2 * linux/fs/ext4/super.c
4 * Copyright (C) 1992, 1993, 1994, 1995
5 * Remy Card (card@masi.ibp.fr)
6 * Laboratoire MASI - Institut Blaise Pascal
7 * Universite Pierre et Marie Curie (Paris VI)
11 * linux/fs/minix/inode.c
13 * Copyright (C) 1991, 1992 Linus Torvalds
15 * Big-endian to little-endian byte-swapping/bitmaps by
16 * David S. Miller (davem@caip.rutgers.edu), 1995
19 #include <linux/module.h>
20 #include <linux/string.h>
22 #include <linux/time.h>
23 #include <linux/vmalloc.h>
24 #include <linux/jbd2.h>
25 #include <linux/slab.h>
26 #include <linux/init.h>
27 #include <linux/blkdev.h>
28 #include <linux/parser.h>
29 #include <linux/buffer_head.h>
30 #include <linux/exportfs.h>
31 #include <linux/vfs.h>
32 #include <linux/random.h>
33 #include <linux/mount.h>
34 #include <linux/namei.h>
35 #include <linux/quotaops.h>
36 #include <linux/seq_file.h>
37 #include <linux/proc_fs.h>
38 #include <linux/ctype.h>
39 #include <linux/log2.h>
40 #include <linux/crc16.h>
41 #include <linux/cleancache.h>
42 #include <asm/uaccess.h>
44 #include <linux/kthread.h>
45 #include <linux/freezer.h>
48 #include "ext4_extents.h" /* Needed for trace points definition */
49 #include "ext4_jbd2.h"
54 #define CREATE_TRACE_POINTS
55 #include <trace/events/ext4.h>
57 static struct proc_dir_entry *ext4_proc_root;
58 static struct kset *ext4_kset;
59 static struct ext4_lazy_init *ext4_li_info;
60 static struct mutex ext4_li_mtx;
61 static struct ext4_features *ext4_feat;
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)
148 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
150 if (!EXT4_HAS_RO_COMPAT_FEATURE(sb,
151 EXT4_FEATURE_RO_COMPAT_METADATA_CSUM))
154 es->s_checksum = ext4_superblock_csum(sb, es);
157 void *ext4_kvmalloc(size_t size, gfp_t flags)
161 ret = kmalloc(size, flags);
163 ret = __vmalloc(size, flags, PAGE_KERNEL);
167 void *ext4_kvzalloc(size_t size, gfp_t flags)
171 ret = kzalloc(size, flags);
173 ret = __vmalloc(size, flags | __GFP_ZERO, PAGE_KERNEL);
177 void ext4_kvfree(void *ptr)
179 if (is_vmalloc_addr(ptr))
186 ext4_fsblk_t ext4_block_bitmap(struct super_block *sb,
187 struct ext4_group_desc *bg)
189 return le32_to_cpu(bg->bg_block_bitmap_lo) |
190 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
191 (ext4_fsblk_t)le32_to_cpu(bg->bg_block_bitmap_hi) << 32 : 0);
194 ext4_fsblk_t ext4_inode_bitmap(struct super_block *sb,
195 struct ext4_group_desc *bg)
197 return le32_to_cpu(bg->bg_inode_bitmap_lo) |
198 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
199 (ext4_fsblk_t)le32_to_cpu(bg->bg_inode_bitmap_hi) << 32 : 0);
202 ext4_fsblk_t ext4_inode_table(struct super_block *sb,
203 struct ext4_group_desc *bg)
205 return le32_to_cpu(bg->bg_inode_table_lo) |
206 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
207 (ext4_fsblk_t)le32_to_cpu(bg->bg_inode_table_hi) << 32 : 0);
210 __u32 ext4_free_group_clusters(struct super_block *sb,
211 struct ext4_group_desc *bg)
213 return le16_to_cpu(bg->bg_free_blocks_count_lo) |
214 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
215 (__u32)le16_to_cpu(bg->bg_free_blocks_count_hi) << 16 : 0);
218 __u32 ext4_free_inodes_count(struct super_block *sb,
219 struct ext4_group_desc *bg)
221 return le16_to_cpu(bg->bg_free_inodes_count_lo) |
222 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
223 (__u32)le16_to_cpu(bg->bg_free_inodes_count_hi) << 16 : 0);
226 __u32 ext4_used_dirs_count(struct super_block *sb,
227 struct ext4_group_desc *bg)
229 return le16_to_cpu(bg->bg_used_dirs_count_lo) |
230 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
231 (__u32)le16_to_cpu(bg->bg_used_dirs_count_hi) << 16 : 0);
234 __u32 ext4_itable_unused_count(struct super_block *sb,
235 struct ext4_group_desc *bg)
237 return le16_to_cpu(bg->bg_itable_unused_lo) |
238 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
239 (__u32)le16_to_cpu(bg->bg_itable_unused_hi) << 16 : 0);
242 void ext4_block_bitmap_set(struct super_block *sb,
243 struct ext4_group_desc *bg, ext4_fsblk_t blk)
245 bg->bg_block_bitmap_lo = cpu_to_le32((u32)blk);
246 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
247 bg->bg_block_bitmap_hi = cpu_to_le32(blk >> 32);
250 void ext4_inode_bitmap_set(struct super_block *sb,
251 struct ext4_group_desc *bg, ext4_fsblk_t blk)
253 bg->bg_inode_bitmap_lo = cpu_to_le32((u32)blk);
254 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
255 bg->bg_inode_bitmap_hi = cpu_to_le32(blk >> 32);
258 void ext4_inode_table_set(struct super_block *sb,
259 struct ext4_group_desc *bg, ext4_fsblk_t blk)
261 bg->bg_inode_table_lo = cpu_to_le32((u32)blk);
262 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
263 bg->bg_inode_table_hi = cpu_to_le32(blk >> 32);
266 void ext4_free_group_clusters_set(struct super_block *sb,
267 struct ext4_group_desc *bg, __u32 count)
269 bg->bg_free_blocks_count_lo = cpu_to_le16((__u16)count);
270 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
271 bg->bg_free_blocks_count_hi = cpu_to_le16(count >> 16);
274 void ext4_free_inodes_set(struct super_block *sb,
275 struct ext4_group_desc *bg, __u32 count)
277 bg->bg_free_inodes_count_lo = cpu_to_le16((__u16)count);
278 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
279 bg->bg_free_inodes_count_hi = cpu_to_le16(count >> 16);
282 void ext4_used_dirs_set(struct super_block *sb,
283 struct ext4_group_desc *bg, __u32 count)
285 bg->bg_used_dirs_count_lo = cpu_to_le16((__u16)count);
286 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
287 bg->bg_used_dirs_count_hi = cpu_to_le16(count >> 16);
290 void ext4_itable_unused_set(struct super_block *sb,
291 struct ext4_group_desc *bg, __u32 count)
293 bg->bg_itable_unused_lo = cpu_to_le16((__u16)count);
294 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
295 bg->bg_itable_unused_hi = cpu_to_le16(count >> 16);
299 /* Just increment the non-pointer handle value */
300 static handle_t *ext4_get_nojournal(void)
302 handle_t *handle = current->journal_info;
303 unsigned long ref_cnt = (unsigned long)handle;
305 BUG_ON(ref_cnt >= EXT4_NOJOURNAL_MAX_REF_COUNT);
308 handle = (handle_t *)ref_cnt;
310 current->journal_info = handle;
315 /* Decrement the non-pointer handle value */
316 static void ext4_put_nojournal(handle_t *handle)
318 unsigned long ref_cnt = (unsigned long)handle;
320 BUG_ON(ref_cnt == 0);
323 handle = (handle_t *)ref_cnt;
325 current->journal_info = handle;
329 * Wrappers for jbd2_journal_start/end.
331 handle_t *ext4_journal_start_sb(struct super_block *sb, int nblocks)
335 trace_ext4_journal_start(sb, nblocks, _RET_IP_);
336 if (sb->s_flags & MS_RDONLY)
337 return ERR_PTR(-EROFS);
339 WARN_ON(sb->s_writers.frozen == SB_FREEZE_COMPLETE);
340 journal = EXT4_SB(sb)->s_journal;
342 return ext4_get_nojournal();
344 * Special case here: if the journal has aborted behind our
345 * backs (eg. EIO in the commit thread), then we still need to
346 * take the FS itself readonly cleanly.
348 if (is_journal_aborted(journal)) {
349 ext4_abort(sb, "Detected aborted journal");
350 return ERR_PTR(-EROFS);
352 return jbd2_journal_start(journal, nblocks);
355 int __ext4_journal_stop(const char *where, unsigned int line, handle_t *handle)
357 struct super_block *sb;
361 if (!ext4_handle_valid(handle)) {
362 ext4_put_nojournal(handle);
365 sb = handle->h_transaction->t_journal->j_private;
367 rc = jbd2_journal_stop(handle);
372 __ext4_std_error(sb, where, line, err);
376 void ext4_journal_abort_handle(const char *caller, unsigned int line,
377 const char *err_fn, struct buffer_head *bh,
378 handle_t *handle, int err)
381 const char *errstr = ext4_decode_error(NULL, err, nbuf);
383 BUG_ON(!ext4_handle_valid(handle));
386 BUFFER_TRACE(bh, "abort");
391 if (is_handle_aborted(handle))
394 printk(KERN_ERR "EXT4-fs: %s:%d: aborting transaction: %s in %s\n",
395 caller, line, errstr, err_fn);
397 jbd2_journal_abort_handle(handle);
400 static void __save_error_info(struct super_block *sb, const char *func,
403 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
405 EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS;
406 es->s_state |= cpu_to_le16(EXT4_ERROR_FS);
407 es->s_last_error_time = cpu_to_le32(get_seconds());
408 strncpy(es->s_last_error_func, func, sizeof(es->s_last_error_func));
409 es->s_last_error_line = cpu_to_le32(line);
410 if (!es->s_first_error_time) {
411 es->s_first_error_time = es->s_last_error_time;
412 strncpy(es->s_first_error_func, func,
413 sizeof(es->s_first_error_func));
414 es->s_first_error_line = cpu_to_le32(line);
415 es->s_first_error_ino = es->s_last_error_ino;
416 es->s_first_error_block = es->s_last_error_block;
419 * Start the daily error reporting function if it hasn't been
422 if (!es->s_error_count)
423 mod_timer(&EXT4_SB(sb)->s_err_report, jiffies + 24*60*60*HZ);
424 le32_add_cpu(&es->s_error_count, 1);
427 static void save_error_info(struct super_block *sb, const char *func,
430 __save_error_info(sb, func, line);
431 ext4_commit_super(sb, 1);
435 * The del_gendisk() function uninitializes the disk-specific data
436 * structures, including the bdi structure, without telling anyone
437 * else. Once this happens, any attempt to call mark_buffer_dirty()
438 * (for example, by ext4_commit_super), will cause a kernel OOPS.
439 * This is a kludge to prevent these oops until we can put in a proper
440 * hook in del_gendisk() to inform the VFS and file system layers.
442 static int block_device_ejected(struct super_block *sb)
444 struct inode *bd_inode = sb->s_bdev->bd_inode;
445 struct backing_dev_info *bdi = bd_inode->i_mapping->backing_dev_info;
447 return bdi->dev == NULL;
450 static void ext4_journal_commit_callback(journal_t *journal, transaction_t *txn)
452 struct super_block *sb = journal->j_private;
453 struct ext4_sb_info *sbi = EXT4_SB(sb);
454 int error = is_journal_aborted(journal);
455 struct ext4_journal_cb_entry *jce;
457 BUG_ON(txn->t_state == T_FINISHED);
458 spin_lock(&sbi->s_md_lock);
459 while (!list_empty(&txn->t_private_list)) {
460 jce = list_entry(txn->t_private_list.next,
461 struct ext4_journal_cb_entry, jce_list);
462 list_del_init(&jce->jce_list);
463 spin_unlock(&sbi->s_md_lock);
464 jce->jce_func(sb, jce, error);
465 spin_lock(&sbi->s_md_lock);
467 spin_unlock(&sbi->s_md_lock);
470 /* Deal with the reporting of failure conditions on a filesystem such as
471 * inconsistencies detected or read IO failures.
473 * On ext2, we can store the error state of the filesystem in the
474 * superblock. That is not possible on ext4, because we may have other
475 * write ordering constraints on the superblock which prevent us from
476 * writing it out straight away; and given that the journal is about to
477 * be aborted, we can't rely on the current, or future, transactions to
478 * write out the superblock safely.
480 * We'll just use the jbd2_journal_abort() error code to record an error in
481 * the journal instead. On recovery, the journal will complain about
482 * that error until we've noted it down and cleared it.
485 static void ext4_handle_error(struct super_block *sb)
487 if (sb->s_flags & MS_RDONLY)
490 if (!test_opt(sb, ERRORS_CONT)) {
491 journal_t *journal = EXT4_SB(sb)->s_journal;
493 EXT4_SB(sb)->s_mount_flags |= EXT4_MF_FS_ABORTED;
495 jbd2_journal_abort(journal, -EIO);
497 if (test_opt(sb, ERRORS_RO)) {
498 ext4_msg(sb, KERN_CRIT, "Remounting filesystem read-only");
499 sb->s_flags |= MS_RDONLY;
501 if (test_opt(sb, ERRORS_PANIC))
502 panic("EXT4-fs (device %s): panic forced after error\n",
506 void __ext4_error(struct super_block *sb, const char *function,
507 unsigned int line, const char *fmt, ...)
509 struct va_format vaf;
515 printk(KERN_CRIT "EXT4-fs error (device %s): %s:%d: comm %s: %pV\n",
516 sb->s_id, function, line, current->comm, &vaf);
518 save_error_info(sb, function, line);
520 ext4_handle_error(sb);
523 void ext4_error_inode(struct inode *inode, const char *function,
524 unsigned int line, ext4_fsblk_t block,
525 const char *fmt, ...)
528 struct va_format vaf;
529 struct ext4_super_block *es = EXT4_SB(inode->i_sb)->s_es;
531 es->s_last_error_ino = cpu_to_le32(inode->i_ino);
532 es->s_last_error_block = cpu_to_le64(block);
533 save_error_info(inode->i_sb, function, line);
538 printk(KERN_CRIT "EXT4-fs error (device %s): %s:%d: "
539 "inode #%lu: block %llu: comm %s: %pV\n",
540 inode->i_sb->s_id, function, line, inode->i_ino,
541 block, current->comm, &vaf);
543 printk(KERN_CRIT "EXT4-fs error (device %s): %s:%d: "
544 "inode #%lu: comm %s: %pV\n",
545 inode->i_sb->s_id, function, line, inode->i_ino,
546 current->comm, &vaf);
549 ext4_handle_error(inode->i_sb);
552 void ext4_error_file(struct file *file, const char *function,
553 unsigned int line, ext4_fsblk_t block,
554 const char *fmt, ...)
557 struct va_format vaf;
558 struct ext4_super_block *es;
559 struct inode *inode = file->f_dentry->d_inode;
560 char pathname[80], *path;
562 es = EXT4_SB(inode->i_sb)->s_es;
563 es->s_last_error_ino = cpu_to_le32(inode->i_ino);
564 save_error_info(inode->i_sb, function, line);
565 path = d_path(&(file->f_path), pathname, sizeof(pathname));
573 "EXT4-fs error (device %s): %s:%d: inode #%lu: "
574 "block %llu: comm %s: path %s: %pV\n",
575 inode->i_sb->s_id, function, line, inode->i_ino,
576 block, current->comm, path, &vaf);
579 "EXT4-fs error (device %s): %s:%d: inode #%lu: "
580 "comm %s: path %s: %pV\n",
581 inode->i_sb->s_id, function, line, inode->i_ino,
582 current->comm, path, &vaf);
585 ext4_handle_error(inode->i_sb);
588 static const char *ext4_decode_error(struct super_block *sb, int errno,
595 errstr = "IO failure";
598 errstr = "Out of memory";
601 if (!sb || (EXT4_SB(sb)->s_journal &&
602 EXT4_SB(sb)->s_journal->j_flags & JBD2_ABORT))
603 errstr = "Journal has aborted";
605 errstr = "Readonly filesystem";
608 /* If the caller passed in an extra buffer for unknown
609 * errors, textualise them now. Else we just return
612 /* Check for truncated error codes... */
613 if (snprintf(nbuf, 16, "error %d", -errno) >= 0)
622 /* __ext4_std_error decodes expected errors from journaling functions
623 * automatically and invokes the appropriate error response. */
625 void __ext4_std_error(struct super_block *sb, const char *function,
626 unsigned int line, int errno)
631 /* Special case: if the error is EROFS, and we're not already
632 * inside a transaction, then there's really no point in logging
634 if (errno == -EROFS && journal_current_handle() == NULL &&
635 (sb->s_flags & MS_RDONLY))
638 errstr = ext4_decode_error(sb, errno, nbuf);
639 printk(KERN_CRIT "EXT4-fs error (device %s) in %s:%d: %s\n",
640 sb->s_id, function, line, errstr);
641 save_error_info(sb, function, line);
643 ext4_handle_error(sb);
647 * ext4_abort is a much stronger failure handler than ext4_error. The
648 * abort function may be used to deal with unrecoverable failures such
649 * as journal IO errors or ENOMEM at a critical moment in log management.
651 * We unconditionally force the filesystem into an ABORT|READONLY state,
652 * unless the error response on the fs has been set to panic in which
653 * case we take the easy way out and panic immediately.
656 void __ext4_abort(struct super_block *sb, const char *function,
657 unsigned int line, const char *fmt, ...)
661 save_error_info(sb, function, line);
663 printk(KERN_CRIT "EXT4-fs error (device %s): %s:%d: ", sb->s_id,
669 if ((sb->s_flags & MS_RDONLY) == 0) {
670 ext4_msg(sb, KERN_CRIT, "Remounting filesystem read-only");
671 sb->s_flags |= MS_RDONLY;
672 EXT4_SB(sb)->s_mount_flags |= EXT4_MF_FS_ABORTED;
673 if (EXT4_SB(sb)->s_journal)
674 jbd2_journal_abort(EXT4_SB(sb)->s_journal, -EIO);
675 save_error_info(sb, function, line);
677 if (test_opt(sb, ERRORS_PANIC))
678 panic("EXT4-fs panic from previous error\n");
681 void ext4_msg(struct super_block *sb, const char *prefix, const char *fmt, ...)
683 struct va_format vaf;
689 printk("%sEXT4-fs (%s): %pV\n", prefix, sb->s_id, &vaf);
693 void __ext4_warning(struct super_block *sb, const char *function,
694 unsigned int line, const char *fmt, ...)
696 struct va_format vaf;
702 printk(KERN_WARNING "EXT4-fs warning (device %s): %s:%d: %pV\n",
703 sb->s_id, function, line, &vaf);
707 void __ext4_grp_locked_error(const char *function, unsigned int line,
708 struct super_block *sb, ext4_group_t grp,
709 unsigned long ino, ext4_fsblk_t block,
710 const char *fmt, ...)
714 struct va_format vaf;
716 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
718 es->s_last_error_ino = cpu_to_le32(ino);
719 es->s_last_error_block = cpu_to_le64(block);
720 __save_error_info(sb, function, line);
726 printk(KERN_CRIT "EXT4-fs error (device %s): %s:%d: group %u, ",
727 sb->s_id, function, line, grp);
729 printk(KERN_CONT "inode %lu: ", ino);
731 printk(KERN_CONT "block %llu:", (unsigned long long) block);
732 printk(KERN_CONT "%pV\n", &vaf);
735 if (test_opt(sb, ERRORS_CONT)) {
736 ext4_commit_super(sb, 0);
740 ext4_unlock_group(sb, grp);
741 ext4_handle_error(sb);
743 * We only get here in the ERRORS_RO case; relocking the group
744 * may be dangerous, but nothing bad will happen since the
745 * filesystem will have already been marked read/only and the
746 * journal has been aborted. We return 1 as a hint to callers
747 * who might what to use the return value from
748 * ext4_grp_locked_error() to distinguish between the
749 * ERRORS_CONT and ERRORS_RO case, and perhaps return more
750 * aggressively from the ext4 function in question, with a
751 * more appropriate error code.
753 ext4_lock_group(sb, grp);
757 void ext4_update_dynamic_rev(struct super_block *sb)
759 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
761 if (le32_to_cpu(es->s_rev_level) > EXT4_GOOD_OLD_REV)
765 "updating to rev %d because of new feature flag, "
766 "running e2fsck is recommended",
769 es->s_first_ino = cpu_to_le32(EXT4_GOOD_OLD_FIRST_INO);
770 es->s_inode_size = cpu_to_le16(EXT4_GOOD_OLD_INODE_SIZE);
771 es->s_rev_level = cpu_to_le32(EXT4_DYNAMIC_REV);
772 /* leave es->s_feature_*compat flags alone */
773 /* es->s_uuid will be set by e2fsck if empty */
776 * The rest of the superblock fields should be zero, and if not it
777 * means they are likely already in use, so leave them alone. We
778 * can leave it up to e2fsck to clean up any inconsistencies there.
