2 * linux/fs/ext4/super.c
4 * Copyright (C) 1992, 1993, 1994, 1995
5 * Remy Card (card@masi.ibp.fr)
6 * Laboratoire MASI - Institut Blaise Pascal
7 * Universite Pierre et Marie Curie (Paris VI)
11 * linux/fs/minix/inode.c
13 * Copyright (C) 1991, 1992 Linus Torvalds
15 * Big-endian to little-endian byte-swapping/bitmaps by
16 * David S. Miller (davem@caip.rutgers.edu), 1995
19 #include <linux/module.h>
20 #include <linux/string.h>
22 #include <linux/time.h>
23 #include <linux/vmalloc.h>
24 #include <linux/jbd2.h>
25 #include <linux/slab.h>
26 #include <linux/init.h>
27 #include <linux/blkdev.h>
28 #include <linux/parser.h>
29 #include <linux/buffer_head.h>
30 #include <linux/exportfs.h>
31 #include <linux/vfs.h>
32 #include <linux/random.h>
33 #include <linux/mount.h>
34 #include <linux/namei.h>
35 #include <linux/quotaops.h>
36 #include <linux/seq_file.h>
37 #include <linux/proc_fs.h>
38 #include <linux/ctype.h>
39 #include <linux/log2.h>
40 #include <linux/crc16.h>
41 #include <linux/cleancache.h>
42 #include <asm/uaccess.h>
44 #include <linux/kthread.h>
45 #include <linux/freezer.h>
48 #include "ext4_extents.h" /* Needed for trace points definition */
49 #include "ext4_jbd2.h"
54 #define CREATE_TRACE_POINTS
55 #include <trace/events/ext4.h>
57 static struct proc_dir_entry *ext4_proc_root;
58 static struct kset *ext4_kset;
59 static struct ext4_lazy_init *ext4_li_info;
60 static struct mutex ext4_li_mtx;
61 static struct ext4_features *ext4_feat;
62 static int ext4_mballoc_ready;
64 static int ext4_load_journal(struct super_block *, struct ext4_super_block *,
65 unsigned long journal_devnum);
66 static int ext4_show_options(struct seq_file *seq, struct dentry *root);
67 static int ext4_commit_super(struct super_block *sb, int sync);
68 static void ext4_mark_recovery_complete(struct super_block *sb,
69 struct ext4_super_block *es);
70 static void ext4_clear_journal_err(struct super_block *sb,
71 struct ext4_super_block *es);
72 static int ext4_sync_fs(struct super_block *sb, int wait);
73 static int ext4_sync_fs_nojournal(struct super_block *sb, int wait);
74 static int ext4_remount(struct super_block *sb, int *flags, char *data);
75 static int ext4_statfs(struct dentry *dentry, struct kstatfs *buf);
76 static int ext4_unfreeze(struct super_block *sb);
77 static int ext4_freeze(struct super_block *sb);
78 static struct dentry *ext4_mount(struct file_system_type *fs_type, int flags,
79 const char *dev_name, void *data);
80 static inline int ext2_feature_set_ok(struct super_block *sb);
81 static inline int ext3_feature_set_ok(struct super_block *sb);
82 static int ext4_feature_set_ok(struct super_block *sb, int readonly);
83 static void ext4_destroy_lazyinit_thread(void);
84 static void ext4_unregister_li_request(struct super_block *sb);
85 static void ext4_clear_request_list(void);
86 static int ext4_reserve_clusters(struct ext4_sb_info *, ext4_fsblk_t);
88 #if !defined(CONFIG_EXT2_FS) && !defined(CONFIG_EXT2_FS_MODULE) && defined(CONFIG_EXT4_USE_FOR_EXT23)
89 static struct file_system_type ext2_fs_type = {
93 .kill_sb = kill_block_super,
94 .fs_flags = FS_REQUIRES_DEV,
96 MODULE_ALIAS_FS("ext2");
98 #define IS_EXT2_SB(sb) ((sb)->s_bdev->bd_holder == &ext2_fs_type)
100 #define IS_EXT2_SB(sb) (0)
104 #if !defined(CONFIG_EXT3_FS) && !defined(CONFIG_EXT3_FS_MODULE) && defined(CONFIG_EXT4_USE_FOR_EXT23)
105 static struct file_system_type ext3_fs_type = {
106 .owner = THIS_MODULE,
109 .kill_sb = kill_block_super,
110 .fs_flags = FS_REQUIRES_DEV,
112 MODULE_ALIAS_FS("ext3");
113 MODULE_ALIAS("ext3");
114 #define IS_EXT3_SB(sb) ((sb)->s_bdev->bd_holder == &ext3_fs_type)
116 #define IS_EXT3_SB(sb) (0)
119 static int ext4_verify_csum_type(struct super_block *sb,
120 struct ext4_super_block *es)
122 if (!EXT4_HAS_RO_COMPAT_FEATURE(sb,
123 EXT4_FEATURE_RO_COMPAT_METADATA_CSUM))
126 return es->s_checksum_type == EXT4_CRC32C_CHKSUM;
129 static __le32 ext4_superblock_csum(struct super_block *sb,
130 struct ext4_super_block *es)
132 struct ext4_sb_info *sbi = EXT4_SB(sb);
133 int offset = offsetof(struct ext4_super_block, s_checksum);
136 csum = ext4_chksum(sbi, ~0, (char *)es, offset);
138 return cpu_to_le32(csum);
141 static int ext4_superblock_csum_verify(struct super_block *sb,
142 struct ext4_super_block *es)
144 if (!EXT4_HAS_RO_COMPAT_FEATURE(sb,
145 EXT4_FEATURE_RO_COMPAT_METADATA_CSUM))
148 return es->s_checksum == ext4_superblock_csum(sb, es);
151 void ext4_superblock_csum_set(struct super_block *sb)
153 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
155 if (!EXT4_HAS_RO_COMPAT_FEATURE(sb,
156 EXT4_FEATURE_RO_COMPAT_METADATA_CSUM))
159 es->s_checksum = ext4_superblock_csum(sb, es);
162 void *ext4_kvmalloc(size_t size, gfp_t flags)
166 ret = kmalloc(size, flags | __GFP_NOWARN);
168 ret = __vmalloc(size, flags, PAGE_KERNEL);
172 void *ext4_kvzalloc(size_t size, gfp_t flags)
176 ret = kzalloc(size, flags | __GFP_NOWARN);
178 ret = __vmalloc(size, flags | __GFP_ZERO, PAGE_KERNEL);
182 void ext4_kvfree(void *ptr)
184 if (is_vmalloc_addr(ptr))
191 ext4_fsblk_t ext4_block_bitmap(struct super_block *sb,
192 struct ext4_group_desc *bg)
194 return le32_to_cpu(bg->bg_block_bitmap_lo) |
195 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
196 (ext4_fsblk_t)le32_to_cpu(bg->bg_block_bitmap_hi) << 32 : 0);
199 ext4_fsblk_t ext4_inode_bitmap(struct super_block *sb,
200 struct ext4_group_desc *bg)
202 return le32_to_cpu(bg->bg_inode_bitmap_lo) |
203 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
204 (ext4_fsblk_t)le32_to_cpu(bg->bg_inode_bitmap_hi) << 32 : 0);
207 ext4_fsblk_t ext4_inode_table(struct super_block *sb,
208 struct ext4_group_desc *bg)
210 return le32_to_cpu(bg->bg_inode_table_lo) |
211 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
212 (ext4_fsblk_t)le32_to_cpu(bg->bg_inode_table_hi) << 32 : 0);
215 __u32 ext4_free_group_clusters(struct super_block *sb,
216 struct ext4_group_desc *bg)
218 return le16_to_cpu(bg->bg_free_blocks_count_lo) |
219 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
220 (__u32)le16_to_cpu(bg->bg_free_blocks_count_hi) << 16 : 0);
223 __u32 ext4_free_inodes_count(struct super_block *sb,
224 struct ext4_group_desc *bg)
226 return le16_to_cpu(bg->bg_free_inodes_count_lo) |
227 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
228 (__u32)le16_to_cpu(bg->bg_free_inodes_count_hi) << 16 : 0);
231 __u32 ext4_used_dirs_count(struct super_block *sb,
232 struct ext4_group_desc *bg)
234 return le16_to_cpu(bg->bg_used_dirs_count_lo) |
235 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
236 (__u32)le16_to_cpu(bg->bg_used_dirs_count_hi) << 16 : 0);
239 __u32 ext4_itable_unused_count(struct super_block *sb,
240 struct ext4_group_desc *bg)
242 return le16_to_cpu(bg->bg_itable_unused_lo) |
243 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
244 (__u32)le16_to_cpu(bg->bg_itable_unused_hi) << 16 : 0);
247 void ext4_block_bitmap_set(struct super_block *sb,
248 struct ext4_group_desc *bg, ext4_fsblk_t blk)
250 bg->bg_block_bitmap_lo = cpu_to_le32((u32)blk);
251 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
252 bg->bg_block_bitmap_hi = cpu_to_le32(blk >> 32);
255 void ext4_inode_bitmap_set(struct super_block *sb,
256 struct ext4_group_desc *bg, ext4_fsblk_t blk)
258 bg->bg_inode_bitmap_lo = cpu_to_le32((u32)blk);
259 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
260 bg->bg_inode_bitmap_hi = cpu_to_le32(blk >> 32);
263 void ext4_inode_table_set(struct super_block *sb,
264 struct ext4_group_desc *bg, ext4_fsblk_t blk)
266 bg->bg_inode_table_lo = cpu_to_le32((u32)blk);
267 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
268 bg->bg_inode_table_hi = cpu_to_le32(blk >> 32);
271 void ext4_free_group_clusters_set(struct super_block *sb,
272 struct ext4_group_desc *bg, __u32 count)
274 bg->bg_free_blocks_count_lo = cpu_to_le16((__u16)count);
275 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
276 bg->bg_free_blocks_count_hi = cpu_to_le16(count >> 16);
279 void ext4_free_inodes_set(struct super_block *sb,
280 struct ext4_group_desc *bg, __u32 count)
282 bg->bg_free_inodes_count_lo = cpu_to_le16((__u16)count);
283 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
284 bg->bg_free_inodes_count_hi = cpu_to_le16(count >> 16);
287 void ext4_used_dirs_set(struct super_block *sb,
288 struct ext4_group_desc *bg, __u32 count)
290 bg->bg_used_dirs_count_lo = cpu_to_le16((__u16)count);
291 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
292 bg->bg_used_dirs_count_hi = cpu_to_le16(count >> 16);
295 void ext4_itable_unused_set(struct super_block *sb,
296 struct ext4_group_desc *bg, __u32 count)
298 bg->bg_itable_unused_lo = cpu_to_le16((__u16)count);
299 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
300 bg->bg_itable_unused_hi = cpu_to_le16(count >> 16);
304 static void __save_error_info(struct super_block *sb, const char *func,
307 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
309 EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS;
310 es->s_state |= cpu_to_le16(EXT4_ERROR_FS);
311 es->s_last_error_time = cpu_to_le32(get_seconds());
312 strncpy(es->s_last_error_func, func, sizeof(es->s_last_error_func));
313 es->s_last_error_line = cpu_to_le32(line);
314 if (!es->s_first_error_time) {
315 es->s_first_error_time = es->s_last_error_time;
316 strncpy(es->s_first_error_func, func,
317 sizeof(es->s_first_error_func));
318 es->s_first_error_line = cpu_to_le32(line);
319 es->s_first_error_ino = es->s_last_error_ino;
320 es->s_first_error_block = es->s_last_error_block;
323 * Start the daily error reporting function if it hasn't been
326 if (!es->s_error_count)
327 mod_timer(&EXT4_SB(sb)->s_err_report, jiffies + 24*60*60*HZ);
328 le32_add_cpu(&es->s_error_count, 1);
331 static void save_error_info(struct super_block *sb, const char *func,
334 __save_error_info(sb, func, line);
335 ext4_commit_super(sb, 1);
339 * The del_gendisk() function uninitializes the disk-specific data
340 * structures, including the bdi structure, without telling anyone
341 * else. Once this happens, any attempt to call mark_buffer_dirty()
342 * (for example, by ext4_commit_super), will cause a kernel OOPS.
343 * This is a kludge to prevent these oops until we can put in a proper
344 * hook in del_gendisk() to inform the VFS and file system layers.
346 static int block_device_ejected(struct super_block *sb)
348 struct inode *bd_inode = sb->s_bdev->bd_inode;
349 struct backing_dev_info *bdi = bd_inode->i_mapping->backing_dev_info;
351 return bdi->dev == NULL;
354 static void ext4_journal_commit_callback(journal_t *journal, transaction_t *txn)
356 struct super_block *sb = journal->j_private;
357 struct ext4_sb_info *sbi = EXT4_SB(sb);
358 int error = is_journal_aborted(journal);
359 struct ext4_journal_cb_entry *jce;
361 BUG_ON(txn->t_state == T_FINISHED);
362 spin_lock(&sbi->s_md_lock);
363 while (!list_empty(&txn->t_private_list)) {
364 jce = list_entry(txn->t_private_list.next,
365 struct ext4_journal_cb_entry, jce_list);
366 list_del_init(&jce->jce_list);
367 spin_unlock(&sbi->s_md_lock);
368 jce->jce_func(sb, jce, error);
369 spin_lock(&sbi->s_md_lock);
371 spin_unlock(&sbi->s_md_lock);
374 /* Deal with the reporting of failure conditions on a filesystem such as
375 * inconsistencies detected or read IO failures.
377 * On ext2, we can store the error state of the filesystem in the
378 * superblock. That is not possible on ext4, because we may have other
379 * write ordering constraints on the superblock which prevent us from
380 * writing it out straight away; and given that the journal is about to
381 * be aborted, we can't rely on the current, or future, transactions to
382 * write out the superblock safely.
384 * We'll just use the jbd2_journal_abort() error code to record an error in
385 * the journal instead. On recovery, the journal will complain about
386 * that error until we've noted it down and cleared it.
389 static void ext4_handle_error(struct super_block *sb)
391 if (sb->s_flags & MS_RDONLY)
394 if (!test_opt(sb, ERRORS_CONT)) {
395 journal_t *journal = EXT4_SB(sb)->s_journal;
397 EXT4_SB(sb)->s_mount_flags |= EXT4_MF_FS_ABORTED;
399 jbd2_journal_abort(journal, -EIO);
401 if (test_opt(sb, ERRORS_RO)) {
402 ext4_msg(sb, KERN_CRIT, "Remounting filesystem read-only");
404 * Make sure updated value of ->s_mount_flags will be visible
405 * before ->s_flags update
408 sb->s_flags |= MS_RDONLY;
410 if (test_opt(sb, ERRORS_PANIC))
411 panic("EXT4-fs (device %s): panic forced after error\n",
415 #define ext4_error_ratelimit(sb) \
416 ___ratelimit(&(EXT4_SB(sb)->s_err_ratelimit_state), \
419 void __ext4_error(struct super_block *sb, const char *function,
420 unsigned int line, const char *fmt, ...)
422 struct va_format vaf;
425 if (ext4_error_ratelimit(sb)) {
430 "EXT4-fs error (device %s): %s:%d: comm %s: %pV\n",
431 sb->s_id, function, line, current->comm, &vaf);
434 save_error_info(sb, function, line);
435 ext4_handle_error(sb);
438 void __ext4_error_inode(struct inode *inode, const char *function,
439 unsigned int line, ext4_fsblk_t block,
440 const char *fmt, ...)
443 struct va_format vaf;
444 struct ext4_super_block *es = EXT4_SB(inode->i_sb)->s_es;
446 es->s_last_error_ino = cpu_to_le32(inode->i_ino);
447 es->s_last_error_block = cpu_to_le64(block);
448 if (ext4_error_ratelimit(inode->i_sb)) {
453 printk(KERN_CRIT "EXT4-fs error (device %s): %s:%d: "
454 "inode #%lu: block %llu: comm %s: %pV\n",
455 inode->i_sb->s_id, function, line, inode->i_ino,
456 block, current->comm, &vaf);
458 printk(KERN_CRIT "EXT4-fs error (device %s): %s:%d: "
459 "inode #%lu: comm %s: %pV\n",
460 inode->i_sb->s_id, function, line, inode->i_ino,
461 current->comm, &vaf);
464 save_error_info(inode->i_sb, function, line);
465 ext4_handle_error(inode->i_sb);
468 void __ext4_error_file(struct file *file, const char *function,
469 unsigned int line, ext4_fsblk_t block,
470 const char *fmt, ...)
473 struct va_format vaf;
474 struct ext4_super_block *es;
475 struct inode *inode = file_inode(file);
476 char pathname[80], *path;
478 es = EXT4_SB(inode->i_sb)->s_es;
479 es->s_last_error_ino = cpu_to_le32(inode->i_ino);
480 if (ext4_error_ratelimit(inode->i_sb)) {
481 path = d_path(&(file->f_path), pathname, sizeof(pathname));
489 "EXT4-fs error (device %s): %s:%d: inode #%lu: "
490 "block %llu: comm %s: path %s: %pV\n",
491 inode->i_sb->s_id, function, line, inode->i_ino,
492 block, current->comm, path, &vaf);
495 "EXT4-fs error (device %s): %s:%d: inode #%lu: "
496 "comm %s: path %s: %pV\n",
497 inode->i_sb->s_id, function, line, inode->i_ino,
498 current->comm, path, &vaf);
501 save_error_info(inode->i_sb, function, line);
502 ext4_handle_error(inode->i_sb);
505 const char *ext4_decode_error(struct super_block *sb, int errno,
512 errstr = "IO failure";
515 errstr = "Out of memory";
518 if (!sb || (EXT4_SB(sb)->s_journal &&
519 EXT4_SB(sb)->s_journal->j_flags & JBD2_ABORT))
520 errstr = "Journal has aborted";
522 errstr = "Readonly filesystem";
525 /* If the caller passed in an extra buffer for unknown
526 * errors, textualise them now. Else we just return
529 /* Check for truncated error codes... */
530 if (snprintf(nbuf, 16, "error %d", -errno) >= 0)
539 /* __ext4_std_error decodes expected errors from journaling functions
540 * automatically and invokes the appropriate error response. */
542 void __ext4_std_error(struct super_block *sb, const char *function,
543 unsigned int line, int errno)
548 /* Special case: if the error is EROFS, and we're not already
549 * inside a transaction, then there's really no point in logging
551 if (errno == -EROFS && journal_current_handle() == NULL &&
552 (sb->s_flags & MS_RDONLY))
555 if (ext4_error_ratelimit(sb)) {
556 errstr = ext4_decode_error(sb, errno, nbuf);
557 printk(KERN_CRIT "EXT4-fs error (device %s) in %s:%d: %s\n",
558 sb->s_id, function, line, errstr);
561 save_error_info(sb, function, line);
562 ext4_handle_error(sb);
566 * ext4_abort is a much stronger failure handler than ext4_error. The
567 * abort function may be used to deal with unrecoverable failures such
568 * as journal IO errors or ENOMEM at a critical moment in log management.
570 * We unconditionally force the filesystem into an ABORT|READONLY state,
571 * unless the error response on the fs has been set to panic in which
572 * case we take the easy way out and panic immediately.
575 void __ext4_abort(struct super_block *sb, const char *function,
576 unsigned int line, const char *fmt, ...)
580 save_error_info(sb, function, line);
582 printk(KERN_CRIT "EXT4-fs error (device %s): %s:%d: ", sb->s_id,
588 if ((sb->s_flags & MS_RDONLY) == 0) {
589 ext4_msg(sb, KERN_CRIT, "Remounting filesystem read-only");
590 EXT4_SB(sb)->s_mount_flags |= EXT4_MF_FS_ABORTED;
592 * Make sure updated value of ->s_mount_flags will be visible
593 * before ->s_flags update
596 sb->s_flags |= MS_RDONLY;
597 if (EXT4_SB(sb)->s_journal)
598 jbd2_journal_abort(EXT4_SB(sb)->s_journal, -EIO);
599 save_error_info(sb, function, line);
601 if (test_opt(sb, ERRORS_PANIC))
602 panic("EXT4-fs panic from previous error\n");
605 void __ext4_msg(struct super_block *sb,
606 const char *prefix, const char *fmt, ...)
608 struct va_format vaf;
611 if (!___ratelimit(&(EXT4_SB(sb)->s_msg_ratelimit_state), "EXT4-fs"))
617 printk("%sEXT4-fs (%s): %pV\n", prefix, sb->s_id, &vaf);
621 void __ext4_warning(struct super_block *sb, const char *function,
622 unsigned int line, const char *fmt, ...)
624 struct va_format vaf;
627 if (!___ratelimit(&(EXT4_SB(sb)->s_warning_ratelimit_state),
634 printk(KERN_WARNING "EXT4-fs warning (device %s): %s:%d: %pV\n",
635 sb->s_id, function, line, &vaf);
639 void __ext4_grp_locked_error(const char *function, unsigned int line,
640 struct super_block *sb, ext4_group_t grp,
641 unsigned long ino, ext4_fsblk_t block,
642 const char *fmt, ...)
646 struct va_format vaf;
648 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
650 es->s_last_error_ino = cpu_to_le32(ino);
651 es->s_last_error_block = cpu_to_le64(block);
652 __save_error_info(sb, function, line);
654 if (ext4_error_ratelimit(sb)) {
658 printk(KERN_CRIT "EXT4-fs error (device %s): %s:%d: group %u, ",
659 sb->s_id, function, line, grp);
661 printk(KERN_CONT "inode %lu: ", ino);
663 printk(KERN_CONT "block %llu:",
664 (unsigned long long) block);
665 printk(KERN_CONT "%pV\n", &vaf);
669 if (test_opt(sb, ERRORS_CONT)) {
670 ext4_commit_super(sb, 0);
674 ext4_unlock_group(sb, grp);
675 ext4_handle_error(sb);
677 * We only get here in the ERRORS_RO case; relocking the group
678 * may be dangerous, but nothing bad will happen since the
679 * filesystem will have already been marked read/only and the
680 * journal has been aborted. We return 1 as a hint to callers
681 * who might what to use the return value from
682 * ext4_grp_locked_error() to distinguish between the
683 * ERRORS_CONT and ERRORS_RO case, and perhaps return more
684 * aggressively from the ext4 function in question, with a
685 * more appropriate error code.
687 ext4_lock_group(sb, grp);
691 void ext4_update_dynamic_rev(struct super_block *sb)
693 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
695 if (le32_to_cpu(es->s_rev_level) > EXT4_GOOD_OLD_REV)
699 "updating to rev %d because of new feature flag, "
700 "running e2fsck is recommended",
703 es->s_first_ino = cpu_to_le32(EXT4_GOOD_OLD_FIRST_INO);
704 es->s_inode_size = cpu_to_le16(EXT4_GOOD_OLD_INODE_SIZE);
705 es->s_rev_level = cpu_to_le32(EXT4_DYNAMIC_REV);
706 /* leave es->s_feature_*compat flags alone */
707 /* es->s_uuid will be set by e2fsck if empty */
710 * The rest of the superblock fields should be zero, and if not it
711 * means they are likely already in use, so leave them alone. We
712 * can leave it up to e2fsck to clean up any inconsistencies there.
