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;
62 static int ext4_mballoc_ready;
64 static int ext4_load_journal(struct super_block *, struct ext4_super_block *,
65 unsigned long journal_devnum);
66 static int ext4_show_options(struct seq_file *seq, struct dentry *root);
67 static int ext4_commit_super(struct super_block *sb, int sync);
68 static void ext4_mark_recovery_complete(struct super_block *sb,
69 struct ext4_super_block *es);
70 static void ext4_clear_journal_err(struct super_block *sb,
71 struct ext4_super_block *es);
72 static int ext4_sync_fs(struct super_block *sb, int wait);
73 static int ext4_sync_fs_nojournal(struct super_block *sb, int wait);
74 static int ext4_remount(struct super_block *sb, int *flags, char *data);
75 static int ext4_statfs(struct dentry *dentry, struct kstatfs *buf);
76 static int ext4_unfreeze(struct super_block *sb);
77 static int ext4_freeze(struct super_block *sb);
78 static struct dentry *ext4_mount(struct file_system_type *fs_type, int flags,
79 const char *dev_name, void *data);
80 static inline int ext2_feature_set_ok(struct super_block *sb);
81 static inline int ext3_feature_set_ok(struct super_block *sb);
82 static int ext4_feature_set_ok(struct super_block *sb, int readonly);
83 static void ext4_destroy_lazyinit_thread(void);
84 static void ext4_unregister_li_request(struct super_block *sb);
85 static void ext4_clear_request_list(void);
86 static int ext4_reserve_clusters(struct ext4_sb_info *, ext4_fsblk_t);
88 #if !defined(CONFIG_EXT2_FS) && !defined(CONFIG_EXT2_FS_MODULE) && defined(CONFIG_EXT4_USE_FOR_EXT23)
89 static struct file_system_type ext2_fs_type = {
93 .kill_sb = kill_block_super,
94 .fs_flags = FS_REQUIRES_DEV,
96 MODULE_ALIAS_FS("ext2");
98 #define IS_EXT2_SB(sb) ((sb)->s_bdev->bd_holder == &ext2_fs_type)
100 #define IS_EXT2_SB(sb) (0)
104 #if !defined(CONFIG_EXT3_FS) && !defined(CONFIG_EXT3_FS_MODULE) && defined(CONFIG_EXT4_USE_FOR_EXT23)
105 static struct file_system_type ext3_fs_type = {
106 .owner = THIS_MODULE,
109 .kill_sb = kill_block_super,
110 .fs_flags = FS_REQUIRES_DEV,
112 MODULE_ALIAS_FS("ext3");
113 MODULE_ALIAS("ext3");
114 #define IS_EXT3_SB(sb) ((sb)->s_bdev->bd_holder == &ext3_fs_type)
116 #define IS_EXT3_SB(sb) (0)
119 static int ext4_verify_csum_type(struct super_block *sb,
120 struct ext4_super_block *es)
122 if (!EXT4_HAS_RO_COMPAT_FEATURE(sb,
123 EXT4_FEATURE_RO_COMPAT_METADATA_CSUM))
126 return es->s_checksum_type == EXT4_CRC32C_CHKSUM;
129 static __le32 ext4_superblock_csum(struct super_block *sb,
130 struct ext4_super_block *es)
132 struct ext4_sb_info *sbi = EXT4_SB(sb);
133 int offset = offsetof(struct ext4_super_block, s_checksum);
136 csum = ext4_chksum(sbi, ~0, (char *)es, offset);
138 return cpu_to_le32(csum);
141 static int ext4_superblock_csum_verify(struct super_block *sb,
142 struct ext4_super_block *es)
144 if (!EXT4_HAS_RO_COMPAT_FEATURE(sb,
145 EXT4_FEATURE_RO_COMPAT_METADATA_CSUM))
148 return es->s_checksum == ext4_superblock_csum(sb, es);
151 void ext4_superblock_csum_set(struct super_block *sb)
153 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
155 if (!EXT4_HAS_RO_COMPAT_FEATURE(sb,
156 EXT4_FEATURE_RO_COMPAT_METADATA_CSUM))
159 es->s_checksum = ext4_superblock_csum(sb, es);
162 void *ext4_kvmalloc(size_t size, gfp_t flags)
166 ret = kmalloc(size, flags | __GFP_NOWARN);
168 ret = __vmalloc(size, flags, PAGE_KERNEL);
172 void *ext4_kvzalloc(size_t size, gfp_t flags)
176 ret = kzalloc(size, flags | __GFP_NOWARN);
178 ret = __vmalloc(size, flags | __GFP_ZERO, PAGE_KERNEL);
182 void ext4_kvfree(void *ptr)
184 if (is_vmalloc_addr(ptr))
191 ext4_fsblk_t ext4_block_bitmap(struct super_block *sb,
192 struct ext4_group_desc *bg)
194 return le32_to_cpu(bg->bg_block_bitmap_lo) |
195 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
196 (ext4_fsblk_t)le32_to_cpu(bg->bg_block_bitmap_hi) << 32 : 0);
199 ext4_fsblk_t ext4_inode_bitmap(struct super_block *sb,
200 struct ext4_group_desc *bg)
202 return le32_to_cpu(bg->bg_inode_bitmap_lo) |
203 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
204 (ext4_fsblk_t)le32_to_cpu(bg->bg_inode_bitmap_hi) << 32 : 0);
207 ext4_fsblk_t ext4_inode_table(struct super_block *sb,
208 struct ext4_group_desc *bg)
210 return le32_to_cpu(bg->bg_inode_table_lo) |
211 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
212 (ext4_fsblk_t)le32_to_cpu(bg->bg_inode_table_hi) << 32 : 0);
215 __u32 ext4_free_group_clusters(struct super_block *sb,
216 struct ext4_group_desc *bg)
218 return le16_to_cpu(bg->bg_free_blocks_count_lo) |
219 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
220 (__u32)le16_to_cpu(bg->bg_free_blocks_count_hi) << 16 : 0);
223 __u32 ext4_free_inodes_count(struct super_block *sb,
224 struct ext4_group_desc *bg)
226 return le16_to_cpu(bg->bg_free_inodes_count_lo) |
227 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
228 (__u32)le16_to_cpu(bg->bg_free_inodes_count_hi) << 16 : 0);
231 __u32 ext4_used_dirs_count(struct super_block *sb,
232 struct ext4_group_desc *bg)
234 return le16_to_cpu(bg->bg_used_dirs_count_lo) |
235 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
236 (__u32)le16_to_cpu(bg->bg_used_dirs_count_hi) << 16 : 0);
239 __u32 ext4_itable_unused_count(struct super_block *sb,
240 struct ext4_group_desc *bg)
242 return le16_to_cpu(bg->bg_itable_unused_lo) |
243 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
244 (__u32)le16_to_cpu(bg->bg_itable_unused_hi) << 16 : 0);
247 void ext4_block_bitmap_set(struct super_block *sb,
248 struct ext4_group_desc *bg, ext4_fsblk_t blk)
250 bg->bg_block_bitmap_lo = cpu_to_le32((u32)blk);
251 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
252 bg->bg_block_bitmap_hi = cpu_to_le32(blk >> 32);
255 void ext4_inode_bitmap_set(struct super_block *sb,
256 struct ext4_group_desc *bg, ext4_fsblk_t blk)
258 bg->bg_inode_bitmap_lo = cpu_to_le32((u32)blk);
259 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
260 bg->bg_inode_bitmap_hi = cpu_to_le32(blk >> 32);
263 void ext4_inode_table_set(struct super_block *sb,
264 struct ext4_group_desc *bg, ext4_fsblk_t blk)
266 bg->bg_inode_table_lo = cpu_to_le32((u32)blk);
267 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
268 bg->bg_inode_table_hi = cpu_to_le32(blk >> 32);
271 void ext4_free_group_clusters_set(struct super_block *sb,
272 struct ext4_group_desc *bg, __u32 count)
274 bg->bg_free_blocks_count_lo = cpu_to_le16((__u16)count);
275 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
276 bg->bg_free_blocks_count_hi = cpu_to_le16(count >> 16);
279 void ext4_free_inodes_set(struct super_block *sb,
280 struct ext4_group_desc *bg, __u32 count)
282 bg->bg_free_inodes_count_lo = cpu_to_le16((__u16)count);
283 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
284 bg->bg_free_inodes_count_hi = cpu_to_le16(count >> 16);
287 void ext4_used_dirs_set(struct super_block *sb,
288 struct ext4_group_desc *bg, __u32 count)
290 bg->bg_used_dirs_count_lo = cpu_to_le16((__u16)count);
291 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
292 bg->bg_used_dirs_count_hi = cpu_to_le16(count >> 16);
295 void ext4_itable_unused_set(struct super_block *sb,
296 struct ext4_group_desc *bg, __u32 count)
298 bg->bg_itable_unused_lo = cpu_to_le16((__u16)count);
299 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
300 bg->bg_itable_unused_hi = cpu_to_le16(count >> 16);
304 static void __save_error_info(struct super_block *sb, const char *func,
307 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
309 EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS;
310 es->s_state |= cpu_to_le16(EXT4_ERROR_FS);
311 es->s_last_error_time = cpu_to_le32(get_seconds());
312 strncpy(es->s_last_error_func, func, sizeof(es->s_last_error_func));
313 es->s_last_error_line = cpu_to_le32(line);
314 if (!es->s_first_error_time) {
315 es->s_first_error_time = es->s_last_error_time;
316 strncpy(es->s_first_error_func, func,
317 sizeof(es->s_first_error_func));
318 es->s_first_error_line = cpu_to_le32(line);
319 es->s_first_error_ino = es->s_last_error_ino;
320 es->s_first_error_block = es->s_last_error_block;
323 * Start the daily error reporting function if it hasn't been
326 if (!es->s_error_count)
327 mod_timer(&EXT4_SB(sb)->s_err_report, jiffies + 24*60*60*HZ);
328 le32_add_cpu(&es->s_error_count, 1);
331 static void save_error_info(struct super_block *sb, const char *func,
334 __save_error_info(sb, func, line);
335 ext4_commit_super(sb, 1);
339 * The del_gendisk() function uninitializes the disk-specific data
340 * structures, including the bdi structure, without telling anyone
341 * else. Once this happens, any attempt to call mark_buffer_dirty()
342 * (for example, by ext4_commit_super), will cause a kernel OOPS.
343 * This is a kludge to prevent these oops until we can put in a proper
344 * hook in del_gendisk() to inform the VFS and file system layers.
346 static int block_device_ejected(struct super_block *sb)
348 struct inode *bd_inode = sb->s_bdev->bd_inode;
349 struct backing_dev_info *bdi = bd_inode->i_mapping->backing_dev_info;
351 return bdi->dev == NULL;
354 static void ext4_journal_commit_callback(journal_t *journal, transaction_t *txn)
356 struct super_block *sb = journal->j_private;
357 struct ext4_sb_info *sbi = EXT4_SB(sb);
358 int error = is_journal_aborted(journal);
359 struct ext4_journal_cb_entry *jce;
361 BUG_ON(txn->t_state == T_FINISHED);
362 spin_lock(&sbi->s_md_lock);
363 while (!list_empty(&txn->t_private_list)) {
364 jce = list_entry(txn->t_private_list.next,
365 struct ext4_journal_cb_entry, jce_list);
366 list_del_init(&jce->jce_list);
367 spin_unlock(&sbi->s_md_lock);
368 jce->jce_func(sb, jce, error);
369 spin_lock(&sbi->s_md_lock);
371 spin_unlock(&sbi->s_md_lock);
374 /* Deal with the reporting of failure conditions on a filesystem such as
375 * inconsistencies detected or read IO failures.
377 * On ext2, we can store the error state of the filesystem in the
378 * superblock. That is not possible on ext4, because we may have other
379 * write ordering constraints on the superblock which prevent us from
380 * writing it out straight away; and given that the journal is about to
381 * be aborted, we can't rely on the current, or future, transactions to
382 * write out the superblock safely.
384 * We'll just use the jbd2_journal_abort() error code to record an error in
385 * the journal instead. On recovery, the journal will complain about
386 * that error until we've noted it down and cleared it.
389 static void ext4_handle_error(struct super_block *sb)
391 if (sb->s_flags & MS_RDONLY)
394 if (!test_opt(sb, ERRORS_CONT)) {
395 journal_t *journal = EXT4_SB(sb)->s_journal;
397 EXT4_SB(sb)->s_mount_flags |= EXT4_MF_FS_ABORTED;
399 jbd2_journal_abort(journal, -EIO);
401 if (test_opt(sb, ERRORS_RO)) {
402 ext4_msg(sb, KERN_CRIT, "Remounting filesystem read-only");
404 * Make sure updated value of ->s_mount_flags will be visible
405 * before ->s_flags update
408 sb->s_flags |= MS_RDONLY;
410 if (test_opt(sb, ERRORS_PANIC))
411 panic("EXT4-fs (device %s): panic forced after error\n",
415 #define ext4_error_ratelimit(sb) \
416 ___ratelimit(&(EXT4_SB(sb)->s_err_ratelimit_state), \
419 void __ext4_error(struct super_block *sb, const char *function,
420 unsigned int line, const char *fmt, ...)
422 struct va_format vaf;
425 if (ext4_error_ratelimit(sb)) {
430 "EXT4-fs error (device %s): %s:%d: comm %s: %pV\n",
431 sb->s_id, function, line, current->comm, &vaf);
434 save_error_info(sb, function, line);
435 ext4_handle_error(sb);
438 void __ext4_error_inode(struct inode *inode, const char *function,
439 unsigned int line, ext4_fsblk_t block,
440 const char *fmt, ...)
443 struct va_format vaf;
444 struct ext4_super_block *es = EXT4_SB(inode->i_sb)->s_es;
446 es->s_last_error_ino = cpu_to_le32(inode->i_ino);
447 es->s_last_error_block = cpu_to_le64(block);
448 if (ext4_error_ratelimit(inode->i_sb)) {
453 printk(KERN_CRIT "EXT4-fs error (device %s): %s:%d: "
454 "inode #%lu: block %llu: comm %s: %pV\n",
455 inode->i_sb->s_id, function, line, inode->i_ino,
456 block, current->comm, &vaf);
458 printk(KERN_CRIT "EXT4-fs error (device %s): %s:%d: "
459 "inode #%lu: comm %s: %pV\n",
460 inode->i_sb->s_id, function, line, inode->i_ino,
461 current->comm, &vaf);
464 save_error_info(inode->i_sb, function, line);
465 ext4_handle_error(inode->i_sb);
468 void __ext4_error_file(struct file *file, const char *function,
469 unsigned int line, ext4_fsblk_t block,
470 const char *fmt, ...)
473 struct va_format vaf;
474 struct ext4_super_block *es;
475 struct inode *inode = file_inode(file);
476 char pathname[80], *path;
478 es = EXT4_SB(inode->i_sb)->s_es;
479 es->s_last_error_ino = cpu_to_le32(inode->i_ino);
480 if (ext4_error_ratelimit(inode->i_sb)) {
481 path = d_path(&(file->f_path), pathname, sizeof(pathname));
489 "EXT4-fs error (device %s): %s:%d: inode #%lu: "
490 "block %llu: comm %s: path %s: %pV\n",
491 inode->i_sb->s_id, function, line, inode->i_ino,
492 block, current->comm, path, &vaf);
495 "EXT4-fs error (device %s): %s:%d: inode #%lu: "
496 "comm %s: path %s: %pV\n",
497 inode->i_sb->s_id, function, line, inode->i_ino,
498 current->comm, path, &vaf);
501 save_error_info(inode->i_sb, function, line);
502 ext4_handle_error(inode->i_sb);
505 const char *ext4_decode_error(struct super_block *sb, int errno,
512 errstr = "IO failure";
515 errstr = "Out of memory";
518 if (!sb || (EXT4_SB(sb)->s_journal &&
519 EXT4_SB(sb)->s_journal->j_flags & JBD2_ABORT))
520 errstr = "Journal has aborted";
522 errstr = "Readonly filesystem";
525 /* If the caller passed in an extra buffer for unknown
526 * errors, textualise them now. Else we just return
529 /* Check for truncated error codes... */
530 if (snprintf(nbuf, 16, "error %d", -errno) >= 0)
539 /* __ext4_std_error decodes expected errors from journaling functions
540 * automatically and invokes the appropriate error response. */
542 void __ext4_std_error(struct super_block *sb, const char *function,
543 unsigned int line, int errno)
548 /* Special case: if the error is EROFS, and we're not already
549 * inside a transaction, then there's really no point in logging
551 if (errno == -EROFS && journal_current_handle() == NULL &&
552 (sb->s_flags & MS_RDONLY))
555 if (ext4_error_ratelimit(sb)) {
556 errstr = ext4_decode_error(sb, errno, nbuf);
557 printk(KERN_CRIT "EXT4-fs error (device %s) in %s:%d: %s\n",
558 sb->s_id, function, line, errstr);
561 save_error_info(sb, function, line);
562 ext4_handle_error(sb);
566 * ext4_abort is a much stronger failure handler than ext4_error. The
567 * abort function may be used to deal with unrecoverable failures such
568 * as journal IO errors or ENOMEM at a critical moment in log management.
570 * We unconditionally force the filesystem into an ABORT|READONLY state,
571 * unless the error response on the fs has been set to panic in which
572 * case we take the easy way out and panic immediately.
575 void __ext4_abort(struct super_block *sb, const char *function,
576 unsigned int line, const char *fmt, ...)
580 save_error_info(sb, function, line);
582 printk(KERN_CRIT "EXT4-fs error (device %s): %s:%d: ", sb->s_id,
588 if ((sb->s_flags & MS_RDONLY) == 0) {
589 ext4_msg(sb, KERN_CRIT, "Remounting filesystem read-only");
590 EXT4_SB(sb)->s_mount_flags |= EXT4_MF_FS_ABORTED;
592 * Make sure updated value of ->s_mount_flags will be visible
593 * before ->s_flags update
596 sb->s_flags |= MS_RDONLY;
597 if (EXT4_SB(sb)->s_journal)
598 jbd2_journal_abort(EXT4_SB(sb)->s_journal, -EIO);
599 save_error_info(sb, function, line);
601 if (test_opt(sb, ERRORS_PANIC))
602 panic("EXT4-fs panic from previous error\n");
605 void __ext4_msg(struct super_block *sb,
606 const char *prefix, const char *fmt, ...)
608 struct va_format vaf;
611 if (!___ratelimit(&(EXT4_SB(sb)->s_msg_ratelimit_state), "EXT4-fs"))
617 printk("%sEXT4-fs (%s): %pV\n", prefix, sb->s_id, &vaf);
621 void __ext4_warning(struct super_block *sb, const char *function,
622 unsigned int line, const char *fmt, ...)
624 struct va_format vaf;
627 if (!___ratelimit(&(EXT4_SB(sb)->s_warning_ratelimit_state),
634 printk(KERN_WARNING "EXT4-fs warning (device %s): %s:%d: %pV\n",
635 sb->s_id, function, line, &vaf);
639 void __ext4_grp_locked_error(const char *function, unsigned int line,
640 struct super_block *sb, ext4_group_t grp,
641 unsigned long ino, ext4_fsblk_t block,
642 const char *fmt, ...)
646 struct va_format vaf;
648 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
650 es->s_last_error_ino = cpu_to_le32(ino);
651 es->s_last_error_block = cpu_to_le64(block);
652 __save_error_info(sb, function, line);
654 if (ext4_error_ratelimit(sb)) {
658 printk(KERN_CRIT "EXT4-fs error (device %s): %s:%d: group %u, ",
659 sb->s_id, function, line, grp);
661 printk(KERN_CONT "inode %lu: ", ino);
663 printk(KERN_CONT "block %llu:",
664 (unsigned long long) block);
665 printk(KERN_CONT "%pV\n", &vaf);
669 if (test_opt(sb, ERRORS_CONT)) {
670 ext4_commit_super(sb, 0);
674 ext4_unlock_group(sb, grp);
675 ext4_handle_error(sb);
677 * We only get here in the ERRORS_RO case; relocking the group
678 * may be dangerous, but nothing bad will happen since the
679 * filesystem will have already been marked read/only and the
680 * journal has been aborted. We return 1 as a hint to callers
681 * who might what to use the return value from
682 * ext4_grp_locked_error() to distinguish between the
683 * ERRORS_CONT and ERRORS_RO case, and perhaps return more
684 * aggressively from the ext4 function in question, with a
685 * more appropriate error code.
687 ext4_lock_group(sb, grp);
691 void ext4_update_dynamic_rev(struct super_block *sb)
693 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
695 if (le32_to_cpu(es->s_rev_level) > EXT4_GOOD_OLD_REV)
699 "updating to rev %d because of new feature flag, "
700 "running e2fsck is recommended",
703 es->s_first_ino = cpu_to_le32(EXT4_GOOD_OLD_FIRST_INO);
704 es->s_inode_size = cpu_to_le16(EXT4_GOOD_OLD_INODE_SIZE);
705 es->s_rev_level = cpu_to_le32(EXT4_DYNAMIC_REV);
706 /* leave es->s_feature_*compat flags alone */
707 /* es->s_uuid will be set by e2fsck if empty */
710 * The rest of the superblock fields should be zero, and if not it
711 * means they are likely already in use, so leave them alone. We
712 * can leave it up to e2fsck to clean up any inconsistencies there.
717 * Open the external journal device
719 static struct block_device *ext4_blkdev_get(dev_t dev, struct super_block *sb)
721 struct block_device *bdev;
722 char b[BDEVNAME_SIZE];
724 bdev = blkdev_get_by_dev(dev, FMODE_READ|FMODE_WRITE|FMODE_EXCL, sb);
730 ext4_msg(sb, KERN_ERR, "failed to open journal device %s: %ld",
731 __bdevname(dev, b), PTR_ERR(bdev));
736 * Release the journal device
738 static void ext4_blkdev_put(struct block_device *bdev)
740 blkdev_put(bdev, FMODE_READ|FMODE_WRITE|FMODE_EXCL);
743 static void ext4_blkdev_remove(struct ext4_sb_info *sbi)
745 struct block_device *bdev;
746 bdev = sbi->journal_bdev;
748 ext4_blkdev_put(bdev);
749 sbi->journal_bdev = NULL;
753 static inline struct inode *orphan_list_entry(struct list_head *l)
755 return &list_entry(l, struct ext4_inode_info, i_orphan)->vfs_inode;
758 static void dump_orphan_list(struct super_block *sb, struct ext4_sb_info *sbi)
762 ext4_msg(sb, KERN_ERR, "sb orphan head is %d",
763 le32_to_cpu(sbi->s_es->s_last_orphan));
765 printk(KERN_ERR "sb_info orphan list:\n");
766 list_for_each(l, &sbi->s_orphan) {
767 struct inode *inode = orphan_list_entry(l);
769 "inode %s:%lu at %p: mode %o, nlink %d, next %d\n",
770 inode->i_sb->s_id, inode->i_ino, inode,
771 inode->i_mode, inode->i_nlink,
776 static void ext4_put_super(struct super_block *sb)
778 struct ext4_sb_info *sbi = EXT4_SB(sb);
779 struct ext4_super_block *es = sbi->s_es;
782 ext4_unregister_li_request(sb);
783 dquot_disable(sb, -1, DQUOT_USAGE_ENABLED | DQUOT_LIMITS_ENABLED);
785 flush_workqueue(sbi->rsv_conversion_wq);
786 destroy_workqueue(sbi->rsv_conversion_wq);
788 if (sbi->s_journal) {
789 err = jbd2_journal_destroy(sbi->s_journal);
790 sbi->s_journal = NULL;
792 ext4_abort(sb, "Couldn't clean up the journal");
795 ext4_es_unregister_shrinker(sbi);
796 del_timer_sync(&sbi->s_err_report);
797 ext4_release_system_zone(sb);
799 ext4_ext_release(sb);
800 ext4_xattr_put_super(sb);
802 if (!(sb->s_flags & MS_RDONLY)) {
803 EXT4_CLEAR_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
804 es->s_state = cpu_to_le16(sbi->s_mount_state);
806 if (!(sb->s_flags & MS_RDONLY))
807 ext4_commit_super(sb, 1);
810 remove_proc_entry("options", sbi->s_proc);
811 remove_proc_entry(sb->s_id, ext4_proc_root);
813 kobject_del(&sbi->s_kobj);
815 for (i = 0; i < sbi->s_gdb_count; i++)
816 brelse(sbi->s_group_desc[i]);
817 ext4_kvfree(sbi->s_group_desc);
818 ext4_kvfree(sbi->s_flex_groups);
819 percpu_counter_destroy(&sbi->s_freeclusters_counter);
820 percpu_counter_destroy(&sbi->s_freeinodes_counter);
821 percpu_counter_destroy(&sbi->s_dirs_counter);
822 percpu_counter_destroy(&sbi->s_dirtyclusters_counter);
823 percpu_counter_destroy(&sbi->s_extent_cache_cnt);
826 for (i = 0; i < MAXQUOTAS; i++)
827 kfree(sbi->s_qf_names[i]);
830 /* Debugging code just in case the in-memory inode orphan list
831 * isn't empty. The on-disk one can be non-empty if we've
832 * detected an error and taken the fs readonly, but the
833 * in-memory list had better be clean by this point. */
834 if (!list_empty(&sbi->s_orphan))
835 dump_orphan_list(sb, sbi);
836 J_ASSERT(list_empty(&sbi->s_orphan));
838 invalidate_bdev(sb->s_bdev);
839 if (sbi->journal_bdev && sbi->journal_bdev != sb->s_bdev) {
841 * Invalidate the journal device's buffers. We don't want them
842 * floating about in memory - the physical journal device may
843 * hotswapped, and it breaks the `ro-after' testing code.
