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);
84 static int ext4_reserve_clusters(struct ext4_sb_info *, ext4_fsblk_t);
86 #if !defined(CONFIG_EXT2_FS) && !defined(CONFIG_EXT2_FS_MODULE) && defined(CONFIG_EXT4_USE_FOR_EXT23)
87 static struct file_system_type ext2_fs_type = {
91 .kill_sb = kill_block_super,
92 .fs_flags = FS_REQUIRES_DEV,
94 MODULE_ALIAS_FS("ext2");
96 #define IS_EXT2_SB(sb) ((sb)->s_bdev->bd_holder == &ext2_fs_type)
98 #define IS_EXT2_SB(sb) (0)
102 #if !defined(CONFIG_EXT3_FS) && !defined(CONFIG_EXT3_FS_MODULE) && defined(CONFIG_EXT4_USE_FOR_EXT23)
103 static struct file_system_type ext3_fs_type = {
104 .owner = THIS_MODULE,
107 .kill_sb = kill_block_super,
108 .fs_flags = FS_REQUIRES_DEV,
110 MODULE_ALIAS_FS("ext3");
111 MODULE_ALIAS("ext3");
112 #define IS_EXT3_SB(sb) ((sb)->s_bdev->bd_holder == &ext3_fs_type)
114 #define IS_EXT3_SB(sb) (0)
117 static int ext4_verify_csum_type(struct super_block *sb,
118 struct ext4_super_block *es)
120 if (!EXT4_HAS_RO_COMPAT_FEATURE(sb,
121 EXT4_FEATURE_RO_COMPAT_METADATA_CSUM))
124 return es->s_checksum_type == EXT4_CRC32C_CHKSUM;
127 static __le32 ext4_superblock_csum(struct super_block *sb,
128 struct ext4_super_block *es)
130 struct ext4_sb_info *sbi = EXT4_SB(sb);
131 int offset = offsetof(struct ext4_super_block, s_checksum);
134 csum = ext4_chksum(sbi, ~0, (char *)es, offset);
136 return cpu_to_le32(csum);
139 int ext4_superblock_csum_verify(struct super_block *sb,
140 struct ext4_super_block *es)
142 if (!EXT4_HAS_RO_COMPAT_FEATURE(sb,
143 EXT4_FEATURE_RO_COMPAT_METADATA_CSUM))
146 return es->s_checksum == ext4_superblock_csum(sb, es);
149 void ext4_superblock_csum_set(struct super_block *sb)
151 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
153 if (!EXT4_HAS_RO_COMPAT_FEATURE(sb,
154 EXT4_FEATURE_RO_COMPAT_METADATA_CSUM))
157 es->s_checksum = ext4_superblock_csum(sb, es);
160 void *ext4_kvmalloc(size_t size, gfp_t flags)
164 ret = kmalloc(size, flags);
166 ret = __vmalloc(size, flags, PAGE_KERNEL);
170 void *ext4_kvzalloc(size_t size, gfp_t flags)
174 ret = kzalloc(size, flags);
176 ret = __vmalloc(size, flags | __GFP_ZERO, PAGE_KERNEL);
180 void ext4_kvfree(void *ptr)
182 if (is_vmalloc_addr(ptr))
189 ext4_fsblk_t ext4_block_bitmap(struct super_block *sb,
190 struct ext4_group_desc *bg)
192 return le32_to_cpu(bg->bg_block_bitmap_lo) |
193 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
194 (ext4_fsblk_t)le32_to_cpu(bg->bg_block_bitmap_hi) << 32 : 0);
197 ext4_fsblk_t ext4_inode_bitmap(struct super_block *sb,
198 struct ext4_group_desc *bg)
200 return le32_to_cpu(bg->bg_inode_bitmap_lo) |
201 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
202 (ext4_fsblk_t)le32_to_cpu(bg->bg_inode_bitmap_hi) << 32 : 0);
205 ext4_fsblk_t ext4_inode_table(struct super_block *sb,
206 struct ext4_group_desc *bg)
208 return le32_to_cpu(bg->bg_inode_table_lo) |
209 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
210 (ext4_fsblk_t)le32_to_cpu(bg->bg_inode_table_hi) << 32 : 0);
213 __u32 ext4_free_group_clusters(struct super_block *sb,
214 struct ext4_group_desc *bg)
216 return le16_to_cpu(bg->bg_free_blocks_count_lo) |
217 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
218 (__u32)le16_to_cpu(bg->bg_free_blocks_count_hi) << 16 : 0);
221 __u32 ext4_free_inodes_count(struct super_block *sb,
222 struct ext4_group_desc *bg)
224 return le16_to_cpu(bg->bg_free_inodes_count_lo) |
225 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
226 (__u32)le16_to_cpu(bg->bg_free_inodes_count_hi) << 16 : 0);
229 __u32 ext4_used_dirs_count(struct super_block *sb,
230 struct ext4_group_desc *bg)
232 return le16_to_cpu(bg->bg_used_dirs_count_lo) |
233 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
234 (__u32)le16_to_cpu(bg->bg_used_dirs_count_hi) << 16 : 0);
237 __u32 ext4_itable_unused_count(struct super_block *sb,
238 struct ext4_group_desc *bg)
240 return le16_to_cpu(bg->bg_itable_unused_lo) |
241 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
242 (__u32)le16_to_cpu(bg->bg_itable_unused_hi) << 16 : 0);
245 void ext4_block_bitmap_set(struct super_block *sb,
246 struct ext4_group_desc *bg, ext4_fsblk_t blk)
248 bg->bg_block_bitmap_lo = cpu_to_le32((u32)blk);
249 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
250 bg->bg_block_bitmap_hi = cpu_to_le32(blk >> 32);
253 void ext4_inode_bitmap_set(struct super_block *sb,
254 struct ext4_group_desc *bg, ext4_fsblk_t blk)
256 bg->bg_inode_bitmap_lo = cpu_to_le32((u32)blk);
257 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
258 bg->bg_inode_bitmap_hi = cpu_to_le32(blk >> 32);
261 void ext4_inode_table_set(struct super_block *sb,
262 struct ext4_group_desc *bg, ext4_fsblk_t blk)
264 bg->bg_inode_table_lo = cpu_to_le32((u32)blk);
265 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
266 bg->bg_inode_table_hi = cpu_to_le32(blk >> 32);
269 void ext4_free_group_clusters_set(struct super_block *sb,
270 struct ext4_group_desc *bg, __u32 count)
272 bg->bg_free_blocks_count_lo = cpu_to_le16((__u16)count);
273 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
274 bg->bg_free_blocks_count_hi = cpu_to_le16(count >> 16);
277 void ext4_free_inodes_set(struct super_block *sb,
278 struct ext4_group_desc *bg, __u32 count)
280 bg->bg_free_inodes_count_lo = cpu_to_le16((__u16)count);
281 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
282 bg->bg_free_inodes_count_hi = cpu_to_le16(count >> 16);
285 void ext4_used_dirs_set(struct super_block *sb,
286 struct ext4_group_desc *bg, __u32 count)
288 bg->bg_used_dirs_count_lo = cpu_to_le16((__u16)count);
289 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
290 bg->bg_used_dirs_count_hi = cpu_to_le16(count >> 16);
293 void ext4_itable_unused_set(struct super_block *sb,
294 struct ext4_group_desc *bg, __u32 count)
296 bg->bg_itable_unused_lo = cpu_to_le16((__u16)count);
297 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
298 bg->bg_itable_unused_hi = cpu_to_le16(count >> 16);
302 static void __save_error_info(struct super_block *sb, const char *func,
305 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
307 EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS;
308 es->s_state |= cpu_to_le16(EXT4_ERROR_FS);
309 es->s_last_error_time = cpu_to_le32(get_seconds());
310 strncpy(es->s_last_error_func, func, sizeof(es->s_last_error_func));
311 es->s_last_error_line = cpu_to_le32(line);
312 if (!es->s_first_error_time) {
313 es->s_first_error_time = es->s_last_error_time;
314 strncpy(es->s_first_error_func, func,
315 sizeof(es->s_first_error_func));
316 es->s_first_error_line = cpu_to_le32(line);
317 es->s_first_error_ino = es->s_last_error_ino;
318 es->s_first_error_block = es->s_last_error_block;
321 * Start the daily error reporting function if it hasn't been
324 if (!es->s_error_count)
325 mod_timer(&EXT4_SB(sb)->s_err_report, jiffies + 24*60*60*HZ);
326 le32_add_cpu(&es->s_error_count, 1);
329 static void save_error_info(struct super_block *sb, const char *func,
332 __save_error_info(sb, func, line);
333 ext4_commit_super(sb, 1);
337 * The del_gendisk() function uninitializes the disk-specific data
338 * structures, including the bdi structure, without telling anyone
339 * else. Once this happens, any attempt to call mark_buffer_dirty()
340 * (for example, by ext4_commit_super), will cause a kernel OOPS.
341 * This is a kludge to prevent these oops until we can put in a proper
342 * hook in del_gendisk() to inform the VFS and file system layers.
344 static int block_device_ejected(struct super_block *sb)
346 struct inode *bd_inode = sb->s_bdev->bd_inode;
347 struct backing_dev_info *bdi = bd_inode->i_mapping->backing_dev_info;
349 return bdi->dev == NULL;
352 static void ext4_journal_commit_callback(journal_t *journal, transaction_t *txn)
354 struct super_block *sb = journal->j_private;
355 struct ext4_sb_info *sbi = EXT4_SB(sb);
356 int error = is_journal_aborted(journal);
357 struct ext4_journal_cb_entry *jce;
359 BUG_ON(txn->t_state == T_FINISHED);
360 spin_lock(&sbi->s_md_lock);
361 while (!list_empty(&txn->t_private_list)) {
362 jce = list_entry(txn->t_private_list.next,
363 struct ext4_journal_cb_entry, jce_list);
364 list_del_init(&jce->jce_list);
365 spin_unlock(&sbi->s_md_lock);
366 jce->jce_func(sb, jce, error);
367 spin_lock(&sbi->s_md_lock);
369 spin_unlock(&sbi->s_md_lock);
372 /* Deal with the reporting of failure conditions on a filesystem such as
373 * inconsistencies detected or read IO failures.
375 * On ext2, we can store the error state of the filesystem in the
376 * superblock. That is not possible on ext4, because we may have other
377 * write ordering constraints on the superblock which prevent us from
378 * writing it out straight away; and given that the journal is about to
379 * be aborted, we can't rely on the current, or future, transactions to
380 * write out the superblock safely.
382 * We'll just use the jbd2_journal_abort() error code to record an error in
383 * the journal instead. On recovery, the journal will complain about
384 * that error until we've noted it down and cleared it.
387 static void ext4_handle_error(struct super_block *sb)
389 if (sb->s_flags & MS_RDONLY)
392 if (!test_opt(sb, ERRORS_CONT)) {
393 journal_t *journal = EXT4_SB(sb)->s_journal;
395 EXT4_SB(sb)->s_mount_flags |= EXT4_MF_FS_ABORTED;
397 jbd2_journal_abort(journal, -EIO);
399 if (test_opt(sb, ERRORS_RO)) {
400 ext4_msg(sb, KERN_CRIT, "Remounting filesystem read-only");
401 sb->s_flags |= MS_RDONLY;
403 if (test_opt(sb, ERRORS_PANIC))
404 panic("EXT4-fs (device %s): panic forced after error\n",
408 void __ext4_error(struct super_block *sb, const char *function,
409 unsigned int line, const char *fmt, ...)
411 struct va_format vaf;
417 printk(KERN_CRIT "EXT4-fs error (device %s): %s:%d: comm %s: %pV\n",
418 sb->s_id, function, line, current->comm, &vaf);
420 save_error_info(sb, function, line);
422 ext4_handle_error(sb);
425 void ext4_error_inode(struct inode *inode, const char *function,
426 unsigned int line, ext4_fsblk_t block,
427 const char *fmt, ...)
430 struct va_format vaf;
431 struct ext4_super_block *es = EXT4_SB(inode->i_sb)->s_es;
433 es->s_last_error_ino = cpu_to_le32(inode->i_ino);
434 es->s_last_error_block = cpu_to_le64(block);
435 save_error_info(inode->i_sb, function, line);
440 printk(KERN_CRIT "EXT4-fs error (device %s): %s:%d: "
441 "inode #%lu: block %llu: comm %s: %pV\n",
442 inode->i_sb->s_id, function, line, inode->i_ino,
443 block, current->comm, &vaf);
445 printk(KERN_CRIT "EXT4-fs error (device %s): %s:%d: "
446 "inode #%lu: comm %s: %pV\n",
447 inode->i_sb->s_id, function, line, inode->i_ino,
448 current->comm, &vaf);
451 ext4_handle_error(inode->i_sb);
454 void ext4_error_file(struct file *file, const char *function,
455 unsigned int line, ext4_fsblk_t block,
456 const char *fmt, ...)
459 struct va_format vaf;
460 struct ext4_super_block *es;
461 struct inode *inode = file_inode(file);
462 char pathname[80], *path;
464 es = EXT4_SB(inode->i_sb)->s_es;
465 es->s_last_error_ino = cpu_to_le32(inode->i_ino);
466 save_error_info(inode->i_sb, function, line);
467 path = d_path(&(file->f_path), pathname, sizeof(pathname));
475 "EXT4-fs error (device %s): %s:%d: inode #%lu: "
476 "block %llu: comm %s: path %s: %pV\n",
477 inode->i_sb->s_id, function, line, inode->i_ino,
478 block, current->comm, path, &vaf);
481 "EXT4-fs error (device %s): %s:%d: inode #%lu: "
482 "comm %s: path %s: %pV\n",
483 inode->i_sb->s_id, function, line, inode->i_ino,
484 current->comm, path, &vaf);
487 ext4_handle_error(inode->i_sb);
490 const char *ext4_decode_error(struct super_block *sb, int errno,
497 errstr = "IO failure";
500 errstr = "Out of memory";
503 if (!sb || (EXT4_SB(sb)->s_journal &&
504 EXT4_SB(sb)->s_journal->j_flags & JBD2_ABORT))
505 errstr = "Journal has aborted";
507 errstr = "Readonly filesystem";
510 /* If the caller passed in an extra buffer for unknown
511 * errors, textualise them now. Else we just return
514 /* Check for truncated error codes... */
515 if (snprintf(nbuf, 16, "error %d", -errno) >= 0)
524 /* __ext4_std_error decodes expected errors from journaling functions
525 * automatically and invokes the appropriate error response. */
527 void __ext4_std_error(struct super_block *sb, const char *function,
528 unsigned int line, int errno)
533 /* Special case: if the error is EROFS, and we're not already
534 * inside a transaction, then there's really no point in logging
536 if (errno == -EROFS && journal_current_handle() == NULL &&
537 (sb->s_flags & MS_RDONLY))
540 errstr = ext4_decode_error(sb, errno, nbuf);
541 printk(KERN_CRIT "EXT4-fs error (device %s) in %s:%d: %s\n",
542 sb->s_id, function, line, errstr);
543 save_error_info(sb, function, line);
545 ext4_handle_error(sb);
549 * ext4_abort is a much stronger failure handler than ext4_error. The
550 * abort function may be used to deal with unrecoverable failures such
551 * as journal IO errors or ENOMEM at a critical moment in log management.
553 * We unconditionally force the filesystem into an ABORT|READONLY state,
554 * unless the error response on the fs has been set to panic in which
555 * case we take the easy way out and panic immediately.
558 void __ext4_abort(struct super_block *sb, const char *function,
559 unsigned int line, const char *fmt, ...)
563 save_error_info(sb, function, line);
565 printk(KERN_CRIT "EXT4-fs error (device %s): %s:%d: ", sb->s_id,
571 if ((sb->s_flags & MS_RDONLY) == 0) {
572 ext4_msg(sb, KERN_CRIT, "Remounting filesystem read-only");
573 sb->s_flags |= MS_RDONLY;
574 EXT4_SB(sb)->s_mount_flags |= EXT4_MF_FS_ABORTED;
575 if (EXT4_SB(sb)->s_journal)
576 jbd2_journal_abort(EXT4_SB(sb)->s_journal, -EIO);
577 save_error_info(sb, function, line);
579 if (test_opt(sb, ERRORS_PANIC))
580 panic("EXT4-fs panic from previous error\n");
583 void ext4_msg(struct super_block *sb, const char *prefix, const char *fmt, ...)
585 struct va_format vaf;
591 printk("%sEXT4-fs (%s): %pV\n", prefix, sb->s_id, &vaf);
595 void __ext4_warning(struct super_block *sb, const char *function,
596 unsigned int line, const char *fmt, ...)
598 struct va_format vaf;
604 printk(KERN_WARNING "EXT4-fs warning (device %s): %s:%d: %pV\n",
605 sb->s_id, function, line, &vaf);
609 void __ext4_grp_locked_error(const char *function, unsigned int line,
610 struct super_block *sb, ext4_group_t grp,
611 unsigned long ino, ext4_fsblk_t block,
612 const char *fmt, ...)
616 struct va_format vaf;
618 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
620 es->s_last_error_ino = cpu_to_le32(ino);
621 es->s_last_error_block = cpu_to_le64(block);
622 __save_error_info(sb, function, line);
628 printk(KERN_CRIT "EXT4-fs error (device %s): %s:%d: group %u, ",
629 sb->s_id, function, line, grp);
631 printk(KERN_CONT "inode %lu: ", ino);
633 printk(KERN_CONT "block %llu:", (unsigned long long) block);
634 printk(KERN_CONT "%pV\n", &vaf);
637 if (test_opt(sb, ERRORS_CONT)) {
638 ext4_commit_super(sb, 0);
642 ext4_unlock_group(sb, grp);
643 ext4_handle_error(sb);
645 * We only get here in the ERRORS_RO case; relocking the group
646 * may be dangerous, but nothing bad will happen since the
647 * filesystem will have already been marked read/only and the
648 * journal has been aborted. We return 1 as a hint to callers
649 * who might what to use the return value from
650 * ext4_grp_locked_error() to distinguish between the
651 * ERRORS_CONT and ERRORS_RO case, and perhaps return more
652 * aggressively from the ext4 function in question, with a
653 * more appropriate error code.
655 ext4_lock_group(sb, grp);
659 void ext4_update_dynamic_rev(struct super_block *sb)
661 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
663 if (le32_to_cpu(es->s_rev_level) > EXT4_GOOD_OLD_REV)
667 "updating to rev %d because of new feature flag, "
668 "running e2fsck is recommended",
671 es->s_first_ino = cpu_to_le32(EXT4_GOOD_OLD_FIRST_INO);
672 es->s_inode_size = cpu_to_le16(EXT4_GOOD_OLD_INODE_SIZE);
673 es->s_rev_level = cpu_to_le32(EXT4_DYNAMIC_REV);
674 /* leave es->s_feature_*compat flags alone */
675 /* es->s_uuid will be set by e2fsck if empty */
678 * The rest of the superblock fields should be zero, and if not it
679 * means they are likely already in use, so leave them alone. We
680 * can leave it up to e2fsck to clean up any inconsistencies there.
685 * Open the external journal device
687 static struct block_device *ext4_blkdev_get(dev_t dev, struct super_block *sb)
689 struct block_device *bdev;
690 char b[BDEVNAME_SIZE];
692 bdev = blkdev_get_by_dev(dev, FMODE_READ|FMODE_WRITE|FMODE_EXCL, sb);
698 ext4_msg(sb, KERN_ERR, "failed to open journal device %s: %ld",
699 __bdevname(dev, b), PTR_ERR(bdev));
704 * Release the journal device
706 static void ext4_blkdev_put(struct block_device *bdev)
708 blkdev_put(bdev, FMODE_READ|FMODE_WRITE|FMODE_EXCL);
711 static void ext4_blkdev_remove(struct ext4_sb_info *sbi)
713 struct block_device *bdev;
714 bdev = sbi->journal_bdev;
716 ext4_blkdev_put(bdev);
717 sbi->journal_bdev = NULL;
721 static inline struct inode *orphan_list_entry(struct list_head *l)
723 return &list_entry(l, struct ext4_inode_info, i_orphan)->vfs_inode;
726 static void dump_orphan_list(struct super_block *sb, struct ext4_sb_info *sbi)
730 ext4_msg(sb, KERN_ERR, "sb orphan head is %d",
731 le32_to_cpu(sbi->s_es->s_last_orphan));
733 printk(KERN_ERR "sb_info orphan list:\n");
734 list_for_each(l, &sbi->s_orphan) {
735 struct inode *inode = orphan_list_entry(l);
737 "inode %s:%lu at %p: mode %o, nlink %d, next %d\n",
738 inode->i_sb->s_id, inode->i_ino, inode,
739 inode->i_mode, inode->i_nlink,
744 static void ext4_put_super(struct super_block *sb)
746 struct ext4_sb_info *sbi = EXT4_SB(sb);
747 struct ext4_super_block *es = sbi->s_es;
750 ext4_unregister_li_request(sb);
751 dquot_disable(sb, -1, DQUOT_USAGE_ENABLED | DQUOT_LIMITS_ENABLED);
753 flush_workqueue(sbi->dio_unwritten_wq);
754 destroy_workqueue(sbi->dio_unwritten_wq);
756 if (sbi->s_journal) {
757 err = jbd2_journal_destroy(sbi->s_journal);
758 sbi->s_journal = NULL;
760 ext4_abort(sb, "Couldn't clean up the journal");
763 ext4_es_unregister_shrinker(sb);
764 del_timer(&sbi->s_err_report);
765 ext4_release_system_zone(sb);
767 ext4_ext_release(sb);
768 ext4_xattr_put_super(sb);
770 if (!(sb->s_flags & MS_RDONLY)) {
771 EXT4_CLEAR_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
772 es->s_state = cpu_to_le16(sbi->s_mount_state);
774 if (!(sb->s_flags & MS_RDONLY))
775 ext4_commit_super(sb, 1);
778 remove_proc_entry("options", sbi->s_proc);
779 remove_proc_entry(sb->s_id, ext4_proc_root);
781 kobject_del(&sbi->s_kobj);
783 for (i = 0; i < sbi->s_gdb_count; i++)
784 brelse(sbi->s_group_desc[i]);
785 ext4_kvfree(sbi->s_group_desc);
786 ext4_kvfree(sbi->s_flex_groups);
787 percpu_counter_destroy(&sbi->s_freeclusters_counter);
788 percpu_counter_destroy(&sbi->s_freeinodes_counter);
789 percpu_counter_destroy(&sbi->s_dirs_counter);
790 percpu_counter_destroy(&sbi->s_dirtyclusters_counter);
791 percpu_counter_destroy(&sbi->s_extent_cache_cnt);
794 for (i = 0; i < MAXQUOTAS; i++)
795 kfree(sbi->s_qf_names[i]);
798 /* Debugging code just in case the in-memory inode orphan list
799 * isn't empty. The on-disk one can be non-empty if we've
800 * detected an error and taken the fs readonly, but the
801 * in-memory list had better be clean by this point. */
802 if (!list_empty(&sbi->s_orphan))
803 dump_orphan_list(sb, sbi);
804 J_ASSERT(list_empty(&sbi->s_orphan));
806 invalidate_bdev(sb->s_bdev);
807 if (sbi->journal_bdev && sbi->journal_bdev != sb->s_bdev) {
809 * Invalidate the journal device's buffers. We don't want them
810 * floating about in memory - the physical journal device may
811 * hotswapped, and it breaks the `ro-after' testing code.