717 * Open the external journal device
719 static struct block_device *ext4_blkdev_get(dev_t dev, struct super_block *sb)
721 struct block_device *bdev;
722 char b[BDEVNAME_SIZE];
724 bdev = blkdev_get_by_dev(dev, FMODE_READ|FMODE_WRITE|FMODE_EXCL, sb);
730 ext4_msg(sb, KERN_ERR, "failed to open journal device %s: %ld",
731 __bdevname(dev, b), PTR_ERR(bdev));
736 * Release the journal device
738 static void ext4_blkdev_put(struct block_device *bdev)
740 blkdev_put(bdev, FMODE_READ|FMODE_WRITE|FMODE_EXCL);
743 static void ext4_blkdev_remove(struct ext4_sb_info *sbi)
745 struct block_device *bdev;
746 bdev = sbi->journal_bdev;
748 ext4_blkdev_put(bdev);
749 sbi->journal_bdev = NULL;
753 static inline struct inode *orphan_list_entry(struct list_head *l)
755 return &list_entry(l, struct ext4_inode_info, i_orphan)->vfs_inode;
758 static void dump_orphan_list(struct super_block *sb, struct ext4_sb_info *sbi)
762 ext4_msg(sb, KERN_ERR, "sb orphan head is %d",
763 le32_to_cpu(sbi->s_es->s_last_orphan));
765 printk(KERN_ERR "sb_info orphan list:\n");
766 list_for_each(l, &sbi->s_orphan) {
767 struct inode *inode = orphan_list_entry(l);
769 "inode %s:%lu at %p: mode %o, nlink %d, next %d\n",
770 inode->i_sb->s_id, inode->i_ino, inode,
771 inode->i_mode, inode->i_nlink,
776 static void ext4_put_super(struct super_block *sb)
778 struct ext4_sb_info *sbi = EXT4_SB(sb);
779 struct ext4_super_block *es = sbi->s_es;
782 ext4_unregister_li_request(sb);
783 dquot_disable(sb, -1, DQUOT_USAGE_ENABLED | DQUOT_LIMITS_ENABLED);
785 flush_workqueue(sbi->rsv_conversion_wq);
786 destroy_workqueue(sbi->rsv_conversion_wq);
788 if (sbi->s_journal) {
789 err = jbd2_journal_destroy(sbi->s_journal);
790 sbi->s_journal = NULL;
792 ext4_abort(sb, "Couldn't clean up the journal");
795 ext4_es_unregister_shrinker(sbi);
796 del_timer_sync(&sbi->s_err_report);
797 ext4_release_system_zone(sb);
799 ext4_ext_release(sb);
800 ext4_xattr_put_super(sb);
802 if (!(sb->s_flags & MS_RDONLY)) {
803 EXT4_CLEAR_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
804 es->s_state = cpu_to_le16(sbi->s_mount_state);
806 if (!(sb->s_flags & MS_RDONLY))
807 ext4_commit_super(sb, 1);
810 remove_proc_entry("options", sbi->s_proc);
811 remove_proc_entry(sb->s_id, ext4_proc_root);
813 kobject_del(&sbi->s_kobj);
815 for (i = 0; i < sbi->s_gdb_count; i++)
816 brelse(sbi->s_group_desc[i]);
817 ext4_kvfree(sbi->s_group_desc);
818 ext4_kvfree(sbi->s_flex_groups);
819 percpu_counter_destroy(&sbi->s_freeclusters_counter);
820 percpu_counter_destroy(&sbi->s_freeinodes_counter);
821 percpu_counter_destroy(&sbi->s_dirs_counter);
822 percpu_counter_destroy(&sbi->s_dirtyclusters_counter);
825 for (i = 0; i < EXT4_MAXQUOTAS; i++)
826 kfree(sbi->s_qf_names[i]);
829 /* Debugging code just in case the in-memory inode orphan list
830 * isn't empty. The on-disk one can be non-empty if we've
831 * detected an error and taken the fs readonly, but the
832 * in-memory list had better be clean by this point. */
833 if (!list_empty(&sbi->s_orphan))
834 dump_orphan_list(sb, sbi);
835 J_ASSERT(list_empty(&sbi->s_orphan));
837 invalidate_bdev(sb->s_bdev);
838 if (sbi->journal_bdev && sbi->journal_bdev != sb->s_bdev) {
840 * Invalidate the journal device's buffers. We don't want them
841 * floating about in memory - the physical journal device may
842 * hotswapped, and it breaks the `ro-after' testing code.
844 sync_blockdev(sbi->journal_bdev);
845 invalidate_bdev(sbi->journal_bdev);
846 ext4_blkdev_remove(sbi);
848 if (sbi->s_mb_cache) {
849 ext4_xattr_destroy_cache(sbi->s_mb_cache);
850 sbi->s_mb_cache = NULL;
853 kthread_stop(sbi->s_mmp_tsk);
854 sb->s_fs_info = NULL;
856 * Now that we are completely done shutting down the
857 * superblock, we need to actually destroy the kobject.
859 kobject_put(&sbi->s_kobj);
860 wait_for_completion(&sbi->s_kobj_unregister);
861 if (sbi->s_chksum_driver)
862 crypto_free_shash(sbi->s_chksum_driver);
863 kfree(sbi->s_blockgroup_lock);
867 static struct kmem_cache *ext4_inode_cachep;
870 * Called inside transaction, so use GFP_NOFS
872 static struct inode *ext4_alloc_inode(struct super_block *sb)
874 struct ext4_inode_info *ei;
876 ei = kmem_cache_alloc(ext4_inode_cachep, GFP_NOFS);
880 ei->vfs_inode.i_version = 1;
881 spin_lock_init(&ei->i_raw_lock);
882 INIT_LIST_HEAD(&ei->i_prealloc_list);
883 spin_lock_init(&ei->i_prealloc_lock);
884 ext4_es_init_tree(&ei->i_es_tree);
885 rwlock_init(&ei->i_es_lock);
886 INIT_LIST_HEAD(&ei->i_es_lru);
889 ei->i_touch_when = 0;
890 ei->i_reserved_data_blocks = 0;
891 ei->i_reserved_meta_blocks = 0;
892 ei->i_allocated_meta_blocks = 0;
893 ei->i_da_metadata_calc_len = 0;
894 ei->i_da_metadata_calc_last_lblock = 0;
895 spin_lock_init(&(ei->i_block_reservation_lock));
897 ei->i_reserved_quota = 0;
900 INIT_LIST_HEAD(&ei->i_rsv_conversion_list);
901 spin_lock_init(&ei->i_completed_io_lock);
903 ei->i_datasync_tid = 0;
904 atomic_set(&ei->i_ioend_count, 0);
905 atomic_set(&ei->i_unwritten, 0);
906 INIT_WORK(&ei->i_rsv_conversion_work, ext4_end_io_rsv_work);
908 return &ei->vfs_inode;
911 static int ext4_drop_inode(struct inode *inode)
913 int drop = generic_drop_inode(inode);
915 trace_ext4_drop_inode(inode, drop);
919 static void ext4_i_callback(struct rcu_head *head)
921 struct inode *inode = container_of(head, struct inode, i_rcu);
922 kmem_cache_free(ext4_inode_cachep, EXT4_I(inode));
925 static void ext4_destroy_inode(struct inode *inode)
927 if (!list_empty(&(EXT4_I(inode)->i_orphan))) {
928 ext4_msg(inode->i_sb, KERN_ERR,
929 "Inode %lu (%p): orphan list check failed!",
930 inode->i_ino, EXT4_I(inode));
931 print_hex_dump(KERN_INFO, "", DUMP_PREFIX_ADDRESS, 16, 4,
932 EXT4_I(inode), sizeof(struct ext4_inode_info),
936 call_rcu(&inode->i_rcu, ext4_i_callback);
939 static void init_once(void *foo)
941 struct ext4_inode_info *ei = (struct ext4_inode_info *) foo;
943 INIT_LIST_HEAD(&ei->i_orphan);
944 init_rwsem(&ei->xattr_sem);
945 init_rwsem(&ei->i_data_sem);
946 inode_init_once(&ei->vfs_inode);
949 static int __init init_inodecache(void)
951 ext4_inode_cachep = kmem_cache_create("ext4_inode_cache",
952 sizeof(struct ext4_inode_info),
953 0, (SLAB_RECLAIM_ACCOUNT|
956 if (ext4_inode_cachep == NULL)
961 static void destroy_inodecache(void)
964 * Make sure all delayed rcu free inodes are flushed before we
968 kmem_cache_destroy(ext4_inode_cachep);
971 void ext4_clear_inode(struct inode *inode)
973 invalidate_inode_buffers(inode);
976 ext4_discard_preallocations(inode);
977 ext4_es_remove_extent(inode, 0, EXT_MAX_BLOCKS);
978 ext4_es_lru_del(inode);
979 if (EXT4_I(inode)->jinode) {
980 jbd2_journal_release_jbd_inode(EXT4_JOURNAL(inode),
981 EXT4_I(inode)->jinode);
982 jbd2_free_inode(EXT4_I(inode)->jinode);
983 EXT4_I(inode)->jinode = NULL;
987 static struct inode *ext4_nfs_get_inode(struct super_block *sb,
988 u64 ino, u32 generation)
992 if (ino < EXT4_FIRST_INO(sb) && ino != EXT4_ROOT_INO)
993 return ERR_PTR(-ESTALE);
994 if (ino > le32_to_cpu(EXT4_SB(sb)->s_es->s_inodes_count))
995 return ERR_PTR(-ESTALE);
997 /* iget isn't really right if the inode is currently unallocated!!
999 * ext4_read_inode will return a bad_inode if the inode had been
1000 * deleted, so we should be safe.
1002 * Currently we don't know the generation for parent directory, so
1003 * a generation of 0 means "accept any"
1005 inode = ext4_iget(sb, ino);
1007 return ERR_CAST(inode);
1008 if (generation && inode->i_generation != generation) {
1010 return ERR_PTR(-ESTALE);
1016 static struct dentry *ext4_fh_to_dentry(struct super_block *sb, struct fid *fid,
1017 int fh_len, int fh_type)
1019 return generic_fh_to_dentry(sb, fid, fh_len, fh_type,
1020 ext4_nfs_get_inode);
1023 static struct dentry *ext4_fh_to_parent(struct super_block *sb, struct fid *fid,
1024 int fh_len, int fh_type)
1026 return generic_fh_to_parent(sb, fid, fh_len, fh_type,
1027 ext4_nfs_get_inode);
1031 * Try to release metadata pages (indirect blocks, directories) which are
1032 * mapped via the block device. Since these pages could have journal heads
1033 * which would prevent try_to_free_buffers() from freeing them, we must use
1034 * jbd2 layer's try_to_free_buffers() function to release them.
1036 static int bdev_try_to_free_page(struct super_block *sb, struct page *page,
1039 journal_t *journal = EXT4_SB(sb)->s_journal;
1041 WARN_ON(PageChecked(page));
1042 if (!page_has_buffers(page))
1045 return jbd2_journal_try_to_free_buffers(journal, page,
1046 wait & ~__GFP_WAIT);
1047 return try_to_free_buffers(page);
1051 #define QTYPE2NAME(t) ((t) == USRQUOTA ? "user" : "group")
1052 #define QTYPE2MOPT(on, t) ((t) == USRQUOTA?((on)##USRJQUOTA):((on)##GRPJQUOTA))
1054 static int ext4_write_dquot(struct dquot *dquot);
1055 static int ext4_acquire_dquot(struct dquot *dquot);
1056 static int ext4_release_dquot(struct dquot *dquot);
1057 static int ext4_mark_dquot_dirty(struct dquot *dquot);
1058 static int ext4_write_info(struct super_block *sb, int type);
1059 static int ext4_quota_on(struct super_block *sb, int type, int format_id,
1061 static int ext4_quota_on_sysfile(struct super_block *sb, int type,
1063 static int ext4_quota_off(struct super_block *sb, int type);
1064 static int ext4_quota_off_sysfile(struct super_block *sb, int type);
1065 static int ext4_quota_on_mount(struct super_block *sb, int type);
1066 static ssize_t ext4_quota_read(struct super_block *sb, int type, char *data,
1067 size_t len, loff_t off);
1068 static ssize_t ext4_quota_write(struct super_block *sb, int type,
1069 const char *data, size_t len, loff_t off);
1070 static int ext4_quota_enable(struct super_block *sb, int type, int format_id,
1071 unsigned int flags);
1072 static int ext4_enable_quotas(struct super_block *sb);
1074 static const struct dquot_operations ext4_quota_operations = {
1075 .get_reserved_space = ext4_get_reserved_space,
1076 .write_dquot = ext4_write_dquot,
1077 .acquire_dquot = ext4_acquire_dquot,
1078 .release_dquot = ext4_release_dquot,
1079 .mark_dirty = ext4_mark_dquot_dirty,
1080 .write_info = ext4_write_info,
1081 .alloc_dquot = dquot_alloc,
1082 .destroy_dquot = dquot_destroy,
1085 static const struct quotactl_ops ext4_qctl_operations = {
1086 .quota_on = ext4_quota_on,
1087 .quota_off = ext4_quota_off,
1088 .quota_sync = dquot_quota_sync,
1089 .get_info = dquot_get_dqinfo,
1090 .set_info = dquot_set_dqinfo,
1091 .get_dqblk = dquot_get_dqblk,
1092 .set_dqblk = dquot_set_dqblk
1095 static const struct quotactl_ops ext4_qctl_sysfile_operations = {
1096 .quota_on_meta = ext4_quota_on_sysfile,
1097 .quota_off = ext4_quota_off_sysfile,
1098 .quota_sync = dquot_quota_sync,
1099 .get_info = dquot_get_dqinfo,
1100 .set_info = dquot_set_dqinfo,
1101 .get_dqblk = dquot_get_dqblk,
1102 .set_dqblk = dquot_set_dqblk
1106 static const struct super_operations ext4_sops = {
1107 .alloc_inode = ext4_alloc_inode,
1108 .destroy_inode = ext4_destroy_inode,
1109 .write_inode = ext4_write_inode,
1110 .dirty_inode = ext4_dirty_inode,
1111 .drop_inode = ext4_drop_inode,
1112 .evict_inode = ext4_evict_inode,
1113 .put_super = ext4_put_super,
1114 .sync_fs = ext4_sync_fs,
1115 .freeze_fs = ext4_freeze,
1116 .unfreeze_fs = ext4_unfreeze,
1117 .statfs = ext4_statfs,
1118 .remount_fs = ext4_remount,
1119 .show_options = ext4_show_options,
1121 .quota_read = ext4_quota_read,
1122 .quota_write = ext4_quota_write,
1124 .bdev_try_to_free_page = bdev_try_to_free_page,
1127 static const struct super_operations ext4_nojournal_sops = {
1128 .alloc_inode = ext4_alloc_inode,
1129 .destroy_inode = ext4_destroy_inode,
1130 .write_inode = ext4_write_inode,
1131 .dirty_inode = ext4_dirty_inode,
1132 .drop_inode = ext4_drop_inode,
1133 .evict_inode = ext4_evict_inode,
1134 .sync_fs = ext4_sync_fs_nojournal,
1135 .put_super = ext4_put_super,
1136 .statfs = ext4_statfs,
1137 .remount_fs = ext4_remount,
1138 .show_options = ext4_show_options,
1140 .quota_read = ext4_quota_read,
1141 .quota_write = ext4_quota_write,
1143 .bdev_try_to_free_page = bdev_try_to_free_page,
1146 static const struct export_operations ext4_export_ops = {
1147 .fh_to_dentry = ext4_fh_to_dentry,
1148 .fh_to_parent = ext4_fh_to_parent,
1149 .get_parent = ext4_get_parent,
1153 Opt_bsd_df, Opt_minix_df, Opt_grpid, Opt_nogrpid,
1154 Opt_resgid, Opt_resuid, Opt_sb, Opt_err_cont, Opt_err_panic, Opt_err_ro,
1155 Opt_nouid32, Opt_debug, Opt_removed,
1156 Opt_user_xattr, Opt_nouser_xattr, Opt_acl, Opt_noacl,
1157 Opt_auto_da_alloc, Opt_noauto_da_alloc, Opt_noload,
1158 Opt_commit, Opt_min_batch_time, Opt_max_batch_time, Opt_journal_dev,
1159 Opt_journal_path, Opt_journal_checksum, Opt_journal_async_commit,
1160 Opt_abort, Opt_data_journal, Opt_data_ordered, Opt_data_writeback,
1161 Opt_data_err_abort, Opt_data_err_ignore,
1162 Opt_usrjquota, Opt_grpjquota, Opt_offusrjquota, Opt_offgrpjquota,
1163 Opt_jqfmt_vfsold, Opt_jqfmt_vfsv0, Opt_jqfmt_vfsv1, Opt_quota,
1164 Opt_noquota, Opt_barrier, Opt_nobarrier, Opt_err,
1165 Opt_usrquota, Opt_grpquota, Opt_i_version,
1166 Opt_stripe, Opt_delalloc, Opt_nodelalloc, Opt_mblk_io_submit,
1167 Opt_nomblk_io_submit, Opt_block_validity, Opt_noblock_validity,
1168 Opt_inode_readahead_blks, Opt_journal_ioprio,
1169 Opt_dioread_nolock, Opt_dioread_lock,
1170 Opt_discard, Opt_nodiscard, Opt_init_itable, Opt_noinit_itable,
1171 Opt_max_dir_size_kb,
1174 static const match_table_t tokens = {
1175 {Opt_bsd_df, "bsddf"},
1176 {Opt_minix_df, "minixdf"},
1177 {Opt_grpid, "grpid"},
1178 {Opt_grpid, "bsdgroups"},
1179 {Opt_nogrpid, "nogrpid"},
1180 {Opt_nogrpid, "sysvgroups"},
1181 {Opt_resgid, "resgid=%u"},
1182 {Opt_resuid, "resuid=%u"},
1184 {Opt_err_cont, "errors=continue"},
1185 {Opt_err_panic, "errors=panic"},
1186 {Opt_err_ro, "errors=remount-ro"},
1187 {Opt_nouid32, "nouid32"},
1188 {Opt_debug, "debug"},
1189 {Opt_removed, "oldalloc"},
1190 {Opt_removed, "orlov"},
1191 {Opt_user_xattr, "user_xattr"},
1192 {Opt_nouser_xattr, "nouser_xattr"},
1194 {Opt_noacl, "noacl"},
1195 {Opt_noload, "norecovery"},
1196 {Opt_noload, "noload"},
1197 {Opt_removed, "nobh"},
1198 {Opt_removed, "bh"},
1199 {Opt_commit, "commit=%u"},
1200 {Opt_min_batch_time, "min_batch_time=%u"},
1201 {Opt_max_batch_time, "max_batch_time=%u"},
1202 {Opt_journal_dev, "journal_dev=%u"},
1203 {Opt_journal_path, "journal_path=%s"},
1204 {Opt_journal_checksum, "journal_checksum"},
1205 {Opt_journal_async_commit, "journal_async_commit"},
1206 {Opt_abort, "abort"},
1207 {Opt_data_journal, "data=journal"},
1208 {Opt_data_ordered, "data=ordered"},
1209 {Opt_data_writeback, "data=writeback"},
1210 {Opt_data_err_abort, "data_err=abort"},
1211 {Opt_data_err_ignore, "data_err=ignore"},
1212 {Opt_offusrjquota, "usrjquota="},
1213 {Opt_usrjquota, "usrjquota=%s"},
1214 {Opt_offgrpjquota, "grpjquota="},
1215 {Opt_grpjquota, "grpjquota=%s"},
1216 {Opt_jqfmt_vfsold, "jqfmt=vfsold"},
1217 {Opt_jqfmt_vfsv0, "jqfmt=vfsv0"},
1218 {Opt_jqfmt_vfsv1, "jqfmt=vfsv1"},
1219 {Opt_grpquota, "grpquota"},
1220 {Opt_noquota, "noquota"},
1221 {Opt_quota, "quota"},
1222 {Opt_usrquota, "usrquota"},
1223 {Opt_barrier, "barrier=%u"},
1224 {Opt_barrier, "barrier"},
1225 {Opt_nobarrier, "nobarrier"},
1226 {Opt_i_version, "i_version"},
1227 {Opt_stripe, "stripe=%u"},
1228 {Opt_delalloc, "delalloc"},
1229 {Opt_nodelalloc, "nodelalloc"},
1230 {Opt_removed, "mblk_io_submit"},
1231 {Opt_removed, "nomblk_io_submit"},
1232 {Opt_block_validity, "block_validity"},
1233 {Opt_noblock_validity, "noblock_validity"},
1234 {Opt_inode_readahead_blks, "inode_readahead_blks=%u"},
1235 {Opt_journal_ioprio, "journal_ioprio=%u"},
1236 {Opt_auto_da_alloc, "auto_da_alloc=%u"},
1237 {Opt_auto_da_alloc, "auto_da_alloc"},
1238 {Opt_noauto_da_alloc, "noauto_da_alloc"},
1239 {Opt_dioread_nolock, "dioread_nolock"},
1240 {Opt_dioread_lock, "dioread_lock"},
1241 {Opt_discard, "discard"},
1242 {Opt_nodiscard, "nodiscard"},
1243 {Opt_init_itable, "init_itable=%u"},
1244 {Opt_init_itable, "init_itable"},
1245 {Opt_noinit_itable, "noinit_itable"},
1246 {Opt_max_dir_size_kb, "max_dir_size_kb=%u"},
1247 {Opt_removed, "check=none"}, /* mount option from ext2/3 */
1248 {Opt_removed, "nocheck"}, /* mount option from ext2/3 */
1249 {Opt_removed, "reservation"}, /* mount option from ext2/3 */
1250 {Opt_removed, "noreservation"}, /* mount option from ext2/3 */
1251 {Opt_removed, "journal=%u"}, /* mount option from ext2/3 */
1255 static ext4_fsblk_t get_sb_block(void **data)
1257 ext4_fsblk_t sb_block;
1258 char *options = (char *) *data;
1260 if (!options || strncmp(options, "sb=", 3) != 0)
1261 return 1; /* Default location */
1264 /* TODO: use simple_strtoll with >32bit ext4 */
1265 sb_block = simple_strtoul(options, &options, 0);
1266 if (*options && *options != ',') {
1267 printk(KERN_ERR "EXT4-fs: Invalid sb specification: %s\n",
1271 if (*options == ',')
1273 *data = (void *) options;
1278 #define DEFAULT_JOURNAL_IOPRIO (IOPRIO_PRIO_VALUE(IOPRIO_CLASS_BE, 3))
1279 static char deprecated_msg[] = "Mount option \"%s\" will be removed by %s\n"
1280 "Contact linux-ext4@vger.kernel.org if you think we should keep it.\n";
1283 static int set_qf_name(struct super_block *sb, int qtype, substring_t *args)
1285 struct ext4_sb_info *sbi = EXT4_SB(sb);
1289 if (sb_any_quota_loaded(sb) &&
1290 !sbi->s_qf_names[qtype]) {
1291 ext4_msg(sb, KERN_ERR,
1292 "Cannot change journaled "
1293 "quota options when quota turned on");
1296 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_QUOTA)) {
1297 ext4_msg(sb, KERN_ERR, "Cannot set journaled quota options "
1298 "when QUOTA feature is enabled");
1301 qname = match_strdup(args);
1303 ext4_msg(sb, KERN_ERR,
1304 "Not enough memory for storing quotafile name");
1307 if (sbi->s_qf_names[qtype]) {
1308 if (strcmp(sbi->s_qf_names[qtype], qname) == 0)
1311 ext4_msg(sb, KERN_ERR,
1312 "%s quota file already specified",
1316 if (strchr(qname, '/')) {
1317 ext4_msg(sb, KERN_ERR,
1318 "quotafile must be on filesystem root");
1321 sbi->s_qf_names[qtype] = qname;
1329 static int clear_qf_name(struct super_block *sb, int qtype)
1332 struct ext4_sb_info *sbi = EXT4_SB(sb);
1334 if (sb_any_quota_loaded(sb) &&
1335 sbi->s_qf_names[qtype]) {
1336 ext4_msg(sb, KERN_ERR, "Cannot change journaled quota options"
1337 " when quota turned on");
1340 kfree(sbi->s_qf_names[qtype]);
1341 sbi->s_qf_names[qtype] = NULL;
1346 #define MOPT_SET 0x0001
1347 #define MOPT_CLEAR 0x0002
1348 #define MOPT_NOSUPPORT 0x0004
1349 #define MOPT_EXPLICIT 0x0008
1350 #define MOPT_CLEAR_ERR 0x0010
1351 #define MOPT_GTE0 0x0020
1354 #define MOPT_QFMT 0x0040
1356 #define MOPT_Q MOPT_NOSUPPORT
1357 #define MOPT_QFMT MOPT_NOSUPPORT
1359 #define MOPT_DATAJ 0x0080
1360 #define MOPT_NO_EXT2 0x0100
1361 #define MOPT_NO_EXT3 0x0200
1362 #define MOPT_EXT4_ONLY (MOPT_NO_EXT2 | MOPT_NO_EXT3)
1363 #define MOPT_STRING 0x0400
1365 static const struct mount_opts {
1369 } ext4_mount_opts[] = {
1370 {Opt_minix_df, EXT4_MOUNT_MINIX_DF, MOPT_SET},
1371 {Opt_bsd_df, EXT4_MOUNT_MINIX_DF, MOPT_CLEAR},
1372 {Opt_grpid, EXT4_MOUNT_GRPID, MOPT_SET},
1373 {Opt_nogrpid, EXT4_MOUNT_GRPID, MOPT_CLEAR},
1374 {Opt_block_validity, EXT4_MOUNT_BLOCK_VALIDITY, MOPT_SET},
1375 {Opt_noblock_validity, EXT4_MOUNT_BLOCK_VALIDITY, MOPT_CLEAR},
1376 {Opt_dioread_nolock, EXT4_MOUNT_DIOREAD_NOLOCK,
1377 MOPT_EXT4_ONLY | MOPT_SET},
1378 {Opt_dioread_lock, EXT4_MOUNT_DIOREAD_NOLOCK,
1379 MOPT_EXT4_ONLY | MOPT_CLEAR},
1380 {Opt_discard, EXT4_MOUNT_DISCARD, MOPT_SET},
1381 {Opt_nodiscard, EXT4_MOUNT_DISCARD, MOPT_CLEAR},
1382 {Opt_delalloc, EXT4_MOUNT_DELALLOC,
1383 MOPT_EXT4_ONLY | MOPT_SET | MOPT_EXPLICIT},
1384 {Opt_nodelalloc, EXT4_MOUNT_DELALLOC,
1385 MOPT_EXT4_ONLY | MOPT_CLEAR},
1386 {Opt_journal_checksum, EXT4_MOUNT_JOURNAL_CHECKSUM,
1387 MOPT_EXT4_ONLY | MOPT_SET},
1388 {Opt_journal_async_commit, (EXT4_MOUNT_JOURNAL_ASYNC_COMMIT |
1389 EXT4_MOUNT_JOURNAL_CHECKSUM),
1390 MOPT_EXT4_ONLY | MOPT_SET},
1391 {Opt_noload, EXT4_MOUNT_NOLOAD, MOPT_NO_EXT2 | MOPT_SET},
1392 {Opt_err_panic, EXT4_MOUNT_ERRORS_PANIC, MOPT_SET | MOPT_CLEAR_ERR},
1393 {Opt_err_ro, EXT4_MOUNT_ERRORS_RO, MOPT_SET | MOPT_CLEAR_ERR},
1394 {Opt_err_cont, EXT4_MOUNT_ERRORS_CONT, MOPT_SET | MOPT_CLEAR_ERR},
1395 {Opt_data_err_abort, EXT4_MOUNT_DATA_ERR_ABORT,
1396 MOPT_NO_EXT2 | MOPT_SET},
1397 {Opt_data_err_ignore, EXT4_MOUNT_DATA_ERR_ABORT,
1398 MOPT_NO_EXT2 | MOPT_CLEAR},
1399 {Opt_barrier, EXT4_MOUNT_BARRIER, MOPT_SET},
1400 {Opt_nobarrier, EXT4_MOUNT_BARRIER, MOPT_CLEAR},
1401 {Opt_noauto_da_alloc, EXT4_MOUNT_NO_AUTO_DA_ALLOC, MOPT_SET},
1402 {Opt_auto_da_alloc, EXT4_MOUNT_NO_AUTO_DA_ALLOC, MOPT_CLEAR},
1403 {Opt_noinit_itable, EXT4_MOUNT_INIT_INODE_TABLE, MOPT_CLEAR},
1404 {Opt_commit, 0, MOPT_GTE0},
1405 {Opt_max_batch_time, 0, MOPT_GTE0},
1406 {Opt_min_batch_time, 0, MOPT_GTE0},
1407 {Opt_inode_readahead_blks, 0, MOPT_GTE0},
1408 {Opt_init_itable, 0, MOPT_GTE0},
1409 {Opt_stripe, 0, MOPT_GTE0},
