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);
865 if (sbi->s_journal) {
866 err = jbd2_journal_destroy(sbi->s_journal);
867 sbi->s_journal = NULL;
869 ext4_abort(sb, "Couldn't clean up the journal");
872 del_timer(&sbi->s_err_report);
873 ext4_release_system_zone(sb);
875 ext4_ext_release(sb);
876 ext4_xattr_put_super(sb);
878 if (!(sb->s_flags & MS_RDONLY)) {
879 EXT4_CLEAR_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
880 es->s_state = cpu_to_le16(sbi->s_mount_state);
882 if (!(sb->s_flags & MS_RDONLY))
883 ext4_commit_super(sb, 1);
886 remove_proc_entry("options", sbi->s_proc);
887 remove_proc_entry(sb->s_id, ext4_proc_root);
889 kobject_del(&sbi->s_kobj);
891 for (i = 0; i < sbi->s_gdb_count; i++)
892 brelse(sbi->s_group_desc[i]);
893 ext4_kvfree(sbi->s_group_desc);
894 ext4_kvfree(sbi->s_flex_groups);
895 percpu_counter_destroy(&sbi->s_freeclusters_counter);
896 percpu_counter_destroy(&sbi->s_freeinodes_counter);
897 percpu_counter_destroy(&sbi->s_dirs_counter);
898 percpu_counter_destroy(&sbi->s_dirtyclusters_counter);
901 for (i = 0; i < MAXQUOTAS; i++)
902 kfree(sbi->s_qf_names[i]);
905 /* Debugging code just in case the in-memory inode orphan list
906 * isn't empty. The on-disk one can be non-empty if we've
907 * detected an error and taken the fs readonly, but the
908 * in-memory list had better be clean by this point. */
909 if (!list_empty(&sbi->s_orphan))
910 dump_orphan_list(sb, sbi);
911 J_ASSERT(list_empty(&sbi->s_orphan));
913 invalidate_bdev(sb->s_bdev);
914 if (sbi->journal_bdev && sbi->journal_bdev != sb->s_bdev) {
916 * Invalidate the journal device's buffers. We don't want them
917 * floating about in memory - the physical journal device may
918 * hotswapped, and it breaks the `ro-after' testing code.
920 sync_blockdev(sbi->journal_bdev);
921 invalidate_bdev(sbi->journal_bdev);
922 ext4_blkdev_remove(sbi);
925 kthread_stop(sbi->s_mmp_tsk);
926 sb->s_fs_info = NULL;
928 * Now that we are completely done shutting down the
929 * superblock, we need to actually destroy the kobject.
932 kobject_put(&sbi->s_kobj);
933 wait_for_completion(&sbi->s_kobj_unregister);
934 if (sbi->s_chksum_driver)
935 crypto_free_shash(sbi->s_chksum_driver);
936 kfree(sbi->s_blockgroup_lock);
940 static struct kmem_cache *ext4_inode_cachep;
943 * Called inside transaction, so use GFP_NOFS
945 static struct inode *ext4_alloc_inode(struct super_block *sb)
947 struct ext4_inode_info *ei;
949 ei = kmem_cache_alloc(ext4_inode_cachep, GFP_NOFS);
953 ei->vfs_inode.i_version = 1;
954 ei->vfs_inode.i_data.writeback_index = 0;
955 memset(&ei->i_cached_extent, 0, sizeof(struct ext4_ext_cache));
956 INIT_LIST_HEAD(&ei->i_prealloc_list);
957 spin_lock_init(&ei->i_prealloc_lock);
958 ei->i_reserved_data_blocks = 0;
959 ei->i_reserved_meta_blocks = 0;
960 ei->i_allocated_meta_blocks = 0;
961 ei->i_da_metadata_calc_len = 0;
962 ei->i_da_metadata_calc_last_lblock = 0;
963 spin_lock_init(&(ei->i_block_reservation_lock));
965 ei->i_reserved_quota = 0;
968 INIT_LIST_HEAD(&ei->i_completed_io_list);
969 spin_lock_init(&ei->i_completed_io_lock);
970 ei->cur_aio_dio = NULL;
972 ei->i_datasync_tid = 0;
973 atomic_set(&ei->i_ioend_count, 0);
974 atomic_set(&ei->i_aiodio_unwritten, 0);
976 return &ei->vfs_inode;
979 static int ext4_drop_inode(struct inode *inode)
981 int drop = generic_drop_inode(inode);
983 trace_ext4_drop_inode(inode, drop);
987 static void ext4_i_callback(struct rcu_head *head)
989 struct inode *inode = container_of(head, struct inode, i_rcu);
990 kmem_cache_free(ext4_inode_cachep, EXT4_I(inode));
993 static void ext4_destroy_inode(struct inode *inode)
995 if (!list_empty(&(EXT4_I(inode)->i_orphan))) {
996 ext4_msg(inode->i_sb, KERN_ERR,
997 "Inode %lu (%p): orphan list check failed!",
998 inode->i_ino, EXT4_I(inode));
999 print_hex_dump(KERN_INFO, "", DUMP_PREFIX_ADDRESS, 16, 4,
1000 EXT4_I(inode), sizeof(struct ext4_inode_info),
1004 call_rcu(&inode->i_rcu, ext4_i_callback);
1007 static void init_once(void *foo)
1009 struct ext4_inode_info *ei = (struct ext4_inode_info *) foo;
1011 INIT_LIST_HEAD(&ei->i_orphan);
1012 #ifdef CONFIG_EXT4_FS_XATTR
1013 init_rwsem(&ei->xattr_sem);
1015 init_rwsem(&ei->i_data_sem);
1016 inode_init_once(&ei->vfs_inode);
1019 static int init_inodecache(void)
1021 ext4_inode_cachep = kmem_cache_create("ext4_inode_cache",
1022 sizeof(struct ext4_inode_info),
1023 0, (SLAB_RECLAIM_ACCOUNT|
1026 if (ext4_inode_cachep == NULL)
1031 static void destroy_inodecache(void)
1033 kmem_cache_destroy(ext4_inode_cachep);
1036 void ext4_clear_inode(struct inode *inode)
1038 invalidate_inode_buffers(inode);
1041 ext4_discard_preallocations(inode);
1042 if (EXT4_I(inode)->jinode) {
1043 jbd2_journal_release_jbd_inode(EXT4_JOURNAL(inode),
1044 EXT4_I(inode)->jinode);
1045 jbd2_free_inode(EXT4_I(inode)->jinode);
1046 EXT4_I(inode)->jinode = NULL;
1050 static struct inode *ext4_nfs_get_inode(struct super_block *sb,
1051 u64 ino, u32 generation)
1053 struct inode *inode;
1055 if (ino < EXT4_FIRST_INO(sb) && ino != EXT4_ROOT_INO)
1056 return ERR_PTR(-ESTALE);
1057 if (ino > le32_to_cpu(EXT4_SB(sb)->s_es->s_inodes_count))
1058 return ERR_PTR(-ESTALE);
1060 /* iget isn't really right if the inode is currently unallocated!!
1062 * ext4_read_inode will return a bad_inode if the inode had been
1063 * deleted, so we should be safe.
1065 * Currently we don't know the generation for parent directory, so
1066 * a generation of 0 means "accept any"
1068 inode = ext4_iget(sb, ino);
1070 return ERR_CAST(inode);
1071 if (generation && inode->i_generation != generation) {
1073 return ERR_PTR(-ESTALE);
1079 static struct dentry *ext4_fh_to_dentry(struct super_block *sb, struct fid *fid,
1080 int fh_len, int fh_type)
1082 return generic_fh_to_dentry(sb, fid, fh_len, fh_type,
1083 ext4_nfs_get_inode);
1086 static struct dentry *ext4_fh_to_parent(struct super_block *sb, struct fid *fid,
1087 int fh_len, int fh_type)
1089 return generic_fh_to_parent(sb, fid, fh_len, fh_type,
1090 ext4_nfs_get_inode);
1094 * Try to release metadata pages (indirect blocks, directories) which are
1095 * mapped via the block device. Since these pages could have journal heads
1096 * which would prevent try_to_free_buffers() from freeing them, we must use
1097 * jbd2 layer's try_to_free_buffers() function to release them.
1099 static int bdev_try_to_free_page(struct super_block *sb, struct page *page,
1102 journal_t *journal = EXT4_SB(sb)->s_journal;
1104 WARN_ON(PageChecked(page));
1105 if (!page_has_buffers(page))
1108 return jbd2_journal_try_to_free_buffers(journal, page,
1109 wait & ~__GFP_WAIT);
1110 return try_to_free_buffers(page);
1114 #define QTYPE2NAME(t) ((t) == USRQUOTA ? "user" : "group")
1115 #define QTYPE2MOPT(on, t) ((t) == USRQUOTA?((on)##USRJQUOTA):((on)##GRPJQUOTA))
1117 static int ext4_write_dquot(struct dquot *dquot);
1118 static int ext4_acquire_dquot(struct dquot *dquot);
1119 static int ext4_release_dquot(struct dquot *dquot);
1120 static int ext4_mark_dquot_dirty(struct dquot *dquot);
1121 static int ext4_write_info(struct super_block *sb, int type);
1122 static int ext4_quota_on(struct super_block *sb, int type, int format_id,
1124 static int ext4_quota_on_sysfile(struct super_block *sb, int type,
1126 static int ext4_quota_off(struct super_block *sb, int type);
1127 static int ext4_quota_off_sysfile(struct super_block *sb, int type);
1128 static int ext4_quota_on_mount(struct super_block *sb, int type);
1129 static ssize_t ext4_quota_read(struct super_block *sb, int type, char *data,
1130 size_t len, loff_t off);
1131 static ssize_t ext4_quota_write(struct super_block *sb, int type,
1132 const char *data, size_t len, loff_t off);
1133 static int ext4_quota_enable(struct super_block *sb, int type, int format_id,
1134 unsigned int flags);
1135 static int ext4_enable_quotas(struct super_block *sb);
1137 static const struct dquot_operations ext4_quota_operations = {
1138 .get_reserved_space = ext4_get_reserved_space,
1139 .write_dquot = ext4_write_dquot,
1140 .acquire_dquot = ext4_acquire_dquot,
1141 .release_dquot = ext4_release_dquot,
1142 .mark_dirty = ext4_mark_dquot_dirty,
1143 .write_info = ext4_write_info,
1144 .alloc_dquot = dquot_alloc,
1145 .destroy_dquot = dquot_destroy,
1148 static const struct quotactl_ops ext4_qctl_operations = {
1149 .quota_on = ext4_quota_on,
1150 .quota_off = ext4_quota_off,
1151 .quota_sync = dquot_quota_sync,
1152 .get_info = dquot_get_dqinfo,
1153 .set_info = dquot_set_dqinfo,
1154 .get_dqblk = dquot_get_dqblk,
1155 .set_dqblk = dquot_set_dqblk
1158 static const struct quotactl_ops ext4_qctl_sysfile_operations = {
1159 .quota_on_meta = ext4_quota_on_sysfile,
1160 .quota_off = ext4_quota_off_sysfile,
1161 .quota_sync = dquot_quota_sync,
1162 .get_info = dquot_get_dqinfo,
1163 .set_info = dquot_set_dqinfo,
1164 .get_dqblk = dquot_get_dqblk,
1165 .set_dqblk = dquot_set_dqblk
1169 static const struct super_operations ext4_sops = {
1170 .alloc_inode = ext4_alloc_inode,
1171 .destroy_inode = ext4_destroy_inode,
1172 .write_inode = ext4_write_inode,
1173 .dirty_inode = ext4_dirty_inode,
1174 .drop_inode = ext4_drop_inode,
1175 .evict_inode = ext4_evict_inode,
1176 .put_super = ext4_put_super,
1177 .sync_fs = ext4_sync_fs,
1178 .freeze_fs = ext4_freeze,
1179 .unfreeze_fs = ext4_unfreeze,
1180 .statfs = ext4_statfs,
1181 .remount_fs = ext4_remount,
1182 .show_options = ext4_show_options,
1184 .quota_read = ext4_quota_read,
1185 .quota_write = ext4_quota_write,
1187 .bdev_try_to_free_page = bdev_try_to_free_page,
1190 static const struct super_operations ext4_nojournal_sops = {
1191 .alloc_inode = ext4_alloc_inode,
1192 .destroy_inode = ext4_destroy_inode,
1193 .write_inode = ext4_write_inode,
1194 .dirty_inode = ext4_dirty_inode,
1195 .drop_inode = ext4_drop_inode,
1196 .evict_inode = ext4_evict_inode,
1197 .put_super = ext4_put_super,
1198 .statfs = ext4_statfs,
1199 .remount_fs = ext4_remount,
1200 .show_options = ext4_show_options,
1202 .quota_read = ext4_quota_read,
1203 .quota_write = ext4_quota_write,
1205 .bdev_try_to_free_page = bdev_try_to_free_page,
1208 static const struct export_operations ext4_export_ops = {
1209 .fh_to_dentry = ext4_fh_to_dentry,
1210 .fh_to_parent = ext4_fh_to_parent,
1211 .get_parent = ext4_get_parent,
1215 Opt_bsd_df, Opt_minix_df, Opt_grpid, Opt_nogrpid,
1216 Opt_resgid, Opt_resuid, Opt_sb, Opt_err_cont, Opt_err_panic, Opt_err_ro,
1217 Opt_nouid32, Opt_debug, Opt_removed,
1218 Opt_user_xattr, Opt_nouser_xattr, Opt_acl, Opt_noacl,
1219 Opt_auto_da_alloc, Opt_noauto_da_alloc, Opt_noload,
1220 Opt_commit, Opt_min_batch_time, Opt_max_batch_time,
1221 Opt_journal_dev, Opt_journal_checksum, Opt_journal_async_commit,
1222 Opt_abort, Opt_data_journal, Opt_data_ordered, Opt_data_writeback,
1223 Opt_data_err_abort, Opt_data_err_ignore,
1224 Opt_usrjquota, Opt_grpjquota, Opt_offusrjquota, Opt_offgrpjquota,
1225 Opt_jqfmt_vfsold, Opt_jqfmt_vfsv0, Opt_jqfmt_vfsv1, Opt_quota,
1226 Opt_noquota, Opt_barrier, Opt_nobarrier, Opt_err,
1227 Opt_usrquota, Opt_grpquota, Opt_i_version,
1228 Opt_stripe, Opt_delalloc, Opt_nodelalloc, Opt_mblk_io_submit,
1229 Opt_nomblk_io_submit, Opt_block_validity, Opt_noblock_validity,
1230 Opt_inode_readahead_blks, Opt_journal_ioprio,
1231 Opt_dioread_nolock, Opt_dioread_lock,
1232 Opt_discard, Opt_nodiscard, Opt_init_itable, Opt_noinit_itable,
1235 static const match_table_t tokens = {
1236 {Opt_bsd_df, "bsddf"},
1237 {Opt_minix_df, "minixdf"},
1238 {Opt_grpid, "grpid"},
1239 {Opt_grpid, "bsdgroups"},
1240 {Opt_nogrpid, "nogrpid"},
1241 {Opt_nogrpid, "sysvgroups"},
1242 {Opt_resgid, "resgid=%u"},
1243 {Opt_resuid, "resuid=%u"},
1245 {Opt_err_cont, "errors=continue"},
1246 {Opt_err_panic, "errors=panic"},
1247 {Opt_err_ro, "errors=remount-ro"},
1248 {Opt_nouid32, "nouid32"},
1249 {Opt_debug, "debug"},
1250 {Opt_removed, "oldalloc"},
1251 {Opt_removed, "orlov"},
1252 {Opt_user_xattr, "user_xattr"},
1253 {Opt_nouser_xattr, "nouser_xattr"},
1255 {Opt_noacl, "noacl"},
1256 {Opt_noload, "norecovery"},
1257 {Opt_noload, "noload"},
1258 {Opt_removed, "nobh"},
1259 {Opt_removed, "bh"},
1260 {Opt_commit, "commit=%u"},
1261 {Opt_min_batch_time, "min_batch_time=%u"},
1262 {Opt_max_batch_time, "max_batch_time=%u"},
1263 {Opt_journal_dev, "journal_dev=%u"},
1264 {Opt_journal_checksum, "journal_checksum"},
1265 {Opt_journal_async_commit, "journal_async_commit"},
1266 {Opt_abort, "abort"},
1267 {Opt_data_journal, "data=journal"},
1268 {Opt_data_ordered, "data=ordered"},
1269 {Opt_data_writeback, "data=writeback"},
1270 {Opt_data_err_abort, "data_err=abort"},
1271 {Opt_data_err_ignore, "data_err=ignore"},
1272 {Opt_offusrjquota, "usrjquota="},
1273 {Opt_usrjquota, "usrjquota=%s"},
1274 {Opt_offgrpjquota, "grpjquota="},
1275 {Opt_grpjquota, "grpjquota=%s"},
1276 {Opt_jqfmt_vfsold, "jqfmt=vfsold"},
1277 {Opt_jqfmt_vfsv0, "jqfmt=vfsv0"},
1278 {Opt_jqfmt_vfsv1, "jqfmt=vfsv1"},
1279 {Opt_grpquota, "grpquota"},
1280 {Opt_noquota, "noquota"},
1281 {Opt_quota, "quota"},
1282 {Opt_usrquota, "usrquota"},
1283 {Opt_barrier, "barrier=%u"},
1284 {Opt_barrier, "barrier"},
1285 {Opt_nobarrier, "nobarrier"},
1286 {Opt_i_version, "i_version"},
1287 {Opt_stripe, "stripe=%u"},
1288 {Opt_delalloc, "delalloc"},
1289 {Opt_nodelalloc, "nodelalloc"},
1290 {Opt_mblk_io_submit, "mblk_io_submit"},
1291 {Opt_nomblk_io_submit, "nomblk_io_submit"},
1292 {Opt_block_validity, "block_validity"},
1293 {Opt_noblock_validity, "noblock_validity"},
1294 {Opt_inode_readahead_blks, "inode_readahead_blks=%u"},
1295 {Opt_journal_ioprio, "journal_ioprio=%u"},
1296 {Opt_auto_da_alloc, "auto_da_alloc=%u"},
1297 {Opt_auto_da_alloc, "auto_da_alloc"},
1298 {Opt_noauto_da_alloc, "noauto_da_alloc"},
1299 {Opt_dioread_nolock, "dioread_nolock"},
1300 {Opt_dioread_lock, "dioread_lock"},
1301 {Opt_discard, "discard"},
1302 {Opt_nodiscard, "nodiscard"},
1303 {Opt_init_itable, "init_itable=%u"},
1304 {Opt_init_itable, "init_itable"},
1305 {Opt_noinit_itable, "noinit_itable"},
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},
1489 static int handle_mount_opt(struct super_block *sb, char *opt, int token,
1490 substring_t *args, unsigned long *journal_devnum,
1491 unsigned int *journal_ioprio, int is_remount)
1493 struct ext4_sb_info *sbi = EXT4_SB(sb);
1494 const struct mount_opts *m;
1500 if (token == Opt_usrjquota)
1501 return set_qf_name(sb, USRQUOTA, &args[0]);
1502 else if (token == Opt_grpjquota)
1503 return set_qf_name(sb, GRPQUOTA, &args[0]);
1504 else if (token == Opt_offusrjquota)
1505 return clear_qf_name(sb, USRQUOTA);
1506 else if (token == Opt_offgrpjquota)
1507 return clear_qf_name(sb, GRPQUOTA);
1509 if (args->from && match_int(args, &arg))
1513 case Opt_nouser_xattr:
1514 ext4_msg(sb, KERN_WARNING, deprecated_msg, opt, "3.5");
1517 return 1; /* handled by get_sb_block() */
1519 ext4_msg(sb, KERN_WARNING,
1520 "Ignoring removed %s option", opt);
1523 uid = make_kuid(current_user_ns(), arg);
1524 if (!uid_valid(uid)) {
1525 ext4_msg(sb, KERN_ERR, "Invalid uid value %d", arg);
1528 sbi->s_resuid = uid;
1531 gid = make_kgid(current_user_ns(), arg);
1532 if (!gid_valid(gid)) {
1533 ext4_msg(sb, KERN_ERR, "Invalid gid value %d", arg);
1536 sbi->s_resgid = gid;
1539 sbi->s_mount_flags |= EXT4_MF_FS_ABORTED;
1542 sb->s_flags |= MS_I_VERSION;
1544 case Opt_journal_dev:
1546 ext4_msg(sb, KERN_ERR,
1547 "Cannot specify journal on remount");
1550 *journal_devnum = arg;
1552 case Opt_journal_ioprio:
1553 if (arg < 0 || arg > 7)
1555 *journal_ioprio = IOPRIO_PRIO_VALUE(IOPRIO_CLASS_BE, arg);
1559 for (m = ext4_mount_opts; m->token != Opt_err; m++) {
1560 if (token != m->token)
1562 if (args->from && (m->flags & MOPT_GTE0) && (arg < 0))
1564 if (m->flags & MOPT_EXPLICIT)