783 * Open the external journal device
785 static struct block_device *ext4_blkdev_get(dev_t dev, struct super_block *sb)
787 struct block_device *bdev;
788 char b[BDEVNAME_SIZE];
790 bdev = blkdev_get_by_dev(dev, FMODE_READ|FMODE_WRITE|FMODE_EXCL, sb);
796 ext4_msg(sb, KERN_ERR, "failed to open journal device %s: %ld",
797 __bdevname(dev, b), PTR_ERR(bdev));
802 * Release the journal device
804 static int ext4_blkdev_put(struct block_device *bdev)
806 return blkdev_put(bdev, FMODE_READ|FMODE_WRITE|FMODE_EXCL);
809 static int ext4_blkdev_remove(struct ext4_sb_info *sbi)
811 struct block_device *bdev;
814 bdev = sbi->journal_bdev;
816 ret = ext4_blkdev_put(bdev);
817 sbi->journal_bdev = NULL;
822 static inline struct inode *orphan_list_entry(struct list_head *l)
824 return &list_entry(l, struct ext4_inode_info, i_orphan)->vfs_inode;
827 static void dump_orphan_list(struct super_block *sb, struct ext4_sb_info *sbi)
831 ext4_msg(sb, KERN_ERR, "sb orphan head is %d",
832 le32_to_cpu(sbi->s_es->s_last_orphan));
834 printk(KERN_ERR "sb_info orphan list:\n");
835 list_for_each(l, &sbi->s_orphan) {
836 struct inode *inode = orphan_list_entry(l);
838 "inode %s:%lu at %p: mode %o, nlink %d, next %d\n",
839 inode->i_sb->s_id, inode->i_ino, inode,
840 inode->i_mode, inode->i_nlink,
845 static void ext4_put_super(struct super_block *sb)
847 struct ext4_sb_info *sbi = EXT4_SB(sb);
848 struct ext4_super_block *es = sbi->s_es;
851 ext4_unregister_li_request(sb);
852 dquot_disable(sb, -1, DQUOT_USAGE_ENABLED | DQUOT_LIMITS_ENABLED);
854 flush_workqueue(sbi->dio_unwritten_wq);
855 destroy_workqueue(sbi->dio_unwritten_wq);
857 if (sbi->s_journal) {
858 err = jbd2_journal_destroy(sbi->s_journal);
859 sbi->s_journal = NULL;
861 ext4_abort(sb, "Couldn't clean up the journal");
864 del_timer(&sbi->s_err_report);
865 ext4_release_system_zone(sb);
867 ext4_ext_release(sb);
868 ext4_xattr_put_super(sb);
870 if (!(sb->s_flags & MS_RDONLY)) {
871 EXT4_CLEAR_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
872 es->s_state = cpu_to_le16(sbi->s_mount_state);
874 if (!(sb->s_flags & MS_RDONLY))
875 ext4_commit_super(sb, 1);
878 remove_proc_entry("options", sbi->s_proc);
879 remove_proc_entry(sb->s_id, ext4_proc_root);
881 kobject_del(&sbi->s_kobj);
883 for (i = 0; i < sbi->s_gdb_count; i++)
884 brelse(sbi->s_group_desc[i]);
885 ext4_kvfree(sbi->s_group_desc);
886 ext4_kvfree(sbi->s_flex_groups);
887 percpu_counter_destroy(&sbi->s_freeclusters_counter);
888 percpu_counter_destroy(&sbi->s_freeinodes_counter);
889 percpu_counter_destroy(&sbi->s_dirs_counter);
890 percpu_counter_destroy(&sbi->s_dirtyclusters_counter);
893 for (i = 0; i < MAXQUOTAS; i++)
894 kfree(sbi->s_qf_names[i]);
897 /* Debugging code just in case the in-memory inode orphan list
898 * isn't empty. The on-disk one can be non-empty if we've
899 * detected an error and taken the fs readonly, but the
900 * in-memory list had better be clean by this point. */
901 if (!list_empty(&sbi->s_orphan))
902 dump_orphan_list(sb, sbi);
903 J_ASSERT(list_empty(&sbi->s_orphan));
905 invalidate_bdev(sb->s_bdev);
906 if (sbi->journal_bdev && sbi->journal_bdev != sb->s_bdev) {
908 * Invalidate the journal device's buffers. We don't want them
909 * floating about in memory - the physical journal device may
910 * hotswapped, and it breaks the `ro-after' testing code.
912 sync_blockdev(sbi->journal_bdev);
913 invalidate_bdev(sbi->journal_bdev);
914 ext4_blkdev_remove(sbi);
917 kthread_stop(sbi->s_mmp_tsk);
918 sb->s_fs_info = NULL;
920 * Now that we are completely done shutting down the
921 * superblock, we need to actually destroy the kobject.
923 kobject_put(&sbi->s_kobj);
924 wait_for_completion(&sbi->s_kobj_unregister);
925 if (sbi->s_chksum_driver)
926 crypto_free_shash(sbi->s_chksum_driver);
927 kfree(sbi->s_blockgroup_lock);
931 static struct kmem_cache *ext4_inode_cachep;
934 * Called inside transaction, so use GFP_NOFS
936 static struct inode *ext4_alloc_inode(struct super_block *sb)
938 struct ext4_inode_info *ei;
940 ei = kmem_cache_alloc(ext4_inode_cachep, GFP_NOFS);
944 ei->vfs_inode.i_version = 1;
945 memset(&ei->i_cached_extent, 0, sizeof(struct ext4_ext_cache));
946 INIT_LIST_HEAD(&ei->i_prealloc_list);
947 spin_lock_init(&ei->i_prealloc_lock);
948 ext4_es_init_tree(&ei->i_es_tree);
949 rwlock_init(&ei->i_es_lock);
950 ei->i_reserved_data_blocks = 0;
951 ei->i_reserved_meta_blocks = 0;
952 ei->i_allocated_meta_blocks = 0;
953 ei->i_da_metadata_calc_len = 0;
954 ei->i_da_metadata_calc_last_lblock = 0;
955 spin_lock_init(&(ei->i_block_reservation_lock));
957 ei->i_reserved_quota = 0;
960 INIT_LIST_HEAD(&ei->i_completed_io_list);
961 spin_lock_init(&ei->i_completed_io_lock);
963 ei->i_datasync_tid = 0;
964 atomic_set(&ei->i_ioend_count, 0);
965 atomic_set(&ei->i_unwritten, 0);
967 return &ei->vfs_inode;
970 static int ext4_drop_inode(struct inode *inode)
972 int drop = generic_drop_inode(inode);
974 trace_ext4_drop_inode(inode, drop);
978 static void ext4_i_callback(struct rcu_head *head)
980 struct inode *inode = container_of(head, struct inode, i_rcu);
981 kmem_cache_free(ext4_inode_cachep, EXT4_I(inode));
984 static void ext4_destroy_inode(struct inode *inode)
986 if (!list_empty(&(EXT4_I(inode)->i_orphan))) {
987 ext4_msg(inode->i_sb, KERN_ERR,
988 "Inode %lu (%p): orphan list check failed!",
989 inode->i_ino, EXT4_I(inode));
990 print_hex_dump(KERN_INFO, "", DUMP_PREFIX_ADDRESS, 16, 4,
991 EXT4_I(inode), sizeof(struct ext4_inode_info),
995 call_rcu(&inode->i_rcu, ext4_i_callback);
998 static void init_once(void *foo)
1000 struct ext4_inode_info *ei = (struct ext4_inode_info *) foo;
1002 INIT_LIST_HEAD(&ei->i_orphan);
1003 init_rwsem(&ei->xattr_sem);
1004 init_rwsem(&ei->i_data_sem);
1005 inode_init_once(&ei->vfs_inode);
1008 static int init_inodecache(void)
1010 ext4_inode_cachep = kmem_cache_create("ext4_inode_cache",
1011 sizeof(struct ext4_inode_info),
1012 0, (SLAB_RECLAIM_ACCOUNT|
1015 if (ext4_inode_cachep == NULL)
1020 static void destroy_inodecache(void)
1023 * Make sure all delayed rcu free inodes are flushed before we
1027 kmem_cache_destroy(ext4_inode_cachep);
1030 void ext4_clear_inode(struct inode *inode)
1032 invalidate_inode_buffers(inode);
1035 ext4_discard_preallocations(inode);
1036 ext4_es_remove_extent(inode, 0, EXT_MAX_BLOCKS);
1037 if (EXT4_I(inode)->jinode) {
1038 jbd2_journal_release_jbd_inode(EXT4_JOURNAL(inode),
1039 EXT4_I(inode)->jinode);
1040 jbd2_free_inode(EXT4_I(inode)->jinode);
1041 EXT4_I(inode)->jinode = NULL;
1045 static struct inode *ext4_nfs_get_inode(struct super_block *sb,
1046 u64 ino, u32 generation)
1048 struct inode *inode;
1050 if (ino < EXT4_FIRST_INO(sb) && ino != EXT4_ROOT_INO)
1051 return ERR_PTR(-ESTALE);
1052 if (ino > le32_to_cpu(EXT4_SB(sb)->s_es->s_inodes_count))
1053 return ERR_PTR(-ESTALE);
1055 /* iget isn't really right if the inode is currently unallocated!!
1057 * ext4_read_inode will return a bad_inode if the inode had been
1058 * deleted, so we should be safe.
1060 * Currently we don't know the generation for parent directory, so
1061 * a generation of 0 means "accept any"
1063 inode = ext4_iget(sb, ino);
1065 return ERR_CAST(inode);
1066 if (generation && inode->i_generation != generation) {
1068 return ERR_PTR(-ESTALE);
1074 static struct dentry *ext4_fh_to_dentry(struct super_block *sb, struct fid *fid,
1075 int fh_len, int fh_type)
1077 return generic_fh_to_dentry(sb, fid, fh_len, fh_type,
1078 ext4_nfs_get_inode);
1081 static struct dentry *ext4_fh_to_parent(struct super_block *sb, struct fid *fid,
1082 int fh_len, int fh_type)
1084 return generic_fh_to_parent(sb, fid, fh_len, fh_type,
1085 ext4_nfs_get_inode);
1089 * Try to release metadata pages (indirect blocks, directories) which are
1090 * mapped via the block device. Since these pages could have journal heads
1091 * which would prevent try_to_free_buffers() from freeing them, we must use
1092 * jbd2 layer's try_to_free_buffers() function to release them.
1094 static int bdev_try_to_free_page(struct super_block *sb, struct page *page,
1097 journal_t *journal = EXT4_SB(sb)->s_journal;
1099 WARN_ON(PageChecked(page));
1100 if (!page_has_buffers(page))
1103 return jbd2_journal_try_to_free_buffers(journal, page,
1104 wait & ~__GFP_WAIT);
1105 return try_to_free_buffers(page);
1109 #define QTYPE2NAME(t) ((t) == USRQUOTA ? "user" : "group")
1110 #define QTYPE2MOPT(on, t) ((t) == USRQUOTA?((on)##USRJQUOTA):((on)##GRPJQUOTA))
1112 static int ext4_write_dquot(struct dquot *dquot);
1113 static int ext4_acquire_dquot(struct dquot *dquot);
1114 static int ext4_release_dquot(struct dquot *dquot);
1115 static int ext4_mark_dquot_dirty(struct dquot *dquot);
1116 static int ext4_write_info(struct super_block *sb, int type);
1117 static int ext4_quota_on(struct super_block *sb, int type, int format_id,
1119 static int ext4_quota_on_sysfile(struct super_block *sb, int type,
1121 static int ext4_quota_off(struct super_block *sb, int type);
1122 static int ext4_quota_off_sysfile(struct super_block *sb, int type);
1123 static int ext4_quota_on_mount(struct super_block *sb, int type);
1124 static ssize_t ext4_quota_read(struct super_block *sb, int type, char *data,
1125 size_t len, loff_t off);
1126 static ssize_t ext4_quota_write(struct super_block *sb, int type,
1127 const char *data, size_t len, loff_t off);
1128 static int ext4_quota_enable(struct super_block *sb, int type, int format_id,
1129 unsigned int flags);
1130 static int ext4_enable_quotas(struct super_block *sb);
1132 static const struct dquot_operations ext4_quota_operations = {
1133 .get_reserved_space = ext4_get_reserved_space,
1134 .write_dquot = ext4_write_dquot,
1135 .acquire_dquot = ext4_acquire_dquot,
1136 .release_dquot = ext4_release_dquot,
1137 .mark_dirty = ext4_mark_dquot_dirty,
1138 .write_info = ext4_write_info,
1139 .alloc_dquot = dquot_alloc,
1140 .destroy_dquot = dquot_destroy,
1143 static const struct quotactl_ops ext4_qctl_operations = {
1144 .quota_on = ext4_quota_on,
1145 .quota_off = ext4_quota_off,
1146 .quota_sync = dquot_quota_sync,
1147 .get_info = dquot_get_dqinfo,
1148 .set_info = dquot_set_dqinfo,
1149 .get_dqblk = dquot_get_dqblk,
1150 .set_dqblk = dquot_set_dqblk
1153 static const struct quotactl_ops ext4_qctl_sysfile_operations = {
1154 .quota_on_meta = ext4_quota_on_sysfile,
1155 .quota_off = ext4_quota_off_sysfile,
1156 .quota_sync = dquot_quota_sync,
1157 .get_info = dquot_get_dqinfo,
1158 .set_info = dquot_set_dqinfo,
1159 .get_dqblk = dquot_get_dqblk,
1160 .set_dqblk = dquot_set_dqblk
1164 static const struct super_operations ext4_sops = {
1165 .alloc_inode = ext4_alloc_inode,
1166 .destroy_inode = ext4_destroy_inode,
1167 .write_inode = ext4_write_inode,
1168 .dirty_inode = ext4_dirty_inode,
1169 .drop_inode = ext4_drop_inode,
1170 .evict_inode = ext4_evict_inode,
1171 .put_super = ext4_put_super,
1172 .sync_fs = ext4_sync_fs,
1173 .freeze_fs = ext4_freeze,
1174 .unfreeze_fs = ext4_unfreeze,
1175 .statfs = ext4_statfs,
1176 .remount_fs = ext4_remount,
1177 .show_options = ext4_show_options,
1179 .quota_read = ext4_quota_read,
1180 .quota_write = ext4_quota_write,
1182 .bdev_try_to_free_page = bdev_try_to_free_page,
1185 static const struct super_operations ext4_nojournal_sops = {
1186 .alloc_inode = ext4_alloc_inode,
1187 .destroy_inode = ext4_destroy_inode,
1188 .write_inode = ext4_write_inode,
1189 .dirty_inode = ext4_dirty_inode,
1190 .drop_inode = ext4_drop_inode,
1191 .evict_inode = ext4_evict_inode,
1192 .put_super = ext4_put_super,
1193 .statfs = ext4_statfs,
1194 .remount_fs = ext4_remount,
1195 .show_options = ext4_show_options,
1197 .quota_read = ext4_quota_read,
1198 .quota_write = ext4_quota_write,
1200 .bdev_try_to_free_page = bdev_try_to_free_page,
1203 static const struct export_operations ext4_export_ops = {
1204 .fh_to_dentry = ext4_fh_to_dentry,
1205 .fh_to_parent = ext4_fh_to_parent,
1206 .get_parent = ext4_get_parent,
1210 Opt_bsd_df, Opt_minix_df, Opt_grpid, Opt_nogrpid,
1211 Opt_resgid, Opt_resuid, Opt_sb, Opt_err_cont, Opt_err_panic, Opt_err_ro,
1212 Opt_nouid32, Opt_debug, Opt_removed,
1213 Opt_user_xattr, Opt_nouser_xattr, Opt_acl, Opt_noacl,
1214 Opt_auto_da_alloc, Opt_noauto_da_alloc, Opt_noload,
1215 Opt_commit, Opt_min_batch_time, Opt_max_batch_time,
1216 Opt_journal_dev, Opt_journal_checksum, Opt_journal_async_commit,
1217 Opt_abort, Opt_data_journal, Opt_data_ordered, Opt_data_writeback,
1218 Opt_data_err_abort, Opt_data_err_ignore,
1219 Opt_usrjquota, Opt_grpjquota, Opt_offusrjquota, Opt_offgrpjquota,
1220 Opt_jqfmt_vfsold, Opt_jqfmt_vfsv0, Opt_jqfmt_vfsv1, Opt_quota,
1221 Opt_noquota, Opt_barrier, Opt_nobarrier, Opt_err,
1222 Opt_usrquota, Opt_grpquota, Opt_i_version,
1223 Opt_stripe, Opt_delalloc, Opt_nodelalloc, Opt_mblk_io_submit,
1224 Opt_nomblk_io_submit, Opt_block_validity, Opt_noblock_validity,
1225 Opt_inode_readahead_blks, Opt_journal_ioprio,
1226 Opt_dioread_nolock, Opt_dioread_lock,
1227 Opt_discard, Opt_nodiscard, Opt_init_itable, Opt_noinit_itable,
1228 Opt_max_dir_size_kb,
1231 static const match_table_t tokens = {
1232 {Opt_bsd_df, "bsddf"},
1233 {Opt_minix_df, "minixdf"},
1234 {Opt_grpid, "grpid"},
1235 {Opt_grpid, "bsdgroups"},
1236 {Opt_nogrpid, "nogrpid"},
1237 {Opt_nogrpid, "sysvgroups"},
1238 {Opt_resgid, "resgid=%u"},
1239 {Opt_resuid, "resuid=%u"},
1241 {Opt_err_cont, "errors=continue"},
1242 {Opt_err_panic, "errors=panic"},
1243 {Opt_err_ro, "errors=remount-ro"},
1244 {Opt_nouid32, "nouid32"},
1245 {Opt_debug, "debug"},
1246 {Opt_removed, "oldalloc"},
1247 {Opt_removed, "orlov"},
1248 {Opt_user_xattr, "user_xattr"},
1249 {Opt_nouser_xattr, "nouser_xattr"},
1251 {Opt_noacl, "noacl"},
1252 {Opt_noload, "norecovery"},
1253 {Opt_noload, "noload"},
1254 {Opt_removed, "nobh"},
1255 {Opt_removed, "bh"},
1256 {Opt_commit, "commit=%u"},
1257 {Opt_min_batch_time, "min_batch_time=%u"},
1258 {Opt_max_batch_time, "max_batch_time=%u"},
1259 {Opt_journal_dev, "journal_dev=%u"},
1260 {Opt_journal_checksum, "journal_checksum"},
1261 {Opt_journal_async_commit, "journal_async_commit"},
1262 {Opt_abort, "abort"},
1263 {Opt_data_journal, "data=journal"},
1264 {Opt_data_ordered, "data=ordered"},
1265 {Opt_data_writeback, "data=writeback"},
1266 {Opt_data_err_abort, "data_err=abort"},
1267 {Opt_data_err_ignore, "data_err=ignore"},
1268 {Opt_offusrjquota, "usrjquota="},
1269 {Opt_usrjquota, "usrjquota=%s"},
1270 {Opt_offgrpjquota, "grpjquota="},
1271 {Opt_grpjquota, "grpjquota=%s"},
1272 {Opt_jqfmt_vfsold, "jqfmt=vfsold"},
1273 {Opt_jqfmt_vfsv0, "jqfmt=vfsv0"},
1274 {Opt_jqfmt_vfsv1, "jqfmt=vfsv1"},
1275 {Opt_grpquota, "grpquota"},
1276 {Opt_noquota, "noquota"},
1277 {Opt_quota, "quota"},
1278 {Opt_usrquota, "usrquota"},
1279 {Opt_barrier, "barrier=%u"},
1280 {Opt_barrier, "barrier"},
1281 {Opt_nobarrier, "nobarrier"},
1282 {Opt_i_version, "i_version"},
1283 {Opt_stripe, "stripe=%u"},
1284 {Opt_delalloc, "delalloc"},
1285 {Opt_nodelalloc, "nodelalloc"},
1286 {Opt_mblk_io_submit, "mblk_io_submit"},
1287 {Opt_nomblk_io_submit, "nomblk_io_submit"},
1288 {Opt_block_validity, "block_validity"},
1289 {Opt_noblock_validity, "noblock_validity"},
1290 {Opt_inode_readahead_blks, "inode_readahead_blks=%u"},
1291 {Opt_journal_ioprio, "journal_ioprio=%u"},
1292 {Opt_auto_da_alloc, "auto_da_alloc=%u"},
1293 {Opt_auto_da_alloc, "auto_da_alloc"},
1294 {Opt_noauto_da_alloc, "noauto_da_alloc"},
1295 {Opt_dioread_nolock, "dioread_nolock"},
1296 {Opt_dioread_lock, "dioread_lock"},
1297 {Opt_discard, "discard"},
1298 {Opt_nodiscard, "nodiscard"},
1299 {Opt_init_itable, "init_itable=%u"},
1300 {Opt_init_itable, "init_itable"},
1301 {Opt_noinit_itable, "noinit_itable"},
1302 {Opt_max_dir_size_kb, "max_dir_size_kb=%u"},
1303 {Opt_removed, "check=none"}, /* mount option from ext2/3 */
1304 {Opt_removed, "nocheck"}, /* mount option from ext2/3 */
1305 {Opt_removed, "reservation"}, /* mount option from ext2/3 */
1306 {Opt_removed, "noreservation"}, /* mount option from ext2/3 */
1307 {Opt_removed, "journal=%u"}, /* mount option from ext2/3 */
1311 static ext4_fsblk_t get_sb_block(void **data)
1313 ext4_fsblk_t sb_block;
1314 char *options = (char *) *data;
1316 if (!options || strncmp(options, "sb=", 3) != 0)
1317 return 1; /* Default location */
1320 /* TODO: use simple_strtoll with >32bit ext4 */
1321 sb_block = simple_strtoul(options, &options, 0);
1322 if (*options && *options != ',') {
1323 printk(KERN_ERR "EXT4-fs: Invalid sb specification: %s\n",
1327 if (*options == ',')
1329 *data = (void *) options;
1334 #define DEFAULT_JOURNAL_IOPRIO (IOPRIO_PRIO_VALUE(IOPRIO_CLASS_BE, 3))
1335 static char deprecated_msg[] = "Mount option \"%s\" will be removed by %s\n"
1336 "Contact linux-ext4@vger.kernel.org if you think we should keep it.\n";
1339 static int set_qf_name(struct super_block *sb, int qtype, substring_t *args)
1341 struct ext4_sb_info *sbi = EXT4_SB(sb);
1344 if (sb_any_quota_loaded(sb) &&
1345 !sbi->s_qf_names[qtype]) {
1346 ext4_msg(sb, KERN_ERR,
1347 "Cannot change journaled "
1348 "quota options when quota turned on");
1351 qname = match_strdup(args);
1353 ext4_msg(sb, KERN_ERR,
1354 "Not enough memory for storing quotafile name");
1357 if (sbi->s_qf_names[qtype] &&
1358 strcmp(sbi->s_qf_names[qtype], qname)) {
1359 ext4_msg(sb, KERN_ERR,
1360 "%s quota file already specified", QTYPE2NAME(qtype));
1364 sbi->s_qf_names[qtype] = qname;
1365 if (strchr(sbi->s_qf_names[qtype], '/')) {
1366 ext4_msg(sb, KERN_ERR,
1367 "quotafile must be on filesystem root");
1368 kfree(sbi->s_qf_names[qtype]);
1369 sbi->s_qf_names[qtype] = NULL;
1376 static int clear_qf_name(struct super_block *sb, int qtype)
1379 struct ext4_sb_info *sbi = EXT4_SB(sb);
1381 if (sb_any_quota_loaded(sb) &&
1382 sbi->s_qf_names[qtype]) {
1383 ext4_msg(sb, KERN_ERR, "Cannot change journaled quota options"
1384 " when quota turned on");
1388 * The space will be released later when all options are confirmed
1391 sbi->s_qf_names[qtype] = NULL;
1396 #define MOPT_SET 0x0001
1397 #define MOPT_CLEAR 0x0002
1398 #define MOPT_NOSUPPORT 0x0004
1399 #define MOPT_EXPLICIT 0x0008
