x86/mm/pat: Don't report PAT on CPUs that don't support it
[linux/fpc-iii.git] / fs / ext4 / super.c
bloba9448db1cf7e87c2bd41daac6fbab7729bc87ecc
1 /*
2 * linux/fs/ext4/super.c
4 * Copyright (C) 1992, 1993, 1994, 1995
5 * Remy Card (card@masi.ibp.fr)
6 * Laboratoire MASI - Institut Blaise Pascal
7 * Universite Pierre et Marie Curie (Paris VI)
9 * from
11 * linux/fs/minix/inode.c
13 * Copyright (C) 1991, 1992 Linus Torvalds
15 * Big-endian to little-endian byte-swapping/bitmaps by
16 * David S. Miller (davem@caip.rutgers.edu), 1995
19 #include <linux/module.h>
20 #include <linux/string.h>
21 #include <linux/fs.h>
22 #include <linux/time.h>
23 #include <linux/vmalloc.h>
24 #include <linux/slab.h>
25 #include <linux/init.h>
26 #include <linux/blkdev.h>
27 #include <linux/backing-dev.h>
28 #include <linux/parser.h>
29 #include <linux/buffer_head.h>
30 #include <linux/exportfs.h>
31 #include <linux/vfs.h>
32 #include <linux/random.h>
33 #include <linux/mount.h>
34 #include <linux/namei.h>
35 #include <linux/quotaops.h>
36 #include <linux/seq_file.h>
37 #include <linux/ctype.h>
38 #include <linux/log2.h>
39 #include <linux/crc16.h>
40 #include <linux/cleancache.h>
41 #include <linux/uaccess.h>
43 #include <linux/kthread.h>
44 #include <linux/freezer.h>
46 #include "ext4.h"
47 #include "ext4_extents.h" /* Needed for trace points definition */
48 #include "ext4_jbd2.h"
49 #include "xattr.h"
50 #include "acl.h"
51 #include "mballoc.h"
53 #define CREATE_TRACE_POINTS
54 #include <trace/events/ext4.h>
56 static struct ext4_lazy_init *ext4_li_info;
57 static struct mutex ext4_li_mtx;
58 static struct ratelimit_state ext4_mount_msg_ratelimit;
60 static int ext4_load_journal(struct super_block *, struct ext4_super_block *,
61 unsigned long journal_devnum);
62 static int ext4_show_options(struct seq_file *seq, struct dentry *root);
63 static int ext4_commit_super(struct super_block *sb, int sync);
64 static void ext4_mark_recovery_complete(struct super_block *sb,
65 struct ext4_super_block *es);
66 static void ext4_clear_journal_err(struct super_block *sb,
67 struct ext4_super_block *es);
68 static int ext4_sync_fs(struct super_block *sb, int wait);
69 static int ext4_remount(struct super_block *sb, int *flags, char *data);
70 static int ext4_statfs(struct dentry *dentry, struct kstatfs *buf);
71 static int ext4_unfreeze(struct super_block *sb);
72 static int ext4_freeze(struct super_block *sb);
73 static struct dentry *ext4_mount(struct file_system_type *fs_type, int flags,
74 const char *dev_name, void *data);
75 static inline int ext2_feature_set_ok(struct super_block *sb);
76 static inline int ext3_feature_set_ok(struct super_block *sb);
77 static int ext4_feature_set_ok(struct super_block *sb, int readonly);
78 static void ext4_destroy_lazyinit_thread(void);
79 static void ext4_unregister_li_request(struct super_block *sb);
80 static void ext4_clear_request_list(void);
81 static struct inode *ext4_get_journal_inode(struct super_block *sb,
82 unsigned int journal_inum);
85 * Lock ordering
87 * Note the difference between i_mmap_sem (EXT4_I(inode)->i_mmap_sem) and
88 * i_mmap_rwsem (inode->i_mmap_rwsem)!
