2 * linux/fs/ext4/super.c
4 * Copyright (C) 1992, 1993, 1994, 1995
5 * Remy Card (card@masi.ibp.fr)
6 * Laboratoire MASI - Institut Blaise Pascal
7 * Universite Pierre et Marie Curie (Paris VI)
11 * linux/fs/minix/inode.c
13 * Copyright (C) 1991, 1992 Linus Torvalds
15 * Big-endian to little-endian byte-swapping/bitmaps by
16 * David S. Miller (davem@caip.rutgers.edu), 1995
19 #include <linux/module.h>
20 #include <linux/string.h>
22 #include <linux/time.h>
23 #include <linux/vmalloc.h>
24 #include <linux/jbd2.h>
25 #include <linux/slab.h>
26 #include <linux/init.h>
27 #include <linux/blkdev.h>
28 #include <linux/parser.h>
29 #include <linux/buffer_head.h>
30 #include <linux/exportfs.h>
31 #include <linux/vfs.h>
32 #include <linux/random.h>
33 #include <linux/mount.h>
34 #include <linux/namei.h>
35 #include <linux/quotaops.h>
36 #include <linux/seq_file.h>
37 #include <linux/proc_fs.h>
38 #include <linux/ctype.h>
39 #include <linux/log2.h>
40 #include <linux/crc16.h>
41 #include <linux/cleancache.h>
42 #include <asm/uaccess.h>
44 #include <linux/kthread.h>
45 #include <linux/freezer.h>
48 #include "ext4_extents.h" /* Needed for trace points definition */
49 #include "ext4_jbd2.h"
54 #define CREATE_TRACE_POINTS
55 #include <trace/events/ext4.h>
57 static struct proc_dir_entry *ext4_proc_root;
58 static struct kset *ext4_kset;
59 static struct ext4_lazy_init *ext4_li_info;
60 static struct mutex ext4_li_mtx;
61 static struct ext4_features *ext4_feat;
63 static int ext4_load_journal(struct super_block *, struct ext4_super_block *,
64 unsigned long journal_devnum);
65 static int ext4_show_options(struct seq_file *seq, struct dentry *root);
66 static int ext4_commit_super(struct super_block *sb, int sync);
67 static void ext4_mark_recovery_complete(struct super_block *sb,
68 struct ext4_super_block *es);
69 static void ext4_clear_journal_err(struct super_block *sb,
70 struct ext4_super_block *es);
71 static int ext4_sync_fs(struct super_block *sb, int wait);
72 static int ext4_remount(struct super_block *sb, int *flags, char *data);
73 static int ext4_statfs(struct dentry *dentry, struct kstatfs *buf);
74 static int ext4_unfreeze(struct super_block *sb);
75 static int ext4_freeze(struct super_block *sb);
76 static struct dentry *ext4_mount(struct file_system_type *fs_type, int flags,
77 const char *dev_name, void *data);
78 static inline int ext2_feature_set_ok(struct super_block *sb);
79 static inline int ext3_feature_set_ok(struct super_block *sb);
80 static int ext4_feature_set_ok(struct super_block *sb, int readonly);
81 static void ext4_destroy_lazyinit_thread(void);
82 static void ext4_unregister_li_request(struct super_block *sb);
83 static void ext4_clear_request_list(void);
85 #if !defined(CONFIG_EXT2_FS) && !defined(CONFIG_EXT2_FS_MODULE) && defined(CONFIG_EXT4_USE_FOR_EXT23)
86 static struct file_system_type ext2_fs_type = {
90 .kill_sb = kill_block_super,
91 .fs_flags = FS_REQUIRES_DEV,
93 MODULE_ALIAS_FS("ext2");
95 #define IS_EXT2_SB(sb) ((sb)->s_bdev->bd_holder == &ext2_fs_type)
97 #define IS_EXT2_SB(sb) (0)
101 #if !defined(CONFIG_EXT3_FS) && !defined(CONFIG_EXT3_FS_MODULE) && defined(CONFIG_EXT4_USE_FOR_EXT23)
102 static struct file_system_type ext3_fs_type = {
103 .owner = THIS_MODULE,
106 .kill_sb = kill_block_super,
107 .fs_flags = FS_REQUIRES_DEV,
109 MODULE_ALIAS_FS("ext3");
110 MODULE_ALIAS("ext3");
111 #define IS_EXT3_SB(sb) ((sb)->s_bdev->bd_holder == &ext3_fs_type)
113 #define IS_EXT3_SB(sb) (0)
116 static int ext4_verify_csum_type(struct super_block *sb,
117 struct ext4_super_block *es)
119 if (!EXT4_HAS_RO_COMPAT_FEATURE(sb,
120 EXT4_FEATURE_RO_COMPAT_METADATA_CSUM))
123 return es->s_checksum_type == EXT4_CRC32C_CHKSUM;
126 static __le32 ext4_superblock_csum(struct super_block *sb,
127 struct ext4_super_block *es)
129 struct ext4_sb_info *sbi = EXT4_SB(sb);
130 int offset = offsetof(struct ext4_super_block, s_checksum);
133 csum = ext4_chksum(sbi, ~0, (char *)es, offset);
135 return cpu_to_le32(csum);
138 int ext4_superblock_csum_verify(struct super_block *sb,
139 struct ext4_super_block *es)
141 if (!EXT4_HAS_RO_COMPAT_FEATURE(sb,
142 EXT4_FEATURE_RO_COMPAT_METADATA_CSUM))
145 return es->s_checksum == ext4_superblock_csum(sb, es);
148 void ext4_superblock_csum_set(struct super_block *sb)
150 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
152 if (!EXT4_HAS_RO_COMPAT_FEATURE(sb,
153 EXT4_FEATURE_RO_COMPAT_METADATA_CSUM))
156 es->s_checksum = ext4_superblock_csum(sb, es);
159 void *ext4_kvmalloc(size_t size, gfp_t flags)
163 ret = kmalloc(size, flags);
165 ret = __vmalloc(size, flags, PAGE_KERNEL);
169 void *ext4_kvzalloc(size_t size, gfp_t flags)
173 ret = kzalloc(size, flags);
175 ret = __vmalloc(size, flags | __GFP_ZERO, PAGE_KERNEL);
179 void ext4_kvfree(void *ptr)
181 if (is_vmalloc_addr(ptr))
188 ext4_fsblk_t ext4_block_bitmap(struct super_block *sb,
189 struct ext4_group_desc *bg)
191 return le32_to_cpu(bg->bg_block_bitmap_lo) |
192 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
193 (ext4_fsblk_t)le32_to_cpu(bg->bg_block_bitmap_hi) << 32 : 0);
196 ext4_fsblk_t ext4_inode_bitmap(struct super_block *sb,
197 struct ext4_group_desc *bg)
199 return le32_to_cpu(bg->bg_inode_bitmap_lo) |
200 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
201 (ext4_fsblk_t)le32_to_cpu(bg->bg_inode_bitmap_hi) << 32 : 0);
204 ext4_fsblk_t ext4_inode_table(struct super_block *sb,
205 struct ext4_group_desc *bg)
207 return le32_to_cpu(bg->bg_inode_table_lo) |
208 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
209 (ext4_fsblk_t)le32_to_cpu(bg->bg_inode_table_hi) << 32 : 0);
212 __u32 ext4_free_group_clusters(struct super_block *sb,
213 struct ext4_group_desc *bg)
215 return le16_to_cpu(bg->bg_free_blocks_count_lo) |
216 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
217 (__u32)le16_to_cpu(bg->bg_free_blocks_count_hi) << 16 : 0);
220 __u32 ext4_free_inodes_count(struct super_block *sb,
221 struct ext4_group_desc *bg)
223 return le16_to_cpu(bg->bg_free_inodes_count_lo) |
224 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
225 (__u32)le16_to_cpu(bg->bg_free_inodes_count_hi) << 16 : 0);
228 __u32 ext4_used_dirs_count(struct super_block *sb,
229 struct ext4_group_desc *bg)
231 return le16_to_cpu(bg->bg_used_dirs_count_lo) |
232 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
233 (__u32)le16_to_cpu(bg->bg_used_dirs_count_hi) << 16 : 0);
236 __u32 ext4_itable_unused_count(struct super_block *sb,
237 struct ext4_group_desc *bg)
239 return le16_to_cpu(bg->bg_itable_unused_lo) |
240 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
241 (__u32)le16_to_cpu(bg->bg_itable_unused_hi) << 16 : 0);
244 void ext4_block_bitmap_set(struct super_block *sb,
245 struct ext4_group_desc *bg, ext4_fsblk_t blk)
247 bg->bg_block_bitmap_lo = cpu_to_le32((u32)blk);
248 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
249 bg->bg_block_bitmap_hi = cpu_to_le32(blk >> 32);
252 void ext4_inode_bitmap_set(struct super_block *sb,
253 struct ext4_group_desc *bg, ext4_fsblk_t blk)
255 bg->bg_inode_bitmap_lo = cpu_to_le32((u32)blk);
256 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
257 bg->bg_inode_bitmap_hi = cpu_to_le32(blk >> 32);
260 void ext4_inode_table_set(struct super_block *sb,
261 struct ext4_group_desc *bg, ext4_fsblk_t blk)
263 bg->bg_inode_table_lo = cpu_to_le32((u32)blk);
264 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
265 bg->bg_inode_table_hi = cpu_to_le32(blk >> 32);
268 void ext4_free_group_clusters_set(struct super_block *sb,
269 struct ext4_group_desc *bg, __u32 count)
271 bg->bg_free_blocks_count_lo = cpu_to_le16((__u16)count);
272 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
273 bg->bg_free_blocks_count_hi = cpu_to_le16(count >> 16);
276 void ext4_free_inodes_set(struct super_block *sb,
277 struct ext4_group_desc *bg, __u32 count)
279 bg->bg_free_inodes_count_lo = cpu_to_le16((__u16)count);
280 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
281 bg->bg_free_inodes_count_hi = cpu_to_le16(count >> 16);
284 void ext4_used_dirs_set(struct super_block *sb,
285 struct ext4_group_desc *bg, __u32 count)
287 bg->bg_used_dirs_count_lo = cpu_to_le16((__u16)count);
288 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
289 bg->bg_used_dirs_count_hi = cpu_to_le16(count >> 16);
292 void ext4_itable_unused_set(struct super_block *sb,
293 struct ext4_group_desc *bg, __u32 count)
295 bg->bg_itable_unused_lo = cpu_to_le16((__u16)count);
296 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
297 bg->bg_itable_unused_hi = cpu_to_le16(count >> 16);
301 static void __save_error_info(struct super_block *sb, const char *func,
304 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
306 EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS;
307 es->s_state |= cpu_to_le16(EXT4_ERROR_FS);
308 es->s_last_error_time = cpu_to_le32(get_seconds());
309 strncpy(es->s_last_error_func, func, sizeof(es->s_last_error_func));
310 es->s_last_error_line = cpu_to_le32(line);
311 if (!es->s_first_error_time) {
312 es->s_first_error_time = es->s_last_error_time;
313 strncpy(es->s_first_error_func, func,
314 sizeof(es->s_first_error_func));
315 es->s_first_error_line = cpu_to_le32(line);
316 es->s_first_error_ino = es->s_last_error_ino;
317 es->s_first_error_block = es->s_last_error_block;
320 * Start the daily error reporting function if it hasn't been
323 if (!es->s_error_count)
324 mod_timer(&EXT4_SB(sb)->s_err_report, jiffies + 24*60*60*HZ);
325 le32_add_cpu(&es->s_error_count, 1);
328 static void save_error_info(struct super_block *sb, const char *func,
331 __save_error_info(sb, func, line);
332 ext4_commit_super(sb, 1);
336 * The del_gendisk() function uninitializes the disk-specific data
337 * structures, including the bdi structure, without telling anyone
338 * else. Once this happens, any attempt to call mark_buffer_dirty()
339 * (for example, by ext4_commit_super), will cause a kernel OOPS.
340 * This is a kludge to prevent these oops until we can put in a proper
341 * hook in del_gendisk() to inform the VFS and file system layers.
343 static int block_device_ejected(struct super_block *sb)
345 struct inode *bd_inode = sb->s_bdev->bd_inode;
346 struct backing_dev_info *bdi = bd_inode->i_mapping->backing_dev_info;
348 return bdi->dev == NULL;
351 static void ext4_journal_commit_callback(journal_t *journal, transaction_t *txn)
353 struct super_block *sb = journal->j_private;
354 struct ext4_sb_info *sbi = EXT4_SB(sb);
355 int error = is_journal_aborted(journal);
356 struct ext4_journal_cb_entry *jce;
358 BUG_ON(txn->t_state == T_FINISHED);
359 spin_lock(&sbi->s_md_lock);
360 while (!list_empty(&txn->t_private_list)) {
361 jce = list_entry(txn->t_private_list.next,
362 struct ext4_journal_cb_entry, jce_list);
363 list_del_init(&jce->jce_list);
364 spin_unlock(&sbi->s_md_lock);
365 jce->jce_func(sb, jce, error);
366 spin_lock(&sbi->s_md_lock);
368 spin_unlock(&sbi->s_md_lock);
371 /* Deal with the reporting of failure conditions on a filesystem such as
372 * inconsistencies detected or read IO failures.
374 * On ext2, we can store the error state of the filesystem in the
375 * superblock. That is not possible on ext4, because we may have other
376 * write ordering constraints on the superblock which prevent us from
377 * writing it out straight away; and given that the journal is about to
378 * be aborted, we can't rely on the current, or future, transactions to
379 * write out the superblock safely.
381 * We'll just use the jbd2_journal_abort() error code to record an error in
382 * the journal instead. On recovery, the journal will complain about
383 * that error until we've noted it down and cleared it.
386 static void ext4_handle_error(struct super_block *sb)
388 if (sb->s_flags & MS_RDONLY)
391 if (!test_opt(sb, ERRORS_CONT)) {
392 journal_t *journal = EXT4_SB(sb)->s_journal;
394 EXT4_SB(sb)->s_mount_flags |= EXT4_MF_FS_ABORTED;
396 jbd2_journal_abort(journal, -EIO);
398 if (test_opt(sb, ERRORS_RO)) {
399 ext4_msg(sb, KERN_CRIT, "Remounting filesystem read-only");
400 sb->s_flags |= MS_RDONLY;
402 if (test_opt(sb, ERRORS_PANIC))
403 panic("EXT4-fs (device %s): panic forced after error\n",
407 void __ext4_error(struct super_block *sb, const char *function,
408 unsigned int line, const char *fmt, ...)
410 struct va_format vaf;
416 printk(KERN_CRIT "EXT4-fs error (device %s): %s:%d: comm %s: %pV\n",
417 sb->s_id, function, line, current->comm, &vaf);
419 save_error_info(sb, function, line);
421 ext4_handle_error(sb);
424 void ext4_error_inode(struct inode *inode, const char *function,
425 unsigned int line, ext4_fsblk_t block,
426 const char *fmt, ...)
429 struct va_format vaf;
430 struct ext4_super_block *es = EXT4_SB(inode->i_sb)->s_es;
432 es->s_last_error_ino = cpu_to_le32(inode->i_ino);
433 es->s_last_error_block = cpu_to_le64(block);
434 save_error_info(inode->i_sb, function, line);
439 printk(KERN_CRIT "EXT4-fs error (device %s): %s:%d: "
440 "inode #%lu: block %llu: comm %s: %pV\n",
441 inode->i_sb->s_id, function, line, inode->i_ino,
442 block, current->comm, &vaf);
444 printk(KERN_CRIT "EXT4-fs error (device %s): %s:%d: "
445 "inode #%lu: comm %s: %pV\n",
446 inode->i_sb->s_id, function, line, inode->i_ino,
447 current->comm, &vaf);
450 ext4_handle_error(inode->i_sb);
453 void ext4_error_file(struct file *file, const char *function,
454 unsigned int line, ext4_fsblk_t block,
455 const char *fmt, ...)
458 struct va_format vaf;
459 struct ext4_super_block *es;
460 struct inode *inode = file_inode(file);
461 char pathname[80], *path;
463 es = EXT4_SB(inode->i_sb)->s_es;
464 es->s_last_error_ino = cpu_to_le32(inode->i_ino);
465 save_error_info(inode->i_sb, function, line);
466 path = d_path(&(file->f_path), pathname, sizeof(pathname));
474 "EXT4-fs error (device %s): %s:%d: inode #%lu: "
475 "block %llu: comm %s: path %s: %pV\n",
476 inode->i_sb->s_id, function, line, inode->i_ino,
477 block, current->comm, path, &vaf);
480 "EXT4-fs error (device %s): %s:%d: inode #%lu: "
481 "comm %s: path %s: %pV\n",
482 inode->i_sb->s_id, function, line, inode->i_ino,
483 current->comm, path, &vaf);
486 ext4_handle_error(inode->i_sb);
489 const char *ext4_decode_error(struct super_block *sb, int errno,
496 errstr = "IO failure";
499 errstr = "Out of memory";
502 if (!sb || (EXT4_SB(sb)->s_journal &&
503 EXT4_SB(sb)->s_journal->j_flags & JBD2_ABORT))
504 errstr = "Journal has aborted";
506 errstr = "Readonly filesystem";
509 /* If the caller passed in an extra buffer for unknown
510 * errors, textualise them now. Else we just return
513 /* Check for truncated error codes... */
514 if (snprintf(nbuf, 16, "error %d", -errno) >= 0)
523 /* __ext4_std_error decodes expected errors from journaling functions
524 * automatically and invokes the appropriate error response. */
526 void __ext4_std_error(struct super_block *sb, const char *function,
527 unsigned int line, int errno)
532 /* Special case: if the error is EROFS, and we're not already
533 * inside a transaction, then there's really no point in logging
535 if (errno == -EROFS && journal_current_handle() == NULL &&
536 (sb->s_flags & MS_RDONLY))
539 errstr = ext4_decode_error(sb, errno, nbuf);
540 printk(KERN_CRIT "EXT4-fs error (device %s) in %s:%d: %s\n",
541 sb->s_id, function, line, errstr);
542 save_error_info(sb, function, line);
544 ext4_handle_error(sb);
548 * ext4_abort is a much stronger failure handler than ext4_error. The
549 * abort function may be used to deal with unrecoverable failures such
550 * as journal IO errors or ENOMEM at a critical moment in log management.
552 * We unconditionally force the filesystem into an ABORT|READONLY state,
553 * unless the error response on the fs has been set to panic in which
554 * case we take the easy way out and panic immediately.
557 void __ext4_abort(struct super_block *sb, const char *function,
558 unsigned int line, const char *fmt, ...)
562 save_error_info(sb, function, line);
564 printk(KERN_CRIT "EXT4-fs error (device %s): %s:%d: ", sb->s_id,
570 if ((sb->s_flags & MS_RDONLY) == 0) {
571 ext4_msg(sb, KERN_CRIT, "Remounting filesystem read-only");
572 sb->s_flags |= MS_RDONLY;
573 EXT4_SB(sb)->s_mount_flags |= EXT4_MF_FS_ABORTED;
574 if (EXT4_SB(sb)->s_journal)
575 jbd2_journal_abort(EXT4_SB(sb)->s_journal, -EIO);
576 save_error_info(sb, function, line);
578 if (test_opt(sb, ERRORS_PANIC))
579 panic("EXT4-fs panic from previous error\n");
582 void ext4_msg(struct super_block *sb, const char *prefix, const char *fmt, ...)
584 struct va_format vaf;
590 printk("%sEXT4-fs (%s): %pV\n", prefix, sb->s_id, &vaf);
594 void __ext4_warning(struct super_block *sb, const char *function,
595 unsigned int line, const char *fmt, ...)
597 struct va_format vaf;
603 printk(KERN_WARNING "EXT4-fs warning (device %s): %s:%d: %pV\n",
604 sb->s_id, function, line, &vaf);
608 void __ext4_grp_locked_error(const char *function, unsigned int line,
609 struct super_block *sb, ext4_group_t grp,
610 unsigned long ino, ext4_fsblk_t block,
611 const char *fmt, ...)
615 struct va_format vaf;
617 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
619 es->s_last_error_ino = cpu_to_le32(ino);
620 es->s_last_error_block = cpu_to_le64(block);
621 __save_error_info(sb, function, line);
627 printk(KERN_CRIT "EXT4-fs error (device %s): %s:%d: group %u, ",
628 sb->s_id, function, line, grp);
630 printk(KERN_CONT "inode %lu: ", ino);
632 printk(KERN_CONT "block %llu:", (unsigned long long) block);
633 printk(KERN_CONT "%pV\n", &vaf);
636 if (test_opt(sb, ERRORS_CONT)) {
637 ext4_commit_super(sb, 0);
641 ext4_unlock_group(sb, grp);
642 ext4_handle_error(sb);
644 * We only get here in the ERRORS_RO case; relocking the group
645 * may be dangerous, but nothing bad will happen since the
646 * filesystem will have already been marked read/only and the
647 * journal has been aborted. We return 1 as a hint to callers
648 * who might what to use the return value from
649 * ext4_grp_locked_error() to distinguish between the
650 * ERRORS_CONT and ERRORS_RO case, and perhaps return more
651 * aggressively from the ext4 function in question, with a
652 * more appropriate error code.
654 ext4_lock_group(sb, grp);
658 void ext4_update_dynamic_rev(struct super_block *sb)
660 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
662 if (le32_to_cpu(es->s_rev_level) > EXT4_GOOD_OLD_REV)
666 "updating to rev %d because of new feature flag, "
667 "running e2fsck is recommended",
670 es->s_first_ino = cpu_to_le32(EXT4_GOOD_OLD_FIRST_INO);
671 es->s_inode_size = cpu_to_le16(EXT4_GOOD_OLD_INODE_SIZE);
672 es->s_rev_level = cpu_to_le32(EXT4_DYNAMIC_REV);
673 /* leave es->s_feature_*compat flags alone */
674 /* es->s_uuid will be set by e2fsck if empty */
677 * The rest of the superblock fields should be zero, and if not it
678 * means they are likely already in use, so leave them alone. We
679 * can leave it up to e2fsck to clean up any inconsistencies there.
