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"
49 #include "ext4_jbd2.h"
54 #define CREATE_TRACE_POINTS
55 #include <trace/events/ext4.h>
57 static struct proc_dir_entry *ext4_proc_root;
58 static struct kset *ext4_kset;
59 static struct ext4_lazy_init *ext4_li_info;
60 static struct mutex ext4_li_mtx;
61 static struct ext4_features *ext4_feat;
63 static int ext4_load_journal(struct super_block *, struct ext4_super_block *,
64 unsigned long journal_devnum);
65 static int ext4_show_options(struct seq_file *seq, struct dentry *root);
66 static int ext4_commit_super(struct super_block *sb, int sync);
67 static void ext4_mark_recovery_complete(struct super_block *sb,
68 struct ext4_super_block *es);
69 static void ext4_clear_journal_err(struct super_block *sb,
70 struct ext4_super_block *es);
71 static int ext4_sync_fs(struct super_block *sb, int wait);
72 static const char *ext4_decode_error(struct super_block *sb, int errno,
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);
87 #if !defined(CONFIG_EXT2_FS) && !defined(CONFIG_EXT2_FS_MODULE) && defined(CONFIG_EXT4_USE_FOR_EXT23)
88 static struct file_system_type ext2_fs_type = {
92 .kill_sb = kill_block_super,
93 .fs_flags = FS_REQUIRES_DEV,
95 #define IS_EXT2_SB(sb) ((sb)->s_bdev->bd_holder == &ext2_fs_type)
97 #define IS_EXT2_SB(sb) (0)
101 #if !defined(CONFIG_EXT3_FS) && !defined(CONFIG_EXT3_FS_MODULE) && defined(CONFIG_EXT4_USE_FOR_EXT23)
102 static struct file_system_type ext3_fs_type = {
103 .owner = THIS_MODULE,
106 .kill_sb = kill_block_super,
107 .fs_flags = FS_REQUIRES_DEV,
109 #define IS_EXT3_SB(sb) ((sb)->s_bdev->bd_holder == &ext3_fs_type)
111 #define IS_EXT3_SB(sb) (0)
114 static int ext4_verify_csum_type(struct super_block *sb,
115 struct ext4_super_block *es)
117 if (!EXT4_HAS_RO_COMPAT_FEATURE(sb,
118 EXT4_FEATURE_RO_COMPAT_METADATA_CSUM))
121 return es->s_checksum_type == EXT4_CRC32C_CHKSUM;
124 static __le32 ext4_superblock_csum(struct super_block *sb,
125 struct ext4_super_block *es)
127 struct ext4_sb_info *sbi = EXT4_SB(sb);
128 int offset = offsetof(struct ext4_super_block, s_checksum);
131 csum = ext4_chksum(sbi, ~0, (char *)es, offset);
133 return cpu_to_le32(csum);
136 int ext4_superblock_csum_verify(struct super_block *sb,
137 struct ext4_super_block *es)
139 if (!EXT4_HAS_RO_COMPAT_FEATURE(sb,
140 EXT4_FEATURE_RO_COMPAT_METADATA_CSUM))
143 return es->s_checksum == ext4_superblock_csum(sb, es);
146 void ext4_superblock_csum_set(struct super_block *sb,
147 struct ext4_super_block *es)
149 if (!EXT4_HAS_RO_COMPAT_FEATURE(sb,
150 EXT4_FEATURE_RO_COMPAT_METADATA_CSUM))
153 es->s_checksum = ext4_superblock_csum(sb, es);
156 void *ext4_kvmalloc(size_t size, gfp_t flags)
160 ret = kmalloc(size, flags);
162 ret = __vmalloc(size, flags, PAGE_KERNEL);
166 void *ext4_kvzalloc(size_t size, gfp_t flags)
170 ret = kzalloc(size, flags);
172 ret = __vmalloc(size, flags | __GFP_ZERO, PAGE_KERNEL);
176 void ext4_kvfree(void *ptr)
178 if (is_vmalloc_addr(ptr))
185 ext4_fsblk_t ext4_block_bitmap(struct super_block *sb,
186 struct ext4_group_desc *bg)
188 return le32_to_cpu(bg->bg_block_bitmap_lo) |
189 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
190 (ext4_fsblk_t)le32_to_cpu(bg->bg_block_bitmap_hi) << 32 : 0);
193 ext4_fsblk_t ext4_inode_bitmap(struct super_block *sb,
194 struct ext4_group_desc *bg)
196 return le32_to_cpu(bg->bg_inode_bitmap_lo) |
197 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
198 (ext4_fsblk_t)le32_to_cpu(bg->bg_inode_bitmap_hi) << 32 : 0);
201 ext4_fsblk_t ext4_inode_table(struct super_block *sb,
202 struct ext4_group_desc *bg)
204 return le32_to_cpu(bg->bg_inode_table_lo) |
205 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
206 (ext4_fsblk_t)le32_to_cpu(bg->bg_inode_table_hi) << 32 : 0);
209 __u32 ext4_free_group_clusters(struct super_block *sb,
210 struct ext4_group_desc *bg)
212 return le16_to_cpu(bg->bg_free_blocks_count_lo) |
213 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
214 (__u32)le16_to_cpu(bg->bg_free_blocks_count_hi) << 16 : 0);
217 __u32 ext4_free_inodes_count(struct super_block *sb,
218 struct ext4_group_desc *bg)
220 return le16_to_cpu(bg->bg_free_inodes_count_lo) |
221 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
222 (__u32)le16_to_cpu(bg->bg_free_inodes_count_hi) << 16 : 0);
225 __u32 ext4_used_dirs_count(struct super_block *sb,
226 struct ext4_group_desc *bg)
228 return le16_to_cpu(bg->bg_used_dirs_count_lo) |
229 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
230 (__u32)le16_to_cpu(bg->bg_used_dirs_count_hi) << 16 : 0);
233 __u32 ext4_itable_unused_count(struct super_block *sb,
234 struct ext4_group_desc *bg)
236 return le16_to_cpu(bg->bg_itable_unused_lo) |
237 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
238 (__u32)le16_to_cpu(bg->bg_itable_unused_hi) << 16 : 0);
241 void ext4_block_bitmap_set(struct super_block *sb,
242 struct ext4_group_desc *bg, ext4_fsblk_t blk)
244 bg->bg_block_bitmap_lo = cpu_to_le32((u32)blk);
245 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
246 bg->bg_block_bitmap_hi = cpu_to_le32(blk >> 32);
249 void ext4_inode_bitmap_set(struct super_block *sb,
250 struct ext4_group_desc *bg, ext4_fsblk_t blk)
252 bg->bg_inode_bitmap_lo = cpu_to_le32((u32)blk);
253 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
254 bg->bg_inode_bitmap_hi = cpu_to_le32(blk >> 32);
257 void ext4_inode_table_set(struct super_block *sb,
258 struct ext4_group_desc *bg, ext4_fsblk_t blk)
260 bg->bg_inode_table_lo = cpu_to_le32((u32)blk);
261 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
262 bg->bg_inode_table_hi = cpu_to_le32(blk >> 32);
265 void ext4_free_group_clusters_set(struct super_block *sb,
266 struct ext4_group_desc *bg, __u32 count)
268 bg->bg_free_blocks_count_lo = cpu_to_le16((__u16)count);
269 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
270 bg->bg_free_blocks_count_hi = cpu_to_le16(count >> 16);
273 void ext4_free_inodes_set(struct super_block *sb,
274 struct ext4_group_desc *bg, __u32 count)
276 bg->bg_free_inodes_count_lo = cpu_to_le16((__u16)count);
277 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
278 bg->bg_free_inodes_count_hi = cpu_to_le16(count >> 16);
281 void ext4_used_dirs_set(struct super_block *sb,
282 struct ext4_group_desc *bg, __u32 count)
284 bg->bg_used_dirs_count_lo = cpu_to_le16((__u16)count);
285 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
286 bg->bg_used_dirs_count_hi = cpu_to_le16(count >> 16);
289 void ext4_itable_unused_set(struct super_block *sb,
290 struct ext4_group_desc *bg, __u32 count)
292 bg->bg_itable_unused_lo = cpu_to_le16((__u16)count);
293 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
294 bg->bg_itable_unused_hi = cpu_to_le16(count >> 16);
298 /* Just increment the non-pointer handle value */
299 static handle_t *ext4_get_nojournal(void)
301 handle_t *handle = current->journal_info;
302 unsigned long ref_cnt = (unsigned long)handle;
304 BUG_ON(ref_cnt >= EXT4_NOJOURNAL_MAX_REF_COUNT);
307 handle = (handle_t *)ref_cnt;
309 current->journal_info = handle;
314 /* Decrement the non-pointer handle value */
315 static void ext4_put_nojournal(handle_t *handle)
317 unsigned long ref_cnt = (unsigned long)handle;
319 BUG_ON(ref_cnt == 0);
322 handle = (handle_t *)ref_cnt;
324 current->journal_info = handle;
328 * Wrappers for jbd2_journal_start/end.
330 * The only special thing we need to do here is to make sure that all
331 * journal_end calls result in the superblock being marked dirty, so
332 * that sync() will call the filesystem's write_super callback if
335 handle_t *ext4_journal_start_sb(struct super_block *sb, int nblocks)
339 trace_ext4_journal_start(sb, nblocks, _RET_IP_);
340 if (sb->s_flags & MS_RDONLY)
341 return ERR_PTR(-EROFS);
343 WARN_ON(sb->s_writers.frozen == SB_FREEZE_COMPLETE);
344 journal = EXT4_SB(sb)->s_journal;
346 return ext4_get_nojournal();
348 * Special case here: if the journal has aborted behind our
349 * backs (eg. EIO in the commit thread), then we still need to
350 * take the FS itself readonly cleanly.
352 if (is_journal_aborted(journal)) {
353 ext4_abort(sb, "Detected aborted journal");
354 return ERR_PTR(-EROFS);
356 return jbd2_journal_start(journal, nblocks);
360 * The only special thing we need to do here is to make sure that all
361 * jbd2_journal_stop calls result in the superblock being marked dirty, so
362 * that sync() will call the filesystem's write_super callback if
365 int __ext4_journal_stop(const char *where, unsigned int line, handle_t *handle)
367 struct super_block *sb;
371 if (!ext4_handle_valid(handle)) {
372 ext4_put_nojournal(handle);
375 sb = handle->h_transaction->t_journal->j_private;
377 rc = jbd2_journal_stop(handle);
382 __ext4_std_error(sb, where, line, err);
386 void ext4_journal_abort_handle(const char *caller, unsigned int line,
387 const char *err_fn, struct buffer_head *bh,
388 handle_t *handle, int err)
391 const char *errstr = ext4_decode_error(NULL, err, nbuf);
393 BUG_ON(!ext4_handle_valid(handle));
396 BUFFER_TRACE(bh, "abort");
401 if (is_handle_aborted(handle))
404 printk(KERN_ERR "EXT4-fs: %s:%d: aborting transaction: %s in %s\n",
405 caller, line, errstr, err_fn);
407 jbd2_journal_abort_handle(handle);
410 static void __save_error_info(struct super_block *sb, const char *func,
413 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
415 EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS;
416 es->s_state |= cpu_to_le16(EXT4_ERROR_FS);
417 es->s_last_error_time = cpu_to_le32(get_seconds());
418 strncpy(es->s_last_error_func, func, sizeof(es->s_last_error_func));
419 es->s_last_error_line = cpu_to_le32(line);
420 if (!es->s_first_error_time) {
421 es->s_first_error_time = es->s_last_error_time;
422 strncpy(es->s_first_error_func, func,
423 sizeof(es->s_first_error_func));
424 es->s_first_error_line = cpu_to_le32(line);
425 es->s_first_error_ino = es->s_last_error_ino;
426 es->s_first_error_block = es->s_last_error_block;
429 * Start the daily error reporting function if it hasn't been
432 if (!es->s_error_count)
433 mod_timer(&EXT4_SB(sb)->s_err_report, jiffies + 24*60*60*HZ);
434 es->s_error_count = cpu_to_le32(le32_to_cpu(es->s_error_count) + 1);
437 static void save_error_info(struct super_block *sb, const char *func,
440 __save_error_info(sb, func, line);
441 ext4_commit_super(sb, 1);
445 * The del_gendisk() function uninitializes the disk-specific data
446 * structures, including the bdi structure, without telling anyone
447 * else. Once this happens, any attempt to call mark_buffer_dirty()
448 * (for example, by ext4_commit_super), will cause a kernel OOPS.
449 * This is a kludge to prevent these oops until we can put in a proper
450 * hook in del_gendisk() to inform the VFS and file system layers.
452 static int block_device_ejected(struct super_block *sb)
454 struct inode *bd_inode = sb->s_bdev->bd_inode;
455 struct backing_dev_info *bdi = bd_inode->i_mapping->backing_dev_info;
457 return bdi->dev == NULL;
460 static void ext4_journal_commit_callback(journal_t *journal, transaction_t *txn)
462 struct super_block *sb = journal->j_private;
463 struct ext4_sb_info *sbi = EXT4_SB(sb);
464 int error = is_journal_aborted(journal);
465 struct ext4_journal_cb_entry *jce, *tmp;
467 spin_lock(&sbi->s_md_lock);
468 list_for_each_entry_safe(jce, tmp, &txn->t_private_list, jce_list) {
469 list_del_init(&jce->jce_list);
470 spin_unlock(&sbi->s_md_lock);
471 jce->jce_func(sb, jce, error);
472 spin_lock(&sbi->s_md_lock);
474 spin_unlock(&sbi->s_md_lock);
477 /* Deal with the reporting of failure conditions on a filesystem such as
478 * inconsistencies detected or read IO failures.
480 * On ext2, we can store the error state of the filesystem in the
481 * superblock. That is not possible on ext4, because we may have other
482 * write ordering constraints on the superblock which prevent us from
483 * writing it out straight away; and given that the journal is about to
484 * be aborted, we can't rely on the current, or future, transactions to
485 * write out the superblock safely.
487 * We'll just use the jbd2_journal_abort() error code to record an error in
488 * the journal instead. On recovery, the journal will complain about
489 * that error until we've noted it down and cleared it.
492 static void ext4_handle_error(struct super_block *sb)
494 if (sb->s_flags & MS_RDONLY)
497 if (!test_opt(sb, ERRORS_CONT)) {
498 journal_t *journal = EXT4_SB(sb)->s_journal;
500 EXT4_SB(sb)->s_mount_flags |= EXT4_MF_FS_ABORTED;
502 jbd2_journal_abort(journal, -EIO);
504 if (test_opt(sb, ERRORS_RO)) {
505 ext4_msg(sb, KERN_CRIT, "Remounting filesystem read-only");
506 sb->s_flags |= MS_RDONLY;
508 if (test_opt(sb, ERRORS_PANIC))
509 panic("EXT4-fs (device %s): panic forced after error\n",
513 void __ext4_error(struct super_block *sb, const char *function,
514 unsigned int line, const char *fmt, ...)
516 struct va_format vaf;
522 printk(KERN_CRIT "EXT4-fs error (device %s): %s:%d: comm %s: %pV\n",
523 sb->s_id, function, line, current->comm, &vaf);
525 save_error_info(sb, function, line);
527 ext4_handle_error(sb);
530 void ext4_error_inode(struct inode *inode, const char *function,
531 unsigned int line, ext4_fsblk_t block,
532 const char *fmt, ...)
535 struct va_format vaf;
536 struct ext4_super_block *es = EXT4_SB(inode->i_sb)->s_es;
538 es->s_last_error_ino = cpu_to_le32(inode->i_ino);
539 es->s_last_error_block = cpu_to_le64(block);
540 save_error_info(inode->i_sb, function, line);
545 printk(KERN_CRIT "EXT4-fs error (device %s): %s:%d: "
546 "inode #%lu: block %llu: comm %s: %pV\n",
547 inode->i_sb->s_id, function, line, inode->i_ino,
548 block, current->comm, &vaf);
550 printk(KERN_CRIT "EXT4-fs error (device %s): %s:%d: "
551 "inode #%lu: comm %s: %pV\n",
552 inode->i_sb->s_id, function, line, inode->i_ino,
553 current->comm, &vaf);
556 ext4_handle_error(inode->i_sb);
559 void ext4_error_file(struct file *file, const char *function,
560 unsigned int line, ext4_fsblk_t block,
561 const char *fmt, ...)
564 struct va_format vaf;
565 struct ext4_super_block *es;
566 struct inode *inode = file->f_dentry->d_inode;
567 char pathname[80], *path;
569 es = EXT4_SB(inode->i_sb)->s_es;
570 es->s_last_error_ino = cpu_to_le32(inode->i_ino);
571 save_error_info(inode->i_sb, function, line);
572 path = d_path(&(file->f_path), pathname, sizeof(pathname));
580 "EXT4-fs error (device %s): %s:%d: inode #%lu: "
581 "block %llu: comm %s: path %s: %pV\n",
582 inode->i_sb->s_id, function, line, inode->i_ino,
583 block, current->comm, path, &vaf);
586 "EXT4-fs error (device %s): %s:%d: inode #%lu: "
587 "comm %s: path %s: %pV\n",
588 inode->i_sb->s_id, function, line, inode->i_ino,
589 current->comm, path, &vaf);
592 ext4_handle_error(inode->i_sb);
595 static const char *ext4_decode_error(struct super_block *sb, int errno,
602 errstr = "IO failure";
605 errstr = "Out of memory";
608 if (!sb || (EXT4_SB(sb)->s_journal &&
609 EXT4_SB(sb)->s_journal->j_flags & JBD2_ABORT))
610 errstr = "Journal has aborted";
612 errstr = "Readonly filesystem";
615 /* If the caller passed in an extra buffer for unknown
616 * errors, textualise them now. Else we just return
619 /* Check for truncated error codes... */
620 if (snprintf(nbuf, 16, "error %d", -errno) >= 0)
629 /* __ext4_std_error decodes expected errors from journaling functions
630 * automatically and invokes the appropriate error response. */
632 void __ext4_std_error(struct super_block *sb, const char *function,
633 unsigned int line, int errno)
638 /* Special case: if the error is EROFS, and we're not already
639 * inside a transaction, then there's really no point in logging
641 if (errno == -EROFS && journal_current_handle() == NULL &&
642 (sb->s_flags & MS_RDONLY))
645 errstr = ext4_decode_error(sb, errno, nbuf);
646 printk(KERN_CRIT "EXT4-fs error (device %s) in %s:%d: %s\n",
647 sb->s_id, function, line, errstr);
648 save_error_info(sb, function, line);
650 ext4_handle_error(sb);
654 * ext4_abort is a much stronger failure handler than ext4_error. The
655 * abort function may be used to deal with unrecoverable failures such
656 * as journal IO errors or ENOMEM at a critical moment in log management.
658 * We unconditionally force the filesystem into an ABORT|READONLY state,
659 * unless the error response on the fs has been set to panic in which
660 * case we take the easy way out and panic immediately.
663 void __ext4_abort(struct super_block *sb, const char *function,
664 unsigned int line, const char *fmt, ...)
668 save_error_info(sb, function, line);
670 printk(KERN_CRIT "EXT4-fs error (device %s): %s:%d: ", sb->s_id,
676 if ((sb->s_flags & MS_RDONLY) == 0) {
677 ext4_msg(sb, KERN_CRIT, "Remounting filesystem read-only");
678 sb->s_flags |= MS_RDONLY;
679 EXT4_SB(sb)->s_mount_flags |= EXT4_MF_FS_ABORTED;
680 if (EXT4_SB(sb)->s_journal)
681 jbd2_journal_abort(EXT4_SB(sb)->s_journal, -EIO);
682 save_error_info(sb, function, line);
684 if (test_opt(sb, ERRORS_PANIC))
685 panic("EXT4-fs panic from previous error\n");
688 void ext4_msg(struct super_block *sb, const char *prefix, const char *fmt, ...)
690 struct va_format vaf;
696 printk("%sEXT4-fs (%s): %pV\n", prefix, sb->s_id, &vaf);
700 void __ext4_warning(struct super_block *sb, const char *function,
701 unsigned int line, const char *fmt, ...)
703 struct va_format vaf;
709 printk(KERN_WARNING "EXT4-fs warning (device %s): %s:%d: %pV\n",
710 sb->s_id, function, line, &vaf);
714 void __ext4_grp_locked_error(const char *function, unsigned int line,
715 struct super_block *sb, ext4_group_t grp,
716 unsigned long ino, ext4_fsblk_t block,
717 const char *fmt, ...)
721 struct va_format vaf;
723 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
725 es->s_last_error_ino = cpu_to_le32(ino);
726 es->s_last_error_block = cpu_to_le64(block);
727 __save_error_info(sb, function, line);
733 printk(KERN_CRIT "EXT4-fs error (device %s): %s:%d: group %u, ",
734 sb->s_id, function, line, grp);
736 printk(KERN_CONT "inode %lu: ", ino);
738 printk(KERN_CONT "block %llu:", (unsigned long long) block);
739 printk(KERN_CONT "%pV\n", &vaf);
742 if (test_opt(sb, ERRORS_CONT)) {
743 ext4_commit_super(sb, 0);
747 ext4_unlock_group(sb, grp);
748 ext4_handle_error(sb);
750 * We only get here in the ERRORS_RO case; relocking the group
751 * may be dangerous, but nothing bad will happen since the
752 * filesystem will have already been marked read/only and the
753 * journal has been aborted. We return 1 as a hint to callers
754 * who might what to use the return value from
755 * ext4_grp_locked_error() to distinguish between the
756 * ERRORS_CONT and ERRORS_RO case, and perhaps return more
757 * aggressively from the ext4 function in question, with a
758 * more appropriate error code.
760 ext4_lock_group(sb, grp);
764 void ext4_update_dynamic_rev(struct super_block *sb)
766 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
768 if (le32_to_cpu(es->s_rev_level) > EXT4_GOOD_OLD_REV)
772 "updating to rev %d because of new feature flag, "
773 "running e2fsck is recommended",
776 es->s_first_ino = cpu_to_le32(EXT4_GOOD_OLD_FIRST_INO);
777 es->s_inode_size = cpu_to_le16(EXT4_GOOD_OLD_INODE_SIZE);
778 es->s_rev_level = cpu_to_le32(EXT4_DYNAMIC_REV);
779 /* leave es->s_feature_*compat flags alone */
780 /* es->s_uuid will be set by e2fsck if empty */
783 * The rest of the superblock fields should be zero, and if not it
784 * means they are likely already in use, so leave them alone. We
785 * can leave it up to e2fsck to clean up any inconsistencies there.
