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/slab.h>
25 #include <linux/init.h>
26 #include <linux/blkdev.h>
27 #include <linux/backing-dev.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/ctype.h>
38 #include <linux/log2.h>
39 #include <linux/crc16.h>
40 #include <linux/dax.h>
41 #include <linux/cleancache.h>
42 #include <linux/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"
55 #define CREATE_TRACE_POINTS
56 #include <trace/events/ext4.h>
58 static struct ext4_lazy_init *ext4_li_info;
59 static struct mutex ext4_li_mtx;
60 static struct ratelimit_state ext4_mount_msg_ratelimit;
62 static int ext4_load_journal(struct super_block *, struct ext4_super_block *,
63 unsigned long journal_devnum);
64 static int ext4_show_options(struct seq_file *seq, struct dentry *root);
65 static int ext4_commit_super(struct super_block *sb, int sync);
66 static int ext4_mark_recovery_complete(struct super_block *sb,
67 struct ext4_super_block *es);
68 static int ext4_clear_journal_err(struct super_block *sb,
69 struct ext4_super_block *es);
70 static int ext4_sync_fs(struct super_block *sb, int wait);
71 static int ext4_remount(struct super_block *sb, int *flags, char *data);
72 static int ext4_statfs(struct dentry *dentry, struct kstatfs *buf);
73 static int ext4_unfreeze(struct super_block *sb);
74 static int ext4_freeze(struct super_block *sb);
75 static struct dentry *ext4_mount(struct file_system_type *fs_type, int flags,
76 const char *dev_name, void *data);
77 static inline int ext2_feature_set_ok(struct super_block *sb);
78 static inline int ext3_feature_set_ok(struct super_block *sb);
79 static int ext4_feature_set_ok(struct super_block *sb, int readonly);
80 static void ext4_destroy_lazyinit_thread(void);
81 static void ext4_unregister_li_request(struct super_block *sb);
82 static void ext4_clear_request_list(void);
83 static struct inode *ext4_get_journal_inode(struct super_block *sb,
84 unsigned int journal_inum);
89 * Note the difference between i_mmap_sem (EXT4_I(inode)->i_mmap_sem) and
90 * i_mmap_rwsem (inode->i_mmap_rwsem)!
93 * mmap_sem -> sb_start_pagefault -> i_mmap_sem (r) -> transaction start ->
94 * page lock -> i_data_sem (rw)
96 * buffered write path:
97 * sb_start_write -> i_mutex -> mmap_sem
98 * sb_start_write -> i_mutex -> transaction start -> page lock ->
102 * sb_start_write -> i_mutex -> EXT4_STATE_DIOREAD_LOCK (w) -> i_mmap_sem (w) ->
103 * i_mmap_rwsem (w) -> page lock
104 * sb_start_write -> i_mutex -> EXT4_STATE_DIOREAD_LOCK (w) -> i_mmap_sem (w) ->
105 * transaction start -> i_data_sem (rw)
108 * sb_start_write -> i_mutex -> EXT4_STATE_DIOREAD_LOCK (r) -> mmap_sem
109 * sb_start_write -> i_mutex -> EXT4_STATE_DIOREAD_LOCK (r) ->
110 * transaction start -> i_data_sem (rw)
113 * transaction start -> page lock(s) -> i_data_sem (rw)
116 #if !defined(CONFIG_EXT2_FS) && !defined(CONFIG_EXT2_FS_MODULE) && defined(CONFIG_EXT4_USE_FOR_EXT2)
117 static struct file_system_type ext2_fs_type = {
118 .owner = THIS_MODULE,
121 .kill_sb = kill_block_super,
122 .fs_flags = FS_REQUIRES_DEV,
124 MODULE_ALIAS_FS("ext2");
125 MODULE_ALIAS("ext2");
126 #define IS_EXT2_SB(sb) ((sb)->s_bdev->bd_holder == &ext2_fs_type)
128 #define IS_EXT2_SB(sb) (0)
132 static struct file_system_type ext3_fs_type = {
133 .owner = THIS_MODULE,
136 .kill_sb = kill_block_super,
137 .fs_flags = FS_REQUIRES_DEV,
139 MODULE_ALIAS_FS("ext3");
140 MODULE_ALIAS("ext3");
141 #define IS_EXT3_SB(sb) ((sb)->s_bdev->bd_holder == &ext3_fs_type)
144 * This works like sb_bread() except it uses ERR_PTR for error
145 * returns. Currently with sb_bread it's impossible to distinguish
146 * between ENOMEM and EIO situations (since both result in a NULL
150 ext4_sb_bread(struct super_block *sb, sector_t block, int op_flags)
152 struct buffer_head *bh = sb_getblk(sb, block);
155 return ERR_PTR(-ENOMEM);
156 if (buffer_uptodate(bh))
158 ll_rw_block(REQ_OP_READ, REQ_META | op_flags, 1, &bh);
160 if (buffer_uptodate(bh))
163 return ERR_PTR(-EIO);
166 static int ext4_verify_csum_type(struct super_block *sb,
167 struct ext4_super_block *es)
169 if (!ext4_has_feature_metadata_csum(sb))
172 return es->s_checksum_type == EXT4_CRC32C_CHKSUM;
175 static __le32 ext4_superblock_csum(struct super_block *sb,
176 struct ext4_super_block *es)
178 struct ext4_sb_info *sbi = EXT4_SB(sb);
179 int offset = offsetof(struct ext4_super_block, s_checksum);
182 csum = ext4_chksum(sbi, ~0, (char *)es, offset);
184 return cpu_to_le32(csum);
187 static int ext4_superblock_csum_verify(struct super_block *sb,
188 struct ext4_super_block *es)
190 if (!ext4_has_metadata_csum(sb))
193 return es->s_checksum == ext4_superblock_csum(sb, es);
196 void ext4_superblock_csum_set(struct super_block *sb)
198 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
200 if (!ext4_has_metadata_csum(sb))
203 es->s_checksum = ext4_superblock_csum(sb, es);
206 void *ext4_kvmalloc(size_t size, gfp_t flags)
210 ret = kmalloc(size, flags | __GFP_NOWARN);
212 ret = __vmalloc(size, flags, PAGE_KERNEL);
216 void *ext4_kvzalloc(size_t size, gfp_t flags)
220 ret = kzalloc(size, flags | __GFP_NOWARN);
222 ret = __vmalloc(size, flags | __GFP_ZERO, PAGE_KERNEL);
226 ext4_fsblk_t ext4_block_bitmap(struct super_block *sb,
227 struct ext4_group_desc *bg)
229 return le32_to_cpu(bg->bg_block_bitmap_lo) |
230 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
231 (ext4_fsblk_t)le32_to_cpu(bg->bg_block_bitmap_hi) << 32 : 0);
234 ext4_fsblk_t ext4_inode_bitmap(struct super_block *sb,
235 struct ext4_group_desc *bg)
237 return le32_to_cpu(bg->bg_inode_bitmap_lo) |
238 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
239 (ext4_fsblk_t)le32_to_cpu(bg->bg_inode_bitmap_hi) << 32 : 0);
242 ext4_fsblk_t ext4_inode_table(struct super_block *sb,
243 struct ext4_group_desc *bg)
245 return le32_to_cpu(bg->bg_inode_table_lo) |
246 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
247 (ext4_fsblk_t)le32_to_cpu(bg->bg_inode_table_hi) << 32 : 0);
250 __u32 ext4_free_group_clusters(struct super_block *sb,
251 struct ext4_group_desc *bg)
253 return le16_to_cpu(bg->bg_free_blocks_count_lo) |
254 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
255 (__u32)le16_to_cpu(bg->bg_free_blocks_count_hi) << 16 : 0);
258 __u32 ext4_free_inodes_count(struct super_block *sb,
259 struct ext4_group_desc *bg)
261 return le16_to_cpu(bg->bg_free_inodes_count_lo) |
262 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
263 (__u32)le16_to_cpu(bg->bg_free_inodes_count_hi) << 16 : 0);
266 __u32 ext4_used_dirs_count(struct super_block *sb,
267 struct ext4_group_desc *bg)
269 return le16_to_cpu(bg->bg_used_dirs_count_lo) |
270 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
271 (__u32)le16_to_cpu(bg->bg_used_dirs_count_hi) << 16 : 0);
274 __u32 ext4_itable_unused_count(struct super_block *sb,
275 struct ext4_group_desc *bg)
277 return le16_to_cpu(bg->bg_itable_unused_lo) |
278 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
279 (__u32)le16_to_cpu(bg->bg_itable_unused_hi) << 16 : 0);
282 void ext4_block_bitmap_set(struct super_block *sb,
283 struct ext4_group_desc *bg, ext4_fsblk_t blk)
285 bg->bg_block_bitmap_lo = cpu_to_le32((u32)blk);
286 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
287 bg->bg_block_bitmap_hi = cpu_to_le32(blk >> 32);
290 void ext4_inode_bitmap_set(struct super_block *sb,
291 struct ext4_group_desc *bg, ext4_fsblk_t blk)
293 bg->bg_inode_bitmap_lo = cpu_to_le32((u32)blk);
294 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
295 bg->bg_inode_bitmap_hi = cpu_to_le32(blk >> 32);
298 void ext4_inode_table_set(struct super_block *sb,
299 struct ext4_group_desc *bg, ext4_fsblk_t blk)
301 bg->bg_inode_table_lo = cpu_to_le32((u32)blk);
302 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
303 bg->bg_inode_table_hi = cpu_to_le32(blk >> 32);
306 void ext4_free_group_clusters_set(struct super_block *sb,
307 struct ext4_group_desc *bg, __u32 count)
309 bg->bg_free_blocks_count_lo = cpu_to_le16((__u16)count);
310 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
311 bg->bg_free_blocks_count_hi = cpu_to_le16(count >> 16);
314 void ext4_free_inodes_set(struct super_block *sb,
315 struct ext4_group_desc *bg, __u32 count)
317 bg->bg_free_inodes_count_lo = cpu_to_le16((__u16)count);
318 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
319 bg->bg_free_inodes_count_hi = cpu_to_le16(count >> 16);
322 void ext4_used_dirs_set(struct super_block *sb,
323 struct ext4_group_desc *bg, __u32 count)
325 bg->bg_used_dirs_count_lo = cpu_to_le16((__u16)count);
326 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
327 bg->bg_used_dirs_count_hi = cpu_to_le16(count >> 16);
330 void ext4_itable_unused_set(struct super_block *sb,
331 struct ext4_group_desc *bg, __u32 count)
333 bg->bg_itable_unused_lo = cpu_to_le16((__u16)count);
334 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
335 bg->bg_itable_unused_hi = cpu_to_le16(count >> 16);
339 static void __save_error_info(struct super_block *sb, const char *func,
342 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
344 EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS;
345 if (bdev_read_only(sb->s_bdev))
347 es->s_state |= cpu_to_le16(EXT4_ERROR_FS);
348 es->s_last_error_time = cpu_to_le32(get_seconds());
349 strncpy(es->s_last_error_func, func, sizeof(es->s_last_error_func));
350 es->s_last_error_line = cpu_to_le32(line);
351 if (!es->s_first_error_time) {
352 es->s_first_error_time = es->s_last_error_time;
353 strncpy(es->s_first_error_func, func,
354 sizeof(es->s_first_error_func));
355 es->s_first_error_line = cpu_to_le32(line);
356 es->s_first_error_ino = es->s_last_error_ino;
357 es->s_first_error_block = es->s_last_error_block;
360 * Start the daily error reporting function if it hasn't been
363 if (!es->s_error_count)
364 mod_timer(&EXT4_SB(sb)->s_err_report, jiffies + 24*60*60*HZ);
365 le32_add_cpu(&es->s_error_count, 1);
368 static void save_error_info(struct super_block *sb, const char *func,
371 __save_error_info(sb, func, line);
372 if (!bdev_read_only(sb->s_bdev))
373 ext4_commit_super(sb, 1);
377 * The del_gendisk() function uninitializes the disk-specific data
378 * structures, including the bdi structure, without telling anyone
379 * else. Once this happens, any attempt to call mark_buffer_dirty()
380 * (for example, by ext4_commit_super), will cause a kernel OOPS.
381 * This is a kludge to prevent these oops until we can put in a proper
382 * hook in del_gendisk() to inform the VFS and file system layers.
384 static int block_device_ejected(struct super_block *sb)
386 struct inode *bd_inode = sb->s_bdev->bd_inode;
387 struct backing_dev_info *bdi = inode_to_bdi(bd_inode);
389 return bdi->dev == NULL;
392 static void ext4_journal_commit_callback(journal_t *journal, transaction_t *txn)
394 struct super_block *sb = journal->j_private;
395 struct ext4_sb_info *sbi = EXT4_SB(sb);
396 int error = is_journal_aborted(journal);
397 struct ext4_journal_cb_entry *jce;
399 BUG_ON(txn->t_state == T_FINISHED);
401 ext4_process_freed_data(sb, txn->t_tid);
403 spin_lock(&sbi->s_md_lock);
404 while (!list_empty(&txn->t_private_list)) {
405 jce = list_entry(txn->t_private_list.next,
406 struct ext4_journal_cb_entry, jce_list);
407 list_del_init(&jce->jce_list);
408 spin_unlock(&sbi->s_md_lock);
409 jce->jce_func(sb, jce, error);
410 spin_lock(&sbi->s_md_lock);
412 spin_unlock(&sbi->s_md_lock);
415 /* Deal with the reporting of failure conditions on a filesystem such as
416 * inconsistencies detected or read IO failures.
418 * On ext2, we can store the error state of the filesystem in the
419 * superblock. That is not possible on ext4, because we may have other
420 * write ordering constraints on the superblock which prevent us from
421 * writing it out straight away; and given that the journal is about to
422 * be aborted, we can't rely on the current, or future, transactions to
423 * write out the superblock safely.
425 * We'll just use the jbd2_journal_abort() error code to record an error in
426 * the journal instead. On recovery, the journal will complain about
427 * that error until we've noted it down and cleared it.
430 static void ext4_handle_error(struct super_block *sb)
435 if (!test_opt(sb, ERRORS_CONT)) {
436 journal_t *journal = EXT4_SB(sb)->s_journal;
438 EXT4_SB(sb)->s_mount_flags |= EXT4_MF_FS_ABORTED;
440 jbd2_journal_abort(journal, -EIO);
442 if (test_opt(sb, ERRORS_RO)) {
443 ext4_msg(sb, KERN_CRIT, "Remounting filesystem read-only");
445 * Make sure updated value of ->s_mount_flags will be visible
446 * before ->s_flags update
449 sb->s_flags |= MS_RDONLY;
451 if (test_opt(sb, ERRORS_PANIC)) {
452 if (EXT4_SB(sb)->s_journal &&
453 !(EXT4_SB(sb)->s_journal->j_flags & JBD2_REC_ERR))
455 panic("EXT4-fs (device %s): panic forced after error\n",
460 #define ext4_error_ratelimit(sb) \
461 ___ratelimit(&(EXT4_SB(sb)->s_err_ratelimit_state), \
464 void __ext4_error(struct super_block *sb, const char *function,
465 unsigned int line, const char *fmt, ...)
467 struct va_format vaf;
470 if (unlikely(ext4_forced_shutdown(EXT4_SB(sb))))
473 if (ext4_error_ratelimit(sb)) {
478 "EXT4-fs error (device %s): %s:%d: comm %s: %pV\n",
479 sb->s_id, function, line, current->comm, &vaf);
482 save_error_info(sb, function, line);
483 ext4_handle_error(sb);
486 void __ext4_error_inode(struct inode *inode, const char *function,
487 unsigned int line, ext4_fsblk_t block,
488 const char *fmt, ...)
491 struct va_format vaf;
492 struct ext4_super_block *es = EXT4_SB(inode->i_sb)->s_es;
494 if (unlikely(ext4_forced_shutdown(EXT4_SB(inode->i_sb))))
497 es->s_last_error_ino = cpu_to_le32(inode->i_ino);
498 es->s_last_error_block = cpu_to_le64(block);
499 if (ext4_error_ratelimit(inode->i_sb)) {
504 printk(KERN_CRIT "EXT4-fs error (device %s): %s:%d: "
505 "inode #%lu: block %llu: comm %s: %pV\n",
506 inode->i_sb->s_id, function, line, inode->i_ino,
507 block, current->comm, &vaf);
509 printk(KERN_CRIT "EXT4-fs error (device %s): %s:%d: "
510 "inode #%lu: comm %s: %pV\n",
511 inode->i_sb->s_id, function, line, inode->i_ino,
512 current->comm, &vaf);
515 save_error_info(inode->i_sb, function, line);
516 ext4_handle_error(inode->i_sb);
519 void __ext4_error_file(struct file *file, const char *function,
520 unsigned int line, ext4_fsblk_t block,
521 const char *fmt, ...)
524 struct va_format vaf;
525 struct ext4_super_block *es;
526 struct inode *inode = file_inode(file);
527 char pathname[80], *path;
529 if (unlikely(ext4_forced_shutdown(EXT4_SB(inode->i_sb))))
532 es = EXT4_SB(inode->i_sb)->s_es;
533 es->s_last_error_ino = cpu_to_le32(inode->i_ino);
534 if (ext4_error_ratelimit(inode->i_sb)) {
535 path = file_path(file, pathname, sizeof(pathname));
543 "EXT4-fs error (device %s): %s:%d: inode #%lu: "
544 "block %llu: comm %s: path %s: %pV\n",
545 inode->i_sb->s_id, function, line, inode->i_ino,
546 block, current->comm, path, &vaf);
549 "EXT4-fs error (device %s): %s:%d: inode #%lu: "
550 "comm %s: path %s: %pV\n",
551 inode->i_sb->s_id, function, line, inode->i_ino,
552 current->comm, path, &vaf);
555 save_error_info(inode->i_sb, function, line);
556 ext4_handle_error(inode->i_sb);
559 const char *ext4_decode_error(struct super_block *sb, int errno,
566 errstr = "Corrupt filesystem";
569 errstr = "Filesystem failed CRC";
572 errstr = "IO failure";
575 errstr = "Out of memory";
578 if (!sb || (EXT4_SB(sb)->s_journal &&
579 EXT4_SB(sb)->s_journal->j_flags & JBD2_ABORT))
580 errstr = "Journal has aborted";
582 errstr = "Readonly filesystem";
585 /* If the caller passed in an extra buffer for unknown
586 * errors, textualise them now. Else we just return
589 /* Check for truncated error codes... */
590 if (snprintf(nbuf, 16, "error %d", -errno) >= 0)
599 /* __ext4_std_error decodes expected errors from journaling functions
600 * automatically and invokes the appropriate error response. */
602 void __ext4_std_error(struct super_block *sb, const char *function,
603 unsigned int line, int errno)
608 if (unlikely(ext4_forced_shutdown(EXT4_SB(sb))))
611 /* Special case: if the error is EROFS, and we're not already
612 * inside a transaction, then there's really no point in logging
614 if (errno == -EROFS && journal_current_handle() == NULL && sb_rdonly(sb))
617 if (ext4_error_ratelimit(sb)) {
618 errstr = ext4_decode_error(sb, errno, nbuf);
619 printk(KERN_CRIT "EXT4-fs error (device %s) in %s:%d: %s\n",
620 sb->s_id, function, line, errstr);
623 save_error_info(sb, function, line);
624 ext4_handle_error(sb);
628 * ext4_abort is a much stronger failure handler than ext4_error. The
629 * abort function may be used to deal with unrecoverable failures such
630 * as journal IO errors or ENOMEM at a critical moment in log management.
632 * We unconditionally force the filesystem into an ABORT|READONLY state,
633 * unless the error response on the fs has been set to panic in which
634 * case we take the easy way out and panic immediately.
637 void __ext4_abort(struct super_block *sb, const char *function,
638 unsigned int line, const char *fmt, ...)
640 struct va_format vaf;
643 if (unlikely(ext4_forced_shutdown(EXT4_SB(sb))))
646 save_error_info(sb, function, line);
650 printk(KERN_CRIT "EXT4-fs error (device %s): %s:%d: %pV\n",
651 sb->s_id, function, line, &vaf);
654 if (sb_rdonly(sb) == 0) {
655 ext4_msg(sb, KERN_CRIT, "Remounting filesystem read-only");
656 EXT4_SB(sb)->s_mount_flags |= EXT4_MF_FS_ABORTED;
658 * Make sure updated value of ->s_mount_flags will be visible
659 * before ->s_flags update
662 sb->s_flags |= MS_RDONLY;
663 if (EXT4_SB(sb)->s_journal)
664 jbd2_journal_abort(EXT4_SB(sb)->s_journal, -EIO);
665 save_error_info(sb, function, line);
667 if (test_opt(sb, ERRORS_PANIC)) {
668 if (EXT4_SB(sb)->s_journal &&
669 !(EXT4_SB(sb)->s_journal->j_flags & JBD2_REC_ERR))
671 panic("EXT4-fs panic from previous error\n");
675 void __ext4_msg(struct super_block *sb,
676 const char *prefix, const char *fmt, ...)
678 struct va_format vaf;
681 if (!___ratelimit(&(EXT4_SB(sb)->s_msg_ratelimit_state), "EXT4-fs"))
687 printk("%sEXT4-fs (%s): %pV\n", prefix, sb->s_id, &vaf);
691 #define ext4_warning_ratelimit(sb) \
692 ___ratelimit(&(EXT4_SB(sb)->s_warning_ratelimit_state), \
695 void __ext4_warning(struct super_block *sb, const char *function,
696 unsigned int line, const char *fmt, ...)
698 struct va_format vaf;
701 if (!ext4_warning_ratelimit(sb))
707 printk(KERN_WARNING "EXT4-fs warning (device %s): %s:%d: %pV\n",
708 sb->s_id, function, line, &vaf);
712 void __ext4_warning_inode(const struct inode *inode, const char *function,
713 unsigned int line, const char *fmt, ...)
715 struct va_format vaf;
718 if (!ext4_warning_ratelimit(inode->i_sb))
724 printk(KERN_WARNING "EXT4-fs warning (device %s): %s:%d: "
725 "inode #%lu: comm %s: %pV\n", inode->i_sb->s_id,
726 function, line, inode->i_ino, current->comm, &vaf);
730 void __ext4_grp_locked_error(const char *function, unsigned int line,
731 struct super_block *sb, ext4_group_t grp,
732 unsigned long ino, ext4_fsblk_t block,
733 const char *fmt, ...)
737 struct va_format vaf;
739 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
741 if (unlikely(ext4_forced_shutdown(EXT4_SB(sb))))
744 es->s_last_error_ino = cpu_to_le32(ino);
745 es->s_last_error_block = cpu_to_le64(block);
746 __save_error_info(sb, function, line);
748 if (ext4_error_ratelimit(sb)) {
752 printk(KERN_CRIT "EXT4-fs error (device %s): %s:%d: group %u, ",
753 sb->s_id, function, line, grp);
755 printk(KERN_CONT "inode %lu: ", ino);
757 printk(KERN_CONT "block %llu:",
758 (unsigned long long) block);
759 printk(KERN_CONT "%pV\n", &vaf);
763 if (test_opt(sb, ERRORS_CONT)) {
764 ext4_commit_super(sb, 0);
768 ext4_unlock_group(sb, grp);
769 ext4_commit_super(sb, 1);
770 ext4_handle_error(sb);
772 * We only get here in the ERRORS_RO case; relocking the group
773 * may be dangerous, but nothing bad will happen since the
774 * filesystem will have already been marked read/only and the
775 * journal has been aborted. We return 1 as a hint to callers
776 * who might what to use the return value from
777 * ext4_grp_locked_error() to distinguish between the
778 * ERRORS_CONT and ERRORS_RO case, and perhaps return more
779 * aggressively from the ext4 function in question, with a
780 * more appropriate error code.
782 ext4_lock_group(sb, grp);
786 void ext4_update_dynamic_rev(struct super_block *sb)
788 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
790 if (le32_to_cpu(es->s_rev_level) > EXT4_GOOD_OLD_REV)
794 "updating to rev %d because of new feature flag, "
795 "running e2fsck is recommended",
798 es->s_first_ino = cpu_to_le32(EXT4_GOOD_OLD_FIRST_INO);
799 es->s_inode_size = cpu_to_le16(EXT4_GOOD_OLD_INODE_SIZE);
800 es->s_rev_level = cpu_to_le32(EXT4_DYNAMIC_REV);
801 /* leave es->s_feature_*compat flags alone */
802 /* es->s_uuid will be set by e2fsck if empty */
805 * The rest of the superblock fields should be zero, and if not it
806 * means they are likely already in use, so leave them alone. We
807 * can leave it up to e2fsck to clean up any inconsistencies there.
812 * Open the external journal device
814 static struct block_device *ext4_blkdev_get(dev_t dev, struct super_block *sb)
816 struct block_device *bdev;
817 char b[BDEVNAME_SIZE];
819 bdev = blkdev_get_by_dev(dev, FMODE_READ|FMODE_WRITE|FMODE_EXCL, sb);
825 ext4_msg(sb, KERN_ERR, "failed to open journal device %s: %ld",
826 __bdevname(dev, b), PTR_ERR(bdev));
831 * Release the journal device
833 static void ext4_blkdev_put(struct block_device *bdev)
835 blkdev_put(bdev, FMODE_READ|FMODE_WRITE|FMODE_EXCL);
838 static void ext4_blkdev_remove(struct ext4_sb_info *sbi)
840 struct block_device *bdev;
841 bdev = sbi->journal_bdev;
843 ext4_blkdev_put(bdev);
844 sbi->journal_bdev = NULL;
848 static inline struct inode *orphan_list_entry(struct list_head *l)
850 return &list_entry(l, struct ext4_inode_info, i_orphan)->vfs_inode;
853 static void dump_orphan_list(struct super_block *sb, struct ext4_sb_info *sbi)
857 ext4_msg(sb, KERN_ERR, "sb orphan head is %d",
858 le32_to_cpu(sbi->s_es->s_last_orphan));
860 printk(KERN_ERR "sb_info orphan list:\n");
861 list_for_each(l, &sbi->s_orphan) {
862 struct inode *inode = orphan_list_entry(l);
864 "inode %s:%lu at %p: mode %o, nlink %d, next %d\n",
865 inode->i_sb->s_id, inode->i_ino, inode,
866 inode->i_mode, inode->i_nlink,
872 static int ext4_quota_off(struct super_block *sb, int type);
874 static inline void ext4_quota_off_umount(struct super_block *sb)
878 /* Use our quota_off function to clear inode flags etc. */
879 for (type = 0; type < EXT4_MAXQUOTAS; type++)
880 ext4_quota_off(sb, type);
884 * This is a helper function which is used in the mount/remount
885 * codepaths (which holds s_umount) to fetch the quota file name.
