0},
1395 {Opt_init_itable, 0, MOPT_GTE0},
1396 {Opt_dax, EXT4_MOUNT_DAX, MOPT_SET},
1397 {Opt_stripe, 0, MOPT_GTE0},
1398 {Opt_resuid, 0, MOPT_GTE0},
1399 {Opt_resgid, 0, MOPT_GTE0},
1400 {Opt_journal_dev, 0, MOPT_NO_EXT2 | MOPT_GTE0},
1401 {Opt_journal_path, 0, MOPT_NO_EXT2 | MOPT_STRING},
1402 {Opt_journal_ioprio, 0, MOPT_NO_EXT2 | MOPT_GTE0},
1403 {Opt_data_journal, EXT4_MOUNT_JOURNAL_DATA, MOPT_NO_EXT2 | MOPT_DATAJ},
1404 {Opt_data_ordered, EXT4_MOUNT_ORDERED_DATA, MOPT_NO_EXT2 | MOPT_DATAJ},
1405 {Opt_data_writeback, EXT4_MOUNT_WRITEBACK_DATA,
1406 MOPT_NO_EXT2 | MOPT_DATAJ},
1407 {Opt_user_xattr, EXT4_MOUNT_XATTR_USER, MOPT_SET},
1408 {Opt_nouser_xattr, EXT4_MOUNT_XATTR_USER, MOPT_CLEAR},
1409 #ifdef CONFIG_EXT4_FS_POSIX_ACL
1410 {Opt_acl, EXT4_MOUNT_POSIX_ACL, MOPT_SET},
1411 {Opt_noacl, EXT4_MOUNT_POSIX_ACL, MOPT_CLEAR},
1413 {Opt_acl, 0, MOPT_NOSUPPORT},
1414 {Opt_noacl, 0, MOPT_NOSUPPORT},
1416 {Opt_nouid32, EXT4_MOUNT_NO_UID32, MOPT_SET},
1417 {Opt_debug, EXT4_MOUNT_DEBUG, MOPT_SET},
1418 {Opt_quota, EXT4_MOUNT_QUOTA | EXT4_MOUNT_USRQUOTA, MOPT_SET | MOPT_Q},
1419 {Opt_usrquota, EXT4_MOUNT_QUOTA | EXT4_MOUNT_USRQUOTA,
1421 {Opt_grpquota, EXT4_MOUNT_QUOTA | EXT4_MOUNT_GRPQUOTA,
1423 {Opt_noquota, (EXT4_MOUNT_QUOTA | EXT4_MOUNT_USRQUOTA |
1424 EXT4_MOUNT_GRPQUOTA), MOPT_CLEAR | MOPT_Q},
1425 {Opt_usrjquota, 0, MOPT_Q},
1426 {Opt_grpjquota, 0, MOPT_Q},
1427 {Opt_offusrjquota, 0, MOPT_Q},
1428 {Opt_offgrpjquota, 0, MOPT_Q},
1429 {Opt_jqfmt_vfsold, QFMT_VFS_OLD, MOPT_QFMT},
1430 {Opt_jqfmt_vfsv0, QFMT_VFS_V0, MOPT_QFMT},
1431 {Opt_jqfmt_vfsv1, QFMT_VFS_V1, MOPT_QFMT},
1432 {Opt_max_dir_size_kb, 0, MOPT_GTE0},
1433 {Opt_test_dummy_encryption, 0, MOPT_GTE0},
1437 static int handle_mount_opt(struct super_block *sb, char *opt, int token,
1438 substring_t *args, unsigned long *journal_devnum,
1439 unsigned int *journal_ioprio, int is_remount)
1441 struct ext4_sb_info *sbi = EXT4_SB(sb);
1442 const struct mount_opts *m;
1448 if (token == Opt_usrjquota)
1449 return set_qf_name(sb, USRQUOTA, &args[0]);
1450 else if (token == Opt_grpjquota)
1451 return set_qf_name(sb, GRPQUOTA, &args[0]);
1452 else if (token == Opt_offusrjquota)
1453 return clear_qf_name(sb, USRQUOTA);
1454 else if (token == Opt_offgrpjquota)
1455 return clear_qf_name(sb, GRPQUOTA);
1459 case Opt_nouser_xattr:
1460 ext4_msg(sb, KERN_WARNING, deprecated_msg, opt, "3.5");
1463 return 1; /* handled by get_sb_block() */
1465 ext4_msg(sb, KERN_WARNING, "Ignoring removed %s option", opt);
1468 sbi->s_mount_flags |= EXT4_MF_FS_ABORTED;
1471 sb->s_flags |= MS_I_VERSION;
1474 sb->s_flags |= MS_LAZYTIME;
1476 case Opt_nolazytime:
1477 sb->s_flags &= ~MS_LAZYTIME;
1481 for (m = ext4_mount_opts; m->token != Opt_err; m++)
1482 if (token == m->token)
1485 if (m->token == Opt_err) {
1486 ext4_msg(sb, KERN_ERR, "Unrecognized mount option \"%s\" "
1487 "or missing value", opt);
1491 if ((m->flags & MOPT_NO_EXT2) && IS_EXT2_SB(sb)) {
1492 ext4_msg(sb, KERN_ERR,
1493 "Mount option \"%s\" incompatible with ext2", opt);
1496 if ((m->flags & MOPT_NO_EXT3) && IS_EXT3_SB(sb)) {
1497 ext4_msg(sb, KERN_ERR,
1498 "Mount option \"%s\" incompatible with ext3", opt);
1502 if (args->from && !(m->flags & MOPT_STRING) && match_int(args, &arg))
1504 if (args->from && (m->flags & MOPT_GTE0) && (arg < 0))
1506 if (m->flags & MOPT_EXPLICIT)
1507 set_opt2(sb, EXPLICIT_DELALLOC);
1508 if (m->flags & MOPT_CLEAR_ERR)
1509 clear_opt(sb, ERRORS_MASK);
1510 if (token == Opt_noquota && sb_any_quota_loaded(sb)) {
1511 ext4_msg(sb, KERN_ERR, "Cannot change quota "
1512 "options when quota turned on");
1516 if (m->flags & MOPT_NOSUPPORT) {
1517 ext4_msg(sb, KERN_ERR, "%s option not supported", opt);
1518 } else if (token == Opt_commit) {
1520 arg = JBD2_DEFAULT_MAX_COMMIT_AGE;
1521 sbi->s_commit_interval = HZ * arg;
1522 } else if (token == Opt_max_batch_time) {
1523 sbi->s_max_batch_time = arg;
1524 } else if (token == Opt_min_batch_time) {
1525 sbi->s_min_batch_time = arg;
1526 } else if (token == Opt_inode_readahead_blks) {
1527 if (arg && (arg > (1 << 30) || !is_power_of_2(arg))) {
1528 ext4_msg(sb, KERN_ERR,
1529 "EXT4-fs: inode_readahead_blks must be "
1530 "0 or a power of 2 smaller than 2^31");
1533 sbi->s_inode_readahead_blks = arg;
1534 } else if (token == Opt_init_itable) {
1535 set_opt(sb, INIT_INODE_TABLE);
1537 arg = EXT4_DEF_LI_WAIT_MULT;
1538 sbi->s_li_wait_mult = arg;
1539 } else if (token == Opt_max_dir_size_kb) {
1540 sbi->s_max_dir_size_kb = arg;
1541 } else if (token == Opt_stripe) {
1542 sbi->s_stripe = arg;
1543 } else if (token == Opt_resuid) {
1544 uid = make_kuid(current_user_ns(), arg);
1545 if (!uid_valid(uid)) {
1546 ext4_msg(sb, KERN_ERR, "Invalid uid value %d", arg);
1549 sbi->s_resuid = uid;
1550 } else if (token == Opt_resgid) {
1551 gid = make_kgid(current_user_ns(), arg);
1552 if (!gid_valid(gid)) {
1553 ext4_msg(sb, KERN_ERR, "Invalid gid value %d", arg);
1556 sbi->s_resgid = gid;
1557 } else if (token == Opt_journal_dev) {
1559 ext4_msg(sb, KERN_ERR,
1560 "Cannot specify journal on remount");
1563 *journal_devnum = arg;
1564 } else if (token == Opt_journal_path) {
1566 struct inode *journal_inode;
1571 ext4_msg(sb, KERN_ERR,
1572 "Cannot specify journal on remount");
1575 journal_path = match_strdup(&args[0]);
1576 if (!journal_path) {
1577 ext4_msg(sb, KERN_ERR, "error: could not dup "
1578 "journal device string");
1582 error = kern_path(journal_path, LOOKUP_FOLLOW, &path);
1584 ext4_msg(sb, KERN_ERR, "error: could not find "
1585 "journal device path: error %d", error);
1586 kfree(journal_path);
1590 journal_inode = d_inode(path.dentry);
1591 if (!S_ISBLK(journal_inode->i_mode)) {
1592 ext4_msg(sb, KERN_ERR, "error: journal path %s "
1593 "is not a block device", journal_path);
1595 kfree(journal_path);
1599 *journal_devnum = new_encode_dev(journal_inode->i_rdev);
1601 kfree(journal_path);
1602 } else if (token == Opt_journal_ioprio) {
1604 ext4_msg(sb, KERN_ERR, "Invalid journal IO priority"
1609 IOPRIO_PRIO_VALUE(IOPRIO_CLASS_BE, arg);
1610 } else if (token == Opt_test_dummy_encryption) {
1611 #ifdef CONFIG_EXT4_FS_ENCRYPTION
1612 sbi->s_mount_flags |= EXT4_MF_TEST_DUMMY_ENCRYPTION;
1613 ext4_msg(sb, KERN_WARNING,
1614 "Test dummy encryption mode enabled");
1616 ext4_msg(sb, KERN_WARNING,
1617 "Test dummy encryption mount option ignored");
1619 } else if (m->flags & MOPT_DATAJ) {
1621 if (!sbi->s_journal)
1622 ext4_msg(sb, KERN_WARNING, "Remounting file system with no journal so ignoring journalled data option");
1623 else if (test_opt(sb, DATA_FLAGS) != m->mount_opt) {
1624 ext4_msg(sb, KERN_ERR,
1625 "Cannot change data mode on remount");
1629 clear_opt(sb, DATA_FLAGS);
1630 sbi->s_mount_opt |= m->mount_opt;
1633 } else if (m->flags & MOPT_QFMT) {
1634 if (sb_any_quota_loaded(sb) &&
1635 sbi->s_jquota_fmt != m->mount_opt) {
1636 ext4_msg(sb, KERN_ERR, "Cannot change journaled "
1637 "quota options when quota turned on");
1640 if (EXT4_HAS_RO_COMPAT_FEATURE(sb,
1641 EXT4_FEATURE_RO_COMPAT_QUOTA)) {
1642 ext4_msg(sb, KERN_ERR,
1643 "Cannot set journaled quota options "
1644 "when QUOTA feature is enabled");
1647 sbi->s_jquota_fmt = m->mount_opt;
1649 #ifndef CONFIG_FS_DAX
1650 } else if (token == Opt_dax) {
1651 ext4_msg(sb, KERN_INFO, "dax option not supported");
1657 if (m->flags & MOPT_CLEAR)
1659 else if (unlikely(!(m->flags & MOPT_SET))) {
1660 ext4_msg(sb, KERN_WARNING,
1661 "buggy handling of option %s", opt);
1666 sbi->s_mount_opt |= m->mount_opt;
1668 sbi->s_mount_opt &= ~m->mount_opt;
1673 static int parse_options(char *options, struct super_block *sb,
1674 unsigned long *journal_devnum,
1675 unsigned int *journal_ioprio,
1678 struct ext4_sb_info *sbi = EXT4_SB(sb);
1680 substring_t args[MAX_OPT_ARGS];
1686 while ((p = strsep(&options, ",")) != NULL) {
1690 * Initialize args struct so we know whether arg was
1691 * found; some options take optional arguments.
1693 args[0].to = args[0].from = NULL;
1694 token = match_token(p, tokens, args);
1695 if (handle_mount_opt(sb, p, token, args, journal_devnum,
1696 journal_ioprio, is_remount) < 0)
1700 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_QUOTA) &&
1701 (test_opt(sb, USRQUOTA) || test_opt(sb, GRPQUOTA))) {
1702 ext4_msg(sb, KERN_ERR, "Cannot set quota options when QUOTA "
1703 "feature is enabled");
1706 if (sbi->s_qf_names[USRQUOTA] || sbi->s_qf_names[GRPQUOTA]) {
1707 if (test_opt(sb, USRQUOTA) && sbi->s_qf_names[USRQUOTA])
1708 clear_opt(sb, USRQUOTA);
1710 if (test_opt(sb, GRPQUOTA) && sbi->s_qf_names[GRPQUOTA])
1711 clear_opt(sb, GRPQUOTA);
1713 if (test_opt(sb, GRPQUOTA) || test_opt(sb, USRQUOTA)) {
1714 ext4_msg(sb, KERN_ERR, "old and new quota "
1719 if (!sbi->s_jquota_fmt) {
1720 ext4_msg(sb, KERN_ERR, "journaled quota format "
1726 if (test_opt(sb, DIOREAD_NOLOCK)) {
1728 BLOCK_SIZE << le32_to_cpu(sbi->s_es->s_log_block_size);
1730 if (blocksize < PAGE_CACHE_SIZE) {
1731 ext4_msg(sb, KERN_ERR, "can't mount with "
1732 "dioread_nolock if block size != PAGE_SIZE");
1736 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA &&
1737 test_opt(sb, JOURNAL_ASYNC_COMMIT)) {
1738 ext4_msg(sb, KERN_ERR, "can't mount with journal_async_commit "
1739 "in data=ordered mode");
1745 static inline void ext4_show_quota_options(struct seq_file *seq,
1746 struct super_block *sb)
1748 #if defined(CONFIG_QUOTA)
1749 struct ext4_sb_info *sbi = EXT4_SB(sb);
1751 if (sbi->s_jquota_fmt) {
1754 switch (sbi->s_jquota_fmt) {
1765 seq_printf(seq, ",jqfmt=%s", fmtname);
1768 if (sbi->s_qf_names[USRQUOTA])
1769 seq_show_option(seq, "usrjquota", sbi->s_qf_names[USRQUOTA]);
1771 if (sbi->s_qf_names[GRPQUOTA])
1772 seq_show_option(seq, "grpjquota", sbi->s_qf_names[GRPQUOTA]);
1776 static const char *token2str(int token)
1778 const struct match_token *t;
1780 for (t = tokens; t->token != Opt_err; t++)
1781 if (t->token == token && !strchr(t->pattern, '='))
1788 * - it's set to a non-default value OR
1789 * - if the per-sb default is different from the global default
1791 static int _ext4_show_options(struct seq_file *seq, struct super_block *sb,
1794 struct ext4_sb_info *sbi = EXT4_SB(sb);
1795 struct ext4_super_block *es = sbi->s_es;
1796 int def_errors, def_mount_opt = nodefs ? 0 : sbi->s_def_mount_opt;
1797 const struct mount_opts *m;
1798 char sep = nodefs ? '\n' : ',';
1800 #define SEQ_OPTS_PUTS(str) seq_printf(seq, "%c" str, sep)
1801 #define SEQ_OPTS_PRINT(str, arg) seq_printf(seq, "%c" str, sep, arg)
1803 if (sbi->s_sb_block != 1)
1804 SEQ_OPTS_PRINT("sb=%llu", sbi->s_sb_block);
1806 for (m = ext4_mount_opts; m->token != Opt_err; m++) {
1807 int want_set = m->flags & MOPT_SET;
1808 if (((m->flags & (MOPT_SET|MOPT_CLEAR)) == 0) ||
1809 (m->flags & MOPT_CLEAR_ERR))
1811 if (!(m->mount_opt & (sbi->s_mount_opt ^ def_mount_opt)))
1812 continue; /* skip if same as the default */
1814 (sbi->s_mount_opt & m->mount_opt) != m->mount_opt) ||
1815 (!want_set && (sbi->s_mount_opt & m->mount_opt)))
1816 continue; /* select Opt_noFoo vs Opt_Foo */
1817 SEQ_OPTS_PRINT("%s", token2str(m->token));
1820 if (nodefs || !uid_eq(sbi->s_resuid, make_kuid(&init_user_ns, EXT4_DEF_RESUID)) ||
1821 le16_to_cpu(es->s_def_resuid) != EXT4_DEF_RESUID)
1822 SEQ_OPTS_PRINT("resuid=%u",
1823 from_kuid_munged(&init_user_ns, sbi->s_resuid));
1824 if (nodefs || !gid_eq(sbi->s_resgid, make_kgid(&init_user_ns, EXT4_DEF_RESGID)) ||
1825 le16_to_cpu(es->s_def_resgid) != EXT4_DEF_RESGID)
1826 SEQ_OPTS_PRINT("resgid=%u",
1827 from_kgid_munged(&init_user_ns, sbi->s_resgid));
1828 def_errors = nodefs ? -1 : le16_to_cpu(es->s_errors);
1829 if (test_opt(sb, ERRORS_RO) && def_errors != EXT4_ERRORS_RO)
1830 SEQ_OPTS_PUTS("errors=remount-ro");
1831 if (test_opt(sb, ERRORS_CONT) && def_errors != EXT4_ERRORS_CONTINUE)
1832 SEQ_OPTS_PUTS("errors=continue");
1833 if (test_opt(sb, ERRORS_PANIC) && def_errors != EXT4_ERRORS_PANIC)
1834 SEQ_OPTS_PUTS("errors=panic");
1835 if (nodefs || sbi->s_commit_interval != JBD2_DEFAULT_MAX_COMMIT_AGE*HZ)
1836 SEQ_OPTS_PRINT("commit=%lu", sbi->s_commit_interval / HZ);
1837 if (nodefs || sbi->s_min_batch_time != EXT4_DEF_MIN_BATCH_TIME)
1838 SEQ_OPTS_PRINT("min_batch_time=%u", sbi->s_min_batch_time);
1839 if (nodefs || sbi->s_max_batch_time != EXT4_DEF_MAX_BATCH_TIME)
1840 SEQ_OPTS_PRINT("max_batch_time=%u", sbi->s_max_batch_time);
1841 if (sb->s_flags & MS_I_VERSION)
1842 SEQ_OPTS_PUTS("i_version");
1843 if (nodefs || sbi->s_stripe)
1844 SEQ_OPTS_PRINT("stripe=%lu", sbi->s_stripe);
1845 if (EXT4_MOUNT_DATA_FLAGS & (sbi->s_mount_opt ^ def_mount_opt)) {
1846 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA)