790 * Open the external journal device
792 static struct block_device *ext4_blkdev_get(dev_t dev, struct super_block *sb)
794 struct block_device *bdev;
795 char b[BDEVNAME_SIZE];
797 bdev = blkdev_get_by_dev(dev, FMODE_READ|FMODE_WRITE|FMODE_EXCL, sb);
803 ext4_msg(sb, KERN_ERR, "failed to open journal device %s: %ld",
804 __bdevname(dev, b), PTR_ERR(bdev));
809 * Release the journal device
811 static int ext4_blkdev_put(struct block_device *bdev)
813 return blkdev_put(bdev, FMODE_READ|FMODE_WRITE|FMODE_EXCL);
816 static int ext4_blkdev_remove(struct ext4_sb_info *sbi)
818 struct block_device *bdev;
821 bdev = sbi->journal_bdev;
823 ret = ext4_blkdev_put(bdev);
824 sbi->journal_bdev = NULL;
829 static inline struct inode *orphan_list_entry(struct list_head *l)
831 return &list_entry(l, struct ext4_inode_info, i_orphan)->vfs_inode;
834 static void dump_orphan_list(struct super_block *sb, struct ext4_sb_info *sbi)
838 ext4_msg(sb, KERN_ERR, "sb orphan head is %d",
839 le32_to_cpu(sbi->s_es->s_last_orphan));
841 printk(KERN_ERR "sb_info orphan list:\n");
842 list_for_each(l, &sbi->s_orphan) {
843 struct inode *inode = orphan_list_entry(l);
845 "inode %s:%lu at %p: mode %o, nlink %d, next %d\n",
846 inode->i_sb->s_id, inode->i_ino, inode,
847 inode->i_mode, inode->i_nlink,
852 static void ext4_put_super(struct super_block *sb)
854 struct ext4_sb_info *sbi = EXT4_SB(sb);
855 struct ext4_super_block *es = sbi->s_es;
858 ext4_unregister_li_request(sb);
859 dquot_disable(sb, -1, DQUOT_USAGE_ENABLED | DQUOT_LIMITS_ENABLED);
861 flush_workqueue(sbi->dio_unwritten_wq);
862 destroy_workqueue(sbi->dio_unwritten_wq);
864 if (sbi->s_journal) {
865 err = jbd2_journal_destroy(sbi->s_journal);
866 sbi->s_journal = NULL;
868 ext4_abort(sb, "Couldn't clean up the journal");
871 del_timer(&sbi->s_err_report);
872 ext4_release_system_zone(sb);
874 ext4_ext_release(sb);
875 ext4_xattr_put_super(sb);
877 if (!(sb->s_flags & MS_RDONLY)) {
878 EXT4_CLEAR_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
879 es->s_state = cpu_to_le16(sbi->s_mount_state);
881 if (!(sb->s_flags & MS_RDONLY))
882 ext4_commit_super(sb, 1);
885 remove_proc_entry("options", sbi->s_proc);
886 remove_proc_entry(sb->s_id, ext4_proc_root);
888 kobject_del(&sbi->s_kobj);
890 for (i = 0; i < sbi->s_gdb_count; i++)
891 brelse(sbi->s_group_desc[i]);
892 ext4_kvfree(sbi->s_group_desc);
893 ext4_kvfree(sbi->s_flex_groups);
894 percpu_counter_destroy(&sbi->s_freeclusters_counter);
895 percpu_counter_destroy(&sbi->s_freeinodes_counter);
896 percpu_counter_destroy(&sbi->s_dirs_counter);
897 percpu_counter_destroy(&sbi->s_dirtyclusters_counter);
900 for (i = 0; i < MAXQUOTAS; i++)
901 kfree(sbi->s_qf_names[i]);
904 /* Debugging code just in case the in-memory inode orphan list
905 * isn't empty. The on-disk one can be non-empty if we've
906 * detected an error and taken the fs readonly, but the
907 * in-memory list had better be clean by this point. */
908 if (!list_empty(&sbi->s_orphan))
909 dump_orphan_list(sb, sbi);
910 J_ASSERT(list_empty(&sbi->s_orphan));
912 invalidate_bdev(sb->s_bdev);
913 if (sbi->journal_bdev && sbi->journal_bdev != sb->s_bdev) {
915 * Invalidate the journal device's buffers. We don't want them
916 * floating about in memory - the physical journal device may
917 * hotswapped, and it breaks the `ro-after' testing code.
919 sync_blockdev(sbi->journal_bdev);
920 invalidate_bdev(sbi->journal_bdev);
921 ext4_blkdev_remove(sbi);
924 kthread_stop(sbi->s_mmp_tsk);
925 sb->s_fs_info = NULL;
927 * Now that we are completely done shutting down the
928 * superblock, we need to actually destroy the kobject.
930 kobject_put(&sbi->s_kobj);
931 wait_for_completion(&sbi->s_kobj_unregister);
932 if (sbi->s_chksum_driver)
933 crypto_free_shash(sbi->s_chksum_driver);
934 kfree(sbi->s_blockgroup_lock);
938 static struct kmem_cache *ext4_inode_cachep;
941 * Called inside transaction, so use GFP_NOFS
943 static struct inode *ext4_alloc_inode(struct super_block *sb)
945 struct ext4_inode_info *ei;
947 ei = kmem_cache_alloc(ext4_inode_cachep, GFP_NOFS);
951 ei->vfs_inode.i_version = 1;
952 ei->vfs_inode.i_data.writeback_index = 0;
953 memset(&ei->i_cached_extent, 0, sizeof(struct ext4_ext_cache));
954 INIT_LIST_HEAD(&ei->i_prealloc_list);
955 spin_lock_init(&ei->i_prealloc_lock);
956 ei->i_reserved_data_blocks = 0;
957 ei->i_reserved_meta_blocks = 0;
958 ei->i_allocated_meta_blocks = 0;
959 ei->i_da_metadata_calc_len = 0;
960 ei->i_da_metadata_calc_last_lblock = 0;
961 spin_lock_init(&(ei->i_block_reservation_lock));
963 ei->i_reserved_quota = 0;
966 INIT_LIST_HEAD(&ei->i_completed_io_list);
967 spin_lock_init(&ei->i_completed_io_lock);
968 ei->cur_aio_dio = NULL;
970 ei->i_datasync_tid = 0;
971 atomic_set(&ei->i_ioend_count, 0);
972 atomic_set(&ei->i_aiodio_unwritten, 0);
974 return &ei->vfs_inode;
977 static int ext4_drop_inode(struct inode *inode)
979 int drop = generic_drop_inode(inode);
981 trace_ext4_drop_inode(inode, drop);
985 static void ext4_i_callback(struct rcu_head *head)
987 struct inode *inode = container_of(head, struct inode, i_rcu);
988 kmem_cache_free(ext4_inode_cachep, EXT4_I(inode));
991 static void ext4_destroy_inode(struct inode *inode)
993 if (!list_empty(&(EXT4_I(inode)->i_orphan))) {
994 ext4_msg(inode->i_sb, KERN_ERR,
995 "Inode %lu (%p): orphan list check failed!",
996 inode->i_ino, EXT4_I(inode));
997 print_hex_dump(KERN_INFO, "", DUMP_PREFIX_ADDRESS, 16, 4,
998 EXT4_I(inode), sizeof(struct ext4_inode_info),
1002 call_rcu(&inode->i_rcu, ext4_i_callback);
1005 static void init_once(void *foo)
1007 struct ext4_inode_info *ei = (struct ext4_inode_info *) foo;
1009 INIT_LIST_HEAD(&ei->i_orphan);
1010 #ifdef CONFIG_EXT4_FS_XATTR
1011 init_rwsem(&ei->xattr_sem);
1013 init_rwsem(&ei->i_data_sem);
1014 inode_init_once(&ei->vfs_inode);
1017 static int init_inodecache(void)
1019 ext4_inode_cachep = kmem_cache_create("ext4_inode_cache",
1020 sizeof(struct ext4_inode_info),
1021 0, (SLAB_RECLAIM_ACCOUNT|
1024 if (ext4_inode_cachep == NULL)
1029 static void destroy_inodecache(void)
1031 kmem_cache_destroy(ext4_inode_cachep);
1034 void ext4_clear_inode(struct inode *inode)
1036 invalidate_inode_buffers(inode);
1039 ext4_discard_preallocations(inode);
1040 if (EXT4_I(inode)->jinode) {
1041 jbd2_journal_release_jbd_inode(EXT4_JOURNAL(inode),
1042 EXT4_I(inode)->jinode);
1043 jbd2_free_inode(EXT4_I(inode)->jinode);
1044 EXT4_I(inode)->jinode = NULL;
1048 static struct inode *ext4_nfs_get_inode(struct super_block *sb,
1049 u64 ino, u32 generation)
1051 struct inode *inode;
1053 if (ino < EXT4_FIRST_INO(sb) && ino != EXT4_ROOT_INO)
1054 return ERR_PTR(-ESTALE);
1055 if (ino > le32_to_cpu(EXT4_SB(sb)->s_es->s_inodes_count))
1056 return ERR_PTR(-ESTALE);
1058 /* iget isn't really right if the inode is currently unallocated!!
1060 * ext4_read_inode will return a bad_inode if the inode had been
1061 * deleted, so we should be safe.
1063 * Currently we don't know the generation for parent directory, so
1064 * a generation of 0 means "accept any"
1066 inode = ext4_iget(sb, ino);
1068 return ERR_CAST(inode);
1069 if (generation && inode->i_generation != generation) {
1071 return ERR_PTR(-ESTALE);
1077 static struct dentry *ext4_fh_to_dentry(struct super_block *sb, struct fid *fid,
1078 int fh_len, int fh_type)
1080 return generic_fh_to_dentry(sb, fid, fh_len, fh_type,
1081 ext4_nfs_get_inode);
1084 static struct dentry *ext4_fh_to_parent(struct super_block *sb, struct fid *fid,
1085 int fh_len, int fh_type)
1087 return generic_fh_to_parent(sb, fid, fh_len, fh_type,
1088 ext4_nfs_get_inode);
1092 * Try to release metadata pages (indirect blocks, directories) which are
1093 * mapped via the block device. Since these pages could have journal heads
1094 * which would prevent try_to_free_buffers() from freeing them, we must use
1095 * jbd2 layer's try_to_free_buffers() function to release them.
1097 static int bdev_try_to_free_page(struct super_block *sb, struct page *page,
1100 journal_t *journal = EXT4_SB(sb)->s_journal;
1102 WARN_ON(PageChecked(page));
1103 if (!page_has_buffers(page))
1106 return jbd2_journal_try_to_free_buffers(journal, page,
1107 wait & ~__GFP_WAIT);
1108 return try_to_free_buffers(page);
1112 #define QTYPE2NAME(t) ((t) == USRQUOTA ? "user" : "group")
1113 #define QTYPE2MOPT(on, t) ((t) == USRQUOTA?((on)##USRJQUOTA):((on)##GRPJQUOTA))
1115 static int ext4_write_dquot(struct dquot *dquot);
1116 static int ext4_acquire_dquot(struct dquot *dquot);
1117 static int ext4_release_dquot(struct dquot *dquot);
1118 static int ext4_mark_dquot_dirty(struct dquot *dquot);
1119 static int ext4_write_info(struct super_block *sb, int type);
1120 static int ext4_quota_on(struct super_block *sb, int type, int format_id,
1122 static int ext4_quota_on_sysfile(struct super_block *sb, int type,
1124 static int ext4_quota_off(struct super_block *sb, int type);
1125 static int ext4_quota_off_sysfile(struct super_block *sb, int type);
1126 static int ext4_quota_on_mount(struct super_block *sb, int type);
1127 static ssize_t ext4_quota_read(struct super_block *sb, int type, char *data,
1128 size_t len, loff_t off);
1129 static ssize_t ext4_quota_write(struct super_block *sb, int type,
1130 const char *data, size_t len, loff_t off);
1131 static int ext4_quota_enable(struct super_block *sb, int type, int format_id,
1132 unsigned int flags);
1133 static int ext4_enable_quotas(struct super_block *sb);
1135 static const struct dquot_operations ext4_quota_operations = {
1136 .get_reserved_space = ext4_get_reserved_space,
1137 .write_dquot = ext4_write_dquot,
1138 .acquire_dquot = ext4_acquire_dquot,
1139 .release_dquot = ext4_release_dquot,
1140 .mark_dirty = ext4_mark_dquot_dirty,
1141 .write_info = ext4_write_info,
1142 .alloc_dquot = dquot_alloc,
1143 .destroy_dquot = dquot_destroy,
1146 static const struct quotactl_ops ext4_qctl_operations = {
1147 .quota_on = ext4_quota_on,
1148 .quota_off = ext4_quota_off,
1149 .quota_sync = dquot_quota_sync,
1150 .get_info = dquot_get_dqinfo,
1151 .set_info = dquot_set_dqinfo,
1152 .get_dqblk = dquot_get_dqblk,
1153 .set_dqblk = dquot_set_dqblk
1156 static const struct quotactl_ops ext4_qctl_sysfile_operations = {
1157 .quota_on_meta = ext4_quota_on_sysfile,
1158 .quota_off = ext4_quota_off_sysfile,
1159 .quota_sync = dquot_quota_sync,
1160 .get_info = dquot_get_dqinfo,
1161 .set_info = dquot_set_dqinfo,
1162 .get_dqblk = dquot_get_dqblk,
1163 .set_dqblk = dquot_set_dqblk
1167 static const struct super_operations ext4_sops = {
1168 .alloc_inode = ext4_alloc_inode,
1169 .destroy_inode = ext4_destroy_inode,
1170 .write_inode = ext4_write_inode,
1171 .dirty_inode = ext4_dirty_inode,
1172 .drop_inode = ext4_drop_inode,
1173 .evict_inode = ext4_evict_inode,
1174 .put_super = ext4_put_super,
1175 .sync_fs = ext4_sync_fs,
1176 .freeze_fs = ext4_freeze,
1177 .unfreeze_fs = ext4_unfreeze,
1178 .statfs = ext4_statfs,
1179 .remount_fs = ext4_remount,
1180 .show_options = ext4_show_options,
1182 .quota_read = ext4_quota_read,
1183 .quota_write = ext4_quota_write,
1185 .bdev_try_to_free_page = bdev_try_to_free_page,
1188 static const struct super_operations ext4_nojournal_sops = {
1189 .alloc_inode = ext4_alloc_inode,
1190 .destroy_inode = ext4_destroy_inode,
1191 .write_inode = ext4_write_inode,
1192 .dirty_inode = ext4_dirty_inode,
1193 .drop_inode = ext4_drop_inode,
1194 .evict_inode = ext4_evict_inode,
1195 .put_super = ext4_put_super,
1196 .statfs = ext4_statfs,
1197 .remount_fs = ext4_remount,
1198 .show_options = ext4_show_options,
1200 .quota_read = ext4_quota_read,
1201 .quota_write = ext4_quota_write,
1203 .bdev_try_to_free_page = bdev_try_to_free_page,
1206 static const struct export_operations ext4_export_ops = {
1207 .fh_to_dentry = ext4_fh_to_dentry,
1208 .fh_to_parent = ext4_fh_to_parent,
1209 .get_parent = ext4_get_parent,
1213 Opt_bsd_df, Opt_minix_df, Opt_grpid, Opt_nogrpid,
1214 Opt_resgid, Opt_resuid, Opt_sb, Opt_err_cont, Opt_err_panic, Opt_err_ro,
1215 Opt_nouid32, Opt_debug, Opt_removed,
1216 Opt_user_xattr, Opt_nouser_xattr, Opt_acl, Opt_noacl,
1217 Opt_auto_da_alloc, Opt_noauto_da_alloc, Opt_noload,
1218 Opt_commit, Opt_min_batch_time, Opt_max_batch_time,
1219 Opt_journal_dev, Opt_journal_checksum, Opt_journal_async_commit,
1220 Opt_abort, Opt_data_journal, Opt_data_ordered, Opt_data_writeback,
1221 Opt_data_err_abort, Opt_data_err_ignore,
1222 Opt_usrjquota, Opt_grpjquota, Opt_offusrjquota, Opt_offgrpjquota,
1223 Opt_jqfmt_vfsold, Opt_jqfmt_vfsv0, Opt_jqfmt_vfsv1, Opt_quota,
1224 Opt_noquota, Opt_barrier, Opt_nobarrier, Opt_err,
1225 Opt_usrquota, Opt_grpquota, Opt_i_version,
1226 Opt_stripe, Opt_delalloc, Opt_nodelalloc, Opt_mblk_io_submit,
1227 Opt_nomblk_io_submit, Opt_block_validity, Opt_noblock_validity,
1228 Opt_inode_readahead_blks, Opt_journal_ioprio,
1229 Opt_dioread_nolock, Opt_dioread_lock,
1230 Opt_discard, Opt_nodiscard, Opt_init_itable, Opt_noinit_itable,
1231 Opt_max_dir_size_kb,
1234 static const match_table_t tokens = {
1235 {Opt_bsd_df, "bsddf"},
1236 {Opt_minix_df, "minixdf"},
1237 {Opt_grpid, "grpid"},
1238 {Opt_grpid, "bsdgroups"},
1239 {Opt_nogrpid, "nogrpid"},
1240 {Opt_nogrpid, "sysvgroups"},
1241 {Opt_resgid, "resgid=%u"},
1242 {Opt_resuid, "resuid=%u"},
1244 {Opt_err_cont, "errors=continue"},
1245 {Opt_err_panic, "errors=panic"},
1246 {Opt_err_ro, "errors=remount-ro"},
1247 {Opt_nouid32, "nouid32"},
1248 {Opt_debug, "debug"},
1249 {Opt_removed, "oldalloc"},
1250 {Opt_removed, "orlov"},
1251 {Opt_user_xattr, "user_xattr"},
1252 {Opt_nouser_xattr, "nouser_xattr"},
1254 {Opt_noacl, "noacl"},
1255 {Opt_noload, "norecovery"},
1256 {Opt_noload, "noload"},
1257 {Opt_removed, "nobh"},
1258 {Opt_removed, "bh"},
1259 {Opt_commit, "commit=%u"},
1260 {Opt_min_batch_time, "min_batch_time=%u"},
1261 {Opt_max_batch_time, "max_batch_time=%u"},
1262 {Opt_journal_dev, "journal_dev=%u"},
1263 {Opt_journal_checksum, "journal_checksum"},
1264 {Opt_journal_async_commit, "journal_async_commit"},
1265 {Opt_abort, "abort"},
1266 {Opt_data_journal, "data=journal"},
1267 {Opt_data_ordered, "data=ordered"},
1268 {Opt_data_writeback, "data=writeback"},
1269 {Opt_data_err_abort, "data_err=abort"},
1270 {Opt_data_err_ignore, "data_err=ignore"},
1271 {Opt_offusrjquota, "usrjquota="},
1272 {Opt_usrjquota, "usrjquota=%s"},
1273 {Opt_offgrpjquota, "grpjquota="},
1274 {Opt_grpjquota, "grpjquota=%s"},
1275 {Opt_jqfmt_vfsold, "jqfmt=vfsold"},
1276 {Opt_jqfmt_vfsv0, "jqfmt=vfsv0"},
1277 {Opt_jqfmt_vfsv1, "jqfmt=vfsv1"},
1278 {Opt_grpquota, "grpquota"},
1279 {Opt_noquota, "noquota"},
1280 {Opt_quota, "quota"},
1281 {Opt_usrquota, "usrquota"},
1282 {Opt_barrier, "barrier=%u"},
1283 {Opt_barrier, "barrier"},
1284 {Opt_nobarrier, "nobarrier"},
1285 {Opt_i_version, "i_version"},
1286 {Opt_stripe, "stripe=%u"},
1287 {Opt_delalloc, "delalloc"},
1288 {Opt_nodelalloc, "nodelalloc"},
1289 {Opt_mblk_io_submit, "mblk_io_submit"},
1290 {Opt_nomblk_io_submit, "nomblk_io_submit"},
1291 {Opt_block_validity, "block_validity"},
1292 {Opt_noblock_validity, "noblock_validity"},
1293 {Opt_inode_readahead_blks, "inode_readahead_blks=%u"},
1294 {Opt_journal_ioprio, "journal_ioprio=%u"},
1295 {Opt_auto_da_alloc, "auto_da_alloc=%u"},
1296 {Opt_auto_da_alloc, "auto_da_alloc"},
1297 {Opt_noauto_da_alloc, "noauto_da_alloc"},
1298 {Opt_dioread_nolock, "dioread_nolock"},
1299 {Opt_dioread_lock, "dioread_lock"},
1300 {Opt_discard, "discard"},
1301 {Opt_nodiscard, "nodiscard"},
1302 {Opt_init_itable, "init_itable=%u"},
1303 {Opt_init_itable, "init_itable"},
1304 {Opt_noinit_itable, "noinit_itable"},
1305 {Opt_max_dir_size_kb, "max_dir_size_kb=%u"},
1306 {Opt_removed, "check=none"}, /* mount option from ext2/3 */
1307 {Opt_removed, "nocheck"}, /* mount option from ext2/3 */
1308 {Opt_removed, "reservation"}, /* mount option from ext2/3 */
1309 {Opt_removed, "noreservation"}, /* mount option from ext2/3 */
1310 {Opt_removed, "journal=%u"}, /* mount option from ext2/3 */
1314 static ext4_fsblk_t get_sb_block(void **data)
1316 ext4_fsblk_t sb_block;
1317 char *options = (char *) *data;
1319 if (!options || strncmp(options, "sb=", 3) != 0)
1320 return 1; /* Default location */
1323 /* TODO: use simple_strtoll with >32bit ext4 */
1324 sb_block = simple_strtoul(options, &options, 0);
1325 if (*options && *options != ',') {
1326 printk(KERN_ERR "EXT4-fs: Invalid sb specification: %s\n",
1330 if (*options == ',')
1332 *data = (void *) options;
1337 #define DEFAULT_JOURNAL_IOPRIO (IOPRIO_PRIO_VALUE(IOPRIO_CLASS_BE, 3))
1338 static char deprecated_msg[] = "Mount option \"%s\" will be removed by %s\n"
1339 "Contact linux-ext4@vger.kernel.org if you think we should keep it.\n";
1342 static int set_qf_name(struct super_block *sb, int qtype, substring_t *args)
1344 struct ext4_sb_info *sbi = EXT4_SB(sb);
1347 if (sb_any_quota_loaded(sb) &&
1348 !sbi->s_qf_names[qtype]) {
1349 ext4_msg(sb, KERN_ERR,
1350 "Cannot change journaled "
1351 "quota options when quota turned on");
1354 qname = match_strdup(args);
1356 ext4_msg(sb, KERN_ERR,
1357 "Not enough memory for storing quotafile name");
1360 if (sbi->s_qf_names[qtype] &&
1361 strcmp(sbi->s_qf_names[qtype], qname)) {
1362 ext4_msg(sb, KERN_ERR,
1363 "%s quota file already specified", QTYPE2NAME(qtype));
1367 sbi->s_qf_names[qtype] = qname;
1368 if (strchr(sbi->s_qf_names[qtype], '/')) {
1369 ext4_msg(sb, KERN_ERR,
1370 "quotafile must be on filesystem root");
1371 kfree(sbi->s_qf_names[qtype]);
1372 sbi->s_qf_names[qtype] = NULL;
1379 static int clear_qf_name(struct super_block *sb, int qtype)
1382 struct ext4_sb_info *sbi = EXT4_SB(sb);
1384 if (sb_any_quota_loaded(sb) &&
1385 sbi->s_qf_names[qtype]) {
1386 ext4_msg(sb, KERN_ERR, "Cannot change journaled quota options"
1387 " when quota turned on");
1391 * The space will be released later when all options are confirmed
1394 sbi->s_qf_names[qtype] = NULL;
1399 #define MOPT_SET 0x0001
1400 #define MOPT_CLEAR 0x0002