684 * Open the external journal device
686 static struct block_device *ext4_blkdev_get(dev_t dev, struct super_block *sb)
688 struct block_device *bdev;
689 char b[BDEVNAME_SIZE];
691 bdev = blkdev_get_by_dev(dev, FMODE_READ|FMODE_WRITE|FMODE_EXCL, sb);
697 ext4_msg(sb, KERN_ERR, "failed to open journal device %s: %ld",
698 __bdevname(dev, b), PTR_ERR(bdev));
703 * Release the journal device
705 static int ext4_blkdev_put(struct block_device *bdev)
707 return blkdev_put(bdev, FMODE_READ|FMODE_WRITE|FMODE_EXCL);
710 static int ext4_blkdev_remove(struct ext4_sb_info *sbi)
712 struct block_device *bdev;
715 bdev = sbi->journal_bdev;
717 ret = ext4_blkdev_put(bdev);
718 sbi->journal_bdev = NULL;
723 static inline struct inode *orphan_list_entry(struct list_head *l)
725 return &list_entry(l, struct ext4_inode_info, i_orphan)->vfs_inode;
728 static void dump_orphan_list(struct super_block *sb, struct ext4_sb_info *sbi)
732 ext4_msg(sb, KERN_ERR, "sb orphan head is %d",
733 le32_to_cpu(sbi->s_es->s_last_orphan));
735 printk(KERN_ERR "sb_info orphan list:\n");
736 list_for_each(l, &sbi->s_orphan) {
737 struct inode *inode = orphan_list_entry(l);
739 "inode %s:%lu at %p: mode %o, nlink %d, next %d\n",
740 inode->i_sb->s_id, inode->i_ino, inode,
741 inode->i_mode, inode->i_nlink,
746 static void ext4_put_super(struct super_block *sb)
748 struct ext4_sb_info *sbi = EXT4_SB(sb);
749 struct ext4_super_block *es = sbi->s_es;
752 ext4_unregister_li_request(sb);
753 dquot_disable(sb, -1, DQUOT_USAGE_ENABLED | DQUOT_LIMITS_ENABLED);
755 flush_workqueue(sbi->dio_unwritten_wq);
756 destroy_workqueue(sbi->dio_unwritten_wq);
758 if (sbi->s_journal) {
759 err = jbd2_journal_destroy(sbi->s_journal);
760 sbi->s_journal = NULL;
762 ext4_abort(sb, "Couldn't clean up the journal");
765 ext4_es_unregister_shrinker(sb);
766 del_timer(&sbi->s_err_report);
767 ext4_release_system_zone(sb);
769 ext4_ext_release(sb);
770 ext4_xattr_put_super(sb);
772 if (!(sb->s_flags & MS_RDONLY)) {
773 EXT4_CLEAR_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
774 es->s_state = cpu_to_le16(sbi->s_mount_state);
776 if (!(sb->s_flags & MS_RDONLY))
777 ext4_commit_super(sb, 1);
780 remove_proc_entry("options", sbi->s_proc);
781 remove_proc_entry(sb->s_id, ext4_proc_root);
783 kobject_del(&sbi->s_kobj);
785 for (i = 0; i < sbi->s_gdb_count; i++)
786 brelse(sbi->s_group_desc[i]);
787 ext4_kvfree(sbi->s_group_desc);
788 ext4_kvfree(sbi->s_flex_groups);
789 percpu_counter_destroy(&sbi->s_freeclusters_counter);
790 percpu_counter_destroy(&sbi->s_freeinodes_counter);
791 percpu_counter_destroy(&sbi->s_dirs_counter);
792 percpu_counter_destroy(&sbi->s_dirtyclusters_counter);
793 percpu_counter_destroy(&sbi->s_extent_cache_cnt);
796 for (i = 0; i < MAXQUOTAS; i++)
797 kfree(sbi->s_qf_names[i]);
800 /* Debugging code just in case the in-memory inode orphan list
801 * isn't empty. The on-disk one can be non-empty if we've
802 * detected an error and taken the fs readonly, but the
803 * in-memory list had better be clean by this point. */
804 if (!list_empty(&sbi->s_orphan))
805 dump_orphan_list(sb, sbi);
806 J_ASSERT(list_empty(&sbi->s_orphan));
808 invalidate_bdev(sb->s_bdev);
809 if (sbi->journal_bdev && sbi->journal_bdev != sb->s_bdev) {
811 * Invalidate the journal device's buffers. We don't want them
812 * floating about in memory - the physical journal device may
813 * hotswapped, and it breaks the `ro-after' testing code.
815 sync_blockdev(sbi->journal_bdev);
816 invalidate_bdev(sbi->journal_bdev);
817 ext4_blkdev_remove(sbi);
820 kthread_stop(sbi->s_mmp_tsk);
821 sb->s_fs_info = NULL;
823 * Now that we are completely done shutting down the
824 * superblock, we need to actually destroy the kobject.
826 kobject_put(&sbi->s_kobj);
827 wait_for_completion(&sbi->s_kobj_unregister);
828 if (sbi->s_chksum_driver)
829 crypto_free_shash(sbi->s_chksum_driver);
830 kfree(sbi->s_blockgroup_lock);
834 static struct kmem_cache *ext4_inode_cachep;
837 * Called inside transaction, so use GFP_NOFS
839 static struct inode *ext4_alloc_inode(struct super_block *sb)
841 struct ext4_inode_info *ei;
843 ei = kmem_cache_alloc(ext4_inode_cachep, GFP_NOFS);
847 ei->vfs_inode.i_version = 1;
848 INIT_LIST_HEAD(&ei->i_prealloc_list);
849 spin_lock_init(&ei->i_prealloc_lock);
850 ext4_es_init_tree(&ei->i_es_tree);
851 rwlock_init(&ei->i_es_lock);
852 INIT_LIST_HEAD(&ei->i_es_lru);
854 ei->i_reserved_data_blocks = 0;
855 ei->i_reserved_meta_blocks = 0;
856 ei->i_allocated_meta_blocks = 0;
857 ei->i_da_metadata_calc_len = 0;
858 ei->i_da_metadata_calc_last_lblock = 0;
859 spin_lock_init(&(ei->i_block_reservation_lock));
861 ei->i_reserved_quota = 0;
864 INIT_LIST_HEAD(&ei->i_completed_io_list);
865 spin_lock_init(&ei->i_completed_io_lock);
867 ei->i_datasync_tid = 0;
868 atomic_set(&ei->i_ioend_count, 0);
869 atomic_set(&ei->i_unwritten, 0);
870 INIT_WORK(&ei->i_unwritten_work, ext4_end_io_work);
872 return &ei->vfs_inode;
875 static int ext4_drop_inode(struct inode *inode)
877 int drop = generic_drop_inode(inode);
879 trace_ext4_drop_inode(inode, drop);
883 static void ext4_i_callback(struct rcu_head *head)
885 struct inode *inode = container_of(head, struct inode, i_rcu);
886 kmem_cache_free(ext4_inode_cachep, EXT4_I(inode));
889 static void ext4_destroy_inode(struct inode *inode)
891 if (!list_empty(&(EXT4_I(inode)->i_orphan))) {
892 ext4_msg(inode->i_sb, KERN_ERR,
893 "Inode %lu (%p): orphan list check failed!",
894 inode->i_ino, EXT4_I(inode));
895 print_hex_dump(KERN_INFO, "", DUMP_PREFIX_ADDRESS, 16, 4,
896 EXT4_I(inode), sizeof(struct ext4_inode_info),
900 call_rcu(&inode->i_rcu, ext4_i_callback);
903 static void init_once(void *foo)
905 struct ext4_inode_info *ei = (struct ext4_inode_info *) foo;
907 INIT_LIST_HEAD(&ei->i_orphan);
908 init_rwsem(&ei->xattr_sem);
909 init_rwsem(&ei->i_data_sem);
910 inode_init_once(&ei->vfs_inode);
913 static int init_inodecache(void)
915 ext4_inode_cachep = kmem_cache_create("ext4_inode_cache",
916 sizeof(struct ext4_inode_info),
917 0, (SLAB_RECLAIM_ACCOUNT|
920 if (ext4_inode_cachep == NULL)
925 static void destroy_inodecache(void)
928 * Make sure all delayed rcu free inodes are flushed before we
932 kmem_cache_destroy(ext4_inode_cachep);
935 void ext4_clear_inode(struct inode *inode)
937 invalidate_inode_buffers(inode);
940 ext4_discard_preallocations(inode);
941 ext4_es_remove_extent(inode, 0, EXT_MAX_BLOCKS);
942 ext4_es_lru_del(inode);
943 if (EXT4_I(inode)->jinode) {
944 jbd2_journal_release_jbd_inode(EXT4_JOURNAL(inode),
945 EXT4_I(inode)->jinode);
946 jbd2_free_inode(EXT4_I(inode)->jinode);
947 EXT4_I(inode)->jinode = NULL;
951 static struct inode *ext4_nfs_get_inode(struct super_block *sb,
952 u64 ino, u32 generation)
956 if (ino < EXT4_FIRST_INO(sb) && ino != EXT4_ROOT_INO)
957 return ERR_PTR(-ESTALE);
958 if (ino > le32_to_cpu(EXT4_SB(sb)->s_es->s_inodes_count))
959 return ERR_PTR(-ESTALE);
961 /* iget isn't really right if the inode is currently unallocated!!
963 * ext4_read_inode will return a bad_inode if the inode had been
964 * deleted, so we should be safe.
966 * Currently we don't know the generation for parent directory, so
967 * a generation of 0 means "accept any"
969 inode = ext4_iget(sb, ino);
971 return ERR_CAST(inode);
972 if (generation && inode->i_generation != generation) {
974 return ERR_PTR(-ESTALE);
980 static struct dentry *ext4_fh_to_dentry(struct super_block *sb, struct fid *fid,
981 int fh_len, int fh_type)
983 return generic_fh_to_dentry(sb, fid, fh_len, fh_type,
987 static struct dentry *ext4_fh_to_parent(struct super_block *sb, struct fid *fid,
988 int fh_len, int fh_type)
990 return generic_fh_to_parent(sb, fid, fh_len, fh_type,
995 * Try to release metadata pages (indirect blocks, directories) which are
996 * mapped via the block device. Since these pages could have journal heads
997 * which would prevent try_to_free_buffers() from freeing them, we must use
998 * jbd2 layer's try_to_free_buffers() function to release them.
1000 static int bdev_try_to_free_page(struct super_block *sb, struct page *page,
1003 journal_t *journal = EXT4_SB(sb)->s_journal;
1005 WARN_ON(PageChecked(page));
1006 if (!page_has_buffers(page))
1009 return jbd2_journal_try_to_free_buffers(journal, page,
1010 wait & ~__GFP_WAIT);
1011 return try_to_free_buffers(page);
1015 #define QTYPE2NAME(t) ((t) == USRQUOTA ? "user" : "group")
1016 #define QTYPE2MOPT(on, t) ((t) == USRQUOTA?((on)##USRJQUOTA):((on)##GRPJQUOTA))
1018 static int ext4_write_dquot(struct dquot *dquot);
1019 static int ext4_acquire_dquot(struct dquot *dquot);
1020 static int ext4_release_dquot(struct dquot *dquot);
1021 static int ext4_mark_dquot_dirty(struct dquot *dquot);
1022 static int ext4_write_info(struct super_block *sb, int type);
1023 static int ext4_quota_on(struct super_block *sb, int type, int format_id,
1025 static int ext4_quota_on_sysfile(struct super_block *sb, int type,
1027 static int ext4_quota_off(struct super_block *sb, int type);
1028 static int ext4_quota_off_sysfile(struct super_block *sb, int type);
1029 static int ext4_quota_on_mount(struct super_block *sb, int type);
1030 static ssize_t ext4_quota_read(struct super_block *sb, int type, char *data,
1031 size_t len, loff_t off);
1032 static ssize_t ext4_quota_write(struct super_block *sb, int type,
1033 const char *data, size_t len, loff_t off);
1034 static int ext4_quota_enable(struct super_block *sb, int type, int format_id,
1035 unsigned int flags);
1036 static int ext4_enable_quotas(struct super_block *sb);
1038 static const struct dquot_operations ext4_quota_operations = {
1039 .get_reserved_space = ext4_get_reserved_space,
1040 .write_dquot = ext4_write_dquot,
1041 .acquire_dquot = ext4_acquire_dquot,
1042 .release_dquot = ext4_release_dquot,
1043 .mark_dirty = ext4_mark_dquot_dirty,
1044 .write_info = ext4_write_info,
1045 .alloc_dquot = dquot_alloc,
1046 .destroy_dquot = dquot_destroy,
1049 static const struct quotactl_ops ext4_qctl_operations = {
1050 .quota_on = ext4_quota_on,
1051 .quota_off = ext4_quota_off,
1052 .quota_sync = dquot_quota_sync,
1053 .get_info = dquot_get_dqinfo,
1054 .set_info = dquot_set_dqinfo,
1055 .get_dqblk = dquot_get_dqblk,
1056 .set_dqblk = dquot_set_dqblk
1059 static const struct quotactl_ops ext4_qctl_sysfile_operations = {
1060 .quota_on_meta = ext4_quota_on_sysfile,
1061 .quota_off = ext4_quota_off_sysfile,
1062 .quota_sync = dquot_quota_sync,
1063 .get_info = dquot_get_dqinfo,
1064 .set_info = dquot_set_dqinfo,
1065 .get_dqblk = dquot_get_dqblk,
1066 .set_dqblk = dquot_set_dqblk
1070 static const struct super_operations ext4_sops = {
1071 .alloc_inode = ext4_alloc_inode,
1072 .destroy_inode = ext4_destroy_inode,
1073 .write_inode = ext4_write_inode,
1074 .dirty_inode = ext4_dirty_inode,
1075 .drop_inode = ext4_drop_inode,
1076 .evict_inode = ext4_evict_inode,
1077 .put_super = ext4_put_super,
1078 .sync_fs = ext4_sync_fs,
1079 .freeze_fs = ext4_freeze,
1080 .unfreeze_fs = ext4_unfreeze,
1081 .statfs = ext4_statfs,
1082 .remount_fs = ext4_remount,
1083 .show_options = ext4_show_options,
1085 .quota_read = ext4_quota_read,
1086 .quota_write = ext4_quota_write,
1088 .bdev_try_to_free_page = bdev_try_to_free_page,
1091 static const struct super_operations ext4_nojournal_sops = {
1092 .alloc_inode = ext4_alloc_inode,
1093 .destroy_inode = ext4_destroy_inode,
1094 .write_inode = ext4_write_inode,
1095 .dirty_inode = ext4_dirty_inode,
1096 .drop_inode = ext4_drop_inode,
1097 .evict_inode = ext4_evict_inode,
1098 .put_super = ext4_put_super,
1099 .statfs = ext4_statfs,
1100 .remount_fs = ext4_remount,
1101 .show_options = ext4_show_options,
1103 .quota_read = ext4_quota_read,
1104 .quota_write = ext4_quota_write,
1106 .bdev_try_to_free_page = bdev_try_to_free_page,
1109 static const struct export_operations ext4_export_ops = {
1110 .fh_to_dentry = ext4_fh_to_dentry,
1111 .fh_to_parent = ext4_fh_to_parent,
1112 .get_parent = ext4_get_parent,
1116 Opt_bsd_df, Opt_minix_df, Opt_grpid, Opt_nogrpid,
1117 Opt_resgid, Opt_resuid, Opt_sb, Opt_err_cont, Opt_err_panic, Opt_err_ro,
1118 Opt_nouid32, Opt_debug, Opt_removed,
1119 Opt_user_xattr, Opt_nouser_xattr, Opt_acl, Opt_noacl,
1120 Opt_auto_da_alloc, Opt_noauto_da_alloc, Opt_noload,
1121 Opt_commit, Opt_min_batch_time, Opt_max_batch_time,
1122 Opt_journal_dev, Opt_journal_checksum, Opt_journal_async_commit,
1123 Opt_abort, Opt_data_journal, Opt_data_ordered, Opt_data_writeback,
1124 Opt_data_err_abort, Opt_data_err_ignore,
1125 Opt_usrjquota, Opt_grpjquota, Opt_offusrjquota, Opt_offgrpjquota,
1126 Opt_jqfmt_vfsold, Opt_jqfmt_vfsv0, Opt_jqfmt_vfsv1, Opt_quota,
1127 Opt_noquota, Opt_barrier, Opt_nobarrier, Opt_err,
1128 Opt_usrquota, Opt_grpquota, Opt_i_version,
1129 Opt_stripe, Opt_delalloc, Opt_nodelalloc, Opt_mblk_io_submit,
1130 Opt_nomblk_io_submit, Opt_block_validity, Opt_noblock_validity,
1131 Opt_inode_readahead_blks, Opt_journal_ioprio,
1132 Opt_dioread_nolock, Opt_dioread_lock,
1133 Opt_discard, Opt_nodiscard, Opt_init_itable, Opt_noinit_itable,
1134 Opt_max_dir_size_kb,
1137 static const match_table_t tokens = {
1138 {Opt_bsd_df, "bsddf"},
1139 {Opt_minix_df, "minixdf"},
1140 {Opt_grpid, "grpid"},
1141 {Opt_grpid, "bsdgroups"},
1142 {Opt_nogrpid, "nogrpid"},
1143 {Opt_nogrpid, "sysvgroups"},
1144 {Opt_resgid, "resgid=%u"},
1145 {Opt_resuid, "resuid=%u"},
1147 {Opt_err_cont, "errors=continue"},
1148 {Opt_err_panic, "errors=panic"},
1149 {Opt_err_ro, "errors=remount-ro"},
1150 {Opt_nouid32, "nouid32"},
1151 {Opt_debug, "debug"},
1152 {Opt_removed, "oldalloc"},
1153 {Opt_removed, "orlov"},
1154 {Opt_user_xattr, "user_xattr"},
1155 {Opt_nouser_xattr, "nouser_xattr"},
1157 {Opt_noacl, "noacl"},
1158 {Opt_noload, "norecovery"},
1159 {Opt_noload, "noload"},
1160 {Opt_removed, "nobh"},
1161 {Opt_removed, "bh"},
1162 {Opt_commit, "commit=%u"},
1163 {Opt_min_batch_time, "min_batch_time=%u"},
1164 {Opt_max_batch_time, "max_batch_time=%u"},
1165 {Opt_journal_dev, "journal_dev=%u"},
1166 {Opt_journal_checksum, "journal_checksum"},
1167 {Opt_journal_async_commit, "journal_async_commit"},
1168 {Opt_abort, "abort"},
1169 {Opt_data_journal, "data=journal"},
1170 {Opt_data_ordered, "data=ordered"},
1171 {Opt_data_writeback, "data=writeback"},
1172 {Opt_data_err_abort, "data_err=abort"},
1173 {Opt_data_err_ignore, "data_err=ignore"},
1174 {Opt_offusrjquota, "usrjquota="},
1175 {Opt_usrjquota, "usrjquota=%s"},
1176 {Opt_offgrpjquota, "grpjquota="},
1177 {Opt_grpjquota, "grpjquota=%s"},
1178 {Opt_jqfmt_vfsold, "jqfmt=vfsold"},
1179 {Opt_jqfmt_vfsv0, "jqfmt=vfsv0"},
1180 {Opt_jqfmt_vfsv1, "jqfmt=vfsv1"},
1181 {Opt_grpquota, "grpquota"},
1182 {Opt_noquota, "noquota"},
1183 {Opt_quota, "quota"},
1184 {Opt_usrquota, "usrquota"},
1185 {Opt_barrier, "barrier=%u"},
1186 {Opt_barrier, "barrier"},
1187 {Opt_nobarrier, "nobarrier"},
1188 {Opt_i_version, "i_version"},
1189 {Opt_stripe, "stripe=%u"},
1190 {Opt_delalloc, "delalloc"},
1191 {Opt_nodelalloc, "nodelalloc"},
1192 {Opt_removed, "mblk_io_submit"},
1193 {Opt_removed, "nomblk_io_submit"},
1194 {Opt_block_validity, "block_validity"},
1195 {Opt_noblock_validity, "noblock_validity"},
1196 {Opt_inode_readahead_blks, "inode_readahead_blks=%u"},
1197 {Opt_journal_ioprio, "journal_ioprio=%u"},
1198 {Opt_auto_da_alloc, "auto_da_alloc=%u"},
1199 {Opt_auto_da_alloc, "auto_da_alloc"},
1200 {Opt_noauto_da_alloc, "noauto_da_alloc"},
1201 {Opt_dioread_nolock, "dioread_nolock"},
1202 {Opt_dioread_lock, "dioread_lock"},
1203 {Opt_discard, "discard"},
1204 {Opt_nodiscard, "nodiscard"},
1205 {Opt_init_itable, "init_itable=%u"},
1206 {Opt_init_itable, "init_itable"},
1207 {Opt_noinit_itable, "noinit_itable"},
1208 {Opt_max_dir_size_kb, "max_dir_size_kb=%u"},
1209 {Opt_removed, "check=none"}, /* mount option from ext2/3 */
1210 {Opt_removed, "nocheck"}, /* mount option from ext2/3 */
1211 {Opt_removed, "reservation"}, /* mount option from ext2/3 */
1212 {Opt_removed, "noreservation"}, /* mount option from ext2/3 */
1213 {Opt_removed, "journal=%u"}, /* mount option from ext2/3 */
1217 static ext4_fsblk_t get_sb_block(void **data)
1219 ext4_fsblk_t sb_block;
1220 char *options = (char *) *data;
1222 if (!options || strncmp(options, "sb=", 3) != 0)
1223 return 1; /* Default location */
1226 /* TODO: use simple_strtoll with >32bit ext4 */
1227 sb_block = simple_strtoul(options, &options, 0);
1228 if (*options && *options != ',') {
1229 printk(KERN_ERR "EXT4-fs: Invalid sb specification: %s\n",
1233 if (*options == ',')
1235 *data = (void *) options;
1240 #define DEFAULT_JOURNAL_IOPRIO (IOPRIO_PRIO_VALUE(IOPRIO_CLASS_BE, 3))
1241 static char deprecated_msg[] = "Mount option \"%s\" will be removed by %s\n"
1242 "Contact linux-ext4@vger.kernel.org if you think we should keep it.\n";
1245 static int set_qf_name(struct super_block *sb, int qtype, substring_t *args)
1247 struct ext4_sb_info *sbi = EXT4_SB(sb);
1251 if (sb_any_quota_loaded(sb) &&
1252 !sbi->s_qf_names[qtype]) {
1253 ext4_msg(sb, KERN_ERR,
1254 "Cannot change journaled "
1255 "quota options when quota turned on");
1258 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_QUOTA)) {
1259 ext4_msg(sb, KERN_ERR, "Cannot set journaled quota options "
1260 "when QUOTA feature is enabled");
1263 qname = match_strdup(args);
1265 ext4_msg(sb, KERN_ERR,
1266 "Not enough memory for storing quotafile name");
1269 if (sbi->s_qf_names[qtype]) {
1270 if (strcmp(sbi->s_qf_names[qtype], qname) == 0)
1273 ext4_msg(sb, KERN_ERR,
1274 "%s quota file already specified",
1278 if (strchr(qname, '/')) {
1279 ext4_msg(sb, KERN_ERR,
1280 "quotafile must be on filesystem root");
1283 sbi->s_qf_names[qtype] = qname;
1291 static int clear_qf_name(struct super_block *sb, int qtype)
1294 struct ext4_sb_info *sbi = EXT4_SB(sb);
1296 if (sb_any_quota_loaded(sb) &&
1297 sbi->s_qf_names[qtype]) {
1298 ext4_msg(sb, KERN_ERR, "Cannot change journaled quota options"
1299 " when quota turned on");
1302 kfree(sbi->s_qf_names[qtype]);
1303 sbi->s_qf_names[qtype] = NULL;
1308 #define MOPT_SET 0x0001
1309 #define MOPT_CLEAR 0x0002
1310 #define MOPT_NOSUPPORT 0x0004
1311 #define MOPT_EXPLICIT 0x0008
1312 #define MOPT_CLEAR_ERR 0x0010
1313 #define MOPT_GTE0 0x0020
1316 #define MOPT_QFMT 0x0040
1318 #define MOPT_Q MOPT_NOSUPPORT
1319 #define MOPT_QFMT MOPT_NOSUPPORT
1321 #define MOPT_DATAJ 0x0080
1322 #define MOPT_NO_EXT2 0x0100
1323 #define MOPT_NO_EXT3 0x0200
1324 #define MOPT_EXT4_ONLY (MOPT_NO_EXT2 | MOPT_NO_EXT3)
1326 static const struct mount_opts {
1330 } ext4_mount_opts[] = {
