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_metadata_csum(sb))
147 return es->s_checksum == ext4_superblock_csum(sb, es);
150 void ext4_superblock_csum_set(struct super_block *sb)
152 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
154 if (!ext4_has_metadata_csum(sb))
157 es->s_checksum = ext4_superblock_csum(sb, es);
160 void *ext4_kvmalloc(size_t size, gfp_t flags)
164 ret = kmalloc(size, flags | __GFP_NOWARN);
166 ret = __vmalloc(size, flags, PAGE_KERNEL);
170 void *ext4_kvzalloc(size_t size, gfp_t flags)
174 ret = kzalloc(size, flags | __GFP_NOWARN);
176 ret = __vmalloc(size, flags | __GFP_ZERO, PAGE_KERNEL);
180 void ext4_kvfree(void *ptr)
182 if (is_vmalloc_addr(ptr))
189 ext4_fsblk_t ext4_block_bitmap(struct super_block *sb,
190 struct ext4_group_desc *bg)
192 return le32_to_cpu(bg->bg_block_bitmap_lo) |
193 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
194 (ext4_fsblk_t)le32_to_cpu(bg->bg_block_bitmap_hi) << 32 : 0);
197 ext4_fsblk_t ext4_inode_bitmap(struct super_block *sb,
198 struct ext4_group_desc *bg)
200 return le32_to_cpu(bg->bg_inode_bitmap_lo) |
201 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
202 (ext4_fsblk_t)le32_to_cpu(bg->bg_inode_bitmap_hi) << 32 : 0);
205 ext4_fsblk_t ext4_inode_table(struct super_block *sb,
206 struct ext4_group_desc *bg)
208 return le32_to_cpu(bg->bg_inode_table_lo) |
209 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
210 (ext4_fsblk_t)le32_to_cpu(bg->bg_inode_table_hi) << 32 : 0);
213 __u32 ext4_free_group_clusters(struct super_block *sb,
214 struct ext4_group_desc *bg)
216 return le16_to_cpu(bg->bg_free_blocks_count_lo) |
217 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
218 (__u32)le16_to_cpu(bg->bg_free_blocks_count_hi) << 16 : 0);
221 __u32 ext4_free_inodes_count(struct super_block *sb,
222 struct ext4_group_desc *bg)
224 return le16_to_cpu(bg->bg_free_inodes_count_lo) |
225 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
226 (__u32)le16_to_cpu(bg->bg_free_inodes_count_hi) << 16 : 0);
229 __u32 ext4_used_dirs_count(struct super_block *sb,
230 struct ext4_group_desc *bg)
232 return le16_to_cpu(bg->bg_used_dirs_count_lo) |
233 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
234 (__u32)le16_to_cpu(bg->bg_used_dirs_count_hi) << 16 : 0);
237 __u32 ext4_itable_unused_count(struct super_block *sb,
238 struct ext4_group_desc *bg)
240 return le16_to_cpu(bg->bg_itable_unused_lo) |
241 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
242 (__u32)le16_to_cpu(bg->bg_itable_unused_hi) << 16 : 0);
245 void ext4_block_bitmap_set(struct super_block *sb,
246 struct ext4_group_desc *bg, ext4_fsblk_t blk)
248 bg->bg_block_bitmap_lo = cpu_to_le32((u32)blk);
249 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
250 bg->bg_block_bitmap_hi = cpu_to_le32(blk >> 32);
253 void ext4_inode_bitmap_set(struct super_block *sb,
254 struct ext4_group_desc *bg, ext4_fsblk_t blk)
256 bg->bg_inode_bitmap_lo = cpu_to_le32((u32)blk);
257 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
258 bg->bg_inode_bitmap_hi = cpu_to_le32(blk >> 32);
261 void ext4_inode_table_set(struct super_block *sb,
262 struct ext4_group_desc *bg, ext4_fsblk_t blk)
264 bg->bg_inode_table_lo = cpu_to_le32((u32)blk);
265 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
266 bg->bg_inode_table_hi = cpu_to_le32(blk >> 32);
269 void ext4_free_group_clusters_set(struct super_block *sb,
270 struct ext4_group_desc *bg, __u32 count)
272 bg->bg_free_blocks_count_lo = cpu_to_le16((__u16)count);
273 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
274 bg->bg_free_blocks_count_hi = cpu_to_le16(count >> 16);
277 void ext4_free_inodes_set(struct super_block *sb,
278 struct ext4_group_desc *bg, __u32 count)
280 bg->bg_free_inodes_count_lo = cpu_to_le16((__u16)count);
281 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
282 bg->bg_free_inodes_count_hi = cpu_to_le16(count >> 16);
285 void ext4_used_dirs_set(struct super_block *sb,
286 struct ext4_group_desc *bg, __u32 count)
288 bg->bg_used_dirs_count_lo = cpu_to_le16((__u16)count);
289 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
290 bg->bg_used_dirs_count_hi = cpu_to_le16(count >> 16);
293 void ext4_itable_unused_set(struct super_block *sb,
294 struct ext4_group_desc *bg, __u32 count)
296 bg->bg_itable_unused_lo = cpu_to_le16((__u16)count);
297 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
298 bg->bg_itable_unused_hi = cpu_to_le16(count >> 16);
302 static void __save_error_info(struct super_block *sb, const char *func,
305 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
307 EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS;
308 es->s_state |= cpu_to_le16(EXT4_ERROR_FS);
309 es->s_last_error_time = cpu_to_le32(get_seconds());
310 strncpy(es->s_last_error_func, func, sizeof(es->s_last_error_func));
311 es->s_last_error_line = cpu_to_le32(line);
312 if (!es->s_first_error_time) {
313 es->s_first_error_time = es->s_last_error_time;
314 strncpy(es->s_first_error_func, func,
315 sizeof(es->s_first_error_func));
316 es->s_first_error_line = cpu_to_le32(line);
317 es->s_first_error_ino = es->s_last_error_ino;
318 es->s_first_error_block = es->s_last_error_block;
321 * Start the daily error reporting function if it hasn't been
324 if (!es->s_error_count)
325 mod_timer(&EXT4_SB(sb)->s_err_report, jiffies + 24*60*60*HZ);
326 le32_add_cpu(&es->s_error_count, 1);
329 static void save_error_info(struct super_block *sb, const char *func,
332 __save_error_info(sb, func, line);
333 ext4_commit_super(sb, 1);
337 * The del_gendisk() function uninitializes the disk-specific data
338 * structures, including the bdi structure, without telling anyone
339 * else. Once this happens, any attempt to call mark_buffer_dirty()
340 * (for example, by ext4_commit_super), will cause a kernel OOPS.
341 * This is a kludge to prevent these oops until we can put in a proper
342 * hook in del_gendisk() to inform the VFS and file system layers.
344 static int block_device_ejected(struct super_block *sb)
346 struct inode *bd_inode = sb->s_bdev->bd_inode;
347 struct backing_dev_info *bdi = bd_inode->i_mapping->backing_dev_info;
349 return bdi->dev == NULL;
352 static void ext4_journal_commit_callback(journal_t *journal, transaction_t *txn)
354 struct super_block *sb = journal->j_private;
355 struct ext4_sb_info *sbi = EXT4_SB(sb);
356 int error = is_journal_aborted(journal);
357 struct ext4_journal_cb_entry *jce;
359 BUG_ON(txn->t_state == T_FINISHED);
360 spin_lock(&sbi->s_md_lock);
361 while (!list_empty(&txn->t_private_list)) {
362 jce = list_entry(txn->t_private_list.next,
363 struct ext4_journal_cb_entry, jce_list);
364 list_del_init(&jce->jce_list);
365 spin_unlock(&sbi->s_md_lock);
366 jce->jce_func(sb, jce, error);
367 spin_lock(&sbi->s_md_lock);
369 spin_unlock(&sbi->s_md_lock);
372 /* Deal with the reporting of failure conditions on a filesystem such as
373 * inconsistencies detected or read IO failures.
375 * On ext2, we can store the error state of the filesystem in the
376 * superblock. That is not possible on ext4, because we may have other
377 * write ordering constraints on the superblock which prevent us from
378 * writing it out straight away; and given that the journal is about to
379 * be aborted, we can't rely on the current, or future, transactions to
380 * write out the superblock safely.
382 * We'll just use the jbd2_journal_abort() error code to record an error in
383 * the journal instead. On recovery, the journal will complain about
384 * that error until we've noted it down and cleared it.
387 static void ext4_handle_error(struct super_block *sb)
389 if (sb->s_flags & MS_RDONLY)
392 if (!test_opt(sb, ERRORS_CONT)) {
393 journal_t *journal = EXT4_SB(sb)->s_journal;
395 EXT4_SB(sb)->s_mount_flags |= EXT4_MF_FS_ABORTED;
397 jbd2_journal_abort(journal, -EIO);
399 if (test_opt(sb, ERRORS_RO)) {
400 ext4_msg(sb, KERN_CRIT, "Remounting filesystem read-only");
402 * Make sure updated value of ->s_mount_flags will be visible
403 * before ->s_flags update
406 sb->s_flags |= MS_RDONLY;
408 if (test_opt(sb, ERRORS_PANIC)) {
409 if (EXT4_SB(sb)->s_journal &&
410 !(EXT4_SB(sb)->s_journal->j_flags & JBD2_REC_ERR))
412 panic("EXT4-fs (device %s): panic forced after error\n",
417 #define ext4_error_ratelimit(sb) \
418 ___ratelimit(&(EXT4_SB(sb)->s_err_ratelimit_state), \
421 void __ext4_error(struct super_block *sb, const char *function,
422 unsigned int line, const char *fmt, ...)
424 struct va_format vaf;
427 if (ext4_error_ratelimit(sb)) {
432 "EXT4-fs error (device %s): %s:%d: comm %s: %pV\n",
433 sb->s_id, function, line, current->comm, &vaf);
436 save_error_info(sb, function, line);
437 ext4_handle_error(sb);
440 void __ext4_error_inode(struct inode *inode, const char *function,
441 unsigned int line, ext4_fsblk_t block,
442 const char *fmt, ...)
445 struct va_format vaf;
446 struct ext4_super_block *es = EXT4_SB(inode->i_sb)->s_es;
448 es->s_last_error_ino = cpu_to_le32(inode->i_ino);
449 es->s_last_error_block = cpu_to_le64(block);
450 if (ext4_error_ratelimit(inode->i_sb)) {
455 printk(KERN_CRIT "EXT4-fs error (device %s): %s:%d: "
456 "inode #%lu: block %llu: comm %s: %pV\n",
457 inode->i_sb->s_id, function, line, inode->i_ino,
458 block, current->comm, &vaf);
460 printk(KERN_CRIT "EXT4-fs error (device %s): %s:%d: "
461 "inode #%lu: comm %s: %pV\n",
462 inode->i_sb->s_id, function, line, inode->i_ino,
463 current->comm, &vaf);
466 save_error_info(inode->i_sb, function, line);
467 ext4_handle_error(inode->i_sb);
470 void __ext4_error_file(struct file *file, const char *function,
471 unsigned int line, ext4_fsblk_t block,
472 const char *fmt, ...)
475 struct va_format vaf;
476 struct ext4_super_block *es;
477 struct inode *inode = file_inode(file);
478 char pathname[80], *path;
480 es = EXT4_SB(inode->i_sb)->s_es;
481 es->s_last_error_ino = cpu_to_le32(inode->i_ino);
482 if (ext4_error_ratelimit(inode->i_sb)) {
483 path = d_path(&(file->f_path), pathname, sizeof(pathname));
491 "EXT4-fs error (device %s): %s:%d: inode #%lu: "
492 "block %llu: comm %s: path %s: %pV\n",
493 inode->i_sb->s_id, function, line, inode->i_ino,
494 block, current->comm, path, &vaf);
497 "EXT4-fs error (device %s): %s:%d: inode #%lu: "
498 "comm %s: path %s: %pV\n",
499 inode->i_sb->s_id, function, line, inode->i_ino,
500 current->comm, path, &vaf);
503 save_error_info(inode->i_sb, function, line);
504 ext4_handle_error(inode->i_sb);
507 const char *ext4_decode_error(struct super_block *sb, int errno,
514 errstr = "IO failure";
517 errstr = "Out of memory";
520 if (!sb || (EXT4_SB(sb)->s_journal &&
521 EXT4_SB(sb)->s_journal->j_flags & JBD2_ABORT))
522 errstr = "Journal has aborted";
524 errstr = "Readonly filesystem";
527 /* If the caller passed in an extra buffer for unknown
528 * errors, textualise them now. Else we just return
531 /* Check for truncated error codes... */
532 if (snprintf(nbuf, 16, "error %d", -errno) >= 0)
541 /* __ext4_std_error decodes expected errors from journaling functions
542 * automatically and invokes the appropriate error response. */
544 void __ext4_std_error(struct super_block *sb, const char *function,
545 unsigned int line, int errno)
550 /* Special case: if the error is EROFS, and we're not already
551 * inside a transaction, then there's really no point in logging
553 if (errno == -EROFS && journal_current_handle() == NULL &&
554 (sb->s_flags & MS_RDONLY))
557 if (ext4_error_ratelimit(sb)) {
558 errstr = ext4_decode_error(sb, errno, nbuf);
559 printk(KERN_CRIT "EXT4-fs error (device %s) in %s:%d: %s\n",
560 sb->s_id, function, line, errstr);
563 save_error_info(sb, function, line);
564 ext4_handle_error(sb);
568 * ext4_abort is a much stronger failure handler than ext4_error. The
569 * abort function may be used to deal with unrecoverable failures such
570 * as journal IO errors or ENOMEM at a critical moment in log management.
572 * We unconditionally force the filesystem into an ABORT|READONLY state,
573 * unless the error response on the fs has been set to panic in which
574 * case we take the easy way out and panic immediately.
577 void __ext4_abort(struct super_block *sb, const char *function,
578 unsigned int line, const char *fmt, ...)
582 save_error_info(sb, function, line);
584 printk(KERN_CRIT "EXT4-fs error (device %s): %s:%d: ", sb->s_id,
590 if ((sb->s_flags & MS_RDONLY) == 0) {
591 ext4_msg(sb, KERN_CRIT, "Remounting filesystem read-only");
592 EXT4_SB(sb)->s_mount_flags |= EXT4_MF_FS_ABORTED;
594 * Make sure updated value of ->s_mount_flags will be visible
595 * before ->s_flags update
598 sb->s_flags |= MS_RDONLY;
599 if (EXT4_SB(sb)->s_journal)
600 jbd2_journal_abort(EXT4_SB(sb)->s_journal, -EIO);
601 save_error_info(sb, function, line);
603 if (test_opt(sb, ERRORS_PANIC)) {
604 if (EXT4_SB(sb)->s_journal &&
605 !(EXT4_SB(sb)->s_journal->j_flags & JBD2_REC_ERR))
607 panic("EXT4-fs panic from previous error\n");
611 void __ext4_msg(struct super_block *sb,
612 const char *prefix, const char *fmt, ...)
614 struct va_format vaf;
617 if (!___ratelimit(&(EXT4_SB(sb)->s_msg_ratelimit_state), "EXT4-fs"))
623 printk("%sEXT4-fs (%s): %pV\n", prefix, sb->s_id, &vaf);
627 void __ext4_warning(struct super_block *sb, const char *function,
628 unsigned int line, const char *fmt, ...)
630 struct va_format vaf;
633 if (!___ratelimit(&(EXT4_SB(sb)->s_warning_ratelimit_state),
640 printk(KERN_WARNING "EXT4-fs warning (device %s): %s:%d: %pV\n",
641 sb->s_id, function, line, &vaf);
645 void __ext4_grp_locked_error(const char *function, unsigned int line,
646 struct super_block *sb, ext4_group_t grp,
647 unsigned long ino, ext4_fsblk_t block,
648 const char *fmt, ...)
652 struct va_format vaf;
654 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
656 es->s_last_error_ino = cpu_to_le32(ino);
657 es->s_last_error_block = cpu_to_le64(block);
658 __save_error_info(sb, function, line);
660 if (ext4_error_ratelimit(sb)) {
664 printk(KERN_CRIT "EXT4-fs error (device %s): %s:%d: group %u, ",
665 sb->s_id, function, line, grp);
667 printk(KERN_CONT "inode %lu: ", ino);
669 printk(KERN_CONT "block %llu:",
670 (unsigned long long) block);
671 printk(KERN_CONT "%pV\n", &vaf);
675 if (test_opt(sb, ERRORS_CONT)) {
676 ext4_commit_super(sb, 0);
680 ext4_unlock_group(sb, grp);
681 ext4_handle_error(sb);
683 * We only get here in the ERRORS_RO case; relocking the group
684 * may be dangerous, but nothing bad will happen since the
685 * filesystem will have already been marked read/only and the
686 * journal has been aborted. We return 1 as a hint to callers
687 * who might what to use the return value from
688 * ext4_grp_locked_error() to distinguish between the
689 * ERRORS_CONT and ERRORS_RO case, and perhaps return more
690 * aggressively from the ext4 function in question, with a
691 * more appropriate error code.
693 ext4_lock_group(sb, grp);
697 void ext4_update_dynamic_rev(struct super_block *sb)
699 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
701 if (le32_to_cpu(es->s_rev_level) > EXT4_GOOD_OLD_REV)
705 "updating to rev %d because of new feature flag, "
706 "running e2fsck is recommended",
709 es->s_first_ino = cpu_to_le32(EXT4_GOOD_OLD_FIRST_INO);
710 es->s_inode_size = cpu_to_le16(EXT4_GOOD_OLD_INODE_SIZE);
711 es->s_rev_level = cpu_to_le32(EXT4_DYNAMIC_REV);
712 /* leave es->s_feature_*compat flags alone */
713 /* es->s_uuid will be set by e2fsck if empty */
716 * The rest of the superblock fields should be zero, and if not it
717 * means they are likely already in use, so leave them alone. We
718 * can leave it up to e2fsck to clean up any inconsistencies there.
723 * Open the external journal device
725 static struct block_device *ext4_blkdev_get(dev_t dev, struct super_block *sb)
727 struct block_device *bdev;
728 char b[BDEVNAME_SIZE];
730 bdev = blkdev_get_by_dev(dev, FMODE_READ|FMODE_WRITE|FMODE_EXCL, sb);
736 ext4_msg(sb, KERN_ERR, "failed to open journal device %s: %ld",
737 __bdevname(dev, b), PTR_ERR(bdev));
742 * Release the journal device
744 static void ext4_blkdev_put(struct block_device *bdev)
746 blkdev_put(bdev, FMODE_READ|FMODE_WRITE|FMODE_EXCL);
749 static void ext4_blkdev_remove(struct ext4_sb_info *sbi)
751 struct block_device *bdev;
752 bdev = sbi->journal_bdev;
754 ext4_blkdev_put(bdev);
755 sbi->journal_bdev = NULL;
759 static inline struct inode *orphan_list_entry(struct list_head *l)
761 return &list_entry(l, struct ext4_inode_info, i_orphan)->vfs_inode;
764 static void dump_orphan_list(struct super_block *sb, struct ext4_sb_info *sbi)
768 ext4_msg(sb, KERN_ERR, "sb orphan head is %d",
769 le32_to_cpu(sbi->s_es->s_last_orphan));
771 printk(KERN_ERR "sb_info orphan list:\n");
772 list_for_each(l, &sbi->s_orphan) {
773 struct inode *inode = orphan_list_entry(l);
775 "inode %s:%lu at %p: mode %o, nlink %d, next %d\n",
776 inode->i_sb->s_id, inode->i_ino, inode,
777 inode->i_mode, inode->i_nlink,
782 static void ext4_put_super(struct super_block *sb)
784 struct ext4_sb_info *sbi = EXT4_SB(sb);
785 struct ext4_super_block *es = sbi->s_es;
788 ext4_unregister_li_request(sb);
789 dquot_disable(sb, -1, DQUOT_USAGE_ENABLED | DQUOT_LIMITS_ENABLED);
791 flush_workqueue(sbi->rsv_conversion_wq);
792 destroy_workqueue(sbi->rsv_conversion_wq);
794 if (sbi->s_journal) {
795 err = jbd2_journal_destroy(sbi->s_journal);
796 sbi->s_journal = NULL;
798 ext4_abort(sb, "Couldn't clean up the journal");
801 ext4_es_unregister_shrinker(sbi);
802 del_timer_sync(&sbi->s_err_report);
803 ext4_release_system_zone(sb);
805 ext4_ext_release(sb);
806 ext4_xattr_put_super(sb);
808 if (!(sb->s_flags & MS_RDONLY)) {
809 EXT4_CLEAR_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
810 es->s_state = cpu_to_le16(sbi->s_mount_state);
812 if (!(sb->s_flags & MS_RDONLY))
813 ext4_commit_super(sb, 1);
816 remove_proc_entry("options", sbi->s_proc);
817 remove_proc_entry(sb->s_id, ext4_proc_root);
819 kobject_del(&sbi->s_kobj);
821 for (i = 0; i < sbi->s_gdb_count; i++)
822 brelse(sbi->s_group_desc[i]);
823 ext4_kvfree(sbi->s_group_desc);
824 ext4_kvfree(sbi->s_flex_groups);
825 percpu_counter_destroy(&sbi->s_freeclusters_counter);
826 percpu_counter_destroy(&sbi->s_freeinodes_counter);
827 percpu_counter_destroy(&sbi->s_dirs_counter);
828 percpu_counter_destroy(&sbi->s_dirtyclusters_counter);
829 percpu_counter_destroy(&sbi->s_extent_cache_cnt);
832 for (i = 0; i < MAXQUOTAS; i++)
833 kfree(sbi->s_qf_names[i]);
836 /* Debugging code just in case the in-memory inode orphan list
837 * isn't empty. The on-disk one can be non-empty if we've
838 * detected an error and taken the fs readonly, but the
839 * in-memory list had better be clean by this point. */
840 if (!list_empty(&sbi->s_orphan))
841 dump_orphan_list(sb, sbi);
842 J_ASSERT(list_empty(&sbi->s_orphan));
844 sync_blockdev(sb->s_bdev);
845 invalidate_bdev(sb->s_bdev);
846 if (sbi->journal_bdev && sbi->journal_bdev != sb->s_bdev) {
848 * Invalidate the journal device's buffers. We don't want them
849 * floating about in memory - the physical journal device may
850 * hotswapped, and it breaks the `ro-after' testing code.
852 sync_blockdev(sbi->journal_bdev);
853 invalidate_bdev(sbi->journal_bdev);
854 ext4_blkdev_remove(sbi);
856 if (sbi->s_mb_cache) {
857 ext4_xattr_destroy_cache(sbi->s_mb_cache);
858 sbi->s_mb_cache = NULL;
861 kthread_stop(sbi->s_mmp_tsk);
862 sb->s_fs_info = NULL;
864 * Now that we are completely done shutting down the
865 * superblock, we need to actually destroy the kobject.
