2 * linux/fs/ext4/super.c
4 * Copyright (C) 1992, 1993, 1994, 1995
5 * Remy Card (card@masi.ibp.fr)
6 * Laboratoire MASI - Institut Blaise Pascal
7 * Universite Pierre et Marie Curie (Paris VI)
11 * linux/fs/minix/inode.c
13 * Copyright (C) 1991, 1992 Linus Torvalds
15 * Big-endian to little-endian byte-swapping/bitmaps by
16 * David S. Miller (davem@caip.rutgers.edu), 1995
19 #include <linux/module.h>
20 #include <linux/string.h>
22 #include <linux/time.h>
23 #include <linux/vmalloc.h>
24 #include <linux/jbd2.h>
25 #include <linux/slab.h>
26 #include <linux/init.h>
27 #include <linux/blkdev.h>
28 #include <linux/parser.h>
29 #include <linux/buffer_head.h>
30 #include <linux/exportfs.h>
31 #include <linux/vfs.h>
32 #include <linux/random.h>
33 #include <linux/mount.h>
34 #include <linux/namei.h>
35 #include <linux/quotaops.h>
36 #include <linux/seq_file.h>
37 #include <linux/proc_fs.h>
38 #include <linux/ctype.h>
39 #include <linux/log2.h>
40 #include <linux/crc16.h>
41 #include <linux/cleancache.h>
42 #include <asm/uaccess.h>
44 #include <linux/kthread.h>
45 #include <linux/freezer.h>
48 #include "ext4_extents.h" /* Needed for trace points definition */
49 #include "ext4_jbd2.h"
54 #define CREATE_TRACE_POINTS
55 #include <trace/events/ext4.h>
57 static struct proc_dir_entry *ext4_proc_root;
58 static struct kset *ext4_kset;
59 static struct ext4_lazy_init *ext4_li_info;
60 static struct mutex ext4_li_mtx;
61 static struct ext4_features *ext4_feat;
62 static int ext4_mballoc_ready;
64 static int ext4_load_journal(struct super_block *, struct ext4_super_block *,
65 unsigned long journal_devnum);
66 static int ext4_show_options(struct seq_file *seq, struct dentry *root);
67 static int ext4_commit_super(struct super_block *sb, int sync);
68 static void ext4_mark_recovery_complete(struct super_block *sb,
69 struct ext4_super_block *es);
70 static void ext4_clear_journal_err(struct super_block *sb,
71 struct ext4_super_block *es);
72 static int ext4_sync_fs(struct super_block *sb, int wait);
73 static int ext4_remount(struct super_block *sb, int *flags, char *data);
74 static int ext4_statfs(struct dentry *dentry, struct kstatfs *buf);
75 static int ext4_unfreeze(struct super_block *sb);
76 static int ext4_freeze(struct super_block *sb);
77 static struct dentry *ext4_mount(struct file_system_type *fs_type, int flags,
78 const char *dev_name, void *data);
79 static inline int ext2_feature_set_ok(struct super_block *sb);
80 static inline int ext3_feature_set_ok(struct super_block *sb);
81 static int ext4_feature_set_ok(struct super_block *sb, int readonly);
82 static void ext4_destroy_lazyinit_thread(void);
83 static void ext4_unregister_li_request(struct super_block *sb);
84 static void ext4_clear_request_list(void);
85 static int ext4_reserve_clusters(struct ext4_sb_info *, ext4_fsblk_t);
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 MODULE_ALIAS_FS("ext2");
97 #define IS_EXT2_SB(sb) ((sb)->s_bdev->bd_holder == &ext2_fs_type)
99 #define IS_EXT2_SB(sb) (0)
103 #if !defined(CONFIG_EXT3_FS) && !defined(CONFIG_EXT3_FS_MODULE) && defined(CONFIG_EXT4_USE_FOR_EXT23)
104 static struct file_system_type ext3_fs_type = {
105 .owner = THIS_MODULE,
108 .kill_sb = kill_block_super,
109 .fs_flags = FS_REQUIRES_DEV,
111 MODULE_ALIAS_FS("ext3");
112 MODULE_ALIAS("ext3");
113 #define IS_EXT3_SB(sb) ((sb)->s_bdev->bd_holder == &ext3_fs_type)
115 #define IS_EXT3_SB(sb) (0)
118 static int ext4_verify_csum_type(struct super_block *sb,
119 struct ext4_super_block *es)
121 if (!EXT4_HAS_RO_COMPAT_FEATURE(sb,
122 EXT4_FEATURE_RO_COMPAT_METADATA_CSUM))
125 return es->s_checksum_type == EXT4_CRC32C_CHKSUM;
128 static __le32 ext4_superblock_csum(struct super_block *sb,
129 struct ext4_super_block *es)
131 struct ext4_sb_info *sbi = EXT4_SB(sb);
132 int offset = offsetof(struct ext4_super_block, s_checksum);
135 csum = ext4_chksum(sbi, ~0, (char *)es, offset);
137 return cpu_to_le32(csum);
140 static int ext4_superblock_csum_verify(struct super_block *sb,
141 struct ext4_super_block *es)
143 if (!EXT4_HAS_RO_COMPAT_FEATURE(sb,
144 EXT4_FEATURE_RO_COMPAT_METADATA_CSUM))
147 return es->s_checksum == ext4_superblock_csum(sb, es);
150 void ext4_superblock_csum_set(struct super_block *sb)
152 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
154 if (!EXT4_HAS_RO_COMPAT_FEATURE(sb,
155 EXT4_FEATURE_RO_COMPAT_METADATA_CSUM))
158 es->s_checksum = ext4_superblock_csum(sb, es);
161 void *ext4_kvmalloc(size_t size, gfp_t flags)
165 ret = kmalloc(size, flags | __GFP_NOWARN);
167 ret = __vmalloc(size, flags, PAGE_KERNEL);
171 void *ext4_kvzalloc(size_t size, gfp_t flags)
175 ret = kzalloc(size, flags | __GFP_NOWARN);
177 ret = __vmalloc(size, flags | __GFP_ZERO, PAGE_KERNEL);
181 void ext4_kvfree(void *ptr)
183 if (is_vmalloc_addr(ptr))
190 ext4_fsblk_t ext4_block_bitmap(struct super_block *sb,
191 struct ext4_group_desc *bg)
193 return le32_to_cpu(bg->bg_block_bitmap_lo) |
194 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
195 (ext4_fsblk_t)le32_to_cpu(bg->bg_block_bitmap_hi) << 32 : 0);
198 ext4_fsblk_t ext4_inode_bitmap(struct super_block *sb,
199 struct ext4_group_desc *bg)
201 return le32_to_cpu(bg->bg_inode_bitmap_lo) |
202 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
203 (ext4_fsblk_t)le32_to_cpu(bg->bg_inode_bitmap_hi) << 32 : 0);
206 ext4_fsblk_t ext4_inode_table(struct super_block *sb,
207 struct ext4_group_desc *bg)
209 return le32_to_cpu(bg->bg_inode_table_lo) |
210 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
211 (ext4_fsblk_t)le32_to_cpu(bg->bg_inode_table_hi) << 32 : 0);
214 __u32 ext4_free_group_clusters(struct super_block *sb,
215 struct ext4_group_desc *bg)
217 return le16_to_cpu(bg->bg_free_blocks_count_lo) |
218 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
219 (__u32)le16_to_cpu(bg->bg_free_blocks_count_hi) << 16 : 0);
222 __u32 ext4_free_inodes_count(struct super_block *sb,
223 struct ext4_group_desc *bg)
225 return le16_to_cpu(bg->bg_free_inodes_count_lo) |
226 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
227 (__u32)le16_to_cpu(bg->bg_free_inodes_count_hi) << 16 : 0);
230 __u32 ext4_used_dirs_count(struct super_block *sb,
231 struct ext4_group_desc *bg)
233 return le16_to_cpu(bg->bg_used_dirs_count_lo) |
234 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
235 (__u32)le16_to_cpu(bg->bg_used_dirs_count_hi) << 16 : 0);
238 __u32 ext4_itable_unused_count(struct super_block *sb,
239 struct ext4_group_desc *bg)
241 return le16_to_cpu(bg->bg_itable_unused_lo) |
242 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
243 (__u32)le16_to_cpu(bg->bg_itable_unused_hi) << 16 : 0);
246 void ext4_block_bitmap_set(struct super_block *sb,
247 struct ext4_group_desc *bg, ext4_fsblk_t blk)
249 bg->bg_block_bitmap_lo = cpu_to_le32((u32)blk);
250 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
251 bg->bg_block_bitmap_hi = cpu_to_le32(blk >> 32);
254 void ext4_inode_bitmap_set(struct super_block *sb,
255 struct ext4_group_desc *bg, ext4_fsblk_t blk)
257 bg->bg_inode_bitmap_lo = cpu_to_le32((u32)blk);
258 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
259 bg->bg_inode_bitmap_hi = cpu_to_le32(blk >> 32);
262 void ext4_inode_table_set(struct super_block *sb,
263 struct ext4_group_desc *bg, ext4_fsblk_t blk)
265 bg->bg_inode_table_lo = cpu_to_le32((u32)blk);
266 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
267 bg->bg_inode_table_hi = cpu_to_le32(blk >> 32);
270 void ext4_free_group_clusters_set(struct super_block *sb,
271 struct ext4_group_desc *bg, __u32 count)
273 bg->bg_free_blocks_count_lo = cpu_to_le16((__u16)count);
274 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
275 bg->bg_free_blocks_count_hi = cpu_to_le16(count >> 16);
278 void ext4_free_inodes_set(struct super_block *sb,
279 struct ext4_group_desc *bg, __u32 count)
281 bg->bg_free_inodes_count_lo = cpu_to_le16((__u16)count);
282 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
283 bg->bg_free_inodes_count_hi = cpu_to_le16(count >> 16);
286 void ext4_used_dirs_set(struct super_block *sb,
287 struct ext4_group_desc *bg, __u32 count)
289 bg->bg_used_dirs_count_lo = cpu_to_le16((__u16)count);
290 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
291 bg->bg_used_dirs_count_hi = cpu_to_le16(count >> 16);
294 void ext4_itable_unused_set(struct super_block *sb,
295 struct ext4_group_desc *bg, __u32 count)
297 bg->bg_itable_unused_lo = cpu_to_le16((__u16)count);
298 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
299 bg->bg_itable_unused_hi = cpu_to_le16(count >> 16);
303 static void __save_error_info(struct super_block *sb, const char *func,
306 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
308 EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS;
309 es->s_state |= cpu_to_le16(EXT4_ERROR_FS);
310 es->s_last_error_time = cpu_to_le32(get_seconds());
311 strncpy(es->s_last_error_func, func, sizeof(es->s_last_error_func));
312 es->s_last_error_line = cpu_to_le32(line);
313 if (!es->s_first_error_time) {
314 es->s_first_error_time = es->s_last_error_time;
315 strncpy(es->s_first_error_func, func,
316 sizeof(es->s_first_error_func));
317 es->s_first_error_line = cpu_to_le32(line);
318 es->s_first_error_ino = es->s_last_error_ino;
319 es->s_first_error_block = es->s_last_error_block;
322 * Start the daily error reporting function if it hasn't been
325 if (!es->s_error_count)
326 mod_timer(&EXT4_SB(sb)->s_err_report, jiffies + 24*60*60*HZ);
327 le32_add_cpu(&es->s_error_count, 1);
330 static void save_error_info(struct super_block *sb, const char *func,
333 __save_error_info(sb, func, line);
334 ext4_commit_super(sb, 1);
338 * The del_gendisk() function uninitializes the disk-specific data
339 * structures, including the bdi structure, without telling anyone
340 * else. Once this happens, any attempt to call mark_buffer_dirty()
341 * (for example, by ext4_commit_super), will cause a kernel OOPS.
342 * This is a kludge to prevent these oops until we can put in a proper
343 * hook in del_gendisk() to inform the VFS and file system layers.
345 static int block_device_ejected(struct super_block *sb)
347 struct inode *bd_inode = sb->s_bdev->bd_inode;
348 struct backing_dev_info *bdi = bd_inode->i_mapping->backing_dev_info;
350 return bdi->dev == NULL;
353 static void ext4_journal_commit_callback(journal_t *journal, transaction_t *txn)
355 struct super_block *sb = journal->j_private;
356 struct ext4_sb_info *sbi = EXT4_SB(sb);
357 int error = is_journal_aborted(journal);
358 struct ext4_journal_cb_entry *jce;
360 BUG_ON(txn->t_state == T_FINISHED);
361 spin_lock(&sbi->s_md_lock);
362 while (!list_empty(&txn->t_private_list)) {
363 jce = list_entry(txn->t_private_list.next,
364 struct ext4_journal_cb_entry, jce_list);
365 list_del_init(&jce->jce_list);
366 spin_unlock(&sbi->s_md_lock);
367 jce->jce_func(sb, jce, error);
368 spin_lock(&sbi->s_md_lock);
370 spin_unlock(&sbi->s_md_lock);
373 /* Deal with the reporting of failure conditions on a filesystem such as
374 * inconsistencies detected or read IO failures.
376 * On ext2, we can store the error state of the filesystem in the
377 * superblock. That is not possible on ext4, because we may have other
378 * write ordering constraints on the superblock which prevent us from
379 * writing it out straight away; and given that the journal is about to
380 * be aborted, we can't rely on the current, or future, transactions to
381 * write out the superblock safely.
383 * We'll just use the jbd2_journal_abort() error code to record an error in
384 * the journal instead. On recovery, the journal will complain about
385 * that error until we've noted it down and cleared it.
388 static void ext4_handle_error(struct super_block *sb)
390 if (sb->s_flags & MS_RDONLY)
393 if (!test_opt(sb, ERRORS_CONT)) {
394 journal_t *journal = EXT4_SB(sb)->s_journal;
396 EXT4_SB(sb)->s_mount_flags |= EXT4_MF_FS_ABORTED;
398 jbd2_journal_abort(journal, -EIO);
400 if (test_opt(sb, ERRORS_RO)) {
401 ext4_msg(sb, KERN_CRIT, "Remounting filesystem read-only");
403 * Make sure updated value of ->s_mount_flags will be visible
404 * before ->s_flags update
407 sb->s_flags |= MS_RDONLY;
409 if (test_opt(sb, ERRORS_PANIC))
410 panic("EXT4-fs (device %s): panic forced after error\n",
414 #define ext4_error_ratelimit(sb) \
415 ___ratelimit(&(EXT4_SB(sb)->s_err_ratelimit_state), \
418 void __ext4_error(struct super_block *sb, const char *function,
419 unsigned int line, const char *fmt, ...)
421 struct va_format vaf;
424 if (ext4_error_ratelimit(sb)) {
429 "EXT4-fs error (device %s): %s:%d: comm %s: %pV\n",
430 sb->s_id, function, line, current->comm, &vaf);
433 save_error_info(sb, function, line);
434 ext4_handle_error(sb);
437 void __ext4_error_inode(struct inode *inode, const char *function,
438 unsigned int line, ext4_fsblk_t block,
439 const char *fmt, ...)
442 struct va_format vaf;
443 struct ext4_super_block *es = EXT4_SB(inode->i_sb)->s_es;
445 es->s_last_error_ino = cpu_to_le32(inode->i_ino);
446 es->s_last_error_block = cpu_to_le64(block);
447 if (ext4_error_ratelimit(inode->i_sb)) {
452 printk(KERN_CRIT "EXT4-fs error (device %s): %s:%d: "
453 "inode #%lu: block %llu: comm %s: %pV\n",
454 inode->i_sb->s_id, function, line, inode->i_ino,
455 block, current->comm, &vaf);
457 printk(KERN_CRIT "EXT4-fs error (device %s): %s:%d: "
458 "inode #%lu: comm %s: %pV\n",
459 inode->i_sb->s_id, function, line, inode->i_ino,
460 current->comm, &vaf);
463 save_error_info(inode->i_sb, function, line);
464 ext4_handle_error(inode->i_sb);
467 void __ext4_error_file(struct file *file, const char *function,
468 unsigned int line, ext4_fsblk_t block,
469 const char *fmt, ...)
472 struct va_format vaf;
473 struct ext4_super_block *es;
474 struct inode *inode = file_inode(file);
475 char pathname[80], *path;
477 es = EXT4_SB(inode->i_sb)->s_es;
478 es->s_last_error_ino = cpu_to_le32(inode->i_ino);
479 if (ext4_error_ratelimit(inode->i_sb)) {
480 path = d_path(&(file->f_path), pathname, sizeof(pathname));
488 "EXT4-fs error (device %s): %s:%d: inode #%lu: "
489 "block %llu: comm %s: path %s: %pV\n",
490 inode->i_sb->s_id, function, line, inode->i_ino,
491 block, current->comm, path, &vaf);
494 "EXT4-fs error (device %s): %s:%d: inode #%lu: "
495 "comm %s: path %s: %pV\n",
496 inode->i_sb->s_id, function, line, inode->i_ino,
497 current->comm, path, &vaf);
500 save_error_info(inode->i_sb, function, line);
501 ext4_handle_error(inode->i_sb);
504 const char *ext4_decode_error(struct super_block *sb, int errno,
511 errstr = "IO failure";
514 errstr = "Out of memory";
517 if (!sb || (EXT4_SB(sb)->s_journal &&
518 EXT4_SB(sb)->s_journal->j_flags & JBD2_ABORT))
519 errstr = "Journal has aborted";
521 errstr = "Readonly filesystem";
524 /* If the caller passed in an extra buffer for unknown
525 * errors, textualise them now. Else we just return
528 /* Check for truncated error codes... */
529 if (snprintf(nbuf, 16, "error %d", -errno) >= 0)
538 /* __ext4_std_error decodes expected errors from journaling functions
539 * automatically and invokes the appropriate error response. */
541 void __ext4_std_error(struct super_block *sb, const char *function,
542 unsigned int line, int errno)
547 /* Special case: if the error is EROFS, and we're not already
548 * inside a transaction, then there's really no point in logging
550 if (errno == -EROFS && journal_current_handle() == NULL &&
551 (sb->s_flags & MS_RDONLY))
554 if (ext4_error_ratelimit(sb)) {
555 errstr = ext4_decode_error(sb, errno, nbuf);
556 printk(KERN_CRIT "EXT4-fs error (device %s) in %s:%d: %s\n",
557 sb->s_id, function, line, errstr);
560 save_error_info(sb, function, line);
561 ext4_handle_error(sb);
565 * ext4_abort is a much stronger failure handler than ext4_error. The
566 * abort function may be used to deal with unrecoverable failures such
567 * as journal IO errors or ENOMEM at a critical moment in log management.
569 * We unconditionally force the filesystem into an ABORT|READONLY state,
570 * unless the error response on the fs has been set to panic in which
571 * case we take the easy way out and panic immediately.
574 void __ext4_abort(struct super_block *sb, const char *function,
575 unsigned int line, const char *fmt, ...)
579 save_error_info(sb, function, line);
581 printk(KERN_CRIT "EXT4-fs error (device %s): %s:%d: ", sb->s_id,
587 if ((sb->s_flags & MS_RDONLY) == 0) {
588 ext4_msg(sb, KERN_CRIT, "Remounting filesystem read-only");
589 EXT4_SB(sb)->s_mount_flags |= EXT4_MF_FS_ABORTED;
591 * Make sure updated value of ->s_mount_flags will be visible
592 * before ->s_flags update
595 sb->s_flags |= MS_RDONLY;
596 if (EXT4_SB(sb)->s_journal)
597 jbd2_journal_abort(EXT4_SB(sb)->s_journal, -EIO);
598 save_error_info(sb, function, line);
600 if (test_opt(sb, ERRORS_PANIC))
601 panic("EXT4-fs panic from previous error\n");
604 void __ext4_msg(struct super_block *sb,
605 const char *prefix, const char *fmt, ...)
607 struct va_format vaf;
610 if (!___ratelimit(&(EXT4_SB(sb)->s_msg_ratelimit_state), "EXT4-fs"))
616 printk("%sEXT4-fs (%s): %pV\n", prefix, sb->s_id, &vaf);
620 void __ext4_warning(struct super_block *sb, const char *function,
621 unsigned int line, const char *fmt, ...)
623 struct va_format vaf;
626 if (!___ratelimit(&(EXT4_SB(sb)->s_warning_ratelimit_state),
633 printk(KERN_WARNING "EXT4-fs warning (device %s): %s:%d: %pV\n",
634 sb->s_id, function, line, &vaf);
638 void __ext4_grp_locked_error(const char *function, unsigned int line,
639 struct super_block *sb, ext4_group_t grp,
640 unsigned long ino, ext4_fsblk_t block,
641 const char *fmt, ...)
645 struct va_format vaf;
647 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
649 es->s_last_error_ino = cpu_to_le32(ino);
650 es->s_last_error_block = cpu_to_le64(block);
651 __save_error_info(sb, function, line);
653 if (ext4_error_ratelimit(sb)) {
657 printk(KERN_CRIT "EXT4-fs error (device %s): %s:%d: group %u, ",
658 sb->s_id, function, line, grp);
660 printk(KERN_CONT "inode %lu: ", ino);
662 printk(KERN_CONT "block %llu:",
663 (unsigned long long) block);
664 printk(KERN_CONT "%pV\n", &vaf);
668 if (test_opt(sb, ERRORS_CONT)) {
669 ext4_commit_super(sb, 0);
673 ext4_unlock_group(sb, grp);
674 ext4_handle_error(sb);
676 * We only get here in the ERRORS_RO case; relocking the group
677 * may be dangerous, but nothing bad will happen since the
678 * filesystem will have already been marked read/only and the
679 * journal has been aborted. We return 1 as a hint to callers
680 * who might what to use the return value from
681 * ext4_grp_locked_error() to distinguish between the
682 * ERRORS_CONT and ERRORS_RO case, and perhaps return more
683 * aggressively from the ext4 function in question, with a
684 * more appropriate error code.
686 ext4_lock_group(sb, grp);
690 void ext4_update_dynamic_rev(struct super_block *sb)
692 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
694 if (le32_to_cpu(es->s_rev_level) > EXT4_GOOD_OLD_REV)
698 "updating to rev %d because of new feature flag, "
699 "running e2fsck is recommended",
702 es->s_first_ino = cpu_to_le32(EXT4_GOOD_OLD_FIRST_INO);
703 es->s_inode_size = cpu_to_le16(EXT4_GOOD_OLD_INODE_SIZE);
704 es->s_rev_level = cpu_to_le32(EXT4_DYNAMIC_REV);
705 /* leave es->s_feature_*compat flags alone */
706 /* es->s_uuid will be set by e2fsck if empty */
709 * The rest of the superblock fields should be zero, and if not it
710 * means they are likely already in use, so leave them alone. We
711 * can leave it up to e2fsck to clean up any inconsistencies there.