1565 set_opt2(sb, EXPLICIT_DELALLOC);
1566 if (m->flags & MOPT_CLEAR_ERR)
1567 clear_opt(sb, ERRORS_MASK);
1568 if (token == Opt_noquota && sb_any_quota_loaded(sb)) {
1569 ext4_msg(sb, KERN_ERR, "Cannot change quota "
1570 "options when quota turned on");
1574 if (m->flags & MOPT_NOSUPPORT) {
1575 ext4_msg(sb, KERN_ERR, "%s option not supported", opt);
1576 } else if (token == Opt_commit) {
1578 arg = JBD2_DEFAULT_MAX_COMMIT_AGE;
1579 sbi->s_commit_interval = HZ * arg;
1580 } else if (token == Opt_max_batch_time) {
1582 arg = EXT4_DEF_MAX_BATCH_TIME;
1583 sbi->s_max_batch_time = arg;
1584 } else if (token == Opt_min_batch_time) {
1585 sbi->s_min_batch_time = arg;
1586 } else if (token == Opt_inode_readahead_blks) {
1587 if (arg > (1 << 30))
1589 if (arg && !is_power_of_2(arg)) {
1590 ext4_msg(sb, KERN_ERR,
1591 "EXT4-fs: inode_readahead_blks"
1592 " must be a power of 2");
1595 sbi->s_inode_readahead_blks = arg;
1596 } else if (token == Opt_init_itable) {
1597 set_opt(sb, INIT_INODE_TABLE);
1599 arg = EXT4_DEF_LI_WAIT_MULT;
1600 sbi->s_li_wait_mult = arg;
1601 } else if (token == Opt_stripe) {
1602 sbi->s_stripe = arg;
1603 } else if (m->flags & MOPT_DATAJ) {
1605 if (!sbi->s_journal)
1606 ext4_msg(sb, KERN_WARNING, "Remounting file system with no journal so ignoring journalled data option");
1607 else if (test_opt(sb, DATA_FLAGS) !=
1609 ext4_msg(sb, KERN_ERR,
1610 "Cannot change data mode on remount");
1614 clear_opt(sb, DATA_FLAGS);
1615 sbi->s_mount_opt |= m->mount_opt;
1618 } else if (m->flags & MOPT_QFMT) {
1619 if (sb_any_quota_loaded(sb) &&
1620 sbi->s_jquota_fmt != m->mount_opt) {
1621 ext4_msg(sb, KERN_ERR, "Cannot "
1622 "change journaled quota options "
1623 "when quota turned on");
1626 sbi->s_jquota_fmt = m->mount_opt;
1631 if (m->flags & MOPT_CLEAR)
1633 else if (unlikely(!(m->flags & MOPT_SET))) {
1634 ext4_msg(sb, KERN_WARNING,
1635 "buggy handling of option %s", opt);
1640 sbi->s_mount_opt |= m->mount_opt;
1642 sbi->s_mount_opt &= ~m->mount_opt;
1646 ext4_msg(sb, KERN_ERR, "Unrecognized mount option \"%s\" "
1647 "or missing value", opt);
1651 static int parse_options(char *options, struct super_block *sb,
1652 unsigned long *journal_devnum,
1653 unsigned int *journal_ioprio,
1657 struct ext4_sb_info *sbi = EXT4_SB(sb);
1660 substring_t args[MAX_OPT_ARGS];
1666 while ((p = strsep(&options, ",")) != NULL) {
1670 * Initialize args struct so we know whether arg was
1671 * found; some options take optional arguments.
1673 args[0].to = args[0].from = 0;
1674 token = match_token(p, tokens, args);
1675 if (handle_mount_opt(sb, p, token, args, journal_devnum,
1676 journal_ioprio, is_remount) < 0)
1680 if (sbi->s_qf_names[USRQUOTA] || sbi->s_qf_names[GRPQUOTA]) {
1681 if (test_opt(sb, USRQUOTA) && sbi->s_qf_names[USRQUOTA])
1682 clear_opt(sb, USRQUOTA);
1684 if (test_opt(sb, GRPQUOTA) && sbi->s_qf_names[GRPQUOTA])
1685 clear_opt(sb, GRPQUOTA);
1687 if (test_opt(sb, GRPQUOTA) || test_opt(sb, USRQUOTA)) {
1688 ext4_msg(sb, KERN_ERR, "old and new quota "
1693 if (!sbi->s_jquota_fmt) {
1694 ext4_msg(sb, KERN_ERR, "journaled quota format "
1699 if (sbi->s_jquota_fmt) {
1700 ext4_msg(sb, KERN_ERR, "journaled quota format "
1701 "specified with no journaling "
1710 static inline void ext4_show_quota_options(struct seq_file *seq,
1711 struct super_block *sb)
1713 #if defined(CONFIG_QUOTA)
1714 struct ext4_sb_info *sbi = EXT4_SB(sb);
1716 if (sbi->s_jquota_fmt) {
1719 switch (sbi->s_jquota_fmt) {
1730 seq_printf(seq, ",jqfmt=%s", fmtname);
1733 if (sbi->s_qf_names[USRQUOTA])
1734 seq_printf(seq, ",usrjquota=%s", sbi->s_qf_names[USRQUOTA]);
1736 if (sbi->s_qf_names[GRPQUOTA])
1737 seq_printf(seq, ",grpjquota=%s", sbi->s_qf_names[GRPQUOTA]);
1739 if (test_opt(sb, USRQUOTA))
1740 seq_puts(seq, ",usrquota");
1742 if (test_opt(sb, GRPQUOTA))
1743 seq_puts(seq, ",grpquota");
1747 static const char *token2str(int token)
1749 static const struct match_token *t;
1751 for (t = tokens; t->token != Opt_err; t++)
1752 if (t->token == token && !strchr(t->pattern, '='))
1759 * - it's set to a non-default value OR
1760 * - if the per-sb default is different from the global default
1762 static int _ext4_show_options(struct seq_file *seq, struct super_block *sb,
1765 struct ext4_sb_info *sbi = EXT4_SB(sb);
1766 struct ext4_super_block *es = sbi->s_es;
1767 int def_errors, def_mount_opt = nodefs ? 0 : sbi->s_def_mount_opt;
1768 const struct mount_opts *m;
1769 char sep = nodefs ? '\n' : ',';
1771 #define SEQ_OPTS_PUTS(str) seq_printf(seq, "%c" str, sep)
1772 #define SEQ_OPTS_PRINT(str, arg) seq_printf(seq, "%c" str, sep, arg)
1774 if (sbi->s_sb_block != 1)
1775 SEQ_OPTS_PRINT("sb=%llu", sbi->s_sb_block);
1777 for (m = ext4_mount_opts; m->token != Opt_err; m++) {
1778 int want_set = m->flags & MOPT_SET;
1779 if (((m->flags & (MOPT_SET|MOPT_CLEAR)) == 0) ||
1780 (m->flags & MOPT_CLEAR_ERR))
1782 if (!(m->mount_opt & (sbi->s_mount_opt ^ def_mount_opt)))
1783 continue; /* skip if same as the default */
1785 (sbi->s_mount_opt & m->mount_opt) != m->mount_opt) ||
1786 (!want_set && (sbi->s_mount_opt & m->mount_opt)))
1787 continue; /* select Opt_noFoo vs Opt_Foo */
1788 SEQ_OPTS_PRINT("%s", token2str(m->token));
1791 if (nodefs || !uid_eq(sbi->s_resuid, make_kuid(&init_user_ns, EXT4_DEF_RESUID)) ||
1792 le16_to_cpu(es->s_def_resuid) != EXT4_DEF_RESUID)
1793 SEQ_OPTS_PRINT("resuid=%u",
1794 from_kuid_munged(&init_user_ns, sbi->s_resuid));
1795 if (nodefs || !gid_eq(sbi->s_resgid, make_kgid(&init_user_ns, EXT4_DEF_RESGID)) ||
1796 le16_to_cpu(es->s_def_resgid) != EXT4_DEF_RESGID)
1797 SEQ_OPTS_PRINT("resgid=%u",
1798 from_kgid_munged(&init_user_ns, sbi->s_resgid));
1799 def_errors = nodefs ? -1 : le16_to_cpu(es->s_errors);
1800 if (test_opt(sb, ERRORS_RO) && def_errors != EXT4_ERRORS_RO)
1801 SEQ_OPTS_PUTS("errors=remount-ro");
1802 if (test_opt(sb, ERRORS_CONT) && def_errors != EXT4_ERRORS_CONTINUE)
1803 SEQ_OPTS_PUTS("errors=continue");
1804 if (test_opt(sb, ERRORS_PANIC) && def_errors != EXT4_ERRORS_PANIC)
1805 SEQ_OPTS_PUTS("errors=panic");
1806 if (nodefs || sbi->s_commit_interval != JBD2_DEFAULT_MAX_COMMIT_AGE*HZ)
1807 SEQ_OPTS_PRINT("commit=%lu", sbi->s_commit_interval / HZ);
1808 if (nodefs || sbi->s_min_batch_time != EXT4_DEF_MIN_BATCH_TIME)
1809 SEQ_OPTS_PRINT("min_batch_time=%u", sbi->s_min_batch_time);
1810 if (nodefs || sbi->s_max_batch_time != EXT4_DEF_MAX_BATCH_TIME)
1811 SEQ_OPTS_PRINT("max_batch_time=%u", sbi->s_max_batch_time);
1812 if (sb->s_flags & MS_I_VERSION)
1813 SEQ_OPTS_PUTS("i_version");
1814 if (nodefs || sbi->s_stripe)
1815 SEQ_OPTS_PRINT("stripe=%lu", sbi->s_stripe);
1816 if (EXT4_MOUNT_DATA_FLAGS & (sbi->s_mount_opt ^ def_mount_opt)) {
1817 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA)
1818 SEQ_OPTS_PUTS("data=journal");
1819 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA)
1820 SEQ_OPTS_PUTS("data=ordered");
1821 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_WRITEBACK_DATA)
1822 SEQ_OPTS_PUTS("data=writeback");
1825 sbi->s_inode_readahead_blks != EXT4_DEF_INODE_READAHEAD_BLKS)
1826 SEQ_OPTS_PRINT("inode_readahead_blks=%u",
1827 sbi->s_inode_readahead_blks);
1829 if (nodefs || (test_opt(sb, INIT_INODE_TABLE) &&
1830 (sbi->s_li_wait_mult != EXT4_DEF_LI_WAIT_MULT)))
1831 SEQ_OPTS_PRINT("init_itable=%u", sbi->s_li_wait_mult);
1833 ext4_show_quota_options(seq, sb);
1837 static int ext4_show_options(struct seq_file *seq, struct dentry *root)
1839 return _ext4_show_options(seq, root->d_sb, 0);
1842 static int options_seq_show(struct seq_file *seq, void *offset)
1844 struct super_block *sb = seq->private;
1847 seq_puts(seq, (sb->s_flags & MS_RDONLY) ? "ro" : "rw");
1848 rc = _ext4_show_options(seq, sb, 1);
1849 seq_puts(seq, "\n");
1853 static int options_open_fs(struct inode *inode, struct file *file)
1855 return single_open(file, options_seq_show, PDE(inode)->data);
1858 static const struct file_operations ext4_seq_options_fops = {
1859 .owner = THIS_MODULE,
1860 .open = options_open_fs,
1862 .llseek = seq_lseek,
1863 .release = single_release,
1866 static int ext4_setup_super(struct super_block *sb, struct ext4_super_block *es,
1869 struct ext4_sb_info *sbi = EXT4_SB(sb);
1872 if (le32_to_cpu(es->s_rev_level) > EXT4_MAX_SUPP_REV) {
1873 ext4_msg(sb, KERN_ERR, "revision level too high, "
1874 "forcing read-only mode");
1879 if (!(sbi->s_mount_state & EXT4_VALID_FS))
1880 ext4_msg(sb, KERN_WARNING, "warning: mounting unchecked fs, "
1881 "running e2fsck is recommended");
1882 else if ((sbi->s_mount_state & EXT4_ERROR_FS))
1883 ext4_msg(sb, KERN_WARNING,
1884 "warning: mounting fs with errors, "
1885 "running e2fsck is recommended");
1886 else if ((__s16) le16_to_cpu(es->s_max_mnt_count) > 0 &&
1887 le16_to_cpu(es->s_mnt_count) >=
1888 (unsigned short) (__s16) le16_to_cpu(es->s_max_mnt_count))
1889 ext4_msg(sb, KERN_WARNING,
1890 "warning: maximal mount count reached, "
1891 "running e2fsck is recommended");
1892 else if (le32_to_cpu(es->s_checkinterval) &&
1893 (le32_to_cpu(es->s_lastcheck) +
1894 le32_to_cpu(es->s_checkinterval) <= get_seconds()))
1895 ext4_msg(sb, KERN_WARNING,
1896 "warning: checktime reached, "
1897 "running e2fsck is recommended");
1898 if (!sbi->s_journal)
1899 es->s_state &= cpu_to_le16(~EXT4_VALID_FS);
1900 if (!(__s16) le16_to_cpu(es->s_max_mnt_count))
1901 es->s_max_mnt_count = cpu_to_le16(EXT4_DFL_MAX_MNT_COUNT);
1902 le16_add_cpu(&es->s_mnt_count, 1);
1903 es->s_mtime = cpu_to_le32(get_seconds());
1904 ext4_update_dynamic_rev(sb);
1906 EXT4_SET_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
1908 ext4_commit_super(sb, 1);
1910 if (test_opt(sb, DEBUG))
1911 printk(KERN_INFO "[EXT4 FS bs=%lu, gc=%u, "
1912 "bpg=%lu, ipg=%lu, mo=%04x, mo2=%04x]\n",
1914 sbi->s_groups_count,
1915 EXT4_BLOCKS_PER_GROUP(sb),
1916 EXT4_INODES_PER_GROUP(sb),
1917 sbi->s_mount_opt, sbi->s_mount_opt2);
1919 cleancache_init_fs(sb);
1923 static int ext4_fill_flex_info(struct super_block *sb)
1925 struct ext4_sb_info *sbi = EXT4_SB(sb);
1926 struct ext4_group_desc *gdp = NULL;
1927 ext4_group_t flex_group_count;
1928 ext4_group_t flex_group;
1929 unsigned int groups_per_flex = 0;
1933 sbi->s_log_groups_per_flex = sbi->s_es->s_log_groups_per_flex;
1934 if (sbi->s_log_groups_per_flex < 1 || sbi->s_log_groups_per_flex > 31) {
1935 sbi->s_log_groups_per_flex = 0;
1938 groups_per_flex = 1 << sbi->s_log_groups_per_flex;
1940 /* We allocate both existing and potentially added groups */
1941 flex_group_count = ((sbi->s_groups_count + groups_per_flex - 1) +
1942 ((le16_to_cpu(sbi->s_es->s_reserved_gdt_blocks) + 1) <<
1943 EXT4_DESC_PER_BLOCK_BITS(sb))) / groups_per_flex;
1944 size = flex_group_count * sizeof(struct flex_groups);
1945 sbi->s_flex_groups = ext4_kvzalloc(size, GFP_KERNEL);
1946 if (sbi->s_flex_groups == NULL) {
1947 ext4_msg(sb, KERN_ERR, "not enough memory for %u flex groups",
1952 for (i = 0; i < sbi->s_groups_count; i++) {
1953 gdp = ext4_get_group_desc(sb, i, NULL);
1955 flex_group = ext4_flex_group(sbi, i);
1956 atomic_add(ext4_free_inodes_count(sb, gdp),
1957 &sbi->s_flex_groups[flex_group].free_inodes);
1958 atomic_add(ext4_free_group_clusters(sb, gdp),
1959 &sbi->s_flex_groups[flex_group].free_clusters);
1960 atomic_add(ext4_used_dirs_count(sb, gdp),
1961 &sbi->s_flex_groups[flex_group].used_dirs);
1969 static __le16 ext4_group_desc_csum(struct ext4_sb_info *sbi, __u32 block_group,
1970 struct ext4_group_desc *gdp)
1974 __le32 le_group = cpu_to_le32(block_group);
1976 if ((sbi->s_es->s_feature_ro_compat &
1977 cpu_to_le32(EXT4_FEATURE_RO_COMPAT_METADATA_CSUM))) {
1978 /* Use new metadata_csum algorithm */
1982 old_csum = gdp->bg_checksum;
1983 gdp->bg_checksum = 0;
1984 csum32 = ext4_chksum(sbi, sbi->s_csum_seed, (__u8 *)&le_group,
1986 csum32 = ext4_chksum(sbi, csum32, (__u8 *)gdp,
1988 gdp->bg_checksum = old_csum;
1990 crc = csum32 & 0xFFFF;
1994 /* old crc16 code */
1995 offset = offsetof(struct ext4_group_desc, bg_checksum);
1997 crc = crc16(~0, sbi->s_es->s_uuid, sizeof(sbi->s_es->s_uuid));
1998 crc = crc16(crc, (__u8 *)&le_group, sizeof(le_group));
1999 crc = crc16(crc, (__u8 *)gdp, offset);
2000 offset += sizeof(gdp->bg_checksum); /* skip checksum */
2001 /* for checksum of struct ext4_group_desc do the rest...*/
2002 if ((sbi->s_es->s_feature_incompat &
2003 cpu_to_le32(EXT4_FEATURE_INCOMPAT_64BIT)) &&
2004 offset < le16_to_cpu(sbi->s_es->s_desc_size))
2005 crc = crc16(crc, (__u8 *)gdp + offset,
2006 le16_to_cpu(sbi->s_es->s_desc_size) -
2010 return cpu_to_le16(crc);
2013 int ext4_group_desc_csum_verify(struct super_block *sb, __u32 block_group,
2014 struct ext4_group_desc *gdp)
2016 if (ext4_has_group_desc_csum(sb) &&
2017 (gdp->bg_checksum != ext4_group_desc_csum(EXT4_SB(sb),
2024 void ext4_group_desc_csum_set(struct super_block *sb, __u32 block_group,
2025 struct ext4_group_desc *gdp)
2027 if (!ext4_has_group_desc_csum(sb))
2029 gdp->bg_checksum = ext4_group_desc_csum(EXT4_SB(sb), block_group, gdp);
2032 /* Called at mount-time, super-block is locked */
2033 static int ext4_check_descriptors(struct super_block *sb,
2034 ext4_group_t *first_not_zeroed)
2036 struct ext4_sb_info *sbi = EXT4_SB(sb);
2037 ext4_fsblk_t first_block = le32_to_cpu(sbi->s_es->s_first_data_block);
2038 ext4_fsblk_t last_block;
2039 ext4_fsblk_t block_bitmap;
2040 ext4_fsblk_t inode_bitmap;
2041 ext4_fsblk_t inode_table;
2042 int flexbg_flag = 0;
2043 ext4_group_t i, grp = sbi->s_groups_count;
2045 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_FLEX_BG))
2048 ext4_debug("Checking group descriptors");
2050 for (i = 0; i < sbi->s_groups_count; i++) {
2051 struct ext4_group_desc *gdp = ext4_get_group_desc(sb, i, NULL);
2053 if (i == sbi->s_groups_count - 1 || flexbg_flag)
2054 last_block = ext4_blocks_count(sbi->s_es) - 1;
2056 last_block = first_block +
2057 (EXT4_BLOCKS_PER_GROUP(sb) - 1);
2059 if ((grp == sbi->s_groups_count) &&
2060 !(gdp->bg_flags & cpu_to_le16(EXT4_BG_INODE_ZEROED)))
2063 block_bitmap = ext4_block_bitmap(sb, gdp);
2064 if (block_bitmap < first_block || block_bitmap > last_block) {
2065 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2066 "Block bitmap for group %u not in group "
2067 "(block %llu)!", i, block_bitmap);
2070 inode_bitmap = ext4_inode_bitmap(sb, gdp);
2071 if (inode_bitmap < first_block || inode_bitmap > last_block) {
2072 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2073 "Inode bitmap for group %u not in group "
2074 "(block %llu)!", i, inode_bitmap);
2077 inode_table = ext4_inode_table(sb, gdp);
2078 if (inode_table < first_block ||
2079 inode_table + sbi->s_itb_per_group - 1 > last_block) {
2080 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2081 "Inode table for group %u not in group "
2082 "(block %llu)!", i, inode_table);
2085 ext4_lock_group(sb, i);
2086 if (!ext4_group_desc_csum_verify(sb, i, gdp)) {
2087 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2088 "Checksum for group %u failed (%u!=%u)",
2089 i, le16_to_cpu(ext4_group_desc_csum(sbi, i,
2090 gdp)), le16_to_cpu(gdp->bg_checksum));
2091 if (!(sb->s_flags & MS_RDONLY)) {
2092 ext4_unlock_group(sb, i);
2096 ext4_unlock_group(sb, i);
2098 first_block += EXT4_BLOCKS_PER_GROUP(sb);
2100 if (NULL != first_not_zeroed)
2101 *first_not_zeroed = grp;
2103 ext4_free_blocks_count_set(sbi->s_es,
2104 EXT4_C2B(sbi, ext4_count_free_clusters(sb)));
2105 sbi->s_es->s_free_inodes_count =cpu_to_le32(ext4_count_free_inodes(sb));
2109 /* ext4_orphan_cleanup() walks a singly-linked list of inodes (starting at
2110 * the superblock) which were deleted from all directories, but held open by
2111 * a process at the time of a crash. We walk the list and try to delete these
2112 * inodes at recovery time (only with a read-write filesystem).
