1 // SPDX-License-Identifier: GPL-2.0
3 * linux/fs/ext4/super.c
5 * Copyright (C) 1992, 1993, 1994, 1995
6 * Remy Card (card@masi.ibp.fr)
7 * Laboratoire MASI - Institut Blaise Pascal
8 * Universite Pierre et Marie Curie (Paris VI)
12 * linux/fs/minix/inode.c
14 * Copyright (C) 1991, 1992 Linus Torvalds
16 * Big-endian to little-endian byte-swapping/bitmaps by
17 * David S. Miller (davem@caip.rutgers.edu), 1995
20 #include <linux/module.h>
21 #include <linux/string.h>
23 #include <linux/time.h>
24 #include <linux/vmalloc.h>
25 #include <linux/slab.h>
26 #include <linux/init.h>
27 #include <linux/blkdev.h>
28 #include <linux/backing-dev.h>
29 #include <linux/parser.h>
30 #include <linux/buffer_head.h>
31 #include <linux/exportfs.h>
32 #include <linux/vfs.h>
33 #include <linux/random.h>
34 #include <linux/mount.h>
35 #include <linux/namei.h>
36 #include <linux/quotaops.h>
37 #include <linux/seq_file.h>
38 #include <linux/ctype.h>
39 #include <linux/log2.h>
40 #include <linux/crc16.h>
41 #include <linux/dax.h>
42 #include <linux/cleancache.h>
43 #include <linux/uaccess.h>
44 #include <linux/iversion.h>
46 #include <linux/kthread.h>
47 #include <linux/freezer.h>
50 #include "ext4_extents.h" /* Needed for trace points definition */
51 #include "ext4_jbd2.h"
57 #define CREATE_TRACE_POINTS
58 #include <trace/events/ext4.h>
60 static struct ext4_lazy_init *ext4_li_info;
61 static struct mutex ext4_li_mtx;
62 static struct ratelimit_state ext4_mount_msg_ratelimit;
64 static int ext4_load_journal(struct super_block *, struct ext4_super_block *,
65 unsigned long journal_devnum);
66 static int ext4_show_options(struct seq_file *seq, struct dentry *root);
67 static int ext4_commit_super(struct super_block *sb, int sync);
68 static int ext4_mark_recovery_complete(struct super_block *sb,
69 struct ext4_super_block *es);
70 static int ext4_clear_journal_err(struct super_block *sb,
71 struct ext4_super_block *es);
72 static int ext4_sync_fs(struct super_block *sb, int wait);
73 static int ext4_remount(struct super_block *sb, int *flags, char *data);
74 static int ext4_statfs(struct dentry *dentry, struct kstatfs *buf);
75 static int ext4_unfreeze(struct super_block *sb);
76 static int ext4_freeze(struct super_block *sb);
77 static struct dentry *ext4_mount(struct file_system_type *fs_type, int flags,
78 const char *dev_name, void *data);
79 static inline int ext2_feature_set_ok(struct super_block *sb);
80 static inline int ext3_feature_set_ok(struct super_block *sb);
81 static int ext4_feature_set_ok(struct super_block *sb, int readonly);
82 static void ext4_destroy_lazyinit_thread(void);
83 static void ext4_unregister_li_request(struct super_block *sb);
84 static void ext4_clear_request_list(void);
85 static struct inode *ext4_get_journal_inode(struct super_block *sb,
86 unsigned int journal_inum);
91 * Note the difference between i_mmap_sem (EXT4_I(inode)->i_mmap_sem) and
92 * i_mmap_rwsem (inode->i_mmap_rwsem)!
95 * mmap_sem -> sb_start_pagefault -> i_mmap_sem (r) -> transaction start ->
96 * page lock -> i_data_sem (rw)
98 * buffered write path:
99 * sb_start_write -> i_mutex -> mmap_sem
100 * sb_start_write -> i_mutex -> transaction start -> page lock ->
104 * sb_start_write -> i_mutex -> i_mmap_sem (w) -> i_mmap_rwsem (w) -> page lock
105 * sb_start_write -> i_mutex -> i_mmap_sem (w) -> transaction start ->
109 * sb_start_write -> i_mutex -> mmap_sem
110 * sb_start_write -> i_mutex -> 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);
338 static void __ext4_update_tstamp(__le32 *lo, __u8 *hi)
340 time64_t now = ktime_get_real_seconds();
342 now = clamp_val(now, 0, (1ull << 40) - 1);
344 *lo = cpu_to_le32(lower_32_bits(now));
345 *hi = upper_32_bits(now);
348 static time64_t __ext4_get_tstamp(__le32 *lo, __u8 *hi)
350 return ((time64_t)(*hi) << 32) + le32_to_cpu(*lo);
352 #define ext4_update_tstamp(es, tstamp) \
353 __ext4_update_tstamp(&(es)->tstamp, &(es)->tstamp ## _hi)
354 #define ext4_get_tstamp(es, tstamp) \
355 __ext4_get_tstamp(&(es)->tstamp, &(es)->tstamp ## _hi)
357 static void __save_error_info(struct super_block *sb, const char *func,
360 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
362 EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS;
363 if (bdev_read_only(sb->s_bdev))
365 es->s_state |= cpu_to_le16(EXT4_ERROR_FS);
366 ext4_update_tstamp(es, s_last_error_time);
367 strncpy(es->s_last_error_func, func, sizeof(es->s_last_error_func));
368 es->s_last_error_line = cpu_to_le32(line);
369 if (!es->s_first_error_time) {
370 es->s_first_error_time = es->s_last_error_time;
371 es->s_first_error_time_hi = es->s_last_error_time_hi;
372 strncpy(es->s_first_error_func, func,
373 sizeof(es->s_first_error_func));
374 es->s_first_error_line = cpu_to_le32(line);
375 es->s_first_error_ino = es->s_last_error_ino;
376 es->s_first_error_block = es->s_last_error_block;
379 * Start the daily error reporting function if it hasn't been
382 if (!es->s_error_count)
383 mod_timer(&EXT4_SB(sb)->s_err_report, jiffies + 24*60*60*HZ);
384 le32_add_cpu(&es->s_error_count, 1);
387 static void save_error_info(struct super_block *sb, const char *func,
390 __save_error_info(sb, func, line);
391 if (!bdev_read_only(sb->s_bdev))
392 ext4_commit_super(sb, 1);
396 * The del_gendisk() function uninitializes the disk-specific data
397 * structures, including the bdi structure, without telling anyone
398 * else. Once this happens, any attempt to call mark_buffer_dirty()
399 * (for example, by ext4_commit_super), will cause a kernel OOPS.
400 * This is a kludge to prevent these oops until we can put in a proper
401 * hook in del_gendisk() to inform the VFS and file system layers.
403 static int block_device_ejected(struct super_block *sb)
405 struct inode *bd_inode = sb->s_bdev->bd_inode;
406 struct backing_dev_info *bdi = inode_to_bdi(bd_inode);
408 return bdi->dev == NULL;
411 static void ext4_journal_commit_callback(journal_t *journal, transaction_t *txn)
413 struct super_block *sb = journal->j_private;
414 struct ext4_sb_info *sbi = EXT4_SB(sb);
415 int error = is_journal_aborted(journal);
416 struct ext4_journal_cb_entry *jce;
418 BUG_ON(txn->t_state == T_FINISHED);
420 ext4_process_freed_data(sb, txn->t_tid);
422 spin_lock(&sbi->s_md_lock);
423 while (!list_empty(&txn->t_private_list)) {
424 jce = list_entry(txn->t_private_list.next,
425 struct ext4_journal_cb_entry, jce_list);
426 list_del_init(&jce->jce_list);
427 spin_unlock(&sbi->s_md_lock);
428 jce->jce_func(sb, jce, error);
429 spin_lock(&sbi->s_md_lock);
431 spin_unlock(&sbi->s_md_lock);
434 static bool system_going_down(void)
436 return system_state == SYSTEM_HALT || system_state == SYSTEM_POWER_OFF
437 || system_state == SYSTEM_RESTART;
440 /* Deal with the reporting of failure conditions on a filesystem such as
441 * inconsistencies detected or read IO failures.
443 * On ext2, we can store the error state of the filesystem in the
444 * superblock. That is not possible on ext4, because we may have other
445 * write ordering constraints on the superblock which prevent us from
446 * writing it out straight away; and given that the journal is about to
447 * be aborted, we can't rely on the current, or future, transactions to
448 * write out the superblock safely.
450 * We'll just use the jbd2_journal_abort() error code to record an error in
451 * the journal instead. On recovery, the journal will complain about
452 * that error until we've noted it down and cleared it.
455 static void ext4_handle_error(struct super_block *sb)
457 journal_t *journal = EXT4_SB(sb)->s_journal;
459 if (test_opt(sb, WARN_ON_ERROR))
462 if (sb_rdonly(sb) || test_opt(sb, ERRORS_CONT))
465 EXT4_SB(sb)->s_mount_flags |= EXT4_MF_FS_ABORTED;
467 jbd2_journal_abort(journal, -EIO);
469 * We force ERRORS_RO behavior when system is rebooting. Otherwise we
470 * could panic during 'reboot -f' as the underlying device got already
473 if (test_opt(sb, ERRORS_RO) || system_going_down()) {
474 ext4_msg(sb, KERN_CRIT, "Remounting filesystem read-only");
476 * Make sure updated value of ->s_mount_flags will be visible
477 * before ->s_flags update
480 sb->s_flags |= SB_RDONLY;
481 } else if (test_opt(sb, ERRORS_PANIC)) {
482 if (EXT4_SB(sb)->s_journal &&
483 !(EXT4_SB(sb)->s_journal->j_flags & JBD2_REC_ERR))
485 panic("EXT4-fs (device %s): panic forced after error\n",
490 #define ext4_error_ratelimit(sb) \
491 ___ratelimit(&(EXT4_SB(sb)->s_err_ratelimit_state), \
494 void __ext4_error(struct super_block *sb, const char *function,
495 unsigned int line, const char *fmt, ...)
497 struct va_format vaf;
500 if (unlikely(ext4_forced_shutdown(EXT4_SB(sb))))
503 trace_ext4_error(sb, function, line);
504 if (ext4_error_ratelimit(sb)) {
509 "EXT4-fs error (device %s): %s:%d: comm %s: %pV\n",
510 sb->s_id, function, line, current->comm, &vaf);
513 save_error_info(sb, function, line);
514 ext4_handle_error(sb);
517 void __ext4_error_inode(struct inode *inode, const char *function,
518 unsigned int line, ext4_fsblk_t block,
519 const char *fmt, ...)
522 struct va_format vaf;
523 struct ext4_super_block *es = EXT4_SB(inode->i_sb)->s_es;
525 if (unlikely(ext4_forced_shutdown(EXT4_SB(inode->i_sb))))
528 trace_ext4_error(inode->i_sb, function, line);
529 es->s_last_error_ino = cpu_to_le32(inode->i_ino);
530 es->s_last_error_block = cpu_to_le64(block);
531 if (ext4_error_ratelimit(inode->i_sb)) {
536 printk(KERN_CRIT "EXT4-fs error (device %s): %s:%d: "
537 "inode #%lu: block %llu: comm %s: %pV\n",
538 inode->i_sb->s_id, function, line, inode->i_ino,
539 block, current->comm, &vaf);
541 printk(KERN_CRIT "EXT4-fs error (device %s): %s:%d: "
542 "inode #%lu: comm %s: %pV\n",
543 inode->i_sb->s_id, function, line, inode->i_ino,
544 current->comm, &vaf);
547 save_error_info(inode->i_sb, function, line);
548 ext4_handle_error(inode->i_sb);
551 void __ext4_error_file(struct file *file, const char *function,
552 unsigned int line, ext4_fsblk_t block,
553 const char *fmt, ...)
556 struct va_format vaf;
557 struct ext4_super_block *es;
558 struct inode *inode = file_inode(file);
559 char pathname[80], *path;
561 if (unlikely(ext4_forced_shutdown(EXT4_SB(inode->i_sb))))
564 trace_ext4_error(inode->i_sb, function, line);
565 es = EXT4_SB(inode->i_sb)->s_es;
566 es->s_last_error_ino = cpu_to_le32(inode->i_ino);
567 if (ext4_error_ratelimit(inode->i_sb)) {
568 path = file_path(file, pathname, sizeof(pathname));
576 "EXT4-fs error (device %s): %s:%d: inode #%lu: "
577 "block %llu: comm %s: path %s: %pV\n",
578 inode->i_sb->s_id, function, line, inode->i_ino,
579 block, current->comm, path, &vaf);
582 "EXT4-fs error (device %s): %s:%d: inode #%lu: "
583 "comm %s: path %s: %pV\n",
584 inode->i_sb->s_id, function, line, inode->i_ino,
585 current->comm, path, &vaf);
588 save_error_info(inode->i_sb, function, line);
589 ext4_handle_error(inode->i_sb);
592 const char *ext4_decode_error(struct super_block *sb, int errno,
599 errstr = "Corrupt filesystem";
602 errstr = "Filesystem failed CRC";
605 errstr = "IO failure";
608 errstr = "Out of memory";
611 if (!sb || (EXT4_SB(sb)->s_journal &&
612 EXT4_SB(sb)->s_journal->j_flags & JBD2_ABORT))
613 errstr = "Journal has aborted";
615 errstr = "Readonly filesystem";
618 /* If the caller passed in an extra buffer for unknown
619 * errors, textualise them now. Else we just return
622 /* Check for truncated error codes... */
623 if (snprintf(nbuf, 16, "error %d", -errno) >= 0)
632 /* __ext4_std_error decodes expected errors from journaling functions
633 * automatically and invokes the appropriate error response. */
635 void __ext4_std_error(struct super_block *sb, const char *function,
636 unsigned int line, int errno)
641 if (unlikely(ext4_forced_shutdown(EXT4_SB(sb))))
644 /* Special case: if the error is EROFS, and we're not already
645 * inside a transaction, then there's really no point in logging
647 if (errno == -EROFS && journal_current_handle() == NULL && sb_rdonly(sb))
650 if (ext4_error_ratelimit(sb)) {
651 errstr = ext4_decode_error(sb, errno, nbuf);
652 printk(KERN_CRIT "EXT4-fs error (device %s) in %s:%d: %s\n",
653 sb->s_id, function, line, errstr);
656 save_error_info(sb, function, line);
657 ext4_handle_error(sb);
661 * ext4_abort is a much stronger failure handler than ext4_error. The
662 * abort function may be used to deal with unrecoverable failures such
663 * as journal IO errors or ENOMEM at a critical moment in log management.
665 * We unconditionally force the filesystem into an ABORT|READONLY state,
666 * unless the error response on the fs has been set to panic in which
667 * case we take the easy way out and panic immediately.
670 void __ext4_abort(struct super_block *sb, const char *function,
671 unsigned int line, const char *fmt, ...)
673 struct va_format vaf;
676 if (unlikely(ext4_forced_shutdown(EXT4_SB(sb))))
679 save_error_info(sb, function, line);
683 printk(KERN_CRIT "EXT4-fs error (device %s): %s:%d: %pV\n",
684 sb->s_id, function, line, &vaf);
687 if (sb_rdonly(sb) == 0) {
688 ext4_msg(sb, KERN_CRIT, "Remounting filesystem read-only");
689 EXT4_SB(sb)->s_mount_flags |= EXT4_MF_FS_ABORTED;
691 * Make sure updated value of ->s_mount_flags will be visible
692 * before ->s_flags update
695 sb->s_flags |= SB_RDONLY;
696 if (EXT4_SB(sb)->s_journal)
697 jbd2_journal_abort(EXT4_SB(sb)->s_journal, -EIO);
698 save_error_info(sb, function, line);
700 if (test_opt(sb, ERRORS_PANIC) && !system_going_down()) {
701 if (EXT4_SB(sb)->s_journal &&
702 !(EXT4_SB(sb)->s_journal->j_flags & JBD2_REC_ERR))
704 panic("EXT4-fs panic from previous error\n");
708 void __ext4_msg(struct super_block *sb,
709 const char *prefix, const char *fmt, ...)
711 struct va_format vaf;
714 if (!___ratelimit(&(EXT4_SB(sb)->s_msg_ratelimit_state), "EXT4-fs"))
720 printk("%sEXT4-fs (%s): %pV\n", prefix, sb->s_id, &vaf);
724 #define ext4_warning_ratelimit(sb) \
725 ___ratelimit(&(EXT4_SB(sb)->s_warning_ratelimit_state), \
728 void __ext4_warning(struct super_block *sb, const char *function,
729 unsigned int line, const char *fmt, ...)
731 struct va_format vaf;
734 if (!ext4_warning_ratelimit(sb))
740 printk(KERN_WARNING "EXT4-fs warning (device %s): %s:%d: %pV\n",
741 sb->s_id, function, line, &vaf);
745 void __ext4_warning_inode(const struct inode *inode, const char *function,
746 unsigned int line, const char *fmt, ...)
748 struct va_format vaf;
751 if (!ext4_warning_ratelimit(inode->i_sb))
757 printk(KERN_WARNING "EXT4-fs warning (device %s): %s:%d: "
758 "inode #%lu: comm %s: %pV\n", inode->i_sb->s_id,
759 function, line, inode->i_ino, current->comm, &vaf);
763 void __ext4_grp_locked_error(const char *function, unsigned int line,
764 struct super_block *sb, ext4_group_t grp,
765 unsigned long ino, ext4_fsblk_t block,
766 const char *fmt, ...)
770 struct va_format vaf;
772 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
774 if (unlikely(ext4_forced_shutdown(EXT4_SB(sb))))
777 trace_ext4_error(sb, function, line);
778 es->s_last_error_ino = cpu_to_le32(ino);
779 es->s_last_error_block = cpu_to_le64(block);
780 __save_error_info(sb, function, line);
782 if (ext4_error_ratelimit(sb)) {
786 printk(KERN_CRIT "EXT4-fs error (device %s): %s:%d: group %u, ",
787 sb->s_id, function, line, grp);
789 printk(KERN_CONT "inode %lu: ", ino);
791 printk(KERN_CONT "block %llu:",
792 (unsigned long long) block);
793 printk(KERN_CONT "%pV\n", &vaf);
797 if (test_opt(sb, WARN_ON_ERROR))
800 if (test_opt(sb, ERRORS_CONT)) {
801 ext4_commit_super(sb, 0);
805 ext4_unlock_group(sb, grp);
806 ext4_commit_super(sb, 1);
807 ext4_handle_error(sb);
809 * We only get here in the ERRORS_RO case; relocking the group
810 * may be dangerous, but nothing bad will happen since the
811 * filesystem will have already been marked read/only and the
812 * journal has been aborted. We return 1 as a hint to callers
813 * who might what to use the return value from
814 * ext4_grp_locked_error() to distinguish between the
815 * ERRORS_CONT and ERRORS_RO case, and perhaps return more
816 * aggressively from the ext4 function in question, with a
817 * more appropriate error code.
819 ext4_lock_group(sb, grp);
823 void ext4_mark_group_bitmap_corrupted(struct super_block *sb,
827 struct ext4_sb_info *sbi = EXT4_SB(sb);
828 struct ext4_group_info *grp = ext4_get_group_info(sb, group);
829 struct ext4_group_desc *gdp = ext4_get_group_desc(sb, group, NULL);
832 if (flags & EXT4_GROUP_INFO_BBITMAP_CORRUPT) {
833 ret = ext4_test_and_set_bit(EXT4_GROUP_INFO_BBITMAP_CORRUPT_BIT,
836 percpu_counter_sub(&sbi->s_freeclusters_counter,
840 if (flags & EXT4_GROUP_INFO_IBITMAP_CORRUPT) {
841 ret = ext4_test_and_set_bit(EXT4_GROUP_INFO_IBITMAP_CORRUPT_BIT,
846 count = ext4_free_inodes_count(sb, gdp);
847 percpu_counter_sub(&sbi->s_freeinodes_counter,
853 void ext4_update_dynamic_rev(struct super_block *sb)
855 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
857 if (le32_to_cpu(es->s_rev_level) > EXT4_GOOD_OLD_REV)
861 "updating to rev %d because of new feature flag, "
862 "running e2fsck is recommended",
865 es->s_first_ino = cpu_to_le32(EXT4_GOOD_OLD_FIRST_INO);
866 es->s_inode_size = cpu_to_le16(EXT4_GOOD_OLD_INODE_SIZE);
867 es->s_rev_level = cpu_to_le32(EXT4_DYNAMIC_REV);
868 /* leave es->s_feature_*compat flags alone */
869 /* es->s_uuid will be set by e2fsck if empty */
872 * The rest of the superblock fields should be zero, and if not it
873 * means they are likely already in use, so leave them alone. We
874 * can leave it up to e2fsck to clean up any inconsistencies there.
879 * Open the external journal device
881 static struct block_device *ext4_blkdev_get(dev_t dev, struct super_block *sb)
883 struct block_device *bdev;
884 char b[BDEVNAME_SIZE];
886 bdev = blkdev_get_by_dev(dev, FMODE_READ|FMODE_WRITE|FMODE_EXCL, sb);
892 ext4_msg(sb, KERN_ERR, "failed to open journal device %s: %ld",
893 __bdevname(dev, b), PTR_ERR(bdev));
898 * Release the journal device
900 static void ext4_blkdev_put(struct block_device *bdev)
902 blkdev_put(bdev, FMODE_READ|FMODE_WRITE|FMODE_EXCL);
905 static void ext4_blkdev_remove(struct ext4_sb_info *sbi)
907 struct block_device *bdev;
908 bdev = sbi->journal_bdev;
910 ext4_blkdev_put(bdev);
911 sbi->journal_bdev = NULL;
915 static inline struct inode *orphan_list_entry(struct list_head *l)
917 return &list_entry(l, struct ext4_inode_info, i_orphan)->vfs_inode;
920 static void dump_orphan_list(struct super_block *sb, struct ext4_sb_info *sbi)
924 ext4_msg(sb, KERN_ERR, "sb orphan head is %d",
925 le32_to_cpu(sbi->s_es->s_last_orphan));
927 printk(KERN_ERR "sb_info orphan list:\n");
928 list_for_each(l, &sbi->s_orphan) {
929 struct inode *inode = orphan_list_entry(l);
931 "inode %s:%lu at %p: mode %o, nlink %d, next %d\n",
932 inode->i_sb->s_id, inode->i_ino, inode,
933 inode->i_mode, inode->i_nlink,
939 static int ext4_quota_off(struct super_block *sb, int type);
941 static inline void ext4_quota_off_umount(struct super_block *sb)
945 /* Use our quota_off function to clear inode flags etc. */
946 for (type = 0; type < EXT4_MAXQUOTAS; type++)
947 ext4_quota_off(sb, type);
951 * This is a helper function which is used in the mount/remount
952 * codepaths (which holds s_umount) to fetch the quota file name.
954 static inline char *get_qf_name(struct super_block *sb,
955 struct ext4_sb_info *sbi,
958 return rcu_dereference_protected(sbi->s_qf_names[type],
959 lockdep_is_held(&sb->s_umount));
962 static inline void ext4_quota_off_umount(struct super_block *sb)
967 static void ext4_put_super(struct super_block *sb)
969 struct ext4_sb_info *sbi = EXT4_SB(sb);
970 struct ext4_super_block *es = sbi->s_es;
971 struct buffer_head **group_desc;
972 struct flex_groups **flex_groups;
976 ext4_unregister_li_request(sb);
977 ext4_quota_off_umount(sb);
979 destroy_workqueue(sbi->rsv_conversion_wq);
981 if (sbi->s_journal) {
982 aborted = is_journal_aborted(sbi->s_journal);
983 err = jbd2_journal_destroy(sbi->s_journal);
984 sbi->s_journal = NULL;
985 if ((err < 0) && !aborted)
986 ext4_abort(sb, "Couldn't clean up the journal");
989 ext4_unregister_sysfs(sb);
990 ext4_es_unregister_shrinker(sbi);
991 del_timer_sync(&sbi->s_err_report);
992 ext4_release_system_zone(sb);
994 ext4_ext_release(sb);
996 if (!sb_rdonly(sb) && !aborted) {
997 ext4_clear_feature_journal_needs_recovery(sb);
998 es->s_state = cpu_to_le16(sbi->s_mount_state);
1001 ext4_commit_super(sb, 1);
1004 group_desc = rcu_dereference(sbi->s_group_desc);
1005 for (i = 0; i < sbi->s_gdb_count; i++)
1006 brelse(group_desc[i]);
1008 flex_groups = rcu_dereference(sbi->s_flex_groups);
1010 for (i = 0; i < sbi->s_flex_groups_allocated; i++)
1011 kvfree(flex_groups[i]);
1012 kvfree(flex_groups);
1015 percpu_counter_destroy(&sbi->s_freeclusters_counter);
1016 percpu_counter_destroy(&sbi->s_freeinodes_counter);
1017 percpu_counter_destroy(&sbi->s_dirs_counter);
1018 percpu_counter_destroy(&sbi->s_dirtyclusters_counter);
1019 percpu_free_rwsem(&sbi->s_writepages_rwsem);
1021 for (i = 0; i < EXT4_MAXQUOTAS; i++)
1022 kfree(get_qf_name(sb, sbi, i));
1025 /* Debugging code just in case the in-memory inode orphan list
1026 * isn't empty. The on-disk one can be non-empty if we've
1027 * detected an error and taken the fs readonly, but the
1028 * in-memory list had better be clean by this point. */
1029 if (!list_empty(&sbi->s_orphan))
1030 dump_orphan_list(sb, sbi);
1031 J_ASSERT(list_empty(&sbi->s_orphan));
1033 sync_blockdev(sb->s_bdev);
1034 invalidate_bdev(sb->s_bdev);
1035 if (sbi->journal_bdev && sbi->journal_bdev != sb->s_bdev) {
1037 * Invalidate the journal device's buffers. We don't want them
1038 * floating about in memory - the physical journal device may
1039 * hotswapped, and it breaks the `ro-after' testing code.
1041 sync_blockdev(sbi->journal_bdev);
1042 invalidate_bdev(sbi->journal_bdev);
1043 ext4_blkdev_remove(sbi);
1045 if (sbi->s_ea_inode_cache) {
1046 ext4_xattr_destroy_cache(sbi->s_ea_inode_cache);
1047 sbi->s_ea_inode_cache = NULL;
1049 if (sbi->s_ea_block_cache) {
1050 ext4_xattr_destroy_cache(sbi->s_ea_block_cache);
1051 sbi->s_ea_block_cache = NULL;
1054 kthread_stop(sbi->s_mmp_tsk);
1056 sb->s_fs_info = NULL;
1058 * Now that we are completely done shutting down the
1059 * superblock, we need to actually destroy the kobject.