716 * Open the external journal device
718 static struct block_device *ext4_blkdev_get(dev_t dev, struct super_block *sb)
720 struct block_device *bdev;
721 char b[BDEVNAME_SIZE];
723 bdev = blkdev_get_by_dev(dev, FMODE_READ|FMODE_WRITE|FMODE_EXCL, sb);
729 ext4_msg(sb, KERN_ERR, "failed to open journal device %s: %ld",
730 __bdevname(dev, b), PTR_ERR(bdev));
735 * Release the journal device
737 static void ext4_blkdev_put(struct block_device *bdev)
739 blkdev_put(bdev, FMODE_READ|FMODE_WRITE|FMODE_EXCL);
742 static void ext4_blkdev_remove(struct ext4_sb_info *sbi)
744 struct block_device *bdev;
745 bdev = sbi->journal_bdev;
747 ext4_blkdev_put(bdev);
748 sbi->journal_bdev = NULL;
752 static inline struct inode *orphan_list_entry(struct list_head *l)
754 return &list_entry(l, struct ext4_inode_info, i_orphan)->vfs_inode;
757 static void dump_orphan_list(struct super_block *sb, struct ext4_sb_info *sbi)
761 ext4_msg(sb, KERN_ERR, "sb orphan head is %d",
762 le32_to_cpu(sbi->s_es->s_last_orphan));
764 printk(KERN_ERR "sb_info orphan list:\n");
765 list_for_each(l, &sbi->s_orphan) {
766 struct inode *inode = orphan_list_entry(l);
768 "inode %s:%lu at %p: mode %o, nlink %d, next %d\n",
769 inode->i_sb->s_id, inode->i_ino, inode,
770 inode->i_mode, inode->i_nlink,
775 static void ext4_put_super(struct super_block *sb)
777 struct ext4_sb_info *sbi = EXT4_SB(sb);
778 struct ext4_super_block *es = sbi->s_es;
781 ext4_unregister_li_request(sb);
782 dquot_disable(sb, -1, DQUOT_USAGE_ENABLED | DQUOT_LIMITS_ENABLED);
784 flush_workqueue(sbi->rsv_conversion_wq);
785 destroy_workqueue(sbi->rsv_conversion_wq);
787 if (sbi->s_journal) {
788 err = jbd2_journal_destroy(sbi->s_journal);
789 sbi->s_journal = NULL;
791 ext4_abort(sb, "Couldn't clean up the journal");
794 ext4_es_unregister_shrinker(sbi);
795 del_timer_sync(&sbi->s_err_report);
796 ext4_release_system_zone(sb);
798 ext4_ext_release(sb);
799 ext4_xattr_put_super(sb);
801 if (!(sb->s_flags & MS_RDONLY)) {
802 EXT4_CLEAR_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
803 es->s_state = cpu_to_le16(sbi->s_mount_state);
805 if (!(sb->s_flags & MS_RDONLY))
806 ext4_commit_super(sb, 1);
809 remove_proc_entry("options", sbi->s_proc);
810 remove_proc_entry(sb->s_id, ext4_proc_root);
812 kobject_del(&sbi->s_kobj);
814 for (i = 0; i < sbi->s_gdb_count; i++)
815 brelse(sbi->s_group_desc[i]);
816 ext4_kvfree(sbi->s_group_desc);
817 ext4_kvfree(sbi->s_flex_groups);
818 percpu_counter_destroy(&sbi->s_freeclusters_counter);
819 percpu_counter_destroy(&sbi->s_freeinodes_counter);
820 percpu_counter_destroy(&sbi->s_dirs_counter);
821 percpu_counter_destroy(&sbi->s_dirtyclusters_counter);
824 for (i = 0; i < EXT4_MAXQUOTAS; i++)
825 kfree(sbi->s_qf_names[i]);
828 /* Debugging code just in case the in-memory inode orphan list
829 * isn't empty. The on-disk one can be non-empty if we've
830 * detected an error and taken the fs readonly, but the
831 * in-memory list had better be clean by this point. */
832 if (!list_empty(&sbi->s_orphan))
833 dump_orphan_list(sb, sbi);
834 J_ASSERT(list_empty(&sbi->s_orphan));
836 invalidate_bdev(sb->s_bdev);
837 if (sbi->journal_bdev && sbi->journal_bdev != sb->s_bdev) {
839 * Invalidate the journal device's buffers. We don't want them
840 * floating about in memory - the physical journal device may
841 * hotswapped, and it breaks the `ro-after' testing code.
843 sync_blockdev(sbi->journal_bdev);
844 invalidate_bdev(sbi->journal_bdev);
845 ext4_blkdev_remove(sbi);
847 if (sbi->s_mb_cache) {
848 ext4_xattr_destroy_cache(sbi->s_mb_cache);
849 sbi->s_mb_cache = NULL;
852 kthread_stop(sbi->s_mmp_tsk);
853 sb->s_fs_info = NULL;
855 * Now that we are completely done shutting down the
856 * superblock, we need to actually destroy the kobject.
858 kobject_put(&sbi->s_kobj);
859 wait_for_completion(&sbi->s_kobj_unregister);
860 if (sbi->s_chksum_driver)
861 crypto_free_shash(sbi->s_chksum_driver);
862 kfree(sbi->s_blockgroup_lock);
866 static struct kmem_cache *ext4_inode_cachep;
869 * Called inside transaction, so use GFP_NOFS
871 static struct inode *ext4_alloc_inode(struct super_block *sb)
873 struct ext4_inode_info *ei;
875 ei = kmem_cache_alloc(ext4_inode_cachep, GFP_NOFS);
879 ei->vfs_inode.i_version = 1;
880 spin_lock_init(&ei->i_raw_lock);
881 INIT_LIST_HEAD(&ei->i_prealloc_list);
882 spin_lock_init(&ei->i_prealloc_lock);
883 ext4_es_init_tree(&ei->i_es_tree);
884 rwlock_init(&ei->i_es_lock);
885 INIT_LIST_HEAD(&ei->i_es_lru);
888 ei->i_touch_when = 0;
889 ei->i_reserved_data_blocks = 0;
890 ei->i_reserved_meta_blocks = 0;
891 ei->i_allocated_meta_blocks = 0;
892 ei->i_da_metadata_calc_len = 0;
893 ei->i_da_metadata_calc_last_lblock = 0;
894 spin_lock_init(&(ei->i_block_reservation_lock));
896 ei->i_reserved_quota = 0;
899 INIT_LIST_HEAD(&ei->i_rsv_conversion_list);
900 spin_lock_init(&ei->i_completed_io_lock);
902 ei->i_datasync_tid = 0;
903 atomic_set(&ei->i_ioend_count, 0);
904 atomic_set(&ei->i_unwritten, 0);
905 INIT_WORK(&ei->i_rsv_conversion_work, ext4_end_io_rsv_work);
907 return &ei->vfs_inode;
910 static int ext4_drop_inode(struct inode *inode)
912 int drop = generic_drop_inode(inode);
914 trace_ext4_drop_inode(inode, drop);
918 static void ext4_i_callback(struct rcu_head *head)
920 struct inode *inode = container_of(head, struct inode, i_rcu);
921 kmem_cache_free(ext4_inode_cachep, EXT4_I(inode));
924 static void ext4_destroy_inode(struct inode *inode)
926 if (!list_empty(&(EXT4_I(inode)->i_orphan))) {
927 ext4_msg(inode->i_sb, KERN_ERR,
928 "Inode %lu (%p): orphan list check failed!",
929 inode->i_ino, EXT4_I(inode));
930 print_hex_dump(KERN_INFO, "", DUMP_PREFIX_ADDRESS, 16, 4,
931 EXT4_I(inode), sizeof(struct ext4_inode_info),
935 call_rcu(&inode->i_rcu, ext4_i_callback);
938 static void init_once(void *foo)
940 struct ext4_inode_info *ei = (struct ext4_inode_info *) foo;
942 INIT_LIST_HEAD(&ei->i_orphan);
943 init_rwsem(&ei->xattr_sem);
944 init_rwsem(&ei->i_data_sem);
945 inode_init_once(&ei->vfs_inode);
948 static int __init init_inodecache(void)
950 ext4_inode_cachep = kmem_cache_create("ext4_inode_cache",
951 sizeof(struct ext4_inode_info),
952 0, (SLAB_RECLAIM_ACCOUNT|
955 if (ext4_inode_cachep == NULL)
960 static void destroy_inodecache(void)
963 * Make sure all delayed rcu free inodes are flushed before we
967 kmem_cache_destroy(ext4_inode_cachep);
970 void ext4_clear_inode(struct inode *inode)
972 invalidate_inode_buffers(inode);
975 ext4_discard_preallocations(inode);
976 ext4_es_remove_extent(inode, 0, EXT_MAX_BLOCKS);
977 ext4_es_lru_del(inode);
978 if (EXT4_I(inode)->jinode) {
979 jbd2_journal_release_jbd_inode(EXT4_JOURNAL(inode),
980 EXT4_I(inode)->jinode);
981 jbd2_free_inode(EXT4_I(inode)->jinode);
982 EXT4_I(inode)->jinode = NULL;
986 static struct inode *ext4_nfs_get_inode(struct super_block *sb,
987 u64 ino, u32 generation)
991 if (ino < EXT4_FIRST_INO(sb) && ino != EXT4_ROOT_INO)
992 return ERR_PTR(-ESTALE);
993 if (ino > le32_to_cpu(EXT4_SB(sb)->s_es->s_inodes_count))
994 return ERR_PTR(-ESTALE);
996 /* iget isn't really right if the inode is currently unallocated!!
998 * ext4_read_inode will return a bad_inode if the inode had been
999 * deleted, so we should be safe.
1001 * Currently we don't know the generation for parent directory, so
1002 * a generation of 0 means "accept any"
1004 inode = ext4_iget(sb, ino);
1006 return ERR_CAST(inode);
1007 if (generation && inode->i_generation != generation) {
1009 return ERR_PTR(-ESTALE);
1015 static struct dentry *ext4_fh_to_dentry(struct super_block *sb, struct fid *fid,
1016 int fh_len, int fh_type)
1018 return generic_fh_to_dentry(sb, fid, fh_len, fh_type,
1019 ext4_nfs_get_inode);
1022 static struct dentry *ext4_fh_to_parent(struct super_block *sb, struct fid *fid,
1023 int fh_len, int fh_type)
1025 return generic_fh_to_parent(sb, fid, fh_len, fh_type,
1026 ext4_nfs_get_inode);
1030 * Try to release metadata pages (indirect blocks, directories) which are
1031 * mapped via the block device. Since these pages could have journal heads
1032 * which would prevent try_to_free_buffers() from freeing them, we must use
1033 * jbd2 layer's try_to_free_buffers() function to release them.
1035 static int bdev_try_to_free_page(struct super_block *sb, struct page *page,
1038 journal_t *journal = EXT4_SB(sb)->s_journal;
1040 WARN_ON(PageChecked(page));
1041 if (!page_has_buffers(page))
1044 return jbd2_journal_try_to_free_buffers(journal, page,
1045 wait & ~__GFP_WAIT);
1046 return try_to_free_buffers(page);
1050 #define QTYPE2NAME(t) ((t) == USRQUOTA ? "user" : "group")
1051 #define QTYPE2MOPT(on, t) ((t) == USRQUOTA?((on)##USRJQUOTA):((on)##GRPJQUOTA))
1053 static int ext4_write_dquot(struct dquot *dquot);
1054 static int ext4_acquire_dquot(struct dquot *dquot);
1055 static int ext4_release_dquot(struct dquot *dquot);
1056 static int ext4_mark_dquot_dirty(struct dquot *dquot);
1057 static int ext4_write_info(struct super_block *sb, int type);
1058 static int ext4_quota_on(struct super_block *sb, int type, int format_id,
1060 static int ext4_quota_on_sysfile(struct super_block *sb, int type,
1062 static int ext4_quota_off(struct super_block *sb, int type);
1063 static int ext4_quota_off_sysfile(struct super_block *sb, int type);
1064 static int ext4_quota_on_mount(struct super_block *sb, int type);
1065 static ssize_t ext4_quota_read(struct super_block *sb, int type, char *data,
1066 size_t len, loff_t off);
1067 static ssize_t ext4_quota_write(struct super_block *sb, int type,
1068 const char *data, size_t len, loff_t off);
1069 static int ext4_quota_enable(struct super_block *sb, int type, int format_id,
1070 unsigned int flags);
1071 static int ext4_enable_quotas(struct super_block *sb);
1073 static const struct dquot_operations ext4_quota_operations = {
1074 .get_reserved_space = ext4_get_reserved_space,
1075 .write_dquot = ext4_write_dquot,
1076 .acquire_dquot = ext4_acquire_dquot,
1077 .release_dquot = ext4_release_dquot,
1078 .mark_dirty = ext4_mark_dquot_dirty,
1079 .write_info = ext4_write_info,
1080 .alloc_dquot = dquot_alloc,
1081 .destroy_dquot = dquot_destroy,
1084 static const struct quotactl_ops ext4_qctl_operations = {
1085 .quota_on = ext4_quota_on,
1086 .quota_off = ext4_quota_off,
1087 .quota_sync = dquot_quota_sync,
1088 .get_info = dquot_get_dqinfo,
1089 .set_info = dquot_set_dqinfo,
1090 .get_dqblk = dquot_get_dqblk,
1091 .set_dqblk = dquot_set_dqblk
1094 static const struct quotactl_ops ext4_qctl_sysfile_operations = {
1095 .quota_on_meta = ext4_quota_on_sysfile,
1096 .quota_off = ext4_quota_off_sysfile,
1097 .quota_sync = dquot_quota_sync,
1098 .get_info = dquot_get_dqinfo,
1099 .set_info = dquot_set_dqinfo,
1100 .get_dqblk = dquot_get_dqblk,
1101 .set_dqblk = dquot_set_dqblk
1105 static const struct super_operations ext4_sops = {
1106 .alloc_inode = ext4_alloc_inode,
1107 .destroy_inode = ext4_destroy_inode,
1108 .write_inode = ext4_write_inode,
1109 .dirty_inode = ext4_dirty_inode,
1110 .drop_inode = ext4_drop_inode,
1111 .evict_inode = ext4_evict_inode,
1112 .put_super = ext4_put_super,
1113 .sync_fs = ext4_sync_fs,
1114 .freeze_fs = ext4_freeze,
1115 .unfreeze_fs = ext4_unfreeze,
1116 .statfs = ext4_statfs,
1117 .remount_fs = ext4_remount,
1118 .show_options = ext4_show_options,
1120 .quota_read = ext4_quota_read,
1121 .quota_write = ext4_quota_write,
1123 .bdev_try_to_free_page = bdev_try_to_free_page,
1126 static const struct super_operations ext4_nojournal_sops = {
1127 .alloc_inode = ext4_alloc_inode,
1128 .destroy_inode = ext4_destroy_inode,
1129 .write_inode = ext4_write_inode,
1130 .dirty_inode = ext4_dirty_inode,
1131 .drop_inode = ext4_drop_inode,
1132 .evict_inode = ext4_evict_inode,
1133 .sync_fs = ext4_sync_fs,
1134 .put_super = ext4_put_super,
1135 .statfs = ext4_statfs,
1136 .remount_fs = ext4_remount,
1137 .show_options = ext4_show_options,
1139 .quota_read = ext4_quota_read,
1140 .quota_write = ext4_quota_write,
1142 .bdev_try_to_free_page = bdev_try_to_free_page,
1145 static const struct export_operations ext4_export_ops = {
1146 .fh_to_dentry = ext4_fh_to_dentry,
1147 .fh_to_parent = ext4_fh_to_parent,
1148 .get_parent = ext4_get_parent,
1152 Opt_bsd_df, Opt_minix_df, Opt_grpid, Opt_nogrpid,
1153 Opt_resgid, Opt_resuid, Opt_sb, Opt_err_cont, Opt_err_panic, Opt_err_ro,
1154 Opt_nouid32, Opt_debug, Opt_removed,
1155 Opt_user_xattr, Opt_nouser_xattr, Opt_acl, Opt_noacl,
1156 Opt_auto_da_alloc, Opt_noauto_da_alloc, Opt_noload,
1157 Opt_commit, Opt_min_batch_time, Opt_max_batch_time, Opt_journal_dev,
1158 Opt_journal_path, Opt_journal_checksum, Opt_journal_async_commit,
1159 Opt_abort, Opt_data_journal, Opt_data_ordered, Opt_data_writeback,
1160 Opt_data_err_abort, Opt_data_err_ignore,
1161 Opt_usrjquota, Opt_grpjquota, Opt_offusrjquota, Opt_offgrpjquota,
1162 Opt_jqfmt_vfsold, Opt_jqfmt_vfsv0, Opt_jqfmt_vfsv1, Opt_quota,
1163 Opt_noquota, Opt_barrier, Opt_nobarrier, Opt_err,
1164 Opt_usrquota, Opt_grpquota, Opt_i_version,
1165 Opt_stripe, Opt_delalloc, Opt_nodelalloc, Opt_mblk_io_submit,
1166 Opt_nomblk_io_submit, Opt_block_validity, Opt_noblock_validity,
1167 Opt_inode_readahead_blks, Opt_journal_ioprio,
1168 Opt_dioread_nolock, Opt_dioread_lock,
1169 Opt_discard, Opt_nodiscard, Opt_init_itable, Opt_noinit_itable,
1170 Opt_max_dir_size_kb,
1173 static const match_table_t tokens = {
1174 {Opt_bsd_df, "bsddf"},
1175 {Opt_minix_df, "minixdf"},
1176 {Opt_grpid, "grpid"},
1177 {Opt_grpid, "bsdgroups"},
1178 {Opt_nogrpid, "nogrpid"},
1179 {Opt_nogrpid, "sysvgroups"},
1180 {Opt_resgid, "resgid=%u"},
1181 {Opt_resuid, "resuid=%u"},
1183 {Opt_err_cont, "errors=continue"},
1184 {Opt_err_panic, "errors=panic"},
1185 {Opt_err_ro, "errors=remount-ro"},
1186 {Opt_nouid32, "nouid32"},
1187 {Opt_debug, "debug"},
1188 {Opt_removed, "oldalloc"},
1189 {Opt_removed, "orlov"},
1190 {Opt_user_xattr, "user_xattr"},
1191 {Opt_nouser_xattr, "nouser_xattr"},
1193 {Opt_noacl, "noacl"},
1194 {Opt_noload, "norecovery"},
1195 {Opt_noload, "noload"},
1196 {Opt_removed, "nobh"},
1197 {Opt_removed, "bh"},
1198 {Opt_commit, "commit=%u"},
1199 {Opt_min_batch_time, "min_batch_time=%u"},
1200 {Opt_max_batch_time, "max_batch_time=%u"},
1201 {Opt_journal_dev, "journal_dev=%u"},
1202 {Opt_journal_path, "journal_path=%s"},
1203 {Opt_journal_checksum, "journal_checksum"},
1204 {Opt_journal_async_commit, "journal_async_commit"},
1205 {Opt_abort, "abort"},
1206 {Opt_data_journal, "data=journal"},
1207 {Opt_data_ordered, "data=ordered"},
1208 {Opt_data_writeback, "data=writeback"},
1209 {Opt_data_err_abort, "data_err=abort"},
1210 {Opt_data_err_ignore, "data_err=ignore"},
1211 {Opt_offusrjquota, "usrjquota="},
1212 {Opt_usrjquota, "usrjquota=%s"},
1213 {Opt_offgrpjquota, "grpjquota="},
1214 {Opt_grpjquota, "grpjquota=%s"},
1215 {Opt_jqfmt_vfsold, "jqfmt=vfsold"},
1216 {Opt_jqfmt_vfsv0, "jqfmt=vfsv0"},
1217 {Opt_jqfmt_vfsv1, "jqfmt=vfsv1"},
1218 {Opt_grpquota, "grpquota"},
1219 {Opt_noquota, "noquota"},
1220 {Opt_quota, "quota"},
1221 {Opt_usrquota, "usrquota"},
1222 {Opt_barrier, "barrier=%u"},
1223 {Opt_barrier, "barrier"},
1224 {Opt_nobarrier, "nobarrier"},
1225 {Opt_i_version, "i_version"},
1226 {Opt_stripe, "stripe=%u"},
1227 {Opt_delalloc, "delalloc"},
1228 {Opt_nodelalloc, "nodelalloc"},
1229 {Opt_removed, "mblk_io_submit"},
1230 {Opt_removed, "nomblk_io_submit"},
1231 {Opt_block_validity, "block_validity"},
1232 {Opt_noblock_validity, "noblock_validity"},
1233 {Opt_inode_readahead_blks, "inode_readahead_blks=%u"},
1234 {Opt_journal_ioprio, "journal_ioprio=%u"},
1235 {Opt_auto_da_alloc, "auto_da_alloc=%u"},
1236 {Opt_auto_da_alloc, "auto_da_alloc"},
1237 {Opt_noauto_da_alloc, "noauto_da_alloc"},
1238 {Opt_dioread_nolock, "dioread_nolock"},
1239 {Opt_dioread_lock, "dioread_lock"},
1240 {Opt_discard, "discard"},
1241 {Opt_nodiscard, "nodiscard"},
1242 {Opt_init_itable, "init_itable=%u"},
1243 {Opt_init_itable, "init_itable"},
1244 {Opt_noinit_itable, "noinit_itable"},
1245 {Opt_max_dir_size_kb, "max_dir_size_kb=%u"},
1246 {Opt_removed, "check=none"}, /* mount option from ext2/3 */
1247 {Opt_removed, "nocheck"}, /* mount option from ext2/3 */
1248 {Opt_removed, "reservation"}, /* mount option from ext2/3 */
1249 {Opt_removed, "noreservation"}, /* mount option from ext2/3 */
1250 {Opt_removed, "journal=%u"}, /* mount option from ext2/3 */
1254 static ext4_fsblk_t get_sb_block(void **data)
1256 ext4_fsblk_t sb_block;
1257 char *options = (char *) *data;
1259 if (!options || strncmp(options, "sb=", 3) != 0)
1260 return 1; /* Default location */
1263 /* TODO: use simple_strtoll with >32bit ext4 */
1264 sb_block = simple_strtoul(options, &options, 0);
1265 if (*options && *options != ',') {
1266 printk(KERN_ERR "EXT4-fs: Invalid sb specification: %s\n",
1270 if (*options == ',')
1272 *data = (void *) options;
1277 #define DEFAULT_JOURNAL_IOPRIO (IOPRIO_PRIO_VALUE(IOPRIO_CLASS_BE, 3))
1278 static char deprecated_msg[] = "Mount option \"%s\" will be removed by %s\n"
1279 "Contact linux-ext4@vger.kernel.org if you think we should keep it.\n";
1282 static int set_qf_name(struct super_block *sb, int qtype, substring_t *args)
1284 struct ext4_sb_info *sbi = EXT4_SB(sb);
1288 if (sb_any_quota_loaded(sb) &&
1289 !sbi->s_qf_names[qtype]) {
1290 ext4_msg(sb, KERN_ERR,
1291 "Cannot change journaled "
1292 "quota options when quota turned on");
1295 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_QUOTA)) {
1296 ext4_msg(sb, KERN_ERR, "Cannot set journaled quota options "
1297 "when QUOTA feature is enabled");
1300 qname = match_strdup(args);
1302 ext4_msg(sb, KERN_ERR,
1303 "Not enough memory for storing quotafile name");
1306 if (sbi->s_qf_names[qtype]) {
1307 if (strcmp(sbi->s_qf_names[qtype], qname) == 0)
1310 ext4_msg(sb, KERN_ERR,
1311 "%s quota file already specified",
1315 if (strchr(qname, '/')) {
1316 ext4_msg(sb, KERN_ERR,
1317 "quotafile must be on filesystem root");
1320 sbi->s_qf_names[qtype] = qname;
1328 static int clear_qf_name(struct super_block *sb, int qtype)
1331 struct ext4_sb_info *sbi = EXT4_SB(sb);
1333 if (sb_any_quota_loaded(sb) &&
1334 sbi->s_qf_names[qtype]) {
1335 ext4_msg(sb, KERN_ERR, "Cannot change journaled quota options"
1336 " when quota turned on");
1339 kfree(sbi->s_qf_names[qtype]);
1340 sbi->s_qf_names[qtype] = NULL;
1345 #define MOPT_SET 0x0001
1346 #define MOPT_CLEAR 0x0002
1347 #define MOPT_NOSUPPORT 0x0004
1348 #define MOPT_EXPLICIT 0x0008
1349 #define MOPT_CLEAR_ERR 0x0010
1350 #define MOPT_GTE0 0x0020
1353 #define MOPT_QFMT 0x0040
1355 #define MOPT_Q MOPT_NOSUPPORT
1356 #define MOPT_QFMT MOPT_NOSUPPORT
1358 #define MOPT_DATAJ 0x0080
1359 #define MOPT_NO_EXT2 0x0100
1360 #define MOPT_NO_EXT3 0x0200
1361 #define MOPT_EXT4_ONLY (MOPT_NO_EXT2 | MOPT_NO_EXT3)
1362 #define MOPT_STRING 0x0400
1364 static const struct mount_opts {
1368 } ext4_mount_opts[] = {
1369 {Opt_minix_df, EXT4_MOUNT_MINIX_DF, MOPT_SET},
1370 {Opt_bsd_df, EXT4_MOUNT_MINIX_DF, MOPT_CLEAR},
1371 {Opt_grpid, EXT4_MOUNT_GRPID, MOPT_SET},
1372 {Opt_nogrpid, EXT4_MOUNT_GRPID, MOPT_CLEAR},
1373 {Opt_block_validity, EXT4_MOUNT_BLOCK_VALIDITY, MOPT_SET},
1374 {Opt_noblock_validity, EXT4_MOUNT_BLOCK_VALIDITY, MOPT_CLEAR},
1375 {Opt_dioread_nolock, EXT4_MOUNT_DIOREAD_NOLOCK,
1376 MOPT_EXT4_ONLY | MOPT_SET},
1377 {Opt_dioread_lock, EXT4_MOUNT_DIOREAD_NOLOCK,
1378 MOPT_EXT4_ONLY | MOPT_CLEAR},
1379 {Opt_discard, EXT4_MOUNT_DISCARD, MOPT_SET},
1380 {Opt_nodiscard, EXT4_MOUNT_DISCARD, MOPT_CLEAR},
1381 {Opt_delalloc, EXT4_MOUNT_DELALLOC,
1382 MOPT_EXT4_ONLY | MOPT_SET | MOPT_EXPLICIT},
1383 {Opt_nodelalloc, EXT4_MOUNT_DELALLOC,
1384 MOPT_EXT4_ONLY | MOPT_CLEAR},
1385 {Opt_journal_checksum, EXT4_MOUNT_JOURNAL_CHECKSUM,
1386 MOPT_EXT4_ONLY | MOPT_SET},
1387 {Opt_journal_async_commit, (EXT4_MOUNT_JOURNAL_ASYNC_COMMIT |
1388 EXT4_MOUNT_JOURNAL_CHECKSUM),
1389 MOPT_EXT4_ONLY | MOPT_SET},
1390 {Opt_noload, EXT4_MOUNT_NOLOAD, MOPT_NO_EXT2 | MOPT_SET},
1391 {Opt_err_panic, EXT4_MOUNT_ERRORS_PANIC, MOPT_SET | MOPT_CLEAR_ERR},
1392 {Opt_err_ro, EXT4_MOUNT_ERRORS_RO, MOPT_SET | MOPT_CLEAR_ERR},
1393 {Opt_err_cont, EXT4_MOUNT_ERRORS_CONT, MOPT_SET | MOPT_CLEAR_ERR},
1394 {Opt_data_err_abort, EXT4_MOUNT_DATA_ERR_ABORT,
1395 MOPT_NO_EXT2 | MOPT_SET},
1396 {Opt_data_err_ignore, EXT4_MOUNT_DATA_ERR_ABORT,
1397 MOPT_NO_EXT2 | MOPT_CLEAR},
1398 {Opt_barrier, EXT4_MOUNT_BARRIER, MOPT_SET},
1399 {Opt_nobarrier, EXT4_MOUNT_BARRIER, MOPT_CLEAR},
1400 {Opt_noauto_da_alloc, EXT4_MOUNT_NO_AUTO_DA_ALLOC, MOPT_SET},
1401 {Opt_auto_da_alloc, EXT4_MOUNT_NO_AUTO_DA_ALLOC, MOPT_CLEAR},
1402 {Opt_noinit_itable, EXT4_MOUNT_INIT_INODE_TABLE, MOPT_CLEAR},
1403 {Opt_commit, 0, MOPT_GTE0},
1404 {Opt_max_batch_time, 0, MOPT_GTE0},
1405 {Opt_min_batch_time, 0, MOPT_GTE0},
1406 {Opt_inode_readahead_blks, 0, MOPT_GTE0},
1407 {Opt_init_itable, 0, MOPT_GTE0},
1408 {Opt_stripe, 0, MOPT_GTE0},