90 * page fault path:
91 * mmap_sem -> sb_start_pagefault -> i_mmap_sem (r) -> transaction start ->
92 * page lock -> i_data_sem (rw)
94 * buffered write path:
95 * sb_start_write -> i_mutex -> mmap_sem
96 * sb_start_write -> i_mutex -> transaction start -> page lock ->
97 * i_data_sem (rw)
99 * truncate:
100 * sb_start_write -> i_mutex -> EXT4_STATE_DIOREAD_LOCK (w) -> i_mmap_sem (w) ->
101 * i_mmap_rwsem (w) -> page lock
102 * sb_start_write -> i_mutex -> EXT4_STATE_DIOREAD_LOCK (w) -> i_mmap_sem (w) ->
103 * transaction start -> i_data_sem (rw)
105 * direct IO:
106 * sb_start_write -> i_mutex -> EXT4_STATE_DIOREAD_LOCK (r) -> mmap_sem
107 * sb_start_write -> i_mutex -> EXT4_STATE_DIOREAD_LOCK (r) ->
108 * transaction start -> i_data_sem (rw)
110 * writepages:
111 * transaction start -> page lock(s) -> i_data_sem (rw)
114 #if !defined(CONFIG_EXT2_FS) && !defined(CONFIG_EXT2_FS_MODULE) && defined(CONFIG_EXT4_USE_FOR_EXT2)
115 static struct file_system_type ext2_fs_type = {
116 .owner = THIS_MODULE,
117 .name = "ext2",
118 .mount = ext4_mount,
119 .kill_sb = kill_block_super,
120 .fs_flags = FS_REQUIRES_DEV,
122 MODULE_ALIAS_FS("ext2");
123 MODULE_ALIAS("ext2");
124 #define IS_EXT2_SB(sb) ((sb)->s_bdev->bd_holder == &ext2_fs_type)
125 #else
126 #define IS_EXT2_SB(sb) (0)
127 #endif
130 static struct file_system_type ext3_fs_type = {
131 .owner = THIS_MODULE,
132 .name = "ext3",
133 .mount = ext4_mount,
134 .kill_sb = kill_block_super,
135 .fs_flags = FS_REQUIRES_DEV,
137 MODULE_ALIAS_FS("ext3");
138 MODULE_ALIAS("ext3");
139 #define IS_EXT3_SB(sb) ((sb)->s_bdev->bd_holder == &ext3_fs_type)
141 static int ext4_verify_csum_type(struct super_block *sb,
142 struct ext4_super_block *es)
144 if (!ext4_has_feature_metadata_csum(sb))
145 return 1;
147 return es->s_checksum_type == EXT4_CRC32C_CHKSUM;
150 static __le32 ext4_superblock_csum(struct super_block *sb,
151 struct ext4_super_block *es)
153 struct ext4_sb_info *sbi = EXT4_SB(sb);
154 int offset = offsetof(struct ext4_super_block, s_checksum);
155 __u32 csum;
157 csum = ext4_chksum(sbi, ~0, (char *)es, offset);
159 return cpu_to_le32(csum);
162 static int ext4_superblock_csum_verify(struct super_block *sb,
163 struct ext4_super_block *es)
165 if (!ext4_has_metadata_csum(sb))
166 return 1;
168 return es->s_checksum == ext4_superblock_csum(sb, es);
171 void ext4_superblock_csum_set(struct super_block *sb)
173 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
175 if (!ext4_has_metadata_csum(sb))
176 return;
178 es->s_checksum = ext4_superblock_csum(sb, es);
181 void *ext4_kvmalloc(size_t size, gfp_t flags)
183 void *ret;
185 ret = kmalloc(size, flags | __GFP_NOWARN);
186 if (!ret)
187 ret = __vmalloc(size, flags, PAGE_KERNEL);
188 return ret;
191 void *ext4_kvzalloc(size_t size, gfp_t flags)
193 void *ret;
195 ret = kzalloc(size, flags | __GFP_NOWARN);
196 if (!ret)
197 ret = __vmalloc(size, flags | __GFP_ZERO, PAGE_KERNEL);
198 return ret;
201 ext4_fsblk_t ext4_block_bitmap(struct super_block *sb,
202 struct ext4_group_desc *bg)
204 return le32_to_cpu(bg->bg_block_bitmap_lo) |
205 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
206 (ext4_fsblk_t)le32_to_cpu(bg->bg_block_bitmap_hi) << 32 : 0);
209 ext4_fsblk_t ext4_inode_bitmap(struct super_block *sb,
210 struct ext4_group_desc *bg)
212 return le32_to_cpu(bg->bg_inode_bitmap_lo) |
213 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
214 (ext4_fsblk_t)le32_to_cpu(bg->bg_inode_bitmap_hi) << 32 : 0);
217 ext4_fsblk_t ext4_inode_table(struct super_block *sb,
218 struct ext4_group_desc *bg)
220 return le32_to_cpu(bg->bg_inode_table_lo) |
221 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
222 (ext4_fsblk_t)le32_to_cpu(bg->bg_inode_table_hi) << 32 : 0);
225 __u32 ext4_free_group_clusters(struct super_block *sb,
226 struct ext4_group_desc *bg)
228 return le16_to_cpu(bg->bg_free_blocks_count_lo) |
229 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
230 (__u32)le16_to_cpu(bg->bg_free_blocks_count_hi) << 16 : 0);
233 __u32 ext4_free_inodes_count(struct super_block *sb,
234 struct ext4_group_desc *bg)
236 return le16_to_cpu(bg->bg_free_inodes_count_lo) |