2114 * In order to keep the orphan inode chain consistent during traversal (in
2115 * case of crash during recovery), we link each inode into the superblock
2116 * orphan list_head and handle it the same way as an inode deletion during
2117 * normal operation (which journals the operations for us).
2119 * We only do an iget() and an iput() on each inode, which is very safe if we
2120 * accidentally point at an in-use or already deleted inode. The worst that
2121 * can happen in this case is that we get a "bit already cleared" message from
2122 * ext4_free_inode(). The only reason we would point at a wrong inode is if
2123 * e2fsck was run on this filesystem, and it must have already done the orphan
2124 * inode cleanup for us, so we can safely abort without any further action.
2126 static void ext4_orphan_cleanup(struct super_block *sb,
2127 struct ext4_super_block *es)
2129 unsigned int s_flags = sb->s_flags;
2130 int nr_orphans = 0, nr_truncates = 0;
2134 if (!es->s_last_orphan) {
2135 jbd_debug(4, "no orphan inodes to clean up\n");
2139 if (bdev_read_only(sb->s_bdev)) {
2140 ext4_msg(sb, KERN_ERR, "write access "
2141 "unavailable, skipping orphan cleanup");
2145 /* Check if feature set would not allow a r/w mount */
2146 if (!ext4_feature_set_ok(sb, 0)) {
2147 ext4_msg(sb, KERN_INFO, "Skipping orphan cleanup due to "
2148 "unknown ROCOMPAT features");
2152 if (EXT4_SB(sb)->s_mount_state & EXT4_ERROR_FS) {
2153 if (es->s_last_orphan)
2154 jbd_debug(1, "Errors on filesystem, "
2155 "clearing orphan list.\n");
2156 es->s_last_orphan = 0;
2157 jbd_debug(1, "Skipping orphan recovery on fs with errors.\n");
2161 if (s_flags & MS_RDONLY) {
2162 ext4_msg(sb, KERN_INFO, "orphan cleanup on readonly fs");
2163 sb->s_flags &= ~MS_RDONLY;
2166 /* Needed for iput() to work correctly and not trash data */
2167 sb->s_flags |= MS_ACTIVE;
2168 /* Turn on quotas so that they are updated correctly */
2169 for (i = 0; i < MAXQUOTAS; i++) {
2170 if (EXT4_SB(sb)->s_qf_names[i]) {
2171 int ret = ext4_quota_on_mount(sb, i);
2173 ext4_msg(sb, KERN_ERR,
2174 "Cannot turn on journaled "
2175 "quota: error %d", ret);
2180 while (es->s_last_orphan) {
2181 struct inode *inode;
2183 inode = ext4_orphan_get(sb, le32_to_cpu(es->s_last_orphan));
2184 if (IS_ERR(inode)) {
2185 es->s_last_orphan = 0;
2189 list_add(&EXT4_I(inode)->i_orphan, &EXT4_SB(sb)->s_orphan);
2190 dquot_initialize(inode);
2191 if (inode->i_nlink) {
2192 ext4_msg(sb, KERN_DEBUG,
2193 "%s: truncating inode %lu to %lld bytes",
2194 __func__, inode->i_ino, inode->i_size);
2195 jbd_debug(2, "truncating inode %lu to %lld bytes\n",
2196 inode->i_ino, inode->i_size);
2197 ext4_truncate(inode);
2200 ext4_msg(sb, KERN_DEBUG,
2201 "%s: deleting unreferenced inode %lu",
2202 __func__, inode->i_ino);
2203 jbd_debug(2, "deleting unreferenced inode %lu\n",
2207 iput(inode); /* The delete magic happens here! */
2210 #define PLURAL(x) (x), ((x) == 1) ? "" : "s"
2213 ext4_msg(sb, KERN_INFO, "%d orphan inode%s deleted",
2214 PLURAL(nr_orphans));
2216 ext4_msg(sb, KERN_INFO, "%d truncate%s cleaned up",
2217 PLURAL(nr_truncates));
2219 /* Turn quotas off */
2220 for (i = 0; i < MAXQUOTAS; i++) {
2221 if (sb_dqopt(sb)->files[i])
2222 dquot_quota_off(sb, i);
2225 sb->s_flags = s_flags; /* Restore MS_RDONLY status */
2229 * Maximal extent format file size.
2230 * Resulting logical blkno at s_maxbytes must fit in our on-disk
2231 * extent format containers, within a sector_t, and within i_blocks
2232 * in the vfs. ext4 inode has 48 bits of i_block in fsblock units,
2233 * so that won't be a limiting factor.
2235 * However there is other limiting factor. We do store extents in the form
2236 * of starting block and length, hence the resulting length of the extent
2237 * covering maximum file size must fit into on-disk format containers as
2238 * well. Given that length is always by 1 unit bigger than max unit (because
2239 * we count 0 as well) we have to lower the s_maxbytes by one fs block.
2241 * Note, this does *not* consider any metadata overhead for vfs i_blocks.
2243 static loff_t ext4_max_size(int blkbits, int has_huge_files)
2246 loff_t upper_limit = MAX_LFS_FILESIZE;
2248 /* small i_blocks in vfs inode? */
2249 if (!has_huge_files || sizeof(blkcnt_t) < sizeof(u64)) {
2251 * CONFIG_LBDAF is not enabled implies the inode
2252 * i_block represent total blocks in 512 bytes
2253 * 32 == size of vfs inode i_blocks * 8
2255 upper_limit = (1LL << 32) - 1;
2257 /* total blocks in file system block size */
2258 upper_limit >>= (blkbits - 9);
2259 upper_limit <<= blkbits;
2263 * 32-bit extent-start container, ee_block. We lower the maxbytes
2264 * by one fs block, so ee_len can cover the extent of maximum file
2267 res = (1LL << 32) - 1;
2270 /* Sanity check against vm- & vfs- imposed limits */
2271 if (res > upper_limit)
2278 * Maximal bitmap file size. There is a direct, and {,double-,triple-}indirect
2279 * block limit, and also a limit of (2^48 - 1) 512-byte sectors in i_blocks.
2280 * We need to be 1 filesystem block less than the 2^48 sector limit.
2282 static loff_t ext4_max_bitmap_size(int bits, int has_huge_files)
2284 loff_t res = EXT4_NDIR_BLOCKS;
2287 /* This is calculated to be the largest file size for a dense, block
2288 * mapped file such that the file's total number of 512-byte sectors,
2289 * including data and all indirect blocks, does not exceed (2^48 - 1).
2291 * __u32 i_blocks_lo and _u16 i_blocks_high represent the total
2292 * number of 512-byte sectors of the file.
2295 if (!has_huge_files || sizeof(blkcnt_t) < sizeof(u64)) {
2297 * !has_huge_files or CONFIG_LBDAF not enabled implies that
2298 * the inode i_block field represents total file blocks in
2299 * 2^32 512-byte sectors == size of vfs inode i_blocks * 8
2301 upper_limit = (1LL << 32) - 1;
2303 /* total blocks in file system block size */
2304 upper_limit >>= (bits - 9);
2308 * We use 48 bit ext4_inode i_blocks
2309 * With EXT4_HUGE_FILE_FL set the i_blocks
2310 * represent total number of blocks in
2311 * file system block size
2313 upper_limit = (1LL << 48) - 1;
2317 /* indirect blocks */
2319 /* double indirect blocks */
2320 meta_blocks += 1 + (1LL << (bits-2));
2321 /* tripple indirect blocks */
2322 meta_blocks += 1 + (1LL << (bits-2)) + (1LL << (2*(bits-2)));
2324 upper_limit -= meta_blocks;
2325 upper_limit <<= bits;
2327 res += 1LL << (bits-2);
2328 res += 1LL << (2*(bits-2));
2329 res += 1LL << (3*(bits-2));
2331 if (res > upper_limit)
2334 if (res > MAX_LFS_FILESIZE)
2335 res = MAX_LFS_FILESIZE;
2340 static ext4_fsblk_t descriptor_loc(struct super_block *sb,
2341 ext4_fsblk_t logical_sb_block, int nr)
2343 struct ext4_sb_info *sbi = EXT4_SB(sb);
2344 ext4_group_t bg, first_meta_bg;
2347 first_meta_bg = le32_to_cpu(sbi->s_es->s_first_meta_bg);
2349 if (!EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_META_BG) ||
2351 return logical_sb_block + nr + 1;
2352 bg = sbi->s_desc_per_block * nr;
2353 if (ext4_bg_has_super(sb, bg))
2356 return (has_super + ext4_group_first_block_no(sb, bg));
2360 * ext4_get_stripe_size: Get the stripe size.
2361 * @sbi: In memory super block info
2363 * If we have specified it via mount option, then
2364 * use the mount option value. If the value specified at mount time is
2365 * greater than the blocks per group use the super block value.
2366 * If the super block value is greater than blocks per group return 0.
2367 * Allocator needs it be less than blocks per group.
2370 static unsigned long ext4_get_stripe_size(struct ext4_sb_info *sbi)
2372 unsigned long stride = le16_to_cpu(sbi->s_es->s_raid_stride);
2373 unsigned long stripe_width =
2374 le32_to_cpu(sbi->s_es->s_raid_stripe_width);
2377 if (sbi->s_stripe && sbi->s_stripe <= sbi->s_blocks_per_group)
2378 ret = sbi->s_stripe;
2379 else if (stripe_width <= sbi->s_blocks_per_group)
2381 else if (stride <= sbi->s_blocks_per_group)
2387 * If the stripe width is 1, this makes no sense and
2388 * we set it to 0 to turn off stripe handling code.
2399 struct attribute attr;
2400 ssize_t (*show)(struct ext4_attr *, struct ext4_sb_info *, char *);
2401 ssize_t (*store)(struct ext4_attr *, struct ext4_sb_info *,
2402 const char *, size_t);
2406 static int parse_strtoul(const char *buf,
2407 unsigned long max, unsigned long *value)
2411 *value = simple_strtoul(skip_spaces(buf), &endp, 0);
2412 endp = skip_spaces(endp);
2413 if (*endp || *value > max)
2419 static ssize_t delayed_allocation_blocks_show(struct ext4_attr *a,
2420 struct ext4_sb_info *sbi,
2423 return snprintf(buf, PAGE_SIZE, "%llu\n",
2425 percpu_counter_sum(&sbi->s_dirtyclusters_counter)));
2428 static ssize_t session_write_kbytes_show(struct ext4_attr *a,
2429 struct ext4_sb_info *sbi, char *buf)
2431 struct super_block *sb = sbi->s_buddy_cache->i_sb;
2433 if (!sb->s_bdev->bd_part)
2434 return snprintf(buf, PAGE_SIZE, "0\n");
2435 return snprintf(buf, PAGE_SIZE, "%lu\n",
2436 (part_stat_read(sb->s_bdev->bd_part, sectors[1]) -
2437 sbi->s_sectors_written_start) >> 1);
2440 static ssize_t lifetime_write_kbytes_show(struct ext4_attr *a,
2441 struct ext4_sb_info *sbi, char *buf)
2443 struct super_block *sb = sbi->s_buddy_cache->i_sb;
2445 if (!sb->s_bdev->bd_part)
2446 return snprintf(buf, PAGE_SIZE, "0\n");
2447 return snprintf(buf, PAGE_SIZE, "%llu\n",
2448 (unsigned long long)(sbi->s_kbytes_written +
2449 ((part_stat_read(sb->s_bdev->bd_part, sectors[1]) -
2450 EXT4_SB(sb)->s_sectors_written_start) >> 1)));
2453 static ssize_t inode_readahead_blks_store(struct ext4_attr *a,
2454 struct ext4_sb_info *sbi,
2455 const char *buf, size_t count)
2459 if (parse_strtoul(buf, 0x40000000, &t))
2462 if (t && !is_power_of_2(t))
2465 sbi->s_inode_readahead_blks = t;
2469 static ssize_t sbi_ui_show(struct ext4_attr *a,
2470 struct ext4_sb_info *sbi, char *buf)
2472 unsigned int *ui = (unsigned int *) (((char *) sbi) + a->offset);
2474 return snprintf(buf, PAGE_SIZE, "%u\n", *ui);
2477 static ssize_t sbi_ui_store(struct ext4_attr *a,
2478 struct ext4_sb_info *sbi,
2479 const char *buf, size_t count)
2481 unsigned int *ui = (unsigned int *) (((char *) sbi) + a->offset);
2484 if (parse_strtoul(buf, 0xffffffff, &t))
2490 static ssize_t trigger_test_error(struct ext4_attr *a,
2491 struct ext4_sb_info *sbi,
2492 const char *buf, size_t count)
2496 if (!capable(CAP_SYS_ADMIN))
2499 if (len && buf[len-1] == '\n')
2503 ext4_error(sbi->s_sb, "%.*s", len, buf);
2507 #define EXT4_ATTR_OFFSET(_name,_mode,_show,_store,_elname) \
2508 static struct ext4_attr ext4_attr_##_name = { \
2509 .attr = {.name = __stringify(_name), .mode = _mode }, \
2512 .offset = offsetof(struct ext4_sb_info, _elname), \
2514 #define EXT4_ATTR(name, mode, show, store) \
2515 static struct ext4_attr ext4_attr_##name = __ATTR(name, mode, show, store)
2517 #define EXT4_INFO_ATTR(name) EXT4_ATTR(name, 0444, NULL, NULL)
2518 #define EXT4_RO_ATTR(name) EXT4_ATTR(name, 0444, name##_show, NULL)
2519 #define EXT4_RW_ATTR(name) EXT4_ATTR(name, 0644, name##_show, name##_store)
2520 #define EXT4_RW_ATTR_SBI_UI(name, elname) \
2521 EXT4_ATTR_OFFSET(name, 0644, sbi_ui_show, sbi_ui_store, elname)
2522 #define ATTR_LIST(name) &ext4_attr_##name.attr
2524 EXT4_RO_ATTR(delayed_allocation_blocks);
2525 EXT4_RO_ATTR(session_write_kbytes);
2526 EXT4_RO_ATTR(lifetime_write_kbytes);
2527 EXT4_ATTR_OFFSET(inode_readahead_blks, 0644, sbi_ui_show,
2528 inode_readahead_blks_store, s_inode_readahead_blks);
2529 EXT4_RW_ATTR_SBI_UI(inode_goal, s_inode_goal);
2530 EXT4_RW_ATTR_SBI_UI(mb_stats, s_mb_stats);
2531 EXT4_RW_ATTR_SBI_UI(mb_max_to_scan, s_mb_max_to_scan);
2532 EXT4_RW_ATTR_SBI_UI(mb_min_to_scan, s_mb_min_to_scan);
2533 EXT4_RW_ATTR_SBI_UI(mb_order2_req, s_mb_order2_reqs);
2534 EXT4_RW_ATTR_SBI_UI(mb_stream_req, s_mb_stream_request);
2535 EXT4_RW_ATTR_SBI_UI(mb_group_prealloc, s_mb_group_prealloc);
2536 EXT4_RW_ATTR_SBI_UI(max_writeback_mb_bump, s_max_writeback_mb_bump);
2537 EXT4_ATTR(trigger_fs_error, 0200, NULL, trigger_test_error);
2539 static struct attribute *ext4_attrs[] = {
2540 ATTR_LIST(delayed_allocation_blocks),
2541 ATTR_LIST(session_write_kbytes),
2542 ATTR_LIST(lifetime_write_kbytes),
2543 ATTR_LIST(inode_readahead_blks),
2544 ATTR_LIST(inode_goal),
2545 ATTR_LIST(mb_stats),
2546 ATTR_LIST(mb_max_to_scan),
2547 ATTR_LIST(mb_min_to_scan),
2548 ATTR_LIST(mb_order2_req),
2549 ATTR_LIST(mb_stream_req),
2550 ATTR_LIST(mb_group_prealloc),
2551 ATTR_LIST(max_writeback_mb_bump),
2552 ATTR_LIST(trigger_fs_error),
2556 /* Features this copy of ext4 supports */
2557 EXT4_INFO_ATTR(lazy_itable_init);
2558 EXT4_INFO_ATTR(batched_discard);
2560 static struct attribute *ext4_feat_attrs[] = {
2561 ATTR_LIST(lazy_itable_init),
2562 ATTR_LIST(batched_discard),
2566 static ssize_t ext4_attr_show(struct kobject *kobj,
2567 struct attribute *attr, char *buf)
2569 struct ext4_sb_info *sbi = container_of(kobj, struct ext4_sb_info,
2571 struct ext4_attr *a = container_of(attr, struct ext4_attr, attr);
2573 return a->show ? a->show(a, sbi, buf) : 0;
2576 static ssize_t ext4_attr_store(struct kobject *kobj,
2577 struct attribute *attr,
2578 const char *buf, size_t len)
2580 struct ext4_sb_info *sbi = container_of(kobj, struct ext4_sb_info,
2582 struct ext4_attr *a = container_of(attr, struct ext4_attr, attr);
2584 return a->store ? a->store(a, sbi, buf, len) : 0;
2587 static void ext4_sb_release(struct kobject *kobj)
2589 struct ext4_sb_info *sbi = container_of(kobj, struct ext4_sb_info,
2591 complete(&sbi->s_kobj_unregister);
2594 static const struct sysfs_ops ext4_attr_ops = {
2595 .show = ext4_attr_show,
2596 .store = ext4_attr_store,
2599 static struct kobj_type ext4_ktype = {
2600 .default_attrs = ext4_attrs,
2601 .sysfs_ops = &ext4_attr_ops,
2602 .release = ext4_sb_release,
2605 static void ext4_feat_release(struct kobject *kobj)
2607 complete(&ext4_feat->f_kobj_unregister);
2610 static struct kobj_type ext4_feat_ktype = {
2611 .default_attrs = ext4_feat_attrs,
2612 .sysfs_ops = &ext4_attr_ops,
2613 .release = ext4_feat_release,
2617 * Check whether this filesystem can be mounted based on
2618 * the features present and the RDONLY/RDWR mount requested.