1409 {Opt_resuid, 0, MOPT_GTE0},
1410 {Opt_resgid, 0, MOPT_GTE0},
1411 {Opt_journal_dev, 0, MOPT_GTE0},
1412 {Opt_journal_path, 0, MOPT_STRING},
1413 {Opt_journal_ioprio, 0, MOPT_GTE0},
1414 {Opt_data_journal, EXT4_MOUNT_JOURNAL_DATA, MOPT_NO_EXT2 | MOPT_DATAJ},
1415 {Opt_data_ordered, EXT4_MOUNT_ORDERED_DATA, MOPT_NO_EXT2 | MOPT_DATAJ},
1416 {Opt_data_writeback, EXT4_MOUNT_WRITEBACK_DATA,
1417 MOPT_NO_EXT2 | MOPT_DATAJ},
1418 {Opt_user_xattr, EXT4_MOUNT_XATTR_USER, MOPT_SET},
1419 {Opt_nouser_xattr, EXT4_MOUNT_XATTR_USER, MOPT_CLEAR},
1420 #ifdef CONFIG_EXT4_FS_POSIX_ACL
1421 {Opt_acl, EXT4_MOUNT_POSIX_ACL, MOPT_SET},
1422 {Opt_noacl, EXT4_MOUNT_POSIX_ACL, MOPT_CLEAR},
1424 {Opt_acl, 0, MOPT_NOSUPPORT},
1425 {Opt_noacl, 0, MOPT_NOSUPPORT},
1427 {Opt_nouid32, EXT4_MOUNT_NO_UID32, MOPT_SET},
1428 {Opt_debug, EXT4_MOUNT_DEBUG, MOPT_SET},
1429 {Opt_quota, EXT4_MOUNT_QUOTA | EXT4_MOUNT_USRQUOTA, MOPT_SET | MOPT_Q},
1430 {Opt_usrquota, EXT4_MOUNT_QUOTA | EXT4_MOUNT_USRQUOTA,
1432 {Opt_grpquota, EXT4_MOUNT_QUOTA | EXT4_MOUNT_GRPQUOTA,
1434 {Opt_noquota, (EXT4_MOUNT_QUOTA | EXT4_MOUNT_USRQUOTA |
1435 EXT4_MOUNT_GRPQUOTA), MOPT_CLEAR | MOPT_Q},
1436 {Opt_usrjquota, 0, MOPT_Q},
1437 {Opt_grpjquota, 0, MOPT_Q},
1438 {Opt_offusrjquota, 0, MOPT_Q},
1439 {Opt_offgrpjquota, 0, MOPT_Q},
1440 {Opt_jqfmt_vfsold, QFMT_VFS_OLD, MOPT_QFMT},
1441 {Opt_jqfmt_vfsv0, QFMT_VFS_V0, MOPT_QFMT},
1442 {Opt_jqfmt_vfsv1, QFMT_VFS_V1, MOPT_QFMT},
1443 {Opt_max_dir_size_kb, 0, MOPT_GTE0},
1447 static int handle_mount_opt(struct super_block *sb, char *opt, int token,
1448 substring_t *args, unsigned long *journal_devnum,
1449 unsigned int *journal_ioprio, int is_remount)
1451 struct ext4_sb_info *sbi = EXT4_SB(sb);
1452 const struct mount_opts *m;
1458 if (token == Opt_usrjquota)
1459 return set_qf_name(sb, USRQUOTA, &args[0]);
1460 else if (token == Opt_grpjquota)
1461 return set_qf_name(sb, GRPQUOTA, &args[0]);
1462 else if (token == Opt_offusrjquota)
1463 return clear_qf_name(sb, USRQUOTA);
1464 else if (token == Opt_offgrpjquota)
1465 return clear_qf_name(sb, GRPQUOTA);
1469 case Opt_nouser_xattr:
1470 ext4_msg(sb, KERN_WARNING, deprecated_msg, opt, "3.5");
1473 return 1; /* handled by get_sb_block() */
1475 ext4_msg(sb, KERN_WARNING, "Ignoring removed %s option", opt);
1478 sbi->s_mount_flags |= EXT4_MF_FS_ABORTED;
1481 sb->s_flags |= MS_I_VERSION;
1485 for (m = ext4_mount_opts; m->token != Opt_err; m++)
1486 if (token == m->token)
1489 if (m->token == Opt_err) {
1490 ext4_msg(sb, KERN_ERR, "Unrecognized mount option \"%s\" "
1491 "or missing value", opt);
1495 if ((m->flags & MOPT_NO_EXT2) && IS_EXT2_SB(sb)) {
1496 ext4_msg(sb, KERN_ERR,
1497 "Mount option \"%s\" incompatible with ext2", opt);
1500 if ((m->flags & MOPT_NO_EXT3) && IS_EXT3_SB(sb)) {
1501 ext4_msg(sb, KERN_ERR,
1502 "Mount option \"%s\" incompatible with ext3", opt);
1506 if (args->from && !(m->flags & MOPT_STRING) && match_int(args, &arg))
1508 if (args->from && (m->flags & MOPT_GTE0) && (arg < 0))
1510 if (m->flags & MOPT_EXPLICIT)
1511 set_opt2(sb, EXPLICIT_DELALLOC);
1512 if (m->flags & MOPT_CLEAR_ERR)
1513 clear_opt(sb, ERRORS_MASK);
1514 if (token == Opt_noquota && sb_any_quota_loaded(sb)) {
1515 ext4_msg(sb, KERN_ERR, "Cannot change quota "
1516 "options when quota turned on");
1520 if (m->flags & MOPT_NOSUPPORT) {
1521 ext4_msg(sb, KERN_ERR, "%s option not supported", opt);
1522 } else if (token == Opt_commit) {
1524 arg = JBD2_DEFAULT_MAX_COMMIT_AGE;
1525 sbi->s_commit_interval = HZ * arg;
1526 } else if (token == Opt_max_batch_time) {
1527 sbi->s_max_batch_time = arg;
1528 } else if (token == Opt_min_batch_time) {
1529 sbi->s_min_batch_time = arg;
1530 } else if (token == Opt_inode_readahead_blks) {
1531 if (arg && (arg > (1 << 30) || !is_power_of_2(arg))) {
1532 ext4_msg(sb, KERN_ERR,
1533 "EXT4-fs: inode_readahead_blks must be "
1534 "0 or a power of 2 smaller than 2^31");
1537 sbi->s_inode_readahead_blks = arg;
1538 } else if (token == Opt_init_itable) {
1539 set_opt(sb, INIT_INODE_TABLE);
1541 arg = EXT4_DEF_LI_WAIT_MULT;
1542 sbi->s_li_wait_mult = arg;
1543 } else if (token == Opt_max_dir_size_kb) {
1544 sbi->s_max_dir_size_kb = arg;
1545 } else if (token == Opt_stripe) {
1546 sbi->s_stripe = arg;
1547 } else if (token == Opt_resuid) {
1548 uid = make_kuid(current_user_ns(), arg);
1549 if (!uid_valid(uid)) {
1550 ext4_msg(sb, KERN_ERR, "Invalid uid value %d", arg);
1553 sbi->s_resuid = uid;
1554 } else if (token == Opt_resgid) {
1555 gid = make_kgid(current_user_ns(), arg);
1556 if (!gid_valid(gid)) {
1557 ext4_msg(sb, KERN_ERR, "Invalid gid value %d", arg);
1560 sbi->s_resgid = gid;
1561 } else if (token == Opt_journal_dev) {
1563 ext4_msg(sb, KERN_ERR,
1564 "Cannot specify journal on remount");
1567 *journal_devnum = arg;
1568 } else if (token == Opt_journal_path) {
1570 struct inode *journal_inode;
1575 ext4_msg(sb, KERN_ERR,
1576 "Cannot specify journal on remount");
1579 journal_path = match_strdup(&args[0]);
1580 if (!journal_path) {
1581 ext4_msg(sb, KERN_ERR, "error: could not dup "
1582 "journal device string");
1586 error = kern_path(journal_path, LOOKUP_FOLLOW, &path);
1588 ext4_msg(sb, KERN_ERR, "error: could not find "
1589 "journal device path: error %d", error);
1590 kfree(journal_path);
1594 journal_inode = path.dentry->d_inode;
1595 if (!S_ISBLK(journal_inode->i_mode)) {
1596 ext4_msg(sb, KERN_ERR, "error: journal path %s "
1597 "is not a block device", journal_path);
1599 kfree(journal_path);
1603 *journal_devnum = new_encode_dev(journal_inode->i_rdev);
1605 kfree(journal_path);
1606 } else if (token == Opt_journal_ioprio) {
1608 ext4_msg(sb, KERN_ERR, "Invalid journal IO priority"
1613 IOPRIO_PRIO_VALUE(IOPRIO_CLASS_BE, arg);
1614 } else if (m->flags & MOPT_DATAJ) {
1616 if (!sbi->s_journal)
1617 ext4_msg(sb, KERN_WARNING, "Remounting file system with no journal so ignoring journalled data option");
1618 else if (test_opt(sb, DATA_FLAGS) != m->mount_opt) {
1619 ext4_msg(sb, KERN_ERR,
1620 "Cannot change data mode on remount");
1624 clear_opt(sb, DATA_FLAGS);
1625 sbi->s_mount_opt |= m->mount_opt;
1628 } else if (m->flags & MOPT_QFMT) {
1629 if (sb_any_quota_loaded(sb) &&
1630 sbi->s_jquota_fmt != m->mount_opt) {
1631 ext4_msg(sb, KERN_ERR, "Cannot change journaled "
1632 "quota options when quota turned on");
1635 if (EXT4_HAS_RO_COMPAT_FEATURE(sb,
1636 EXT4_FEATURE_RO_COMPAT_QUOTA)) {
1637 ext4_msg(sb, KERN_ERR,
1638 "Cannot set journaled quota options "
1639 "when QUOTA feature is enabled");
1642 sbi->s_jquota_fmt = m->mount_opt;
1647 if (m->flags & MOPT_CLEAR)
1649 else if (unlikely(!(m->flags & MOPT_SET))) {
1650 ext4_msg(sb, KERN_WARNING,
1651 "buggy handling of option %s", opt);
1656 sbi->s_mount_opt |= m->mount_opt;
1658 sbi->s_mount_opt &= ~m->mount_opt;
1663 static int parse_options(char *options, struct super_block *sb,
1664 unsigned long *journal_devnum,
1665 unsigned int *journal_ioprio,
1668 struct ext4_sb_info *sbi = EXT4_SB(sb);
1670 substring_t args[MAX_OPT_ARGS];
1676 while ((p = strsep(&options, ",")) != NULL) {
1680 * Initialize args struct so we know whether arg was
1681 * found; some options take optional arguments.
1683 args[0].to = args[0].from = NULL;
1684 token = match_token(p, tokens, args);
1685 if (handle_mount_opt(sb, p, token, args, journal_devnum,
1686 journal_ioprio, is_remount) < 0)
1690 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_QUOTA) &&
1691 (test_opt(sb, USRQUOTA) || test_opt(sb, GRPQUOTA))) {
1692 ext4_msg(sb, KERN_ERR, "Cannot set quota options when QUOTA "
1693 "feature is enabled");
1696 if (sbi->s_qf_names[USRQUOTA] || sbi->s_qf_names[GRPQUOTA]) {
1697 if (test_opt(sb, USRQUOTA) && sbi->s_qf_names[USRQUOTA])
1698 clear_opt(sb, USRQUOTA);
1700 if (test_opt(sb, GRPQUOTA) && sbi->s_qf_names[GRPQUOTA])
1701 clear_opt(sb, GRPQUOTA);
1703 if (test_opt(sb, GRPQUOTA) || test_opt(sb, USRQUOTA)) {
1704 ext4_msg(sb, KERN_ERR, "old and new quota "
1709 if (!sbi->s_jquota_fmt) {
1710 ext4_msg(sb, KERN_ERR, "journaled quota format "
1716 if (test_opt(sb, DIOREAD_NOLOCK)) {
1718 BLOCK_SIZE << le32_to_cpu(sbi->s_es->s_log_block_size);
1720 if (blocksize < PAGE_CACHE_SIZE) {
1721 ext4_msg(sb, KERN_ERR, "can't mount with "
1722 "dioread_nolock if block size != PAGE_SIZE");
1729 static inline void ext4_show_quota_options(struct seq_file *seq,
1730 struct super_block *sb)
1732 #if defined(CONFIG_QUOTA)
1733 struct ext4_sb_info *sbi = EXT4_SB(sb);
1735 if (sbi->s_jquota_fmt) {
1738 switch (sbi->s_jquota_fmt) {
1749 seq_printf(seq, ",jqfmt=%s", fmtname);
1752 if (sbi->s_qf_names[USRQUOTA])
1753 seq_printf(seq, ",usrjquota=%s", sbi->s_qf_names[USRQUOTA]);
1755 if (sbi->s_qf_names[GRPQUOTA])
1756 seq_printf(seq, ",grpjquota=%s", sbi->s_qf_names[GRPQUOTA]);
1760 static const char *token2str(int token)
1762 const struct match_token *t;
1764 for (t = tokens; t->token != Opt_err; t++)
1765 if (t->token == token && !strchr(t->pattern, '='))
1772 * - it's set to a non-default value OR
1773 * - if the per-sb default is different from the global default
1775 static int _ext4_show_options(struct seq_file *seq, struct super_block *sb,
1778 struct ext4_sb_info *sbi = EXT4_SB(sb);
1779 struct ext4_super_block *es = sbi->s_es;
1780 int def_errors, def_mount_opt = nodefs ? 0 : sbi->s_def_mount_opt;
1781 const struct mount_opts *m;
1782 char sep = nodefs ? '\n' : ',';
1784 #define SEQ_OPTS_PUTS(str) seq_printf(seq, "%c" str, sep)
1785 #define SEQ_OPTS_PRINT(str, arg) seq_printf(seq, "%c" str, sep, arg)
1787 if (sbi->s_sb_block != 1)
1788 SEQ_OPTS_PRINT("sb=%llu", sbi->s_sb_block);
1790 for (m = ext4_mount_opts; m->token != Opt_err; m++) {
1791 int want_set = m->flags & MOPT_SET;
1792 if (((m->flags & (MOPT_SET|MOPT_CLEAR)) == 0) ||
1793 (m->flags & MOPT_CLEAR_ERR))
1795 if (!(m->mount_opt & (sbi->s_mount_opt ^ def_mount_opt)))
1796 continue; /* skip if same as the default */
1798 (sbi->s_mount_opt & m->mount_opt) != m->mount_opt) ||
1799 (!want_set && (sbi->s_mount_opt & m->mount_opt)))
1800 continue; /* select Opt_noFoo vs Opt_Foo */
1801 SEQ_OPTS_PRINT("%s", token2str(m->token));
1804 if (nodefs || !uid_eq(sbi->s_resuid, make_kuid(&init_user_ns, EXT4_DEF_RESUID)) ||
1805 le16_to_cpu(es->s_def_resuid) != EXT4_DEF_RESUID)
1806 SEQ_OPTS_PRINT("resuid=%u",
1807 from_kuid_munged(&init_user_ns, sbi->s_resuid));
1808 if (nodefs || !gid_eq(sbi->s_resgid, make_kgid(&init_user_ns, EXT4_DEF_RESGID)) ||
1809 le16_to_cpu(es->s_def_resgid) != EXT4_DEF_RESGID)
1810 SEQ_OPTS_PRINT("resgid=%u",
1811 from_kgid_munged(&init_user_ns, sbi->s_resgid));
1812 def_errors = nodefs ? -1 : le16_to_cpu(es->s_errors);
1813 if (test_opt(sb, ERRORS_RO) && def_errors != EXT4_ERRORS_RO)
1814 SEQ_OPTS_PUTS("errors=remount-ro");
1815 if (test_opt(sb, ERRORS_CONT) && def_errors != EXT4_ERRORS_CONTINUE)
1816 SEQ_OPTS_PUTS("errors=continue");
1817 if (test_opt(sb, ERRORS_PANIC) && def_errors != EXT4_ERRORS_PANIC)
1818 SEQ_OPTS_PUTS("errors=panic");
1819 if (nodefs || sbi->s_commit_interval != JBD2_DEFAULT_MAX_COMMIT_AGE*HZ)
1820 SEQ_OPTS_PRINT("commit=%lu", sbi->s_commit_interval / HZ);
1821 if (nodefs || sbi->s_min_batch_time != EXT4_DEF_MIN_BATCH_TIME)
1822 SEQ_OPTS_PRINT("min_batch_time=%u", sbi->s_min_batch_time);
1823 if (nodefs || sbi->s_max_batch_time != EXT4_DEF_MAX_BATCH_TIME)
1824 SEQ_OPTS_PRINT("max_batch_time=%u", sbi->s_max_batch_time);
1825 if (sb->s_flags & MS_I_VERSION)
1826 SEQ_OPTS_PUTS("i_version");
1827 if (nodefs || sbi->s_stripe)
1828 SEQ_OPTS_PRINT("stripe=%lu", sbi->s_stripe);
1829 if (EXT4_MOUNT_DATA_FLAGS & (sbi->s_mount_opt ^ def_mount_opt)) {
1830 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA)
1831 SEQ_OPTS_PUTS("data=journal");
1832 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA)
1833 SEQ_OPTS_PUTS("data=ordered");
1834 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_WRITEBACK_DATA)
1835 SEQ_OPTS_PUTS("data=writeback");
1838 sbi->s_inode_readahead_blks != EXT4_DEF_INODE_READAHEAD_BLKS)
1839 SEQ_OPTS_PRINT("inode_readahead_blks=%u",
1840 sbi->s_inode_readahead_blks);
1842 if (nodefs || (test_opt(sb, INIT_INODE_TABLE) &&
1843 (sbi->s_li_wait_mult != EXT4_DEF_LI_WAIT_MULT)))
1844 SEQ_OPTS_PRINT("init_itable=%u", sbi->s_li_wait_mult);
1845 if (nodefs || sbi->s_max_dir_size_kb)
1846 SEQ_OPTS_PRINT("max_dir_size_kb=%u", sbi->s_max_dir_size_kb);
1848 ext4_show_quota_options(seq, sb);
1852 static int ext4_show_options(struct seq_file *seq, struct dentry *root)
1854 return _ext4_show_options(seq, root->d_sb, 0);
1857 static int options_seq_show(struct seq_file *seq, void *offset)
1859 struct super_block *sb = seq->private;
1862 seq_puts(seq, (sb->s_flags & MS_RDONLY) ? "ro" : "rw");
1863 rc = _ext4_show_options(seq, sb, 1);
1864 seq_puts(seq, "\n");
1868 static int options_open_fs(struct inode *inode, struct file *file)
1870 return single_open(file, options_seq_show, PDE_DATA(inode));
1873 static const struct file_operations ext4_seq_options_fops = {
1874 .owner = THIS_MODULE,
1875 .open = options_open_fs,
1877 .llseek = seq_lseek,
1878 .release = single_release,
1881 static int ext4_setup_super(struct super_block *sb, struct ext4_super_block *es,
1884 struct ext4_sb_info *sbi = EXT4_SB(sb);
1887 if (le32_to_cpu(es->s_rev_level) > EXT4_MAX_SUPP_REV) {
1888 ext4_msg(sb, KERN_ERR, "revision level too high, "
1889 "forcing read-only mode");
1894 if (!(sbi->s_mount_state & EXT4_VALID_FS))
1895 ext4_msg(sb, KERN_WARNING, "warning: mounting unchecked fs, "
1896 "running e2fsck is recommended");
1897 else if (sbi->s_mount_state & EXT4_ERROR_FS)
1898 ext4_msg(sb, KERN_WARNING,
1899 "warning: mounting fs with errors, "
1900 "running e2fsck is recommended");
1901 else if ((__s16) le16_to_cpu(es->s_max_mnt_count) > 0 &&
1902 le16_to_cpu(es->s_mnt_count) >=
1903 (unsigned short) (__s16) le16_to_cpu(es->s_max_mnt_count))
1904 ext4_msg(sb, KERN_WARNING,
1905 "warning: maximal mount count reached, "
1906 "running e2fsck is recommended");
1907 else if (le32_to_cpu(es->s_checkinterval) &&
1908 (le32_to_cpu(es->s_lastcheck) +
1909 le32_to_cpu(es->s_checkinterval) <= get_seconds()))
1910 ext4_msg(sb, KERN_WARNING,
1911 "warning: checktime reached, "
1912 "running e2fsck is recommended");
1913 if (!sbi->s_journal)
1914 es->s_state &= cpu_to_le16(~EXT4_VALID_FS);
1915 if (!(__s16) le16_to_cpu(es->s_max_mnt_count))
1916 es->s_max_mnt_count = cpu_to_le16(EXT4_DFL_MAX_MNT_COUNT);
1917 le16_add_cpu(&es->s_mnt_count, 1);
1918 es->s_mtime = cpu_to_le32(get_seconds());
1919 ext4_update_dynamic_rev(sb);
1921 EXT4_SET_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
1923 ext4_commit_super(sb, 1);
1925 if (test_opt(sb, DEBUG))
1926 printk(KERN_INFO "[EXT4 FS bs=%lu, gc=%u, "
1927 "bpg=%lu, ipg=%lu, mo=%04x, mo2=%04x]\n",
1929 sbi->s_groups_count,
1930 EXT4_BLOCKS_PER_GROUP(sb),
1931 EXT4_INODES_PER_GROUP(sb),
1932 sbi->s_mount_opt, sbi->s_mount_opt2);
1934 cleancache_init_fs(sb);
1938 int ext4_alloc_flex_bg_array(struct super_block *sb, ext4_group_t ngroup)
1940 struct ext4_sb_info *sbi = EXT4_SB(sb);
1941 struct flex_groups *new_groups;
1944 if (!sbi->s_log_groups_per_flex)
1947 size = ext4_flex_group(sbi, ngroup - 1) + 1;
1948 if (size <= sbi->s_flex_groups_allocated)
1951 size = roundup_pow_of_two(size * sizeof(struct flex_groups));
1952 new_groups = ext4_kvzalloc(size, GFP_KERNEL);
1954 ext4_msg(sb, KERN_ERR, "not enough memory for %d flex groups",
1955 size / (int) sizeof(struct flex_groups));
1959 if (sbi->s_flex_groups) {
1960 memcpy(new_groups, sbi->s_flex_groups,
1961 (sbi->s_flex_groups_allocated *
1962 sizeof(struct flex_groups)));
1963 ext4_kvfree(sbi->s_flex_groups);
1965 sbi->s_flex_groups = new_groups;
1966 sbi->s_flex_groups_allocated = size / sizeof(struct flex_groups);
1970 static int ext4_fill_flex_info(struct super_block *sb)
1972 struct ext4_sb_info *sbi = EXT4_SB(sb);
1973 struct ext4_group_desc *gdp = NULL;
1974 ext4_group_t flex_group;
1977 sbi->s_log_groups_per_flex = sbi->s_es->s_log_groups_per_flex;
1978 if (sbi->s_log_groups_per_flex < 1 || sbi->s_log_groups_per_flex > 31) {
1979 sbi->s_log_groups_per_flex = 0;
1983 err = ext4_alloc_flex_bg_array(sb, sbi->s_groups_count);
1987 for (i = 0; i < sbi->s_groups_count; i++) {
1988 gdp = ext4_get_group_desc(sb, i, NULL);
1990 flex_group = ext4_flex_group(sbi, i);
1991 atomic_add(ext4_free_inodes_count(sb, gdp),
1992 &sbi->s_flex_groups[flex_group].free_inodes);
1993 atomic64_add(ext4_free_group_clusters(sb, gdp),
1994 &sbi->s_flex_groups[flex_group].free_clusters);
1995 atomic_add(ext4_used_dirs_count(sb, gdp),
1996 &sbi->s_flex_groups[flex_group].used_dirs);
2004 static __le16 ext4_group_desc_csum(struct ext4_sb_info *sbi, __u32 block_group,
2005 struct ext4_group_desc *gdp)
2009 __le32 le_group = cpu_to_le32(block_group);
2011 if ((sbi->s_es->s_feature_ro_compat &
2012 cpu_to_le32(EXT4_FEATURE_RO_COMPAT_METADATA_CSUM))) {
2013 /* Use new metadata_csum algorithm */
2017 save_csum = gdp->bg_checksum;
2018 gdp->bg_checksum = 0;
2019 csum32 = ext4_chksum(sbi, sbi->s_csum_seed, (__u8 *)&le_group,
2021 csum32 = ext4_chksum(sbi, csum32, (__u8 *)gdp,
2023 gdp->bg_checksum = save_csum;
2025 crc = csum32 & 0xFFFF;
2029 /* old crc16 code */
2030 offset = offsetof(struct ext4_group_desc, bg_checksum);
2032 crc = crc16(~0, sbi->s_es->s_uuid, sizeof(sbi->s_es->s_uuid));
2033 crc = crc16(crc, (__u8 *)&le_group, sizeof(le_group));
2034 crc = crc16(crc, (__u8 *)gdp, offset);
2035 offset += sizeof(gdp->bg_checksum); /* skip checksum */
2036 /* for checksum of struct ext4_group_desc do the rest...*/
2037 if ((sbi->s_es->s_feature_incompat &
2038 cpu_to_le32(EXT4_FEATURE_INCOMPAT_64BIT)) &&
2039 offset < le16_to_cpu(sbi->s_es->s_desc_size))
2040 crc = crc16(crc, (__u8 *)gdp + offset,
2041 le16_to_cpu(sbi->s_es->s_desc_size) -
2045 return cpu_to_le16(crc);
2048 int ext4_group_desc_csum_verify(struct super_block *sb, __u32 block_group,
2049 struct ext4_group_desc *gdp)
2051 if (ext4_has_group_desc_csum(sb) &&
2052 (gdp->bg_checksum != ext4_group_desc_csum(EXT4_SB(sb),
2059 void ext4_group_desc_csum_set(struct super_block *sb, __u32 block_group,
2060 struct ext4_group_desc *gdp)
2062 if (!ext4_has_group_desc_csum(sb))
2064 gdp->bg_checksum = ext4_group_desc_csum(EXT4_SB(sb), block_group, gdp);
2067 /* Called at mount-time, super-block is locked */
2068 static int ext4_check_descriptors(struct super_block *sb,
2069 ext4_group_t *first_not_zeroed)
2071 struct ext4_sb_info *sbi = EXT4_SB(sb);
2072 ext4_fsblk_t first_block = le32_to_cpu(sbi->s_es->s_first_data_block);
2073 ext4_fsblk_t last_block;
2074 ext4_fsblk_t block_bitmap;
2075 ext4_fsblk_t inode_bitmap;
2076 ext4_fsblk_t inode_table;
2077 int flexbg_flag = 0;
2078 ext4_group_t i, grp = sbi->s_groups_count;
2080 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_FLEX_BG))
2083 ext4_debug("Checking group descriptors");
2085 for (i = 0; i < sbi->s_groups_count; i++) {
2086 struct ext4_group_desc *gdp = ext4_get_group_desc(sb, i, NULL);
2088 if (i == sbi->s_groups_count - 1 || flexbg_flag)
2089 last_block = ext4_blocks_count(sbi->s_es) - 1;
2091 last_block = first_block +
2092 (EXT4_BLOCKS_PER_GROUP(sb) - 1);
2094 if ((grp == sbi->s_groups_count) &&
2095 !(gdp->bg_flags & cpu_to_le16(EXT4_BG_INODE_ZEROED)))
2098 block_bitmap = ext4_block_bitmap(sb, gdp);
2099 if (block_bitmap < first_block || block_bitmap > last_block) {
2100 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2101 "Block bitmap for group %u not in group "
2102 "(block %llu)!", i, block_bitmap);
2105 inode_bitmap = ext4_inode_bitmap(sb, gdp);
2106 if (inode_bitmap < first_block || inode_bitmap > last_block) {
2107 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2108 "Inode bitmap for group %u not in group "
2109 "(block %llu)!", i, inode_bitmap);
2112 inode_table = ext4_inode_table(sb, gdp);
2113 if (inode_table < first_block ||
2114 inode_table + sbi->s_itb_per_group - 1 > last_block) {
2115 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2116 "Inode table for group %u not in group "
2117 "(block %llu)!", i, inode_table);
2120 ext4_lock_group(sb, i);
2121 if (!ext4_group_desc_csum_verify(sb, i, gdp)) {
2122 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2123 "Checksum for group %u failed (%u!=%u)",
2124 i, le16_to_cpu(ext4_group_desc_csum(sbi, i,
2125 gdp)), le16_to_cpu(gdp->bg_checksum));
2126 if (!(sb->s_flags & MS_RDONLY)) {
2127 ext4_unlock_group(sb, i);
2131 ext4_unlock_group(sb, i);
2133 first_block += EXT4_BLOCKS_PER_GROUP(sb);
2135 if (NULL != first_not_zeroed)
2136 *first_not_zeroed = grp;
2140 /* ext4_orphan_cleanup() walks a singly-linked list of inodes (starting at
2141 * the superblock) which were deleted from all directories, but held open by
2142 * a process at the time of a crash. We walk the list and try to delete these
2143 * inodes at recovery time (only with a read-write filesystem).