1061 kobject_put(&sbi->s_kobj);
1062 wait_for_completion(&sbi->s_kobj_unregister);
1063 if (sbi->s_chksum_driver)
1064 crypto_free_shash(sbi->s_chksum_driver);
1065 kfree(sbi->s_blockgroup_lock);
1066 fs_put_dax(sbi->s_daxdev);
1070 static struct kmem_cache *ext4_inode_cachep;
1073 * Called inside transaction, so use GFP_NOFS
1075 static struct inode *ext4_alloc_inode(struct super_block *sb)
1077 struct ext4_inode_info *ei;
1079 ei = kmem_cache_alloc(ext4_inode_cachep, GFP_NOFS);
1083 inode_set_iversion(&ei->vfs_inode, 1);
1084 spin_lock_init(&ei->i_raw_lock);
1085 INIT_LIST_HEAD(&ei->i_prealloc_list);
1086 spin_lock_init(&ei->i_prealloc_lock);
1087 ext4_es_init_tree(&ei->i_es_tree);
1088 rwlock_init(&ei->i_es_lock);
1089 INIT_LIST_HEAD(&ei->i_es_list);
1090 ei->i_es_all_nr = 0;
1091 ei->i_es_shk_nr = 0;
1092 ei->i_es_shrink_lblk = 0;
1093 ei->i_reserved_data_blocks = 0;
1094 ei->i_da_metadata_calc_len = 0;
1095 ei->i_da_metadata_calc_last_lblock = 0;
1096 spin_lock_init(&(ei->i_block_reservation_lock));
1098 ei->i_reserved_quota = 0;
1099 memset(&ei->i_dquot, 0, sizeof(ei->i_dquot));
1102 INIT_LIST_HEAD(&ei->i_rsv_conversion_list);
1103 spin_lock_init(&ei->i_completed_io_lock);
1105 ei->i_datasync_tid = 0;
1106 atomic_set(&ei->i_unwritten, 0);
1107 INIT_WORK(&ei->i_rsv_conversion_work, ext4_end_io_rsv_work);
1108 return &ei->vfs_inode;
1111 static int ext4_drop_inode(struct inode *inode)
1113 int drop = generic_drop_inode(inode);
1115 trace_ext4_drop_inode(inode, drop);
1119 static void ext4_i_callback(struct rcu_head *head)
1121 struct inode *inode = container_of(head, struct inode, i_rcu);
1122 kmem_cache_free(ext4_inode_cachep, EXT4_I(inode));
1125 static void ext4_destroy_inode(struct inode *inode)
1127 if (!list_empty(&(EXT4_I(inode)->i_orphan))) {
1128 ext4_msg(inode->i_sb, KERN_ERR,
1129 "Inode %lu (%p): orphan list check failed!",
1130 inode->i_ino, EXT4_I(inode));
1131 print_hex_dump(KERN_INFO, "", DUMP_PREFIX_ADDRESS, 16, 4,
1132 EXT4_I(inode), sizeof(struct ext4_inode_info),
1136 call_rcu(&inode->i_rcu, ext4_i_callback);
1139 static void init_once(void *foo)
1141 struct ext4_inode_info *ei = (struct ext4_inode_info *) foo;
1143 INIT_LIST_HEAD(&ei->i_orphan);
1144 init_rwsem(&ei->xattr_sem);
1145 init_rwsem(&ei->i_data_sem);
1146 init_rwsem(&ei->i_mmap_sem);
1147 inode_init_once(&ei->vfs_inode);
1150 static int __init init_inodecache(void)
1152 ext4_inode_cachep = kmem_cache_create_usercopy("ext4_inode_cache",
1153 sizeof(struct ext4_inode_info), 0,
1154 (SLAB_RECLAIM_ACCOUNT|SLAB_MEM_SPREAD|
1156 offsetof(struct ext4_inode_info, i_data),
1157 sizeof_field(struct ext4_inode_info, i_data),
1159 if (ext4_inode_cachep == NULL)
1164 static void destroy_inodecache(void)
1167 * Make sure all delayed rcu free inodes are flushed before we
1171 kmem_cache_destroy(ext4_inode_cachep);
1174 void ext4_clear_inode(struct inode *inode)
1176 invalidate_inode_buffers(inode);
1179 ext4_discard_preallocations(inode);
1180 ext4_es_remove_extent(inode, 0, EXT_MAX_BLOCKS);
1181 if (EXT4_I(inode)->jinode) {
1182 jbd2_journal_release_jbd_inode(EXT4_JOURNAL(inode),
1183 EXT4_I(inode)->jinode);
1184 jbd2_free_inode(EXT4_I(inode)->jinode);
1185 EXT4_I(inode)->jinode = NULL;
1187 fscrypt_put_encryption_info(inode);
1190 static struct inode *ext4_nfs_get_inode(struct super_block *sb,
1191 u64 ino, u32 generation)
1193 struct inode *inode;
1196 * Currently we don't know the generation for parent directory, so
1197 * a generation of 0 means "accept any"
1199 inode = ext4_iget(sb, ino, EXT4_IGET_HANDLE);
1201 return ERR_CAST(inode);
1202 if (generation && inode->i_generation != generation) {
1204 return ERR_PTR(-ESTALE);
1210 static struct dentry *ext4_fh_to_dentry(struct super_block *sb, struct fid *fid,
1211 int fh_len, int fh_type)
1213 return generic_fh_to_dentry(sb, fid, fh_len, fh_type,
1214 ext4_nfs_get_inode);
1217 static struct dentry *ext4_fh_to_parent(struct super_block *sb, struct fid *fid,
1218 int fh_len, int fh_type)
1220 return generic_fh_to_parent(sb, fid, fh_len, fh_type,
1221 ext4_nfs_get_inode);
1224 static int ext4_nfs_commit_metadata(struct inode *inode)
1226 struct writeback_control wbc = {
1227 .sync_mode = WB_SYNC_ALL
1230 trace_ext4_nfs_commit_metadata(inode);
1231 return ext4_write_inode(inode, &wbc);
1235 * Try to release metadata pages (indirect blocks, directories) which are
1236 * mapped via the block device. Since these pages could have journal heads
1237 * which would prevent try_to_free_buffers() from freeing them, we must use
1238 * jbd2 layer's try_to_free_buffers() function to release them.
1240 static int bdev_try_to_free_page(struct super_block *sb, struct page *page,
1243 journal_t *journal = EXT4_SB(sb)->s_journal;
1245 WARN_ON(PageChecked(page));
1246 if (!page_has_buffers(page))
1249 return jbd2_journal_try_to_free_buffers(journal, page,
1250 wait & ~__GFP_DIRECT_RECLAIM);
1251 return try_to_free_buffers(page);
1254 #ifdef CONFIG_EXT4_FS_ENCRYPTION
1255 static int ext4_get_context(struct inode *inode, void *ctx, size_t len)
1257 return ext4_xattr_get(inode, EXT4_XATTR_INDEX_ENCRYPTION,
1258 EXT4_XATTR_NAME_ENCRYPTION_CONTEXT, ctx, len);
1261 static int ext4_set_context(struct inode *inode, const void *ctx, size_t len,
1264 handle_t *handle = fs_data;
1265 int res, res2, credits, retries = 0;
1268 * Encrypting the root directory is not allowed because e2fsck expects
1269 * lost+found to exist and be unencrypted, and encrypting the root
1270 * directory would imply encrypting the lost+found directory as well as
1271 * the filename "lost+found" itself.
1273 if (inode->i_ino == EXT4_ROOT_INO)
1276 if (WARN_ON_ONCE(IS_DAX(inode) && i_size_read(inode)))
1279 res = ext4_convert_inline_data(inode);
1284 * If a journal handle was specified, then the encryption context is
1285 * being set on a new inode via inheritance and is part of a larger
1286 * transaction to create the inode. Otherwise the encryption context is
1287 * being set on an existing inode in its own transaction. Only in the
1288 * latter case should the "retry on ENOSPC" logic be used.
1292 res = ext4_xattr_set_handle(handle, inode,
1293 EXT4_XATTR_INDEX_ENCRYPTION,
1294 EXT4_XATTR_NAME_ENCRYPTION_CONTEXT,
1297 ext4_set_inode_flag(inode, EXT4_INODE_ENCRYPT);
1298 ext4_clear_inode_state(inode,
1299 EXT4_STATE_MAY_INLINE_DATA);
1301 * Update inode->i_flags - S_ENCRYPTED will be enabled,
1302 * S_DAX may be disabled
1304 ext4_set_inode_flags(inode);
1309 res = dquot_initialize(inode);
1313 res = ext4_xattr_set_credits(inode, len, false /* is_create */,
1318 handle = ext4_journal_start(inode, EXT4_HT_MISC, credits);
1320 return PTR_ERR(handle);
1322 res = ext4_xattr_set_handle(handle, inode, EXT4_XATTR_INDEX_ENCRYPTION,
1323 EXT4_XATTR_NAME_ENCRYPTION_CONTEXT,
1326 ext4_set_inode_flag(inode, EXT4_INODE_ENCRYPT);
1328 * Update inode->i_flags - S_ENCRYPTED will be enabled,
1329 * S_DAX may be disabled
1331 ext4_set_inode_flags(inode);
1332 res = ext4_mark_inode_dirty(handle, inode);
1334 EXT4_ERROR_INODE(inode, "Failed to mark inode dirty");
1336 res2 = ext4_journal_stop(handle);
1338 if (res == -ENOSPC && ext4_should_retry_alloc(inode->i_sb, &retries))
1345 static bool ext4_dummy_context(struct inode *inode)
1347 return DUMMY_ENCRYPTION_ENABLED(EXT4_SB(inode->i_sb));
1350 static const struct fscrypt_operations ext4_cryptops = {
1351 .key_prefix = "ext4:",
1352 .get_context = ext4_get_context,
1353 .set_context = ext4_set_context,
1354 .dummy_context = ext4_dummy_context,
1355 .empty_dir = ext4_empty_dir,
1356 .max_namelen = EXT4_NAME_LEN,
1361 static const char * const quotatypes[] = INITQFNAMES;
1362 #define QTYPE2NAME(t) (quotatypes[t])
1364 static int ext4_write_dquot(struct dquot *dquot);
1365 static int ext4_acquire_dquot(struct dquot *dquot);
1366 static int ext4_release_dquot(struct dquot *dquot);
1367 static int ext4_mark_dquot_dirty(struct dquot *dquot);
1368 static int ext4_write_info(struct super_block *sb, int type);
1369 static int ext4_quota_on(struct super_block *sb, int type, int format_id,
1370 const struct path *path);
1371 static int ext4_quota_on_mount(struct super_block *sb, int type);
1372 static ssize_t ext4_quota_read(struct super_block *sb, int type, char *data,
1373 size_t len, loff_t off);
1374 static ssize_t ext4_quota_write(struct super_block *sb, int type,
1375 const char *data, size_t len, loff_t off);
1376 static int ext4_quota_enable(struct super_block *sb, int type, int format_id,
1377 unsigned int flags);
1378 static int ext4_enable_quotas(struct super_block *sb);
1379 static int ext4_get_next_id(struct super_block *sb, struct kqid *qid);
1381 static struct dquot **ext4_get_dquots(struct inode *inode)
1383 return EXT4_I(inode)->i_dquot;
1386 static const struct dquot_operations ext4_quota_operations = {
1387 .get_reserved_space = ext4_get_reserved_space,
1388 .write_dquot = ext4_write_dquot,
1389 .acquire_dquot = ext4_acquire_dquot,
1390 .release_dquot = ext4_release_dquot,
1391 .mark_dirty = ext4_mark_dquot_dirty,
1392 .write_info = ext4_write_info,
1393 .alloc_dquot = dquot_alloc,
1394 .destroy_dquot = dquot_destroy,
1395 .get_projid = ext4_get_projid,
1396 .get_inode_usage = ext4_get_inode_usage,
1397 .get_next_id = ext4_get_next_id,
1400 static const struct quotactl_ops ext4_qctl_operations = {
1401 .quota_on = ext4_quota_on,
1402 .quota_off = ext4_quota_off,
1403 .quota_sync = dquot_quota_sync,
1404 .get_state = dquot_get_state,
1405 .set_info = dquot_set_dqinfo,
1406 .get_dqblk = dquot_get_dqblk,
1407 .set_dqblk = dquot_set_dqblk,
1408 .get_nextdqblk = dquot_get_next_dqblk,
1412 static const struct super_operations ext4_sops = {
1413 .alloc_inode = ext4_alloc_inode,
1414 .destroy_inode = ext4_destroy_inode,
1415 .write_inode = ext4_write_inode,
1416 .dirty_inode = ext4_dirty_inode,
1417 .drop_inode = ext4_drop_inode,
1418 .evict_inode = ext4_evict_inode,
1419 .put_super = ext4_put_super,
1420 .sync_fs = ext4_sync_fs,
1421 .freeze_fs = ext4_freeze,
1422 .unfreeze_fs = ext4_unfreeze,
1423 .statfs = ext4_statfs,
1424 .remount_fs = ext4_remount,
1425 .show_options = ext4_show_options,
1427 .quota_read = ext4_quota_read,
1428 .quota_write = ext4_quota_write,
1429 .get_dquots = ext4_get_dquots,
1431 .bdev_try_to_free_page = bdev_try_to_free_page,
1434 static const struct export_operations ext4_export_ops = {
1435 .fh_to_dentry = ext4_fh_to_dentry,
1436 .fh_to_parent = ext4_fh_to_parent,
1437 .get_parent = ext4_get_parent,
1438 .commit_metadata = ext4_nfs_commit_metadata,
1442 Opt_bsd_df, Opt_minix_df, Opt_grpid, Opt_nogrpid,
1443 Opt_resgid, Opt_resuid, Opt_sb, Opt_err_cont, Opt_err_panic, Opt_err_ro,
1444 Opt_nouid32, Opt_debug, Opt_removed,
1445 Opt_user_xattr, Opt_nouser_xattr, Opt_acl, Opt_noacl,
1446 Opt_auto_da_alloc, Opt_noauto_da_alloc, Opt_noload,
1447 Opt_commit, Opt_min_batch_time, Opt_max_batch_time, Opt_journal_dev,
1448 Opt_journal_path, Opt_journal_checksum, Opt_journal_async_commit,
1449 Opt_abort, Opt_data_journal, Opt_data_ordered, Opt_data_writeback,
1450 Opt_data_err_abort, Opt_data_err_ignore, Opt_test_dummy_encryption,
1451 Opt_usrjquota, Opt_grpjquota, Opt_offusrjquota, Opt_offgrpjquota,
1452 Opt_jqfmt_vfsold, Opt_jqfmt_vfsv0, Opt_jqfmt_vfsv1, Opt_quota,
1453 Opt_noquota, Opt_barrier, Opt_nobarrier, Opt_err,
1454 Opt_usrquota, Opt_grpquota, Opt_prjquota, Opt_i_version, Opt_dax,
1455 Opt_stripe, Opt_delalloc, Opt_nodelalloc, Opt_warn_on_error,
1456 Opt_nowarn_on_error, Opt_mblk_io_submit,
1457 Opt_lazytime, Opt_nolazytime, Opt_debug_want_extra_isize,
1458 Opt_nomblk_io_submit, Opt_block_validity, Opt_noblock_validity,
1459 Opt_inode_readahead_blks, Opt_journal_ioprio,
1460 Opt_dioread_nolock, Opt_dioread_lock,
1461 Opt_discard, Opt_nodiscard, Opt_init_itable, Opt_noinit_itable,
1462 Opt_max_dir_size_kb, Opt_nojournal_checksum, Opt_nombcache,
1465 static const match_table_t tokens = {
1466 {Opt_bsd_df, "bsddf"},
1467 {Opt_minix_df, "minixdf"},
1468 {Opt_grpid, "grpid"},
1469 {Opt_grpid, "bsdgroups"},
1470 {Opt_nogrpid, "nogrpid"},
1471 {Opt_nogrpid, "sysvgroups"},
1472 {Opt_resgid, "resgid=%u"},
1473 {Opt_resuid, "resuid=%u"},
1475 {Opt_err_cont, "errors=continue"},
1476 {Opt_err_panic, "errors=panic"},
1477 {Opt_err_ro, "errors=remount-ro"},
1478 {Opt_nouid32, "nouid32"},
1479 {Opt_debug, "debug"},
1480 {Opt_removed, "oldalloc"},
1481 {Opt_removed, "orlov"},
1482 {Opt_user_xattr, "user_xattr"},
1483 {Opt_nouser_xattr, "nouser_xattr"},
1485 {Opt_noacl, "noacl"},
1486 {Opt_noload, "norecovery"},
1487 {Opt_noload, "noload"},
1488 {Opt_removed, "nobh"},
1489 {Opt_removed, "bh"},
1490 {Opt_commit, "commit=%u"},
1491 {Opt_min_batch_time, "min_batch_time=%u"},
1492 {Opt_max_batch_time, "max_batch_time=%u"},
1493 {Opt_journal_dev, "journal_dev=%u"},
1494 {Opt_journal_path, "journal_path=%s"},
1495 {Opt_journal_checksum, "journal_checksum"},
1496 {Opt_nojournal_checksum, "nojournal_checksum"},
1497 {Opt_journal_async_commit, "journal_async_commit"},
1498 {Opt_abort, "abort"},
1499 {Opt_data_journal, "data=journal"},
1500 {Opt_data_ordered, "data=ordered"},
1501 {Opt_data_writeback, "data=writeback"},
1502 {Opt_data_err_abort, "data_err=abort"},
1503 {Opt_data_err_ignore, "data_err=ignore"},
1504 {Opt_offusrjquota, "usrjquota="},
1505 {Opt_usrjquota, "usrjquota=%s"},
1506 {Opt_offgrpjquota, "grpjquota="},
1507 {Opt_grpjquota, "grpjquota=%s"},
1508 {Opt_jqfmt_vfsold, "jqfmt=vfsold"},
1509 {Opt_jqfmt_vfsv0, "jqfmt=vfsv0"},
1510 {Opt_jqfmt_vfsv1, "jqfmt=vfsv1"},
1511 {Opt_grpquota, "grpquota"},
1512 {Opt_noquota, "noquota"},
1513 {Opt_quota, "quota"},
1514 {Opt_usrquota, "usrquota"},
1515 {Opt_prjquota, "prjquota"},
1516 {Opt_barrier, "barrier=%u"},
1517 {Opt_barrier, "barrier"},
1518 {Opt_nobarrier, "nobarrier"},
1519 {Opt_i_version, "i_version"},
1521 {Opt_stripe, "stripe=%u"},
1522 {Opt_delalloc, "delalloc"},
1523 {Opt_warn_on_error, "warn_on_error"},
1524 {Opt_nowarn_on_error, "nowarn_on_error"},
1525 {Opt_lazytime, "lazytime"},
1526 {Opt_nolazytime, "nolazytime"},
1527 {Opt_debug_want_extra_isize, "debug_want_extra_isize=%u"},
1528 {Opt_nodelalloc, "nodelalloc"},
1529 {Opt_removed, "mblk_io_submit"},
1530 {Opt_removed, "nomblk_io_submit"},
1531 {Opt_block_validity, "block_validity"},
1532 {Opt_noblock_validity, "noblock_validity"},
1533 {Opt_inode_readahead_blks, "inode_readahead_blks=%u"},
1534 {Opt_journal_ioprio, "journal_ioprio=%u"},
1535 {Opt_auto_da_alloc, "auto_da_alloc=%u"},
1536 {Opt_auto_da_alloc, "auto_da_alloc"},
1537 {Opt_noauto_da_alloc, "noauto_da_alloc"},
1538 {Opt_dioread_nolock, "dioread_nolock"},
1539 {Opt_dioread_lock, "dioread_lock"},
1540 {Opt_discard, "discard"},
1541 {Opt_nodiscard, "nodiscard"},
1542 {Opt_init_itable, "init_itable=%u"},
1543 {Opt_init_itable, "init_itable"},
1544 {Opt_noinit_itable, "noinit_itable"},
1545 {Opt_max_dir_size_kb, "max_dir_size_kb=%u"},
1546 {Opt_test_dummy_encryption, "test_dummy_encryption"},
1547 {Opt_nombcache, "nombcache"},
1548 {Opt_nombcache, "no_mbcache"}, /* for backward compatibility */
1549 {Opt_removed, "check=none"}, /* mount option from ext2/3 */
1550 {Opt_removed, "nocheck"}, /* mount option from ext2/3 */
1551 {Opt_removed, "reservation"}, /* mount option from ext2/3 */
1552 {Opt_removed, "noreservation"}, /* mount option from ext2/3 */
1553 {Opt_removed, "journal=%u"}, /* mount option from ext2/3 */
1557 static ext4_fsblk_t get_sb_block(void **data)
1559 ext4_fsblk_t sb_block;
1560 char *options = (char *) *data;
1562 if (!options || strncmp(options, "sb=", 3) != 0)
1563 return 1; /* Default location */
1566 /* TODO: use simple_strtoll with >32bit ext4 */
1567 sb_block = simple_strtoul(options, &options, 0);
1568 if (*options && *options != ',') {
1569 printk(KERN_ERR "EXT4-fs: Invalid sb specification: %s\n",
1573 if (*options == ',')
1575 *data = (void *) options;
1580 #define DEFAULT_JOURNAL_IOPRIO (IOPRIO_PRIO_VALUE(IOPRIO_CLASS_BE, 3))
1581 static const char deprecated_msg[] =
1582 "Mount option \"%s\" will be removed by %s\n"
1583 "Contact linux-ext4@vger.kernel.org if you think we should keep it.\n";
1586 static int set_qf_name(struct super_block *sb, int qtype, substring_t *args)
1588 struct ext4_sb_info *sbi = EXT4_SB(sb);
1589 char *qname, *old_qname = get_qf_name(sb, sbi, qtype);
1592 if (sb_any_quota_loaded(sb) && !old_qname) {
1593 ext4_msg(sb, KERN_ERR,
1594 "Cannot change journaled "
1595 "quota options when quota turned on");
1598 if (ext4_has_feature_quota(sb)) {
1599 ext4_msg(sb, KERN_INFO, "Journaled quota options "
1600 "ignored when QUOTA feature is enabled");
1603 qname = match_strdup(args);
1605 ext4_msg(sb, KERN_ERR,
1606 "Not enough memory for storing quotafile name");
1610 if (strcmp(old_qname, qname) == 0)
1613 ext4_msg(sb, KERN_ERR,
1614 "%s quota file already specified",
1618 if (strchr(qname, '/')) {
1619 ext4_msg(sb, KERN_ERR,
1620 "quotafile must be on filesystem root");
1623 rcu_assign_pointer(sbi->s_qf_names[qtype], qname);
1631 static int clear_qf_name(struct super_block *sb, int qtype)
1634 struct ext4_sb_info *sbi = EXT4_SB(sb);
1635 char *old_qname = get_qf_name(sb, sbi, qtype);
1637 if (sb_any_quota_loaded(sb) && old_qname) {
1638 ext4_msg(sb, KERN_ERR, "Cannot change journaled quota options"
1639 " when quota turned on");
1642 rcu_assign_pointer(sbi->s_qf_names[qtype], NULL);
1649 #define MOPT_SET 0x0001
1650 #define MOPT_CLEAR 0x0002
1651 #define MOPT_NOSUPPORT 0x0004
1652 #define MOPT_EXPLICIT 0x0008
1653 #define MOPT_CLEAR_ERR 0x0010
1654 #define MOPT_GTE0 0x0020
1657 #define MOPT_QFMT 0x0040
1659 #define MOPT_Q MOPT_NOSUPPORT
1660 #define MOPT_QFMT MOPT_NOSUPPORT
1662 #define MOPT_DATAJ 0x0080
1663 #define MOPT_NO_EXT2 0x0100
1664 #define MOPT_NO_EXT3 0x0200
1665 #define MOPT_EXT4_ONLY (MOPT_NO_EXT2 | MOPT_NO_EXT3)
1666 #define MOPT_STRING 0x0400
1668 static const struct mount_opts {
1672 } ext4_mount_opts[] = {
1673 {Opt_minix_df, EXT4_MOUNT_MINIX_DF, MOPT_SET},
1674 {Opt_bsd_df, EXT4_MOUNT_MINIX_DF, MOPT_CLEAR},
1675 {Opt_grpid, EXT4_MOUNT_GRPID, MOPT_SET},
1676 {Opt_nogrpid, EXT4_MOUNT_GRPID, MOPT_CLEAR},
1677 {Opt_block_validity, EXT4_MOUNT_BLOCK_VALIDITY, MOPT_SET},
1678 {Opt_noblock_validity, EXT4_MOUNT_BLOCK_VALIDITY, MOPT_CLEAR},
1679 {Opt_dioread_nolock, EXT4_MOUNT_DIOREAD_NOLOCK,
1680 MOPT_EXT4_ONLY | MOPT_SET},
1681 {Opt_dioread_lock, EXT4_MOUNT_DIOREAD_NOLOCK,
1682 MOPT_EXT4_ONLY | MOPT_CLEAR},
1683 {Opt_discard, EXT4_MOUNT_DISCARD, MOPT_SET},
1684 {Opt_nodiscard, EXT4_MOUNT_DISCARD, MOPT_CLEAR},
1685 {Opt_delalloc, EXT4_MOUNT_DELALLOC,
1686 MOPT_EXT4_ONLY | MOPT_SET | MOPT_EXPLICIT},
1687 {Opt_nodelalloc, EXT4_MOUNT_DELALLOC,
1688 MOPT_EXT4_ONLY | MOPT_CLEAR},
1689 {Opt_warn_on_error, EXT4_MOUNT_WARN_ON_ERROR, MOPT_SET},
1690 {Opt_nowarn_on_error, EXT4_MOUNT_WARN_ON_ERROR, MOPT_CLEAR},
1691 {Opt_nojournal_checksum, EXT4_MOUNT_JOURNAL_CHECKSUM,
1692 MOPT_EXT4_ONLY | MOPT_CLEAR},
1693 {Opt_journal_checksum, EXT4_MOUNT_JOURNAL_CHECKSUM,
1694 MOPT_EXT4_ONLY | MOPT_SET | MOPT_EXPLICIT},
1695 {Opt_journal_async_commit, (EXT4_MOUNT_JOURNAL_ASYNC_COMMIT |
1696 EXT4_MOUNT_JOURNAL_CHECKSUM),
1697 MOPT_EXT4_ONLY | MOPT_SET | MOPT_EXPLICIT},
1698 {Opt_noload, EXT4_MOUNT_NOLOAD, MOPT_NO_EXT2 | MOPT_SET},
1699 {Opt_err_panic, EXT4_MOUNT_ERRORS_PANIC, MOPT_SET | MOPT_CLEAR_ERR},
1700 {Opt_err_ro, EXT4_MOUNT_ERRORS_RO, MOPT_SET | MOPT_CLEAR_ERR},
1701 {Opt_err_cont, EXT4_MOUNT_ERRORS_CONT, MOPT_SET | MOPT_CLEAR_ERR},
1702 {Opt_data_err_abort, EXT4_MOUNT_DATA_ERR_ABORT,
1704 {Opt_data_err_ignore, EXT4_MOUNT_DATA_ERR_ABORT,
1706 {Opt_barrier, EXT4_MOUNT_BARRIER, MOPT_SET},
1707 {Opt_nobarrier, EXT4_MOUNT_BARRIER, MOPT_CLEAR},