1847 SEQ_OPTS_PUTS("data=journal");
1848 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA)
1849 SEQ_OPTS_PUTS("data=ordered");
1850 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_WRITEBACK_DATA)
1851 SEQ_OPTS_PUTS("data=writeback");
1854 sbi->s_inode_readahead_blks != EXT4_DEF_INODE_READAHEAD_BLKS)
1855 SEQ_OPTS_PRINT("inode_readahead_blks=%u",
1856 sbi->s_inode_readahead_blks);
1858 if (nodefs || (test_opt(sb, INIT_INODE_TABLE) &&
1859 (sbi->s_li_wait_mult != EXT4_DEF_LI_WAIT_MULT)))
1860 SEQ_OPTS_PRINT("init_itable=%u", sbi->s_li_wait_mult);
1861 if (nodefs || sbi->s_max_dir_size_kb)
1862 SEQ_OPTS_PRINT("max_dir_size_kb=%u", sbi->s_max_dir_size_kb);
1864 ext4_show_quota_options(seq, sb);
1868 static int ext4_show_options(struct seq_file *seq, struct dentry *root)
1870 return _ext4_show_options(seq, root->d_sb, 0);
1873 int ext4_seq_options_show(struct seq_file *seq, void *offset)
1875 struct super_block *sb = seq->private;
1878 seq_puts(seq, (sb->s_flags & MS_RDONLY) ? "ro" : "rw");
1879 rc = _ext4_show_options(seq, sb, 1);
1880 seq_puts(seq, "\n");
1884 static int ext4_setup_super(struct super_block *sb, struct ext4_super_block *es,
1887 struct ext4_sb_info *sbi = EXT4_SB(sb);
1890 if (le32_to_cpu(es->s_rev_level) > EXT4_MAX_SUPP_REV) {
1891 ext4_msg(sb, KERN_ERR, "revision level too high, "
1892 "forcing read-only mode");
1897 if (!(sbi->s_mount_state & EXT4_VALID_FS))
1898 ext4_msg(sb, KERN_WARNING, "warning: mounting unchecked fs, "
1899 "running e2fsck is recommended");
1900 else if (sbi->s_mount_state & EXT4_ERROR_FS)
1901 ext4_msg(sb, KERN_WARNING,
1902 "warning: mounting fs with errors, "
1903 "running e2fsck is recommended");
1904 else if ((__s16) le16_to_cpu(es->s_max_mnt_count) > 0 &&
1905 le16_to_cpu(es->s_mnt_count) >=
1906 (unsigned short) (__s16) le16_to_cpu(es->s_max_mnt_count))
1907 ext4_msg(sb, KERN_WARNING,
1908 "warning: maximal mount count reached, "
1909 "running e2fsck is recommended");
1910 else if (le32_to_cpu(es->s_checkinterval) &&
1911 (le32_to_cpu(es->s_lastcheck) +
1912 le32_to_cpu(es->s_checkinterval) <= get_seconds()))
1913 ext4_msg(sb, KERN_WARNING,
1914 "warning: checktime reached, "
1915 "running e2fsck is recommended");
1916 if (!sbi->s_journal)
1917 es->s_state &= cpu_to_le16(~EXT4_VALID_FS);
1918 if (!(__s16) le16_to_cpu(es->s_max_mnt_count))
1919 es->s_max_mnt_count = cpu_to_le16(EXT4_DFL_MAX_MNT_COUNT);
1920 le16_add_cpu(&es->s_mnt_count, 1);
1921 es->s_mtime = cpu_to_le32(get_seconds());
1922 ext4_update_dynamic_rev(sb);
1924 EXT4_SET_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
1926 ext4_commit_super(sb, 1);
1928 if (test_opt(sb, DEBUG))
1929 printk(KERN_INFO "[EXT4 FS bs=%lu, gc=%u, "
1930 "bpg=%lu, ipg=%lu, mo=%04x, mo2=%04x]\n",
1932 sbi->s_groups_count,
1933 EXT4_BLOCKS_PER_GROUP(sb),
1934 EXT4_INODES_PER_GROUP(sb),
1935 sbi->s_mount_opt, sbi->s_mount_opt2);
1937 cleancache_init_fs(sb);
1941 int ext4_alloc_flex_bg_array(struct super_block *sb, ext4_group_t ngroup)
1943 struct ext4_sb_info *sbi = EXT4_SB(sb);
1944 struct flex_groups *new_groups;
1947 if (!sbi->s_log_groups_per_flex)
1950 size = ext4_flex_group(sbi, ngroup - 1) + 1;
1951 if (size <= sbi->s_flex_groups_allocated)
1954 size = roundup_pow_of_two(size * sizeof(struct flex_groups));
1955 new_groups = ext4_kvzalloc(size, GFP_KERNEL);
1957 ext4_msg(sb, KERN_ERR, "not enough memory for %d flex groups",
1958 size / (int) sizeof(struct flex_groups));
1962 if (sbi->s_flex_groups) {
1963 memcpy(new_groups, sbi->s_flex_groups,
1964 (sbi->s_flex_groups_allocated *
1965 sizeof(struct flex_groups)));
1966 kvfree(sbi->s_flex_groups);
1968 sbi->s_flex_groups = new_groups;
1969 sbi->s_flex_groups_allocated = size / sizeof(struct flex_groups);
1973 static int ext4_fill_flex_info(struct super_block *sb)
1975 struct ext4_sb_info *sbi = EXT4_SB(sb);
1976 struct ext4_group_desc *gdp = NULL;
1977 ext4_group_t flex_group;
1980 sbi->s_log_groups_per_flex = sbi->s_es->s_log_groups_per_flex;
1981 if (sbi->s_log_groups_per_flex < 1 || sbi->s_log_groups_per_flex > 31) {
1982 sbi->s_log_groups_per_flex = 0;
1986 err = ext4_alloc_flex_bg_array(sb, sbi->s_groups_count);
1990 for (i = 0; i < sbi->s_groups_count; i++) {
1991 gdp = ext4_get_group_desc(sb, i, NULL);
1993 flex_group = ext4_flex_group(sbi, i);
1994 atomic_add(ext4_free_inodes_count(sb, gdp),
1995 &sbi->s_flex_groups[flex_group].free_inodes);
1996 atomic64_add(ext4_free_group_clusters(sb, gdp),
1997 &sbi->s_flex_groups[flex_group].free_clusters);
1998 atomic_add(ext4_used_dirs_count(sb, gdp),
1999 &sbi->s_flex_groups[flex_group].used_dirs);
2007 static __le16 ext4_group_desc_csum(struct ext4_sb_info *sbi, __u32 block_group,
2008 struct ext4_group_desc *gdp)
2012 __le32 le_group = cpu_to_le32(block_group);
2014 if (ext4_has_metadata_csum(sbi->s_sb)) {
2015 /* Use new metadata_csum algorithm */
2019 save_csum = gdp->bg_checksum;
2020 gdp->bg_checksum = 0;
2021 csum32 = ext4_chksum(sbi, sbi->s_csum_seed, (__u8 *)&le_group,
2023 csum32 = ext4_chksum(sbi, csum32, (__u8 *)gdp,
2025 gdp->bg_checksum = save_csum;
2027 crc = csum32 & 0xFFFF;
2031 /* old crc16 code */
2032 if (!(sbi->s_es->s_feature_ro_compat &
2033 cpu_to_le32(EXT4_FEATURE_RO_COMPAT_GDT_CSUM)))
2036 offset = offsetof(struct ext4_group_desc, bg_checksum);
2038 crc = crc16(~0, sbi->s_es->s_uuid, sizeof(sbi->s_es->s_uuid));
2039 crc = crc16(crc, (__u8 *)&le_group, sizeof(le_group));
2040 crc = crc16(crc, (__u8 *)gdp, offset);
2041 offset += sizeof(gdp->bg_checksum); /* skip checksum */
2042 /* for checksum of struct ext4_group_desc do the rest...*/
2043 if ((sbi->s_es->s_feature_incompat &
2044 cpu_to_le32(EXT4_FEATURE_INCOMPAT_64BIT)) &&
2045 offset < le16_to_cpu(sbi->s_es->s_desc_size))
2046 crc = crc16(crc, (__u8 *)gdp + offset,
2047 le16_to_cpu(sbi->s_es->s_desc_size) -
2051 return cpu_to_le16(crc);
2054 int ext4_group_desc_csum_verify(struct super_block *sb, __u32 block_group,
2055 struct ext4_group_desc *gdp)
2057 if (ext4_has_group_desc_csum(sb) &&
2058 (gdp->bg_checksum != ext4_group_desc_csum(EXT4_SB(sb),
2065 void ext4_group_desc_csum_set(struct super_block *sb, __u32 block_group,
2066 struct ext4_group_desc *gdp)
2068 if (!ext4_has_group_desc_csum(sb))
2070 gdp->bg_checksum = ext4_group_desc_csum(EXT4_SB(sb), block_group, gdp);
2073 /* Called at mount-time, super-block is locked */
2074 static int ext4_check_descriptors(struct super_block *sb,
2075 ext4_group_t *first_not_zeroed)
2077 struct ext4_sb_info *sbi = EXT4_SB(sb);
2078 ext4_fsblk_t first_block = le32_to_cpu(sbi->s_es->s_first_data_block);
2079 ext4_fsblk_t last_block;
2080 ext4_fsblk_t block_bitmap;
2081 ext4_fsblk_t inode_bitmap;
2082 ext4_fsblk_t inode_table;
2083 int flexbg_flag = 0;
2084 ext4_group_t i, grp = sbi->s_groups_count;
2086 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_FLEX_BG))
2089 ext4_debug("Checking group descriptors");
2091 for (i = 0; i < sbi->s_groups_count; i++) {
2092 struct ext4_group_desc *gdp = ext4_get_group_desc(sb, i, NULL);
2094 if (i == sbi->s_groups_count - 1 || flexbg_flag)
2095 last_block = ext4_blocks_count(sbi->s_es) - 1;
2097 last_block = first_block +
2098 (EXT4_BLOCKS_PER_GROUP(sb) - 1);
2100 if ((grp == sbi->s_groups_count) &&
2101 !(gdp->bg_flags & cpu_to_le16(EXT4_BG_INODE_ZEROED)))
2104 block_bitmap = ext4_block_bitmap(sb, gdp);
2105 if (block_bitmap < first_block || block_bitmap > last_block) {
2106 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2107 "Block bitmap for group %u not in group "
2108 "(block %llu)!", i, block_bitmap);
2111 inode_bitmap = ext4_inode_bitmap(sb, gdp);
2112 if (inode_bitmap < first_block || inode_bitmap > last_block) {
2113 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2114 "Inode bitmap for group %u not in group "
2115 "(block %llu)!", i, inode_bitmap);
2118 inode_table = ext4_inode_table(sb, gdp);
2119 if (inode_table < first_block ||
2120 inode_table + sbi->s_itb_per_group - 1 > last_block) {
2121 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2122 "Inode table for group %u not in group "
2123 "(block %llu)!", i, inode_table);
2126 ext4_lock_group(sb, i);
2127 if (!ext4_group_desc_csum_verify(sb, i, gdp)) {
2128 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2129 "Checksum for group %u failed (%u!=%u)",
2130 i, le16_to_cpu(ext4_group_desc_csum(sbi, i,
2131 gdp)), le16_to_cpu(gdp->bg_checksum));
2132 if (!(sb->s_flags & MS_RDONLY)) {
2133 ext4_unlock_group(sb, i);
2137 ext4_unlock_group(sb, i);
2139 first_block += EXT4_BLOCKS_PER_GROUP(sb);
2141 if (NULL != first_not_zeroed)
2142 *first_not_zeroed = grp;
2146 /* ext4_orphan_cleanup() walks a singly-linked list of inodes (starting at
2147 * the superblock) which were deleted from all directories, but held open by
2148 * a process at the time of a crash. We walk the list and try to delete these
2149 * inodes at recovery time (only with a read-write filesystem).
2151 * In order to keep the orphan inode chain consistent during traversal (in
2152 * case of crash during recovery), we link each inode into the superblock
2153 * orphan list_head and handle it the same way as an inode deletion during
2154 * normal operation (which journals the operations for us).
2156 * We only do an iget() and an iput() on each inode, which is very safe if we
2157 * accidentally point at an in-use or already deleted inode. The worst that
2158 * can happen in this case is that we get a "bit already cleared" message from
2159 * ext4_free_inode(). The only reason we would point at a wrong inode is if
2160 * e2fsck was run on this filesystem, and it must have already done the orphan
2161 * inode cleanup for us, so we can safely abort without any further action.
2163 static void ext4_orphan_cleanup(struct super_block *sb,
2164 struct ext4_super_block *es)
2166 unsigned int s_flags = sb->s_flags;
2167 int nr_orphans = 0, nr_truncates = 0;
2171 if (!es->s_last_orphan) {
2172 jbd_debug(4, "no orphan inodes to clean up\n");
2176 if (bdev_read_only(sb->s_bdev)) {
2177 ext4_msg(sb, KERN_ERR, "write access "
2178 "unavailable, skipping orphan cleanup");
2182 /* Check if feature set would not allow a r/w mount */
2183 if (!ext4_feature_set_ok(sb, 0)) {
2184 ext4_msg(sb, KERN_INFO, "Skipping orphan cleanup due to "
2185 "unknown ROCOMPAT features");
2189 if (EXT4_SB(sb)->s_mount_state & EXT4_ERROR_FS) {
2190 /* don't clear list on RO mount w/ errors */
2191 if (es->s_last_orphan && !(s_flags & MS_RDONLY)) {
2192 ext4_msg(sb, KERN_INFO, "Errors on filesystem, "
2193 "clearing orphan list.\n");
2194 es->s_last_orphan = 0;
2196 jbd_debug(1, "Skipping orphan recovery on fs with errors.\n");
2200 if (s_flags & MS_RDONLY) {
2201 ext4_msg(sb, KERN_INFO, "orphan cleanup on readonly fs");
2202 sb->s_flags &= ~MS_RDONLY;
2205 /* Needed for iput() to work correctly and not trash data */
2206 sb->s_flags |= MS_ACTIVE;
2207 /* Turn on quotas so that they are updated correctly */
2208 for (i = 0; i < EXT4_MAXQUOTAS; i++) {
2209 if (EXT4_SB(sb)->s_qf_names[i]) {
2210 int ret = ext4_quota_on_mount(sb, i);
2212 ext4_msg(sb, KERN_ERR,
2213 "Cannot turn on journaled "
2214 "quota: error %d", ret);
2219 while (es->s_last_orphan) {
2220 struct inode *inode;
2222 inode = ext4_orphan_get(sb, le32_to_cpu(es->s_last_orphan));
2223 if (IS_ERR(inode)) {
2224 es->s_last_orphan = 0;
2228 list_add(&EXT4_I(inode)->i_orphan, &EXT4_SB(sb)->s_orphan);
2229 dquot_initialize(inode);
2230 if (inode->i_nlink) {
2231 if (test_opt(sb, DEBUG))
2232 ext4_msg(sb, KERN_DEBUG,
2233 "%s: truncating inode %lu to %lld bytes",
2234 __func__, inode->i_ino, inode->i_size);
2235 jbd_debug(2, "truncating inode %lu to %lld bytes\n",
2236 inode->i_ino, inode->i_size);
2237 mutex_lock(&inode->i_mutex);
2238 truncate_inode_pages(inode->i_mapping, inode->i_size);
2239 ext4_truncate(inode);
2240 mutex_unlock(&inode->i_mutex);
2243 if (test_opt(sb, DEBUG))
2244 ext4_msg(sb, KERN_DEBUG,
2245 "%s: deleting unreferenced inode %lu",
2246 __func__, inode->i_ino);
2247 jbd_debug(2, "deleting unreferenced inode %lu\n",
2251 iput(inode); /* The delete magic happens here! */
2254 #define PLURAL(x) (x), ((x) == 1) ? "" : "s"
2257 ext4_msg(sb, KERN_INFO, "%d orphan inode%s deleted",
2258 PLURAL(nr_orphans));
2260 ext4_msg(sb, KERN_INFO, "%d truncate%s cleaned up",
2261 PLURAL(nr_truncates));
2263 /* Turn quotas off */
2264 for (i = 0; i < EXT4_MAXQUOTAS; i++) {
2265 if (sb_dqopt(sb)->files[i])
2266 dquot_quota_off(sb, i);
2269 sb->s_flags = s_flags; /* Restore MS_RDONLY status */
2273 * Maximal extent format file size.