845 sync_blockdev(sbi->journal_bdev);
846 invalidate_bdev(sbi->journal_bdev);
847 ext4_blkdev_remove(sbi);
849 if (sbi->s_mb_cache) {
850 ext4_xattr_destroy_cache(sbi->s_mb_cache);
851 sbi->s_mb_cache = NULL;
854 kthread_stop(sbi->s_mmp_tsk);
855 sb->s_fs_info = NULL;
857 * Now that we are completely done shutting down the
858 * superblock, we need to actually destroy the kobject.
860 kobject_put(&sbi->s_kobj);
861 wait_for_completion(&sbi->s_kobj_unregister);
862 if (sbi->s_chksum_driver)
863 crypto_free_shash(sbi->s_chksum_driver);
864 kfree(sbi->s_blockgroup_lock);
868 static struct kmem_cache *ext4_inode_cachep;
871 * Called inside transaction, so use GFP_NOFS
873 static struct inode *ext4_alloc_inode(struct super_block *sb)
875 struct ext4_inode_info *ei;
877 ei = kmem_cache_alloc(ext4_inode_cachep, GFP_NOFS);
881 ei->vfs_inode.i_version = 1;
882 spin_lock_init(&ei->i_raw_lock);
883 INIT_LIST_HEAD(&ei->i_prealloc_list);
884 spin_lock_init(&ei->i_prealloc_lock);
885 ext4_es_init_tree(&ei->i_es_tree);
886 rwlock_init(&ei->i_es_lock);
887 INIT_LIST_HEAD(&ei->i_es_lru);
889 ei->i_touch_when = 0;
890 ei->i_reserved_data_blocks = 0;
891 ei->i_reserved_meta_blocks = 0;
892 ei->i_allocated_meta_blocks = 0;
893 ei->i_da_metadata_calc_len = 0;
894 ei->i_da_metadata_calc_last_lblock = 0;
895 spin_lock_init(&(ei->i_block_reservation_lock));
897 ei->i_reserved_quota = 0;
900 INIT_LIST_HEAD(&ei->i_rsv_conversion_list);
901 spin_lock_init(&ei->i_completed_io_lock);
903 ei->i_datasync_tid = 0;
904 atomic_set(&ei->i_ioend_count, 0);
905 atomic_set(&ei->i_unwritten, 0);
906 INIT_WORK(&ei->i_rsv_conversion_work, ext4_end_io_rsv_work);
908 return &ei->vfs_inode;
911 static int ext4_drop_inode(struct inode *inode)
913 int drop = generic_drop_inode(inode);
915 trace_ext4_drop_inode(inode, drop);
919 static void ext4_i_callback(struct rcu_head *head)
921 struct inode *inode = container_of(head, struct inode, i_rcu);
922 kmem_cache_free(ext4_inode_cachep, EXT4_I(inode));
925 static void ext4_destroy_inode(struct inode *inode)
927 if (!list_empty(&(EXT4_I(inode)->i_orphan))) {
928 ext4_msg(inode->i_sb, KERN_ERR,
929 "Inode %lu (%p): orphan list check failed!",
930 inode->i_ino, EXT4_I(inode));
931 print_hex_dump(KERN_INFO, "", DUMP_PREFIX_ADDRESS, 16, 4,
932 EXT4_I(inode), sizeof(struct ext4_inode_info),
936 call_rcu(&inode->i_rcu, ext4_i_callback);
939 static void init_once(void *foo)
941 struct ext4_inode_info *ei = (struct ext4_inode_info *) foo;
943 INIT_LIST_HEAD(&ei->i_orphan);
944 init_rwsem(&ei->xattr_sem);
945 init_rwsem(&ei->i_data_sem);
946 inode_init_once(&ei->vfs_inode);
949 static int __init init_inodecache(void)
951 ext4_inode_cachep = kmem_cache_create("ext4_inode_cache",
952 sizeof(struct ext4_inode_info),
953 0, (SLAB_RECLAIM_ACCOUNT|
956 if (ext4_inode_cachep == NULL)
961 static void destroy_inodecache(void)
964 * Make sure all delayed rcu free inodes are flushed before we
968 kmem_cache_destroy(ext4_inode_cachep);
971 void ext4_clear_inode(struct inode *inode)
973 invalidate_inode_buffers(inode);
976 ext4_discard_preallocations(inode);
977 ext4_es_remove_extent(inode, 0, EXT_MAX_BLOCKS);
978 ext4_es_lru_del(inode);
979 if (EXT4_I(inode)->jinode) {
980 jbd2_journal_release_jbd_inode(EXT4_JOURNAL(inode),
981 EXT4_I(inode)->jinode);
982 jbd2_free_inode(EXT4_I(inode)->jinode);
983 EXT4_I(inode)->jinode = NULL;
987 static struct inode *ext4_nfs_get_inode(struct super_block *sb,
988 u64 ino, u32 generation)
992 if (ino < EXT4_FIRST_INO(sb) && ino != EXT4_ROOT_INO)
993 return ERR_PTR(-ESTALE);
994 if (ino > le32_to_cpu(EXT4_SB(sb)->s_es->s_inodes_count))
995 return ERR_PTR(-ESTALE);
997 /* iget isn't really right if the inode is currently unallocated!!
999 * ext4_read_inode will return a bad_inode if the inode had been
1000 * deleted, so we should be safe.
1002 * Currently we don't know the generation for parent directory, so
1003 * a generation of 0 means "accept any"
1005 inode = ext4_iget(sb, ino);
1007 return ERR_CAST(inode);
1008 if (generation && inode->i_generation != generation) {
1010 return ERR_PTR(-ESTALE);
1016 static struct dentry *ext4_fh_to_dentry(struct super_block *sb, struct fid *fid,
1017 int fh_len, int fh_type)
1019 return generic_fh_to_dentry(sb, fid, fh_len, fh_type,
1020 ext4_nfs_get_inode);
1023 static struct dentry *ext4_fh_to_parent(struct super_block *sb, struct fid *fid,
1024 int fh_len, int fh_type)
1026 return generic_fh_to_parent(sb, fid, fh_len, fh_type,
1027 ext4_nfs_get_inode);
1031 * Try to release metadata pages (indirect blocks, directories) which are
1032 * mapped via the block device. Since these pages could have journal heads
1033 * which would prevent try_to_free_buffers() from freeing them, we must use
1034 * jbd2 layer's try_to_free_buffers() function to release them.
1036 static int bdev_try_to_free_page(struct super_block *sb, struct page *page,
1039 journal_t *journal = EXT4_SB(sb)->s_journal;
1041 WARN_ON(PageChecked(page));
1042 if (!page_has_buffers(page))
1045 return jbd2_journal_try_to_free_buffers(journal, page,
1046 wait & ~__GFP_WAIT);
1047 return try_to_free_buffers(page);
1051 #define QTYPE2NAME(t) ((t) == USRQUOTA ? "user" : "group")
1052 #define QTYPE2MOPT(on, t) ((t) == USRQUOTA?((on)##USRJQUOTA):((on)##GRPJQUOTA))
1054 static int ext4_write_dquot(struct dquot *dquot);
1055 static int ext4_acquire_dquot(struct dquot *dquot);
1056 static int ext4_release_dquot(struct dquot *dquot);
1057 static int ext4_mark_dquot_dirty(struct dquot *dquot);
1058 static int ext4_write_info(struct super_block *sb, int type);
1059 static int ext4_quota_on(struct super_block *sb, int type, int format_id,
1061 static int ext4_quota_on_sysfile(struct super_block *sb, int type,
1063 static int ext4_quota_off(struct super_block *sb, int type);
1064 static int ext4_quota_off_sysfile(struct super_block *sb, int type);
1065 static int ext4_quota_on_mount(struct super_block *sb, int type);
1066 static ssize_t ext4_quota_read(struct super_block *sb, int type, char *data,
1067 size_t len, loff_t off);
1068 static ssize_t ext4_quota_write(struct super_block *sb, int type,
1069 const char *data, size_t len, loff_t off);
1070 static int ext4_quota_enable(struct super_block *sb, int type, int format_id,
1071 unsigned int flags);
1072 static int ext4_enable_quotas(struct super_block *sb);
1074 static const struct dquot_operations ext4_quota_operations = {
1075 .get_reserved_space = ext4_get_reserved_space,
1076 .write_dquot = ext4_write_dquot,
1077 .acquire_dquot = ext4_acquire_dquot,
1078 .release_dquot = ext4_release_dquot,
1079 .mark_dirty = ext4_mark_dquot_dirty,
1080 .write_info = ext4_write_info,
1081 .alloc_dquot = dquot_alloc,
1082 .destroy_dquot = dquot_destroy,
1085 static const struct quotactl_ops ext4_qctl_operations = {
1086 .quota_on = ext4_quota_on,
1087 .quota_off = ext4_quota_off,
1088 .quota_sync = dquot_quota_sync,
1089 .get_info = dquot_get_dqinfo,
1090 .set_info = dquot_set_dqinfo,
1091 .get_dqblk = dquot_get_dqblk,
1092 .set_dqblk = dquot_set_dqblk
1095 static const struct quotactl_ops ext4_qctl_sysfile_operations = {
1096 .quota_on_meta = ext4_quota_on_sysfile,
1097 .quota_off = ext4_quota_off_sysfile,
1098 .quota_sync = dquot_quota_sync,
1099 .get_info = dquot_get_dqinfo,
1100 .set_info = dquot_set_dqinfo,
1101 .get_dqblk = dquot_get_dqblk,
1102 .set_dqblk = dquot_set_dqblk
1106 static const struct super_operations ext4_sops = {
1107 .alloc_inode = ext4_alloc_inode,
1108 .destroy_inode = ext4_destroy_inode,
1109 .write_inode = ext4_write_inode,
1110 .dirty_inode = ext4_dirty_inode,
1111 .drop_inode = ext4_drop_inode,
1112 .evict_inode = ext4_evict_inode,
1113 .put_super = ext4_put_super,
1114 .sync_fs = ext4_sync_fs,
1115 .freeze_fs = ext4_freeze,
1116 .unfreeze_fs = ext4_unfreeze,
1117 .statfs = ext4_statfs,
1118 .remount_fs = ext4_remount,
1119 .show_options = ext4_show_options,
1121 .quota_read = ext4_quota_read,
1122 .quota_write = ext4_quota_write,
1124 .bdev_try_to_free_page = bdev_try_to_free_page,
1127 static const struct super_operations ext4_nojournal_sops = {
1128 .alloc_inode = ext4_alloc_inode,
1129 .destroy_inode = ext4_destroy_inode,
1130 .write_inode = ext4_write_inode,
1131 .dirty_inode = ext4_dirty_inode,
1132 .drop_inode = ext4_drop_inode,
1133 .evict_inode = ext4_evict_inode,
1134 .sync_fs = ext4_sync_fs_nojournal,
1135 .put_super = ext4_put_super,
1136 .statfs = ext4_statfs,
1137 .remount_fs = ext4_remount,
1138 .show_options = ext4_show_options,
1140 .quota_read = ext4_quota_read,
1141 .quota_write = ext4_quota_write,
1143 .bdev_try_to_free_page = bdev_try_to_free_page,
1146 static const struct export_operations ext4_export_ops = {
1147 .fh_to_dentry = ext4_fh_to_dentry,
1148 .fh_to_parent = ext4_fh_to_parent,
1149 .get_parent = ext4_get_parent,
1153 Opt_bsd_df, Opt_minix_df, Opt_grpid, Opt_nogrpid,
1154 Opt_resgid, Opt_resuid, Opt_sb, Opt_err_cont, Opt_err_panic, Opt_err_ro,
1155 Opt_nouid32, Opt_debug, Opt_removed,
1156 Opt_user_xattr, Opt_nouser_xattr, Opt_acl, Opt_noacl,
1157 Opt_auto_da_alloc, Opt_noauto_da_alloc, Opt_noload,
1158 Opt_commit, Opt_min_batch_time, Opt_max_batch_time, Opt_journal_dev,
1159 Opt_journal_path, Opt_journal_checksum, Opt_journal_async_commit,
1160 Opt_abort, Opt_data_journal, Opt_data_ordered, Opt_data_writeback,
1161 Opt_data_err_abort, Opt_data_err_ignore,
1162 Opt_usrjquota, Opt_grpjquota, Opt_offusrjquota, Opt_offgrpjquota,
1163 Opt_jqfmt_vfsold, Opt_jqfmt_vfsv0, Opt_jqfmt_vfsv1, Opt_quota,
1164 Opt_noquota, Opt_barrier, Opt_nobarrier, Opt_err,
1165 Opt_usrquota, Opt_grpquota, Opt_i_version,
1166 Opt_stripe, Opt_delalloc, Opt_nodelalloc, Opt_mblk_io_submit,
1167 Opt_nomblk_io_submit, Opt_block_validity, Opt_noblock_validity,
1168 Opt_inode_readahead_blks, Opt_journal_ioprio,
1169 Opt_dioread_nolock, Opt_dioread_lock,
1170 Opt_discard, Opt_nodiscard, Opt_init_itable, Opt_noinit_itable,
1171 Opt_max_dir_size_kb,
1174 static const match_table_t tokens = {
1175 {Opt_bsd_df, "bsddf"},
1176 {Opt_minix_df, "minixdf"},
1177 {Opt_grpid, "grpid"},
1178 {Opt_grpid, "bsdgroups"},
1179 {Opt_nogrpid, "nogrpid"},
1180 {Opt_nogrpid, "sysvgroups"},
1181 {Opt_resgid, "resgid=%u"},
1182 {Opt_resuid, "resuid=%u"},
1184 {Opt_err_cont, "errors=continue"},
1185 {Opt_err_panic, "errors=panic"},
1186 {Opt_err_ro, "errors=remount-ro"},
1187 {Opt_nouid32, "nouid32"},
1188 {Opt_debug, "debug"},
1189 {Opt_removed, "oldalloc"},
1190 {Opt_removed, "orlov"},
1191 {Opt_user_xattr, "user_xattr"},
1192 {Opt_nouser_xattr, "nouser_xattr"},
1194 {Opt_noacl, "noacl"},
1195 {Opt_noload, "norecovery"},
1196 {Opt_noload, "noload"},
1197 {Opt_removed, "nobh"},
1198 {Opt_removed, "bh"},
1199 {Opt_commit, "commit=%u"},
1200 {Opt_min_batch_time, "min_batch_time=%u"},
1201 {Opt_max_batch_time, "max_batch_time=%u"},
1202 {Opt_journal_dev, "journal_dev=%u"},
1203 {Opt_journal_path, "journal_path=%s"},
1204 {Opt_journal_checksum, "journal_checksum"},
1205 {Opt_journal_async_commit, "journal_async_commit"},
1206 {Opt_abort, "abort"},
1207 {Opt_data_journal, "data=journal"},
1208 {Opt_data_ordered, "data=ordered"},
1209 {Opt_data_writeback, "data=writeback"},
1210 {Opt_data_err_abort, "data_err=abort"},
1211 {Opt_data_err_ignore, "data_err=ignore"},
1212 {Opt_offusrjquota, "usrjquota="},
1213 {Opt_usrjquota, "usrjquota=%s"},
1214 {Opt_offgrpjquota, "grpjquota="},
1215 {Opt_grpjquota, "grpjquota=%s"},
1216 {Opt_jqfmt_vfsold, "jqfmt=vfsold"},
1217 {Opt_jqfmt_vfsv0, "jqfmt=vfsv0"},
1218 {Opt_jqfmt_vfsv1, "jqfmt=vfsv1"},
1219 {Opt_grpquota, "grpquota"},
1220 {Opt_noquota, "noquota"},
1221 {Opt_quota, "quota"},
1222 {Opt_usrquota, "usrquota"},
1223 {Opt_barrier, "barrier=%u"},
1224 {Opt_barrier, "barrier"},
1225 {Opt_nobarrier, "nobarrier"},
1226 {Opt_i_version, "i_version"},
1227 {Opt_stripe, "stripe=%u"},
1228 {Opt_delalloc, "delalloc"},
1229 {Opt_nodelalloc, "nodelalloc"},
1230 {Opt_removed, "mblk_io_submit"},
1231 {Opt_removed, "nomblk_io_submit"},
1232 {Opt_block_validity, "block_validity"},
1233 {Opt_noblock_validity, "noblock_validity"},
1234 {Opt_inode_readahead_blks, "inode_readahead_blks=%u"},
1235 {Opt_journal_ioprio, "journal_ioprio=%u"},
1236 {Opt_auto_da_alloc, "auto_da_alloc=%u"},
1237 {Opt_auto_da_alloc, "auto_da_alloc"},
1238 {Opt_noauto_da_alloc, "noauto_da_alloc"},
1239 {Opt_dioread_nolock, "dioread_nolock"},
1240 {Opt_dioread_lock, "dioread_lock"},
1241 {Opt_discard, "discard"},
1242 {Opt_nodiscard, "nodiscard"},
1243 {Opt_init_itable, "init_itable=%u"},
1244 {Opt_init_itable, "init_itable"},
1245 {Opt_noinit_itable, "noinit_itable"},
1246 {Opt_max_dir_size_kb, "max_dir_size_kb=%u"},
1247 {Opt_removed, "check=none"}, /* mount option from ext2/3 */
1248 {Opt_removed, "nocheck"}, /* mount option from ext2/3 */
1249 {Opt_removed, "reservation"}, /* mount option from ext2/3 */
1250 {Opt_removed, "noreservation"}, /* mount option from ext2/3 */
1251 {Opt_removed, "journal=%u"}, /* mount option from ext2/3 */
1255 static ext4_fsblk_t get_sb_block(void **data)
1257 ext4_fsblk_t sb_block;
1258 char *options = (char *) *data;
1260 if (!options || strncmp(options, "sb=", 3) != 0)
1261 return 1; /* Default location */
1264 /* TODO: use simple_strtoll with >32bit ext4 */
1265 sb_block = simple_strtoul(options, &options, 0);
1266 if (*options && *options != ',') {
1267 printk(KERN_ERR "EXT4-fs: Invalid sb specification: %s\n",
1271 if (*options == ',')
1273 *data = (void *) options;
1278 #define DEFAULT_JOURNAL_IOPRIO (IOPRIO_PRIO_VALUE(IOPRIO_CLASS_BE, 3))
1279 static char deprecated_msg[] = "Mount option \"%s\" will be removed by %s\n"
1280 "Contact linux-ext4@vger.kernel.org if you think we should keep it.\n";
1283 static int set_qf_name(struct super_block *sb, int qtype, substring_t *args)
1285 struct ext4_sb_info *sbi = EXT4_SB(sb);
1289 if (sb_any_quota_loaded(sb) &&
1290 !sbi->s_qf_names[qtype]) {
1291 ext4_msg(sb, KERN_ERR,
1292 "Cannot change journaled "
1293 "quota options when quota turned on");
1296 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_QUOTA)) {
1297 ext4_msg(sb, KERN_ERR, "Cannot set journaled quota options "
1298 "when QUOTA feature is enabled");
1301 qname = match_strdup(args);
1303 ext4_msg(sb, KERN_ERR,
1304 "Not enough memory for storing quotafile name");
1307 if (sbi->s_qf_names[qtype]) {
1308 if (strcmp(sbi->s_qf_names[qtype], qname) == 0)
1311 ext4_msg(sb, KERN_ERR,
1312 "%s quota file already specified",
1316 if (strchr(qname, '/')) {
1317 ext4_msg(sb, KERN_ERR,
1318 "quotafile must be on filesystem root");
1321 sbi->s_qf_names[qtype] = qname;
1329 static int clear_qf_name(struct super_block *sb, int qtype)
1332 struct ext4_sb_info *sbi = EXT4_SB(sb);
1334 if (sb_any_quota_loaded(sb) &&
1335 sbi->s_qf_names[qtype]) {
1336 ext4_msg(sb, KERN_ERR, "Cannot change journaled quota options"
1337 " when quota turned on");
1340 kfree(sbi->s_qf_names[qtype]);
1341 sbi->s_qf_names[qtype] = NULL;
1346 #define MOPT_SET 0x0001
1347 #define MOPT_CLEAR 0x0002
1348 #define MOPT_NOSUPPORT 0x0004
1349 #define MOPT_EXPLICIT 0x0008
1350 #define MOPT_CLEAR_ERR 0x0010
1351 #define MOPT_GTE0 0x0020
1354 #define MOPT_QFMT 0x0040
1356 #define MOPT_Q MOPT_NOSUPPORT
1357 #define MOPT_QFMT MOPT_NOSUPPORT
1359 #define MOPT_DATAJ 0x0080
1360 #define MOPT_NO_EXT2 0x0100
1361 #define MOPT_NO_EXT3 0x0200
1362 #define MOPT_EXT4_ONLY (MOPT_NO_EXT2 | MOPT_NO_EXT3)
1363 #define MOPT_STRING 0x0400
1365 static const struct mount_opts {
1369 } ext4_mount_opts[] = {
1370 {Opt_minix_df, EXT4_MOUNT_MINIX_DF, MOPT_SET},
1371 {Opt_bsd_df, EXT4_MOUNT_MINIX_DF, MOPT_CLEAR},
1372 {Opt_grpid, EXT4_MOUNT_GRPID, MOPT_SET},
1373 {Opt_nogrpid, EXT4_MOUNT_GRPID, MOPT_CLEAR},
1374 {Opt_block_validity, EXT4_MOUNT_BLOCK_VALIDITY, MOPT_SET},
1375 {Opt_noblock_validity, EXT4_MOUNT_BLOCK_VALIDITY, MOPT_CLEAR},
1376 {Opt_dioread_nolock, EXT4_MOUNT_DIOREAD_NOLOCK,
1377 MOPT_EXT4_ONLY | MOPT_SET},
1378 {Opt_dioread_lock, EXT4_MOUNT_DIOREAD_NOLOCK,
1379 MOPT_EXT4_ONLY | MOPT_CLEAR},
1380 {Opt_discard, EXT4_MOUNT_DISCARD, MOPT_SET},
1381 {Opt_nodiscard, EXT4_MOUNT_DISCARD, MOPT_CLEAR},
1382 {Opt_delalloc, EXT4_MOUNT_DELALLOC,
1383 MOPT_EXT4_ONLY | MOPT_SET | MOPT_EXPLICIT},
1384 {Opt_nodelalloc, EXT4_MOUNT_DELALLOC,
1385 MOPT_EXT4_ONLY | MOPT_CLEAR},
1386 {Opt_journal_checksum, EXT4_MOUNT_JOURNAL_CHECKSUM,
1387 MOPT_EXT4_ONLY | MOPT_SET},
1388 {Opt_journal_async_commit, (EXT4_MOUNT_JOURNAL_ASYNC_COMMIT |
1389 EXT4_MOUNT_JOURNAL_CHECKSUM),
1390 MOPT_EXT4_ONLY | MOPT_SET},
1391 {Opt_noload, EXT4_MOUNT_NOLOAD, MOPT_NO_EXT2 | MOPT_SET},
1392 {Opt_err_panic, EXT4_MOUNT_ERRORS_PANIC, MOPT_SET | MOPT_CLEAR_ERR},
1393 {Opt_err_ro, EXT4_MOUNT_ERRORS_RO, MOPT_SET | MOPT_CLEAR_ERR},
1394 {Opt_err_cont, EXT4_MOUNT_ERRORS_CONT, MOPT_SET | MOPT_CLEAR_ERR},
1395 {Opt_data_err_abort, EXT4_MOUNT_DATA_ERR_ABORT,
1396 MOPT_NO_EXT2 | MOPT_SET},
1397 {Opt_data_err_ignore, EXT4_MOUNT_DATA_ERR_ABORT,
1398 MOPT_NO_EXT2 | MOPT_CLEAR},
1399 {Opt_barrier, EXT4_MOUNT_BARRIER, MOPT_SET},
1400 {Opt_nobarrier, EXT4_MOUNT_BARRIER, MOPT_CLEAR},
1401 {Opt_noauto_da_alloc, EXT4_MOUNT_NO_AUTO_DA_ALLOC, MOPT_SET},
1402 {Opt_auto_da_alloc, EXT4_MOUNT_NO_AUTO_DA_ALLOC, MOPT_CLEAR},
1403 {Opt_noinit_itable, EXT4_MOUNT_INIT_INODE_TABLE, MOPT_CLEAR},
1404 {Opt_commit, 0, MOPT_GTE0},
1405 {Opt_max_batch_time, 0, MOPT_GTE0},
1406 {Opt_min_batch_time, 0, MOPT_GTE0},
1407 {Opt_inode_readahead_blks, 0, MOPT_GTE0},
1408 {Opt_init_itable, 0, MOPT_GTE0},
1409 {Opt_stripe, 0, MOPT_GTE0},
1410 {Opt_resuid, 0, MOPT_GTE0},
1411 {Opt_resgid, 0, MOPT_GTE0},
1412 {Opt_journal_dev, 0, MOPT_GTE0},
1413 {Opt_journal_path, 0, MOPT_STRING},
1414 {Opt_journal_ioprio, 0, MOPT_GTE0},
1415 {Opt_data_journal, EXT4_MOUNT_JOURNAL_DATA, MOPT_NO_EXT2 | MOPT_DATAJ},
1416 {Opt_data_ordered, EXT4_MOUNT_ORDERED_DATA, MOPT_NO_EXT2 | MOPT_DATAJ},
1417 {Opt_data_writeback, EXT4_MOUNT_WRITEBACK_DATA,
1418 MOPT_NO_EXT2 | MOPT_DATAJ},
1419 {Opt_user_xattr, EXT4_MOUNT_XATTR_USER, MOPT_SET},
1420 {Opt_nouser_xattr, EXT4_MOUNT_XATTR_USER, MOPT_CLEAR},
1421 #ifdef CONFIG_EXT4_FS_POSIX_ACL
1422 {Opt_acl, EXT4_MOUNT_POSIX_ACL, MOPT_SET},
1423 {Opt_noacl, EXT4_MOUNT_POSIX_ACL, MOPT_CLEAR},
1425 {Opt_acl, 0, MOPT_NOSUPPORT},
1426 {Opt_noacl, 0, MOPT_NOSUPPORT},
1428 {Opt_nouid32, EXT4_MOUNT_NO_UID32, MOPT_SET},
1429 {Opt_debug, EXT4_MOUNT_DEBUG, MOPT_SET},
1430 {Opt_quota, EXT4_MOUNT_QUOTA | EXT4_MOUNT_USRQUOTA, MOPT_SET | MOPT_Q},
1431 {Opt_usrquota, EXT4_MOUNT_QUOTA | EXT4_MOUNT_USRQUOTA,
1433 {Opt_grpquota, EXT4_MOUNT_QUOTA | EXT4_MOUNT_GRPQUOTA,
1435 {Opt_noquota, (EXT4_MOUNT_QUOTA | EXT4_MOUNT_USRQUOTA |
1436 EXT4_MOUNT_GRPQUOTA), MOPT_CLEAR | MOPT_Q},
1437 {Opt_usrjquota, 0, MOPT_Q},
1438 {Opt_grpjquota, 0, MOPT_Q},
1439 {Opt_offusrjquota, 0, MOPT_Q},
1440 {Opt_offgrpjquota, 0, MOPT_Q},
1441 {Opt_jqfmt_vfsold, QFMT_VFS_OLD, MOPT_QFMT},
1442 {Opt_jqfmt_vfsv0, QFMT_VFS_V0, MOPT_QFMT},
1443 {Opt_jqfmt_vfsv1, QFMT_VFS_V1, MOPT_QFMT},
1444 {Opt_max_dir_size_kb, 0, MOPT_GTE0},
1448 static int handle_mount_opt(struct super_block *sb, char *opt, int token,
1449 substring_t *args, unsigned long *journal_devnum,
1450 unsigned int *journal_ioprio, int is_remount)
1452 struct ext4_sb_info *sbi = EXT4_SB(sb);
1453 const struct mount_opts *m;
1459 if (token == Opt_usrjquota)
1460 return set_qf_name(sb, USRQUOTA, &args[0]);
1461 else if (token == Opt_grpjquota)