2619 * Returns 1 if this filesystem can be mounted as requested,
2620 * 0 if it cannot be.
2622 static int ext4_feature_set_ok(struct super_block *sb, int readonly)
2624 if (EXT4_HAS_INCOMPAT_FEATURE(sb, ~EXT4_FEATURE_INCOMPAT_SUPP)) {
2625 ext4_msg(sb, KERN_ERR,
2626 "Couldn't mount because of "
2627 "unsupported optional features (%x)",
2628 (le32_to_cpu(EXT4_SB(sb)->s_es->s_feature_incompat) &
2629 ~EXT4_FEATURE_INCOMPAT_SUPP));
2636 /* Check that feature set is OK for a read-write mount */
2637 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, ~EXT4_FEATURE_RO_COMPAT_SUPP)) {
2638 ext4_msg(sb, KERN_ERR, "couldn't mount RDWR because of "
2639 "unsupported optional features (%x)",
2640 (le32_to_cpu(EXT4_SB(sb)->s_es->s_feature_ro_compat) &
2641 ~EXT4_FEATURE_RO_COMPAT_SUPP));
2645 * Large file size enabled file system can only be mounted
2646 * read-write on 32-bit systems if kernel is built with CONFIG_LBDAF
2648 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_HUGE_FILE)) {
2649 if (sizeof(blkcnt_t) < sizeof(u64)) {
2650 ext4_msg(sb, KERN_ERR, "Filesystem with huge files "
2651 "cannot be mounted RDWR without "
2656 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_BIGALLOC) &&
2657 !EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_EXTENTS)) {
2658 ext4_msg(sb, KERN_ERR,
2659 "Can't support bigalloc feature without "
2660 "extents feature\n");
2664 #ifndef CONFIG_QUOTA
2665 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_QUOTA) &&
2667 ext4_msg(sb, KERN_ERR,
2668 "Filesystem with quota feature cannot be mounted RDWR "
2669 "without CONFIG_QUOTA");
2672 #endif /* CONFIG_QUOTA */
2677 * This function is called once a day if we have errors logged
2678 * on the file system
2680 static void print_daily_error_info(unsigned long arg)
2682 struct super_block *sb = (struct super_block *) arg;
2683 struct ext4_sb_info *sbi;
2684 struct ext4_super_block *es;
2689 if (es->s_error_count)
2690 ext4_msg(sb, KERN_NOTICE, "error count: %u",
2691 le32_to_cpu(es->s_error_count));
2692 if (es->s_first_error_time) {
2693 printk(KERN_NOTICE "EXT4-fs (%s): initial error at %u: %.*s:%d",
2694 sb->s_id, le32_to_cpu(es->s_first_error_time),
2695 (int) sizeof(es->s_first_error_func),
2696 es->s_first_error_func,
2697 le32_to_cpu(es->s_first_error_line));
2698 if (es->s_first_error_ino)
2699 printk(": inode %u",
2700 le32_to_cpu(es->s_first_error_ino));
2701 if (es->s_first_error_block)
2702 printk(": block %llu", (unsigned long long)
2703 le64_to_cpu(es->s_first_error_block));
2706 if (es->s_last_error_time) {
2707 printk(KERN_NOTICE "EXT4-fs (%s): last error at %u: %.*s:%d",
2708 sb->s_id, le32_to_cpu(es->s_last_error_time),
2709 (int) sizeof(es->s_last_error_func),
2710 es->s_last_error_func,
2711 le32_to_cpu(es->s_last_error_line));
2712 if (es->s_last_error_ino)
2713 printk(": inode %u",
2714 le32_to_cpu(es->s_last_error_ino));
2715 if (es->s_last_error_block)
2716 printk(": block %llu", (unsigned long long)
2717 le64_to_cpu(es->s_last_error_block));
2720 mod_timer(&sbi->s_err_report, jiffies + 24*60*60*HZ); /* Once a day */
2723 /* Find next suitable group and run ext4_init_inode_table */
2724 static int ext4_run_li_request(struct ext4_li_request *elr)
2726 struct ext4_group_desc *gdp = NULL;
2727 ext4_group_t group, ngroups;
2728 struct super_block *sb;
2729 unsigned long timeout = 0;
2733 ngroups = EXT4_SB(sb)->s_groups_count;
2736 for (group = elr->lr_next_group; group < ngroups; group++) {
2737 gdp = ext4_get_group_desc(sb, group, NULL);
2743 if (!(gdp->bg_flags & cpu_to_le16(EXT4_BG_INODE_ZEROED)))
2747 if (group == ngroups)
2752 ret = ext4_init_inode_table(sb, group,
2753 elr->lr_timeout ? 0 : 1);
2754 if (elr->lr_timeout == 0) {
2755 timeout = (jiffies - timeout) *
2756 elr->lr_sbi->s_li_wait_mult;
2757 elr->lr_timeout = timeout;
2759 elr->lr_next_sched = jiffies + elr->lr_timeout;
2760 elr->lr_next_group = group + 1;
2768 * Remove lr_request from the list_request and free the
2769 * request structure. Should be called with li_list_mtx held
2771 static void ext4_remove_li_request(struct ext4_li_request *elr)
2773 struct ext4_sb_info *sbi;
2780 list_del(&elr->lr_request);
2781 sbi->s_li_request = NULL;
2785 static void ext4_unregister_li_request(struct super_block *sb)
2787 mutex_lock(&ext4_li_mtx);
2788 if (!ext4_li_info) {
2789 mutex_unlock(&ext4_li_mtx);
2793 mutex_lock(&ext4_li_info->li_list_mtx);
2794 ext4_remove_li_request(EXT4_SB(sb)->s_li_request);
2795 mutex_unlock(&ext4_li_info->li_list_mtx);
2796 mutex_unlock(&ext4_li_mtx);
2799 static struct task_struct *ext4_lazyinit_task;
2802 * This is the function where ext4lazyinit thread lives. It walks
2803 * through the request list searching for next scheduled filesystem.
2804 * When such a fs is found, run the lazy initialization request
2805 * (ext4_rn_li_request) and keep track of the time spend in this
2806 * function. Based on that time we compute next schedule time of
2807 * the request. When walking through the list is complete, compute
2808 * next waking time and put itself into sleep.
2810 static int ext4_lazyinit_thread(void *arg)
2812 struct ext4_lazy_init *eli = (struct ext4_lazy_init *)arg;
2813 struct list_head *pos, *n;
2814 struct ext4_li_request *elr;
2815 unsigned long next_wakeup, cur;
2817 BUG_ON(NULL == eli);
2821 next_wakeup = MAX_JIFFY_OFFSET;
2823 mutex_lock(&eli->li_list_mtx);
2824 if (list_empty(&eli->li_request_list)) {
2825 mutex_unlock(&eli->li_list_mtx);
2829 list_for_each_safe(pos, n, &eli->li_request_list) {
2830 elr = list_entry(pos, struct ext4_li_request,
2833 if (time_after_eq(jiffies, elr->lr_next_sched)) {
2834 if (ext4_run_li_request(elr) != 0) {
2835 /* error, remove the lazy_init job */
2836 ext4_remove_li_request(elr);
2841 if (time_before(elr->lr_next_sched, next_wakeup))
2842 next_wakeup = elr->lr_next_sched;
2844 mutex_unlock(&eli->li_list_mtx);
2849 if ((time_after_eq(cur, next_wakeup)) ||
2850 (MAX_JIFFY_OFFSET == next_wakeup)) {
2855 schedule_timeout_interruptible(next_wakeup - cur);
2857 if (kthread_should_stop()) {
2858 ext4_clear_request_list();
2865 * It looks like the request list is empty, but we need
2866 * to check it under the li_list_mtx lock, to prevent any
2867 * additions into it, and of course we should lock ext4_li_mtx
2868 * to atomically free the list and ext4_li_info, because at
2869 * this point another ext4 filesystem could be registering
2872 mutex_lock(&ext4_li_mtx);
2873 mutex_lock(&eli->li_list_mtx);
2874 if (!list_empty(&eli->li_request_list)) {
2875 mutex_unlock(&eli->li_list_mtx);
2876 mutex_unlock(&ext4_li_mtx);
2879 mutex_unlock(&eli->li_list_mtx);
2880 kfree(ext4_li_info);
2881 ext4_li_info = NULL;
2882 mutex_unlock(&ext4_li_mtx);
2887 static void ext4_clear_request_list(void)
2889 struct list_head *pos, *n;
2890 struct ext4_li_request *elr;
2892 mutex_lock(&ext4_li_info->li_list_mtx);
2893 list_for_each_safe(pos, n, &ext4_li_info->li_request_list) {
2894 elr = list_entry(pos, struct ext4_li_request,
2896 ext4_remove_li_request(elr);
2898 mutex_unlock(&ext4_li_info->li_list_mtx);
2901 static int ext4_run_lazyinit_thread(void)
2903 ext4_lazyinit_task = kthread_run(ext4_lazyinit_thread,
2904 ext4_li_info, "ext4lazyinit");
2905 if (IS_ERR(ext4_lazyinit_task)) {
2906 int err = PTR_ERR(ext4_lazyinit_task);
2907 ext4_clear_request_list();
2908 kfree(ext4_li_info);
2909 ext4_li_info = NULL;
2910 printk(KERN_CRIT "EXT4-fs: error %d creating inode table "
2911 "initialization thread\n",
2915 ext4_li_info->li_state |= EXT4_LAZYINIT_RUNNING;
2920 * Check whether it make sense to run itable init. thread or not.
2921 * If there is at least one uninitialized inode table, return
2922 * corresponding group number, else the loop goes through all
2923 * groups and return total number of groups.
2925 static ext4_group_t ext4_has_uninit_itable(struct super_block *sb)
2927 ext4_group_t group, ngroups = EXT4_SB(sb)->s_groups_count;
2928 struct ext4_group_desc *gdp = NULL;
2930 for (group = 0; group < ngroups; group++) {
2931 gdp = ext4_get_group_desc(sb, group, NULL);
2935 if (!(gdp->bg_flags & cpu_to_le16(EXT4_BG_INODE_ZEROED)))
2942 static int ext4_li_info_new(void)
2944 struct ext4_lazy_init *eli = NULL;
2946 eli = kzalloc(sizeof(*eli), GFP_KERNEL);
2950 INIT_LIST_HEAD(&eli->li_request_list);
2951 mutex_init(&eli->li_list_mtx);
2953 eli->li_state |= EXT4_LAZYINIT_QUIT;
2960 static struct ext4_li_request *ext4_li_request_new(struct super_block *sb,
2963 struct ext4_sb_info *sbi = EXT4_SB(sb);
2964 struct ext4_li_request *elr;
2967 elr = kzalloc(sizeof(*elr), GFP_KERNEL);
2973 elr->lr_next_group = start;
2976 * Randomize first schedule time of the request to
2977 * spread the inode table initialization requests
2980 get_random_bytes(&rnd, sizeof(rnd));
2981 elr->lr_next_sched = jiffies + (unsigned long)rnd %
2982 (EXT4_DEF_LI_MAX_START_DELAY * HZ);
2987 static int ext4_register_li_request(struct super_block *sb,
2988 ext4_group_t first_not_zeroed)
2990 struct ext4_sb_info *sbi = EXT4_SB(sb);
2991 struct ext4_li_request *elr;
2992 ext4_group_t ngroups = EXT4_SB(sb)->s_groups_count;
2995 if (sbi->s_li_request != NULL) {
2997 * Reset timeout so it can be computed again, because
2998 * s_li_wait_mult might have changed.
3000 sbi->s_li_request->lr_timeout = 0;
3004 if (first_not_zeroed == ngroups ||
3005 (sb->s_flags & MS_RDONLY) ||
3006 !test_opt(sb, INIT_INODE_TABLE))
3009 elr = ext4_li_request_new(sb, first_not_zeroed);
3013 mutex_lock(&ext4_li_mtx);
3015 if (NULL == ext4_li_info) {
3016 ret = ext4_li_info_new();
3021 mutex_lock(&ext4_li_info->li_list_mtx);
3022 list_add(&elr->lr_request, &ext4_li_info->li_request_list);
3023 mutex_unlock(&ext4_li_info->li_list_mtx);
3025 sbi->s_li_request = elr;
3027 * set elr to NULL here since it has been inserted to
3028 * the request_list and the removal and free of it is
3029 * handled by ext4_clear_request_list from now on.
3033 if (!(ext4_li_info->li_state & EXT4_LAZYINIT_RUNNING)) {
3034 ret = ext4_run_lazyinit_thread();
3039 mutex_unlock(&ext4_li_mtx);
3046 * We do not need to lock anything since this is called on
3049 static void ext4_destroy_lazyinit_thread(void)
3052 * If thread exited earlier
3053 * there's nothing to be done.
3055 if (!ext4_li_info || !ext4_lazyinit_task)
3058 kthread_stop(ext4_lazyinit_task);
3061 static int set_journal_csum_feature_set(struct super_block *sb)
3064 int compat, incompat;
3065 struct ext4_sb_info *sbi = EXT4_SB(sb);
3067 if (EXT4_HAS_RO_COMPAT_FEATURE(sb,
3068 EXT4_FEATURE_RO_COMPAT_METADATA_CSUM)) {
3069 /* journal checksum v2 */
3071 incompat = JBD2_FEATURE_INCOMPAT_CSUM_V2;
3073 /* journal checksum v1 */
3074 compat = JBD2_FEATURE_COMPAT_CHECKSUM;
3078 if (test_opt(sb, JOURNAL_ASYNC_COMMIT)) {
3079 ret = jbd2_journal_set_features(sbi->s_journal,
3081 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT |
3083 } else if (test_opt(sb, JOURNAL_CHECKSUM)) {
3084 ret = jbd2_journal_set_features(sbi->s_journal,
3087 jbd2_journal_clear_features(sbi->s_journal, 0, 0,
3088 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT);
3090 jbd2_journal_clear_features(sbi->s_journal,
3091 JBD2_FEATURE_COMPAT_CHECKSUM, 0,
3092 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT |
3093 JBD2_FEATURE_INCOMPAT_CSUM_V2);
3100 * Note: calculating the overhead so we can be compatible with
3101 * historical BSD practice is quite difficult in the face of
3102 * clusters/bigalloc. This is because multiple metadata blocks from
3103 * different block group can end up in the same allocation cluster.
3104 * Calculating the exact overhead in the face of clustered allocation
3105 * requires either O(all block bitmaps) in memory or O(number of block
3106 * groups**2) in time. We will still calculate the superblock for
3107 * older file systems --- and if we come across with a bigalloc file
3108 * system with zero in s_overhead_clusters the estimate will be close to
3109 * correct especially for very large cluster sizes --- but for newer
3110 * file systems, it's better to calculate this figure once at mkfs
3111 * time, and store it in the superblock. If the superblock value is
3112 * present (even for non-bigalloc file systems), we will use it.