2274 * Resulting logical blkno at s_maxbytes must fit in our on-disk
2275 * extent format containers, within a sector_t, and within i_blocks
2276 * in the vfs. ext4 inode has 48 bits of i_block in fsblock units,
2277 * so that won't be a limiting factor.
2279 * However there is other limiting factor. We do store extents in the form
2280 * of starting block and length, hence the resulting length of the extent
2281 * covering maximum file size must fit into on-disk format containers as
2282 * well. Given that length is always by 1 unit bigger than max unit (because
2283 * we count 0 as well) we have to lower the s_maxbytes by one fs block.
2285 * Note, this does *not* consider any metadata overhead for vfs i_blocks.
2287 static loff_t ext4_max_size(int blkbits, int has_huge_files)
2290 loff_t upper_limit = MAX_LFS_FILESIZE;
2292 /* small i_blocks in vfs inode? */
2293 if (!has_huge_files || sizeof(blkcnt_t) < sizeof(u64)) {
2295 * CONFIG_LBDAF is not enabled implies the inode
2296 * i_block represent total blocks in 512 bytes
2297 * 32 == size of vfs inode i_blocks * 8
2299 upper_limit = (1LL << 32) - 1;
2301 /* total blocks in file system block size */
2302 upper_limit >>= (blkbits - 9);
2303 upper_limit <<= blkbits;
2307 * 32-bit extent-start container, ee_block. We lower the maxbytes
2308 * by one fs block, so ee_len can cover the extent of maximum file
2311 res = (1LL << 32) - 1;
2314 /* Sanity check against vm- & vfs- imposed limits */
2315 if (res > upper_limit)
2322 * Maximal bitmap file size. There is a direct, and {,double-,triple-}indirect
2323 * block limit, and also a limit of (2^48 - 1) 512-byte sectors in i_blocks.
2324 * We need to be 1 filesystem block less than the 2^48 sector limit.
2326 static loff_t ext4_max_bitmap_size(int bits, int has_huge_files)
2328 loff_t res = EXT4_NDIR_BLOCKS;
2331 /* This is calculated to be the largest file size for a dense, block
2332 * mapped file such that the file's total number of 512-byte sectors,
2333 * including data and all indirect blocks, does not exceed (2^48 - 1).
2335 * __u32 i_blocks_lo and _u16 i_blocks_high represent the total
2336 * number of 512-byte sectors of the file.
2339 if (!has_huge_files || sizeof(blkcnt_t) < sizeof(u64)) {
2341 * !has_huge_files or CONFIG_LBDAF not enabled implies that
2342 * the inode i_block field represents total file blocks in
2343 * 2^32 512-byte sectors == size of vfs inode i_blocks * 8
2345 upper_limit = (1LL << 32) - 1;
2347 /* total blocks in file system block size */
2348 upper_limit >>= (bits - 9);
2352 * We use 48 bit ext4_inode i_blocks
2353 * With EXT4_HUGE_FILE_FL set the i_blocks
2354 * represent total number of blocks in
2355 * file system block size
2357 upper_limit = (1LL << 48) - 1;
2361 /* indirect blocks */
2363 /* double indirect blocks */
2364 meta_blocks += 1 + (1LL << (bits-2));
2365 /* tripple indirect blocks */
2366 meta_blocks += 1 + (1LL << (bits-2)) + (1LL << (2*(bits-2)));
2368 upper_limit -= meta_blocks;
2369 upper_limit <<= bits;
2371 res += 1LL << (bits-2);
2372 res += 1LL << (2*(bits-2));
2373 res += 1LL << (3*(bits-2));
2375 if (res > upper_limit)
2378 if (res > MAX_LFS_FILESIZE)
2379 res = MAX_LFS_FILESIZE;
2384 static ext4_fsblk_t descriptor_loc(struct super_block *sb,
2385 ext4_fsblk_t logical_sb_block, int nr)
2387 struct ext4_sb_info *sbi = EXT4_SB(sb);
2388 ext4_group_t bg, first_meta_bg;
2391 first_meta_bg = le32_to_cpu(sbi->s_es->s_first_meta_bg);
2393 if (!EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_META_BG) ||
2395 return logical_sb_block + nr + 1;
2396 bg = sbi->s_desc_per_block * nr;
2397 if (ext4_bg_has_super(sb, bg))
2401 * If we have a meta_bg fs with 1k blocks, group 0's GDT is at
2402 * block 2, not 1. If s_first_data_block == 0 (bigalloc is enabled
2403 * on modern mke2fs or blksize > 1k on older mke2fs) then we must
2406 if (sb->s_blocksize == 1024 && nr == 0 &&
2407 le32_to_cpu(EXT4_SB(sb)->s_es->s_first_data_block) == 0)
2410 return (has_super + ext4_group_first_block_no(sb, bg));
2414 * ext4_get_stripe_size: Get the stripe size.
2415 * @sbi: In memory super block info
2417 * If we have specified it via mount option, then
2418 * use the mount option value. If the value specified at mount time is
2419 * greater than the blocks per group use the super block value.
2420 * If the super block value is greater than blocks per group return 0.
2421 * Allocator needs it be less than blocks per group.
2424 static unsigned long ext4_get_stripe_size(struct ext4_sb_info *sbi)
2426 unsigned long stride = le16_to_cpu(sbi->s_es->s_raid_stride);
2427 unsigned long stripe_width =
2428 le32_to_cpu(sbi->s_es->s_raid_stripe_width);
2431 if (sbi->s_stripe && sbi->s_stripe <= sbi->s_blocks_per_group)
2432 ret = sbi->s_stripe;
2433 else if (stripe_width <= sbi->s_blocks_per_group)
2435 else if (stride <= sbi->s_blocks_per_group)
2441 * If the stripe width is 1, this makes no sense and
2442 * we set it to 0 to turn off stripe handling code.
2451 * Check whether this filesystem can be mounted based on
2452 * the features present and the RDONLY/RDWR mount requested.
2453 * Returns 1 if this filesystem can be mounted as requested,
2454 * 0 if it cannot be.
2456 static int ext4_feature_set_ok(struct super_block *sb, int readonly)
2458 if (EXT4_HAS_INCOMPAT_FEATURE(sb, ~EXT4_FEATURE_INCOMPAT_SUPP)) {
2459 ext4_msg(sb, KERN_ERR,
2460 "Couldn't mount because of "
2461 "unsupported optional features (%x)",
2462 (le32_to_cpu(EXT4_SB(sb)->s_es->s_feature_incompat) &
2463 ~EXT4_FEATURE_INCOMPAT_SUPP));
2470 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_READONLY)) {
2471 ext4_msg(sb, KERN_INFO, "filesystem is read-only");
2472 sb->s_flags |= MS_RDONLY;
2476 /* Check that feature set is OK for a read-write mount */
2477 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, ~EXT4_FEATURE_RO_COMPAT_SUPP)) {
2478 ext4_msg(sb, KERN_ERR, "couldn't mount RDWR because of "
2479 "unsupported optional features (%x)",
2480 (le32_to_cpu(EXT4_SB(sb)->s_es->s_feature_ro_compat) &
2481 ~EXT4_FEATURE_RO_COMPAT_SUPP));
2485 * Large file size enabled file system can only be mounted
2486 * read-write on 32-bit systems if kernel is built with CONFIG_LBDAF
2488 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_HUGE_FILE)) {
2489 if (sizeof(blkcnt_t) < sizeof(u64)) {
2490 ext4_msg(sb, KERN_ERR, "Filesystem with huge files "
2491 "cannot be mounted RDWR without "
2496 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_BIGALLOC) &&
2497 !EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_EXTENTS)) {
2498 ext4_msg(sb, KERN_ERR,
2499 "Can't support bigalloc feature without "
2500 "extents feature\n");
2504 #ifndef CONFIG_QUOTA
2505 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_QUOTA) &&
2507 ext4_msg(sb, KERN_ERR,
2508 "Filesystem with quota feature cannot be mounted RDWR "
2509 "without CONFIG_QUOTA");
2512 #endif /* CONFIG_QUOTA */
2517 * This function is called once a day if we have errors logged
2518 * on the file system
2520 static void print_daily_error_info(unsigned long arg)
2522 struct super_block *sb = (struct super_block *) arg;
2523 struct ext4_sb_info *sbi;
2524 struct ext4_super_block *es;
2529 if (es->s_error_count)
2530 /* fsck newer than v1.41.13 is needed to clean this condition. */
2531 ext4_msg(sb, KERN_NOTICE, "error count since last fsck: %u",
2532 le32_to_cpu(es->s_error_count));
2533 if (es->s_first_error_time) {
2534 printk(KERN_NOTICE "EXT4-fs (%s): initial error at time %u: %.*s:%d",
2535 sb->s_id, le32_to_cpu(es->s_first_error_time),
2536 (int) sizeof(es->s_first_error_func),
2537 es->s_first_error_func,
2538 le32_to_cpu(es->s_first_error_line));
2539 if (es->s_first_error_ino)
2540 printk(": inode %u",
2541 le32_to_cpu(es->s_first_error_ino));
2542 if (es->s_first_error_block)
2543 printk(": block %llu", (unsigned long long)
2544 le64_to_cpu(es->s_first_error_block));
2547 if (es->s_last_error_time) {
2548 printk(KERN_NOTICE "EXT4-fs (%s): last error at time %u: %.*s:%d",
2549 sb->s_id, le32_to_cpu(es->s_last_error_time),
2550 (int) sizeof(es->s_last_error_func),
2551 es->s_last_error_func,
2552 le32_to_cpu(es->s_last_error_line));
2553 if (es->s_last_error_ino)
2554 printk(": inode %u",
2555 le32_to_cpu(es->s_last_error_ino));
2556 if (es->s_last_error_block)
2557 printk(": block %llu", (unsigned long long)
2558 le64_to_cpu(es->s_last_error_block));
2561 mod_timer(&sbi->s_err_report, jiffies + 24*60*60*HZ); /* Once a day */
2564 /* Find next suitable group and run ext4_init_inode_table */
2565 static int ext4_run_li_request(struct ext4_li_request *elr)
2567 struct ext4_group_desc *gdp = NULL;
2568 ext4_group_t group, ngroups;
2569 struct super_block *sb;
2570 unsigned long timeout = 0;
2574 ngroups = EXT4_SB(sb)->s_groups_count;
2577 for (group = elr->lr_next_group; group < ngroups; group++) {
2578 gdp = ext4_get_group_desc(sb, group, NULL);
2584 if (!(gdp->bg_flags & cpu_to_le16(EXT4_BG_INODE_ZEROED)))
2588 if (group >= ngroups)
2593 ret = ext4_init_inode_table(sb, group,
2594 elr->lr_timeout ? 0 : 1);
2595 if (elr->lr_timeout == 0) {
2596 timeout = (jiffies - timeout) *
2597 elr->lr_sbi->s_li_wait_mult;
2598 elr->lr_timeout = timeout;
2600 elr->lr_next_sched = jiffies + elr->lr_timeout;
2601 elr->lr_next_group = group + 1;
2609 * Remove lr_request from the list_request and free the
2610 * request structure. Should be called with li_list_mtx held
2612 static void ext4_remove_li_request(struct ext4_li_request *elr)
2614 struct ext4_sb_info *sbi;
2621 list_del(&elr->lr_request);
2622 sbi->s_li_request = NULL;
2626 static void ext4_unregister_li_request(struct super_block *sb)
2628 mutex_lock(&ext4_li_mtx);
2629 if (!ext4_li_info) {
2630 mutex_unlock(&ext4_li_mtx);
2634 mutex_lock(&ext4_li_info->li_list_mtx);
2635 ext4_remove_li_request(EXT4_SB(sb)->s_li_request);
2636 mutex_unlock(&ext4_li_info->li_list_mtx);
2637 mutex_unlock(&ext4_li_mtx);
2640 static struct task_struct *ext4_lazyinit_task;
2643 * This is the function where ext4lazyinit thread lives. It walks
2644 * through the request list searching for next scheduled filesystem.
2645 * When such a fs is found, run the lazy initialization request
2646 * (ext4_rn_li_request) and keep track of the time spend in this
2647 * function. Based on that time we compute next schedule time of
2648 * the request. When walking through the list is complete, compute
2649 * next waking time and put itself into sleep.
2651 static int ext4_lazyinit_thread(void *arg)
2653 struct ext4_lazy_init *eli = (struct ext4_lazy_init *)arg;
2654 struct list_head *pos, *n;
2655 struct ext4_li_request *elr;
2656 unsigned long next_wakeup, cur;
2658 BUG_ON(NULL == eli);
2662 next_wakeup = MAX_JIFFY_OFFSET;
2664 mutex_lock(&eli->li_list_mtx);
2665 if (list_empty(&eli->li_request_list)) {
2666 mutex_unlock(&eli->li_list_mtx);
2670 list_for_each_safe(pos, n, &eli->li_request_list) {
2671 elr = list_entry(pos, struct ext4_li_request,
2674 if (time_after_eq(jiffies, elr->lr_next_sched)) {
2675 if (ext4_run_li_request(elr) != 0) {
2676 /* error, remove the lazy_init job */
2677 ext4_remove_li_request(elr);
2682 if (time_before(elr->lr_next_sched, next_wakeup))
2683 next_wakeup = elr->lr_next_sched;
2685 mutex_unlock(&eli->li_list_mtx);
2690 if ((time_after_eq(cur, next_wakeup)) ||
2691 (MAX_JIFFY_OFFSET == next_wakeup)) {
2696 schedule_timeout_interruptible(next_wakeup - cur);
2698 if (kthread_should_stop()) {
2699 ext4_clear_request_list();
2706 * It looks like the request list is empty, but we need
2707 * to check it under the li_list_mtx lock, to prevent any
2708 * additions into it, and of course we should lock ext4_li_mtx
2709 * to atomically free the list and ext4_li_info, because at
2710 * this point another ext4 filesystem could be registering
2713 mutex_lock(&ext4_li_mtx);
2714 mutex_lock(&eli->li_list_mtx);
2715 if (!list_empty(&eli->li_request_list)) {
2716 mutex_unlock(&eli->li_list_mtx);
2717 mutex_unlock(&ext4_li_mtx);
2720 mutex_unlock(&eli->li_list_mtx);
2721 kfree(ext4_li_info);
2722 ext4_li_info = NULL;
2723 mutex_unlock(&ext4_li_mtx);
2728 static void ext4_clear_request_list(void)
2730 struct list_head *pos, *n;
2731 struct ext4_li_request *elr;
2733 mutex_lock(&ext4_li_info->li_list_mtx);
2734 list_for_each_safe(pos, n, &ext4_li_info->li_request_list) {
2735 elr = list_entry(pos, struct ext4_li_request,
2737 ext4_remove_li_request(elr);
2739 mutex_unlock(&ext4_li_info->li_list_mtx);
2742 static int ext4_run_lazyinit_thread(void)
2744 ext4_lazyinit_task = kthread_run(ext4_lazyinit_thread,
2745 ext4_li_info, "ext4lazyinit");
2746 if (IS_ERR(ext4_lazyinit_task)) {
2747 int err = PTR_ERR(ext4_lazyinit_task);
2748 ext4_clear_request_list();
2749 kfree(ext4_li_info);
2750 ext4_li_info = NULL;
2751 printk(KERN_CRIT "EXT4-fs: error %d creating inode table "
2752 "initialization thread\n",
2756 ext4_li_info->li_state |= EXT4_LAZYINIT_RUNNING;
2761 * Check whether it make sense to run itable init. thread or not.
2762 * If there is at least one uninitialized inode table, return
2763 * corresponding group number, else the loop goes through all
2764 * groups and return total number of groups.