237 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
238 (__u32)le16_to_cpu(bg->bg_free_inodes_count_hi) << 16 : 0);
241 __u32 ext4_used_dirs_count(struct super_block *sb,
242 struct ext4_group_desc *bg)
244 return le16_to_cpu(bg->bg_used_dirs_count_lo) |
245 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
246 (__u32)le16_to_cpu(bg->bg_used_dirs_count_hi) << 16 : 0);
249 __u32 ext4_itable_unused_count(struct super_block *sb,
250 struct ext4_group_desc *bg)
252 return le16_to_cpu(bg->bg_itable_unused_lo) |
253 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
254 (__u32)le16_to_cpu(bg->bg_itable_unused_hi) << 16 : 0);
257 void ext4_block_bitmap_set(struct super_block *sb,
258 struct ext4_group_desc *bg, ext4_fsblk_t blk)
260 bg->bg_block_bitmap_lo = cpu_to_le32((u32)blk);
261 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
262 bg->bg_block_bitmap_hi = cpu_to_le32(blk >> 32);
265 void ext4_inode_bitmap_set(struct super_block *sb,
266 struct ext4_group_desc *bg, ext4_fsblk_t blk)
268 bg->bg_inode_bitmap_lo = cpu_to_le32((u32)blk);
269 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
270 bg->bg_inode_bitmap_hi = cpu_to_le32(blk >> 32);
273 void ext4_inode_table_set(struct super_block *sb,
274 struct ext4_group_desc *bg, ext4_fsblk_t blk)
276 bg->bg_inode_table_lo = cpu_to_le32((u32)blk);
277 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
278 bg->bg_inode_table_hi = cpu_to_le32(blk >> 32);
281 void ext4_free_group_clusters_set(struct super_block *sb,
282 struct ext4_group_desc *bg, __u32 count)
284 bg->bg_free_blocks_count_lo = cpu_to_le16((__u16)count);
285 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
286 bg->bg_free_blocks_count_hi = cpu_to_le16(count >> 16);
289 void ext4_free_inodes_set(struct super_block *sb,
290 struct ext4_group_desc *bg, __u32 count)
292 bg->bg_free_inodes_count_lo = cpu_to_le16((__u16)count);
293 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
294 bg->bg_free_inodes_count_hi = cpu_to_le16(count >> 16);
297 void ext4_used_dirs_set(struct super_block *sb,
298 struct ext4_group_desc *bg, __u32 count)
300 bg->bg_used_dirs_count_lo = cpu_to_le16((__u16)count);
301 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
302 bg->bg_used_dirs_count_hi = cpu_to_le16(count >> 16);
305 void ext4_itable_unused_set(struct super_block *sb,
306 struct ext4_group_desc *bg, __u32 count)
308 bg->bg_itable_unused_lo = cpu_to_le16((__u16)count);
309 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
310 bg->bg_itable_unused_hi = cpu_to_le16(count >> 16);
314 static void __save_error_info(struct super_block *sb, const char *func,
315 unsigned int line)
317 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
319 EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS;
320 if (bdev_read_only(sb->s_bdev))
321 return;
322 es->s_state |= cpu_to_le16(EXT4_ERROR_FS);
323 es->s_last_error_time = cpu_to_le32(get_seconds());
324 strncpy(es->s_last_error_func, func, sizeof(es->s_last_error_func));
325 es->s_last_error_line = cpu_to_le32(line);
326 if (!es->s_first_error_time) {
327 es->s_first_error_time = es->s_last_error_time;
328 strncpy(es->s_first_error_func, func,
329 sizeof(es->s_first_error_func));
330 es->s_first_error_line = cpu_to_le32(line);
331 es->s_first_error_ino = es->s_last_error_ino;
332 es->s_first_error_block = es->s_last_error_block;
335 * Start the daily error reporting function if it hasn't been
336 * started already
338 if (!es->s_error_count)
339 mod_timer(&EXT4_SB(sb)->s_err_report, jiffies + 24*60*60*HZ);
340 le32_add_cpu(&es->s_error_count, 1);
343 static void save_error_info(struct super_block *sb, const char *func,
344 unsigned int line)
346 __save_error_info(sb, func, line);
347 ext4_commit_super(sb, 1);
351 * The del_gendisk() function uninitializes the disk-specific data
352 * structures, including the bdi structure, without telling anyone
353 * else. Once this happens, any attempt to call mark_buffer_dirty()
354 * (for example, by ext4_commit_super), will cause a kernel OOPS.
355 * This is a kludge to prevent these oops until we can put in a proper
356 * hook in del_gendisk() to inform the VFS and file system layers.
