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_remount(struct super_block *sb, int *flags, char *data);
74 static int ext4_statfs(struct dentry *dentry, struct kstatfs *buf);
75 static int ext4_unfreeze(struct super_block *sb);
76 static int ext4_freeze(struct super_block *sb);
77 static struct dentry *ext4_mount(struct file_system_type *fs_type, int flags,
78 const char *dev_name, void *data);
79 static inline int ext2_feature_set_ok(struct super_block *sb);
80 static inline int ext3_feature_set_ok(struct super_block *sb);
81 static int ext4_feature_set_ok(struct super_block *sb, int readonly);
82 static void ext4_destroy_lazyinit_thread(void);
83 static void ext4_unregister_li_request(struct super_block *sb);
84 static void ext4_clear_request_list(void);
85 static int ext4_reserve_clusters(struct ext4_sb_info *, ext4_fsblk_t);
87 #if !defined(CONFIG_EXT2_FS) && !defined(CONFIG_EXT2_FS_MODULE) && defined(CONFIG_EXT4_USE_FOR_EXT23)
88 static struct file_system_type ext2_fs_type = {
92 .kill_sb = kill_block_super,
93 .fs_flags = FS_REQUIRES_DEV,
95 MODULE_ALIAS_FS("ext2");
97 #define IS_EXT2_SB(sb) ((sb)->s_bdev->bd_holder == &ext2_fs_type)
99 #define IS_EXT2_SB(sb) (0)
103 #if !defined(CONFIG_EXT3_FS) && !defined(CONFIG_EXT3_FS_MODULE) && defined(CONFIG_EXT4_USE_FOR_EXT23)
104 static struct file_system_type ext3_fs_type = {
105 .owner = THIS_MODULE,
108 .kill_sb = kill_block_super,
109 .fs_flags = FS_REQUIRES_DEV,
111 MODULE_ALIAS_FS("ext3");
112 MODULE_ALIAS("ext3");
113 #define IS_EXT3_SB(sb) ((sb)->s_bdev->bd_holder == &ext3_fs_type)
115 #define IS_EXT3_SB(sb) (0)
118 static int ext4_verify_csum_type(struct super_block *sb,
119 struct ext4_super_block *es)
121 if (!EXT4_HAS_RO_COMPAT_FEATURE(sb,
122 EXT4_FEATURE_RO_COMPAT_METADATA_CSUM))
125 return es->s_checksum_type == EXT4_CRC32C_CHKSUM;
128 static __le32 ext4_superblock_csum(struct super_block *sb,
129 struct ext4_super_block *es)
131 struct ext4_sb_info *sbi = EXT4_SB(sb);
132 int offset = offsetof(struct ext4_super_block, s_checksum);
135 csum = ext4_chksum(sbi, ~0, (char *)es, offset);
137 return cpu_to_le32(csum);
140 static int ext4_superblock_csum_verify(struct super_block *sb,
141 struct ext4_super_block *es)
143 if (!ext4_has_metadata_csum(sb))
146 return es->s_checksum == ext4_superblock_csum(sb, es);
149 void ext4_superblock_csum_set(struct super_block *sb)
151 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
153 if (!ext4_has_metadata_csum(sb))
156 es->s_checksum = ext4_superblock_csum(sb, es);
159 void *ext4_kvmalloc(size_t size, gfp_t flags)
163 ret = kmalloc(size, flags | __GFP_NOWARN);
165 ret = __vmalloc(size, flags, PAGE_KERNEL);
169 void *ext4_kvzalloc(size_t size, gfp_t flags)
173 ret = kzalloc(size, flags | __GFP_NOWARN);
175 ret = __vmalloc(size, flags | __GFP_ZERO, PAGE_KERNEL);
179 ext4_fsblk_t ext4_block_bitmap(struct super_block *sb,
180 struct ext4_group_desc *bg)
182 return le32_to_cpu(bg->bg_block_bitmap_lo) |
183 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
184 (ext4_fsblk_t)le32_to_cpu(bg->bg_block_bitmap_hi) << 32 : 0);
187 ext4_fsblk_t ext4_inode_bitmap(struct super_block *sb,
188 struct ext4_group_desc *bg)
190 return le32_to_cpu(bg->bg_inode_bitmap_lo) |
191 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
192 (ext4_fsblk_t)le32_to_cpu(bg->bg_inode_bitmap_hi) << 32 : 0);
195 ext4_fsblk_t ext4_inode_table(struct super_block *sb,
196 struct ext4_group_desc *bg)
198 return le32_to_cpu(bg->bg_inode_table_lo) |
199 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
200 (ext4_fsblk_t)le32_to_cpu(bg->bg_inode_table_hi) << 32 : 0);
203 __u32 ext4_free_group_clusters(struct super_block *sb,
204 struct ext4_group_desc *bg)
206 return le16_to_cpu(bg->bg_free_blocks_count_lo) |
207 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
208 (__u32)le16_to_cpu(bg->bg_free_blocks_count_hi) << 16 : 0);
211 __u32 ext4_free_inodes_count(struct super_block *sb,
212 struct ext4_group_desc *bg)
214 return le16_to_cpu(bg->bg_free_inodes_count_lo) |
215 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
216 (__u32)le16_to_cpu(bg->bg_free_inodes_count_hi) << 16 : 0);
219 __u32 ext4_used_dirs_count(struct super_block *sb,
220 struct ext4_group_desc *bg)
222 return le16_to_cpu(bg->bg_used_dirs_count_lo) |
223 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
224 (__u32)le16_to_cpu(bg->bg_used_dirs_count_hi) << 16 : 0);
227 __u32 ext4_itable_unused_count(struct super_block *sb,
228 struct ext4_group_desc *bg)
230 return le16_to_cpu(bg->bg_itable_unused_lo) |
231 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
232 (__u32)le16_to_cpu(bg->bg_itable_unused_hi) << 16 : 0);
235 void ext4_block_bitmap_set(struct super_block *sb,
236 struct ext4_group_desc *bg, ext4_fsblk_t blk)
238 bg->bg_block_bitmap_lo = cpu_to_le32((u32)blk);
239 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
240 bg->bg_block_bitmap_hi = cpu_to_le32(blk >> 32);
243 void ext4_inode_bitmap_set(struct super_block *sb,
244 struct ext4_group_desc *bg, ext4_fsblk_t blk)
246 bg->bg_inode_bitmap_lo = cpu_to_le32((u32)blk);
247 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
248 bg->bg_inode_bitmap_hi = cpu_to_le32(blk >> 32);
251 void ext4_inode_table_set(struct super_block *sb,
252 struct ext4_group_desc *bg, ext4_fsblk_t blk)
254 bg->bg_inode_table_lo = cpu_to_le32((u32)blk);
255 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
256 bg->bg_inode_table_hi = cpu_to_le32(blk >> 32);
259 void ext4_free_group_clusters_set(struct super_block *sb,
260 struct ext4_group_desc *bg, __u32 count)
262 bg->bg_free_blocks_count_lo = cpu_to_le16((__u16)count);
263 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
264 bg->bg_free_blocks_count_hi = cpu_to_le16(count >> 16);
267 void ext4_free_inodes_set(struct super_block *sb,
268 struct ext4_group_desc *bg, __u32 count)
270 bg->bg_free_inodes_count_lo = cpu_to_le16((__u16)count);
271 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
272 bg->bg_free_inodes_count_hi = cpu_to_le16(count >> 16);
275 void ext4_used_dirs_set(struct super_block *sb,
276 struct ext4_group_desc *bg, __u32 count)
278 bg->bg_used_dirs_count_lo = cpu_to_le16((__u16)count);
279 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
280 bg->bg_used_dirs_count_hi = cpu_to_le16(count >> 16);
283 void ext4_itable_unused_set(struct super_block *sb,
284 struct ext4_group_desc *bg, __u32 count)
286 bg->bg_itable_unused_lo = cpu_to_le16((__u16)count);
287 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
288 bg->bg_itable_unused_hi = cpu_to_le16(count >> 16);
292 static void __save_error_info(struct super_block *sb, const char *func,
295 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
297 EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS;
298 es->s_state |= cpu_to_le16(EXT4_ERROR_FS);
299 es->s_last_error_time = cpu_to_le32(get_seconds());
300 strncpy(es->s_last_error_func, func, sizeof(es->s_last_error_func));
301 es->s_last_error_line = cpu_to_le32(line);
302 if (!es->s_first_error_time) {
303 es->s_first_error_time = es->s_last_error_time;
304 strncpy(es->s_first_error_func, func,
305 sizeof(es->s_first_error_func));
306 es->s_first_error_line = cpu_to_le32(line);
307 es->s_first_error_ino = es->s_last_error_ino;
308 es->s_first_error_block = es->s_last_error_block;
311 * Start the daily error reporting function if it hasn't been
314 if (!es->s_error_count)
315 mod_timer(&EXT4_SB(sb)->s_err_report, jiffies + 24*60*60*HZ);
316 le32_add_cpu(&es->s_error_count, 1);
319 static void save_error_info(struct super_block *sb, const char *func,
322 __save_error_info(sb, func, line);
323 ext4_commit_super(sb, 1);
327 * The del_gendisk() function uninitializes the disk-specific data
328 * structures, including the bdi structure, without telling anyone
329 * else. Once this happens, any attempt to call mark_buffer_dirty()
330 * (for example, by ext4_commit_super), will cause a kernel OOPS.
331 * This is a kludge to prevent these oops until we can put in a proper
332 * hook in del_gendisk() to inform the VFS and file system layers.
334 static int block_device_ejected(struct super_block *sb)
336 struct inode *bd_inode = sb->s_bdev->bd_inode;
337 struct backing_dev_info *bdi = bd_inode->i_mapping->backing_dev_info;
339 return bdi->dev == NULL;
342 static void ext4_journal_commit_callback(journal_t *journal, transaction_t *txn)
344 struct super_block *sb = journal->j_private;
345 struct ext4_sb_info *sbi = EXT4_SB(sb);
346 int error = is_journal_aborted(journal);
347 struct ext4_journal_cb_entry *jce;
349 BUG_ON(txn->t_state == T_FINISHED);
350 spin_lock(&sbi->s_md_lock);
351 while (!list_empty(&txn->t_private_list)) {
352 jce = list_entry(txn->t_private_list.next,
353 struct ext4_journal_cb_entry, jce_list);
354 list_del_init(&jce->jce_list);
355 spin_unlock(&sbi->s_md_lock);
356 jce->jce_func(sb, jce, error);
357 spin_lock(&sbi->s_md_lock);
359 spin_unlock(&sbi->s_md_lock);
362 /* Deal with the reporting of failure conditions on a filesystem such as
363 * inconsistencies detected or read IO failures.
365 * On ext2, we can store the error state of the filesystem in the
366 * superblock. That is not possible on ext4, because we may have other
367 * write ordering constraints on the superblock which prevent us from
368 * writing it out straight away; and given that the journal is about to
369 * be aborted, we can't rely on the current, or future, transactions to
370 * write out the superblock safely.
372 * We'll just use the jbd2_journal_abort() error code to record an error in
373 * the journal instead. On recovery, the journal will complain about
374 * that error until we've noted it down and cleared it.
377 static void ext4_handle_error(struct super_block *sb)
379 if (sb->s_flags & MS_RDONLY)
382 if (!test_opt(sb, ERRORS_CONT)) {
383 journal_t *journal = EXT4_SB(sb)->s_journal;
385 EXT4_SB(sb)->s_mount_flags |= EXT4_MF_FS_ABORTED;
387 jbd2_journal_abort(journal, -EIO);
389 if (test_opt(sb, ERRORS_RO)) {
390 ext4_msg(sb, KERN_CRIT, "Remounting filesystem read-only");
392 * Make sure updated value of ->s_mount_flags will be visible
393 * before ->s_flags update
396 sb->s_flags |= MS_RDONLY;
398 if (test_opt(sb, ERRORS_PANIC))
399 panic("EXT4-fs (device %s): panic forced after error\n",
403 #define ext4_error_ratelimit(sb) \
404 ___ratelimit(&(EXT4_SB(sb)->s_err_ratelimit_state), \
407 void __ext4_error(struct super_block *sb, const char *function,
408 unsigned int line, const char *fmt, ...)
410 struct va_format vaf;
413 if (ext4_error_ratelimit(sb)) {
418 "EXT4-fs error (device %s): %s:%d: comm %s: %pV\n",
419 sb->s_id, function, line, current->comm, &vaf);
422 save_error_info(sb, function, line);
423 ext4_handle_error(sb);
426 void __ext4_error_inode(struct inode *inode, const char *function,
427 unsigned int line, ext4_fsblk_t block,
428 const char *fmt, ...)
431 struct va_format vaf;
432 struct ext4_super_block *es = EXT4_SB(inode->i_sb)->s_es;
434 es->s_last_error_ino = cpu_to_le32(inode->i_ino);
435 es->s_last_error_block = cpu_to_le64(block);
436 if (ext4_error_ratelimit(inode->i_sb)) {
441 printk(KERN_CRIT "EXT4-fs error (device %s): %s:%d: "
442 "inode #%lu: block %llu: comm %s: %pV\n",
443 inode->i_sb->s_id, function, line, inode->i_ino,
444 block, current->comm, &vaf);
446 printk(KERN_CRIT "EXT4-fs error (device %s): %s:%d: "
447 "inode #%lu: comm %s: %pV\n",
448 inode->i_sb->s_id, function, line, inode->i_ino,
449 current->comm, &vaf);
452 save_error_info(inode->i_sb, function, line);
453 ext4_handle_error(inode->i_sb);
456 void __ext4_error_file(struct file *file, const char *function,
457 unsigned int line, ext4_fsblk_t block,
458 const char *fmt, ...)
461 struct va_format vaf;
462 struct ext4_super_block *es;
463 struct inode *inode = file_inode(file);
464 char pathname[80], *path;
466 es = EXT4_SB(inode->i_sb)->s_es;
467 es->s_last_error_ino = cpu_to_le32(inode->i_ino);
468 if (ext4_error_ratelimit(inode->i_sb)) {
469 path = d_path(&(file->f_path), pathname, sizeof(pathname));
477 "EXT4-fs error (device %s): %s:%d: inode #%lu: "
478 "block %llu: comm %s: path %s: %pV\n",
479 inode->i_sb->s_id, function, line, inode->i_ino,
480 block, current->comm, path, &vaf);
483 "EXT4-fs error (device %s): %s:%d: inode #%lu: "
484 "comm %s: path %s: %pV\n",
485 inode->i_sb->s_id, function, line, inode->i_ino,
486 current->comm, path, &vaf);
489 save_error_info(inode->i_sb, function, line);
490 ext4_handle_error(inode->i_sb);
493 const char *ext4_decode_error(struct super_block *sb, int errno,
500 errstr = "IO failure";
503 errstr = "Out of memory";
506 if (!sb || (EXT4_SB(sb)->s_journal &&
507 EXT4_SB(sb)->s_journal->j_flags & JBD2_ABORT))
508 errstr = "Journal has aborted";
510 errstr = "Readonly filesystem";
513 /* If the caller passed in an extra buffer for unknown
514 * errors, textualise them now. Else we just return
517 /* Check for truncated error codes... */
518 if (snprintf(nbuf, 16, "error %d", -errno) >= 0)
527 /* __ext4_std_error decodes expected errors from journaling functions
528 * automatically and invokes the appropriate error response. */
530 void __ext4_std_error(struct super_block *sb, const char *function,
531 unsigned int line, int errno)
536 /* Special case: if the error is EROFS, and we're not already
537 * inside a transaction, then there's really no point in logging
539 if (errno == -EROFS && journal_current_handle() == NULL &&
540 (sb->s_flags & MS_RDONLY))
543 if (ext4_error_ratelimit(sb)) {
544 errstr = ext4_decode_error(sb, errno, nbuf);
545 printk(KERN_CRIT "EXT4-fs error (device %s) in %s:%d: %s\n",
546 sb->s_id, function, line, errstr);
549 save_error_info(sb, function, line);
550 ext4_handle_error(sb);
554 * ext4_abort is a much stronger failure handler than ext4_error. The
555 * abort function may be used to deal with unrecoverable failures such
556 * as journal IO errors or ENOMEM at a critical moment in log management.
558 * We unconditionally force the filesystem into an ABORT|READONLY state,
559 * unless the error response on the fs has been set to panic in which
560 * case we take the easy way out and panic immediately.
563 void __ext4_abort(struct super_block *sb, const char *function,
564 unsigned int line, const char *fmt, ...)
568 save_error_info(sb, function, line);
570 printk(KERN_CRIT "EXT4-fs error (device %s): %s:%d: ", sb->s_id,
576 if ((sb->s_flags & MS_RDONLY) == 0) {
577 ext4_msg(sb, KERN_CRIT, "Remounting filesystem read-only");
578 EXT4_SB(sb)->s_mount_flags |= EXT4_MF_FS_ABORTED;
580 * Make sure updated value of ->s_mount_flags will be visible
581 * before ->s_flags update
584 sb->s_flags |= MS_RDONLY;
585 if (EXT4_SB(sb)->s_journal)
586 jbd2_journal_abort(EXT4_SB(sb)->s_journal, -EIO);
587 save_error_info(sb, function, line);
589 if (test_opt(sb, ERRORS_PANIC))
590 panic("EXT4-fs panic from previous error\n");
593 void __ext4_msg(struct super_block *sb,
594 const char *prefix, const char *fmt, ...)
596 struct va_format vaf;
599 if (!___ratelimit(&(EXT4_SB(sb)->s_msg_ratelimit_state), "EXT4-fs"))
605 printk("%sEXT4-fs (%s): %pV\n", prefix, sb->s_id, &vaf);
609 void __ext4_warning(struct super_block *sb, const char *function,
610 unsigned int line, const char *fmt, ...)
612 struct va_format vaf;
615 if (!___ratelimit(&(EXT4_SB(sb)->s_warning_ratelimit_state),
622 printk(KERN_WARNING "EXT4-fs warning (device %s): %s:%d: %pV\n",
623 sb->s_id, function, line, &vaf);
627 void __ext4_grp_locked_error(const char *function, unsigned int line,
628 struct super_block *sb, ext4_group_t grp,
629 unsigned long ino, ext4_fsblk_t block,
630 const char *fmt, ...)
634 struct va_format vaf;
636 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
638 es->s_last_error_ino = cpu_to_le32(ino);
639 es->s_last_error_block = cpu_to_le64(block);
640 __save_error_info(sb, function, line);
642 if (ext4_error_ratelimit(sb)) {
646 printk(KERN_CRIT "EXT4-fs error (device %s): %s:%d: group %u, ",
647 sb->s_id, function, line, grp);
649 printk(KERN_CONT "inode %lu: ", ino);
651 printk(KERN_CONT "block %llu:",
652 (unsigned long long) block);
653 printk(KERN_CONT "%pV\n", &vaf);
657 if (test_opt(sb, ERRORS_CONT)) {
658 ext4_commit_super(sb, 0);
662 ext4_unlock_group(sb, grp);
663 ext4_handle_error(sb);
665 * We only get here in the ERRORS_RO case; relocking the group
666 * may be dangerous, but nothing bad will happen since the
667 * filesystem will have already been marked read/only and the
668 * journal has been aborted. We return 1 as a hint to callers
669 * who might what to use the return value from
670 * ext4_grp_locked_error() to distinguish between the
671 * ERRORS_CONT and ERRORS_RO case, and perhaps return more
672 * aggressively from the ext4 function in question, with a
673 * more appropriate error code.
675 ext4_lock_group(sb, grp);
679 void ext4_update_dynamic_rev(struct super_block *sb)
681 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
683 if (le32_to_cpu(es->s_rev_level) > EXT4_GOOD_OLD_REV)
687 "updating to rev %d because of new feature flag, "
688 "running e2fsck is recommended",
691 es->s_first_ino = cpu_to_le32(EXT4_GOOD_OLD_FIRST_INO);
692 es->s_inode_size = cpu_to_le16(EXT4_GOOD_OLD_INODE_SIZE);
693 es->s_rev_level = cpu_to_le32(EXT4_DYNAMIC_REV);
694 /* leave es->s_feature_*compat flags alone */
695 /* es->s_uuid will be set by e2fsck if empty */
698 * The rest of the superblock fields should be zero, and if not it
699 * means they are likely already in use, so leave them alone. We
700 * can leave it up to e2fsck to clean up any inconsistencies there.
705 * Open the external journal device
707 static struct block_device *ext4_blkdev_get(dev_t dev, struct super_block *sb)
709 struct block_device *bdev;
710 char b[BDEVNAME_SIZE];
712 bdev = blkdev_get_by_dev(dev, FMODE_READ|FMODE_WRITE|FMODE_EXCL, sb);
718 ext4_msg(sb, KERN_ERR, "failed to open journal device %s: %ld",
719 __bdevname(dev, b), PTR_ERR(bdev));
724 * Release the journal device
726 static void ext4_blkdev_put(struct block_device *bdev)
728 blkdev_put(bdev, FMODE_READ|FMODE_WRITE|FMODE_EXCL);
731 static void ext4_blkdev_remove(struct ext4_sb_info *sbi)
733 struct block_device *bdev;
734 bdev = sbi->journal_bdev;
736 ext4_blkdev_put(bdev);
737 sbi->journal_bdev = NULL;
741 static inline struct inode *orphan_list_entry(struct list_head *l)
743 return &list_entry(l, struct ext4_inode_info, i_orphan)->vfs_inode;
746 static void dump_orphan_list(struct super_block *sb, struct ext4_sb_info *sbi)
750 ext4_msg(sb, KERN_ERR, "sb orphan head is %d",
751 le32_to_cpu(sbi->s_es->s_last_orphan));
753 printk(KERN_ERR "sb_info orphan list:\n");
754 list_for_each(l, &sbi->s_orphan) {
755 struct inode *inode = orphan_list_entry(l);
757 "inode %s:%lu at %p: mode %o, nlink %d, next %d\n",
758 inode->i_sb->s_id, inode->i_ino, inode,
759 inode->i_mode, inode->i_nlink,
764 static void ext4_put_super(struct super_block *sb)
766 struct ext4_sb_info *sbi = EXT4_SB(sb);
767 struct ext4_super_block *es = sbi->s_es;
770 ext4_unregister_li_request(sb);
771 dquot_disable(sb, -1, DQUOT_USAGE_ENABLED | DQUOT_LIMITS_ENABLED);
773 flush_workqueue(sbi->rsv_conversion_wq);
774 destroy_workqueue(sbi->rsv_conversion_wq);
776 if (sbi->s_journal) {
777 err = jbd2_journal_destroy(sbi->s_journal);
778 sbi->s_journal = NULL;
780 ext4_abort(sb, "Couldn't clean up the journal");
783 ext4_es_unregister_shrinker(sbi);
784 del_timer_sync(&sbi->s_err_report);
785 ext4_release_system_zone(sb);
787 ext4_ext_release(sb);
788 ext4_xattr_put_super(sb);
790 if (!(sb->s_flags & MS_RDONLY)) {
791 EXT4_CLEAR_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
792 es->s_state = cpu_to_le16(sbi->s_mount_state);
794 if (!(sb->s_flags & MS_RDONLY))
795 ext4_commit_super(sb, 1);
798 remove_proc_entry("options", sbi->s_proc);
799 remove_proc_entry(sb->s_id, ext4_proc_root);
801 kobject_del(&sbi->s_kobj);
803 for (i = 0; i < sbi->s_gdb_count; i++)
804 brelse(sbi->s_group_desc[i]);
805 kvfree(sbi->s_group_desc);
806 kvfree(sbi->s_flex_groups);
807 percpu_counter_destroy(&sbi->s_freeclusters_counter);
808 percpu_counter_destroy(&sbi->s_freeinodes_counter);
809 percpu_counter_destroy(&sbi->s_dirs_counter);
810 percpu_counter_destroy(&sbi->s_dirtyclusters_counter);
813 for (i = 0; i < EXT4_MAXQUOTAS; i++)
814 kfree(sbi->s_qf_names[i]);
817 /* Debugging code just in case the in-memory inode orphan list
818 * isn't empty. The on-disk one can be non-empty if we've
819 * detected an error and taken the fs readonly, but the
820 * in-memory list had better be clean by this point. */
821 if (!list_empty(&sbi->s_orphan))
822 dump_orphan_list(sb, sbi);
823 J_ASSERT(list_empty(&sbi->s_orphan));
825 invalidate_bdev(sb->s_bdev);
826 if (sbi->journal_bdev && sbi->journal_bdev != sb->s_bdev) {
828 * Invalidate the journal device's buffers. We don't want them
829 * floating about in memory - the physical journal device may
830 * hotswapped, and it breaks the `ro-after' testing code.
