2 * linux/fs/ext4/super.c
4 * Copyright (C) 1992, 1993, 1994, 1995
5 * Remy Card (card@masi.ibp.fr)
6 * Laboratoire MASI - Institut Blaise Pascal
7 * Universite Pierre et Marie Curie (Paris VI)
11 * linux/fs/minix/inode.c
13 * Copyright (C) 1991, 1992 Linus Torvalds
15 * Big-endian to little-endian byte-swapping/bitmaps by
16 * David S. Miller (davem@caip.rutgers.edu), 1995
19 #include <linux/module.h>
20 #include <linux/string.h>
22 #include <linux/time.h>
23 #include <linux/vmalloc.h>
24 #include <linux/jbd2.h>
25 #include <linux/slab.h>
26 #include <linux/init.h>
27 #include <linux/blkdev.h>
28 #include <linux/parser.h>
29 #include <linux/buffer_head.h>
30 #include <linux/exportfs.h>
31 #include <linux/vfs.h>
32 #include <linux/random.h>
33 #include <linux/mount.h>
34 #include <linux/namei.h>
35 #include <linux/quotaops.h>
36 #include <linux/seq_file.h>
37 #include <linux/proc_fs.h>
38 #include <linux/ctype.h>
39 #include <linux/log2.h>
40 #include <linux/crc16.h>
41 #include <linux/cleancache.h>
42 #include <asm/uaccess.h>
44 #include <linux/kthread.h>
45 #include <linux/freezer.h>
48 #include "ext4_extents.h" /* Needed for trace points definition */
49 #include "ext4_jbd2.h"
54 #define CREATE_TRACE_POINTS
55 #include <trace/events/ext4.h>
57 static struct proc_dir_entry *ext4_proc_root;
58 static struct kset *ext4_kset;
59 static struct ext4_lazy_init *ext4_li_info;
60 static struct mutex ext4_li_mtx;
61 static struct ext4_features *ext4_feat;
63 static int ext4_load_journal(struct super_block *, struct ext4_super_block *,
64 unsigned long journal_devnum);
65 static int ext4_show_options(struct seq_file *seq, struct dentry *root);
66 static int ext4_commit_super(struct super_block *sb, int sync);
67 static void ext4_mark_recovery_complete(struct super_block *sb,
68 struct ext4_super_block *es);
69 static void ext4_clear_journal_err(struct super_block *sb,
70 struct ext4_super_block *es);
71 static int ext4_sync_fs(struct super_block *sb, int wait);
72 static int ext4_remount(struct super_block *sb, int *flags, char *data);
73 static int ext4_statfs(struct dentry *dentry, struct kstatfs *buf);
74 static int ext4_unfreeze(struct super_block *sb);
75 static int ext4_freeze(struct super_block *sb);
76 static struct dentry *ext4_mount(struct file_system_type *fs_type, int flags,
77 const char *dev_name, void *data);
78 static inline int ext2_feature_set_ok(struct super_block *sb);
79 static inline int ext3_feature_set_ok(struct super_block *sb);
80 static int ext4_feature_set_ok(struct super_block *sb, int readonly);
81 static void ext4_destroy_lazyinit_thread(void);
82 static void ext4_unregister_li_request(struct super_block *sb);
83 static void ext4_clear_request_list(void);
85 #if !defined(CONFIG_EXT2_FS) && !defined(CONFIG_EXT2_FS_MODULE) && defined(CONFIG_EXT4_USE_FOR_EXT23)
86 static struct file_system_type ext2_fs_type = {
90 .kill_sb = kill_block_super,
91 .fs_flags = FS_REQUIRES_DEV,
93 #define IS_EXT2_SB(sb) ((sb)->s_bdev->bd_holder == &ext2_fs_type)
95 #define IS_EXT2_SB(sb) (0)
99 #if !defined(CONFIG_EXT3_FS) && !defined(CONFIG_EXT3_FS_MODULE) && defined(CONFIG_EXT4_USE_FOR_EXT23)
100 static struct file_system_type ext3_fs_type = {
101 .owner = THIS_MODULE,
104 .kill_sb = kill_block_super,
105 .fs_flags = FS_REQUIRES_DEV,
107 #define IS_EXT3_SB(sb) ((sb)->s_bdev->bd_holder == &ext3_fs_type)
109 #define IS_EXT3_SB(sb) (0)
112 static int ext4_verify_csum_type(struct super_block *sb,
113 struct ext4_super_block *es)
115 if (!EXT4_HAS_RO_COMPAT_FEATURE(sb,
116 EXT4_FEATURE_RO_COMPAT_METADATA_CSUM))
119 return es->s_checksum_type == EXT4_CRC32C_CHKSUM;
122 static __le32 ext4_superblock_csum(struct super_block *sb,
123 struct ext4_super_block *es)
125 struct ext4_sb_info *sbi = EXT4_SB(sb);
126 int offset = offsetof(struct ext4_super_block, s_checksum);
129 csum = ext4_chksum(sbi, ~0, (char *)es, offset);
131 return cpu_to_le32(csum);
134 int ext4_superblock_csum_verify(struct super_block *sb,
135 struct ext4_super_block *es)
137 if (!EXT4_HAS_RO_COMPAT_FEATURE(sb,
138 EXT4_FEATURE_RO_COMPAT_METADATA_CSUM))
141 return es->s_checksum == ext4_superblock_csum(sb, es);
144 void ext4_superblock_csum_set(struct super_block *sb)
146 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
148 if (!EXT4_HAS_RO_COMPAT_FEATURE(sb,
149 EXT4_FEATURE_RO_COMPAT_METADATA_CSUM))
152 es->s_checksum = ext4_superblock_csum(sb, es);
155 void *ext4_kvmalloc(size_t size, gfp_t flags)
159 ret = kmalloc(size, flags);
161 ret = __vmalloc(size, flags, PAGE_KERNEL);
165 void *ext4_kvzalloc(size_t size, gfp_t flags)
169 ret = kzalloc(size, flags);
171 ret = __vmalloc(size, flags | __GFP_ZERO, PAGE_KERNEL);
175 void ext4_kvfree(void *ptr)
177 if (is_vmalloc_addr(ptr))
184 ext4_fsblk_t ext4_block_bitmap(struct super_block *sb,
185 struct ext4_group_desc *bg)
187 return le32_to_cpu(bg->bg_block_bitmap_lo) |
188 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
189 (ext4_fsblk_t)le32_to_cpu(bg->bg_block_bitmap_hi) << 32 : 0);
192 ext4_fsblk_t ext4_inode_bitmap(struct super_block *sb,
193 struct ext4_group_desc *bg)
195 return le32_to_cpu(bg->bg_inode_bitmap_lo) |
196 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
197 (ext4_fsblk_t)le32_to_cpu(bg->bg_inode_bitmap_hi) << 32 : 0);
200 ext4_fsblk_t ext4_inode_table(struct super_block *sb,
201 struct ext4_group_desc *bg)
203 return le32_to_cpu(bg->bg_inode_table_lo) |
204 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
205 (ext4_fsblk_t)le32_to_cpu(bg->bg_inode_table_hi) << 32 : 0);
208 __u32 ext4_free_group_clusters(struct super_block *sb,
209 struct ext4_group_desc *bg)
211 return le16_to_cpu(bg->bg_free_blocks_count_lo) |
212 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
213 (__u32)le16_to_cpu(bg->bg_free_blocks_count_hi) << 16 : 0);
216 __u32 ext4_free_inodes_count(struct super_block *sb,
217 struct ext4_group_desc *bg)
219 return le16_to_cpu(bg->bg_free_inodes_count_lo) |
220 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
221 (__u32)le16_to_cpu(bg->bg_free_inodes_count_hi) << 16 : 0);
224 __u32 ext4_used_dirs_count(struct super_block *sb,
225 struct ext4_group_desc *bg)
227 return le16_to_cpu(bg->bg_used_dirs_count_lo) |
228 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
229 (__u32)le16_to_cpu(bg->bg_used_dirs_count_hi) << 16 : 0);
232 __u32 ext4_itable_unused_count(struct super_block *sb,
233 struct ext4_group_desc *bg)
235 return le16_to_cpu(bg->bg_itable_unused_lo) |
236 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
237 (__u32)le16_to_cpu(bg->bg_itable_unused_hi) << 16 : 0);
240 void ext4_block_bitmap_set(struct super_block *sb,
241 struct ext4_group_desc *bg, ext4_fsblk_t blk)
243 bg->bg_block_bitmap_lo = cpu_to_le32((u32)blk);
244 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
245 bg->bg_block_bitmap_hi = cpu_to_le32(blk >> 32);
248 void ext4_inode_bitmap_set(struct super_block *sb,
249 struct ext4_group_desc *bg, ext4_fsblk_t blk)
251 bg->bg_inode_bitmap_lo = cpu_to_le32((u32)blk);
252 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
253 bg->bg_inode_bitmap_hi = cpu_to_le32(blk >> 32);
256 void ext4_inode_table_set(struct super_block *sb,
257 struct ext4_group_desc *bg, ext4_fsblk_t blk)
259 bg->bg_inode_table_lo = cpu_to_le32((u32)blk);
260 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
261 bg->bg_inode_table_hi = cpu_to_le32(blk >> 32);
264 void ext4_free_group_clusters_set(struct super_block *sb,
265 struct ext4_group_desc *bg, __u32 count)
267 bg->bg_free_blocks_count_lo = cpu_to_le16((__u16)count);
268 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
269 bg->bg_free_blocks_count_hi = cpu_to_le16(count >> 16);
272 void ext4_free_inodes_set(struct super_block *sb,
273 struct ext4_group_desc *bg, __u32 count)
275 bg->bg_free_inodes_count_lo = cpu_to_le16((__u16)count);
276 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
277 bg->bg_free_inodes_count_hi = cpu_to_le16(count >> 16);
280 void ext4_used_dirs_set(struct super_block *sb,
281 struct ext4_group_desc *bg, __u32 count)
283 bg->bg_used_dirs_count_lo = cpu_to_le16((__u16)count);
284 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
285 bg->bg_used_dirs_count_hi = cpu_to_le16(count >> 16);
288 void ext4_itable_unused_set(struct super_block *sb,
289 struct ext4_group_desc *bg, __u32 count)
291 bg->bg_itable_unused_lo = cpu_to_le16((__u16)count);
292 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
293 bg->bg_itable_unused_hi = cpu_to_le16(count >> 16);
297 static void __save_error_info(struct super_block *sb, const char *func,
300 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
302 EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS;
303 es->s_state |= cpu_to_le16(EXT4_ERROR_FS);
304 es->s_last_error_time = cpu_to_le32(get_seconds());
305 strncpy(es->s_last_error_func, func, sizeof(es->s_last_error_func));
306 es->s_last_error_line = cpu_to_le32(line);
307 if (!es->s_first_error_time) {
308 es->s_first_error_time = es->s_last_error_time;
309 strncpy(es->s_first_error_func, func,
310 sizeof(es->s_first_error_func));
311 es->s_first_error_line = cpu_to_le32(line);
312 es->s_first_error_ino = es->s_last_error_ino;
313 es->s_first_error_block = es->s_last_error_block;
316 * Start the daily error reporting function if it hasn't been
319 if (!es->s_error_count)
320 mod_timer(&EXT4_SB(sb)->s_err_report, jiffies + 24*60*60*HZ);
321 le32_add_cpu(&es->s_error_count, 1);
324 static void save_error_info(struct super_block *sb, const char *func,
327 __save_error_info(sb, func, line);
328 ext4_commit_super(sb, 1);
332 * The del_gendisk() function uninitializes the disk-specific data
333 * structures, including the bdi structure, without telling anyone
334 * else. Once this happens, any attempt to call mark_buffer_dirty()
335 * (for example, by ext4_commit_super), will cause a kernel OOPS.
336 * This is a kludge to prevent these oops until we can put in a proper
337 * hook in del_gendisk() to inform the VFS and file system layers.
339 static int block_device_ejected(struct super_block *sb)
341 struct inode *bd_inode = sb->s_bdev->bd_inode;
342 struct backing_dev_info *bdi = bd_inode->i_mapping->backing_dev_info;
344 return bdi->dev == NULL;
347 static void ext4_journal_commit_callback(journal_t *journal, transaction_t *txn)
349 struct super_block *sb = journal->j_private;
350 struct ext4_sb_info *sbi = EXT4_SB(sb);
351 int error = is_journal_aborted(journal);
352 struct ext4_journal_cb_entry *jce, *tmp;
354 spin_lock(&sbi->s_md_lock);
355 list_for_each_entry_safe(jce, tmp, &txn->t_private_list, jce_list) {
356 list_del_init(&jce->jce_list);
357 spin_unlock(&sbi->s_md_lock);
358 jce->jce_func(sb, jce, error);
359 spin_lock(&sbi->s_md_lock);
361 spin_unlock(&sbi->s_md_lock);
364 /* Deal with the reporting of failure conditions on a filesystem such as
365 * inconsistencies detected or read IO failures.
367 * On ext2, we can store the error state of the filesystem in the
368 * superblock. That is not possible on ext4, because we may have other
369 * write ordering constraints on the superblock which prevent us from
370 * writing it out straight away; and given that the journal is about to
371 * be aborted, we can't rely on the current, or future, transactions to
372 * write out the superblock safely.
374 * We'll just use the jbd2_journal_abort() error code to record an error in
375 * the journal instead. On recovery, the journal will complain about
376 * that error until we've noted it down and cleared it.
379 static void ext4_handle_error(struct super_block *sb)
381 if (sb->s_flags & MS_RDONLY)
384 if (!test_opt(sb, ERRORS_CONT)) {
385 journal_t *journal = EXT4_SB(sb)->s_journal;
387 EXT4_SB(sb)->s_mount_flags |= EXT4_MF_FS_ABORTED;
389 jbd2_journal_abort(journal, -EIO);
391 if (test_opt(sb, ERRORS_RO)) {
392 ext4_msg(sb, KERN_CRIT, "Remounting filesystem read-only");
393 sb->s_flags |= MS_RDONLY;
395 if (test_opt(sb, ERRORS_PANIC))
396 panic("EXT4-fs (device %s): panic forced after error\n",
400 void __ext4_error(struct super_block *sb, const char *function,
401 unsigned int line, const char *fmt, ...)
403 struct va_format vaf;
409 printk(KERN_CRIT "EXT4-fs error (device %s): %s:%d: comm %s: %pV\n",
410 sb->s_id, function, line, current->comm, &vaf);
412 save_error_info(sb, function, line);
414 ext4_handle_error(sb);
417 void ext4_error_inode(struct inode *inode, const char *function,
418 unsigned int line, ext4_fsblk_t block,
419 const char *fmt, ...)
422 struct va_format vaf;
423 struct ext4_super_block *es = EXT4_SB(inode->i_sb)->s_es;
425 es->s_last_error_ino = cpu_to_le32(inode->i_ino);
426 es->s_last_error_block = cpu_to_le64(block);
427 save_error_info(inode->i_sb, function, line);
432 printk(KERN_CRIT "EXT4-fs error (device %s): %s:%d: "
433 "inode #%lu: block %llu: comm %s: %pV\n",
434 inode->i_sb->s_id, function, line, inode->i_ino,
435 block, current->comm, &vaf);
437 printk(KERN_CRIT "EXT4-fs error (device %s): %s:%d: "
438 "inode #%lu: comm %s: %pV\n",
439 inode->i_sb->s_id, function, line, inode->i_ino,
440 current->comm, &vaf);
443 ext4_handle_error(inode->i_sb);
446 void ext4_error_file(struct file *file, const char *function,
447 unsigned int line, ext4_fsblk_t block,
448 const char *fmt, ...)
451 struct va_format vaf;
452 struct ext4_super_block *es;
453 struct inode *inode = file_inode(file);
454 char pathname[80], *path;
456 es = EXT4_SB(inode->i_sb)->s_es;
457 es->s_last_error_ino = cpu_to_le32(inode->i_ino);
458 save_error_info(inode->i_sb, function, line);
459 path = d_path(&(file->f_path), pathname, sizeof(pathname));
467 "EXT4-fs error (device %s): %s:%d: inode #%lu: "
468 "block %llu: comm %s: path %s: %pV\n",
469 inode->i_sb->s_id, function, line, inode->i_ino,
470 block, current->comm, path, &vaf);
473 "EXT4-fs error (device %s): %s:%d: inode #%lu: "
474 "comm %s: path %s: %pV\n",
475 inode->i_sb->s_id, function, line, inode->i_ino,
476 current->comm, path, &vaf);
479 ext4_handle_error(inode->i_sb);
482 const char *ext4_decode_error(struct super_block *sb, int errno,
489 errstr = "IO failure";
492 errstr = "Out of memory";
495 if (!sb || (EXT4_SB(sb)->s_journal &&
496 EXT4_SB(sb)->s_journal->j_flags & JBD2_ABORT))
497 errstr = "Journal has aborted";
499 errstr = "Readonly filesystem";
502 /* If the caller passed in an extra buffer for unknown
503 * errors, textualise them now. Else we just return
506 /* Check for truncated error codes... */
507 if (snprintf(nbuf, 16, "error %d", -errno) >= 0)
516 /* __ext4_std_error decodes expected errors from journaling functions
517 * automatically and invokes the appropriate error response. */
519 void __ext4_std_error(struct super_block *sb, const char *function,
520 unsigned int line, int errno)
525 /* Special case: if the error is EROFS, and we're not already
526 * inside a transaction, then there's really no point in logging
528 if (errno == -EROFS && journal_current_handle() == NULL &&
529 (sb->s_flags & MS_RDONLY))
532 errstr = ext4_decode_error(sb, errno, nbuf);
533 printk(KERN_CRIT "EXT4-fs error (device %s) in %s:%d: %s\n",
534 sb->s_id, function, line, errstr);
535 save_error_info(sb, function, line);
537 ext4_handle_error(sb);
541 * ext4_abort is a much stronger failure handler than ext4_error. The
542 * abort function may be used to deal with unrecoverable failures such
543 * as journal IO errors or ENOMEM at a critical moment in log management.
545 * We unconditionally force the filesystem into an ABORT|READONLY state,
546 * unless the error response on the fs has been set to panic in which
547 * case we take the easy way out and panic immediately.
550 void __ext4_abort(struct super_block *sb, const char *function,
551 unsigned int line, const char *fmt, ...)
555 save_error_info(sb, function, line);
557 printk(KERN_CRIT "EXT4-fs error (device %s): %s:%d: ", sb->s_id,
563 if ((sb->s_flags & MS_RDONLY) == 0) {
564 ext4_msg(sb, KERN_CRIT, "Remounting filesystem read-only");
565 sb->s_flags |= MS_RDONLY;
566 EXT4_SB(sb)->s_mount_flags |= EXT4_MF_FS_ABORTED;
567 if (EXT4_SB(sb)->s_journal)
568 jbd2_journal_abort(EXT4_SB(sb)->s_journal, -EIO);
569 save_error_info(sb, function, line);
571 if (test_opt(sb, ERRORS_PANIC))
572 panic("EXT4-fs panic from previous error\n");
575 void ext4_msg(struct super_block *sb, const char *prefix, const char *fmt, ...)
577 struct va_format vaf;
583 printk("%sEXT4-fs (%s): %pV\n", prefix, sb->s_id, &vaf);
587 void __ext4_warning(struct super_block *sb, const char *function,
588 unsigned int line, const char *fmt, ...)
590 struct va_format vaf;
596 printk(KERN_WARNING "EXT4-fs warning (device %s): %s:%d: %pV\n",
597 sb->s_id, function, line, &vaf);
601 void __ext4_grp_locked_error(const char *function, unsigned int line,
602 struct super_block *sb, ext4_group_t grp,
603 unsigned long ino, ext4_fsblk_t block,
604 const char *fmt, ...)
608 struct va_format vaf;
610 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
612 es->s_last_error_ino = cpu_to_le32(ino);
613 es->s_last_error_block = cpu_to_le64(block);
614 __save_error_info(sb, function, line);
620 printk(KERN_CRIT "EXT4-fs error (device %s): %s:%d: group %u, ",
621 sb->s_id, function, line, grp);
623 printk(KERN_CONT "inode %lu: ", ino);
625 printk(KERN_CONT "block %llu:", (unsigned long long) block);
626 printk(KERN_CONT "%pV\n", &vaf);
629 if (test_opt(sb, ERRORS_CONT)) {
630 ext4_commit_super(sb, 0);
634 ext4_unlock_group(sb, grp);
635 ext4_handle_error(sb);
637 * We only get here in the ERRORS_RO case; relocking the group
638 * may be dangerous, but nothing bad will happen since the
639 * filesystem will have already been marked read/only and the
640 * journal has been aborted. We return 1 as a hint to callers
641 * who might what to use the return value from
642 * ext4_grp_locked_error() to distinguish between the
643 * ERRORS_CONT and ERRORS_RO case, and perhaps return more
644 * aggressively from the ext4 function in question, with a
645 * more appropriate error code.
647 ext4_lock_group(sb, grp);
651 void ext4_update_dynamic_rev(struct super_block *sb)
653 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
655 if (le32_to_cpu(es->s_rev_level) > EXT4_GOOD_OLD_REV)
659 "updating to rev %d because of new feature flag, "
660 "running e2fsck is recommended",
663 es->s_first_ino = cpu_to_le32(EXT4_GOOD_OLD_FIRST_INO);
664 es->s_inode_size = cpu_to_le16(EXT4_GOOD_OLD_INODE_SIZE);
665 es->s_rev_level = cpu_to_le32(EXT4_DYNAMIC_REV);
666 /* leave es->s_feature_*compat flags alone */
667 /* es->s_uuid will be set by e2fsck if empty */
670 * The rest of the superblock fields should be zero, and if not it
671 * means they are likely already in use, so leave them alone. We
672 * can leave it up to e2fsck to clean up any inconsistencies there.