1401 #define MOPT_NOSUPPORT 0x0004
1402 #define MOPT_EXPLICIT 0x0008
1403 #define MOPT_CLEAR_ERR 0x0010
1404 #define MOPT_GTE0 0x0020
1407 #define MOPT_QFMT 0x0040
1409 #define MOPT_Q MOPT_NOSUPPORT
1410 #define MOPT_QFMT MOPT_NOSUPPORT
1412 #define MOPT_DATAJ 0x0080
1414 static const struct mount_opts {
1418 } ext4_mount_opts[] = {
1419 {Opt_minix_df, EXT4_MOUNT_MINIX_DF, MOPT_SET},
1420 {Opt_bsd_df, EXT4_MOUNT_MINIX_DF, MOPT_CLEAR},
1421 {Opt_grpid, EXT4_MOUNT_GRPID, MOPT_SET},
1422 {Opt_nogrpid, EXT4_MOUNT_GRPID, MOPT_CLEAR},
1423 {Opt_mblk_io_submit, EXT4_MOUNT_MBLK_IO_SUBMIT, MOPT_SET},
1424 {Opt_nomblk_io_submit, EXT4_MOUNT_MBLK_IO_SUBMIT, MOPT_CLEAR},
1425 {Opt_block_validity, EXT4_MOUNT_BLOCK_VALIDITY, MOPT_SET},
1426 {Opt_noblock_validity, EXT4_MOUNT_BLOCK_VALIDITY, MOPT_CLEAR},
1427 {Opt_dioread_nolock, EXT4_MOUNT_DIOREAD_NOLOCK, MOPT_SET},
1428 {Opt_dioread_lock, EXT4_MOUNT_DIOREAD_NOLOCK, MOPT_CLEAR},
1429 {Opt_discard, EXT4_MOUNT_DISCARD, MOPT_SET},
1430 {Opt_nodiscard, EXT4_MOUNT_DISCARD, MOPT_CLEAR},
1431 {Opt_delalloc, EXT4_MOUNT_DELALLOC, MOPT_SET | MOPT_EXPLICIT},
1432 {Opt_nodelalloc, EXT4_MOUNT_DELALLOC, MOPT_CLEAR | MOPT_EXPLICIT},
1433 {Opt_journal_checksum, EXT4_MOUNT_JOURNAL_CHECKSUM, MOPT_SET},
1434 {Opt_journal_async_commit, (EXT4_MOUNT_JOURNAL_ASYNC_COMMIT |
1435 EXT4_MOUNT_JOURNAL_CHECKSUM), MOPT_SET},
1436 {Opt_noload, EXT4_MOUNT_NOLOAD, MOPT_SET},
1437 {Opt_err_panic, EXT4_MOUNT_ERRORS_PANIC, MOPT_SET | MOPT_CLEAR_ERR},
1438 {Opt_err_ro, EXT4_MOUNT_ERRORS_RO, MOPT_SET | MOPT_CLEAR_ERR},
1439 {Opt_err_cont, EXT4_MOUNT_ERRORS_CONT, MOPT_SET | MOPT_CLEAR_ERR},
1440 {Opt_data_err_abort, EXT4_MOUNT_DATA_ERR_ABORT, MOPT_SET},
1441 {Opt_data_err_ignore, EXT4_MOUNT_DATA_ERR_ABORT, MOPT_CLEAR},
1442 {Opt_barrier, EXT4_MOUNT_BARRIER, MOPT_SET},
1443 {Opt_nobarrier, EXT4_MOUNT_BARRIER, MOPT_CLEAR},
1444 {Opt_noauto_da_alloc, EXT4_MOUNT_NO_AUTO_DA_ALLOC, MOPT_SET},
1445 {Opt_auto_da_alloc, EXT4_MOUNT_NO_AUTO_DA_ALLOC, MOPT_CLEAR},
1446 {Opt_noinit_itable, EXT4_MOUNT_INIT_INODE_TABLE, MOPT_CLEAR},
1447 {Opt_commit, 0, MOPT_GTE0},
1448 {Opt_max_batch_time, 0, MOPT_GTE0},
1449 {Opt_min_batch_time, 0, MOPT_GTE0},
1450 {Opt_inode_readahead_blks, 0, MOPT_GTE0},
1451 {Opt_init_itable, 0, MOPT_GTE0},
1452 {Opt_stripe, 0, MOPT_GTE0},
1453 {Opt_data_journal, EXT4_MOUNT_JOURNAL_DATA, MOPT_DATAJ},
1454 {Opt_data_ordered, EXT4_MOUNT_ORDERED_DATA, MOPT_DATAJ},
1455 {Opt_data_writeback, EXT4_MOUNT_WRITEBACK_DATA, MOPT_DATAJ},
1456 #ifdef CONFIG_EXT4_FS_XATTR
1457 {Opt_user_xattr, EXT4_MOUNT_XATTR_USER, MOPT_SET},
1458 {Opt_nouser_xattr, EXT4_MOUNT_XATTR_USER, MOPT_CLEAR},
1460 {Opt_user_xattr, 0, MOPT_NOSUPPORT},
1461 {Opt_nouser_xattr, 0, MOPT_NOSUPPORT},
1463 #ifdef CONFIG_EXT4_FS_POSIX_ACL
1464 {Opt_acl, EXT4_MOUNT_POSIX_ACL, MOPT_SET},
1465 {Opt_noacl, EXT4_MOUNT_POSIX_ACL, MOPT_CLEAR},
1467 {Opt_acl, 0, MOPT_NOSUPPORT},
1468 {Opt_noacl, 0, MOPT_NOSUPPORT},
1470 {Opt_nouid32, EXT4_MOUNT_NO_UID32, MOPT_SET},
1471 {Opt_debug, EXT4_MOUNT_DEBUG, MOPT_SET},
1472 {Opt_quota, EXT4_MOUNT_QUOTA | EXT4_MOUNT_USRQUOTA, MOPT_SET | MOPT_Q},
1473 {Opt_usrquota, EXT4_MOUNT_QUOTA | EXT4_MOUNT_USRQUOTA,
1475 {Opt_grpquota, EXT4_MOUNT_QUOTA | EXT4_MOUNT_GRPQUOTA,
1477 {Opt_noquota, (EXT4_MOUNT_QUOTA | EXT4_MOUNT_USRQUOTA |
1478 EXT4_MOUNT_GRPQUOTA), MOPT_CLEAR | MOPT_Q},
1479 {Opt_usrjquota, 0, MOPT_Q},
1480 {Opt_grpjquota, 0, MOPT_Q},
1481 {Opt_offusrjquota, 0, MOPT_Q},
1482 {Opt_offgrpjquota, 0, MOPT_Q},
1483 {Opt_jqfmt_vfsold, QFMT_VFS_OLD, MOPT_QFMT},
1484 {Opt_jqfmt_vfsv0, QFMT_VFS_V0, MOPT_QFMT},
1485 {Opt_jqfmt_vfsv1, QFMT_VFS_V1, MOPT_QFMT},
1486 {Opt_max_dir_size_kb, 0, MOPT_GTE0},
1490 static int handle_mount_opt(struct super_block *sb, char *opt, int token,
1491 substring_t *args, unsigned long *journal_devnum,
1492 unsigned int *journal_ioprio, int is_remount)
1494 struct ext4_sb_info *sbi = EXT4_SB(sb);
1495 const struct mount_opts *m;
1501 if (token == Opt_usrjquota)
1502 return set_qf_name(sb, USRQUOTA, &args[0]);
1503 else if (token == Opt_grpjquota)
1504 return set_qf_name(sb, GRPQUOTA, &args[0]);
1505 else if (token == Opt_offusrjquota)
1506 return clear_qf_name(sb, USRQUOTA);
1507 else if (token == Opt_offgrpjquota)
1508 return clear_qf_name(sb, GRPQUOTA);
1510 if (args->from && match_int(args, &arg))
1514 case Opt_nouser_xattr:
1515 ext4_msg(sb, KERN_WARNING, deprecated_msg, opt, "3.5");
1518 return 1; /* handled by get_sb_block() */
1520 ext4_msg(sb, KERN_WARNING,
1521 "Ignoring removed %s option", opt);
1524 uid = make_kuid(current_user_ns(), arg);
1525 if (!uid_valid(uid)) {
1526 ext4_msg(sb, KERN_ERR, "Invalid uid value %d", arg);
1529 sbi->s_resuid = uid;
1532 gid = make_kgid(current_user_ns(), arg);
1533 if (!gid_valid(gid)) {
1534 ext4_msg(sb, KERN_ERR, "Invalid gid value %d", arg);
1537 sbi->s_resgid = gid;
1540 sbi->s_mount_flags |= EXT4_MF_FS_ABORTED;
1543 sb->s_flags |= MS_I_VERSION;
1545 case Opt_journal_dev:
1547 ext4_msg(sb, KERN_ERR,
1548 "Cannot specify journal on remount");
1551 *journal_devnum = arg;
1553 case Opt_journal_ioprio:
1554 if (arg < 0 || arg > 7)
1556 *journal_ioprio = IOPRIO_PRIO_VALUE(IOPRIO_CLASS_BE, arg);
1560 for (m = ext4_mount_opts; m->token != Opt_err; m++) {
1561 if (token != m->token)
1563 if (args->from && (m->flags & MOPT_GTE0) && (arg < 0))
1565 if (m->flags & MOPT_EXPLICIT)
1566 set_opt2(sb, EXPLICIT_DELALLOC);
1567 if (m->flags & MOPT_CLEAR_ERR)
1568 clear_opt(sb, ERRORS_MASK);
1569 if (token == Opt_noquota && sb_any_quota_loaded(sb)) {
1570 ext4_msg(sb, KERN_ERR, "Cannot change quota "
1571 "options when quota turned on");
1575 if (m->flags & MOPT_NOSUPPORT) {
1576 ext4_msg(sb, KERN_ERR, "%s option not supported", opt);
1577 } else if (token == Opt_commit) {
1579 arg = JBD2_DEFAULT_MAX_COMMIT_AGE;
1580 sbi->s_commit_interval = HZ * arg;
1581 } else if (token == Opt_max_batch_time) {
1583 arg = EXT4_DEF_MAX_BATCH_TIME;
1584 sbi->s_max_batch_time = arg;
1585 } else if (token == Opt_min_batch_time) {
1586 sbi->s_min_batch_time = arg;
1587 } else if (token == Opt_inode_readahead_blks) {
1588 if (arg > (1 << 30))
1590 if (arg && !is_power_of_2(arg)) {
1591 ext4_msg(sb, KERN_ERR,
1592 "EXT4-fs: inode_readahead_blks"
1593 " must be a power of 2");
1596 sbi->s_inode_readahead_blks = arg;
1597 } else if (token == Opt_init_itable) {
1598 set_opt(sb, INIT_INODE_TABLE);
1600 arg = EXT4_DEF_LI_WAIT_MULT;
1601 sbi->s_li_wait_mult = arg;
1602 } else if (token == Opt_max_dir_size_kb) {
1603 sbi->s_max_dir_size_kb = arg;
1604 } else if (token == Opt_stripe) {
1605 sbi->s_stripe = arg;
1606 } else if (m->flags & MOPT_DATAJ) {
1608 if (!sbi->s_journal)
1609 ext4_msg(sb, KERN_WARNING, "Remounting file system with no journal so ignoring journalled data option");
1610 else if (test_opt(sb, DATA_FLAGS) !=
1612 ext4_msg(sb, KERN_ERR,
1613 "Cannot change data mode on remount");
1617 clear_opt(sb, DATA_FLAGS);
1618 sbi->s_mount_opt |= m->mount_opt;
1621 } else if (m->flags & MOPT_QFMT) {
1622 if (sb_any_quota_loaded(sb) &&
1623 sbi->s_jquota_fmt != m->mount_opt) {
1624 ext4_msg(sb, KERN_ERR, "Cannot "
1625 "change journaled quota options "
1626 "when quota turned on");
1629 sbi->s_jquota_fmt = m->mount_opt;
1634 if (m->flags & MOPT_CLEAR)
1636 else if (unlikely(!(m->flags & MOPT_SET))) {
1637 ext4_msg(sb, KERN_WARNING,
1638 "buggy handling of option %s", opt);
1643 sbi->s_mount_opt |= m->mount_opt;
1645 sbi->s_mount_opt &= ~m->mount_opt;
1649 ext4_msg(sb, KERN_ERR, "Unrecognized mount option \"%s\" "
1650 "or missing value", opt);
1654 static int parse_options(char *options, struct super_block *sb,
1655 unsigned long *journal_devnum,
1656 unsigned int *journal_ioprio,
1660 struct ext4_sb_info *sbi = EXT4_SB(sb);
1663 substring_t args[MAX_OPT_ARGS];
1669 while ((p = strsep(&options, ",")) != NULL) {
1673 * Initialize args struct so we know whether arg was
1674 * found; some options take optional arguments.
1676 args[0].to = args[0].from = NULL;
1677 token = match_token(p, tokens, args);
1678 if (handle_mount_opt(sb, p, token, args, journal_devnum,
1679 journal_ioprio, is_remount) < 0)
1683 if (sbi->s_qf_names[USRQUOTA] || sbi->s_qf_names[GRPQUOTA]) {
1684 if (test_opt(sb, USRQUOTA) && sbi->s_qf_names[USRQUOTA])
1685 clear_opt(sb, USRQUOTA);
1687 if (test_opt(sb, GRPQUOTA) && sbi->s_qf_names[GRPQUOTA])
1688 clear_opt(sb, GRPQUOTA);
1690 if (test_opt(sb, GRPQUOTA) || test_opt(sb, USRQUOTA)) {
1691 ext4_msg(sb, KERN_ERR, "old and new quota "
1696 if (!sbi->s_jquota_fmt) {
1697 ext4_msg(sb, KERN_ERR, "journaled quota format "
1702 if (sbi->s_jquota_fmt) {
1703 ext4_msg(sb, KERN_ERR, "journaled quota format "
1704 "specified with no journaling "
1713 static inline void ext4_show_quota_options(struct seq_file *seq,
1714 struct super_block *sb)
1716 #if defined(CONFIG_QUOTA)
1717 struct ext4_sb_info *sbi = EXT4_SB(sb);
1719 if (sbi->s_jquota_fmt) {
1722 switch (sbi->s_jquota_fmt) {
1733 seq_printf(seq, ",jqfmt=%s", fmtname);
1736 if (sbi->s_qf_names[USRQUOTA])
1737 seq_printf(seq, ",usrjquota=%s", sbi->s_qf_names[USRQUOTA]);
1739 if (sbi->s_qf_names[GRPQUOTA])
1740 seq_printf(seq, ",grpjquota=%s", sbi->s_qf_names[GRPQUOTA]);
1742 if (test_opt(sb, USRQUOTA))
1743 seq_puts(seq, ",usrquota");
1745 if (test_opt(sb, GRPQUOTA))
1746 seq_puts(seq, ",grpquota");
1750 static const char *token2str(int token)
1752 static const struct match_token *t;
1754 for (t = tokens; t->token != Opt_err; t++)
1755 if (t->token == token && !strchr(t->pattern, '='))
1762 * - it's set to a non-default value OR
1763 * - if the per-sb default is different from the global default
1765 static int _ext4_show_options(struct seq_file *seq, struct super_block *sb,
1768 struct ext4_sb_info *sbi = EXT4_SB(sb);
1769 struct ext4_super_block *es = sbi->s_es;
1770 int def_errors, def_mount_opt = nodefs ? 0 : sbi->s_def_mount_opt;
1771 const struct mount_opts *m;
1772 char sep = nodefs ? '\n' : ',';
1774 #define SEQ_OPTS_PUTS(str) seq_printf(seq, "%c" str, sep)
1775 #define SEQ_OPTS_PRINT(str, arg) seq_printf(seq, "%c" str, sep, arg)
1777 if (sbi->s_sb_block != 1)
1778 SEQ_OPTS_PRINT("sb=%llu", sbi->s_sb_block);
1780 for (m = ext4_mount_opts; m->token != Opt_err; m++) {
1781 int want_set = m->flags & MOPT_SET;
1782 if (((m->flags & (MOPT_SET|MOPT_CLEAR)) == 0) ||
1783 (m->flags & MOPT_CLEAR_ERR))
1785 if (!(m->mount_opt & (sbi->s_mount_opt ^ def_mount_opt)))
1786 continue; /* skip if same as the default */
1788 (sbi->s_mount_opt & m->mount_opt) != m->mount_opt) ||
1789 (!want_set && (sbi->s_mount_opt & m->mount_opt)))
1790 continue; /* select Opt_noFoo vs Opt_Foo */
1791 SEQ_OPTS_PRINT("%s", token2str(m->token));
1794 if (nodefs || !uid_eq(sbi->s_resuid, make_kuid(&init_user_ns, EXT4_DEF_RESUID)) ||
1795 le16_to_cpu(es->s_def_resuid) != EXT4_DEF_RESUID)
1796 SEQ_OPTS_PRINT("resuid=%u",
1797 from_kuid_munged(&init_user_ns, sbi->s_resuid));
1798 if (nodefs || !gid_eq(sbi->s_resgid, make_kgid(&init_user_ns, EXT4_DEF_RESGID)) ||
1799 le16_to_cpu(es->s_def_resgid) != EXT4_DEF_RESGID)
1800 SEQ_OPTS_PRINT("resgid=%u",
1801 from_kgid_munged(&init_user_ns, sbi->s_resgid));
1802 def_errors = nodefs ? -1 : le16_to_cpu(es->s_errors);
1803 if (test_opt(sb, ERRORS_RO) && def_errors != EXT4_ERRORS_RO)
1804 SEQ_OPTS_PUTS("errors=remount-ro");
1805 if (test_opt(sb, ERRORS_CONT) && def_errors != EXT4_ERRORS_CONTINUE)
1806 SEQ_OPTS_PUTS("errors=continue");
1807 if (test_opt(sb, ERRORS_PANIC) && def_errors != EXT4_ERRORS_PANIC)
1808 SEQ_OPTS_PUTS("errors=panic");
1809 if (nodefs || sbi->s_commit_interval != JBD2_DEFAULT_MAX_COMMIT_AGE*HZ)
1810 SEQ_OPTS_PRINT("commit=%lu", sbi->s_commit_interval / HZ);
1811 if (nodefs || sbi->s_min_batch_time != EXT4_DEF_MIN_BATCH_TIME)
1812 SEQ_OPTS_PRINT("min_batch_time=%u", sbi->s_min_batch_time);
1813 if (nodefs || sbi->s_max_batch_time != EXT4_DEF_MAX_BATCH_TIME)
1814 SEQ_OPTS_PRINT("max_batch_time=%u", sbi->s_max_batch_time);
1815 if (sb->s_flags & MS_I_VERSION)
1816 SEQ_OPTS_PUTS("i_version");
1817 if (nodefs || sbi->s_stripe)
1818 SEQ_OPTS_PRINT("stripe=%lu", sbi->s_stripe);
1819 if (EXT4_MOUNT_DATA_FLAGS & (sbi->s_mount_opt ^ def_mount_opt)) {
1820 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA)
1821 SEQ_OPTS_PUTS("data=journal");
1822 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA)
1823 SEQ_OPTS_PUTS("data=ordered");
1824 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_WRITEBACK_DATA)
1825 SEQ_OPTS_PUTS("data=writeback");
1828 sbi->s_inode_readahead_blks != EXT4_DEF_INODE_READAHEAD_BLKS)
1829 SEQ_OPTS_PRINT("inode_readahead_blks=%u",
1830 sbi->s_inode_readahead_blks);
1832 if (nodefs || (test_opt(sb, INIT_INODE_TABLE) &&
1833 (sbi->s_li_wait_mult != EXT4_DEF_LI_WAIT_MULT)))
1834 SEQ_OPTS_PRINT("init_itable=%u", sbi->s_li_wait_mult);
1835 if (nodefs || sbi->s_max_dir_size_kb)
1836 SEQ_OPTS_PRINT("max_dir_size_kb=%u", sbi->s_max_dir_size_kb);
1838 ext4_show_quota_options(seq, sb);
1842 static int ext4_show_options(struct seq_file *seq, struct dentry *root)
1844 return _ext4_show_options(seq, root->d_sb, 0);
1847 static int options_seq_show(struct seq_file *seq, void *offset)
1849 struct super_block *sb = seq->private;
1852 seq_puts(seq, (sb->s_flags & MS_RDONLY) ? "ro" : "rw");
1853 rc = _ext4_show_options(seq, sb, 1);
1854 seq_puts(seq, "\n");
1858 static int options_open_fs(struct inode *inode, struct file *file)
1860 return single_open(file, options_seq_show, PDE(inode)->data);
1863 static const struct file_operations ext4_seq_options_fops = {
1864 .owner = THIS_MODULE,
1865 .open = options_open_fs,
1867 .llseek = seq_lseek,
1868 .release = single_release,
1871 static int ext4_setup_super(struct super_block *sb, struct ext4_super_block *es,
1874 struct ext4_sb_info *sbi = EXT4_SB(sb);
1877 if (le32_to_cpu(es->s_rev_level) > EXT4_MAX_SUPP_REV) {
1878 ext4_msg(sb, KERN_ERR, "revision level too high, "
1879 "forcing read-only mode");
1884 if (!(sbi->s_mount_state & EXT4_VALID_FS))
1885 ext4_msg(sb, KERN_WARNING, "warning: mounting unchecked fs, "
1886 "running e2fsck is recommended");
1887 else if ((sbi->s_mount_state & EXT4_ERROR_FS))
1888 ext4_msg(sb, KERN_WARNING,
1889 "warning: mounting fs with errors, "
1890 "running e2fsck is recommended");
1891 else if ((__s16) le16_to_cpu(es->s_max_mnt_count) > 0 &&
1892 le16_to_cpu(es->s_mnt_count) >=
1893 (unsigned short) (__s16) le16_to_cpu(es->s_max_mnt_count))
1894 ext4_msg(sb, KERN_WARNING,
1895 "warning: maximal mount count reached, "
1896 "running e2fsck is recommended");
1897 else if (le32_to_cpu(es->s_checkinterval) &&
1898 (le32_to_cpu(es->s_lastcheck) +
1899 le32_to_cpu(es->s_checkinterval) <= get_seconds()))
1900 ext4_msg(sb, KERN_WARNING,
1901 "warning: checktime reached, "
1902 "running e2fsck is recommended");
1903 if (!sbi->s_journal)
1904 es->s_state &= cpu_to_le16(~EXT4_VALID_FS);
1905 if (!(__s16) le16_to_cpu(es->s_max_mnt_count))
1906 es->s_max_mnt_count = cpu_to_le16(EXT4_DFL_MAX_MNT_COUNT);
1907 le16_add_cpu(&es->s_mnt_count, 1);
1908 es->s_mtime = cpu_to_le32(get_seconds());
1909 ext4_update_dynamic_rev(sb);
1911 EXT4_SET_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
1913 ext4_commit_super(sb, 1);
1915 if (test_opt(sb, DEBUG))
1916 printk(KERN_INFO "[EXT4 FS bs=%lu, gc=%u, "
1917 "bpg=%lu, ipg=%lu, mo=%04x, mo2=%04x]\n",
1919 sbi->s_groups_count,
1920 EXT4_BLOCKS_PER_GROUP(sb),
1921 EXT4_INODES_PER_GROUP(sb),
1922 sbi->s_mount_opt, sbi->s_mount_opt2);
1924 cleancache_init_fs(sb);
1928 int ext4_alloc_flex_bg_array(struct super_block *sb, ext4_group_t ngroup)
1930 struct ext4_sb_info *sbi = EXT4_SB(sb);
1931 struct flex_groups *new_groups;
1934 if (!sbi->s_log_groups_per_flex)
1937 size = ext4_flex_group(sbi, ngroup - 1) + 1;
1938 if (size <= sbi->s_flex_groups_allocated)
1941 size = roundup_pow_of_two(size * sizeof(struct flex_groups));
1942 new_groups = ext4_kvzalloc(size, GFP_KERNEL);
1944 ext4_msg(sb, KERN_ERR, "not enough memory for %d flex groups",
1945 size / (int) sizeof(struct flex_groups));
1949 if (sbi->s_flex_groups) {
1950 memcpy(new_groups, sbi->s_flex_groups,
1951 (sbi->s_flex_groups_allocated *
1952 sizeof(struct flex_groups)));
1953 ext4_kvfree(sbi->s_flex_groups);
1955 sbi->s_flex_groups = new_groups;
1956 sbi->s_flex_groups_allocated = size / sizeof(struct flex_groups);
1960 static int ext4_fill_flex_info(struct super_block *sb)
1962 struct ext4_sb_info *sbi = EXT4_SB(sb);
1963 struct ext4_group_desc *gdp = NULL;
1964 ext4_group_t flex_group;
1965 unsigned int groups_per_flex = 0;
1968 sbi->s_log_groups_per_flex = sbi->s_es->s_log_groups_per_flex;
1969 if (sbi->s_log_groups_per_flex < 1 || sbi->s_log_groups_per_flex > 31) {
1970 sbi->s_log_groups_per_flex = 0;
1973 groups_per_flex = 1 << sbi->s_log_groups_per_flex;
1975 err = ext4_alloc_flex_bg_array(sb, sbi->s_groups_count);
1979 for (i = 0; i < sbi->s_groups_count; i++) {
1980 gdp = ext4_get_group_desc(sb, i, NULL);
1982 flex_group = ext4_flex_group(sbi, i);
1983 atomic_add(ext4_free_inodes_count(sb, gdp),
1984 &sbi->s_flex_groups[flex_group].free_inodes);
1985 atomic_add(ext4_free_group_clusters(sb, gdp),
1986 &sbi->s_flex_groups[flex_group].free_clusters);
1987 atomic_add(ext4_used_dirs_count(sb, gdp),
1988 &sbi->s_flex_groups[flex_group].used_dirs);
1996 static __le16 ext4_group_desc_csum(struct ext4_sb_info *sbi, __u32 block_group,
1997 struct ext4_group_desc *gdp)
2001 __le32 le_group = cpu_to_le32(block_group);
2003 if ((sbi->s_es->s_feature_ro_compat &
2004 cpu_to_le32(EXT4_FEATURE_RO_COMPAT_METADATA_CSUM))) {
2005 /* Use new metadata_csum algorithm */
2009 old_csum = gdp->bg_checksum;
2010 gdp->bg_checksum = 0;
2011 csum32 = ext4_chksum(sbi, sbi->s_csum_seed, (__u8 *)&le_group,
2013 csum32 = ext4_chksum(sbi, csum32, (__u8 *)gdp,
2015 gdp->bg_checksum = old_csum;
2017 crc = csum32 & 0xFFFF;
2021 /* old crc16 code */
2022 offset = offsetof(struct ext4_group_desc, bg_checksum);
2024 crc = crc16(~0, sbi->s_es->s_uuid, sizeof(sbi->s_es->s_uuid));