1331 {Opt_minix_df, EXT4_MOUNT_MINIX_DF, MOPT_SET},
1332 {Opt_bsd_df, EXT4_MOUNT_MINIX_DF, MOPT_CLEAR},
1333 {Opt_grpid, EXT4_MOUNT_GRPID, MOPT_SET},
1334 {Opt_nogrpid, EXT4_MOUNT_GRPID, MOPT_CLEAR},
1335 {Opt_block_validity, EXT4_MOUNT_BLOCK_VALIDITY, MOPT_SET},
1336 {Opt_noblock_validity, EXT4_MOUNT_BLOCK_VALIDITY, MOPT_CLEAR},
1337 {Opt_dioread_nolock, EXT4_MOUNT_DIOREAD_NOLOCK,
1338 MOPT_EXT4_ONLY | MOPT_SET},
1339 {Opt_dioread_lock, EXT4_MOUNT_DIOREAD_NOLOCK,
1340 MOPT_EXT4_ONLY | MOPT_CLEAR},
1341 {Opt_discard, EXT4_MOUNT_DISCARD, MOPT_SET},
1342 {Opt_nodiscard, EXT4_MOUNT_DISCARD, MOPT_CLEAR},
1343 {Opt_delalloc, EXT4_MOUNT_DELALLOC,
1344 MOPT_EXT4_ONLY | MOPT_SET | MOPT_EXPLICIT},
1345 {Opt_nodelalloc, EXT4_MOUNT_DELALLOC,
1346 MOPT_EXT4_ONLY | MOPT_CLEAR | MOPT_EXPLICIT},
1347 {Opt_journal_checksum, EXT4_MOUNT_JOURNAL_CHECKSUM,
1348 MOPT_EXT4_ONLY | MOPT_SET},
1349 {Opt_journal_async_commit, (EXT4_MOUNT_JOURNAL_ASYNC_COMMIT |
1350 EXT4_MOUNT_JOURNAL_CHECKSUM),
1351 MOPT_EXT4_ONLY | MOPT_SET},
1352 {Opt_noload, EXT4_MOUNT_NOLOAD, MOPT_NO_EXT2 | MOPT_SET},
1353 {Opt_err_panic, EXT4_MOUNT_ERRORS_PANIC, MOPT_SET | MOPT_CLEAR_ERR},
1354 {Opt_err_ro, EXT4_MOUNT_ERRORS_RO, MOPT_SET | MOPT_CLEAR_ERR},
1355 {Opt_err_cont, EXT4_MOUNT_ERRORS_CONT, MOPT_SET | MOPT_CLEAR_ERR},
1356 {Opt_data_err_abort, EXT4_MOUNT_DATA_ERR_ABORT,
1357 MOPT_NO_EXT2 | MOPT_SET},
1358 {Opt_data_err_ignore, EXT4_MOUNT_DATA_ERR_ABORT,
1359 MOPT_NO_EXT2 | MOPT_CLEAR},
1360 {Opt_barrier, EXT4_MOUNT_BARRIER, MOPT_SET},
1361 {Opt_nobarrier, EXT4_MOUNT_BARRIER, MOPT_CLEAR},
1362 {Opt_noauto_da_alloc, EXT4_MOUNT_NO_AUTO_DA_ALLOC, MOPT_SET},
1363 {Opt_auto_da_alloc, EXT4_MOUNT_NO_AUTO_DA_ALLOC, MOPT_CLEAR},
1364 {Opt_noinit_itable, EXT4_MOUNT_INIT_INODE_TABLE, MOPT_CLEAR},
1365 {Opt_commit, 0, MOPT_GTE0},
1366 {Opt_max_batch_time, 0, MOPT_GTE0},
1367 {Opt_min_batch_time, 0, MOPT_GTE0},
1368 {Opt_inode_readahead_blks, 0, MOPT_GTE0},
1369 {Opt_init_itable, 0, MOPT_GTE0},
1370 {Opt_stripe, 0, MOPT_GTE0},
1371 {Opt_resuid, 0, MOPT_GTE0},
1372 {Opt_resgid, 0, MOPT_GTE0},
1373 {Opt_journal_dev, 0, MOPT_GTE0},
1374 {Opt_journal_ioprio, 0, MOPT_GTE0},
1375 {Opt_data_journal, EXT4_MOUNT_JOURNAL_DATA, MOPT_NO_EXT2 | MOPT_DATAJ},
1376 {Opt_data_ordered, EXT4_MOUNT_ORDERED_DATA, MOPT_NO_EXT2 | MOPT_DATAJ},
1377 {Opt_data_writeback, EXT4_MOUNT_WRITEBACK_DATA,
1378 MOPT_NO_EXT2 | MOPT_DATAJ},
1379 {Opt_user_xattr, EXT4_MOUNT_XATTR_USER, MOPT_SET},
1380 {Opt_nouser_xattr, EXT4_MOUNT_XATTR_USER, MOPT_CLEAR},
1381 #ifdef CONFIG_EXT4_FS_POSIX_ACL
1382 {Opt_acl, EXT4_MOUNT_POSIX_ACL, MOPT_SET},
1383 {Opt_noacl, EXT4_MOUNT_POSIX_ACL, MOPT_CLEAR},
1385 {Opt_acl, 0, MOPT_NOSUPPORT},
1386 {Opt_noacl, 0, MOPT_NOSUPPORT},
1388 {Opt_nouid32, EXT4_MOUNT_NO_UID32, MOPT_SET},
1389 {Opt_debug, EXT4_MOUNT_DEBUG, MOPT_SET},
1390 {Opt_quota, EXT4_MOUNT_QUOTA | EXT4_MOUNT_USRQUOTA, MOPT_SET | MOPT_Q},
1391 {Opt_usrquota, EXT4_MOUNT_QUOTA | EXT4_MOUNT_USRQUOTA,
1393 {Opt_grpquota, EXT4_MOUNT_QUOTA | EXT4_MOUNT_GRPQUOTA,
1395 {Opt_noquota, (EXT4_MOUNT_QUOTA | EXT4_MOUNT_USRQUOTA |
1396 EXT4_MOUNT_GRPQUOTA), MOPT_CLEAR | MOPT_Q},
1397 {Opt_usrjquota, 0, MOPT_Q},
1398 {Opt_grpjquota, 0, MOPT_Q},
1399 {Opt_offusrjquota, 0, MOPT_Q},
1400 {Opt_offgrpjquota, 0, MOPT_Q},
1401 {Opt_jqfmt_vfsold, QFMT_VFS_OLD, MOPT_QFMT},
1402 {Opt_jqfmt_vfsv0, QFMT_VFS_V0, MOPT_QFMT},
1403 {Opt_jqfmt_vfsv1, QFMT_VFS_V1, MOPT_QFMT},
1404 {Opt_max_dir_size_kb, 0, MOPT_GTE0},
1408 static int handle_mount_opt(struct super_block *sb, char *opt, int token,
1409 substring_t *args, unsigned long *journal_devnum,
1410 unsigned int *journal_ioprio, int is_remount)
1412 struct ext4_sb_info *sbi = EXT4_SB(sb);
1413 const struct mount_opts *m;
1419 if (token == Opt_usrjquota)
1420 return set_qf_name(sb, USRQUOTA, &args[0]);
1421 else if (token == Opt_grpjquota)
1422 return set_qf_name(sb, GRPQUOTA, &args[0]);
1423 else if (token == Opt_offusrjquota)
1424 return clear_qf_name(sb, USRQUOTA);
1425 else if (token == Opt_offgrpjquota)
1426 return clear_qf_name(sb, GRPQUOTA);
1430 case Opt_nouser_xattr:
1431 ext4_msg(sb, KERN_WARNING, deprecated_msg, opt, "3.5");
1434 return 1; /* handled by get_sb_block() */
1436 ext4_msg(sb, KERN_WARNING, "Ignoring removed %s option", opt);
1439 sbi->s_mount_flags |= EXT4_MF_FS_ABORTED;
1442 sb->s_flags |= MS_I_VERSION;
1446 for (m = ext4_mount_opts; m->token != Opt_err; m++)
1447 if (token == m->token)
1450 if (m->token == Opt_err) {
1451 ext4_msg(sb, KERN_ERR, "Unrecognized mount option \"%s\" "
1452 "or missing value", opt);
1456 if ((m->flags & MOPT_NO_EXT2) && IS_EXT2_SB(sb)) {
1457 ext4_msg(sb, KERN_ERR,
1458 "Mount option \"%s\" incompatible with ext2", opt);
1461 if ((m->flags & MOPT_NO_EXT3) && IS_EXT3_SB(sb)) {
1462 ext4_msg(sb, KERN_ERR,
1463 "Mount option \"%s\" incompatible with ext3", opt);
1467 if (args->from && match_int(args, &arg))
1469 if (args->from && (m->flags & MOPT_GTE0) && (arg < 0))
1471 if (m->flags & MOPT_EXPLICIT)
1472 set_opt2(sb, EXPLICIT_DELALLOC);
1473 if (m->flags & MOPT_CLEAR_ERR)
1474 clear_opt(sb, ERRORS_MASK);
1475 if (token == Opt_noquota && sb_any_quota_loaded(sb)) {
1476 ext4_msg(sb, KERN_ERR, "Cannot change quota "
1477 "options when quota turned on");
1481 if (m->flags & MOPT_NOSUPPORT) {
1482 ext4_msg(sb, KERN_ERR, "%s option not supported", opt);
1483 } else if (token == Opt_commit) {
1485 arg = JBD2_DEFAULT_MAX_COMMIT_AGE;
1486 sbi->s_commit_interval = HZ * arg;
1487 } else if (token == Opt_max_batch_time) {
1489 arg = EXT4_DEF_MAX_BATCH_TIME;
1490 sbi->s_max_batch_time = arg;
1491 } else if (token == Opt_min_batch_time) {
1492 sbi->s_min_batch_time = arg;
1493 } else if (token == Opt_inode_readahead_blks) {
1494 if (arg && (arg > (1 << 30) || !is_power_of_2(arg))) {
1495 ext4_msg(sb, KERN_ERR,
1496 "EXT4-fs: inode_readahead_blks must be "
1497 "0 or a power of 2 smaller than 2^31");
1500 sbi->s_inode_readahead_blks = arg;
1501 } else if (token == Opt_init_itable) {
1502 set_opt(sb, INIT_INODE_TABLE);
1504 arg = EXT4_DEF_LI_WAIT_MULT;
1505 sbi->s_li_wait_mult = arg;
1506 } else if (token == Opt_max_dir_size_kb) {
1507 sbi->s_max_dir_size_kb = arg;
1508 } else if (token == Opt_stripe) {
1509 sbi->s_stripe = arg;
1510 } else if (token == Opt_resuid) {
1511 uid = make_kuid(current_user_ns(), arg);
1512 if (!uid_valid(uid)) {
1513 ext4_msg(sb, KERN_ERR, "Invalid uid value %d", arg);
1516 sbi->s_resuid = uid;
1517 } else if (token == Opt_resgid) {
1518 gid = make_kgid(current_user_ns(), arg);
1519 if (!gid_valid(gid)) {
1520 ext4_msg(sb, KERN_ERR, "Invalid gid value %d", arg);
1523 sbi->s_resgid = gid;
1524 } else if (token == Opt_journal_dev) {
1526 ext4_msg(sb, KERN_ERR,
1527 "Cannot specify journal on remount");
1530 *journal_devnum = arg;
1531 } else if (token == Opt_journal_ioprio) {
1533 ext4_msg(sb, KERN_ERR, "Invalid journal IO priority"
1538 IOPRIO_PRIO_VALUE(IOPRIO_CLASS_BE, arg);
1539 } else if (m->flags & MOPT_DATAJ) {
1541 if (!sbi->s_journal)
1542 ext4_msg(sb, KERN_WARNING, "Remounting file system with no journal so ignoring journalled data option");
1543 else if (test_opt(sb, DATA_FLAGS) != m->mount_opt) {
1544 ext4_msg(sb, KERN_ERR,
1545 "Cannot change data mode on remount");
1549 clear_opt(sb, DATA_FLAGS);
1550 sbi->s_mount_opt |= m->mount_opt;
1553 } else if (m->flags & MOPT_QFMT) {
1554 if (sb_any_quota_loaded(sb) &&
1555 sbi->s_jquota_fmt != m->mount_opt) {
1556 ext4_msg(sb, KERN_ERR, "Cannot change journaled "
1557 "quota options when quota turned on");
1560 if (EXT4_HAS_RO_COMPAT_FEATURE(sb,
1561 EXT4_FEATURE_RO_COMPAT_QUOTA)) {
1562 ext4_msg(sb, KERN_ERR,
1563 "Cannot set journaled quota options "
1564 "when QUOTA feature is enabled");
1567 sbi->s_jquota_fmt = m->mount_opt;
1572 if (m->flags & MOPT_CLEAR)
1574 else if (unlikely(!(m->flags & MOPT_SET))) {
1575 ext4_msg(sb, KERN_WARNING,
1576 "buggy handling of option %s", opt);
1581 sbi->s_mount_opt |= m->mount_opt;
1583 sbi->s_mount_opt &= ~m->mount_opt;
1588 static int parse_options(char *options, struct super_block *sb,
1589 unsigned long *journal_devnum,
1590 unsigned int *journal_ioprio,
1593 struct ext4_sb_info *sbi = EXT4_SB(sb);
1595 substring_t args[MAX_OPT_ARGS];
1601 while ((p = strsep(&options, ",")) != NULL) {
1605 * Initialize args struct so we know whether arg was
1606 * found; some options take optional arguments.
1608 args[0].to = args[0].from = NULL;
1609 token = match_token(p, tokens, args);
1610 if (handle_mount_opt(sb, p, token, args, journal_devnum,
1611 journal_ioprio, is_remount) < 0)
1615 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_QUOTA) &&
1616 (test_opt(sb, USRQUOTA) || test_opt(sb, GRPQUOTA))) {
1617 ext4_msg(sb, KERN_ERR, "Cannot set quota options when QUOTA "
1618 "feature is enabled");
1621 if (sbi->s_qf_names[USRQUOTA] || sbi->s_qf_names[GRPQUOTA]) {
1622 if (test_opt(sb, USRQUOTA) && sbi->s_qf_names[USRQUOTA])
1623 clear_opt(sb, USRQUOTA);
1625 if (test_opt(sb, GRPQUOTA) && sbi->s_qf_names[GRPQUOTA])
1626 clear_opt(sb, GRPQUOTA);
1628 if (test_opt(sb, GRPQUOTA) || test_opt(sb, USRQUOTA)) {
1629 ext4_msg(sb, KERN_ERR, "old and new quota "
1634 if (!sbi->s_jquota_fmt) {
1635 ext4_msg(sb, KERN_ERR, "journaled quota format "
1640 if (sbi->s_jquota_fmt) {
1641 ext4_msg(sb, KERN_ERR, "journaled quota format "
1642 "specified with no journaling "
1648 if (test_opt(sb, DIOREAD_NOLOCK)) {
1650 BLOCK_SIZE << le32_to_cpu(sbi->s_es->s_log_block_size);
1652 if (blocksize < PAGE_CACHE_SIZE) {
1653 ext4_msg(sb, KERN_ERR, "can't mount with "
1654 "dioread_nolock if block size != PAGE_SIZE");
1661 static inline void ext4_show_quota_options(struct seq_file *seq,
1662 struct super_block *sb)
1664 #if defined(CONFIG_QUOTA)
1665 struct ext4_sb_info *sbi = EXT4_SB(sb);
1667 if (sbi->s_jquota_fmt) {
1670 switch (sbi->s_jquota_fmt) {
1681 seq_printf(seq, ",jqfmt=%s", fmtname);
1684 if (sbi->s_qf_names[USRQUOTA])
1685 seq_printf(seq, ",usrjquota=%s", sbi->s_qf_names[USRQUOTA]);
1687 if (sbi->s_qf_names[GRPQUOTA])
1688 seq_printf(seq, ",grpjquota=%s", sbi->s_qf_names[GRPQUOTA]);
1690 if (test_opt(sb, USRQUOTA))
1691 seq_puts(seq, ",usrquota");
1693 if (test_opt(sb, GRPQUOTA))
1694 seq_puts(seq, ",grpquota");
1698 static const char *token2str(int token)
1700 const struct match_token *t;
1702 for (t = tokens; t->token != Opt_err; t++)
1703 if (t->token == token && !strchr(t->pattern, '='))
1710 * - it's set to a non-default value OR
1711 * - if the per-sb default is different from the global default
1713 static int _ext4_show_options(struct seq_file *seq, struct super_block *sb,
1716 struct ext4_sb_info *sbi = EXT4_SB(sb);
1717 struct ext4_super_block *es = sbi->s_es;
1718 int def_errors, def_mount_opt = nodefs ? 0 : sbi->s_def_mount_opt;
1719 const struct mount_opts *m;
1720 char sep = nodefs ? '\n' : ',';
1722 #define SEQ_OPTS_PUTS(str) seq_printf(seq, "%c" str, sep)
1723 #define SEQ_OPTS_PRINT(str, arg) seq_printf(seq, "%c" str, sep, arg)
1725 if (sbi->s_sb_block != 1)
1726 SEQ_OPTS_PRINT("sb=%llu", sbi->s_sb_block);
1728 for (m = ext4_mount_opts; m->token != Opt_err; m++) {
1729 int want_set = m->flags & MOPT_SET;
1730 if (((m->flags & (MOPT_SET|MOPT_CLEAR)) == 0) ||
1731 (m->flags & MOPT_CLEAR_ERR))
1733 if (!(m->mount_opt & (sbi->s_mount_opt ^ def_mount_opt)))
1734 continue; /* skip if same as the default */
1736 (sbi->s_mount_opt & m->mount_opt) != m->mount_opt) ||
1737 (!want_set && (sbi->s_mount_opt & m->mount_opt)))
1738 continue; /* select Opt_noFoo vs Opt_Foo */
1739 SEQ_OPTS_PRINT("%s", token2str(m->token));
1742 if (nodefs || !uid_eq(sbi->s_resuid, make_kuid(&init_user_ns, EXT4_DEF_RESUID)) ||
1743 le16_to_cpu(es->s_def_resuid) != EXT4_DEF_RESUID)
1744 SEQ_OPTS_PRINT("resuid=%u",
1745 from_kuid_munged(&init_user_ns, sbi->s_resuid));
1746 if (nodefs || !gid_eq(sbi->s_resgid, make_kgid(&init_user_ns, EXT4_DEF_RESGID)) ||
1747 le16_to_cpu(es->s_def_resgid) != EXT4_DEF_RESGID)
1748 SEQ_OPTS_PRINT("resgid=%u",
1749 from_kgid_munged(&init_user_ns, sbi->s_resgid));
1750 def_errors = nodefs ? -1 : le16_to_cpu(es->s_errors);
1751 if (test_opt(sb, ERRORS_RO) && def_errors != EXT4_ERRORS_RO)
1752 SEQ_OPTS_PUTS("errors=remount-ro");
1753 if (test_opt(sb, ERRORS_CONT) && def_errors != EXT4_ERRORS_CONTINUE)
1754 SEQ_OPTS_PUTS("errors=continue");
1755 if (test_opt(sb, ERRORS_PANIC) && def_errors != EXT4_ERRORS_PANIC)
1756 SEQ_OPTS_PUTS("errors=panic");
1757 if (nodefs || sbi->s_commit_interval != JBD2_DEFAULT_MAX_COMMIT_AGE*HZ)
1758 SEQ_OPTS_PRINT("commit=%lu", sbi->s_commit_interval / HZ);
1759 if (nodefs || sbi->s_min_batch_time != EXT4_DEF_MIN_BATCH_TIME)
1760 SEQ_OPTS_PRINT("min_batch_time=%u", sbi->s_min_batch_time);
1761 if (nodefs || sbi->s_max_batch_time != EXT4_DEF_MAX_BATCH_TIME)
1762 SEQ_OPTS_PRINT("max_batch_time=%u", sbi->s_max_batch_time);
1763 if (sb->s_flags & MS_I_VERSION)
1764 SEQ_OPTS_PUTS("i_version");
1765 if (nodefs || sbi->s_stripe)
1766 SEQ_OPTS_PRINT("stripe=%lu", sbi->s_stripe);
1767 if (EXT4_MOUNT_DATA_FLAGS & (sbi->s_mount_opt ^ def_mount_opt)) {
1768 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA)
1769 SEQ_OPTS_PUTS("data=journal");
1770 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA)
1771 SEQ_OPTS_PUTS("data=ordered");
1772 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_WRITEBACK_DATA)
1773 SEQ_OPTS_PUTS("data=writeback");
1776 sbi->s_inode_readahead_blks != EXT4_DEF_INODE_READAHEAD_BLKS)
1777 SEQ_OPTS_PRINT("inode_readahead_blks=%u",
1778 sbi->s_inode_readahead_blks);
1780 if (nodefs || (test_opt(sb, INIT_INODE_TABLE) &&
1781 (sbi->s_li_wait_mult != EXT4_DEF_LI_WAIT_MULT)))
1782 SEQ_OPTS_PRINT("init_itable=%u", sbi->s_li_wait_mult);
1783 if (nodefs || sbi->s_max_dir_size_kb)
1784 SEQ_OPTS_PRINT("max_dir_size_kb=%u", sbi->s_max_dir_size_kb);
1786 ext4_show_quota_options(seq, sb);
1790 static int ext4_show_options(struct seq_file *seq, struct dentry *root)
1792 return _ext4_show_options(seq, root->d_sb, 0);
1795 static int options_seq_show(struct seq_file *seq, void *offset)
1797 struct super_block *sb = seq->private;
1800 seq_puts(seq, (sb->s_flags & MS_RDONLY) ? "ro" : "rw");
1801 rc = _ext4_show_options(seq, sb, 1);
1802 seq_puts(seq, "\n");
1806 static int options_open_fs(struct inode *inode, struct file *file)
1808 return single_open(file, options_seq_show, PDE(inode)->data);
1811 static const struct file_operations ext4_seq_options_fops = {
1812 .owner = THIS_MODULE,
1813 .open = options_open_fs,
1815 .llseek = seq_lseek,
1816 .release = single_release,
1819 static int ext4_setup_super(struct super_block *sb, struct ext4_super_block *es,
1822 struct ext4_sb_info *sbi = EXT4_SB(sb);
1825 if (le32_to_cpu(es->s_rev_level) > EXT4_MAX_SUPP_REV) {
1826 ext4_msg(sb, KERN_ERR, "revision level too high, "
1827 "forcing read-only mode");
1832 if (!(sbi->s_mount_state & EXT4_VALID_FS))
1833 ext4_msg(sb, KERN_WARNING, "warning: mounting unchecked fs, "
1834 "running e2fsck is recommended");
1835 else if ((sbi->s_mount_state & EXT4_ERROR_FS))
1836 ext4_msg(sb, KERN_WARNING,
1837 "warning: mounting fs with errors, "
1838 "running e2fsck is recommended");
1839 else if ((__s16) le16_to_cpu(es->s_max_mnt_count) > 0 &&
1840 le16_to_cpu(es->s_mnt_count) >=
1841 (unsigned short) (__s16) le16_to_cpu(es->s_max_mnt_count))
1842 ext4_msg(sb, KERN_WARNING,
1843 "warning: maximal mount count reached, "
1844 "running e2fsck is recommended");
1845 else if (le32_to_cpu(es->s_checkinterval) &&
1846 (le32_to_cpu(es->s_lastcheck) +
1847 le32_to_cpu(es->s_checkinterval) <= get_seconds()))
1848 ext4_msg(sb, KERN_WARNING,
1849 "warning: checktime reached, "
1850 "running e2fsck is recommended");
1851 if (!sbi->s_journal)
1852 es->s_state &= cpu_to_le16(~EXT4_VALID_FS);
1853 if (!(__s16) le16_to_cpu(es->s_max_mnt_count))
1854 es->s_max_mnt_count = cpu_to_le16(EXT4_DFL_MAX_MNT_COUNT);
1855 le16_add_cpu(&es->s_mnt_count, 1);
1856 es->s_mtime = cpu_to_le32(get_seconds());
1857 ext4_update_dynamic_rev(sb);
1859 EXT4_SET_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
1861 ext4_commit_super(sb, 1);
1863 if (test_opt(sb, DEBUG))
1864 printk(KERN_INFO "[EXT4 FS bs=%lu, gc=%u, "
1865 "bpg=%lu, ipg=%lu, mo=%04x, mo2=%04x]\n",
1867 sbi->s_groups_count,
1868 EXT4_BLOCKS_PER_GROUP(sb),
1869 EXT4_INODES_PER_GROUP(sb),
1870 sbi->s_mount_opt, sbi->s_mount_opt2);
1872 cleancache_init_fs(sb);
1876 int ext4_alloc_flex_bg_array(struct super_block *sb, ext4_group_t ngroup)
1878 struct ext4_sb_info *sbi = EXT4_SB(sb);
1879 struct flex_groups *new_groups;
1882 if (!sbi->s_log_groups_per_flex)
1885 size = ext4_flex_group(sbi, ngroup - 1) + 1;
1886 if (size <= sbi->s_flex_groups_allocated)
1889 size = roundup_pow_of_two(size * sizeof(struct flex_groups));
1890 new_groups = ext4_kvzalloc(size, GFP_KERNEL);
1892 ext4_msg(sb, KERN_ERR, "not enough memory for %d flex groups",
1893 size / (int) sizeof(struct flex_groups));
1897 if (sbi->s_flex_groups) {
1898 memcpy(new_groups, sbi->s_flex_groups,
1899 (sbi->s_flex_groups_allocated *
1900 sizeof(struct flex_groups)));
1901 ext4_kvfree(sbi->s_flex_groups);
1903 sbi->s_flex_groups = new_groups;
1904 sbi->s_flex_groups_allocated = size / sizeof(struct flex_groups);
1908 static int ext4_fill_flex_info(struct super_block *sb)
1910 struct ext4_sb_info *sbi = EXT4_SB(sb);
1911 struct ext4_group_desc *gdp = NULL;
1912 ext4_group_t flex_group;
1913 unsigned int groups_per_flex = 0;
1916 sbi->s_log_groups_per_flex = sbi->s_es->s_log_groups_per_flex;
1917 if (sbi->s_log_groups_per_flex < 1 || sbi->s_log_groups_per_flex > 31) {
1918 sbi->s_log_groups_per_flex = 0;
1921 groups_per_flex = 1U << sbi->s_log_groups_per_flex;
1923 err = ext4_alloc_flex_bg_array(sb, sbi->s_groups_count);
1927 for (i = 0; i < sbi->s_groups_count; i++) {
1928 gdp = ext4_get_group_desc(sb, i, NULL);
1930 flex_group = ext4_flex_group(sbi, i);
1931 atomic_add(ext4_free_inodes_count(sb, gdp),
1932 &sbi->s_flex_groups[flex_group].free_inodes);
1933 atomic64_add(ext4_free_group_clusters(sb, gdp),
1934 &sbi->s_flex_groups[flex_group].free_clusters);
1935 atomic_add(ext4_used_dirs_count(sb, gdp),
1936 &sbi->s_flex_groups[flex_group].used_dirs);
1944 static __le16 ext4_group_desc_csum(struct ext4_sb_info *sbi, __u32 block_group,
1945 struct ext4_group_desc *gdp)
1949 __le32 le_group = cpu_to_le32(block_group);
1951 if ((sbi->s_es->s_feature_ro_compat &
1952 cpu_to_le32(EXT4_FEATURE_RO_COMPAT_METADATA_CSUM))) {
1953 /* Use new metadata_csum algorithm */
1957 old_csum = gdp->bg_checksum;
1958 gdp->bg_checksum = 0;