1400 #define MOPT_CLEAR_ERR 0x0010
1401 #define MOPT_GTE0 0x0020
1404 #define MOPT_QFMT 0x0040
1406 #define MOPT_Q MOPT_NOSUPPORT
1407 #define MOPT_QFMT MOPT_NOSUPPORT
1409 #define MOPT_DATAJ 0x0080
1411 static const struct mount_opts {
1415 } ext4_mount_opts[] = {
1416 {Opt_minix_df, EXT4_MOUNT_MINIX_DF, MOPT_SET},
1417 {Opt_bsd_df, EXT4_MOUNT_MINIX_DF, MOPT_CLEAR},
1418 {Opt_grpid, EXT4_MOUNT_GRPID, MOPT_SET},
1419 {Opt_nogrpid, EXT4_MOUNT_GRPID, MOPT_CLEAR},
1420 {Opt_mblk_io_submit, EXT4_MOUNT_MBLK_IO_SUBMIT, MOPT_SET},
1421 {Opt_nomblk_io_submit, EXT4_MOUNT_MBLK_IO_SUBMIT, MOPT_CLEAR},
1422 {Opt_block_validity, EXT4_MOUNT_BLOCK_VALIDITY, MOPT_SET},
1423 {Opt_noblock_validity, EXT4_MOUNT_BLOCK_VALIDITY, MOPT_CLEAR},
1424 {Opt_dioread_nolock, EXT4_MOUNT_DIOREAD_NOLOCK, MOPT_SET},
1425 {Opt_dioread_lock, EXT4_MOUNT_DIOREAD_NOLOCK, MOPT_CLEAR},
1426 {Opt_discard, EXT4_MOUNT_DISCARD, MOPT_SET},
1427 {Opt_nodiscard, EXT4_MOUNT_DISCARD, MOPT_CLEAR},
1428 {Opt_delalloc, EXT4_MOUNT_DELALLOC, MOPT_SET | MOPT_EXPLICIT},
1429 {Opt_nodelalloc, EXT4_MOUNT_DELALLOC, MOPT_CLEAR | MOPT_EXPLICIT},
1430 {Opt_journal_checksum, EXT4_MOUNT_JOURNAL_CHECKSUM, MOPT_SET},
1431 {Opt_journal_async_commit, (EXT4_MOUNT_JOURNAL_ASYNC_COMMIT |
1432 EXT4_MOUNT_JOURNAL_CHECKSUM), MOPT_SET},
1433 {Opt_noload, EXT4_MOUNT_NOLOAD, MOPT_SET},
1434 {Opt_err_panic, EXT4_MOUNT_ERRORS_PANIC, MOPT_SET | MOPT_CLEAR_ERR},
1435 {Opt_err_ro, EXT4_MOUNT_ERRORS_RO, MOPT_SET | MOPT_CLEAR_ERR},
1436 {Opt_err_cont, EXT4_MOUNT_ERRORS_CONT, MOPT_SET | MOPT_CLEAR_ERR},
1437 {Opt_data_err_abort, EXT4_MOUNT_DATA_ERR_ABORT, MOPT_SET},
1438 {Opt_data_err_ignore, EXT4_MOUNT_DATA_ERR_ABORT, MOPT_CLEAR},
1439 {Opt_barrier, EXT4_MOUNT_BARRIER, MOPT_SET},
1440 {Opt_nobarrier, EXT4_MOUNT_BARRIER, MOPT_CLEAR},
1441 {Opt_noauto_da_alloc, EXT4_MOUNT_NO_AUTO_DA_ALLOC, MOPT_SET},
1442 {Opt_auto_da_alloc, EXT4_MOUNT_NO_AUTO_DA_ALLOC, MOPT_CLEAR},
1443 {Opt_noinit_itable, EXT4_MOUNT_INIT_INODE_TABLE, MOPT_CLEAR},
1444 {Opt_commit, 0, MOPT_GTE0},
1445 {Opt_max_batch_time, 0, MOPT_GTE0},
1446 {Opt_min_batch_time, 0, MOPT_GTE0},
1447 {Opt_inode_readahead_blks, 0, MOPT_GTE0},
1448 {Opt_init_itable, 0, MOPT_GTE0},
1449 {Opt_stripe, 0, MOPT_GTE0},
1450 {Opt_data_journal, EXT4_MOUNT_JOURNAL_DATA, MOPT_DATAJ},
1451 {Opt_data_ordered, EXT4_MOUNT_ORDERED_DATA, MOPT_DATAJ},
1452 {Opt_data_writeback, EXT4_MOUNT_WRITEBACK_DATA, MOPT_DATAJ},
1453 {Opt_user_xattr, EXT4_MOUNT_XATTR_USER, MOPT_SET},
1454 {Opt_nouser_xattr, EXT4_MOUNT_XATTR_USER, MOPT_CLEAR},
1455 #ifdef CONFIG_EXT4_FS_POSIX_ACL
1456 {Opt_acl, EXT4_MOUNT_POSIX_ACL, MOPT_SET},
1457 {Opt_noacl, EXT4_MOUNT_POSIX_ACL, MOPT_CLEAR},
1459 {Opt_acl, 0, MOPT_NOSUPPORT},
1460 {Opt_noacl, 0, MOPT_NOSUPPORT},
1462 {Opt_nouid32, EXT4_MOUNT_NO_UID32, MOPT_SET},
1463 {Opt_debug, EXT4_MOUNT_DEBUG, MOPT_SET},
1464 {Opt_quota, EXT4_MOUNT_QUOTA | EXT4_MOUNT_USRQUOTA, MOPT_SET | MOPT_Q},
1465 {Opt_usrquota, EXT4_MOUNT_QUOTA | EXT4_MOUNT_USRQUOTA,
1467 {Opt_grpquota, EXT4_MOUNT_QUOTA | EXT4_MOUNT_GRPQUOTA,
1469 {Opt_noquota, (EXT4_MOUNT_QUOTA | EXT4_MOUNT_USRQUOTA |
1470 EXT4_MOUNT_GRPQUOTA), MOPT_CLEAR | MOPT_Q},
1471 {Opt_usrjquota, 0, MOPT_Q},
1472 {Opt_grpjquota, 0, MOPT_Q},
1473 {Opt_offusrjquota, 0, MOPT_Q},
1474 {Opt_offgrpjquota, 0, MOPT_Q},
1475 {Opt_jqfmt_vfsold, QFMT_VFS_OLD, MOPT_QFMT},
1476 {Opt_jqfmt_vfsv0, QFMT_VFS_V0, MOPT_QFMT},
1477 {Opt_jqfmt_vfsv1, QFMT_VFS_V1, MOPT_QFMT},
1478 {Opt_max_dir_size_kb, 0, MOPT_GTE0},
1482 static int handle_mount_opt(struct super_block *sb, char *opt, int token,
1483 substring_t *args, unsigned long *journal_devnum,
1484 unsigned int *journal_ioprio, int is_remount)
1486 struct ext4_sb_info *sbi = EXT4_SB(sb);
1487 const struct mount_opts *m;
1493 if (token == Opt_usrjquota)
1494 return set_qf_name(sb, USRQUOTA, &args[0]);
1495 else if (token == Opt_grpjquota)
1496 return set_qf_name(sb, GRPQUOTA, &args[0]);
1497 else if (token == Opt_offusrjquota)
1498 return clear_qf_name(sb, USRQUOTA);
1499 else if (token == Opt_offgrpjquota)
1500 return clear_qf_name(sb, GRPQUOTA);
1502 if (args->from && match_int(args, &arg))
1506 case Opt_nouser_xattr:
1507 ext4_msg(sb, KERN_WARNING, deprecated_msg, opt, "3.5");
1510 return 1; /* handled by get_sb_block() */
1512 ext4_msg(sb, KERN_WARNING,
1513 "Ignoring removed %s option", opt);
1516 uid = make_kuid(current_user_ns(), arg);
1517 if (!uid_valid(uid)) {
1518 ext4_msg(sb, KERN_ERR, "Invalid uid value %d", arg);
1521 sbi->s_resuid = uid;
1524 gid = make_kgid(current_user_ns(), arg);
1525 if (!gid_valid(gid)) {
1526 ext4_msg(sb, KERN_ERR, "Invalid gid value %d", arg);
1529 sbi->s_resgid = gid;
1532 sbi->s_mount_flags |= EXT4_MF_FS_ABORTED;
1535 sb->s_flags |= MS_I_VERSION;
1537 case Opt_journal_dev:
1539 ext4_msg(sb, KERN_ERR,
1540 "Cannot specify journal on remount");
1543 *journal_devnum = arg;
1545 case Opt_journal_ioprio:
1546 if (arg < 0 || arg > 7)
1548 *journal_ioprio = IOPRIO_PRIO_VALUE(IOPRIO_CLASS_BE, arg);
1552 for (m = ext4_mount_opts; m->token != Opt_err; m++) {
1553 if (token != m->token)
1555 if (args->from && (m->flags & MOPT_GTE0) && (arg < 0))
1557 if (m->flags & MOPT_EXPLICIT)
1558 set_opt2(sb, EXPLICIT_DELALLOC);
1559 if (m->flags & MOPT_CLEAR_ERR)
1560 clear_opt(sb, ERRORS_MASK);
1561 if (token == Opt_noquota && sb_any_quota_loaded(sb)) {
1562 ext4_msg(sb, KERN_ERR, "Cannot change quota "
1563 "options when quota turned on");
1567 if (m->flags & MOPT_NOSUPPORT) {
1568 ext4_msg(sb, KERN_ERR, "%s option not supported", opt);
1569 } else if (token == Opt_commit) {
1571 arg = JBD2_DEFAULT_MAX_COMMIT_AGE;
1572 sbi->s_commit_interval = HZ * arg;
1573 } else if (token == Opt_max_batch_time) {
1575 arg = EXT4_DEF_MAX_BATCH_TIME;
1576 sbi->s_max_batch_time = arg;
1577 } else if (token == Opt_min_batch_time) {
1578 sbi->s_min_batch_time = arg;
1579 } else if (token == Opt_inode_readahead_blks) {
1580 if (arg > (1 << 30))
1582 if (arg && !is_power_of_2(arg)) {
1583 ext4_msg(sb, KERN_ERR,
1584 "EXT4-fs: inode_readahead_blks"
1585 " must be a power of 2");
1588 sbi->s_inode_readahead_blks = arg;
1589 } else if (token == Opt_init_itable) {
1590 set_opt(sb, INIT_INODE_TABLE);
1592 arg = EXT4_DEF_LI_WAIT_MULT;
1593 sbi->s_li_wait_mult = arg;
1594 } else if (token == Opt_max_dir_size_kb) {
1595 sbi->s_max_dir_size_kb = arg;
1596 } else if (token == Opt_stripe) {
1597 sbi->s_stripe = arg;
1598 } else if (m->flags & MOPT_DATAJ) {
1600 if (!sbi->s_journal)
1601 ext4_msg(sb, KERN_WARNING, "Remounting file system with no journal so ignoring journalled data option");
1602 else if (test_opt(sb, DATA_FLAGS) !=
1604 ext4_msg(sb, KERN_ERR,
1605 "Cannot change data mode on remount");
1609 clear_opt(sb, DATA_FLAGS);
1610 sbi->s_mount_opt |= m->mount_opt;
1613 } else if (m->flags & MOPT_QFMT) {
1614 if (sb_any_quota_loaded(sb) &&
1615 sbi->s_jquota_fmt != m->mount_opt) {
1616 ext4_msg(sb, KERN_ERR, "Cannot "
1617 "change journaled quota options "
1618 "when quota turned on");
1621 sbi->s_jquota_fmt = m->mount_opt;
1626 if (m->flags & MOPT_CLEAR)
1628 else if (unlikely(!(m->flags & MOPT_SET))) {
1629 ext4_msg(sb, KERN_WARNING,
1630 "buggy handling of option %s", opt);
1635 sbi->s_mount_opt |= m->mount_opt;
1637 sbi->s_mount_opt &= ~m->mount_opt;
1641 ext4_msg(sb, KERN_ERR, "Unrecognized mount option \"%s\" "
1642 "or missing value", opt);
1646 static int parse_options(char *options, struct super_block *sb,
1647 unsigned long *journal_devnum,
1648 unsigned int *journal_ioprio,
1651 struct ext4_sb_info *sbi = EXT4_SB(sb);
1653 substring_t args[MAX_OPT_ARGS];
1659 while ((p = strsep(&options, ",")) != NULL) {
1663 * Initialize args struct so we know whether arg was
1664 * found; some options take optional arguments.
1666 args[0].to = args[0].from = NULL;
1667 token = match_token(p, tokens, args);
1668 if (handle_mount_opt(sb, p, token, args, journal_devnum,
1669 journal_ioprio, is_remount) < 0)
1673 if (sbi->s_qf_names[USRQUOTA] || sbi->s_qf_names[GRPQUOTA]) {
1674 if (test_opt(sb, USRQUOTA) && sbi->s_qf_names[USRQUOTA])
1675 clear_opt(sb, USRQUOTA);
1677 if (test_opt(sb, GRPQUOTA) && sbi->s_qf_names[GRPQUOTA])
1678 clear_opt(sb, GRPQUOTA);
1680 if (test_opt(sb, GRPQUOTA) || test_opt(sb, USRQUOTA)) {
1681 ext4_msg(sb, KERN_ERR, "old and new quota "
1686 if (!sbi->s_jquota_fmt) {
1687 ext4_msg(sb, KERN_ERR, "journaled quota format "
1692 if (sbi->s_jquota_fmt) {
1693 ext4_msg(sb, KERN_ERR, "journaled quota format "
1694 "specified with no journaling "
1700 if (test_opt(sb, DIOREAD_NOLOCK)) {
1702 BLOCK_SIZE << le32_to_cpu(sbi->s_es->s_log_block_size);
1704 if (blocksize < PAGE_CACHE_SIZE) {
1705 ext4_msg(sb, KERN_ERR, "can't mount with "
1706 "dioread_nolock if block size != PAGE_SIZE");
1713 static inline void ext4_show_quota_options(struct seq_file *seq,
1714 struct super_block *sb)
1716 #if defined(CONFIG_QUOTA)
1717 struct ext4_sb_info *sbi = EXT4_SB(sb);
1719 if (sbi->s_jquota_fmt) {
1722 switch (sbi->s_jquota_fmt) {
1733 seq_printf(seq, ",jqfmt=%s", fmtname);
1736 if (sbi->s_qf_names[USRQUOTA])
1737 seq_printf(seq, ",usrjquota=%s", sbi->s_qf_names[USRQUOTA]);
1739 if (sbi->s_qf_names[GRPQUOTA])
1740 seq_printf(seq, ",grpjquota=%s", sbi->s_qf_names[GRPQUOTA]);
1742 if (test_opt(sb, USRQUOTA))
1743 seq_puts(seq, ",usrquota");
1745 if (test_opt(sb, GRPQUOTA))
1746 seq_puts(seq, ",grpquota");
1750 static const char *token2str(int token)
1752 const struct match_token *t;
1754 for (t = tokens; t->token != Opt_err; t++)
1755 if (t->token == token && !strchr(t->pattern, '='))
1762 * - it's set to a non-default value OR
1763 * - if the per-sb default is different from the global default
1765 static int _ext4_show_options(struct seq_file *seq, struct super_block *sb,
1768 struct ext4_sb_info *sbi = EXT4_SB(sb);
1769 struct ext4_super_block *es = sbi->s_es;
1770 int def_errors, def_mount_opt = nodefs ? 0 : sbi->s_def_mount_opt;
1771 const struct mount_opts *m;
1772 char sep = nodefs ? '\n' : ',';
1774 #define SEQ_OPTS_PUTS(str) seq_printf(seq, "%c" str, sep)
1775 #define SEQ_OPTS_PRINT(str, arg) seq_printf(seq, "%c" str, sep, arg)
1777 if (sbi->s_sb_block != 1)
1778 SEQ_OPTS_PRINT("sb=%llu", sbi->s_sb_block);
1780 for (m = ext4_mount_opts; m->token != Opt_err; m++) {
1781 int want_set = m->flags & MOPT_SET;
1782 if (((m->flags & (MOPT_SET|MOPT_CLEAR)) == 0) ||
1783 (m->flags & MOPT_CLEAR_ERR))
1785 if (!(m->mount_opt & (sbi->s_mount_opt ^ def_mount_opt)))
1786 continue; /* skip if same as the default */
1788 (sbi->s_mount_opt & m->mount_opt) != m->mount_opt) ||
1789 (!want_set && (sbi->s_mount_opt & m->mount_opt)))
1790 continue; /* select Opt_noFoo vs Opt_Foo */
1791 SEQ_OPTS_PRINT("%s", token2str(m->token));
1794 if (nodefs || !uid_eq(sbi->s_resuid, make_kuid(&init_user_ns, EXT4_DEF_RESUID)) ||
1795 le16_to_cpu(es->s_def_resuid) != EXT4_DEF_RESUID)
1796 SEQ_OPTS_PRINT("resuid=%u",
1797 from_kuid_munged(&init_user_ns, sbi->s_resuid));
1798 if (nodefs || !gid_eq(sbi->s_resgid, make_kgid(&init_user_ns, EXT4_DEF_RESGID)) ||
1799 le16_to_cpu(es->s_def_resgid) != EXT4_DEF_RESGID)
1800 SEQ_OPTS_PRINT("resgid=%u",
1801 from_kgid_munged(&init_user_ns, sbi->s_resgid));
1802 def_errors = nodefs ? -1 : le16_to_cpu(es->s_errors);
1803 if (test_opt(sb, ERRORS_RO) && def_errors != EXT4_ERRORS_RO)
1804 SEQ_OPTS_PUTS("errors=remount-ro");
1805 if (test_opt(sb, ERRORS_CONT) && def_errors != EXT4_ERRORS_CONTINUE)
1806 SEQ_OPTS_PUTS("errors=continue");
1807 if (test_opt(sb, ERRORS_PANIC) && def_errors != EXT4_ERRORS_PANIC)
1808 SEQ_OPTS_PUTS("errors=panic");
1809 if (nodefs || sbi->s_commit_interval != JBD2_DEFAULT_MAX_COMMIT_AGE*HZ)
1810 SEQ_OPTS_PRINT("commit=%lu", sbi->s_commit_interval / HZ);
1811 if (nodefs || sbi->s_min_batch_time != EXT4_DEF_MIN_BATCH_TIME)
1812 SEQ_OPTS_PRINT("min_batch_time=%u", sbi->s_min_batch_time);
1813 if (nodefs || sbi->s_max_batch_time != EXT4_DEF_MAX_BATCH_TIME)
1814 SEQ_OPTS_PRINT("max_batch_time=%u", sbi->s_max_batch_time);
1815 if (sb->s_flags & MS_I_VERSION)
1816 SEQ_OPTS_PUTS("i_version");
1817 if (nodefs || sbi->s_stripe)
1818 SEQ_OPTS_PRINT("stripe=%lu", sbi->s_stripe);
1819 if (EXT4_MOUNT_DATA_FLAGS & (sbi->s_mount_opt ^ def_mount_opt)) {
1820 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA)
1821 SEQ_OPTS_PUTS("data=journal");
1822 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA)
1823 SEQ_OPTS_PUTS("data=ordered");
1824 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_WRITEBACK_DATA)
1825 SEQ_OPTS_PUTS("data=writeback");
1828 sbi->s_inode_readahead_blks != EXT4_DEF_INODE_READAHEAD_BLKS)
1829 SEQ_OPTS_PRINT("inode_readahead_blks=%u",
1830 sbi->s_inode_readahead_blks);
1832 if (nodefs || (test_opt(sb, INIT_INODE_TABLE) &&
1833 (sbi->s_li_wait_mult != EXT4_DEF_LI_WAIT_MULT)))
1834 SEQ_OPTS_PRINT("init_itable=%u", sbi->s_li_wait_mult);
1835 if (nodefs || sbi->s_max_dir_size_kb)
1836 SEQ_OPTS_PRINT("max_dir_size_kb=%u", sbi->s_max_dir_size_kb);
1838 ext4_show_quota_options(seq, sb);
1842 static int ext4_show_options(struct seq_file *seq, struct dentry *root)
1844 return _ext4_show_options(seq, root->d_sb, 0);
1847 static int options_seq_show(struct seq_file *seq, void *offset)
1849 struct super_block *sb = seq->private;
1852 seq_puts(seq, (sb->s_flags & MS_RDONLY) ? "ro" : "rw");
1853 rc = _ext4_show_options(seq, sb, 1);
1854 seq_puts(seq, "\n");
1858 static int options_open_fs(struct inode *inode, struct file *file)
1860 return single_open(file, options_seq_show, PDE(inode)->data);
1863 static const struct file_operations ext4_seq_options_fops = {
1864 .owner = THIS_MODULE,
1865 .open = options_open_fs,
1867 .llseek = seq_lseek,
1868 .release = single_release,
1871 static int ext4_setup_super(struct super_block *sb, struct ext4_super_block *es,
1874 struct ext4_sb_info *sbi = EXT4_SB(sb);
1877 if (le32_to_cpu(es->s_rev_level) > EXT4_MAX_SUPP_REV) {
1878 ext4_msg(sb, KERN_ERR, "revision level too high, "
1879 "forcing read-only mode");
1884 if (!(sbi->s_mount_state & EXT4_VALID_FS))
1885 ext4_msg(sb, KERN_WARNING, "warning: mounting unchecked fs, "
1886 "running e2fsck is recommended");
1887 else if ((sbi->s_mount_state & EXT4_ERROR_FS))
1888 ext4_msg(sb, KERN_WARNING,
1889 "warning: mounting fs with errors, "
1890 "running e2fsck is recommended");
1891 else if ((__s16) le16_to_cpu(es->s_max_mnt_count) > 0 &&
1892 le16_to_cpu(es->s_mnt_count) >=
1893 (unsigned short) (__s16) le16_to_cpu(es->s_max_mnt_count))
1894 ext4_msg(sb, KERN_WARNING,
1895 "warning: maximal mount count reached, "
1896 "running e2fsck is recommended");
1897 else if (le32_to_cpu(es->s_checkinterval) &&
1898 (le32_to_cpu(es->s_lastcheck) +
1899 le32_to_cpu(es->s_checkinterval) <= get_seconds()))
1900 ext4_msg(sb, KERN_WARNING,
1901 "warning: checktime reached, "
1902 "running e2fsck is recommended");
1903 if (!sbi->s_journal)
1904 es->s_state &= cpu_to_le16(~EXT4_VALID_FS);
1905 if (!(__s16) le16_to_cpu(es->s_max_mnt_count))
1906 es->s_max_mnt_count = cpu_to_le16(EXT4_DFL_MAX_MNT_COUNT);
1907 le16_add_cpu(&es->s_mnt_count, 1);
1908 es->s_mtime = cpu_to_le32(get_seconds());
1909 ext4_update_dynamic_rev(sb);
1911 EXT4_SET_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
1913 ext4_commit_super(sb, 1);
1915 if (test_opt(sb, DEBUG))
1916 printk(KERN_INFO "[EXT4 FS bs=%lu, gc=%u, "
1917 "bpg=%lu, ipg=%lu, mo=%04x, mo2=%04x]\n",
1919 sbi->s_groups_count,
1920 EXT4_BLOCKS_PER_GROUP(sb),
1921 EXT4_INODES_PER_GROUP(sb),
1922 sbi->s_mount_opt, sbi->s_mount_opt2);
1924 cleancache_init_fs(sb);
1928 int ext4_alloc_flex_bg_array(struct super_block *sb, ext4_group_t ngroup)
1930 struct ext4_sb_info *sbi = EXT4_SB(sb);
1931 struct flex_groups *new_groups;
1934 if (!sbi->s_log_groups_per_flex)
1937 size = ext4_flex_group(sbi, ngroup - 1) + 1;
1938 if (size <= sbi->s_flex_groups_allocated)
1941 size = roundup_pow_of_two(size * sizeof(struct flex_groups));
1942 new_groups = ext4_kvzalloc(size, GFP_KERNEL);
1944 ext4_msg(sb, KERN_ERR, "not enough memory for %d flex groups",
1945 size / (int) sizeof(struct flex_groups));
1949 if (sbi->s_flex_groups) {
1950 memcpy(new_groups, sbi->s_flex_groups,
1951 (sbi->s_flex_groups_allocated *
1952 sizeof(struct flex_groups)));
1953 ext4_kvfree(sbi->s_flex_groups);
1955 sbi->s_flex_groups = new_groups;
1956 sbi->s_flex_groups_allocated = size / sizeof(struct flex_groups);
1960 static int ext4_fill_flex_info(struct super_block *sb)
1962 struct ext4_sb_info *sbi = EXT4_SB(sb);
1963 struct ext4_group_desc *gdp = NULL;
1964 ext4_group_t flex_group;
1965 unsigned int groups_per_flex = 0;
1968 sbi->s_log_groups_per_flex = sbi->s_es->s_log_groups_per_flex;
1969 if (sbi->s_log_groups_per_flex < 1 || sbi->s_log_groups_per_flex > 31) {
1970 sbi->s_log_groups_per_flex = 0;
1973 groups_per_flex = 1U << sbi->s_log_groups_per_flex;
1975 err = ext4_alloc_flex_bg_array(sb, sbi->s_groups_count);
1979 for (i = 0; i < sbi->s_groups_count; i++) {
1980 gdp = ext4_get_group_desc(sb, i, NULL);
1982 flex_group = ext4_flex_group(sbi, i);
1983 atomic_add(ext4_free_inodes_count(sb, gdp),
1984 &sbi->s_flex_groups[flex_group].free_inodes);
1985 atomic64_add(ext4_free_group_clusters(sb, gdp),
1986 &sbi->s_flex_groups[flex_group].free_clusters);
1987 atomic_add(ext4_used_dirs_count(sb, gdp),
1988 &sbi->s_flex_groups[flex_group].used_dirs);
1996 static __le16 ext4_group_desc_csum(struct ext4_sb_info *sbi, __u32 block_group,
1997 struct ext4_group_desc *gdp)
2001 __le32 le_group = cpu_to_le32(block_group);
2003 if ((sbi->s_es->s_feature_ro_compat &
2004 cpu_to_le32(EXT4_FEATURE_RO_COMPAT_METADATA_CSUM))) {
2005 /* Use new metadata_csum algorithm */
2009 old_csum = gdp->bg_checksum;
2010 gdp->bg_checksum = 0;
2011 csum32 = ext4_chksum(sbi, sbi->s_csum_seed, (__u8 *)&le_group,
2013 csum32 = ext4_chksum(sbi, csum32, (__u8 *)gdp,
2015 gdp->bg_checksum = old_csum;
2017 crc = csum32 & 0xFFFF;
2021 /* old crc16 code */
2022 offset = offsetof(struct ext4_group_desc, bg_checksum);
2024 crc = crc16(~0, sbi->s_es->s_uuid, sizeof(sbi->s_es->s_uuid));
2025 crc = crc16(crc, (__u8 *)&le_group, sizeof(le_group));
2026 crc = crc16(crc, (__u8 *)gdp, offset);