2766 static ext4_group_t ext4_has_uninit_itable(struct super_block *sb)
2768 ext4_group_t group, ngroups = EXT4_SB(sb)->s_groups_count;
2769 struct ext4_group_desc *gdp = NULL;
2771 for (group = 0; group < ngroups; group++) {
2772 gdp = ext4_get_group_desc(sb, group, NULL);
2776 if (!(gdp->bg_flags & cpu_to_le16(EXT4_BG_INODE_ZEROED)))
2783 static int ext4_li_info_new(void)
2785 struct ext4_lazy_init *eli = NULL;
2787 eli = kzalloc(sizeof(*eli), GFP_KERNEL);
2791 INIT_LIST_HEAD(&eli->li_request_list);
2792 mutex_init(&eli->li_list_mtx);
2794 eli->li_state |= EXT4_LAZYINIT_QUIT;
2801 static struct ext4_li_request *ext4_li_request_new(struct super_block *sb,
2804 struct ext4_sb_info *sbi = EXT4_SB(sb);
2805 struct ext4_li_request *elr;
2807 elr = kzalloc(sizeof(*elr), GFP_KERNEL);
2813 elr->lr_next_group = start;
2816 * Randomize first schedule time of the request to
2817 * spread the inode table initialization requests
2820 elr->lr_next_sched = jiffies + (prandom_u32() %
2821 (EXT4_DEF_LI_MAX_START_DELAY * HZ));
2825 int ext4_register_li_request(struct super_block *sb,
2826 ext4_group_t first_not_zeroed)
2828 struct ext4_sb_info *sbi = EXT4_SB(sb);
2829 struct ext4_li_request *elr = NULL;
2830 ext4_group_t ngroups = EXT4_SB(sb)->s_groups_count;
2833 mutex_lock(&ext4_li_mtx);
2834 if (sbi->s_li_request != NULL) {
2836 * Reset timeout so it can be computed again, because
2837 * s_li_wait_mult might have changed.
2839 sbi->s_li_request->lr_timeout = 0;
2843 if (first_not_zeroed == ngroups ||
2844 (sb->s_flags & MS_RDONLY) ||
2845 !test_opt(sb, INIT_INODE_TABLE))
2848 elr = ext4_li_request_new(sb, first_not_zeroed);
2854 if (NULL == ext4_li_info) {
2855 ret = ext4_li_info_new();
2860 mutex_lock(&ext4_li_info->li_list_mtx);
2861 list_add(&elr->lr_request, &ext4_li_info->li_request_list);
2862 mutex_unlock(&ext4_li_info->li_list_mtx);
2864 sbi->s_li_request = elr;
2866 * set elr to NULL here since it has been inserted to
2867 * the request_list and the removal and free of it is
2868 * handled by ext4_clear_request_list from now on.
2872 if (!(ext4_li_info->li_state & EXT4_LAZYINIT_RUNNING)) {
2873 ret = ext4_run_lazyinit_thread();
2878 mutex_unlock(&ext4_li_mtx);
2885 * We do not need to lock anything since this is called on
2888 static void ext4_destroy_lazyinit_thread(void)
2891 * If thread exited earlier
2892 * there's nothing to be done.
2894 if (!ext4_li_info || !ext4_lazyinit_task)
2897 kthread_stop(ext4_lazyinit_task);
2900 static int set_journal_csum_feature_set(struct super_block *sb)
2903 int compat, incompat;
2904 struct ext4_sb_info *sbi = EXT4_SB(sb);
2906 if (ext4_has_metadata_csum(sb)) {
2907 /* journal checksum v3 */
2909 incompat = JBD2_FEATURE_INCOMPAT_CSUM_V3;
2911 /* journal checksum v1 */
2912 compat = JBD2_FEATURE_COMPAT_CHECKSUM;
2916 jbd2_journal_clear_features(sbi->s_journal,
2917 JBD2_FEATURE_COMPAT_CHECKSUM, 0,
2918 JBD2_FEATURE_INCOMPAT_CSUM_V3 |
2919 JBD2_FEATURE_INCOMPAT_CSUM_V2);
2920 if (test_opt(sb, JOURNAL_ASYNC_COMMIT)) {
2921 ret = jbd2_journal_set_features(sbi->s_journal,
2923 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT |
2925 } else if (test_opt(sb, JOURNAL_CHECKSUM)) {
2926 ret = jbd2_journal_set_features(sbi->s_journal,
2929 jbd2_journal_clear_features(sbi->s_journal, 0, 0,
2930 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT);
2932 jbd2_journal_clear_features(sbi->s_journal, 0, 0,
2933 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT);
2940 * Note: calculating the overhead so we can be compatible with
2941 * historical BSD practice is quite difficult in the face of
2942 * clusters/bigalloc. This is because multiple metadata blocks from
2943 * different block group can end up in the same allocation cluster.
2944 * Calculating the exact overhead in the face of clustered allocation
2945 * requires either O(all block bitmaps) in memory or O(number of block
2946 * groups**2) in time. We will still calculate the superblock for
2947 * older file systems --- and if we come across with a bigalloc file
2948 * system with zero in s_overhead_clusters the estimate will be close to
2949 * correct especially for very large cluster sizes --- but for newer
2950 * file systems, it's better to calculate this figure once at mkfs
2951 * time, and store it in the superblock. If the superblock value is
2952 * present (even for non-bigalloc file systems), we will use it.
2954 static int count_overhead(struct super_block *sb, ext4_group_t grp,
2957 struct ext4_sb_info *sbi = EXT4_SB(sb);
2958 struct ext4_group_desc *gdp;
2959 ext4_fsblk_t first_block, last_block, b;
2960 ext4_group_t i, ngroups = ext4_get_groups_count(sb);
2961 int s, j, count = 0;
2963 if (!EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_BIGALLOC))
2964 return (ext4_bg_has_super(sb, grp) + ext4_bg_num_gdb(sb, grp) +
2965 sbi->s_itb_per_group + 2);
2967 first_block = le32_to_cpu(sbi->s_es->s_first_data_block) +
2968 (grp * EXT4_BLOCKS_PER_GROUP(sb));
2969 last_block = first_block + EXT4_BLOCKS_PER_GROUP(sb) - 1;
2970 for (i = 0; i < ngroups; i++) {
2971 gdp = ext4_get_group_desc(sb, i, NULL);
2972 b = ext4_block_bitmap(sb, gdp);
2973 if (b >= first_block && b <= last_block) {
2974 ext4_set_bit(EXT4_B2C(sbi, b - first_block), buf);
2977 b = ext4_inode_bitmap(sb, gdp);
2978 if (b >= first_block && b <= last_block) {
2979 ext4_set_bit(EXT4_B2C(sbi, b - first_block), buf);
2982 b = ext4_inode_table(sb, gdp);
2983 if (b >= first_block && b + sbi->s_itb_per_group <= last_block)
2984 for (j = 0; j < sbi->s_itb_per_group; j++, b++) {
2985 int c = EXT4_B2C(sbi, b - first_block);
2986 ext4_set_bit(c, buf);
2992 if (ext4_bg_has_super(sb, grp)) {
2993 ext4_set_bit(s++, buf);
2996 for (j = ext4_bg_num_gdb(sb, grp); j > 0; j--) {
2997 ext4_set_bit(EXT4_B2C(sbi, s++), buf);
3003 return EXT4_CLUSTERS_PER_GROUP(sb) -
3004 ext4_count_free(buf, EXT4_CLUSTERS_PER_GROUP(sb) / 8);
3008 * Compute the overhead and stash it in sbi->s_overhead
3010 int ext4_calculate_overhead(struct super_block *sb)
3012 struct ext4_sb_info *sbi = EXT4_SB(sb);
3013 struct ext4_super_block *es = sbi->s_es;
3014 ext4_group_t i, ngroups = ext4_get_groups_count(sb);
3015 ext4_fsblk_t overhead = 0;
3016 char *buf = (char *) get_zeroed_page(GFP_NOFS);
3022 * Compute the overhead (FS structures). This is constant
3023 * for a given filesystem unless the number of block groups
3024 * changes so we cache the previous value until it does.
3028 * All of the blocks before first_data_block are overhead
3030 overhead = EXT4_B2C(sbi, le32_to_cpu(es->s_first_data_block));
3033 * Add the overhead found in each block group
3035 for (i = 0; i < ngroups; i++) {
3038 blks = count_overhead(sb, i, buf);
3041 memset(buf, 0, PAGE_SIZE);
3044 /* Add the internal journal blocks as well */
3045 if (sbi->s_journal && !sbi->journal_bdev)
3046 overhead += EXT4_NUM_B2C(sbi, sbi->s_journal->j_maxlen);
3048 sbi->s_overhead = overhead;
3050 free_page((unsigned long) buf);
3054 static void ext4_set_resv_clusters(struct super_block *sb)
3056 ext4_fsblk_t resv_clusters;
3057 struct ext4_sb_info *sbi = EXT4_SB(sb);
3060 * There's no need to reserve anything when we aren't using extents.
3061 * The space estimates are exact, there are no unwritten extents,
3062 * hole punching doesn't need new metadata... This is needed especially
3063 * to keep ext2/3 backward compatibility.
3065 if (!EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_EXTENTS))
3068 * By default we reserve 2% or 4096 clusters, whichever is smaller.
3069 * This should cover the situations where we can not afford to run
3070 * out of space like for example punch hole, or converting
3071 * unwritten extents in delalloc path. In most cases such
3072 * allocation would require 1, or 2 blocks, higher numbers are
3075 resv_clusters = (ext4_blocks_count(sbi->s_es) >>
3076 sbi->s_cluster_bits);
3078 do_div(resv_clusters, 50);
3079 resv_clusters = min_t(ext4_fsblk_t, resv_clusters, 4096);
3081 atomic64_set(&sbi->s_resv_clusters, resv_clusters);
3084 static int ext4_fill_super(struct super_block *sb, void *data, int silent)
3086 char *orig_data = kstrdup(data, GFP_KERNEL);
3087 struct buffer_head *bh;
3088 struct ext4_super_block *es = NULL;
3089 struct ext4_sb_info *sbi;
3091 ext4_fsblk_t sb_block = get_sb_block(&data);
3092 ext4_fsblk_t logical_sb_block;
3093 unsigned long offset = 0;
3094 unsigned long journal_devnum = 0;
3095 unsigned long def_mount_opts;
3099 int blocksize, clustersize;
3100 unsigned int db_count;
3102 int needs_recovery, has_huge_files, has_bigalloc;
3105 unsigned int journal_ioprio = DEFAULT_JOURNAL_IOPRIO;
3106 ext4_group_t first_not_zeroed;
3108 sbi = kzalloc(sizeof(*sbi), GFP_KERNEL);
3112 sbi->s_blockgroup_lock =
3113 kzalloc(sizeof(struct blockgroup_lock), GFP_KERNEL);
3114 if (!sbi->s_blockgroup_lock) {
3118 sb->s_fs_info = sbi;
3120 sbi->s_inode_readahead_blks = EXT4_DEF_INODE_READAHEAD_BLKS;
3121 sbi->s_sb_block = sb_block;
3122 if (sb->s_bdev->bd_part)
3123 sbi->s_sectors_written_start =
3124 part_stat_read(sb->s_bdev->bd_part, sectors[1]);
3126 /* Cleanup superblock name */
3127 strreplace(sb->s_id, '/', '!');
3129 /* -EINVAL is default */
3131 blocksize = sb_min_blocksize(sb, EXT4_MIN_BLOCK_SIZE);
3133 ext4_msg(sb, KERN_ERR, "unable to set blocksize");
3138 * The ext4 superblock will not be buffer aligned for other than 1kB
3139 * block sizes. We need to calculate the offset from buffer start.
3141 if (blocksize != EXT4_MIN_BLOCK_SIZE) {
3142 logical_sb_block = sb_block * EXT4_MIN_BLOCK_SIZE;
3143 offset = do_div(logical_sb_block, blocksize);
3145 logical_sb_block = sb_block;
3148 if (!(bh = sb_bread_unmovable(sb, logical_sb_block))) {
3149 ext4_msg(sb, KERN_ERR, "unable to read superblock");
3153 * Note: s_es must be initialized as soon as possible because
3154 * some ext4 macro-instructions depend on its value
3156 es = (struct ext4_super_block *) (bh->b_data + offset);
3158 sb->s_magic = le16_to_cpu(es->s_magic);
3159 if (sb->s_magic != EXT4_SUPER_MAGIC)
3161 sbi->s_kbytes_written = le64_to_cpu(es->s_kbytes_written);
3163 /* Warn if metadata_csum and gdt_csum are both set. */
3164 if (EXT4_HAS_RO_COMPAT_FEATURE(sb,
3165 EXT4_FEATURE_RO_COMPAT_METADATA_CSUM) &&
3166 EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_GDT_CSUM))
3167 ext4_warning(sb, "metadata_csum and uninit_bg are "
3168 "redundant flags; please run fsck.");
3170 /* Check for a known checksum algorithm */
3171 if (!ext4_verify_csum_type(sb, es)) {
3172 ext4_msg(sb, KERN_ERR, "VFS: Found ext4 filesystem with "
3173 "unknown checksum algorithm.");
3178 /* Load the checksum driver */
3179 if (EXT4_HAS_RO_COMPAT_FEATURE(sb,
3180 EXT4_FEATURE_RO_COMPAT_METADATA_CSUM)) {
3181 sbi->s_chksum_driver = crypto_alloc_shash("crc32c", 0, 0);
3182 if (IS_ERR(sbi->s_chksum_driver)) {
3183 ext4_msg(sb, KERN_ERR, "Cannot load crc32c driver.");
3184 ret = PTR_ERR(sbi->s_chksum_driver);
3185 sbi->s_chksum_driver = NULL;
3190 /* Check superblock checksum */
3191 if (!ext4_superblock_csum_verify(sb, es)) {
3192 ext4_msg(sb, KERN_ERR, "VFS: Found ext4 filesystem with "
3193 "invalid superblock checksum. Run e2fsck?");
3198 /* Precompute checksum seed for all metadata */
3199 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_CSUM_SEED))
3200 sbi->s_csum_seed = le32_to_cpu(es->s_checksum_seed);
3201 else if (ext4_has_metadata_csum(sb))
3202 sbi->s_csum_seed = ext4_chksum(sbi, ~0, es->s_uuid,
3203 sizeof(es->s_uuid));
3205 /* Set defaults before we parse the mount options */
3206 def_mount_opts = le32_to_cpu(es->s_default_mount_opts);
3207 set_opt(sb, INIT_INODE_TABLE);
3208 if (def_mount_opts & EXT4_DEFM_DEBUG)
3210 if (def_mount_opts & EXT4_DEFM_BSDGROUPS)
3212 if (def_mount_opts & EXT4_DEFM_UID16)
3213 set_opt(sb, NO_UID32);
3214 /* xattr user namespace & acls are now defaulted on */
3215 set_opt(sb, XATTR_USER);
3216 #ifdef CONFIG_EXT4_FS_POSIX_ACL
3217 set_opt(sb, POSIX_ACL);
3219 /* don't forget to enable journal_csum when metadata_csum is enabled. */
3220 if (ext4_has_metadata_csum(sb))
3221 set_opt(sb, JOURNAL_CHECKSUM);
3223 if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_DATA)
3224 set_opt(sb, JOURNAL_DATA);
3225 else if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_ORDERED)
3226 set_opt(sb, ORDERED_DATA);
3227 else if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_WBACK)
3228 set_opt(sb, WRITEBACK_DATA);
3230 if (le16_to_cpu(sbi->s_es->s_errors) == EXT4_ERRORS_PANIC)
3231 set_opt(sb, ERRORS_PANIC);
3232 else if (le16_to_cpu(sbi->s_es->s_errors) == EXT4_ERRORS_CONTINUE)
3233 set_opt(sb, ERRORS_CONT);
3235 set_opt(sb, ERRORS_RO);
3236 /* block_validity enabled by default; disable with noblock_validity */
3237 set_opt(sb, BLOCK_VALIDITY);
3238 if (def_mount_opts & EXT4_DEFM_DISCARD)
3239 set_opt(sb, DISCARD);
3241 sbi->s_resuid = make_kuid(&init_user_ns, le16_to_cpu(es->s_def_resuid));
3242 sbi->s_resgid = make_kgid(&init_user_ns, le16_to_cpu(es->s_def_resgid));
3243 sbi->s_commit_interval = JBD2_DEFAULT_MAX_COMMIT_AGE * HZ;
3244 sbi->s_min_batch_time = EXT4_DEF_MIN_BATCH_TIME;
3245 sbi->s_max_batch_time = EXT4_DEF_MAX_BATCH_TIME;
3247 if ((def_mount_opts & EXT4_DEFM_NOBARRIER) == 0)
3248 set_opt(sb, BARRIER);
3251 * enable delayed allocation by default
3252 * Use -o nodelalloc to turn it off
3254 if (!IS_EXT3_SB(sb) && !IS_EXT2_SB(sb) &&
3255 ((def_mount_opts & EXT4_DEFM_NODELALLOC) == 0))
3256 set_opt(sb, DELALLOC);
3259 * set default s_li_wait_mult for lazyinit, for the case there is
3260 * no mount option specified.