1959 csum32 = ext4_chksum(sbi, sbi->s_csum_seed, (__u8 *)&le_group,
1961 csum32 = ext4_chksum(sbi, csum32, (__u8 *)gdp,
1963 gdp->bg_checksum = old_csum;
1965 crc = csum32 & 0xFFFF;
1969 /* old crc16 code */
1970 offset = offsetof(struct ext4_group_desc, bg_checksum);
1972 crc = crc16(~0, sbi->s_es->s_uuid, sizeof(sbi->s_es->s_uuid));
1973 crc = crc16(crc, (__u8 *)&le_group, sizeof(le_group));
1974 crc = crc16(crc, (__u8 *)gdp, offset);
1975 offset += sizeof(gdp->bg_checksum); /* skip checksum */
1976 /* for checksum of struct ext4_group_desc do the rest...*/
1977 if ((sbi->s_es->s_feature_incompat &
1978 cpu_to_le32(EXT4_FEATURE_INCOMPAT_64BIT)) &&
1979 offset < le16_to_cpu(sbi->s_es->s_desc_size))
1980 crc = crc16(crc, (__u8 *)gdp + offset,
1981 le16_to_cpu(sbi->s_es->s_desc_size) -
1985 return cpu_to_le16(crc);
1988 int ext4_group_desc_csum_verify(struct super_block *sb, __u32 block_group,
1989 struct ext4_group_desc *gdp)
1991 if (ext4_has_group_desc_csum(sb) &&
1992 (gdp->bg_checksum != ext4_group_desc_csum(EXT4_SB(sb),
1999 void ext4_group_desc_csum_set(struct super_block *sb, __u32 block_group,
2000 struct ext4_group_desc *gdp)
2002 if (!ext4_has_group_desc_csum(sb))
2004 gdp->bg_checksum = ext4_group_desc_csum(EXT4_SB(sb), block_group, gdp);
2007 /* Called at mount-time, super-block is locked */
2008 static int ext4_check_descriptors(struct super_block *sb,
2009 ext4_group_t *first_not_zeroed)
2011 struct ext4_sb_info *sbi = EXT4_SB(sb);
2012 ext4_fsblk_t first_block = le32_to_cpu(sbi->s_es->s_first_data_block);
2013 ext4_fsblk_t last_block;
2014 ext4_fsblk_t block_bitmap;
2015 ext4_fsblk_t inode_bitmap;
2016 ext4_fsblk_t inode_table;
2017 int flexbg_flag = 0;
2018 ext4_group_t i, grp = sbi->s_groups_count;
2020 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_FLEX_BG))
2023 ext4_debug("Checking group descriptors");
2025 for (i = 0; i < sbi->s_groups_count; i++) {
2026 struct ext4_group_desc *gdp = ext4_get_group_desc(sb, i, NULL);
2028 if (i == sbi->s_groups_count - 1 || flexbg_flag)
2029 last_block = ext4_blocks_count(sbi->s_es) - 1;
2031 last_block = first_block +
2032 (EXT4_BLOCKS_PER_GROUP(sb) - 1);
2034 if ((grp == sbi->s_groups_count) &&
2035 !(gdp->bg_flags & cpu_to_le16(EXT4_BG_INODE_ZEROED)))
2038 block_bitmap = ext4_block_bitmap(sb, gdp);
2039 if (block_bitmap < first_block || block_bitmap > last_block) {
2040 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2041 "Block bitmap for group %u not in group "
2042 "(block %llu)!", i, block_bitmap);
2045 inode_bitmap = ext4_inode_bitmap(sb, gdp);
2046 if (inode_bitmap < first_block || inode_bitmap > last_block) {
2047 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2048 "Inode bitmap for group %u not in group "
2049 "(block %llu)!", i, inode_bitmap);
2052 inode_table = ext4_inode_table(sb, gdp);
2053 if (inode_table < first_block ||
2054 inode_table + sbi->s_itb_per_group - 1 > last_block) {
2055 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2056 "Inode table for group %u not in group "
2057 "(block %llu)!", i, inode_table);
2060 ext4_lock_group(sb, i);
2061 if (!ext4_group_desc_csum_verify(sb, i, gdp)) {
2062 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2063 "Checksum for group %u failed (%u!=%u)",
2064 i, le16_to_cpu(ext4_group_desc_csum(sbi, i,
2065 gdp)), le16_to_cpu(gdp->bg_checksum));
2066 if (!(sb->s_flags & MS_RDONLY)) {
2067 ext4_unlock_group(sb, i);
2071 ext4_unlock_group(sb, i);
2073 first_block += EXT4_BLOCKS_PER_GROUP(sb);
2075 if (NULL != first_not_zeroed)
2076 *first_not_zeroed = grp;
2078 ext4_free_blocks_count_set(sbi->s_es,
2079 EXT4_C2B(sbi, ext4_count_free_clusters(sb)));
2080 sbi->s_es->s_free_inodes_count =cpu_to_le32(ext4_count_free_inodes(sb));
2084 /* ext4_orphan_cleanup() walks a singly-linked list of inodes (starting at
2085 * the superblock) which were deleted from all directories, but held open by
2086 * a process at the time of a crash. We walk the list and try to delete these
2087 * inodes at recovery time (only with a read-write filesystem).
2089 * In order to keep the orphan inode chain consistent during traversal (in
2090 * case of crash during recovery), we link each inode into the superblock
2091 * orphan list_head and handle it the same way as an inode deletion during
2092 * normal operation (which journals the operations for us).
2094 * We only do an iget() and an iput() on each inode, which is very safe if we
2095 * accidentally point at an in-use or already deleted inode. The worst that
2096 * can happen in this case is that we get a "bit already cleared" message from
2097 * ext4_free_inode(). The only reason we would point at a wrong inode is if
2098 * e2fsck was run on this filesystem, and it must have already done the orphan
2099 * inode cleanup for us, so we can safely abort without any further action.
2101 static void ext4_orphan_cleanup(struct super_block *sb,
2102 struct ext4_super_block *es)
2104 unsigned int s_flags = sb->s_flags;
2105 int nr_orphans = 0, nr_truncates = 0;
2109 if (!es->s_last_orphan) {
2110 jbd_debug(4, "no orphan inodes to clean up\n");
2114 if (bdev_read_only(sb->s_bdev)) {
2115 ext4_msg(sb, KERN_ERR, "write access "
2116 "unavailable, skipping orphan cleanup");
2120 /* Check if feature set would not allow a r/w mount */
2121 if (!ext4_feature_set_ok(sb, 0)) {
2122 ext4_msg(sb, KERN_INFO, "Skipping orphan cleanup due to "
2123 "unknown ROCOMPAT features");
2127 if (EXT4_SB(sb)->s_mount_state & EXT4_ERROR_FS) {
2128 /* don't clear list on RO mount w/ errors */
2129 if (es->s_last_orphan && !(s_flags & MS_RDONLY)) {
2130 jbd_debug(1, "Errors on filesystem, "
2131 "clearing orphan list.\n");
2132 es->s_last_orphan = 0;
2134 jbd_debug(1, "Skipping orphan recovery on fs with errors.\n");
2138 if (s_flags & MS_RDONLY) {
2139 ext4_msg(sb, KERN_INFO, "orphan cleanup on readonly fs");
2140 sb->s_flags &= ~MS_RDONLY;
2143 /* Needed for iput() to work correctly and not trash data */
2144 sb->s_flags |= MS_ACTIVE;
2145 /* Turn on quotas so that they are updated correctly */
2146 for (i = 0; i < MAXQUOTAS; i++) {
2147 if (EXT4_SB(sb)->s_qf_names[i]) {
2148 int ret = ext4_quota_on_mount(sb, i);
2150 ext4_msg(sb, KERN_ERR,
2151 "Cannot turn on journaled "
2152 "quota: error %d", ret);
2157 while (es->s_last_orphan) {
2158 struct inode *inode;
2160 inode = ext4_orphan_get(sb, le32_to_cpu(es->s_last_orphan));
2161 if (IS_ERR(inode)) {
2162 es->s_last_orphan = 0;
2166 list_add(&EXT4_I(inode)->i_orphan, &EXT4_SB(sb)->s_orphan);
2167 dquot_initialize(inode);
2168 if (inode->i_nlink) {
2169 ext4_msg(sb, KERN_DEBUG,
2170 "%s: truncating inode %lu to %lld bytes",
2171 __func__, inode->i_ino, inode->i_size);
2172 jbd_debug(2, "truncating inode %lu to %lld bytes\n",
2173 inode->i_ino, inode->i_size);
2174 mutex_lock(&inode->i_mutex);
2175 ext4_truncate(inode);
2176 mutex_unlock(&inode->i_mutex);
2179 ext4_msg(sb, KERN_DEBUG,
2180 "%s: deleting unreferenced inode %lu",
2181 __func__, inode->i_ino);
2182 jbd_debug(2, "deleting unreferenced inode %lu\n",
2186 iput(inode); /* The delete magic happens here! */
2189 #define PLURAL(x) (x), ((x) == 1) ? "" : "s"
2192 ext4_msg(sb, KERN_INFO, "%d orphan inode%s deleted",
2193 PLURAL(nr_orphans));
2195 ext4_msg(sb, KERN_INFO, "%d truncate%s cleaned up",
2196 PLURAL(nr_truncates));
2198 /* Turn quotas off */
2199 for (i = 0; i < MAXQUOTAS; i++) {
2200 if (sb_dqopt(sb)->files[i])
2201 dquot_quota_off(sb, i);
2204 sb->s_flags = s_flags; /* Restore MS_RDONLY status */
2208 * Maximal extent format file size.
2209 * Resulting logical blkno at s_maxbytes must fit in our on-disk
2210 * extent format containers, within a sector_t, and within i_blocks
2211 * in the vfs. ext4 inode has 48 bits of i_block in fsblock units,
2212 * so that won't be a limiting factor.
2214 * However there is other limiting factor. We do store extents in the form
2215 * of starting block and length, hence the resulting length of the extent
2216 * covering maximum file size must fit into on-disk format containers as
2217 * well. Given that length is always by 1 unit bigger than max unit (because
2218 * we count 0 as well) we have to lower the s_maxbytes by one fs block.
2220 * Note, this does *not* consider any metadata overhead for vfs i_blocks.
2222 static loff_t ext4_max_size(int blkbits, int has_huge_files)
2225 loff_t upper_limit = MAX_LFS_FILESIZE;
2227 /* small i_blocks in vfs inode? */
2228 if (!has_huge_files || sizeof(blkcnt_t) < sizeof(u64)) {
2230 * CONFIG_LBDAF is not enabled implies the inode
2231 * i_block represent total blocks in 512 bytes
2232 * 32 == size of vfs inode i_blocks * 8
2234 upper_limit = (1LL << 32) - 1;
2236 /* total blocks in file system block size */
2237 upper_limit >>= (blkbits - 9);
2238 upper_limit <<= blkbits;
2242 * 32-bit extent-start container, ee_block. We lower the maxbytes
2243 * by one fs block, so ee_len can cover the extent of maximum file
2246 res = (1LL << 32) - 1;
2249 /* Sanity check against vm- & vfs- imposed limits */
2250 if (res > upper_limit)
2257 * Maximal bitmap file size. There is a direct, and {,double-,triple-}indirect
2258 * block limit, and also a limit of (2^48 - 1) 512-byte sectors in i_blocks.
2259 * We need to be 1 filesystem block less than the 2^48 sector limit.
2261 static loff_t ext4_max_bitmap_size(int bits, int has_huge_files)
2263 loff_t res = EXT4_NDIR_BLOCKS;
2266 /* This is calculated to be the largest file size for a dense, block
2267 * mapped file such that the file's total number of 512-byte sectors,
2268 * including data and all indirect blocks, does not exceed (2^48 - 1).
2270 * __u32 i_blocks_lo and _u16 i_blocks_high represent the total
2271 * number of 512-byte sectors of the file.
2274 if (!has_huge_files || sizeof(blkcnt_t) < sizeof(u64)) {
2276 * !has_huge_files or CONFIG_LBDAF not enabled implies that
2277 * the inode i_block field represents total file blocks in
2278 * 2^32 512-byte sectors == size of vfs inode i_blocks * 8
2280 upper_limit = (1LL << 32) - 1;
2282 /* total blocks in file system block size */
2283 upper_limit >>= (bits - 9);
2287 * We use 48 bit ext4_inode i_blocks
2288 * With EXT4_HUGE_FILE_FL set the i_blocks
2289 * represent total number of blocks in
2290 * file system block size
2292 upper_limit = (1LL << 48) - 1;
2296 /* indirect blocks */
2298 /* double indirect blocks */
2299 meta_blocks += 1 + (1LL << (bits-2));
2300 /* tripple indirect blocks */
2301 meta_blocks += 1 + (1LL << (bits-2)) + (1LL << (2*(bits-2)));
2303 upper_limit -= meta_blocks;
2304 upper_limit <<= bits;
2306 res += 1LL << (bits-2);
2307 res += 1LL << (2*(bits-2));
2308 res += 1LL << (3*(bits-2));
2310 if (res > upper_limit)
2313 if (res > MAX_LFS_FILESIZE)
2314 res = MAX_LFS_FILESIZE;
2319 static ext4_fsblk_t descriptor_loc(struct super_block *sb,
2320 ext4_fsblk_t logical_sb_block, int nr)
2322 struct ext4_sb_info *sbi = EXT4_SB(sb);
2323 ext4_group_t bg, first_meta_bg;
2326 first_meta_bg = le32_to_cpu(sbi->s_es->s_first_meta_bg);
2328 if (!EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_META_BG) ||
2330 return logical_sb_block + nr + 1;
2331 bg = sbi->s_desc_per_block * nr;
2332 if (ext4_bg_has_super(sb, bg))
2335 return (has_super + ext4_group_first_block_no(sb, bg));
2339 * ext4_get_stripe_size: Get the stripe size.
2340 * @sbi: In memory super block info
2342 * If we have specified it via mount option, then
2343 * use the mount option value. If the value specified at mount time is
2344 * greater than the blocks per group use the super block value.
2345 * If the super block value is greater than blocks per group return 0.
2346 * Allocator needs it be less than blocks per group.
2349 static unsigned long ext4_get_stripe_size(struct ext4_sb_info *sbi)
2351 unsigned long stride = le16_to_cpu(sbi->s_es->s_raid_stride);
2352 unsigned long stripe_width =
2353 le32_to_cpu(sbi->s_es->s_raid_stripe_width);
2356 if (sbi->s_stripe && sbi->s_stripe <= sbi->s_blocks_per_group)
2357 ret = sbi->s_stripe;
2358 else if (stripe_width <= sbi->s_blocks_per_group)
2360 else if (stride <= sbi->s_blocks_per_group)
2366 * If the stripe width is 1, this makes no sense and
2367 * we set it to 0 to turn off stripe handling code.
2378 struct attribute attr;
2379 ssize_t (*show)(struct ext4_attr *, struct ext4_sb_info *, char *);
2380 ssize_t (*store)(struct ext4_attr *, struct ext4_sb_info *,
2381 const char *, size_t);
2385 static int parse_strtoul(const char *buf,
2386 unsigned long max, unsigned long *value)
2390 *value = simple_strtoul(skip_spaces(buf), &endp, 0);
2391 endp = skip_spaces(endp);
2392 if (*endp || *value > max)
2398 static ssize_t delayed_allocation_blocks_show(struct ext4_attr *a,
2399 struct ext4_sb_info *sbi,
2402 return snprintf(buf, PAGE_SIZE, "%llu\n",
2404 percpu_counter_sum(&sbi->s_dirtyclusters_counter)));
2407 static ssize_t session_write_kbytes_show(struct ext4_attr *a,
2408 struct ext4_sb_info *sbi, char *buf)
2410 struct super_block *sb = sbi->s_buddy_cache->i_sb;
2412 if (!sb->s_bdev->bd_part)
2413 return snprintf(buf, PAGE_SIZE, "0\n");
2414 return snprintf(buf, PAGE_SIZE, "%lu\n",
2415 (part_stat_read(sb->s_bdev->bd_part, sectors[1]) -
2416 sbi->s_sectors_written_start) >> 1);
2419 static ssize_t lifetime_write_kbytes_show(struct ext4_attr *a,
2420 struct ext4_sb_info *sbi, char *buf)
2422 struct super_block *sb = sbi->s_buddy_cache->i_sb;
2424 if (!sb->s_bdev->bd_part)
2425 return snprintf(buf, PAGE_SIZE, "0\n");
2426 return snprintf(buf, PAGE_SIZE, "%llu\n",
2427 (unsigned long long)(sbi->s_kbytes_written +
2428 ((part_stat_read(sb->s_bdev->bd_part, sectors[1]) -
2429 EXT4_SB(sb)->s_sectors_written_start) >> 1)));
2432 static ssize_t inode_readahead_blks_store(struct ext4_attr *a,
2433 struct ext4_sb_info *sbi,
2434 const char *buf, size_t count)
2438 if (parse_strtoul(buf, 0x40000000, &t))
2441 if (t && !is_power_of_2(t))
2444 sbi->s_inode_readahead_blks = t;
2448 static ssize_t sbi_ui_show(struct ext4_attr *a,
2449 struct ext4_sb_info *sbi, char *buf)
2451 unsigned int *ui = (unsigned int *) (((char *) sbi) + a->offset);
2453 return snprintf(buf, PAGE_SIZE, "%u\n", *ui);
2456 static ssize_t sbi_ui_store(struct ext4_attr *a,
2457 struct ext4_sb_info *sbi,
2458 const char *buf, size_t count)
2460 unsigned int *ui = (unsigned int *) (((char *) sbi) + a->offset);
2463 if (parse_strtoul(buf, 0xffffffff, &t))
2469 static ssize_t trigger_test_error(struct ext4_attr *a,
2470 struct ext4_sb_info *sbi,
2471 const char *buf, size_t count)
2475 if (!capable(CAP_SYS_ADMIN))
2478 if (len && buf[len-1] == '\n')
2482 ext4_error(sbi->s_sb, "%.*s", len, buf);
2486 #define EXT4_ATTR_OFFSET(_name,_mode,_show,_store,_elname) \
2487 static struct ext4_attr ext4_attr_##_name = { \
2488 .attr = {.name = __stringify(_name), .mode = _mode }, \
2491 .offset = offsetof(struct ext4_sb_info, _elname), \
2493 #define EXT4_ATTR(name, mode, show, store) \
2494 static struct ext4_attr ext4_attr_##name = __ATTR(name, mode, show, store)
2496 #define EXT4_INFO_ATTR(name) EXT4_ATTR(name, 0444, NULL, NULL)
2497 #define EXT4_RO_ATTR(name) EXT4_ATTR(name, 0444, name##_show, NULL)
2498 #define EXT4_RW_ATTR(name) EXT4_ATTR(name, 0644, name##_show, name##_store)
2499 #define EXT4_RW_ATTR_SBI_UI(name, elname) \
2500 EXT4_ATTR_OFFSET(name, 0644, sbi_ui_show, sbi_ui_store, elname)
2501 #define ATTR_LIST(name) &ext4_attr_##name.attr
2503 EXT4_RO_ATTR(delayed_allocation_blocks);
2504 EXT4_RO_ATTR(session_write_kbytes);
2505 EXT4_RO_ATTR(lifetime_write_kbytes);
2506 EXT4_ATTR_OFFSET(inode_readahead_blks, 0644, sbi_ui_show,
2507 inode_readahead_blks_store, s_inode_readahead_blks);
2508 EXT4_RW_ATTR_SBI_UI(inode_goal, s_inode_goal);
2509 EXT4_RW_ATTR_SBI_UI(mb_stats, s_mb_stats);
2510 EXT4_RW_ATTR_SBI_UI(mb_max_to_scan, s_mb_max_to_scan);
2511 EXT4_RW_ATTR_SBI_UI(mb_min_to_scan, s_mb_min_to_scan);
2512 EXT4_RW_ATTR_SBI_UI(mb_order2_req, s_mb_order2_reqs);
2513 EXT4_RW_ATTR_SBI_UI(mb_stream_req, s_mb_stream_request);
2514 EXT4_RW_ATTR_SBI_UI(mb_group_prealloc, s_mb_group_prealloc);
2515 EXT4_RW_ATTR_SBI_UI(max_writeback_mb_bump, s_max_writeback_mb_bump);
2516 EXT4_RW_ATTR_SBI_UI(extent_max_zeroout_kb, s_extent_max_zeroout_kb);
2517 EXT4_ATTR(trigger_fs_error, 0200, NULL, trigger_test_error);
2519 static struct attribute *ext4_attrs[] = {
2520 ATTR_LIST(delayed_allocation_blocks),
2521 ATTR_LIST(session_write_kbytes),
2522 ATTR_LIST(lifetime_write_kbytes),
2523 ATTR_LIST(inode_readahead_blks),
2524 ATTR_LIST(inode_goal),
2525 ATTR_LIST(mb_stats),
2526 ATTR_LIST(mb_max_to_scan),
2527 ATTR_LIST(mb_min_to_scan),
2528 ATTR_LIST(mb_order2_req),
2529 ATTR_LIST(mb_stream_req),
2530 ATTR_LIST(mb_group_prealloc),
2531 ATTR_LIST(max_writeback_mb_bump),
2532 ATTR_LIST(extent_max_zeroout_kb),
2533 ATTR_LIST(trigger_fs_error),
2537 /* Features this copy of ext4 supports */
2538 EXT4_INFO_ATTR(lazy_itable_init);
2539 EXT4_INFO_ATTR(batched_discard);
2540 EXT4_INFO_ATTR(meta_bg_resize);
2542 static struct attribute *ext4_feat_attrs[] = {
2543 ATTR_LIST(lazy_itable_init),
2544 ATTR_LIST(batched_discard),
2545 ATTR_LIST(meta_bg_resize),
2549 static ssize_t ext4_attr_show(struct kobject *kobj,
2550 struct attribute *attr, char *buf)
2552 struct ext4_sb_info *sbi = container_of(kobj, struct ext4_sb_info,
2554 struct ext4_attr *a = container_of(attr, struct ext4_attr, attr);
2556 return a->show ? a->show(a, sbi, buf) : 0;
2559 static ssize_t ext4_attr_store(struct kobject *kobj,
2560 struct attribute *attr,
2561 const char *buf, size_t len)
2563 struct ext4_sb_info *sbi = container_of(kobj, struct ext4_sb_info,
2565 struct ext4_attr *a = container_of(attr, struct ext4_attr, attr);
2567 return a->store ? a->store(a, sbi, buf, len) : 0;
2570 static void ext4_sb_release(struct kobject *kobj)
2572 struct ext4_sb_info *sbi = container_of(kobj, struct ext4_sb_info,
2574 complete(&sbi->s_kobj_unregister);
2577 static const struct sysfs_ops ext4_attr_ops = {
2578 .show = ext4_attr_show,
2579 .store = ext4_attr_store,
2582 static struct kobj_type ext4_ktype = {
2583 .default_attrs = ext4_attrs,
2584 .sysfs_ops = &ext4_attr_ops,
2585 .release = ext4_sb_release,
2588 static void ext4_feat_release(struct kobject *kobj)
2590 complete(&ext4_feat->f_kobj_unregister);
2593 static struct kobj_type ext4_feat_ktype = {
2594 .default_attrs = ext4_feat_attrs,
2595 .sysfs_ops = &ext4_attr_ops,
2596 .release = ext4_feat_release,
2600 * Check whether this filesystem can be mounted based on
2601 * the features present and the RDONLY/RDWR mount requested.