832 sync_blockdev(sbi->journal_bdev);
833 invalidate_bdev(sbi->journal_bdev);
834 ext4_blkdev_remove(sbi);
836 if (sbi->s_mb_cache) {
837 ext4_xattr_destroy_cache(sbi->s_mb_cache);
838 sbi->s_mb_cache = NULL;
841 kthread_stop(sbi->s_mmp_tsk);
842 sb->s_fs_info = NULL;
844 * Now that we are completely done shutting down the
845 * superblock, we need to actually destroy the kobject.
847 kobject_put(&sbi->s_kobj);
848 wait_for_completion(&sbi->s_kobj_unregister);
849 if (sbi->s_chksum_driver)
850 crypto_free_shash(sbi->s_chksum_driver);
851 kfree(sbi->s_blockgroup_lock);
855 static struct kmem_cache *ext4_inode_cachep;
858 * Called inside transaction, so use GFP_NOFS
860 static struct inode *ext4_alloc_inode(struct super_block *sb)
862 struct ext4_inode_info *ei;
864 ei = kmem_cache_alloc(ext4_inode_cachep, GFP_NOFS);
868 ei->vfs_inode.i_version = 1;
869 spin_lock_init(&ei->i_raw_lock);
870 INIT_LIST_HEAD(&ei->i_prealloc_list);
871 spin_lock_init(&ei->i_prealloc_lock);
872 ext4_es_init_tree(&ei->i_es_tree);
873 rwlock_init(&ei->i_es_lock);
874 INIT_LIST_HEAD(&ei->i_es_list);
877 ei->i_es_shrink_lblk = 0;
878 ei->i_reserved_data_blocks = 0;
879 ei->i_reserved_meta_blocks = 0;
880 ei->i_allocated_meta_blocks = 0;
881 ei->i_da_metadata_calc_len = 0;
882 ei->i_da_metadata_calc_last_lblock = 0;
883 spin_lock_init(&(ei->i_block_reservation_lock));
885 ei->i_reserved_quota = 0;
888 INIT_LIST_HEAD(&ei->i_rsv_conversion_list);
889 spin_lock_init(&ei->i_completed_io_lock);
891 ei->i_datasync_tid = 0;
892 atomic_set(&ei->i_ioend_count, 0);
893 atomic_set(&ei->i_unwritten, 0);
894 INIT_WORK(&ei->i_rsv_conversion_work, ext4_end_io_rsv_work);
896 return &ei->vfs_inode;
899 static int ext4_drop_inode(struct inode *inode)
901 int drop = generic_drop_inode(inode);
903 trace_ext4_drop_inode(inode, drop);
907 static void ext4_i_callback(struct rcu_head *head)
909 struct inode *inode = container_of(head, struct inode, i_rcu);
910 kmem_cache_free(ext4_inode_cachep, EXT4_I(inode));
913 static void ext4_destroy_inode(struct inode *inode)
915 if (!list_empty(&(EXT4_I(inode)->i_orphan))) {
916 ext4_msg(inode->i_sb, KERN_ERR,
917 "Inode %lu (%p): orphan list check failed!",
918 inode->i_ino, EXT4_I(inode));
919 print_hex_dump(KERN_INFO, "", DUMP_PREFIX_ADDRESS, 16, 4,
920 EXT4_I(inode), sizeof(struct ext4_inode_info),
924 call_rcu(&inode->i_rcu, ext4_i_callback);
927 static void init_once(void *foo)
929 struct ext4_inode_info *ei = (struct ext4_inode_info *) foo;
931 INIT_LIST_HEAD(&ei->i_orphan);
932 init_rwsem(&ei->xattr_sem);
933 init_rwsem(&ei->i_data_sem);
934 inode_init_once(&ei->vfs_inode);
937 static int __init init_inodecache(void)
939 ext4_inode_cachep = kmem_cache_create("ext4_inode_cache",
940 sizeof(struct ext4_inode_info),
941 0, (SLAB_RECLAIM_ACCOUNT|
944 if (ext4_inode_cachep == NULL)
949 static void destroy_inodecache(void)
952 * Make sure all delayed rcu free inodes are flushed before we
956 kmem_cache_destroy(ext4_inode_cachep);
959 void ext4_clear_inode(struct inode *inode)
961 invalidate_inode_buffers(inode);
964 ext4_discard_preallocations(inode);
965 ext4_es_remove_extent(inode, 0, EXT_MAX_BLOCKS);
966 if (EXT4_I(inode)->jinode) {
967 jbd2_journal_release_jbd_inode(EXT4_JOURNAL(inode),
968 EXT4_I(inode)->jinode);
969 jbd2_free_inode(EXT4_I(inode)->jinode);
970 EXT4_I(inode)->jinode = NULL;
974 static struct inode *ext4_nfs_get_inode(struct super_block *sb,
975 u64 ino, u32 generation)
979 if (ino < EXT4_FIRST_INO(sb) && ino != EXT4_ROOT_INO)
980 return ERR_PTR(-ESTALE);
981 if (ino > le32_to_cpu(EXT4_SB(sb)->s_es->s_inodes_count))
982 return ERR_PTR(-ESTALE);
984 /* iget isn't really right if the inode is currently unallocated!!
986 * ext4_read_inode will return a bad_inode if the inode had been
987 * deleted, so we should be safe.
989 * Currently we don't know the generation for parent directory, so
990 * a generation of 0 means "accept any"
992 inode = ext4_iget_normal(sb, ino);
994 return ERR_CAST(inode);
995 if (generation && inode->i_generation != generation) {
997 return ERR_PTR(-ESTALE);
1003 static struct dentry *ext4_fh_to_dentry(struct super_block *sb, struct fid *fid,
1004 int fh_len, int fh_type)
1006 return generic_fh_to_dentry(sb, fid, fh_len, fh_type,
1007 ext4_nfs_get_inode);
1010 static struct dentry *ext4_fh_to_parent(struct super_block *sb, struct fid *fid,
1011 int fh_len, int fh_type)
1013 return generic_fh_to_parent(sb, fid, fh_len, fh_type,
1014 ext4_nfs_get_inode);
1018 * Try to release metadata pages (indirect blocks, directories) which are
1019 * mapped via the block device. Since these pages could have journal heads
1020 * which would prevent try_to_free_buffers() from freeing them, we must use
1021 * jbd2 layer's try_to_free_buffers() function to release them.
1023 static int bdev_try_to_free_page(struct super_block *sb, struct page *page,
1026 journal_t *journal = EXT4_SB(sb)->s_journal;
1028 WARN_ON(PageChecked(page));
1029 if (!page_has_buffers(page))
1032 return jbd2_journal_try_to_free_buffers(journal, page,
1033 wait & ~__GFP_WAIT);
1034 return try_to_free_buffers(page);
1038 #define QTYPE2NAME(t) ((t) == USRQUOTA ? "user" : "group")
1039 #define QTYPE2MOPT(on, t) ((t) == USRQUOTA?((on)##USRJQUOTA):((on)##GRPJQUOTA))
1041 static int ext4_write_dquot(struct dquot *dquot);
1042 static int ext4_acquire_dquot(struct dquot *dquot);
1043 static int ext4_release_dquot(struct dquot *dquot);
1044 static int ext4_mark_dquot_dirty(struct dquot *dquot);
1045 static int ext4_write_info(struct super_block *sb, int type);
1046 static int ext4_quota_on(struct super_block *sb, int type, int format_id,
1048 static int ext4_quota_on_sysfile(struct super_block *sb, int type,
1050 static int ext4_quota_off(struct super_block *sb, int type);
1051 static int ext4_quota_off_sysfile(struct super_block *sb, int type);
1052 static int ext4_quota_on_mount(struct super_block *sb, int type);
1053 static ssize_t ext4_quota_read(struct super_block *sb, int type, char *data,
1054 size_t len, loff_t off);
1055 static ssize_t ext4_quota_write(struct super_block *sb, int type,
1056 const char *data, size_t len, loff_t off);
1057 static int ext4_quota_enable(struct super_block *sb, int type, int format_id,
1058 unsigned int flags);
1059 static int ext4_enable_quotas(struct super_block *sb);
1061 static const struct dquot_operations ext4_quota_operations = {
1062 .get_reserved_space = ext4_get_reserved_space,
1063 .write_dquot = ext4_write_dquot,
1064 .acquire_dquot = ext4_acquire_dquot,
1065 .release_dquot = ext4_release_dquot,
1066 .mark_dirty = ext4_mark_dquot_dirty,
1067 .write_info = ext4_write_info,
1068 .alloc_dquot = dquot_alloc,
1069 .destroy_dquot = dquot_destroy,
1072 static const struct quotactl_ops ext4_qctl_operations = {
1073 .quota_on = ext4_quota_on,
1074 .quota_off = ext4_quota_off,
1075 .quota_sync = dquot_quota_sync,
1076 .get_info = dquot_get_dqinfo,
1077 .set_info = dquot_set_dqinfo,
1078 .get_dqblk = dquot_get_dqblk,
1079 .set_dqblk = dquot_set_dqblk
1082 static const struct quotactl_ops ext4_qctl_sysfile_operations = {
1083 .quota_on_meta = ext4_quota_on_sysfile,
1084 .quota_off = ext4_quota_off_sysfile,
1085 .quota_sync = dquot_quota_sync,
1086 .get_info = dquot_get_dqinfo,
1087 .set_info = dquot_set_dqinfo,
1088 .get_dqblk = dquot_get_dqblk,
1089 .set_dqblk = dquot_set_dqblk
1093 static const struct super_operations ext4_sops = {
1094 .alloc_inode = ext4_alloc_inode,
1095 .destroy_inode = ext4_destroy_inode,
1096 .write_inode = ext4_write_inode,
1097 .dirty_inode = ext4_dirty_inode,
1098 .drop_inode = ext4_drop_inode,
1099 .evict_inode = ext4_evict_inode,
1100 .put_super = ext4_put_super,
1101 .sync_fs = ext4_sync_fs,
1102 .freeze_fs = ext4_freeze,
1103 .unfreeze_fs = ext4_unfreeze,
1104 .statfs = ext4_statfs,
1105 .remount_fs = ext4_remount,
1106 .show_options = ext4_show_options,
1108 .quota_read = ext4_quota_read,
1109 .quota_write = ext4_quota_write,
1111 .bdev_try_to_free_page = bdev_try_to_free_page,
1114 static const struct export_operations ext4_export_ops = {
1115 .fh_to_dentry = ext4_fh_to_dentry,
1116 .fh_to_parent = ext4_fh_to_parent,
1117 .get_parent = ext4_get_parent,
1121 Opt_bsd_df, Opt_minix_df, Opt_grpid, Opt_nogrpid,
1122 Opt_resgid, Opt_resuid, Opt_sb, Opt_err_cont, Opt_err_panic, Opt_err_ro,
1123 Opt_nouid32, Opt_debug, Opt_removed,
1124 Opt_user_xattr, Opt_nouser_xattr, Opt_acl, Opt_noacl,
1125 Opt_auto_da_alloc, Opt_noauto_da_alloc, Opt_noload,
1126 Opt_commit, Opt_min_batch_time, Opt_max_batch_time, Opt_journal_dev,
1127 Opt_journal_path, Opt_journal_checksum, Opt_journal_async_commit,
1128 Opt_abort, Opt_data_journal, Opt_data_ordered, Opt_data_writeback,
1129 Opt_data_err_abort, Opt_data_err_ignore,
1130 Opt_usrjquota, Opt_grpjquota, Opt_offusrjquota, Opt_offgrpjquota,
1131 Opt_jqfmt_vfsold, Opt_jqfmt_vfsv0, Opt_jqfmt_vfsv1, Opt_quota,
1132 Opt_noquota, Opt_barrier, Opt_nobarrier, Opt_err,
1133 Opt_usrquota, Opt_grpquota, Opt_i_version,
1134 Opt_stripe, Opt_delalloc, Opt_nodelalloc, Opt_mblk_io_submit,
1135 Opt_nomblk_io_submit, Opt_block_validity, Opt_noblock_validity,
1136 Opt_inode_readahead_blks, Opt_journal_ioprio,
1137 Opt_dioread_nolock, Opt_dioread_lock,
1138 Opt_discard, Opt_nodiscard, Opt_init_itable, Opt_noinit_itable,
1139 Opt_max_dir_size_kb, Opt_nojournal_checksum,
1142 static const match_table_t tokens = {
1143 {Opt_bsd_df, "bsddf"},
1144 {Opt_minix_df, "minixdf"},
1145 {Opt_grpid, "grpid"},
1146 {Opt_grpid, "bsdgroups"},
1147 {Opt_nogrpid, "nogrpid"},
1148 {Opt_nogrpid, "sysvgroups"},
1149 {Opt_resgid, "resgid=%u"},
1150 {Opt_resuid, "resuid=%u"},
1152 {Opt_err_cont, "errors=continue"},
1153 {Opt_err_panic, "errors=panic"},
1154 {Opt_err_ro, "errors=remount-ro"},
1155 {Opt_nouid32, "nouid32"},
1156 {Opt_debug, "debug"},
1157 {Opt_removed, "oldalloc"},
1158 {Opt_removed, "orlov"},
1159 {Opt_user_xattr, "user_xattr"},
1160 {Opt_nouser_xattr, "nouser_xattr"},
1162 {Opt_noacl, "noacl"},
1163 {Opt_noload, "norecovery"},
1164 {Opt_noload, "noload"},
1165 {Opt_removed, "nobh"},
1166 {Opt_removed, "bh"},
1167 {Opt_commit, "commit=%u"},
1168 {Opt_min_batch_time, "min_batch_time=%u"},
1169 {Opt_max_batch_time, "max_batch_time=%u"},
1170 {Opt_journal_dev, "journal_dev=%u"},
1171 {Opt_journal_path, "journal_path=%s"},
1172 {Opt_journal_checksum, "journal_checksum"},
1173 {Opt_nojournal_checksum, "nojournal_checksum"},
1174 {Opt_journal_async_commit, "journal_async_commit"},
1175 {Opt_abort, "abort"},
1176 {Opt_data_journal, "data=journal"},
1177 {Opt_data_ordered, "data=ordered"},
1178 {Opt_data_writeback, "data=writeback"},
1179 {Opt_data_err_abort, "data_err=abort"},
1180 {Opt_data_err_ignore, "data_err=ignore"},
1181 {Opt_offusrjquota, "usrjquota="},
1182 {Opt_usrjquota, "usrjquota=%s"},
1183 {Opt_offgrpjquota, "grpjquota="},
1184 {Opt_grpjquota, "grpjquota=%s"},
1185 {Opt_jqfmt_vfsold, "jqfmt=vfsold"},
1186 {Opt_jqfmt_vfsv0, "jqfmt=vfsv0"},
1187 {Opt_jqfmt_vfsv1, "jqfmt=vfsv1"},
1188 {Opt_grpquota, "grpquota"},
1189 {Opt_noquota, "noquota"},
1190 {Opt_quota, "quota"},
1191 {Opt_usrquota, "usrquota"},
1192 {Opt_barrier, "barrier=%u"},
1193 {Opt_barrier, "barrier"},
1194 {Opt_nobarrier, "nobarrier"},
1195 {Opt_i_version, "i_version"},
1196 {Opt_stripe, "stripe=%u"},
1197 {Opt_delalloc, "delalloc"},
1198 {Opt_nodelalloc, "nodelalloc"},
1199 {Opt_removed, "mblk_io_submit"},
1200 {Opt_removed, "nomblk_io_submit"},
1201 {Opt_block_validity, "block_validity"},
1202 {Opt_noblock_validity, "noblock_validity"},
1203 {Opt_inode_readahead_blks, "inode_readahead_blks=%u"},
1204 {Opt_journal_ioprio, "journal_ioprio=%u"},
1205 {Opt_auto_da_alloc, "auto_da_alloc=%u"},
1206 {Opt_auto_da_alloc, "auto_da_alloc"},
1207 {Opt_noauto_da_alloc, "noauto_da_alloc"},
1208 {Opt_dioread_nolock, "dioread_nolock"},
1209 {Opt_dioread_lock, "dioread_lock"},
1210 {Opt_discard, "discard"},
1211 {Opt_nodiscard, "nodiscard"},
1212 {Opt_init_itable, "init_itable=%u"},
1213 {Opt_init_itable, "init_itable"},
1214 {Opt_noinit_itable, "noinit_itable"},
1215 {Opt_max_dir_size_kb, "max_dir_size_kb=%u"},
1216 {Opt_removed, "check=none"}, /* mount option from ext2/3 */
1217 {Opt_removed, "nocheck"}, /* mount option from ext2/3 */
1218 {Opt_removed, "reservation"}, /* mount option from ext2/3 */
1219 {Opt_removed, "noreservation"}, /* mount option from ext2/3 */
1220 {Opt_removed, "journal=%u"}, /* mount option from ext2/3 */
1224 static ext4_fsblk_t get_sb_block(void **data)
1226 ext4_fsblk_t sb_block;
1227 char *options = (char *) *data;
1229 if (!options || strncmp(options, "sb=", 3) != 0)
1230 return 1; /* Default location */
1233 /* TODO: use simple_strtoll with >32bit ext4 */
1234 sb_block = simple_strtoul(options, &options, 0);
1235 if (*options && *options != ',') {
1236 printk(KERN_ERR "EXT4-fs: Invalid sb specification: %s\n",
1240 if (*options == ',')
1242 *data = (void *) options;
1247 #define DEFAULT_JOURNAL_IOPRIO (IOPRIO_PRIO_VALUE(IOPRIO_CLASS_BE, 3))
1248 static char deprecated_msg[] = "Mount option \"%s\" will be removed by %s\n"
1249 "Contact linux-ext4@vger.kernel.org if you think we should keep it.\n";
1252 static int set_qf_name(struct super_block *sb, int qtype, substring_t *args)
1254 struct ext4_sb_info *sbi = EXT4_SB(sb);
1258 if (sb_any_quota_loaded(sb) &&
1259 !sbi->s_qf_names[qtype]) {
1260 ext4_msg(sb, KERN_ERR,
1261 "Cannot change journaled "
1262 "quota options when quota turned on");
1265 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_QUOTA)) {
1266 ext4_msg(sb, KERN_ERR, "Cannot set journaled quota options "
1267 "when QUOTA feature is enabled");
1270 qname = match_strdup(args);
1272 ext4_msg(sb, KERN_ERR,
1273 "Not enough memory for storing quotafile name");
1276 if (sbi->s_qf_names[qtype]) {
1277 if (strcmp(sbi->s_qf_names[qtype], qname) == 0)
1280 ext4_msg(sb, KERN_ERR,
1281 "%s quota file already specified",
1285 if (strchr(qname, '/')) {
1286 ext4_msg(sb, KERN_ERR,
1287 "quotafile must be on filesystem root");
1290 sbi->s_qf_names[qtype] = qname;
1298 static int clear_qf_name(struct super_block *sb, int qtype)
1301 struct ext4_sb_info *sbi = EXT4_SB(sb);
1303 if (sb_any_quota_loaded(sb) &&
1304 sbi->s_qf_names[qtype]) {
1305 ext4_msg(sb, KERN_ERR, "Cannot change journaled quota options"
1306 " when quota turned on");
1309 kfree(sbi->s_qf_names[qtype]);
1310 sbi->s_qf_names[qtype] = NULL;
1315 #define MOPT_SET 0x0001
1316 #define MOPT_CLEAR 0x0002
1317 #define MOPT_NOSUPPORT 0x0004
1318 #define MOPT_EXPLICIT 0x0008
1319 #define MOPT_CLEAR_ERR 0x0010
1320 #define MOPT_GTE0 0x0020
1323 #define MOPT_QFMT 0x0040
1325 #define MOPT_Q MOPT_NOSUPPORT
1326 #define MOPT_QFMT MOPT_NOSUPPORT
1328 #define MOPT_DATAJ 0x0080
1329 #define MOPT_NO_EXT2 0x0100
1330 #define MOPT_NO_EXT3 0x0200
1331 #define MOPT_EXT4_ONLY (MOPT_NO_EXT2 | MOPT_NO_EXT3)
1332 #define MOPT_STRING 0x0400
1334 static const struct mount_opts {
1338 } ext4_mount_opts[] = {
1339 {Opt_minix_df, EXT4_MOUNT_MINIX_DF, MOPT_SET},
1340 {Opt_bsd_df, EXT4_MOUNT_MINIX_DF, MOPT_CLEAR},
1341 {Opt_grpid, EXT4_MOUNT_GRPID, MOPT_SET},
1342 {Opt_nogrpid, EXT4_MOUNT_GRPID, MOPT_CLEAR},
1343 {Opt_block_validity, EXT4_MOUNT_BLOCK_VALIDITY, MOPT_SET},
1344 {Opt_noblock_validity, EXT4_MOUNT_BLOCK_VALIDITY, MOPT_CLEAR},
1345 {Opt_dioread_nolock, EXT4_MOUNT_DIOREAD_NOLOCK,
1346 MOPT_EXT4_ONLY | MOPT_SET},
1347 {Opt_dioread_lock, EXT4_MOUNT_DIOREAD_NOLOCK,
1348 MOPT_EXT4_ONLY | MOPT_CLEAR},
1349 {Opt_discard, EXT4_MOUNT_DISCARD, MOPT_SET},
1350 {Opt_nodiscard, EXT4_MOUNT_DISCARD, MOPT_CLEAR},
1351 {Opt_delalloc, EXT4_MOUNT_DELALLOC,
1352 MOPT_EXT4_ONLY | MOPT_SET | MOPT_EXPLICIT},
1353 {Opt_nodelalloc, EXT4_MOUNT_DELALLOC,
1354 MOPT_EXT4_ONLY | MOPT_CLEAR},
1355 {Opt_nojournal_checksum, EXT4_MOUNT_JOURNAL_CHECKSUM,
1356 MOPT_EXT4_ONLY | MOPT_CLEAR},
1357 {Opt_journal_checksum, EXT4_MOUNT_JOURNAL_CHECKSUM,
1358 MOPT_EXT4_ONLY | MOPT_SET},
1359 {Opt_journal_async_commit, (EXT4_MOUNT_JOURNAL_ASYNC_COMMIT |
1360 EXT4_MOUNT_JOURNAL_CHECKSUM),
1361 MOPT_EXT4_ONLY | MOPT_SET},
1362 {Opt_noload, EXT4_MOUNT_NOLOAD, MOPT_NO_EXT2 | MOPT_SET},
1363 {Opt_err_panic, EXT4_MOUNT_ERRORS_PANIC, MOPT_SET | MOPT_CLEAR_ERR},
1364 {Opt_err_ro, EXT4_MOUNT_ERRORS_RO, MOPT_SET | MOPT_CLEAR_ERR},
1365 {Opt_err_cont, EXT4_MOUNT_ERRORS_CONT, MOPT_SET | MOPT_CLEAR_ERR},
1366 {Opt_data_err_abort, EXT4_MOUNT_DATA_ERR_ABORT,
1367 MOPT_NO_EXT2 | MOPT_SET},
1368 {Opt_data_err_ignore, EXT4_MOUNT_DATA_ERR_ABORT,
1369 MOPT_NO_EXT2 | MOPT_CLEAR},
1370 {Opt_barrier, EXT4_MOUNT_BARRIER, MOPT_SET},
1371 {Opt_nobarrier, EXT4_MOUNT_BARRIER, MOPT_CLEAR},
1372 {Opt_noauto_da_alloc, EXT4_MOUNT_NO_AUTO_DA_ALLOC, MOPT_SET},
1373 {Opt_auto_da_alloc, EXT4_MOUNT_NO_AUTO_DA_ALLOC, MOPT_CLEAR},
1374 {Opt_noinit_itable, EXT4_MOUNT_INIT_INODE_TABLE, MOPT_CLEAR},
1375 {Opt_commit, 0, MOPT_GTE0},
1376 {Opt_max_batch_time, 0, MOPT_GTE0},
1377 {Opt_min_batch_time, 0, MOPT_GTE0},
1378 {Opt_inode_readahead_blks, 0, MOPT_GTE0},
1379 {Opt_init_itable, 0, MOPT_GTE0},
1380 {Opt_stripe, 0, MOPT_GTE0},
1381 {Opt_resuid, 0, MOPT_GTE0},
1382 {Opt_resgid, 0, MOPT_GTE0},
1383 {Opt_journal_dev, 0, MOPT_GTE0},
1384 {Opt_journal_path, 0, MOPT_STRING},
1385 {Opt_journal_ioprio, 0, MOPT_GTE0},
1386 {Opt_data_journal, EXT4_MOUNT_JOURNAL_DATA, MOPT_NO_EXT2 | MOPT_DATAJ},
1387 {Opt_data_ordered, EXT4_MOUNT_ORDERED_DATA, MOPT_NO_EXT2 | MOPT_DATAJ},
1388 {Opt_data_writeback, EXT4_MOUNT_WRITEBACK_DATA,
1389 MOPT_NO_EXT2 | MOPT_DATAJ},
1390 {Opt_user_xattr, EXT4_MOUNT_XATTR_USER, MOPT_SET},
1391 {Opt_nouser_xattr, EXT4_MOUNT_XATTR_USER, MOPT_CLEAR},
1392 #ifdef CONFIG_EXT4_FS_POSIX_ACL
1393 {Opt_acl, EXT4_MOUNT_POSIX_ACL, MOPT_SET},
1394 {Opt_noacl, EXT4_MOUNT_POSIX_ACL, MOPT_CLEAR},
1396 {Opt_acl, 0, MOPT_NOSUPPORT},
1397 {Opt_noacl, 0, MOPT_NOSUPPORT},
1399 {Opt_nouid32, EXT4_MOUNT_NO_UID32, MOPT_SET},
1400 {Opt_debug, EXT4_MOUNT_DEBUG, MOPT_SET},
1401 {Opt_quota, EXT4_MOUNT_QUOTA | EXT4_MOUNT_USRQUOTA, MOPT_SET | MOPT_Q},
1402 {Opt_usrquota, EXT4_MOUNT_QUOTA | EXT4_MOUNT_USRQUOTA,
1404 {Opt_grpquota, EXT4_MOUNT_QUOTA | EXT4_MOUNT_GRPQUOTA,
1406 {Opt_noquota, (EXT4_MOUNT_QUOTA | EXT4_MOUNT_USRQUOTA |
1407 EXT4_MOUNT_GRPQUOTA), MOPT_CLEAR | MOPT_Q},
1408 {Opt_usrjquota, 0, MOPT_Q},
1409 {Opt_grpjquota, 0, MOPT_Q},
1410 {Opt_offusrjquota, 0, MOPT_Q},
1411 {Opt_offgrpjquota, 0, MOPT_Q},
1412 {Opt_jqfmt_vfsold, QFMT_VFS_OLD, MOPT_QFMT},
1413 {Opt_jqfmt_vfsv0, QFMT_VFS_V0, MOPT_QFMT},
1414 {Opt_jqfmt_vfsv1, QFMT_VFS_V1, MOPT_QFMT},
1415 {Opt_max_dir_size_kb, 0, MOPT_GTE0},
1419 static int handle_mount_opt(struct super_block *sb, char *opt, int token,
1420 substring_t *args, unsigned long *journal_devnum,
1421 unsigned int *journal_ioprio, int is_remount)
1423 struct ext4_sb_info *sbi = EXT4_SB(sb);
1424 const struct mount_opts *m;
1430 if (token == Opt_usrjquota)
1431 return set_qf_name(sb, USRQUOTA, &args[0]);
1432 else if (token == Opt_grpjquota)
1433 return set_qf_name(sb, GRPQUOTA, &args[0]);
1434 else if (token == Opt_offusrjquota)
1435 return clear_qf_name(sb, USRQUOTA);
1436 else if (token == Opt_offgrpjquota)