887 static inline char *get_qf_name(struct super_block *sb,
888 struct ext4_sb_info *sbi,
891 return rcu_dereference_protected(sbi->s_qf_names[type],
892 lockdep_is_held(&sb->s_umount));
895 static inline void ext4_quota_off_umount(struct super_block *sb)
900 static void ext4_put_super(struct super_block *sb)
902 struct ext4_sb_info *sbi = EXT4_SB(sb);
903 struct ext4_super_block *es = sbi->s_es;
904 struct buffer_head **group_desc;
905 struct flex_groups **flex_groups;
909 ext4_unregister_li_request(sb);
910 ext4_quota_off_umount(sb);
912 flush_workqueue(sbi->rsv_conversion_wq);
913 destroy_workqueue(sbi->rsv_conversion_wq);
915 if (sbi->s_journal) {
916 aborted = is_journal_aborted(sbi->s_journal);
917 err = jbd2_journal_destroy(sbi->s_journal);
918 sbi->s_journal = NULL;
919 if ((err < 0) && !aborted)
920 ext4_abort(sb, "Couldn't clean up the journal");
923 ext4_unregister_sysfs(sb);
924 ext4_es_unregister_shrinker(sbi);
925 del_timer_sync(&sbi->s_err_report);
926 ext4_release_system_zone(sb);
928 ext4_ext_release(sb);
930 if (!sb_rdonly(sb) && !aborted) {
931 ext4_clear_feature_journal_needs_recovery(sb);
932 es->s_state = cpu_to_le16(sbi->s_mount_state);
935 ext4_commit_super(sb, 1);
938 group_desc = rcu_dereference(sbi->s_group_desc);
939 for (i = 0; i < sbi->s_gdb_count; i++)
940 brelse(group_desc[i]);
942 flex_groups = rcu_dereference(sbi->s_flex_groups);
944 for (i = 0; i < sbi->s_flex_groups_allocated; i++)
945 kvfree(flex_groups[i]);
949 percpu_counter_destroy(&sbi->s_freeclusters_counter);
950 percpu_counter_destroy(&sbi->s_freeinodes_counter);
951 percpu_counter_destroy(&sbi->s_dirs_counter);
952 percpu_counter_destroy(&sbi->s_dirtyclusters_counter);
953 percpu_free_rwsem(&sbi->s_writepages_rwsem);
955 for (i = 0; i < EXT4_MAXQUOTAS; i++)
956 kfree(get_qf_name(sb, sbi, i));
959 /* Debugging code just in case the in-memory inode orphan list
960 * isn't empty. The on-disk one can be non-empty if we've
961 * detected an error and taken the fs readonly, but the
962 * in-memory list had better be clean by this point. */
963 if (!list_empty(&sbi->s_orphan))
964 dump_orphan_list(sb, sbi);
965 J_ASSERT(list_empty(&sbi->s_orphan));
967 sync_blockdev(sb->s_bdev);
968 invalidate_bdev(sb->s_bdev);
969 if (sbi->journal_bdev && sbi->journal_bdev != sb->s_bdev) {
971 * Invalidate the journal device's buffers. We don't want them
972 * floating about in memory - the physical journal device may
973 * hotswapped, and it breaks the `ro-after' testing code.
975 sync_blockdev(sbi->journal_bdev);
976 invalidate_bdev(sbi->journal_bdev);
977 ext4_blkdev_remove(sbi);
979 if (sbi->s_ea_inode_cache) {
980 ext4_xattr_destroy_cache(sbi->s_ea_inode_cache);
981 sbi->s_ea_inode_cache = NULL;
983 if (sbi->s_ea_block_cache) {
984 ext4_xattr_destroy_cache(sbi->s_ea_block_cache);
985 sbi->s_ea_block_cache = NULL;
988 kthread_stop(sbi->s_mmp_tsk);
990 sb->s_fs_info = NULL;
992 * Now that we are completely done shutting down the
993 * superblock, we need to actually destroy the kobject.
995 kobject_put(&sbi->s_kobj);
996 wait_for_completion(&sbi->s_kobj_unregister);
997 if (sbi->s_chksum_driver)
998 crypto_free_shash(sbi->s_chksum_driver);
999 kfree(sbi->s_blockgroup_lock);
1000 fs_put_dax(sbi->s_daxdev);
1004 static struct kmem_cache *ext4_inode_cachep;
1007 * Called inside transaction, so use GFP_NOFS
1009 static struct inode *ext4_alloc_inode(struct super_block *sb)
1011 struct ext4_inode_info *ei;
1013 ei = kmem_cache_alloc(ext4_inode_cachep, GFP_NOFS);
1017 ei->vfs_inode.i_version = 1;
1018 spin_lock_init(&ei->i_raw_lock);
1019 INIT_LIST_HEAD(&ei->i_prealloc_list);
1020 spin_lock_init(&ei->i_prealloc_lock);
1021 ext4_es_init_tree(&ei->i_es_tree);
1022 rwlock_init(&ei->i_es_lock);
1023 INIT_LIST_HEAD(&ei->i_es_list);
1024 ei->i_es_all_nr = 0;
1025 ei->i_es_shk_nr = 0;
1026 ei->i_es_shrink_lblk = 0;
1027 ei->i_reserved_data_blocks = 0;
1028 ei->i_da_metadata_calc_len = 0;
1029 ei->i_da_metadata_calc_last_lblock = 0;
1030 spin_lock_init(&(ei->i_block_reservation_lock));
1032 ei->i_reserved_quota = 0;
1033 memset(&ei->i_dquot, 0, sizeof(ei->i_dquot));
1036 INIT_LIST_HEAD(&ei->i_rsv_conversion_list);
1037 spin_lock_init(&ei->i_completed_io_lock);
1039 ei->i_datasync_tid = 0;
1040 atomic_set(&ei->i_unwritten, 0);
1041 INIT_WORK(&ei->i_rsv_conversion_work, ext4_end_io_rsv_work);
1042 return &ei->vfs_inode;
1045 static int ext4_drop_inode(struct inode *inode)
1047 int drop = generic_drop_inode(inode);
1049 trace_ext4_drop_inode(inode, drop);
1053 static void ext4_i_callback(struct rcu_head *head)
1055 struct inode *inode = container_of(head, struct inode, i_rcu);
1056 kmem_cache_free(ext4_inode_cachep, EXT4_I(inode));
1059 static void ext4_destroy_inode(struct inode *inode)
1061 if (!list_empty(&(EXT4_I(inode)->i_orphan))) {
1062 ext4_msg(inode->i_sb, KERN_ERR,
1063 "Inode %lu (%p): orphan list check failed!",
1064 inode->i_ino, EXT4_I(inode));
1065 print_hex_dump(KERN_INFO, "", DUMP_PREFIX_ADDRESS, 16, 4,
1066 EXT4_I(inode), sizeof(struct ext4_inode_info),
1070 call_rcu(&inode->i_rcu, ext4_i_callback);
1073 static void init_once(void *foo)
1075 struct ext4_inode_info *ei = (struct ext4_inode_info *) foo;
1077 INIT_LIST_HEAD(&ei->i_orphan);
1078 init_rwsem(&ei->xattr_sem);
1079 init_rwsem(&ei->i_data_sem);
1080 init_rwsem(&ei->i_mmap_sem);
1081 inode_init_once(&ei->vfs_inode);
1084 static int __init init_inodecache(void)
1086 ext4_inode_cachep = kmem_cache_create("ext4_inode_cache",
1087 sizeof(struct ext4_inode_info),
1088 0, (SLAB_RECLAIM_ACCOUNT|
1089 SLAB_MEM_SPREAD|SLAB_ACCOUNT),
1091 if (ext4_inode_cachep == NULL)
1096 static void destroy_inodecache(void)
1099 * Make sure all delayed rcu free inodes are flushed before we
1103 kmem_cache_destroy(ext4_inode_cachep);
1106 void ext4_clear_inode(struct inode *inode)
1108 invalidate_inode_buffers(inode);
1111 ext4_discard_preallocations(inode);
1112 ext4_es_remove_extent(inode, 0, EXT_MAX_BLOCKS);
1113 if (EXT4_I(inode)->jinode) {
1114 jbd2_journal_release_jbd_inode(EXT4_JOURNAL(inode),
1115 EXT4_I(inode)->jinode);
1116 jbd2_free_inode(EXT4_I(inode)->jinode);
1117 EXT4_I(inode)->jinode = NULL;
1119 #ifdef CONFIG_EXT4_FS_ENCRYPTION
1120 fscrypt_put_encryption_info(inode, NULL);
1124 static struct inode *ext4_nfs_get_inode(struct super_block *sb,
1125 u64 ino, u32 generation)
1127 struct inode *inode;
1130 * Currently we don't know the generation for parent directory, so
1131 * a generation of 0 means "accept any"
1133 inode = ext4_iget(sb, ino, EXT4_IGET_HANDLE);
1135 return ERR_CAST(inode);
1136 if (generation && inode->i_generation != generation) {
1138 return ERR_PTR(-ESTALE);
1144 static struct dentry *ext4_fh_to_dentry(struct super_block *sb, struct fid *fid,
1145 int fh_len, int fh_type)
1147 return generic_fh_to_dentry(sb, fid, fh_len, fh_type,
1148 ext4_nfs_get_inode);
1151 static struct dentry *ext4_fh_to_parent(struct super_block *sb, struct fid *fid,
1152 int fh_len, int fh_type)
1154 return generic_fh_to_parent(sb, fid, fh_len, fh_type,
1155 ext4_nfs_get_inode);
1158 static int ext4_nfs_commit_metadata(struct inode *inode)
1160 struct writeback_control wbc = {
1161 .sync_mode = WB_SYNC_ALL
1164 trace_ext4_nfs_commit_metadata(inode);
1165 return ext4_write_inode(inode, &wbc);
1169 * Try to release metadata pages (indirect blocks, directories) which are
1170 * mapped via the block device. Since these pages could have journal heads
1171 * which would prevent try_to_free_buffers() from freeing them, we must use
1172 * jbd2 layer's try_to_free_buffers() function to release them.
1174 static int bdev_try_to_free_page(struct super_block *sb, struct page *page,
1177 journal_t *journal = EXT4_SB(sb)->s_journal;
1179 WARN_ON(PageChecked(page));
1180 if (!page_has_buffers(page))
1183 return jbd2_journal_try_to_free_buffers(journal, page,
1184 wait & ~__GFP_DIRECT_RECLAIM);
1185 return try_to_free_buffers(page);
1188 #ifdef CONFIG_EXT4_FS_ENCRYPTION
1189 static int ext4_get_context(struct inode *inode, void *ctx, size_t len)
1191 return ext4_xattr_get(inode, EXT4_XATTR_INDEX_ENCRYPTION,
1192 EXT4_XATTR_NAME_ENCRYPTION_CONTEXT, ctx, len);
1195 static int ext4_set_context(struct inode *inode, const void *ctx, size_t len,
1198 handle_t *handle = fs_data;
1199 int res, res2, credits, retries = 0;
1202 * Encrypting the root directory is not allowed because e2fsck expects
1203 * lost+found to exist and be unencrypted, and encrypting the root
1204 * directory would imply encrypting the lost+found directory as well as
1205 * the filename "lost+found" itself.
1207 if (inode->i_ino == EXT4_ROOT_INO)
1210 res = ext4_convert_inline_data(inode);
1215 * If a journal handle was specified, then the encryption context is
1216 * being set on a new inode via inheritance and is part of a larger
1217 * transaction to create the inode. Otherwise the encryption context is
1218 * being set on an existing inode in its own transaction. Only in the
1219 * latter case should the "retry on ENOSPC" logic be used.
1223 res = ext4_xattr_set_handle(handle, inode,
1224 EXT4_XATTR_INDEX_ENCRYPTION,
1225 EXT4_XATTR_NAME_ENCRYPTION_CONTEXT,
1228 ext4_set_inode_flag(inode, EXT4_INODE_ENCRYPT);
1229 ext4_clear_inode_state(inode,
1230 EXT4_STATE_MAY_INLINE_DATA);
1232 * Update inode->i_flags - e.g. S_DAX may get disabled
1234 ext4_set_inode_flags(inode);
1239 res = dquot_initialize(inode);
1243 res = ext4_xattr_set_credits(inode, len, false /* is_create */,
1248 handle = ext4_journal_start(inode, EXT4_HT_MISC, credits);
1250 return PTR_ERR(handle);
1252 res = ext4_xattr_set_handle(handle, inode, EXT4_XATTR_INDEX_ENCRYPTION,
1253 EXT4_XATTR_NAME_ENCRYPTION_CONTEXT,
1256 ext4_set_inode_flag(inode, EXT4_INODE_ENCRYPT);
1257 /* Update inode->i_flags - e.g. S_DAX may get disabled */
1258 ext4_set_inode_flags(inode);
1259 res = ext4_mark_inode_dirty(handle, inode);
1261 EXT4_ERROR_INODE(inode, "Failed to mark inode dirty");
1263 res2 = ext4_journal_stop(handle);
1265 if (res == -ENOSPC && ext4_should_retry_alloc(inode->i_sb, &retries))
1272 static bool ext4_dummy_context(struct inode *inode)
1274 return DUMMY_ENCRYPTION_ENABLED(EXT4_SB(inode->i_sb));
1277 static unsigned ext4_max_namelen(struct inode *inode)
1279 return S_ISLNK(inode->i_mode) ? inode->i_sb->s_blocksize :
1283 static const struct fscrypt_operations ext4_cryptops = {
1284 .key_prefix = "ext4:",
1285 .get_context = ext4_get_context,
1286 .set_context = ext4_set_context,
1287 .dummy_context = ext4_dummy_context,
1288 .is_encrypted = ext4_encrypted_inode,
1289 .empty_dir = ext4_empty_dir,
1290 .max_namelen = ext4_max_namelen,
1293 static const struct fscrypt_operations ext4_cryptops = {
1294 .is_encrypted = ext4_encrypted_inode,
1299 static const char * const quotatypes[] = INITQFNAMES;
1300 #define QTYPE2NAME(t) (quotatypes[t])
1302 static int ext4_write_dquot(struct dquot *dquot);
1303 static int ext4_acquire_dquot(struct dquot *dquot);
1304 static int ext4_release_dquot(struct dquot *dquot);
1305 static int ext4_mark_dquot_dirty(struct dquot *dquot);
1306 static int ext4_write_info(struct super_block *sb, int type);
1307 static int ext4_quota_on(struct super_block *sb, int type, int format_id,
1308 const struct path *path);
1309 static int ext4_quota_on_mount(struct super_block *sb, int type);
1310 static ssize_t ext4_quota_read(struct super_block *sb, int type, char *data,
1311 size_t len, loff_t off);
1312 static ssize_t ext4_quota_write(struct super_block *sb, int type,
1313 const char *data, size_t len, loff_t off);
1314 static int ext4_quota_enable(struct super_block *sb, int type, int format_id,
1315 unsigned int flags);
1316 static int ext4_enable_quotas(struct super_block *sb);
1317 static int ext4_get_next_id(struct super_block *sb, struct kqid *qid);
1319 static struct dquot **ext4_get_dquots(struct inode *inode)
1321 return EXT4_I(inode)->i_dquot;
1324 static const struct dquot_operations ext4_quota_operations = {
1325 .get_reserved_space = ext4_get_reserved_space,
1326 .write_dquot = ext4_write_dquot,
1327 .acquire_dquot = ext4_acquire_dquot,
1328 .release_dquot = ext4_release_dquot,
1329 .mark_dirty = ext4_mark_dquot_dirty,
1330 .write_info = ext4_write_info,
1331 .alloc_dquot = dquot_alloc,
1332 .destroy_dquot = dquot_destroy,
1333 .get_projid = ext4_get_projid,
1334 .get_inode_usage = ext4_get_inode_usage,
1335 .get_next_id = ext4_get_next_id,
1338 static const struct quotactl_ops ext4_qctl_operations = {
1339 .quota_on = ext4_quota_on,
1340 .quota_off = ext4_quota_off,
1341 .quota_sync = dquot_quota_sync,
1342 .get_state = dquot_get_state,
1343 .set_info = dquot_set_dqinfo,
1344 .get_dqblk = dquot_get_dqblk,
1345 .set_dqblk = dquot_set_dqblk,
1346 .get_nextdqblk = dquot_get_next_dqblk,
1350 static const struct super_operations ext4_sops = {
1351 .alloc_inode = ext4_alloc_inode,
1352 .destroy_inode = ext4_destroy_inode,
1353 .write_inode = ext4_write_inode,
1354 .dirty_inode = ext4_dirty_inode,
1355 .drop_inode = ext4_drop_inode,
1356 .evict_inode = ext4_evict_inode,
1357 .put_super = ext4_put_super,
1358 .sync_fs = ext4_sync_fs,
1359 .freeze_fs = ext4_freeze,
1360 .unfreeze_fs = ext4_unfreeze,
1361 .statfs = ext4_statfs,
1362 .remount_fs = ext4_remount,
1363 .show_options = ext4_show_options,
1365 .quota_read = ext4_quota_read,
1366 .quota_write = ext4_quota_write,
1367 .get_dquots = ext4_get_dquots,
1369 .bdev_try_to_free_page = bdev_try_to_free_page,
1372 static const struct export_operations ext4_export_ops = {
1373 .fh_to_dentry = ext4_fh_to_dentry,
1374 .fh_to_parent = ext4_fh_to_parent,
1375 .get_parent = ext4_get_parent,
1376 .commit_metadata = ext4_nfs_commit_metadata,
1380 Opt_bsd_df, Opt_minix_df, Opt_grpid, Opt_nogrpid,
1381 Opt_resgid, Opt_resuid, Opt_sb, Opt_err_cont, Opt_err_panic, Opt_err_ro,
1382 Opt_nouid32, Opt_debug, Opt_removed,
1383 Opt_user_xattr, Opt_nouser_xattr, Opt_acl, Opt_noacl,
1384 Opt_auto_da_alloc, Opt_noauto_da_alloc, Opt_noload,
1385 Opt_commit, Opt_min_batch_time, Opt_max_batch_time, Opt_journal_dev,
1386 Opt_journal_path, Opt_journal_checksum, Opt_journal_async_commit,
1387 Opt_abort, Opt_data_journal, Opt_data_ordered, Opt_data_writeback,
1388 Opt_data_err_abort, Opt_data_err_ignore, Opt_test_dummy_encryption,
1389 Opt_usrjquota, Opt_grpjquota, Opt_offusrjquota, Opt_offgrpjquota,
1390 Opt_jqfmt_vfsold, Opt_jqfmt_vfsv0, Opt_jqfmt_vfsv1, Opt_quota,
1391 Opt_noquota, Opt_barrier, Opt_nobarrier, Opt_err,
1392 Opt_usrquota, Opt_grpquota, Opt_prjquota, Opt_i_version, Opt_dax,
1393 Opt_stripe, Opt_delalloc, Opt_nodelalloc, Opt_mblk_io_submit,
1394 Opt_lazytime, Opt_nolazytime, Opt_debug_want_extra_isize,
1395 Opt_nomblk_io_submit, Opt_block_validity, Opt_noblock_validity,
1396 Opt_inode_readahead_blks, Opt_journal_ioprio,
1397 Opt_dioread_nolock, Opt_dioread_lock,
1398 Opt_discard, Opt_nodiscard, Opt_init_itable, Opt_noinit_itable,
1399 Opt_max_dir_size_kb, Opt_nojournal_checksum, Opt_nombcache,
1402 static const match_table_t tokens = {
1403 {Opt_bsd_df, "bsddf"},
1404 {Opt_minix_df, "minixdf"},
1405 {Opt_grpid, "grpid"},
1406 {Opt_grpid, "bsdgroups"},
1407 {Opt_nogrpid, "nogrpid"},
1408 {Opt_nogrpid, "sysvgroups"},
1409 {Opt_resgid, "resgid=%u"},
1410 {Opt_resuid, "resuid=%u"},
1412 {Opt_err_cont, "errors=continue"},
1413 {Opt_err_panic, "errors=panic"},
1414 {Opt_err_ro, "errors=remount-ro"},
1415 {Opt_nouid32, "nouid32"},
1416 {Opt_debug, "debug"},
1417 {Opt_removed, "oldalloc"},
1418 {Opt_removed, "orlov"},
1419 {Opt_user_xattr, "user_xattr"},
1420 {Opt_nouser_xattr, "nouser_xattr"},
1422 {Opt_noacl, "noacl"},
1423 {Opt_noload, "norecovery"},
1424 {Opt_noload, "noload"},
1425 {Opt_removed, "nobh"},
1426 {Opt_removed, "bh"},
1427 {Opt_commit, "commit=%u"},
1428 {Opt_min_batch_time, "min_batch_time=%u"},
1429 {Opt_max_batch_time, "max_batch_time=%u"},
1430 {Opt_journal_dev, "journal_dev=%u"},
1431 {Opt_journal_path, "journal_path=%s"},
1432 {Opt_journal_checksum, "journal_checksum"},
1433 {Opt_nojournal_checksum, "nojournal_checksum"},
1434 {Opt_journal_async_commit, "journal_async_commit"},
1435 {Opt_abort, "abort"},
1436 {Opt_data_journal, "data=journal"},
1437 {Opt_data_ordered, "data=ordered"},
1438 {Opt_data_writeback, "data=writeback"},
1439 {Opt_data_err_abort, "data_err=abort"},
1440 {Opt_data_err_ignore, "data_err=ignore"},
1441 {Opt_offusrjquota, "usrjquota="},
1442 {Opt_usrjquota, "usrjquota=%s"},
1443 {Opt_offgrpjquota, "grpjquota="},
1444 {Opt_grpjquota, "grpjquota=%s"},
1445 {Opt_jqfmt_vfsold, "jqfmt=vfsold"},
1446 {Opt_jqfmt_vfsv0, "jqfmt=vfsv0"},
1447 {Opt_jqfmt_vfsv1, "jqfmt=vfsv1"},
1448 {Opt_grpquota, "grpquota"},
1449 {Opt_noquota, "noquota"},
1450 {Opt_quota, "quota"},
1451 {Opt_usrquota, "usrquota"},
1452 {Opt_prjquota, "prjquota"},
1453 {Opt_barrier, "barrier=%u"},
1454 {Opt_barrier, "barrier"},
1455 {Opt_nobarrier, "nobarrier"},
1456 {Opt_i_version, "i_version"},
1458 {Opt_stripe, "stripe=%u"},
1459 {Opt_delalloc, "delalloc"},
1460 {Opt_lazytime, "lazytime"},
1461 {Opt_nolazytime, "nolazytime"},
1462 {Opt_debug_want_extra_isize, "debug_want_extra_isize=%u"},
1463 {Opt_nodelalloc, "nodelalloc"},
1464 {Opt_removed, "mblk_io_submit"},
1465 {Opt_removed, "nomblk_io_submit"},
1466 {Opt_block_validity, "block_validity"},
1467 {Opt_noblock_validity, "noblock_validity"},
1468 {Opt_inode_readahead_blks, "inode_readahead_blks=%u"},
1469 {Opt_journal_ioprio, "journal_ioprio=%u"},
1470 {Opt_auto_da_alloc, "auto_da_alloc=%u"},
1471 {Opt_auto_da_alloc, "auto_da_alloc"},
1472 {Opt_noauto_da_alloc, "noauto_da_alloc"},
1473 {Opt_dioread_nolock, "dioread_nolock"},
1474 {Opt_dioread_lock, "dioread_lock"},
1475 {Opt_discard, "discard"},
1476 {Opt_nodiscard, "nodiscard"},
1477 {Opt_init_itable, "init_itable=%u"},
1478 {Opt_init_itable, "init_itable"},
1479 {Opt_noinit_itable, "noinit_itable"},
1480 {Opt_max_dir_size_kb, "max_dir_size_kb=%u"},
1481 {Opt_test_dummy_encryption, "test_dummy_encryption"},
1482 {Opt_nombcache, "nombcache"},
1483 {Opt_nombcache, "no_mbcache"}, /* for backward compatibility */
1484 {Opt_removed, "check=none"}, /* mount option from ext2/3 */
1485 {Opt_removed, "nocheck"}, /* mount option from ext2/3 */
1486 {Opt_removed, "reservation"}, /* mount option from ext2/3 */
1487 {Opt_removed, "noreservation"}, /* mount option from ext2/3 */
1488 {Opt_removed, "journal=%u"}, /* mount option from ext2/3 */
1492 static ext4_fsblk_t get_sb_block(void **data)
1494 ext4_fsblk_t sb_block;
1495 char *options = (char *) *data;
1497 if (!options || strncmp(options, "sb=", 3) != 0)
1498 return 1; /* Default location */
1501 /* TODO: use simple_strtoll with >32bit ext4 */
1502 sb_block = simple_strtoul(options, &options, 0);
1503 if (*options && *options != ',') {
1504 printk(KERN_ERR "EXT4-fs: Invalid sb specification: %s\n",
1508 if (*options == ',')
1510 *data = (void *) options;
1515 #define DEFAULT_JOURNAL_IOPRIO (IOPRIO_PRIO_VALUE(IOPRIO_CLASS_BE, 3))
1516 static const char deprecated_msg[] =
1517 "Mount option \"%s\" will be removed by %s\n"
1518 "Contact linux-ext4@vger.kernel.org if you think we should keep it.\n";
1521 static int set_qf_name(struct super_block *sb, int qtype, substring_t *args)
1523 struct ext4_sb_info *sbi = EXT4_SB(sb);
1524 char *qname, *old_qname = get_qf_name(sb, sbi, qtype);
1527 if (sb_any_quota_loaded(sb) && !old_qname) {
1528 ext4_msg(sb, KERN_ERR,
1529 "Cannot change journaled "
1530 "quota options when quota turned on");
1533 if (ext4_has_feature_quota(sb)) {
1534 ext4_msg(sb, KERN_INFO, "Journaled quota options "
1535 "ignored when QUOTA feature is enabled");
1538 qname = match_strdup(args);
1540 ext4_msg(sb, KERN_ERR,
1541 "Not enough memory for storing quotafile name");
1545 if (strcmp(old_qname, qname) == 0)
1548 ext4_msg(sb, KERN_ERR,
1549 "%s quota file already specified",
1553 if (strchr(qname, '/')) {
1554 ext4_msg(sb, KERN_ERR,
1555 "quotafile must be on filesystem root");
1558 rcu_assign_pointer(sbi->s_qf_names[qtype], qname);
1566 static int clear_qf_name(struct super_block *sb, int qtype)
1569 struct ext4_sb_info *sbi = EXT4_SB(sb);
1570 char *old_qname = get_qf_name(sb, sbi, qtype);
1572 if (sb_any_quota_loaded(sb) && old_qname) {
1573 ext4_msg(sb, KERN_ERR, "Cannot change journaled quota options"
1574 " when quota turned on");
1577 rcu_assign_pointer(sbi->s_qf_names[qtype], NULL);
1584 #define MOPT_SET 0x0001
1585 #define MOPT_CLEAR 0x0002
1586 #define MOPT_NOSUPPORT 0x0004
1587 #define MOPT_EXPLICIT 0x0008
1588 #define MOPT_CLEAR_ERR 0x0010
1589 #define MOPT_GTE0 0x0020
1592 #define MOPT_QFMT 0x0040