2145 * In order to keep the orphan inode chain consistent during traversal (in
2146 * case of crash during recovery), we link each inode into the superblock
2147 * orphan list_head and handle it the same way as an inode deletion during
2148 * normal operation (which journals the operations for us).
2150 * We only do an iget() and an iput() on each inode, which is very safe if we
2151 * accidentally point at an in-use or already deleted inode. The worst that
2152 * can happen in this case is that we get a "bit already cleared" message from
2153 * ext4_free_inode(). The only reason we would point at a wrong inode is if
2154 * e2fsck was run on this filesystem, and it must have already done the orphan
2155 * inode cleanup for us, so we can safely abort without any further action.
2157 static void ext4_orphan_cleanup(struct super_block *sb,
2158 struct ext4_super_block *es)
2160 unsigned int s_flags = sb->s_flags;
2161 int nr_orphans = 0, nr_truncates = 0;
2165 if (!es->s_last_orphan) {
2166 jbd_debug(4, "no orphan inodes to clean up\n");
2170 if (bdev_read_only(sb->s_bdev)) {
2171 ext4_msg(sb, KERN_ERR, "write access "
2172 "unavailable, skipping orphan cleanup");
2176 /* Check if feature set would not allow a r/w mount */
2177 if (!ext4_feature_set_ok(sb, 0)) {
2178 ext4_msg(sb, KERN_INFO, "Skipping orphan cleanup due to "
2179 "unknown ROCOMPAT features");
2183 if (EXT4_SB(sb)->s_mount_state & EXT4_ERROR_FS) {
2184 /* don't clear list on RO mount w/ errors */
2185 if (es->s_last_orphan && !(s_flags & MS_RDONLY)) {
2186 ext4_msg(sb, KERN_INFO, "Errors on filesystem, "
2187 "clearing orphan list.\n");
2188 es->s_last_orphan = 0;
2190 jbd_debug(1, "Skipping orphan recovery on fs with errors.\n");
2194 if (s_flags & MS_RDONLY) {
2195 ext4_msg(sb, KERN_INFO, "orphan cleanup on readonly fs");
2196 sb->s_flags &= ~MS_RDONLY;
2199 /* Needed for iput() to work correctly and not trash data */
2200 sb->s_flags |= MS_ACTIVE;
2201 /* Turn on quotas so that they are updated correctly */
2202 for (i = 0; i < EXT4_MAXQUOTAS; i++) {
2203 if (EXT4_SB(sb)->s_qf_names[i]) {
2204 int ret = ext4_quota_on_mount(sb, i);
2206 ext4_msg(sb, KERN_ERR,
2207 "Cannot turn on journaled "
2208 "quota: error %d", ret);
2213 while (es->s_last_orphan) {
2214 struct inode *inode;
2216 inode = ext4_orphan_get(sb, le32_to_cpu(es->s_last_orphan));
2217 if (IS_ERR(inode)) {
2218 es->s_last_orphan = 0;
2222 list_add(&EXT4_I(inode)->i_orphan, &EXT4_SB(sb)->s_orphan);
2223 dquot_initialize(inode);
2224 if (inode->i_nlink) {
2225 if (test_opt(sb, DEBUG))
2226 ext4_msg(sb, KERN_DEBUG,
2227 "%s: truncating inode %lu to %lld bytes",
2228 __func__, inode->i_ino, inode->i_size);
2229 jbd_debug(2, "truncating inode %lu to %lld bytes\n",
2230 inode->i_ino, inode->i_size);
2231 mutex_lock(&inode->i_mutex);
2232 truncate_inode_pages(inode->i_mapping, inode->i_size);
2233 ext4_truncate(inode);
2234 mutex_unlock(&inode->i_mutex);
2237 if (test_opt(sb, DEBUG))
2238 ext4_msg(sb, KERN_DEBUG,
2239 "%s: deleting unreferenced inode %lu",
2240 __func__, inode->i_ino);
2241 jbd_debug(2, "deleting unreferenced inode %lu\n",
2245 iput(inode); /* The delete magic happens here! */
2248 #define PLURAL(x) (x), ((x) == 1) ? "" : "s"
2251 ext4_msg(sb, KERN_INFO, "%d orphan inode%s deleted",
2252 PLURAL(nr_orphans));
2254 ext4_msg(sb, KERN_INFO, "%d truncate%s cleaned up",
2255 PLURAL(nr_truncates));
2257 /* Turn quotas off */
2258 for (i = 0; i < EXT4_MAXQUOTAS; i++) {
2259 if (sb_dqopt(sb)->files[i])
2260 dquot_quota_off(sb, i);
2263 sb->s_flags = s_flags; /* Restore MS_RDONLY status */
2267 * Maximal extent format file size.
2268 * Resulting logical blkno at s_maxbytes must fit in our on-disk
2269 * extent format containers, within a sector_t, and within i_blocks
2270 * in the vfs. ext4 inode has 48 bits of i_block in fsblock units,
2271 * so that won't be a limiting factor.
2273 * However there is other limiting factor. We do store extents in the form
2274 * of starting block and length, hence the resulting length of the extent
2275 * covering maximum file size must fit into on-disk format containers as
2276 * well. Given that length is always by 1 unit bigger than max unit (because
2277 * we count 0 as well) we have to lower the s_maxbytes by one fs block.
2279 * Note, this does *not* consider any metadata overhead for vfs i_blocks.
2281 static loff_t ext4_max_size(int blkbits, int has_huge_files)
2284 loff_t upper_limit = MAX_LFS_FILESIZE;
2286 /* small i_blocks in vfs inode? */
2287 if (!has_huge_files || sizeof(blkcnt_t) < sizeof(u64)) {
2289 * CONFIG_LBDAF is not enabled implies the inode
2290 * i_block represent total blocks in 512 bytes
2291 * 32 == size of vfs inode i_blocks * 8
2293 upper_limit = (1LL << 32) - 1;
2295 /* total blocks in file system block size */
2296 upper_limit >>= (blkbits - 9);
2297 upper_limit <<= blkbits;
2301 * 32-bit extent-start container, ee_block. We lower the maxbytes
2302 * by one fs block, so ee_len can cover the extent of maximum file
2305 res = (1LL << 32) - 1;
2308 /* Sanity check against vm- & vfs- imposed limits */
2309 if (res > upper_limit)
2316 * Maximal bitmap file size. There is a direct, and {,double-,triple-}indirect
2317 * block limit, and also a limit of (2^48 - 1) 512-byte sectors in i_blocks.
2318 * We need to be 1 filesystem block less than the 2^48 sector limit.
2320 static loff_t ext4_max_bitmap_size(int bits, int has_huge_files)
2322 loff_t res = EXT4_NDIR_BLOCKS;
2325 /* This is calculated to be the largest file size for a dense, block
2326 * mapped file such that the file's total number of 512-byte sectors,
2327 * including data and all indirect blocks, does not exceed (2^48 - 1).
2329 * __u32 i_blocks_lo and _u16 i_blocks_high represent the total
2330 * number of 512-byte sectors of the file.
2333 if (!has_huge_files || sizeof(blkcnt_t) < sizeof(u64)) {
2335 * !has_huge_files or CONFIG_LBDAF not enabled implies that
2336 * the inode i_block field represents total file blocks in
2337 * 2^32 512-byte sectors == size of vfs inode i_blocks * 8
2339 upper_limit = (1LL << 32) - 1;
2341 /* total blocks in file system block size */
2342 upper_limit >>= (bits - 9);
2346 * We use 48 bit ext4_inode i_blocks
2347 * With EXT4_HUGE_FILE_FL set the i_blocks
2348 * represent total number of blocks in
2349 * file system block size
2351 upper_limit = (1LL << 48) - 1;
2355 /* indirect blocks */
2357 /* double indirect blocks */
2358 meta_blocks += 1 + (1LL << (bits-2));
2359 /* tripple indirect blocks */
2360 meta_blocks += 1 + (1LL << (bits-2)) + (1LL << (2*(bits-2)));
2362 upper_limit -= meta_blocks;
2363 upper_limit <<= bits;
2365 res += 1LL << (bits-2);
2366 res += 1LL << (2*(bits-2));
2367 res += 1LL << (3*(bits-2));
2369 if (res > upper_limit)
2372 if (res > MAX_LFS_FILESIZE)
2373 res = MAX_LFS_FILESIZE;
2378 static ext4_fsblk_t descriptor_loc(struct super_block *sb,
2379 ext4_fsblk_t logical_sb_block, int nr)
2381 struct ext4_sb_info *sbi = EXT4_SB(sb);
2382 ext4_group_t bg, first_meta_bg;
2385 first_meta_bg = le32_to_cpu(sbi->s_es->s_first_meta_bg);
2387 if (!EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_META_BG) ||
2389 return logical_sb_block + nr + 1;
2390 bg = sbi->s_desc_per_block * nr;
2391 if (ext4_bg_has_super(sb, bg))
2395 * If we have a meta_bg fs with 1k blocks, group 0's GDT is at
2396 * block 2, not 1. If s_first_data_block == 0 (bigalloc is enabled
2397 * on modern mke2fs or blksize > 1k on older mke2fs) then we must
2400 if (sb->s_blocksize == 1024 && nr == 0 &&
2401 le32_to_cpu(EXT4_SB(sb)->s_es->s_first_data_block) == 0)
2404 return (has_super + ext4_group_first_block_no(sb, bg));
2408 * ext4_get_stripe_size: Get the stripe size.
2409 * @sbi: In memory super block info
2411 * If we have specified it via mount option, then
2412 * use the mount option value. If the value specified at mount time is
2413 * greater than the blocks per group use the super block value.
2414 * If the super block value is greater than blocks per group return 0.
2415 * Allocator needs it be less than blocks per group.
2418 static unsigned long ext4_get_stripe_size(struct ext4_sb_info *sbi)
2420 unsigned long stride = le16_to_cpu(sbi->s_es->s_raid_stride);
2421 unsigned long stripe_width =
2422 le32_to_cpu(sbi->s_es->s_raid_stripe_width);
2425 if (sbi->s_stripe && sbi->s_stripe <= sbi->s_blocks_per_group)
2426 ret = sbi->s_stripe;
2427 else if (stripe_width <= sbi->s_blocks_per_group)
2429 else if (stride <= sbi->s_blocks_per_group)
2435 * If the stripe width is 1, this makes no sense and
2436 * we set it to 0 to turn off stripe handling code.
2447 struct attribute attr;
2448 ssize_t (*show)(struct ext4_attr *, struct ext4_sb_info *, char *);
2449 ssize_t (*store)(struct ext4_attr *, struct ext4_sb_info *,
2450 const char *, size_t);
2457 static int parse_strtoull(const char *buf,
2458 unsigned long long max, unsigned long long *value)
2462 ret = kstrtoull(skip_spaces(buf), 0, value);
2463 if (!ret && *value > max)
2468 static ssize_t delayed_allocation_blocks_show(struct ext4_attr *a,
2469 struct ext4_sb_info *sbi,
2472 return snprintf(buf, PAGE_SIZE, "%llu\n",
2474 percpu_counter_sum(&sbi->s_dirtyclusters_counter)));
2477 static ssize_t session_write_kbytes_show(struct ext4_attr *a,
2478 struct ext4_sb_info *sbi, char *buf)
2480 struct super_block *sb = sbi->s_buddy_cache->i_sb;
2482 if (!sb->s_bdev->bd_part)
2483 return snprintf(buf, PAGE_SIZE, "0\n");
2484 return snprintf(buf, PAGE_SIZE, "%lu\n",
2485 (part_stat_read(sb->s_bdev->bd_part, sectors[1]) -
2486 sbi->s_sectors_written_start) >> 1);
2489 static ssize_t lifetime_write_kbytes_show(struct ext4_attr *a,
2490 struct ext4_sb_info *sbi, char *buf)
2492 struct super_block *sb = sbi->s_buddy_cache->i_sb;
2494 if (!sb->s_bdev->bd_part)
2495 return snprintf(buf, PAGE_SIZE, "0\n");
2496 return snprintf(buf, PAGE_SIZE, "%llu\n",
2497 (unsigned long long)(sbi->s_kbytes_written +
2498 ((part_stat_read(sb->s_bdev->bd_part, sectors[1]) -
2499 EXT4_SB(sb)->s_sectors_written_start) >> 1)));
2502 static ssize_t inode_readahead_blks_store(struct ext4_attr *a,
2503 struct ext4_sb_info *sbi,
2504 const char *buf, size_t count)
2509 ret = kstrtoul(skip_spaces(buf), 0, &t);
2513 if (t && (!is_power_of_2(t) || t > 0x40000000))
2516 sbi->s_inode_readahead_blks = t;
2520 static ssize_t sbi_ui_show(struct ext4_attr *a,
2521 struct ext4_sb_info *sbi, char *buf)
2523 unsigned int *ui = (unsigned int *) (((char *) sbi) + a->u.offset);
2525 return snprintf(buf, PAGE_SIZE, "%u\n", *ui);
2528 static ssize_t sbi_ui_store(struct ext4_attr *a,
2529 struct ext4_sb_info *sbi,
2530 const char *buf, size_t count)
2532 unsigned int *ui = (unsigned int *) (((char *) sbi) + a->u.offset);
2536 ret = kstrtoul(skip_spaces(buf), 0, &t);
2543 static ssize_t es_ui_show(struct ext4_attr *a,
2544 struct ext4_sb_info *sbi, char *buf)
2547 unsigned int *ui = (unsigned int *) (((char *) sbi->s_es) +
2550 return snprintf(buf, PAGE_SIZE, "%u\n", *ui);
2553 static ssize_t reserved_clusters_show(struct ext4_attr *a,
2554 struct ext4_sb_info *sbi, char *buf)
2556 return snprintf(buf, PAGE_SIZE, "%llu\n",
2557 (unsigned long long) atomic64_read(&sbi->s_resv_clusters));
2560 static ssize_t reserved_clusters_store(struct ext4_attr *a,
2561 struct ext4_sb_info *sbi,
2562 const char *buf, size_t count)
2564 unsigned long long val;
2567 if (parse_strtoull(buf, -1ULL, &val))
2569 ret = ext4_reserve_clusters(sbi, val);
2571 return ret ? ret : count;
2574 static ssize_t trigger_test_error(struct ext4_attr *a,
2575 struct ext4_sb_info *sbi,
2576 const char *buf, size_t count)
2580 if (!capable(CAP_SYS_ADMIN))
2583 if (len && buf[len-1] == '\n')
2587 ext4_error(sbi->s_sb, "%.*s", len, buf);
2591 static ssize_t sbi_deprecated_show(struct ext4_attr *a,
2592 struct ext4_sb_info *sbi, char *buf)
2594 return snprintf(buf, PAGE_SIZE, "%d\n", a->u.deprecated_val);
2597 #define EXT4_ATTR_OFFSET(_name,_mode,_show,_store,_elname) \
2598 static struct ext4_attr ext4_attr_##_name = { \
2599 .attr = {.name = __stringify(_name), .mode = _mode }, \
2603 .offset = offsetof(struct ext4_sb_info, _elname),\
2607 #define EXT4_ATTR_OFFSET_ES(_name,_mode,_show,_store,_elname) \
2608 static struct ext4_attr ext4_attr_##_name = { \
2609 .attr = {.name = __stringify(_name), .mode = _mode }, \
2613 .offset = offsetof(struct ext4_super_block, _elname), \
2617 #define EXT4_ATTR(name, mode, show, store) \
2618 static struct ext4_attr ext4_attr_##name = __ATTR(name, mode, show, store)
2620 #define EXT4_INFO_ATTR(name) EXT4_ATTR(name, 0444, NULL, NULL)
2621 #define EXT4_RO_ATTR(name) EXT4_ATTR(name, 0444, name##_show, NULL)
2622 #define EXT4_RW_ATTR(name) EXT4_ATTR(name, 0644, name##_show, name##_store)
2624 #define EXT4_RO_ATTR_ES_UI(name, elname) \
2625 EXT4_ATTR_OFFSET_ES(name, 0444, es_ui_show, NULL, elname)
2626 #define EXT4_RW_ATTR_SBI_UI(name, elname) \
2627 EXT4_ATTR_OFFSET(name, 0644, sbi_ui_show, sbi_ui_store, elname)
2629 #define ATTR_LIST(name) &ext4_attr_##name.attr
2630 #define EXT4_DEPRECATED_ATTR(_name, _val) \
2631 static struct ext4_attr ext4_attr_##_name = { \
2632 .attr = {.name = __stringify(_name), .mode = 0444 }, \
2633 .show = sbi_deprecated_show, \
2635 .deprecated_val = _val, \
2639 EXT4_RO_ATTR(delayed_allocation_blocks);
2640 EXT4_RO_ATTR(session_write_kbytes);
2641 EXT4_RO_ATTR(lifetime_write_kbytes);
2642 EXT4_RW_ATTR(reserved_clusters);
2643 EXT4_ATTR_OFFSET(inode_readahead_blks, 0644, sbi_ui_show,
2644 inode_readahead_blks_store, s_inode_readahead_blks);
2645 EXT4_RW_ATTR_SBI_UI(inode_goal, s_inode_goal);
2646 EXT4_RW_ATTR_SBI_UI(mb_stats, s_mb_stats);
2647 EXT4_RW_ATTR_SBI_UI(mb_max_to_scan, s_mb_max_to_scan);
2648 EXT4_RW_ATTR_SBI_UI(mb_min_to_scan, s_mb_min_to_scan);
2649 EXT4_RW_ATTR_SBI_UI(mb_order2_req, s_mb_order2_reqs);
2650 EXT4_RW_ATTR_SBI_UI(mb_stream_req, s_mb_stream_request);
2651 EXT4_RW_ATTR_SBI_UI(mb_group_prealloc, s_mb_group_prealloc);
2652 EXT4_DEPRECATED_ATTR(max_writeback_mb_bump, 128);
2653 EXT4_RW_ATTR_SBI_UI(extent_max_zeroout_kb, s_extent_max_zeroout_kb);
2654 EXT4_ATTR(trigger_fs_error, 0200, NULL, trigger_test_error);
2655 EXT4_RW_ATTR_SBI_UI(err_ratelimit_interval_ms, s_err_ratelimit_state.interval);
2656 EXT4_RW_ATTR_SBI_UI(err_ratelimit_burst, s_err_ratelimit_state.burst);
2657 EXT4_RW_ATTR_SBI_UI(warning_ratelimit_interval_ms, s_warning_ratelimit_state.interval);
2658 EXT4_RW_ATTR_SBI_UI(warning_ratelimit_burst, s_warning_ratelimit_state.burst);
2659 EXT4_RW_ATTR_SBI_UI(msg_ratelimit_interval_ms, s_msg_ratelimit_state.interval);
2660 EXT4_RW_ATTR_SBI_UI(msg_ratelimit_burst, s_msg_ratelimit_state.burst);
2661 EXT4_RO_ATTR_ES_UI(errors_count, s_error_count);
2662 EXT4_RO_ATTR_ES_UI(first_error_time, s_first_error_time);
2663 EXT4_RO_ATTR_ES_UI(last_error_time, s_last_error_time);
2665 static struct attribute *ext4_attrs[] = {
2666 ATTR_LIST(delayed_allocation_blocks),
2667 ATTR_LIST(session_write_kbytes),
2668 ATTR_LIST(lifetime_write_kbytes),
2669 ATTR_LIST(reserved_clusters),
2670 ATTR_LIST(inode_readahead_blks),
2671 ATTR_LIST(inode_goal),
2672 ATTR_LIST(mb_stats),
2673 ATTR_LIST(mb_max_to_scan),
2674 ATTR_LIST(mb_min_to_scan),
2675 ATTR_LIST(mb_order2_req),
2676 ATTR_LIST(mb_stream_req),
2677 ATTR_LIST(mb_group_prealloc),
2678 ATTR_LIST(max_writeback_mb_bump),
2679 ATTR_LIST(extent_max_zeroout_kb),
2680 ATTR_LIST(trigger_fs_error),
2681 ATTR_LIST(err_ratelimit_interval_ms),
2682 ATTR_LIST(err_ratelimit_burst),
2683 ATTR_LIST(warning_ratelimit_interval_ms),
2684 ATTR_LIST(warning_ratelimit_burst),
2685 ATTR_LIST(msg_ratelimit_interval_ms),
2686 ATTR_LIST(msg_ratelimit_burst),
2687 ATTR_LIST(errors_count),
2688 ATTR_LIST(first_error_time),
2689 ATTR_LIST(last_error_time),
2693 /* Features this copy of ext4 supports */
2694 EXT4_INFO_ATTR(lazy_itable_init);
2695 EXT4_INFO_ATTR(batched_discard);
2696 EXT4_INFO_ATTR(meta_bg_resize);
2698 static struct attribute *ext4_feat_attrs[] = {
2699 ATTR_LIST(lazy_itable_init),
2700 ATTR_LIST(batched_discard),
2701 ATTR_LIST(meta_bg_resize),
2705 static ssize_t ext4_attr_show(struct kobject *kobj,
2706 struct attribute *attr, char *buf)
2708 struct ext4_sb_info *sbi = container_of(kobj, struct ext4_sb_info,
2710 struct ext4_attr *a = container_of(attr, struct ext4_attr, attr);
2712 return a->show ? a->show(a, sbi, buf) : 0;
2715 static ssize_t ext4_attr_store(struct kobject *kobj,
2716 struct attribute *attr,
2717 const char *buf, size_t len)
2719 struct ext4_sb_info *sbi = container_of(kobj, struct ext4_sb_info,
2721 struct ext4_attr *a = container_of(attr, struct ext4_attr, attr);
2723 return a->store ? a->store(a, sbi, buf, len) : 0;
2726 static void ext4_sb_release(struct kobject *kobj)
2728 struct ext4_sb_info *sbi = container_of(kobj, struct ext4_sb_info,
2730 complete(&sbi->s_kobj_unregister);
2733 static const struct sysfs_ops ext4_attr_ops = {
2734 .show = ext4_attr_show,
2735 .store = ext4_attr_store,
2738 static struct kobj_type ext4_ktype = {
2739 .default_attrs = ext4_attrs,
2740 .sysfs_ops = &ext4_attr_ops,
2741 .release = ext4_sb_release,
2744 static void ext4_feat_release(struct kobject *kobj)
2746 complete(&ext4_feat->f_kobj_unregister);
2749 static ssize_t ext4_feat_show(struct kobject *kobj,
2750 struct attribute *attr, char *buf)
2752 return snprintf(buf, PAGE_SIZE, "supported\n");
2756 * We can not use ext4_attr_show/store because it relies on the kobject
2757 * being embedded in the ext4_sb_info structure which is definitely not
2758 * true in this case.