1437 return clear_qf_name(sb, GRPQUOTA);
1441 case Opt_nouser_xattr:
1442 ext4_msg(sb, KERN_WARNING, deprecated_msg, opt, "3.5");
1445 return 1; /* handled by get_sb_block() */
1447 ext4_msg(sb, KERN_WARNING, "Ignoring removed %s option", opt);
1450 sbi->s_mount_flags |= EXT4_MF_FS_ABORTED;
1453 sb->s_flags |= MS_I_VERSION;
1457 for (m = ext4_mount_opts; m->token != Opt_err; m++)
1458 if (token == m->token)
1461 if (m->token == Opt_err) {
1462 ext4_msg(sb, KERN_ERR, "Unrecognized mount option \"%s\" "
1463 "or missing value", opt);
1467 if ((m->flags & MOPT_NO_EXT2) && IS_EXT2_SB(sb)) {
1468 ext4_msg(sb, KERN_ERR,
1469 "Mount option \"%s\" incompatible with ext2", opt);
1472 if ((m->flags & MOPT_NO_EXT3) && IS_EXT3_SB(sb)) {
1473 ext4_msg(sb, KERN_ERR,
1474 "Mount option \"%s\" incompatible with ext3", opt);
1478 if (args->from && !(m->flags & MOPT_STRING) && match_int(args, &arg))
1480 if (args->from && (m->flags & MOPT_GTE0) && (arg < 0))
1482 if (m->flags & MOPT_EXPLICIT)
1483 set_opt2(sb, EXPLICIT_DELALLOC);
1484 if (m->flags & MOPT_CLEAR_ERR)
1485 clear_opt(sb, ERRORS_MASK);
1486 if (token == Opt_noquota && sb_any_quota_loaded(sb)) {
1487 ext4_msg(sb, KERN_ERR, "Cannot change quota "
1488 "options when quota turned on");
1492 if (m->flags & MOPT_NOSUPPORT) {
1493 ext4_msg(sb, KERN_ERR, "%s option not supported", opt);
1494 } else if (token == Opt_commit) {
1496 arg = JBD2_DEFAULT_MAX_COMMIT_AGE;
1497 sbi->s_commit_interval = HZ * arg;
1498 } else if (token == Opt_max_batch_time) {
1499 sbi->s_max_batch_time = arg;
1500 } else if (token == Opt_min_batch_time) {
1501 sbi->s_min_batch_time = arg;
1502 } else if (token == Opt_inode_readahead_blks) {
1503 if (arg && (arg > (1 << 30) || !is_power_of_2(arg))) {
1504 ext4_msg(sb, KERN_ERR,
1505 "EXT4-fs: inode_readahead_blks must be "
1506 "0 or a power of 2 smaller than 2^31");
1509 sbi->s_inode_readahead_blks = arg;
1510 } else if (token == Opt_init_itable) {
1511 set_opt(sb, INIT_INODE_TABLE);
1513 arg = EXT4_DEF_LI_WAIT_MULT;
1514 sbi->s_li_wait_mult = arg;
1515 } else if (token == Opt_max_dir_size_kb) {
1516 sbi->s_max_dir_size_kb = arg;
1517 } else if (token == Opt_stripe) {
1518 sbi->s_stripe = arg;
1519 } else if (token == Opt_resuid) {
1520 uid = make_kuid(current_user_ns(), arg);
1521 if (!uid_valid(uid)) {
1522 ext4_msg(sb, KERN_ERR, "Invalid uid value %d", arg);
1525 sbi->s_resuid = uid;
1526 } else if (token == Opt_resgid) {
1527 gid = make_kgid(current_user_ns(), arg);
1528 if (!gid_valid(gid)) {
1529 ext4_msg(sb, KERN_ERR, "Invalid gid value %d", arg);
1532 sbi->s_resgid = gid;
1533 } else if (token == Opt_journal_dev) {
1535 ext4_msg(sb, KERN_ERR,
1536 "Cannot specify journal on remount");
1539 *journal_devnum = arg;
1540 } else if (token == Opt_journal_path) {
1542 struct inode *journal_inode;
1547 ext4_msg(sb, KERN_ERR,
1548 "Cannot specify journal on remount");
1551 journal_path = match_strdup(&args[0]);
1552 if (!journal_path) {
1553 ext4_msg(sb, KERN_ERR, "error: could not dup "
1554 "journal device string");
1558 error = kern_path(journal_path, LOOKUP_FOLLOW, &path);
1560 ext4_msg(sb, KERN_ERR, "error: could not find "
1561 "journal device path: error %d", error);
1562 kfree(journal_path);
1566 journal_inode = path.dentry->d_inode;
1567 if (!S_ISBLK(journal_inode->i_mode)) {
1568 ext4_msg(sb, KERN_ERR, "error: journal path %s "
1569 "is not a block device", journal_path);
1571 kfree(journal_path);
1575 *journal_devnum = new_encode_dev(journal_inode->i_rdev);
1577 kfree(journal_path);
1578 } else if (token == Opt_journal_ioprio) {
1580 ext4_msg(sb, KERN_ERR, "Invalid journal IO priority"
1585 IOPRIO_PRIO_VALUE(IOPRIO_CLASS_BE, arg);
1586 } else if (m->flags & MOPT_DATAJ) {
1588 if (!sbi->s_journal)
1589 ext4_msg(sb, KERN_WARNING, "Remounting file system with no journal so ignoring journalled data option");
1590 else if (test_opt(sb, DATA_FLAGS) != m->mount_opt) {
1591 ext4_msg(sb, KERN_ERR,
1592 "Cannot change data mode on remount");
1596 clear_opt(sb, DATA_FLAGS);
1597 sbi->s_mount_opt |= m->mount_opt;
1600 } else if (m->flags & MOPT_QFMT) {
1601 if (sb_any_quota_loaded(sb) &&
1602 sbi->s_jquota_fmt != m->mount_opt) {
1603 ext4_msg(sb, KERN_ERR, "Cannot change journaled "
1604 "quota options when quota turned on");
1607 if (EXT4_HAS_RO_COMPAT_FEATURE(sb,
1608 EXT4_FEATURE_RO_COMPAT_QUOTA)) {
1609 ext4_msg(sb, KERN_ERR,
1610 "Cannot set journaled quota options "
1611 "when QUOTA feature is enabled");
1614 sbi->s_jquota_fmt = m->mount_opt;
1619 if (m->flags & MOPT_CLEAR)
1621 else if (unlikely(!(m->flags & MOPT_SET))) {
1622 ext4_msg(sb, KERN_WARNING,
1623 "buggy handling of option %s", opt);
1628 sbi->s_mount_opt |= m->mount_opt;
1630 sbi->s_mount_opt &= ~m->mount_opt;
1635 static int parse_options(char *options, struct super_block *sb,
1636 unsigned long *journal_devnum,
1637 unsigned int *journal_ioprio,
1640 struct ext4_sb_info *sbi = EXT4_SB(sb);
1642 substring_t args[MAX_OPT_ARGS];
1648 while ((p = strsep(&options, ",")) != NULL) {
1652 * Initialize args struct so we know whether arg was
1653 * found; some options take optional arguments.
1655 args[0].to = args[0].from = NULL;
1656 token = match_token(p, tokens, args);
1657 if (handle_mount_opt(sb, p, token, args, journal_devnum,
1658 journal_ioprio, is_remount) < 0)
1662 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_QUOTA) &&
1663 (test_opt(sb, USRQUOTA) || test_opt(sb, GRPQUOTA))) {
1664 ext4_msg(sb, KERN_ERR, "Cannot set quota options when QUOTA "
1665 "feature is enabled");
1668 if (sbi->s_qf_names[USRQUOTA] || sbi->s_qf_names[GRPQUOTA]) {
1669 if (test_opt(sb, USRQUOTA) && sbi->s_qf_names[USRQUOTA])
1670 clear_opt(sb, USRQUOTA);
1672 if (test_opt(sb, GRPQUOTA) && sbi->s_qf_names[GRPQUOTA])
1673 clear_opt(sb, GRPQUOTA);
1675 if (test_opt(sb, GRPQUOTA) || test_opt(sb, USRQUOTA)) {
1676 ext4_msg(sb, KERN_ERR, "old and new quota "
1681 if (!sbi->s_jquota_fmt) {
1682 ext4_msg(sb, KERN_ERR, "journaled quota format "
1688 if (test_opt(sb, DIOREAD_NOLOCK)) {
1690 BLOCK_SIZE << le32_to_cpu(sbi->s_es->s_log_block_size);
1692 if (blocksize < PAGE_CACHE_SIZE) {
1693 ext4_msg(sb, KERN_ERR, "can't mount with "
1694 "dioread_nolock if block size != PAGE_SIZE");
1698 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA &&
1699 test_opt(sb, JOURNAL_ASYNC_COMMIT)) {
1700 ext4_msg(sb, KERN_ERR, "can't mount with journal_async_commit "
1701 "in data=ordered mode");
1707 static inline void ext4_show_quota_options(struct seq_file *seq,
1708 struct super_block *sb)
1710 #if defined(CONFIG_QUOTA)
1711 struct ext4_sb_info *sbi = EXT4_SB(sb);
1713 if (sbi->s_jquota_fmt) {
1716 switch (sbi->s_jquota_fmt) {
1727 seq_printf(seq, ",jqfmt=%s", fmtname);
1730 if (sbi->s_qf_names[USRQUOTA])
1731 seq_printf(seq, ",usrjquota=%s", sbi->s_qf_names[USRQUOTA]);
1733 if (sbi->s_qf_names[GRPQUOTA])
1734 seq_printf(seq, ",grpjquota=%s", sbi->s_qf_names[GRPQUOTA]);
1738 static const char *token2str(int token)
1740 const struct match_token *t;
1742 for (t = tokens; t->token != Opt_err; t++)
1743 if (t->token == token && !strchr(t->pattern, '='))
1750 * - it's set to a non-default value OR
1751 * - if the per-sb default is different from the global default
1753 static int _ext4_show_options(struct seq_file *seq, struct super_block *sb,
1756 struct ext4_sb_info *sbi = EXT4_SB(sb);
1757 struct ext4_super_block *es = sbi->s_es;
1758 int def_errors, def_mount_opt = nodefs ? 0 : sbi->s_def_mount_opt;
1759 const struct mount_opts *m;
1760 char sep = nodefs ? '\n' : ',';
1762 #define SEQ_OPTS_PUTS(str) seq_printf(seq, "%c" str, sep)
1763 #define SEQ_OPTS_PRINT(str, arg) seq_printf(seq, "%c" str, sep, arg)
1765 if (sbi->s_sb_block != 1)
1766 SEQ_OPTS_PRINT("sb=%llu", sbi->s_sb_block);
1768 for (m = ext4_mount_opts; m->token != Opt_err; m++) {
1769 int want_set = m->flags & MOPT_SET;
1770 if (((m->flags & (MOPT_SET|MOPT_CLEAR)) == 0) ||
1771 (m->flags & MOPT_CLEAR_ERR))
1773 if (!(m->mount_opt & (sbi->s_mount_opt ^ def_mount_opt)))
1774 continue; /* skip if same as the default */
1776 (sbi->s_mount_opt & m->mount_opt) != m->mount_opt) ||
1777 (!want_set && (sbi->s_mount_opt & m->mount_opt)))
1778 continue; /* select Opt_noFoo vs Opt_Foo */
1779 SEQ_OPTS_PRINT("%s", token2str(m->token));
1782 if (nodefs || !uid_eq(sbi->s_resuid, make_kuid(&init_user_ns, EXT4_DEF_RESUID)) ||
1783 le16_to_cpu(es->s_def_resuid) != EXT4_DEF_RESUID)
1784 SEQ_OPTS_PRINT("resuid=%u",
1785 from_kuid_munged(&init_user_ns, sbi->s_resuid));
1786 if (nodefs || !gid_eq(sbi->s_resgid, make_kgid(&init_user_ns, EXT4_DEF_RESGID)) ||
1787 le16_to_cpu(es->s_def_resgid) != EXT4_DEF_RESGID)
1788 SEQ_OPTS_PRINT("resgid=%u",
1789 from_kgid_munged(&init_user_ns, sbi->s_resgid));
1790 def_errors = nodefs ? -1 : le16_to_cpu(es->s_errors);
1791 if (test_opt(sb, ERRORS_RO) && def_errors != EXT4_ERRORS_RO)
1792 SEQ_OPTS_PUTS("errors=remount-ro");
1793 if (test_opt(sb, ERRORS_CONT) && def_errors != EXT4_ERRORS_CONTINUE)
1794 SEQ_OPTS_PUTS("errors=continue");
1795 if (test_opt(sb, ERRORS_PANIC) && def_errors != EXT4_ERRORS_PANIC)
1796 SEQ_OPTS_PUTS("errors=panic");
1797 if (nodefs || sbi->s_commit_interval != JBD2_DEFAULT_MAX_COMMIT_AGE*HZ)
1798 SEQ_OPTS_PRINT("commit=%lu", sbi->s_commit_interval / HZ);
1799 if (nodefs || sbi->s_min_batch_time != EXT4_DEF_MIN_BATCH_TIME)
1800 SEQ_OPTS_PRINT("min_batch_time=%u", sbi->s_min_batch_time);
1801 if (nodefs || sbi->s_max_batch_time != EXT4_DEF_MAX_BATCH_TIME)
1802 SEQ_OPTS_PRINT("max_batch_time=%u", sbi->s_max_batch_time);
1803 if (sb->s_flags & MS_I_VERSION)
1804 SEQ_OPTS_PUTS("i_version");
1805 if (nodefs || sbi->s_stripe)
1806 SEQ_OPTS_PRINT("stripe=%lu", sbi->s_stripe);
1807 if (EXT4_MOUNT_DATA_FLAGS & (sbi->s_mount_opt ^ def_mount_opt)) {
1808 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA)
1809 SEQ_OPTS_PUTS("data=journal");
1810 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA)
1811 SEQ_OPTS_PUTS("data=ordered");
1812 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_WRITEBACK_DATA)
1813 SEQ_OPTS_PUTS("data=writeback");
1816 sbi->s_inode_readahead_blks != EXT4_DEF_INODE_READAHEAD_BLKS)
1817 SEQ_OPTS_PRINT("inode_readahead_blks=%u",
1818 sbi->s_inode_readahead_blks);
1820 if (nodefs || (test_opt(sb, INIT_INODE_TABLE) &&
1821 (sbi->s_li_wait_mult != EXT4_DEF_LI_WAIT_MULT)))
1822 SEQ_OPTS_PRINT("init_itable=%u", sbi->s_li_wait_mult);
1823 if (nodefs || sbi->s_max_dir_size_kb)
1824 SEQ_OPTS_PRINT("max_dir_size_kb=%u", sbi->s_max_dir_size_kb);
1826 ext4_show_quota_options(seq, sb);
1830 static int ext4_show_options(struct seq_file *seq, struct dentry *root)
1832 return _ext4_show_options(seq, root->d_sb, 0);
1835 static int options_seq_show(struct seq_file *seq, void *offset)
1837 struct super_block *sb = seq->private;
1840 seq_puts(seq, (sb->s_flags & MS_RDONLY) ? "ro" : "rw");
1841 rc = _ext4_show_options(seq, sb, 1);
1842 seq_puts(seq, "\n");
1846 static int options_open_fs(struct inode *inode, struct file *file)
1848 return single_open(file, options_seq_show, PDE_DATA(inode));
1851 static const struct file_operations ext4_seq_options_fops = {
1852 .owner = THIS_MODULE,
1853 .open = options_open_fs,
1855 .llseek = seq_lseek,
1856 .release = single_release,
1859 static int ext4_setup_super(struct super_block *sb, struct ext4_super_block *es,
1862 struct ext4_sb_info *sbi = EXT4_SB(sb);
1865 if (le32_to_cpu(es->s_rev_level) > EXT4_MAX_SUPP_REV) {
1866 ext4_msg(sb, KERN_ERR, "revision level too high, "
1867 "forcing read-only mode");
1872 if (!(sbi->s_mount_state & EXT4_VALID_FS))
1873 ext4_msg(sb, KERN_WARNING, "warning: mounting unchecked fs, "
1874 "running e2fsck is recommended");
1875 else if (sbi->s_mount_state & EXT4_ERROR_FS)
1876 ext4_msg(sb, KERN_WARNING,
1877 "warning: mounting fs with errors, "
1878 "running e2fsck is recommended");
1879 else if ((__s16) le16_to_cpu(es->s_max_mnt_count) > 0 &&
1880 le16_to_cpu(es->s_mnt_count) >=
1881 (unsigned short) (__s16) le16_to_cpu(es->s_max_mnt_count))
1882 ext4_msg(sb, KERN_WARNING,
1883 "warning: maximal mount count reached, "
1884 "running e2fsck is recommended");
1885 else if (le32_to_cpu(es->s_checkinterval) &&
1886 (le32_to_cpu(es->s_lastcheck) +
1887 le32_to_cpu(es->s_checkinterval) <= get_seconds()))
1888 ext4_msg(sb, KERN_WARNING,
1889 "warning: checktime reached, "
1890 "running e2fsck is recommended");
1891 if (!sbi->s_journal)
1892 es->s_state &= cpu_to_le16(~EXT4_VALID_FS);
1893 if (!(__s16) le16_to_cpu(es->s_max_mnt_count))
1894 es->s_max_mnt_count = cpu_to_le16(EXT4_DFL_MAX_MNT_COUNT);
1895 le16_add_cpu(&es->s_mnt_count, 1);
1896 es->s_mtime = cpu_to_le32(get_seconds());
1897 ext4_update_dynamic_rev(sb);
1899 EXT4_SET_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
1901 ext4_commit_super(sb, 1);
1903 if (test_opt(sb, DEBUG))
1904 printk(KERN_INFO "[EXT4 FS bs=%lu, gc=%u, "
1905 "bpg=%lu, ipg=%lu, mo=%04x, mo2=%04x]\n",
1907 sbi->s_groups_count,
1908 EXT4_BLOCKS_PER_GROUP(sb),
1909 EXT4_INODES_PER_GROUP(sb),
1910 sbi->s_mount_opt, sbi->s_mount_opt2);
1912 cleancache_init_fs(sb);
1916 int ext4_alloc_flex_bg_array(struct super_block *sb, ext4_group_t ngroup)
1918 struct ext4_sb_info *sbi = EXT4_SB(sb);
1919 struct flex_groups *new_groups;
1922 if (!sbi->s_log_groups_per_flex)
1925 size = ext4_flex_group(sbi, ngroup - 1) + 1;
1926 if (size <= sbi->s_flex_groups_allocated)
1929 size = roundup_pow_of_two(size * sizeof(struct flex_groups));
1930 new_groups = ext4_kvzalloc(size, GFP_KERNEL);
1932 ext4_msg(sb, KERN_ERR, "not enough memory for %d flex groups",
1933 size / (int) sizeof(struct flex_groups));
1937 if (sbi->s_flex_groups) {
1938 memcpy(new_groups, sbi->s_flex_groups,
1939 (sbi->s_flex_groups_allocated *
1940 sizeof(struct flex_groups)));
1941 kvfree(sbi->s_flex_groups);
1943 sbi->s_flex_groups = new_groups;
1944 sbi->s_flex_groups_allocated = size / sizeof(struct flex_groups);
1948 static int ext4_fill_flex_info(struct super_block *sb)
1950 struct ext4_sb_info *sbi = EXT4_SB(sb);
1951 struct ext4_group_desc *gdp = NULL;
1952 ext4_group_t flex_group;
1955 sbi->s_log_groups_per_flex = sbi->s_es->s_log_groups_per_flex;
1956 if (sbi->s_log_groups_per_flex < 1 || sbi->s_log_groups_per_flex > 31) {
1957 sbi->s_log_groups_per_flex = 0;
1961 err = ext4_alloc_flex_bg_array(sb, sbi->s_groups_count);
1965 for (i = 0; i < sbi->s_groups_count; i++) {
1966 gdp = ext4_get_group_desc(sb, i, NULL);
1968 flex_group = ext4_flex_group(sbi, i);
1969 atomic_add(ext4_free_inodes_count(sb, gdp),
1970 &sbi->s_flex_groups[flex_group].free_inodes);
1971 atomic64_add(ext4_free_group_clusters(sb, gdp),
1972 &sbi->s_flex_groups[flex_group].free_clusters);
1973 atomic_add(ext4_used_dirs_count(sb, gdp),
1974 &sbi->s_flex_groups[flex_group].used_dirs);
1982 static __le16 ext4_group_desc_csum(struct ext4_sb_info *sbi, __u32 block_group,
1983 struct ext4_group_desc *gdp)
1987 __le32 le_group = cpu_to_le32(block_group);
1989 if (ext4_has_metadata_csum(sbi->s_sb)) {
1990 /* Use new metadata_csum algorithm */
1994 save_csum = gdp->bg_checksum;
1995 gdp->bg_checksum = 0;
1996 csum32 = ext4_chksum(sbi, sbi->s_csum_seed, (__u8 *)&le_group,
1998 csum32 = ext4_chksum(sbi, csum32, (__u8 *)gdp,
2000 gdp->bg_checksum = save_csum;
2002 crc = csum32 & 0xFFFF;
2006 /* old crc16 code */
2007 if (!(sbi->s_es->s_feature_ro_compat &
2008 cpu_to_le32(EXT4_FEATURE_RO_COMPAT_GDT_CSUM)))
2011 offset = offsetof(struct ext4_group_desc, bg_checksum);
2013 crc = crc16(~0, sbi->s_es->s_uuid, sizeof(sbi->s_es->s_uuid));
2014 crc = crc16(crc, (__u8 *)&le_group, sizeof(le_group));
2015 crc = crc16(crc, (__u8 *)gdp, offset);
2016 offset += sizeof(gdp->bg_checksum); /* skip checksum */
2017 /* for checksum of struct ext4_group_desc do the rest...*/
2018 if ((sbi->s_es->s_feature_incompat &
2019 cpu_to_le32(EXT4_FEATURE_INCOMPAT_64BIT)) &&
2020 offset < le16_to_cpu(sbi->s_es->s_desc_size))
2021 crc = crc16(crc, (__u8 *)gdp + offset,
2022 le16_to_cpu(sbi->s_es->s_desc_size) -
2026 return cpu_to_le16(crc);
2029 int ext4_group_desc_csum_verify(struct super_block *sb, __u32 block_group,
2030 struct ext4_group_desc *gdp)
2032 if (ext4_has_group_desc_csum(sb) &&
2033 (gdp->bg_checksum != ext4_group_desc_csum(EXT4_SB(sb),
2040 void ext4_group_desc_csum_set(struct super_block *sb, __u32 block_group,
2041 struct ext4_group_desc *gdp)
2043 if (!ext4_has_group_desc_csum(sb))
2045 gdp->bg_checksum = ext4_group_desc_csum(EXT4_SB(sb), block_group, gdp);
2048 /* Called at mount-time, super-block is locked */
2049 static int ext4_check_descriptors(struct super_block *sb,
2050 ext4_group_t *first_not_zeroed)
2052 struct ext4_sb_info *sbi = EXT4_SB(sb);
2053 ext4_fsblk_t first_block = le32_to_cpu(sbi->s_es->s_first_data_block);
2054 ext4_fsblk_t last_block;
2055 ext4_fsblk_t block_bitmap;
2056 ext4_fsblk_t inode_bitmap;
2057 ext4_fsblk_t inode_table;
2058 int flexbg_flag = 0;
2059 ext4_group_t i, grp = sbi->s_groups_count;
2061 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_FLEX_BG))
2064 ext4_debug("Checking group descriptors");
2066 for (i = 0; i < sbi->s_groups_count; i++) {
2067 struct ext4_group_desc *gdp = ext4_get_group_desc(sb, i, NULL);
2069 if (i == sbi->s_groups_count - 1 || flexbg_flag)
2070 last_block = ext4_blocks_count(sbi->s_es) - 1;
2072 last_block = first_block +
2073 (EXT4_BLOCKS_PER_GROUP(sb) - 1);
2075 if ((grp == sbi->s_groups_count) &&
2076 !(gdp->bg_flags & cpu_to_le16(EXT4_BG_INODE_ZEROED)))
2079 block_bitmap = ext4_block_bitmap(sb, gdp);
2080 if (block_bitmap < first_block || block_bitmap > last_block) {
2081 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2082 "Block bitmap for group %u not in group "
2083 "(block %llu)!", i, block_bitmap);
2086 inode_bitmap = ext4_inode_bitmap(sb, gdp);
2087 if (inode_bitmap < first_block || inode_bitmap > last_block) {
2088 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2089 "Inode bitmap for group %u not in group "
2090 "(block %llu)!", i, inode_bitmap);
2093 inode_table = ext4_inode_table(sb, gdp);
2094 if (inode_table < first_block ||
2095 inode_table + sbi->s_itb_per_group - 1 > last_block) {
2096 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2097 "Inode table for group %u not in group "
2098 "(block %llu)!", i, inode_table);
2101 ext4_lock_group(sb, i);
2102 if (!ext4_group_desc_csum_verify(sb, i, gdp)) {
2103 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2104 "Checksum for group %u failed (%u!=%u)",
2105 i, le16_to_cpu(ext4_group_desc_csum(sbi, i,
2106 gdp)), le16_to_cpu(gdp->bg_checksum));
2107 if (!(sb->s_flags & MS_RDONLY)) {
2108 ext4_unlock_group(sb, i);
2112 ext4_unlock_group(sb, i);
2114 first_block += EXT4_BLOCKS_PER_GROUP(sb);
2116 if (NULL != first_not_zeroed)
2117 *first_not_zeroed = grp;
2121 /* ext4_orphan_cleanup() walks a singly-linked list of inodes (starting at
2122 * the superblock) which were deleted from all directories, but held open by
2123 * a process at the time of a crash. We walk the list and try to delete these
2124 * inodes at recovery time (only with a read-write filesystem).