1462 return set_qf_name(sb, GRPQUOTA, &args[0]);
1463 else if (token == Opt_offusrjquota)
1464 return clear_qf_name(sb, USRQUOTA);
1465 else if (token == Opt_offgrpjquota)
1466 return clear_qf_name(sb, GRPQUOTA);
1470 case Opt_nouser_xattr:
1471 ext4_msg(sb, KERN_WARNING, deprecated_msg, opt, "3.5");
1474 return 1; /* handled by get_sb_block() */
1476 ext4_msg(sb, KERN_WARNING, "Ignoring removed %s option", opt);
1479 sbi->s_mount_flags |= EXT4_MF_FS_ABORTED;
1482 sb->s_flags |= MS_I_VERSION;
1486 for (m = ext4_mount_opts; m->token != Opt_err; m++)
1487 if (token == m->token)
1490 if (m->token == Opt_err) {
1491 ext4_msg(sb, KERN_ERR, "Unrecognized mount option \"%s\" "
1492 "or missing value", opt);
1496 if ((m->flags & MOPT_NO_EXT2) && IS_EXT2_SB(sb)) {
1497 ext4_msg(sb, KERN_ERR,
1498 "Mount option \"%s\" incompatible with ext2", opt);
1501 if ((m->flags & MOPT_NO_EXT3) && IS_EXT3_SB(sb)) {
1502 ext4_msg(sb, KERN_ERR,
1503 "Mount option \"%s\" incompatible with ext3", opt);
1507 if (args->from && !(m->flags & MOPT_STRING) && match_int(args, &arg))
1509 if (args->from && (m->flags & MOPT_GTE0) && (arg < 0))
1511 if (m->flags & MOPT_EXPLICIT)
1512 set_opt2(sb, EXPLICIT_DELALLOC);
1513 if (m->flags & MOPT_CLEAR_ERR)
1514 clear_opt(sb, ERRORS_MASK);
1515 if (token == Opt_noquota && sb_any_quota_loaded(sb)) {
1516 ext4_msg(sb, KERN_ERR, "Cannot change quota "
1517 "options when quota turned on");
1521 if (m->flags & MOPT_NOSUPPORT) {
1522 ext4_msg(sb, KERN_ERR, "%s option not supported", opt);
1523 } else if (token == Opt_commit) {
1525 arg = JBD2_DEFAULT_MAX_COMMIT_AGE;
1526 sbi->s_commit_interval = HZ * arg;
1527 } else if (token == Opt_max_batch_time) {
1529 arg = EXT4_DEF_MAX_BATCH_TIME;
1530 sbi->s_max_batch_time = arg;
1531 } else if (token == Opt_min_batch_time) {
1532 sbi->s_min_batch_time = arg;
1533 } else if (token == Opt_inode_readahead_blks) {
1534 if (arg && (arg > (1 << 30) || !is_power_of_2(arg))) {
1535 ext4_msg(sb, KERN_ERR,
1536 "EXT4-fs: inode_readahead_blks must be "
1537 "0 or a power of 2 smaller than 2^31");
1540 sbi->s_inode_readahead_blks = arg;
1541 } else if (token == Opt_init_itable) {
1542 set_opt(sb, INIT_INODE_TABLE);
1544 arg = EXT4_DEF_LI_WAIT_MULT;
1545 sbi->s_li_wait_mult = arg;
1546 } else if (token == Opt_max_dir_size_kb) {
1547 sbi->s_max_dir_size_kb = arg;
1548 } else if (token == Opt_stripe) {
1549 sbi->s_stripe = arg;
1550 } else if (token == Opt_resuid) {
1551 uid = make_kuid(current_user_ns(), arg);
1552 if (!uid_valid(uid)) {
1553 ext4_msg(sb, KERN_ERR, "Invalid uid value %d", arg);
1556 sbi->s_resuid = uid;
1557 } else if (token == Opt_resgid) {
1558 gid = make_kgid(current_user_ns(), arg);
1559 if (!gid_valid(gid)) {
1560 ext4_msg(sb, KERN_ERR, "Invalid gid value %d", arg);
1563 sbi->s_resgid = gid;
1564 } else if (token == Opt_journal_dev) {
1566 ext4_msg(sb, KERN_ERR,
1567 "Cannot specify journal on remount");
1570 *journal_devnum = arg;
1571 } else if (token == Opt_journal_path) {
1573 struct inode *journal_inode;
1578 ext4_msg(sb, KERN_ERR,
1579 "Cannot specify journal on remount");
1582 journal_path = match_strdup(&args[0]);
1583 if (!journal_path) {
1584 ext4_msg(sb, KERN_ERR, "error: could not dup "
1585 "journal device string");
1589 error = kern_path(journal_path, LOOKUP_FOLLOW, &path);
1591 ext4_msg(sb, KERN_ERR, "error: could not find "
1592 "journal device path: error %d", error);
1593 kfree(journal_path);
1597 journal_inode = path.dentry->d_inode;
1598 if (!S_ISBLK(journal_inode->i_mode)) {
1599 ext4_msg(sb, KERN_ERR, "error: journal path %s "
1600 "is not a block device", journal_path);
1602 kfree(journal_path);
1606 *journal_devnum = new_encode_dev(journal_inode->i_rdev);
1608 kfree(journal_path);
1609 } else if (token == Opt_journal_ioprio) {
1611 ext4_msg(sb, KERN_ERR, "Invalid journal IO priority"
1616 IOPRIO_PRIO_VALUE(IOPRIO_CLASS_BE, arg);
1617 } else if (m->flags & MOPT_DATAJ) {
1619 if (!sbi->s_journal)
1620 ext4_msg(sb, KERN_WARNING, "Remounting file system with no journal so ignoring journalled data option");
1621 else if (test_opt(sb, DATA_FLAGS) != m->mount_opt) {
1622 ext4_msg(sb, KERN_ERR,
1623 "Cannot change data mode on remount");
1627 clear_opt(sb, DATA_FLAGS);
1628 sbi->s_mount_opt |= m->mount_opt;
1631 } else if (m->flags & MOPT_QFMT) {
1632 if (sb_any_quota_loaded(sb) &&
1633 sbi->s_jquota_fmt != m->mount_opt) {
1634 ext4_msg(sb, KERN_ERR, "Cannot change journaled "
1635 "quota options when quota turned on");
1638 if (EXT4_HAS_RO_COMPAT_FEATURE(sb,
1639 EXT4_FEATURE_RO_COMPAT_QUOTA)) {
1640 ext4_msg(sb, KERN_ERR,
1641 "Cannot set journaled quota options "
1642 "when QUOTA feature is enabled");
1645 sbi->s_jquota_fmt = m->mount_opt;
1650 if (m->flags & MOPT_CLEAR)
1652 else if (unlikely(!(m->flags & MOPT_SET))) {
1653 ext4_msg(sb, KERN_WARNING,
1654 "buggy handling of option %s", opt);
1659 sbi->s_mount_opt |= m->mount_opt;
1661 sbi->s_mount_opt &= ~m->mount_opt;
1666 static int parse_options(char *options, struct super_block *sb,
1667 unsigned long *journal_devnum,
1668 unsigned int *journal_ioprio,
1671 struct ext4_sb_info *sbi = EXT4_SB(sb);
1673 substring_t args[MAX_OPT_ARGS];
1679 while ((p = strsep(&options, ",")) != NULL) {
1683 * Initialize args struct so we know whether arg was
1684 * found; some options take optional arguments.
1686 args[0].to = args[0].from = NULL;
1687 token = match_token(p, tokens, args);
1688 if (handle_mount_opt(sb, p, token, args, journal_devnum,
1689 journal_ioprio, is_remount) < 0)
1693 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_QUOTA) &&
1694 (test_opt(sb, USRQUOTA) || test_opt(sb, GRPQUOTA))) {
1695 ext4_msg(sb, KERN_ERR, "Cannot set quota options when QUOTA "
1696 "feature is enabled");
1699 if (sbi->s_qf_names[USRQUOTA] || sbi->s_qf_names[GRPQUOTA]) {
1700 if (test_opt(sb, USRQUOTA) && sbi->s_qf_names[USRQUOTA])
1701 clear_opt(sb, USRQUOTA);
1703 if (test_opt(sb, GRPQUOTA) && sbi->s_qf_names[GRPQUOTA])
1704 clear_opt(sb, GRPQUOTA);
1706 if (test_opt(sb, GRPQUOTA) || test_opt(sb, USRQUOTA)) {
1707 ext4_msg(sb, KERN_ERR, "old and new quota "
1712 if (!sbi->s_jquota_fmt) {
1713 ext4_msg(sb, KERN_ERR, "journaled quota format "
1718 if (sbi->s_jquota_fmt) {
1719 ext4_msg(sb, KERN_ERR, "journaled quota format "
1720 "specified with no journaling "
1726 if (test_opt(sb, DIOREAD_NOLOCK)) {
1728 BLOCK_SIZE << le32_to_cpu(sbi->s_es->s_log_block_size);
1730 if (blocksize < PAGE_CACHE_SIZE) {
1731 ext4_msg(sb, KERN_ERR, "can't mount with "
1732 "dioread_nolock if block size != PAGE_SIZE");
1739 static inline void ext4_show_quota_options(struct seq_file *seq,
1740 struct super_block *sb)
1742 #if defined(CONFIG_QUOTA)
1743 struct ext4_sb_info *sbi = EXT4_SB(sb);
1745 if (sbi->s_jquota_fmt) {
1748 switch (sbi->s_jquota_fmt) {
1759 seq_printf(seq, ",jqfmt=%s", fmtname);
1762 if (sbi->s_qf_names[USRQUOTA])
1763 seq_printf(seq, ",usrjquota=%s", sbi->s_qf_names[USRQUOTA]);
1765 if (sbi->s_qf_names[GRPQUOTA])
1766 seq_printf(seq, ",grpjquota=%s", sbi->s_qf_names[GRPQUOTA]);
1770 static const char *token2str(int token)
1772 const struct match_token *t;
1774 for (t = tokens; t->token != Opt_err; t++)
1775 if (t->token == token && !strchr(t->pattern, '='))
1782 * - it's set to a non-default value OR
1783 * - if the per-sb default is different from the global default
1785 static int _ext4_show_options(struct seq_file *seq, struct super_block *sb,
1788 struct ext4_sb_info *sbi = EXT4_SB(sb);
1789 struct ext4_super_block *es = sbi->s_es;
1790 int def_errors, def_mount_opt = nodefs ? 0 : sbi->s_def_mount_opt;
1791 const struct mount_opts *m;
1792 char sep = nodefs ? '\n' : ',';
1794 #define SEQ_OPTS_PUTS(str) seq_printf(seq, "%c" str, sep)
1795 #define SEQ_OPTS_PRINT(str, arg) seq_printf(seq, "%c" str, sep, arg)
1797 if (sbi->s_sb_block != 1)
1798 SEQ_OPTS_PRINT("sb=%llu", sbi->s_sb_block);
1800 for (m = ext4_mount_opts; m->token != Opt_err; m++) {
1801 int want_set = m->flags & MOPT_SET;
1802 if (((m->flags & (MOPT_SET|MOPT_CLEAR)) == 0) ||
1803 (m->flags & MOPT_CLEAR_ERR))
1805 if (!(m->mount_opt & (sbi->s_mount_opt ^ def_mount_opt)))
1806 continue; /* skip if same as the default */
1808 (sbi->s_mount_opt & m->mount_opt) != m->mount_opt) ||
1809 (!want_set && (sbi->s_mount_opt & m->mount_opt)))
1810 continue; /* select Opt_noFoo vs Opt_Foo */
1811 SEQ_OPTS_PRINT("%s", token2str(m->token));
1814 if (nodefs || !uid_eq(sbi->s_resuid, make_kuid(&init_user_ns, EXT4_DEF_RESUID)) ||
1815 le16_to_cpu(es->s_def_resuid) != EXT4_DEF_RESUID)
1816 SEQ_OPTS_PRINT("resuid=%u",
1817 from_kuid_munged(&init_user_ns, sbi->s_resuid));
1818 if (nodefs || !gid_eq(sbi->s_resgid, make_kgid(&init_user_ns, EXT4_DEF_RESGID)) ||
1819 le16_to_cpu(es->s_def_resgid) != EXT4_DEF_RESGID)
1820 SEQ_OPTS_PRINT("resgid=%u",
1821 from_kgid_munged(&init_user_ns, sbi->s_resgid));
1822 def_errors = nodefs ? -1 : le16_to_cpu(es->s_errors);
1823 if (test_opt(sb, ERRORS_RO) && def_errors != EXT4_ERRORS_RO)
1824 SEQ_OPTS_PUTS("errors=remount-ro");
1825 if (test_opt(sb, ERRORS_CONT) && def_errors != EXT4_ERRORS_CONTINUE)
1826 SEQ_OPTS_PUTS("errors=continue");
1827 if (test_opt(sb, ERRORS_PANIC) && def_errors != EXT4_ERRORS_PANIC)
1828 SEQ_OPTS_PUTS("errors=panic");
1829 if (nodefs || sbi->s_commit_interval != JBD2_DEFAULT_MAX_COMMIT_AGE*HZ)
1830 SEQ_OPTS_PRINT("commit=%lu", sbi->s_commit_interval / HZ);
1831 if (nodefs || sbi->s_min_batch_time != EXT4_DEF_MIN_BATCH_TIME)
1832 SEQ_OPTS_PRINT("min_batch_time=%u", sbi->s_min_batch_time);
1833 if (nodefs || sbi->s_max_batch_time != EXT4_DEF_MAX_BATCH_TIME)
1834 SEQ_OPTS_PRINT("max_batch_time=%u", sbi->s_max_batch_time);
1835 if (sb->s_flags & MS_I_VERSION)
1836 SEQ_OPTS_PUTS("i_version");
1837 if (nodefs || sbi->s_stripe)
1838 SEQ_OPTS_PRINT("stripe=%lu", sbi->s_stripe);
1839 if (EXT4_MOUNT_DATA_FLAGS & (sbi->s_mount_opt ^ def_mount_opt)) {
1840 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA)
1841 SEQ_OPTS_PUTS("data=journal");
1842 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA)
1843 SEQ_OPTS_PUTS("data=ordered");
1844 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_WRITEBACK_DATA)
1845 SEQ_OPTS_PUTS("data=writeback");
1848 sbi->s_inode_readahead_blks != EXT4_DEF_INODE_READAHEAD_BLKS)
1849 SEQ_OPTS_PRINT("inode_readahead_blks=%u",
1850 sbi->s_inode_readahead_blks);
1852 if (nodefs || (test_opt(sb, INIT_INODE_TABLE) &&
1853 (sbi->s_li_wait_mult != EXT4_DEF_LI_WAIT_MULT)))
1854 SEQ_OPTS_PRINT("init_itable=%u", sbi->s_li_wait_mult);
1855 if (nodefs || sbi->s_max_dir_size_kb)
1856 SEQ_OPTS_PRINT("max_dir_size_kb=%u", sbi->s_max_dir_size_kb);
1858 ext4_show_quota_options(seq, sb);
1862 static int ext4_show_options(struct seq_file *seq, struct dentry *root)
1864 return _ext4_show_options(seq, root->d_sb, 0);
1867 static int options_seq_show(struct seq_file *seq, void *offset)
1869 struct super_block *sb = seq->private;
1872 seq_puts(seq, (sb->s_flags & MS_RDONLY) ? "ro" : "rw");
1873 rc = _ext4_show_options(seq, sb, 1);
1874 seq_puts(seq, "\n");
1878 static int options_open_fs(struct inode *inode, struct file *file)
1880 return single_open(file, options_seq_show, PDE_DATA(inode));
1883 static const struct file_operations ext4_seq_options_fops = {
1884 .owner = THIS_MODULE,
1885 .open = options_open_fs,
1887 .llseek = seq_lseek,
1888 .release = single_release,
1891 static int ext4_setup_super(struct super_block *sb, struct ext4_super_block *es,
1894 struct ext4_sb_info *sbi = EXT4_SB(sb);
1897 if (le32_to_cpu(es->s_rev_level) > EXT4_MAX_SUPP_REV) {
1898 ext4_msg(sb, KERN_ERR, "revision level too high, "
1899 "forcing read-only mode");
1904 if (!(sbi->s_mount_state & EXT4_VALID_FS))
1905 ext4_msg(sb, KERN_WARNING, "warning: mounting unchecked fs, "
1906 "running e2fsck is recommended");
1907 else if (sbi->s_mount_state & EXT4_ERROR_FS)
1908 ext4_msg(sb, KERN_WARNING,
1909 "warning: mounting fs with errors, "
1910 "running e2fsck is recommended");
1911 else if ((__s16) le16_to_cpu(es->s_max_mnt_count) > 0 &&
1912 le16_to_cpu(es->s_mnt_count) >=
1913 (unsigned short) (__s16) le16_to_cpu(es->s_max_mnt_count))
1914 ext4_msg(sb, KERN_WARNING,
1915 "warning: maximal mount count reached, "
1916 "running e2fsck is recommended");
1917 else if (le32_to_cpu(es->s_checkinterval) &&
1918 (le32_to_cpu(es->s_lastcheck) +
1919 le32_to_cpu(es->s_checkinterval) <= get_seconds()))
1920 ext4_msg(sb, KERN_WARNING,
1921 "warning: checktime reached, "
1922 "running e2fsck is recommended");
1923 if (!sbi->s_journal)
1924 es->s_state &= cpu_to_le16(~EXT4_VALID_FS);
1925 if (!(__s16) le16_to_cpu(es->s_max_mnt_count))
1926 es->s_max_mnt_count = cpu_to_le16(EXT4_DFL_MAX_MNT_COUNT);
1927 le16_add_cpu(&es->s_mnt_count, 1);
1928 es->s_mtime = cpu_to_le32(get_seconds());
1929 ext4_update_dynamic_rev(sb);
1931 EXT4_SET_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
1933 ext4_commit_super(sb, 1);
1935 if (test_opt(sb, DEBUG))
1936 printk(KERN_INFO "[EXT4 FS bs=%lu, gc=%u, "
1937 "bpg=%lu, ipg=%lu, mo=%04x, mo2=%04x]\n",
1939 sbi->s_groups_count,
1940 EXT4_BLOCKS_PER_GROUP(sb),
1941 EXT4_INODES_PER_GROUP(sb),
1942 sbi->s_mount_opt, sbi->s_mount_opt2);
1944 cleancache_init_fs(sb);
1948 int ext4_alloc_flex_bg_array(struct super_block *sb, ext4_group_t ngroup)
1950 struct ext4_sb_info *sbi = EXT4_SB(sb);
1951 struct flex_groups *new_groups;
1954 if (!sbi->s_log_groups_per_flex)
1957 size = ext4_flex_group(sbi, ngroup - 1) + 1;
1958 if (size <= sbi->s_flex_groups_allocated)
1961 size = roundup_pow_of_two(size * sizeof(struct flex_groups));
1962 new_groups = ext4_kvzalloc(size, GFP_KERNEL);
1964 ext4_msg(sb, KERN_ERR, "not enough memory for %d flex groups",
1965 size / (int) sizeof(struct flex_groups));
1969 if (sbi->s_flex_groups) {
1970 memcpy(new_groups, sbi->s_flex_groups,
1971 (sbi->s_flex_groups_allocated *
1972 sizeof(struct flex_groups)));
1973 ext4_kvfree(sbi->s_flex_groups);
1975 sbi->s_flex_groups = new_groups;
1976 sbi->s_flex_groups_allocated = size / sizeof(struct flex_groups);
1980 static int ext4_fill_flex_info(struct super_block *sb)
1982 struct ext4_sb_info *sbi = EXT4_SB(sb);
1983 struct ext4_group_desc *gdp = NULL;
1984 ext4_group_t flex_group;
1987 sbi->s_log_groups_per_flex = sbi->s_es->s_log_groups_per_flex;
1988 if (sbi->s_log_groups_per_flex < 1 || sbi->s_log_groups_per_flex > 31) {
1989 sbi->s_log_groups_per_flex = 0;
1993 err = ext4_alloc_flex_bg_array(sb, sbi->s_groups_count);
1997 for (i = 0; i < sbi->s_groups_count; i++) {
1998 gdp = ext4_get_group_desc(sb, i, NULL);
2000 flex_group = ext4_flex_group(sbi, i);
2001 atomic_add(ext4_free_inodes_count(sb, gdp),
2002 &sbi->s_flex_groups[flex_group].free_inodes);
2003 atomic64_add(ext4_free_group_clusters(sb, gdp),
2004 &sbi->s_flex_groups[flex_group].free_clusters);
2005 atomic_add(ext4_used_dirs_count(sb, gdp),
2006 &sbi->s_flex_groups[flex_group].used_dirs);
2014 static __le16 ext4_group_desc_csum(struct ext4_sb_info *sbi, __u32 block_group,
2015 struct ext4_group_desc *gdp)
2019 __le32 le_group = cpu_to_le32(block_group);
2021 if ((sbi->s_es->s_feature_ro_compat &
2022 cpu_to_le32(EXT4_FEATURE_RO_COMPAT_METADATA_CSUM))) {
2023 /* Use new metadata_csum algorithm */
2027 save_csum = gdp->bg_checksum;
2028 gdp->bg_checksum = 0;
2029 csum32 = ext4_chksum(sbi, sbi->s_csum_seed, (__u8 *)&le_group,
2031 csum32 = ext4_chksum(sbi, csum32, (__u8 *)gdp,
2033 gdp->bg_checksum = save_csum;
2035 crc = csum32 & 0xFFFF;
2039 /* old crc16 code */
2040 offset = offsetof(struct ext4_group_desc, bg_checksum);
2042 crc = crc16(~0, sbi->s_es->s_uuid, sizeof(sbi->s_es->s_uuid));
2043 crc = crc16(crc, (__u8 *)&le_group, sizeof(le_group));
2044 crc = crc16(crc, (__u8 *)gdp, offset);
2045 offset += sizeof(gdp->bg_checksum); /* skip checksum */
2046 /* for checksum of struct ext4_group_desc do the rest...*/
2047 if ((sbi->s_es->s_feature_incompat &
2048 cpu_to_le32(EXT4_FEATURE_INCOMPAT_64BIT)) &&
2049 offset < le16_to_cpu(sbi->s_es->s_desc_size))
2050 crc = crc16(crc, (__u8 *)gdp + offset,
2051 le16_to_cpu(sbi->s_es->s_desc_size) -
2055 return cpu_to_le16(crc);
2058 int ext4_group_desc_csum_verify(struct super_block *sb, __u32 block_group,
2059 struct ext4_group_desc *gdp)
2061 if (ext4_has_group_desc_csum(sb) &&
2062 (gdp->bg_checksum != ext4_group_desc_csum(EXT4_SB(sb),
2069 void ext4_group_desc_csum_set(struct super_block *sb, __u32 block_group,
2070 struct ext4_group_desc *gdp)
2072 if (!ext4_has_group_desc_csum(sb))
2074 gdp->bg_checksum = ext4_group_desc_csum(EXT4_SB(sb), block_group, gdp);
2077 /* Called at mount-time, super-block is locked */
2078 static int ext4_check_descriptors(struct super_block *sb,
2079 ext4_group_t *first_not_zeroed)
2081 struct ext4_sb_info *sbi = EXT4_SB(sb);
2082 ext4_fsblk_t first_block = le32_to_cpu(sbi->s_es->s_first_data_block);
2083 ext4_fsblk_t last_block;
2084 ext4_fsblk_t block_bitmap;
2085 ext4_fsblk_t inode_bitmap;
2086 ext4_fsblk_t inode_table;
2087 int flexbg_flag = 0;
2088 ext4_group_t i, grp = sbi->s_groups_count;
2090 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_FLEX_BG))
2093 ext4_debug("Checking group descriptors");
2095 for (i = 0; i < sbi->s_groups_count; i++) {
2096 struct ext4_group_desc *gdp = ext4_get_group_desc(sb, i, NULL);
2098 if (i == sbi->s_groups_count - 1 || flexbg_flag)
2099 last_block = ext4_blocks_count(sbi->s_es) - 1;
2101 last_block = first_block +
2102 (EXT4_BLOCKS_PER_GROUP(sb) - 1);
2104 if ((grp == sbi->s_groups_count) &&
2105 !(gdp->bg_flags & cpu_to_le16(EXT4_BG_INODE_ZEROED)))
2108 block_bitmap = ext4_block_bitmap(sb, gdp);
2109 if (block_bitmap < first_block || block_bitmap > last_block) {
2110 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2111 "Block bitmap for group %u not in group "
2112 "(block %llu)!", i, block_bitmap);
2115 inode_bitmap = ext4_inode_bitmap(sb, gdp);
2116 if (inode_bitmap < first_block || inode_bitmap > last_block) {
2117 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2118 "Inode bitmap for group %u not in group "
2119 "(block %llu)!", i, inode_bitmap);
2122 inode_table = ext4_inode_table(sb, gdp);
2123 if (inode_table < first_block ||
2124 inode_table + sbi->s_itb_per_group - 1 > last_block) {
2125 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2126 "Inode table for group %u not in group "
2127 "(block %llu)!", i, inode_table);
2130 ext4_lock_group(sb, i);
2131 if (!ext4_group_desc_csum_verify(sb, i, gdp)) {
2132 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2133 "Checksum for group %u failed (%u!=%u)",
2134 i, le16_to_cpu(ext4_group_desc_csum(sbi, i,
2135 gdp)), le16_to_cpu(gdp->bg_checksum));
2136 if (!(sb->s_flags & MS_RDONLY)) {
2137 ext4_unlock_group(sb, i);
2141 ext4_unlock_group(sb, i);
2143 first_block += EXT4_BLOCKS_PER_GROUP(sb);
2145 if (NULL != first_not_zeroed)
2146 *first_not_zeroed = grp;
2148 ext4_free_blocks_count_set(sbi->s_es,
2149 EXT4_C2B(sbi, ext4_count_free_clusters(sb)));
2150 sbi->s_es->s_free_inodes_count =cpu_to_le32(ext4_count_free_inodes(sb));
2154 /* ext4_orphan_cleanup() walks a singly-linked list of inodes (starting at
2155 * the superblock) which were deleted from all directories, but held open by
2156 * a process at the time of a crash. We walk the list and try to delete these
2157 * inodes at recovery time (only with a read-write filesystem).