1410 {Opt_resuid, 0, MOPT_GTE0},
1411 {Opt_resgid, 0, MOPT_GTE0},
1412 {Opt_journal_dev, 0, MOPT_GTE0},
1413 {Opt_journal_path, 0, MOPT_STRING},
1414 {Opt_journal_ioprio, 0, MOPT_GTE0},
1415 {Opt_data_journal, EXT4_MOUNT_JOURNAL_DATA, MOPT_NO_EXT2 | MOPT_DATAJ},
1416 {Opt_data_ordered, EXT4_MOUNT_ORDERED_DATA, MOPT_NO_EXT2 | MOPT_DATAJ},
1417 {Opt_data_writeback, EXT4_MOUNT_WRITEBACK_DATA,
1418 MOPT_NO_EXT2 | MOPT_DATAJ},
1419 {Opt_user_xattr, EXT4_MOUNT_XATTR_USER, MOPT_SET},
1420 {Opt_nouser_xattr, EXT4_MOUNT_XATTR_USER, MOPT_CLEAR},
1421 #ifdef CONFIG_EXT4_FS_POSIX_ACL
1422 {Opt_acl, EXT4_MOUNT_POSIX_ACL, MOPT_SET},
1423 {Opt_noacl, EXT4_MOUNT_POSIX_ACL, MOPT_CLEAR},
1425 {Opt_acl, 0, MOPT_NOSUPPORT},
1426 {Opt_noacl, 0, MOPT_NOSUPPORT},
1428 {Opt_nouid32, EXT4_MOUNT_NO_UID32, MOPT_SET},
1429 {Opt_debug, EXT4_MOUNT_DEBUG, MOPT_SET},
1430 {Opt_quota, EXT4_MOUNT_QUOTA | EXT4_MOUNT_USRQUOTA, MOPT_SET | MOPT_Q},
1431 {Opt_usrquota, EXT4_MOUNT_QUOTA | EXT4_MOUNT_USRQUOTA,
1433 {Opt_grpquota, EXT4_MOUNT_QUOTA | EXT4_MOUNT_GRPQUOTA,
1435 {Opt_noquota, (EXT4_MOUNT_QUOTA | EXT4_MOUNT_USRQUOTA |
1436 EXT4_MOUNT_GRPQUOTA), MOPT_CLEAR | MOPT_Q},
1437 {Opt_usrjquota, 0, MOPT_Q},
1438 {Opt_grpjquota, 0, MOPT_Q},
1439 {Opt_offusrjquota, 0, MOPT_Q},
1440 {Opt_offgrpjquota, 0, MOPT_Q},
1441 {Opt_jqfmt_vfsold, QFMT_VFS_OLD, MOPT_QFMT},
1442 {Opt_jqfmt_vfsv0, QFMT_VFS_V0, MOPT_QFMT},
1443 {Opt_jqfmt_vfsv1, QFMT_VFS_V1, MOPT_QFMT},
1444 {Opt_max_dir_size_kb, 0, MOPT_GTE0},
1448 static int handle_mount_opt(struct super_block *sb, char *opt, int token,
1449 substring_t *args, unsigned long *journal_devnum,
1450 unsigned int *journal_ioprio, int is_remount)
1452 struct ext4_sb_info *sbi = EXT4_SB(sb);
1453 const struct mount_opts *m;
1459 if (token == Opt_usrjquota)
1460 return set_qf_name(sb, USRQUOTA, &args[0]);
1461 else if (token == Opt_grpjquota)
1462 return set_qf_name(sb, GRPQUOTA, &args[0]);
1463 else if (token == Opt_offusrjquota)
1464 return clear_qf_name(sb, USRQUOTA);
1465 else if (token == Opt_offgrpjquota)
1466 return clear_qf_name(sb, GRPQUOTA);
1470 case Opt_nouser_xattr:
1471 ext4_msg(sb, KERN_WARNING, deprecated_msg, opt, "3.5");
1474 return 1; /* handled by get_sb_block() */
1476 ext4_msg(sb, KERN_WARNING, "Ignoring removed %s option", opt);
1479 sbi->s_mount_flags |= EXT4_MF_FS_ABORTED;
1482 sb->s_flags |= MS_I_VERSION;
1486 for (m = ext4_mount_opts; m->token != Opt_err; m++)
1487 if (token == m->token)
1490 if (m->token == Opt_err) {
1491 ext4_msg(sb, KERN_ERR, "Unrecognized mount option \"%s\" "
1492 "or missing value", opt);
1496 if ((m->flags & MOPT_NO_EXT2) && IS_EXT2_SB(sb)) {
1497 ext4_msg(sb, KERN_ERR,
1498 "Mount option \"%s\" incompatible with ext2", opt);
1501 if ((m->flags & MOPT_NO_EXT3) && IS_EXT3_SB(sb)) {
1502 ext4_msg(sb, KERN_ERR,
1503 "Mount option \"%s\" incompatible with ext3", opt);
1507 if (args->from && !(m->flags & MOPT_STRING) && match_int(args, &arg))
1509 if (args->from && (m->flags & MOPT_GTE0) && (arg < 0))
1511 if (m->flags & MOPT_EXPLICIT)
1512 set_opt2(sb, EXPLICIT_DELALLOC);
1513 if (m->flags & MOPT_CLEAR_ERR)
1514 clear_opt(sb, ERRORS_MASK);
1515 if (token == Opt_noquota && sb_any_quota_loaded(sb)) {
1516 ext4_msg(sb, KERN_ERR, "Cannot change quota "
1517 "options when quota turned on");
1521 if (m->flags & MOPT_NOSUPPORT) {
1522 ext4_msg(sb, KERN_ERR, "%s option not supported", opt);
1523 } else if (token == Opt_commit) {
1525 arg = JBD2_DEFAULT_MAX_COMMIT_AGE;
1526 sbi->s_commit_interval = HZ * arg;
1527 } else if (token == Opt_max_batch_time) {
1528 sbi->s_max_batch_time = arg;
1529 } else if (token == Opt_min_batch_time) {
1530 sbi->s_min_batch_time = arg;
1531 } else if (token == Opt_inode_readahead_blks) {
1532 if (arg && (arg > (1 << 30) || !is_power_of_2(arg))) {
1533 ext4_msg(sb, KERN_ERR,
1534 "EXT4-fs: inode_readahead_blks must be "
1535 "0 or a power of 2 smaller than 2^31");
1538 sbi->s_inode_readahead_blks = arg;
1539 } else if (token == Opt_init_itable) {
1540 set_opt(sb, INIT_INODE_TABLE);
1542 arg = EXT4_DEF_LI_WAIT_MULT;
1543 sbi->s_li_wait_mult = arg;
1544 } else if (token == Opt_max_dir_size_kb) {
1545 sbi->s_max_dir_size_kb = arg;
1546 } else if (token == Opt_stripe) {
1547 sbi->s_stripe = arg;
1548 } else if (token == Opt_resuid) {
1549 uid = make_kuid(current_user_ns(), arg);
1550 if (!uid_valid(uid)) {
1551 ext4_msg(sb, KERN_ERR, "Invalid uid value %d", arg);
1554 sbi->s_resuid = uid;
1555 } else if (token == Opt_resgid) {
1556 gid = make_kgid(current_user_ns(), arg);
1557 if (!gid_valid(gid)) {
1558 ext4_msg(sb, KERN_ERR, "Invalid gid value %d", arg);
1561 sbi->s_resgid = gid;
1562 } else if (token == Opt_journal_dev) {
1564 ext4_msg(sb, KERN_ERR,
1565 "Cannot specify journal on remount");
1568 *journal_devnum = arg;
1569 } else if (token == Opt_journal_path) {
1571 struct inode *journal_inode;
1576 ext4_msg(sb, KERN_ERR,
1577 "Cannot specify journal on remount");
1580 journal_path = match_strdup(&args[0]);
1581 if (!journal_path) {
1582 ext4_msg(sb, KERN_ERR, "error: could not dup "
1583 "journal device string");
1587 error = kern_path(journal_path, LOOKUP_FOLLOW, &path);
1589 ext4_msg(sb, KERN_ERR, "error: could not find "
1590 "journal device path: error %d", error);
1591 kfree(journal_path);
1595 journal_inode = path.dentry->d_inode;
1596 if (!S_ISBLK(journal_inode->i_mode)) {
1597 ext4_msg(sb, KERN_ERR, "error: journal path %s "
1598 "is not a block device", journal_path);
1600 kfree(journal_path);
1604 *journal_devnum = new_encode_dev(journal_inode->i_rdev);
1606 kfree(journal_path);
1607 } else if (token == Opt_journal_ioprio) {
1609 ext4_msg(sb, KERN_ERR, "Invalid journal IO priority"
1614 IOPRIO_PRIO_VALUE(IOPRIO_CLASS_BE, arg);
1615 } else if (m->flags & MOPT_DATAJ) {
1617 if (!sbi->s_journal)
1618 ext4_msg(sb, KERN_WARNING, "Remounting file system with no journal so ignoring journalled data option");
1619 else if (test_opt(sb, DATA_FLAGS) != m->mount_opt) {
1620 ext4_msg(sb, KERN_ERR,
1621 "Cannot change data mode on remount");
1625 clear_opt(sb, DATA_FLAGS);
1626 sbi->s_mount_opt |= m->mount_opt;
1629 } else if (m->flags & MOPT_QFMT) {
1630 if (sb_any_quota_loaded(sb) &&
1631 sbi->s_jquota_fmt != m->mount_opt) {
1632 ext4_msg(sb, KERN_ERR, "Cannot change journaled "
1633 "quota options when quota turned on");
1636 if (EXT4_HAS_RO_COMPAT_FEATURE(sb,
1637 EXT4_FEATURE_RO_COMPAT_QUOTA)) {
1638 ext4_msg(sb, KERN_ERR,
1639 "Cannot set journaled quota options "
1640 "when QUOTA feature is enabled");
1643 sbi->s_jquota_fmt = m->mount_opt;
1648 if (m->flags & MOPT_CLEAR)
1650 else if (unlikely(!(m->flags & MOPT_SET))) {
1651 ext4_msg(sb, KERN_WARNING,
1652 "buggy handling of option %s", opt);
1657 sbi->s_mount_opt |= m->mount_opt;
1659 sbi->s_mount_opt &= ~m->mount_opt;
1664 static int parse_options(char *options, struct super_block *sb,
1665 unsigned long *journal_devnum,
1666 unsigned int *journal_ioprio,
1669 struct ext4_sb_info *sbi = EXT4_SB(sb);
1671 substring_t args[MAX_OPT_ARGS];
1677 while ((p = strsep(&options, ",")) != NULL) {
1681 * Initialize args struct so we know whether arg was
1682 * found; some options take optional arguments.
1684 args[0].to = args[0].from = NULL;
1685 token = match_token(p, tokens, args);
1686 if (handle_mount_opt(sb, p, token, args, journal_devnum,
1687 journal_ioprio, is_remount) < 0)
1691 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_QUOTA) &&
1692 (test_opt(sb, USRQUOTA) || test_opt(sb, GRPQUOTA))) {
1693 ext4_msg(sb, KERN_ERR, "Cannot set quota options when QUOTA "
1694 "feature is enabled");
1697 if (sbi->s_qf_names[USRQUOTA] || sbi->s_qf_names[GRPQUOTA]) {
1698 if (test_opt(sb, USRQUOTA) && sbi->s_qf_names[USRQUOTA])
1699 clear_opt(sb, USRQUOTA);
1701 if (test_opt(sb, GRPQUOTA) && sbi->s_qf_names[GRPQUOTA])
1702 clear_opt(sb, GRPQUOTA);
1704 if (test_opt(sb, GRPQUOTA) || test_opt(sb, USRQUOTA)) {
1705 ext4_msg(sb, KERN_ERR, "old and new quota "
1710 if (!sbi->s_jquota_fmt) {
1711 ext4_msg(sb, KERN_ERR, "journaled quota format "
1717 if (test_opt(sb, DIOREAD_NOLOCK)) {
1719 BLOCK_SIZE << le32_to_cpu(sbi->s_es->s_log_block_size);
1721 if (blocksize < PAGE_CACHE_SIZE) {
1722 ext4_msg(sb, KERN_ERR, "can't mount with "
1723 "dioread_nolock if block size != PAGE_SIZE");
1730 static inline void ext4_show_quota_options(struct seq_file *seq,
1731 struct super_block *sb)
1733 #if defined(CONFIG_QUOTA)
1734 struct ext4_sb_info *sbi = EXT4_SB(sb);
1736 if (sbi->s_jquota_fmt) {
1739 switch (sbi->s_jquota_fmt) {
1750 seq_printf(seq, ",jqfmt=%s", fmtname);
1753 if (sbi->s_qf_names[USRQUOTA])
1754 seq_printf(seq, ",usrjquota=%s", sbi->s_qf_names[USRQUOTA]);
1756 if (sbi->s_qf_names[GRPQUOTA])
1757 seq_printf(seq, ",grpjquota=%s", sbi->s_qf_names[GRPQUOTA]);
1761 static const char *token2str(int token)
1763 const struct match_token *t;
1765 for (t = tokens; t->token != Opt_err; t++)
1766 if (t->token == token && !strchr(t->pattern, '='))
1773 * - it's set to a non-default value OR
1774 * - if the per-sb default is different from the global default
1776 static int _ext4_show_options(struct seq_file *seq, struct super_block *sb,
1779 struct ext4_sb_info *sbi = EXT4_SB(sb);
1780 struct ext4_super_block *es = sbi->s_es;
1781 int def_errors, def_mount_opt = nodefs ? 0 : sbi->s_def_mount_opt;
1782 const struct mount_opts *m;
1783 char sep = nodefs ? '\n' : ',';
1785 #define SEQ_OPTS_PUTS(str) seq_printf(seq, "%c" str, sep)
1786 #define SEQ_OPTS_PRINT(str, arg) seq_printf(seq, "%c" str, sep, arg)
1788 if (sbi->s_sb_block != 1)
1789 SEQ_OPTS_PRINT("sb=%llu", sbi->s_sb_block);
1791 for (m = ext4_mount_opts; m->token != Opt_err; m++) {
1792 int want_set = m->flags & MOPT_SET;
1793 if (((m->flags & (MOPT_SET|MOPT_CLEAR)) == 0) ||
1794 (m->flags & MOPT_CLEAR_ERR))
1796 if (!(m->mount_opt & (sbi->s_mount_opt ^ def_mount_opt)))
1797 continue; /* skip if same as the default */
1799 (sbi->s_mount_opt & m->mount_opt) != m->mount_opt) ||
1800 (!want_set && (sbi->s_mount_opt & m->mount_opt)))
1801 continue; /* select Opt_noFoo vs Opt_Foo */
1802 SEQ_OPTS_PRINT("%s", token2str(m->token));
1805 if (nodefs || !uid_eq(sbi->s_resuid, make_kuid(&init_user_ns, EXT4_DEF_RESUID)) ||
1806 le16_to_cpu(es->s_def_resuid) != EXT4_DEF_RESUID)
1807 SEQ_OPTS_PRINT("resuid=%u",
1808 from_kuid_munged(&init_user_ns, sbi->s_resuid));
1809 if (nodefs || !gid_eq(sbi->s_resgid, make_kgid(&init_user_ns, EXT4_DEF_RESGID)) ||
1810 le16_to_cpu(es->s_def_resgid) != EXT4_DEF_RESGID)
1811 SEQ_OPTS_PRINT("resgid=%u",
1812 from_kgid_munged(&init_user_ns, sbi->s_resgid));
1813 def_errors = nodefs ? -1 : le16_to_cpu(es->s_errors);
1814 if (test_opt(sb, ERRORS_RO) && def_errors != EXT4_ERRORS_RO)
1815 SEQ_OPTS_PUTS("errors=remount-ro");
1816 if (test_opt(sb, ERRORS_CONT) && def_errors != EXT4_ERRORS_CONTINUE)
1817 SEQ_OPTS_PUTS("errors=continue");
1818 if (test_opt(sb, ERRORS_PANIC) && def_errors != EXT4_ERRORS_PANIC)
1819 SEQ_OPTS_PUTS("errors=panic");
1820 if (nodefs || sbi->s_commit_interval != JBD2_DEFAULT_MAX_COMMIT_AGE*HZ)
1821 SEQ_OPTS_PRINT("commit=%lu", sbi->s_commit_interval / HZ);
1822 if (nodefs || sbi->s_min_batch_time != EXT4_DEF_MIN_BATCH_TIME)
1823 SEQ_OPTS_PRINT("min_batch_time=%u", sbi->s_min_batch_time);
1824 if (nodefs || sbi->s_max_batch_time != EXT4_DEF_MAX_BATCH_TIME)
1825 SEQ_OPTS_PRINT("max_batch_time=%u", sbi->s_max_batch_time);
1826 if (sb->s_flags & MS_I_VERSION)
1827 SEQ_OPTS_PUTS("i_version");
1828 if (nodefs || sbi->s_stripe)
1829 SEQ_OPTS_PRINT("stripe=%lu", sbi->s_stripe);
1830 if (EXT4_MOUNT_DATA_FLAGS & (sbi->s_mount_opt ^ def_mount_opt)) {
1831 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA)
1832 SEQ_OPTS_PUTS("data=journal");
1833 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA)
1834 SEQ_OPTS_PUTS("data=ordered");
1835 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_WRITEBACK_DATA)
1836 SEQ_OPTS_PUTS("data=writeback");
1839 sbi->s_inode_readahead_blks != EXT4_DEF_INODE_READAHEAD_BLKS)
1840 SEQ_OPTS_PRINT("inode_readahead_blks=%u",
1841 sbi->s_inode_readahead_blks);
1843 if (nodefs || (test_opt(sb, INIT_INODE_TABLE) &&
1844 (sbi->s_li_wait_mult != EXT4_DEF_LI_WAIT_MULT)))
1845 SEQ_OPTS_PRINT("init_itable=%u", sbi->s_li_wait_mult);
1846 if (nodefs || sbi->s_max_dir_size_kb)
1847 SEQ_OPTS_PRINT("max_dir_size_kb=%u", sbi->s_max_dir_size_kb);
1849 ext4_show_quota_options(seq, sb);
1853 static int ext4_show_options(struct seq_file *seq, struct dentry *root)
1855 return _ext4_show_options(seq, root->d_sb, 0);
1858 static int options_seq_show(struct seq_file *seq, void *offset)
1860 struct super_block *sb = seq->private;
1863 seq_puts(seq, (sb->s_flags & MS_RDONLY) ? "ro" : "rw");
1864 rc = _ext4_show_options(seq, sb, 1);
1865 seq_puts(seq, "\n");
1869 static int options_open_fs(struct inode *inode, struct file *file)
1871 return single_open(file, options_seq_show, PDE_DATA(inode));
1874 static const struct file_operations ext4_seq_options_fops = {
1875 .owner = THIS_MODULE,
1876 .open = options_open_fs,
1878 .llseek = seq_lseek,
1879 .release = single_release,
1882 static int ext4_setup_super(struct super_block *sb, struct ext4_super_block *es,
1885 struct ext4_sb_info *sbi = EXT4_SB(sb);
1888 if (le32_to_cpu(es->s_rev_level) > EXT4_MAX_SUPP_REV) {
1889 ext4_msg(sb, KERN_ERR, "revision level too high, "
1890 "forcing read-only mode");
1895 if (!(sbi->s_mount_state & EXT4_VALID_FS))
1896 ext4_msg(sb, KERN_WARNING, "warning: mounting unchecked fs, "
1897 "running e2fsck is recommended");
1898 else if (sbi->s_mount_state & EXT4_ERROR_FS)
1899 ext4_msg(sb, KERN_WARNING,
1900 "warning: mounting fs with errors, "
1901 "running e2fsck is recommended");
1902 else if ((__s16) le16_to_cpu(es->s_max_mnt_count) > 0 &&
1903 le16_to_cpu(es->s_mnt_count) >=
1904 (unsigned short) (__s16) le16_to_cpu(es->s_max_mnt_count))
1905 ext4_msg(sb, KERN_WARNING,
1906 "warning: maximal mount count reached, "
1907 "running e2fsck is recommended");
1908 else if (le32_to_cpu(es->s_checkinterval) &&
1909 (le32_to_cpu(es->s_lastcheck) +
1910 le32_to_cpu(es->s_checkinterval) <= get_seconds()))
1911 ext4_msg(sb, KERN_WARNING,
1912 "warning: checktime reached, "
1913 "running e2fsck is recommended");
1914 if (!sbi->s_journal)
1915 es->s_state &= cpu_to_le16(~EXT4_VALID_FS);
1916 if (!(__s16) le16_to_cpu(es->s_max_mnt_count))
1917 es->s_max_mnt_count = cpu_to_le16(EXT4_DFL_MAX_MNT_COUNT);
1918 le16_add_cpu(&es->s_mnt_count, 1);
1919 es->s_mtime = cpu_to_le32(get_seconds());
1920 ext4_update_dynamic_rev(sb);
1922 EXT4_SET_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
1924 ext4_commit_super(sb, 1);
1926 if (test_opt(sb, DEBUG))
1927 printk(KERN_INFO "[EXT4 FS bs=%lu, gc=%u, "
1928 "bpg=%lu, ipg=%lu, mo=%04x, mo2=%04x]\n",
1930 sbi->s_groups_count,
1931 EXT4_BLOCKS_PER_GROUP(sb),
1932 EXT4_INODES_PER_GROUP(sb),
1933 sbi->s_mount_opt, sbi->s_mount_opt2);
1935 cleancache_init_fs(sb);
1939 int ext4_alloc_flex_bg_array(struct super_block *sb, ext4_group_t ngroup)
1941 struct ext4_sb_info *sbi = EXT4_SB(sb);
1942 struct flex_groups *new_groups;
1945 if (!sbi->s_log_groups_per_flex)
1948 size = ext4_flex_group(sbi, ngroup - 1) + 1;
1949 if (size <= sbi->s_flex_groups_allocated)
1952 size = roundup_pow_of_two(size * sizeof(struct flex_groups));
1953 new_groups = ext4_kvzalloc(size, GFP_KERNEL);
1955 ext4_msg(sb, KERN_ERR, "not enough memory for %d flex groups",
1956 size / (int) sizeof(struct flex_groups));
1960 if (sbi->s_flex_groups) {
1961 memcpy(new_groups, sbi->s_flex_groups,
1962 (sbi->s_flex_groups_allocated *
1963 sizeof(struct flex_groups)));
1964 ext4_kvfree(sbi->s_flex_groups);
1966 sbi->s_flex_groups = new_groups;
1967 sbi->s_flex_groups_allocated = size / sizeof(struct flex_groups);
1971 static int ext4_fill_flex_info(struct super_block *sb)
1973 struct ext4_sb_info *sbi = EXT4_SB(sb);
1974 struct ext4_group_desc *gdp = NULL;
1975 ext4_group_t flex_group;
1978 sbi->s_log_groups_per_flex = sbi->s_es->s_log_groups_per_flex;
1979 if (sbi->s_log_groups_per_flex < 1 || sbi->s_log_groups_per_flex > 31) {
1980 sbi->s_log_groups_per_flex = 0;
1984 err = ext4_alloc_flex_bg_array(sb, sbi->s_groups_count);
1988 for (i = 0; i < sbi->s_groups_count; i++) {
1989 gdp = ext4_get_group_desc(sb, i, NULL);
1991 flex_group = ext4_flex_group(sbi, i);
1992 atomic_add(ext4_free_inodes_count(sb, gdp),
1993 &sbi->s_flex_groups[flex_group].free_inodes);
1994 atomic64_add(ext4_free_group_clusters(sb, gdp),
1995 &sbi->s_flex_groups[flex_group].free_clusters);
1996 atomic_add(ext4_used_dirs_count(sb, gdp),
1997 &sbi->s_flex_groups[flex_group].used_dirs);
2005 static __le16 ext4_group_desc_csum(struct ext4_sb_info *sbi, __u32 block_group,
2006 struct ext4_group_desc *gdp)
2010 __le32 le_group = cpu_to_le32(block_group);
2012 if ((sbi->s_es->s_feature_ro_compat &
2013 cpu_to_le32(EXT4_FEATURE_RO_COMPAT_METADATA_CSUM))) {
2014 /* Use new metadata_csum algorithm */
2018 save_csum = gdp->bg_checksum;
2019 gdp->bg_checksum = 0;
2020 csum32 = ext4_chksum(sbi, sbi->s_csum_seed, (__u8 *)&le_group,
2022 csum32 = ext4_chksum(sbi, csum32, (__u8 *)gdp,
2024 gdp->bg_checksum = save_csum;
2026 crc = csum32 & 0xFFFF;
2030 /* old crc16 code */
2031 offset = offsetof(struct ext4_group_desc, bg_checksum);
2033 crc = crc16(~0, sbi->s_es->s_uuid, sizeof(sbi->s_es->s_uuid));
2034 crc = crc16(crc, (__u8 *)&le_group, sizeof(le_group));
2035 crc = crc16(crc, (__u8 *)gdp, offset);
2036 offset += sizeof(gdp->bg_checksum); /* skip checksum */
2037 /* for checksum of struct ext4_group_desc do the rest...*/
2038 if ((sbi->s_es->s_feature_incompat &
2039 cpu_to_le32(EXT4_FEATURE_INCOMPAT_64BIT)) &&
2040 offset < le16_to_cpu(sbi->s_es->s_desc_size))
2041 crc = crc16(crc, (__u8 *)gdp + offset,
2042 le16_to_cpu(sbi->s_es->s_desc_size) -
2046 return cpu_to_le16(crc);
2049 int ext4_group_desc_csum_verify(struct super_block *sb, __u32 block_group,
2050 struct ext4_group_desc *gdp)
2052 if (ext4_has_group_desc_csum(sb) &&
2053 (gdp->bg_checksum != ext4_group_desc_csum(EXT4_SB(sb),
2060 void ext4_group_desc_csum_set(struct super_block *sb, __u32 block_group,
2061 struct ext4_group_desc *gdp)
2063 if (!ext4_has_group_desc_csum(sb))
2065 gdp->bg_checksum = ext4_group_desc_csum(EXT4_SB(sb), block_group, gdp);
2068 /* Called at mount-time, super-block is locked */
2069 static int ext4_check_descriptors(struct super_block *sb,
2070 ext4_group_t *first_not_zeroed)
2072 struct ext4_sb_info *sbi = EXT4_SB(sb);
2073 ext4_fsblk_t first_block = le32_to_cpu(sbi->s_es->s_first_data_block);
2074 ext4_fsblk_t last_block;
2075 ext4_fsblk_t block_bitmap;
2076 ext4_fsblk_t inode_bitmap;
2077 ext4_fsblk_t inode_table;
2078 int flexbg_flag = 0;
2079 ext4_group_t i, grp = sbi->s_groups_count;
2081 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_FLEX_BG))
2084 ext4_debug("Checking group descriptors");
2086 for (i = 0; i < sbi->s_groups_count; i++) {
2087 struct ext4_group_desc *gdp = ext4_get_group_desc(sb, i, NULL);
2089 if (i == sbi->s_groups_count - 1 || flexbg_flag)
2090 last_block = ext4_blocks_count(sbi->s_es) - 1;
2092 last_block = first_block +
2093 (EXT4_BLOCKS_PER_GROUP(sb) - 1);
2095 if ((grp == sbi->s_groups_count) &&
2096 !(gdp->bg_flags & cpu_to_le16(EXT4_BG_INODE_ZEROED)))
2099 block_bitmap = ext4_block_bitmap(sb, gdp);
2100 if (block_bitmap < first_block || block_bitmap > last_block) {
2101 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2102 "Block bitmap for group %u not in group "
2103 "(block %llu)!", i, block_bitmap);
2106 inode_bitmap = ext4_inode_bitmap(sb, gdp);
2107 if (inode_bitmap < first_block || inode_bitmap > last_block) {
2108 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2109 "Inode bitmap for group %u not in group "
2110 "(block %llu)!", i, inode_bitmap);
2113 inode_table = ext4_inode_table(sb, gdp);
2114 if (inode_table < first_block ||
2115 inode_table + sbi->s_itb_per_group - 1 > last_block) {
2116 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2117 "Inode table for group %u not in group "
2118 "(block %llu)!", i, inode_table);
2121 ext4_lock_group(sb, i);
2122 if (!ext4_group_desc_csum_verify(sb, i, gdp)) {
2123 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2124 "Checksum for group %u failed (%u!=%u)",
2125 i, le16_to_cpu(ext4_group_desc_csum(sbi, i,
2126 gdp)), le16_to_cpu(gdp->bg_checksum));
2127 if (!(sb->s_flags & MS_RDONLY)) {
2128 ext4_unlock_group(sb, i);
2132 ext4_unlock_group(sb, i);
2134 first_block += EXT4_BLOCKS_PER_GROUP(sb);
2136 if (NULL != first_not_zeroed)
2137 *first_not_zeroed = grp;
2141 /* ext4_orphan_cleanup() walks a singly-linked list of inodes (starting at
2142 * the superblock) which were deleted from all directories, but held open by
2143 * a process at the time of a crash. We walk the list and try to delete these
2144 * inodes at recovery time (only with a read-write filesystem).