1594 #define MOPT_Q MOPT_NOSUPPORT
1595 #define MOPT_QFMT MOPT_NOSUPPORT
1597 #define MOPT_DATAJ 0x0080
1598 #define MOPT_NO_EXT2 0x0100
1599 #define MOPT_NO_EXT3 0x0200
1600 #define MOPT_EXT4_ONLY (MOPT_NO_EXT2 | MOPT_NO_EXT3)
1601 #define MOPT_STRING 0x0400
1603 static const struct mount_opts {
1607 } ext4_mount_opts[] = {
1608 {Opt_minix_df, EXT4_MOUNT_MINIX_DF, MOPT_SET},
1609 {Opt_bsd_df, EXT4_MOUNT_MINIX_DF, MOPT_CLEAR},
1610 {Opt_grpid, EXT4_MOUNT_GRPID, MOPT_SET},
1611 {Opt_nogrpid, EXT4_MOUNT_GRPID, MOPT_CLEAR},
1612 {Opt_block_validity, EXT4_MOUNT_BLOCK_VALIDITY, MOPT_SET},
1613 {Opt_noblock_validity, EXT4_MOUNT_BLOCK_VALIDITY, MOPT_CLEAR},
1614 {Opt_dioread_nolock, EXT4_MOUNT_DIOREAD_NOLOCK,
1615 MOPT_EXT4_ONLY | MOPT_SET},
1616 {Opt_dioread_lock, EXT4_MOUNT_DIOREAD_NOLOCK,
1617 MOPT_EXT4_ONLY | MOPT_CLEAR},
1618 {Opt_discard, EXT4_MOUNT_DISCARD, MOPT_SET},
1619 {Opt_nodiscard, EXT4_MOUNT_DISCARD, MOPT_CLEAR},
1620 {Opt_delalloc, EXT4_MOUNT_DELALLOC,
1621 MOPT_EXT4_ONLY | MOPT_SET | MOPT_EXPLICIT},
1622 {Opt_nodelalloc, EXT4_MOUNT_DELALLOC,
1623 MOPT_EXT4_ONLY | MOPT_CLEAR},
1624 {Opt_nojournal_checksum, EXT4_MOUNT_JOURNAL_CHECKSUM,
1625 MOPT_EXT4_ONLY | MOPT_CLEAR},
1626 {Opt_journal_checksum, EXT4_MOUNT_JOURNAL_CHECKSUM,
1627 MOPT_EXT4_ONLY | MOPT_SET | MOPT_EXPLICIT},
1628 {Opt_journal_async_commit, (EXT4_MOUNT_JOURNAL_ASYNC_COMMIT |
1629 EXT4_MOUNT_JOURNAL_CHECKSUM),
1630 MOPT_EXT4_ONLY | MOPT_SET | MOPT_EXPLICIT},
1631 {Opt_noload, EXT4_MOUNT_NOLOAD, MOPT_NO_EXT2 | MOPT_SET},
1632 {Opt_err_panic, EXT4_MOUNT_ERRORS_PANIC, MOPT_SET | MOPT_CLEAR_ERR},
1633 {Opt_err_ro, EXT4_MOUNT_ERRORS_RO, MOPT_SET | MOPT_CLEAR_ERR},
1634 {Opt_err_cont, EXT4_MOUNT_ERRORS_CONT, MOPT_SET | MOPT_CLEAR_ERR},
1635 {Opt_data_err_abort, EXT4_MOUNT_DATA_ERR_ABORT,
1637 {Opt_data_err_ignore, EXT4_MOUNT_DATA_ERR_ABORT,
1639 {Opt_barrier, EXT4_MOUNT_BARRIER, MOPT_SET},
1640 {Opt_nobarrier, EXT4_MOUNT_BARRIER, MOPT_CLEAR},
1641 {Opt_noauto_da_alloc, EXT4_MOUNT_NO_AUTO_DA_ALLOC, MOPT_SET},
1642 {Opt_auto_da_alloc, EXT4_MOUNT_NO_AUTO_DA_ALLOC, MOPT_CLEAR},
1643 {Opt_noinit_itable, EXT4_MOUNT_INIT_INODE_TABLE, MOPT_CLEAR},
1644 {Opt_commit, 0, MOPT_GTE0},
1645 {Opt_max_batch_time, 0, MOPT_GTE0},
1646 {Opt_min_batch_time, 0, MOPT_GTE0},
1647 {Opt_inode_readahead_blks, 0, MOPT_GTE0},
1648 {Opt_init_itable, 0, MOPT_GTE0},
1649 {Opt_dax, EXT4_MOUNT_DAX, MOPT_SET},
1650 {Opt_stripe, 0, MOPT_GTE0},
1651 {Opt_resuid, 0, MOPT_GTE0},
1652 {Opt_resgid, 0, MOPT_GTE0},
1653 {Opt_journal_dev, 0, MOPT_NO_EXT2 | MOPT_GTE0},
1654 {Opt_journal_path, 0, MOPT_NO_EXT2 | MOPT_STRING},
1655 {Opt_journal_ioprio, 0, MOPT_NO_EXT2 | MOPT_GTE0},
1656 {Opt_data_journal, EXT4_MOUNT_JOURNAL_DATA, MOPT_NO_EXT2 | MOPT_DATAJ},
1657 {Opt_data_ordered, EXT4_MOUNT_ORDERED_DATA, MOPT_NO_EXT2 | MOPT_DATAJ},
1658 {Opt_data_writeback, EXT4_MOUNT_WRITEBACK_DATA,
1659 MOPT_NO_EXT2 | MOPT_DATAJ},
1660 {Opt_user_xattr, EXT4_MOUNT_XATTR_USER, MOPT_SET},
1661 {Opt_nouser_xattr, EXT4_MOUNT_XATTR_USER, MOPT_CLEAR},
1662 #ifdef CONFIG_EXT4_FS_POSIX_ACL
1663 {Opt_acl, EXT4_MOUNT_POSIX_ACL, MOPT_SET},
1664 {Opt_noacl, EXT4_MOUNT_POSIX_ACL, MOPT_CLEAR},
1666 {Opt_acl, 0, MOPT_NOSUPPORT},
1667 {Opt_noacl, 0, MOPT_NOSUPPORT},
1669 {Opt_nouid32, EXT4_MOUNT_NO_UID32, MOPT_SET},
1670 {Opt_debug, EXT4_MOUNT_DEBUG, MOPT_SET},
1671 {Opt_debug_want_extra_isize, 0, MOPT_GTE0},
1672 {Opt_quota, EXT4_MOUNT_QUOTA | EXT4_MOUNT_USRQUOTA, MOPT_SET | MOPT_Q},
1673 {Opt_usrquota, EXT4_MOUNT_QUOTA | EXT4_MOUNT_USRQUOTA,
1675 {Opt_grpquota, EXT4_MOUNT_QUOTA | EXT4_MOUNT_GRPQUOTA,
1677 {Opt_prjquota, EXT4_MOUNT_QUOTA | EXT4_MOUNT_PRJQUOTA,
1679 {Opt_noquota, (EXT4_MOUNT_QUOTA | EXT4_MOUNT_USRQUOTA |
1680 EXT4_MOUNT_GRPQUOTA | EXT4_MOUNT_PRJQUOTA),
1681 MOPT_CLEAR | MOPT_Q},
1682 {Opt_usrjquota, 0, MOPT_Q | MOPT_STRING},
1683 {Opt_grpjquota, 0, MOPT_Q | MOPT_STRING},
1684 {Opt_offusrjquota, 0, MOPT_Q},
1685 {Opt_offgrpjquota, 0, MOPT_Q},
1686 {Opt_jqfmt_vfsold, QFMT_VFS_OLD, MOPT_QFMT},
1687 {Opt_jqfmt_vfsv0, QFMT_VFS_V0, MOPT_QFMT},
1688 {Opt_jqfmt_vfsv1, QFMT_VFS_V1, MOPT_QFMT},
1689 {Opt_max_dir_size_kb, 0, MOPT_GTE0},
1690 {Opt_test_dummy_encryption, 0, MOPT_GTE0},
1691 {Opt_nombcache, EXT4_MOUNT_NO_MBCACHE, MOPT_SET},
1695 static int handle_mount_opt(struct super_block *sb, char *opt, int token,
1696 substring_t *args, unsigned long *journal_devnum,
1697 unsigned int *journal_ioprio, int is_remount)
1699 struct ext4_sb_info *sbi = EXT4_SB(sb);
1700 const struct mount_opts *m;
1706 if (token == Opt_usrjquota)
1707 return set_qf_name(sb, USRQUOTA, &args[0]);
1708 else if (token == Opt_grpjquota)
1709 return set_qf_name(sb, GRPQUOTA, &args[0]);
1710 else if (token == Opt_offusrjquota)
1711 return clear_qf_name(sb, USRQUOTA);
1712 else if (token == Opt_offgrpjquota)
1713 return clear_qf_name(sb, GRPQUOTA);
1717 case Opt_nouser_xattr:
1718 ext4_msg(sb, KERN_WARNING, deprecated_msg, opt, "3.5");
1721 return 1; /* handled by get_sb_block() */
1723 ext4_msg(sb, KERN_WARNING, "Ignoring removed %s option", opt);
1726 sbi->s_mount_flags |= EXT4_MF_FS_ABORTED;
1729 sb->s_flags |= SB_I_VERSION;
1732 sb->s_flags |= MS_LAZYTIME;
1734 case Opt_nolazytime:
1735 sb->s_flags &= ~MS_LAZYTIME;
1739 for (m = ext4_mount_opts; m->token != Opt_err; m++)
1740 if (token == m->token)
1743 if (m->token == Opt_err) {
1744 ext4_msg(sb, KERN_ERR, "Unrecognized mount option \"%s\" "
1745 "or missing value", opt);
1749 if ((m->flags & MOPT_NO_EXT2) && IS_EXT2_SB(sb)) {
1750 ext4_msg(sb, KERN_ERR,
1751 "Mount option \"%s\" incompatible with ext2", opt);
1754 if ((m->flags & MOPT_NO_EXT3) && IS_EXT3_SB(sb)) {
1755 ext4_msg(sb, KERN_ERR,
1756 "Mount option \"%s\" incompatible with ext3", opt);
1760 if (args->from && !(m->flags & MOPT_STRING) && match_int(args, &arg))
1762 if (args->from && (m->flags & MOPT_GTE0) && (arg < 0))
1764 if (m->flags & MOPT_EXPLICIT) {
1765 if (m->mount_opt & EXT4_MOUNT_DELALLOC) {
1766 set_opt2(sb, EXPLICIT_DELALLOC);
1767 } else if (m->mount_opt & EXT4_MOUNT_JOURNAL_CHECKSUM) {
1768 set_opt2(sb, EXPLICIT_JOURNAL_CHECKSUM);
1772 if (m->flags & MOPT_CLEAR_ERR)
1773 clear_opt(sb, ERRORS_MASK);
1774 if (token == Opt_noquota && sb_any_quota_loaded(sb)) {
1775 ext4_msg(sb, KERN_ERR, "Cannot change quota "
1776 "options when quota turned on");
1780 if (m->flags & MOPT_NOSUPPORT) {
1781 ext4_msg(sb, KERN_ERR, "%s option not supported", opt);
1782 } else if (token == Opt_commit) {
1784 arg = JBD2_DEFAULT_MAX_COMMIT_AGE;
1785 sbi->s_commit_interval = HZ * arg;
1786 } else if (token == Opt_debug_want_extra_isize) {
1789 (arg > (sbi->s_inode_size - EXT4_GOOD_OLD_INODE_SIZE))) {
1790 ext4_msg(sb, KERN_ERR,
1791 "Invalid want_extra_isize %d", arg);
1794 sbi->s_want_extra_isize = arg;
1795 } else if (token == Opt_max_batch_time) {
1796 sbi->s_max_batch_time = arg;
1797 } else if (token == Opt_min_batch_time) {
1798 sbi->s_min_batch_time = arg;
1799 } else if (token == Opt_inode_readahead_blks) {
1800 if (arg && (arg > (1 << 30) || !is_power_of_2(arg))) {
1801 ext4_msg(sb, KERN_ERR,
1802 "EXT4-fs: inode_readahead_blks must be "
1803 "0 or a power of 2 smaller than 2^31");
1806 sbi->s_inode_readahead_blks = arg;
1807 } else if (token == Opt_init_itable) {
1808 set_opt(sb, INIT_INODE_TABLE);
1810 arg = EXT4_DEF_LI_WAIT_MULT;
1811 sbi->s_li_wait_mult = arg;
1812 } else if (token == Opt_max_dir_size_kb) {
1813 sbi->s_max_dir_size_kb = arg;
1814 } else if (token == Opt_stripe) {
1815 sbi->s_stripe = arg;
1816 } else if (token == Opt_resuid) {
1817 uid = make_kuid(current_user_ns(), arg);
1818 if (!uid_valid(uid)) {
1819 ext4_msg(sb, KERN_ERR, "Invalid uid value %d", arg);
1822 sbi->s_resuid = uid;
1823 } else if (token == Opt_resgid) {
1824 gid = make_kgid(current_user_ns(), arg);
1825 if (!gid_valid(gid)) {
1826 ext4_msg(sb, KERN_ERR, "Invalid gid value %d", arg);
1829 sbi->s_resgid = gid;
1830 } else if (token == Opt_journal_dev) {
1832 ext4_msg(sb, KERN_ERR,
1833 "Cannot specify journal on remount");
1836 *journal_devnum = arg;
1837 } else if (token == Opt_journal_path) {
1839 struct inode *journal_inode;
1844 ext4_msg(sb, KERN_ERR,
1845 "Cannot specify journal on remount");
1848 journal_path = match_strdup(&args[0]);
1849 if (!journal_path) {
1850 ext4_msg(sb, KERN_ERR, "error: could not dup "
1851 "journal device string");
1855 error = kern_path(journal_path, LOOKUP_FOLLOW, &path);
1857 ext4_msg(sb, KERN_ERR, "error: could not find "
1858 "journal device path: error %d", error);
1859 kfree(journal_path);
1863 journal_inode = d_inode(path.dentry);
1864 if (!S_ISBLK(journal_inode->i_mode)) {
1865 ext4_msg(sb, KERN_ERR, "error: journal path %s "
1866 "is not a block device", journal_path);
1868 kfree(journal_path);
1872 *journal_devnum = new_encode_dev(journal_inode->i_rdev);
1874 kfree(journal_path);
1875 } else if (token == Opt_journal_ioprio) {
1877 ext4_msg(sb, KERN_ERR, "Invalid journal IO priority"
1882 IOPRIO_PRIO_VALUE(IOPRIO_CLASS_BE, arg);
1883 } else if (token == Opt_test_dummy_encryption) {
1884 #ifdef CONFIG_EXT4_FS_ENCRYPTION
1885 sbi->s_mount_flags |= EXT4_MF_TEST_DUMMY_ENCRYPTION;
1886 ext4_msg(sb, KERN_WARNING,
1887 "Test dummy encryption mode enabled");
1889 ext4_msg(sb, KERN_WARNING,
1890 "Test dummy encryption mount option ignored");
1892 } else if (m->flags & MOPT_DATAJ) {
1894 if (!sbi->s_journal)
1895 ext4_msg(sb, KERN_WARNING, "Remounting file system with no journal so ignoring journalled data option");
1896 else if (test_opt(sb, DATA_FLAGS) != m->mount_opt) {
1897 ext4_msg(sb, KERN_ERR,
1898 "Cannot change data mode on remount");
1902 clear_opt(sb, DATA_FLAGS);
1903 sbi->s_mount_opt |= m->mount_opt;
1906 } else if (m->flags & MOPT_QFMT) {
1907 if (sb_any_quota_loaded(sb) &&
1908 sbi->s_jquota_fmt != m->mount_opt) {
1909 ext4_msg(sb, KERN_ERR, "Cannot change journaled "
1910 "quota options when quota turned on");
1913 if (ext4_has_feature_quota(sb)) {
1914 ext4_msg(sb, KERN_INFO,
1915 "Quota format mount options ignored "
1916 "when QUOTA feature is enabled");
1919 sbi->s_jquota_fmt = m->mount_opt;
1921 } else if (token == Opt_dax) {
1922 #ifdef CONFIG_FS_DAX
1923 ext4_msg(sb, KERN_WARNING,
1924 "DAX enabled. Warning: EXPERIMENTAL, use at your own risk");
1925 sbi->s_mount_opt |= m->mount_opt;
1927 ext4_msg(sb, KERN_INFO, "dax option not supported");
1930 } else if (token == Opt_data_err_abort) {
1931 sbi->s_mount_opt |= m->mount_opt;
1932 } else if (token == Opt_data_err_ignore) {
1933 sbi->s_mount_opt &= ~m->mount_opt;
1937 if (m->flags & MOPT_CLEAR)
1939 else if (unlikely(!(m->flags & MOPT_SET))) {
1940 ext4_msg(sb, KERN_WARNING,
1941 "buggy handling of option %s", opt);
1946 sbi->s_mount_opt |= m->mount_opt;
1948 sbi->s_mount_opt &= ~m->mount_opt;
1953 static int parse_options(char *options, struct super_block *sb,
1954 unsigned long *journal_devnum,
1955 unsigned int *journal_ioprio,
1958 struct ext4_sb_info *sbi = EXT4_SB(sb);
1959 char *p, __maybe_unused *usr_qf_name, __maybe_unused *grp_qf_name;
1960 substring_t args[MAX_OPT_ARGS];
1966 while ((p = strsep(&options, ",")) != NULL) {
1970 * Initialize args struct so we know whether arg was
1971 * found; some options take optional arguments.
1973 args[0].to = args[0].from = NULL;
1974 token = match_token(p, tokens, args);
1975 if (handle_mount_opt(sb, p, token, args, journal_devnum,
1976 journal_ioprio, is_remount) < 0)
1981 * We do the test below only for project quotas. 'usrquota' and
1982 * 'grpquota' mount options are allowed even without quota feature
1983 * to support legacy quotas in quota files.
1985 if (test_opt(sb, PRJQUOTA) && !ext4_has_feature_project(sb)) {
1986 ext4_msg(sb, KERN_ERR, "Project quota feature not enabled. "
1987 "Cannot enable project quota enforcement.");
1990 usr_qf_name = get_qf_name(sb, sbi, USRQUOTA);
1991 grp_qf_name = get_qf_name(sb, sbi, GRPQUOTA);
1992 if (usr_qf_name || grp_qf_name) {
1993 if (test_opt(sb, USRQUOTA) && usr_qf_name)
1994 clear_opt(sb, USRQUOTA);
1996 if (test_opt(sb, GRPQUOTA) && grp_qf_name)
1997 clear_opt(sb, GRPQUOTA);
1999 if (test_opt(sb, GRPQUOTA) || test_opt(sb, USRQUOTA)) {
2000 ext4_msg(sb, KERN_ERR, "old and new quota "
2005 if (!sbi->s_jquota_fmt) {
2006 ext4_msg(sb, KERN_ERR, "journaled quota format "
2012 if (test_opt(sb, DIOREAD_NOLOCK)) {
2014 BLOCK_SIZE << le32_to_cpu(sbi->s_es->s_log_block_size);
2016 if (blocksize < PAGE_SIZE) {
2017 ext4_msg(sb, KERN_ERR, "can't mount with "
2018 "dioread_nolock if block size != PAGE_SIZE");
2025 static inline void ext4_show_quota_options(struct seq_file *seq,
2026 struct super_block *sb)
2028 #if defined(CONFIG_QUOTA)
2029 struct ext4_sb_info *sbi = EXT4_SB(sb);
2030 char *usr_qf_name, *grp_qf_name;
2032 if (sbi->s_jquota_fmt) {
2035 switch (sbi->s_jquota_fmt) {
2046 seq_printf(seq, ",jqfmt=%s", fmtname);
2050 usr_qf_name = rcu_dereference(sbi->s_qf_names[USRQUOTA]);
2051 grp_qf_name = rcu_dereference(sbi->s_qf_names[GRPQUOTA]);
2053 seq_show_option(seq, "usrjquota", usr_qf_name);
2055 seq_show_option(seq, "grpjquota", grp_qf_name);
2060 static const char *token2str(int token)
2062 const struct match_token *t;
2064 for (t = tokens; t->token != Opt_err; t++)
2065 if (t->token == token && !strchr(t->pattern, '='))
2072 * - it's set to a non-default value OR
2073 * - if the per-sb default is different from the global default
2075 static int _ext4_show_options(struct seq_file *seq, struct super_block *sb,
2078 struct ext4_sb_info *sbi = EXT4_SB(sb);
2079 struct ext4_super_block *es = sbi->s_es;
2080 int def_errors, def_mount_opt = nodefs ? 0 : sbi->s_def_mount_opt;
2081 const struct mount_opts *m;
2082 char sep = nodefs ? '\n' : ',';
2084 #define SEQ_OPTS_PUTS(str) seq_printf(seq, "%c" str, sep)
2085 #define SEQ_OPTS_PRINT(str, arg) seq_printf(seq, "%c" str, sep, arg)
2087 if (sbi->s_sb_block != 1)
2088 SEQ_OPTS_PRINT("sb=%llu", sbi->s_sb_block);
2090 for (m = ext4_mount_opts; m->token != Opt_err; m++) {
2091 int want_set = m->flags & MOPT_SET;
2092 if (((m->flags & (MOPT_SET|MOPT_CLEAR)) == 0) ||
2093 (m->flags & MOPT_CLEAR_ERR))
2095 if (!(m->mount_opt & (sbi->s_mount_opt ^ def_mount_opt)))
2096 continue; /* skip if same as the default */
2098 (sbi->s_mount_opt & m->mount_opt) != m->mount_opt) ||
2099 (!want_set && (sbi->s_mount_opt & m->mount_opt)))
2100 continue; /* select Opt_noFoo vs Opt_Foo */
2101 SEQ_OPTS_PRINT("%s", token2str(m->token));
2104 if (nodefs || !uid_eq(sbi->s_resuid, make_kuid(&init_user_ns, EXT4_DEF_RESUID)) ||
2105 le16_to_cpu(es->s_def_resuid) != EXT4_DEF_RESUID)
2106 SEQ_OPTS_PRINT("resuid=%u",
2107 from_kuid_munged(&init_user_ns, sbi->s_resuid));
2108 if (nodefs || !gid_eq(sbi->s_resgid, make_kgid(&init_user_ns, EXT4_DEF_RESGID)) ||
2109 le16_to_cpu(es->s_def_resgid) != EXT4_DEF_RESGID)
2110 SEQ_OPTS_PRINT("resgid=%u",
2111 from_kgid_munged(&init_user_ns, sbi->s_resgid));
2112 def_errors = nodefs ? -1 : le16_to_cpu(es->s_errors);
2113 if (test_opt(sb, ERRORS_RO) && def_errors != EXT4_ERRORS_RO)
2114 SEQ_OPTS_PUTS("errors=remount-ro");
2115 if (test_opt(sb, ERRORS_CONT) && def_errors != EXT4_ERRORS_CONTINUE)
2116 SEQ_OPTS_PUTS("errors=continue");
2117 if (test_opt(sb, ERRORS_PANIC) && def_errors != EXT4_ERRORS_PANIC)
2118 SEQ_OPTS_PUTS("errors=panic");
2119 if (nodefs || sbi->s_commit_interval != JBD2_DEFAULT_MAX_COMMIT_AGE*HZ)
2120 SEQ_OPTS_PRINT("commit=%lu", sbi->s_commit_interval / HZ);
2121 if (nodefs || sbi->s_min_batch_time != EXT4_DEF_MIN_BATCH_TIME)
2122 SEQ_OPTS_PRINT("min_batch_time=%u", sbi->s_min_batch_time);
2123 if (nodefs || sbi->s_max_batch_time != EXT4_DEF_MAX_BATCH_TIME)
2124 SEQ_OPTS_PRINT("max_batch_time=%u", sbi->s_max_batch_time);
2125 if (sb->s_flags & SB_I_VERSION)
2126 SEQ_OPTS_PUTS("i_version");
2127 if (nodefs || sbi->s_stripe)
2128 SEQ_OPTS_PRINT("stripe=%lu", sbi->s_stripe);
2129 if (EXT4_MOUNT_DATA_FLAGS & (sbi->s_mount_opt ^ def_mount_opt)) {
2130 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA)
2131 SEQ_OPTS_PUTS("data=journal");
2132 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA)
2133 SEQ_OPTS_PUTS("data=ordered");
2134 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_WRITEBACK_DATA)
2135 SEQ_OPTS_PUTS("data=writeback");
2138 sbi->s_inode_readahead_blks != EXT4_DEF_INODE_READAHEAD_BLKS)
2139 SEQ_OPTS_PRINT("inode_readahead_blks=%u",
2140 sbi->s_inode_readahead_blks);
2142 if (nodefs || (test_opt(sb, INIT_INODE_TABLE) &&
2143 (sbi->s_li_wait_mult != EXT4_DEF_LI_WAIT_MULT)))
2144 SEQ_OPTS_PRINT("init_itable=%u", sbi->s_li_wait_mult);
2145 if (nodefs || sbi->s_max_dir_size_kb)
2146 SEQ_OPTS_PRINT("max_dir_size_kb=%u", sbi->s_max_dir_size_kb);
2147 if (test_opt(sb, DATA_ERR_ABORT))
2148 SEQ_OPTS_PUTS("data_err=abort");
2149 if (DUMMY_ENCRYPTION_ENABLED(sbi))
2150 SEQ_OPTS_PUTS("test_dummy_encryption");
2152 ext4_show_quota_options(seq, sb);
2156 static int ext4_show_options(struct seq_file *seq, struct dentry *root)
2158 return _ext4_show_options(seq, root->d_sb, 0);
2161 int ext4_seq_options_show(struct seq_file *seq, void *offset)
2163 struct super_block *sb = seq->private;
2166 seq_puts(seq, sb_rdonly(sb) ? "ro" : "rw");
2167 rc = _ext4_show_options(seq, sb, 1);
2168 seq_puts(seq, "\n");
2172 static int ext4_setup_super(struct super_block *sb, struct ext4_super_block *es,
2175 struct ext4_sb_info *sbi = EXT4_SB(sb);
2178 if (le32_to_cpu(es->s_rev_level) > EXT4_MAX_SUPP_REV) {
2179 ext4_msg(sb, KERN_ERR, "revision level too high, "
2180 "forcing read-only mode");
2185 if (!(sbi->s_mount_state & EXT4_VALID_FS))
2186 ext4_msg(sb, KERN_WARNING, "warning: mounting unchecked fs, "
2187 "running e2fsck is recommended");
2188 else if (sbi->s_mount_state & EXT4_ERROR_FS)
2189 ext4_msg(sb, KERN_WARNING,
2190 "warning: mounting fs with errors, "
2191 "running e2fsck is recommended");
2192 else if ((__s16) le16_to_cpu(es->s_max_mnt_count) > 0 &&
2193 le16_to_cpu(es->s_mnt_count) >=
2194 (unsigned short) (__s16) le16_to_cpu(es->s_max_mnt_count))
2195 ext4_msg(sb, KERN_WARNING,
2196 "warning: maximal mount count reached, "
2197 "running e2fsck is recommended");
2198 else if (le32_to_cpu(es->s_checkinterval) &&
2199 (le32_to_cpu(es->s_lastcheck) +
2200 le32_to_cpu(es->s_checkinterval) <= get_seconds()))
2201 ext4_msg(sb, KERN_WARNING,
2202 "warning: checktime reached, "
2203 "running e2fsck is recommended");
2204 if (!sbi->s_journal)
2205 es->s_state &= cpu_to_le16(~EXT4_VALID_FS);
2206 if (!(__s16) le16_to_cpu(es->s_max_mnt_count))
2207 es->s_max_mnt_count = cpu_to_le16(EXT4_DFL_MAX_MNT_COUNT);
2208 le16_add_cpu(&es->s_mnt_count, 1);
2209 es->s_mtime = cpu_to_le32(get_seconds());
2210 ext4_update_dynamic_rev(sb);
2212 ext4_set_feature_journal_needs_recovery(sb);
2214 ext4_commit_super(sb, 1);
2216 if (test_opt(sb, DEBUG))
2217 printk(KERN_INFO "[EXT4 FS bs=%lu, gc=%u, "
2218 "bpg=%lu, ipg=%lu, mo=%04x, mo2=%04x]\n",
2220 sbi->s_groups_count,
2221 EXT4_BLOCKS_PER_GROUP(sb),
2222 EXT4_INODES_PER_GROUP(sb),
2223 sbi->s_mount_opt, sbi->s_mount_opt2);
2225 cleancache_init_fs(sb);
2229 int ext4_alloc_flex_bg_array(struct super_block *sb, ext4_group_t ngroup)
2231 struct ext4_sb_info *sbi = EXT4_SB(sb);
2232 struct flex_groups **old_groups, **new_groups;
2235 if (!sbi->s_log_groups_per_flex)
2238 size = ext4_flex_group(sbi, ngroup - 1) + 1;
2239 if (size <= sbi->s_flex_groups_allocated)
2242 new_groups = kvzalloc(roundup_pow_of_two(size *
2243 sizeof(*sbi->s_flex_groups)), GFP_KERNEL);
2245 ext4_msg(sb, KERN_ERR,
2246 "not enough memory for %d flex group pointers", size);
2249 for (i = sbi->s_flex_groups_allocated; i < size; i++) {
2250 new_groups[i] = kvzalloc(roundup_pow_of_two(
2251 sizeof(struct flex_groups)),
2253 if (!new_groups[i]) {
2254 for (j = sbi->s_flex_groups_allocated; j < i; j++)
2255 kvfree(new_groups[j]);
2257 ext4_msg(sb, KERN_ERR,
2258 "not enough memory for %d flex groups", size);
2263 old_groups = rcu_dereference(sbi->s_flex_groups);
2265 memcpy(new_groups, old_groups,
2266 (sbi->s_flex_groups_allocated *
2267 sizeof(struct flex_groups *)));
2269 rcu_assign_pointer(sbi->s_flex_groups, new_groups);
2270 sbi->s_flex_groups_allocated = size;
2272 ext4_kvfree_array_rcu(old_groups);
2276 static int ext4_fill_flex_info(struct super_block *sb)
2278 struct ext4_sb_info *sbi = EXT4_SB(sb);
2279 struct ext4_group_desc *gdp = NULL;
2280 struct flex_groups *fg;
2281 ext4_group_t flex_group;
2284 sbi->s_log_groups_per_flex = sbi->s_es->s_log_groups_per_flex;
2285 if (sbi->s_log_groups_per_flex < 1 || sbi->s_log_groups_per_flex > 31) {
2286 sbi->s_log_groups_per_flex = 0;
2290 err = ext4_alloc_flex_bg_array(sb, sbi->s_groups_count);