3114 static int count_overhead(struct super_block *sb, ext4_group_t grp,
3117 struct ext4_sb_info *sbi = EXT4_SB(sb);
3118 struct ext4_group_desc *gdp;
3119 ext4_fsblk_t first_block, last_block, b;
3120 ext4_group_t i, ngroups = ext4_get_groups_count(sb);
3121 int s, j, count = 0;
3123 if (!EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_BIGALLOC))
3124 return (ext4_bg_has_super(sb, grp) + ext4_bg_num_gdb(sb, grp) +
3125 sbi->s_itb_per_group + 2);
3127 first_block = le32_to_cpu(sbi->s_es->s_first_data_block) +
3128 (grp * EXT4_BLOCKS_PER_GROUP(sb));
3129 last_block = first_block + EXT4_BLOCKS_PER_GROUP(sb) - 1;
3130 for (i = 0; i < ngroups; i++) {
3131 gdp = ext4_get_group_desc(sb, i, NULL);
3132 b = ext4_block_bitmap(sb, gdp);
3133 if (b >= first_block && b <= last_block) {
3134 ext4_set_bit(EXT4_B2C(sbi, b - first_block), buf);
3137 b = ext4_inode_bitmap(sb, gdp);
3138 if (b >= first_block && b <= last_block) {
3139 ext4_set_bit(EXT4_B2C(sbi, b - first_block), buf);
3142 b = ext4_inode_table(sb, gdp);
3143 if (b >= first_block && b + sbi->s_itb_per_group <= last_block)
3144 for (j = 0; j < sbi->s_itb_per_group; j++, b++) {
3145 int c = EXT4_B2C(sbi, b - first_block);
3146 ext4_set_bit(c, buf);
3152 if (ext4_bg_has_super(sb, grp)) {
3153 ext4_set_bit(s++, buf);
3156 for (j = ext4_bg_num_gdb(sb, grp); j > 0; j--) {
3157 ext4_set_bit(EXT4_B2C(sbi, s++), buf);
3163 return EXT4_CLUSTERS_PER_GROUP(sb) -
3164 ext4_count_free(buf, EXT4_CLUSTERS_PER_GROUP(sb) / 8);
3168 * Compute the overhead and stash it in sbi->s_overhead
3170 int ext4_calculate_overhead(struct super_block *sb)
3172 struct ext4_sb_info *sbi = EXT4_SB(sb);
3173 struct ext4_super_block *es = sbi->s_es;
3174 ext4_group_t i, ngroups = ext4_get_groups_count(sb);
3175 ext4_fsblk_t overhead = 0;
3176 char *buf = (char *) get_zeroed_page(GFP_KERNEL);
3178 memset(buf, 0, PAGE_SIZE);
3183 * Compute the overhead (FS structures). This is constant
3184 * for a given filesystem unless the number of block groups
3185 * changes so we cache the previous value until it does.
3189 * All of the blocks before first_data_block are overhead
3191 overhead = EXT4_B2C(sbi, le32_to_cpu(es->s_first_data_block));
3194 * Add the overhead found in each block group
3196 for (i = 0; i < ngroups; i++) {
3199 blks = count_overhead(sb, i, buf);
3202 memset(buf, 0, PAGE_SIZE);
3205 sbi->s_overhead = overhead;
3207 free_page((unsigned long) buf);
3211 static int ext4_fill_super(struct super_block *sb, void *data, int silent)
3213 char *orig_data = kstrdup(data, GFP_KERNEL);
3214 struct buffer_head *bh;
3215 struct ext4_super_block *es = NULL;
3216 struct ext4_sb_info *sbi;
3218 ext4_fsblk_t sb_block = get_sb_block(&data);
3219 ext4_fsblk_t logical_sb_block;
3220 unsigned long offset = 0;
3221 unsigned long journal_devnum = 0;
3222 unsigned long def_mount_opts;
3227 int blocksize, clustersize;
3228 unsigned int db_count;
3230 int needs_recovery, has_huge_files, has_bigalloc;
3233 unsigned int journal_ioprio = DEFAULT_JOURNAL_IOPRIO;
3234 ext4_group_t first_not_zeroed;
3236 sbi = kzalloc(sizeof(*sbi), GFP_KERNEL);
3240 sbi->s_blockgroup_lock =
3241 kzalloc(sizeof(struct blockgroup_lock), GFP_KERNEL);
3242 if (!sbi->s_blockgroup_lock) {
3246 sb->s_fs_info = sbi;
3248 sbi->s_mount_opt = 0;
3249 sbi->s_resuid = make_kuid(&init_user_ns, EXT4_DEF_RESUID);
3250 sbi->s_resgid = make_kgid(&init_user_ns, EXT4_DEF_RESGID);
3251 sbi->s_inode_readahead_blks = EXT4_DEF_INODE_READAHEAD_BLKS;
3252 sbi->s_sb_block = sb_block;
3253 if (sb->s_bdev->bd_part)
3254 sbi->s_sectors_written_start =
3255 part_stat_read(sb->s_bdev->bd_part, sectors[1]);
3257 /* Cleanup superblock name */
3258 for (cp = sb->s_id; (cp = strchr(cp, '/'));)
3262 blocksize = sb_min_blocksize(sb, EXT4_MIN_BLOCK_SIZE);
3264 ext4_msg(sb, KERN_ERR, "unable to set blocksize");
3269 * The ext4 superblock will not be buffer aligned for other than 1kB
3270 * block sizes. We need to calculate the offset from buffer start.
3272 if (blocksize != EXT4_MIN_BLOCK_SIZE) {
3273 logical_sb_block = sb_block * EXT4_MIN_BLOCK_SIZE;
3274 offset = do_div(logical_sb_block, blocksize);
3276 logical_sb_block = sb_block;
3279 if (!(bh = sb_bread(sb, logical_sb_block))) {
3280 ext4_msg(sb, KERN_ERR, "unable to read superblock");
3284 * Note: s_es must be initialized as soon as possible because
3285 * some ext4 macro-instructions depend on its value
3287 es = (struct ext4_super_block *) (bh->b_data + offset);
3289 sb->s_magic = le16_to_cpu(es->s_magic);
3290 if (sb->s_magic != EXT4_SUPER_MAGIC)
3292 sbi->s_kbytes_written = le64_to_cpu(es->s_kbytes_written);
3294 /* Warn if metadata_csum and gdt_csum are both set. */
3295 if (EXT4_HAS_RO_COMPAT_FEATURE(sb,
3296 EXT4_FEATURE_RO_COMPAT_METADATA_CSUM) &&
3297 EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_GDT_CSUM))
3298 ext4_warning(sb, KERN_INFO "metadata_csum and uninit_bg are "
3299 "redundant flags; please run fsck.");
3301 /* Check for a known checksum algorithm */
3302 if (!ext4_verify_csum_type(sb, es)) {
3303 ext4_msg(sb, KERN_ERR, "VFS: Found ext4 filesystem with "
3304 "unknown checksum algorithm.");
3309 /* Load the checksum driver */
3310 if (EXT4_HAS_RO_COMPAT_FEATURE(sb,
3311 EXT4_FEATURE_RO_COMPAT_METADATA_CSUM)) {
3312 sbi->s_chksum_driver = crypto_alloc_shash("crc32c", 0, 0);
3313 if (IS_ERR(sbi->s_chksum_driver)) {
3314 ext4_msg(sb, KERN_ERR, "Cannot load crc32c driver.");
3315 ret = PTR_ERR(sbi->s_chksum_driver);
3316 sbi->s_chksum_driver = NULL;
3321 /* Check superblock checksum */
3322 if (!ext4_superblock_csum_verify(sb, es)) {
3323 ext4_msg(sb, KERN_ERR, "VFS: Found ext4 filesystem with "
3324 "invalid superblock checksum. Run e2fsck?");
3329 /* Precompute checksum seed for all metadata */
3330 if (EXT4_HAS_RO_COMPAT_FEATURE(sb,
3331 EXT4_FEATURE_RO_COMPAT_METADATA_CSUM))
3332 sbi->s_csum_seed = ext4_chksum(sbi, ~0, es->s_uuid,
3333 sizeof(es->s_uuid));
3335 /* Set defaults before we parse the mount options */
3336 def_mount_opts = le32_to_cpu(es->s_default_mount_opts);
3337 set_opt(sb, INIT_INODE_TABLE);
3338 if (def_mount_opts & EXT4_DEFM_DEBUG)
3340 if (def_mount_opts & EXT4_DEFM_BSDGROUPS)
3342 if (def_mount_opts & EXT4_DEFM_UID16)
3343 set_opt(sb, NO_UID32);
3344 /* xattr user namespace & acls are now defaulted on */
3345 #ifdef CONFIG_EXT4_FS_XATTR
3346 set_opt(sb, XATTR_USER);
3348 #ifdef CONFIG_EXT4_FS_POSIX_ACL
3349 set_opt(sb, POSIX_ACL);
3351 set_opt(sb, MBLK_IO_SUBMIT);
3352 if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_DATA)
3353 set_opt(sb, JOURNAL_DATA);
3354 else if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_ORDERED)
3355 set_opt(sb, ORDERED_DATA);
3356 else if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_WBACK)
3357 set_opt(sb, WRITEBACK_DATA);
3359 if (le16_to_cpu(sbi->s_es->s_errors) == EXT4_ERRORS_PANIC)
3360 set_opt(sb, ERRORS_PANIC);
3361 else if (le16_to_cpu(sbi->s_es->s_errors) == EXT4_ERRORS_CONTINUE)
3362 set_opt(sb, ERRORS_CONT);
3364 set_opt(sb, ERRORS_RO);
3365 if (def_mount_opts & EXT4_DEFM_BLOCK_VALIDITY)
3366 set_opt(sb, BLOCK_VALIDITY);
3367 if (def_mount_opts & EXT4_DEFM_DISCARD)
3368 set_opt(sb, DISCARD);
3370 sbi->s_resuid = make_kuid(&init_user_ns, le16_to_cpu(es->s_def_resuid));
3371 sbi->s_resgid = make_kgid(&init_user_ns, le16_to_cpu(es->s_def_resgid));
3372 sbi->s_commit_interval = JBD2_DEFAULT_MAX_COMMIT_AGE * HZ;
3373 sbi->s_min_batch_time = EXT4_DEF_MIN_BATCH_TIME;
3374 sbi->s_max_batch_time = EXT4_DEF_MAX_BATCH_TIME;
3376 if ((def_mount_opts & EXT4_DEFM_NOBARRIER) == 0)
3377 set_opt(sb, BARRIER);
3380 * enable delayed allocation by default
3381 * Use -o nodelalloc to turn it off
3383 if (!IS_EXT3_SB(sb) &&
3384 ((def_mount_opts & EXT4_DEFM_NODELALLOC) == 0))
3385 set_opt(sb, DELALLOC);
3388 * set default s_li_wait_mult for lazyinit, for the case there is
3389 * no mount option specified.
3391 sbi->s_li_wait_mult = EXT4_DEF_LI_WAIT_MULT;
3393 if (!parse_options((char *) sbi->s_es->s_mount_opts, sb,
3394 &journal_devnum, &journal_ioprio, 0)) {
3395 ext4_msg(sb, KERN_WARNING,
3396 "failed to parse options in superblock: %s",
3397 sbi->s_es->s_mount_opts);
3399 sbi->s_def_mount_opt = sbi->s_mount_opt;
3400 if (!parse_options((char *) data, sb, &journal_devnum,
3401 &journal_ioprio, 0))
3404 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA) {
3405 printk_once(KERN_WARNING "EXT4-fs: Warning: mounting "
3406 "with data=journal disables delayed "
3407 "allocation and O_DIRECT support!\n");
3408 if (test_opt2(sb, EXPLICIT_DELALLOC)) {
3409 ext4_msg(sb, KERN_ERR, "can't mount with "
3410 "both data=journal and delalloc");
3413 if (test_opt(sb, DIOREAD_NOLOCK)) {
3414 ext4_msg(sb, KERN_ERR, "can't mount with "
3415 "both data=journal and delalloc");
3418 if (test_opt(sb, DELALLOC))
3419 clear_opt(sb, DELALLOC);
3422 blocksize = BLOCK_SIZE << le32_to_cpu(es->s_log_block_size);
3423 if (test_opt(sb, DIOREAD_NOLOCK)) {
3424 if (blocksize < PAGE_SIZE) {
3425 ext4_msg(sb, KERN_ERR, "can't mount with "
3426 "dioread_nolock if block size != PAGE_SIZE");
3431 sb->s_flags = (sb->s_flags & ~MS_POSIXACL) |
3432 (test_opt(sb, POSIX_ACL) ? MS_POSIXACL : 0);
3434 if (le32_to_cpu(es->s_rev_level) == EXT4_GOOD_OLD_REV &&
3435 (EXT4_HAS_COMPAT_FEATURE(sb, ~0U) ||
3436 EXT4_HAS_RO_COMPAT_FEATURE(sb, ~0U) ||
3437 EXT4_HAS_INCOMPAT_FEATURE(sb, ~0U)))
3438 ext4_msg(sb, KERN_WARNING,
3439 "feature flags set on rev 0 fs, "
3440 "running e2fsck is recommended");
3442 if (IS_EXT2_SB(sb)) {
3443 if (ext2_feature_set_ok(sb))
3444 ext4_msg(sb, KERN_INFO, "mounting ext2 file system "
3445 "using the ext4 subsystem");
3447 ext4_msg(sb, KERN_ERR, "couldn't mount as ext2 due "
3448 "to feature incompatibilities");
3453 if (IS_EXT3_SB(sb)) {
3454 if (ext3_feature_set_ok(sb))
3455 ext4_msg(sb, KERN_INFO, "mounting ext3 file system "
3456 "using the ext4 subsystem");
3458 ext4_msg(sb, KERN_ERR, "couldn't mount as ext3 due "
3459 "to feature incompatibilities");
3465 * Check feature flags regardless of the revision level, since we
3466 * previously didn't change the revision level when setting the flags,
3467 * so there is a chance incompat flags are set on a rev 0 filesystem.
3469 if (!ext4_feature_set_ok(sb, (sb->s_flags & MS_RDONLY)))
3472 if (blocksize < EXT4_MIN_BLOCK_SIZE ||
3473 blocksize > EXT4_MAX_BLOCK_SIZE) {
3474 ext4_msg(sb, KERN_ERR,
3475 "Unsupported filesystem blocksize %d", blocksize);
3479 if (sb->s_blocksize != blocksize) {
3480 /* Validate the filesystem blocksize */
3481 if (!sb_set_blocksize(sb, blocksize)) {
3482 ext4_msg(sb, KERN_ERR, "bad block size %d",
3488 logical_sb_block = sb_block * EXT4_MIN_BLOCK_SIZE;
3489 offset = do_div(logical_sb_block, blocksize);
3490 bh = sb_bread(sb, logical_sb_block);
3492 ext4_msg(sb, KERN_ERR,
3493 "Can't read superblock on 2nd try");
3496 es = (struct ext4_super_block *)(bh->b_data + offset);
3498 if (es->s_magic != cpu_to_le16(EXT4_SUPER_MAGIC)) {
3499 ext4_msg(sb, KERN_ERR,
3500 "Magic mismatch, very weird!");
3505 has_huge_files = EXT4_HAS_RO_COMPAT_FEATURE(sb,
3506 EXT4_FEATURE_RO_COMPAT_HUGE_FILE);
3507 sbi->s_bitmap_maxbytes = ext4_max_bitmap_size(sb->s_blocksize_bits,
3509 sb->s_maxbytes = ext4_max_size(sb->s_blocksize_bits, has_huge_files);
3511 if (le32_to_cpu(es->s_rev_level) == EXT4_GOOD_OLD_REV) {
3512 sbi->s_inode_size = EXT4_GOOD_OLD_INODE_SIZE;
3513 sbi->s_first_ino = EXT4_GOOD_OLD_FIRST_INO;
3515 sbi->s_inode_size = le16_to_cpu(es->s_inode_size);
3516 sbi->s_first_ino = le32_to_cpu(es->s_first_ino);
3517 if ((sbi->s_inode_size < EXT4_GOOD_OLD_INODE_SIZE) ||
3518 (!is_power_of_2(sbi->s_inode_size)) ||
3519 (sbi->s_inode_size > blocksize)) {
3520 ext4_msg(sb, KERN_ERR,
3521 "unsupported inode size: %d",
3525 if (sbi->s_inode_size > EXT4_GOOD_OLD_INODE_SIZE)
3526 sb->s_time_gran = 1 << (EXT4_EPOCH_BITS - 2);
3529 sbi->s_desc_size = le16_to_cpu(es->s_desc_size);
3530 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_64BIT)) {
3531 if (sbi->s_desc_size < EXT4_MIN_DESC_SIZE_64BIT ||
3532 sbi->s_desc_size > EXT4_MAX_DESC_SIZE ||
3533 !is_power_of_2(sbi->s_desc_size)) {
3534 ext4_msg(sb, KERN_ERR,
3535 "unsupported descriptor size %lu",
3540 sbi->s_desc_size = EXT4_MIN_DESC_SIZE;
3542 sbi->s_blocks_per_group = le32_to_cpu(es->s_blocks_per_group);
3543 sbi->s_inodes_per_group = le32_to_cpu(es->s_inodes_per_group);
3544 if (EXT4_INODE_SIZE(sb) == 0 || EXT4_INODES_PER_GROUP(sb) == 0)
3547 sbi->s_inodes_per_block = blocksize / EXT4_INODE_SIZE(sb);
3548 if (sbi->s_inodes_per_block == 0)
3550 sbi->s_itb_per_group = sbi->s_inodes_per_group /
3551 sbi->s_inodes_per_block;
3552 sbi->s_desc_per_block = blocksize / EXT4_DESC_SIZE(sb);
3554 sbi->s_mount_state = le16_to_cpu(es->s_state);
3555 sbi->s_addr_per_block_bits = ilog2(EXT4_ADDR_PER_BLOCK(sb));
3556 sbi->s_desc_per_block_bits = ilog2(EXT4_DESC_PER_BLOCK(sb));
3558 for (i = 0; i < 4; i++)
3559 sbi->s_hash_seed[i] = le32_to_cpu(es->s_hash_seed[i]);
3560 sbi->s_def_hash_version = es->s_def_hash_version;
3561 i = le32_to_cpu(es->s_flags);
3562 if (i & EXT2_FLAGS_UNSIGNED_HASH)
3563 sbi->s_hash_unsigned = 3;
3564 else if ((i & EXT2_FLAGS_SIGNED_HASH) == 0) {
3565 #ifdef __CHAR_UNSIGNED__
3566 es->s_flags |= cpu_to_le32(EXT2_FLAGS_UNSIGNED_HASH);
3567 sbi->s_hash_unsigned = 3;
3569 es->s_flags |= cpu_to_le32(EXT2_FLAGS_SIGNED_HASH);
3573 /* Handle clustersize */
3574 clustersize = BLOCK_SIZE << le32_to_cpu(es->s_log_cluster_size);
3575 has_bigalloc = EXT4_HAS_RO_COMPAT_FEATURE(sb,
3576 EXT4_FEATURE_RO_COMPAT_BIGALLOC);
3578 if (clustersize < blocksize) {
3579 ext4_msg(sb, KERN_ERR,
3580 "cluster size (%d) smaller than "
3581 "block size (%d)", clustersize, blocksize);
3584 sbi->s_cluster_bits = le32_to_cpu(es->s_log_cluster_size) -
3585 le32_to_cpu(es->s_log_block_size);
3586 sbi->s_clusters_per_group =
3587 le32_to_cpu(es->s_clusters_per_group);
3588 if (sbi->s_clusters_per_group > blocksize * 8) {
3589 ext4_msg(sb, KERN_ERR,
3590 "#clusters per group too big: %lu",
3591 sbi->s_clusters_per_group);
3594 if (sbi->s_blocks_per_group !=
3595 (sbi->s_clusters_per_group * (clustersize / blocksize))) {
3596 ext4_msg(sb, KERN_ERR, "blocks per group (%lu) and "
3597 "clusters per group (%lu) inconsistent",
3598 sbi->s_blocks_per_group,
3599 sbi->s_clusters_per_group);
3603 if (clustersize != blocksize) {
3604 ext4_warning(sb, "fragment/cluster size (%d) != "
3605 "block size (%d)", clustersize,
3607 clustersize = blocksize;
3609 if (sbi->s_blocks_per_group > blocksize * 8) {
3610 ext4_msg(sb, KERN_ERR,
3611 "#blocks per group too big: %lu",
3612 sbi->s_blocks_per_group);
3615 sbi->s_clusters_per_group = sbi->s_blocks_per_group;
3616 sbi->s_cluster_bits = 0;
3618 sbi->s_cluster_ratio = clustersize / blocksize;
3620 if (sbi->s_inodes_per_group > blocksize * 8) {
3621 ext4_msg(sb, KERN_ERR,
3622 "#inodes per group too big: %lu",
3623 sbi->s_inodes_per_group);
3628 * Test whether we have more sectors than will fit in sector_t,
3629 * and whether the max offset is addressable by the page cache.
