2 * linux/fs/ext4/super.c
4 * Copyright (C) 1992, 1993, 1994, 1995
5 * Remy Card (card@masi.ibp.fr)
6 * Laboratoire MASI - Institut Blaise Pascal
7 * Universite Pierre et Marie Curie (Paris VI)
11 * linux/fs/minix/inode.c
13 * Copyright (C) 1991, 1992 Linus Torvalds
15 * Big-endian to little-endian byte-swapping/bitmaps by
16 * David S. Miller (davem@caip.rutgers.edu), 1995
19 #include <linux/module.h>
20 #include <linux/string.h>
22 #include <linux/time.h>
23 #include <linux/vmalloc.h>
24 #include <linux/jbd2.h>
25 #include <linux/slab.h>
26 #include <linux/init.h>
27 #include <linux/blkdev.h>
28 #include <linux/parser.h>
29 #include <linux/buffer_head.h>
30 #include <linux/exportfs.h>
31 #include <linux/vfs.h>
32 #include <linux/random.h>
33 #include <linux/mount.h>
34 #include <linux/namei.h>
35 #include <linux/quotaops.h>
36 #include <linux/seq_file.h>
37 #include <linux/proc_fs.h>
38 #include <linux/ctype.h>
39 #include <linux/log2.h>
40 #include <linux/crc16.h>
41 #include <linux/cleancache.h>
42 #include <asm/uaccess.h>
44 #include <linux/kthread.h>
45 #include <linux/freezer.h>
48 #include "ext4_extents.h" /* Needed for trace points definition */
49 #include "ext4_jbd2.h"
54 #define CREATE_TRACE_POINTS
55 #include <trace/events/ext4.h>
57 static struct proc_dir_entry *ext4_proc_root;
58 static struct kset *ext4_kset;
59 static struct ext4_lazy_init *ext4_li_info;
60 static struct mutex ext4_li_mtx;
61 static struct ext4_features *ext4_feat;
63 static int ext4_load_journal(struct super_block *, struct ext4_super_block *,
64 unsigned long journal_devnum);
65 static int ext4_show_options(struct seq_file *seq, struct dentry *root);
66 static int ext4_commit_super(struct super_block *sb, int sync);
67 static void ext4_mark_recovery_complete(struct super_block *sb,
68 struct ext4_super_block *es);
69 static void ext4_clear_journal_err(struct super_block *sb,
70 struct ext4_super_block *es);
71 static int ext4_sync_fs(struct super_block *sb, int wait);
72 static int ext4_remount(struct super_block *sb, int *flags, char *data);
73 static int ext4_statfs(struct dentry *dentry, struct kstatfs *buf);
74 static int ext4_unfreeze(struct super_block *sb);
75 static int ext4_freeze(struct super_block *sb);
76 static struct dentry *ext4_mount(struct file_system_type *fs_type, int flags,
77 const char *dev_name, void *data);
78 static inline int ext2_feature_set_ok(struct super_block *sb);
79 static inline int ext3_feature_set_ok(struct super_block *sb);
80 static int ext4_feature_set_ok(struct super_block *sb, int readonly);
81 static void ext4_destroy_lazyinit_thread(void);
82 static void ext4_unregister_li_request(struct super_block *sb);
83 static void ext4_clear_request_list(void);
84 static int ext4_reserve_clusters(struct ext4_sb_info *, ext4_fsblk_t);
86 #if !defined(CONFIG_EXT2_FS) && !defined(CONFIG_EXT2_FS_MODULE) && defined(CONFIG_EXT4_USE_FOR_EXT23)
87 static struct file_system_type ext2_fs_type = {
91 .kill_sb = kill_block_super,
92 .fs_flags = FS_REQUIRES_DEV,
94 MODULE_ALIAS_FS("ext2");
96 #define IS_EXT2_SB(sb) ((sb)->s_bdev->bd_holder == &ext2_fs_type)
98 #define IS_EXT2_SB(sb) (0)
102 #if !defined(CONFIG_EXT3_FS) && !defined(CONFIG_EXT3_FS_MODULE) && defined(CONFIG_EXT4_USE_FOR_EXT23)
103 static struct file_system_type ext3_fs_type = {
104 .owner = THIS_MODULE,
107 .kill_sb = kill_block_super,
108 .fs_flags = FS_REQUIRES_DEV,
110 MODULE_ALIAS_FS("ext3");
111 MODULE_ALIAS("ext3");
112 #define IS_EXT3_SB(sb) ((sb)->s_bdev->bd_holder == &ext3_fs_type)
114 #define IS_EXT3_SB(sb) (0)
117 static int ext4_verify_csum_type(struct super_block *sb,
118 struct ext4_super_block *es)
120 if (!EXT4_HAS_RO_COMPAT_FEATURE(sb,
121 EXT4_FEATURE_RO_COMPAT_METADATA_CSUM))
124 return es->s_checksum_type == EXT4_CRC32C_CHKSUM;
127 static __le32 ext4_superblock_csum(struct super_block *sb,
128 struct ext4_super_block *es)
130 struct ext4_sb_info *sbi = EXT4_SB(sb);
131 int offset = offsetof(struct ext4_super_block, s_checksum);
134 csum = ext4_chksum(sbi, ~0, (char *)es, offset);
136 return cpu_to_le32(csum);
139 int ext4_superblock_csum_verify(struct super_block *sb,
140 struct ext4_super_block *es)
142 if (!EXT4_HAS_RO_COMPAT_FEATURE(sb,
143 EXT4_FEATURE_RO_COMPAT_METADATA_CSUM))
146 return es->s_checksum == ext4_superblock_csum(sb, es);
149 void ext4_superblock_csum_set(struct super_block *sb)
151 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
153 if (!EXT4_HAS_RO_COMPAT_FEATURE(sb,
154 EXT4_FEATURE_RO_COMPAT_METADATA_CSUM))
157 es->s_checksum = ext4_superblock_csum(sb, es);
160 void *ext4_kvmalloc(size_t size, gfp_t flags)
164 ret = kmalloc(size, flags);
166 ret = __vmalloc(size, flags, PAGE_KERNEL);
170 void *ext4_kvzalloc(size_t size, gfp_t flags)
174 ret = kzalloc(size, flags);
176 ret = __vmalloc(size, flags | __GFP_ZERO, PAGE_KERNEL);
180 void ext4_kvfree(void *ptr)
182 if (is_vmalloc_addr(ptr))
189 ext4_fsblk_t ext4_block_bitmap(struct super_block *sb,
190 struct ext4_group_desc *bg)
192 return le32_to_cpu(bg->bg_block_bitmap_lo) |
193 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
194 (ext4_fsblk_t)le32_to_cpu(bg->bg_block_bitmap_hi) << 32 : 0);
197 ext4_fsblk_t ext4_inode_bitmap(struct super_block *sb,
198 struct ext4_group_desc *bg)
200 return le32_to_cpu(bg->bg_inode_bitmap_lo) |
201 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
202 (ext4_fsblk_t)le32_to_cpu(bg->bg_inode_bitmap_hi) << 32 : 0);
205 ext4_fsblk_t ext4_inode_table(struct super_block *sb,
206 struct ext4_group_desc *bg)
208 return le32_to_cpu(bg->bg_inode_table_lo) |
209 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
210 (ext4_fsblk_t)le32_to_cpu(bg->bg_inode_table_hi) << 32 : 0);
213 __u32 ext4_free_group_clusters(struct super_block *sb,
214 struct ext4_group_desc *bg)
216 return le16_to_cpu(bg->bg_free_blocks_count_lo) |
217 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
218 (__u32)le16_to_cpu(bg->bg_free_blocks_count_hi) << 16 : 0);
221 __u32 ext4_free_inodes_count(struct super_block *sb,
222 struct ext4_group_desc *bg)
224 return le16_to_cpu(bg->bg_free_inodes_count_lo) |
225 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
226 (__u32)le16_to_cpu(bg->bg_free_inodes_count_hi) << 16 : 0);
229 __u32 ext4_used_dirs_count(struct super_block *sb,
230 struct ext4_group_desc *bg)
232 return le16_to_cpu(bg->bg_used_dirs_count_lo) |
233 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
234 (__u32)le16_to_cpu(bg->bg_used_dirs_count_hi) << 16 : 0);
237 __u32 ext4_itable_unused_count(struct super_block *sb,
238 struct ext4_group_desc *bg)
240 return le16_to_cpu(bg->bg_itable_unused_lo) |
241 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
242 (__u32)le16_to_cpu(bg->bg_itable_unused_hi) << 16 : 0);
245 void ext4_block_bitmap_set(struct super_block *sb,
246 struct ext4_group_desc *bg, ext4_fsblk_t blk)
248 bg->bg_block_bitmap_lo = cpu_to_le32((u32)blk);
249 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
250 bg->bg_block_bitmap_hi = cpu_to_le32(blk >> 32);
253 void ext4_inode_bitmap_set(struct super_block *sb,
254 struct ext4_group_desc *bg, ext4_fsblk_t blk)
256 bg->bg_inode_bitmap_lo = cpu_to_le32((u32)blk);
257 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
258 bg->bg_inode_bitmap_hi = cpu_to_le32(blk >> 32);
261 void ext4_inode_table_set(struct super_block *sb,
262 struct ext4_group_desc *bg, ext4_fsblk_t blk)
264 bg->bg_inode_table_lo = cpu_to_le32((u32)blk);
265 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
266 bg->bg_inode_table_hi = cpu_to_le32(blk >> 32);
269 void ext4_free_group_clusters_set(struct super_block *sb,
270 struct ext4_group_desc *bg, __u32 count)
272 bg->bg_free_blocks_count_lo = cpu_to_le16((__u16)count);
273 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
274 bg->bg_free_blocks_count_hi = cpu_to_le16(count >> 16);
277 void ext4_free_inodes_set(struct super_block *sb,
278 struct ext4_group_desc *bg, __u32 count)
280 bg->bg_free_inodes_count_lo = cpu_to_le16((__u16)count);
281 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
282 bg->bg_free_inodes_count_hi = cpu_to_le16(count >> 16);
285 void ext4_used_dirs_set(struct super_block *sb,
286 struct ext4_group_desc *bg, __u32 count)
288 bg->bg_used_dirs_count_lo = cpu_to_le16((__u16)count);
289 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
290 bg->bg_used_dirs_count_hi = cpu_to_le16(count >> 16);
293 void ext4_itable_unused_set(struct super_block *sb,
294 struct ext4_group_desc *bg, __u32 count)
296 bg->bg_itable_unused_lo = cpu_to_le16((__u16)count);
297 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
298 bg->bg_itable_unused_hi = cpu_to_le16(count >> 16);
302 static void __save_error_info(struct super_block *sb, const char *func,
305 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
307 EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS;
308 if (bdev_read_only(sb->s_bdev))
310 es->s_state |= cpu_to_le16(EXT4_ERROR_FS);
311 es->s_last_error_time = cpu_to_le32(get_seconds());
312 strncpy(es->s_last_error_func, func, sizeof(es->s_last_error_func));
313 es->s_last_error_line = cpu_to_le32(line);
314 if (!es->s_first_error_time) {
315 es->s_first_error_time = es->s_last_error_time;
316 strncpy(es->s_first_error_func, func,
317 sizeof(es->s_first_error_func));
318 es->s_first_error_line = cpu_to_le32(line);
319 es->s_first_error_ino = es->s_last_error_ino;
320 es->s_first_error_block = es->s_last_error_block;
323 * Start the daily error reporting function if it hasn't been
326 if (!es->s_error_count)
327 mod_timer(&EXT4_SB(sb)->s_err_report, jiffies + 24*60*60*HZ);
328 le32_add_cpu(&es->s_error_count, 1);
331 static void save_error_info(struct super_block *sb, const char *func,
334 __save_error_info(sb, func, line);
335 ext4_commit_super(sb, 1);
339 * The del_gendisk() function uninitializes the disk-specific data
340 * structures, including the bdi structure, without telling anyone
341 * else. Once this happens, any attempt to call mark_buffer_dirty()
342 * (for example, by ext4_commit_super), will cause a kernel OOPS.
343 * This is a kludge to prevent these oops until we can put in a proper
344 * hook in del_gendisk() to inform the VFS and file system layers.
346 static int block_device_ejected(struct super_block *sb)
348 struct inode *bd_inode = sb->s_bdev->bd_inode;
349 struct backing_dev_info *bdi = bd_inode->i_mapping->backing_dev_info;
351 return bdi->dev == NULL;
354 static void ext4_journal_commit_callback(journal_t *journal, transaction_t *txn)
356 struct super_block *sb = journal->j_private;
357 struct ext4_sb_info *sbi = EXT4_SB(sb);
358 int error = is_journal_aborted(journal);
359 struct ext4_journal_cb_entry *jce;
361 BUG_ON(txn->t_state == T_FINISHED);
362 spin_lock(&sbi->s_md_lock);
363 while (!list_empty(&txn->t_private_list)) {
364 jce = list_entry(txn->t_private_list.next,
365 struct ext4_journal_cb_entry, jce_list);
366 list_del_init(&jce->jce_list);
367 spin_unlock(&sbi->s_md_lock);
368 jce->jce_func(sb, jce, error);
369 spin_lock(&sbi->s_md_lock);
371 spin_unlock(&sbi->s_md_lock);
374 /* Deal with the reporting of failure conditions on a filesystem such as
375 * inconsistencies detected or read IO failures.
377 * On ext2, we can store the error state of the filesystem in the
378 * superblock. That is not possible on ext4, because we may have other
379 * write ordering constraints on the superblock which prevent us from
380 * writing it out straight away; and given that the journal is about to
381 * be aborted, we can't rely on the current, or future, transactions to
382 * write out the superblock safely.
384 * We'll just use the jbd2_journal_abort() error code to record an error in
385 * the journal instead. On recovery, the journal will complain about
386 * that error until we've noted it down and cleared it.
389 static void ext4_handle_error(struct super_block *sb)
391 if (sb->s_flags & MS_RDONLY)
394 if (!test_opt(sb, ERRORS_CONT)) {
395 journal_t *journal = EXT4_SB(sb)->s_journal;
397 EXT4_SB(sb)->s_mount_flags |= EXT4_MF_FS_ABORTED;
399 jbd2_journal_abort(journal, -EIO);
401 if (test_opt(sb, ERRORS_RO)) {
402 ext4_msg(sb, KERN_CRIT, "Remounting filesystem read-only");
403 sb->s_flags |= MS_RDONLY;
405 if (test_opt(sb, ERRORS_PANIC))
406 panic("EXT4-fs (device %s): panic forced after error\n",
410 void __ext4_error(struct super_block *sb, const char *function,
411 unsigned int line, const char *fmt, ...)
413 struct va_format vaf;
419 printk(KERN_CRIT "EXT4-fs error (device %s): %s:%d: comm %s: %pV\n",
420 sb->s_id, function, line, current->comm, &vaf);
422 save_error_info(sb, function, line);
424 ext4_handle_error(sb);
427 void ext4_error_inode(struct inode *inode, const char *function,
428 unsigned int line, ext4_fsblk_t block,
429 const char *fmt, ...)
432 struct va_format vaf;
433 struct ext4_super_block *es = EXT4_SB(inode->i_sb)->s_es;
435 es->s_last_error_ino = cpu_to_le32(inode->i_ino);
436 es->s_last_error_block = cpu_to_le64(block);
437 save_error_info(inode->i_sb, function, line);
442 printk(KERN_CRIT "EXT4-fs error (device %s): %s:%d: "
443 "inode #%lu: block %llu: comm %s: %pV\n",
444 inode->i_sb->s_id, function, line, inode->i_ino,
445 block, current->comm, &vaf);
447 printk(KERN_CRIT "EXT4-fs error (device %s): %s:%d: "
448 "inode #%lu: comm %s: %pV\n",
449 inode->i_sb->s_id, function, line, inode->i_ino,
450 current->comm, &vaf);
453 ext4_handle_error(inode->i_sb);
456 void ext4_error_file(struct file *file, const char *function,
457 unsigned int line, ext4_fsblk_t block,
458 const char *fmt, ...)
461 struct va_format vaf;
462 struct ext4_super_block *es;
463 struct inode *inode = file_inode(file);
464 char pathname[80], *path;
466 es = EXT4_SB(inode->i_sb)->s_es;
467 es->s_last_error_ino = cpu_to_le32(inode->i_ino);
468 save_error_info(inode->i_sb, function, line);
469 path = d_path(&(file->f_path), pathname, sizeof(pathname));
477 "EXT4-fs error (device %s): %s:%d: inode #%lu: "
478 "block %llu: comm %s: path %s: %pV\n",
479 inode->i_sb->s_id, function, line, inode->i_ino,
480 block, current->comm, path, &vaf);
483 "EXT4-fs error (device %s): %s:%d: inode #%lu: "
484 "comm %s: path %s: %pV\n",
485 inode->i_sb->s_id, function, line, inode->i_ino,
486 current->comm, path, &vaf);
489 ext4_handle_error(inode->i_sb);
492 const char *ext4_decode_error(struct super_block *sb, int errno,
499 errstr = "IO failure";
502 errstr = "Out of memory";
505 if (!sb || (EXT4_SB(sb)->s_journal &&
506 EXT4_SB(sb)->s_journal->j_flags & JBD2_ABORT))
507 errstr = "Journal has aborted";
509 errstr = "Readonly filesystem";
512 /* If the caller passed in an extra buffer for unknown
513 * errors, textualise them now. Else we just return
516 /* Check for truncated error codes... */
517 if (snprintf(nbuf, 16, "error %d", -errno) >= 0)
526 /* __ext4_std_error decodes expected errors from journaling functions
527 * automatically and invokes the appropriate error response. */
529 void __ext4_std_error(struct super_block *sb, const char *function,
530 unsigned int line, int errno)
535 /* Special case: if the error is EROFS, and we're not already
536 * inside a transaction, then there's really no point in logging
538 if (errno == -EROFS && journal_current_handle() == NULL &&
539 (sb->s_flags & MS_RDONLY))
542 errstr = ext4_decode_error(sb, errno, nbuf);
543 printk(KERN_CRIT "EXT4-fs error (device %s) in %s:%d: %s\n",
544 sb->s_id, function, line, errstr);
545 save_error_info(sb, function, line);
547 ext4_handle_error(sb);
551 * ext4_abort is a much stronger failure handler than ext4_error. The
552 * abort function may be used to deal with unrecoverable failures such
553 * as journal IO errors or ENOMEM at a critical moment in log management.
555 * We unconditionally force the filesystem into an ABORT|READONLY state,
556 * unless the error response on the fs has been set to panic in which
557 * case we take the easy way out and panic immediately.
560 void __ext4_abort(struct super_block *sb, const char *function,
561 unsigned int line, const char *fmt, ...)
565 save_error_info(sb, function, line);
567 printk(KERN_CRIT "EXT4-fs error (device %s): %s:%d: ", sb->s_id,
573 if ((sb->s_flags & MS_RDONLY) == 0) {
574 ext4_msg(sb, KERN_CRIT, "Remounting filesystem read-only");
575 sb->s_flags |= MS_RDONLY;
576 EXT4_SB(sb)->s_mount_flags |= EXT4_MF_FS_ABORTED;
577 if (EXT4_SB(sb)->s_journal)
578 jbd2_journal_abort(EXT4_SB(sb)->s_journal, -EIO);
579 save_error_info(sb, function, line);
581 if (test_opt(sb, ERRORS_PANIC))
582 panic("EXT4-fs panic from previous error\n");
585 void ext4_msg(struct super_block *sb, const char *prefix, const char *fmt, ...)
587 struct va_format vaf;
593 printk("%sEXT4-fs (%s): %pV\n", prefix, sb->s_id, &vaf);
597 void __ext4_warning(struct super_block *sb, const char *function,
598 unsigned int line, const char *fmt, ...)
600 struct va_format vaf;
606 printk(KERN_WARNING "EXT4-fs warning (device %s): %s:%d: %pV\n",
607 sb->s_id, function, line, &vaf);
611 void __ext4_grp_locked_error(const char *function, unsigned int line,
612 struct super_block *sb, ext4_group_t grp,
613 unsigned long ino, ext4_fsblk_t block,
614 const char *fmt, ...)
618 struct va_format vaf;
620 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
622 es->s_last_error_ino = cpu_to_le32(ino);
623 es->s_last_error_block = cpu_to_le64(block);
624 __save_error_info(sb, function, line);
630 printk(KERN_CRIT "EXT4-fs error (device %s): %s:%d: group %u, ",
631 sb->s_id, function, line, grp);
633 printk(KERN_CONT "inode %lu: ", ino);
635 printk(KERN_CONT "block %llu:", (unsigned long long) block);
636 printk(KERN_CONT "%pV\n", &vaf);
639 if (test_opt(sb, ERRORS_CONT)) {
640 ext4_commit_super(sb, 0);
644 ext4_unlock_group(sb, grp);
645 ext4_handle_error(sb);
647 * We only get here in the ERRORS_RO case; relocking the group
648 * may be dangerous, but nothing bad will happen since the
649 * filesystem will have already been marked read/only and the
650 * journal has been aborted. We return 1 as a hint to callers
651 * who might what to use the return value from
652 * ext4_grp_locked_error() to distinguish between the
653 * ERRORS_CONT and ERRORS_RO case, and perhaps return more
654 * aggressively from the ext4 function in question, with a
655 * more appropriate error code.
657 ext4_lock_group(sb, grp);
661 void ext4_update_dynamic_rev(struct super_block *sb)
663 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
665 if (le32_to_cpu(es->s_rev_level) > EXT4_GOOD_OLD_REV)
669 "updating to rev %d because of new feature flag, "
670 "running e2fsck is recommended",
673 es->s_first_ino = cpu_to_le32(EXT4_GOOD_OLD_FIRST_INO);
674 es->s_inode_size = cpu_to_le16(EXT4_GOOD_OLD_INODE_SIZE);
675 es->s_rev_level = cpu_to_le32(EXT4_DYNAMIC_REV);
676 /* leave es->s_feature_*compat flags alone */
677 /* es->s_uuid will be set by e2fsck if empty */
680 * The rest of the superblock fields should be zero, and if not it
681 * means they are likely already in use, so leave them alone. We
682 * can leave it up to e2fsck to clean up any inconsistencies there.