677 * Open the external journal device
679 static struct block_device *ext4_blkdev_get(dev_t dev, struct super_block *sb)
681 struct block_device *bdev;
682 char b[BDEVNAME_SIZE];
684 bdev = blkdev_get_by_dev(dev, FMODE_READ|FMODE_WRITE|FMODE_EXCL, sb);
690 ext4_msg(sb, KERN_ERR, "failed to open journal device %s: %ld",
691 __bdevname(dev, b), PTR_ERR(bdev));
696 * Release the journal device
698 static int ext4_blkdev_put(struct block_device *bdev)
700 return blkdev_put(bdev, FMODE_READ|FMODE_WRITE|FMODE_EXCL);
703 static int ext4_blkdev_remove(struct ext4_sb_info *sbi)
705 struct block_device *bdev;
708 bdev = sbi->journal_bdev;
710 ret = ext4_blkdev_put(bdev);
711 sbi->journal_bdev = NULL;
716 static inline struct inode *orphan_list_entry(struct list_head *l)
718 return &list_entry(l, struct ext4_inode_info, i_orphan)->vfs_inode;
721 static void dump_orphan_list(struct super_block *sb, struct ext4_sb_info *sbi)
725 ext4_msg(sb, KERN_ERR, "sb orphan head is %d",
726 le32_to_cpu(sbi->s_es->s_last_orphan));
728 printk(KERN_ERR "sb_info orphan list:\n");
729 list_for_each(l, &sbi->s_orphan) {
730 struct inode *inode = orphan_list_entry(l);
732 "inode %s:%lu at %p: mode %o, nlink %d, next %d\n",
733 inode->i_sb->s_id, inode->i_ino, inode,
734 inode->i_mode, inode->i_nlink,
739 static void ext4_put_super(struct super_block *sb)
741 struct ext4_sb_info *sbi = EXT4_SB(sb);
742 struct ext4_super_block *es = sbi->s_es;
745 ext4_unregister_li_request(sb);
746 dquot_disable(sb, -1, DQUOT_USAGE_ENABLED | DQUOT_LIMITS_ENABLED);
748 flush_workqueue(sbi->dio_unwritten_wq);
749 destroy_workqueue(sbi->dio_unwritten_wq);
751 if (sbi->s_journal) {
752 err = jbd2_journal_destroy(sbi->s_journal);
753 sbi->s_journal = NULL;
755 ext4_abort(sb, "Couldn't clean up the journal");
758 ext4_es_unregister_shrinker(sb);
759 del_timer(&sbi->s_err_report);
760 ext4_release_system_zone(sb);
762 ext4_ext_release(sb);
763 ext4_xattr_put_super(sb);
765 if (!(sb->s_flags & MS_RDONLY)) {
766 EXT4_CLEAR_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
767 es->s_state = cpu_to_le16(sbi->s_mount_state);
769 if (!(sb->s_flags & MS_RDONLY))
770 ext4_commit_super(sb, 1);
773 remove_proc_entry("options", sbi->s_proc);
774 remove_proc_entry(sb->s_id, ext4_proc_root);
776 kobject_del(&sbi->s_kobj);
778 for (i = 0; i < sbi->s_gdb_count; i++)
779 brelse(sbi->s_group_desc[i]);
780 ext4_kvfree(sbi->s_group_desc);
781 ext4_kvfree(sbi->s_flex_groups);
782 percpu_counter_destroy(&sbi->s_freeclusters_counter);
783 percpu_counter_destroy(&sbi->s_freeinodes_counter);
784 percpu_counter_destroy(&sbi->s_dirs_counter);
785 percpu_counter_destroy(&sbi->s_dirtyclusters_counter);
788 for (i = 0; i < MAXQUOTAS; i++)
789 kfree(sbi->s_qf_names[i]);
792 /* Debugging code just in case the in-memory inode orphan list
793 * isn't empty. The on-disk one can be non-empty if we've
794 * detected an error and taken the fs readonly, but the
795 * in-memory list had better be clean by this point. */
796 if (!list_empty(&sbi->s_orphan))
797 dump_orphan_list(sb, sbi);
798 J_ASSERT(list_empty(&sbi->s_orphan));
800 invalidate_bdev(sb->s_bdev);
801 if (sbi->journal_bdev && sbi->journal_bdev != sb->s_bdev) {
803 * Invalidate the journal device's buffers. We don't want them
804 * floating about in memory - the physical journal device may
805 * hotswapped, and it breaks the `ro-after' testing code.
807 sync_blockdev(sbi->journal_bdev);
808 invalidate_bdev(sbi->journal_bdev);
809 ext4_blkdev_remove(sbi);
812 kthread_stop(sbi->s_mmp_tsk);
813 sb->s_fs_info = NULL;
815 * Now that we are completely done shutting down the
816 * superblock, we need to actually destroy the kobject.
818 kobject_put(&sbi->s_kobj);
819 wait_for_completion(&sbi->s_kobj_unregister);
820 if (sbi->s_chksum_driver)
821 crypto_free_shash(sbi->s_chksum_driver);
822 kfree(sbi->s_blockgroup_lock);
826 static struct kmem_cache *ext4_inode_cachep;
829 * Called inside transaction, so use GFP_NOFS
831 static struct inode *ext4_alloc_inode(struct super_block *sb)
833 struct ext4_inode_info *ei;
835 ei = kmem_cache_alloc(ext4_inode_cachep, GFP_NOFS);
839 ei->vfs_inode.i_version = 1;
840 INIT_LIST_HEAD(&ei->i_prealloc_list);
841 spin_lock_init(&ei->i_prealloc_lock);
842 ext4_es_init_tree(&ei->i_es_tree);
843 rwlock_init(&ei->i_es_lock);
844 INIT_LIST_HEAD(&ei->i_es_lru);
846 ei->i_reserved_data_blocks = 0;
847 ei->i_reserved_meta_blocks = 0;
848 ei->i_allocated_meta_blocks = 0;
849 ei->i_da_metadata_calc_len = 0;
850 ei->i_da_metadata_calc_last_lblock = 0;
851 spin_lock_init(&(ei->i_block_reservation_lock));
853 ei->i_reserved_quota = 0;
856 INIT_LIST_HEAD(&ei->i_completed_io_list);
857 spin_lock_init(&ei->i_completed_io_lock);
859 ei->i_datasync_tid = 0;
860 atomic_set(&ei->i_ioend_count, 0);
861 atomic_set(&ei->i_unwritten, 0);
862 INIT_WORK(&ei->i_unwritten_work, ext4_end_io_work);
864 return &ei->vfs_inode;
867 static int ext4_drop_inode(struct inode *inode)
869 int drop = generic_drop_inode(inode);
871 trace_ext4_drop_inode(inode, drop);
875 static void ext4_i_callback(struct rcu_head *head)
877 struct inode *inode = container_of(head, struct inode, i_rcu);
878 kmem_cache_free(ext4_inode_cachep, EXT4_I(inode));
881 static void ext4_destroy_inode(struct inode *inode)
883 if (!list_empty(&(EXT4_I(inode)->i_orphan))) {
884 ext4_msg(inode->i_sb, KERN_ERR,
885 "Inode %lu (%p): orphan list check failed!",
886 inode->i_ino, EXT4_I(inode));
887 print_hex_dump(KERN_INFO, "", DUMP_PREFIX_ADDRESS, 16, 4,
888 EXT4_I(inode), sizeof(struct ext4_inode_info),
892 call_rcu(&inode->i_rcu, ext4_i_callback);
895 static void init_once(void *foo)
897 struct ext4_inode_info *ei = (struct ext4_inode_info *) foo;
899 INIT_LIST_HEAD(&ei->i_orphan);
900 init_rwsem(&ei->xattr_sem);
901 init_rwsem(&ei->i_data_sem);
902 inode_init_once(&ei->vfs_inode);
905 static int init_inodecache(void)
907 ext4_inode_cachep = kmem_cache_create("ext4_inode_cache",
908 sizeof(struct ext4_inode_info),
909 0, (SLAB_RECLAIM_ACCOUNT|
912 if (ext4_inode_cachep == NULL)
917 static void destroy_inodecache(void)
920 * Make sure all delayed rcu free inodes are flushed before we
924 kmem_cache_destroy(ext4_inode_cachep);
927 void ext4_clear_inode(struct inode *inode)
929 invalidate_inode_buffers(inode);
932 ext4_discard_preallocations(inode);
933 ext4_es_remove_extent(inode, 0, EXT_MAX_BLOCKS);
934 ext4_es_lru_del(inode);
935 if (EXT4_I(inode)->jinode) {
936 jbd2_journal_release_jbd_inode(EXT4_JOURNAL(inode),
937 EXT4_I(inode)->jinode);
938 jbd2_free_inode(EXT4_I(inode)->jinode);
939 EXT4_I(inode)->jinode = NULL;
943 static struct inode *ext4_nfs_get_inode(struct super_block *sb,
944 u64 ino, u32 generation)
948 if (ino < EXT4_FIRST_INO(sb) && ino != EXT4_ROOT_INO)
949 return ERR_PTR(-ESTALE);
950 if (ino > le32_to_cpu(EXT4_SB(sb)->s_es->s_inodes_count))
951 return ERR_PTR(-ESTALE);
953 /* iget isn't really right if the inode is currently unallocated!!
955 * ext4_read_inode will return a bad_inode if the inode had been
956 * deleted, so we should be safe.
958 * Currently we don't know the generation for parent directory, so
959 * a generation of 0 means "accept any"
961 inode = ext4_iget(sb, ino);
963 return ERR_CAST(inode);
964 if (generation && inode->i_generation != generation) {
966 return ERR_PTR(-ESTALE);
972 static struct dentry *ext4_fh_to_dentry(struct super_block *sb, struct fid *fid,
973 int fh_len, int fh_type)
975 return generic_fh_to_dentry(sb, fid, fh_len, fh_type,
979 static struct dentry *ext4_fh_to_parent(struct super_block *sb, struct fid *fid,
980 int fh_len, int fh_type)
982 return generic_fh_to_parent(sb, fid, fh_len, fh_type,
987 * Try to release metadata pages (indirect blocks, directories) which are
988 * mapped via the block device. Since these pages could have journal heads
989 * which would prevent try_to_free_buffers() from freeing them, we must use
990 * jbd2 layer's try_to_free_buffers() function to release them.
992 static int bdev_try_to_free_page(struct super_block *sb, struct page *page,
995 journal_t *journal = EXT4_SB(sb)->s_journal;
997 WARN_ON(PageChecked(page));
998 if (!page_has_buffers(page))
1001 return jbd2_journal_try_to_free_buffers(journal, page,
1002 wait & ~__GFP_WAIT);
1003 return try_to_free_buffers(page);
1007 #define QTYPE2NAME(t) ((t) == USRQUOTA ? "user" : "group")
1008 #define QTYPE2MOPT(on, t) ((t) == USRQUOTA?((on)##USRJQUOTA):((on)##GRPJQUOTA))
1010 static int ext4_write_dquot(struct dquot *dquot);
1011 static int ext4_acquire_dquot(struct dquot *dquot);
1012 static int ext4_release_dquot(struct dquot *dquot);
1013 static int ext4_mark_dquot_dirty(struct dquot *dquot);
1014 static int ext4_write_info(struct super_block *sb, int type);
1015 static int ext4_quota_on(struct super_block *sb, int type, int format_id,
1017 static int ext4_quota_on_sysfile(struct super_block *sb, int type,
1019 static int ext4_quota_off(struct super_block *sb, int type);
1020 static int ext4_quota_off_sysfile(struct super_block *sb, int type);
1021 static int ext4_quota_on_mount(struct super_block *sb, int type);
1022 static ssize_t ext4_quota_read(struct super_block *sb, int type, char *data,
1023 size_t len, loff_t off);
1024 static ssize_t ext4_quota_write(struct super_block *sb, int type,
1025 const char *data, size_t len, loff_t off);
1026 static int ext4_quota_enable(struct super_block *sb, int type, int format_id,
1027 unsigned int flags);
1028 static int ext4_enable_quotas(struct super_block *sb);
1030 static const struct dquot_operations ext4_quota_operations = {
1031 .get_reserved_space = ext4_get_reserved_space,
1032 .write_dquot = ext4_write_dquot,
1033 .acquire_dquot = ext4_acquire_dquot,
1034 .release_dquot = ext4_release_dquot,
1035 .mark_dirty = ext4_mark_dquot_dirty,
1036 .write_info = ext4_write_info,
1037 .alloc_dquot = dquot_alloc,
1038 .destroy_dquot = dquot_destroy,
1041 static const struct quotactl_ops ext4_qctl_operations = {
1042 .quota_on = ext4_quota_on,
1043 .quota_off = ext4_quota_off,
1044 .quota_sync = dquot_quota_sync,
1045 .get_info = dquot_get_dqinfo,
1046 .set_info = dquot_set_dqinfo,
1047 .get_dqblk = dquot_get_dqblk,
1048 .set_dqblk = dquot_set_dqblk
1051 static const struct quotactl_ops ext4_qctl_sysfile_operations = {
1052 .quota_on_meta = ext4_quota_on_sysfile,
1053 .quota_off = ext4_quota_off_sysfile,
1054 .quota_sync = dquot_quota_sync,
1055 .get_info = dquot_get_dqinfo,
1056 .set_info = dquot_set_dqinfo,
1057 .get_dqblk = dquot_get_dqblk,
1058 .set_dqblk = dquot_set_dqblk
1062 static const struct super_operations ext4_sops = {
1063 .alloc_inode = ext4_alloc_inode,
1064 .destroy_inode = ext4_destroy_inode,
1065 .write_inode = ext4_write_inode,
1066 .dirty_inode = ext4_dirty_inode,
1067 .drop_inode = ext4_drop_inode,
1068 .evict_inode = ext4_evict_inode,
1069 .put_super = ext4_put_super,
1070 .sync_fs = ext4_sync_fs,
1071 .freeze_fs = ext4_freeze,
1072 .unfreeze_fs = ext4_unfreeze,
1073 .statfs = ext4_statfs,
1074 .remount_fs = ext4_remount,
1075 .show_options = ext4_show_options,
1077 .quota_read = ext4_quota_read,
1078 .quota_write = ext4_quota_write,
1080 .bdev_try_to_free_page = bdev_try_to_free_page,
1083 static const struct super_operations ext4_nojournal_sops = {
1084 .alloc_inode = ext4_alloc_inode,
1085 .destroy_inode = ext4_destroy_inode,
1086 .write_inode = ext4_write_inode,
1087 .dirty_inode = ext4_dirty_inode,
1088 .drop_inode = ext4_drop_inode,
1089 .evict_inode = ext4_evict_inode,
1090 .put_super = ext4_put_super,
1091 .statfs = ext4_statfs,
1092 .remount_fs = ext4_remount,
1093 .show_options = ext4_show_options,
1095 .quota_read = ext4_quota_read,
1096 .quota_write = ext4_quota_write,
1098 .bdev_try_to_free_page = bdev_try_to_free_page,
1101 static const struct export_operations ext4_export_ops = {
1102 .fh_to_dentry = ext4_fh_to_dentry,
1103 .fh_to_parent = ext4_fh_to_parent,
1104 .get_parent = ext4_get_parent,
1108 Opt_bsd_df, Opt_minix_df, Opt_grpid, Opt_nogrpid,
1109 Opt_resgid, Opt_resuid, Opt_sb, Opt_err_cont, Opt_err_panic, Opt_err_ro,
1110 Opt_nouid32, Opt_debug, Opt_removed,
1111 Opt_user_xattr, Opt_nouser_xattr, Opt_acl, Opt_noacl,
1112 Opt_auto_da_alloc, Opt_noauto_da_alloc, Opt_noload,
1113 Opt_commit, Opt_min_batch_time, Opt_max_batch_time,
1114 Opt_journal_dev, Opt_journal_checksum, Opt_journal_async_commit,
1115 Opt_abort, Opt_data_journal, Opt_data_ordered, Opt_data_writeback,
1116 Opt_data_err_abort, Opt_data_err_ignore,
1117 Opt_usrjquota, Opt_grpjquota, Opt_offusrjquota, Opt_offgrpjquota,
1118 Opt_jqfmt_vfsold, Opt_jqfmt_vfsv0, Opt_jqfmt_vfsv1, Opt_quota,
1119 Opt_noquota, Opt_barrier, Opt_nobarrier, Opt_err,
1120 Opt_usrquota, Opt_grpquota, Opt_i_version,
1121 Opt_stripe, Opt_delalloc, Opt_nodelalloc, Opt_mblk_io_submit,
1122 Opt_nomblk_io_submit, Opt_block_validity, Opt_noblock_validity,
1123 Opt_inode_readahead_blks, Opt_journal_ioprio,
1124 Opt_dioread_nolock, Opt_dioread_lock,
1125 Opt_discard, Opt_nodiscard, Opt_init_itable, Opt_noinit_itable,
1126 Opt_max_dir_size_kb,
1129 static const match_table_t tokens = {
1130 {Opt_bsd_df, "bsddf"},
1131 {Opt_minix_df, "minixdf"},
1132 {Opt_grpid, "grpid"},
1133 {Opt_grpid, "bsdgroups"},
1134 {Opt_nogrpid, "nogrpid"},
1135 {Opt_nogrpid, "sysvgroups"},
1136 {Opt_resgid, "resgid=%u"},
1137 {Opt_resuid, "resuid=%u"},
1139 {Opt_err_cont, "errors=continue"},
1140 {Opt_err_panic, "errors=panic"},
1141 {Opt_err_ro, "errors=remount-ro"},
1142 {Opt_nouid32, "nouid32"},
1143 {Opt_debug, "debug"},
1144 {Opt_removed, "oldalloc"},
1145 {Opt_removed, "orlov"},
1146 {Opt_user_xattr, "user_xattr"},
1147 {Opt_nouser_xattr, "nouser_xattr"},
1149 {Opt_noacl, "noacl"},
1150 {Opt_noload, "norecovery"},
1151 {Opt_noload, "noload"},
1152 {Opt_removed, "nobh"},
1153 {Opt_removed, "bh"},
1154 {Opt_commit, "commit=%u"},
1155 {Opt_min_batch_time, "min_batch_time=%u"},
1156 {Opt_max_batch_time, "max_batch_time=%u"},
1157 {Opt_journal_dev, "journal_dev=%u"},
1158 {Opt_journal_checksum, "journal_checksum"},
1159 {Opt_journal_async_commit, "journal_async_commit"},
1160 {Opt_abort, "abort"},
1161 {Opt_data_journal, "data=journal"},
1162 {Opt_data_ordered, "data=ordered"},
1163 {Opt_data_writeback, "data=writeback"},
1164 {Opt_data_err_abort, "data_err=abort"},
1165 {Opt_data_err_ignore, "data_err=ignore"},
1166 {Opt_offusrjquota, "usrjquota="},
1167 {Opt_usrjquota, "usrjquota=%s"},
1168 {Opt_offgrpjquota, "grpjquota="},
1169 {Opt_grpjquota, "grpjquota=%s"},
1170 {Opt_jqfmt_vfsold, "jqfmt=vfsold"},
1171 {Opt_jqfmt_vfsv0, "jqfmt=vfsv0"},
1172 {Opt_jqfmt_vfsv1, "jqfmt=vfsv1"},
1173 {Opt_grpquota, "grpquota"},
1174 {Opt_noquota, "noquota"},
1175 {Opt_quota, "quota"},
1176 {Opt_usrquota, "usrquota"},
1177 {Opt_barrier, "barrier=%u"},
1178 {Opt_barrier, "barrier"},
1179 {Opt_nobarrier, "nobarrier"},
1180 {Opt_i_version, "i_version"},
1181 {Opt_stripe, "stripe=%u"},
1182 {Opt_delalloc, "delalloc"},
1183 {Opt_nodelalloc, "nodelalloc"},
1184 {Opt_removed, "mblk_io_submit"},
1185 {Opt_removed, "nomblk_io_submit"},
1186 {Opt_block_validity, "block_validity"},
1187 {Opt_noblock_validity, "noblock_validity"},
1188 {Opt_inode_readahead_blks, "inode_readahead_blks=%u"},
1189 {Opt_journal_ioprio, "journal_ioprio=%u"},
1190 {Opt_auto_da_alloc, "auto_da_alloc=%u"},
1191 {Opt_auto_da_alloc, "auto_da_alloc"},
1192 {Opt_noauto_da_alloc, "noauto_da_alloc"},
1193 {Opt_dioread_nolock, "dioread_nolock"},
1194 {Opt_dioread_lock, "dioread_lock"},
1195 {Opt_discard, "discard"},
1196 {Opt_nodiscard, "nodiscard"},
1197 {Opt_init_itable, "init_itable=%u"},
1198 {Opt_init_itable, "init_itable"},
1199 {Opt_noinit_itable, "noinit_itable"},
1200 {Opt_max_dir_size_kb, "max_dir_size_kb=%u"},
1201 {Opt_removed, "check=none"}, /* mount option from ext2/3 */
1202 {Opt_removed, "nocheck"}, /* mount option from ext2/3 */
1203 {Opt_removed, "reservation"}, /* mount option from ext2/3 */
1204 {Opt_removed, "noreservation"}, /* mount option from ext2/3 */
1205 {Opt_removed, "journal=%u"}, /* mount option from ext2/3 */
1209 static ext4_fsblk_t get_sb_block(void **data)
1211 ext4_fsblk_t sb_block;
1212 char *options = (char *) *data;
1214 if (!options || strncmp(options, "sb=", 3) != 0)
1215 return 1; /* Default location */
1218 /* TODO: use simple_strtoll with >32bit ext4 */
1219 sb_block = simple_strtoul(options, &options, 0);
1220 if (*options && *options != ',') {
1221 printk(KERN_ERR "EXT4-fs: Invalid sb specification: %s\n",
1225 if (*options == ',')
1227 *data = (void *) options;
1232 #define DEFAULT_JOURNAL_IOPRIO (IOPRIO_PRIO_VALUE(IOPRIO_CLASS_BE, 3))
1233 static char deprecated_msg[] = "Mount option \"%s\" will be removed by %s\n"
1234 "Contact linux-ext4@vger.kernel.org if you think we should keep it.\n";
1237 static int set_qf_name(struct super_block *sb, int qtype, substring_t *args)
1239 struct ext4_sb_info *sbi = EXT4_SB(sb);
1243 if (sb_any_quota_loaded(sb) &&
1244 !sbi->s_qf_names[qtype]) {
1245 ext4_msg(sb, KERN_ERR,
1246 "Cannot change journaled "
1247 "quota options when quota turned on");
1250 qname = match_strdup(args);
1252 ext4_msg(sb, KERN_ERR,
1253 "Not enough memory for storing quotafile name");
1256 if (sbi->s_qf_names[qtype]) {
1257 if (strcmp(sbi->s_qf_names[qtype], qname) == 0)
1260 ext4_msg(sb, KERN_ERR,
1261 "%s quota file already specified",
1265 if (strchr(qname, '/')) {
1266 ext4_msg(sb, KERN_ERR,
1267 "quotafile must be on filesystem root");
1270 sbi->s_qf_names[qtype] = qname;
1278 static int clear_qf_name(struct super_block *sb, int qtype)
1281 struct ext4_sb_info *sbi = EXT4_SB(sb);
1283 if (sb_any_quota_loaded(sb) &&
1284 sbi->s_qf_names[qtype]) {
1285 ext4_msg(sb, KERN_ERR, "Cannot change journaled quota options"
1286 " when quota turned on");
1289 kfree(sbi->s_qf_names[qtype]);
1290 sbi->s_qf_names[qtype] = NULL;
1295 #define MOPT_SET 0x0001
1296 #define MOPT_CLEAR 0x0002
1297 #define MOPT_NOSUPPORT 0x0004
1298 #define MOPT_EXPLICIT 0x0008
1299 #define MOPT_CLEAR_ERR 0x0010
1300 #define MOPT_GTE0 0x0020
1303 #define MOPT_QFMT 0x0040
1305 #define MOPT_Q MOPT_NOSUPPORT
1306 #define MOPT_QFMT MOPT_NOSUPPORT
1308 #define MOPT_DATAJ 0x0080
1309 #define MOPT_NO_EXT2 0x0100
1310 #define MOPT_NO_EXT3 0x0200
1311 #define MOPT_EXT4_ONLY (MOPT_NO_EXT2 | MOPT_NO_EXT3)
1313 static const struct mount_opts {
1317 } ext4_mount_opts[] = {