2146 * In order to keep the orphan inode chain consistent during traversal (in
2147 * case of crash during recovery), we link each inode into the superblock
2148 * orphan list_head and handle it the same way as an inode deletion during
2149 * normal operation (which journals the operations for us).
2151 * We only do an iget() and an iput() on each inode, which is very safe if we
2152 * accidentally point at an in-use or already deleted inode. The worst that
2153 * can happen in this case is that we get a "bit already cleared" message from
2154 * ext4_free_inode(). The only reason we would point at a wrong inode is if
2155 * e2fsck was run on this filesystem, and it must have already done the orphan
2156 * inode cleanup for us, so we can safely abort without any further action.
2158 static void ext4_orphan_cleanup(struct super_block *sb,
2159 struct ext4_super_block *es)
2161 unsigned int s_flags = sb->s_flags;
2162 int nr_orphans = 0, nr_truncates = 0;
2166 if (!es->s_last_orphan) {
2167 jbd_debug(4, "no orphan inodes to clean up\n");
2171 if (bdev_read_only(sb->s_bdev)) {
2172 ext4_msg(sb, KERN_ERR, "write access "
2173 "unavailable, skipping orphan cleanup");
2177 /* Check if feature set would not allow a r/w mount */
2178 if (!ext4_feature_set_ok(sb, 0)) {
2179 ext4_msg(sb, KERN_INFO, "Skipping orphan cleanup due to "
2180 "unknown ROCOMPAT features");
2184 if (EXT4_SB(sb)->s_mount_state & EXT4_ERROR_FS) {
2185 /* don't clear list on RO mount w/ errors */
2186 if (es->s_last_orphan && !(s_flags & MS_RDONLY)) {
2187 ext4_msg(sb, KERN_INFO, "Errors on filesystem, "
2188 "clearing orphan list.\n");
2189 es->s_last_orphan = 0;
2191 jbd_debug(1, "Skipping orphan recovery on fs with errors.\n");
2195 if (s_flags & MS_RDONLY) {
2196 ext4_msg(sb, KERN_INFO, "orphan cleanup on readonly fs");
2197 sb->s_flags &= ~MS_RDONLY;
2200 /* Needed for iput() to work correctly and not trash data */
2201 sb->s_flags |= MS_ACTIVE;
2202 /* Turn on quotas so that they are updated correctly */
2203 for (i = 0; i < EXT4_MAXQUOTAS; i++) {
2204 if (EXT4_SB(sb)->s_qf_names[i]) {
2205 int ret = ext4_quota_on_mount(sb, i);
2207 ext4_msg(sb, KERN_ERR,
2208 "Cannot turn on journaled "
2209 "quota: error %d", ret);
2214 while (es->s_last_orphan) {
2215 struct inode *inode;
2217 inode = ext4_orphan_get(sb, le32_to_cpu(es->s_last_orphan));
2218 if (IS_ERR(inode)) {
2219 es->s_last_orphan = 0;
2223 list_add(&EXT4_I(inode)->i_orphan, &EXT4_SB(sb)->s_orphan);
2224 dquot_initialize(inode);
2225 if (inode->i_nlink) {
2226 if (test_opt(sb, DEBUG))
2227 ext4_msg(sb, KERN_DEBUG,
2228 "%s: truncating inode %lu to %lld bytes",
2229 __func__, inode->i_ino, inode->i_size);
2230 jbd_debug(2, "truncating inode %lu to %lld bytes\n",
2231 inode->i_ino, inode->i_size);
2232 mutex_lock(&inode->i_mutex);
2233 truncate_inode_pages(inode->i_mapping, inode->i_size);
2234 ext4_truncate(inode);
2235 mutex_unlock(&inode->i_mutex);
2238 if (test_opt(sb, DEBUG))
2239 ext4_msg(sb, KERN_DEBUG,
2240 "%s: deleting unreferenced inode %lu",
2241 __func__, inode->i_ino);
2242 jbd_debug(2, "deleting unreferenced inode %lu\n",
2246 iput(inode); /* The delete magic happens here! */
2249 #define PLURAL(x) (x), ((x) == 1) ? "" : "s"
2252 ext4_msg(sb, KERN_INFO, "%d orphan inode%s deleted",
2253 PLURAL(nr_orphans));
2255 ext4_msg(sb, KERN_INFO, "%d truncate%s cleaned up",
2256 PLURAL(nr_truncates));
2258 /* Turn quotas off */
2259 for (i = 0; i < EXT4_MAXQUOTAS; i++) {
2260 if (sb_dqopt(sb)->files[i])
2261 dquot_quota_off(sb, i);
2264 sb->s_flags = s_flags; /* Restore MS_RDONLY status */
2268 * Maximal extent format file size.
2269 * Resulting logical blkno at s_maxbytes must fit in our on-disk
2270 * extent format containers, within a sector_t, and within i_blocks
2271 * in the vfs. ext4 inode has 48 bits of i_block in fsblock units,
2272 * so that won't be a limiting factor.
2274 * However there is other limiting factor. We do store extents in the form
2275 * of starting block and length, hence the resulting length of the extent
2276 * covering maximum file size must fit into on-disk format containers as
2277 * well. Given that length is always by 1 unit bigger than max unit (because
2278 * we count 0 as well) we have to lower the s_maxbytes by one fs block.
2280 * Note, this does *not* consider any metadata overhead for vfs i_blocks.
2282 static loff_t ext4_max_size(int blkbits, int has_huge_files)
2285 loff_t upper_limit = MAX_LFS_FILESIZE;
2287 /* small i_blocks in vfs inode? */
2288 if (!has_huge_files || sizeof(blkcnt_t) < sizeof(u64)) {
2290 * CONFIG_LBDAF is not enabled implies the inode
2291 * i_block represent total blocks in 512 bytes
2292 * 32 == size of vfs inode i_blocks * 8
2294 upper_limit = (1LL << 32) - 1;
2296 /* total blocks in file system block size */
2297 upper_limit >>= (blkbits - 9);
2298 upper_limit <<= blkbits;
2302 * 32-bit extent-start container, ee_block. We lower the maxbytes
2303 * by one fs block, so ee_len can cover the extent of maximum file
2306 res = (1LL << 32) - 1;
2309 /* Sanity check against vm- & vfs- imposed limits */
2310 if (res > upper_limit)
2317 * Maximal bitmap file size. There is a direct, and {,double-,triple-}indirect
2318 * block limit, and also a limit of (2^48 - 1) 512-byte sectors in i_blocks.
2319 * We need to be 1 filesystem block less than the 2^48 sector limit.
2321 static loff_t ext4_max_bitmap_size(int bits, int has_huge_files)
2323 loff_t res = EXT4_NDIR_BLOCKS;
2326 /* This is calculated to be the largest file size for a dense, block
2327 * mapped file such that the file's total number of 512-byte sectors,
2328 * including data and all indirect blocks, does not exceed (2^48 - 1).
2330 * __u32 i_blocks_lo and _u16 i_blocks_high represent the total
2331 * number of 512-byte sectors of the file.
2334 if (!has_huge_files || sizeof(blkcnt_t) < sizeof(u64)) {
2336 * !has_huge_files or CONFIG_LBDAF not enabled implies that
2337 * the inode i_block field represents total file blocks in
2338 * 2^32 512-byte sectors == size of vfs inode i_blocks * 8
2340 upper_limit = (1LL << 32) - 1;
2342 /* total blocks in file system block size */
2343 upper_limit >>= (bits - 9);
2347 * We use 48 bit ext4_inode i_blocks
2348 * With EXT4_HUGE_FILE_FL set the i_blocks
2349 * represent total number of blocks in
2350 * file system block size
2352 upper_limit = (1LL << 48) - 1;
2356 /* indirect blocks */
2358 /* double indirect blocks */
2359 meta_blocks += 1 + (1LL << (bits-2));
2360 /* tripple indirect blocks */
2361 meta_blocks += 1 + (1LL << (bits-2)) + (1LL << (2*(bits-2)));
2363 upper_limit -= meta_blocks;
2364 upper_limit <<= bits;
2366 res += 1LL << (bits-2);
2367 res += 1LL << (2*(bits-2));
2368 res += 1LL << (3*(bits-2));
2370 if (res > upper_limit)
2373 if (res > MAX_LFS_FILESIZE)
2374 res = MAX_LFS_FILESIZE;
2379 static ext4_fsblk_t descriptor_loc(struct super_block *sb,
2380 ext4_fsblk_t logical_sb_block, int nr)
2382 struct ext4_sb_info *sbi = EXT4_SB(sb);
2383 ext4_group_t bg, first_meta_bg;
2386 first_meta_bg = le32_to_cpu(sbi->s_es->s_first_meta_bg);
2388 if (!EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_META_BG) ||
2390 return logical_sb_block + nr + 1;
2391 bg = sbi->s_desc_per_block * nr;
2392 if (ext4_bg_has_super(sb, bg))
2396 * If we have a meta_bg fs with 1k blocks, group 0's GDT is at
2397 * block 2, not 1. If s_first_data_block == 0 (bigalloc is enabled
2398 * on modern mke2fs or blksize > 1k on older mke2fs) then we must
2401 if (sb->s_blocksize == 1024 && nr == 0 &&
2402 le32_to_cpu(EXT4_SB(sb)->s_es->s_first_data_block) == 0)
2405 return (has_super + ext4_group_first_block_no(sb, bg));
2409 * ext4_get_stripe_size: Get the stripe size.
2410 * @sbi: In memory super block info
2412 * If we have specified it via mount option, then
2413 * use the mount option value. If the value specified at mount time is
2414 * greater than the blocks per group use the super block value.
2415 * If the super block value is greater than blocks per group return 0.
2416 * Allocator needs it be less than blocks per group.
2419 static unsigned long ext4_get_stripe_size(struct ext4_sb_info *sbi)
2421 unsigned long stride = le16_to_cpu(sbi->s_es->s_raid_stride);
2422 unsigned long stripe_width =
2423 le32_to_cpu(sbi->s_es->s_raid_stripe_width);
2426 if (sbi->s_stripe && sbi->s_stripe <= sbi->s_blocks_per_group)
2427 ret = sbi->s_stripe;
2428 else if (stripe_width <= sbi->s_blocks_per_group)
2430 else if (stride <= sbi->s_blocks_per_group)
2436 * If the stripe width is 1, this makes no sense and
2437 * we set it to 0 to turn off stripe handling code.
2448 struct attribute attr;
2449 ssize_t (*show)(struct ext4_attr *, struct ext4_sb_info *, char *);
2450 ssize_t (*store)(struct ext4_attr *, struct ext4_sb_info *,
2451 const char *, size_t);
2458 static int parse_strtoull(const char *buf,
2459 unsigned long long max, unsigned long long *value)
2463 ret = kstrtoull(skip_spaces(buf), 0, value);
2464 if (!ret && *value > max)
2469 static ssize_t delayed_allocation_blocks_show(struct ext4_attr *a,
2470 struct ext4_sb_info *sbi,
2473 return snprintf(buf, PAGE_SIZE, "%llu\n",
2475 percpu_counter_sum(&sbi->s_dirtyclusters_counter)));
2478 static ssize_t session_write_kbytes_show(struct ext4_attr *a,
2479 struct ext4_sb_info *sbi, char *buf)
2481 struct super_block *sb = sbi->s_buddy_cache->i_sb;
2483 if (!sb->s_bdev->bd_part)
2484 return snprintf(buf, PAGE_SIZE, "0\n");
2485 return snprintf(buf, PAGE_SIZE, "%lu\n",
2486 (part_stat_read(sb->s_bdev->bd_part, sectors[1]) -
2487 sbi->s_sectors_written_start) >> 1);
2490 static ssize_t lifetime_write_kbytes_show(struct ext4_attr *a,
2491 struct ext4_sb_info *sbi, char *buf)
2493 struct super_block *sb = sbi->s_buddy_cache->i_sb;
2495 if (!sb->s_bdev->bd_part)
2496 return snprintf(buf, PAGE_SIZE, "0\n");
2497 return snprintf(buf, PAGE_SIZE, "%llu\n",
2498 (unsigned long long)(sbi->s_kbytes_written +
2499 ((part_stat_read(sb->s_bdev->bd_part, sectors[1]) -
2500 EXT4_SB(sb)->s_sectors_written_start) >> 1)));
2503 static ssize_t inode_readahead_blks_store(struct ext4_attr *a,
2504 struct ext4_sb_info *sbi,
2505 const char *buf, size_t count)
2510 ret = kstrtoul(skip_spaces(buf), 0, &t);
2514 if (t && (!is_power_of_2(t) || t > 0x40000000))
2517 sbi->s_inode_readahead_blks = t;
2521 static ssize_t sbi_ui_show(struct ext4_attr *a,
2522 struct ext4_sb_info *sbi, char *buf)
2524 unsigned int *ui = (unsigned int *) (((char *) sbi) + a->u.offset);
2526 return snprintf(buf, PAGE_SIZE, "%u\n", *ui);
2529 static ssize_t sbi_ui_store(struct ext4_attr *a,
2530 struct ext4_sb_info *sbi,
2531 const char *buf, size_t count)
2533 unsigned int *ui = (unsigned int *) (((char *) sbi) + a->u.offset);
2537 ret = kstrtoul(skip_spaces(buf), 0, &t);
2544 static ssize_t es_ui_show(struct ext4_attr *a,
2545 struct ext4_sb_info *sbi, char *buf)
2548 unsigned int *ui = (unsigned int *) (((char *) sbi->s_es) +
2551 return snprintf(buf, PAGE_SIZE, "%u\n", *ui);
2554 static ssize_t reserved_clusters_show(struct ext4_attr *a,
2555 struct ext4_sb_info *sbi, char *buf)
2557 return snprintf(buf, PAGE_SIZE, "%llu\n",
2558 (unsigned long long) atomic64_read(&sbi->s_resv_clusters));
2561 static ssize_t reserved_clusters_store(struct ext4_attr *a,
2562 struct ext4_sb_info *sbi,
2563 const char *buf, size_t count)
2565 unsigned long long val;
2568 if (parse_strtoull(buf, -1ULL, &val))
2570 ret = ext4_reserve_clusters(sbi, val);
2572 return ret ? ret : count;
2575 static ssize_t trigger_test_error(struct ext4_attr *a,
2576 struct ext4_sb_info *sbi,
2577 const char *buf, size_t count)
2581 if (!capable(CAP_SYS_ADMIN))
2584 if (len && buf[len-1] == '\n')
2588 ext4_error(sbi->s_sb, "%.*s", len, buf);
2592 static ssize_t sbi_deprecated_show(struct ext4_attr *a,
2593 struct ext4_sb_info *sbi, char *buf)
2595 return snprintf(buf, PAGE_SIZE, "%d\n", a->u.deprecated_val);
2598 #define EXT4_ATTR_OFFSET(_name,_mode,_show,_store,_elname) \
2599 static struct ext4_attr ext4_attr_##_name = { \
2600 .attr = {.name = __stringify(_name), .mode = _mode }, \
2604 .offset = offsetof(struct ext4_sb_info, _elname),\
2608 #define EXT4_ATTR_OFFSET_ES(_name,_mode,_show,_store,_elname) \
2609 static struct ext4_attr ext4_attr_##_name = { \
2610 .attr = {.name = __stringify(_name), .mode = _mode }, \
2614 .offset = offsetof(struct ext4_super_block, _elname), \
2618 #define EXT4_ATTR(name, mode, show, store) \
2619 static struct ext4_attr ext4_attr_##name = __ATTR(name, mode, show, store)
2621 #define EXT4_INFO_ATTR(name) EXT4_ATTR(name, 0444, NULL, NULL)
2622 #define EXT4_RO_ATTR(name) EXT4_ATTR(name, 0444, name##_show, NULL)
2623 #define EXT4_RW_ATTR(name) EXT4_ATTR(name, 0644, name##_show, name##_store)
2625 #define EXT4_RO_ATTR_ES_UI(name, elname) \
2626 EXT4_ATTR_OFFSET_ES(name, 0444, es_ui_show, NULL, elname)
2627 #define EXT4_RW_ATTR_SBI_UI(name, elname) \
2628 EXT4_ATTR_OFFSET(name, 0644, sbi_ui_show, sbi_ui_store, elname)
2630 #define ATTR_LIST(name) &ext4_attr_##name.attr
2631 #define EXT4_DEPRECATED_ATTR(_name, _val) \
2632 static struct ext4_attr ext4_attr_##_name = { \
2633 .attr = {.name = __stringify(_name), .mode = 0444 }, \
2634 .show = sbi_deprecated_show, \
2636 .deprecated_val = _val, \
2640 EXT4_RO_ATTR(delayed_allocation_blocks);
2641 EXT4_RO_ATTR(session_write_kbytes);
2642 EXT4_RO_ATTR(lifetime_write_kbytes);
2643 EXT4_RW_ATTR(reserved_clusters);
2644 EXT4_ATTR_OFFSET(inode_readahead_blks, 0644, sbi_ui_show,
2645 inode_readahead_blks_store, s_inode_readahead_blks);
2646 EXT4_RW_ATTR_SBI_UI(inode_goal, s_inode_goal);
2647 EXT4_RW_ATTR_SBI_UI(mb_stats, s_mb_stats);
2648 EXT4_RW_ATTR_SBI_UI(mb_max_to_scan, s_mb_max_to_scan);
2649 EXT4_RW_ATTR_SBI_UI(mb_min_to_scan, s_mb_min_to_scan);
2650 EXT4_RW_ATTR_SBI_UI(mb_order2_req, s_mb_order2_reqs);
2651 EXT4_RW_ATTR_SBI_UI(mb_stream_req, s_mb_stream_request);
2652 EXT4_RW_ATTR_SBI_UI(mb_group_prealloc, s_mb_group_prealloc);
2653 EXT4_DEPRECATED_ATTR(max_writeback_mb_bump, 128);
2654 EXT4_RW_ATTR_SBI_UI(extent_max_zeroout_kb, s_extent_max_zeroout_kb);
2655 EXT4_ATTR(trigger_fs_error, 0200, NULL, trigger_test_error);
2656 EXT4_RW_ATTR_SBI_UI(err_ratelimit_interval_ms, s_err_ratelimit_state.interval);
2657 EXT4_RW_ATTR_SBI_UI(err_ratelimit_burst, s_err_ratelimit_state.burst);
2658 EXT4_RW_ATTR_SBI_UI(warning_ratelimit_interval_ms, s_warning_ratelimit_state.interval);
2659 EXT4_RW_ATTR_SBI_UI(warning_ratelimit_burst, s_warning_ratelimit_state.burst);
2660 EXT4_RW_ATTR_SBI_UI(msg_ratelimit_interval_ms, s_msg_ratelimit_state.interval);
2661 EXT4_RW_ATTR_SBI_UI(msg_ratelimit_burst, s_msg_ratelimit_state.burst);
2662 EXT4_RO_ATTR_ES_UI(errors_count, s_error_count);
2663 EXT4_RO_ATTR_ES_UI(first_error_time, s_first_error_time);
2664 EXT4_RO_ATTR_ES_UI(last_error_time, s_last_error_time);
2666 static struct attribute *ext4_attrs[] = {
2667 ATTR_LIST(delayed_allocation_blocks),
2668 ATTR_LIST(session_write_kbytes),
2669 ATTR_LIST(lifetime_write_kbytes),
2670 ATTR_LIST(reserved_clusters),
2671 ATTR_LIST(inode_readahead_blks),
2672 ATTR_LIST(inode_goal),
2673 ATTR_LIST(mb_stats),
2674 ATTR_LIST(mb_max_to_scan),
2675 ATTR_LIST(mb_min_to_scan),
2676 ATTR_LIST(mb_order2_req),
2677 ATTR_LIST(mb_stream_req),
2678 ATTR_LIST(mb_group_prealloc),
2679 ATTR_LIST(max_writeback_mb_bump),
2680 ATTR_LIST(extent_max_zeroout_kb),
2681 ATTR_LIST(trigger_fs_error),
2682 ATTR_LIST(err_ratelimit_interval_ms),
2683 ATTR_LIST(err_ratelimit_burst),
2684 ATTR_LIST(warning_ratelimit_interval_ms),
2685 ATTR_LIST(warning_ratelimit_burst),
2686 ATTR_LIST(msg_ratelimit_interval_ms),
2687 ATTR_LIST(msg_ratelimit_burst),
2688 ATTR_LIST(errors_count),
2689 ATTR_LIST(first_error_time),
2690 ATTR_LIST(last_error_time),
2694 /* Features this copy of ext4 supports */
2695 EXT4_INFO_ATTR(lazy_itable_init);
2696 EXT4_INFO_ATTR(batched_discard);
2697 EXT4_INFO_ATTR(meta_bg_resize);
2699 static struct attribute *ext4_feat_attrs[] = {
2700 ATTR_LIST(lazy_itable_init),
2701 ATTR_LIST(batched_discard),
2702 ATTR_LIST(meta_bg_resize),
2706 static ssize_t ext4_attr_show(struct kobject *kobj,
2707 struct attribute *attr, char *buf)
2709 struct ext4_sb_info *sbi = container_of(kobj, struct ext4_sb_info,
2711 struct ext4_attr *a = container_of(attr, struct ext4_attr, attr);
2713 return a->show ? a->show(a, sbi, buf) : 0;
2716 static ssize_t ext4_attr_store(struct kobject *kobj,
2717 struct attribute *attr,
2718 const char *buf, size_t len)
2720 struct ext4_sb_info *sbi = container_of(kobj, struct ext4_sb_info,
2722 struct ext4_attr *a = container_of(attr, struct ext4_attr, attr);
2724 return a->store ? a->store(a, sbi, buf, len) : 0;
2727 static void ext4_sb_release(struct kobject *kobj)
2729 struct ext4_sb_info *sbi = container_of(kobj, struct ext4_sb_info,
2731 complete(&sbi->s_kobj_unregister);
2734 static const struct sysfs_ops ext4_attr_ops = {
2735 .show = ext4_attr_show,
2736 .store = ext4_attr_store,
2739 static struct kobj_type ext4_ktype = {
2740 .default_attrs = ext4_attrs,
2741 .sysfs_ops = &ext4_attr_ops,
2742 .release = ext4_sb_release,
2745 static void ext4_feat_release(struct kobject *kobj)
2747 complete(&ext4_feat->f_kobj_unregister);
2750 static ssize_t ext4_feat_show(struct kobject *kobj,
2751 struct attribute *attr, char *buf)
2753 return snprintf(buf, PAGE_SIZE, "supported\n");
2757 * We can not use ext4_attr_show/store because it relies on the kobject
2758 * being embedded in the ext4_sb_info structure which is definitely not
2759 * true in this case.