2760 static const struct sysfs_ops ext4_feat_ops = {
2761 .show = ext4_feat_show,
2765 static struct kobj_type ext4_feat_ktype = {
2766 .default_attrs = ext4_feat_attrs,
2767 .sysfs_ops = &ext4_feat_ops,
2768 .release = ext4_feat_release,
2772 * Check whether this filesystem can be mounted based on
2773 * the features present and the RDONLY/RDWR mount requested.
2774 * Returns 1 if this filesystem can be mounted as requested,
2775 * 0 if it cannot be.
2777 static int ext4_feature_set_ok(struct super_block *sb, int readonly)
2779 if (EXT4_HAS_INCOMPAT_FEATURE(sb, ~EXT4_FEATURE_INCOMPAT_SUPP)) {
2780 ext4_msg(sb, KERN_ERR,
2781 "Couldn't mount because of "
2782 "unsupported optional features (%x)",
2783 (le32_to_cpu(EXT4_SB(sb)->s_es->s_feature_incompat) &
2784 ~EXT4_FEATURE_INCOMPAT_SUPP));
2791 /* Check that feature set is OK for a read-write mount */
2792 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, ~EXT4_FEATURE_RO_COMPAT_SUPP)) {
2793 ext4_msg(sb, KERN_ERR, "couldn't mount RDWR because of "
2794 "unsupported optional features (%x)",
2795 (le32_to_cpu(EXT4_SB(sb)->s_es->s_feature_ro_compat) &
2796 ~EXT4_FEATURE_RO_COMPAT_SUPP));
2800 * Large file size enabled file system can only be mounted
2801 * read-write on 32-bit systems if kernel is built with CONFIG_LBDAF
2803 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_HUGE_FILE)) {
2804 if (sizeof(blkcnt_t) < sizeof(u64)) {
2805 ext4_msg(sb, KERN_ERR, "Filesystem with huge files "
2806 "cannot be mounted RDWR without "
2811 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_BIGALLOC) &&
2812 !EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_EXTENTS)) {
2813 ext4_msg(sb, KERN_ERR,
2814 "Can't support bigalloc feature without "
2815 "extents feature\n");
2819 #ifndef CONFIG_QUOTA
2820 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_QUOTA) &&
2822 ext4_msg(sb, KERN_ERR,
2823 "Filesystem with quota feature cannot be mounted RDWR "
2824 "without CONFIG_QUOTA");
2827 #endif /* CONFIG_QUOTA */
2832 * This function is called once a day if we have errors logged
2833 * on the file system
2835 static void print_daily_error_info(unsigned long arg)
2837 struct super_block *sb = (struct super_block *) arg;
2838 struct ext4_sb_info *sbi;
2839 struct ext4_super_block *es;
2844 if (es->s_error_count)
2845 /* fsck newer than v1.41.13 is needed to clean this condition. */
2846 ext4_msg(sb, KERN_NOTICE, "error count since last fsck: %u",
2847 le32_to_cpu(es->s_error_count));
2848 if (es->s_first_error_time) {
2849 printk(KERN_NOTICE "EXT4-fs (%s): initial error at time %u: %.*s:%d",
2850 sb->s_id, le32_to_cpu(es->s_first_error_time),
2851 (int) sizeof(es->s_first_error_func),
2852 es->s_first_error_func,
2853 le32_to_cpu(es->s_first_error_line));
2854 if (es->s_first_error_ino)
2855 printk(": inode %u",
2856 le32_to_cpu(es->s_first_error_ino));
2857 if (es->s_first_error_block)
2858 printk(": block %llu", (unsigned long long)
2859 le64_to_cpu(es->s_first_error_block));
2862 if (es->s_last_error_time) {
2863 printk(KERN_NOTICE "EXT4-fs (%s): last error at time %u: %.*s:%d",
2864 sb->s_id, le32_to_cpu(es->s_last_error_time),
2865 (int) sizeof(es->s_last_error_func),
2866 es->s_last_error_func,
2867 le32_to_cpu(es->s_last_error_line));
2868 if (es->s_last_error_ino)
2869 printk(": inode %u",
2870 le32_to_cpu(es->s_last_error_ino));
2871 if (es->s_last_error_block)
2872 printk(": block %llu", (unsigned long long)
2873 le64_to_cpu(es->s_last_error_block));
2876 mod_timer(&sbi->s_err_report, jiffies + 24*60*60*HZ); /* Once a day */
2879 /* Find next suitable group and run ext4_init_inode_table */
2880 static int ext4_run_li_request(struct ext4_li_request *elr)
2882 struct ext4_group_desc *gdp = NULL;
2883 ext4_group_t group, ngroups;
2884 struct super_block *sb;
2885 unsigned long timeout = 0;
2889 ngroups = EXT4_SB(sb)->s_groups_count;
2892 for (group = elr->lr_next_group; group < ngroups; group++) {
2893 gdp = ext4_get_group_desc(sb, group, NULL);
2899 if (!(gdp->bg_flags & cpu_to_le16(EXT4_BG_INODE_ZEROED)))
2903 if (group >= ngroups)
2908 ret = ext4_init_inode_table(sb, group,
2909 elr->lr_timeout ? 0 : 1);
2910 if (elr->lr_timeout == 0) {
2911 timeout = (jiffies - timeout) *
2912 elr->lr_sbi->s_li_wait_mult;
2913 elr->lr_timeout = timeout;
2915 elr->lr_next_sched = jiffies + elr->lr_timeout;
2916 elr->lr_next_group = group + 1;
2924 * Remove lr_request from the list_request and free the
2925 * request structure. Should be called with li_list_mtx held
2927 static void ext4_remove_li_request(struct ext4_li_request *elr)
2929 struct ext4_sb_info *sbi;
2936 list_del(&elr->lr_request);
2937 sbi->s_li_request = NULL;
2941 static void ext4_unregister_li_request(struct super_block *sb)
2943 mutex_lock(&ext4_li_mtx);
2944 if (!ext4_li_info) {
2945 mutex_unlock(&ext4_li_mtx);
2949 mutex_lock(&ext4_li_info->li_list_mtx);
2950 ext4_remove_li_request(EXT4_SB(sb)->s_li_request);
2951 mutex_unlock(&ext4_li_info->li_list_mtx);
2952 mutex_unlock(&ext4_li_mtx);
2955 static struct task_struct *ext4_lazyinit_task;
2958 * This is the function where ext4lazyinit thread lives. It walks
2959 * through the request list searching for next scheduled filesystem.
2960 * When such a fs is found, run the lazy initialization request
2961 * (ext4_rn_li_request) and keep track of the time spend in this
2962 * function. Based on that time we compute next schedule time of
2963 * the request. When walking through the list is complete, compute
2964 * next waking time and put itself into sleep.
2966 static int ext4_lazyinit_thread(void *arg)
2968 struct ext4_lazy_init *eli = (struct ext4_lazy_init *)arg;
2969 struct list_head *pos, *n;
2970 struct ext4_li_request *elr;
2971 unsigned long next_wakeup, cur;
2973 BUG_ON(NULL == eli);
2977 next_wakeup = MAX_JIFFY_OFFSET;
2979 mutex_lock(&eli->li_list_mtx);
2980 if (list_empty(&eli->li_request_list)) {
2981 mutex_unlock(&eli->li_list_mtx);
2985 list_for_each_safe(pos, n, &eli->li_request_list) {
2986 elr = list_entry(pos, struct ext4_li_request,
2989 if (time_after_eq(jiffies, elr->lr_next_sched)) {
2990 if (ext4_run_li_request(elr) != 0) {
2991 /* error, remove the lazy_init job */
2992 ext4_remove_li_request(elr);
2997 if (time_before(elr->lr_next_sched, next_wakeup))
2998 next_wakeup = elr->lr_next_sched;
3000 mutex_unlock(&eli->li_list_mtx);
3005 if ((time_after_eq(cur, next_wakeup)) ||
3006 (MAX_JIFFY_OFFSET == next_wakeup)) {
3011 schedule_timeout_interruptible(next_wakeup - cur);
3013 if (kthread_should_stop()) {
3014 ext4_clear_request_list();
3021 * It looks like the request list is empty, but we need
3022 * to check it under the li_list_mtx lock, to prevent any
3023 * additions into it, and of course we should lock ext4_li_mtx
3024 * to atomically free the list and ext4_li_info, because at
3025 * this point another ext4 filesystem could be registering
3028 mutex_lock(&ext4_li_mtx);
3029 mutex_lock(&eli->li_list_mtx);
3030 if (!list_empty(&eli->li_request_list)) {
3031 mutex_unlock(&eli->li_list_mtx);
3032 mutex_unlock(&ext4_li_mtx);
3035 mutex_unlock(&eli->li_list_mtx);
3036 kfree(ext4_li_info);
3037 ext4_li_info = NULL;
3038 mutex_unlock(&ext4_li_mtx);
3043 static void ext4_clear_request_list(void)
3045 struct list_head *pos, *n;
3046 struct ext4_li_request *elr;
3048 mutex_lock(&ext4_li_info->li_list_mtx);
3049 list_for_each_safe(pos, n, &ext4_li_info->li_request_list) {
3050 elr = list_entry(pos, struct ext4_li_request,
3052 ext4_remove_li_request(elr);
3054 mutex_unlock(&ext4_li_info->li_list_mtx);
3057 static int ext4_run_lazyinit_thread(void)
3059 ext4_lazyinit_task = kthread_run(ext4_lazyinit_thread,
3060 ext4_li_info, "ext4lazyinit");
3061 if (IS_ERR(ext4_lazyinit_task)) {
3062 int err = PTR_ERR(ext4_lazyinit_task);
3063 ext4_clear_request_list();
3064 kfree(ext4_li_info);
3065 ext4_li_info = NULL;
3066 printk(KERN_CRIT "EXT4-fs: error %d creating inode table "
3067 "initialization thread\n",
3071 ext4_li_info->li_state |= EXT4_LAZYINIT_RUNNING;
3076 * Check whether it make sense to run itable init. thread or not.
3077 * If there is at least one uninitialized inode table, return
3078 * corresponding group number, else the loop goes through all
3079 * groups and return total number of groups.
3081 static ext4_group_t ext4_has_uninit_itable(struct super_block *sb)
3083 ext4_group_t group, ngroups = EXT4_SB(sb)->s_groups_count;
3084 struct ext4_group_desc *gdp = NULL;
3086 for (group = 0; group < ngroups; group++) {
3087 gdp = ext4_get_group_desc(sb, group, NULL);
3091 if (!(gdp->bg_flags & cpu_to_le16(EXT4_BG_INODE_ZEROED)))
3098 static int ext4_li_info_new(void)
3100 struct ext4_lazy_init *eli = NULL;
3102 eli = kzalloc(sizeof(*eli), GFP_KERNEL);
3106 INIT_LIST_HEAD(&eli->li_request_list);
3107 mutex_init(&eli->li_list_mtx);
3109 eli->li_state |= EXT4_LAZYINIT_QUIT;
3116 static struct ext4_li_request *ext4_li_request_new(struct super_block *sb,
3119 struct ext4_sb_info *sbi = EXT4_SB(sb);
3120 struct ext4_li_request *elr;
3122 elr = kzalloc(sizeof(*elr), GFP_KERNEL);
3128 elr->lr_next_group = start;
3131 * Randomize first schedule time of the request to
3132 * spread the inode table initialization requests
3135 elr->lr_next_sched = jiffies + (prandom_u32() %
3136 (EXT4_DEF_LI_MAX_START_DELAY * HZ));
3140 int ext4_register_li_request(struct super_block *sb,
3141 ext4_group_t first_not_zeroed)
3143 struct ext4_sb_info *sbi = EXT4_SB(sb);
3144 struct ext4_li_request *elr = NULL;
3145 ext4_group_t ngroups = EXT4_SB(sb)->s_groups_count;
3148 mutex_lock(&ext4_li_mtx);
3149 if (sbi->s_li_request != NULL) {
3151 * Reset timeout so it can be computed again, because
3152 * s_li_wait_mult might have changed.
3154 sbi->s_li_request->lr_timeout = 0;
3158 if (first_not_zeroed == ngroups ||
3159 (sb->s_flags & MS_RDONLY) ||
3160 !test_opt(sb, INIT_INODE_TABLE))
3163 elr = ext4_li_request_new(sb, first_not_zeroed);
3169 if (NULL == ext4_li_info) {
3170 ret = ext4_li_info_new();
3175 mutex_lock(&ext4_li_info->li_list_mtx);
3176 list_add(&elr->lr_request, &ext4_li_info->li_request_list);
3177 mutex_unlock(&ext4_li_info->li_list_mtx);
3179 sbi->s_li_request = elr;
3181 * set elr to NULL here since it has been inserted to
3182 * the request_list and the removal and free of it is
3183 * handled by ext4_clear_request_list from now on.
3187 if (!(ext4_li_info->li_state & EXT4_LAZYINIT_RUNNING)) {
3188 ret = ext4_run_lazyinit_thread();
3193 mutex_unlock(&ext4_li_mtx);
3200 * We do not need to lock anything since this is called on
3203 static void ext4_destroy_lazyinit_thread(void)
3206 * If thread exited earlier
3207 * there's nothing to be done.
3209 if (!ext4_li_info || !ext4_lazyinit_task)
3212 kthread_stop(ext4_lazyinit_task);
3215 static int set_journal_csum_feature_set(struct super_block *sb)
3218 int compat, incompat;
3219 struct ext4_sb_info *sbi = EXT4_SB(sb);
3221 if (EXT4_HAS_RO_COMPAT_FEATURE(sb,
3222 EXT4_FEATURE_RO_COMPAT_METADATA_CSUM)) {
3223 /* journal checksum v3 */
3225 incompat = JBD2_FEATURE_INCOMPAT_CSUM_V3;
3227 /* journal checksum v1 */
3228 compat = JBD2_FEATURE_COMPAT_CHECKSUM;
3232 jbd2_journal_clear_features(sbi->s_journal,
3233 JBD2_FEATURE_COMPAT_CHECKSUM, 0,
3234 JBD2_FEATURE_INCOMPAT_CSUM_V3 |
3235 JBD2_FEATURE_INCOMPAT_CSUM_V2);
3236 if (test_opt(sb, JOURNAL_ASYNC_COMMIT)) {
3237 ret = jbd2_journal_set_features(sbi->s_journal,
3239 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT |
3241 } else if (test_opt(sb, JOURNAL_CHECKSUM)) {
3242 ret = jbd2_journal_set_features(sbi->s_journal,
3245 jbd2_journal_clear_features(sbi->s_journal, 0, 0,
3246 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT);
3248 jbd2_journal_clear_features(sbi->s_journal, 0, 0,
3249 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT);
3256 * Note: calculating the overhead so we can be compatible with
3257 * historical BSD practice is quite difficult in the face of
3258 * clusters/bigalloc. This is because multiple metadata blocks from
3259 * different block group can end up in the same allocation cluster.
3260 * Calculating the exact overhead in the face of clustered allocation
3261 * requires either O(all block bitmaps) in memory or O(number of block
3262 * groups**2) in time. We will still calculate the superblock for
3263 * older file systems --- and if we come across with a bigalloc file
3264 * system with zero in s_overhead_clusters the estimate will be close to
3265 * correct especially for very large cluster sizes --- but for newer
3266 * file systems, it's better to calculate this figure once at mkfs
3267 * time, and store it in the superblock. If the superblock value is
3268 * present (even for non-bigalloc file systems), we will use it.
3270 static int count_overhead(struct super_block *sb, ext4_group_t grp,
3273 struct ext4_sb_info *sbi = EXT4_SB(sb);
3274 struct ext4_group_desc *gdp;
3275 ext4_fsblk_t first_block, last_block, b;
3276 ext4_group_t i, ngroups = ext4_get_groups_count(sb);
3277 int s, j, count = 0;
3279 if (!EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_BIGALLOC))
3280 return (ext4_bg_has_super(sb, grp) + ext4_bg_num_gdb(sb, grp) +
3281 sbi->s_itb_per_group + 2);
3283 first_block = le32_to_cpu(sbi->s_es->s_first_data_block) +
3284 (grp * EXT4_BLOCKS_PER_GROUP(sb));
3285 last_block = first_block + EXT4_BLOCKS_PER_GROUP(sb) - 1;
3286 for (i = 0; i < ngroups; i++) {
3287 gdp = ext4_get_group_desc(sb, i, NULL);
3288 b = ext4_block_bitmap(sb, gdp);
3289 if (b >= first_block && b <= last_block) {
3290 ext4_set_bit(EXT4_B2C(sbi, b - first_block), buf);
3293 b = ext4_inode_bitmap(sb, gdp);
3294 if (b >= first_block && b <= last_block) {
3295 ext4_set_bit(EXT4_B2C(sbi, b - first_block), buf);
3298 b = ext4_inode_table(sb, gdp);
3299 if (b >= first_block && b + sbi->s_itb_per_group <= last_block)
3300 for (j = 0; j < sbi->s_itb_per_group; j++, b++) {
3301 int c = EXT4_B2C(sbi, b - first_block);
3302 ext4_set_bit(c, buf);
3308 if (ext4_bg_has_super(sb, grp)) {
3309 ext4_set_bit(s++, buf);
3312 for (j = ext4_bg_num_gdb(sb, grp); j > 0; j--) {
3313 ext4_set_bit(EXT4_B2C(sbi, s++), buf);
3319 return EXT4_CLUSTERS_PER_GROUP(sb) -
3320 ext4_count_free(buf, EXT4_CLUSTERS_PER_GROUP(sb) / 8);
3324 * Compute the overhead and stash it in sbi->s_overhead
3326 int ext4_calculate_overhead(struct super_block *sb)
3328 struct ext4_sb_info *sbi = EXT4_SB(sb);
3329 struct ext4_super_block *es = sbi->s_es;
3330 ext4_group_t i, ngroups = ext4_get_groups_count(sb);
3331 ext4_fsblk_t overhead = 0;
3332 char *buf = (char *) get_zeroed_page(GFP_KERNEL);
3338 * Compute the overhead (FS structures). This is constant
3339 * for a given filesystem unless the number of block groups
3340 * changes so we cache the previous value until it does.
3344 * All of the blocks before first_data_block are overhead
3346 overhead = EXT4_B2C(sbi, le32_to_cpu(es->s_first_data_block));
3349 * Add the overhead found in each block group
3351 for (i = 0; i < ngroups; i++) {
3354 blks = count_overhead(sb, i, buf);
3357 memset(buf, 0, PAGE_SIZE);
3360 /* Add the journal blocks as well */
3362 overhead += EXT4_NUM_B2C(sbi, sbi->s_journal->j_maxlen);
3364 sbi->s_overhead = overhead;
3366 free_page((unsigned long) buf);
3371 static ext4_fsblk_t ext4_calculate_resv_clusters(struct super_block *sb)
3373 ext4_fsblk_t resv_clusters;
3376 * There's no need to reserve anything when we aren't using extents.
3377 * The space estimates are exact, there are no unwritten extents,
3378 * hole punching doesn't need new metadata... This is needed especially
3379 * to keep ext2/3 backward compatibility.
3381 if (!EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_EXTENTS))
3384 * By default we reserve 2% or 4096 clusters, whichever is smaller.
3385 * This should cover the situations where we can not afford to run
3386 * out of space like for example punch hole, or converting
3387 * unwritten extents in delalloc path. In most cases such
3388 * allocation would require 1, or 2 blocks, higher numbers are
3391 resv_clusters = ext4_blocks_count(EXT4_SB(sb)->s_es) >>
3392 EXT4_SB(sb)->s_cluster_bits;
3394 do_div(resv_clusters, 50);
3395 resv_clusters = min_t(ext4_fsblk_t, resv_clusters, 4096);
3397 return resv_clusters;
3401 static int ext4_reserve_clusters(struct ext4_sb_info *sbi, ext4_fsblk_t count)
3403 ext4_fsblk_t clusters = ext4_blocks_count(sbi->s_es) >>
3404 sbi->s_cluster_bits;
3406 if (count >= clusters)
3409 atomic64_set(&sbi->s_resv_clusters, count);
3413 static int ext4_fill_super(struct super_block *sb, void *data, int silent)
3415 char *orig_data = kstrdup(data, GFP_KERNEL);
3416 struct buffer_head *bh;
3417 struct ext4_super_block *es = NULL;
3418 struct ext4_sb_info *sbi;
3420 ext4_fsblk_t sb_block = get_sb_block(&data);
3421 ext4_fsblk_t logical_sb_block;
3422 unsigned long offset = 0;
3423 unsigned long journal_devnum = 0;
3424 unsigned long def_mount_opts;
3429 int blocksize, clustersize;
3430 unsigned int db_count;
3432 int needs_recovery, has_huge_files, has_bigalloc;
3435 unsigned int journal_ioprio = DEFAULT_JOURNAL_IOPRIO;
3436 ext4_group_t first_not_zeroed;
3438 sbi = kzalloc(sizeof(*sbi), GFP_KERNEL);
3442 sbi->s_blockgroup_lock =
3443 kzalloc(sizeof(struct blockgroup_lock), GFP_KERNEL);
3444 if (!sbi->s_blockgroup_lock) {
3448 sb->s_fs_info = sbi;
3450 sbi->s_inode_readahead_blks = EXT4_DEF_INODE_READAHEAD_BLKS;
3451 sbi->s_sb_block = sb_block;
3452 if (sb->s_bdev->bd_part)
3453 sbi->s_sectors_written_start =
3454 part_stat_read(sb->s_bdev->bd_part, sectors[1]);
3456 /* Cleanup superblock name */
3457 for (cp = sb->s_id; (cp = strchr(cp, '/'));)
3460 /* -EINVAL is default */
3462 blocksize = sb_min_blocksize(sb, EXT4_MIN_BLOCK_SIZE);
3464 ext4_msg(sb, KERN_ERR, "unable to set blocksize");
3469 * The ext4 superblock will not be buffer aligned for other than 1kB
3470 * block sizes. We need to calculate the offset from buffer start.