813 sync_blockdev(sbi->journal_bdev);
814 invalidate_bdev(sbi->journal_bdev);
815 ext4_blkdev_remove(sbi);
818 kthread_stop(sbi->s_mmp_tsk);
819 sb->s_fs_info = NULL;
821 * Now that we are completely done shutting down the
822 * superblock, we need to actually destroy the kobject.
824 kobject_put(&sbi->s_kobj);
825 wait_for_completion(&sbi->s_kobj_unregister);
826 if (sbi->s_chksum_driver)
827 crypto_free_shash(sbi->s_chksum_driver);
828 kfree(sbi->s_blockgroup_lock);
832 static struct kmem_cache *ext4_inode_cachep;
835 * Called inside transaction, so use GFP_NOFS
837 static struct inode *ext4_alloc_inode(struct super_block *sb)
839 struct ext4_inode_info *ei;
841 ei = kmem_cache_alloc(ext4_inode_cachep, GFP_NOFS);
845 ei->vfs_inode.i_version = 1;
846 INIT_LIST_HEAD(&ei->i_prealloc_list);
847 spin_lock_init(&ei->i_prealloc_lock);
848 ext4_es_init_tree(&ei->i_es_tree);
849 rwlock_init(&ei->i_es_lock);
850 INIT_LIST_HEAD(&ei->i_es_lru);
852 ei->i_reserved_data_blocks = 0;
853 ei->i_reserved_meta_blocks = 0;
854 ei->i_allocated_meta_blocks = 0;
855 ei->i_da_metadata_calc_len = 0;
856 ei->i_da_metadata_calc_last_lblock = 0;
857 spin_lock_init(&(ei->i_block_reservation_lock));
859 ei->i_reserved_quota = 0;
862 INIT_LIST_HEAD(&ei->i_completed_io_list);
863 spin_lock_init(&ei->i_completed_io_lock);
865 ei->i_datasync_tid = 0;
866 atomic_set(&ei->i_ioend_count, 0);
867 atomic_set(&ei->i_unwritten, 0);
868 INIT_WORK(&ei->i_unwritten_work, ext4_end_io_work);
870 return &ei->vfs_inode;
873 static int ext4_drop_inode(struct inode *inode)
875 int drop = generic_drop_inode(inode);
877 trace_ext4_drop_inode(inode, drop);
881 static void ext4_i_callback(struct rcu_head *head)
883 struct inode *inode = container_of(head, struct inode, i_rcu);
884 kmem_cache_free(ext4_inode_cachep, EXT4_I(inode));
887 static void ext4_destroy_inode(struct inode *inode)
889 if (!list_empty(&(EXT4_I(inode)->i_orphan))) {
890 ext4_msg(inode->i_sb, KERN_ERR,
891 "Inode %lu (%p): orphan list check failed!",
892 inode->i_ino, EXT4_I(inode));
893 print_hex_dump(KERN_INFO, "", DUMP_PREFIX_ADDRESS, 16, 4,
894 EXT4_I(inode), sizeof(struct ext4_inode_info),
898 call_rcu(&inode->i_rcu, ext4_i_callback);
901 static void init_once(void *foo)
903 struct ext4_inode_info *ei = (struct ext4_inode_info *) foo;
905 INIT_LIST_HEAD(&ei->i_orphan);
906 init_rwsem(&ei->xattr_sem);
907 init_rwsem(&ei->i_data_sem);
908 inode_init_once(&ei->vfs_inode);
911 static int init_inodecache(void)
913 ext4_inode_cachep = kmem_cache_create("ext4_inode_cache",
914 sizeof(struct ext4_inode_info),
915 0, (SLAB_RECLAIM_ACCOUNT|
918 if (ext4_inode_cachep == NULL)
923 static void destroy_inodecache(void)
926 * Make sure all delayed rcu free inodes are flushed before we
930 kmem_cache_destroy(ext4_inode_cachep);
933 void ext4_clear_inode(struct inode *inode)
935 invalidate_inode_buffers(inode);
938 ext4_discard_preallocations(inode);
939 ext4_es_remove_extent(inode, 0, EXT_MAX_BLOCKS);
940 ext4_es_lru_del(inode);
941 if (EXT4_I(inode)->jinode) {
942 jbd2_journal_release_jbd_inode(EXT4_JOURNAL(inode),
943 EXT4_I(inode)->jinode);
944 jbd2_free_inode(EXT4_I(inode)->jinode);
945 EXT4_I(inode)->jinode = NULL;
949 static struct inode *ext4_nfs_get_inode(struct super_block *sb,
950 u64 ino, u32 generation)
954 if (ino < EXT4_FIRST_INO(sb) && ino != EXT4_ROOT_INO)
955 return ERR_PTR(-ESTALE);
956 if (ino > le32_to_cpu(EXT4_SB(sb)->s_es->s_inodes_count))
957 return ERR_PTR(-ESTALE);
959 /* iget isn't really right if the inode is currently unallocated!!
961 * ext4_read_inode will return a bad_inode if the inode had been
962 * deleted, so we should be safe.
964 * Currently we don't know the generation for parent directory, so
965 * a generation of 0 means "accept any"
967 inode = ext4_iget(sb, ino);
969 return ERR_CAST(inode);
970 if (generation && inode->i_generation != generation) {
972 return ERR_PTR(-ESTALE);
978 static struct dentry *ext4_fh_to_dentry(struct super_block *sb, struct fid *fid,
979 int fh_len, int fh_type)
981 return generic_fh_to_dentry(sb, fid, fh_len, fh_type,
985 static struct dentry *ext4_fh_to_parent(struct super_block *sb, struct fid *fid,
986 int fh_len, int fh_type)
988 return generic_fh_to_parent(sb, fid, fh_len, fh_type,
993 * Try to release metadata pages (indirect blocks, directories) which are
994 * mapped via the block device. Since these pages could have journal heads
995 * which would prevent try_to_free_buffers() from freeing them, we must use
996 * jbd2 layer's try_to_free_buffers() function to release them.
998 static int bdev_try_to_free_page(struct super_block *sb, struct page *page,
1001 journal_t *journal = EXT4_SB(sb)->s_journal;
1003 WARN_ON(PageChecked(page));
1004 if (!page_has_buffers(page))
1007 return jbd2_journal_try_to_free_buffers(journal, page,
1008 wait & ~__GFP_WAIT);
1009 return try_to_free_buffers(page);
1013 #define QTYPE2NAME(t) ((t) == USRQUOTA ? "user" : "group")
1014 #define QTYPE2MOPT(on, t) ((t) == USRQUOTA?((on)##USRJQUOTA):((on)##GRPJQUOTA))
1016 static int ext4_write_dquot(struct dquot *dquot);
1017 static int ext4_acquire_dquot(struct dquot *dquot);
1018 static int ext4_release_dquot(struct dquot *dquot);
1019 static int ext4_mark_dquot_dirty(struct dquot *dquot);
1020 static int ext4_write_info(struct super_block *sb, int type);
1021 static int ext4_quota_on(struct super_block *sb, int type, int format_id,
1023 static int ext4_quota_on_sysfile(struct super_block *sb, int type,
1025 static int ext4_quota_off(struct super_block *sb, int type);
1026 static int ext4_quota_off_sysfile(struct super_block *sb, int type);
1027 static int ext4_quota_on_mount(struct super_block *sb, int type);
1028 static ssize_t ext4_quota_read(struct super_block *sb, int type, char *data,
1029 size_t len, loff_t off);
1030 static ssize_t ext4_quota_write(struct super_block *sb, int type,
1031 const char *data, size_t len, loff_t off);
1032 static int ext4_quota_enable(struct super_block *sb, int type, int format_id,
1033 unsigned int flags);
1034 static int ext4_enable_quotas(struct super_block *sb);
1036 static const struct dquot_operations ext4_quota_operations = {
1037 .get_reserved_space = ext4_get_reserved_space,
1038 .write_dquot = ext4_write_dquot,
1039 .acquire_dquot = ext4_acquire_dquot,
1040 .release_dquot = ext4_release_dquot,
1041 .mark_dirty = ext4_mark_dquot_dirty,
1042 .write_info = ext4_write_info,
1043 .alloc_dquot = dquot_alloc,
1044 .destroy_dquot = dquot_destroy,
1047 static const struct quotactl_ops ext4_qctl_operations = {
1048 .quota_on = ext4_quota_on,
1049 .quota_off = ext4_quota_off,
1050 .quota_sync = dquot_quota_sync,
1051 .get_info = dquot_get_dqinfo,
1052 .set_info = dquot_set_dqinfo,
1053 .get_dqblk = dquot_get_dqblk,
1054 .set_dqblk = dquot_set_dqblk
1057 static const struct quotactl_ops ext4_qctl_sysfile_operations = {
1058 .quota_on_meta = ext4_quota_on_sysfile,
1059 .quota_off = ext4_quota_off_sysfile,
1060 .quota_sync = dquot_quota_sync,
1061 .get_info = dquot_get_dqinfo,
1062 .set_info = dquot_set_dqinfo,
1063 .get_dqblk = dquot_get_dqblk,
1064 .set_dqblk = dquot_set_dqblk
1068 static const struct super_operations ext4_sops = {
1069 .alloc_inode = ext4_alloc_inode,
1070 .destroy_inode = ext4_destroy_inode,
1071 .write_inode = ext4_write_inode,
1072 .dirty_inode = ext4_dirty_inode,
1073 .drop_inode = ext4_drop_inode,
1074 .evict_inode = ext4_evict_inode,
1075 .put_super = ext4_put_super,
1076 .sync_fs = ext4_sync_fs,
1077 .freeze_fs = ext4_freeze,
1078 .unfreeze_fs = ext4_unfreeze,
1079 .statfs = ext4_statfs,
1080 .remount_fs = ext4_remount,
1081 .show_options = ext4_show_options,
1083 .quota_read = ext4_quota_read,
1084 .quota_write = ext4_quota_write,
1086 .bdev_try_to_free_page = bdev_try_to_free_page,
1089 static const struct super_operations ext4_nojournal_sops = {
1090 .alloc_inode = ext4_alloc_inode,
1091 .destroy_inode = ext4_destroy_inode,
1092 .write_inode = ext4_write_inode,
1093 .dirty_inode = ext4_dirty_inode,
1094 .drop_inode = ext4_drop_inode,
1095 .evict_inode = ext4_evict_inode,
1096 .put_super = ext4_put_super,
1097 .statfs = ext4_statfs,
1098 .remount_fs = ext4_remount,
1099 .show_options = ext4_show_options,
1101 .quota_read = ext4_quota_read,
1102 .quota_write = ext4_quota_write,
1104 .bdev_try_to_free_page = bdev_try_to_free_page,
1107 static const struct export_operations ext4_export_ops = {
1108 .fh_to_dentry = ext4_fh_to_dentry,
1109 .fh_to_parent = ext4_fh_to_parent,
1110 .get_parent = ext4_get_parent,
1114 Opt_bsd_df, Opt_minix_df, Opt_grpid, Opt_nogrpid,
1115 Opt_resgid, Opt_resuid, Opt_sb, Opt_err_cont, Opt_err_panic, Opt_err_ro,
1116 Opt_nouid32, Opt_debug, Opt_removed,
1117 Opt_user_xattr, Opt_nouser_xattr, Opt_acl, Opt_noacl,
1118 Opt_auto_da_alloc, Opt_noauto_da_alloc, Opt_noload,
1119 Opt_commit, Opt_min_batch_time, Opt_max_batch_time,
1120 Opt_journal_dev, Opt_journal_checksum, Opt_journal_async_commit,
1121 Opt_abort, Opt_data_journal, Opt_data_ordered, Opt_data_writeback,
1122 Opt_data_err_abort, Opt_data_err_ignore,
1123 Opt_usrjquota, Opt_grpjquota, Opt_offusrjquota, Opt_offgrpjquota,
1124 Opt_jqfmt_vfsold, Opt_jqfmt_vfsv0, Opt_jqfmt_vfsv1, Opt_quota,
1125 Opt_noquota, Opt_barrier, Opt_nobarrier, Opt_err,
1126 Opt_usrquota, Opt_grpquota, Opt_i_version,
1127 Opt_stripe, Opt_delalloc, Opt_nodelalloc, Opt_mblk_io_submit,
1128 Opt_nomblk_io_submit, Opt_block_validity, Opt_noblock_validity,
1129 Opt_inode_readahead_blks, Opt_journal_ioprio,
1130 Opt_dioread_nolock, Opt_dioread_lock,
1131 Opt_discard, Opt_nodiscard, Opt_init_itable, Opt_noinit_itable,
1132 Opt_max_dir_size_kb,
1135 static const match_table_t tokens = {
1136 {Opt_bsd_df, "bsddf"},
1137 {Opt_minix_df, "minixdf"},
1138 {Opt_grpid, "grpid"},
1139 {Opt_grpid, "bsdgroups"},
1140 {Opt_nogrpid, "nogrpid"},
1141 {Opt_nogrpid, "sysvgroups"},
1142 {Opt_resgid, "resgid=%u"},
1143 {Opt_resuid, "resuid=%u"},
1145 {Opt_err_cont, "errors=continue"},
1146 {Opt_err_panic, "errors=panic"},
1147 {Opt_err_ro, "errors=remount-ro"},
1148 {Opt_nouid32, "nouid32"},
1149 {Opt_debug, "debug"},
1150 {Opt_removed, "oldalloc"},
1151 {Opt_removed, "orlov"},
1152 {Opt_user_xattr, "user_xattr"},
1153 {Opt_nouser_xattr, "nouser_xattr"},
1155 {Opt_noacl, "noacl"},
1156 {Opt_noload, "norecovery"},
1157 {Opt_noload, "noload"},
1158 {Opt_removed, "nobh"},
1159 {Opt_removed, "bh"},
1160 {Opt_commit, "commit=%u"},
1161 {Opt_min_batch_time, "min_batch_time=%u"},
1162 {Opt_max_batch_time, "max_batch_time=%u"},
1163 {Opt_journal_dev, "journal_dev=%u"},
1164 {Opt_journal_checksum, "journal_checksum"},
1165 {Opt_journal_async_commit, "journal_async_commit"},
1166 {Opt_abort, "abort"},
1167 {Opt_data_journal, "data=journal"},
1168 {Opt_data_ordered, "data=ordered"},
1169 {Opt_data_writeback, "data=writeback"},
1170 {Opt_data_err_abort, "data_err=abort"},
1171 {Opt_data_err_ignore, "data_err=ignore"},
1172 {Opt_offusrjquota, "usrjquota="},
1173 {Opt_usrjquota, "usrjquota=%s"},
1174 {Opt_offgrpjquota, "grpjquota="},
1175 {Opt_grpjquota, "grpjquota=%s"},
1176 {Opt_jqfmt_vfsold, "jqfmt=vfsold"},
1177 {Opt_jqfmt_vfsv0, "jqfmt=vfsv0"},
1178 {Opt_jqfmt_vfsv1, "jqfmt=vfsv1"},
1179 {Opt_grpquota, "grpquota"},
1180 {Opt_noquota, "noquota"},
1181 {Opt_quota, "quota"},
1182 {Opt_usrquota, "usrquota"},
1183 {Opt_barrier, "barrier=%u"},
1184 {Opt_barrier, "barrier"},
1185 {Opt_nobarrier, "nobarrier"},
1186 {Opt_i_version, "i_version"},
1187 {Opt_stripe, "stripe=%u"},
1188 {Opt_delalloc, "delalloc"},
1189 {Opt_nodelalloc, "nodelalloc"},
1190 {Opt_removed, "mblk_io_submit"},
1191 {Opt_removed, "nomblk_io_submit"},
1192 {Opt_block_validity, "block_validity"},
1193 {Opt_noblock_validity, "noblock_validity"},
1194 {Opt_inode_readahead_blks, "inode_readahead_blks=%u"},
1195 {Opt_journal_ioprio, "journal_ioprio=%u"},
1196 {Opt_auto_da_alloc, "auto_da_alloc=%u"},
1197 {Opt_auto_da_alloc, "auto_da_alloc"},
1198 {Opt_noauto_da_alloc, "noauto_da_alloc"},
1199 {Opt_dioread_nolock, "dioread_nolock"},
1200 {Opt_dioread_lock, "dioread_lock"},
1201 {Opt_discard, "discard"},
1202 {Opt_nodiscard, "nodiscard"},
1203 {Opt_init_itable, "init_itable=%u"},
1204 {Opt_init_itable, "init_itable"},
1205 {Opt_noinit_itable, "noinit_itable"},
1206 {Opt_max_dir_size_kb, "max_dir_size_kb=%u"},
1207 {Opt_removed, "check=none"}, /* mount option from ext2/3 */
1208 {Opt_removed, "nocheck"}, /* mount option from ext2/3 */
1209 {Opt_removed, "reservation"}, /* mount option from ext2/3 */
1210 {Opt_removed, "noreservation"}, /* mount option from ext2/3 */
1211 {Opt_removed, "journal=%u"}, /* mount option from ext2/3 */
1215 static ext4_fsblk_t get_sb_block(void **data)
1217 ext4_fsblk_t sb_block;
1218 char *options = (char *) *data;
1220 if (!options || strncmp(options, "sb=", 3) != 0)
1221 return 1; /* Default location */
1224 /* TODO: use simple_strtoll with >32bit ext4 */
1225 sb_block = simple_strtoul(options, &options, 0);
1226 if (*options && *options != ',') {
1227 printk(KERN_ERR "EXT4-fs: Invalid sb specification: %s\n",
1231 if (*options == ',')
1233 *data = (void *) options;
1238 #define DEFAULT_JOURNAL_IOPRIO (IOPRIO_PRIO_VALUE(IOPRIO_CLASS_BE, 3))
1239 static char deprecated_msg[] = "Mount option \"%s\" will be removed by %s\n"
1240 "Contact linux-ext4@vger.kernel.org if you think we should keep it.\n";
1243 static int set_qf_name(struct super_block *sb, int qtype, substring_t *args)
1245 struct ext4_sb_info *sbi = EXT4_SB(sb);
1249 if (sb_any_quota_loaded(sb) &&
1250 !sbi->s_qf_names[qtype]) {
1251 ext4_msg(sb, KERN_ERR,
1252 "Cannot change journaled "
1253 "quota options when quota turned on");
1256 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_QUOTA)) {
1257 ext4_msg(sb, KERN_ERR, "Cannot set journaled quota options "
1258 "when QUOTA feature is enabled");
1261 qname = match_strdup(args);
1263 ext4_msg(sb, KERN_ERR,
1264 "Not enough memory for storing quotafile name");
1267 if (sbi->s_qf_names[qtype]) {
1268 if (strcmp(sbi->s_qf_names[qtype], qname) == 0)
1271 ext4_msg(sb, KERN_ERR,
1272 "%s quota file already specified",
1276 if (strchr(qname, '/')) {
1277 ext4_msg(sb, KERN_ERR,
1278 "quotafile must be on filesystem root");
1281 sbi->s_qf_names[qtype] = qname;
1289 static int clear_qf_name(struct super_block *sb, int qtype)
1292 struct ext4_sb_info *sbi = EXT4_SB(sb);
1294 if (sb_any_quota_loaded(sb) &&
1295 sbi->s_qf_names[qtype]) {
1296 ext4_msg(sb, KERN_ERR, "Cannot change journaled quota options"
1297 " when quota turned on");
1300 kfree(sbi->s_qf_names[qtype]);
1301 sbi->s_qf_names[qtype] = NULL;
1306 #define MOPT_SET 0x0001
1307 #define MOPT_CLEAR 0x0002
1308 #define MOPT_NOSUPPORT 0x0004
1309 #define MOPT_EXPLICIT 0x0008
1310 #define MOPT_CLEAR_ERR 0x0010
1311 #define MOPT_GTE0 0x0020
1314 #define MOPT_QFMT 0x0040
1316 #define MOPT_Q MOPT_NOSUPPORT
1317 #define MOPT_QFMT MOPT_NOSUPPORT
1319 #define MOPT_DATAJ 0x0080
1320 #define MOPT_NO_EXT2 0x0100
1321 #define MOPT_NO_EXT3 0x0200
1322 #define MOPT_EXT4_ONLY (MOPT_NO_EXT2 | MOPT_NO_EXT3)
1324 static const struct mount_opts {
1328 } ext4_mount_opts[] = {
1329 {Opt_minix_df, EXT4_MOUNT_MINIX_DF, MOPT_SET},
1330 {Opt_bsd_df, EXT4_MOUNT_MINIX_DF, MOPT_CLEAR},
1331 {Opt_grpid, EXT4_MOUNT_GRPID, MOPT_SET},
1332 {Opt_nogrpid, EXT4_MOUNT_GRPID, MOPT_CLEAR},
1333 {Opt_block_validity, EXT4_MOUNT_BLOCK_VALIDITY, MOPT_SET},
1334 {Opt_noblock_validity, EXT4_MOUNT_BLOCK_VALIDITY, MOPT_CLEAR},
1335 {Opt_dioread_nolock, EXT4_MOUNT_DIOREAD_NOLOCK,
1336 MOPT_EXT4_ONLY | MOPT_SET},
1337 {Opt_dioread_lock, EXT4_MOUNT_DIOREAD_NOLOCK,
1338 MOPT_EXT4_ONLY | MOPT_CLEAR},
1339 {Opt_discard, EXT4_MOUNT_DISCARD, MOPT_SET},
1340 {Opt_nodiscard, EXT4_MOUNT_DISCARD, MOPT_CLEAR},
1341 {Opt_delalloc, EXT4_MOUNT_DELALLOC,
1342 MOPT_EXT4_ONLY | MOPT_SET | MOPT_EXPLICIT},
1343 {Opt_nodelalloc, EXT4_MOUNT_DELALLOC,
1344 MOPT_EXT4_ONLY | MOPT_CLEAR | MOPT_EXPLICIT},
1345 {Opt_journal_checksum, EXT4_MOUNT_JOURNAL_CHECKSUM,
1346 MOPT_EXT4_ONLY | MOPT_SET},
1347 {Opt_journal_async_commit, (EXT4_MOUNT_JOURNAL_ASYNC_COMMIT |
1348 EXT4_MOUNT_JOURNAL_CHECKSUM),
1349 MOPT_EXT4_ONLY | MOPT_SET},
1350 {Opt_noload, EXT4_MOUNT_NOLOAD, MOPT_NO_EXT2 | MOPT_SET},
1351 {Opt_err_panic, EXT4_MOUNT_ERRORS_PANIC, MOPT_SET | MOPT_CLEAR_ERR},
1352 {Opt_err_ro, EXT4_MOUNT_ERRORS_RO, MOPT_SET | MOPT_CLEAR_ERR},
1353 {Opt_err_cont, EXT4_MOUNT_ERRORS_CONT, MOPT_SET | MOPT_CLEAR_ERR},
1354 {Opt_data_err_abort, EXT4_MOUNT_DATA_ERR_ABORT,
1355 MOPT_NO_EXT2 | MOPT_SET},
1356 {Opt_data_err_ignore, EXT4_MOUNT_DATA_ERR_ABORT,
1357 MOPT_NO_EXT2 | MOPT_CLEAR},
1358 {Opt_barrier, EXT4_MOUNT_BARRIER, MOPT_SET},
1359 {Opt_nobarrier, EXT4_MOUNT_BARRIER, MOPT_CLEAR},
1360 {Opt_noauto_da_alloc, EXT4_MOUNT_NO_AUTO_DA_ALLOC, MOPT_SET},
1361 {Opt_auto_da_alloc, EXT4_MOUNT_NO_AUTO_DA_ALLOC, MOPT_CLEAR},
1362 {Opt_noinit_itable, EXT4_MOUNT_INIT_INODE_TABLE, MOPT_CLEAR},
1363 {Opt_commit, 0, MOPT_GTE0},
1364 {Opt_max_batch_time, 0, MOPT_GTE0},
1365 {Opt_min_batch_time, 0, MOPT_GTE0},
1366 {Opt_inode_readahead_blks, 0, MOPT_GTE0},
1367 {Opt_init_itable, 0, MOPT_GTE0},
1368 {Opt_stripe, 0, MOPT_GTE0},
1369 {Opt_resuid, 0, MOPT_GTE0},
1370 {Opt_resgid, 0, MOPT_GTE0},
1371 {Opt_journal_dev, 0, MOPT_GTE0},
1372 {Opt_journal_ioprio, 0, MOPT_GTE0},
1373 {Opt_data_journal, EXT4_MOUNT_JOURNAL_DATA, MOPT_NO_EXT2 | MOPT_DATAJ},
1374 {Opt_data_ordered, EXT4_MOUNT_ORDERED_DATA, MOPT_NO_EXT2 | MOPT_DATAJ},
1375 {Opt_data_writeback, EXT4_MOUNT_WRITEBACK_DATA,
1376 MOPT_NO_EXT2 | MOPT_DATAJ},
1377 {Opt_user_xattr, EXT4_MOUNT_XATTR_USER, MOPT_SET},
1378 {Opt_nouser_xattr, EXT4_MOUNT_XATTR_USER, MOPT_CLEAR},
1379 #ifdef CONFIG_EXT4_FS_POSIX_ACL
1380 {Opt_acl, EXT4_MOUNT_POSIX_ACL, MOPT_SET},
1381 {Opt_noacl, EXT4_MOUNT_POSIX_ACL, MOPT_CLEAR},
1383 {Opt_acl, 0, MOPT_NOSUPPORT},
1384 {Opt_noacl, 0, MOPT_NOSUPPORT},
1386 {Opt_nouid32, EXT4_MOUNT_NO_UID32, MOPT_SET},
1387 {Opt_debug, EXT4_MOUNT_DEBUG, MOPT_SET},
1388 {Opt_quota, EXT4_MOUNT_QUOTA | EXT4_MOUNT_USRQUOTA, MOPT_SET | MOPT_Q},
1389 {Opt_usrquota, EXT4_MOUNT_QUOTA | EXT4_MOUNT_USRQUOTA,
1391 {Opt_grpquota, EXT4_MOUNT_QUOTA | EXT4_MOUNT_GRPQUOTA,
1393 {Opt_noquota, (EXT4_MOUNT_QUOTA | EXT4_MOUNT_USRQUOTA |
1394 EXT4_MOUNT_GRPQUOTA), MOPT_CLEAR | MOPT_Q},
1395 {Opt_usrjquota, 0, MOPT_Q},
1396 {Opt_grpjquota, 0, MOPT_Q},
1397 {Opt_offusrjquota, 0, MOPT_Q},
1398 {Opt_offgrpjquota, 0, MOPT_Q},
1399 {Opt_jqfmt_vfsold, QFMT_VFS_OLD, MOPT_QFMT},
1400 {Opt_jqfmt_vfsv0, QFMT_VFS_V0, MOPT_QFMT},
1401 {Opt_jqfmt_vfsv1, QFMT_VFS_V1, MOPT_QFMT},
1402 {Opt_max_dir_size_kb, 0, MOPT_GTE0},
1406 static int handle_mount_opt(struct super_block *sb, char *opt, int token,
1407 substring_t *args, unsigned long *journal_devnum,
1408 unsigned int *journal_ioprio, int is_remount)
1410 struct ext4_sb_info *sbi = EXT4_SB(sb);
1411 const struct mount_opts *m;