1708 {Opt_noauto_da_alloc, EXT4_MOUNT_NO_AUTO_DA_ALLOC, MOPT_SET},
1709 {Opt_auto_da_alloc, EXT4_MOUNT_NO_AUTO_DA_ALLOC, MOPT_CLEAR},
1710 {Opt_noinit_itable, EXT4_MOUNT_INIT_INODE_TABLE, MOPT_CLEAR},
1711 {Opt_commit, 0, MOPT_GTE0},
1712 {Opt_max_batch_time, 0, MOPT_GTE0},
1713 {Opt_min_batch_time, 0, MOPT_GTE0},
1714 {Opt_inode_readahead_blks, 0, MOPT_GTE0},
1715 {Opt_init_itable, 0, MOPT_GTE0},
1716 {Opt_dax, EXT4_MOUNT_DAX, MOPT_SET},
1717 {Opt_stripe, 0, MOPT_GTE0},
1718 {Opt_resuid, 0, MOPT_GTE0},
1719 {Opt_resgid, 0, MOPT_GTE0},
1720 {Opt_journal_dev, 0, MOPT_NO_EXT2 | MOPT_GTE0},
1721 {Opt_journal_path, 0, MOPT_NO_EXT2 | MOPT_STRING},
1722 {Opt_journal_ioprio, 0, MOPT_NO_EXT2 | MOPT_GTE0},
1723 {Opt_data_journal, EXT4_MOUNT_JOURNAL_DATA, MOPT_NO_EXT2 | MOPT_DATAJ},
1724 {Opt_data_ordered, EXT4_MOUNT_ORDERED_DATA, MOPT_NO_EXT2 | MOPT_DATAJ},
1725 {Opt_data_writeback, EXT4_MOUNT_WRITEBACK_DATA,
1726 MOPT_NO_EXT2 | MOPT_DATAJ},
1727 {Opt_user_xattr, EXT4_MOUNT_XATTR_USER, MOPT_SET},
1728 {Opt_nouser_xattr, EXT4_MOUNT_XATTR_USER, MOPT_CLEAR},
1729 #ifdef CONFIG_EXT4_FS_POSIX_ACL
1730 {Opt_acl, EXT4_MOUNT_POSIX_ACL, MOPT_SET},
1731 {Opt_noacl, EXT4_MOUNT_POSIX_ACL, MOPT_CLEAR},
1733 {Opt_acl, 0, MOPT_NOSUPPORT},
1734 {Opt_noacl, 0, MOPT_NOSUPPORT},
1736 {Opt_nouid32, EXT4_MOUNT_NO_UID32, MOPT_SET},
1737 {Opt_debug, EXT4_MOUNT_DEBUG, MOPT_SET},
1738 {Opt_debug_want_extra_isize, 0, MOPT_GTE0},
1739 {Opt_quota, EXT4_MOUNT_QUOTA | EXT4_MOUNT_USRQUOTA, MOPT_SET | MOPT_Q},
1740 {Opt_usrquota, EXT4_MOUNT_QUOTA | EXT4_MOUNT_USRQUOTA,
1742 {Opt_grpquota, EXT4_MOUNT_QUOTA | EXT4_MOUNT_GRPQUOTA,
1744 {Opt_prjquota, EXT4_MOUNT_QUOTA | EXT4_MOUNT_PRJQUOTA,
1746 {Opt_noquota, (EXT4_MOUNT_QUOTA | EXT4_MOUNT_USRQUOTA |
1747 EXT4_MOUNT_GRPQUOTA | EXT4_MOUNT_PRJQUOTA),
1748 MOPT_CLEAR | MOPT_Q},
1749 {Opt_usrjquota, 0, MOPT_Q | MOPT_STRING},
1750 {Opt_grpjquota, 0, MOPT_Q | MOPT_STRING},
1751 {Opt_offusrjquota, 0, MOPT_Q},
1752 {Opt_offgrpjquota, 0, MOPT_Q},
1753 {Opt_jqfmt_vfsold, QFMT_VFS_OLD, MOPT_QFMT},
1754 {Opt_jqfmt_vfsv0, QFMT_VFS_V0, MOPT_QFMT},
1755 {Opt_jqfmt_vfsv1, QFMT_VFS_V1, MOPT_QFMT},
1756 {Opt_max_dir_size_kb, 0, MOPT_GTE0},
1757 {Opt_test_dummy_encryption, 0, MOPT_GTE0},
1758 {Opt_nombcache, EXT4_MOUNT_NO_MBCACHE, MOPT_SET},
1762 static int handle_mount_opt(struct super_block *sb, char *opt, int token,
1763 substring_t *args, unsigned long *journal_devnum,
1764 unsigned int *journal_ioprio, int is_remount)
1766 struct ext4_sb_info *sbi = EXT4_SB(sb);
1767 const struct mount_opts *m;
1773 if (token == Opt_usrjquota)
1774 return set_qf_name(sb, USRQUOTA, &args[0]);
1775 else if (token == Opt_grpjquota)
1776 return set_qf_name(sb, GRPQUOTA, &args[0]);
1777 else if (token == Opt_offusrjquota)
1778 return clear_qf_name(sb, USRQUOTA);
1779 else if (token == Opt_offgrpjquota)
1780 return clear_qf_name(sb, GRPQUOTA);
1784 case Opt_nouser_xattr:
1785 ext4_msg(sb, KERN_WARNING, deprecated_msg, opt, "3.5");
1788 return 1; /* handled by get_sb_block() */
1790 ext4_msg(sb, KERN_WARNING, "Ignoring removed %s option", opt);
1793 sbi->s_mount_flags |= EXT4_MF_FS_ABORTED;
1796 sb->s_flags |= SB_I_VERSION;
1799 sb->s_flags |= SB_LAZYTIME;
1801 case Opt_nolazytime:
1802 sb->s_flags &= ~SB_LAZYTIME;
1806 for (m = ext4_mount_opts; m->token != Opt_err; m++)
1807 if (token == m->token)
1810 if (m->token == Opt_err) {
1811 ext4_msg(sb, KERN_ERR, "Unrecognized mount option \"%s\" "
1812 "or missing value", opt);
1816 if ((m->flags & MOPT_NO_EXT2) && IS_EXT2_SB(sb)) {
1817 ext4_msg(sb, KERN_ERR,
1818 "Mount option \"%s\" incompatible with ext2", opt);
1821 if ((m->flags & MOPT_NO_EXT3) && IS_EXT3_SB(sb)) {
1822 ext4_msg(sb, KERN_ERR,
1823 "Mount option \"%s\" incompatible with ext3", opt);
1827 if (args->from && !(m->flags & MOPT_STRING) && match_int(args, &arg))
1829 if (args->from && (m->flags & MOPT_GTE0) && (arg < 0))
1831 if (m->flags & MOPT_EXPLICIT) {
1832 if (m->mount_opt & EXT4_MOUNT_DELALLOC) {
1833 set_opt2(sb, EXPLICIT_DELALLOC);
1834 } else if (m->mount_opt & EXT4_MOUNT_JOURNAL_CHECKSUM) {
1835 set_opt2(sb, EXPLICIT_JOURNAL_CHECKSUM);
1839 if (m->flags & MOPT_CLEAR_ERR)
1840 clear_opt(sb, ERRORS_MASK);
1841 if (token == Opt_noquota && sb_any_quota_loaded(sb)) {
1842 ext4_msg(sb, KERN_ERR, "Cannot change quota "
1843 "options when quota turned on");
1847 if (m->flags & MOPT_NOSUPPORT) {
1848 ext4_msg(sb, KERN_ERR, "%s option not supported", opt);
1849 } else if (token == Opt_commit) {
1851 arg = JBD2_DEFAULT_MAX_COMMIT_AGE;
1852 sbi->s_commit_interval = HZ * arg;
1853 } else if (token == Opt_debug_want_extra_isize) {
1856 (arg > (sbi->s_inode_size - EXT4_GOOD_OLD_INODE_SIZE))) {
1857 ext4_msg(sb, KERN_ERR,
1858 "Invalid want_extra_isize %d", arg);
1861 sbi->s_want_extra_isize = arg;
1862 } else if (token == Opt_max_batch_time) {
1863 sbi->s_max_batch_time = arg;
1864 } else if (token == Opt_min_batch_time) {
1865 sbi->s_min_batch_time = arg;
1866 } else if (token == Opt_inode_readahead_blks) {
1867 if (arg && (arg > (1 << 30) || !is_power_of_2(arg))) {
1868 ext4_msg(sb, KERN_ERR,
1869 "EXT4-fs: inode_readahead_blks must be "
1870 "0 or a power of 2 smaller than 2^31");
1873 sbi->s_inode_readahead_blks = arg;
1874 } else if (token == Opt_init_itable) {
1875 set_opt(sb, INIT_INODE_TABLE);
1877 arg = EXT4_DEF_LI_WAIT_MULT;
1878 sbi->s_li_wait_mult = arg;
1879 } else if (token == Opt_max_dir_size_kb) {
1880 sbi->s_max_dir_size_kb = arg;
1881 } else if (token == Opt_stripe) {
1882 sbi->s_stripe = arg;
1883 } else if (token == Opt_resuid) {
1884 uid = make_kuid(current_user_ns(), arg);
1885 if (!uid_valid(uid)) {
1886 ext4_msg(sb, KERN_ERR, "Invalid uid value %d", arg);
1889 sbi->s_resuid = uid;
1890 } else if (token == Opt_resgid) {
1891 gid = make_kgid(current_user_ns(), arg);
1892 if (!gid_valid(gid)) {
1893 ext4_msg(sb, KERN_ERR, "Invalid gid value %d", arg);
1896 sbi->s_resgid = gid;
1897 } else if (token == Opt_journal_dev) {
1899 ext4_msg(sb, KERN_ERR,
1900 "Cannot specify journal on remount");
1903 *journal_devnum = arg;
1904 } else if (token == Opt_journal_path) {
1906 struct inode *journal_inode;
1911 ext4_msg(sb, KERN_ERR,
1912 "Cannot specify journal on remount");
1915 journal_path = match_strdup(&args[0]);
1916 if (!journal_path) {
1917 ext4_msg(sb, KERN_ERR, "error: could not dup "
1918 "journal device string");
1922 error = kern_path(journal_path, LOOKUP_FOLLOW, &path);
1924 ext4_msg(sb, KERN_ERR, "error: could not find "
1925 "journal device path: error %d", error);
1926 kfree(journal_path);
1930 journal_inode = d_inode(path.dentry);
1931 if (!S_ISBLK(journal_inode->i_mode)) {
1932 ext4_msg(sb, KERN_ERR, "error: journal path %s "
1933 "is not a block device", journal_path);
1935 kfree(journal_path);
1939 *journal_devnum = new_encode_dev(journal_inode->i_rdev);
1941 kfree(journal_path);
1942 } else if (token == Opt_journal_ioprio) {
1944 ext4_msg(sb, KERN_ERR, "Invalid journal IO priority"
1949 IOPRIO_PRIO_VALUE(IOPRIO_CLASS_BE, arg);
1950 } else if (token == Opt_test_dummy_encryption) {
1951 #ifdef CONFIG_EXT4_FS_ENCRYPTION
1952 sbi->s_mount_flags |= EXT4_MF_TEST_DUMMY_ENCRYPTION;
1953 ext4_msg(sb, KERN_WARNING,
1954 "Test dummy encryption mode enabled");
1956 ext4_msg(sb, KERN_WARNING,
1957 "Test dummy encryption mount option ignored");
1959 } else if (m->flags & MOPT_DATAJ) {
1961 if (!sbi->s_journal)
1962 ext4_msg(sb, KERN_WARNING, "Remounting file system with no journal so ignoring journalled data option");
1963 else if (test_opt(sb, DATA_FLAGS) != m->mount_opt) {
1964 ext4_msg(sb, KERN_ERR,
1965 "Cannot change data mode on remount");
1969 clear_opt(sb, DATA_FLAGS);
1970 sbi->s_mount_opt |= m->mount_opt;
1973 } else if (m->flags & MOPT_QFMT) {
1974 if (sb_any_quota_loaded(sb) &&
1975 sbi->s_jquota_fmt != m->mount_opt) {
1976 ext4_msg(sb, KERN_ERR, "Cannot change journaled "
1977 "quota options when quota turned on");
1980 if (ext4_has_feature_quota(sb)) {
1981 ext4_msg(sb, KERN_INFO,
1982 "Quota format mount options ignored "
1983 "when QUOTA feature is enabled");
1986 sbi->s_jquota_fmt = m->mount_opt;
1988 } else if (token == Opt_dax) {
1989 #ifdef CONFIG_FS_DAX
1990 if (is_remount && test_opt(sb, DAX)) {
1991 ext4_msg(sb, KERN_ERR, "can't mount with "
1992 "both data=journal and dax");
1995 if (is_remount && !(sbi->s_mount_opt & EXT4_MOUNT_DAX)) {
1996 ext4_msg(sb, KERN_ERR, "can't change "
1997 "dax mount option while remounting");
2000 ext4_msg(sb, KERN_WARNING,
2001 "DAX enabled. Warning: EXPERIMENTAL, use at your own risk");
2002 sbi->s_mount_opt |= m->mount_opt;
2004 ext4_msg(sb, KERN_INFO, "dax option not supported");
2007 } else if (token == Opt_data_err_abort) {
2008 sbi->s_mount_opt |= m->mount_opt;
2009 } else if (token == Opt_data_err_ignore) {
2010 sbi->s_mount_opt &= ~m->mount_opt;
2014 if (m->flags & MOPT_CLEAR)
2016 else if (unlikely(!(m->flags & MOPT_SET))) {
2017 ext4_msg(sb, KERN_WARNING,
2018 "buggy handling of option %s", opt);
2023 sbi->s_mount_opt |= m->mount_opt;
2025 sbi->s_mount_opt &= ~m->mount_opt;
2030 static int parse_options(char *options, struct super_block *sb,
2031 unsigned long *journal_devnum,
2032 unsigned int *journal_ioprio,
2035 struct ext4_sb_info *sbi = EXT4_SB(sb);
2036 char *p, __maybe_unused *usr_qf_name, __maybe_unused *grp_qf_name;
2037 substring_t args[MAX_OPT_ARGS];
2043 while ((p = strsep(&options, ",")) != NULL) {
2047 * Initialize args struct so we know whether arg was
2048 * found; some options take optional arguments.
2050 args[0].to = args[0].from = NULL;
2051 token = match_token(p, tokens, args);
2052 if (handle_mount_opt(sb, p, token, args, journal_devnum,
2053 journal_ioprio, is_remount) < 0)
2058 * We do the test below only for project quotas. 'usrquota' and
2059 * 'grpquota' mount options are allowed even without quota feature
2060 * to support legacy quotas in quota files.
2062 if (test_opt(sb, PRJQUOTA) && !ext4_has_feature_project(sb)) {
2063 ext4_msg(sb, KERN_ERR, "Project quota feature not enabled. "
2064 "Cannot enable project quota enforcement.");
2067 usr_qf_name = get_qf_name(sb, sbi, USRQUOTA);
2068 grp_qf_name = get_qf_name(sb, sbi, GRPQUOTA);
2069 if (usr_qf_name || grp_qf_name) {
2070 if (test_opt(sb, USRQUOTA) && usr_qf_name)
2071 clear_opt(sb, USRQUOTA);
2073 if (test_opt(sb, GRPQUOTA) && grp_qf_name)
2074 clear_opt(sb, GRPQUOTA);
2076 if (test_opt(sb, GRPQUOTA) || test_opt(sb, USRQUOTA)) {
2077 ext4_msg(sb, KERN_ERR, "old and new quota "
2082 if (!sbi->s_jquota_fmt) {
2083 ext4_msg(sb, KERN_ERR, "journaled quota format "
2089 if (test_opt(sb, DIOREAD_NOLOCK)) {
2091 BLOCK_SIZE << le32_to_cpu(sbi->s_es->s_log_block_size);
2093 if (blocksize < PAGE_SIZE) {
2094 ext4_msg(sb, KERN_ERR, "can't mount with "
2095 "dioread_nolock if block size != PAGE_SIZE");
2102 static inline void ext4_show_quota_options(struct seq_file *seq,
2103 struct super_block *sb)
2105 #if defined(CONFIG_QUOTA)
2106 struct ext4_sb_info *sbi = EXT4_SB(sb);
2107 char *usr_qf_name, *grp_qf_name;
2109 if (sbi->s_jquota_fmt) {
2112 switch (sbi->s_jquota_fmt) {
2123 seq_printf(seq, ",jqfmt=%s", fmtname);
2127 usr_qf_name = rcu_dereference(sbi->s_qf_names[USRQUOTA]);
2128 grp_qf_name = rcu_dereference(sbi->s_qf_names[GRPQUOTA]);
2130 seq_show_option(seq, "usrjquota", usr_qf_name);
2132 seq_show_option(seq, "grpjquota", grp_qf_name);
2137 static const char *token2str(int token)
2139 const struct match_token *t;
2141 for (t = tokens; t->token != Opt_err; t++)
2142 if (t->token == token && !strchr(t->pattern, '='))
2149 * - it's set to a non-default value OR
2150 * - if the per-sb default is different from the global default
2152 static int _ext4_show_options(struct seq_file *seq, struct super_block *sb,
2155 struct ext4_sb_info *sbi = EXT4_SB(sb);
2156 struct ext4_super_block *es = sbi->s_es;
2157 int def_errors, def_mount_opt = sbi->s_def_mount_opt;
2158 const struct mount_opts *m;
2159 char sep = nodefs ? '\n' : ',';
2161 #define SEQ_OPTS_PUTS(str) seq_printf(seq, "%c" str, sep)
2162 #define SEQ_OPTS_PRINT(str, arg) seq_printf(seq, "%c" str, sep, arg)
2164 if (sbi->s_sb_block != 1)
2165 SEQ_OPTS_PRINT("sb=%llu", sbi->s_sb_block);
2167 for (m = ext4_mount_opts; m->token != Opt_err; m++) {
2168 int want_set = m->flags & MOPT_SET;
2169 if (((m->flags & (MOPT_SET|MOPT_CLEAR)) == 0) ||
2170 (m->flags & MOPT_CLEAR_ERR))
2172 if (!nodefs && !(m->mount_opt & (sbi->s_mount_opt ^ def_mount_opt)))
2173 continue; /* skip if same as the default */
2175 (sbi->s_mount_opt & m->mount_opt) != m->mount_opt) ||
2176 (!want_set && (sbi->s_mount_opt & m->mount_opt)))
2177 continue; /* select Opt_noFoo vs Opt_Foo */
2178 SEQ_OPTS_PRINT("%s", token2str(m->token));
2181 if (nodefs || !uid_eq(sbi->s_resuid, make_kuid(&init_user_ns, EXT4_DEF_RESUID)) ||
2182 le16_to_cpu(es->s_def_resuid) != EXT4_DEF_RESUID)
2183 SEQ_OPTS_PRINT("resuid=%u",
2184 from_kuid_munged(&init_user_ns, sbi->s_resuid));
2185 if (nodefs || !gid_eq(sbi->s_resgid, make_kgid(&init_user_ns, EXT4_DEF_RESGID)) ||
2186 le16_to_cpu(es->s_def_resgid) != EXT4_DEF_RESGID)
2187 SEQ_OPTS_PRINT("resgid=%u",
2188 from_kgid_munged(&init_user_ns, sbi->s_resgid));
2189 def_errors = nodefs ? -1 : le16_to_cpu(es->s_errors);
2190 if (test_opt(sb, ERRORS_RO) && def_errors != EXT4_ERRORS_RO)
2191 SEQ_OPTS_PUTS("errors=remount-ro");
2192 if (test_opt(sb, ERRORS_CONT) && def_errors != EXT4_ERRORS_CONTINUE)
2193 SEQ_OPTS_PUTS("errors=continue");
2194 if (test_opt(sb, ERRORS_PANIC) && def_errors != EXT4_ERRORS_PANIC)
2195 SEQ_OPTS_PUTS("errors=panic");
2196 if (nodefs || sbi->s_commit_interval != JBD2_DEFAULT_MAX_COMMIT_AGE*HZ)
2197 SEQ_OPTS_PRINT("commit=%lu", sbi->s_commit_interval / HZ);
2198 if (nodefs || sbi->s_min_batch_time != EXT4_DEF_MIN_BATCH_TIME)
2199 SEQ_OPTS_PRINT("min_batch_time=%u", sbi->s_min_batch_time);
2200 if (nodefs || sbi->s_max_batch_time != EXT4_DEF_MAX_BATCH_TIME)
2201 SEQ_OPTS_PRINT("max_batch_time=%u", sbi->s_max_batch_time);
2202 if (sb->s_flags & SB_I_VERSION)
2203 SEQ_OPTS_PUTS("i_version");
2204 if (nodefs || sbi->s_stripe)
2205 SEQ_OPTS_PRINT("stripe=%lu", sbi->s_stripe);
2206 if (nodefs || EXT4_MOUNT_DATA_FLAGS &
2207 (sbi->s_mount_opt ^ def_mount_opt)) {
2208 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA)
2209 SEQ_OPTS_PUTS("data=journal");
2210 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA)
2211 SEQ_OPTS_PUTS("data=ordered");
2212 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_WRITEBACK_DATA)
2213 SEQ_OPTS_PUTS("data=writeback");
2216 sbi->s_inode_readahead_blks != EXT4_DEF_INODE_READAHEAD_BLKS)
2217 SEQ_OPTS_PRINT("inode_readahead_blks=%u",
2218 sbi->s_inode_readahead_blks);
2220 if (test_opt(sb, INIT_INODE_TABLE) && (nodefs ||
2221 (sbi->s_li_wait_mult != EXT4_DEF_LI_WAIT_MULT)))
2222 SEQ_OPTS_PRINT("init_itable=%u", sbi->s_li_wait_mult);
2223 if (nodefs || sbi->s_max_dir_size_kb)
2224 SEQ_OPTS_PRINT("max_dir_size_kb=%u", sbi->s_max_dir_size_kb);
2225 if (test_opt(sb, DATA_ERR_ABORT))
2226 SEQ_OPTS_PUTS("data_err=abort");
2227 if (DUMMY_ENCRYPTION_ENABLED(sbi))
2228 SEQ_OPTS_PUTS("test_dummy_encryption");
2230 ext4_show_quota_options(seq, sb);
2234 static int ext4_show_options(struct seq_file *seq, struct dentry *root)
2236 return _ext4_show_options(seq, root->d_sb, 0);
2239 int ext4_seq_options_show(struct seq_file *seq, void *offset)
2241 struct super_block *sb = seq->private;
2244 seq_puts(seq, sb_rdonly(sb) ? "ro" : "rw");
2245 rc = _ext4_show_options(seq, sb, 1);
2246 seq_puts(seq, "\n");
2250 static int ext4_setup_super(struct super_block *sb, struct ext4_super_block *es,
2253 struct ext4_sb_info *sbi = EXT4_SB(sb);
2256 if (le32_to_cpu(es->s_rev_level) > EXT4_MAX_SUPP_REV) {
2257 ext4_msg(sb, KERN_ERR, "revision level too high, "
2258 "forcing read-only mode");
2264 if (!(sbi->s_mount_state & EXT4_VALID_FS))
2265 ext4_msg(sb, KERN_WARNING, "warning: mounting unchecked fs, "
2266 "running e2fsck is recommended");
2267 else if (sbi->s_mount_state & EXT4_ERROR_FS)
2268 ext4_msg(sb, KERN_WARNING,
2269 "warning: mounting fs with errors, "
2270 "running e2fsck is recommended");
2271 else if ((__s16) le16_to_cpu(es->s_max_mnt_count) > 0 &&
2272 le16_to_cpu(es->s_mnt_count) >=
2273 (unsigned short) (__s16) le16_to_cpu(es->s_max_mnt_count))
2274 ext4_msg(sb, KERN_WARNING,
2275 "warning: maximal mount count reached, "
2276 "running e2fsck is recommended");
2277 else if (le32_to_cpu(es->s_checkinterval) &&
2278 (ext4_get_tstamp(es, s_lastcheck) +
2279 le32_to_cpu(es->s_checkinterval) <= ktime_get_real_seconds()))
2280 ext4_msg(sb, KERN_WARNING,
2281 "warning: checktime reached, "
2282 "running e2fsck is recommended");
2283 if (!sbi->s_journal)
2284 es->s_state &= cpu_to_le16(~EXT4_VALID_FS);
2285 if (!(__s16) le16_to_cpu(es->s_max_mnt_count))
2286 es->s_max_mnt_count = cpu_to_le16(EXT4_DFL_MAX_MNT_COUNT);
2287 le16_add_cpu(&es->s_mnt_count, 1);
2288 ext4_update_tstamp(es, s_mtime);
2290 ext4_set_feature_journal_needs_recovery(sb);
2292 err = ext4_commit_super(sb, 1);
2294 if (test_opt(sb, DEBUG))
2295 printk(KERN_INFO "[EXT4 FS bs=%lu, gc=%u, "
2296 "bpg=%lu, ipg=%lu, mo=%04x, mo2=%04x]\n",
2298 sbi->s_groups_count,
2299 EXT4_BLOCKS_PER_GROUP(sb),
2300 EXT4_INODES_PER_GROUP(sb),
2301 sbi->s_mount_opt, sbi->s_mount_opt2);
2303 cleancache_init_fs(sb);
2307 int ext4_alloc_flex_bg_array(struct super_block *sb, ext4_group_t ngroup)
2309 struct ext4_sb_info *sbi = EXT4_SB(sb);
2310 struct flex_groups **old_groups, **new_groups;
2313 if (!sbi->s_log_groups_per_flex)
2316 size = ext4_flex_group(sbi, ngroup - 1) + 1;
2317 if (size <= sbi->s_flex_groups_allocated)
2320 new_groups = kvzalloc(roundup_pow_of_two(size *
2321 sizeof(*sbi->s_flex_groups)), GFP_KERNEL);
2323 ext4_msg(sb, KERN_ERR,
2324 "not enough memory for %d flex group pointers", size);
2327 for (i = sbi->s_flex_groups_allocated; i < size; i++) {
2328 new_groups[i] = kvzalloc(roundup_pow_of_two(
2329 sizeof(struct flex_groups)),
2331 if (!new_groups[i]) {
2332 for (j = sbi->s_flex_groups_allocated; j < i; j++)
2333 kvfree(new_groups[j]);
2335 ext4_msg(sb, KERN_ERR,
2336 "not enough memory for %d flex groups", size);
2341 old_groups = rcu_dereference(sbi->s_flex_groups);
2343 memcpy(new_groups, old_groups,
2344 (sbi->s_flex_groups_allocated *
2345 sizeof(struct flex_groups *)));
2347 rcu_assign_pointer(sbi->s_flex_groups, new_groups);
2348 sbi->s_flex_groups_allocated = size;
2350 ext4_kvfree_array_rcu(old_groups);
2354 static int ext4_fill_flex_info(struct super_block *sb)
2356 struct ext4_sb_info *sbi = EXT4_SB(sb);
2357 struct ext4_group_desc *gdp = NULL;
2358 struct flex_groups *fg;
2359 ext4_group_t flex_group;
2362 sbi->s_log_groups_per_flex = sbi->s_es->s_log_groups_per_flex;
2363 if (sbi->s_log_groups_per_flex < 1 || sbi->s_log_groups_per_flex > 31) {
2364 sbi->s_log_groups_per_flex = 0;
2368 err = ext4_alloc_flex_bg_array(sb, sbi->s_groups_count);
2372 for (i = 0; i < sbi->s_groups_count; i++) {
2373 gdp = ext4_get_group_desc(sb, i, NULL);
2375 flex_group = ext4_flex_group(sbi, i);
2376 fg = sbi_array_rcu_deref(sbi, s_flex_groups, flex_group);
2377 atomic_add(ext4_free_inodes_count(sb, gdp), &fg->free_inodes);
2378 atomic64_add(ext4_free_group_clusters(sb, gdp),
2379 &fg->free_clusters);
2380 atomic_add(ext4_used_dirs_count(sb, gdp), &fg->used_dirs);
2388 static __le16 ext4_group_desc_csum(struct super_block *sb, __u32 block_group,
2389 struct ext4_group_desc *gdp)
2391 int offset = offsetof(struct ext4_group_desc, bg_checksum);
2393 __le32 le_group = cpu_to_le32(block_group);
2394 struct ext4_sb_info *sbi = EXT4_SB(sb);
2396 if (ext4_has_metadata_csum(sbi->s_sb)) {
2397 /* Use new metadata_csum algorithm */