687 * Open the external journal device
689 static struct block_device *ext4_blkdev_get(dev_t dev, struct super_block *sb)
691 struct block_device *bdev;
692 char b[BDEVNAME_SIZE];
694 bdev = blkdev_get_by_dev(dev, FMODE_READ|FMODE_WRITE|FMODE_EXCL, sb);
700 ext4_msg(sb, KERN_ERR, "failed to open journal device %s: %ld",
701 __bdevname(dev, b), PTR_ERR(bdev));
706 * Release the journal device
708 static void ext4_blkdev_put(struct block_device *bdev)
710 blkdev_put(bdev, FMODE_READ|FMODE_WRITE|FMODE_EXCL);
713 static void ext4_blkdev_remove(struct ext4_sb_info *sbi)
715 struct block_device *bdev;
716 bdev = sbi->journal_bdev;
718 ext4_blkdev_put(bdev);
719 sbi->journal_bdev = NULL;
723 static inline struct inode *orphan_list_entry(struct list_head *l)
725 return &list_entry(l, struct ext4_inode_info, i_orphan)->vfs_inode;
728 static void dump_orphan_list(struct super_block *sb, struct ext4_sb_info *sbi)
732 ext4_msg(sb, KERN_ERR, "sb orphan head is %d",
733 le32_to_cpu(sbi->s_es->s_last_orphan));
735 printk(KERN_ERR "sb_info orphan list:\n");
736 list_for_each(l, &sbi->s_orphan) {
737 struct inode *inode = orphan_list_entry(l);
739 "inode %s:%lu at %p: mode %o, nlink %d, next %d\n",
740 inode->i_sb->s_id, inode->i_ino, inode,
741 inode->i_mode, inode->i_nlink,
746 static void ext4_put_super(struct super_block *sb)
748 struct ext4_sb_info *sbi = EXT4_SB(sb);
749 struct ext4_super_block *es = sbi->s_es;
752 ext4_unregister_li_request(sb);
753 dquot_disable(sb, -1, DQUOT_USAGE_ENABLED | DQUOT_LIMITS_ENABLED);
755 flush_workqueue(sbi->dio_unwritten_wq);
756 destroy_workqueue(sbi->dio_unwritten_wq);
758 if (sbi->s_journal) {
759 err = jbd2_journal_destroy(sbi->s_journal);
760 sbi->s_journal = NULL;
762 ext4_abort(sb, "Couldn't clean up the journal");
765 ext4_es_unregister_shrinker(sb);
766 del_timer(&sbi->s_err_report);
767 ext4_release_system_zone(sb);
769 ext4_ext_release(sb);
770 ext4_xattr_put_super(sb);
772 if (!(sb->s_flags & MS_RDONLY)) {
773 EXT4_CLEAR_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
774 es->s_state = cpu_to_le16(sbi->s_mount_state);
776 if (!(sb->s_flags & MS_RDONLY))
777 ext4_commit_super(sb, 1);
780 remove_proc_entry("options", sbi->s_proc);
781 remove_proc_entry(sb->s_id, ext4_proc_root);
783 kobject_del(&sbi->s_kobj);
785 for (i = 0; i < sbi->s_gdb_count; i++)
786 brelse(sbi->s_group_desc[i]);
787 ext4_kvfree(sbi->s_group_desc);
788 ext4_kvfree(sbi->s_flex_groups);
789 percpu_counter_destroy(&sbi->s_freeclusters_counter);
790 percpu_counter_destroy(&sbi->s_freeinodes_counter);
791 percpu_counter_destroy(&sbi->s_dirs_counter);
792 percpu_counter_destroy(&sbi->s_dirtyclusters_counter);
793 percpu_counter_destroy(&sbi->s_extent_cache_cnt);
796 for (i = 0; i < MAXQUOTAS; i++)
797 kfree(sbi->s_qf_names[i]);
800 /* Debugging code just in case the in-memory inode orphan list
801 * isn't empty. The on-disk one can be non-empty if we've
802 * detected an error and taken the fs readonly, but the
803 * in-memory list had better be clean by this point. */
804 if (!list_empty(&sbi->s_orphan))
805 dump_orphan_list(sb, sbi);
806 J_ASSERT(list_empty(&sbi->s_orphan));
808 sync_blockdev(sb->s_bdev);
809 invalidate_bdev(sb->s_bdev);
810 if (sbi->journal_bdev && sbi->journal_bdev != sb->s_bdev) {
812 * Invalidate the journal device's buffers. We don't want them
813 * floating about in memory - the physical journal device may
814 * hotswapped, and it breaks the `ro-after' testing code.
816 sync_blockdev(sbi->journal_bdev);
817 invalidate_bdev(sbi->journal_bdev);
818 ext4_blkdev_remove(sbi);
821 kthread_stop(sbi->s_mmp_tsk);
822 sb->s_fs_info = NULL;
824 * Now that we are completely done shutting down the
825 * superblock, we need to actually destroy the kobject.
827 kobject_put(&sbi->s_kobj);
828 wait_for_completion(&sbi->s_kobj_unregister);
829 if (sbi->s_chksum_driver)
830 crypto_free_shash(sbi->s_chksum_driver);
831 kfree(sbi->s_blockgroup_lock);
835 static struct kmem_cache *ext4_inode_cachep;
838 * Called inside transaction, so use GFP_NOFS
840 static struct inode *ext4_alloc_inode(struct super_block *sb)
842 struct ext4_inode_info *ei;
844 ei = kmem_cache_alloc(ext4_inode_cachep, GFP_NOFS);
848 ei->vfs_inode.i_version = 1;
849 INIT_LIST_HEAD(&ei->i_prealloc_list);
850 spin_lock_init(&ei->i_prealloc_lock);
851 ext4_es_init_tree(&ei->i_es_tree);
852 rwlock_init(&ei->i_es_lock);
853 INIT_LIST_HEAD(&ei->i_es_lru);
855 ei->i_reserved_data_blocks = 0;
856 ei->i_reserved_meta_blocks = 0;
857 ei->i_allocated_meta_blocks = 0;
858 ei->i_da_metadata_calc_len = 0;
859 ei->i_da_metadata_calc_last_lblock = 0;
860 spin_lock_init(&(ei->i_block_reservation_lock));
862 ei->i_reserved_quota = 0;
865 INIT_LIST_HEAD(&ei->i_completed_io_list);
866 spin_lock_init(&ei->i_completed_io_lock);
868 ei->i_datasync_tid = 0;
869 atomic_set(&ei->i_ioend_count, 0);
870 atomic_set(&ei->i_unwritten, 0);
871 INIT_WORK(&ei->i_unwritten_work, ext4_end_io_work);
873 return &ei->vfs_inode;
876 static int ext4_drop_inode(struct inode *inode)
878 int drop = generic_drop_inode(inode);
880 trace_ext4_drop_inode(inode, drop);
884 static void ext4_i_callback(struct rcu_head *head)
886 struct inode *inode = container_of(head, struct inode, i_rcu);
887 kmem_cache_free(ext4_inode_cachep, EXT4_I(inode));
890 static void ext4_destroy_inode(struct inode *inode)
892 if (!list_empty(&(EXT4_I(inode)->i_orphan))) {
893 ext4_msg(inode->i_sb, KERN_ERR,
894 "Inode %lu (%p): orphan list check failed!",
895 inode->i_ino, EXT4_I(inode));
896 print_hex_dump(KERN_INFO, "", DUMP_PREFIX_ADDRESS, 16, 4,
897 EXT4_I(inode), sizeof(struct ext4_inode_info),
901 call_rcu(&inode->i_rcu, ext4_i_callback);
904 static void init_once(void *foo)
906 struct ext4_inode_info *ei = (struct ext4_inode_info *) foo;
908 INIT_LIST_HEAD(&ei->i_orphan);
909 init_rwsem(&ei->xattr_sem);
910 init_rwsem(&ei->i_data_sem);
911 inode_init_once(&ei->vfs_inode);
914 static int init_inodecache(void)
916 ext4_inode_cachep = kmem_cache_create("ext4_inode_cache",
917 sizeof(struct ext4_inode_info),
918 0, (SLAB_RECLAIM_ACCOUNT|
921 if (ext4_inode_cachep == NULL)
926 static void destroy_inodecache(void)
929 * Make sure all delayed rcu free inodes are flushed before we
933 kmem_cache_destroy(ext4_inode_cachep);
936 void ext4_clear_inode(struct inode *inode)
938 invalidate_inode_buffers(inode);
941 ext4_discard_preallocations(inode);
942 ext4_es_remove_extent(inode, 0, EXT_MAX_BLOCKS);
943 ext4_es_lru_del(inode);
944 if (EXT4_I(inode)->jinode) {
945 jbd2_journal_release_jbd_inode(EXT4_JOURNAL(inode),
946 EXT4_I(inode)->jinode);
947 jbd2_free_inode(EXT4_I(inode)->jinode);
948 EXT4_I(inode)->jinode = NULL;
952 static struct inode *ext4_nfs_get_inode(struct super_block *sb,
953 u64 ino, u32 generation)
957 if (ino < EXT4_FIRST_INO(sb) && ino != EXT4_ROOT_INO)
958 return ERR_PTR(-ESTALE);
959 if (ino > le32_to_cpu(EXT4_SB(sb)->s_es->s_inodes_count))
960 return ERR_PTR(-ESTALE);
962 /* iget isn't really right if the inode is currently unallocated!!
964 * ext4_read_inode will return a bad_inode if the inode had been
965 * deleted, so we should be safe.
967 * Currently we don't know the generation for parent directory, so
968 * a generation of 0 means "accept any"
970 inode = ext4_iget_normal(sb, ino);
972 return ERR_CAST(inode);
973 if (generation && inode->i_generation != generation) {
975 return ERR_PTR(-ESTALE);
981 static struct dentry *ext4_fh_to_dentry(struct super_block *sb, struct fid *fid,
982 int fh_len, int fh_type)
984 return generic_fh_to_dentry(sb, fid, fh_len, fh_type,
988 static struct dentry *ext4_fh_to_parent(struct super_block *sb, struct fid *fid,
989 int fh_len, int fh_type)
991 return generic_fh_to_parent(sb, fid, fh_len, fh_type,
996 * Try to release metadata pages (indirect blocks, directories) which are
997 * mapped via the block device. Since these pages could have journal heads
998 * which would prevent try_to_free_buffers() from freeing them, we must use
999 * jbd2 layer's try_to_free_buffers() function to release them.
1001 static int bdev_try_to_free_page(struct super_block *sb, struct page *page,
1004 journal_t *journal = EXT4_SB(sb)->s_journal;
1006 WARN_ON(PageChecked(page));
1007 if (!page_has_buffers(page))
1010 return jbd2_journal_try_to_free_buffers(journal, page,
1011 wait & ~__GFP_WAIT);
1012 return try_to_free_buffers(page);
1016 #define QTYPE2NAME(t) ((t) == USRQUOTA ? "user" : "group")
1017 #define QTYPE2MOPT(on, t) ((t) == USRQUOTA?((on)##USRJQUOTA):((on)##GRPJQUOTA))
1019 static int ext4_write_dquot(struct dquot *dquot);
1020 static int ext4_acquire_dquot(struct dquot *dquot);
1021 static int ext4_release_dquot(struct dquot *dquot);
1022 static int ext4_mark_dquot_dirty(struct dquot *dquot);
1023 static int ext4_write_info(struct super_block *sb, int type);
1024 static int ext4_quota_on(struct super_block *sb, int type, int format_id,
1026 static int ext4_quota_on_sysfile(struct super_block *sb, int type,
1028 static int ext4_quota_off(struct super_block *sb, int type);
1029 static int ext4_quota_off_sysfile(struct super_block *sb, int type);
1030 static int ext4_quota_on_mount(struct super_block *sb, int type);
1031 static ssize_t ext4_quota_read(struct super_block *sb, int type, char *data,
1032 size_t len, loff_t off);
1033 static ssize_t ext4_quota_write(struct super_block *sb, int type,
1034 const char *data, size_t len, loff_t off);
1035 static int ext4_quota_enable(struct super_block *sb, int type, int format_id,
1036 unsigned int flags);
1037 static int ext4_enable_quotas(struct super_block *sb);
1039 static const struct dquot_operations ext4_quota_operations = {
1040 .get_reserved_space = ext4_get_reserved_space,
1041 .write_dquot = ext4_write_dquot,
1042 .acquire_dquot = ext4_acquire_dquot,
1043 .release_dquot = ext4_release_dquot,
1044 .mark_dirty = ext4_mark_dquot_dirty,
1045 .write_info = ext4_write_info,
1046 .alloc_dquot = dquot_alloc,
1047 .destroy_dquot = dquot_destroy,
1050 static const struct quotactl_ops ext4_qctl_operations = {
1051 .quota_on = ext4_quota_on,
1052 .quota_off = ext4_quota_off,
1053 .quota_sync = dquot_quota_sync,
1054 .get_info = dquot_get_dqinfo,
1055 .set_info = dquot_set_dqinfo,
1056 .get_dqblk = dquot_get_dqblk,
1057 .set_dqblk = dquot_set_dqblk
1060 static const struct quotactl_ops ext4_qctl_sysfile_operations = {
1061 .quota_on_meta = ext4_quota_on_sysfile,
1062 .quota_off = ext4_quota_off_sysfile,
1063 .quota_sync = dquot_quota_sync,
1064 .get_info = dquot_get_dqinfo,
1065 .set_info = dquot_set_dqinfo,
1066 .get_dqblk = dquot_get_dqblk,
1067 .set_dqblk = dquot_set_dqblk
1071 static const struct super_operations ext4_sops = {
1072 .alloc_inode = ext4_alloc_inode,
1073 .destroy_inode = ext4_destroy_inode,
1074 .write_inode = ext4_write_inode,
1075 .dirty_inode = ext4_dirty_inode,
1076 .drop_inode = ext4_drop_inode,
1077 .evict_inode = ext4_evict_inode,
1078 .put_super = ext4_put_super,
1079 .sync_fs = ext4_sync_fs,
1080 .freeze_fs = ext4_freeze,
1081 .unfreeze_fs = ext4_unfreeze,
1082 .statfs = ext4_statfs,
1083 .remount_fs = ext4_remount,
1084 .show_options = ext4_show_options,
1086 .quota_read = ext4_quota_read,
1087 .quota_write = ext4_quota_write,
1089 .bdev_try_to_free_page = bdev_try_to_free_page,
1092 static const struct super_operations ext4_nojournal_sops = {
1093 .alloc_inode = ext4_alloc_inode,
1094 .destroy_inode = ext4_destroy_inode,
1095 .write_inode = ext4_write_inode,
1096 .dirty_inode = ext4_dirty_inode,
1097 .drop_inode = ext4_drop_inode,
1098 .evict_inode = ext4_evict_inode,
1099 .put_super = ext4_put_super,
1100 .statfs = ext4_statfs,
1101 .remount_fs = ext4_remount,
1102 .show_options = ext4_show_options,
1104 .quota_read = ext4_quota_read,
1105 .quota_write = ext4_quota_write,
1107 .bdev_try_to_free_page = bdev_try_to_free_page,
1110 static const struct export_operations ext4_export_ops = {
1111 .fh_to_dentry = ext4_fh_to_dentry,
1112 .fh_to_parent = ext4_fh_to_parent,
1113 .get_parent = ext4_get_parent,
1117 Opt_bsd_df, Opt_minix_df, Opt_grpid, Opt_nogrpid,
1118 Opt_resgid, Opt_resuid, Opt_sb, Opt_err_cont, Opt_err_panic, Opt_err_ro,
1119 Opt_nouid32, Opt_debug, Opt_removed,
1120 Opt_user_xattr, Opt_nouser_xattr, Opt_acl, Opt_noacl,
1121 Opt_auto_da_alloc, Opt_noauto_da_alloc, Opt_noload,
1122 Opt_commit, Opt_min_batch_time, Opt_max_batch_time,
1123 Opt_journal_dev, Opt_journal_checksum, Opt_journal_async_commit,
1124 Opt_abort, Opt_data_journal, Opt_data_ordered, Opt_data_writeback,
1125 Opt_data_err_abort, Opt_data_err_ignore,
1126 Opt_usrjquota, Opt_grpjquota, Opt_offusrjquota, Opt_offgrpjquota,
1127 Opt_jqfmt_vfsold, Opt_jqfmt_vfsv0, Opt_jqfmt_vfsv1, Opt_quota,
1128 Opt_noquota, Opt_barrier, Opt_nobarrier, Opt_err,
1129 Opt_usrquota, Opt_grpquota, Opt_i_version,
1130 Opt_stripe, Opt_delalloc, Opt_nodelalloc, Opt_mblk_io_submit,
1131 Opt_nomblk_io_submit, Opt_block_validity, Opt_noblock_validity,
1132 Opt_inode_readahead_blks, Opt_journal_ioprio,
1133 Opt_dioread_nolock, Opt_dioread_lock,
1134 Opt_discard, Opt_nodiscard, Opt_init_itable, Opt_noinit_itable,
1135 Opt_max_dir_size_kb,
1138 static const match_table_t tokens = {
1139 {Opt_bsd_df, "bsddf"},
1140 {Opt_minix_df, "minixdf"},
1141 {Opt_grpid, "grpid"},
1142 {Opt_grpid, "bsdgroups"},
1143 {Opt_nogrpid, "nogrpid"},
1144 {Opt_nogrpid, "sysvgroups"},
1145 {Opt_resgid, "resgid=%u"},
1146 {Opt_resuid, "resuid=%u"},
1148 {Opt_err_cont, "errors=continue"},
1149 {Opt_err_panic, "errors=panic"},
1150 {Opt_err_ro, "errors=remount-ro"},
1151 {Opt_nouid32, "nouid32"},
1152 {Opt_debug, "debug"},
1153 {Opt_removed, "oldalloc"},
1154 {Opt_removed, "orlov"},
1155 {Opt_user_xattr, "user_xattr"},
1156 {Opt_nouser_xattr, "nouser_xattr"},
1158 {Opt_noacl, "noacl"},
1159 {Opt_noload, "norecovery"},
1160 {Opt_noload, "noload"},
1161 {Opt_removed, "nobh"},
1162 {Opt_removed, "bh"},
1163 {Opt_commit, "commit=%u"},
1164 {Opt_min_batch_time, "min_batch_time=%u"},
1165 {Opt_max_batch_time, "max_batch_time=%u"},
1166 {Opt_journal_dev, "journal_dev=%u"},
1167 {Opt_journal_checksum, "journal_checksum"},
1168 {Opt_journal_async_commit, "journal_async_commit"},
1169 {Opt_abort, "abort"},
1170 {Opt_data_journal, "data=journal"},
1171 {Opt_data_ordered, "data=ordered"},
1172 {Opt_data_writeback, "data=writeback"},
1173 {Opt_data_err_abort, "data_err=abort"},
1174 {Opt_data_err_ignore, "data_err=ignore"},
1175 {Opt_offusrjquota, "usrjquota="},
1176 {Opt_usrjquota, "usrjquota=%s"},
1177 {Opt_offgrpjquota, "grpjquota="},
1178 {Opt_grpjquota, "grpjquota=%s"},
1179 {Opt_jqfmt_vfsold, "jqfmt=vfsold"},
1180 {Opt_jqfmt_vfsv0, "jqfmt=vfsv0"},
1181 {Opt_jqfmt_vfsv1, "jqfmt=vfsv1"},
1182 {Opt_grpquota, "grpquota"},
1183 {Opt_noquota, "noquota"},
1184 {Opt_quota, "quota"},
1185 {Opt_usrquota, "usrquota"},
1186 {Opt_barrier, "barrier=%u"},
1187 {Opt_barrier, "barrier"},
1188 {Opt_nobarrier, "nobarrier"},
1189 {Opt_i_version, "i_version"},
1190 {Opt_stripe, "stripe=%u"},
1191 {Opt_delalloc, "delalloc"},
1192 {Opt_nodelalloc, "nodelalloc"},
1193 {Opt_removed, "mblk_io_submit"},
1194 {Opt_removed, "nomblk_io_submit"},
1195 {Opt_block_validity, "block_validity"},
1196 {Opt_noblock_validity, "noblock_validity"},
1197 {Opt_inode_readahead_blks, "inode_readahead_blks=%u"},
1198 {Opt_journal_ioprio, "journal_ioprio=%u"},
1199 {Opt_auto_da_alloc, "auto_da_alloc=%u"},
1200 {Opt_auto_da_alloc, "auto_da_alloc"},
1201 {Opt_noauto_da_alloc, "noauto_da_alloc"},
1202 {Opt_dioread_nolock, "dioread_nolock"},
1203 {Opt_dioread_lock, "dioread_lock"},
1204 {Opt_discard, "discard"},
1205 {Opt_nodiscard, "nodiscard"},
1206 {Opt_init_itable, "init_itable=%u"},
1207 {Opt_init_itable, "init_itable"},
1208 {Opt_noinit_itable, "noinit_itable"},
1209 {Opt_max_dir_size_kb, "max_dir_size_kb=%u"},
1210 {Opt_removed, "check=none"}, /* mount option from ext2/3 */
1211 {Opt_removed, "nocheck"}, /* mount option from ext2/3 */
1212 {Opt_removed, "reservation"}, /* mount option from ext2/3 */
1213 {Opt_removed, "noreservation"}, /* mount option from ext2/3 */
1214 {Opt_removed, "journal=%u"}, /* mount option from ext2/3 */
1218 static ext4_fsblk_t get_sb_block(void **data)
1220 ext4_fsblk_t sb_block;
1221 char *options = (char *) *data;
1223 if (!options || strncmp(options, "sb=", 3) != 0)
1224 return 1; /* Default location */
1227 /* TODO: use simple_strtoll with >32bit ext4 */
1228 sb_block = simple_strtoul(options, &options, 0);
1229 if (*options && *options != ',') {
1230 printk(KERN_ERR "EXT4-fs: Invalid sb specification: %s\n",
1234 if (*options == ',')
1236 *data = (void *) options;
1241 #define DEFAULT_JOURNAL_IOPRIO (IOPRIO_PRIO_VALUE(IOPRIO_CLASS_BE, 3))
1242 static char deprecated_msg[] = "Mount option \"%s\" will be removed by %s\n"
1243 "Contact linux-ext4@vger.kernel.org if you think we should keep it.\n";
1246 static int set_qf_name(struct super_block *sb, int qtype, substring_t *args)
1248 struct ext4_sb_info *sbi = EXT4_SB(sb);
1252 if (sb_any_quota_loaded(sb) &&
1253 !sbi->s_qf_names[qtype]) {
1254 ext4_msg(sb, KERN_ERR,
1255 "Cannot change journaled "
1256 "quota options when quota turned on");
1259 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_QUOTA)) {
1260 ext4_msg(sb, KERN_ERR, "Cannot set journaled quota options "
1261 "when QUOTA feature is enabled");
1264 qname = match_strdup(args);
1266 ext4_msg(sb, KERN_ERR,
1267 "Not enough memory for storing quotafile name");
1270 if (sbi->s_qf_names[qtype]) {
1271 if (strcmp(sbi->s_qf_names[qtype], qname) == 0)
1274 ext4_msg(sb, KERN_ERR,
1275 "%s quota file already specified",
1279 if (strchr(qname, '/')) {
1280 ext4_msg(sb, KERN_ERR,
1281 "quotafile must be on filesystem root");
1284 sbi->s_qf_names[qtype] = qname;
1292 static int clear_qf_name(struct super_block *sb, int qtype)
1295 struct ext4_sb_info *sbi = EXT4_SB(sb);
1297 if (sb_any_quota_loaded(sb) &&
1298 sbi->s_qf_names[qtype]) {
1299 ext4_msg(sb, KERN_ERR, "Cannot change journaled quota options"
1300 " when quota turned on");
1303 kfree(sbi->s_qf_names[qtype]);
1304 sbi->s_qf_names[qtype] = NULL;
1309 #define MOPT_SET 0x0001
1310 #define MOPT_CLEAR 0x0002
1311 #define MOPT_NOSUPPORT 0x0004
1312 #define MOPT_EXPLICIT 0x0008
1313 #define MOPT_CLEAR_ERR 0x0010
1314 #define MOPT_GTE0 0x0020
1317 #define MOPT_QFMT 0x0040
1319 #define MOPT_Q MOPT_NOSUPPORT
1320 #define MOPT_QFMT MOPT_NOSUPPORT
1322 #define MOPT_DATAJ 0x0080
1323 #define MOPT_NO_EXT2 0x0100
1324 #define MOPT_NO_EXT3 0x0200
1325 #define MOPT_EXT4_ONLY (MOPT_NO_EXT2 | MOPT_NO_EXT3)
1327 static const struct mount_opts {
1331 } ext4_mount_opts[] = {
1332 {Opt_minix_df, EXT4_MOUNT_MINIX_DF, MOPT_SET},
1333 {Opt_bsd_df, EXT4_MOUNT_MINIX_DF, MOPT_CLEAR},