1417 if (token == Opt_usrjquota)
1418 return set_qf_name(sb, USRQUOTA, &args[0]);
1419 else if (token == Opt_grpjquota)
1420 return set_qf_name(sb, GRPQUOTA, &args[0]);
1421 else if (token == Opt_offusrjquota)
1422 return clear_qf_name(sb, USRQUOTA);
1423 else if (token == Opt_offgrpjquota)
1424 return clear_qf_name(sb, GRPQUOTA);
1428 case Opt_nouser_xattr:
1429 ext4_msg(sb, KERN_WARNING, deprecated_msg, opt, "3.5");
1432 return 1; /* handled by get_sb_block() */
1434 ext4_msg(sb, KERN_WARNING, "Ignoring removed %s option", opt);
1437 sbi->s_mount_flags |= EXT4_MF_FS_ABORTED;
1440 sb->s_flags |= MS_I_VERSION;
1444 for (m = ext4_mount_opts; m->token != Opt_err; m++)
1445 if (token == m->token)
1448 if (m->token == Opt_err) {
1449 ext4_msg(sb, KERN_ERR, "Unrecognized mount option \"%s\" "
1450 "or missing value", opt);
1454 if ((m->flags & MOPT_NO_EXT2) && IS_EXT2_SB(sb)) {
1455 ext4_msg(sb, KERN_ERR,
1456 "Mount option \"%s\" incompatible with ext2", opt);
1459 if ((m->flags & MOPT_NO_EXT3) && IS_EXT3_SB(sb)) {
1460 ext4_msg(sb, KERN_ERR,
1461 "Mount option \"%s\" incompatible with ext3", opt);
1465 if (args->from && match_int(args, &arg))
1467 if (args->from && (m->flags & MOPT_GTE0) && (arg < 0))
1469 if (m->flags & MOPT_EXPLICIT)
1470 set_opt2(sb, EXPLICIT_DELALLOC);
1471 if (m->flags & MOPT_CLEAR_ERR)
1472 clear_opt(sb, ERRORS_MASK);
1473 if (token == Opt_noquota && sb_any_quota_loaded(sb)) {
1474 ext4_msg(sb, KERN_ERR, "Cannot change quota "
1475 "options when quota turned on");
1479 if (m->flags & MOPT_NOSUPPORT) {
1480 ext4_msg(sb, KERN_ERR, "%s option not supported", opt);
1481 } else if (token == Opt_commit) {
1483 arg = JBD2_DEFAULT_MAX_COMMIT_AGE;
1484 sbi->s_commit_interval = HZ * arg;
1485 } else if (token == Opt_max_batch_time) {
1487 arg = EXT4_DEF_MAX_BATCH_TIME;
1488 sbi->s_max_batch_time = arg;
1489 } else if (token == Opt_min_batch_time) {
1490 sbi->s_min_batch_time = arg;
1491 } else if (token == Opt_inode_readahead_blks) {
1492 if (arg && (arg > (1 << 30) || !is_power_of_2(arg))) {
1493 ext4_msg(sb, KERN_ERR,
1494 "EXT4-fs: inode_readahead_blks must be "
1495 "0 or a power of 2 smaller than 2^31");
1498 sbi->s_inode_readahead_blks = arg;
1499 } else if (token == Opt_init_itable) {
1500 set_opt(sb, INIT_INODE_TABLE);
1502 arg = EXT4_DEF_LI_WAIT_MULT;
1503 sbi->s_li_wait_mult = arg;
1504 } else if (token == Opt_max_dir_size_kb) {
1505 sbi->s_max_dir_size_kb = arg;
1506 } else if (token == Opt_stripe) {
1507 sbi->s_stripe = arg;
1508 } else if (token == Opt_resuid) {
1509 uid = make_kuid(current_user_ns(), arg);
1510 if (!uid_valid(uid)) {
1511 ext4_msg(sb, KERN_ERR, "Invalid uid value %d", arg);
1514 sbi->s_resuid = uid;
1515 } else if (token == Opt_resgid) {
1516 gid = make_kgid(current_user_ns(), arg);
1517 if (!gid_valid(gid)) {
1518 ext4_msg(sb, KERN_ERR, "Invalid gid value %d", arg);
1521 sbi->s_resgid = gid;
1522 } else if (token == Opt_journal_dev) {
1524 ext4_msg(sb, KERN_ERR,
1525 "Cannot specify journal on remount");
1528 *journal_devnum = arg;
1529 } else if (token == Opt_journal_ioprio) {
1531 ext4_msg(sb, KERN_ERR, "Invalid journal IO priority"
1536 IOPRIO_PRIO_VALUE(IOPRIO_CLASS_BE, arg);
1537 } else if (m->flags & MOPT_DATAJ) {
1539 if (!sbi->s_journal)
1540 ext4_msg(sb, KERN_WARNING, "Remounting file system with no journal so ignoring journalled data option");
1541 else if (test_opt(sb, DATA_FLAGS) != m->mount_opt) {
1542 ext4_msg(sb, KERN_ERR,
1543 "Cannot change data mode on remount");
1547 clear_opt(sb, DATA_FLAGS);
1548 sbi->s_mount_opt |= m->mount_opt;
1551 } else if (m->flags & MOPT_QFMT) {
1552 if (sb_any_quota_loaded(sb) &&
1553 sbi->s_jquota_fmt != m->mount_opt) {
1554 ext4_msg(sb, KERN_ERR, "Cannot change journaled "
1555 "quota options when quota turned on");
1558 if (EXT4_HAS_RO_COMPAT_FEATURE(sb,
1559 EXT4_FEATURE_RO_COMPAT_QUOTA)) {
1560 ext4_msg(sb, KERN_ERR,
1561 "Cannot set journaled quota options "
1562 "when QUOTA feature is enabled");
1565 sbi->s_jquota_fmt = m->mount_opt;
1570 if (m->flags & MOPT_CLEAR)
1572 else if (unlikely(!(m->flags & MOPT_SET))) {
1573 ext4_msg(sb, KERN_WARNING,
1574 "buggy handling of option %s", opt);
1579 sbi->s_mount_opt |= m->mount_opt;
1581 sbi->s_mount_opt &= ~m->mount_opt;
1586 static int parse_options(char *options, struct super_block *sb,
1587 unsigned long *journal_devnum,
1588 unsigned int *journal_ioprio,
1591 struct ext4_sb_info *sbi = EXT4_SB(sb);
1593 substring_t args[MAX_OPT_ARGS];
1599 while ((p = strsep(&options, ",")) != NULL) {
1603 * Initialize args struct so we know whether arg was
1604 * found; some options take optional arguments.
1606 args[0].to = args[0].from = NULL;
1607 token = match_token(p, tokens, args);
1608 if (handle_mount_opt(sb, p, token, args, journal_devnum,
1609 journal_ioprio, is_remount) < 0)
1613 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_QUOTA) &&
1614 (test_opt(sb, USRQUOTA) || test_opt(sb, GRPQUOTA))) {
1615 ext4_msg(sb, KERN_ERR, "Cannot set quota options when QUOTA "
1616 "feature is enabled");
1619 if (sbi->s_qf_names[USRQUOTA] || sbi->s_qf_names[GRPQUOTA]) {
1620 if (test_opt(sb, USRQUOTA) && sbi->s_qf_names[USRQUOTA])
1621 clear_opt(sb, USRQUOTA);
1623 if (test_opt(sb, GRPQUOTA) && sbi->s_qf_names[GRPQUOTA])
1624 clear_opt(sb, GRPQUOTA);
1626 if (test_opt(sb, GRPQUOTA) || test_opt(sb, USRQUOTA)) {
1627 ext4_msg(sb, KERN_ERR, "old and new quota "
1632 if (!sbi->s_jquota_fmt) {
1633 ext4_msg(sb, KERN_ERR, "journaled quota format "
1638 if (sbi->s_jquota_fmt) {
1639 ext4_msg(sb, KERN_ERR, "journaled quota format "
1640 "specified with no journaling "
1646 if (test_opt(sb, DIOREAD_NOLOCK)) {
1648 BLOCK_SIZE << le32_to_cpu(sbi->s_es->s_log_block_size);
1650 if (blocksize < PAGE_CACHE_SIZE) {
1651 ext4_msg(sb, KERN_ERR, "can't mount with "
1652 "dioread_nolock if block size != PAGE_SIZE");
1659 static inline void ext4_show_quota_options(struct seq_file *seq,
1660 struct super_block *sb)
1662 #if defined(CONFIG_QUOTA)
1663 struct ext4_sb_info *sbi = EXT4_SB(sb);
1665 if (sbi->s_jquota_fmt) {
1668 switch (sbi->s_jquota_fmt) {
1679 seq_printf(seq, ",jqfmt=%s", fmtname);
1682 if (sbi->s_qf_names[USRQUOTA])
1683 seq_printf(seq, ",usrjquota=%s", sbi->s_qf_names[USRQUOTA]);
1685 if (sbi->s_qf_names[GRPQUOTA])
1686 seq_printf(seq, ",grpjquota=%s", sbi->s_qf_names[GRPQUOTA]);
1688 if (test_opt(sb, USRQUOTA))
1689 seq_puts(seq, ",usrquota");
1691 if (test_opt(sb, GRPQUOTA))
1692 seq_puts(seq, ",grpquota");
1696 static const char *token2str(int token)
1698 const struct match_token *t;
1700 for (t = tokens; t->token != Opt_err; t++)
1701 if (t->token == token && !strchr(t->pattern, '='))
1708 * - it's set to a non-default value OR
1709 * - if the per-sb default is different from the global default
1711 static int _ext4_show_options(struct seq_file *seq, struct super_block *sb,
1714 struct ext4_sb_info *sbi = EXT4_SB(sb);
1715 struct ext4_super_block *es = sbi->s_es;
1716 int def_errors, def_mount_opt = nodefs ? 0 : sbi->s_def_mount_opt;
1717 const struct mount_opts *m;
1718 char sep = nodefs ? '\n' : ',';
1720 #define SEQ_OPTS_PUTS(str) seq_printf(seq, "%c" str, sep)
1721 #define SEQ_OPTS_PRINT(str, arg) seq_printf(seq, "%c" str, sep, arg)
1723 if (sbi->s_sb_block != 1)
1724 SEQ_OPTS_PRINT("sb=%llu", sbi->s_sb_block);
1726 for (m = ext4_mount_opts; m->token != Opt_err; m++) {
1727 int want_set = m->flags & MOPT_SET;
1728 if (((m->flags & (MOPT_SET|MOPT_CLEAR)) == 0) ||
1729 (m->flags & MOPT_CLEAR_ERR))
1731 if (!(m->mount_opt & (sbi->s_mount_opt ^ def_mount_opt)))
1732 continue; /* skip if same as the default */
1734 (sbi->s_mount_opt & m->mount_opt) != m->mount_opt) ||
1735 (!want_set && (sbi->s_mount_opt & m->mount_opt)))
1736 continue; /* select Opt_noFoo vs Opt_Foo */
1737 SEQ_OPTS_PRINT("%s", token2str(m->token));
1740 if (nodefs || !uid_eq(sbi->s_resuid, make_kuid(&init_user_ns, EXT4_DEF_RESUID)) ||
1741 le16_to_cpu(es->s_def_resuid) != EXT4_DEF_RESUID)
1742 SEQ_OPTS_PRINT("resuid=%u",
1743 from_kuid_munged(&init_user_ns, sbi->s_resuid));
1744 if (nodefs || !gid_eq(sbi->s_resgid, make_kgid(&init_user_ns, EXT4_DEF_RESGID)) ||
1745 le16_to_cpu(es->s_def_resgid) != EXT4_DEF_RESGID)
1746 SEQ_OPTS_PRINT("resgid=%u",
1747 from_kgid_munged(&init_user_ns, sbi->s_resgid));
1748 def_errors = nodefs ? -1 : le16_to_cpu(es->s_errors);
1749 if (test_opt(sb, ERRORS_RO) && def_errors != EXT4_ERRORS_RO)
1750 SEQ_OPTS_PUTS("errors=remount-ro");
1751 if (test_opt(sb, ERRORS_CONT) && def_errors != EXT4_ERRORS_CONTINUE)
1752 SEQ_OPTS_PUTS("errors=continue");
1753 if (test_opt(sb, ERRORS_PANIC) && def_errors != EXT4_ERRORS_PANIC)
1754 SEQ_OPTS_PUTS("errors=panic");
1755 if (nodefs || sbi->s_commit_interval != JBD2_DEFAULT_MAX_COMMIT_AGE*HZ)
1756 SEQ_OPTS_PRINT("commit=%lu", sbi->s_commit_interval / HZ);
1757 if (nodefs || sbi->s_min_batch_time != EXT4_DEF_MIN_BATCH_TIME)
1758 SEQ_OPTS_PRINT("min_batch_time=%u", sbi->s_min_batch_time);
1759 if (nodefs || sbi->s_max_batch_time != EXT4_DEF_MAX_BATCH_TIME)
1760 SEQ_OPTS_PRINT("max_batch_time=%u", sbi->s_max_batch_time);
1761 if (sb->s_flags & MS_I_VERSION)
1762 SEQ_OPTS_PUTS("i_version");
1763 if (nodefs || sbi->s_stripe)
1764 SEQ_OPTS_PRINT("stripe=%lu", sbi->s_stripe);
1765 if (EXT4_MOUNT_DATA_FLAGS & (sbi->s_mount_opt ^ def_mount_opt)) {
1766 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA)
1767 SEQ_OPTS_PUTS("data=journal");
1768 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA)
1769 SEQ_OPTS_PUTS("data=ordered");
1770 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_WRITEBACK_DATA)
1771 SEQ_OPTS_PUTS("data=writeback");
1774 sbi->s_inode_readahead_blks != EXT4_DEF_INODE_READAHEAD_BLKS)
1775 SEQ_OPTS_PRINT("inode_readahead_blks=%u",
1776 sbi->s_inode_readahead_blks);
1778 if (nodefs || (test_opt(sb, INIT_INODE_TABLE) &&
1779 (sbi->s_li_wait_mult != EXT4_DEF_LI_WAIT_MULT)))
1780 SEQ_OPTS_PRINT("init_itable=%u", sbi->s_li_wait_mult);
1781 if (nodefs || sbi->s_max_dir_size_kb)
1782 SEQ_OPTS_PRINT("max_dir_size_kb=%u", sbi->s_max_dir_size_kb);
1784 ext4_show_quota_options(seq, sb);
1788 static int ext4_show_options(struct seq_file *seq, struct dentry *root)
1790 return _ext4_show_options(seq, root->d_sb, 0);
1793 static int options_seq_show(struct seq_file *seq, void *offset)
1795 struct super_block *sb = seq->private;
1798 seq_puts(seq, (sb->s_flags & MS_RDONLY) ? "ro" : "rw");
1799 rc = _ext4_show_options(seq, sb, 1);
1800 seq_puts(seq, "\n");
1804 static int options_open_fs(struct inode *inode, struct file *file)
1806 return single_open(file, options_seq_show, PDE_DATA(inode));
1809 static const struct file_operations ext4_seq_options_fops = {
1810 .owner = THIS_MODULE,
1811 .open = options_open_fs,
1813 .llseek = seq_lseek,
1814 .release = single_release,
1817 static int ext4_setup_super(struct super_block *sb, struct ext4_super_block *es,
1820 struct ext4_sb_info *sbi = EXT4_SB(sb);
1823 if (le32_to_cpu(es->s_rev_level) > EXT4_MAX_SUPP_REV) {
1824 ext4_msg(sb, KERN_ERR, "revision level too high, "
1825 "forcing read-only mode");
1830 if (!(sbi->s_mount_state & EXT4_VALID_FS))
1831 ext4_msg(sb, KERN_WARNING, "warning: mounting unchecked fs, "
1832 "running e2fsck is recommended");
1833 else if ((sbi->s_mount_state & EXT4_ERROR_FS))
1834 ext4_msg(sb, KERN_WARNING,
1835 "warning: mounting fs with errors, "
1836 "running e2fsck is recommended");
1837 else if ((__s16) le16_to_cpu(es->s_max_mnt_count) > 0 &&
1838 le16_to_cpu(es->s_mnt_count) >=
1839 (unsigned short) (__s16) le16_to_cpu(es->s_max_mnt_count))
1840 ext4_msg(sb, KERN_WARNING,
1841 "warning: maximal mount count reached, "
1842 "running e2fsck is recommended");
1843 else if (le32_to_cpu(es->s_checkinterval) &&
1844 (le32_to_cpu(es->s_lastcheck) +
1845 le32_to_cpu(es->s_checkinterval) <= get_seconds()))
1846 ext4_msg(sb, KERN_WARNING,
1847 "warning: checktime reached, "
1848 "running e2fsck is recommended");
1849 if (!sbi->s_journal)
1850 es->s_state &= cpu_to_le16(~EXT4_VALID_FS);
1851 if (!(__s16) le16_to_cpu(es->s_max_mnt_count))
1852 es->s_max_mnt_count = cpu_to_le16(EXT4_DFL_MAX_MNT_COUNT);
1853 le16_add_cpu(&es->s_mnt_count, 1);
1854 es->s_mtime = cpu_to_le32(get_seconds());
1855 ext4_update_dynamic_rev(sb);
1857 EXT4_SET_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
1859 ext4_commit_super(sb, 1);
1861 if (test_opt(sb, DEBUG))
1862 printk(KERN_INFO "[EXT4 FS bs=%lu, gc=%u, "
1863 "bpg=%lu, ipg=%lu, mo=%04x, mo2=%04x]\n",
1865 sbi->s_groups_count,
1866 EXT4_BLOCKS_PER_GROUP(sb),
1867 EXT4_INODES_PER_GROUP(sb),
1868 sbi->s_mount_opt, sbi->s_mount_opt2);
1870 cleancache_init_fs(sb);
1874 int ext4_alloc_flex_bg_array(struct super_block *sb, ext4_group_t ngroup)
1876 struct ext4_sb_info *sbi = EXT4_SB(sb);
1877 struct flex_groups *new_groups;
1880 if (!sbi->s_log_groups_per_flex)
1883 size = ext4_flex_group(sbi, ngroup - 1) + 1;
1884 if (size <= sbi->s_flex_groups_allocated)
1887 size = roundup_pow_of_two(size * sizeof(struct flex_groups));
1888 new_groups = ext4_kvzalloc(size, GFP_KERNEL);
1890 ext4_msg(sb, KERN_ERR, "not enough memory for %d flex groups",
1891 size / (int) sizeof(struct flex_groups));
1895 if (sbi->s_flex_groups) {
1896 memcpy(new_groups, sbi->s_flex_groups,
1897 (sbi->s_flex_groups_allocated *
1898 sizeof(struct flex_groups)));
1899 ext4_kvfree(sbi->s_flex_groups);
1901 sbi->s_flex_groups = new_groups;
1902 sbi->s_flex_groups_allocated = size / sizeof(struct flex_groups);
1906 static int ext4_fill_flex_info(struct super_block *sb)
1908 struct ext4_sb_info *sbi = EXT4_SB(sb);
1909 struct ext4_group_desc *gdp = NULL;
1910 ext4_group_t flex_group;
1911 unsigned int groups_per_flex = 0;
1914 sbi->s_log_groups_per_flex = sbi->s_es->s_log_groups_per_flex;
1915 if (sbi->s_log_groups_per_flex < 1 || sbi->s_log_groups_per_flex > 31) {
1916 sbi->s_log_groups_per_flex = 0;
1919 groups_per_flex = 1U << sbi->s_log_groups_per_flex;
1921 err = ext4_alloc_flex_bg_array(sb, sbi->s_groups_count);
1925 for (i = 0; i < sbi->s_groups_count; i++) {
1926 gdp = ext4_get_group_desc(sb, i, NULL);
1928 flex_group = ext4_flex_group(sbi, i);
1929 atomic_add(ext4_free_inodes_count(sb, gdp),
1930 &sbi->s_flex_groups[flex_group].free_inodes);
1931 atomic64_add(ext4_free_group_clusters(sb, gdp),
1932 &sbi->s_flex_groups[flex_group].free_clusters);
1933 atomic_add(ext4_used_dirs_count(sb, gdp),
1934 &sbi->s_flex_groups[flex_group].used_dirs);
1942 static __le16 ext4_group_desc_csum(struct ext4_sb_info *sbi, __u32 block_group,
1943 struct ext4_group_desc *gdp)
1947 __le32 le_group = cpu_to_le32(block_group);
1949 if ((sbi->s_es->s_feature_ro_compat &
1950 cpu_to_le32(EXT4_FEATURE_RO_COMPAT_METADATA_CSUM))) {
1951 /* Use new metadata_csum algorithm */
1955 save_csum = gdp->bg_checksum;
1956 gdp->bg_checksum = 0;
1957 csum32 = ext4_chksum(sbi, sbi->s_csum_seed, (__u8 *)&le_group,
1959 csum32 = ext4_chksum(sbi, csum32, (__u8 *)gdp,
1961 gdp->bg_checksum = save_csum;
1963 crc = csum32 & 0xFFFF;
1967 /* old crc16 code */
1968 offset = offsetof(struct ext4_group_desc, bg_checksum);
1970 crc = crc16(~0, sbi->s_es->s_uuid, sizeof(sbi->s_es->s_uuid));
1971 crc = crc16(crc, (__u8 *)&le_group, sizeof(le_group));
1972 crc = crc16(crc, (__u8 *)gdp, offset);
1973 offset += sizeof(gdp->bg_checksum); /* skip checksum */
1974 /* for checksum of struct ext4_group_desc do the rest...*/
1975 if ((sbi->s_es->s_feature_incompat &
1976 cpu_to_le32(EXT4_FEATURE_INCOMPAT_64BIT)) &&
1977 offset < le16_to_cpu(sbi->s_es->s_desc_size))
1978 crc = crc16(crc, (__u8 *)gdp + offset,
1979 le16_to_cpu(sbi->s_es->s_desc_size) -
1983 return cpu_to_le16(crc);
1986 int ext4_group_desc_csum_verify(struct super_block *sb, __u32 block_group,
1987 struct ext4_group_desc *gdp)
1989 if (ext4_has_group_desc_csum(sb) &&
1990 (gdp->bg_checksum != ext4_group_desc_csum(EXT4_SB(sb),
1997 void ext4_group_desc_csum_set(struct super_block *sb, __u32 block_group,
1998 struct ext4_group_desc *gdp)
2000 if (!ext4_has_group_desc_csum(sb))
2002 gdp->bg_checksum = ext4_group_desc_csum(EXT4_SB(sb), block_group, gdp);
2005 /* Called at mount-time, super-block is locked */
2006 static int ext4_check_descriptors(struct super_block *sb,
2007 ext4_group_t *first_not_zeroed)
2009 struct ext4_sb_info *sbi = EXT4_SB(sb);
2010 ext4_fsblk_t first_block = le32_to_cpu(sbi->s_es->s_first_data_block);
2011 ext4_fsblk_t last_block;
2012 ext4_fsblk_t block_bitmap;
2013 ext4_fsblk_t inode_bitmap;
2014 ext4_fsblk_t inode_table;
2015 int flexbg_flag = 0;
2016 ext4_group_t i, grp = sbi->s_groups_count;
2018 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_FLEX_BG))
2021 ext4_debug("Checking group descriptors");
2023 for (i = 0; i < sbi->s_groups_count; i++) {
2024 struct ext4_group_desc *gdp = ext4_get_group_desc(sb, i, NULL);
2026 if (i == sbi->s_groups_count - 1 || flexbg_flag)
2027 last_block = ext4_blocks_count(sbi->s_es) - 1;
2029 last_block = first_block +
2030 (EXT4_BLOCKS_PER_GROUP(sb) - 1);
2032 if ((grp == sbi->s_groups_count) &&
2033 !(gdp->bg_flags & cpu_to_le16(EXT4_BG_INODE_ZEROED)))
2036 block_bitmap = ext4_block_bitmap(sb, gdp);
2037 if (block_bitmap < first_block || block_bitmap > last_block) {
2038 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2039 "Block bitmap for group %u not in group "
2040 "(block %llu)!", i, block_bitmap);
2043 inode_bitmap = ext4_inode_bitmap(sb, gdp);
2044 if (inode_bitmap < first_block || inode_bitmap > last_block) {
2045 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2046 "Inode bitmap for group %u not in group "
2047 "(block %llu)!", i, inode_bitmap);
2050 inode_table = ext4_inode_table(sb, gdp);
2051 if (inode_table < first_block ||
2052 inode_table + sbi->s_itb_per_group - 1 > last_block) {
2053 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2054 "Inode table for group %u not in group "
2055 "(block %llu)!", i, inode_table);
2058 ext4_lock_group(sb, i);
2059 if (!ext4_group_desc_csum_verify(sb, i, gdp)) {
2060 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2061 "Checksum for group %u failed (%u!=%u)",
2062 i, le16_to_cpu(ext4_group_desc_csum(sbi, i,
2063 gdp)), le16_to_cpu(gdp->bg_checksum));
2064 if (!(sb->s_flags & MS_RDONLY)) {
2065 ext4_unlock_group(sb, i);
2069 ext4_unlock_group(sb, i);
2071 first_block += EXT4_BLOCKS_PER_GROUP(sb);
2073 if (NULL != first_not_zeroed)
2074 *first_not_zeroed = grp;
2076 ext4_free_blocks_count_set(sbi->s_es,
2077 EXT4_C2B(sbi, ext4_count_free_clusters(sb)));
2078 sbi->s_es->s_free_inodes_count =cpu_to_le32(ext4_count_free_inodes(sb));
2082 /* ext4_orphan_cleanup() walks a singly-linked list of inodes (starting at
2083 * the superblock) which were deleted from all directories, but held open by
2084 * a process at the time of a crash. We walk the list and try to delete these
2085 * inodes at recovery time (only with a read-write filesystem).