2294 for (i = 0; i < sbi->s_groups_count; i++) {
2295 gdp = ext4_get_group_desc(sb, i, NULL);
2297 flex_group = ext4_flex_group(sbi, i);
2298 fg = sbi_array_rcu_deref(sbi, s_flex_groups, flex_group);
2299 atomic_add(ext4_free_inodes_count(sb, gdp), &fg->free_inodes);
2300 atomic64_add(ext4_free_group_clusters(sb, gdp),
2301 &fg->free_clusters);
2302 atomic_add(ext4_used_dirs_count(sb, gdp), &fg->used_dirs);
2310 static __le16 ext4_group_desc_csum(struct super_block *sb, __u32 block_group,
2311 struct ext4_group_desc *gdp)
2313 int offset = offsetof(struct ext4_group_desc, bg_checksum);
2315 __le32 le_group = cpu_to_le32(block_group);
2316 struct ext4_sb_info *sbi = EXT4_SB(sb);
2318 if (ext4_has_metadata_csum(sbi->s_sb)) {
2319 /* Use new metadata_csum algorithm */
2321 __u16 dummy_csum = 0;
2323 csum32 = ext4_chksum(sbi, sbi->s_csum_seed, (__u8 *)&le_group,
2325 csum32 = ext4_chksum(sbi, csum32, (__u8 *)gdp, offset);
2326 csum32 = ext4_chksum(sbi, csum32, (__u8 *)&dummy_csum,
2327 sizeof(dummy_csum));
2328 offset += sizeof(dummy_csum);
2329 if (offset < sbi->s_desc_size)
2330 csum32 = ext4_chksum(sbi, csum32, (__u8 *)gdp + offset,
2331 sbi->s_desc_size - offset);
2333 crc = csum32 & 0xFFFF;
2337 /* old crc16 code */
2338 if (!ext4_has_feature_gdt_csum(sb))
2341 crc = crc16(~0, sbi->s_es->s_uuid, sizeof(sbi->s_es->s_uuid));
2342 crc = crc16(crc, (__u8 *)&le_group, sizeof(le_group));
2343 crc = crc16(crc, (__u8 *)gdp, offset);
2344 offset += sizeof(gdp->bg_checksum); /* skip checksum */
2345 /* for checksum of struct ext4_group_desc do the rest...*/
2346 if (ext4_has_feature_64bit(sb) &&
2347 offset < le16_to_cpu(sbi->s_es->s_desc_size))
2348 crc = crc16(crc, (__u8 *)gdp + offset,
2349 le16_to_cpu(sbi->s_es->s_desc_size) -
2353 return cpu_to_le16(crc);
2356 int ext4_group_desc_csum_verify(struct super_block *sb, __u32 block_group,
2357 struct ext4_group_desc *gdp)
2359 if (ext4_has_group_desc_csum(sb) &&
2360 (gdp->bg_checksum != ext4_group_desc_csum(sb, block_group, gdp)))
2366 void ext4_group_desc_csum_set(struct super_block *sb, __u32 block_group,
2367 struct ext4_group_desc *gdp)
2369 if (!ext4_has_group_desc_csum(sb))
2371 gdp->bg_checksum = ext4_group_desc_csum(sb, block_group, gdp);
2374 /* Called at mount-time, super-block is locked */
2375 static int ext4_check_descriptors(struct super_block *sb,
2376 ext4_fsblk_t sb_block,
2377 ext4_group_t *first_not_zeroed)
2379 struct ext4_sb_info *sbi = EXT4_SB(sb);
2380 ext4_fsblk_t first_block = le32_to_cpu(sbi->s_es->s_first_data_block);
2381 ext4_fsblk_t last_block;
2382 ext4_fsblk_t last_bg_block = sb_block + ext4_bg_num_gdb(sb, 0);
2383 ext4_fsblk_t block_bitmap;
2384 ext4_fsblk_t inode_bitmap;
2385 ext4_fsblk_t inode_table;
2386 int flexbg_flag = 0;
2387 ext4_group_t i, grp = sbi->s_groups_count;
2389 if (ext4_has_feature_flex_bg(sb))
2392 ext4_debug("Checking group descriptors");
2394 for (i = 0; i < sbi->s_groups_count; i++) {
2395 struct ext4_group_desc *gdp = ext4_get_group_desc(sb, i, NULL);
2397 if (i == sbi->s_groups_count - 1 || flexbg_flag)
2398 last_block = ext4_blocks_count(sbi->s_es) - 1;
2400 last_block = first_block +
2401 (EXT4_BLOCKS_PER_GROUP(sb) - 1);
2403 if ((grp == sbi->s_groups_count) &&
2404 !(gdp->bg_flags & cpu_to_le16(EXT4_BG_INODE_ZEROED)))
2407 block_bitmap = ext4_block_bitmap(sb, gdp);
2408 if (block_bitmap == sb_block) {
2409 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2410 "Block bitmap for group %u overlaps "
2415 if (block_bitmap >= sb_block + 1 &&
2416 block_bitmap <= last_bg_block) {
2417 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2418 "Block bitmap for group %u overlaps "
2419 "block group descriptors", i);
2423 if (block_bitmap < first_block || block_bitmap > last_block) {
2424 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2425 "Block bitmap for group %u not in group "
2426 "(block %llu)!", i, block_bitmap);
2429 inode_bitmap = ext4_inode_bitmap(sb, gdp);
2430 if (inode_bitmap == sb_block) {
2431 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2432 "Inode bitmap for group %u overlaps "
2437 if (inode_bitmap >= sb_block + 1 &&
2438 inode_bitmap <= last_bg_block) {
2439 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2440 "Inode bitmap for group %u overlaps "
2441 "block group descriptors", i);
2445 if (inode_bitmap < first_block || inode_bitmap > last_block) {
2446 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2447 "Inode bitmap for group %u not in group "
2448 "(block %llu)!", i, inode_bitmap);
2451 inode_table = ext4_inode_table(sb, gdp);
2452 if (inode_table == sb_block) {
2453 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2454 "Inode table for group %u overlaps "
2459 if (inode_table >= sb_block + 1 &&
2460 inode_table <= last_bg_block) {
2461 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2462 "Inode table for group %u overlaps "
2463 "block group descriptors", i);
2467 if (inode_table < first_block ||
2468 inode_table + sbi->s_itb_per_group - 1 > last_block) {
2469 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2470 "Inode table for group %u not in group "
2471 "(block %llu)!", i, inode_table);
2474 ext4_lock_group(sb, i);
2475 if (!ext4_group_desc_csum_verify(sb, i, gdp)) {
2476 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2477 "Checksum for group %u failed (%u!=%u)",
2478 i, le16_to_cpu(ext4_group_desc_csum(sb, i,
2479 gdp)), le16_to_cpu(gdp->bg_checksum));
2480 if (!sb_rdonly(sb)) {
2481 ext4_unlock_group(sb, i);
2485 ext4_unlock_group(sb, i);
2487 first_block += EXT4_BLOCKS_PER_GROUP(sb);
2489 if (NULL != first_not_zeroed)
2490 *first_not_zeroed = grp;
2494 /* ext4_orphan_cleanup() walks a singly-linked list of inodes (starting at
2495 * the superblock) which were deleted from all directories, but held open by
2496 * a process at the time of a crash. We walk the list and try to delete these
2497 * inodes at recovery time (only with a read-write filesystem).
2499 * In order to keep the orphan inode chain consistent during traversal (in
2500 * case of crash during recovery), we link each inode into the superblock
2501 * orphan list_head and handle it the same way as an inode deletion during
2502 * normal operation (which journals the operations for us).
2504 * We only do an iget() and an iput() on each inode, which is very safe if we
2505 * accidentally point at an in-use or already deleted inode. The worst that
2506 * can happen in this case is that we get a "bit already cleared" message from
2507 * ext4_free_inode(). The only reason we would point at a wrong inode is if
2508 * e2fsck was run on this filesystem, and it must have already done the orphan
2509 * inode cleanup for us, so we can safely abort without any further action.
2511 static void ext4_orphan_cleanup(struct super_block *sb,
2512 struct ext4_super_block *es)
2514 unsigned int s_flags = sb->s_flags;
2515 int ret, nr_orphans = 0, nr_truncates = 0;
2517 int quota_update = 0;
2520 if (!es->s_last_orphan) {
2521 jbd_debug(4, "no orphan inodes to clean up\n");
2525 if (bdev_read_only(sb->s_bdev)) {
2526 ext4_msg(sb, KERN_ERR, "write access "
2527 "unavailable, skipping orphan cleanup");
2531 /* Check if feature set would not allow a r/w mount */
2532 if (!ext4_feature_set_ok(sb, 0)) {
2533 ext4_msg(sb, KERN_INFO, "Skipping orphan cleanup due to "
2534 "unknown ROCOMPAT features");
2538 if (EXT4_SB(sb)->s_mount_state & EXT4_ERROR_FS) {
2539 /* don't clear list on RO mount w/ errors */
2540 if (es->s_last_orphan && !(s_flags & MS_RDONLY)) {
2541 ext4_msg(sb, KERN_INFO, "Errors on filesystem, "
2542 "clearing orphan list.\n");
2543 es->s_last_orphan = 0;
2545 jbd_debug(1, "Skipping orphan recovery on fs with errors.\n");
2549 if (s_flags & MS_RDONLY) {
2550 ext4_msg(sb, KERN_INFO, "orphan cleanup on readonly fs");
2551 sb->s_flags &= ~MS_RDONLY;
2554 /* Needed for iput() to work correctly and not trash data */
2555 sb->s_flags |= MS_ACTIVE;
2558 * Turn on quotas which were not enabled for read-only mounts if
2559 * filesystem has quota feature, so that they are updated correctly.
2561 if (ext4_has_feature_quota(sb) && (s_flags & MS_RDONLY)) {
2562 int ret = ext4_enable_quotas(sb);
2567 ext4_msg(sb, KERN_ERR,
2568 "Cannot turn on quotas: error %d", ret);
2571 /* Turn on journaled quotas used for old sytle */
2572 for (i = 0; i < EXT4_MAXQUOTAS; i++) {
2573 if (EXT4_SB(sb)->s_qf_names[i]) {
2574 int ret = ext4_quota_on_mount(sb, i);
2579 ext4_msg(sb, KERN_ERR,
2580 "Cannot turn on journaled "
2581 "quota: type %d: error %d", i, ret);
2586 while (es->s_last_orphan) {
2587 struct inode *inode;
2590 * We may have encountered an error during cleanup; if
2591 * so, skip the rest.
2593 if (EXT4_SB(sb)->s_mount_state & EXT4_ERROR_FS) {
2594 jbd_debug(1, "Skipping orphan recovery on fs with errors.\n");
2595 es->s_last_orphan = 0;
2599 inode = ext4_orphan_get(sb, le32_to_cpu(es->s_last_orphan));
2600 if (IS_ERR(inode)) {
2601 es->s_last_orphan = 0;
2605 list_add(&EXT4_I(inode)->i_orphan, &EXT4_SB(sb)->s_orphan);
2606 dquot_initialize(inode);
2607 if (inode->i_nlink) {
2608 if (test_opt(sb, DEBUG))
2609 ext4_msg(sb, KERN_DEBUG,
2610 "%s: truncating inode %lu to %lld bytes",
2611 __func__, inode->i_ino, inode->i_size);
2612 jbd_debug(2, "truncating inode %lu to %lld bytes\n",
2613 inode->i_ino, inode->i_size);
2615 truncate_inode_pages(inode->i_mapping, inode->i_size);
2616 ret = ext4_truncate(inode);
2619 * We need to clean up the in-core orphan list
2620 * manually if ext4_truncate() failed to get a
2621 * transaction handle.
2623 ext4_orphan_del(NULL, inode);
2624 ext4_std_error(inode->i_sb, ret);
2626 inode_unlock(inode);
2629 if (test_opt(sb, DEBUG))
2630 ext4_msg(sb, KERN_DEBUG,
2631 "%s: deleting unreferenced inode %lu",
2632 __func__, inode->i_ino);
2633 jbd_debug(2, "deleting unreferenced inode %lu\n",
2637 iput(inode); /* The delete magic happens here! */
2640 #define PLURAL(x) (x), ((x) == 1) ? "" : "s"
2643 ext4_msg(sb, KERN_INFO, "%d orphan inode%s deleted",
2644 PLURAL(nr_orphans));
2646 ext4_msg(sb, KERN_INFO, "%d truncate%s cleaned up",
2647 PLURAL(nr_truncates));
2649 /* Turn off quotas if they were enabled for orphan cleanup */
2651 for (i = 0; i < EXT4_MAXQUOTAS; i++) {
2652 if (sb_dqopt(sb)->files[i])
2653 dquot_quota_off(sb, i);
2657 sb->s_flags = s_flags; /* Restore MS_RDONLY status */
2661 * Maximal extent format file size.
2662 * Resulting logical blkno at s_maxbytes must fit in our on-disk
2663 * extent format containers, within a sector_t, and within i_blocks
2664 * in the vfs. ext4 inode has 48 bits of i_block in fsblock units,
2665 * so that won't be a limiting factor.
2667 * However there is other limiting factor. We do store extents in the form
2668 * of starting block and length, hence the resulting length of the extent
2669 * covering maximum file size must fit into on-disk format containers as
2670 * well. Given that length is always by 1 unit bigger than max unit (because
2671 * we count 0 as well) we have to lower the s_maxbytes by one fs block.
2673 * Note, this does *not* consider any metadata overhead for vfs i_blocks.
2675 static loff_t ext4_max_size(int blkbits, int has_huge_files)
2678 loff_t upper_limit = MAX_LFS_FILESIZE;
2680 /* small i_blocks in vfs inode? */
2681 if (!has_huge_files || sizeof(blkcnt_t) < sizeof(u64)) {
2683 * CONFIG_LBDAF is not enabled implies the inode
2684 * i_block represent total blocks in 512 bytes
2685 * 32 == size of vfs inode i_blocks * 8
2687 upper_limit = (1LL << 32) - 1;
2689 /* total blocks in file system block size */
2690 upper_limit >>= (blkbits - 9);
2691 upper_limit <<= blkbits;
2695 * 32-bit extent-start container, ee_block. We lower the maxbytes
2696 * by one fs block, so ee_len can cover the extent of maximum file
2699 res = (1LL << 32) - 1;
2702 /* Sanity check against vm- & vfs- imposed limits */
2703 if (res > upper_limit)
2710 * Maximal bitmap file size. There is a direct, and {,double-,triple-}indirect
2711 * block limit, and also a limit of (2^48 - 1) 512-byte sectors in i_blocks.
2712 * We need to be 1 filesystem block less than the 2^48 sector limit.
2714 static loff_t ext4_max_bitmap_size(int bits, int has_huge_files)
2716 loff_t res = EXT4_NDIR_BLOCKS;
2719 /* This is calculated to be the largest file size for a dense, block
2720 * mapped file such that the file's total number of 512-byte sectors,
2721 * including data and all indirect blocks, does not exceed (2^48 - 1).
2723 * __u32 i_blocks_lo and _u16 i_blocks_high represent the total
2724 * number of 512-byte sectors of the file.
2727 if (!has_huge_files || sizeof(blkcnt_t) < sizeof(u64)) {
2729 * !has_huge_files or CONFIG_LBDAF not enabled implies that
2730 * the inode i_block field represents total file blocks in
2731 * 2^32 512-byte sectors == size of vfs inode i_blocks * 8
2733 upper_limit = (1LL << 32) - 1;
2735 /* total blocks in file system block size */
2736 upper_limit >>= (bits - 9);
2740 * We use 48 bit ext4_inode i_blocks
2741 * With EXT4_HUGE_FILE_FL set the i_blocks
2742 * represent total number of blocks in
2743 * file system block size
2745 upper_limit = (1LL << 48) - 1;
2749 /* indirect blocks */
2751 /* double indirect blocks */
2752 meta_blocks += 1 + (1LL << (bits-2));
2753 /* tripple indirect blocks */
2754 meta_blocks += 1 + (1LL << (bits-2)) + (1LL << (2*(bits-2)));
2756 upper_limit -= meta_blocks;
2757 upper_limit <<= bits;
2759 res += 1LL << (bits-2);
2760 res += 1LL << (2*(bits-2));
2761 res += 1LL << (3*(bits-2));
2763 if (res > upper_limit)
2766 if (res > MAX_LFS_FILESIZE)
2767 res = MAX_LFS_FILESIZE;
2772 static ext4_fsblk_t descriptor_loc(struct super_block *sb,
2773 ext4_fsblk_t logical_sb_block, int nr)
2775 struct ext4_sb_info *sbi = EXT4_SB(sb);
2776 ext4_group_t bg, first_meta_bg;
2779 first_meta_bg = le32_to_cpu(sbi->s_es->s_first_meta_bg);
2781 if (!ext4_has_feature_meta_bg(sb) || nr < first_meta_bg)
2782 return logical_sb_block + nr + 1;
2783 bg = sbi->s_desc_per_block * nr;
2784 if (ext4_bg_has_super(sb, bg))
2788 * If we have a meta_bg fs with 1k blocks, group 0's GDT is at
2789 * block 2, not 1. If s_first_data_block == 0 (bigalloc is enabled
2790 * on modern mke2fs or blksize > 1k on older mke2fs) then we must
2793 if (sb->s_blocksize == 1024 && nr == 0 &&
2794 le32_to_cpu(EXT4_SB(sb)->s_es->s_first_data_block) == 0)
2797 return (has_super + ext4_group_first_block_no(sb, bg));
2801 * ext4_get_stripe_size: Get the stripe size.
2802 * @sbi: In memory super block info
2804 * If we have specified it via mount option, then
2805 * use the mount option value. If the value specified at mount time is
2806 * greater than the blocks per group use the super block value.
2807 * If the super block value is greater than blocks per group return 0.
2808 * Allocator needs it be less than blocks per group.
2811 static unsigned long ext4_get_stripe_size(struct ext4_sb_info *sbi)
2813 unsigned long stride = le16_to_cpu(sbi->s_es->s_raid_stride);
2814 unsigned long stripe_width =
2815 le32_to_cpu(sbi->s_es->s_raid_stripe_width);
2818 if (sbi->s_stripe && sbi->s_stripe <= sbi->s_blocks_per_group)
2819 ret = sbi->s_stripe;
2820 else if (stripe_width && stripe_width <= sbi->s_blocks_per_group)
2822 else if (stride && stride <= sbi->s_blocks_per_group)
2828 * If the stripe width is 1, this makes no sense and
2829 * we set it to 0 to turn off stripe handling code.
2838 * Check whether this filesystem can be mounted based on
2839 * the features present and the RDONLY/RDWR mount requested.
2840 * Returns 1 if this filesystem can be mounted as requested,
2841 * 0 if it cannot be.
2843 static int ext4_feature_set_ok(struct super_block *sb, int readonly)
2845 if (ext4_has_unknown_ext4_incompat_features(sb)) {
2846 ext4_msg(sb, KERN_ERR,
2847 "Couldn't mount because of "
2848 "unsupported optional features (%x)",
2849 (le32_to_cpu(EXT4_SB(sb)->s_es->s_feature_incompat) &
2850 ~EXT4_FEATURE_INCOMPAT_SUPP));
2857 if (ext4_has_feature_readonly(sb)) {
2858 ext4_msg(sb, KERN_INFO, "filesystem is read-only");
2859 sb->s_flags |= MS_RDONLY;
2863 /* Check that feature set is OK for a read-write mount */
2864 if (ext4_has_unknown_ext4_ro_compat_features(sb)) {
2865 ext4_msg(sb, KERN_ERR, "couldn't mount RDWR because of "
2866 "unsupported optional features (%x)",
2867 (le32_to_cpu(EXT4_SB(sb)->s_es->s_feature_ro_compat) &
2868 ~EXT4_FEATURE_RO_COMPAT_SUPP));
2872 * Large file size enabled file system can only be mounted
2873 * read-write on 32-bit systems if kernel is built with CONFIG_LBDAF
2875 if (ext4_has_feature_huge_file(sb)) {
2876 if (sizeof(blkcnt_t) < sizeof(u64)) {
2877 ext4_msg(sb, KERN_ERR, "Filesystem with huge files "
2878 "cannot be mounted RDWR without "
2883 if (ext4_has_feature_bigalloc(sb) && !ext4_has_feature_extents(sb)) {
2884 ext4_msg(sb, KERN_ERR,
2885 "Can't support bigalloc feature without "
2886 "extents feature\n");
2890 #if !IS_ENABLED(CONFIG_QUOTA) || !IS_ENABLED(CONFIG_QFMT_V2)
2891 if (!readonly && (ext4_has_feature_quota(sb) ||
2892 ext4_has_feature_project(sb))) {
2893 ext4_msg(sb, KERN_ERR,
2894 "The kernel was not built with CONFIG_QUOTA and CONFIG_QFMT_V2");
2897 #endif /* CONFIG_QUOTA */
2902 * This function is called once a day if we have errors logged
2903 * on the file system
2905 static void print_daily_error_info(unsigned long arg)
2907 struct super_block *sb = (struct super_block *) arg;
2908 struct ext4_sb_info *sbi;
2909 struct ext4_super_block *es;
2914 if (es->s_error_count)
2915 /* fsck newer than v1.41.13 is needed to clean this condition. */
2916 ext4_msg(sb, KERN_NOTICE, "error count since last fsck: %u",
2917 le32_to_cpu(es->s_error_count));
2918 if (es->s_first_error_time) {
2919 printk(KERN_NOTICE "EXT4-fs (%s): initial error at time %u: %.*s:%d",
2920 sb->s_id, le32_to_cpu(es->s_first_error_time),
2921 (int) sizeof(es->s_first_error_func),
2922 es->s_first_error_func,
2923 le32_to_cpu(es->s_first_error_line));
2924 if (es->s_first_error_ino)
2925 printk(KERN_CONT ": inode %u",
2926 le32_to_cpu(es->s_first_error_ino));
2927 if (es->s_first_error_block)
2928 printk(KERN_CONT ": block %llu", (unsigned long long)
2929 le64_to_cpu(es->s_first_error_block));
2930 printk(KERN_CONT "\n");
2932 if (es->s_last_error_time) {
2933 printk(KERN_NOTICE "EXT4-fs (%s): last error at time %u: %.*s:%d",
2934 sb->s_id, le32_to_cpu(es->s_last_error_time),
2935 (int) sizeof(es->s_last_error_func),
2936 es->s_last_error_func,
2937 le32_to_cpu(es->s_last_error_line));
2938 if (es->s_last_error_ino)
2939 printk(KERN_CONT ": inode %u",
2940 le32_to_cpu(es->s_last_error_ino));
2941 if (es->s_last_error_block)
2942 printk(KERN_CONT ": block %llu", (unsigned long long)
2943 le64_to_cpu(es->s_last_error_block));
2944 printk(KERN_CONT "\n");
2946 mod_timer(&sbi->s_err_report, jiffies + 24*60*60*HZ); /* Once a day */
2949 /* Find next suitable group and run ext4_init_inode_table */
2950 static int ext4_run_li_request(struct ext4_li_request *elr)
2952 struct ext4_group_desc *gdp = NULL;
2953 ext4_group_t group, ngroups;
2954 struct super_block *sb;
2959 ngroups = EXT4_SB(sb)->s_groups_count;
2961 for (group = elr->lr_next_group; group < ngroups; group++) {
2962 gdp = ext4_get_group_desc(sb, group, NULL);
2968 if (!(gdp->bg_flags & cpu_to_le16(EXT4_BG_INODE_ZEROED)))
2972 if (group >= ngroups)
2976 start_time = ktime_get_real_ns();
2977 ret = ext4_init_inode_table(sb, group,
2978 elr->lr_timeout ? 0 : 1);
2979 if (elr->lr_timeout == 0) {
2980 elr->lr_timeout = nsecs_to_jiffies((ktime_get_real_ns() - start_time) *
2981 elr->lr_sbi->s_li_wait_mult);
2983 elr->lr_next_sched = jiffies + elr->lr_timeout;
2984 elr->lr_next_group = group + 1;
2990 * Remove lr_request from the list_request and free the
2991 * request structure. Should be called with li_list_mtx held
2993 static void ext4_remove_li_request(struct ext4_li_request *elr)
2995 struct ext4_sb_info *sbi;
3002 list_del(&elr->lr_request);
3003 sbi->s_li_request = NULL;
3007 static void ext4_unregister_li_request(struct super_block *sb)
3009 mutex_lock(&ext4_li_mtx);
3010 if (!ext4_li_info) {
3011 mutex_unlock(&ext4_li_mtx);
3015 mutex_lock(&ext4_li_info->li_list_mtx);
3016 ext4_remove_li_request(EXT4_SB(sb)->s_li_request);
3017 mutex_unlock(&ext4_li_info->li_list_mtx);
3018 mutex_unlock(&ext4_li_mtx);
3021 static struct task_struct *ext4_lazyinit_task;
3024 * This is the function where ext4lazyinit thread lives. It walks
3025 * through the request list searching for next scheduled filesystem.