3472 if (blocksize != EXT4_MIN_BLOCK_SIZE) {
3473 logical_sb_block = sb_block * EXT4_MIN_BLOCK_SIZE;
3474 offset = do_div(logical_sb_block, blocksize);
3476 logical_sb_block = sb_block;
3479 if (!(bh = sb_bread_unmovable(sb, logical_sb_block))) {
3480 ext4_msg(sb, KERN_ERR, "unable to read superblock");
3484 * Note: s_es must be initialized as soon as possible because
3485 * some ext4 macro-instructions depend on its value
3487 es = (struct ext4_super_block *) (bh->b_data + offset);
3489 sb->s_magic = le16_to_cpu(es->s_magic);
3490 if (sb->s_magic != EXT4_SUPER_MAGIC)
3492 sbi->s_kbytes_written = le64_to_cpu(es->s_kbytes_written);
3494 /* Warn if metadata_csum and gdt_csum are both set. */
3495 if (EXT4_HAS_RO_COMPAT_FEATURE(sb,
3496 EXT4_FEATURE_RO_COMPAT_METADATA_CSUM) &&
3497 EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_GDT_CSUM))
3498 ext4_warning(sb, KERN_INFO "metadata_csum and uninit_bg are "
3499 "redundant flags; please run fsck.");
3501 /* Check for a known checksum algorithm */
3502 if (!ext4_verify_csum_type(sb, es)) {
3503 ext4_msg(sb, KERN_ERR, "VFS: Found ext4 filesystem with "
3504 "unknown checksum algorithm.");
3509 /* Load the checksum driver */
3510 if (EXT4_HAS_RO_COMPAT_FEATURE(sb,
3511 EXT4_FEATURE_RO_COMPAT_METADATA_CSUM)) {
3512 sbi->s_chksum_driver = crypto_alloc_shash("crc32c", 0, 0);
3513 if (IS_ERR(sbi->s_chksum_driver)) {
3514 ext4_msg(sb, KERN_ERR, "Cannot load crc32c driver.");
3515 ret = PTR_ERR(sbi->s_chksum_driver);
3516 sbi->s_chksum_driver = NULL;
3521 /* Check superblock checksum */
3522 if (!ext4_superblock_csum_verify(sb, es)) {
3523 ext4_msg(sb, KERN_ERR, "VFS: Found ext4 filesystem with "
3524 "invalid superblock checksum. Run e2fsck?");
3529 /* Precompute checksum seed for all metadata */
3530 if (EXT4_HAS_RO_COMPAT_FEATURE(sb,
3531 EXT4_FEATURE_RO_COMPAT_METADATA_CSUM))
3532 sbi->s_csum_seed = ext4_chksum(sbi, ~0, es->s_uuid,
3533 sizeof(es->s_uuid));
3535 /* Set defaults before we parse the mount options */
3536 def_mount_opts = le32_to_cpu(es->s_default_mount_opts);
3537 set_opt(sb, INIT_INODE_TABLE);
3538 if (def_mount_opts & EXT4_DEFM_DEBUG)
3540 if (def_mount_opts & EXT4_DEFM_BSDGROUPS)
3542 if (def_mount_opts & EXT4_DEFM_UID16)
3543 set_opt(sb, NO_UID32);
3544 /* xattr user namespace & acls are now defaulted on */
3545 set_opt(sb, XATTR_USER);
3546 #ifdef CONFIG_EXT4_FS_POSIX_ACL
3547 set_opt(sb, POSIX_ACL);
3549 if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_DATA)
3550 set_opt(sb, JOURNAL_DATA);
3551 else if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_ORDERED)
3552 set_opt(sb, ORDERED_DATA);
3553 else if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_WBACK)
3554 set_opt(sb, WRITEBACK_DATA);
3556 if (le16_to_cpu(sbi->s_es->s_errors) == EXT4_ERRORS_PANIC)
3557 set_opt(sb, ERRORS_PANIC);
3558 else if (le16_to_cpu(sbi->s_es->s_errors) == EXT4_ERRORS_CONTINUE)
3559 set_opt(sb, ERRORS_CONT);
3561 set_opt(sb, ERRORS_RO);
3562 /* block_validity enabled by default; disable with noblock_validity */
3563 set_opt(sb, BLOCK_VALIDITY);
3564 if (def_mount_opts & EXT4_DEFM_DISCARD)
3565 set_opt(sb, DISCARD);
3567 sbi->s_resuid = make_kuid(&init_user_ns, le16_to_cpu(es->s_def_resuid));
3568 sbi->s_resgid = make_kgid(&init_user_ns, le16_to_cpu(es->s_def_resgid));
3569 sbi->s_commit_interval = JBD2_DEFAULT_MAX_COMMIT_AGE * HZ;
3570 sbi->s_min_batch_time = EXT4_DEF_MIN_BATCH_TIME;
3571 sbi->s_max_batch_time = EXT4_DEF_MAX_BATCH_TIME;
3573 if ((def_mount_opts & EXT4_DEFM_NOBARRIER) == 0)
3574 set_opt(sb, BARRIER);
3577 * enable delayed allocation by default
3578 * Use -o nodelalloc to turn it off
3580 if (!IS_EXT3_SB(sb) && !IS_EXT2_SB(sb) &&
3581 ((def_mount_opts & EXT4_DEFM_NODELALLOC) == 0))
3582 set_opt(sb, DELALLOC);
3585 * set default s_li_wait_mult for lazyinit, for the case there is
3586 * no mount option specified.
3588 sbi->s_li_wait_mult = EXT4_DEF_LI_WAIT_MULT;
3590 if (!parse_options((char *) sbi->s_es->s_mount_opts, sb,
3591 &journal_devnum, &journal_ioprio, 0)) {
3592 ext4_msg(sb, KERN_WARNING,
3593 "failed to parse options in superblock: %s",
3594 sbi->s_es->s_mount_opts);
3596 sbi->s_def_mount_opt = sbi->s_mount_opt;
3597 if (!parse_options((char *) data, sb, &journal_devnum,
3598 &journal_ioprio, 0))
3601 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA) {
3602 printk_once(KERN_WARNING "EXT4-fs: Warning: mounting "
3603 "with data=journal disables delayed "
3604 "allocation and O_DIRECT support!\n");
3605 if (test_opt2(sb, EXPLICIT_DELALLOC)) {
3606 ext4_msg(sb, KERN_ERR, "can't mount with "
3607 "both data=journal and delalloc");
3610 if (test_opt(sb, DIOREAD_NOLOCK)) {
3611 ext4_msg(sb, KERN_ERR, "can't mount with "
3612 "both data=journal and dioread_nolock");
3615 if (test_opt(sb, DELALLOC))
3616 clear_opt(sb, DELALLOC);
3619 sb->s_flags = (sb->s_flags & ~MS_POSIXACL) |
3620 (test_opt(sb, POSIX_ACL) ? MS_POSIXACL : 0);
3622 if (le32_to_cpu(es->s_rev_level) == EXT4_GOOD_OLD_REV &&
3623 (EXT4_HAS_COMPAT_FEATURE(sb, ~0U) ||
3624 EXT4_HAS_RO_COMPAT_FEATURE(sb, ~0U) ||
3625 EXT4_HAS_INCOMPAT_FEATURE(sb, ~0U)))
3626 ext4_msg(sb, KERN_WARNING,
3627 "feature flags set on rev 0 fs, "
3628 "running e2fsck is recommended");
3630 if (es->s_creator_os == cpu_to_le32(EXT4_OS_HURD)) {
3631 set_opt2(sb, HURD_COMPAT);
3632 if (EXT4_HAS_INCOMPAT_FEATURE(sb,
3633 EXT4_FEATURE_INCOMPAT_64BIT)) {
3634 ext4_msg(sb, KERN_ERR,
3635 "The Hurd can't support 64-bit file systems");
3640 if (IS_EXT2_SB(sb)) {
3641 if (ext2_feature_set_ok(sb))
3642 ext4_msg(sb, KERN_INFO, "mounting ext2 file system "
3643 "using the ext4 subsystem");
3645 ext4_msg(sb, KERN_ERR, "couldn't mount as ext2 due "
3646 "to feature incompatibilities");
3651 if (IS_EXT3_SB(sb)) {
3652 if (ext3_feature_set_ok(sb))
3653 ext4_msg(sb, KERN_INFO, "mounting ext3 file system "
3654 "using the ext4 subsystem");
3656 ext4_msg(sb, KERN_ERR, "couldn't mount as ext3 due "
3657 "to feature incompatibilities");
3663 * Check feature flags regardless of the revision level, since we
3664 * previously didn't change the revision level when setting the flags,
3665 * so there is a chance incompat flags are set on a rev 0 filesystem.
3667 if (!ext4_feature_set_ok(sb, (sb->s_flags & MS_RDONLY)))
3670 blocksize = BLOCK_SIZE << le32_to_cpu(es->s_log_block_size);
3671 if (blocksize < EXT4_MIN_BLOCK_SIZE ||
3672 blocksize > EXT4_MAX_BLOCK_SIZE) {
3673 ext4_msg(sb, KERN_ERR,
3674 "Unsupported filesystem blocksize %d", blocksize);
3678 if (sb->s_blocksize != blocksize) {
3679 /* Validate the filesystem blocksize */
3680 if (!sb_set_blocksize(sb, blocksize)) {
3681 ext4_msg(sb, KERN_ERR, "bad block size %d",
3687 logical_sb_block = sb_block * EXT4_MIN_BLOCK_SIZE;
3688 offset = do_div(logical_sb_block, blocksize);
3689 bh = sb_bread_unmovable(sb, logical_sb_block);
3691 ext4_msg(sb, KERN_ERR,
3692 "Can't read superblock on 2nd try");
3695 es = (struct ext4_super_block *)(bh->b_data + offset);
3697 if (es->s_magic != cpu_to_le16(EXT4_SUPER_MAGIC)) {
3698 ext4_msg(sb, KERN_ERR,
3699 "Magic mismatch, very weird!");
3704 has_huge_files = EXT4_HAS_RO_COMPAT_FEATURE(sb,
3705 EXT4_FEATURE_RO_COMPAT_HUGE_FILE);
3706 sbi->s_bitmap_maxbytes = ext4_max_bitmap_size(sb->s_blocksize_bits,
3708 sb->s_maxbytes = ext4_max_size(sb->s_blocksize_bits, has_huge_files);
3710 if (le32_to_cpu(es->s_rev_level) == EXT4_GOOD_OLD_REV) {
3711 sbi->s_inode_size = EXT4_GOOD_OLD_INODE_SIZE;
3712 sbi->s_first_ino = EXT4_GOOD_OLD_FIRST_INO;
3714 sbi->s_inode_size = le16_to_cpu(es->s_inode_size);
3715 sbi->s_first_ino = le32_to_cpu(es->s_first_ino);
3716 if ((sbi->s_inode_size < EXT4_GOOD_OLD_INODE_SIZE) ||
3717 (!is_power_of_2(sbi->s_inode_size)) ||
3718 (sbi->s_inode_size > blocksize)) {
3719 ext4_msg(sb, KERN_ERR,
3720 "unsupported inode size: %d",
3724 if (sbi->s_inode_size > EXT4_GOOD_OLD_INODE_SIZE)
3725 sb->s_time_gran = 1 << (EXT4_EPOCH_BITS - 2);
3728 sbi->s_desc_size = le16_to_cpu(es->s_desc_size);
3729 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_64BIT)) {
3730 if (sbi->s_desc_size < EXT4_MIN_DESC_SIZE_64BIT ||
3731 sbi->s_desc_size > EXT4_MAX_DESC_SIZE ||
3732 !is_power_of_2(sbi->s_desc_size)) {
3733 ext4_msg(sb, KERN_ERR,
3734 "unsupported descriptor size %lu",
3739 sbi->s_desc_size = EXT4_MIN_DESC_SIZE;
3741 sbi->s_blocks_per_group = le32_to_cpu(es->s_blocks_per_group);
3742 sbi->s_inodes_per_group = le32_to_cpu(es->s_inodes_per_group);
3743 if (EXT4_INODE_SIZE(sb) == 0 || EXT4_INODES_PER_GROUP(sb) == 0)
3746 sbi->s_inodes_per_block = blocksize / EXT4_INODE_SIZE(sb);
3747 if (sbi->s_inodes_per_block == 0)
3749 sbi->s_itb_per_group = sbi->s_inodes_per_group /
3750 sbi->s_inodes_per_block;
3751 sbi->s_desc_per_block = blocksize / EXT4_DESC_SIZE(sb);
3753 sbi->s_mount_state = le16_to_cpu(es->s_state);
3754 sbi->s_addr_per_block_bits = ilog2(EXT4_ADDR_PER_BLOCK(sb));
3755 sbi->s_desc_per_block_bits = ilog2(EXT4_DESC_PER_BLOCK(sb));
3757 for (i = 0; i < 4; i++)
3758 sbi->s_hash_seed[i] = le32_to_cpu(es->s_hash_seed[i]);
3759 sbi->s_def_hash_version = es->s_def_hash_version;
3760 if (EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_DIR_INDEX)) {
3761 i = le32_to_cpu(es->s_flags);
3762 if (i & EXT2_FLAGS_UNSIGNED_HASH)
3763 sbi->s_hash_unsigned = 3;
3764 else if ((i & EXT2_FLAGS_SIGNED_HASH) == 0) {
3765 #ifdef __CHAR_UNSIGNED__
3766 if (!(sb->s_flags & MS_RDONLY))
3768 cpu_to_le32(EXT2_FLAGS_UNSIGNED_HASH);
3769 sbi->s_hash_unsigned = 3;
3771 if (!(sb->s_flags & MS_RDONLY))
3773 cpu_to_le32(EXT2_FLAGS_SIGNED_HASH);
3778 /* Handle clustersize */
3779 clustersize = BLOCK_SIZE << le32_to_cpu(es->s_log_cluster_size);
3780 has_bigalloc = EXT4_HAS_RO_COMPAT_FEATURE(sb,
3781 EXT4_FEATURE_RO_COMPAT_BIGALLOC);
3783 if (clustersize < blocksize) {
3784 ext4_msg(sb, KERN_ERR,
3785 "cluster size (%d) smaller than "
3786 "block size (%d)", clustersize, blocksize);
3789 sbi->s_cluster_bits = le32_to_cpu(es->s_log_cluster_size) -
3790 le32_to_cpu(es->s_log_block_size);
3791 sbi->s_clusters_per_group =
3792 le32_to_cpu(es->s_clusters_per_group);
3793 if (sbi->s_clusters_per_group > blocksize * 8) {
3794 ext4_msg(sb, KERN_ERR,
3795 "#clusters per group too big: %lu",
3796 sbi->s_clusters_per_group);
3799 if (sbi->s_blocks_per_group !=
3800 (sbi->s_clusters_per_group * (clustersize / blocksize))) {
3801 ext4_msg(sb, KERN_ERR, "blocks per group (%lu) and "
3802 "clusters per group (%lu) inconsistent",
3803 sbi->s_blocks_per_group,
3804 sbi->s_clusters_per_group);
3808 if (clustersize != blocksize) {
3809 ext4_warning(sb, "fragment/cluster size (%d) != "
3810 "block size (%d)", clustersize,
3812 clustersize = blocksize;
3814 if (sbi->s_blocks_per_group > blocksize * 8) {
3815 ext4_msg(sb, KERN_ERR,
3816 "#blocks per group too big: %lu",
3817 sbi->s_blocks_per_group);
3820 sbi->s_clusters_per_group = sbi->s_blocks_per_group;
3821 sbi->s_cluster_bits = 0;
3823 sbi->s_cluster_ratio = clustersize / blocksize;
3825 if (sbi->s_inodes_per_group > blocksize * 8) {
3826 ext4_msg(sb, KERN_ERR,
3827 "#inodes per group too big: %lu",
3828 sbi->s_inodes_per_group);
3832 /* Do we have standard group size of clustersize * 8 blocks ? */
3833 if (sbi->s_blocks_per_group == clustersize << 3)
3834 set_opt2(sb, STD_GROUP_SIZE);
3837 * Test whether we have more sectors than will fit in sector_t,
3838 * and whether the max offset is addressable by the page cache.
3840 err = generic_check_addressable(sb->s_blocksize_bits,
3841 ext4_blocks_count(es));
3843 ext4_msg(sb, KERN_ERR, "filesystem"
3844 " too large to mount safely on this system");
3845 if (sizeof(sector_t) < 8)
3846 ext4_msg(sb, KERN_WARNING, "CONFIG_LBDAF not enabled");
3850 if (EXT4_BLOCKS_PER_GROUP(sb) == 0)
3853 /* check blocks count against device size */
3854 blocks_count = sb->s_bdev->bd_inode->i_size >> sb->s_blocksize_bits;
3855 if (blocks_count && ext4_blocks_count(es) > blocks_count) {
3856 ext4_msg(sb, KERN_WARNING, "bad geometry: block count %llu "
3857 "exceeds size of device (%llu blocks)",
3858 ext4_blocks_count(es), blocks_count);
3863 * It makes no sense for the first data block to be beyond the end
3864 * of the filesystem.
3866 if (le32_to_cpu(es->s_first_data_block) >= ext4_blocks_count(es)) {
3867 ext4_msg(sb, KERN_WARNING, "bad geometry: first data "
3868 "block %u is beyond end of filesystem (%llu)",
3869 le32_to_cpu(es->s_first_data_block),
3870 ext4_blocks_count(es));
3873 blocks_count = (ext4_blocks_count(es) -
3874 le32_to_cpu(es->s_first_data_block) +
3875 EXT4_BLOCKS_PER_GROUP(sb) - 1);
3876 do_div(blocks_count, EXT4_BLOCKS_PER_GROUP(sb));
3877 if (blocks_count > ((uint64_t)1<<32) - EXT4_DESC_PER_BLOCK(sb)) {
3878 ext4_msg(sb, KERN_WARNING, "groups count too large: %u "
3879 "(block count %llu, first data block %u, "
3880 "blocks per group %lu)", sbi->s_groups_count,
3881 ext4_blocks_count(es),
3882 le32_to_cpu(es->s_first_data_block),
3883 EXT4_BLOCKS_PER_GROUP(sb));
3886 sbi->s_groups_count = blocks_count;
3887 sbi->s_blockfile_groups = min_t(ext4_group_t, sbi->s_groups_count,
3888 (EXT4_MAX_BLOCK_FILE_PHYS / EXT4_BLOCKS_PER_GROUP(sb)));
3889 db_count = (sbi->s_groups_count + EXT4_DESC_PER_BLOCK(sb) - 1) /
3890 EXT4_DESC_PER_BLOCK(sb);
3891 sbi->s_group_desc = ext4_kvmalloc(db_count *
3892 sizeof(struct buffer_head *),
3894 if (sbi->s_group_desc == NULL) {
3895 ext4_msg(sb, KERN_ERR, "not enough memory");
3901 sbi->s_proc = proc_mkdir(sb->s_id, ext4_proc_root);
3904 proc_create_data("options", S_IRUGO, sbi->s_proc,
3905 &ext4_seq_options_fops, sb);
3907 bgl_lock_init(sbi->s_blockgroup_lock);
3909 for (i = 0; i < db_count; i++) {
3910 block = descriptor_loc(sb, logical_sb_block, i);
3911 sbi->s_group_desc[i] = sb_bread_unmovable(sb, block);
3912 if (!sbi->s_group_desc[i]) {
3913 ext4_msg(sb, KERN_ERR,
3914 "can't read group descriptor %d", i);
3919 if (!ext4_check_descriptors(sb, &first_not_zeroed)) {
3920 ext4_msg(sb, KERN_ERR, "group descriptors corrupted!");
3924 sbi->s_gdb_count = db_count;
3925 get_random_bytes(&sbi->s_next_generation, sizeof(u32));
3926 spin_lock_init(&sbi->s_next_gen_lock);
3928 init_timer(&sbi->s_err_report);
3929 sbi->s_err_report.function = print_daily_error_info;
3930 sbi->s_err_report.data = (unsigned long) sb;
3932 /* Register extent status tree shrinker */
3933 if (ext4_es_register_shrinker(sbi))
3936 sbi->s_stripe = ext4_get_stripe_size(sbi);
3937 sbi->s_extent_max_zeroout_kb = 32;
3940 * set up enough so that it can read an inode
3942 if (!test_opt(sb, NOLOAD) &&
3943 EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL))
3944 sb->s_op = &ext4_sops;
3946 sb->s_op = &ext4_nojournal_sops;
3947 sb->s_export_op = &ext4_export_ops;
3948 sb->s_xattr = ext4_xattr_handlers;
3950 sb->dq_op = &ext4_quota_operations;
3951 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_QUOTA))
3952 sb->s_qcop = &ext4_qctl_sysfile_operations;
3954 sb->s_qcop = &ext4_qctl_operations;
3956 memcpy(sb->s_uuid, es->s_uuid, sizeof(es->s_uuid));
3958 INIT_LIST_HEAD(&sbi->s_orphan); /* unlinked but open files */
3959 mutex_init(&sbi->s_orphan_lock);
3963 needs_recovery = (es->s_last_orphan != 0 ||
3964 EXT4_HAS_INCOMPAT_FEATURE(sb,
3965 EXT4_FEATURE_INCOMPAT_RECOVER));
3967 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_MMP) &&
3968 !(sb->s_flags & MS_RDONLY))
3969 if (ext4_multi_mount_protect(sb, le64_to_cpu(es->s_mmp_block)))
3973 * The first inode we look at is the journal inode. Don't try
3974 * root first: it may be modified in the journal!
3976 if (!test_opt(sb, NOLOAD) &&
3977 EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL)) {
3978 if (ext4_load_journal(sb, es, journal_devnum))
3980 } else if (test_opt(sb, NOLOAD) && !(sb->s_flags & MS_RDONLY) &&
3981 EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER)) {
3982 ext4_msg(sb, KERN_ERR, "required journal recovery "
3983 "suppressed and not mounted read-only");
3984 goto failed_mount_wq;
3986 clear_opt(sb, DATA_FLAGS);
3987 sbi->s_journal = NULL;
3992 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_64BIT) &&
3993 !jbd2_journal_set_features(EXT4_SB(sb)->s_journal, 0, 0,
3994 JBD2_FEATURE_INCOMPAT_64BIT)) {
3995 ext4_msg(sb, KERN_ERR, "Failed to set 64-bit journal feature");
3996 goto failed_mount_wq;
3999 if (!set_journal_csum_feature_set(sb)) {
4000 ext4_msg(sb, KERN_ERR, "Failed to set journal checksum "
4002 goto failed_mount_wq;
4005 /* We have now updated the journal if required, so we can
4006 * validate the data journaling mode. */
4007 switch (test_opt(sb, DATA_FLAGS)) {
4009 /* No mode set, assume a default based on the journal
4010 * capabilities: ORDERED_DATA if the journal can
4011 * cope, else JOURNAL_DATA
4013 if (jbd2_journal_check_available_features
4014 (sbi->s_journal, 0, 0, JBD2_FEATURE_INCOMPAT_REVOKE))
4015 set_opt(sb, ORDERED_DATA);
4017 set_opt(sb, JOURNAL_DATA);
4020 case EXT4_MOUNT_ORDERED_DATA:
4021 case EXT4_MOUNT_WRITEBACK_DATA:
4022 if (!jbd2_journal_check_available_features
4023 (sbi->s_journal, 0, 0, JBD2_FEATURE_INCOMPAT_REVOKE)) {
4024 ext4_msg(sb, KERN_ERR, "Journal does not support "
4025 "requested data journaling mode");
4026 goto failed_mount_wq;
4031 set_task_ioprio(sbi->s_journal->j_task, journal_ioprio);
4033 sbi->s_journal->j_commit_callback = ext4_journal_commit_callback;
4036 if (ext4_mballoc_ready) {
4037 sbi->s_mb_cache = ext4_xattr_create_cache(sb->s_id);
4038 if (!sbi->s_mb_cache) {
4039 ext4_msg(sb, KERN_ERR, "Failed to create an mb_cache");
4040 goto failed_mount_wq;
4045 * Get the # of file system overhead blocks from the
4046 * superblock if present.