2602 * Returns 1 if this filesystem can be mounted as requested,
2603 * 0 if it cannot be.
2605 static int ext4_feature_set_ok(struct super_block *sb, int readonly)
2607 if (EXT4_HAS_INCOMPAT_FEATURE(sb, ~EXT4_FEATURE_INCOMPAT_SUPP)) {
2608 ext4_msg(sb, KERN_ERR,
2609 "Couldn't mount because of "
2610 "unsupported optional features (%x)",
2611 (le32_to_cpu(EXT4_SB(sb)->s_es->s_feature_incompat) &
2612 ~EXT4_FEATURE_INCOMPAT_SUPP));
2619 /* Check that feature set is OK for a read-write mount */
2620 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, ~EXT4_FEATURE_RO_COMPAT_SUPP)) {
2621 ext4_msg(sb, KERN_ERR, "couldn't mount RDWR because of "
2622 "unsupported optional features (%x)",
2623 (le32_to_cpu(EXT4_SB(sb)->s_es->s_feature_ro_compat) &
2624 ~EXT4_FEATURE_RO_COMPAT_SUPP));
2628 * Large file size enabled file system can only be mounted
2629 * read-write on 32-bit systems if kernel is built with CONFIG_LBDAF
2631 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_HUGE_FILE)) {
2632 if (sizeof(blkcnt_t) < sizeof(u64)) {
2633 ext4_msg(sb, KERN_ERR, "Filesystem with huge files "
2634 "cannot be mounted RDWR without "
2639 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_BIGALLOC) &&
2640 !EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_EXTENTS)) {
2641 ext4_msg(sb, KERN_ERR,
2642 "Can't support bigalloc feature without "
2643 "extents feature\n");
2647 #ifndef CONFIG_QUOTA
2648 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_QUOTA) &&
2650 ext4_msg(sb, KERN_ERR,
2651 "Filesystem with quota feature cannot be mounted RDWR "
2652 "without CONFIG_QUOTA");
2655 #endif /* CONFIG_QUOTA */
2660 * This function is called once a day if we have errors logged
2661 * on the file system
2663 static void print_daily_error_info(unsigned long arg)
2665 struct super_block *sb = (struct super_block *) arg;
2666 struct ext4_sb_info *sbi;
2667 struct ext4_super_block *es;
2672 if (es->s_error_count)
2673 ext4_msg(sb, KERN_NOTICE, "error count: %u",
2674 le32_to_cpu(es->s_error_count));
2675 if (es->s_first_error_time) {
2676 printk(KERN_NOTICE "EXT4-fs (%s): initial error at %u: %.*s:%d",
2677 sb->s_id, le32_to_cpu(es->s_first_error_time),
2678 (int) sizeof(es->s_first_error_func),
2679 es->s_first_error_func,
2680 le32_to_cpu(es->s_first_error_line));
2681 if (es->s_first_error_ino)
2682 printk(": inode %u",
2683 le32_to_cpu(es->s_first_error_ino));
2684 if (es->s_first_error_block)
2685 printk(": block %llu", (unsigned long long)
2686 le64_to_cpu(es->s_first_error_block));
2689 if (es->s_last_error_time) {
2690 printk(KERN_NOTICE "EXT4-fs (%s): last error at %u: %.*s:%d",
2691 sb->s_id, le32_to_cpu(es->s_last_error_time),
2692 (int) sizeof(es->s_last_error_func),
2693 es->s_last_error_func,
2694 le32_to_cpu(es->s_last_error_line));
2695 if (es->s_last_error_ino)
2696 printk(": inode %u",
2697 le32_to_cpu(es->s_last_error_ino));
2698 if (es->s_last_error_block)
2699 printk(": block %llu", (unsigned long long)
2700 le64_to_cpu(es->s_last_error_block));
2703 mod_timer(&sbi->s_err_report, jiffies + 24*60*60*HZ); /* Once a day */
2706 /* Find next suitable group and run ext4_init_inode_table */
2707 static int ext4_run_li_request(struct ext4_li_request *elr)
2709 struct ext4_group_desc *gdp = NULL;
2710 ext4_group_t group, ngroups;
2711 struct super_block *sb;
2712 unsigned long timeout = 0;
2716 ngroups = EXT4_SB(sb)->s_groups_count;
2719 for (group = elr->lr_next_group; group < ngroups; group++) {
2720 gdp = ext4_get_group_desc(sb, group, NULL);
2726 if (!(gdp->bg_flags & cpu_to_le16(EXT4_BG_INODE_ZEROED)))
2730 if (group >= ngroups)
2735 ret = ext4_init_inode_table(sb, group,
2736 elr->lr_timeout ? 0 : 1);
2737 if (elr->lr_timeout == 0) {
2738 timeout = (jiffies - timeout) *
2739 elr->lr_sbi->s_li_wait_mult;
2740 elr->lr_timeout = timeout;
2742 elr->lr_next_sched = jiffies + elr->lr_timeout;
2743 elr->lr_next_group = group + 1;
2751 * Remove lr_request from the list_request and free the
2752 * request structure. Should be called with li_list_mtx held
2754 static void ext4_remove_li_request(struct ext4_li_request *elr)
2756 struct ext4_sb_info *sbi;
2763 list_del(&elr->lr_request);
2764 sbi->s_li_request = NULL;
2768 static void ext4_unregister_li_request(struct super_block *sb)
2770 mutex_lock(&ext4_li_mtx);
2771 if (!ext4_li_info) {
2772 mutex_unlock(&ext4_li_mtx);
2776 mutex_lock(&ext4_li_info->li_list_mtx);
2777 ext4_remove_li_request(EXT4_SB(sb)->s_li_request);
2778 mutex_unlock(&ext4_li_info->li_list_mtx);
2779 mutex_unlock(&ext4_li_mtx);
2782 static struct task_struct *ext4_lazyinit_task;
2785 * This is the function where ext4lazyinit thread lives. It walks
2786 * through the request list searching for next scheduled filesystem.
2787 * When such a fs is found, run the lazy initialization request
2788 * (ext4_rn_li_request) and keep track of the time spend in this
2789 * function. Based on that time we compute next schedule time of
2790 * the request. When walking through the list is complete, compute
2791 * next waking time and put itself into sleep.
2793 static int ext4_lazyinit_thread(void *arg)
2795 struct ext4_lazy_init *eli = (struct ext4_lazy_init *)arg;
2796 struct list_head *pos, *n;
2797 struct ext4_li_request *elr;
2798 unsigned long next_wakeup, cur;
2800 BUG_ON(NULL == eli);
2804 next_wakeup = MAX_JIFFY_OFFSET;
2806 mutex_lock(&eli->li_list_mtx);
2807 if (list_empty(&eli->li_request_list)) {
2808 mutex_unlock(&eli->li_list_mtx);
2812 list_for_each_safe(pos, n, &eli->li_request_list) {
2813 elr = list_entry(pos, struct ext4_li_request,
2816 if (time_after_eq(jiffies, elr->lr_next_sched)) {
2817 if (ext4_run_li_request(elr) != 0) {
2818 /* error, remove the lazy_init job */
2819 ext4_remove_li_request(elr);
2824 if (time_before(elr->lr_next_sched, next_wakeup))
2825 next_wakeup = elr->lr_next_sched;
2827 mutex_unlock(&eli->li_list_mtx);
2832 if ((time_after_eq(cur, next_wakeup)) ||
2833 (MAX_JIFFY_OFFSET == next_wakeup)) {
2838 schedule_timeout_interruptible(next_wakeup - cur);
2840 if (kthread_should_stop()) {
2841 ext4_clear_request_list();
2848 * It looks like the request list is empty, but we need
2849 * to check it under the li_list_mtx lock, to prevent any
2850 * additions into it, and of course we should lock ext4_li_mtx
2851 * to atomically free the list and ext4_li_info, because at
2852 * this point another ext4 filesystem could be registering
2855 mutex_lock(&ext4_li_mtx);
2856 mutex_lock(&eli->li_list_mtx);
2857 if (!list_empty(&eli->li_request_list)) {
2858 mutex_unlock(&eli->li_list_mtx);
2859 mutex_unlock(&ext4_li_mtx);
2862 mutex_unlock(&eli->li_list_mtx);
2863 kfree(ext4_li_info);
2864 ext4_li_info = NULL;
2865 mutex_unlock(&ext4_li_mtx);
2870 static void ext4_clear_request_list(void)
2872 struct list_head *pos, *n;
2873 struct ext4_li_request *elr;
2875 mutex_lock(&ext4_li_info->li_list_mtx);
2876 list_for_each_safe(pos, n, &ext4_li_info->li_request_list) {
2877 elr = list_entry(pos, struct ext4_li_request,
2879 ext4_remove_li_request(elr);
2881 mutex_unlock(&ext4_li_info->li_list_mtx);
2884 static int ext4_run_lazyinit_thread(void)
2886 ext4_lazyinit_task = kthread_run(ext4_lazyinit_thread,
2887 ext4_li_info, "ext4lazyinit");
2888 if (IS_ERR(ext4_lazyinit_task)) {
2889 int err = PTR_ERR(ext4_lazyinit_task);
2890 ext4_clear_request_list();
2891 kfree(ext4_li_info);
2892 ext4_li_info = NULL;
2893 printk(KERN_CRIT "EXT4-fs: error %d creating inode table "
2894 "initialization thread\n",
2898 ext4_li_info->li_state |= EXT4_LAZYINIT_RUNNING;
2903 * Check whether it make sense to run itable init. thread or not.
2904 * If there is at least one uninitialized inode table, return
2905 * corresponding group number, else the loop goes through all
2906 * groups and return total number of groups.
2908 static ext4_group_t ext4_has_uninit_itable(struct super_block *sb)
2910 ext4_group_t group, ngroups = EXT4_SB(sb)->s_groups_count;
2911 struct ext4_group_desc *gdp = NULL;
2913 for (group = 0; group < ngroups; group++) {
2914 gdp = ext4_get_group_desc(sb, group, NULL);
2918 if (!(gdp->bg_flags & cpu_to_le16(EXT4_BG_INODE_ZEROED)))
2925 static int ext4_li_info_new(void)
2927 struct ext4_lazy_init *eli = NULL;
2929 eli = kzalloc(sizeof(*eli), GFP_KERNEL);
2933 INIT_LIST_HEAD(&eli->li_request_list);
2934 mutex_init(&eli->li_list_mtx);
2936 eli->li_state |= EXT4_LAZYINIT_QUIT;
2943 static struct ext4_li_request *ext4_li_request_new(struct super_block *sb,
2946 struct ext4_sb_info *sbi = EXT4_SB(sb);
2947 struct ext4_li_request *elr;
2950 elr = kzalloc(sizeof(*elr), GFP_KERNEL);
2956 elr->lr_next_group = start;
2959 * Randomize first schedule time of the request to
2960 * spread the inode table initialization requests
2963 get_random_bytes(&rnd, sizeof(rnd));
2964 elr->lr_next_sched = jiffies + (unsigned long)rnd %
2965 (EXT4_DEF_LI_MAX_START_DELAY * HZ);
2970 int ext4_register_li_request(struct super_block *sb,
2971 ext4_group_t first_not_zeroed)
2973 struct ext4_sb_info *sbi = EXT4_SB(sb);
2974 struct ext4_li_request *elr = NULL;
2975 ext4_group_t ngroups = EXT4_SB(sb)->s_groups_count;
2978 mutex_lock(&ext4_li_mtx);
2979 if (sbi->s_li_request != NULL) {
2981 * Reset timeout so it can be computed again, because
2982 * s_li_wait_mult might have changed.
2984 sbi->s_li_request->lr_timeout = 0;
2988 if (first_not_zeroed == ngroups ||
2989 (sb->s_flags & MS_RDONLY) ||
2990 !test_opt(sb, INIT_INODE_TABLE))
2993 elr = ext4_li_request_new(sb, first_not_zeroed);
2999 if (NULL == ext4_li_info) {
3000 ret = ext4_li_info_new();
3005 mutex_lock(&ext4_li_info->li_list_mtx);
3006 list_add(&elr->lr_request, &ext4_li_info->li_request_list);
3007 mutex_unlock(&ext4_li_info->li_list_mtx);
3009 sbi->s_li_request = elr;
3011 * set elr to NULL here since it has been inserted to
3012 * the request_list and the removal and free of it is
3013 * handled by ext4_clear_request_list from now on.
3017 if (!(ext4_li_info->li_state & EXT4_LAZYINIT_RUNNING)) {
3018 ret = ext4_run_lazyinit_thread();
3023 mutex_unlock(&ext4_li_mtx);
3030 * We do not need to lock anything since this is called on
3033 static void ext4_destroy_lazyinit_thread(void)
3036 * If thread exited earlier
3037 * there's nothing to be done.
3039 if (!ext4_li_info || !ext4_lazyinit_task)
3042 kthread_stop(ext4_lazyinit_task);
3045 static int set_journal_csum_feature_set(struct super_block *sb)
3048 int compat, incompat;
3049 struct ext4_sb_info *sbi = EXT4_SB(sb);
3051 if (EXT4_HAS_RO_COMPAT_FEATURE(sb,
3052 EXT4_FEATURE_RO_COMPAT_METADATA_CSUM)) {
3053 /* journal checksum v2 */
3055 incompat = JBD2_FEATURE_INCOMPAT_CSUM_V2;
3057 /* journal checksum v1 */
3058 compat = JBD2_FEATURE_COMPAT_CHECKSUM;
3062 if (test_opt(sb, JOURNAL_ASYNC_COMMIT)) {
3063 ret = jbd2_journal_set_features(sbi->s_journal,
3065 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT |
3067 } else if (test_opt(sb, JOURNAL_CHECKSUM)) {
3068 ret = jbd2_journal_set_features(sbi->s_journal,
3071 jbd2_journal_clear_features(sbi->s_journal, 0, 0,
3072 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT);
3074 jbd2_journal_clear_features(sbi->s_journal,
3075 JBD2_FEATURE_COMPAT_CHECKSUM, 0,
3076 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT |
3077 JBD2_FEATURE_INCOMPAT_CSUM_V2);
3084 * Note: calculating the overhead so we can be compatible with
3085 * historical BSD practice is quite difficult in the face of
3086 * clusters/bigalloc. This is because multiple metadata blocks from
3087 * different block group can end up in the same allocation cluster.
3088 * Calculating the exact overhead in the face of clustered allocation
3089 * requires either O(all block bitmaps) in memory or O(number of block
3090 * groups**2) in time. We will still calculate the superblock for
3091 * older file systems --- and if we come across with a bigalloc file
3092 * system with zero in s_overhead_clusters the estimate will be close to
3093 * correct especially for very large cluster sizes --- but for newer
3094 * file systems, it's better to calculate this figure once at mkfs
3095 * time, and store it in the superblock. If the superblock value is
3096 * present (even for non-bigalloc file systems), we will use it.
3098 static int count_overhead(struct super_block *sb, ext4_group_t grp,
3101 struct ext4_sb_info *sbi = EXT4_SB(sb);
3102 struct ext4_group_desc *gdp;
3103 ext4_fsblk_t first_block, last_block, b;
3104 ext4_group_t i, ngroups = ext4_get_groups_count(sb);
3105 int s, j, count = 0;
3107 if (!EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_BIGALLOC))
3108 return (ext4_bg_has_super(sb, grp) + ext4_bg_num_gdb(sb, grp) +
3109 sbi->s_itb_per_group + 2);
3111 first_block = le32_to_cpu(sbi->s_es->s_first_data_block) +
3112 (grp * EXT4_BLOCKS_PER_GROUP(sb));
3113 last_block = first_block + EXT4_BLOCKS_PER_GROUP(sb) - 1;
3114 for (i = 0; i < ngroups; i++) {
3115 gdp = ext4_get_group_desc(sb, i, NULL);
3116 b = ext4_block_bitmap(sb, gdp);
3117 if (b >= first_block && b <= last_block) {
3118 ext4_set_bit(EXT4_B2C(sbi, b - first_block), buf);
3121 b = ext4_inode_bitmap(sb, gdp);
3122 if (b >= first_block && b <= last_block) {
3123 ext4_set_bit(EXT4_B2C(sbi, b - first_block), buf);
3126 b = ext4_inode_table(sb, gdp);
3127 if (b >= first_block && b + sbi->s_itb_per_group <= last_block)
3128 for (j = 0; j < sbi->s_itb_per_group; j++, b++) {
3129 int c = EXT4_B2C(sbi, b - first_block);
3130 ext4_set_bit(c, buf);
3136 if (ext4_bg_has_super(sb, grp)) {
3137 ext4_set_bit(s++, buf);
3140 for (j = ext4_bg_num_gdb(sb, grp); j > 0; j--) {
3141 ext4_set_bit(EXT4_B2C(sbi, s++), buf);
3147 return EXT4_CLUSTERS_PER_GROUP(sb) -
3148 ext4_count_free(buf, EXT4_CLUSTERS_PER_GROUP(sb) / 8);
3152 * Compute the overhead and stash it in sbi->s_overhead
3154 int ext4_calculate_overhead(struct super_block *sb)
3156 struct ext4_sb_info *sbi = EXT4_SB(sb);
3157 struct ext4_super_block *es = sbi->s_es;
3158 ext4_group_t i, ngroups = ext4_get_groups_count(sb);
3159 ext4_fsblk_t overhead = 0;
3160 char *buf = (char *) get_zeroed_page(GFP_KERNEL);
3166 * Compute the overhead (FS structures). This is constant
3167 * for a given filesystem unless the number of block groups
3168 * changes so we cache the previous value until it does.
3172 * All of the blocks before first_data_block are overhead
3174 overhead = EXT4_B2C(sbi, le32_to_cpu(es->s_first_data_block));
3177 * Add the overhead found in each block group
3179 for (i = 0; i < ngroups; i++) {
3182 blks = count_overhead(sb, i, buf);
3185 memset(buf, 0, PAGE_SIZE);
3188 /* Add the journal blocks as well */
3190 overhead += EXT4_NUM_B2C(sbi, sbi->s_journal->j_maxlen);
3192 sbi->s_overhead = overhead;
3194 free_page((unsigned long) buf);
3198 static int ext4_fill_super(struct super_block *sb, void *data, int silent)
3200 char *orig_data = kstrdup(data, GFP_KERNEL);
3201 struct buffer_head *bh;
3202 struct ext4_super_block *es = NULL;
3203 struct ext4_sb_info *sbi;
3205 ext4_fsblk_t sb_block = get_sb_block(&data);
3206 ext4_fsblk_t logical_sb_block;
3207 unsigned long offset = 0;
3208 unsigned long journal_devnum = 0;
3209 unsigned long def_mount_opts;
3214 int blocksize, clustersize;
3215 unsigned int db_count;
3217 int needs_recovery, has_huge_files, has_bigalloc;
3220 unsigned int journal_ioprio = DEFAULT_JOURNAL_IOPRIO;
3221 ext4_group_t first_not_zeroed;
3223 sbi = kzalloc(sizeof(*sbi), GFP_KERNEL);
3227 sbi->s_blockgroup_lock =
3228 kzalloc(sizeof(struct blockgroup_lock), GFP_KERNEL);
3229 if (!sbi->s_blockgroup_lock) {
3233 sb->s_fs_info = sbi;
3235 sbi->s_inode_readahead_blks = EXT4_DEF_INODE_READAHEAD_BLKS;
3236 sbi->s_sb_block = sb_block;
3237 if (sb->s_bdev->bd_part)
3238 sbi->s_sectors_written_start =
3239 part_stat_read(sb->s_bdev->bd_part, sectors[1]);
3241 /* Cleanup superblock name */
3242 for (cp = sb->s_id; (cp = strchr(cp, '/'));)
3245 /* -EINVAL is default */
3247 blocksize = sb_min_blocksize(sb, EXT4_MIN_BLOCK_SIZE);
3249 ext4_msg(sb, KERN_ERR, "unable to set blocksize");
3254 * The ext4 superblock will not be buffer aligned for other than 1kB
3255 * block sizes. We need to calculate the offset from buffer start.