3026 * When such a fs is found, run the lazy initialization request
3027 * (ext4_rn_li_request) and keep track of the time spend in this
3028 * function. Based on that time we compute next schedule time of
3029 * the request. When walking through the list is complete, compute
3030 * next waking time and put itself into sleep.
3032 static int ext4_lazyinit_thread(void *arg)
3034 struct ext4_lazy_init *eli = (struct ext4_lazy_init *)arg;
3035 struct list_head *pos, *n;
3036 struct ext4_li_request *elr;
3037 unsigned long next_wakeup, cur;
3039 BUG_ON(NULL == eli);
3044 next_wakeup = MAX_JIFFY_OFFSET;
3046 mutex_lock(&eli->li_list_mtx);
3047 if (list_empty(&eli->li_request_list)) {
3048 mutex_unlock(&eli->li_list_mtx);
3051 list_for_each_safe(pos, n, &eli->li_request_list) {
3054 elr = list_entry(pos, struct ext4_li_request,
3057 if (time_before(jiffies, elr->lr_next_sched)) {
3058 if (time_before(elr->lr_next_sched, next_wakeup))
3059 next_wakeup = elr->lr_next_sched;
3062 if (down_read_trylock(&elr->lr_super->s_umount)) {
3063 if (sb_start_write_trylock(elr->lr_super)) {
3066 * We hold sb->s_umount, sb can not
3067 * be removed from the list, it is
3068 * now safe to drop li_list_mtx
3070 mutex_unlock(&eli->li_list_mtx);
3071 err = ext4_run_li_request(elr);
3072 sb_end_write(elr->lr_super);
3073 mutex_lock(&eli->li_list_mtx);
3076 up_read((&elr->lr_super->s_umount));
3078 /* error, remove the lazy_init job */
3080 ext4_remove_li_request(elr);
3084 elr->lr_next_sched = jiffies +
3086 % (EXT4_DEF_LI_MAX_START_DELAY * HZ));
3088 if (time_before(elr->lr_next_sched, next_wakeup))
3089 next_wakeup = elr->lr_next_sched;
3091 mutex_unlock(&eli->li_list_mtx);
3096 if ((time_after_eq(cur, next_wakeup)) ||
3097 (MAX_JIFFY_OFFSET == next_wakeup)) {
3102 schedule_timeout_interruptible(next_wakeup - cur);
3104 if (kthread_should_stop()) {
3105 ext4_clear_request_list();
3112 * It looks like the request list is empty, but we need
3113 * to check it under the li_list_mtx lock, to prevent any
3114 * additions into it, and of course we should lock ext4_li_mtx
3115 * to atomically free the list and ext4_li_info, because at
3116 * this point another ext4 filesystem could be registering
3119 mutex_lock(&ext4_li_mtx);
3120 mutex_lock(&eli->li_list_mtx);
3121 if (!list_empty(&eli->li_request_list)) {
3122 mutex_unlock(&eli->li_list_mtx);
3123 mutex_unlock(&ext4_li_mtx);
3126 mutex_unlock(&eli->li_list_mtx);
3127 kfree(ext4_li_info);
3128 ext4_li_info = NULL;
3129 mutex_unlock(&ext4_li_mtx);
3134 static void ext4_clear_request_list(void)
3136 struct list_head *pos, *n;
3137 struct ext4_li_request *elr;
3139 mutex_lock(&ext4_li_info->li_list_mtx);
3140 list_for_each_safe(pos, n, &ext4_li_info->li_request_list) {
3141 elr = list_entry(pos, struct ext4_li_request,
3143 ext4_remove_li_request(elr);
3145 mutex_unlock(&ext4_li_info->li_list_mtx);
3148 static int ext4_run_lazyinit_thread(void)
3150 ext4_lazyinit_task = kthread_run(ext4_lazyinit_thread,
3151 ext4_li_info, "ext4lazyinit");
3152 if (IS_ERR(ext4_lazyinit_task)) {
3153 int err = PTR_ERR(ext4_lazyinit_task);
3154 ext4_clear_request_list();
3155 kfree(ext4_li_info);
3156 ext4_li_info = NULL;
3157 printk(KERN_CRIT "EXT4-fs: error %d creating inode table "
3158 "initialization thread\n",
3162 ext4_li_info->li_state |= EXT4_LAZYINIT_RUNNING;
3167 * Check whether it make sense to run itable init. thread or not.
3168 * If there is at least one uninitialized inode table, return
3169 * corresponding group number, else the loop goes through all
3170 * groups and return total number of groups.
3172 static ext4_group_t ext4_has_uninit_itable(struct super_block *sb)
3174 ext4_group_t group, ngroups = EXT4_SB(sb)->s_groups_count;
3175 struct ext4_group_desc *gdp = NULL;
3177 if (!ext4_has_group_desc_csum(sb))
3180 for (group = 0; group < ngroups; group++) {
3181 gdp = ext4_get_group_desc(sb, group, NULL);
3185 if (!(gdp->bg_flags & cpu_to_le16(EXT4_BG_INODE_ZEROED)))
3192 static int ext4_li_info_new(void)
3194 struct ext4_lazy_init *eli = NULL;
3196 eli = kzalloc(sizeof(*eli), GFP_KERNEL);
3200 INIT_LIST_HEAD(&eli->li_request_list);
3201 mutex_init(&eli->li_list_mtx);
3203 eli->li_state |= EXT4_LAZYINIT_QUIT;
3210 static struct ext4_li_request *ext4_li_request_new(struct super_block *sb,
3213 struct ext4_sb_info *sbi = EXT4_SB(sb);
3214 struct ext4_li_request *elr;
3216 elr = kzalloc(sizeof(*elr), GFP_KERNEL);
3222 elr->lr_next_group = start;
3225 * Randomize first schedule time of the request to
3226 * spread the inode table initialization requests
3229 elr->lr_next_sched = jiffies + (prandom_u32() %
3230 (EXT4_DEF_LI_MAX_START_DELAY * HZ));
3234 int ext4_register_li_request(struct super_block *sb,
3235 ext4_group_t first_not_zeroed)
3237 struct ext4_sb_info *sbi = EXT4_SB(sb);
3238 struct ext4_li_request *elr = NULL;
3239 ext4_group_t ngroups = EXT4_SB(sb)->s_groups_count;
3242 mutex_lock(&ext4_li_mtx);
3243 if (sbi->s_li_request != NULL) {
3245 * Reset timeout so it can be computed again, because
3246 * s_li_wait_mult might have changed.
3248 sbi->s_li_request->lr_timeout = 0;
3252 if (first_not_zeroed == ngroups || sb_rdonly(sb) ||
3253 !test_opt(sb, INIT_INODE_TABLE))
3256 elr = ext4_li_request_new(sb, first_not_zeroed);
3262 if (NULL == ext4_li_info) {
3263 ret = ext4_li_info_new();
3268 mutex_lock(&ext4_li_info->li_list_mtx);
3269 list_add(&elr->lr_request, &ext4_li_info->li_request_list);
3270 mutex_unlock(&ext4_li_info->li_list_mtx);
3272 sbi->s_li_request = elr;
3274 * set elr to NULL here since it has been inserted to
3275 * the request_list and the removal and free of it is
3276 * handled by ext4_clear_request_list from now on.
3280 if (!(ext4_li_info->li_state & EXT4_LAZYINIT_RUNNING)) {
3281 ret = ext4_run_lazyinit_thread();
3286 mutex_unlock(&ext4_li_mtx);
3293 * We do not need to lock anything since this is called on
3296 static void ext4_destroy_lazyinit_thread(void)
3299 * If thread exited earlier
3300 * there's nothing to be done.
3302 if (!ext4_li_info || !ext4_lazyinit_task)
3305 kthread_stop(ext4_lazyinit_task);
3308 static int set_journal_csum_feature_set(struct super_block *sb)
3311 int compat, incompat;
3312 struct ext4_sb_info *sbi = EXT4_SB(sb);
3314 if (ext4_has_metadata_csum(sb)) {
3315 /* journal checksum v3 */
3317 incompat = JBD2_FEATURE_INCOMPAT_CSUM_V3;
3319 /* journal checksum v1 */
3320 compat = JBD2_FEATURE_COMPAT_CHECKSUM;
3324 jbd2_journal_clear_features(sbi->s_journal,
3325 JBD2_FEATURE_COMPAT_CHECKSUM, 0,
3326 JBD2_FEATURE_INCOMPAT_CSUM_V3 |
3327 JBD2_FEATURE_INCOMPAT_CSUM_V2);
3328 if (test_opt(sb, JOURNAL_ASYNC_COMMIT)) {
3329 ret = jbd2_journal_set_features(sbi->s_journal,
3331 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT |
3333 } else if (test_opt(sb, JOURNAL_CHECKSUM)) {
3334 ret = jbd2_journal_set_features(sbi->s_journal,
3337 jbd2_journal_clear_features(sbi->s_journal, 0, 0,
3338 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT);
3340 jbd2_journal_clear_features(sbi->s_journal, 0, 0,
3341 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT);
3348 * Note: calculating the overhead so we can be compatible with
3349 * historical BSD practice is quite difficult in the face of
3350 * clusters/bigalloc. This is because multiple metadata blocks from
3351 * different block group can end up in the same allocation cluster.
3352 * Calculating the exact overhead in the face of clustered allocation
3353 * requires either O(all block bitmaps) in memory or O(number of block
3354 * groups**2) in time. We will still calculate the superblock for
3355 * older file systems --- and if we come across with a bigalloc file
3356 * system with zero in s_overhead_clusters the estimate will be close to
3357 * correct especially for very large cluster sizes --- but for newer
3358 * file systems, it's better to calculate this figure once at mkfs
3359 * time, and store it in the superblock. If the superblock value is
3360 * present (even for non-bigalloc file systems), we will use it.
3362 static int count_overhead(struct super_block *sb, ext4_group_t grp,
3365 struct ext4_sb_info *sbi = EXT4_SB(sb);
3366 struct ext4_group_desc *gdp;
3367 ext4_fsblk_t first_block, last_block, b;
3368 ext4_group_t i, ngroups = ext4_get_groups_count(sb);
3369 int s, j, count = 0;
3370 int has_super = ext4_bg_has_super(sb, grp);
3372 if (!ext4_has_feature_bigalloc(sb))
3373 return (has_super + ext4_bg_num_gdb(sb, grp) +
3374 (has_super ? le16_to_cpu(sbi->s_es->s_reserved_gdt_blocks) : 0) +
3375 sbi->s_itb_per_group + 2);
3377 first_block = le32_to_cpu(sbi->s_es->s_first_data_block) +
3378 (grp * EXT4_BLOCKS_PER_GROUP(sb));
3379 last_block = first_block + EXT4_BLOCKS_PER_GROUP(sb) - 1;
3380 for (i = 0; i < ngroups; i++) {
3381 gdp = ext4_get_group_desc(sb, i, NULL);
3382 b = ext4_block_bitmap(sb, gdp);
3383 if (b >= first_block && b <= last_block) {
3384 ext4_set_bit(EXT4_B2C(sbi, b - first_block), buf);
3387 b = ext4_inode_bitmap(sb, gdp);
3388 if (b >= first_block && b <= last_block) {
3389 ext4_set_bit(EXT4_B2C(sbi, b - first_block), buf);
3392 b = ext4_inode_table(sb, gdp);
3393 if (b >= first_block && b + sbi->s_itb_per_group <= last_block)
3394 for (j = 0; j < sbi->s_itb_per_group; j++, b++) {
3395 int c = EXT4_B2C(sbi, b - first_block);
3396 ext4_set_bit(c, buf);
3402 if (ext4_bg_has_super(sb, grp)) {
3403 ext4_set_bit(s++, buf);
3406 j = ext4_bg_num_gdb(sb, grp);
3407 if (s + j > EXT4_BLOCKS_PER_GROUP(sb)) {
3408 ext4_error(sb, "Invalid number of block group "
3409 "descriptor blocks: %d", j);
3410 j = EXT4_BLOCKS_PER_GROUP(sb) - s;
3414 ext4_set_bit(EXT4_B2C(sbi, s++), buf);
3418 return EXT4_CLUSTERS_PER_GROUP(sb) -
3419 ext4_count_free(buf, EXT4_CLUSTERS_PER_GROUP(sb) / 8);
3423 * Compute the overhead and stash it in sbi->s_overhead
3425 int ext4_calculate_overhead(struct super_block *sb)
3427 struct ext4_sb_info *sbi = EXT4_SB(sb);
3428 struct ext4_super_block *es = sbi->s_es;
3429 struct inode *j_inode;
3430 unsigned int j_blocks, j_inum = le32_to_cpu(es->s_journal_inum);
3431 ext4_group_t i, ngroups = ext4_get_groups_count(sb);
3432 ext4_fsblk_t overhead = 0;
3433 char *buf = (char *) get_zeroed_page(GFP_NOFS);
3439 * Compute the overhead (FS structures). This is constant
3440 * for a given filesystem unless the number of block groups
3441 * changes so we cache the previous value until it does.
3445 * All of the blocks before first_data_block are overhead
3447 overhead = EXT4_B2C(sbi, le32_to_cpu(es->s_first_data_block));
3450 * Add the overhead found in each block group
3452 for (i = 0; i < ngroups; i++) {
3455 blks = count_overhead(sb, i, buf);
3458 memset(buf, 0, PAGE_SIZE);
3463 * Add the internal journal blocks whether the journal has been
3466 if (sbi->s_journal && !sbi->journal_bdev)
3467 overhead += EXT4_NUM_B2C(sbi, sbi->s_journal->j_maxlen);
3468 else if (ext4_has_feature_journal(sb) && !sbi->s_journal && j_inum) {
3469 /* j_inum for internal journal is non-zero */
3470 j_inode = ext4_get_journal_inode(sb, j_inum);
3472 j_blocks = j_inode->i_size >> sb->s_blocksize_bits;
3473 overhead += EXT4_NUM_B2C(sbi, j_blocks);
3476 ext4_msg(sb, KERN_ERR, "can't get journal size");
3479 sbi->s_overhead = overhead;
3481 free_page((unsigned long) buf);
3485 static void ext4_set_resv_clusters(struct super_block *sb)
3487 ext4_fsblk_t resv_clusters;
3488 struct ext4_sb_info *sbi = EXT4_SB(sb);
3491 * There's no need to reserve anything when we aren't using extents.
3492 * The space estimates are exact, there are no unwritten extents,
3493 * hole punching doesn't need new metadata... This is needed especially
3494 * to keep ext2/3 backward compatibility.
3496 if (!ext4_has_feature_extents(sb))
3499 * By default we reserve 2% or 4096 clusters, whichever is smaller.
3500 * This should cover the situations where we can not afford to run
3501 * out of space like for example punch hole, or converting
3502 * unwritten extents in delalloc path. In most cases such
3503 * allocation would require 1, or 2 blocks, higher numbers are
3506 resv_clusters = (ext4_blocks_count(sbi->s_es) >>
3507 sbi->s_cluster_bits);
3509 do_div(resv_clusters, 50);
3510 resv_clusters = min_t(ext4_fsblk_t, resv_clusters, 4096);
3512 atomic64_set(&sbi->s_resv_clusters, resv_clusters);
3515 static int ext4_fill_super(struct super_block *sb, void *data, int silent)
3517 struct dax_device *dax_dev = fs_dax_get_by_bdev(sb->s_bdev);
3518 char *orig_data = kstrdup(data, GFP_KERNEL);
3519 struct buffer_head *bh, **group_desc;
3520 struct ext4_super_block *es = NULL;
3521 struct ext4_sb_info *sbi = kzalloc(sizeof(*sbi), GFP_KERNEL);
3522 struct flex_groups **flex_groups;
3524 ext4_fsblk_t sb_block = get_sb_block(&data);
3525 ext4_fsblk_t logical_sb_block;
3526 unsigned long offset = 0;
3527 unsigned long journal_devnum = 0;
3528 unsigned long def_mount_opts;
3532 int blocksize, clustersize;
3533 unsigned int db_count;
3535 int needs_recovery, has_huge_files, has_bigalloc;
3538 unsigned int journal_ioprio = DEFAULT_JOURNAL_IOPRIO;
3539 ext4_group_t first_not_zeroed;
3541 if ((data && !orig_data) || !sbi)
3544 sbi->s_daxdev = dax_dev;
3545 sbi->s_blockgroup_lock =
3546 kzalloc(sizeof(struct blockgroup_lock), GFP_KERNEL);
3547 if (!sbi->s_blockgroup_lock)
3550 sb->s_fs_info = sbi;
3552 sbi->s_inode_readahead_blks = EXT4_DEF_INODE_READAHEAD_BLKS;
3553 sbi->s_sb_block = sb_block;
3554 if (sb->s_bdev->bd_part)
3555 sbi->s_sectors_written_start =
3556 part_stat_read(sb->s_bdev->bd_part, sectors[1]);
3558 /* Cleanup superblock name */
3559 strreplace(sb->s_id, '/', '!');
3561 /* -EINVAL is default */
3563 blocksize = sb_min_blocksize(sb, EXT4_MIN_BLOCK_SIZE);
3565 ext4_msg(sb, KERN_ERR, "unable to set blocksize");
3570 * The ext4 superblock will not be buffer aligned for other than 1kB
3571 * block sizes. We need to calculate the offset from buffer start.
3573 if (blocksize != EXT4_MIN_BLOCK_SIZE) {
3574 logical_sb_block = sb_block * EXT4_MIN_BLOCK_SIZE;
3575 offset = do_div(logical_sb_block, blocksize);
3577 logical_sb_block = sb_block;
3580 if (!(bh = sb_bread_unmovable(sb, logical_sb_block))) {
3581 ext4_msg(sb, KERN_ERR, "unable to read superblock");
3585 * Note: s_es must be initialized as soon as possible because
3586 * some ext4 macro-instructions depend on its value
3588 es = (struct ext4_super_block *) (bh->b_data + offset);
3590 sb->s_magic = le16_to_cpu(es->s_magic);
3591 if (sb->s_magic != EXT4_SUPER_MAGIC)
3593 sbi->s_kbytes_written = le64_to_cpu(es->s_kbytes_written);
3595 /* Warn if metadata_csum and gdt_csum are both set. */
3596 if (ext4_has_feature_metadata_csum(sb) &&
3597 ext4_has_feature_gdt_csum(sb))
3598 ext4_warning(sb, "metadata_csum and uninit_bg are "
3599 "redundant flags; please run fsck.");
3601 /* Check for a known checksum algorithm */
3602 if (!ext4_verify_csum_type(sb, es)) {
3603 ext4_msg(sb, KERN_ERR, "VFS: Found ext4 filesystem with "
3604 "unknown checksum algorithm.");
3609 /* Load the checksum driver */
3610 sbi->s_chksum_driver = crypto_alloc_shash("crc32c", 0, 0);
3611 if (IS_ERR(sbi->s_chksum_driver)) {
3612 ext4_msg(sb, KERN_ERR, "Cannot load crc32c driver.");
3613 ret = PTR_ERR(sbi->s_chksum_driver);
3614 sbi->s_chksum_driver = NULL;
3618 /* Check superblock checksum */
3619 if (!ext4_superblock_csum_verify(sb, es)) {
3620 ext4_msg(sb, KERN_ERR, "VFS: Found ext4 filesystem with "
3621 "invalid superblock checksum. Run e2fsck?");
3627 /* Precompute checksum seed for all metadata */
3628 if (ext4_has_feature_csum_seed(sb))
3629 sbi->s_csum_seed = le32_to_cpu(es->s_checksum_seed);
3630 else if (ext4_has_metadata_csum(sb) || ext4_has_feature_ea_inode(sb))
3631 sbi->s_csum_seed = ext4_chksum(sbi, ~0, es->s_uuid,
3632 sizeof(es->s_uuid));
3634 /* Set defaults before we parse the mount options */
3635 def_mount_opts = le32_to_cpu(es->s_default_mount_opts);
3636 set_opt(sb, INIT_INODE_TABLE);
3637 if (def_mount_opts & EXT4_DEFM_DEBUG)
3639 if (def_mount_opts & EXT4_DEFM_BSDGROUPS)
3641 if (def_mount_opts & EXT4_DEFM_UID16)
3642 set_opt(sb, NO_UID32);
3643 /* xattr user namespace & acls are now defaulted on */
3644 set_opt(sb, XATTR_USER);
3645 #ifdef CONFIG_EXT4_FS_POSIX_ACL
3646 set_opt(sb, POSIX_ACL);
3648 /* don't forget to enable journal_csum when metadata_csum is enabled. */
3649 if (ext4_has_metadata_csum(sb))
3650 set_opt(sb, JOURNAL_CHECKSUM);
3652 if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_DATA)
3653 set_opt(sb, JOURNAL_DATA);
3654 else if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_ORDERED)
3655 set_opt(sb, ORDERED_DATA);
3656 else if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_WBACK)
3657 set_opt(sb, WRITEBACK_DATA);
3659 if (le16_to_cpu(sbi->s_es->s_errors) == EXT4_ERRORS_PANIC)
3660 set_opt(sb, ERRORS_PANIC);
3661 else if (le16_to_cpu(sbi->s_es->s_errors) == EXT4_ERRORS_CONTINUE)
3662 set_opt(sb, ERRORS_CONT);
3664 set_opt(sb, ERRORS_RO);
3665 /* block_validity enabled by default; disable with noblock_validity */
3666 set_opt(sb, BLOCK_VALIDITY);
3667 if (def_mount_opts & EXT4_DEFM_DISCARD)
3668 set_opt(sb, DISCARD);
3670 sbi->s_resuid = make_kuid(&init_user_ns, le16_to_cpu(es->s_def_resuid));
3671 sbi->s_resgid = make_kgid(&init_user_ns, le16_to_cpu(es->s_def_resgid));
3672 sbi->s_commit_interval = JBD2_DEFAULT_MAX_COMMIT_AGE * HZ;
3673 sbi->s_min_batch_time = EXT4_DEF_MIN_BATCH_TIME;
3674 sbi->s_max_batch_time = EXT4_DEF_MAX_BATCH_TIME;
3676 if ((def_mount_opts & EXT4_DEFM_NOBARRIER) == 0)
3677 set_opt(sb, BARRIER);
3680 * enable delayed allocation by default
3681 * Use -o nodelalloc to turn it off
3683 if (!IS_EXT3_SB(sb) && !IS_EXT2_SB(sb) &&
3684 ((def_mount_opts & EXT4_DEFM_NODELALLOC) == 0))
3685 set_opt(sb, DELALLOC);
3688 * set default s_li_wait_mult for lazyinit, for the case there is
3689 * no mount option specified.