2126 * In order to keep the orphan inode chain consistent during traversal (in
2127 * case of crash during recovery), we link each inode into the superblock
2128 * orphan list_head and handle it the same way as an inode deletion during
2129 * normal operation (which journals the operations for us).
2131 * We only do an iget() and an iput() on each inode, which is very safe if we
2132 * accidentally point at an in-use or already deleted inode. The worst that
2133 * can happen in this case is that we get a "bit already cleared" message from
2134 * ext4_free_inode(). The only reason we would point at a wrong inode is if
2135 * e2fsck was run on this filesystem, and it must have already done the orphan
2136 * inode cleanup for us, so we can safely abort without any further action.
2138 static void ext4_orphan_cleanup(struct super_block *sb,
2139 struct ext4_super_block *es)
2141 unsigned int s_flags = sb->s_flags;
2142 int nr_orphans = 0, nr_truncates = 0;
2146 if (!es->s_last_orphan) {
2147 jbd_debug(4, "no orphan inodes to clean up\n");
2151 if (bdev_read_only(sb->s_bdev)) {
2152 ext4_msg(sb, KERN_ERR, "write access "
2153 "unavailable, skipping orphan cleanup");
2157 /* Check if feature set would not allow a r/w mount */
2158 if (!ext4_feature_set_ok(sb, 0)) {
2159 ext4_msg(sb, KERN_INFO, "Skipping orphan cleanup due to "
2160 "unknown ROCOMPAT features");
2164 if (EXT4_SB(sb)->s_mount_state & EXT4_ERROR_FS) {
2165 /* don't clear list on RO mount w/ errors */
2166 if (es->s_last_orphan && !(s_flags & MS_RDONLY)) {
2167 ext4_msg(sb, KERN_INFO, "Errors on filesystem, "
2168 "clearing orphan list.\n");
2169 es->s_last_orphan = 0;
2171 jbd_debug(1, "Skipping orphan recovery on fs with errors.\n");
2175 if (s_flags & MS_RDONLY) {
2176 ext4_msg(sb, KERN_INFO, "orphan cleanup on readonly fs");
2177 sb->s_flags &= ~MS_RDONLY;
2180 /* Needed for iput() to work correctly and not trash data */
2181 sb->s_flags |= MS_ACTIVE;
2182 /* Turn on quotas so that they are updated correctly */
2183 for (i = 0; i < EXT4_MAXQUOTAS; i++) {
2184 if (EXT4_SB(sb)->s_qf_names[i]) {
2185 int ret = ext4_quota_on_mount(sb, i);
2187 ext4_msg(sb, KERN_ERR,
2188 "Cannot turn on journaled "
2189 "quota: error %d", ret);
2194 while (es->s_last_orphan) {
2195 struct inode *inode;
2197 inode = ext4_orphan_get(sb, le32_to_cpu(es->s_last_orphan));
2198 if (IS_ERR(inode)) {
2199 es->s_last_orphan = 0;
2203 list_add(&EXT4_I(inode)->i_orphan, &EXT4_SB(sb)->s_orphan);
2204 dquot_initialize(inode);
2205 if (inode->i_nlink) {
2206 if (test_opt(sb, DEBUG))
2207 ext4_msg(sb, KERN_DEBUG,
2208 "%s: truncating inode %lu to %lld bytes",
2209 __func__, inode->i_ino, inode->i_size);
2210 jbd_debug(2, "truncating inode %lu to %lld bytes\n",
2211 inode->i_ino, inode->i_size);
2212 mutex_lock(&inode->i_mutex);
2213 truncate_inode_pages(inode->i_mapping, inode->i_size);
2214 ext4_truncate(inode);
2215 mutex_unlock(&inode->i_mutex);
2218 if (test_opt(sb, DEBUG))
2219 ext4_msg(sb, KERN_DEBUG,
2220 "%s: deleting unreferenced inode %lu",
2221 __func__, inode->i_ino);
2222 jbd_debug(2, "deleting unreferenced inode %lu\n",
2226 iput(inode); /* The delete magic happens here! */
2229 #define PLURAL(x) (x), ((x) == 1) ? "" : "s"
2232 ext4_msg(sb, KERN_INFO, "%d orphan inode%s deleted",
2233 PLURAL(nr_orphans));
2235 ext4_msg(sb, KERN_INFO, "%d truncate%s cleaned up",
2236 PLURAL(nr_truncates));
2238 /* Turn quotas off */
2239 for (i = 0; i < EXT4_MAXQUOTAS; i++) {
2240 if (sb_dqopt(sb)->files[i])
2241 dquot_quota_off(sb, i);
2244 sb->s_flags = s_flags; /* Restore MS_RDONLY status */
2248 * Maximal extent format file size.
2249 * Resulting logical blkno at s_maxbytes must fit in our on-disk
2250 * extent format containers, within a sector_t, and within i_blocks
2251 * in the vfs. ext4 inode has 48 bits of i_block in fsblock units,
2252 * so that won't be a limiting factor.
2254 * However there is other limiting factor. We do store extents in the form
2255 * of starting block and length, hence the resulting length of the extent
2256 * covering maximum file size must fit into on-disk format containers as
2257 * well. Given that length is always by 1 unit bigger than max unit (because
2258 * we count 0 as well) we have to lower the s_maxbytes by one fs block.
2260 * Note, this does *not* consider any metadata overhead for vfs i_blocks.
2262 static loff_t ext4_max_size(int blkbits, int has_huge_files)
2265 loff_t upper_limit = MAX_LFS_FILESIZE;
2267 /* small i_blocks in vfs inode? */
2268 if (!has_huge_files || sizeof(blkcnt_t) < sizeof(u64)) {
2270 * CONFIG_LBDAF is not enabled implies the inode
2271 * i_block represent total blocks in 512 bytes
2272 * 32 == size of vfs inode i_blocks * 8
2274 upper_limit = (1LL << 32) - 1;
2276 /* total blocks in file system block size */
2277 upper_limit >>= (blkbits - 9);
2278 upper_limit <<= blkbits;
2282 * 32-bit extent-start container, ee_block. We lower the maxbytes
2283 * by one fs block, so ee_len can cover the extent of maximum file
2286 res = (1LL << 32) - 1;
2289 /* Sanity check against vm- & vfs- imposed limits */
2290 if (res > upper_limit)
2297 * Maximal bitmap file size. There is a direct, and {,double-,triple-}indirect
2298 * block limit, and also a limit of (2^48 - 1) 512-byte sectors in i_blocks.
2299 * We need to be 1 filesystem block less than the 2^48 sector limit.
2301 static loff_t ext4_max_bitmap_size(int bits, int has_huge_files)
2303 loff_t res = EXT4_NDIR_BLOCKS;
2306 /* This is calculated to be the largest file size for a dense, block
2307 * mapped file such that the file's total number of 512-byte sectors,
2308 * including data and all indirect blocks, does not exceed (2^48 - 1).
2310 * __u32 i_blocks_lo and _u16 i_blocks_high represent the total
2311 * number of 512-byte sectors of the file.
2314 if (!has_huge_files || sizeof(blkcnt_t) < sizeof(u64)) {
2316 * !has_huge_files or CONFIG_LBDAF not enabled implies that
2317 * the inode i_block field represents total file blocks in
2318 * 2^32 512-byte sectors == size of vfs inode i_blocks * 8
2320 upper_limit = (1LL << 32) - 1;
2322 /* total blocks in file system block size */
2323 upper_limit >>= (bits - 9);
2327 * We use 48 bit ext4_inode i_blocks
2328 * With EXT4_HUGE_FILE_FL set the i_blocks
2329 * represent total number of blocks in
2330 * file system block size
2332 upper_limit = (1LL << 48) - 1;
2336 /* indirect blocks */
2338 /* double indirect blocks */
2339 meta_blocks += 1 + (1LL << (bits-2));
2340 /* tripple indirect blocks */
2341 meta_blocks += 1 + (1LL << (bits-2)) + (1LL << (2*(bits-2)));
2343 upper_limit -= meta_blocks;
2344 upper_limit <<= bits;
2346 res += 1LL << (bits-2);
2347 res += 1LL << (2*(bits-2));
2348 res += 1LL << (3*(bits-2));
2350 if (res > upper_limit)
2353 if (res > MAX_LFS_FILESIZE)
2354 res = MAX_LFS_FILESIZE;
2359 static ext4_fsblk_t descriptor_loc(struct super_block *sb,
2360 ext4_fsblk_t logical_sb_block, int nr)
2362 struct ext4_sb_info *sbi = EXT4_SB(sb);
2363 ext4_group_t bg, first_meta_bg;
2366 first_meta_bg = le32_to_cpu(sbi->s_es->s_first_meta_bg);
2368 if (!EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_META_BG) ||
2370 return logical_sb_block + nr + 1;
2371 bg = sbi->s_desc_per_block * nr;
2372 if (ext4_bg_has_super(sb, bg))
2376 * If we have a meta_bg fs with 1k blocks, group 0's GDT is at
2377 * block 2, not 1. If s_first_data_block == 0 (bigalloc is enabled
2378 * on modern mke2fs or blksize > 1k on older mke2fs) then we must
2381 if (sb->s_blocksize == 1024 && nr == 0 &&
2382 le32_to_cpu(EXT4_SB(sb)->s_es->s_first_data_block) == 0)
2385 return (has_super + ext4_group_first_block_no(sb, bg));
2389 * ext4_get_stripe_size: Get the stripe size.
2390 * @sbi: In memory super block info
2392 * If we have specified it via mount option, then
2393 * use the mount option value. If the value specified at mount time is
2394 * greater than the blocks per group use the super block value.
2395 * If the super block value is greater than blocks per group return 0.
2396 * Allocator needs it be less than blocks per group.
2399 static unsigned long ext4_get_stripe_size(struct ext4_sb_info *sbi)
2401 unsigned long stride = le16_to_cpu(sbi->s_es->s_raid_stride);
2402 unsigned long stripe_width =
2403 le32_to_cpu(sbi->s_es->s_raid_stripe_width);
2406 if (sbi->s_stripe && sbi->s_stripe <= sbi->s_blocks_per_group)
2407 ret = sbi->s_stripe;
2408 else if (stripe_width <= sbi->s_blocks_per_group)
2410 else if (stride <= sbi->s_blocks_per_group)
2416 * If the stripe width is 1, this makes no sense and
2417 * we set it to 0 to turn off stripe handling code.
2428 struct attribute attr;
2429 ssize_t (*show)(struct ext4_attr *, struct ext4_sb_info *, char *);
2430 ssize_t (*store)(struct ext4_attr *, struct ext4_sb_info *,
2431 const char *, size_t);
2438 static int parse_strtoull(const char *buf,
2439 unsigned long long max, unsigned long long *value)
2443 ret = kstrtoull(skip_spaces(buf), 0, value);
2444 if (!ret && *value > max)
2449 static ssize_t delayed_allocation_blocks_show(struct ext4_attr *a,
2450 struct ext4_sb_info *sbi,
2453 return snprintf(buf, PAGE_SIZE, "%llu\n",
2455 percpu_counter_sum(&sbi->s_dirtyclusters_counter)));
2458 static ssize_t session_write_kbytes_show(struct ext4_attr *a,
2459 struct ext4_sb_info *sbi, char *buf)
2461 struct super_block *sb = sbi->s_buddy_cache->i_sb;
2463 if (!sb->s_bdev->bd_part)
2464 return snprintf(buf, PAGE_SIZE, "0\n");
2465 return snprintf(buf, PAGE_SIZE, "%lu\n",
2466 (part_stat_read(sb->s_bdev->bd_part, sectors[1]) -
2467 sbi->s_sectors_written_start) >> 1);
2470 static ssize_t lifetime_write_kbytes_show(struct ext4_attr *a,
2471 struct ext4_sb_info *sbi, char *buf)
2473 struct super_block *sb = sbi->s_buddy_cache->i_sb;
2475 if (!sb->s_bdev->bd_part)
2476 return snprintf(buf, PAGE_SIZE, "0\n");
2477 return snprintf(buf, PAGE_SIZE, "%llu\n",
2478 (unsigned long long)(sbi->s_kbytes_written +
2479 ((part_stat_read(sb->s_bdev->bd_part, sectors[1]) -
2480 EXT4_SB(sb)->s_sectors_written_start) >> 1)));
2483 static ssize_t inode_readahead_blks_store(struct ext4_attr *a,
2484 struct ext4_sb_info *sbi,
2485 const char *buf, size_t count)
2490 ret = kstrtoul(skip_spaces(buf), 0, &t);
2494 if (t && (!is_power_of_2(t) || t > 0x40000000))
2497 sbi->s_inode_readahead_blks = t;
2501 static ssize_t sbi_ui_show(struct ext4_attr *a,
2502 struct ext4_sb_info *sbi, char *buf)
2504 unsigned int *ui = (unsigned int *) (((char *) sbi) + a->u.offset);
2506 return snprintf(buf, PAGE_SIZE, "%u\n", *ui);
2509 static ssize_t sbi_ui_store(struct ext4_attr *a,
2510 struct ext4_sb_info *sbi,
2511 const char *buf, size_t count)
2513 unsigned int *ui = (unsigned int *) (((char *) sbi) + a->u.offset);
2517 ret = kstrtoul(skip_spaces(buf), 0, &t);
2524 static ssize_t es_ui_show(struct ext4_attr *a,
2525 struct ext4_sb_info *sbi, char *buf)
2528 unsigned int *ui = (unsigned int *) (((char *) sbi->s_es) +
2531 return snprintf(buf, PAGE_SIZE, "%u\n", *ui);
2534 static ssize_t reserved_clusters_show(struct ext4_attr *a,
2535 struct ext4_sb_info *sbi, char *buf)
2537 return snprintf(buf, PAGE_SIZE, "%llu\n",
2538 (unsigned long long) atomic64_read(&sbi->s_resv_clusters));
2541 static ssize_t reserved_clusters_store(struct ext4_attr *a,
2542 struct ext4_sb_info *sbi,
2543 const char *buf, size_t count)
2545 unsigned long long val;
2548 if (parse_strtoull(buf, -1ULL, &val))
2550 ret = ext4_reserve_clusters(sbi, val);
2552 return ret ? ret : count;
2555 static ssize_t trigger_test_error(struct ext4_attr *a,
2556 struct ext4_sb_info *sbi,
2557 const char *buf, size_t count)
2561 if (!capable(CAP_SYS_ADMIN))
2564 if (len && buf[len-1] == '\n')
2568 ext4_error(sbi->s_sb, "%.*s", len, buf);
2572 static ssize_t sbi_deprecated_show(struct ext4_attr *a,
2573 struct ext4_sb_info *sbi, char *buf)
2575 return snprintf(buf, PAGE_SIZE, "%d\n", a->u.deprecated_val);
2578 #define EXT4_ATTR_OFFSET(_name,_mode,_show,_store,_elname) \
2579 static struct ext4_attr ext4_attr_##_name = { \
2580 .attr = {.name = __stringify(_name), .mode = _mode }, \
2584 .offset = offsetof(struct ext4_sb_info, _elname),\
2588 #define EXT4_ATTR_OFFSET_ES(_name,_mode,_show,_store,_elname) \
2589 static struct ext4_attr ext4_attr_##_name = { \
2590 .attr = {.name = __stringify(_name), .mode = _mode }, \
2594 .offset = offsetof(struct ext4_super_block, _elname), \
2598 #define EXT4_ATTR(name, mode, show, store) \
2599 static struct ext4_attr ext4_attr_##name = __ATTR(name, mode, show, store)
2601 #define EXT4_INFO_ATTR(name) EXT4_ATTR(name, 0444, NULL, NULL)
2602 #define EXT4_RO_ATTR(name) EXT4_ATTR(name, 0444, name##_show, NULL)
2603 #define EXT4_RW_ATTR(name) EXT4_ATTR(name, 0644, name##_show, name##_store)
2605 #define EXT4_RO_ATTR_ES_UI(name, elname) \
2606 EXT4_ATTR_OFFSET_ES(name, 0444, es_ui_show, NULL, elname)
2607 #define EXT4_RW_ATTR_SBI_UI(name, elname) \
2608 EXT4_ATTR_OFFSET(name, 0644, sbi_ui_show, sbi_ui_store, elname)
2610 #define ATTR_LIST(name) &ext4_attr_##name.attr
2611 #define EXT4_DEPRECATED_ATTR(_name, _val) \
2612 static struct ext4_attr ext4_attr_##_name = { \
2613 .attr = {.name = __stringify(_name), .mode = 0444 }, \
2614 .show = sbi_deprecated_show, \
2616 .deprecated_val = _val, \
2620 EXT4_RO_ATTR(delayed_allocation_blocks);
2621 EXT4_RO_ATTR(session_write_kbytes);
2622 EXT4_RO_ATTR(lifetime_write_kbytes);
2623 EXT4_RW_ATTR(reserved_clusters);
2624 EXT4_ATTR_OFFSET(inode_readahead_blks, 0644, sbi_ui_show,
2625 inode_readahead_blks_store, s_inode_readahead_blks);
2626 EXT4_RW_ATTR_SBI_UI(inode_goal, s_inode_goal);
2627 EXT4_RW_ATTR_SBI_UI(mb_stats, s_mb_stats);
2628 EXT4_RW_ATTR_SBI_UI(mb_max_to_scan, s_mb_max_to_scan);
2629 EXT4_RW_ATTR_SBI_UI(mb_min_to_scan, s_mb_min_to_scan);
2630 EXT4_RW_ATTR_SBI_UI(mb_order2_req, s_mb_order2_reqs);
2631 EXT4_RW_ATTR_SBI_UI(mb_stream_req, s_mb_stream_request);
2632 EXT4_RW_ATTR_SBI_UI(mb_group_prealloc, s_mb_group_prealloc);
2633 EXT4_DEPRECATED_ATTR(max_writeback_mb_bump, 128);
2634 EXT4_RW_ATTR_SBI_UI(extent_max_zeroout_kb, s_extent_max_zeroout_kb);
2635 EXT4_ATTR(trigger_fs_error, 0200, NULL, trigger_test_error);
2636 EXT4_RW_ATTR_SBI_UI(err_ratelimit_interval_ms, s_err_ratelimit_state.interval);
2637 EXT4_RW_ATTR_SBI_UI(err_ratelimit_burst, s_err_ratelimit_state.burst);
2638 EXT4_RW_ATTR_SBI_UI(warning_ratelimit_interval_ms, s_warning_ratelimit_state.interval);
2639 EXT4_RW_ATTR_SBI_UI(warning_ratelimit_burst, s_warning_ratelimit_state.burst);
2640 EXT4_RW_ATTR_SBI_UI(msg_ratelimit_interval_ms, s_msg_ratelimit_state.interval);
2641 EXT4_RW_ATTR_SBI_UI(msg_ratelimit_burst, s_msg_ratelimit_state.burst);
2642 EXT4_RO_ATTR_ES_UI(errors_count, s_error_count);
2643 EXT4_RO_ATTR_ES_UI(first_error_time, s_first_error_time);
2644 EXT4_RO_ATTR_ES_UI(last_error_time, s_last_error_time);
2646 static struct attribute *ext4_attrs[] = {
2647 ATTR_LIST(delayed_allocation_blocks),
2648 ATTR_LIST(session_write_kbytes),
2649 ATTR_LIST(lifetime_write_kbytes),
2650 ATTR_LIST(reserved_clusters),
2651 ATTR_LIST(inode_readahead_blks),
2652 ATTR_LIST(inode_goal),
2653 ATTR_LIST(mb_stats),
2654 ATTR_LIST(mb_max_to_scan),
2655 ATTR_LIST(mb_min_to_scan),
2656 ATTR_LIST(mb_order2_req),
2657 ATTR_LIST(mb_stream_req),
2658 ATTR_LIST(mb_group_prealloc),
2659 ATTR_LIST(max_writeback_mb_bump),
2660 ATTR_LIST(extent_max_zeroout_kb),
2661 ATTR_LIST(trigger_fs_error),
2662 ATTR_LIST(err_ratelimit_interval_ms),
2663 ATTR_LIST(err_ratelimit_burst),
2664 ATTR_LIST(warning_ratelimit_interval_ms),
2665 ATTR_LIST(warning_ratelimit_burst),
2666 ATTR_LIST(msg_ratelimit_interval_ms),
2667 ATTR_LIST(msg_ratelimit_burst),
2668 ATTR_LIST(errors_count),
2669 ATTR_LIST(first_error_time),
2670 ATTR_LIST(last_error_time),
2674 /* Features this copy of ext4 supports */
2675 EXT4_INFO_ATTR(lazy_itable_init);
2676 EXT4_INFO_ATTR(batched_discard);
2677 EXT4_INFO_ATTR(meta_bg_resize);
2679 static struct attribute *ext4_feat_attrs[] = {
2680 ATTR_LIST(lazy_itable_init),
2681 ATTR_LIST(batched_discard),
2682 ATTR_LIST(meta_bg_resize),
2686 static ssize_t ext4_attr_show(struct kobject *kobj,
2687 struct attribute *attr, char *buf)
2689 struct ext4_sb_info *sbi = container_of(kobj, struct ext4_sb_info,
2691 struct ext4_attr *a = container_of(attr, struct ext4_attr, attr);
2693 return a->show ? a->show(a, sbi, buf) : 0;
2696 static ssize_t ext4_attr_store(struct kobject *kobj,
2697 struct attribute *attr,
2698 const char *buf, size_t len)
2700 struct ext4_sb_info *sbi = container_of(kobj, struct ext4_sb_info,
2702 struct ext4_attr *a = container_of(attr, struct ext4_attr, attr);
2704 return a->store ? a->store(a, sbi, buf, len) : 0;
2707 static void ext4_sb_release(struct kobject *kobj)
2709 struct ext4_sb_info *sbi = container_of(kobj, struct ext4_sb_info,
2711 complete(&sbi->s_kobj_unregister);
2714 static const struct sysfs_ops ext4_attr_ops = {
2715 .show = ext4_attr_show,
2716 .store = ext4_attr_store,
2719 static struct kobj_type ext4_ktype = {
2720 .default_attrs = ext4_attrs,
2721 .sysfs_ops = &ext4_attr_ops,
2722 .release = ext4_sb_release,
2725 static void ext4_feat_release(struct kobject *kobj)
2727 complete(&ext4_feat->f_kobj_unregister);
2730 static ssize_t ext4_feat_show(struct kobject *kobj,
2731 struct attribute *attr, char *buf)
2733 return snprintf(buf, PAGE_SIZE, "supported\n");
2737 * We can not use ext4_attr_show/store because it relies on the kobject
2738 * being embedded in the ext4_sb_info structure which is definitely not
2739 * true in this case.