358 static int block_device_ejected(struct super_block *sb)
360 struct inode *bd_inode = sb->s_bdev->bd_inode;
361 struct backing_dev_info *bdi = inode_to_bdi(bd_inode);
363 return bdi->dev == NULL;
366 static void ext4_journal_commit_callback(journal_t *journal, transaction_t *txn)
368 struct super_block *sb = journal->j_private;
369 struct ext4_sb_info *sbi = EXT4_SB(sb);
370 int error = is_journal_aborted(journal);
371 struct ext4_journal_cb_entry *jce;
373 BUG_ON(txn->t_state == T_FINISHED);
374 spin_lock(&sbi->s_md_lock);
375 while (!list_empty(&txn->t_private_list)) {
376 jce = list_entry(txn->t_private_list.next,
377 struct ext4_journal_cb_entry, jce_list);
378 list_del_init(&jce->jce_list);
379 spin_unlock(&sbi->s_md_lock);
380 jce->jce_func(sb, jce, error);
381 spin_lock(&sbi->s_md_lock);
383 spin_unlock(&sbi->s_md_lock);
386 /* Deal with the reporting of failure conditions on a filesystem such as
387 * inconsistencies detected or read IO failures.
389 * On ext2, we can store the error state of the filesystem in the
390 * superblock. That is not possible on ext4, because we may have other
391 * write ordering constraints on the superblock which prevent us from
392 * writing it out straight away; and given that the journal is about to
393 * be aborted, we can't rely on the current, or future, transactions to
394 * write out the superblock safely.
396 * We'll just use the jbd2_journal_abort() error code to record an error in
397 * the journal instead. On recovery, the journal will complain about
398 * that error until we've noted it down and cleared it.
401 static void ext4_handle_error(struct super_block *sb)
403 if (sb->s_flags & MS_RDONLY)
404 return;
406 if (!test_opt(sb, ERRORS_CONT)) {
407 journal_t *journal = EXT4_SB(sb)->s_journal;
409 EXT4_SB(sb)->s_mount_flags |= EXT4_MF_FS_ABORTED;
410 if (journal)
411 jbd2_journal_abort(journal, -EIO);
413 if (test_opt(sb, ERRORS_RO)) {
414 ext4_msg(sb, KERN_CRIT, "Remounting filesystem read-only");
416 * Make sure updated value of ->s_mount_flags will be visible
417 * before ->s_flags update
419 smp_wmb();
420 sb->s_flags |= MS_RDONLY;
422 if (test_opt(sb, ERRORS_PANIC)) {
423 if (EXT4_SB(sb)->s_journal &&
424 !(EXT4_SB(sb)->s_journal->j_flags & JBD2_REC_ERR))
425 return;
426 panic("EXT4-fs (device %s): panic forced after error\n",
427 sb->s_id);
431 #define ext4_error_ratelimit(sb) \
432 ___ratelimit(&(EXT4_SB(sb)->s_err_ratelimit_state), \
433 "EXT4-fs error")
435 void __ext4_error(struct super_block *sb, const char *function,
436 unsigned int line, const char *fmt, ...)
438 struct va_format vaf;
439 va_list args;
441 if (unlikely(ext4_forced_shutdown(EXT4_SB(sb))))
442 return;
444 if (ext4_error_ratelimit(sb)) {
445 va_start(args, fmt);
446 vaf.fmt = fmt;
447 vaf.va = &args;
448 printk(KERN_CRIT
449 "EXT4-fs error (device %s): %s:%d: comm %s: %pV\n",
450 sb->s_id, function, line, current->comm, &vaf);
451 va_end(args);
453 save_error_info(sb, function, line);
454 ext4_handle_error(sb);
457 void __ext4_error_inode(struct inode *inode, const char *function,
458 unsigned int line, ext4_fsblk_t block,
459 const char *fmt, ...)
461 va_list args;
462 struct va_format vaf;
463 struct ext4_super_block *es = EXT4_SB(inode->i_sb)->s_es;
465 if (unlikely(ext4_forced_shutdown(EXT4_SB(inode->i_sb))))
466 return;
468 es->s_last_error_ino = cpu_to_le32(inode->i_ino);
469 es->s_last_error_block = cpu_to_le64(block);
470 if (ext4_error_ratelimit(inode->i_sb)) {
471 va_start(args, fmt);
472 vaf.fmt = fmt;
473 vaf.va = &args;
474 if (block)
475 printk(KERN_CRIT "EXT4-fs error (device %s): %s:%d: "
476 "inode #%lu: block %llu: comm %s: %pV\n",
477 inode->i_sb->s_id, function, line, inode->i_ino,
478 block, current->comm, &vaf);
479 else
480 printk(KERN_CRIT "EXT4-fs error (device %s): %s:%d: "
481 "inode #%lu: comm %s: %pV\n",
482 inode->i_sb->s_id, function, line, inode->i_ino,
483 current->comm, &vaf);
484 va_end(args);
486 save_error_info(inode->i_sb, function, line);
487 ext4_handle_error(inode->i_sb);
490 void __ext4_error_file(struct file *file, const char *function,
491 unsigned int line, ext4_fsblk_t block,
492 const char *fmt, ...)