2159 * In order to keep the orphan inode chain consistent during traversal (in
2160 * case of crash during recovery), we link each inode into the superblock
2161 * orphan list_head and handle it the same way as an inode deletion during
2162 * normal operation (which journals the operations for us).
2164 * We only do an iget() and an iput() on each inode, which is very safe if we
2165 * accidentally point at an in-use or already deleted inode. The worst that
2166 * can happen in this case is that we get a "bit already cleared" message from
2167 * ext4_free_inode(). The only reason we would point at a wrong inode is if
2168 * e2fsck was run on this filesystem, and it must have already done the orphan
2169 * inode cleanup for us, so we can safely abort without any further action.
2171 static void ext4_orphan_cleanup(struct super_block *sb,
2172 struct ext4_super_block *es)
2174 unsigned int s_flags = sb->s_flags;
2175 int nr_orphans = 0, nr_truncates = 0;
2179 if (!es->s_last_orphan) {
2180 jbd_debug(4, "no orphan inodes to clean up\n");
2184 if (bdev_read_only(sb->s_bdev)) {
2185 ext4_msg(sb, KERN_ERR, "write access "
2186 "unavailable, skipping orphan cleanup");
2190 /* Check if feature set would not allow a r/w mount */
2191 if (!ext4_feature_set_ok(sb, 0)) {
2192 ext4_msg(sb, KERN_INFO, "Skipping orphan cleanup due to "
2193 "unknown ROCOMPAT features");
2197 if (EXT4_SB(sb)->s_mount_state & EXT4_ERROR_FS) {
2198 /* don't clear list on RO mount w/ errors */
2199 if (es->s_last_orphan && !(s_flags & MS_RDONLY)) {
2200 jbd_debug(1, "Errors on filesystem, "
2201 "clearing orphan list.\n");
2202 es->s_last_orphan = 0;
2204 jbd_debug(1, "Skipping orphan recovery on fs with errors.\n");
2208 if (s_flags & MS_RDONLY) {
2209 ext4_msg(sb, KERN_INFO, "orphan cleanup on readonly fs");
2210 sb->s_flags &= ~MS_RDONLY;
2213 /* Needed for iput() to work correctly and not trash data */
2214 sb->s_flags |= MS_ACTIVE;
2215 /* Turn on quotas so that they are updated correctly */
2216 for (i = 0; i < MAXQUOTAS; i++) {
2217 if (EXT4_SB(sb)->s_qf_names[i]) {
2218 int ret = ext4_quota_on_mount(sb, i);
2220 ext4_msg(sb, KERN_ERR,
2221 "Cannot turn on journaled "
2222 "quota: error %d", ret);
2227 while (es->s_last_orphan) {
2228 struct inode *inode;
2230 inode = ext4_orphan_get(sb, le32_to_cpu(es->s_last_orphan));
2231 if (IS_ERR(inode)) {
2232 es->s_last_orphan = 0;
2236 list_add(&EXT4_I(inode)->i_orphan, &EXT4_SB(sb)->s_orphan);
2237 dquot_initialize(inode);
2238 if (inode->i_nlink) {
2239 if (test_opt(sb, DEBUG))
2240 ext4_msg(sb, KERN_DEBUG,
2241 "%s: truncating inode %lu to %lld bytes",
2242 __func__, inode->i_ino, inode->i_size);
2243 jbd_debug(2, "truncating inode %lu to %lld bytes\n",
2244 inode->i_ino, inode->i_size);
2245 mutex_lock(&inode->i_mutex);
2246 truncate_inode_pages(inode->i_mapping, inode->i_size);
2247 ext4_truncate(inode);
2248 mutex_unlock(&inode->i_mutex);
2251 if (test_opt(sb, DEBUG))
2252 ext4_msg(sb, KERN_DEBUG,
2253 "%s: deleting unreferenced inode %lu",
2254 __func__, inode->i_ino);
2255 jbd_debug(2, "deleting unreferenced inode %lu\n",
2259 iput(inode); /* The delete magic happens here! */
2262 #define PLURAL(x) (x), ((x) == 1) ? "" : "s"
2265 ext4_msg(sb, KERN_INFO, "%d orphan inode%s deleted",
2266 PLURAL(nr_orphans));
2268 ext4_msg(sb, KERN_INFO, "%d truncate%s cleaned up",
2269 PLURAL(nr_truncates));
2271 /* Turn quotas off */
2272 for (i = 0; i < MAXQUOTAS; i++) {
2273 if (sb_dqopt(sb)->files[i])
2274 dquot_quota_off(sb, i);
2277 sb->s_flags = s_flags; /* Restore MS_RDONLY status */
2281 * Maximal extent format file size.
2282 * Resulting logical blkno at s_maxbytes must fit in our on-disk
2283 * extent format containers, within a sector_t, and within i_blocks
2284 * in the vfs. ext4 inode has 48 bits of i_block in fsblock units,
2285 * so that won't be a limiting factor.
2287 * However there is other limiting factor. We do store extents in the form
2288 * of starting block and length, hence the resulting length of the extent
2289 * covering maximum file size must fit into on-disk format containers as
2290 * well. Given that length is always by 1 unit bigger than max unit (because
2291 * we count 0 as well) we have to lower the s_maxbytes by one fs block.
2293 * Note, this does *not* consider any metadata overhead for vfs i_blocks.
2295 static loff_t ext4_max_size(int blkbits, int has_huge_files)
2298 loff_t upper_limit = MAX_LFS_FILESIZE;
2300 /* small i_blocks in vfs inode? */
2301 if (!has_huge_files || sizeof(blkcnt_t) < sizeof(u64)) {
2303 * CONFIG_LBDAF is not enabled implies the inode
2304 * i_block represent total blocks in 512 bytes
2305 * 32 == size of vfs inode i_blocks * 8
2307 upper_limit = (1LL << 32) - 1;
2309 /* total blocks in file system block size */
2310 upper_limit >>= (blkbits - 9);
2311 upper_limit <<= blkbits;
2315 * 32-bit extent-start container, ee_block. We lower the maxbytes
2316 * by one fs block, so ee_len can cover the extent of maximum file
2319 res = (1LL << 32) - 1;
2322 /* Sanity check against vm- & vfs- imposed limits */
2323 if (res > upper_limit)
2330 * Maximal bitmap file size. There is a direct, and {,double-,triple-}indirect
2331 * block limit, and also a limit of (2^48 - 1) 512-byte sectors in i_blocks.
2332 * We need to be 1 filesystem block less than the 2^48 sector limit.
2334 static loff_t ext4_max_bitmap_size(int bits, int has_huge_files)
2336 loff_t res = EXT4_NDIR_BLOCKS;
2339 /* This is calculated to be the largest file size for a dense, block
2340 * mapped file such that the file's total number of 512-byte sectors,
2341 * including data and all indirect blocks, does not exceed (2^48 - 1).
2343 * __u32 i_blocks_lo and _u16 i_blocks_high represent the total
2344 * number of 512-byte sectors of the file.
2347 if (!has_huge_files || sizeof(blkcnt_t) < sizeof(u64)) {
2349 * !has_huge_files or CONFIG_LBDAF not enabled implies that
2350 * the inode i_block field represents total file blocks in
2351 * 2^32 512-byte sectors == size of vfs inode i_blocks * 8
2353 upper_limit = (1LL << 32) - 1;
2355 /* total blocks in file system block size */
2356 upper_limit >>= (bits - 9);
2360 * We use 48 bit ext4_inode i_blocks
2361 * With EXT4_HUGE_FILE_FL set the i_blocks
2362 * represent total number of blocks in
2363 * file system block size
2365 upper_limit = (1LL << 48) - 1;
2369 /* indirect blocks */
2371 /* double indirect blocks */
2372 meta_blocks += 1 + (1LL << (bits-2));
2373 /* tripple indirect blocks */
2374 meta_blocks += 1 + (1LL << (bits-2)) + (1LL << (2*(bits-2)));
2376 upper_limit -= meta_blocks;
2377 upper_limit <<= bits;
2379 res += 1LL << (bits-2);
2380 res += 1LL << (2*(bits-2));
2381 res += 1LL << (3*(bits-2));
2383 if (res > upper_limit)
2386 if (res > MAX_LFS_FILESIZE)
2387 res = MAX_LFS_FILESIZE;
2392 static ext4_fsblk_t descriptor_loc(struct super_block *sb,
2393 ext4_fsblk_t logical_sb_block, int nr)
2395 struct ext4_sb_info *sbi = EXT4_SB(sb);
2396 ext4_group_t bg, first_meta_bg;
2399 first_meta_bg = le32_to_cpu(sbi->s_es->s_first_meta_bg);
2401 if (!EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_META_BG) ||
2403 return logical_sb_block + nr + 1;
2404 bg = sbi->s_desc_per_block * nr;
2405 if (ext4_bg_has_super(sb, bg))
2409 * If we have a meta_bg fs with 1k blocks, group 0's GDT is at
2410 * block 2, not 1. If s_first_data_block == 0 (bigalloc is enabled
2411 * on modern mke2fs or blksize > 1k on older mke2fs) then we must
2414 if (sb->s_blocksize == 1024 && nr == 0 &&
2415 le32_to_cpu(EXT4_SB(sb)->s_es->s_first_data_block) == 0)
2418 return (has_super + ext4_group_first_block_no(sb, bg));
2422 * ext4_get_stripe_size: Get the stripe size.
2423 * @sbi: In memory super block info
2425 * If we have specified it via mount option, then
2426 * use the mount option value. If the value specified at mount time is
2427 * greater than the blocks per group use the super block value.
2428 * If the super block value is greater than blocks per group return 0.
2429 * Allocator needs it be less than blocks per group.
2432 static unsigned long ext4_get_stripe_size(struct ext4_sb_info *sbi)
2434 unsigned long stride = le16_to_cpu(sbi->s_es->s_raid_stride);
2435 unsigned long stripe_width =
2436 le32_to_cpu(sbi->s_es->s_raid_stripe_width);
2439 if (sbi->s_stripe && sbi->s_stripe <= sbi->s_blocks_per_group)
2440 ret = sbi->s_stripe;
2441 else if (stripe_width <= sbi->s_blocks_per_group)
2443 else if (stride <= sbi->s_blocks_per_group)
2449 * If the stripe width is 1, this makes no sense and
2450 * we set it to 0 to turn off stripe handling code.
2461 struct attribute attr;
2462 ssize_t (*show)(struct ext4_attr *, struct ext4_sb_info *, char *);
2463 ssize_t (*store)(struct ext4_attr *, struct ext4_sb_info *,
2464 const char *, size_t);
2471 static int parse_strtoull(const char *buf,
2472 unsigned long long max, unsigned long long *value)
2476 ret = kstrtoull(skip_spaces(buf), 0, value);
2477 if (!ret && *value > max)
2482 static ssize_t delayed_allocation_blocks_show(struct ext4_attr *a,
2483 struct ext4_sb_info *sbi,
2486 return snprintf(buf, PAGE_SIZE, "%llu\n",
2488 percpu_counter_sum(&sbi->s_dirtyclusters_counter)));
2491 static ssize_t session_write_kbytes_show(struct ext4_attr *a,
2492 struct ext4_sb_info *sbi, char *buf)
2494 struct super_block *sb = sbi->s_buddy_cache->i_sb;
2496 if (!sb->s_bdev->bd_part)
2497 return snprintf(buf, PAGE_SIZE, "0\n");
2498 return snprintf(buf, PAGE_SIZE, "%lu\n",
2499 (part_stat_read(sb->s_bdev->bd_part, sectors[1]) -
2500 sbi->s_sectors_written_start) >> 1);
2503 static ssize_t lifetime_write_kbytes_show(struct ext4_attr *a,
2504 struct ext4_sb_info *sbi, char *buf)
2506 struct super_block *sb = sbi->s_buddy_cache->i_sb;
2508 if (!sb->s_bdev->bd_part)
2509 return snprintf(buf, PAGE_SIZE, "0\n");
2510 return snprintf(buf, PAGE_SIZE, "%llu\n",
2511 (unsigned long long)(sbi->s_kbytes_written +
2512 ((part_stat_read(sb->s_bdev->bd_part, sectors[1]) -
2513 EXT4_SB(sb)->s_sectors_written_start) >> 1)));
2516 static ssize_t inode_readahead_blks_store(struct ext4_attr *a,
2517 struct ext4_sb_info *sbi,
2518 const char *buf, size_t count)
2523 ret = kstrtoul(skip_spaces(buf), 0, &t);
2527 if (t && (!is_power_of_2(t) || t > 0x40000000))
2530 sbi->s_inode_readahead_blks = t;
2534 static ssize_t sbi_ui_show(struct ext4_attr *a,
2535 struct ext4_sb_info *sbi, char *buf)
2537 unsigned int *ui = (unsigned int *) (((char *) sbi) + a->u.offset);
2539 return snprintf(buf, PAGE_SIZE, "%u\n", *ui);
2542 static ssize_t sbi_ui_store(struct ext4_attr *a,
2543 struct ext4_sb_info *sbi,
2544 const char *buf, size_t count)
2546 unsigned int *ui = (unsigned int *) (((char *) sbi) + a->u.offset);
2550 ret = kstrtoul(skip_spaces(buf), 0, &t);
2557 static ssize_t reserved_clusters_show(struct ext4_attr *a,
2558 struct ext4_sb_info *sbi, char *buf)
2560 return snprintf(buf, PAGE_SIZE, "%llu\n",
2561 (unsigned long long) atomic64_read(&sbi->s_resv_clusters));
2564 static ssize_t reserved_clusters_store(struct ext4_attr *a,
2565 struct ext4_sb_info *sbi,
2566 const char *buf, size_t count)
2568 unsigned long long val;
2571 if (parse_strtoull(buf, -1ULL, &val))
2573 ret = ext4_reserve_clusters(sbi, val);
2575 return ret ? ret : count;
2578 static ssize_t trigger_test_error(struct ext4_attr *a,
2579 struct ext4_sb_info *sbi,
2580 const char *buf, size_t count)
2584 if (!capable(CAP_SYS_ADMIN))
2587 if (len && buf[len-1] == '\n')
2591 ext4_error(sbi->s_sb, "%.*s", len, buf);
2595 static ssize_t sbi_deprecated_show(struct ext4_attr *a,
2596 struct ext4_sb_info *sbi, char *buf)
2598 return snprintf(buf, PAGE_SIZE, "%d\n", a->u.deprecated_val);
2601 #define EXT4_ATTR_OFFSET(_name,_mode,_show,_store,_elname) \
2602 static struct ext4_attr ext4_attr_##_name = { \
2603 .attr = {.name = __stringify(_name), .mode = _mode }, \
2607 .offset = offsetof(struct ext4_sb_info, _elname),\
2610 #define EXT4_ATTR(name, mode, show, store) \
2611 static struct ext4_attr ext4_attr_##name = __ATTR(name, mode, show, store)
2613 #define EXT4_INFO_ATTR(name) EXT4_ATTR(name, 0444, NULL, NULL)
2614 #define EXT4_RO_ATTR(name) EXT4_ATTR(name, 0444, name##_show, NULL)
2615 #define EXT4_RW_ATTR(name) EXT4_ATTR(name, 0644, name##_show, name##_store)
2616 #define EXT4_RW_ATTR_SBI_UI(name, elname) \
2617 EXT4_ATTR_OFFSET(name, 0644, sbi_ui_show, sbi_ui_store, elname)
2618 #define ATTR_LIST(name) &ext4_attr_##name.attr
2619 #define EXT4_DEPRECATED_ATTR(_name, _val) \
2620 static struct ext4_attr ext4_attr_##_name = { \
2621 .attr = {.name = __stringify(_name), .mode = 0444 }, \
2622 .show = sbi_deprecated_show, \
2624 .deprecated_val = _val, \
2628 EXT4_RO_ATTR(delayed_allocation_blocks);
2629 EXT4_RO_ATTR(session_write_kbytes);
2630 EXT4_RO_ATTR(lifetime_write_kbytes);
2631 EXT4_RW_ATTR(reserved_clusters);
2632 EXT4_ATTR_OFFSET(inode_readahead_blks, 0644, sbi_ui_show,
2633 inode_readahead_blks_store, s_inode_readahead_blks);
2634 EXT4_RW_ATTR_SBI_UI(inode_goal, s_inode_goal);
2635 EXT4_RW_ATTR_SBI_UI(mb_stats, s_mb_stats);
2636 EXT4_RW_ATTR_SBI_UI(mb_max_to_scan, s_mb_max_to_scan);
2637 EXT4_RW_ATTR_SBI_UI(mb_min_to_scan, s_mb_min_to_scan);
2638 EXT4_RW_ATTR_SBI_UI(mb_order2_req, s_mb_order2_reqs);
2639 EXT4_RW_ATTR_SBI_UI(mb_stream_req, s_mb_stream_request);
2640 EXT4_RW_ATTR_SBI_UI(mb_group_prealloc, s_mb_group_prealloc);
2641 EXT4_DEPRECATED_ATTR(max_writeback_mb_bump, 128);
2642 EXT4_RW_ATTR_SBI_UI(extent_max_zeroout_kb, s_extent_max_zeroout_kb);
2643 EXT4_ATTR(trigger_fs_error, 0200, NULL, trigger_test_error);
2644 EXT4_RW_ATTR_SBI_UI(err_ratelimit_interval_ms, s_err_ratelimit_state.interval);
2645 EXT4_RW_ATTR_SBI_UI(err_ratelimit_burst, s_err_ratelimit_state.burst);
2646 EXT4_RW_ATTR_SBI_UI(warning_ratelimit_interval_ms, s_warning_ratelimit_state.interval);
2647 EXT4_RW_ATTR_SBI_UI(warning_ratelimit_burst, s_warning_ratelimit_state.burst);
2648 EXT4_RW_ATTR_SBI_UI(msg_ratelimit_interval_ms, s_msg_ratelimit_state.interval);
2649 EXT4_RW_ATTR_SBI_UI(msg_ratelimit_burst, s_msg_ratelimit_state.burst);
2651 static struct attribute *ext4_attrs[] = {
2652 ATTR_LIST(delayed_allocation_blocks),
2653 ATTR_LIST(session_write_kbytes),
2654 ATTR_LIST(lifetime_write_kbytes),
2655 ATTR_LIST(reserved_clusters),
2656 ATTR_LIST(inode_readahead_blks),
2657 ATTR_LIST(inode_goal),
2658 ATTR_LIST(mb_stats),
2659 ATTR_LIST(mb_max_to_scan),
2660 ATTR_LIST(mb_min_to_scan),
2661 ATTR_LIST(mb_order2_req),
2662 ATTR_LIST(mb_stream_req),
2663 ATTR_LIST(mb_group_prealloc),
2664 ATTR_LIST(max_writeback_mb_bump),
2665 ATTR_LIST(extent_max_zeroout_kb),
2666 ATTR_LIST(trigger_fs_error),
2667 ATTR_LIST(err_ratelimit_interval_ms),
2668 ATTR_LIST(err_ratelimit_burst),
2669 ATTR_LIST(warning_ratelimit_interval_ms),
2670 ATTR_LIST(warning_ratelimit_burst),
2671 ATTR_LIST(msg_ratelimit_interval_ms),
2672 ATTR_LIST(msg_ratelimit_burst),
2676 /* Features this copy of ext4 supports */
2677 EXT4_INFO_ATTR(lazy_itable_init);
2678 EXT4_INFO_ATTR(batched_discard);
2679 EXT4_INFO_ATTR(meta_bg_resize);
2681 static struct attribute *ext4_feat_attrs[] = {
2682 ATTR_LIST(lazy_itable_init),
2683 ATTR_LIST(batched_discard),
2684 ATTR_LIST(meta_bg_resize),
2688 static ssize_t ext4_attr_show(struct kobject *kobj,
2689 struct attribute *attr, char *buf)
2691 struct ext4_sb_info *sbi = container_of(kobj, struct ext4_sb_info,
2693 struct ext4_attr *a = container_of(attr, struct ext4_attr, attr);
2695 return a->show ? a->show(a, sbi, buf) : 0;
2698 static ssize_t ext4_attr_store(struct kobject *kobj,
2699 struct attribute *attr,
2700 const char *buf, size_t len)
2702 struct ext4_sb_info *sbi = container_of(kobj, struct ext4_sb_info,
2704 struct ext4_attr *a = container_of(attr, struct ext4_attr, attr);
2706 return a->store ? a->store(a, sbi, buf, len) : 0;
2709 static void ext4_sb_release(struct kobject *kobj)
2711 struct ext4_sb_info *sbi = container_of(kobj, struct ext4_sb_info,
2713 complete(&sbi->s_kobj_unregister);
2716 static const struct sysfs_ops ext4_attr_ops = {
2717 .show = ext4_attr_show,
2718 .store = ext4_attr_store,
2721 static struct kobj_type ext4_ktype = {
2722 .default_attrs = ext4_attrs,
2723 .sysfs_ops = &ext4_attr_ops,
2724 .release = ext4_sb_release,
2727 static void ext4_feat_release(struct kobject *kobj)
2729 complete(&ext4_feat->f_kobj_unregister);
2732 static struct kobj_type ext4_feat_ktype = {
2733 .default_attrs = ext4_feat_attrs,
2734 .sysfs_ops = &ext4_attr_ops,
2735 .release = ext4_feat_release,
2739 * Check whether this filesystem can be mounted based on
2740 * the features present and the RDONLY/RDWR mount requested.