2741 static const struct sysfs_ops ext4_feat_ops = {
2742 .show = ext4_feat_show,
2746 static struct kobj_type ext4_feat_ktype = {
2747 .default_attrs = ext4_feat_attrs,
2748 .sysfs_ops = &ext4_feat_ops,
2749 .release = ext4_feat_release,
2753 * Check whether this filesystem can be mounted based on
2754 * the features present and the RDONLY/RDWR mount requested.
2755 * Returns 1 if this filesystem can be mounted as requested,
2756 * 0 if it cannot be.
2758 static int ext4_feature_set_ok(struct super_block *sb, int readonly)
2760 if (EXT4_HAS_INCOMPAT_FEATURE(sb, ~EXT4_FEATURE_INCOMPAT_SUPP)) {
2761 ext4_msg(sb, KERN_ERR,
2762 "Couldn't mount because of "
2763 "unsupported optional features (%x)",
2764 (le32_to_cpu(EXT4_SB(sb)->s_es->s_feature_incompat) &
2765 ~EXT4_FEATURE_INCOMPAT_SUPP));
2772 /* Check that feature set is OK for a read-write mount */
2773 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, ~EXT4_FEATURE_RO_COMPAT_SUPP)) {
2774 ext4_msg(sb, KERN_ERR, "couldn't mount RDWR because of "
2775 "unsupported optional features (%x)",
2776 (le32_to_cpu(EXT4_SB(sb)->s_es->s_feature_ro_compat) &
2777 ~EXT4_FEATURE_RO_COMPAT_SUPP));
2781 * Large file size enabled file system can only be mounted
2782 * read-write on 32-bit systems if kernel is built with CONFIG_LBDAF
2784 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_HUGE_FILE)) {
2785 if (sizeof(blkcnt_t) < sizeof(u64)) {
2786 ext4_msg(sb, KERN_ERR, "Filesystem with huge files "
2787 "cannot be mounted RDWR without "
2792 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_BIGALLOC) &&
2793 !EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_EXTENTS)) {
2794 ext4_msg(sb, KERN_ERR,
2795 "Can't support bigalloc feature without "
2796 "extents feature\n");
2800 #ifndef CONFIG_QUOTA
2801 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_QUOTA) &&
2803 ext4_msg(sb, KERN_ERR,
2804 "Filesystem with quota feature cannot be mounted RDWR "
2805 "without CONFIG_QUOTA");
2808 #endif /* CONFIG_QUOTA */
2813 * This function is called once a day if we have errors logged
2814 * on the file system
2816 static void print_daily_error_info(unsigned long arg)
2818 struct super_block *sb = (struct super_block *) arg;
2819 struct ext4_sb_info *sbi;
2820 struct ext4_super_block *es;
2825 if (es->s_error_count)
2826 /* fsck newer than v1.41.13 is needed to clean this condition. */
2827 ext4_msg(sb, KERN_NOTICE, "error count since last fsck: %u",
2828 le32_to_cpu(es->s_error_count));
2829 if (es->s_first_error_time) {
2830 printk(KERN_NOTICE "EXT4-fs (%s): initial error at time %u: %.*s:%d",
2831 sb->s_id, le32_to_cpu(es->s_first_error_time),
2832 (int) sizeof(es->s_first_error_func),
2833 es->s_first_error_func,
2834 le32_to_cpu(es->s_first_error_line));
2835 if (es->s_first_error_ino)
2836 printk(": inode %u",
2837 le32_to_cpu(es->s_first_error_ino));
2838 if (es->s_first_error_block)
2839 printk(": block %llu", (unsigned long long)
2840 le64_to_cpu(es->s_first_error_block));
2843 if (es->s_last_error_time) {
2844 printk(KERN_NOTICE "EXT4-fs (%s): last error at time %u: %.*s:%d",
2845 sb->s_id, le32_to_cpu(es->s_last_error_time),
2846 (int) sizeof(es->s_last_error_func),
2847 es->s_last_error_func,
2848 le32_to_cpu(es->s_last_error_line));
2849 if (es->s_last_error_ino)
2850 printk(": inode %u",
2851 le32_to_cpu(es->s_last_error_ino));
2852 if (es->s_last_error_block)
2853 printk(": block %llu", (unsigned long long)
2854 le64_to_cpu(es->s_last_error_block));
2857 mod_timer(&sbi->s_err_report, jiffies + 24*60*60*HZ); /* Once a day */
2860 /* Find next suitable group and run ext4_init_inode_table */
2861 static int ext4_run_li_request(struct ext4_li_request *elr)
2863 struct ext4_group_desc *gdp = NULL;
2864 ext4_group_t group, ngroups;
2865 struct super_block *sb;
2866 unsigned long timeout = 0;
2870 ngroups = EXT4_SB(sb)->s_groups_count;
2873 for (group = elr->lr_next_group; group < ngroups; group++) {
2874 gdp = ext4_get_group_desc(sb, group, NULL);
2880 if (!(gdp->bg_flags & cpu_to_le16(EXT4_BG_INODE_ZEROED)))
2884 if (group >= ngroups)
2889 ret = ext4_init_inode_table(sb, group,
2890 elr->lr_timeout ? 0 : 1);
2891 if (elr->lr_timeout == 0) {
2892 timeout = (jiffies - timeout) *
2893 elr->lr_sbi->s_li_wait_mult;
2894 elr->lr_timeout = timeout;
2896 elr->lr_next_sched = jiffies + elr->lr_timeout;
2897 elr->lr_next_group = group + 1;
2905 * Remove lr_request from the list_request and free the
2906 * request structure. Should be called with li_list_mtx held
2908 static void ext4_remove_li_request(struct ext4_li_request *elr)
2910 struct ext4_sb_info *sbi;
2917 list_del(&elr->lr_request);
2918 sbi->s_li_request = NULL;
2922 static void ext4_unregister_li_request(struct super_block *sb)
2924 mutex_lock(&ext4_li_mtx);
2925 if (!ext4_li_info) {
2926 mutex_unlock(&ext4_li_mtx);
2930 mutex_lock(&ext4_li_info->li_list_mtx);
2931 ext4_remove_li_request(EXT4_SB(sb)->s_li_request);
2932 mutex_unlock(&ext4_li_info->li_list_mtx);
2933 mutex_unlock(&ext4_li_mtx);
2936 static struct task_struct *ext4_lazyinit_task;
2939 * This is the function where ext4lazyinit thread lives. It walks
2940 * through the request list searching for next scheduled filesystem.
2941 * When such a fs is found, run the lazy initialization request
2942 * (ext4_rn_li_request) and keep track of the time spend in this
2943 * function. Based on that time we compute next schedule time of
2944 * the request. When walking through the list is complete, compute
2945 * next waking time and put itself into sleep.
2947 static int ext4_lazyinit_thread(void *arg)
2949 struct ext4_lazy_init *eli = (struct ext4_lazy_init *)arg;
2950 struct list_head *pos, *n;
2951 struct ext4_li_request *elr;
2952 unsigned long next_wakeup, cur;
2954 BUG_ON(NULL == eli);
2958 next_wakeup = MAX_JIFFY_OFFSET;
2960 mutex_lock(&eli->li_list_mtx);
2961 if (list_empty(&eli->li_request_list)) {
2962 mutex_unlock(&eli->li_list_mtx);
2966 list_for_each_safe(pos, n, &eli->li_request_list) {
2967 elr = list_entry(pos, struct ext4_li_request,
2970 if (time_after_eq(jiffies, elr->lr_next_sched)) {
2971 if (ext4_run_li_request(elr) != 0) {
2972 /* error, remove the lazy_init job */
2973 ext4_remove_li_request(elr);
2978 if (time_before(elr->lr_next_sched, next_wakeup))
2979 next_wakeup = elr->lr_next_sched;
2981 mutex_unlock(&eli->li_list_mtx);
2986 if ((time_after_eq(cur, next_wakeup)) ||
2987 (MAX_JIFFY_OFFSET == next_wakeup)) {
2992 schedule_timeout_interruptible(next_wakeup - cur);
2994 if (kthread_should_stop()) {
2995 ext4_clear_request_list();
3002 * It looks like the request list is empty, but we need
3003 * to check it under the li_list_mtx lock, to prevent any
3004 * additions into it, and of course we should lock ext4_li_mtx
3005 * to atomically free the list and ext4_li_info, because at
3006 * this point another ext4 filesystem could be registering
3009 mutex_lock(&ext4_li_mtx);
3010 mutex_lock(&eli->li_list_mtx);
3011 if (!list_empty(&eli->li_request_list)) {
3012 mutex_unlock(&eli->li_list_mtx);
3013 mutex_unlock(&ext4_li_mtx);
3016 mutex_unlock(&eli->li_list_mtx);
3017 kfree(ext4_li_info);
3018 ext4_li_info = NULL;
3019 mutex_unlock(&ext4_li_mtx);
3024 static void ext4_clear_request_list(void)
3026 struct list_head *pos, *n;
3027 struct ext4_li_request *elr;
3029 mutex_lock(&ext4_li_info->li_list_mtx);
3030 list_for_each_safe(pos, n, &ext4_li_info->li_request_list) {
3031 elr = list_entry(pos, struct ext4_li_request,
3033 ext4_remove_li_request(elr);
3035 mutex_unlock(&ext4_li_info->li_list_mtx);
3038 static int ext4_run_lazyinit_thread(void)
3040 ext4_lazyinit_task = kthread_run(ext4_lazyinit_thread,
3041 ext4_li_info, "ext4lazyinit");
3042 if (IS_ERR(ext4_lazyinit_task)) {
3043 int err = PTR_ERR(ext4_lazyinit_task);
3044 ext4_clear_request_list();
3045 kfree(ext4_li_info);
3046 ext4_li_info = NULL;
3047 printk(KERN_CRIT "EXT4-fs: error %d creating inode table "
3048 "initialization thread\n",
3052 ext4_li_info->li_state |= EXT4_LAZYINIT_RUNNING;
3057 * Check whether it make sense to run itable init. thread or not.
3058 * If there is at least one uninitialized inode table, return
3059 * corresponding group number, else the loop goes through all
3060 * groups and return total number of groups.
3062 static ext4_group_t ext4_has_uninit_itable(struct super_block *sb)
3064 ext4_group_t group, ngroups = EXT4_SB(sb)->s_groups_count;
3065 struct ext4_group_desc *gdp = NULL;
3067 for (group = 0; group < ngroups; group++) {
3068 gdp = ext4_get_group_desc(sb, group, NULL);
3072 if (!(gdp->bg_flags & cpu_to_le16(EXT4_BG_INODE_ZEROED)))
3079 static int ext4_li_info_new(void)
3081 struct ext4_lazy_init *eli = NULL;
3083 eli = kzalloc(sizeof(*eli), GFP_KERNEL);
3087 INIT_LIST_HEAD(&eli->li_request_list);
3088 mutex_init(&eli->li_list_mtx);
3090 eli->li_state |= EXT4_LAZYINIT_QUIT;
3097 static struct ext4_li_request *ext4_li_request_new(struct super_block *sb,
3100 struct ext4_sb_info *sbi = EXT4_SB(sb);
3101 struct ext4_li_request *elr;
3103 elr = kzalloc(sizeof(*elr), GFP_KERNEL);
3109 elr->lr_next_group = start;
3112 * Randomize first schedule time of the request to
3113 * spread the inode table initialization requests
3116 elr->lr_next_sched = jiffies + (prandom_u32() %
3117 (EXT4_DEF_LI_MAX_START_DELAY * HZ));
3121 int ext4_register_li_request(struct super_block *sb,
3122 ext4_group_t first_not_zeroed)
3124 struct ext4_sb_info *sbi = EXT4_SB(sb);
3125 struct ext4_li_request *elr = NULL;
3126 ext4_group_t ngroups = EXT4_SB(sb)->s_groups_count;
3129 mutex_lock(&ext4_li_mtx);
3130 if (sbi->s_li_request != NULL) {
3132 * Reset timeout so it can be computed again, because
3133 * s_li_wait_mult might have changed.
3135 sbi->s_li_request->lr_timeout = 0;
3139 if (first_not_zeroed == ngroups ||
3140 (sb->s_flags & MS_RDONLY) ||
3141 !test_opt(sb, INIT_INODE_TABLE))
3144 elr = ext4_li_request_new(sb, first_not_zeroed);
3150 if (NULL == ext4_li_info) {
3151 ret = ext4_li_info_new();
3156 mutex_lock(&ext4_li_info->li_list_mtx);
3157 list_add(&elr->lr_request, &ext4_li_info->li_request_list);
3158 mutex_unlock(&ext4_li_info->li_list_mtx);
3160 sbi->s_li_request = elr;
3162 * set elr to NULL here since it has been inserted to
3163 * the request_list and the removal and free of it is
3164 * handled by ext4_clear_request_list from now on.
3168 if (!(ext4_li_info->li_state & EXT4_LAZYINIT_RUNNING)) {
3169 ret = ext4_run_lazyinit_thread();
3174 mutex_unlock(&ext4_li_mtx);
3181 * We do not need to lock anything since this is called on
3184 static void ext4_destroy_lazyinit_thread(void)
3187 * If thread exited earlier
3188 * there's nothing to be done.
3190 if (!ext4_li_info || !ext4_lazyinit_task)
3193 kthread_stop(ext4_lazyinit_task);
3196 static int set_journal_csum_feature_set(struct super_block *sb)
3199 int compat, incompat;
3200 struct ext4_sb_info *sbi = EXT4_SB(sb);
3202 if (ext4_has_metadata_csum(sb)) {
3203 /* journal checksum v3 */
3205 incompat = JBD2_FEATURE_INCOMPAT_CSUM_V3;
3207 /* journal checksum v1 */
3208 compat = JBD2_FEATURE_COMPAT_CHECKSUM;
3212 jbd2_journal_clear_features(sbi->s_journal,
3213 JBD2_FEATURE_COMPAT_CHECKSUM, 0,
3214 JBD2_FEATURE_INCOMPAT_CSUM_V3 |
3215 JBD2_FEATURE_INCOMPAT_CSUM_V2);
3216 if (test_opt(sb, JOURNAL_ASYNC_COMMIT)) {
3217 ret = jbd2_journal_set_features(sbi->s_journal,
3219 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT |
3221 } else if (test_opt(sb, JOURNAL_CHECKSUM)) {
3222 ret = jbd2_journal_set_features(sbi->s_journal,
3225 jbd2_journal_clear_features(sbi->s_journal, 0, 0,
3226 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT);
3228 jbd2_journal_clear_features(sbi->s_journal, 0, 0,
3229 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT);
3236 * Note: calculating the overhead so we can be compatible with
3237 * historical BSD practice is quite difficult in the face of
3238 * clusters/bigalloc. This is because multiple metadata blocks from
3239 * different block group can end up in the same allocation cluster.
3240 * Calculating the exact overhead in the face of clustered allocation
3241 * requires either O(all block bitmaps) in memory or O(number of block
3242 * groups**2) in time. We will still calculate the superblock for
3243 * older file systems --- and if we come across with a bigalloc file
3244 * system with zero in s_overhead_clusters the estimate will be close to
3245 * correct especially for very large cluster sizes --- but for newer
3246 * file systems, it's better to calculate this figure once at mkfs
3247 * time, and store it in the superblock. If the superblock value is
3248 * present (even for non-bigalloc file systems), we will use it.
3250 static int count_overhead(struct super_block *sb, ext4_group_t grp,
3253 struct ext4_sb_info *sbi = EXT4_SB(sb);
3254 struct ext4_group_desc *gdp;
3255 ext4_fsblk_t first_block, last_block, b;
3256 ext4_group_t i, ngroups = ext4_get_groups_count(sb);
3257 int s, j, count = 0;
3259 if (!EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_BIGALLOC))
3260 return (ext4_bg_has_super(sb, grp) + ext4_bg_num_gdb(sb, grp) +
3261 sbi->s_itb_per_group + 2);
3263 first_block = le32_to_cpu(sbi->s_es->s_first_data_block) +
3264 (grp * EXT4_BLOCKS_PER_GROUP(sb));
3265 last_block = first_block + EXT4_BLOCKS_PER_GROUP(sb) - 1;
3266 for (i = 0; i < ngroups; i++) {
3267 gdp = ext4_get_group_desc(sb, i, NULL);
3268 b = ext4_block_bitmap(sb, gdp);
3269 if (b >= first_block && b <= last_block) {
3270 ext4_set_bit(EXT4_B2C(sbi, b - first_block), buf);
3273 b = ext4_inode_bitmap(sb, gdp);
3274 if (b >= first_block && b <= last_block) {
3275 ext4_set_bit(EXT4_B2C(sbi, b - first_block), buf);
3278 b = ext4_inode_table(sb, gdp);
3279 if (b >= first_block && b + sbi->s_itb_per_group <= last_block)
3280 for (j = 0; j < sbi->s_itb_per_group; j++, b++) {
3281 int c = EXT4_B2C(sbi, b - first_block);
3282 ext4_set_bit(c, buf);
3288 if (ext4_bg_has_super(sb, grp)) {
3289 ext4_set_bit(s++, buf);
3292 for (j = ext4_bg_num_gdb(sb, grp); j > 0; j--) {
3293 ext4_set_bit(EXT4_B2C(sbi, s++), buf);
3299 return EXT4_CLUSTERS_PER_GROUP(sb) -
3300 ext4_count_free(buf, EXT4_CLUSTERS_PER_GROUP(sb) / 8);
3304 * Compute the overhead and stash it in sbi->s_overhead
3306 int ext4_calculate_overhead(struct super_block *sb)
3308 struct ext4_sb_info *sbi = EXT4_SB(sb);
3309 struct ext4_super_block *es = sbi->s_es;
3310 ext4_group_t i, ngroups = ext4_get_groups_count(sb);
3311 ext4_fsblk_t overhead = 0;
3312 char *buf = (char *) get_zeroed_page(GFP_NOFS);
3318 * Compute the overhead (FS structures). This is constant
3319 * for a given filesystem unless the number of block groups
3320 * changes so we cache the previous value until it does.
3324 * All of the blocks before first_data_block are overhead
3326 overhead = EXT4_B2C(sbi, le32_to_cpu(es->s_first_data_block));
3329 * Add the overhead found in each block group
3331 for (i = 0; i < ngroups; i++) {
3334 blks = count_overhead(sb, i, buf);
3337 memset(buf, 0, PAGE_SIZE);
3340 /* Add the internal journal blocks as well */
3341 if (sbi->s_journal && !sbi->journal_bdev)
3342 overhead += EXT4_NUM_B2C(sbi, sbi->s_journal->j_maxlen);
3344 sbi->s_overhead = overhead;
3346 free_page((unsigned long) buf);
3351 static ext4_fsblk_t ext4_calculate_resv_clusters(struct super_block *sb)
3353 ext4_fsblk_t resv_clusters;
3356 * There's no need to reserve anything when we aren't using extents.
3357 * The space estimates are exact, there are no unwritten extents,
3358 * hole punching doesn't need new metadata... This is needed especially
3359 * to keep ext2/3 backward compatibility.
3361 if (!EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_EXTENTS))
3364 * By default we reserve 2% or 4096 clusters, whichever is smaller.
3365 * This should cover the situations where we can not afford to run
3366 * out of space like for example punch hole, or converting
3367 * unwritten extents in delalloc path. In most cases such
3368 * allocation would require 1, or 2 blocks, higher numbers are
3371 resv_clusters = ext4_blocks_count(EXT4_SB(sb)->s_es) >>
3372 EXT4_SB(sb)->s_cluster_bits;
3374 do_div(resv_clusters, 50);
3375 resv_clusters = min_t(ext4_fsblk_t, resv_clusters, 4096);
3377 return resv_clusters;
3381 static int ext4_reserve_clusters(struct ext4_sb_info *sbi, ext4_fsblk_t count)
3383 ext4_fsblk_t clusters = ext4_blocks_count(sbi->s_es) >>
3384 sbi->s_cluster_bits;
3386 if (count >= clusters)
3389 atomic64_set(&sbi->s_resv_clusters, count);
3393 static int ext4_fill_super(struct super_block *sb, void *data, int silent)
3395 char *orig_data = kstrdup(data, GFP_KERNEL);
3396 struct buffer_head *bh;
3397 struct ext4_super_block *es = NULL;
3398 struct ext4_sb_info *sbi;
3400 ext4_fsblk_t sb_block = get_sb_block(&data);
3401 ext4_fsblk_t logical_sb_block;
3402 unsigned long offset = 0;
3403 unsigned long journal_devnum = 0;
3404 unsigned long def_mount_opts;
3409 int blocksize, clustersize;
3410 unsigned int db_count;
3412 int needs_recovery, has_huge_files, has_bigalloc;
3415 unsigned int journal_ioprio = DEFAULT_JOURNAL_IOPRIO;
3416 ext4_group_t first_not_zeroed;
3418 sbi = kzalloc(sizeof(*sbi), GFP_KERNEL);
3422 sbi->s_blockgroup_lock =
3423 kzalloc(sizeof(struct blockgroup_lock), GFP_KERNEL);
3424 if (!sbi->s_blockgroup_lock) {
3428 sb->s_fs_info = sbi;
3430 sbi->s_inode_readahead_blks = EXT4_DEF_INODE_READAHEAD_BLKS;
3431 sbi->s_sb_block = sb_block;
3432 if (sb->s_bdev->bd_part)
3433 sbi->s_sectors_written_start =
3434 part_stat_read(sb->s_bdev->bd_part, sectors[1]);
3436 /* Cleanup superblock name */
3437 for (cp = sb->s_id; (cp = strchr(cp, '/'));)
3440 /* -EINVAL is default */
3442 blocksize = sb_min_blocksize(sb, EXT4_MIN_BLOCK_SIZE);
3444 ext4_msg(sb, KERN_ERR, "unable to set blocksize");
3449 * The ext4 superblock will not be buffer aligned for other than 1kB
3450 * block sizes. We need to calculate the offset from buffer start.