867 kobject_put(&sbi->s_kobj);
868 wait_for_completion(&sbi->s_kobj_unregister);
869 if (sbi->s_chksum_driver)
870 crypto_free_shash(sbi->s_chksum_driver);
871 kfree(sbi->s_blockgroup_lock);
875 static struct kmem_cache *ext4_inode_cachep;
878 * Called inside transaction, so use GFP_NOFS
880 static struct inode *ext4_alloc_inode(struct super_block *sb)
882 struct ext4_inode_info *ei;
884 ei = kmem_cache_alloc(ext4_inode_cachep, GFP_NOFS);
888 ei->vfs_inode.i_version = 1;
889 spin_lock_init(&ei->i_raw_lock);
890 INIT_LIST_HEAD(&ei->i_prealloc_list);
891 spin_lock_init(&ei->i_prealloc_lock);
892 ext4_es_init_tree(&ei->i_es_tree);
893 rwlock_init(&ei->i_es_lock);
894 INIT_LIST_HEAD(&ei->i_es_lru);
896 ei->i_touch_when = 0;
897 ei->i_reserved_data_blocks = 0;
898 ei->i_reserved_meta_blocks = 0;
899 ei->i_allocated_meta_blocks = 0;
900 ei->i_da_metadata_calc_len = 0;
901 ei->i_da_metadata_calc_last_lblock = 0;
902 spin_lock_init(&(ei->i_block_reservation_lock));
904 ei->i_reserved_quota = 0;
907 INIT_LIST_HEAD(&ei->i_rsv_conversion_list);
908 spin_lock_init(&ei->i_completed_io_lock);
910 ei->i_datasync_tid = 0;
911 atomic_set(&ei->i_ioend_count, 0);
912 atomic_set(&ei->i_unwritten, 0);
913 INIT_WORK(&ei->i_rsv_conversion_work, ext4_end_io_rsv_work);
915 return &ei->vfs_inode;
918 static int ext4_drop_inode(struct inode *inode)
920 int drop = generic_drop_inode(inode);
922 trace_ext4_drop_inode(inode, drop);
926 static void ext4_i_callback(struct rcu_head *head)
928 struct inode *inode = container_of(head, struct inode, i_rcu);
929 kmem_cache_free(ext4_inode_cachep, EXT4_I(inode));
932 static void ext4_destroy_inode(struct inode *inode)
934 if (!list_empty(&(EXT4_I(inode)->i_orphan))) {
935 ext4_msg(inode->i_sb, KERN_ERR,
936 "Inode %lu (%p): orphan list check failed!",
937 inode->i_ino, EXT4_I(inode));
938 print_hex_dump(KERN_INFO, "", DUMP_PREFIX_ADDRESS, 16, 4,
939 EXT4_I(inode), sizeof(struct ext4_inode_info),
943 call_rcu(&inode->i_rcu, ext4_i_callback);
946 static void init_once(void *foo)
948 struct ext4_inode_info *ei = (struct ext4_inode_info *) foo;
950 INIT_LIST_HEAD(&ei->i_orphan);
951 init_rwsem(&ei->xattr_sem);
952 init_rwsem(&ei->i_data_sem);
953 inode_init_once(&ei->vfs_inode);
956 static int __init init_inodecache(void)
958 ext4_inode_cachep = kmem_cache_create("ext4_inode_cache",
959 sizeof(struct ext4_inode_info),
960 0, (SLAB_RECLAIM_ACCOUNT|
963 if (ext4_inode_cachep == NULL)
968 static void destroy_inodecache(void)
971 * Make sure all delayed rcu free inodes are flushed before we
975 kmem_cache_destroy(ext4_inode_cachep);
978 void ext4_clear_inode(struct inode *inode)
980 invalidate_inode_buffers(inode);
983 ext4_discard_preallocations(inode);
984 ext4_es_remove_extent(inode, 0, EXT_MAX_BLOCKS);
985 ext4_es_lru_del(inode);
986 if (EXT4_I(inode)->jinode) {
987 jbd2_journal_release_jbd_inode(EXT4_JOURNAL(inode),
988 EXT4_I(inode)->jinode);
989 jbd2_free_inode(EXT4_I(inode)->jinode);
990 EXT4_I(inode)->jinode = NULL;
994 static struct inode *ext4_nfs_get_inode(struct super_block *sb,
995 u64 ino, u32 generation)
999 if (ino < EXT4_FIRST_INO(sb) && ino != EXT4_ROOT_INO)
1000 return ERR_PTR(-ESTALE);
1001 if (ino > le32_to_cpu(EXT4_SB(sb)->s_es->s_inodes_count))
1002 return ERR_PTR(-ESTALE);
1004 /* iget isn't really right if the inode is currently unallocated!!
1006 * ext4_read_inode will return a bad_inode if the inode had been
1007 * deleted, so we should be safe.
1009 * Currently we don't know the generation for parent directory, so
1010 * a generation of 0 means "accept any"
1012 inode = ext4_iget_normal(sb, ino);
1014 return ERR_CAST(inode);
1015 if (generation && inode->i_generation != generation) {
1017 return ERR_PTR(-ESTALE);
1023 static struct dentry *ext4_fh_to_dentry(struct super_block *sb, struct fid *fid,
1024 int fh_len, int fh_type)
1026 return generic_fh_to_dentry(sb, fid, fh_len, fh_type,
1027 ext4_nfs_get_inode);
1030 static struct dentry *ext4_fh_to_parent(struct super_block *sb, struct fid *fid,
1031 int fh_len, int fh_type)
1033 return generic_fh_to_parent(sb, fid, fh_len, fh_type,
1034 ext4_nfs_get_inode);
1038 * Try to release metadata pages (indirect blocks, directories) which are
1039 * mapped via the block device. Since these pages could have journal heads
1040 * which would prevent try_to_free_buffers() from freeing them, we must use
1041 * jbd2 layer's try_to_free_buffers() function to release them.
1043 static int bdev_try_to_free_page(struct super_block *sb, struct page *page,
1046 journal_t *journal = EXT4_SB(sb)->s_journal;
1048 WARN_ON(PageChecked(page));
1049 if (!page_has_buffers(page))
1052 return jbd2_journal_try_to_free_buffers(journal, page,
1053 wait & ~__GFP_WAIT);
1054 return try_to_free_buffers(page);
1058 #define QTYPE2NAME(t) ((t) == USRQUOTA ? "user" : "group")
1059 #define QTYPE2MOPT(on, t) ((t) == USRQUOTA?((on)##USRJQUOTA):((on)##GRPJQUOTA))
1061 static int ext4_write_dquot(struct dquot *dquot);
1062 static int ext4_acquire_dquot(struct dquot *dquot);
1063 static int ext4_release_dquot(struct dquot *dquot);
1064 static int ext4_mark_dquot_dirty(struct dquot *dquot);
1065 static int ext4_write_info(struct super_block *sb, int type);
1066 static int ext4_quota_on(struct super_block *sb, int type, int format_id,
1068 static int ext4_quota_on_sysfile(struct super_block *sb, int type,
1070 static int ext4_quota_off(struct super_block *sb, int type);
1071 static int ext4_quota_off_sysfile(struct super_block *sb, int type);
1072 static int ext4_quota_on_mount(struct super_block *sb, int type);
1073 static ssize_t ext4_quota_read(struct super_block *sb, int type, char *data,
1074 size_t len, loff_t off);
1075 static ssize_t ext4_quota_write(struct super_block *sb, int type,
1076 const char *data, size_t len, loff_t off);
1077 static int ext4_quota_enable(struct super_block *sb, int type, int format_id,
1078 unsigned int flags);
1079 static int ext4_enable_quotas(struct super_block *sb);
1081 static const struct dquot_operations ext4_quota_operations = {
1082 .get_reserved_space = ext4_get_reserved_space,
1083 .write_dquot = ext4_write_dquot,
1084 .acquire_dquot = ext4_acquire_dquot,
1085 .release_dquot = ext4_release_dquot,
1086 .mark_dirty = ext4_mark_dquot_dirty,
1087 .write_info = ext4_write_info,
1088 .alloc_dquot = dquot_alloc,
1089 .destroy_dquot = dquot_destroy,
1092 static const struct quotactl_ops ext4_qctl_operations = {
1093 .quota_on = ext4_quota_on,
1094 .quota_off = ext4_quota_off,
1095 .quota_sync = dquot_quota_sync,
1096 .get_info = dquot_get_dqinfo,
1097 .set_info = dquot_set_dqinfo,
1098 .get_dqblk = dquot_get_dqblk,
1099 .set_dqblk = dquot_set_dqblk
1102 static const struct quotactl_ops ext4_qctl_sysfile_operations = {
1103 .quota_on_meta = ext4_quota_on_sysfile,
1104 .quota_off = ext4_quota_off_sysfile,
1105 .quota_sync = dquot_quota_sync,
1106 .get_info = dquot_get_dqinfo,
1107 .set_info = dquot_set_dqinfo,
1108 .get_dqblk = dquot_get_dqblk,
1109 .set_dqblk = dquot_set_dqblk
1113 static const struct super_operations ext4_sops = {
1114 .alloc_inode = ext4_alloc_inode,
1115 .destroy_inode = ext4_destroy_inode,
1116 .write_inode = ext4_write_inode,
1117 .dirty_inode = ext4_dirty_inode,
1118 .drop_inode = ext4_drop_inode,
1119 .evict_inode = ext4_evict_inode,
1120 .put_super = ext4_put_super,
1121 .sync_fs = ext4_sync_fs,
1122 .freeze_fs = ext4_freeze,
1123 .unfreeze_fs = ext4_unfreeze,
1124 .statfs = ext4_statfs,
1125 .remount_fs = ext4_remount,
1126 .show_options = ext4_show_options,
1128 .quota_read = ext4_quota_read,
1129 .quota_write = ext4_quota_write,
1131 .bdev_try_to_free_page = bdev_try_to_free_page,
1134 static const struct super_operations ext4_nojournal_sops = {
1135 .alloc_inode = ext4_alloc_inode,
1136 .destroy_inode = ext4_destroy_inode,
1137 .write_inode = ext4_write_inode,
1138 .dirty_inode = ext4_dirty_inode,
1139 .drop_inode = ext4_drop_inode,
1140 .evict_inode = ext4_evict_inode,
1141 .sync_fs = ext4_sync_fs_nojournal,
1142 .put_super = ext4_put_super,
1143 .statfs = ext4_statfs,
1144 .remount_fs = ext4_remount,
1145 .show_options = ext4_show_options,
1147 .quota_read = ext4_quota_read,
1148 .quota_write = ext4_quota_write,
1150 .bdev_try_to_free_page = bdev_try_to_free_page,
1153 static const struct export_operations ext4_export_ops = {
1154 .fh_to_dentry = ext4_fh_to_dentry,
1155 .fh_to_parent = ext4_fh_to_parent,
1156 .get_parent = ext4_get_parent,
1160 Opt_bsd_df, Opt_minix_df, Opt_grpid, Opt_nogrpid,
1161 Opt_resgid, Opt_resuid, Opt_sb, Opt_err_cont, Opt_err_panic, Opt_err_ro,
1162 Opt_nouid32, Opt_debug, Opt_removed,
1163 Opt_user_xattr, Opt_nouser_xattr, Opt_acl, Opt_noacl,
1164 Opt_auto_da_alloc, Opt_noauto_da_alloc, Opt_noload,
1165 Opt_commit, Opt_min_batch_time, Opt_max_batch_time, Opt_journal_dev,
1166 Opt_journal_path, Opt_journal_checksum, Opt_journal_async_commit,
1167 Opt_abort, Opt_data_journal, Opt_data_ordered, Opt_data_writeback,
1168 Opt_data_err_abort, Opt_data_err_ignore,
1169 Opt_usrjquota, Opt_grpjquota, Opt_offusrjquota, Opt_offgrpjquota,
1170 Opt_jqfmt_vfsold, Opt_jqfmt_vfsv0, Opt_jqfmt_vfsv1, Opt_quota,
1171 Opt_noquota, Opt_barrier, Opt_nobarrier, Opt_err,
1172 Opt_usrquota, Opt_grpquota, Opt_i_version,
1173 Opt_stripe, Opt_delalloc, Opt_nodelalloc, Opt_mblk_io_submit,
1174 Opt_nomblk_io_submit, Opt_block_validity, Opt_noblock_validity,
1175 Opt_inode_readahead_blks, Opt_journal_ioprio,
1176 Opt_dioread_nolock, Opt_dioread_lock,
1177 Opt_discard, Opt_nodiscard, Opt_init_itable, Opt_noinit_itable,
1178 Opt_max_dir_size_kb,
1181 static const match_table_t tokens = {
1182 {Opt_bsd_df, "bsddf"},
1183 {Opt_minix_df, "minixdf"},
1184 {Opt_grpid, "grpid"},
1185 {Opt_grpid, "bsdgroups"},
1186 {Opt_nogrpid, "nogrpid"},
1187 {Opt_nogrpid, "sysvgroups"},
1188 {Opt_resgid, "resgid=%u"},
1189 {Opt_resuid, "resuid=%u"},
1191 {Opt_err_cont, "errors=continue"},
1192 {Opt_err_panic, "errors=panic"},
1193 {Opt_err_ro, "errors=remount-ro"},
1194 {Opt_nouid32, "nouid32"},
1195 {Opt_debug, "debug"},
1196 {Opt_removed, "oldalloc"},
1197 {Opt_removed, "orlov"},
1198 {Opt_user_xattr, "user_xattr"},
1199 {Opt_nouser_xattr, "nouser_xattr"},
1201 {Opt_noacl, "noacl"},
1202 {Opt_noload, "norecovery"},
1203 {Opt_noload, "noload"},
1204 {Opt_removed, "nobh"},
1205 {Opt_removed, "bh"},
1206 {Opt_commit, "commit=%u"},
1207 {Opt_min_batch_time, "min_batch_time=%u"},
1208 {Opt_max_batch_time, "max_batch_time=%u"},
1209 {Opt_journal_dev, "journal_dev=%u"},
1210 {Opt_journal_path, "journal_path=%s"},
1211 {Opt_journal_checksum, "journal_checksum"},
1212 {Opt_journal_async_commit, "journal_async_commit"},
1213 {Opt_abort, "abort"},
1214 {Opt_data_journal, "data=journal"},
1215 {Opt_data_ordered, "data=ordered"},
1216 {Opt_data_writeback, "data=writeback"},
1217 {Opt_data_err_abort, "data_err=abort"},
1218 {Opt_data_err_ignore, "data_err=ignore"},
1219 {Opt_offusrjquota, "usrjquota="},
1220 {Opt_usrjquota, "usrjquota=%s"},
1221 {Opt_offgrpjquota, "grpjquota="},
1222 {Opt_grpjquota, "grpjquota=%s"},
1223 {Opt_jqfmt_vfsold, "jqfmt=vfsold"},
1224 {Opt_jqfmt_vfsv0, "jqfmt=vfsv0"},
1225 {Opt_jqfmt_vfsv1, "jqfmt=vfsv1"},
1226 {Opt_grpquota, "grpquota"},
1227 {Opt_noquota, "noquota"},
1228 {Opt_quota, "quota"},
1229 {Opt_usrquota, "usrquota"},
1230 {Opt_barrier, "barrier=%u"},
1231 {Opt_barrier, "barrier"},
1232 {Opt_nobarrier, "nobarrier"},
1233 {Opt_i_version, "i_version"},
1234 {Opt_stripe, "stripe=%u"},
1235 {Opt_delalloc, "delalloc"},
1236 {Opt_nodelalloc, "nodelalloc"},
1237 {Opt_removed, "mblk_io_submit"},
1238 {Opt_removed, "nomblk_io_submit"},
1239 {Opt_block_validity, "block_validity"},
1240 {Opt_noblock_validity, "noblock_validity"},
1241 {Opt_inode_readahead_blks, "inode_readahead_blks=%u"},
1242 {Opt_journal_ioprio, "journal_ioprio=%u"},
1243 {Opt_auto_da_alloc, "auto_da_alloc=%u"},
1244 {Opt_auto_da_alloc, "auto_da_alloc"},
1245 {Opt_noauto_da_alloc, "noauto_da_alloc"},
1246 {Opt_dioread_nolock, "dioread_nolock"},
1247 {Opt_dioread_lock, "dioread_lock"},
1248 {Opt_discard, "discard"},
1249 {Opt_nodiscard, "nodiscard"},
1250 {Opt_init_itable, "init_itable=%u"},
1251 {Opt_init_itable, "init_itable"},
1252 {Opt_noinit_itable, "noinit_itable"},
1253 {Opt_max_dir_size_kb, "max_dir_size_kb=%u"},
1254 {Opt_removed, "check=none"}, /* mount option from ext2/3 */
1255 {Opt_removed, "nocheck"}, /* mount option from ext2/3 */
1256 {Opt_removed, "reservation"}, /* mount option from ext2/3 */
1257 {Opt_removed, "noreservation"}, /* mount option from ext2/3 */
1258 {Opt_removed, "journal=%u"}, /* mount option from ext2/3 */
1262 static ext4_fsblk_t get_sb_block(void **data)
1264 ext4_fsblk_t sb_block;
1265 char *options = (char *) *data;
1267 if (!options || strncmp(options, "sb=", 3) != 0)
1268 return 1; /* Default location */
1271 /* TODO: use simple_strtoll with >32bit ext4 */
1272 sb_block = simple_strtoul(options, &options, 0);
1273 if (*options && *options != ',') {
1274 printk(KERN_ERR "EXT4-fs: Invalid sb specification: %s\n",
1278 if (*options == ',')
1280 *data = (void *) options;
1285 #define DEFAULT_JOURNAL_IOPRIO (IOPRIO_PRIO_VALUE(IOPRIO_CLASS_BE, 3))
1286 static char deprecated_msg[] = "Mount option \"%s\" will be removed by %s\n"
1287 "Contact linux-ext4@vger.kernel.org if you think we should keep it.\n";
1290 static int set_qf_name(struct super_block *sb, int qtype, substring_t *args)
1292 struct ext4_sb_info *sbi = EXT4_SB(sb);
1296 if (sb_any_quota_loaded(sb) &&
1297 !sbi->s_qf_names[qtype]) {
1298 ext4_msg(sb, KERN_ERR,
1299 "Cannot change journaled "
1300 "quota options when quota turned on");
1303 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_QUOTA)) {
1304 ext4_msg(sb, KERN_ERR, "Cannot set journaled quota options "
1305 "when QUOTA feature is enabled");
1308 qname = match_strdup(args);
1310 ext4_msg(sb, KERN_ERR,
1311 "Not enough memory for storing quotafile name");
1314 if (sbi->s_qf_names[qtype]) {
1315 if (strcmp(sbi->s_qf_names[qtype], qname) == 0)
1318 ext4_msg(sb, KERN_ERR,
1319 "%s quota file already specified",
1323 if (strchr(qname, '/')) {
1324 ext4_msg(sb, KERN_ERR,
1325 "quotafile must be on filesystem root");
1328 sbi->s_qf_names[qtype] = qname;
1336 static int clear_qf_name(struct super_block *sb, int qtype)
1339 struct ext4_sb_info *sbi = EXT4_SB(sb);
1341 if (sb_any_quota_loaded(sb) &&
1342 sbi->s_qf_names[qtype]) {
1343 ext4_msg(sb, KERN_ERR, "Cannot change journaled quota options"
1344 " when quota turned on");
1347 kfree(sbi->s_qf_names[qtype]);
1348 sbi->s_qf_names[qtype] = NULL;
1353 #define MOPT_SET 0x0001
1354 #define MOPT_CLEAR 0x0002
1355 #define MOPT_NOSUPPORT 0x0004
1356 #define MOPT_EXPLICIT 0x0008
1357 #define MOPT_CLEAR_ERR 0x0010
1358 #define MOPT_GTE0 0x0020
1361 #define MOPT_QFMT 0x0040
1363 #define MOPT_Q MOPT_NOSUPPORT
1364 #define MOPT_QFMT MOPT_NOSUPPORT
1366 #define MOPT_DATAJ 0x0080
1367 #define MOPT_NO_EXT2 0x0100
1368 #define MOPT_NO_EXT3 0x0200
1369 #define MOPT_EXT4_ONLY (MOPT_NO_EXT2 | MOPT_NO_EXT3)
1370 #define MOPT_STRING 0x0400
1372 static const struct mount_opts {
1376 } ext4_mount_opts[] = {
1377 {Opt_minix_df, EXT4_MOUNT_MINIX_DF, MOPT_SET},
1378 {Opt_bsd_df, EXT4_MOUNT_MINIX_DF, MOPT_CLEAR},
1379 {Opt_grpid, EXT4_MOUNT_GRPID, MOPT_SET},
1380 {Opt_nogrpid, EXT4_MOUNT_GRPID, MOPT_CLEAR},
1381 {Opt_block_validity, EXT4_MOUNT_BLOCK_VALIDITY, MOPT_SET},
1382 {Opt_noblock_validity, EXT4_MOUNT_BLOCK_VALIDITY, MOPT_CLEAR},
1383 {Opt_dioread_nolock, EXT4_MOUNT_DIOREAD_NOLOCK,
1384 MOPT_EXT4_ONLY | MOPT_SET},
1385 {Opt_dioread_lock, EXT4_MOUNT_DIOREAD_NOLOCK,
1386 MOPT_EXT4_ONLY | MOPT_CLEAR},
1387 {Opt_discard, EXT4_MOUNT_DISCARD, MOPT_SET},
1388 {Opt_nodiscard, EXT4_MOUNT_DISCARD, MOPT_CLEAR},
1389 {Opt_delalloc, EXT4_MOUNT_DELALLOC,
1390 MOPT_EXT4_ONLY | MOPT_SET | MOPT_EXPLICIT},
1391 {Opt_nodelalloc, EXT4_MOUNT_DELALLOC,
1392 MOPT_EXT4_ONLY | MOPT_CLEAR},
1393 {Opt_journal_checksum, EXT4_MOUNT_JOURNAL_CHECKSUM,
1394 MOPT_EXT4_ONLY | MOPT_SET},
1395 {Opt_journal_async_commit, (EXT4_MOUNT_JOURNAL_ASYNC_COMMIT |
1396 EXT4_MOUNT_JOURNAL_CHECKSUM),
1397 MOPT_EXT4_ONLY | MOPT_SET},
1398 {Opt_noload, EXT4_MOUNT_NOLOAD, MOPT_NO_EXT2 | MOPT_SET},
1399 {Opt_err_panic, EXT4_MOUNT_ERRORS_PANIC, MOPT_SET | MOPT_CLEAR_ERR},
1400 {Opt_err_ro, EXT4_MOUNT_ERRORS_RO, MOPT_SET | MOPT_CLEAR_ERR},
1401 {Opt_err_cont, EXT4_MOUNT_ERRORS_CONT, MOPT_SET | MOPT_CLEAR_ERR},
1402 {Opt_data_err_abort, EXT4_MOUNT_DATA_ERR_ABORT,
1403 MOPT_NO_EXT2 | MOPT_SET},
1404 {Opt_data_err_ignore, EXT4_MOUNT_DATA_ERR_ABORT,
1405 MOPT_NO_EXT2 | MOPT_CLEAR},
1406 {Opt_barrier, EXT4_MOUNT_BARRIER, MOPT_SET},
1407 {Opt_nobarrier, EXT4_MOUNT_BARRIER, MOPT_CLEAR},
1408 {Opt_noauto_da_alloc, EXT4_MOUNT_NO_AUTO_DA_ALLOC, MOPT_SET},
1409 {Opt_auto_da_alloc, EXT4_MOUNT_NO_AUTO_DA_ALLOC, MOPT_CLEAR},
1410 {Opt_noinit_itable, EXT4_MOUNT_INIT_INODE_TABLE, MOPT_CLEAR},
1411 {Opt_commit, 0, MOPT_GTE0},
1412 {Opt_max_batch_time, 0, MOPT_GTE0},
1413 {Opt_min_batch_time, 0, MOPT_GTE0},
1414 {Opt_inode_readahead_blks, 0, MOPT_GTE0},
1415 {Opt_init_itable, 0, MOPT_GTE0},
1416 {Opt_stripe, 0, MOPT_GTE0},
1417 {Opt_resuid, 0, MOPT_GTE0},
1418 {Opt_resgid, 0, MOPT_GTE0},
1419 {Opt_journal_dev, 0, MOPT_GTE0},
1420 {Opt_journal_path, 0, MOPT_STRING},
1421 {Opt_journal_ioprio, 0, MOPT_GTE0},
1422 {Opt_data_journal, EXT4_MOUNT_JOURNAL_DATA, MOPT_NO_EXT2 | MOPT_DATAJ},
1423 {Opt_data_ordered, EXT4_MOUNT_ORDERED_DATA, MOPT_NO_EXT2 | MOPT_DATAJ},
1424 {Opt_data_writeback, EXT4_MOUNT_WRITEBACK_DATA,
1425 MOPT_NO_EXT2 | MOPT_DATAJ},
1426 {Opt_user_xattr, EXT4_MOUNT_XATTR_USER, MOPT_SET},
1427 {Opt_nouser_xattr, EXT4_MOUNT_XATTR_USER, MOPT_CLEAR},
1428 #ifdef CONFIG_EXT4_FS_POSIX_ACL
1429 {Opt_acl, EXT4_MOUNT_POSIX_ACL, MOPT_SET},
1430 {Opt_noacl, EXT4_MOUNT_POSIX_ACL, MOPT_CLEAR},
1432 {Opt_acl, 0, MOPT_NOSUPPORT},
1433 {Opt_noacl, 0, MOPT_NOSUPPORT},
1435 {Opt_nouid32, EXT4_MOUNT_NO_UID32, MOPT_SET},
1436 {Opt_debug, EXT4_MOUNT_DEBUG, MOPT_SET},
1437 {Opt_quota, EXT4_MOUNT_QUOTA | EXT4_MOUNT_USRQUOTA, MOPT_SET | MOPT_Q},
1438 {Opt_usrquota, EXT4_MOUNT_QUOTA | EXT4_MOUNT_USRQUOTA,
1440 {Opt_grpquota, EXT4_MOUNT_QUOTA | EXT4_MOUNT_GRPQUOTA,
1442 {Opt_noquota, (EXT4_MOUNT_QUOTA | EXT4_MOUNT_USRQUOTA |
1443 EXT4_MOUNT_GRPQUOTA), MOPT_CLEAR | MOPT_Q},
1444 {Opt_usrjquota, 0, MOPT_Q},
1445 {Opt_grpjquota, 0, MOPT_Q},
1446 {Opt_offusrjquota, 0, MOPT_Q},
1447 {Opt_offgrpjquota, 0, MOPT_Q},
1448 {Opt_jqfmt_vfsold, QFMT_VFS_OLD, MOPT_QFMT},
1449 {Opt_jqfmt_vfsv0, QFMT_VFS_V0, MOPT_QFMT},
1450 {Opt_jqfmt_vfsv1, QFMT_VFS_V1, MOPT_QFMT},
1451 {Opt_max_dir_size_kb, 0, MOPT_GTE0},
1455 static int handle_mount_opt(struct super_block *sb, char *opt, int token,
1456 substring_t *args, unsigned long *journal_devnum,
1457 unsigned int *journal_ioprio, int is_remount)
1459 struct ext4_sb_info *sbi = EXT4_SB(sb);
1460 const struct mount_opts *m;
1466 if (token == Opt_usrjquota)
1467 return set_qf_name(sb, USRQUOTA, &args[0]);
1468 else if (token == Opt_grpjquota)
1469 return set_qf_name(sb, GRPQUOTA, &args[0]);
1470 else if (token == Opt_offusrjquota)
1471 return clear_qf_name(sb, USRQUOTA);
1472 else if (token == Opt_offgrpjquota)
1473 return clear_qf_name(sb, GRPQUOTA);
1477 case Opt_nouser_xattr:
1478 ext4_msg(sb, KERN_WARNING, deprecated_msg, opt, "3.5");
1481 return 1; /* handled by get_sb_block() */