2399 __u16 dummy_csum = 0;
2401 csum32 = ext4_chksum(sbi, sbi->s_csum_seed, (__u8 *)&le_group,
2403 csum32 = ext4_chksum(sbi, csum32, (__u8 *)gdp, offset);
2404 csum32 = ext4_chksum(sbi, csum32, (__u8 *)&dummy_csum,
2405 sizeof(dummy_csum));
2406 offset += sizeof(dummy_csum);
2407 if (offset < sbi->s_desc_size)
2408 csum32 = ext4_chksum(sbi, csum32, (__u8 *)gdp + offset,
2409 sbi->s_desc_size - offset);
2411 crc = csum32 & 0xFFFF;
2415 /* old crc16 code */
2416 if (!ext4_has_feature_gdt_csum(sb))
2419 crc = crc16(~0, sbi->s_es->s_uuid, sizeof(sbi->s_es->s_uuid));
2420 crc = crc16(crc, (__u8 *)&le_group, sizeof(le_group));
2421 crc = crc16(crc, (__u8 *)gdp, offset);
2422 offset += sizeof(gdp->bg_checksum); /* skip checksum */
2423 /* for checksum of struct ext4_group_desc do the rest...*/
2424 if (ext4_has_feature_64bit(sb) &&
2425 offset < le16_to_cpu(sbi->s_es->s_desc_size))
2426 crc = crc16(crc, (__u8 *)gdp + offset,
2427 le16_to_cpu(sbi->s_es->s_desc_size) -
2431 return cpu_to_le16(crc);
2434 int ext4_group_desc_csum_verify(struct super_block *sb, __u32 block_group,
2435 struct ext4_group_desc *gdp)
2437 if (ext4_has_group_desc_csum(sb) &&
2438 (gdp->bg_checksum != ext4_group_desc_csum(sb, block_group, gdp)))
2444 void ext4_group_desc_csum_set(struct super_block *sb, __u32 block_group,
2445 struct ext4_group_desc *gdp)
2447 if (!ext4_has_group_desc_csum(sb))
2449 gdp->bg_checksum = ext4_group_desc_csum(sb, block_group, gdp);
2452 /* Called at mount-time, super-block is locked */
2453 static int ext4_check_descriptors(struct super_block *sb,
2454 ext4_fsblk_t sb_block,
2455 ext4_group_t *first_not_zeroed)
2457 struct ext4_sb_info *sbi = EXT4_SB(sb);
2458 ext4_fsblk_t first_block = le32_to_cpu(sbi->s_es->s_first_data_block);
2459 ext4_fsblk_t last_block;
2460 ext4_fsblk_t last_bg_block = sb_block + ext4_bg_num_gdb(sb, 0);
2461 ext4_fsblk_t block_bitmap;
2462 ext4_fsblk_t inode_bitmap;
2463 ext4_fsblk_t inode_table;
2464 int flexbg_flag = 0;
2465 ext4_group_t i, grp = sbi->s_groups_count;
2467 if (ext4_has_feature_flex_bg(sb))
2470 ext4_debug("Checking group descriptors");
2472 for (i = 0; i < sbi->s_groups_count; i++) {
2473 struct ext4_group_desc *gdp = ext4_get_group_desc(sb, i, NULL);
2475 if (i == sbi->s_groups_count - 1 || flexbg_flag)
2476 last_block = ext4_blocks_count(sbi->s_es) - 1;
2478 last_block = first_block +
2479 (EXT4_BLOCKS_PER_GROUP(sb) - 1);
2481 if ((grp == sbi->s_groups_count) &&
2482 !(gdp->bg_flags & cpu_to_le16(EXT4_BG_INODE_ZEROED)))
2485 block_bitmap = ext4_block_bitmap(sb, gdp);
2486 if (block_bitmap == sb_block) {
2487 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2488 "Block bitmap for group %u overlaps "
2493 if (block_bitmap >= sb_block + 1 &&
2494 block_bitmap <= last_bg_block) {
2495 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2496 "Block bitmap for group %u overlaps "
2497 "block group descriptors", i);
2501 if (block_bitmap < first_block || block_bitmap > last_block) {
2502 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2503 "Block bitmap for group %u not in group "
2504 "(block %llu)!", i, block_bitmap);
2507 inode_bitmap = ext4_inode_bitmap(sb, gdp);
2508 if (inode_bitmap == sb_block) {
2509 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2510 "Inode bitmap for group %u overlaps "
2515 if (inode_bitmap >= sb_block + 1 &&
2516 inode_bitmap <= last_bg_block) {
2517 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2518 "Inode bitmap for group %u overlaps "
2519 "block group descriptors", i);
2523 if (inode_bitmap < first_block || inode_bitmap > last_block) {
2524 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2525 "Inode bitmap for group %u not in group "
2526 "(block %llu)!", i, inode_bitmap);
2529 inode_table = ext4_inode_table(sb, gdp);
2530 if (inode_table == sb_block) {
2531 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2532 "Inode table for group %u overlaps "
2537 if (inode_table >= sb_block + 1 &&
2538 inode_table <= last_bg_block) {
2539 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2540 "Inode table for group %u overlaps "
2541 "block group descriptors", i);
2545 if (inode_table < first_block ||
2546 inode_table + sbi->s_itb_per_group - 1 > last_block) {
2547 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2548 "Inode table for group %u not in group "
2549 "(block %llu)!", i, inode_table);
2552 ext4_lock_group(sb, i);
2553 if (!ext4_group_desc_csum_verify(sb, i, gdp)) {
2554 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2555 "Checksum for group %u failed (%u!=%u)",
2556 i, le16_to_cpu(ext4_group_desc_csum(sb, i,
2557 gdp)), le16_to_cpu(gdp->bg_checksum));
2558 if (!sb_rdonly(sb)) {
2559 ext4_unlock_group(sb, i);
2563 ext4_unlock_group(sb, i);
2565 first_block += EXT4_BLOCKS_PER_GROUP(sb);
2567 if (NULL != first_not_zeroed)
2568 *first_not_zeroed = grp;
2572 /* ext4_orphan_cleanup() walks a singly-linked list of inodes (starting at
2573 * the superblock) which were deleted from all directories, but held open by
2574 * a process at the time of a crash. We walk the list and try to delete these
2575 * inodes at recovery time (only with a read-write filesystem).
2577 * In order to keep the orphan inode chain consistent during traversal (in
2578 * case of crash during recovery), we link each inode into the superblock
2579 * orphan list_head and handle it the same way as an inode deletion during
2580 * normal operation (which journals the operations for us).
2582 * We only do an iget() and an iput() on each inode, which is very safe if we
2583 * accidentally point at an in-use or already deleted inode. The worst that
2584 * can happen in this case is that we get a "bit already cleared" message from
2585 * ext4_free_inode(). The only reason we would point at a wrong inode is if
2586 * e2fsck was run on this filesystem, and it must have already done the orphan
2587 * inode cleanup for us, so we can safely abort without any further action.
2589 static void ext4_orphan_cleanup(struct super_block *sb,
2590 struct ext4_super_block *es)
2592 unsigned int s_flags = sb->s_flags;
2593 int ret, nr_orphans = 0, nr_truncates = 0;
2595 int quota_update = 0;
2598 if (!es->s_last_orphan) {
2599 jbd_debug(4, "no orphan inodes to clean up\n");
2603 if (bdev_read_only(sb->s_bdev)) {
2604 ext4_msg(sb, KERN_ERR, "write access "
2605 "unavailable, skipping orphan cleanup");
2609 /* Check if feature set would not allow a r/w mount */
2610 if (!ext4_feature_set_ok(sb, 0)) {
2611 ext4_msg(sb, KERN_INFO, "Skipping orphan cleanup due to "
2612 "unknown ROCOMPAT features");
2616 if (EXT4_SB(sb)->s_mount_state & EXT4_ERROR_FS) {
2617 /* don't clear list on RO mount w/ errors */
2618 if (es->s_last_orphan && !(s_flags & SB_RDONLY)) {
2619 ext4_msg(sb, KERN_INFO, "Errors on filesystem, "
2620 "clearing orphan list.\n");
2621 es->s_last_orphan = 0;
2623 jbd_debug(1, "Skipping orphan recovery on fs with errors.\n");
2627 if (s_flags & SB_RDONLY) {
2628 ext4_msg(sb, KERN_INFO, "orphan cleanup on readonly fs");
2629 sb->s_flags &= ~SB_RDONLY;
2633 * Turn on quotas which were not enabled for read-only mounts if
2634 * filesystem has quota feature, so that they are updated correctly.
2636 if (ext4_has_feature_quota(sb) && (s_flags & SB_RDONLY)) {
2637 int ret = ext4_enable_quotas(sb);
2642 ext4_msg(sb, KERN_ERR,
2643 "Cannot turn on quotas: error %d", ret);
2646 /* Turn on journaled quotas used for old sytle */
2647 for (i = 0; i < EXT4_MAXQUOTAS; i++) {
2648 if (EXT4_SB(sb)->s_qf_names[i]) {
2649 int ret = ext4_quota_on_mount(sb, i);
2654 ext4_msg(sb, KERN_ERR,
2655 "Cannot turn on journaled "
2656 "quota: type %d: error %d", i, ret);
2661 while (es->s_last_orphan) {
2662 struct inode *inode;
2665 * We may have encountered an error during cleanup; if
2666 * so, skip the rest.
2668 if (EXT4_SB(sb)->s_mount_state & EXT4_ERROR_FS) {
2669 jbd_debug(1, "Skipping orphan recovery on fs with errors.\n");
2670 es->s_last_orphan = 0;
2674 inode = ext4_orphan_get(sb, le32_to_cpu(es->s_last_orphan));
2675 if (IS_ERR(inode)) {
2676 es->s_last_orphan = 0;
2680 list_add(&EXT4_I(inode)->i_orphan, &EXT4_SB(sb)->s_orphan);
2681 dquot_initialize(inode);
2682 if (inode->i_nlink) {
2683 if (test_opt(sb, DEBUG))
2684 ext4_msg(sb, KERN_DEBUG,
2685 "%s: truncating inode %lu to %lld bytes",
2686 __func__, inode->i_ino, inode->i_size);
2687 jbd_debug(2, "truncating inode %lu to %lld bytes\n",
2688 inode->i_ino, inode->i_size);
2690 truncate_inode_pages(inode->i_mapping, inode->i_size);
2691 ret = ext4_truncate(inode);
2694 * We need to clean up the in-core orphan list
2695 * manually if ext4_truncate() failed to get a
2696 * transaction handle.
2698 ext4_orphan_del(NULL, inode);
2699 ext4_std_error(inode->i_sb, ret);
2701 inode_unlock(inode);
2704 if (test_opt(sb, DEBUG))
2705 ext4_msg(sb, KERN_DEBUG,
2706 "%s: deleting unreferenced inode %lu",
2707 __func__, inode->i_ino);
2708 jbd_debug(2, "deleting unreferenced inode %lu\n",
2712 iput(inode); /* The delete magic happens here! */
2715 #define PLURAL(x) (x), ((x) == 1) ? "" : "s"
2718 ext4_msg(sb, KERN_INFO, "%d orphan inode%s deleted",
2719 PLURAL(nr_orphans));
2721 ext4_msg(sb, KERN_INFO, "%d truncate%s cleaned up",
2722 PLURAL(nr_truncates));
2724 /* Turn off quotas if they were enabled for orphan cleanup */
2726 for (i = 0; i < EXT4_MAXQUOTAS; i++) {
2727 if (sb_dqopt(sb)->files[i])
2728 dquot_quota_off(sb, i);
2732 sb->s_flags = s_flags; /* Restore SB_RDONLY status */
2736 * Maximal extent format file size.
2737 * Resulting logical blkno at s_maxbytes must fit in our on-disk
2738 * extent format containers, within a sector_t, and within i_blocks
2739 * in the vfs. ext4 inode has 48 bits of i_block in fsblock units,
2740 * so that won't be a limiting factor.
2742 * However there is other limiting factor. We do store extents in the form
2743 * of starting block and length, hence the resulting length of the extent
2744 * covering maximum file size must fit into on-disk format containers as
2745 * well. Given that length is always by 1 unit bigger than max unit (because
2746 * we count 0 as well) we have to lower the s_maxbytes by one fs block.
2748 * Note, this does *not* consider any metadata overhead for vfs i_blocks.
2750 static loff_t ext4_max_size(int blkbits, int has_huge_files)
2753 loff_t upper_limit = MAX_LFS_FILESIZE;
2755 /* small i_blocks in vfs inode? */
2756 if (!has_huge_files || sizeof(blkcnt_t) < sizeof(u64)) {
2758 * CONFIG_LBDAF is not enabled implies the inode
2759 * i_block represent total blocks in 512 bytes
2760 * 32 == size of vfs inode i_blocks * 8
2762 upper_limit = (1LL << 32) - 1;
2764 /* total blocks in file system block size */
2765 upper_limit >>= (blkbits - 9);
2766 upper_limit <<= blkbits;
2770 * 32-bit extent-start container, ee_block. We lower the maxbytes
2771 * by one fs block, so ee_len can cover the extent of maximum file
2774 res = (1LL << 32) - 1;
2777 /* Sanity check against vm- & vfs- imposed limits */
2778 if (res > upper_limit)
2785 * Maximal bitmap file size. There is a direct, and {,double-,triple-}indirect
2786 * block limit, and also a limit of (2^48 - 1) 512-byte sectors in i_blocks.
2787 * We need to be 1 filesystem block less than the 2^48 sector limit.
2789 static loff_t ext4_max_bitmap_size(int bits, int has_huge_files)
2791 loff_t res = EXT4_NDIR_BLOCKS;
2794 /* This is calculated to be the largest file size for a dense, block
2795 * mapped file such that the file's total number of 512-byte sectors,
2796 * including data and all indirect blocks, does not exceed (2^48 - 1).
2798 * __u32 i_blocks_lo and _u16 i_blocks_high represent the total
2799 * number of 512-byte sectors of the file.
2802 if (!has_huge_files || sizeof(blkcnt_t) < sizeof(u64)) {
2804 * !has_huge_files or CONFIG_LBDAF not enabled implies that
2805 * the inode i_block field represents total file blocks in
2806 * 2^32 512-byte sectors == size of vfs inode i_blocks * 8
2808 upper_limit = (1LL << 32) - 1;
2810 /* total blocks in file system block size */
2811 upper_limit >>= (bits - 9);
2815 * We use 48 bit ext4_inode i_blocks
2816 * With EXT4_HUGE_FILE_FL set the i_blocks
2817 * represent total number of blocks in
2818 * file system block size
2820 upper_limit = (1LL << 48) - 1;
2824 /* indirect blocks */
2826 /* double indirect blocks */
2827 meta_blocks += 1 + (1LL << (bits-2));
2828 /* tripple indirect blocks */
2829 meta_blocks += 1 + (1LL << (bits-2)) + (1LL << (2*(bits-2)));
2831 upper_limit -= meta_blocks;
2832 upper_limit <<= bits;
2834 res += 1LL << (bits-2);
2835 res += 1LL << (2*(bits-2));
2836 res += 1LL << (3*(bits-2));
2838 if (res > upper_limit)
2841 if (res > MAX_LFS_FILESIZE)
2842 res = MAX_LFS_FILESIZE;
2847 static ext4_fsblk_t descriptor_loc(struct super_block *sb,
2848 ext4_fsblk_t logical_sb_block, int nr)
2850 struct ext4_sb_info *sbi = EXT4_SB(sb);
2851 ext4_group_t bg, first_meta_bg;
2854 first_meta_bg = le32_to_cpu(sbi->s_es->s_first_meta_bg);
2856 if (!ext4_has_feature_meta_bg(sb) || nr < first_meta_bg)
2857 return logical_sb_block + nr + 1;
2858 bg = sbi->s_desc_per_block * nr;
2859 if (ext4_bg_has_super(sb, bg))
2863 * If we have a meta_bg fs with 1k blocks, group 0's GDT is at
2864 * block 2, not 1. If s_first_data_block == 0 (bigalloc is enabled
2865 * on modern mke2fs or blksize > 1k on older mke2fs) then we must
2868 if (sb->s_blocksize == 1024 && nr == 0 &&
2869 le32_to_cpu(sbi->s_es->s_first_data_block) == 0)
2872 return (has_super + ext4_group_first_block_no(sb, bg));
2876 * ext4_get_stripe_size: Get the stripe size.
2877 * @sbi: In memory super block info
2879 * If we have specified it via mount option, then
2880 * use the mount option value. If the value specified at mount time is
2881 * greater than the blocks per group use the super block value.
2882 * If the super block value is greater than blocks per group return 0.
2883 * Allocator needs it be less than blocks per group.
2886 static unsigned long ext4_get_stripe_size(struct ext4_sb_info *sbi)
2888 unsigned long stride = le16_to_cpu(sbi->s_es->s_raid_stride);
2889 unsigned long stripe_width =
2890 le32_to_cpu(sbi->s_es->s_raid_stripe_width);
2893 if (sbi->s_stripe && sbi->s_stripe <= sbi->s_blocks_per_group)
2894 ret = sbi->s_stripe;
2895 else if (stripe_width && stripe_width <= sbi->s_blocks_per_group)
2897 else if (stride && stride <= sbi->s_blocks_per_group)
2903 * If the stripe width is 1, this makes no sense and
2904 * we set it to 0 to turn off stripe handling code.
2913 * Check whether this filesystem can be mounted based on
2914 * the features present and the RDONLY/RDWR mount requested.
2915 * Returns 1 if this filesystem can be mounted as requested,
2916 * 0 if it cannot be.
2918 static int ext4_feature_set_ok(struct super_block *sb, int readonly)
2920 if (ext4_has_unknown_ext4_incompat_features(sb)) {
2921 ext4_msg(sb, KERN_ERR,
2922 "Couldn't mount because of "
2923 "unsupported optional features (%x)",
2924 (le32_to_cpu(EXT4_SB(sb)->s_es->s_feature_incompat) &
2925 ~EXT4_FEATURE_INCOMPAT_SUPP));
2932 if (ext4_has_feature_readonly(sb)) {
2933 ext4_msg(sb, KERN_INFO, "filesystem is read-only");
2934 sb->s_flags |= SB_RDONLY;
2938 /* Check that feature set is OK for a read-write mount */
2939 if (ext4_has_unknown_ext4_ro_compat_features(sb)) {
2940 ext4_msg(sb, KERN_ERR, "couldn't mount RDWR because of "
2941 "unsupported optional features (%x)",
2942 (le32_to_cpu(EXT4_SB(sb)->s_es->s_feature_ro_compat) &
2943 ~EXT4_FEATURE_RO_COMPAT_SUPP));
2947 * Large file size enabled file system can only be mounted
2948 * read-write on 32-bit systems if kernel is built with CONFIG_LBDAF
2950 if (ext4_has_feature_huge_file(sb)) {
2951 if (sizeof(blkcnt_t) < sizeof(u64)) {
2952 ext4_msg(sb, KERN_ERR, "Filesystem with huge files "
2953 "cannot be mounted RDWR without "
2958 if (ext4_has_feature_bigalloc(sb) && !ext4_has_feature_extents(sb)) {
2959 ext4_msg(sb, KERN_ERR,
2960 "Can't support bigalloc feature without "
2961 "extents feature\n");
2965 #if !IS_ENABLED(CONFIG_QUOTA) || !IS_ENABLED(CONFIG_QFMT_V2)
2966 if (!readonly && (ext4_has_feature_quota(sb) ||
2967 ext4_has_feature_project(sb))) {
2968 ext4_msg(sb, KERN_ERR,
2969 "The kernel was not built with CONFIG_QUOTA and CONFIG_QFMT_V2");
2972 #endif /* CONFIG_QUOTA */
2977 * This function is called once a day if we have errors logged
2978 * on the file system
2980 static void print_daily_error_info(struct timer_list *t)
2982 struct ext4_sb_info *sbi = from_timer(sbi, t, s_err_report);
2983 struct super_block *sb = sbi->s_sb;
2984 struct ext4_super_block *es = sbi->s_es;
2986 if (es->s_error_count)
2987 /* fsck newer than v1.41.13 is needed to clean this condition. */
2988 ext4_msg(sb, KERN_NOTICE, "error count since last fsck: %u",
2989 le32_to_cpu(es->s_error_count));
2990 if (es->s_first_error_time) {
2991 printk(KERN_NOTICE "EXT4-fs (%s): initial error at time %llu: %.*s:%d",
2993 ext4_get_tstamp(es, s_first_error_time),
2994 (int) sizeof(es->s_first_error_func),
2995 es->s_first_error_func,
2996 le32_to_cpu(es->s_first_error_line));
2997 if (es->s_first_error_ino)
2998 printk(KERN_CONT ": inode %u",
2999 le32_to_cpu(es->s_first_error_ino));
3000 if (es->s_first_error_block)
3001 printk(KERN_CONT ": block %llu", (unsigned long long)
3002 le64_to_cpu(es->s_first_error_block));
3003 printk(KERN_CONT "\n");
3005 if (es->s_last_error_time) {
3006 printk(KERN_NOTICE "EXT4-fs (%s): last error at time %llu: %.*s:%d",
3008 ext4_get_tstamp(es, s_last_error_time),
3009 (int) sizeof(es->s_last_error_func),
3010 es->s_last_error_func,
3011 le32_to_cpu(es->s_last_error_line));
3012 if (es->s_last_error_ino)
3013 printk(KERN_CONT ": inode %u",
3014 le32_to_cpu(es->s_last_error_ino));
3015 if (es->s_last_error_block)
3016 printk(KERN_CONT ": block %llu", (unsigned long long)
3017 le64_to_cpu(es->s_last_error_block));
3018 printk(KERN_CONT "\n");
3020 mod_timer(&sbi->s_err_report, jiffies + 24*60*60*HZ); /* Once a day */
3023 /* Find next suitable group and run ext4_init_inode_table */
3024 static int ext4_run_li_request(struct ext4_li_request *elr)
3026 struct ext4_group_desc *gdp = NULL;
3027 ext4_group_t group, ngroups;
3028 struct super_block *sb;
3033 ngroups = EXT4_SB(sb)->s_groups_count;
3035 for (group = elr->lr_next_group; group < ngroups; group++) {
3036 gdp = ext4_get_group_desc(sb, group, NULL);
3042 if (!(gdp->bg_flags & cpu_to_le16(EXT4_BG_INODE_ZEROED)))
3046 if (group >= ngroups)
3050 start_time = ktime_get_real_ns();
3051 ret = ext4_init_inode_table(sb, group,
3052 elr->lr_timeout ? 0 : 1);
3053 if (elr->lr_timeout == 0) {
3054 elr->lr_timeout = nsecs_to_jiffies((ktime_get_real_ns() - start_time) *
3055 elr->lr_sbi->s_li_wait_mult);
3057 elr->lr_next_sched = jiffies + elr->lr_timeout;
3058 elr->lr_next_group = group + 1;
3064 * Remove lr_request from the list_request and free the
3065 * request structure. Should be called with li_list_mtx held
3067 static void ext4_remove_li_request(struct ext4_li_request *elr)
3069 struct ext4_sb_info *sbi;
3076 list_del(&elr->lr_request);
3077 sbi->s_li_request = NULL;
3081 static void ext4_unregister_li_request(struct super_block *sb)
3083 mutex_lock(&ext4_li_mtx);
3084 if (!ext4_li_info) {
3085 mutex_unlock(&ext4_li_mtx);
3089 mutex_lock(&ext4_li_info->li_list_mtx);
3090 ext4_remove_li_request(EXT4_SB(sb)->s_li_request);
3091 mutex_unlock(&ext4_li_info->li_list_mtx);
3092 mutex_unlock(&ext4_li_mtx);
3095 static struct task_struct *ext4_lazyinit_task;
3098 * This is the function where ext4lazyinit thread lives. It walks
3099 * through the request list searching for next scheduled filesystem.