2025 crc = crc16(crc, (__u8 *)&le_group, sizeof(le_group));
2026 crc = crc16(crc, (__u8 *)gdp, offset);
2027 offset += sizeof(gdp->bg_checksum); /* skip checksum */
2028 /* for checksum of struct ext4_group_desc do the rest...*/
2029 if ((sbi->s_es->s_feature_incompat &
2030 cpu_to_le32(EXT4_FEATURE_INCOMPAT_64BIT)) &&
2031 offset < le16_to_cpu(sbi->s_es->s_desc_size))
2032 crc = crc16(crc, (__u8 *)gdp + offset,
2033 le16_to_cpu(sbi->s_es->s_desc_size) -
2037 return cpu_to_le16(crc);
2040 int ext4_group_desc_csum_verify(struct super_block *sb, __u32 block_group,
2041 struct ext4_group_desc *gdp)
2043 if (ext4_has_group_desc_csum(sb) &&
2044 (gdp->bg_checksum != ext4_group_desc_csum(EXT4_SB(sb),
2051 void ext4_group_desc_csum_set(struct super_block *sb, __u32 block_group,
2052 struct ext4_group_desc *gdp)
2054 if (!ext4_has_group_desc_csum(sb))
2056 gdp->bg_checksum = ext4_group_desc_csum(EXT4_SB(sb), block_group, gdp);
2059 /* Called at mount-time, super-block is locked */
2060 static int ext4_check_descriptors(struct super_block *sb,
2061 ext4_group_t *first_not_zeroed)
2063 struct ext4_sb_info *sbi = EXT4_SB(sb);
2064 ext4_fsblk_t first_block = le32_to_cpu(sbi->s_es->s_first_data_block);
2065 ext4_fsblk_t last_block;
2066 ext4_fsblk_t block_bitmap;
2067 ext4_fsblk_t inode_bitmap;
2068 ext4_fsblk_t inode_table;
2069 int flexbg_flag = 0;
2070 ext4_group_t i, grp = sbi->s_groups_count;
2072 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_FLEX_BG))
2075 ext4_debug("Checking group descriptors");
2077 for (i = 0; i < sbi->s_groups_count; i++) {
2078 struct ext4_group_desc *gdp = ext4_get_group_desc(sb, i, NULL);
2080 if (i == sbi->s_groups_count - 1 || flexbg_flag)
2081 last_block = ext4_blocks_count(sbi->s_es) - 1;
2083 last_block = first_block +
2084 (EXT4_BLOCKS_PER_GROUP(sb) - 1);
2086 if ((grp == sbi->s_groups_count) &&
2087 !(gdp->bg_flags & cpu_to_le16(EXT4_BG_INODE_ZEROED)))
2090 block_bitmap = ext4_block_bitmap(sb, gdp);
2091 if (block_bitmap < first_block || block_bitmap > last_block) {
2092 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2093 "Block bitmap for group %u not in group "
2094 "(block %llu)!", i, block_bitmap);
2097 inode_bitmap = ext4_inode_bitmap(sb, gdp);
2098 if (inode_bitmap < first_block || inode_bitmap > last_block) {
2099 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2100 "Inode bitmap for group %u not in group "
2101 "(block %llu)!", i, inode_bitmap);
2104 inode_table = ext4_inode_table(sb, gdp);
2105 if (inode_table < first_block ||
2106 inode_table + sbi->s_itb_per_group - 1 > last_block) {
2107 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2108 "Inode table for group %u not in group "
2109 "(block %llu)!", i, inode_table);
2112 ext4_lock_group(sb, i);
2113 if (!ext4_group_desc_csum_verify(sb, i, gdp)) {
2114 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2115 "Checksum for group %u failed (%u!=%u)",
2116 i, le16_to_cpu(ext4_group_desc_csum(sbi, i,
2117 gdp)), le16_to_cpu(gdp->bg_checksum));
2118 if (!(sb->s_flags & MS_RDONLY)) {
2119 ext4_unlock_group(sb, i);
2123 ext4_unlock_group(sb, i);
2125 first_block += EXT4_BLOCKS_PER_GROUP(sb);
2127 if (NULL != first_not_zeroed)
2128 *first_not_zeroed = grp;
2130 ext4_free_blocks_count_set(sbi->s_es,
2131 EXT4_C2B(sbi, ext4_count_free_clusters(sb)));
2132 sbi->s_es->s_free_inodes_count =cpu_to_le32(ext4_count_free_inodes(sb));
2136 /* ext4_orphan_cleanup() walks a singly-linked list of inodes (starting at
2137 * the superblock) which were deleted from all directories, but held open by
2138 * a process at the time of a crash. We walk the list and try to delete these
2139 * inodes at recovery time (only with a read-write filesystem).
2141 * In order to keep the orphan inode chain consistent during traversal (in
2142 * case of crash during recovery), we link each inode into the superblock
2143 * orphan list_head and handle it the same way as an inode deletion during
2144 * normal operation (which journals the operations for us).
2146 * We only do an iget() and an iput() on each inode, which is very safe if we
2147 * accidentally point at an in-use or already deleted inode. The worst that
2148 * can happen in this case is that we get a "bit already cleared" message from
2149 * ext4_free_inode(). The only reason we would point at a wrong inode is if
2150 * e2fsck was run on this filesystem, and it must have already done the orphan
2151 * inode cleanup for us, so we can safely abort without any further action.
2153 static void ext4_orphan_cleanup(struct super_block *sb,
2154 struct ext4_super_block *es)
2156 unsigned int s_flags = sb->s_flags;
2157 int nr_orphans = 0, nr_truncates = 0;
2161 if (!es->s_last_orphan) {
2162 jbd_debug(4, "no orphan inodes to clean up\n");
2166 if (bdev_read_only(sb->s_bdev)) {
2167 ext4_msg(sb, KERN_ERR, "write access "
2168 "unavailable, skipping orphan cleanup");
2172 /* Check if feature set would not allow a r/w mount */
2173 if (!ext4_feature_set_ok(sb, 0)) {
2174 ext4_msg(sb, KERN_INFO, "Skipping orphan cleanup due to "
2175 "unknown ROCOMPAT features");
2179 if (EXT4_SB(sb)->s_mount_state & EXT4_ERROR_FS) {
2180 if (es->s_last_orphan)
2181 jbd_debug(1, "Errors on filesystem, "
2182 "clearing orphan list.\n");
2183 es->s_last_orphan = 0;
2184 jbd_debug(1, "Skipping orphan recovery on fs with errors.\n");
2188 if (s_flags & MS_RDONLY) {
2189 ext4_msg(sb, KERN_INFO, "orphan cleanup on readonly fs");
2190 sb->s_flags &= ~MS_RDONLY;
2193 /* Needed for iput() to work correctly and not trash data */
2194 sb->s_flags |= MS_ACTIVE;
2195 /* Turn on quotas so that they are updated correctly */
2196 for (i = 0; i < MAXQUOTAS; i++) {
2197 if (EXT4_SB(sb)->s_qf_names[i]) {
2198 int ret = ext4_quota_on_mount(sb, i);
2200 ext4_msg(sb, KERN_ERR,
2201 "Cannot turn on journaled "
2202 "quota: error %d", ret);
2207 while (es->s_last_orphan) {
2208 struct inode *inode;
2210 inode = ext4_orphan_get(sb, le32_to_cpu(es->s_last_orphan));
2211 if (IS_ERR(inode)) {
2212 es->s_last_orphan = 0;
2216 list_add(&EXT4_I(inode)->i_orphan, &EXT4_SB(sb)->s_orphan);
2217 dquot_initialize(inode);
2218 if (inode->i_nlink) {
2219 ext4_msg(sb, KERN_DEBUG,
2220 "%s: truncating inode %lu to %lld bytes",
2221 __func__, inode->i_ino, inode->i_size);
2222 jbd_debug(2, "truncating inode %lu to %lld bytes\n",
2223 inode->i_ino, inode->i_size);
2224 ext4_truncate(inode);
2227 ext4_msg(sb, KERN_DEBUG,
2228 "%s: deleting unreferenced inode %lu",
2229 __func__, inode->i_ino);
2230 jbd_debug(2, "deleting unreferenced inode %lu\n",
2234 iput(inode); /* The delete magic happens here! */
2237 #define PLURAL(x) (x), ((x) == 1) ? "" : "s"
2240 ext4_msg(sb, KERN_INFO, "%d orphan inode%s deleted",
2241 PLURAL(nr_orphans));
2243 ext4_msg(sb, KERN_INFO, "%d truncate%s cleaned up",
2244 PLURAL(nr_truncates));
2246 /* Turn quotas off */
2247 for (i = 0; i < MAXQUOTAS; i++) {
2248 if (sb_dqopt(sb)->files[i])
2249 dquot_quota_off(sb, i);
2252 sb->s_flags = s_flags; /* Restore MS_RDONLY status */
2256 * Maximal extent format file size.
2257 * Resulting logical blkno at s_maxbytes must fit in our on-disk
2258 * extent format containers, within a sector_t, and within i_blocks
2259 * in the vfs. ext4 inode has 48 bits of i_block in fsblock units,
2260 * so that won't be a limiting factor.
2262 * However there is other limiting factor. We do store extents in the form
2263 * of starting block and length, hence the resulting length of the extent
2264 * covering maximum file size must fit into on-disk format containers as
2265 * well. Given that length is always by 1 unit bigger than max unit (because
2266 * we count 0 as well) we have to lower the s_maxbytes by one fs block.
2268 * Note, this does *not* consider any metadata overhead for vfs i_blocks.
2270 static loff_t ext4_max_size(int blkbits, int has_huge_files)
2273 loff_t upper_limit = MAX_LFS_FILESIZE;
2275 /* small i_blocks in vfs inode? */
2276 if (!has_huge_files || sizeof(blkcnt_t) < sizeof(u64)) {
2278 * CONFIG_LBDAF is not enabled implies the inode
2279 * i_block represent total blocks in 512 bytes
2280 * 32 == size of vfs inode i_blocks * 8
2282 upper_limit = (1LL << 32) - 1;
2284 /* total blocks in file system block size */
2285 upper_limit >>= (blkbits - 9);
2286 upper_limit <<= blkbits;
2290 * 32-bit extent-start container, ee_block. We lower the maxbytes
2291 * by one fs block, so ee_len can cover the extent of maximum file
2294 res = (1LL << 32) - 1;
2297 /* Sanity check against vm- & vfs- imposed limits */
2298 if (res > upper_limit)
2305 * Maximal bitmap file size. There is a direct, and {,double-,triple-}indirect
2306 * block limit, and also a limit of (2^48 - 1) 512-byte sectors in i_blocks.
2307 * We need to be 1 filesystem block less than the 2^48 sector limit.
2309 static loff_t ext4_max_bitmap_size(int bits, int has_huge_files)
2311 loff_t res = EXT4_NDIR_BLOCKS;
2314 /* This is calculated to be the largest file size for a dense, block
2315 * mapped file such that the file's total number of 512-byte sectors,
2316 * including data and all indirect blocks, does not exceed (2^48 - 1).
2318 * __u32 i_blocks_lo and _u16 i_blocks_high represent the total
2319 * number of 512-byte sectors of the file.
2322 if (!has_huge_files || sizeof(blkcnt_t) < sizeof(u64)) {
2324 * !has_huge_files or CONFIG_LBDAF not enabled implies that
2325 * the inode i_block field represents total file blocks in
2326 * 2^32 512-byte sectors == size of vfs inode i_blocks * 8
2328 upper_limit = (1LL << 32) - 1;
2330 /* total blocks in file system block size */
2331 upper_limit >>= (bits - 9);
2335 * We use 48 bit ext4_inode i_blocks
2336 * With EXT4_HUGE_FILE_FL set the i_blocks
2337 * represent total number of blocks in
2338 * file system block size
2340 upper_limit = (1LL << 48) - 1;
2344 /* indirect blocks */
2346 /* double indirect blocks */
2347 meta_blocks += 1 + (1LL << (bits-2));
2348 /* tripple indirect blocks */
2349 meta_blocks += 1 + (1LL << (bits-2)) + (1LL << (2*(bits-2)));
2351 upper_limit -= meta_blocks;
2352 upper_limit <<= bits;
2354 res += 1LL << (bits-2);
2355 res += 1LL << (2*(bits-2));
2356 res += 1LL << (3*(bits-2));
2358 if (res > upper_limit)
2361 if (res > MAX_LFS_FILESIZE)
2362 res = MAX_LFS_FILESIZE;
2367 static ext4_fsblk_t descriptor_loc(struct super_block *sb,
2368 ext4_fsblk_t logical_sb_block, int nr)
2370 struct ext4_sb_info *sbi = EXT4_SB(sb);
2371 ext4_group_t bg, first_meta_bg;
2374 first_meta_bg = le32_to_cpu(sbi->s_es->s_first_meta_bg);
2376 if (!EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_META_BG) ||
2378 return logical_sb_block + nr + 1;
2379 bg = sbi->s_desc_per_block * nr;
2380 if (ext4_bg_has_super(sb, bg))
2383 return (has_super + ext4_group_first_block_no(sb, bg));
2387 * ext4_get_stripe_size: Get the stripe size.
2388 * @sbi: In memory super block info
2390 * If we have specified it via mount option, then
2391 * use the mount option value. If the value specified at mount time is
2392 * greater than the blocks per group use the super block value.
2393 * If the super block value is greater than blocks per group return 0.
2394 * Allocator needs it be less than blocks per group.
2397 static unsigned long ext4_get_stripe_size(struct ext4_sb_info *sbi)
2399 unsigned long stride = le16_to_cpu(sbi->s_es->s_raid_stride);
2400 unsigned long stripe_width =
2401 le32_to_cpu(sbi->s_es->s_raid_stripe_width);
2404 if (sbi->s_stripe && sbi->s_stripe <= sbi->s_blocks_per_group)
2405 ret = sbi->s_stripe;
2406 else if (stripe_width <= sbi->s_blocks_per_group)
2408 else if (stride <= sbi->s_blocks_per_group)
2414 * If the stripe width is 1, this makes no sense and
2415 * we set it to 0 to turn off stripe handling code.
2426 struct attribute attr;
2427 ssize_t (*show)(struct ext4_attr *, struct ext4_sb_info *, char *);
2428 ssize_t (*store)(struct ext4_attr *, struct ext4_sb_info *,
2429 const char *, size_t);
2433 static int parse_strtoul(const char *buf,
2434 unsigned long max, unsigned long *value)
2438 *value = simple_strtoul(skip_spaces(buf), &endp, 0);
2439 endp = skip_spaces(endp);
2440 if (*endp || *value > max)
2446 static ssize_t delayed_allocation_blocks_show(struct ext4_attr *a,
2447 struct ext4_sb_info *sbi,
2450 return snprintf(buf, PAGE_SIZE, "%llu\n",
2452 percpu_counter_sum(&sbi->s_dirtyclusters_counter)));
2455 static ssize_t session_write_kbytes_show(struct ext4_attr *a,
2456 struct ext4_sb_info *sbi, char *buf)
2458 struct super_block *sb = sbi->s_buddy_cache->i_sb;
2460 if (!sb->s_bdev->bd_part)
2461 return snprintf(buf, PAGE_SIZE, "0\n");
2462 return snprintf(buf, PAGE_SIZE, "%lu\n",
2463 (part_stat_read(sb->s_bdev->bd_part, sectors[1]) -
2464 sbi->s_sectors_written_start) >> 1);
2467 static ssize_t lifetime_write_kbytes_show(struct ext4_attr *a,
2468 struct ext4_sb_info *sbi, char *buf)
2470 struct super_block *sb = sbi->s_buddy_cache->i_sb;
2472 if (!sb->s_bdev->bd_part)
2473 return snprintf(buf, PAGE_SIZE, "0\n");
2474 return snprintf(buf, PAGE_SIZE, "%llu\n",
2475 (unsigned long long)(sbi->s_kbytes_written +
2476 ((part_stat_read(sb->s_bdev->bd_part, sectors[1]) -
2477 EXT4_SB(sb)->s_sectors_written_start) >> 1)));
2480 static ssize_t inode_readahead_blks_store(struct ext4_attr *a,
2481 struct ext4_sb_info *sbi,
2482 const char *buf, size_t count)
2486 if (parse_strtoul(buf, 0x40000000, &t))
2489 if (t && !is_power_of_2(t))
2492 sbi->s_inode_readahead_blks = t;
2496 static ssize_t sbi_ui_show(struct ext4_attr *a,
2497 struct ext4_sb_info *sbi, char *buf)
2499 unsigned int *ui = (unsigned int *) (((char *) sbi) + a->offset);
2501 return snprintf(buf, PAGE_SIZE, "%u\n", *ui);
2504 static ssize_t sbi_ui_store(struct ext4_attr *a,
2505 struct ext4_sb_info *sbi,
2506 const char *buf, size_t count)
2508 unsigned int *ui = (unsigned int *) (((char *) sbi) + a->offset);
2511 if (parse_strtoul(buf, 0xffffffff, &t))
2517 static ssize_t trigger_test_error(struct ext4_attr *a,
2518 struct ext4_sb_info *sbi,
2519 const char *buf, size_t count)
2523 if (!capable(CAP_SYS_ADMIN))
2526 if (len && buf[len-1] == '\n')
2530 ext4_error(sbi->s_sb, "%.*s", len, buf);
2534 #define EXT4_ATTR_OFFSET(_name,_mode,_show,_store,_elname) \
2535 static struct ext4_attr ext4_attr_##_name = { \
2536 .attr = {.name = __stringify(_name), .mode = _mode }, \
2539 .offset = offsetof(struct ext4_sb_info, _elname), \
2541 #define EXT4_ATTR(name, mode, show, store) \
2542 static struct ext4_attr ext4_attr_##name = __ATTR(name, mode, show, store)
2544 #define EXT4_INFO_ATTR(name) EXT4_ATTR(name, 0444, NULL, NULL)
2545 #define EXT4_RO_ATTR(name) EXT4_ATTR(name, 0444, name##_show, NULL)
2546 #define EXT4_RW_ATTR(name) EXT4_ATTR(name, 0644, name##_show, name##_store)
2547 #define EXT4_RW_ATTR_SBI_UI(name, elname) \
2548 EXT4_ATTR_OFFSET(name, 0644, sbi_ui_show, sbi_ui_store, elname)
2549 #define ATTR_LIST(name) &ext4_attr_##name.attr
2551 EXT4_RO_ATTR(delayed_allocation_blocks);
2552 EXT4_RO_ATTR(session_write_kbytes);
2553 EXT4_RO_ATTR(lifetime_write_kbytes);
2554 EXT4_ATTR_OFFSET(inode_readahead_blks, 0644, sbi_ui_show,
2555 inode_readahead_blks_store, s_inode_readahead_blks);
2556 EXT4_RW_ATTR_SBI_UI(inode_goal, s_inode_goal);
2557 EXT4_RW_ATTR_SBI_UI(mb_stats, s_mb_stats);
2558 EXT4_RW_ATTR_SBI_UI(mb_max_to_scan, s_mb_max_to_scan);
2559 EXT4_RW_ATTR_SBI_UI(mb_min_to_scan, s_mb_min_to_scan);
2560 EXT4_RW_ATTR_SBI_UI(mb_order2_req, s_mb_order2_reqs);
2561 EXT4_RW_ATTR_SBI_UI(mb_stream_req, s_mb_stream_request);
2562 EXT4_RW_ATTR_SBI_UI(mb_group_prealloc, s_mb_group_prealloc);
2563 EXT4_RW_ATTR_SBI_UI(max_writeback_mb_bump, s_max_writeback_mb_bump);
2564 EXT4_RW_ATTR_SBI_UI(extent_max_zeroout_kb, s_extent_max_zeroout_kb);
2565 EXT4_ATTR(trigger_fs_error, 0200, NULL, trigger_test_error);
2567 static struct attribute *ext4_attrs[] = {
2568 ATTR_LIST(delayed_allocation_blocks),
2569 ATTR_LIST(session_write_kbytes),
2570 ATTR_LIST(lifetime_write_kbytes),
2571 ATTR_LIST(inode_readahead_blks),
2572 ATTR_LIST(inode_goal),
2573 ATTR_LIST(mb_stats),
2574 ATTR_LIST(mb_max_to_scan),
2575 ATTR_LIST(mb_min_to_scan),
2576 ATTR_LIST(mb_order2_req),
2577 ATTR_LIST(mb_stream_req),
2578 ATTR_LIST(mb_group_prealloc),
2579 ATTR_LIST(max_writeback_mb_bump),
2580 ATTR_LIST(extent_max_zeroout_kb),
2581 ATTR_LIST(trigger_fs_error),
2585 /* Features this copy of ext4 supports */
2586 EXT4_INFO_ATTR(lazy_itable_init);
2587 EXT4_INFO_ATTR(batched_discard);
2589 static struct attribute *ext4_feat_attrs[] = {
2590 ATTR_LIST(lazy_itable_init),
2591 ATTR_LIST(batched_discard),
2595 static ssize_t ext4_attr_show(struct kobject *kobj,
2596 struct attribute *attr, char *buf)
2598 struct ext4_sb_info *sbi = container_of(kobj, struct ext4_sb_info,
2600 struct ext4_attr *a = container_of(attr, struct ext4_attr, attr);
2602 return a->show ? a->show(a, sbi, buf) : 0;
2605 static ssize_t ext4_attr_store(struct kobject *kobj,
2606 struct attribute *attr,
2607 const char *buf, size_t len)
2609 struct ext4_sb_info *sbi = container_of(kobj, struct ext4_sb_info,
2611 struct ext4_attr *a = container_of(attr, struct ext4_attr, attr);
2613 return a->store ? a->store(a, sbi, buf, len) : 0;
2616 static void ext4_sb_release(struct kobject *kobj)
2618 struct ext4_sb_info *sbi = container_of(kobj, struct ext4_sb_info,
2620 complete(&sbi->s_kobj_unregister);
2623 static const struct sysfs_ops ext4_attr_ops = {
2624 .show = ext4_attr_show,
2625 .store = ext4_attr_store,
2628 static struct kobj_type ext4_ktype = {
2629 .default_attrs = ext4_attrs,
2630 .sysfs_ops = &ext4_attr_ops,
2631 .release = ext4_sb_release,
2634 static void ext4_feat_release(struct kobject *kobj)
2636 complete(&ext4_feat->f_kobj_unregister);
2639 static struct kobj_type ext4_feat_ktype = {
2640 .default_attrs = ext4_feat_attrs,
2641 .sysfs_ops = &ext4_attr_ops,
2642 .release = ext4_feat_release,
2646 * Check whether this filesystem can be mounted based on
2647 * the features present and the RDONLY/RDWR mount requested.