1334 {Opt_grpid, EXT4_MOUNT_GRPID, MOPT_SET},
1335 {Opt_nogrpid, EXT4_MOUNT_GRPID, MOPT_CLEAR},
1336 {Opt_block_validity, EXT4_MOUNT_BLOCK_VALIDITY, MOPT_SET},
1337 {Opt_noblock_validity, EXT4_MOUNT_BLOCK_VALIDITY, MOPT_CLEAR},
1338 {Opt_dioread_nolock, EXT4_MOUNT_DIOREAD_NOLOCK,
1339 MOPT_EXT4_ONLY | MOPT_SET},
1340 {Opt_dioread_lock, EXT4_MOUNT_DIOREAD_NOLOCK,
1341 MOPT_EXT4_ONLY | MOPT_CLEAR},
1342 {Opt_discard, EXT4_MOUNT_DISCARD, MOPT_SET},
1343 {Opt_nodiscard, EXT4_MOUNT_DISCARD, MOPT_CLEAR},
1344 {Opt_delalloc, EXT4_MOUNT_DELALLOC,
1345 MOPT_EXT4_ONLY | MOPT_SET | MOPT_EXPLICIT},
1346 {Opt_nodelalloc, EXT4_MOUNT_DELALLOC,
1347 MOPT_EXT4_ONLY | MOPT_CLEAR},
1348 {Opt_journal_checksum, EXT4_MOUNT_JOURNAL_CHECKSUM,
1349 MOPT_EXT4_ONLY | MOPT_SET},
1350 {Opt_journal_async_commit, (EXT4_MOUNT_JOURNAL_ASYNC_COMMIT |
1351 EXT4_MOUNT_JOURNAL_CHECKSUM),
1352 MOPT_EXT4_ONLY | MOPT_SET},
1353 {Opt_noload, EXT4_MOUNT_NOLOAD, MOPT_NO_EXT2 | MOPT_SET},
1354 {Opt_err_panic, EXT4_MOUNT_ERRORS_PANIC, MOPT_SET | MOPT_CLEAR_ERR},
1355 {Opt_err_ro, EXT4_MOUNT_ERRORS_RO, MOPT_SET | MOPT_CLEAR_ERR},
1356 {Opt_err_cont, EXT4_MOUNT_ERRORS_CONT, MOPT_SET | MOPT_CLEAR_ERR},
1357 {Opt_data_err_abort, EXT4_MOUNT_DATA_ERR_ABORT,
1358 MOPT_NO_EXT2 | MOPT_SET},
1359 {Opt_data_err_ignore, EXT4_MOUNT_DATA_ERR_ABORT,
1360 MOPT_NO_EXT2 | MOPT_CLEAR},
1361 {Opt_barrier, EXT4_MOUNT_BARRIER, MOPT_SET},
1362 {Opt_nobarrier, EXT4_MOUNT_BARRIER, MOPT_CLEAR},
1363 {Opt_noauto_da_alloc, EXT4_MOUNT_NO_AUTO_DA_ALLOC, MOPT_SET},
1364 {Opt_auto_da_alloc, EXT4_MOUNT_NO_AUTO_DA_ALLOC, MOPT_CLEAR},
1365 {Opt_noinit_itable, EXT4_MOUNT_INIT_INODE_TABLE, MOPT_CLEAR},
1366 {Opt_commit, 0, MOPT_GTE0},
1367 {Opt_max_batch_time, 0, MOPT_GTE0},
1368 {Opt_min_batch_time, 0, MOPT_GTE0},
1369 {Opt_inode_readahead_blks, 0, MOPT_GTE0},
1370 {Opt_init_itable, 0, MOPT_GTE0},
1371 {Opt_stripe, 0, MOPT_GTE0},
1372 {Opt_resuid, 0, MOPT_GTE0},
1373 {Opt_resgid, 0, MOPT_GTE0},
1374 {Opt_journal_dev, 0, MOPT_GTE0},
1375 {Opt_journal_ioprio, 0, MOPT_GTE0},
1376 {Opt_data_journal, EXT4_MOUNT_JOURNAL_DATA, MOPT_NO_EXT2 | MOPT_DATAJ},
1377 {Opt_data_ordered, EXT4_MOUNT_ORDERED_DATA, MOPT_NO_EXT2 | MOPT_DATAJ},
1378 {Opt_data_writeback, EXT4_MOUNT_WRITEBACK_DATA,
1379 MOPT_NO_EXT2 | MOPT_DATAJ},
1380 {Opt_user_xattr, EXT4_MOUNT_XATTR_USER, MOPT_SET},
1381 {Opt_nouser_xattr, EXT4_MOUNT_XATTR_USER, MOPT_CLEAR},
1382 #ifdef CONFIG_EXT4_FS_POSIX_ACL
1383 {Opt_acl, EXT4_MOUNT_POSIX_ACL, MOPT_SET},
1384 {Opt_noacl, EXT4_MOUNT_POSIX_ACL, MOPT_CLEAR},
1386 {Opt_acl, 0, MOPT_NOSUPPORT},
1387 {Opt_noacl, 0, MOPT_NOSUPPORT},
1389 {Opt_nouid32, EXT4_MOUNT_NO_UID32, MOPT_SET},
1390 {Opt_debug, EXT4_MOUNT_DEBUG, MOPT_SET},
1391 {Opt_quota, EXT4_MOUNT_QUOTA | EXT4_MOUNT_USRQUOTA, MOPT_SET | MOPT_Q},
1392 {Opt_usrquota, EXT4_MOUNT_QUOTA | EXT4_MOUNT_USRQUOTA,
1394 {Opt_grpquota, EXT4_MOUNT_QUOTA | EXT4_MOUNT_GRPQUOTA,
1396 {Opt_noquota, (EXT4_MOUNT_QUOTA | EXT4_MOUNT_USRQUOTA |
1397 EXT4_MOUNT_GRPQUOTA), MOPT_CLEAR | MOPT_Q},
1398 {Opt_usrjquota, 0, MOPT_Q},
1399 {Opt_grpjquota, 0, MOPT_Q},
1400 {Opt_offusrjquota, 0, MOPT_Q},
1401 {Opt_offgrpjquota, 0, MOPT_Q},
1402 {Opt_jqfmt_vfsold, QFMT_VFS_OLD, MOPT_QFMT},
1403 {Opt_jqfmt_vfsv0, QFMT_VFS_V0, MOPT_QFMT},
1404 {Opt_jqfmt_vfsv1, QFMT_VFS_V1, MOPT_QFMT},
1405 {Opt_max_dir_size_kb, 0, MOPT_GTE0},
1409 static int handle_mount_opt(struct super_block *sb, char *opt, int token,
1410 substring_t *args, unsigned long *journal_devnum,
1411 unsigned int *journal_ioprio, int is_remount)
1413 struct ext4_sb_info *sbi = EXT4_SB(sb);
1414 const struct mount_opts *m;
1420 if (token == Opt_usrjquota)
1421 return set_qf_name(sb, USRQUOTA, &args[0]);
1422 else if (token == Opt_grpjquota)
1423 return set_qf_name(sb, GRPQUOTA, &args[0]);
1424 else if (token == Opt_offusrjquota)
1425 return clear_qf_name(sb, USRQUOTA);
1426 else if (token == Opt_offgrpjquota)
1427 return clear_qf_name(sb, GRPQUOTA);
1431 case Opt_nouser_xattr:
1432 ext4_msg(sb, KERN_WARNING, deprecated_msg, opt, "3.5");
1435 return 1; /* handled by get_sb_block() */
1437 ext4_msg(sb, KERN_WARNING, "Ignoring removed %s option", opt);
1440 sbi->s_mount_flags |= EXT4_MF_FS_ABORTED;
1443 sb->s_flags |= MS_I_VERSION;
1447 for (m = ext4_mount_opts; m->token != Opt_err; m++)
1448 if (token == m->token)
1451 if (m->token == Opt_err) {
1452 ext4_msg(sb, KERN_ERR, "Unrecognized mount option \"%s\" "
1453 "or missing value", opt);
1457 if ((m->flags & MOPT_NO_EXT2) && IS_EXT2_SB(sb)) {
1458 ext4_msg(sb, KERN_ERR,
1459 "Mount option \"%s\" incompatible with ext2", opt);
1462 if ((m->flags & MOPT_NO_EXT3) && IS_EXT3_SB(sb)) {
1463 ext4_msg(sb, KERN_ERR,
1464 "Mount option \"%s\" incompatible with ext3", opt);
1468 if (args->from && match_int(args, &arg))
1470 if (args->from && (m->flags & MOPT_GTE0) && (arg < 0))
1472 if (m->flags & MOPT_EXPLICIT)
1473 set_opt2(sb, EXPLICIT_DELALLOC);
1474 if (m->flags & MOPT_CLEAR_ERR)
1475 clear_opt(sb, ERRORS_MASK);
1476 if (token == Opt_noquota && sb_any_quota_loaded(sb)) {
1477 ext4_msg(sb, KERN_ERR, "Cannot change quota "
1478 "options when quota turned on");
1482 if (m->flags & MOPT_NOSUPPORT) {
1483 ext4_msg(sb, KERN_ERR, "%s option not supported", opt);
1484 } else if (token == Opt_commit) {
1486 arg = JBD2_DEFAULT_MAX_COMMIT_AGE;
1487 sbi->s_commit_interval = HZ * arg;
1488 } else if (token == Opt_max_batch_time) {
1489 sbi->s_max_batch_time = arg;
1490 } else if (token == Opt_min_batch_time) {
1491 sbi->s_min_batch_time = arg;
1492 } else if (token == Opt_inode_readahead_blks) {
1493 if (arg && (arg > (1 << 30) || !is_power_of_2(arg))) {
1494 ext4_msg(sb, KERN_ERR,
1495 "EXT4-fs: inode_readahead_blks must be "
1496 "0 or a power of 2 smaller than 2^31");
1499 sbi->s_inode_readahead_blks = arg;
1500 } else if (token == Opt_init_itable) {
1501 set_opt(sb, INIT_INODE_TABLE);
1503 arg = EXT4_DEF_LI_WAIT_MULT;
1504 sbi->s_li_wait_mult = arg;
1505 } else if (token == Opt_max_dir_size_kb) {
1506 sbi->s_max_dir_size_kb = arg;
1507 } else if (token == Opt_stripe) {
1508 sbi->s_stripe = arg;
1509 } else if (token == Opt_resuid) {
1510 uid = make_kuid(current_user_ns(), arg);
1511 if (!uid_valid(uid)) {
1512 ext4_msg(sb, KERN_ERR, "Invalid uid value %d", arg);
1515 sbi->s_resuid = uid;
1516 } else if (token == Opt_resgid) {
1517 gid = make_kgid(current_user_ns(), arg);
1518 if (!gid_valid(gid)) {
1519 ext4_msg(sb, KERN_ERR, "Invalid gid value %d", arg);
1522 sbi->s_resgid = gid;
1523 } else if (token == Opt_journal_dev) {
1525 ext4_msg(sb, KERN_ERR,
1526 "Cannot specify journal on remount");
1529 *journal_devnum = arg;
1530 } else if (token == Opt_journal_ioprio) {
1532 ext4_msg(sb, KERN_ERR, "Invalid journal IO priority"
1537 IOPRIO_PRIO_VALUE(IOPRIO_CLASS_BE, arg);
1538 } else if (m->flags & MOPT_DATAJ) {
1540 if (!sbi->s_journal)
1541 ext4_msg(sb, KERN_WARNING, "Remounting file system with no journal so ignoring journalled data option");
1542 else if (test_opt(sb, DATA_FLAGS) != m->mount_opt) {
1543 ext4_msg(sb, KERN_ERR,
1544 "Cannot change data mode on remount");
1548 clear_opt(sb, DATA_FLAGS);
1549 sbi->s_mount_opt |= m->mount_opt;
1552 } else if (m->flags & MOPT_QFMT) {
1553 if (sb_any_quota_loaded(sb) &&
1554 sbi->s_jquota_fmt != m->mount_opt) {
1555 ext4_msg(sb, KERN_ERR, "Cannot change journaled "
1556 "quota options when quota turned on");
1559 if (EXT4_HAS_RO_COMPAT_FEATURE(sb,
1560 EXT4_FEATURE_RO_COMPAT_QUOTA)) {
1561 ext4_msg(sb, KERN_ERR,
1562 "Cannot set journaled quota options "
1563 "when QUOTA feature is enabled");
1566 sbi->s_jquota_fmt = m->mount_opt;
1571 if (m->flags & MOPT_CLEAR)
1573 else if (unlikely(!(m->flags & MOPT_SET))) {
1574 ext4_msg(sb, KERN_WARNING,
1575 "buggy handling of option %s", opt);
1580 sbi->s_mount_opt |= m->mount_opt;
1582 sbi->s_mount_opt &= ~m->mount_opt;
1587 static int parse_options(char *options, struct super_block *sb,
1588 unsigned long *journal_devnum,
1589 unsigned int *journal_ioprio,
1592 struct ext4_sb_info *sbi = EXT4_SB(sb);
1594 substring_t args[MAX_OPT_ARGS];
1600 while ((p = strsep(&options, ",")) != NULL) {
1604 * Initialize args struct so we know whether arg was
1605 * found; some options take optional arguments.
1607 args[0].to = args[0].from = NULL;
1608 token = match_token(p, tokens, args);
1609 if (handle_mount_opt(sb, p, token, args, journal_devnum,
1610 journal_ioprio, is_remount) < 0)
1614 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_QUOTA) &&
1615 (test_opt(sb, USRQUOTA) || test_opt(sb, GRPQUOTA))) {
1616 ext4_msg(sb, KERN_ERR, "Cannot set quota options when QUOTA "
1617 "feature is enabled");
1620 if (sbi->s_qf_names[USRQUOTA] || sbi->s_qf_names[GRPQUOTA]) {
1621 if (test_opt(sb, USRQUOTA) && sbi->s_qf_names[USRQUOTA])
1622 clear_opt(sb, USRQUOTA);
1624 if (test_opt(sb, GRPQUOTA) && sbi->s_qf_names[GRPQUOTA])
1625 clear_opt(sb, GRPQUOTA);
1627 if (test_opt(sb, GRPQUOTA) || test_opt(sb, USRQUOTA)) {
1628 ext4_msg(sb, KERN_ERR, "old and new quota "
1633 if (!sbi->s_jquota_fmt) {
1634 ext4_msg(sb, KERN_ERR, "journaled quota format "
1640 if (test_opt(sb, DIOREAD_NOLOCK)) {
1642 BLOCK_SIZE << le32_to_cpu(sbi->s_es->s_log_block_size);
1644 if (blocksize < PAGE_CACHE_SIZE) {
1645 ext4_msg(sb, KERN_ERR, "can't mount with "
1646 "dioread_nolock if block size != PAGE_SIZE");
1653 static inline void ext4_show_quota_options(struct seq_file *seq,
1654 struct super_block *sb)
1656 #if defined(CONFIG_QUOTA)
1657 struct ext4_sb_info *sbi = EXT4_SB(sb);
1659 if (sbi->s_jquota_fmt) {
1662 switch (sbi->s_jquota_fmt) {
1673 seq_printf(seq, ",jqfmt=%s", fmtname);
1676 if (sbi->s_qf_names[USRQUOTA])
1677 seq_printf(seq, ",usrjquota=%s", sbi->s_qf_names[USRQUOTA]);
1679 if (sbi->s_qf_names[GRPQUOTA])
1680 seq_printf(seq, ",grpjquota=%s", sbi->s_qf_names[GRPQUOTA]);
1684 static const char *token2str(int token)
1686 const struct match_token *t;
1688 for (t = tokens; t->token != Opt_err; t++)
1689 if (t->token == token && !strchr(t->pattern, '='))
1696 * - it's set to a non-default value OR
1697 * - if the per-sb default is different from the global default
1699 static int _ext4_show_options(struct seq_file *seq, struct super_block *sb,
1702 struct ext4_sb_info *sbi = EXT4_SB(sb);
1703 struct ext4_super_block *es = sbi->s_es;
1704 int def_errors, def_mount_opt = nodefs ? 0 : sbi->s_def_mount_opt;
1705 const struct mount_opts *m;
1706 char sep = nodefs ? '\n' : ',';
1708 #define SEQ_OPTS_PUTS(str) seq_printf(seq, "%c" str, sep)
1709 #define SEQ_OPTS_PRINT(str, arg) seq_printf(seq, "%c" str, sep, arg)
1711 if (sbi->s_sb_block != 1)
1712 SEQ_OPTS_PRINT("sb=%llu", sbi->s_sb_block);
1714 for (m = ext4_mount_opts; m->token != Opt_err; m++) {
1715 int want_set = m->flags & MOPT_SET;
1716 if (((m->flags & (MOPT_SET|MOPT_CLEAR)) == 0) ||
1717 (m->flags & MOPT_CLEAR_ERR))
1719 if (!(m->mount_opt & (sbi->s_mount_opt ^ def_mount_opt)))
1720 continue; /* skip if same as the default */
1722 (sbi->s_mount_opt & m->mount_opt) != m->mount_opt) ||
1723 (!want_set && (sbi->s_mount_opt & m->mount_opt)))
1724 continue; /* select Opt_noFoo vs Opt_Foo */
1725 SEQ_OPTS_PRINT("%s", token2str(m->token));
1728 if (nodefs || !uid_eq(sbi->s_resuid, make_kuid(&init_user_ns, EXT4_DEF_RESUID)) ||
1729 le16_to_cpu(es->s_def_resuid) != EXT4_DEF_RESUID)
1730 SEQ_OPTS_PRINT("resuid=%u",
1731 from_kuid_munged(&init_user_ns, sbi->s_resuid));
1732 if (nodefs || !gid_eq(sbi->s_resgid, make_kgid(&init_user_ns, EXT4_DEF_RESGID)) ||
1733 le16_to_cpu(es->s_def_resgid) != EXT4_DEF_RESGID)
1734 SEQ_OPTS_PRINT("resgid=%u",
1735 from_kgid_munged(&init_user_ns, sbi->s_resgid));
1736 def_errors = nodefs ? -1 : le16_to_cpu(es->s_errors);
1737 if (test_opt(sb, ERRORS_RO) && def_errors != EXT4_ERRORS_RO)
1738 SEQ_OPTS_PUTS("errors=remount-ro");
1739 if (test_opt(sb, ERRORS_CONT) && def_errors != EXT4_ERRORS_CONTINUE)
1740 SEQ_OPTS_PUTS("errors=continue");
1741 if (test_opt(sb, ERRORS_PANIC) && def_errors != EXT4_ERRORS_PANIC)
1742 SEQ_OPTS_PUTS("errors=panic");
1743 if (nodefs || sbi->s_commit_interval != JBD2_DEFAULT_MAX_COMMIT_AGE*HZ)
1744 SEQ_OPTS_PRINT("commit=%lu", sbi->s_commit_interval / HZ);
1745 if (nodefs || sbi->s_min_batch_time != EXT4_DEF_MIN_BATCH_TIME)
1746 SEQ_OPTS_PRINT("min_batch_time=%u", sbi->s_min_batch_time);
1747 if (nodefs || sbi->s_max_batch_time != EXT4_DEF_MAX_BATCH_TIME)
1748 SEQ_OPTS_PRINT("max_batch_time=%u", sbi->s_max_batch_time);
1749 if (sb->s_flags & MS_I_VERSION)
1750 SEQ_OPTS_PUTS("i_version");
1751 if (nodefs || sbi->s_stripe)
1752 SEQ_OPTS_PRINT("stripe=%lu", sbi->s_stripe);
1753 if (EXT4_MOUNT_DATA_FLAGS & (sbi->s_mount_opt ^ def_mount_opt)) {
1754 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA)
1755 SEQ_OPTS_PUTS("data=journal");
1756 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA)
1757 SEQ_OPTS_PUTS("data=ordered");
1758 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_WRITEBACK_DATA)
1759 SEQ_OPTS_PUTS("data=writeback");
1762 sbi->s_inode_readahead_blks != EXT4_DEF_INODE_READAHEAD_BLKS)
1763 SEQ_OPTS_PRINT("inode_readahead_blks=%u",
1764 sbi->s_inode_readahead_blks);
1766 if (nodefs || (test_opt(sb, INIT_INODE_TABLE) &&
1767 (sbi->s_li_wait_mult != EXT4_DEF_LI_WAIT_MULT)))
1768 SEQ_OPTS_PRINT("init_itable=%u", sbi->s_li_wait_mult);
1769 if (nodefs || sbi->s_max_dir_size_kb)
1770 SEQ_OPTS_PRINT("max_dir_size_kb=%u", sbi->s_max_dir_size_kb);
1772 ext4_show_quota_options(seq, sb);
1776 static int ext4_show_options(struct seq_file *seq, struct dentry *root)
1778 return _ext4_show_options(seq, root->d_sb, 0);
1781 static int options_seq_show(struct seq_file *seq, void *offset)
1783 struct super_block *sb = seq->private;
1786 seq_puts(seq, (sb->s_flags & MS_RDONLY) ? "ro" : "rw");
1787 rc = _ext4_show_options(seq, sb, 1);
1788 seq_puts(seq, "\n");
1792 static int options_open_fs(struct inode *inode, struct file *file)
1794 return single_open(file, options_seq_show, PDE_DATA(inode));
1797 static const struct file_operations ext4_seq_options_fops = {
1798 .owner = THIS_MODULE,
1799 .open = options_open_fs,
1801 .llseek = seq_lseek,
1802 .release = single_release,
1805 static int ext4_setup_super(struct super_block *sb, struct ext4_super_block *es,
1808 struct ext4_sb_info *sbi = EXT4_SB(sb);
1811 if (le32_to_cpu(es->s_rev_level) > EXT4_MAX_SUPP_REV) {
1812 ext4_msg(sb, KERN_ERR, "revision level too high, "
1813 "forcing read-only mode");
1818 if (!(sbi->s_mount_state & EXT4_VALID_FS))
1819 ext4_msg(sb, KERN_WARNING, "warning: mounting unchecked fs, "
1820 "running e2fsck is recommended");
1821 else if ((sbi->s_mount_state & EXT4_ERROR_FS))
1822 ext4_msg(sb, KERN_WARNING,
1823 "warning: mounting fs with errors, "
1824 "running e2fsck is recommended");
1825 else if ((__s16) le16_to_cpu(es->s_max_mnt_count) > 0 &&
1826 le16_to_cpu(es->s_mnt_count) >=
1827 (unsigned short) (__s16) le16_to_cpu(es->s_max_mnt_count))
1828 ext4_msg(sb, KERN_WARNING,
1829 "warning: maximal mount count reached, "
1830 "running e2fsck is recommended");
1831 else if (le32_to_cpu(es->s_checkinterval) &&
1832 (le32_to_cpu(es->s_lastcheck) +
1833 le32_to_cpu(es->s_checkinterval) <= get_seconds()))
1834 ext4_msg(sb, KERN_WARNING,
1835 "warning: checktime reached, "
1836 "running e2fsck is recommended");
1837 if (!sbi->s_journal)
1838 es->s_state &= cpu_to_le16(~EXT4_VALID_FS);
1839 if (!(__s16) le16_to_cpu(es->s_max_mnt_count))
1840 es->s_max_mnt_count = cpu_to_le16(EXT4_DFL_MAX_MNT_COUNT);
1841 le16_add_cpu(&es->s_mnt_count, 1);
1842 es->s_mtime = cpu_to_le32(get_seconds());
1843 ext4_update_dynamic_rev(sb);
1845 EXT4_SET_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
1847 ext4_commit_super(sb, 1);
1849 if (test_opt(sb, DEBUG))
1850 printk(KERN_INFO "[EXT4 FS bs=%lu, gc=%u, "
1851 "bpg=%lu, ipg=%lu, mo=%04x, mo2=%04x]\n",
1853 sbi->s_groups_count,
1854 EXT4_BLOCKS_PER_GROUP(sb),
1855 EXT4_INODES_PER_GROUP(sb),
1856 sbi->s_mount_opt, sbi->s_mount_opt2);
1858 cleancache_init_fs(sb);
1862 int ext4_alloc_flex_bg_array(struct super_block *sb, ext4_group_t ngroup)
1864 struct ext4_sb_info *sbi = EXT4_SB(sb);
1865 struct flex_groups *new_groups;
1868 if (!sbi->s_log_groups_per_flex)
1871 size = ext4_flex_group(sbi, ngroup - 1) + 1;
1872 if (size <= sbi->s_flex_groups_allocated)
1875 size = roundup_pow_of_two(size * sizeof(struct flex_groups));
1876 new_groups = ext4_kvzalloc(size, GFP_KERNEL);
1878 ext4_msg(sb, KERN_ERR, "not enough memory for %d flex groups",
1879 size / (int) sizeof(struct flex_groups));
1883 if (sbi->s_flex_groups) {
1884 memcpy(new_groups, sbi->s_flex_groups,
1885 (sbi->s_flex_groups_allocated *
1886 sizeof(struct flex_groups)));
1887 ext4_kvfree(sbi->s_flex_groups);
1889 sbi->s_flex_groups = new_groups;
1890 sbi->s_flex_groups_allocated = size / sizeof(struct flex_groups);
1894 static int ext4_fill_flex_info(struct super_block *sb)
1896 struct ext4_sb_info *sbi = EXT4_SB(sb);
1897 struct ext4_group_desc *gdp = NULL;
1898 ext4_group_t flex_group;
1899 unsigned int groups_per_flex = 0;
1902 sbi->s_log_groups_per_flex = sbi->s_es->s_log_groups_per_flex;
1903 if (sbi->s_log_groups_per_flex < 1 || sbi->s_log_groups_per_flex > 31) {
1904 sbi->s_log_groups_per_flex = 0;
1907 groups_per_flex = 1U << sbi->s_log_groups_per_flex;
1909 err = ext4_alloc_flex_bg_array(sb, sbi->s_groups_count);
1913 for (i = 0; i < sbi->s_groups_count; i++) {
1914 gdp = ext4_get_group_desc(sb, i, NULL);
1916 flex_group = ext4_flex_group(sbi, i);
1917 atomic_add(ext4_free_inodes_count(sb, gdp),
1918 &sbi->s_flex_groups[flex_group].free_inodes);
1919 atomic64_add(ext4_free_group_clusters(sb, gdp),
1920 &sbi->s_flex_groups[flex_group].free_clusters);
1921 atomic_add(ext4_used_dirs_count(sb, gdp),
1922 &sbi->s_flex_groups[flex_group].used_dirs);
1930 static __le16 ext4_group_desc_csum(struct ext4_sb_info *sbi, __u32 block_group,
1931 struct ext4_group_desc *gdp)
1935 __le32 le_group = cpu_to_le32(block_group);
1937 if ((sbi->s_es->s_feature_ro_compat &
1938 cpu_to_le32(EXT4_FEATURE_RO_COMPAT_METADATA_CSUM))) {
1939 /* Use new metadata_csum algorithm */
1943 save_csum = gdp->bg_checksum;
1944 gdp->bg_checksum = 0;
1945 csum32 = ext4_chksum(sbi, sbi->s_csum_seed, (__u8 *)&le_group,
1947 csum32 = ext4_chksum(sbi, csum32, (__u8 *)gdp,
1949 gdp->bg_checksum = save_csum;
1951 crc = csum32 & 0xFFFF;
1955 /* old crc16 code */
1956 if (!(sbi->s_es->s_feature_ro_compat &
1957 cpu_to_le32(EXT4_FEATURE_RO_COMPAT_GDT_CSUM)))
1960 offset = offsetof(struct ext4_group_desc, bg_checksum);
1962 crc = crc16(~0, sbi->s_es->s_uuid, sizeof(sbi->s_es->s_uuid));
1963 crc = crc16(crc, (__u8 *)&le_group, sizeof(le_group));
1964 crc = crc16(crc, (__u8 *)gdp, offset);