1483 ext4_msg(sb, KERN_WARNING, "Ignoring removed %s option", opt);
1486 sbi->s_mount_flags |= EXT4_MF_FS_ABORTED;
1489 sb->s_flags |= MS_I_VERSION;
1493 for (m = ext4_mount_opts; m->token != Opt_err; m++)
1494 if (token == m->token)
1497 if (m->token == Opt_err) {
1498 ext4_msg(sb, KERN_ERR, "Unrecognized mount option \"%s\" "
1499 "or missing value", opt);
1503 if ((m->flags & MOPT_NO_EXT2) && IS_EXT2_SB(sb)) {
1504 ext4_msg(sb, KERN_ERR,
1505 "Mount option \"%s\" incompatible with ext2", opt);
1508 if ((m->flags & MOPT_NO_EXT3) && IS_EXT3_SB(sb)) {
1509 ext4_msg(sb, KERN_ERR,
1510 "Mount option \"%s\" incompatible with ext3", opt);
1514 if (args->from && !(m->flags & MOPT_STRING) && match_int(args, &arg))
1516 if (args->from && (m->flags & MOPT_GTE0) && (arg < 0))
1518 if (m->flags & MOPT_EXPLICIT)
1519 set_opt2(sb, EXPLICIT_DELALLOC);
1520 if (m->flags & MOPT_CLEAR_ERR)
1521 clear_opt(sb, ERRORS_MASK);
1522 if (token == Opt_noquota && sb_any_quota_loaded(sb)) {
1523 ext4_msg(sb, KERN_ERR, "Cannot change quota "
1524 "options when quota turned on");
1528 if (m->flags & MOPT_NOSUPPORT) {
1529 ext4_msg(sb, KERN_ERR, "%s option not supported", opt);
1530 } else if (token == Opt_commit) {
1532 arg = JBD2_DEFAULT_MAX_COMMIT_AGE;
1533 sbi->s_commit_interval = HZ * arg;
1534 } else if (token == Opt_max_batch_time) {
1535 sbi->s_max_batch_time = arg;
1536 } else if (token == Opt_min_batch_time) {
1537 sbi->s_min_batch_time = arg;
1538 } else if (token == Opt_inode_readahead_blks) {
1539 if (arg && (arg > (1 << 30) || !is_power_of_2(arg))) {
1540 ext4_msg(sb, KERN_ERR,
1541 "EXT4-fs: inode_readahead_blks must be "
1542 "0 or a power of 2 smaller than 2^31");
1545 sbi->s_inode_readahead_blks = arg;
1546 } else if (token == Opt_init_itable) {
1547 set_opt(sb, INIT_INODE_TABLE);
1549 arg = EXT4_DEF_LI_WAIT_MULT;
1550 sbi->s_li_wait_mult = arg;
1551 } else if (token == Opt_max_dir_size_kb) {
1552 sbi->s_max_dir_size_kb = arg;
1553 } else if (token == Opt_stripe) {
1554 sbi->s_stripe = arg;
1555 } else if (token == Opt_resuid) {
1556 uid = make_kuid(current_user_ns(), arg);
1557 if (!uid_valid(uid)) {
1558 ext4_msg(sb, KERN_ERR, "Invalid uid value %d", arg);
1561 sbi->s_resuid = uid;
1562 } else if (token == Opt_resgid) {
1563 gid = make_kgid(current_user_ns(), arg);
1564 if (!gid_valid(gid)) {
1565 ext4_msg(sb, KERN_ERR, "Invalid gid value %d", arg);
1568 sbi->s_resgid = gid;
1569 } else if (token == Opt_journal_dev) {
1571 ext4_msg(sb, KERN_ERR,
1572 "Cannot specify journal on remount");
1575 *journal_devnum = arg;
1576 } else if (token == Opt_journal_path) {
1578 struct inode *journal_inode;
1583 ext4_msg(sb, KERN_ERR,
1584 "Cannot specify journal on remount");
1587 journal_path = match_strdup(&args[0]);
1588 if (!journal_path) {
1589 ext4_msg(sb, KERN_ERR, "error: could not dup "
1590 "journal device string");
1594 error = kern_path(journal_path, LOOKUP_FOLLOW, &path);
1596 ext4_msg(sb, KERN_ERR, "error: could not find "
1597 "journal device path: error %d", error);
1598 kfree(journal_path);
1602 journal_inode = path.dentry->d_inode;
1603 if (!S_ISBLK(journal_inode->i_mode)) {
1604 ext4_msg(sb, KERN_ERR, "error: journal path %s "
1605 "is not a block device", journal_path);
1607 kfree(journal_path);
1611 *journal_devnum = new_encode_dev(journal_inode->i_rdev);
1613 kfree(journal_path);
1614 } else if (token == Opt_journal_ioprio) {
1616 ext4_msg(sb, KERN_ERR, "Invalid journal IO priority"
1621 IOPRIO_PRIO_VALUE(IOPRIO_CLASS_BE, arg);
1622 } else if (m->flags & MOPT_DATAJ) {
1624 if (!sbi->s_journal)
1625 ext4_msg(sb, KERN_WARNING, "Remounting file system with no journal so ignoring journalled data option");
1626 else if (test_opt(sb, DATA_FLAGS) != m->mount_opt) {
1627 ext4_msg(sb, KERN_ERR,
1628 "Cannot change data mode on remount");
1632 clear_opt(sb, DATA_FLAGS);
1633 sbi->s_mount_opt |= m->mount_opt;
1636 } else if (m->flags & MOPT_QFMT) {
1637 if (sb_any_quota_loaded(sb) &&
1638 sbi->s_jquota_fmt != m->mount_opt) {
1639 ext4_msg(sb, KERN_ERR, "Cannot change journaled "
1640 "quota options when quota turned on");
1643 if (EXT4_HAS_RO_COMPAT_FEATURE(sb,
1644 EXT4_FEATURE_RO_COMPAT_QUOTA)) {
1645 ext4_msg(sb, KERN_ERR,
1646 "Cannot set journaled quota options "
1647 "when QUOTA feature is enabled");
1650 sbi->s_jquota_fmt = m->mount_opt;
1655 if (m->flags & MOPT_CLEAR)
1657 else if (unlikely(!(m->flags & MOPT_SET))) {
1658 ext4_msg(sb, KERN_WARNING,
1659 "buggy handling of option %s", opt);
1664 sbi->s_mount_opt |= m->mount_opt;
1666 sbi->s_mount_opt &= ~m->mount_opt;
1671 static int parse_options(char *options, struct super_block *sb,
1672 unsigned long *journal_devnum,
1673 unsigned int *journal_ioprio,
1676 struct ext4_sb_info *sbi = EXT4_SB(sb);
1678 substring_t args[MAX_OPT_ARGS];
1684 while ((p = strsep(&options, ",")) != NULL) {
1688 * Initialize args struct so we know whether arg was
1689 * found; some options take optional arguments.
1691 args[0].to = args[0].from = NULL;
1692 token = match_token(p, tokens, args);
1693 if (handle_mount_opt(sb, p, token, args, journal_devnum,
1694 journal_ioprio, is_remount) < 0)
1698 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_QUOTA) &&
1699 (test_opt(sb, USRQUOTA) || test_opt(sb, GRPQUOTA))) {
1700 ext4_msg(sb, KERN_ERR, "Cannot set quota options when QUOTA "
1701 "feature is enabled");
1704 if (sbi->s_qf_names[USRQUOTA] || sbi->s_qf_names[GRPQUOTA]) {
1705 if (test_opt(sb, USRQUOTA) && sbi->s_qf_names[USRQUOTA])
1706 clear_opt(sb, USRQUOTA);
1708 if (test_opt(sb, GRPQUOTA) && sbi->s_qf_names[GRPQUOTA])
1709 clear_opt(sb, GRPQUOTA);
1711 if (test_opt(sb, GRPQUOTA) || test_opt(sb, USRQUOTA)) {
1712 ext4_msg(sb, KERN_ERR, "old and new quota "
1717 if (!sbi->s_jquota_fmt) {
1718 ext4_msg(sb, KERN_ERR, "journaled quota format "
1724 if (test_opt(sb, DIOREAD_NOLOCK)) {
1726 BLOCK_SIZE << le32_to_cpu(sbi->s_es->s_log_block_size);
1728 if (blocksize < PAGE_CACHE_SIZE) {
1729 ext4_msg(sb, KERN_ERR, "can't mount with "
1730 "dioread_nolock if block size != PAGE_SIZE");
1737 static inline void ext4_show_quota_options(struct seq_file *seq,
1738 struct super_block *sb)
1740 #if defined(CONFIG_QUOTA)
1741 struct ext4_sb_info *sbi = EXT4_SB(sb);
1743 if (sbi->s_jquota_fmt) {
1746 switch (sbi->s_jquota_fmt) {
1757 seq_printf(seq, ",jqfmt=%s", fmtname);
1760 if (sbi->s_qf_names[USRQUOTA])
1761 seq_show_option(seq, "usrjquota", sbi->s_qf_names[USRQUOTA]);
1763 if (sbi->s_qf_names[GRPQUOTA])
1764 seq_show_option(seq, "grpjquota", sbi->s_qf_names[GRPQUOTA]);
1768 static const char *token2str(int token)
1770 const struct match_token *t;
1772 for (t = tokens; t->token != Opt_err; t++)
1773 if (t->token == token && !strchr(t->pattern, '='))
1780 * - it's set to a non-default value OR
1781 * - if the per-sb default is different from the global default
1783 static int _ext4_show_options(struct seq_file *seq, struct super_block *sb,
1786 struct ext4_sb_info *sbi = EXT4_SB(sb);
1787 struct ext4_super_block *es = sbi->s_es;
1788 int def_errors, def_mount_opt = nodefs ? 0 : sbi->s_def_mount_opt;
1789 const struct mount_opts *m;
1790 char sep = nodefs ? '\n' : ',';
1792 #define SEQ_OPTS_PUTS(str) seq_printf(seq, "%c" str, sep)
1793 #define SEQ_OPTS_PRINT(str, arg) seq_printf(seq, "%c" str, sep, arg)
1795 if (sbi->s_sb_block != 1)
1796 SEQ_OPTS_PRINT("sb=%llu", sbi->s_sb_block);
1798 for (m = ext4_mount_opts; m->token != Opt_err; m++) {
1799 int want_set = m->flags & MOPT_SET;
1800 if (((m->flags & (MOPT_SET|MOPT_CLEAR)) == 0) ||
1801 (m->flags & MOPT_CLEAR_ERR))
1803 if (!(m->mount_opt & (sbi->s_mount_opt ^ def_mount_opt)))
1804 continue; /* skip if same as the default */
1806 (sbi->s_mount_opt & m->mount_opt) != m->mount_opt) ||
1807 (!want_set && (sbi->s_mount_opt & m->mount_opt)))
1808 continue; /* select Opt_noFoo vs Opt_Foo */
1809 SEQ_OPTS_PRINT("%s", token2str(m->token));
1812 if (nodefs || !uid_eq(sbi->s_resuid, make_kuid(&init_user_ns, EXT4_DEF_RESUID)) ||
1813 le16_to_cpu(es->s_def_resuid) != EXT4_DEF_RESUID)
1814 SEQ_OPTS_PRINT("resuid=%u",
1815 from_kuid_munged(&init_user_ns, sbi->s_resuid));
1816 if (nodefs || !gid_eq(sbi->s_resgid, make_kgid(&init_user_ns, EXT4_DEF_RESGID)) ||
1817 le16_to_cpu(es->s_def_resgid) != EXT4_DEF_RESGID)
1818 SEQ_OPTS_PRINT("resgid=%u",
1819 from_kgid_munged(&init_user_ns, sbi->s_resgid));
1820 def_errors = nodefs ? -1 : le16_to_cpu(es->s_errors);
1821 if (test_opt(sb, ERRORS_RO) && def_errors != EXT4_ERRORS_RO)
1822 SEQ_OPTS_PUTS("errors=remount-ro");
1823 if (test_opt(sb, ERRORS_CONT) && def_errors != EXT4_ERRORS_CONTINUE)
1824 SEQ_OPTS_PUTS("errors=continue");
1825 if (test_opt(sb, ERRORS_PANIC) && def_errors != EXT4_ERRORS_PANIC)
1826 SEQ_OPTS_PUTS("errors=panic");
1827 if (nodefs || sbi->s_commit_interval != JBD2_DEFAULT_MAX_COMMIT_AGE*HZ)
1828 SEQ_OPTS_PRINT("commit=%lu", sbi->s_commit_interval / HZ);
1829 if (nodefs || sbi->s_min_batch_time != EXT4_DEF_MIN_BATCH_TIME)
1830 SEQ_OPTS_PRINT("min_batch_time=%u", sbi->s_min_batch_time);
1831 if (nodefs || sbi->s_max_batch_time != EXT4_DEF_MAX_BATCH_TIME)
1832 SEQ_OPTS_PRINT("max_batch_time=%u", sbi->s_max_batch_time);
1833 if (sb->s_flags & MS_I_VERSION)
1834 SEQ_OPTS_PUTS("i_version");
1835 if (nodefs || sbi->s_stripe)
1836 SEQ_OPTS_PRINT("stripe=%lu", sbi->s_stripe);
1837 if (EXT4_MOUNT_DATA_FLAGS & (sbi->s_mount_opt ^ def_mount_opt)) {
1838 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA)
1839 SEQ_OPTS_PUTS("data=journal");
1840 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA)
1841 SEQ_OPTS_PUTS("data=ordered");
1842 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_WRITEBACK_DATA)
1843 SEQ_OPTS_PUTS("data=writeback");
1846 sbi->s_inode_readahead_blks != EXT4_DEF_INODE_READAHEAD_BLKS)
1847 SEQ_OPTS_PRINT("inode_readahead_blks=%u",
1848 sbi->s_inode_readahead_blks);
1850 if (nodefs || (test_opt(sb, INIT_INODE_TABLE) &&
1851 (sbi->s_li_wait_mult != EXT4_DEF_LI_WAIT_MULT)))
1852 SEQ_OPTS_PRINT("init_itable=%u", sbi->s_li_wait_mult);
1853 if (nodefs || sbi->s_max_dir_size_kb)
1854 SEQ_OPTS_PRINT("max_dir_size_kb=%u", sbi->s_max_dir_size_kb);
1856 ext4_show_quota_options(seq, sb);
1860 static int ext4_show_options(struct seq_file *seq, struct dentry *root)
1862 return _ext4_show_options(seq, root->d_sb, 0);
1865 static int options_seq_show(struct seq_file *seq, void *offset)
1867 struct super_block *sb = seq->private;
1870 seq_puts(seq, (sb->s_flags & MS_RDONLY) ? "ro" : "rw");
1871 rc = _ext4_show_options(seq, sb, 1);
1872 seq_puts(seq, "\n");
1876 static int options_open_fs(struct inode *inode, struct file *file)
1878 return single_open(file, options_seq_show, PDE_DATA(inode));
1881 static const struct file_operations ext4_seq_options_fops = {
1882 .owner = THIS_MODULE,
1883 .open = options_open_fs,
1885 .llseek = seq_lseek,
1886 .release = single_release,
1889 static int ext4_setup_super(struct super_block *sb, struct ext4_super_block *es,
1892 struct ext4_sb_info *sbi = EXT4_SB(sb);
1895 if (le32_to_cpu(es->s_rev_level) > EXT4_MAX_SUPP_REV) {
1896 ext4_msg(sb, KERN_ERR, "revision level too high, "
1897 "forcing read-only mode");
1902 if (!(sbi->s_mount_state & EXT4_VALID_FS))
1903 ext4_msg(sb, KERN_WARNING, "warning: mounting unchecked fs, "
1904 "running e2fsck is recommended");
1905 else if (sbi->s_mount_state & EXT4_ERROR_FS)
1906 ext4_msg(sb, KERN_WARNING,
1907 "warning: mounting fs with errors, "
1908 "running e2fsck is recommended");
1909 else if ((__s16) le16_to_cpu(es->s_max_mnt_count) > 0 &&
1910 le16_to_cpu(es->s_mnt_count) >=
1911 (unsigned short) (__s16) le16_to_cpu(es->s_max_mnt_count))
1912 ext4_msg(sb, KERN_WARNING,
1913 "warning: maximal mount count reached, "
1914 "running e2fsck is recommended");
1915 else if (le32_to_cpu(es->s_checkinterval) &&
1916 (le32_to_cpu(es->s_lastcheck) +
1917 le32_to_cpu(es->s_checkinterval) <= get_seconds()))
1918 ext4_msg(sb, KERN_WARNING,
1919 "warning: checktime reached, "
1920 "running e2fsck is recommended");
1921 if (!sbi->s_journal)
1922 es->s_state &= cpu_to_le16(~EXT4_VALID_FS);
1923 if (!(__s16) le16_to_cpu(es->s_max_mnt_count))
1924 es->s_max_mnt_count = cpu_to_le16(EXT4_DFL_MAX_MNT_COUNT);
1925 le16_add_cpu(&es->s_mnt_count, 1);
1926 es->s_mtime = cpu_to_le32(get_seconds());
1927 ext4_update_dynamic_rev(sb);
1929 EXT4_SET_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
1931 ext4_commit_super(sb, 1);
1933 if (test_opt(sb, DEBUG))
1934 printk(KERN_INFO "[EXT4 FS bs=%lu, gc=%u, "
1935 "bpg=%lu, ipg=%lu, mo=%04x, mo2=%04x]\n",
1937 sbi->s_groups_count,
1938 EXT4_BLOCKS_PER_GROUP(sb),
1939 EXT4_INODES_PER_GROUP(sb),
1940 sbi->s_mount_opt, sbi->s_mount_opt2);
1942 cleancache_init_fs(sb);
1946 int ext4_alloc_flex_bg_array(struct super_block *sb, ext4_group_t ngroup)
1948 struct ext4_sb_info *sbi = EXT4_SB(sb);
1949 struct flex_groups *new_groups;
1952 if (!sbi->s_log_groups_per_flex)
1955 size = ext4_flex_group(sbi, ngroup - 1) + 1;
1956 if (size <= sbi->s_flex_groups_allocated)
1959 size = roundup_pow_of_two(size * sizeof(struct flex_groups));
1960 new_groups = ext4_kvzalloc(size, GFP_KERNEL);
1962 ext4_msg(sb, KERN_ERR, "not enough memory for %d flex groups",
1963 size / (int) sizeof(struct flex_groups));
1967 if (sbi->s_flex_groups) {
1968 memcpy(new_groups, sbi->s_flex_groups,
1969 (sbi->s_flex_groups_allocated *
1970 sizeof(struct flex_groups)));
1971 ext4_kvfree(sbi->s_flex_groups);
1973 sbi->s_flex_groups = new_groups;
1974 sbi->s_flex_groups_allocated = size / sizeof(struct flex_groups);
1978 static int ext4_fill_flex_info(struct super_block *sb)
1980 struct ext4_sb_info *sbi = EXT4_SB(sb);
1981 struct ext4_group_desc *gdp = NULL;
1982 ext4_group_t flex_group;
1985 sbi->s_log_groups_per_flex = sbi->s_es->s_log_groups_per_flex;
1986 if (sbi->s_log_groups_per_flex < 1 || sbi->s_log_groups_per_flex > 31) {
1987 sbi->s_log_groups_per_flex = 0;
1991 err = ext4_alloc_flex_bg_array(sb, sbi->s_groups_count);
1995 for (i = 0; i < sbi->s_groups_count; i++) {
1996 gdp = ext4_get_group_desc(sb, i, NULL);
1998 flex_group = ext4_flex_group(sbi, i);
1999 atomic_add(ext4_free_inodes_count(sb, gdp),
2000 &sbi->s_flex_groups[flex_group].free_inodes);
2001 atomic64_add(ext4_free_group_clusters(sb, gdp),
2002 &sbi->s_flex_groups[flex_group].free_clusters);
2003 atomic_add(ext4_used_dirs_count(sb, gdp),
2004 &sbi->s_flex_groups[flex_group].used_dirs);
2012 static __le16 ext4_group_desc_csum(struct ext4_sb_info *sbi, __u32 block_group,
2013 struct ext4_group_desc *gdp)
2017 __le32 le_group = cpu_to_le32(block_group);
2019 if (ext4_has_metadata_csum(sbi->s_sb)) {
2020 /* Use new metadata_csum algorithm */
2024 save_csum = gdp->bg_checksum;
2025 gdp->bg_checksum = 0;
2026 csum32 = ext4_chksum(sbi, sbi->s_csum_seed, (__u8 *)&le_group,
2028 csum32 = ext4_chksum(sbi, csum32, (__u8 *)gdp,
2030 gdp->bg_checksum = save_csum;
2032 crc = csum32 & 0xFFFF;
2036 /* old crc16 code */
2037 if (!(sbi->s_es->s_feature_ro_compat &
2038 cpu_to_le32(EXT4_FEATURE_RO_COMPAT_GDT_CSUM)))
2041 offset = offsetof(struct ext4_group_desc, bg_checksum);
2043 crc = crc16(~0, sbi->s_es->s_uuid, sizeof(sbi->s_es->s_uuid));
2044 crc = crc16(crc, (__u8 *)&le_group, sizeof(le_group));
2045 crc = crc16(crc, (__u8 *)gdp, offset);
2046 offset += sizeof(gdp->bg_checksum); /* skip checksum */
2047 /* for checksum of struct ext4_group_desc do the rest...*/
2048 if ((sbi->s_es->s_feature_incompat &
2049 cpu_to_le32(EXT4_FEATURE_INCOMPAT_64BIT)) &&
2050 offset < le16_to_cpu(sbi->s_es->s_desc_size))
2051 crc = crc16(crc, (__u8 *)gdp + offset,
2052 le16_to_cpu(sbi->s_es->s_desc_size) -
2056 return cpu_to_le16(crc);
2059 int ext4_group_desc_csum_verify(struct super_block *sb, __u32 block_group,
2060 struct ext4_group_desc *gdp)
2062 if (ext4_has_group_desc_csum(sb) &&
2063 (gdp->bg_checksum != ext4_group_desc_csum(EXT4_SB(sb),
2070 void ext4_group_desc_csum_set(struct super_block *sb, __u32 block_group,
2071 struct ext4_group_desc *gdp)
2073 if (!ext4_has_group_desc_csum(sb))
2075 gdp->bg_checksum = ext4_group_desc_csum(EXT4_SB(sb), block_group, gdp);
2078 /* Called at mount-time, super-block is locked */
2079 static int ext4_check_descriptors(struct super_block *sb,
2080 ext4_fsblk_t sb_block,
2081 ext4_group_t *first_not_zeroed)
2083 struct ext4_sb_info *sbi = EXT4_SB(sb);
2084 ext4_fsblk_t first_block = le32_to_cpu(sbi->s_es->s_first_data_block);
2085 ext4_fsblk_t last_block;
2086 ext4_fsblk_t block_bitmap;
2087 ext4_fsblk_t inode_bitmap;
2088 ext4_fsblk_t inode_table;
2089 int flexbg_flag = 0;
2090 ext4_group_t i, grp = sbi->s_groups_count;
2092 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_FLEX_BG))
2095 ext4_debug("Checking group descriptors");
2097 for (i = 0; i < sbi->s_groups_count; i++) {
2098 struct ext4_group_desc *gdp = ext4_get_group_desc(sb, i, NULL);
2100 if (i == sbi->s_groups_count - 1 || flexbg_flag)
2101 last_block = ext4_blocks_count(sbi->s_es) - 1;
2103 last_block = first_block +
2104 (EXT4_BLOCKS_PER_GROUP(sb) - 1);
2106 if ((grp == sbi->s_groups_count) &&
2107 !(gdp->bg_flags & cpu_to_le16(EXT4_BG_INODE_ZEROED)))
2110 block_bitmap = ext4_block_bitmap(sb, gdp);
2111 if (block_bitmap == sb_block) {
2112 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2113 "Block bitmap for group %u overlaps "
2116 if (block_bitmap < first_block || block_bitmap > last_block) {
2117 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2118 "Block bitmap for group %u not in group "
2119 "(block %llu)!", i, block_bitmap);
2122 inode_bitmap = ext4_inode_bitmap(sb, gdp);
2123 if (inode_bitmap == sb_block) {
2124 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2125 "Inode bitmap for group %u overlaps "
2128 if (inode_bitmap < first_block || inode_bitmap > last_block) {
2129 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2130 "Inode bitmap for group %u not in group "
2131 "(block %llu)!", i, inode_bitmap);
2134 inode_table = ext4_inode_table(sb, gdp);
2135 if (inode_table == sb_block) {
2136 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2137 "Inode table for group %u overlaps "
2140 if (inode_table < first_block ||
2141 inode_table + sbi->s_itb_per_group - 1 > last_block) {
2142 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2143 "Inode table for group %u not in group "
2144 "(block %llu)!", i, inode_table);
2147 ext4_lock_group(sb, i);
2148 if (!ext4_group_desc_csum_verify(sb, i, gdp)) {
2149 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2150 "Checksum for group %u failed (%u!=%u)",
2151 i, le16_to_cpu(ext4_group_desc_csum(sbi, i,
2152 gdp)), le16_to_cpu(gdp->bg_checksum));
2153 if (!(sb->s_flags & MS_RDONLY)) {
2154 ext4_unlock_group(sb, i);
2158 ext4_unlock_group(sb, i);
2160 first_block += EXT4_BLOCKS_PER_GROUP(sb);
2162 if (NULL != first_not_zeroed)
2163 *first_not_zeroed = grp;
2165 ext4_free_blocks_count_set(sbi->s_es,
2166 EXT4_C2B(sbi, ext4_count_free_clusters(sb)));
2167 sbi->s_es->s_free_inodes_count =cpu_to_le32(ext4_count_free_inodes(sb));
2171 /* ext4_orphan_cleanup() walks a singly-linked list of inodes (starting at
2172 * the superblock) which were deleted from all directories, but held open by
2173 * a process at the time of a crash. We walk the list and try to delete these
2174 * inodes at recovery time (only with a read-write filesystem).