3100 * When such a fs is found, run the lazy initialization request
3101 * (ext4_rn_li_request) and keep track of the time spend in this
3102 * function. Based on that time we compute next schedule time of
3103 * the request. When walking through the list is complete, compute
3104 * next waking time and put itself into sleep.
3106 static int ext4_lazyinit_thread(void *arg)
3108 struct ext4_lazy_init *eli = (struct ext4_lazy_init *)arg;
3109 struct list_head *pos, *n;
3110 struct ext4_li_request *elr;
3111 unsigned long next_wakeup, cur;
3113 BUG_ON(NULL == eli);
3117 next_wakeup = MAX_JIFFY_OFFSET;
3119 mutex_lock(&eli->li_list_mtx);
3120 if (list_empty(&eli->li_request_list)) {
3121 mutex_unlock(&eli->li_list_mtx);
3124 list_for_each_safe(pos, n, &eli->li_request_list) {
3127 elr = list_entry(pos, struct ext4_li_request,
3130 if (time_before(jiffies, elr->lr_next_sched)) {
3131 if (time_before(elr->lr_next_sched, next_wakeup))
3132 next_wakeup = elr->lr_next_sched;
3135 if (down_read_trylock(&elr->lr_super->s_umount)) {
3136 if (sb_start_write_trylock(elr->lr_super)) {
3139 * We hold sb->s_umount, sb can not
3140 * be removed from the list, it is
3141 * now safe to drop li_list_mtx
3143 mutex_unlock(&eli->li_list_mtx);
3144 err = ext4_run_li_request(elr);
3145 sb_end_write(elr->lr_super);
3146 mutex_lock(&eli->li_list_mtx);
3149 up_read((&elr->lr_super->s_umount));
3151 /* error, remove the lazy_init job */
3153 ext4_remove_li_request(elr);
3157 elr->lr_next_sched = jiffies +
3159 % (EXT4_DEF_LI_MAX_START_DELAY * HZ));
3161 if (time_before(elr->lr_next_sched, next_wakeup))
3162 next_wakeup = elr->lr_next_sched;
3164 mutex_unlock(&eli->li_list_mtx);
3169 if ((time_after_eq(cur, next_wakeup)) ||
3170 (MAX_JIFFY_OFFSET == next_wakeup)) {
3175 schedule_timeout_interruptible(next_wakeup - cur);
3177 if (kthread_should_stop()) {
3178 ext4_clear_request_list();
3185 * It looks like the request list is empty, but we need
3186 * to check it under the li_list_mtx lock, to prevent any
3187 * additions into it, and of course we should lock ext4_li_mtx
3188 * to atomically free the list and ext4_li_info, because at
3189 * this point another ext4 filesystem could be registering
3192 mutex_lock(&ext4_li_mtx);
3193 mutex_lock(&eli->li_list_mtx);
3194 if (!list_empty(&eli->li_request_list)) {
3195 mutex_unlock(&eli->li_list_mtx);
3196 mutex_unlock(&ext4_li_mtx);
3199 mutex_unlock(&eli->li_list_mtx);
3200 kfree(ext4_li_info);
3201 ext4_li_info = NULL;
3202 mutex_unlock(&ext4_li_mtx);
3207 static void ext4_clear_request_list(void)
3209 struct list_head *pos, *n;
3210 struct ext4_li_request *elr;
3212 mutex_lock(&ext4_li_info->li_list_mtx);
3213 list_for_each_safe(pos, n, &ext4_li_info->li_request_list) {
3214 elr = list_entry(pos, struct ext4_li_request,
3216 ext4_remove_li_request(elr);
3218 mutex_unlock(&ext4_li_info->li_list_mtx);
3221 static int ext4_run_lazyinit_thread(void)
3223 ext4_lazyinit_task = kthread_run(ext4_lazyinit_thread,
3224 ext4_li_info, "ext4lazyinit");
3225 if (IS_ERR(ext4_lazyinit_task)) {
3226 int err = PTR_ERR(ext4_lazyinit_task);
3227 ext4_clear_request_list();
3228 kfree(ext4_li_info);
3229 ext4_li_info = NULL;
3230 printk(KERN_CRIT "EXT4-fs: error %d creating inode table "
3231 "initialization thread\n",
3235 ext4_li_info->li_state |= EXT4_LAZYINIT_RUNNING;
3240 * Check whether it make sense to run itable init. thread or not.
3241 * If there is at least one uninitialized inode table, return
3242 * corresponding group number, else the loop goes through all
3243 * groups and return total number of groups.
3245 static ext4_group_t ext4_has_uninit_itable(struct super_block *sb)
3247 ext4_group_t group, ngroups = EXT4_SB(sb)->s_groups_count;
3248 struct ext4_group_desc *gdp = NULL;
3250 if (!ext4_has_group_desc_csum(sb))
3253 for (group = 0; group < ngroups; group++) {
3254 gdp = ext4_get_group_desc(sb, group, NULL);
3258 if (!(gdp->bg_flags & cpu_to_le16(EXT4_BG_INODE_ZEROED)))
3265 static int ext4_li_info_new(void)
3267 struct ext4_lazy_init *eli = NULL;
3269 eli = kzalloc(sizeof(*eli), GFP_KERNEL);
3273 INIT_LIST_HEAD(&eli->li_request_list);
3274 mutex_init(&eli->li_list_mtx);
3276 eli->li_state |= EXT4_LAZYINIT_QUIT;
3283 static struct ext4_li_request *ext4_li_request_new(struct super_block *sb,
3286 struct ext4_sb_info *sbi = EXT4_SB(sb);
3287 struct ext4_li_request *elr;
3289 elr = kzalloc(sizeof(*elr), GFP_KERNEL);
3295 elr->lr_next_group = start;
3298 * Randomize first schedule time of the request to
3299 * spread the inode table initialization requests
3302 elr->lr_next_sched = jiffies + (prandom_u32() %
3303 (EXT4_DEF_LI_MAX_START_DELAY * HZ));
3307 int ext4_register_li_request(struct super_block *sb,
3308 ext4_group_t first_not_zeroed)
3310 struct ext4_sb_info *sbi = EXT4_SB(sb);
3311 struct ext4_li_request *elr = NULL;
3312 ext4_group_t ngroups = sbi->s_groups_count;
3315 mutex_lock(&ext4_li_mtx);
3316 if (sbi->s_li_request != NULL) {
3318 * Reset timeout so it can be computed again, because
3319 * s_li_wait_mult might have changed.
3321 sbi->s_li_request->lr_timeout = 0;
3325 if (first_not_zeroed == ngroups || sb_rdonly(sb) ||
3326 !test_opt(sb, INIT_INODE_TABLE))
3329 elr = ext4_li_request_new(sb, first_not_zeroed);
3335 if (NULL == ext4_li_info) {
3336 ret = ext4_li_info_new();
3341 mutex_lock(&ext4_li_info->li_list_mtx);
3342 list_add(&elr->lr_request, &ext4_li_info->li_request_list);
3343 mutex_unlock(&ext4_li_info->li_list_mtx);
3345 sbi->s_li_request = elr;
3347 * set elr to NULL here since it has been inserted to
3348 * the request_list and the removal and free of it is
3349 * handled by ext4_clear_request_list from now on.
3353 if (!(ext4_li_info->li_state & EXT4_LAZYINIT_RUNNING)) {
3354 ret = ext4_run_lazyinit_thread();
3359 mutex_unlock(&ext4_li_mtx);
3366 * We do not need to lock anything since this is called on
3369 static void ext4_destroy_lazyinit_thread(void)
3372 * If thread exited earlier
3373 * there's nothing to be done.
3375 if (!ext4_li_info || !ext4_lazyinit_task)
3378 kthread_stop(ext4_lazyinit_task);
3381 static int set_journal_csum_feature_set(struct super_block *sb)
3384 int compat, incompat;
3385 struct ext4_sb_info *sbi = EXT4_SB(sb);
3387 if (ext4_has_metadata_csum(sb)) {
3388 /* journal checksum v3 */
3390 incompat = JBD2_FEATURE_INCOMPAT_CSUM_V3;
3392 /* journal checksum v1 */
3393 compat = JBD2_FEATURE_COMPAT_CHECKSUM;
3397 jbd2_journal_clear_features(sbi->s_journal,
3398 JBD2_FEATURE_COMPAT_CHECKSUM, 0,
3399 JBD2_FEATURE_INCOMPAT_CSUM_V3 |
3400 JBD2_FEATURE_INCOMPAT_CSUM_V2);
3401 if (test_opt(sb, JOURNAL_ASYNC_COMMIT)) {
3402 ret = jbd2_journal_set_features(sbi->s_journal,
3404 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT |
3406 } else if (test_opt(sb, JOURNAL_CHECKSUM)) {
3407 ret = jbd2_journal_set_features(sbi->s_journal,
3410 jbd2_journal_clear_features(sbi->s_journal, 0, 0,
3411 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT);
3413 jbd2_journal_clear_features(sbi->s_journal, 0, 0,
3414 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT);
3421 * Note: calculating the overhead so we can be compatible with
3422 * historical BSD practice is quite difficult in the face of
3423 * clusters/bigalloc. This is because multiple metadata blocks from
3424 * different block group can end up in the same allocation cluster.
3425 * Calculating the exact overhead in the face of clustered allocation
3426 * requires either O(all block bitmaps) in memory or O(number of block
3427 * groups**2) in time. We will still calculate the superblock for
3428 * older file systems --- and if we come across with a bigalloc file
3429 * system with zero in s_overhead_clusters the estimate will be close to
3430 * correct especially for very large cluster sizes --- but for newer
3431 * file systems, it's better to calculate this figure once at mkfs
3432 * time, and store it in the superblock. If the superblock value is
3433 * present (even for non-bigalloc file systems), we will use it.
3435 static int count_overhead(struct super_block *sb, ext4_group_t grp,
3438 struct ext4_sb_info *sbi = EXT4_SB(sb);
3439 struct ext4_group_desc *gdp;
3440 ext4_fsblk_t first_block, last_block, b;
3441 ext4_group_t i, ngroups = ext4_get_groups_count(sb);
3442 int s, j, count = 0;
3443 int has_super = ext4_bg_has_super(sb, grp);
3445 if (!ext4_has_feature_bigalloc(sb))
3446 return (has_super + ext4_bg_num_gdb(sb, grp) +
3447 (has_super ? le16_to_cpu(sbi->s_es->s_reserved_gdt_blocks) : 0) +
3448 sbi->s_itb_per_group + 2);
3450 first_block = le32_to_cpu(sbi->s_es->s_first_data_block) +
3451 (grp * EXT4_BLOCKS_PER_GROUP(sb));
3452 last_block = first_block + EXT4_BLOCKS_PER_GROUP(sb) - 1;
3453 for (i = 0; i < ngroups; i++) {
3454 gdp = ext4_get_group_desc(sb, i, NULL);
3455 b = ext4_block_bitmap(sb, gdp);
3456 if (b >= first_block && b <= last_block) {
3457 ext4_set_bit(EXT4_B2C(sbi, b - first_block), buf);
3460 b = ext4_inode_bitmap(sb, gdp);
3461 if (b >= first_block && b <= last_block) {
3462 ext4_set_bit(EXT4_B2C(sbi, b - first_block), buf);
3465 b = ext4_inode_table(sb, gdp);
3466 if (b >= first_block && b + sbi->s_itb_per_group <= last_block)
3467 for (j = 0; j < sbi->s_itb_per_group; j++, b++) {
3468 int c = EXT4_B2C(sbi, b - first_block);
3469 ext4_set_bit(c, buf);
3475 if (ext4_bg_has_super(sb, grp)) {
3476 ext4_set_bit(s++, buf);
3479 j = ext4_bg_num_gdb(sb, grp);
3480 if (s + j > EXT4_BLOCKS_PER_GROUP(sb)) {
3481 ext4_error(sb, "Invalid number of block group "
3482 "descriptor blocks: %d", j);
3483 j = EXT4_BLOCKS_PER_GROUP(sb) - s;
3487 ext4_set_bit(EXT4_B2C(sbi, s++), buf);
3491 return EXT4_CLUSTERS_PER_GROUP(sb) -
3492 ext4_count_free(buf, EXT4_CLUSTERS_PER_GROUP(sb) / 8);
3496 * Compute the overhead and stash it in sbi->s_overhead
3498 int ext4_calculate_overhead(struct super_block *sb)
3500 struct ext4_sb_info *sbi = EXT4_SB(sb);
3501 struct ext4_super_block *es = sbi->s_es;
3502 struct inode *j_inode;
3503 unsigned int j_blocks, j_inum = le32_to_cpu(es->s_journal_inum);
3504 ext4_group_t i, ngroups = ext4_get_groups_count(sb);
3505 ext4_fsblk_t overhead = 0;
3506 char *buf = (char *) get_zeroed_page(GFP_NOFS);
3512 * Compute the overhead (FS structures). This is constant
3513 * for a given filesystem unless the number of block groups
3514 * changes so we cache the previous value until it does.
3518 * All of the blocks before first_data_block are overhead
3520 overhead = EXT4_B2C(sbi, le32_to_cpu(es->s_first_data_block));
3523 * Add the overhead found in each block group
3525 for (i = 0; i < ngroups; i++) {
3528 blks = count_overhead(sb, i, buf);
3531 memset(buf, 0, PAGE_SIZE);
3536 * Add the internal journal blocks whether the journal has been
3539 if (sbi->s_journal && !sbi->journal_bdev)
3540 overhead += EXT4_NUM_B2C(sbi, sbi->s_journal->j_maxlen);
3541 else if (ext4_has_feature_journal(sb) && !sbi->s_journal && j_inum) {
3542 /* j_inum for internal journal is non-zero */
3543 j_inode = ext4_get_journal_inode(sb, j_inum);
3545 j_blocks = j_inode->i_size >> sb->s_blocksize_bits;
3546 overhead += EXT4_NUM_B2C(sbi, j_blocks);
3549 ext4_msg(sb, KERN_ERR, "can't get journal size");
3552 sbi->s_overhead = overhead;
3554 free_page((unsigned long) buf);
3558 static void ext4_set_resv_clusters(struct super_block *sb)
3560 ext4_fsblk_t resv_clusters;
3561 struct ext4_sb_info *sbi = EXT4_SB(sb);
3564 * There's no need to reserve anything when we aren't using extents.
3565 * The space estimates are exact, there are no unwritten extents,
3566 * hole punching doesn't need new metadata... This is needed especially
3567 * to keep ext2/3 backward compatibility.
3569 if (!ext4_has_feature_extents(sb))
3572 * By default we reserve 2% or 4096 clusters, whichever is smaller.
3573 * This should cover the situations where we can not afford to run
3574 * out of space like for example punch hole, or converting
3575 * unwritten extents in delalloc path. In most cases such
3576 * allocation would require 1, or 2 blocks, higher numbers are
3579 resv_clusters = (ext4_blocks_count(sbi->s_es) >>
3580 sbi->s_cluster_bits);
3582 do_div(resv_clusters, 50);
3583 resv_clusters = min_t(ext4_fsblk_t, resv_clusters, 4096);
3585 atomic64_set(&sbi->s_resv_clusters, resv_clusters);
3588 static int ext4_fill_super(struct super_block *sb, void *data, int silent)
3590 struct dax_device *dax_dev = fs_dax_get_by_bdev(sb->s_bdev);
3591 char *orig_data = kstrdup(data, GFP_KERNEL);
3592 struct buffer_head *bh, **group_desc;
3593 struct ext4_super_block *es = NULL;
3594 struct ext4_sb_info *sbi = kzalloc(sizeof(*sbi), GFP_KERNEL);
3595 struct flex_groups **flex_groups;
3597 ext4_fsblk_t sb_block = get_sb_block(&data);
3598 ext4_fsblk_t logical_sb_block;
3599 unsigned long offset = 0;
3600 unsigned long journal_devnum = 0;
3601 unsigned long def_mount_opts;
3605 int blocksize, clustersize;
3606 unsigned int db_count;
3608 int needs_recovery, has_huge_files, has_bigalloc;
3611 unsigned int journal_ioprio = DEFAULT_JOURNAL_IOPRIO;
3612 ext4_group_t first_not_zeroed;
3614 if ((data && !orig_data) || !sbi)
3617 sbi->s_daxdev = dax_dev;
3618 sbi->s_blockgroup_lock =
3619 kzalloc(sizeof(struct blockgroup_lock), GFP_KERNEL);
3620 if (!sbi->s_blockgroup_lock)
3623 sb->s_fs_info = sbi;
3625 sbi->s_inode_readahead_blks = EXT4_DEF_INODE_READAHEAD_BLKS;
3626 sbi->s_sb_block = sb_block;
3627 if (sb->s_bdev->bd_part)
3628 sbi->s_sectors_written_start =
3629 part_stat_read(sb->s_bdev->bd_part, sectors[STAT_WRITE]);
3631 /* Cleanup superblock name */
3632 strreplace(sb->s_id, '/', '!');
3634 /* -EINVAL is default */
3636 blocksize = sb_min_blocksize(sb, EXT4_MIN_BLOCK_SIZE);
3638 ext4_msg(sb, KERN_ERR, "unable to set blocksize");
3643 * The ext4 superblock will not be buffer aligned for other than 1kB
3644 * block sizes. We need to calculate the offset from buffer start.
3646 if (blocksize != EXT4_MIN_BLOCK_SIZE) {
3647 logical_sb_block = sb_block * EXT4_MIN_BLOCK_SIZE;
3648 offset = do_div(logical_sb_block, blocksize);
3650 logical_sb_block = sb_block;
3653 if (!(bh = sb_bread_unmovable(sb, logical_sb_block))) {
3654 ext4_msg(sb, KERN_ERR, "unable to read superblock");
3658 * Note: s_es must be initialized as soon as possible because
3659 * some ext4 macro-instructions depend on its value
3661 es = (struct ext4_super_block *) (bh->b_data + offset);
3663 sb->s_magic = le16_to_cpu(es->s_magic);
3664 if (sb->s_magic != EXT4_SUPER_MAGIC)
3666 sbi->s_kbytes_written = le64_to_cpu(es->s_kbytes_written);
3668 /* Warn if metadata_csum and gdt_csum are both set. */
3669 if (ext4_has_feature_metadata_csum(sb) &&
3670 ext4_has_feature_gdt_csum(sb))
3671 ext4_warning(sb, "metadata_csum and uninit_bg are "
3672 "redundant flags; please run fsck.");
3674 /* Check for a known checksum algorithm */
3675 if (!ext4_verify_csum_type(sb, es)) {
3676 ext4_msg(sb, KERN_ERR, "VFS: Found ext4 filesystem with "
3677 "unknown checksum algorithm.");
3682 /* Load the checksum driver */
3683 sbi->s_chksum_driver = crypto_alloc_shash("crc32c", 0, 0);
3684 if (IS_ERR(sbi->s_chksum_driver)) {
3685 ext4_msg(sb, KERN_ERR, "Cannot load crc32c driver.");
3686 ret = PTR_ERR(sbi->s_chksum_driver);
3687 sbi->s_chksum_driver = NULL;
3691 /* Check superblock checksum */
3692 if (!ext4_superblock_csum_verify(sb, es)) {
3693 ext4_msg(sb, KERN_ERR, "VFS: Found ext4 filesystem with "
3694 "invalid superblock checksum. Run e2fsck?");
3700 /* Precompute checksum seed for all metadata */
3701 if (ext4_has_feature_csum_seed(sb))
3702 sbi->s_csum_seed = le32_to_cpu(es->s_checksum_seed);
3703 else if (ext4_has_metadata_csum(sb) || ext4_has_feature_ea_inode(sb))
3704 sbi->s_csum_seed = ext4_chksum(sbi, ~0, es->s_uuid,
3705 sizeof(es->s_uuid));
3707 /* Set defaults before we parse the mount options */
3708 def_mount_opts = le32_to_cpu(es->s_default_mount_opts);
3709 set_opt(sb, INIT_INODE_TABLE);
3710 if (def_mount_opts & EXT4_DEFM_DEBUG)
3712 if (def_mount_opts & EXT4_DEFM_BSDGROUPS)
3714 if (def_mount_opts & EXT4_DEFM_UID16)
3715 set_opt(sb, NO_UID32);
3716 /* xattr user namespace & acls are now defaulted on */
3717 set_opt(sb, XATTR_USER);
3718 #ifdef CONFIG_EXT4_FS_POSIX_ACL
3719 set_opt(sb, POSIX_ACL);
3721 /* don't forget to enable journal_csum when metadata_csum is enabled. */
3722 if (ext4_has_metadata_csum(sb))
3723 set_opt(sb, JOURNAL_CHECKSUM);
3725 if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_DATA)
3726 set_opt(sb, JOURNAL_DATA);
3727 else if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_ORDERED)
3728 set_opt(sb, ORDERED_DATA);
3729 else if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_WBACK)
3730 set_opt(sb, WRITEBACK_DATA);
3732 if (le16_to_cpu(sbi->s_es->s_errors) == EXT4_ERRORS_PANIC)
3733 set_opt(sb, ERRORS_PANIC);
3734 else if (le16_to_cpu(sbi->s_es->s_errors) == EXT4_ERRORS_CONTINUE)
3735 set_opt(sb, ERRORS_CONT);
3737 set_opt(sb, ERRORS_RO);
3738 /* block_validity enabled by default; disable with noblock_validity */
3739 set_opt(sb, BLOCK_VALIDITY);
3740 if (def_mount_opts & EXT4_DEFM_DISCARD)
3741 set_opt(sb, DISCARD);
3743 sbi->s_resuid = make_kuid(&init_user_ns, le16_to_cpu(es->s_def_resuid));
3744 sbi->s_resgid = make_kgid(&init_user_ns, le16_to_cpu(es->s_def_resgid));
3745 sbi->s_commit_interval = JBD2_DEFAULT_MAX_COMMIT_AGE * HZ;
3746 sbi->s_min_batch_time = EXT4_DEF_MIN_BATCH_TIME;
3747 sbi->s_max_batch_time = EXT4_DEF_MAX_BATCH_TIME;
3749 if ((def_mount_opts & EXT4_DEFM_NOBARRIER) == 0)
3750 set_opt(sb, BARRIER);
3753 * enable delayed allocation by default
3754 * Use -o nodelalloc to turn it off
3756 if (!IS_EXT3_SB(sb) && !IS_EXT2_SB(sb) &&
3757 ((def_mount_opts & EXT4_DEFM_NODELALLOC) == 0))
3758 set_opt(sb, DELALLOC);
3761 * set default s_li_wait_mult for lazyinit, for the case there is
3762 * no mount option specified.
3764 sbi->s_li_wait_mult = EXT4_DEF_LI_WAIT_MULT;
3766 blocksize = BLOCK_SIZE << le32_to_cpu(es->s_log_block_size);
3767 if (blocksize < EXT4_MIN_BLOCK_SIZE ||
3768 blocksize > EXT4_MAX_BLOCK_SIZE) {
3769 ext4_msg(sb, KERN_ERR,
3770 "Unsupported filesystem blocksize %d (%d log_block_size)",
3771 blocksize, le32_to_cpu(es->s_log_block_size));
3775 if (le32_to_cpu(es->s_rev_level) == EXT4_GOOD_OLD_REV) {
3776 sbi->s_inode_size = EXT4_GOOD_OLD_INODE_SIZE;
3777 sbi->s_first_ino = EXT4_GOOD_OLD_FIRST_INO;
3779 sbi->s_inode_size = le16_to_cpu(es->s_inode_size);
3780 sbi->s_first_ino = le32_to_cpu(es->s_first_ino);
3781 if (sbi->s_first_ino < EXT4_GOOD_OLD_FIRST_INO) {
3782 ext4_msg(sb, KERN_ERR, "invalid first ino: %u",
3786 if ((sbi->s_inode_size < EXT4_GOOD_OLD_INODE_SIZE) ||
3787 (!is_power_of_2(sbi->s_inode_size)) ||
3788 (sbi->s_inode_size > blocksize)) {
3789 ext4_msg(sb, KERN_ERR,
3790 "unsupported inode size: %d",
3792 ext4_msg(sb, KERN_ERR, "blocksize: %d", blocksize);
3796 * i_atime_extra is the last extra field available for
3797 * [acm]times in struct ext4_inode. Checking for that
3798 * field should suffice to ensure we have extra space
3801 if (sbi->s_inode_size >= offsetof(struct ext4_inode, i_atime_extra) +
3802 sizeof(((struct ext4_inode *)0)->i_atime_extra)) {
3803 sb->s_time_gran = 1;
3805 sb->s_time_gran = NSEC_PER_SEC;
3808 if (sbi->s_inode_size > EXT4_GOOD_OLD_INODE_SIZE) {
3809 sbi->s_want_extra_isize = sizeof(struct ext4_inode) -
3810 EXT4_GOOD_OLD_INODE_SIZE;
3811 if (ext4_has_feature_extra_isize(sb)) {
3812 unsigned v, max = (sbi->s_inode_size -
3813 EXT4_GOOD_OLD_INODE_SIZE);
3815 v = le16_to_cpu(es->s_want_extra_isize);
3817 ext4_msg(sb, KERN_ERR,
3818 "bad s_want_extra_isize: %d", v);
3821 if (sbi->s_want_extra_isize < v)
3822 sbi->s_want_extra_isize = v;
3824 v = le16_to_cpu(es->s_min_extra_isize);
3826 ext4_msg(sb, KERN_ERR,
3827 "bad s_min_extra_isize: %d", v);
3830 if (sbi->s_want_extra_isize < v)
3831 sbi->s_want_extra_isize = v;
3835 if (sbi->s_es->s_mount_opts[0]) {
3836 char *s_mount_opts = kstrndup(sbi->s_es->s_mount_opts,
3837 sizeof(sbi->s_es->s_mount_opts),
3841 if (!parse_options(s_mount_opts, sb, &journal_devnum,
3842 &journal_ioprio, 0)) {
3843 ext4_msg(sb, KERN_WARNING,
3844 "failed to parse options in superblock: %s",
3847 kfree(s_mount_opts);
3849 sbi->s_def_mount_opt = sbi->s_mount_opt;
3850 if (!parse_options((char *) data, sb, &journal_devnum,
3851 &journal_ioprio, 0))
3854 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA) {
3855 printk_once(KERN_WARNING "EXT4-fs: Warning: mounting "
3856 "with data=journal disables delayed "
3857 "allocation and O_DIRECT support!\n");
3858 if (test_opt2(sb, EXPLICIT_DELALLOC)) {
3859 ext4_msg(sb, KERN_ERR, "can't mount with "
3860 "both data=journal and delalloc");
3863 if (test_opt(sb, DIOREAD_NOLOCK)) {
3864 ext4_msg(sb, KERN_ERR, "can't mount with "
3865 "both data=journal and dioread_nolock");
3868 if (test_opt(sb, DAX)) {
3869 ext4_msg(sb, KERN_ERR, "can't mount with "
3870 "both data=journal and dax");
3873 if (ext4_has_feature_encrypt(sb)) {
3874 ext4_msg(sb, KERN_WARNING,
3875 "encrypted files will use data=ordered "
3876 "instead of data journaling mode");
3878 if (test_opt(sb, DELALLOC))
3879 clear_opt(sb, DELALLOC);
3881 sb->s_iflags |= SB_I_CGROUPWB;
3884 sb->s_flags = (sb->s_flags & ~SB_POSIXACL) |
3885 (test_opt(sb, POSIX_ACL) ? SB_POSIXACL : 0);
3887 if (le32_to_cpu(es->s_rev_level) == EXT4_GOOD_OLD_REV &&
3888 (ext4_has_compat_features(sb) ||
3889 ext4_has_ro_compat_features(sb) ||
3890 ext4_has_incompat_features(sb)))
3891 ext4_msg(sb, KERN_WARNING,
3892 "feature flags set on rev 0 fs, "
3893 "running e2fsck is recommended");
3895 if (es->s_creator_os == cpu_to_le32(EXT4_OS_HURD)) {
3896 set_opt2(sb, HURD_COMPAT);
3897 if (ext4_has_feature_64bit(sb)) {
3898 ext4_msg(sb, KERN_ERR,
3899 "The Hurd can't support 64-bit file systems");
3904 * ea_inode feature uses l_i_version field which is not
3905 * available in HURD_COMPAT mode.