3262 sbi->s_li_wait_mult = EXT4_DEF_LI_WAIT_MULT;
3264 if (!parse_options((char *) sbi->s_es->s_mount_opts, sb,
3265 &journal_devnum, &journal_ioprio, 0)) {
3266 ext4_msg(sb, KERN_WARNING,
3267 "failed to parse options in superblock: %s",
3268 sbi->s_es->s_mount_opts);
3270 sbi->s_def_mount_opt = sbi->s_mount_opt;
3271 if (!parse_options((char *) data, sb, &journal_devnum,
3272 &journal_ioprio, 0))
3275 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA) {
3276 printk_once(KERN_WARNING "EXT4-fs: Warning: mounting "
3277 "with data=journal disables delayed "
3278 "allocation and O_DIRECT support!\n");
3279 if (test_opt2(sb, EXPLICIT_DELALLOC)) {
3280 ext4_msg(sb, KERN_ERR, "can't mount with "
3281 "both data=journal and delalloc");
3284 if (test_opt(sb, DIOREAD_NOLOCK)) {
3285 ext4_msg(sb, KERN_ERR, "can't mount with "
3286 "both data=journal and dioread_nolock");
3289 if (test_opt(sb, DAX)) {
3290 ext4_msg(sb, KERN_ERR, "can't mount with "
3291 "both data=journal and dax");
3294 if (test_opt(sb, DELALLOC))
3295 clear_opt(sb, DELALLOC);
3297 sb->s_iflags |= SB_I_CGROUPWB;
3300 sb->s_flags = (sb->s_flags & ~MS_POSIXACL) |
3301 (test_opt(sb, POSIX_ACL) ? MS_POSIXACL : 0);
3303 if (le32_to_cpu(es->s_rev_level) == EXT4_GOOD_OLD_REV &&
3304 (EXT4_HAS_COMPAT_FEATURE(sb, ~0U) ||
3305 EXT4_HAS_RO_COMPAT_FEATURE(sb, ~0U) ||
3306 EXT4_HAS_INCOMPAT_FEATURE(sb, ~0U)))
3307 ext4_msg(sb, KERN_WARNING,
3308 "feature flags set on rev 0 fs, "
3309 "running e2fsck is recommended");
3311 if (es->s_creator_os == cpu_to_le32(EXT4_OS_HURD)) {
3312 set_opt2(sb, HURD_COMPAT);
3313 if (EXT4_HAS_INCOMPAT_FEATURE(sb,
3314 EXT4_FEATURE_INCOMPAT_64BIT)) {
3315 ext4_msg(sb, KERN_ERR,
3316 "The Hurd can't support 64-bit file systems");
3321 if (IS_EXT2_SB(sb)) {
3322 if (ext2_feature_set_ok(sb))
3323 ext4_msg(sb, KERN_INFO, "mounting ext2 file system "
3324 "using the ext4 subsystem");
3326 ext4_msg(sb, KERN_ERR, "couldn't mount as ext2 due "
3327 "to feature incompatibilities");
3332 if (IS_EXT3_SB(sb)) {
3333 if (ext3_feature_set_ok(sb))
3334 ext4_msg(sb, KERN_INFO, "mounting ext3 file system "
3335 "using the ext4 subsystem");
3337 ext4_msg(sb, KERN_ERR, "couldn't mount as ext3 due "
3338 "to feature incompatibilities");
3344 * Check feature flags regardless of the revision level, since we
3345 * previously didn't change the revision level when setting the flags,
3346 * so there is a chance incompat flags are set on a rev 0 filesystem.
3348 if (!ext4_feature_set_ok(sb, (sb->s_flags & MS_RDONLY)))
3351 blocksize = BLOCK_SIZE << le32_to_cpu(es->s_log_block_size);
3352 if (blocksize < EXT4_MIN_BLOCK_SIZE ||
3353 blocksize > EXT4_MAX_BLOCK_SIZE) {
3354 ext4_msg(sb, KERN_ERR,
3355 "Unsupported filesystem blocksize %d", blocksize);
3359 if (sbi->s_mount_opt & EXT4_MOUNT_DAX) {
3360 if (blocksize != PAGE_SIZE) {
3361 ext4_msg(sb, KERN_ERR,
3362 "error: unsupported blocksize for dax");
3365 if (!sb->s_bdev->bd_disk->fops->direct_access) {
3366 ext4_msg(sb, KERN_ERR,
3367 "error: device does not support dax");
3372 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_ENCRYPT) &&
3373 es->s_encryption_level) {
3374 ext4_msg(sb, KERN_ERR, "Unsupported encryption level %d",
3375 es->s_encryption_level);
3379 if (sb->s_blocksize != blocksize) {
3380 /* Validate the filesystem blocksize */
3381 if (!sb_set_blocksize(sb, blocksize)) {
3382 ext4_msg(sb, KERN_ERR, "bad block size %d",
3388 logical_sb_block = sb_block * EXT4_MIN_BLOCK_SIZE;
3389 offset = do_div(logical_sb_block, blocksize);
3390 bh = sb_bread_unmovable(sb, logical_sb_block);
3392 ext4_msg(sb, KERN_ERR,
3393 "Can't read superblock on 2nd try");
3396 es = (struct ext4_super_block *)(bh->b_data + offset);
3398 if (es->s_magic != cpu_to_le16(EXT4_SUPER_MAGIC)) {
3399 ext4_msg(sb, KERN_ERR,
3400 "Magic mismatch, very weird!");
3405 has_huge_files = EXT4_HAS_RO_COMPAT_FEATURE(sb,
3406 EXT4_FEATURE_RO_COMPAT_HUGE_FILE);
3407 sbi->s_bitmap_maxbytes = ext4_max_bitmap_size(sb->s_blocksize_bits,
3409 sb->s_maxbytes = ext4_max_size(sb->s_blocksize_bits, has_huge_files);
3411 if (le32_to_cpu(es->s_rev_level) == EXT4_GOOD_OLD_REV) {
3412 sbi->s_inode_size = EXT4_GOOD_OLD_INODE_SIZE;
3413 sbi->s_first_ino = EXT4_GOOD_OLD_FIRST_INO;
3415 sbi->s_inode_size = le16_to_cpu(es->s_inode_size);
3416 sbi->s_first_ino = le32_to_cpu(es->s_first_ino);
3417 if ((sbi->s_inode_size < EXT4_GOOD_OLD_INODE_SIZE) ||
3418 (!is_power_of_2(sbi->s_inode_size)) ||
3419 (sbi->s_inode_size > blocksize)) {
3420 ext4_msg(sb, KERN_ERR,
3421 "unsupported inode size: %d",
3425 if (sbi->s_inode_size > EXT4_GOOD_OLD_INODE_SIZE)
3426 sb->s_time_gran = 1 << (EXT4_EPOCH_BITS - 2);
3429 sbi->s_desc_size = le16_to_cpu(es->s_desc_size);
3430 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_64BIT)) {
3431 if (sbi->s_desc_size < EXT4_MIN_DESC_SIZE_64BIT ||
3432 sbi->s_desc_size > EXT4_MAX_DESC_SIZE ||
3433 !is_power_of_2(sbi->s_desc_size)) {
3434 ext4_msg(sb, KERN_ERR,
3435 "unsupported descriptor size %lu",
3440 sbi->s_desc_size = EXT4_MIN_DESC_SIZE;
3442 sbi->s_blocks_per_group = le32_to_cpu(es->s_blocks_per_group);
3443 sbi->s_inodes_per_group = le32_to_cpu(es->s_inodes_per_group);
3444 if (EXT4_INODE_SIZE(sb) == 0 || EXT4_INODES_PER_GROUP(sb) == 0)
3447 sbi->s_inodes_per_block = blocksize / EXT4_INODE_SIZE(sb);
3448 if (sbi->s_inodes_per_block == 0)
3450 sbi->s_itb_per_group = sbi->s_inodes_per_group /
3451 sbi->s_inodes_per_block;
3452 sbi->s_desc_per_block = blocksize / EXT4_DESC_SIZE(sb);
3454 sbi->s_mount_state = le16_to_cpu(es->s_state);
3455 sbi->s_addr_per_block_bits = ilog2(EXT4_ADDR_PER_BLOCK(sb));
3456 sbi->s_desc_per_block_bits = ilog2(EXT4_DESC_PER_BLOCK(sb));
3458 for (i = 0; i < 4; i++)
3459 sbi->s_hash_seed[i] = le32_to_cpu(es->s_hash_seed[i]);
3460 sbi->s_def_hash_version = es->s_def_hash_version;
3461 if (EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_DIR_INDEX)) {
3462 i = le32_to_cpu(es->s_flags);
3463 if (i & EXT2_FLAGS_UNSIGNED_HASH)
3464 sbi->s_hash_unsigned = 3;
3465 else if ((i & EXT2_FLAGS_SIGNED_HASH) == 0) {
3466 #ifdef __CHAR_UNSIGNED__
3467 if (!(sb->s_flags & MS_RDONLY))
3469 cpu_to_le32(EXT2_FLAGS_UNSIGNED_HASH);
3470 sbi->s_hash_unsigned = 3;
3472 if (!(sb->s_flags & MS_RDONLY))
3474 cpu_to_le32(EXT2_FLAGS_SIGNED_HASH);
3479 /* Handle clustersize */
3480 clustersize = BLOCK_SIZE << le32_to_cpu(es->s_log_cluster_size);
3481 has_bigalloc = EXT4_HAS_RO_COMPAT_FEATURE(sb,
3482 EXT4_FEATURE_RO_COMPAT_BIGALLOC);
3484 if (clustersize < blocksize) {
3485 ext4_msg(sb, KERN_ERR,
3486 "cluster size (%d) smaller than "
3487 "block size (%d)", clustersize, blocksize);
3490 sbi->s_cluster_bits = le32_to_cpu(es->s_log_cluster_size) -
3491 le32_to_cpu(es->s_log_block_size);
3492 sbi->s_clusters_per_group =
3493 le32_to_cpu(es->s_clusters_per_group);
3494 if (sbi->s_clusters_per_group > blocksize * 8) {
3495 ext4_msg(sb, KERN_ERR,
3496 "#clusters per group too big: %lu",
3497 sbi->s_clusters_per_group);
3500 if (sbi->s_blocks_per_group !=
3501 (sbi->s_clusters_per_group * (clustersize / blocksize))) {
3502 ext4_msg(sb, KERN_ERR, "blocks per group (%lu) and "
3503 "clusters per group (%lu) inconsistent",
3504 sbi->s_blocks_per_group,
3505 sbi->s_clusters_per_group);
3509 if (clustersize != blocksize) {
3510 ext4_warning(sb, "fragment/cluster size (%d) != "
3511 "block size (%d)", clustersize,
3513 clustersize = blocksize;
3515 if (sbi->s_blocks_per_group > blocksize * 8) {
3516 ext4_msg(sb, KERN_ERR,
3517 "#blocks per group too big: %lu",
3518 sbi->s_blocks_per_group);
3521 sbi->s_clusters_per_group = sbi->s_blocks_per_group;
3522 sbi->s_cluster_bits = 0;
3524 sbi->s_cluster_ratio = clustersize / blocksize;
3526 if (sbi->s_inodes_per_group > blocksize * 8) {
3527 ext4_msg(sb, KERN_ERR,
3528 "#inodes per group too big: %lu",
3529 sbi->s_inodes_per_group);
3533 /* Do we have standard group size of clustersize * 8 blocks ? */
3534 if (sbi->s_blocks_per_group == clustersize << 3)
3535 set_opt2(sb, STD_GROUP_SIZE);
3538 * Test whether we have more sectors than will fit in sector_t,
3539 * and whether the max offset is addressable by the page cache.
3541 err = generic_check_addressable(sb->s_blocksize_bits,
3542 ext4_blocks_count(es));
3544 ext4_msg(sb, KERN_ERR, "filesystem"
3545 " too large to mount safely on this system");
3546 if (sizeof(sector_t) < 8)
3547 ext4_msg(sb, KERN_WARNING, "CONFIG_LBDAF not enabled");
3551 if (EXT4_BLOCKS_PER_GROUP(sb) == 0)
3554 /* check blocks count against device size */
3555 blocks_count = sb->s_bdev->bd_inode->i_size >> sb->s_blocksize_bits;
3556 if (blocks_count && ext4_blocks_count(es) > blocks_count) {
3557 ext4_msg(sb, KERN_WARNING, "bad geometry: block count %llu "
3558 "exceeds size of device (%llu blocks)",
3559 ext4_blocks_count(es), blocks_count);
3564 * It makes no sense for the first data block to be beyond the end
3565 * of the filesystem.
3567 if (le32_to_cpu(es->s_first_data_block) >= ext4_blocks_count(es)) {
3568 ext4_msg(sb, KERN_WARNING, "bad geometry: first data "
3569 "block %u is beyond end of filesystem (%llu)",
3570 le32_to_cpu(es->s_first_data_block),
3571 ext4_blocks_count(es));
3574 blocks_count = (ext4_blocks_count(es) -
3575 le32_to_cpu(es->s_first_data_block) +
3576 EXT4_BLOCKS_PER_GROUP(sb) - 1);
3577 do_div(blocks_count, EXT4_BLOCKS_PER_GROUP(sb));
3578 if (blocks_count > ((uint64_t)1<<32) - EXT4_DESC_PER_BLOCK(sb)) {
3579 ext4_msg(sb, KERN_WARNING, "groups count too large: %u "
3580 "(block count %llu, first data block %u, "
3581 "blocks per group %lu)", sbi->s_groups_count,
3582 ext4_blocks_count(es),
3583 le32_to_cpu(es->s_first_data_block),
3584 EXT4_BLOCKS_PER_GROUP(sb));
3587 sbi->s_groups_count = blocks_count;
3588 sbi->s_blockfile_groups = min_t(ext4_group_t, sbi->s_groups_count,
3589 (EXT4_MAX_BLOCK_FILE_PHYS / EXT4_BLOCKS_PER_GROUP(sb)));
3590 db_count = (sbi->s_groups_count + EXT4_DESC_PER_BLOCK(sb) - 1) /
3591 EXT4_DESC_PER_BLOCK(sb);
3592 sbi->s_group_desc = ext4_kvmalloc(db_count *
3593 sizeof(struct buffer_head *),
3595 if (sbi->s_group_desc == NULL) {
3596 ext4_msg(sb, KERN_ERR, "not enough memory");
3601 bgl_lock_init(sbi->s_blockgroup_lock);
3603 for (i = 0; i < db_count; i++) {
3604 block = descriptor_loc(sb, logical_sb_block, i);
3605 sbi->s_group_desc[i] = sb_bread_unmovable(sb, block);
3606 if (!sbi->s_group_desc[i]) {
3607 ext4_msg(sb, KERN_ERR,
3608 "can't read group descriptor %d", i);
3613 if (!ext4_check_descriptors(sb, &first_not_zeroed)) {
3614 ext4_msg(sb, KERN_ERR, "group descriptors corrupted!");
3618 sbi->s_gdb_count = db_count;
3619 get_random_bytes(&sbi->s_next_generation, sizeof(u32));
3620 spin_lock_init(&sbi->s_next_gen_lock);
3622 setup_timer(&sbi->s_err_report, print_daily_error_info,
3623 (unsigned long) sb);
3625 /* Register extent status tree shrinker */
3626 if (ext4_es_register_shrinker(sbi))
3629 sbi->s_stripe = ext4_get_stripe_size(sbi);
3630 sbi->s_extent_max_zeroout_kb = 32;
3633 * set up enough so that it can read an inode
3635 sb->s_op = &ext4_sops;
3636 sb->s_export_op = &ext4_export_ops;
3637 sb->s_xattr = ext4_xattr_handlers;
3639 sb->dq_op = &ext4_quota_operations;
3640 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_QUOTA))
3641 sb->s_qcop = &dquot_quotactl_sysfile_ops;
3643 sb->s_qcop = &ext4_qctl_operations;
3644 sb->s_quota_types = QTYPE_MASK_USR | QTYPE_MASK_GRP;
3646 memcpy(sb->s_uuid, es->s_uuid, sizeof(es->s_uuid));
3648 INIT_LIST_HEAD(&sbi->s_orphan); /* unlinked but open files */
3649 mutex_init(&sbi->s_orphan_lock);
3653 needs_recovery = (es->s_last_orphan != 0 ||
3654 EXT4_HAS_INCOMPAT_FEATURE(sb,
3655 EXT4_FEATURE_INCOMPAT_RECOVER));
3657 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_MMP) &&
3658 !(sb->s_flags & MS_RDONLY))
3659 if (ext4_multi_mount_protect(sb, le64_to_cpu(es->s_mmp_block)))
3660 goto failed_mount3a;
3663 * The first inode we look at is the journal inode. Don't try
3664 * root first: it may be modified in the journal!