2087 * In order to keep the orphan inode chain consistent during traversal (in
2088 * case of crash during recovery), we link each inode into the superblock
2089 * orphan list_head and handle it the same way as an inode deletion during
2090 * normal operation (which journals the operations for us).
2092 * We only do an iget() and an iput() on each inode, which is very safe if we
2093 * accidentally point at an in-use or already deleted inode. The worst that
2094 * can happen in this case is that we get a "bit already cleared" message from
2095 * ext4_free_inode(). The only reason we would point at a wrong inode is if
2096 * e2fsck was run on this filesystem, and it must have already done the orphan
2097 * inode cleanup for us, so we can safely abort without any further action.
2099 static void ext4_orphan_cleanup(struct super_block *sb,
2100 struct ext4_super_block *es)
2102 unsigned int s_flags = sb->s_flags;
2103 int nr_orphans = 0, nr_truncates = 0;
2107 if (!es->s_last_orphan) {
2108 jbd_debug(4, "no orphan inodes to clean up\n");
2112 if (bdev_read_only(sb->s_bdev)) {
2113 ext4_msg(sb, KERN_ERR, "write access "
2114 "unavailable, skipping orphan cleanup");
2118 /* Check if feature set would not allow a r/w mount */
2119 if (!ext4_feature_set_ok(sb, 0)) {
2120 ext4_msg(sb, KERN_INFO, "Skipping orphan cleanup due to "
2121 "unknown ROCOMPAT features");
2125 if (EXT4_SB(sb)->s_mount_state & EXT4_ERROR_FS) {
2126 /* don't clear list on RO mount w/ errors */
2127 if (es->s_last_orphan && !(s_flags & MS_RDONLY)) {
2128 jbd_debug(1, "Errors on filesystem, "
2129 "clearing orphan list.\n");
2130 es->s_last_orphan = 0;
2132 jbd_debug(1, "Skipping orphan recovery on fs with errors.\n");
2136 if (s_flags & MS_RDONLY) {
2137 ext4_msg(sb, KERN_INFO, "orphan cleanup on readonly fs");
2138 sb->s_flags &= ~MS_RDONLY;
2141 /* Needed for iput() to work correctly and not trash data */
2142 sb->s_flags |= MS_ACTIVE;
2143 /* Turn on quotas so that they are updated correctly */
2144 for (i = 0; i < MAXQUOTAS; i++) {
2145 if (EXT4_SB(sb)->s_qf_names[i]) {
2146 int ret = ext4_quota_on_mount(sb, i);
2148 ext4_msg(sb, KERN_ERR,
2149 "Cannot turn on journaled "
2150 "quota: error %d", ret);
2155 while (es->s_last_orphan) {
2156 struct inode *inode;
2158 inode = ext4_orphan_get(sb, le32_to_cpu(es->s_last_orphan));
2159 if (IS_ERR(inode)) {
2160 es->s_last_orphan = 0;
2164 list_add(&EXT4_I(inode)->i_orphan, &EXT4_SB(sb)->s_orphan);
2165 dquot_initialize(inode);
2166 if (inode->i_nlink) {
2167 if (test_opt(sb, DEBUG))
2168 ext4_msg(sb, KERN_DEBUG,
2169 "%s: truncating inode %lu to %lld bytes",
2170 __func__, inode->i_ino, inode->i_size);
2171 jbd_debug(2, "truncating inode %lu to %lld bytes\n",
2172 inode->i_ino, inode->i_size);
2173 mutex_lock(&inode->i_mutex);
2174 truncate_inode_pages(inode->i_mapping, inode->i_size);
2175 ext4_truncate(inode);
2176 mutex_unlock(&inode->i_mutex);
2179 if (test_opt(sb, DEBUG))
2180 ext4_msg(sb, KERN_DEBUG,
2181 "%s: deleting unreferenced inode %lu",
2182 __func__, inode->i_ino);
2183 jbd_debug(2, "deleting unreferenced inode %lu\n",
2187 iput(inode); /* The delete magic happens here! */
2190 #define PLURAL(x) (x), ((x) == 1) ? "" : "s"
2193 ext4_msg(sb, KERN_INFO, "%d orphan inode%s deleted",
2194 PLURAL(nr_orphans));
2196 ext4_msg(sb, KERN_INFO, "%d truncate%s cleaned up",
2197 PLURAL(nr_truncates));
2199 /* Turn quotas off */
2200 for (i = 0; i < MAXQUOTAS; i++) {
2201 if (sb_dqopt(sb)->files[i])
2202 dquot_quota_off(sb, i);
2205 sb->s_flags = s_flags; /* Restore MS_RDONLY status */
2209 * Maximal extent format file size.
2210 * Resulting logical blkno at s_maxbytes must fit in our on-disk
2211 * extent format containers, within a sector_t, and within i_blocks
2212 * in the vfs. ext4 inode has 48 bits of i_block in fsblock units,
2213 * so that won't be a limiting factor.
2215 * However there is other limiting factor. We do store extents in the form
2216 * of starting block and length, hence the resulting length of the extent
2217 * covering maximum file size must fit into on-disk format containers as
2218 * well. Given that length is always by 1 unit bigger than max unit (because
2219 * we count 0 as well) we have to lower the s_maxbytes by one fs block.
2221 * Note, this does *not* consider any metadata overhead for vfs i_blocks.
2223 static loff_t ext4_max_size(int blkbits, int has_huge_files)
2226 loff_t upper_limit = MAX_LFS_FILESIZE;
2228 /* small i_blocks in vfs inode? */
2229 if (!has_huge_files || sizeof(blkcnt_t) < sizeof(u64)) {
2231 * CONFIG_LBDAF is not enabled implies the inode
2232 * i_block represent total blocks in 512 bytes
2233 * 32 == size of vfs inode i_blocks * 8
2235 upper_limit = (1LL << 32) - 1;
2237 /* total blocks in file system block size */
2238 upper_limit >>= (blkbits - 9);
2239 upper_limit <<= blkbits;
2243 * 32-bit extent-start container, ee_block. We lower the maxbytes
2244 * by one fs block, so ee_len can cover the extent of maximum file
2247 res = (1LL << 32) - 1;
2250 /* Sanity check against vm- & vfs- imposed limits */
2251 if (res > upper_limit)
2258 * Maximal bitmap file size. There is a direct, and {,double-,triple-}indirect
2259 * block limit, and also a limit of (2^48 - 1) 512-byte sectors in i_blocks.
2260 * We need to be 1 filesystem block less than the 2^48 sector limit.
2262 static loff_t ext4_max_bitmap_size(int bits, int has_huge_files)
2264 loff_t res = EXT4_NDIR_BLOCKS;
2267 /* This is calculated to be the largest file size for a dense, block
2268 * mapped file such that the file's total number of 512-byte sectors,
2269 * including data and all indirect blocks, does not exceed (2^48 - 1).
2271 * __u32 i_blocks_lo and _u16 i_blocks_high represent the total
2272 * number of 512-byte sectors of the file.
2275 if (!has_huge_files || sizeof(blkcnt_t) < sizeof(u64)) {
2277 * !has_huge_files or CONFIG_LBDAF not enabled implies that
2278 * the inode i_block field represents total file blocks in
2279 * 2^32 512-byte sectors == size of vfs inode i_blocks * 8
2281 upper_limit = (1LL << 32) - 1;
2283 /* total blocks in file system block size */
2284 upper_limit >>= (bits - 9);
2288 * We use 48 bit ext4_inode i_blocks
2289 * With EXT4_HUGE_FILE_FL set the i_blocks
2290 * represent total number of blocks in
2291 * file system block size
2293 upper_limit = (1LL << 48) - 1;
2297 /* indirect blocks */
2299 /* double indirect blocks */
2300 meta_blocks += 1 + (1LL << (bits-2));
2301 /* tripple indirect blocks */
2302 meta_blocks += 1 + (1LL << (bits-2)) + (1LL << (2*(bits-2)));
2304 upper_limit -= meta_blocks;
2305 upper_limit <<= bits;
2307 res += 1LL << (bits-2);
2308 res += 1LL << (2*(bits-2));
2309 res += 1LL << (3*(bits-2));
2311 if (res > upper_limit)
2314 if (res > MAX_LFS_FILESIZE)
2315 res = MAX_LFS_FILESIZE;
2320 static ext4_fsblk_t descriptor_loc(struct super_block *sb,
2321 ext4_fsblk_t logical_sb_block, int nr)
2323 struct ext4_sb_info *sbi = EXT4_SB(sb);
2324 ext4_group_t bg, first_meta_bg;
2327 first_meta_bg = le32_to_cpu(sbi->s_es->s_first_meta_bg);
2329 if (!EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_META_BG) ||
2331 return logical_sb_block + nr + 1;
2332 bg = sbi->s_desc_per_block * nr;
2333 if (ext4_bg_has_super(sb, bg))
2336 return (has_super + ext4_group_first_block_no(sb, bg));
2340 * ext4_get_stripe_size: Get the stripe size.
2341 * @sbi: In memory super block info
2343 * If we have specified it via mount option, then
2344 * use the mount option value. If the value specified at mount time is
2345 * greater than the blocks per group use the super block value.
2346 * If the super block value is greater than blocks per group return 0.
2347 * Allocator needs it be less than blocks per group.
2350 static unsigned long ext4_get_stripe_size(struct ext4_sb_info *sbi)
2352 unsigned long stride = le16_to_cpu(sbi->s_es->s_raid_stride);
2353 unsigned long stripe_width =
2354 le32_to_cpu(sbi->s_es->s_raid_stripe_width);
2357 if (sbi->s_stripe && sbi->s_stripe <= sbi->s_blocks_per_group)
2358 ret = sbi->s_stripe;
2359 else if (stripe_width <= sbi->s_blocks_per_group)
2361 else if (stride <= sbi->s_blocks_per_group)
2367 * If the stripe width is 1, this makes no sense and
2368 * we set it to 0 to turn off stripe handling code.
2379 struct attribute attr;
2380 ssize_t (*show)(struct ext4_attr *, struct ext4_sb_info *, char *);
2381 ssize_t (*store)(struct ext4_attr *, struct ext4_sb_info *,
2382 const char *, size_t);
2389 static int parse_strtoull(const char *buf,
2390 unsigned long long max, unsigned long long *value)
2394 ret = kstrtoull(skip_spaces(buf), 0, value);
2395 if (!ret && *value > max)
2400 static ssize_t delayed_allocation_blocks_show(struct ext4_attr *a,
2401 struct ext4_sb_info *sbi,
2404 return snprintf(buf, PAGE_SIZE, "%llu\n",
2406 percpu_counter_sum(&sbi->s_dirtyclusters_counter)));
2409 static ssize_t session_write_kbytes_show(struct ext4_attr *a,
2410 struct ext4_sb_info *sbi, char *buf)
2412 struct super_block *sb = sbi->s_buddy_cache->i_sb;
2414 if (!sb->s_bdev->bd_part)
2415 return snprintf(buf, PAGE_SIZE, "0\n");
2416 return snprintf(buf, PAGE_SIZE, "%lu\n",
2417 (part_stat_read(sb->s_bdev->bd_part, sectors[1]) -
2418 sbi->s_sectors_written_start) >> 1);
2421 static ssize_t lifetime_write_kbytes_show(struct ext4_attr *a,
2422 struct ext4_sb_info *sbi, char *buf)
2424 struct super_block *sb = sbi->s_buddy_cache->i_sb;
2426 if (!sb->s_bdev->bd_part)
2427 return snprintf(buf, PAGE_SIZE, "0\n");
2428 return snprintf(buf, PAGE_SIZE, "%llu\n",
2429 (unsigned long long)(sbi->s_kbytes_written +
2430 ((part_stat_read(sb->s_bdev->bd_part, sectors[1]) -
2431 EXT4_SB(sb)->s_sectors_written_start) >> 1)));
2434 static ssize_t inode_readahead_blks_store(struct ext4_attr *a,
2435 struct ext4_sb_info *sbi,
2436 const char *buf, size_t count)
2441 ret = kstrtoul(skip_spaces(buf), 0, &t);
2445 if (t && (!is_power_of_2(t) || t > 0x40000000))
2448 sbi->s_inode_readahead_blks = t;
2452 static ssize_t sbi_ui_show(struct ext4_attr *a,
2453 struct ext4_sb_info *sbi, char *buf)
2455 unsigned int *ui = (unsigned int *) (((char *) sbi) + a->u.offset);
2457 return snprintf(buf, PAGE_SIZE, "%u\n", *ui);
2460 static ssize_t sbi_ui_store(struct ext4_attr *a,
2461 struct ext4_sb_info *sbi,
2462 const char *buf, size_t count)
2464 unsigned int *ui = (unsigned int *) (((char *) sbi) + a->u.offset);
2468 ret = kstrtoul(skip_spaces(buf), 0, &t);
2475 static ssize_t reserved_clusters_show(struct ext4_attr *a,
2476 struct ext4_sb_info *sbi, char *buf)
2478 return snprintf(buf, PAGE_SIZE, "%llu\n",
2479 (unsigned long long) atomic64_read(&sbi->s_resv_clusters));
2482 static ssize_t reserved_clusters_store(struct ext4_attr *a,
2483 struct ext4_sb_info *sbi,
2484 const char *buf, size_t count)
2486 unsigned long long val;
2489 if (parse_strtoull(buf, -1ULL, &val))
2491 ret = ext4_reserve_clusters(sbi, val);
2493 return ret ? ret : count;
2496 static ssize_t trigger_test_error(struct ext4_attr *a,
2497 struct ext4_sb_info *sbi,
2498 const char *buf, size_t count)
2502 if (!capable(CAP_SYS_ADMIN))
2505 if (len && buf[len-1] == '\n')
2509 ext4_error(sbi->s_sb, "%.*s", len, buf);
2513 static ssize_t sbi_deprecated_show(struct ext4_attr *a,
2514 struct ext4_sb_info *sbi, char *buf)
2516 return snprintf(buf, PAGE_SIZE, "%d\n", a->u.deprecated_val);
2519 #define EXT4_ATTR_OFFSET(_name,_mode,_show,_store,_elname) \
2520 static struct ext4_attr ext4_attr_##_name = { \
2521 .attr = {.name = __stringify(_name), .mode = _mode }, \
2525 .offset = offsetof(struct ext4_sb_info, _elname),\
2528 #define EXT4_ATTR(name, mode, show, store) \
2529 static struct ext4_attr ext4_attr_##name = __ATTR(name, mode, show, store)
2531 #define EXT4_INFO_ATTR(name) EXT4_ATTR(name, 0444, NULL, NULL)
2532 #define EXT4_RO_ATTR(name) EXT4_ATTR(name, 0444, name##_show, NULL)
2533 #define EXT4_RW_ATTR(name) EXT4_ATTR(name, 0644, name##_show, name##_store)
2534 #define EXT4_RW_ATTR_SBI_UI(name, elname) \
2535 EXT4_ATTR_OFFSET(name, 0644, sbi_ui_show, sbi_ui_store, elname)
2536 #define ATTR_LIST(name) &ext4_attr_##name.attr
2537 #define EXT4_DEPRECATED_ATTR(_name, _val) \
2538 static struct ext4_attr ext4_attr_##_name = { \
2539 .attr = {.name = __stringify(_name), .mode = 0444 }, \
2540 .show = sbi_deprecated_show, \
2542 .deprecated_val = _val, \
2546 EXT4_RO_ATTR(delayed_allocation_blocks);
2547 EXT4_RO_ATTR(session_write_kbytes);
2548 EXT4_RO_ATTR(lifetime_write_kbytes);
2549 EXT4_RW_ATTR(reserved_clusters);
2550 EXT4_ATTR_OFFSET(inode_readahead_blks, 0644, sbi_ui_show,
2551 inode_readahead_blks_store, s_inode_readahead_blks);
2552 EXT4_RW_ATTR_SBI_UI(inode_goal, s_inode_goal);
2553 EXT4_RW_ATTR_SBI_UI(mb_stats, s_mb_stats);
2554 EXT4_RW_ATTR_SBI_UI(mb_max_to_scan, s_mb_max_to_scan);
2555 EXT4_RW_ATTR_SBI_UI(mb_min_to_scan, s_mb_min_to_scan);
2556 EXT4_RW_ATTR_SBI_UI(mb_order2_req, s_mb_order2_reqs);
2557 EXT4_RW_ATTR_SBI_UI(mb_stream_req, s_mb_stream_request);
2558 EXT4_RW_ATTR_SBI_UI(mb_group_prealloc, s_mb_group_prealloc);
2559 EXT4_DEPRECATED_ATTR(max_writeback_mb_bump, 128);
2560 EXT4_RW_ATTR_SBI_UI(extent_max_zeroout_kb, s_extent_max_zeroout_kb);
2561 EXT4_ATTR(trigger_fs_error, 0200, NULL, trigger_test_error);
2563 static struct attribute *ext4_attrs[] = {
2564 ATTR_LIST(delayed_allocation_blocks),
2565 ATTR_LIST(session_write_kbytes),
2566 ATTR_LIST(lifetime_write_kbytes),
2567 ATTR_LIST(reserved_clusters),
2568 ATTR_LIST(inode_readahead_blks),
2569 ATTR_LIST(inode_goal),
2570 ATTR_LIST(mb_stats),
2571 ATTR_LIST(mb_max_to_scan),
2572 ATTR_LIST(mb_min_to_scan),
2573 ATTR_LIST(mb_order2_req),
2574 ATTR_LIST(mb_stream_req),
2575 ATTR_LIST(mb_group_prealloc),
2576 ATTR_LIST(max_writeback_mb_bump),
2577 ATTR_LIST(extent_max_zeroout_kb),
2578 ATTR_LIST(trigger_fs_error),
2582 /* Features this copy of ext4 supports */
2583 EXT4_INFO_ATTR(lazy_itable_init);
2584 EXT4_INFO_ATTR(batched_discard);
2585 EXT4_INFO_ATTR(meta_bg_resize);
2587 static struct attribute *ext4_feat_attrs[] = {
2588 ATTR_LIST(lazy_itable_init),
2589 ATTR_LIST(batched_discard),
2590 ATTR_LIST(meta_bg_resize),
2594 static ssize_t ext4_attr_show(struct kobject *kobj,
2595 struct attribute *attr, char *buf)
2597 struct ext4_sb_info *sbi = container_of(kobj, struct ext4_sb_info,
2599 struct ext4_attr *a = container_of(attr, struct ext4_attr, attr);
2601 return a->show ? a->show(a, sbi, buf) : 0;
2604 static ssize_t ext4_attr_store(struct kobject *kobj,
2605 struct attribute *attr,
2606 const char *buf, size_t len)
2608 struct ext4_sb_info *sbi = container_of(kobj, struct ext4_sb_info,
2610 struct ext4_attr *a = container_of(attr, struct ext4_attr, attr);
2612 return a->store ? a->store(a, sbi, buf, len) : 0;
2615 static void ext4_sb_release(struct kobject *kobj)
2617 struct ext4_sb_info *sbi = container_of(kobj, struct ext4_sb_info,
2619 complete(&sbi->s_kobj_unregister);
2622 static const struct sysfs_ops ext4_attr_ops = {
2623 .show = ext4_attr_show,
2624 .store = ext4_attr_store,
2627 static struct kobj_type ext4_ktype = {
2628 .default_attrs = ext4_attrs,
2629 .sysfs_ops = &ext4_attr_ops,
2630 .release = ext4_sb_release,
2633 static void ext4_feat_release(struct kobject *kobj)
2635 complete(&ext4_feat->f_kobj_unregister);
2638 static struct kobj_type ext4_feat_ktype = {
2639 .default_attrs = ext4_feat_attrs,
2640 .sysfs_ops = &ext4_attr_ops,
2641 .release = ext4_feat_release,
2645 * Check whether this filesystem can be mounted based on
2646 * the features present and the RDONLY/RDWR mount requested.