494 va_list args;
495 struct va_format vaf;
496 struct ext4_super_block *es;
497 struct inode *inode = file_inode(file);
498 char pathname[80], *path;
500 if (unlikely(ext4_forced_shutdown(EXT4_SB(inode->i_sb))))
501 return;
503 es = EXT4_SB(inode->i_sb)->s_es;
504 es->s_last_error_ino = cpu_to_le32(inode->i_ino);
505 if (ext4_error_ratelimit(inode->i_sb)) {
506 path = file_path(file, pathname, sizeof(pathname));
507 if (IS_ERR(path))
508 path = "(unknown)";
509 va_start(args, fmt);
510 vaf.fmt = fmt;
511 vaf.va = &args;
512 if (block)
513 printk(KERN_CRIT
514 "EXT4-fs error (device %s): %s:%d: inode #%lu: "
515 "block %llu: comm %s: path %s: %pV\n",
516 inode->i_sb->s_id, function, line, inode->i_ino,
517 block, current->comm, path, &vaf);
518 else
519 printk(KERN_CRIT
520 "EXT4-fs error (device %s): %s:%d: inode #%lu: "
521 "comm %s: path %s: %pV\n",
522 inode->i_sb->s_id, function, line, inode->i_ino,
523 current->comm, path, &vaf);
524 va_end(args);
526 save_error_info(inode->i_sb, function, line);
527 ext4_handle_error(inode->i_sb);
530 const char *ext4_decode_error(struct super_block *sb, int errno,
531 char nbuf[16])
533 char *errstr = NULL;
535 switch (errno) {
536 case -EFSCORRUPTED:
537 errstr = "Corrupt filesystem";
538 break;
539 case -EFSBADCRC:
540 errstr = "Filesystem failed CRC";
541 break;
542 case -EIO:
543 errstr = "IO failure";
544 break;
545 case -ENOMEM:
546 errstr = "Out of memory";
547 break;
548 case -EROFS:
549 if (!sb || (EXT4_SB(sb)->s_journal &&
550 EXT4_SB(sb)->s_journal->j_flags & JBD2_ABORT))
551 errstr = "Journal has aborted";
552 else
553 errstr = "Readonly filesystem";
554 break;
555 default:
556 /* If the caller passed in an extra buffer for unknown
557 * errors, textualise them now. Else we just return
558 * NULL. */
559 if (nbuf) {
560 /* Check for truncated error codes... */
561 if (snprintf(nbuf, 16, "error %d", -errno) >= 0)
562 errstr = nbuf;
564 break;
567 return errstr;
570 /* __ext4_std_error decodes expected errors from journaling functions
571 * automatically and invokes the appropriate error response. */
573 void __ext4_std_error(struct super_block *sb, const char *function,
574 unsigned int line, int errno)
576 char nbuf[16];
577 const char *errstr;
579 if (unlikely(ext4_forced_shutdown(EXT4_SB(sb))))
580 return;
582 /* Special case: if the error is EROFS, and we're not already
583 * inside a transaction, then there's really no point in logging
584 * an error. */
585 if (errno == -EROFS && journal_current_handle() == NULL &&
586 (sb->s_flags & MS_RDONLY))
587 return;
589 if (ext4_error_ratelimit(sb)) {
590 errstr = ext4_decode_error(sb, errno, nbuf);
591 printk(KERN_CRIT "EXT4-fs error (device %s) in %s:%d: %s\n",
592 sb->s_id, function, line, errstr);
595 save_error_info(sb, function, line);
596 ext4_handle_error(sb);
600 * ext4_abort is a much stronger failure handler than ext4_error. The
601 * abort function may be used to deal with unrecoverable failures such
602 * as journal IO errors or ENOMEM at a critical moment in log management.
604 * We unconditionally force the filesystem into an ABORT|READONLY state,
605 * unless the error response on the fs has been set to panic in which
606 * case we take the easy way out and panic immediately.
609 void __ext4_abort(struct super_block *sb, const char *function,
610 unsigned int line, const char *fmt, ...)