3666 if (!test_opt(sb, NOLOAD) &&
3667 EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL)) {
3668 if (ext4_load_journal(sb, es, journal_devnum))
3669 goto failed_mount3a;
3670 } else if (test_opt(sb, NOLOAD) && !(sb->s_flags & MS_RDONLY) &&
3671 EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER)) {
3672 ext4_msg(sb, KERN_ERR, "required journal recovery "
3673 "suppressed and not mounted read-only");
3674 goto failed_mount_wq;
3676 clear_opt(sb, DATA_FLAGS);
3677 sbi->s_journal = NULL;
3682 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_64BIT) &&
3683 !jbd2_journal_set_features(EXT4_SB(sb)->s_journal, 0, 0,
3684 JBD2_FEATURE_INCOMPAT_64BIT)) {
3685 ext4_msg(sb, KERN_ERR, "Failed to set 64-bit journal feature");
3686 goto failed_mount_wq;
3689 if (!set_journal_csum_feature_set(sb)) {
3690 ext4_msg(sb, KERN_ERR, "Failed to set journal checksum "
3692 goto failed_mount_wq;
3695 /* We have now updated the journal if required, so we can
3696 * validate the data journaling mode. */
3697 switch (test_opt(sb, DATA_FLAGS)) {
3699 /* No mode set, assume a default based on the journal
3700 * capabilities: ORDERED_DATA if the journal can
3701 * cope, else JOURNAL_DATA
3703 if (jbd2_journal_check_available_features
3704 (sbi->s_journal, 0, 0, JBD2_FEATURE_INCOMPAT_REVOKE))
3705 set_opt(sb, ORDERED_DATA);
3707 set_opt(sb, JOURNAL_DATA);
3710 case EXT4_MOUNT_ORDERED_DATA:
3711 case EXT4_MOUNT_WRITEBACK_DATA:
3712 if (!jbd2_journal_check_available_features
3713 (sbi->s_journal, 0, 0, JBD2_FEATURE_INCOMPAT_REVOKE)) {
3714 ext4_msg(sb, KERN_ERR, "Journal does not support "
3715 "requested data journaling mode");
3716 goto failed_mount_wq;
3721 set_task_ioprio(sbi->s_journal->j_task, journal_ioprio);
3723 sbi->s_journal->j_commit_callback = ext4_journal_commit_callback;
3726 if (ext4_mballoc_ready) {
3727 sbi->s_mb_cache = ext4_xattr_create_cache(sb->s_id);
3728 if (!sbi->s_mb_cache) {
3729 ext4_msg(sb, KERN_ERR, "Failed to create an mb_cache");
3730 goto failed_mount_wq;
3734 if ((DUMMY_ENCRYPTION_ENABLED(sbi) ||
3735 EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_ENCRYPT)) &&
3736 (blocksize != PAGE_CACHE_SIZE)) {
3737 ext4_msg(sb, KERN_ERR,
3738 "Unsupported blocksize for fs encryption");
3739 goto failed_mount_wq;
3742 if (DUMMY_ENCRYPTION_ENABLED(sbi) &&
3743 !(sb->s_flags & MS_RDONLY) &&
3744 !EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_ENCRYPT)) {
3745 EXT4_SET_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_ENCRYPT);
3746 ext4_commit_super(sb, 1);
3750 * Get the # of file system overhead blocks from the
3751 * superblock if present.
3753 if (es->s_overhead_clusters)
3754 sbi->s_overhead = le32_to_cpu(es->s_overhead_clusters);
3756 err = ext4_calculate_overhead(sb);
3758 goto failed_mount_wq;
3762 * The maximum number of concurrent works can be high and
3763 * concurrency isn't really necessary. Limit it to 1.
3765 EXT4_SB(sb)->rsv_conversion_wq =
3766 alloc_workqueue("ext4-rsv-conversion", WQ_MEM_RECLAIM | WQ_UNBOUND, 1);
3767 if (!EXT4_SB(sb)->rsv_conversion_wq) {
3768 printk(KERN_ERR "EXT4-fs: failed to create workqueue\n");
3774 * The jbd2_journal_load will have done any necessary log recovery,
3775 * so we can safely mount the rest of the filesystem now.
3778 root = ext4_iget(sb, EXT4_ROOT_INO);
3780 ext4_msg(sb, KERN_ERR, "get root inode failed");
3781 ret = PTR_ERR(root);
3785 if (!S_ISDIR(root->i_mode) || !root->i_blocks || !root->i_size) {
3786 ext4_msg(sb, KERN_ERR, "corrupt root inode, run e2fsck");
3790 sb->s_root = d_make_root(root);
3792 ext4_msg(sb, KERN_ERR, "get root dentry failed");
3797 if (ext4_setup_super(sb, es, sb->s_flags & MS_RDONLY))
3798 sb->s_flags |= MS_RDONLY;
3800 /* determine the minimum size of new large inodes, if present */
3801 if (sbi->s_inode_size > EXT4_GOOD_OLD_INODE_SIZE) {
3802 sbi->s_want_extra_isize = sizeof(struct ext4_inode) -
3803 EXT4_GOOD_OLD_INODE_SIZE;
3804 if (EXT4_HAS_RO_COMPAT_FEATURE(sb,
3805 EXT4_FEATURE_RO_COMPAT_EXTRA_ISIZE)) {
3806 if (sbi->s_want_extra_isize <
3807 le16_to_cpu(es->s_want_extra_isize))
3808 sbi->s_want_extra_isize =
3809 le16_to_cpu(es->s_want_extra_isize);
3810 if (sbi->s_want_extra_isize <
3811 le16_to_cpu(es->s_min_extra_isize))
3812 sbi->s_want_extra_isize =
3813 le16_to_cpu(es->s_min_extra_isize);
3816 /* Check if enough inode space is available */
3817 if (EXT4_GOOD_OLD_INODE_SIZE + sbi->s_want_extra_isize >
3818 sbi->s_inode_size) {
3819 sbi->s_want_extra_isize = sizeof(struct ext4_inode) -
3820 EXT4_GOOD_OLD_INODE_SIZE;
3821 ext4_msg(sb, KERN_INFO, "required extra inode space not"
3825 ext4_set_resv_clusters(sb);
3827 err = ext4_setup_system_zone(sb);
3829 ext4_msg(sb, KERN_ERR, "failed to initialize system "
3831 goto failed_mount4a;
3835 err = ext4_mb_init(sb);
3837 ext4_msg(sb, KERN_ERR, "failed to initialize mballoc (%d)",
3842 block = ext4_count_free_clusters(sb);
3843 ext4_free_blocks_count_set(sbi->s_es,
3844 EXT4_C2B(sbi, block));
3845 err = percpu_counter_init(&sbi->s_freeclusters_counter, block,
3848 unsigned long freei = ext4_count_free_inodes(sb);
3849 sbi->s_es->s_free_inodes_count = cpu_to_le32(freei);
3850 err = percpu_counter_init(&sbi->s_freeinodes_counter, freei,
3854 err = percpu_counter_init(&sbi->s_dirs_counter,
3855 ext4_count_dirs(sb), GFP_KERNEL);
3857 err = percpu_counter_init(&sbi->s_dirtyclusters_counter, 0,
3860 ext4_msg(sb, KERN_ERR, "insufficient memory");
3864 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_FLEX_BG))
3865 if (!ext4_fill_flex_info(sb)) {
3866 ext4_msg(sb, KERN_ERR,
3867 "unable to initialize "
3868 "flex_bg meta info!");
3872 err = ext4_register_li_request(sb, first_not_zeroed);
3876 err = ext4_register_sysfs(sb);
3881 /* Enable quota usage during mount. */
3882 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_QUOTA) &&
3883 !(sb->s_flags & MS_RDONLY)) {
3884 err = ext4_enable_quotas(sb);
3888 #endif /* CONFIG_QUOTA */
3890 EXT4_SB(sb)->s_mount_state |= EXT4_ORPHAN_FS;
3891 ext4_orphan_cleanup(sb, es);
3892 EXT4_SB(sb)->s_mount_state &= ~EXT4_ORPHAN_FS;
3893 if (needs_recovery) {
3894 ext4_msg(sb, KERN_INFO, "recovery complete");
3895 ext4_mark_recovery_complete(sb, es);
3897 if (EXT4_SB(sb)->s_journal) {
3898 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA)
3899 descr = " journalled data mode";
3900 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA)
3901 descr = " ordered data mode";
3903 descr = " writeback data mode";
3905 descr = "out journal";
3907 if (test_opt(sb, DISCARD)) {
3908 struct request_queue *q = bdev_get_queue(sb->s_bdev);
3909 if (!blk_queue_discard(q))
3910 ext4_msg(sb, KERN_WARNING,
3911 "mounting with \"discard\" option, but "
3912 "the device does not support discard");
3915 if (___ratelimit(&ext4_mount_msg_ratelimit, "EXT4-fs mount"))
3916 ext4_msg(sb, KERN_INFO, "mounted filesystem with%s. "
3917 "Opts: %s%s%s", descr, sbi->s_es->s_mount_opts,
3918 *sbi->s_es->s_mount_opts ? "; " : "", orig_data);
3920 if (es->s_error_count)
3921 mod_timer(&sbi->s_err_report, jiffies + 300*HZ); /* 5 minutes */
3923 /* Enable message ratelimiting. Default is 10 messages per 5 secs. */
3924 ratelimit_state_init(&sbi->s_err_ratelimit_state, 5 * HZ, 10);
3925 ratelimit_state_init(&sbi->s_warning_ratelimit_state, 5 * HZ, 10);
3926 ratelimit_state_init(&sbi->s_msg_ratelimit_state, 5 * HZ, 10);
3933 ext4_msg(sb, KERN_ERR, "VFS: Can't find ext4 filesystem");
3938 ext4_unregister_sysfs(sb);
3941 ext4_unregister_li_request(sb);
3943 ext4_mb_release(sb);
3944 if (sbi->s_flex_groups)
3945 kvfree(sbi->s_flex_groups);
3946 percpu_counter_destroy(&sbi->s_freeclusters_counter);
3947 percpu_counter_destroy(&sbi->s_freeinodes_counter);
3948 percpu_counter_destroy(&sbi->s_dirs_counter);
3949 percpu_counter_destroy(&sbi->s_dirtyclusters_counter);
3951 ext4_ext_release(sb);
3952 ext4_release_system_zone(sb);
3957 ext4_msg(sb, KERN_ERR, "mount failed");
3958 if (EXT4_SB(sb)->rsv_conversion_wq)
3959 destroy_workqueue(EXT4_SB(sb)->rsv_conversion_wq);
3961 if (sbi->s_journal) {
3962 jbd2_journal_destroy(sbi->s_journal);
3963 sbi->s_journal = NULL;
3966 ext4_es_unregister_shrinker(sbi);
3968 del_timer_sync(&sbi->s_err_report);
3970 kthread_stop(sbi->s_mmp_tsk);
3972 for (i = 0; i < db_count; i++)
3973 brelse(sbi->s_group_desc[i]);
3974 kvfree(sbi->s_group_desc);
3976 if (sbi->s_chksum_driver)
3977 crypto_free_shash(sbi->s_chksum_driver);
3979 for (i = 0; i < EXT4_MAXQUOTAS; i++)
3980 kfree(sbi->s_qf_names[i]);
3982 ext4_blkdev_remove(sbi);
3985 sb->s_fs_info = NULL;
3986 kfree(sbi->s_blockgroup_lock);
3990 return err ? err : ret;
3994 * Setup any per-fs journal parameters now. We'll do this both on
3995 * initial mount, once the journal has been initialised but before we've
3996 * done any recovery; and again on any subsequent remount.
3998 static void ext4_init_journal_params(struct super_block *sb, journal_t *journal)
4000 struct ext4_sb_info *sbi = EXT4_SB(sb);
4002 journal->j_commit_interval = sbi->s_commit_interval;
4003 journal->j_min_batch_time = sbi->s_min_batch_time;
4004 journal->j_max_batch_time = sbi->s_max_batch_time;
4006 write_lock(&journal->j_state_lock);
4007 if (test_opt(sb, BARRIER))
4008 journal->j_flags |= JBD2_BARRIER;
4010 journal->j_flags &= ~JBD2_BARRIER;
4011 if (test_opt(sb, DATA_ERR_ABORT))
4012 journal->j_flags |= JBD2_ABORT_ON_SYNCDATA_ERR;
4014 journal->j_flags &= ~JBD2_ABORT_ON_SYNCDATA_ERR;
4015 write_unlock(&journal->j_state_lock);
4018 static journal_t *ext4_get_journal(struct super_block *sb,
4019 unsigned int journal_inum)
4021 struct inode *journal_inode;
4024 BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL));
4026 /* First, test for the existence of a valid inode on disk. Bad
4027 * things happen if we iget() an unused inode, as the subsequent
4028 * iput() will try to delete it. */
4030 journal_inode = ext4_iget(sb, journal_inum);
4031 if (IS_ERR(journal_inode)) {
4032 ext4_msg(sb, KERN_ERR, "no journal found");
4035 if (!journal_inode->i_nlink) {
4036 make_bad_inode(journal_inode);
4037 iput(journal_inode);
4038 ext4_msg(sb, KERN_ERR, "journal inode is deleted");
4042 jbd_debug(2, "Journal inode found at %p: %lld bytes\n",
4043 journal_inode, journal_inode->i_size);
4044 if (!S_ISREG(journal_inode->i_mode)) {
4045 ext4_msg(sb, KERN_ERR, "invalid journal inode");
4046 iput(journal_inode);
4050 journal = jbd2_journal_init_inode(journal_inode);
4052 ext4_msg(sb, KERN_ERR, "Could not load journal inode");
4053 iput(journal_inode);
4056 journal->j_private = sb;
4057 ext4_init_journal_params(sb, journal);
4061 static journal_t *ext4_get_dev_journal(struct super_block *sb,
4064 struct buffer_head *bh;
4068 int hblock, blocksize;
4069 ext4_fsblk_t sb_block;
4070 unsigned long offset;
4071 struct ext4_super_block *es;
4072 struct block_device *bdev;
4074 BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL));
4076 bdev = ext4_blkdev_get(j_dev, sb);
4080 blocksize = sb->s_blocksize;
4081 hblock = bdev_logical_block_size(bdev);
4082 if (blocksize < hblock) {
4083 ext4_msg(sb, KERN_ERR,
4084 "blocksize too small for journal device");
4088 sb_block = EXT4_MIN_BLOCK_SIZE / blocksize;
4089 offset = EXT4_MIN_BLOCK_SIZE % blocksize;
4090 set_blocksize(bdev, blocksize);
4091 if (!(bh = __bread(bdev, sb_block, blocksize))) {
4092 ext4_msg(sb, KERN_ERR, "couldn't read superblock of "
4093 "external journal");
4097 es = (struct ext4_super_block *) (bh->b_data + offset);
4098 if ((le16_to_cpu(es->s_magic) != EXT4_SUPER_MAGIC) ||
4099 !(le32_to_cpu(es->s_feature_incompat) &
4100 EXT4_FEATURE_INCOMPAT_JOURNAL_DEV)) {
4101 ext4_msg(sb, KERN_ERR, "external journal has "
4107 if ((le32_to_cpu(es->s_feature_ro_compat) &
4108 EXT4_FEATURE_RO_COMPAT_METADATA_CSUM) &&
4109 es->s_checksum != ext4_superblock_csum(sb, es)) {
4110 ext4_msg(sb, KERN_ERR, "external journal has "
4111 "corrupt superblock");
4116 if (memcmp(EXT4_SB(sb)->s_es->s_journal_uuid, es->s_uuid, 16)) {
4117 ext4_msg(sb, KERN_ERR, "journal UUID does not match");
4122 len = ext4_blocks_count(es);
4123 start = sb_block + 1;
4124 brelse(bh); /* we're done with the superblock */
4126 journal = jbd2_journal_init_dev(bdev, sb->s_bdev,
4127 start, len, blocksize);
4129 ext4_msg(sb, KERN_ERR, "failed to create device journal");
4132 journal->j_private = sb;
4133 ll_rw_block(READ | REQ_META | REQ_PRIO, 1, &journal->j_sb_buffer);
4134 wait_on_buffer(journal->j_sb_buffer);
4135 if (!buffer_uptodate(journal->j_sb_buffer)) {
4136 ext4_msg(sb, KERN_ERR, "I/O error on journal device");
4139 if (be32_to_cpu(journal->j_superblock->s_nr_users) != 1) {
4140 ext4_msg(sb, KERN_ERR, "External journal has more than one "
4141 "user (unsupported) - %d",
4142 be32_to_cpu(journal->j_superblock->s_nr_users));
4145 EXT4_SB(sb)->journal_bdev = bdev;
4146 ext4_init_journal_params(sb, journal);
4150 jbd2_journal_destroy(journal);
4152 ext4_blkdev_put(bdev);
4156 static int ext4_load_journal(struct super_block *sb,
4157 struct ext4_super_block *es,
4158 unsigned long journal_devnum)
4161 unsigned int journal_inum = le32_to_cpu(es->s_journal_inum);
4164 int really_read_only;
4166 BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL));
4168 if (journal_devnum &&
4169 journal_devnum != le32_to_cpu(es->s_journal_dev)) {
4170 ext4_msg(sb, KERN_INFO, "external journal device major/minor "
4171 "numbers have changed");
4172 journal_dev = new_decode_dev(journal_devnum);
4174 journal_dev = new_decode_dev(le32_to_cpu(es->s_journal_dev));
4176 really_read_only = bdev_read_only(sb->s_bdev);
4179 * Are we loading a blank journal or performing recovery after a
4180 * crash? For recovery, we need to check in advance whether we
4181 * can get read-write access to the device.