2648 * Returns 1 if this filesystem can be mounted as requested,
2649 * 0 if it cannot be.
2651 static int ext4_feature_set_ok(struct super_block *sb, int readonly)
2653 if (EXT4_HAS_INCOMPAT_FEATURE(sb, ~EXT4_FEATURE_INCOMPAT_SUPP)) {
2654 ext4_msg(sb, KERN_ERR,
2655 "Couldn't mount because of "
2656 "unsupported optional features (%x)",
2657 (le32_to_cpu(EXT4_SB(sb)->s_es->s_feature_incompat) &
2658 ~EXT4_FEATURE_INCOMPAT_SUPP));
2665 /* Check that feature set is OK for a read-write mount */
2666 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, ~EXT4_FEATURE_RO_COMPAT_SUPP)) {
2667 ext4_msg(sb, KERN_ERR, "couldn't mount RDWR because of "
2668 "unsupported optional features (%x)",
2669 (le32_to_cpu(EXT4_SB(sb)->s_es->s_feature_ro_compat) &
2670 ~EXT4_FEATURE_RO_COMPAT_SUPP));
2674 * Large file size enabled file system can only be mounted
2675 * read-write on 32-bit systems if kernel is built with CONFIG_LBDAF
2677 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_HUGE_FILE)) {
2678 if (sizeof(blkcnt_t) < sizeof(u64)) {
2679 ext4_msg(sb, KERN_ERR, "Filesystem with huge files "
2680 "cannot be mounted RDWR without "
2685 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_BIGALLOC) &&
2686 !EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_EXTENTS)) {
2687 ext4_msg(sb, KERN_ERR,
2688 "Can't support bigalloc feature without "
2689 "extents feature\n");
2693 #ifndef CONFIG_QUOTA
2694 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_QUOTA) &&
2696 ext4_msg(sb, KERN_ERR,
2697 "Filesystem with quota feature cannot be mounted RDWR "
2698 "without CONFIG_QUOTA");
2701 #endif /* CONFIG_QUOTA */
2706 * This function is called once a day if we have errors logged
2707 * on the file system
2709 static void print_daily_error_info(unsigned long arg)
2711 struct super_block *sb = (struct super_block *) arg;
2712 struct ext4_sb_info *sbi;
2713 struct ext4_super_block *es;
2718 if (es->s_error_count)
2719 ext4_msg(sb, KERN_NOTICE, "error count: %u",
2720 le32_to_cpu(es->s_error_count));
2721 if (es->s_first_error_time) {
2722 printk(KERN_NOTICE "EXT4-fs (%s): initial error at %u: %.*s:%d",
2723 sb->s_id, le32_to_cpu(es->s_first_error_time),
2724 (int) sizeof(es->s_first_error_func),
2725 es->s_first_error_func,
2726 le32_to_cpu(es->s_first_error_line));
2727 if (es->s_first_error_ino)
2728 printk(": inode %u",
2729 le32_to_cpu(es->s_first_error_ino));
2730 if (es->s_first_error_block)
2731 printk(": block %llu", (unsigned long long)
2732 le64_to_cpu(es->s_first_error_block));
2735 if (es->s_last_error_time) {
2736 printk(KERN_NOTICE "EXT4-fs (%s): last error at %u: %.*s:%d",
2737 sb->s_id, le32_to_cpu(es->s_last_error_time),
2738 (int) sizeof(es->s_last_error_func),
2739 es->s_last_error_func,
2740 le32_to_cpu(es->s_last_error_line));
2741 if (es->s_last_error_ino)
2742 printk(": inode %u",
2743 le32_to_cpu(es->s_last_error_ino));
2744 if (es->s_last_error_block)
2745 printk(": block %llu", (unsigned long long)
2746 le64_to_cpu(es->s_last_error_block));
2749 mod_timer(&sbi->s_err_report, jiffies + 24*60*60*HZ); /* Once a day */
2752 /* Find next suitable group and run ext4_init_inode_table */
2753 static int ext4_run_li_request(struct ext4_li_request *elr)
2755 struct ext4_group_desc *gdp = NULL;
2756 ext4_group_t group, ngroups;
2757 struct super_block *sb;
2758 unsigned long timeout = 0;
2762 ngroups = EXT4_SB(sb)->s_groups_count;
2765 for (group = elr->lr_next_group; group < ngroups; group++) {
2766 gdp = ext4_get_group_desc(sb, group, NULL);
2772 if (!(gdp->bg_flags & cpu_to_le16(EXT4_BG_INODE_ZEROED)))
2776 if (group == ngroups)
2781 ret = ext4_init_inode_table(sb, group,
2782 elr->lr_timeout ? 0 : 1);
2783 if (elr->lr_timeout == 0) {
2784 timeout = (jiffies - timeout) *
2785 elr->lr_sbi->s_li_wait_mult;
2786 elr->lr_timeout = timeout;
2788 elr->lr_next_sched = jiffies + elr->lr_timeout;
2789 elr->lr_next_group = group + 1;
2797 * Remove lr_request from the list_request and free the
2798 * request structure. Should be called with li_list_mtx held
2800 static void ext4_remove_li_request(struct ext4_li_request *elr)
2802 struct ext4_sb_info *sbi;
2809 list_del(&elr->lr_request);
2810 sbi->s_li_request = NULL;
2814 static void ext4_unregister_li_request(struct super_block *sb)
2816 mutex_lock(&ext4_li_mtx);
2817 if (!ext4_li_info) {
2818 mutex_unlock(&ext4_li_mtx);
2822 mutex_lock(&ext4_li_info->li_list_mtx);
2823 ext4_remove_li_request(EXT4_SB(sb)->s_li_request);
2824 mutex_unlock(&ext4_li_info->li_list_mtx);
2825 mutex_unlock(&ext4_li_mtx);
2828 static struct task_struct *ext4_lazyinit_task;
2831 * This is the function where ext4lazyinit thread lives. It walks
2832 * through the request list searching for next scheduled filesystem.
2833 * When such a fs is found, run the lazy initialization request
2834 * (ext4_rn_li_request) and keep track of the time spend in this
2835 * function. Based on that time we compute next schedule time of
2836 * the request. When walking through the list is complete, compute
2837 * next waking time and put itself into sleep.
2839 static int ext4_lazyinit_thread(void *arg)
2841 struct ext4_lazy_init *eli = (struct ext4_lazy_init *)arg;
2842 struct list_head *pos, *n;
2843 struct ext4_li_request *elr;
2844 unsigned long next_wakeup, cur;
2846 BUG_ON(NULL == eli);
2850 next_wakeup = MAX_JIFFY_OFFSET;
2852 mutex_lock(&eli->li_list_mtx);
2853 if (list_empty(&eli->li_request_list)) {
2854 mutex_unlock(&eli->li_list_mtx);
2858 list_for_each_safe(pos, n, &eli->li_request_list) {
2859 elr = list_entry(pos, struct ext4_li_request,
2862 if (time_after_eq(jiffies, elr->lr_next_sched)) {
2863 if (ext4_run_li_request(elr) != 0) {
2864 /* error, remove the lazy_init job */
2865 ext4_remove_li_request(elr);
2870 if (time_before(elr->lr_next_sched, next_wakeup))
2871 next_wakeup = elr->lr_next_sched;
2873 mutex_unlock(&eli->li_list_mtx);
2878 if ((time_after_eq(cur, next_wakeup)) ||
2879 (MAX_JIFFY_OFFSET == next_wakeup)) {
2884 schedule_timeout_interruptible(next_wakeup - cur);
2886 if (kthread_should_stop()) {
2887 ext4_clear_request_list();
2894 * It looks like the request list is empty, but we need
2895 * to check it under the li_list_mtx lock, to prevent any
2896 * additions into it, and of course we should lock ext4_li_mtx
2897 * to atomically free the list and ext4_li_info, because at
2898 * this point another ext4 filesystem could be registering
2901 mutex_lock(&ext4_li_mtx);
2902 mutex_lock(&eli->li_list_mtx);
2903 if (!list_empty(&eli->li_request_list)) {
2904 mutex_unlock(&eli->li_list_mtx);
2905 mutex_unlock(&ext4_li_mtx);
2908 mutex_unlock(&eli->li_list_mtx);
2909 kfree(ext4_li_info);
2910 ext4_li_info = NULL;
2911 mutex_unlock(&ext4_li_mtx);
2916 static void ext4_clear_request_list(void)
2918 struct list_head *pos, *n;
2919 struct ext4_li_request *elr;
2921 mutex_lock(&ext4_li_info->li_list_mtx);
2922 list_for_each_safe(pos, n, &ext4_li_info->li_request_list) {
2923 elr = list_entry(pos, struct ext4_li_request,
2925 ext4_remove_li_request(elr);
2927 mutex_unlock(&ext4_li_info->li_list_mtx);
2930 static int ext4_run_lazyinit_thread(void)
2932 ext4_lazyinit_task = kthread_run(ext4_lazyinit_thread,
2933 ext4_li_info, "ext4lazyinit");
2934 if (IS_ERR(ext4_lazyinit_task)) {
2935 int err = PTR_ERR(ext4_lazyinit_task);
2936 ext4_clear_request_list();
2937 kfree(ext4_li_info);
2938 ext4_li_info = NULL;
2939 printk(KERN_CRIT "EXT4-fs: error %d creating inode table "
2940 "initialization thread\n",
2944 ext4_li_info->li_state |= EXT4_LAZYINIT_RUNNING;
2949 * Check whether it make sense to run itable init. thread or not.
2950 * If there is at least one uninitialized inode table, return
2951 * corresponding group number, else the loop goes through all
2952 * groups and return total number of groups.
2954 static ext4_group_t ext4_has_uninit_itable(struct super_block *sb)
2956 ext4_group_t group, ngroups = EXT4_SB(sb)->s_groups_count;
2957 struct ext4_group_desc *gdp = NULL;
2959 for (group = 0; group < ngroups; group++) {
2960 gdp = ext4_get_group_desc(sb, group, NULL);
2964 if (!(gdp->bg_flags & cpu_to_le16(EXT4_BG_INODE_ZEROED)))
2971 static int ext4_li_info_new(void)
2973 struct ext4_lazy_init *eli = NULL;
2975 eli = kzalloc(sizeof(*eli), GFP_KERNEL);
2979 INIT_LIST_HEAD(&eli->li_request_list);
2980 mutex_init(&eli->li_list_mtx);
2982 eli->li_state |= EXT4_LAZYINIT_QUIT;
2989 static struct ext4_li_request *ext4_li_request_new(struct super_block *sb,
2992 struct ext4_sb_info *sbi = EXT4_SB(sb);
2993 struct ext4_li_request *elr;
2996 elr = kzalloc(sizeof(*elr), GFP_KERNEL);
3002 elr->lr_next_group = start;
3005 * Randomize first schedule time of the request to
3006 * spread the inode table initialization requests
3009 get_random_bytes(&rnd, sizeof(rnd));
3010 elr->lr_next_sched = jiffies + (unsigned long)rnd %
3011 (EXT4_DEF_LI_MAX_START_DELAY * HZ);
3016 static int ext4_register_li_request(struct super_block *sb,
3017 ext4_group_t first_not_zeroed)
3019 struct ext4_sb_info *sbi = EXT4_SB(sb);
3020 struct ext4_li_request *elr;
3021 ext4_group_t ngroups = EXT4_SB(sb)->s_groups_count;
3024 if (sbi->s_li_request != NULL) {
3026 * Reset timeout so it can be computed again, because
3027 * s_li_wait_mult might have changed.
3029 sbi->s_li_request->lr_timeout = 0;
3033 if (first_not_zeroed == ngroups ||
3034 (sb->s_flags & MS_RDONLY) ||
3035 !test_opt(sb, INIT_INODE_TABLE))
3038 elr = ext4_li_request_new(sb, first_not_zeroed);
3042 mutex_lock(&ext4_li_mtx);
3044 if (NULL == ext4_li_info) {
3045 ret = ext4_li_info_new();
3050 mutex_lock(&ext4_li_info->li_list_mtx);
3051 list_add(&elr->lr_request, &ext4_li_info->li_request_list);
3052 mutex_unlock(&ext4_li_info->li_list_mtx);
3054 sbi->s_li_request = elr;
3056 * set elr to NULL here since it has been inserted to
3057 * the request_list and the removal and free of it is
3058 * handled by ext4_clear_request_list from now on.
3062 if (!(ext4_li_info->li_state & EXT4_LAZYINIT_RUNNING)) {
3063 ret = ext4_run_lazyinit_thread();
3068 mutex_unlock(&ext4_li_mtx);
3075 * We do not need to lock anything since this is called on
3078 static void ext4_destroy_lazyinit_thread(void)
3081 * If thread exited earlier
3082 * there's nothing to be done.
3084 if (!ext4_li_info || !ext4_lazyinit_task)
3087 kthread_stop(ext4_lazyinit_task);
3090 static int set_journal_csum_feature_set(struct super_block *sb)
3093 int compat, incompat;
3094 struct ext4_sb_info *sbi = EXT4_SB(sb);
3096 if (EXT4_HAS_RO_COMPAT_FEATURE(sb,
3097 EXT4_FEATURE_RO_COMPAT_METADATA_CSUM)) {
3098 /* journal checksum v2 */
3100 incompat = JBD2_FEATURE_INCOMPAT_CSUM_V2;
3102 /* journal checksum v1 */
3103 compat = JBD2_FEATURE_COMPAT_CHECKSUM;
3107 if (test_opt(sb, JOURNAL_ASYNC_COMMIT)) {
3108 ret = jbd2_journal_set_features(sbi->s_journal,
3110 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT |
3112 } else if (test_opt(sb, JOURNAL_CHECKSUM)) {
3113 ret = jbd2_journal_set_features(sbi->s_journal,
3116 jbd2_journal_clear_features(sbi->s_journal, 0, 0,
3117 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT);
3119 jbd2_journal_clear_features(sbi->s_journal,
3120 JBD2_FEATURE_COMPAT_CHECKSUM, 0,
3121 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT |
3122 JBD2_FEATURE_INCOMPAT_CSUM_V2);
3129 * Note: calculating the overhead so we can be compatible with
3130 * historical BSD practice is quite difficult in the face of
3131 * clusters/bigalloc. This is because multiple metadata blocks from
3132 * different block group can end up in the same allocation cluster.
3133 * Calculating the exact overhead in the face of clustered allocation
3134 * requires either O(all block bitmaps) in memory or O(number of block
3135 * groups**2) in time. We will still calculate the superblock for
3136 * older file systems --- and if we come across with a bigalloc file
3137 * system with zero in s_overhead_clusters the estimate will be close to
3138 * correct especially for very large cluster sizes --- but for newer
3139 * file systems, it's better to calculate this figure once at mkfs
3140 * time, and store it in the superblock. If the superblock value is
3141 * present (even for non-bigalloc file systems), we will use it.
3143 static int count_overhead(struct super_block *sb, ext4_group_t grp,
3146 struct ext4_sb_info *sbi = EXT4_SB(sb);
3147 struct ext4_group_desc *gdp;
3148 ext4_fsblk_t first_block, last_block, b;
3149 ext4_group_t i, ngroups = ext4_get_groups_count(sb);
3150 int s, j, count = 0;
3152 if (!EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_BIGALLOC))
3153 return (ext4_bg_has_super(sb, grp) + ext4_bg_num_gdb(sb, grp) +
3154 sbi->s_itb_per_group + 2);
3156 first_block = le32_to_cpu(sbi->s_es->s_first_data_block) +
3157 (grp * EXT4_BLOCKS_PER_GROUP(sb));
3158 last_block = first_block + EXT4_BLOCKS_PER_GROUP(sb) - 1;
3159 for (i = 0; i < ngroups; i++) {
3160 gdp = ext4_get_group_desc(sb, i, NULL);
3161 b = ext4_block_bitmap(sb, gdp);
3162 if (b >= first_block && b <= last_block) {
3163 ext4_set_bit(EXT4_B2C(sbi, b - first_block), buf);
3166 b = ext4_inode_bitmap(sb, gdp);
3167 if (b >= first_block && b <= last_block) {
3168 ext4_set_bit(EXT4_B2C(sbi, b - first_block), buf);
3171 b = ext4_inode_table(sb, gdp);
3172 if (b >= first_block && b + sbi->s_itb_per_group <= last_block)
3173 for (j = 0; j < sbi->s_itb_per_group; j++, b++) {
3174 int c = EXT4_B2C(sbi, b - first_block);
3175 ext4_set_bit(c, buf);
3181 if (ext4_bg_has_super(sb, grp)) {
3182 ext4_set_bit(s++, buf);
3185 for (j = ext4_bg_num_gdb(sb, grp); j > 0; j--) {
3186 ext4_set_bit(EXT4_B2C(sbi, s++), buf);
3192 return EXT4_CLUSTERS_PER_GROUP(sb) -
3193 ext4_count_free(buf, EXT4_CLUSTERS_PER_GROUP(sb) / 8);
3197 * Compute the overhead and stash it in sbi->s_overhead
3199 int ext4_calculate_overhead(struct super_block *sb)
3201 struct ext4_sb_info *sbi = EXT4_SB(sb);
3202 struct ext4_super_block *es = sbi->s_es;
3203 ext4_group_t i, ngroups = ext4_get_groups_count(sb);
3204 ext4_fsblk_t overhead = 0;
3205 char *buf = (char *) get_zeroed_page(GFP_KERNEL);
3207 memset(buf, 0, PAGE_SIZE);
3212 * Compute the overhead (FS structures). This is constant
3213 * for a given filesystem unless the number of block groups
3214 * changes so we cache the previous value until it does.
3218 * All of the blocks before first_data_block are overhead
3220 overhead = EXT4_B2C(sbi, le32_to_cpu(es->s_first_data_block));
3223 * Add the overhead found in each block group
3225 for (i = 0; i < ngroups; i++) {
3228 blks = count_overhead(sb, i, buf);
3231 memset(buf, 0, PAGE_SIZE);
3234 sbi->s_overhead = overhead;
3236 free_page((unsigned long) buf);
3240 static int ext4_fill_super(struct super_block *sb, void *data, int silent)
3242 char *orig_data = kstrdup(data, GFP_KERNEL);
3243 struct buffer_head *bh;
3244 struct ext4_super_block *es = NULL;
3245 struct ext4_sb_info *sbi;
3247 ext4_fsblk_t sb_block = get_sb_block(&data);
3248 ext4_fsblk_t logical_sb_block;
3249 unsigned long offset = 0;
3250 unsigned long journal_devnum = 0;
3251 unsigned long def_mount_opts;
3256 int blocksize, clustersize;
3257 unsigned int db_count;
3259 int needs_recovery, has_huge_files, has_bigalloc;
3262 unsigned int journal_ioprio = DEFAULT_JOURNAL_IOPRIO;
3263 ext4_group_t first_not_zeroed;
3265 sbi = kzalloc(sizeof(*sbi), GFP_KERNEL);
3269 sbi->s_blockgroup_lock =
3270 kzalloc(sizeof(struct blockgroup_lock), GFP_KERNEL);
3271 if (!sbi->s_blockgroup_lock) {
3275 sb->s_fs_info = sbi;
3277 sbi->s_mount_opt = 0;
3278 sbi->s_resuid = make_kuid(&init_user_ns, EXT4_DEF_RESUID);
3279 sbi->s_resgid = make_kgid(&init_user_ns, EXT4_DEF_RESGID);
3280 sbi->s_inode_readahead_blks = EXT4_DEF_INODE_READAHEAD_BLKS;
3281 sbi->s_sb_block = sb_block;
3282 if (sb->s_bdev->bd_part)
3283 sbi->s_sectors_written_start =
3284 part_stat_read(sb->s_bdev->bd_part, sectors[1]);
3286 /* Cleanup superblock name */
3287 for (cp = sb->s_id; (cp = strchr(cp, '/'));)
3291 blocksize = sb_min_blocksize(sb, EXT4_MIN_BLOCK_SIZE);
3293 ext4_msg(sb, KERN_ERR, "unable to set blocksize");
3298 * The ext4 superblock will not be buffer aligned for other than 1kB
3299 * block sizes. We need to calculate the offset from buffer start.