2027 offset += sizeof(gdp->bg_checksum); /* skip checksum */
2028 /* for checksum of struct ext4_group_desc do the rest...*/
2029 if ((sbi->s_es->s_feature_incompat &
2030 cpu_to_le32(EXT4_FEATURE_INCOMPAT_64BIT)) &&
2031 offset < le16_to_cpu(sbi->s_es->s_desc_size))
2032 crc = crc16(crc, (__u8 *)gdp + offset,
2033 le16_to_cpu(sbi->s_es->s_desc_size) -
2037 return cpu_to_le16(crc);
2040 int ext4_group_desc_csum_verify(struct super_block *sb, __u32 block_group,
2041 struct ext4_group_desc *gdp)
2043 if (ext4_has_group_desc_csum(sb) &&
2044 (gdp->bg_checksum != ext4_group_desc_csum(EXT4_SB(sb),
2051 void ext4_group_desc_csum_set(struct super_block *sb, __u32 block_group,
2052 struct ext4_group_desc *gdp)
2054 if (!ext4_has_group_desc_csum(sb))
2056 gdp->bg_checksum = ext4_group_desc_csum(EXT4_SB(sb), block_group, gdp);
2059 /* Called at mount-time, super-block is locked */
2060 static int ext4_check_descriptors(struct super_block *sb,
2061 ext4_group_t *first_not_zeroed)
2063 struct ext4_sb_info *sbi = EXT4_SB(sb);
2064 ext4_fsblk_t first_block = le32_to_cpu(sbi->s_es->s_first_data_block);
2065 ext4_fsblk_t last_block;
2066 ext4_fsblk_t block_bitmap;
2067 ext4_fsblk_t inode_bitmap;
2068 ext4_fsblk_t inode_table;
2069 int flexbg_flag = 0;
2070 ext4_group_t i, grp = sbi->s_groups_count;
2072 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_FLEX_BG))
2075 ext4_debug("Checking group descriptors");
2077 for (i = 0; i < sbi->s_groups_count; i++) {
2078 struct ext4_group_desc *gdp = ext4_get_group_desc(sb, i, NULL);
2080 if (i == sbi->s_groups_count - 1 || flexbg_flag)
2081 last_block = ext4_blocks_count(sbi->s_es) - 1;
2083 last_block = first_block +
2084 (EXT4_BLOCKS_PER_GROUP(sb) - 1);
2086 if ((grp == sbi->s_groups_count) &&
2087 !(gdp->bg_flags & cpu_to_le16(EXT4_BG_INODE_ZEROED)))
2090 block_bitmap = ext4_block_bitmap(sb, gdp);
2091 if (block_bitmap < first_block || block_bitmap > last_block) {
2092 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2093 "Block bitmap for group %u not in group "
2094 "(block %llu)!", i, block_bitmap);
2097 inode_bitmap = ext4_inode_bitmap(sb, gdp);
2098 if (inode_bitmap < first_block || inode_bitmap > last_block) {
2099 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2100 "Inode bitmap for group %u not in group "
2101 "(block %llu)!", i, inode_bitmap);
2104 inode_table = ext4_inode_table(sb, gdp);
2105 if (inode_table < first_block ||
2106 inode_table + sbi->s_itb_per_group - 1 > last_block) {
2107 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2108 "Inode table for group %u not in group "
2109 "(block %llu)!", i, inode_table);
2112 ext4_lock_group(sb, i);
2113 if (!ext4_group_desc_csum_verify(sb, i, gdp)) {
2114 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2115 "Checksum for group %u failed (%u!=%u)",
2116 i, le16_to_cpu(ext4_group_desc_csum(sbi, i,
2117 gdp)), le16_to_cpu(gdp->bg_checksum));
2118 if (!(sb->s_flags & MS_RDONLY)) {
2119 ext4_unlock_group(sb, i);
2123 ext4_unlock_group(sb, i);
2125 first_block += EXT4_BLOCKS_PER_GROUP(sb);
2127 if (NULL != first_not_zeroed)
2128 *first_not_zeroed = grp;
2130 ext4_free_blocks_count_set(sbi->s_es,
2131 EXT4_C2B(sbi, ext4_count_free_clusters(sb)));
2132 sbi->s_es->s_free_inodes_count =cpu_to_le32(ext4_count_free_inodes(sb));
2136 /* ext4_orphan_cleanup() walks a singly-linked list of inodes (starting at
2137 * the superblock) which were deleted from all directories, but held open by
2138 * a process at the time of a crash. We walk the list and try to delete these
2139 * inodes at recovery time (only with a read-write filesystem).
2141 * In order to keep the orphan inode chain consistent during traversal (in
2142 * case of crash during recovery), we link each inode into the superblock
2143 * orphan list_head and handle it the same way as an inode deletion during
2144 * normal operation (which journals the operations for us).
2146 * We only do an iget() and an iput() on each inode, which is very safe if we
2147 * accidentally point at an in-use or already deleted inode. The worst that
2148 * can happen in this case is that we get a "bit already cleared" message from
2149 * ext4_free_inode(). The only reason we would point at a wrong inode is if
2150 * e2fsck was run on this filesystem, and it must have already done the orphan
2151 * inode cleanup for us, so we can safely abort without any further action.
2153 static void ext4_orphan_cleanup(struct super_block *sb,
2154 struct ext4_super_block *es)
2156 unsigned int s_flags = sb->s_flags;
2157 int nr_orphans = 0, nr_truncates = 0;
2161 if (!es->s_last_orphan) {
2162 jbd_debug(4, "no orphan inodes to clean up\n");
2166 if (bdev_read_only(sb->s_bdev)) {
2167 ext4_msg(sb, KERN_ERR, "write access "
2168 "unavailable, skipping orphan cleanup");
2172 /* Check if feature set would not allow a r/w mount */
2173 if (!ext4_feature_set_ok(sb, 0)) {
2174 ext4_msg(sb, KERN_INFO, "Skipping orphan cleanup due to "
2175 "unknown ROCOMPAT features");
2179 if (EXT4_SB(sb)->s_mount_state & EXT4_ERROR_FS) {
2180 /* don't clear list on RO mount w/ errors */
2181 if (es->s_last_orphan && !(s_flags & MS_RDONLY)) {
2182 jbd_debug(1, "Errors on filesystem, "
2183 "clearing orphan list.\n");
2184 es->s_last_orphan = 0;
2186 jbd_debug(1, "Skipping orphan recovery on fs with errors.\n");
2190 if (s_flags & MS_RDONLY) {
2191 ext4_msg(sb, KERN_INFO, "orphan cleanup on readonly fs");
2192 sb->s_flags &= ~MS_RDONLY;
2195 /* Needed for iput() to work correctly and not trash data */
2196 sb->s_flags |= MS_ACTIVE;
2197 /* Turn on quotas so that they are updated correctly */
2198 for (i = 0; i < MAXQUOTAS; i++) {
2199 if (EXT4_SB(sb)->s_qf_names[i]) {
2200 int ret = ext4_quota_on_mount(sb, i);
2202 ext4_msg(sb, KERN_ERR,
2203 "Cannot turn on journaled "
2204 "quota: error %d", ret);
2209 while (es->s_last_orphan) {
2210 struct inode *inode;
2212 inode = ext4_orphan_get(sb, le32_to_cpu(es->s_last_orphan));
2213 if (IS_ERR(inode)) {
2214 es->s_last_orphan = 0;
2218 list_add(&EXT4_I(inode)->i_orphan, &EXT4_SB(sb)->s_orphan);
2219 dquot_initialize(inode);
2220 if (inode->i_nlink) {
2221 ext4_msg(sb, KERN_DEBUG,
2222 "%s: truncating inode %lu to %lld bytes",
2223 __func__, inode->i_ino, inode->i_size);
2224 jbd_debug(2, "truncating inode %lu to %lld bytes\n",
2225 inode->i_ino, inode->i_size);
2226 mutex_lock(&inode->i_mutex);
2227 ext4_truncate(inode);
2228 mutex_unlock(&inode->i_mutex);
2231 ext4_msg(sb, KERN_DEBUG,
2232 "%s: deleting unreferenced inode %lu",
2233 __func__, inode->i_ino);
2234 jbd_debug(2, "deleting unreferenced inode %lu\n",
2238 iput(inode); /* The delete magic happens here! */
2241 #define PLURAL(x) (x), ((x) == 1) ? "" : "s"
2244 ext4_msg(sb, KERN_INFO, "%d orphan inode%s deleted",
2245 PLURAL(nr_orphans));
2247 ext4_msg(sb, KERN_INFO, "%d truncate%s cleaned up",
2248 PLURAL(nr_truncates));
2250 /* Turn quotas off */
2251 for (i = 0; i < MAXQUOTAS; i++) {
2252 if (sb_dqopt(sb)->files[i])
2253 dquot_quota_off(sb, i);
2256 sb->s_flags = s_flags; /* Restore MS_RDONLY status */
2260 * Maximal extent format file size.
2261 * Resulting logical blkno at s_maxbytes must fit in our on-disk
2262 * extent format containers, within a sector_t, and within i_blocks
2263 * in the vfs. ext4 inode has 48 bits of i_block in fsblock units,
2264 * so that won't be a limiting factor.
2266 * However there is other limiting factor. We do store extents in the form
2267 * of starting block and length, hence the resulting length of the extent
2268 * covering maximum file size must fit into on-disk format containers as
2269 * well. Given that length is always by 1 unit bigger than max unit (because
2270 * we count 0 as well) we have to lower the s_maxbytes by one fs block.
2272 * Note, this does *not* consider any metadata overhead for vfs i_blocks.
2274 static loff_t ext4_max_size(int blkbits, int has_huge_files)
2277 loff_t upper_limit = MAX_LFS_FILESIZE;
2279 /* small i_blocks in vfs inode? */
2280 if (!has_huge_files || sizeof(blkcnt_t) < sizeof(u64)) {
2282 * CONFIG_LBDAF is not enabled implies the inode
2283 * i_block represent total blocks in 512 bytes
2284 * 32 == size of vfs inode i_blocks * 8
2286 upper_limit = (1LL << 32) - 1;
2288 /* total blocks in file system block size */
2289 upper_limit >>= (blkbits - 9);
2290 upper_limit <<= blkbits;
2294 * 32-bit extent-start container, ee_block. We lower the maxbytes
2295 * by one fs block, so ee_len can cover the extent of maximum file
2298 res = (1LL << 32) - 1;
2301 /* Sanity check against vm- & vfs- imposed limits */
2302 if (res > upper_limit)
2309 * Maximal bitmap file size. There is a direct, and {,double-,triple-}indirect
2310 * block limit, and also a limit of (2^48 - 1) 512-byte sectors in i_blocks.
2311 * We need to be 1 filesystem block less than the 2^48 sector limit.
2313 static loff_t ext4_max_bitmap_size(int bits, int has_huge_files)
2315 loff_t res = EXT4_NDIR_BLOCKS;
2318 /* This is calculated to be the largest file size for a dense, block
2319 * mapped file such that the file's total number of 512-byte sectors,
2320 * including data and all indirect blocks, does not exceed (2^48 - 1).
2322 * __u32 i_blocks_lo and _u16 i_blocks_high represent the total
2323 * number of 512-byte sectors of the file.
2326 if (!has_huge_files || sizeof(blkcnt_t) < sizeof(u64)) {
2328 * !has_huge_files or CONFIG_LBDAF not enabled implies that
2329 * the inode i_block field represents total file blocks in
2330 * 2^32 512-byte sectors == size of vfs inode i_blocks * 8
2332 upper_limit = (1LL << 32) - 1;
2334 /* total blocks in file system block size */
2335 upper_limit >>= (bits - 9);
2339 * We use 48 bit ext4_inode i_blocks
2340 * With EXT4_HUGE_FILE_FL set the i_blocks
2341 * represent total number of blocks in
2342 * file system block size
2344 upper_limit = (1LL << 48) - 1;
2348 /* indirect blocks */
2350 /* double indirect blocks */
2351 meta_blocks += 1 + (1LL << (bits-2));
2352 /* tripple indirect blocks */
2353 meta_blocks += 1 + (1LL << (bits-2)) + (1LL << (2*(bits-2)));
2355 upper_limit -= meta_blocks;
2356 upper_limit <<= bits;
2358 res += 1LL << (bits-2);
2359 res += 1LL << (2*(bits-2));
2360 res += 1LL << (3*(bits-2));
2362 if (res > upper_limit)
2365 if (res > MAX_LFS_FILESIZE)
2366 res = MAX_LFS_FILESIZE;
2371 static ext4_fsblk_t descriptor_loc(struct super_block *sb,
2372 ext4_fsblk_t logical_sb_block, int nr)
2374 struct ext4_sb_info *sbi = EXT4_SB(sb);
2375 ext4_group_t bg, first_meta_bg;
2378 first_meta_bg = le32_to_cpu(sbi->s_es->s_first_meta_bg);
2380 if (!EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_META_BG) ||
2382 return logical_sb_block + nr + 1;
2383 bg = sbi->s_desc_per_block * nr;
2384 if (ext4_bg_has_super(sb, bg))
2387 return (has_super + ext4_group_first_block_no(sb, bg));
2391 * ext4_get_stripe_size: Get the stripe size.
2392 * @sbi: In memory super block info
2394 * If we have specified it via mount option, then
2395 * use the mount option value. If the value specified at mount time is
2396 * greater than the blocks per group use the super block value.
2397 * If the super block value is greater than blocks per group return 0.
2398 * Allocator needs it be less than blocks per group.
2401 static unsigned long ext4_get_stripe_size(struct ext4_sb_info *sbi)
2403 unsigned long stride = le16_to_cpu(sbi->s_es->s_raid_stride);
2404 unsigned long stripe_width =
2405 le32_to_cpu(sbi->s_es->s_raid_stripe_width);
2408 if (sbi->s_stripe && sbi->s_stripe <= sbi->s_blocks_per_group)
2409 ret = sbi->s_stripe;
2410 else if (stripe_width <= sbi->s_blocks_per_group)
2412 else if (stride <= sbi->s_blocks_per_group)
2418 * If the stripe width is 1, this makes no sense and
2419 * we set it to 0 to turn off stripe handling code.
2430 struct attribute attr;
2431 ssize_t (*show)(struct ext4_attr *, struct ext4_sb_info *, char *);
2432 ssize_t (*store)(struct ext4_attr *, struct ext4_sb_info *,
2433 const char *, size_t);
2437 static int parse_strtoul(const char *buf,
2438 unsigned long max, unsigned long *value)
2442 *value = simple_strtoul(skip_spaces(buf), &endp, 0);
2443 endp = skip_spaces(endp);
2444 if (*endp || *value > max)
2450 static ssize_t delayed_allocation_blocks_show(struct ext4_attr *a,
2451 struct ext4_sb_info *sbi,
2454 return snprintf(buf, PAGE_SIZE, "%llu\n",
2456 percpu_counter_sum(&sbi->s_dirtyclusters_counter)));
2459 static ssize_t session_write_kbytes_show(struct ext4_attr *a,
2460 struct ext4_sb_info *sbi, char *buf)
2462 struct super_block *sb = sbi->s_buddy_cache->i_sb;
2464 if (!sb->s_bdev->bd_part)
2465 return snprintf(buf, PAGE_SIZE, "0\n");
2466 return snprintf(buf, PAGE_SIZE, "%lu\n",
2467 (part_stat_read(sb->s_bdev->bd_part, sectors[1]) -
2468 sbi->s_sectors_written_start) >> 1);
2471 static ssize_t lifetime_write_kbytes_show(struct ext4_attr *a,
2472 struct ext4_sb_info *sbi, char *buf)
2474 struct super_block *sb = sbi->s_buddy_cache->i_sb;
2476 if (!sb->s_bdev->bd_part)
2477 return snprintf(buf, PAGE_SIZE, "0\n");
2478 return snprintf(buf, PAGE_SIZE, "%llu\n",
2479 (unsigned long long)(sbi->s_kbytes_written +
2480 ((part_stat_read(sb->s_bdev->bd_part, sectors[1]) -
2481 EXT4_SB(sb)->s_sectors_written_start) >> 1)));
2484 static ssize_t inode_readahead_blks_store(struct ext4_attr *a,
2485 struct ext4_sb_info *sbi,
2486 const char *buf, size_t count)
2490 if (parse_strtoul(buf, 0x40000000, &t))
2493 if (t && !is_power_of_2(t))
2496 sbi->s_inode_readahead_blks = t;
2500 static ssize_t sbi_ui_show(struct ext4_attr *a,
2501 struct ext4_sb_info *sbi, char *buf)
2503 unsigned int *ui = (unsigned int *) (((char *) sbi) + a->offset);
2505 return snprintf(buf, PAGE_SIZE, "%u\n", *ui);
2508 static ssize_t sbi_ui_store(struct ext4_attr *a,
2509 struct ext4_sb_info *sbi,
2510 const char *buf, size_t count)
2512 unsigned int *ui = (unsigned int *) (((char *) sbi) + a->offset);
2515 if (parse_strtoul(buf, 0xffffffff, &t))
2521 static ssize_t trigger_test_error(struct ext4_attr *a,
2522 struct ext4_sb_info *sbi,
2523 const char *buf, size_t count)
2527 if (!capable(CAP_SYS_ADMIN))
2530 if (len && buf[len-1] == '\n')
2534 ext4_error(sbi->s_sb, "%.*s", len, buf);
2538 #define EXT4_ATTR_OFFSET(_name,_mode,_show,_store,_elname) \
2539 static struct ext4_attr ext4_attr_##_name = { \
2540 .attr = {.name = __stringify(_name), .mode = _mode }, \
2543 .offset = offsetof(struct ext4_sb_info, _elname), \
2545 #define EXT4_ATTR(name, mode, show, store) \
2546 static struct ext4_attr ext4_attr_##name = __ATTR(name, mode, show, store)
2548 #define EXT4_INFO_ATTR(name) EXT4_ATTR(name, 0444, NULL, NULL)
2549 #define EXT4_RO_ATTR(name) EXT4_ATTR(name, 0444, name##_show, NULL)
2550 #define EXT4_RW_ATTR(name) EXT4_ATTR(name, 0644, name##_show, name##_store)
2551 #define EXT4_RW_ATTR_SBI_UI(name, elname) \
2552 EXT4_ATTR_OFFSET(name, 0644, sbi_ui_show, sbi_ui_store, elname)
2553 #define ATTR_LIST(name) &ext4_attr_##name.attr
2555 EXT4_RO_ATTR(delayed_allocation_blocks);
2556 EXT4_RO_ATTR(session_write_kbytes);
2557 EXT4_RO_ATTR(lifetime_write_kbytes);
2558 EXT4_ATTR_OFFSET(inode_readahead_blks, 0644, sbi_ui_show,
2559 inode_readahead_blks_store, s_inode_readahead_blks);
2560 EXT4_RW_ATTR_SBI_UI(inode_goal, s_inode_goal);
2561 EXT4_RW_ATTR_SBI_UI(mb_stats, s_mb_stats);
2562 EXT4_RW_ATTR_SBI_UI(mb_max_to_scan, s_mb_max_to_scan);
2563 EXT4_RW_ATTR_SBI_UI(mb_min_to_scan, s_mb_min_to_scan);
2564 EXT4_RW_ATTR_SBI_UI(mb_order2_req, s_mb_order2_reqs);
2565 EXT4_RW_ATTR_SBI_UI(mb_stream_req, s_mb_stream_request);
2566 EXT4_RW_ATTR_SBI_UI(mb_group_prealloc, s_mb_group_prealloc);
2567 EXT4_RW_ATTR_SBI_UI(max_writeback_mb_bump, s_max_writeback_mb_bump);
2568 EXT4_RW_ATTR_SBI_UI(extent_max_zeroout_kb, s_extent_max_zeroout_kb);
2569 EXT4_ATTR(trigger_fs_error, 0200, NULL, trigger_test_error);
2571 static struct attribute *ext4_attrs[] = {
2572 ATTR_LIST(delayed_allocation_blocks),
2573 ATTR_LIST(session_write_kbytes),
2574 ATTR_LIST(lifetime_write_kbytes),
2575 ATTR_LIST(inode_readahead_blks),
2576 ATTR_LIST(inode_goal),
2577 ATTR_LIST(mb_stats),
2578 ATTR_LIST(mb_max_to_scan),
2579 ATTR_LIST(mb_min_to_scan),
2580 ATTR_LIST(mb_order2_req),
2581 ATTR_LIST(mb_stream_req),
2582 ATTR_LIST(mb_group_prealloc),
2583 ATTR_LIST(max_writeback_mb_bump),
2584 ATTR_LIST(extent_max_zeroout_kb),
2585 ATTR_LIST(trigger_fs_error),
2589 /* Features this copy of ext4 supports */
2590 EXT4_INFO_ATTR(lazy_itable_init);
2591 EXT4_INFO_ATTR(batched_discard);
2592 EXT4_INFO_ATTR(meta_bg_resize);
2594 static struct attribute *ext4_feat_attrs[] = {
2595 ATTR_LIST(lazy_itable_init),
2596 ATTR_LIST(batched_discard),
2597 ATTR_LIST(meta_bg_resize),
2601 static ssize_t ext4_attr_show(struct kobject *kobj,
2602 struct attribute *attr, char *buf)
2604 struct ext4_sb_info *sbi = container_of(kobj, struct ext4_sb_info,
2606 struct ext4_attr *a = container_of(attr, struct ext4_attr, attr);
2608 return a->show ? a->show(a, sbi, buf) : 0;
2611 static ssize_t ext4_attr_store(struct kobject *kobj,
2612 struct attribute *attr,
2613 const char *buf, size_t len)
2615 struct ext4_sb_info *sbi = container_of(kobj, struct ext4_sb_info,
2617 struct ext4_attr *a = container_of(attr, struct ext4_attr, attr);
2619 return a->store ? a->store(a, sbi, buf, len) : 0;
2622 static void ext4_sb_release(struct kobject *kobj)
2624 struct ext4_sb_info *sbi = container_of(kobj, struct ext4_sb_info,
2626 complete(&sbi->s_kobj_unregister);
2629 static const struct sysfs_ops ext4_attr_ops = {
2630 .show = ext4_attr_show,
2631 .store = ext4_attr_store,
2634 static struct kobj_type ext4_ktype = {
2635 .default_attrs = ext4_attrs,
2636 .sysfs_ops = &ext4_attr_ops,
2637 .release = ext4_sb_release,
2640 static void ext4_feat_release(struct kobject *kobj)
2642 complete(&ext4_feat->f_kobj_unregister);
2645 static struct kobj_type ext4_feat_ktype = {
2646 .default_attrs = ext4_feat_attrs,
2647 .sysfs_ops = &ext4_attr_ops,
2648 .release = ext4_feat_release,
2652 * Check whether this filesystem can be mounted based on
2653 * the features present and the RDONLY/RDWR mount requested.