1318 {Opt_minix_df, EXT4_MOUNT_MINIX_DF, MOPT_SET},
1319 {Opt_bsd_df, EXT4_MOUNT_MINIX_DF, MOPT_CLEAR},
1320 {Opt_grpid, EXT4_MOUNT_GRPID, MOPT_SET},
1321 {Opt_nogrpid, EXT4_MOUNT_GRPID, MOPT_CLEAR},
1322 {Opt_block_validity, EXT4_MOUNT_BLOCK_VALIDITY, MOPT_SET},
1323 {Opt_noblock_validity, EXT4_MOUNT_BLOCK_VALIDITY, MOPT_CLEAR},
1324 {Opt_dioread_nolock, EXT4_MOUNT_DIOREAD_NOLOCK,
1325 MOPT_EXT4_ONLY | MOPT_SET},
1326 {Opt_dioread_lock, EXT4_MOUNT_DIOREAD_NOLOCK,
1327 MOPT_EXT4_ONLY | MOPT_CLEAR},
1328 {Opt_discard, EXT4_MOUNT_DISCARD, MOPT_SET},
1329 {Opt_nodiscard, EXT4_MOUNT_DISCARD, MOPT_CLEAR},
1330 {Opt_delalloc, EXT4_MOUNT_DELALLOC,
1331 MOPT_EXT4_ONLY | MOPT_SET | MOPT_EXPLICIT},
1332 {Opt_nodelalloc, EXT4_MOUNT_DELALLOC,
1333 MOPT_EXT4_ONLY | MOPT_CLEAR | MOPT_EXPLICIT},
1334 {Opt_journal_checksum, EXT4_MOUNT_JOURNAL_CHECKSUM,
1335 MOPT_EXT4_ONLY | MOPT_SET},
1336 {Opt_journal_async_commit, (EXT4_MOUNT_JOURNAL_ASYNC_COMMIT |
1337 EXT4_MOUNT_JOURNAL_CHECKSUM),
1338 MOPT_EXT4_ONLY | MOPT_SET},
1339 {Opt_noload, EXT4_MOUNT_NOLOAD, MOPT_NO_EXT2 | MOPT_SET},
1340 {Opt_err_panic, EXT4_MOUNT_ERRORS_PANIC, MOPT_SET | MOPT_CLEAR_ERR},
1341 {Opt_err_ro, EXT4_MOUNT_ERRORS_RO, MOPT_SET | MOPT_CLEAR_ERR},
1342 {Opt_err_cont, EXT4_MOUNT_ERRORS_CONT, MOPT_SET | MOPT_CLEAR_ERR},
1343 {Opt_data_err_abort, EXT4_MOUNT_DATA_ERR_ABORT,
1344 MOPT_NO_EXT2 | MOPT_SET},
1345 {Opt_data_err_ignore, EXT4_MOUNT_DATA_ERR_ABORT,
1346 MOPT_NO_EXT2 | MOPT_CLEAR},
1347 {Opt_barrier, EXT4_MOUNT_BARRIER, MOPT_SET},
1348 {Opt_nobarrier, EXT4_MOUNT_BARRIER, MOPT_CLEAR},
1349 {Opt_noauto_da_alloc, EXT4_MOUNT_NO_AUTO_DA_ALLOC, MOPT_SET},
1350 {Opt_auto_da_alloc, EXT4_MOUNT_NO_AUTO_DA_ALLOC, MOPT_CLEAR},
1351 {Opt_noinit_itable, EXT4_MOUNT_INIT_INODE_TABLE, MOPT_CLEAR},
1352 {Opt_commit, 0, MOPT_GTE0},
1353 {Opt_max_batch_time, 0, MOPT_GTE0},
1354 {Opt_min_batch_time, 0, MOPT_GTE0},
1355 {Opt_inode_readahead_blks, 0, MOPT_GTE0},
1356 {Opt_init_itable, 0, MOPT_GTE0},
1357 {Opt_stripe, 0, MOPT_GTE0},
1358 {Opt_resuid, 0, MOPT_GTE0},
1359 {Opt_resgid, 0, MOPT_GTE0},
1360 {Opt_journal_dev, 0, MOPT_GTE0},
1361 {Opt_journal_ioprio, 0, MOPT_GTE0},
1362 {Opt_data_journal, EXT4_MOUNT_JOURNAL_DATA, MOPT_NO_EXT2 | MOPT_DATAJ},
1363 {Opt_data_ordered, EXT4_MOUNT_ORDERED_DATA, MOPT_NO_EXT2 | MOPT_DATAJ},
1364 {Opt_data_writeback, EXT4_MOUNT_WRITEBACK_DATA,
1365 MOPT_NO_EXT2 | MOPT_DATAJ},
1366 {Opt_user_xattr, EXT4_MOUNT_XATTR_USER, MOPT_SET},
1367 {Opt_nouser_xattr, EXT4_MOUNT_XATTR_USER, MOPT_CLEAR},
1368 #ifdef CONFIG_EXT4_FS_POSIX_ACL
1369 {Opt_acl, EXT4_MOUNT_POSIX_ACL, MOPT_SET},
1370 {Opt_noacl, EXT4_MOUNT_POSIX_ACL, MOPT_CLEAR},
1372 {Opt_acl, 0, MOPT_NOSUPPORT},
1373 {Opt_noacl, 0, MOPT_NOSUPPORT},
1375 {Opt_nouid32, EXT4_MOUNT_NO_UID32, MOPT_SET},
1376 {Opt_debug, EXT4_MOUNT_DEBUG, MOPT_SET},
1377 {Opt_quota, EXT4_MOUNT_QUOTA | EXT4_MOUNT_USRQUOTA, MOPT_SET | MOPT_Q},
1378 {Opt_usrquota, EXT4_MOUNT_QUOTA | EXT4_MOUNT_USRQUOTA,
1380 {Opt_grpquota, EXT4_MOUNT_QUOTA | EXT4_MOUNT_GRPQUOTA,
1382 {Opt_noquota, (EXT4_MOUNT_QUOTA | EXT4_MOUNT_USRQUOTA |
1383 EXT4_MOUNT_GRPQUOTA), MOPT_CLEAR | MOPT_Q},
1384 {Opt_usrjquota, 0, MOPT_Q},
1385 {Opt_grpjquota, 0, MOPT_Q},
1386 {Opt_offusrjquota, 0, MOPT_Q},
1387 {Opt_offgrpjquota, 0, MOPT_Q},
1388 {Opt_jqfmt_vfsold, QFMT_VFS_OLD, MOPT_QFMT},
1389 {Opt_jqfmt_vfsv0, QFMT_VFS_V0, MOPT_QFMT},
1390 {Opt_jqfmt_vfsv1, QFMT_VFS_V1, MOPT_QFMT},
1391 {Opt_max_dir_size_kb, 0, MOPT_GTE0},
1395 static int handle_mount_opt(struct super_block *sb, char *opt, int token,
1396 substring_t *args, unsigned long *journal_devnum,
1397 unsigned int *journal_ioprio, int is_remount)
1399 struct ext4_sb_info *sbi = EXT4_SB(sb);
1400 const struct mount_opts *m;
1406 if (token == Opt_usrjquota)
1407 return set_qf_name(sb, USRQUOTA, &args[0]);
1408 else if (token == Opt_grpjquota)
1409 return set_qf_name(sb, GRPQUOTA, &args[0]);
1410 else if (token == Opt_offusrjquota)
1411 return clear_qf_name(sb, USRQUOTA);
1412 else if (token == Opt_offgrpjquota)
1413 return clear_qf_name(sb, GRPQUOTA);
1417 case Opt_nouser_xattr:
1418 ext4_msg(sb, KERN_WARNING, deprecated_msg, opt, "3.5");
1421 return 1; /* handled by get_sb_block() */
1423 ext4_msg(sb, KERN_WARNING, "Ignoring removed %s option", opt);
1426 sbi->s_mount_flags |= EXT4_MF_FS_ABORTED;
1429 sb->s_flags |= MS_I_VERSION;
1433 for (m = ext4_mount_opts; m->token != Opt_err; m++)
1434 if (token == m->token)
1437 if (m->token == Opt_err) {
1438 ext4_msg(sb, KERN_ERR, "Unrecognized mount option \"%s\" "
1439 "or missing value", opt);
1443 if ((m->flags & MOPT_NO_EXT2) && IS_EXT2_SB(sb)) {
1444 ext4_msg(sb, KERN_ERR,
1445 "Mount option \"%s\" incompatible with ext2", opt);
1448 if ((m->flags & MOPT_NO_EXT3) && IS_EXT3_SB(sb)) {
1449 ext4_msg(sb, KERN_ERR,
1450 "Mount option \"%s\" incompatible with ext3", opt);
1454 if (args->from && match_int(args, &arg))
1456 if (args->from && (m->flags & MOPT_GTE0) && (arg < 0))
1458 if (m->flags & MOPT_EXPLICIT)
1459 set_opt2(sb, EXPLICIT_DELALLOC);
1460 if (m->flags & MOPT_CLEAR_ERR)
1461 clear_opt(sb, ERRORS_MASK);
1462 if (token == Opt_noquota && sb_any_quota_loaded(sb)) {
1463 ext4_msg(sb, KERN_ERR, "Cannot change quota "
1464 "options when quota turned on");
1468 if (m->flags & MOPT_NOSUPPORT) {
1469 ext4_msg(sb, KERN_ERR, "%s option not supported", opt);
1470 } else if (token == Opt_commit) {
1472 arg = JBD2_DEFAULT_MAX_COMMIT_AGE;
1473 sbi->s_commit_interval = HZ * arg;
1474 } else if (token == Opt_max_batch_time) {
1476 arg = EXT4_DEF_MAX_BATCH_TIME;
1477 sbi->s_max_batch_time = arg;
1478 } else if (token == Opt_min_batch_time) {
1479 sbi->s_min_batch_time = arg;
1480 } else if (token == Opt_inode_readahead_blks) {
1481 if (arg && (arg > (1 << 30) || !is_power_of_2(arg))) {
1482 ext4_msg(sb, KERN_ERR,
1483 "EXT4-fs: inode_readahead_blks must be "
1484 "0 or a power of 2 smaller than 2^31");
1487 sbi->s_inode_readahead_blks = arg;
1488 } else if (token == Opt_init_itable) {
1489 set_opt(sb, INIT_INODE_TABLE);
1491 arg = EXT4_DEF_LI_WAIT_MULT;
1492 sbi->s_li_wait_mult = arg;
1493 } else if (token == Opt_max_dir_size_kb) {
1494 sbi->s_max_dir_size_kb = arg;
1495 } else if (token == Opt_stripe) {
1496 sbi->s_stripe = arg;
1497 } else if (token == Opt_resuid) {
1498 uid = make_kuid(current_user_ns(), arg);
1499 if (!uid_valid(uid)) {
1500 ext4_msg(sb, KERN_ERR, "Invalid uid value %d", arg);
1503 sbi->s_resuid = uid;
1504 } else if (token == Opt_resgid) {
1505 gid = make_kgid(current_user_ns(), arg);
1506 if (!gid_valid(gid)) {
1507 ext4_msg(sb, KERN_ERR, "Invalid gid value %d", arg);
1510 sbi->s_resgid = gid;
1511 } else if (token == Opt_journal_dev) {
1513 ext4_msg(sb, KERN_ERR,
1514 "Cannot specify journal on remount");
1517 *journal_devnum = arg;
1518 } else if (token == Opt_journal_ioprio) {
1520 ext4_msg(sb, KERN_ERR, "Invalid journal IO priority"
1525 IOPRIO_PRIO_VALUE(IOPRIO_CLASS_BE, arg);
1526 } else if (m->flags & MOPT_DATAJ) {
1528 if (!sbi->s_journal)
1529 ext4_msg(sb, KERN_WARNING, "Remounting file system with no journal so ignoring journalled data option");
1530 else if (test_opt(sb, DATA_FLAGS) != m->mount_opt) {
1531 ext4_msg(sb, KERN_ERR,
1532 "Cannot change data mode on remount");
1536 clear_opt(sb, DATA_FLAGS);
1537 sbi->s_mount_opt |= m->mount_opt;
1540 } else if (m->flags & MOPT_QFMT) {
1541 if (sb_any_quota_loaded(sb) &&
1542 sbi->s_jquota_fmt != m->mount_opt) {
1543 ext4_msg(sb, KERN_ERR, "Cannot change journaled "
1544 "quota options when quota turned on");
1547 sbi->s_jquota_fmt = m->mount_opt;
1552 if (m->flags & MOPT_CLEAR)
1554 else if (unlikely(!(m->flags & MOPT_SET))) {
1555 ext4_msg(sb, KERN_WARNING,
1556 "buggy handling of option %s", opt);
1561 sbi->s_mount_opt |= m->mount_opt;
1563 sbi->s_mount_opt &= ~m->mount_opt;
1568 static int parse_options(char *options, struct super_block *sb,
1569 unsigned long *journal_devnum,
1570 unsigned int *journal_ioprio,
1573 struct ext4_sb_info *sbi = EXT4_SB(sb);
1575 substring_t args[MAX_OPT_ARGS];
1581 while ((p = strsep(&options, ",")) != NULL) {
1585 * Initialize args struct so we know whether arg was
1586 * found; some options take optional arguments.
1588 args[0].to = args[0].from = NULL;
1589 token = match_token(p, tokens, args);
1590 if (handle_mount_opt(sb, p, token, args, journal_devnum,
1591 journal_ioprio, is_remount) < 0)
1595 if (sbi->s_qf_names[USRQUOTA] || sbi->s_qf_names[GRPQUOTA]) {
1596 if (test_opt(sb, USRQUOTA) && sbi->s_qf_names[USRQUOTA])
1597 clear_opt(sb, USRQUOTA);
1599 if (test_opt(sb, GRPQUOTA) && sbi->s_qf_names[GRPQUOTA])
1600 clear_opt(sb, GRPQUOTA);
1602 if (test_opt(sb, GRPQUOTA) || test_opt(sb, USRQUOTA)) {
1603 ext4_msg(sb, KERN_ERR, "old and new quota "
1608 if (!sbi->s_jquota_fmt) {
1609 ext4_msg(sb, KERN_ERR, "journaled quota format "
1614 if (sbi->s_jquota_fmt) {
1615 ext4_msg(sb, KERN_ERR, "journaled quota format "
1616 "specified with no journaling "
1622 if (test_opt(sb, DIOREAD_NOLOCK)) {
1624 BLOCK_SIZE << le32_to_cpu(sbi->s_es->s_log_block_size);
1626 if (blocksize < PAGE_CACHE_SIZE) {
1627 ext4_msg(sb, KERN_ERR, "can't mount with "
1628 "dioread_nolock if block size != PAGE_SIZE");
1635 static inline void ext4_show_quota_options(struct seq_file *seq,
1636 struct super_block *sb)
1638 #if defined(CONFIG_QUOTA)
1639 struct ext4_sb_info *sbi = EXT4_SB(sb);
1641 if (sbi->s_jquota_fmt) {
1644 switch (sbi->s_jquota_fmt) {
1655 seq_printf(seq, ",jqfmt=%s", fmtname);
1658 if (sbi->s_qf_names[USRQUOTA])
1659 seq_printf(seq, ",usrjquota=%s", sbi->s_qf_names[USRQUOTA]);
1661 if (sbi->s_qf_names[GRPQUOTA])
1662 seq_printf(seq, ",grpjquota=%s", sbi->s_qf_names[GRPQUOTA]);
1664 if (test_opt(sb, USRQUOTA))
1665 seq_puts(seq, ",usrquota");
1667 if (test_opt(sb, GRPQUOTA))
1668 seq_puts(seq, ",grpquota");
1672 static const char *token2str(int token)
1674 const struct match_token *t;
1676 for (t = tokens; t->token != Opt_err; t++)
1677 if (t->token == token && !strchr(t->pattern, '='))
1684 * - it's set to a non-default value OR
1685 * - if the per-sb default is different from the global default
1687 static int _ext4_show_options(struct seq_file *seq, struct super_block *sb,
1690 struct ext4_sb_info *sbi = EXT4_SB(sb);
1691 struct ext4_super_block *es = sbi->s_es;
1692 int def_errors, def_mount_opt = nodefs ? 0 : sbi->s_def_mount_opt;
1693 const struct mount_opts *m;
1694 char sep = nodefs ? '\n' : ',';
1696 #define SEQ_OPTS_PUTS(str) seq_printf(seq, "%c" str, sep)
1697 #define SEQ_OPTS_PRINT(str, arg) seq_printf(seq, "%c" str, sep, arg)
1699 if (sbi->s_sb_block != 1)
1700 SEQ_OPTS_PRINT("sb=%llu", sbi->s_sb_block);
1702 for (m = ext4_mount_opts; m->token != Opt_err; m++) {
1703 int want_set = m->flags & MOPT_SET;
1704 if (((m->flags & (MOPT_SET|MOPT_CLEAR)) == 0) ||
1705 (m->flags & MOPT_CLEAR_ERR))
1707 if (!(m->mount_opt & (sbi->s_mount_opt ^ def_mount_opt)))
1708 continue; /* skip if same as the default */
1710 (sbi->s_mount_opt & m->mount_opt) != m->mount_opt) ||
1711 (!want_set && (sbi->s_mount_opt & m->mount_opt)))
1712 continue; /* select Opt_noFoo vs Opt_Foo */
1713 SEQ_OPTS_PRINT("%s", token2str(m->token));
1716 if (nodefs || !uid_eq(sbi->s_resuid, make_kuid(&init_user_ns, EXT4_DEF_RESUID)) ||
1717 le16_to_cpu(es->s_def_resuid) != EXT4_DEF_RESUID)
1718 SEQ_OPTS_PRINT("resuid=%u",
1719 from_kuid_munged(&init_user_ns, sbi->s_resuid));
1720 if (nodefs || !gid_eq(sbi->s_resgid, make_kgid(&init_user_ns, EXT4_DEF_RESGID)) ||
1721 le16_to_cpu(es->s_def_resgid) != EXT4_DEF_RESGID)
1722 SEQ_OPTS_PRINT("resgid=%u",
1723 from_kgid_munged(&init_user_ns, sbi->s_resgid));
1724 def_errors = nodefs ? -1 : le16_to_cpu(es->s_errors);
1725 if (test_opt(sb, ERRORS_RO) && def_errors != EXT4_ERRORS_RO)
1726 SEQ_OPTS_PUTS("errors=remount-ro");
1727 if (test_opt(sb, ERRORS_CONT) && def_errors != EXT4_ERRORS_CONTINUE)
1728 SEQ_OPTS_PUTS("errors=continue");
1729 if (test_opt(sb, ERRORS_PANIC) && def_errors != EXT4_ERRORS_PANIC)
1730 SEQ_OPTS_PUTS("errors=panic");
1731 if (nodefs || sbi->s_commit_interval != JBD2_DEFAULT_MAX_COMMIT_AGE*HZ)
1732 SEQ_OPTS_PRINT("commit=%lu", sbi->s_commit_interval / HZ);
1733 if (nodefs || sbi->s_min_batch_time != EXT4_DEF_MIN_BATCH_TIME)
1734 SEQ_OPTS_PRINT("min_batch_time=%u", sbi->s_min_batch_time);
1735 if (nodefs || sbi->s_max_batch_time != EXT4_DEF_MAX_BATCH_TIME)
1736 SEQ_OPTS_PRINT("max_batch_time=%u", sbi->s_max_batch_time);
1737 if (sb->s_flags & MS_I_VERSION)
1738 SEQ_OPTS_PUTS("i_version");
1739 if (nodefs || sbi->s_stripe)
1740 SEQ_OPTS_PRINT("stripe=%lu", sbi->s_stripe);
1741 if (EXT4_MOUNT_DATA_FLAGS & (sbi->s_mount_opt ^ def_mount_opt)) {
1742 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA)
1743 SEQ_OPTS_PUTS("data=journal");
1744 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA)
1745 SEQ_OPTS_PUTS("data=ordered");
1746 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_WRITEBACK_DATA)
1747 SEQ_OPTS_PUTS("data=writeback");
1750 sbi->s_inode_readahead_blks != EXT4_DEF_INODE_READAHEAD_BLKS)
1751 SEQ_OPTS_PRINT("inode_readahead_blks=%u",
1752 sbi->s_inode_readahead_blks);
1754 if (nodefs || (test_opt(sb, INIT_INODE_TABLE) &&
1755 (sbi->s_li_wait_mult != EXT4_DEF_LI_WAIT_MULT)))
1756 SEQ_OPTS_PRINT("init_itable=%u", sbi->s_li_wait_mult);
1757 if (nodefs || sbi->s_max_dir_size_kb)
1758 SEQ_OPTS_PRINT("max_dir_size_kb=%u", sbi->s_max_dir_size_kb);
1760 ext4_show_quota_options(seq, sb);
1764 static int ext4_show_options(struct seq_file *seq, struct dentry *root)
1766 return _ext4_show_options(seq, root->d_sb, 0);
1769 static int options_seq_show(struct seq_file *seq, void *offset)
1771 struct super_block *sb = seq->private;
1774 seq_puts(seq, (sb->s_flags & MS_RDONLY) ? "ro" : "rw");
1775 rc = _ext4_show_options(seq, sb, 1);
1776 seq_puts(seq, "\n");
1780 static int options_open_fs(struct inode *inode, struct file *file)
1782 return single_open(file, options_seq_show, PDE(inode)->data);
1785 static const struct file_operations ext4_seq_options_fops = {
1786 .owner = THIS_MODULE,
1787 .open = options_open_fs,
1789 .llseek = seq_lseek,
1790 .release = single_release,
1793 static int ext4_setup_super(struct super_block *sb, struct ext4_super_block *es,
1796 struct ext4_sb_info *sbi = EXT4_SB(sb);
1799 if (le32_to_cpu(es->s_rev_level) > EXT4_MAX_SUPP_REV) {
1800 ext4_msg(sb, KERN_ERR, "revision level too high, "
1801 "forcing read-only mode");
1806 if (!(sbi->s_mount_state & EXT4_VALID_FS))
1807 ext4_msg(sb, KERN_WARNING, "warning: mounting unchecked fs, "
1808 "running e2fsck is recommended");
1809 else if ((sbi->s_mount_state & EXT4_ERROR_FS))
1810 ext4_msg(sb, KERN_WARNING,
1811 "warning: mounting fs with errors, "
1812 "running e2fsck is recommended");
1813 else if ((__s16) le16_to_cpu(es->s_max_mnt_count) > 0 &&
1814 le16_to_cpu(es->s_mnt_count) >=
1815 (unsigned short) (__s16) le16_to_cpu(es->s_max_mnt_count))
1816 ext4_msg(sb, KERN_WARNING,
1817 "warning: maximal mount count reached, "
1818 "running e2fsck is recommended");
1819 else if (le32_to_cpu(es->s_checkinterval) &&
1820 (le32_to_cpu(es->s_lastcheck) +
1821 le32_to_cpu(es->s_checkinterval) <= get_seconds()))
1822 ext4_msg(sb, KERN_WARNING,
1823 "warning: checktime reached, "
1824 "running e2fsck is recommended");
1825 if (!sbi->s_journal)
1826 es->s_state &= cpu_to_le16(~EXT4_VALID_FS);
1827 if (!(__s16) le16_to_cpu(es->s_max_mnt_count))
1828 es->s_max_mnt_count = cpu_to_le16(EXT4_DFL_MAX_MNT_COUNT);
1829 le16_add_cpu(&es->s_mnt_count, 1);
1830 es->s_mtime = cpu_to_le32(get_seconds());
1831 ext4_update_dynamic_rev(sb);
1833 EXT4_SET_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
1835 ext4_commit_super(sb, 1);
1837 if (test_opt(sb, DEBUG))
1838 printk(KERN_INFO "[EXT4 FS bs=%lu, gc=%u, "
1839 "bpg=%lu, ipg=%lu, mo=%04x, mo2=%04x]\n",
1841 sbi->s_groups_count,
1842 EXT4_BLOCKS_PER_GROUP(sb),
1843 EXT4_INODES_PER_GROUP(sb),
1844 sbi->s_mount_opt, sbi->s_mount_opt2);
1846 cleancache_init_fs(sb);
1850 int ext4_alloc_flex_bg_array(struct super_block *sb, ext4_group_t ngroup)
1852 struct ext4_sb_info *sbi = EXT4_SB(sb);
1853 struct flex_groups *new_groups;
1856 if (!sbi->s_log_groups_per_flex)
1859 size = ext4_flex_group(sbi, ngroup - 1) + 1;
1860 if (size <= sbi->s_flex_groups_allocated)
1863 size = roundup_pow_of_two(size * sizeof(struct flex_groups));
1864 new_groups = ext4_kvzalloc(size, GFP_KERNEL);
1866 ext4_msg(sb, KERN_ERR, "not enough memory for %d flex groups",
1867 size / (int) sizeof(struct flex_groups));
1871 if (sbi->s_flex_groups) {
1872 memcpy(new_groups, sbi->s_flex_groups,
1873 (sbi->s_flex_groups_allocated *
1874 sizeof(struct flex_groups)));
1875 ext4_kvfree(sbi->s_flex_groups);
1877 sbi->s_flex_groups = new_groups;
1878 sbi->s_flex_groups_allocated = size / sizeof(struct flex_groups);
1882 static int ext4_fill_flex_info(struct super_block *sb)
1884 struct ext4_sb_info *sbi = EXT4_SB(sb);
1885 struct ext4_group_desc *gdp = NULL;
1886 ext4_group_t flex_group;
1887 unsigned int groups_per_flex = 0;
1890 sbi->s_log_groups_per_flex = sbi->s_es->s_log_groups_per_flex;
1891 if (sbi->s_log_groups_per_flex < 1 || sbi->s_log_groups_per_flex > 31) {
1892 sbi->s_log_groups_per_flex = 0;
1895 groups_per_flex = 1U << sbi->s_log_groups_per_flex;
1897 err = ext4_alloc_flex_bg_array(sb, sbi->s_groups_count);
1901 for (i = 0; i < sbi->s_groups_count; i++) {
1902 gdp = ext4_get_group_desc(sb, i, NULL);
1904 flex_group = ext4_flex_group(sbi, i);
1905 atomic_add(ext4_free_inodes_count(sb, gdp),
1906 &sbi->s_flex_groups[flex_group].free_inodes);
1907 atomic_add(ext4_free_group_clusters(sb, gdp),
1908 &sbi->s_flex_groups[flex_group].free_clusters);
1909 atomic_add(ext4_used_dirs_count(sb, gdp),
1910 &sbi->s_flex_groups[flex_group].used_dirs);
1918 static __le16 ext4_group_desc_csum(struct ext4_sb_info *sbi, __u32 block_group,
1919 struct ext4_group_desc *gdp)
1923 __le32 le_group = cpu_to_le32(block_group);
1925 if ((sbi->s_es->s_feature_ro_compat &
1926 cpu_to_le32(EXT4_FEATURE_RO_COMPAT_METADATA_CSUM))) {
1927 /* Use new metadata_csum algorithm */
1931 old_csum = gdp->bg_checksum;
1932 gdp->bg_checksum = 0;
1933 csum32 = ext4_chksum(sbi, sbi->s_csum_seed, (__u8 *)&le_group,
1935 csum32 = ext4_chksum(sbi, csum32, (__u8 *)gdp,
1937 gdp->bg_checksum = old_csum;
1939 crc = csum32 & 0xFFFF;
1943 /* old crc16 code */