2761 static const struct sysfs_ops ext4_feat_ops = {
2762 .show = ext4_feat_show,
2766 static struct kobj_type ext4_feat_ktype = {
2767 .default_attrs = ext4_feat_attrs,
2768 .sysfs_ops = &ext4_feat_ops,
2769 .release = ext4_feat_release,
2773 * Check whether this filesystem can be mounted based on
2774 * the features present and the RDONLY/RDWR mount requested.
2775 * Returns 1 if this filesystem can be mounted as requested,
2776 * 0 if it cannot be.
2778 static int ext4_feature_set_ok(struct super_block *sb, int readonly)
2780 if (EXT4_HAS_INCOMPAT_FEATURE(sb, ~EXT4_FEATURE_INCOMPAT_SUPP)) {
2781 ext4_msg(sb, KERN_ERR,
2782 "Couldn't mount because of "
2783 "unsupported optional features (%x)",
2784 (le32_to_cpu(EXT4_SB(sb)->s_es->s_feature_incompat) &
2785 ~EXT4_FEATURE_INCOMPAT_SUPP));
2792 /* Check that feature set is OK for a read-write mount */
2793 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, ~EXT4_FEATURE_RO_COMPAT_SUPP)) {
2794 ext4_msg(sb, KERN_ERR, "couldn't mount RDWR because of "
2795 "unsupported optional features (%x)",
2796 (le32_to_cpu(EXT4_SB(sb)->s_es->s_feature_ro_compat) &
2797 ~EXT4_FEATURE_RO_COMPAT_SUPP));
2801 * Large file size enabled file system can only be mounted
2802 * read-write on 32-bit systems if kernel is built with CONFIG_LBDAF
2804 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_HUGE_FILE)) {
2805 if (sizeof(blkcnt_t) < sizeof(u64)) {
2806 ext4_msg(sb, KERN_ERR, "Filesystem with huge files "
2807 "cannot be mounted RDWR without "
2812 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_BIGALLOC) &&
2813 !EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_EXTENTS)) {
2814 ext4_msg(sb, KERN_ERR,
2815 "Can't support bigalloc feature without "
2816 "extents feature\n");
2820 #ifndef CONFIG_QUOTA
2821 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_QUOTA) &&
2823 ext4_msg(sb, KERN_ERR,
2824 "Filesystem with quota feature cannot be mounted RDWR "
2825 "without CONFIG_QUOTA");
2828 #endif /* CONFIG_QUOTA */
2833 * This function is called once a day if we have errors logged
2834 * on the file system
2836 static void print_daily_error_info(unsigned long arg)
2838 struct super_block *sb = (struct super_block *) arg;
2839 struct ext4_sb_info *sbi;
2840 struct ext4_super_block *es;
2845 if (es->s_error_count)
2846 /* fsck newer than v1.41.13 is needed to clean this condition. */
2847 ext4_msg(sb, KERN_NOTICE, "error count since last fsck: %u",
2848 le32_to_cpu(es->s_error_count));
2849 if (es->s_first_error_time) {
2850 printk(KERN_NOTICE "EXT4-fs (%s): initial error at time %u: %.*s:%d",
2851 sb->s_id, le32_to_cpu(es->s_first_error_time),
2852 (int) sizeof(es->s_first_error_func),
2853 es->s_first_error_func,
2854 le32_to_cpu(es->s_first_error_line));
2855 if (es->s_first_error_ino)
2856 printk(": inode %u",
2857 le32_to_cpu(es->s_first_error_ino));
2858 if (es->s_first_error_block)
2859 printk(": block %llu", (unsigned long long)
2860 le64_to_cpu(es->s_first_error_block));
2863 if (es->s_last_error_time) {
2864 printk(KERN_NOTICE "EXT4-fs (%s): last error at time %u: %.*s:%d",
2865 sb->s_id, le32_to_cpu(es->s_last_error_time),
2866 (int) sizeof(es->s_last_error_func),
2867 es->s_last_error_func,
2868 le32_to_cpu(es->s_last_error_line));
2869 if (es->s_last_error_ino)
2870 printk(": inode %u",
2871 le32_to_cpu(es->s_last_error_ino));
2872 if (es->s_last_error_block)
2873 printk(": block %llu", (unsigned long long)
2874 le64_to_cpu(es->s_last_error_block));
2877 mod_timer(&sbi->s_err_report, jiffies + 24*60*60*HZ); /* Once a day */
2880 /* Find next suitable group and run ext4_init_inode_table */
2881 static int ext4_run_li_request(struct ext4_li_request *elr)
2883 struct ext4_group_desc *gdp = NULL;
2884 ext4_group_t group, ngroups;
2885 struct super_block *sb;
2886 unsigned long timeout = 0;
2890 ngroups = EXT4_SB(sb)->s_groups_count;
2893 for (group = elr->lr_next_group; group < ngroups; group++) {
2894 gdp = ext4_get_group_desc(sb, group, NULL);
2900 if (!(gdp->bg_flags & cpu_to_le16(EXT4_BG_INODE_ZEROED)))
2904 if (group >= ngroups)
2909 ret = ext4_init_inode_table(sb, group,
2910 elr->lr_timeout ? 0 : 1);
2911 if (elr->lr_timeout == 0) {
2912 timeout = (jiffies - timeout) *
2913 elr->lr_sbi->s_li_wait_mult;
2914 elr->lr_timeout = timeout;
2916 elr->lr_next_sched = jiffies + elr->lr_timeout;
2917 elr->lr_next_group = group + 1;
2925 * Remove lr_request from the list_request and free the
2926 * request structure. Should be called with li_list_mtx held
2928 static void ext4_remove_li_request(struct ext4_li_request *elr)
2930 struct ext4_sb_info *sbi;
2937 list_del(&elr->lr_request);
2938 sbi->s_li_request = NULL;
2942 static void ext4_unregister_li_request(struct super_block *sb)
2944 mutex_lock(&ext4_li_mtx);
2945 if (!ext4_li_info) {
2946 mutex_unlock(&ext4_li_mtx);
2950 mutex_lock(&ext4_li_info->li_list_mtx);
2951 ext4_remove_li_request(EXT4_SB(sb)->s_li_request);
2952 mutex_unlock(&ext4_li_info->li_list_mtx);
2953 mutex_unlock(&ext4_li_mtx);
2956 static struct task_struct *ext4_lazyinit_task;
2959 * This is the function where ext4lazyinit thread lives. It walks
2960 * through the request list searching for next scheduled filesystem.
2961 * When such a fs is found, run the lazy initialization request
2962 * (ext4_rn_li_request) and keep track of the time spend in this
2963 * function. Based on that time we compute next schedule time of
2964 * the request. When walking through the list is complete, compute
2965 * next waking time and put itself into sleep.
2967 static int ext4_lazyinit_thread(void *arg)
2969 struct ext4_lazy_init *eli = (struct ext4_lazy_init *)arg;
2970 struct list_head *pos, *n;
2971 struct ext4_li_request *elr;
2972 unsigned long next_wakeup, cur;
2974 BUG_ON(NULL == eli);
2978 next_wakeup = MAX_JIFFY_OFFSET;
2980 mutex_lock(&eli->li_list_mtx);
2981 if (list_empty(&eli->li_request_list)) {
2982 mutex_unlock(&eli->li_list_mtx);
2986 list_for_each_safe(pos, n, &eli->li_request_list) {
2987 elr = list_entry(pos, struct ext4_li_request,
2990 if (time_after_eq(jiffies, elr->lr_next_sched)) {
2991 if (ext4_run_li_request(elr) != 0) {
2992 /* error, remove the lazy_init job */
2993 ext4_remove_li_request(elr);
2998 if (time_before(elr->lr_next_sched, next_wakeup))
2999 next_wakeup = elr->lr_next_sched;
3001 mutex_unlock(&eli->li_list_mtx);
3006 if ((time_after_eq(cur, next_wakeup)) ||
3007 (MAX_JIFFY_OFFSET == next_wakeup)) {
3012 schedule_timeout_interruptible(next_wakeup - cur);
3014 if (kthread_should_stop()) {
3015 ext4_clear_request_list();
3022 * It looks like the request list is empty, but we need
3023 * to check it under the li_list_mtx lock, to prevent any
3024 * additions into it, and of course we should lock ext4_li_mtx
3025 * to atomically free the list and ext4_li_info, because at
3026 * this point another ext4 filesystem could be registering
3029 mutex_lock(&ext4_li_mtx);
3030 mutex_lock(&eli->li_list_mtx);
3031 if (!list_empty(&eli->li_request_list)) {
3032 mutex_unlock(&eli->li_list_mtx);
3033 mutex_unlock(&ext4_li_mtx);
3036 mutex_unlock(&eli->li_list_mtx);
3037 kfree(ext4_li_info);
3038 ext4_li_info = NULL;
3039 mutex_unlock(&ext4_li_mtx);
3044 static void ext4_clear_request_list(void)
3046 struct list_head *pos, *n;
3047 struct ext4_li_request *elr;
3049 mutex_lock(&ext4_li_info->li_list_mtx);
3050 list_for_each_safe(pos, n, &ext4_li_info->li_request_list) {
3051 elr = list_entry(pos, struct ext4_li_request,
3053 ext4_remove_li_request(elr);
3055 mutex_unlock(&ext4_li_info->li_list_mtx);
3058 static int ext4_run_lazyinit_thread(void)
3060 ext4_lazyinit_task = kthread_run(ext4_lazyinit_thread,
3061 ext4_li_info, "ext4lazyinit");
3062 if (IS_ERR(ext4_lazyinit_task)) {
3063 int err = PTR_ERR(ext4_lazyinit_task);
3064 ext4_clear_request_list();
3065 kfree(ext4_li_info);
3066 ext4_li_info = NULL;
3067 printk(KERN_CRIT "EXT4-fs: error %d creating inode table "
3068 "initialization thread\n",
3072 ext4_li_info->li_state |= EXT4_LAZYINIT_RUNNING;
3077 * Check whether it make sense to run itable init. thread or not.
3078 * If there is at least one uninitialized inode table, return
3079 * corresponding group number, else the loop goes through all
3080 * groups and return total number of groups.
3082 static ext4_group_t ext4_has_uninit_itable(struct super_block *sb)
3084 ext4_group_t group, ngroups = EXT4_SB(sb)->s_groups_count;
3085 struct ext4_group_desc *gdp = NULL;
3087 for (group = 0; group < ngroups; group++) {
3088 gdp = ext4_get_group_desc(sb, group, NULL);
3092 if (!(gdp->bg_flags & cpu_to_le16(EXT4_BG_INODE_ZEROED)))
3099 static int ext4_li_info_new(void)
3101 struct ext4_lazy_init *eli = NULL;
3103 eli = kzalloc(sizeof(*eli), GFP_KERNEL);
3107 INIT_LIST_HEAD(&eli->li_request_list);
3108 mutex_init(&eli->li_list_mtx);
3110 eli->li_state |= EXT4_LAZYINIT_QUIT;
3117 static struct ext4_li_request *ext4_li_request_new(struct super_block *sb,
3120 struct ext4_sb_info *sbi = EXT4_SB(sb);
3121 struct ext4_li_request *elr;
3123 elr = kzalloc(sizeof(*elr), GFP_KERNEL);
3129 elr->lr_next_group = start;
3132 * Randomize first schedule time of the request to
3133 * spread the inode table initialization requests
3136 elr->lr_next_sched = jiffies + (prandom_u32() %
3137 (EXT4_DEF_LI_MAX_START_DELAY * HZ));
3141 int ext4_register_li_request(struct super_block *sb,
3142 ext4_group_t first_not_zeroed)
3144 struct ext4_sb_info *sbi = EXT4_SB(sb);
3145 struct ext4_li_request *elr = NULL;
3146 ext4_group_t ngroups = EXT4_SB(sb)->s_groups_count;
3149 mutex_lock(&ext4_li_mtx);
3150 if (sbi->s_li_request != NULL) {
3152 * Reset timeout so it can be computed again, because
3153 * s_li_wait_mult might have changed.
3155 sbi->s_li_request->lr_timeout = 0;
3159 if (first_not_zeroed == ngroups ||
3160 (sb->s_flags & MS_RDONLY) ||
3161 !test_opt(sb, INIT_INODE_TABLE))
3164 elr = ext4_li_request_new(sb, first_not_zeroed);
3170 if (NULL == ext4_li_info) {
3171 ret = ext4_li_info_new();
3176 mutex_lock(&ext4_li_info->li_list_mtx);
3177 list_add(&elr->lr_request, &ext4_li_info->li_request_list);
3178 mutex_unlock(&ext4_li_info->li_list_mtx);
3180 sbi->s_li_request = elr;
3182 * set elr to NULL here since it has been inserted to
3183 * the request_list and the removal and free of it is
3184 * handled by ext4_clear_request_list from now on.
3188 if (!(ext4_li_info->li_state & EXT4_LAZYINIT_RUNNING)) {
3189 ret = ext4_run_lazyinit_thread();
3194 mutex_unlock(&ext4_li_mtx);
3201 * We do not need to lock anything since this is called on
3204 static void ext4_destroy_lazyinit_thread(void)
3207 * If thread exited earlier
3208 * there's nothing to be done.
3210 if (!ext4_li_info || !ext4_lazyinit_task)
3213 kthread_stop(ext4_lazyinit_task);
3216 static int set_journal_csum_feature_set(struct super_block *sb)
3219 int compat, incompat;
3220 struct ext4_sb_info *sbi = EXT4_SB(sb);
3222 if (EXT4_HAS_RO_COMPAT_FEATURE(sb,
3223 EXT4_FEATURE_RO_COMPAT_METADATA_CSUM)) {
3224 /* journal checksum v3 */
3226 incompat = JBD2_FEATURE_INCOMPAT_CSUM_V3;
3228 /* journal checksum v1 */
3229 compat = JBD2_FEATURE_COMPAT_CHECKSUM;
3233 jbd2_journal_clear_features(sbi->s_journal,
3234 JBD2_FEATURE_COMPAT_CHECKSUM, 0,
3235 JBD2_FEATURE_INCOMPAT_CSUM_V3 |
3236 JBD2_FEATURE_INCOMPAT_CSUM_V2);
3237 if (test_opt(sb, JOURNAL_ASYNC_COMMIT)) {
3238 ret = jbd2_journal_set_features(sbi->s_journal,
3240 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT |
3242 } else if (test_opt(sb, JOURNAL_CHECKSUM)) {
3243 ret = jbd2_journal_set_features(sbi->s_journal,
3246 jbd2_journal_clear_features(sbi->s_journal, 0, 0,
3247 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT);
3249 jbd2_journal_clear_features(sbi->s_journal, 0, 0,
3250 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT);
3257 * Note: calculating the overhead so we can be compatible with
3258 * historical BSD practice is quite difficult in the face of
3259 * clusters/bigalloc. This is because multiple metadata blocks from
3260 * different block group can end up in the same allocation cluster.
3261 * Calculating the exact overhead in the face of clustered allocation
3262 * requires either O(all block bitmaps) in memory or O(number of block
3263 * groups**2) in time. We will still calculate the superblock for
3264 * older file systems --- and if we come across with a bigalloc file
3265 * system with zero in s_overhead_clusters the estimate will be close to
3266 * correct especially for very large cluster sizes --- but for newer
3267 * file systems, it's better to calculate this figure once at mkfs
3268 * time, and store it in the superblock. If the superblock value is
3269 * present (even for non-bigalloc file systems), we will use it.
3271 static int count_overhead(struct super_block *sb, ext4_group_t grp,
3274 struct ext4_sb_info *sbi = EXT4_SB(sb);
3275 struct ext4_group_desc *gdp;
3276 ext4_fsblk_t first_block, last_block, b;
3277 ext4_group_t i, ngroups = ext4_get_groups_count(sb);
3278 int s, j, count = 0;
3280 if (!EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_BIGALLOC))
3281 return (ext4_bg_has_super(sb, grp) + ext4_bg_num_gdb(sb, grp) +
3282 sbi->s_itb_per_group + 2);
3284 first_block = le32_to_cpu(sbi->s_es->s_first_data_block) +
3285 (grp * EXT4_BLOCKS_PER_GROUP(sb));
3286 last_block = first_block + EXT4_BLOCKS_PER_GROUP(sb) - 1;
3287 for (i = 0; i < ngroups; i++) {
3288 gdp = ext4_get_group_desc(sb, i, NULL);
3289 b = ext4_block_bitmap(sb, gdp);
3290 if (b >= first_block && b <= last_block) {
3291 ext4_set_bit(EXT4_B2C(sbi, b - first_block), buf);
3294 b = ext4_inode_bitmap(sb, gdp);
3295 if (b >= first_block && b <= last_block) {
3296 ext4_set_bit(EXT4_B2C(sbi, b - first_block), buf);
3299 b = ext4_inode_table(sb, gdp);
3300 if (b >= first_block && b + sbi->s_itb_per_group <= last_block)
3301 for (j = 0; j < sbi->s_itb_per_group; j++, b++) {
3302 int c = EXT4_B2C(sbi, b - first_block);
3303 ext4_set_bit(c, buf);
3309 if (ext4_bg_has_super(sb, grp)) {
3310 ext4_set_bit(s++, buf);
3313 for (j = ext4_bg_num_gdb(sb, grp); j > 0; j--) {
3314 ext4_set_bit(EXT4_B2C(sbi, s++), buf);
3320 return EXT4_CLUSTERS_PER_GROUP(sb) -
3321 ext4_count_free(buf, EXT4_CLUSTERS_PER_GROUP(sb) / 8);
3325 * Compute the overhead and stash it in sbi->s_overhead
3327 int ext4_calculate_overhead(struct super_block *sb)
3329 struct ext4_sb_info *sbi = EXT4_SB(sb);
3330 struct ext4_super_block *es = sbi->s_es;
3331 ext4_group_t i, ngroups = ext4_get_groups_count(sb);
3332 ext4_fsblk_t overhead = 0;
3333 char *buf = (char *) get_zeroed_page(GFP_KERNEL);
3339 * Compute the overhead (FS structures). This is constant
3340 * for a given filesystem unless the number of block groups
3341 * changes so we cache the previous value until it does.
3345 * All of the blocks before first_data_block are overhead
3347 overhead = EXT4_B2C(sbi, le32_to_cpu(es->s_first_data_block));
3350 * Add the overhead found in each block group
3352 for (i = 0; i < ngroups; i++) {
3355 blks = count_overhead(sb, i, buf);
3358 memset(buf, 0, PAGE_SIZE);
3361 /* Add the journal blocks as well */
3363 overhead += EXT4_NUM_B2C(sbi, sbi->s_journal->j_maxlen);
3365 sbi->s_overhead = overhead;
3367 free_page((unsigned long) buf);
3372 static ext4_fsblk_t ext4_calculate_resv_clusters(struct super_block *sb)
3374 ext4_fsblk_t resv_clusters;
3377 * There's no need to reserve anything when we aren't using extents.
3378 * The space estimates are exact, there are no unwritten extents,
3379 * hole punching doesn't need new metadata... This is needed especially
3380 * to keep ext2/3 backward compatibility.
3382 if (!EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_EXTENTS))
3385 * By default we reserve 2% or 4096 clusters, whichever is smaller.
3386 * This should cover the situations where we can not afford to run
3387 * out of space like for example punch hole, or converting
3388 * unwritten extents in delalloc path. In most cases such
3389 * allocation would require 1, or 2 blocks, higher numbers are
3392 resv_clusters = ext4_blocks_count(EXT4_SB(sb)->s_es) >>
3393 EXT4_SB(sb)->s_cluster_bits;
3395 do_div(resv_clusters, 50);
3396 resv_clusters = min_t(ext4_fsblk_t, resv_clusters, 4096);
3398 return resv_clusters;
3402 static int ext4_reserve_clusters(struct ext4_sb_info *sbi, ext4_fsblk_t count)
3404 ext4_fsblk_t clusters = ext4_blocks_count(sbi->s_es) >>
3405 sbi->s_cluster_bits;
3407 if (count >= clusters)
3410 atomic64_set(&sbi->s_resv_clusters, count);
3414 static int ext4_fill_super(struct super_block *sb, void *data, int silent)
3416 char *orig_data = kstrdup(data, GFP_KERNEL);
3417 struct buffer_head *bh;
3418 struct ext4_super_block *es = NULL;
3419 struct ext4_sb_info *sbi;
3421 ext4_fsblk_t sb_block = get_sb_block(&data);
3422 ext4_fsblk_t logical_sb_block;
3423 unsigned long offset = 0;
3424 unsigned long journal_devnum = 0;
3425 unsigned long def_mount_opts;
3430 int blocksize, clustersize;
3431 unsigned int db_count;
3433 int needs_recovery, has_huge_files, has_bigalloc;
3436 unsigned int journal_ioprio = DEFAULT_JOURNAL_IOPRIO;
3437 ext4_group_t first_not_zeroed;
3439 sbi = kzalloc(sizeof(*sbi), GFP_KERNEL);
3443 sbi->s_blockgroup_lock =
3444 kzalloc(sizeof(struct blockgroup_lock), GFP_KERNEL);
3445 if (!sbi->s_blockgroup_lock) {
3449 sb->s_fs_info = sbi;
3451 sbi->s_inode_readahead_blks = EXT4_DEF_INODE_READAHEAD_BLKS;
3452 sbi->s_sb_block = sb_block;
3453 if (sb->s_bdev->bd_part)
3454 sbi->s_sectors_written_start =
3455 part_stat_read(sb->s_bdev->bd_part, sectors[1]);
3457 /* Cleanup superblock name */
3458 for (cp = sb->s_id; (cp = strchr(cp, '/'));)
3461 /* -EINVAL is default */
3463 blocksize = sb_min_blocksize(sb, EXT4_MIN_BLOCK_SIZE);
3465 ext4_msg(sb, KERN_ERR, "unable to set blocksize");
3470 * The ext4 superblock will not be buffer aligned for other than 1kB
3471 * block sizes. We need to calculate the offset from buffer start.