2176 * In order to keep the orphan inode chain consistent during traversal (in
2177 * case of crash during recovery), we link each inode into the superblock
2178 * orphan list_head and handle it the same way as an inode deletion during
2179 * normal operation (which journals the operations for us).
2181 * We only do an iget() and an iput() on each inode, which is very safe if we
2182 * accidentally point at an in-use or already deleted inode. The worst that
2183 * can happen in this case is that we get a "bit already cleared" message from
2184 * ext4_free_inode(). The only reason we would point at a wrong inode is if
2185 * e2fsck was run on this filesystem, and it must have already done the orphan
2186 * inode cleanup for us, so we can safely abort without any further action.
2188 static void ext4_orphan_cleanup(struct super_block *sb,
2189 struct ext4_super_block *es)
2191 unsigned int s_flags = sb->s_flags;
2192 int nr_orphans = 0, nr_truncates = 0;
2196 if (!es->s_last_orphan) {
2197 jbd_debug(4, "no orphan inodes to clean up\n");
2201 if (bdev_read_only(sb->s_bdev)) {
2202 ext4_msg(sb, KERN_ERR, "write access "
2203 "unavailable, skipping orphan cleanup");
2207 /* Check if feature set would not allow a r/w mount */
2208 if (!ext4_feature_set_ok(sb, 0)) {
2209 ext4_msg(sb, KERN_INFO, "Skipping orphan cleanup due to "
2210 "unknown ROCOMPAT features");
2214 if (EXT4_SB(sb)->s_mount_state & EXT4_ERROR_FS) {
2215 /* don't clear list on RO mount w/ errors */
2216 if (es->s_last_orphan && !(s_flags & MS_RDONLY)) {
2217 jbd_debug(1, "Errors on filesystem, "
2218 "clearing orphan list.\n");
2219 es->s_last_orphan = 0;
2221 jbd_debug(1, "Skipping orphan recovery on fs with errors.\n");
2225 if (s_flags & MS_RDONLY) {
2226 ext4_msg(sb, KERN_INFO, "orphan cleanup on readonly fs");
2227 sb->s_flags &= ~MS_RDONLY;
2230 /* Needed for iput() to work correctly and not trash data */
2231 sb->s_flags |= MS_ACTIVE;
2232 /* Turn on quotas so that they are updated correctly */
2233 for (i = 0; i < MAXQUOTAS; i++) {
2234 if (EXT4_SB(sb)->s_qf_names[i]) {
2235 int ret = ext4_quota_on_mount(sb, i);
2237 ext4_msg(sb, KERN_ERR,
2238 "Cannot turn on journaled "
2239 "quota: error %d", ret);
2244 while (es->s_last_orphan) {
2245 struct inode *inode;
2248 * We may have encountered an error during cleanup; if
2249 * so, skip the rest.
2251 if (EXT4_SB(sb)->s_mount_state & EXT4_ERROR_FS) {
2252 jbd_debug(1, "Skipping orphan recovery on fs with errors.\n");
2253 es->s_last_orphan = 0;
2257 inode = ext4_orphan_get(sb, le32_to_cpu(es->s_last_orphan));
2258 if (IS_ERR(inode)) {
2259 es->s_last_orphan = 0;
2263 list_add(&EXT4_I(inode)->i_orphan, &EXT4_SB(sb)->s_orphan);
2264 dquot_initialize(inode);
2265 if (inode->i_nlink) {
2266 if (test_opt(sb, DEBUG))
2267 ext4_msg(sb, KERN_DEBUG,
2268 "%s: truncating inode %lu to %lld bytes",
2269 __func__, inode->i_ino, inode->i_size);
2270 jbd_debug(2, "truncating inode %lu to %lld bytes\n",
2271 inode->i_ino, inode->i_size);
2272 mutex_lock(&inode->i_mutex);
2273 truncate_inode_pages(inode->i_mapping, inode->i_size);
2274 ext4_truncate(inode);
2275 mutex_unlock(&inode->i_mutex);
2278 if (test_opt(sb, DEBUG))
2279 ext4_msg(sb, KERN_DEBUG,
2280 "%s: deleting unreferenced inode %lu",
2281 __func__, inode->i_ino);
2282 jbd_debug(2, "deleting unreferenced inode %lu\n",
2286 iput(inode); /* The delete magic happens here! */
2289 #define PLURAL(x) (x), ((x) == 1) ? "" : "s"
2292 ext4_msg(sb, KERN_INFO, "%d orphan inode%s deleted",
2293 PLURAL(nr_orphans));
2295 ext4_msg(sb, KERN_INFO, "%d truncate%s cleaned up",
2296 PLURAL(nr_truncates));
2298 /* Turn quotas off */
2299 for (i = 0; i < MAXQUOTAS; i++) {
2300 if (sb_dqopt(sb)->files[i])
2301 dquot_quota_off(sb, i);
2304 sb->s_flags = s_flags; /* Restore MS_RDONLY status */
2308 * Maximal extent format file size.
2309 * Resulting logical blkno at s_maxbytes must fit in our on-disk
2310 * extent format containers, within a sector_t, and within i_blocks
2311 * in the vfs. ext4 inode has 48 bits of i_block in fsblock units,
2312 * so that won't be a limiting factor.
2314 * However there is other limiting factor. We do store extents in the form
2315 * of starting block and length, hence the resulting length of the extent
2316 * covering maximum file size must fit into on-disk format containers as
2317 * well. Given that length is always by 1 unit bigger than max unit (because
2318 * we count 0 as well) we have to lower the s_maxbytes by one fs block.
2320 * Note, this does *not* consider any metadata overhead for vfs i_blocks.
2322 static loff_t ext4_max_size(int blkbits, int has_huge_files)
2325 loff_t upper_limit = MAX_LFS_FILESIZE;
2327 /* small i_blocks in vfs inode? */
2328 if (!has_huge_files || sizeof(blkcnt_t) < sizeof(u64)) {
2330 * CONFIG_LBDAF is not enabled implies the inode
2331 * i_block represent total blocks in 512 bytes
2332 * 32 == size of vfs inode i_blocks * 8
2334 upper_limit = (1LL << 32) - 1;
2336 /* total blocks in file system block size */
2337 upper_limit >>= (blkbits - 9);
2338 upper_limit <<= blkbits;
2342 * 32-bit extent-start container, ee_block. We lower the maxbytes
2343 * by one fs block, so ee_len can cover the extent of maximum file
2346 res = (1LL << 32) - 1;
2349 /* Sanity check against vm- & vfs- imposed limits */
2350 if (res > upper_limit)
2357 * Maximal bitmap file size. There is a direct, and {,double-,triple-}indirect
2358 * block limit, and also a limit of (2^48 - 1) 512-byte sectors in i_blocks.
2359 * We need to be 1 filesystem block less than the 2^48 sector limit.
2361 static loff_t ext4_max_bitmap_size(int bits, int has_huge_files)
2363 loff_t res = EXT4_NDIR_BLOCKS;
2366 /* This is calculated to be the largest file size for a dense, block
2367 * mapped file such that the file's total number of 512-byte sectors,
2368 * including data and all indirect blocks, does not exceed (2^48 - 1).
2370 * __u32 i_blocks_lo and _u16 i_blocks_high represent the total
2371 * number of 512-byte sectors of the file.
2374 if (!has_huge_files || sizeof(blkcnt_t) < sizeof(u64)) {
2376 * !has_huge_files or CONFIG_LBDAF not enabled implies that
2377 * the inode i_block field represents total file blocks in
2378 * 2^32 512-byte sectors == size of vfs inode i_blocks * 8
2380 upper_limit = (1LL << 32) - 1;
2382 /* total blocks in file system block size */
2383 upper_limit >>= (bits - 9);
2387 * We use 48 bit ext4_inode i_blocks
2388 * With EXT4_HUGE_FILE_FL set the i_blocks
2389 * represent total number of blocks in
2390 * file system block size
2392 upper_limit = (1LL << 48) - 1;
2396 /* indirect blocks */
2398 /* double indirect blocks */
2399 meta_blocks += 1 + (1LL << (bits-2));
2400 /* tripple indirect blocks */
2401 meta_blocks += 1 + (1LL << (bits-2)) + (1LL << (2*(bits-2)));
2403 upper_limit -= meta_blocks;
2404 upper_limit <<= bits;
2406 res += 1LL << (bits-2);
2407 res += 1LL << (2*(bits-2));
2408 res += 1LL << (3*(bits-2));
2410 if (res > upper_limit)
2413 if (res > MAX_LFS_FILESIZE)
2414 res = MAX_LFS_FILESIZE;
2419 static ext4_fsblk_t descriptor_loc(struct super_block *sb,
2420 ext4_fsblk_t logical_sb_block, int nr)
2422 struct ext4_sb_info *sbi = EXT4_SB(sb);
2423 ext4_group_t bg, first_meta_bg;
2426 first_meta_bg = le32_to_cpu(sbi->s_es->s_first_meta_bg);
2428 if (!EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_META_BG) ||
2430 return logical_sb_block + nr + 1;
2431 bg = sbi->s_desc_per_block * nr;
2432 if (ext4_bg_has_super(sb, bg))
2436 * If we have a meta_bg fs with 1k blocks, group 0's GDT is at
2437 * block 2, not 1. If s_first_data_block == 0 (bigalloc is enabled
2438 * on modern mke2fs or blksize > 1k on older mke2fs) then we must
2441 if (sb->s_blocksize == 1024 && nr == 0 &&
2442 le32_to_cpu(EXT4_SB(sb)->s_es->s_first_data_block) == 0)
2445 return (has_super + ext4_group_first_block_no(sb, bg));
2449 * ext4_get_stripe_size: Get the stripe size.
2450 * @sbi: In memory super block info
2452 * If we have specified it via mount option, then
2453 * use the mount option value. If the value specified at mount time is
2454 * greater than the blocks per group use the super block value.
2455 * If the super block value is greater than blocks per group return 0.
2456 * Allocator needs it be less than blocks per group.
2459 static unsigned long ext4_get_stripe_size(struct ext4_sb_info *sbi)
2461 unsigned long stride = le16_to_cpu(sbi->s_es->s_raid_stride);
2462 unsigned long stripe_width =
2463 le32_to_cpu(sbi->s_es->s_raid_stripe_width);
2466 if (sbi->s_stripe && sbi->s_stripe <= sbi->s_blocks_per_group)
2467 ret = sbi->s_stripe;
2468 else if (stripe_width <= sbi->s_blocks_per_group)
2470 else if (stride <= sbi->s_blocks_per_group)
2476 * If the stripe width is 1, this makes no sense and
2477 * we set it to 0 to turn off stripe handling code.
2488 struct attribute attr;
2489 ssize_t (*show)(struct ext4_attr *, struct ext4_sb_info *, char *);
2490 ssize_t (*store)(struct ext4_attr *, struct ext4_sb_info *,
2491 const char *, size_t);
2498 static int parse_strtoull(const char *buf,
2499 unsigned long long max, unsigned long long *value)
2503 ret = kstrtoull(skip_spaces(buf), 0, value);
2504 if (!ret && *value > max)
2509 static ssize_t delayed_allocation_blocks_show(struct ext4_attr *a,
2510 struct ext4_sb_info *sbi,
2513 return snprintf(buf, PAGE_SIZE, "%llu\n",
2515 percpu_counter_sum(&sbi->s_dirtyclusters_counter)));
2518 static ssize_t session_write_kbytes_show(struct ext4_attr *a,
2519 struct ext4_sb_info *sbi, char *buf)
2521 struct super_block *sb = sbi->s_buddy_cache->i_sb;
2523 if (!sb->s_bdev->bd_part)
2524 return snprintf(buf, PAGE_SIZE, "0\n");
2525 return snprintf(buf, PAGE_SIZE, "%lu\n",
2526 (part_stat_read(sb->s_bdev->bd_part, sectors[1]) -
2527 sbi->s_sectors_written_start) >> 1);
2530 static ssize_t lifetime_write_kbytes_show(struct ext4_attr *a,
2531 struct ext4_sb_info *sbi, char *buf)
2533 struct super_block *sb = sbi->s_buddy_cache->i_sb;
2535 if (!sb->s_bdev->bd_part)
2536 return snprintf(buf, PAGE_SIZE, "0\n");
2537 return snprintf(buf, PAGE_SIZE, "%llu\n",
2538 (unsigned long long)(sbi->s_kbytes_written +
2539 ((part_stat_read(sb->s_bdev->bd_part, sectors[1]) -
2540 EXT4_SB(sb)->s_sectors_written_start) >> 1)));
2543 static ssize_t inode_readahead_blks_store(struct ext4_attr *a,
2544 struct ext4_sb_info *sbi,
2545 const char *buf, size_t count)
2550 ret = kstrtoul(skip_spaces(buf), 0, &t);
2554 if (t && (!is_power_of_2(t) || t > 0x40000000))
2557 sbi->s_inode_readahead_blks = t;
2561 static ssize_t sbi_ui_show(struct ext4_attr *a,
2562 struct ext4_sb_info *sbi, char *buf)
2564 unsigned int *ui = (unsigned int *) (((char *) sbi) + a->u.offset);
2566 return snprintf(buf, PAGE_SIZE, "%u\n", *ui);
2569 static ssize_t sbi_ui_store(struct ext4_attr *a,
2570 struct ext4_sb_info *sbi,
2571 const char *buf, size_t count)
2573 unsigned int *ui = (unsigned int *) (((char *) sbi) + a->u.offset);
2577 ret = kstrtoul(skip_spaces(buf), 0, &t);
2584 static ssize_t reserved_clusters_show(struct ext4_attr *a,
2585 struct ext4_sb_info *sbi, char *buf)
2587 return snprintf(buf, PAGE_SIZE, "%llu\n",
2588 (unsigned long long) atomic64_read(&sbi->s_resv_clusters));
2591 static ssize_t reserved_clusters_store(struct ext4_attr *a,
2592 struct ext4_sb_info *sbi,
2593 const char *buf, size_t count)
2595 unsigned long long val;
2598 if (parse_strtoull(buf, -1ULL, &val))
2600 ret = ext4_reserve_clusters(sbi, val);
2602 return ret ? ret : count;
2605 static ssize_t trigger_test_error(struct ext4_attr *a,
2606 struct ext4_sb_info *sbi,
2607 const char *buf, size_t count)
2611 if (!capable(CAP_SYS_ADMIN))
2614 if (len && buf[len-1] == '\n')
2618 ext4_error(sbi->s_sb, "%.*s", len, buf);
2622 static ssize_t sbi_deprecated_show(struct ext4_attr *a,
2623 struct ext4_sb_info *sbi, char *buf)
2625 return snprintf(buf, PAGE_SIZE, "%d\n", a->u.deprecated_val);
2628 #define EXT4_ATTR_OFFSET(_name,_mode,_show,_store,_elname) \
2629 static struct ext4_attr ext4_attr_##_name = { \
2630 .attr = {.name = __stringify(_name), .mode = _mode }, \
2634 .offset = offsetof(struct ext4_sb_info, _elname),\
2637 #define EXT4_ATTR(name, mode, show, store) \
2638 static struct ext4_attr ext4_attr_##name = __ATTR(name, mode, show, store)
2640 #define EXT4_INFO_ATTR(name) EXT4_ATTR(name, 0444, NULL, NULL)
2641 #define EXT4_RO_ATTR(name) EXT4_ATTR(name, 0444, name##_show, NULL)
2642 #define EXT4_RW_ATTR(name) EXT4_ATTR(name, 0644, name##_show, name##_store)
2643 #define EXT4_RW_ATTR_SBI_UI(name, elname) \
2644 EXT4_ATTR_OFFSET(name, 0644, sbi_ui_show, sbi_ui_store, elname)
2645 #define ATTR_LIST(name) &ext4_attr_##name.attr
2646 #define EXT4_DEPRECATED_ATTR(_name, _val) \
2647 static struct ext4_attr ext4_attr_##_name = { \
2648 .attr = {.name = __stringify(_name), .mode = 0444 }, \
2649 .show = sbi_deprecated_show, \
2651 .deprecated_val = _val, \
2655 EXT4_RO_ATTR(delayed_allocation_blocks);
2656 EXT4_RO_ATTR(session_write_kbytes);
2657 EXT4_RO_ATTR(lifetime_write_kbytes);
2658 EXT4_RW_ATTR(reserved_clusters);
2659 EXT4_ATTR_OFFSET(inode_readahead_blks, 0644, sbi_ui_show,
2660 inode_readahead_blks_store, s_inode_readahead_blks);
2661 EXT4_RW_ATTR_SBI_UI(inode_goal, s_inode_goal);
2662 EXT4_RW_ATTR_SBI_UI(mb_stats, s_mb_stats);
2663 EXT4_RW_ATTR_SBI_UI(mb_max_to_scan, s_mb_max_to_scan);
2664 EXT4_RW_ATTR_SBI_UI(mb_min_to_scan, s_mb_min_to_scan);
2665 EXT4_RW_ATTR_SBI_UI(mb_order2_req, s_mb_order2_reqs);
2666 EXT4_RW_ATTR_SBI_UI(mb_stream_req, s_mb_stream_request);
2667 EXT4_RW_ATTR_SBI_UI(mb_group_prealloc, s_mb_group_prealloc);
2668 EXT4_DEPRECATED_ATTR(max_writeback_mb_bump, 128);
2669 EXT4_RW_ATTR_SBI_UI(extent_max_zeroout_kb, s_extent_max_zeroout_kb);
2670 EXT4_ATTR(trigger_fs_error, 0200, NULL, trigger_test_error);
2671 EXT4_RW_ATTR_SBI_UI(err_ratelimit_interval_ms, s_err_ratelimit_state.interval);
2672 EXT4_RW_ATTR_SBI_UI(err_ratelimit_burst, s_err_ratelimit_state.burst);
2673 EXT4_RW_ATTR_SBI_UI(warning_ratelimit_interval_ms, s_warning_ratelimit_state.interval);
2674 EXT4_RW_ATTR_SBI_UI(warning_ratelimit_burst, s_warning_ratelimit_state.burst);
2675 EXT4_RW_ATTR_SBI_UI(msg_ratelimit_interval_ms, s_msg_ratelimit_state.interval);
2676 EXT4_RW_ATTR_SBI_UI(msg_ratelimit_burst, s_msg_ratelimit_state.burst);
2678 static struct attribute *ext4_attrs[] = {
2679 ATTR_LIST(delayed_allocation_blocks),
2680 ATTR_LIST(session_write_kbytes),
2681 ATTR_LIST(lifetime_write_kbytes),
2682 ATTR_LIST(reserved_clusters),
2683 ATTR_LIST(inode_readahead_blks),
2684 ATTR_LIST(inode_goal),
2685 ATTR_LIST(mb_stats),
2686 ATTR_LIST(mb_max_to_scan),
2687 ATTR_LIST(mb_min_to_scan),
2688 ATTR_LIST(mb_order2_req),
2689 ATTR_LIST(mb_stream_req),
2690 ATTR_LIST(mb_group_prealloc),
2691 ATTR_LIST(max_writeback_mb_bump),
2692 ATTR_LIST(extent_max_zeroout_kb),
2693 ATTR_LIST(trigger_fs_error),
2694 ATTR_LIST(err_ratelimit_interval_ms),
2695 ATTR_LIST(err_ratelimit_burst),
2696 ATTR_LIST(warning_ratelimit_interval_ms),
2697 ATTR_LIST(warning_ratelimit_burst),
2698 ATTR_LIST(msg_ratelimit_interval_ms),
2699 ATTR_LIST(msg_ratelimit_burst),
2703 /* Features this copy of ext4 supports */
2704 EXT4_INFO_ATTR(lazy_itable_init);
2705 EXT4_INFO_ATTR(batched_discard);
2706 EXT4_INFO_ATTR(meta_bg_resize);
2708 static struct attribute *ext4_feat_attrs[] = {
2709 ATTR_LIST(lazy_itable_init),
2710 ATTR_LIST(batched_discard),
2711 ATTR_LIST(meta_bg_resize),
2715 static ssize_t ext4_attr_show(struct kobject *kobj,
2716 struct attribute *attr, char *buf)
2718 struct ext4_sb_info *sbi = container_of(kobj, struct ext4_sb_info,
2720 struct ext4_attr *a = container_of(attr, struct ext4_attr, attr);
2722 return a->show ? a->show(a, sbi, buf) : 0;
2725 static ssize_t ext4_attr_store(struct kobject *kobj,
2726 struct attribute *attr,
2727 const char *buf, size_t len)
2729 struct ext4_sb_info *sbi = container_of(kobj, struct ext4_sb_info,
2731 struct ext4_attr *a = container_of(attr, struct ext4_attr, attr);
2733 return a->store ? a->store(a, sbi, buf, len) : 0;
2736 static void ext4_sb_release(struct kobject *kobj)
2738 struct ext4_sb_info *sbi = container_of(kobj, struct ext4_sb_info,
2740 complete(&sbi->s_kobj_unregister);
2743 static const struct sysfs_ops ext4_attr_ops = {
2744 .show = ext4_attr_show,
2745 .store = ext4_attr_store,
2748 static struct kobj_type ext4_ktype = {
2749 .default_attrs = ext4_attrs,
2750 .sysfs_ops = &ext4_attr_ops,
2751 .release = ext4_sb_release,
2754 static void ext4_feat_release(struct kobject *kobj)
2756 complete(&ext4_feat->f_kobj_unregister);
2759 static struct kobj_type ext4_feat_ktype = {
2760 .default_attrs = ext4_feat_attrs,
2761 .sysfs_ops = &ext4_attr_ops,
2762 .release = ext4_feat_release,
2766 * Check whether this filesystem can be mounted based on
2767 * the features present and the RDONLY/RDWR mount requested.