1944 offset = offsetof(struct ext4_group_desc, bg_checksum);
1946 crc = crc16(~0, sbi->s_es->s_uuid, sizeof(sbi->s_es->s_uuid));
1947 crc = crc16(crc, (__u8 *)&le_group, sizeof(le_group));
1948 crc = crc16(crc, (__u8 *)gdp, offset);
1949 offset += sizeof(gdp->bg_checksum); /* skip checksum */
1950 /* for checksum of struct ext4_group_desc do the rest...*/
1951 if ((sbi->s_es->s_feature_incompat &
1952 cpu_to_le32(EXT4_FEATURE_INCOMPAT_64BIT)) &&
1953 offset < le16_to_cpu(sbi->s_es->s_desc_size))
1954 crc = crc16(crc, (__u8 *)gdp + offset,
1955 le16_to_cpu(sbi->s_es->s_desc_size) -
1959 return cpu_to_le16(crc);
1962 int ext4_group_desc_csum_verify(struct super_block *sb, __u32 block_group,
1963 struct ext4_group_desc *gdp)
1965 if (ext4_has_group_desc_csum(sb) &&
1966 (gdp->bg_checksum != ext4_group_desc_csum(EXT4_SB(sb),
1973 void ext4_group_desc_csum_set(struct super_block *sb, __u32 block_group,
1974 struct ext4_group_desc *gdp)
1976 if (!ext4_has_group_desc_csum(sb))
1978 gdp->bg_checksum = ext4_group_desc_csum(EXT4_SB(sb), block_group, gdp);
1981 /* Called at mount-time, super-block is locked */
1982 static int ext4_check_descriptors(struct super_block *sb,
1983 ext4_group_t *first_not_zeroed)
1985 struct ext4_sb_info *sbi = EXT4_SB(sb);
1986 ext4_fsblk_t first_block = le32_to_cpu(sbi->s_es->s_first_data_block);
1987 ext4_fsblk_t last_block;
1988 ext4_fsblk_t block_bitmap;
1989 ext4_fsblk_t inode_bitmap;
1990 ext4_fsblk_t inode_table;
1991 int flexbg_flag = 0;
1992 ext4_group_t i, grp = sbi->s_groups_count;
1994 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_FLEX_BG))
1997 ext4_debug("Checking group descriptors");
1999 for (i = 0; i < sbi->s_groups_count; i++) {
2000 struct ext4_group_desc *gdp = ext4_get_group_desc(sb, i, NULL);
2002 if (i == sbi->s_groups_count - 1 || flexbg_flag)
2003 last_block = ext4_blocks_count(sbi->s_es) - 1;
2005 last_block = first_block +
2006 (EXT4_BLOCKS_PER_GROUP(sb) - 1);
2008 if ((grp == sbi->s_groups_count) &&
2009 !(gdp->bg_flags & cpu_to_le16(EXT4_BG_INODE_ZEROED)))
2012 block_bitmap = ext4_block_bitmap(sb, gdp);
2013 if (block_bitmap < first_block || block_bitmap > last_block) {
2014 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2015 "Block bitmap for group %u not in group "
2016 "(block %llu)!", i, block_bitmap);
2019 inode_bitmap = ext4_inode_bitmap(sb, gdp);
2020 if (inode_bitmap < first_block || inode_bitmap > last_block) {
2021 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2022 "Inode bitmap for group %u not in group "
2023 "(block %llu)!", i, inode_bitmap);
2026 inode_table = ext4_inode_table(sb, gdp);
2027 if (inode_table < first_block ||
2028 inode_table + sbi->s_itb_per_group - 1 > last_block) {
2029 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2030 "Inode table for group %u not in group "
2031 "(block %llu)!", i, inode_table);
2034 ext4_lock_group(sb, i);
2035 if (!ext4_group_desc_csum_verify(sb, i, gdp)) {
2036 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2037 "Checksum for group %u failed (%u!=%u)",
2038 i, le16_to_cpu(ext4_group_desc_csum(sbi, i,
2039 gdp)), le16_to_cpu(gdp->bg_checksum));
2040 if (!(sb->s_flags & MS_RDONLY)) {
2041 ext4_unlock_group(sb, i);
2045 ext4_unlock_group(sb, i);
2047 first_block += EXT4_BLOCKS_PER_GROUP(sb);
2049 if (NULL != first_not_zeroed)
2050 *first_not_zeroed = grp;
2052 ext4_free_blocks_count_set(sbi->s_es,
2053 EXT4_C2B(sbi, ext4_count_free_clusters(sb)));
2054 sbi->s_es->s_free_inodes_count =cpu_to_le32(ext4_count_free_inodes(sb));
2058 /* ext4_orphan_cleanup() walks a singly-linked list of inodes (starting at
2059 * the superblock) which were deleted from all directories, but held open by
2060 * a process at the time of a crash. We walk the list and try to delete these
2061 * inodes at recovery time (only with a read-write filesystem).
2063 * In order to keep the orphan inode chain consistent during traversal (in
2064 * case of crash during recovery), we link each inode into the superblock
2065 * orphan list_head and handle it the same way as an inode deletion during
2066 * normal operation (which journals the operations for us).
2068 * We only do an iget() and an iput() on each inode, which is very safe if we
2069 * accidentally point at an in-use or already deleted inode. The worst that
2070 * can happen in this case is that we get a "bit already cleared" message from
2071 * ext4_free_inode(). The only reason we would point at a wrong inode is if
2072 * e2fsck was run on this filesystem, and it must have already done the orphan
2073 * inode cleanup for us, so we can safely abort without any further action.
2075 static void ext4_orphan_cleanup(struct super_block *sb,
2076 struct ext4_super_block *es)
2078 unsigned int s_flags = sb->s_flags;
2079 int nr_orphans = 0, nr_truncates = 0;
2083 if (!es->s_last_orphan) {
2084 jbd_debug(4, "no orphan inodes to clean up\n");
2088 if (bdev_read_only(sb->s_bdev)) {
2089 ext4_msg(sb, KERN_ERR, "write access "
2090 "unavailable, skipping orphan cleanup");
2094 /* Check if feature set would not allow a r/w mount */
2095 if (!ext4_feature_set_ok(sb, 0)) {
2096 ext4_msg(sb, KERN_INFO, "Skipping orphan cleanup due to "
2097 "unknown ROCOMPAT features");
2101 if (EXT4_SB(sb)->s_mount_state & EXT4_ERROR_FS) {
2102 /* don't clear list on RO mount w/ errors */
2103 if (es->s_last_orphan && !(s_flags & MS_RDONLY)) {
2104 jbd_debug(1, "Errors on filesystem, "
2105 "clearing orphan list.\n");
2106 es->s_last_orphan = 0;
2108 jbd_debug(1, "Skipping orphan recovery on fs with errors.\n");
2112 if (s_flags & MS_RDONLY) {
2113 ext4_msg(sb, KERN_INFO, "orphan cleanup on readonly fs");
2114 sb->s_flags &= ~MS_RDONLY;
2117 /* Needed for iput() to work correctly and not trash data */
2118 sb->s_flags |= MS_ACTIVE;
2119 /* Turn on quotas so that they are updated correctly */
2120 for (i = 0; i < MAXQUOTAS; i++) {
2121 if (EXT4_SB(sb)->s_qf_names[i]) {
2122 int ret = ext4_quota_on_mount(sb, i);
2124 ext4_msg(sb, KERN_ERR,
2125 "Cannot turn on journaled "
2126 "quota: error %d", ret);
2131 while (es->s_last_orphan) {
2132 struct inode *inode;
2134 inode = ext4_orphan_get(sb, le32_to_cpu(es->s_last_orphan));
2135 if (IS_ERR(inode)) {
2136 es->s_last_orphan = 0;
2140 list_add(&EXT4_I(inode)->i_orphan, &EXT4_SB(sb)->s_orphan);
2141 dquot_initialize(inode);
2142 if (inode->i_nlink) {
2143 ext4_msg(sb, KERN_DEBUG,
2144 "%s: truncating inode %lu to %lld bytes",
2145 __func__, inode->i_ino, inode->i_size);
2146 jbd_debug(2, "truncating inode %lu to %lld bytes\n",
2147 inode->i_ino, inode->i_size);
2148 mutex_lock(&inode->i_mutex);
2149 ext4_truncate(inode);
2150 mutex_unlock(&inode->i_mutex);
2153 ext4_msg(sb, KERN_DEBUG,
2154 "%s: deleting unreferenced inode %lu",
2155 __func__, inode->i_ino);
2156 jbd_debug(2, "deleting unreferenced inode %lu\n",
2160 iput(inode); /* The delete magic happens here! */
2163 #define PLURAL(x) (x), ((x) == 1) ? "" : "s"
2166 ext4_msg(sb, KERN_INFO, "%d orphan inode%s deleted",
2167 PLURAL(nr_orphans));
2169 ext4_msg(sb, KERN_INFO, "%d truncate%s cleaned up",
2170 PLURAL(nr_truncates));
2172 /* Turn quotas off */
2173 for (i = 0; i < MAXQUOTAS; i++) {
2174 if (sb_dqopt(sb)->files[i])
2175 dquot_quota_off(sb, i);
2178 sb->s_flags = s_flags; /* Restore MS_RDONLY status */
2182 * Maximal extent format file size.
2183 * Resulting logical blkno at s_maxbytes must fit in our on-disk
2184 * extent format containers, within a sector_t, and within i_blocks
2185 * in the vfs. ext4 inode has 48 bits of i_block in fsblock units,
2186 * so that won't be a limiting factor.
2188 * However there is other limiting factor. We do store extents in the form
2189 * of starting block and length, hence the resulting length of the extent
2190 * covering maximum file size must fit into on-disk format containers as
2191 * well. Given that length is always by 1 unit bigger than max unit (because
2192 * we count 0 as well) we have to lower the s_maxbytes by one fs block.
2194 * Note, this does *not* consider any metadata overhead for vfs i_blocks.
2196 static loff_t ext4_max_size(int blkbits, int has_huge_files)
2199 loff_t upper_limit = MAX_LFS_FILESIZE;
2201 /* small i_blocks in vfs inode? */
2202 if (!has_huge_files || sizeof(blkcnt_t) < sizeof(u64)) {
2204 * CONFIG_LBDAF is not enabled implies the inode
2205 * i_block represent total blocks in 512 bytes
2206 * 32 == size of vfs inode i_blocks * 8
2208 upper_limit = (1LL << 32) - 1;
2210 /* total blocks in file system block size */
2211 upper_limit >>= (blkbits - 9);
2212 upper_limit <<= blkbits;
2216 * 32-bit extent-start container, ee_block. We lower the maxbytes
2217 * by one fs block, so ee_len can cover the extent of maximum file
2220 res = (1LL << 32) - 1;
2223 /* Sanity check against vm- & vfs- imposed limits */
2224 if (res > upper_limit)
2231 * Maximal bitmap file size. There is a direct, and {,double-,triple-}indirect
2232 * block limit, and also a limit of (2^48 - 1) 512-byte sectors in i_blocks.
2233 * We need to be 1 filesystem block less than the 2^48 sector limit.
2235 static loff_t ext4_max_bitmap_size(int bits, int has_huge_files)
2237 loff_t res = EXT4_NDIR_BLOCKS;
2240 /* This is calculated to be the largest file size for a dense, block
2241 * mapped file such that the file's total number of 512-byte sectors,
2242 * including data and all indirect blocks, does not exceed (2^48 - 1).
2244 * __u32 i_blocks_lo and _u16 i_blocks_high represent the total
2245 * number of 512-byte sectors of the file.
2248 if (!has_huge_files || sizeof(blkcnt_t) < sizeof(u64)) {
2250 * !has_huge_files or CONFIG_LBDAF not enabled implies that
2251 * the inode i_block field represents total file blocks in
2252 * 2^32 512-byte sectors == size of vfs inode i_blocks * 8
2254 upper_limit = (1LL << 32) - 1;
2256 /* total blocks in file system block size */
2257 upper_limit >>= (bits - 9);
2261 * We use 48 bit ext4_inode i_blocks
2262 * With EXT4_HUGE_FILE_FL set the i_blocks
2263 * represent total number of blocks in
2264 * file system block size
2266 upper_limit = (1LL << 48) - 1;
2270 /* indirect blocks */
2272 /* double indirect blocks */
2273 meta_blocks += 1 + (1LL << (bits-2));
2274 /* tripple indirect blocks */
2275 meta_blocks += 1 + (1LL << (bits-2)) + (1LL << (2*(bits-2)));
2277 upper_limit -= meta_blocks;
2278 upper_limit <<= bits;
2280 res += 1LL << (bits-2);
2281 res += 1LL << (2*(bits-2));
2282 res += 1LL << (3*(bits-2));
2284 if (res > upper_limit)
2287 if (res > MAX_LFS_FILESIZE)
2288 res = MAX_LFS_FILESIZE;
2293 static ext4_fsblk_t descriptor_loc(struct super_block *sb,
2294 ext4_fsblk_t logical_sb_block, int nr)
2296 struct ext4_sb_info *sbi = EXT4_SB(sb);
2297 ext4_group_t bg, first_meta_bg;
2300 first_meta_bg = le32_to_cpu(sbi->s_es->s_first_meta_bg);
2302 if (!EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_META_BG) ||
2304 return logical_sb_block + nr + 1;
2305 bg = sbi->s_desc_per_block * nr;
2306 if (ext4_bg_has_super(sb, bg))
2309 return (has_super + ext4_group_first_block_no(sb, bg));
2313 * ext4_get_stripe_size: Get the stripe size.
2314 * @sbi: In memory super block info
2316 * If we have specified it via mount option, then
2317 * use the mount option value. If the value specified at mount time is
2318 * greater than the blocks per group use the super block value.
2319 * If the super block value is greater than blocks per group return 0.
2320 * Allocator needs it be less than blocks per group.
2323 static unsigned long ext4_get_stripe_size(struct ext4_sb_info *sbi)
2325 unsigned long stride = le16_to_cpu(sbi->s_es->s_raid_stride);
2326 unsigned long stripe_width =
2327 le32_to_cpu(sbi->s_es->s_raid_stripe_width);
2330 if (sbi->s_stripe && sbi->s_stripe <= sbi->s_blocks_per_group)
2331 ret = sbi->s_stripe;
2332 else if (stripe_width <= sbi->s_blocks_per_group)
2334 else if (stride <= sbi->s_blocks_per_group)
2340 * If the stripe width is 1, this makes no sense and
2341 * we set it to 0 to turn off stripe handling code.
2352 struct attribute attr;
2353 ssize_t (*show)(struct ext4_attr *, struct ext4_sb_info *, char *);
2354 ssize_t (*store)(struct ext4_attr *, struct ext4_sb_info *,
2355 const char *, size_t);
2359 static int parse_strtoul(const char *buf,
2360 unsigned long max, unsigned long *value)
2364 *value = simple_strtoul(skip_spaces(buf), &endp, 0);
2365 endp = skip_spaces(endp);
2366 if (*endp || *value > max)
2372 static ssize_t delayed_allocation_blocks_show(struct ext4_attr *a,
2373 struct ext4_sb_info *sbi,
2376 return snprintf(buf, PAGE_SIZE, "%llu\n",
2378 percpu_counter_sum(&sbi->s_dirtyclusters_counter)));
2381 static ssize_t session_write_kbytes_show(struct ext4_attr *a,
2382 struct ext4_sb_info *sbi, char *buf)
2384 struct super_block *sb = sbi->s_buddy_cache->i_sb;
2386 if (!sb->s_bdev->bd_part)
2387 return snprintf(buf, PAGE_SIZE, "0\n");
2388 return snprintf(buf, PAGE_SIZE, "%lu\n",
2389 (part_stat_read(sb->s_bdev->bd_part, sectors[1]) -
2390 sbi->s_sectors_written_start) >> 1);
2393 static ssize_t lifetime_write_kbytes_show(struct ext4_attr *a,
2394 struct ext4_sb_info *sbi, char *buf)
2396 struct super_block *sb = sbi->s_buddy_cache->i_sb;
2398 if (!sb->s_bdev->bd_part)
2399 return snprintf(buf, PAGE_SIZE, "0\n");
2400 return snprintf(buf, PAGE_SIZE, "%llu\n",
2401 (unsigned long long)(sbi->s_kbytes_written +
2402 ((part_stat_read(sb->s_bdev->bd_part, sectors[1]) -
2403 EXT4_SB(sb)->s_sectors_written_start) >> 1)));
2406 static ssize_t inode_readahead_blks_store(struct ext4_attr *a,
2407 struct ext4_sb_info *sbi,
2408 const char *buf, size_t count)
2412 if (parse_strtoul(buf, 0x40000000, &t))
2415 if (t && !is_power_of_2(t))
2418 sbi->s_inode_readahead_blks = t;
2422 static ssize_t sbi_ui_show(struct ext4_attr *a,
2423 struct ext4_sb_info *sbi, char *buf)
2425 unsigned int *ui = (unsigned int *) (((char *) sbi) + a->offset);
2427 return snprintf(buf, PAGE_SIZE, "%u\n", *ui);
2430 static ssize_t sbi_ui_store(struct ext4_attr *a,
2431 struct ext4_sb_info *sbi,
2432 const char *buf, size_t count)
2434 unsigned int *ui = (unsigned int *) (((char *) sbi) + a->offset);
2437 if (parse_strtoul(buf, 0xffffffff, &t))
2443 static ssize_t trigger_test_error(struct ext4_attr *a,
2444 struct ext4_sb_info *sbi,
2445 const char *buf, size_t count)
2449 if (!capable(CAP_SYS_ADMIN))
2452 if (len && buf[len-1] == '\n')
2456 ext4_error(sbi->s_sb, "%.*s", len, buf);
2460 #define EXT4_ATTR_OFFSET(_name,_mode,_show,_store,_elname) \
2461 static struct ext4_attr ext4_attr_##_name = { \
2462 .attr = {.name = __stringify(_name), .mode = _mode }, \
2465 .offset = offsetof(struct ext4_sb_info, _elname), \
2467 #define EXT4_ATTR(name, mode, show, store) \
2468 static struct ext4_attr ext4_attr_##name = __ATTR(name, mode, show, store)
2470 #define EXT4_INFO_ATTR(name) EXT4_ATTR(name, 0444, NULL, NULL)
2471 #define EXT4_RO_ATTR(name) EXT4_ATTR(name, 0444, name##_show, NULL)
2472 #define EXT4_RW_ATTR(name) EXT4_ATTR(name, 0644, name##_show, name##_store)
2473 #define EXT4_RW_ATTR_SBI_UI(name, elname) \
2474 EXT4_ATTR_OFFSET(name, 0644, sbi_ui_show, sbi_ui_store, elname)
2475 #define ATTR_LIST(name) &ext4_attr_##name.attr
2477 EXT4_RO_ATTR(delayed_allocation_blocks);
2478 EXT4_RO_ATTR(session_write_kbytes);
2479 EXT4_RO_ATTR(lifetime_write_kbytes);
2480 EXT4_ATTR_OFFSET(inode_readahead_blks, 0644, sbi_ui_show,
2481 inode_readahead_blks_store, s_inode_readahead_blks);
2482 EXT4_RW_ATTR_SBI_UI(inode_goal, s_inode_goal);
2483 EXT4_RW_ATTR_SBI_UI(mb_stats, s_mb_stats);
2484 EXT4_RW_ATTR_SBI_UI(mb_max_to_scan, s_mb_max_to_scan);
2485 EXT4_RW_ATTR_SBI_UI(mb_min_to_scan, s_mb_min_to_scan);
2486 EXT4_RW_ATTR_SBI_UI(mb_order2_req, s_mb_order2_reqs);
2487 EXT4_RW_ATTR_SBI_UI(mb_stream_req, s_mb_stream_request);
2488 EXT4_RW_ATTR_SBI_UI(mb_group_prealloc, s_mb_group_prealloc);
2489 EXT4_RW_ATTR_SBI_UI(max_writeback_mb_bump, s_max_writeback_mb_bump);
2490 EXT4_RW_ATTR_SBI_UI(extent_max_zeroout_kb, s_extent_max_zeroout_kb);
2491 EXT4_ATTR(trigger_fs_error, 0200, NULL, trigger_test_error);
2493 static struct attribute *ext4_attrs[] = {
2494 ATTR_LIST(delayed_allocation_blocks),
2495 ATTR_LIST(session_write_kbytes),
2496 ATTR_LIST(lifetime_write_kbytes),
2497 ATTR_LIST(inode_readahead_blks),
2498 ATTR_LIST(inode_goal),
2499 ATTR_LIST(mb_stats),
2500 ATTR_LIST(mb_max_to_scan),
2501 ATTR_LIST(mb_min_to_scan),
2502 ATTR_LIST(mb_order2_req),
2503 ATTR_LIST(mb_stream_req),
2504 ATTR_LIST(mb_group_prealloc),
2505 ATTR_LIST(max_writeback_mb_bump),
2506 ATTR_LIST(extent_max_zeroout_kb),
2507 ATTR_LIST(trigger_fs_error),
2511 /* Features this copy of ext4 supports */
2512 EXT4_INFO_ATTR(lazy_itable_init);
2513 EXT4_INFO_ATTR(batched_discard);
2514 EXT4_INFO_ATTR(meta_bg_resize);
2516 static struct attribute *ext4_feat_attrs[] = {
2517 ATTR_LIST(lazy_itable_init),
2518 ATTR_LIST(batched_discard),
2519 ATTR_LIST(meta_bg_resize),
2523 static ssize_t ext4_attr_show(struct kobject *kobj,
2524 struct attribute *attr, char *buf)
2526 struct ext4_sb_info *sbi = container_of(kobj, struct ext4_sb_info,
2528 struct ext4_attr *a = container_of(attr, struct ext4_attr, attr);
2530 return a->show ? a->show(a, sbi, buf) : 0;
2533 static ssize_t ext4_attr_store(struct kobject *kobj,
2534 struct attribute *attr,
2535 const char *buf, size_t len)
2537 struct ext4_sb_info *sbi = container_of(kobj, struct ext4_sb_info,
2539 struct ext4_attr *a = container_of(attr, struct ext4_attr, attr);
2541 return a->store ? a->store(a, sbi, buf, len) : 0;
2544 static void ext4_sb_release(struct kobject *kobj)
2546 struct ext4_sb_info *sbi = container_of(kobj, struct ext4_sb_info,
2548 complete(&sbi->s_kobj_unregister);
2551 static const struct sysfs_ops ext4_attr_ops = {
2552 .show = ext4_attr_show,
2553 .store = ext4_attr_store,
2556 static struct kobj_type ext4_ktype = {
2557 .default_attrs = ext4_attrs,
2558 .sysfs_ops = &ext4_attr_ops,
2559 .release = ext4_sb_release,
2562 static void ext4_feat_release(struct kobject *kobj)
2564 complete(&ext4_feat->f_kobj_unregister);
2567 static struct kobj_type ext4_feat_ktype = {
2568 .default_attrs = ext4_feat_attrs,
2569 .sysfs_ops = &ext4_attr_ops,
2570 .release = ext4_feat_release,
2574 * Check whether this filesystem can be mounted based on
2575 * the features present and the RDONLY/RDWR mount requested.