1965 offset += sizeof(gdp->bg_checksum); /* skip checksum */
1966 /* for checksum of struct ext4_group_desc do the rest...*/
1967 if ((sbi->s_es->s_feature_incompat &
1968 cpu_to_le32(EXT4_FEATURE_INCOMPAT_64BIT)) &&
1969 offset < le16_to_cpu(sbi->s_es->s_desc_size))
1970 crc = crc16(crc, (__u8 *)gdp + offset,
1971 le16_to_cpu(sbi->s_es->s_desc_size) -
1975 return cpu_to_le16(crc);
1978 int ext4_group_desc_csum_verify(struct super_block *sb, __u32 block_group,
1979 struct ext4_group_desc *gdp)
1981 if (ext4_has_group_desc_csum(sb) &&
1982 (gdp->bg_checksum != ext4_group_desc_csum(EXT4_SB(sb),
1989 void ext4_group_desc_csum_set(struct super_block *sb, __u32 block_group,
1990 struct ext4_group_desc *gdp)
1992 if (!ext4_has_group_desc_csum(sb))
1994 gdp->bg_checksum = ext4_group_desc_csum(EXT4_SB(sb), block_group, gdp);
1997 /* Called at mount-time, super-block is locked */
1998 static int ext4_check_descriptors(struct super_block *sb,
1999 ext4_group_t *first_not_zeroed)
2001 struct ext4_sb_info *sbi = EXT4_SB(sb);
2002 ext4_fsblk_t first_block = le32_to_cpu(sbi->s_es->s_first_data_block);
2003 ext4_fsblk_t last_block;
2004 ext4_fsblk_t block_bitmap;
2005 ext4_fsblk_t inode_bitmap;
2006 ext4_fsblk_t inode_table;
2007 int flexbg_flag = 0;
2008 ext4_group_t i, grp = sbi->s_groups_count;
2010 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_FLEX_BG))
2013 ext4_debug("Checking group descriptors");
2015 for (i = 0; i < sbi->s_groups_count; i++) {
2016 struct ext4_group_desc *gdp = ext4_get_group_desc(sb, i, NULL);
2018 if (i == sbi->s_groups_count - 1 || flexbg_flag)
2019 last_block = ext4_blocks_count(sbi->s_es) - 1;
2021 last_block = first_block +
2022 (EXT4_BLOCKS_PER_GROUP(sb) - 1);
2024 if ((grp == sbi->s_groups_count) &&
2025 !(gdp->bg_flags & cpu_to_le16(EXT4_BG_INODE_ZEROED)))
2028 block_bitmap = ext4_block_bitmap(sb, gdp);
2029 if (block_bitmap < first_block || block_bitmap > last_block) {
2030 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2031 "Block bitmap for group %u not in group "
2032 "(block %llu)!", i, block_bitmap);
2035 inode_bitmap = ext4_inode_bitmap(sb, gdp);
2036 if (inode_bitmap < first_block || inode_bitmap > last_block) {
2037 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2038 "Inode bitmap for group %u not in group "
2039 "(block %llu)!", i, inode_bitmap);
2042 inode_table = ext4_inode_table(sb, gdp);
2043 if (inode_table < first_block ||
2044 inode_table + sbi->s_itb_per_group - 1 > last_block) {
2045 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2046 "Inode table for group %u not in group "
2047 "(block %llu)!", i, inode_table);
2050 ext4_lock_group(sb, i);
2051 if (!ext4_group_desc_csum_verify(sb, i, gdp)) {
2052 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2053 "Checksum for group %u failed (%u!=%u)",
2054 i, le16_to_cpu(ext4_group_desc_csum(sbi, i,
2055 gdp)), le16_to_cpu(gdp->bg_checksum));
2056 if (!(sb->s_flags & MS_RDONLY)) {
2057 ext4_unlock_group(sb, i);
2061 ext4_unlock_group(sb, i);
2063 first_block += EXT4_BLOCKS_PER_GROUP(sb);
2065 if (NULL != first_not_zeroed)
2066 *first_not_zeroed = grp;
2068 ext4_free_blocks_count_set(sbi->s_es,
2069 EXT4_C2B(sbi, ext4_count_free_clusters(sb)));
2070 sbi->s_es->s_free_inodes_count =cpu_to_le32(ext4_count_free_inodes(sb));
2074 /* ext4_orphan_cleanup() walks a singly-linked list of inodes (starting at
2075 * the superblock) which were deleted from all directories, but held open by
2076 * a process at the time of a crash. We walk the list and try to delete these
2077 * inodes at recovery time (only with a read-write filesystem).
2079 * In order to keep the orphan inode chain consistent during traversal (in
2080 * case of crash during recovery), we link each inode into the superblock
2081 * orphan list_head and handle it the same way as an inode deletion during
2082 * normal operation (which journals the operations for us).
2084 * We only do an iget() and an iput() on each inode, which is very safe if we
2085 * accidentally point at an in-use or already deleted inode. The worst that
2086 * can happen in this case is that we get a "bit already cleared" message from
2087 * ext4_free_inode(). The only reason we would point at a wrong inode is if
2088 * e2fsck was run on this filesystem, and it must have already done the orphan
2089 * inode cleanup for us, so we can safely abort without any further action.
2091 static void ext4_orphan_cleanup(struct super_block *sb,
2092 struct ext4_super_block *es)
2094 unsigned int s_flags = sb->s_flags;
2095 int nr_orphans = 0, nr_truncates = 0;
2099 if (!es->s_last_orphan) {
2100 jbd_debug(4, "no orphan inodes to clean up\n");
2104 if (bdev_read_only(sb->s_bdev)) {
2105 ext4_msg(sb, KERN_ERR, "write access "
2106 "unavailable, skipping orphan cleanup");
2110 /* Check if feature set would not allow a r/w mount */
2111 if (!ext4_feature_set_ok(sb, 0)) {
2112 ext4_msg(sb, KERN_INFO, "Skipping orphan cleanup due to "
2113 "unknown ROCOMPAT features");
2117 if (EXT4_SB(sb)->s_mount_state & EXT4_ERROR_FS) {
2118 /* don't clear list on RO mount w/ errors */
2119 if (es->s_last_orphan && !(s_flags & MS_RDONLY)) {
2120 jbd_debug(1, "Errors on filesystem, "
2121 "clearing orphan list.\n");
2122 es->s_last_orphan = 0;
2124 jbd_debug(1, "Skipping orphan recovery on fs with errors.\n");
2128 if (s_flags & MS_RDONLY) {
2129 ext4_msg(sb, KERN_INFO, "orphan cleanup on readonly fs");
2130 sb->s_flags &= ~MS_RDONLY;
2133 /* Needed for iput() to work correctly and not trash data */
2134 sb->s_flags |= MS_ACTIVE;
2135 /* Turn on quotas so that they are updated correctly */
2136 for (i = 0; i < MAXQUOTAS; i++) {
2137 if (EXT4_SB(sb)->s_qf_names[i]) {
2138 int ret = ext4_quota_on_mount(sb, i);
2140 ext4_msg(sb, KERN_ERR,
2141 "Cannot turn on journaled "
2142 "quota: error %d", ret);
2147 while (es->s_last_orphan) {
2148 struct inode *inode;
2150 inode = ext4_orphan_get(sb, le32_to_cpu(es->s_last_orphan));
2151 if (IS_ERR(inode)) {
2152 es->s_last_orphan = 0;
2156 list_add(&EXT4_I(inode)->i_orphan, &EXT4_SB(sb)->s_orphan);
2157 dquot_initialize(inode);
2158 if (inode->i_nlink) {
2159 ext4_msg(sb, KERN_DEBUG,
2160 "%s: truncating inode %lu to %lld bytes",
2161 __func__, inode->i_ino, inode->i_size);
2162 jbd_debug(2, "truncating inode %lu to %lld bytes\n",
2163 inode->i_ino, inode->i_size);
2164 mutex_lock(&inode->i_mutex);
2165 ext4_truncate(inode);
2166 mutex_unlock(&inode->i_mutex);
2169 ext4_msg(sb, KERN_DEBUG,
2170 "%s: deleting unreferenced inode %lu",
2171 __func__, inode->i_ino);
2172 jbd_debug(2, "deleting unreferenced inode %lu\n",
2176 iput(inode); /* The delete magic happens here! */
2179 #define PLURAL(x) (x), ((x) == 1) ? "" : "s"
2182 ext4_msg(sb, KERN_INFO, "%d orphan inode%s deleted",
2183 PLURAL(nr_orphans));
2185 ext4_msg(sb, KERN_INFO, "%d truncate%s cleaned up",
2186 PLURAL(nr_truncates));
2188 /* Turn quotas off */
2189 for (i = 0; i < MAXQUOTAS; i++) {
2190 if (sb_dqopt(sb)->files[i])
2191 dquot_quota_off(sb, i);
2194 sb->s_flags = s_flags; /* Restore MS_RDONLY status */
2198 * Maximal extent format file size.
2199 * Resulting logical blkno at s_maxbytes must fit in our on-disk
2200 * extent format containers, within a sector_t, and within i_blocks
2201 * in the vfs. ext4 inode has 48 bits of i_block in fsblock units,
2202 * so that won't be a limiting factor.
2204 * However there is other limiting factor. We do store extents in the form
2205 * of starting block and length, hence the resulting length of the extent
2206 * covering maximum file size must fit into on-disk format containers as
2207 * well. Given that length is always by 1 unit bigger than max unit (because
2208 * we count 0 as well) we have to lower the s_maxbytes by one fs block.
2210 * Note, this does *not* consider any metadata overhead for vfs i_blocks.
2212 static loff_t ext4_max_size(int blkbits, int has_huge_files)
2215 loff_t upper_limit = MAX_LFS_FILESIZE;
2217 /* small i_blocks in vfs inode? */
2218 if (!has_huge_files || sizeof(blkcnt_t) < sizeof(u64)) {
2220 * CONFIG_LBDAF is not enabled implies the inode
2221 * i_block represent total blocks in 512 bytes
2222 * 32 == size of vfs inode i_blocks * 8
2224 upper_limit = (1LL << 32) - 1;
2226 /* total blocks in file system block size */
2227 upper_limit >>= (blkbits - 9);
2228 upper_limit <<= blkbits;
2232 * 32-bit extent-start container, ee_block. We lower the maxbytes
2233 * by one fs block, so ee_len can cover the extent of maximum file
2236 res = (1LL << 32) - 1;
2239 /* Sanity check against vm- & vfs- imposed limits */
2240 if (res > upper_limit)
2247 * Maximal bitmap file size. There is a direct, and {,double-,triple-}indirect
2248 * block limit, and also a limit of (2^48 - 1) 512-byte sectors in i_blocks.
2249 * We need to be 1 filesystem block less than the 2^48 sector limit.
2251 static loff_t ext4_max_bitmap_size(int bits, int has_huge_files)
2253 loff_t res = EXT4_NDIR_BLOCKS;
2256 /* This is calculated to be the largest file size for a dense, block
2257 * mapped file such that the file's total number of 512-byte sectors,
2258 * including data and all indirect blocks, does not exceed (2^48 - 1).
2260 * __u32 i_blocks_lo and _u16 i_blocks_high represent the total
2261 * number of 512-byte sectors of the file.
2264 if (!has_huge_files || sizeof(blkcnt_t) < sizeof(u64)) {
2266 * !has_huge_files or CONFIG_LBDAF not enabled implies that
2267 * the inode i_block field represents total file blocks in
2268 * 2^32 512-byte sectors == size of vfs inode i_blocks * 8
2270 upper_limit = (1LL << 32) - 1;
2272 /* total blocks in file system block size */
2273 upper_limit >>= (bits - 9);
2277 * We use 48 bit ext4_inode i_blocks
2278 * With EXT4_HUGE_FILE_FL set the i_blocks
2279 * represent total number of blocks in
2280 * file system block size
2282 upper_limit = (1LL << 48) - 1;
2286 /* indirect blocks */
2288 /* double indirect blocks */
2289 meta_blocks += 1 + (1LL << (bits-2));
2290 /* tripple indirect blocks */
2291 meta_blocks += 1 + (1LL << (bits-2)) + (1LL << (2*(bits-2)));
2293 upper_limit -= meta_blocks;
2294 upper_limit <<= bits;
2296 res += 1LL << (bits-2);
2297 res += 1LL << (2*(bits-2));
2298 res += 1LL << (3*(bits-2));
2300 if (res > upper_limit)
2303 if (res > MAX_LFS_FILESIZE)
2304 res = MAX_LFS_FILESIZE;
2309 static ext4_fsblk_t descriptor_loc(struct super_block *sb,
2310 ext4_fsblk_t logical_sb_block, int nr)
2312 struct ext4_sb_info *sbi = EXT4_SB(sb);
2313 ext4_group_t bg, first_meta_bg;
2316 first_meta_bg = le32_to_cpu(sbi->s_es->s_first_meta_bg);
2318 if (!EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_META_BG) ||
2320 return logical_sb_block + nr + 1;
2321 bg = sbi->s_desc_per_block * nr;
2322 if (ext4_bg_has_super(sb, bg))
2325 return (has_super + ext4_group_first_block_no(sb, bg));
2329 * ext4_get_stripe_size: Get the stripe size.
2330 * @sbi: In memory super block info
2332 * If we have specified it via mount option, then
2333 * use the mount option value. If the value specified at mount time is
2334 * greater than the blocks per group use the super block value.
2335 * If the super block value is greater than blocks per group return 0.
2336 * Allocator needs it be less than blocks per group.
2339 static unsigned long ext4_get_stripe_size(struct ext4_sb_info *sbi)
2341 unsigned long stride = le16_to_cpu(sbi->s_es->s_raid_stride);
2342 unsigned long stripe_width =
2343 le32_to_cpu(sbi->s_es->s_raid_stripe_width);
2346 if (sbi->s_stripe && sbi->s_stripe <= sbi->s_blocks_per_group)
2347 ret = sbi->s_stripe;
2348 else if (stripe_width <= sbi->s_blocks_per_group)
2350 else if (stride <= sbi->s_blocks_per_group)
2356 * If the stripe width is 1, this makes no sense and
2357 * we set it to 0 to turn off stripe handling code.
2368 struct attribute attr;
2369 ssize_t (*show)(struct ext4_attr *, struct ext4_sb_info *, char *);
2370 ssize_t (*store)(struct ext4_attr *, struct ext4_sb_info *,
2371 const char *, size_t);
2375 static int parse_strtoull(const char *buf,
2376 unsigned long long max, unsigned long long *value)
2380 ret = kstrtoull(skip_spaces(buf), 0, value);
2381 if (!ret && *value > max)
2386 static ssize_t delayed_allocation_blocks_show(struct ext4_attr *a,
2387 struct ext4_sb_info *sbi,
2390 return snprintf(buf, PAGE_SIZE, "%llu\n",
2392 percpu_counter_sum(&sbi->s_dirtyclusters_counter)));
2395 static ssize_t session_write_kbytes_show(struct ext4_attr *a,
2396 struct ext4_sb_info *sbi, char *buf)
2398 struct super_block *sb = sbi->s_buddy_cache->i_sb;
2400 if (!sb->s_bdev->bd_part)
2401 return snprintf(buf, PAGE_SIZE, "0\n");
2402 return snprintf(buf, PAGE_SIZE, "%lu\n",
2403 (part_stat_read(sb->s_bdev->bd_part, sectors[1]) -
2404 sbi->s_sectors_written_start) >> 1);
2407 static ssize_t lifetime_write_kbytes_show(struct ext4_attr *a,
2408 struct ext4_sb_info *sbi, char *buf)
2410 struct super_block *sb = sbi->s_buddy_cache->i_sb;
2412 if (!sb->s_bdev->bd_part)
2413 return snprintf(buf, PAGE_SIZE, "0\n");
2414 return snprintf(buf, PAGE_SIZE, "%llu\n",
2415 (unsigned long long)(sbi->s_kbytes_written +
2416 ((part_stat_read(sb->s_bdev->bd_part, sectors[1]) -
2417 EXT4_SB(sb)->s_sectors_written_start) >> 1)));
2420 static ssize_t inode_readahead_blks_store(struct ext4_attr *a,
2421 struct ext4_sb_info *sbi,
2422 const char *buf, size_t count)
2427 ret = kstrtoul(skip_spaces(buf), 0, &t);
2431 if (t && (!is_power_of_2(t) || t > 0x40000000))
2434 sbi->s_inode_readahead_blks = t;
2438 static ssize_t sbi_ui_show(struct ext4_attr *a,
2439 struct ext4_sb_info *sbi, char *buf)
2441 unsigned int *ui = (unsigned int *) (((char *) sbi) + a->offset);
2443 return snprintf(buf, PAGE_SIZE, "%u\n", *ui);
2446 static ssize_t sbi_ui_store(struct ext4_attr *a,
2447 struct ext4_sb_info *sbi,
2448 const char *buf, size_t count)
2450 unsigned int *ui = (unsigned int *) (((char *) sbi) + a->offset);
2454 ret = kstrtoul(skip_spaces(buf), 0, &t);
2461 static ssize_t reserved_clusters_show(struct ext4_attr *a,
2462 struct ext4_sb_info *sbi, char *buf)
2464 return snprintf(buf, PAGE_SIZE, "%llu\n",
2465 (unsigned long long) atomic64_read(&sbi->s_resv_clusters));
2468 static ssize_t reserved_clusters_store(struct ext4_attr *a,
2469 struct ext4_sb_info *sbi,
2470 const char *buf, size_t count)
2472 unsigned long long val;
2475 if (parse_strtoull(buf, -1ULL, &val))
2477 ret = ext4_reserve_clusters(sbi, val);
2479 return ret ? ret : count;
2482 static ssize_t trigger_test_error(struct ext4_attr *a,
2483 struct ext4_sb_info *sbi,
2484 const char *buf, size_t count)
2488 if (!capable(CAP_SYS_ADMIN))
2491 if (len && buf[len-1] == '\n')
2495 ext4_error(sbi->s_sb, "%.*s", len, buf);
2499 #define EXT4_ATTR_OFFSET(_name,_mode,_show,_store,_elname) \
2500 static struct ext4_attr ext4_attr_##_name = { \
2501 .attr = {.name = __stringify(_name), .mode = _mode }, \
2504 .offset = offsetof(struct ext4_sb_info, _elname), \
2506 #define EXT4_ATTR(name, mode, show, store) \
2507 static struct ext4_attr ext4_attr_##name = __ATTR(name, mode, show, store)
2509 #define EXT4_INFO_ATTR(name) EXT4_ATTR(name, 0444, NULL, NULL)
2510 #define EXT4_RO_ATTR(name) EXT4_ATTR(name, 0444, name##_show, NULL)
2511 #define EXT4_RW_ATTR(name) EXT4_ATTR(name, 0644, name##_show, name##_store)
2512 #define EXT4_RW_ATTR_SBI_UI(name, elname) \
2513 EXT4_ATTR_OFFSET(name, 0644, sbi_ui_show, sbi_ui_store, elname)
2514 #define ATTR_LIST(name) &ext4_attr_##name.attr
2516 EXT4_RO_ATTR(delayed_allocation_blocks);
2517 EXT4_RO_ATTR(session_write_kbytes);
2518 EXT4_RO_ATTR(lifetime_write_kbytes);
2519 EXT4_RW_ATTR(reserved_clusters);
2520 EXT4_ATTR_OFFSET(inode_readahead_blks, 0644, sbi_ui_show,
2521 inode_readahead_blks_store, s_inode_readahead_blks);
2522 EXT4_RW_ATTR_SBI_UI(inode_goal, s_inode_goal);
2523 EXT4_RW_ATTR_SBI_UI(mb_stats, s_mb_stats);
2524 EXT4_RW_ATTR_SBI_UI(mb_max_to_scan, s_mb_max_to_scan);
2525 EXT4_RW_ATTR_SBI_UI(mb_min_to_scan, s_mb_min_to_scan);
2526 EXT4_RW_ATTR_SBI_UI(mb_order2_req, s_mb_order2_reqs);
2527 EXT4_RW_ATTR_SBI_UI(mb_stream_req, s_mb_stream_request);
2528 EXT4_RW_ATTR_SBI_UI(mb_group_prealloc, s_mb_group_prealloc);
2529 EXT4_RW_ATTR_SBI_UI(max_writeback_mb_bump, s_max_writeback_mb_bump);
2530 EXT4_RW_ATTR_SBI_UI(extent_max_zeroout_kb, s_extent_max_zeroout_kb);
2531 EXT4_ATTR(trigger_fs_error, 0200, NULL, trigger_test_error);
2533 static struct attribute *ext4_attrs[] = {
2534 ATTR_LIST(delayed_allocation_blocks),
2535 ATTR_LIST(session_write_kbytes),
2536 ATTR_LIST(lifetime_write_kbytes),
2537 ATTR_LIST(reserved_clusters),
2538 ATTR_LIST(inode_readahead_blks),
2539 ATTR_LIST(inode_goal),
2540 ATTR_LIST(mb_stats),
2541 ATTR_LIST(mb_max_to_scan),
2542 ATTR_LIST(mb_min_to_scan),
2543 ATTR_LIST(mb_order2_req),
2544 ATTR_LIST(mb_stream_req),
2545 ATTR_LIST(mb_group_prealloc),
2546 ATTR_LIST(max_writeback_mb_bump),
2547 ATTR_LIST(extent_max_zeroout_kb),
2548 ATTR_LIST(trigger_fs_error),
2552 /* Features this copy of ext4 supports */
2553 EXT4_INFO_ATTR(lazy_itable_init);
2554 EXT4_INFO_ATTR(batched_discard);
2555 EXT4_INFO_ATTR(meta_bg_resize);
2557 static struct attribute *ext4_feat_attrs[] = {
2558 ATTR_LIST(lazy_itable_init),
2559 ATTR_LIST(batched_discard),
2560 ATTR_LIST(meta_bg_resize),
2564 static ssize_t ext4_attr_show(struct kobject *kobj,
2565 struct attribute *attr, char *buf)
2567 struct ext4_sb_info *sbi = container_of(kobj, struct ext4_sb_info,
2569 struct ext4_attr *a = container_of(attr, struct ext4_attr, attr);
2571 return a->show ? a->show(a, sbi, buf) : 0;
2574 static ssize_t ext4_attr_store(struct kobject *kobj,
2575 struct attribute *attr,
2576 const char *buf, size_t len)
2578 struct ext4_sb_info *sbi = container_of(kobj, struct ext4_sb_info,
2580 struct ext4_attr *a = container_of(attr, struct ext4_attr, attr);
2582 return a->store ? a->store(a, sbi, buf, len) : 0;
2585 static void ext4_sb_release(struct kobject *kobj)
2587 struct ext4_sb_info *sbi = container_of(kobj, struct ext4_sb_info,
2589 complete(&sbi->s_kobj_unregister);
2592 static const struct sysfs_ops ext4_attr_ops = {
2593 .show = ext4_attr_show,
2594 .store = ext4_attr_store,
2597 static struct kobj_type ext4_ktype = {
2598 .default_attrs = ext4_attrs,
2599 .sysfs_ops = &ext4_attr_ops,
2600 .release = ext4_sb_release,
2603 static void ext4_feat_release(struct kobject *kobj)
2605 complete(&ext4_feat->f_kobj_unregister);
2608 static struct kobj_type ext4_feat_ktype = {
2609 .default_attrs = ext4_feat_attrs,
2610 .sysfs_ops = &ext4_attr_ops,
2611 .release = ext4_feat_release,
2615 * Check whether this filesystem can be mounted based on
2616 * the features present and the RDONLY/RDWR mount requested.