2654 * Returns 1 if this filesystem can be mounted as requested,
2655 * 0 if it cannot be.
2657 static int ext4_feature_set_ok(struct super_block *sb, int readonly)
2659 if (EXT4_HAS_INCOMPAT_FEATURE(sb, ~EXT4_FEATURE_INCOMPAT_SUPP)) {
2660 ext4_msg(sb, KERN_ERR,
2661 "Couldn't mount because of "
2662 "unsupported optional features (%x)",
2663 (le32_to_cpu(EXT4_SB(sb)->s_es->s_feature_incompat) &
2664 ~EXT4_FEATURE_INCOMPAT_SUPP));
2671 /* Check that feature set is OK for a read-write mount */
2672 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, ~EXT4_FEATURE_RO_COMPAT_SUPP)) {
2673 ext4_msg(sb, KERN_ERR, "couldn't mount RDWR because of "
2674 "unsupported optional features (%x)",
2675 (le32_to_cpu(EXT4_SB(sb)->s_es->s_feature_ro_compat) &
2676 ~EXT4_FEATURE_RO_COMPAT_SUPP));
2680 * Large file size enabled file system can only be mounted
2681 * read-write on 32-bit systems if kernel is built with CONFIG_LBDAF
2683 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_HUGE_FILE)) {
2684 if (sizeof(blkcnt_t) < sizeof(u64)) {
2685 ext4_msg(sb, KERN_ERR, "Filesystem with huge files "
2686 "cannot be mounted RDWR without "
2691 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_BIGALLOC) &&
2692 !EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_EXTENTS)) {
2693 ext4_msg(sb, KERN_ERR,
2694 "Can't support bigalloc feature without "
2695 "extents feature\n");
2699 #ifndef CONFIG_QUOTA
2700 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_QUOTA) &&
2702 ext4_msg(sb, KERN_ERR,
2703 "Filesystem with quota feature cannot be mounted RDWR "
2704 "without CONFIG_QUOTA");
2707 #endif /* CONFIG_QUOTA */
2712 * This function is called once a day if we have errors logged
2713 * on the file system
2715 static void print_daily_error_info(unsigned long arg)
2717 struct super_block *sb = (struct super_block *) arg;
2718 struct ext4_sb_info *sbi;
2719 struct ext4_super_block *es;
2724 if (es->s_error_count)
2725 ext4_msg(sb, KERN_NOTICE, "error count: %u",
2726 le32_to_cpu(es->s_error_count));
2727 if (es->s_first_error_time) {
2728 printk(KERN_NOTICE "EXT4-fs (%s): initial error at %u: %.*s:%d",
2729 sb->s_id, le32_to_cpu(es->s_first_error_time),
2730 (int) sizeof(es->s_first_error_func),
2731 es->s_first_error_func,
2732 le32_to_cpu(es->s_first_error_line));
2733 if (es->s_first_error_ino)
2734 printk(": inode %u",
2735 le32_to_cpu(es->s_first_error_ino));
2736 if (es->s_first_error_block)
2737 printk(": block %llu", (unsigned long long)
2738 le64_to_cpu(es->s_first_error_block));
2741 if (es->s_last_error_time) {
2742 printk(KERN_NOTICE "EXT4-fs (%s): last error at %u: %.*s:%d",
2743 sb->s_id, le32_to_cpu(es->s_last_error_time),
2744 (int) sizeof(es->s_last_error_func),
2745 es->s_last_error_func,
2746 le32_to_cpu(es->s_last_error_line));
2747 if (es->s_last_error_ino)
2748 printk(": inode %u",
2749 le32_to_cpu(es->s_last_error_ino));
2750 if (es->s_last_error_block)
2751 printk(": block %llu", (unsigned long long)
2752 le64_to_cpu(es->s_last_error_block));
2755 mod_timer(&sbi->s_err_report, jiffies + 24*60*60*HZ); /* Once a day */
2758 /* Find next suitable group and run ext4_init_inode_table */
2759 static int ext4_run_li_request(struct ext4_li_request *elr)
2761 struct ext4_group_desc *gdp = NULL;
2762 ext4_group_t group, ngroups;
2763 struct super_block *sb;
2764 unsigned long timeout = 0;
2768 ngroups = EXT4_SB(sb)->s_groups_count;
2771 for (group = elr->lr_next_group; group < ngroups; group++) {
2772 gdp = ext4_get_group_desc(sb, group, NULL);
2778 if (!(gdp->bg_flags & cpu_to_le16(EXT4_BG_INODE_ZEROED)))
2782 if (group == ngroups)
2787 ret = ext4_init_inode_table(sb, group,
2788 elr->lr_timeout ? 0 : 1);
2789 if (elr->lr_timeout == 0) {
2790 timeout = (jiffies - timeout) *
2791 elr->lr_sbi->s_li_wait_mult;
2792 elr->lr_timeout = timeout;
2794 elr->lr_next_sched = jiffies + elr->lr_timeout;
2795 elr->lr_next_group = group + 1;
2803 * Remove lr_request from the list_request and free the
2804 * request structure. Should be called with li_list_mtx held
2806 static void ext4_remove_li_request(struct ext4_li_request *elr)
2808 struct ext4_sb_info *sbi;
2815 list_del(&elr->lr_request);
2816 sbi->s_li_request = NULL;
2820 static void ext4_unregister_li_request(struct super_block *sb)
2822 mutex_lock(&ext4_li_mtx);
2823 if (!ext4_li_info) {
2824 mutex_unlock(&ext4_li_mtx);
2828 mutex_lock(&ext4_li_info->li_list_mtx);
2829 ext4_remove_li_request(EXT4_SB(sb)->s_li_request);
2830 mutex_unlock(&ext4_li_info->li_list_mtx);
2831 mutex_unlock(&ext4_li_mtx);
2834 static struct task_struct *ext4_lazyinit_task;
2837 * This is the function where ext4lazyinit thread lives. It walks
2838 * through the request list searching for next scheduled filesystem.
2839 * When such a fs is found, run the lazy initialization request
2840 * (ext4_rn_li_request) and keep track of the time spend in this
2841 * function. Based on that time we compute next schedule time of
2842 * the request. When walking through the list is complete, compute
2843 * next waking time and put itself into sleep.
2845 static int ext4_lazyinit_thread(void *arg)
2847 struct ext4_lazy_init *eli = (struct ext4_lazy_init *)arg;
2848 struct list_head *pos, *n;
2849 struct ext4_li_request *elr;
2850 unsigned long next_wakeup, cur;
2852 BUG_ON(NULL == eli);
2856 next_wakeup = MAX_JIFFY_OFFSET;
2858 mutex_lock(&eli->li_list_mtx);
2859 if (list_empty(&eli->li_request_list)) {
2860 mutex_unlock(&eli->li_list_mtx);
2864 list_for_each_safe(pos, n, &eli->li_request_list) {
2865 elr = list_entry(pos, struct ext4_li_request,
2868 if (time_after_eq(jiffies, elr->lr_next_sched)) {
2869 if (ext4_run_li_request(elr) != 0) {
2870 /* error, remove the lazy_init job */
2871 ext4_remove_li_request(elr);
2876 if (time_before(elr->lr_next_sched, next_wakeup))
2877 next_wakeup = elr->lr_next_sched;
2879 mutex_unlock(&eli->li_list_mtx);
2884 if ((time_after_eq(cur, next_wakeup)) ||
2885 (MAX_JIFFY_OFFSET == next_wakeup)) {
2890 schedule_timeout_interruptible(next_wakeup - cur);
2892 if (kthread_should_stop()) {
2893 ext4_clear_request_list();
2900 * It looks like the request list is empty, but we need
2901 * to check it under the li_list_mtx lock, to prevent any
2902 * additions into it, and of course we should lock ext4_li_mtx
2903 * to atomically free the list and ext4_li_info, because at
2904 * this point another ext4 filesystem could be registering
2907 mutex_lock(&ext4_li_mtx);
2908 mutex_lock(&eli->li_list_mtx);
2909 if (!list_empty(&eli->li_request_list)) {
2910 mutex_unlock(&eli->li_list_mtx);
2911 mutex_unlock(&ext4_li_mtx);
2914 mutex_unlock(&eli->li_list_mtx);
2915 kfree(ext4_li_info);
2916 ext4_li_info = NULL;
2917 mutex_unlock(&ext4_li_mtx);
2922 static void ext4_clear_request_list(void)
2924 struct list_head *pos, *n;
2925 struct ext4_li_request *elr;
2927 mutex_lock(&ext4_li_info->li_list_mtx);
2928 list_for_each_safe(pos, n, &ext4_li_info->li_request_list) {
2929 elr = list_entry(pos, struct ext4_li_request,
2931 ext4_remove_li_request(elr);
2933 mutex_unlock(&ext4_li_info->li_list_mtx);
2936 static int ext4_run_lazyinit_thread(void)
2938 ext4_lazyinit_task = kthread_run(ext4_lazyinit_thread,
2939 ext4_li_info, "ext4lazyinit");
2940 if (IS_ERR(ext4_lazyinit_task)) {
2941 int err = PTR_ERR(ext4_lazyinit_task);
2942 ext4_clear_request_list();
2943 kfree(ext4_li_info);
2944 ext4_li_info = NULL;
2945 printk(KERN_CRIT "EXT4-fs: error %d creating inode table "
2946 "initialization thread\n",
2950 ext4_li_info->li_state |= EXT4_LAZYINIT_RUNNING;
2955 * Check whether it make sense to run itable init. thread or not.
2956 * If there is at least one uninitialized inode table, return
2957 * corresponding group number, else the loop goes through all
2958 * groups and return total number of groups.
2960 static ext4_group_t ext4_has_uninit_itable(struct super_block *sb)
2962 ext4_group_t group, ngroups = EXT4_SB(sb)->s_groups_count;
2963 struct ext4_group_desc *gdp = NULL;
2965 for (group = 0; group < ngroups; group++) {
2966 gdp = ext4_get_group_desc(sb, group, NULL);
2970 if (!(gdp->bg_flags & cpu_to_le16(EXT4_BG_INODE_ZEROED)))
2977 static int ext4_li_info_new(void)
2979 struct ext4_lazy_init *eli = NULL;
2981 eli = kzalloc(sizeof(*eli), GFP_KERNEL);
2985 INIT_LIST_HEAD(&eli->li_request_list);
2986 mutex_init(&eli->li_list_mtx);
2988 eli->li_state |= EXT4_LAZYINIT_QUIT;
2995 static struct ext4_li_request *ext4_li_request_new(struct super_block *sb,
2998 struct ext4_sb_info *sbi = EXT4_SB(sb);
2999 struct ext4_li_request *elr;
3002 elr = kzalloc(sizeof(*elr), GFP_KERNEL);
3008 elr->lr_next_group = start;
3011 * Randomize first schedule time of the request to
3012 * spread the inode table initialization requests
3015 get_random_bytes(&rnd, sizeof(rnd));
3016 elr->lr_next_sched = jiffies + (unsigned long)rnd %
3017 (EXT4_DEF_LI_MAX_START_DELAY * HZ);
3022 static int ext4_register_li_request(struct super_block *sb,
3023 ext4_group_t first_not_zeroed)
3025 struct ext4_sb_info *sbi = EXT4_SB(sb);
3026 struct ext4_li_request *elr;
3027 ext4_group_t ngroups = EXT4_SB(sb)->s_groups_count;
3030 if (sbi->s_li_request != NULL) {
3032 * Reset timeout so it can be computed again, because
3033 * s_li_wait_mult might have changed.
3035 sbi->s_li_request->lr_timeout = 0;
3039 if (first_not_zeroed == ngroups ||
3040 (sb->s_flags & MS_RDONLY) ||
3041 !test_opt(sb, INIT_INODE_TABLE))
3044 elr = ext4_li_request_new(sb, first_not_zeroed);
3048 mutex_lock(&ext4_li_mtx);
3050 if (NULL == ext4_li_info) {
3051 ret = ext4_li_info_new();
3056 mutex_lock(&ext4_li_info->li_list_mtx);
3057 list_add(&elr->lr_request, &ext4_li_info->li_request_list);
3058 mutex_unlock(&ext4_li_info->li_list_mtx);
3060 sbi->s_li_request = elr;
3062 * set elr to NULL here since it has been inserted to
3063 * the request_list and the removal and free of it is
3064 * handled by ext4_clear_request_list from now on.
3068 if (!(ext4_li_info->li_state & EXT4_LAZYINIT_RUNNING)) {
3069 ret = ext4_run_lazyinit_thread();
3074 mutex_unlock(&ext4_li_mtx);
3081 * We do not need to lock anything since this is called on
3084 static void ext4_destroy_lazyinit_thread(void)
3087 * If thread exited earlier
3088 * there's nothing to be done.
3090 if (!ext4_li_info || !ext4_lazyinit_task)
3093 kthread_stop(ext4_lazyinit_task);
3096 static int set_journal_csum_feature_set(struct super_block *sb)
3099 int compat, incompat;
3100 struct ext4_sb_info *sbi = EXT4_SB(sb);
3102 if (EXT4_HAS_RO_COMPAT_FEATURE(sb,
3103 EXT4_FEATURE_RO_COMPAT_METADATA_CSUM)) {
3104 /* journal checksum v2 */
3106 incompat = JBD2_FEATURE_INCOMPAT_CSUM_V2;
3108 /* journal checksum v1 */
3109 compat = JBD2_FEATURE_COMPAT_CHECKSUM;
3113 if (test_opt(sb, JOURNAL_ASYNC_COMMIT)) {
3114 ret = jbd2_journal_set_features(sbi->s_journal,
3116 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT |
3118 } else if (test_opt(sb, JOURNAL_CHECKSUM)) {
3119 ret = jbd2_journal_set_features(sbi->s_journal,
3122 jbd2_journal_clear_features(sbi->s_journal, 0, 0,
3123 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT);
3125 jbd2_journal_clear_features(sbi->s_journal,
3126 JBD2_FEATURE_COMPAT_CHECKSUM, 0,
3127 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT |
3128 JBD2_FEATURE_INCOMPAT_CSUM_V2);
3135 * Note: calculating the overhead so we can be compatible with
3136 * historical BSD practice is quite difficult in the face of
3137 * clusters/bigalloc. This is because multiple metadata blocks from
3138 * different block group can end up in the same allocation cluster.
3139 * Calculating the exact overhead in the face of clustered allocation
3140 * requires either O(all block bitmaps) in memory or O(number of block
3141 * groups**2) in time. We will still calculate the superblock for
3142 * older file systems --- and if we come across with a bigalloc file
3143 * system with zero in s_overhead_clusters the estimate will be close to
3144 * correct especially for very large cluster sizes --- but for newer
3145 * file systems, it's better to calculate this figure once at mkfs
3146 * time, and store it in the superblock. If the superblock value is
3147 * present (even for non-bigalloc file systems), we will use it.
3149 static int count_overhead(struct super_block *sb, ext4_group_t grp,
3152 struct ext4_sb_info *sbi = EXT4_SB(sb);
3153 struct ext4_group_desc *gdp;
3154 ext4_fsblk_t first_block, last_block, b;
3155 ext4_group_t i, ngroups = ext4_get_groups_count(sb);
3156 int s, j, count = 0;
3158 if (!EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_BIGALLOC))
3159 return (ext4_bg_has_super(sb, grp) + ext4_bg_num_gdb(sb, grp) +
3160 sbi->s_itb_per_group + 2);
3162 first_block = le32_to_cpu(sbi->s_es->s_first_data_block) +
3163 (grp * EXT4_BLOCKS_PER_GROUP(sb));
3164 last_block = first_block + EXT4_BLOCKS_PER_GROUP(sb) - 1;
3165 for (i = 0; i < ngroups; i++) {
3166 gdp = ext4_get_group_desc(sb, i, NULL);
3167 b = ext4_block_bitmap(sb, gdp);
3168 if (b >= first_block && b <= last_block) {
3169 ext4_set_bit(EXT4_B2C(sbi, b - first_block), buf);
3172 b = ext4_inode_bitmap(sb, gdp);
3173 if (b >= first_block && b <= last_block) {
3174 ext4_set_bit(EXT4_B2C(sbi, b - first_block), buf);
3177 b = ext4_inode_table(sb, gdp);
3178 if (b >= first_block && b + sbi->s_itb_per_group <= last_block)
3179 for (j = 0; j < sbi->s_itb_per_group; j++, b++) {
3180 int c = EXT4_B2C(sbi, b - first_block);
3181 ext4_set_bit(c, buf);
3187 if (ext4_bg_has_super(sb, grp)) {
3188 ext4_set_bit(s++, buf);
3191 for (j = ext4_bg_num_gdb(sb, grp); j > 0; j--) {
3192 ext4_set_bit(EXT4_B2C(sbi, s++), buf);
3198 return EXT4_CLUSTERS_PER_GROUP(sb) -
3199 ext4_count_free(buf, EXT4_CLUSTERS_PER_GROUP(sb) / 8);
3203 * Compute the overhead and stash it in sbi->s_overhead
3205 int ext4_calculate_overhead(struct super_block *sb)
3207 struct ext4_sb_info *sbi = EXT4_SB(sb);
3208 struct ext4_super_block *es = sbi->s_es;
3209 ext4_group_t i, ngroups = ext4_get_groups_count(sb);
3210 ext4_fsblk_t overhead = 0;
3211 char *buf = (char *) get_zeroed_page(GFP_KERNEL);
3217 * Compute the overhead (FS structures). This is constant
3218 * for a given filesystem unless the number of block groups
3219 * changes so we cache the previous value until it does.
3223 * All of the blocks before first_data_block are overhead
3225 overhead = EXT4_B2C(sbi, le32_to_cpu(es->s_first_data_block));
3228 * Add the overhead found in each block group
3230 for (i = 0; i < ngroups; i++) {
3233 blks = count_overhead(sb, i, buf);
3236 memset(buf, 0, PAGE_SIZE);
3239 /* Add the journal blocks as well */
3241 overhead += EXT4_NUM_B2C(sbi, sbi->s_journal->j_maxlen);
3243 sbi->s_overhead = overhead;
3245 free_page((unsigned long) buf);
3249 static int ext4_fill_super(struct super_block *sb, void *data, int silent)
3251 char *orig_data = kstrdup(data, GFP_KERNEL);
3252 struct buffer_head *bh;
3253 struct ext4_super_block *es = NULL;
3254 struct ext4_sb_info *sbi;
3256 ext4_fsblk_t sb_block = get_sb_block(&data);
3257 ext4_fsblk_t logical_sb_block;
3258 unsigned long offset = 0;
3259 unsigned long journal_devnum = 0;
3260 unsigned long def_mount_opts;
3265 int blocksize, clustersize;
3266 unsigned int db_count;
3268 int needs_recovery, has_huge_files, has_bigalloc;
3271 unsigned int journal_ioprio = DEFAULT_JOURNAL_IOPRIO;
3272 ext4_group_t first_not_zeroed;
3274 sbi = kzalloc(sizeof(*sbi), GFP_KERNEL);
3278 sbi->s_blockgroup_lock =
3279 kzalloc(sizeof(struct blockgroup_lock), GFP_KERNEL);
3280 if (!sbi->s_blockgroup_lock) {
3284 sb->s_fs_info = sbi;
3286 sbi->s_inode_readahead_blks = EXT4_DEF_INODE_READAHEAD_BLKS;
3287 sbi->s_sb_block = sb_block;
3288 if (sb->s_bdev->bd_part)
3289 sbi->s_sectors_written_start =
3290 part_stat_read(sb->s_bdev->bd_part, sectors[1]);
3292 /* Cleanup superblock name */
3293 for (cp = sb->s_id; (cp = strchr(cp, '/'));)
3296 /* -EINVAL is default */
3298 blocksize = sb_min_blocksize(sb, EXT4_MIN_BLOCK_SIZE);
3300 ext4_msg(sb, KERN_ERR, "unable to set blocksize");
3305 * The ext4 superblock will not be buffer aligned for other than 1kB
3306 * block sizes. We need to calculate the offset from buffer start.