2647 * Returns 1 if this filesystem can be mounted as requested,
2648 * 0 if it cannot be.
2650 static int ext4_feature_set_ok(struct super_block *sb, int readonly)
2652 if (EXT4_HAS_INCOMPAT_FEATURE(sb, ~EXT4_FEATURE_INCOMPAT_SUPP)) {
2653 ext4_msg(sb, KERN_ERR,
2654 "Couldn't mount because of "
2655 "unsupported optional features (%x)",
2656 (le32_to_cpu(EXT4_SB(sb)->s_es->s_feature_incompat) &
2657 ~EXT4_FEATURE_INCOMPAT_SUPP));
2664 /* Check that feature set is OK for a read-write mount */
2665 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, ~EXT4_FEATURE_RO_COMPAT_SUPP)) {
2666 ext4_msg(sb, KERN_ERR, "couldn't mount RDWR because of "
2667 "unsupported optional features (%x)",
2668 (le32_to_cpu(EXT4_SB(sb)->s_es->s_feature_ro_compat) &
2669 ~EXT4_FEATURE_RO_COMPAT_SUPP));
2673 * Large file size enabled file system can only be mounted
2674 * read-write on 32-bit systems if kernel is built with CONFIG_LBDAF
2676 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_HUGE_FILE)) {
2677 if (sizeof(blkcnt_t) < sizeof(u64)) {
2678 ext4_msg(sb, KERN_ERR, "Filesystem with huge files "
2679 "cannot be mounted RDWR without "
2684 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_BIGALLOC) &&
2685 !EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_EXTENTS)) {
2686 ext4_msg(sb, KERN_ERR,
2687 "Can't support bigalloc feature without "
2688 "extents feature\n");
2692 #ifndef CONFIG_QUOTA
2693 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_QUOTA) &&
2695 ext4_msg(sb, KERN_ERR,
2696 "Filesystem with quota feature cannot be mounted RDWR "
2697 "without CONFIG_QUOTA");
2700 #endif /* CONFIG_QUOTA */
2705 * This function is called once a day if we have errors logged
2706 * on the file system
2708 static void print_daily_error_info(unsigned long arg)
2710 struct super_block *sb = (struct super_block *) arg;
2711 struct ext4_sb_info *sbi;
2712 struct ext4_super_block *es;
2717 if (es->s_error_count)
2718 ext4_msg(sb, KERN_NOTICE, "error count: %u",
2719 le32_to_cpu(es->s_error_count));
2720 if (es->s_first_error_time) {
2721 printk(KERN_NOTICE "EXT4-fs (%s): initial error at %u: %.*s:%d",
2722 sb->s_id, le32_to_cpu(es->s_first_error_time),
2723 (int) sizeof(es->s_first_error_func),
2724 es->s_first_error_func,
2725 le32_to_cpu(es->s_first_error_line));
2726 if (es->s_first_error_ino)
2727 printk(": inode %u",
2728 le32_to_cpu(es->s_first_error_ino));
2729 if (es->s_first_error_block)
2730 printk(": block %llu", (unsigned long long)
2731 le64_to_cpu(es->s_first_error_block));
2734 if (es->s_last_error_time) {
2735 printk(KERN_NOTICE "EXT4-fs (%s): last error at %u: %.*s:%d",
2736 sb->s_id, le32_to_cpu(es->s_last_error_time),
2737 (int) sizeof(es->s_last_error_func),
2738 es->s_last_error_func,
2739 le32_to_cpu(es->s_last_error_line));
2740 if (es->s_last_error_ino)
2741 printk(": inode %u",
2742 le32_to_cpu(es->s_last_error_ino));
2743 if (es->s_last_error_block)
2744 printk(": block %llu", (unsigned long long)
2745 le64_to_cpu(es->s_last_error_block));
2748 mod_timer(&sbi->s_err_report, jiffies + 24*60*60*HZ); /* Once a day */
2751 /* Find next suitable group and run ext4_init_inode_table */
2752 static int ext4_run_li_request(struct ext4_li_request *elr)
2754 struct ext4_group_desc *gdp = NULL;
2755 ext4_group_t group, ngroups;
2756 struct super_block *sb;
2757 unsigned long timeout = 0;
2761 ngroups = EXT4_SB(sb)->s_groups_count;
2764 for (group = elr->lr_next_group; group < ngroups; group++) {
2765 gdp = ext4_get_group_desc(sb, group, NULL);
2771 if (!(gdp->bg_flags & cpu_to_le16(EXT4_BG_INODE_ZEROED)))
2775 if (group >= ngroups)
2780 ret = ext4_init_inode_table(sb, group,
2781 elr->lr_timeout ? 0 : 1);
2782 if (elr->lr_timeout == 0) {
2783 timeout = (jiffies - timeout) *
2784 elr->lr_sbi->s_li_wait_mult;
2785 elr->lr_timeout = timeout;
2787 elr->lr_next_sched = jiffies + elr->lr_timeout;
2788 elr->lr_next_group = group + 1;
2796 * Remove lr_request from the list_request and free the
2797 * request structure. Should be called with li_list_mtx held
2799 static void ext4_remove_li_request(struct ext4_li_request *elr)
2801 struct ext4_sb_info *sbi;
2808 list_del(&elr->lr_request);
2809 sbi->s_li_request = NULL;
2813 static void ext4_unregister_li_request(struct super_block *sb)
2815 mutex_lock(&ext4_li_mtx);
2816 if (!ext4_li_info) {
2817 mutex_unlock(&ext4_li_mtx);
2821 mutex_lock(&ext4_li_info->li_list_mtx);
2822 ext4_remove_li_request(EXT4_SB(sb)->s_li_request);
2823 mutex_unlock(&ext4_li_info->li_list_mtx);
2824 mutex_unlock(&ext4_li_mtx);
2827 static struct task_struct *ext4_lazyinit_task;
2830 * This is the function where ext4lazyinit thread lives. It walks
2831 * through the request list searching for next scheduled filesystem.
2832 * When such a fs is found, run the lazy initialization request
2833 * (ext4_rn_li_request) and keep track of the time spend in this
2834 * function. Based on that time we compute next schedule time of
2835 * the request. When walking through the list is complete, compute
2836 * next waking time and put itself into sleep.
2838 static int ext4_lazyinit_thread(void *arg)
2840 struct ext4_lazy_init *eli = (struct ext4_lazy_init *)arg;
2841 struct list_head *pos, *n;
2842 struct ext4_li_request *elr;
2843 unsigned long next_wakeup, cur;
2845 BUG_ON(NULL == eli);
2849 next_wakeup = MAX_JIFFY_OFFSET;
2851 mutex_lock(&eli->li_list_mtx);
2852 if (list_empty(&eli->li_request_list)) {
2853 mutex_unlock(&eli->li_list_mtx);
2857 list_for_each_safe(pos, n, &eli->li_request_list) {
2858 elr = list_entry(pos, struct ext4_li_request,
2861 if (time_after_eq(jiffies, elr->lr_next_sched)) {
2862 if (ext4_run_li_request(elr) != 0) {
2863 /* error, remove the lazy_init job */
2864 ext4_remove_li_request(elr);
2869 if (time_before(elr->lr_next_sched, next_wakeup))
2870 next_wakeup = elr->lr_next_sched;
2872 mutex_unlock(&eli->li_list_mtx);
2877 if ((time_after_eq(cur, next_wakeup)) ||
2878 (MAX_JIFFY_OFFSET == next_wakeup)) {
2883 schedule_timeout_interruptible(next_wakeup - cur);
2885 if (kthread_should_stop()) {
2886 ext4_clear_request_list();
2893 * It looks like the request list is empty, but we need
2894 * to check it under the li_list_mtx lock, to prevent any
2895 * additions into it, and of course we should lock ext4_li_mtx
2896 * to atomically free the list and ext4_li_info, because at
2897 * this point another ext4 filesystem could be registering
2900 mutex_lock(&ext4_li_mtx);
2901 mutex_lock(&eli->li_list_mtx);
2902 if (!list_empty(&eli->li_request_list)) {
2903 mutex_unlock(&eli->li_list_mtx);
2904 mutex_unlock(&ext4_li_mtx);
2907 mutex_unlock(&eli->li_list_mtx);
2908 kfree(ext4_li_info);
2909 ext4_li_info = NULL;
2910 mutex_unlock(&ext4_li_mtx);
2915 static void ext4_clear_request_list(void)
2917 struct list_head *pos, *n;
2918 struct ext4_li_request *elr;
2920 mutex_lock(&ext4_li_info->li_list_mtx);
2921 list_for_each_safe(pos, n, &ext4_li_info->li_request_list) {
2922 elr = list_entry(pos, struct ext4_li_request,
2924 ext4_remove_li_request(elr);
2926 mutex_unlock(&ext4_li_info->li_list_mtx);
2929 static int ext4_run_lazyinit_thread(void)
2931 ext4_lazyinit_task = kthread_run(ext4_lazyinit_thread,
2932 ext4_li_info, "ext4lazyinit");
2933 if (IS_ERR(ext4_lazyinit_task)) {
2934 int err = PTR_ERR(ext4_lazyinit_task);
2935 ext4_clear_request_list();
2936 kfree(ext4_li_info);
2937 ext4_li_info = NULL;
2938 printk(KERN_CRIT "EXT4-fs: error %d creating inode table "
2939 "initialization thread\n",
2943 ext4_li_info->li_state |= EXT4_LAZYINIT_RUNNING;
2948 * Check whether it make sense to run itable init. thread or not.
2949 * If there is at least one uninitialized inode table, return
2950 * corresponding group number, else the loop goes through all
2951 * groups and return total number of groups.
2953 static ext4_group_t ext4_has_uninit_itable(struct super_block *sb)
2955 ext4_group_t group, ngroups = EXT4_SB(sb)->s_groups_count;
2956 struct ext4_group_desc *gdp = NULL;
2958 for (group = 0; group < ngroups; group++) {
2959 gdp = ext4_get_group_desc(sb, group, NULL);
2963 if (!(gdp->bg_flags & cpu_to_le16(EXT4_BG_INODE_ZEROED)))
2970 static int ext4_li_info_new(void)
2972 struct ext4_lazy_init *eli = NULL;
2974 eli = kzalloc(sizeof(*eli), GFP_KERNEL);
2978 INIT_LIST_HEAD(&eli->li_request_list);
2979 mutex_init(&eli->li_list_mtx);
2981 eli->li_state |= EXT4_LAZYINIT_QUIT;
2988 static struct ext4_li_request *ext4_li_request_new(struct super_block *sb,
2991 struct ext4_sb_info *sbi = EXT4_SB(sb);
2992 struct ext4_li_request *elr;
2995 elr = kzalloc(sizeof(*elr), GFP_KERNEL);
3001 elr->lr_next_group = start;
3004 * Randomize first schedule time of the request to
3005 * spread the inode table initialization requests
3008 get_random_bytes(&rnd, sizeof(rnd));
3009 elr->lr_next_sched = jiffies + (unsigned long)rnd %
3010 (EXT4_DEF_LI_MAX_START_DELAY * HZ);
3015 int ext4_register_li_request(struct super_block *sb,
3016 ext4_group_t first_not_zeroed)
3018 struct ext4_sb_info *sbi = EXT4_SB(sb);
3019 struct ext4_li_request *elr = NULL;
3020 ext4_group_t ngroups = EXT4_SB(sb)->s_groups_count;
3023 mutex_lock(&ext4_li_mtx);
3024 if (sbi->s_li_request != NULL) {
3026 * Reset timeout so it can be computed again, because
3027 * s_li_wait_mult might have changed.
3029 sbi->s_li_request->lr_timeout = 0;
3033 if (first_not_zeroed == ngroups ||
3034 (sb->s_flags & MS_RDONLY) ||
3035 !test_opt(sb, INIT_INODE_TABLE))
3038 elr = ext4_li_request_new(sb, first_not_zeroed);
3044 if (NULL == ext4_li_info) {
3045 ret = ext4_li_info_new();
3050 mutex_lock(&ext4_li_info->li_list_mtx);
3051 list_add(&elr->lr_request, &ext4_li_info->li_request_list);
3052 mutex_unlock(&ext4_li_info->li_list_mtx);
3054 sbi->s_li_request = elr;
3056 * set elr to NULL here since it has been inserted to
3057 * the request_list and the removal and free of it is
3058 * handled by ext4_clear_request_list from now on.
3062 if (!(ext4_li_info->li_state & EXT4_LAZYINIT_RUNNING)) {
3063 ret = ext4_run_lazyinit_thread();
3068 mutex_unlock(&ext4_li_mtx);
3075 * We do not need to lock anything since this is called on
3078 static void ext4_destroy_lazyinit_thread(void)
3081 * If thread exited earlier
3082 * there's nothing to be done.
3084 if (!ext4_li_info || !ext4_lazyinit_task)
3087 kthread_stop(ext4_lazyinit_task);
3090 static int set_journal_csum_feature_set(struct super_block *sb)
3093 int compat, incompat;
3094 struct ext4_sb_info *sbi = EXT4_SB(sb);
3096 if (EXT4_HAS_RO_COMPAT_FEATURE(sb,
3097 EXT4_FEATURE_RO_COMPAT_METADATA_CSUM)) {
3098 /* journal checksum v2 */
3100 incompat = JBD2_FEATURE_INCOMPAT_CSUM_V2;
3102 /* journal checksum v1 */
3103 compat = JBD2_FEATURE_COMPAT_CHECKSUM;
3107 if (test_opt(sb, JOURNAL_ASYNC_COMMIT)) {
3108 ret = jbd2_journal_set_features(sbi->s_journal,
3110 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT |
3112 } else if (test_opt(sb, JOURNAL_CHECKSUM)) {
3113 ret = jbd2_journal_set_features(sbi->s_journal,
3116 jbd2_journal_clear_features(sbi->s_journal, 0, 0,
3117 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT);
3119 jbd2_journal_clear_features(sbi->s_journal,
3120 JBD2_FEATURE_COMPAT_CHECKSUM, 0,
3121 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT |
3122 JBD2_FEATURE_INCOMPAT_CSUM_V2);
3129 * Note: calculating the overhead so we can be compatible with
3130 * historical BSD practice is quite difficult in the face of
3131 * clusters/bigalloc. This is because multiple metadata blocks from
3132 * different block group can end up in the same allocation cluster.
3133 * Calculating the exact overhead in the face of clustered allocation
3134 * requires either O(all block bitmaps) in memory or O(number of block
3135 * groups**2) in time. We will still calculate the superblock for
3136 * older file systems --- and if we come across with a bigalloc file
3137 * system with zero in s_overhead_clusters the estimate will be close to
3138 * correct especially for very large cluster sizes --- but for newer
3139 * file systems, it's better to calculate this figure once at mkfs
3140 * time, and store it in the superblock. If the superblock value is
3141 * present (even for non-bigalloc file systems), we will use it.
3143 static int count_overhead(struct super_block *sb, ext4_group_t grp,
3146 struct ext4_sb_info *sbi = EXT4_SB(sb);
3147 struct ext4_group_desc *gdp;
3148 ext4_fsblk_t first_block, last_block, b;
3149 ext4_group_t i, ngroups = ext4_get_groups_count(sb);
3150 int s, j, count = 0;
3152 if (!EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_BIGALLOC))
3153 return (ext4_bg_has_super(sb, grp) + ext4_bg_num_gdb(sb, grp) +
3154 sbi->s_itb_per_group + 2);
3156 first_block = le32_to_cpu(sbi->s_es->s_first_data_block) +
3157 (grp * EXT4_BLOCKS_PER_GROUP(sb));
3158 last_block = first_block + EXT4_BLOCKS_PER_GROUP(sb) - 1;
3159 for (i = 0; i < ngroups; i++) {
3160 gdp = ext4_get_group_desc(sb, i, NULL);
3161 b = ext4_block_bitmap(sb, gdp);
3162 if (b >= first_block && b <= last_block) {
3163 ext4_set_bit(EXT4_B2C(sbi, b - first_block), buf);
3166 b = ext4_inode_bitmap(sb, gdp);
3167 if (b >= first_block && b <= last_block) {
3168 ext4_set_bit(EXT4_B2C(sbi, b - first_block), buf);
3171 b = ext4_inode_table(sb, gdp);
3172 if (b >= first_block && b + sbi->s_itb_per_group <= last_block)
3173 for (j = 0; j < sbi->s_itb_per_group; j++, b++) {
3174 int c = EXT4_B2C(sbi, b - first_block);
3175 ext4_set_bit(c, buf);
3181 if (ext4_bg_has_super(sb, grp)) {
3182 ext4_set_bit(s++, buf);
3185 for (j = ext4_bg_num_gdb(sb, grp); j > 0; j--) {
3186 ext4_set_bit(EXT4_B2C(sbi, s++), buf);
3192 return EXT4_CLUSTERS_PER_GROUP(sb) -
3193 ext4_count_free(buf, EXT4_CLUSTERS_PER_GROUP(sb) / 8);
3197 * Compute the overhead and stash it in sbi->s_overhead
3199 int ext4_calculate_overhead(struct super_block *sb)
3201 struct ext4_sb_info *sbi = EXT4_SB(sb);
3202 struct ext4_super_block *es = sbi->s_es;
3203 ext4_group_t i, ngroups = ext4_get_groups_count(sb);
3204 ext4_fsblk_t overhead = 0;
3205 char *buf = (char *) get_zeroed_page(GFP_KERNEL);
3211 * Compute the overhead (FS structures). This is constant
3212 * for a given filesystem unless the number of block groups
3213 * changes so we cache the previous value until it does.
3217 * All of the blocks before first_data_block are overhead
3219 overhead = EXT4_B2C(sbi, le32_to_cpu(es->s_first_data_block));
3222 * Add the overhead found in each block group
3224 for (i = 0; i < ngroups; i++) {
3227 blks = count_overhead(sb, i, buf);
3230 memset(buf, 0, PAGE_SIZE);
3233 /* Add the journal blocks as well */
3235 overhead += EXT4_NUM_B2C(sbi, sbi->s_journal->j_maxlen);
3237 sbi->s_overhead = overhead;
3239 free_page((unsigned long) buf);
3244 static ext4_fsblk_t ext4_calculate_resv_clusters(struct ext4_sb_info *sbi)
3246 ext4_fsblk_t resv_clusters;
3249 * By default we reserve 2% or 4096 clusters, whichever is smaller.
3250 * This should cover the situations where we can not afford to run
3251 * out of space like for example punch hole, or converting
3252 * uninitialized extents in delalloc path. In most cases such
3253 * allocation would require 1, or 2 blocks, higher numbers are
3256 resv_clusters = ext4_blocks_count(sbi->s_es) >> sbi->s_cluster_bits;
3258 do_div(resv_clusters, 50);
3259 resv_clusters = min_t(ext4_fsblk_t, resv_clusters, 4096);
3261 return resv_clusters;
3265 static int ext4_reserve_clusters(struct ext4_sb_info *sbi, ext4_fsblk_t count)
3267 ext4_fsblk_t clusters = ext4_blocks_count(sbi->s_es) >>
3268 sbi->s_cluster_bits;
3270 if (count >= clusters)
3273 atomic64_set(&sbi->s_resv_clusters, count);
3277 static int ext4_fill_super(struct super_block *sb, void *data, int silent)
3279 char *orig_data = kstrdup(data, GFP_KERNEL);
3280 struct buffer_head *bh;
3281 struct ext4_super_block *es = NULL;
3282 struct ext4_sb_info *sbi;
3284 ext4_fsblk_t sb_block = get_sb_block(&data);
3285 ext4_fsblk_t logical_sb_block;
3286 unsigned long offset = 0;
3287 unsigned long journal_devnum = 0;
3288 unsigned long def_mount_opts;
3293 int blocksize, clustersize;
3294 unsigned int db_count;
3296 int needs_recovery, has_huge_files, has_bigalloc;
3299 unsigned int journal_ioprio = DEFAULT_JOURNAL_IOPRIO;
3300 ext4_group_t first_not_zeroed;
3302 sbi = kzalloc(sizeof(*sbi), GFP_KERNEL);
3306 sbi->s_blockgroup_lock =
3307 kzalloc(sizeof(struct blockgroup_lock), GFP_KERNEL);
3308 if (!sbi->s_blockgroup_lock) {
3312 sb->s_fs_info = sbi;
3314 sbi->s_inode_readahead_blks = EXT4_DEF_INODE_READAHEAD_BLKS;
3315 sbi->s_sb_block = sb_block;
3316 if (sb->s_bdev->bd_part)
3317 sbi->s_sectors_written_start =
3318 part_stat_read(sb->s_bdev->bd_part, sectors[1]);
3320 /* Cleanup superblock name */
3321 for (cp = sb->s_id; (cp = strchr(cp, '/'));)
3324 /* -EINVAL is default */
3326 blocksize = sb_min_blocksize(sb, EXT4_MIN_BLOCK_SIZE);
3328 ext4_msg(sb, KERN_ERR, "unable to set blocksize");
3333 * The ext4 superblock will not be buffer aligned for other than 1kB
3334 * block sizes. We need to calculate the offset from buffer start.