3907 if (ext4_has_feature_ea_inode(sb)) {
3908 ext4_msg(sb, KERN_ERR,
3909 "ea_inode feature is not supported for Hurd");
3914 if (IS_EXT2_SB(sb)) {
3915 if (ext2_feature_set_ok(sb))
3916 ext4_msg(sb, KERN_INFO, "mounting ext2 file system "
3917 "using the ext4 subsystem");
3920 * If we're probing be silent, if this looks like
3921 * it's actually an ext[34] filesystem.
3923 if (silent && ext4_feature_set_ok(sb, sb_rdonly(sb)))
3925 ext4_msg(sb, KERN_ERR, "couldn't mount as ext2 due "
3926 "to feature incompatibilities");
3931 if (IS_EXT3_SB(sb)) {
3932 if (ext3_feature_set_ok(sb))
3933 ext4_msg(sb, KERN_INFO, "mounting ext3 file system "
3934 "using the ext4 subsystem");
3937 * If we're probing be silent, if this looks like
3938 * it's actually an ext4 filesystem.
3940 if (silent && ext4_feature_set_ok(sb, sb_rdonly(sb)))
3942 ext4_msg(sb, KERN_ERR, "couldn't mount as ext3 due "
3943 "to feature incompatibilities");
3949 * Check feature flags regardless of the revision level, since we
3950 * previously didn't change the revision level when setting the flags,
3951 * so there is a chance incompat flags are set on a rev 0 filesystem.
3953 if (!ext4_feature_set_ok(sb, (sb_rdonly(sb))))
3956 if (le32_to_cpu(es->s_log_block_size) >
3957 (EXT4_MAX_BLOCK_LOG_SIZE - EXT4_MIN_BLOCK_LOG_SIZE)) {
3958 ext4_msg(sb, KERN_ERR,
3959 "Invalid log block size: %u",
3960 le32_to_cpu(es->s_log_block_size));
3963 if (le32_to_cpu(es->s_log_cluster_size) >
3964 (EXT4_MAX_CLUSTER_LOG_SIZE - EXT4_MIN_BLOCK_LOG_SIZE)) {
3965 ext4_msg(sb, KERN_ERR,
3966 "Invalid log cluster size: %u",
3967 le32_to_cpu(es->s_log_cluster_size));
3971 if (le16_to_cpu(sbi->s_es->s_reserved_gdt_blocks) > (blocksize / 4)) {
3972 ext4_msg(sb, KERN_ERR,
3973 "Number of reserved GDT blocks insanely large: %d",
3974 le16_to_cpu(sbi->s_es->s_reserved_gdt_blocks));
3978 if (sbi->s_mount_opt & EXT4_MOUNT_DAX) {
3979 if (ext4_has_feature_inline_data(sb)) {
3980 ext4_msg(sb, KERN_ERR, "Cannot use DAX on a filesystem"
3981 " that may contain inline data");
3982 sbi->s_mount_opt &= ~EXT4_MOUNT_DAX;
3984 if (!bdev_dax_supported(sb->s_bdev, blocksize)) {
3985 ext4_msg(sb, KERN_ERR,
3986 "DAX unsupported by block device. Turning off DAX.");
3987 sbi->s_mount_opt &= ~EXT4_MOUNT_DAX;
3991 if (ext4_has_feature_encrypt(sb) && es->s_encryption_level) {
3992 ext4_msg(sb, KERN_ERR, "Unsupported encryption level %d",
3993 es->s_encryption_level);
3997 if (sb->s_blocksize != blocksize) {
3998 /* Validate the filesystem blocksize */
3999 if (!sb_set_blocksize(sb, blocksize)) {
4000 ext4_msg(sb, KERN_ERR, "bad block size %d",
4006 logical_sb_block = sb_block * EXT4_MIN_BLOCK_SIZE;
4007 offset = do_div(logical_sb_block, blocksize);
4008 bh = sb_bread_unmovable(sb, logical_sb_block);
4010 ext4_msg(sb, KERN_ERR,
4011 "Can't read superblock on 2nd try");
4014 es = (struct ext4_super_block *)(bh->b_data + offset);
4016 if (es->s_magic != cpu_to_le16(EXT4_SUPER_MAGIC)) {
4017 ext4_msg(sb, KERN_ERR,
4018 "Magic mismatch, very weird!");
4023 has_huge_files = ext4_has_feature_huge_file(sb);
4024 sbi->s_bitmap_maxbytes = ext4_max_bitmap_size(sb->s_blocksize_bits,
4026 sb->s_maxbytes = ext4_max_size(sb->s_blocksize_bits, has_huge_files);
4028 sbi->s_desc_size = le16_to_cpu(es->s_desc_size);
4029 if (ext4_has_feature_64bit(sb)) {
4030 if (sbi->s_desc_size < EXT4_MIN_DESC_SIZE_64BIT ||
4031 sbi->s_desc_size > EXT4_MAX_DESC_SIZE ||
4032 !is_power_of_2(sbi->s_desc_size)) {
4033 ext4_msg(sb, KERN_ERR,
4034 "unsupported descriptor size %lu",
4039 sbi->s_desc_size = EXT4_MIN_DESC_SIZE;
4041 sbi->s_blocks_per_group = le32_to_cpu(es->s_blocks_per_group);
4042 sbi->s_inodes_per_group = le32_to_cpu(es->s_inodes_per_group);
4044 sbi->s_inodes_per_block = blocksize / EXT4_INODE_SIZE(sb);
4045 if (sbi->s_inodes_per_block == 0)
4047 if (sbi->s_inodes_per_group < sbi->s_inodes_per_block ||
4048 sbi->s_inodes_per_group > blocksize * 8) {
4049 ext4_msg(sb, KERN_ERR, "invalid inodes per group: %lu\n",
4050 sbi->s_inodes_per_group);
4053 sbi->s_itb_per_group = sbi->s_inodes_per_group /
4054 sbi->s_inodes_per_block;
4055 sbi->s_desc_per_block = blocksize / EXT4_DESC_SIZE(sb);
4057 sbi->s_mount_state = le16_to_cpu(es->s_state);
4058 sbi->s_addr_per_block_bits = ilog2(EXT4_ADDR_PER_BLOCK(sb));
4059 sbi->s_desc_per_block_bits = ilog2(EXT4_DESC_PER_BLOCK(sb));
4061 for (i = 0; i < 4; i++)
4062 sbi->s_hash_seed[i] = le32_to_cpu(es->s_hash_seed[i]);
4063 sbi->s_def_hash_version = es->s_def_hash_version;
4064 if (ext4_has_feature_dir_index(sb)) {
4065 i = le32_to_cpu(es->s_flags);
4066 if (i & EXT2_FLAGS_UNSIGNED_HASH)
4067 sbi->s_hash_unsigned = 3;
4068 else if ((i & EXT2_FLAGS_SIGNED_HASH) == 0) {
4069 #ifdef __CHAR_UNSIGNED__
4072 cpu_to_le32(EXT2_FLAGS_UNSIGNED_HASH);
4073 sbi->s_hash_unsigned = 3;
4077 cpu_to_le32(EXT2_FLAGS_SIGNED_HASH);
4082 /* Handle clustersize */
4083 clustersize = BLOCK_SIZE << le32_to_cpu(es->s_log_cluster_size);
4084 has_bigalloc = ext4_has_feature_bigalloc(sb);
4086 if (clustersize < blocksize) {
4087 ext4_msg(sb, KERN_ERR,
4088 "cluster size (%d) smaller than "
4089 "block size (%d)", clustersize, blocksize);
4092 sbi->s_cluster_bits = le32_to_cpu(es->s_log_cluster_size) -
4093 le32_to_cpu(es->s_log_block_size);
4094 sbi->s_clusters_per_group =
4095 le32_to_cpu(es->s_clusters_per_group);
4096 if (sbi->s_clusters_per_group > blocksize * 8) {
4097 ext4_msg(sb, KERN_ERR,
4098 "#clusters per group too big: %lu",
4099 sbi->s_clusters_per_group);
4102 if (sbi->s_blocks_per_group !=
4103 (sbi->s_clusters_per_group * (clustersize / blocksize))) {
4104 ext4_msg(sb, KERN_ERR, "blocks per group (%lu) and "
4105 "clusters per group (%lu) inconsistent",
4106 sbi->s_blocks_per_group,
4107 sbi->s_clusters_per_group);
4111 if (clustersize != blocksize) {
4112 ext4_msg(sb, KERN_ERR,
4113 "fragment/cluster size (%d) != "
4114 "block size (%d)", clustersize, blocksize);
4117 if (sbi->s_blocks_per_group > blocksize * 8) {
4118 ext4_msg(sb, KERN_ERR,
4119 "#blocks per group too big: %lu",
4120 sbi->s_blocks_per_group);
4123 sbi->s_clusters_per_group = sbi->s_blocks_per_group;
4124 sbi->s_cluster_bits = 0;
4126 sbi->s_cluster_ratio = clustersize / blocksize;
4128 /* Do we have standard group size of clustersize * 8 blocks ? */
4129 if (sbi->s_blocks_per_group == clustersize << 3)
4130 set_opt2(sb, STD_GROUP_SIZE);
4133 * Test whether we have more sectors than will fit in sector_t,
4134 * and whether the max offset is addressable by the page cache.
4136 err = generic_check_addressable(sb->s_blocksize_bits,
4137 ext4_blocks_count(es));
4139 ext4_msg(sb, KERN_ERR, "filesystem"
4140 " too large to mount safely on this system");
4141 if (sizeof(sector_t) < 8)
4142 ext4_msg(sb, KERN_WARNING, "CONFIG_LBDAF not enabled");
4146 if (EXT4_BLOCKS_PER_GROUP(sb) == 0)
4149 /* check blocks count against device size */
4150 blocks_count = sb->s_bdev->bd_inode->i_size >> sb->s_blocksize_bits;
4151 if (blocks_count && ext4_blocks_count(es) > blocks_count) {
4152 ext4_msg(sb, KERN_WARNING, "bad geometry: block count %llu "
4153 "exceeds size of device (%llu blocks)",
4154 ext4_blocks_count(es), blocks_count);
4159 * It makes no sense for the first data block to be beyond the end
4160 * of the filesystem.
4162 if (le32_to_cpu(es->s_first_data_block) >= ext4_blocks_count(es)) {
4163 ext4_msg(sb, KERN_WARNING, "bad geometry: first data "
4164 "block %u is beyond end of filesystem (%llu)",
4165 le32_to_cpu(es->s_first_data_block),
4166 ext4_blocks_count(es));
4169 if ((es->s_first_data_block == 0) && (es->s_log_block_size == 0) &&
4170 (sbi->s_cluster_ratio == 1)) {
4171 ext4_msg(sb, KERN_WARNING, "bad geometry: first data "
4172 "block is 0 with a 1k block and cluster size");
4176 blocks_count = (ext4_blocks_count(es) -
4177 le32_to_cpu(es->s_first_data_block) +
4178 EXT4_BLOCKS_PER_GROUP(sb) - 1);
4179 do_div(blocks_count, EXT4_BLOCKS_PER_GROUP(sb));
4180 if (blocks_count > ((uint64_t)1<<32) - EXT4_DESC_PER_BLOCK(sb)) {
4181 ext4_msg(sb, KERN_WARNING, "groups count too large: %llu "
4182 "(block count %llu, first data block %u, "
4183 "blocks per group %lu)", blocks_count,
4184 ext4_blocks_count(es),
4185 le32_to_cpu(es->s_first_data_block),
4186 EXT4_BLOCKS_PER_GROUP(sb));
4189 sbi->s_groups_count = blocks_count;
4190 sbi->s_blockfile_groups = min_t(ext4_group_t, sbi->s_groups_count,
4191 (EXT4_MAX_BLOCK_FILE_PHYS / EXT4_BLOCKS_PER_GROUP(sb)));
4192 if (((u64)sbi->s_groups_count * sbi->s_inodes_per_group) !=
4193 le32_to_cpu(es->s_inodes_count)) {
4194 ext4_msg(sb, KERN_ERR, "inodes count not valid: %u vs %llu",
4195 le32_to_cpu(es->s_inodes_count),
4196 ((u64)sbi->s_groups_count * sbi->s_inodes_per_group));
4200 db_count = (sbi->s_groups_count + EXT4_DESC_PER_BLOCK(sb) - 1) /
4201 EXT4_DESC_PER_BLOCK(sb);
4202 if (ext4_has_feature_meta_bg(sb)) {
4203 if (le32_to_cpu(es->s_first_meta_bg) > db_count) {
4204 ext4_msg(sb, KERN_WARNING,
4205 "first meta block group too large: %u "
4206 "(group descriptor block count %u)",
4207 le32_to_cpu(es->s_first_meta_bg), db_count);
4211 rcu_assign_pointer(sbi->s_group_desc,
4212 kvmalloc_array(db_count,
4213 sizeof(struct buffer_head *),
4215 if (sbi->s_group_desc == NULL) {
4216 ext4_msg(sb, KERN_ERR, "not enough memory");
4221 bgl_lock_init(sbi->s_blockgroup_lock);
4223 /* Pre-read the descriptors into the buffer cache */
4224 for (i = 0; i < db_count; i++) {
4225 block = descriptor_loc(sb, logical_sb_block, i);
4226 sb_breadahead_unmovable(sb, block);
4229 for (i = 0; i < db_count; i++) {
4230 struct buffer_head *bh;
4232 block = descriptor_loc(sb, logical_sb_block, i);
4233 bh = sb_bread_unmovable(sb, block);
4235 ext4_msg(sb, KERN_ERR,
4236 "can't read group descriptor %d", i);
4241 rcu_dereference(sbi->s_group_desc)[i] = bh;
4244 sbi->s_gdb_count = db_count;
4245 if (!ext4_check_descriptors(sb, logical_sb_block, &first_not_zeroed)) {
4246 ext4_msg(sb, KERN_ERR, "group descriptors corrupted!");
4247 ret = -EFSCORRUPTED;
4251 timer_setup(&sbi->s_err_report, print_daily_error_info, 0);
4253 /* Register extent status tree shrinker */
4254 if (ext4_es_register_shrinker(sbi))
4257 sbi->s_stripe = ext4_get_stripe_size(sbi);
4258 sbi->s_extent_max_zeroout_kb = 32;
4261 * set up enough so that it can read an inode
4263 sb->s_op = &ext4_sops;
4264 sb->s_export_op = &ext4_export_ops;
4265 sb->s_xattr = ext4_xattr_handlers;
4266 #ifdef CONFIG_EXT4_FS_ENCRYPTION
4267 sb->s_cop = &ext4_cryptops;
4270 sb->dq_op = &ext4_quota_operations;
4271 if (ext4_has_feature_quota(sb))
4272 sb->s_qcop = &dquot_quotactl_sysfile_ops;
4274 sb->s_qcop = &ext4_qctl_operations;
4275 sb->s_quota_types = QTYPE_MASK_USR | QTYPE_MASK_GRP | QTYPE_MASK_PRJ;
4277 memcpy(&sb->s_uuid, es->s_uuid, sizeof(es->s_uuid));
4279 INIT_LIST_HEAD(&sbi->s_orphan); /* unlinked but open files */
4280 mutex_init(&sbi->s_orphan_lock);
4284 needs_recovery = (es->s_last_orphan != 0 ||
4285 ext4_has_feature_journal_needs_recovery(sb));
4287 if (ext4_has_feature_mmp(sb) && !sb_rdonly(sb))
4288 if (ext4_multi_mount_protect(sb, le64_to_cpu(es->s_mmp_block)))
4289 goto failed_mount3a;
4292 * The first inode we look at is the journal inode. Don't try
4293 * root first: it may be modified in the journal!
4295 if (!test_opt(sb, NOLOAD) && ext4_has_feature_journal(sb)) {
4296 err = ext4_load_journal(sb, es, journal_devnum);
4298 goto failed_mount3a;
4299 } else if (test_opt(sb, NOLOAD) && !sb_rdonly(sb) &&
4300 ext4_has_feature_journal_needs_recovery(sb)) {
4301 ext4_msg(sb, KERN_ERR, "required journal recovery "
4302 "suppressed and not mounted read-only");
4303 goto failed_mount_wq;
4305 /* Nojournal mode, all journal mount options are illegal */
4306 if (test_opt2(sb, EXPLICIT_JOURNAL_CHECKSUM)) {
4307 ext4_msg(sb, KERN_ERR, "can't mount with "
4308 "journal_checksum, fs mounted w/o journal");
4309 goto failed_mount_wq;
4311 if (test_opt(sb, JOURNAL_ASYNC_COMMIT)) {
4312 ext4_msg(sb, KERN_ERR, "can't mount with "
4313 "journal_async_commit, fs mounted w/o journal");
4314 goto failed_mount_wq;
4316 if (sbi->s_commit_interval != JBD2_DEFAULT_MAX_COMMIT_AGE*HZ) {
4317 ext4_msg(sb, KERN_ERR, "can't mount with "
4318 "commit=%lu, fs mounted w/o journal",
4319 sbi->s_commit_interval / HZ);
4320 goto failed_mount_wq;
4322 if (EXT4_MOUNT_DATA_FLAGS &
4323 (sbi->s_mount_opt ^ sbi->s_def_mount_opt)) {
4324 ext4_msg(sb, KERN_ERR, "can't mount with "
4325 "data=, fs mounted w/o journal");
4326 goto failed_mount_wq;
4328 sbi->s_def_mount_opt &= ~EXT4_MOUNT_JOURNAL_CHECKSUM;
4329 clear_opt(sb, JOURNAL_CHECKSUM);
4330 clear_opt(sb, DATA_FLAGS);
4331 sbi->s_journal = NULL;
4336 if (ext4_has_feature_64bit(sb) &&
4337 !jbd2_journal_set_features(EXT4_SB(sb)->s_journal, 0, 0,
4338 JBD2_FEATURE_INCOMPAT_64BIT)) {
4339 ext4_msg(sb, KERN_ERR, "Failed to set 64-bit journal feature");
4340 goto failed_mount_wq;
4343 if (!set_journal_csum_feature_set(sb)) {
4344 ext4_msg(sb, KERN_ERR, "Failed to set journal checksum "
4346 goto failed_mount_wq;
4349 /* We have now updated the journal if required, so we can
4350 * validate the data journaling mode. */
4351 switch (test_opt(sb, DATA_FLAGS)) {
4353 /* No mode set, assume a default based on the journal
4354 * capabilities: ORDERED_DATA if the journal can
4355 * cope, else JOURNAL_DATA
4357 if (jbd2_journal_check_available_features
4358 (sbi->s_journal, 0, 0, JBD2_FEATURE_INCOMPAT_REVOKE)) {
4359 set_opt(sb, ORDERED_DATA);
4360 sbi->s_def_mount_opt |= EXT4_MOUNT_ORDERED_DATA;
4362 set_opt(sb, JOURNAL_DATA);
4363 sbi->s_def_mount_opt |= EXT4_MOUNT_JOURNAL_DATA;
4367 case EXT4_MOUNT_ORDERED_DATA:
4368 case EXT4_MOUNT_WRITEBACK_DATA:
4369 if (!jbd2_journal_check_available_features
4370 (sbi->s_journal, 0, 0, JBD2_FEATURE_INCOMPAT_REVOKE)) {
4371 ext4_msg(sb, KERN_ERR, "Journal does not support "
4372 "requested data journaling mode");
4373 goto failed_mount_wq;
4379 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA &&
4380 test_opt(sb, JOURNAL_ASYNC_COMMIT)) {
4381 ext4_msg(sb, KERN_ERR, "can't mount with "
4382 "journal_async_commit in data=ordered mode");
4383 goto failed_mount_wq;
4386 set_task_ioprio(sbi->s_journal->j_task, journal_ioprio);
4388 sbi->s_journal->j_commit_callback = ext4_journal_commit_callback;
4391 if (!test_opt(sb, NO_MBCACHE)) {
4392 sbi->s_ea_block_cache = ext4_xattr_create_cache();
4393 if (!sbi->s_ea_block_cache) {
4394 ext4_msg(sb, KERN_ERR,
4395 "Failed to create ea_block_cache");
4396 goto failed_mount_wq;
4399 if (ext4_has_feature_ea_inode(sb)) {
4400 sbi->s_ea_inode_cache = ext4_xattr_create_cache();
4401 if (!sbi->s_ea_inode_cache) {
4402 ext4_msg(sb, KERN_ERR,
4403 "Failed to create ea_inode_cache");
4404 goto failed_mount_wq;
4409 if ((DUMMY_ENCRYPTION_ENABLED(sbi) || ext4_has_feature_encrypt(sb)) &&
4410 (blocksize != PAGE_SIZE)) {
4411 ext4_msg(sb, KERN_ERR,
4412 "Unsupported blocksize for fs encryption");
4413 goto failed_mount_wq;
4416 if (DUMMY_ENCRYPTION_ENABLED(sbi) && !sb_rdonly(sb) &&
4417 !ext4_has_feature_encrypt(sb)) {
4418 ext4_set_feature_encrypt(sb);
4419 ext4_commit_super(sb, 1);
4423 * Get the # of file system overhead blocks from the
4424 * superblock if present.
4426 sbi->s_overhead = le32_to_cpu(es->s_overhead_clusters);
4427 /* ignore the precalculated value if it is ridiculous */
4428 if (sbi->s_overhead > ext4_blocks_count(es))
4429 sbi->s_overhead = 0;
4431 * If the bigalloc feature is not enabled recalculating the
4432 * overhead doesn't take long, so we might as well just redo
4433 * it to make sure we are using the correct value.
4435 if (!ext4_has_feature_bigalloc(sb))
4436 sbi->s_overhead = 0;
4437 if (sbi->s_overhead == 0) {
4438 err = ext4_calculate_overhead(sb);
4440 goto failed_mount_wq;
4444 * The maximum number of concurrent works can be high and
4445 * concurrency isn't really necessary. Limit it to 1.
4447 EXT4_SB(sb)->rsv_conversion_wq =
4448 alloc_workqueue("ext4-rsv-conversion", WQ_MEM_RECLAIM | WQ_UNBOUND, 1);
4449 if (!EXT4_SB(sb)->rsv_conversion_wq) {
4450 printk(KERN_ERR "EXT4-fs: failed to create workqueue\n");
4456 * The jbd2_journal_load will have done any necessary log recovery,
4457 * so we can safely mount the rest of the filesystem now.