3301 if (blocksize != EXT4_MIN_BLOCK_SIZE) {
3302 logical_sb_block = sb_block * EXT4_MIN_BLOCK_SIZE;
3303 offset = do_div(logical_sb_block, blocksize);
3305 logical_sb_block = sb_block;
3308 if (!(bh = sb_bread(sb, logical_sb_block))) {
3309 ext4_msg(sb, KERN_ERR, "unable to read superblock");
3313 * Note: s_es must be initialized as soon as possible because
3314 * some ext4 macro-instructions depend on its value
3316 es = (struct ext4_super_block *) (bh->b_data + offset);
3318 sb->s_magic = le16_to_cpu(es->s_magic);
3319 if (sb->s_magic != EXT4_SUPER_MAGIC)
3321 sbi->s_kbytes_written = le64_to_cpu(es->s_kbytes_written);
3323 /* Warn if metadata_csum and gdt_csum are both set. */
3324 if (EXT4_HAS_RO_COMPAT_FEATURE(sb,
3325 EXT4_FEATURE_RO_COMPAT_METADATA_CSUM) &&
3326 EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_GDT_CSUM))
3327 ext4_warning(sb, KERN_INFO "metadata_csum and uninit_bg are "
3328 "redundant flags; please run fsck.");
3330 /* Check for a known checksum algorithm */
3331 if (!ext4_verify_csum_type(sb, es)) {
3332 ext4_msg(sb, KERN_ERR, "VFS: Found ext4 filesystem with "
3333 "unknown checksum algorithm.");
3338 /* Load the checksum driver */
3339 if (EXT4_HAS_RO_COMPAT_FEATURE(sb,
3340 EXT4_FEATURE_RO_COMPAT_METADATA_CSUM)) {
3341 sbi->s_chksum_driver = crypto_alloc_shash("crc32c", 0, 0);
3342 if (IS_ERR(sbi->s_chksum_driver)) {
3343 ext4_msg(sb, KERN_ERR, "Cannot load crc32c driver.");
3344 ret = PTR_ERR(sbi->s_chksum_driver);
3345 sbi->s_chksum_driver = NULL;
3350 /* Check superblock checksum */
3351 if (!ext4_superblock_csum_verify(sb, es)) {
3352 ext4_msg(sb, KERN_ERR, "VFS: Found ext4 filesystem with "
3353 "invalid superblock checksum. Run e2fsck?");
3358 /* Precompute checksum seed for all metadata */
3359 if (EXT4_HAS_RO_COMPAT_FEATURE(sb,
3360 EXT4_FEATURE_RO_COMPAT_METADATA_CSUM))
3361 sbi->s_csum_seed = ext4_chksum(sbi, ~0, es->s_uuid,
3362 sizeof(es->s_uuid));
3364 /* Set defaults before we parse the mount options */
3365 def_mount_opts = le32_to_cpu(es->s_default_mount_opts);
3366 set_opt(sb, INIT_INODE_TABLE);
3367 if (def_mount_opts & EXT4_DEFM_DEBUG)
3369 if (def_mount_opts & EXT4_DEFM_BSDGROUPS)
3371 if (def_mount_opts & EXT4_DEFM_UID16)
3372 set_opt(sb, NO_UID32);
3373 /* xattr user namespace & acls are now defaulted on */
3374 #ifdef CONFIG_EXT4_FS_XATTR
3375 set_opt(sb, XATTR_USER);
3377 #ifdef CONFIG_EXT4_FS_POSIX_ACL
3378 set_opt(sb, POSIX_ACL);
3380 set_opt(sb, MBLK_IO_SUBMIT);
3381 if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_DATA)
3382 set_opt(sb, JOURNAL_DATA);
3383 else if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_ORDERED)
3384 set_opt(sb, ORDERED_DATA);
3385 else if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_WBACK)
3386 set_opt(sb, WRITEBACK_DATA);
3388 if (le16_to_cpu(sbi->s_es->s_errors) == EXT4_ERRORS_PANIC)
3389 set_opt(sb, ERRORS_PANIC);
3390 else if (le16_to_cpu(sbi->s_es->s_errors) == EXT4_ERRORS_CONTINUE)
3391 set_opt(sb, ERRORS_CONT);
3393 set_opt(sb, ERRORS_RO);
3394 if (def_mount_opts & EXT4_DEFM_BLOCK_VALIDITY)
3395 set_opt(sb, BLOCK_VALIDITY);
3396 if (def_mount_opts & EXT4_DEFM_DISCARD)
3397 set_opt(sb, DISCARD);
3399 sbi->s_resuid = make_kuid(&init_user_ns, le16_to_cpu(es->s_def_resuid));
3400 sbi->s_resgid = make_kgid(&init_user_ns, le16_to_cpu(es->s_def_resgid));
3401 sbi->s_commit_interval = JBD2_DEFAULT_MAX_COMMIT_AGE * HZ;
3402 sbi->s_min_batch_time = EXT4_DEF_MIN_BATCH_TIME;
3403 sbi->s_max_batch_time = EXT4_DEF_MAX_BATCH_TIME;
3405 if ((def_mount_opts & EXT4_DEFM_NOBARRIER) == 0)
3406 set_opt(sb, BARRIER);
3409 * enable delayed allocation by default
3410 * Use -o nodelalloc to turn it off
3412 if (!IS_EXT3_SB(sb) &&
3413 ((def_mount_opts & EXT4_DEFM_NODELALLOC) == 0))
3414 set_opt(sb, DELALLOC);
3417 * set default s_li_wait_mult for lazyinit, for the case there is
3418 * no mount option specified.
3420 sbi->s_li_wait_mult = EXT4_DEF_LI_WAIT_MULT;
3422 if (!parse_options((char *) sbi->s_es->s_mount_opts, sb,
3423 &journal_devnum, &journal_ioprio, 0)) {
3424 ext4_msg(sb, KERN_WARNING,
3425 "failed to parse options in superblock: %s",
3426 sbi->s_es->s_mount_opts);
3428 sbi->s_def_mount_opt = sbi->s_mount_opt;
3429 if (!parse_options((char *) data, sb, &journal_devnum,
3430 &journal_ioprio, 0))
3433 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA) {
3434 printk_once(KERN_WARNING "EXT4-fs: Warning: mounting "
3435 "with data=journal disables delayed "
3436 "allocation and O_DIRECT support!\n");
3437 if (test_opt2(sb, EXPLICIT_DELALLOC)) {
3438 ext4_msg(sb, KERN_ERR, "can't mount with "
3439 "both data=journal and delalloc");
3442 if (test_opt(sb, DIOREAD_NOLOCK)) {
3443 ext4_msg(sb, KERN_ERR, "can't mount with "
3444 "both data=journal and delalloc");
3447 if (test_opt(sb, DELALLOC))
3448 clear_opt(sb, DELALLOC);
3451 blocksize = BLOCK_SIZE << le32_to_cpu(es->s_log_block_size);
3452 if (test_opt(sb, DIOREAD_NOLOCK)) {
3453 if (blocksize < PAGE_SIZE) {
3454 ext4_msg(sb, KERN_ERR, "can't mount with "
3455 "dioread_nolock if block size != PAGE_SIZE");
3460 sb->s_flags = (sb->s_flags & ~MS_POSIXACL) |
3461 (test_opt(sb, POSIX_ACL) ? MS_POSIXACL : 0);
3463 if (le32_to_cpu(es->s_rev_level) == EXT4_GOOD_OLD_REV &&
3464 (EXT4_HAS_COMPAT_FEATURE(sb, ~0U) ||
3465 EXT4_HAS_RO_COMPAT_FEATURE(sb, ~0U) ||
3466 EXT4_HAS_INCOMPAT_FEATURE(sb, ~0U)))
3467 ext4_msg(sb, KERN_WARNING,
3468 "feature flags set on rev 0 fs, "
3469 "running e2fsck is recommended");
3471 if (IS_EXT2_SB(sb)) {
3472 if (ext2_feature_set_ok(sb))
3473 ext4_msg(sb, KERN_INFO, "mounting ext2 file system "
3474 "using the ext4 subsystem");
3476 ext4_msg(sb, KERN_ERR, "couldn't mount as ext2 due "
3477 "to feature incompatibilities");
3482 if (IS_EXT3_SB(sb)) {
3483 if (ext3_feature_set_ok(sb))
3484 ext4_msg(sb, KERN_INFO, "mounting ext3 file system "
3485 "using the ext4 subsystem");
3487 ext4_msg(sb, KERN_ERR, "couldn't mount as ext3 due "
3488 "to feature incompatibilities");
3494 * Check feature flags regardless of the revision level, since we
3495 * previously didn't change the revision level when setting the flags,
3496 * so there is a chance incompat flags are set on a rev 0 filesystem.
3498 if (!ext4_feature_set_ok(sb, (sb->s_flags & MS_RDONLY)))
3501 if (blocksize < EXT4_MIN_BLOCK_SIZE ||
3502 blocksize > EXT4_MAX_BLOCK_SIZE) {
3503 ext4_msg(sb, KERN_ERR,
3504 "Unsupported filesystem blocksize %d", blocksize);
3508 if (sb->s_blocksize != blocksize) {
3509 /* Validate the filesystem blocksize */
3510 if (!sb_set_blocksize(sb, blocksize)) {
3511 ext4_msg(sb, KERN_ERR, "bad block size %d",
3517 logical_sb_block = sb_block * EXT4_MIN_BLOCK_SIZE;
3518 offset = do_div(logical_sb_block, blocksize);
3519 bh = sb_bread(sb, logical_sb_block);
3521 ext4_msg(sb, KERN_ERR,
3522 "Can't read superblock on 2nd try");
3525 es = (struct ext4_super_block *)(bh->b_data + offset);
3527 if (es->s_magic != cpu_to_le16(EXT4_SUPER_MAGIC)) {
3528 ext4_msg(sb, KERN_ERR,
3529 "Magic mismatch, very weird!");
3534 has_huge_files = EXT4_HAS_RO_COMPAT_FEATURE(sb,
3535 EXT4_FEATURE_RO_COMPAT_HUGE_FILE);
3536 sbi->s_bitmap_maxbytes = ext4_max_bitmap_size(sb->s_blocksize_bits,
3538 sb->s_maxbytes = ext4_max_size(sb->s_blocksize_bits, has_huge_files);
3540 if (le32_to_cpu(es->s_rev_level) == EXT4_GOOD_OLD_REV) {
3541 sbi->s_inode_size = EXT4_GOOD_OLD_INODE_SIZE;
3542 sbi->s_first_ino = EXT4_GOOD_OLD_FIRST_INO;
3544 sbi->s_inode_size = le16_to_cpu(es->s_inode_size);
3545 sbi->s_first_ino = le32_to_cpu(es->s_first_ino);
3546 if ((sbi->s_inode_size < EXT4_GOOD_OLD_INODE_SIZE) ||
3547 (!is_power_of_2(sbi->s_inode_size)) ||
3548 (sbi->s_inode_size > blocksize)) {
3549 ext4_msg(sb, KERN_ERR,
3550 "unsupported inode size: %d",
3554 if (sbi->s_inode_size > EXT4_GOOD_OLD_INODE_SIZE)
3555 sb->s_time_gran = 1 << (EXT4_EPOCH_BITS - 2);
3558 sbi->s_desc_size = le16_to_cpu(es->s_desc_size);
3559 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_64BIT)) {
3560 if (sbi->s_desc_size < EXT4_MIN_DESC_SIZE_64BIT ||
3561 sbi->s_desc_size > EXT4_MAX_DESC_SIZE ||
3562 !is_power_of_2(sbi->s_desc_size)) {
3563 ext4_msg(sb, KERN_ERR,
3564 "unsupported descriptor size %lu",
3569 sbi->s_desc_size = EXT4_MIN_DESC_SIZE;
3571 sbi->s_blocks_per_group = le32_to_cpu(es->s_blocks_per_group);
3572 sbi->s_inodes_per_group = le32_to_cpu(es->s_inodes_per_group);
3573 if (EXT4_INODE_SIZE(sb) == 0 || EXT4_INODES_PER_GROUP(sb) == 0)
3576 sbi->s_inodes_per_block = blocksize / EXT4_INODE_SIZE(sb);
3577 if (sbi->s_inodes_per_block == 0)
3579 sbi->s_itb_per_group = sbi->s_inodes_per_group /
3580 sbi->s_inodes_per_block;
3581 sbi->s_desc_per_block = blocksize / EXT4_DESC_SIZE(sb);
3583 sbi->s_mount_state = le16_to_cpu(es->s_state);
3584 sbi->s_addr_per_block_bits = ilog2(EXT4_ADDR_PER_BLOCK(sb));
3585 sbi->s_desc_per_block_bits = ilog2(EXT4_DESC_PER_BLOCK(sb));
3587 for (i = 0; i < 4; i++)
3588 sbi->s_hash_seed[i] = le32_to_cpu(es->s_hash_seed[i]);
3589 sbi->s_def_hash_version = es->s_def_hash_version;
3590 i = le32_to_cpu(es->s_flags);
3591 if (i & EXT2_FLAGS_UNSIGNED_HASH)
3592 sbi->s_hash_unsigned = 3;
3593 else if ((i & EXT2_FLAGS_SIGNED_HASH) == 0) {
3594 #ifdef __CHAR_UNSIGNED__
3595 es->s_flags |= cpu_to_le32(EXT2_FLAGS_UNSIGNED_HASH);
3596 sbi->s_hash_unsigned = 3;
3598 es->s_flags |= cpu_to_le32(EXT2_FLAGS_SIGNED_HASH);
3602 /* Handle clustersize */
3603 clustersize = BLOCK_SIZE << le32_to_cpu(es->s_log_cluster_size);
3604 has_bigalloc = EXT4_HAS_RO_COMPAT_FEATURE(sb,
3605 EXT4_FEATURE_RO_COMPAT_BIGALLOC);
3607 if (clustersize < blocksize) {
3608 ext4_msg(sb, KERN_ERR,
3609 "cluster size (%d) smaller than "
3610 "block size (%d)", clustersize, blocksize);
3613 sbi->s_cluster_bits = le32_to_cpu(es->s_log_cluster_size) -
3614 le32_to_cpu(es->s_log_block_size);
3615 sbi->s_clusters_per_group =
3616 le32_to_cpu(es->s_clusters_per_group);
3617 if (sbi->s_clusters_per_group > blocksize * 8) {
3618 ext4_msg(sb, KERN_ERR,
3619 "#clusters per group too big: %lu",
3620 sbi->s_clusters_per_group);
3623 if (sbi->s_blocks_per_group !=
3624 (sbi->s_clusters_per_group * (clustersize / blocksize))) {
3625 ext4_msg(sb, KERN_ERR, "blocks per group (%lu) and "
3626 "clusters per group (%lu) inconsistent",
3627 sbi->s_blocks_per_group,
3628 sbi->s_clusters_per_group);
3632 if (clustersize != blocksize) {
3633 ext4_warning(sb, "fragment/cluster size (%d) != "
3634 "block size (%d)", clustersize,
3636 clustersize = blocksize;
3638 if (sbi->s_blocks_per_group > blocksize * 8) {
3639 ext4_msg(sb, KERN_ERR,
3640 "#blocks per group too big: %lu",
3641 sbi->s_blocks_per_group);
3644 sbi->s_clusters_per_group = sbi->s_blocks_per_group;
3645 sbi->s_cluster_bits = 0;
3647 sbi->s_cluster_ratio = clustersize / blocksize;
3649 if (sbi->s_inodes_per_group > blocksize * 8) {
3650 ext4_msg(sb, KERN_ERR,
3651 "#inodes per group too big: %lu",
3652 sbi->s_inodes_per_group);
3657 * Test whether we have more sectors than will fit in sector_t,
3658 * and whether the max offset is addressable by the page cache.
3660 err = generic_check_addressable(sb->s_blocksize_bits,
3661 ext4_blocks_count(es));
3663 ext4_msg(sb, KERN_ERR, "filesystem"
3664 " too large to mount safely on this system");
3665 if (sizeof(sector_t) < 8)
3666 ext4_msg(sb, KERN_WARNING, "CONFIG_LBDAF not enabled");
3671 if (EXT4_BLOCKS_PER_GROUP(sb) == 0)
3674 /* check blocks count against device size */
3675 blocks_count = sb->s_bdev->bd_inode->i_size >> sb->s_blocksize_bits;
3676 if (blocks_count && ext4_blocks_count(es) > blocks_count) {
3677 ext4_msg(sb, KERN_WARNING, "bad geometry: block count %llu "
3678 "exceeds size of device (%llu blocks)",
3679 ext4_blocks_count(es), blocks_count);
3684 * It makes no sense for the first data block to be beyond the end
3685 * of the filesystem.
3687 if (le32_to_cpu(es->s_first_data_block) >= ext4_blocks_count(es)) {
3688 ext4_msg(sb, KERN_WARNING, "bad geometry: first data "
3689 "block %u is beyond end of filesystem (%llu)",
3690 le32_to_cpu(es->s_first_data_block),
3691 ext4_blocks_count(es));
3694 blocks_count = (ext4_blocks_count(es) -
3695 le32_to_cpu(es->s_first_data_block) +
3696 EXT4_BLOCKS_PER_GROUP(sb) - 1);
3697 do_div(blocks_count, EXT4_BLOCKS_PER_GROUP(sb));
3698 if (blocks_count > ((uint64_t)1<<32) - EXT4_DESC_PER_BLOCK(sb)) {
3699 ext4_msg(sb, KERN_WARNING, "groups count too large: %u "
3700 "(block count %llu, first data block %u, "
3701 "blocks per group %lu)", sbi->s_groups_count,
3702 ext4_blocks_count(es),
3703 le32_to_cpu(es->s_first_data_block),
3704 EXT4_BLOCKS_PER_GROUP(sb));
3707 sbi->s_groups_count = blocks_count;
3708 sbi->s_blockfile_groups = min_t(ext4_group_t, sbi->s_groups_count,
3709 (EXT4_MAX_BLOCK_FILE_PHYS / EXT4_BLOCKS_PER_GROUP(sb)));
3710 db_count = (sbi->s_groups_count + EXT4_DESC_PER_BLOCK(sb) - 1) /
3711 EXT4_DESC_PER_BLOCK(sb);
3712 sbi->s_group_desc = ext4_kvmalloc(db_count *
3713 sizeof(struct buffer_head *),
3715 if (sbi->s_group_desc == NULL) {
3716 ext4_msg(sb, KERN_ERR, "not enough memory");
3722 sbi->s_proc = proc_mkdir(sb->s_id, ext4_proc_root);
3725 proc_create_data("options", S_IRUGO, sbi->s_proc,
3726 &ext4_seq_options_fops, sb);
3728 bgl_lock_init(sbi->s_blockgroup_lock);
3730 for (i = 0; i < db_count; i++) {
3731 block = descriptor_loc(sb, logical_sb_block, i);
3732 sbi->s_group_desc[i] = sb_bread(sb, block);
3733 if (!sbi->s_group_desc[i]) {
3734 ext4_msg(sb, KERN_ERR,
3735 "can't read group descriptor %d", i);
3740 if (!ext4_check_descriptors(sb, &first_not_zeroed)) {
3741 ext4_msg(sb, KERN_ERR, "group descriptors corrupted!");
3744 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_FLEX_BG))
3745 if (!ext4_fill_flex_info(sb)) {
3746 ext4_msg(sb, KERN_ERR,
3747 "unable to initialize "
3748 "flex_bg meta info!");
3752 sbi->s_gdb_count = db_count;
3753 get_random_bytes(&sbi->s_next_generation, sizeof(u32));
3754 spin_lock_init(&sbi->s_next_gen_lock);
3756 init_timer(&sbi->s_err_report);
3757 sbi->s_err_report.function = print_daily_error_info;
3758 sbi->s_err_report.data = (unsigned long) sb;
3760 err = percpu_counter_init(&sbi->s_freeclusters_counter,
3761 ext4_count_free_clusters(sb));
3763 err = percpu_counter_init(&sbi->s_freeinodes_counter,
3764 ext4_count_free_inodes(sb));
3767 err = percpu_counter_init(&sbi->s_dirs_counter,
3768 ext4_count_dirs(sb));
3771 err = percpu_counter_init(&sbi->s_dirtyclusters_counter, 0);
3774 ext4_msg(sb, KERN_ERR, "insufficient memory");
3779 sbi->s_stripe = ext4_get_stripe_size(sbi);
3780 sbi->s_max_writeback_mb_bump = 128;
3781 sbi->s_extent_max_zeroout_kb = 32;
3784 * set up enough so that it can read an inode
3786 if (!test_opt(sb, NOLOAD) &&
3787 EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL))
3788 sb->s_op = &ext4_sops;
3790 sb->s_op = &ext4_nojournal_sops;
3791 sb->s_export_op = &ext4_export_ops;
3792 sb->s_xattr = ext4_xattr_handlers;
3794 sb->s_qcop = &ext4_qctl_operations;
3795 sb->dq_op = &ext4_quota_operations;
3797 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_QUOTA)) {
3798 /* Use qctl operations for hidden quota files. */
3799 sb->s_qcop = &ext4_qctl_sysfile_operations;
3802 memcpy(sb->s_uuid, es->s_uuid, sizeof(es->s_uuid));
3804 INIT_LIST_HEAD(&sbi->s_orphan); /* unlinked but open files */
3805 mutex_init(&sbi->s_orphan_lock);
3806 sbi->s_resize_flags = 0;
3810 needs_recovery = (es->s_last_orphan != 0 ||
3811 EXT4_HAS_INCOMPAT_FEATURE(sb,
3812 EXT4_FEATURE_INCOMPAT_RECOVER));
3814 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_MMP) &&
3815 !(sb->s_flags & MS_RDONLY))
3816 if (ext4_multi_mount_protect(sb, le64_to_cpu(es->s_mmp_block)))
3820 * The first inode we look at is the journal inode. Don't try
3821 * root first: it may be modified in the journal!
3823 if (!test_opt(sb, NOLOAD) &&
3824 EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL)) {
3825 if (ext4_load_journal(sb, es, journal_devnum))
3827 } else if (test_opt(sb, NOLOAD) && !(sb->s_flags & MS_RDONLY) &&
3828 EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER)) {
3829 ext4_msg(sb, KERN_ERR, "required journal recovery "
3830 "suppressed and not mounted read-only");
3831 goto failed_mount_wq;
3833 clear_opt(sb, DATA_FLAGS);
3834 sbi->s_journal = NULL;
3839 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_64BIT) &&
3840 !jbd2_journal_set_features(EXT4_SB(sb)->s_journal, 0, 0,
3841 JBD2_FEATURE_INCOMPAT_64BIT)) {
3842 ext4_msg(sb, KERN_ERR, "Failed to set 64-bit journal feature");
3843 goto failed_mount_wq;
3846 if (!set_journal_csum_feature_set(sb)) {
3847 ext4_msg(sb, KERN_ERR, "Failed to set journal checksum "
3849 goto failed_mount_wq;
3852 /* We have now updated the journal if required, so we can
3853 * validate the data journaling mode. */
3854 switch (test_opt(sb, DATA_FLAGS)) {
3856 /* No mode set, assume a default based on the journal
3857 * capabilities: ORDERED_DATA if the journal can
3858 * cope, else JOURNAL_DATA
3860 if (jbd2_journal_check_available_features
3861 (sbi->s_journal, 0, 0, JBD2_FEATURE_INCOMPAT_REVOKE))
3862 set_opt(sb, ORDERED_DATA);
3864 set_opt(sb, JOURNAL_DATA);
3867 case EXT4_MOUNT_ORDERED_DATA:
3868 case EXT4_MOUNT_WRITEBACK_DATA:
3869 if (!jbd2_journal_check_available_features
3870 (sbi->s_journal, 0, 0, JBD2_FEATURE_INCOMPAT_REVOKE)) {
3871 ext4_msg(sb, KERN_ERR, "Journal does not support "
3872 "requested data journaling mode");
3873 goto failed_mount_wq;
3878 set_task_ioprio(sbi->s_journal->j_task, journal_ioprio);
3880 sbi->s_journal->j_commit_callback = ext4_journal_commit_callback;
3883 * The journal may have updated the bg summary counts, so we
3884 * need to update the global counters.