3691 sbi->s_li_wait_mult = EXT4_DEF_LI_WAIT_MULT;
3693 blocksize = BLOCK_SIZE << le32_to_cpu(es->s_log_block_size);
3694 if (blocksize < EXT4_MIN_BLOCK_SIZE ||
3695 blocksize > EXT4_MAX_BLOCK_SIZE) {
3696 ext4_msg(sb, KERN_ERR,
3697 "Unsupported filesystem blocksize %d (%d log_block_size)",
3698 blocksize, le32_to_cpu(es->s_log_block_size));
3702 if (le32_to_cpu(es->s_rev_level) == EXT4_GOOD_OLD_REV) {
3703 sbi->s_inode_size = EXT4_GOOD_OLD_INODE_SIZE;
3704 sbi->s_first_ino = EXT4_GOOD_OLD_FIRST_INO;
3706 sbi->s_inode_size = le16_to_cpu(es->s_inode_size);
3707 sbi->s_first_ino = le32_to_cpu(es->s_first_ino);
3708 if (sbi->s_first_ino < EXT4_GOOD_OLD_FIRST_INO) {
3709 ext4_msg(sb, KERN_ERR, "invalid first ino: %u",
3713 if ((sbi->s_inode_size < EXT4_GOOD_OLD_INODE_SIZE) ||
3714 (!is_power_of_2(sbi->s_inode_size)) ||
3715 (sbi->s_inode_size > blocksize)) {
3716 ext4_msg(sb, KERN_ERR,
3717 "unsupported inode size: %d",
3719 ext4_msg(sb, KERN_ERR, "blocksize: %d", blocksize);
3723 * i_atime_extra is the last extra field available for
3724 * [acm]times in struct ext4_inode. Checking for that
3725 * field should suffice to ensure we have extra space
3728 if (sbi->s_inode_size >= offsetof(struct ext4_inode, i_atime_extra) +
3729 sizeof(((struct ext4_inode *)0)->i_atime_extra)) {
3730 sb->s_time_gran = 1;
3732 sb->s_time_gran = NSEC_PER_SEC;
3735 if (sbi->s_inode_size > EXT4_GOOD_OLD_INODE_SIZE) {
3736 sbi->s_want_extra_isize = sizeof(struct ext4_inode) -
3737 EXT4_GOOD_OLD_INODE_SIZE;
3738 if (ext4_has_feature_extra_isize(sb)) {
3739 unsigned v, max = (sbi->s_inode_size -
3740 EXT4_GOOD_OLD_INODE_SIZE);
3742 v = le16_to_cpu(es->s_want_extra_isize);
3744 ext4_msg(sb, KERN_ERR,
3745 "bad s_want_extra_isize: %d", v);
3748 if (sbi->s_want_extra_isize < v)
3749 sbi->s_want_extra_isize = v;
3751 v = le16_to_cpu(es->s_min_extra_isize);
3753 ext4_msg(sb, KERN_ERR,
3754 "bad s_min_extra_isize: %d", v);
3757 if (sbi->s_want_extra_isize < v)
3758 sbi->s_want_extra_isize = v;
3762 if (sbi->s_es->s_mount_opts[0]) {
3763 char *s_mount_opts = kstrndup(sbi->s_es->s_mount_opts,
3764 sizeof(sbi->s_es->s_mount_opts),
3768 if (!parse_options(s_mount_opts, sb, &journal_devnum,
3769 &journal_ioprio, 0)) {
3770 ext4_msg(sb, KERN_WARNING,
3771 "failed to parse options in superblock: %s",
3774 kfree(s_mount_opts);
3776 sbi->s_def_mount_opt = sbi->s_mount_opt;
3777 if (!parse_options((char *) data, sb, &journal_devnum,
3778 &journal_ioprio, 0))
3781 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA) {
3782 printk_once(KERN_WARNING "EXT4-fs: Warning: mounting "
3783 "with data=journal disables delayed "
3784 "allocation and O_DIRECT support!\n");
3785 if (test_opt2(sb, EXPLICIT_DELALLOC)) {
3786 ext4_msg(sb, KERN_ERR, "can't mount with "
3787 "both data=journal and delalloc");
3790 if (test_opt(sb, DIOREAD_NOLOCK)) {
3791 ext4_msg(sb, KERN_ERR, "can't mount with "
3792 "both data=journal and dioread_nolock");
3795 if (test_opt(sb, DAX)) {
3796 ext4_msg(sb, KERN_ERR, "can't mount with "
3797 "both data=journal and dax");
3800 if (ext4_has_feature_encrypt(sb)) {
3801 ext4_msg(sb, KERN_WARNING,
3802 "encrypted files will use data=ordered "
3803 "instead of data journaling mode");
3805 if (test_opt(sb, DELALLOC))
3806 clear_opt(sb, DELALLOC);
3808 sb->s_iflags |= SB_I_CGROUPWB;
3811 sb->s_flags = (sb->s_flags & ~MS_POSIXACL) |
3812 (test_opt(sb, POSIX_ACL) ? MS_POSIXACL : 0);
3814 if (le32_to_cpu(es->s_rev_level) == EXT4_GOOD_OLD_REV &&
3815 (ext4_has_compat_features(sb) ||
3816 ext4_has_ro_compat_features(sb) ||
3817 ext4_has_incompat_features(sb)))
3818 ext4_msg(sb, KERN_WARNING,
3819 "feature flags set on rev 0 fs, "
3820 "running e2fsck is recommended");
3822 if (es->s_creator_os == cpu_to_le32(EXT4_OS_HURD)) {
3823 set_opt2(sb, HURD_COMPAT);
3824 if (ext4_has_feature_64bit(sb)) {
3825 ext4_msg(sb, KERN_ERR,
3826 "The Hurd can't support 64-bit file systems");
3831 * ea_inode feature uses l_i_version field which is not
3832 * available in HURD_COMPAT mode.
3834 if (ext4_has_feature_ea_inode(sb)) {
3835 ext4_msg(sb, KERN_ERR,
3836 "ea_inode feature is not supported for Hurd");
3841 if (IS_EXT2_SB(sb)) {
3842 if (ext2_feature_set_ok(sb))
3843 ext4_msg(sb, KERN_INFO, "mounting ext2 file system "
3844 "using the ext4 subsystem");
3847 * If we're probing be silent, if this looks like
3848 * it's actually an ext[34] filesystem.
3850 if (silent && ext4_feature_set_ok(sb, sb_rdonly(sb)))
3852 ext4_msg(sb, KERN_ERR, "couldn't mount as ext2 due "
3853 "to feature incompatibilities");
3858 if (IS_EXT3_SB(sb)) {
3859 if (ext3_feature_set_ok(sb))
3860 ext4_msg(sb, KERN_INFO, "mounting ext3 file system "
3861 "using the ext4 subsystem");
3864 * If we're probing be silent, if this looks like
3865 * it's actually an ext4 filesystem.
3867 if (silent && ext4_feature_set_ok(sb, sb_rdonly(sb)))
3869 ext4_msg(sb, KERN_ERR, "couldn't mount as ext3 due "
3870 "to feature incompatibilities");
3876 * Check feature flags regardless of the revision level, since we
3877 * previously didn't change the revision level when setting the flags,
3878 * so there is a chance incompat flags are set on a rev 0 filesystem.
3880 if (!ext4_feature_set_ok(sb, (sb_rdonly(sb))))
3883 if (le32_to_cpu(es->s_log_block_size) >
3884 (EXT4_MAX_BLOCK_LOG_SIZE - EXT4_MIN_BLOCK_LOG_SIZE)) {
3885 ext4_msg(sb, KERN_ERR,
3886 "Invalid log block size: %u",
3887 le32_to_cpu(es->s_log_block_size));
3890 if (le32_to_cpu(es->s_log_cluster_size) >
3891 (EXT4_MAX_CLUSTER_LOG_SIZE - EXT4_MIN_BLOCK_LOG_SIZE)) {
3892 ext4_msg(sb, KERN_ERR,
3893 "Invalid log cluster size: %u",
3894 le32_to_cpu(es->s_log_cluster_size));
3898 if (le16_to_cpu(sbi->s_es->s_reserved_gdt_blocks) > (blocksize / 4)) {
3899 ext4_msg(sb, KERN_ERR,
3900 "Number of reserved GDT blocks insanely large: %d",
3901 le16_to_cpu(sbi->s_es->s_reserved_gdt_blocks));
3905 if (sbi->s_mount_opt & EXT4_MOUNT_DAX) {
3906 if (ext4_has_feature_inline_data(sb)) {
3907 ext4_msg(sb, KERN_ERR, "Cannot use DAX on a filesystem"
3908 " that may contain inline data");
3911 if (!bdev_dax_supported(sb->s_bdev, blocksize))
3915 if (ext4_has_feature_encrypt(sb) && es->s_encryption_level) {
3916 ext4_msg(sb, KERN_ERR, "Unsupported encryption level %d",
3917 es->s_encryption_level);
3921 if (sb->s_blocksize != blocksize) {
3922 /* Validate the filesystem blocksize */
3923 if (!sb_set_blocksize(sb, blocksize)) {
3924 ext4_msg(sb, KERN_ERR, "bad block size %d",
3930 logical_sb_block = sb_block * EXT4_MIN_BLOCK_SIZE;
3931 offset = do_div(logical_sb_block, blocksize);
3932 bh = sb_bread_unmovable(sb, logical_sb_block);
3934 ext4_msg(sb, KERN_ERR,
3935 "Can't read superblock on 2nd try");
3938 es = (struct ext4_super_block *)(bh->b_data + offset);
3940 if (es->s_magic != cpu_to_le16(EXT4_SUPER_MAGIC)) {
3941 ext4_msg(sb, KERN_ERR,
3942 "Magic mismatch, very weird!");
3947 has_huge_files = ext4_has_feature_huge_file(sb);
3948 sbi->s_bitmap_maxbytes = ext4_max_bitmap_size(sb->s_blocksize_bits,
3950 sb->s_maxbytes = ext4_max_size(sb->s_blocksize_bits, has_huge_files);
3952 sbi->s_desc_size = le16_to_cpu(es->s_desc_size);
3953 if (ext4_has_feature_64bit(sb)) {
3954 if (sbi->s_desc_size < EXT4_MIN_DESC_SIZE_64BIT ||
3955 sbi->s_desc_size > EXT4_MAX_DESC_SIZE ||
3956 !is_power_of_2(sbi->s_desc_size)) {
3957 ext4_msg(sb, KERN_ERR,
3958 "unsupported descriptor size %lu",
3963 sbi->s_desc_size = EXT4_MIN_DESC_SIZE;
3965 sbi->s_blocks_per_group = le32_to_cpu(es->s_blocks_per_group);
3966 sbi->s_inodes_per_group = le32_to_cpu(es->s_inodes_per_group);
3968 sbi->s_inodes_per_block = blocksize / EXT4_INODE_SIZE(sb);
3969 if (sbi->s_inodes_per_block == 0)
3971 if (sbi->s_inodes_per_group < sbi->s_inodes_per_block ||
3972 sbi->s_inodes_per_group > blocksize * 8) {
3973 ext4_msg(sb, KERN_ERR, "invalid inodes per group: %lu\n",
3974 sbi->s_inodes_per_group);
3977 sbi->s_itb_per_group = sbi->s_inodes_per_group /
3978 sbi->s_inodes_per_block;
3979 sbi->s_desc_per_block = blocksize / EXT4_DESC_SIZE(sb);
3981 sbi->s_mount_state = le16_to_cpu(es->s_state);
3982 sbi->s_addr_per_block_bits = ilog2(EXT4_ADDR_PER_BLOCK(sb));
3983 sbi->s_desc_per_block_bits = ilog2(EXT4_DESC_PER_BLOCK(sb));
3985 for (i = 0; i < 4; i++)
3986 sbi->s_hash_seed[i] = le32_to_cpu(es->s_hash_seed[i]);
3987 sbi->s_def_hash_version = es->s_def_hash_version;
3988 if (ext4_has_feature_dir_index(sb)) {
3989 i = le32_to_cpu(es->s_flags);
3990 if (i & EXT2_FLAGS_UNSIGNED_HASH)
3991 sbi->s_hash_unsigned = 3;
3992 else if ((i & EXT2_FLAGS_SIGNED_HASH) == 0) {
3993 #ifdef __CHAR_UNSIGNED__
3996 cpu_to_le32(EXT2_FLAGS_UNSIGNED_HASH);
3997 sbi->s_hash_unsigned = 3;
4001 cpu_to_le32(EXT2_FLAGS_SIGNED_HASH);
4006 /* Handle clustersize */
4007 clustersize = BLOCK_SIZE << le32_to_cpu(es->s_log_cluster_size);
4008 has_bigalloc = ext4_has_feature_bigalloc(sb);
4010 if (clustersize < blocksize) {
4011 ext4_msg(sb, KERN_ERR,
4012 "cluster size (%d) smaller than "
4013 "block size (%d)", clustersize, blocksize);
4016 sbi->s_cluster_bits = le32_to_cpu(es->s_log_cluster_size) -
4017 le32_to_cpu(es->s_log_block_size);
4018 sbi->s_clusters_per_group =
4019 le32_to_cpu(es->s_clusters_per_group);
4020 if (sbi->s_clusters_per_group > blocksize * 8) {
4021 ext4_msg(sb, KERN_ERR,
4022 "#clusters per group too big: %lu",
4023 sbi->s_clusters_per_group);
4026 if (sbi->s_blocks_per_group !=
4027 (sbi->s_clusters_per_group * (clustersize / blocksize))) {
4028 ext4_msg(sb, KERN_ERR, "blocks per group (%lu) and "
4029 "clusters per group (%lu) inconsistent",
4030 sbi->s_blocks_per_group,
4031 sbi->s_clusters_per_group);
4035 if (clustersize != blocksize) {
4036 ext4_msg(sb, KERN_ERR,
4037 "fragment/cluster size (%d) != "
4038 "block size (%d)", clustersize, blocksize);
4041 if (sbi->s_blocks_per_group > blocksize * 8) {
4042 ext4_msg(sb, KERN_ERR,
4043 "#blocks per group too big: %lu",
4044 sbi->s_blocks_per_group);
4047 sbi->s_clusters_per_group = sbi->s_blocks_per_group;
4048 sbi->s_cluster_bits = 0;
4050 sbi->s_cluster_ratio = clustersize / blocksize;
4052 /* Do we have standard group size of clustersize * 8 blocks ? */
4053 if (sbi->s_blocks_per_group == clustersize << 3)
4054 set_opt2(sb, STD_GROUP_SIZE);
4057 * Test whether we have more sectors than will fit in sector_t,
4058 * and whether the max offset is addressable by the page cache.
4060 err = generic_check_addressable(sb->s_blocksize_bits,
4061 ext4_blocks_count(es));
4063 ext4_msg(sb, KERN_ERR, "filesystem"
4064 " too large to mount safely on this system");
4065 if (sizeof(sector_t) < 8)
4066 ext4_msg(sb, KERN_WARNING, "CONFIG_LBDAF not enabled");
4070 if (EXT4_BLOCKS_PER_GROUP(sb) == 0)
4073 /* check blocks count against device size */
4074 blocks_count = sb->s_bdev->bd_inode->i_size >> sb->s_blocksize_bits;
4075 if (blocks_count && ext4_blocks_count(es) > blocks_count) {
4076 ext4_msg(sb, KERN_WARNING, "bad geometry: block count %llu "
4077 "exceeds size of device (%llu blocks)",
4078 ext4_blocks_count(es), blocks_count);
4083 * It makes no sense for the first data block to be beyond the end
4084 * of the filesystem.
4086 if (le32_to_cpu(es->s_first_data_block) >= ext4_blocks_count(es)) {
4087 ext4_msg(sb, KERN_WARNING, "bad geometry: first data "
4088 "block %u is beyond end of filesystem (%llu)",
4089 le32_to_cpu(es->s_first_data_block),
4090 ext4_blocks_count(es));
4093 if ((es->s_first_data_block == 0) && (es->s_log_block_size == 0) &&
4094 (sbi->s_cluster_ratio == 1)) {
4095 ext4_msg(sb, KERN_WARNING, "bad geometry: first data "
4096 "block is 0 with a 1k block and cluster size");
4100 blocks_count = (ext4_blocks_count(es) -
4101 le32_to_cpu(es->s_first_data_block) +
4102 EXT4_BLOCKS_PER_GROUP(sb) - 1);
4103 do_div(blocks_count, EXT4_BLOCKS_PER_GROUP(sb));
4104 if (blocks_count > ((uint64_t)1<<32) - EXT4_DESC_PER_BLOCK(sb)) {
4105 ext4_msg(sb, KERN_WARNING, "groups count too large: %llu "
4106 "(block count %llu, first data block %u, "
4107 "blocks per group %lu)", blocks_count,
4108 ext4_blocks_count(es),
4109 le32_to_cpu(es->s_first_data_block),
4110 EXT4_BLOCKS_PER_GROUP(sb));
4113 sbi->s_groups_count = blocks_count;
4114 sbi->s_blockfile_groups = min_t(ext4_group_t, sbi->s_groups_count,
4115 (EXT4_MAX_BLOCK_FILE_PHYS / EXT4_BLOCKS_PER_GROUP(sb)));
4116 if (((u64)sbi->s_groups_count * sbi->s_inodes_per_group) !=
4117 le32_to_cpu(es->s_inodes_count)) {
4118 ext4_msg(sb, KERN_ERR, "inodes count not valid: %u vs %llu",
4119 le32_to_cpu(es->s_inodes_count),
4120 ((u64)sbi->s_groups_count * sbi->s_inodes_per_group));
4124 db_count = (sbi->s_groups_count + EXT4_DESC_PER_BLOCK(sb) - 1) /
4125 EXT4_DESC_PER_BLOCK(sb);
4126 if (ext4_has_feature_meta_bg(sb)) {
4127 if (le32_to_cpu(es->s_first_meta_bg) > db_count) {
4128 ext4_msg(sb, KERN_WARNING,
4129 "first meta block group too large: %u "
4130 "(group descriptor block count %u)",
4131 le32_to_cpu(es->s_first_meta_bg), db_count);
4135 rcu_assign_pointer(sbi->s_group_desc,
4136 kvmalloc_array(db_count,
4137 sizeof(struct buffer_head *),
4139 if (sbi->s_group_desc == NULL) {
4140 ext4_msg(sb, KERN_ERR, "not enough memory");
4145 bgl_lock_init(sbi->s_blockgroup_lock);
4147 /* Pre-read the descriptors into the buffer cache */
4148 for (i = 0; i < db_count; i++) {
4149 block = descriptor_loc(sb, logical_sb_block, i);
4150 sb_breadahead_unmovable(sb, block);
4153 for (i = 0; i < db_count; i++) {
4154 struct buffer_head *bh;
4156 block = descriptor_loc(sb, logical_sb_block, i);
4157 bh = sb_bread_unmovable(sb, block);
4159 ext4_msg(sb, KERN_ERR,
4160 "can't read group descriptor %d", i);
4165 rcu_dereference(sbi->s_group_desc)[i] = bh;
4168 sbi->s_gdb_count = db_count;
4169 if (!ext4_check_descriptors(sb, logical_sb_block, &first_not_zeroed)) {
4170 ext4_msg(sb, KERN_ERR, "group descriptors corrupted!");
4171 ret = -EFSCORRUPTED;
4175 get_random_bytes(&sbi->s_next_generation, sizeof(u32));
4176 spin_lock_init(&sbi->s_next_gen_lock);
4178 setup_timer(&sbi->s_err_report, print_daily_error_info,
4179 (unsigned long) sb);
4181 /* Register extent status tree shrinker */
4182 if (ext4_es_register_shrinker(sbi))
4185 sbi->s_stripe = ext4_get_stripe_size(sbi);
4186 sbi->s_extent_max_zeroout_kb = 32;
4189 * set up enough so that it can read an inode
4191 sb->s_op = &ext4_sops;
4192 sb->s_export_op = &ext4_export_ops;
4193 sb->s_xattr = ext4_xattr_handlers;
4194 sb->s_cop = &ext4_cryptops;
4196 sb->dq_op = &ext4_quota_operations;
4197 if (ext4_has_feature_quota(sb))
4198 sb->s_qcop = &dquot_quotactl_sysfile_ops;
4200 sb->s_qcop = &ext4_qctl_operations;
4201 sb->s_quota_types = QTYPE_MASK_USR | QTYPE_MASK_GRP | QTYPE_MASK_PRJ;
4203 memcpy(&sb->s_uuid, es->s_uuid, sizeof(es->s_uuid));
4205 INIT_LIST_HEAD(&sbi->s_orphan); /* unlinked but open files */
4206 mutex_init(&sbi->s_orphan_lock);
4210 needs_recovery = (es->s_last_orphan != 0 ||
4211 ext4_has_feature_journal_needs_recovery(sb));
4213 if (ext4_has_feature_mmp(sb) && !sb_rdonly(sb))
4214 if (ext4_multi_mount_protect(sb, le64_to_cpu(es->s_mmp_block)))
4215 goto failed_mount3a;
4218 * The first inode we look at is the journal inode. Don't try
4219 * root first: it may be modified in the journal!
4221 if (!test_opt(sb, NOLOAD) && ext4_has_feature_journal(sb)) {
4222 err = ext4_load_journal(sb, es, journal_devnum);
4224 goto failed_mount3a;
4225 } else if (test_opt(sb, NOLOAD) && !sb_rdonly(sb) &&
4226 ext4_has_feature_journal_needs_recovery(sb)) {
4227 ext4_msg(sb, KERN_ERR, "required journal recovery "
4228 "suppressed and not mounted read-only");
4229 goto failed_mount_wq;
4231 /* Nojournal mode, all journal mount options are illegal */
4232 if (test_opt2(sb, EXPLICIT_JOURNAL_CHECKSUM)) {
4233 ext4_msg(sb, KERN_ERR, "can't mount with "
4234 "journal_checksum, fs mounted w/o journal");
4235 goto failed_mount_wq;
4237 if (test_opt(sb, JOURNAL_ASYNC_COMMIT)) {
4238 ext4_msg(sb, KERN_ERR, "can't mount with "
4239 "journal_async_commit, fs mounted w/o journal");
4240 goto failed_mount_wq;
4242 if (sbi->s_commit_interval != JBD2_DEFAULT_MAX_COMMIT_AGE*HZ) {
4243 ext4_msg(sb, KERN_ERR, "can't mount with "
4244 "commit=%lu, fs mounted w/o journal",
4245 sbi->s_commit_interval / HZ);
4246 goto failed_mount_wq;
4248 if (EXT4_MOUNT_DATA_FLAGS &
4249 (sbi->s_mount_opt ^ sbi->s_def_mount_opt)) {
4250 ext4_msg(sb, KERN_ERR, "can't mount with "
4251 "data=, fs mounted w/o journal");
4252 goto failed_mount_wq;
4254 sbi->s_def_mount_opt &= ~EXT4_MOUNT_JOURNAL_CHECKSUM;
4255 clear_opt(sb, JOURNAL_CHECKSUM);
4256 clear_opt(sb, DATA_FLAGS);
4257 sbi->s_journal = NULL;
4262 if (ext4_has_feature_64bit(sb) &&
4263 !jbd2_journal_set_features(EXT4_SB(sb)->s_journal, 0, 0,
4264 JBD2_FEATURE_INCOMPAT_64BIT)) {
4265 ext4_msg(sb, KERN_ERR, "Failed to set 64-bit journal feature");
4266 goto failed_mount_wq;
4269 if (!set_journal_csum_feature_set(sb)) {
4270 ext4_msg(sb, KERN_ERR, "Failed to set journal checksum "
4272 goto failed_mount_wq;
4275 /* We have now updated the journal if required, so we can
4276 * validate the data journaling mode. */
4277 switch (test_opt(sb, DATA_FLAGS)) {
4279 /* No mode set, assume a default based on the journal
4280 * capabilities: ORDERED_DATA if the journal can
4281 * cope, else JOURNAL_DATA
4283 if (jbd2_journal_check_available_features
4284 (sbi->s_journal, 0, 0, JBD2_FEATURE_INCOMPAT_REVOKE))
4285 set_opt(sb, ORDERED_DATA);
4287 set_opt(sb, JOURNAL_DATA);
4290 case EXT4_MOUNT_ORDERED_DATA:
4291 case EXT4_MOUNT_WRITEBACK_DATA:
4292 if (!jbd2_journal_check_available_features
4293 (sbi->s_journal, 0, 0, JBD2_FEATURE_INCOMPAT_REVOKE)) {
4294 ext4_msg(sb, KERN_ERR, "Journal does not support "
4295 "requested data journaling mode");
4296 goto failed_mount_wq;
4302 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA &&
4303 test_opt(sb, JOURNAL_ASYNC_COMMIT)) {
4304 ext4_msg(sb, KERN_ERR, "can't mount with "
4305 "journal_async_commit in data=ordered mode");
4306 goto failed_mount_wq;
4309 set_task_ioprio(sbi->s_journal->j_task, journal_ioprio);
4311 sbi->s_journal->j_commit_callback = ext4_journal_commit_callback;
4314 if (!test_opt(sb, NO_MBCACHE)) {
4315 sbi->s_ea_block_cache = ext4_xattr_create_cache();
4316 if (!sbi->s_ea_block_cache) {
4317 ext4_msg(sb, KERN_ERR,
4318 "Failed to create ea_block_cache");
4319 goto failed_mount_wq;
4322 if (ext4_has_feature_ea_inode(sb)) {
4323 sbi->s_ea_inode_cache = ext4_xattr_create_cache();
4324 if (!sbi->s_ea_inode_cache) {
4325 ext4_msg(sb, KERN_ERR,
4326 "Failed to create ea_inode_cache");
4327 goto failed_mount_wq;
4332 if ((DUMMY_ENCRYPTION_ENABLED(sbi) || ext4_has_feature_encrypt(sb)) &&
4333 (blocksize != PAGE_SIZE)) {
4334 ext4_msg(sb, KERN_ERR,
4335 "Unsupported blocksize for fs encryption");
4336 goto failed_mount_wq;
4339 if (DUMMY_ENCRYPTION_ENABLED(sbi) && !sb_rdonly(sb) &&
4340 !ext4_has_feature_encrypt(sb)) {
4341 ext4_set_feature_encrypt(sb);
4342 ext4_commit_super(sb, 1);
4346 * Get the # of file system overhead blocks from the
4347 * superblock if present.
4349 sbi->s_overhead = le32_to_cpu(es->s_overhead_clusters);
4350 /* ignore the precalculated value if it is ridiculous */
4351 if (sbi->s_overhead > ext4_blocks_count(es))
4352 sbi->s_overhead = 0;
4354 * If the bigalloc feature is not enabled recalculating the
4355 * overhead doesn't take long, so we might as well just redo
4356 * it to make sure we are using the correct value.
4358 if (!ext4_has_feature_bigalloc(sb))
4359 sbi->s_overhead = 0;
4360 if (sbi->s_overhead == 0) {
4361 err = ext4_calculate_overhead(sb);
4363 goto failed_mount_wq;
4367 * The maximum number of concurrent works can be high and
4368 * concurrency isn't really necessary. Limit it to 1.
4370 EXT4_SB(sb)->rsv_conversion_wq =
4371 alloc_workqueue("ext4-rsv-conversion", WQ_MEM_RECLAIM | WQ_UNBOUND, 1);
4372 if (!EXT4_SB(sb)->rsv_conversion_wq) {
4373 printk(KERN_ERR "EXT4-fs: failed to create workqueue\n");
4379 * The jbd2_journal_load will have done any necessary log recovery,
4380 * so we can safely mount the rest of the filesystem now.