612 struct va_format vaf;
613 va_list args;
615 if (unlikely(ext4_forced_shutdown(EXT4_SB(sb))))
616 return;
618 save_error_info(sb, function, line);
619 va_start(args, fmt);
620 vaf.fmt = fmt;
621 vaf.va = &args;
622 printk(KERN_CRIT "EXT4-fs error (device %s): %s:%d: %pV\n",
623 sb->s_id, function, line, &vaf);
624 va_end(args);
626 if ((sb->s_flags & MS_RDONLY) == 0) {
627 ext4_msg(sb, KERN_CRIT, "Remounting filesystem read-only");
628 EXT4_SB(sb)->s_mount_flags |= EXT4_MF_FS_ABORTED;
630 * Make sure updated value of ->s_mount_flags will be visible
631 * before ->s_flags update
633 smp_wmb();
634 sb->s_flags |= MS_RDONLY;
635 if (EXT4_SB(sb)->s_journal)
636 jbd2_journal_abort(EXT4_SB(sb)->s_journal, -EIO);
637 save_error_info(sb, function, line);
639 if (test_opt(sb, ERRORS_PANIC)) {
640 if (EXT4_SB(sb)->s_journal &&
641 !(EXT4_SB(sb)->s_journal->j_flags & JBD2_REC_ERR))
642 return;
643 panic("EXT4-fs panic from previous error\n");
647 void __ext4_msg(struct super_block *sb,
648 const char *prefix, const char *fmt, ...)
650 struct va_format vaf;
651 va_list args;
653 if (!___ratelimit(&(EXT4_SB(sb)->s_msg_ratelimit_state), "EXT4-fs"))
654 return;
656 va_start(args, fmt);
657 vaf.fmt = fmt;
658 vaf.va = &args;
659 printk("%sEXT4-fs (%s): %pV\n", prefix, sb->s_id, &vaf);
660 va_end(args);
663 #define ext4_warning_ratelimit(sb) \
664 ___ratelimit(&(EXT4_SB(sb)->s_warning_ratelimit_state), \
665 "EXT4-fs warning")
667 void __ext4_warning(struct super_block *sb, const char *function,
668 unsigned int line, const char *fmt, ...)
670 struct va_format vaf;
671 va_list args;
673 if (!ext4_warning_ratelimit(sb))
674 return;
676 va_start(args, fmt);
677 vaf.fmt = fmt;
678 vaf.va = &args;
679 printk(KERN_WARNING "EXT4-fs warning (device %s): %s:%d: %pV\n",
680 sb->s_id, function, line, &vaf);
681 va_end(args);
684 void __ext4_warning_inode(const struct inode *inode, const char *function,
685 unsigned int line, const char *fmt, ...)
687 struct va_format vaf;
688 va_list args;
690 if (!ext4_warning_ratelimit(inode->i_sb))
691 return;
693 va_start(args, fmt);
694 vaf.fmt = fmt;
695 vaf.va = &args;
696 printk(KERN_WARNING "EXT4-fs warning (device %s): %s:%d: "
697 "inode #%lu: comm %s: %pV\n", inode->i_sb->s_id,
698 function, line, inode->i_ino, current->comm, &vaf);
699 va_end(args);
702 void __ext4_grp_locked_error(const char *function, unsigned int line,
703 struct super_block *sb, ext4_group_t grp,
704 unsigned long ino, ext4_fsblk_t block,
705 const char *fmt, ...)
706 __releases(bitlock)
707 __acquires(bitlock)
709 struct va_format vaf;
710 va_list args;
711 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
713 if (unlikely(ext4_forced_shutdown(EXT4_SB(sb))))
714 return;
716 es->s_last_error_ino = cpu_to_le32(ino);
717 es->s_last_error_block = cpu_to_le64(block);
718 __save_error_info(sb, function, line);
720 if (ext4_error_ratelimit(sb)) {
721 va_start(args, fmt);
722 vaf.fmt = fmt;
723 vaf.va = &args;
724 printk(KERN_CRIT "EXT4-fs error (device %s): %s:%d: group %u, ",
725 sb->s_id, function, line, grp);
726 if (ino)
727 printk(KERN_CONT "inode %lu: ", ino);
728 if (block)
729 printk(KERN_CONT "block %llu:",
730 (unsigned long long) block);
731 printk(KERN_CONT "%pV\n", &vaf);
732 va_end(args);
735 if (test_opt(sb, ERRORS_CONT)) {
736 ext4_commit_super(sb, 0);
737 return;
740 ext4_unlock_group(sb, grp);
741 ext4_handle_error(sb);
743 * We only get here in the ERRORS_RO case; relocking the group
744 * may be dangerous, but nothing bad will happen since the
745 * filesystem will have already been marked read/only and the
746 * journal has been aborted. We return 1 as a hint to callers
747 * who might what to use the return value from
748 * ext4_grp_locked_error() to distinguish between the
749 * ERRORS_CONT and ERRORS_RO case, and perhaps return more
750 * aggressively from the ext4 function in question, with a
751 * more appropriate error code.