3308 if (blocksize != EXT4_MIN_BLOCK_SIZE) {
3309 logical_sb_block = sb_block * EXT4_MIN_BLOCK_SIZE;
3310 offset = do_div(logical_sb_block, blocksize);
3312 logical_sb_block = sb_block;
3315 if (!(bh = sb_bread(sb, logical_sb_block))) {
3316 ext4_msg(sb, KERN_ERR, "unable to read superblock");
3320 * Note: s_es must be initialized as soon as possible because
3321 * some ext4 macro-instructions depend on its value
3323 es = (struct ext4_super_block *) (bh->b_data + offset);
3325 sb->s_magic = le16_to_cpu(es->s_magic);
3326 if (sb->s_magic != EXT4_SUPER_MAGIC)
3328 sbi->s_kbytes_written = le64_to_cpu(es->s_kbytes_written);
3330 /* Warn if metadata_csum and gdt_csum are both set. */
3331 if (EXT4_HAS_RO_COMPAT_FEATURE(sb,
3332 EXT4_FEATURE_RO_COMPAT_METADATA_CSUM) &&
3333 EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_GDT_CSUM))
3334 ext4_warning(sb, KERN_INFO "metadata_csum and uninit_bg are "
3335 "redundant flags; please run fsck.");
3337 /* Check for a known checksum algorithm */
3338 if (!ext4_verify_csum_type(sb, es)) {
3339 ext4_msg(sb, KERN_ERR, "VFS: Found ext4 filesystem with "
3340 "unknown checksum algorithm.");
3345 /* Load the checksum driver */
3346 if (EXT4_HAS_RO_COMPAT_FEATURE(sb,
3347 EXT4_FEATURE_RO_COMPAT_METADATA_CSUM)) {
3348 sbi->s_chksum_driver = crypto_alloc_shash("crc32c", 0, 0);
3349 if (IS_ERR(sbi->s_chksum_driver)) {
3350 ext4_msg(sb, KERN_ERR, "Cannot load crc32c driver.");
3351 ret = PTR_ERR(sbi->s_chksum_driver);
3352 sbi->s_chksum_driver = NULL;
3357 /* Check superblock checksum */
3358 if (!ext4_superblock_csum_verify(sb, es)) {
3359 ext4_msg(sb, KERN_ERR, "VFS: Found ext4 filesystem with "
3360 "invalid superblock checksum. Run e2fsck?");
3365 /* Precompute checksum seed for all metadata */
3366 if (EXT4_HAS_RO_COMPAT_FEATURE(sb,
3367 EXT4_FEATURE_RO_COMPAT_METADATA_CSUM))
3368 sbi->s_csum_seed = ext4_chksum(sbi, ~0, es->s_uuid,
3369 sizeof(es->s_uuid));
3371 /* Set defaults before we parse the mount options */
3372 def_mount_opts = le32_to_cpu(es->s_default_mount_opts);
3373 set_opt(sb, INIT_INODE_TABLE);
3374 if (def_mount_opts & EXT4_DEFM_DEBUG)
3376 if (def_mount_opts & EXT4_DEFM_BSDGROUPS)
3378 if (def_mount_opts & EXT4_DEFM_UID16)
3379 set_opt(sb, NO_UID32);
3380 /* xattr user namespace & acls are now defaulted on */
3381 set_opt(sb, XATTR_USER);
3382 #ifdef CONFIG_EXT4_FS_POSIX_ACL
3383 set_opt(sb, POSIX_ACL);
3385 set_opt(sb, MBLK_IO_SUBMIT);
3386 if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_DATA)
3387 set_opt(sb, JOURNAL_DATA);
3388 else if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_ORDERED)
3389 set_opt(sb, ORDERED_DATA);
3390 else if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_WBACK)
3391 set_opt(sb, WRITEBACK_DATA);
3393 if (le16_to_cpu(sbi->s_es->s_errors) == EXT4_ERRORS_PANIC)
3394 set_opt(sb, ERRORS_PANIC);
3395 else if (le16_to_cpu(sbi->s_es->s_errors) == EXT4_ERRORS_CONTINUE)
3396 set_opt(sb, ERRORS_CONT);
3398 set_opt(sb, ERRORS_RO);
3399 if (def_mount_opts & EXT4_DEFM_BLOCK_VALIDITY)
3400 set_opt(sb, BLOCK_VALIDITY);
3401 if (def_mount_opts & EXT4_DEFM_DISCARD)
3402 set_opt(sb, DISCARD);
3404 sbi->s_resuid = make_kuid(&init_user_ns, le16_to_cpu(es->s_def_resuid));
3405 sbi->s_resgid = make_kgid(&init_user_ns, le16_to_cpu(es->s_def_resgid));
3406 sbi->s_commit_interval = JBD2_DEFAULT_MAX_COMMIT_AGE * HZ;
3407 sbi->s_min_batch_time = EXT4_DEF_MIN_BATCH_TIME;
3408 sbi->s_max_batch_time = EXT4_DEF_MAX_BATCH_TIME;
3410 if ((def_mount_opts & EXT4_DEFM_NOBARRIER) == 0)
3411 set_opt(sb, BARRIER);
3414 * enable delayed allocation by default
3415 * Use -o nodelalloc to turn it off
3417 if (!IS_EXT3_SB(sb) && !IS_EXT2_SB(sb) &&
3418 ((def_mount_opts & EXT4_DEFM_NODELALLOC) == 0))
3419 set_opt(sb, DELALLOC);
3422 * set default s_li_wait_mult for lazyinit, for the case there is
3423 * no mount option specified.
3425 sbi->s_li_wait_mult = EXT4_DEF_LI_WAIT_MULT;
3427 if (!parse_options((char *) sbi->s_es->s_mount_opts, sb,
3428 &journal_devnum, &journal_ioprio, 0)) {
3429 ext4_msg(sb, KERN_WARNING,
3430 "failed to parse options in superblock: %s",
3431 sbi->s_es->s_mount_opts);
3433 sbi->s_def_mount_opt = sbi->s_mount_opt;
3434 if (!parse_options((char *) data, sb, &journal_devnum,
3435 &journal_ioprio, 0))
3438 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA) {
3439 printk_once(KERN_WARNING "EXT4-fs: Warning: mounting "
3440 "with data=journal disables delayed "
3441 "allocation and O_DIRECT support!\n");
3442 if (test_opt2(sb, EXPLICIT_DELALLOC)) {
3443 ext4_msg(sb, KERN_ERR, "can't mount with "
3444 "both data=journal and delalloc");
3447 if (test_opt(sb, DIOREAD_NOLOCK)) {
3448 ext4_msg(sb, KERN_ERR, "can't mount with "
3449 "both data=journal and delalloc");
3452 if (test_opt(sb, DELALLOC))
3453 clear_opt(sb, DELALLOC);
3456 sb->s_flags = (sb->s_flags & ~MS_POSIXACL) |
3457 (test_opt(sb, POSIX_ACL) ? MS_POSIXACL : 0);
3459 if (le32_to_cpu(es->s_rev_level) == EXT4_GOOD_OLD_REV &&
3460 (EXT4_HAS_COMPAT_FEATURE(sb, ~0U) ||
3461 EXT4_HAS_RO_COMPAT_FEATURE(sb, ~0U) ||
3462 EXT4_HAS_INCOMPAT_FEATURE(sb, ~0U)))
3463 ext4_msg(sb, KERN_WARNING,
3464 "feature flags set on rev 0 fs, "
3465 "running e2fsck is recommended");
3467 if (IS_EXT2_SB(sb)) {
3468 if (ext2_feature_set_ok(sb))
3469 ext4_msg(sb, KERN_INFO, "mounting ext2 file system "
3470 "using the ext4 subsystem");
3472 ext4_msg(sb, KERN_ERR, "couldn't mount as ext2 due "
3473 "to feature incompatibilities");
3478 if (IS_EXT3_SB(sb)) {
3479 if (ext3_feature_set_ok(sb))
3480 ext4_msg(sb, KERN_INFO, "mounting ext3 file system "
3481 "using the ext4 subsystem");
3483 ext4_msg(sb, KERN_ERR, "couldn't mount as ext3 due "
3484 "to feature incompatibilities");
3490 * Check feature flags regardless of the revision level, since we
3491 * previously didn't change the revision level when setting the flags,
3492 * so there is a chance incompat flags are set on a rev 0 filesystem.
3494 if (!ext4_feature_set_ok(sb, (sb->s_flags & MS_RDONLY)))
3497 blocksize = BLOCK_SIZE << le32_to_cpu(es->s_log_block_size);
3498 if (blocksize < EXT4_MIN_BLOCK_SIZE ||
3499 blocksize > EXT4_MAX_BLOCK_SIZE) {
3500 ext4_msg(sb, KERN_ERR,
3501 "Unsupported filesystem blocksize %d", blocksize);
3505 if (sb->s_blocksize != blocksize) {
3506 /* Validate the filesystem blocksize */
3507 if (!sb_set_blocksize(sb, blocksize)) {
3508 ext4_msg(sb, KERN_ERR, "bad block size %d",
3514 logical_sb_block = sb_block * EXT4_MIN_BLOCK_SIZE;
3515 offset = do_div(logical_sb_block, blocksize);
3516 bh = sb_bread(sb, logical_sb_block);
3518 ext4_msg(sb, KERN_ERR,
3519 "Can't read superblock on 2nd try");
3522 es = (struct ext4_super_block *)(bh->b_data + offset);
3524 if (es->s_magic != cpu_to_le16(EXT4_SUPER_MAGIC)) {
3525 ext4_msg(sb, KERN_ERR,
3526 "Magic mismatch, very weird!");
3531 has_huge_files = EXT4_HAS_RO_COMPAT_FEATURE(sb,
3532 EXT4_FEATURE_RO_COMPAT_HUGE_FILE);
3533 sbi->s_bitmap_maxbytes = ext4_max_bitmap_size(sb->s_blocksize_bits,
3535 sb->s_maxbytes = ext4_max_size(sb->s_blocksize_bits, has_huge_files);
3537 if (le32_to_cpu(es->s_rev_level) == EXT4_GOOD_OLD_REV) {
3538 sbi->s_inode_size = EXT4_GOOD_OLD_INODE_SIZE;
3539 sbi->s_first_ino = EXT4_GOOD_OLD_FIRST_INO;
3541 sbi->s_inode_size = le16_to_cpu(es->s_inode_size);
3542 sbi->s_first_ino = le32_to_cpu(es->s_first_ino);
3543 if ((sbi->s_inode_size < EXT4_GOOD_OLD_INODE_SIZE) ||
3544 (!is_power_of_2(sbi->s_inode_size)) ||
3545 (sbi->s_inode_size > blocksize)) {
3546 ext4_msg(sb, KERN_ERR,
3547 "unsupported inode size: %d",
3551 if (sbi->s_inode_size > EXT4_GOOD_OLD_INODE_SIZE)
3552 sb->s_time_gran = 1 << (EXT4_EPOCH_BITS - 2);
3555 sbi->s_desc_size = le16_to_cpu(es->s_desc_size);
3556 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_64BIT)) {
3557 if (sbi->s_desc_size < EXT4_MIN_DESC_SIZE_64BIT ||
3558 sbi->s_desc_size > EXT4_MAX_DESC_SIZE ||
3559 !is_power_of_2(sbi->s_desc_size)) {
3560 ext4_msg(sb, KERN_ERR,
3561 "unsupported descriptor size %lu",
3566 sbi->s_desc_size = EXT4_MIN_DESC_SIZE;
3568 sbi->s_blocks_per_group = le32_to_cpu(es->s_blocks_per_group);
3569 sbi->s_inodes_per_group = le32_to_cpu(es->s_inodes_per_group);
3570 if (EXT4_INODE_SIZE(sb) == 0 || EXT4_INODES_PER_GROUP(sb) == 0)
3573 sbi->s_inodes_per_block = blocksize / EXT4_INODE_SIZE(sb);
3574 if (sbi->s_inodes_per_block == 0)
3576 sbi->s_itb_per_group = sbi->s_inodes_per_group /
3577 sbi->s_inodes_per_block;
3578 sbi->s_desc_per_block = blocksize / EXT4_DESC_SIZE(sb);
3580 sbi->s_mount_state = le16_to_cpu(es->s_state);
3581 sbi->s_addr_per_block_bits = ilog2(EXT4_ADDR_PER_BLOCK(sb));
3582 sbi->s_desc_per_block_bits = ilog2(EXT4_DESC_PER_BLOCK(sb));
3584 for (i = 0; i < 4; i++)
3585 sbi->s_hash_seed[i] = le32_to_cpu(es->s_hash_seed[i]);
3586 sbi->s_def_hash_version = es->s_def_hash_version;
3587 i = le32_to_cpu(es->s_flags);
3588 if (i & EXT2_FLAGS_UNSIGNED_HASH)
3589 sbi->s_hash_unsigned = 3;
3590 else if ((i & EXT2_FLAGS_SIGNED_HASH) == 0) {
3591 #ifdef __CHAR_UNSIGNED__
3592 es->s_flags |= cpu_to_le32(EXT2_FLAGS_UNSIGNED_HASH);
3593 sbi->s_hash_unsigned = 3;
3595 es->s_flags |= cpu_to_le32(EXT2_FLAGS_SIGNED_HASH);
3599 /* Handle clustersize */
3600 clustersize = BLOCK_SIZE << le32_to_cpu(es->s_log_cluster_size);
3601 has_bigalloc = EXT4_HAS_RO_COMPAT_FEATURE(sb,
3602 EXT4_FEATURE_RO_COMPAT_BIGALLOC);
3604 if (clustersize < blocksize) {
3605 ext4_msg(sb, KERN_ERR,
3606 "cluster size (%d) smaller than "
3607 "block size (%d)", clustersize, blocksize);
3610 sbi->s_cluster_bits = le32_to_cpu(es->s_log_cluster_size) -
3611 le32_to_cpu(es->s_log_block_size);
3612 sbi->s_clusters_per_group =
3613 le32_to_cpu(es->s_clusters_per_group);
3614 if (sbi->s_clusters_per_group > blocksize * 8) {
3615 ext4_msg(sb, KERN_ERR,
3616 "#clusters per group too big: %lu",
3617 sbi->s_clusters_per_group);
3620 if (sbi->s_blocks_per_group !=
3621 (sbi->s_clusters_per_group * (clustersize / blocksize))) {
3622 ext4_msg(sb, KERN_ERR, "blocks per group (%lu) and "
3623 "clusters per group (%lu) inconsistent",
3624 sbi->s_blocks_per_group,
3625 sbi->s_clusters_per_group);
3629 if (clustersize != blocksize) {
3630 ext4_warning(sb, "fragment/cluster size (%d) != "
3631 "block size (%d)", clustersize,
3633 clustersize = blocksize;
3635 if (sbi->s_blocks_per_group > blocksize * 8) {
3636 ext4_msg(sb, KERN_ERR,
3637 "#blocks per group too big: %lu",
3638 sbi->s_blocks_per_group);
3641 sbi->s_clusters_per_group = sbi->s_blocks_per_group;
3642 sbi->s_cluster_bits = 0;
3644 sbi->s_cluster_ratio = clustersize / blocksize;
3646 if (sbi->s_inodes_per_group > blocksize * 8) {
3647 ext4_msg(sb, KERN_ERR,
3648 "#inodes per group too big: %lu",
3649 sbi->s_inodes_per_group);
3654 * Test whether we have more sectors than will fit in sector_t,
3655 * and whether the max offset is addressable by the page cache.
3657 err = generic_check_addressable(sb->s_blocksize_bits,
3658 ext4_blocks_count(es));
3660 ext4_msg(sb, KERN_ERR, "filesystem"
3661 " too large to mount safely on this system");
3662 if (sizeof(sector_t) < 8)
3663 ext4_msg(sb, KERN_WARNING, "CONFIG_LBDAF not enabled");
3667 if (EXT4_BLOCKS_PER_GROUP(sb) == 0)
3670 /* check blocks count against device size */
3671 blocks_count = sb->s_bdev->bd_inode->i_size >> sb->s_blocksize_bits;
3672 if (blocks_count && ext4_blocks_count(es) > blocks_count) {
3673 ext4_msg(sb, KERN_WARNING, "bad geometry: block count %llu "
3674 "exceeds size of device (%llu blocks)",
3675 ext4_blocks_count(es), blocks_count);
3680 * It makes no sense for the first data block to be beyond the end
3681 * of the filesystem.
3683 if (le32_to_cpu(es->s_first_data_block) >= ext4_blocks_count(es)) {
3684 ext4_msg(sb, KERN_WARNING, "bad geometry: first data "
3685 "block %u is beyond end of filesystem (%llu)",
3686 le32_to_cpu(es->s_first_data_block),
3687 ext4_blocks_count(es));
3690 blocks_count = (ext4_blocks_count(es) -
3691 le32_to_cpu(es->s_first_data_block) +
3692 EXT4_BLOCKS_PER_GROUP(sb) - 1);
3693 do_div(blocks_count, EXT4_BLOCKS_PER_GROUP(sb));
3694 if (blocks_count > ((uint64_t)1<<32) - EXT4_DESC_PER_BLOCK(sb)) {
3695 ext4_msg(sb, KERN_WARNING, "groups count too large: %u "
3696 "(block count %llu, first data block %u, "
3697 "blocks per group %lu)", sbi->s_groups_count,
3698 ext4_blocks_count(es),
3699 le32_to_cpu(es->s_first_data_block),
3700 EXT4_BLOCKS_PER_GROUP(sb));
3703 sbi->s_groups_count = blocks_count;
3704 sbi->s_blockfile_groups = min_t(ext4_group_t, sbi->s_groups_count,
3705 (EXT4_MAX_BLOCK_FILE_PHYS / EXT4_BLOCKS_PER_GROUP(sb)));
3706 db_count = (sbi->s_groups_count + EXT4_DESC_PER_BLOCK(sb) - 1) /
3707 EXT4_DESC_PER_BLOCK(sb);
3708 sbi->s_group_desc = ext4_kvmalloc(db_count *
3709 sizeof(struct buffer_head *),
3711 if (sbi->s_group_desc == NULL) {
3712 ext4_msg(sb, KERN_ERR, "not enough memory");
3718 sbi->s_proc = proc_mkdir(sb->s_id, ext4_proc_root);
3721 proc_create_data("options", S_IRUGO, sbi->s_proc,
3722 &ext4_seq_options_fops, sb);
3724 bgl_lock_init(sbi->s_blockgroup_lock);
3726 for (i = 0; i < db_count; i++) {
3727 block = descriptor_loc(sb, logical_sb_block, i);
3728 sbi->s_group_desc[i] = sb_bread(sb, block);
3729 if (!sbi->s_group_desc[i]) {
3730 ext4_msg(sb, KERN_ERR,
3731 "can't read group descriptor %d", i);
3736 if (!ext4_check_descriptors(sb, &first_not_zeroed)) {
3737 ext4_msg(sb, KERN_ERR, "group descriptors corrupted!");
3740 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_FLEX_BG))
3741 if (!ext4_fill_flex_info(sb)) {
3742 ext4_msg(sb, KERN_ERR,
3743 "unable to initialize "
3744 "flex_bg meta info!");
3748 sbi->s_gdb_count = db_count;
3749 get_random_bytes(&sbi->s_next_generation, sizeof(u32));
3750 spin_lock_init(&sbi->s_next_gen_lock);
3752 init_timer(&sbi->s_err_report);
3753 sbi->s_err_report.function = print_daily_error_info;
3754 sbi->s_err_report.data = (unsigned long) sb;
3756 err = percpu_counter_init(&sbi->s_freeclusters_counter,
3757 ext4_count_free_clusters(sb));
3759 err = percpu_counter_init(&sbi->s_freeinodes_counter,
3760 ext4_count_free_inodes(sb));
3763 err = percpu_counter_init(&sbi->s_dirs_counter,
3764 ext4_count_dirs(sb));
3767 err = percpu_counter_init(&sbi->s_dirtyclusters_counter, 0);
3770 ext4_msg(sb, KERN_ERR, "insufficient memory");
3774 sbi->s_stripe = ext4_get_stripe_size(sbi);
3775 sbi->s_max_writeback_mb_bump = 128;
3776 sbi->s_extent_max_zeroout_kb = 32;
3779 * set up enough so that it can read an inode
3781 if (!test_opt(sb, NOLOAD) &&
3782 EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL))
3783 sb->s_op = &ext4_sops;
3785 sb->s_op = &ext4_nojournal_sops;
3786 sb->s_export_op = &ext4_export_ops;
3787 sb->s_xattr = ext4_xattr_handlers;
3789 sb->s_qcop = &ext4_qctl_operations;
3790 sb->dq_op = &ext4_quota_operations;
3792 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_QUOTA)) {
3793 /* Use qctl operations for hidden quota files. */
3794 sb->s_qcop = &ext4_qctl_sysfile_operations;
3797 memcpy(sb->s_uuid, es->s_uuid, sizeof(es->s_uuid));
3799 INIT_LIST_HEAD(&sbi->s_orphan); /* unlinked but open files */
3800 mutex_init(&sbi->s_orphan_lock);
3804 needs_recovery = (es->s_last_orphan != 0 ||
3805 EXT4_HAS_INCOMPAT_FEATURE(sb,
3806 EXT4_FEATURE_INCOMPAT_RECOVER));
3808 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_MMP) &&
3809 !(sb->s_flags & MS_RDONLY))
3810 if (ext4_multi_mount_protect(sb, le64_to_cpu(es->s_mmp_block)))
3814 * The first inode we look at is the journal inode. Don't try
3815 * root first: it may be modified in the journal!
3817 if (!test_opt(sb, NOLOAD) &&
3818 EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL)) {
3819 if (ext4_load_journal(sb, es, journal_devnum))
3821 } else if (test_opt(sb, NOLOAD) && !(sb->s_flags & MS_RDONLY) &&
3822 EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER)) {
3823 ext4_msg(sb, KERN_ERR, "required journal recovery "
3824 "suppressed and not mounted read-only");
3825 goto failed_mount_wq;
3827 clear_opt(sb, DATA_FLAGS);
3828 sbi->s_journal = NULL;
3833 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_64BIT) &&
3834 !jbd2_journal_set_features(EXT4_SB(sb)->s_journal, 0, 0,
3835 JBD2_FEATURE_INCOMPAT_64BIT)) {
3836 ext4_msg(sb, KERN_ERR, "Failed to set 64-bit journal feature");
3837 goto failed_mount_wq;
3840 if (!set_journal_csum_feature_set(sb)) {
3841 ext4_msg(sb, KERN_ERR, "Failed to set journal checksum "
3843 goto failed_mount_wq;
3846 /* We have now updated the journal if required, so we can
3847 * validate the data journaling mode. */
3848 switch (test_opt(sb, DATA_FLAGS)) {
3850 /* No mode set, assume a default based on the journal
3851 * capabilities: ORDERED_DATA if the journal can
3852 * cope, else JOURNAL_DATA
3854 if (jbd2_journal_check_available_features
3855 (sbi->s_journal, 0, 0, JBD2_FEATURE_INCOMPAT_REVOKE))
3856 set_opt(sb, ORDERED_DATA);
3858 set_opt(sb, JOURNAL_DATA);
3861 case EXT4_MOUNT_ORDERED_DATA:
3862 case EXT4_MOUNT_WRITEBACK_DATA:
3863 if (!jbd2_journal_check_available_features
3864 (sbi->s_journal, 0, 0, JBD2_FEATURE_INCOMPAT_REVOKE)) {
3865 ext4_msg(sb, KERN_ERR, "Journal does not support "
3866 "requested data journaling mode");
3867 goto failed_mount_wq;
3872 set_task_ioprio(sbi->s_journal->j_task, journal_ioprio);
3874 sbi->s_journal->j_commit_callback = ext4_journal_commit_callback;
3877 * The journal may have updated the bg summary counts, so we
3878 * need to update the global counters.