2768 * Returns 1 if this filesystem can be mounted as requested,
2769 * 0 if it cannot be.
2771 static int ext4_feature_set_ok(struct super_block *sb, int readonly)
2773 if (EXT4_HAS_INCOMPAT_FEATURE(sb, ~EXT4_FEATURE_INCOMPAT_SUPP)) {
2774 ext4_msg(sb, KERN_ERR,
2775 "Couldn't mount because of "
2776 "unsupported optional features (%x)",
2777 (le32_to_cpu(EXT4_SB(sb)->s_es->s_feature_incompat) &
2778 ~EXT4_FEATURE_INCOMPAT_SUPP));
2785 /* Check that feature set is OK for a read-write mount */
2786 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, ~EXT4_FEATURE_RO_COMPAT_SUPP)) {
2787 ext4_msg(sb, KERN_ERR, "couldn't mount RDWR because of "
2788 "unsupported optional features (%x)",
2789 (le32_to_cpu(EXT4_SB(sb)->s_es->s_feature_ro_compat) &
2790 ~EXT4_FEATURE_RO_COMPAT_SUPP));
2794 * Large file size enabled file system can only be mounted
2795 * read-write on 32-bit systems if kernel is built with CONFIG_LBDAF
2797 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_HUGE_FILE)) {
2798 if (sizeof(blkcnt_t) < sizeof(u64)) {
2799 ext4_msg(sb, KERN_ERR, "Filesystem with huge files "
2800 "cannot be mounted RDWR without "
2805 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_BIGALLOC) &&
2806 !EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_EXTENTS)) {
2807 ext4_msg(sb, KERN_ERR,
2808 "Can't support bigalloc feature without "
2809 "extents feature\n");
2813 #ifndef CONFIG_QUOTA
2814 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_QUOTA) &&
2816 ext4_msg(sb, KERN_ERR,
2817 "Filesystem with quota feature cannot be mounted RDWR "
2818 "without CONFIG_QUOTA");
2821 #endif /* CONFIG_QUOTA */
2826 * This function is called once a day if we have errors logged
2827 * on the file system
2829 static void print_daily_error_info(unsigned long arg)
2831 struct super_block *sb = (struct super_block *) arg;
2832 struct ext4_sb_info *sbi;
2833 struct ext4_super_block *es;
2838 if (es->s_error_count)
2839 /* fsck newer than v1.41.13 is needed to clean this condition. */
2840 ext4_msg(sb, KERN_NOTICE, "error count since last fsck: %u",
2841 le32_to_cpu(es->s_error_count));
2842 if (es->s_first_error_time) {
2843 printk(KERN_NOTICE "EXT4-fs (%s): initial error at time %u: %.*s:%d",
2844 sb->s_id, le32_to_cpu(es->s_first_error_time),
2845 (int) sizeof(es->s_first_error_func),
2846 es->s_first_error_func,
2847 le32_to_cpu(es->s_first_error_line));
2848 if (es->s_first_error_ino)
2849 printk(": inode %u",
2850 le32_to_cpu(es->s_first_error_ino));
2851 if (es->s_first_error_block)
2852 printk(": block %llu", (unsigned long long)
2853 le64_to_cpu(es->s_first_error_block));
2856 if (es->s_last_error_time) {
2857 printk(KERN_NOTICE "EXT4-fs (%s): last error at time %u: %.*s:%d",
2858 sb->s_id, le32_to_cpu(es->s_last_error_time),
2859 (int) sizeof(es->s_last_error_func),
2860 es->s_last_error_func,
2861 le32_to_cpu(es->s_last_error_line));
2862 if (es->s_last_error_ino)
2863 printk(": inode %u",
2864 le32_to_cpu(es->s_last_error_ino));
2865 if (es->s_last_error_block)
2866 printk(": block %llu", (unsigned long long)
2867 le64_to_cpu(es->s_last_error_block));
2870 mod_timer(&sbi->s_err_report, jiffies + 24*60*60*HZ); /* Once a day */
2873 /* Find next suitable group and run ext4_init_inode_table */
2874 static int ext4_run_li_request(struct ext4_li_request *elr)
2876 struct ext4_group_desc *gdp = NULL;
2877 ext4_group_t group, ngroups;
2878 struct super_block *sb;
2879 unsigned long timeout = 0;
2883 ngroups = EXT4_SB(sb)->s_groups_count;
2886 for (group = elr->lr_next_group; group < ngroups; group++) {
2887 gdp = ext4_get_group_desc(sb, group, NULL);
2893 if (!(gdp->bg_flags & cpu_to_le16(EXT4_BG_INODE_ZEROED)))
2897 if (group >= ngroups)
2902 ret = ext4_init_inode_table(sb, group,
2903 elr->lr_timeout ? 0 : 1);
2904 if (elr->lr_timeout == 0) {
2905 timeout = (jiffies - timeout) *
2906 elr->lr_sbi->s_li_wait_mult;
2907 elr->lr_timeout = timeout;
2909 elr->lr_next_sched = jiffies + elr->lr_timeout;
2910 elr->lr_next_group = group + 1;
2918 * Remove lr_request from the list_request and free the
2919 * request structure. Should be called with li_list_mtx held
2921 static void ext4_remove_li_request(struct ext4_li_request *elr)
2923 struct ext4_sb_info *sbi;
2930 list_del(&elr->lr_request);
2931 sbi->s_li_request = NULL;
2935 static void ext4_unregister_li_request(struct super_block *sb)
2937 mutex_lock(&ext4_li_mtx);
2938 if (!ext4_li_info) {
2939 mutex_unlock(&ext4_li_mtx);
2943 mutex_lock(&ext4_li_info->li_list_mtx);
2944 ext4_remove_li_request(EXT4_SB(sb)->s_li_request);
2945 mutex_unlock(&ext4_li_info->li_list_mtx);
2946 mutex_unlock(&ext4_li_mtx);
2949 static struct task_struct *ext4_lazyinit_task;
2952 * This is the function where ext4lazyinit thread lives. It walks
2953 * through the request list searching for next scheduled filesystem.
2954 * When such a fs is found, run the lazy initialization request
2955 * (ext4_rn_li_request) and keep track of the time spend in this
2956 * function. Based on that time we compute next schedule time of
2957 * the request. When walking through the list is complete, compute
2958 * next waking time and put itself into sleep.
2960 static int ext4_lazyinit_thread(void *arg)
2962 struct ext4_lazy_init *eli = (struct ext4_lazy_init *)arg;
2963 struct list_head *pos, *n;
2964 struct ext4_li_request *elr;
2965 unsigned long next_wakeup, cur;
2967 BUG_ON(NULL == eli);
2971 next_wakeup = MAX_JIFFY_OFFSET;
2973 mutex_lock(&eli->li_list_mtx);
2974 if (list_empty(&eli->li_request_list)) {
2975 mutex_unlock(&eli->li_list_mtx);
2979 list_for_each_safe(pos, n, &eli->li_request_list) {
2980 elr = list_entry(pos, struct ext4_li_request,
2983 if (time_after_eq(jiffies, elr->lr_next_sched)) {
2984 if (ext4_run_li_request(elr) != 0) {
2985 /* error, remove the lazy_init job */
2986 ext4_remove_li_request(elr);
2991 if (time_before(elr->lr_next_sched, next_wakeup))
2992 next_wakeup = elr->lr_next_sched;
2994 mutex_unlock(&eli->li_list_mtx);
2999 if ((time_after_eq(cur, next_wakeup)) ||
3000 (MAX_JIFFY_OFFSET == next_wakeup)) {
3005 schedule_timeout_interruptible(next_wakeup - cur);
3007 if (kthread_should_stop()) {
3008 ext4_clear_request_list();
3015 * It looks like the request list is empty, but we need
3016 * to check it under the li_list_mtx lock, to prevent any
3017 * additions into it, and of course we should lock ext4_li_mtx
3018 * to atomically free the list and ext4_li_info, because at
3019 * this point another ext4 filesystem could be registering
3022 mutex_lock(&ext4_li_mtx);
3023 mutex_lock(&eli->li_list_mtx);
3024 if (!list_empty(&eli->li_request_list)) {
3025 mutex_unlock(&eli->li_list_mtx);
3026 mutex_unlock(&ext4_li_mtx);
3029 mutex_unlock(&eli->li_list_mtx);
3030 kfree(ext4_li_info);
3031 ext4_li_info = NULL;
3032 mutex_unlock(&ext4_li_mtx);
3037 static void ext4_clear_request_list(void)
3039 struct list_head *pos, *n;
3040 struct ext4_li_request *elr;
3042 mutex_lock(&ext4_li_info->li_list_mtx);
3043 list_for_each_safe(pos, n, &ext4_li_info->li_request_list) {
3044 elr = list_entry(pos, struct ext4_li_request,
3046 ext4_remove_li_request(elr);
3048 mutex_unlock(&ext4_li_info->li_list_mtx);
3051 static int ext4_run_lazyinit_thread(void)
3053 ext4_lazyinit_task = kthread_run(ext4_lazyinit_thread,
3054 ext4_li_info, "ext4lazyinit");
3055 if (IS_ERR(ext4_lazyinit_task)) {
3056 int err = PTR_ERR(ext4_lazyinit_task);
3057 ext4_clear_request_list();
3058 kfree(ext4_li_info);
3059 ext4_li_info = NULL;
3060 printk(KERN_CRIT "EXT4-fs: error %d creating inode table "
3061 "initialization thread\n",
3065 ext4_li_info->li_state |= EXT4_LAZYINIT_RUNNING;
3070 * Check whether it make sense to run itable init. thread or not.
3071 * If there is at least one uninitialized inode table, return
3072 * corresponding group number, else the loop goes through all
3073 * groups and return total number of groups.
3075 static ext4_group_t ext4_has_uninit_itable(struct super_block *sb)
3077 ext4_group_t group, ngroups = EXT4_SB(sb)->s_groups_count;
3078 struct ext4_group_desc *gdp = NULL;
3080 for (group = 0; group < ngroups; group++) {
3081 gdp = ext4_get_group_desc(sb, group, NULL);
3085 if (!(gdp->bg_flags & cpu_to_le16(EXT4_BG_INODE_ZEROED)))
3092 static int ext4_li_info_new(void)
3094 struct ext4_lazy_init *eli = NULL;
3096 eli = kzalloc(sizeof(*eli), GFP_KERNEL);
3100 INIT_LIST_HEAD(&eli->li_request_list);
3101 mutex_init(&eli->li_list_mtx);
3103 eli->li_state |= EXT4_LAZYINIT_QUIT;
3110 static struct ext4_li_request *ext4_li_request_new(struct super_block *sb,
3113 struct ext4_sb_info *sbi = EXT4_SB(sb);
3114 struct ext4_li_request *elr;
3116 elr = kzalloc(sizeof(*elr), GFP_KERNEL);
3122 elr->lr_next_group = start;
3125 * Randomize first schedule time of the request to
3126 * spread the inode table initialization requests
3129 elr->lr_next_sched = jiffies + (prandom_u32() %
3130 (EXT4_DEF_LI_MAX_START_DELAY * HZ));
3134 int ext4_register_li_request(struct super_block *sb,
3135 ext4_group_t first_not_zeroed)
3137 struct ext4_sb_info *sbi = EXT4_SB(sb);
3138 struct ext4_li_request *elr = NULL;
3139 ext4_group_t ngroups = EXT4_SB(sb)->s_groups_count;
3142 mutex_lock(&ext4_li_mtx);
3143 if (sbi->s_li_request != NULL) {
3145 * Reset timeout so it can be computed again, because
3146 * s_li_wait_mult might have changed.
3148 sbi->s_li_request->lr_timeout = 0;
3152 if (first_not_zeroed == ngroups ||
3153 (sb->s_flags & MS_RDONLY) ||
3154 !test_opt(sb, INIT_INODE_TABLE))
3157 elr = ext4_li_request_new(sb, first_not_zeroed);
3163 if (NULL == ext4_li_info) {
3164 ret = ext4_li_info_new();
3169 mutex_lock(&ext4_li_info->li_list_mtx);
3170 list_add(&elr->lr_request, &ext4_li_info->li_request_list);
3171 mutex_unlock(&ext4_li_info->li_list_mtx);
3173 sbi->s_li_request = elr;
3175 * set elr to NULL here since it has been inserted to
3176 * the request_list and the removal and free of it is
3177 * handled by ext4_clear_request_list from now on.
3181 if (!(ext4_li_info->li_state & EXT4_LAZYINIT_RUNNING)) {
3182 ret = ext4_run_lazyinit_thread();
3187 mutex_unlock(&ext4_li_mtx);
3194 * We do not need to lock anything since this is called on
3197 static void ext4_destroy_lazyinit_thread(void)
3200 * If thread exited earlier
3201 * there's nothing to be done.
3203 if (!ext4_li_info || !ext4_lazyinit_task)
3206 kthread_stop(ext4_lazyinit_task);
3209 static int set_journal_csum_feature_set(struct super_block *sb)
3212 int compat, incompat;
3213 struct ext4_sb_info *sbi = EXT4_SB(sb);
3215 if (ext4_has_metadata_csum(sb)) {
3216 /* journal checksum v3 */
3218 incompat = JBD2_FEATURE_INCOMPAT_CSUM_V3;
3220 /* journal checksum v1 */
3221 compat = JBD2_FEATURE_COMPAT_CHECKSUM;
3225 if (test_opt(sb, JOURNAL_ASYNC_COMMIT)) {
3226 ret = jbd2_journal_set_features(sbi->s_journal,
3228 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT |
3230 } else if (test_opt(sb, JOURNAL_CHECKSUM)) {
3231 ret = jbd2_journal_set_features(sbi->s_journal,
3234 jbd2_journal_clear_features(sbi->s_journal, 0, 0,
3235 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT);
3237 jbd2_journal_clear_features(sbi->s_journal,
3238 JBD2_FEATURE_COMPAT_CHECKSUM, 0,
3239 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT |
3240 JBD2_FEATURE_INCOMPAT_CSUM_V3 |
3241 JBD2_FEATURE_INCOMPAT_CSUM_V2);
3248 * Note: calculating the overhead so we can be compatible with
3249 * historical BSD practice is quite difficult in the face of
3250 * clusters/bigalloc. This is because multiple metadata blocks from
3251 * different block group can end up in the same allocation cluster.
3252 * Calculating the exact overhead in the face of clustered allocation
3253 * requires either O(all block bitmaps) in memory or O(number of block
3254 * groups**2) in time. We will still calculate the superblock for
3255 * older file systems --- and if we come across with a bigalloc file
3256 * system with zero in s_overhead_clusters the estimate will be close to
3257 * correct especially for very large cluster sizes --- but for newer
3258 * file systems, it's better to calculate this figure once at mkfs
3259 * time, and store it in the superblock. If the superblock value is
3260 * present (even for non-bigalloc file systems), we will use it.
3262 static int count_overhead(struct super_block *sb, ext4_group_t grp,
3265 struct ext4_sb_info *sbi = EXT4_SB(sb);
3266 struct ext4_group_desc *gdp;
3267 ext4_fsblk_t first_block, last_block, b;
3268 ext4_group_t i, ngroups = ext4_get_groups_count(sb);
3269 int s, j, count = 0;
3271 if (!EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_BIGALLOC))
3272 return (ext4_bg_has_super(sb, grp) + ext4_bg_num_gdb(sb, grp) +
3273 sbi->s_itb_per_group + 2);
3275 first_block = le32_to_cpu(sbi->s_es->s_first_data_block) +
3276 (grp * EXT4_BLOCKS_PER_GROUP(sb));
3277 last_block = first_block + EXT4_BLOCKS_PER_GROUP(sb) - 1;
3278 for (i = 0; i < ngroups; i++) {
3279 gdp = ext4_get_group_desc(sb, i, NULL);
3280 b = ext4_block_bitmap(sb, gdp);
3281 if (b >= first_block && b <= last_block) {
3282 ext4_set_bit(EXT4_B2C(sbi, b - first_block), buf);
3285 b = ext4_inode_bitmap(sb, gdp);
3286 if (b >= first_block && b <= last_block) {
3287 ext4_set_bit(EXT4_B2C(sbi, b - first_block), buf);
3290 b = ext4_inode_table(sb, gdp);
3291 if (b >= first_block && b + sbi->s_itb_per_group <= last_block)
3292 for (j = 0; j < sbi->s_itb_per_group; j++, b++) {
3293 int c = EXT4_B2C(sbi, b - first_block);
3294 ext4_set_bit(c, buf);
3300 if (ext4_bg_has_super(sb, grp)) {
3301 ext4_set_bit(s++, buf);
3304 for (j = ext4_bg_num_gdb(sb, grp); j > 0; j--) {
3305 ext4_set_bit(EXT4_B2C(sbi, s++), buf);
3311 return EXT4_CLUSTERS_PER_GROUP(sb) -
3312 ext4_count_free(buf, EXT4_CLUSTERS_PER_GROUP(sb) / 8);
3316 * Compute the overhead and stash it in sbi->s_overhead
3318 int ext4_calculate_overhead(struct super_block *sb)
3320 struct ext4_sb_info *sbi = EXT4_SB(sb);
3321 struct ext4_super_block *es = sbi->s_es;
3322 ext4_group_t i, ngroups = ext4_get_groups_count(sb);
3323 ext4_fsblk_t overhead = 0;
3324 char *buf = (char *) get_zeroed_page(GFP_KERNEL);
3330 * Compute the overhead (FS structures). This is constant
3331 * for a given filesystem unless the number of block groups
3332 * changes so we cache the previous value until it does.
3336 * All of the blocks before first_data_block are overhead
3338 overhead = EXT4_B2C(sbi, le32_to_cpu(es->s_first_data_block));
3341 * Add the overhead found in each block group
3343 for (i = 0; i < ngroups; i++) {
3346 blks = count_overhead(sb, i, buf);
3349 memset(buf, 0, PAGE_SIZE);
3352 /* Add the journal blocks as well */
3354 overhead += EXT4_NUM_B2C(sbi, sbi->s_journal->j_maxlen);
3356 sbi->s_overhead = overhead;
3358 free_page((unsigned long) buf);
3363 static ext4_fsblk_t ext4_calculate_resv_clusters(struct super_block *sb)
3365 ext4_fsblk_t resv_clusters;
3368 * There's no need to reserve anything when we aren't using extents.
3369 * The space estimates are exact, there are no unwritten extents,
3370 * hole punching doesn't need new metadata... This is needed especially
3371 * to keep ext2/3 backward compatibility.
3373 if (!EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_EXTENTS))
3376 * By default we reserve 2% or 4096 clusters, whichever is smaller.
3377 * This should cover the situations where we can not afford to run
3378 * out of space like for example punch hole, or converting
3379 * unwritten extents in delalloc path. In most cases such
3380 * allocation would require 1, or 2 blocks, higher numbers are
3383 resv_clusters = ext4_blocks_count(EXT4_SB(sb)->s_es) >>
3384 EXT4_SB(sb)->s_cluster_bits;
3386 do_div(resv_clusters, 50);
3387 resv_clusters = min_t(ext4_fsblk_t, resv_clusters, 4096);
3389 return resv_clusters;
3393 static int ext4_reserve_clusters(struct ext4_sb_info *sbi, ext4_fsblk_t count)
3395 ext4_fsblk_t clusters = ext4_blocks_count(sbi->s_es) >>
3396 sbi->s_cluster_bits;
3398 if (count >= clusters)
3401 atomic64_set(&sbi->s_resv_clusters, count);
3405 static int ext4_fill_super(struct super_block *sb, void *data, int silent)
3407 char *orig_data = kstrdup(data, GFP_KERNEL);
3408 struct buffer_head *bh;
3409 struct ext4_super_block *es = NULL;
3410 struct ext4_sb_info *sbi;
3412 ext4_fsblk_t sb_block = get_sb_block(&data);
3413 ext4_fsblk_t logical_sb_block;
3414 unsigned long offset = 0;
3415 unsigned long journal_devnum = 0;
3416 unsigned long def_mount_opts;
3421 int blocksize, clustersize;
3422 unsigned int db_count;
3424 int needs_recovery, has_huge_files, has_bigalloc;
3427 unsigned int journal_ioprio = DEFAULT_JOURNAL_IOPRIO;
3428 ext4_group_t first_not_zeroed;
3430 sbi = kzalloc(sizeof(*sbi), GFP_KERNEL);
3434 sbi->s_blockgroup_lock =
3435 kzalloc(sizeof(struct blockgroup_lock), GFP_KERNEL);
3436 if (!sbi->s_blockgroup_lock) {
3440 sb->s_fs_info = sbi;
3442 sbi->s_inode_readahead_blks = EXT4_DEF_INODE_READAHEAD_BLKS;
3443 sbi->s_sb_block = sb_block;
3444 if (sb->s_bdev->bd_part)
3445 sbi->s_sectors_written_start =
3446 part_stat_read(sb->s_bdev->bd_part, sectors[1]);
3448 /* Cleanup superblock name */
3449 for (cp = sb->s_id; (cp = strchr(cp, '/'));)
3452 /* -EINVAL is default */
3454 blocksize = sb_min_blocksize(sb, EXT4_MIN_BLOCK_SIZE);
3456 ext4_msg(sb, KERN_ERR, "unable to set blocksize");
3461 * The ext4 superblock will not be buffer aligned for other than 1kB
3462 * block sizes. We need to calculate the offset from buffer start.
3464 if (blocksize != EXT4_MIN_BLOCK_SIZE) {
3465 logical_sb_block = sb_block * EXT4_MIN_BLOCK_SIZE;
3466 offset = do_div(logical_sb_block, blocksize);
3468 logical_sb_block = sb_block;
3471 if (!(bh = sb_bread(sb, logical_sb_block))) {
3472 ext4_msg(sb, KERN_ERR, "unable to read superblock");
3476 * Note: s_es must be initialized as soon as possible because
3477 * some ext4 macro-instructions depend on its value
3479 es = (struct ext4_super_block *) (bh->b_data + offset);
3481 sb->s_magic = le16_to_cpu(es->s_magic);
3482 if (sb->s_magic != EXT4_SUPER_MAGIC)
3484 sbi->s_kbytes_written = le64_to_cpu(es->s_kbytes_written);
3486 /* Warn if metadata_csum and gdt_csum are both set. */
3487 if (EXT4_HAS_RO_COMPAT_FEATURE(sb,
3488 EXT4_FEATURE_RO_COMPAT_METADATA_CSUM) &&
3489 EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_GDT_CSUM))
3490 ext4_warning(sb, KERN_INFO "metadata_csum and uninit_bg are "
3491 "redundant flags; please run fsck.");
3493 /* Check for a known checksum algorithm */
3494 if (!ext4_verify_csum_type(sb, es)) {
3495 ext4_msg(sb, KERN_ERR, "VFS: Found ext4 filesystem with "
3496 "unknown checksum algorithm.");
3501 /* Load the checksum driver */
3502 if (EXT4_HAS_RO_COMPAT_FEATURE(sb,
3503 EXT4_FEATURE_RO_COMPAT_METADATA_CSUM)) {
3504 sbi->s_chksum_driver = crypto_alloc_shash("crc32c", 0, 0);
3505 if (IS_ERR(sbi->s_chksum_driver)) {
3506 ext4_msg(sb, KERN_ERR, "Cannot load crc32c driver.");
3507 ret = PTR_ERR(sbi->s_chksum_driver);
3508 sbi->s_chksum_driver = NULL;
3513 /* Check superblock checksum */
3514 if (!ext4_superblock_csum_verify(sb, es)) {
3515 ext4_msg(sb, KERN_ERR, "VFS: Found ext4 filesystem with "
3516 "invalid superblock checksum. Run e2fsck?");
3521 /* Precompute checksum seed for all metadata */
3522 if (ext4_has_metadata_csum(sb))
3523 sbi->s_csum_seed = ext4_chksum(sbi, ~0, es->s_uuid,
3524 sizeof(es->s_uuid));
3526 /* Set defaults before we parse the mount options */
3527 def_mount_opts = le32_to_cpu(es->s_default_mount_opts);
3528 set_opt(sb, INIT_INODE_TABLE);
3529 if (def_mount_opts & EXT4_DEFM_DEBUG)
3531 if (def_mount_opts & EXT4_DEFM_BSDGROUPS)
3533 if (def_mount_opts & EXT4_DEFM_UID16)
3534 set_opt(sb, NO_UID32);
3535 /* xattr user namespace & acls are now defaulted on */
3536 set_opt(sb, XATTR_USER);
3537 #ifdef CONFIG_EXT4_FS_POSIX_ACL
3538 set_opt(sb, POSIX_ACL);
3540 /* don't forget to enable journal_csum when metadata_csum is enabled. */
3541 if (ext4_has_metadata_csum(sb))
3542 set_opt(sb, JOURNAL_CHECKSUM);
3544 if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_DATA)
3545 set_opt(sb, JOURNAL_DATA);
3546 else if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_ORDERED)
3547 set_opt(sb, ORDERED_DATA);
3548 else if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_WBACK)
3549 set_opt(sb, WRITEBACK_DATA);
3551 if (le16_to_cpu(sbi->s_es->s_errors) == EXT4_ERRORS_PANIC)
3552 set_opt(sb, ERRORS_PANIC);
3553 else if (le16_to_cpu(sbi->s_es->s_errors) == EXT4_ERRORS_CONTINUE)
3554 set_opt(sb, ERRORS_CONT);
3556 set_opt(sb, ERRORS_RO);
3557 if (def_mount_opts & EXT4_DEFM_BLOCK_VALIDITY)
3558 set_opt(sb, BLOCK_VALIDITY);
3559 if (def_mount_opts & EXT4_DEFM_DISCARD)
3560 set_opt(sb, DISCARD);
3562 sbi->s_resuid = make_kuid(&init_user_ns, le16_to_cpu(es->s_def_resuid));
3563 sbi->s_resgid = make_kgid(&init_user_ns, le16_to_cpu(es->s_def_resgid));
3564 sbi->s_commit_interval = JBD2_DEFAULT_MAX_COMMIT_AGE * HZ;
3565 sbi->s_min_batch_time = EXT4_DEF_MIN_BATCH_TIME;
3566 sbi->s_max_batch_time = EXT4_DEF_MAX_BATCH_TIME;
3568 if ((def_mount_opts & EXT4_DEFM_NOBARRIER) == 0)
3569 set_opt(sb, BARRIER);
3572 * enable delayed allocation by default
3573 * Use -o nodelalloc to turn it off
3575 if (!IS_EXT3_SB(sb) && !IS_EXT2_SB(sb) &&
3576 ((def_mount_opts & EXT4_DEFM_NODELALLOC) == 0))
3577 set_opt(sb, DELALLOC);
3580 * set default s_li_wait_mult for lazyinit, for the case there is
3581 * no mount option specified.