3257 if (blocksize != EXT4_MIN_BLOCK_SIZE) {
3258 logical_sb_block = sb_block * EXT4_MIN_BLOCK_SIZE;
3259 offset = do_div(logical_sb_block, blocksize);
3261 logical_sb_block = sb_block;
3264 if (!(bh = sb_bread(sb, logical_sb_block))) {
3265 ext4_msg(sb, KERN_ERR, "unable to read superblock");
3269 * Note: s_es must be initialized as soon as possible because
3270 * some ext4 macro-instructions depend on its value
3272 es = (struct ext4_super_block *) (bh->b_data + offset);
3274 sb->s_magic = le16_to_cpu(es->s_magic);
3275 if (sb->s_magic != EXT4_SUPER_MAGIC)
3277 sbi->s_kbytes_written = le64_to_cpu(es->s_kbytes_written);
3279 /* Warn if metadata_csum and gdt_csum are both set. */
3280 if (EXT4_HAS_RO_COMPAT_FEATURE(sb,
3281 EXT4_FEATURE_RO_COMPAT_METADATA_CSUM) &&
3282 EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_GDT_CSUM))
3283 ext4_warning(sb, KERN_INFO "metadata_csum and uninit_bg are "
3284 "redundant flags; please run fsck.");
3286 /* Check for a known checksum algorithm */
3287 if (!ext4_verify_csum_type(sb, es)) {
3288 ext4_msg(sb, KERN_ERR, "VFS: Found ext4 filesystem with "
3289 "unknown checksum algorithm.");
3294 /* Load the checksum driver */
3295 if (EXT4_HAS_RO_COMPAT_FEATURE(sb,
3296 EXT4_FEATURE_RO_COMPAT_METADATA_CSUM)) {
3297 sbi->s_chksum_driver = crypto_alloc_shash("crc32c", 0, 0);
3298 if (IS_ERR(sbi->s_chksum_driver)) {
3299 ext4_msg(sb, KERN_ERR, "Cannot load crc32c driver.");
3300 ret = PTR_ERR(sbi->s_chksum_driver);
3301 sbi->s_chksum_driver = NULL;
3306 /* Check superblock checksum */
3307 if (!ext4_superblock_csum_verify(sb, es)) {
3308 ext4_msg(sb, KERN_ERR, "VFS: Found ext4 filesystem with "
3309 "invalid superblock checksum. Run e2fsck?");
3314 /* Precompute checksum seed for all metadata */
3315 if (EXT4_HAS_RO_COMPAT_FEATURE(sb,
3316 EXT4_FEATURE_RO_COMPAT_METADATA_CSUM))
3317 sbi->s_csum_seed = ext4_chksum(sbi, ~0, es->s_uuid,
3318 sizeof(es->s_uuid));
3320 /* Set defaults before we parse the mount options */
3321 def_mount_opts = le32_to_cpu(es->s_default_mount_opts);
3322 set_opt(sb, INIT_INODE_TABLE);
3323 if (def_mount_opts & EXT4_DEFM_DEBUG)
3325 if (def_mount_opts & EXT4_DEFM_BSDGROUPS)
3327 if (def_mount_opts & EXT4_DEFM_UID16)
3328 set_opt(sb, NO_UID32);
3329 /* xattr user namespace & acls are now defaulted on */
3330 set_opt(sb, XATTR_USER);
3331 #ifdef CONFIG_EXT4_FS_POSIX_ACL
3332 set_opt(sb, POSIX_ACL);
3334 if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_DATA)
3335 set_opt(sb, JOURNAL_DATA);
3336 else if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_ORDERED)
3337 set_opt(sb, ORDERED_DATA);
3338 else if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_WBACK)
3339 set_opt(sb, WRITEBACK_DATA);
3341 if (le16_to_cpu(sbi->s_es->s_errors) == EXT4_ERRORS_PANIC)
3342 set_opt(sb, ERRORS_PANIC);
3343 else if (le16_to_cpu(sbi->s_es->s_errors) == EXT4_ERRORS_CONTINUE)
3344 set_opt(sb, ERRORS_CONT);
3346 set_opt(sb, ERRORS_RO);
3347 if (def_mount_opts & EXT4_DEFM_BLOCK_VALIDITY)
3348 set_opt(sb, BLOCK_VALIDITY);
3349 if (def_mount_opts & EXT4_DEFM_DISCARD)
3350 set_opt(sb, DISCARD);
3352 sbi->s_resuid = make_kuid(&init_user_ns, le16_to_cpu(es->s_def_resuid));
3353 sbi->s_resgid = make_kgid(&init_user_ns, le16_to_cpu(es->s_def_resgid));
3354 sbi->s_commit_interval = JBD2_DEFAULT_MAX_COMMIT_AGE * HZ;
3355 sbi->s_min_batch_time = EXT4_DEF_MIN_BATCH_TIME;
3356 sbi->s_max_batch_time = EXT4_DEF_MAX_BATCH_TIME;
3358 if ((def_mount_opts & EXT4_DEFM_NOBARRIER) == 0)
3359 set_opt(sb, BARRIER);
3362 * enable delayed allocation by default
3363 * Use -o nodelalloc to turn it off
3365 if (!IS_EXT3_SB(sb) && !IS_EXT2_SB(sb) &&
3366 ((def_mount_opts & EXT4_DEFM_NODELALLOC) == 0))
3367 set_opt(sb, DELALLOC);
3370 * set default s_li_wait_mult for lazyinit, for the case there is
3371 * no mount option specified.
3373 sbi->s_li_wait_mult = EXT4_DEF_LI_WAIT_MULT;
3375 if (!parse_options((char *) sbi->s_es->s_mount_opts, sb,
3376 &journal_devnum, &journal_ioprio, 0)) {
3377 ext4_msg(sb, KERN_WARNING,
3378 "failed to parse options in superblock: %s",
3379 sbi->s_es->s_mount_opts);
3381 sbi->s_def_mount_opt = sbi->s_mount_opt;
3382 if (!parse_options((char *) data, sb, &journal_devnum,
3383 &journal_ioprio, 0))
3386 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA) {
3387 printk_once(KERN_WARNING "EXT4-fs: Warning: mounting "
3388 "with data=journal disables delayed "
3389 "allocation and O_DIRECT support!\n");
3390 if (test_opt2(sb, EXPLICIT_DELALLOC)) {
3391 ext4_msg(sb, KERN_ERR, "can't mount with "
3392 "both data=journal and delalloc");
3395 if (test_opt(sb, DIOREAD_NOLOCK)) {
3396 ext4_msg(sb, KERN_ERR, "can't mount with "
3397 "both data=journal and delalloc");
3400 if (test_opt(sb, DELALLOC))
3401 clear_opt(sb, DELALLOC);
3404 sb->s_flags = (sb->s_flags & ~MS_POSIXACL) |
3405 (test_opt(sb, POSIX_ACL) ? MS_POSIXACL : 0);
3407 if (le32_to_cpu(es->s_rev_level) == EXT4_GOOD_OLD_REV &&
3408 (EXT4_HAS_COMPAT_FEATURE(sb, ~0U) ||
3409 EXT4_HAS_RO_COMPAT_FEATURE(sb, ~0U) ||
3410 EXT4_HAS_INCOMPAT_FEATURE(sb, ~0U)))
3411 ext4_msg(sb, KERN_WARNING,
3412 "feature flags set on rev 0 fs, "
3413 "running e2fsck is recommended");
3415 if (IS_EXT2_SB(sb)) {
3416 if (ext2_feature_set_ok(sb))
3417 ext4_msg(sb, KERN_INFO, "mounting ext2 file system "
3418 "using the ext4 subsystem");
3420 ext4_msg(sb, KERN_ERR, "couldn't mount as ext2 due "
3421 "to feature incompatibilities");
3426 if (IS_EXT3_SB(sb)) {
3427 if (ext3_feature_set_ok(sb))
3428 ext4_msg(sb, KERN_INFO, "mounting ext3 file system "
3429 "using the ext4 subsystem");
3431 ext4_msg(sb, KERN_ERR, "couldn't mount as ext3 due "
3432 "to feature incompatibilities");
3438 * Check feature flags regardless of the revision level, since we
3439 * previously didn't change the revision level when setting the flags,
3440 * so there is a chance incompat flags are set on a rev 0 filesystem.
3442 if (!ext4_feature_set_ok(sb, (sb->s_flags & MS_RDONLY)))
3445 blocksize = BLOCK_SIZE << le32_to_cpu(es->s_log_block_size);
3446 if (blocksize < EXT4_MIN_BLOCK_SIZE ||
3447 blocksize > EXT4_MAX_BLOCK_SIZE) {
3448 ext4_msg(sb, KERN_ERR,
3449 "Unsupported filesystem blocksize %d", blocksize);
3453 if (sb->s_blocksize != blocksize) {
3454 /* Validate the filesystem blocksize */
3455 if (!sb_set_blocksize(sb, blocksize)) {
3456 ext4_msg(sb, KERN_ERR, "bad block size %d",
3462 logical_sb_block = sb_block * EXT4_MIN_BLOCK_SIZE;
3463 offset = do_div(logical_sb_block, blocksize);
3464 bh = sb_bread(sb, logical_sb_block);
3466 ext4_msg(sb, KERN_ERR,
3467 "Can't read superblock on 2nd try");
3470 es = (struct ext4_super_block *)(bh->b_data + offset);
3472 if (es->s_magic != cpu_to_le16(EXT4_SUPER_MAGIC)) {
3473 ext4_msg(sb, KERN_ERR,
3474 "Magic mismatch, very weird!");
3479 has_huge_files = EXT4_HAS_RO_COMPAT_FEATURE(sb,
3480 EXT4_FEATURE_RO_COMPAT_HUGE_FILE);
3481 sbi->s_bitmap_maxbytes = ext4_max_bitmap_size(sb->s_blocksize_bits,
3483 sb->s_maxbytes = ext4_max_size(sb->s_blocksize_bits, has_huge_files);
3485 if (le32_to_cpu(es->s_rev_level) == EXT4_GOOD_OLD_REV) {
3486 sbi->s_inode_size = EXT4_GOOD_OLD_INODE_SIZE;
3487 sbi->s_first_ino = EXT4_GOOD_OLD_FIRST_INO;
3489 sbi->s_inode_size = le16_to_cpu(es->s_inode_size);
3490 sbi->s_first_ino = le32_to_cpu(es->s_first_ino);
3491 if ((sbi->s_inode_size < EXT4_GOOD_OLD_INODE_SIZE) ||
3492 (!is_power_of_2(sbi->s_inode_size)) ||
3493 (sbi->s_inode_size > blocksize)) {
3494 ext4_msg(sb, KERN_ERR,
3495 "unsupported inode size: %d",
3499 if (sbi->s_inode_size > EXT4_GOOD_OLD_INODE_SIZE)
3500 sb->s_time_gran = 1 << (EXT4_EPOCH_BITS - 2);
3503 sbi->s_desc_size = le16_to_cpu(es->s_desc_size);
3504 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_64BIT)) {
3505 if (sbi->s_desc_size < EXT4_MIN_DESC_SIZE_64BIT ||
3506 sbi->s_desc_size > EXT4_MAX_DESC_SIZE ||
3507 !is_power_of_2(sbi->s_desc_size)) {
3508 ext4_msg(sb, KERN_ERR,
3509 "unsupported descriptor size %lu",
3514 sbi->s_desc_size = EXT4_MIN_DESC_SIZE;
3516 sbi->s_blocks_per_group = le32_to_cpu(es->s_blocks_per_group);
3517 sbi->s_inodes_per_group = le32_to_cpu(es->s_inodes_per_group);
3518 if (EXT4_INODE_SIZE(sb) == 0 || EXT4_INODES_PER_GROUP(sb) == 0)
3521 sbi->s_inodes_per_block = blocksize / EXT4_INODE_SIZE(sb);
3522 if (sbi->s_inodes_per_block == 0)
3524 sbi->s_itb_per_group = sbi->s_inodes_per_group /
3525 sbi->s_inodes_per_block;
3526 sbi->s_desc_per_block = blocksize / EXT4_DESC_SIZE(sb);
3528 sbi->s_mount_state = le16_to_cpu(es->s_state);
3529 sbi->s_addr_per_block_bits = ilog2(EXT4_ADDR_PER_BLOCK(sb));
3530 sbi->s_desc_per_block_bits = ilog2(EXT4_DESC_PER_BLOCK(sb));
3532 /* Do we have standard group size of blocksize * 8 blocks ? */
3533 if (sbi->s_blocks_per_group == blocksize << 3)
3534 set_opt2(sb, STD_GROUP_SIZE);
3536 for (i = 0; i < 4; i++)
3537 sbi->s_hash_seed[i] = le32_to_cpu(es->s_hash_seed[i]);
3538 sbi->s_def_hash_version = es->s_def_hash_version;
3539 i = le32_to_cpu(es->s_flags);
3540 if (i & EXT2_FLAGS_UNSIGNED_HASH)
3541 sbi->s_hash_unsigned = 3;
3542 else if ((i & EXT2_FLAGS_SIGNED_HASH) == 0) {
3543 #ifdef __CHAR_UNSIGNED__
3544 es->s_flags |= cpu_to_le32(EXT2_FLAGS_UNSIGNED_HASH);
3545 sbi->s_hash_unsigned = 3;
3547 es->s_flags |= cpu_to_le32(EXT2_FLAGS_SIGNED_HASH);
3551 /* Handle clustersize */
3552 clustersize = BLOCK_SIZE << le32_to_cpu(es->s_log_cluster_size);
3553 has_bigalloc = EXT4_HAS_RO_COMPAT_FEATURE(sb,
3554 EXT4_FEATURE_RO_COMPAT_BIGALLOC);
3556 if (clustersize < blocksize) {
3557 ext4_msg(sb, KERN_ERR,
3558 "cluster size (%d) smaller than "
3559 "block size (%d)", clustersize, blocksize);
3562 sbi->s_cluster_bits = le32_to_cpu(es->s_log_cluster_size) -
3563 le32_to_cpu(es->s_log_block_size);
3564 sbi->s_clusters_per_group =
3565 le32_to_cpu(es->s_clusters_per_group);
3566 if (sbi->s_clusters_per_group > blocksize * 8) {
3567 ext4_msg(sb, KERN_ERR,
3568 "#clusters per group too big: %lu",
3569 sbi->s_clusters_per_group);
3572 if (sbi->s_blocks_per_group !=
3573 (sbi->s_clusters_per_group * (clustersize / blocksize))) {
3574 ext4_msg(sb, KERN_ERR, "blocks per group (%lu) and "
3575 "clusters per group (%lu) inconsistent",
3576 sbi->s_blocks_per_group,
3577 sbi->s_clusters_per_group);
3581 if (clustersize != blocksize) {
3582 ext4_warning(sb, "fragment/cluster size (%d) != "
3583 "block size (%d)", clustersize,
3585 clustersize = blocksize;
3587 if (sbi->s_blocks_per_group > blocksize * 8) {
3588 ext4_msg(sb, KERN_ERR,
3589 "#blocks per group too big: %lu",
3590 sbi->s_blocks_per_group);
3593 sbi->s_clusters_per_group = sbi->s_blocks_per_group;
3594 sbi->s_cluster_bits = 0;
3596 sbi->s_cluster_ratio = clustersize / blocksize;
3598 if (sbi->s_inodes_per_group > blocksize * 8) {
3599 ext4_msg(sb, KERN_ERR,
3600 "#inodes per group too big: %lu",
3601 sbi->s_inodes_per_group);
3606 * Test whether we have more sectors than will fit in sector_t,
3607 * and whether the max offset is addressable by the page cache.
3609 err = generic_check_addressable(sb->s_blocksize_bits,
3610 ext4_blocks_count(es));
3612 ext4_msg(sb, KERN_ERR, "filesystem"
3613 " too large to mount safely on this system");
3614 if (sizeof(sector_t) < 8)
3615 ext4_msg(sb, KERN_WARNING, "CONFIG_LBDAF not enabled");
3619 if (EXT4_BLOCKS_PER_GROUP(sb) == 0)
3622 /* check blocks count against device size */
3623 blocks_count = sb->s_bdev->bd_inode->i_size >> sb->s_blocksize_bits;
3624 if (blocks_count && ext4_blocks_count(es) > blocks_count) {
3625 ext4_msg(sb, KERN_WARNING, "bad geometry: block count %llu "
3626 "exceeds size of device (%llu blocks)",
3627 ext4_blocks_count(es), blocks_count);
3632 * It makes no sense for the first data block to be beyond the end
3633 * of the filesystem.
3635 if (le32_to_cpu(es->s_first_data_block) >= ext4_blocks_count(es)) {
3636 ext4_msg(sb, KERN_WARNING, "bad geometry: first data "
3637 "block %u is beyond end of filesystem (%llu)",
3638 le32_to_cpu(es->s_first_data_block),
3639 ext4_blocks_count(es));
3642 blocks_count = (ext4_blocks_count(es) -
3643 le32_to_cpu(es->s_first_data_block) +
3644 EXT4_BLOCKS_PER_GROUP(sb) - 1);
3645 do_div(blocks_count, EXT4_BLOCKS_PER_GROUP(sb));
3646 if (blocks_count > ((uint64_t)1<<32) - EXT4_DESC_PER_BLOCK(sb)) {
3647 ext4_msg(sb, KERN_WARNING, "groups count too large: %u "
3648 "(block count %llu, first data block %u, "
3649 "blocks per group %lu)", sbi->s_groups_count,
3650 ext4_blocks_count(es),
3651 le32_to_cpu(es->s_first_data_block),
3652 EXT4_BLOCKS_PER_GROUP(sb));
3655 sbi->s_groups_count = blocks_count;
3656 sbi->s_blockfile_groups = min_t(ext4_group_t, sbi->s_groups_count,
3657 (EXT4_MAX_BLOCK_FILE_PHYS / EXT4_BLOCKS_PER_GROUP(sb)));
3658 db_count = (sbi->s_groups_count + EXT4_DESC_PER_BLOCK(sb) - 1) /
3659 EXT4_DESC_PER_BLOCK(sb);
3660 sbi->s_group_desc = ext4_kvmalloc(db_count *
3661 sizeof(struct buffer_head *),
3663 if (sbi->s_group_desc == NULL) {
3664 ext4_msg(sb, KERN_ERR, "not enough memory");
3670 sbi->s_proc = proc_mkdir(sb->s_id, ext4_proc_root);
3673 proc_create_data("options", S_IRUGO, sbi->s_proc,
3674 &ext4_seq_options_fops, sb);
3676 bgl_lock_init(sbi->s_blockgroup_lock);
3678 for (i = 0; i < db_count; i++) {
3679 block = descriptor_loc(sb, logical_sb_block, i);
3680 sbi->s_group_desc[i] = sb_bread(sb, block);
3681 if (!sbi->s_group_desc[i]) {
3682 ext4_msg(sb, KERN_ERR,
3683 "can't read group descriptor %d", i);
3688 if (!ext4_check_descriptors(sb, &first_not_zeroed)) {
3689 ext4_msg(sb, KERN_ERR, "group descriptors corrupted!");
3692 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_FLEX_BG))
3693 if (!ext4_fill_flex_info(sb)) {
3694 ext4_msg(sb, KERN_ERR,
3695 "unable to initialize "
3696 "flex_bg meta info!");
3700 sbi->s_gdb_count = db_count;
3701 get_random_bytes(&sbi->s_next_generation, sizeof(u32));
3702 spin_lock_init(&sbi->s_next_gen_lock);
3704 init_timer(&sbi->s_err_report);
3705 sbi->s_err_report.function = print_daily_error_info;
3706 sbi->s_err_report.data = (unsigned long) sb;
3708 /* Register extent status tree shrinker */
3709 ext4_es_register_shrinker(sb);
3711 err = percpu_counter_init(&sbi->s_freeclusters_counter,
3712 ext4_count_free_clusters(sb));
3714 err = percpu_counter_init(&sbi->s_freeinodes_counter,
3715 ext4_count_free_inodes(sb));
3718 err = percpu_counter_init(&sbi->s_dirs_counter,
3719 ext4_count_dirs(sb));
3722 err = percpu_counter_init(&sbi->s_dirtyclusters_counter, 0);
3725 err = percpu_counter_init(&sbi->s_extent_cache_cnt, 0);
3728 ext4_msg(sb, KERN_ERR, "insufficient memory");
3732 sbi->s_stripe = ext4_get_stripe_size(sbi);
3733 sbi->s_max_writeback_mb_bump = 128;
3734 sbi->s_extent_max_zeroout_kb = 32;
3737 * set up enough so that it can read an inode
3739 if (!test_opt(sb, NOLOAD) &&
3740 EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL))
3741 sb->s_op = &ext4_sops;
3743 sb->s_op = &ext4_nojournal_sops;
3744 sb->s_export_op = &ext4_export_ops;
3745 sb->s_xattr = ext4_xattr_handlers;
3747 sb->dq_op = &ext4_quota_operations;
3748 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_QUOTA))
3749 sb->s_qcop = &ext4_qctl_sysfile_operations;
3751 sb->s_qcop = &ext4_qctl_operations;
3753 memcpy(sb->s_uuid, es->s_uuid, sizeof(es->s_uuid));
3755 INIT_LIST_HEAD(&sbi->s_orphan); /* unlinked but open files */
3756 mutex_init(&sbi->s_orphan_lock);
3760 needs_recovery = (es->s_last_orphan != 0 ||
3761 EXT4_HAS_INCOMPAT_FEATURE(sb,
3762 EXT4_FEATURE_INCOMPAT_RECOVER));
3764 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_MMP) &&
3765 !(sb->s_flags & MS_RDONLY))
3766 if (ext4_multi_mount_protect(sb, le64_to_cpu(es->s_mmp_block)))
3770 * The first inode we look at is the journal inode. Don't try
3771 * root first: it may be modified in the journal!
3773 if (!test_opt(sb, NOLOAD) &&
3774 EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL)) {
3775 if (ext4_load_journal(sb, es, journal_devnum))
3777 } else if (test_opt(sb, NOLOAD) && !(sb->s_flags & MS_RDONLY) &&
3778 EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER)) {
3779 ext4_msg(sb, KERN_ERR, "required journal recovery "
3780 "suppressed and not mounted read-only");
3781 goto failed_mount_wq;
3783 clear_opt(sb, DATA_FLAGS);
3784 sbi->s_journal = NULL;
3789 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_64BIT) &&
3790 !jbd2_journal_set_features(EXT4_SB(sb)->s_journal, 0, 0,
3791 JBD2_FEATURE_INCOMPAT_64BIT)) {
3792 ext4_msg(sb, KERN_ERR, "Failed to set 64-bit journal feature");
3793 goto failed_mount_wq;
3796 if (!set_journal_csum_feature_set(sb)) {
3797 ext4_msg(sb, KERN_ERR, "Failed to set journal checksum "
3799 goto failed_mount_wq;
3802 /* We have now updated the journal if required, so we can
3803 * validate the data journaling mode. */
3804 switch (test_opt(sb, DATA_FLAGS)) {
3806 /* No mode set, assume a default based on the journal
3807 * capabilities: ORDERED_DATA if the journal can
3808 * cope, else JOURNAL_DATA
3810 if (jbd2_journal_check_available_features
3811 (sbi->s_journal, 0, 0, JBD2_FEATURE_INCOMPAT_REVOKE))
3812 set_opt(sb, ORDERED_DATA);
3814 set_opt(sb, JOURNAL_DATA);
3817 case EXT4_MOUNT_ORDERED_DATA:
3818 case EXT4_MOUNT_WRITEBACK_DATA:
3819 if (!jbd2_journal_check_available_features
3820 (sbi->s_journal, 0, 0, JBD2_FEATURE_INCOMPAT_REVOKE)) {
3821 ext4_msg(sb, KERN_ERR, "Journal does not support "
3822 "requested data journaling mode");
3823 goto failed_mount_wq;
3828 set_task_ioprio(sbi->s_journal->j_task, journal_ioprio);
3830 sbi->s_journal->j_commit_callback = ext4_journal_commit_callback;
3833 * The journal may have updated the bg summary counts, so we
3834 * need to update the global counters.