3336 if (blocksize != EXT4_MIN_BLOCK_SIZE) {
3337 logical_sb_block = sb_block * EXT4_MIN_BLOCK_SIZE;
3338 offset = do_div(logical_sb_block, blocksize);
3340 logical_sb_block = sb_block;
3343 if (!(bh = sb_bread(sb, logical_sb_block))) {
3344 ext4_msg(sb, KERN_ERR, "unable to read superblock");
3348 * Note: s_es must be initialized as soon as possible because
3349 * some ext4 macro-instructions depend on its value
3351 es = (struct ext4_super_block *) (bh->b_data + offset);
3353 sb->s_magic = le16_to_cpu(es->s_magic);
3354 if (sb->s_magic != EXT4_SUPER_MAGIC)
3356 sbi->s_kbytes_written = le64_to_cpu(es->s_kbytes_written);
3358 /* Warn if metadata_csum and gdt_csum are both set. */
3359 if (EXT4_HAS_RO_COMPAT_FEATURE(sb,
3360 EXT4_FEATURE_RO_COMPAT_METADATA_CSUM) &&
3361 EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_GDT_CSUM))
3362 ext4_warning(sb, KERN_INFO "metadata_csum and uninit_bg are "
3363 "redundant flags; please run fsck.");
3365 /* Check for a known checksum algorithm */
3366 if (!ext4_verify_csum_type(sb, es)) {
3367 ext4_msg(sb, KERN_ERR, "VFS: Found ext4 filesystem with "
3368 "unknown checksum algorithm.");
3373 /* Load the checksum driver */
3374 if (EXT4_HAS_RO_COMPAT_FEATURE(sb,
3375 EXT4_FEATURE_RO_COMPAT_METADATA_CSUM)) {
3376 sbi->s_chksum_driver = crypto_alloc_shash("crc32c", 0, 0);
3377 if (IS_ERR(sbi->s_chksum_driver)) {
3378 ext4_msg(sb, KERN_ERR, "Cannot load crc32c driver.");
3379 ret = PTR_ERR(sbi->s_chksum_driver);
3380 sbi->s_chksum_driver = NULL;
3385 /* Check superblock checksum */
3386 if (!ext4_superblock_csum_verify(sb, es)) {
3387 ext4_msg(sb, KERN_ERR, "VFS: Found ext4 filesystem with "
3388 "invalid superblock checksum. Run e2fsck?");
3393 /* Precompute checksum seed for all metadata */
3394 if (EXT4_HAS_RO_COMPAT_FEATURE(sb,
3395 EXT4_FEATURE_RO_COMPAT_METADATA_CSUM))
3396 sbi->s_csum_seed = ext4_chksum(sbi, ~0, es->s_uuid,
3397 sizeof(es->s_uuid));
3399 /* Set defaults before we parse the mount options */
3400 def_mount_opts = le32_to_cpu(es->s_default_mount_opts);
3401 set_opt(sb, INIT_INODE_TABLE);
3402 if (def_mount_opts & EXT4_DEFM_DEBUG)
3404 if (def_mount_opts & EXT4_DEFM_BSDGROUPS)
3406 if (def_mount_opts & EXT4_DEFM_UID16)
3407 set_opt(sb, NO_UID32);
3408 /* xattr user namespace & acls are now defaulted on */
3409 set_opt(sb, XATTR_USER);
3410 #ifdef CONFIG_EXT4_FS_POSIX_ACL
3411 set_opt(sb, POSIX_ACL);
3413 if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_DATA)
3414 set_opt(sb, JOURNAL_DATA);
3415 else if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_ORDERED)
3416 set_opt(sb, ORDERED_DATA);
3417 else if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_WBACK)
3418 set_opt(sb, WRITEBACK_DATA);
3420 if (le16_to_cpu(sbi->s_es->s_errors) == EXT4_ERRORS_PANIC)
3421 set_opt(sb, ERRORS_PANIC);
3422 else if (le16_to_cpu(sbi->s_es->s_errors) == EXT4_ERRORS_CONTINUE)
3423 set_opt(sb, ERRORS_CONT);
3425 set_opt(sb, ERRORS_RO);
3426 if (def_mount_opts & EXT4_DEFM_BLOCK_VALIDITY)
3427 set_opt(sb, BLOCK_VALIDITY);
3428 if (def_mount_opts & EXT4_DEFM_DISCARD)
3429 set_opt(sb, DISCARD);
3431 sbi->s_resuid = make_kuid(&init_user_ns, le16_to_cpu(es->s_def_resuid));
3432 sbi->s_resgid = make_kgid(&init_user_ns, le16_to_cpu(es->s_def_resgid));
3433 sbi->s_commit_interval = JBD2_DEFAULT_MAX_COMMIT_AGE * HZ;
3434 sbi->s_min_batch_time = EXT4_DEF_MIN_BATCH_TIME;
3435 sbi->s_max_batch_time = EXT4_DEF_MAX_BATCH_TIME;
3437 if ((def_mount_opts & EXT4_DEFM_NOBARRIER) == 0)
3438 set_opt(sb, BARRIER);
3441 * enable delayed allocation by default
3442 * Use -o nodelalloc to turn it off
3444 if (!IS_EXT3_SB(sb) && !IS_EXT2_SB(sb) &&
3445 ((def_mount_opts & EXT4_DEFM_NODELALLOC) == 0))
3446 set_opt(sb, DELALLOC);
3449 * set default s_li_wait_mult for lazyinit, for the case there is
3450 * no mount option specified.
3452 sbi->s_li_wait_mult = EXT4_DEF_LI_WAIT_MULT;
3454 if (!parse_options((char *) sbi->s_es->s_mount_opts, sb,
3455 &journal_devnum, &journal_ioprio, 0)) {
3456 ext4_msg(sb, KERN_WARNING,
3457 "failed to parse options in superblock: %s",
3458 sbi->s_es->s_mount_opts);
3460 sbi->s_def_mount_opt = sbi->s_mount_opt;
3461 if (!parse_options((char *) data, sb, &journal_devnum,
3462 &journal_ioprio, 0))
3465 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA) {
3466 printk_once(KERN_WARNING "EXT4-fs: Warning: mounting "
3467 "with data=journal disables delayed "
3468 "allocation and O_DIRECT support!\n");
3469 if (test_opt2(sb, EXPLICIT_DELALLOC)) {
3470 ext4_msg(sb, KERN_ERR, "can't mount with "
3471 "both data=journal and delalloc");
3474 if (test_opt(sb, DIOREAD_NOLOCK)) {
3475 ext4_msg(sb, KERN_ERR, "can't mount with "
3476 "both data=journal and delalloc");
3479 if (test_opt(sb, DELALLOC))
3480 clear_opt(sb, DELALLOC);
3483 sb->s_flags = (sb->s_flags & ~MS_POSIXACL) |
3484 (test_opt(sb, POSIX_ACL) ? MS_POSIXACL : 0);
3486 if (le32_to_cpu(es->s_rev_level) == EXT4_GOOD_OLD_REV &&
3487 (EXT4_HAS_COMPAT_FEATURE(sb, ~0U) ||
3488 EXT4_HAS_RO_COMPAT_FEATURE(sb, ~0U) ||
3489 EXT4_HAS_INCOMPAT_FEATURE(sb, ~0U)))
3490 ext4_msg(sb, KERN_WARNING,
3491 "feature flags set on rev 0 fs, "
3492 "running e2fsck is recommended");
3494 if (IS_EXT2_SB(sb)) {
3495 if (ext2_feature_set_ok(sb))
3496 ext4_msg(sb, KERN_INFO, "mounting ext2 file system "
3497 "using the ext4 subsystem");
3499 ext4_msg(sb, KERN_ERR, "couldn't mount as ext2 due "
3500 "to feature incompatibilities");
3505 if (IS_EXT3_SB(sb)) {
3506 if (ext3_feature_set_ok(sb))
3507 ext4_msg(sb, KERN_INFO, "mounting ext3 file system "
3508 "using the ext4 subsystem");
3510 ext4_msg(sb, KERN_ERR, "couldn't mount as ext3 due "
3511 "to feature incompatibilities");
3517 * Check feature flags regardless of the revision level, since we
3518 * previously didn't change the revision level when setting the flags,
3519 * so there is a chance incompat flags are set on a rev 0 filesystem.
3521 if (!ext4_feature_set_ok(sb, (sb->s_flags & MS_RDONLY)))
3524 blocksize = BLOCK_SIZE << le32_to_cpu(es->s_log_block_size);
3525 if (blocksize < EXT4_MIN_BLOCK_SIZE ||
3526 blocksize > EXT4_MAX_BLOCK_SIZE) {
3527 ext4_msg(sb, KERN_ERR,
3528 "Unsupported filesystem blocksize %d", blocksize);
3532 if (sb->s_blocksize != blocksize) {
3533 /* Validate the filesystem blocksize */
3534 if (!sb_set_blocksize(sb, blocksize)) {
3535 ext4_msg(sb, KERN_ERR, "bad block size %d",
3541 logical_sb_block = sb_block * EXT4_MIN_BLOCK_SIZE;
3542 offset = do_div(logical_sb_block, blocksize);
3543 bh = sb_bread(sb, logical_sb_block);
3545 ext4_msg(sb, KERN_ERR,
3546 "Can't read superblock on 2nd try");
3549 es = (struct ext4_super_block *)(bh->b_data + offset);
3551 if (es->s_magic != cpu_to_le16(EXT4_SUPER_MAGIC)) {
3552 ext4_msg(sb, KERN_ERR,
3553 "Magic mismatch, very weird!");
3558 has_huge_files = EXT4_HAS_RO_COMPAT_FEATURE(sb,
3559 EXT4_FEATURE_RO_COMPAT_HUGE_FILE);
3560 sbi->s_bitmap_maxbytes = ext4_max_bitmap_size(sb->s_blocksize_bits,
3562 sb->s_maxbytes = ext4_max_size(sb->s_blocksize_bits, has_huge_files);
3564 if (le32_to_cpu(es->s_rev_level) == EXT4_GOOD_OLD_REV) {
3565 sbi->s_inode_size = EXT4_GOOD_OLD_INODE_SIZE;
3566 sbi->s_first_ino = EXT4_GOOD_OLD_FIRST_INO;
3568 sbi->s_inode_size = le16_to_cpu(es->s_inode_size);
3569 sbi->s_first_ino = le32_to_cpu(es->s_first_ino);
3570 if ((sbi->s_inode_size < EXT4_GOOD_OLD_INODE_SIZE) ||
3571 (!is_power_of_2(sbi->s_inode_size)) ||
3572 (sbi->s_inode_size > blocksize)) {
3573 ext4_msg(sb, KERN_ERR,
3574 "unsupported inode size: %d",
3578 if (sbi->s_inode_size > EXT4_GOOD_OLD_INODE_SIZE)
3579 sb->s_time_gran = 1 << (EXT4_EPOCH_BITS - 2);
3582 sbi->s_desc_size = le16_to_cpu(es->s_desc_size);
3583 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_64BIT)) {
3584 if (sbi->s_desc_size < EXT4_MIN_DESC_SIZE_64BIT ||
3585 sbi->s_desc_size > EXT4_MAX_DESC_SIZE ||
3586 !is_power_of_2(sbi->s_desc_size)) {
3587 ext4_msg(sb, KERN_ERR,
3588 "unsupported descriptor size %lu",
3593 sbi->s_desc_size = EXT4_MIN_DESC_SIZE;
3595 sbi->s_blocks_per_group = le32_to_cpu(es->s_blocks_per_group);
3596 sbi->s_inodes_per_group = le32_to_cpu(es->s_inodes_per_group);
3597 if (EXT4_INODE_SIZE(sb) == 0 || EXT4_INODES_PER_GROUP(sb) == 0)
3600 sbi->s_inodes_per_block = blocksize / EXT4_INODE_SIZE(sb);
3601 if (sbi->s_inodes_per_block == 0)
3603 sbi->s_itb_per_group = sbi->s_inodes_per_group /
3604 sbi->s_inodes_per_block;
3605 sbi->s_desc_per_block = blocksize / EXT4_DESC_SIZE(sb);
3607 sbi->s_mount_state = le16_to_cpu(es->s_state);
3608 sbi->s_addr_per_block_bits = ilog2(EXT4_ADDR_PER_BLOCK(sb));
3609 sbi->s_desc_per_block_bits = ilog2(EXT4_DESC_PER_BLOCK(sb));
3611 /* Do we have standard group size of blocksize * 8 blocks ? */
3612 if (sbi->s_blocks_per_group == blocksize << 3)
3613 set_opt2(sb, STD_GROUP_SIZE);
3615 for (i = 0; i < 4; i++)
3616 sbi->s_hash_seed[i] = le32_to_cpu(es->s_hash_seed[i]);
3617 sbi->s_def_hash_version = es->s_def_hash_version;
3618 i = le32_to_cpu(es->s_flags);
3619 if (i & EXT2_FLAGS_UNSIGNED_HASH)
3620 sbi->s_hash_unsigned = 3;
3621 else if ((i & EXT2_FLAGS_SIGNED_HASH) == 0) {
3622 #ifdef __CHAR_UNSIGNED__
3623 es->s_flags |= cpu_to_le32(EXT2_FLAGS_UNSIGNED_HASH);
3624 sbi->s_hash_unsigned = 3;
3626 es->s_flags |= cpu_to_le32(EXT2_FLAGS_SIGNED_HASH);
3630 /* Handle clustersize */
3631 clustersize = BLOCK_SIZE << le32_to_cpu(es->s_log_cluster_size);
3632 has_bigalloc = EXT4_HAS_RO_COMPAT_FEATURE(sb,
3633 EXT4_FEATURE_RO_COMPAT_BIGALLOC);
3635 if (clustersize < blocksize) {
3636 ext4_msg(sb, KERN_ERR,
3637 "cluster size (%d) smaller than "
3638 "block size (%d)", clustersize, blocksize);
3641 sbi->s_cluster_bits = le32_to_cpu(es->s_log_cluster_size) -
3642 le32_to_cpu(es->s_log_block_size);
3643 sbi->s_clusters_per_group =
3644 le32_to_cpu(es->s_clusters_per_group);
3645 if (sbi->s_clusters_per_group > blocksize * 8) {
3646 ext4_msg(sb, KERN_ERR,
3647 "#clusters per group too big: %lu",
3648 sbi->s_clusters_per_group);
3651 if (sbi->s_blocks_per_group !=
3652 (sbi->s_clusters_per_group * (clustersize / blocksize))) {
3653 ext4_msg(sb, KERN_ERR, "blocks per group (%lu) and "
3654 "clusters per group (%lu) inconsistent",
3655 sbi->s_blocks_per_group,
3656 sbi->s_clusters_per_group);
3660 if (clustersize != blocksize) {
3661 ext4_warning(sb, "fragment/cluster size (%d) != "
3662 "block size (%d)", clustersize,
3664 clustersize = blocksize;
3666 if (sbi->s_blocks_per_group > blocksize * 8) {
3667 ext4_msg(sb, KERN_ERR,
3668 "#blocks per group too big: %lu",
3669 sbi->s_blocks_per_group);
3672 sbi->s_clusters_per_group = sbi->s_blocks_per_group;
3673 sbi->s_cluster_bits = 0;
3675 sbi->s_cluster_ratio = clustersize / blocksize;
3677 if (sbi->s_inodes_per_group > blocksize * 8) {
3678 ext4_msg(sb, KERN_ERR,
3679 "#inodes per group too big: %lu",
3680 sbi->s_inodes_per_group);
3685 * Test whether we have more sectors than will fit in sector_t,
3686 * and whether the max offset is addressable by the page cache.
3688 err = generic_check_addressable(sb->s_blocksize_bits,
3689 ext4_blocks_count(es));
3691 ext4_msg(sb, KERN_ERR, "filesystem"
3692 " too large to mount safely on this system");
3693 if (sizeof(sector_t) < 8)
3694 ext4_msg(sb, KERN_WARNING, "CONFIG_LBDAF not enabled");
3698 if (EXT4_BLOCKS_PER_GROUP(sb) == 0)
3701 /* check blocks count against device size */
3702 blocks_count = sb->s_bdev->bd_inode->i_size >> sb->s_blocksize_bits;
3703 if (blocks_count && ext4_blocks_count(es) > blocks_count) {
3704 ext4_msg(sb, KERN_WARNING, "bad geometry: block count %llu "
3705 "exceeds size of device (%llu blocks)",
3706 ext4_blocks_count(es), blocks_count);
3711 * It makes no sense for the first data block to be beyond the end
3712 * of the filesystem.
3714 if (le32_to_cpu(es->s_first_data_block) >= ext4_blocks_count(es)) {
3715 ext4_msg(sb, KERN_WARNING, "bad geometry: first data "
3716 "block %u is beyond end of filesystem (%llu)",
3717 le32_to_cpu(es->s_first_data_block),
3718 ext4_blocks_count(es));
3721 blocks_count = (ext4_blocks_count(es) -
3722 le32_to_cpu(es->s_first_data_block) +
3723 EXT4_BLOCKS_PER_GROUP(sb) - 1);
3724 do_div(blocks_count, EXT4_BLOCKS_PER_GROUP(sb));
3725 if (blocks_count > ((uint64_t)1<<32) - EXT4_DESC_PER_BLOCK(sb)) {
3726 ext4_msg(sb, KERN_WARNING, "groups count too large: %u "
3727 "(block count %llu, first data block %u, "
3728 "blocks per group %lu)", sbi->s_groups_count,
3729 ext4_blocks_count(es),
3730 le32_to_cpu(es->s_first_data_block),
3731 EXT4_BLOCKS_PER_GROUP(sb));
3734 sbi->s_groups_count = blocks_count;
3735 sbi->s_blockfile_groups = min_t(ext4_group_t, sbi->s_groups_count,
3736 (EXT4_MAX_BLOCK_FILE_PHYS / EXT4_BLOCKS_PER_GROUP(sb)));
3737 db_count = (sbi->s_groups_count + EXT4_DESC_PER_BLOCK(sb) - 1) /
3738 EXT4_DESC_PER_BLOCK(sb);
3739 sbi->s_group_desc = ext4_kvmalloc(db_count *
3740 sizeof(struct buffer_head *),
3742 if (sbi->s_group_desc == NULL) {
3743 ext4_msg(sb, KERN_ERR, "not enough memory");
3749 sbi->s_proc = proc_mkdir(sb->s_id, ext4_proc_root);
3752 proc_create_data("options", S_IRUGO, sbi->s_proc,
3753 &ext4_seq_options_fops, sb);
3755 bgl_lock_init(sbi->s_blockgroup_lock);
3757 for (i = 0; i < db_count; i++) {
3758 block = descriptor_loc(sb, logical_sb_block, i);
3759 sbi->s_group_desc[i] = sb_bread(sb, block);
3760 if (!sbi->s_group_desc[i]) {
3761 ext4_msg(sb, KERN_ERR,
3762 "can't read group descriptor %d", i);
3767 if (!ext4_check_descriptors(sb, &first_not_zeroed)) {
3768 ext4_msg(sb, KERN_ERR, "group descriptors corrupted!");
3771 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_FLEX_BG))
3772 if (!ext4_fill_flex_info(sb)) {
3773 ext4_msg(sb, KERN_ERR,
3774 "unable to initialize "
3775 "flex_bg meta info!");
3779 sbi->s_gdb_count = db_count;
3780 get_random_bytes(&sbi->s_next_generation, sizeof(u32));
3781 spin_lock_init(&sbi->s_next_gen_lock);
3783 init_timer(&sbi->s_err_report);
3784 sbi->s_err_report.function = print_daily_error_info;
3785 sbi->s_err_report.data = (unsigned long) sb;
3787 /* Register extent status tree shrinker */
3788 ext4_es_register_shrinker(sb);
3790 err = percpu_counter_init(&sbi->s_freeclusters_counter,
3791 ext4_count_free_clusters(sb));
3793 err = percpu_counter_init(&sbi->s_freeinodes_counter,
3794 ext4_count_free_inodes(sb));
3797 err = percpu_counter_init(&sbi->s_dirs_counter,
3798 ext4_count_dirs(sb));
3801 err = percpu_counter_init(&sbi->s_dirtyclusters_counter, 0);
3804 err = percpu_counter_init(&sbi->s_extent_cache_cnt, 0);
3807 ext4_msg(sb, KERN_ERR, "insufficient memory");
3811 sbi->s_stripe = ext4_get_stripe_size(sbi);
3812 sbi->s_extent_max_zeroout_kb = 32;
3815 * set up enough so that it can read an inode
3817 if (!test_opt(sb, NOLOAD) &&
3818 EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL))
3819 sb->s_op = &ext4_sops;
3821 sb->s_op = &ext4_nojournal_sops;
3822 sb->s_export_op = &ext4_export_ops;
3823 sb->s_xattr = ext4_xattr_handlers;
3825 sb->dq_op = &ext4_quota_operations;
3826 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_QUOTA))
3827 sb->s_qcop = &ext4_qctl_sysfile_operations;
3829 sb->s_qcop = &ext4_qctl_operations;
3831 memcpy(sb->s_uuid, es->s_uuid, sizeof(es->s_uuid));
3833 INIT_LIST_HEAD(&sbi->s_orphan); /* unlinked but open files */
3834 mutex_init(&sbi->s_orphan_lock);
3838 needs_recovery = (es->s_last_orphan != 0 ||
3839 EXT4_HAS_INCOMPAT_FEATURE(sb,
3840 EXT4_FEATURE_INCOMPAT_RECOVER));
3842 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_MMP) &&
3843 !(sb->s_flags & MS_RDONLY))
3844 if (ext4_multi_mount_protect(sb, le64_to_cpu(es->s_mmp_block)))
3848 * The first inode we look at is the journal inode. Don't try
3849 * root first: it may be modified in the journal!
3851 if (!test_opt(sb, NOLOAD) &&
3852 EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL)) {
3853 if (ext4_load_journal(sb, es, journal_devnum))
3855 } else if (test_opt(sb, NOLOAD) && !(sb->s_flags & MS_RDONLY) &&
3856 EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER)) {
3857 ext4_msg(sb, KERN_ERR, "required journal recovery "
3858 "suppressed and not mounted read-only");
3859 goto failed_mount_wq;
3861 clear_opt(sb, DATA_FLAGS);
3862 sbi->s_journal = NULL;
3867 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_64BIT) &&
3868 !jbd2_journal_set_features(EXT4_SB(sb)->s_journal, 0, 0,
3869 JBD2_FEATURE_INCOMPAT_64BIT)) {
3870 ext4_msg(sb, KERN_ERR, "Failed to set 64-bit journal feature");
3871 goto failed_mount_wq;
3874 if (!set_journal_csum_feature_set(sb)) {
3875 ext4_msg(sb, KERN_ERR, "Failed to set journal checksum "
3877 goto failed_mount_wq;
3880 /* We have now updated the journal if required, so we can
3881 * validate the data journaling mode. */
3882 switch (test_opt(sb, DATA_FLAGS)) {
3884 /* No mode set, assume a default based on the journal
3885 * capabilities: ORDERED_DATA if the journal can
3886 * cope, else JOURNAL_DATA
3888 if (jbd2_journal_check_available_features
3889 (sbi->s_journal, 0, 0, JBD2_FEATURE_INCOMPAT_REVOKE))
3890 set_opt(sb, ORDERED_DATA);
3892 set_opt(sb, JOURNAL_DATA);
3895 case EXT4_MOUNT_ORDERED_DATA:
3896 case EXT4_MOUNT_WRITEBACK_DATA:
3897 if (!jbd2_journal_check_available_features
3898 (sbi->s_journal, 0, 0, JBD2_FEATURE_INCOMPAT_REVOKE)) {
3899 ext4_msg(sb, KERN_ERR, "Journal does not support "
3900 "requested data journaling mode");
3901 goto failed_mount_wq;
3906 set_task_ioprio(sbi->s_journal->j_task, journal_ioprio);
3908 sbi->s_journal->j_commit_callback = ext4_journal_commit_callback;
3911 * The journal may have updated the bg summary counts, so we
3912 * need to update the global counters.