2741 * Returns 1 if this filesystem can be mounted as requested,
2742 * 0 if it cannot be.
2744 static int ext4_feature_set_ok(struct super_block *sb, int readonly)
2746 if (EXT4_HAS_INCOMPAT_FEATURE(sb, ~EXT4_FEATURE_INCOMPAT_SUPP)) {
2747 ext4_msg(sb, KERN_ERR,
2748 "Couldn't mount because of "
2749 "unsupported optional features (%x)",
2750 (le32_to_cpu(EXT4_SB(sb)->s_es->s_feature_incompat) &
2751 ~EXT4_FEATURE_INCOMPAT_SUPP));
2758 /* Check that feature set is OK for a read-write mount */
2759 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, ~EXT4_FEATURE_RO_COMPAT_SUPP)) {
2760 ext4_msg(sb, KERN_ERR, "couldn't mount RDWR because of "
2761 "unsupported optional features (%x)",
2762 (le32_to_cpu(EXT4_SB(sb)->s_es->s_feature_ro_compat) &
2763 ~EXT4_FEATURE_RO_COMPAT_SUPP));
2767 * Large file size enabled file system can only be mounted
2768 * read-write on 32-bit systems if kernel is built with CONFIG_LBDAF
2770 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_HUGE_FILE)) {
2771 if (sizeof(blkcnt_t) < sizeof(u64)) {
2772 ext4_msg(sb, KERN_ERR, "Filesystem with huge files "
2773 "cannot be mounted RDWR without "
2778 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_BIGALLOC) &&
2779 !EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_EXTENTS)) {
2780 ext4_msg(sb, KERN_ERR,
2781 "Can't support bigalloc feature without "
2782 "extents feature\n");
2786 #ifndef CONFIG_QUOTA
2787 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_QUOTA) &&
2789 ext4_msg(sb, KERN_ERR,
2790 "Filesystem with quota feature cannot be mounted RDWR "
2791 "without CONFIG_QUOTA");
2794 #endif /* CONFIG_QUOTA */
2799 * This function is called once a day if we have errors logged
2800 * on the file system
2802 static void print_daily_error_info(unsigned long arg)
2804 struct super_block *sb = (struct super_block *) arg;
2805 struct ext4_sb_info *sbi;
2806 struct ext4_super_block *es;
2811 if (es->s_error_count)
2812 /* fsck newer than v1.41.13 is needed to clean this condition. */
2813 ext4_msg(sb, KERN_NOTICE, "error count since last fsck: %u",
2814 le32_to_cpu(es->s_error_count));
2815 if (es->s_first_error_time) {
2816 printk(KERN_NOTICE "EXT4-fs (%s): initial error at time %u: %.*s:%d",
2817 sb->s_id, le32_to_cpu(es->s_first_error_time),
2818 (int) sizeof(es->s_first_error_func),
2819 es->s_first_error_func,
2820 le32_to_cpu(es->s_first_error_line));
2821 if (es->s_first_error_ino)
2822 printk(": inode %u",
2823 le32_to_cpu(es->s_first_error_ino));
2824 if (es->s_first_error_block)
2825 printk(": block %llu", (unsigned long long)
2826 le64_to_cpu(es->s_first_error_block));
2829 if (es->s_last_error_time) {
2830 printk(KERN_NOTICE "EXT4-fs (%s): last error at time %u: %.*s:%d",
2831 sb->s_id, le32_to_cpu(es->s_last_error_time),
2832 (int) sizeof(es->s_last_error_func),
2833 es->s_last_error_func,
2834 le32_to_cpu(es->s_last_error_line));
2835 if (es->s_last_error_ino)
2836 printk(": inode %u",
2837 le32_to_cpu(es->s_last_error_ino));
2838 if (es->s_last_error_block)
2839 printk(": block %llu", (unsigned long long)
2840 le64_to_cpu(es->s_last_error_block));
2843 mod_timer(&sbi->s_err_report, jiffies + 24*60*60*HZ); /* Once a day */
2846 /* Find next suitable group and run ext4_init_inode_table */
2847 static int ext4_run_li_request(struct ext4_li_request *elr)
2849 struct ext4_group_desc *gdp = NULL;
2850 ext4_group_t group, ngroups;
2851 struct super_block *sb;
2852 unsigned long timeout = 0;
2856 ngroups = EXT4_SB(sb)->s_groups_count;
2859 for (group = elr->lr_next_group; group < ngroups; group++) {
2860 gdp = ext4_get_group_desc(sb, group, NULL);
2866 if (!(gdp->bg_flags & cpu_to_le16(EXT4_BG_INODE_ZEROED)))
2870 if (group >= ngroups)
2875 ret = ext4_init_inode_table(sb, group,
2876 elr->lr_timeout ? 0 : 1);
2877 if (elr->lr_timeout == 0) {
2878 timeout = (jiffies - timeout) *
2879 elr->lr_sbi->s_li_wait_mult;
2880 elr->lr_timeout = timeout;
2882 elr->lr_next_sched = jiffies + elr->lr_timeout;
2883 elr->lr_next_group = group + 1;
2891 * Remove lr_request from the list_request and free the
2892 * request structure. Should be called with li_list_mtx held
2894 static void ext4_remove_li_request(struct ext4_li_request *elr)
2896 struct ext4_sb_info *sbi;
2903 list_del(&elr->lr_request);
2904 sbi->s_li_request = NULL;
2908 static void ext4_unregister_li_request(struct super_block *sb)
2910 mutex_lock(&ext4_li_mtx);
2911 if (!ext4_li_info) {
2912 mutex_unlock(&ext4_li_mtx);
2916 mutex_lock(&ext4_li_info->li_list_mtx);
2917 ext4_remove_li_request(EXT4_SB(sb)->s_li_request);
2918 mutex_unlock(&ext4_li_info->li_list_mtx);
2919 mutex_unlock(&ext4_li_mtx);
2922 static struct task_struct *ext4_lazyinit_task;
2925 * This is the function where ext4lazyinit thread lives. It walks
2926 * through the request list searching for next scheduled filesystem.
2927 * When such a fs is found, run the lazy initialization request
2928 * (ext4_rn_li_request) and keep track of the time spend in this
2929 * function. Based on that time we compute next schedule time of
2930 * the request. When walking through the list is complete, compute
2931 * next waking time and put itself into sleep.
2933 static int ext4_lazyinit_thread(void *arg)
2935 struct ext4_lazy_init *eli = (struct ext4_lazy_init *)arg;
2936 struct list_head *pos, *n;
2937 struct ext4_li_request *elr;
2938 unsigned long next_wakeup, cur;
2940 BUG_ON(NULL == eli);
2944 next_wakeup = MAX_JIFFY_OFFSET;
2946 mutex_lock(&eli->li_list_mtx);
2947 if (list_empty(&eli->li_request_list)) {
2948 mutex_unlock(&eli->li_list_mtx);
2952 list_for_each_safe(pos, n, &eli->li_request_list) {
2953 elr = list_entry(pos, struct ext4_li_request,
2956 if (time_after_eq(jiffies, elr->lr_next_sched)) {
2957 if (ext4_run_li_request(elr) != 0) {
2958 /* error, remove the lazy_init job */
2959 ext4_remove_li_request(elr);
2964 if (time_before(elr->lr_next_sched, next_wakeup))
2965 next_wakeup = elr->lr_next_sched;
2967 mutex_unlock(&eli->li_list_mtx);
2972 if ((time_after_eq(cur, next_wakeup)) ||
2973 (MAX_JIFFY_OFFSET == next_wakeup)) {
2978 schedule_timeout_interruptible(next_wakeup - cur);
2980 if (kthread_should_stop()) {
2981 ext4_clear_request_list();
2988 * It looks like the request list is empty, but we need
2989 * to check it under the li_list_mtx lock, to prevent any
2990 * additions into it, and of course we should lock ext4_li_mtx
2991 * to atomically free the list and ext4_li_info, because at
2992 * this point another ext4 filesystem could be registering
2995 mutex_lock(&ext4_li_mtx);
2996 mutex_lock(&eli->li_list_mtx);
2997 if (!list_empty(&eli->li_request_list)) {
2998 mutex_unlock(&eli->li_list_mtx);
2999 mutex_unlock(&ext4_li_mtx);
3002 mutex_unlock(&eli->li_list_mtx);
3003 kfree(ext4_li_info);
3004 ext4_li_info = NULL;
3005 mutex_unlock(&ext4_li_mtx);
3010 static void ext4_clear_request_list(void)
3012 struct list_head *pos, *n;
3013 struct ext4_li_request *elr;
3015 mutex_lock(&ext4_li_info->li_list_mtx);
3016 list_for_each_safe(pos, n, &ext4_li_info->li_request_list) {
3017 elr = list_entry(pos, struct ext4_li_request,
3019 ext4_remove_li_request(elr);
3021 mutex_unlock(&ext4_li_info->li_list_mtx);
3024 static int ext4_run_lazyinit_thread(void)
3026 ext4_lazyinit_task = kthread_run(ext4_lazyinit_thread,
3027 ext4_li_info, "ext4lazyinit");
3028 if (IS_ERR(ext4_lazyinit_task)) {
3029 int err = PTR_ERR(ext4_lazyinit_task);
3030 ext4_clear_request_list();
3031 kfree(ext4_li_info);
3032 ext4_li_info = NULL;
3033 printk(KERN_CRIT "EXT4-fs: error %d creating inode table "
3034 "initialization thread\n",
3038 ext4_li_info->li_state |= EXT4_LAZYINIT_RUNNING;
3043 * Check whether it make sense to run itable init. thread or not.
3044 * If there is at least one uninitialized inode table, return
3045 * corresponding group number, else the loop goes through all
3046 * groups and return total number of groups.
3048 static ext4_group_t ext4_has_uninit_itable(struct super_block *sb)
3050 ext4_group_t group, ngroups = EXT4_SB(sb)->s_groups_count;
3051 struct ext4_group_desc *gdp = NULL;
3053 for (group = 0; group < ngroups; group++) {
3054 gdp = ext4_get_group_desc(sb, group, NULL);
3058 if (!(gdp->bg_flags & cpu_to_le16(EXT4_BG_INODE_ZEROED)))
3065 static int ext4_li_info_new(void)
3067 struct ext4_lazy_init *eli = NULL;
3069 eli = kzalloc(sizeof(*eli), GFP_KERNEL);
3073 INIT_LIST_HEAD(&eli->li_request_list);
3074 mutex_init(&eli->li_list_mtx);
3076 eli->li_state |= EXT4_LAZYINIT_QUIT;
3083 static struct ext4_li_request *ext4_li_request_new(struct super_block *sb,
3086 struct ext4_sb_info *sbi = EXT4_SB(sb);
3087 struct ext4_li_request *elr;
3089 elr = kzalloc(sizeof(*elr), GFP_KERNEL);
3095 elr->lr_next_group = start;
3098 * Randomize first schedule time of the request to
3099 * spread the inode table initialization requests
3102 elr->lr_next_sched = jiffies + (prandom_u32() %
3103 (EXT4_DEF_LI_MAX_START_DELAY * HZ));
3107 int ext4_register_li_request(struct super_block *sb,
3108 ext4_group_t first_not_zeroed)
3110 struct ext4_sb_info *sbi = EXT4_SB(sb);
3111 struct ext4_li_request *elr = NULL;
3112 ext4_group_t ngroups = EXT4_SB(sb)->s_groups_count;
3115 mutex_lock(&ext4_li_mtx);
3116 if (sbi->s_li_request != NULL) {
3118 * Reset timeout so it can be computed again, because
3119 * s_li_wait_mult might have changed.
3121 sbi->s_li_request->lr_timeout = 0;
3125 if (first_not_zeroed == ngroups ||
3126 (sb->s_flags & MS_RDONLY) ||
3127 !test_opt(sb, INIT_INODE_TABLE))
3130 elr = ext4_li_request_new(sb, first_not_zeroed);
3136 if (NULL == ext4_li_info) {
3137 ret = ext4_li_info_new();
3142 mutex_lock(&ext4_li_info->li_list_mtx);
3143 list_add(&elr->lr_request, &ext4_li_info->li_request_list);
3144 mutex_unlock(&ext4_li_info->li_list_mtx);
3146 sbi->s_li_request = elr;
3148 * set elr to NULL here since it has been inserted to
3149 * the request_list and the removal and free of it is
3150 * handled by ext4_clear_request_list from now on.
3154 if (!(ext4_li_info->li_state & EXT4_LAZYINIT_RUNNING)) {
3155 ret = ext4_run_lazyinit_thread();
3160 mutex_unlock(&ext4_li_mtx);
3167 * We do not need to lock anything since this is called on
3170 static void ext4_destroy_lazyinit_thread(void)
3173 * If thread exited earlier
3174 * there's nothing to be done.
3176 if (!ext4_li_info || !ext4_lazyinit_task)
3179 kthread_stop(ext4_lazyinit_task);
3182 static int set_journal_csum_feature_set(struct super_block *sb)
3185 int compat, incompat;
3186 struct ext4_sb_info *sbi = EXT4_SB(sb);
3188 if (EXT4_HAS_RO_COMPAT_FEATURE(sb,
3189 EXT4_FEATURE_RO_COMPAT_METADATA_CSUM)) {
3190 /* journal checksum v2 */
3192 incompat = JBD2_FEATURE_INCOMPAT_CSUM_V2;
3194 /* journal checksum v1 */
3195 compat = JBD2_FEATURE_COMPAT_CHECKSUM;
3199 if (test_opt(sb, JOURNAL_ASYNC_COMMIT)) {
3200 ret = jbd2_journal_set_features(sbi->s_journal,
3202 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT |
3204 } else if (test_opt(sb, JOURNAL_CHECKSUM)) {
3205 ret = jbd2_journal_set_features(sbi->s_journal,
3208 jbd2_journal_clear_features(sbi->s_journal, 0, 0,
3209 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT);
3211 jbd2_journal_clear_features(sbi->s_journal,
3212 JBD2_FEATURE_COMPAT_CHECKSUM, 0,
3213 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT |
3214 JBD2_FEATURE_INCOMPAT_CSUM_V2);
3221 * Note: calculating the overhead so we can be compatible with
3222 * historical BSD practice is quite difficult in the face of
3223 * clusters/bigalloc. This is because multiple metadata blocks from
3224 * different block group can end up in the same allocation cluster.
3225 * Calculating the exact overhead in the face of clustered allocation
3226 * requires either O(all block bitmaps) in memory or O(number of block
3227 * groups**2) in time. We will still calculate the superblock for
3228 * older file systems --- and if we come across with a bigalloc file
3229 * system with zero in s_overhead_clusters the estimate will be close to
3230 * correct especially for very large cluster sizes --- but for newer
3231 * file systems, it's better to calculate this figure once at mkfs
3232 * time, and store it in the superblock. If the superblock value is
3233 * present (even for non-bigalloc file systems), we will use it.
3235 static int count_overhead(struct super_block *sb, ext4_group_t grp,
3238 struct ext4_sb_info *sbi = EXT4_SB(sb);
3239 struct ext4_group_desc *gdp;
3240 ext4_fsblk_t first_block, last_block, b;
3241 ext4_group_t i, ngroups = ext4_get_groups_count(sb);
3242 int s, j, count = 0;
3244 if (!EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_BIGALLOC))
3245 return (ext4_bg_has_super(sb, grp) + ext4_bg_num_gdb(sb, grp) +
3246 sbi->s_itb_per_group + 2);
3248 first_block = le32_to_cpu(sbi->s_es->s_first_data_block) +
3249 (grp * EXT4_BLOCKS_PER_GROUP(sb));
3250 last_block = first_block + EXT4_BLOCKS_PER_GROUP(sb) - 1;
3251 for (i = 0; i < ngroups; i++) {
3252 gdp = ext4_get_group_desc(sb, i, NULL);
3253 b = ext4_block_bitmap(sb, gdp);
3254 if (b >= first_block && b <= last_block) {
3255 ext4_set_bit(EXT4_B2C(sbi, b - first_block), buf);
3258 b = ext4_inode_bitmap(sb, gdp);
3259 if (b >= first_block && b <= last_block) {
3260 ext4_set_bit(EXT4_B2C(sbi, b - first_block), buf);
3263 b = ext4_inode_table(sb, gdp);
3264 if (b >= first_block && b + sbi->s_itb_per_group <= last_block)
3265 for (j = 0; j < sbi->s_itb_per_group; j++, b++) {
3266 int c = EXT4_B2C(sbi, b - first_block);
3267 ext4_set_bit(c, buf);
3273 if (ext4_bg_has_super(sb, grp)) {
3274 ext4_set_bit(s++, buf);
3277 for (j = ext4_bg_num_gdb(sb, grp); j > 0; j--) {
3278 ext4_set_bit(EXT4_B2C(sbi, s++), buf);
3284 return EXT4_CLUSTERS_PER_GROUP(sb) -
3285 ext4_count_free(buf, EXT4_CLUSTERS_PER_GROUP(sb) / 8);
3289 * Compute the overhead and stash it in sbi->s_overhead
3291 int ext4_calculate_overhead(struct super_block *sb)
3293 struct ext4_sb_info *sbi = EXT4_SB(sb);
3294 struct ext4_super_block *es = sbi->s_es;
3295 ext4_group_t i, ngroups = ext4_get_groups_count(sb);
3296 ext4_fsblk_t overhead = 0;
3297 char *buf = (char *) get_zeroed_page(GFP_KERNEL);
3303 * Compute the overhead (FS structures). This is constant
3304 * for a given filesystem unless the number of block groups
3305 * changes so we cache the previous value until it does.
3309 * All of the blocks before first_data_block are overhead
3311 overhead = EXT4_B2C(sbi, le32_to_cpu(es->s_first_data_block));
3314 * Add the overhead found in each block group
3316 for (i = 0; i < ngroups; i++) {
3319 blks = count_overhead(sb, i, buf);
3322 memset(buf, 0, PAGE_SIZE);
3325 /* Add the journal blocks as well */
3327 overhead += EXT4_NUM_B2C(sbi, sbi->s_journal->j_maxlen);
3329 sbi->s_overhead = overhead;
3331 free_page((unsigned long) buf);
3336 static ext4_fsblk_t ext4_calculate_resv_clusters(struct super_block *sb)
3338 ext4_fsblk_t resv_clusters;
3341 * There's no need to reserve anything when we aren't using extents.
3342 * The space estimates are exact, there are no unwritten extents,
3343 * hole punching doesn't need new metadata... This is needed especially
3344 * to keep ext2/3 backward compatibility.
3346 if (!EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_EXTENTS))
3349 * By default we reserve 2% or 4096 clusters, whichever is smaller.
3350 * This should cover the situations where we can not afford to run
3351 * out of space like for example punch hole, or converting
3352 * unwritten extents in delalloc path. In most cases such
3353 * allocation would require 1, or 2 blocks, higher numbers are
3356 resv_clusters = ext4_blocks_count(EXT4_SB(sb)->s_es) >>
3357 EXT4_SB(sb)->s_cluster_bits;
3359 do_div(resv_clusters, 50);
3360 resv_clusters = min_t(ext4_fsblk_t, resv_clusters, 4096);
3362 return resv_clusters;
3366 static int ext4_reserve_clusters(struct ext4_sb_info *sbi, ext4_fsblk_t count)
3368 ext4_fsblk_t clusters = ext4_blocks_count(sbi->s_es) >>
3369 sbi->s_cluster_bits;
3371 if (count >= clusters)
3374 atomic64_set(&sbi->s_resv_clusters, count);
3378 static int ext4_fill_super(struct super_block *sb, void *data, int silent)
3380 char *orig_data = kstrdup(data, GFP_KERNEL);
3381 struct buffer_head *bh;
3382 struct ext4_super_block *es = NULL;
3383 struct ext4_sb_info *sbi;
3385 ext4_fsblk_t sb_block = get_sb_block(&data);
3386 ext4_fsblk_t logical_sb_block;
3387 unsigned long offset = 0;
3388 unsigned long journal_devnum = 0;
3389 unsigned long def_mount_opts;
3394 int blocksize, clustersize;
3395 unsigned int db_count;
3397 int needs_recovery, has_huge_files, has_bigalloc;
3400 unsigned int journal_ioprio = DEFAULT_JOURNAL_IOPRIO;
3401 ext4_group_t first_not_zeroed;
3403 sbi = kzalloc(sizeof(*sbi), GFP_KERNEL);
3407 sbi->s_blockgroup_lock =
3408 kzalloc(sizeof(struct blockgroup_lock), GFP_KERNEL);
3409 if (!sbi->s_blockgroup_lock) {
3413 sb->s_fs_info = sbi;
3415 sbi->s_inode_readahead_blks = EXT4_DEF_INODE_READAHEAD_BLKS;
3416 sbi->s_sb_block = sb_block;
3417 if (sb->s_bdev->bd_part)
3418 sbi->s_sectors_written_start =
3419 part_stat_read(sb->s_bdev->bd_part, sectors[1]);
3421 /* Cleanup superblock name */
3422 for (cp = sb->s_id; (cp = strchr(cp, '/'));)
3425 /* -EINVAL is default */
3427 blocksize = sb_min_blocksize(sb, EXT4_MIN_BLOCK_SIZE);
3429 ext4_msg(sb, KERN_ERR, "unable to set blocksize");
3434 * The ext4 superblock will not be buffer aligned for other than 1kB
3435 * block sizes. We need to calculate the offset from buffer start.