3473 if (blocksize != EXT4_MIN_BLOCK_SIZE) {
3474 logical_sb_block = sb_block * EXT4_MIN_BLOCK_SIZE;
3475 offset = do_div(logical_sb_block, blocksize);
3477 logical_sb_block = sb_block;
3480 if (!(bh = sb_bread_unmovable(sb, logical_sb_block))) {
3481 ext4_msg(sb, KERN_ERR, "unable to read superblock");
3485 * Note: s_es must be initialized as soon as possible because
3486 * some ext4 macro-instructions depend on its value
3488 es = (struct ext4_super_block *) (bh->b_data + offset);
3490 sb->s_magic = le16_to_cpu(es->s_magic);
3491 if (sb->s_magic != EXT4_SUPER_MAGIC)
3493 sbi->s_kbytes_written = le64_to_cpu(es->s_kbytes_written);
3495 /* Warn if metadata_csum and gdt_csum are both set. */
3496 if (EXT4_HAS_RO_COMPAT_FEATURE(sb,
3497 EXT4_FEATURE_RO_COMPAT_METADATA_CSUM) &&
3498 EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_GDT_CSUM))
3499 ext4_warning(sb, KERN_INFO "metadata_csum and uninit_bg are "
3500 "redundant flags; please run fsck.");
3502 /* Check for a known checksum algorithm */
3503 if (!ext4_verify_csum_type(sb, es)) {
3504 ext4_msg(sb, KERN_ERR, "VFS: Found ext4 filesystem with "
3505 "unknown checksum algorithm.");
3510 /* Load the checksum driver */
3511 if (EXT4_HAS_RO_COMPAT_FEATURE(sb,
3512 EXT4_FEATURE_RO_COMPAT_METADATA_CSUM)) {
3513 sbi->s_chksum_driver = crypto_alloc_shash("crc32c", 0, 0);
3514 if (IS_ERR(sbi->s_chksum_driver)) {
3515 ext4_msg(sb, KERN_ERR, "Cannot load crc32c driver.");
3516 ret = PTR_ERR(sbi->s_chksum_driver);
3517 sbi->s_chksum_driver = NULL;
3522 /* Check superblock checksum */
3523 if (!ext4_superblock_csum_verify(sb, es)) {
3524 ext4_msg(sb, KERN_ERR, "VFS: Found ext4 filesystem with "
3525 "invalid superblock checksum. Run e2fsck?");
3530 /* Precompute checksum seed for all metadata */
3531 if (EXT4_HAS_RO_COMPAT_FEATURE(sb,
3532 EXT4_FEATURE_RO_COMPAT_METADATA_CSUM))
3533 sbi->s_csum_seed = ext4_chksum(sbi, ~0, es->s_uuid,
3534 sizeof(es->s_uuid));
3536 /* Set defaults before we parse the mount options */
3537 def_mount_opts = le32_to_cpu(es->s_default_mount_opts);
3538 set_opt(sb, INIT_INODE_TABLE);
3539 if (def_mount_opts & EXT4_DEFM_DEBUG)
3541 if (def_mount_opts & EXT4_DEFM_BSDGROUPS)
3543 if (def_mount_opts & EXT4_DEFM_UID16)
3544 set_opt(sb, NO_UID32);
3545 /* xattr user namespace & acls are now defaulted on */
3546 set_opt(sb, XATTR_USER);
3547 #ifdef CONFIG_EXT4_FS_POSIX_ACL
3548 set_opt(sb, POSIX_ACL);
3550 if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_DATA)
3551 set_opt(sb, JOURNAL_DATA);
3552 else if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_ORDERED)
3553 set_opt(sb, ORDERED_DATA);
3554 else if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_WBACK)
3555 set_opt(sb, WRITEBACK_DATA);
3557 if (le16_to_cpu(sbi->s_es->s_errors) == EXT4_ERRORS_PANIC)
3558 set_opt(sb, ERRORS_PANIC);
3559 else if (le16_to_cpu(sbi->s_es->s_errors) == EXT4_ERRORS_CONTINUE)
3560 set_opt(sb, ERRORS_CONT);
3562 set_opt(sb, ERRORS_RO);
3563 /* block_validity enabled by default; disable with noblock_validity */
3564 set_opt(sb, BLOCK_VALIDITY);
3565 if (def_mount_opts & EXT4_DEFM_DISCARD)
3566 set_opt(sb, DISCARD);
3568 sbi->s_resuid = make_kuid(&init_user_ns, le16_to_cpu(es->s_def_resuid));
3569 sbi->s_resgid = make_kgid(&init_user_ns, le16_to_cpu(es->s_def_resgid));
3570 sbi->s_commit_interval = JBD2_DEFAULT_MAX_COMMIT_AGE * HZ;
3571 sbi->s_min_batch_time = EXT4_DEF_MIN_BATCH_TIME;
3572 sbi->s_max_batch_time = EXT4_DEF_MAX_BATCH_TIME;
3574 if ((def_mount_opts & EXT4_DEFM_NOBARRIER) == 0)
3575 set_opt(sb, BARRIER);
3578 * enable delayed allocation by default
3579 * Use -o nodelalloc to turn it off
3581 if (!IS_EXT3_SB(sb) && !IS_EXT2_SB(sb) &&
3582 ((def_mount_opts & EXT4_DEFM_NODELALLOC) == 0))
3583 set_opt(sb, DELALLOC);
3586 * set default s_li_wait_mult for lazyinit, for the case there is
3587 * no mount option specified.
3589 sbi->s_li_wait_mult = EXT4_DEF_LI_WAIT_MULT;
3591 if (!parse_options((char *) sbi->s_es->s_mount_opts, sb,
3592 &journal_devnum, &journal_ioprio, 0)) {
3593 ext4_msg(sb, KERN_WARNING,
3594 "failed to parse options in superblock: %s",
3595 sbi->s_es->s_mount_opts);
3597 sbi->s_def_mount_opt = sbi->s_mount_opt;
3598 if (!parse_options((char *) data, sb, &journal_devnum,
3599 &journal_ioprio, 0))
3602 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA) {
3603 printk_once(KERN_WARNING "EXT4-fs: Warning: mounting "
3604 "with data=journal disables delayed "
3605 "allocation and O_DIRECT support!\n");
3606 if (test_opt2(sb, EXPLICIT_DELALLOC)) {
3607 ext4_msg(sb, KERN_ERR, "can't mount with "
3608 "both data=journal and delalloc");
3611 if (test_opt(sb, DIOREAD_NOLOCK)) {
3612 ext4_msg(sb, KERN_ERR, "can't mount with "
3613 "both data=journal and dioread_nolock");
3616 if (test_opt(sb, DELALLOC))
3617 clear_opt(sb, DELALLOC);
3620 sb->s_flags = (sb->s_flags & ~MS_POSIXACL) |
3621 (test_opt(sb, POSIX_ACL) ? MS_POSIXACL : 0);
3623 if (le32_to_cpu(es->s_rev_level) == EXT4_GOOD_OLD_REV &&
3624 (EXT4_HAS_COMPAT_FEATURE(sb, ~0U) ||
3625 EXT4_HAS_RO_COMPAT_FEATURE(sb, ~0U) ||
3626 EXT4_HAS_INCOMPAT_FEATURE(sb, ~0U)))
3627 ext4_msg(sb, KERN_WARNING,
3628 "feature flags set on rev 0 fs, "
3629 "running e2fsck is recommended");
3631 if (es->s_creator_os == cpu_to_le32(EXT4_OS_HURD)) {
3632 set_opt2(sb, HURD_COMPAT);
3633 if (EXT4_HAS_INCOMPAT_FEATURE(sb,
3634 EXT4_FEATURE_INCOMPAT_64BIT)) {
3635 ext4_msg(sb, KERN_ERR,
3636 "The Hurd can't support 64-bit file systems");
3641 if (IS_EXT2_SB(sb)) {
3642 if (ext2_feature_set_ok(sb))
3643 ext4_msg(sb, KERN_INFO, "mounting ext2 file system "
3644 "using the ext4 subsystem");
3646 ext4_msg(sb, KERN_ERR, "couldn't mount as ext2 due "
3647 "to feature incompatibilities");
3652 if (IS_EXT3_SB(sb)) {
3653 if (ext3_feature_set_ok(sb))
3654 ext4_msg(sb, KERN_INFO, "mounting ext3 file system "
3655 "using the ext4 subsystem");
3657 ext4_msg(sb, KERN_ERR, "couldn't mount as ext3 due "
3658 "to feature incompatibilities");
3664 * Check feature flags regardless of the revision level, since we
3665 * previously didn't change the revision level when setting the flags,
3666 * so there is a chance incompat flags are set on a rev 0 filesystem.
3668 if (!ext4_feature_set_ok(sb, (sb->s_flags & MS_RDONLY)))
3671 blocksize = BLOCK_SIZE << le32_to_cpu(es->s_log_block_size);
3672 if (blocksize < EXT4_MIN_BLOCK_SIZE ||
3673 blocksize > EXT4_MAX_BLOCK_SIZE) {
3674 ext4_msg(sb, KERN_ERR,
3675 "Unsupported filesystem blocksize %d", blocksize);
3679 if (sb->s_blocksize != blocksize) {
3680 /* Validate the filesystem blocksize */
3681 if (!sb_set_blocksize(sb, blocksize)) {
3682 ext4_msg(sb, KERN_ERR, "bad block size %d",
3688 logical_sb_block = sb_block * EXT4_MIN_BLOCK_SIZE;
3689 offset = do_div(logical_sb_block, blocksize);
3690 bh = sb_bread_unmovable(sb, logical_sb_block);
3692 ext4_msg(sb, KERN_ERR,
3693 "Can't read superblock on 2nd try");
3696 es = (struct ext4_super_block *)(bh->b_data + offset);
3698 if (es->s_magic != cpu_to_le16(EXT4_SUPER_MAGIC)) {
3699 ext4_msg(sb, KERN_ERR,
3700 "Magic mismatch, very weird!");
3705 has_huge_files = EXT4_HAS_RO_COMPAT_FEATURE(sb,
3706 EXT4_FEATURE_RO_COMPAT_HUGE_FILE);
3707 sbi->s_bitmap_maxbytes = ext4_max_bitmap_size(sb->s_blocksize_bits,
3709 sb->s_maxbytes = ext4_max_size(sb->s_blocksize_bits, has_huge_files);
3711 if (le32_to_cpu(es->s_rev_level) == EXT4_GOOD_OLD_REV) {
3712 sbi->s_inode_size = EXT4_GOOD_OLD_INODE_SIZE;
3713 sbi->s_first_ino = EXT4_GOOD_OLD_FIRST_INO;
3715 sbi->s_inode_size = le16_to_cpu(es->s_inode_size);
3716 sbi->s_first_ino = le32_to_cpu(es->s_first_ino);
3717 if ((sbi->s_inode_size < EXT4_GOOD_OLD_INODE_SIZE) ||
3718 (!is_power_of_2(sbi->s_inode_size)) ||
3719 (sbi->s_inode_size > blocksize)) {
3720 ext4_msg(sb, KERN_ERR,
3721 "unsupported inode size: %d",
3725 if (sbi->s_inode_size > EXT4_GOOD_OLD_INODE_SIZE)
3726 sb->s_time_gran = 1 << (EXT4_EPOCH_BITS - 2);
3729 sbi->s_desc_size = le16_to_cpu(es->s_desc_size);
3730 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_64BIT)) {
3731 if (sbi->s_desc_size < EXT4_MIN_DESC_SIZE_64BIT ||
3732 sbi->s_desc_size > EXT4_MAX_DESC_SIZE ||
3733 !is_power_of_2(sbi->s_desc_size)) {
3734 ext4_msg(sb, KERN_ERR,
3735 "unsupported descriptor size %lu",
3740 sbi->s_desc_size = EXT4_MIN_DESC_SIZE;
3742 sbi->s_blocks_per_group = le32_to_cpu(es->s_blocks_per_group);
3743 sbi->s_inodes_per_group = le32_to_cpu(es->s_inodes_per_group);
3744 if (EXT4_INODE_SIZE(sb) == 0 || EXT4_INODES_PER_GROUP(sb) == 0)
3747 sbi->s_inodes_per_block = blocksize / EXT4_INODE_SIZE(sb);
3748 if (sbi->s_inodes_per_block == 0)
3750 sbi->s_itb_per_group = sbi->s_inodes_per_group /
3751 sbi->s_inodes_per_block;
3752 sbi->s_desc_per_block = blocksize / EXT4_DESC_SIZE(sb);
3754 sbi->s_mount_state = le16_to_cpu(es->s_state);
3755 sbi->s_addr_per_block_bits = ilog2(EXT4_ADDR_PER_BLOCK(sb));
3756 sbi->s_desc_per_block_bits = ilog2(EXT4_DESC_PER_BLOCK(sb));
3758 for (i = 0; i < 4; i++)
3759 sbi->s_hash_seed[i] = le32_to_cpu(es->s_hash_seed[i]);
3760 sbi->s_def_hash_version = es->s_def_hash_version;
3761 if (EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_DIR_INDEX)) {
3762 i = le32_to_cpu(es->s_flags);
3763 if (i & EXT2_FLAGS_UNSIGNED_HASH)
3764 sbi->s_hash_unsigned = 3;
3765 else if ((i & EXT2_FLAGS_SIGNED_HASH) == 0) {
3766 #ifdef __CHAR_UNSIGNED__
3767 if (!(sb->s_flags & MS_RDONLY))
3769 cpu_to_le32(EXT2_FLAGS_UNSIGNED_HASH);
3770 sbi->s_hash_unsigned = 3;
3772 if (!(sb->s_flags & MS_RDONLY))
3774 cpu_to_le32(EXT2_FLAGS_SIGNED_HASH);
3779 /* Handle clustersize */
3780 clustersize = BLOCK_SIZE << le32_to_cpu(es->s_log_cluster_size);
3781 has_bigalloc = EXT4_HAS_RO_COMPAT_FEATURE(sb,
3782 EXT4_FEATURE_RO_COMPAT_BIGALLOC);
3784 if (clustersize < blocksize) {
3785 ext4_msg(sb, KERN_ERR,
3786 "cluster size (%d) smaller than "
3787 "block size (%d)", clustersize, blocksize);
3790 sbi->s_cluster_bits = le32_to_cpu(es->s_log_cluster_size) -
3791 le32_to_cpu(es->s_log_block_size);
3792 sbi->s_clusters_per_group =
3793 le32_to_cpu(es->s_clusters_per_group);
3794 if (sbi->s_clusters_per_group > blocksize * 8) {
3795 ext4_msg(sb, KERN_ERR,
3796 "#clusters per group too big: %lu",
3797 sbi->s_clusters_per_group);
3800 if (sbi->s_blocks_per_group !=
3801 (sbi->s_clusters_per_group * (clustersize / blocksize))) {
3802 ext4_msg(sb, KERN_ERR, "blocks per group (%lu) and "
3803 "clusters per group (%lu) inconsistent",
3804 sbi->s_blocks_per_group,
3805 sbi->s_clusters_per_group);
3809 if (clustersize != blocksize) {
3810 ext4_warning(sb, "fragment/cluster size (%d) != "
3811 "block size (%d)", clustersize,
3813 clustersize = blocksize;
3815 if (sbi->s_blocks_per_group > blocksize * 8) {
3816 ext4_msg(sb, KERN_ERR,
3817 "#blocks per group too big: %lu",
3818 sbi->s_blocks_per_group);
3821 sbi->s_clusters_per_group = sbi->s_blocks_per_group;
3822 sbi->s_cluster_bits = 0;
3824 sbi->s_cluster_ratio = clustersize / blocksize;
3826 if (sbi->s_inodes_per_group > blocksize * 8) {
3827 ext4_msg(sb, KERN_ERR,
3828 "#inodes per group too big: %lu",
3829 sbi->s_inodes_per_group);
3833 /* Do we have standard group size of clustersize * 8 blocks ? */
3834 if (sbi->s_blocks_per_group == clustersize << 3)
3835 set_opt2(sb, STD_GROUP_SIZE);
3838 * Test whether we have more sectors than will fit in sector_t,
3839 * and whether the max offset is addressable by the page cache.
3841 err = generic_check_addressable(sb->s_blocksize_bits,
3842 ext4_blocks_count(es));
3844 ext4_msg(sb, KERN_ERR, "filesystem"
3845 " too large to mount safely on this system");
3846 if (sizeof(sector_t) < 8)
3847 ext4_msg(sb, KERN_WARNING, "CONFIG_LBDAF not enabled");
3851 if (EXT4_BLOCKS_PER_GROUP(sb) == 0)
3854 /* check blocks count against device size */
3855 blocks_count = sb->s_bdev->bd_inode->i_size >> sb->s_blocksize_bits;
3856 if (blocks_count && ext4_blocks_count(es) > blocks_count) {
3857 ext4_msg(sb, KERN_WARNING, "bad geometry: block count %llu "
3858 "exceeds size of device (%llu blocks)",
3859 ext4_blocks_count(es), blocks_count);
3864 * It makes no sense for the first data block to be beyond the end
3865 * of the filesystem.
3867 if (le32_to_cpu(es->s_first_data_block) >= ext4_blocks_count(es)) {
3868 ext4_msg(sb, KERN_WARNING, "bad geometry: first data "
3869 "block %u is beyond end of filesystem (%llu)",
3870 le32_to_cpu(es->s_first_data_block),
3871 ext4_blocks_count(es));
3874 blocks_count = (ext4_blocks_count(es) -
3875 le32_to_cpu(es->s_first_data_block) +
3876 EXT4_BLOCKS_PER_GROUP(sb) - 1);
3877 do_div(blocks_count, EXT4_BLOCKS_PER_GROUP(sb));
3878 if (blocks_count > ((uint64_t)1<<32) - EXT4_DESC_PER_BLOCK(sb)) {
3879 ext4_msg(sb, KERN_WARNING, "groups count too large: %u "
3880 "(block count %llu, first data block %u, "
3881 "blocks per group %lu)", sbi->s_groups_count,
3882 ext4_blocks_count(es),
3883 le32_to_cpu(es->s_first_data_block),
3884 EXT4_BLOCKS_PER_GROUP(sb));
3887 sbi->s_groups_count = blocks_count;
3888 sbi->s_blockfile_groups = min_t(ext4_group_t, sbi->s_groups_count,
3889 (EXT4_MAX_BLOCK_FILE_PHYS / EXT4_BLOCKS_PER_GROUP(sb)));
3890 db_count = (sbi->s_groups_count + EXT4_DESC_PER_BLOCK(sb) - 1) /
3891 EXT4_DESC_PER_BLOCK(sb);
3892 sbi->s_group_desc = ext4_kvmalloc(db_count *
3893 sizeof(struct buffer_head *),
3895 if (sbi->s_group_desc == NULL) {
3896 ext4_msg(sb, KERN_ERR, "not enough memory");
3902 sbi->s_proc = proc_mkdir(sb->s_id, ext4_proc_root);
3905 proc_create_data("options", S_IRUGO, sbi->s_proc,
3906 &ext4_seq_options_fops, sb);
3908 bgl_lock_init(sbi->s_blockgroup_lock);
3910 for (i = 0; i < db_count; i++) {
3911 block = descriptor_loc(sb, logical_sb_block, i);
3912 sbi->s_group_desc[i] = sb_bread_unmovable(sb, block);
3913 if (!sbi->s_group_desc[i]) {
3914 ext4_msg(sb, KERN_ERR,
3915 "can't read group descriptor %d", i);
3920 if (!ext4_check_descriptors(sb, &first_not_zeroed)) {
3921 ext4_msg(sb, KERN_ERR, "group descriptors corrupted!");
3925 sbi->s_gdb_count = db_count;
3926 get_random_bytes(&sbi->s_next_generation, sizeof(u32));
3927 spin_lock_init(&sbi->s_next_gen_lock);
3929 init_timer(&sbi->s_err_report);
3930 sbi->s_err_report.function = print_daily_error_info;
3931 sbi->s_err_report.data = (unsigned long) sb;
3933 /* Register extent status tree shrinker */
3934 if (ext4_es_register_shrinker(sbi))
3937 sbi->s_stripe = ext4_get_stripe_size(sbi);
3938 sbi->s_extent_max_zeroout_kb = 32;
3941 * set up enough so that it can read an inode
3943 if (!test_opt(sb, NOLOAD) &&
3944 EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL))
3945 sb->s_op = &ext4_sops;
3947 sb->s_op = &ext4_nojournal_sops;
3948 sb->s_export_op = &ext4_export_ops;
3949 sb->s_xattr = ext4_xattr_handlers;
3951 sb->dq_op = &ext4_quota_operations;
3952 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_QUOTA))
3953 sb->s_qcop = &ext4_qctl_sysfile_operations;
3955 sb->s_qcop = &ext4_qctl_operations;
3957 memcpy(sb->s_uuid, es->s_uuid, sizeof(es->s_uuid));
3959 INIT_LIST_HEAD(&sbi->s_orphan); /* unlinked but open files */
3960 mutex_init(&sbi->s_orphan_lock);
3964 needs_recovery = (es->s_last_orphan != 0 ||
3965 EXT4_HAS_INCOMPAT_FEATURE(sb,
3966 EXT4_FEATURE_INCOMPAT_RECOVER));
3968 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_MMP) &&
3969 !(sb->s_flags & MS_RDONLY))
3970 if (ext4_multi_mount_protect(sb, le64_to_cpu(es->s_mmp_block)))
3974 * The first inode we look at is the journal inode. Don't try
3975 * root first: it may be modified in the journal!
3977 if (!test_opt(sb, NOLOAD) &&
3978 EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL)) {
3979 if (ext4_load_journal(sb, es, journal_devnum))
3981 } else if (test_opt(sb, NOLOAD) && !(sb->s_flags & MS_RDONLY) &&
3982 EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER)) {
3983 ext4_msg(sb, KERN_ERR, "required journal recovery "
3984 "suppressed and not mounted read-only");
3985 goto failed_mount_wq;
3987 clear_opt(sb, DATA_FLAGS);
3988 sbi->s_journal = NULL;
3993 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_64BIT) &&
3994 !jbd2_journal_set_features(EXT4_SB(sb)->s_journal, 0, 0,
3995 JBD2_FEATURE_INCOMPAT_64BIT)) {
3996 ext4_msg(sb, KERN_ERR, "Failed to set 64-bit journal feature");
3997 goto failed_mount_wq;
4000 if (!set_journal_csum_feature_set(sb)) {
4001 ext4_msg(sb, KERN_ERR, "Failed to set journal checksum "
4003 goto failed_mount_wq;
4006 /* We have now updated the journal if required, so we can
4007 * validate the data journaling mode. */
4008 switch (test_opt(sb, DATA_FLAGS)) {
4010 /* No mode set, assume a default based on the journal
4011 * capabilities: ORDERED_DATA if the journal can
4012 * cope, else JOURNAL_DATA
4014 if (jbd2_journal_check_available_features
4015 (sbi->s_journal, 0, 0, JBD2_FEATURE_INCOMPAT_REVOKE))
4016 set_opt(sb, ORDERED_DATA);
4018 set_opt(sb, JOURNAL_DATA);
4021 case EXT4_MOUNT_ORDERED_DATA:
4022 case EXT4_MOUNT_WRITEBACK_DATA:
4023 if (!jbd2_journal_check_available_features
4024 (sbi->s_journal, 0, 0, JBD2_FEATURE_INCOMPAT_REVOKE)) {
4025 ext4_msg(sb, KERN_ERR, "Journal does not support "
4026 "requested data journaling mode");
4027 goto failed_mount_wq;
4032 set_task_ioprio(sbi->s_journal->j_task, journal_ioprio);
4034 sbi->s_journal->j_commit_callback = ext4_journal_commit_callback;
4037 if (ext4_mballoc_ready) {
4038 sbi->s_mb_cache = ext4_xattr_create_cache(sb->s_id);
4039 if (!sbi->s_mb_cache) {
4040 ext4_msg(sb, KERN_ERR, "Failed to create an mb_cache");
4041 goto failed_mount_wq;
4046 * Get the # of file system overhead blocks from the
4047 * superblock if present.