3914 percpu_counter_set(&sbi->s_freeclusters_counter,
3915 ext4_count_free_clusters(sb));
3916 percpu_counter_set(&sbi->s_freeinodes_counter,
3917 ext4_count_free_inodes(sb));
3918 percpu_counter_set(&sbi->s_dirs_counter,
3919 ext4_count_dirs(sb));
3920 percpu_counter_set(&sbi->s_dirtyclusters_counter, 0);
3924 * Get the # of file system overhead blocks from the
3925 * superblock if present.
3927 if (es->s_overhead_clusters)
3928 sbi->s_overhead = le32_to_cpu(es->s_overhead_clusters);
3930 err = ext4_calculate_overhead(sb);
3932 goto failed_mount_wq;
3936 * The maximum number of concurrent works can be high and
3937 * concurrency isn't really necessary. Limit it to 1.
3939 EXT4_SB(sb)->dio_unwritten_wq =
3940 alloc_workqueue("ext4-dio-unwritten", WQ_MEM_RECLAIM | WQ_UNBOUND, 1);
3941 if (!EXT4_SB(sb)->dio_unwritten_wq) {
3942 printk(KERN_ERR "EXT4-fs: failed to create DIO workqueue\n");
3944 goto failed_mount_wq;
3948 * The jbd2_journal_load will have done any necessary log recovery,
3949 * so we can safely mount the rest of the filesystem now.
3952 root = ext4_iget(sb, EXT4_ROOT_INO);
3954 ext4_msg(sb, KERN_ERR, "get root inode failed");
3955 ret = PTR_ERR(root);
3959 if (!S_ISDIR(root->i_mode) || !root->i_blocks || !root->i_size) {
3960 ext4_msg(sb, KERN_ERR, "corrupt root inode, run e2fsck");
3964 sb->s_root = d_make_root(root);
3966 ext4_msg(sb, KERN_ERR, "get root dentry failed");
3971 if (ext4_setup_super(sb, es, sb->s_flags & MS_RDONLY))
3972 sb->s_flags |= MS_RDONLY;
3974 /* determine the minimum size of new large inodes, if present */
3975 if (sbi->s_inode_size > EXT4_GOOD_OLD_INODE_SIZE) {
3976 sbi->s_want_extra_isize = sizeof(struct ext4_inode) -
3977 EXT4_GOOD_OLD_INODE_SIZE;
3978 if (EXT4_HAS_RO_COMPAT_FEATURE(sb,
3979 EXT4_FEATURE_RO_COMPAT_EXTRA_ISIZE)) {
3980 if (sbi->s_want_extra_isize <
3981 le16_to_cpu(es->s_want_extra_isize))
3982 sbi->s_want_extra_isize =
3983 le16_to_cpu(es->s_want_extra_isize);
3984 if (sbi->s_want_extra_isize <
3985 le16_to_cpu(es->s_min_extra_isize))
3986 sbi->s_want_extra_isize =
3987 le16_to_cpu(es->s_min_extra_isize);
3990 /* Check if enough inode space is available */
3991 if (EXT4_GOOD_OLD_INODE_SIZE + sbi->s_want_extra_isize >
3992 sbi->s_inode_size) {
3993 sbi->s_want_extra_isize = sizeof(struct ext4_inode) -
3994 EXT4_GOOD_OLD_INODE_SIZE;
3995 ext4_msg(sb, KERN_INFO, "required extra inode space not"
3999 err = ext4_reserve_clusters(sbi, ext4_calculate_resv_clusters(sbi));
4001 ext4_msg(sb, KERN_ERR, "failed to reserve %llu clusters for "
4002 "reserved pool", ext4_calculate_resv_clusters(sbi));
4003 goto failed_mount4a;
4006 err = ext4_setup_system_zone(sb);
4008 ext4_msg(sb, KERN_ERR, "failed to initialize system "
4010 goto failed_mount4a;
4014 err = ext4_mb_init(sb);
4016 ext4_msg(sb, KERN_ERR, "failed to initialize mballoc (%d)",
4021 err = ext4_register_li_request(sb, first_not_zeroed);
4025 sbi->s_kobj.kset = ext4_kset;
4026 init_completion(&sbi->s_kobj_unregister);
4027 err = kobject_init_and_add(&sbi->s_kobj, &ext4_ktype, NULL,
4033 /* Enable quota usage during mount. */
4034 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_QUOTA) &&
4035 !(sb->s_flags & MS_RDONLY)) {
4036 err = ext4_enable_quotas(sb);
4040 #endif /* CONFIG_QUOTA */
4042 EXT4_SB(sb)->s_mount_state |= EXT4_ORPHAN_FS;
4043 ext4_orphan_cleanup(sb, es);
4044 EXT4_SB(sb)->s_mount_state &= ~EXT4_ORPHAN_FS;
4045 if (needs_recovery) {
4046 ext4_msg(sb, KERN_INFO, "recovery complete");
4047 ext4_mark_recovery_complete(sb, es);
4049 if (EXT4_SB(sb)->s_journal) {
4050 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA)
4051 descr = " journalled data mode";
4052 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA)
4053 descr = " ordered data mode";
4055 descr = " writeback data mode";
4057 descr = "out journal";
4059 if (test_opt(sb, DISCARD)) {
4060 struct request_queue *q = bdev_get_queue(sb->s_bdev);
4061 if (!blk_queue_discard(q))
4062 ext4_msg(sb, KERN_WARNING,
4063 "mounting with \"discard\" option, but "
4064 "the device does not support discard");
4067 ext4_msg(sb, KERN_INFO, "mounted filesystem with%s. "
4068 "Opts: %s%s%s", descr, sbi->s_es->s_mount_opts,
4069 *sbi->s_es->s_mount_opts ? "; " : "", orig_data);
4071 if (es->s_error_count)
4072 mod_timer(&sbi->s_err_report, jiffies + 300*HZ); /* 5 minutes */
4079 ext4_msg(sb, KERN_ERR, "VFS: Can't find ext4 filesystem");
4084 kobject_del(&sbi->s_kobj);
4087 ext4_unregister_li_request(sb);
4089 ext4_mb_release(sb);
4091 ext4_ext_release(sb);
4092 ext4_release_system_zone(sb);
4097 ext4_msg(sb, KERN_ERR, "mount failed");
4098 destroy_workqueue(EXT4_SB(sb)->dio_unwritten_wq);
4100 if (sbi->s_journal) {
4101 jbd2_journal_destroy(sbi->s_journal);
4102 sbi->s_journal = NULL;
4105 ext4_es_unregister_shrinker(sb);
4106 del_timer(&sbi->s_err_report);
4107 if (sbi->s_flex_groups)
4108 ext4_kvfree(sbi->s_flex_groups);
4109 percpu_counter_destroy(&sbi->s_freeclusters_counter);
4110 percpu_counter_destroy(&sbi->s_freeinodes_counter);
4111 percpu_counter_destroy(&sbi->s_dirs_counter);
4112 percpu_counter_destroy(&sbi->s_dirtyclusters_counter);
4113 percpu_counter_destroy(&sbi->s_extent_cache_cnt);
4115 kthread_stop(sbi->s_mmp_tsk);
4117 for (i = 0; i < db_count; i++)
4118 brelse(sbi->s_group_desc[i]);
4119 ext4_kvfree(sbi->s_group_desc);
4121 if (sbi->s_chksum_driver)
4122 crypto_free_shash(sbi->s_chksum_driver);
4124 remove_proc_entry("options", sbi->s_proc);
4125 remove_proc_entry(sb->s_id, ext4_proc_root);
4128 for (i = 0; i < MAXQUOTAS; i++)
4129 kfree(sbi->s_qf_names[i]);
4131 ext4_blkdev_remove(sbi);
4134 sb->s_fs_info = NULL;
4135 kfree(sbi->s_blockgroup_lock);
4139 return err ? err : ret;
4143 * Setup any per-fs journal parameters now. We'll do this both on
4144 * initial mount, once the journal has been initialised but before we've
4145 * done any recovery; and again on any subsequent remount.
4147 static void ext4_init_journal_params(struct super_block *sb, journal_t *journal)
4149 struct ext4_sb_info *sbi = EXT4_SB(sb);
4151 journal->j_commit_interval = sbi->s_commit_interval;
4152 journal->j_min_batch_time = sbi->s_min_batch_time;
4153 journal->j_max_batch_time = sbi->s_max_batch_time;
4155 write_lock(&journal->j_state_lock);
4156 if (test_opt(sb, BARRIER))
4157 journal->j_flags |= JBD2_BARRIER;
4159 journal->j_flags &= ~JBD2_BARRIER;
4160 if (test_opt(sb, DATA_ERR_ABORT))
4161 journal->j_flags |= JBD2_ABORT_ON_SYNCDATA_ERR;
4163 journal->j_flags &= ~JBD2_ABORT_ON_SYNCDATA_ERR;
4164 write_unlock(&journal->j_state_lock);
4167 static journal_t *ext4_get_journal(struct super_block *sb,
4168 unsigned int journal_inum)
4170 struct inode *journal_inode;
4173 BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL));
4175 /* First, test for the existence of a valid inode on disk. Bad
4176 * things happen if we iget() an unused inode, as the subsequent
4177 * iput() will try to delete it. */
4179 journal_inode = ext4_iget(sb, journal_inum);
4180 if (IS_ERR(journal_inode)) {
4181 ext4_msg(sb, KERN_ERR, "no journal found");
4184 if (!journal_inode->i_nlink) {
4185 make_bad_inode(journal_inode);
4186 iput(journal_inode);
4187 ext4_msg(sb, KERN_ERR, "journal inode is deleted");
4191 jbd_debug(2, "Journal inode found at %p: %lld bytes\n",
4192 journal_inode, journal_inode->i_size);
4193 if (!S_ISREG(journal_inode->i_mode)) {
4194 ext4_msg(sb, KERN_ERR, "invalid journal inode");
4195 iput(journal_inode);
4199 journal = jbd2_journal_init_inode(journal_inode);
4201 ext4_msg(sb, KERN_ERR, "Could not load journal inode");
4202 iput(journal_inode);
4205 journal->j_private = sb;
4206 ext4_init_journal_params(sb, journal);
4210 static journal_t *ext4_get_dev_journal(struct super_block *sb,
4213 struct buffer_head *bh;
4217 int hblock, blocksize;
4218 ext4_fsblk_t sb_block;
4219 unsigned long offset;
4220 struct ext4_super_block *es;
4221 struct block_device *bdev;
4223 BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL));
4225 bdev = ext4_blkdev_get(j_dev, sb);
4229 blocksize = sb->s_blocksize;
4230 hblock = bdev_logical_block_size(bdev);
4231 if (blocksize < hblock) {
4232 ext4_msg(sb, KERN_ERR,
4233 "blocksize too small for journal device");
4237 sb_block = EXT4_MIN_BLOCK_SIZE / blocksize;
4238 offset = EXT4_MIN_BLOCK_SIZE % blocksize;
4239 set_blocksize(bdev, blocksize);
4240 if (!(bh = __bread(bdev, sb_block, blocksize))) {
4241 ext4_msg(sb, KERN_ERR, "couldn't read superblock of "
4242 "external journal");
4246 es = (struct ext4_super_block *) (bh->b_data + offset);
4247 if ((le16_to_cpu(es->s_magic) != EXT4_SUPER_MAGIC) ||
4248 !(le32_to_cpu(es->s_feature_incompat) &
4249 EXT4_FEATURE_INCOMPAT_JOURNAL_DEV)) {
4250 ext4_msg(sb, KERN_ERR, "external journal has "
4256 if (memcmp(EXT4_SB(sb)->s_es->s_journal_uuid, es->s_uuid, 16)) {
4257 ext4_msg(sb, KERN_ERR, "journal UUID does not match");
4262 len = ext4_blocks_count(es);
4263 start = sb_block + 1;
4264 brelse(bh); /* we're done with the superblock */
4266 journal = jbd2_journal_init_dev(bdev, sb->s_bdev,
4267 start, len, blocksize);
4269 ext4_msg(sb, KERN_ERR, "failed to create device journal");
4272 journal->j_private = sb;
4273 ll_rw_block(READ | REQ_META | REQ_PRIO, 1, &journal->j_sb_buffer);
4274 wait_on_buffer(journal->j_sb_buffer);
4275 if (!buffer_uptodate(journal->j_sb_buffer)) {
4276 ext4_msg(sb, KERN_ERR, "I/O error on journal device");
4279 if (be32_to_cpu(journal->j_superblock->s_nr_users) != 1) {
4280 ext4_msg(sb, KERN_ERR, "External journal has more than one "
4281 "user (unsupported) - %d",
4282 be32_to_cpu(journal->j_superblock->s_nr_users));
4285 EXT4_SB(sb)->journal_bdev = bdev;
4286 ext4_init_journal_params(sb, journal);
4290 jbd2_journal_destroy(journal);
4292 ext4_blkdev_put(bdev);
4296 static int ext4_load_journal(struct super_block *sb,
4297 struct ext4_super_block *es,
4298 unsigned long journal_devnum)
4301 unsigned int journal_inum = le32_to_cpu(es->s_journal_inum);
4304 int really_read_only;
4306 BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL));
4308 if (journal_devnum &&
4309 journal_devnum != le32_to_cpu(es->s_journal_dev)) {
4310 ext4_msg(sb, KERN_INFO, "external journal device major/minor "
4311 "numbers have changed");
4312 journal_dev = new_decode_dev(journal_devnum);
4314 journal_dev = new_decode_dev(le32_to_cpu(es->s_journal_dev));
4316 really_read_only = bdev_read_only(sb->s_bdev);
4319 * Are we loading a blank journal or performing recovery after a
4320 * crash? For recovery, we need to check in advance whether we
4321 * can get read-write access to the device.
4323 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER)) {
4324 if (sb->s_flags & MS_RDONLY) {
4325 ext4_msg(sb, KERN_INFO, "INFO: recovery "
4326 "required on readonly filesystem");
4327 if (really_read_only) {
4328 ext4_msg(sb, KERN_ERR, "write access "
4329 "unavailable, cannot proceed");
4332 ext4_msg(sb, KERN_INFO, "write access will "
4333 "be enabled during recovery");
4337 if (journal_inum && journal_dev) {
4338 ext4_msg(sb, KERN_ERR, "filesystem has both journal "
4339 "and inode journals!");
4344 if (!(journal = ext4_get_journal(sb, journal_inum)))
4347 if (!(journal = ext4_get_dev_journal(sb, journal_dev)))
4351 if (!(journal->j_flags & JBD2_BARRIER))
4352 ext4_msg(sb, KERN_INFO, "barriers disabled");
4354 if (!EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER))
4355 err = jbd2_journal_wipe(journal, !really_read_only);
4357 char *save = kmalloc(EXT4_S_ERR_LEN, GFP_KERNEL);
4359 memcpy(save, ((char *) es) +
4360 EXT4_S_ERR_START, EXT4_S_ERR_LEN);
4361 err = jbd2_journal_load(journal);
4363 memcpy(((char *) es) + EXT4_S_ERR_START,
4364 save, EXT4_S_ERR_LEN);
4369 ext4_msg(sb, KERN_ERR, "error loading journal");
4370 jbd2_journal_destroy(journal);
4374 EXT4_SB(sb)->s_journal = journal;
4375 ext4_clear_journal_err(sb, es);
4377 if (!really_read_only && journal_devnum &&
4378 journal_devnum != le32_to_cpu(es->s_journal_dev)) {
4379 es->s_journal_dev = cpu_to_le32(journal_devnum);
4381 /* Make sure we flush the recovery flag to disk. */
4382 ext4_commit_super(sb, 1);
4388 static int ext4_commit_super(struct super_block *sb, int sync)
4390 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
4391 struct buffer_head *sbh = EXT4_SB(sb)->s_sbh;
4394 if (!sbh || block_device_ejected(sb))
4396 if (buffer_write_io_error(sbh)) {
4398 * Oh, dear. A previous attempt to write the
4399 * superblock failed. This could happen because the
4400 * USB device was yanked out. Or it could happen to
4401 * be a transient write error and maybe the block will
4402 * be remapped. Nothing we can do but to retry the
4403 * write and hope for the best.
4405 ext4_msg(sb, KERN_ERR, "previous I/O error to "
4406 "superblock detected");
4407 clear_buffer_write_io_error(sbh);
4408 set_buffer_uptodate(sbh);
4411 * If the file system is mounted read-only, don't update the
4412 * superblock write time. This avoids updating the superblock
4413 * write time when we are mounting the root file system
4414 * read/only but we need to replay the journal; at that point,
4415 * for people who are east of GMT and who make their clock
4416 * tick in localtime for Windows bug-for-bug compatibility,
4417 * the clock is set in the future, and this will cause e2fsck
4418 * to complain and force a full file system check.
4420 if (!(sb->s_flags & MS_RDONLY))
4421 es->s_wtime = cpu_to_le32(get_seconds());
4422 if (sb->s_bdev->bd_part)
4423 es->s_kbytes_written =
4424 cpu_to_le64(EXT4_SB(sb)->s_kbytes_written +
4425 ((part_stat_read(sb->s_bdev->bd_part, sectors[1]) -
4426 EXT4_SB(sb)->s_sectors_written_start) >> 1));
4428 es->s_kbytes_written =
4429 cpu_to_le64(EXT4_SB(sb)->s_kbytes_written);
4430 ext4_free_blocks_count_set(es,
4431 EXT4_C2B(EXT4_SB(sb), percpu_counter_sum_positive(
4432 &EXT4_SB(sb)->s_freeclusters_counter)));
4433 es->s_free_inodes_count =
4434 cpu_to_le32(percpu_counter_sum_positive(
4435 &EXT4_SB(sb)->s_freeinodes_counter));
4436 BUFFER_TRACE(sbh, "marking dirty");
4437 ext4_superblock_csum_set(sb);
4438 mark_buffer_dirty(sbh);
4440 error = sync_dirty_buffer(sbh);
4444 error = buffer_write_io_error(sbh);
4446 ext4_msg(sb, KERN_ERR, "I/O error while writing "
4448 clear_buffer_write_io_error(sbh);
4449 set_buffer_uptodate(sbh);
4456 * Have we just finished recovery? If so, and if we are mounting (or
4457 * remounting) the filesystem readonly, then we will end up with a
4458 * consistent fs on disk. Record that fact.
4460 static void ext4_mark_recovery_complete(struct super_block *sb,
4461 struct ext4_super_block *es)
4463 journal_t *journal = EXT4_SB(sb)->s_journal;
4465 if (!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL)) {
4466 BUG_ON(journal != NULL);
4469 jbd2_journal_lock_updates(journal);
4470 if (jbd2_journal_flush(journal) < 0)
4473 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER) &&
4474 sb->s_flags & MS_RDONLY) {
4475 EXT4_CLEAR_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
4476 ext4_commit_super(sb, 1);
4480 jbd2_journal_unlock_updates(journal);
4484 * If we are mounting (or read-write remounting) a filesystem whose journal
4485 * has recorded an error from a previous lifetime, move that error to the
4486 * main filesystem now.
4488 static void ext4_clear_journal_err(struct super_block *sb,
4489 struct ext4_super_block *es)
4495 BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL));
4497 journal = EXT4_SB(sb)->s_journal;
4500 * Now check for any error status which may have been recorded in the
4501 * journal by a prior ext4_error() or ext4_abort()
4504 j_errno = jbd2_journal_errno(journal);
4508 errstr = ext4_decode_error(sb, j_errno, nbuf);
4509 ext4_warning(sb, "Filesystem error recorded "
4510 "from previous mount: %s", errstr);
4511 ext4_warning(sb, "Marking fs in need of filesystem check.");
4513 EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS;
4514 es->s_state |= cpu_to_le16(EXT4_ERROR_FS);
4515 ext4_commit_super(sb, 1);
4517 jbd2_journal_clear_err(journal);
4518 jbd2_journal_update_sb_errno(journal);
4523 * Force the running and committing transactions to commit,
4524 * and wait on the commit.
4526 int ext4_force_commit(struct super_block *sb)
4530 if (sb->s_flags & MS_RDONLY)
4533 journal = EXT4_SB(sb)->s_journal;
4534 return ext4_journal_force_commit(journal);
4537 static int ext4_sync_fs(struct super_block *sb, int wait)
4541 struct ext4_sb_info *sbi = EXT4_SB(sb);
4543 trace_ext4_sync_fs(sb, wait);
4544 flush_workqueue(sbi->dio_unwritten_wq);
4546 * Writeback quota in non-journalled quota case - journalled quota has
4549 dquot_writeback_dquots(sb, -1);
4550 if (jbd2_journal_start_commit(sbi->s_journal, &target)) {
4552 jbd2_log_wait_commit(sbi->s_journal, target);
4558 * LVM calls this function before a (read-only) snapshot is created. This
4559 * gives us a chance to flush the journal completely and mark the fs clean.
4561 * Note that only this function cannot bring a filesystem to be in a clean
4562 * state independently. It relies on upper layer to stop all data & metadata
4565 static int ext4_freeze(struct super_block *sb)
4570 if (sb->s_flags & MS_RDONLY)
4573 journal = EXT4_SB(sb)->s_journal;
4575 /* Now we set up the journal barrier. */
4576 jbd2_journal_lock_updates(journal);
4579 * Don't clear the needs_recovery flag if we failed to flush
4582 error = jbd2_journal_flush(journal);
4586 /* Journal blocked and flushed, clear needs_recovery flag. */
4587 EXT4_CLEAR_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
4588 error = ext4_commit_super(sb, 1);
4590 /* we rely on upper layer to stop further updates */
4591 jbd2_journal_unlock_updates(EXT4_SB(sb)->s_journal);
4596 * Called by LVM after the snapshot is done. We need to reset the RECOVER
4597 * flag here, even though the filesystem is not technically dirty yet.
4599 static int ext4_unfreeze(struct super_block *sb)
4601 if (sb->s_flags & MS_RDONLY)
4604 /* Reset the needs_recovery flag before the fs is unlocked. */
4605 EXT4_SET_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
4606 ext4_commit_super(sb, 1);
4611 * Structure to save mount options for ext4_remount's benefit
4613 struct ext4_mount_options {
4614 unsigned long s_mount_opt;
4615 unsigned long s_mount_opt2;
4618 unsigned long s_commit_interval;
4619 u32 s_min_batch_time, s_max_batch_time;
4622 char *s_qf_names[MAXQUOTAS];
4626 static int ext4_remount(struct super_block *sb, int *flags, char *data)
4628 struct ext4_super_block *es;
4629 struct ext4_sb_info *sbi = EXT4_SB(sb);
4630 unsigned long old_sb_flags;
4631 struct ext4_mount_options old_opts;
4632 int enable_quota = 0;
4634 unsigned int journal_ioprio = DEFAULT_JOURNAL_IOPRIO;
4639 char *orig_data = kstrdup(data, GFP_KERNEL);
4641 /* Store the original options */
4642 old_sb_flags = sb->s_flags;
4643 old_opts.s_mount_opt = sbi->s_mount_opt;
4644 old_opts.s_mount_opt2 = sbi->s_mount_opt2;
4645 old_opts.s_resuid = sbi->s_resuid;
4646 old_opts.s_resgid = sbi->s_resgid;
4647 old_opts.s_commit_interval = sbi->s_commit_interval;
4648 old_opts.s_min_batch_time = sbi->s_min_batch_time;
4649 old_opts.s_max_batch_time = sbi->s_max_batch_time;
4651 old_opts.s_jquota_fmt = sbi->s_jquota_fmt;
4652 for (i = 0; i < MAXQUOTAS; i++)
4653 if (sbi->s_qf_names[i]) {
4654 old_opts.s_qf_names[i] = kstrdup(sbi->s_qf_names[i],
4656 if (!old_opts.s_qf_names[i]) {
4657 for (j = 0; j < i; j++)
4658 kfree(old_opts.s_qf_names[j]);
4663 old_opts.s_qf_names[i] = NULL;
4665 if (sbi->s_journal && sbi->s_journal->j_task->io_context)
4666 journal_ioprio = sbi->s_journal->j_task->io_context->ioprio;
4669 * Allow the "check" option to be passed as a remount option.