2617 * Returns 1 if this filesystem can be mounted as requested,
2618 * 0 if it cannot be.
2620 static int ext4_feature_set_ok(struct super_block *sb, int readonly)
2622 if (EXT4_HAS_INCOMPAT_FEATURE(sb, ~EXT4_FEATURE_INCOMPAT_SUPP)) {
2623 ext4_msg(sb, KERN_ERR,
2624 "Couldn't mount because of "
2625 "unsupported optional features (%x)",
2626 (le32_to_cpu(EXT4_SB(sb)->s_es->s_feature_incompat) &
2627 ~EXT4_FEATURE_INCOMPAT_SUPP));
2634 /* Check that feature set is OK for a read-write mount */
2635 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, ~EXT4_FEATURE_RO_COMPAT_SUPP)) {
2636 ext4_msg(sb, KERN_ERR, "couldn't mount RDWR because of "
2637 "unsupported optional features (%x)",
2638 (le32_to_cpu(EXT4_SB(sb)->s_es->s_feature_ro_compat) &
2639 ~EXT4_FEATURE_RO_COMPAT_SUPP));
2643 * Large file size enabled file system can only be mounted
2644 * read-write on 32-bit systems if kernel is built with CONFIG_LBDAF
2646 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_HUGE_FILE)) {
2647 if (sizeof(blkcnt_t) < sizeof(u64)) {
2648 ext4_msg(sb, KERN_ERR, "Filesystem with huge files "
2649 "cannot be mounted RDWR without "
2654 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_BIGALLOC) &&
2655 !EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_EXTENTS)) {
2656 ext4_msg(sb, KERN_ERR,
2657 "Can't support bigalloc feature without "
2658 "extents feature\n");
2662 #ifndef CONFIG_QUOTA
2663 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_QUOTA) &&
2665 ext4_msg(sb, KERN_ERR,
2666 "Filesystem with quota feature cannot be mounted RDWR "
2667 "without CONFIG_QUOTA");
2670 #endif /* CONFIG_QUOTA */
2675 * This function is called once a day if we have errors logged
2676 * on the file system
2678 static void print_daily_error_info(unsigned long arg)
2680 struct super_block *sb = (struct super_block *) arg;
2681 struct ext4_sb_info *sbi;
2682 struct ext4_super_block *es;
2687 if (es->s_error_count)
2688 /* fsck newer than v1.41.13 is needed to clean this condition. */
2689 ext4_msg(sb, KERN_NOTICE, "error count since last fsck: %u",
2690 le32_to_cpu(es->s_error_count));
2691 if (es->s_first_error_time) {
2692 printk(KERN_NOTICE "EXT4-fs (%s): initial error at time %u: %.*s:%d",
2693 sb->s_id, le32_to_cpu(es->s_first_error_time),
2694 (int) sizeof(es->s_first_error_func),
2695 es->s_first_error_func,
2696 le32_to_cpu(es->s_first_error_line));
2697 if (es->s_first_error_ino)
2698 printk(": inode %u",
2699 le32_to_cpu(es->s_first_error_ino));
2700 if (es->s_first_error_block)
2701 printk(": block %llu", (unsigned long long)
2702 le64_to_cpu(es->s_first_error_block));
2705 if (es->s_last_error_time) {
2706 printk(KERN_NOTICE "EXT4-fs (%s): last error at time %u: %.*s:%d",
2707 sb->s_id, le32_to_cpu(es->s_last_error_time),
2708 (int) sizeof(es->s_last_error_func),
2709 es->s_last_error_func,
2710 le32_to_cpu(es->s_last_error_line));
2711 if (es->s_last_error_ino)
2712 printk(": inode %u",
2713 le32_to_cpu(es->s_last_error_ino));
2714 if (es->s_last_error_block)
2715 printk(": block %llu", (unsigned long long)
2716 le64_to_cpu(es->s_last_error_block));
2719 mod_timer(&sbi->s_err_report, jiffies + 24*60*60*HZ); /* Once a day */
2722 /* Find next suitable group and run ext4_init_inode_table */
2723 static int ext4_run_li_request(struct ext4_li_request *elr)
2725 struct ext4_group_desc *gdp = NULL;
2726 ext4_group_t group, ngroups;
2727 struct super_block *sb;
2728 unsigned long timeout = 0;
2732 ngroups = EXT4_SB(sb)->s_groups_count;
2735 for (group = elr->lr_next_group; group < ngroups; group++) {
2736 gdp = ext4_get_group_desc(sb, group, NULL);
2742 if (!(gdp->bg_flags & cpu_to_le16(EXT4_BG_INODE_ZEROED)))
2746 if (group >= ngroups)
2751 ret = ext4_init_inode_table(sb, group,
2752 elr->lr_timeout ? 0 : 1);
2753 if (elr->lr_timeout == 0) {
2754 timeout = (jiffies - timeout) *
2755 elr->lr_sbi->s_li_wait_mult;
2756 elr->lr_timeout = timeout;
2758 elr->lr_next_sched = jiffies + elr->lr_timeout;
2759 elr->lr_next_group = group + 1;
2767 * Remove lr_request from the list_request and free the
2768 * request structure. Should be called with li_list_mtx held
2770 static void ext4_remove_li_request(struct ext4_li_request *elr)
2772 struct ext4_sb_info *sbi;
2779 list_del(&elr->lr_request);
2780 sbi->s_li_request = NULL;
2784 static void ext4_unregister_li_request(struct super_block *sb)
2786 mutex_lock(&ext4_li_mtx);
2787 if (!ext4_li_info) {
2788 mutex_unlock(&ext4_li_mtx);
2792 mutex_lock(&ext4_li_info->li_list_mtx);
2793 ext4_remove_li_request(EXT4_SB(sb)->s_li_request);
2794 mutex_unlock(&ext4_li_info->li_list_mtx);
2795 mutex_unlock(&ext4_li_mtx);
2798 static struct task_struct *ext4_lazyinit_task;
2801 * This is the function where ext4lazyinit thread lives. It walks
2802 * through the request list searching for next scheduled filesystem.
2803 * When such a fs is found, run the lazy initialization request
2804 * (ext4_rn_li_request) and keep track of the time spend in this
2805 * function. Based on that time we compute next schedule time of
2806 * the request. When walking through the list is complete, compute
2807 * next waking time and put itself into sleep.
2809 static int ext4_lazyinit_thread(void *arg)
2811 struct ext4_lazy_init *eli = (struct ext4_lazy_init *)arg;
2812 struct list_head *pos, *n;
2813 struct ext4_li_request *elr;
2814 unsigned long next_wakeup, cur;
2816 BUG_ON(NULL == eli);
2820 next_wakeup = MAX_JIFFY_OFFSET;
2822 mutex_lock(&eli->li_list_mtx);
2823 if (list_empty(&eli->li_request_list)) {
2824 mutex_unlock(&eli->li_list_mtx);
2828 list_for_each_safe(pos, n, &eli->li_request_list) {
2829 elr = list_entry(pos, struct ext4_li_request,
2832 if (time_after_eq(jiffies, elr->lr_next_sched)) {
2833 if (ext4_run_li_request(elr) != 0) {
2834 /* error, remove the lazy_init job */
2835 ext4_remove_li_request(elr);
2840 if (time_before(elr->lr_next_sched, next_wakeup))
2841 next_wakeup = elr->lr_next_sched;
2843 mutex_unlock(&eli->li_list_mtx);
2848 if ((time_after_eq(cur, next_wakeup)) ||
2849 (MAX_JIFFY_OFFSET == next_wakeup)) {
2854 schedule_timeout_interruptible(next_wakeup - cur);
2856 if (kthread_should_stop()) {
2857 ext4_clear_request_list();
2864 * It looks like the request list is empty, but we need
2865 * to check it under the li_list_mtx lock, to prevent any
2866 * additions into it, and of course we should lock ext4_li_mtx
2867 * to atomically free the list and ext4_li_info, because at
2868 * this point another ext4 filesystem could be registering
2871 mutex_lock(&ext4_li_mtx);
2872 mutex_lock(&eli->li_list_mtx);
2873 if (!list_empty(&eli->li_request_list)) {
2874 mutex_unlock(&eli->li_list_mtx);
2875 mutex_unlock(&ext4_li_mtx);
2878 mutex_unlock(&eli->li_list_mtx);
2879 kfree(ext4_li_info);
2880 ext4_li_info = NULL;
2881 mutex_unlock(&ext4_li_mtx);
2886 static void ext4_clear_request_list(void)
2888 struct list_head *pos, *n;
2889 struct ext4_li_request *elr;
2891 mutex_lock(&ext4_li_info->li_list_mtx);
2892 list_for_each_safe(pos, n, &ext4_li_info->li_request_list) {
2893 elr = list_entry(pos, struct ext4_li_request,
2895 ext4_remove_li_request(elr);
2897 mutex_unlock(&ext4_li_info->li_list_mtx);
2900 static int ext4_run_lazyinit_thread(void)
2902 ext4_lazyinit_task = kthread_run(ext4_lazyinit_thread,
2903 ext4_li_info, "ext4lazyinit");
2904 if (IS_ERR(ext4_lazyinit_task)) {
2905 int err = PTR_ERR(ext4_lazyinit_task);
2906 ext4_clear_request_list();
2907 kfree(ext4_li_info);
2908 ext4_li_info = NULL;
2909 printk(KERN_CRIT "EXT4-fs: error %d creating inode table "
2910 "initialization thread\n",
2914 ext4_li_info->li_state |= EXT4_LAZYINIT_RUNNING;
2919 * Check whether it make sense to run itable init. thread or not.
2920 * If there is at least one uninitialized inode table, return
2921 * corresponding group number, else the loop goes through all
2922 * groups and return total number of groups.
2924 static ext4_group_t ext4_has_uninit_itable(struct super_block *sb)
2926 ext4_group_t group, ngroups = EXT4_SB(sb)->s_groups_count;
2927 struct ext4_group_desc *gdp = NULL;
2929 for (group = 0; group < ngroups; group++) {
2930 gdp = ext4_get_group_desc(sb, group, NULL);
2934 if (!(gdp->bg_flags & cpu_to_le16(EXT4_BG_INODE_ZEROED)))
2941 static int ext4_li_info_new(void)
2943 struct ext4_lazy_init *eli = NULL;
2945 eli = kzalloc(sizeof(*eli), GFP_KERNEL);
2949 INIT_LIST_HEAD(&eli->li_request_list);
2950 mutex_init(&eli->li_list_mtx);
2952 eli->li_state |= EXT4_LAZYINIT_QUIT;
2959 static struct ext4_li_request *ext4_li_request_new(struct super_block *sb,
2962 struct ext4_sb_info *sbi = EXT4_SB(sb);
2963 struct ext4_li_request *elr;
2966 elr = kzalloc(sizeof(*elr), GFP_KERNEL);
2972 elr->lr_next_group = start;
2975 * Randomize first schedule time of the request to
2976 * spread the inode table initialization requests
2979 get_random_bytes(&rnd, sizeof(rnd));
2980 elr->lr_next_sched = jiffies + (unsigned long)rnd %
2981 (EXT4_DEF_LI_MAX_START_DELAY * HZ);
2986 int ext4_register_li_request(struct super_block *sb,
2987 ext4_group_t first_not_zeroed)
2989 struct ext4_sb_info *sbi = EXT4_SB(sb);
2990 struct ext4_li_request *elr = NULL;
2991 ext4_group_t ngroups = EXT4_SB(sb)->s_groups_count;
2994 mutex_lock(&ext4_li_mtx);
2995 if (sbi->s_li_request != NULL) {
2997 * Reset timeout so it can be computed again, because
2998 * s_li_wait_mult might have changed.
3000 sbi->s_li_request->lr_timeout = 0;
3004 if (first_not_zeroed == ngroups ||
3005 (sb->s_flags & MS_RDONLY) ||
3006 !test_opt(sb, INIT_INODE_TABLE))
3009 elr = ext4_li_request_new(sb, first_not_zeroed);
3015 if (NULL == ext4_li_info) {
3016 ret = ext4_li_info_new();
3021 mutex_lock(&ext4_li_info->li_list_mtx);
3022 list_add(&elr->lr_request, &ext4_li_info->li_request_list);
3023 mutex_unlock(&ext4_li_info->li_list_mtx);
3025 sbi->s_li_request = elr;
3027 * set elr to NULL here since it has been inserted to
3028 * the request_list and the removal and free of it is
3029 * handled by ext4_clear_request_list from now on.
3033 if (!(ext4_li_info->li_state & EXT4_LAZYINIT_RUNNING)) {
3034 ret = ext4_run_lazyinit_thread();
3039 mutex_unlock(&ext4_li_mtx);
3046 * We do not need to lock anything since this is called on
3049 static void ext4_destroy_lazyinit_thread(void)
3052 * If thread exited earlier
3053 * there's nothing to be done.
3055 if (!ext4_li_info || !ext4_lazyinit_task)
3058 kthread_stop(ext4_lazyinit_task);
3061 static int set_journal_csum_feature_set(struct super_block *sb)
3064 int compat, incompat;
3065 struct ext4_sb_info *sbi = EXT4_SB(sb);
3067 if (EXT4_HAS_RO_COMPAT_FEATURE(sb,
3068 EXT4_FEATURE_RO_COMPAT_METADATA_CSUM)) {
3069 /* journal checksum v2 */
3071 incompat = JBD2_FEATURE_INCOMPAT_CSUM_V2;
3073 /* journal checksum v1 */
3074 compat = JBD2_FEATURE_COMPAT_CHECKSUM;
3078 if (test_opt(sb, JOURNAL_ASYNC_COMMIT)) {
3079 ret = jbd2_journal_set_features(sbi->s_journal,
3081 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT |
3083 } else if (test_opt(sb, JOURNAL_CHECKSUM)) {
3084 ret = jbd2_journal_set_features(sbi->s_journal,
3087 jbd2_journal_clear_features(sbi->s_journal, 0, 0,
3088 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT);
3090 jbd2_journal_clear_features(sbi->s_journal,
3091 JBD2_FEATURE_COMPAT_CHECKSUM, 0,
3092 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT |
3093 JBD2_FEATURE_INCOMPAT_CSUM_V2);
3100 * Note: calculating the overhead so we can be compatible with
3101 * historical BSD practice is quite difficult in the face of
3102 * clusters/bigalloc. This is because multiple metadata blocks from
3103 * different block group can end up in the same allocation cluster.
3104 * Calculating the exact overhead in the face of clustered allocation
3105 * requires either O(all block bitmaps) in memory or O(number of block
3106 * groups**2) in time. We will still calculate the superblock for
3107 * older file systems --- and if we come across with a bigalloc file
3108 * system with zero in s_overhead_clusters the estimate will be close to
3109 * correct especially for very large cluster sizes --- but for newer
3110 * file systems, it's better to calculate this figure once at mkfs
3111 * time, and store it in the superblock. If the superblock value is
3112 * present (even for non-bigalloc file systems), we will use it.
3114 static int count_overhead(struct super_block *sb, ext4_group_t grp,
3117 struct ext4_sb_info *sbi = EXT4_SB(sb);
3118 struct ext4_group_desc *gdp;
3119 ext4_fsblk_t first_block, last_block, b;
3120 ext4_group_t i, ngroups = ext4_get_groups_count(sb);
3121 int s, j, count = 0;
3123 if (!EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_BIGALLOC))
3124 return (ext4_bg_has_super(sb, grp) + ext4_bg_num_gdb(sb, grp) +
3125 sbi->s_itb_per_group + 2);
3127 first_block = le32_to_cpu(sbi->s_es->s_first_data_block) +
3128 (grp * EXT4_BLOCKS_PER_GROUP(sb));
3129 last_block = first_block + EXT4_BLOCKS_PER_GROUP(sb) - 1;
3130 for (i = 0; i < ngroups; i++) {
3131 gdp = ext4_get_group_desc(sb, i, NULL);
3132 b = ext4_block_bitmap(sb, gdp);
3133 if (b >= first_block && b <= last_block) {
3134 ext4_set_bit(EXT4_B2C(sbi, b - first_block), buf);
3137 b = ext4_inode_bitmap(sb, gdp);
3138 if (b >= first_block && b <= last_block) {
3139 ext4_set_bit(EXT4_B2C(sbi, b - first_block), buf);
3142 b = ext4_inode_table(sb, gdp);
3143 if (b >= first_block && b + sbi->s_itb_per_group <= last_block)
3144 for (j = 0; j < sbi->s_itb_per_group; j++, b++) {
3145 int c = EXT4_B2C(sbi, b - first_block);
3146 ext4_set_bit(c, buf);
3152 if (ext4_bg_has_super(sb, grp)) {
3153 ext4_set_bit(s++, buf);
3156 for (j = ext4_bg_num_gdb(sb, grp); j > 0; j--) {
3157 ext4_set_bit(EXT4_B2C(sbi, s++), buf);
3163 return EXT4_CLUSTERS_PER_GROUP(sb) -
3164 ext4_count_free(buf, EXT4_CLUSTERS_PER_GROUP(sb) / 8);
3168 * Compute the overhead and stash it in sbi->s_overhead
3170 int ext4_calculate_overhead(struct super_block *sb)
3172 struct ext4_sb_info *sbi = EXT4_SB(sb);
3173 struct ext4_super_block *es = sbi->s_es;
3174 ext4_group_t i, ngroups = ext4_get_groups_count(sb);
3175 ext4_fsblk_t overhead = 0;
3176 char *buf = (char *) get_zeroed_page(GFP_KERNEL);
3182 * Compute the overhead (FS structures). This is constant
3183 * for a given filesystem unless the number of block groups
3184 * changes so we cache the previous value until it does.
3188 * All of the blocks before first_data_block are overhead
3190 overhead = EXT4_B2C(sbi, le32_to_cpu(es->s_first_data_block));
3193 * Add the overhead found in each block group
3195 for (i = 0; i < ngroups; i++) {
3198 blks = count_overhead(sb, i, buf);
3201 memset(buf, 0, PAGE_SIZE);
3204 /* Add the journal blocks as well */
3206 overhead += EXT4_NUM_B2C(sbi, sbi->s_journal->j_maxlen);
3208 sbi->s_overhead = overhead;
3210 free_page((unsigned long) buf);
3215 static ext4_fsblk_t ext4_calculate_resv_clusters(struct super_block *sb)
3217 ext4_fsblk_t resv_clusters;
3220 * There's no need to reserve anything when we aren't using extents.
3221 * The space estimates are exact, there are no unwritten extents,
3222 * hole punching doesn't need new metadata... This is needed especially
3223 * to keep ext2/3 backward compatibility.
3225 if (!EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_EXTENTS))
3228 * By default we reserve 2% or 4096 clusters, whichever is smaller.
3229 * This should cover the situations where we can not afford to run
3230 * out of space like for example punch hole, or converting
3231 * uninitialized extents in delalloc path. In most cases such
3232 * allocation would require 1, or 2 blocks, higher numbers are
3235 resv_clusters = ext4_blocks_count(EXT4_SB(sb)->s_es) >>
3236 EXT4_SB(sb)->s_cluster_bits;
3238 do_div(resv_clusters, 50);
3239 resv_clusters = min_t(ext4_fsblk_t, resv_clusters, 4096);
3241 return resv_clusters;
3245 static int ext4_reserve_clusters(struct ext4_sb_info *sbi, ext4_fsblk_t count)
3247 ext4_fsblk_t clusters = ext4_blocks_count(sbi->s_es) >>
3248 sbi->s_cluster_bits;
3250 if (count >= clusters)
3253 atomic64_set(&sbi->s_resv_clusters, count);
3257 static int ext4_fill_super(struct super_block *sb, void *data, int silent)
3259 char *orig_data = kstrdup(data, GFP_KERNEL);
3260 struct buffer_head *bh;
3261 struct ext4_super_block *es = NULL;
3262 struct ext4_sb_info *sbi;
3264 ext4_fsblk_t sb_block = get_sb_block(&data);
3265 ext4_fsblk_t logical_sb_block;
3266 unsigned long offset = 0;
3267 unsigned long journal_devnum = 0;
3268 unsigned long def_mount_opts;
3273 int blocksize, clustersize;
3274 unsigned int db_count;
3276 int needs_recovery, has_huge_files, has_bigalloc;
3279 unsigned int journal_ioprio = DEFAULT_JOURNAL_IOPRIO;
3280 ext4_group_t first_not_zeroed;
3282 sbi = kzalloc(sizeof(*sbi), GFP_KERNEL);
3286 sbi->s_blockgroup_lock =
3287 kzalloc(sizeof(struct blockgroup_lock), GFP_KERNEL);
3288 if (!sbi->s_blockgroup_lock) {
3292 sb->s_fs_info = sbi;
3294 sbi->s_inode_readahead_blks = EXT4_DEF_INODE_READAHEAD_BLKS;
3295 sbi->s_sb_block = sb_block;
3296 if (sb->s_bdev->bd_part)
3297 sbi->s_sectors_written_start =
3298 part_stat_read(sb->s_bdev->bd_part, sectors[1]);
3300 /* Cleanup superblock name */
3301 for (cp = sb->s_id; (cp = strchr(cp, '/'));)
3304 /* -EINVAL is default */
3306 blocksize = sb_min_blocksize(sb, EXT4_MIN_BLOCK_SIZE);
3308 ext4_msg(sb, KERN_ERR, "unable to set blocksize");
3313 * The ext4 superblock will not be buffer aligned for other than 1kB
3314 * block sizes. We need to calculate the offset from buffer start.