3437 if (blocksize != EXT4_MIN_BLOCK_SIZE) {
3438 logical_sb_block = sb_block * EXT4_MIN_BLOCK_SIZE;
3439 offset = do_div(logical_sb_block, blocksize);
3441 logical_sb_block = sb_block;
3444 if (!(bh = sb_bread(sb, logical_sb_block))) {
3445 ext4_msg(sb, KERN_ERR, "unable to read superblock");
3449 * Note: s_es must be initialized as soon as possible because
3450 * some ext4 macro-instructions depend on its value
3452 es = (struct ext4_super_block *) (bh->b_data + offset);
3454 sb->s_magic = le16_to_cpu(es->s_magic);
3455 if (sb->s_magic != EXT4_SUPER_MAGIC)
3457 sbi->s_kbytes_written = le64_to_cpu(es->s_kbytes_written);
3459 /* Warn if metadata_csum and gdt_csum are both set. */
3460 if (EXT4_HAS_RO_COMPAT_FEATURE(sb,
3461 EXT4_FEATURE_RO_COMPAT_METADATA_CSUM) &&
3462 EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_GDT_CSUM))
3463 ext4_warning(sb, KERN_INFO "metadata_csum and uninit_bg are "
3464 "redundant flags; please run fsck.");
3466 /* Check for a known checksum algorithm */
3467 if (!ext4_verify_csum_type(sb, es)) {
3468 ext4_msg(sb, KERN_ERR, "VFS: Found ext4 filesystem with "
3469 "unknown checksum algorithm.");
3474 /* Load the checksum driver */
3475 if (EXT4_HAS_RO_COMPAT_FEATURE(sb,
3476 EXT4_FEATURE_RO_COMPAT_METADATA_CSUM)) {
3477 sbi->s_chksum_driver = crypto_alloc_shash("crc32c", 0, 0);
3478 if (IS_ERR(sbi->s_chksum_driver)) {
3479 ext4_msg(sb, KERN_ERR, "Cannot load crc32c driver.");
3480 ret = PTR_ERR(sbi->s_chksum_driver);
3481 sbi->s_chksum_driver = NULL;
3486 /* Check superblock checksum */
3487 if (!ext4_superblock_csum_verify(sb, es)) {
3488 ext4_msg(sb, KERN_ERR, "VFS: Found ext4 filesystem with "
3489 "invalid superblock checksum. Run e2fsck?");
3494 /* Precompute checksum seed for all metadata */
3495 if (EXT4_HAS_RO_COMPAT_FEATURE(sb,
3496 EXT4_FEATURE_RO_COMPAT_METADATA_CSUM))
3497 sbi->s_csum_seed = ext4_chksum(sbi, ~0, es->s_uuid,
3498 sizeof(es->s_uuid));
3500 /* Set defaults before we parse the mount options */
3501 def_mount_opts = le32_to_cpu(es->s_default_mount_opts);
3502 set_opt(sb, INIT_INODE_TABLE);
3503 if (def_mount_opts & EXT4_DEFM_DEBUG)
3505 if (def_mount_opts & EXT4_DEFM_BSDGROUPS)
3507 if (def_mount_opts & EXT4_DEFM_UID16)
3508 set_opt(sb, NO_UID32);
3509 /* xattr user namespace & acls are now defaulted on */
3510 set_opt(sb, XATTR_USER);
3511 #ifdef CONFIG_EXT4_FS_POSIX_ACL
3512 set_opt(sb, POSIX_ACL);
3514 if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_DATA)
3515 set_opt(sb, JOURNAL_DATA);
3516 else if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_ORDERED)
3517 set_opt(sb, ORDERED_DATA);
3518 else if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_WBACK)
3519 set_opt(sb, WRITEBACK_DATA);
3521 if (le16_to_cpu(sbi->s_es->s_errors) == EXT4_ERRORS_PANIC)
3522 set_opt(sb, ERRORS_PANIC);
3523 else if (le16_to_cpu(sbi->s_es->s_errors) == EXT4_ERRORS_CONTINUE)
3524 set_opt(sb, ERRORS_CONT);
3526 set_opt(sb, ERRORS_RO);
3527 if (def_mount_opts & EXT4_DEFM_BLOCK_VALIDITY)
3528 set_opt(sb, BLOCK_VALIDITY);
3529 if (def_mount_opts & EXT4_DEFM_DISCARD)
3530 set_opt(sb, DISCARD);
3532 sbi->s_resuid = make_kuid(&init_user_ns, le16_to_cpu(es->s_def_resuid));
3533 sbi->s_resgid = make_kgid(&init_user_ns, le16_to_cpu(es->s_def_resgid));
3534 sbi->s_commit_interval = JBD2_DEFAULT_MAX_COMMIT_AGE * HZ;
3535 sbi->s_min_batch_time = EXT4_DEF_MIN_BATCH_TIME;
3536 sbi->s_max_batch_time = EXT4_DEF_MAX_BATCH_TIME;
3538 if ((def_mount_opts & EXT4_DEFM_NOBARRIER) == 0)
3539 set_opt(sb, BARRIER);
3542 * enable delayed allocation by default
3543 * Use -o nodelalloc to turn it off
3545 if (!IS_EXT3_SB(sb) && !IS_EXT2_SB(sb) &&
3546 ((def_mount_opts & EXT4_DEFM_NODELALLOC) == 0))
3547 set_opt(sb, DELALLOC);
3550 * set default s_li_wait_mult for lazyinit, for the case there is
3551 * no mount option specified.
3553 sbi->s_li_wait_mult = EXT4_DEF_LI_WAIT_MULT;
3555 if (!parse_options((char *) sbi->s_es->s_mount_opts, sb,
3556 &journal_devnum, &journal_ioprio, 0)) {
3557 ext4_msg(sb, KERN_WARNING,
3558 "failed to parse options in superblock: %s",
3559 sbi->s_es->s_mount_opts);
3561 sbi->s_def_mount_opt = sbi->s_mount_opt;
3562 if (!parse_options((char *) data, sb, &journal_devnum,
3563 &journal_ioprio, 0))
3566 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA) {
3567 printk_once(KERN_WARNING "EXT4-fs: Warning: mounting "
3568 "with data=journal disables delayed "
3569 "allocation and O_DIRECT support!\n");
3570 if (test_opt2(sb, EXPLICIT_DELALLOC)) {
3571 ext4_msg(sb, KERN_ERR, "can't mount with "
3572 "both data=journal and delalloc");
3575 if (test_opt(sb, DIOREAD_NOLOCK)) {
3576 ext4_msg(sb, KERN_ERR, "can't mount with "
3577 "both data=journal and dioread_nolock");
3580 if (test_opt(sb, DELALLOC))
3581 clear_opt(sb, DELALLOC);
3584 sb->s_flags = (sb->s_flags & ~MS_POSIXACL) |
3585 (test_opt(sb, POSIX_ACL) ? MS_POSIXACL : 0);
3587 if (le32_to_cpu(es->s_rev_level) == EXT4_GOOD_OLD_REV &&
3588 (EXT4_HAS_COMPAT_FEATURE(sb, ~0U) ||
3589 EXT4_HAS_RO_COMPAT_FEATURE(sb, ~0U) ||
3590 EXT4_HAS_INCOMPAT_FEATURE(sb, ~0U)))
3591 ext4_msg(sb, KERN_WARNING,
3592 "feature flags set on rev 0 fs, "
3593 "running e2fsck is recommended");
3595 if (es->s_creator_os == cpu_to_le32(EXT4_OS_HURD)) {
3596 set_opt2(sb, HURD_COMPAT);
3597 if (EXT4_HAS_INCOMPAT_FEATURE(sb,
3598 EXT4_FEATURE_INCOMPAT_64BIT)) {
3599 ext4_msg(sb, KERN_ERR,
3600 "The Hurd can't support 64-bit file systems");
3605 if (IS_EXT2_SB(sb)) {
3606 if (ext2_feature_set_ok(sb))
3607 ext4_msg(sb, KERN_INFO, "mounting ext2 file system "
3608 "using the ext4 subsystem");
3610 ext4_msg(sb, KERN_ERR, "couldn't mount as ext2 due "
3611 "to feature incompatibilities");
3616 if (IS_EXT3_SB(sb)) {
3617 if (ext3_feature_set_ok(sb))
3618 ext4_msg(sb, KERN_INFO, "mounting ext3 file system "
3619 "using the ext4 subsystem");
3621 ext4_msg(sb, KERN_ERR, "couldn't mount as ext3 due "
3622 "to feature incompatibilities");
3628 * Check feature flags regardless of the revision level, since we
3629 * previously didn't change the revision level when setting the flags,
3630 * so there is a chance incompat flags are set on a rev 0 filesystem.
3632 if (!ext4_feature_set_ok(sb, (sb->s_flags & MS_RDONLY)))
3635 blocksize = BLOCK_SIZE << le32_to_cpu(es->s_log_block_size);
3636 if (blocksize < EXT4_MIN_BLOCK_SIZE ||
3637 blocksize > EXT4_MAX_BLOCK_SIZE) {
3638 ext4_msg(sb, KERN_ERR,
3639 "Unsupported filesystem blocksize %d", blocksize);
3643 if (sb->s_blocksize != blocksize) {
3644 /* Validate the filesystem blocksize */
3645 if (!sb_set_blocksize(sb, blocksize)) {
3646 ext4_msg(sb, KERN_ERR, "bad block size %d",
3652 logical_sb_block = sb_block * EXT4_MIN_BLOCK_SIZE;
3653 offset = do_div(logical_sb_block, blocksize);
3654 bh = sb_bread(sb, logical_sb_block);
3656 ext4_msg(sb, KERN_ERR,
3657 "Can't read superblock on 2nd try");
3660 es = (struct ext4_super_block *)(bh->b_data + offset);
3662 if (es->s_magic != cpu_to_le16(EXT4_SUPER_MAGIC)) {
3663 ext4_msg(sb, KERN_ERR,
3664 "Magic mismatch, very weird!");
3669 has_huge_files = EXT4_HAS_RO_COMPAT_FEATURE(sb,
3670 EXT4_FEATURE_RO_COMPAT_HUGE_FILE);
3671 sbi->s_bitmap_maxbytes = ext4_max_bitmap_size(sb->s_blocksize_bits,
3673 sb->s_maxbytes = ext4_max_size(sb->s_blocksize_bits, has_huge_files);
3675 if (le32_to_cpu(es->s_rev_level) == EXT4_GOOD_OLD_REV) {
3676 sbi->s_inode_size = EXT4_GOOD_OLD_INODE_SIZE;
3677 sbi->s_first_ino = EXT4_GOOD_OLD_FIRST_INO;
3679 sbi->s_inode_size = le16_to_cpu(es->s_inode_size);
3680 sbi->s_first_ino = le32_to_cpu(es->s_first_ino);
3681 if ((sbi->s_inode_size < EXT4_GOOD_OLD_INODE_SIZE) ||
3682 (!is_power_of_2(sbi->s_inode_size)) ||
3683 (sbi->s_inode_size > blocksize)) {
3684 ext4_msg(sb, KERN_ERR,
3685 "unsupported inode size: %d",
3689 if (sbi->s_inode_size > EXT4_GOOD_OLD_INODE_SIZE)
3690 sb->s_time_gran = 1 << (EXT4_EPOCH_BITS - 2);
3693 sbi->s_desc_size = le16_to_cpu(es->s_desc_size);
3694 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_64BIT)) {
3695 if (sbi->s_desc_size < EXT4_MIN_DESC_SIZE_64BIT ||
3696 sbi->s_desc_size > EXT4_MAX_DESC_SIZE ||
3697 !is_power_of_2(sbi->s_desc_size)) {
3698 ext4_msg(sb, KERN_ERR,
3699 "unsupported descriptor size %lu",
3704 sbi->s_desc_size = EXT4_MIN_DESC_SIZE;
3706 sbi->s_blocks_per_group = le32_to_cpu(es->s_blocks_per_group);
3707 sbi->s_inodes_per_group = le32_to_cpu(es->s_inodes_per_group);
3708 if (EXT4_INODE_SIZE(sb) == 0 || EXT4_INODES_PER_GROUP(sb) == 0)
3711 sbi->s_inodes_per_block = blocksize / EXT4_INODE_SIZE(sb);
3712 if (sbi->s_inodes_per_block == 0)
3714 sbi->s_itb_per_group = sbi->s_inodes_per_group /
3715 sbi->s_inodes_per_block;
3716 sbi->s_desc_per_block = blocksize / EXT4_DESC_SIZE(sb);
3718 sbi->s_mount_state = le16_to_cpu(es->s_state);
3719 sbi->s_addr_per_block_bits = ilog2(EXT4_ADDR_PER_BLOCK(sb));
3720 sbi->s_desc_per_block_bits = ilog2(EXT4_DESC_PER_BLOCK(sb));
3722 for (i = 0; i < 4; i++)
3723 sbi->s_hash_seed[i] = le32_to_cpu(es->s_hash_seed[i]);
3724 sbi->s_def_hash_version = es->s_def_hash_version;
3725 if (EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_DIR_INDEX)) {
3726 i = le32_to_cpu(es->s_flags);
3727 if (i & EXT2_FLAGS_UNSIGNED_HASH)
3728 sbi->s_hash_unsigned = 3;
3729 else if ((i & EXT2_FLAGS_SIGNED_HASH) == 0) {
3730 #ifdef __CHAR_UNSIGNED__
3731 if (!(sb->s_flags & MS_RDONLY))
3733 cpu_to_le32(EXT2_FLAGS_UNSIGNED_HASH);
3734 sbi->s_hash_unsigned = 3;
3736 if (!(sb->s_flags & MS_RDONLY))
3738 cpu_to_le32(EXT2_FLAGS_SIGNED_HASH);
3743 /* Handle clustersize */
3744 clustersize = BLOCK_SIZE << le32_to_cpu(es->s_log_cluster_size);
3745 has_bigalloc = EXT4_HAS_RO_COMPAT_FEATURE(sb,
3746 EXT4_FEATURE_RO_COMPAT_BIGALLOC);
3748 if (clustersize < blocksize) {
3749 ext4_msg(sb, KERN_ERR,
3750 "cluster size (%d) smaller than "
3751 "block size (%d)", clustersize, blocksize);
3754 sbi->s_cluster_bits = le32_to_cpu(es->s_log_cluster_size) -
3755 le32_to_cpu(es->s_log_block_size);
3756 sbi->s_clusters_per_group =
3757 le32_to_cpu(es->s_clusters_per_group);
3758 if (sbi->s_clusters_per_group > blocksize * 8) {
3759 ext4_msg(sb, KERN_ERR,
3760 "#clusters per group too big: %lu",
3761 sbi->s_clusters_per_group);
3764 if (sbi->s_blocks_per_group !=
3765 (sbi->s_clusters_per_group * (clustersize / blocksize))) {
3766 ext4_msg(sb, KERN_ERR, "blocks per group (%lu) and "
3767 "clusters per group (%lu) inconsistent",
3768 sbi->s_blocks_per_group,
3769 sbi->s_clusters_per_group);
3773 if (clustersize != blocksize) {
3774 ext4_warning(sb, "fragment/cluster size (%d) != "
3775 "block size (%d)", clustersize,
3777 clustersize = blocksize;
3779 if (sbi->s_blocks_per_group > blocksize * 8) {
3780 ext4_msg(sb, KERN_ERR,
3781 "#blocks per group too big: %lu",
3782 sbi->s_blocks_per_group);
3785 sbi->s_clusters_per_group = sbi->s_blocks_per_group;
3786 sbi->s_cluster_bits = 0;
3788 sbi->s_cluster_ratio = clustersize / blocksize;
3790 if (sbi->s_inodes_per_group > blocksize * 8) {
3791 ext4_msg(sb, KERN_ERR,
3792 "#inodes per group too big: %lu",
3793 sbi->s_inodes_per_group);
3797 /* Do we have standard group size of clustersize * 8 blocks ? */
3798 if (sbi->s_blocks_per_group == clustersize << 3)
3799 set_opt2(sb, STD_GROUP_SIZE);
3802 * Test whether we have more sectors than will fit in sector_t,
3803 * and whether the max offset is addressable by the page cache.
3805 err = generic_check_addressable(sb->s_blocksize_bits,
3806 ext4_blocks_count(es));
3808 ext4_msg(sb, KERN_ERR, "filesystem"
3809 " too large to mount safely on this system");
3810 if (sizeof(sector_t) < 8)
3811 ext4_msg(sb, KERN_WARNING, "CONFIG_LBDAF not enabled");
3815 if (EXT4_BLOCKS_PER_GROUP(sb) == 0)
3818 /* check blocks count against device size */
3819 blocks_count = sb->s_bdev->bd_inode->i_size >> sb->s_blocksize_bits;
3820 if (blocks_count && ext4_blocks_count(es) > blocks_count) {
3821 ext4_msg(sb, KERN_WARNING, "bad geometry: block count %llu "
3822 "exceeds size of device (%llu blocks)",
3823 ext4_blocks_count(es), blocks_count);
3828 * It makes no sense for the first data block to be beyond the end
3829 * of the filesystem.
3831 if (le32_to_cpu(es->s_first_data_block) >= ext4_blocks_count(es)) {
3832 ext4_msg(sb, KERN_WARNING, "bad geometry: first data "
3833 "block %u is beyond end of filesystem (%llu)",
3834 le32_to_cpu(es->s_first_data_block),
3835 ext4_blocks_count(es));
3838 blocks_count = (ext4_blocks_count(es) -
3839 le32_to_cpu(es->s_first_data_block) +
3840 EXT4_BLOCKS_PER_GROUP(sb) - 1);
3841 do_div(blocks_count, EXT4_BLOCKS_PER_GROUP(sb));
3842 if (blocks_count > ((uint64_t)1<<32) - EXT4_DESC_PER_BLOCK(sb)) {
3843 ext4_msg(sb, KERN_WARNING, "groups count too large: %u "
3844 "(block count %llu, first data block %u, "
3845 "blocks per group %lu)", sbi->s_groups_count,
3846 ext4_blocks_count(es),
3847 le32_to_cpu(es->s_first_data_block),
3848 EXT4_BLOCKS_PER_GROUP(sb));
3851 sbi->s_groups_count = blocks_count;
3852 sbi->s_blockfile_groups = min_t(ext4_group_t, sbi->s_groups_count,
3853 (EXT4_MAX_BLOCK_FILE_PHYS / EXT4_BLOCKS_PER_GROUP(sb)));
3854 db_count = (sbi->s_groups_count + EXT4_DESC_PER_BLOCK(sb) - 1) /
3855 EXT4_DESC_PER_BLOCK(sb);
3856 sbi->s_group_desc = ext4_kvmalloc(db_count *
3857 sizeof(struct buffer_head *),
3859 if (sbi->s_group_desc == NULL) {
3860 ext4_msg(sb, KERN_ERR, "not enough memory");
3866 sbi->s_proc = proc_mkdir(sb->s_id, ext4_proc_root);
3869 proc_create_data("options", S_IRUGO, sbi->s_proc,
3870 &ext4_seq_options_fops, sb);
3872 bgl_lock_init(sbi->s_blockgroup_lock);
3874 for (i = 0; i < db_count; i++) {
3875 block = descriptor_loc(sb, logical_sb_block, i);
3876 sbi->s_group_desc[i] = sb_bread(sb, block);
3877 if (!sbi->s_group_desc[i]) {
3878 ext4_msg(sb, KERN_ERR,
3879 "can't read group descriptor %d", i);
3886 * set up enough so that it can read an inode,
3887 * and create new inode for buddy allocator
3889 sbi->s_gdb_count = db_count;
3890 if (!test_opt(sb, NOLOAD) &&
3891 EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL))
3892 sb->s_op = &ext4_sops;
3894 sb->s_op = &ext4_nojournal_sops;
3897 err = ext4_mb_init(sb);
3899 ext4_msg(sb, KERN_ERR, "failed to initialize mballoc (%d)",
3904 if (!ext4_check_descriptors(sb, &first_not_zeroed)) {
3905 ext4_msg(sb, KERN_ERR, "group descriptors corrupted!");
3906 goto failed_mount2a;
3908 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_FLEX_BG))
3909 if (!ext4_fill_flex_info(sb)) {
3910 ext4_msg(sb, KERN_ERR,
3911 "unable to initialize "
3912 "flex_bg meta info!");
3913 goto failed_mount2a;
3916 get_random_bytes(&sbi->s_next_generation, sizeof(u32));
3917 spin_lock_init(&sbi->s_next_gen_lock);
3919 init_timer(&sbi->s_err_report);
3920 sbi->s_err_report.function = print_daily_error_info;
3921 sbi->s_err_report.data = (unsigned long) sb;
3923 /* Register extent status tree shrinker */
3924 ext4_es_register_shrinker(sbi);
3926 err = percpu_counter_init(&sbi->s_freeclusters_counter,
3927 ext4_count_free_clusters(sb));
3929 err = percpu_counter_init(&sbi->s_freeinodes_counter,
3930 ext4_count_free_inodes(sb));
3933 err = percpu_counter_init(&sbi->s_dirs_counter,
3934 ext4_count_dirs(sb));
3937 err = percpu_counter_init(&sbi->s_dirtyclusters_counter, 0);
3940 err = percpu_counter_init(&sbi->s_extent_cache_cnt, 0);
3943 ext4_msg(sb, KERN_ERR, "insufficient memory");
3947 sbi->s_stripe = ext4_get_stripe_size(sbi);
3948 sbi->s_extent_max_zeroout_kb = 32;
3950 sb->s_export_op = &ext4_export_ops;
3951 sb->s_xattr = ext4_xattr_handlers;
3953 sb->dq_op = &ext4_quota_operations;
3954 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_QUOTA))
3955 sb->s_qcop = &ext4_qctl_sysfile_operations;
3957 sb->s_qcop = &ext4_qctl_operations;
3959 memcpy(sb->s_uuid, es->s_uuid, sizeof(es->s_uuid));
3961 INIT_LIST_HEAD(&sbi->s_orphan); /* unlinked but open files */
3962 mutex_init(&sbi->s_orphan_lock);
3966 needs_recovery = (es->s_last_orphan != 0 ||
3967 EXT4_HAS_INCOMPAT_FEATURE(sb,
3968 EXT4_FEATURE_INCOMPAT_RECOVER));
3970 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_MMP) &&
3971 !(sb->s_flags & MS_RDONLY))
3972 if (ext4_multi_mount_protect(sb, le64_to_cpu(es->s_mmp_block)))
3976 * The first inode we look at is the journal inode. Don't try
3977 * root first: it may be modified in the journal!
3979 if (!test_opt(sb, NOLOAD) &&
3980 EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL)) {
3981 if (ext4_load_journal(sb, es, journal_devnum))
3983 } else if (test_opt(sb, NOLOAD) && !(sb->s_flags & MS_RDONLY) &&
3984 EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER)) {
3985 ext4_msg(sb, KERN_ERR, "required journal recovery "
3986 "suppressed and not mounted read-only");
3987 goto failed_mount_wq;
3989 clear_opt(sb, DATA_FLAGS);
3990 sbi->s_journal = NULL;
3995 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_64BIT) &&
3996 !jbd2_journal_set_features(EXT4_SB(sb)->s_journal, 0, 0,
3997 JBD2_FEATURE_INCOMPAT_64BIT)) {
3998 ext4_msg(sb, KERN_ERR, "Failed to set 64-bit journal feature");
3999 goto failed_mount_wq;
4002 if (!set_journal_csum_feature_set(sb)) {
4003 ext4_msg(sb, KERN_ERR, "Failed to set journal checksum "
4005 goto failed_mount_wq;
4008 /* We have now updated the journal if required, so we can
4009 * validate the data journaling mode. */
4010 switch (test_opt(sb, DATA_FLAGS)) {
4012 /* No mode set, assume a default based on the journal
4013 * capabilities: ORDERED_DATA if the journal can
4014 * cope, else JOURNAL_DATA
4016 if (jbd2_journal_check_available_features
4017 (sbi->s_journal, 0, 0, JBD2_FEATURE_INCOMPAT_REVOKE))
4018 set_opt(sb, ORDERED_DATA);
4020 set_opt(sb, JOURNAL_DATA);
4023 case EXT4_MOUNT_ORDERED_DATA:
4024 case EXT4_MOUNT_WRITEBACK_DATA:
4025 if (!jbd2_journal_check_available_features
4026 (sbi->s_journal, 0, 0, JBD2_FEATURE_INCOMPAT_REVOKE)) {
4027 ext4_msg(sb, KERN_ERR, "Journal does not support "
4028 "requested data journaling mode");
4029 goto failed_mount_wq;
4034 set_task_ioprio(sbi->s_journal->j_task, journal_ioprio);
4036 sbi->s_journal->j_commit_callback = ext4_journal_commit_callback;
4039 * The journal may have updated the bg summary counts, so we
4040 * need to update the global counters.