2576 * Returns 1 if this filesystem can be mounted as requested,
2577 * 0 if it cannot be.
2579 static int ext4_feature_set_ok(struct super_block *sb, int readonly)
2581 if (EXT4_HAS_INCOMPAT_FEATURE(sb, ~EXT4_FEATURE_INCOMPAT_SUPP)) {
2582 ext4_msg(sb, KERN_ERR,
2583 "Couldn't mount because of "
2584 "unsupported optional features (%x)",
2585 (le32_to_cpu(EXT4_SB(sb)->s_es->s_feature_incompat) &
2586 ~EXT4_FEATURE_INCOMPAT_SUPP));
2593 /* Check that feature set is OK for a read-write mount */
2594 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, ~EXT4_FEATURE_RO_COMPAT_SUPP)) {
2595 ext4_msg(sb, KERN_ERR, "couldn't mount RDWR because of "
2596 "unsupported optional features (%x)",
2597 (le32_to_cpu(EXT4_SB(sb)->s_es->s_feature_ro_compat) &
2598 ~EXT4_FEATURE_RO_COMPAT_SUPP));
2602 * Large file size enabled file system can only be mounted
2603 * read-write on 32-bit systems if kernel is built with CONFIG_LBDAF
2605 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_HUGE_FILE)) {
2606 if (sizeof(blkcnt_t) < sizeof(u64)) {
2607 ext4_msg(sb, KERN_ERR, "Filesystem with huge files "
2608 "cannot be mounted RDWR without "
2613 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_BIGALLOC) &&
2614 !EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_EXTENTS)) {
2615 ext4_msg(sb, KERN_ERR,
2616 "Can't support bigalloc feature without "
2617 "extents feature\n");
2621 #ifndef CONFIG_QUOTA
2622 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_QUOTA) &&
2624 ext4_msg(sb, KERN_ERR,
2625 "Filesystem with quota feature cannot be mounted RDWR "
2626 "without CONFIG_QUOTA");
2629 #endif /* CONFIG_QUOTA */
2634 * This function is called once a day if we have errors logged
2635 * on the file system
2637 static void print_daily_error_info(unsigned long arg)
2639 struct super_block *sb = (struct super_block *) arg;
2640 struct ext4_sb_info *sbi;
2641 struct ext4_super_block *es;
2646 if (es->s_error_count)
2647 ext4_msg(sb, KERN_NOTICE, "error count: %u",
2648 le32_to_cpu(es->s_error_count));
2649 if (es->s_first_error_time) {
2650 printk(KERN_NOTICE "EXT4-fs (%s): initial error at %u: %.*s:%d",
2651 sb->s_id, le32_to_cpu(es->s_first_error_time),
2652 (int) sizeof(es->s_first_error_func),
2653 es->s_first_error_func,
2654 le32_to_cpu(es->s_first_error_line));
2655 if (es->s_first_error_ino)
2656 printk(": inode %u",
2657 le32_to_cpu(es->s_first_error_ino));
2658 if (es->s_first_error_block)
2659 printk(": block %llu", (unsigned long long)
2660 le64_to_cpu(es->s_first_error_block));
2663 if (es->s_last_error_time) {
2664 printk(KERN_NOTICE "EXT4-fs (%s): last error at %u: %.*s:%d",
2665 sb->s_id, le32_to_cpu(es->s_last_error_time),
2666 (int) sizeof(es->s_last_error_func),
2667 es->s_last_error_func,
2668 le32_to_cpu(es->s_last_error_line));
2669 if (es->s_last_error_ino)
2670 printk(": inode %u",
2671 le32_to_cpu(es->s_last_error_ino));
2672 if (es->s_last_error_block)
2673 printk(": block %llu", (unsigned long long)
2674 le64_to_cpu(es->s_last_error_block));
2677 mod_timer(&sbi->s_err_report, jiffies + 24*60*60*HZ); /* Once a day */
2680 /* Find next suitable group and run ext4_init_inode_table */
2681 static int ext4_run_li_request(struct ext4_li_request *elr)
2683 struct ext4_group_desc *gdp = NULL;
2684 ext4_group_t group, ngroups;
2685 struct super_block *sb;
2686 unsigned long timeout = 0;
2690 ngroups = EXT4_SB(sb)->s_groups_count;
2693 for (group = elr->lr_next_group; group < ngroups; group++) {
2694 gdp = ext4_get_group_desc(sb, group, NULL);
2700 if (!(gdp->bg_flags & cpu_to_le16(EXT4_BG_INODE_ZEROED)))
2704 if (group >= ngroups)
2709 ret = ext4_init_inode_table(sb, group,
2710 elr->lr_timeout ? 0 : 1);
2711 if (elr->lr_timeout == 0) {
2712 timeout = (jiffies - timeout) *
2713 elr->lr_sbi->s_li_wait_mult;
2714 elr->lr_timeout = timeout;
2716 elr->lr_next_sched = jiffies + elr->lr_timeout;
2717 elr->lr_next_group = group + 1;
2725 * Remove lr_request from the list_request and free the
2726 * request structure. Should be called with li_list_mtx held
2728 static void ext4_remove_li_request(struct ext4_li_request *elr)
2730 struct ext4_sb_info *sbi;
2737 list_del(&elr->lr_request);
2738 sbi->s_li_request = NULL;
2742 static void ext4_unregister_li_request(struct super_block *sb)
2744 mutex_lock(&ext4_li_mtx);
2745 if (!ext4_li_info) {
2746 mutex_unlock(&ext4_li_mtx);
2750 mutex_lock(&ext4_li_info->li_list_mtx);
2751 ext4_remove_li_request(EXT4_SB(sb)->s_li_request);
2752 mutex_unlock(&ext4_li_info->li_list_mtx);
2753 mutex_unlock(&ext4_li_mtx);
2756 static struct task_struct *ext4_lazyinit_task;
2759 * This is the function where ext4lazyinit thread lives. It walks
2760 * through the request list searching for next scheduled filesystem.
2761 * When such a fs is found, run the lazy initialization request
2762 * (ext4_rn_li_request) and keep track of the time spend in this
2763 * function. Based on that time we compute next schedule time of
2764 * the request. When walking through the list is complete, compute
2765 * next waking time and put itself into sleep.
2767 static int ext4_lazyinit_thread(void *arg)
2769 struct ext4_lazy_init *eli = (struct ext4_lazy_init *)arg;
2770 struct list_head *pos, *n;
2771 struct ext4_li_request *elr;
2772 unsigned long next_wakeup, cur;
2774 BUG_ON(NULL == eli);
2778 next_wakeup = MAX_JIFFY_OFFSET;
2780 mutex_lock(&eli->li_list_mtx);
2781 if (list_empty(&eli->li_request_list)) {
2782 mutex_unlock(&eli->li_list_mtx);
2786 list_for_each_safe(pos, n, &eli->li_request_list) {
2787 elr = list_entry(pos, struct ext4_li_request,
2790 if (time_after_eq(jiffies, elr->lr_next_sched)) {
2791 if (ext4_run_li_request(elr) != 0) {
2792 /* error, remove the lazy_init job */
2793 ext4_remove_li_request(elr);
2798 if (time_before(elr->lr_next_sched, next_wakeup))
2799 next_wakeup = elr->lr_next_sched;
2801 mutex_unlock(&eli->li_list_mtx);
2806 if ((time_after_eq(cur, next_wakeup)) ||
2807 (MAX_JIFFY_OFFSET == next_wakeup)) {
2812 schedule_timeout_interruptible(next_wakeup - cur);
2814 if (kthread_should_stop()) {
2815 ext4_clear_request_list();
2822 * It looks like the request list is empty, but we need
2823 * to check it under the li_list_mtx lock, to prevent any
2824 * additions into it, and of course we should lock ext4_li_mtx
2825 * to atomically free the list and ext4_li_info, because at
2826 * this point another ext4 filesystem could be registering
2829 mutex_lock(&ext4_li_mtx);
2830 mutex_lock(&eli->li_list_mtx);
2831 if (!list_empty(&eli->li_request_list)) {
2832 mutex_unlock(&eli->li_list_mtx);
2833 mutex_unlock(&ext4_li_mtx);
2836 mutex_unlock(&eli->li_list_mtx);
2837 kfree(ext4_li_info);
2838 ext4_li_info = NULL;
2839 mutex_unlock(&ext4_li_mtx);
2844 static void ext4_clear_request_list(void)
2846 struct list_head *pos, *n;
2847 struct ext4_li_request *elr;
2849 mutex_lock(&ext4_li_info->li_list_mtx);
2850 list_for_each_safe(pos, n, &ext4_li_info->li_request_list) {
2851 elr = list_entry(pos, struct ext4_li_request,
2853 ext4_remove_li_request(elr);
2855 mutex_unlock(&ext4_li_info->li_list_mtx);
2858 static int ext4_run_lazyinit_thread(void)
2860 ext4_lazyinit_task = kthread_run(ext4_lazyinit_thread,
2861 ext4_li_info, "ext4lazyinit");
2862 if (IS_ERR(ext4_lazyinit_task)) {
2863 int err = PTR_ERR(ext4_lazyinit_task);
2864 ext4_clear_request_list();
2865 kfree(ext4_li_info);
2866 ext4_li_info = NULL;
2867 printk(KERN_CRIT "EXT4-fs: error %d creating inode table "
2868 "initialization thread\n",
2872 ext4_li_info->li_state |= EXT4_LAZYINIT_RUNNING;
2877 * Check whether it make sense to run itable init. thread or not.
2878 * If there is at least one uninitialized inode table, return
2879 * corresponding group number, else the loop goes through all
2880 * groups and return total number of groups.
2882 static ext4_group_t ext4_has_uninit_itable(struct super_block *sb)
2884 ext4_group_t group, ngroups = EXT4_SB(sb)->s_groups_count;
2885 struct ext4_group_desc *gdp = NULL;
2887 for (group = 0; group < ngroups; group++) {
2888 gdp = ext4_get_group_desc(sb, group, NULL);
2892 if (!(gdp->bg_flags & cpu_to_le16(EXT4_BG_INODE_ZEROED)))
2899 static int ext4_li_info_new(void)
2901 struct ext4_lazy_init *eli = NULL;
2903 eli = kzalloc(sizeof(*eli), GFP_KERNEL);
2907 INIT_LIST_HEAD(&eli->li_request_list);
2908 mutex_init(&eli->li_list_mtx);
2910 eli->li_state |= EXT4_LAZYINIT_QUIT;
2917 static struct ext4_li_request *ext4_li_request_new(struct super_block *sb,
2920 struct ext4_sb_info *sbi = EXT4_SB(sb);
2921 struct ext4_li_request *elr;
2924 elr = kzalloc(sizeof(*elr), GFP_KERNEL);
2930 elr->lr_next_group = start;
2933 * Randomize first schedule time of the request to
2934 * spread the inode table initialization requests
2937 get_random_bytes(&rnd, sizeof(rnd));
2938 elr->lr_next_sched = jiffies + (unsigned long)rnd %
2939 (EXT4_DEF_LI_MAX_START_DELAY * HZ);
2944 int ext4_register_li_request(struct super_block *sb,
2945 ext4_group_t first_not_zeroed)
2947 struct ext4_sb_info *sbi = EXT4_SB(sb);
2948 struct ext4_li_request *elr = NULL;
2949 ext4_group_t ngroups = EXT4_SB(sb)->s_groups_count;
2952 mutex_lock(&ext4_li_mtx);
2953 if (sbi->s_li_request != NULL) {
2955 * Reset timeout so it can be computed again, because
2956 * s_li_wait_mult might have changed.
2958 sbi->s_li_request->lr_timeout = 0;
2962 if (first_not_zeroed == ngroups ||
2963 (sb->s_flags & MS_RDONLY) ||
2964 !test_opt(sb, INIT_INODE_TABLE))
2967 elr = ext4_li_request_new(sb, first_not_zeroed);
2973 if (NULL == ext4_li_info) {
2974 ret = ext4_li_info_new();
2979 mutex_lock(&ext4_li_info->li_list_mtx);
2980 list_add(&elr->lr_request, &ext4_li_info->li_request_list);
2981 mutex_unlock(&ext4_li_info->li_list_mtx);
2983 sbi->s_li_request = elr;
2985 * set elr to NULL here since it has been inserted to
2986 * the request_list and the removal and free of it is
2987 * handled by ext4_clear_request_list from now on.
2991 if (!(ext4_li_info->li_state & EXT4_LAZYINIT_RUNNING)) {
2992 ret = ext4_run_lazyinit_thread();
2997 mutex_unlock(&ext4_li_mtx);
3004 * We do not need to lock anything since this is called on
3007 static void ext4_destroy_lazyinit_thread(void)
3010 * If thread exited earlier
3011 * there's nothing to be done.
3013 if (!ext4_li_info || !ext4_lazyinit_task)
3016 kthread_stop(ext4_lazyinit_task);
3019 static int set_journal_csum_feature_set(struct super_block *sb)
3022 int compat, incompat;
3023 struct ext4_sb_info *sbi = EXT4_SB(sb);
3025 if (EXT4_HAS_RO_COMPAT_FEATURE(sb,
3026 EXT4_FEATURE_RO_COMPAT_METADATA_CSUM)) {
3027 /* journal checksum v2 */
3029 incompat = JBD2_FEATURE_INCOMPAT_CSUM_V2;
3031 /* journal checksum v1 */
3032 compat = JBD2_FEATURE_COMPAT_CHECKSUM;
3036 if (test_opt(sb, JOURNAL_ASYNC_COMMIT)) {
3037 ret = jbd2_journal_set_features(sbi->s_journal,
3039 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT |
3041 } else if (test_opt(sb, JOURNAL_CHECKSUM)) {
3042 ret = jbd2_journal_set_features(sbi->s_journal,
3045 jbd2_journal_clear_features(sbi->s_journal, 0, 0,
3046 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT);
3048 jbd2_journal_clear_features(sbi->s_journal,
3049 JBD2_FEATURE_COMPAT_CHECKSUM, 0,
3050 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT |
3051 JBD2_FEATURE_INCOMPAT_CSUM_V2);
3058 * Note: calculating the overhead so we can be compatible with
3059 * historical BSD practice is quite difficult in the face of
3060 * clusters/bigalloc. This is because multiple metadata blocks from
3061 * different block group can end up in the same allocation cluster.
3062 * Calculating the exact overhead in the face of clustered allocation
3063 * requires either O(all block bitmaps) in memory or O(number of block
3064 * groups**2) in time. We will still calculate the superblock for
3065 * older file systems --- and if we come across with a bigalloc file
3066 * system with zero in s_overhead_clusters the estimate will be close to
3067 * correct especially for very large cluster sizes --- but for newer
3068 * file systems, it's better to calculate this figure once at mkfs
3069 * time, and store it in the superblock. If the superblock value is
3070 * present (even for non-bigalloc file systems), we will use it.
3072 static int count_overhead(struct super_block *sb, ext4_group_t grp,
3075 struct ext4_sb_info *sbi = EXT4_SB(sb);
3076 struct ext4_group_desc *gdp;
3077 ext4_fsblk_t first_block, last_block, b;
3078 ext4_group_t i, ngroups = ext4_get_groups_count(sb);
3079 int s, j, count = 0;
3081 if (!EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_BIGALLOC))
3082 return (ext4_bg_has_super(sb, grp) + ext4_bg_num_gdb(sb, grp) +
3083 sbi->s_itb_per_group + 2);
3085 first_block = le32_to_cpu(sbi->s_es->s_first_data_block) +
3086 (grp * EXT4_BLOCKS_PER_GROUP(sb));
3087 last_block = first_block + EXT4_BLOCKS_PER_GROUP(sb) - 1;
3088 for (i = 0; i < ngroups; i++) {
3089 gdp = ext4_get_group_desc(sb, i, NULL);
3090 b = ext4_block_bitmap(sb, gdp);
3091 if (b >= first_block && b <= last_block) {
3092 ext4_set_bit(EXT4_B2C(sbi, b - first_block), buf);
3095 b = ext4_inode_bitmap(sb, gdp);
3096 if (b >= first_block && b <= last_block) {
3097 ext4_set_bit(EXT4_B2C(sbi, b - first_block), buf);
3100 b = ext4_inode_table(sb, gdp);
3101 if (b >= first_block && b + sbi->s_itb_per_group <= last_block)
3102 for (j = 0; j < sbi->s_itb_per_group; j++, b++) {
3103 int c = EXT4_B2C(sbi, b - first_block);
3104 ext4_set_bit(c, buf);
3110 if (ext4_bg_has_super(sb, grp)) {
3111 ext4_set_bit(s++, buf);
3114 for (j = ext4_bg_num_gdb(sb, grp); j > 0; j--) {
3115 ext4_set_bit(EXT4_B2C(sbi, s++), buf);
3121 return EXT4_CLUSTERS_PER_GROUP(sb) -
3122 ext4_count_free(buf, EXT4_CLUSTERS_PER_GROUP(sb) / 8);
3126 * Compute the overhead and stash it in sbi->s_overhead
3128 int ext4_calculate_overhead(struct super_block *sb)
3130 struct ext4_sb_info *sbi = EXT4_SB(sb);
3131 struct ext4_super_block *es = sbi->s_es;
3132 ext4_group_t i, ngroups = ext4_get_groups_count(sb);
3133 ext4_fsblk_t overhead = 0;
3134 char *buf = (char *) get_zeroed_page(GFP_KERNEL);
3140 * Compute the overhead (FS structures). This is constant
3141 * for a given filesystem unless the number of block groups
3142 * changes so we cache the previous value until it does.
3146 * All of the blocks before first_data_block are overhead
3148 overhead = EXT4_B2C(sbi, le32_to_cpu(es->s_first_data_block));
3151 * Add the overhead found in each block group
3153 for (i = 0; i < ngroups; i++) {
3156 blks = count_overhead(sb, i, buf);
3159 memset(buf, 0, PAGE_SIZE);
3162 /* Add the journal blocks as well */
3164 overhead += EXT4_B2C(sbi, sbi->s_journal->j_maxlen);
3166 sbi->s_overhead = overhead;
3168 free_page((unsigned long) buf);
3172 static int ext4_fill_super(struct super_block *sb, void *data, int silent)
3174 char *orig_data = kstrdup(data, GFP_KERNEL);
3175 struct buffer_head *bh;
3176 struct ext4_super_block *es = NULL;
3177 struct ext4_sb_info *sbi;
3179 ext4_fsblk_t sb_block = get_sb_block(&data);
3180 ext4_fsblk_t logical_sb_block;
3181 unsigned long offset = 0;
3182 unsigned long journal_devnum = 0;
3183 unsigned long def_mount_opts;
3188 int blocksize, clustersize;
3189 unsigned int db_count;
3191 int needs_recovery, has_huge_files, has_bigalloc;
3194 unsigned int journal_ioprio = DEFAULT_JOURNAL_IOPRIO;
3195 ext4_group_t first_not_zeroed;
3197 sbi = kzalloc(sizeof(*sbi), GFP_KERNEL);
3201 sbi->s_blockgroup_lock =
3202 kzalloc(sizeof(struct blockgroup_lock), GFP_KERNEL);
3203 if (!sbi->s_blockgroup_lock) {
3207 sb->s_fs_info = sbi;
3209 sbi->s_inode_readahead_blks = EXT4_DEF_INODE_READAHEAD_BLKS;
3210 sbi->s_sb_block = sb_block;
3211 if (sb->s_bdev->bd_part)
3212 sbi->s_sectors_written_start =
3213 part_stat_read(sb->s_bdev->bd_part, sectors[1]);
3215 /* Cleanup superblock name */
3216 for (cp = sb->s_id; (cp = strchr(cp, '/'));)
3219 /* -EINVAL is default */
3221 blocksize = sb_min_blocksize(sb, EXT4_MIN_BLOCK_SIZE);
3223 ext4_msg(sb, KERN_ERR, "unable to set blocksize");
3228 * The ext4 superblock will not be buffer aligned for other than 1kB
3229 * block sizes. We need to calculate the offset from buffer start.