4049 if (es->s_overhead_clusters)
4050 sbi->s_overhead = le32_to_cpu(es->s_overhead_clusters);
4052 err = ext4_calculate_overhead(sb);
4054 goto failed_mount_wq;
4058 * The maximum number of concurrent works can be high and
4059 * concurrency isn't really necessary. Limit it to 1.
4061 EXT4_SB(sb)->rsv_conversion_wq =
4062 alloc_workqueue("ext4-rsv-conversion", WQ_MEM_RECLAIM | WQ_UNBOUND, 1);
4063 if (!EXT4_SB(sb)->rsv_conversion_wq) {
4064 printk(KERN_ERR "EXT4-fs: failed to create workqueue\n");
4070 * The jbd2_journal_load will have done any necessary log recovery,
4071 * so we can safely mount the rest of the filesystem now.
4074 root = ext4_iget(sb, EXT4_ROOT_INO);
4076 ext4_msg(sb, KERN_ERR, "get root inode failed");
4077 ret = PTR_ERR(root);
4081 if (!S_ISDIR(root->i_mode) || !root->i_blocks || !root->i_size) {
4082 ext4_msg(sb, KERN_ERR, "corrupt root inode, run e2fsck");
4086 sb->s_root = d_make_root(root);
4088 ext4_msg(sb, KERN_ERR, "get root dentry failed");
4093 if (ext4_setup_super(sb, es, sb->s_flags & MS_RDONLY))
4094 sb->s_flags |= MS_RDONLY;
4096 /* determine the minimum size of new large inodes, if present */
4097 if (sbi->s_inode_size > EXT4_GOOD_OLD_INODE_SIZE) {
4098 sbi->s_want_extra_isize = sizeof(struct ext4_inode) -
4099 EXT4_GOOD_OLD_INODE_SIZE;
4100 if (EXT4_HAS_RO_COMPAT_FEATURE(sb,
4101 EXT4_FEATURE_RO_COMPAT_EXTRA_ISIZE)) {
4102 if (sbi->s_want_extra_isize <
4103 le16_to_cpu(es->s_want_extra_isize))
4104 sbi->s_want_extra_isize =
4105 le16_to_cpu(es->s_want_extra_isize);
4106 if (sbi->s_want_extra_isize <
4107 le16_to_cpu(es->s_min_extra_isize))
4108 sbi->s_want_extra_isize =
4109 le16_to_cpu(es->s_min_extra_isize);
4112 /* Check if enough inode space is available */
4113 if (EXT4_GOOD_OLD_INODE_SIZE + sbi->s_want_extra_isize >
4114 sbi->s_inode_size) {
4115 sbi->s_want_extra_isize = sizeof(struct ext4_inode) -
4116 EXT4_GOOD_OLD_INODE_SIZE;
4117 ext4_msg(sb, KERN_INFO, "required extra inode space not"
4121 err = ext4_reserve_clusters(sbi, ext4_calculate_resv_clusters(sb));
4123 ext4_msg(sb, KERN_ERR, "failed to reserve %llu clusters for "
4124 "reserved pool", ext4_calculate_resv_clusters(sb));
4125 goto failed_mount4a;
4128 err = ext4_setup_system_zone(sb);
4130 ext4_msg(sb, KERN_ERR, "failed to initialize system "
4132 goto failed_mount4a;
4136 err = ext4_mb_init(sb);
4138 ext4_msg(sb, KERN_ERR, "failed to initialize mballoc (%d)",
4143 block = ext4_count_free_clusters(sb);
4144 ext4_free_blocks_count_set(sbi->s_es,
4145 EXT4_C2B(sbi, block));
4146 err = percpu_counter_init(&sbi->s_freeclusters_counter, block);
4148 unsigned long freei = ext4_count_free_inodes(sb);
4149 sbi->s_es->s_free_inodes_count = cpu_to_le32(freei);
4150 err = percpu_counter_init(&sbi->s_freeinodes_counter, freei);
4153 err = percpu_counter_init(&sbi->s_dirs_counter,
4154 ext4_count_dirs(sb));
4156 err = percpu_counter_init(&sbi->s_dirtyclusters_counter, 0);
4158 ext4_msg(sb, KERN_ERR, "insufficient memory");
4162 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_FLEX_BG))
4163 if (!ext4_fill_flex_info(sb)) {
4164 ext4_msg(sb, KERN_ERR,
4165 "unable to initialize "
4166 "flex_bg meta info!");
4170 err = ext4_register_li_request(sb, first_not_zeroed);
4174 sbi->s_kobj.kset = ext4_kset;
4175 init_completion(&sbi->s_kobj_unregister);
4176 err = kobject_init_and_add(&sbi->s_kobj, &ext4_ktype, NULL,
4182 /* Enable quota usage during mount. */
4183 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_QUOTA) &&
4184 !(sb->s_flags & MS_RDONLY)) {
4185 err = ext4_enable_quotas(sb);
4189 #endif /* CONFIG_QUOTA */
4191 EXT4_SB(sb)->s_mount_state |= EXT4_ORPHAN_FS;
4192 ext4_orphan_cleanup(sb, es);
4193 EXT4_SB(sb)->s_mount_state &= ~EXT4_ORPHAN_FS;
4194 if (needs_recovery) {
4195 ext4_msg(sb, KERN_INFO, "recovery complete");
4196 ext4_mark_recovery_complete(sb, es);
4198 if (EXT4_SB(sb)->s_journal) {
4199 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA)
4200 descr = " journalled data mode";
4201 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA)
4202 descr = " ordered data mode";
4204 descr = " writeback data mode";
4206 descr = "out journal";
4208 if (test_opt(sb, DISCARD)) {
4209 struct request_queue *q = bdev_get_queue(sb->s_bdev);
4210 if (!blk_queue_discard(q))
4211 ext4_msg(sb, KERN_WARNING,
4212 "mounting with \"discard\" option, but "
4213 "the device does not support discard");
4216 ext4_msg(sb, KERN_INFO, "mounted filesystem with%s. "
4217 "Opts: %s%s%s", descr, sbi->s_es->s_mount_opts,
4218 *sbi->s_es->s_mount_opts ? "; " : "", orig_data);
4220 if (es->s_error_count)
4221 mod_timer(&sbi->s_err_report, jiffies + 300*HZ); /* 5 minutes */
4223 /* Enable message ratelimiting. Default is 10 messages per 5 secs. */
4224 ratelimit_state_init(&sbi->s_err_ratelimit_state, 5 * HZ, 10);
4225 ratelimit_state_init(&sbi->s_warning_ratelimit_state, 5 * HZ, 10);
4226 ratelimit_state_init(&sbi->s_msg_ratelimit_state, 5 * HZ, 10);
4233 ext4_msg(sb, KERN_ERR, "VFS: Can't find ext4 filesystem");
4238 kobject_del(&sbi->s_kobj);
4241 ext4_unregister_li_request(sb);
4243 ext4_mb_release(sb);
4244 if (sbi->s_flex_groups)
4245 ext4_kvfree(sbi->s_flex_groups);
4246 percpu_counter_destroy(&sbi->s_freeclusters_counter);
4247 percpu_counter_destroy(&sbi->s_freeinodes_counter);
4248 percpu_counter_destroy(&sbi->s_dirs_counter);
4249 percpu_counter_destroy(&sbi->s_dirtyclusters_counter);
4251 ext4_ext_release(sb);
4252 ext4_release_system_zone(sb);
4257 ext4_msg(sb, KERN_ERR, "mount failed");
4258 if (EXT4_SB(sb)->rsv_conversion_wq)
4259 destroy_workqueue(EXT4_SB(sb)->rsv_conversion_wq);
4261 if (sbi->s_journal) {
4262 jbd2_journal_destroy(sbi->s_journal);
4263 sbi->s_journal = NULL;
4265 ext4_es_unregister_shrinker(sbi);
4267 del_timer_sync(&sbi->s_err_report);
4269 kthread_stop(sbi->s_mmp_tsk);
4271 for (i = 0; i < db_count; i++)
4272 brelse(sbi->s_group_desc[i]);
4273 ext4_kvfree(sbi->s_group_desc);
4275 if (sbi->s_chksum_driver)
4276 crypto_free_shash(sbi->s_chksum_driver);
4278 remove_proc_entry("options", sbi->s_proc);
4279 remove_proc_entry(sb->s_id, ext4_proc_root);
4282 for (i = 0; i < EXT4_MAXQUOTAS; i++)
4283 kfree(sbi->s_qf_names[i]);
4285 ext4_blkdev_remove(sbi);
4288 sb->s_fs_info = NULL;
4289 kfree(sbi->s_blockgroup_lock);
4293 return err ? err : ret;
4297 * Setup any per-fs journal parameters now. We'll do this both on
4298 * initial mount, once the journal has been initialised but before we've
4299 * done any recovery; and again on any subsequent remount.
4301 static void ext4_init_journal_params(struct super_block *sb, journal_t *journal)
4303 struct ext4_sb_info *sbi = EXT4_SB(sb);
4305 journal->j_commit_interval = sbi->s_commit_interval;
4306 journal->j_min_batch_time = sbi->s_min_batch_time;
4307 journal->j_max_batch_time = sbi->s_max_batch_time;
4309 write_lock(&journal->j_state_lock);
4310 if (test_opt(sb, BARRIER))
4311 journal->j_flags |= JBD2_BARRIER;
4313 journal->j_flags &= ~JBD2_BARRIER;
4314 if (test_opt(sb, DATA_ERR_ABORT))
4315 journal->j_flags |= JBD2_ABORT_ON_SYNCDATA_ERR;
4317 journal->j_flags &= ~JBD2_ABORT_ON_SYNCDATA_ERR;
4318 write_unlock(&journal->j_state_lock);
4321 static journal_t *ext4_get_journal(struct super_block *sb,
4322 unsigned int journal_inum)
4324 struct inode *journal_inode;
4327 BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL));
4329 /* First, test for the existence of a valid inode on disk. Bad
4330 * things happen if we iget() an unused inode, as the subsequent
4331 * iput() will try to delete it. */
4333 journal_inode = ext4_iget(sb, journal_inum);
4334 if (IS_ERR(journal_inode)) {
4335 ext4_msg(sb, KERN_ERR, "no journal found");
4338 if (!journal_inode->i_nlink) {
4339 make_bad_inode(journal_inode);
4340 iput(journal_inode);
4341 ext4_msg(sb, KERN_ERR, "journal inode is deleted");
4345 jbd_debug(2, "Journal inode found at %p: %lld bytes\n",
4346 journal_inode, journal_inode->i_size);
4347 if (!S_ISREG(journal_inode->i_mode)) {
4348 ext4_msg(sb, KERN_ERR, "invalid journal inode");
4349 iput(journal_inode);
4353 journal = jbd2_journal_init_inode(journal_inode);
4355 ext4_msg(sb, KERN_ERR, "Could not load journal inode");
4356 iput(journal_inode);
4359 journal->j_private = sb;
4360 ext4_init_journal_params(sb, journal);
4364 static journal_t *ext4_get_dev_journal(struct super_block *sb,
4367 struct buffer_head *bh;
4371 int hblock, blocksize;
4372 ext4_fsblk_t sb_block;
4373 unsigned long offset;
4374 struct ext4_super_block *es;
4375 struct block_device *bdev;
4377 BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL));
4379 bdev = ext4_blkdev_get(j_dev, sb);
4383 blocksize = sb->s_blocksize;
4384 hblock = bdev_logical_block_size(bdev);
4385 if (blocksize < hblock) {
4386 ext4_msg(sb, KERN_ERR,
4387 "blocksize too small for journal device");
4391 sb_block = EXT4_MIN_BLOCK_SIZE / blocksize;
4392 offset = EXT4_MIN_BLOCK_SIZE % blocksize;
4393 set_blocksize(bdev, blocksize);
4394 if (!(bh = __bread(bdev, sb_block, blocksize))) {
4395 ext4_msg(sb, KERN_ERR, "couldn't read superblock of "
4396 "external journal");
4400 es = (struct ext4_super_block *) (bh->b_data + offset);
4401 if ((le16_to_cpu(es->s_magic) != EXT4_SUPER_MAGIC) ||
4402 !(le32_to_cpu(es->s_feature_incompat) &
4403 EXT4_FEATURE_INCOMPAT_JOURNAL_DEV)) {
4404 ext4_msg(sb, KERN_ERR, "external journal has "
4410 if ((le32_to_cpu(es->s_feature_ro_compat) &
4411 EXT4_FEATURE_RO_COMPAT_METADATA_CSUM) &&
4412 es->s_checksum != ext4_superblock_csum(sb, es)) {
4413 ext4_msg(sb, KERN_ERR, "external journal has "
4414 "corrupt superblock");
4419 if (memcmp(EXT4_SB(sb)->s_es->s_journal_uuid, es->s_uuid, 16)) {
4420 ext4_msg(sb, KERN_ERR, "journal UUID does not match");
4425 len = ext4_blocks_count(es);
4426 start = sb_block + 1;
4427 brelse(bh); /* we're done with the superblock */
4429 journal = jbd2_journal_init_dev(bdev, sb->s_bdev,
4430 start, len, blocksize);
4432 ext4_msg(sb, KERN_ERR, "failed to create device journal");
4435 journal->j_private = sb;
4436 ll_rw_block(READ | REQ_META | REQ_PRIO, 1, &journal->j_sb_buffer);
4437 wait_on_buffer(journal->j_sb_buffer);
4438 if (!buffer_uptodate(journal->j_sb_buffer)) {
4439 ext4_msg(sb, KERN_ERR, "I/O error on journal device");
4442 if (be32_to_cpu(journal->j_superblock->s_nr_users) != 1) {
4443 ext4_msg(sb, KERN_ERR, "External journal has more than one "
4444 "user (unsupported) - %d",
4445 be32_to_cpu(journal->j_superblock->s_nr_users));
4448 EXT4_SB(sb)->journal_bdev = bdev;
4449 ext4_init_journal_params(sb, journal);
4453 jbd2_journal_destroy(journal);
4455 ext4_blkdev_put(bdev);
4459 static int ext4_load_journal(struct super_block *sb,
4460 struct ext4_super_block *es,
4461 unsigned long journal_devnum)
4464 unsigned int journal_inum = le32_to_cpu(es->s_journal_inum);
4467 int really_read_only;
4469 BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL));
4471 if (journal_devnum &&
4472 journal_devnum != le32_to_cpu(es->s_journal_dev)) {
4473 ext4_msg(sb, KERN_INFO, "external journal device major/minor "
4474 "numbers have changed");
4475 journal_dev = new_decode_dev(journal_devnum);
4477 journal_dev = new_decode_dev(le32_to_cpu(es->s_journal_dev));
4479 really_read_only = bdev_read_only(sb->s_bdev);
4482 * Are we loading a blank journal or performing recovery after a
4483 * crash? For recovery, we need to check in advance whether we
4484 * can get read-write access to the device.
4486 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER)) {
4487 if (sb->s_flags & MS_RDONLY) {
4488 ext4_msg(sb, KERN_INFO, "INFO: recovery "
4489 "required on readonly filesystem");
4490 if (really_read_only) {
4491 ext4_msg(sb, KERN_ERR, "write access "
4492 "unavailable, cannot proceed");
4495 ext4_msg(sb, KERN_INFO, "write access will "
4496 "be enabled during recovery");
4500 if (journal_inum && journal_dev) {
4501 ext4_msg(sb, KERN_ERR, "filesystem has both journal "
4502 "and inode journals!");
4507 if (!(journal = ext4_get_journal(sb, journal_inum)))
4510 if (!(journal = ext4_get_dev_journal(sb, journal_dev)))
4514 if (!(journal->j_flags & JBD2_BARRIER))
4515 ext4_msg(sb, KERN_INFO, "barriers disabled");
4517 if (!EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER))
4518 err = jbd2_journal_wipe(journal, !really_read_only);
4520 char *save = kmalloc(EXT4_S_ERR_LEN, GFP_KERNEL);
4522 memcpy(save, ((char *) es) +
4523 EXT4_S_ERR_START, EXT4_S_ERR_LEN);
4524 err = jbd2_journal_load(journal);
4526 memcpy(((char *) es) + EXT4_S_ERR_START,
4527 save, EXT4_S_ERR_LEN);
4532 ext4_msg(sb, KERN_ERR, "error loading journal");
4533 jbd2_journal_destroy(journal);
4537 EXT4_SB(sb)->s_journal = journal;
4538 ext4_clear_journal_err(sb, es);
4540 if (!really_read_only && journal_devnum &&
4541 journal_devnum != le32_to_cpu(es->s_journal_dev)) {
4542 es->s_journal_dev = cpu_to_le32(journal_devnum);
4544 /* Make sure we flush the recovery flag to disk. */
4545 ext4_commit_super(sb, 1);
4551 static int ext4_commit_super(struct super_block *sb, int sync)
4553 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
4554 struct buffer_head *sbh = EXT4_SB(sb)->s_sbh;
4557 if (!sbh || block_device_ejected(sb))
4559 if (buffer_write_io_error(sbh)) {
4561 * Oh, dear. A previous attempt to write the
4562 * superblock failed. This could happen because the
4563 * USB device was yanked out. Or it could happen to
4564 * be a transient write error and maybe the block will
4565 * be remapped. Nothing we can do but to retry the
4566 * write and hope for the best.
4568 ext4_msg(sb, KERN_ERR, "previous I/O error to "
4569 "superblock detected");
4570 clear_buffer_write_io_error(sbh);
4571 set_buffer_uptodate(sbh);
4574 * If the file system is mounted read-only, don't update the
4575 * superblock write time. This avoids updating the superblock
4576 * write time when we are mounting the root file system
4577 * read/only but we need to replay the journal; at that point,
4578 * for people who are east of GMT and who make their clock
4579 * tick in localtime for Windows bug-for-bug compatibility,
4580 * the clock is set in the future, and this will cause e2fsck
4581 * to complain and force a full file system check.
4583 if (!(sb->s_flags & MS_RDONLY))
4584 es->s_wtime = cpu_to_le32(get_seconds());
4585 if (sb->s_bdev->bd_part)
4586 es->s_kbytes_written =
4587 cpu_to_le64(EXT4_SB(sb)->s_kbytes_written +
4588 ((part_stat_read(sb->s_bdev->bd_part, sectors[1]) -
4589 EXT4_SB(sb)->s_sectors_written_start) >> 1));
4591 es->s_kbytes_written =
4592 cpu_to_le64(EXT4_SB(sb)->s_kbytes_written);
4593 if (percpu_counter_initialized(&EXT4_SB(sb)->s_freeclusters_counter))
4594 ext4_free_blocks_count_set(es,
4595 EXT4_C2B(EXT4_SB(sb), percpu_counter_sum_positive(
4596 &EXT4_SB(sb)->s_freeclusters_counter)));
4597 if (percpu_counter_initialized(&EXT4_SB(sb)->s_freeinodes_counter))
4598 es->s_free_inodes_count =
4599 cpu_to_le32(percpu_counter_sum_positive(
4600 &EXT4_SB(sb)->s_freeinodes_counter));
4601 BUFFER_TRACE(sbh, "marking dirty");
4602 ext4_superblock_csum_set(sb);
4603 mark_buffer_dirty(sbh);
4605 error = sync_dirty_buffer(sbh);
4609 error = buffer_write_io_error(sbh);
4611 ext4_msg(sb, KERN_ERR, "I/O error while writing "
4613 clear_buffer_write_io_error(sbh);
4614 set_buffer_uptodate(sbh);
4621 * Have we just finished recovery? If so, and if we are mounting (or
4622 * remounting) the filesystem readonly, then we will end up with a
4623 * consistent fs on disk. Record that fact.
4625 static void ext4_mark_recovery_complete(struct super_block *sb,
4626 struct ext4_super_block *es)
4628 journal_t *journal = EXT4_SB(sb)->s_journal;
4630 if (!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL)) {
4631 BUG_ON(journal != NULL);
4634 jbd2_journal_lock_updates(journal);
4635 if (jbd2_journal_flush(journal) < 0)
4638 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER) &&
4639 sb->s_flags & MS_RDONLY) {
4640 EXT4_CLEAR_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
4641 ext4_commit_super(sb, 1);
4645 jbd2_journal_unlock_updates(journal);
4649 * If we are mounting (or read-write remounting) a filesystem whose journal
4650 * has recorded an error from a previous lifetime, move that error to the
4651 * main filesystem now.
4653 static void ext4_clear_journal_err(struct super_block *sb,
4654 struct ext4_super_block *es)
4660 BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL));
4662 journal = EXT4_SB(sb)->s_journal;
4665 * Now check for any error status which may have been recorded in the
4666 * journal by a prior ext4_error() or ext4_abort()
4669 j_errno = jbd2_journal_errno(journal);
4673 errstr = ext4_decode_error(sb, j_errno, nbuf);
4674 ext4_warning(sb, "Filesystem error recorded "
4675 "from previous mount: %s", errstr);
4676 ext4_warning(sb, "Marking fs in need of filesystem check.");
4678 EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS;
4679 es->s_state |= cpu_to_le16(EXT4_ERROR_FS);
4680 ext4_commit_super(sb, 1);
4682 jbd2_journal_clear_err(journal);
4683 jbd2_journal_update_sb_errno(journal);
4688 * Force the running and committing transactions to commit,
4689 * and wait on the commit.
4691 int ext4_force_commit(struct super_block *sb)
4695 if (sb->s_flags & MS_RDONLY)
4698 journal = EXT4_SB(sb)->s_journal;
4699 return ext4_journal_force_commit(journal);
4702 static int ext4_sync_fs(struct super_block *sb, int wait)
4706 bool needs_barrier = false;
4707 struct ext4_sb_info *sbi = EXT4_SB(sb);
4709 trace_ext4_sync_fs(sb, wait);
4710 flush_workqueue(sbi->rsv_conversion_wq);
4712 * Writeback quota in non-journalled quota case - journalled quota has
4715 dquot_writeback_dquots(sb, -1);
4717 * Data writeback is possible w/o journal transaction, so barrier must
4718 * being sent at the end of the function. But we can skip it if
4719 * transaction_commit will do it for us.
4721 target = jbd2_get_latest_transaction(sbi->s_journal);
4722 if (wait && sbi->s_journal->j_flags & JBD2_BARRIER &&
4723 !jbd2_trans_will_send_data_barrier(sbi->s_journal, target))
4724 needs_barrier = true;
4726 if (jbd2_journal_start_commit(sbi->s_journal, &target)) {
4728 ret = jbd2_log_wait_commit(sbi->s_journal, target);
4730 if (needs_barrier) {
4732 err = blkdev_issue_flush(sb->s_bdev, GFP_KERNEL, NULL);
4740 static int ext4_sync_fs_nojournal(struct super_block *sb, int wait)
4744 trace_ext4_sync_fs(sb, wait);
4745 flush_workqueue(EXT4_SB(sb)->rsv_conversion_wq);
4746 dquot_writeback_dquots(sb, -1);
4747 if (wait && test_opt(sb, BARRIER))
4748 ret = blkdev_issue_flush(sb->s_bdev, GFP_KERNEL, NULL);
4754 * LVM calls this function before a (read-only) snapshot is created. This
4755 * gives us a chance to flush the journal completely and mark the fs clean.
4757 * Note that only this function cannot bring a filesystem to be in a clean
4758 * state independently. It relies on upper layer to stop all data & metadata
4761 static int ext4_freeze(struct super_block *sb)
4766 if (sb->s_flags & MS_RDONLY)
4769 journal = EXT4_SB(sb)->s_journal;
4771 /* Now we set up the journal barrier. */
4772 jbd2_journal_lock_updates(journal);
4775 * Don't clear the needs_recovery flag if we failed to flush
4778 error = jbd2_journal_flush(journal);
4782 /* Journal blocked and flushed, clear needs_recovery flag. */
4783 EXT4_CLEAR_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
4784 error = ext4_commit_super(sb, 1);
4786 /* we rely on upper layer to stop further updates */
4787 jbd2_journal_unlock_updates(EXT4_SB(sb)->s_journal);
4792 * Called by LVM after the snapshot is done. We need to reset the RECOVER
4793 * flag here, even though the filesystem is not technically dirty yet.