3886 percpu_counter_set(&sbi->s_freeclusters_counter,
3887 ext4_count_free_clusters(sb));
3888 percpu_counter_set(&sbi->s_freeinodes_counter,
3889 ext4_count_free_inodes(sb));
3890 percpu_counter_set(&sbi->s_dirs_counter,
3891 ext4_count_dirs(sb));
3892 percpu_counter_set(&sbi->s_dirtyclusters_counter, 0);
3896 * Get the # of file system overhead blocks from the
3897 * superblock if present.
3899 if (es->s_overhead_clusters)
3900 sbi->s_overhead = le32_to_cpu(es->s_overhead_clusters);
3902 ret = ext4_calculate_overhead(sb);
3904 goto failed_mount_wq;
3908 * The maximum number of concurrent works can be high and
3909 * concurrency isn't really necessary. Limit it to 1.
3911 EXT4_SB(sb)->dio_unwritten_wq =
3912 alloc_workqueue("ext4-dio-unwritten", WQ_MEM_RECLAIM | WQ_UNBOUND, 1);
3913 if (!EXT4_SB(sb)->dio_unwritten_wq) {
3914 printk(KERN_ERR "EXT4-fs: failed to create DIO workqueue\n");
3915 goto failed_mount_wq;
3919 * The jbd2_journal_load will have done any necessary log recovery,
3920 * so we can safely mount the rest of the filesystem now.
3923 root = ext4_iget(sb, EXT4_ROOT_INO);
3925 ext4_msg(sb, KERN_ERR, "get root inode failed");
3926 ret = PTR_ERR(root);
3930 if (!S_ISDIR(root->i_mode) || !root->i_blocks || !root->i_size) {
3931 ext4_msg(sb, KERN_ERR, "corrupt root inode, run e2fsck");
3935 sb->s_root = d_make_root(root);
3937 ext4_msg(sb, KERN_ERR, "get root dentry failed");
3942 if (ext4_setup_super(sb, es, sb->s_flags & MS_RDONLY))
3943 sb->s_flags |= MS_RDONLY;
3945 /* determine the minimum size of new large inodes, if present */
3946 if (sbi->s_inode_size > EXT4_GOOD_OLD_INODE_SIZE) {
3947 sbi->s_want_extra_isize = sizeof(struct ext4_inode) -
3948 EXT4_GOOD_OLD_INODE_SIZE;
3949 if (EXT4_HAS_RO_COMPAT_FEATURE(sb,
3950 EXT4_FEATURE_RO_COMPAT_EXTRA_ISIZE)) {
3951 if (sbi->s_want_extra_isize <
3952 le16_to_cpu(es->s_want_extra_isize))
3953 sbi->s_want_extra_isize =
3954 le16_to_cpu(es->s_want_extra_isize);
3955 if (sbi->s_want_extra_isize <
3956 le16_to_cpu(es->s_min_extra_isize))
3957 sbi->s_want_extra_isize =
3958 le16_to_cpu(es->s_min_extra_isize);
3961 /* Check if enough inode space is available */
3962 if (EXT4_GOOD_OLD_INODE_SIZE + sbi->s_want_extra_isize >
3963 sbi->s_inode_size) {
3964 sbi->s_want_extra_isize = sizeof(struct ext4_inode) -
3965 EXT4_GOOD_OLD_INODE_SIZE;
3966 ext4_msg(sb, KERN_INFO, "required extra inode space not"
3970 err = ext4_setup_system_zone(sb);
3972 ext4_msg(sb, KERN_ERR, "failed to initialize system "
3974 goto failed_mount4a;
3978 err = ext4_mb_init(sb);
3980 ext4_msg(sb, KERN_ERR, "failed to initialize mballoc (%d)",
3985 err = ext4_register_li_request(sb, first_not_zeroed);
3989 sbi->s_kobj.kset = ext4_kset;
3990 init_completion(&sbi->s_kobj_unregister);
3991 err = kobject_init_and_add(&sbi->s_kobj, &ext4_ktype, NULL,
3996 EXT4_SB(sb)->s_mount_state |= EXT4_ORPHAN_FS;
3997 ext4_orphan_cleanup(sb, es);
3998 EXT4_SB(sb)->s_mount_state &= ~EXT4_ORPHAN_FS;
3999 if (needs_recovery) {
4000 ext4_msg(sb, KERN_INFO, "recovery complete");
4001 ext4_mark_recovery_complete(sb, es);
4003 if (EXT4_SB(sb)->s_journal) {
4004 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA)
4005 descr = " journalled data mode";
4006 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA)
4007 descr = " ordered data mode";
4009 descr = " writeback data mode";
4011 descr = "out journal";
4014 /* Enable quota usage during mount. */
4015 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_QUOTA) &&
4016 !(sb->s_flags & MS_RDONLY)) {
4017 ret = ext4_enable_quotas(sb);
4021 #endif /* CONFIG_QUOTA */
4023 ext4_msg(sb, KERN_INFO, "mounted filesystem with%s. "
4024 "Opts: %s%s%s", descr, sbi->s_es->s_mount_opts,
4025 *sbi->s_es->s_mount_opts ? "; " : "", orig_data);
4027 if (es->s_error_count)
4028 mod_timer(&sbi->s_err_report, jiffies + 300*HZ); /* 5 minutes */
4035 ext4_msg(sb, KERN_ERR, "VFS: Can't find ext4 filesystem");
4039 ext4_unregister_li_request(sb);
4041 ext4_mb_release(sb);
4043 ext4_ext_release(sb);
4044 ext4_release_system_zone(sb);
4049 ext4_msg(sb, KERN_ERR, "mount failed");
4050 destroy_workqueue(EXT4_SB(sb)->dio_unwritten_wq);
4052 if (sbi->s_journal) {
4053 jbd2_journal_destroy(sbi->s_journal);
4054 sbi->s_journal = NULL;
4057 del_timer(&sbi->s_err_report);
4058 if (sbi->s_flex_groups)
4059 ext4_kvfree(sbi->s_flex_groups);
4060 percpu_counter_destroy(&sbi->s_freeclusters_counter);
4061 percpu_counter_destroy(&sbi->s_freeinodes_counter);
4062 percpu_counter_destroy(&sbi->s_dirs_counter);
4063 percpu_counter_destroy(&sbi->s_dirtyclusters_counter);
4065 kthread_stop(sbi->s_mmp_tsk);
4067 for (i = 0; i < db_count; i++)
4068 brelse(sbi->s_group_desc[i]);
4069 ext4_kvfree(sbi->s_group_desc);
4071 if (sbi->s_chksum_driver)
4072 crypto_free_shash(sbi->s_chksum_driver);
4074 remove_proc_entry("options", sbi->s_proc);
4075 remove_proc_entry(sb->s_id, ext4_proc_root);
4078 for (i = 0; i < MAXQUOTAS; i++)
4079 kfree(sbi->s_qf_names[i]);
4081 ext4_blkdev_remove(sbi);
4084 sb->s_fs_info = NULL;
4085 kfree(sbi->s_blockgroup_lock);
4093 * Setup any per-fs journal parameters now. We'll do this both on
4094 * initial mount, once the journal has been initialised but before we've
4095 * done any recovery; and again on any subsequent remount.
4097 static void ext4_init_journal_params(struct super_block *sb, journal_t *journal)
4099 struct ext4_sb_info *sbi = EXT4_SB(sb);
4101 journal->j_commit_interval = sbi->s_commit_interval;
4102 journal->j_min_batch_time = sbi->s_min_batch_time;
4103 journal->j_max_batch_time = sbi->s_max_batch_time;
4105 write_lock(&journal->j_state_lock);
4106 if (test_opt(sb, BARRIER))
4107 journal->j_flags |= JBD2_BARRIER;
4109 journal->j_flags &= ~JBD2_BARRIER;
4110 if (test_opt(sb, DATA_ERR_ABORT))
4111 journal->j_flags |= JBD2_ABORT_ON_SYNCDATA_ERR;
4113 journal->j_flags &= ~JBD2_ABORT_ON_SYNCDATA_ERR;
4114 write_unlock(&journal->j_state_lock);
4117 static journal_t *ext4_get_journal(struct super_block *sb,
4118 unsigned int journal_inum)
4120 struct inode *journal_inode;
4123 BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL));
4125 /* First, test for the existence of a valid inode on disk. Bad
4126 * things happen if we iget() an unused inode, as the subsequent
4127 * iput() will try to delete it. */
4129 journal_inode = ext4_iget(sb, journal_inum);
4130 if (IS_ERR(journal_inode)) {
4131 ext4_msg(sb, KERN_ERR, "no journal found");
4134 if (!journal_inode->i_nlink) {
4135 make_bad_inode(journal_inode);
4136 iput(journal_inode);
4137 ext4_msg(sb, KERN_ERR, "journal inode is deleted");
4141 jbd_debug(2, "Journal inode found at %p: %lld bytes\n",
4142 journal_inode, journal_inode->i_size);
4143 if (!S_ISREG(journal_inode->i_mode)) {
4144 ext4_msg(sb, KERN_ERR, "invalid journal inode");
4145 iput(journal_inode);
4149 journal = jbd2_journal_init_inode(journal_inode);
4151 ext4_msg(sb, KERN_ERR, "Could not load journal inode");
4152 iput(journal_inode);
4155 journal->j_private = sb;
4156 ext4_init_journal_params(sb, journal);
4160 static journal_t *ext4_get_dev_journal(struct super_block *sb,
4163 struct buffer_head *bh;
4167 int hblock, blocksize;
4168 ext4_fsblk_t sb_block;
4169 unsigned long offset;
4170 struct ext4_super_block *es;
4171 struct block_device *bdev;
4173 BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL));
4175 bdev = ext4_blkdev_get(j_dev, sb);
4179 blocksize = sb->s_blocksize;
4180 hblock = bdev_logical_block_size(bdev);
4181 if (blocksize < hblock) {
4182 ext4_msg(sb, KERN_ERR,
4183 "blocksize too small for journal device");
4187 sb_block = EXT4_MIN_BLOCK_SIZE / blocksize;
4188 offset = EXT4_MIN_BLOCK_SIZE % blocksize;
4189 set_blocksize(bdev, blocksize);
4190 if (!(bh = __bread(bdev, sb_block, blocksize))) {
4191 ext4_msg(sb, KERN_ERR, "couldn't read superblock of "
4192 "external journal");
4196 es = (struct ext4_super_block *) (bh->b_data + offset);
4197 if ((le16_to_cpu(es->s_magic) != EXT4_SUPER_MAGIC) ||
4198 !(le32_to_cpu(es->s_feature_incompat) &
4199 EXT4_FEATURE_INCOMPAT_JOURNAL_DEV)) {
4200 ext4_msg(sb, KERN_ERR, "external journal has "
4206 if (memcmp(EXT4_SB(sb)->s_es->s_journal_uuid, es->s_uuid, 16)) {
4207 ext4_msg(sb, KERN_ERR, "journal UUID does not match");
4212 len = ext4_blocks_count(es);
4213 start = sb_block + 1;
4214 brelse(bh); /* we're done with the superblock */
4216 journal = jbd2_journal_init_dev(bdev, sb->s_bdev,
4217 start, len, blocksize);
4219 ext4_msg(sb, KERN_ERR, "failed to create device journal");
4222 journal->j_private = sb;
4223 ll_rw_block(READ, 1, &journal->j_sb_buffer);
4224 wait_on_buffer(journal->j_sb_buffer);
4225 if (!buffer_uptodate(journal->j_sb_buffer)) {
4226 ext4_msg(sb, KERN_ERR, "I/O error on journal device");
4229 if (be32_to_cpu(journal->j_superblock->s_nr_users) != 1) {
4230 ext4_msg(sb, KERN_ERR, "External journal has more than one "
4231 "user (unsupported) - %d",
4232 be32_to_cpu(journal->j_superblock->s_nr_users));
4235 EXT4_SB(sb)->journal_bdev = bdev;
4236 ext4_init_journal_params(sb, journal);
4240 jbd2_journal_destroy(journal);
4242 ext4_blkdev_put(bdev);
4246 static int ext4_load_journal(struct super_block *sb,
4247 struct ext4_super_block *es,
4248 unsigned long journal_devnum)
4251 unsigned int journal_inum = le32_to_cpu(es->s_journal_inum);
4254 int really_read_only;
4256 BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL));
4258 if (journal_devnum &&
4259 journal_devnum != le32_to_cpu(es->s_journal_dev)) {
4260 ext4_msg(sb, KERN_INFO, "external journal device major/minor "
4261 "numbers have changed");
4262 journal_dev = new_decode_dev(journal_devnum);
4264 journal_dev = new_decode_dev(le32_to_cpu(es->s_journal_dev));
4266 really_read_only = bdev_read_only(sb->s_bdev);
4269 * Are we loading a blank journal or performing recovery after a
4270 * crash? For recovery, we need to check in advance whether we
4271 * can get read-write access to the device.
4273 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER)) {
4274 if (sb->s_flags & MS_RDONLY) {
4275 ext4_msg(sb, KERN_INFO, "INFO: recovery "
4276 "required on readonly filesystem");
4277 if (really_read_only) {
4278 ext4_msg(sb, KERN_ERR, "write access "
4279 "unavailable, cannot proceed");
4282 ext4_msg(sb, KERN_INFO, "write access will "
4283 "be enabled during recovery");
4287 if (journal_inum && journal_dev) {
4288 ext4_msg(sb, KERN_ERR, "filesystem has both journal "
4289 "and inode journals!");
4294 if (!(journal = ext4_get_journal(sb, journal_inum)))
4297 if (!(journal = ext4_get_dev_journal(sb, journal_dev)))
4301 if (!(journal->j_flags & JBD2_BARRIER))
4302 ext4_msg(sb, KERN_INFO, "barriers disabled");
4304 if (!EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER))
4305 err = jbd2_journal_wipe(journal, !really_read_only);
4307 char *save = kmalloc(EXT4_S_ERR_LEN, GFP_KERNEL);
4309 memcpy(save, ((char *) es) +
4310 EXT4_S_ERR_START, EXT4_S_ERR_LEN);
4311 err = jbd2_journal_load(journal);
4313 memcpy(((char *) es) + EXT4_S_ERR_START,
4314 save, EXT4_S_ERR_LEN);
4319 ext4_msg(sb, KERN_ERR, "error loading journal");
4320 jbd2_journal_destroy(journal);
4324 EXT4_SB(sb)->s_journal = journal;
4325 ext4_clear_journal_err(sb, es);
4327 if (!really_read_only && journal_devnum &&
4328 journal_devnum != le32_to_cpu(es->s_journal_dev)) {
4329 es->s_journal_dev = cpu_to_le32(journal_devnum);
4331 /* Make sure we flush the recovery flag to disk. */
4332 ext4_commit_super(sb, 1);
4338 static int ext4_commit_super(struct super_block *sb, int sync)
4340 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
4341 struct buffer_head *sbh = EXT4_SB(sb)->s_sbh;
4344 if (!sbh || block_device_ejected(sb))
4346 if (buffer_write_io_error(sbh)) {
4348 * Oh, dear. A previous attempt to write the
4349 * superblock failed. This could happen because the
4350 * USB device was yanked out. Or it could happen to
4351 * be a transient write error and maybe the block will
4352 * be remapped. Nothing we can do but to retry the
4353 * write and hope for the best.
4355 ext4_msg(sb, KERN_ERR, "previous I/O error to "
4356 "superblock detected");
4357 clear_buffer_write_io_error(sbh);
4358 set_buffer_uptodate(sbh);
4361 * If the file system is mounted read-only, don't update the
4362 * superblock write time. This avoids updating the superblock
4363 * write time when we are mounting the root file system
4364 * read/only but we need to replay the journal; at that point,
4365 * for people who are east of GMT and who make their clock
4366 * tick in localtime for Windows bug-for-bug compatibility,
4367 * the clock is set in the future, and this will cause e2fsck
4368 * to complain and force a full file system check.
4370 if (!(sb->s_flags & MS_RDONLY))
4371 es->s_wtime = cpu_to_le32(get_seconds());
4372 if (sb->s_bdev->bd_part)
4373 es->s_kbytes_written =
4374 cpu_to_le64(EXT4_SB(sb)->s_kbytes_written +
4375 ((part_stat_read(sb->s_bdev->bd_part, sectors[1]) -
4376 EXT4_SB(sb)->s_sectors_written_start) >> 1));
4378 es->s_kbytes_written =
4379 cpu_to_le64(EXT4_SB(sb)->s_kbytes_written);
4380 ext4_free_blocks_count_set(es,
4381 EXT4_C2B(EXT4_SB(sb), percpu_counter_sum_positive(
4382 &EXT4_SB(sb)->s_freeclusters_counter)));
4383 es->s_free_inodes_count =
4384 cpu_to_le32(percpu_counter_sum_positive(
4385 &EXT4_SB(sb)->s_freeinodes_counter));
4386 BUFFER_TRACE(sbh, "marking dirty");
4387 ext4_superblock_csum_set(sb, es);
4388 mark_buffer_dirty(sbh);
4390 error = sync_dirty_buffer(sbh);
4394 error = buffer_write_io_error(sbh);
4396 ext4_msg(sb, KERN_ERR, "I/O error while writing "
4398 clear_buffer_write_io_error(sbh);
4399 set_buffer_uptodate(sbh);
4406 * Have we just finished recovery? If so, and if we are mounting (or
4407 * remounting) the filesystem readonly, then we will end up with a
4408 * consistent fs on disk. Record that fact.
4410 static void ext4_mark_recovery_complete(struct super_block *sb,
4411 struct ext4_super_block *es)
4413 journal_t *journal = EXT4_SB(sb)->s_journal;
4415 if (!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL)) {
4416 BUG_ON(journal != NULL);
4419 jbd2_journal_lock_updates(journal);
4420 if (jbd2_journal_flush(journal) < 0)
4423 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER) &&
4424 sb->s_flags & MS_RDONLY) {
4425 EXT4_CLEAR_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
4426 ext4_commit_super(sb, 1);
4430 jbd2_journal_unlock_updates(journal);
4434 * If we are mounting (or read-write remounting) a filesystem whose journal
4435 * has recorded an error from a previous lifetime, move that error to the
4436 * main filesystem now.
4438 static void ext4_clear_journal_err(struct super_block *sb,
4439 struct ext4_super_block *es)
4445 BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL));
4447 journal = EXT4_SB(sb)->s_journal;
4450 * Now check for any error status which may have been recorded in the
4451 * journal by a prior ext4_error() or ext4_abort()
4454 j_errno = jbd2_journal_errno(journal);
4458 errstr = ext4_decode_error(sb, j_errno, nbuf);
4459 ext4_warning(sb, "Filesystem error recorded "
4460 "from previous mount: %s", errstr);
4461 ext4_warning(sb, "Marking fs in need of filesystem check.");
4463 EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS;
4464 es->s_state |= cpu_to_le16(EXT4_ERROR_FS);
4465 ext4_commit_super(sb, 1);
4467 jbd2_journal_clear_err(journal);
4468 jbd2_journal_update_sb_errno(journal);
4473 * Force the running and committing transactions to commit,
4474 * and wait on the commit.
4476 int ext4_force_commit(struct super_block *sb)
4481 if (sb->s_flags & MS_RDONLY)
4484 journal = EXT4_SB(sb)->s_journal;
4486 ret = ext4_journal_force_commit(journal);
4491 static int ext4_sync_fs(struct super_block *sb, int wait)
4495 struct ext4_sb_info *sbi = EXT4_SB(sb);
4497 trace_ext4_sync_fs(sb, wait);
4498 flush_workqueue(sbi->dio_unwritten_wq);
4500 * Writeback quota in non-journalled quota case - journalled quota has
4503 dquot_writeback_dquots(sb, -1);
4504 if (jbd2_journal_start_commit(sbi->s_journal, &target)) {
4506 jbd2_log_wait_commit(sbi->s_journal, target);
4512 * LVM calls this function before a (read-only) snapshot is created. This
4513 * gives us a chance to flush the journal completely and mark the fs clean.
4515 * Note that only this function cannot bring a filesystem to be in a clean
4516 * state independently. It relies on upper layer to stop all data & metadata
4519 static int ext4_freeze(struct super_block *sb)
4524 if (sb->s_flags & MS_RDONLY)
4527 journal = EXT4_SB(sb)->s_journal;
4529 /* Now we set up the journal barrier. */
4530 jbd2_journal_lock_updates(journal);
4533 * Don't clear the needs_recovery flag if we failed to flush
4536 error = jbd2_journal_flush(journal);
4540 /* Journal blocked and flushed, clear needs_recovery flag. */
4541 EXT4_CLEAR_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
4542 error = ext4_commit_super(sb, 1);
4544 /* we rely on upper layer to stop further updates */
4545 jbd2_journal_unlock_updates(EXT4_SB(sb)->s_journal);
4550 * Called by LVM after the snapshot is done. We need to reset the RECOVER
4551 * flag here, even though the filesystem is not technically dirty yet.
4553 static int ext4_unfreeze(struct super_block *sb)
4555 if (sb->s_flags & MS_RDONLY)
4558 /* Reset the needs_recovery flag before the fs is unlocked. */
4559 EXT4_SET_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
4560 ext4_commit_super(sb, 1);
4565 * Structure to save mount options for ext4_remount's benefit
4567 struct ext4_mount_options {
4568 unsigned long s_mount_opt;
4569 unsigned long s_mount_opt2;
4572 unsigned long s_commit_interval;
4573 u32 s_min_batch_time, s_max_batch_time;
4576 char *s_qf_names[MAXQUOTAS];
4580 static int ext4_remount(struct super_block *sb, int *flags, char *data)
4582 struct ext4_super_block *es;
4583 struct ext4_sb_info *sbi = EXT4_SB(sb);
4584 unsigned long old_sb_flags;
4585 struct ext4_mount_options old_opts;
4586 int enable_quota = 0;
4588 unsigned int journal_ioprio = DEFAULT_JOURNAL_IOPRIO;
4593 char *orig_data = kstrdup(data, GFP_KERNEL);
4595 /* Store the original options */
4596 old_sb_flags = sb->s_flags;
4597 old_opts.s_mount_opt = sbi->s_mount_opt;
4598 old_opts.s_mount_opt2 = sbi->s_mount_opt2;
4599 old_opts.s_resuid = sbi->s_resuid;
4600 old_opts.s_resgid = sbi->s_resgid;
4601 old_opts.s_commit_interval = sbi->s_commit_interval;
4602 old_opts.s_min_batch_time = sbi->s_min_batch_time;
4603 old_opts.s_max_batch_time = sbi->s_max_batch_time;
4605 old_opts.s_jquota_fmt = sbi->s_jquota_fmt;
4606 for (i = 0; i < MAXQUOTAS; i++)
4607 old_opts.s_qf_names[i] = sbi->s_qf_names[i];
4609 if (sbi->s_journal && sbi->s_journal->j_task->io_context)
4610 journal_ioprio = sbi->s_journal->j_task->io_context->ioprio;
4613 * Allow the "check" option to be passed as a remount option.