4042 percpu_counter_set(&sbi->s_freeclusters_counter,
4043 ext4_count_free_clusters(sb));
4044 percpu_counter_set(&sbi->s_freeinodes_counter,
4045 ext4_count_free_inodes(sb));
4046 percpu_counter_set(&sbi->s_dirs_counter,
4047 ext4_count_dirs(sb));
4048 percpu_counter_set(&sbi->s_dirtyclusters_counter, 0);
4051 if (ext4_mballoc_ready) {
4052 sbi->s_mb_cache = ext4_xattr_create_cache(sb->s_id);
4053 if (!sbi->s_mb_cache) {
4054 ext4_msg(sb, KERN_ERR, "Failed to create an mb_cache");
4055 goto failed_mount_wq;
4060 * Get the # of file system overhead blocks from the
4061 * superblock if present.
4063 if (es->s_overhead_clusters)
4064 sbi->s_overhead = le32_to_cpu(es->s_overhead_clusters);
4066 err = ext4_calculate_overhead(sb);
4068 goto failed_mount_wq;
4072 * The maximum number of concurrent works can be high and
4073 * concurrency isn't really necessary. Limit it to 1.
4075 EXT4_SB(sb)->rsv_conversion_wq =
4076 alloc_workqueue("ext4-rsv-conversion", WQ_MEM_RECLAIM | WQ_UNBOUND, 1);
4077 if (!EXT4_SB(sb)->rsv_conversion_wq) {
4078 printk(KERN_ERR "EXT4-fs: failed to create workqueue\n");
4084 * The jbd2_journal_load will have done any necessary log recovery,
4085 * so we can safely mount the rest of the filesystem now.
4088 root = ext4_iget(sb, EXT4_ROOT_INO);
4090 ext4_msg(sb, KERN_ERR, "get root inode failed");
4091 ret = PTR_ERR(root);
4095 if (!S_ISDIR(root->i_mode) || !root->i_blocks || !root->i_size) {
4096 ext4_msg(sb, KERN_ERR, "corrupt root inode, run e2fsck");
4100 sb->s_root = d_make_root(root);
4102 ext4_msg(sb, KERN_ERR, "get root dentry failed");
4107 if (ext4_setup_super(sb, es, sb->s_flags & MS_RDONLY))
4108 sb->s_flags |= MS_RDONLY;
4110 /* determine the minimum size of new large inodes, if present */
4111 if (sbi->s_inode_size > EXT4_GOOD_OLD_INODE_SIZE) {
4112 sbi->s_want_extra_isize = sizeof(struct ext4_inode) -
4113 EXT4_GOOD_OLD_INODE_SIZE;
4114 if (EXT4_HAS_RO_COMPAT_FEATURE(sb,
4115 EXT4_FEATURE_RO_COMPAT_EXTRA_ISIZE)) {
4116 if (sbi->s_want_extra_isize <
4117 le16_to_cpu(es->s_want_extra_isize))
4118 sbi->s_want_extra_isize =
4119 le16_to_cpu(es->s_want_extra_isize);
4120 if (sbi->s_want_extra_isize <
4121 le16_to_cpu(es->s_min_extra_isize))
4122 sbi->s_want_extra_isize =
4123 le16_to_cpu(es->s_min_extra_isize);
4126 /* Check if enough inode space is available */
4127 if (EXT4_GOOD_OLD_INODE_SIZE + sbi->s_want_extra_isize >
4128 sbi->s_inode_size) {
4129 sbi->s_want_extra_isize = sizeof(struct ext4_inode) -
4130 EXT4_GOOD_OLD_INODE_SIZE;
4131 ext4_msg(sb, KERN_INFO, "required extra inode space not"
4135 err = ext4_reserve_clusters(sbi, ext4_calculate_resv_clusters(sb));
4137 ext4_msg(sb, KERN_ERR, "failed to reserve %llu clusters for "
4138 "reserved pool", ext4_calculate_resv_clusters(sb));
4142 err = ext4_setup_system_zone(sb);
4144 ext4_msg(sb, KERN_ERR, "failed to initialize system "
4149 err = ext4_register_li_request(sb, first_not_zeroed);
4153 sbi->s_kobj.kset = ext4_kset;
4154 init_completion(&sbi->s_kobj_unregister);
4155 err = kobject_init_and_add(&sbi->s_kobj, &ext4_ktype, NULL,
4161 /* Enable quota usage during mount. */
4162 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_QUOTA) &&
4163 !(sb->s_flags & MS_RDONLY)) {
4164 err = ext4_enable_quotas(sb);
4168 #endif /* CONFIG_QUOTA */
4170 EXT4_SB(sb)->s_mount_state |= EXT4_ORPHAN_FS;
4171 ext4_orphan_cleanup(sb, es);
4172 EXT4_SB(sb)->s_mount_state &= ~EXT4_ORPHAN_FS;
4173 if (needs_recovery) {
4174 ext4_msg(sb, KERN_INFO, "recovery complete");
4175 ext4_mark_recovery_complete(sb, es);
4177 if (EXT4_SB(sb)->s_journal) {
4178 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA)
4179 descr = " journalled data mode";
4180 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA)
4181 descr = " ordered data mode";
4183 descr = " writeback data mode";
4185 descr = "out journal";
4187 if (test_opt(sb, DISCARD)) {
4188 struct request_queue *q = bdev_get_queue(sb->s_bdev);
4189 if (!blk_queue_discard(q))
4190 ext4_msg(sb, KERN_WARNING,
4191 "mounting with \"discard\" option, but "
4192 "the device does not support discard");
4195 ext4_msg(sb, KERN_INFO, "mounted filesystem with%s. "
4196 "Opts: %s%s%s", descr, sbi->s_es->s_mount_opts,
4197 *sbi->s_es->s_mount_opts ? "; " : "", orig_data);
4199 if (es->s_error_count)
4200 mod_timer(&sbi->s_err_report, jiffies + 300*HZ); /* 5 minutes */
4202 /* Enable message ratelimiting. Default is 10 messages per 5 secs. */
4203 ratelimit_state_init(&sbi->s_err_ratelimit_state, 5 * HZ, 10);
4204 ratelimit_state_init(&sbi->s_warning_ratelimit_state, 5 * HZ, 10);
4205 ratelimit_state_init(&sbi->s_msg_ratelimit_state, 5 * HZ, 10);
4212 ext4_msg(sb, KERN_ERR, "VFS: Can't find ext4 filesystem");
4217 kobject_del(&sbi->s_kobj);
4220 ext4_unregister_li_request(sb);
4222 ext4_release_system_zone(sb);
4227 ext4_msg(sb, KERN_ERR, "mount failed");
4228 if (EXT4_SB(sb)->rsv_conversion_wq)
4229 destroy_workqueue(EXT4_SB(sb)->rsv_conversion_wq);
4231 if (sbi->s_journal) {
4232 jbd2_journal_destroy(sbi->s_journal);
4233 sbi->s_journal = NULL;
4236 ext4_es_unregister_shrinker(sbi);
4237 del_timer_sync(&sbi->s_err_report);
4238 if (sbi->s_flex_groups)
4239 ext4_kvfree(sbi->s_flex_groups);
4240 percpu_counter_destroy(&sbi->s_freeclusters_counter);
4241 percpu_counter_destroy(&sbi->s_freeinodes_counter);
4242 percpu_counter_destroy(&sbi->s_dirs_counter);
4243 percpu_counter_destroy(&sbi->s_dirtyclusters_counter);
4244 percpu_counter_destroy(&sbi->s_extent_cache_cnt);
4246 kthread_stop(sbi->s_mmp_tsk);
4248 ext4_mb_release(sb);
4250 for (i = 0; i < db_count; i++)
4251 brelse(sbi->s_group_desc[i]);
4252 ext4_kvfree(sbi->s_group_desc);
4254 ext4_ext_release(sb);
4255 if (sbi->s_chksum_driver)
4256 crypto_free_shash(sbi->s_chksum_driver);
4258 remove_proc_entry("options", sbi->s_proc);
4259 remove_proc_entry(sb->s_id, ext4_proc_root);
4262 for (i = 0; i < MAXQUOTAS; i++)
4263 kfree(sbi->s_qf_names[i]);
4265 ext4_blkdev_remove(sbi);
4268 sb->s_fs_info = NULL;
4269 kfree(sbi->s_blockgroup_lock);
4273 return err ? err : ret;
4277 * Setup any per-fs journal parameters now. We'll do this both on
4278 * initial mount, once the journal has been initialised but before we've
4279 * done any recovery; and again on any subsequent remount.
4281 static void ext4_init_journal_params(struct super_block *sb, journal_t *journal)
4283 struct ext4_sb_info *sbi = EXT4_SB(sb);
4285 journal->j_commit_interval = sbi->s_commit_interval;
4286 journal->j_min_batch_time = sbi->s_min_batch_time;
4287 journal->j_max_batch_time = sbi->s_max_batch_time;
4289 write_lock(&journal->j_state_lock);
4290 if (test_opt(sb, BARRIER))
4291 journal->j_flags |= JBD2_BARRIER;
4293 journal->j_flags &= ~JBD2_BARRIER;
4294 if (test_opt(sb, DATA_ERR_ABORT))
4295 journal->j_flags |= JBD2_ABORT_ON_SYNCDATA_ERR;
4297 journal->j_flags &= ~JBD2_ABORT_ON_SYNCDATA_ERR;
4298 write_unlock(&journal->j_state_lock);
4301 static journal_t *ext4_get_journal(struct super_block *sb,
4302 unsigned int journal_inum)
4304 struct inode *journal_inode;
4307 BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL));
4309 /* First, test for the existence of a valid inode on disk. Bad
4310 * things happen if we iget() an unused inode, as the subsequent
4311 * iput() will try to delete it. */
4313 journal_inode = ext4_iget(sb, journal_inum);
4314 if (IS_ERR(journal_inode)) {
4315 ext4_msg(sb, KERN_ERR, "no journal found");
4318 if (!journal_inode->i_nlink) {
4319 make_bad_inode(journal_inode);
4320 iput(journal_inode);
4321 ext4_msg(sb, KERN_ERR, "journal inode is deleted");
4325 jbd_debug(2, "Journal inode found at %p: %lld bytes\n",
4326 journal_inode, journal_inode->i_size);
4327 if (!S_ISREG(journal_inode->i_mode)) {
4328 ext4_msg(sb, KERN_ERR, "invalid journal inode");
4329 iput(journal_inode);
4333 journal = jbd2_journal_init_inode(journal_inode);
4335 ext4_msg(sb, KERN_ERR, "Could not load journal inode");
4336 iput(journal_inode);
4339 journal->j_private = sb;
4340 ext4_init_journal_params(sb, journal);
4344 static journal_t *ext4_get_dev_journal(struct super_block *sb,
4347 struct buffer_head *bh;
4351 int hblock, blocksize;
4352 ext4_fsblk_t sb_block;
4353 unsigned long offset;
4354 struct ext4_super_block *es;
4355 struct block_device *bdev;
4357 BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL));
4359 bdev = ext4_blkdev_get(j_dev, sb);
4363 blocksize = sb->s_blocksize;
4364 hblock = bdev_logical_block_size(bdev);
4365 if (blocksize < hblock) {
4366 ext4_msg(sb, KERN_ERR,
4367 "blocksize too small for journal device");
4371 sb_block = EXT4_MIN_BLOCK_SIZE / blocksize;
4372 offset = EXT4_MIN_BLOCK_SIZE % blocksize;
4373 set_blocksize(bdev, blocksize);
4374 if (!(bh = __bread(bdev, sb_block, blocksize))) {
4375 ext4_msg(sb, KERN_ERR, "couldn't read superblock of "
4376 "external journal");
4380 es = (struct ext4_super_block *) (bh->b_data + offset);
4381 if ((le16_to_cpu(es->s_magic) != EXT4_SUPER_MAGIC) ||
4382 !(le32_to_cpu(es->s_feature_incompat) &
4383 EXT4_FEATURE_INCOMPAT_JOURNAL_DEV)) {
4384 ext4_msg(sb, KERN_ERR, "external journal has "
4390 if (memcmp(EXT4_SB(sb)->s_es->s_journal_uuid, es->s_uuid, 16)) {
4391 ext4_msg(sb, KERN_ERR, "journal UUID does not match");
4396 len = ext4_blocks_count(es);
4397 start = sb_block + 1;
4398 brelse(bh); /* we're done with the superblock */
4400 journal = jbd2_journal_init_dev(bdev, sb->s_bdev,
4401 start, len, blocksize);
4403 ext4_msg(sb, KERN_ERR, "failed to create device journal");
4406 journal->j_private = sb;
4407 ll_rw_block(READ | REQ_META | REQ_PRIO, 1, &journal->j_sb_buffer);
4408 wait_on_buffer(journal->j_sb_buffer);
4409 if (!buffer_uptodate(journal->j_sb_buffer)) {
4410 ext4_msg(sb, KERN_ERR, "I/O error on journal device");
4413 if (be32_to_cpu(journal->j_superblock->s_nr_users) != 1) {
4414 ext4_msg(sb, KERN_ERR, "External journal has more than one "
4415 "user (unsupported) - %d",
4416 be32_to_cpu(journal->j_superblock->s_nr_users));
4419 EXT4_SB(sb)->journal_bdev = bdev;
4420 ext4_init_journal_params(sb, journal);
4424 jbd2_journal_destroy(journal);
4426 ext4_blkdev_put(bdev);
4430 static int ext4_load_journal(struct super_block *sb,
4431 struct ext4_super_block *es,
4432 unsigned long journal_devnum)
4435 unsigned int journal_inum = le32_to_cpu(es->s_journal_inum);
4438 int really_read_only;
4440 BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL));
4442 if (journal_devnum &&
4443 journal_devnum != le32_to_cpu(es->s_journal_dev)) {
4444 ext4_msg(sb, KERN_INFO, "external journal device major/minor "
4445 "numbers have changed");
4446 journal_dev = new_decode_dev(journal_devnum);
4448 journal_dev = new_decode_dev(le32_to_cpu(es->s_journal_dev));
4450 really_read_only = bdev_read_only(sb->s_bdev);
4453 * Are we loading a blank journal or performing recovery after a
4454 * crash? For recovery, we need to check in advance whether we
4455 * can get read-write access to the device.
4457 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER)) {
4458 if (sb->s_flags & MS_RDONLY) {
4459 ext4_msg(sb, KERN_INFO, "INFO: recovery "
4460 "required on readonly filesystem");
4461 if (really_read_only) {
4462 ext4_msg(sb, KERN_ERR, "write access "
4463 "unavailable, cannot proceed");
4466 ext4_msg(sb, KERN_INFO, "write access will "
4467 "be enabled during recovery");
4471 if (journal_inum && journal_dev) {
4472 ext4_msg(sb, KERN_ERR, "filesystem has both journal "
4473 "and inode journals!");
4478 if (!(journal = ext4_get_journal(sb, journal_inum)))
4481 if (!(journal = ext4_get_dev_journal(sb, journal_dev)))
4485 if (!(journal->j_flags & JBD2_BARRIER))
4486 ext4_msg(sb, KERN_INFO, "barriers disabled");
4488 if (!EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER))
4489 err = jbd2_journal_wipe(journal, !really_read_only);
4491 char *save = kmalloc(EXT4_S_ERR_LEN, GFP_KERNEL);
4493 memcpy(save, ((char *) es) +
4494 EXT4_S_ERR_START, EXT4_S_ERR_LEN);
4495 err = jbd2_journal_load(journal);
4497 memcpy(((char *) es) + EXT4_S_ERR_START,
4498 save, EXT4_S_ERR_LEN);
4503 ext4_msg(sb, KERN_ERR, "error loading journal");
4504 jbd2_journal_destroy(journal);
4508 EXT4_SB(sb)->s_journal = journal;
4509 ext4_clear_journal_err(sb, es);
4511 if (!really_read_only && journal_devnum &&
4512 journal_devnum != le32_to_cpu(es->s_journal_dev)) {
4513 es->s_journal_dev = cpu_to_le32(journal_devnum);
4515 /* Make sure we flush the recovery flag to disk. */
4516 ext4_commit_super(sb, 1);
4522 static int ext4_commit_super(struct super_block *sb, int sync)
4524 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
4525 struct buffer_head *sbh = EXT4_SB(sb)->s_sbh;
4528 if (!sbh || block_device_ejected(sb))
4530 if (buffer_write_io_error(sbh)) {
4532 * Oh, dear. A previous attempt to write the
4533 * superblock failed. This could happen because the
4534 * USB device was yanked out. Or it could happen to
4535 * be a transient write error and maybe the block will
4536 * be remapped. Nothing we can do but to retry the
4537 * write and hope for the best.
4539 ext4_msg(sb, KERN_ERR, "previous I/O error to "
4540 "superblock detected");
4541 clear_buffer_write_io_error(sbh);
4542 set_buffer_uptodate(sbh);
4545 * If the file system is mounted read-only, don't update the
4546 * superblock write time. This avoids updating the superblock
4547 * write time when we are mounting the root file system
4548 * read/only but we need to replay the journal; at that point,
4549 * for people who are east of GMT and who make their clock
4550 * tick in localtime for Windows bug-for-bug compatibility,
4551 * the clock is set in the future, and this will cause e2fsck
4552 * to complain and force a full file system check.
4554 if (!(sb->s_flags & MS_RDONLY))
4555 es->s_wtime = cpu_to_le32(get_seconds());
4556 if (sb->s_bdev->bd_part)
4557 es->s_kbytes_written =
4558 cpu_to_le64(EXT4_SB(sb)->s_kbytes_written +
4559 ((part_stat_read(sb->s_bdev->bd_part, sectors[1]) -
4560 EXT4_SB(sb)->s_sectors_written_start) >> 1));
4562 es->s_kbytes_written =
4563 cpu_to_le64(EXT4_SB(sb)->s_kbytes_written);
4564 ext4_free_blocks_count_set(es,
4565 EXT4_C2B(EXT4_SB(sb), percpu_counter_sum_positive(
4566 &EXT4_SB(sb)->s_freeclusters_counter)));
4567 es->s_free_inodes_count =
4568 cpu_to_le32(percpu_counter_sum_positive(
4569 &EXT4_SB(sb)->s_freeinodes_counter));
4570 BUFFER_TRACE(sbh, "marking dirty");
4571 ext4_superblock_csum_set(sb);
4572 mark_buffer_dirty(sbh);
4574 error = sync_dirty_buffer(sbh);
4578 error = buffer_write_io_error(sbh);
4580 ext4_msg(sb, KERN_ERR, "I/O error while writing "
4582 clear_buffer_write_io_error(sbh);
4583 set_buffer_uptodate(sbh);
4590 * Have we just finished recovery? If so, and if we are mounting (or
4591 * remounting) the filesystem readonly, then we will end up with a
4592 * consistent fs on disk. Record that fact.
4594 static void ext4_mark_recovery_complete(struct super_block *sb,
4595 struct ext4_super_block *es)
4597 journal_t *journal = EXT4_SB(sb)->s_journal;
4599 if (!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL)) {
4600 BUG_ON(journal != NULL);
4603 jbd2_journal_lock_updates(journal);
4604 if (jbd2_journal_flush(journal) < 0)
4607 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER) &&
4608 sb->s_flags & MS_RDONLY) {
4609 EXT4_CLEAR_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
4610 ext4_commit_super(sb, 1);
4614 jbd2_journal_unlock_updates(journal);
4618 * If we are mounting (or read-write remounting) a filesystem whose journal
4619 * has recorded an error from a previous lifetime, move that error to the
4620 * main filesystem now.
4622 static void ext4_clear_journal_err(struct super_block *sb,
4623 struct ext4_super_block *es)
4629 BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL));
4631 journal = EXT4_SB(sb)->s_journal;
4634 * Now check for any error status which may have been recorded in the
4635 * journal by a prior ext4_error() or ext4_abort()
4638 j_errno = jbd2_journal_errno(journal);
4642 errstr = ext4_decode_error(sb, j_errno, nbuf);
4643 ext4_warning(sb, "Filesystem error recorded "
4644 "from previous mount: %s", errstr);
4645 ext4_warning(sb, "Marking fs in need of filesystem check.");
4647 EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS;
4648 es->s_state |= cpu_to_le16(EXT4_ERROR_FS);
4649 ext4_commit_super(sb, 1);
4651 jbd2_journal_clear_err(journal);
4652 jbd2_journal_update_sb_errno(journal);
4657 * Force the running and committing transactions to commit,
4658 * and wait on the commit.
4660 int ext4_force_commit(struct super_block *sb)
4664 if (sb->s_flags & MS_RDONLY)
4667 journal = EXT4_SB(sb)->s_journal;
4668 return ext4_journal_force_commit(journal);
4671 static int ext4_sync_fs(struct super_block *sb, int wait)
4675 bool needs_barrier = false;
4676 struct ext4_sb_info *sbi = EXT4_SB(sb);
4678 trace_ext4_sync_fs(sb, wait);
4679 flush_workqueue(sbi->rsv_conversion_wq);
4681 * Writeback quota in non-journalled quota case - journalled quota has
4684 dquot_writeback_dquots(sb, -1);
4686 * Data writeback is possible w/o journal transaction, so barrier must
4687 * being sent at the end of the function. But we can skip it if
4688 * transaction_commit will do it for us.
4690 target = jbd2_get_latest_transaction(sbi->s_journal);
4691 if (wait && sbi->s_journal->j_flags & JBD2_BARRIER &&
4692 !jbd2_trans_will_send_data_barrier(sbi->s_journal, target))
4693 needs_barrier = true;
4695 if (jbd2_journal_start_commit(sbi->s_journal, &target)) {
4697 ret = jbd2_log_wait_commit(sbi->s_journal, target);
4699 if (needs_barrier) {
4701 err = blkdev_issue_flush(sb->s_bdev, GFP_KERNEL, NULL);
4709 static int ext4_sync_fs_nojournal(struct super_block *sb, int wait)
4713 trace_ext4_sync_fs(sb, wait);
4714 flush_workqueue(EXT4_SB(sb)->rsv_conversion_wq);
4715 dquot_writeback_dquots(sb, -1);
4716 if (wait && test_opt(sb, BARRIER))
4717 ret = blkdev_issue_flush(sb->s_bdev, GFP_KERNEL, NULL);
4723 * LVM calls this function before a (read-only) snapshot is created. This
4724 * gives us a chance to flush the journal completely and mark the fs clean.
4726 * Note that only this function cannot bring a filesystem to be in a clean
4727 * state independently. It relies on upper layer to stop all data & metadata
4730 static int ext4_freeze(struct super_block *sb)
4735 if (sb->s_flags & MS_RDONLY)
4738 journal = EXT4_SB(sb)->s_journal;
4740 /* Now we set up the journal barrier. */
4741 jbd2_journal_lock_updates(journal);
4744 * Don't clear the needs_recovery flag if we failed to flush
4747 error = jbd2_journal_flush(journal);
4751 /* Journal blocked and flushed, clear needs_recovery flag. */
4752 EXT4_CLEAR_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
4753 error = ext4_commit_super(sb, 1);
4755 /* we rely on upper layer to stop further updates */
4756 jbd2_journal_unlock_updates(EXT4_SB(sb)->s_journal);
4761 * Called by LVM after the snapshot is done. We need to reset the RECOVER
4762 * flag here, even though the filesystem is not technically dirty yet.