3452 if (blocksize != EXT4_MIN_BLOCK_SIZE) {
3453 logical_sb_block = sb_block * EXT4_MIN_BLOCK_SIZE;
3454 offset = do_div(logical_sb_block, blocksize);
3456 logical_sb_block = sb_block;
3459 if (!(bh = sb_bread_unmovable(sb, logical_sb_block))) {
3460 ext4_msg(sb, KERN_ERR, "unable to read superblock");
3464 * Note: s_es must be initialized as soon as possible because
3465 * some ext4 macro-instructions depend on its value
3467 es = (struct ext4_super_block *) (bh->b_data + offset);
3469 sb->s_magic = le16_to_cpu(es->s_magic);
3470 if (sb->s_magic != EXT4_SUPER_MAGIC)
3472 sbi->s_kbytes_written = le64_to_cpu(es->s_kbytes_written);
3474 /* Warn if metadata_csum and gdt_csum are both set. */
3475 if (EXT4_HAS_RO_COMPAT_FEATURE(sb,
3476 EXT4_FEATURE_RO_COMPAT_METADATA_CSUM) &&
3477 EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_GDT_CSUM))
3478 ext4_warning(sb, "metadata_csum and uninit_bg are "
3479 "redundant flags; please run fsck.");
3481 /* Check for a known checksum algorithm */
3482 if (!ext4_verify_csum_type(sb, es)) {
3483 ext4_msg(sb, KERN_ERR, "VFS: Found ext4 filesystem with "
3484 "unknown checksum algorithm.");
3489 /* Load the checksum driver */
3490 if (EXT4_HAS_RO_COMPAT_FEATURE(sb,
3491 EXT4_FEATURE_RO_COMPAT_METADATA_CSUM)) {
3492 sbi->s_chksum_driver = crypto_alloc_shash("crc32c", 0, 0);
3493 if (IS_ERR(sbi->s_chksum_driver)) {
3494 ext4_msg(sb, KERN_ERR, "Cannot load crc32c driver.");
3495 ret = PTR_ERR(sbi->s_chksum_driver);
3496 sbi->s_chksum_driver = NULL;
3501 /* Check superblock checksum */
3502 if (!ext4_superblock_csum_verify(sb, es)) {
3503 ext4_msg(sb, KERN_ERR, "VFS: Found ext4 filesystem with "
3504 "invalid superblock checksum. Run e2fsck?");
3509 /* Precompute checksum seed for all metadata */
3510 if (ext4_has_metadata_csum(sb))
3511 sbi->s_csum_seed = ext4_chksum(sbi, ~0, es->s_uuid,
3512 sizeof(es->s_uuid));
3514 /* Set defaults before we parse the mount options */
3515 def_mount_opts = le32_to_cpu(es->s_default_mount_opts);
3516 set_opt(sb, INIT_INODE_TABLE);
3517 if (def_mount_opts & EXT4_DEFM_DEBUG)
3519 if (def_mount_opts & EXT4_DEFM_BSDGROUPS)
3521 if (def_mount_opts & EXT4_DEFM_UID16)
3522 set_opt(sb, NO_UID32);
3523 /* xattr user namespace & acls are now defaulted on */
3524 set_opt(sb, XATTR_USER);
3525 #ifdef CONFIG_EXT4_FS_POSIX_ACL
3526 set_opt(sb, POSIX_ACL);
3528 /* don't forget to enable journal_csum when metadata_csum is enabled. */
3529 if (ext4_has_metadata_csum(sb))
3530 set_opt(sb, JOURNAL_CHECKSUM);
3532 if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_DATA)
3533 set_opt(sb, JOURNAL_DATA);
3534 else if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_ORDERED)
3535 set_opt(sb, ORDERED_DATA);
3536 else if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_WBACK)
3537 set_opt(sb, WRITEBACK_DATA);
3539 if (le16_to_cpu(sbi->s_es->s_errors) == EXT4_ERRORS_PANIC)
3540 set_opt(sb, ERRORS_PANIC);
3541 else if (le16_to_cpu(sbi->s_es->s_errors) == EXT4_ERRORS_CONTINUE)
3542 set_opt(sb, ERRORS_CONT);
3544 set_opt(sb, ERRORS_RO);
3545 /* block_validity enabled by default; disable with noblock_validity */
3546 set_opt(sb, BLOCK_VALIDITY);
3547 if (def_mount_opts & EXT4_DEFM_DISCARD)
3548 set_opt(sb, DISCARD);
3550 sbi->s_resuid = make_kuid(&init_user_ns, le16_to_cpu(es->s_def_resuid));
3551 sbi->s_resgid = make_kgid(&init_user_ns, le16_to_cpu(es->s_def_resgid));
3552 sbi->s_commit_interval = JBD2_DEFAULT_MAX_COMMIT_AGE * HZ;
3553 sbi->s_min_batch_time = EXT4_DEF_MIN_BATCH_TIME;
3554 sbi->s_max_batch_time = EXT4_DEF_MAX_BATCH_TIME;
3556 if ((def_mount_opts & EXT4_DEFM_NOBARRIER) == 0)
3557 set_opt(sb, BARRIER);
3560 * enable delayed allocation by default
3561 * Use -o nodelalloc to turn it off
3563 if (!IS_EXT3_SB(sb) && !IS_EXT2_SB(sb) &&
3564 ((def_mount_opts & EXT4_DEFM_NODELALLOC) == 0))
3565 set_opt(sb, DELALLOC);
3568 * set default s_li_wait_mult for lazyinit, for the case there is
3569 * no mount option specified.
3571 sbi->s_li_wait_mult = EXT4_DEF_LI_WAIT_MULT;
3573 if (!parse_options((char *) sbi->s_es->s_mount_opts, sb,
3574 &journal_devnum, &journal_ioprio, 0)) {
3575 ext4_msg(sb, KERN_WARNING,
3576 "failed to parse options in superblock: %s",
3577 sbi->s_es->s_mount_opts);
3579 sbi->s_def_mount_opt = sbi->s_mount_opt;
3580 if (!parse_options((char *) data, sb, &journal_devnum,
3581 &journal_ioprio, 0))
3584 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA) {
3585 printk_once(KERN_WARNING "EXT4-fs: Warning: mounting "
3586 "with data=journal disables delayed "
3587 "allocation and O_DIRECT support!\n");
3588 if (test_opt2(sb, EXPLICIT_DELALLOC)) {
3589 ext4_msg(sb, KERN_ERR, "can't mount with "
3590 "both data=journal and delalloc");
3593 if (test_opt(sb, DIOREAD_NOLOCK)) {
3594 ext4_msg(sb, KERN_ERR, "can't mount with "
3595 "both data=journal and dioread_nolock");
3598 if (test_opt(sb, DELALLOC))
3599 clear_opt(sb, DELALLOC);
3602 sb->s_flags = (sb->s_flags & ~MS_POSIXACL) |
3603 (test_opt(sb, POSIX_ACL) ? MS_POSIXACL : 0);
3605 if (le32_to_cpu(es->s_rev_level) == EXT4_GOOD_OLD_REV &&
3606 (EXT4_HAS_COMPAT_FEATURE(sb, ~0U) ||
3607 EXT4_HAS_RO_COMPAT_FEATURE(sb, ~0U) ||
3608 EXT4_HAS_INCOMPAT_FEATURE(sb, ~0U)))
3609 ext4_msg(sb, KERN_WARNING,
3610 "feature flags set on rev 0 fs, "
3611 "running e2fsck is recommended");
3613 if (es->s_creator_os == cpu_to_le32(EXT4_OS_HURD)) {
3614 set_opt2(sb, HURD_COMPAT);
3615 if (EXT4_HAS_INCOMPAT_FEATURE(sb,
3616 EXT4_FEATURE_INCOMPAT_64BIT)) {
3617 ext4_msg(sb, KERN_ERR,
3618 "The Hurd can't support 64-bit file systems");
3623 if (IS_EXT2_SB(sb)) {
3624 if (ext2_feature_set_ok(sb))
3625 ext4_msg(sb, KERN_INFO, "mounting ext2 file system "
3626 "using the ext4 subsystem");
3628 ext4_msg(sb, KERN_ERR, "couldn't mount as ext2 due "
3629 "to feature incompatibilities");
3634 if (IS_EXT3_SB(sb)) {
3635 if (ext3_feature_set_ok(sb))
3636 ext4_msg(sb, KERN_INFO, "mounting ext3 file system "
3637 "using the ext4 subsystem");
3639 ext4_msg(sb, KERN_ERR, "couldn't mount as ext3 due "
3640 "to feature incompatibilities");
3646 * Check feature flags regardless of the revision level, since we
3647 * previously didn't change the revision level when setting the flags,
3648 * so there is a chance incompat flags are set on a rev 0 filesystem.
3650 if (!ext4_feature_set_ok(sb, (sb->s_flags & MS_RDONLY)))
3653 blocksize = BLOCK_SIZE << le32_to_cpu(es->s_log_block_size);
3654 if (blocksize < EXT4_MIN_BLOCK_SIZE ||
3655 blocksize > EXT4_MAX_BLOCK_SIZE) {
3656 ext4_msg(sb, KERN_ERR,
3657 "Unsupported filesystem blocksize %d", blocksize);
3661 if (sb->s_blocksize != blocksize) {
3662 /* Validate the filesystem blocksize */
3663 if (!sb_set_blocksize(sb, blocksize)) {
3664 ext4_msg(sb, KERN_ERR, "bad block size %d",
3670 logical_sb_block = sb_block * EXT4_MIN_BLOCK_SIZE;
3671 offset = do_div(logical_sb_block, blocksize);
3672 bh = sb_bread_unmovable(sb, logical_sb_block);
3674 ext4_msg(sb, KERN_ERR,
3675 "Can't read superblock on 2nd try");
3678 es = (struct ext4_super_block *)(bh->b_data + offset);
3680 if (es->s_magic != cpu_to_le16(EXT4_SUPER_MAGIC)) {
3681 ext4_msg(sb, KERN_ERR,
3682 "Magic mismatch, very weird!");
3687 has_huge_files = EXT4_HAS_RO_COMPAT_FEATURE(sb,
3688 EXT4_FEATURE_RO_COMPAT_HUGE_FILE);
3689 sbi->s_bitmap_maxbytes = ext4_max_bitmap_size(sb->s_blocksize_bits,
3691 sb->s_maxbytes = ext4_max_size(sb->s_blocksize_bits, has_huge_files);
3693 if (le32_to_cpu(es->s_rev_level) == EXT4_GOOD_OLD_REV) {
3694 sbi->s_inode_size = EXT4_GOOD_OLD_INODE_SIZE;
3695 sbi->s_first_ino = EXT4_GOOD_OLD_FIRST_INO;
3697 sbi->s_inode_size = le16_to_cpu(es->s_inode_size);
3698 sbi->s_first_ino = le32_to_cpu(es->s_first_ino);
3699 if ((sbi->s_inode_size < EXT4_GOOD_OLD_INODE_SIZE) ||
3700 (!is_power_of_2(sbi->s_inode_size)) ||
3701 (sbi->s_inode_size > blocksize)) {
3702 ext4_msg(sb, KERN_ERR,
3703 "unsupported inode size: %d",
3707 if (sbi->s_inode_size > EXT4_GOOD_OLD_INODE_SIZE)
3708 sb->s_time_gran = 1 << (EXT4_EPOCH_BITS - 2);
3711 sbi->s_desc_size = le16_to_cpu(es->s_desc_size);
3712 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_64BIT)) {
3713 if (sbi->s_desc_size < EXT4_MIN_DESC_SIZE_64BIT ||
3714 sbi->s_desc_size > EXT4_MAX_DESC_SIZE ||
3715 !is_power_of_2(sbi->s_desc_size)) {
3716 ext4_msg(sb, KERN_ERR,
3717 "unsupported descriptor size %lu",
3722 sbi->s_desc_size = EXT4_MIN_DESC_SIZE;
3724 sbi->s_blocks_per_group = le32_to_cpu(es->s_blocks_per_group);
3725 sbi->s_inodes_per_group = le32_to_cpu(es->s_inodes_per_group);
3726 if (EXT4_INODE_SIZE(sb) == 0 || EXT4_INODES_PER_GROUP(sb) == 0)
3729 sbi->s_inodes_per_block = blocksize / EXT4_INODE_SIZE(sb);
3730 if (sbi->s_inodes_per_block == 0)
3732 sbi->s_itb_per_group = sbi->s_inodes_per_group /
3733 sbi->s_inodes_per_block;
3734 sbi->s_desc_per_block = blocksize / EXT4_DESC_SIZE(sb);
3736 sbi->s_mount_state = le16_to_cpu(es->s_state);
3737 sbi->s_addr_per_block_bits = ilog2(EXT4_ADDR_PER_BLOCK(sb));
3738 sbi->s_desc_per_block_bits = ilog2(EXT4_DESC_PER_BLOCK(sb));
3740 for (i = 0; i < 4; i++)
3741 sbi->s_hash_seed[i] = le32_to_cpu(es->s_hash_seed[i]);
3742 sbi->s_def_hash_version = es->s_def_hash_version;
3743 if (EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_DIR_INDEX)) {
3744 i = le32_to_cpu(es->s_flags);
3745 if (i & EXT2_FLAGS_UNSIGNED_HASH)
3746 sbi->s_hash_unsigned = 3;
3747 else if ((i & EXT2_FLAGS_SIGNED_HASH) == 0) {
3748 #ifdef __CHAR_UNSIGNED__
3749 if (!(sb->s_flags & MS_RDONLY))
3751 cpu_to_le32(EXT2_FLAGS_UNSIGNED_HASH);
3752 sbi->s_hash_unsigned = 3;
3754 if (!(sb->s_flags & MS_RDONLY))
3756 cpu_to_le32(EXT2_FLAGS_SIGNED_HASH);
3761 /* Handle clustersize */
3762 clustersize = BLOCK_SIZE << le32_to_cpu(es->s_log_cluster_size);
3763 has_bigalloc = EXT4_HAS_RO_COMPAT_FEATURE(sb,
3764 EXT4_FEATURE_RO_COMPAT_BIGALLOC);
3766 if (clustersize < blocksize) {
3767 ext4_msg(sb, KERN_ERR,
3768 "cluster size (%d) smaller than "
3769 "block size (%d)", clustersize, blocksize);
3772 sbi->s_cluster_bits = le32_to_cpu(es->s_log_cluster_size) -
3773 le32_to_cpu(es->s_log_block_size);
3774 sbi->s_clusters_per_group =
3775 le32_to_cpu(es->s_clusters_per_group);
3776 if (sbi->s_clusters_per_group > blocksize * 8) {
3777 ext4_msg(sb, KERN_ERR,
3778 "#clusters per group too big: %lu",
3779 sbi->s_clusters_per_group);
3782 if (sbi->s_blocks_per_group !=
3783 (sbi->s_clusters_per_group * (clustersize / blocksize))) {
3784 ext4_msg(sb, KERN_ERR, "blocks per group (%lu) and "
3785 "clusters per group (%lu) inconsistent",
3786 sbi->s_blocks_per_group,
3787 sbi->s_clusters_per_group);
3791 if (clustersize != blocksize) {
3792 ext4_warning(sb, "fragment/cluster size (%d) != "
3793 "block size (%d)", clustersize,
3795 clustersize = blocksize;
3797 if (sbi->s_blocks_per_group > blocksize * 8) {
3798 ext4_msg(sb, KERN_ERR,
3799 "#blocks per group too big: %lu",
3800 sbi->s_blocks_per_group);
3803 sbi->s_clusters_per_group = sbi->s_blocks_per_group;
3804 sbi->s_cluster_bits = 0;
3806 sbi->s_cluster_ratio = clustersize / blocksize;
3808 if (sbi->s_inodes_per_group > blocksize * 8) {
3809 ext4_msg(sb, KERN_ERR,
3810 "#inodes per group too big: %lu",
3811 sbi->s_inodes_per_group);
3815 /* Do we have standard group size of clustersize * 8 blocks ? */
3816 if (sbi->s_blocks_per_group == clustersize << 3)
3817 set_opt2(sb, STD_GROUP_SIZE);
3820 * Test whether we have more sectors than will fit in sector_t,
3821 * and whether the max offset is addressable by the page cache.
3823 err = generic_check_addressable(sb->s_blocksize_bits,
3824 ext4_blocks_count(es));
3826 ext4_msg(sb, KERN_ERR, "filesystem"
3827 " too large to mount safely on this system");
3828 if (sizeof(sector_t) < 8)
3829 ext4_msg(sb, KERN_WARNING, "CONFIG_LBDAF not enabled");
3833 if (EXT4_BLOCKS_PER_GROUP(sb) == 0)
3836 /* check blocks count against device size */
3837 blocks_count = sb->s_bdev->bd_inode->i_size >> sb->s_blocksize_bits;
3838 if (blocks_count && ext4_blocks_count(es) > blocks_count) {
3839 ext4_msg(sb, KERN_WARNING, "bad geometry: block count %llu "
3840 "exceeds size of device (%llu blocks)",
3841 ext4_blocks_count(es), blocks_count);
3846 * It makes no sense for the first data block to be beyond the end
3847 * of the filesystem.
3849 if (le32_to_cpu(es->s_first_data_block) >= ext4_blocks_count(es)) {
3850 ext4_msg(sb, KERN_WARNING, "bad geometry: first data "
3851 "block %u is beyond end of filesystem (%llu)",
3852 le32_to_cpu(es->s_first_data_block),
3853 ext4_blocks_count(es));
3856 blocks_count = (ext4_blocks_count(es) -
3857 le32_to_cpu(es->s_first_data_block) +
3858 EXT4_BLOCKS_PER_GROUP(sb) - 1);
3859 do_div(blocks_count, EXT4_BLOCKS_PER_GROUP(sb));
3860 if (blocks_count > ((uint64_t)1<<32) - EXT4_DESC_PER_BLOCK(sb)) {
3861 ext4_msg(sb, KERN_WARNING, "groups count too large: %u "
3862 "(block count %llu, first data block %u, "
3863 "blocks per group %lu)", sbi->s_groups_count,
3864 ext4_blocks_count(es),
3865 le32_to_cpu(es->s_first_data_block),
3866 EXT4_BLOCKS_PER_GROUP(sb));
3869 sbi->s_groups_count = blocks_count;
3870 sbi->s_blockfile_groups = min_t(ext4_group_t, sbi->s_groups_count,
3871 (EXT4_MAX_BLOCK_FILE_PHYS / EXT4_BLOCKS_PER_GROUP(sb)));
3872 db_count = (sbi->s_groups_count + EXT4_DESC_PER_BLOCK(sb) - 1) /
3873 EXT4_DESC_PER_BLOCK(sb);
3874 sbi->s_group_desc = ext4_kvmalloc(db_count *
3875 sizeof(struct buffer_head *),
3877 if (sbi->s_group_desc == NULL) {
3878 ext4_msg(sb, KERN_ERR, "not enough memory");
3884 sbi->s_proc = proc_mkdir(sb->s_id, ext4_proc_root);
3887 proc_create_data("options", S_IRUGO, sbi->s_proc,
3888 &ext4_seq_options_fops, sb);
3890 bgl_lock_init(sbi->s_blockgroup_lock);
3892 for (i = 0; i < db_count; i++) {
3893 block = descriptor_loc(sb, logical_sb_block, i);
3894 sbi->s_group_desc[i] = sb_bread_unmovable(sb, block);
3895 if (!sbi->s_group_desc[i]) {
3896 ext4_msg(sb, KERN_ERR,
3897 "can't read group descriptor %d", i);
3902 if (!ext4_check_descriptors(sb, &first_not_zeroed)) {
3903 ext4_msg(sb, KERN_ERR, "group descriptors corrupted!");
3907 sbi->s_gdb_count = db_count;
3908 get_random_bytes(&sbi->s_next_generation, sizeof(u32));
3909 spin_lock_init(&sbi->s_next_gen_lock);
3911 init_timer(&sbi->s_err_report);
3912 sbi->s_err_report.function = print_daily_error_info;
3913 sbi->s_err_report.data = (unsigned long) sb;
3915 /* Register extent status tree shrinker */
3916 if (ext4_es_register_shrinker(sbi))
3919 sbi->s_stripe = ext4_get_stripe_size(sbi);
3920 sbi->s_extent_max_zeroout_kb = 32;
3923 * set up enough so that it can read an inode
3925 sb->s_op = &ext4_sops;
3926 sb->s_export_op = &ext4_export_ops;
3927 sb->s_xattr = ext4_xattr_handlers;
3929 sb->dq_op = &ext4_quota_operations;
3930 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_QUOTA))
3931 sb->s_qcop = &ext4_qctl_sysfile_operations;
3933 sb->s_qcop = &ext4_qctl_operations;
3935 memcpy(sb->s_uuid, es->s_uuid, sizeof(es->s_uuid));
3937 INIT_LIST_HEAD(&sbi->s_orphan); /* unlinked but open files */
3938 mutex_init(&sbi->s_orphan_lock);
3942 needs_recovery = (es->s_last_orphan != 0 ||
3943 EXT4_HAS_INCOMPAT_FEATURE(sb,
3944 EXT4_FEATURE_INCOMPAT_RECOVER));
3946 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_MMP) &&
3947 !(sb->s_flags & MS_RDONLY))
3948 if (ext4_multi_mount_protect(sb, le64_to_cpu(es->s_mmp_block)))
3949 goto failed_mount3a;
3952 * The first inode we look at is the journal inode. Don't try
3953 * root first: it may be modified in the journal!
3955 if (!test_opt(sb, NOLOAD) &&
3956 EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL)) {
3957 if (ext4_load_journal(sb, es, journal_devnum))
3958 goto failed_mount3a;
3959 } else if (test_opt(sb, NOLOAD) && !(sb->s_flags & MS_RDONLY) &&
3960 EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER)) {
3961 ext4_msg(sb, KERN_ERR, "required journal recovery "
3962 "suppressed and not mounted read-only");
3963 goto failed_mount_wq;
3965 clear_opt(sb, DATA_FLAGS);
3966 sbi->s_journal = NULL;
3971 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_64BIT) &&
3972 !jbd2_journal_set_features(EXT4_SB(sb)->s_journal, 0, 0,
3973 JBD2_FEATURE_INCOMPAT_64BIT)) {
3974 ext4_msg(sb, KERN_ERR, "Failed to set 64-bit journal feature");
3975 goto failed_mount_wq;
3978 if (!set_journal_csum_feature_set(sb)) {
3979 ext4_msg(sb, KERN_ERR, "Failed to set journal checksum "
3981 goto failed_mount_wq;
3984 /* We have now updated the journal if required, so we can
3985 * validate the data journaling mode. */
3986 switch (test_opt(sb, DATA_FLAGS)) {
3988 /* No mode set, assume a default based on the journal
3989 * capabilities: ORDERED_DATA if the journal can
3990 * cope, else JOURNAL_DATA
3992 if (jbd2_journal_check_available_features
3993 (sbi->s_journal, 0, 0, JBD2_FEATURE_INCOMPAT_REVOKE))
3994 set_opt(sb, ORDERED_DATA);
3996 set_opt(sb, JOURNAL_DATA);
3999 case EXT4_MOUNT_ORDERED_DATA:
4000 case EXT4_MOUNT_WRITEBACK_DATA:
4001 if (!jbd2_journal_check_available_features
4002 (sbi->s_journal, 0, 0, JBD2_FEATURE_INCOMPAT_REVOKE)) {
4003 ext4_msg(sb, KERN_ERR, "Journal does not support "
4004 "requested data journaling mode");
4005 goto failed_mount_wq;
4010 set_task_ioprio(sbi->s_journal->j_task, journal_ioprio);
4012 sbi->s_journal->j_commit_callback = ext4_journal_commit_callback;
4015 if (ext4_mballoc_ready) {
4016 sbi->s_mb_cache = ext4_xattr_create_cache(sb->s_id);
4017 if (!sbi->s_mb_cache) {
4018 ext4_msg(sb, KERN_ERR, "Failed to create an mb_cache");
4019 goto failed_mount_wq;
4024 * Get the # of file system overhead blocks from the
4025 * superblock if present.
4027 if (es->s_overhead_clusters)
4028 sbi->s_overhead = le32_to_cpu(es->s_overhead_clusters);
4030 err = ext4_calculate_overhead(sb);
4032 goto failed_mount_wq;
4036 * The maximum number of concurrent works can be high and
4037 * concurrency isn't really necessary. Limit it to 1.
4039 EXT4_SB(sb)->rsv_conversion_wq =
4040 alloc_workqueue("ext4-rsv-conversion", WQ_MEM_RECLAIM | WQ_UNBOUND, 1);
4041 if (!EXT4_SB(sb)->rsv_conversion_wq) {
4042 printk(KERN_ERR "EXT4-fs: failed to create workqueue\n");
4048 * The jbd2_journal_load will have done any necessary log recovery,
4049 * so we can safely mount the rest of the filesystem now.