753 ext4_lock_group(sb, grp);
754 return;
757 void ext4_update_dynamic_rev(struct super_block *sb)
759 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
761 if (le32_to_cpu(es->s_rev_level) > EXT4_GOOD_OLD_REV)
762 return;
764 ext4_warning(sb,
765 "updating to rev %d because of new feature flag, "
766 "running e2fsck is recommended",
767 EXT4_DYNAMIC_REV);
769 es->s_first_ino = cpu_to_le32(EXT4_GOOD_OLD_FIRST_INO);
770 es->s_inode_size = cpu_to_le16(EXT4_GOOD_OLD_INODE_SIZE);
771 es->s_rev_level = cpu_to_le32(EXT4_DYNAMIC_REV);
772 /* leave es->s_feature_*compat flags alone */
773 /* es->s_uuid will be set by e2fsck if empty */
776 * The rest of the superblock fields should be zero, and if not it
777 * means they are likely already in use, so leave them alone. We
778 * can leave it up to e2fsck to clean up any inconsistencies there.
783 * Open the external journal device
785 static struct block_device *ext4_blkdev_get(dev_t dev, struct super_block *sb)
787 struct block_device *bdev;
788 char b[BDEVNAME_SIZE];
790 bdev = blkdev_get_by_dev(dev, FMODE_READ|FMODE_WRITE|FMODE_EXCL, sb);
791 if (IS_ERR(bdev))
792 goto fail;
793 return bdev;
795 fail:
796 ext4_msg(sb, KERN_ERR, "failed to open journal device %s: %ld",
797 __bdevname(dev, b), PTR_ERR(bdev));
798 return NULL;
802 * Release the journal device
804 static void ext4_blkdev_put(struct block_device *bdev)
806 blkdev_put(bdev, FMODE_READ|FMODE_WRITE|FMODE_EXCL);
809 static void ext4_blkdev_remove(struct ext4_sb_info *sbi)
811 struct block_device *bdev;
812 bdev = sbi->journal_bdev;
813 if (bdev) {
814 ext4_blkdev_put(bdev);
815 sbi->journal_bdev = NULL;
819 static inline struct inode *orphan_list_entry(struct list_head *l)
821 return &list_entry(l, struct ext4_inode_info, i_orphan)->vfs_inode;
824 static void dump_orphan_list(struct super_block *sb, struct ext4_sb_info *sbi)
826 struct list_head *l;
828 ext4_msg(sb, KERN_ERR, "sb orphan head is %d",
829 le32_to_cpu(sbi->s_es->s_last_orphan));
831 printk(KERN_ERR "sb_info orphan list:\n");
832 list_for_each(l, &sbi->s_orphan) {
833 struct inode *inode = orphan_list_entry(l);
834 printk(KERN_ERR " "
835 "inode %s:%lu at %p: mode %o, nlink %d, next %d\n",
836 inode->i_sb->s_id, inode->i_ino, inode,
837 inode->i_mode, inode->i_nlink,
838 NEXT_ORPHAN(inode));
842 static void ext4_put_super(struct super_block *sb)
844 struct ext4_sb_info *sbi = EXT4_SB(sb);
845 struct ext4_super_block *es = sbi->s_es;
846 int aborted = 0;
847 int i, err;
849 ext4_unregister_li_request(sb);
850 dquot_disable(sb, -1, DQUOT_USAGE_ENABLED | DQUOT_LIMITS_ENABLED);
852 flush_workqueue(sbi->rsv_conversion_wq);
853 destroy_workqueue(sbi->rsv_conversion_wq);
855 if (sbi->s_journal) {
856 aborted = is_journal_aborted(sbi->s_journal);
857 err = jbd2_journal_destroy(sbi->s_journal);
858 sbi->s_journal = NULL;
859 if ((err < 0) && !aborted)
860 ext4_abort(sb, "Couldn't clean up the journal");
863 ext4_unregister_sysfs(sb);
864 ext4_es_unregister_shrinker(sbi);
865 del_timer_sync(&sbi->s_err_report);
866 ext4_release_system_zone(sb);
867 ext4_mb_release(sb);
868 ext4_ext_release(sb);
870 if (!(sb->s_flags & MS_RDONLY) && !aborted) {
871 ext4_clear_feature_journal_needs_recovery(sb);
872 es->s_state = cpu_to_le16(sbi->s_mount_state);
874 if (!(sb->s_flags & MS_RDONLY))
875 ext4_commit_super(sb, 1);
877 for (i = 0; i < sbi->s_gdb_count; i++)
878 brelse(sbi->s_group_desc[i]);
879 kvfree(sbi->s_group_desc);
880 kvfree(sbi->s_flex_groups);
881 percpu_counter_destroy(&sbi->s_freeclusters_counter);
882 percpu_counter_destroy(&sbi->s_freeinodes_counter);
883 percpu_counter_destroy(&sbi->s_dirs_counter);
884 percpu_counter_destroy(&sbi->s_dirtyclusters_counter);
885 percpu_free_rwsem(&sbi->s_journal_flag_rwsem);
886 #ifdef CONFIG_QUOTA
887 for (i = 0; i < EXT4_MAXQUOTAS; i++)
888 kfree(sbi->s_qf_names[i]);
889 #endif
891 /* Debugging code just in case the in-memory inode orphan list
892 * isn't empty. The on-disk one can be non-empty if we've
893 * detected an error and taken the fs readonly, but the
894 * in-memory list had better be clean by this point. */
895 if (!list_empty(&sbi->s_orphan))
896 dump_orphan_list(sb, sbi);
897 J_ASSERT(list_empty(&sbi->s_orphan));
899 sync_blockdev(sb->s_bdev);
900 invalidate_bdev(sb->s_bdev);
901 if (sbi->journal_bdev && sbi->journal_bdev != sb->s_bdev) {
903 * Invalidate the journal device's buffers. We don't want them
904 * floating about in memory - the physical journal device may
905 * hotswapped, and it breaks the `ro-after' testing code.