3231 if (blocksize != EXT4_MIN_BLOCK_SIZE) {
3232 logical_sb_block = sb_block * EXT4_MIN_BLOCK_SIZE;
3233 offset = do_div(logical_sb_block, blocksize);
3235 logical_sb_block = sb_block;
3238 if (!(bh = sb_bread(sb, logical_sb_block))) {
3239 ext4_msg(sb, KERN_ERR, "unable to read superblock");
3243 * Note: s_es must be initialized as soon as possible because
3244 * some ext4 macro-instructions depend on its value
3246 es = (struct ext4_super_block *) (bh->b_data + offset);
3248 sb->s_magic = le16_to_cpu(es->s_magic);
3249 if (sb->s_magic != EXT4_SUPER_MAGIC)
3251 sbi->s_kbytes_written = le64_to_cpu(es->s_kbytes_written);
3253 /* Warn if metadata_csum and gdt_csum are both set. */
3254 if (EXT4_HAS_RO_COMPAT_FEATURE(sb,
3255 EXT4_FEATURE_RO_COMPAT_METADATA_CSUM) &&
3256 EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_GDT_CSUM))
3257 ext4_warning(sb, KERN_INFO "metadata_csum and uninit_bg are "
3258 "redundant flags; please run fsck.");
3260 /* Check for a known checksum algorithm */
3261 if (!ext4_verify_csum_type(sb, es)) {
3262 ext4_msg(sb, KERN_ERR, "VFS: Found ext4 filesystem with "
3263 "unknown checksum algorithm.");
3268 /* Load the checksum driver */
3269 if (EXT4_HAS_RO_COMPAT_FEATURE(sb,
3270 EXT4_FEATURE_RO_COMPAT_METADATA_CSUM)) {
3271 sbi->s_chksum_driver = crypto_alloc_shash("crc32c", 0, 0);
3272 if (IS_ERR(sbi->s_chksum_driver)) {
3273 ext4_msg(sb, KERN_ERR, "Cannot load crc32c driver.");
3274 ret = PTR_ERR(sbi->s_chksum_driver);
3275 sbi->s_chksum_driver = NULL;
3280 /* Check superblock checksum */
3281 if (!ext4_superblock_csum_verify(sb, es)) {
3282 ext4_msg(sb, KERN_ERR, "VFS: Found ext4 filesystem with "
3283 "invalid superblock checksum. Run e2fsck?");
3288 /* Precompute checksum seed for all metadata */
3289 if (EXT4_HAS_RO_COMPAT_FEATURE(sb,
3290 EXT4_FEATURE_RO_COMPAT_METADATA_CSUM))
3291 sbi->s_csum_seed = ext4_chksum(sbi, ~0, es->s_uuid,
3292 sizeof(es->s_uuid));
3294 /* Set defaults before we parse the mount options */
3295 def_mount_opts = le32_to_cpu(es->s_default_mount_opts);
3296 set_opt(sb, INIT_INODE_TABLE);
3297 if (def_mount_opts & EXT4_DEFM_DEBUG)
3299 if (def_mount_opts & EXT4_DEFM_BSDGROUPS)
3301 if (def_mount_opts & EXT4_DEFM_UID16)
3302 set_opt(sb, NO_UID32);
3303 /* xattr user namespace & acls are now defaulted on */
3304 set_opt(sb, XATTR_USER);
3305 #ifdef CONFIG_EXT4_FS_POSIX_ACL
3306 set_opt(sb, POSIX_ACL);
3308 if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_DATA)
3309 set_opt(sb, JOURNAL_DATA);
3310 else if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_ORDERED)
3311 set_opt(sb, ORDERED_DATA);
3312 else if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_WBACK)
3313 set_opt(sb, WRITEBACK_DATA);
3315 if (le16_to_cpu(sbi->s_es->s_errors) == EXT4_ERRORS_PANIC)
3316 set_opt(sb, ERRORS_PANIC);
3317 else if (le16_to_cpu(sbi->s_es->s_errors) == EXT4_ERRORS_CONTINUE)
3318 set_opt(sb, ERRORS_CONT);
3320 set_opt(sb, ERRORS_RO);
3321 if (def_mount_opts & EXT4_DEFM_BLOCK_VALIDITY)
3322 set_opt(sb, BLOCK_VALIDITY);
3323 if (def_mount_opts & EXT4_DEFM_DISCARD)
3324 set_opt(sb, DISCARD);
3326 sbi->s_resuid = make_kuid(&init_user_ns, le16_to_cpu(es->s_def_resuid));
3327 sbi->s_resgid = make_kgid(&init_user_ns, le16_to_cpu(es->s_def_resgid));
3328 sbi->s_commit_interval = JBD2_DEFAULT_MAX_COMMIT_AGE * HZ;
3329 sbi->s_min_batch_time = EXT4_DEF_MIN_BATCH_TIME;
3330 sbi->s_max_batch_time = EXT4_DEF_MAX_BATCH_TIME;
3332 if ((def_mount_opts & EXT4_DEFM_NOBARRIER) == 0)
3333 set_opt(sb, BARRIER);
3336 * enable delayed allocation by default
3337 * Use -o nodelalloc to turn it off
3339 if (!IS_EXT3_SB(sb) && !IS_EXT2_SB(sb) &&
3340 ((def_mount_opts & EXT4_DEFM_NODELALLOC) == 0))
3341 set_opt(sb, DELALLOC);
3344 * set default s_li_wait_mult for lazyinit, for the case there is
3345 * no mount option specified.
3347 sbi->s_li_wait_mult = EXT4_DEF_LI_WAIT_MULT;
3349 if (!parse_options((char *) sbi->s_es->s_mount_opts, sb,
3350 &journal_devnum, &journal_ioprio, 0)) {
3351 ext4_msg(sb, KERN_WARNING,
3352 "failed to parse options in superblock: %s",
3353 sbi->s_es->s_mount_opts);
3355 sbi->s_def_mount_opt = sbi->s_mount_opt;
3356 if (!parse_options((char *) data, sb, &journal_devnum,
3357 &journal_ioprio, 0))
3360 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA) {
3361 printk_once(KERN_WARNING "EXT4-fs: Warning: mounting "
3362 "with data=journal disables delayed "
3363 "allocation and O_DIRECT support!\n");
3364 if (test_opt2(sb, EXPLICIT_DELALLOC)) {
3365 ext4_msg(sb, KERN_ERR, "can't mount with "
3366 "both data=journal and delalloc");
3369 if (test_opt(sb, DIOREAD_NOLOCK)) {
3370 ext4_msg(sb, KERN_ERR, "can't mount with "
3371 "both data=journal and delalloc");
3374 if (test_opt(sb, DELALLOC))
3375 clear_opt(sb, DELALLOC);
3378 sb->s_flags = (sb->s_flags & ~MS_POSIXACL) |
3379 (test_opt(sb, POSIX_ACL) ? MS_POSIXACL : 0);
3381 if (le32_to_cpu(es->s_rev_level) == EXT4_GOOD_OLD_REV &&
3382 (EXT4_HAS_COMPAT_FEATURE(sb, ~0U) ||
3383 EXT4_HAS_RO_COMPAT_FEATURE(sb, ~0U) ||
3384 EXT4_HAS_INCOMPAT_FEATURE(sb, ~0U)))
3385 ext4_msg(sb, KERN_WARNING,
3386 "feature flags set on rev 0 fs, "
3387 "running e2fsck is recommended");
3389 if (IS_EXT2_SB(sb)) {
3390 if (ext2_feature_set_ok(sb))
3391 ext4_msg(sb, KERN_INFO, "mounting ext2 file system "
3392 "using the ext4 subsystem");
3394 ext4_msg(sb, KERN_ERR, "couldn't mount as ext2 due "
3395 "to feature incompatibilities");
3400 if (IS_EXT3_SB(sb)) {
3401 if (ext3_feature_set_ok(sb))
3402 ext4_msg(sb, KERN_INFO, "mounting ext3 file system "
3403 "using the ext4 subsystem");
3405 ext4_msg(sb, KERN_ERR, "couldn't mount as ext3 due "
3406 "to feature incompatibilities");
3412 * Check feature flags regardless of the revision level, since we
3413 * previously didn't change the revision level when setting the flags,
3414 * so there is a chance incompat flags are set on a rev 0 filesystem.
3416 if (!ext4_feature_set_ok(sb, (sb->s_flags & MS_RDONLY)))
3419 blocksize = BLOCK_SIZE << le32_to_cpu(es->s_log_block_size);
3420 if (blocksize < EXT4_MIN_BLOCK_SIZE ||
3421 blocksize > EXT4_MAX_BLOCK_SIZE) {
3422 ext4_msg(sb, KERN_ERR,
3423 "Unsupported filesystem blocksize %d", blocksize);
3427 if (sb->s_blocksize != blocksize) {
3428 /* Validate the filesystem blocksize */
3429 if (!sb_set_blocksize(sb, blocksize)) {
3430 ext4_msg(sb, KERN_ERR, "bad block size %d",
3436 logical_sb_block = sb_block * EXT4_MIN_BLOCK_SIZE;
3437 offset = do_div(logical_sb_block, blocksize);
3438 bh = sb_bread(sb, logical_sb_block);
3440 ext4_msg(sb, KERN_ERR,
3441 "Can't read superblock on 2nd try");
3444 es = (struct ext4_super_block *)(bh->b_data + offset);
3446 if (es->s_magic != cpu_to_le16(EXT4_SUPER_MAGIC)) {
3447 ext4_msg(sb, KERN_ERR,
3448 "Magic mismatch, very weird!");
3453 has_huge_files = EXT4_HAS_RO_COMPAT_FEATURE(sb,
3454 EXT4_FEATURE_RO_COMPAT_HUGE_FILE);
3455 sbi->s_bitmap_maxbytes = ext4_max_bitmap_size(sb->s_blocksize_bits,
3457 sb->s_maxbytes = ext4_max_size(sb->s_blocksize_bits, has_huge_files);
3459 if (le32_to_cpu(es->s_rev_level) == EXT4_GOOD_OLD_REV) {
3460 sbi->s_inode_size = EXT4_GOOD_OLD_INODE_SIZE;
3461 sbi->s_first_ino = EXT4_GOOD_OLD_FIRST_INO;
3463 sbi->s_inode_size = le16_to_cpu(es->s_inode_size);
3464 sbi->s_first_ino = le32_to_cpu(es->s_first_ino);
3465 if ((sbi->s_inode_size < EXT4_GOOD_OLD_INODE_SIZE) ||
3466 (!is_power_of_2(sbi->s_inode_size)) ||
3467 (sbi->s_inode_size > blocksize)) {
3468 ext4_msg(sb, KERN_ERR,
3469 "unsupported inode size: %d",
3473 if (sbi->s_inode_size > EXT4_GOOD_OLD_INODE_SIZE)
3474 sb->s_time_gran = 1 << (EXT4_EPOCH_BITS - 2);
3477 sbi->s_desc_size = le16_to_cpu(es->s_desc_size);
3478 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_64BIT)) {
3479 if (sbi->s_desc_size < EXT4_MIN_DESC_SIZE_64BIT ||
3480 sbi->s_desc_size > EXT4_MAX_DESC_SIZE ||
3481 !is_power_of_2(sbi->s_desc_size)) {
3482 ext4_msg(sb, KERN_ERR,
3483 "unsupported descriptor size %lu",
3488 sbi->s_desc_size = EXT4_MIN_DESC_SIZE;
3490 sbi->s_blocks_per_group = le32_to_cpu(es->s_blocks_per_group);
3491 sbi->s_inodes_per_group = le32_to_cpu(es->s_inodes_per_group);
3492 if (EXT4_INODE_SIZE(sb) == 0 || EXT4_INODES_PER_GROUP(sb) == 0)
3495 sbi->s_inodes_per_block = blocksize / EXT4_INODE_SIZE(sb);
3496 if (sbi->s_inodes_per_block == 0)
3498 sbi->s_itb_per_group = sbi->s_inodes_per_group /
3499 sbi->s_inodes_per_block;
3500 sbi->s_desc_per_block = blocksize / EXT4_DESC_SIZE(sb);
3502 sbi->s_mount_state = le16_to_cpu(es->s_state);
3503 sbi->s_addr_per_block_bits = ilog2(EXT4_ADDR_PER_BLOCK(sb));
3504 sbi->s_desc_per_block_bits = ilog2(EXT4_DESC_PER_BLOCK(sb));
3506 for (i = 0; i < 4; i++)
3507 sbi->s_hash_seed[i] = le32_to_cpu(es->s_hash_seed[i]);
3508 sbi->s_def_hash_version = es->s_def_hash_version;
3509 i = le32_to_cpu(es->s_flags);
3510 if (i & EXT2_FLAGS_UNSIGNED_HASH)
3511 sbi->s_hash_unsigned = 3;
3512 else if ((i & EXT2_FLAGS_SIGNED_HASH) == 0) {
3513 #ifdef __CHAR_UNSIGNED__
3514 es->s_flags |= cpu_to_le32(EXT2_FLAGS_UNSIGNED_HASH);
3515 sbi->s_hash_unsigned = 3;
3517 es->s_flags |= cpu_to_le32(EXT2_FLAGS_SIGNED_HASH);
3521 /* Handle clustersize */
3522 clustersize = BLOCK_SIZE << le32_to_cpu(es->s_log_cluster_size);
3523 has_bigalloc = EXT4_HAS_RO_COMPAT_FEATURE(sb,
3524 EXT4_FEATURE_RO_COMPAT_BIGALLOC);
3526 if (clustersize < blocksize) {
3527 ext4_msg(sb, KERN_ERR,
3528 "cluster size (%d) smaller than "
3529 "block size (%d)", clustersize, blocksize);
3532 sbi->s_cluster_bits = le32_to_cpu(es->s_log_cluster_size) -
3533 le32_to_cpu(es->s_log_block_size);
3534 sbi->s_clusters_per_group =
3535 le32_to_cpu(es->s_clusters_per_group);
3536 if (sbi->s_clusters_per_group > blocksize * 8) {
3537 ext4_msg(sb, KERN_ERR,
3538 "#clusters per group too big: %lu",
3539 sbi->s_clusters_per_group);
3542 if (sbi->s_blocks_per_group !=
3543 (sbi->s_clusters_per_group * (clustersize / blocksize))) {
3544 ext4_msg(sb, KERN_ERR, "blocks per group (%lu) and "
3545 "clusters per group (%lu) inconsistent",
3546 sbi->s_blocks_per_group,
3547 sbi->s_clusters_per_group);
3551 if (clustersize != blocksize) {
3552 ext4_warning(sb, "fragment/cluster size (%d) != "
3553 "block size (%d)", clustersize,
3555 clustersize = blocksize;
3557 if (sbi->s_blocks_per_group > blocksize * 8) {
3558 ext4_msg(sb, KERN_ERR,
3559 "#blocks per group too big: %lu",
3560 sbi->s_blocks_per_group);
3563 sbi->s_clusters_per_group = sbi->s_blocks_per_group;
3564 sbi->s_cluster_bits = 0;
3566 sbi->s_cluster_ratio = clustersize / blocksize;
3568 if (sbi->s_inodes_per_group > blocksize * 8) {
3569 ext4_msg(sb, KERN_ERR,
3570 "#inodes per group too big: %lu",
3571 sbi->s_inodes_per_group);
3576 * Test whether we have more sectors than will fit in sector_t,
3577 * and whether the max offset is addressable by the page cache.
3579 err = generic_check_addressable(sb->s_blocksize_bits,
3580 ext4_blocks_count(es));
3582 ext4_msg(sb, KERN_ERR, "filesystem"
3583 " too large to mount safely on this system");
3584 if (sizeof(sector_t) < 8)
3585 ext4_msg(sb, KERN_WARNING, "CONFIG_LBDAF not enabled");
3589 if (EXT4_BLOCKS_PER_GROUP(sb) == 0)
3592 /* check blocks count against device size */
3593 blocks_count = sb->s_bdev->bd_inode->i_size >> sb->s_blocksize_bits;
3594 if (blocks_count && ext4_blocks_count(es) > blocks_count) {
3595 ext4_msg(sb, KERN_WARNING, "bad geometry: block count %llu "
3596 "exceeds size of device (%llu blocks)",
3597 ext4_blocks_count(es), blocks_count);
3602 * It makes no sense for the first data block to be beyond the end
3603 * of the filesystem.
3605 if (le32_to_cpu(es->s_first_data_block) >= ext4_blocks_count(es)) {
3606 ext4_msg(sb, KERN_WARNING, "bad geometry: first data "
3607 "block %u is beyond end of filesystem (%llu)",
3608 le32_to_cpu(es->s_first_data_block),
3609 ext4_blocks_count(es));
3612 blocks_count = (ext4_blocks_count(es) -
3613 le32_to_cpu(es->s_first_data_block) +
3614 EXT4_BLOCKS_PER_GROUP(sb) - 1);
3615 do_div(blocks_count, EXT4_BLOCKS_PER_GROUP(sb));
3616 if (blocks_count > ((uint64_t)1<<32) - EXT4_DESC_PER_BLOCK(sb)) {
3617 ext4_msg(sb, KERN_WARNING, "groups count too large: %u "
3618 "(block count %llu, first data block %u, "
3619 "blocks per group %lu)", sbi->s_groups_count,
3620 ext4_blocks_count(es),
3621 le32_to_cpu(es->s_first_data_block),
3622 EXT4_BLOCKS_PER_GROUP(sb));
3625 sbi->s_groups_count = blocks_count;
3626 sbi->s_blockfile_groups = min_t(ext4_group_t, sbi->s_groups_count,
3627 (EXT4_MAX_BLOCK_FILE_PHYS / EXT4_BLOCKS_PER_GROUP(sb)));
3628 db_count = (sbi->s_groups_count + EXT4_DESC_PER_BLOCK(sb) - 1) /
3629 EXT4_DESC_PER_BLOCK(sb);
3630 sbi->s_group_desc = ext4_kvmalloc(db_count *
3631 sizeof(struct buffer_head *),
3633 if (sbi->s_group_desc == NULL) {
3634 ext4_msg(sb, KERN_ERR, "not enough memory");
3640 sbi->s_proc = proc_mkdir(sb->s_id, ext4_proc_root);
3643 proc_create_data("options", S_IRUGO, sbi->s_proc,
3644 &ext4_seq_options_fops, sb);
3646 bgl_lock_init(sbi->s_blockgroup_lock);
3648 for (i = 0; i < db_count; i++) {
3649 block = descriptor_loc(sb, logical_sb_block, i);
3650 sbi->s_group_desc[i] = sb_bread(sb, block);
3651 if (!sbi->s_group_desc[i]) {
3652 ext4_msg(sb, KERN_ERR,
3653 "can't read group descriptor %d", i);
3658 if (!ext4_check_descriptors(sb, &first_not_zeroed)) {
3659 ext4_msg(sb, KERN_ERR, "group descriptors corrupted!");
3662 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_FLEX_BG))
3663 if (!ext4_fill_flex_info(sb)) {
3664 ext4_msg(sb, KERN_ERR,
3665 "unable to initialize "
3666 "flex_bg meta info!");
3670 sbi->s_gdb_count = db_count;
3671 get_random_bytes(&sbi->s_next_generation, sizeof(u32));
3672 spin_lock_init(&sbi->s_next_gen_lock);
3674 init_timer(&sbi->s_err_report);
3675 sbi->s_err_report.function = print_daily_error_info;
3676 sbi->s_err_report.data = (unsigned long) sb;
3678 err = percpu_counter_init(&sbi->s_freeclusters_counter,
3679 ext4_count_free_clusters(sb));
3681 err = percpu_counter_init(&sbi->s_freeinodes_counter,
3682 ext4_count_free_inodes(sb));
3685 err = percpu_counter_init(&sbi->s_dirs_counter,
3686 ext4_count_dirs(sb));
3689 err = percpu_counter_init(&sbi->s_dirtyclusters_counter, 0);
3692 ext4_msg(sb, KERN_ERR, "insufficient memory");
3696 sbi->s_stripe = ext4_get_stripe_size(sbi);
3697 sbi->s_max_writeback_mb_bump = 128;
3698 sbi->s_extent_max_zeroout_kb = 32;
3700 /* Register extent status tree shrinker */
3701 ext4_es_register_shrinker(sb);
3704 * set up enough so that it can read an inode
3706 if (!test_opt(sb, NOLOAD) &&
3707 EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL))
3708 sb->s_op = &ext4_sops;
3710 sb->s_op = &ext4_nojournal_sops;
3711 sb->s_export_op = &ext4_export_ops;
3712 sb->s_xattr = ext4_xattr_handlers;
3714 sb->s_qcop = &ext4_qctl_operations;
3715 sb->dq_op = &ext4_quota_operations;
3717 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_QUOTA)) {
3718 /* Use qctl operations for hidden quota files. */
3719 sb->s_qcop = &ext4_qctl_sysfile_operations;
3722 memcpy(sb->s_uuid, es->s_uuid, sizeof(es->s_uuid));
3724 INIT_LIST_HEAD(&sbi->s_orphan); /* unlinked but open files */
3725 mutex_init(&sbi->s_orphan_lock);
3729 needs_recovery = (es->s_last_orphan != 0 ||
3730 EXT4_HAS_INCOMPAT_FEATURE(sb,
3731 EXT4_FEATURE_INCOMPAT_RECOVER));
3733 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_MMP) &&
3734 !(sb->s_flags & MS_RDONLY))
3735 if (ext4_multi_mount_protect(sb, le64_to_cpu(es->s_mmp_block)))
3739 * The first inode we look at is the journal inode. Don't try
3740 * root first: it may be modified in the journal!
3742 if (!test_opt(sb, NOLOAD) &&
3743 EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL)) {
3744 if (ext4_load_journal(sb, es, journal_devnum))
3746 } else if (test_opt(sb, NOLOAD) && !(sb->s_flags & MS_RDONLY) &&
3747 EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER)) {
3748 ext4_msg(sb, KERN_ERR, "required journal recovery "
3749 "suppressed and not mounted read-only");
3750 goto failed_mount_wq;
3752 clear_opt(sb, DATA_FLAGS);
3753 sbi->s_journal = NULL;
3758 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_64BIT) &&
3759 !jbd2_journal_set_features(EXT4_SB(sb)->s_journal, 0, 0,
3760 JBD2_FEATURE_INCOMPAT_64BIT)) {
3761 ext4_msg(sb, KERN_ERR, "Failed to set 64-bit journal feature");
3762 goto failed_mount_wq;
3765 if (!set_journal_csum_feature_set(sb)) {
3766 ext4_msg(sb, KERN_ERR, "Failed to set journal checksum "
3768 goto failed_mount_wq;
3771 /* We have now updated the journal if required, so we can
3772 * validate the data journaling mode. */
3773 switch (test_opt(sb, DATA_FLAGS)) {
3775 /* No mode set, assume a default based on the journal
3776 * capabilities: ORDERED_DATA if the journal can
3777 * cope, else JOURNAL_DATA
3779 if (jbd2_journal_check_available_features
3780 (sbi->s_journal, 0, 0, JBD2_FEATURE_INCOMPAT_REVOKE))
3781 set_opt(sb, ORDERED_DATA);
3783 set_opt(sb, JOURNAL_DATA);
3786 case EXT4_MOUNT_ORDERED_DATA:
3787 case EXT4_MOUNT_WRITEBACK_DATA:
3788 if (!jbd2_journal_check_available_features
3789 (sbi->s_journal, 0, 0, JBD2_FEATURE_INCOMPAT_REVOKE)) {
3790 ext4_msg(sb, KERN_ERR, "Journal does not support "
3791 "requested data journaling mode");
3792 goto failed_mount_wq;
3797 set_task_ioprio(sbi->s_journal->j_task, journal_ioprio);
3799 sbi->s_journal->j_commit_callback = ext4_journal_commit_callback;
3802 * The journal may have updated the bg summary counts, so we
3803 * need to update the global counters.