4183 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER)) {
4184 if (sb->s_flags & MS_RDONLY) {
4185 ext4_msg(sb, KERN_INFO, "INFO: recovery "
4186 "required on readonly filesystem");
4187 if (really_read_only) {
4188 ext4_msg(sb, KERN_ERR, "write access "
4189 "unavailable, cannot proceed");
4192 ext4_msg(sb, KERN_INFO, "write access will "
4193 "be enabled during recovery");
4197 if (journal_inum && journal_dev) {
4198 ext4_msg(sb, KERN_ERR, "filesystem has both journal "
4199 "and inode journals!");
4204 if (!(journal = ext4_get_journal(sb, journal_inum)))
4207 if (!(journal = ext4_get_dev_journal(sb, journal_dev)))
4211 if (!(journal->j_flags & JBD2_BARRIER))
4212 ext4_msg(sb, KERN_INFO, "barriers disabled");
4214 if (!EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER))
4215 err = jbd2_journal_wipe(journal, !really_read_only);
4217 char *save = kmalloc(EXT4_S_ERR_LEN, GFP_KERNEL);
4219 memcpy(save, ((char *) es) +
4220 EXT4_S_ERR_START, EXT4_S_ERR_LEN);
4221 err = jbd2_journal_load(journal);
4223 memcpy(((char *) es) + EXT4_S_ERR_START,
4224 save, EXT4_S_ERR_LEN);
4229 ext4_msg(sb, KERN_ERR, "error loading journal");
4230 jbd2_journal_destroy(journal);
4234 EXT4_SB(sb)->s_journal = journal;
4235 ext4_clear_journal_err(sb, es);
4237 if (!really_read_only && journal_devnum &&
4238 journal_devnum != le32_to_cpu(es->s_journal_dev)) {
4239 es->s_journal_dev = cpu_to_le32(journal_devnum);
4241 /* Make sure we flush the recovery flag to disk. */
4242 ext4_commit_super(sb, 1);
4248 static int ext4_commit_super(struct super_block *sb, int sync)
4250 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
4251 struct buffer_head *sbh = EXT4_SB(sb)->s_sbh;
4254 if (!sbh || block_device_ejected(sb))
4256 if (buffer_write_io_error(sbh)) {
4258 * Oh, dear. A previous attempt to write the
4259 * superblock failed. This could happen because the
4260 * USB device was yanked out. Or it could happen to
4261 * be a transient write error and maybe the block will
4262 * be remapped. Nothing we can do but to retry the
4263 * write and hope for the best.
4265 ext4_msg(sb, KERN_ERR, "previous I/O error to "
4266 "superblock detected");
4267 clear_buffer_write_io_error(sbh);
4268 set_buffer_uptodate(sbh);
4271 * If the file system is mounted read-only, don't update the
4272 * superblock write time. This avoids updating the superblock
4273 * write time when we are mounting the root file system
4274 * read/only but we need to replay the journal; at that point,
4275 * for people who are east of GMT and who make their clock
4276 * tick in localtime for Windows bug-for-bug compatibility,
4277 * the clock is set in the future, and this will cause e2fsck
4278 * to complain and force a full file system check.
4280 if (!(sb->s_flags & MS_RDONLY))
4281 es->s_wtime = cpu_to_le32(get_seconds());
4282 if (sb->s_bdev->bd_part)
4283 es->s_kbytes_written =
4284 cpu_to_le64(EXT4_SB(sb)->s_kbytes_written +
4285 ((part_stat_read(sb->s_bdev->bd_part, sectors[1]) -
4286 EXT4_SB(sb)->s_sectors_written_start) >> 1));
4288 es->s_kbytes_written =
4289 cpu_to_le64(EXT4_SB(sb)->s_kbytes_written);
4290 if (percpu_counter_initialized(&EXT4_SB(sb)->s_freeclusters_counter))
4291 ext4_free_blocks_count_set(es,
4292 EXT4_C2B(EXT4_SB(sb), percpu_counter_sum_positive(
4293 &EXT4_SB(sb)->s_freeclusters_counter)));
4294 if (percpu_counter_initialized(&EXT4_SB(sb)->s_freeinodes_counter))
4295 es->s_free_inodes_count =
4296 cpu_to_le32(percpu_counter_sum_positive(
4297 &EXT4_SB(sb)->s_freeinodes_counter));
4298 BUFFER_TRACE(sbh, "marking dirty");
4299 ext4_superblock_csum_set(sb);
4300 mark_buffer_dirty(sbh);
4302 error = __sync_dirty_buffer(sbh,
4303 test_opt(sb, BARRIER) ? WRITE_FUA : WRITE_SYNC);
4307 error = buffer_write_io_error(sbh);
4309 ext4_msg(sb, KERN_ERR, "I/O error while writing "
4311 clear_buffer_write_io_error(sbh);
4312 set_buffer_uptodate(sbh);
4319 * Have we just finished recovery? If so, and if we are mounting (or
4320 * remounting) the filesystem readonly, then we will end up with a
4321 * consistent fs on disk. Record that fact.
4323 static void ext4_mark_recovery_complete(struct super_block *sb,
4324 struct ext4_super_block *es)
4326 journal_t *journal = EXT4_SB(sb)->s_journal;
4328 if (!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL)) {
4329 BUG_ON(journal != NULL);
4332 jbd2_journal_lock_updates(journal);
4333 if (jbd2_journal_flush(journal) < 0)
4336 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER) &&
4337 sb->s_flags & MS_RDONLY) {
4338 EXT4_CLEAR_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
4339 ext4_commit_super(sb, 1);
4343 jbd2_journal_unlock_updates(journal);
4347 * If we are mounting (or read-write remounting) a filesystem whose journal
4348 * has recorded an error from a previous lifetime, move that error to the
4349 * main filesystem now.
4351 static void ext4_clear_journal_err(struct super_block *sb,
4352 struct ext4_super_block *es)
4358 BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL));
4360 journal = EXT4_SB(sb)->s_journal;
4363 * Now check for any error status which may have been recorded in the
4364 * journal by a prior ext4_error() or ext4_abort()
4367 j_errno = jbd2_journal_errno(journal);
4371 errstr = ext4_decode_error(sb, j_errno, nbuf);
4372 ext4_warning(sb, "Filesystem error recorded "
4373 "from previous mount: %s", errstr);
4374 ext4_warning(sb, "Marking fs in need of filesystem check.");
4376 EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS;
4377 es->s_state |= cpu_to_le16(EXT4_ERROR_FS);
4378 ext4_commit_super(sb, 1);
4380 jbd2_journal_clear_err(journal);
4381 jbd2_journal_update_sb_errno(journal);
4386 * Force the running and committing transactions to commit,
4387 * and wait on the commit.
4389 int ext4_force_commit(struct super_block *sb)
4393 if (sb->s_flags & MS_RDONLY)
4396 journal = EXT4_SB(sb)->s_journal;
4397 return ext4_journal_force_commit(journal);
4400 static int ext4_sync_fs(struct super_block *sb, int wait)
4404 bool needs_barrier = false;
4405 struct ext4_sb_info *sbi = EXT4_SB(sb);
4407 trace_ext4_sync_fs(sb, wait);
4408 flush_workqueue(sbi->rsv_conversion_wq);
4410 * Writeback quota in non-journalled quota case - journalled quota has
4413 dquot_writeback_dquots(sb, -1);
4415 * Data writeback is possible w/o journal transaction, so barrier must
4416 * being sent at the end of the function. But we can skip it if
4417 * transaction_commit will do it for us.
4419 if (sbi->s_journal) {
4420 target = jbd2_get_latest_transaction(sbi->s_journal);
4421 if (wait && sbi->s_journal->j_flags & JBD2_BARRIER &&
4422 !jbd2_trans_will_send_data_barrier(sbi->s_journal, target))
4423 needs_barrier = true;
4425 if (jbd2_journal_start_commit(sbi->s_journal, &target)) {
4427 ret = jbd2_log_wait_commit(sbi->s_journal,
4430 } else if (wait && test_opt(sb, BARRIER))
4431 needs_barrier = true;
4432 if (needs_barrier) {
4434 err = blkdev_issue_flush(sb->s_bdev, GFP_KERNEL, NULL);
4443 * LVM calls this function before a (read-only) snapshot is created. This
4444 * gives us a chance to flush the journal completely and mark the fs clean.
4446 * Note that only this function cannot bring a filesystem to be in a clean
4447 * state independently. It relies on upper layer to stop all data & metadata
4450 static int ext4_freeze(struct super_block *sb)
4455 if (sb->s_flags & MS_RDONLY)
4458 journal = EXT4_SB(sb)->s_journal;
4461 /* Now we set up the journal barrier. */
4462 jbd2_journal_lock_updates(journal);
4465 * Don't clear the needs_recovery flag if we failed to
4466 * flush the journal.
4468 error = jbd2_journal_flush(journal);
4472 /* Journal blocked and flushed, clear needs_recovery flag. */
4473 EXT4_CLEAR_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
4476 error = ext4_commit_super(sb, 1);
4479 /* we rely on upper layer to stop further updates */
4480 jbd2_journal_unlock_updates(journal);
4485 * Called by LVM after the snapshot is done. We need to reset the RECOVER
4486 * flag here, even though the filesystem is not technically dirty yet.
4488 static int ext4_unfreeze(struct super_block *sb)
4490 if (sb->s_flags & MS_RDONLY)
4493 if (EXT4_SB(sb)->s_journal) {
4494 /* Reset the needs_recovery flag before the fs is unlocked. */
4495 EXT4_SET_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
4498 ext4_commit_super(sb, 1);
4503 * Structure to save mount options for ext4_remount's benefit
4505 struct ext4_mount_options {
4506 unsigned long s_mount_opt;
4507 unsigned long s_mount_opt2;
4510 unsigned long s_commit_interval;
4511 u32 s_min_batch_time, s_max_batch_time;
4514 char *s_qf_names[EXT4_MAXQUOTAS];
4518 static int ext4_remount(struct super_block *sb, int *flags, char *data)
4520 struct ext4_super_block *es;
4521 struct ext4_sb_info *sbi = EXT4_SB(sb);
4522 unsigned long old_sb_flags;
4523 struct ext4_mount_options old_opts;
4524 int enable_quota = 0;
4526 unsigned int journal_ioprio = DEFAULT_JOURNAL_IOPRIO;
4531 char *orig_data = kstrdup(data, GFP_KERNEL);
4533 /* Store the original options */
4534 old_sb_flags = sb->s_flags;
4535 old_opts.s_mount_opt = sbi->s_mount_opt;
4536 old_opts.s_mount_opt2 = sbi->s_mount_opt2;
4537 old_opts.s_resuid = sbi->s_resuid;
4538 old_opts.s_resgid = sbi->s_resgid;
4539 old_opts.s_commit_interval = sbi->s_commit_interval;
4540 old_opts.s_min_batch_time = sbi->s_min_batch_time;
4541 old_opts.s_max_batch_time = sbi->s_max_batch_time;
4543 old_opts.s_jquota_fmt = sbi->s_jquota_fmt;
4544 for (i = 0; i < EXT4_MAXQUOTAS; i++)
4545 if (sbi->s_qf_names[i]) {
4546 old_opts.s_qf_names[i] = kstrdup(sbi->s_qf_names[i],
4548 if (!old_opts.s_qf_names[i]) {
4549 for (j = 0; j < i; j++)
4550 kfree(old_opts.s_qf_names[j]);
4555 old_opts.s_qf_names[i] = NULL;
4557 if (sbi->s_journal && sbi->s_journal->j_task->io_context)
4558 journal_ioprio = sbi->s_journal->j_task->io_context->ioprio;
4560 if (!parse_options(data, sb, NULL, &journal_ioprio, 1)) {
4565 if ((old_opts.s_mount_opt & EXT4_MOUNT_JOURNAL_CHECKSUM) ^
4566 test_opt(sb, JOURNAL_CHECKSUM)) {
4567 ext4_msg(sb, KERN_ERR, "changing journal_checksum "
4568 "during remount not supported; ignoring");
4569 sbi->s_mount_opt ^= EXT4_MOUNT_JOURNAL_CHECKSUM;
4572 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA) {
4573 if (test_opt2(sb, EXPLICIT_DELALLOC)) {
4574 ext4_msg(sb, KERN_ERR, "can't mount with "
4575 "both data=journal and delalloc");
4579 if (test_opt(sb, DIOREAD_NOLOCK)) {
4580 ext4_msg(sb, KERN_ERR, "can't mount with "
4581 "both data=journal and dioread_nolock");
4585 if (test_opt(sb, DAX)) {
4586 ext4_msg(sb, KERN_ERR, "can't mount with "
4587 "both data=journal and dax");
4593 if ((sbi->s_mount_opt ^ old_opts.s_mount_opt) & EXT4_MOUNT_DAX) {
4594 ext4_msg(sb, KERN_WARNING, "warning: refusing change of "
4595 "dax flag with busy inodes while remounting");
4596 sbi->s_mount_opt ^= EXT4_MOUNT_DAX;
4599 if (sbi->s_mount_flags & EXT4_MF_FS_ABORTED)
4600 ext4_abort(sb, "Abort forced by user");
4602 sb->s_flags = (sb->s_flags & ~MS_POSIXACL) |
4603 (test_opt(sb, POSIX_ACL) ? MS_POSIXACL : 0);
4607 if (sbi->s_journal) {
4608 ext4_init_journal_params(sb, sbi->s_journal);
4609 set_task_ioprio(sbi->s_journal->j_task, journal_ioprio);
4612 if (*flags & MS_LAZYTIME)
4613 sb->s_flags |= MS_LAZYTIME;
4615 if ((*flags & MS_RDONLY) != (sb->s_flags & MS_RDONLY)) {
4616 if (sbi->s_mount_flags & EXT4_MF_FS_ABORTED) {
4621 if (*flags & MS_RDONLY) {
4622 err = sync_filesystem(sb);
4625 err = dquot_suspend(sb, -1);
4630 * First of all, the unconditional stuff we have to do
4631 * to disable replay of the journal when we next remount
4633 sb->s_flags |= MS_RDONLY;
4636 * OK, test if we are remounting a valid rw partition
4637 * readonly, and if so set the rdonly flag and then
4638 * mark the partition as valid again.
4640 if (!(es->s_state & cpu_to_le16(EXT4_VALID_FS)) &&
4641 (sbi->s_mount_state & EXT4_VALID_FS))
4642 es->s_state = cpu_to_le16(sbi->s_mount_state);
4645 ext4_mark_recovery_complete(sb, es);
4647 /* Make sure we can mount this feature set readwrite */
4648 if (EXT4_HAS_RO_COMPAT_FEATURE(sb,
4649 EXT4_FEATURE_RO_COMPAT_READONLY) ||
4650 !ext4_feature_set_ok(sb, 0)) {
4655 * Make sure the group descriptor checksums
4656 * are sane. If they aren't, refuse to remount r/w.
4658 for (g = 0; g < sbi->s_groups_count; g++) {
4659 struct ext4_group_desc *gdp =
4660 ext4_get_group_desc(sb, g, NULL);
4662 if (!ext4_group_desc_csum_verify(sb, g, gdp)) {
4663 ext4_msg(sb, KERN_ERR,
4664 "ext4_remount: Checksum for group %u failed (%u!=%u)",
4665 g, le16_to_cpu(ext4_group_desc_csum(sbi, g, gdp)),
4666 le16_to_cpu(gdp->bg_checksum));
4673 * If we have an unprocessed orphan list hanging
4674 * around from a previously readonly bdev mount,
4675 * require a full umount/remount for now.