3880 percpu_counter_set(&sbi->s_freeclusters_counter,
3881 ext4_count_free_clusters(sb));
3882 percpu_counter_set(&sbi->s_freeinodes_counter,
3883 ext4_count_free_inodes(sb));
3884 percpu_counter_set(&sbi->s_dirs_counter,
3885 ext4_count_dirs(sb));
3886 percpu_counter_set(&sbi->s_dirtyclusters_counter, 0);
3890 * Get the # of file system overhead blocks from the
3891 * superblock if present.
3893 if (es->s_overhead_clusters)
3894 sbi->s_overhead = le32_to_cpu(es->s_overhead_clusters);
3896 err = ext4_calculate_overhead(sb);
3898 goto failed_mount_wq;
3902 * The maximum number of concurrent works can be high and
3903 * concurrency isn't really necessary. Limit it to 1.
3905 EXT4_SB(sb)->dio_unwritten_wq =
3906 alloc_workqueue("ext4-dio-unwritten", WQ_MEM_RECLAIM | WQ_UNBOUND, 1);
3907 if (!EXT4_SB(sb)->dio_unwritten_wq) {
3908 printk(KERN_ERR "EXT4-fs: failed to create DIO workqueue\n");
3910 goto failed_mount_wq;
3914 * The jbd2_journal_load will have done any necessary log recovery,
3915 * so we can safely mount the rest of the filesystem now.
3918 root = ext4_iget(sb, EXT4_ROOT_INO);
3920 ext4_msg(sb, KERN_ERR, "get root inode failed");
3921 ret = PTR_ERR(root);
3925 if (!S_ISDIR(root->i_mode) || !root->i_blocks || !root->i_size) {
3926 ext4_msg(sb, KERN_ERR, "corrupt root inode, run e2fsck");
3930 sb->s_root = d_make_root(root);
3932 ext4_msg(sb, KERN_ERR, "get root dentry failed");
3937 if (ext4_setup_super(sb, es, sb->s_flags & MS_RDONLY))
3938 sb->s_flags |= MS_RDONLY;
3940 /* determine the minimum size of new large inodes, if present */
3941 if (sbi->s_inode_size > EXT4_GOOD_OLD_INODE_SIZE) {
3942 sbi->s_want_extra_isize = sizeof(struct ext4_inode) -
3943 EXT4_GOOD_OLD_INODE_SIZE;
3944 if (EXT4_HAS_RO_COMPAT_FEATURE(sb,
3945 EXT4_FEATURE_RO_COMPAT_EXTRA_ISIZE)) {
3946 if (sbi->s_want_extra_isize <
3947 le16_to_cpu(es->s_want_extra_isize))
3948 sbi->s_want_extra_isize =
3949 le16_to_cpu(es->s_want_extra_isize);
3950 if (sbi->s_want_extra_isize <
3951 le16_to_cpu(es->s_min_extra_isize))
3952 sbi->s_want_extra_isize =
3953 le16_to_cpu(es->s_min_extra_isize);
3956 /* Check if enough inode space is available */
3957 if (EXT4_GOOD_OLD_INODE_SIZE + sbi->s_want_extra_isize >
3958 sbi->s_inode_size) {
3959 sbi->s_want_extra_isize = sizeof(struct ext4_inode) -
3960 EXT4_GOOD_OLD_INODE_SIZE;
3961 ext4_msg(sb, KERN_INFO, "required extra inode space not"
3965 err = ext4_setup_system_zone(sb);
3967 ext4_msg(sb, KERN_ERR, "failed to initialize system "
3969 goto failed_mount4a;
3973 err = ext4_mb_init(sb);
3975 ext4_msg(sb, KERN_ERR, "failed to initialize mballoc (%d)",
3980 err = ext4_register_li_request(sb, first_not_zeroed);
3984 sbi->s_kobj.kset = ext4_kset;
3985 init_completion(&sbi->s_kobj_unregister);
3986 err = kobject_init_and_add(&sbi->s_kobj, &ext4_ktype, NULL,
3991 EXT4_SB(sb)->s_mount_state |= EXT4_ORPHAN_FS;
3992 ext4_orphan_cleanup(sb, es);
3993 EXT4_SB(sb)->s_mount_state &= ~EXT4_ORPHAN_FS;
3994 if (needs_recovery) {
3995 ext4_msg(sb, KERN_INFO, "recovery complete");
3996 ext4_mark_recovery_complete(sb, es);
3998 if (EXT4_SB(sb)->s_journal) {
3999 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA)
4000 descr = " journalled data mode";
4001 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA)
4002 descr = " ordered data mode";
4004 descr = " writeback data mode";
4006 descr = "out journal";
4009 /* Enable quota usage during mount. */
4010 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_QUOTA) &&
4011 !(sb->s_flags & MS_RDONLY)) {
4012 err = ext4_enable_quotas(sb);
4016 #endif /* CONFIG_QUOTA */
4018 if (test_opt(sb, DISCARD)) {
4019 struct request_queue *q = bdev_get_queue(sb->s_bdev);
4020 if (!blk_queue_discard(q))
4021 ext4_msg(sb, KERN_WARNING,
4022 "mounting with \"discard\" option, but "
4023 "the device does not support discard");
4026 ext4_msg(sb, KERN_INFO, "mounted filesystem with%s. "
4027 "Opts: %s%s%s", descr, sbi->s_es->s_mount_opts,
4028 *sbi->s_es->s_mount_opts ? "; " : "", orig_data);
4030 if (es->s_error_count)
4031 mod_timer(&sbi->s_err_report, jiffies + 300*HZ); /* 5 minutes */
4038 ext4_msg(sb, KERN_ERR, "VFS: Can't find ext4 filesystem");
4043 kobject_del(&sbi->s_kobj);
4046 ext4_unregister_li_request(sb);
4048 ext4_mb_release(sb);
4050 ext4_ext_release(sb);
4051 ext4_release_system_zone(sb);
4056 ext4_msg(sb, KERN_ERR, "mount failed");
4057 destroy_workqueue(EXT4_SB(sb)->dio_unwritten_wq);
4059 if (sbi->s_journal) {
4060 jbd2_journal_destroy(sbi->s_journal);
4061 sbi->s_journal = NULL;
4064 del_timer(&sbi->s_err_report);
4065 if (sbi->s_flex_groups)
4066 ext4_kvfree(sbi->s_flex_groups);
4067 percpu_counter_destroy(&sbi->s_freeclusters_counter);
4068 percpu_counter_destroy(&sbi->s_freeinodes_counter);
4069 percpu_counter_destroy(&sbi->s_dirs_counter);
4070 percpu_counter_destroy(&sbi->s_dirtyclusters_counter);
4072 kthread_stop(sbi->s_mmp_tsk);
4074 for (i = 0; i < db_count; i++)
4075 brelse(sbi->s_group_desc[i]);
4076 ext4_kvfree(sbi->s_group_desc);
4078 if (sbi->s_chksum_driver)
4079 crypto_free_shash(sbi->s_chksum_driver);
4081 remove_proc_entry("options", sbi->s_proc);
4082 remove_proc_entry(sb->s_id, ext4_proc_root);
4085 for (i = 0; i < MAXQUOTAS; i++)
4086 kfree(sbi->s_qf_names[i]);
4088 ext4_blkdev_remove(sbi);
4091 sb->s_fs_info = NULL;
4092 kfree(sbi->s_blockgroup_lock);
4096 return err ? err : ret;
4100 * Setup any per-fs journal parameters now. We'll do this both on
4101 * initial mount, once the journal has been initialised but before we've
4102 * done any recovery; and again on any subsequent remount.
4104 static void ext4_init_journal_params(struct super_block *sb, journal_t *journal)
4106 struct ext4_sb_info *sbi = EXT4_SB(sb);
4108 journal->j_commit_interval = sbi->s_commit_interval;
4109 journal->j_min_batch_time = sbi->s_min_batch_time;
4110 journal->j_max_batch_time = sbi->s_max_batch_time;
4112 write_lock(&journal->j_state_lock);
4113 if (test_opt(sb, BARRIER))
4114 journal->j_flags |= JBD2_BARRIER;
4116 journal->j_flags &= ~JBD2_BARRIER;
4117 if (test_opt(sb, DATA_ERR_ABORT))
4118 journal->j_flags |= JBD2_ABORT_ON_SYNCDATA_ERR;
4120 journal->j_flags &= ~JBD2_ABORT_ON_SYNCDATA_ERR;
4121 write_unlock(&journal->j_state_lock);
4124 static journal_t *ext4_get_journal(struct super_block *sb,
4125 unsigned int journal_inum)
4127 struct inode *journal_inode;
4130 BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL));
4132 /* First, test for the existence of a valid inode on disk. Bad
4133 * things happen if we iget() an unused inode, as the subsequent
4134 * iput() will try to delete it. */
4136 journal_inode = ext4_iget(sb, journal_inum);
4137 if (IS_ERR(journal_inode)) {
4138 ext4_msg(sb, KERN_ERR, "no journal found");
4141 if (!journal_inode->i_nlink) {
4142 make_bad_inode(journal_inode);
4143 iput(journal_inode);
4144 ext4_msg(sb, KERN_ERR, "journal inode is deleted");
4148 jbd_debug(2, "Journal inode found at %p: %lld bytes\n",
4149 journal_inode, journal_inode->i_size);
4150 if (!S_ISREG(journal_inode->i_mode)) {
4151 ext4_msg(sb, KERN_ERR, "invalid journal inode");
4152 iput(journal_inode);
4156 journal = jbd2_journal_init_inode(journal_inode);
4158 ext4_msg(sb, KERN_ERR, "Could not load journal inode");
4159 iput(journal_inode);
4162 journal->j_private = sb;
4163 ext4_init_journal_params(sb, journal);
4167 static journal_t *ext4_get_dev_journal(struct super_block *sb,
4170 struct buffer_head *bh;
4174 int hblock, blocksize;
4175 ext4_fsblk_t sb_block;
4176 unsigned long offset;
4177 struct ext4_super_block *es;
4178 struct block_device *bdev;
4180 BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL));
4182 bdev = ext4_blkdev_get(j_dev, sb);
4186 blocksize = sb->s_blocksize;
4187 hblock = bdev_logical_block_size(bdev);
4188 if (blocksize < hblock) {
4189 ext4_msg(sb, KERN_ERR,
4190 "blocksize too small for journal device");
4194 sb_block = EXT4_MIN_BLOCK_SIZE / blocksize;
4195 offset = EXT4_MIN_BLOCK_SIZE % blocksize;
4196 set_blocksize(bdev, blocksize);
4197 if (!(bh = __bread(bdev, sb_block, blocksize))) {
4198 ext4_msg(sb, KERN_ERR, "couldn't read superblock of "
4199 "external journal");
4203 es = (struct ext4_super_block *) (bh->b_data + offset);
4204 if ((le16_to_cpu(es->s_magic) != EXT4_SUPER_MAGIC) ||
4205 !(le32_to_cpu(es->s_feature_incompat) &
4206 EXT4_FEATURE_INCOMPAT_JOURNAL_DEV)) {
4207 ext4_msg(sb, KERN_ERR, "external journal has "
4213 if (memcmp(EXT4_SB(sb)->s_es->s_journal_uuid, es->s_uuid, 16)) {
4214 ext4_msg(sb, KERN_ERR, "journal UUID does not match");
4219 len = ext4_blocks_count(es);
4220 start = sb_block + 1;
4221 brelse(bh); /* we're done with the superblock */
4223 journal = jbd2_journal_init_dev(bdev, sb->s_bdev,
4224 start, len, blocksize);
4226 ext4_msg(sb, KERN_ERR, "failed to create device journal");
4229 journal->j_private = sb;
4230 ll_rw_block(READ, 1, &journal->j_sb_buffer);
4231 wait_on_buffer(journal->j_sb_buffer);
4232 if (!buffer_uptodate(journal->j_sb_buffer)) {
4233 ext4_msg(sb, KERN_ERR, "I/O error on journal device");
4236 if (be32_to_cpu(journal->j_superblock->s_nr_users) != 1) {
4237 ext4_msg(sb, KERN_ERR, "External journal has more than one "
4238 "user (unsupported) - %d",
4239 be32_to_cpu(journal->j_superblock->s_nr_users));
4242 EXT4_SB(sb)->journal_bdev = bdev;
4243 ext4_init_journal_params(sb, journal);
4247 jbd2_journal_destroy(journal);
4249 ext4_blkdev_put(bdev);
4253 static int ext4_load_journal(struct super_block *sb,
4254 struct ext4_super_block *es,
4255 unsigned long journal_devnum)
4258 unsigned int journal_inum = le32_to_cpu(es->s_journal_inum);
4261 int really_read_only;
4263 BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL));
4265 if (journal_devnum &&
4266 journal_devnum != le32_to_cpu(es->s_journal_dev)) {
4267 ext4_msg(sb, KERN_INFO, "external journal device major/minor "
4268 "numbers have changed");
4269 journal_dev = new_decode_dev(journal_devnum);
4271 journal_dev = new_decode_dev(le32_to_cpu(es->s_journal_dev));
4273 really_read_only = bdev_read_only(sb->s_bdev);
4276 * Are we loading a blank journal or performing recovery after a
4277 * crash? For recovery, we need to check in advance whether we
4278 * can get read-write access to the device.
4280 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER)) {
4281 if (sb->s_flags & MS_RDONLY) {
4282 ext4_msg(sb, KERN_INFO, "INFO: recovery "
4283 "required on readonly filesystem");
4284 if (really_read_only) {
4285 ext4_msg(sb, KERN_ERR, "write access "
4286 "unavailable, cannot proceed");
4289 ext4_msg(sb, KERN_INFO, "write access will "
4290 "be enabled during recovery");
4294 if (journal_inum && journal_dev) {
4295 ext4_msg(sb, KERN_ERR, "filesystem has both journal "
4296 "and inode journals!");
4301 if (!(journal = ext4_get_journal(sb, journal_inum)))
4304 if (!(journal = ext4_get_dev_journal(sb, journal_dev)))
4308 if (!(journal->j_flags & JBD2_BARRIER))
4309 ext4_msg(sb, KERN_INFO, "barriers disabled");
4311 if (!EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER))
4312 err = jbd2_journal_wipe(journal, !really_read_only);
4314 char *save = kmalloc(EXT4_S_ERR_LEN, GFP_KERNEL);
4316 memcpy(save, ((char *) es) +
4317 EXT4_S_ERR_START, EXT4_S_ERR_LEN);
4318 err = jbd2_journal_load(journal);
4320 memcpy(((char *) es) + EXT4_S_ERR_START,
4321 save, EXT4_S_ERR_LEN);
4326 ext4_msg(sb, KERN_ERR, "error loading journal");
4327 jbd2_journal_destroy(journal);
4331 EXT4_SB(sb)->s_journal = journal;
4332 ext4_clear_journal_err(sb, es);
4334 if (!really_read_only && journal_devnum &&
4335 journal_devnum != le32_to_cpu(es->s_journal_dev)) {
4336 es->s_journal_dev = cpu_to_le32(journal_devnum);
4338 /* Make sure we flush the recovery flag to disk. */
4339 ext4_commit_super(sb, 1);
4345 static int ext4_commit_super(struct super_block *sb, int sync)
4347 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
4348 struct buffer_head *sbh = EXT4_SB(sb)->s_sbh;
4351 if (!sbh || block_device_ejected(sb))
4353 if (buffer_write_io_error(sbh)) {
4355 * Oh, dear. A previous attempt to write the
4356 * superblock failed. This could happen because the
4357 * USB device was yanked out. Or it could happen to
4358 * be a transient write error and maybe the block will
4359 * be remapped. Nothing we can do but to retry the
4360 * write and hope for the best.
4362 ext4_msg(sb, KERN_ERR, "previous I/O error to "
4363 "superblock detected");
4364 clear_buffer_write_io_error(sbh);
4365 set_buffer_uptodate(sbh);
4368 * If the file system is mounted read-only, don't update the
4369 * superblock write time. This avoids updating the superblock
4370 * write time when we are mounting the root file system
4371 * read/only but we need to replay the journal; at that point,
4372 * for people who are east of GMT and who make their clock
4373 * tick in localtime for Windows bug-for-bug compatibility,
4374 * the clock is set in the future, and this will cause e2fsck
4375 * to complain and force a full file system check.
4377 if (!(sb->s_flags & MS_RDONLY))
4378 es->s_wtime = cpu_to_le32(get_seconds());
4379 if (sb->s_bdev->bd_part)
4380 es->s_kbytes_written =
4381 cpu_to_le64(EXT4_SB(sb)->s_kbytes_written +
4382 ((part_stat_read(sb->s_bdev->bd_part, sectors[1]) -
4383 EXT4_SB(sb)->s_sectors_written_start) >> 1));
4385 es->s_kbytes_written =
4386 cpu_to_le64(EXT4_SB(sb)->s_kbytes_written);
4387 ext4_free_blocks_count_set(es,
4388 EXT4_C2B(EXT4_SB(sb), percpu_counter_sum_positive(
4389 &EXT4_SB(sb)->s_freeclusters_counter)));
4390 es->s_free_inodes_count =
4391 cpu_to_le32(percpu_counter_sum_positive(
4392 &EXT4_SB(sb)->s_freeinodes_counter));
4393 BUFFER_TRACE(sbh, "marking dirty");
4394 ext4_superblock_csum_set(sb);
4395 mark_buffer_dirty(sbh);
4397 error = sync_dirty_buffer(sbh);
4401 error = buffer_write_io_error(sbh);
4403 ext4_msg(sb, KERN_ERR, "I/O error while writing "
4405 clear_buffer_write_io_error(sbh);
4406 set_buffer_uptodate(sbh);
4413 * Have we just finished recovery? If so, and if we are mounting (or
4414 * remounting) the filesystem readonly, then we will end up with a
4415 * consistent fs on disk. Record that fact.
4417 static void ext4_mark_recovery_complete(struct super_block *sb,
4418 struct ext4_super_block *es)
4420 journal_t *journal = EXT4_SB(sb)->s_journal;
4422 if (!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL)) {
4423 BUG_ON(journal != NULL);
4426 jbd2_journal_lock_updates(journal);
4427 if (jbd2_journal_flush(journal) < 0)
4430 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER) &&
4431 sb->s_flags & MS_RDONLY) {
4432 EXT4_CLEAR_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
4433 ext4_commit_super(sb, 1);
4437 jbd2_journal_unlock_updates(journal);
4441 * If we are mounting (or read-write remounting) a filesystem whose journal
4442 * has recorded an error from a previous lifetime, move that error to the
4443 * main filesystem now.
4445 static void ext4_clear_journal_err(struct super_block *sb,
4446 struct ext4_super_block *es)
4452 BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL));
4454 journal = EXT4_SB(sb)->s_journal;
4457 * Now check for any error status which may have been recorded in the
4458 * journal by a prior ext4_error() or ext4_abort()
4461 j_errno = jbd2_journal_errno(journal);
4465 errstr = ext4_decode_error(sb, j_errno, nbuf);
4466 ext4_warning(sb, "Filesystem error recorded "
4467 "from previous mount: %s", errstr);
4468 ext4_warning(sb, "Marking fs in need of filesystem check.");
4470 EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS;
4471 es->s_state |= cpu_to_le16(EXT4_ERROR_FS);
4472 ext4_commit_super(sb, 1);
4474 jbd2_journal_clear_err(journal);
4475 jbd2_journal_update_sb_errno(journal);
4480 * Force the running and committing transactions to commit,
4481 * and wait on the commit.
4483 int ext4_force_commit(struct super_block *sb)
4488 if (sb->s_flags & MS_RDONLY)
4491 journal = EXT4_SB(sb)->s_journal;
4493 ret = ext4_journal_force_commit(journal);
4498 static int ext4_sync_fs(struct super_block *sb, int wait)
4502 struct ext4_sb_info *sbi = EXT4_SB(sb);
4504 trace_ext4_sync_fs(sb, wait);
4505 flush_workqueue(sbi->dio_unwritten_wq);
4507 * Writeback quota in non-journalled quota case - journalled quota has
4510 dquot_writeback_dquots(sb, -1);
4511 if (jbd2_journal_start_commit(sbi->s_journal, &target)) {
4513 jbd2_log_wait_commit(sbi->s_journal, target);
4519 * LVM calls this function before a (read-only) snapshot is created. This
4520 * gives us a chance to flush the journal completely and mark the fs clean.
4522 * Note that only this function cannot bring a filesystem to be in a clean
4523 * state independently. It relies on upper layer to stop all data & metadata
4526 static int ext4_freeze(struct super_block *sb)
4531 if (sb->s_flags & MS_RDONLY)
4534 journal = EXT4_SB(sb)->s_journal;
4536 /* Now we set up the journal barrier. */
4537 jbd2_journal_lock_updates(journal);
4540 * Don't clear the needs_recovery flag if we failed to flush
4543 error = jbd2_journal_flush(journal);
4547 /* Journal blocked and flushed, clear needs_recovery flag. */
4548 EXT4_CLEAR_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
4549 error = ext4_commit_super(sb, 1);
4551 /* we rely on upper layer to stop further updates */
4552 jbd2_journal_unlock_updates(EXT4_SB(sb)->s_journal);
4557 * Called by LVM after the snapshot is done. We need to reset the RECOVER
4558 * flag here, even though the filesystem is not technically dirty yet.
4560 static int ext4_unfreeze(struct super_block *sb)
4562 if (sb->s_flags & MS_RDONLY)
4565 /* Reset the needs_recovery flag before the fs is unlocked. */
4566 EXT4_SET_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
4567 ext4_commit_super(sb, 1);
4572 * Structure to save mount options for ext4_remount's benefit
4574 struct ext4_mount_options {
4575 unsigned long s_mount_opt;
4576 unsigned long s_mount_opt2;
4579 unsigned long s_commit_interval;
4580 u32 s_min_batch_time, s_max_batch_time;
4583 char *s_qf_names[MAXQUOTAS];
4587 static int ext4_remount(struct super_block *sb, int *flags, char *data)
4589 struct ext4_super_block *es;
4590 struct ext4_sb_info *sbi = EXT4_SB(sb);
4591 unsigned long old_sb_flags;
4592 struct ext4_mount_options old_opts;
4593 int enable_quota = 0;
4595 unsigned int journal_ioprio = DEFAULT_JOURNAL_IOPRIO;
4600 char *orig_data = kstrdup(data, GFP_KERNEL);
4602 /* Store the original options */
4603 old_sb_flags = sb->s_flags;
4604 old_opts.s_mount_opt = sbi->s_mount_opt;
4605 old_opts.s_mount_opt2 = sbi->s_mount_opt2;
4606 old_opts.s_resuid = sbi->s_resuid;
4607 old_opts.s_resgid = sbi->s_resgid;
4608 old_opts.s_commit_interval = sbi->s_commit_interval;
4609 old_opts.s_min_batch_time = sbi->s_min_batch_time;
4610 old_opts.s_max_batch_time = sbi->s_max_batch_time;
4612 old_opts.s_jquota_fmt = sbi->s_jquota_fmt;
4613 for (i = 0; i < MAXQUOTAS; i++)
4614 old_opts.s_qf_names[i] = sbi->s_qf_names[i];
4616 if (sbi->s_journal && sbi->s_journal->j_task->io_context)
4617 journal_ioprio = sbi->s_journal->j_task->io_context->ioprio;
4620 * Allow the "check" option to be passed as a remount option.