3805 percpu_counter_set(&sbi->s_freeclusters_counter,
3806 ext4_count_free_clusters(sb));
3807 percpu_counter_set(&sbi->s_freeinodes_counter,
3808 ext4_count_free_inodes(sb));
3809 percpu_counter_set(&sbi->s_dirs_counter,
3810 ext4_count_dirs(sb));
3811 percpu_counter_set(&sbi->s_dirtyclusters_counter, 0);
3815 * Get the # of file system overhead blocks from the
3816 * superblock if present.
3818 if (es->s_overhead_clusters)
3819 sbi->s_overhead = le32_to_cpu(es->s_overhead_clusters);
3821 err = ext4_calculate_overhead(sb);
3823 goto failed_mount_wq;
3827 * The maximum number of concurrent works can be high and
3828 * concurrency isn't really necessary. Limit it to 1.
3830 EXT4_SB(sb)->dio_unwritten_wq =
3831 alloc_workqueue("ext4-dio-unwritten", WQ_MEM_RECLAIM | WQ_UNBOUND, 1);
3832 if (!EXT4_SB(sb)->dio_unwritten_wq) {
3833 printk(KERN_ERR "EXT4-fs: failed to create DIO workqueue\n");
3835 goto failed_mount_wq;
3839 * The jbd2_journal_load will have done any necessary log recovery,
3840 * so we can safely mount the rest of the filesystem now.
3843 root = ext4_iget(sb, EXT4_ROOT_INO);
3845 ext4_msg(sb, KERN_ERR, "get root inode failed");
3846 ret = PTR_ERR(root);
3850 if (!S_ISDIR(root->i_mode) || !root->i_blocks || !root->i_size) {
3851 ext4_msg(sb, KERN_ERR, "corrupt root inode, run e2fsck");
3855 sb->s_root = d_make_root(root);
3857 ext4_msg(sb, KERN_ERR, "get root dentry failed");
3862 if (ext4_setup_super(sb, es, sb->s_flags & MS_RDONLY))
3863 sb->s_flags |= MS_RDONLY;
3865 /* determine the minimum size of new large inodes, if present */
3866 if (sbi->s_inode_size > EXT4_GOOD_OLD_INODE_SIZE) {
3867 sbi->s_want_extra_isize = sizeof(struct ext4_inode) -
3868 EXT4_GOOD_OLD_INODE_SIZE;
3869 if (EXT4_HAS_RO_COMPAT_FEATURE(sb,
3870 EXT4_FEATURE_RO_COMPAT_EXTRA_ISIZE)) {
3871 if (sbi->s_want_extra_isize <
3872 le16_to_cpu(es->s_want_extra_isize))
3873 sbi->s_want_extra_isize =
3874 le16_to_cpu(es->s_want_extra_isize);
3875 if (sbi->s_want_extra_isize <
3876 le16_to_cpu(es->s_min_extra_isize))
3877 sbi->s_want_extra_isize =
3878 le16_to_cpu(es->s_min_extra_isize);
3881 /* Check if enough inode space is available */
3882 if (EXT4_GOOD_OLD_INODE_SIZE + sbi->s_want_extra_isize >
3883 sbi->s_inode_size) {
3884 sbi->s_want_extra_isize = sizeof(struct ext4_inode) -
3885 EXT4_GOOD_OLD_INODE_SIZE;
3886 ext4_msg(sb, KERN_INFO, "required extra inode space not"
3890 err = ext4_setup_system_zone(sb);
3892 ext4_msg(sb, KERN_ERR, "failed to initialize system "
3894 goto failed_mount4a;
3898 err = ext4_mb_init(sb);
3900 ext4_msg(sb, KERN_ERR, "failed to initialize mballoc (%d)",
3905 err = ext4_register_li_request(sb, first_not_zeroed);
3909 sbi->s_kobj.kset = ext4_kset;
3910 init_completion(&sbi->s_kobj_unregister);
3911 err = kobject_init_and_add(&sbi->s_kobj, &ext4_ktype, NULL,
3916 EXT4_SB(sb)->s_mount_state |= EXT4_ORPHAN_FS;
3917 ext4_orphan_cleanup(sb, es);
3918 EXT4_SB(sb)->s_mount_state &= ~EXT4_ORPHAN_FS;
3919 if (needs_recovery) {
3920 ext4_msg(sb, KERN_INFO, "recovery complete");
3921 ext4_mark_recovery_complete(sb, es);
3923 if (EXT4_SB(sb)->s_journal) {
3924 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA)
3925 descr = " journalled data mode";
3926 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA)
3927 descr = " ordered data mode";
3929 descr = " writeback data mode";
3931 descr = "out journal";
3934 /* Enable quota usage during mount. */
3935 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_QUOTA) &&
3936 !(sb->s_flags & MS_RDONLY)) {
3937 err = ext4_enable_quotas(sb);
3941 #endif /* CONFIG_QUOTA */
3943 if (test_opt(sb, DISCARD)) {
3944 struct request_queue *q = bdev_get_queue(sb->s_bdev);
3945 if (!blk_queue_discard(q))
3946 ext4_msg(sb, KERN_WARNING,
3947 "mounting with \"discard\" option, but "
3948 "the device does not support discard");
3951 ext4_msg(sb, KERN_INFO, "mounted filesystem with%s. "
3952 "Opts: %s%s%s", descr, sbi->s_es->s_mount_opts,
3953 *sbi->s_es->s_mount_opts ? "; " : "", orig_data);
3955 if (es->s_error_count)
3956 mod_timer(&sbi->s_err_report, jiffies + 300*HZ); /* 5 minutes */
3963 ext4_msg(sb, KERN_ERR, "VFS: Can't find ext4 filesystem");
3968 kobject_del(&sbi->s_kobj);
3971 ext4_unregister_li_request(sb);
3973 ext4_mb_release(sb);
3975 ext4_ext_release(sb);
3976 ext4_release_system_zone(sb);
3981 ext4_msg(sb, KERN_ERR, "mount failed");
3982 destroy_workqueue(EXT4_SB(sb)->dio_unwritten_wq);
3984 if (sbi->s_journal) {
3985 jbd2_journal_destroy(sbi->s_journal);
3986 sbi->s_journal = NULL;
3989 del_timer(&sbi->s_err_report);
3990 if (sbi->s_flex_groups)
3991 ext4_kvfree(sbi->s_flex_groups);
3992 percpu_counter_destroy(&sbi->s_freeclusters_counter);
3993 percpu_counter_destroy(&sbi->s_freeinodes_counter);
3994 percpu_counter_destroy(&sbi->s_dirs_counter);
3995 percpu_counter_destroy(&sbi->s_dirtyclusters_counter);
3997 kthread_stop(sbi->s_mmp_tsk);
3999 for (i = 0; i < db_count; i++)
4000 brelse(sbi->s_group_desc[i]);
4001 ext4_kvfree(sbi->s_group_desc);
4003 if (sbi->s_chksum_driver)
4004 crypto_free_shash(sbi->s_chksum_driver);
4006 remove_proc_entry("options", sbi->s_proc);
4007 remove_proc_entry(sb->s_id, ext4_proc_root);
4010 for (i = 0; i < MAXQUOTAS; i++)
4011 kfree(sbi->s_qf_names[i]);
4013 ext4_blkdev_remove(sbi);
4016 sb->s_fs_info = NULL;
4017 kfree(sbi->s_blockgroup_lock);
4021 return err ? err : ret;
4025 * Setup any per-fs journal parameters now. We'll do this both on
4026 * initial mount, once the journal has been initialised but before we've
4027 * done any recovery; and again on any subsequent remount.
4029 static void ext4_init_journal_params(struct super_block *sb, journal_t *journal)
4031 struct ext4_sb_info *sbi = EXT4_SB(sb);
4033 journal->j_commit_interval = sbi->s_commit_interval;
4034 journal->j_min_batch_time = sbi->s_min_batch_time;
4035 journal->j_max_batch_time = sbi->s_max_batch_time;
4037 write_lock(&journal->j_state_lock);
4038 if (test_opt(sb, BARRIER))
4039 journal->j_flags |= JBD2_BARRIER;
4041 journal->j_flags &= ~JBD2_BARRIER;
4042 if (test_opt(sb, DATA_ERR_ABORT))
4043 journal->j_flags |= JBD2_ABORT_ON_SYNCDATA_ERR;
4045 journal->j_flags &= ~JBD2_ABORT_ON_SYNCDATA_ERR;
4046 write_unlock(&journal->j_state_lock);
4049 static journal_t *ext4_get_journal(struct super_block *sb,
4050 unsigned int journal_inum)
4052 struct inode *journal_inode;
4055 BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL));
4057 /* First, test for the existence of a valid inode on disk. Bad
4058 * things happen if we iget() an unused inode, as the subsequent
4059 * iput() will try to delete it. */
4061 journal_inode = ext4_iget(sb, journal_inum);
4062 if (IS_ERR(journal_inode)) {
4063 ext4_msg(sb, KERN_ERR, "no journal found");
4066 if (!journal_inode->i_nlink) {
4067 make_bad_inode(journal_inode);
4068 iput(journal_inode);
4069 ext4_msg(sb, KERN_ERR, "journal inode is deleted");
4073 jbd_debug(2, "Journal inode found at %p: %lld bytes\n",
4074 journal_inode, journal_inode->i_size);
4075 if (!S_ISREG(journal_inode->i_mode)) {
4076 ext4_msg(sb, KERN_ERR, "invalid journal inode");
4077 iput(journal_inode);
4081 journal = jbd2_journal_init_inode(journal_inode);
4083 ext4_msg(sb, KERN_ERR, "Could not load journal inode");
4084 iput(journal_inode);
4087 journal->j_private = sb;
4088 ext4_init_journal_params(sb, journal);
4092 static journal_t *ext4_get_dev_journal(struct super_block *sb,
4095 struct buffer_head *bh;
4099 int hblock, blocksize;
4100 ext4_fsblk_t sb_block;
4101 unsigned long offset;
4102 struct ext4_super_block *es;
4103 struct block_device *bdev;
4105 BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL));
4107 bdev = ext4_blkdev_get(j_dev, sb);
4111 blocksize = sb->s_blocksize;
4112 hblock = bdev_logical_block_size(bdev);
4113 if (blocksize < hblock) {
4114 ext4_msg(sb, KERN_ERR,
4115 "blocksize too small for journal device");
4119 sb_block = EXT4_MIN_BLOCK_SIZE / blocksize;
4120 offset = EXT4_MIN_BLOCK_SIZE % blocksize;
4121 set_blocksize(bdev, blocksize);
4122 if (!(bh = __bread(bdev, sb_block, blocksize))) {
4123 ext4_msg(sb, KERN_ERR, "couldn't read superblock of "
4124 "external journal");
4128 es = (struct ext4_super_block *) (bh->b_data + offset);
4129 if ((le16_to_cpu(es->s_magic) != EXT4_SUPER_MAGIC) ||
4130 !(le32_to_cpu(es->s_feature_incompat) &
4131 EXT4_FEATURE_INCOMPAT_JOURNAL_DEV)) {
4132 ext4_msg(sb, KERN_ERR, "external journal has "
4138 if (memcmp(EXT4_SB(sb)->s_es->s_journal_uuid, es->s_uuid, 16)) {
4139 ext4_msg(sb, KERN_ERR, "journal UUID does not match");
4144 len = ext4_blocks_count(es);
4145 start = sb_block + 1;
4146 brelse(bh); /* we're done with the superblock */
4148 journal = jbd2_journal_init_dev(bdev, sb->s_bdev,
4149 start, len, blocksize);
4151 ext4_msg(sb, KERN_ERR, "failed to create device journal");
4154 journal->j_private = sb;
4155 ll_rw_block(READ, 1, &journal->j_sb_buffer);
4156 wait_on_buffer(journal->j_sb_buffer);
4157 if (!buffer_uptodate(journal->j_sb_buffer)) {
4158 ext4_msg(sb, KERN_ERR, "I/O error on journal device");
4161 if (be32_to_cpu(journal->j_superblock->s_nr_users) != 1) {
4162 ext4_msg(sb, KERN_ERR, "External journal has more than one "
4163 "user (unsupported) - %d",
4164 be32_to_cpu(journal->j_superblock->s_nr_users));
4167 EXT4_SB(sb)->journal_bdev = bdev;
4168 ext4_init_journal_params(sb, journal);
4172 jbd2_journal_destroy(journal);
4174 ext4_blkdev_put(bdev);
4178 static int ext4_load_journal(struct super_block *sb,
4179 struct ext4_super_block *es,
4180 unsigned long journal_devnum)
4183 unsigned int journal_inum = le32_to_cpu(es->s_journal_inum);
4186 int really_read_only;
4188 BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL));
4190 if (journal_devnum &&
4191 journal_devnum != le32_to_cpu(es->s_journal_dev)) {
4192 ext4_msg(sb, KERN_INFO, "external journal device major/minor "
4193 "numbers have changed");
4194 journal_dev = new_decode_dev(journal_devnum);
4196 journal_dev = new_decode_dev(le32_to_cpu(es->s_journal_dev));
4198 really_read_only = bdev_read_only(sb->s_bdev);
4201 * Are we loading a blank journal or performing recovery after a
4202 * crash? For recovery, we need to check in advance whether we
4203 * can get read-write access to the device.
4205 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER)) {
4206 if (sb->s_flags & MS_RDONLY) {
4207 ext4_msg(sb, KERN_INFO, "INFO: recovery "
4208 "required on readonly filesystem");
4209 if (really_read_only) {
4210 ext4_msg(sb, KERN_ERR, "write access "
4211 "unavailable, cannot proceed");
4214 ext4_msg(sb, KERN_INFO, "write access will "
4215 "be enabled during recovery");
4219 if (journal_inum && journal_dev) {
4220 ext4_msg(sb, KERN_ERR, "filesystem has both journal "
4221 "and inode journals!");
4226 if (!(journal = ext4_get_journal(sb, journal_inum)))
4229 if (!(journal = ext4_get_dev_journal(sb, journal_dev)))
4233 if (!(journal->j_flags & JBD2_BARRIER))
4234 ext4_msg(sb, KERN_INFO, "barriers disabled");
4236 if (!EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER))
4237 err = jbd2_journal_wipe(journal, !really_read_only);
4239 char *save = kmalloc(EXT4_S_ERR_LEN, GFP_KERNEL);
4241 memcpy(save, ((char *) es) +
4242 EXT4_S_ERR_START, EXT4_S_ERR_LEN);
4243 err = jbd2_journal_load(journal);
4245 memcpy(((char *) es) + EXT4_S_ERR_START,
4246 save, EXT4_S_ERR_LEN);
4251 ext4_msg(sb, KERN_ERR, "error loading journal");
4252 jbd2_journal_destroy(journal);
4256 EXT4_SB(sb)->s_journal = journal;
4257 ext4_clear_journal_err(sb, es);
4259 if (!really_read_only && journal_devnum &&
4260 journal_devnum != le32_to_cpu(es->s_journal_dev)) {
4261 es->s_journal_dev = cpu_to_le32(journal_devnum);
4263 /* Make sure we flush the recovery flag to disk. */
4264 ext4_commit_super(sb, 1);
4270 static int ext4_commit_super(struct super_block *sb, int sync)
4272 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
4273 struct buffer_head *sbh = EXT4_SB(sb)->s_sbh;
4276 if (!sbh || block_device_ejected(sb))
4278 if (buffer_write_io_error(sbh)) {
4280 * Oh, dear. A previous attempt to write the
4281 * superblock failed. This could happen because the
4282 * USB device was yanked out. Or it could happen to
4283 * be a transient write error and maybe the block will
4284 * be remapped. Nothing we can do but to retry the
4285 * write and hope for the best.
4287 ext4_msg(sb, KERN_ERR, "previous I/O error to "
4288 "superblock detected");
4289 clear_buffer_write_io_error(sbh);
4290 set_buffer_uptodate(sbh);
4293 * If the file system is mounted read-only, don't update the
4294 * superblock write time. This avoids updating the superblock
4295 * write time when we are mounting the root file system
4296 * read/only but we need to replay the journal; at that point,
4297 * for people who are east of GMT and who make their clock
4298 * tick in localtime for Windows bug-for-bug compatibility,
4299 * the clock is set in the future, and this will cause e2fsck
4300 * to complain and force a full file system check.
4302 if (!(sb->s_flags & MS_RDONLY))
4303 es->s_wtime = cpu_to_le32(get_seconds());
4304 if (sb->s_bdev->bd_part)
4305 es->s_kbytes_written =
4306 cpu_to_le64(EXT4_SB(sb)->s_kbytes_written +
4307 ((part_stat_read(sb->s_bdev->bd_part, sectors[1]) -
4308 EXT4_SB(sb)->s_sectors_written_start) >> 1));
4310 es->s_kbytes_written =
4311 cpu_to_le64(EXT4_SB(sb)->s_kbytes_written);
4312 ext4_free_blocks_count_set(es,
4313 EXT4_C2B(EXT4_SB(sb), percpu_counter_sum_positive(
4314 &EXT4_SB(sb)->s_freeclusters_counter)));
4315 es->s_free_inodes_count =
4316 cpu_to_le32(percpu_counter_sum_positive(
4317 &EXT4_SB(sb)->s_freeinodes_counter));
4318 BUFFER_TRACE(sbh, "marking dirty");
4319 ext4_superblock_csum_set(sb);
4320 mark_buffer_dirty(sbh);
4322 error = sync_dirty_buffer(sbh);
4326 error = buffer_write_io_error(sbh);
4328 ext4_msg(sb, KERN_ERR, "I/O error while writing "
4330 clear_buffer_write_io_error(sbh);
4331 set_buffer_uptodate(sbh);
4338 * Have we just finished recovery? If so, and if we are mounting (or
4339 * remounting) the filesystem readonly, then we will end up with a
4340 * consistent fs on disk. Record that fact.
4342 static void ext4_mark_recovery_complete(struct super_block *sb,
4343 struct ext4_super_block *es)
4345 journal_t *journal = EXT4_SB(sb)->s_journal;
4347 if (!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL)) {
4348 BUG_ON(journal != NULL);
4351 jbd2_journal_lock_updates(journal);
4352 if (jbd2_journal_flush(journal) < 0)
4355 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER) &&
4356 sb->s_flags & MS_RDONLY) {
4357 EXT4_CLEAR_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
4358 ext4_commit_super(sb, 1);
4362 jbd2_journal_unlock_updates(journal);
4366 * If we are mounting (or read-write remounting) a filesystem whose journal
4367 * has recorded an error from a previous lifetime, move that error to the
4368 * main filesystem now.
4370 static void ext4_clear_journal_err(struct super_block *sb,
4371 struct ext4_super_block *es)
4377 BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL));
4379 journal = EXT4_SB(sb)->s_journal;
4382 * Now check for any error status which may have been recorded in the
4383 * journal by a prior ext4_error() or ext4_abort()
4386 j_errno = jbd2_journal_errno(journal);
4390 errstr = ext4_decode_error(sb, j_errno, nbuf);
4391 ext4_warning(sb, "Filesystem error recorded "
4392 "from previous mount: %s", errstr);
4393 ext4_warning(sb, "Marking fs in need of filesystem check.");
4395 EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS;
4396 es->s_state |= cpu_to_le16(EXT4_ERROR_FS);
4397 ext4_commit_super(sb, 1);
4399 jbd2_journal_clear_err(journal);
4400 jbd2_journal_update_sb_errno(journal);
4405 * Force the running and committing transactions to commit,
4406 * and wait on the commit.
4408 int ext4_force_commit(struct super_block *sb)
4412 if (sb->s_flags & MS_RDONLY)
4415 journal = EXT4_SB(sb)->s_journal;
4416 return ext4_journal_force_commit(journal);
4419 static int ext4_sync_fs(struct super_block *sb, int wait)
4423 struct ext4_sb_info *sbi = EXT4_SB(sb);
4425 trace_ext4_sync_fs(sb, wait);
4426 flush_workqueue(sbi->dio_unwritten_wq);
4428 * Writeback quota in non-journalled quota case - journalled quota has
4431 dquot_writeback_dquots(sb, -1);
4432 if (jbd2_journal_start_commit(sbi->s_journal, &target)) {
4434 jbd2_log_wait_commit(sbi->s_journal, target);
4440 * LVM calls this function before a (read-only) snapshot is created. This
4441 * gives us a chance to flush the journal completely and mark the fs clean.
4443 * Note that only this function cannot bring a filesystem to be in a clean
4444 * state independently. It relies on upper layer to stop all data & metadata
4447 static int ext4_freeze(struct super_block *sb)
4452 if (sb->s_flags & MS_RDONLY)
4455 journal = EXT4_SB(sb)->s_journal;
4457 /* Now we set up the journal barrier. */
4458 jbd2_journal_lock_updates(journal);
4461 * Don't clear the needs_recovery flag if we failed to flush
4464 error = jbd2_journal_flush(journal);
4468 /* Journal blocked and flushed, clear needs_recovery flag. */
4469 EXT4_CLEAR_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
4470 error = ext4_commit_super(sb, 1);
4472 /* we rely on upper layer to stop further updates */
4473 jbd2_journal_unlock_updates(EXT4_SB(sb)->s_journal);
4478 * Called by LVM after the snapshot is done. We need to reset the RECOVER
4479 * flag here, even though the filesystem is not technically dirty yet.
4481 static int ext4_unfreeze(struct super_block *sb)
4483 if (sb->s_flags & MS_RDONLY)
4486 /* Reset the needs_recovery flag before the fs is unlocked. */
4487 EXT4_SET_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
4488 ext4_commit_super(sb, 1);
4493 * Structure to save mount options for ext4_remount's benefit
4495 struct ext4_mount_options {
4496 unsigned long s_mount_opt;
4497 unsigned long s_mount_opt2;
4500 unsigned long s_commit_interval;
4501 u32 s_min_batch_time, s_max_batch_time;
4504 char *s_qf_names[MAXQUOTAS];
4508 static int ext4_remount(struct super_block *sb, int *flags, char *data)
4510 struct ext4_super_block *es;
4511 struct ext4_sb_info *sbi = EXT4_SB(sb);
4512 unsigned long old_sb_flags;
4513 struct ext4_mount_options old_opts;
4514 int enable_quota = 0;
4516 unsigned int journal_ioprio = DEFAULT_JOURNAL_IOPRIO;
4521 char *orig_data = kstrdup(data, GFP_KERNEL);
4523 /* Store the original options */
4524 old_sb_flags = sb->s_flags;
4525 old_opts.s_mount_opt = sbi->s_mount_opt;
4526 old_opts.s_mount_opt2 = sbi->s_mount_opt2;
4527 old_opts.s_resuid = sbi->s_resuid;
4528 old_opts.s_resgid = sbi->s_resgid;
4529 old_opts.s_commit_interval = sbi->s_commit_interval;
4530 old_opts.s_min_batch_time = sbi->s_min_batch_time;
4531 old_opts.s_max_batch_time = sbi->s_max_batch_time;
4533 old_opts.s_jquota_fmt = sbi->s_jquota_fmt;
4534 for (i = 0; i < MAXQUOTAS; i++)
4535 if (sbi->s_qf_names[i]) {
4536 old_opts.s_qf_names[i] = kstrdup(sbi->s_qf_names[i],
4538 if (!old_opts.s_qf_names[i]) {
4539 for (j = 0; j < i; j++)
4540 kfree(old_opts.s_qf_names[j]);
4544 old_opts.s_qf_names[i] = NULL;
4546 if (sbi->s_journal && sbi->s_journal->j_task->io_context)
4547 journal_ioprio = sbi->s_journal->j_task->io_context->ioprio;
4550 * Allow the "check" option to be passed as a remount option.