4460 root = ext4_iget(sb, EXT4_ROOT_INO, EXT4_IGET_SPECIAL);
4462 ext4_msg(sb, KERN_ERR, "get root inode failed");
4463 ret = PTR_ERR(root);
4467 if (!S_ISDIR(root->i_mode) || !root->i_blocks || !root->i_size) {
4468 ext4_msg(sb, KERN_ERR, "corrupt root inode, run e2fsck");
4472 sb->s_root = d_make_root(root);
4474 ext4_msg(sb, KERN_ERR, "get root dentry failed");
4479 ret = ext4_setup_super(sb, es, sb_rdonly(sb));
4480 if (ret == -EROFS) {
4481 sb->s_flags |= SB_RDONLY;
4484 goto failed_mount4a;
4486 ext4_set_resv_clusters(sb);
4488 if (test_opt(sb, BLOCK_VALIDITY)) {
4489 err = ext4_setup_system_zone(sb);
4491 ext4_msg(sb, KERN_ERR, "failed to initialize system "
4493 goto failed_mount4a;
4498 err = ext4_mb_init(sb);
4500 ext4_msg(sb, KERN_ERR, "failed to initialize mballoc (%d)",
4505 block = ext4_count_free_clusters(sb);
4506 ext4_free_blocks_count_set(sbi->s_es,
4507 EXT4_C2B(sbi, block));
4508 ext4_superblock_csum_set(sb);
4509 err = percpu_counter_init(&sbi->s_freeclusters_counter, block,
4512 unsigned long freei = ext4_count_free_inodes(sb);
4513 sbi->s_es->s_free_inodes_count = cpu_to_le32(freei);
4514 ext4_superblock_csum_set(sb);
4515 err = percpu_counter_init(&sbi->s_freeinodes_counter, freei,
4519 err = percpu_counter_init(&sbi->s_dirs_counter,
4520 ext4_count_dirs(sb), GFP_KERNEL);
4522 err = percpu_counter_init(&sbi->s_dirtyclusters_counter, 0,
4525 err = percpu_init_rwsem(&sbi->s_writepages_rwsem);
4528 ext4_msg(sb, KERN_ERR, "insufficient memory");
4532 if (ext4_has_feature_flex_bg(sb))
4533 if (!ext4_fill_flex_info(sb)) {
4534 ext4_msg(sb, KERN_ERR,
4535 "unable to initialize "
4536 "flex_bg meta info!");
4541 err = ext4_register_li_request(sb, first_not_zeroed);
4545 err = ext4_register_sysfs(sb);
4550 /* Enable quota usage during mount. */
4551 if (ext4_has_feature_quota(sb) && !sb_rdonly(sb)) {
4552 err = ext4_enable_quotas(sb);
4556 #endif /* CONFIG_QUOTA */
4558 EXT4_SB(sb)->s_mount_state |= EXT4_ORPHAN_FS;
4559 ext4_orphan_cleanup(sb, es);
4560 EXT4_SB(sb)->s_mount_state &= ~EXT4_ORPHAN_FS;
4561 if (needs_recovery) {
4562 ext4_msg(sb, KERN_INFO, "recovery complete");
4563 err = ext4_mark_recovery_complete(sb, es);
4567 if (EXT4_SB(sb)->s_journal) {
4568 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA)
4569 descr = " journalled data mode";
4570 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA)
4571 descr = " ordered data mode";
4573 descr = " writeback data mode";
4575 descr = "out journal";
4577 if (test_opt(sb, DISCARD)) {
4578 struct request_queue *q = bdev_get_queue(sb->s_bdev);
4579 if (!blk_queue_discard(q))
4580 ext4_msg(sb, KERN_WARNING,
4581 "mounting with \"discard\" option, but "
4582 "the device does not support discard");
4585 if (___ratelimit(&ext4_mount_msg_ratelimit, "EXT4-fs mount"))
4586 ext4_msg(sb, KERN_INFO, "mounted filesystem with%s. "
4587 "Opts: %.*s%s%s", descr,
4588 (int) sizeof(sbi->s_es->s_mount_opts),
4589 sbi->s_es->s_mount_opts,
4590 *sbi->s_es->s_mount_opts ? "; " : "", orig_data);
4592 if (es->s_error_count)
4593 mod_timer(&sbi->s_err_report, jiffies + 300*HZ); /* 5 minutes */
4595 /* Enable message ratelimiting. Default is 10 messages per 5 secs. */
4596 ratelimit_state_init(&sbi->s_err_ratelimit_state, 5 * HZ, 10);
4597 ratelimit_state_init(&sbi->s_warning_ratelimit_state, 5 * HZ, 10);
4598 ratelimit_state_init(&sbi->s_msg_ratelimit_state, 5 * HZ, 10);
4605 ext4_msg(sb, KERN_ERR, "VFS: Can't find ext4 filesystem");
4609 ext4_unregister_sysfs(sb);
4610 kobject_put(&sbi->s_kobj);
4612 ext4_unregister_li_request(sb);
4614 ext4_mb_release(sb);
4616 flex_groups = rcu_dereference(sbi->s_flex_groups);
4618 for (i = 0; i < sbi->s_flex_groups_allocated; i++)
4619 kvfree(flex_groups[i]);
4620 kvfree(flex_groups);
4623 percpu_counter_destroy(&sbi->s_freeclusters_counter);
4624 percpu_counter_destroy(&sbi->s_freeinodes_counter);
4625 percpu_counter_destroy(&sbi->s_dirs_counter);
4626 percpu_counter_destroy(&sbi->s_dirtyclusters_counter);
4627 percpu_free_rwsem(&sbi->s_writepages_rwsem);
4629 ext4_ext_release(sb);
4630 ext4_release_system_zone(sb);
4635 ext4_msg(sb, KERN_ERR, "mount failed");
4636 if (EXT4_SB(sb)->rsv_conversion_wq)
4637 destroy_workqueue(EXT4_SB(sb)->rsv_conversion_wq);
4639 if (sbi->s_ea_inode_cache) {
4640 ext4_xattr_destroy_cache(sbi->s_ea_inode_cache);
4641 sbi->s_ea_inode_cache = NULL;
4643 if (sbi->s_ea_block_cache) {
4644 ext4_xattr_destroy_cache(sbi->s_ea_block_cache);
4645 sbi->s_ea_block_cache = NULL;
4647 if (sbi->s_journal) {
4648 jbd2_journal_destroy(sbi->s_journal);
4649 sbi->s_journal = NULL;
4652 ext4_es_unregister_shrinker(sbi);
4654 del_timer_sync(&sbi->s_err_report);
4656 kthread_stop(sbi->s_mmp_tsk);
4659 group_desc = rcu_dereference(sbi->s_group_desc);
4660 for (i = 0; i < db_count; i++)
4661 brelse(group_desc[i]);
4665 if (sbi->s_chksum_driver)
4666 crypto_free_shash(sbi->s_chksum_driver);
4668 for (i = 0; i < EXT4_MAXQUOTAS; i++)
4669 kfree(sbi->s_qf_names[i]);
4671 ext4_blkdev_remove(sbi);
4674 sb->s_fs_info = NULL;
4675 kfree(sbi->s_blockgroup_lock);
4679 fs_put_dax(dax_dev);
4680 return err ? err : ret;
4684 * Setup any per-fs journal parameters now. We'll do this both on
4685 * initial mount, once the journal has been initialised but before we've
4686 * done any recovery; and again on any subsequent remount.
4688 static void ext4_init_journal_params(struct super_block *sb, journal_t *journal)
4690 struct ext4_sb_info *sbi = EXT4_SB(sb);
4692 journal->j_commit_interval = sbi->s_commit_interval;
4693 journal->j_min_batch_time = sbi->s_min_batch_time;
4694 journal->j_max_batch_time = sbi->s_max_batch_time;
4696 write_lock(&journal->j_state_lock);
4697 if (test_opt(sb, BARRIER))
4698 journal->j_flags |= JBD2_BARRIER;
4700 journal->j_flags &= ~JBD2_BARRIER;
4701 if (test_opt(sb, DATA_ERR_ABORT))
4702 journal->j_flags |= JBD2_ABORT_ON_SYNCDATA_ERR;
4704 journal->j_flags &= ~JBD2_ABORT_ON_SYNCDATA_ERR;
4705 write_unlock(&journal->j_state_lock);
4708 static struct inode *ext4_get_journal_inode(struct super_block *sb,
4709 unsigned int journal_inum)
4711 struct inode *journal_inode;
4714 * Test for the existence of a valid inode on disk. Bad things
4715 * happen if we iget() an unused inode, as the subsequent iput()
4716 * will try to delete it.
4718 journal_inode = ext4_iget(sb, journal_inum, EXT4_IGET_SPECIAL);
4719 if (IS_ERR(journal_inode)) {
4720 ext4_msg(sb, KERN_ERR, "no journal found");
4723 if (!journal_inode->i_nlink) {
4724 make_bad_inode(journal_inode);
4725 iput(journal_inode);
4726 ext4_msg(sb, KERN_ERR, "journal inode is deleted");
4730 jbd_debug(2, "Journal inode found at %p: %lld bytes\n",
4731 journal_inode, journal_inode->i_size);
4732 if (!S_ISREG(journal_inode->i_mode)) {
4733 ext4_msg(sb, KERN_ERR, "invalid journal inode");
4734 iput(journal_inode);
4737 return journal_inode;
4740 static journal_t *ext4_get_journal(struct super_block *sb,
4741 unsigned int journal_inum)
4743 struct inode *journal_inode;
4746 if (WARN_ON_ONCE(!ext4_has_feature_journal(sb)))
4749 journal_inode = ext4_get_journal_inode(sb, journal_inum);
4753 journal = jbd2_journal_init_inode(journal_inode);
4755 ext4_msg(sb, KERN_ERR, "Could not load journal inode");
4756 iput(journal_inode);
4759 journal->j_private = sb;
4760 ext4_init_journal_params(sb, journal);
4764 static journal_t *ext4_get_dev_journal(struct super_block *sb,
4767 struct buffer_head *bh;
4771 int hblock, blocksize;
4772 ext4_fsblk_t sb_block;
4773 unsigned long offset;
4774 struct ext4_super_block *es;
4775 struct block_device *bdev;
4777 if (WARN_ON_ONCE(!ext4_has_feature_journal(sb)))
4780 bdev = ext4_blkdev_get(j_dev, sb);
4784 blocksize = sb->s_blocksize;
4785 hblock = bdev_logical_block_size(bdev);
4786 if (blocksize < hblock) {
4787 ext4_msg(sb, KERN_ERR,
4788 "blocksize too small for journal device");
4792 sb_block = EXT4_MIN_BLOCK_SIZE / blocksize;
4793 offset = EXT4_MIN_BLOCK_SIZE % blocksize;
4794 set_blocksize(bdev, blocksize);
4795 if (!(bh = __bread(bdev, sb_block, blocksize))) {
4796 ext4_msg(sb, KERN_ERR, "couldn't read superblock of "
4797 "external journal");
4801 es = (struct ext4_super_block *) (bh->b_data + offset);
4802 if ((le16_to_cpu(es->s_magic) != EXT4_SUPER_MAGIC) ||
4803 !(le32_to_cpu(es->s_feature_incompat) &
4804 EXT4_FEATURE_INCOMPAT_JOURNAL_DEV)) {
4805 ext4_msg(sb, KERN_ERR, "external journal has "
4811 if ((le32_to_cpu(es->s_feature_ro_compat) &
4812 EXT4_FEATURE_RO_COMPAT_METADATA_CSUM) &&
4813 es->s_checksum != ext4_superblock_csum(sb, es)) {
4814 ext4_msg(sb, KERN_ERR, "external journal has "
4815 "corrupt superblock");
4820 if (memcmp(EXT4_SB(sb)->s_es->s_journal_uuid, es->s_uuid, 16)) {
4821 ext4_msg(sb, KERN_ERR, "journal UUID does not match");
4826 len = ext4_blocks_count(es);
4827 start = sb_block + 1;
4828 brelse(bh); /* we're done with the superblock */
4830 journal = jbd2_journal_init_dev(bdev, sb->s_bdev,
4831 start, len, blocksize);
4833 ext4_msg(sb, KERN_ERR, "failed to create device journal");
4836 journal->j_private = sb;
4837 ll_rw_block(REQ_OP_READ, REQ_META | REQ_PRIO, 1, &journal->j_sb_buffer);
4838 wait_on_buffer(journal->j_sb_buffer);
4839 if (!buffer_uptodate(journal->j_sb_buffer)) {
4840 ext4_msg(sb, KERN_ERR, "I/O error on journal device");
4843 if (be32_to_cpu(journal->j_superblock->s_nr_users) != 1) {
4844 ext4_msg(sb, KERN_ERR, "External journal has more than one "
4845 "user (unsupported) - %d",
4846 be32_to_cpu(journal->j_superblock->s_nr_users));
4849 EXT4_SB(sb)->journal_bdev = bdev;
4850 ext4_init_journal_params(sb, journal);
4854 jbd2_journal_destroy(journal);
4856 ext4_blkdev_put(bdev);
4860 static int ext4_load_journal(struct super_block *sb,
4861 struct ext4_super_block *es,
4862 unsigned long journal_devnum)
4865 unsigned int journal_inum = le32_to_cpu(es->s_journal_inum);
4868 int really_read_only;
4871 if (WARN_ON_ONCE(!ext4_has_feature_journal(sb)))
4872 return -EFSCORRUPTED;
4874 if (journal_devnum &&
4875 journal_devnum != le32_to_cpu(es->s_journal_dev)) {
4876 ext4_msg(sb, KERN_INFO, "external journal device major/minor "
4877 "numbers have changed");
4878 journal_dev = new_decode_dev(journal_devnum);
4880 journal_dev = new_decode_dev(le32_to_cpu(es->s_journal_dev));
4882 if (journal_inum && journal_dev) {
4883 ext4_msg(sb, KERN_ERR,
4884 "filesystem has both journal inode and journal device!");
4889 journal = ext4_get_journal(sb, journal_inum);
4893 journal = ext4_get_dev_journal(sb, journal_dev);
4898 journal_dev_ro = bdev_read_only(journal->j_dev);
4899 really_read_only = bdev_read_only(sb->s_bdev) | journal_dev_ro;
4901 if (journal_dev_ro && !sb_rdonly(sb)) {
4902 ext4_msg(sb, KERN_ERR,
4903 "journal device read-only, try mounting with '-o ro'");
4909 * Are we loading a blank journal or performing recovery after a
4910 * crash? For recovery, we need to check in advance whether we
4911 * can get read-write access to the device.
4913 if (ext4_has_feature_journal_needs_recovery(sb)) {
4914 if (sb_rdonly(sb)) {
4915 ext4_msg(sb, KERN_INFO, "INFO: recovery "
4916 "required on readonly filesystem");
4917 if (really_read_only) {
4918 ext4_msg(sb, KERN_ERR, "write access "
4919 "unavailable, cannot proceed "
4920 "(try mounting with noload)");
4924 ext4_msg(sb, KERN_INFO, "write access will "
4925 "be enabled during recovery");
4929 if (!(journal->j_flags & JBD2_BARRIER))
4930 ext4_msg(sb, KERN_INFO, "barriers disabled");
4932 if (!ext4_has_feature_journal_needs_recovery(sb))
4933 err = jbd2_journal_wipe(journal, !really_read_only);
4935 char *save = kmalloc(EXT4_S_ERR_LEN, GFP_KERNEL);
4937 memcpy(save, ((char *) es) +
4938 EXT4_S_ERR_START, EXT4_S_ERR_LEN);
4939 err = jbd2_journal_load(journal);
4941 memcpy(((char *) es) + EXT4_S_ERR_START,
4942 save, EXT4_S_ERR_LEN);
4947 ext4_msg(sb, KERN_ERR, "error loading journal");
4951 EXT4_SB(sb)->s_journal = journal;
4952 err = ext4_clear_journal_err(sb, es);
4954 EXT4_SB(sb)->s_journal = NULL;
4955 jbd2_journal_destroy(journal);
4959 if (!really_read_only && journal_devnum &&
4960 journal_devnum != le32_to_cpu(es->s_journal_dev)) {
4961 es->s_journal_dev = cpu_to_le32(journal_devnum);
4963 /* Make sure we flush the recovery flag to disk. */
4964 ext4_commit_super(sb, 1);
4970 jbd2_journal_destroy(journal);
4974 static int ext4_commit_super(struct super_block *sb, int sync)
4976 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
4977 struct buffer_head *sbh = EXT4_SB(sb)->s_sbh;
4982 if (block_device_ejected(sb))
4986 * If the file system is mounted read-only, don't update the
4987 * superblock write time. This avoids updating the superblock
4988 * write time when we are mounting the root file system
4989 * read/only but we need to replay the journal; at that point,
4990 * for people who are east of GMT and who make their clock
4991 * tick in localtime for Windows bug-for-bug compatibility,
4992 * the clock is set in the future, and this will cause e2fsck
4993 * to complain and force a full file system check.
4995 if (!(sb->s_flags & SB_RDONLY))
4996 ext4_update_tstamp(es, s_wtime);
4997 if (sb->s_bdev->bd_part)
4998 es->s_kbytes_written =
4999 cpu_to_le64(EXT4_SB(sb)->s_kbytes_written +
5000 ((part_stat_read(sb->s_bdev->bd_part,
5001 sectors[STAT_WRITE]) -
5002 EXT4_SB(sb)->s_sectors_written_start) >> 1));
5004 es->s_kbytes_written =
5005 cpu_to_le64(EXT4_SB(sb)->s_kbytes_written);
5006 if (percpu_counter_initialized(&EXT4_SB(sb)->s_freeclusters_counter))
5007 ext4_free_blocks_count_set(es,
5008 EXT4_C2B(EXT4_SB(sb), percpu_counter_sum_positive(
5009 &EXT4_SB(sb)->s_freeclusters_counter)));
5010 if (percpu_counter_initialized(&EXT4_SB(sb)->s_freeinodes_counter))
5011 es->s_free_inodes_count =
5012 cpu_to_le32(percpu_counter_sum_positive(
5013 &EXT4_SB(sb)->s_freeinodes_counter));
5014 BUFFER_TRACE(sbh, "marking dirty");
5015 ext4_superblock_csum_set(sb);
5018 if (buffer_write_io_error(sbh) || !buffer_uptodate(sbh)) {
5020 * Oh, dear. A previous attempt to write the
5021 * superblock failed. This could happen because the
5022 * USB device was yanked out. Or it could happen to
5023 * be a transient write error and maybe the block will
5024 * be remapped. Nothing we can do but to retry the
5025 * write and hope for the best.
5027 ext4_msg(sb, KERN_ERR, "previous I/O error to "
5028 "superblock detected");
5029 clear_buffer_write_io_error(sbh);
5030 set_buffer_uptodate(sbh);
5032 mark_buffer_dirty(sbh);
5035 error = __sync_dirty_buffer(sbh,
5036 REQ_SYNC | (test_opt(sb, BARRIER) ? REQ_FUA : 0));
5037 if (buffer_write_io_error(sbh)) {
5038 ext4_msg(sb, KERN_ERR, "I/O error while writing "
5040 clear_buffer_write_io_error(sbh);
5041 set_buffer_uptodate(sbh);
5048 * Have we just finished recovery? If so, and if we are mounting (or
5049 * remounting) the filesystem readonly, then we will end up with a
5050 * consistent fs on disk. Record that fact.
5052 static int ext4_mark_recovery_complete(struct super_block *sb,
5053 struct ext4_super_block *es)
5056 journal_t *journal = EXT4_SB(sb)->s_journal;
5058 if (!ext4_has_feature_journal(sb)) {
5059 if (journal != NULL) {
5060 ext4_error(sb, "Journal got removed while the fs was "
5062 return -EFSCORRUPTED;
5066 jbd2_journal_lock_updates(journal);
5067 err = jbd2_journal_flush(journal);
5071 if (ext4_has_feature_journal_needs_recovery(sb) && sb_rdonly(sb)) {
5072 ext4_clear_feature_journal_needs_recovery(sb);
5073 ext4_commit_super(sb, 1);
5076 jbd2_journal_unlock_updates(journal);
5081 * If we are mounting (or read-write remounting) a filesystem whose journal
5082 * has recorded an error from a previous lifetime, move that error to the
5083 * main filesystem now.
5085 static int ext4_clear_journal_err(struct super_block *sb,
5086 struct ext4_super_block *es)
5092 if (!ext4_has_feature_journal(sb)) {
5093 ext4_error(sb, "Journal got removed while the fs was mounted!");
5094 return -EFSCORRUPTED;
5097 journal = EXT4_SB(sb)->s_journal;
5100 * Now check for any error status which may have been recorded in the
5101 * journal by a prior ext4_error() or ext4_abort()
5104 j_errno = jbd2_journal_errno(journal);
5108 errstr = ext4_decode_error(sb, j_errno, nbuf);
5109 ext4_warning(sb, "Filesystem error recorded "
5110 "from previous mount: %s", errstr);
5111 ext4_warning(sb, "Marking fs in need of filesystem check.");
5113 EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS;
5114 es->s_state |= cpu_to_le16(EXT4_ERROR_FS);
5115 ext4_commit_super(sb, 1);
5117 jbd2_journal_clear_err(journal);
5118 jbd2_journal_update_sb_errno(journal);
5124 * Force the running and committing transactions to commit,
5125 * and wait on the commit.
5127 int ext4_force_commit(struct super_block *sb)
5134 journal = EXT4_SB(sb)->s_journal;
5135 return ext4_journal_force_commit(journal);
5138 static int ext4_sync_fs(struct super_block *sb, int wait)
5142 bool needs_barrier = false;
5143 struct ext4_sb_info *sbi = EXT4_SB(sb);
5145 if (unlikely(ext4_forced_shutdown(sbi)))
5148 trace_ext4_sync_fs(sb, wait);
5149 flush_workqueue(sbi->rsv_conversion_wq);
5151 * Writeback quota in non-journalled quota case - journalled quota has
5154 dquot_writeback_dquots(sb, -1);
5156 * Data writeback is possible w/o journal transaction, so barrier must
5157 * being sent at the end of the function. But we can skip it if
5158 * transaction_commit will do it for us.
5160 if (sbi->s_journal) {
5161 target = jbd2_get_latest_transaction(sbi->s_journal);
5162 if (wait && sbi->s_journal->j_flags & JBD2_BARRIER &&
5163 !jbd2_trans_will_send_data_barrier(sbi->s_journal, target))
5164 needs_barrier = true;
5166 if (jbd2_journal_start_commit(sbi->s_journal, &target)) {
5168 ret = jbd2_log_wait_commit(sbi->s_journal,
5171 } else if (wait && test_opt(sb, BARRIER))
5172 needs_barrier = true;
5173 if (needs_barrier) {
5175 err = blkdev_issue_flush(sb->s_bdev, GFP_KERNEL, NULL);
5184 * LVM calls this function before a (read-only) snapshot is created. This
5185 * gives us a chance to flush the journal completely and mark the fs clean.
5187 * Note that only this function cannot bring a filesystem to be in a clean
5188 * state independently. It relies on upper layer to stop all data & metadata
5191 static int ext4_freeze(struct super_block *sb)
5199 journal = EXT4_SB(sb)->s_journal;
5202 /* Now we set up the journal barrier. */
5203 jbd2_journal_lock_updates(journal);
5206 * Don't clear the needs_recovery flag if we failed to
5207 * flush the journal.
5209 error = jbd2_journal_flush(journal);
5213 /* Journal blocked and flushed, clear needs_recovery flag. */
5214 ext4_clear_feature_journal_needs_recovery(sb);
5217 error = ext4_commit_super(sb, 1);
5220 /* we rely on upper layer to stop further updates */
5221 jbd2_journal_unlock_updates(journal);
5226 * Called by LVM after the snapshot is done. We need to reset the RECOVER
5227 * flag here, even though the filesystem is not technically dirty yet.
5229 static int ext4_unfreeze(struct super_block *sb)
5231 if (sb_rdonly(sb) || ext4_forced_shutdown(EXT4_SB(sb)))
5234 if (EXT4_SB(sb)->s_journal) {
5235 /* Reset the needs_recovery flag before the fs is unlocked. */
5236 ext4_set_feature_journal_needs_recovery(sb);
5239 ext4_commit_super(sb, 1);
5244 * Structure to save mount options for ext4_remount's benefit
5246 struct ext4_mount_options {
5247 unsigned long s_mount_opt;
5248 unsigned long s_mount_opt2;
5251 unsigned long s_commit_interval;
5252 u32 s_min_batch_time, s_max_batch_time;
5255 char *s_qf_names[EXT4_MAXQUOTAS];
5259 static int ext4_remount(struct super_block *sb, int *flags, char *data)
5261 struct ext4_super_block *es;
5262 struct ext4_sb_info *sbi = EXT4_SB(sb);
5263 unsigned long old_sb_flags, vfs_flags;
5264 struct ext4_mount_options old_opts;
5265 int enable_quota = 0;
5267 unsigned int journal_ioprio = DEFAULT_JOURNAL_IOPRIO;
5271 char *to_free[EXT4_MAXQUOTAS];
5273 char *orig_data = kstrdup(data, GFP_KERNEL);
5275 if (data && !orig_data)
5278 /* Store the original options */
5279 old_sb_flags = sb->s_flags;
5280 old_opts.s_mount_opt = sbi->s_mount_opt;
5281 old_opts.s_mount_opt2 = sbi->s_mount_opt2;
5282 old_opts.s_resuid = sbi->s_resuid;
5283 old_opts.s_resgid = sbi->s_resgid;
5284 old_opts.s_commit_interval = sbi->s_commit_interval;
5285 old_opts.s_min_batch_time = sbi->s_min_batch_time;
5286 old_opts.s_max_batch_time = sbi->s_max_batch_time;
5288 old_opts.s_jquota_fmt = sbi->s_jquota_fmt;
5289 for (i = 0; i < EXT4_MAXQUOTAS; i++)
5290 if (sbi->s_qf_names[i]) {
5291 char *qf_name = get_qf_name(sb, sbi, i);
5293 old_opts.s_qf_names[i] = kstrdup(qf_name, GFP_KERNEL);
5294 if (!old_opts.s_qf_names[i]) {
5295 for (j = 0; j < i; j++)
5296 kfree(old_opts.s_qf_names[j]);
5301 old_opts.s_qf_names[i] = NULL;
5303 if (sbi->s_journal && sbi->s_journal->j_task->io_context)
5304 journal_ioprio = sbi->s_journal->j_task->io_context->ioprio;
5307 * Some options can be enabled by ext4 and/or by VFS mount flag
5308 * either way we need to make sure it matches in both *flags and
5309 * s_flags. Copy those selected flags from *flags to s_flags
5311 vfs_flags = SB_LAZYTIME | SB_I_VERSION;
5312 sb->s_flags = (sb->s_flags & ~vfs_flags) | (*flags & vfs_flags);
5314 if (!parse_options(data, sb, NULL, &journal_ioprio, 1)) {
5319 if ((old_opts.s_mount_opt & EXT4_MOUNT_JOURNAL_CHECKSUM) ^
5320 test_opt(sb, JOURNAL_CHECKSUM)) {
5321 ext4_msg(sb, KERN_ERR, "changing journal_checksum "
5322 "during remount not supported; ignoring");
5323 sbi->s_mount_opt ^= EXT4_MOUNT_JOURNAL_CHECKSUM;
5326 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA) {
5327 if (test_opt2(sb, EXPLICIT_DELALLOC)) {
5328 ext4_msg(sb, KERN_ERR, "can't mount with "
5329 "both data=journal and delalloc");
5333 if (test_opt(sb, DIOREAD_NOLOCK)) {
5334 ext4_msg(sb, KERN_ERR, "can't mount with "
5335 "both data=journal and dioread_nolock");
5339 } else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA) {
5340 if (test_opt(sb, JOURNAL_ASYNC_COMMIT)) {
5341 ext4_msg(sb, KERN_ERR, "can't mount with "
5342 "journal_async_commit in data=ordered mode");
5348 if ((sbi->s_mount_opt ^ old_opts.s_mount_opt) & EXT4_MOUNT_NO_MBCACHE) {
5349 ext4_msg(sb, KERN_ERR, "can't enable nombcache during remount");
5354 if (sbi->s_mount_flags & EXT4_MF_FS_ABORTED)
5355 ext4_abort(sb, "Abort forced by user");
5357 sb->s_flags = (sb->s_flags & ~SB_POSIXACL) |
5358 (test_opt(sb, POSIX_ACL) ? SB_POSIXACL : 0);
5362 if (sbi->s_journal) {
5363 ext4_init_journal_params(sb, sbi->s_journal);
5364 set_task_ioprio(sbi->s_journal->j_task, journal_ioprio);
5367 if ((bool)(*flags & SB_RDONLY) != sb_rdonly(sb)) {
5368 if (sbi->s_mount_flags & EXT4_MF_FS_ABORTED) {
5373 if (*flags & SB_RDONLY) {
5374 err = sync_filesystem(sb);
5377 err = dquot_suspend(sb, -1);
5382 * First of all, the unconditional stuff we have to do
5383 * to disable replay of the journal when we next remount
5385 sb->s_flags |= SB_RDONLY;
5388 * OK, test if we are remounting a valid rw partition
5389 * readonly, and if so set the rdonly flag and then
5390 * mark the partition as valid again.