4048 if (es->s_overhead_clusters)
4049 sbi->s_overhead = le32_to_cpu(es->s_overhead_clusters);
4051 err = ext4_calculate_overhead(sb);
4053 goto failed_mount_wq;
4057 * The maximum number of concurrent works can be high and
4058 * concurrency isn't really necessary. Limit it to 1.
4060 EXT4_SB(sb)->rsv_conversion_wq =
4061 alloc_workqueue("ext4-rsv-conversion", WQ_MEM_RECLAIM | WQ_UNBOUND, 1);
4062 if (!EXT4_SB(sb)->rsv_conversion_wq) {
4063 printk(KERN_ERR "EXT4-fs: failed to create workqueue\n");
4069 * The jbd2_journal_load will have done any necessary log recovery,
4070 * so we can safely mount the rest of the filesystem now.
4073 root = ext4_iget(sb, EXT4_ROOT_INO);
4075 ext4_msg(sb, KERN_ERR, "get root inode failed");
4076 ret = PTR_ERR(root);
4080 if (!S_ISDIR(root->i_mode) || !root->i_blocks || !root->i_size) {
4081 ext4_msg(sb, KERN_ERR, "corrupt root inode, run e2fsck");
4085 sb->s_root = d_make_root(root);
4087 ext4_msg(sb, KERN_ERR, "get root dentry failed");
4092 if (ext4_setup_super(sb, es, sb->s_flags & MS_RDONLY))
4093 sb->s_flags |= MS_RDONLY;
4095 /* determine the minimum size of new large inodes, if present */
4096 if (sbi->s_inode_size > EXT4_GOOD_OLD_INODE_SIZE) {
4097 sbi->s_want_extra_isize = sizeof(struct ext4_inode) -
4098 EXT4_GOOD_OLD_INODE_SIZE;
4099 if (EXT4_HAS_RO_COMPAT_FEATURE(sb,
4100 EXT4_FEATURE_RO_COMPAT_EXTRA_ISIZE)) {
4101 if (sbi->s_want_extra_isize <
4102 le16_to_cpu(es->s_want_extra_isize))
4103 sbi->s_want_extra_isize =
4104 le16_to_cpu(es->s_want_extra_isize);
4105 if (sbi->s_want_extra_isize <
4106 le16_to_cpu(es->s_min_extra_isize))
4107 sbi->s_want_extra_isize =
4108 le16_to_cpu(es->s_min_extra_isize);
4111 /* Check if enough inode space is available */
4112 if (EXT4_GOOD_OLD_INODE_SIZE + sbi->s_want_extra_isize >
4113 sbi->s_inode_size) {
4114 sbi->s_want_extra_isize = sizeof(struct ext4_inode) -
4115 EXT4_GOOD_OLD_INODE_SIZE;
4116 ext4_msg(sb, KERN_INFO, "required extra inode space not"
4120 err = ext4_reserve_clusters(sbi, ext4_calculate_resv_clusters(sb));
4122 ext4_msg(sb, KERN_ERR, "failed to reserve %llu clusters for "
4123 "reserved pool", ext4_calculate_resv_clusters(sb));
4124 goto failed_mount4a;
4127 err = ext4_setup_system_zone(sb);
4129 ext4_msg(sb, KERN_ERR, "failed to initialize system "
4131 goto failed_mount4a;
4135 err = ext4_mb_init(sb);
4137 ext4_msg(sb, KERN_ERR, "failed to initialize mballoc (%d)",
4142 block = ext4_count_free_clusters(sb);
4143 ext4_free_blocks_count_set(sbi->s_es,
4144 EXT4_C2B(sbi, block));
4145 err = percpu_counter_init(&sbi->s_freeclusters_counter, block);
4147 unsigned long freei = ext4_count_free_inodes(sb);
4148 sbi->s_es->s_free_inodes_count = cpu_to_le32(freei);
4149 err = percpu_counter_init(&sbi->s_freeinodes_counter, freei);
4152 err = percpu_counter_init(&sbi->s_dirs_counter,
4153 ext4_count_dirs(sb));
4155 err = percpu_counter_init(&sbi->s_dirtyclusters_counter, 0);
4157 ext4_msg(sb, KERN_ERR, "insufficient memory");
4161 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_FLEX_BG))
4162 if (!ext4_fill_flex_info(sb)) {
4163 ext4_msg(sb, KERN_ERR,
4164 "unable to initialize "
4165 "flex_bg meta info!");
4169 err = ext4_register_li_request(sb, first_not_zeroed);
4173 sbi->s_kobj.kset = ext4_kset;
4174 init_completion(&sbi->s_kobj_unregister);
4175 err = kobject_init_and_add(&sbi->s_kobj, &ext4_ktype, NULL,
4181 /* Enable quota usage during mount. */
4182 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_QUOTA) &&
4183 !(sb->s_flags & MS_RDONLY)) {
4184 err = ext4_enable_quotas(sb);
4188 #endif /* CONFIG_QUOTA */
4190 EXT4_SB(sb)->s_mount_state |= EXT4_ORPHAN_FS;
4191 ext4_orphan_cleanup(sb, es);
4192 EXT4_SB(sb)->s_mount_state &= ~EXT4_ORPHAN_FS;
4193 if (needs_recovery) {
4194 ext4_msg(sb, KERN_INFO, "recovery complete");
4195 ext4_mark_recovery_complete(sb, es);
4197 if (EXT4_SB(sb)->s_journal) {
4198 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA)
4199 descr = " journalled data mode";
4200 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA)
4201 descr = " ordered data mode";
4203 descr = " writeback data mode";
4205 descr = "out journal";
4207 if (test_opt(sb, DISCARD)) {
4208 struct request_queue *q = bdev_get_queue(sb->s_bdev);
4209 if (!blk_queue_discard(q))
4210 ext4_msg(sb, KERN_WARNING,
4211 "mounting with \"discard\" option, but "
4212 "the device does not support discard");
4215 ext4_msg(sb, KERN_INFO, "mounted filesystem with%s. "
4216 "Opts: %s%s%s", descr, sbi->s_es->s_mount_opts,
4217 *sbi->s_es->s_mount_opts ? "; " : "", orig_data);
4219 if (es->s_error_count)
4220 mod_timer(&sbi->s_err_report, jiffies + 300*HZ); /* 5 minutes */
4222 /* Enable message ratelimiting. Default is 10 messages per 5 secs. */
4223 ratelimit_state_init(&sbi->s_err_ratelimit_state, 5 * HZ, 10);
4224 ratelimit_state_init(&sbi->s_warning_ratelimit_state, 5 * HZ, 10);
4225 ratelimit_state_init(&sbi->s_msg_ratelimit_state, 5 * HZ, 10);
4232 ext4_msg(sb, KERN_ERR, "VFS: Can't find ext4 filesystem");
4237 kobject_del(&sbi->s_kobj);
4240 ext4_unregister_li_request(sb);
4242 ext4_mb_release(sb);
4243 if (sbi->s_flex_groups)
4244 ext4_kvfree(sbi->s_flex_groups);
4245 percpu_counter_destroy(&sbi->s_freeclusters_counter);
4246 percpu_counter_destroy(&sbi->s_freeinodes_counter);
4247 percpu_counter_destroy(&sbi->s_dirs_counter);
4248 percpu_counter_destroy(&sbi->s_dirtyclusters_counter);
4250 ext4_ext_release(sb);
4251 ext4_release_system_zone(sb);
4256 ext4_msg(sb, KERN_ERR, "mount failed");
4257 if (EXT4_SB(sb)->rsv_conversion_wq)
4258 destroy_workqueue(EXT4_SB(sb)->rsv_conversion_wq);
4260 if (sbi->s_journal) {
4261 jbd2_journal_destroy(sbi->s_journal);
4262 sbi->s_journal = NULL;
4264 ext4_es_unregister_shrinker(sbi);
4266 del_timer_sync(&sbi->s_err_report);
4268 kthread_stop(sbi->s_mmp_tsk);
4270 for (i = 0; i < db_count; i++)
4271 brelse(sbi->s_group_desc[i]);
4272 ext4_kvfree(sbi->s_group_desc);
4274 if (sbi->s_chksum_driver)
4275 crypto_free_shash(sbi->s_chksum_driver);
4277 remove_proc_entry("options", sbi->s_proc);
4278 remove_proc_entry(sb->s_id, ext4_proc_root);
4281 for (i = 0; i < EXT4_MAXQUOTAS; i++)
4282 kfree(sbi->s_qf_names[i]);
4284 ext4_blkdev_remove(sbi);
4287 sb->s_fs_info = NULL;
4288 kfree(sbi->s_blockgroup_lock);
4292 return err ? err : ret;
4296 * Setup any per-fs journal parameters now. We'll do this both on
4297 * initial mount, once the journal has been initialised but before we've
4298 * done any recovery; and again on any subsequent remount.
4300 static void ext4_init_journal_params(struct super_block *sb, journal_t *journal)
4302 struct ext4_sb_info *sbi = EXT4_SB(sb);
4304 journal->j_commit_interval = sbi->s_commit_interval;
4305 journal->j_min_batch_time = sbi->s_min_batch_time;
4306 journal->j_max_batch_time = sbi->s_max_batch_time;
4308 write_lock(&journal->j_state_lock);
4309 if (test_opt(sb, BARRIER))
4310 journal->j_flags |= JBD2_BARRIER;
4312 journal->j_flags &= ~JBD2_BARRIER;
4313 if (test_opt(sb, DATA_ERR_ABORT))
4314 journal->j_flags |= JBD2_ABORT_ON_SYNCDATA_ERR;
4316 journal->j_flags &= ~JBD2_ABORT_ON_SYNCDATA_ERR;
4317 write_unlock(&journal->j_state_lock);
4320 static journal_t *ext4_get_journal(struct super_block *sb,
4321 unsigned int journal_inum)
4323 struct inode *journal_inode;
4326 BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL));
4328 /* First, test for the existence of a valid inode on disk. Bad
4329 * things happen if we iget() an unused inode, as the subsequent
4330 * iput() will try to delete it. */
4332 journal_inode = ext4_iget(sb, journal_inum);
4333 if (IS_ERR(journal_inode)) {
4334 ext4_msg(sb, KERN_ERR, "no journal found");
4337 if (!journal_inode->i_nlink) {
4338 make_bad_inode(journal_inode);
4339 iput(journal_inode);
4340 ext4_msg(sb, KERN_ERR, "journal inode is deleted");
4344 jbd_debug(2, "Journal inode found at %p: %lld bytes\n",
4345 journal_inode, journal_inode->i_size);
4346 if (!S_ISREG(journal_inode->i_mode)) {
4347 ext4_msg(sb, KERN_ERR, "invalid journal inode");
4348 iput(journal_inode);
4352 journal = jbd2_journal_init_inode(journal_inode);
4354 ext4_msg(sb, KERN_ERR, "Could not load journal inode");
4355 iput(journal_inode);
4358 journal->j_private = sb;
4359 ext4_init_journal_params(sb, journal);
4363 static journal_t *ext4_get_dev_journal(struct super_block *sb,
4366 struct buffer_head *bh;
4370 int hblock, blocksize;
4371 ext4_fsblk_t sb_block;
4372 unsigned long offset;
4373 struct ext4_super_block *es;
4374 struct block_device *bdev;
4376 BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL));
4378 bdev = ext4_blkdev_get(j_dev, sb);
4382 blocksize = sb->s_blocksize;
4383 hblock = bdev_logical_block_size(bdev);
4384 if (blocksize < hblock) {
4385 ext4_msg(sb, KERN_ERR,
4386 "blocksize too small for journal device");
4390 sb_block = EXT4_MIN_BLOCK_SIZE / blocksize;
4391 offset = EXT4_MIN_BLOCK_SIZE % blocksize;
4392 set_blocksize(bdev, blocksize);
4393 if (!(bh = __bread(bdev, sb_block, blocksize))) {
4394 ext4_msg(sb, KERN_ERR, "couldn't read superblock of "
4395 "external journal");
4399 es = (struct ext4_super_block *) (bh->b_data + offset);
4400 if ((le16_to_cpu(es->s_magic) != EXT4_SUPER_MAGIC) ||
4401 !(le32_to_cpu(es->s_feature_incompat) &
4402 EXT4_FEATURE_INCOMPAT_JOURNAL_DEV)) {
4403 ext4_msg(sb, KERN_ERR, "external journal has "
4409 if ((le32_to_cpu(es->s_feature_ro_compat) &
4410 EXT4_FEATURE_RO_COMPAT_METADATA_CSUM) &&
4411 es->s_checksum != ext4_superblock_csum(sb, es)) {
4412 ext4_msg(sb, KERN_ERR, "external journal has "
4413 "corrupt superblock");
4418 if (memcmp(EXT4_SB(sb)->s_es->s_journal_uuid, es->s_uuid, 16)) {
4419 ext4_msg(sb, KERN_ERR, "journal UUID does not match");
4424 len = ext4_blocks_count(es);
4425 start = sb_block + 1;
4426 brelse(bh); /* we're done with the superblock */
4428 journal = jbd2_journal_init_dev(bdev, sb->s_bdev,
4429 start, len, blocksize);
4431 ext4_msg(sb, KERN_ERR, "failed to create device journal");
4434 journal->j_private = sb;
4435 ll_rw_block(READ | REQ_META | REQ_PRIO, 1, &journal->j_sb_buffer);
4436 wait_on_buffer(journal->j_sb_buffer);
4437 if (!buffer_uptodate(journal->j_sb_buffer)) {
4438 ext4_msg(sb, KERN_ERR, "I/O error on journal device");
4441 if (be32_to_cpu(journal->j_superblock->s_nr_users) != 1) {
4442 ext4_msg(sb, KERN_ERR, "External journal has more than one "
4443 "user (unsupported) - %d",
4444 be32_to_cpu(journal->j_superblock->s_nr_users));
4447 EXT4_SB(sb)->journal_bdev = bdev;
4448 ext4_init_journal_params(sb, journal);
4452 jbd2_journal_destroy(journal);
4454 ext4_blkdev_put(bdev);
4458 static int ext4_load_journal(struct super_block *sb,
4459 struct ext4_super_block *es,
4460 unsigned long journal_devnum)
4463 unsigned int journal_inum = le32_to_cpu(es->s_journal_inum);
4466 int really_read_only;
4468 BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL));
4470 if (journal_devnum &&
4471 journal_devnum != le32_to_cpu(es->s_journal_dev)) {
4472 ext4_msg(sb, KERN_INFO, "external journal device major/minor "
4473 "numbers have changed");
4474 journal_dev = new_decode_dev(journal_devnum);
4476 journal_dev = new_decode_dev(le32_to_cpu(es->s_journal_dev));
4478 really_read_only = bdev_read_only(sb->s_bdev);
4481 * Are we loading a blank journal or performing recovery after a
4482 * crash? For recovery, we need to check in advance whether we
4483 * can get read-write access to the device.
4485 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER)) {
4486 if (sb->s_flags & MS_RDONLY) {
4487 ext4_msg(sb, KERN_INFO, "INFO: recovery "
4488 "required on readonly filesystem");
4489 if (really_read_only) {
4490 ext4_msg(sb, KERN_ERR, "write access "
4491 "unavailable, cannot proceed");
4494 ext4_msg(sb, KERN_INFO, "write access will "
4495 "be enabled during recovery");
4499 if (journal_inum && journal_dev) {
4500 ext4_msg(sb, KERN_ERR, "filesystem has both journal "
4501 "and inode journals!");
4506 if (!(journal = ext4_get_journal(sb, journal_inum)))
4509 if (!(journal = ext4_get_dev_journal(sb, journal_dev)))
4513 if (!(journal->j_flags & JBD2_BARRIER))
4514 ext4_msg(sb, KERN_INFO, "barriers disabled");
4516 if (!EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER))
4517 err = jbd2_journal_wipe(journal, !really_read_only);
4519 char *save = kmalloc(EXT4_S_ERR_LEN, GFP_KERNEL);
4521 memcpy(save, ((char *) es) +
4522 EXT4_S_ERR_START, EXT4_S_ERR_LEN);
4523 err = jbd2_journal_load(journal);
4525 memcpy(((char *) es) + EXT4_S_ERR_START,
4526 save, EXT4_S_ERR_LEN);
4531 ext4_msg(sb, KERN_ERR, "error loading journal");
4532 jbd2_journal_destroy(journal);
4536 EXT4_SB(sb)->s_journal = journal;
4537 ext4_clear_journal_err(sb, es);
4539 if (!really_read_only && journal_devnum &&
4540 journal_devnum != le32_to_cpu(es->s_journal_dev)) {
4541 es->s_journal_dev = cpu_to_le32(journal_devnum);
4543 /* Make sure we flush the recovery flag to disk. */
4544 ext4_commit_super(sb, 1);
4550 static int ext4_commit_super(struct super_block *sb, int sync)
4552 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
4553 struct buffer_head *sbh = EXT4_SB(sb)->s_sbh;
4556 if (!sbh || block_device_ejected(sb))
4558 if (buffer_write_io_error(sbh)) {
4560 * Oh, dear. A previous attempt to write the
4561 * superblock failed. This could happen because the
4562 * USB device was yanked out. Or it could happen to
4563 * be a transient write error and maybe the block will
4564 * be remapped. Nothing we can do but to retry the
4565 * write and hope for the best.
4567 ext4_msg(sb, KERN_ERR, "previous I/O error to "
4568 "superblock detected");
4569 clear_buffer_write_io_error(sbh);
4570 set_buffer_uptodate(sbh);
4573 * If the file system is mounted read-only, don't update the
4574 * superblock write time. This avoids updating the superblock
4575 * write time when we are mounting the root file system
4576 * read/only but we need to replay the journal; at that point,
4577 * for people who are east of GMT and who make their clock
4578 * tick in localtime for Windows bug-for-bug compatibility,
4579 * the clock is set in the future, and this will cause e2fsck
4580 * to complain and force a full file system check.
4582 if (!(sb->s_flags & MS_RDONLY))
4583 es->s_wtime = cpu_to_le32(get_seconds());
4584 if (sb->s_bdev->bd_part)
4585 es->s_kbytes_written =
4586 cpu_to_le64(EXT4_SB(sb)->s_kbytes_written +
4587 ((part_stat_read(sb->s_bdev->bd_part, sectors[1]) -
4588 EXT4_SB(sb)->s_sectors_written_start) >> 1));
4590 es->s_kbytes_written =
4591 cpu_to_le64(EXT4_SB(sb)->s_kbytes_written);
4592 if (percpu_counter_initialized(&EXT4_SB(sb)->s_freeclusters_counter))
4593 ext4_free_blocks_count_set(es,
4594 EXT4_C2B(EXT4_SB(sb), percpu_counter_sum_positive(
4595 &EXT4_SB(sb)->s_freeclusters_counter)));
4596 if (percpu_counter_initialized(&EXT4_SB(sb)->s_freeinodes_counter))
4597 es->s_free_inodes_count =
4598 cpu_to_le32(percpu_counter_sum_positive(
4599 &EXT4_SB(sb)->s_freeinodes_counter));
4600 BUFFER_TRACE(sbh, "marking dirty");
4601 ext4_superblock_csum_set(sb);
4602 mark_buffer_dirty(sbh);
4604 error = sync_dirty_buffer(sbh);
4608 error = buffer_write_io_error(sbh);
4610 ext4_msg(sb, KERN_ERR, "I/O error while writing "
4612 clear_buffer_write_io_error(sbh);
4613 set_buffer_uptodate(sbh);
4620 * Have we just finished recovery? If so, and if we are mounting (or
4621 * remounting) the filesystem readonly, then we will end up with a
4622 * consistent fs on disk. Record that fact.
4624 static void ext4_mark_recovery_complete(struct super_block *sb,
4625 struct ext4_super_block *es)
4627 journal_t *journal = EXT4_SB(sb)->s_journal;
4629 if (!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL)) {
4630 BUG_ON(journal != NULL);
4633 jbd2_journal_lock_updates(journal);
4634 if (jbd2_journal_flush(journal) < 0)
4637 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER) &&
4638 sb->s_flags & MS_RDONLY) {
4639 EXT4_CLEAR_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
4640 ext4_commit_super(sb, 1);
4644 jbd2_journal_unlock_updates(journal);
4648 * If we are mounting (or read-write remounting) a filesystem whose journal
4649 * has recorded an error from a previous lifetime, move that error to the
4650 * main filesystem now.
4652 static void ext4_clear_journal_err(struct super_block *sb,
4653 struct ext4_super_block *es)
4659 BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL));
4661 journal = EXT4_SB(sb)->s_journal;
4664 * Now check for any error status which may have been recorded in the
4665 * journal by a prior ext4_error() or ext4_abort()
4668 j_errno = jbd2_journal_errno(journal);
4672 errstr = ext4_decode_error(sb, j_errno, nbuf);
4673 ext4_warning(sb, "Filesystem error recorded "
4674 "from previous mount: %s", errstr);
4675 ext4_warning(sb, "Marking fs in need of filesystem check.");
4677 EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS;
4678 es->s_state |= cpu_to_le16(EXT4_ERROR_FS);
4679 ext4_commit_super(sb, 1);
4681 jbd2_journal_clear_err(journal);
4682 jbd2_journal_update_sb_errno(journal);
4687 * Force the running and committing transactions to commit,
4688 * and wait on the commit.
4690 int ext4_force_commit(struct super_block *sb)
4694 if (sb->s_flags & MS_RDONLY)
4697 journal = EXT4_SB(sb)->s_journal;
4698 return ext4_journal_force_commit(journal);
4701 static int ext4_sync_fs(struct super_block *sb, int wait)
4705 bool needs_barrier = false;
4706 struct ext4_sb_info *sbi = EXT4_SB(sb);
4708 trace_ext4_sync_fs(sb, wait);
4709 flush_workqueue(sbi->rsv_conversion_wq);
4711 * Writeback quota in non-journalled quota case - journalled quota has
4714 dquot_writeback_dquots(sb, -1);
4716 * Data writeback is possible w/o journal transaction, so barrier must
4717 * being sent at the end of the function. But we can skip it if
4718 * transaction_commit will do it for us.
4720 if (sbi->s_journal) {
4721 target = jbd2_get_latest_transaction(sbi->s_journal);
4722 if (wait && sbi->s_journal->j_flags & JBD2_BARRIER &&
4723 !jbd2_trans_will_send_data_barrier(sbi->s_journal, target))
4724 needs_barrier = true;
4726 if (jbd2_journal_start_commit(sbi->s_journal, &target)) {
4728 ret = jbd2_log_wait_commit(sbi->s_journal,
4731 } else if (wait && test_opt(sb, BARRIER))
4732 needs_barrier = true;
4733 if (needs_barrier) {
4735 err = blkdev_issue_flush(sb->s_bdev, GFP_KERNEL, NULL);
4744 * LVM calls this function before a (read-only) snapshot is created. This
4745 * gives us a chance to flush the journal completely and mark the fs clean.
4747 * Note that only this function cannot bring a filesystem to be in a clean
4748 * state independently. It relies on upper layer to stop all data & metadata
4751 static int ext4_freeze(struct super_block *sb)
4756 if (sb->s_flags & MS_RDONLY)
4759 journal = EXT4_SB(sb)->s_journal;
4761 /* Now we set up the journal barrier. */
4762 jbd2_journal_lock_updates(journal);
4765 * Don't clear the needs_recovery flag if we failed to flush
4768 error = jbd2_journal_flush(journal);
4772 /* Journal blocked and flushed, clear needs_recovery flag. */
4773 EXT4_CLEAR_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
4774 error = ext4_commit_super(sb, 1);
4776 /* we rely on upper layer to stop further updates */
4777 jbd2_journal_unlock_updates(EXT4_SB(sb)->s_journal);
4782 * Called by LVM after the snapshot is done. We need to reset the RECOVER
4783 * flag here, even though the filesystem is not technically dirty yet.