4052 root = ext4_iget(sb, EXT4_ROOT_INO);
4054 ext4_msg(sb, KERN_ERR, "get root inode failed");
4055 ret = PTR_ERR(root);
4059 if (!S_ISDIR(root->i_mode) || !root->i_blocks || !root->i_size) {
4060 ext4_msg(sb, KERN_ERR, "corrupt root inode, run e2fsck");
4064 sb->s_root = d_make_root(root);
4066 ext4_msg(sb, KERN_ERR, "get root dentry failed");
4071 if (ext4_setup_super(sb, es, sb->s_flags & MS_RDONLY))
4072 sb->s_flags |= MS_RDONLY;
4074 /* determine the minimum size of new large inodes, if present */
4075 if (sbi->s_inode_size > EXT4_GOOD_OLD_INODE_SIZE) {
4076 sbi->s_want_extra_isize = sizeof(struct ext4_inode) -
4077 EXT4_GOOD_OLD_INODE_SIZE;
4078 if (EXT4_HAS_RO_COMPAT_FEATURE(sb,
4079 EXT4_FEATURE_RO_COMPAT_EXTRA_ISIZE)) {
4080 if (sbi->s_want_extra_isize <
4081 le16_to_cpu(es->s_want_extra_isize))
4082 sbi->s_want_extra_isize =
4083 le16_to_cpu(es->s_want_extra_isize);
4084 if (sbi->s_want_extra_isize <
4085 le16_to_cpu(es->s_min_extra_isize))
4086 sbi->s_want_extra_isize =
4087 le16_to_cpu(es->s_min_extra_isize);
4090 /* Check if enough inode space is available */
4091 if (EXT4_GOOD_OLD_INODE_SIZE + sbi->s_want_extra_isize >
4092 sbi->s_inode_size) {
4093 sbi->s_want_extra_isize = sizeof(struct ext4_inode) -
4094 EXT4_GOOD_OLD_INODE_SIZE;
4095 ext4_msg(sb, KERN_INFO, "required extra inode space not"
4099 err = ext4_reserve_clusters(sbi, ext4_calculate_resv_clusters(sb));
4101 ext4_msg(sb, KERN_ERR, "failed to reserve %llu clusters for "
4102 "reserved pool", ext4_calculate_resv_clusters(sb));
4103 goto failed_mount4a;
4106 err = ext4_setup_system_zone(sb);
4108 ext4_msg(sb, KERN_ERR, "failed to initialize system "
4110 goto failed_mount4a;
4114 err = ext4_mb_init(sb);
4116 ext4_msg(sb, KERN_ERR, "failed to initialize mballoc (%d)",
4121 block = ext4_count_free_clusters(sb);
4122 ext4_free_blocks_count_set(sbi->s_es,
4123 EXT4_C2B(sbi, block));
4124 err = percpu_counter_init(&sbi->s_freeclusters_counter, block,
4127 unsigned long freei = ext4_count_free_inodes(sb);
4128 sbi->s_es->s_free_inodes_count = cpu_to_le32(freei);
4129 err = percpu_counter_init(&sbi->s_freeinodes_counter, freei,
4133 err = percpu_counter_init(&sbi->s_dirs_counter,
4134 ext4_count_dirs(sb), GFP_KERNEL);
4136 err = percpu_counter_init(&sbi->s_dirtyclusters_counter, 0,
4139 ext4_msg(sb, KERN_ERR, "insufficient memory");
4143 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_FLEX_BG))
4144 if (!ext4_fill_flex_info(sb)) {
4145 ext4_msg(sb, KERN_ERR,
4146 "unable to initialize "
4147 "flex_bg meta info!");
4151 err = ext4_register_li_request(sb, first_not_zeroed);
4155 sbi->s_kobj.kset = ext4_kset;
4156 init_completion(&sbi->s_kobj_unregister);
4157 err = kobject_init_and_add(&sbi->s_kobj, &ext4_ktype, NULL,
4163 /* Enable quota usage during mount. */
4164 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_QUOTA) &&
4165 !(sb->s_flags & MS_RDONLY)) {
4166 err = ext4_enable_quotas(sb);
4170 #endif /* CONFIG_QUOTA */
4172 EXT4_SB(sb)->s_mount_state |= EXT4_ORPHAN_FS;
4173 ext4_orphan_cleanup(sb, es);
4174 EXT4_SB(sb)->s_mount_state &= ~EXT4_ORPHAN_FS;
4175 if (needs_recovery) {
4176 ext4_msg(sb, KERN_INFO, "recovery complete");
4177 ext4_mark_recovery_complete(sb, es);
4179 if (EXT4_SB(sb)->s_journal) {
4180 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA)
4181 descr = " journalled data mode";
4182 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA)
4183 descr = " ordered data mode";
4185 descr = " writeback data mode";
4187 descr = "out journal";
4189 if (test_opt(sb, DISCARD)) {
4190 struct request_queue *q = bdev_get_queue(sb->s_bdev);
4191 if (!blk_queue_discard(q))
4192 ext4_msg(sb, KERN_WARNING,
4193 "mounting with \"discard\" option, but "
4194 "the device does not support discard");
4197 ext4_msg(sb, KERN_INFO, "mounted filesystem with%s. "
4198 "Opts: %s%s%s", descr, sbi->s_es->s_mount_opts,
4199 *sbi->s_es->s_mount_opts ? "; " : "", orig_data);
4201 if (es->s_error_count)
4202 mod_timer(&sbi->s_err_report, jiffies + 300*HZ); /* 5 minutes */
4204 /* Enable message ratelimiting. Default is 10 messages per 5 secs. */
4205 ratelimit_state_init(&sbi->s_err_ratelimit_state, 5 * HZ, 10);
4206 ratelimit_state_init(&sbi->s_warning_ratelimit_state, 5 * HZ, 10);
4207 ratelimit_state_init(&sbi->s_msg_ratelimit_state, 5 * HZ, 10);
4214 ext4_msg(sb, KERN_ERR, "VFS: Can't find ext4 filesystem");
4219 kobject_del(&sbi->s_kobj);
4222 ext4_unregister_li_request(sb);
4224 ext4_mb_release(sb);
4225 if (sbi->s_flex_groups)
4226 kvfree(sbi->s_flex_groups);
4227 percpu_counter_destroy(&sbi->s_freeclusters_counter);
4228 percpu_counter_destroy(&sbi->s_freeinodes_counter);
4229 percpu_counter_destroy(&sbi->s_dirs_counter);
4230 percpu_counter_destroy(&sbi->s_dirtyclusters_counter);
4232 ext4_ext_release(sb);
4233 ext4_release_system_zone(sb);
4238 ext4_msg(sb, KERN_ERR, "mount failed");
4239 if (EXT4_SB(sb)->rsv_conversion_wq)
4240 destroy_workqueue(EXT4_SB(sb)->rsv_conversion_wq);
4242 if (sbi->s_journal) {
4243 jbd2_journal_destroy(sbi->s_journal);
4244 sbi->s_journal = NULL;
4247 ext4_es_unregister_shrinker(sbi);
4249 del_timer_sync(&sbi->s_err_report);
4251 kthread_stop(sbi->s_mmp_tsk);
4253 for (i = 0; i < db_count; i++)
4254 brelse(sbi->s_group_desc[i]);
4255 kvfree(sbi->s_group_desc);
4257 if (sbi->s_chksum_driver)
4258 crypto_free_shash(sbi->s_chksum_driver);
4260 remove_proc_entry("options", sbi->s_proc);
4261 remove_proc_entry(sb->s_id, ext4_proc_root);
4264 for (i = 0; i < EXT4_MAXQUOTAS; i++)
4265 kfree(sbi->s_qf_names[i]);
4267 ext4_blkdev_remove(sbi);
4270 sb->s_fs_info = NULL;
4271 kfree(sbi->s_blockgroup_lock);
4275 return err ? err : ret;
4279 * Setup any per-fs journal parameters now. We'll do this both on
4280 * initial mount, once the journal has been initialised but before we've
4281 * done any recovery; and again on any subsequent remount.
4283 static void ext4_init_journal_params(struct super_block *sb, journal_t *journal)
4285 struct ext4_sb_info *sbi = EXT4_SB(sb);
4287 journal->j_commit_interval = sbi->s_commit_interval;
4288 journal->j_min_batch_time = sbi->s_min_batch_time;
4289 journal->j_max_batch_time = sbi->s_max_batch_time;
4291 write_lock(&journal->j_state_lock);
4292 if (test_opt(sb, BARRIER))
4293 journal->j_flags |= JBD2_BARRIER;
4295 journal->j_flags &= ~JBD2_BARRIER;
4296 if (test_opt(sb, DATA_ERR_ABORT))
4297 journal->j_flags |= JBD2_ABORT_ON_SYNCDATA_ERR;
4299 journal->j_flags &= ~JBD2_ABORT_ON_SYNCDATA_ERR;
4300 write_unlock(&journal->j_state_lock);
4303 static journal_t *ext4_get_journal(struct super_block *sb,
4304 unsigned int journal_inum)
4306 struct inode *journal_inode;
4309 BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL));
4311 /* First, test for the existence of a valid inode on disk. Bad
4312 * things happen if we iget() an unused inode, as the subsequent
4313 * iput() will try to delete it. */
4315 journal_inode = ext4_iget(sb, journal_inum);
4316 if (IS_ERR(journal_inode)) {
4317 ext4_msg(sb, KERN_ERR, "no journal found");
4320 if (!journal_inode->i_nlink) {
4321 make_bad_inode(journal_inode);
4322 iput(journal_inode);
4323 ext4_msg(sb, KERN_ERR, "journal inode is deleted");
4327 jbd_debug(2, "Journal inode found at %p: %lld bytes\n",
4328 journal_inode, journal_inode->i_size);
4329 if (!S_ISREG(journal_inode->i_mode)) {
4330 ext4_msg(sb, KERN_ERR, "invalid journal inode");
4331 iput(journal_inode);
4335 journal = jbd2_journal_init_inode(journal_inode);
4337 ext4_msg(sb, KERN_ERR, "Could not load journal inode");
4338 iput(journal_inode);
4341 journal->j_private = sb;
4342 ext4_init_journal_params(sb, journal);
4346 static journal_t *ext4_get_dev_journal(struct super_block *sb,
4349 struct buffer_head *bh;
4353 int hblock, blocksize;
4354 ext4_fsblk_t sb_block;
4355 unsigned long offset;
4356 struct ext4_super_block *es;
4357 struct block_device *bdev;
4359 BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL));
4361 bdev = ext4_blkdev_get(j_dev, sb);
4365 blocksize = sb->s_blocksize;
4366 hblock = bdev_logical_block_size(bdev);
4367 if (blocksize < hblock) {
4368 ext4_msg(sb, KERN_ERR,
4369 "blocksize too small for journal device");
4373 sb_block = EXT4_MIN_BLOCK_SIZE / blocksize;
4374 offset = EXT4_MIN_BLOCK_SIZE % blocksize;
4375 set_blocksize(bdev, blocksize);
4376 if (!(bh = __bread(bdev, sb_block, blocksize))) {
4377 ext4_msg(sb, KERN_ERR, "couldn't read superblock of "
4378 "external journal");
4382 es = (struct ext4_super_block *) (bh->b_data + offset);
4383 if ((le16_to_cpu(es->s_magic) != EXT4_SUPER_MAGIC) ||
4384 !(le32_to_cpu(es->s_feature_incompat) &
4385 EXT4_FEATURE_INCOMPAT_JOURNAL_DEV)) {
4386 ext4_msg(sb, KERN_ERR, "external journal has "
4392 if ((le32_to_cpu(es->s_feature_ro_compat) &
4393 EXT4_FEATURE_RO_COMPAT_METADATA_CSUM) &&
4394 es->s_checksum != ext4_superblock_csum(sb, es)) {
4395 ext4_msg(sb, KERN_ERR, "external journal has "
4396 "corrupt superblock");
4401 if (memcmp(EXT4_SB(sb)->s_es->s_journal_uuid, es->s_uuid, 16)) {
4402 ext4_msg(sb, KERN_ERR, "journal UUID does not match");
4407 len = ext4_blocks_count(es);
4408 start = sb_block + 1;
4409 brelse(bh); /* we're done with the superblock */
4411 journal = jbd2_journal_init_dev(bdev, sb->s_bdev,
4412 start, len, blocksize);
4414 ext4_msg(sb, KERN_ERR, "failed to create device journal");
4417 journal->j_private = sb;
4418 ll_rw_block(READ | REQ_META | REQ_PRIO, 1, &journal->j_sb_buffer);
4419 wait_on_buffer(journal->j_sb_buffer);
4420 if (!buffer_uptodate(journal->j_sb_buffer)) {
4421 ext4_msg(sb, KERN_ERR, "I/O error on journal device");
4424 if (be32_to_cpu(journal->j_superblock->s_nr_users) != 1) {
4425 ext4_msg(sb, KERN_ERR, "External journal has more than one "
4426 "user (unsupported) - %d",
4427 be32_to_cpu(journal->j_superblock->s_nr_users));
4430 EXT4_SB(sb)->journal_bdev = bdev;
4431 ext4_init_journal_params(sb, journal);
4435 jbd2_journal_destroy(journal);
4437 ext4_blkdev_put(bdev);
4441 static int ext4_load_journal(struct super_block *sb,
4442 struct ext4_super_block *es,
4443 unsigned long journal_devnum)
4446 unsigned int journal_inum = le32_to_cpu(es->s_journal_inum);
4449 int really_read_only;
4451 BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL));
4453 if (journal_devnum &&
4454 journal_devnum != le32_to_cpu(es->s_journal_dev)) {
4455 ext4_msg(sb, KERN_INFO, "external journal device major/minor "
4456 "numbers have changed");
4457 journal_dev = new_decode_dev(journal_devnum);
4459 journal_dev = new_decode_dev(le32_to_cpu(es->s_journal_dev));
4461 really_read_only = bdev_read_only(sb->s_bdev);
4464 * Are we loading a blank journal or performing recovery after a
4465 * crash? For recovery, we need to check in advance whether we
4466 * can get read-write access to the device.
4468 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER)) {
4469 if (sb->s_flags & MS_RDONLY) {
4470 ext4_msg(sb, KERN_INFO, "INFO: recovery "
4471 "required on readonly filesystem");
4472 if (really_read_only) {
4473 ext4_msg(sb, KERN_ERR, "write access "
4474 "unavailable, cannot proceed");
4477 ext4_msg(sb, KERN_INFO, "write access will "
4478 "be enabled during recovery");
4482 if (journal_inum && journal_dev) {
4483 ext4_msg(sb, KERN_ERR, "filesystem has both journal "
4484 "and inode journals!");
4489 if (!(journal = ext4_get_journal(sb, journal_inum)))
4492 if (!(journal = ext4_get_dev_journal(sb, journal_dev)))
4496 if (!(journal->j_flags & JBD2_BARRIER))
4497 ext4_msg(sb, KERN_INFO, "barriers disabled");
4499 if (!EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER))
4500 err = jbd2_journal_wipe(journal, !really_read_only);
4502 char *save = kmalloc(EXT4_S_ERR_LEN, GFP_KERNEL);
4504 memcpy(save, ((char *) es) +
4505 EXT4_S_ERR_START, EXT4_S_ERR_LEN);
4506 err = jbd2_journal_load(journal);
4508 memcpy(((char *) es) + EXT4_S_ERR_START,
4509 save, EXT4_S_ERR_LEN);
4514 ext4_msg(sb, KERN_ERR, "error loading journal");
4515 jbd2_journal_destroy(journal);
4519 EXT4_SB(sb)->s_journal = journal;
4520 ext4_clear_journal_err(sb, es);
4522 if (!really_read_only && journal_devnum &&
4523 journal_devnum != le32_to_cpu(es->s_journal_dev)) {
4524 es->s_journal_dev = cpu_to_le32(journal_devnum);
4526 /* Make sure we flush the recovery flag to disk. */
4527 ext4_commit_super(sb, 1);
4533 static int ext4_commit_super(struct super_block *sb, int sync)
4535 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
4536 struct buffer_head *sbh = EXT4_SB(sb)->s_sbh;
4539 if (!sbh || block_device_ejected(sb))
4541 if (buffer_write_io_error(sbh)) {
4543 * Oh, dear. A previous attempt to write the
4544 * superblock failed. This could happen because the
4545 * USB device was yanked out. Or it could happen to
4546 * be a transient write error and maybe the block will
4547 * be remapped. Nothing we can do but to retry the
4548 * write and hope for the best.
4550 ext4_msg(sb, KERN_ERR, "previous I/O error to "
4551 "superblock detected");
4552 clear_buffer_write_io_error(sbh);
4553 set_buffer_uptodate(sbh);
4556 * If the file system is mounted read-only, don't update the
4557 * superblock write time. This avoids updating the superblock
4558 * write time when we are mounting the root file system
4559 * read/only but we need to replay the journal; at that point,
4560 * for people who are east of GMT and who make their clock
4561 * tick in localtime for Windows bug-for-bug compatibility,
4562 * the clock is set in the future, and this will cause e2fsck
4563 * to complain and force a full file system check.
4565 if (!(sb->s_flags & MS_RDONLY))
4566 es->s_wtime = cpu_to_le32(get_seconds());
4567 if (sb->s_bdev->bd_part)
4568 es->s_kbytes_written =
4569 cpu_to_le64(EXT4_SB(sb)->s_kbytes_written +
4570 ((part_stat_read(sb->s_bdev->bd_part, sectors[1]) -
4571 EXT4_SB(sb)->s_sectors_written_start) >> 1));
4573 es->s_kbytes_written =
4574 cpu_to_le64(EXT4_SB(sb)->s_kbytes_written);
4575 if (percpu_counter_initialized(&EXT4_SB(sb)->s_freeclusters_counter))
4576 ext4_free_blocks_count_set(es,
4577 EXT4_C2B(EXT4_SB(sb), percpu_counter_sum_positive(
4578 &EXT4_SB(sb)->s_freeclusters_counter)));
4579 if (percpu_counter_initialized(&EXT4_SB(sb)->s_freeinodes_counter))
4580 es->s_free_inodes_count =
4581 cpu_to_le32(percpu_counter_sum_positive(
4582 &EXT4_SB(sb)->s_freeinodes_counter));
4583 BUFFER_TRACE(sbh, "marking dirty");
4584 ext4_superblock_csum_set(sb);
4585 mark_buffer_dirty(sbh);
4587 error = sync_dirty_buffer(sbh);
4591 error = buffer_write_io_error(sbh);
4593 ext4_msg(sb, KERN_ERR, "I/O error while writing "
4595 clear_buffer_write_io_error(sbh);
4596 set_buffer_uptodate(sbh);
4603 * Have we just finished recovery? If so, and if we are mounting (or
4604 * remounting) the filesystem readonly, then we will end up with a
4605 * consistent fs on disk. Record that fact.
4607 static void ext4_mark_recovery_complete(struct super_block *sb,
4608 struct ext4_super_block *es)
4610 journal_t *journal = EXT4_SB(sb)->s_journal;
4612 if (!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL)) {
4613 BUG_ON(journal != NULL);
4616 jbd2_journal_lock_updates(journal);
4617 if (jbd2_journal_flush(journal) < 0)
4620 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER) &&
4621 sb->s_flags & MS_RDONLY) {
4622 EXT4_CLEAR_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
4623 ext4_commit_super(sb, 1);
4627 jbd2_journal_unlock_updates(journal);
4631 * If we are mounting (or read-write remounting) a filesystem whose journal
4632 * has recorded an error from a previous lifetime, move that error to the
4633 * main filesystem now.
4635 static void ext4_clear_journal_err(struct super_block *sb,
4636 struct ext4_super_block *es)
4642 BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL));
4644 journal = EXT4_SB(sb)->s_journal;
4647 * Now check for any error status which may have been recorded in the
4648 * journal by a prior ext4_error() or ext4_abort()
4651 j_errno = jbd2_journal_errno(journal);
4655 errstr = ext4_decode_error(sb, j_errno, nbuf);
4656 ext4_warning(sb, "Filesystem error recorded "
4657 "from previous mount: %s", errstr);
4658 ext4_warning(sb, "Marking fs in need of filesystem check.");
4660 EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS;
4661 es->s_state |= cpu_to_le16(EXT4_ERROR_FS);
4662 ext4_commit_super(sb, 1);
4664 jbd2_journal_clear_err(journal);
4665 jbd2_journal_update_sb_errno(journal);
4670 * Force the running and committing transactions to commit,
4671 * and wait on the commit.
4673 int ext4_force_commit(struct super_block *sb)
4677 if (sb->s_flags & MS_RDONLY)
4680 journal = EXT4_SB(sb)->s_journal;
4681 return ext4_journal_force_commit(journal);
4684 static int ext4_sync_fs(struct super_block *sb, int wait)
4688 bool needs_barrier = false;
4689 struct ext4_sb_info *sbi = EXT4_SB(sb);
4691 trace_ext4_sync_fs(sb, wait);
4692 flush_workqueue(sbi->rsv_conversion_wq);
4694 * Writeback quota in non-journalled quota case - journalled quota has
4697 dquot_writeback_dquots(sb, -1);
4699 * Data writeback is possible w/o journal transaction, so barrier must
4700 * being sent at the end of the function. But we can skip it if
4701 * transaction_commit will do it for us.
4703 if (sbi->s_journal) {
4704 target = jbd2_get_latest_transaction(sbi->s_journal);
4705 if (wait && sbi->s_journal->j_flags & JBD2_BARRIER &&
4706 !jbd2_trans_will_send_data_barrier(sbi->s_journal, target))
4707 needs_barrier = true;
4709 if (jbd2_journal_start_commit(sbi->s_journal, &target)) {
4711 ret = jbd2_log_wait_commit(sbi->s_journal,
4714 } else if (wait && test_opt(sb, BARRIER))
4715 needs_barrier = true;
4716 if (needs_barrier) {
4718 err = blkdev_issue_flush(sb->s_bdev, GFP_KERNEL, NULL);
4727 * LVM calls this function before a (read-only) snapshot is created. This
4728 * gives us a chance to flush the journal completely and mark the fs clean.
4730 * Note that only this function cannot bring a filesystem to be in a clean
4731 * state independently. It relies on upper layer to stop all data & metadata
4734 static int ext4_freeze(struct super_block *sb)
4739 if (sb->s_flags & MS_RDONLY)
4742 journal = EXT4_SB(sb)->s_journal;
4745 /* Now we set up the journal barrier. */
4746 jbd2_journal_lock_updates(journal);
4749 * Don't clear the needs_recovery flag if we failed to
4750 * flush the journal.
4752 error = jbd2_journal_flush(journal);
4757 /* Journal blocked and flushed, clear needs_recovery flag. */
4758 EXT4_CLEAR_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
4759 error = ext4_commit_super(sb, 1);
4762 /* we rely on upper layer to stop further updates */
4763 jbd2_journal_unlock_updates(journal);
4768 * Called by LVM after the snapshot is done. We need to reset the RECOVER
4769 * flag here, even though the filesystem is not technically dirty yet.
4771 static int ext4_unfreeze(struct super_block *sb)
4773 if (sb->s_flags & MS_RDONLY)
4776 /* Reset the needs_recovery flag before the fs is unlocked. */
4777 EXT4_SET_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
4778 ext4_commit_super(sb, 1);
4783 * Structure to save mount options for ext4_remount's benefit
4785 struct ext4_mount_options {
4786 unsigned long s_mount_opt;
4787 unsigned long s_mount_opt2;
4790 unsigned long s_commit_interval;
4791 u32 s_min_batch_time, s_max_batch_time;
4794 char *s_qf_names[EXT4_MAXQUOTAS];
4798 static int ext4_remount(struct super_block *sb, int *flags, char *data)
4800 struct ext4_super_block *es;
4801 struct ext4_sb_info *sbi = EXT4_SB(sb);
4802 unsigned long old_sb_flags;
4803 struct ext4_mount_options old_opts;
4804 int enable_quota = 0;
4806 unsigned int journal_ioprio = DEFAULT_JOURNAL_IOPRIO;
4811 char *orig_data = kstrdup(data, GFP_KERNEL);
4813 /* Store the original options */
4814 old_sb_flags = sb->s_flags;
4815 old_opts.s_mount_opt = sbi->s_mount_opt;
4816 old_opts.s_mount_opt2 = sbi->s_mount_opt2;
4817 old_opts.s_resuid = sbi->s_resuid;
4818 old_opts.s_resgid = sbi->s_resgid;
4819 old_opts.s_commit_interval = sbi->s_commit_interval;
4820 old_opts.s_min_batch_time = sbi->s_min_batch_time;
4821 old_opts.s_max_batch_time = sbi->s_max_batch_time;
4823 old_opts.s_jquota_fmt = sbi->s_jquota_fmt;
4824 for (i = 0; i < EXT4_MAXQUOTAS; i++)
4825 if (sbi->s_qf_names[i]) {
4826 old_opts.s_qf_names[i] = kstrdup(sbi->s_qf_names[i],
4828 if (!old_opts.s_qf_names[i]) {
4829 for (j = 0; j < i; j++)
4830 kfree(old_opts.s_qf_names[j]);
4835 old_opts.s_qf_names[i] = NULL;
4837 if (sbi->s_journal && sbi->s_journal->j_task->io_context)
4838 journal_ioprio = sbi->s_journal->j_task->io_context->ioprio;
4841 * Allow the "check" option to be passed as a remount option.