4671 if (!parse_options(data, sb, NULL, &journal_ioprio, 1)) {
4676 if (sbi->s_mount_flags & EXT4_MF_FS_ABORTED)
4677 ext4_abort(sb, "Abort forced by user");
4679 sb->s_flags = (sb->s_flags & ~MS_POSIXACL) |
4680 (test_opt(sb, POSIX_ACL) ? MS_POSIXACL : 0);
4684 if (sbi->s_journal) {
4685 ext4_init_journal_params(sb, sbi->s_journal);
4686 set_task_ioprio(sbi->s_journal->j_task, journal_ioprio);
4689 if ((*flags & MS_RDONLY) != (sb->s_flags & MS_RDONLY)) {
4690 if (sbi->s_mount_flags & EXT4_MF_FS_ABORTED) {
4695 if (*flags & MS_RDONLY) {
4696 err = dquot_suspend(sb, -1);
4701 * First of all, the unconditional stuff we have to do
4702 * to disable replay of the journal when we next remount
4704 sb->s_flags |= MS_RDONLY;
4707 * OK, test if we are remounting a valid rw partition
4708 * readonly, and if so set the rdonly flag and then
4709 * mark the partition as valid again.
4711 if (!(es->s_state & cpu_to_le16(EXT4_VALID_FS)) &&
4712 (sbi->s_mount_state & EXT4_VALID_FS))
4713 es->s_state = cpu_to_le16(sbi->s_mount_state);
4716 ext4_mark_recovery_complete(sb, es);
4718 /* Make sure we can mount this feature set readwrite */
4719 if (!ext4_feature_set_ok(sb, 0)) {
4724 * Make sure the group descriptor checksums
4725 * are sane. If they aren't, refuse to remount r/w.
4727 for (g = 0; g < sbi->s_groups_count; g++) {
4728 struct ext4_group_desc *gdp =
4729 ext4_get_group_desc(sb, g, NULL);
4731 if (!ext4_group_desc_csum_verify(sb, g, gdp)) {
4732 ext4_msg(sb, KERN_ERR,
4733 "ext4_remount: Checksum for group %u failed (%u!=%u)",
4734 g, le16_to_cpu(ext4_group_desc_csum(sbi, g, gdp)),
4735 le16_to_cpu(gdp->bg_checksum));
4742 * If we have an unprocessed orphan list hanging
4743 * around from a previously readonly bdev mount,
4744 * require a full umount/remount for now.
4746 if (es->s_last_orphan) {
4747 ext4_msg(sb, KERN_WARNING, "Couldn't "
4748 "remount RDWR because of unprocessed "
4749 "orphan inode list. Please "
4750 "umount/remount instead");
4756 * Mounting a RDONLY partition read-write, so reread
4757 * and store the current valid flag. (It may have
4758 * been changed by e2fsck since we originally mounted
4762 ext4_clear_journal_err(sb, es);
4763 sbi->s_mount_state = le16_to_cpu(es->s_state);
4764 if (!ext4_setup_super(sb, es, 0))
4765 sb->s_flags &= ~MS_RDONLY;
4766 if (EXT4_HAS_INCOMPAT_FEATURE(sb,
4767 EXT4_FEATURE_INCOMPAT_MMP))
4768 if (ext4_multi_mount_protect(sb,
4769 le64_to_cpu(es->s_mmp_block))) {
4778 * Reinitialize lazy itable initialization thread based on
4781 if ((sb->s_flags & MS_RDONLY) || !test_opt(sb, INIT_INODE_TABLE))
4782 ext4_unregister_li_request(sb);
4784 ext4_group_t first_not_zeroed;
4785 first_not_zeroed = ext4_has_uninit_itable(sb);
4786 ext4_register_li_request(sb, first_not_zeroed);
4789 ext4_setup_system_zone(sb);
4790 if (sbi->s_journal == NULL && !(old_sb_flags & MS_RDONLY))
4791 ext4_commit_super(sb, 1);
4794 /* Release old quota file names */
4795 for (i = 0; i < MAXQUOTAS; i++)
4796 kfree(old_opts.s_qf_names[i]);
4798 if (sb_any_quota_suspended(sb))
4799 dquot_resume(sb, -1);
4800 else if (EXT4_HAS_RO_COMPAT_FEATURE(sb,
4801 EXT4_FEATURE_RO_COMPAT_QUOTA)) {
4802 err = ext4_enable_quotas(sb);
4809 ext4_msg(sb, KERN_INFO, "re-mounted. Opts: %s", orig_data);
4814 sb->s_flags = old_sb_flags;
4815 sbi->s_mount_opt = old_opts.s_mount_opt;
4816 sbi->s_mount_opt2 = old_opts.s_mount_opt2;
4817 sbi->s_resuid = old_opts.s_resuid;
4818 sbi->s_resgid = old_opts.s_resgid;
4819 sbi->s_commit_interval = old_opts.s_commit_interval;
4820 sbi->s_min_batch_time = old_opts.s_min_batch_time;
4821 sbi->s_max_batch_time = old_opts.s_max_batch_time;
4823 sbi->s_jquota_fmt = old_opts.s_jquota_fmt;
4824 for (i = 0; i < MAXQUOTAS; i++) {
4825 kfree(sbi->s_qf_names[i]);
4826 sbi->s_qf_names[i] = old_opts.s_qf_names[i];
4833 static int ext4_statfs(struct dentry *dentry, struct kstatfs *buf)
4835 struct super_block *sb = dentry->d_sb;
4836 struct ext4_sb_info *sbi = EXT4_SB(sb);
4837 struct ext4_super_block *es = sbi->s_es;
4838 ext4_fsblk_t overhead = 0, resv_blocks;
4841 resv_blocks = EXT4_C2B(sbi, atomic64_read(&sbi->s_resv_clusters));
4843 if (!test_opt(sb, MINIX_DF))
4844 overhead = sbi->s_overhead;
4846 buf->f_type = EXT4_SUPER_MAGIC;
4847 buf->f_bsize = sb->s_blocksize;
4848 buf->f_blocks = ext4_blocks_count(es) - EXT4_C2B(sbi, overhead);
4849 bfree = percpu_counter_sum_positive(&sbi->s_freeclusters_counter) -
4850 percpu_counter_sum_positive(&sbi->s_dirtyclusters_counter);
4851 /* prevent underflow in case that few free space is available */
4852 buf->f_bfree = EXT4_C2B(sbi, max_t(s64, bfree, 0));
4853 buf->f_bavail = buf->f_bfree -
4854 (ext4_r_blocks_count(es) + resv_blocks);
4855 if (buf->f_bfree < (ext4_r_blocks_count(es) + resv_blocks))
4857 buf->f_files = le32_to_cpu(es->s_inodes_count);
4858 buf->f_ffree = percpu_counter_sum_positive(&sbi->s_freeinodes_counter);
4859 buf->f_namelen = EXT4_NAME_LEN;
4860 fsid = le64_to_cpup((void *)es->s_uuid) ^
4861 le64_to_cpup((void *)es->s_uuid + sizeof(u64));
4862 buf->f_fsid.val[0] = fsid & 0xFFFFFFFFUL;
4863 buf->f_fsid.val[1] = (fsid >> 32) & 0xFFFFFFFFUL;
4868 /* Helper function for writing quotas on sync - we need to start transaction
4869 * before quota file is locked for write. Otherwise the are possible deadlocks:
4870 * Process 1 Process 2
4871 * ext4_create() quota_sync()
4872 * jbd2_journal_start() write_dquot()
4873 * dquot_initialize() down(dqio_mutex)
4874 * down(dqio_mutex) jbd2_journal_start()
4880 static inline struct inode *dquot_to_inode(struct dquot *dquot)
4882 return sb_dqopt(dquot->dq_sb)->files[dquot->dq_id.type];
4885 static int ext4_write_dquot(struct dquot *dquot)
4889 struct inode *inode;
4891 inode = dquot_to_inode(dquot);
4892 handle = ext4_journal_start(inode, EXT4_HT_QUOTA,
4893 EXT4_QUOTA_TRANS_BLOCKS(dquot->dq_sb));
4895 return PTR_ERR(handle);
4896 ret = dquot_commit(dquot);
4897 err = ext4_journal_stop(handle);
4903 static int ext4_acquire_dquot(struct dquot *dquot)
4908 handle = ext4_journal_start(dquot_to_inode(dquot), EXT4_HT_QUOTA,
4909 EXT4_QUOTA_INIT_BLOCKS(dquot->dq_sb));
4911 return PTR_ERR(handle);
4912 ret = dquot_acquire(dquot);
4913 err = ext4_journal_stop(handle);
4919 static int ext4_release_dquot(struct dquot *dquot)
4924 handle = ext4_journal_start(dquot_to_inode(dquot), EXT4_HT_QUOTA,
4925 EXT4_QUOTA_DEL_BLOCKS(dquot->dq_sb));
4926 if (IS_ERR(handle)) {
4927 /* Release dquot anyway to avoid endless cycle in dqput() */
4928 dquot_release(dquot);
4929 return PTR_ERR(handle);
4931 ret = dquot_release(dquot);
4932 err = ext4_journal_stop(handle);
4938 static int ext4_mark_dquot_dirty(struct dquot *dquot)
4940 struct super_block *sb = dquot->dq_sb;
4941 struct ext4_sb_info *sbi = EXT4_SB(sb);
4943 /* Are we journaling quotas? */
4944 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_QUOTA) ||
4945 sbi->s_qf_names[USRQUOTA] || sbi->s_qf_names[GRPQUOTA]) {
4946 dquot_mark_dquot_dirty(dquot);
4947 return ext4_write_dquot(dquot);
4949 return dquot_mark_dquot_dirty(dquot);
4953 static int ext4_write_info(struct super_block *sb, int type)
4958 /* Data block + inode block */
4959 handle = ext4_journal_start(sb->s_root->d_inode, EXT4_HT_QUOTA, 2);
4961 return PTR_ERR(handle);
4962 ret = dquot_commit_info(sb, type);
4963 err = ext4_journal_stop(handle);
4970 * Turn on quotas during mount time - we need to find
4971 * the quota file and such...
4973 static int ext4_quota_on_mount(struct super_block *sb, int type)
4975 return dquot_quota_on_mount(sb, EXT4_SB(sb)->s_qf_names[type],
4976 EXT4_SB(sb)->s_jquota_fmt, type);
4980 * Standard function to be called on quota_on
4982 static int ext4_quota_on(struct super_block *sb, int type, int format_id,
4987 if (!test_opt(sb, QUOTA))
4990 /* Quotafile not on the same filesystem? */
4991 if (path->dentry->d_sb != sb)
4993 /* Journaling quota? */
4994 if (EXT4_SB(sb)->s_qf_names[type]) {
4995 /* Quotafile not in fs root? */
4996 if (path->dentry->d_parent != sb->s_root)
4997 ext4_msg(sb, KERN_WARNING,
4998 "Quota file not on filesystem root. "
4999 "Journaled quota will not work");
5003 * When we journal data on quota file, we have to flush journal to see
5004 * all updates to the file when we bypass pagecache...
5006 if (EXT4_SB(sb)->s_journal &&
5007 ext4_should_journal_data(path->dentry->d_inode)) {
5009 * We don't need to lock updates but journal_flush() could
5010 * otherwise be livelocked...
5012 jbd2_journal_lock_updates(EXT4_SB(sb)->s_journal);
5013 err = jbd2_journal_flush(EXT4_SB(sb)->s_journal);
5014 jbd2_journal_unlock_updates(EXT4_SB(sb)->s_journal);
5019 return dquot_quota_on(sb, type, format_id, path);
5022 static int ext4_quota_enable(struct super_block *sb, int type, int format_id,
5026 struct inode *qf_inode;
5027 unsigned long qf_inums[MAXQUOTAS] = {
5028 le32_to_cpu(EXT4_SB(sb)->s_es->s_usr_quota_inum),
5029 le32_to_cpu(EXT4_SB(sb)->s_es->s_grp_quota_inum)
5032 BUG_ON(!EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_QUOTA));
5034 if (!qf_inums[type])
5037 qf_inode = ext4_iget(sb, qf_inums[type]);
5038 if (IS_ERR(qf_inode)) {
5039 ext4_error(sb, "Bad quota inode # %lu", qf_inums[type]);
5040 return PTR_ERR(qf_inode);
5043 /* Don't account quota for quota files to avoid recursion */
5044 qf_inode->i_flags |= S_NOQUOTA;
5045 err = dquot_enable(qf_inode, type, format_id, flags);
5051 /* Enable usage tracking for all quota types. */
5052 static int ext4_enable_quotas(struct super_block *sb)
5055 unsigned long qf_inums[MAXQUOTAS] = {
5056 le32_to_cpu(EXT4_SB(sb)->s_es->s_usr_quota_inum),
5057 le32_to_cpu(EXT4_SB(sb)->s_es->s_grp_quota_inum)
5060 sb_dqopt(sb)->flags |= DQUOT_QUOTA_SYS_FILE;
5061 for (type = 0; type < MAXQUOTAS; type++) {
5062 if (qf_inums[type]) {
5063 err = ext4_quota_enable(sb, type, QFMT_VFS_V1,
5064 DQUOT_USAGE_ENABLED);
5067 "Failed to enable quota tracking "
5068 "(type=%d, err=%d). Please run "
5069 "e2fsck to fix.", type, err);
5078 * quota_on function that is used when QUOTA feature is set.
5080 static int ext4_quota_on_sysfile(struct super_block *sb, int type,
5083 if (!EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_QUOTA))
5087 * USAGE was enabled at mount time. Only need to enable LIMITS now.
5089 return ext4_quota_enable(sb, type, format_id, DQUOT_LIMITS_ENABLED);
5092 static int ext4_quota_off(struct super_block *sb, int type)
5094 struct inode *inode = sb_dqopt(sb)->files[type];
5097 /* Force all delayed allocation blocks to be allocated.
5098 * Caller already holds s_umount sem */
5099 if (test_opt(sb, DELALLOC))
5100 sync_filesystem(sb);
5105 /* Update modification times of quota files when userspace can
5106 * start looking at them */
5107 handle = ext4_journal_start(inode, EXT4_HT_QUOTA, 1);
5110 inode->i_mtime = inode->i_ctime = CURRENT_TIME;
5111 ext4_mark_inode_dirty(handle, inode);
5112 ext4_journal_stop(handle);
5115 return dquot_quota_off(sb, type);
5119 * quota_off function that is used when QUOTA feature is set.
5121 static int ext4_quota_off_sysfile(struct super_block *sb, int type)
5123 if (!EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_QUOTA))
5126 /* Disable only the limits. */
5127 return dquot_disable(sb, type, DQUOT_LIMITS_ENABLED);
5130 /* Read data from quotafile - avoid pagecache and such because we cannot afford
5131 * acquiring the locks... As quota files are never truncated and quota code
5132 * itself serializes the operations (and no one else should touch the files)
5133 * we don't have to be afraid of races */
5134 static ssize_t ext4_quota_read(struct super_block *sb, int type, char *data,
5135 size_t len, loff_t off)
5137 struct inode *inode = sb_dqopt(sb)->files[type];
5138 ext4_lblk_t blk = off >> EXT4_BLOCK_SIZE_BITS(sb);
5140 int offset = off & (sb->s_blocksize - 1);
5143 struct buffer_head *bh;
5144 loff_t i_size = i_size_read(inode);
5148 if (off+len > i_size)
5151 while (toread > 0) {
5152 tocopy = sb->s_blocksize - offset < toread ?
5153 sb->s_blocksize - offset : toread;
5154 bh = ext4_bread(NULL, inode, blk, 0, &err);
5157 if (!bh) /* A hole? */
5158 memset(data, 0, tocopy);
5160 memcpy(data, bh->b_data+offset, tocopy);
5170 /* Write to quotafile (we know the transaction is already started and has
5171 * enough credits) */
5172 static ssize_t ext4_quota_write(struct super_block *sb, int type,
5173 const char *data, size_t len, loff_t off)
5175 struct inode *inode = sb_dqopt(sb)->files[type];
5176 ext4_lblk_t blk = off >> EXT4_BLOCK_SIZE_BITS(sb);
5178 int offset = off & (sb->s_blocksize - 1);
5179 struct buffer_head *bh;
5180 handle_t *handle = journal_current_handle();
5182 if (EXT4_SB(sb)->s_journal && !handle) {
5183 ext4_msg(sb, KERN_WARNING, "Quota write (off=%llu, len=%llu)"
5184 " cancelled because transaction is not started",
5185 (unsigned long long)off, (unsigned long long)len);
5189 * Since we account only one data block in transaction credits,
5190 * then it is impossible to cross a block boundary.
5192 if (sb->s_blocksize - offset < len) {
5193 ext4_msg(sb, KERN_WARNING, "Quota write (off=%llu, len=%llu)"
5194 " cancelled because not block aligned",
5195 (unsigned long long)off, (unsigned long long)len);
5199 bh = ext4_bread(handle, inode, blk, 1, &err);
5202 err = ext4_journal_get_write_access(handle, bh);
5208 memcpy(bh->b_data+offset, data, len);
5209 flush_dcache_page(bh->b_page);
5211 err = ext4_handle_dirty_metadata(handle, NULL, bh);
5216 if (inode->i_size < off + len) {
5217 i_size_write(inode, off + len);
5218 EXT4_I(inode)->i_disksize = inode->i_size;
5219 ext4_mark_inode_dirty(handle, inode);
5226 static struct dentry *ext4_mount(struct file_system_type *fs_type, int flags,
5227 const char *dev_name, void *data)
5229 return mount_bdev(fs_type, flags, dev_name, data, ext4_fill_super);
5232 #if !defined(CONFIG_EXT2_FS) && !defined(CONFIG_EXT2_FS_MODULE) && defined(CONFIG_EXT4_USE_FOR_EXT23)
5233 static inline void register_as_ext2(void)
5235 int err = register_filesystem(&ext2_fs_type);
5238 "EXT4-fs: Unable to register as ext2 (%d)\n", err);
5241 static inline void unregister_as_ext2(void)
5243 unregister_filesystem(&ext2_fs_type);
5246 static inline int ext2_feature_set_ok(struct super_block *sb)
5248 if (EXT4_HAS_INCOMPAT_FEATURE(sb, ~EXT2_FEATURE_INCOMPAT_SUPP))
5250 if (sb->s_flags & MS_RDONLY)
5252 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, ~EXT2_FEATURE_RO_COMPAT_SUPP))
5257 static inline void register_as_ext2(void) { }
5258 static inline void unregister_as_ext2(void) { }
5259 static inline int ext2_feature_set_ok(struct super_block *sb) { return 0; }
5262 #if !defined(CONFIG_EXT3_FS) && !defined(CONFIG_EXT3_FS_MODULE) && defined(CONFIG_EXT4_USE_FOR_EXT23)
5263 static inline void register_as_ext3(void)
5265 int err = register_filesystem(&ext3_fs_type);
5268 "EXT4-fs: Unable to register as ext3 (%d)\n", err);
5271 static inline void unregister_as_ext3(void)
5273 unregister_filesystem(&ext3_fs_type);
5276 static inline int ext3_feature_set_ok(struct super_block *sb)
5278 if (EXT4_HAS_INCOMPAT_FEATURE(sb, ~EXT3_FEATURE_INCOMPAT_SUPP))
5280 if (!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL))
5282 if (sb->s_flags & MS_RDONLY)
5284 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, ~EXT3_FEATURE_RO_COMPAT_SUPP))
5289 static inline void register_as_ext3(void) { }
5290 static inline void unregister_as_ext3(void) { }
5291 static inline int ext3_feature_set_ok(struct super_block *sb) { return 0; }
5294 static struct file_system_type ext4_fs_type = {
5295 .owner = THIS_MODULE,
5297 .mount = ext4_mount,
5298 .kill_sb = kill_block_super,
5299 .fs_flags = FS_REQUIRES_DEV,
5301 MODULE_ALIAS_FS("ext4");
5303 static int __init ext4_init_feat_adverts(void)
5305 struct ext4_features *ef;
5308 ef = kzalloc(sizeof(struct ext4_features), GFP_KERNEL);
5312 ef->f_kobj.kset = ext4_kset;
5313 init_completion(&ef->f_kobj_unregister);
5314 ret = kobject_init_and_add(&ef->f_kobj, &ext4_feat_ktype, NULL,
5327 static void ext4_exit_feat_adverts(void)
5329 kobject_put(&ext4_feat->f_kobj);
5330 wait_for_completion(&ext4_feat->f_kobj_unregister);
5334 /* Shared across all ext4 file systems */
5335 wait_queue_head_t ext4__ioend_wq[EXT4_WQ_HASH_SZ];
5336 struct mutex ext4__aio_mutex[EXT4_WQ_HASH_SZ];
5338 static int __init ext4_init_fs(void)
5342 ext4_li_info = NULL;
5343 mutex_init(&ext4_li_mtx);
5345 /* Build-time check for flags consistency */
5346 ext4_check_flag_values();
5348 for (i = 0; i < EXT4_WQ_HASH_SZ; i++) {
5349 mutex_init(&ext4__aio_mutex[i]);
5350 init_waitqueue_head(&ext4__ioend_wq[i]);
5353 err = ext4_init_es();
5357 err = ext4_init_pageio();
5361 err = ext4_init_system_zone();
5364 ext4_kset = kset_create_and_add("ext4", NULL, fs_kobj);
5369 ext4_proc_root = proc_mkdir("fs/ext4", NULL);
5371 err = ext4_init_feat_adverts();
5375 err = ext4_init_mballoc();
5379 err = ext4_init_xattr();
5382 err = init_inodecache();
5387 err = register_filesystem(&ext4_fs_type);
5393 unregister_as_ext2();
5394 unregister_as_ext3();
5395 destroy_inodecache();
5399 ext4_exit_mballoc();
5401 ext4_exit_feat_adverts();
5404 remove_proc_entry("fs/ext4", NULL);
5405 kset_unregister(ext4_kset);
5407 ext4_exit_system_zone();
5416 static void __exit ext4_exit_fs(void)
5418 ext4_destroy_lazyinit_thread();
5419 unregister_as_ext2();
5420 unregister_as_ext3();
5421 unregister_filesystem(&ext4_fs_type);
5422 destroy_inodecache();
5424 ext4_exit_mballoc();
5425 ext4_exit_feat_adverts();
5426 remove_proc_entry("fs/ext4", NULL);
5427 kset_unregister(ext4_kset);
5428 ext4_exit_system_zone();
5432 MODULE_AUTHOR("Remy Card, Stephen Tweedie, Andrew Morton, Andreas Dilger, Theodore Ts'o and others");
5433 MODULE_DESCRIPTION("Fourth Extended Filesystem");
5434 MODULE_LICENSE("GPL");
5435 module_init(ext4_init_fs)
5436 module_exit(ext4_exit_fs)