4785 static int ext4_unfreeze(struct super_block *sb)
4787 if (sb->s_flags & MS_RDONLY)
4790 /* Reset the needs_recovery flag before the fs is unlocked. */
4791 EXT4_SET_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
4792 ext4_commit_super(sb, 1);
4797 * Structure to save mount options for ext4_remount's benefit
4799 struct ext4_mount_options {
4800 unsigned long s_mount_opt;
4801 unsigned long s_mount_opt2;
4804 unsigned long s_commit_interval;
4805 u32 s_min_batch_time, s_max_batch_time;
4808 char *s_qf_names[EXT4_MAXQUOTAS];
4812 static int ext4_remount(struct super_block *sb, int *flags, char *data)
4814 struct ext4_super_block *es;
4815 struct ext4_sb_info *sbi = EXT4_SB(sb);
4816 unsigned long old_sb_flags;
4817 struct ext4_mount_options old_opts;
4818 int enable_quota = 0;
4820 unsigned int journal_ioprio = DEFAULT_JOURNAL_IOPRIO;
4825 char *orig_data = kstrdup(data, GFP_KERNEL);
4827 /* Store the original options */
4828 old_sb_flags = sb->s_flags;
4829 old_opts.s_mount_opt = sbi->s_mount_opt;
4830 old_opts.s_mount_opt2 = sbi->s_mount_opt2;
4831 old_opts.s_resuid = sbi->s_resuid;
4832 old_opts.s_resgid = sbi->s_resgid;
4833 old_opts.s_commit_interval = sbi->s_commit_interval;
4834 old_opts.s_min_batch_time = sbi->s_min_batch_time;
4835 old_opts.s_max_batch_time = sbi->s_max_batch_time;
4837 old_opts.s_jquota_fmt = sbi->s_jquota_fmt;
4838 for (i = 0; i < EXT4_MAXQUOTAS; i++)
4839 if (sbi->s_qf_names[i]) {
4840 old_opts.s_qf_names[i] = kstrdup(sbi->s_qf_names[i],
4842 if (!old_opts.s_qf_names[i]) {
4843 for (j = 0; j < i; j++)
4844 kfree(old_opts.s_qf_names[j]);
4849 old_opts.s_qf_names[i] = NULL;
4851 if (sbi->s_journal && sbi->s_journal->j_task->io_context)
4852 journal_ioprio = sbi->s_journal->j_task->io_context->ioprio;
4855 * Allow the "check" option to be passed as a remount option.
4857 if (!parse_options(data, sb, NULL, &journal_ioprio, 1)) {
4862 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA) {
4863 if (test_opt2(sb, EXPLICIT_DELALLOC)) {
4864 ext4_msg(sb, KERN_ERR, "can't mount with "
4865 "both data=journal and delalloc");
4869 if (test_opt(sb, DIOREAD_NOLOCK)) {
4870 ext4_msg(sb, KERN_ERR, "can't mount with "
4871 "both data=journal and dioread_nolock");
4877 if (sbi->s_mount_flags & EXT4_MF_FS_ABORTED)
4878 ext4_abort(sb, "Abort forced by user");
4880 sb->s_flags = (sb->s_flags & ~MS_POSIXACL) |
4881 (test_opt(sb, POSIX_ACL) ? MS_POSIXACL : 0);
4885 if (sbi->s_journal) {
4886 ext4_init_journal_params(sb, sbi->s_journal);
4887 set_task_ioprio(sbi->s_journal->j_task, journal_ioprio);
4890 if ((*flags & MS_RDONLY) != (sb->s_flags & MS_RDONLY)) {
4891 if (sbi->s_mount_flags & EXT4_MF_FS_ABORTED) {
4896 if (*flags & MS_RDONLY) {
4897 err = sync_filesystem(sb);
4900 err = dquot_suspend(sb, -1);
4905 * First of all, the unconditional stuff we have to do
4906 * to disable replay of the journal when we next remount
4908 sb->s_flags |= MS_RDONLY;
4911 * OK, test if we are remounting a valid rw partition
4912 * readonly, and if so set the rdonly flag and then
4913 * mark the partition as valid again.
4915 if (!(es->s_state & cpu_to_le16(EXT4_VALID_FS)) &&
4916 (sbi->s_mount_state & EXT4_VALID_FS))
4917 es->s_state = cpu_to_le16(sbi->s_mount_state);
4920 ext4_mark_recovery_complete(sb, es);
4922 /* Make sure we can mount this feature set readwrite */
4923 if (!ext4_feature_set_ok(sb, 0)) {
4928 * Make sure the group descriptor checksums
4929 * are sane. If they aren't, refuse to remount r/w.
4931 for (g = 0; g < sbi->s_groups_count; g++) {
4932 struct ext4_group_desc *gdp =
4933 ext4_get_group_desc(sb, g, NULL);
4935 if (!ext4_group_desc_csum_verify(sb, g, gdp)) {
4936 ext4_msg(sb, KERN_ERR,
4937 "ext4_remount: Checksum for group %u failed (%u!=%u)",
4938 g, le16_to_cpu(ext4_group_desc_csum(sbi, g, gdp)),
4939 le16_to_cpu(gdp->bg_checksum));
4946 * If we have an unprocessed orphan list hanging
4947 * around from a previously readonly bdev mount,
4948 * require a full umount/remount for now.
4950 if (es->s_last_orphan) {
4951 ext4_msg(sb, KERN_WARNING, "Couldn't "
4952 "remount RDWR because of unprocessed "
4953 "orphan inode list. Please "
4954 "umount/remount instead");
4960 * Mounting a RDONLY partition read-write, so reread
4961 * and store the current valid flag. (It may have
4962 * been changed by e2fsck since we originally mounted
4966 ext4_clear_journal_err(sb, es);
4967 sbi->s_mount_state = le16_to_cpu(es->s_state);
4968 if (!ext4_setup_super(sb, es, 0))
4969 sb->s_flags &= ~MS_RDONLY;
4970 if (EXT4_HAS_INCOMPAT_FEATURE(sb,
4971 EXT4_FEATURE_INCOMPAT_MMP))
4972 if (ext4_multi_mount_protect(sb,
4973 le64_to_cpu(es->s_mmp_block))) {
4982 * Reinitialize lazy itable initialization thread based on
4985 if ((sb->s_flags & MS_RDONLY) || !test_opt(sb, INIT_INODE_TABLE))
4986 ext4_unregister_li_request(sb);
4988 ext4_group_t first_not_zeroed;
4989 first_not_zeroed = ext4_has_uninit_itable(sb);
4990 ext4_register_li_request(sb, first_not_zeroed);
4993 ext4_setup_system_zone(sb);
4994 if (sbi->s_journal == NULL && !(old_sb_flags & MS_RDONLY))
4995 ext4_commit_super(sb, 1);
4998 /* Release old quota file names */
4999 for (i = 0; i < EXT4_MAXQUOTAS; i++)
5000 kfree(old_opts.s_qf_names[i]);
5002 if (sb_any_quota_suspended(sb))
5003 dquot_resume(sb, -1);
5004 else if (EXT4_HAS_RO_COMPAT_FEATURE(sb,
5005 EXT4_FEATURE_RO_COMPAT_QUOTA)) {
5006 err = ext4_enable_quotas(sb);
5013 ext4_msg(sb, KERN_INFO, "re-mounted. Opts: %s", orig_data);
5018 sb->s_flags = old_sb_flags;
5019 sbi->s_mount_opt = old_opts.s_mount_opt;
5020 sbi->s_mount_opt2 = old_opts.s_mount_opt2;
5021 sbi->s_resuid = old_opts.s_resuid;
5022 sbi->s_resgid = old_opts.s_resgid;
5023 sbi->s_commit_interval = old_opts.s_commit_interval;
5024 sbi->s_min_batch_time = old_opts.s_min_batch_time;
5025 sbi->s_max_batch_time = old_opts.s_max_batch_time;
5027 sbi->s_jquota_fmt = old_opts.s_jquota_fmt;
5028 for (i = 0; i < EXT4_MAXQUOTAS; i++) {
5029 kfree(sbi->s_qf_names[i]);
5030 sbi->s_qf_names[i] = old_opts.s_qf_names[i];
5037 static int ext4_statfs(struct dentry *dentry, struct kstatfs *buf)
5039 struct super_block *sb = dentry->d_sb;
5040 struct ext4_sb_info *sbi = EXT4_SB(sb);
5041 struct ext4_super_block *es = sbi->s_es;
5042 ext4_fsblk_t overhead = 0, resv_blocks;
5045 resv_blocks = EXT4_C2B(sbi, atomic64_read(&sbi->s_resv_clusters));
5047 if (!test_opt(sb, MINIX_DF))
5048 overhead = sbi->s_overhead;
5050 buf->f_type = EXT4_SUPER_MAGIC;
5051 buf->f_bsize = sb->s_blocksize;
5052 buf->f_blocks = ext4_blocks_count(es) - EXT4_C2B(sbi, overhead);
5053 bfree = percpu_counter_sum_positive(&sbi->s_freeclusters_counter) -
5054 percpu_counter_sum_positive(&sbi->s_dirtyclusters_counter);
5055 /* prevent underflow in case that few free space is available */
5056 buf->f_bfree = EXT4_C2B(sbi, max_t(s64, bfree, 0));
5057 buf->f_bavail = buf->f_bfree -
5058 (ext4_r_blocks_count(es) + resv_blocks);
5059 if (buf->f_bfree < (ext4_r_blocks_count(es) + resv_blocks))
5061 buf->f_files = le32_to_cpu(es->s_inodes_count);
5062 buf->f_ffree = percpu_counter_sum_positive(&sbi->s_freeinodes_counter);
5063 buf->f_namelen = EXT4_NAME_LEN;
5064 fsid = le64_to_cpup((void *)es->s_uuid) ^
5065 le64_to_cpup((void *)es->s_uuid + sizeof(u64));
5066 buf->f_fsid.val[0] = fsid & 0xFFFFFFFFUL;
5067 buf->f_fsid.val[1] = (fsid >> 32) & 0xFFFFFFFFUL;
5072 /* Helper function for writing quotas on sync - we need to start transaction
5073 * before quota file is locked for write. Otherwise the are possible deadlocks:
5074 * Process 1 Process 2
5075 * ext4_create() quota_sync()
5076 * jbd2_journal_start() write_dquot()
5077 * dquot_initialize() down(dqio_mutex)
5078 * down(dqio_mutex) jbd2_journal_start()
5084 static inline struct inode *dquot_to_inode(struct dquot *dquot)
5086 return sb_dqopt(dquot->dq_sb)->files[dquot->dq_id.type];
5089 static int ext4_write_dquot(struct dquot *dquot)
5093 struct inode *inode;
5095 inode = dquot_to_inode(dquot);
5096 handle = ext4_journal_start(inode, EXT4_HT_QUOTA,
5097 EXT4_QUOTA_TRANS_BLOCKS(dquot->dq_sb));
5099 return PTR_ERR(handle);
5100 ret = dquot_commit(dquot);
5101 err = ext4_journal_stop(handle);
5107 static int ext4_acquire_dquot(struct dquot *dquot)
5112 handle = ext4_journal_start(dquot_to_inode(dquot), EXT4_HT_QUOTA,
5113 EXT4_QUOTA_INIT_BLOCKS(dquot->dq_sb));
5115 return PTR_ERR(handle);
5116 ret = dquot_acquire(dquot);
5117 err = ext4_journal_stop(handle);
5123 static int ext4_release_dquot(struct dquot *dquot)
5128 handle = ext4_journal_start(dquot_to_inode(dquot), EXT4_HT_QUOTA,
5129 EXT4_QUOTA_DEL_BLOCKS(dquot->dq_sb));
5130 if (IS_ERR(handle)) {
5131 /* Release dquot anyway to avoid endless cycle in dqput() */
5132 dquot_release(dquot);
5133 return PTR_ERR(handle);
5135 ret = dquot_release(dquot);
5136 err = ext4_journal_stop(handle);
5142 static int ext4_mark_dquot_dirty(struct dquot *dquot)
5144 struct super_block *sb = dquot->dq_sb;
5145 struct ext4_sb_info *sbi = EXT4_SB(sb);
5147 /* Are we journaling quotas? */
5148 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_QUOTA) ||
5149 sbi->s_qf_names[USRQUOTA] || sbi->s_qf_names[GRPQUOTA]) {
5150 dquot_mark_dquot_dirty(dquot);
5151 return ext4_write_dquot(dquot);
5153 return dquot_mark_dquot_dirty(dquot);
5157 static int ext4_write_info(struct super_block *sb, int type)
5162 /* Data block + inode block */
5163 handle = ext4_journal_start(sb->s_root->d_inode, EXT4_HT_QUOTA, 2);
5165 return PTR_ERR(handle);
5166 ret = dquot_commit_info(sb, type);
5167 err = ext4_journal_stop(handle);
5174 * Turn on quotas during mount time - we need to find
5175 * the quota file and such...
5177 static int ext4_quota_on_mount(struct super_block *sb, int type)
5179 return dquot_quota_on_mount(sb, EXT4_SB(sb)->s_qf_names[type],
5180 EXT4_SB(sb)->s_jquota_fmt, type);
5184 * Standard function to be called on quota_on
5186 static int ext4_quota_on(struct super_block *sb, int type, int format_id,
5191 if (!test_opt(sb, QUOTA))
5194 /* Quotafile not on the same filesystem? */
5195 if (path->dentry->d_sb != sb)
5197 /* Journaling quota? */
5198 if (EXT4_SB(sb)->s_qf_names[type]) {
5199 /* Quotafile not in fs root? */
5200 if (path->dentry->d_parent != sb->s_root)
5201 ext4_msg(sb, KERN_WARNING,
5202 "Quota file not on filesystem root. "
5203 "Journaled quota will not work");
5207 * When we journal data on quota file, we have to flush journal to see
5208 * all updates to the file when we bypass pagecache...
5210 if (EXT4_SB(sb)->s_journal &&
5211 ext4_should_journal_data(path->dentry->d_inode)) {
5213 * We don't need to lock updates but journal_flush() could
5214 * otherwise be livelocked...
5216 jbd2_journal_lock_updates(EXT4_SB(sb)->s_journal);
5217 err = jbd2_journal_flush(EXT4_SB(sb)->s_journal);
5218 jbd2_journal_unlock_updates(EXT4_SB(sb)->s_journal);
5223 return dquot_quota_on(sb, type, format_id, path);
5226 static int ext4_quota_enable(struct super_block *sb, int type, int format_id,
5230 struct inode *qf_inode;
5231 unsigned long qf_inums[EXT4_MAXQUOTAS] = {
5232 le32_to_cpu(EXT4_SB(sb)->s_es->s_usr_quota_inum),
5233 le32_to_cpu(EXT4_SB(sb)->s_es->s_grp_quota_inum)
5236 BUG_ON(!EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_QUOTA));
5238 if (!qf_inums[type])
5241 qf_inode = ext4_iget(sb, qf_inums[type]);
5242 if (IS_ERR(qf_inode)) {
5243 ext4_error(sb, "Bad quota inode # %lu", qf_inums[type]);
5244 return PTR_ERR(qf_inode);
5247 /* Don't account quota for quota files to avoid recursion */
5248 qf_inode->i_flags |= S_NOQUOTA;
5249 err = dquot_enable(qf_inode, type, format_id, flags);
5255 /* Enable usage tracking for all quota types. */
5256 static int ext4_enable_quotas(struct super_block *sb)
5259 unsigned long qf_inums[EXT4_MAXQUOTAS] = {
5260 le32_to_cpu(EXT4_SB(sb)->s_es->s_usr_quota_inum),
5261 le32_to_cpu(EXT4_SB(sb)->s_es->s_grp_quota_inum)
5264 sb_dqopt(sb)->flags |= DQUOT_QUOTA_SYS_FILE;
5265 for (type = 0; type < EXT4_MAXQUOTAS; type++) {
5266 if (qf_inums[type]) {
5267 err = ext4_quota_enable(sb, type, QFMT_VFS_V1,
5268 DQUOT_USAGE_ENABLED);
5271 "Failed to enable quota tracking "
5272 "(type=%d, err=%d). Please run "
5273 "e2fsck to fix.", type, err);
5282 * quota_on function that is used when QUOTA feature is set.
5284 static int ext4_quota_on_sysfile(struct super_block *sb, int type,
5287 if (!EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_QUOTA))
5291 * USAGE was enabled at mount time. Only need to enable LIMITS now.
5293 return ext4_quota_enable(sb, type, format_id, DQUOT_LIMITS_ENABLED);
5296 static int ext4_quota_off(struct super_block *sb, int type)
5298 struct inode *inode = sb_dqopt(sb)->files[type];
5301 /* Force all delayed allocation blocks to be allocated.
5302 * Caller already holds s_umount sem */
5303 if (test_opt(sb, DELALLOC))
5304 sync_filesystem(sb);
5309 /* Update modification times of quota files when userspace can
5310 * start looking at them */
5311 handle = ext4_journal_start(inode, EXT4_HT_QUOTA, 1);
5314 inode->i_mtime = inode->i_ctime = CURRENT_TIME;
5315 ext4_mark_inode_dirty(handle, inode);
5316 ext4_journal_stop(handle);
5319 return dquot_quota_off(sb, type);
5323 * quota_off function that is used when QUOTA feature is set.
5325 static int ext4_quota_off_sysfile(struct super_block *sb, int type)
5327 if (!EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_QUOTA))
5330 /* Disable only the limits. */
5331 return dquot_disable(sb, type, DQUOT_LIMITS_ENABLED);
5334 /* Read data from quotafile - avoid pagecache and such because we cannot afford
5335 * acquiring the locks... As quota files are never truncated and quota code
5336 * itself serializes the operations (and no one else should touch the files)
5337 * we don't have to be afraid of races */
5338 static ssize_t ext4_quota_read(struct super_block *sb, int type, char *data,
5339 size_t len, loff_t off)
5341 struct inode *inode = sb_dqopt(sb)->files[type];
5342 ext4_lblk_t blk = off >> EXT4_BLOCK_SIZE_BITS(sb);
5343 int offset = off & (sb->s_blocksize - 1);
5346 struct buffer_head *bh;
5347 loff_t i_size = i_size_read(inode);
5351 if (off+len > i_size)
5354 while (toread > 0) {
5355 tocopy = sb->s_blocksize - offset < toread ?
5356 sb->s_blocksize - offset : toread;
5357 bh = ext4_bread(NULL, inode, blk, 0);
5360 if (!bh) /* A hole? */
5361 memset(data, 0, tocopy);
5363 memcpy(data, bh->b_data+offset, tocopy);
5373 /* Write to quotafile (we know the transaction is already started and has
5374 * enough credits) */
5375 static ssize_t ext4_quota_write(struct super_block *sb, int type,
5376 const char *data, size_t len, loff_t off)
5378 struct inode *inode = sb_dqopt(sb)->files[type];
5379 ext4_lblk_t blk = off >> EXT4_BLOCK_SIZE_BITS(sb);
5380 int err, offset = off & (sb->s_blocksize - 1);
5381 struct buffer_head *bh;
5382 handle_t *handle = journal_current_handle();
5384 if (EXT4_SB(sb)->s_journal && !handle) {
5385 ext4_msg(sb, KERN_WARNING, "Quota write (off=%llu, len=%llu)"
5386 " cancelled because transaction is not started",
5387 (unsigned long long)off, (unsigned long long)len);
5391 * Since we account only one data block in transaction credits,
5392 * then it is impossible to cross a block boundary.
5394 if (sb->s_blocksize - offset < len) {
5395 ext4_msg(sb, KERN_WARNING, "Quota write (off=%llu, len=%llu)"
5396 " cancelled because not block aligned",
5397 (unsigned long long)off, (unsigned long long)len);
5401 bh = ext4_bread(handle, inode, blk, 1);
5406 BUFFER_TRACE(bh, "get write access");
5407 err = ext4_journal_get_write_access(handle, bh);
5413 memcpy(bh->b_data+offset, data, len);
5414 flush_dcache_page(bh->b_page);
5416 err = ext4_handle_dirty_metadata(handle, NULL, bh);
5419 if (inode->i_size < off + len) {
5420 i_size_write(inode, off + len);
5421 EXT4_I(inode)->i_disksize = inode->i_size;
5422 ext4_mark_inode_dirty(handle, inode);
5429 static struct dentry *ext4_mount(struct file_system_type *fs_type, int flags,
5430 const char *dev_name, void *data)
5432 return mount_bdev(fs_type, flags, dev_name, data, ext4_fill_super);
5435 #if !defined(CONFIG_EXT2_FS) && !defined(CONFIG_EXT2_FS_MODULE) && defined(CONFIG_EXT4_USE_FOR_EXT23)
5436 static inline void register_as_ext2(void)
5438 int err = register_filesystem(&ext2_fs_type);
5441 "EXT4-fs: Unable to register as ext2 (%d)\n", err);
5444 static inline void unregister_as_ext2(void)
5446 unregister_filesystem(&ext2_fs_type);
5449 static inline int ext2_feature_set_ok(struct super_block *sb)
5451 if (EXT4_HAS_INCOMPAT_FEATURE(sb, ~EXT2_FEATURE_INCOMPAT_SUPP))
5453 if (sb->s_flags & MS_RDONLY)
5455 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, ~EXT2_FEATURE_RO_COMPAT_SUPP))
5460 static inline void register_as_ext2(void) { }
5461 static inline void unregister_as_ext2(void) { }
5462 static inline int ext2_feature_set_ok(struct super_block *sb) { return 0; }
5465 #if !defined(CONFIG_EXT3_FS) && !defined(CONFIG_EXT3_FS_MODULE) && defined(CONFIG_EXT4_USE_FOR_EXT23)
5466 static inline void register_as_ext3(void)
5468 int err = register_filesystem(&ext3_fs_type);
5471 "EXT4-fs: Unable to register as ext3 (%d)\n", err);
5474 static inline void unregister_as_ext3(void)
5476 unregister_filesystem(&ext3_fs_type);
5479 static inline int ext3_feature_set_ok(struct super_block *sb)
5481 if (EXT4_HAS_INCOMPAT_FEATURE(sb, ~EXT3_FEATURE_INCOMPAT_SUPP))
5483 if (!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL))
5485 if (sb->s_flags & MS_RDONLY)
5487 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, ~EXT3_FEATURE_RO_COMPAT_SUPP))
5492 static inline void register_as_ext3(void) { }
5493 static inline void unregister_as_ext3(void) { }
5494 static inline int ext3_feature_set_ok(struct super_block *sb) { return 0; }
5497 static struct file_system_type ext4_fs_type = {
5498 .owner = THIS_MODULE,
5500 .mount = ext4_mount,
5501 .kill_sb = kill_block_super,
5502 .fs_flags = FS_REQUIRES_DEV,
5504 MODULE_ALIAS_FS("ext4");
5506 static int __init ext4_init_feat_adverts(void)
5508 struct ext4_features *ef;
5511 ef = kzalloc(sizeof(struct ext4_features), GFP_KERNEL);
5515 ef->f_kobj.kset = ext4_kset;
5516 init_completion(&ef->f_kobj_unregister);
5517 ret = kobject_init_and_add(&ef->f_kobj, &ext4_feat_ktype, NULL,
5530 static void ext4_exit_feat_adverts(void)
5532 kobject_put(&ext4_feat->f_kobj);
5533 wait_for_completion(&ext4_feat->f_kobj_unregister);
5537 /* Shared across all ext4 file systems */
5538 wait_queue_head_t ext4__ioend_wq[EXT4_WQ_HASH_SZ];
5539 struct mutex ext4__aio_mutex[EXT4_WQ_HASH_SZ];
5541 static int __init ext4_init_fs(void)
5545 ext4_li_info = NULL;
5546 mutex_init(&ext4_li_mtx);
5548 /* Build-time check for flags consistency */
5549 ext4_check_flag_values();
5551 for (i = 0; i < EXT4_WQ_HASH_SZ; i++) {
5552 mutex_init(&ext4__aio_mutex[i]);
5553 init_waitqueue_head(&ext4__ioend_wq[i]);
5556 err = ext4_init_es();
5560 err = ext4_init_pageio();
5564 err = ext4_init_system_zone();
5567 ext4_kset = kset_create_and_add("ext4", NULL, fs_kobj);
5572 ext4_proc_root = proc_mkdir("fs/ext4", NULL);
5574 err = ext4_init_feat_adverts();
5578 err = ext4_init_mballoc();
5582 ext4_mballoc_ready = 1;
5583 err = init_inodecache();
5588 err = register_filesystem(&ext4_fs_type);
5594 unregister_as_ext2();
5595 unregister_as_ext3();
5596 destroy_inodecache();
5598 ext4_mballoc_ready = 0;
5599 ext4_exit_mballoc();
5601 ext4_exit_feat_adverts();
5604 remove_proc_entry("fs/ext4", NULL);
5605 kset_unregister(ext4_kset);
5607 ext4_exit_system_zone();
5616 static void __exit ext4_exit_fs(void)
5618 ext4_destroy_lazyinit_thread();
5619 unregister_as_ext2();
5620 unregister_as_ext3();
5621 unregister_filesystem(&ext4_fs_type);
5622 destroy_inodecache();
5623 ext4_exit_mballoc();
5624 ext4_exit_feat_adverts();
5625 remove_proc_entry("fs/ext4", NULL);
5626 kset_unregister(ext4_kset);
5627 ext4_exit_system_zone();
5632 MODULE_AUTHOR("Remy Card, Stephen Tweedie, Andrew Morton, Andreas Dilger, Theodore Ts'o and others");
5633 MODULE_DESCRIPTION("Fourth Extended Filesystem");
5634 MODULE_LICENSE("GPL");
5635 module_init(ext4_init_fs)
5636 module_exit(ext4_exit_fs)