4383 root = ext4_iget(sb, EXT4_ROOT_INO, EXT4_IGET_SPECIAL);
4385 ext4_msg(sb, KERN_ERR, "get root inode failed");
4386 ret = PTR_ERR(root);
4390 if (!S_ISDIR(root->i_mode) || !root->i_blocks || !root->i_size) {
4391 ext4_msg(sb, KERN_ERR, "corrupt root inode, run e2fsck");
4395 sb->s_root = d_make_root(root);
4397 ext4_msg(sb, KERN_ERR, "get root dentry failed");
4402 if (ext4_setup_super(sb, es, sb_rdonly(sb)))
4403 sb->s_flags |= MS_RDONLY;
4405 ext4_set_resv_clusters(sb);
4407 err = ext4_setup_system_zone(sb);
4409 ext4_msg(sb, KERN_ERR, "failed to initialize system "
4411 goto failed_mount4a;
4415 err = ext4_mb_init(sb);
4417 ext4_msg(sb, KERN_ERR, "failed to initialize mballoc (%d)",
4422 block = ext4_count_free_clusters(sb);
4423 ext4_free_blocks_count_set(sbi->s_es,
4424 EXT4_C2B(sbi, block));
4425 ext4_superblock_csum_set(sb);
4426 err = percpu_counter_init(&sbi->s_freeclusters_counter, block,
4429 unsigned long freei = ext4_count_free_inodes(sb);
4430 sbi->s_es->s_free_inodes_count = cpu_to_le32(freei);
4431 ext4_superblock_csum_set(sb);
4432 err = percpu_counter_init(&sbi->s_freeinodes_counter, freei,
4436 err = percpu_counter_init(&sbi->s_dirs_counter,
4437 ext4_count_dirs(sb), GFP_KERNEL);
4439 err = percpu_counter_init(&sbi->s_dirtyclusters_counter, 0,
4442 err = percpu_init_rwsem(&sbi->s_writepages_rwsem);
4445 ext4_msg(sb, KERN_ERR, "insufficient memory");
4449 if (ext4_has_feature_flex_bg(sb))
4450 if (!ext4_fill_flex_info(sb)) {
4451 ext4_msg(sb, KERN_ERR,
4452 "unable to initialize "
4453 "flex_bg meta info!");
4457 err = ext4_register_li_request(sb, first_not_zeroed);
4461 err = ext4_register_sysfs(sb);
4466 /* Enable quota usage during mount. */
4467 if (ext4_has_feature_quota(sb) && !sb_rdonly(sb)) {
4468 err = ext4_enable_quotas(sb);
4472 #endif /* CONFIG_QUOTA */
4474 EXT4_SB(sb)->s_mount_state |= EXT4_ORPHAN_FS;
4475 ext4_orphan_cleanup(sb, es);
4476 EXT4_SB(sb)->s_mount_state &= ~EXT4_ORPHAN_FS;
4477 if (needs_recovery) {
4478 ext4_msg(sb, KERN_INFO, "recovery complete");
4479 err = ext4_mark_recovery_complete(sb, es);
4483 if (EXT4_SB(sb)->s_journal) {
4484 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA)
4485 descr = " journalled data mode";
4486 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA)
4487 descr = " ordered data mode";
4489 descr = " writeback data mode";
4491 descr = "out journal";
4493 if (test_opt(sb, DISCARD)) {
4494 struct request_queue *q = bdev_get_queue(sb->s_bdev);
4495 if (!blk_queue_discard(q))
4496 ext4_msg(sb, KERN_WARNING,
4497 "mounting with \"discard\" option, but "
4498 "the device does not support discard");
4501 if (___ratelimit(&ext4_mount_msg_ratelimit, "EXT4-fs mount"))
4502 ext4_msg(sb, KERN_INFO, "mounted filesystem with%s. "
4503 "Opts: %.*s%s%s", descr,
4504 (int) sizeof(sbi->s_es->s_mount_opts),
4505 sbi->s_es->s_mount_opts,
4506 *sbi->s_es->s_mount_opts ? "; " : "", orig_data);
4508 if (es->s_error_count)
4509 mod_timer(&sbi->s_err_report, jiffies + 300*HZ); /* 5 minutes */
4511 /* Enable message ratelimiting. Default is 10 messages per 5 secs. */
4512 ratelimit_state_init(&sbi->s_err_ratelimit_state, 5 * HZ, 10);
4513 ratelimit_state_init(&sbi->s_warning_ratelimit_state, 5 * HZ, 10);
4514 ratelimit_state_init(&sbi->s_msg_ratelimit_state, 5 * HZ, 10);
4521 ext4_msg(sb, KERN_ERR, "VFS: Can't find ext4 filesystem");
4525 ext4_unregister_sysfs(sb);
4526 kobject_put(&sbi->s_kobj);
4528 ext4_unregister_li_request(sb);
4530 ext4_mb_release(sb);
4532 flex_groups = rcu_dereference(sbi->s_flex_groups);
4534 for (i = 0; i < sbi->s_flex_groups_allocated; i++)
4535 kvfree(flex_groups[i]);
4536 kvfree(flex_groups);
4539 percpu_counter_destroy(&sbi->s_freeclusters_counter);
4540 percpu_counter_destroy(&sbi->s_freeinodes_counter);
4541 percpu_counter_destroy(&sbi->s_dirs_counter);
4542 percpu_counter_destroy(&sbi->s_dirtyclusters_counter);
4543 percpu_free_rwsem(&sbi->s_writepages_rwsem);
4545 ext4_ext_release(sb);
4546 ext4_release_system_zone(sb);
4551 ext4_msg(sb, KERN_ERR, "mount failed");
4552 if (EXT4_SB(sb)->rsv_conversion_wq)
4553 destroy_workqueue(EXT4_SB(sb)->rsv_conversion_wq);
4555 if (sbi->s_ea_inode_cache) {
4556 ext4_xattr_destroy_cache(sbi->s_ea_inode_cache);
4557 sbi->s_ea_inode_cache = NULL;
4559 if (sbi->s_ea_block_cache) {
4560 ext4_xattr_destroy_cache(sbi->s_ea_block_cache);
4561 sbi->s_ea_block_cache = NULL;
4563 if (sbi->s_journal) {
4564 jbd2_journal_destroy(sbi->s_journal);
4565 sbi->s_journal = NULL;
4568 ext4_es_unregister_shrinker(sbi);
4570 del_timer_sync(&sbi->s_err_report);
4572 kthread_stop(sbi->s_mmp_tsk);
4575 group_desc = rcu_dereference(sbi->s_group_desc);
4576 for (i = 0; i < db_count; i++)
4577 brelse(group_desc[i]);
4581 if (sbi->s_chksum_driver)
4582 crypto_free_shash(sbi->s_chksum_driver);
4584 for (i = 0; i < EXT4_MAXQUOTAS; i++)
4585 kfree(sbi->s_qf_names[i]);
4587 ext4_blkdev_remove(sbi);
4590 sb->s_fs_info = NULL;
4591 kfree(sbi->s_blockgroup_lock);
4595 fs_put_dax(dax_dev);
4596 return err ? err : ret;
4600 * Setup any per-fs journal parameters now. We'll do this both on
4601 * initial mount, once the journal has been initialised but before we've
4602 * done any recovery; and again on any subsequent remount.
4604 static void ext4_init_journal_params(struct super_block *sb, journal_t *journal)
4606 struct ext4_sb_info *sbi = EXT4_SB(sb);
4608 journal->j_commit_interval = sbi->s_commit_interval;
4609 journal->j_min_batch_time = sbi->s_min_batch_time;
4610 journal->j_max_batch_time = sbi->s_max_batch_time;
4612 write_lock(&journal->j_state_lock);
4613 if (test_opt(sb, BARRIER))
4614 journal->j_flags |= JBD2_BARRIER;
4616 journal->j_flags &= ~JBD2_BARRIER;
4617 if (test_opt(sb, DATA_ERR_ABORT))
4618 journal->j_flags |= JBD2_ABORT_ON_SYNCDATA_ERR;
4620 journal->j_flags &= ~JBD2_ABORT_ON_SYNCDATA_ERR;
4621 write_unlock(&journal->j_state_lock);
4624 static struct inode *ext4_get_journal_inode(struct super_block *sb,
4625 unsigned int journal_inum)
4627 struct inode *journal_inode;
4630 * Test for the existence of a valid inode on disk. Bad things
4631 * happen if we iget() an unused inode, as the subsequent iput()
4632 * will try to delete it.
4634 journal_inode = ext4_iget(sb, journal_inum, EXT4_IGET_SPECIAL);
4635 if (IS_ERR(journal_inode)) {
4636 ext4_msg(sb, KERN_ERR, "no journal found");
4639 if (!journal_inode->i_nlink) {
4640 make_bad_inode(journal_inode);
4641 iput(journal_inode);
4642 ext4_msg(sb, KERN_ERR, "journal inode is deleted");
4646 jbd_debug(2, "Journal inode found at %p: %lld bytes\n",
4647 journal_inode, journal_inode->i_size);
4648 if (!S_ISREG(journal_inode->i_mode)) {
4649 ext4_msg(sb, KERN_ERR, "invalid journal inode");
4650 iput(journal_inode);
4653 return journal_inode;
4656 static journal_t *ext4_get_journal(struct super_block *sb,
4657 unsigned int journal_inum)
4659 struct inode *journal_inode;
4662 if (WARN_ON_ONCE(!ext4_has_feature_journal(sb)))
4665 journal_inode = ext4_get_journal_inode(sb, journal_inum);
4669 journal = jbd2_journal_init_inode(journal_inode);
4671 ext4_msg(sb, KERN_ERR, "Could not load journal inode");
4672 iput(journal_inode);
4675 journal->j_private = sb;
4676 ext4_init_journal_params(sb, journal);
4680 static journal_t *ext4_get_dev_journal(struct super_block *sb,
4683 struct buffer_head *bh;
4687 int hblock, blocksize;
4688 ext4_fsblk_t sb_block;
4689 unsigned long offset;
4690 struct ext4_super_block *es;
4691 struct block_device *bdev;
4693 if (WARN_ON_ONCE(!ext4_has_feature_journal(sb)))
4696 bdev = ext4_blkdev_get(j_dev, sb);
4700 blocksize = sb->s_blocksize;
4701 hblock = bdev_logical_block_size(bdev);
4702 if (blocksize < hblock) {
4703 ext4_msg(sb, KERN_ERR,
4704 "blocksize too small for journal device");
4708 sb_block = EXT4_MIN_BLOCK_SIZE / blocksize;
4709 offset = EXT4_MIN_BLOCK_SIZE % blocksize;
4710 set_blocksize(bdev, blocksize);
4711 if (!(bh = __bread(bdev, sb_block, blocksize))) {
4712 ext4_msg(sb, KERN_ERR, "couldn't read superblock of "
4713 "external journal");
4717 es = (struct ext4_super_block *) (bh->b_data + offset);
4718 if ((le16_to_cpu(es->s_magic) != EXT4_SUPER_MAGIC) ||
4719 !(le32_to_cpu(es->s_feature_incompat) &
4720 EXT4_FEATURE_INCOMPAT_JOURNAL_DEV)) {
4721 ext4_msg(sb, KERN_ERR, "external journal has "
4727 if ((le32_to_cpu(es->s_feature_ro_compat) &
4728 EXT4_FEATURE_RO_COMPAT_METADATA_CSUM) &&
4729 es->s_checksum != ext4_superblock_csum(sb, es)) {
4730 ext4_msg(sb, KERN_ERR, "external journal has "
4731 "corrupt superblock");
4736 if (memcmp(EXT4_SB(sb)->s_es->s_journal_uuid, es->s_uuid, 16)) {
4737 ext4_msg(sb, KERN_ERR, "journal UUID does not match");
4742 len = ext4_blocks_count(es);
4743 start = sb_block + 1;
4744 brelse(bh); /* we're done with the superblock */
4746 journal = jbd2_journal_init_dev(bdev, sb->s_bdev,
4747 start, len, blocksize);
4749 ext4_msg(sb, KERN_ERR, "failed to create device journal");
4752 journal->j_private = sb;
4753 ll_rw_block(REQ_OP_READ, REQ_META | REQ_PRIO, 1, &journal->j_sb_buffer);
4754 wait_on_buffer(journal->j_sb_buffer);
4755 if (!buffer_uptodate(journal->j_sb_buffer)) {
4756 ext4_msg(sb, KERN_ERR, "I/O error on journal device");
4759 if (be32_to_cpu(journal->j_superblock->s_nr_users) != 1) {
4760 ext4_msg(sb, KERN_ERR, "External journal has more than one "
4761 "user (unsupported) - %d",
4762 be32_to_cpu(journal->j_superblock->s_nr_users));
4765 EXT4_SB(sb)->journal_bdev = bdev;
4766 ext4_init_journal_params(sb, journal);
4770 jbd2_journal_destroy(journal);
4772 ext4_blkdev_put(bdev);
4776 static int ext4_load_journal(struct super_block *sb,
4777 struct ext4_super_block *es,
4778 unsigned long journal_devnum)
4781 unsigned int journal_inum = le32_to_cpu(es->s_journal_inum);
4784 int really_read_only;
4786 if (WARN_ON_ONCE(!ext4_has_feature_journal(sb)))
4787 return -EFSCORRUPTED;
4789 if (journal_devnum &&
4790 journal_devnum != le32_to_cpu(es->s_journal_dev)) {
4791 ext4_msg(sb, KERN_INFO, "external journal device major/minor "
4792 "numbers have changed");
4793 journal_dev = new_decode_dev(journal_devnum);
4795 journal_dev = new_decode_dev(le32_to_cpu(es->s_journal_dev));
4797 really_read_only = bdev_read_only(sb->s_bdev);
4800 * Are we loading a blank journal or performing recovery after a
4801 * crash? For recovery, we need to check in advance whether we
4802 * can get read-write access to the device.
4804 if (ext4_has_feature_journal_needs_recovery(sb)) {
4805 if (sb_rdonly(sb)) {
4806 ext4_msg(sb, KERN_INFO, "INFO: recovery "
4807 "required on readonly filesystem");
4808 if (really_read_only) {
4809 ext4_msg(sb, KERN_ERR, "write access "
4810 "unavailable, cannot proceed");
4813 ext4_msg(sb, KERN_INFO, "write access will "
4814 "be enabled during recovery");
4818 if (journal_inum && journal_dev) {
4819 ext4_msg(sb, KERN_ERR, "filesystem has both journal "
4820 "and inode journals!");
4825 if (!(journal = ext4_get_journal(sb, journal_inum)))
4828 if (!(journal = ext4_get_dev_journal(sb, journal_dev)))
4832 if (!(journal->j_flags & JBD2_BARRIER))
4833 ext4_msg(sb, KERN_INFO, "barriers disabled");
4835 if (!ext4_has_feature_journal_needs_recovery(sb))
4836 err = jbd2_journal_wipe(journal, !really_read_only);
4838 char *save = kmalloc(EXT4_S_ERR_LEN, GFP_KERNEL);
4840 memcpy(save, ((char *) es) +
4841 EXT4_S_ERR_START, EXT4_S_ERR_LEN);
4842 err = jbd2_journal_load(journal);
4844 memcpy(((char *) es) + EXT4_S_ERR_START,
4845 save, EXT4_S_ERR_LEN);
4850 ext4_msg(sb, KERN_ERR, "error loading journal");
4851 jbd2_journal_destroy(journal);
4855 EXT4_SB(sb)->s_journal = journal;
4856 err = ext4_clear_journal_err(sb, es);
4858 EXT4_SB(sb)->s_journal = NULL;
4859 jbd2_journal_destroy(journal);
4863 if (!really_read_only && journal_devnum &&
4864 journal_devnum != le32_to_cpu(es->s_journal_dev)) {
4865 es->s_journal_dev = cpu_to_le32(journal_devnum);
4867 /* Make sure we flush the recovery flag to disk. */
4868 ext4_commit_super(sb, 1);
4874 static int ext4_commit_super(struct super_block *sb, int sync)
4876 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
4877 struct buffer_head *sbh = EXT4_SB(sb)->s_sbh;
4882 if (block_device_ejected(sb))
4886 * If the file system is mounted read-only, don't update the
4887 * superblock write time. This avoids updating the superblock
4888 * write time when we are mounting the root file system
4889 * read/only but we need to replay the journal; at that point,
4890 * for people who are east of GMT and who make their clock
4891 * tick in localtime for Windows bug-for-bug compatibility,
4892 * the clock is set in the future, and this will cause e2fsck
4893 * to complain and force a full file system check.
4895 if (!(sb->s_flags & MS_RDONLY))
4896 es->s_wtime = cpu_to_le32(get_seconds());
4897 if (sb->s_bdev->bd_part)
4898 es->s_kbytes_written =
4899 cpu_to_le64(EXT4_SB(sb)->s_kbytes_written +
4900 ((part_stat_read(sb->s_bdev->bd_part, sectors[1]) -
4901 EXT4_SB(sb)->s_sectors_written_start) >> 1));
4903 es->s_kbytes_written =
4904 cpu_to_le64(EXT4_SB(sb)->s_kbytes_written);
4905 if (percpu_counter_initialized(&EXT4_SB(sb)->s_freeclusters_counter))
4906 ext4_free_blocks_count_set(es,
4907 EXT4_C2B(EXT4_SB(sb), percpu_counter_sum_positive(
4908 &EXT4_SB(sb)->s_freeclusters_counter)));
4909 if (percpu_counter_initialized(&EXT4_SB(sb)->s_freeinodes_counter))
4910 es->s_free_inodes_count =
4911 cpu_to_le32(percpu_counter_sum_positive(
4912 &EXT4_SB(sb)->s_freeinodes_counter));
4913 BUFFER_TRACE(sbh, "marking dirty");
4914 ext4_superblock_csum_set(sb);
4917 if (buffer_write_io_error(sbh) || !buffer_uptodate(sbh)) {
4919 * Oh, dear. A previous attempt to write the
4920 * superblock failed. This could happen because the
4921 * USB device was yanked out. Or it could happen to
4922 * be a transient write error and maybe the block will
4923 * be remapped. Nothing we can do but to retry the
4924 * write and hope for the best.
4926 ext4_msg(sb, KERN_ERR, "previous I/O error to "
4927 "superblock detected");
4928 clear_buffer_write_io_error(sbh);
4929 set_buffer_uptodate(sbh);
4931 mark_buffer_dirty(sbh);
4934 error = __sync_dirty_buffer(sbh,
4935 REQ_SYNC | (test_opt(sb, BARRIER) ? REQ_FUA : 0));
4939 error = buffer_write_io_error(sbh);
4941 ext4_msg(sb, KERN_ERR, "I/O error while writing "
4943 clear_buffer_write_io_error(sbh);
4944 set_buffer_uptodate(sbh);
4951 * Have we just finished recovery? If so, and if we are mounting (or
4952 * remounting) the filesystem readonly, then we will end up with a
4953 * consistent fs on disk. Record that fact.
4955 static int ext4_mark_recovery_complete(struct super_block *sb,
4956 struct ext4_super_block *es)
4959 journal_t *journal = EXT4_SB(sb)->s_journal;
4961 if (!ext4_has_feature_journal(sb)) {
4962 if (journal != NULL) {
4963 ext4_error(sb, "Journal got removed while the fs was "
4965 return -EFSCORRUPTED;
4969 jbd2_journal_lock_updates(journal);
4970 err = jbd2_journal_flush(journal);
4974 if (ext4_has_feature_journal_needs_recovery(sb) && sb_rdonly(sb)) {
4975 ext4_clear_feature_journal_needs_recovery(sb);
4976 ext4_commit_super(sb, 1);
4979 jbd2_journal_unlock_updates(journal);
4984 * If we are mounting (or read-write remounting) a filesystem whose journal
4985 * has recorded an error from a previous lifetime, move that error to the
4986 * main filesystem now.
4988 static int ext4_clear_journal_err(struct super_block *sb,
4989 struct ext4_super_block *es)
4995 if (!ext4_has_feature_journal(sb)) {
4996 ext4_error(sb, "Journal got removed while the fs was mounted!");
4997 return -EFSCORRUPTED;
5000 journal = EXT4_SB(sb)->s_journal;
5003 * Now check for any error status which may have been recorded in the
5004 * journal by a prior ext4_error() or ext4_abort()
5007 j_errno = jbd2_journal_errno(journal);
5011 errstr = ext4_decode_error(sb, j_errno, nbuf);
5012 ext4_warning(sb, "Filesystem error recorded "
5013 "from previous mount: %s", errstr);
5014 ext4_warning(sb, "Marking fs in need of filesystem check.");
5016 EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS;
5017 es->s_state |= cpu_to_le16(EXT4_ERROR_FS);
5018 ext4_commit_super(sb, 1);
5020 jbd2_journal_clear_err(journal);
5021 jbd2_journal_update_sb_errno(journal);
5027 * Force the running and committing transactions to commit,
5028 * and wait on the commit.
5030 int ext4_force_commit(struct super_block *sb)
5037 journal = EXT4_SB(sb)->s_journal;
5038 return ext4_journal_force_commit(journal);
5041 static int ext4_sync_fs(struct super_block *sb, int wait)
5045 bool needs_barrier = false;
5046 struct ext4_sb_info *sbi = EXT4_SB(sb);
5048 if (unlikely(ext4_forced_shutdown(EXT4_SB(sb))))
5051 trace_ext4_sync_fs(sb, wait);
5052 flush_workqueue(sbi->rsv_conversion_wq);
5054 * Writeback quota in non-journalled quota case - journalled quota has
5057 dquot_writeback_dquots(sb, -1);
5059 * Data writeback is possible w/o journal transaction, so barrier must
5060 * being sent at the end of the function. But we can skip it if
5061 * transaction_commit will do it for us.
5063 if (sbi->s_journal) {
5064 target = jbd2_get_latest_transaction(sbi->s_journal);
5065 if (wait && sbi->s_journal->j_flags & JBD2_BARRIER &&
5066 !jbd2_trans_will_send_data_barrier(sbi->s_journal, target))
5067 needs_barrier = true;
5069 if (jbd2_journal_start_commit(sbi->s_journal, &target)) {
5071 ret = jbd2_log_wait_commit(sbi->s_journal,
5074 } else if (wait && test_opt(sb, BARRIER))
5075 needs_barrier = true;
5076 if (needs_barrier) {
5078 err = blkdev_issue_flush(sb->s_bdev, GFP_KERNEL, NULL);
5087 * LVM calls this function before a (read-only) snapshot is created. This
5088 * gives us a chance to flush the journal completely and mark the fs clean.
5090 * Note that only this function cannot bring a filesystem to be in a clean
5091 * state independently. It relies on upper layer to stop all data & metadata
5094 static int ext4_freeze(struct super_block *sb)
5102 journal = EXT4_SB(sb)->s_journal;
5105 /* Now we set up the journal barrier. */
5106 jbd2_journal_lock_updates(journal);
5109 * Don't clear the needs_recovery flag if we failed to
5110 * flush the journal.
5112 error = jbd2_journal_flush(journal);
5116 /* Journal blocked and flushed, clear needs_recovery flag. */
5117 ext4_clear_feature_journal_needs_recovery(sb);
5120 error = ext4_commit_super(sb, 1);
5123 /* we rely on upper layer to stop further updates */
5124 jbd2_journal_unlock_updates(journal);
5129 * Called by LVM after the snapshot is done. We need to reset the RECOVER
5130 * flag here, even though the filesystem is not technically dirty yet.
5132 static int ext4_unfreeze(struct super_block *sb)
5134 if (sb_rdonly(sb) || ext4_forced_shutdown(EXT4_SB(sb)))
5137 if (EXT4_SB(sb)->s_journal) {
5138 /* Reset the needs_recovery flag before the fs is unlocked. */
5139 ext4_set_feature_journal_needs_recovery(sb);
5142 ext4_commit_super(sb, 1);
5147 * Structure to save mount options for ext4_remount's benefit
5149 struct ext4_mount_options {
5150 unsigned long s_mount_opt;
5151 unsigned long s_mount_opt2;
5154 unsigned long s_commit_interval;
5155 u32 s_min_batch_time, s_max_batch_time;
5158 char *s_qf_names[EXT4_MAXQUOTAS];
5162 static int ext4_remount(struct super_block *sb, int *flags, char *data)
5164 struct ext4_super_block *es;
5165 struct ext4_sb_info *sbi = EXT4_SB(sb);
5166 unsigned long old_sb_flags;
5167 struct ext4_mount_options old_opts;
5168 int enable_quota = 0;
5170 unsigned int journal_ioprio = DEFAULT_JOURNAL_IOPRIO;
5174 char *to_free[EXT4_MAXQUOTAS];
5176 char *orig_data = kstrdup(data, GFP_KERNEL);
5178 /* Store the original options */
5179 old_sb_flags = sb->s_flags;
5180 old_opts.s_mount_opt = sbi->s_mount_opt;
5181 old_opts.s_mount_opt2 = sbi->s_mount_opt2;
5182 old_opts.s_resuid = sbi->s_resuid;
5183 old_opts.s_resgid = sbi->s_resgid;
5184 old_opts.s_commit_interval = sbi->s_commit_interval;
5185 old_opts.s_min_batch_time = sbi->s_min_batch_time;
5186 old_opts.s_max_batch_time = sbi->s_max_batch_time;
5188 old_opts.s_jquota_fmt = sbi->s_jquota_fmt;
5189 for (i = 0; i < EXT4_MAXQUOTAS; i++)
5190 if (sbi->s_qf_names[i]) {
5191 char *qf_name = get_qf_name(sb, sbi, i);
5193 old_opts.s_qf_names[i] = kstrdup(qf_name, GFP_KERNEL);
5194 if (!old_opts.s_qf_names[i]) {
5195 for (j = 0; j < i; j++)
5196 kfree(old_opts.s_qf_names[j]);
5201 old_opts.s_qf_names[i] = NULL;
5203 if (sbi->s_journal && sbi->s_journal->j_task->io_context)
5204 journal_ioprio = sbi->s_journal->j_task->io_context->ioprio;
5206 if (!parse_options(data, sb, NULL, &journal_ioprio, 1)) {
5211 if ((old_opts.s_mount_opt & EXT4_MOUNT_JOURNAL_CHECKSUM) ^
5212 test_opt(sb, JOURNAL_CHECKSUM)) {
5213 ext4_msg(sb, KERN_ERR, "changing journal_checksum "
5214 "during remount not supported; ignoring");
5215 sbi->s_mount_opt ^= EXT4_MOUNT_JOURNAL_CHECKSUM;
5218 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA) {
5219 if (test_opt2(sb, EXPLICIT_DELALLOC)) {
5220 ext4_msg(sb, KERN_ERR, "can't mount with "
5221 "both data=journal and delalloc");
5225 if (test_opt(sb, DIOREAD_NOLOCK)) {
5226 ext4_msg(sb, KERN_ERR, "can't mount with "
5227 "both data=journal and dioread_nolock");
5231 if (test_opt(sb, DAX)) {
5232 ext4_msg(sb, KERN_ERR, "can't mount with "
5233 "both data=journal and dax");
5237 } else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA) {
5238 if (test_opt(sb, JOURNAL_ASYNC_COMMIT)) {
5239 ext4_msg(sb, KERN_ERR, "can't mount with "
5240 "journal_async_commit in data=ordered mode");
5246 if ((sbi->s_mount_opt ^ old_opts.s_mount_opt) & EXT4_MOUNT_NO_MBCACHE) {
5247 ext4_msg(sb, KERN_ERR, "can't enable nombcache during remount");
5252 if ((sbi->s_mount_opt ^ old_opts.s_mount_opt) & EXT4_MOUNT_DAX) {
5253 ext4_msg(sb, KERN_WARNING, "warning: refusing change of "
5254 "dax flag with busy inodes while remounting");
5255 sbi->s_mount_opt ^= EXT4_MOUNT_DAX;
5258 if (sbi->s_mount_flags & EXT4_MF_FS_ABORTED)
5259 ext4_abort(sb, "Abort forced by user");
5261 sb->s_flags = (sb->s_flags & ~MS_POSIXACL) |
5262 (test_opt(sb, POSIX_ACL) ? MS_POSIXACL : 0);
5266 if (sbi->s_journal) {
5267 ext4_init_journal_params(sb, sbi->s_journal);
5268 set_task_ioprio(sbi->s_journal->j_task, journal_ioprio);
5271 if (*flags & MS_LAZYTIME)
5272 sb->s_flags |= MS_LAZYTIME;
5274 if ((bool)(*flags & MS_RDONLY) != sb_rdonly(sb)) {
5275 if (sbi->s_mount_flags & EXT4_MF_FS_ABORTED) {
5280 if (*flags & MS_RDONLY) {
5281 err = sync_filesystem(sb);
5284 err = dquot_suspend(sb, -1);
5289 * First of all, the unconditional stuff we have to do
5290 * to disable replay of the journal when we next remount
5292 sb->s_flags |= MS_RDONLY;
5295 * OK, test if we are remounting a valid rw partition
5296 * readonly, and if so set the rdonly flag and then
5297 * mark the partition as valid again.