3583 sbi->s_li_wait_mult = EXT4_DEF_LI_WAIT_MULT;
3585 if (!parse_options((char *) sbi->s_es->s_mount_opts, sb,
3586 &journal_devnum, &journal_ioprio, 0)) {
3587 ext4_msg(sb, KERN_WARNING,
3588 "failed to parse options in superblock: %s",
3589 sbi->s_es->s_mount_opts);
3591 sbi->s_def_mount_opt = sbi->s_mount_opt;
3592 if (!parse_options((char *) data, sb, &journal_devnum,
3593 &journal_ioprio, 0))
3596 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA) {
3597 printk_once(KERN_WARNING "EXT4-fs: Warning: mounting "
3598 "with data=journal disables delayed "
3599 "allocation and O_DIRECT support!\n");
3600 if (test_opt2(sb, EXPLICIT_DELALLOC)) {
3601 ext4_msg(sb, KERN_ERR, "can't mount with "
3602 "both data=journal and delalloc");
3605 if (test_opt(sb, DIOREAD_NOLOCK)) {
3606 ext4_msg(sb, KERN_ERR, "can't mount with "
3607 "both data=journal and dioread_nolock");
3610 if (test_opt(sb, DELALLOC))
3611 clear_opt(sb, DELALLOC);
3614 sb->s_flags = (sb->s_flags & ~MS_POSIXACL) |
3615 (test_opt(sb, POSIX_ACL) ? MS_POSIXACL : 0);
3617 if (le32_to_cpu(es->s_rev_level) == EXT4_GOOD_OLD_REV &&
3618 (EXT4_HAS_COMPAT_FEATURE(sb, ~0U) ||
3619 EXT4_HAS_RO_COMPAT_FEATURE(sb, ~0U) ||
3620 EXT4_HAS_INCOMPAT_FEATURE(sb, ~0U)))
3621 ext4_msg(sb, KERN_WARNING,
3622 "feature flags set on rev 0 fs, "
3623 "running e2fsck is recommended");
3625 if (es->s_creator_os == cpu_to_le32(EXT4_OS_HURD)) {
3626 set_opt2(sb, HURD_COMPAT);
3627 if (EXT4_HAS_INCOMPAT_FEATURE(sb,
3628 EXT4_FEATURE_INCOMPAT_64BIT)) {
3629 ext4_msg(sb, KERN_ERR,
3630 "The Hurd can't support 64-bit file systems");
3635 if (IS_EXT2_SB(sb)) {
3636 if (ext2_feature_set_ok(sb))
3637 ext4_msg(sb, KERN_INFO, "mounting ext2 file system "
3638 "using the ext4 subsystem");
3640 ext4_msg(sb, KERN_ERR, "couldn't mount as ext2 due "
3641 "to feature incompatibilities");
3646 if (IS_EXT3_SB(sb)) {
3647 if (ext3_feature_set_ok(sb))
3648 ext4_msg(sb, KERN_INFO, "mounting ext3 file system "
3649 "using the ext4 subsystem");
3651 ext4_msg(sb, KERN_ERR, "couldn't mount as ext3 due "
3652 "to feature incompatibilities");
3658 * Check feature flags regardless of the revision level, since we
3659 * previously didn't change the revision level when setting the flags,
3660 * so there is a chance incompat flags are set on a rev 0 filesystem.
3662 if (!ext4_feature_set_ok(sb, (sb->s_flags & MS_RDONLY)))
3665 blocksize = BLOCK_SIZE << le32_to_cpu(es->s_log_block_size);
3666 if (blocksize < EXT4_MIN_BLOCK_SIZE ||
3667 blocksize > EXT4_MAX_BLOCK_SIZE) {
3668 ext4_msg(sb, KERN_ERR,
3669 "Unsupported filesystem blocksize %d (%d log_block_size)",
3670 blocksize, le32_to_cpu(es->s_log_block_size));
3673 if (le32_to_cpu(es->s_log_block_size) >
3674 (EXT4_MAX_BLOCK_LOG_SIZE - EXT4_MIN_BLOCK_LOG_SIZE)) {
3675 ext4_msg(sb, KERN_ERR,
3676 "Invalid log block size: %u",
3677 le32_to_cpu(es->s_log_block_size));
3681 if (le16_to_cpu(sbi->s_es->s_reserved_gdt_blocks) > (blocksize / 4)) {
3682 ext4_msg(sb, KERN_ERR,
3683 "Number of reserved GDT blocks insanely large: %d",
3684 le16_to_cpu(sbi->s_es->s_reserved_gdt_blocks));
3688 if (sb->s_blocksize != blocksize) {
3689 /* Validate the filesystem blocksize */
3690 if (!sb_set_blocksize(sb, blocksize)) {
3691 ext4_msg(sb, KERN_ERR, "bad block size %d",
3697 logical_sb_block = sb_block * EXT4_MIN_BLOCK_SIZE;
3698 offset = do_div(logical_sb_block, blocksize);
3699 bh = sb_bread(sb, logical_sb_block);
3701 ext4_msg(sb, KERN_ERR,
3702 "Can't read superblock on 2nd try");
3705 es = (struct ext4_super_block *)(bh->b_data + offset);
3707 if (es->s_magic != cpu_to_le16(EXT4_SUPER_MAGIC)) {
3708 ext4_msg(sb, KERN_ERR,
3709 "Magic mismatch, very weird!");
3714 has_huge_files = EXT4_HAS_RO_COMPAT_FEATURE(sb,
3715 EXT4_FEATURE_RO_COMPAT_HUGE_FILE);
3716 sbi->s_bitmap_maxbytes = ext4_max_bitmap_size(sb->s_blocksize_bits,
3718 sb->s_maxbytes = ext4_max_size(sb->s_blocksize_bits, has_huge_files);
3720 if (le32_to_cpu(es->s_rev_level) == EXT4_GOOD_OLD_REV) {
3721 sbi->s_inode_size = EXT4_GOOD_OLD_INODE_SIZE;
3722 sbi->s_first_ino = EXT4_GOOD_OLD_FIRST_INO;
3724 sbi->s_inode_size = le16_to_cpu(es->s_inode_size);
3725 sbi->s_first_ino = le32_to_cpu(es->s_first_ino);
3726 if ((sbi->s_inode_size < EXT4_GOOD_OLD_INODE_SIZE) ||
3727 (!is_power_of_2(sbi->s_inode_size)) ||
3728 (sbi->s_inode_size > blocksize)) {
3729 ext4_msg(sb, KERN_ERR,
3730 "unsupported inode size: %d",
3734 if (sbi->s_inode_size > EXT4_GOOD_OLD_INODE_SIZE)
3735 sb->s_time_gran = 1 << (EXT4_EPOCH_BITS - 2);
3738 sbi->s_desc_size = le16_to_cpu(es->s_desc_size);
3739 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_64BIT)) {
3740 if (sbi->s_desc_size < EXT4_MIN_DESC_SIZE_64BIT ||
3741 sbi->s_desc_size > EXT4_MAX_DESC_SIZE ||
3742 !is_power_of_2(sbi->s_desc_size)) {
3743 ext4_msg(sb, KERN_ERR,
3744 "unsupported descriptor size %lu",
3749 sbi->s_desc_size = EXT4_MIN_DESC_SIZE;
3751 sbi->s_blocks_per_group = le32_to_cpu(es->s_blocks_per_group);
3752 sbi->s_inodes_per_group = le32_to_cpu(es->s_inodes_per_group);
3753 if (EXT4_INODE_SIZE(sb) == 0 || EXT4_INODES_PER_GROUP(sb) == 0)
3756 sbi->s_inodes_per_block = blocksize / EXT4_INODE_SIZE(sb);
3757 if (sbi->s_inodes_per_block == 0)
3759 sbi->s_itb_per_group = sbi->s_inodes_per_group /
3760 sbi->s_inodes_per_block;
3761 sbi->s_desc_per_block = blocksize / EXT4_DESC_SIZE(sb);
3763 sbi->s_mount_state = le16_to_cpu(es->s_state);
3764 sbi->s_addr_per_block_bits = ilog2(EXT4_ADDR_PER_BLOCK(sb));
3765 sbi->s_desc_per_block_bits = ilog2(EXT4_DESC_PER_BLOCK(sb));
3767 for (i = 0; i < 4; i++)
3768 sbi->s_hash_seed[i] = le32_to_cpu(es->s_hash_seed[i]);
3769 sbi->s_def_hash_version = es->s_def_hash_version;
3770 if (EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_DIR_INDEX)) {
3771 i = le32_to_cpu(es->s_flags);
3772 if (i & EXT2_FLAGS_UNSIGNED_HASH)
3773 sbi->s_hash_unsigned = 3;
3774 else if ((i & EXT2_FLAGS_SIGNED_HASH) == 0) {
3775 #ifdef __CHAR_UNSIGNED__
3776 if (!(sb->s_flags & MS_RDONLY))
3778 cpu_to_le32(EXT2_FLAGS_UNSIGNED_HASH);
3779 sbi->s_hash_unsigned = 3;
3781 if (!(sb->s_flags & MS_RDONLY))
3783 cpu_to_le32(EXT2_FLAGS_SIGNED_HASH);
3788 /* Handle clustersize */
3789 clustersize = BLOCK_SIZE << le32_to_cpu(es->s_log_cluster_size);
3790 has_bigalloc = EXT4_HAS_RO_COMPAT_FEATURE(sb,
3791 EXT4_FEATURE_RO_COMPAT_BIGALLOC);
3793 if (clustersize < blocksize) {
3794 ext4_msg(sb, KERN_ERR,
3795 "cluster size (%d) smaller than "
3796 "block size (%d)", clustersize, blocksize);
3799 if (le32_to_cpu(es->s_log_cluster_size) >
3800 (EXT4_MAX_CLUSTER_LOG_SIZE - EXT4_MIN_BLOCK_LOG_SIZE)) {
3801 ext4_msg(sb, KERN_ERR,
3802 "Invalid log cluster size: %u",
3803 le32_to_cpu(es->s_log_cluster_size));
3806 sbi->s_cluster_bits = le32_to_cpu(es->s_log_cluster_size) -
3807 le32_to_cpu(es->s_log_block_size);
3808 sbi->s_clusters_per_group =
3809 le32_to_cpu(es->s_clusters_per_group);
3810 if (sbi->s_clusters_per_group > blocksize * 8) {
3811 ext4_msg(sb, KERN_ERR,
3812 "#clusters per group too big: %lu",
3813 sbi->s_clusters_per_group);
3816 if (sbi->s_blocks_per_group !=
3817 (sbi->s_clusters_per_group * (clustersize / blocksize))) {
3818 ext4_msg(sb, KERN_ERR, "blocks per group (%lu) and "
3819 "clusters per group (%lu) inconsistent",
3820 sbi->s_blocks_per_group,
3821 sbi->s_clusters_per_group);
3825 if (clustersize != blocksize) {
3826 ext4_warning(sb, "fragment/cluster size (%d) != "
3827 "block size (%d)", clustersize,
3829 clustersize = blocksize;
3831 if (sbi->s_blocks_per_group > blocksize * 8) {
3832 ext4_msg(sb, KERN_ERR,
3833 "#blocks per group too big: %lu",
3834 sbi->s_blocks_per_group);
3837 sbi->s_clusters_per_group = sbi->s_blocks_per_group;
3838 sbi->s_cluster_bits = 0;
3840 sbi->s_cluster_ratio = clustersize / blocksize;
3842 if (sbi->s_inodes_per_group > blocksize * 8) {
3843 ext4_msg(sb, KERN_ERR,
3844 "#inodes per group too big: %lu",
3845 sbi->s_inodes_per_group);
3849 /* Do we have standard group size of clustersize * 8 blocks ? */
3850 if (sbi->s_blocks_per_group == clustersize << 3)
3851 set_opt2(sb, STD_GROUP_SIZE);
3854 * Test whether we have more sectors than will fit in sector_t,
3855 * and whether the max offset is addressable by the page cache.
3857 err = generic_check_addressable(sb->s_blocksize_bits,
3858 ext4_blocks_count(es));
3860 ext4_msg(sb, KERN_ERR, "filesystem"
3861 " too large to mount safely on this system");
3862 if (sizeof(sector_t) < 8)
3863 ext4_msg(sb, KERN_WARNING, "CONFIG_LBDAF not enabled");
3867 if (EXT4_BLOCKS_PER_GROUP(sb) == 0)
3870 /* check blocks count against device size */
3871 blocks_count = sb->s_bdev->bd_inode->i_size >> sb->s_blocksize_bits;
3872 if (blocks_count && ext4_blocks_count(es) > blocks_count) {
3873 ext4_msg(sb, KERN_WARNING, "bad geometry: block count %llu "
3874 "exceeds size of device (%llu blocks)",
3875 ext4_blocks_count(es), blocks_count);
3880 * It makes no sense for the first data block to be beyond the end
3881 * of the filesystem.
3883 if (le32_to_cpu(es->s_first_data_block) >= ext4_blocks_count(es)) {
3884 ext4_msg(sb, KERN_WARNING, "bad geometry: first data "
3885 "block %u is beyond end of filesystem (%llu)",
3886 le32_to_cpu(es->s_first_data_block),
3887 ext4_blocks_count(es));
3890 blocks_count = (ext4_blocks_count(es) -
3891 le32_to_cpu(es->s_first_data_block) +
3892 EXT4_BLOCKS_PER_GROUP(sb) - 1);
3893 do_div(blocks_count, EXT4_BLOCKS_PER_GROUP(sb));
3894 if (blocks_count > ((uint64_t)1<<32) - EXT4_DESC_PER_BLOCK(sb)) {
3895 ext4_msg(sb, KERN_WARNING, "groups count too large: %u "
3896 "(block count %llu, first data block %u, "
3897 "blocks per group %lu)", sbi->s_groups_count,
3898 ext4_blocks_count(es),
3899 le32_to_cpu(es->s_first_data_block),
3900 EXT4_BLOCKS_PER_GROUP(sb));
3903 sbi->s_groups_count = blocks_count;
3904 sbi->s_blockfile_groups = min_t(ext4_group_t, sbi->s_groups_count,
3905 (EXT4_MAX_BLOCK_FILE_PHYS / EXT4_BLOCKS_PER_GROUP(sb)));
3906 db_count = (sbi->s_groups_count + EXT4_DESC_PER_BLOCK(sb) - 1) /
3907 EXT4_DESC_PER_BLOCK(sb);
3908 if (EXT4_HAS_INCOMPAT_FEATURE(sb,EXT4_FEATURE_INCOMPAT_META_BG)) {
3909 if (le32_to_cpu(es->s_first_meta_bg) >= db_count) {
3910 ext4_msg(sb, KERN_WARNING,
3911 "first meta block group too large: %u "
3912 "(group descriptor block count %u)",
3913 le32_to_cpu(es->s_first_meta_bg), db_count);
3917 sbi->s_group_desc = ext4_kvmalloc(db_count *
3918 sizeof(struct buffer_head *),
3920 if (sbi->s_group_desc == NULL) {
3921 ext4_msg(sb, KERN_ERR, "not enough memory");
3927 sbi->s_proc = proc_mkdir(sb->s_id, ext4_proc_root);
3930 proc_create_data("options", S_IRUGO, sbi->s_proc,
3931 &ext4_seq_options_fops, sb);
3933 bgl_lock_init(sbi->s_blockgroup_lock);
3935 for (i = 0; i < db_count; i++) {
3936 block = descriptor_loc(sb, logical_sb_block, i);
3937 sbi->s_group_desc[i] = sb_bread(sb, block);
3938 if (!sbi->s_group_desc[i]) {
3939 ext4_msg(sb, KERN_ERR,
3940 "can't read group descriptor %d", i);
3945 if (!ext4_check_descriptors(sb, logical_sb_block, &first_not_zeroed)) {
3946 ext4_msg(sb, KERN_ERR, "group descriptors corrupted!");
3949 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_FLEX_BG))
3950 if (!ext4_fill_flex_info(sb)) {
3951 ext4_msg(sb, KERN_ERR,
3952 "unable to initialize "
3953 "flex_bg meta info!");
3957 sbi->s_gdb_count = db_count;
3958 get_random_bytes(&sbi->s_next_generation, sizeof(u32));
3959 spin_lock_init(&sbi->s_next_gen_lock);
3961 init_timer(&sbi->s_err_report);
3962 sbi->s_err_report.function = print_daily_error_info;
3963 sbi->s_err_report.data = (unsigned long) sb;
3965 /* Register extent status tree shrinker */
3966 ext4_es_register_shrinker(sbi);
3968 err = percpu_counter_init(&sbi->s_freeclusters_counter,
3969 ext4_count_free_clusters(sb));
3971 err = percpu_counter_init(&sbi->s_freeinodes_counter,
3972 ext4_count_free_inodes(sb));
3975 err = percpu_counter_init(&sbi->s_dirs_counter,
3976 ext4_count_dirs(sb));
3979 err = percpu_counter_init(&sbi->s_dirtyclusters_counter, 0);
3982 err = percpu_counter_init(&sbi->s_extent_cache_cnt, 0);
3985 ext4_msg(sb, KERN_ERR, "insufficient memory");
3989 sbi->s_stripe = ext4_get_stripe_size(sbi);
3990 sbi->s_extent_max_zeroout_kb = 32;
3993 * set up enough so that it can read an inode
3995 if (!test_opt(sb, NOLOAD) &&
3996 EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL))
3997 sb->s_op = &ext4_sops;
3999 sb->s_op = &ext4_nojournal_sops;
4000 sb->s_export_op = &ext4_export_ops;
4001 sb->s_xattr = ext4_xattr_handlers;
4003 sb->dq_op = &ext4_quota_operations;
4004 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_QUOTA))
4005 sb->s_qcop = &ext4_qctl_sysfile_operations;
4007 sb->s_qcop = &ext4_qctl_operations;
4009 memcpy(sb->s_uuid, es->s_uuid, sizeof(es->s_uuid));
4011 INIT_LIST_HEAD(&sbi->s_orphan); /* unlinked but open files */
4012 mutex_init(&sbi->s_orphan_lock);
4016 needs_recovery = (es->s_last_orphan != 0 ||
4017 EXT4_HAS_INCOMPAT_FEATURE(sb,
4018 EXT4_FEATURE_INCOMPAT_RECOVER));
4020 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_MMP) &&
4021 !(sb->s_flags & MS_RDONLY))
4022 if (ext4_multi_mount_protect(sb, le64_to_cpu(es->s_mmp_block)))
4026 * The first inode we look at is the journal inode. Don't try
4027 * root first: it may be modified in the journal!
4029 if (!test_opt(sb, NOLOAD) &&
4030 EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL)) {
4031 if (ext4_load_journal(sb, es, journal_devnum))
4033 } else if (test_opt(sb, NOLOAD) && !(sb->s_flags & MS_RDONLY) &&
4034 EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER)) {
4035 ext4_msg(sb, KERN_ERR, "required journal recovery "
4036 "suppressed and not mounted read-only");
4037 goto failed_mount_wq;
4039 clear_opt(sb, DATA_FLAGS);
4040 sbi->s_journal = NULL;
4045 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_64BIT) &&
4046 !jbd2_journal_set_features(EXT4_SB(sb)->s_journal, 0, 0,
4047 JBD2_FEATURE_INCOMPAT_64BIT)) {
4048 ext4_msg(sb, KERN_ERR, "Failed to set 64-bit journal feature");
4049 goto failed_mount_wq;
4052 if (!set_journal_csum_feature_set(sb)) {
4053 ext4_msg(sb, KERN_ERR, "Failed to set journal checksum "
4055 goto failed_mount_wq;
4058 /* We have now updated the journal if required, so we can
4059 * validate the data journaling mode. */
4060 switch (test_opt(sb, DATA_FLAGS)) {
4062 /* No mode set, assume a default based on the journal
4063 * capabilities: ORDERED_DATA if the journal can
4064 * cope, else JOURNAL_DATA
4066 if (jbd2_journal_check_available_features
4067 (sbi->s_journal, 0, 0, JBD2_FEATURE_INCOMPAT_REVOKE))
4068 set_opt(sb, ORDERED_DATA);
4070 set_opt(sb, JOURNAL_DATA);
4073 case EXT4_MOUNT_ORDERED_DATA:
4074 case EXT4_MOUNT_WRITEBACK_DATA:
4075 if (!jbd2_journal_check_available_features
4076 (sbi->s_journal, 0, 0, JBD2_FEATURE_INCOMPAT_REVOKE)) {
4077 ext4_msg(sb, KERN_ERR, "Journal does not support "
4078 "requested data journaling mode");
4079 goto failed_mount_wq;
4084 set_task_ioprio(sbi->s_journal->j_task, journal_ioprio);
4086 sbi->s_journal->j_commit_callback = ext4_journal_commit_callback;
4089 * The journal may have updated the bg summary counts, so we
4090 * need to update the global counters.
4092 percpu_counter_set(&sbi->s_freeclusters_counter,
4093 ext4_count_free_clusters(sb));
4094 percpu_counter_set(&sbi->s_freeinodes_counter,
4095 ext4_count_free_inodes(sb));
4096 percpu_counter_set(&sbi->s_dirs_counter,
4097 ext4_count_dirs(sb));
4098 percpu_counter_set(&sbi->s_dirtyclusters_counter, 0);
4101 if (ext4_mballoc_ready) {
4102 sbi->s_mb_cache = ext4_xattr_create_cache(sb->s_id);
4103 if (!sbi->s_mb_cache) {
4104 ext4_msg(sb, KERN_ERR, "Failed to create an mb_cache");
4105 goto failed_mount_wq;
4110 * Get the # of file system overhead blocks from the
4111 * superblock if present.
4113 if (es->s_overhead_clusters)
4114 sbi->s_overhead = le32_to_cpu(es->s_overhead_clusters);
4116 err = ext4_calculate_overhead(sb);
4118 goto failed_mount_wq;
4122 * The maximum number of concurrent works can be high and
4123 * concurrency isn't really necessary. Limit it to 1.