4795 static int ext4_unfreeze(struct super_block *sb)
4797 if (sb->s_flags & MS_RDONLY)
4800 /* Reset the needs_recovery flag before the fs is unlocked. */
4801 EXT4_SET_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
4802 ext4_commit_super(sb, 1);
4807 * Structure to save mount options for ext4_remount's benefit
4809 struct ext4_mount_options {
4810 unsigned long s_mount_opt;
4811 unsigned long s_mount_opt2;
4814 unsigned long s_commit_interval;
4815 u32 s_min_batch_time, s_max_batch_time;
4818 char *s_qf_names[EXT4_MAXQUOTAS];
4822 static int ext4_remount(struct super_block *sb, int *flags, char *data)
4824 struct ext4_super_block *es;
4825 struct ext4_sb_info *sbi = EXT4_SB(sb);
4826 unsigned long old_sb_flags;
4827 struct ext4_mount_options old_opts;
4828 int enable_quota = 0;
4830 unsigned int journal_ioprio = DEFAULT_JOURNAL_IOPRIO;
4835 char *orig_data = kstrdup(data, GFP_KERNEL);
4837 /* Store the original options */
4838 old_sb_flags = sb->s_flags;
4839 old_opts.s_mount_opt = sbi->s_mount_opt;
4840 old_opts.s_mount_opt2 = sbi->s_mount_opt2;
4841 old_opts.s_resuid = sbi->s_resuid;
4842 old_opts.s_resgid = sbi->s_resgid;
4843 old_opts.s_commit_interval = sbi->s_commit_interval;
4844 old_opts.s_min_batch_time = sbi->s_min_batch_time;
4845 old_opts.s_max_batch_time = sbi->s_max_batch_time;
4847 old_opts.s_jquota_fmt = sbi->s_jquota_fmt;
4848 for (i = 0; i < EXT4_MAXQUOTAS; i++)
4849 if (sbi->s_qf_names[i]) {
4850 old_opts.s_qf_names[i] = kstrdup(sbi->s_qf_names[i],
4852 if (!old_opts.s_qf_names[i]) {
4853 for (j = 0; j < i; j++)
4854 kfree(old_opts.s_qf_names[j]);
4859 old_opts.s_qf_names[i] = NULL;
4861 if (sbi->s_journal && sbi->s_journal->j_task->io_context)
4862 journal_ioprio = sbi->s_journal->j_task->io_context->ioprio;
4865 * Allow the "check" option to be passed as a remount option.
4867 if (!parse_options(data, sb, NULL, &journal_ioprio, 1)) {
4872 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA) {
4873 if (test_opt2(sb, EXPLICIT_DELALLOC)) {
4874 ext4_msg(sb, KERN_ERR, "can't mount with "
4875 "both data=journal and delalloc");
4879 if (test_opt(sb, DIOREAD_NOLOCK)) {
4880 ext4_msg(sb, KERN_ERR, "can't mount with "
4881 "both data=journal and dioread_nolock");
4887 if (sbi->s_mount_flags & EXT4_MF_FS_ABORTED)
4888 ext4_abort(sb, "Abort forced by user");
4890 sb->s_flags = (sb->s_flags & ~MS_POSIXACL) |
4891 (test_opt(sb, POSIX_ACL) ? MS_POSIXACL : 0);
4895 if (sbi->s_journal) {
4896 ext4_init_journal_params(sb, sbi->s_journal);
4897 set_task_ioprio(sbi->s_journal->j_task, journal_ioprio);
4900 if ((*flags & MS_RDONLY) != (sb->s_flags & MS_RDONLY)) {
4901 if (sbi->s_mount_flags & EXT4_MF_FS_ABORTED) {
4906 if (*flags & MS_RDONLY) {
4907 err = sync_filesystem(sb);
4910 err = dquot_suspend(sb, -1);
4915 * First of all, the unconditional stuff we have to do
4916 * to disable replay of the journal when we next remount
4918 sb->s_flags |= MS_RDONLY;
4921 * OK, test if we are remounting a valid rw partition
4922 * readonly, and if so set the rdonly flag and then
4923 * mark the partition as valid again.
4925 if (!(es->s_state & cpu_to_le16(EXT4_VALID_FS)) &&
4926 (sbi->s_mount_state & EXT4_VALID_FS))
4927 es->s_state = cpu_to_le16(sbi->s_mount_state);
4930 ext4_mark_recovery_complete(sb, es);
4932 /* Make sure we can mount this feature set readwrite */
4933 if (!ext4_feature_set_ok(sb, 0)) {
4938 * Make sure the group descriptor checksums
4939 * are sane. If they aren't, refuse to remount r/w.
4941 for (g = 0; g < sbi->s_groups_count; g++) {
4942 struct ext4_group_desc *gdp =
4943 ext4_get_group_desc(sb, g, NULL);
4945 if (!ext4_group_desc_csum_verify(sb, g, gdp)) {
4946 ext4_msg(sb, KERN_ERR,
4947 "ext4_remount: Checksum for group %u failed (%u!=%u)",
4948 g, le16_to_cpu(ext4_group_desc_csum(sbi, g, gdp)),
4949 le16_to_cpu(gdp->bg_checksum));
4956 * If we have an unprocessed orphan list hanging
4957 * around from a previously readonly bdev mount,
4958 * require a full umount/remount for now.
4960 if (es->s_last_orphan) {
4961 ext4_msg(sb, KERN_WARNING, "Couldn't "
4962 "remount RDWR because of unprocessed "
4963 "orphan inode list. Please "
4964 "umount/remount instead");
4970 * Mounting a RDONLY partition read-write, so reread
4971 * and store the current valid flag. (It may have
4972 * been changed by e2fsck since we originally mounted
4976 ext4_clear_journal_err(sb, es);
4977 sbi->s_mount_state = le16_to_cpu(es->s_state);
4978 if (!ext4_setup_super(sb, es, 0))
4979 sb->s_flags &= ~MS_RDONLY;
4980 if (EXT4_HAS_INCOMPAT_FEATURE(sb,
4981 EXT4_FEATURE_INCOMPAT_MMP))
4982 if (ext4_multi_mount_protect(sb,
4983 le64_to_cpu(es->s_mmp_block))) {
4992 * Reinitialize lazy itable initialization thread based on
4995 if ((sb->s_flags & MS_RDONLY) || !test_opt(sb, INIT_INODE_TABLE))
4996 ext4_unregister_li_request(sb);
4998 ext4_group_t first_not_zeroed;
4999 first_not_zeroed = ext4_has_uninit_itable(sb);
5000 ext4_register_li_request(sb, first_not_zeroed);
5003 ext4_setup_system_zone(sb);
5004 if (sbi->s_journal == NULL && !(old_sb_flags & MS_RDONLY))
5005 ext4_commit_super(sb, 1);
5008 /* Release old quota file names */
5009 for (i = 0; i < EXT4_MAXQUOTAS; i++)
5010 kfree(old_opts.s_qf_names[i]);
5012 if (sb_any_quota_suspended(sb))
5013 dquot_resume(sb, -1);
5014 else if (EXT4_HAS_RO_COMPAT_FEATURE(sb,
5015 EXT4_FEATURE_RO_COMPAT_QUOTA)) {
5016 err = ext4_enable_quotas(sb);
5023 ext4_msg(sb, KERN_INFO, "re-mounted. Opts: %s", orig_data);
5028 sb->s_flags = old_sb_flags;
5029 sbi->s_mount_opt = old_opts.s_mount_opt;
5030 sbi->s_mount_opt2 = old_opts.s_mount_opt2;
5031 sbi->s_resuid = old_opts.s_resuid;
5032 sbi->s_resgid = old_opts.s_resgid;
5033 sbi->s_commit_interval = old_opts.s_commit_interval;
5034 sbi->s_min_batch_time = old_opts.s_min_batch_time;
5035 sbi->s_max_batch_time = old_opts.s_max_batch_time;
5037 sbi->s_jquota_fmt = old_opts.s_jquota_fmt;
5038 for (i = 0; i < EXT4_MAXQUOTAS; i++) {
5039 kfree(sbi->s_qf_names[i]);
5040 sbi->s_qf_names[i] = old_opts.s_qf_names[i];
5047 static int ext4_statfs(struct dentry *dentry, struct kstatfs *buf)
5049 struct super_block *sb = dentry->d_sb;
5050 struct ext4_sb_info *sbi = EXT4_SB(sb);
5051 struct ext4_super_block *es = sbi->s_es;
5052 ext4_fsblk_t overhead = 0, resv_blocks;
5055 resv_blocks = EXT4_C2B(sbi, atomic64_read(&sbi->s_resv_clusters));
5057 if (!test_opt(sb, MINIX_DF))
5058 overhead = sbi->s_overhead;
5060 buf->f_type = EXT4_SUPER_MAGIC;
5061 buf->f_bsize = sb->s_blocksize;
5062 buf->f_blocks = ext4_blocks_count(es) - EXT4_C2B(sbi, overhead);
5063 bfree = percpu_counter_sum_positive(&sbi->s_freeclusters_counter) -
5064 percpu_counter_sum_positive(&sbi->s_dirtyclusters_counter);
5065 /* prevent underflow in case that few free space is available */
5066 buf->f_bfree = EXT4_C2B(sbi, max_t(s64, bfree, 0));
5067 buf->f_bavail = buf->f_bfree -
5068 (ext4_r_blocks_count(es) + resv_blocks);
5069 if (buf->f_bfree < (ext4_r_blocks_count(es) + resv_blocks))
5071 buf->f_files = le32_to_cpu(es->s_inodes_count);
5072 buf->f_ffree = percpu_counter_sum_positive(&sbi->s_freeinodes_counter);
5073 buf->f_namelen = EXT4_NAME_LEN;
5074 fsid = le64_to_cpup((void *)es->s_uuid) ^
5075 le64_to_cpup((void *)es->s_uuid + sizeof(u64));
5076 buf->f_fsid.val[0] = fsid & 0xFFFFFFFFUL;
5077 buf->f_fsid.val[1] = (fsid >> 32) & 0xFFFFFFFFUL;
5082 /* Helper function for writing quotas on sync - we need to start transaction
5083 * before quota file is locked for write. Otherwise the are possible deadlocks:
5084 * Process 1 Process 2
5085 * ext4_create() quota_sync()
5086 * jbd2_journal_start() write_dquot()
5087 * dquot_initialize() down(dqio_mutex)
5088 * down(dqio_mutex) jbd2_journal_start()
5094 static inline struct inode *dquot_to_inode(struct dquot *dquot)
5096 return sb_dqopt(dquot->dq_sb)->files[dquot->dq_id.type];
5099 static int ext4_write_dquot(struct dquot *dquot)
5103 struct inode *inode;
5105 inode = dquot_to_inode(dquot);
5106 handle = ext4_journal_start(inode, EXT4_HT_QUOTA,
5107 EXT4_QUOTA_TRANS_BLOCKS(dquot->dq_sb));
5109 return PTR_ERR(handle);
5110 ret = dquot_commit(dquot);
5111 err = ext4_journal_stop(handle);
5117 static int ext4_acquire_dquot(struct dquot *dquot)
5122 handle = ext4_journal_start(dquot_to_inode(dquot), EXT4_HT_QUOTA,
5123 EXT4_QUOTA_INIT_BLOCKS(dquot->dq_sb));
5125 return PTR_ERR(handle);
5126 ret = dquot_acquire(dquot);
5127 err = ext4_journal_stop(handle);
5133 static int ext4_release_dquot(struct dquot *dquot)
5138 handle = ext4_journal_start(dquot_to_inode(dquot), EXT4_HT_QUOTA,
5139 EXT4_QUOTA_DEL_BLOCKS(dquot->dq_sb));
5140 if (IS_ERR(handle)) {
5141 /* Release dquot anyway to avoid endless cycle in dqput() */
5142 dquot_release(dquot);
5143 return PTR_ERR(handle);
5145 ret = dquot_release(dquot);
5146 err = ext4_journal_stop(handle);
5152 static int ext4_mark_dquot_dirty(struct dquot *dquot)
5154 struct super_block *sb = dquot->dq_sb;
5155 struct ext4_sb_info *sbi = EXT4_SB(sb);
5157 /* Are we journaling quotas? */
5158 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_QUOTA) ||
5159 sbi->s_qf_names[USRQUOTA] || sbi->s_qf_names[GRPQUOTA]) {
5160 dquot_mark_dquot_dirty(dquot);
5161 return ext4_write_dquot(dquot);
5163 return dquot_mark_dquot_dirty(dquot);
5167 static int ext4_write_info(struct super_block *sb, int type)
5172 /* Data block + inode block */
5173 handle = ext4_journal_start(sb->s_root->d_inode, EXT4_HT_QUOTA, 2);
5175 return PTR_ERR(handle);
5176 ret = dquot_commit_info(sb, type);
5177 err = ext4_journal_stop(handle);
5184 * Turn on quotas during mount time - we need to find
5185 * the quota file and such...
5187 static int ext4_quota_on_mount(struct super_block *sb, int type)
5189 return dquot_quota_on_mount(sb, EXT4_SB(sb)->s_qf_names[type],
5190 EXT4_SB(sb)->s_jquota_fmt, type);
5194 * Standard function to be called on quota_on
5196 static int ext4_quota_on(struct super_block *sb, int type, int format_id,
5201 if (!test_opt(sb, QUOTA))
5204 /* Quotafile not on the same filesystem? */
5205 if (path->dentry->d_sb != sb)
5207 /* Journaling quota? */
5208 if (EXT4_SB(sb)->s_qf_names[type]) {
5209 /* Quotafile not in fs root? */
5210 if (path->dentry->d_parent != sb->s_root)
5211 ext4_msg(sb, KERN_WARNING,
5212 "Quota file not on filesystem root. "
5213 "Journaled quota will not work");
5217 * When we journal data on quota file, we have to flush journal to see
5218 * all updates to the file when we bypass pagecache...
5220 if (EXT4_SB(sb)->s_journal &&
5221 ext4_should_journal_data(path->dentry->d_inode)) {
5223 * We don't need to lock updates but journal_flush() could
5224 * otherwise be livelocked...
5226 jbd2_journal_lock_updates(EXT4_SB(sb)->s_journal);
5227 err = jbd2_journal_flush(EXT4_SB(sb)->s_journal);
5228 jbd2_journal_unlock_updates(EXT4_SB(sb)->s_journal);
5233 return dquot_quota_on(sb, type, format_id, path);
5236 static int ext4_quota_enable(struct super_block *sb, int type, int format_id,
5240 struct inode *qf_inode;
5241 unsigned long qf_inums[EXT4_MAXQUOTAS] = {
5242 le32_to_cpu(EXT4_SB(sb)->s_es->s_usr_quota_inum),
5243 le32_to_cpu(EXT4_SB(sb)->s_es->s_grp_quota_inum)
5246 BUG_ON(!EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_QUOTA));
5248 if (!qf_inums[type])
5251 qf_inode = ext4_iget(sb, qf_inums[type]);
5252 if (IS_ERR(qf_inode)) {
5253 ext4_error(sb, "Bad quota inode # %lu", qf_inums[type]);
5254 return PTR_ERR(qf_inode);
5257 /* Don't account quota for quota files to avoid recursion */
5258 qf_inode->i_flags |= S_NOQUOTA;
5259 err = dquot_enable(qf_inode, type, format_id, flags);
5265 /* Enable usage tracking for all quota types. */
5266 static int ext4_enable_quotas(struct super_block *sb)
5269 unsigned long qf_inums[EXT4_MAXQUOTAS] = {
5270 le32_to_cpu(EXT4_SB(sb)->s_es->s_usr_quota_inum),
5271 le32_to_cpu(EXT4_SB(sb)->s_es->s_grp_quota_inum)
5274 sb_dqopt(sb)->flags |= DQUOT_QUOTA_SYS_FILE;
5275 for (type = 0; type < EXT4_MAXQUOTAS; type++) {
5276 if (qf_inums[type]) {
5277 err = ext4_quota_enable(sb, type, QFMT_VFS_V1,
5278 DQUOT_USAGE_ENABLED);
5281 "Failed to enable quota tracking "
5282 "(type=%d, err=%d). Please run "
5283 "e2fsck to fix.", type, err);
5292 * quota_on function that is used when QUOTA feature is set.
5294 static int ext4_quota_on_sysfile(struct super_block *sb, int type,
5297 if (!EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_QUOTA))
5301 * USAGE was enabled at mount time. Only need to enable LIMITS now.
5303 return ext4_quota_enable(sb, type, format_id, DQUOT_LIMITS_ENABLED);
5306 static int ext4_quota_off(struct super_block *sb, int type)
5308 struct inode *inode = sb_dqopt(sb)->files[type];
5311 /* Force all delayed allocation blocks to be allocated.
5312 * Caller already holds s_umount sem */
5313 if (test_opt(sb, DELALLOC))
5314 sync_filesystem(sb);
5319 /* Update modification times of quota files when userspace can
5320 * start looking at them */
5321 handle = ext4_journal_start(inode, EXT4_HT_QUOTA, 1);
5324 inode->i_mtime = inode->i_ctime = CURRENT_TIME;
5325 ext4_mark_inode_dirty(handle, inode);
5326 ext4_journal_stop(handle);
5329 return dquot_quota_off(sb, type);
5333 * quota_off function that is used when QUOTA feature is set.
5335 static int ext4_quota_off_sysfile(struct super_block *sb, int type)
5337 if (!EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_QUOTA))
5340 /* Disable only the limits. */
5341 return dquot_disable(sb, type, DQUOT_LIMITS_ENABLED);
5344 /* Read data from quotafile - avoid pagecache and such because we cannot afford
5345 * acquiring the locks... As quota files are never truncated and quota code
5346 * itself serializes the operations (and no one else should touch the files)
5347 * we don't have to be afraid of races */
5348 static ssize_t ext4_quota_read(struct super_block *sb, int type, char *data,
5349 size_t len, loff_t off)
5351 struct inode *inode = sb_dqopt(sb)->files[type];
5352 ext4_lblk_t blk = off >> EXT4_BLOCK_SIZE_BITS(sb);
5353 int offset = off & (sb->s_blocksize - 1);
5356 struct buffer_head *bh;
5357 loff_t i_size = i_size_read(inode);
5361 if (off+len > i_size)
5364 while (toread > 0) {
5365 tocopy = sb->s_blocksize - offset < toread ?
5366 sb->s_blocksize - offset : toread;
5367 bh = ext4_bread(NULL, inode, blk, 0);
5370 if (!bh) /* A hole? */
5371 memset(data, 0, tocopy);
5373 memcpy(data, bh->b_data+offset, tocopy);
5383 /* Write to quotafile (we know the transaction is already started and has
5384 * enough credits) */
5385 static ssize_t ext4_quota_write(struct super_block *sb, int type,
5386 const char *data, size_t len, loff_t off)
5388 struct inode *inode = sb_dqopt(sb)->files[type];
5389 ext4_lblk_t blk = off >> EXT4_BLOCK_SIZE_BITS(sb);
5390 int err, offset = off & (sb->s_blocksize - 1);
5391 struct buffer_head *bh;
5392 handle_t *handle = journal_current_handle();
5394 if (EXT4_SB(sb)->s_journal && !handle) {
5395 ext4_msg(sb, KERN_WARNING, "Quota write (off=%llu, len=%llu)"
5396 " cancelled because transaction is not started",
5397 (unsigned long long)off, (unsigned long long)len);
5401 * Since we account only one data block in transaction credits,
5402 * then it is impossible to cross a block boundary.
5404 if (sb->s_blocksize - offset < len) {
5405 ext4_msg(sb, KERN_WARNING, "Quota write (off=%llu, len=%llu)"
5406 " cancelled because not block aligned",
5407 (unsigned long long)off, (unsigned long long)len);
5411 bh = ext4_bread(handle, inode, blk, 1);
5416 BUFFER_TRACE(bh, "get write access");
5417 err = ext4_journal_get_write_access(handle, bh);
5423 memcpy(bh->b_data+offset, data, len);
5424 flush_dcache_page(bh->b_page);
5426 err = ext4_handle_dirty_metadata(handle, NULL, bh);
5429 if (inode->i_size < off + len) {
5430 i_size_write(inode, off + len);
5431 EXT4_I(inode)->i_disksize = inode->i_size;
5432 ext4_mark_inode_dirty(handle, inode);
5439 static struct dentry *ext4_mount(struct file_system_type *fs_type, int flags,
5440 const char *dev_name, void *data)
5442 return mount_bdev(fs_type, flags, dev_name, data, ext4_fill_super);
5445 #if !defined(CONFIG_EXT2_FS) && !defined(CONFIG_EXT2_FS_MODULE) && defined(CONFIG_EXT4_USE_FOR_EXT23)
5446 static inline void register_as_ext2(void)
5448 int err = register_filesystem(&ext2_fs_type);
5451 "EXT4-fs: Unable to register as ext2 (%d)\n", err);
5454 static inline void unregister_as_ext2(void)
5456 unregister_filesystem(&ext2_fs_type);
5459 static inline int ext2_feature_set_ok(struct super_block *sb)
5461 if (EXT4_HAS_INCOMPAT_FEATURE(sb, ~EXT2_FEATURE_INCOMPAT_SUPP))
5463 if (sb->s_flags & MS_RDONLY)
5465 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, ~EXT2_FEATURE_RO_COMPAT_SUPP))
5470 static inline void register_as_ext2(void) { }
5471 static inline void unregister_as_ext2(void) { }
5472 static inline int ext2_feature_set_ok(struct super_block *sb) { return 0; }
5475 #if !defined(CONFIG_EXT3_FS) && !defined(CONFIG_EXT3_FS_MODULE) && defined(CONFIG_EXT4_USE_FOR_EXT23)
5476 static inline void register_as_ext3(void)
5478 int err = register_filesystem(&ext3_fs_type);
5481 "EXT4-fs: Unable to register as ext3 (%d)\n", err);
5484 static inline void unregister_as_ext3(void)
5486 unregister_filesystem(&ext3_fs_type);
5489 static inline int ext3_feature_set_ok(struct super_block *sb)
5491 if (EXT4_HAS_INCOMPAT_FEATURE(sb, ~EXT3_FEATURE_INCOMPAT_SUPP))
5493 if (!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL))
5495 if (sb->s_flags & MS_RDONLY)
5497 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, ~EXT3_FEATURE_RO_COMPAT_SUPP))
5502 static inline void register_as_ext3(void) { }
5503 static inline void unregister_as_ext3(void) { }
5504 static inline int ext3_feature_set_ok(struct super_block *sb) { return 0; }
5507 static struct file_system_type ext4_fs_type = {
5508 .owner = THIS_MODULE,
5510 .mount = ext4_mount,
5511 .kill_sb = kill_block_super,
5512 .fs_flags = FS_REQUIRES_DEV,
5514 MODULE_ALIAS_FS("ext4");
5516 static int __init ext4_init_feat_adverts(void)
5518 struct ext4_features *ef;
5521 ef = kzalloc(sizeof(struct ext4_features), GFP_KERNEL);
5525 ef->f_kobj.kset = ext4_kset;
5526 init_completion(&ef->f_kobj_unregister);
5527 ret = kobject_init_and_add(&ef->f_kobj, &ext4_feat_ktype, NULL,
5540 static void ext4_exit_feat_adverts(void)
5542 kobject_put(&ext4_feat->f_kobj);
5543 wait_for_completion(&ext4_feat->f_kobj_unregister);
5547 /* Shared across all ext4 file systems */
5548 wait_queue_head_t ext4__ioend_wq[EXT4_WQ_HASH_SZ];
5549 struct mutex ext4__aio_mutex[EXT4_WQ_HASH_SZ];
5551 static int __init ext4_init_fs(void)
5555 ext4_li_info = NULL;
5556 mutex_init(&ext4_li_mtx);
5558 /* Build-time check for flags consistency */
5559 ext4_check_flag_values();
5561 for (i = 0; i < EXT4_WQ_HASH_SZ; i++) {
5562 mutex_init(&ext4__aio_mutex[i]);
5563 init_waitqueue_head(&ext4__ioend_wq[i]);
5566 err = ext4_init_es();
5570 err = ext4_init_pageio();
5574 err = ext4_init_system_zone();
5577 ext4_kset = kset_create_and_add("ext4", NULL, fs_kobj);
5582 ext4_proc_root = proc_mkdir("fs/ext4", NULL);
5584 err = ext4_init_feat_adverts();
5588 err = ext4_init_mballoc();
5592 ext4_mballoc_ready = 1;
5593 err = init_inodecache();
5598 err = register_filesystem(&ext4_fs_type);
5604 unregister_as_ext2();
5605 unregister_as_ext3();
5606 destroy_inodecache();
5608 ext4_mballoc_ready = 0;
5609 ext4_exit_mballoc();
5611 ext4_exit_feat_adverts();
5614 remove_proc_entry("fs/ext4", NULL);
5615 kset_unregister(ext4_kset);
5617 ext4_exit_system_zone();
5626 static void __exit ext4_exit_fs(void)
5628 ext4_destroy_lazyinit_thread();
5629 unregister_as_ext2();
5630 unregister_as_ext3();
5631 unregister_filesystem(&ext4_fs_type);
5632 destroy_inodecache();
5633 ext4_exit_mballoc();
5634 ext4_exit_feat_adverts();
5635 remove_proc_entry("fs/ext4", NULL);
5636 kset_unregister(ext4_kset);
5637 ext4_exit_system_zone();
5642 MODULE_AUTHOR("Remy Card, Stephen Tweedie, Andrew Morton, Andreas Dilger, Theodore Ts'o and others");
5643 MODULE_DESCRIPTION("Fourth Extended Filesystem");
5644 MODULE_LICENSE("GPL");
5645 module_init(ext4_init_fs)
5646 module_exit(ext4_exit_fs)