5299 if (!(es->s_state & cpu_to_le16(EXT4_VALID_FS)) &&
5300 (sbi->s_mount_state & EXT4_VALID_FS))
5301 es->s_state = cpu_to_le16(sbi->s_mount_state);
5303 if (sbi->s_journal) {
5305 * We let remount-ro finish even if marking fs
5306 * as clean failed...
5308 ext4_mark_recovery_complete(sb, es);
5311 kthread_stop(sbi->s_mmp_tsk);
5313 /* Make sure we can mount this feature set readwrite */
5314 if (ext4_has_feature_readonly(sb) ||
5315 !ext4_feature_set_ok(sb, 0)) {
5320 * Make sure the group descriptor checksums
5321 * are sane. If they aren't, refuse to remount r/w.
5323 for (g = 0; g < sbi->s_groups_count; g++) {
5324 struct ext4_group_desc *gdp =
5325 ext4_get_group_desc(sb, g, NULL);
5327 if (!ext4_group_desc_csum_verify(sb, g, gdp)) {
5328 ext4_msg(sb, KERN_ERR,
5329 "ext4_remount: Checksum for group %u failed (%u!=%u)",
5330 g, le16_to_cpu(ext4_group_desc_csum(sb, g, gdp)),
5331 le16_to_cpu(gdp->bg_checksum));
5338 * If we have an unprocessed orphan list hanging
5339 * around from a previously readonly bdev mount,
5340 * require a full umount/remount for now.
5342 if (es->s_last_orphan) {
5343 ext4_msg(sb, KERN_WARNING, "Couldn't "
5344 "remount RDWR because of unprocessed "
5345 "orphan inode list. Please "
5346 "umount/remount instead");
5352 * Mounting a RDONLY partition read-write, so reread
5353 * and store the current valid flag. (It may have
5354 * been changed by e2fsck since we originally mounted
5357 if (sbi->s_journal) {
5358 err = ext4_clear_journal_err(sb, es);
5362 sbi->s_mount_state = le16_to_cpu(es->s_state);
5363 if (!ext4_setup_super(sb, es, 0))
5364 sb->s_flags &= ~MS_RDONLY;
5365 if (ext4_has_feature_mmp(sb))
5366 if (ext4_multi_mount_protect(sb,
5367 le64_to_cpu(es->s_mmp_block))) {
5376 * Reinitialize lazy itable initialization thread based on
5379 if (sb_rdonly(sb) || !test_opt(sb, INIT_INODE_TABLE))
5380 ext4_unregister_li_request(sb);
5382 ext4_group_t first_not_zeroed;
5383 first_not_zeroed = ext4_has_uninit_itable(sb);
5384 ext4_register_li_request(sb, first_not_zeroed);
5387 err = ext4_setup_system_zone(sb);
5391 if (sbi->s_journal == NULL && !(old_sb_flags & MS_RDONLY))
5392 ext4_commit_super(sb, 1);
5395 /* Release old quota file names */
5396 for (i = 0; i < EXT4_MAXQUOTAS; i++)
5397 kfree(old_opts.s_qf_names[i]);
5399 if (sb_any_quota_suspended(sb))
5400 dquot_resume(sb, -1);
5401 else if (ext4_has_feature_quota(sb)) {
5402 err = ext4_enable_quotas(sb);
5409 *flags = (*flags & ~MS_LAZYTIME) | (sb->s_flags & MS_LAZYTIME);
5410 ext4_msg(sb, KERN_INFO, "re-mounted. Opts: %s", orig_data);
5415 sb->s_flags = old_sb_flags;
5416 sbi->s_mount_opt = old_opts.s_mount_opt;
5417 sbi->s_mount_opt2 = old_opts.s_mount_opt2;
5418 sbi->s_resuid = old_opts.s_resuid;
5419 sbi->s_resgid = old_opts.s_resgid;
5420 sbi->s_commit_interval = old_opts.s_commit_interval;
5421 sbi->s_min_batch_time = old_opts.s_min_batch_time;
5422 sbi->s_max_batch_time = old_opts.s_max_batch_time;
5424 sbi->s_jquota_fmt = old_opts.s_jquota_fmt;
5425 for (i = 0; i < EXT4_MAXQUOTAS; i++) {
5426 to_free[i] = get_qf_name(sb, sbi, i);
5427 rcu_assign_pointer(sbi->s_qf_names[i], old_opts.s_qf_names[i]);
5430 for (i = 0; i < EXT4_MAXQUOTAS; i++)
5438 static int ext4_statfs_project(struct super_block *sb,
5439 kprojid_t projid, struct kstatfs *buf)
5442 struct dquot *dquot;
5446 qid = make_kqid_projid(projid);
5447 dquot = dqget(sb, qid);
5449 return PTR_ERR(dquot);
5450 spin_lock(&dquot->dq_dqb_lock);
5452 limit = (dquot->dq_dqb.dqb_bsoftlimit ?
5453 dquot->dq_dqb.dqb_bsoftlimit :
5454 dquot->dq_dqb.dqb_bhardlimit) >> sb->s_blocksize_bits;
5455 if (limit && buf->f_blocks > limit) {
5456 curblock = dquot->dq_dqb.dqb_curspace >> sb->s_blocksize_bits;
5457 buf->f_blocks = limit;
5458 buf->f_bfree = buf->f_bavail =
5459 (buf->f_blocks > curblock) ?
5460 (buf->f_blocks - curblock) : 0;
5463 limit = dquot->dq_dqb.dqb_isoftlimit ?
5464 dquot->dq_dqb.dqb_isoftlimit :
5465 dquot->dq_dqb.dqb_ihardlimit;
5466 if (limit && buf->f_files > limit) {
5467 buf->f_files = limit;
5469 (buf->f_files > dquot->dq_dqb.dqb_curinodes) ?
5470 (buf->f_files - dquot->dq_dqb.dqb_curinodes) : 0;
5473 spin_unlock(&dquot->dq_dqb_lock);
5479 static int ext4_statfs(struct dentry *dentry, struct kstatfs *buf)
5481 struct super_block *sb = dentry->d_sb;
5482 struct ext4_sb_info *sbi = EXT4_SB(sb);
5483 struct ext4_super_block *es = sbi->s_es;
5484 ext4_fsblk_t overhead = 0, resv_blocks;
5487 resv_blocks = EXT4_C2B(sbi, atomic64_read(&sbi->s_resv_clusters));
5489 if (!test_opt(sb, MINIX_DF))
5490 overhead = sbi->s_overhead;
5492 buf->f_type = EXT4_SUPER_MAGIC;
5493 buf->f_bsize = sb->s_blocksize;
5494 buf->f_blocks = ext4_blocks_count(es) - EXT4_C2B(sbi, overhead);
5495 bfree = percpu_counter_sum_positive(&sbi->s_freeclusters_counter) -
5496 percpu_counter_sum_positive(&sbi->s_dirtyclusters_counter);
5497 /* prevent underflow in case that few free space is available */
5498 buf->f_bfree = EXT4_C2B(sbi, max_t(s64, bfree, 0));
5499 buf->f_bavail = buf->f_bfree -
5500 (ext4_r_blocks_count(es) + resv_blocks);
5501 if (buf->f_bfree < (ext4_r_blocks_count(es) + resv_blocks))
5503 buf->f_files = le32_to_cpu(es->s_inodes_count);
5504 buf->f_ffree = percpu_counter_sum_positive(&sbi->s_freeinodes_counter);
5505 buf->f_namelen = EXT4_NAME_LEN;
5506 fsid = le64_to_cpup((void *)es->s_uuid) ^
5507 le64_to_cpup((void *)es->s_uuid + sizeof(u64));
5508 buf->f_fsid.val[0] = fsid & 0xFFFFFFFFUL;
5509 buf->f_fsid.val[1] = (fsid >> 32) & 0xFFFFFFFFUL;
5512 if (ext4_test_inode_flag(dentry->d_inode, EXT4_INODE_PROJINHERIT) &&
5513 sb_has_quota_limits_enabled(sb, PRJQUOTA))
5514 ext4_statfs_project(sb, EXT4_I(dentry->d_inode)->i_projid, buf);
5523 * Helper functions so that transaction is started before we acquire dqio_sem
5524 * to keep correct lock ordering of transaction > dqio_sem
5526 static inline struct inode *dquot_to_inode(struct dquot *dquot)
5528 return sb_dqopt(dquot->dq_sb)->files[dquot->dq_id.type];
5531 static int ext4_write_dquot(struct dquot *dquot)
5535 struct inode *inode;
5537 inode = dquot_to_inode(dquot);
5538 handle = ext4_journal_start(inode, EXT4_HT_QUOTA,
5539 EXT4_QUOTA_TRANS_BLOCKS(dquot->dq_sb));
5541 return PTR_ERR(handle);
5542 ret = dquot_commit(dquot);
5543 err = ext4_journal_stop(handle);
5549 static int ext4_acquire_dquot(struct dquot *dquot)
5554 handle = ext4_journal_start(dquot_to_inode(dquot), EXT4_HT_QUOTA,
5555 EXT4_QUOTA_INIT_BLOCKS(dquot->dq_sb));
5557 return PTR_ERR(handle);
5558 ret = dquot_acquire(dquot);
5559 err = ext4_journal_stop(handle);
5565 static int ext4_release_dquot(struct dquot *dquot)
5570 handle = ext4_journal_start(dquot_to_inode(dquot), EXT4_HT_QUOTA,
5571 EXT4_QUOTA_DEL_BLOCKS(dquot->dq_sb));
5572 if (IS_ERR(handle)) {
5573 /* Release dquot anyway to avoid endless cycle in dqput() */
5574 dquot_release(dquot);
5575 return PTR_ERR(handle);
5577 ret = dquot_release(dquot);
5578 err = ext4_journal_stop(handle);
5584 static int ext4_mark_dquot_dirty(struct dquot *dquot)
5586 struct super_block *sb = dquot->dq_sb;
5587 struct ext4_sb_info *sbi = EXT4_SB(sb);
5589 /* Are we journaling quotas? */
5590 if (ext4_has_feature_quota(sb) ||
5591 sbi->s_qf_names[USRQUOTA] || sbi->s_qf_names[GRPQUOTA]) {
5592 dquot_mark_dquot_dirty(dquot);
5593 return ext4_write_dquot(dquot);
5595 return dquot_mark_dquot_dirty(dquot);
5599 static int ext4_write_info(struct super_block *sb, int type)
5604 /* Data block + inode block */
5605 handle = ext4_journal_start_sb(sb, EXT4_HT_QUOTA, 2);
5607 return PTR_ERR(handle);
5608 ret = dquot_commit_info(sb, type);
5609 err = ext4_journal_stop(handle);
5616 * Turn on quotas during mount time - we need to find
5617 * the quota file and such...
5619 static int ext4_quota_on_mount(struct super_block *sb, int type)
5621 return dquot_quota_on_mount(sb, get_qf_name(sb, EXT4_SB(sb), type),
5622 EXT4_SB(sb)->s_jquota_fmt, type);
5625 static void lockdep_set_quota_inode(struct inode *inode, int subclass)
5627 struct ext4_inode_info *ei = EXT4_I(inode);
5629 /* The first argument of lockdep_set_subclass has to be
5630 * *exactly* the same as the argument to init_rwsem() --- in
5631 * this case, in init_once() --- or lockdep gets unhappy
5632 * because the name of the lock is set using the
5633 * stringification of the argument to init_rwsem().
5635 (void) ei; /* shut up clang warning if !CONFIG_LOCKDEP */
5636 lockdep_set_subclass(&ei->i_data_sem, subclass);
5640 * Standard function to be called on quota_on
5642 static int ext4_quota_on(struct super_block *sb, int type, int format_id,
5643 const struct path *path)
5647 if (!test_opt(sb, QUOTA))
5650 /* Quotafile not on the same filesystem? */
5651 if (path->dentry->d_sb != sb)
5654 /* Quota already enabled for this file? */
5655 if (IS_NOQUOTA(d_inode(path->dentry)))
5658 /* Journaling quota? */
5659 if (EXT4_SB(sb)->s_qf_names[type]) {
5660 /* Quotafile not in fs root? */
5661 if (path->dentry->d_parent != sb->s_root)
5662 ext4_msg(sb, KERN_WARNING,
5663 "Quota file not on filesystem root. "
5664 "Journaled quota will not work");
5665 sb_dqopt(sb)->flags |= DQUOT_NOLIST_DIRTY;
5668 * Clear the flag just in case mount options changed since
5671 sb_dqopt(sb)->flags &= ~DQUOT_NOLIST_DIRTY;
5675 * When we journal data on quota file, we have to flush journal to see
5676 * all updates to the file when we bypass pagecache...
5678 if (EXT4_SB(sb)->s_journal &&
5679 ext4_should_journal_data(d_inode(path->dentry))) {
5681 * We don't need to lock updates but journal_flush() could
5682 * otherwise be livelocked...
5684 jbd2_journal_lock_updates(EXT4_SB(sb)->s_journal);
5685 err = jbd2_journal_flush(EXT4_SB(sb)->s_journal);
5686 jbd2_journal_unlock_updates(EXT4_SB(sb)->s_journal);
5691 lockdep_set_quota_inode(path->dentry->d_inode, I_DATA_SEM_QUOTA);
5692 err = dquot_quota_on(sb, type, format_id, path);
5694 struct inode *inode = d_inode(path->dentry);
5698 * Set inode flags to prevent userspace from messing with quota
5699 * files. If this fails, we return success anyway since quotas
5700 * are already enabled and this is not a hard failure.
5703 handle = ext4_journal_start(inode, EXT4_HT_QUOTA, 1);
5706 EXT4_I(inode)->i_flags |= EXT4_NOATIME_FL | EXT4_IMMUTABLE_FL;
5707 inode_set_flags(inode, S_NOATIME | S_IMMUTABLE,
5708 S_NOATIME | S_IMMUTABLE);
5709 ext4_mark_inode_dirty(handle, inode);
5710 ext4_journal_stop(handle);
5712 inode_unlock(inode);
5714 dquot_quota_off(sb, type);
5717 lockdep_set_quota_inode(path->dentry->d_inode,
5722 static int ext4_quota_enable(struct super_block *sb, int type, int format_id,
5726 struct inode *qf_inode;
5727 unsigned long qf_inums[EXT4_MAXQUOTAS] = {
5728 le32_to_cpu(EXT4_SB(sb)->s_es->s_usr_quota_inum),
5729 le32_to_cpu(EXT4_SB(sb)->s_es->s_grp_quota_inum),
5730 le32_to_cpu(EXT4_SB(sb)->s_es->s_prj_quota_inum)
5733 BUG_ON(!ext4_has_feature_quota(sb));
5735 if (!qf_inums[type])
5738 qf_inode = ext4_iget(sb, qf_inums[type], EXT4_IGET_SPECIAL);
5739 if (IS_ERR(qf_inode)) {
5740 ext4_error(sb, "Bad quota inode # %lu", qf_inums[type]);
5741 return PTR_ERR(qf_inode);
5744 /* Don't account quota for quota files to avoid recursion */
5745 qf_inode->i_flags |= S_NOQUOTA;
5746 lockdep_set_quota_inode(qf_inode, I_DATA_SEM_QUOTA);
5747 err = dquot_enable(qf_inode, type, format_id, flags);
5749 lockdep_set_quota_inode(qf_inode, I_DATA_SEM_NORMAL);
5755 /* Enable usage tracking for all quota types. */
5756 static int ext4_enable_quotas(struct super_block *sb)
5759 unsigned long qf_inums[EXT4_MAXQUOTAS] = {
5760 le32_to_cpu(EXT4_SB(sb)->s_es->s_usr_quota_inum),
5761 le32_to_cpu(EXT4_SB(sb)->s_es->s_grp_quota_inum),
5762 le32_to_cpu(EXT4_SB(sb)->s_es->s_prj_quota_inum)
5764 bool quota_mopt[EXT4_MAXQUOTAS] = {
5765 test_opt(sb, USRQUOTA),
5766 test_opt(sb, GRPQUOTA),
5767 test_opt(sb, PRJQUOTA),
5770 sb_dqopt(sb)->flags |= DQUOT_QUOTA_SYS_FILE | DQUOT_NOLIST_DIRTY;
5771 for (type = 0; type < EXT4_MAXQUOTAS; type++) {
5772 if (qf_inums[type]) {
5773 err = ext4_quota_enable(sb, type, QFMT_VFS_V1,
5774 DQUOT_USAGE_ENABLED |
5775 (quota_mopt[type] ? DQUOT_LIMITS_ENABLED : 0));
5777 for (type--; type >= 0; type--)
5778 dquot_quota_off(sb, type);
5781 "Failed to enable quota tracking "
5782 "(type=%d, err=%d). Please run "
5783 "e2fsck to fix.", type, err);
5791 static int ext4_quota_off(struct super_block *sb, int type)
5793 struct inode *inode = sb_dqopt(sb)->files[type];
5797 /* Force all delayed allocation blocks to be allocated.
5798 * Caller already holds s_umount sem */
5799 if (test_opt(sb, DELALLOC))
5800 sync_filesystem(sb);
5802 if (!inode || !igrab(inode))
5805 err = dquot_quota_off(sb, type);
5806 if (err || ext4_has_feature_quota(sb))
5811 * Update modification times of quota files when userspace can
5812 * start looking at them. If we fail, we return success anyway since
5813 * this is not a hard failure and quotas are already disabled.
5815 handle = ext4_journal_start(inode, EXT4_HT_QUOTA, 1);
5818 EXT4_I(inode)->i_flags &= ~(EXT4_NOATIME_FL | EXT4_IMMUTABLE_FL);
5819 inode_set_flags(inode, 0, S_NOATIME | S_IMMUTABLE);
5820 inode->i_mtime = inode->i_ctime = current_time(inode);
5821 ext4_mark_inode_dirty(handle, inode);
5822 ext4_journal_stop(handle);
5824 inode_unlock(inode);
5826 lockdep_set_quota_inode(inode, I_DATA_SEM_NORMAL);
5830 return dquot_quota_off(sb, type);
5833 /* Read data from quotafile - avoid pagecache and such because we cannot afford
5834 * acquiring the locks... As quota files are never truncated and quota code
5835 * itself serializes the operations (and no one else should touch the files)
5836 * we don't have to be afraid of races */
5837 static ssize_t ext4_quota_read(struct super_block *sb, int type, char *data,
5838 size_t len, loff_t off)
5840 struct inode *inode = sb_dqopt(sb)->files[type];
5841 ext4_lblk_t blk = off >> EXT4_BLOCK_SIZE_BITS(sb);
5842 int offset = off & (sb->s_blocksize - 1);
5845 struct buffer_head *bh;
5846 loff_t i_size = i_size_read(inode);
5850 if (off+len > i_size)
5853 while (toread > 0) {
5854 tocopy = sb->s_blocksize - offset < toread ?
5855 sb->s_blocksize - offset : toread;
5856 bh = ext4_bread(NULL, inode, blk, 0);
5859 if (!bh) /* A hole? */
5860 memset(data, 0, tocopy);
5862 memcpy(data, bh->b_data+offset, tocopy);
5872 /* Write to quotafile (we know the transaction is already started and has
5873 * enough credits) */
5874 static ssize_t ext4_quota_write(struct super_block *sb, int type,
5875 const char *data, size_t len, loff_t off)
5877 struct inode *inode = sb_dqopt(sb)->files[type];
5878 ext4_lblk_t blk = off >> EXT4_BLOCK_SIZE_BITS(sb);
5879 int err, offset = off & (sb->s_blocksize - 1);
5881 struct buffer_head *bh;
5882 handle_t *handle = journal_current_handle();
5885 ext4_msg(sb, KERN_WARNING, "Quota write (off=%llu, len=%llu)"
5886 " cancelled because transaction is not started",
5887 (unsigned long long)off, (unsigned long long)len);
5891 * Since we account only one data block in transaction credits,
5892 * then it is impossible to cross a block boundary.
5894 if (sb->s_blocksize - offset < len) {
5895 ext4_msg(sb, KERN_WARNING, "Quota write (off=%llu, len=%llu)"
5896 " cancelled because not block aligned",
5897 (unsigned long long)off, (unsigned long long)len);
5902 bh = ext4_bread(handle, inode, blk,
5903 EXT4_GET_BLOCKS_CREATE |
5904 EXT4_GET_BLOCKS_METADATA_NOFAIL);
5905 } while (IS_ERR(bh) && (PTR_ERR(bh) == -ENOSPC) &&
5906 ext4_should_retry_alloc(inode->i_sb, &retries));
5911 BUFFER_TRACE(bh, "get write access");
5912 err = ext4_journal_get_write_access(handle, bh);
5918 memcpy(bh->b_data+offset, data, len);
5919 flush_dcache_page(bh->b_page);
5921 err = ext4_handle_dirty_metadata(handle, NULL, bh);
5924 if (inode->i_size < off + len) {
5925 i_size_write(inode, off + len);
5926 EXT4_I(inode)->i_disksize = inode->i_size;
5927 ext4_mark_inode_dirty(handle, inode);
5932 static int ext4_get_next_id(struct super_block *sb, struct kqid *qid)
5934 const struct quota_format_ops *ops;
5936 if (!sb_has_quota_loaded(sb, qid->type))
5938 ops = sb_dqopt(sb)->ops[qid->type];
5939 if (!ops || !ops->get_next_id)
5941 return dquot_get_next_id(sb, qid);
5945 static struct dentry *ext4_mount(struct file_system_type *fs_type, int flags,
5946 const char *dev_name, void *data)
5948 return mount_bdev(fs_type, flags, dev_name, data, ext4_fill_super);
5951 #if !defined(CONFIG_EXT2_FS) && !defined(CONFIG_EXT2_FS_MODULE) && defined(CONFIG_EXT4_USE_FOR_EXT2)
5952 static inline void register_as_ext2(void)
5954 int err = register_filesystem(&ext2_fs_type);
5957 "EXT4-fs: Unable to register as ext2 (%d)\n", err);
5960 static inline void unregister_as_ext2(void)
5962 unregister_filesystem(&ext2_fs_type);
5965 static inline int ext2_feature_set_ok(struct super_block *sb)
5967 if (ext4_has_unknown_ext2_incompat_features(sb))
5971 if (ext4_has_unknown_ext2_ro_compat_features(sb))
5976 static inline void register_as_ext2(void) { }
5977 static inline void unregister_as_ext2(void) { }
5978 static inline int ext2_feature_set_ok(struct super_block *sb) { return 0; }
5981 static inline void register_as_ext3(void)
5983 int err = register_filesystem(&ext3_fs_type);
5986 "EXT4-fs: Unable to register as ext3 (%d)\n", err);
5989 static inline void unregister_as_ext3(void)
5991 unregister_filesystem(&ext3_fs_type);
5994 static inline int ext3_feature_set_ok(struct super_block *sb)
5996 if (ext4_has_unknown_ext3_incompat_features(sb))
5998 if (!ext4_has_feature_journal(sb))
6002 if (ext4_has_unknown_ext3_ro_compat_features(sb))
6007 static struct file_system_type ext4_fs_type = {
6008 .owner = THIS_MODULE,
6010 .mount = ext4_mount,
6011 .kill_sb = kill_block_super,
6012 .fs_flags = FS_REQUIRES_DEV,
6014 MODULE_ALIAS_FS("ext4");
6016 /* Shared across all ext4 file systems */
6017 wait_queue_head_t ext4__ioend_wq[EXT4_WQ_HASH_SZ];
6019 static int __init ext4_init_fs(void)
6023 ratelimit_state_init(&ext4_mount_msg_ratelimit, 30 * HZ, 64);
6024 ext4_li_info = NULL;
6025 mutex_init(&ext4_li_mtx);
6027 /* Build-time check for flags consistency */
6028 ext4_check_flag_values();
6030 for (i = 0; i < EXT4_WQ_HASH_SZ; i++)
6031 init_waitqueue_head(&ext4__ioend_wq[i]);
6033 err = ext4_init_es();
6037 err = ext4_init_pageio();
6041 err = ext4_init_system_zone();
6045 err = ext4_init_sysfs();
6049 err = ext4_init_mballoc();
6052 err = init_inodecache();
6057 err = register_filesystem(&ext4_fs_type);
6063 unregister_as_ext2();
6064 unregister_as_ext3();
6065 destroy_inodecache();
6067 ext4_exit_mballoc();
6071 ext4_exit_system_zone();
6080 static void __exit ext4_exit_fs(void)
6082 ext4_destroy_lazyinit_thread();
6083 unregister_as_ext2();
6084 unregister_as_ext3();
6085 unregister_filesystem(&ext4_fs_type);
6086 destroy_inodecache();
6087 ext4_exit_mballoc();
6089 ext4_exit_system_zone();
6094 MODULE_AUTHOR("Remy Card, Stephen Tweedie, Andrew Morton, Andreas Dilger, Theodore Ts'o and others");
6095 MODULE_DESCRIPTION("Fourth Extended Filesystem");
6096 MODULE_LICENSE("GPL");
6097 MODULE_SOFTDEP("pre: crc32c");
6098 module_init(ext4_init_fs)
6099 module_exit(ext4_exit_fs)