4125 EXT4_SB(sb)->rsv_conversion_wq =
4126 alloc_workqueue("ext4-rsv-conversion", WQ_MEM_RECLAIM | WQ_UNBOUND, 1);
4127 if (!EXT4_SB(sb)->rsv_conversion_wq) {
4128 printk(KERN_ERR "EXT4-fs: failed to create workqueue\n");
4134 * The jbd2_journal_load will have done any necessary log recovery,
4135 * so we can safely mount the rest of the filesystem now.
4138 root = ext4_iget(sb, EXT4_ROOT_INO);
4140 ext4_msg(sb, KERN_ERR, "get root inode failed");
4141 ret = PTR_ERR(root);
4145 if (!S_ISDIR(root->i_mode) || !root->i_blocks || !root->i_size) {
4146 ext4_msg(sb, KERN_ERR, "corrupt root inode, run e2fsck");
4150 sb->s_root = d_make_root(root);
4152 ext4_msg(sb, KERN_ERR, "get root dentry failed");
4157 if (ext4_setup_super(sb, es, sb->s_flags & MS_RDONLY))
4158 sb->s_flags |= MS_RDONLY;
4160 /* determine the minimum size of new large inodes, if present */
4161 if (sbi->s_inode_size > EXT4_GOOD_OLD_INODE_SIZE) {
4162 sbi->s_want_extra_isize = sizeof(struct ext4_inode) -
4163 EXT4_GOOD_OLD_INODE_SIZE;
4164 if (EXT4_HAS_RO_COMPAT_FEATURE(sb,
4165 EXT4_FEATURE_RO_COMPAT_EXTRA_ISIZE)) {
4166 if (sbi->s_want_extra_isize <
4167 le16_to_cpu(es->s_want_extra_isize))
4168 sbi->s_want_extra_isize =
4169 le16_to_cpu(es->s_want_extra_isize);
4170 if (sbi->s_want_extra_isize <
4171 le16_to_cpu(es->s_min_extra_isize))
4172 sbi->s_want_extra_isize =
4173 le16_to_cpu(es->s_min_extra_isize);
4176 /* Check if enough inode space is available */
4177 if (EXT4_GOOD_OLD_INODE_SIZE + sbi->s_want_extra_isize >
4178 sbi->s_inode_size) {
4179 sbi->s_want_extra_isize = sizeof(struct ext4_inode) -
4180 EXT4_GOOD_OLD_INODE_SIZE;
4181 ext4_msg(sb, KERN_INFO, "required extra inode space not"
4185 err = ext4_reserve_clusters(sbi, ext4_calculate_resv_clusters(sb));
4187 ext4_msg(sb, KERN_ERR, "failed to reserve %llu clusters for "
4188 "reserved pool", ext4_calculate_resv_clusters(sb));
4189 goto failed_mount4a;
4192 err = ext4_setup_system_zone(sb);
4194 ext4_msg(sb, KERN_ERR, "failed to initialize system "
4196 goto failed_mount4a;
4200 err = ext4_mb_init(sb);
4202 ext4_msg(sb, KERN_ERR, "failed to initialize mballoc (%d)",
4207 err = ext4_register_li_request(sb, first_not_zeroed);
4211 sbi->s_kobj.kset = ext4_kset;
4212 init_completion(&sbi->s_kobj_unregister);
4213 err = kobject_init_and_add(&sbi->s_kobj, &ext4_ktype, NULL,
4219 /* Enable quota usage during mount. */
4220 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_QUOTA) &&
4221 !(sb->s_flags & MS_RDONLY)) {
4222 err = ext4_enable_quotas(sb);
4226 #endif /* CONFIG_QUOTA */
4228 EXT4_SB(sb)->s_mount_state |= EXT4_ORPHAN_FS;
4229 ext4_orphan_cleanup(sb, es);
4230 EXT4_SB(sb)->s_mount_state &= ~EXT4_ORPHAN_FS;
4231 if (needs_recovery) {
4232 ext4_msg(sb, KERN_INFO, "recovery complete");
4233 ext4_mark_recovery_complete(sb, es);
4235 if (EXT4_SB(sb)->s_journal) {
4236 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA)
4237 descr = " journalled data mode";
4238 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA)
4239 descr = " ordered data mode";
4241 descr = " writeback data mode";
4243 descr = "out journal";
4245 if (test_opt(sb, DISCARD)) {
4246 struct request_queue *q = bdev_get_queue(sb->s_bdev);
4247 if (!blk_queue_discard(q))
4248 ext4_msg(sb, KERN_WARNING,
4249 "mounting with \"discard\" option, but "
4250 "the device does not support discard");
4253 ext4_msg(sb, KERN_INFO, "mounted filesystem with%s. "
4254 "Opts: %s%s%s", descr, sbi->s_es->s_mount_opts,
4255 *sbi->s_es->s_mount_opts ? "; " : "", orig_data);
4257 if (es->s_error_count)
4258 mod_timer(&sbi->s_err_report, jiffies + 300*HZ); /* 5 minutes */
4260 /* Enable message ratelimiting. Default is 10 messages per 5 secs. */
4261 ratelimit_state_init(&sbi->s_err_ratelimit_state, 5 * HZ, 10);
4262 ratelimit_state_init(&sbi->s_warning_ratelimit_state, 5 * HZ, 10);
4263 ratelimit_state_init(&sbi->s_msg_ratelimit_state, 5 * HZ, 10);
4270 ext4_msg(sb, KERN_ERR, "VFS: Can't find ext4 filesystem");
4275 kobject_del(&sbi->s_kobj);
4278 ext4_unregister_li_request(sb);
4280 ext4_mb_release(sb);
4282 ext4_ext_release(sb);
4283 ext4_release_system_zone(sb);
4288 ext4_msg(sb, KERN_ERR, "mount failed");
4289 if (EXT4_SB(sb)->rsv_conversion_wq)
4290 destroy_workqueue(EXT4_SB(sb)->rsv_conversion_wq);
4292 if (sbi->s_journal) {
4293 jbd2_journal_destroy(sbi->s_journal);
4294 sbi->s_journal = NULL;
4297 ext4_es_unregister_shrinker(sbi);
4298 del_timer_sync(&sbi->s_err_report);
4299 if (sbi->s_flex_groups)
4300 ext4_kvfree(sbi->s_flex_groups);
4301 percpu_counter_destroy(&sbi->s_freeclusters_counter);
4302 percpu_counter_destroy(&sbi->s_freeinodes_counter);
4303 percpu_counter_destroy(&sbi->s_dirs_counter);
4304 percpu_counter_destroy(&sbi->s_dirtyclusters_counter);
4305 percpu_counter_destroy(&sbi->s_extent_cache_cnt);
4307 kthread_stop(sbi->s_mmp_tsk);
4309 for (i = 0; i < db_count; i++)
4310 brelse(sbi->s_group_desc[i]);
4311 ext4_kvfree(sbi->s_group_desc);
4313 if (sbi->s_chksum_driver)
4314 crypto_free_shash(sbi->s_chksum_driver);
4316 remove_proc_entry("options", sbi->s_proc);
4317 remove_proc_entry(sb->s_id, ext4_proc_root);
4320 for (i = 0; i < MAXQUOTAS; i++)
4321 kfree(sbi->s_qf_names[i]);
4323 ext4_blkdev_remove(sbi);
4326 sb->s_fs_info = NULL;
4327 kfree(sbi->s_blockgroup_lock);
4331 return err ? err : ret;
4335 * Setup any per-fs journal parameters now. We'll do this both on
4336 * initial mount, once the journal has been initialised but before we've
4337 * done any recovery; and again on any subsequent remount.
4339 static void ext4_init_journal_params(struct super_block *sb, journal_t *journal)
4341 struct ext4_sb_info *sbi = EXT4_SB(sb);
4343 journal->j_commit_interval = sbi->s_commit_interval;
4344 journal->j_min_batch_time = sbi->s_min_batch_time;
4345 journal->j_max_batch_time = sbi->s_max_batch_time;
4347 write_lock(&journal->j_state_lock);
4348 if (test_opt(sb, BARRIER))
4349 journal->j_flags |= JBD2_BARRIER;
4351 journal->j_flags &= ~JBD2_BARRIER;
4352 if (test_opt(sb, DATA_ERR_ABORT))
4353 journal->j_flags |= JBD2_ABORT_ON_SYNCDATA_ERR;
4355 journal->j_flags &= ~JBD2_ABORT_ON_SYNCDATA_ERR;
4356 write_unlock(&journal->j_state_lock);
4359 static journal_t *ext4_get_journal(struct super_block *sb,
4360 unsigned int journal_inum)
4362 struct inode *journal_inode;
4365 BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL));
4367 /* First, test for the existence of a valid inode on disk. Bad
4368 * things happen if we iget() an unused inode, as the subsequent
4369 * iput() will try to delete it. */
4371 journal_inode = ext4_iget(sb, journal_inum);
4372 if (IS_ERR(journal_inode)) {
4373 ext4_msg(sb, KERN_ERR, "no journal found");
4376 if (!journal_inode->i_nlink) {
4377 make_bad_inode(journal_inode);
4378 iput(journal_inode);
4379 ext4_msg(sb, KERN_ERR, "journal inode is deleted");
4383 jbd_debug(2, "Journal inode found at %p: %lld bytes\n",
4384 journal_inode, journal_inode->i_size);
4385 if (!S_ISREG(journal_inode->i_mode)) {
4386 ext4_msg(sb, KERN_ERR, "invalid journal inode");
4387 iput(journal_inode);
4391 journal = jbd2_journal_init_inode(journal_inode);
4393 ext4_msg(sb, KERN_ERR, "Could not load journal inode");
4394 iput(journal_inode);
4397 journal->j_private = sb;
4398 ext4_init_journal_params(sb, journal);
4402 static journal_t *ext4_get_dev_journal(struct super_block *sb,
4405 struct buffer_head *bh;
4409 int hblock, blocksize;
4410 ext4_fsblk_t sb_block;
4411 unsigned long offset;
4412 struct ext4_super_block *es;
4413 struct block_device *bdev;
4415 BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL));
4417 bdev = ext4_blkdev_get(j_dev, sb);
4421 blocksize = sb->s_blocksize;
4422 hblock = bdev_logical_block_size(bdev);
4423 if (blocksize < hblock) {
4424 ext4_msg(sb, KERN_ERR,
4425 "blocksize too small for journal device");
4429 sb_block = EXT4_MIN_BLOCK_SIZE / blocksize;
4430 offset = EXT4_MIN_BLOCK_SIZE % blocksize;
4431 set_blocksize(bdev, blocksize);
4432 if (!(bh = __bread(bdev, sb_block, blocksize))) {
4433 ext4_msg(sb, KERN_ERR, "couldn't read superblock of "
4434 "external journal");
4438 es = (struct ext4_super_block *) (bh->b_data + offset);
4439 if ((le16_to_cpu(es->s_magic) != EXT4_SUPER_MAGIC) ||
4440 !(le32_to_cpu(es->s_feature_incompat) &
4441 EXT4_FEATURE_INCOMPAT_JOURNAL_DEV)) {
4442 ext4_msg(sb, KERN_ERR, "external journal has "
4448 if (memcmp(EXT4_SB(sb)->s_es->s_journal_uuid, es->s_uuid, 16)) {
4449 ext4_msg(sb, KERN_ERR, "journal UUID does not match");
4454 len = ext4_blocks_count(es);
4455 start = sb_block + 1;
4456 brelse(bh); /* we're done with the superblock */
4458 journal = jbd2_journal_init_dev(bdev, sb->s_bdev,
4459 start, len, blocksize);
4461 ext4_msg(sb, KERN_ERR, "failed to create device journal");
4464 journal->j_private = sb;
4465 ll_rw_block(READ | REQ_META | REQ_PRIO, 1, &journal->j_sb_buffer);
4466 wait_on_buffer(journal->j_sb_buffer);
4467 if (!buffer_uptodate(journal->j_sb_buffer)) {
4468 ext4_msg(sb, KERN_ERR, "I/O error on journal device");
4471 if (be32_to_cpu(journal->j_superblock->s_nr_users) != 1) {
4472 ext4_msg(sb, KERN_ERR, "External journal has more than one "
4473 "user (unsupported) - %d",
4474 be32_to_cpu(journal->j_superblock->s_nr_users));
4477 EXT4_SB(sb)->journal_bdev = bdev;
4478 ext4_init_journal_params(sb, journal);
4482 jbd2_journal_destroy(journal);
4484 ext4_blkdev_put(bdev);
4488 static int ext4_load_journal(struct super_block *sb,
4489 struct ext4_super_block *es,
4490 unsigned long journal_devnum)
4493 unsigned int journal_inum = le32_to_cpu(es->s_journal_inum);
4496 int really_read_only;
4498 BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL));
4500 if (journal_devnum &&
4501 journal_devnum != le32_to_cpu(es->s_journal_dev)) {
4502 ext4_msg(sb, KERN_INFO, "external journal device major/minor "
4503 "numbers have changed");
4504 journal_dev = new_decode_dev(journal_devnum);
4506 journal_dev = new_decode_dev(le32_to_cpu(es->s_journal_dev));
4508 really_read_only = bdev_read_only(sb->s_bdev);
4511 * Are we loading a blank journal or performing recovery after a
4512 * crash? For recovery, we need to check in advance whether we
4513 * can get read-write access to the device.
4515 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER)) {
4516 if (sb->s_flags & MS_RDONLY) {
4517 ext4_msg(sb, KERN_INFO, "INFO: recovery "
4518 "required on readonly filesystem");
4519 if (really_read_only) {
4520 ext4_msg(sb, KERN_ERR, "write access "
4521 "unavailable, cannot proceed");
4524 ext4_msg(sb, KERN_INFO, "write access will "
4525 "be enabled during recovery");
4529 if (journal_inum && journal_dev) {
4530 ext4_msg(sb, KERN_ERR, "filesystem has both journal "
4531 "and inode journals!");
4536 if (!(journal = ext4_get_journal(sb, journal_inum)))
4539 if (!(journal = ext4_get_dev_journal(sb, journal_dev)))
4543 if (!(journal->j_flags & JBD2_BARRIER))
4544 ext4_msg(sb, KERN_INFO, "barriers disabled");
4546 if (!EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER))
4547 err = jbd2_journal_wipe(journal, !really_read_only);
4549 char *save = kmalloc(EXT4_S_ERR_LEN, GFP_KERNEL);
4551 memcpy(save, ((char *) es) +
4552 EXT4_S_ERR_START, EXT4_S_ERR_LEN);
4553 err = jbd2_journal_load(journal);
4555 memcpy(((char *) es) + EXT4_S_ERR_START,
4556 save, EXT4_S_ERR_LEN);
4561 ext4_msg(sb, KERN_ERR, "error loading journal");
4562 jbd2_journal_destroy(journal);
4566 EXT4_SB(sb)->s_journal = journal;
4567 ext4_clear_journal_err(sb, es);
4569 if (!really_read_only && journal_devnum &&
4570 journal_devnum != le32_to_cpu(es->s_journal_dev)) {
4571 es->s_journal_dev = cpu_to_le32(journal_devnum);
4573 /* Make sure we flush the recovery flag to disk. */
4574 ext4_commit_super(sb, 1);
4580 static int ext4_commit_super(struct super_block *sb, int sync)
4582 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
4583 struct buffer_head *sbh = EXT4_SB(sb)->s_sbh;
4586 if (!sbh || block_device_ejected(sb))
4588 if (buffer_write_io_error(sbh)) {
4590 * Oh, dear. A previous attempt to write the
4591 * superblock failed. This could happen because the
4592 * USB device was yanked out. Or it could happen to
4593 * be a transient write error and maybe the block will
4594 * be remapped. Nothing we can do but to retry the
4595 * write and hope for the best.
4597 ext4_msg(sb, KERN_ERR, "previous I/O error to "
4598 "superblock detected");
4599 clear_buffer_write_io_error(sbh);
4600 set_buffer_uptodate(sbh);
4603 * If the file system is mounted read-only, don't update the
4604 * superblock write time. This avoids updating the superblock
4605 * write time when we are mounting the root file system
4606 * read/only but we need to replay the journal; at that point,
4607 * for people who are east of GMT and who make their clock
4608 * tick in localtime for Windows bug-for-bug compatibility,
4609 * the clock is set in the future, and this will cause e2fsck
4610 * to complain and force a full file system check.
4612 if (!(sb->s_flags & MS_RDONLY))
4613 es->s_wtime = cpu_to_le32(get_seconds());
4614 if (sb->s_bdev->bd_part)
4615 es->s_kbytes_written =
4616 cpu_to_le64(EXT4_SB(sb)->s_kbytes_written +
4617 ((part_stat_read(sb->s_bdev->bd_part, sectors[1]) -
4618 EXT4_SB(sb)->s_sectors_written_start) >> 1));
4620 es->s_kbytes_written =
4621 cpu_to_le64(EXT4_SB(sb)->s_kbytes_written);
4622 ext4_free_blocks_count_set(es,
4623 EXT4_C2B(EXT4_SB(sb), percpu_counter_sum_positive(
4624 &EXT4_SB(sb)->s_freeclusters_counter)));
4625 es->s_free_inodes_count =
4626 cpu_to_le32(percpu_counter_sum_positive(
4627 &EXT4_SB(sb)->s_freeinodes_counter));
4628 BUFFER_TRACE(sbh, "marking dirty");
4629 ext4_superblock_csum_set(sb);
4630 mark_buffer_dirty(sbh);
4632 error = sync_dirty_buffer(sbh);
4636 error = buffer_write_io_error(sbh);
4638 ext4_msg(sb, KERN_ERR, "I/O error while writing "
4640 clear_buffer_write_io_error(sbh);
4641 set_buffer_uptodate(sbh);
4648 * Have we just finished recovery? If so, and if we are mounting (or
4649 * remounting) the filesystem readonly, then we will end up with a
4650 * consistent fs on disk. Record that fact.
4652 static void ext4_mark_recovery_complete(struct super_block *sb,
4653 struct ext4_super_block *es)
4655 journal_t *journal = EXT4_SB(sb)->s_journal;
4657 if (!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL)) {
4658 BUG_ON(journal != NULL);
4661 jbd2_journal_lock_updates(journal);
4662 if (jbd2_journal_flush(journal) < 0)
4665 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER) &&
4666 sb->s_flags & MS_RDONLY) {
4667 EXT4_CLEAR_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
4668 ext4_commit_super(sb, 1);
4672 jbd2_journal_unlock_updates(journal);
4676 * If we are mounting (or read-write remounting) a filesystem whose journal
4677 * has recorded an error from a previous lifetime, move that error to the
4678 * main filesystem now.
4680 static void ext4_clear_journal_err(struct super_block *sb,
4681 struct ext4_super_block *es)
4687 BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL));
4689 journal = EXT4_SB(sb)->s_journal;
4692 * Now check for any error status which may have been recorded in the
4693 * journal by a prior ext4_error() or ext4_abort()
4696 j_errno = jbd2_journal_errno(journal);
4700 errstr = ext4_decode_error(sb, j_errno, nbuf);
4701 ext4_warning(sb, "Filesystem error recorded "
4702 "from previous mount: %s", errstr);
4703 ext4_warning(sb, "Marking fs in need of filesystem check.");
4705 EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS;
4706 es->s_state |= cpu_to_le16(EXT4_ERROR_FS);
4707 ext4_commit_super(sb, 1);
4709 jbd2_journal_clear_err(journal);
4710 jbd2_journal_update_sb_errno(journal);
4715 * Force the running and committing transactions to commit,
4716 * and wait on the commit.
4718 int ext4_force_commit(struct super_block *sb)
4722 if (sb->s_flags & MS_RDONLY)
4725 journal = EXT4_SB(sb)->s_journal;
4726 return ext4_journal_force_commit(journal);
4729 static int ext4_sync_fs(struct super_block *sb, int wait)
4733 bool needs_barrier = false;
4734 struct ext4_sb_info *sbi = EXT4_SB(sb);
4736 trace_ext4_sync_fs(sb, wait);
4737 flush_workqueue(sbi->rsv_conversion_wq);
4739 * Writeback quota in non-journalled quota case - journalled quota has
4742 dquot_writeback_dquots(sb, -1);
4744 * Data writeback is possible w/o journal transaction, so barrier must
4745 * being sent at the end of the function. But we can skip it if
4746 * transaction_commit will do it for us.
4748 target = jbd2_get_latest_transaction(sbi->s_journal);
4749 if (wait && sbi->s_journal->j_flags & JBD2_BARRIER &&
4750 !jbd2_trans_will_send_data_barrier(sbi->s_journal, target))
4751 needs_barrier = true;
4753 if (jbd2_journal_start_commit(sbi->s_journal, &target)) {
4755 ret = jbd2_log_wait_commit(sbi->s_journal, target);
4757 if (needs_barrier) {
4759 err = blkdev_issue_flush(sb->s_bdev, GFP_KERNEL, NULL);
4767 static int ext4_sync_fs_nojournal(struct super_block *sb, int wait)
4771 trace_ext4_sync_fs(sb, wait);
4772 flush_workqueue(EXT4_SB(sb)->rsv_conversion_wq);
4773 dquot_writeback_dquots(sb, -1);
4774 if (wait && test_opt(sb, BARRIER))
4775 ret = blkdev_issue_flush(sb->s_bdev, GFP_KERNEL, NULL);
4781 * LVM calls this function before a (read-only) snapshot is created. This
4782 * gives us a chance to flush the journal completely and mark the fs clean.
4784 * Note that only this function cannot bring a filesystem to be in a clean
4785 * state independently. It relies on upper layer to stop all data & metadata
4788 static int ext4_freeze(struct super_block *sb)
4793 if (sb->s_flags & MS_RDONLY)
4796 journal = EXT4_SB(sb)->s_journal;
4798 /* Now we set up the journal barrier. */
4799 jbd2_journal_lock_updates(journal);
4802 * Don't clear the needs_recovery flag if we failed to flush
4805 error = jbd2_journal_flush(journal);
4809 /* Journal blocked and flushed, clear needs_recovery flag. */
4810 EXT4_CLEAR_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
4811 error = ext4_commit_super(sb, 1);
4813 /* we rely on upper layer to stop further updates */
4814 jbd2_journal_unlock_updates(EXT4_SB(sb)->s_journal);
4819 * Called by LVM after the snapshot is done. We need to reset the RECOVER
4820 * flag here, even though the filesystem is not technically dirty yet.
4822 static int ext4_unfreeze(struct super_block *sb)
4824 if (sb->s_flags & MS_RDONLY)
4827 /* Reset the needs_recovery flag before the fs is unlocked. */
4828 EXT4_SET_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
4829 ext4_commit_super(sb, 1);
4834 * Structure to save mount options for ext4_remount's benefit
4836 struct ext4_mount_options {
4837 unsigned long s_mount_opt;
4838 unsigned long s_mount_opt2;
4841 unsigned long s_commit_interval;
4842 u32 s_min_batch_time, s_max_batch_time;
4845 char *s_qf_names[MAXQUOTAS];
4849 static int ext4_remount(struct super_block *sb, int *flags, char *data)
4851 struct ext4_super_block *es;
4852 struct ext4_sb_info *sbi = EXT4_SB(sb);
4853 unsigned long old_sb_flags;
4854 struct ext4_mount_options old_opts;
4855 int enable_quota = 0;
4857 unsigned int journal_ioprio = DEFAULT_JOURNAL_IOPRIO;
4862 char *orig_data = kstrdup(data, GFP_KERNEL);
4864 /* Store the original options */
4865 old_sb_flags = sb->s_flags;
4866 old_opts.s_mount_opt = sbi->s_mount_opt;
4867 old_opts.s_mount_opt2 = sbi->s_mount_opt2;
4868 old_opts.s_resuid = sbi->s_resuid;
4869 old_opts.s_resgid = sbi->s_resgid;
4870 old_opts.s_commit_interval = sbi->s_commit_interval;
4871 old_opts.s_min_batch_time = sbi->s_min_batch_time;
4872 old_opts.s_max_batch_time = sbi->s_max_batch_time;
4874 old_opts.s_jquota_fmt = sbi->s_jquota_fmt;
4875 for (i = 0; i < MAXQUOTAS; i++)
4876 if (sbi->s_qf_names[i]) {
4877 old_opts.s_qf_names[i] = kstrdup(sbi->s_qf_names[i],
4879 if (!old_opts.s_qf_names[i]) {
4880 for (j = 0; j < i; j++)
4881 kfree(old_opts.s_qf_names[j]);
4886 old_opts.s_qf_names[i] = NULL;
4888 if (sbi->s_journal && sbi->s_journal->j_task->io_context)
4889 journal_ioprio = sbi->s_journal->j_task->io_context->ioprio;
4892 * Allow the "check" option to be passed as a remount option.