4677 if (es->s_last_orphan) {
4678 ext4_msg(sb, KERN_WARNING, "Couldn't "
4679 "remount RDWR because of unprocessed "
4680 "orphan inode list. Please "
4681 "umount/remount instead");
4687 * Mounting a RDONLY partition read-write, so reread
4688 * and store the current valid flag. (It may have
4689 * been changed by e2fsck since we originally mounted
4693 ext4_clear_journal_err(sb, es);
4694 sbi->s_mount_state = le16_to_cpu(es->s_state);
4695 if (!ext4_setup_super(sb, es, 0))
4696 sb->s_flags &= ~MS_RDONLY;
4697 if (EXT4_HAS_INCOMPAT_FEATURE(sb,
4698 EXT4_FEATURE_INCOMPAT_MMP))
4699 if (ext4_multi_mount_protect(sb,
4700 le64_to_cpu(es->s_mmp_block))) {
4709 * Reinitialize lazy itable initialization thread based on
4712 if ((sb->s_flags & MS_RDONLY) || !test_opt(sb, INIT_INODE_TABLE))
4713 ext4_unregister_li_request(sb);
4715 ext4_group_t first_not_zeroed;
4716 first_not_zeroed = ext4_has_uninit_itable(sb);
4717 ext4_register_li_request(sb, first_not_zeroed);
4720 ext4_setup_system_zone(sb);
4721 if (sbi->s_journal == NULL && !(old_sb_flags & MS_RDONLY))
4722 ext4_commit_super(sb, 1);
4725 /* Release old quota file names */
4726 for (i = 0; i < EXT4_MAXQUOTAS; i++)
4727 kfree(old_opts.s_qf_names[i]);
4729 if (sb_any_quota_suspended(sb))
4730 dquot_resume(sb, -1);
4731 else if (EXT4_HAS_RO_COMPAT_FEATURE(sb,
4732 EXT4_FEATURE_RO_COMPAT_QUOTA)) {
4733 err = ext4_enable_quotas(sb);
4740 *flags = (*flags & ~MS_LAZYTIME) | (sb->s_flags & MS_LAZYTIME);
4741 ext4_msg(sb, KERN_INFO, "re-mounted. Opts: %s", orig_data);
4746 sb->s_flags = old_sb_flags;
4747 sbi->s_mount_opt = old_opts.s_mount_opt;
4748 sbi->s_mount_opt2 = old_opts.s_mount_opt2;
4749 sbi->s_resuid = old_opts.s_resuid;
4750 sbi->s_resgid = old_opts.s_resgid;
4751 sbi->s_commit_interval = old_opts.s_commit_interval;
4752 sbi->s_min_batch_time = old_opts.s_min_batch_time;
4753 sbi->s_max_batch_time = old_opts.s_max_batch_time;
4755 sbi->s_jquota_fmt = old_opts.s_jquota_fmt;
4756 for (i = 0; i < EXT4_MAXQUOTAS; i++) {
4757 kfree(sbi->s_qf_names[i]);
4758 sbi->s_qf_names[i] = old_opts.s_qf_names[i];
4765 static int ext4_statfs(struct dentry *dentry, struct kstatfs *buf)
4767 struct super_block *sb = dentry->d_sb;
4768 struct ext4_sb_info *sbi = EXT4_SB(sb);
4769 struct ext4_super_block *es = sbi->s_es;
4770 ext4_fsblk_t overhead = 0, resv_blocks;
4773 resv_blocks = EXT4_C2B(sbi, atomic64_read(&sbi->s_resv_clusters));
4775 if (!test_opt(sb, MINIX_DF))
4776 overhead = sbi->s_overhead;
4778 buf->f_type = EXT4_SUPER_MAGIC;
4779 buf->f_bsize = sb->s_blocksize;
4780 buf->f_blocks = ext4_blocks_count(es) - EXT4_C2B(sbi, overhead);
4781 bfree = percpu_counter_sum_positive(&sbi->s_freeclusters_counter) -
4782 percpu_counter_sum_positive(&sbi->s_dirtyclusters_counter);
4783 /* prevent underflow in case that few free space is available */
4784 buf->f_bfree = EXT4_C2B(sbi, max_t(s64, bfree, 0));
4785 buf->f_bavail = buf->f_bfree -
4786 (ext4_r_blocks_count(es) + resv_blocks);
4787 if (buf->f_bfree < (ext4_r_blocks_count(es) + resv_blocks))
4789 buf->f_files = le32_to_cpu(es->s_inodes_count);
4790 buf->f_ffree = percpu_counter_sum_positive(&sbi->s_freeinodes_counter);
4791 buf->f_namelen = EXT4_NAME_LEN;
4792 fsid = le64_to_cpup((void *)es->s_uuid) ^
4793 le64_to_cpup((void *)es->s_uuid + sizeof(u64));
4794 buf->f_fsid.val[0] = fsid & 0xFFFFFFFFUL;
4795 buf->f_fsid.val[1] = (fsid >> 32) & 0xFFFFFFFFUL;
4800 /* Helper function for writing quotas on sync - we need to start transaction
4801 * before quota file is locked for write. Otherwise the are possible deadlocks:
4802 * Process 1 Process 2
4803 * ext4_create() quota_sync()
4804 * jbd2_journal_start() write_dquot()
4805 * dquot_initialize() down(dqio_mutex)
4806 * down(dqio_mutex) jbd2_journal_start()
4812 static inline struct inode *dquot_to_inode(struct dquot *dquot)
4814 return sb_dqopt(dquot->dq_sb)->files[dquot->dq_id.type];
4817 static int ext4_write_dquot(struct dquot *dquot)
4821 struct inode *inode;
4823 inode = dquot_to_inode(dquot);
4824 handle = ext4_journal_start(inode, EXT4_HT_QUOTA,
4825 EXT4_QUOTA_TRANS_BLOCKS(dquot->dq_sb));
4827 return PTR_ERR(handle);
4828 ret = dquot_commit(dquot);
4829 err = ext4_journal_stop(handle);
4835 static int ext4_acquire_dquot(struct dquot *dquot)
4840 handle = ext4_journal_start(dquot_to_inode(dquot), EXT4_HT_QUOTA,
4841 EXT4_QUOTA_INIT_BLOCKS(dquot->dq_sb));
4843 return PTR_ERR(handle);
4844 ret = dquot_acquire(dquot);
4845 err = ext4_journal_stop(handle);
4851 static int ext4_release_dquot(struct dquot *dquot)
4856 handle = ext4_journal_start(dquot_to_inode(dquot), EXT4_HT_QUOTA,
4857 EXT4_QUOTA_DEL_BLOCKS(dquot->dq_sb));
4858 if (IS_ERR(handle)) {
4859 /* Release dquot anyway to avoid endless cycle in dqput() */
4860 dquot_release(dquot);
4861 return PTR_ERR(handle);
4863 ret = dquot_release(dquot);
4864 err = ext4_journal_stop(handle);
4870 static int ext4_mark_dquot_dirty(struct dquot *dquot)
4872 struct super_block *sb = dquot->dq_sb;
4873 struct ext4_sb_info *sbi = EXT4_SB(sb);
4875 /* Are we journaling quotas? */
4876 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_QUOTA) ||
4877 sbi->s_qf_names[USRQUOTA] || sbi->s_qf_names[GRPQUOTA]) {
4878 dquot_mark_dquot_dirty(dquot);
4879 return ext4_write_dquot(dquot);
4881 return dquot_mark_dquot_dirty(dquot);
4885 static int ext4_write_info(struct super_block *sb, int type)
4890 /* Data block + inode block */
4891 handle = ext4_journal_start(d_inode(sb->s_root), EXT4_HT_QUOTA, 2);
4893 return PTR_ERR(handle);
4894 ret = dquot_commit_info(sb, type);
4895 err = ext4_journal_stop(handle);
4902 * Turn on quotas during mount time - we need to find
4903 * the quota file and such...
4905 static int ext4_quota_on_mount(struct super_block *sb, int type)
4907 return dquot_quota_on_mount(sb, EXT4_SB(sb)->s_qf_names[type],
4908 EXT4_SB(sb)->s_jquota_fmt, type);
4912 * Standard function to be called on quota_on
4914 static int ext4_quota_on(struct super_block *sb, int type, int format_id,
4919 if (!test_opt(sb, QUOTA))
4922 /* Quotafile not on the same filesystem? */
4923 if (path->dentry->d_sb != sb)
4925 /* Journaling quota? */
4926 if (EXT4_SB(sb)->s_qf_names[type]) {
4927 /* Quotafile not in fs root? */
4928 if (path->dentry->d_parent != sb->s_root)
4929 ext4_msg(sb, KERN_WARNING,
4930 "Quota file not on filesystem root. "
4931 "Journaled quota will not work");
4935 * When we journal data on quota file, we have to flush journal to see
4936 * all updates to the file when we bypass pagecache...
4938 if (EXT4_SB(sb)->s_journal &&
4939 ext4_should_journal_data(d_inode(path->dentry))) {
4941 * We don't need to lock updates but journal_flush() could
4942 * otherwise be livelocked...
4944 jbd2_journal_lock_updates(EXT4_SB(sb)->s_journal);
4945 err = jbd2_journal_flush(EXT4_SB(sb)->s_journal);
4946 jbd2_journal_unlock_updates(EXT4_SB(sb)->s_journal);
4951 return dquot_quota_on(sb, type, format_id, path);
4954 static int ext4_quota_enable(struct super_block *sb, int type, int format_id,
4958 struct inode *qf_inode;
4959 unsigned long qf_inums[EXT4_MAXQUOTAS] = {
4960 le32_to_cpu(EXT4_SB(sb)->s_es->s_usr_quota_inum),
4961 le32_to_cpu(EXT4_SB(sb)->s_es->s_grp_quota_inum)
4964 BUG_ON(!EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_QUOTA));
4966 if (!qf_inums[type])
4969 qf_inode = ext4_iget(sb, qf_inums[type]);
4970 if (IS_ERR(qf_inode)) {
4971 ext4_error(sb, "Bad quota inode # %lu", qf_inums[type]);
4972 return PTR_ERR(qf_inode);
4975 /* Don't account quota for quota files to avoid recursion */
4976 qf_inode->i_flags |= S_NOQUOTA;
4977 err = dquot_enable(qf_inode, type, format_id, flags);
4983 /* Enable usage tracking for all quota types. */
4984 static int ext4_enable_quotas(struct super_block *sb)
4987 unsigned long qf_inums[EXT4_MAXQUOTAS] = {
4988 le32_to_cpu(EXT4_SB(sb)->s_es->s_usr_quota_inum),
4989 le32_to_cpu(EXT4_SB(sb)->s_es->s_grp_quota_inum)
4992 sb_dqopt(sb)->flags |= DQUOT_QUOTA_SYS_FILE;
4993 for (type = 0; type < EXT4_MAXQUOTAS; type++) {
4994 if (qf_inums[type]) {
4995 err = ext4_quota_enable(sb, type, QFMT_VFS_V1,
4996 DQUOT_USAGE_ENABLED);
4999 "Failed to enable quota tracking "
5000 "(type=%d, err=%d). Please run "
5001 "e2fsck to fix.", type, err);
5009 static int ext4_quota_off(struct super_block *sb, int type)
5011 struct inode *inode = sb_dqopt(sb)->files[type];
5014 /* Force all delayed allocation blocks to be allocated.
5015 * Caller already holds s_umount sem */
5016 if (test_opt(sb, DELALLOC))
5017 sync_filesystem(sb);
5022 /* Update modification times of quota files when userspace can
5023 * start looking at them */
5024 handle = ext4_journal_start(inode, EXT4_HT_QUOTA, 1);
5027 inode->i_mtime = inode->i_ctime = CURRENT_TIME;
5028 ext4_mark_inode_dirty(handle, inode);
5029 ext4_journal_stop(handle);
5032 return dquot_quota_off(sb, type);
5035 /* Read data from quotafile - avoid pagecache and such because we cannot afford
5036 * acquiring the locks... As quota files are never truncated and quota code
5037 * itself serializes the operations (and no one else should touch the files)
5038 * we don't have to be afraid of races */
5039 static ssize_t ext4_quota_read(struct super_block *sb, int type, char *data,
5040 size_t len, loff_t off)
5042 struct inode *inode = sb_dqopt(sb)->files[type];
5043 ext4_lblk_t blk = off >> EXT4_BLOCK_SIZE_BITS(sb);
5044 int offset = off & (sb->s_blocksize - 1);
5047 struct buffer_head *bh;
5048 loff_t i_size = i_size_read(inode);
5052 if (off+len > i_size)
5055 while (toread > 0) {
5056 tocopy = sb->s_blocksize - offset < toread ?
5057 sb->s_blocksize - offset : toread;
5058 bh = ext4_bread(NULL, inode, blk, 0);
5061 if (!bh) /* A hole? */
5062 memset(data, 0, tocopy);
5064 memcpy(data, bh->b_data+offset, tocopy);
5074 /* Write to quotafile (we know the transaction is already started and has
5075 * enough credits) */
5076 static ssize_t ext4_quota_write(struct super_block *sb, int type,
5077 const char *data, size_t len, loff_t off)
5079 struct inode *inode = sb_dqopt(sb)->files[type];
5080 ext4_lblk_t blk = off >> EXT4_BLOCK_SIZE_BITS(sb);
5081 int err, offset = off & (sb->s_blocksize - 1);
5083 struct buffer_head *bh;
5084 handle_t *handle = journal_current_handle();
5086 if (EXT4_SB(sb)->s_journal && !handle) {
5087 ext4_msg(sb, KERN_WARNING, "Quota write (off=%llu, len=%llu)"
5088 " cancelled because transaction is not started",
5089 (unsigned long long)off, (unsigned long long)len);
5093 * Since we account only one data block in transaction credits,
5094 * then it is impossible to cross a block boundary.
5096 if (sb->s_blocksize - offset < len) {
5097 ext4_msg(sb, KERN_WARNING, "Quota write (off=%llu, len=%llu)"
5098 " cancelled because not block aligned",
5099 (unsigned long long)off, (unsigned long long)len);
5104 bh = ext4_bread(handle, inode, blk,
5105 EXT4_GET_BLOCKS_CREATE |
5106 EXT4_GET_BLOCKS_METADATA_NOFAIL);
5107 } while (IS_ERR(bh) && (PTR_ERR(bh) == -ENOSPC) &&
5108 ext4_should_retry_alloc(inode->i_sb, &retries));
5113 BUFFER_TRACE(bh, "get write access");
5114 err = ext4_journal_get_write_access(handle, bh);
5120 memcpy(bh->b_data+offset, data, len);
5121 flush_dcache_page(bh->b_page);
5123 err = ext4_handle_dirty_metadata(handle, NULL, bh);
5126 if (inode->i_size < off + len) {
5127 i_size_write(inode, off + len);
5128 EXT4_I(inode)->i_disksize = inode->i_size;
5129 ext4_mark_inode_dirty(handle, inode);
5136 static struct dentry *ext4_mount(struct file_system_type *fs_type, int flags,
5137 const char *dev_name, void *data)
5139 return mount_bdev(fs_type, flags, dev_name, data, ext4_fill_super);
5142 #if !defined(CONFIG_EXT2_FS) && !defined(CONFIG_EXT2_FS_MODULE) && defined(CONFIG_EXT4_USE_FOR_EXT2)
5143 static inline void register_as_ext2(void)
5145 int err = register_filesystem(&ext2_fs_type);
5148 "EXT4-fs: Unable to register as ext2 (%d)\n", err);
5151 static inline void unregister_as_ext2(void)
5153 unregister_filesystem(&ext2_fs_type);
5156 static inline int ext2_feature_set_ok(struct super_block *sb)
5158 if (EXT4_HAS_INCOMPAT_FEATURE(sb, ~EXT2_FEATURE_INCOMPAT_SUPP))
5160 if (sb->s_flags & MS_RDONLY)
5162 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, ~EXT2_FEATURE_RO_COMPAT_SUPP))
5167 static inline void register_as_ext2(void) { }
5168 static inline void unregister_as_ext2(void) { }
5169 static inline int ext2_feature_set_ok(struct super_block *sb) { return 0; }
5172 static inline void register_as_ext3(void)
5174 int err = register_filesystem(&ext3_fs_type);
5177 "EXT4-fs: Unable to register as ext3 (%d)\n", err);
5180 static inline void unregister_as_ext3(void)
5182 unregister_filesystem(&ext3_fs_type);
5185 static inline int ext3_feature_set_ok(struct super_block *sb)
5187 if (EXT4_HAS_INCOMPAT_FEATURE(sb, ~EXT3_FEATURE_INCOMPAT_SUPP))
5189 if (!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL))
5191 if (sb->s_flags & MS_RDONLY)
5193 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, ~EXT3_FEATURE_RO_COMPAT_SUPP))
5198 static struct file_system_type ext4_fs_type = {
5199 .owner = THIS_MODULE,
5201 .mount = ext4_mount,
5202 .kill_sb = kill_block_super,
5203 .fs_flags = FS_REQUIRES_DEV,
5205 MODULE_ALIAS_FS("ext4");
5207 /* Shared across all ext4 file systems */
5208 wait_queue_head_t ext4__ioend_wq[EXT4_WQ_HASH_SZ];
5209 struct mutex ext4__aio_mutex[EXT4_WQ_HASH_SZ];
5211 static int __init ext4_init_fs(void)
5215 ratelimit_state_init(&ext4_mount_msg_ratelimit, 30 * HZ, 64);
5216 ext4_li_info = NULL;
5217 mutex_init(&ext4_li_mtx);
5219 /* Build-time check for flags consistency */
5220 ext4_check_flag_values();
5222 for (i = 0; i < EXT4_WQ_HASH_SZ; i++) {
5223 mutex_init(&ext4__aio_mutex[i]);
5224 init_waitqueue_head(&ext4__ioend_wq[i]);
5227 err = ext4_init_es();
5231 err = ext4_init_pageio();
5235 err = ext4_init_system_zone();
5239 err = ext4_init_sysfs();
5243 err = ext4_init_mballoc();
5247 ext4_mballoc_ready = 1;
5248 err = init_inodecache();
5253 err = register_filesystem(&ext4_fs_type);
5259 unregister_as_ext2();
5260 unregister_as_ext3();
5261 destroy_inodecache();
5263 ext4_mballoc_ready = 0;
5264 ext4_exit_mballoc();
5268 ext4_exit_system_zone();
5277 static void __exit ext4_exit_fs(void)
5280 ext4_destroy_lazyinit_thread();
5281 unregister_as_ext2();
5282 unregister_as_ext3();
5283 unregister_filesystem(&ext4_fs_type);
5284 destroy_inodecache();
5285 ext4_exit_mballoc();
5287 ext4_exit_system_zone();
5292 MODULE_AUTHOR("Remy Card, Stephen Tweedie, Andrew Morton, Andreas Dilger, Theodore Ts'o and others");
5293 MODULE_DESCRIPTION("Fourth Extended Filesystem");
5294 MODULE_LICENSE("GPL");
5295 module_init(ext4_init_fs)
5296 module_exit(ext4_exit_fs)