3631 err = generic_check_addressable(sb->s_blocksize_bits,
3632 ext4_blocks_count(es));
3634 ext4_msg(sb, KERN_ERR, "filesystem"
3635 " too large to mount safely on this system");
3636 if (sizeof(sector_t) < 8)
3637 ext4_msg(sb, KERN_WARNING, "CONFIG_LBDAF not enabled");
3642 if (EXT4_BLOCKS_PER_GROUP(sb) == 0)
3645 /* check blocks count against device size */
3646 blocks_count = sb->s_bdev->bd_inode->i_size >> sb->s_blocksize_bits;
3647 if (blocks_count && ext4_blocks_count(es) > blocks_count) {
3648 ext4_msg(sb, KERN_WARNING, "bad geometry: block count %llu "
3649 "exceeds size of device (%llu blocks)",
3650 ext4_blocks_count(es), blocks_count);
3655 * It makes no sense for the first data block to be beyond the end
3656 * of the filesystem.
3658 if (le32_to_cpu(es->s_first_data_block) >= ext4_blocks_count(es)) {
3659 ext4_msg(sb, KERN_WARNING, "bad geometry: first data "
3660 "block %u is beyond end of filesystem (%llu)",
3661 le32_to_cpu(es->s_first_data_block),
3662 ext4_blocks_count(es));
3665 blocks_count = (ext4_blocks_count(es) -
3666 le32_to_cpu(es->s_first_data_block) +
3667 EXT4_BLOCKS_PER_GROUP(sb) - 1);
3668 do_div(blocks_count, EXT4_BLOCKS_PER_GROUP(sb));
3669 if (blocks_count > ((uint64_t)1<<32) - EXT4_DESC_PER_BLOCK(sb)) {
3670 ext4_msg(sb, KERN_WARNING, "groups count too large: %u "
3671 "(block count %llu, first data block %u, "
3672 "blocks per group %lu)", sbi->s_groups_count,
3673 ext4_blocks_count(es),
3674 le32_to_cpu(es->s_first_data_block),
3675 EXT4_BLOCKS_PER_GROUP(sb));
3678 sbi->s_groups_count = blocks_count;
3679 sbi->s_blockfile_groups = min_t(ext4_group_t, sbi->s_groups_count,
3680 (EXT4_MAX_BLOCK_FILE_PHYS / EXT4_BLOCKS_PER_GROUP(sb)));
3681 db_count = (sbi->s_groups_count + EXT4_DESC_PER_BLOCK(sb) - 1) /
3682 EXT4_DESC_PER_BLOCK(sb);
3683 sbi->s_group_desc = ext4_kvmalloc(db_count *
3684 sizeof(struct buffer_head *),
3686 if (sbi->s_group_desc == NULL) {
3687 ext4_msg(sb, KERN_ERR, "not enough memory");
3693 sbi->s_proc = proc_mkdir(sb->s_id, ext4_proc_root);
3696 proc_create_data("options", S_IRUGO, sbi->s_proc,
3697 &ext4_seq_options_fops, sb);
3699 bgl_lock_init(sbi->s_blockgroup_lock);
3701 for (i = 0; i < db_count; i++) {
3702 block = descriptor_loc(sb, logical_sb_block, i);
3703 sbi->s_group_desc[i] = sb_bread(sb, block);
3704 if (!sbi->s_group_desc[i]) {
3705 ext4_msg(sb, KERN_ERR,
3706 "can't read group descriptor %d", i);
3711 if (!ext4_check_descriptors(sb, &first_not_zeroed)) {
3712 ext4_msg(sb, KERN_ERR, "group descriptors corrupted!");
3715 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_FLEX_BG))
3716 if (!ext4_fill_flex_info(sb)) {
3717 ext4_msg(sb, KERN_ERR,
3718 "unable to initialize "
3719 "flex_bg meta info!");
3723 sbi->s_gdb_count = db_count;
3724 get_random_bytes(&sbi->s_next_generation, sizeof(u32));
3725 spin_lock_init(&sbi->s_next_gen_lock);
3727 init_timer(&sbi->s_err_report);
3728 sbi->s_err_report.function = print_daily_error_info;
3729 sbi->s_err_report.data = (unsigned long) sb;
3731 err = percpu_counter_init(&sbi->s_freeclusters_counter,
3732 ext4_count_free_clusters(sb));
3734 err = percpu_counter_init(&sbi->s_freeinodes_counter,
3735 ext4_count_free_inodes(sb));
3738 err = percpu_counter_init(&sbi->s_dirs_counter,
3739 ext4_count_dirs(sb));
3742 err = percpu_counter_init(&sbi->s_dirtyclusters_counter, 0);
3745 ext4_msg(sb, KERN_ERR, "insufficient memory");
3750 sbi->s_stripe = ext4_get_stripe_size(sbi);
3751 sbi->s_max_writeback_mb_bump = 128;
3754 * set up enough so that it can read an inode
3756 if (!test_opt(sb, NOLOAD) &&
3757 EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL))
3758 sb->s_op = &ext4_sops;
3760 sb->s_op = &ext4_nojournal_sops;
3761 sb->s_export_op = &ext4_export_ops;
3762 sb->s_xattr = ext4_xattr_handlers;
3764 sb->s_qcop = &ext4_qctl_operations;
3765 sb->dq_op = &ext4_quota_operations;
3767 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_QUOTA)) {
3768 /* Use qctl operations for hidden quota files. */
3769 sb->s_qcop = &ext4_qctl_sysfile_operations;
3772 memcpy(sb->s_uuid, es->s_uuid, sizeof(es->s_uuid));
3774 INIT_LIST_HEAD(&sbi->s_orphan); /* unlinked but open files */
3775 mutex_init(&sbi->s_orphan_lock);
3776 sbi->s_resize_flags = 0;
3780 needs_recovery = (es->s_last_orphan != 0 ||
3781 EXT4_HAS_INCOMPAT_FEATURE(sb,
3782 EXT4_FEATURE_INCOMPAT_RECOVER));
3784 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_MMP) &&
3785 !(sb->s_flags & MS_RDONLY))
3786 if (ext4_multi_mount_protect(sb, le64_to_cpu(es->s_mmp_block)))
3790 * The first inode we look at is the journal inode. Don't try
3791 * root first: it may be modified in the journal!
3793 if (!test_opt(sb, NOLOAD) &&
3794 EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL)) {
3795 if (ext4_load_journal(sb, es, journal_devnum))
3797 } else if (test_opt(sb, NOLOAD) && !(sb->s_flags & MS_RDONLY) &&
3798 EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER)) {
3799 ext4_msg(sb, KERN_ERR, "required journal recovery "
3800 "suppressed and not mounted read-only");
3801 goto failed_mount_wq;
3803 clear_opt(sb, DATA_FLAGS);
3804 sbi->s_journal = NULL;
3809 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_64BIT) &&
3810 !jbd2_journal_set_features(EXT4_SB(sb)->s_journal, 0, 0,
3811 JBD2_FEATURE_INCOMPAT_64BIT)) {
3812 ext4_msg(sb, KERN_ERR, "Failed to set 64-bit journal feature");
3813 goto failed_mount_wq;
3816 if (!set_journal_csum_feature_set(sb)) {
3817 ext4_msg(sb, KERN_ERR, "Failed to set journal checksum "
3819 goto failed_mount_wq;
3822 /* We have now updated the journal if required, so we can
3823 * validate the data journaling mode. */
3824 switch (test_opt(sb, DATA_FLAGS)) {
3826 /* No mode set, assume a default based on the journal
3827 * capabilities: ORDERED_DATA if the journal can
3828 * cope, else JOURNAL_DATA
3830 if (jbd2_journal_check_available_features
3831 (sbi->s_journal, 0, 0, JBD2_FEATURE_INCOMPAT_REVOKE))
3832 set_opt(sb, ORDERED_DATA);
3834 set_opt(sb, JOURNAL_DATA);
3837 case EXT4_MOUNT_ORDERED_DATA:
3838 case EXT4_MOUNT_WRITEBACK_DATA:
3839 if (!jbd2_journal_check_available_features
3840 (sbi->s_journal, 0, 0, JBD2_FEATURE_INCOMPAT_REVOKE)) {
3841 ext4_msg(sb, KERN_ERR, "Journal does not support "
3842 "requested data journaling mode");
3843 goto failed_mount_wq;
3848 set_task_ioprio(sbi->s_journal->j_task, journal_ioprio);
3850 sbi->s_journal->j_commit_callback = ext4_journal_commit_callback;
3853 * The journal may have updated the bg summary counts, so we
3854 * need to update the global counters.
3856 percpu_counter_set(&sbi->s_freeclusters_counter,
3857 ext4_count_free_clusters(sb));
3858 percpu_counter_set(&sbi->s_freeinodes_counter,
3859 ext4_count_free_inodes(sb));
3860 percpu_counter_set(&sbi->s_dirs_counter,
3861 ext4_count_dirs(sb));
3862 percpu_counter_set(&sbi->s_dirtyclusters_counter, 0);
3866 * Get the # of file system overhead blocks from the
3867 * superblock if present.
3869 if (es->s_overhead_clusters)
3870 sbi->s_overhead = le32_to_cpu(es->s_overhead_clusters);
3872 ret = ext4_calculate_overhead(sb);
3874 goto failed_mount_wq;
3878 * The maximum number of concurrent works can be high and
3879 * concurrency isn't really necessary. Limit it to 1.
3881 EXT4_SB(sb)->dio_unwritten_wq =
3882 alloc_workqueue("ext4-dio-unwritten", WQ_MEM_RECLAIM | WQ_UNBOUND, 1);
3883 if (!EXT4_SB(sb)->dio_unwritten_wq) {
3884 printk(KERN_ERR "EXT4-fs: failed to create DIO workqueue\n");
3885 goto failed_mount_wq;
3889 * The jbd2_journal_load will have done any necessary log recovery,
3890 * so we can safely mount the rest of the filesystem now.
3893 root = ext4_iget(sb, EXT4_ROOT_INO);
3895 ext4_msg(sb, KERN_ERR, "get root inode failed");
3896 ret = PTR_ERR(root);
3900 if (!S_ISDIR(root->i_mode) || !root->i_blocks || !root->i_size) {
3901 ext4_msg(sb, KERN_ERR, "corrupt root inode, run e2fsck");
3905 sb->s_root = d_make_root(root);
3907 ext4_msg(sb, KERN_ERR, "get root dentry failed");
3912 if (ext4_setup_super(sb, es, sb->s_flags & MS_RDONLY))
3913 sb->s_flags |= MS_RDONLY;
3915 /* determine the minimum size of new large inodes, if present */
3916 if (sbi->s_inode_size > EXT4_GOOD_OLD_INODE_SIZE) {
3917 sbi->s_want_extra_isize = sizeof(struct ext4_inode) -
3918 EXT4_GOOD_OLD_INODE_SIZE;
3919 if (EXT4_HAS_RO_COMPAT_FEATURE(sb,
3920 EXT4_FEATURE_RO_COMPAT_EXTRA_ISIZE)) {
3921 if (sbi->s_want_extra_isize <
3922 le16_to_cpu(es->s_want_extra_isize))
3923 sbi->s_want_extra_isize =
3924 le16_to_cpu(es->s_want_extra_isize);
3925 if (sbi->s_want_extra_isize <
3926 le16_to_cpu(es->s_min_extra_isize))
3927 sbi->s_want_extra_isize =
3928 le16_to_cpu(es->s_min_extra_isize);
3931 /* Check if enough inode space is available */
3932 if (EXT4_GOOD_OLD_INODE_SIZE + sbi->s_want_extra_isize >
3933 sbi->s_inode_size) {
3934 sbi->s_want_extra_isize = sizeof(struct ext4_inode) -
3935 EXT4_GOOD_OLD_INODE_SIZE;
3936 ext4_msg(sb, KERN_INFO, "required extra inode space not"
3940 err = ext4_setup_system_zone(sb);
3942 ext4_msg(sb, KERN_ERR, "failed to initialize system "
3944 goto failed_mount4a;
3948 err = ext4_mb_init(sb);
3950 ext4_msg(sb, KERN_ERR, "failed to initialize mballoc (%d)",
3955 err = ext4_register_li_request(sb, first_not_zeroed);
3959 sbi->s_kobj.kset = ext4_kset;
3960 init_completion(&sbi->s_kobj_unregister);
3961 err = kobject_init_and_add(&sbi->s_kobj, &ext4_ktype, NULL,
3966 EXT4_SB(sb)->s_mount_state |= EXT4_ORPHAN_FS;
3967 ext4_orphan_cleanup(sb, es);
3968 EXT4_SB(sb)->s_mount_state &= ~EXT4_ORPHAN_FS;
3969 if (needs_recovery) {
3970 ext4_msg(sb, KERN_INFO, "recovery complete");
3971 ext4_mark_recovery_complete(sb, es);
3973 if (EXT4_SB(sb)->s_journal) {
3974 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA)
3975 descr = " journalled data mode";
3976 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA)
3977 descr = " ordered data mode";
3979 descr = " writeback data mode";
3981 descr = "out journal";
3984 /* Enable quota usage during mount. */
3985 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_QUOTA) &&
3986 !(sb->s_flags & MS_RDONLY)) {
3987 ret = ext4_enable_quotas(sb);
3991 #endif /* CONFIG_QUOTA */
3993 ext4_msg(sb, KERN_INFO, "mounted filesystem with%s. "
3994 "Opts: %s%s%s", descr, sbi->s_es->s_mount_opts,
3995 *sbi->s_es->s_mount_opts ? "; " : "", orig_data);
3997 if (es->s_error_count)
3998 mod_timer(&sbi->s_err_report, jiffies + 300*HZ); /* 5 minutes */
4005 ext4_msg(sb, KERN_ERR, "VFS: Can't find ext4 filesystem");
4009 ext4_unregister_li_request(sb);
4011 ext4_mb_release(sb);
4013 ext4_ext_release(sb);
4014 ext4_release_system_zone(sb);
4019 ext4_msg(sb, KERN_ERR, "mount failed");
4020 destroy_workqueue(EXT4_SB(sb)->dio_unwritten_wq);
4022 if (sbi->s_journal) {
4023 jbd2_journal_destroy(sbi->s_journal);
4024 sbi->s_journal = NULL;
4027 del_timer(&sbi->s_err_report);
4028 if (sbi->s_flex_groups)
4029 ext4_kvfree(sbi->s_flex_groups);
4030 percpu_counter_destroy(&sbi->s_freeclusters_counter);
4031 percpu_counter_destroy(&sbi->s_freeinodes_counter);
4032 percpu_counter_destroy(&sbi->s_dirs_counter);
4033 percpu_counter_destroy(&sbi->s_dirtyclusters_counter);
4035 kthread_stop(sbi->s_mmp_tsk);
4037 for (i = 0; i < db_count; i++)
4038 brelse(sbi->s_group_desc[i]);
4039 ext4_kvfree(sbi->s_group_desc);
4041 if (sbi->s_chksum_driver)
4042 crypto_free_shash(sbi->s_chksum_driver);
4044 remove_proc_entry("options", sbi->s_proc);
4045 remove_proc_entry(sb->s_id, ext4_proc_root);
4048 for (i = 0; i < MAXQUOTAS; i++)
4049 kfree(sbi->s_qf_names[i]);
4051 ext4_blkdev_remove(sbi);
4054 sb->s_fs_info = NULL;
4055 kfree(sbi->s_blockgroup_lock);
4063 * Setup any per-fs journal parameters now. We'll do this both on
4064 * initial mount, once the journal has been initialised but before we've
4065 * done any recovery; and again on any subsequent remount.