4894 if (!parse_options(data, sb, NULL, &journal_ioprio, 1)) {
4899 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA) {
4900 if (test_opt2(sb, EXPLICIT_DELALLOC)) {
4901 ext4_msg(sb, KERN_ERR, "can't mount with "
4902 "both data=journal and delalloc");
4906 if (test_opt(sb, DIOREAD_NOLOCK)) {
4907 ext4_msg(sb, KERN_ERR, "can't mount with "
4908 "both data=journal and dioread_nolock");
4914 if (sbi->s_mount_flags & EXT4_MF_FS_ABORTED)
4915 ext4_abort(sb, "Abort forced by user");
4917 sb->s_flags = (sb->s_flags & ~MS_POSIXACL) |
4918 (test_opt(sb, POSIX_ACL) ? MS_POSIXACL : 0);
4922 if (sbi->s_journal) {
4923 ext4_init_journal_params(sb, sbi->s_journal);
4924 set_task_ioprio(sbi->s_journal->j_task, journal_ioprio);
4927 if ((*flags & MS_RDONLY) != (sb->s_flags & MS_RDONLY)) {
4928 if (sbi->s_mount_flags & EXT4_MF_FS_ABORTED) {
4933 if (*flags & MS_RDONLY) {
4934 err = sync_filesystem(sb);
4937 err = dquot_suspend(sb, -1);
4942 * First of all, the unconditional stuff we have to do
4943 * to disable replay of the journal when we next remount
4945 sb->s_flags |= MS_RDONLY;
4948 * OK, test if we are remounting a valid rw partition
4949 * readonly, and if so set the rdonly flag and then
4950 * mark the partition as valid again.
4952 if (!(es->s_state & cpu_to_le16(EXT4_VALID_FS)) &&
4953 (sbi->s_mount_state & EXT4_VALID_FS))
4954 es->s_state = cpu_to_le16(sbi->s_mount_state);
4957 ext4_mark_recovery_complete(sb, es);
4959 /* Make sure we can mount this feature set readwrite */
4960 if (!ext4_feature_set_ok(sb, 0)) {
4965 * Make sure the group descriptor checksums
4966 * are sane. If they aren't, refuse to remount r/w.
4968 for (g = 0; g < sbi->s_groups_count; g++) {
4969 struct ext4_group_desc *gdp =
4970 ext4_get_group_desc(sb, g, NULL);
4972 if (!ext4_group_desc_csum_verify(sb, g, gdp)) {
4973 ext4_msg(sb, KERN_ERR,
4974 "ext4_remount: Checksum for group %u failed (%u!=%u)",
4975 g, le16_to_cpu(ext4_group_desc_csum(sbi, g, gdp)),
4976 le16_to_cpu(gdp->bg_checksum));
4983 * If we have an unprocessed orphan list hanging
4984 * around from a previously readonly bdev mount,
4985 * require a full umount/remount for now.
4987 if (es->s_last_orphan) {
4988 ext4_msg(sb, KERN_WARNING, "Couldn't "
4989 "remount RDWR because of unprocessed "
4990 "orphan inode list. Please "
4991 "umount/remount instead");
4997 * Mounting a RDONLY partition read-write, so reread
4998 * and store the current valid flag. (It may have
4999 * been changed by e2fsck since we originally mounted
5003 ext4_clear_journal_err(sb, es);
5004 sbi->s_mount_state = le16_to_cpu(es->s_state);
5005 if (!ext4_setup_super(sb, es, 0))
5006 sb->s_flags &= ~MS_RDONLY;
5007 if (EXT4_HAS_INCOMPAT_FEATURE(sb,
5008 EXT4_FEATURE_INCOMPAT_MMP))
5009 if (ext4_multi_mount_protect(sb,
5010 le64_to_cpu(es->s_mmp_block))) {
5019 * Reinitialize lazy itable initialization thread based on
5022 if ((sb->s_flags & MS_RDONLY) || !test_opt(sb, INIT_INODE_TABLE))
5023 ext4_unregister_li_request(sb);
5025 ext4_group_t first_not_zeroed;
5026 first_not_zeroed = ext4_has_uninit_itable(sb);
5027 ext4_register_li_request(sb, first_not_zeroed);
5030 ext4_setup_system_zone(sb);
5031 if (sbi->s_journal == NULL && !(old_sb_flags & MS_RDONLY))
5032 ext4_commit_super(sb, 1);
5035 /* Release old quota file names */
5036 for (i = 0; i < MAXQUOTAS; i++)
5037 kfree(old_opts.s_qf_names[i]);
5039 if (sb_any_quota_suspended(sb))
5040 dquot_resume(sb, -1);
5041 else if (EXT4_HAS_RO_COMPAT_FEATURE(sb,
5042 EXT4_FEATURE_RO_COMPAT_QUOTA)) {
5043 err = ext4_enable_quotas(sb);
5050 ext4_msg(sb, KERN_INFO, "re-mounted. Opts: %s", orig_data);
5055 sb->s_flags = old_sb_flags;
5056 sbi->s_mount_opt = old_opts.s_mount_opt;
5057 sbi->s_mount_opt2 = old_opts.s_mount_opt2;
5058 sbi->s_resuid = old_opts.s_resuid;
5059 sbi->s_resgid = old_opts.s_resgid;
5060 sbi->s_commit_interval = old_opts.s_commit_interval;
5061 sbi->s_min_batch_time = old_opts.s_min_batch_time;
5062 sbi->s_max_batch_time = old_opts.s_max_batch_time;
5064 sbi->s_jquota_fmt = old_opts.s_jquota_fmt;
5065 for (i = 0; i < MAXQUOTAS; i++) {
5066 kfree(sbi->s_qf_names[i]);
5067 sbi->s_qf_names[i] = old_opts.s_qf_names[i];
5074 static int ext4_statfs(struct dentry *dentry, struct kstatfs *buf)
5076 struct super_block *sb = dentry->d_sb;
5077 struct ext4_sb_info *sbi = EXT4_SB(sb);
5078 struct ext4_super_block *es = sbi->s_es;
5079 ext4_fsblk_t overhead = 0, resv_blocks;
5082 resv_blocks = EXT4_C2B(sbi, atomic64_read(&sbi->s_resv_clusters));
5084 if (!test_opt(sb, MINIX_DF))
5085 overhead = sbi->s_overhead;
5087 buf->f_type = EXT4_SUPER_MAGIC;
5088 buf->f_bsize = sb->s_blocksize;
5089 buf->f_blocks = ext4_blocks_count(es) - EXT4_C2B(sbi, overhead);
5090 bfree = percpu_counter_sum_positive(&sbi->s_freeclusters_counter) -
5091 percpu_counter_sum_positive(&sbi->s_dirtyclusters_counter);
5092 /* prevent underflow in case that few free space is available */
5093 buf->f_bfree = EXT4_C2B(sbi, max_t(s64, bfree, 0));
5094 buf->f_bavail = buf->f_bfree -
5095 (ext4_r_blocks_count(es) + resv_blocks);
5096 if (buf->f_bfree < (ext4_r_blocks_count(es) + resv_blocks))
5098 buf->f_files = le32_to_cpu(es->s_inodes_count);
5099 buf->f_ffree = percpu_counter_sum_positive(&sbi->s_freeinodes_counter);
5100 buf->f_namelen = EXT4_NAME_LEN;
5101 fsid = le64_to_cpup((void *)es->s_uuid) ^
5102 le64_to_cpup((void *)es->s_uuid + sizeof(u64));
5103 buf->f_fsid.val[0] = fsid & 0xFFFFFFFFUL;
5104 buf->f_fsid.val[1] = (fsid >> 32) & 0xFFFFFFFFUL;
5109 /* Helper function for writing quotas on sync - we need to start transaction
5110 * before quota file is locked for write. Otherwise the are possible deadlocks:
5111 * Process 1 Process 2
5112 * ext4_create() quota_sync()
5113 * jbd2_journal_start() write_dquot()
5114 * dquot_initialize() down(dqio_mutex)
5115 * down(dqio_mutex) jbd2_journal_start()
5121 static inline struct inode *dquot_to_inode(struct dquot *dquot)
5123 return sb_dqopt(dquot->dq_sb)->files[dquot->dq_id.type];
5126 static int ext4_write_dquot(struct dquot *dquot)
5130 struct inode *inode;
5132 inode = dquot_to_inode(dquot);
5133 handle = ext4_journal_start(inode, EXT4_HT_QUOTA,
5134 EXT4_QUOTA_TRANS_BLOCKS(dquot->dq_sb));
5136 return PTR_ERR(handle);
5137 ret = dquot_commit(dquot);
5138 err = ext4_journal_stop(handle);
5144 static int ext4_acquire_dquot(struct dquot *dquot)
5149 handle = ext4_journal_start(dquot_to_inode(dquot), EXT4_HT_QUOTA,
5150 EXT4_QUOTA_INIT_BLOCKS(dquot->dq_sb));
5152 return PTR_ERR(handle);
5153 ret = dquot_acquire(dquot);
5154 err = ext4_journal_stop(handle);
5160 static int ext4_release_dquot(struct dquot *dquot)
5165 handle = ext4_journal_start(dquot_to_inode(dquot), EXT4_HT_QUOTA,
5166 EXT4_QUOTA_DEL_BLOCKS(dquot->dq_sb));
5167 if (IS_ERR(handle)) {
5168 /* Release dquot anyway to avoid endless cycle in dqput() */
5169 dquot_release(dquot);
5170 return PTR_ERR(handle);
5172 ret = dquot_release(dquot);
5173 err = ext4_journal_stop(handle);
5179 static int ext4_mark_dquot_dirty(struct dquot *dquot)
5181 struct super_block *sb = dquot->dq_sb;
5182 struct ext4_sb_info *sbi = EXT4_SB(sb);
5184 /* Are we journaling quotas? */
5185 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_QUOTA) ||
5186 sbi->s_qf_names[USRQUOTA] || sbi->s_qf_names[GRPQUOTA]) {
5187 dquot_mark_dquot_dirty(dquot);
5188 return ext4_write_dquot(dquot);
5190 return dquot_mark_dquot_dirty(dquot);
5194 static int ext4_write_info(struct super_block *sb, int type)
5199 /* Data block + inode block */
5200 handle = ext4_journal_start(sb->s_root->d_inode, EXT4_HT_QUOTA, 2);
5202 return PTR_ERR(handle);
5203 ret = dquot_commit_info(sb, type);
5204 err = ext4_journal_stop(handle);
5211 * Turn on quotas during mount time - we need to find
5212 * the quota file and such...
5214 static int ext4_quota_on_mount(struct super_block *sb, int type)
5216 return dquot_quota_on_mount(sb, EXT4_SB(sb)->s_qf_names[type],
5217 EXT4_SB(sb)->s_jquota_fmt, type);
5220 static void lockdep_set_quota_inode(struct inode *inode, int subclass)
5222 struct ext4_inode_info *ei = EXT4_I(inode);
5224 /* The first argument of lockdep_set_subclass has to be
5225 * *exactly* the same as the argument to init_rwsem() --- in
5226 * this case, in init_once() --- or lockdep gets unhappy
5227 * because the name of the lock is set using the
5228 * stringification of the argument to init_rwsem().
5230 (void) ei; /* shut up clang warning if !CONFIG_LOCKDEP */
5231 lockdep_set_subclass(&ei->i_data_sem, subclass);
5235 * Standard function to be called on quota_on
5237 static int ext4_quota_on(struct super_block *sb, int type, int format_id,
5242 if (!test_opt(sb, QUOTA))
5245 /* Quotafile not on the same filesystem? */
5246 if (path->dentry->d_sb != sb)
5248 /* Journaling quota? */
5249 if (EXT4_SB(sb)->s_qf_names[type]) {
5250 /* Quotafile not in fs root? */
5251 if (path->dentry->d_parent != sb->s_root)
5252 ext4_msg(sb, KERN_WARNING,
5253 "Quota file not on filesystem root. "
5254 "Journaled quota will not work");
5258 * When we journal data on quota file, we have to flush journal to see
5259 * all updates to the file when we bypass pagecache...
5261 if (EXT4_SB(sb)->s_journal &&
5262 ext4_should_journal_data(path->dentry->d_inode)) {
5264 * We don't need to lock updates but journal_flush() could
5265 * otherwise be livelocked...
5267 jbd2_journal_lock_updates(EXT4_SB(sb)->s_journal);
5268 err = jbd2_journal_flush(EXT4_SB(sb)->s_journal);
5269 jbd2_journal_unlock_updates(EXT4_SB(sb)->s_journal);
5273 lockdep_set_quota_inode(path->dentry->d_inode, I_DATA_SEM_QUOTA);
5274 err = dquot_quota_on(sb, type, format_id, path);
5276 lockdep_set_quota_inode(path->dentry->d_inode,
5281 static int ext4_quota_enable(struct super_block *sb, int type, int format_id,
5285 struct inode *qf_inode;
5286 unsigned long qf_inums[MAXQUOTAS] = {
5287 le32_to_cpu(EXT4_SB(sb)->s_es->s_usr_quota_inum),
5288 le32_to_cpu(EXT4_SB(sb)->s_es->s_grp_quota_inum)
5291 BUG_ON(!EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_QUOTA));
5293 if (!qf_inums[type])
5296 qf_inode = ext4_iget(sb, qf_inums[type]);
5297 if (IS_ERR(qf_inode)) {
5298 ext4_error(sb, "Bad quota inode # %lu", qf_inums[type]);
5299 return PTR_ERR(qf_inode);
5302 /* Don't account quota for quota files to avoid recursion */
5303 qf_inode->i_flags |= S_NOQUOTA;
5304 lockdep_set_quota_inode(qf_inode, I_DATA_SEM_QUOTA);
5305 err = dquot_enable(qf_inode, type, format_id, flags);
5308 lockdep_set_quota_inode(qf_inode, I_DATA_SEM_NORMAL);
5313 /* Enable usage tracking for all quota types. */
5314 static int ext4_enable_quotas(struct super_block *sb)
5317 unsigned long qf_inums[MAXQUOTAS] = {
5318 le32_to_cpu(EXT4_SB(sb)->s_es->s_usr_quota_inum),
5319 le32_to_cpu(EXT4_SB(sb)->s_es->s_grp_quota_inum)
5322 sb_dqopt(sb)->flags |= DQUOT_QUOTA_SYS_FILE;
5323 for (type = 0; type < MAXQUOTAS; type++) {
5324 if (qf_inums[type]) {
5325 err = ext4_quota_enable(sb, type, QFMT_VFS_V1,
5326 DQUOT_USAGE_ENABLED);
5329 "Failed to enable quota tracking "
5330 "(type=%d, err=%d). Please run "
5331 "e2fsck to fix.", type, err);
5340 * quota_on function that is used when QUOTA feature is set.
5342 static int ext4_quota_on_sysfile(struct super_block *sb, int type,
5345 if (!EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_QUOTA))
5349 * USAGE was enabled at mount time. Only need to enable LIMITS now.
5351 return ext4_quota_enable(sb, type, format_id, DQUOT_LIMITS_ENABLED);
5354 static int ext4_quota_off(struct super_block *sb, int type)
5356 struct inode *inode = sb_dqopt(sb)->files[type];
5359 /* Force all delayed allocation blocks to be allocated.
5360 * Caller already holds s_umount sem */
5361 if (test_opt(sb, DELALLOC))
5362 sync_filesystem(sb);
5367 /* Update modification times of quota files when userspace can
5368 * start looking at them */
5369 handle = ext4_journal_start(inode, EXT4_HT_QUOTA, 1);
5372 inode->i_mtime = inode->i_ctime = CURRENT_TIME;
5373 ext4_mark_inode_dirty(handle, inode);
5374 ext4_journal_stop(handle);
5377 return dquot_quota_off(sb, type);
5381 * quota_off function that is used when QUOTA feature is set.
5383 static int ext4_quota_off_sysfile(struct super_block *sb, int type)
5385 if (!EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_QUOTA))
5388 /* Disable only the limits. */
5389 return dquot_disable(sb, type, DQUOT_LIMITS_ENABLED);
5392 /* Read data from quotafile - avoid pagecache and such because we cannot afford
5393 * acquiring the locks... As quota files are never truncated and quota code
5394 * itself serializes the operations (and no one else should touch the files)
5395 * we don't have to be afraid of races */
5396 static ssize_t ext4_quota_read(struct super_block *sb, int type, char *data,
5397 size_t len, loff_t off)
5399 struct inode *inode = sb_dqopt(sb)->files[type];
5400 ext4_lblk_t blk = off >> EXT4_BLOCK_SIZE_BITS(sb);
5402 int offset = off & (sb->s_blocksize - 1);
5405 struct buffer_head *bh;
5406 loff_t i_size = i_size_read(inode);
5410 if (off+len > i_size)
5413 while (toread > 0) {
5414 tocopy = sb->s_blocksize - offset < toread ?
5415 sb->s_blocksize - offset : toread;
5416 bh = ext4_bread(NULL, inode, blk, 0, &err);
5419 if (!bh) /* A hole? */
5420 memset(data, 0, tocopy);
5422 memcpy(data, bh->b_data+offset, tocopy);
5432 /* Write to quotafile (we know the transaction is already started and has
5433 * enough credits) */
5434 static ssize_t ext4_quota_write(struct super_block *sb, int type,
5435 const char *data, size_t len, loff_t off)
5437 struct inode *inode = sb_dqopt(sb)->files[type];
5438 ext4_lblk_t blk = off >> EXT4_BLOCK_SIZE_BITS(sb);
5440 int offset = off & (sb->s_blocksize - 1);
5441 struct buffer_head *bh;
5442 handle_t *handle = journal_current_handle();
5444 if (EXT4_SB(sb)->s_journal && !handle) {
5445 ext4_msg(sb, KERN_WARNING, "Quota write (off=%llu, len=%llu)"
5446 " cancelled because transaction is not started",
5447 (unsigned long long)off, (unsigned long long)len);
5451 * Since we account only one data block in transaction credits,
5452 * then it is impossible to cross a block boundary.
5454 if (sb->s_blocksize - offset < len) {
5455 ext4_msg(sb, KERN_WARNING, "Quota write (off=%llu, len=%llu)"
5456 " cancelled because not block aligned",
5457 (unsigned long long)off, (unsigned long long)len);
5461 bh = ext4_bread(handle, inode, blk, 1, &err);
5464 BUFFER_TRACE(bh, "get write access");
5465 err = ext4_journal_get_write_access(handle, bh);
5471 memcpy(bh->b_data+offset, data, len);
5472 flush_dcache_page(bh->b_page);
5474 err = ext4_handle_dirty_metadata(handle, NULL, bh);
5479 if (inode->i_size < off + len) {
5480 i_size_write(inode, off + len);
5481 EXT4_I(inode)->i_disksize = inode->i_size;
5482 ext4_mark_inode_dirty(handle, inode);
5489 static struct dentry *ext4_mount(struct file_system_type *fs_type, int flags,
5490 const char *dev_name, void *data)
5492 return mount_bdev(fs_type, flags, dev_name, data, ext4_fill_super);
5495 #if !defined(CONFIG_EXT2_FS) && !defined(CONFIG_EXT2_FS_MODULE) && defined(CONFIG_EXT4_USE_FOR_EXT23)
5496 static inline void register_as_ext2(void)
5498 int err = register_filesystem(&ext2_fs_type);
5501 "EXT4-fs: Unable to register as ext2 (%d)\n", err);
5504 static inline void unregister_as_ext2(void)
5506 unregister_filesystem(&ext2_fs_type);
5509 static inline int ext2_feature_set_ok(struct super_block *sb)
5511 if (EXT4_HAS_INCOMPAT_FEATURE(sb, ~EXT2_FEATURE_INCOMPAT_SUPP))
5513 if (sb->s_flags & MS_RDONLY)
5515 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, ~EXT2_FEATURE_RO_COMPAT_SUPP))
5520 static inline void register_as_ext2(void) { }
5521 static inline void unregister_as_ext2(void) { }
5522 static inline int ext2_feature_set_ok(struct super_block *sb) { return 0; }
5525 #if !defined(CONFIG_EXT3_FS) && !defined(CONFIG_EXT3_FS_MODULE) && defined(CONFIG_EXT4_USE_FOR_EXT23)
5526 static inline void register_as_ext3(void)
5528 int err = register_filesystem(&ext3_fs_type);
5531 "EXT4-fs: Unable to register as ext3 (%d)\n", err);
5534 static inline void unregister_as_ext3(void)
5536 unregister_filesystem(&ext3_fs_type);
5539 static inline int ext3_feature_set_ok(struct super_block *sb)
5541 if (EXT4_HAS_INCOMPAT_FEATURE(sb, ~EXT3_FEATURE_INCOMPAT_SUPP))
5543 if (!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL))
5545 if (sb->s_flags & MS_RDONLY)
5547 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, ~EXT3_FEATURE_RO_COMPAT_SUPP))
5552 static inline void register_as_ext3(void) { }
5553 static inline void unregister_as_ext3(void) { }
5554 static inline int ext3_feature_set_ok(struct super_block *sb) { return 0; }
5557 static struct file_system_type ext4_fs_type = {
5558 .owner = THIS_MODULE,
5560 .mount = ext4_mount,
5561 .kill_sb = kill_block_super,
5562 .fs_flags = FS_REQUIRES_DEV,
5564 MODULE_ALIAS_FS("ext4");
5566 static int __init ext4_init_feat_adverts(void)
5568 struct ext4_features *ef;
5571 ef = kzalloc(sizeof(struct ext4_features), GFP_KERNEL);
5575 ef->f_kobj.kset = ext4_kset;
5576 init_completion(&ef->f_kobj_unregister);
5577 ret = kobject_init_and_add(&ef->f_kobj, &ext4_feat_ktype, NULL,
5590 static void ext4_exit_feat_adverts(void)
5592 kobject_put(&ext4_feat->f_kobj);
5593 wait_for_completion(&ext4_feat->f_kobj_unregister);
5597 /* Shared across all ext4 file systems */
5598 wait_queue_head_t ext4__ioend_wq[EXT4_WQ_HASH_SZ];
5599 struct mutex ext4__aio_mutex[EXT4_WQ_HASH_SZ];
5601 static int __init ext4_init_fs(void)
5605 ext4_li_info = NULL;
5606 mutex_init(&ext4_li_mtx);
5608 /* Build-time check for flags consistency */
5609 ext4_check_flag_values();
5611 for (i = 0; i < EXT4_WQ_HASH_SZ; i++) {
5612 mutex_init(&ext4__aio_mutex[i]);
5613 init_waitqueue_head(&ext4__ioend_wq[i]);
5616 err = ext4_init_es();
5620 err = ext4_init_pageio();
5624 err = ext4_init_system_zone();
5627 ext4_kset = kset_create_and_add("ext4", NULL, fs_kobj);
5632 ext4_proc_root = proc_mkdir("fs/ext4", NULL);
5634 err = ext4_init_feat_adverts();
5638 err = ext4_init_mballoc();
5642 ext4_mballoc_ready = 1;
5643 err = init_inodecache();
5648 err = register_filesystem(&ext4_fs_type);
5654 unregister_as_ext2();
5655 unregister_as_ext3();
5656 destroy_inodecache();
5658 ext4_mballoc_ready = 0;
5659 ext4_exit_mballoc();
5661 ext4_exit_feat_adverts();
5664 remove_proc_entry("fs/ext4", NULL);
5665 kset_unregister(ext4_kset);
5667 ext4_exit_system_zone();
5676 static void __exit ext4_exit_fs(void)
5678 ext4_destroy_lazyinit_thread();
5679 unregister_as_ext2();
5680 unregister_as_ext3();
5681 unregister_filesystem(&ext4_fs_type);
5682 destroy_inodecache();
5683 ext4_exit_mballoc();
5684 ext4_exit_feat_adverts();
5685 remove_proc_entry("fs/ext4", NULL);
5686 kset_unregister(ext4_kset);
5687 ext4_exit_system_zone();
5692 MODULE_AUTHOR("Remy Card, Stephen Tweedie, Andrew Morton, Andreas Dilger, Theodore Ts'o and others");
5693 MODULE_DESCRIPTION("Fourth Extended Filesystem");
5694 MODULE_LICENSE("GPL");
5695 module_init(ext4_init_fs)
5696 module_exit(ext4_exit_fs)