5392 if (!(es->s_state & cpu_to_le16(EXT4_VALID_FS)) &&
5393 (sbi->s_mount_state & EXT4_VALID_FS))
5394 es->s_state = cpu_to_le16(sbi->s_mount_state);
5396 if (sbi->s_journal) {
5398 * We let remount-ro finish even if marking fs
5399 * as clean failed...
5401 ext4_mark_recovery_complete(sb, es);
5404 kthread_stop(sbi->s_mmp_tsk);
5406 /* Make sure we can mount this feature set readwrite */
5407 if (ext4_has_feature_readonly(sb) ||
5408 !ext4_feature_set_ok(sb, 0)) {
5413 * Make sure the group descriptor checksums
5414 * are sane. If they aren't, refuse to remount r/w.
5416 for (g = 0; g < sbi->s_groups_count; g++) {
5417 struct ext4_group_desc *gdp =
5418 ext4_get_group_desc(sb, g, NULL);
5420 if (!ext4_group_desc_csum_verify(sb, g, gdp)) {
5421 ext4_msg(sb, KERN_ERR,
5422 "ext4_remount: Checksum for group %u failed (%u!=%u)",
5423 g, le16_to_cpu(ext4_group_desc_csum(sb, g, gdp)),
5424 le16_to_cpu(gdp->bg_checksum));
5431 * If we have an unprocessed orphan list hanging
5432 * around from a previously readonly bdev mount,
5433 * require a full umount/remount for now.
5435 if (es->s_last_orphan) {
5436 ext4_msg(sb, KERN_WARNING, "Couldn't "
5437 "remount RDWR because of unprocessed "
5438 "orphan inode list. Please "
5439 "umount/remount instead");
5445 * Mounting a RDONLY partition read-write, so reread
5446 * and store the current valid flag. (It may have
5447 * been changed by e2fsck since we originally mounted
5450 if (sbi->s_journal) {
5451 err = ext4_clear_journal_err(sb, es);
5455 sbi->s_mount_state = le16_to_cpu(es->s_state);
5457 err = ext4_setup_super(sb, es, 0);
5461 sb->s_flags &= ~SB_RDONLY;
5462 if (ext4_has_feature_mmp(sb))
5463 if (ext4_multi_mount_protect(sb,
5464 le64_to_cpu(es->s_mmp_block))) {
5473 * Reinitialize lazy itable initialization thread based on
5476 if (sb_rdonly(sb) || !test_opt(sb, INIT_INODE_TABLE))
5477 ext4_unregister_li_request(sb);
5479 ext4_group_t first_not_zeroed;
5480 first_not_zeroed = ext4_has_uninit_itable(sb);
5481 ext4_register_li_request(sb, first_not_zeroed);
5485 * Handle creation of system zone data early because it can fail.
5486 * Releasing of existing data is done when we are sure remount will
5489 if (test_opt(sb, BLOCK_VALIDITY) && !sbi->system_blks) {
5490 err = ext4_setup_system_zone(sb);
5495 if (sbi->s_journal == NULL && !(old_sb_flags & SB_RDONLY)) {
5496 err = ext4_commit_super(sb, 1);
5502 /* Release old quota file names */
5503 for (i = 0; i < EXT4_MAXQUOTAS; i++)
5504 kfree(old_opts.s_qf_names[i]);
5506 if (sb_any_quota_suspended(sb))
5507 dquot_resume(sb, -1);
5508 else if (ext4_has_feature_quota(sb)) {
5509 err = ext4_enable_quotas(sb);
5515 if (!test_opt(sb, BLOCK_VALIDITY) && sbi->system_blks)
5516 ext4_release_system_zone(sb);
5519 * Some options can be enabled by ext4 and/or by VFS mount flag
5520 * either way we need to make sure it matches in both *flags and
5521 * s_flags. Copy those selected flags from s_flags to *flags
5523 *flags = (*flags & ~vfs_flags) | (sb->s_flags & vfs_flags);
5525 ext4_msg(sb, KERN_INFO, "re-mounted. Opts: %s", orig_data);
5530 sb->s_flags = old_sb_flags;
5531 sbi->s_mount_opt = old_opts.s_mount_opt;
5532 sbi->s_mount_opt2 = old_opts.s_mount_opt2;
5533 sbi->s_resuid = old_opts.s_resuid;
5534 sbi->s_resgid = old_opts.s_resgid;
5535 sbi->s_commit_interval = old_opts.s_commit_interval;
5536 sbi->s_min_batch_time = old_opts.s_min_batch_time;
5537 sbi->s_max_batch_time = old_opts.s_max_batch_time;
5538 if (!test_opt(sb, BLOCK_VALIDITY) && sbi->system_blks)
5539 ext4_release_system_zone(sb);
5541 sbi->s_jquota_fmt = old_opts.s_jquota_fmt;
5542 for (i = 0; i < EXT4_MAXQUOTAS; i++) {
5543 to_free[i] = get_qf_name(sb, sbi, i);
5544 rcu_assign_pointer(sbi->s_qf_names[i], old_opts.s_qf_names[i]);
5547 for (i = 0; i < EXT4_MAXQUOTAS; i++)
5555 static int ext4_statfs_project(struct super_block *sb,
5556 kprojid_t projid, struct kstatfs *buf)
5559 struct dquot *dquot;
5563 qid = make_kqid_projid(projid);
5564 dquot = dqget(sb, qid);
5566 return PTR_ERR(dquot);
5567 spin_lock(&dquot->dq_dqb_lock);
5569 limit = (dquot->dq_dqb.dqb_bsoftlimit ?
5570 dquot->dq_dqb.dqb_bsoftlimit :
5571 dquot->dq_dqb.dqb_bhardlimit) >> sb->s_blocksize_bits;
5572 if (limit && buf->f_blocks > limit) {
5573 curblock = (dquot->dq_dqb.dqb_curspace +
5574 dquot->dq_dqb.dqb_rsvspace) >> sb->s_blocksize_bits;
5575 buf->f_blocks = limit;
5576 buf->f_bfree = buf->f_bavail =
5577 (buf->f_blocks > curblock) ?
5578 (buf->f_blocks - curblock) : 0;
5581 limit = dquot->dq_dqb.dqb_isoftlimit ?
5582 dquot->dq_dqb.dqb_isoftlimit :
5583 dquot->dq_dqb.dqb_ihardlimit;
5584 if (limit && buf->f_files > limit) {
5585 buf->f_files = limit;
5587 (buf->f_files > dquot->dq_dqb.dqb_curinodes) ?
5588 (buf->f_files - dquot->dq_dqb.dqb_curinodes) : 0;
5591 spin_unlock(&dquot->dq_dqb_lock);
5597 static int ext4_statfs(struct dentry *dentry, struct kstatfs *buf)
5599 struct super_block *sb = dentry->d_sb;
5600 struct ext4_sb_info *sbi = EXT4_SB(sb);
5601 struct ext4_super_block *es = sbi->s_es;
5602 ext4_fsblk_t overhead = 0, resv_blocks;
5605 resv_blocks = EXT4_C2B(sbi, atomic64_read(&sbi->s_resv_clusters));
5607 if (!test_opt(sb, MINIX_DF))
5608 overhead = sbi->s_overhead;
5610 buf->f_type = EXT4_SUPER_MAGIC;
5611 buf->f_bsize = sb->s_blocksize;
5612 buf->f_blocks = ext4_blocks_count(es) - EXT4_C2B(sbi, overhead);
5613 bfree = percpu_counter_sum_positive(&sbi->s_freeclusters_counter) -
5614 percpu_counter_sum_positive(&sbi->s_dirtyclusters_counter);
5615 /* prevent underflow in case that few free space is available */
5616 buf->f_bfree = EXT4_C2B(sbi, max_t(s64, bfree, 0));
5617 buf->f_bavail = buf->f_bfree -
5618 (ext4_r_blocks_count(es) + resv_blocks);
5619 if (buf->f_bfree < (ext4_r_blocks_count(es) + resv_blocks))
5621 buf->f_files = le32_to_cpu(es->s_inodes_count);
5622 buf->f_ffree = percpu_counter_sum_positive(&sbi->s_freeinodes_counter);
5623 buf->f_namelen = EXT4_NAME_LEN;
5624 fsid = le64_to_cpup((void *)es->s_uuid) ^
5625 le64_to_cpup((void *)es->s_uuid + sizeof(u64));
5626 buf->f_fsid.val[0] = fsid & 0xFFFFFFFFUL;
5627 buf->f_fsid.val[1] = (fsid >> 32) & 0xFFFFFFFFUL;
5630 if (ext4_test_inode_flag(dentry->d_inode, EXT4_INODE_PROJINHERIT) &&
5631 sb_has_quota_limits_enabled(sb, PRJQUOTA))
5632 ext4_statfs_project(sb, EXT4_I(dentry->d_inode)->i_projid, buf);
5641 * Helper functions so that transaction is started before we acquire dqio_sem
5642 * to keep correct lock ordering of transaction > dqio_sem
5644 static inline struct inode *dquot_to_inode(struct dquot *dquot)
5646 return sb_dqopt(dquot->dq_sb)->files[dquot->dq_id.type];
5649 static int ext4_write_dquot(struct dquot *dquot)
5653 struct inode *inode;
5655 inode = dquot_to_inode(dquot);
5656 handle = ext4_journal_start(inode, EXT4_HT_QUOTA,
5657 EXT4_QUOTA_TRANS_BLOCKS(dquot->dq_sb));
5659 return PTR_ERR(handle);
5660 ret = dquot_commit(dquot);
5661 err = ext4_journal_stop(handle);
5667 static int ext4_acquire_dquot(struct dquot *dquot)
5672 handle = ext4_journal_start(dquot_to_inode(dquot), EXT4_HT_QUOTA,
5673 EXT4_QUOTA_INIT_BLOCKS(dquot->dq_sb));
5675 return PTR_ERR(handle);
5676 ret = dquot_acquire(dquot);
5677 err = ext4_journal_stop(handle);
5683 static int ext4_release_dquot(struct dquot *dquot)
5688 handle = ext4_journal_start(dquot_to_inode(dquot), EXT4_HT_QUOTA,
5689 EXT4_QUOTA_DEL_BLOCKS(dquot->dq_sb));
5690 if (IS_ERR(handle)) {
5691 /* Release dquot anyway to avoid endless cycle in dqput() */
5692 dquot_release(dquot);
5693 return PTR_ERR(handle);
5695 ret = dquot_release(dquot);
5696 err = ext4_journal_stop(handle);
5702 static int ext4_mark_dquot_dirty(struct dquot *dquot)
5704 struct super_block *sb = dquot->dq_sb;
5705 struct ext4_sb_info *sbi = EXT4_SB(sb);
5707 /* Are we journaling quotas? */
5708 if (ext4_has_feature_quota(sb) ||
5709 sbi->s_qf_names[USRQUOTA] || sbi->s_qf_names[GRPQUOTA]) {
5710 dquot_mark_dquot_dirty(dquot);
5711 return ext4_write_dquot(dquot);
5713 return dquot_mark_dquot_dirty(dquot);
5717 static int ext4_write_info(struct super_block *sb, int type)
5722 /* Data block + inode block */
5723 handle = ext4_journal_start(d_inode(sb->s_root), EXT4_HT_QUOTA, 2);
5725 return PTR_ERR(handle);
5726 ret = dquot_commit_info(sb, type);
5727 err = ext4_journal_stop(handle);
5734 * Turn on quotas during mount time - we need to find
5735 * the quota file and such...
5737 static int ext4_quota_on_mount(struct super_block *sb, int type)
5739 return dquot_quota_on_mount(sb, get_qf_name(sb, EXT4_SB(sb), type),
5740 EXT4_SB(sb)->s_jquota_fmt, type);
5743 static void lockdep_set_quota_inode(struct inode *inode, int subclass)
5745 struct ext4_inode_info *ei = EXT4_I(inode);
5747 /* The first argument of lockdep_set_subclass has to be
5748 * *exactly* the same as the argument to init_rwsem() --- in
5749 * this case, in init_once() --- or lockdep gets unhappy
5750 * because the name of the lock is set using the
5751 * stringification of the argument to init_rwsem().
5753 (void) ei; /* shut up clang warning if !CONFIG_LOCKDEP */
5754 lockdep_set_subclass(&ei->i_data_sem, subclass);
5758 * Standard function to be called on quota_on
5760 static int ext4_quota_on(struct super_block *sb, int type, int format_id,
5761 const struct path *path)
5765 if (!test_opt(sb, QUOTA))
5768 /* Quotafile not on the same filesystem? */
5769 if (path->dentry->d_sb != sb)
5772 /* Quota already enabled for this file? */
5773 if (IS_NOQUOTA(d_inode(path->dentry)))
5776 /* Journaling quota? */
5777 if (EXT4_SB(sb)->s_qf_names[type]) {
5778 /* Quotafile not in fs root? */
5779 if (path->dentry->d_parent != sb->s_root)
5780 ext4_msg(sb, KERN_WARNING,
5781 "Quota file not on filesystem root. "
5782 "Journaled quota will not work");
5783 sb_dqopt(sb)->flags |= DQUOT_NOLIST_DIRTY;
5786 * Clear the flag just in case mount options changed since
5789 sb_dqopt(sb)->flags &= ~DQUOT_NOLIST_DIRTY;
5793 * When we journal data on quota file, we have to flush journal to see
5794 * all updates to the file when we bypass pagecache...
5796 if (EXT4_SB(sb)->s_journal &&
5797 ext4_should_journal_data(d_inode(path->dentry))) {
5799 * We don't need to lock updates but journal_flush() could
5800 * otherwise be livelocked...
5802 jbd2_journal_lock_updates(EXT4_SB(sb)->s_journal);
5803 err = jbd2_journal_flush(EXT4_SB(sb)->s_journal);
5804 jbd2_journal_unlock_updates(EXT4_SB(sb)->s_journal);
5809 lockdep_set_quota_inode(path->dentry->d_inode, I_DATA_SEM_QUOTA);
5810 err = dquot_quota_on(sb, type, format_id, path);
5812 struct inode *inode = d_inode(path->dentry);
5816 * Set inode flags to prevent userspace from messing with quota
5817 * files. If this fails, we return success anyway since quotas
5818 * are already enabled and this is not a hard failure.
5821 handle = ext4_journal_start(inode, EXT4_HT_QUOTA, 1);
5824 EXT4_I(inode)->i_flags |= EXT4_NOATIME_FL | EXT4_IMMUTABLE_FL;
5825 inode_set_flags(inode, S_NOATIME | S_IMMUTABLE,
5826 S_NOATIME | S_IMMUTABLE);
5827 ext4_mark_inode_dirty(handle, inode);
5828 ext4_journal_stop(handle);
5830 inode_unlock(inode);
5832 dquot_quota_off(sb, type);
5835 lockdep_set_quota_inode(path->dentry->d_inode,
5840 static int ext4_quota_enable(struct super_block *sb, int type, int format_id,
5844 struct inode *qf_inode;
5845 unsigned long qf_inums[EXT4_MAXQUOTAS] = {
5846 le32_to_cpu(EXT4_SB(sb)->s_es->s_usr_quota_inum),
5847 le32_to_cpu(EXT4_SB(sb)->s_es->s_grp_quota_inum),
5848 le32_to_cpu(EXT4_SB(sb)->s_es->s_prj_quota_inum)
5851 BUG_ON(!ext4_has_feature_quota(sb));
5853 if (!qf_inums[type])
5856 qf_inode = ext4_iget(sb, qf_inums[type], EXT4_IGET_SPECIAL);
5857 if (IS_ERR(qf_inode)) {
5858 ext4_error(sb, "Bad quota inode # %lu", qf_inums[type]);
5859 return PTR_ERR(qf_inode);
5862 /* Don't account quota for quota files to avoid recursion */
5863 qf_inode->i_flags |= S_NOQUOTA;
5864 lockdep_set_quota_inode(qf_inode, I_DATA_SEM_QUOTA);
5865 err = dquot_enable(qf_inode, type, format_id, flags);
5867 lockdep_set_quota_inode(qf_inode, I_DATA_SEM_NORMAL);
5873 /* Enable usage tracking for all quota types. */
5874 static int ext4_enable_quotas(struct super_block *sb)
5877 unsigned long qf_inums[EXT4_MAXQUOTAS] = {
5878 le32_to_cpu(EXT4_SB(sb)->s_es->s_usr_quota_inum),
5879 le32_to_cpu(EXT4_SB(sb)->s_es->s_grp_quota_inum),
5880 le32_to_cpu(EXT4_SB(sb)->s_es->s_prj_quota_inum)
5882 bool quota_mopt[EXT4_MAXQUOTAS] = {
5883 test_opt(sb, USRQUOTA),
5884 test_opt(sb, GRPQUOTA),
5885 test_opt(sb, PRJQUOTA),
5888 sb_dqopt(sb)->flags |= DQUOT_QUOTA_SYS_FILE | DQUOT_NOLIST_DIRTY;
5889 for (type = 0; type < EXT4_MAXQUOTAS; type++) {
5890 if (qf_inums[type]) {
5891 err = ext4_quota_enable(sb, type, QFMT_VFS_V1,
5892 DQUOT_USAGE_ENABLED |
5893 (quota_mopt[type] ? DQUOT_LIMITS_ENABLED : 0));
5896 "Failed to enable quota tracking "
5897 "(type=%d, err=%d). Please run "
5898 "e2fsck to fix.", type, err);
5899 for (type--; type >= 0; type--) {
5900 struct inode *inode;
5902 inode = sb_dqopt(sb)->files[type];
5904 inode = igrab(inode);
5905 dquot_quota_off(sb, type);
5907 lockdep_set_quota_inode(inode,
5920 static int ext4_quota_off(struct super_block *sb, int type)
5922 struct inode *inode = sb_dqopt(sb)->files[type];
5926 /* Force all delayed allocation blocks to be allocated.
5927 * Caller already holds s_umount sem */
5928 if (test_opt(sb, DELALLOC))
5929 sync_filesystem(sb);
5931 if (!inode || !igrab(inode))
5934 err = dquot_quota_off(sb, type);
5935 if (err || ext4_has_feature_quota(sb))
5940 * Update modification times of quota files when userspace can
5941 * start looking at them. If we fail, we return success anyway since
5942 * this is not a hard failure and quotas are already disabled.
5944 handle = ext4_journal_start(inode, EXT4_HT_QUOTA, 1);
5947 EXT4_I(inode)->i_flags &= ~(EXT4_NOATIME_FL | EXT4_IMMUTABLE_FL);
5948 inode_set_flags(inode, 0, S_NOATIME | S_IMMUTABLE);
5949 inode->i_mtime = inode->i_ctime = current_time(inode);
5950 ext4_mark_inode_dirty(handle, inode);
5951 ext4_journal_stop(handle);
5953 inode_unlock(inode);
5955 lockdep_set_quota_inode(inode, I_DATA_SEM_NORMAL);
5959 return dquot_quota_off(sb, type);
5962 /* Read data from quotafile - avoid pagecache and such because we cannot afford
5963 * acquiring the locks... As quota files are never truncated and quota code
5964 * itself serializes the operations (and no one else should touch the files)
5965 * we don't have to be afraid of races */
5966 static ssize_t ext4_quota_read(struct super_block *sb, int type, char *data,
5967 size_t len, loff_t off)
5969 struct inode *inode = sb_dqopt(sb)->files[type];
5970 ext4_lblk_t blk = off >> EXT4_BLOCK_SIZE_BITS(sb);
5971 int offset = off & (sb->s_blocksize - 1);
5974 struct buffer_head *bh;
5975 loff_t i_size = i_size_read(inode);
5979 if (off+len > i_size)
5982 while (toread > 0) {
5983 tocopy = sb->s_blocksize - offset < toread ?
5984 sb->s_blocksize - offset : toread;
5985 bh = ext4_bread(NULL, inode, blk, 0);
5988 if (!bh) /* A hole? */
5989 memset(data, 0, tocopy);
5991 memcpy(data, bh->b_data+offset, tocopy);
6001 /* Write to quotafile (we know the transaction is already started and has
6002 * enough credits) */
6003 static ssize_t ext4_quota_write(struct super_block *sb, int type,
6004 const char *data, size_t len, loff_t off)
6006 struct inode *inode = sb_dqopt(sb)->files[type];
6007 ext4_lblk_t blk = off >> EXT4_BLOCK_SIZE_BITS(sb);
6008 int err, offset = off & (sb->s_blocksize - 1);
6010 struct buffer_head *bh;
6011 handle_t *handle = journal_current_handle();
6014 ext4_msg(sb, KERN_WARNING, "Quota write (off=%llu, len=%llu)"
6015 " cancelled because transaction is not started",
6016 (unsigned long long)off, (unsigned long long)len);
6020 * Since we account only one data block in transaction credits,
6021 * then it is impossible to cross a block boundary.
6023 if (sb->s_blocksize - offset < len) {
6024 ext4_msg(sb, KERN_WARNING, "Quota write (off=%llu, len=%llu)"
6025 " cancelled because not block aligned",
6026 (unsigned long long)off, (unsigned long long)len);
6031 bh = ext4_bread(handle, inode, blk,
6032 EXT4_GET_BLOCKS_CREATE |
6033 EXT4_GET_BLOCKS_METADATA_NOFAIL);
6034 } while (IS_ERR(bh) && (PTR_ERR(bh) == -ENOSPC) &&
6035 ext4_should_retry_alloc(inode->i_sb, &retries));
6040 BUFFER_TRACE(bh, "get write access");
6041 err = ext4_journal_get_write_access(handle, bh);
6047 memcpy(bh->b_data+offset, data, len);
6048 flush_dcache_page(bh->b_page);
6050 err = ext4_handle_dirty_metadata(handle, NULL, bh);
6053 if (inode->i_size < off + len) {
6054 i_size_write(inode, off + len);
6055 EXT4_I(inode)->i_disksize = inode->i_size;
6056 ext4_mark_inode_dirty(handle, inode);
6061 static int ext4_get_next_id(struct super_block *sb, struct kqid *qid)
6063 const struct quota_format_ops *ops;
6065 if (!sb_has_quota_loaded(sb, qid->type))
6067 ops = sb_dqopt(sb)->ops[qid->type];
6068 if (!ops || !ops->get_next_id)
6070 return dquot_get_next_id(sb, qid);
6074 static struct dentry *ext4_mount(struct file_system_type *fs_type, int flags,
6075 const char *dev_name, void *data)
6077 return mount_bdev(fs_type, flags, dev_name, data, ext4_fill_super);
6080 #if !defined(CONFIG_EXT2_FS) && !defined(CONFIG_EXT2_FS_MODULE) && defined(CONFIG_EXT4_USE_FOR_EXT2)
6081 static inline void register_as_ext2(void)
6083 int err = register_filesystem(&ext2_fs_type);
6086 "EXT4-fs: Unable to register as ext2 (%d)\n", err);
6089 static inline void unregister_as_ext2(void)
6091 unregister_filesystem(&ext2_fs_type);
6094 static inline int ext2_feature_set_ok(struct super_block *sb)
6096 if (ext4_has_unknown_ext2_incompat_features(sb))
6100 if (ext4_has_unknown_ext2_ro_compat_features(sb))
6105 static inline void register_as_ext2(void) { }
6106 static inline void unregister_as_ext2(void) { }
6107 static inline int ext2_feature_set_ok(struct super_block *sb) { return 0; }
6110 static inline void register_as_ext3(void)
6112 int err = register_filesystem(&ext3_fs_type);
6115 "EXT4-fs: Unable to register as ext3 (%d)\n", err);
6118 static inline void unregister_as_ext3(void)
6120 unregister_filesystem(&ext3_fs_type);
6123 static inline int ext3_feature_set_ok(struct super_block *sb)
6125 if (ext4_has_unknown_ext3_incompat_features(sb))
6127 if (!ext4_has_feature_journal(sb))
6131 if (ext4_has_unknown_ext3_ro_compat_features(sb))
6136 static struct file_system_type ext4_fs_type = {
6137 .owner = THIS_MODULE,
6139 .mount = ext4_mount,
6140 .kill_sb = kill_block_super,
6141 .fs_flags = FS_REQUIRES_DEV,
6143 MODULE_ALIAS_FS("ext4");
6145 /* Shared across all ext4 file systems */
6146 wait_queue_head_t ext4__ioend_wq[EXT4_WQ_HASH_SZ];
6148 static int __init ext4_init_fs(void)
6152 ratelimit_state_init(&ext4_mount_msg_ratelimit, 30 * HZ, 64);
6153 ext4_li_info = NULL;
6154 mutex_init(&ext4_li_mtx);
6156 /* Build-time check for flags consistency */
6157 ext4_check_flag_values();
6159 for (i = 0; i < EXT4_WQ_HASH_SZ; i++)
6160 init_waitqueue_head(&ext4__ioend_wq[i]);
6162 err = ext4_init_es();
6166 err = ext4_init_pageio();
6170 err = ext4_init_system_zone();
6174 err = ext4_init_sysfs();
6178 err = ext4_init_mballoc();
6181 err = init_inodecache();
6186 err = register_filesystem(&ext4_fs_type);
6192 unregister_as_ext2();
6193 unregister_as_ext3();
6194 destroy_inodecache();
6196 ext4_exit_mballoc();
6200 ext4_exit_system_zone();
6209 static void __exit ext4_exit_fs(void)
6211 ext4_destroy_lazyinit_thread();
6212 unregister_as_ext2();
6213 unregister_as_ext3();
6214 unregister_filesystem(&ext4_fs_type);
6215 destroy_inodecache();
6216 ext4_exit_mballoc();
6218 ext4_exit_system_zone();
6223 MODULE_AUTHOR("Remy Card, Stephen Tweedie, Andrew Morton, Andreas Dilger, Theodore Ts'o and others");
6224 MODULE_DESCRIPTION("Fourth Extended Filesystem");
6225 MODULE_LICENSE("GPL");
6226 MODULE_SOFTDEP("pre: crc32c");
6227 module_init(ext4_init_fs)
6228 module_exit(ext4_exit_fs)