4067 static void ext4_init_journal_params(struct super_block *sb, journal_t *journal)
4069 struct ext4_sb_info *sbi = EXT4_SB(sb);
4071 journal->j_commit_interval = sbi->s_commit_interval;
4072 journal->j_min_batch_time = sbi->s_min_batch_time;
4073 journal->j_max_batch_time = sbi->s_max_batch_time;
4075 write_lock(&journal->j_state_lock);
4076 if (test_opt(sb, BARRIER))
4077 journal->j_flags |= JBD2_BARRIER;
4079 journal->j_flags &= ~JBD2_BARRIER;
4080 if (test_opt(sb, DATA_ERR_ABORT))
4081 journal->j_flags |= JBD2_ABORT_ON_SYNCDATA_ERR;
4083 journal->j_flags &= ~JBD2_ABORT_ON_SYNCDATA_ERR;
4084 write_unlock(&journal->j_state_lock);
4087 static journal_t *ext4_get_journal(struct super_block *sb,
4088 unsigned int journal_inum)
4090 struct inode *journal_inode;
4093 BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL));
4095 /* First, test for the existence of a valid inode on disk. Bad
4096 * things happen if we iget() an unused inode, as the subsequent
4097 * iput() will try to delete it. */
4099 journal_inode = ext4_iget(sb, journal_inum);
4100 if (IS_ERR(journal_inode)) {
4101 ext4_msg(sb, KERN_ERR, "no journal found");
4104 if (!journal_inode->i_nlink) {
4105 make_bad_inode(journal_inode);
4106 iput(journal_inode);
4107 ext4_msg(sb, KERN_ERR, "journal inode is deleted");
4111 jbd_debug(2, "Journal inode found at %p: %lld bytes\n",
4112 journal_inode, journal_inode->i_size);
4113 if (!S_ISREG(journal_inode->i_mode)) {
4114 ext4_msg(sb, KERN_ERR, "invalid journal inode");
4115 iput(journal_inode);
4119 journal = jbd2_journal_init_inode(journal_inode);
4121 ext4_msg(sb, KERN_ERR, "Could not load journal inode");
4122 iput(journal_inode);
4125 journal->j_private = sb;
4126 ext4_init_journal_params(sb, journal);
4130 static journal_t *ext4_get_dev_journal(struct super_block *sb,
4133 struct buffer_head *bh;
4137 int hblock, blocksize;
4138 ext4_fsblk_t sb_block;
4139 unsigned long offset;
4140 struct ext4_super_block *es;
4141 struct block_device *bdev;
4143 BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL));
4145 bdev = ext4_blkdev_get(j_dev, sb);
4149 blocksize = sb->s_blocksize;
4150 hblock = bdev_logical_block_size(bdev);
4151 if (blocksize < hblock) {
4152 ext4_msg(sb, KERN_ERR,
4153 "blocksize too small for journal device");
4157 sb_block = EXT4_MIN_BLOCK_SIZE / blocksize;
4158 offset = EXT4_MIN_BLOCK_SIZE % blocksize;
4159 set_blocksize(bdev, blocksize);
4160 if (!(bh = __bread(bdev, sb_block, blocksize))) {
4161 ext4_msg(sb, KERN_ERR, "couldn't read superblock of "
4162 "external journal");
4166 es = (struct ext4_super_block *) (bh->b_data + offset);
4167 if ((le16_to_cpu(es->s_magic) != EXT4_SUPER_MAGIC) ||
4168 !(le32_to_cpu(es->s_feature_incompat) &
4169 EXT4_FEATURE_INCOMPAT_JOURNAL_DEV)) {
4170 ext4_msg(sb, KERN_ERR, "external journal has "
4176 if (memcmp(EXT4_SB(sb)->s_es->s_journal_uuid, es->s_uuid, 16)) {
4177 ext4_msg(sb, KERN_ERR, "journal UUID does not match");
4182 len = ext4_blocks_count(es);
4183 start = sb_block + 1;
4184 brelse(bh); /* we're done with the superblock */
4186 journal = jbd2_journal_init_dev(bdev, sb->s_bdev,
4187 start, len, blocksize);
4189 ext4_msg(sb, KERN_ERR, "failed to create device journal");
4192 journal->j_private = sb;
4193 ll_rw_block(READ, 1, &journal->j_sb_buffer);
4194 wait_on_buffer(journal->j_sb_buffer);
4195 if (!buffer_uptodate(journal->j_sb_buffer)) {
4196 ext4_msg(sb, KERN_ERR, "I/O error on journal device");
4199 if (be32_to_cpu(journal->j_superblock->s_nr_users) != 1) {
4200 ext4_msg(sb, KERN_ERR, "External journal has more than one "
4201 "user (unsupported) - %d",
4202 be32_to_cpu(journal->j_superblock->s_nr_users));
4205 EXT4_SB(sb)->journal_bdev = bdev;
4206 ext4_init_journal_params(sb, journal);
4210 jbd2_journal_destroy(journal);
4212 ext4_blkdev_put(bdev);
4216 static int ext4_load_journal(struct super_block *sb,
4217 struct ext4_super_block *es,
4218 unsigned long journal_devnum)
4221 unsigned int journal_inum = le32_to_cpu(es->s_journal_inum);
4224 int really_read_only;
4226 BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL));
4228 if (journal_devnum &&
4229 journal_devnum != le32_to_cpu(es->s_journal_dev)) {
4230 ext4_msg(sb, KERN_INFO, "external journal device major/minor "
4231 "numbers have changed");
4232 journal_dev = new_decode_dev(journal_devnum);
4234 journal_dev = new_decode_dev(le32_to_cpu(es->s_journal_dev));
4236 really_read_only = bdev_read_only(sb->s_bdev);
4239 * Are we loading a blank journal or performing recovery after a
4240 * crash? For recovery, we need to check in advance whether we
4241 * can get read-write access to the device.
4243 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER)) {
4244 if (sb->s_flags & MS_RDONLY) {
4245 ext4_msg(sb, KERN_INFO, "INFO: recovery "
4246 "required on readonly filesystem");
4247 if (really_read_only) {
4248 ext4_msg(sb, KERN_ERR, "write access "
4249 "unavailable, cannot proceed");
4252 ext4_msg(sb, KERN_INFO, "write access will "
4253 "be enabled during recovery");
4257 if (journal_inum && journal_dev) {
4258 ext4_msg(sb, KERN_ERR, "filesystem has both journal "
4259 "and inode journals!");
4264 if (!(journal = ext4_get_journal(sb, journal_inum)))
4267 if (!(journal = ext4_get_dev_journal(sb, journal_dev)))
4271 if (!(journal->j_flags & JBD2_BARRIER))
4272 ext4_msg(sb, KERN_INFO, "barriers disabled");
4274 if (!EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER))
4275 err = jbd2_journal_wipe(journal, !really_read_only);
4277 char *save = kmalloc(EXT4_S_ERR_LEN, GFP_KERNEL);
4279 memcpy(save, ((char *) es) +
4280 EXT4_S_ERR_START, EXT4_S_ERR_LEN);
4281 err = jbd2_journal_load(journal);
4283 memcpy(((char *) es) + EXT4_S_ERR_START,
4284 save, EXT4_S_ERR_LEN);
4289 ext4_msg(sb, KERN_ERR, "error loading journal");
4290 jbd2_journal_destroy(journal);
4294 EXT4_SB(sb)->s_journal = journal;
4295 ext4_clear_journal_err(sb, es);
4297 if (!really_read_only && journal_devnum &&
4298 journal_devnum != le32_to_cpu(es->s_journal_dev)) {
4299 es->s_journal_dev = cpu_to_le32(journal_devnum);
4301 /* Make sure we flush the recovery flag to disk. */
4302 ext4_commit_super(sb, 1);
4308 static int ext4_commit_super(struct super_block *sb, int sync)
4310 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
4311 struct buffer_head *sbh = EXT4_SB(sb)->s_sbh;
4314 if (!sbh || block_device_ejected(sb))
4316 if (buffer_write_io_error(sbh)) {
4318 * Oh, dear. A previous attempt to write the
4319 * superblock failed. This could happen because the
4320 * USB device was yanked out. Or it could happen to
4321 * be a transient write error and maybe the block will
4322 * be remapped. Nothing we can do but to retry the
4323 * write and hope for the best.
4325 ext4_msg(sb, KERN_ERR, "previous I/O error to "
4326 "superblock detected");
4327 clear_buffer_write_io_error(sbh);
4328 set_buffer_uptodate(sbh);
4331 * If the file system is mounted read-only, don't update the
4332 * superblock write time. This avoids updating the superblock
4333 * write time when we are mounting the root file system
4334 * read/only but we need to replay the journal; at that point,
4335 * for people who are east of GMT and who make their clock
4336 * tick in localtime for Windows bug-for-bug compatibility,
4337 * the clock is set in the future, and this will cause e2fsck
4338 * to complain and force a full file system check.
4340 if (!(sb->s_flags & MS_RDONLY))
4341 es->s_wtime = cpu_to_le32(get_seconds());
4342 if (sb->s_bdev->bd_part)
4343 es->s_kbytes_written =
4344 cpu_to_le64(EXT4_SB(sb)->s_kbytes_written +
4345 ((part_stat_read(sb->s_bdev->bd_part, sectors[1]) -
4346 EXT4_SB(sb)->s_sectors_written_start) >> 1));
4348 es->s_kbytes_written =
4349 cpu_to_le64(EXT4_SB(sb)->s_kbytes_written);
4350 ext4_free_blocks_count_set(es,
4351 EXT4_C2B(EXT4_SB(sb), percpu_counter_sum_positive(
4352 &EXT4_SB(sb)->s_freeclusters_counter)));
4353 es->s_free_inodes_count =
4354 cpu_to_le32(percpu_counter_sum_positive(
4355 &EXT4_SB(sb)->s_freeinodes_counter));
4356 BUFFER_TRACE(sbh, "marking dirty");
4357 ext4_superblock_csum_set(sb, es);
4358 mark_buffer_dirty(sbh);
4360 error = sync_dirty_buffer(sbh);
4364 error = buffer_write_io_error(sbh);
4366 ext4_msg(sb, KERN_ERR, "I/O error while writing "
4368 clear_buffer_write_io_error(sbh);
4369 set_buffer_uptodate(sbh);
4376 * Have we just finished recovery? If so, and if we are mounting (or
4377 * remounting) the filesystem readonly, then we will end up with a
4378 * consistent fs on disk. Record that fact.
4380 static void ext4_mark_recovery_complete(struct super_block *sb,
4381 struct ext4_super_block *es)
4383 journal_t *journal = EXT4_SB(sb)->s_journal;
4385 if (!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL)) {
4386 BUG_ON(journal != NULL);
4389 jbd2_journal_lock_updates(journal);
4390 if (jbd2_journal_flush(journal) < 0)
4393 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER) &&
4394 sb->s_flags & MS_RDONLY) {
4395 EXT4_CLEAR_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
4396 ext4_commit_super(sb, 1);
4400 jbd2_journal_unlock_updates(journal);
4404 * If we are mounting (or read-write remounting) a filesystem whose journal
4405 * has recorded an error from a previous lifetime, move that error to the
4406 * main filesystem now.
4408 static void ext4_clear_journal_err(struct super_block *sb,
4409 struct ext4_super_block *es)
4415 BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL));
4417 journal = EXT4_SB(sb)->s_journal;
4420 * Now check for any error status which may have been recorded in the
4421 * journal by a prior ext4_error() or ext4_abort()
4424 j_errno = jbd2_journal_errno(journal);
4428 errstr = ext4_decode_error(sb, j_errno, nbuf);
4429 ext4_warning(sb, "Filesystem error recorded "
4430 "from previous mount: %s", errstr);
4431 ext4_warning(sb, "Marking fs in need of filesystem check.");
4433 EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS;
4434 es->s_state |= cpu_to_le16(EXT4_ERROR_FS);
4435 ext4_commit_super(sb, 1);
4437 jbd2_journal_clear_err(journal);
4438 jbd2_journal_update_sb_errno(journal);
4443 * Force the running and committing transactions to commit,
4444 * and wait on the commit.
4446 int ext4_force_commit(struct super_block *sb)
4451 if (sb->s_flags & MS_RDONLY)
4454 journal = EXT4_SB(sb)->s_journal;
4456 ret = ext4_journal_force_commit(journal);
4461 static int ext4_sync_fs(struct super_block *sb, int wait)
4465 struct ext4_sb_info *sbi = EXT4_SB(sb);
4467 trace_ext4_sync_fs(sb, wait);
4468 flush_workqueue(sbi->dio_unwritten_wq);
4470 * Writeback quota in non-journalled quota case - journalled quota has
4473 dquot_writeback_dquots(sb, -1);
4474 if (jbd2_journal_start_commit(sbi->s_journal, &target)) {
4476 jbd2_log_wait_commit(sbi->s_journal, target);
4482 * LVM calls this function before a (read-only) snapshot is created. This
4483 * gives us a chance to flush the journal completely and mark the fs clean.
4485 * Note that only this function cannot bring a filesystem to be in a clean
4486 * state independently. It relies on upper layer to stop all data & metadata
4489 static int ext4_freeze(struct super_block *sb)
4494 if (sb->s_flags & MS_RDONLY)
4497 journal = EXT4_SB(sb)->s_journal;
4499 /* Now we set up the journal barrier. */
4500 jbd2_journal_lock_updates(journal);
4503 * Don't clear the needs_recovery flag if we failed to flush
4506 error = jbd2_journal_flush(journal);
4510 /* Journal blocked and flushed, clear needs_recovery flag. */
4511 EXT4_CLEAR_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
4512 error = ext4_commit_super(sb, 1);
4514 /* we rely on upper layer to stop further updates */
4515 jbd2_journal_unlock_updates(EXT4_SB(sb)->s_journal);
4520 * Called by LVM after the snapshot is done. We need to reset the RECOVER
4521 * flag here, even though the filesystem is not technically dirty yet.
4523 static int ext4_unfreeze(struct super_block *sb)
4525 if (sb->s_flags & MS_RDONLY)
4529 /* Reset the needs_recovery flag before the fs is unlocked. */
4530 EXT4_SET_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
4531 ext4_commit_super(sb, 1);
4537 * Structure to save mount options for ext4_remount's benefit
4539 struct ext4_mount_options {
4540 unsigned long s_mount_opt;
4541 unsigned long s_mount_opt2;
4544 unsigned long s_commit_interval;
4545 u32 s_min_batch_time, s_max_batch_time;
4548 char *s_qf_names[MAXQUOTAS];
4552 static int ext4_remount(struct super_block *sb, int *flags, char *data)
4554 struct ext4_super_block *es;
4555 struct ext4_sb_info *sbi = EXT4_SB(sb);
4556 unsigned long old_sb_flags;
4557 struct ext4_mount_options old_opts;
4558 int enable_quota = 0;
4560 unsigned int journal_ioprio = DEFAULT_JOURNAL_IOPRIO;
4565 char *orig_data = kstrdup(data, GFP_KERNEL);
4567 /* Store the original options */
4569 old_sb_flags = sb->s_flags;
4570 old_opts.s_mount_opt = sbi->s_mount_opt;
4571 old_opts.s_mount_opt2 = sbi->s_mount_opt2;
4572 old_opts.s_resuid = sbi->s_resuid;
4573 old_opts.s_resgid = sbi->s_resgid;
4574 old_opts.s_commit_interval = sbi->s_commit_interval;
4575 old_opts.s_min_batch_time = sbi->s_min_batch_time;
4576 old_opts.s_max_batch_time = sbi->s_max_batch_time;
4578 old_opts.s_jquota_fmt = sbi->s_jquota_fmt;
4579 for (i = 0; i < MAXQUOTAS; i++)
4580 old_opts.s_qf_names[i] = sbi->s_qf_names[i];
4582 if (sbi->s_journal && sbi->s_journal->j_task->io_context)
4583 journal_ioprio = sbi->s_journal->j_task->io_context->ioprio;
4586 * Allow the "check" option to be passed as a remount option.
4588 if (!parse_options(data, sb, NULL, &journal_ioprio, 1)) {
4593 if (sbi->s_mount_flags & EXT4_MF_FS_ABORTED)
4594 ext4_abort(sb, "Abort forced by user");
4596 sb->s_flags = (sb->s_flags & ~MS_POSIXACL) |
4597 (test_opt(sb, POSIX_ACL) ? MS_POSIXACL : 0);
4601 if (sbi->s_journal) {
4602 ext4_init_journal_params(sb, sbi->s_journal);
4603 set_task_ioprio(sbi->s_journal->j_task, journal_ioprio);
4606 if ((*flags & MS_RDONLY) != (sb->s_flags & MS_RDONLY)) {
4607 if (sbi->s_mount_flags & EXT4_MF_FS_ABORTED) {
4612 if (*flags & MS_RDONLY) {
4613 err = dquot_suspend(sb, -1);
4618 * First of all, the unconditional stuff we have to do
4619 * to disable replay of the journal when we next remount
4621 sb->s_flags |= MS_RDONLY;
4624 * OK, test if we are remounting a valid rw partition
4625 * readonly, and if so set the rdonly flag and then
4626 * mark the partition as valid again.
4628 if (!(es->s_state & cpu_to_le16(EXT4_VALID_FS)) &&
4629 (sbi->s_mount_state & EXT4_VALID_FS))
4630 es->s_state = cpu_to_le16(sbi->s_mount_state);
4633 ext4_mark_recovery_complete(sb, es);
4635 /* Make sure we can mount this feature set readwrite */
4636 if (!ext4_feature_set_ok(sb, 0)) {
4641 * Make sure the group descriptor checksums
4642 * are sane. If they aren't, refuse to remount r/w.
4644 for (g = 0; g < sbi->s_groups_count; g++) {
4645 struct ext4_group_desc *gdp =
4646 ext4_get_group_desc(sb, g, NULL);
4648 if (!ext4_group_desc_csum_verify(sb, g, gdp)) {
4649 ext4_msg(sb, KERN_ERR,
4650 "ext4_remount: Checksum for group %u failed (%u!=%u)",
4651 g, le16_to_cpu(ext4_group_desc_csum(sbi, g, gdp)),
4652 le16_to_cpu(gdp->bg_checksum));
4659 * If we have an unprocessed orphan list hanging
4660 * around from a previously readonly bdev mount,
4661 * require a full umount/remount for now.