3836 percpu_counter_set(&sbi->s_freeclusters_counter,
3837 ext4_count_free_clusters(sb));
3838 percpu_counter_set(&sbi->s_freeinodes_counter,
3839 ext4_count_free_inodes(sb));
3840 percpu_counter_set(&sbi->s_dirs_counter,
3841 ext4_count_dirs(sb));
3842 percpu_counter_set(&sbi->s_dirtyclusters_counter, 0);
3846 * Get the # of file system overhead blocks from the
3847 * superblock if present.
3849 if (es->s_overhead_clusters)
3850 sbi->s_overhead = le32_to_cpu(es->s_overhead_clusters);
3852 err = ext4_calculate_overhead(sb);
3854 goto failed_mount_wq;
3858 * The maximum number of concurrent works can be high and
3859 * concurrency isn't really necessary. Limit it to 1.
3861 EXT4_SB(sb)->dio_unwritten_wq =
3862 alloc_workqueue("ext4-dio-unwritten", WQ_MEM_RECLAIM | WQ_UNBOUND, 1);
3863 if (!EXT4_SB(sb)->dio_unwritten_wq) {
3864 printk(KERN_ERR "EXT4-fs: failed to create DIO workqueue\n");
3866 goto failed_mount_wq;
3870 * The jbd2_journal_load will have done any necessary log recovery,
3871 * so we can safely mount the rest of the filesystem now.
3874 root = ext4_iget(sb, EXT4_ROOT_INO);
3876 ext4_msg(sb, KERN_ERR, "get root inode failed");
3877 ret = PTR_ERR(root);
3881 if (!S_ISDIR(root->i_mode) || !root->i_blocks || !root->i_size) {
3882 ext4_msg(sb, KERN_ERR, "corrupt root inode, run e2fsck");
3886 sb->s_root = d_make_root(root);
3888 ext4_msg(sb, KERN_ERR, "get root dentry failed");
3893 if (ext4_setup_super(sb, es, sb->s_flags & MS_RDONLY))
3894 sb->s_flags |= MS_RDONLY;
3896 /* determine the minimum size of new large inodes, if present */
3897 if (sbi->s_inode_size > EXT4_GOOD_OLD_INODE_SIZE) {
3898 sbi->s_want_extra_isize = sizeof(struct ext4_inode) -
3899 EXT4_GOOD_OLD_INODE_SIZE;
3900 if (EXT4_HAS_RO_COMPAT_FEATURE(sb,
3901 EXT4_FEATURE_RO_COMPAT_EXTRA_ISIZE)) {
3902 if (sbi->s_want_extra_isize <
3903 le16_to_cpu(es->s_want_extra_isize))
3904 sbi->s_want_extra_isize =
3905 le16_to_cpu(es->s_want_extra_isize);
3906 if (sbi->s_want_extra_isize <
3907 le16_to_cpu(es->s_min_extra_isize))
3908 sbi->s_want_extra_isize =
3909 le16_to_cpu(es->s_min_extra_isize);
3912 /* Check if enough inode space is available */
3913 if (EXT4_GOOD_OLD_INODE_SIZE + sbi->s_want_extra_isize >
3914 sbi->s_inode_size) {
3915 sbi->s_want_extra_isize = sizeof(struct ext4_inode) -
3916 EXT4_GOOD_OLD_INODE_SIZE;
3917 ext4_msg(sb, KERN_INFO, "required extra inode space not"
3921 err = ext4_setup_system_zone(sb);
3923 ext4_msg(sb, KERN_ERR, "failed to initialize system "
3925 goto failed_mount4a;
3929 err = ext4_mb_init(sb);
3931 ext4_msg(sb, KERN_ERR, "failed to initialize mballoc (%d)",
3936 err = ext4_register_li_request(sb, first_not_zeroed);
3940 sbi->s_kobj.kset = ext4_kset;
3941 init_completion(&sbi->s_kobj_unregister);
3942 err = kobject_init_and_add(&sbi->s_kobj, &ext4_ktype, NULL,
3948 /* Enable quota usage during mount. */
3949 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_QUOTA) &&
3950 !(sb->s_flags & MS_RDONLY)) {
3951 err = ext4_enable_quotas(sb);
3955 #endif /* CONFIG_QUOTA */
3957 EXT4_SB(sb)->s_mount_state |= EXT4_ORPHAN_FS;
3958 ext4_orphan_cleanup(sb, es);
3959 EXT4_SB(sb)->s_mount_state &= ~EXT4_ORPHAN_FS;
3960 if (needs_recovery) {
3961 ext4_msg(sb, KERN_INFO, "recovery complete");
3962 ext4_mark_recovery_complete(sb, es);
3964 if (EXT4_SB(sb)->s_journal) {
3965 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA)
3966 descr = " journalled data mode";
3967 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA)
3968 descr = " ordered data mode";
3970 descr = " writeback data mode";
3972 descr = "out journal";
3974 if (test_opt(sb, DISCARD)) {
3975 struct request_queue *q = bdev_get_queue(sb->s_bdev);
3976 if (!blk_queue_discard(q))
3977 ext4_msg(sb, KERN_WARNING,
3978 "mounting with \"discard\" option, but "
3979 "the device does not support discard");
3982 ext4_msg(sb, KERN_INFO, "mounted filesystem with%s. "
3983 "Opts: %s%s%s", descr, sbi->s_es->s_mount_opts,
3984 *sbi->s_es->s_mount_opts ? "; " : "", orig_data);
3986 if (es->s_error_count)
3987 mod_timer(&sbi->s_err_report, jiffies + 300*HZ); /* 5 minutes */
3994 ext4_msg(sb, KERN_ERR, "VFS: Can't find ext4 filesystem");
3999 kobject_del(&sbi->s_kobj);
4002 ext4_unregister_li_request(sb);
4004 ext4_mb_release(sb);
4006 ext4_ext_release(sb);
4007 ext4_release_system_zone(sb);
4012 ext4_msg(sb, KERN_ERR, "mount failed");
4013 destroy_workqueue(EXT4_SB(sb)->dio_unwritten_wq);
4015 if (sbi->s_journal) {
4016 jbd2_journal_destroy(sbi->s_journal);
4017 sbi->s_journal = NULL;
4020 ext4_es_unregister_shrinker(sb);
4021 del_timer(&sbi->s_err_report);
4022 if (sbi->s_flex_groups)
4023 ext4_kvfree(sbi->s_flex_groups);
4024 percpu_counter_destroy(&sbi->s_freeclusters_counter);
4025 percpu_counter_destroy(&sbi->s_freeinodes_counter);
4026 percpu_counter_destroy(&sbi->s_dirs_counter);
4027 percpu_counter_destroy(&sbi->s_dirtyclusters_counter);
4028 percpu_counter_destroy(&sbi->s_extent_cache_cnt);
4030 kthread_stop(sbi->s_mmp_tsk);
4032 for (i = 0; i < db_count; i++)
4033 brelse(sbi->s_group_desc[i]);
4034 ext4_kvfree(sbi->s_group_desc);
4036 if (sbi->s_chksum_driver)
4037 crypto_free_shash(sbi->s_chksum_driver);
4039 remove_proc_entry("options", sbi->s_proc);
4040 remove_proc_entry(sb->s_id, ext4_proc_root);
4043 for (i = 0; i < MAXQUOTAS; i++)
4044 kfree(sbi->s_qf_names[i]);
4046 ext4_blkdev_remove(sbi);
4049 sb->s_fs_info = NULL;
4050 kfree(sbi->s_blockgroup_lock);
4054 return err ? err : ret;
4058 * Setup any per-fs journal parameters now. We'll do this both on
4059 * initial mount, once the journal has been initialised but before we've
4060 * done any recovery; and again on any subsequent remount.
4062 static void ext4_init_journal_params(struct super_block *sb, journal_t *journal)
4064 struct ext4_sb_info *sbi = EXT4_SB(sb);
4066 journal->j_commit_interval = sbi->s_commit_interval;
4067 journal->j_min_batch_time = sbi->s_min_batch_time;
4068 journal->j_max_batch_time = sbi->s_max_batch_time;
4070 write_lock(&journal->j_state_lock);
4071 if (test_opt(sb, BARRIER))
4072 journal->j_flags |= JBD2_BARRIER;
4074 journal->j_flags &= ~JBD2_BARRIER;
4075 if (test_opt(sb, DATA_ERR_ABORT))
4076 journal->j_flags |= JBD2_ABORT_ON_SYNCDATA_ERR;
4078 journal->j_flags &= ~JBD2_ABORT_ON_SYNCDATA_ERR;
4079 write_unlock(&journal->j_state_lock);
4082 static journal_t *ext4_get_journal(struct super_block *sb,
4083 unsigned int journal_inum)
4085 struct inode *journal_inode;
4088 BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL));
4090 /* First, test for the existence of a valid inode on disk. Bad
4091 * things happen if we iget() an unused inode, as the subsequent
4092 * iput() will try to delete it. */
4094 journal_inode = ext4_iget(sb, journal_inum);
4095 if (IS_ERR(journal_inode)) {
4096 ext4_msg(sb, KERN_ERR, "no journal found");
4099 if (!journal_inode->i_nlink) {
4100 make_bad_inode(journal_inode);
4101 iput(journal_inode);
4102 ext4_msg(sb, KERN_ERR, "journal inode is deleted");
4106 jbd_debug(2, "Journal inode found at %p: %lld bytes\n",
4107 journal_inode, journal_inode->i_size);
4108 if (!S_ISREG(journal_inode->i_mode)) {
4109 ext4_msg(sb, KERN_ERR, "invalid journal inode");
4110 iput(journal_inode);
4114 journal = jbd2_journal_init_inode(journal_inode);
4116 ext4_msg(sb, KERN_ERR, "Could not load journal inode");
4117 iput(journal_inode);
4120 journal->j_private = sb;
4121 ext4_init_journal_params(sb, journal);
4125 static journal_t *ext4_get_dev_journal(struct super_block *sb,
4128 struct buffer_head *bh;
4132 int hblock, blocksize;
4133 ext4_fsblk_t sb_block;
4134 unsigned long offset;
4135 struct ext4_super_block *es;
4136 struct block_device *bdev;
4138 BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL));
4140 bdev = ext4_blkdev_get(j_dev, sb);
4144 blocksize = sb->s_blocksize;
4145 hblock = bdev_logical_block_size(bdev);
4146 if (blocksize < hblock) {
4147 ext4_msg(sb, KERN_ERR,
4148 "blocksize too small for journal device");
4152 sb_block = EXT4_MIN_BLOCK_SIZE / blocksize;
4153 offset = EXT4_MIN_BLOCK_SIZE % blocksize;
4154 set_blocksize(bdev, blocksize);
4155 if (!(bh = __bread(bdev, sb_block, blocksize))) {
4156 ext4_msg(sb, KERN_ERR, "couldn't read superblock of "
4157 "external journal");
4161 es = (struct ext4_super_block *) (bh->b_data + offset);
4162 if ((le16_to_cpu(es->s_magic) != EXT4_SUPER_MAGIC) ||
4163 !(le32_to_cpu(es->s_feature_incompat) &
4164 EXT4_FEATURE_INCOMPAT_JOURNAL_DEV)) {
4165 ext4_msg(sb, KERN_ERR, "external journal has "
4171 if (memcmp(EXT4_SB(sb)->s_es->s_journal_uuid, es->s_uuid, 16)) {
4172 ext4_msg(sb, KERN_ERR, "journal UUID does not match");
4177 len = ext4_blocks_count(es);
4178 start = sb_block + 1;
4179 brelse(bh); /* we're done with the superblock */
4181 journal = jbd2_journal_init_dev(bdev, sb->s_bdev,
4182 start, len, blocksize);
4184 ext4_msg(sb, KERN_ERR, "failed to create device journal");
4187 journal->j_private = sb;
4188 ll_rw_block(READ, 1, &journal->j_sb_buffer);
4189 wait_on_buffer(journal->j_sb_buffer);
4190 if (!buffer_uptodate(journal->j_sb_buffer)) {
4191 ext4_msg(sb, KERN_ERR, "I/O error on journal device");
4194 if (be32_to_cpu(journal->j_superblock->s_nr_users) != 1) {
4195 ext4_msg(sb, KERN_ERR, "External journal has more than one "
4196 "user (unsupported) - %d",
4197 be32_to_cpu(journal->j_superblock->s_nr_users));
4200 EXT4_SB(sb)->journal_bdev = bdev;
4201 ext4_init_journal_params(sb, journal);
4205 jbd2_journal_destroy(journal);
4207 ext4_blkdev_put(bdev);
4211 static int ext4_load_journal(struct super_block *sb,
4212 struct ext4_super_block *es,
4213 unsigned long journal_devnum)
4216 unsigned int journal_inum = le32_to_cpu(es->s_journal_inum);
4219 int really_read_only;
4221 BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL));
4223 if (journal_devnum &&
4224 journal_devnum != le32_to_cpu(es->s_journal_dev)) {
4225 ext4_msg(sb, KERN_INFO, "external journal device major/minor "
4226 "numbers have changed");
4227 journal_dev = new_decode_dev(journal_devnum);
4229 journal_dev = new_decode_dev(le32_to_cpu(es->s_journal_dev));
4231 really_read_only = bdev_read_only(sb->s_bdev);
4234 * Are we loading a blank journal or performing recovery after a
4235 * crash? For recovery, we need to check in advance whether we
4236 * can get read-write access to the device.
4238 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER)) {
4239 if (sb->s_flags & MS_RDONLY) {
4240 ext4_msg(sb, KERN_INFO, "INFO: recovery "
4241 "required on readonly filesystem");
4242 if (really_read_only) {
4243 ext4_msg(sb, KERN_ERR, "write access "
4244 "unavailable, cannot proceed");
4247 ext4_msg(sb, KERN_INFO, "write access will "
4248 "be enabled during recovery");
4252 if (journal_inum && journal_dev) {
4253 ext4_msg(sb, KERN_ERR, "filesystem has both journal "
4254 "and inode journals!");
4259 if (!(journal = ext4_get_journal(sb, journal_inum)))
4262 if (!(journal = ext4_get_dev_journal(sb, journal_dev)))
4266 if (!(journal->j_flags & JBD2_BARRIER))
4267 ext4_msg(sb, KERN_INFO, "barriers disabled");
4269 if (!EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER))
4270 err = jbd2_journal_wipe(journal, !really_read_only);
4272 char *save = kmalloc(EXT4_S_ERR_LEN, GFP_KERNEL);
4274 memcpy(save, ((char *) es) +
4275 EXT4_S_ERR_START, EXT4_S_ERR_LEN);
4276 err = jbd2_journal_load(journal);
4278 memcpy(((char *) es) + EXT4_S_ERR_START,
4279 save, EXT4_S_ERR_LEN);
4284 ext4_msg(sb, KERN_ERR, "error loading journal");
4285 jbd2_journal_destroy(journal);
4289 EXT4_SB(sb)->s_journal = journal;
4290 ext4_clear_journal_err(sb, es);
4292 if (!really_read_only && journal_devnum &&
4293 journal_devnum != le32_to_cpu(es->s_journal_dev)) {
4294 es->s_journal_dev = cpu_to_le32(journal_devnum);
4296 /* Make sure we flush the recovery flag to disk. */
4297 ext4_commit_super(sb, 1);
4303 static int ext4_commit_super(struct super_block *sb, int sync)
4305 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
4306 struct buffer_head *sbh = EXT4_SB(sb)->s_sbh;
4309 if (!sbh || block_device_ejected(sb))
4311 if (buffer_write_io_error(sbh)) {
4313 * Oh, dear. A previous attempt to write the
4314 * superblock failed. This could happen because the
4315 * USB device was yanked out. Or it could happen to
4316 * be a transient write error and maybe the block will
4317 * be remapped. Nothing we can do but to retry the
4318 * write and hope for the best.
4320 ext4_msg(sb, KERN_ERR, "previous I/O error to "
4321 "superblock detected");
4322 clear_buffer_write_io_error(sbh);
4323 set_buffer_uptodate(sbh);
4326 * If the file system is mounted read-only, don't update the
4327 * superblock write time. This avoids updating the superblock
4328 * write time when we are mounting the root file system
4329 * read/only but we need to replay the journal; at that point,
4330 * for people who are east of GMT and who make their clock
4331 * tick in localtime for Windows bug-for-bug compatibility,
4332 * the clock is set in the future, and this will cause e2fsck
4333 * to complain and force a full file system check.
4335 if (!(sb->s_flags & MS_RDONLY))
4336 es->s_wtime = cpu_to_le32(get_seconds());
4337 if (sb->s_bdev->bd_part)
4338 es->s_kbytes_written =
4339 cpu_to_le64(EXT4_SB(sb)->s_kbytes_written +
4340 ((part_stat_read(sb->s_bdev->bd_part, sectors[1]) -
4341 EXT4_SB(sb)->s_sectors_written_start) >> 1));
4343 es->s_kbytes_written =
4344 cpu_to_le64(EXT4_SB(sb)->s_kbytes_written);
4345 ext4_free_blocks_count_set(es,
4346 EXT4_C2B(EXT4_SB(sb), percpu_counter_sum_positive(
4347 &EXT4_SB(sb)->s_freeclusters_counter)));
4348 es->s_free_inodes_count =
4349 cpu_to_le32(percpu_counter_sum_positive(
4350 &EXT4_SB(sb)->s_freeinodes_counter));
4351 BUFFER_TRACE(sbh, "marking dirty");
4352 ext4_superblock_csum_set(sb);
4353 mark_buffer_dirty(sbh);
4355 error = sync_dirty_buffer(sbh);
4359 error = buffer_write_io_error(sbh);
4361 ext4_msg(sb, KERN_ERR, "I/O error while writing "
4363 clear_buffer_write_io_error(sbh);
4364 set_buffer_uptodate(sbh);
4371 * Have we just finished recovery? If so, and if we are mounting (or
4372 * remounting) the filesystem readonly, then we will end up with a
4373 * consistent fs on disk. Record that fact.
4375 static void ext4_mark_recovery_complete(struct super_block *sb,
4376 struct ext4_super_block *es)
4378 journal_t *journal = EXT4_SB(sb)->s_journal;
4380 if (!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL)) {
4381 BUG_ON(journal != NULL);
4384 jbd2_journal_lock_updates(journal);
4385 if (jbd2_journal_flush(journal) < 0)
4388 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER) &&
4389 sb->s_flags & MS_RDONLY) {
4390 EXT4_CLEAR_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
4391 ext4_commit_super(sb, 1);
4395 jbd2_journal_unlock_updates(journal);
4399 * If we are mounting (or read-write remounting) a filesystem whose journal
4400 * has recorded an error from a previous lifetime, move that error to the
4401 * main filesystem now.
4403 static void ext4_clear_journal_err(struct super_block *sb,
4404 struct ext4_super_block *es)
4410 BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL));
4412 journal = EXT4_SB(sb)->s_journal;
4415 * Now check for any error status which may have been recorded in the
4416 * journal by a prior ext4_error() or ext4_abort()
4419 j_errno = jbd2_journal_errno(journal);
4423 errstr = ext4_decode_error(sb, j_errno, nbuf);
4424 ext4_warning(sb, "Filesystem error recorded "
4425 "from previous mount: %s", errstr);
4426 ext4_warning(sb, "Marking fs in need of filesystem check.");
4428 EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS;
4429 es->s_state |= cpu_to_le16(EXT4_ERROR_FS);
4430 ext4_commit_super(sb, 1);
4432 jbd2_journal_clear_err(journal);
4433 jbd2_journal_update_sb_errno(journal);
4438 * Force the running and committing transactions to commit,
4439 * and wait on the commit.
4441 int ext4_force_commit(struct super_block *sb)
4445 if (sb->s_flags & MS_RDONLY)
4448 journal = EXT4_SB(sb)->s_journal;
4449 return ext4_journal_force_commit(journal);
4452 static int ext4_sync_fs(struct super_block *sb, int wait)
4456 struct ext4_sb_info *sbi = EXT4_SB(sb);
4458 trace_ext4_sync_fs(sb, wait);
4459 flush_workqueue(sbi->dio_unwritten_wq);
4461 * Writeback quota in non-journalled quota case - journalled quota has
4464 dquot_writeback_dquots(sb, -1);
4465 if (jbd2_journal_start_commit(sbi->s_journal, &target)) {
4467 jbd2_log_wait_commit(sbi->s_journal, target);
4473 * LVM calls this function before a (read-only) snapshot is created. This
4474 * gives us a chance to flush the journal completely and mark the fs clean.
4476 * Note that only this function cannot bring a filesystem to be in a clean
4477 * state independently. It relies on upper layer to stop all data & metadata
4480 static int ext4_freeze(struct super_block *sb)
4485 if (sb->s_flags & MS_RDONLY)
4488 journal = EXT4_SB(sb)->s_journal;
4490 /* Now we set up the journal barrier. */
4491 jbd2_journal_lock_updates(journal);
4494 * Don't clear the needs_recovery flag if we failed to flush
4497 error = jbd2_journal_flush(journal);
4501 /* Journal blocked and flushed, clear needs_recovery flag. */
4502 EXT4_CLEAR_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
4503 error = ext4_commit_super(sb, 1);
4505 /* we rely on upper layer to stop further updates */
4506 jbd2_journal_unlock_updates(EXT4_SB(sb)->s_journal);
4511 * Called by LVM after the snapshot is done. We need to reset the RECOVER
4512 * flag here, even though the filesystem is not technically dirty yet.
4514 static int ext4_unfreeze(struct super_block *sb)
4516 if (sb->s_flags & MS_RDONLY)
4519 /* Reset the needs_recovery flag before the fs is unlocked. */
4520 EXT4_SET_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
4521 ext4_commit_super(sb, 1);
4526 * Structure to save mount options for ext4_remount's benefit
4528 struct ext4_mount_options {
4529 unsigned long s_mount_opt;
4530 unsigned long s_mount_opt2;
4533 unsigned long s_commit_interval;
4534 u32 s_min_batch_time, s_max_batch_time;
4537 char *s_qf_names[MAXQUOTAS];
4541 static int ext4_remount(struct super_block *sb, int *flags, char *data)
4543 struct ext4_super_block *es;
4544 struct ext4_sb_info *sbi = EXT4_SB(sb);
4545 unsigned long old_sb_flags;
4546 struct ext4_mount_options old_opts;
4547 int enable_quota = 0;
4549 unsigned int journal_ioprio = DEFAULT_JOURNAL_IOPRIO;
4554 char *orig_data = kstrdup(data, GFP_KERNEL);
4556 /* Store the original options */
4557 old_sb_flags = sb->s_flags;
4558 old_opts.s_mount_opt = sbi->s_mount_opt;
4559 old_opts.s_mount_opt2 = sbi->s_mount_opt2;
4560 old_opts.s_resuid = sbi->s_resuid;
4561 old_opts.s_resgid = sbi->s_resgid;
4562 old_opts.s_commit_interval = sbi->s_commit_interval;
4563 old_opts.s_min_batch_time = sbi->s_min_batch_time;
4564 old_opts.s_max_batch_time = sbi->s_max_batch_time;
4566 old_opts.s_jquota_fmt = sbi->s_jquota_fmt;
4567 for (i = 0; i < MAXQUOTAS; i++)
4568 if (sbi->s_qf_names[i]) {
4569 old_opts.s_qf_names[i] = kstrdup(sbi->s_qf_names[i],
4571 if (!old_opts.s_qf_names[i]) {
4572 for (j = 0; j < i; j++)
4573 kfree(old_opts.s_qf_names[j]);
4578 old_opts.s_qf_names[i] = NULL;
4580 if (sbi->s_journal && sbi->s_journal->j_task->io_context)
4581 journal_ioprio = sbi->s_journal->j_task->io_context->ioprio;
4584 * Allow the "check" option to be passed as a remount option.