4552 if (!parse_options(data, sb, NULL, &journal_ioprio, 1)) {
4557 if (sbi->s_mount_flags & EXT4_MF_FS_ABORTED)
4558 ext4_abort(sb, "Abort forced by user");
4560 sb->s_flags = (sb->s_flags & ~MS_POSIXACL) |
4561 (test_opt(sb, POSIX_ACL) ? MS_POSIXACL : 0);
4565 if (sbi->s_journal) {
4566 ext4_init_journal_params(sb, sbi->s_journal);
4567 set_task_ioprio(sbi->s_journal->j_task, journal_ioprio);
4570 if ((*flags & MS_RDONLY) != (sb->s_flags & MS_RDONLY)) {
4571 if (sbi->s_mount_flags & EXT4_MF_FS_ABORTED) {
4576 if (*flags & MS_RDONLY) {
4577 err = dquot_suspend(sb, -1);
4582 * First of all, the unconditional stuff we have to do
4583 * to disable replay of the journal when we next remount
4585 sb->s_flags |= MS_RDONLY;
4588 * OK, test if we are remounting a valid rw partition
4589 * readonly, and if so set the rdonly flag and then
4590 * mark the partition as valid again.
4592 if (!(es->s_state & cpu_to_le16(EXT4_VALID_FS)) &&
4593 (sbi->s_mount_state & EXT4_VALID_FS))
4594 es->s_state = cpu_to_le16(sbi->s_mount_state);
4597 ext4_mark_recovery_complete(sb, es);
4599 /* Make sure we can mount this feature set readwrite */
4600 if (!ext4_feature_set_ok(sb, 0)) {
4605 * Make sure the group descriptor checksums
4606 * are sane. If they aren't, refuse to remount r/w.
4608 for (g = 0; g < sbi->s_groups_count; g++) {
4609 struct ext4_group_desc *gdp =
4610 ext4_get_group_desc(sb, g, NULL);
4612 if (!ext4_group_desc_csum_verify(sb, g, gdp)) {
4613 ext4_msg(sb, KERN_ERR,
4614 "ext4_remount: Checksum for group %u failed (%u!=%u)",
4615 g, le16_to_cpu(ext4_group_desc_csum(sbi, g, gdp)),
4616 le16_to_cpu(gdp->bg_checksum));
4623 * If we have an unprocessed orphan list hanging
4624 * around from a previously readonly bdev mount,
4625 * require a full umount/remount for now.
4627 if (es->s_last_orphan) {
4628 ext4_msg(sb, KERN_WARNING, "Couldn't "
4629 "remount RDWR because of unprocessed "
4630 "orphan inode list. Please "
4631 "umount/remount instead");
4637 * Mounting a RDONLY partition read-write, so reread
4638 * and store the current valid flag. (It may have
4639 * been changed by e2fsck since we originally mounted
4643 ext4_clear_journal_err(sb, es);
4644 sbi->s_mount_state = le16_to_cpu(es->s_state);
4645 if (!ext4_setup_super(sb, es, 0))
4646 sb->s_flags &= ~MS_RDONLY;
4647 if (EXT4_HAS_INCOMPAT_FEATURE(sb,
4648 EXT4_FEATURE_INCOMPAT_MMP))
4649 if (ext4_multi_mount_protect(sb,
4650 le64_to_cpu(es->s_mmp_block))) {
4659 * Reinitialize lazy itable initialization thread based on
4662 if ((sb->s_flags & MS_RDONLY) || !test_opt(sb, INIT_INODE_TABLE))
4663 ext4_unregister_li_request(sb);
4665 ext4_group_t first_not_zeroed;
4666 first_not_zeroed = ext4_has_uninit_itable(sb);
4667 ext4_register_li_request(sb, first_not_zeroed);
4670 ext4_setup_system_zone(sb);
4671 if (sbi->s_journal == NULL && !(old_sb_flags & MS_RDONLY))
4672 ext4_commit_super(sb, 1);
4675 /* Release old quota file names */
4676 for (i = 0; i < MAXQUOTAS; i++)
4677 kfree(old_opts.s_qf_names[i]);
4679 if (sb_any_quota_suspended(sb))
4680 dquot_resume(sb, -1);
4681 else if (EXT4_HAS_RO_COMPAT_FEATURE(sb,
4682 EXT4_FEATURE_RO_COMPAT_QUOTA)) {
4683 err = ext4_enable_quotas(sb);
4690 ext4_msg(sb, KERN_INFO, "re-mounted. Opts: %s", orig_data);
4695 sb->s_flags = old_sb_flags;
4696 sbi->s_mount_opt = old_opts.s_mount_opt;
4697 sbi->s_mount_opt2 = old_opts.s_mount_opt2;
4698 sbi->s_resuid = old_opts.s_resuid;
4699 sbi->s_resgid = old_opts.s_resgid;
4700 sbi->s_commit_interval = old_opts.s_commit_interval;
4701 sbi->s_min_batch_time = old_opts.s_min_batch_time;
4702 sbi->s_max_batch_time = old_opts.s_max_batch_time;
4704 sbi->s_jquota_fmt = old_opts.s_jquota_fmt;
4705 for (i = 0; i < MAXQUOTAS; i++) {
4706 kfree(sbi->s_qf_names[i]);
4707 sbi->s_qf_names[i] = old_opts.s_qf_names[i];
4714 static int ext4_statfs(struct dentry *dentry, struct kstatfs *buf)
4716 struct super_block *sb = dentry->d_sb;
4717 struct ext4_sb_info *sbi = EXT4_SB(sb);
4718 struct ext4_super_block *es = sbi->s_es;
4719 ext4_fsblk_t overhead = 0;
4723 if (!test_opt(sb, MINIX_DF))
4724 overhead = sbi->s_overhead;
4726 buf->f_type = EXT4_SUPER_MAGIC;
4727 buf->f_bsize = sb->s_blocksize;
4728 buf->f_blocks = ext4_blocks_count(es) - EXT4_C2B(sbi, overhead);
4729 bfree = percpu_counter_sum_positive(&sbi->s_freeclusters_counter) -
4730 percpu_counter_sum_positive(&sbi->s_dirtyclusters_counter);
4731 /* prevent underflow in case that few free space is available */
4732 buf->f_bfree = EXT4_C2B(sbi, max_t(s64, bfree, 0));
4733 buf->f_bavail = buf->f_bfree - ext4_r_blocks_count(es);
4734 if (buf->f_bfree < ext4_r_blocks_count(es))
4736 buf->f_files = le32_to_cpu(es->s_inodes_count);
4737 buf->f_ffree = percpu_counter_sum_positive(&sbi->s_freeinodes_counter);
4738 buf->f_namelen = EXT4_NAME_LEN;
4739 fsid = le64_to_cpup((void *)es->s_uuid) ^
4740 le64_to_cpup((void *)es->s_uuid + sizeof(u64));
4741 buf->f_fsid.val[0] = fsid & 0xFFFFFFFFUL;
4742 buf->f_fsid.val[1] = (fsid >> 32) & 0xFFFFFFFFUL;
4747 /* Helper function for writing quotas on sync - we need to start transaction
4748 * before quota file is locked for write. Otherwise the are possible deadlocks:
4749 * Process 1 Process 2
4750 * ext4_create() quota_sync()
4751 * jbd2_journal_start() write_dquot()
4752 * dquot_initialize() down(dqio_mutex)
4753 * down(dqio_mutex) jbd2_journal_start()
4759 static inline struct inode *dquot_to_inode(struct dquot *dquot)
4761 return sb_dqopt(dquot->dq_sb)->files[dquot->dq_id.type];
4764 static int ext4_write_dquot(struct dquot *dquot)
4768 struct inode *inode;
4770 inode = dquot_to_inode(dquot);
4771 handle = ext4_journal_start(inode, EXT4_HT_QUOTA,
4772 EXT4_QUOTA_TRANS_BLOCKS(dquot->dq_sb));
4774 return PTR_ERR(handle);
4775 ret = dquot_commit(dquot);
4776 err = ext4_journal_stop(handle);
4782 static int ext4_acquire_dquot(struct dquot *dquot)
4787 handle = ext4_journal_start(dquot_to_inode(dquot), EXT4_HT_QUOTA,
4788 EXT4_QUOTA_INIT_BLOCKS(dquot->dq_sb));
4790 return PTR_ERR(handle);
4791 ret = dquot_acquire(dquot);
4792 err = ext4_journal_stop(handle);
4798 static int ext4_release_dquot(struct dquot *dquot)
4803 handle = ext4_journal_start(dquot_to_inode(dquot), EXT4_HT_QUOTA,
4804 EXT4_QUOTA_DEL_BLOCKS(dquot->dq_sb));
4805 if (IS_ERR(handle)) {
4806 /* Release dquot anyway to avoid endless cycle in dqput() */
4807 dquot_release(dquot);
4808 return PTR_ERR(handle);
4810 ret = dquot_release(dquot);
4811 err = ext4_journal_stop(handle);
4817 static int ext4_mark_dquot_dirty(struct dquot *dquot)
4819 /* Are we journaling quotas? */
4820 if (EXT4_SB(dquot->dq_sb)->s_qf_names[USRQUOTA] ||
4821 EXT4_SB(dquot->dq_sb)->s_qf_names[GRPQUOTA]) {
4822 dquot_mark_dquot_dirty(dquot);
4823 return ext4_write_dquot(dquot);
4825 return dquot_mark_dquot_dirty(dquot);
4829 static int ext4_write_info(struct super_block *sb, int type)
4834 /* Data block + inode block */
4835 handle = ext4_journal_start(sb->s_root->d_inode, EXT4_HT_QUOTA, 2);
4837 return PTR_ERR(handle);
4838 ret = dquot_commit_info(sb, type);
4839 err = ext4_journal_stop(handle);
4846 * Turn on quotas during mount time - we need to find
4847 * the quota file and such...
4849 static int ext4_quota_on_mount(struct super_block *sb, int type)
4851 return dquot_quota_on_mount(sb, EXT4_SB(sb)->s_qf_names[type],
4852 EXT4_SB(sb)->s_jquota_fmt, type);
4856 * Standard function to be called on quota_on
4858 static int ext4_quota_on(struct super_block *sb, int type, int format_id,
4863 if (!test_opt(sb, QUOTA))
4866 /* Quotafile not on the same filesystem? */
4867 if (path->dentry->d_sb != sb)
4869 /* Journaling quota? */
4870 if (EXT4_SB(sb)->s_qf_names[type]) {
4871 /* Quotafile not in fs root? */
4872 if (path->dentry->d_parent != sb->s_root)
4873 ext4_msg(sb, KERN_WARNING,
4874 "Quota file not on filesystem root. "
4875 "Journaled quota will not work");
4879 * When we journal data on quota file, we have to flush journal to see
4880 * all updates to the file when we bypass pagecache...
4882 if (EXT4_SB(sb)->s_journal &&
4883 ext4_should_journal_data(path->dentry->d_inode)) {
4885 * We don't need to lock updates but journal_flush() could
4886 * otherwise be livelocked...
4888 jbd2_journal_lock_updates(EXT4_SB(sb)->s_journal);
4889 err = jbd2_journal_flush(EXT4_SB(sb)->s_journal);
4890 jbd2_journal_unlock_updates(EXT4_SB(sb)->s_journal);
4895 return dquot_quota_on(sb, type, format_id, path);
4898 static int ext4_quota_enable(struct super_block *sb, int type, int format_id,
4902 struct inode *qf_inode;
4903 unsigned long qf_inums[MAXQUOTAS] = {
4904 le32_to_cpu(EXT4_SB(sb)->s_es->s_usr_quota_inum),
4905 le32_to_cpu(EXT4_SB(sb)->s_es->s_grp_quota_inum)
4908 BUG_ON(!EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_QUOTA));
4910 if (!qf_inums[type])
4913 qf_inode = ext4_iget(sb, qf_inums[type]);
4914 if (IS_ERR(qf_inode)) {
4915 ext4_error(sb, "Bad quota inode # %lu", qf_inums[type]);
4916 return PTR_ERR(qf_inode);
4919 err = dquot_enable(qf_inode, type, format_id, flags);
4925 /* Enable usage tracking for all quota types. */
4926 static int ext4_enable_quotas(struct super_block *sb)
4929 unsigned long qf_inums[MAXQUOTAS] = {
4930 le32_to_cpu(EXT4_SB(sb)->s_es->s_usr_quota_inum),
4931 le32_to_cpu(EXT4_SB(sb)->s_es->s_grp_quota_inum)
4934 sb_dqopt(sb)->flags |= DQUOT_QUOTA_SYS_FILE;
4935 for (type = 0; type < MAXQUOTAS; type++) {
4936 if (qf_inums[type]) {
4937 err = ext4_quota_enable(sb, type, QFMT_VFS_V1,
4938 DQUOT_USAGE_ENABLED);
4941 "Failed to enable quota tracking "
4942 "(type=%d, err=%d). Please run "
4943 "e2fsck to fix.", type, err);
4952 * quota_on function that is used when QUOTA feature is set.
4954 static int ext4_quota_on_sysfile(struct super_block *sb, int type,
4957 if (!EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_QUOTA))
4961 * USAGE was enabled at mount time. Only need to enable LIMITS now.
4963 return ext4_quota_enable(sb, type, format_id, DQUOT_LIMITS_ENABLED);
4966 static int ext4_quota_off(struct super_block *sb, int type)
4968 struct inode *inode = sb_dqopt(sb)->files[type];
4971 /* Force all delayed allocation blocks to be allocated.
4972 * Caller already holds s_umount sem */
4973 if (test_opt(sb, DELALLOC))
4974 sync_filesystem(sb);
4979 /* Update modification times of quota files when userspace can
4980 * start looking at them */
4981 handle = ext4_journal_start(inode, EXT4_HT_QUOTA, 1);
4984 inode->i_mtime = inode->i_ctime = CURRENT_TIME;
4985 ext4_mark_inode_dirty(handle, inode);
4986 ext4_journal_stop(handle);
4989 return dquot_quota_off(sb, type);
4993 * quota_off function that is used when QUOTA feature is set.
4995 static int ext4_quota_off_sysfile(struct super_block *sb, int type)
4997 if (!EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_QUOTA))
5000 /* Disable only the limits. */
5001 return dquot_disable(sb, type, DQUOT_LIMITS_ENABLED);
5004 /* Read data from quotafile - avoid pagecache and such because we cannot afford
5005 * acquiring the locks... As quota files are never truncated and quota code
5006 * itself serializes the operations (and no one else should touch the files)
5007 * we don't have to be afraid of races */
5008 static ssize_t ext4_quota_read(struct super_block *sb, int type, char *data,
5009 size_t len, loff_t off)
5011 struct inode *inode = sb_dqopt(sb)->files[type];
5012 ext4_lblk_t blk = off >> EXT4_BLOCK_SIZE_BITS(sb);
5014 int offset = off & (sb->s_blocksize - 1);
5017 struct buffer_head *bh;
5018 loff_t i_size = i_size_read(inode);
5022 if (off+len > i_size)
5025 while (toread > 0) {
5026 tocopy = sb->s_blocksize - offset < toread ?
5027 sb->s_blocksize - offset : toread;
5028 bh = ext4_bread(NULL, inode, blk, 0, &err);
5031 if (!bh) /* A hole? */
5032 memset(data, 0, tocopy);
5034 memcpy(data, bh->b_data+offset, tocopy);
5044 /* Write to quotafile (we know the transaction is already started and has
5045 * enough credits) */
5046 static ssize_t ext4_quota_write(struct super_block *sb, int type,
5047 const char *data, size_t len, loff_t off)
5049 struct inode *inode = sb_dqopt(sb)->files[type];
5050 ext4_lblk_t blk = off >> EXT4_BLOCK_SIZE_BITS(sb);
5052 int offset = off & (sb->s_blocksize - 1);
5053 struct buffer_head *bh;
5054 handle_t *handle = journal_current_handle();
5056 if (EXT4_SB(sb)->s_journal && !handle) {
5057 ext4_msg(sb, KERN_WARNING, "Quota write (off=%llu, len=%llu)"
5058 " cancelled because transaction is not started",
5059 (unsigned long long)off, (unsigned long long)len);
5063 * Since we account only one data block in transaction credits,
5064 * then it is impossible to cross a block boundary.
5066 if (sb->s_blocksize - offset < len) {
5067 ext4_msg(sb, KERN_WARNING, "Quota write (off=%llu, len=%llu)"
5068 " cancelled because not block aligned",
5069 (unsigned long long)off, (unsigned long long)len);
5073 bh = ext4_bread(handle, inode, blk, 1, &err);
5076 err = ext4_journal_get_write_access(handle, bh);
5082 memcpy(bh->b_data+offset, data, len);
5083 flush_dcache_page(bh->b_page);
5085 err = ext4_handle_dirty_metadata(handle, NULL, bh);
5090 if (inode->i_size < off + len) {
5091 i_size_write(inode, off + len);
5092 EXT4_I(inode)->i_disksize = inode->i_size;
5093 ext4_mark_inode_dirty(handle, inode);
5100 static struct dentry *ext4_mount(struct file_system_type *fs_type, int flags,
5101 const char *dev_name, void *data)
5103 return mount_bdev(fs_type, flags, dev_name, data, ext4_fill_super);
5106 #if !defined(CONFIG_EXT2_FS) && !defined(CONFIG_EXT2_FS_MODULE) && defined(CONFIG_EXT4_USE_FOR_EXT23)
5107 static inline void register_as_ext2(void)
5109 int err = register_filesystem(&ext2_fs_type);
5112 "EXT4-fs: Unable to register as ext2 (%d)\n", err);
5115 static inline void unregister_as_ext2(void)
5117 unregister_filesystem(&ext2_fs_type);
5120 static inline int ext2_feature_set_ok(struct super_block *sb)
5122 if (EXT4_HAS_INCOMPAT_FEATURE(sb, ~EXT2_FEATURE_INCOMPAT_SUPP))
5124 if (sb->s_flags & MS_RDONLY)
5126 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, ~EXT2_FEATURE_RO_COMPAT_SUPP))
5130 MODULE_ALIAS("ext2");
5132 static inline void register_as_ext2(void) { }
5133 static inline void unregister_as_ext2(void) { }
5134 static inline int ext2_feature_set_ok(struct super_block *sb) { return 0; }
5137 #if !defined(CONFIG_EXT3_FS) && !defined(CONFIG_EXT3_FS_MODULE) && defined(CONFIG_EXT4_USE_FOR_EXT23)
5138 static inline void register_as_ext3(void)
5140 int err = register_filesystem(&ext3_fs_type);
5143 "EXT4-fs: Unable to register as ext3 (%d)\n", err);
5146 static inline void unregister_as_ext3(void)
5148 unregister_filesystem(&ext3_fs_type);
5151 static inline int ext3_feature_set_ok(struct super_block *sb)
5153 if (EXT4_HAS_INCOMPAT_FEATURE(sb, ~EXT3_FEATURE_INCOMPAT_SUPP))
5155 if (!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL))
5157 if (sb->s_flags & MS_RDONLY)
5159 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, ~EXT3_FEATURE_RO_COMPAT_SUPP))
5163 MODULE_ALIAS("ext3");
5165 static inline void register_as_ext3(void) { }
5166 static inline void unregister_as_ext3(void) { }
5167 static inline int ext3_feature_set_ok(struct super_block *sb) { return 0; }
5170 static struct file_system_type ext4_fs_type = {
5171 .owner = THIS_MODULE,
5173 .mount = ext4_mount,
5174 .kill_sb = kill_block_super,
5175 .fs_flags = FS_REQUIRES_DEV,
5178 static int __init ext4_init_feat_adverts(void)
5180 struct ext4_features *ef;
5183 ef = kzalloc(sizeof(struct ext4_features), GFP_KERNEL);
5187 ef->f_kobj.kset = ext4_kset;
5188 init_completion(&ef->f_kobj_unregister);
5189 ret = kobject_init_and_add(&ef->f_kobj, &ext4_feat_ktype, NULL,
5202 static void ext4_exit_feat_adverts(void)
5204 kobject_put(&ext4_feat->f_kobj);
5205 wait_for_completion(&ext4_feat->f_kobj_unregister);
5209 /* Shared across all ext4 file systems */
5210 wait_queue_head_t ext4__ioend_wq[EXT4_WQ_HASH_SZ];
5211 struct mutex ext4__aio_mutex[EXT4_WQ_HASH_SZ];
5213 static int __init ext4_init_fs(void)
5217 ext4_li_info = NULL;
5218 mutex_init(&ext4_li_mtx);
5220 /* Build-time check for flags consistency */
5221 ext4_check_flag_values();
5223 for (i = 0; i < EXT4_WQ_HASH_SZ; i++) {
5224 mutex_init(&ext4__aio_mutex[i]);
5225 init_waitqueue_head(&ext4__ioend_wq[i]);
5228 err = ext4_init_es();
5232 err = ext4_init_pageio();
5236 err = ext4_init_system_zone();
5239 ext4_kset = kset_create_and_add("ext4", NULL, fs_kobj);
5244 ext4_proc_root = proc_mkdir("fs/ext4", NULL);
5246 err = ext4_init_feat_adverts();
5250 err = ext4_init_mballoc();
5254 err = ext4_init_xattr();
5257 err = init_inodecache();
5262 err = register_filesystem(&ext4_fs_type);
5268 unregister_as_ext2();
5269 unregister_as_ext3();
5270 destroy_inodecache();
5274 ext4_exit_mballoc();
5276 ext4_exit_feat_adverts();
5279 remove_proc_entry("fs/ext4", NULL);
5280 kset_unregister(ext4_kset);
5282 ext4_exit_system_zone();
5291 static void __exit ext4_exit_fs(void)
5293 ext4_destroy_lazyinit_thread();
5294 unregister_as_ext2();
5295 unregister_as_ext3();
5296 unregister_filesystem(&ext4_fs_type);
5297 destroy_inodecache();
5299 ext4_exit_mballoc();
5300 ext4_exit_feat_adverts();
5301 remove_proc_entry("fs/ext4", NULL);
5302 kset_unregister(ext4_kset);
5303 ext4_exit_system_zone();
5307 MODULE_AUTHOR("Remy Card, Stephen Tweedie, Andrew Morton, Andreas Dilger, Theodore Ts'o and others");
5308 MODULE_DESCRIPTION("Fourth Extended Filesystem");
5309 MODULE_LICENSE("GPL");
5310 module_init(ext4_init_fs)
5311 module_exit(ext4_exit_fs)