4615 if (!parse_options(data, sb, NULL, &journal_ioprio, 1)) {
4620 if (sbi->s_mount_flags & EXT4_MF_FS_ABORTED)
4621 ext4_abort(sb, "Abort forced by user");
4623 sb->s_flags = (sb->s_flags & ~MS_POSIXACL) |
4624 (test_opt(sb, POSIX_ACL) ? MS_POSIXACL : 0);
4628 if (sbi->s_journal) {
4629 ext4_init_journal_params(sb, sbi->s_journal);
4630 set_task_ioprio(sbi->s_journal->j_task, journal_ioprio);
4633 if ((*flags & MS_RDONLY) != (sb->s_flags & MS_RDONLY)) {
4634 if (sbi->s_mount_flags & EXT4_MF_FS_ABORTED) {
4639 if (*flags & MS_RDONLY) {
4640 err = dquot_suspend(sb, -1);
4645 * First of all, the unconditional stuff we have to do
4646 * to disable replay of the journal when we next remount
4648 sb->s_flags |= MS_RDONLY;
4651 * OK, test if we are remounting a valid rw partition
4652 * readonly, and if so set the rdonly flag and then
4653 * mark the partition as valid again.
4655 if (!(es->s_state & cpu_to_le16(EXT4_VALID_FS)) &&
4656 (sbi->s_mount_state & EXT4_VALID_FS))
4657 es->s_state = cpu_to_le16(sbi->s_mount_state);
4660 ext4_mark_recovery_complete(sb, es);
4662 /* Make sure we can mount this feature set readwrite */
4663 if (!ext4_feature_set_ok(sb, 0)) {
4668 * Make sure the group descriptor checksums
4669 * are sane. If they aren't, refuse to remount r/w.
4671 for (g = 0; g < sbi->s_groups_count; g++) {
4672 struct ext4_group_desc *gdp =
4673 ext4_get_group_desc(sb, g, NULL);
4675 if (!ext4_group_desc_csum_verify(sb, g, gdp)) {
4676 ext4_msg(sb, KERN_ERR,
4677 "ext4_remount: Checksum for group %u failed (%u!=%u)",
4678 g, le16_to_cpu(ext4_group_desc_csum(sbi, g, gdp)),
4679 le16_to_cpu(gdp->bg_checksum));
4686 * If we have an unprocessed orphan list hanging
4687 * around from a previously readonly bdev mount,
4688 * require a full umount/remount for now.
4690 if (es->s_last_orphan) {
4691 ext4_msg(sb, KERN_WARNING, "Couldn't "
4692 "remount RDWR because of unprocessed "
4693 "orphan inode list. Please "
4694 "umount/remount instead");
4700 * Mounting a RDONLY partition read-write, so reread
4701 * and store the current valid flag. (It may have
4702 * been changed by e2fsck since we originally mounted
4706 ext4_clear_journal_err(sb, es);
4707 sbi->s_mount_state = le16_to_cpu(es->s_state);
4708 if (!ext4_setup_super(sb, es, 0))
4709 sb->s_flags &= ~MS_RDONLY;
4710 if (EXT4_HAS_INCOMPAT_FEATURE(sb,
4711 EXT4_FEATURE_INCOMPAT_MMP))
4712 if (ext4_multi_mount_protect(sb,
4713 le64_to_cpu(es->s_mmp_block))) {
4722 * Reinitialize lazy itable initialization thread based on
4725 if ((sb->s_flags & MS_RDONLY) || !test_opt(sb, INIT_INODE_TABLE))
4726 ext4_unregister_li_request(sb);
4728 ext4_group_t first_not_zeroed;
4729 first_not_zeroed = ext4_has_uninit_itable(sb);
4730 ext4_register_li_request(sb, first_not_zeroed);
4733 ext4_setup_system_zone(sb);
4734 if (sbi->s_journal == NULL)
4735 ext4_commit_super(sb, 1);
4738 /* Release old quota file names */
4739 for (i = 0; i < MAXQUOTAS; i++)
4740 if (old_opts.s_qf_names[i] &&
4741 old_opts.s_qf_names[i] != sbi->s_qf_names[i])
4742 kfree(old_opts.s_qf_names[i]);
4744 if (sb_any_quota_suspended(sb))
4745 dquot_resume(sb, -1);
4746 else if (EXT4_HAS_RO_COMPAT_FEATURE(sb,
4747 EXT4_FEATURE_RO_COMPAT_QUOTA)) {
4748 err = ext4_enable_quotas(sb);
4755 ext4_msg(sb, KERN_INFO, "re-mounted. Opts: %s", orig_data);
4760 sb->s_flags = old_sb_flags;
4761 sbi->s_mount_opt = old_opts.s_mount_opt;
4762 sbi->s_mount_opt2 = old_opts.s_mount_opt2;
4763 sbi->s_resuid = old_opts.s_resuid;
4764 sbi->s_resgid = old_opts.s_resgid;
4765 sbi->s_commit_interval = old_opts.s_commit_interval;
4766 sbi->s_min_batch_time = old_opts.s_min_batch_time;
4767 sbi->s_max_batch_time = old_opts.s_max_batch_time;
4769 sbi->s_jquota_fmt = old_opts.s_jquota_fmt;
4770 for (i = 0; i < MAXQUOTAS; i++) {
4771 if (sbi->s_qf_names[i] &&
4772 old_opts.s_qf_names[i] != sbi->s_qf_names[i])
4773 kfree(sbi->s_qf_names[i]);
4774 sbi->s_qf_names[i] = old_opts.s_qf_names[i];
4781 static int ext4_statfs(struct dentry *dentry, struct kstatfs *buf)
4783 struct super_block *sb = dentry->d_sb;
4784 struct ext4_sb_info *sbi = EXT4_SB(sb);
4785 struct ext4_super_block *es = sbi->s_es;
4786 ext4_fsblk_t overhead = 0;
4790 if (!test_opt(sb, MINIX_DF))
4791 overhead = sbi->s_overhead;
4793 buf->f_type = EXT4_SUPER_MAGIC;
4794 buf->f_bsize = sb->s_blocksize;
4795 buf->f_blocks = ext4_blocks_count(es) - EXT4_C2B(sbi, sbi->s_overhead);
4796 bfree = percpu_counter_sum_positive(&sbi->s_freeclusters_counter) -
4797 percpu_counter_sum_positive(&sbi->s_dirtyclusters_counter);
4798 /* prevent underflow in case that few free space is available */
4799 buf->f_bfree = EXT4_C2B(sbi, max_t(s64, bfree, 0));
4800 buf->f_bavail = buf->f_bfree - ext4_r_blocks_count(es);
4801 if (buf->f_bfree < ext4_r_blocks_count(es))
4803 buf->f_files = le32_to_cpu(es->s_inodes_count);
4804 buf->f_ffree = percpu_counter_sum_positive(&sbi->s_freeinodes_counter);
4805 buf->f_namelen = EXT4_NAME_LEN;
4806 fsid = le64_to_cpup((void *)es->s_uuid) ^
4807 le64_to_cpup((void *)es->s_uuid + sizeof(u64));
4808 buf->f_fsid.val[0] = fsid & 0xFFFFFFFFUL;
4809 buf->f_fsid.val[1] = (fsid >> 32) & 0xFFFFFFFFUL;
4814 /* Helper function for writing quotas on sync - we need to start transaction
4815 * before quota file is locked for write. Otherwise the are possible deadlocks:
4816 * Process 1 Process 2
4817 * ext4_create() quota_sync()
4818 * jbd2_journal_start() write_dquot()
4819 * dquot_initialize() down(dqio_mutex)
4820 * down(dqio_mutex) jbd2_journal_start()
4826 static inline struct inode *dquot_to_inode(struct dquot *dquot)
4828 return sb_dqopt(dquot->dq_sb)->files[dquot->dq_type];
4831 static int ext4_write_dquot(struct dquot *dquot)
4835 struct inode *inode;
4837 inode = dquot_to_inode(dquot);
4838 handle = ext4_journal_start(inode,
4839 EXT4_QUOTA_TRANS_BLOCKS(dquot->dq_sb));
4841 return PTR_ERR(handle);
4842 ret = dquot_commit(dquot);
4843 err = ext4_journal_stop(handle);
4849 static int ext4_acquire_dquot(struct dquot *dquot)
4854 handle = ext4_journal_start(dquot_to_inode(dquot),
4855 EXT4_QUOTA_INIT_BLOCKS(dquot->dq_sb));
4857 return PTR_ERR(handle);
4858 ret = dquot_acquire(dquot);
4859 err = ext4_journal_stop(handle);
4865 static int ext4_release_dquot(struct dquot *dquot)
4870 handle = ext4_journal_start(dquot_to_inode(dquot),
4871 EXT4_QUOTA_DEL_BLOCKS(dquot->dq_sb));
4872 if (IS_ERR(handle)) {
4873 /* Release dquot anyway to avoid endless cycle in dqput() */
4874 dquot_release(dquot);
4875 return PTR_ERR(handle);
4877 ret = dquot_release(dquot);
4878 err = ext4_journal_stop(handle);
4884 static int ext4_mark_dquot_dirty(struct dquot *dquot)
4886 /* Are we journaling quotas? */
4887 if (EXT4_SB(dquot->dq_sb)->s_qf_names[USRQUOTA] ||
4888 EXT4_SB(dquot->dq_sb)->s_qf_names[GRPQUOTA]) {
4889 dquot_mark_dquot_dirty(dquot);
4890 return ext4_write_dquot(dquot);
4892 return dquot_mark_dquot_dirty(dquot);
4896 static int ext4_write_info(struct super_block *sb, int type)
4901 /* Data block + inode block */
4902 handle = ext4_journal_start(sb->s_root->d_inode, 2);
4904 return PTR_ERR(handle);
4905 ret = dquot_commit_info(sb, type);
4906 err = ext4_journal_stop(handle);
4913 * Turn on quotas during mount time - we need to find
4914 * the quota file and such...
4916 static int ext4_quota_on_mount(struct super_block *sb, int type)
4918 return dquot_quota_on_mount(sb, EXT4_SB(sb)->s_qf_names[type],
4919 EXT4_SB(sb)->s_jquota_fmt, type);
4923 * Standard function to be called on quota_on
4925 static int ext4_quota_on(struct super_block *sb, int type, int format_id,
4930 if (!test_opt(sb, QUOTA))
4933 /* Quotafile not on the same filesystem? */
4934 if (path->dentry->d_sb != sb)
4936 /* Journaling quota? */
4937 if (EXT4_SB(sb)->s_qf_names[type]) {
4938 /* Quotafile not in fs root? */
4939 if (path->dentry->d_parent != sb->s_root)
4940 ext4_msg(sb, KERN_WARNING,
4941 "Quota file not on filesystem root. "
4942 "Journaled quota will not work");
4946 * When we journal data on quota file, we have to flush journal to see
4947 * all updates to the file when we bypass pagecache...
4949 if (EXT4_SB(sb)->s_journal &&
4950 ext4_should_journal_data(path->dentry->d_inode)) {
4952 * We don't need to lock updates but journal_flush() could
4953 * otherwise be livelocked...
4955 jbd2_journal_lock_updates(EXT4_SB(sb)->s_journal);
4956 err = jbd2_journal_flush(EXT4_SB(sb)->s_journal);
4957 jbd2_journal_unlock_updates(EXT4_SB(sb)->s_journal);
4962 return dquot_quota_on(sb, type, format_id, path);
4965 static int ext4_quota_enable(struct super_block *sb, int type, int format_id,
4969 struct inode *qf_inode;
4970 unsigned long qf_inums[MAXQUOTAS] = {
4971 le32_to_cpu(EXT4_SB(sb)->s_es->s_usr_quota_inum),
4972 le32_to_cpu(EXT4_SB(sb)->s_es->s_grp_quota_inum)
4975 BUG_ON(!EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_QUOTA));
4977 if (!qf_inums[type])
4980 qf_inode = ext4_iget(sb, qf_inums[type]);
4981 if (IS_ERR(qf_inode)) {
4982 ext4_error(sb, "Bad quota inode # %lu", qf_inums[type]);
4983 return PTR_ERR(qf_inode);
4986 err = dquot_enable(qf_inode, type, format_id, flags);
4992 /* Enable usage tracking for all quota types. */
4993 static int ext4_enable_quotas(struct super_block *sb)
4996 unsigned long qf_inums[MAXQUOTAS] = {
4997 le32_to_cpu(EXT4_SB(sb)->s_es->s_usr_quota_inum),
4998 le32_to_cpu(EXT4_SB(sb)->s_es->s_grp_quota_inum)
5001 sb_dqopt(sb)->flags |= DQUOT_QUOTA_SYS_FILE;
5002 for (type = 0; type < MAXQUOTAS; type++) {
5003 if (qf_inums[type]) {
5004 err = ext4_quota_enable(sb, type, QFMT_VFS_V1,
5005 DQUOT_USAGE_ENABLED);
5008 "Failed to enable quota (type=%d) "
5009 "tracking. Please run e2fsck to fix.",
5019 * quota_on function that is used when QUOTA feature is set.
5021 static int ext4_quota_on_sysfile(struct super_block *sb, int type,
5024 if (!EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_QUOTA))
5028 * USAGE was enabled at mount time. Only need to enable LIMITS now.
5030 return ext4_quota_enable(sb, type, format_id, DQUOT_LIMITS_ENABLED);
5033 static int ext4_quota_off(struct super_block *sb, int type)
5035 struct inode *inode = sb_dqopt(sb)->files[type];
5038 /* Force all delayed allocation blocks to be allocated.
5039 * Caller already holds s_umount sem */
5040 if (test_opt(sb, DELALLOC))
5041 sync_filesystem(sb);
5046 /* Update modification times of quota files when userspace can
5047 * start looking at them */
5048 handle = ext4_journal_start(inode, 1);
5051 inode->i_mtime = inode->i_ctime = CURRENT_TIME;
5052 ext4_mark_inode_dirty(handle, inode);
5053 ext4_journal_stop(handle);
5056 return dquot_quota_off(sb, type);
5060 * quota_off function that is used when QUOTA feature is set.
5062 static int ext4_quota_off_sysfile(struct super_block *sb, int type)
5064 if (!EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_QUOTA))
5067 /* Disable only the limits. */
5068 return dquot_disable(sb, type, DQUOT_LIMITS_ENABLED);
5071 /* Read data from quotafile - avoid pagecache and such because we cannot afford
5072 * acquiring the locks... As quota files are never truncated and quota code
5073 * itself serializes the operations (and no one else should touch the files)
5074 * we don't have to be afraid of races */
5075 static ssize_t ext4_quota_read(struct super_block *sb, int type, char *data,
5076 size_t len, loff_t off)
5078 struct inode *inode = sb_dqopt(sb)->files[type];
5079 ext4_lblk_t blk = off >> EXT4_BLOCK_SIZE_BITS(sb);
5081 int offset = off & (sb->s_blocksize - 1);
5084 struct buffer_head *bh;
5085 loff_t i_size = i_size_read(inode);
5089 if (off+len > i_size)
5092 while (toread > 0) {
5093 tocopy = sb->s_blocksize - offset < toread ?
5094 sb->s_blocksize - offset : toread;
5095 bh = ext4_bread(NULL, inode, blk, 0, &err);
5098 if (!bh) /* A hole? */
5099 memset(data, 0, tocopy);
5101 memcpy(data, bh->b_data+offset, tocopy);
5111 /* Write to quotafile (we know the transaction is already started and has
5112 * enough credits) */
5113 static ssize_t ext4_quota_write(struct super_block *sb, int type,
5114 const char *data, size_t len, loff_t off)
5116 struct inode *inode = sb_dqopt(sb)->files[type];
5117 ext4_lblk_t blk = off >> EXT4_BLOCK_SIZE_BITS(sb);
5119 int offset = off & (sb->s_blocksize - 1);
5120 struct buffer_head *bh;
5121 handle_t *handle = journal_current_handle();
5123 if (EXT4_SB(sb)->s_journal && !handle) {
5124 ext4_msg(sb, KERN_WARNING, "Quota write (off=%llu, len=%llu)"
5125 " cancelled because transaction is not started",
5126 (unsigned long long)off, (unsigned long long)len);
5130 * Since we account only one data block in transaction credits,
5131 * then it is impossible to cross a block boundary.
5133 if (sb->s_blocksize - offset < len) {
5134 ext4_msg(sb, KERN_WARNING, "Quota write (off=%llu, len=%llu)"
5135 " cancelled because not block aligned",
5136 (unsigned long long)off, (unsigned long long)len);
5140 bh = ext4_bread(handle, inode, blk, 1, &err);
5143 err = ext4_journal_get_write_access(handle, bh);
5149 memcpy(bh->b_data+offset, data, len);
5150 flush_dcache_page(bh->b_page);
5152 err = ext4_handle_dirty_metadata(handle, NULL, bh);
5157 if (inode->i_size < off + len) {
5158 i_size_write(inode, off + len);
5159 EXT4_I(inode)->i_disksize = inode->i_size;
5160 ext4_mark_inode_dirty(handle, inode);
5167 static struct dentry *ext4_mount(struct file_system_type *fs_type, int flags,
5168 const char *dev_name, void *data)
5170 return mount_bdev(fs_type, flags, dev_name, data, ext4_fill_super);
5173 #if !defined(CONFIG_EXT2_FS) && !defined(CONFIG_EXT2_FS_MODULE) && defined(CONFIG_EXT4_USE_FOR_EXT23)
5174 static inline void register_as_ext2(void)
5176 int err = register_filesystem(&ext2_fs_type);
5179 "EXT4-fs: Unable to register as ext2 (%d)\n", err);
5182 static inline void unregister_as_ext2(void)
5184 unregister_filesystem(&ext2_fs_type);
5187 static inline int ext2_feature_set_ok(struct super_block *sb)
5189 if (EXT4_HAS_INCOMPAT_FEATURE(sb, ~EXT2_FEATURE_INCOMPAT_SUPP))
5191 if (sb->s_flags & MS_RDONLY)
5193 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, ~EXT2_FEATURE_RO_COMPAT_SUPP))
5197 MODULE_ALIAS("ext2");
5199 static inline void register_as_ext2(void) { }
5200 static inline void unregister_as_ext2(void) { }
5201 static inline int ext2_feature_set_ok(struct super_block *sb) { return 0; }
5204 #if !defined(CONFIG_EXT3_FS) && !defined(CONFIG_EXT3_FS_MODULE) && defined(CONFIG_EXT4_USE_FOR_EXT23)
5205 static inline void register_as_ext3(void)
5207 int err = register_filesystem(&ext3_fs_type);
5210 "EXT4-fs: Unable to register as ext3 (%d)\n", err);
5213 static inline void unregister_as_ext3(void)
5215 unregister_filesystem(&ext3_fs_type);
5218 static inline int ext3_feature_set_ok(struct super_block *sb)
5220 if (EXT4_HAS_INCOMPAT_FEATURE(sb, ~EXT3_FEATURE_INCOMPAT_SUPP))
5222 if (!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL))
5224 if (sb->s_flags & MS_RDONLY)
5226 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, ~EXT3_FEATURE_RO_COMPAT_SUPP))
5230 MODULE_ALIAS("ext3");
5232 static inline void register_as_ext3(void) { }
5233 static inline void unregister_as_ext3(void) { }
5234 static inline int ext3_feature_set_ok(struct super_block *sb) { return 0; }
5237 static struct file_system_type ext4_fs_type = {
5238 .owner = THIS_MODULE,
5240 .mount = ext4_mount,
5241 .kill_sb = kill_block_super,
5242 .fs_flags = FS_REQUIRES_DEV,
5245 static int __init ext4_init_feat_adverts(void)
5247 struct ext4_features *ef;
5250 ef = kzalloc(sizeof(struct ext4_features), GFP_KERNEL);
5254 ef->f_kobj.kset = ext4_kset;
5255 init_completion(&ef->f_kobj_unregister);
5256 ret = kobject_init_and_add(&ef->f_kobj, &ext4_feat_ktype, NULL,
5269 static void ext4_exit_feat_adverts(void)
5271 kobject_put(&ext4_feat->f_kobj);
5272 wait_for_completion(&ext4_feat->f_kobj_unregister);
5276 /* Shared across all ext4 file systems */
5277 wait_queue_head_t ext4__ioend_wq[EXT4_WQ_HASH_SZ];
5278 struct mutex ext4__aio_mutex[EXT4_WQ_HASH_SZ];
5280 static int __init ext4_init_fs(void)
5284 ext4_li_info = NULL;
5285 mutex_init(&ext4_li_mtx);
5287 ext4_check_flag_values();
5289 for (i = 0; i < EXT4_WQ_HASH_SZ; i++) {
5290 mutex_init(&ext4__aio_mutex[i]);
5291 init_waitqueue_head(&ext4__ioend_wq[i]);
5294 err = ext4_init_pageio();
5297 err = ext4_init_system_zone();
5300 ext4_kset = kset_create_and_add("ext4", NULL, fs_kobj);
5305 ext4_proc_root = proc_mkdir("fs/ext4", NULL);
5307 err = ext4_init_feat_adverts();
5311 err = ext4_init_mballoc();
5315 err = ext4_init_xattr();
5318 err = init_inodecache();
5323 err = register_filesystem(&ext4_fs_type);
5329 unregister_as_ext2();
5330 unregister_as_ext3();
5331 destroy_inodecache();
5335 ext4_exit_mballoc();
5337 ext4_exit_feat_adverts();
5340 remove_proc_entry("fs/ext4", NULL);
5341 kset_unregister(ext4_kset);
5343 ext4_exit_system_zone();
5349 static void __exit ext4_exit_fs(void)
5351 ext4_destroy_lazyinit_thread();
5352 unregister_as_ext2();
5353 unregister_as_ext3();
5354 unregister_filesystem(&ext4_fs_type);
5355 destroy_inodecache();
5357 ext4_exit_mballoc();
5358 ext4_exit_feat_adverts();
5359 remove_proc_entry("fs/ext4", NULL);
5360 kset_unregister(ext4_kset);
5361 ext4_exit_system_zone();
5365 MODULE_AUTHOR("Remy Card, Stephen Tweedie, Andrew Morton, Andreas Dilger, Theodore Ts'o and others");
5366 MODULE_DESCRIPTION("Fourth Extended Filesystem");
5367 MODULE_LICENSE("GPL");
5368 module_init(ext4_init_fs)
5369 module_exit(ext4_exit_fs)