4586 if (!parse_options(data, sb, NULL, &journal_ioprio, 1)) {
4591 if (sbi->s_mount_flags & EXT4_MF_FS_ABORTED)
4592 ext4_abort(sb, "Abort forced by user");
4594 sb->s_flags = (sb->s_flags & ~MS_POSIXACL) |
4595 (test_opt(sb, POSIX_ACL) ? MS_POSIXACL : 0);
4599 if (sbi->s_journal) {
4600 ext4_init_journal_params(sb, sbi->s_journal);
4601 set_task_ioprio(sbi->s_journal->j_task, journal_ioprio);
4604 if ((*flags & MS_RDONLY) != (sb->s_flags & MS_RDONLY)) {
4605 if (sbi->s_mount_flags & EXT4_MF_FS_ABORTED) {
4610 if (*flags & MS_RDONLY) {
4611 err = dquot_suspend(sb, -1);
4616 * First of all, the unconditional stuff we have to do
4617 * to disable replay of the journal when we next remount
4619 sb->s_flags |= MS_RDONLY;
4622 * OK, test if we are remounting a valid rw partition
4623 * readonly, and if so set the rdonly flag and then
4624 * mark the partition as valid again.
4626 if (!(es->s_state & cpu_to_le16(EXT4_VALID_FS)) &&
4627 (sbi->s_mount_state & EXT4_VALID_FS))
4628 es->s_state = cpu_to_le16(sbi->s_mount_state);
4631 ext4_mark_recovery_complete(sb, es);
4633 /* Make sure we can mount this feature set readwrite */
4634 if (!ext4_feature_set_ok(sb, 0)) {
4639 * Make sure the group descriptor checksums
4640 * are sane. If they aren't, refuse to remount r/w.
4642 for (g = 0; g < sbi->s_groups_count; g++) {
4643 struct ext4_group_desc *gdp =
4644 ext4_get_group_desc(sb, g, NULL);
4646 if (!ext4_group_desc_csum_verify(sb, g, gdp)) {
4647 ext4_msg(sb, KERN_ERR,
4648 "ext4_remount: Checksum for group %u failed (%u!=%u)",
4649 g, le16_to_cpu(ext4_group_desc_csum(sbi, g, gdp)),
4650 le16_to_cpu(gdp->bg_checksum));
4657 * If we have an unprocessed orphan list hanging
4658 * around from a previously readonly bdev mount,
4659 * require a full umount/remount for now.
4661 if (es->s_last_orphan) {
4662 ext4_msg(sb, KERN_WARNING, "Couldn't "
4663 "remount RDWR because of unprocessed "
4664 "orphan inode list. Please "
4665 "umount/remount instead");
4671 * Mounting a RDONLY partition read-write, so reread
4672 * and store the current valid flag. (It may have
4673 * been changed by e2fsck since we originally mounted
4677 ext4_clear_journal_err(sb, es);
4678 sbi->s_mount_state = le16_to_cpu(es->s_state);
4679 if (!ext4_setup_super(sb, es, 0))
4680 sb->s_flags &= ~MS_RDONLY;
4681 if (EXT4_HAS_INCOMPAT_FEATURE(sb,
4682 EXT4_FEATURE_INCOMPAT_MMP))
4683 if (ext4_multi_mount_protect(sb,
4684 le64_to_cpu(es->s_mmp_block))) {
4693 * Reinitialize lazy itable initialization thread based on
4696 if ((sb->s_flags & MS_RDONLY) || !test_opt(sb, INIT_INODE_TABLE))
4697 ext4_unregister_li_request(sb);
4699 ext4_group_t first_not_zeroed;
4700 first_not_zeroed = ext4_has_uninit_itable(sb);
4701 ext4_register_li_request(sb, first_not_zeroed);
4704 ext4_setup_system_zone(sb);
4705 if (sbi->s_journal == NULL && !(old_sb_flags & MS_RDONLY))
4706 ext4_commit_super(sb, 1);
4709 /* Release old quota file names */
4710 for (i = 0; i < MAXQUOTAS; i++)
4711 kfree(old_opts.s_qf_names[i]);
4713 if (sb_any_quota_suspended(sb))
4714 dquot_resume(sb, -1);
4715 else if (EXT4_HAS_RO_COMPAT_FEATURE(sb,
4716 EXT4_FEATURE_RO_COMPAT_QUOTA)) {
4717 err = ext4_enable_quotas(sb);
4724 ext4_msg(sb, KERN_INFO, "re-mounted. Opts: %s", orig_data);
4729 sb->s_flags = old_sb_flags;
4730 sbi->s_mount_opt = old_opts.s_mount_opt;
4731 sbi->s_mount_opt2 = old_opts.s_mount_opt2;
4732 sbi->s_resuid = old_opts.s_resuid;
4733 sbi->s_resgid = old_opts.s_resgid;
4734 sbi->s_commit_interval = old_opts.s_commit_interval;
4735 sbi->s_min_batch_time = old_opts.s_min_batch_time;
4736 sbi->s_max_batch_time = old_opts.s_max_batch_time;
4738 sbi->s_jquota_fmt = old_opts.s_jquota_fmt;
4739 for (i = 0; i < MAXQUOTAS; i++) {
4740 kfree(sbi->s_qf_names[i]);
4741 sbi->s_qf_names[i] = old_opts.s_qf_names[i];
4748 static int ext4_statfs(struct dentry *dentry, struct kstatfs *buf)
4750 struct super_block *sb = dentry->d_sb;
4751 struct ext4_sb_info *sbi = EXT4_SB(sb);
4752 struct ext4_super_block *es = sbi->s_es;
4753 ext4_fsblk_t overhead = 0;
4757 if (!test_opt(sb, MINIX_DF))
4758 overhead = sbi->s_overhead;
4760 buf->f_type = EXT4_SUPER_MAGIC;
4761 buf->f_bsize = sb->s_blocksize;
4762 buf->f_blocks = ext4_blocks_count(es) - EXT4_C2B(sbi, overhead);
4763 bfree = percpu_counter_sum_positive(&sbi->s_freeclusters_counter) -
4764 percpu_counter_sum_positive(&sbi->s_dirtyclusters_counter);
4765 /* prevent underflow in case that few free space is available */
4766 buf->f_bfree = EXT4_C2B(sbi, max_t(s64, bfree, 0));
4767 buf->f_bavail = buf->f_bfree - ext4_r_blocks_count(es);
4768 if (buf->f_bfree < ext4_r_blocks_count(es))
4770 buf->f_files = le32_to_cpu(es->s_inodes_count);
4771 buf->f_ffree = percpu_counter_sum_positive(&sbi->s_freeinodes_counter);
4772 buf->f_namelen = EXT4_NAME_LEN;
4773 fsid = le64_to_cpup((void *)es->s_uuid) ^
4774 le64_to_cpup((void *)es->s_uuid + sizeof(u64));
4775 buf->f_fsid.val[0] = fsid & 0xFFFFFFFFUL;
4776 buf->f_fsid.val[1] = (fsid >> 32) & 0xFFFFFFFFUL;
4781 /* Helper function for writing quotas on sync - we need to start transaction
4782 * before quota file is locked for write. Otherwise the are possible deadlocks:
4783 * Process 1 Process 2
4784 * ext4_create() quota_sync()
4785 * jbd2_journal_start() write_dquot()
4786 * dquot_initialize() down(dqio_mutex)
4787 * down(dqio_mutex) jbd2_journal_start()
4793 static inline struct inode *dquot_to_inode(struct dquot *dquot)
4795 return sb_dqopt(dquot->dq_sb)->files[dquot->dq_id.type];
4798 static int ext4_write_dquot(struct dquot *dquot)
4802 struct inode *inode;
4804 inode = dquot_to_inode(dquot);
4805 handle = ext4_journal_start(inode, EXT4_HT_QUOTA,
4806 EXT4_QUOTA_TRANS_BLOCKS(dquot->dq_sb));
4808 return PTR_ERR(handle);
4809 ret = dquot_commit(dquot);
4810 err = ext4_journal_stop(handle);
4816 static int ext4_acquire_dquot(struct dquot *dquot)
4821 handle = ext4_journal_start(dquot_to_inode(dquot), EXT4_HT_QUOTA,
4822 EXT4_QUOTA_INIT_BLOCKS(dquot->dq_sb));
4824 return PTR_ERR(handle);
4825 ret = dquot_acquire(dquot);
4826 err = ext4_journal_stop(handle);
4832 static int ext4_release_dquot(struct dquot *dquot)
4837 handle = ext4_journal_start(dquot_to_inode(dquot), EXT4_HT_QUOTA,
4838 EXT4_QUOTA_DEL_BLOCKS(dquot->dq_sb));
4839 if (IS_ERR(handle)) {
4840 /* Release dquot anyway to avoid endless cycle in dqput() */
4841 dquot_release(dquot);
4842 return PTR_ERR(handle);
4844 ret = dquot_release(dquot);
4845 err = ext4_journal_stop(handle);
4851 static int ext4_mark_dquot_dirty(struct dquot *dquot)
4853 struct super_block *sb = dquot->dq_sb;
4854 struct ext4_sb_info *sbi = EXT4_SB(sb);
4856 /* Are we journaling quotas? */
4857 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_QUOTA) ||
4858 sbi->s_qf_names[USRQUOTA] || sbi->s_qf_names[GRPQUOTA]) {
4859 dquot_mark_dquot_dirty(dquot);
4860 return ext4_write_dquot(dquot);
4862 return dquot_mark_dquot_dirty(dquot);
4866 static int ext4_write_info(struct super_block *sb, int type)
4871 /* Data block + inode block */
4872 handle = ext4_journal_start(sb->s_root->d_inode, EXT4_HT_QUOTA, 2);
4874 return PTR_ERR(handle);
4875 ret = dquot_commit_info(sb, type);
4876 err = ext4_journal_stop(handle);
4883 * Turn on quotas during mount time - we need to find
4884 * the quota file and such...
4886 static int ext4_quota_on_mount(struct super_block *sb, int type)
4888 return dquot_quota_on_mount(sb, EXT4_SB(sb)->s_qf_names[type],
4889 EXT4_SB(sb)->s_jquota_fmt, type);
4893 * Standard function to be called on quota_on
4895 static int ext4_quota_on(struct super_block *sb, int type, int format_id,
4900 if (!test_opt(sb, QUOTA))
4903 /* Quotafile not on the same filesystem? */
4904 if (path->dentry->d_sb != sb)
4906 /* Journaling quota? */
4907 if (EXT4_SB(sb)->s_qf_names[type]) {
4908 /* Quotafile not in fs root? */
4909 if (path->dentry->d_parent != sb->s_root)
4910 ext4_msg(sb, KERN_WARNING,
4911 "Quota file not on filesystem root. "
4912 "Journaled quota will not work");
4916 * When we journal data on quota file, we have to flush journal to see
4917 * all updates to the file when we bypass pagecache...
4919 if (EXT4_SB(sb)->s_journal &&
4920 ext4_should_journal_data(path->dentry->d_inode)) {
4922 * We don't need to lock updates but journal_flush() could
4923 * otherwise be livelocked...
4925 jbd2_journal_lock_updates(EXT4_SB(sb)->s_journal);
4926 err = jbd2_journal_flush(EXT4_SB(sb)->s_journal);
4927 jbd2_journal_unlock_updates(EXT4_SB(sb)->s_journal);
4932 return dquot_quota_on(sb, type, format_id, path);
4935 static int ext4_quota_enable(struct super_block *sb, int type, int format_id,
4939 struct inode *qf_inode;
4940 unsigned long qf_inums[MAXQUOTAS] = {
4941 le32_to_cpu(EXT4_SB(sb)->s_es->s_usr_quota_inum),
4942 le32_to_cpu(EXT4_SB(sb)->s_es->s_grp_quota_inum)
4945 BUG_ON(!EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_QUOTA));
4947 if (!qf_inums[type])
4950 qf_inode = ext4_iget(sb, qf_inums[type]);
4951 if (IS_ERR(qf_inode)) {
4952 ext4_error(sb, "Bad quota inode # %lu", qf_inums[type]);
4953 return PTR_ERR(qf_inode);
4956 err = dquot_enable(qf_inode, type, format_id, flags);
4962 /* Enable usage tracking for all quota types. */
4963 static int ext4_enable_quotas(struct super_block *sb)
4966 unsigned long qf_inums[MAXQUOTAS] = {
4967 le32_to_cpu(EXT4_SB(sb)->s_es->s_usr_quota_inum),
4968 le32_to_cpu(EXT4_SB(sb)->s_es->s_grp_quota_inum)
4971 sb_dqopt(sb)->flags |= DQUOT_QUOTA_SYS_FILE;
4972 for (type = 0; type < MAXQUOTAS; type++) {
4973 if (qf_inums[type]) {
4974 err = ext4_quota_enable(sb, type, QFMT_VFS_V1,
4975 DQUOT_USAGE_ENABLED);
4978 "Failed to enable quota tracking "
4979 "(type=%d, err=%d). Please run "
4980 "e2fsck to fix.", type, err);
4989 * quota_on function that is used when QUOTA feature is set.
4991 static int ext4_quota_on_sysfile(struct super_block *sb, int type,
4994 if (!EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_QUOTA))
4998 * USAGE was enabled at mount time. Only need to enable LIMITS now.
5000 return ext4_quota_enable(sb, type, format_id, DQUOT_LIMITS_ENABLED);
5003 static int ext4_quota_off(struct super_block *sb, int type)
5005 struct inode *inode = sb_dqopt(sb)->files[type];
5008 /* Force all delayed allocation blocks to be allocated.
5009 * Caller already holds s_umount sem */
5010 if (test_opt(sb, DELALLOC))
5011 sync_filesystem(sb);
5016 /* Update modification times of quota files when userspace can
5017 * start looking at them */
5018 handle = ext4_journal_start(inode, EXT4_HT_QUOTA, 1);
5021 inode->i_mtime = inode->i_ctime = CURRENT_TIME;
5022 ext4_mark_inode_dirty(handle, inode);
5023 ext4_journal_stop(handle);
5026 return dquot_quota_off(sb, type);
5030 * quota_off function that is used when QUOTA feature is set.
5032 static int ext4_quota_off_sysfile(struct super_block *sb, int type)
5034 if (!EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_QUOTA))
5037 /* Disable only the limits. */
5038 return dquot_disable(sb, type, DQUOT_LIMITS_ENABLED);
5041 /* Read data from quotafile - avoid pagecache and such because we cannot afford
5042 * acquiring the locks... As quota files are never truncated and quota code
5043 * itself serializes the operations (and no one else should touch the files)
5044 * we don't have to be afraid of races */
5045 static ssize_t ext4_quota_read(struct super_block *sb, int type, char *data,
5046 size_t len, loff_t off)
5048 struct inode *inode = sb_dqopt(sb)->files[type];
5049 ext4_lblk_t blk = off >> EXT4_BLOCK_SIZE_BITS(sb);
5051 int offset = off & (sb->s_blocksize - 1);
5054 struct buffer_head *bh;
5055 loff_t i_size = i_size_read(inode);
5059 if (off+len > i_size)
5062 while (toread > 0) {
5063 tocopy = sb->s_blocksize - offset < toread ?
5064 sb->s_blocksize - offset : toread;
5065 bh = ext4_bread(NULL, inode, blk, 0, &err);
5068 if (!bh) /* A hole? */
5069 memset(data, 0, tocopy);
5071 memcpy(data, bh->b_data+offset, tocopy);
5081 /* Write to quotafile (we know the transaction is already started and has
5082 * enough credits) */
5083 static ssize_t ext4_quota_write(struct super_block *sb, int type,
5084 const char *data, size_t len, loff_t off)
5086 struct inode *inode = sb_dqopt(sb)->files[type];
5087 ext4_lblk_t blk = off >> EXT4_BLOCK_SIZE_BITS(sb);
5089 int offset = off & (sb->s_blocksize - 1);
5090 struct buffer_head *bh;
5091 handle_t *handle = journal_current_handle();
5093 if (EXT4_SB(sb)->s_journal && !handle) {
5094 ext4_msg(sb, KERN_WARNING, "Quota write (off=%llu, len=%llu)"
5095 " cancelled because transaction is not started",
5096 (unsigned long long)off, (unsigned long long)len);
5100 * Since we account only one data block in transaction credits,
5101 * then it is impossible to cross a block boundary.
5103 if (sb->s_blocksize - offset < len) {
5104 ext4_msg(sb, KERN_WARNING, "Quota write (off=%llu, len=%llu)"
5105 " cancelled because not block aligned",
5106 (unsigned long long)off, (unsigned long long)len);
5110 bh = ext4_bread(handle, inode, blk, 1, &err);
5113 err = ext4_journal_get_write_access(handle, bh);
5119 memcpy(bh->b_data+offset, data, len);
5120 flush_dcache_page(bh->b_page);
5122 err = ext4_handle_dirty_metadata(handle, NULL, bh);
5127 if (inode->i_size < off + len) {
5128 i_size_write(inode, off + len);
5129 EXT4_I(inode)->i_disksize = inode->i_size;
5130 ext4_mark_inode_dirty(handle, inode);
5137 static struct dentry *ext4_mount(struct file_system_type *fs_type, int flags,
5138 const char *dev_name, void *data)
5140 return mount_bdev(fs_type, flags, dev_name, data, ext4_fill_super);
5143 #if !defined(CONFIG_EXT2_FS) && !defined(CONFIG_EXT2_FS_MODULE) && defined(CONFIG_EXT4_USE_FOR_EXT23)
5144 static inline void register_as_ext2(void)
5146 int err = register_filesystem(&ext2_fs_type);
5149 "EXT4-fs: Unable to register as ext2 (%d)\n", err);
5152 static inline void unregister_as_ext2(void)
5154 unregister_filesystem(&ext2_fs_type);
5157 static inline int ext2_feature_set_ok(struct super_block *sb)
5159 if (EXT4_HAS_INCOMPAT_FEATURE(sb, ~EXT2_FEATURE_INCOMPAT_SUPP))
5161 if (sb->s_flags & MS_RDONLY)
5163 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, ~EXT2_FEATURE_RO_COMPAT_SUPP))
5168 static inline void register_as_ext2(void) { }
5169 static inline void unregister_as_ext2(void) { }
5170 static inline int ext2_feature_set_ok(struct super_block *sb) { return 0; }
5173 #if !defined(CONFIG_EXT3_FS) && !defined(CONFIG_EXT3_FS_MODULE) && defined(CONFIG_EXT4_USE_FOR_EXT23)
5174 static inline void register_as_ext3(void)
5176 int err = register_filesystem(&ext3_fs_type);
5179 "EXT4-fs: Unable to register as ext3 (%d)\n", err);
5182 static inline void unregister_as_ext3(void)
5184 unregister_filesystem(&ext3_fs_type);
5187 static inline int ext3_feature_set_ok(struct super_block *sb)
5189 if (EXT4_HAS_INCOMPAT_FEATURE(sb, ~EXT3_FEATURE_INCOMPAT_SUPP))
5191 if (!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL))
5193 if (sb->s_flags & MS_RDONLY)
5195 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, ~EXT3_FEATURE_RO_COMPAT_SUPP))
5200 static inline void register_as_ext3(void) { }
5201 static inline void unregister_as_ext3(void) { }
5202 static inline int ext3_feature_set_ok(struct super_block *sb) { return 0; }
5205 static struct file_system_type ext4_fs_type = {
5206 .owner = THIS_MODULE,
5208 .mount = ext4_mount,
5209 .kill_sb = kill_block_super,
5210 .fs_flags = FS_REQUIRES_DEV,
5212 MODULE_ALIAS_FS("ext4");
5214 static int __init ext4_init_feat_adverts(void)
5216 struct ext4_features *ef;
5219 ef = kzalloc(sizeof(struct ext4_features), GFP_KERNEL);
5223 ef->f_kobj.kset = ext4_kset;
5224 init_completion(&ef->f_kobj_unregister);
5225 ret = kobject_init_and_add(&ef->f_kobj, &ext4_feat_ktype, NULL,
5238 static void ext4_exit_feat_adverts(void)
5240 kobject_put(&ext4_feat->f_kobj);
5241 wait_for_completion(&ext4_feat->f_kobj_unregister);
5245 /* Shared across all ext4 file systems */
5246 wait_queue_head_t ext4__ioend_wq[EXT4_WQ_HASH_SZ];
5247 struct mutex ext4__aio_mutex[EXT4_WQ_HASH_SZ];
5249 static int __init ext4_init_fs(void)
5253 ext4_li_info = NULL;
5254 mutex_init(&ext4_li_mtx);
5256 /* Build-time check for flags consistency */
5257 ext4_check_flag_values();
5259 for (i = 0; i < EXT4_WQ_HASH_SZ; i++) {
5260 mutex_init(&ext4__aio_mutex[i]);
5261 init_waitqueue_head(&ext4__ioend_wq[i]);
5264 err = ext4_init_es();
5268 err = ext4_init_pageio();
5272 err = ext4_init_system_zone();
5275 ext4_kset = kset_create_and_add("ext4", NULL, fs_kobj);
5280 ext4_proc_root = proc_mkdir("fs/ext4", NULL);
5282 err = ext4_init_feat_adverts();
5286 err = ext4_init_mballoc();
5290 err = ext4_init_xattr();
5293 err = init_inodecache();
5298 err = register_filesystem(&ext4_fs_type);
5304 unregister_as_ext2();
5305 unregister_as_ext3();
5306 destroy_inodecache();
5310 ext4_exit_mballoc();
5312 ext4_exit_feat_adverts();
5315 remove_proc_entry("fs/ext4", NULL);
5316 kset_unregister(ext4_kset);
5318 ext4_exit_system_zone();
5327 static void __exit ext4_exit_fs(void)
5329 ext4_destroy_lazyinit_thread();
5330 unregister_as_ext2();
5331 unregister_as_ext3();
5332 unregister_filesystem(&ext4_fs_type);
5333 destroy_inodecache();
5335 ext4_exit_mballoc();
5336 ext4_exit_feat_adverts();
5337 remove_proc_entry("fs/ext4", NULL);
5338 kset_unregister(ext4_kset);
5339 ext4_exit_system_zone();
5343 MODULE_AUTHOR("Remy Card, Stephen Tweedie, Andrew Morton, Andreas Dilger, Theodore Ts'o and others");
5344 MODULE_DESCRIPTION("Fourth Extended Filesystem");
5345 MODULE_LICENSE("GPL");
5346 module_init(ext4_init_fs)
5347 module_exit(ext4_exit_fs)