4663 if (es->s_last_orphan) {
4664 ext4_msg(sb, KERN_WARNING, "Couldn't "
4665 "remount RDWR because of unprocessed "
4666 "orphan inode list. Please "
4667 "umount/remount instead");
4673 * Mounting a RDONLY partition read-write, so reread
4674 * and store the current valid flag. (It may have
4675 * been changed by e2fsck since we originally mounted
4679 ext4_clear_journal_err(sb, es);
4680 sbi->s_mount_state = le16_to_cpu(es->s_state);
4681 if (!ext4_setup_super(sb, es, 0))
4682 sb->s_flags &= ~MS_RDONLY;
4683 if (EXT4_HAS_INCOMPAT_FEATURE(sb,
4684 EXT4_FEATURE_INCOMPAT_MMP))
4685 if (ext4_multi_mount_protect(sb,
4686 le64_to_cpu(es->s_mmp_block))) {
4695 * Reinitialize lazy itable initialization thread based on
4698 if ((sb->s_flags & MS_RDONLY) || !test_opt(sb, INIT_INODE_TABLE))
4699 ext4_unregister_li_request(sb);
4701 ext4_group_t first_not_zeroed;
4702 first_not_zeroed = ext4_has_uninit_itable(sb);
4703 ext4_register_li_request(sb, first_not_zeroed);
4706 ext4_setup_system_zone(sb);
4707 if (sbi->s_journal == NULL)
4708 ext4_commit_super(sb, 1);
4712 /* Release old quota file names */
4713 for (i = 0; i < MAXQUOTAS; i++)
4714 if (old_opts.s_qf_names[i] &&
4715 old_opts.s_qf_names[i] != sbi->s_qf_names[i])
4716 kfree(old_opts.s_qf_names[i]);
4718 if (sb_any_quota_suspended(sb))
4719 dquot_resume(sb, -1);
4720 else if (EXT4_HAS_RO_COMPAT_FEATURE(sb,
4721 EXT4_FEATURE_RO_COMPAT_QUOTA)) {
4722 err = ext4_enable_quotas(sb);
4731 ext4_msg(sb, KERN_INFO, "re-mounted. Opts: %s", orig_data);
4736 sb->s_flags = old_sb_flags;
4737 sbi->s_mount_opt = old_opts.s_mount_opt;
4738 sbi->s_mount_opt2 = old_opts.s_mount_opt2;
4739 sbi->s_resuid = old_opts.s_resuid;
4740 sbi->s_resgid = old_opts.s_resgid;
4741 sbi->s_commit_interval = old_opts.s_commit_interval;
4742 sbi->s_min_batch_time = old_opts.s_min_batch_time;
4743 sbi->s_max_batch_time = old_opts.s_max_batch_time;
4745 sbi->s_jquota_fmt = old_opts.s_jquota_fmt;
4746 for (i = 0; i < MAXQUOTAS; i++) {
4747 if (sbi->s_qf_names[i] &&
4748 old_opts.s_qf_names[i] != sbi->s_qf_names[i])
4749 kfree(sbi->s_qf_names[i]);
4750 sbi->s_qf_names[i] = old_opts.s_qf_names[i];
4758 static int ext4_statfs(struct dentry *dentry, struct kstatfs *buf)
4760 struct super_block *sb = dentry->d_sb;
4761 struct ext4_sb_info *sbi = EXT4_SB(sb);
4762 struct ext4_super_block *es = sbi->s_es;
4763 ext4_fsblk_t overhead = 0;
4767 if (!test_opt(sb, MINIX_DF))
4768 overhead = sbi->s_overhead;
4770 buf->f_type = EXT4_SUPER_MAGIC;
4771 buf->f_bsize = sb->s_blocksize;
4772 buf->f_blocks = ext4_blocks_count(es) - EXT4_C2B(sbi, sbi->s_overhead);
4773 bfree = percpu_counter_sum_positive(&sbi->s_freeclusters_counter) -
4774 percpu_counter_sum_positive(&sbi->s_dirtyclusters_counter);
4775 /* prevent underflow in case that few free space is available */
4776 buf->f_bfree = EXT4_C2B(sbi, max_t(s64, bfree, 0));
4777 buf->f_bavail = buf->f_bfree - ext4_r_blocks_count(es);
4778 if (buf->f_bfree < ext4_r_blocks_count(es))
4780 buf->f_files = le32_to_cpu(es->s_inodes_count);
4781 buf->f_ffree = percpu_counter_sum_positive(&sbi->s_freeinodes_counter);
4782 buf->f_namelen = EXT4_NAME_LEN;
4783 fsid = le64_to_cpup((void *)es->s_uuid) ^
4784 le64_to_cpup((void *)es->s_uuid + sizeof(u64));
4785 buf->f_fsid.val[0] = fsid & 0xFFFFFFFFUL;
4786 buf->f_fsid.val[1] = (fsid >> 32) & 0xFFFFFFFFUL;
4791 /* Helper function for writing quotas on sync - we need to start transaction
4792 * before quota file is locked for write. Otherwise the are possible deadlocks:
4793 * Process 1 Process 2
4794 * ext4_create() quota_sync()
4795 * jbd2_journal_start() write_dquot()
4796 * dquot_initialize() down(dqio_mutex)
4797 * down(dqio_mutex) jbd2_journal_start()
4803 static inline struct inode *dquot_to_inode(struct dquot *dquot)
4805 return sb_dqopt(dquot->dq_sb)->files[dquot->dq_type];
4808 static int ext4_write_dquot(struct dquot *dquot)
4812 struct inode *inode;
4814 inode = dquot_to_inode(dquot);
4815 handle = ext4_journal_start(inode,
4816 EXT4_QUOTA_TRANS_BLOCKS(dquot->dq_sb));
4818 return PTR_ERR(handle);
4819 ret = dquot_commit(dquot);
4820 err = ext4_journal_stop(handle);
4826 static int ext4_acquire_dquot(struct dquot *dquot)
4831 handle = ext4_journal_start(dquot_to_inode(dquot),
4832 EXT4_QUOTA_INIT_BLOCKS(dquot->dq_sb));
4834 return PTR_ERR(handle);
4835 ret = dquot_acquire(dquot);
4836 err = ext4_journal_stop(handle);
4842 static int ext4_release_dquot(struct dquot *dquot)
4847 handle = ext4_journal_start(dquot_to_inode(dquot),
4848 EXT4_QUOTA_DEL_BLOCKS(dquot->dq_sb));
4849 if (IS_ERR(handle)) {
4850 /* Release dquot anyway to avoid endless cycle in dqput() */
4851 dquot_release(dquot);
4852 return PTR_ERR(handle);
4854 ret = dquot_release(dquot);
4855 err = ext4_journal_stop(handle);
4861 static int ext4_mark_dquot_dirty(struct dquot *dquot)
4863 /* Are we journaling quotas? */
4864 if (EXT4_SB(dquot->dq_sb)->s_qf_names[USRQUOTA] ||
4865 EXT4_SB(dquot->dq_sb)->s_qf_names[GRPQUOTA]) {
4866 dquot_mark_dquot_dirty(dquot);
4867 return ext4_write_dquot(dquot);
4869 return dquot_mark_dquot_dirty(dquot);
4873 static int ext4_write_info(struct super_block *sb, int type)
4878 /* Data block + inode block */
4879 handle = ext4_journal_start(sb->s_root->d_inode, 2);
4881 return PTR_ERR(handle);
4882 ret = dquot_commit_info(sb, type);
4883 err = ext4_journal_stop(handle);
4890 * Turn on quotas during mount time - we need to find
4891 * the quota file and such...
4893 static int ext4_quota_on_mount(struct super_block *sb, int type)
4895 return dquot_quota_on_mount(sb, EXT4_SB(sb)->s_qf_names[type],
4896 EXT4_SB(sb)->s_jquota_fmt, type);
4900 * Standard function to be called on quota_on
4902 static int ext4_quota_on(struct super_block *sb, int type, int format_id,
4907 if (!test_opt(sb, QUOTA))
4910 /* Quotafile not on the same filesystem? */
4911 if (path->dentry->d_sb != sb)
4913 /* Journaling quota? */
4914 if (EXT4_SB(sb)->s_qf_names[type]) {
4915 /* Quotafile not in fs root? */
4916 if (path->dentry->d_parent != sb->s_root)
4917 ext4_msg(sb, KERN_WARNING,
4918 "Quota file not on filesystem root. "
4919 "Journaled quota will not work");
4923 * When we journal data on quota file, we have to flush journal to see
4924 * all updates to the file when we bypass pagecache...
4926 if (EXT4_SB(sb)->s_journal &&
4927 ext4_should_journal_data(path->dentry->d_inode)) {
4929 * We don't need to lock updates but journal_flush() could
4930 * otherwise be livelocked...
4932 jbd2_journal_lock_updates(EXT4_SB(sb)->s_journal);
4933 err = jbd2_journal_flush(EXT4_SB(sb)->s_journal);
4934 jbd2_journal_unlock_updates(EXT4_SB(sb)->s_journal);
4939 return dquot_quota_on(sb, type, format_id, path);
4942 static int ext4_quota_enable(struct super_block *sb, int type, int format_id,
4946 struct inode *qf_inode;
4947 unsigned long qf_inums[MAXQUOTAS] = {
4948 le32_to_cpu(EXT4_SB(sb)->s_es->s_usr_quota_inum),
4949 le32_to_cpu(EXT4_SB(sb)->s_es->s_grp_quota_inum)
4952 BUG_ON(!EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_QUOTA));
4954 if (!qf_inums[type])
4957 qf_inode = ext4_iget(sb, qf_inums[type]);
4958 if (IS_ERR(qf_inode)) {
4959 ext4_error(sb, "Bad quota inode # %lu", qf_inums[type]);
4960 return PTR_ERR(qf_inode);
4963 err = dquot_enable(qf_inode, type, format_id, flags);
4969 /* Enable usage tracking for all quota types. */
4970 static int ext4_enable_quotas(struct super_block *sb)
4973 unsigned long qf_inums[MAXQUOTAS] = {
4974 le32_to_cpu(EXT4_SB(sb)->s_es->s_usr_quota_inum),
4975 le32_to_cpu(EXT4_SB(sb)->s_es->s_grp_quota_inum)
4978 sb_dqopt(sb)->flags |= DQUOT_QUOTA_SYS_FILE;
4979 for (type = 0; type < MAXQUOTAS; type++) {
4980 if (qf_inums[type]) {
4981 err = ext4_quota_enable(sb, type, QFMT_VFS_V1,
4982 DQUOT_USAGE_ENABLED);
4985 "Failed to enable quota (type=%d) "
4986 "tracking. Please run e2fsck to fix.",
4996 * quota_on function that is used when QUOTA feature is set.
4998 static int ext4_quota_on_sysfile(struct super_block *sb, int type,
5001 if (!EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_QUOTA))
5005 * USAGE was enabled at mount time. Only need to enable LIMITS now.
5007 return ext4_quota_enable(sb, type, format_id, DQUOT_LIMITS_ENABLED);
5010 static int ext4_quota_off(struct super_block *sb, int type)
5012 struct inode *inode = sb_dqopt(sb)->files[type];
5015 /* Force all delayed allocation blocks to be allocated.
5016 * Caller already holds s_umount sem */
5017 if (test_opt(sb, DELALLOC))
5018 sync_filesystem(sb);
5023 /* Update modification times of quota files when userspace can
5024 * start looking at them */
5025 handle = ext4_journal_start(inode, 1);
5028 inode->i_mtime = inode->i_ctime = CURRENT_TIME;
5029 ext4_mark_inode_dirty(handle, inode);
5030 ext4_journal_stop(handle);
5033 return dquot_quota_off(sb, type);
5037 * quota_off function that is used when QUOTA feature is set.
5039 static int ext4_quota_off_sysfile(struct super_block *sb, int type)
5041 if (!EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_QUOTA))
5044 /* Disable only the limits. */
5045 return dquot_disable(sb, type, DQUOT_LIMITS_ENABLED);
5048 /* Read data from quotafile - avoid pagecache and such because we cannot afford
5049 * acquiring the locks... As quota files are never truncated and quota code
5050 * itself serializes the operations (and no one else should touch the files)
5051 * we don't have to be afraid of races */
5052 static ssize_t ext4_quota_read(struct super_block *sb, int type, char *data,
5053 size_t len, loff_t off)
5055 struct inode *inode = sb_dqopt(sb)->files[type];
5056 ext4_lblk_t blk = off >> EXT4_BLOCK_SIZE_BITS(sb);
5058 int offset = off & (sb->s_blocksize - 1);
5061 struct buffer_head *bh;
5062 loff_t i_size = i_size_read(inode);
5066 if (off+len > i_size)
5069 while (toread > 0) {
5070 tocopy = sb->s_blocksize - offset < toread ?
5071 sb->s_blocksize - offset : toread;
5072 bh = ext4_bread(NULL, inode, blk, 0, &err);
5075 if (!bh) /* A hole? */
5076 memset(data, 0, tocopy);
5078 memcpy(data, bh->b_data+offset, tocopy);
5088 /* Write to quotafile (we know the transaction is already started and has
5089 * enough credits) */
5090 static ssize_t ext4_quota_write(struct super_block *sb, int type,
5091 const char *data, size_t len, loff_t off)
5093 struct inode *inode = sb_dqopt(sb)->files[type];
5094 ext4_lblk_t blk = off >> EXT4_BLOCK_SIZE_BITS(sb);
5096 int offset = off & (sb->s_blocksize - 1);
5097 struct buffer_head *bh;
5098 handle_t *handle = journal_current_handle();
5100 if (EXT4_SB(sb)->s_journal && !handle) {
5101 ext4_msg(sb, KERN_WARNING, "Quota write (off=%llu, len=%llu)"
5102 " cancelled because transaction is not started",
5103 (unsigned long long)off, (unsigned long long)len);
5107 * Since we account only one data block in transaction credits,
5108 * then it is impossible to cross a block boundary.
5110 if (sb->s_blocksize - offset < len) {
5111 ext4_msg(sb, KERN_WARNING, "Quota write (off=%llu, len=%llu)"
5112 " cancelled because not block aligned",
5113 (unsigned long long)off, (unsigned long long)len);
5117 bh = ext4_bread(handle, inode, blk, 1, &err);
5120 err = ext4_journal_get_write_access(handle, bh);
5126 memcpy(bh->b_data+offset, data, len);
5127 flush_dcache_page(bh->b_page);
5129 err = ext4_handle_dirty_metadata(handle, NULL, bh);
5134 if (inode->i_size < off + len) {
5135 i_size_write(inode, off + len);
5136 EXT4_I(inode)->i_disksize = inode->i_size;
5137 ext4_mark_inode_dirty(handle, inode);
5144 static struct dentry *ext4_mount(struct file_system_type *fs_type, int flags,
5145 const char *dev_name, void *data)
5147 return mount_bdev(fs_type, flags, dev_name, data, ext4_fill_super);
5150 #if !defined(CONFIG_EXT2_FS) && !defined(CONFIG_EXT2_FS_MODULE) && defined(CONFIG_EXT4_USE_FOR_EXT23)
5151 static inline void register_as_ext2(void)
5153 int err = register_filesystem(&ext2_fs_type);
5156 "EXT4-fs: Unable to register as ext2 (%d)\n", err);
5159 static inline void unregister_as_ext2(void)
5161 unregister_filesystem(&ext2_fs_type);
5164 static inline int ext2_feature_set_ok(struct super_block *sb)
5166 if (EXT4_HAS_INCOMPAT_FEATURE(sb, ~EXT2_FEATURE_INCOMPAT_SUPP))
5168 if (sb->s_flags & MS_RDONLY)
5170 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, ~EXT2_FEATURE_RO_COMPAT_SUPP))
5174 MODULE_ALIAS("ext2");
5176 static inline void register_as_ext2(void) { }
5177 static inline void unregister_as_ext2(void) { }
5178 static inline int ext2_feature_set_ok(struct super_block *sb) { return 0; }
5181 #if !defined(CONFIG_EXT3_FS) && !defined(CONFIG_EXT3_FS_MODULE) && defined(CONFIG_EXT4_USE_FOR_EXT23)
5182 static inline void register_as_ext3(void)
5184 int err = register_filesystem(&ext3_fs_type);
5187 "EXT4-fs: Unable to register as ext3 (%d)\n", err);
5190 static inline void unregister_as_ext3(void)
5192 unregister_filesystem(&ext3_fs_type);
5195 static inline int ext3_feature_set_ok(struct super_block *sb)
5197 if (EXT4_HAS_INCOMPAT_FEATURE(sb, ~EXT3_FEATURE_INCOMPAT_SUPP))
5199 if (!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL))
5201 if (sb->s_flags & MS_RDONLY)
5203 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, ~EXT3_FEATURE_RO_COMPAT_SUPP))
5207 MODULE_ALIAS("ext3");
5209 static inline void register_as_ext3(void) { }
5210 static inline void unregister_as_ext3(void) { }
5211 static inline int ext3_feature_set_ok(struct super_block *sb) { return 0; }
5214 static struct file_system_type ext4_fs_type = {
5215 .owner = THIS_MODULE,
5217 .mount = ext4_mount,
5218 .kill_sb = kill_block_super,
5219 .fs_flags = FS_REQUIRES_DEV,
5222 static int __init ext4_init_feat_adverts(void)
5224 struct ext4_features *ef;
5227 ef = kzalloc(sizeof(struct ext4_features), GFP_KERNEL);
5231 ef->f_kobj.kset = ext4_kset;
5232 init_completion(&ef->f_kobj_unregister);
5233 ret = kobject_init_and_add(&ef->f_kobj, &ext4_feat_ktype, NULL,
5246 static void ext4_exit_feat_adverts(void)
5248 kobject_put(&ext4_feat->f_kobj);
5249 wait_for_completion(&ext4_feat->f_kobj_unregister);
5253 /* Shared across all ext4 file systems */
5254 wait_queue_head_t ext4__ioend_wq[EXT4_WQ_HASH_SZ];
5255 struct mutex ext4__aio_mutex[EXT4_WQ_HASH_SZ];
5257 static int __init ext4_init_fs(void)
5261 ext4_li_info = NULL;
5262 mutex_init(&ext4_li_mtx);
5264 ext4_check_flag_values();
5266 for (i = 0; i < EXT4_WQ_HASH_SZ; i++) {
5267 mutex_init(&ext4__aio_mutex[i]);
5268 init_waitqueue_head(&ext4__ioend_wq[i]);
5271 err = ext4_init_pageio();
5274 err = ext4_init_system_zone();
5277 ext4_kset = kset_create_and_add("ext4", NULL, fs_kobj);
5280 ext4_proc_root = proc_mkdir("fs/ext4", NULL);
5282 err = ext4_init_feat_adverts();
5286 err = ext4_init_mballoc();
5290 err = ext4_init_xattr();
5293 err = init_inodecache();
5298 err = register_filesystem(&ext4_fs_type);
5304 unregister_as_ext2();
5305 unregister_as_ext3();
5306 destroy_inodecache();
5310 ext4_exit_mballoc();
5312 ext4_exit_feat_adverts();
5315 remove_proc_entry("fs/ext4", NULL);
5316 kset_unregister(ext4_kset);
5318 ext4_exit_system_zone();
5324 static void __exit ext4_exit_fs(void)
5326 ext4_destroy_lazyinit_thread();
5327 unregister_as_ext2();
5328 unregister_as_ext3();
5329 unregister_filesystem(&ext4_fs_type);
5330 destroy_inodecache();
5332 ext4_exit_mballoc();
5333 ext4_exit_feat_adverts();
5334 remove_proc_entry("fs/ext4", NULL);
5335 kset_unregister(ext4_kset);
5336 ext4_exit_system_zone();
5340 MODULE_AUTHOR("Remy Card, Stephen Tweedie, Andrew Morton, Andreas Dilger, Theodore Ts'o and others");
5341 MODULE_DESCRIPTION("Fourth Extended Filesystem");
5342 MODULE_LICENSE("GPL");
5343 module_init(ext4_init_fs)
5344 module_exit(ext4_exit_fs)