3316 if (blocksize != EXT4_MIN_BLOCK_SIZE) {
3317 logical_sb_block = sb_block * EXT4_MIN_BLOCK_SIZE;
3318 offset = do_div(logical_sb_block, blocksize);
3320 logical_sb_block = sb_block;
3323 if (!(bh = sb_bread(sb, logical_sb_block))) {
3324 ext4_msg(sb, KERN_ERR, "unable to read superblock");
3328 * Note: s_es must be initialized as soon as possible because
3329 * some ext4 macro-instructions depend on its value
3331 es = (struct ext4_super_block *) (bh->b_data + offset);
3333 sb->s_magic = le16_to_cpu(es->s_magic);
3334 if (sb->s_magic != EXT4_SUPER_MAGIC)
3336 sbi->s_kbytes_written = le64_to_cpu(es->s_kbytes_written);
3338 /* Warn if metadata_csum and gdt_csum are both set. */
3339 if (EXT4_HAS_RO_COMPAT_FEATURE(sb,
3340 EXT4_FEATURE_RO_COMPAT_METADATA_CSUM) &&
3341 EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_GDT_CSUM))
3342 ext4_warning(sb, KERN_INFO "metadata_csum and uninit_bg are "
3343 "redundant flags; please run fsck.");
3345 /* Check for a known checksum algorithm */
3346 if (!ext4_verify_csum_type(sb, es)) {
3347 ext4_msg(sb, KERN_ERR, "VFS: Found ext4 filesystem with "
3348 "unknown checksum algorithm.");
3353 /* Load the checksum driver */
3354 if (EXT4_HAS_RO_COMPAT_FEATURE(sb,
3355 EXT4_FEATURE_RO_COMPAT_METADATA_CSUM)) {
3356 sbi->s_chksum_driver = crypto_alloc_shash("crc32c", 0, 0);
3357 if (IS_ERR(sbi->s_chksum_driver)) {
3358 ext4_msg(sb, KERN_ERR, "Cannot load crc32c driver.");
3359 ret = PTR_ERR(sbi->s_chksum_driver);
3360 sbi->s_chksum_driver = NULL;
3365 /* Check superblock checksum */
3366 if (!ext4_superblock_csum_verify(sb, es)) {
3367 ext4_msg(sb, KERN_ERR, "VFS: Found ext4 filesystem with "
3368 "invalid superblock checksum. Run e2fsck?");
3373 /* Precompute checksum seed for all metadata */
3374 if (EXT4_HAS_RO_COMPAT_FEATURE(sb,
3375 EXT4_FEATURE_RO_COMPAT_METADATA_CSUM))
3376 sbi->s_csum_seed = ext4_chksum(sbi, ~0, es->s_uuid,
3377 sizeof(es->s_uuid));
3379 /* Set defaults before we parse the mount options */
3380 def_mount_opts = le32_to_cpu(es->s_default_mount_opts);
3381 set_opt(sb, INIT_INODE_TABLE);
3382 if (def_mount_opts & EXT4_DEFM_DEBUG)
3384 if (def_mount_opts & EXT4_DEFM_BSDGROUPS)
3386 if (def_mount_opts & EXT4_DEFM_UID16)
3387 set_opt(sb, NO_UID32);
3388 /* xattr user namespace & acls are now defaulted on */
3389 set_opt(sb, XATTR_USER);
3390 #ifdef CONFIG_EXT4_FS_POSIX_ACL
3391 set_opt(sb, POSIX_ACL);
3393 if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_DATA)
3394 set_opt(sb, JOURNAL_DATA);
3395 else if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_ORDERED)
3396 set_opt(sb, ORDERED_DATA);
3397 else if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_WBACK)
3398 set_opt(sb, WRITEBACK_DATA);
3400 if (le16_to_cpu(sbi->s_es->s_errors) == EXT4_ERRORS_PANIC)
3401 set_opt(sb, ERRORS_PANIC);
3402 else if (le16_to_cpu(sbi->s_es->s_errors) == EXT4_ERRORS_CONTINUE)
3403 set_opt(sb, ERRORS_CONT);
3405 set_opt(sb, ERRORS_RO);
3406 if (def_mount_opts & EXT4_DEFM_BLOCK_VALIDITY)
3407 set_opt(sb, BLOCK_VALIDITY);
3408 if (def_mount_opts & EXT4_DEFM_DISCARD)
3409 set_opt(sb, DISCARD);
3411 sbi->s_resuid = make_kuid(&init_user_ns, le16_to_cpu(es->s_def_resuid));
3412 sbi->s_resgid = make_kgid(&init_user_ns, le16_to_cpu(es->s_def_resgid));
3413 sbi->s_commit_interval = JBD2_DEFAULT_MAX_COMMIT_AGE * HZ;
3414 sbi->s_min_batch_time = EXT4_DEF_MIN_BATCH_TIME;
3415 sbi->s_max_batch_time = EXT4_DEF_MAX_BATCH_TIME;
3417 if ((def_mount_opts & EXT4_DEFM_NOBARRIER) == 0)
3418 set_opt(sb, BARRIER);
3421 * enable delayed allocation by default
3422 * Use -o nodelalloc to turn it off
3424 if (!IS_EXT3_SB(sb) && !IS_EXT2_SB(sb) &&
3425 ((def_mount_opts & EXT4_DEFM_NODELALLOC) == 0))
3426 set_opt(sb, DELALLOC);
3429 * set default s_li_wait_mult for lazyinit, for the case there is
3430 * no mount option specified.
3432 sbi->s_li_wait_mult = EXT4_DEF_LI_WAIT_MULT;
3434 if (!parse_options((char *) sbi->s_es->s_mount_opts, sb,
3435 &journal_devnum, &journal_ioprio, 0)) {
3436 ext4_msg(sb, KERN_WARNING,
3437 "failed to parse options in superblock: %s",
3438 sbi->s_es->s_mount_opts);
3440 sbi->s_def_mount_opt = sbi->s_mount_opt;
3441 if (!parse_options((char *) data, sb, &journal_devnum,
3442 &journal_ioprio, 0))
3445 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA) {
3446 printk_once(KERN_WARNING "EXT4-fs: Warning: mounting "
3447 "with data=journal disables delayed "
3448 "allocation and O_DIRECT support!\n");
3449 if (test_opt2(sb, EXPLICIT_DELALLOC)) {
3450 ext4_msg(sb, KERN_ERR, "can't mount with "
3451 "both data=journal and delalloc");
3454 if (test_opt(sb, DIOREAD_NOLOCK)) {
3455 ext4_msg(sb, KERN_ERR, "can't mount with "
3456 "both data=journal and dioread_nolock");
3459 if (test_opt(sb, DELALLOC))
3460 clear_opt(sb, DELALLOC);
3463 sb->s_flags = (sb->s_flags & ~MS_POSIXACL) |
3464 (test_opt(sb, POSIX_ACL) ? MS_POSIXACL : 0);
3466 if (le32_to_cpu(es->s_rev_level) == EXT4_GOOD_OLD_REV &&
3467 (EXT4_HAS_COMPAT_FEATURE(sb, ~0U) ||
3468 EXT4_HAS_RO_COMPAT_FEATURE(sb, ~0U) ||
3469 EXT4_HAS_INCOMPAT_FEATURE(sb, ~0U)))
3470 ext4_msg(sb, KERN_WARNING,
3471 "feature flags set on rev 0 fs, "
3472 "running e2fsck is recommended");
3474 if (IS_EXT2_SB(sb)) {
3475 if (ext2_feature_set_ok(sb))
3476 ext4_msg(sb, KERN_INFO, "mounting ext2 file system "
3477 "using the ext4 subsystem");
3479 ext4_msg(sb, KERN_ERR, "couldn't mount as ext2 due "
3480 "to feature incompatibilities");
3485 if (IS_EXT3_SB(sb)) {
3486 if (ext3_feature_set_ok(sb))
3487 ext4_msg(sb, KERN_INFO, "mounting ext3 file system "
3488 "using the ext4 subsystem");
3490 ext4_msg(sb, KERN_ERR, "couldn't mount as ext3 due "
3491 "to feature incompatibilities");
3497 * Check feature flags regardless of the revision level, since we
3498 * previously didn't change the revision level when setting the flags,
3499 * so there is a chance incompat flags are set on a rev 0 filesystem.
3501 if (!ext4_feature_set_ok(sb, (sb->s_flags & MS_RDONLY)))
3504 blocksize = BLOCK_SIZE << le32_to_cpu(es->s_log_block_size);
3505 if (blocksize < EXT4_MIN_BLOCK_SIZE ||
3506 blocksize > EXT4_MAX_BLOCK_SIZE) {
3507 ext4_msg(sb, KERN_ERR,
3508 "Unsupported filesystem blocksize %d", blocksize);
3512 if (sb->s_blocksize != blocksize) {
3513 /* Validate the filesystem blocksize */
3514 if (!sb_set_blocksize(sb, blocksize)) {
3515 ext4_msg(sb, KERN_ERR, "bad block size %d",
3521 logical_sb_block = sb_block * EXT4_MIN_BLOCK_SIZE;
3522 offset = do_div(logical_sb_block, blocksize);
3523 bh = sb_bread(sb, logical_sb_block);
3525 ext4_msg(sb, KERN_ERR,
3526 "Can't read superblock on 2nd try");
3529 es = (struct ext4_super_block *)(bh->b_data + offset);
3531 if (es->s_magic != cpu_to_le16(EXT4_SUPER_MAGIC)) {
3532 ext4_msg(sb, KERN_ERR,
3533 "Magic mismatch, very weird!");
3538 has_huge_files = EXT4_HAS_RO_COMPAT_FEATURE(sb,
3539 EXT4_FEATURE_RO_COMPAT_HUGE_FILE);
3540 sbi->s_bitmap_maxbytes = ext4_max_bitmap_size(sb->s_blocksize_bits,
3542 sb->s_maxbytes = ext4_max_size(sb->s_blocksize_bits, has_huge_files);
3544 if (le32_to_cpu(es->s_rev_level) == EXT4_GOOD_OLD_REV) {
3545 sbi->s_inode_size = EXT4_GOOD_OLD_INODE_SIZE;
3546 sbi->s_first_ino = EXT4_GOOD_OLD_FIRST_INO;
3548 sbi->s_inode_size = le16_to_cpu(es->s_inode_size);
3549 sbi->s_first_ino = le32_to_cpu(es->s_first_ino);
3550 if ((sbi->s_inode_size < EXT4_GOOD_OLD_INODE_SIZE) ||
3551 (!is_power_of_2(sbi->s_inode_size)) ||
3552 (sbi->s_inode_size > blocksize)) {
3553 ext4_msg(sb, KERN_ERR,
3554 "unsupported inode size: %d",
3558 if (sbi->s_inode_size > EXT4_GOOD_OLD_INODE_SIZE)
3559 sb->s_time_gran = 1 << (EXT4_EPOCH_BITS - 2);
3562 sbi->s_desc_size = le16_to_cpu(es->s_desc_size);
3563 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_64BIT)) {
3564 if (sbi->s_desc_size < EXT4_MIN_DESC_SIZE_64BIT ||
3565 sbi->s_desc_size > EXT4_MAX_DESC_SIZE ||
3566 !is_power_of_2(sbi->s_desc_size)) {
3567 ext4_msg(sb, KERN_ERR,
3568 "unsupported descriptor size %lu",
3573 sbi->s_desc_size = EXT4_MIN_DESC_SIZE;
3575 sbi->s_blocks_per_group = le32_to_cpu(es->s_blocks_per_group);
3576 sbi->s_inodes_per_group = le32_to_cpu(es->s_inodes_per_group);
3577 if (EXT4_INODE_SIZE(sb) == 0 || EXT4_INODES_PER_GROUP(sb) == 0)
3580 sbi->s_inodes_per_block = blocksize / EXT4_INODE_SIZE(sb);
3581 if (sbi->s_inodes_per_block == 0)
3583 sbi->s_itb_per_group = sbi->s_inodes_per_group /
3584 sbi->s_inodes_per_block;
3585 sbi->s_desc_per_block = blocksize / EXT4_DESC_SIZE(sb);
3587 sbi->s_mount_state = le16_to_cpu(es->s_state);
3588 sbi->s_addr_per_block_bits = ilog2(EXT4_ADDR_PER_BLOCK(sb));
3589 sbi->s_desc_per_block_bits = ilog2(EXT4_DESC_PER_BLOCK(sb));
3591 for (i = 0; i < 4; i++)
3592 sbi->s_hash_seed[i] = le32_to_cpu(es->s_hash_seed[i]);
3593 sbi->s_def_hash_version = es->s_def_hash_version;
3594 if (EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_DIR_INDEX)) {
3595 i = le32_to_cpu(es->s_flags);
3596 if (i & EXT2_FLAGS_UNSIGNED_HASH)
3597 sbi->s_hash_unsigned = 3;
3598 else if ((i & EXT2_FLAGS_SIGNED_HASH) == 0) {
3599 #ifdef __CHAR_UNSIGNED__
3600 if (!(sb->s_flags & MS_RDONLY))
3602 cpu_to_le32(EXT2_FLAGS_UNSIGNED_HASH);
3603 sbi->s_hash_unsigned = 3;
3605 if (!(sb->s_flags & MS_RDONLY))
3607 cpu_to_le32(EXT2_FLAGS_SIGNED_HASH);
3612 /* Handle clustersize */
3613 clustersize = BLOCK_SIZE << le32_to_cpu(es->s_log_cluster_size);
3614 has_bigalloc = EXT4_HAS_RO_COMPAT_FEATURE(sb,
3615 EXT4_FEATURE_RO_COMPAT_BIGALLOC);
3617 if (clustersize < blocksize) {
3618 ext4_msg(sb, KERN_ERR,
3619 "cluster size (%d) smaller than "
3620 "block size (%d)", clustersize, blocksize);
3623 sbi->s_cluster_bits = le32_to_cpu(es->s_log_cluster_size) -
3624 le32_to_cpu(es->s_log_block_size);
3625 sbi->s_clusters_per_group =
3626 le32_to_cpu(es->s_clusters_per_group);
3627 if (sbi->s_clusters_per_group > blocksize * 8) {
3628 ext4_msg(sb, KERN_ERR,
3629 "#clusters per group too big: %lu",
3630 sbi->s_clusters_per_group);
3633 if (sbi->s_blocks_per_group !=
3634 (sbi->s_clusters_per_group * (clustersize / blocksize))) {
3635 ext4_msg(sb, KERN_ERR, "blocks per group (%lu) and "
3636 "clusters per group (%lu) inconsistent",
3637 sbi->s_blocks_per_group,
3638 sbi->s_clusters_per_group);
3642 if (clustersize != blocksize) {
3643 ext4_warning(sb, "fragment/cluster size (%d) != "
3644 "block size (%d)", clustersize,
3646 clustersize = blocksize;
3648 if (sbi->s_blocks_per_group > blocksize * 8) {
3649 ext4_msg(sb, KERN_ERR,
3650 "#blocks per group too big: %lu",
3651 sbi->s_blocks_per_group);
3654 sbi->s_clusters_per_group = sbi->s_blocks_per_group;
3655 sbi->s_cluster_bits = 0;
3657 sbi->s_cluster_ratio = clustersize / blocksize;
3659 if (sbi->s_inodes_per_group > blocksize * 8) {
3660 ext4_msg(sb, KERN_ERR,
3661 "#inodes per group too big: %lu",
3662 sbi->s_inodes_per_group);
3666 /* Do we have standard group size of clustersize * 8 blocks ? */
3667 if (sbi->s_blocks_per_group == clustersize << 3)
3668 set_opt2(sb, STD_GROUP_SIZE);
3671 * Test whether we have more sectors than will fit in sector_t,
3672 * and whether the max offset is addressable by the page cache.
3674 err = generic_check_addressable(sb->s_blocksize_bits,
3675 ext4_blocks_count(es));
3677 ext4_msg(sb, KERN_ERR, "filesystem"
3678 " too large to mount safely on this system");
3679 if (sizeof(sector_t) < 8)
3680 ext4_msg(sb, KERN_WARNING, "CONFIG_LBDAF not enabled");
3684 if (EXT4_BLOCKS_PER_GROUP(sb) == 0)
3687 /* check blocks count against device size */
3688 blocks_count = sb->s_bdev->bd_inode->i_size >> sb->s_blocksize_bits;
3689 if (blocks_count && ext4_blocks_count(es) > blocks_count) {
3690 ext4_msg(sb, KERN_WARNING, "bad geometry: block count %llu "
3691 "exceeds size of device (%llu blocks)",
3692 ext4_blocks_count(es), blocks_count);
3697 * It makes no sense for the first data block to be beyond the end
3698 * of the filesystem.
3700 if (le32_to_cpu(es->s_first_data_block) >= ext4_blocks_count(es)) {
3701 ext4_msg(sb, KERN_WARNING, "bad geometry: first data "
3702 "block %u is beyond end of filesystem (%llu)",
3703 le32_to_cpu(es->s_first_data_block),
3704 ext4_blocks_count(es));
3707 blocks_count = (ext4_blocks_count(es) -
3708 le32_to_cpu(es->s_first_data_block) +
3709 EXT4_BLOCKS_PER_GROUP(sb) - 1);
3710 do_div(blocks_count, EXT4_BLOCKS_PER_GROUP(sb));
3711 if (blocks_count > ((uint64_t)1<<32) - EXT4_DESC_PER_BLOCK(sb)) {
3712 ext4_msg(sb, KERN_WARNING, "groups count too large: %u "
3713 "(block count %llu, first data block %u, "
3714 "blocks per group %lu)", sbi->s_groups_count,
3715 ext4_blocks_count(es),
3716 le32_to_cpu(es->s_first_data_block),
3717 EXT4_BLOCKS_PER_GROUP(sb));
3720 sbi->s_groups_count = blocks_count;
3721 sbi->s_blockfile_groups = min_t(ext4_group_t, sbi->s_groups_count,
3722 (EXT4_MAX_BLOCK_FILE_PHYS / EXT4_BLOCKS_PER_GROUP(sb)));
3723 db_count = (sbi->s_groups_count + EXT4_DESC_PER_BLOCK(sb) - 1) /
3724 EXT4_DESC_PER_BLOCK(sb);
3725 sbi->s_group_desc = ext4_kvmalloc(db_count *
3726 sizeof(struct buffer_head *),
3728 if (sbi->s_group_desc == NULL) {
3729 ext4_msg(sb, KERN_ERR, "not enough memory");
3735 sbi->s_proc = proc_mkdir(sb->s_id, ext4_proc_root);
3738 proc_create_data("options", S_IRUGO, sbi->s_proc,
3739 &ext4_seq_options_fops, sb);
3741 bgl_lock_init(sbi->s_blockgroup_lock);
3743 for (i = 0; i < db_count; i++) {
3744 block = descriptor_loc(sb, logical_sb_block, i);
3745 sbi->s_group_desc[i] = sb_bread(sb, block);
3746 if (!sbi->s_group_desc[i]) {
3747 ext4_msg(sb, KERN_ERR,
3748 "can't read group descriptor %d", i);
3753 if (!ext4_check_descriptors(sb, &first_not_zeroed)) {
3754 ext4_msg(sb, KERN_ERR, "group descriptors corrupted!");
3757 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_FLEX_BG))
3758 if (!ext4_fill_flex_info(sb)) {
3759 ext4_msg(sb, KERN_ERR,
3760 "unable to initialize "
3761 "flex_bg meta info!");
3765 sbi->s_gdb_count = db_count;
3766 get_random_bytes(&sbi->s_next_generation, sizeof(u32));
3767 spin_lock_init(&sbi->s_next_gen_lock);
3769 init_timer(&sbi->s_err_report);
3770 sbi->s_err_report.function = print_daily_error_info;
3771 sbi->s_err_report.data = (unsigned long) sb;
3773 /* Register extent status tree shrinker */
3774 ext4_es_register_shrinker(sb);
3776 err = percpu_counter_init(&sbi->s_freeclusters_counter,
3777 ext4_count_free_clusters(sb));
3779 err = percpu_counter_init(&sbi->s_freeinodes_counter,
3780 ext4_count_free_inodes(sb));
3783 err = percpu_counter_init(&sbi->s_dirs_counter,
3784 ext4_count_dirs(sb));
3787 err = percpu_counter_init(&sbi->s_dirtyclusters_counter, 0);
3790 err = percpu_counter_init(&sbi->s_extent_cache_cnt, 0);
3793 ext4_msg(sb, KERN_ERR, "insufficient memory");
3797 sbi->s_stripe = ext4_get_stripe_size(sbi);
3798 sbi->s_max_writeback_mb_bump = 128;
3799 sbi->s_extent_max_zeroout_kb = 32;
3802 * set up enough so that it can read an inode
3804 if (!test_opt(sb, NOLOAD) &&
3805 EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL))
3806 sb->s_op = &ext4_sops;
3808 sb->s_op = &ext4_nojournal_sops;
3809 sb->s_export_op = &ext4_export_ops;
3810 sb->s_xattr = ext4_xattr_handlers;
3812 sb->dq_op = &ext4_quota_operations;
3813 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_QUOTA))
3814 sb->s_qcop = &ext4_qctl_sysfile_operations;
3816 sb->s_qcop = &ext4_qctl_operations;
3818 memcpy(sb->s_uuid, es->s_uuid, sizeof(es->s_uuid));
3820 INIT_LIST_HEAD(&sbi->s_orphan); /* unlinked but open files */
3821 mutex_init(&sbi->s_orphan_lock);
3825 needs_recovery = (es->s_last_orphan != 0 ||
3826 EXT4_HAS_INCOMPAT_FEATURE(sb,
3827 EXT4_FEATURE_INCOMPAT_RECOVER));
3829 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_MMP) &&
3830 !(sb->s_flags & MS_RDONLY))
3831 if (ext4_multi_mount_protect(sb, le64_to_cpu(es->s_mmp_block)))
3835 * The first inode we look at is the journal inode. Don't try
3836 * root first: it may be modified in the journal!
3838 if (!test_opt(sb, NOLOAD) &&
3839 EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL)) {
3840 if (ext4_load_journal(sb, es, journal_devnum))
3842 } else if (test_opt(sb, NOLOAD) && !(sb->s_flags & MS_RDONLY) &&
3843 EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER)) {
3844 ext4_msg(sb, KERN_ERR, "required journal recovery "
3845 "suppressed and not mounted read-only");
3846 goto failed_mount_wq;
3848 clear_opt(sb, DATA_FLAGS);
3849 sbi->s_journal = NULL;
3854 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_64BIT) &&
3855 !jbd2_journal_set_features(EXT4_SB(sb)->s_journal, 0, 0,
3856 JBD2_FEATURE_INCOMPAT_64BIT)) {
3857 ext4_msg(sb, KERN_ERR, "Failed to set 64-bit journal feature");
3858 goto failed_mount_wq;
3861 if (!set_journal_csum_feature_set(sb)) {
3862 ext4_msg(sb, KERN_ERR, "Failed to set journal checksum "
3864 goto failed_mount_wq;
3867 /* We have now updated the journal if required, so we can
3868 * validate the data journaling mode. */
3869 switch (test_opt(sb, DATA_FLAGS)) {
3871 /* No mode set, assume a default based on the journal
3872 * capabilities: ORDERED_DATA if the journal can
3873 * cope, else JOURNAL_DATA
3875 if (jbd2_journal_check_available_features
3876 (sbi->s_journal, 0, 0, JBD2_FEATURE_INCOMPAT_REVOKE))
3877 set_opt(sb, ORDERED_DATA);
3879 set_opt(sb, JOURNAL_DATA);
3882 case EXT4_MOUNT_ORDERED_DATA:
3883 case EXT4_MOUNT_WRITEBACK_DATA:
3884 if (!jbd2_journal_check_available_features
3885 (sbi->s_journal, 0, 0, JBD2_FEATURE_INCOMPAT_REVOKE)) {
3886 ext4_msg(sb, KERN_ERR, "Journal does not support "
3887 "requested data journaling mode");
3888 goto failed_mount_wq;
3893 set_task_ioprio(sbi->s_journal->j_task, journal_ioprio);
3895 sbi->s_journal->j_commit_callback = ext4_journal_commit_callback;
3898 * The journal may have updated the bg summary counts, so we
3899 * need to update the global counters.