4764 static int ext4_unfreeze(struct super_block *sb)
4766 if (sb->s_flags & MS_RDONLY)
4769 /* Reset the needs_recovery flag before the fs is unlocked. */
4770 EXT4_SET_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
4771 ext4_commit_super(sb, 1);
4776 * Structure to save mount options for ext4_remount's benefit
4778 struct ext4_mount_options {
4779 unsigned long s_mount_opt;
4780 unsigned long s_mount_opt2;
4783 unsigned long s_commit_interval;
4784 u32 s_min_batch_time, s_max_batch_time;
4787 char *s_qf_names[MAXQUOTAS];
4791 static int ext4_remount(struct super_block *sb, int *flags, char *data)
4793 struct ext4_super_block *es;
4794 struct ext4_sb_info *sbi = EXT4_SB(sb);
4795 unsigned long old_sb_flags;
4796 struct ext4_mount_options old_opts;
4797 int enable_quota = 0;
4799 unsigned int journal_ioprio = DEFAULT_JOURNAL_IOPRIO;
4804 char *orig_data = kstrdup(data, GFP_KERNEL);
4806 /* Store the original options */
4807 old_sb_flags = sb->s_flags;
4808 old_opts.s_mount_opt = sbi->s_mount_opt;
4809 old_opts.s_mount_opt2 = sbi->s_mount_opt2;
4810 old_opts.s_resuid = sbi->s_resuid;
4811 old_opts.s_resgid = sbi->s_resgid;
4812 old_opts.s_commit_interval = sbi->s_commit_interval;
4813 old_opts.s_min_batch_time = sbi->s_min_batch_time;
4814 old_opts.s_max_batch_time = sbi->s_max_batch_time;
4816 old_opts.s_jquota_fmt = sbi->s_jquota_fmt;
4817 for (i = 0; i < MAXQUOTAS; i++)
4818 if (sbi->s_qf_names[i]) {
4819 old_opts.s_qf_names[i] = kstrdup(sbi->s_qf_names[i],
4821 if (!old_opts.s_qf_names[i]) {
4822 for (j = 0; j < i; j++)
4823 kfree(old_opts.s_qf_names[j]);
4828 old_opts.s_qf_names[i] = NULL;
4830 if (sbi->s_journal && sbi->s_journal->j_task->io_context)
4831 journal_ioprio = sbi->s_journal->j_task->io_context->ioprio;
4834 * Allow the "check" option to be passed as a remount option.
4836 if (!parse_options(data, sb, NULL, &journal_ioprio, 1)) {
4841 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA) {
4842 if (test_opt2(sb, EXPLICIT_DELALLOC)) {
4843 ext4_msg(sb, KERN_ERR, "can't mount with "
4844 "both data=journal and delalloc");
4848 if (test_opt(sb, DIOREAD_NOLOCK)) {
4849 ext4_msg(sb, KERN_ERR, "can't mount with "
4850 "both data=journal and dioread_nolock");
4856 if (sbi->s_mount_flags & EXT4_MF_FS_ABORTED)
4857 ext4_abort(sb, "Abort forced by user");
4859 sb->s_flags = (sb->s_flags & ~MS_POSIXACL) |
4860 (test_opt(sb, POSIX_ACL) ? MS_POSIXACL : 0);
4864 if (sbi->s_journal) {
4865 ext4_init_journal_params(sb, sbi->s_journal);
4866 set_task_ioprio(sbi->s_journal->j_task, journal_ioprio);
4869 if ((*flags & MS_RDONLY) != (sb->s_flags & MS_RDONLY)) {
4870 if (sbi->s_mount_flags & EXT4_MF_FS_ABORTED) {
4875 if (*flags & MS_RDONLY) {
4876 err = sync_filesystem(sb);
4879 err = dquot_suspend(sb, -1);
4884 * First of all, the unconditional stuff we have to do
4885 * to disable replay of the journal when we next remount
4887 sb->s_flags |= MS_RDONLY;
4890 * OK, test if we are remounting a valid rw partition
4891 * readonly, and if so set the rdonly flag and then
4892 * mark the partition as valid again.
4894 if (!(es->s_state & cpu_to_le16(EXT4_VALID_FS)) &&
4895 (sbi->s_mount_state & EXT4_VALID_FS))
4896 es->s_state = cpu_to_le16(sbi->s_mount_state);
4899 ext4_mark_recovery_complete(sb, es);
4901 /* Make sure we can mount this feature set readwrite */
4902 if (!ext4_feature_set_ok(sb, 0)) {
4907 * Make sure the group descriptor checksums
4908 * are sane. If they aren't, refuse to remount r/w.
4910 for (g = 0; g < sbi->s_groups_count; g++) {
4911 struct ext4_group_desc *gdp =
4912 ext4_get_group_desc(sb, g, NULL);
4914 if (!ext4_group_desc_csum_verify(sb, g, gdp)) {
4915 ext4_msg(sb, KERN_ERR,
4916 "ext4_remount: Checksum for group %u failed (%u!=%u)",
4917 g, le16_to_cpu(ext4_group_desc_csum(sbi, g, gdp)),
4918 le16_to_cpu(gdp->bg_checksum));
4925 * If we have an unprocessed orphan list hanging
4926 * around from a previously readonly bdev mount,
4927 * require a full umount/remount for now.
4929 if (es->s_last_orphan) {
4930 ext4_msg(sb, KERN_WARNING, "Couldn't "
4931 "remount RDWR because of unprocessed "
4932 "orphan inode list. Please "
4933 "umount/remount instead");
4939 * Mounting a RDONLY partition read-write, so reread
4940 * and store the current valid flag. (It may have
4941 * been changed by e2fsck since we originally mounted
4945 ext4_clear_journal_err(sb, es);
4946 sbi->s_mount_state = le16_to_cpu(es->s_state);
4947 if (!ext4_setup_super(sb, es, 0))
4948 sb->s_flags &= ~MS_RDONLY;
4949 if (EXT4_HAS_INCOMPAT_FEATURE(sb,
4950 EXT4_FEATURE_INCOMPAT_MMP))
4951 if (ext4_multi_mount_protect(sb,
4952 le64_to_cpu(es->s_mmp_block))) {
4961 * Reinitialize lazy itable initialization thread based on
4964 if ((sb->s_flags & MS_RDONLY) || !test_opt(sb, INIT_INODE_TABLE))
4965 ext4_unregister_li_request(sb);
4967 ext4_group_t first_not_zeroed;
4968 first_not_zeroed = ext4_has_uninit_itable(sb);
4969 ext4_register_li_request(sb, first_not_zeroed);
4972 ext4_setup_system_zone(sb);
4973 if (sbi->s_journal == NULL && !(old_sb_flags & MS_RDONLY))
4974 ext4_commit_super(sb, 1);
4977 /* Release old quota file names */
4978 for (i = 0; i < MAXQUOTAS; i++)
4979 kfree(old_opts.s_qf_names[i]);
4981 if (sb_any_quota_suspended(sb))
4982 dquot_resume(sb, -1);
4983 else if (EXT4_HAS_RO_COMPAT_FEATURE(sb,
4984 EXT4_FEATURE_RO_COMPAT_QUOTA)) {
4985 err = ext4_enable_quotas(sb);
4992 ext4_msg(sb, KERN_INFO, "re-mounted. Opts: %s", orig_data);
4997 sb->s_flags = old_sb_flags;
4998 sbi->s_mount_opt = old_opts.s_mount_opt;
4999 sbi->s_mount_opt2 = old_opts.s_mount_opt2;
5000 sbi->s_resuid = old_opts.s_resuid;
5001 sbi->s_resgid = old_opts.s_resgid;
5002 sbi->s_commit_interval = old_opts.s_commit_interval;
5003 sbi->s_min_batch_time = old_opts.s_min_batch_time;
5004 sbi->s_max_batch_time = old_opts.s_max_batch_time;
5006 sbi->s_jquota_fmt = old_opts.s_jquota_fmt;
5007 for (i = 0; i < MAXQUOTAS; i++) {
5008 kfree(sbi->s_qf_names[i]);
5009 sbi->s_qf_names[i] = old_opts.s_qf_names[i];
5016 static int ext4_statfs(struct dentry *dentry, struct kstatfs *buf)
5018 struct super_block *sb = dentry->d_sb;
5019 struct ext4_sb_info *sbi = EXT4_SB(sb);
5020 struct ext4_super_block *es = sbi->s_es;
5021 ext4_fsblk_t overhead = 0, resv_blocks;
5024 resv_blocks = EXT4_C2B(sbi, atomic64_read(&sbi->s_resv_clusters));
5026 if (!test_opt(sb, MINIX_DF))
5027 overhead = sbi->s_overhead;
5029 buf->f_type = EXT4_SUPER_MAGIC;
5030 buf->f_bsize = sb->s_blocksize;
5031 buf->f_blocks = ext4_blocks_count(es) - EXT4_C2B(sbi, overhead);
5032 bfree = percpu_counter_sum_positive(&sbi->s_freeclusters_counter) -
5033 percpu_counter_sum_positive(&sbi->s_dirtyclusters_counter);
5034 /* prevent underflow in case that few free space is available */
5035 buf->f_bfree = EXT4_C2B(sbi, max_t(s64, bfree, 0));
5036 buf->f_bavail = buf->f_bfree -
5037 (ext4_r_blocks_count(es) + resv_blocks);
5038 if (buf->f_bfree < (ext4_r_blocks_count(es) + resv_blocks))
5040 buf->f_files = le32_to_cpu(es->s_inodes_count);
5041 buf->f_ffree = percpu_counter_sum_positive(&sbi->s_freeinodes_counter);
5042 buf->f_namelen = EXT4_NAME_LEN;
5043 fsid = le64_to_cpup((void *)es->s_uuid) ^
5044 le64_to_cpup((void *)es->s_uuid + sizeof(u64));
5045 buf->f_fsid.val[0] = fsid & 0xFFFFFFFFUL;
5046 buf->f_fsid.val[1] = (fsid >> 32) & 0xFFFFFFFFUL;
5051 /* Helper function for writing quotas on sync - we need to start transaction
5052 * before quota file is locked for write. Otherwise the are possible deadlocks:
5053 * Process 1 Process 2
5054 * ext4_create() quota_sync()
5055 * jbd2_journal_start() write_dquot()
5056 * dquot_initialize() down(dqio_mutex)
5057 * down(dqio_mutex) jbd2_journal_start()
5063 static inline struct inode *dquot_to_inode(struct dquot *dquot)
5065 return sb_dqopt(dquot->dq_sb)->files[dquot->dq_id.type];
5068 static int ext4_write_dquot(struct dquot *dquot)
5072 struct inode *inode;
5074 inode = dquot_to_inode(dquot);
5075 handle = ext4_journal_start(inode, EXT4_HT_QUOTA,
5076 EXT4_QUOTA_TRANS_BLOCKS(dquot->dq_sb));
5078 return PTR_ERR(handle);
5079 ret = dquot_commit(dquot);
5080 err = ext4_journal_stop(handle);
5086 static int ext4_acquire_dquot(struct dquot *dquot)
5091 handle = ext4_journal_start(dquot_to_inode(dquot), EXT4_HT_QUOTA,
5092 EXT4_QUOTA_INIT_BLOCKS(dquot->dq_sb));
5094 return PTR_ERR(handle);
5095 ret = dquot_acquire(dquot);
5096 err = ext4_journal_stop(handle);
5102 static int ext4_release_dquot(struct dquot *dquot)
5107 handle = ext4_journal_start(dquot_to_inode(dquot), EXT4_HT_QUOTA,
5108 EXT4_QUOTA_DEL_BLOCKS(dquot->dq_sb));
5109 if (IS_ERR(handle)) {
5110 /* Release dquot anyway to avoid endless cycle in dqput() */
5111 dquot_release(dquot);
5112 return PTR_ERR(handle);
5114 ret = dquot_release(dquot);
5115 err = ext4_journal_stop(handle);
5121 static int ext4_mark_dquot_dirty(struct dquot *dquot)
5123 struct super_block *sb = dquot->dq_sb;
5124 struct ext4_sb_info *sbi = EXT4_SB(sb);
5126 /* Are we journaling quotas? */
5127 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_QUOTA) ||
5128 sbi->s_qf_names[USRQUOTA] || sbi->s_qf_names[GRPQUOTA]) {
5129 dquot_mark_dquot_dirty(dquot);
5130 return ext4_write_dquot(dquot);
5132 return dquot_mark_dquot_dirty(dquot);
5136 static int ext4_write_info(struct super_block *sb, int type)
5141 /* Data block + inode block */
5142 handle = ext4_journal_start(sb->s_root->d_inode, EXT4_HT_QUOTA, 2);
5144 return PTR_ERR(handle);
5145 ret = dquot_commit_info(sb, type);
5146 err = ext4_journal_stop(handle);
5153 * Turn on quotas during mount time - we need to find
5154 * the quota file and such...
5156 static int ext4_quota_on_mount(struct super_block *sb, int type)
5158 return dquot_quota_on_mount(sb, EXT4_SB(sb)->s_qf_names[type],
5159 EXT4_SB(sb)->s_jquota_fmt, type);
5163 * Standard function to be called on quota_on
5165 static int ext4_quota_on(struct super_block *sb, int type, int format_id,
5170 if (!test_opt(sb, QUOTA))
5173 /* Quotafile not on the same filesystem? */
5174 if (path->dentry->d_sb != sb)
5176 /* Journaling quota? */
5177 if (EXT4_SB(sb)->s_qf_names[type]) {
5178 /* Quotafile not in fs root? */
5179 if (path->dentry->d_parent != sb->s_root)
5180 ext4_msg(sb, KERN_WARNING,
5181 "Quota file not on filesystem root. "
5182 "Journaled quota will not work");
5186 * When we journal data on quota file, we have to flush journal to see
5187 * all updates to the file when we bypass pagecache...
5189 if (EXT4_SB(sb)->s_journal &&
5190 ext4_should_journal_data(path->dentry->d_inode)) {
5192 * We don't need to lock updates but journal_flush() could
5193 * otherwise be livelocked...
5195 jbd2_journal_lock_updates(EXT4_SB(sb)->s_journal);
5196 err = jbd2_journal_flush(EXT4_SB(sb)->s_journal);
5197 jbd2_journal_unlock_updates(EXT4_SB(sb)->s_journal);
5202 return dquot_quota_on(sb, type, format_id, path);
5205 static int ext4_quota_enable(struct super_block *sb, int type, int format_id,
5209 struct inode *qf_inode;
5210 unsigned long qf_inums[MAXQUOTAS] = {
5211 le32_to_cpu(EXT4_SB(sb)->s_es->s_usr_quota_inum),
5212 le32_to_cpu(EXT4_SB(sb)->s_es->s_grp_quota_inum)
5215 BUG_ON(!EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_QUOTA));
5217 if (!qf_inums[type])
5220 qf_inode = ext4_iget(sb, qf_inums[type]);
5221 if (IS_ERR(qf_inode)) {
5222 ext4_error(sb, "Bad quota inode # %lu", qf_inums[type]);
5223 return PTR_ERR(qf_inode);
5226 /* Don't account quota for quota files to avoid recursion */
5227 qf_inode->i_flags |= S_NOQUOTA;
5228 err = dquot_enable(qf_inode, type, format_id, flags);
5234 /* Enable usage tracking for all quota types. */
5235 static int ext4_enable_quotas(struct super_block *sb)
5238 unsigned long qf_inums[MAXQUOTAS] = {
5239 le32_to_cpu(EXT4_SB(sb)->s_es->s_usr_quota_inum),
5240 le32_to_cpu(EXT4_SB(sb)->s_es->s_grp_quota_inum)
5243 sb_dqopt(sb)->flags |= DQUOT_QUOTA_SYS_FILE;
5244 for (type = 0; type < MAXQUOTAS; type++) {
5245 if (qf_inums[type]) {
5246 err = ext4_quota_enable(sb, type, QFMT_VFS_V1,
5247 DQUOT_USAGE_ENABLED);
5250 "Failed to enable quota tracking "
5251 "(type=%d, err=%d). Please run "
5252 "e2fsck to fix.", type, err);
5261 * quota_on function that is used when QUOTA feature is set.
5263 static int ext4_quota_on_sysfile(struct super_block *sb, int type,
5266 if (!EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_QUOTA))
5270 * USAGE was enabled at mount time. Only need to enable LIMITS now.
5272 return ext4_quota_enable(sb, type, format_id, DQUOT_LIMITS_ENABLED);
5275 static int ext4_quota_off(struct super_block *sb, int type)
5277 struct inode *inode = sb_dqopt(sb)->files[type];
5280 /* Force all delayed allocation blocks to be allocated.
5281 * Caller already holds s_umount sem */
5282 if (test_opt(sb, DELALLOC))
5283 sync_filesystem(sb);
5288 /* Update modification times of quota files when userspace can
5289 * start looking at them */
5290 handle = ext4_journal_start(inode, EXT4_HT_QUOTA, 1);
5293 inode->i_mtime = inode->i_ctime = CURRENT_TIME;
5294 ext4_mark_inode_dirty(handle, inode);
5295 ext4_journal_stop(handle);
5298 return dquot_quota_off(sb, type);
5302 * quota_off function that is used when QUOTA feature is set.
5304 static int ext4_quota_off_sysfile(struct super_block *sb, int type)
5306 if (!EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_QUOTA))
5309 /* Disable only the limits. */
5310 return dquot_disable(sb, type, DQUOT_LIMITS_ENABLED);
5313 /* Read data from quotafile - avoid pagecache and such because we cannot afford
5314 * acquiring the locks... As quota files are never truncated and quota code
5315 * itself serializes the operations (and no one else should touch the files)
5316 * we don't have to be afraid of races */
5317 static ssize_t ext4_quota_read(struct super_block *sb, int type, char *data,
5318 size_t len, loff_t off)
5320 struct inode *inode = sb_dqopt(sb)->files[type];
5321 ext4_lblk_t blk = off >> EXT4_BLOCK_SIZE_BITS(sb);
5323 int offset = off & (sb->s_blocksize - 1);
5326 struct buffer_head *bh;
5327 loff_t i_size = i_size_read(inode);
5331 if (off+len > i_size)
5334 while (toread > 0) {
5335 tocopy = sb->s_blocksize - offset < toread ?
5336 sb->s_blocksize - offset : toread;
5337 bh = ext4_bread(NULL, inode, blk, 0, &err);
5340 if (!bh) /* A hole? */
5341 memset(data, 0, tocopy);
5343 memcpy(data, bh->b_data+offset, tocopy);
5353 /* Write to quotafile (we know the transaction is already started and has
5354 * enough credits) */
5355 static ssize_t ext4_quota_write(struct super_block *sb, int type,
5356 const char *data, size_t len, loff_t off)
5358 struct inode *inode = sb_dqopt(sb)->files[type];
5359 ext4_lblk_t blk = off >> EXT4_BLOCK_SIZE_BITS(sb);
5361 int offset = off & (sb->s_blocksize - 1);
5362 struct buffer_head *bh;
5363 handle_t *handle = journal_current_handle();
5365 if (EXT4_SB(sb)->s_journal && !handle) {
5366 ext4_msg(sb, KERN_WARNING, "Quota write (off=%llu, len=%llu)"
5367 " cancelled because transaction is not started",
5368 (unsigned long long)off, (unsigned long long)len);
5372 * Since we account only one data block in transaction credits,
5373 * then it is impossible to cross a block boundary.
5375 if (sb->s_blocksize - offset < len) {
5376 ext4_msg(sb, KERN_WARNING, "Quota write (off=%llu, len=%llu)"
5377 " cancelled because not block aligned",
5378 (unsigned long long)off, (unsigned long long)len);
5382 bh = ext4_bread(handle, inode, blk, 1, &err);
5385 BUFFER_TRACE(bh, "get write access");
5386 err = ext4_journal_get_write_access(handle, bh);
5392 memcpy(bh->b_data+offset, data, len);
5393 flush_dcache_page(bh->b_page);
5395 err = ext4_handle_dirty_metadata(handle, NULL, bh);
5400 if (inode->i_size < off + len) {
5401 i_size_write(inode, off + len);
5402 EXT4_I(inode)->i_disksize = inode->i_size;
5403 ext4_mark_inode_dirty(handle, inode);
5410 static struct dentry *ext4_mount(struct file_system_type *fs_type, int flags,
5411 const char *dev_name, void *data)
5413 return mount_bdev(fs_type, flags, dev_name, data, ext4_fill_super);
5416 #if !defined(CONFIG_EXT2_FS) && !defined(CONFIG_EXT2_FS_MODULE) && defined(CONFIG_EXT4_USE_FOR_EXT23)
5417 static inline void register_as_ext2(void)
5419 int err = register_filesystem(&ext2_fs_type);
5422 "EXT4-fs: Unable to register as ext2 (%d)\n", err);
5425 static inline void unregister_as_ext2(void)
5427 unregister_filesystem(&ext2_fs_type);
5430 static inline int ext2_feature_set_ok(struct super_block *sb)
5432 if (EXT4_HAS_INCOMPAT_FEATURE(sb, ~EXT2_FEATURE_INCOMPAT_SUPP))
5434 if (sb->s_flags & MS_RDONLY)
5436 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, ~EXT2_FEATURE_RO_COMPAT_SUPP))
5441 static inline void register_as_ext2(void) { }
5442 static inline void unregister_as_ext2(void) { }
5443 static inline int ext2_feature_set_ok(struct super_block *sb) { return 0; }
5446 #if !defined(CONFIG_EXT3_FS) && !defined(CONFIG_EXT3_FS_MODULE) && defined(CONFIG_EXT4_USE_FOR_EXT23)
5447 static inline void register_as_ext3(void)
5449 int err = register_filesystem(&ext3_fs_type);
5452 "EXT4-fs: Unable to register as ext3 (%d)\n", err);
5455 static inline void unregister_as_ext3(void)
5457 unregister_filesystem(&ext3_fs_type);
5460 static inline int ext3_feature_set_ok(struct super_block *sb)
5462 if (EXT4_HAS_INCOMPAT_FEATURE(sb, ~EXT3_FEATURE_INCOMPAT_SUPP))
5464 if (!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL))
5466 if (sb->s_flags & MS_RDONLY)
5468 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, ~EXT3_FEATURE_RO_COMPAT_SUPP))
5473 static inline void register_as_ext3(void) { }
5474 static inline void unregister_as_ext3(void) { }
5475 static inline int ext3_feature_set_ok(struct super_block *sb) { return 0; }
5478 static struct file_system_type ext4_fs_type = {
5479 .owner = THIS_MODULE,
5481 .mount = ext4_mount,
5482 .kill_sb = kill_block_super,
5483 .fs_flags = FS_REQUIRES_DEV,
5485 MODULE_ALIAS_FS("ext4");
5487 static int __init ext4_init_feat_adverts(void)
5489 struct ext4_features *ef;
5492 ef = kzalloc(sizeof(struct ext4_features), GFP_KERNEL);
5496 ef->f_kobj.kset = ext4_kset;
5497 init_completion(&ef->f_kobj_unregister);
5498 ret = kobject_init_and_add(&ef->f_kobj, &ext4_feat_ktype, NULL,
5511 static void ext4_exit_feat_adverts(void)
5513 kobject_put(&ext4_feat->f_kobj);
5514 wait_for_completion(&ext4_feat->f_kobj_unregister);
5518 /* Shared across all ext4 file systems */
5519 wait_queue_head_t ext4__ioend_wq[EXT4_WQ_HASH_SZ];
5520 struct mutex ext4__aio_mutex[EXT4_WQ_HASH_SZ];
5522 static int __init ext4_init_fs(void)
5526 ext4_li_info = NULL;
5527 mutex_init(&ext4_li_mtx);
5529 /* Build-time check for flags consistency */
5530 ext4_check_flag_values();
5532 for (i = 0; i < EXT4_WQ_HASH_SZ; i++) {
5533 mutex_init(&ext4__aio_mutex[i]);
5534 init_waitqueue_head(&ext4__ioend_wq[i]);
5537 err = ext4_init_es();
5541 err = ext4_init_pageio();
5545 err = ext4_init_system_zone();
5548 ext4_kset = kset_create_and_add("ext4", NULL, fs_kobj);
5553 ext4_proc_root = proc_mkdir("fs/ext4", NULL);
5555 err = ext4_init_feat_adverts();
5559 err = ext4_init_mballoc();
5563 ext4_mballoc_ready = 1;
5564 err = init_inodecache();
5569 err = register_filesystem(&ext4_fs_type);
5575 unregister_as_ext2();
5576 unregister_as_ext3();
5577 destroy_inodecache();
5579 ext4_mballoc_ready = 0;
5580 ext4_exit_mballoc();
5582 ext4_exit_feat_adverts();
5585 remove_proc_entry("fs/ext4", NULL);
5586 kset_unregister(ext4_kset);
5588 ext4_exit_system_zone();
5597 static void __exit ext4_exit_fs(void)
5599 ext4_destroy_lazyinit_thread();
5600 unregister_as_ext2();
5601 unregister_as_ext3();
5602 unregister_filesystem(&ext4_fs_type);
5603 destroy_inodecache();
5604 ext4_exit_mballoc();
5605 ext4_exit_feat_adverts();
5606 remove_proc_entry("fs/ext4", NULL);
5607 kset_unregister(ext4_kset);
5608 ext4_exit_system_zone();
5613 MODULE_AUTHOR("Remy Card, Stephen Tweedie, Andrew Morton, Andreas Dilger, Theodore Ts'o and others");
5614 MODULE_DESCRIPTION("Fourth Extended Filesystem");
5615 MODULE_LICENSE("GPL");
5616 module_init(ext4_init_fs)
5617 module_exit(ext4_exit_fs)