4622 if (!parse_options(data, sb, NULL, &journal_ioprio, 1)) {
4627 if (sbi->s_mount_flags & EXT4_MF_FS_ABORTED)
4628 ext4_abort(sb, "Abort forced by user");
4630 sb->s_flags = (sb->s_flags & ~MS_POSIXACL) |
4631 (test_opt(sb, POSIX_ACL) ? MS_POSIXACL : 0);
4635 if (sbi->s_journal) {
4636 ext4_init_journal_params(sb, sbi->s_journal);
4637 set_task_ioprio(sbi->s_journal->j_task, journal_ioprio);
4640 if ((*flags & MS_RDONLY) != (sb->s_flags & MS_RDONLY)) {
4641 if (sbi->s_mount_flags & EXT4_MF_FS_ABORTED) {
4646 if (*flags & MS_RDONLY) {
4647 err = dquot_suspend(sb, -1);
4652 * First of all, the unconditional stuff we have to do
4653 * to disable replay of the journal when we next remount
4655 sb->s_flags |= MS_RDONLY;
4658 * OK, test if we are remounting a valid rw partition
4659 * readonly, and if so set the rdonly flag and then
4660 * mark the partition as valid again.
4662 if (!(es->s_state & cpu_to_le16(EXT4_VALID_FS)) &&
4663 (sbi->s_mount_state & EXT4_VALID_FS))
4664 es->s_state = cpu_to_le16(sbi->s_mount_state);
4667 ext4_mark_recovery_complete(sb, es);
4669 /* Make sure we can mount this feature set readwrite */
4670 if (!ext4_feature_set_ok(sb, 0)) {
4675 * Make sure the group descriptor checksums
4676 * are sane. If they aren't, refuse to remount r/w.
4678 for (g = 0; g < sbi->s_groups_count; g++) {
4679 struct ext4_group_desc *gdp =
4680 ext4_get_group_desc(sb, g, NULL);
4682 if (!ext4_group_desc_csum_verify(sb, g, gdp)) {
4683 ext4_msg(sb, KERN_ERR,
4684 "ext4_remount: Checksum for group %u failed (%u!=%u)",
4685 g, le16_to_cpu(ext4_group_desc_csum(sbi, g, gdp)),
4686 le16_to_cpu(gdp->bg_checksum));
4693 * If we have an unprocessed orphan list hanging
4694 * around from a previously readonly bdev mount,
4695 * require a full umount/remount for now.
4697 if (es->s_last_orphan) {
4698 ext4_msg(sb, KERN_WARNING, "Couldn't "
4699 "remount RDWR because of unprocessed "
4700 "orphan inode list. Please "
4701 "umount/remount instead");
4707 * Mounting a RDONLY partition read-write, so reread
4708 * and store the current valid flag. (It may have
4709 * been changed by e2fsck since we originally mounted
4713 ext4_clear_journal_err(sb, es);
4714 sbi->s_mount_state = le16_to_cpu(es->s_state);
4715 if (!ext4_setup_super(sb, es, 0))
4716 sb->s_flags &= ~MS_RDONLY;
4717 if (EXT4_HAS_INCOMPAT_FEATURE(sb,
4718 EXT4_FEATURE_INCOMPAT_MMP))
4719 if (ext4_multi_mount_protect(sb,
4720 le64_to_cpu(es->s_mmp_block))) {
4729 * Reinitialize lazy itable initialization thread based on
4732 if ((sb->s_flags & MS_RDONLY) || !test_opt(sb, INIT_INODE_TABLE))
4733 ext4_unregister_li_request(sb);
4735 ext4_group_t first_not_zeroed;
4736 first_not_zeroed = ext4_has_uninit_itable(sb);
4737 ext4_register_li_request(sb, first_not_zeroed);
4740 ext4_setup_system_zone(sb);
4741 if (sbi->s_journal == NULL && !(old_sb_flags & MS_RDONLY))
4742 ext4_commit_super(sb, 1);
4745 /* Release old quota file names */
4746 for (i = 0; i < MAXQUOTAS; i++)
4747 if (old_opts.s_qf_names[i] &&
4748 old_opts.s_qf_names[i] != sbi->s_qf_names[i])
4749 kfree(old_opts.s_qf_names[i]);
4751 if (sb_any_quota_suspended(sb))
4752 dquot_resume(sb, -1);
4753 else if (EXT4_HAS_RO_COMPAT_FEATURE(sb,
4754 EXT4_FEATURE_RO_COMPAT_QUOTA)) {
4755 err = ext4_enable_quotas(sb);
4762 ext4_msg(sb, KERN_INFO, "re-mounted. Opts: %s", orig_data);
4767 sb->s_flags = old_sb_flags;
4768 sbi->s_mount_opt = old_opts.s_mount_opt;
4769 sbi->s_mount_opt2 = old_opts.s_mount_opt2;
4770 sbi->s_resuid = old_opts.s_resuid;
4771 sbi->s_resgid = old_opts.s_resgid;
4772 sbi->s_commit_interval = old_opts.s_commit_interval;
4773 sbi->s_min_batch_time = old_opts.s_min_batch_time;
4774 sbi->s_max_batch_time = old_opts.s_max_batch_time;
4776 sbi->s_jquota_fmt = old_opts.s_jquota_fmt;
4777 for (i = 0; i < MAXQUOTAS; i++) {
4778 if (sbi->s_qf_names[i] &&
4779 old_opts.s_qf_names[i] != sbi->s_qf_names[i])
4780 kfree(sbi->s_qf_names[i]);
4781 sbi->s_qf_names[i] = old_opts.s_qf_names[i];
4788 static int ext4_statfs(struct dentry *dentry, struct kstatfs *buf)
4790 struct super_block *sb = dentry->d_sb;
4791 struct ext4_sb_info *sbi = EXT4_SB(sb);
4792 struct ext4_super_block *es = sbi->s_es;
4793 ext4_fsblk_t overhead = 0;
4797 if (!test_opt(sb, MINIX_DF))
4798 overhead = sbi->s_overhead;
4800 buf->f_type = EXT4_SUPER_MAGIC;
4801 buf->f_bsize = sb->s_blocksize;
4802 buf->f_blocks = ext4_blocks_count(es) - EXT4_C2B(sbi, overhead);
4803 bfree = percpu_counter_sum_positive(&sbi->s_freeclusters_counter) -
4804 percpu_counter_sum_positive(&sbi->s_dirtyclusters_counter);
4805 /* prevent underflow in case that few free space is available */
4806 buf->f_bfree = EXT4_C2B(sbi, max_t(s64, bfree, 0));
4807 buf->f_bavail = buf->f_bfree - ext4_r_blocks_count(es);
4808 if (buf->f_bfree < ext4_r_blocks_count(es))
4810 buf->f_files = le32_to_cpu(es->s_inodes_count);
4811 buf->f_ffree = percpu_counter_sum_positive(&sbi->s_freeinodes_counter);
4812 buf->f_namelen = EXT4_NAME_LEN;
4813 fsid = le64_to_cpup((void *)es->s_uuid) ^
4814 le64_to_cpup((void *)es->s_uuid + sizeof(u64));
4815 buf->f_fsid.val[0] = fsid & 0xFFFFFFFFUL;
4816 buf->f_fsid.val[1] = (fsid >> 32) & 0xFFFFFFFFUL;
4821 /* Helper function for writing quotas on sync - we need to start transaction
4822 * before quota file is locked for write. Otherwise the are possible deadlocks:
4823 * Process 1 Process 2
4824 * ext4_create() quota_sync()
4825 * jbd2_journal_start() write_dquot()
4826 * dquot_initialize() down(dqio_mutex)
4827 * down(dqio_mutex) jbd2_journal_start()
4833 static inline struct inode *dquot_to_inode(struct dquot *dquot)
4835 return sb_dqopt(dquot->dq_sb)->files[dquot->dq_id.type];
4838 static int ext4_write_dquot(struct dquot *dquot)
4842 struct inode *inode;
4844 inode = dquot_to_inode(dquot);
4845 handle = ext4_journal_start(inode,
4846 EXT4_QUOTA_TRANS_BLOCKS(dquot->dq_sb));
4848 return PTR_ERR(handle);
4849 ret = dquot_commit(dquot);
4850 err = ext4_journal_stop(handle);
4856 static int ext4_acquire_dquot(struct dquot *dquot)
4861 handle = ext4_journal_start(dquot_to_inode(dquot),
4862 EXT4_QUOTA_INIT_BLOCKS(dquot->dq_sb));
4864 return PTR_ERR(handle);
4865 ret = dquot_acquire(dquot);
4866 err = ext4_journal_stop(handle);
4872 static int ext4_release_dquot(struct dquot *dquot)
4877 handle = ext4_journal_start(dquot_to_inode(dquot),
4878 EXT4_QUOTA_DEL_BLOCKS(dquot->dq_sb));
4879 if (IS_ERR(handle)) {
4880 /* Release dquot anyway to avoid endless cycle in dqput() */
4881 dquot_release(dquot);
4882 return PTR_ERR(handle);
4884 ret = dquot_release(dquot);
4885 err = ext4_journal_stop(handle);
4891 static int ext4_mark_dquot_dirty(struct dquot *dquot)
4893 /* Are we journaling quotas? */
4894 if (EXT4_SB(dquot->dq_sb)->s_qf_names[USRQUOTA] ||
4895 EXT4_SB(dquot->dq_sb)->s_qf_names[GRPQUOTA]) {
4896 dquot_mark_dquot_dirty(dquot);
4897 return ext4_write_dquot(dquot);
4899 return dquot_mark_dquot_dirty(dquot);
4903 static int ext4_write_info(struct super_block *sb, int type)
4908 /* Data block + inode block */
4909 handle = ext4_journal_start(sb->s_root->d_inode, 2);
4911 return PTR_ERR(handle);
4912 ret = dquot_commit_info(sb, type);
4913 err = ext4_journal_stop(handle);
4920 * Turn on quotas during mount time - we need to find
4921 * the quota file and such...
4923 static int ext4_quota_on_mount(struct super_block *sb, int type)
4925 return dquot_quota_on_mount(sb, EXT4_SB(sb)->s_qf_names[type],
4926 EXT4_SB(sb)->s_jquota_fmt, type);
4930 * Standard function to be called on quota_on
4932 static int ext4_quota_on(struct super_block *sb, int type, int format_id,
4937 if (!test_opt(sb, QUOTA))
4940 /* Quotafile not on the same filesystem? */
4941 if (path->dentry->d_sb != sb)
4943 /* Journaling quota? */
4944 if (EXT4_SB(sb)->s_qf_names[type]) {
4945 /* Quotafile not in fs root? */
4946 if (path->dentry->d_parent != sb->s_root)
4947 ext4_msg(sb, KERN_WARNING,
4948 "Quota file not on filesystem root. "
4949 "Journaled quota will not work");
4953 * When we journal data on quota file, we have to flush journal to see
4954 * all updates to the file when we bypass pagecache...
4956 if (EXT4_SB(sb)->s_journal &&
4957 ext4_should_journal_data(path->dentry->d_inode)) {
4959 * We don't need to lock updates but journal_flush() could
4960 * otherwise be livelocked...
4962 jbd2_journal_lock_updates(EXT4_SB(sb)->s_journal);
4963 err = jbd2_journal_flush(EXT4_SB(sb)->s_journal);
4964 jbd2_journal_unlock_updates(EXT4_SB(sb)->s_journal);
4969 return dquot_quota_on(sb, type, format_id, path);
4972 static int ext4_quota_enable(struct super_block *sb, int type, int format_id,
4976 struct inode *qf_inode;
4977 unsigned long qf_inums[MAXQUOTAS] = {
4978 le32_to_cpu(EXT4_SB(sb)->s_es->s_usr_quota_inum),
4979 le32_to_cpu(EXT4_SB(sb)->s_es->s_grp_quota_inum)
4982 BUG_ON(!EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_QUOTA));
4984 if (!qf_inums[type])
4987 qf_inode = ext4_iget(sb, qf_inums[type]);
4988 if (IS_ERR(qf_inode)) {
4989 ext4_error(sb, "Bad quota inode # %lu", qf_inums[type]);
4990 return PTR_ERR(qf_inode);
4993 err = dquot_enable(qf_inode, type, format_id, flags);
4999 /* Enable usage tracking for all quota types. */
5000 static int ext4_enable_quotas(struct super_block *sb)
5003 unsigned long qf_inums[MAXQUOTAS] = {
5004 le32_to_cpu(EXT4_SB(sb)->s_es->s_usr_quota_inum),
5005 le32_to_cpu(EXT4_SB(sb)->s_es->s_grp_quota_inum)
5008 sb_dqopt(sb)->flags |= DQUOT_QUOTA_SYS_FILE;
5009 for (type = 0; type < MAXQUOTAS; type++) {
5010 if (qf_inums[type]) {
5011 err = ext4_quota_enable(sb, type, QFMT_VFS_V1,
5012 DQUOT_USAGE_ENABLED);
5015 "Failed to enable quota tracking "
5016 "(type=%d, err=%d). Please run "
5017 "e2fsck to fix.", type, err);
5026 * quota_on function that is used when QUOTA feature is set.
5028 static int ext4_quota_on_sysfile(struct super_block *sb, int type,
5031 if (!EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_QUOTA))
5035 * USAGE was enabled at mount time. Only need to enable LIMITS now.
5037 return ext4_quota_enable(sb, type, format_id, DQUOT_LIMITS_ENABLED);
5040 static int ext4_quota_off(struct super_block *sb, int type)
5042 struct inode *inode = sb_dqopt(sb)->files[type];
5045 /* Force all delayed allocation blocks to be allocated.
5046 * Caller already holds s_umount sem */
5047 if (test_opt(sb, DELALLOC))
5048 sync_filesystem(sb);
5053 /* Update modification times of quota files when userspace can
5054 * start looking at them */
5055 handle = ext4_journal_start(inode, 1);
5058 inode->i_mtime = inode->i_ctime = CURRENT_TIME;
5059 ext4_mark_inode_dirty(handle, inode);
5060 ext4_journal_stop(handle);
5063 return dquot_quota_off(sb, type);
5067 * quota_off function that is used when QUOTA feature is set.
5069 static int ext4_quota_off_sysfile(struct super_block *sb, int type)
5071 if (!EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_QUOTA))
5074 /* Disable only the limits. */
5075 return dquot_disable(sb, type, DQUOT_LIMITS_ENABLED);
5078 /* Read data from quotafile - avoid pagecache and such because we cannot afford
5079 * acquiring the locks... As quota files are never truncated and quota code
5080 * itself serializes the operations (and no one else should touch the files)
5081 * we don't have to be afraid of races */
5082 static ssize_t ext4_quota_read(struct super_block *sb, int type, char *data,
5083 size_t len, loff_t off)
5085 struct inode *inode = sb_dqopt(sb)->files[type];
5086 ext4_lblk_t blk = off >> EXT4_BLOCK_SIZE_BITS(sb);
5088 int offset = off & (sb->s_blocksize - 1);
5091 struct buffer_head *bh;
5092 loff_t i_size = i_size_read(inode);
5096 if (off+len > i_size)
5099 while (toread > 0) {
5100 tocopy = sb->s_blocksize - offset < toread ?
5101 sb->s_blocksize - offset : toread;
5102 bh = ext4_bread(NULL, inode, blk, 0, &err);
5105 if (!bh) /* A hole? */
5106 memset(data, 0, tocopy);
5108 memcpy(data, bh->b_data+offset, tocopy);
5118 /* Write to quotafile (we know the transaction is already started and has
5119 * enough credits) */
5120 static ssize_t ext4_quota_write(struct super_block *sb, int type,
5121 const char *data, size_t len, loff_t off)
5123 struct inode *inode = sb_dqopt(sb)->files[type];
5124 ext4_lblk_t blk = off >> EXT4_BLOCK_SIZE_BITS(sb);
5126 int offset = off & (sb->s_blocksize - 1);
5127 struct buffer_head *bh;
5128 handle_t *handle = journal_current_handle();
5130 if (EXT4_SB(sb)->s_journal && !handle) {
5131 ext4_msg(sb, KERN_WARNING, "Quota write (off=%llu, len=%llu)"
5132 " cancelled because transaction is not started",
5133 (unsigned long long)off, (unsigned long long)len);
5137 * Since we account only one data block in transaction credits,
5138 * then it is impossible to cross a block boundary.
5140 if (sb->s_blocksize - offset < len) {
5141 ext4_msg(sb, KERN_WARNING, "Quota write (off=%llu, len=%llu)"
5142 " cancelled because not block aligned",
5143 (unsigned long long)off, (unsigned long long)len);
5147 bh = ext4_bread(handle, inode, blk, 1, &err);
5150 err = ext4_journal_get_write_access(handle, bh);
5156 memcpy(bh->b_data+offset, data, len);
5157 flush_dcache_page(bh->b_page);
5159 err = ext4_handle_dirty_metadata(handle, NULL, bh);
5164 if (inode->i_size < off + len) {
5165 i_size_write(inode, off + len);
5166 EXT4_I(inode)->i_disksize = inode->i_size;
5167 ext4_mark_inode_dirty(handle, inode);
5174 static struct dentry *ext4_mount(struct file_system_type *fs_type, int flags,
5175 const char *dev_name, void *data)
5177 return mount_bdev(fs_type, flags, dev_name, data, ext4_fill_super);
5180 #if !defined(CONFIG_EXT2_FS) && !defined(CONFIG_EXT2_FS_MODULE) && defined(CONFIG_EXT4_USE_FOR_EXT23)
5181 static inline void register_as_ext2(void)
5183 int err = register_filesystem(&ext2_fs_type);
5186 "EXT4-fs: Unable to register as ext2 (%d)\n", err);
5189 static inline void unregister_as_ext2(void)
5191 unregister_filesystem(&ext2_fs_type);
5194 static inline int ext2_feature_set_ok(struct super_block *sb)
5196 if (EXT4_HAS_INCOMPAT_FEATURE(sb, ~EXT2_FEATURE_INCOMPAT_SUPP))
5198 if (sb->s_flags & MS_RDONLY)
5200 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, ~EXT2_FEATURE_RO_COMPAT_SUPP))
5204 MODULE_ALIAS("ext2");
5206 static inline void register_as_ext2(void) { }
5207 static inline void unregister_as_ext2(void) { }
5208 static inline int ext2_feature_set_ok(struct super_block *sb) { return 0; }
5211 #if !defined(CONFIG_EXT3_FS) && !defined(CONFIG_EXT3_FS_MODULE) && defined(CONFIG_EXT4_USE_FOR_EXT23)
5212 static inline void register_as_ext3(void)
5214 int err = register_filesystem(&ext3_fs_type);
5217 "EXT4-fs: Unable to register as ext3 (%d)\n", err);
5220 static inline void unregister_as_ext3(void)
5222 unregister_filesystem(&ext3_fs_type);
5225 static inline int ext3_feature_set_ok(struct super_block *sb)
5227 if (EXT4_HAS_INCOMPAT_FEATURE(sb, ~EXT3_FEATURE_INCOMPAT_SUPP))
5229 if (!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL))
5231 if (sb->s_flags & MS_RDONLY)
5233 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, ~EXT3_FEATURE_RO_COMPAT_SUPP))
5237 MODULE_ALIAS("ext3");
5239 static inline void register_as_ext3(void) { }
5240 static inline void unregister_as_ext3(void) { }
5241 static inline int ext3_feature_set_ok(struct super_block *sb) { return 0; }
5244 static struct file_system_type ext4_fs_type = {
5245 .owner = THIS_MODULE,
5247 .mount = ext4_mount,
5248 .kill_sb = kill_block_super,
5249 .fs_flags = FS_REQUIRES_DEV,
5252 static int __init ext4_init_feat_adverts(void)
5254 struct ext4_features *ef;
5257 ef = kzalloc(sizeof(struct ext4_features), GFP_KERNEL);
5261 ef->f_kobj.kset = ext4_kset;
5262 init_completion(&ef->f_kobj_unregister);
5263 ret = kobject_init_and_add(&ef->f_kobj, &ext4_feat_ktype, NULL,
5276 static void ext4_exit_feat_adverts(void)
5278 kobject_put(&ext4_feat->f_kobj);
5279 wait_for_completion(&ext4_feat->f_kobj_unregister);
5283 /* Shared across all ext4 file systems */
5284 wait_queue_head_t ext4__ioend_wq[EXT4_WQ_HASH_SZ];
5285 struct mutex ext4__aio_mutex[EXT4_WQ_HASH_SZ];
5287 static int __init ext4_init_fs(void)
5291 ext4_li_info = NULL;
5292 mutex_init(&ext4_li_mtx);
5294 /* Build-time check for flags consistency */
5295 ext4_check_flag_values();
5297 for (i = 0; i < EXT4_WQ_HASH_SZ; i++) {
5298 mutex_init(&ext4__aio_mutex[i]);
5299 init_waitqueue_head(&ext4__ioend_wq[i]);
5302 err = ext4_init_es();
5306 err = ext4_init_pageio();
5310 err = ext4_init_system_zone();
5313 ext4_kset = kset_create_and_add("ext4", NULL, fs_kobj);
5318 ext4_proc_root = proc_mkdir("fs/ext4", NULL);
5320 err = ext4_init_feat_adverts();
5324 err = ext4_init_mballoc();
5328 err = ext4_init_xattr();
5331 err = init_inodecache();
5336 err = register_filesystem(&ext4_fs_type);
5342 unregister_as_ext2();
5343 unregister_as_ext3();
5344 destroy_inodecache();
5348 ext4_exit_mballoc();
5350 ext4_exit_feat_adverts();
5353 remove_proc_entry("fs/ext4", NULL);
5354 kset_unregister(ext4_kset);
5356 ext4_exit_system_zone();
5365 static void __exit ext4_exit_fs(void)
5367 ext4_destroy_lazyinit_thread();
5368 unregister_as_ext2();
5369 unregister_as_ext3();
5370 unregister_filesystem(&ext4_fs_type);
5371 destroy_inodecache();
5373 ext4_exit_mballoc();
5374 ext4_exit_feat_adverts();
5375 remove_proc_entry("fs/ext4", NULL);
5376 kset_unregister(ext4_kset);
5377 ext4_exit_system_zone();
5381 MODULE_AUTHOR("Remy Card, Stephen Tweedie, Andrew Morton, Andreas Dilger, Theodore Ts'o and others");
5382 MODULE_DESCRIPTION("Fourth Extended Filesystem");
5383 MODULE_LICENSE("GPL");
5384 module_init(ext4_init_fs)
5385 module_exit(ext4_exit_fs)