3901 percpu_counter_set(&sbi->s_freeclusters_counter,
3902 ext4_count_free_clusters(sb));
3903 percpu_counter_set(&sbi->s_freeinodes_counter,
3904 ext4_count_free_inodes(sb));
3905 percpu_counter_set(&sbi->s_dirs_counter,
3906 ext4_count_dirs(sb));
3907 percpu_counter_set(&sbi->s_dirtyclusters_counter, 0);
3911 * Get the # of file system overhead blocks from the
3912 * superblock if present.
3914 if (es->s_overhead_clusters)
3915 sbi->s_overhead = le32_to_cpu(es->s_overhead_clusters);
3917 err = ext4_calculate_overhead(sb);
3919 goto failed_mount_wq;
3923 * The maximum number of concurrent works can be high and
3924 * concurrency isn't really necessary. Limit it to 1.
3926 EXT4_SB(sb)->dio_unwritten_wq =
3927 alloc_workqueue("ext4-dio-unwritten", WQ_MEM_RECLAIM | WQ_UNBOUND, 1);
3928 if (!EXT4_SB(sb)->dio_unwritten_wq) {
3929 printk(KERN_ERR "EXT4-fs: failed to create DIO workqueue\n");
3931 goto failed_mount_wq;
3935 * The jbd2_journal_load will have done any necessary log recovery,
3936 * so we can safely mount the rest of the filesystem now.
3939 root = ext4_iget(sb, EXT4_ROOT_INO);
3941 ext4_msg(sb, KERN_ERR, "get root inode failed");
3942 ret = PTR_ERR(root);
3946 if (!S_ISDIR(root->i_mode) || !root->i_blocks || !root->i_size) {
3947 ext4_msg(sb, KERN_ERR, "corrupt root inode, run e2fsck");
3951 sb->s_root = d_make_root(root);
3953 ext4_msg(sb, KERN_ERR, "get root dentry failed");
3958 if (ext4_setup_super(sb, es, sb->s_flags & MS_RDONLY))
3959 sb->s_flags |= MS_RDONLY;
3961 /* determine the minimum size of new large inodes, if present */
3962 if (sbi->s_inode_size > EXT4_GOOD_OLD_INODE_SIZE) {
3963 sbi->s_want_extra_isize = sizeof(struct ext4_inode) -
3964 EXT4_GOOD_OLD_INODE_SIZE;
3965 if (EXT4_HAS_RO_COMPAT_FEATURE(sb,
3966 EXT4_FEATURE_RO_COMPAT_EXTRA_ISIZE)) {
3967 if (sbi->s_want_extra_isize <
3968 le16_to_cpu(es->s_want_extra_isize))
3969 sbi->s_want_extra_isize =
3970 le16_to_cpu(es->s_want_extra_isize);
3971 if (sbi->s_want_extra_isize <
3972 le16_to_cpu(es->s_min_extra_isize))
3973 sbi->s_want_extra_isize =
3974 le16_to_cpu(es->s_min_extra_isize);
3977 /* Check if enough inode space is available */
3978 if (EXT4_GOOD_OLD_INODE_SIZE + sbi->s_want_extra_isize >
3979 sbi->s_inode_size) {
3980 sbi->s_want_extra_isize = sizeof(struct ext4_inode) -
3981 EXT4_GOOD_OLD_INODE_SIZE;
3982 ext4_msg(sb, KERN_INFO, "required extra inode space not"
3986 err = ext4_reserve_clusters(sbi, ext4_calculate_resv_clusters(sb));
3988 ext4_msg(sb, KERN_ERR, "failed to reserve %llu clusters for "
3989 "reserved pool", ext4_calculate_resv_clusters(sb));
3990 goto failed_mount4a;
3993 err = ext4_setup_system_zone(sb);
3995 ext4_msg(sb, KERN_ERR, "failed to initialize system "
3997 goto failed_mount4a;
4001 err = ext4_mb_init(sb);
4003 ext4_msg(sb, KERN_ERR, "failed to initialize mballoc (%d)",
4008 err = ext4_register_li_request(sb, first_not_zeroed);
4012 sbi->s_kobj.kset = ext4_kset;
4013 init_completion(&sbi->s_kobj_unregister);
4014 err = kobject_init_and_add(&sbi->s_kobj, &ext4_ktype, NULL,
4020 /* Enable quota usage during mount. */
4021 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_QUOTA) &&
4022 !(sb->s_flags & MS_RDONLY)) {
4023 err = ext4_enable_quotas(sb);
4027 #endif /* CONFIG_QUOTA */
4029 EXT4_SB(sb)->s_mount_state |= EXT4_ORPHAN_FS;
4030 ext4_orphan_cleanup(sb, es);
4031 EXT4_SB(sb)->s_mount_state &= ~EXT4_ORPHAN_FS;
4032 if (needs_recovery) {
4033 ext4_msg(sb, KERN_INFO, "recovery complete");
4034 ext4_mark_recovery_complete(sb, es);
4036 if (EXT4_SB(sb)->s_journal) {
4037 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA)
4038 descr = " journalled data mode";
4039 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA)
4040 descr = " ordered data mode";
4042 descr = " writeback data mode";
4044 descr = "out journal";
4046 if (test_opt(sb, DISCARD)) {
4047 struct request_queue *q = bdev_get_queue(sb->s_bdev);
4048 if (!blk_queue_discard(q))
4049 ext4_msg(sb, KERN_WARNING,
4050 "mounting with \"discard\" option, but "
4051 "the device does not support discard");
4054 ext4_msg(sb, KERN_INFO, "mounted filesystem with%s. "
4055 "Opts: %s%s%s", descr, sbi->s_es->s_mount_opts,
4056 *sbi->s_es->s_mount_opts ? "; " : "", orig_data);
4058 if (es->s_error_count)
4059 mod_timer(&sbi->s_err_report, jiffies + 300*HZ); /* 5 minutes */
4066 ext4_msg(sb, KERN_ERR, "VFS: Can't find ext4 filesystem");
4071 kobject_del(&sbi->s_kobj);
4074 ext4_unregister_li_request(sb);
4076 ext4_mb_release(sb);
4078 ext4_ext_release(sb);
4079 ext4_release_system_zone(sb);
4084 ext4_msg(sb, KERN_ERR, "mount failed");
4085 destroy_workqueue(EXT4_SB(sb)->dio_unwritten_wq);
4087 if (sbi->s_journal) {
4088 jbd2_journal_destroy(sbi->s_journal);
4089 sbi->s_journal = NULL;
4092 ext4_es_unregister_shrinker(sb);
4093 del_timer(&sbi->s_err_report);
4094 if (sbi->s_flex_groups)
4095 ext4_kvfree(sbi->s_flex_groups);
4096 percpu_counter_destroy(&sbi->s_freeclusters_counter);
4097 percpu_counter_destroy(&sbi->s_freeinodes_counter);
4098 percpu_counter_destroy(&sbi->s_dirs_counter);
4099 percpu_counter_destroy(&sbi->s_dirtyclusters_counter);
4100 percpu_counter_destroy(&sbi->s_extent_cache_cnt);
4102 kthread_stop(sbi->s_mmp_tsk);
4104 for (i = 0; i < db_count; i++)
4105 brelse(sbi->s_group_desc[i]);
4106 ext4_kvfree(sbi->s_group_desc);
4108 if (sbi->s_chksum_driver)
4109 crypto_free_shash(sbi->s_chksum_driver);
4111 remove_proc_entry("options", sbi->s_proc);
4112 remove_proc_entry(sb->s_id, ext4_proc_root);
4115 for (i = 0; i < MAXQUOTAS; i++)
4116 kfree(sbi->s_qf_names[i]);
4118 ext4_blkdev_remove(sbi);
4121 sb->s_fs_info = NULL;
4122 kfree(sbi->s_blockgroup_lock);
4126 return err ? err : ret;
4130 * Setup any per-fs journal parameters now. We'll do this both on
4131 * initial mount, once the journal has been initialised but before we've
4132 * done any recovery; and again on any subsequent remount.
4134 static void ext4_init_journal_params(struct super_block *sb, journal_t *journal)
4136 struct ext4_sb_info *sbi = EXT4_SB(sb);
4138 journal->j_commit_interval = sbi->s_commit_interval;
4139 journal->j_min_batch_time = sbi->s_min_batch_time;
4140 journal->j_max_batch_time = sbi->s_max_batch_time;
4142 write_lock(&journal->j_state_lock);
4143 if (test_opt(sb, BARRIER))
4144 journal->j_flags |= JBD2_BARRIER;
4146 journal->j_flags &= ~JBD2_BARRIER;
4147 if (test_opt(sb, DATA_ERR_ABORT))
4148 journal->j_flags |= JBD2_ABORT_ON_SYNCDATA_ERR;
4150 journal->j_flags &= ~JBD2_ABORT_ON_SYNCDATA_ERR;
4151 write_unlock(&journal->j_state_lock);
4154 static journal_t *ext4_get_journal(struct super_block *sb,
4155 unsigned int journal_inum)
4157 struct inode *journal_inode;
4160 BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL));
4162 /* First, test for the existence of a valid inode on disk. Bad
4163 * things happen if we iget() an unused inode, as the subsequent
4164 * iput() will try to delete it. */
4166 journal_inode = ext4_iget(sb, journal_inum);
4167 if (IS_ERR(journal_inode)) {
4168 ext4_msg(sb, KERN_ERR, "no journal found");
4171 if (!journal_inode->i_nlink) {
4172 make_bad_inode(journal_inode);
4173 iput(journal_inode);
4174 ext4_msg(sb, KERN_ERR, "journal inode is deleted");
4178 jbd_debug(2, "Journal inode found at %p: %lld bytes\n",
4179 journal_inode, journal_inode->i_size);
4180 if (!S_ISREG(journal_inode->i_mode)) {
4181 ext4_msg(sb, KERN_ERR, "invalid journal inode");
4182 iput(journal_inode);
4186 journal = jbd2_journal_init_inode(journal_inode);
4188 ext4_msg(sb, KERN_ERR, "Could not load journal inode");
4189 iput(journal_inode);
4192 journal->j_private = sb;
4193 ext4_init_journal_params(sb, journal);
4197 static journal_t *ext4_get_dev_journal(struct super_block *sb,
4200 struct buffer_head *bh;
4204 int hblock, blocksize;
4205 ext4_fsblk_t sb_block;
4206 unsigned long offset;
4207 struct ext4_super_block *es;
4208 struct block_device *bdev;
4210 BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL));
4212 bdev = ext4_blkdev_get(j_dev, sb);
4216 blocksize = sb->s_blocksize;
4217 hblock = bdev_logical_block_size(bdev);
4218 if (blocksize < hblock) {
4219 ext4_msg(sb, KERN_ERR,
4220 "blocksize too small for journal device");
4224 sb_block = EXT4_MIN_BLOCK_SIZE / blocksize;
4225 offset = EXT4_MIN_BLOCK_SIZE % blocksize;
4226 set_blocksize(bdev, blocksize);
4227 if (!(bh = __bread(bdev, sb_block, blocksize))) {
4228 ext4_msg(sb, KERN_ERR, "couldn't read superblock of "
4229 "external journal");
4233 es = (struct ext4_super_block *) (bh->b_data + offset);
4234 if ((le16_to_cpu(es->s_magic) != EXT4_SUPER_MAGIC) ||
4235 !(le32_to_cpu(es->s_feature_incompat) &
4236 EXT4_FEATURE_INCOMPAT_JOURNAL_DEV)) {
4237 ext4_msg(sb, KERN_ERR, "external journal has "
4243 if (memcmp(EXT4_SB(sb)->s_es->s_journal_uuid, es->s_uuid, 16)) {
4244 ext4_msg(sb, KERN_ERR, "journal UUID does not match");
4249 len = ext4_blocks_count(es);
4250 start = sb_block + 1;
4251 brelse(bh); /* we're done with the superblock */
4253 journal = jbd2_journal_init_dev(bdev, sb->s_bdev,
4254 start, len, blocksize);
4256 ext4_msg(sb, KERN_ERR, "failed to create device journal");
4259 journal->j_private = sb;
4260 ll_rw_block(READ | REQ_META | REQ_PRIO, 1, &journal->j_sb_buffer);
4261 wait_on_buffer(journal->j_sb_buffer);
4262 if (!buffer_uptodate(journal->j_sb_buffer)) {
4263 ext4_msg(sb, KERN_ERR, "I/O error on journal device");
4266 if (be32_to_cpu(journal->j_superblock->s_nr_users) != 1) {
4267 ext4_msg(sb, KERN_ERR, "External journal has more than one "
4268 "user (unsupported) - %d",
4269 be32_to_cpu(journal->j_superblock->s_nr_users));
4272 EXT4_SB(sb)->journal_bdev = bdev;
4273 ext4_init_journal_params(sb, journal);
4277 jbd2_journal_destroy(journal);
4279 ext4_blkdev_put(bdev);
4283 static int ext4_load_journal(struct super_block *sb,
4284 struct ext4_super_block *es,
4285 unsigned long journal_devnum)
4288 unsigned int journal_inum = le32_to_cpu(es->s_journal_inum);
4291 int really_read_only;
4293 BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL));
4295 if (journal_devnum &&
4296 journal_devnum != le32_to_cpu(es->s_journal_dev)) {
4297 ext4_msg(sb, KERN_INFO, "external journal device major/minor "
4298 "numbers have changed");
4299 journal_dev = new_decode_dev(journal_devnum);
4301 journal_dev = new_decode_dev(le32_to_cpu(es->s_journal_dev));
4303 really_read_only = bdev_read_only(sb->s_bdev);
4306 * Are we loading a blank journal or performing recovery after a
4307 * crash? For recovery, we need to check in advance whether we
4308 * can get read-write access to the device.
4310 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER)) {
4311 if (sb->s_flags & MS_RDONLY) {
4312 ext4_msg(sb, KERN_INFO, "INFO: recovery "
4313 "required on readonly filesystem");
4314 if (really_read_only) {
4315 ext4_msg(sb, KERN_ERR, "write access "
4316 "unavailable, cannot proceed");
4319 ext4_msg(sb, KERN_INFO, "write access will "
4320 "be enabled during recovery");
4324 if (journal_inum && journal_dev) {
4325 ext4_msg(sb, KERN_ERR, "filesystem has both journal "
4326 "and inode journals!");
4331 if (!(journal = ext4_get_journal(sb, journal_inum)))
4334 if (!(journal = ext4_get_dev_journal(sb, journal_dev)))
4338 if (!(journal->j_flags & JBD2_BARRIER))
4339 ext4_msg(sb, KERN_INFO, "barriers disabled");
4341 if (!EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER))
4342 err = jbd2_journal_wipe(journal, !really_read_only);
4344 char *save = kmalloc(EXT4_S_ERR_LEN, GFP_KERNEL);
4346 memcpy(save, ((char *) es) +
4347 EXT4_S_ERR_START, EXT4_S_ERR_LEN);
4348 err = jbd2_journal_load(journal);
4350 memcpy(((char *) es) + EXT4_S_ERR_START,
4351 save, EXT4_S_ERR_LEN);
4356 ext4_msg(sb, KERN_ERR, "error loading journal");
4357 jbd2_journal_destroy(journal);
4361 EXT4_SB(sb)->s_journal = journal;
4362 ext4_clear_journal_err(sb, es);
4364 if (!really_read_only && journal_devnum &&
4365 journal_devnum != le32_to_cpu(es->s_journal_dev)) {
4366 es->s_journal_dev = cpu_to_le32(journal_devnum);
4368 /* Make sure we flush the recovery flag to disk. */
4369 ext4_commit_super(sb, 1);
4375 static int ext4_commit_super(struct super_block *sb, int sync)
4377 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
4378 struct buffer_head *sbh = EXT4_SB(sb)->s_sbh;
4381 if (!sbh || block_device_ejected(sb))
4383 if (buffer_write_io_error(sbh)) {
4385 * Oh, dear. A previous attempt to write the
4386 * superblock failed. This could happen because the
4387 * USB device was yanked out. Or it could happen to
4388 * be a transient write error and maybe the block will
4389 * be remapped. Nothing we can do but to retry the
4390 * write and hope for the best.
4392 ext4_msg(sb, KERN_ERR, "previous I/O error to "
4393 "superblock detected");
4394 clear_buffer_write_io_error(sbh);
4395 set_buffer_uptodate(sbh);
4398 * If the file system is mounted read-only, don't update the
4399 * superblock write time. This avoids updating the superblock
4400 * write time when we are mounting the root file system
4401 * read/only but we need to replay the journal; at that point,
4402 * for people who are east of GMT and who make their clock
4403 * tick in localtime for Windows bug-for-bug compatibility,
4404 * the clock is set in the future, and this will cause e2fsck
4405 * to complain and force a full file system check.
4407 if (!(sb->s_flags & MS_RDONLY))
4408 es->s_wtime = cpu_to_le32(get_seconds());
4409 if (sb->s_bdev->bd_part)
4410 es->s_kbytes_written =
4411 cpu_to_le64(EXT4_SB(sb)->s_kbytes_written +
4412 ((part_stat_read(sb->s_bdev->bd_part, sectors[1]) -
4413 EXT4_SB(sb)->s_sectors_written_start) >> 1));
4415 es->s_kbytes_written =
4416 cpu_to_le64(EXT4_SB(sb)->s_kbytes_written);
4417 ext4_free_blocks_count_set(es,
4418 EXT4_C2B(EXT4_SB(sb), percpu_counter_sum_positive(
4419 &EXT4_SB(sb)->s_freeclusters_counter)));
4420 es->s_free_inodes_count =
4421 cpu_to_le32(percpu_counter_sum_positive(
4422 &EXT4_SB(sb)->s_freeinodes_counter));
4423 BUFFER_TRACE(sbh, "marking dirty");
4424 ext4_superblock_csum_set(sb);
4425 mark_buffer_dirty(sbh);
4427 error = sync_dirty_buffer(sbh);
4431 error = buffer_write_io_error(sbh);
4433 ext4_msg(sb, KERN_ERR, "I/O error while writing "
4435 clear_buffer_write_io_error(sbh);
4436 set_buffer_uptodate(sbh);
4443 * Have we just finished recovery? If so, and if we are mounting (or
4444 * remounting) the filesystem readonly, then we will end up with a
4445 * consistent fs on disk. Record that fact.
4447 static void ext4_mark_recovery_complete(struct super_block *sb,
4448 struct ext4_super_block *es)
4450 journal_t *journal = EXT4_SB(sb)->s_journal;
4452 if (!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL)) {
4453 BUG_ON(journal != NULL);
4456 jbd2_journal_lock_updates(journal);
4457 if (jbd2_journal_flush(journal) < 0)
4460 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER) &&
4461 sb->s_flags & MS_RDONLY) {
4462 EXT4_CLEAR_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
4463 ext4_commit_super(sb, 1);
4467 jbd2_journal_unlock_updates(journal);
4471 * If we are mounting (or read-write remounting) a filesystem whose journal
4472 * has recorded an error from a previous lifetime, move that error to the
4473 * main filesystem now.
4475 static void ext4_clear_journal_err(struct super_block *sb,
4476 struct ext4_super_block *es)
4482 BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL));
4484 journal = EXT4_SB(sb)->s_journal;
4487 * Now check for any error status which may have been recorded in the
4488 * journal by a prior ext4_error() or ext4_abort()
4491 j_errno = jbd2_journal_errno(journal);
4495 errstr = ext4_decode_error(sb, j_errno, nbuf);
4496 ext4_warning(sb, "Filesystem error recorded "
4497 "from previous mount: %s", errstr);
4498 ext4_warning(sb, "Marking fs in need of filesystem check.");
4500 EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS;
4501 es->s_state |= cpu_to_le16(EXT4_ERROR_FS);
4502 ext4_commit_super(sb, 1);
4504 jbd2_journal_clear_err(journal);
4505 jbd2_journal_update_sb_errno(journal);
4510 * Force the running and committing transactions to commit,
4511 * and wait on the commit.
4513 int ext4_force_commit(struct super_block *sb)
4517 if (sb->s_flags & MS_RDONLY)
4520 journal = EXT4_SB(sb)->s_journal;
4521 return ext4_journal_force_commit(journal);
4524 static int ext4_sync_fs(struct super_block *sb, int wait)
4528 struct ext4_sb_info *sbi = EXT4_SB(sb);
4530 trace_ext4_sync_fs(sb, wait);
4531 flush_workqueue(sbi->dio_unwritten_wq);
4533 * Writeback quota in non-journalled quota case - journalled quota has
4536 dquot_writeback_dquots(sb, -1);
4537 if (jbd2_journal_start_commit(sbi->s_journal, &target)) {
4539 jbd2_log_wait_commit(sbi->s_journal, target);
4545 * LVM calls this function before a (read-only) snapshot is created. This
4546 * gives us a chance to flush the journal completely and mark the fs clean.
4548 * Note that only this function cannot bring a filesystem to be in a clean
4549 * state independently. It relies on upper layer to stop all data & metadata
4552 static int ext4_freeze(struct super_block *sb)
4557 if (sb->s_flags & MS_RDONLY)
4560 journal = EXT4_SB(sb)->s_journal;
4562 /* Now we set up the journal barrier. */
4563 jbd2_journal_lock_updates(journal);
4566 * Don't clear the needs_recovery flag if we failed to flush
4569 error = jbd2_journal_flush(journal);
4573 /* Journal blocked and flushed, clear needs_recovery flag. */
4574 EXT4_CLEAR_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
4575 error = ext4_commit_super(sb, 1);
4577 /* we rely on upper layer to stop further updates */
4578 jbd2_journal_unlock_updates(EXT4_SB(sb)->s_journal);
4583 * Called by LVM after the snapshot is done. We need to reset the RECOVER
4584 * flag here, even though the filesystem is not technically dirty yet.
4586 static int ext4_unfreeze(struct super_block *sb)
4588 if (sb->s_flags & MS_RDONLY)
4591 /* Reset the needs_recovery flag before the fs is unlocked. */
4592 EXT4_SET_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
4593 ext4_commit_super(sb, 1);
4598 * Structure to save mount options for ext4_remount's benefit
4600 struct ext4_mount_options {
4601 unsigned long s_mount_opt;
4602 unsigned long s_mount_opt2;
4605 unsigned long s_commit_interval;
4606 u32 s_min_batch_time, s_max_batch_time;
4609 char *s_qf_names[MAXQUOTAS];
4613 static int ext4_remount(struct super_block *sb, int *flags, char *data)
4615 struct ext4_super_block *es;
4616 struct ext4_sb_info *sbi = EXT4_SB(sb);
4617 unsigned long old_sb_flags;
4618 struct ext4_mount_options old_opts;
4619 int enable_quota = 0;
4621 unsigned int journal_ioprio = DEFAULT_JOURNAL_IOPRIO;
4626 char *orig_data = kstrdup(data, GFP_KERNEL);
4628 sync_filesystem(sb);
4630 /* Store the original options */
4631 old_sb_flags = sb->s_flags;
4632 old_opts.s_mount_opt = sbi->s_mount_opt;
4633 old_opts.s_mount_opt2 = sbi->s_mount_opt2;
4634 old_opts.s_resuid = sbi->s_resuid;
4635 old_opts.s_resgid = sbi->s_resgid;
4636 old_opts.s_commit_interval = sbi->s_commit_interval;
4637 old_opts.s_min_batch_time = sbi->s_min_batch_time;
4638 old_opts.s_max_batch_time = sbi->s_max_batch_time;
4640 old_opts.s_jquota_fmt = sbi->s_jquota_fmt;
4641 for (i = 0; i < MAXQUOTAS; i++)
4642 if (sbi->s_qf_names[i]) {
4643 old_opts.s_qf_names[i] = kstrdup(sbi->s_qf_names[i],
4645 if (!old_opts.s_qf_names[i]) {
4646 for (j = 0; j < i; j++)
4647 kfree(old_opts.s_qf_names[j]);
4652 old_opts.s_qf_names[i] = NULL;
4654 if (sbi->s_journal && sbi->s_journal->j_task->io_context)
4655 journal_ioprio = sbi->s_journal->j_task->io_context->ioprio;
4658 * Allow the "check" option to be passed as a remount option.