4843 if (!parse_options(data, sb, NULL, &journal_ioprio, 1)) {
4848 if ((old_opts.s_mount_opt & EXT4_MOUNT_JOURNAL_CHECKSUM) ^
4849 test_opt(sb, JOURNAL_CHECKSUM)) {
4850 ext4_msg(sb, KERN_ERR, "changing journal_checksum "
4851 "during remount not supported");
4856 if ((old_opts.s_mount_opt & EXT4_MOUNT_JOURNAL_CHECKSUM) ^
4857 test_opt(sb, JOURNAL_CHECKSUM)) {
4858 ext4_msg(sb, KERN_ERR, "changing journal_checksum "
4859 "during remount not supported");
4864 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA) {
4865 if (test_opt2(sb, EXPLICIT_DELALLOC)) {
4866 ext4_msg(sb, KERN_ERR, "can't mount with "
4867 "both data=journal and delalloc");
4871 if (test_opt(sb, DIOREAD_NOLOCK)) {
4872 ext4_msg(sb, KERN_ERR, "can't mount with "
4873 "both data=journal and dioread_nolock");
4879 if (sbi->s_mount_flags & EXT4_MF_FS_ABORTED)
4880 ext4_abort(sb, "Abort forced by user");
4882 sb->s_flags = (sb->s_flags & ~MS_POSIXACL) |
4883 (test_opt(sb, POSIX_ACL) ? MS_POSIXACL : 0);
4887 if (sbi->s_journal) {
4888 ext4_init_journal_params(sb, sbi->s_journal);
4889 set_task_ioprio(sbi->s_journal->j_task, journal_ioprio);
4892 if ((*flags & MS_RDONLY) != (sb->s_flags & MS_RDONLY)) {
4893 if (sbi->s_mount_flags & EXT4_MF_FS_ABORTED) {
4898 if (*flags & MS_RDONLY) {
4899 err = sync_filesystem(sb);
4902 err = dquot_suspend(sb, -1);
4907 * First of all, the unconditional stuff we have to do
4908 * to disable replay of the journal when we next remount
4910 sb->s_flags |= MS_RDONLY;
4913 * OK, test if we are remounting a valid rw partition
4914 * readonly, and if so set the rdonly flag and then
4915 * mark the partition as valid again.
4917 if (!(es->s_state & cpu_to_le16(EXT4_VALID_FS)) &&
4918 (sbi->s_mount_state & EXT4_VALID_FS))
4919 es->s_state = cpu_to_le16(sbi->s_mount_state);
4922 ext4_mark_recovery_complete(sb, es);
4924 /* Make sure we can mount this feature set readwrite */
4925 if (!ext4_feature_set_ok(sb, 0)) {
4930 * Make sure the group descriptor checksums
4931 * are sane. If they aren't, refuse to remount r/w.
4933 for (g = 0; g < sbi->s_groups_count; g++) {
4934 struct ext4_group_desc *gdp =
4935 ext4_get_group_desc(sb, g, NULL);
4937 if (!ext4_group_desc_csum_verify(sb, g, gdp)) {
4938 ext4_msg(sb, KERN_ERR,
4939 "ext4_remount: Checksum for group %u failed (%u!=%u)",
4940 g, le16_to_cpu(ext4_group_desc_csum(sbi, g, gdp)),
4941 le16_to_cpu(gdp->bg_checksum));
4948 * If we have an unprocessed orphan list hanging
4949 * around from a previously readonly bdev mount,
4950 * require a full umount/remount for now.
4952 if (es->s_last_orphan) {
4953 ext4_msg(sb, KERN_WARNING, "Couldn't "
4954 "remount RDWR because of unprocessed "
4955 "orphan inode list. Please "
4956 "umount/remount instead");
4962 * Mounting a RDONLY partition read-write, so reread
4963 * and store the current valid flag. (It may have
4964 * been changed by e2fsck since we originally mounted
4968 ext4_clear_journal_err(sb, es);
4969 sbi->s_mount_state = le16_to_cpu(es->s_state);
4970 if (!ext4_setup_super(sb, es, 0))
4971 sb->s_flags &= ~MS_RDONLY;
4972 if (EXT4_HAS_INCOMPAT_FEATURE(sb,
4973 EXT4_FEATURE_INCOMPAT_MMP))
4974 if (ext4_multi_mount_protect(sb,
4975 le64_to_cpu(es->s_mmp_block))) {
4984 * Reinitialize lazy itable initialization thread based on
4987 if ((sb->s_flags & MS_RDONLY) || !test_opt(sb, INIT_INODE_TABLE))
4988 ext4_unregister_li_request(sb);
4990 ext4_group_t first_not_zeroed;
4991 first_not_zeroed = ext4_has_uninit_itable(sb);
4992 ext4_register_li_request(sb, first_not_zeroed);
4995 ext4_setup_system_zone(sb);
4996 if (sbi->s_journal == NULL && !(old_sb_flags & MS_RDONLY))
4997 ext4_commit_super(sb, 1);
5000 /* Release old quota file names */
5001 for (i = 0; i < EXT4_MAXQUOTAS; i++)
5002 kfree(old_opts.s_qf_names[i]);
5004 if (sb_any_quota_suspended(sb))
5005 dquot_resume(sb, -1);
5006 else if (EXT4_HAS_RO_COMPAT_FEATURE(sb,
5007 EXT4_FEATURE_RO_COMPAT_QUOTA)) {
5008 err = ext4_enable_quotas(sb);
5015 ext4_msg(sb, KERN_INFO, "re-mounted. Opts: %s", orig_data);
5020 sb->s_flags = old_sb_flags;
5021 sbi->s_mount_opt = old_opts.s_mount_opt;
5022 sbi->s_mount_opt2 = old_opts.s_mount_opt2;
5023 sbi->s_resuid = old_opts.s_resuid;
5024 sbi->s_resgid = old_opts.s_resgid;
5025 sbi->s_commit_interval = old_opts.s_commit_interval;
5026 sbi->s_min_batch_time = old_opts.s_min_batch_time;
5027 sbi->s_max_batch_time = old_opts.s_max_batch_time;
5029 sbi->s_jquota_fmt = old_opts.s_jquota_fmt;
5030 for (i = 0; i < EXT4_MAXQUOTAS; i++) {
5031 kfree(sbi->s_qf_names[i]);
5032 sbi->s_qf_names[i] = old_opts.s_qf_names[i];
5039 static int ext4_statfs(struct dentry *dentry, struct kstatfs *buf)
5041 struct super_block *sb = dentry->d_sb;
5042 struct ext4_sb_info *sbi = EXT4_SB(sb);
5043 struct ext4_super_block *es = sbi->s_es;
5044 ext4_fsblk_t overhead = 0, resv_blocks;
5047 resv_blocks = EXT4_C2B(sbi, atomic64_read(&sbi->s_resv_clusters));
5049 if (!test_opt(sb, MINIX_DF))
5050 overhead = sbi->s_overhead;
5052 buf->f_type = EXT4_SUPER_MAGIC;
5053 buf->f_bsize = sb->s_blocksize;
5054 buf->f_blocks = ext4_blocks_count(es) - EXT4_C2B(sbi, overhead);
5055 bfree = percpu_counter_sum_positive(&sbi->s_freeclusters_counter) -
5056 percpu_counter_sum_positive(&sbi->s_dirtyclusters_counter);
5057 /* prevent underflow in case that few free space is available */
5058 buf->f_bfree = EXT4_C2B(sbi, max_t(s64, bfree, 0));
5059 buf->f_bavail = buf->f_bfree -
5060 (ext4_r_blocks_count(es) + resv_blocks);
5061 if (buf->f_bfree < (ext4_r_blocks_count(es) + resv_blocks))
5063 buf->f_files = le32_to_cpu(es->s_inodes_count);
5064 buf->f_ffree = percpu_counter_sum_positive(&sbi->s_freeinodes_counter);
5065 buf->f_namelen = EXT4_NAME_LEN;
5066 fsid = le64_to_cpup((void *)es->s_uuid) ^
5067 le64_to_cpup((void *)es->s_uuid + sizeof(u64));
5068 buf->f_fsid.val[0] = fsid & 0xFFFFFFFFUL;
5069 buf->f_fsid.val[1] = (fsid >> 32) & 0xFFFFFFFFUL;
5074 /* Helper function for writing quotas on sync - we need to start transaction
5075 * before quota file is locked for write. Otherwise the are possible deadlocks:
5076 * Process 1 Process 2
5077 * ext4_create() quota_sync()
5078 * jbd2_journal_start() write_dquot()
5079 * dquot_initialize() down(dqio_mutex)
5080 * down(dqio_mutex) jbd2_journal_start()
5086 static inline struct inode *dquot_to_inode(struct dquot *dquot)
5088 return sb_dqopt(dquot->dq_sb)->files[dquot->dq_id.type];
5091 static int ext4_write_dquot(struct dquot *dquot)
5095 struct inode *inode;
5097 inode = dquot_to_inode(dquot);
5098 handle = ext4_journal_start(inode, EXT4_HT_QUOTA,
5099 EXT4_QUOTA_TRANS_BLOCKS(dquot->dq_sb));
5101 return PTR_ERR(handle);
5102 ret = dquot_commit(dquot);
5103 err = ext4_journal_stop(handle);
5109 static int ext4_acquire_dquot(struct dquot *dquot)
5114 handle = ext4_journal_start(dquot_to_inode(dquot), EXT4_HT_QUOTA,
5115 EXT4_QUOTA_INIT_BLOCKS(dquot->dq_sb));
5117 return PTR_ERR(handle);
5118 ret = dquot_acquire(dquot);
5119 err = ext4_journal_stop(handle);
5125 static int ext4_release_dquot(struct dquot *dquot)
5130 handle = ext4_journal_start(dquot_to_inode(dquot), EXT4_HT_QUOTA,
5131 EXT4_QUOTA_DEL_BLOCKS(dquot->dq_sb));
5132 if (IS_ERR(handle)) {
5133 /* Release dquot anyway to avoid endless cycle in dqput() */
5134 dquot_release(dquot);
5135 return PTR_ERR(handle);
5137 ret = dquot_release(dquot);
5138 err = ext4_journal_stop(handle);
5144 static int ext4_mark_dquot_dirty(struct dquot *dquot)
5146 struct super_block *sb = dquot->dq_sb;
5147 struct ext4_sb_info *sbi = EXT4_SB(sb);
5149 /* Are we journaling quotas? */
5150 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_QUOTA) ||
5151 sbi->s_qf_names[USRQUOTA] || sbi->s_qf_names[GRPQUOTA]) {
5152 dquot_mark_dquot_dirty(dquot);
5153 return ext4_write_dquot(dquot);
5155 return dquot_mark_dquot_dirty(dquot);
5159 static int ext4_write_info(struct super_block *sb, int type)
5164 /* Data block + inode block */
5165 handle = ext4_journal_start(sb->s_root->d_inode, EXT4_HT_QUOTA, 2);
5167 return PTR_ERR(handle);
5168 ret = dquot_commit_info(sb, type);
5169 err = ext4_journal_stop(handle);
5176 * Turn on quotas during mount time - we need to find
5177 * the quota file and such...
5179 static int ext4_quota_on_mount(struct super_block *sb, int type)
5181 return dquot_quota_on_mount(sb, EXT4_SB(sb)->s_qf_names[type],
5182 EXT4_SB(sb)->s_jquota_fmt, type);
5186 * Standard function to be called on quota_on
5188 static int ext4_quota_on(struct super_block *sb, int type, int format_id,
5193 if (!test_opt(sb, QUOTA))
5196 /* Quotafile not on the same filesystem? */
5197 if (path->dentry->d_sb != sb)
5199 /* Journaling quota? */
5200 if (EXT4_SB(sb)->s_qf_names[type]) {
5201 /* Quotafile not in fs root? */
5202 if (path->dentry->d_parent != sb->s_root)
5203 ext4_msg(sb, KERN_WARNING,
5204 "Quota file not on filesystem root. "
5205 "Journaled quota will not work");
5209 * When we journal data on quota file, we have to flush journal to see
5210 * all updates to the file when we bypass pagecache...
5212 if (EXT4_SB(sb)->s_journal &&
5213 ext4_should_journal_data(path->dentry->d_inode)) {
5215 * We don't need to lock updates but journal_flush() could
5216 * otherwise be livelocked...
5218 jbd2_journal_lock_updates(EXT4_SB(sb)->s_journal);
5219 err = jbd2_journal_flush(EXT4_SB(sb)->s_journal);
5220 jbd2_journal_unlock_updates(EXT4_SB(sb)->s_journal);
5225 return dquot_quota_on(sb, type, format_id, path);
5228 static int ext4_quota_enable(struct super_block *sb, int type, int format_id,
5232 struct inode *qf_inode;
5233 unsigned long qf_inums[EXT4_MAXQUOTAS] = {
5234 le32_to_cpu(EXT4_SB(sb)->s_es->s_usr_quota_inum),
5235 le32_to_cpu(EXT4_SB(sb)->s_es->s_grp_quota_inum)
5238 BUG_ON(!EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_QUOTA));
5240 if (!qf_inums[type])
5243 qf_inode = ext4_iget(sb, qf_inums[type]);
5244 if (IS_ERR(qf_inode)) {
5245 ext4_error(sb, "Bad quota inode # %lu", qf_inums[type]);
5246 return PTR_ERR(qf_inode);
5249 /* Don't account quota for quota files to avoid recursion */
5250 qf_inode->i_flags |= S_NOQUOTA;
5251 err = dquot_enable(qf_inode, type, format_id, flags);
5257 /* Enable usage tracking for all quota types. */
5258 static int ext4_enable_quotas(struct super_block *sb)
5261 unsigned long qf_inums[EXT4_MAXQUOTAS] = {
5262 le32_to_cpu(EXT4_SB(sb)->s_es->s_usr_quota_inum),
5263 le32_to_cpu(EXT4_SB(sb)->s_es->s_grp_quota_inum)
5266 sb_dqopt(sb)->flags |= DQUOT_QUOTA_SYS_FILE;
5267 for (type = 0; type < EXT4_MAXQUOTAS; type++) {
5268 if (qf_inums[type]) {
5269 err = ext4_quota_enable(sb, type, QFMT_VFS_V1,
5270 DQUOT_USAGE_ENABLED);
5273 "Failed to enable quota tracking "
5274 "(type=%d, err=%d). Please run "
5275 "e2fsck to fix.", type, err);
5284 * quota_on function that is used when QUOTA feature is set.
5286 static int ext4_quota_on_sysfile(struct super_block *sb, int type,
5289 if (!EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_QUOTA))
5293 * USAGE was enabled at mount time. Only need to enable LIMITS now.
5295 return ext4_quota_enable(sb, type, format_id, DQUOT_LIMITS_ENABLED);
5298 static int ext4_quota_off(struct super_block *sb, int type)
5300 struct inode *inode = sb_dqopt(sb)->files[type];
5303 /* Force all delayed allocation blocks to be allocated.
5304 * Caller already holds s_umount sem */
5305 if (test_opt(sb, DELALLOC))
5306 sync_filesystem(sb);
5311 /* Update modification times of quota files when userspace can
5312 * start looking at them */
5313 handle = ext4_journal_start(inode, EXT4_HT_QUOTA, 1);
5316 inode->i_mtime = inode->i_ctime = CURRENT_TIME;
5317 ext4_mark_inode_dirty(handle, inode);
5318 ext4_journal_stop(handle);
5321 return dquot_quota_off(sb, type);
5325 * quota_off function that is used when QUOTA feature is set.
5327 static int ext4_quota_off_sysfile(struct super_block *sb, int type)
5329 if (!EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_QUOTA))
5332 /* Disable only the limits. */
5333 return dquot_disable(sb, type, DQUOT_LIMITS_ENABLED);
5336 /* Read data from quotafile - avoid pagecache and such because we cannot afford
5337 * acquiring the locks... As quota files are never truncated and quota code
5338 * itself serializes the operations (and no one else should touch the files)
5339 * we don't have to be afraid of races */
5340 static ssize_t ext4_quota_read(struct super_block *sb, int type, char *data,
5341 size_t len, loff_t off)
5343 struct inode *inode = sb_dqopt(sb)->files[type];
5344 ext4_lblk_t blk = off >> EXT4_BLOCK_SIZE_BITS(sb);
5345 int offset = off & (sb->s_blocksize - 1);
5348 struct buffer_head *bh;
5349 loff_t i_size = i_size_read(inode);
5353 if (off+len > i_size)
5356 while (toread > 0) {
5357 tocopy = sb->s_blocksize - offset < toread ?
5358 sb->s_blocksize - offset : toread;
5359 bh = ext4_bread(NULL, inode, blk, 0);
5362 if (!bh) /* A hole? */
5363 memset(data, 0, tocopy);
5365 memcpy(data, bh->b_data+offset, tocopy);
5375 /* Write to quotafile (we know the transaction is already started and has
5376 * enough credits) */
5377 static ssize_t ext4_quota_write(struct super_block *sb, int type,
5378 const char *data, size_t len, loff_t off)
5380 struct inode *inode = sb_dqopt(sb)->files[type];
5381 ext4_lblk_t blk = off >> EXT4_BLOCK_SIZE_BITS(sb);
5382 int err, offset = off & (sb->s_blocksize - 1);
5383 struct buffer_head *bh;
5384 handle_t *handle = journal_current_handle();
5386 if (EXT4_SB(sb)->s_journal && !handle) {
5387 ext4_msg(sb, KERN_WARNING, "Quota write (off=%llu, len=%llu)"
5388 " cancelled because transaction is not started",
5389 (unsigned long long)off, (unsigned long long)len);
5393 * Since we account only one data block in transaction credits,
5394 * then it is impossible to cross a block boundary.
5396 if (sb->s_blocksize - offset < len) {
5397 ext4_msg(sb, KERN_WARNING, "Quota write (off=%llu, len=%llu)"
5398 " cancelled because not block aligned",
5399 (unsigned long long)off, (unsigned long long)len);
5403 bh = ext4_bread(handle, inode, blk, 1);
5408 BUFFER_TRACE(bh, "get write access");
5409 err = ext4_journal_get_write_access(handle, bh);
5415 memcpy(bh->b_data+offset, data, len);
5416 flush_dcache_page(bh->b_page);
5418 err = ext4_handle_dirty_metadata(handle, NULL, bh);
5421 if (inode->i_size < off + len) {
5422 i_size_write(inode, off + len);
5423 EXT4_I(inode)->i_disksize = inode->i_size;
5424 ext4_mark_inode_dirty(handle, inode);
5431 static struct dentry *ext4_mount(struct file_system_type *fs_type, int flags,
5432 const char *dev_name, void *data)
5434 return mount_bdev(fs_type, flags, dev_name, data, ext4_fill_super);
5437 #if !defined(CONFIG_EXT2_FS) && !defined(CONFIG_EXT2_FS_MODULE) && defined(CONFIG_EXT4_USE_FOR_EXT23)
5438 static inline void register_as_ext2(void)
5440 int err = register_filesystem(&ext2_fs_type);
5443 "EXT4-fs: Unable to register as ext2 (%d)\n", err);
5446 static inline void unregister_as_ext2(void)
5448 unregister_filesystem(&ext2_fs_type);
5451 static inline int ext2_feature_set_ok(struct super_block *sb)
5453 if (EXT4_HAS_INCOMPAT_FEATURE(sb, ~EXT2_FEATURE_INCOMPAT_SUPP))
5455 if (sb->s_flags & MS_RDONLY)
5457 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, ~EXT2_FEATURE_RO_COMPAT_SUPP))
5462 static inline void register_as_ext2(void) { }
5463 static inline void unregister_as_ext2(void) { }
5464 static inline int ext2_feature_set_ok(struct super_block *sb) { return 0; }
5467 #if !defined(CONFIG_EXT3_FS) && !defined(CONFIG_EXT3_FS_MODULE) && defined(CONFIG_EXT4_USE_FOR_EXT23)
5468 static inline void register_as_ext3(void)
5470 int err = register_filesystem(&ext3_fs_type);
5473 "EXT4-fs: Unable to register as ext3 (%d)\n", err);
5476 static inline void unregister_as_ext3(void)
5478 unregister_filesystem(&ext3_fs_type);
5481 static inline int ext3_feature_set_ok(struct super_block *sb)
5483 if (EXT4_HAS_INCOMPAT_FEATURE(sb, ~EXT3_FEATURE_INCOMPAT_SUPP))
5485 if (!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL))
5487 if (sb->s_flags & MS_RDONLY)
5489 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, ~EXT3_FEATURE_RO_COMPAT_SUPP))
5494 static inline void register_as_ext3(void) { }
5495 static inline void unregister_as_ext3(void) { }
5496 static inline int ext3_feature_set_ok(struct super_block *sb) { return 0; }
5499 static struct file_system_type ext4_fs_type = {
5500 .owner = THIS_MODULE,
5502 .mount = ext4_mount,
5503 .kill_sb = kill_block_super,
5504 .fs_flags = FS_REQUIRES_DEV,
5506 MODULE_ALIAS_FS("ext4");
5508 static int __init ext4_init_feat_adverts(void)
5510 struct ext4_features *ef;
5513 ef = kzalloc(sizeof(struct ext4_features), GFP_KERNEL);
5517 ef->f_kobj.kset = ext4_kset;
5518 init_completion(&ef->f_kobj_unregister);
5519 ret = kobject_init_and_add(&ef->f_kobj, &ext4_feat_ktype, NULL,
5532 static void ext4_exit_feat_adverts(void)
5534 kobject_put(&ext4_feat->f_kobj);
5535 wait_for_completion(&ext4_feat->f_kobj_unregister);
5539 /* Shared across all ext4 file systems */
5540 wait_queue_head_t ext4__ioend_wq[EXT4_WQ_HASH_SZ];
5541 struct mutex ext4__aio_mutex[EXT4_WQ_HASH_SZ];
5543 static int __init ext4_init_fs(void)
5547 ext4_li_info = NULL;
5548 mutex_init(&ext4_li_mtx);
5550 /* Build-time check for flags consistency */
5551 ext4_check_flag_values();
5553 for (i = 0; i < EXT4_WQ_HASH_SZ; i++) {
5554 mutex_init(&ext4__aio_mutex[i]);
5555 init_waitqueue_head(&ext4__ioend_wq[i]);
5558 err = ext4_init_es();
5562 err = ext4_init_pageio();
5566 err = ext4_init_system_zone();
5569 ext4_kset = kset_create_and_add("ext4", NULL, fs_kobj);
5574 ext4_proc_root = proc_mkdir("fs/ext4", NULL);
5576 err = ext4_init_feat_adverts();
5580 err = ext4_init_mballoc();
5584 ext4_mballoc_ready = 1;
5585 err = init_inodecache();
5590 err = register_filesystem(&ext4_fs_type);
5596 unregister_as_ext2();
5597 unregister_as_ext3();
5598 destroy_inodecache();
5600 ext4_mballoc_ready = 0;
5601 ext4_exit_mballoc();
5603 ext4_exit_feat_adverts();
5606 remove_proc_entry("fs/ext4", NULL);
5607 kset_unregister(ext4_kset);
5609 ext4_exit_system_zone();
5618 static void __exit ext4_exit_fs(void)
5620 ext4_destroy_lazyinit_thread();
5621 unregister_as_ext2();
5622 unregister_as_ext3();
5623 unregister_filesystem(&ext4_fs_type);
5624 destroy_inodecache();
5625 ext4_exit_mballoc();
5626 ext4_exit_feat_adverts();
5627 remove_proc_entry("fs/ext4", NULL);
5628 kset_unregister(ext4_kset);
5629 ext4_exit_system_zone();
5634 MODULE_AUTHOR("Remy Card, Stephen Tweedie, Andrew Morton, Andreas Dilger, Theodore Ts'o and others");
5635 MODULE_DESCRIPTION("Fourth Extended Filesystem");
5636 MODULE_LICENSE("GPL");
5637 module_init(ext4_init_fs)
5638 module_exit(ext4_exit_fs)