4660 if (!parse_options(data, sb, NULL, &journal_ioprio, 1)) {
4665 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA) {
4666 if (test_opt2(sb, EXPLICIT_DELALLOC)) {
4667 ext4_msg(sb, KERN_ERR, "can't mount with "
4668 "both data=journal and delalloc");
4672 if (test_opt(sb, DIOREAD_NOLOCK)) {
4673 ext4_msg(sb, KERN_ERR, "can't mount with "
4674 "both data=journal and dioread_nolock");
4680 if (sbi->s_mount_flags & EXT4_MF_FS_ABORTED)
4681 ext4_abort(sb, "Abort forced by user");
4683 sb->s_flags = (sb->s_flags & ~MS_POSIXACL) |
4684 (test_opt(sb, POSIX_ACL) ? MS_POSIXACL : 0);
4688 if (sbi->s_journal) {
4689 ext4_init_journal_params(sb, sbi->s_journal);
4690 set_task_ioprio(sbi->s_journal->j_task, journal_ioprio);
4693 if ((*flags & MS_RDONLY) != (sb->s_flags & MS_RDONLY)) {
4694 if (sbi->s_mount_flags & EXT4_MF_FS_ABORTED) {
4699 if (*flags & MS_RDONLY) {
4700 err = dquot_suspend(sb, -1);
4705 * First of all, the unconditional stuff we have to do
4706 * to disable replay of the journal when we next remount
4708 sb->s_flags |= MS_RDONLY;
4711 * OK, test if we are remounting a valid rw partition
4712 * readonly, and if so set the rdonly flag and then
4713 * mark the partition as valid again.
4715 if (!(es->s_state & cpu_to_le16(EXT4_VALID_FS)) &&
4716 (sbi->s_mount_state & EXT4_VALID_FS))
4717 es->s_state = cpu_to_le16(sbi->s_mount_state);
4720 ext4_mark_recovery_complete(sb, es);
4722 /* Make sure we can mount this feature set readwrite */
4723 if (!ext4_feature_set_ok(sb, 0)) {
4728 * Make sure the group descriptor checksums
4729 * are sane. If they aren't, refuse to remount r/w.
4731 for (g = 0; g < sbi->s_groups_count; g++) {
4732 struct ext4_group_desc *gdp =
4733 ext4_get_group_desc(sb, g, NULL);
4735 if (!ext4_group_desc_csum_verify(sb, g, gdp)) {
4736 ext4_msg(sb, KERN_ERR,
4737 "ext4_remount: Checksum for group %u failed (%u!=%u)",
4738 g, le16_to_cpu(ext4_group_desc_csum(sbi, g, gdp)),
4739 le16_to_cpu(gdp->bg_checksum));
4746 * If we have an unprocessed orphan list hanging
4747 * around from a previously readonly bdev mount,
4748 * require a full umount/remount for now.
4750 if (es->s_last_orphan) {
4751 ext4_msg(sb, KERN_WARNING, "Couldn't "
4752 "remount RDWR because of unprocessed "
4753 "orphan inode list. Please "
4754 "umount/remount instead");
4760 * Mounting a RDONLY partition read-write, so reread
4761 * and store the current valid flag. (It may have
4762 * been changed by e2fsck since we originally mounted
4766 ext4_clear_journal_err(sb, es);
4767 sbi->s_mount_state = le16_to_cpu(es->s_state);
4768 if (!ext4_setup_super(sb, es, 0))
4769 sb->s_flags &= ~MS_RDONLY;
4770 if (EXT4_HAS_INCOMPAT_FEATURE(sb,
4771 EXT4_FEATURE_INCOMPAT_MMP))
4772 if (ext4_multi_mount_protect(sb,
4773 le64_to_cpu(es->s_mmp_block))) {
4782 * Reinitialize lazy itable initialization thread based on
4785 if ((sb->s_flags & MS_RDONLY) || !test_opt(sb, INIT_INODE_TABLE))
4786 ext4_unregister_li_request(sb);
4788 ext4_group_t first_not_zeroed;
4789 first_not_zeroed = ext4_has_uninit_itable(sb);
4790 ext4_register_li_request(sb, first_not_zeroed);
4793 ext4_setup_system_zone(sb);
4794 if (sbi->s_journal == NULL && !(old_sb_flags & MS_RDONLY))
4795 ext4_commit_super(sb, 1);
4798 /* Release old quota file names */
4799 for (i = 0; i < MAXQUOTAS; i++)
4800 kfree(old_opts.s_qf_names[i]);
4802 if (sb_any_quota_suspended(sb))
4803 dquot_resume(sb, -1);
4804 else if (EXT4_HAS_RO_COMPAT_FEATURE(sb,
4805 EXT4_FEATURE_RO_COMPAT_QUOTA)) {
4806 err = ext4_enable_quotas(sb);
4813 ext4_msg(sb, KERN_INFO, "re-mounted. Opts: %s", orig_data);
4818 sb->s_flags = old_sb_flags;
4819 sbi->s_mount_opt = old_opts.s_mount_opt;
4820 sbi->s_mount_opt2 = old_opts.s_mount_opt2;
4821 sbi->s_resuid = old_opts.s_resuid;
4822 sbi->s_resgid = old_opts.s_resgid;
4823 sbi->s_commit_interval = old_opts.s_commit_interval;
4824 sbi->s_min_batch_time = old_opts.s_min_batch_time;
4825 sbi->s_max_batch_time = old_opts.s_max_batch_time;
4827 sbi->s_jquota_fmt = old_opts.s_jquota_fmt;
4828 for (i = 0; i < MAXQUOTAS; i++) {
4829 kfree(sbi->s_qf_names[i]);
4830 sbi->s_qf_names[i] = old_opts.s_qf_names[i];
4837 static int ext4_statfs(struct dentry *dentry, struct kstatfs *buf)
4839 struct super_block *sb = dentry->d_sb;
4840 struct ext4_sb_info *sbi = EXT4_SB(sb);
4841 struct ext4_super_block *es = sbi->s_es;
4842 ext4_fsblk_t overhead = 0, resv_blocks;
4845 resv_blocks = EXT4_C2B(sbi, atomic64_read(&sbi->s_resv_clusters));
4847 if (!test_opt(sb, MINIX_DF))
4848 overhead = sbi->s_overhead;
4850 buf->f_type = EXT4_SUPER_MAGIC;
4851 buf->f_bsize = sb->s_blocksize;
4852 buf->f_blocks = ext4_blocks_count(es) - EXT4_C2B(sbi, overhead);
4853 bfree = percpu_counter_sum_positive(&sbi->s_freeclusters_counter) -
4854 percpu_counter_sum_positive(&sbi->s_dirtyclusters_counter);
4855 /* prevent underflow in case that few free space is available */
4856 buf->f_bfree = EXT4_C2B(sbi, max_t(s64, bfree, 0));
4857 buf->f_bavail = buf->f_bfree -
4858 (ext4_r_blocks_count(es) + resv_blocks);
4859 if (buf->f_bfree < (ext4_r_blocks_count(es) + resv_blocks))
4861 buf->f_files = le32_to_cpu(es->s_inodes_count);
4862 buf->f_ffree = percpu_counter_sum_positive(&sbi->s_freeinodes_counter);
4863 buf->f_namelen = EXT4_NAME_LEN;
4864 fsid = le64_to_cpup((void *)es->s_uuid) ^
4865 le64_to_cpup((void *)es->s_uuid + sizeof(u64));
4866 buf->f_fsid.val[0] = fsid & 0xFFFFFFFFUL;
4867 buf->f_fsid.val[1] = (fsid >> 32) & 0xFFFFFFFFUL;
4872 /* Helper function for writing quotas on sync - we need to start transaction
4873 * before quota file is locked for write. Otherwise the are possible deadlocks:
4874 * Process 1 Process 2
4875 * ext4_create() quota_sync()
4876 * jbd2_journal_start() write_dquot()
4877 * dquot_initialize() down(dqio_mutex)
4878 * down(dqio_mutex) jbd2_journal_start()
4884 static inline struct inode *dquot_to_inode(struct dquot *dquot)
4886 return sb_dqopt(dquot->dq_sb)->files[dquot->dq_id.type];
4889 static int ext4_write_dquot(struct dquot *dquot)
4893 struct inode *inode;
4895 inode = dquot_to_inode(dquot);
4896 handle = ext4_journal_start(inode, EXT4_HT_QUOTA,
4897 EXT4_QUOTA_TRANS_BLOCKS(dquot->dq_sb));
4899 return PTR_ERR(handle);
4900 ret = dquot_commit(dquot);
4901 err = ext4_journal_stop(handle);
4907 static int ext4_acquire_dquot(struct dquot *dquot)
4912 handle = ext4_journal_start(dquot_to_inode(dquot), EXT4_HT_QUOTA,
4913 EXT4_QUOTA_INIT_BLOCKS(dquot->dq_sb));
4915 return PTR_ERR(handle);
4916 ret = dquot_acquire(dquot);
4917 err = ext4_journal_stop(handle);
4923 static int ext4_release_dquot(struct dquot *dquot)
4928 handle = ext4_journal_start(dquot_to_inode(dquot), EXT4_HT_QUOTA,
4929 EXT4_QUOTA_DEL_BLOCKS(dquot->dq_sb));
4930 if (IS_ERR(handle)) {
4931 /* Release dquot anyway to avoid endless cycle in dqput() */
4932 dquot_release(dquot);
4933 return PTR_ERR(handle);
4935 ret = dquot_release(dquot);
4936 err = ext4_journal_stop(handle);
4942 static int ext4_mark_dquot_dirty(struct dquot *dquot)
4944 struct super_block *sb = dquot->dq_sb;
4945 struct ext4_sb_info *sbi = EXT4_SB(sb);
4947 /* Are we journaling quotas? */
4948 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_QUOTA) ||
4949 sbi->s_qf_names[USRQUOTA] || sbi->s_qf_names[GRPQUOTA]) {
4950 dquot_mark_dquot_dirty(dquot);
4951 return ext4_write_dquot(dquot);
4953 return dquot_mark_dquot_dirty(dquot);
4957 static int ext4_write_info(struct super_block *sb, int type)
4962 /* Data block + inode block */
4963 handle = ext4_journal_start(sb->s_root->d_inode, EXT4_HT_QUOTA, 2);
4965 return PTR_ERR(handle);
4966 ret = dquot_commit_info(sb, type);
4967 err = ext4_journal_stop(handle);
4974 * Turn on quotas during mount time - we need to find
4975 * the quota file and such...
4977 static int ext4_quota_on_mount(struct super_block *sb, int type)
4979 return dquot_quota_on_mount(sb, EXT4_SB(sb)->s_qf_names[type],
4980 EXT4_SB(sb)->s_jquota_fmt, type);
4984 * Standard function to be called on quota_on
4986 static int ext4_quota_on(struct super_block *sb, int type, int format_id,
4991 if (!test_opt(sb, QUOTA))
4994 /* Quotafile not on the same filesystem? */
4995 if (path->dentry->d_sb != sb)
4997 /* Journaling quota? */
4998 if (EXT4_SB(sb)->s_qf_names[type]) {
4999 /* Quotafile not in fs root? */
5000 if (path->dentry->d_parent != sb->s_root)
5001 ext4_msg(sb, KERN_WARNING,
5002 "Quota file not on filesystem root. "
5003 "Journaled quota will not work");
5007 * When we journal data on quota file, we have to flush journal to see
5008 * all updates to the file when we bypass pagecache...
5010 if (EXT4_SB(sb)->s_journal &&
5011 ext4_should_journal_data(path->dentry->d_inode)) {
5013 * We don't need to lock updates but journal_flush() could
5014 * otherwise be livelocked...
5016 jbd2_journal_lock_updates(EXT4_SB(sb)->s_journal);
5017 err = jbd2_journal_flush(EXT4_SB(sb)->s_journal);
5018 jbd2_journal_unlock_updates(EXT4_SB(sb)->s_journal);
5023 return dquot_quota_on(sb, type, format_id, path);
5026 static int ext4_quota_enable(struct super_block *sb, int type, int format_id,
5030 struct inode *qf_inode;
5031 unsigned long qf_inums[MAXQUOTAS] = {
5032 le32_to_cpu(EXT4_SB(sb)->s_es->s_usr_quota_inum),
5033 le32_to_cpu(EXT4_SB(sb)->s_es->s_grp_quota_inum)
5036 BUG_ON(!EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_QUOTA));
5038 if (!qf_inums[type])
5041 qf_inode = ext4_iget(sb, qf_inums[type]);
5042 if (IS_ERR(qf_inode)) {
5043 ext4_error(sb, "Bad quota inode # %lu", qf_inums[type]);
5044 return PTR_ERR(qf_inode);
5047 /* Don't account quota for quota files to avoid recursion */
5048 qf_inode->i_flags |= S_NOQUOTA;
5049 err = dquot_enable(qf_inode, type, format_id, flags);
5055 /* Enable usage tracking for all quota types. */
5056 static int ext4_enable_quotas(struct super_block *sb)
5059 unsigned long qf_inums[MAXQUOTAS] = {
5060 le32_to_cpu(EXT4_SB(sb)->s_es->s_usr_quota_inum),
5061 le32_to_cpu(EXT4_SB(sb)->s_es->s_grp_quota_inum)
5064 sb_dqopt(sb)->flags |= DQUOT_QUOTA_SYS_FILE;
5065 for (type = 0; type < MAXQUOTAS; type++) {
5066 if (qf_inums[type]) {
5067 err = ext4_quota_enable(sb, type, QFMT_VFS_V1,
5068 DQUOT_USAGE_ENABLED);
5071 "Failed to enable quota tracking "
5072 "(type=%d, err=%d). Please run "
5073 "e2fsck to fix.", type, err);
5082 * quota_on function that is used when QUOTA feature is set.
5084 static int ext4_quota_on_sysfile(struct super_block *sb, int type,
5087 if (!EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_QUOTA))
5091 * USAGE was enabled at mount time. Only need to enable LIMITS now.
5093 return ext4_quota_enable(sb, type, format_id, DQUOT_LIMITS_ENABLED);
5096 static int ext4_quota_off(struct super_block *sb, int type)
5098 struct inode *inode = sb_dqopt(sb)->files[type];
5101 /* Force all delayed allocation blocks to be allocated.
5102 * Caller already holds s_umount sem */
5103 if (test_opt(sb, DELALLOC))
5104 sync_filesystem(sb);
5109 /* Update modification times of quota files when userspace can
5110 * start looking at them */
5111 handle = ext4_journal_start(inode, EXT4_HT_QUOTA, 1);
5114 inode->i_mtime = inode->i_ctime = CURRENT_TIME;
5115 ext4_mark_inode_dirty(handle, inode);
5116 ext4_journal_stop(handle);
5119 return dquot_quota_off(sb, type);
5123 * quota_off function that is used when QUOTA feature is set.
5125 static int ext4_quota_off_sysfile(struct super_block *sb, int type)
5127 if (!EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_QUOTA))
5130 /* Disable only the limits. */
5131 return dquot_disable(sb, type, DQUOT_LIMITS_ENABLED);
5134 /* Read data from quotafile - avoid pagecache and such because we cannot afford
5135 * acquiring the locks... As quota files are never truncated and quota code
5136 * itself serializes the operations (and no one else should touch the files)
5137 * we don't have to be afraid of races */
5138 static ssize_t ext4_quota_read(struct super_block *sb, int type, char *data,
5139 size_t len, loff_t off)
5141 struct inode *inode = sb_dqopt(sb)->files[type];
5142 ext4_lblk_t blk = off >> EXT4_BLOCK_SIZE_BITS(sb);
5144 int offset = off & (sb->s_blocksize - 1);
5147 struct buffer_head *bh;
5148 loff_t i_size = i_size_read(inode);
5152 if (off+len > i_size)
5155 while (toread > 0) {
5156 tocopy = sb->s_blocksize - offset < toread ?
5157 sb->s_blocksize - offset : toread;
5158 bh = ext4_bread(NULL, inode, blk, 0, &err);
5161 if (!bh) /* A hole? */
5162 memset(data, 0, tocopy);
5164 memcpy(data, bh->b_data+offset, tocopy);
5174 /* Write to quotafile (we know the transaction is already started and has
5175 * enough credits) */
5176 static ssize_t ext4_quota_write(struct super_block *sb, int type,
5177 const char *data, size_t len, loff_t off)
5179 struct inode *inode = sb_dqopt(sb)->files[type];
5180 ext4_lblk_t blk = off >> EXT4_BLOCK_SIZE_BITS(sb);
5182 int offset = off & (sb->s_blocksize - 1);
5183 struct buffer_head *bh;
5184 handle_t *handle = journal_current_handle();
5186 if (EXT4_SB(sb)->s_journal && !handle) {
5187 ext4_msg(sb, KERN_WARNING, "Quota write (off=%llu, len=%llu)"
5188 " cancelled because transaction is not started",
5189 (unsigned long long)off, (unsigned long long)len);
5193 * Since we account only one data block in transaction credits,
5194 * then it is impossible to cross a block boundary.
5196 if (sb->s_blocksize - offset < len) {
5197 ext4_msg(sb, KERN_WARNING, "Quota write (off=%llu, len=%llu)"
5198 " cancelled because not block aligned",
5199 (unsigned long long)off, (unsigned long long)len);
5203 bh = ext4_bread(handle, inode, blk, 1, &err);
5206 err = ext4_journal_get_write_access(handle, bh);
5212 memcpy(bh->b_data+offset, data, len);
5213 flush_dcache_page(bh->b_page);
5215 err = ext4_handle_dirty_metadata(handle, NULL, bh);
5220 if (inode->i_size < off + len) {
5221 i_size_write(inode, off + len);
5222 EXT4_I(inode)->i_disksize = inode->i_size;
5223 ext4_mark_inode_dirty(handle, inode);
5230 static struct dentry *ext4_mount(struct file_system_type *fs_type, int flags,
5231 const char *dev_name, void *data)
5233 return mount_bdev(fs_type, flags, dev_name, data, ext4_fill_super);
5236 #if !defined(CONFIG_EXT2_FS) && !defined(CONFIG_EXT2_FS_MODULE) && defined(CONFIG_EXT4_USE_FOR_EXT23)
5237 static inline void register_as_ext2(void)
5239 int err = register_filesystem(&ext2_fs_type);
5242 "EXT4-fs: Unable to register as ext2 (%d)\n", err);
5245 static inline void unregister_as_ext2(void)
5247 unregister_filesystem(&ext2_fs_type);
5250 static inline int ext2_feature_set_ok(struct super_block *sb)
5252 if (EXT4_HAS_INCOMPAT_FEATURE(sb, ~EXT2_FEATURE_INCOMPAT_SUPP))
5254 if (sb->s_flags & MS_RDONLY)
5256 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, ~EXT2_FEATURE_RO_COMPAT_SUPP))
5261 static inline void register_as_ext2(void) { }
5262 static inline void unregister_as_ext2(void) { }
5263 static inline int ext2_feature_set_ok(struct super_block *sb) { return 0; }
5266 #if !defined(CONFIG_EXT3_FS) && !defined(CONFIG_EXT3_FS_MODULE) && defined(CONFIG_EXT4_USE_FOR_EXT23)
5267 static inline void register_as_ext3(void)
5269 int err = register_filesystem(&ext3_fs_type);
5272 "EXT4-fs: Unable to register as ext3 (%d)\n", err);
5275 static inline void unregister_as_ext3(void)
5277 unregister_filesystem(&ext3_fs_type);
5280 static inline int ext3_feature_set_ok(struct super_block *sb)
5282 if (EXT4_HAS_INCOMPAT_FEATURE(sb, ~EXT3_FEATURE_INCOMPAT_SUPP))
5284 if (!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL))
5286 if (sb->s_flags & MS_RDONLY)
5288 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, ~EXT3_FEATURE_RO_COMPAT_SUPP))
5293 static inline void register_as_ext3(void) { }
5294 static inline void unregister_as_ext3(void) { }
5295 static inline int ext3_feature_set_ok(struct super_block *sb) { return 0; }
5298 static struct file_system_type ext4_fs_type = {
5299 .owner = THIS_MODULE,
5301 .mount = ext4_mount,
5302 .kill_sb = kill_block_super,
5303 .fs_flags = FS_REQUIRES_DEV,
5305 MODULE_ALIAS_FS("ext4");
5307 static int __init ext4_init_feat_adverts(void)
5309 struct ext4_features *ef;
5312 ef = kzalloc(sizeof(struct ext4_features), GFP_KERNEL);
5316 ef->f_kobj.kset = ext4_kset;
5317 init_completion(&ef->f_kobj_unregister);
5318 ret = kobject_init_and_add(&ef->f_kobj, &ext4_feat_ktype, NULL,
5331 static void ext4_exit_feat_adverts(void)
5333 kobject_put(&ext4_feat->f_kobj);
5334 wait_for_completion(&ext4_feat->f_kobj_unregister);
5338 /* Shared across all ext4 file systems */
5339 wait_queue_head_t ext4__ioend_wq[EXT4_WQ_HASH_SZ];
5340 struct mutex ext4__aio_mutex[EXT4_WQ_HASH_SZ];
5342 static int __init ext4_init_fs(void)
5346 ext4_li_info = NULL;
5347 mutex_init(&ext4_li_mtx);
5349 /* Build-time check for flags consistency */
5350 ext4_check_flag_values();
5352 for (i = 0; i < EXT4_WQ_HASH_SZ; i++) {
5353 mutex_init(&ext4__aio_mutex[i]);
5354 init_waitqueue_head(&ext4__ioend_wq[i]);
5357 err = ext4_init_es();
5361 err = ext4_init_pageio();
5365 err = ext4_init_system_zone();
5368 ext4_kset = kset_create_and_add("ext4", NULL, fs_kobj);
5373 ext4_proc_root = proc_mkdir("fs/ext4", NULL);
5375 err = ext4_init_feat_adverts();
5379 err = ext4_init_mballoc();
5383 err = ext4_init_xattr();
5386 err = init_inodecache();
5391 err = register_filesystem(&ext4_fs_type);
5397 unregister_as_ext2();
5398 unregister_as_ext3();
5399 destroy_inodecache();
5403 ext4_exit_mballoc();
5405 ext4_exit_feat_adverts();
5408 remove_proc_entry("fs/ext4", NULL);
5409 kset_unregister(ext4_kset);
5411 ext4_exit_system_zone();
5420 static void __exit ext4_exit_fs(void)
5422 ext4_destroy_lazyinit_thread();
5423 unregister_as_ext2();
5424 unregister_as_ext3();
5425 unregister_filesystem(&ext4_fs_type);
5426 destroy_inodecache();
5428 ext4_exit_mballoc();
5429 ext4_exit_feat_adverts();
5430 remove_proc_entry("fs/ext4", NULL);
5431 kset_unregister(ext4_kset);
5432 ext4_exit_system_zone();
5437 MODULE_AUTHOR("Remy Card, Stephen Tweedie, Andrew Morton, Andreas Dilger, Theodore Ts'o and others");
5438 MODULE_DESCRIPTION("Fourth Extended Filesystem");
5439 MODULE_LICENSE("GPL");
5440 module_init(ext4_init_fs)
5441 module_exit(ext4_exit_fs)