907 sync_blockdev(sbi->journal_bdev);
908 invalidate_bdev(sbi->journal_bdev);
909 ext4_blkdev_remove(sbi);
911 if (sbi->s_mb_cache) {
912 ext4_xattr_destroy_cache(sbi->s_mb_cache);
913 sbi->s_mb_cache = NULL;
915 if (sbi->s_mmp_tsk)
916 kthread_stop(sbi->s_mmp_tsk);
917 brelse(sbi->s_sbh);
918 sb->s_fs_info = NULL;
920 * Now that we are completely done shutting down the
921 * superblock, we need to actually destroy the kobject.
923 kobject_put(&sbi->s_kobj);
924 wait_for_completion(&sbi->s_kobj_unregister);
925 if (sbi->s_chksum_driver)
926 crypto_free_shash(sbi->s_chksum_driver);
927 kfree(sbi->s_blockgroup_lock);
928 kfree(sbi);
931 static struct kmem_cache *ext4_inode_cachep;
934 * Called inside transaction, so use GFP_NOFS
936 static struct inode *ext4_alloc_inode(struct super_block *sb)
938 struct ext4_inode_info *ei;
940 ei = kmem_cache_alloc(ext4_inode_cachep, GFP_NOFS);
941 if (!ei)
942 return NULL;
944 ei->vfs_inode.i_version = 1;
945 spin_lock_init(&ei->i_raw_lock);
946 INIT_LIST_HEAD(&ei->i_prealloc_list);
947 spin_lock_init(&ei->i_prealloc_lock);
948 ext4_es_init_tree(&ei->i_es_tree);
949 rwlock_init(&ei->i_es_lock);
950 INIT_LIST_HEAD(&ei->i_es_list);
951 ei->i_es_all_nr = 0;
952 ei->i_es_shk_nr = 0;
953 ei->i_es_shrink_lblk = 0;
954 ei->i_reserved_data_blocks = 0;
955 ei->i_reserved_meta_blocks = 0;
956 ei->i_allocated_meta_blocks = 0;
957 ei->i_da_metadata_calc_len = 0;
958 ei->i_da_metadata_calc_last_lblock = 0;
959 spin_lock_init(&(ei->i_block_reservation_lock));
960 #ifdef CONFIG_QUOTA
961 ei->i_reserved_quota = 0;
962 memset(&ei->i_dquot, 0, sizeof(ei->i_dquot));
963 #endif
964 ei->jinode = NULL;
965 INIT_LIST_HEAD(&ei->i_rsv_conversion_list);
966 spin_lock_init(&ei->i_completed_io_lock);
967 ei->i_sync_tid = 0;
968 ei->i_datasync_tid = 0;
969 atomic_set(&ei->i_unwritten, 0);
970 INIT_WORK(&ei->i_rsv_conversion_work, ext4_end_io_rsv_work);
971 return &ei->vfs_inode;
974 static int ext4_drop_inode(struct inode *inode)
976 int drop = generic_drop_inode(inode);
978 trace_ext4_drop_inode(inode, drop);
979 return drop;
982 static void ext4_i_callback(struct rcu_head *head)
984 struct inode *inode = container_of(head, struct inode, i_rcu);
985 kmem_cache_free(ext4_inode_cachep, EXT4_I(inode));
988 static void ext4_destroy_inode(struct inode *inode)
990 if (!list_empty(&(EXT4_I(inode)->i_orphan))) {
991 ext4_msg(inode->i_sb, KERN_ERR,
992 "Inode %lu (%p): orphan list check failed!",
993 inode->i_ino, EXT4_I(inode));
994 print_hex_dump(KERN_INFO, "", DUMP_PREFIX_ADDRESS, 16, 4,
995 EXT4_I(inode), sizeof(struct ext4_inode_info),
996 true);
997 dump_stack();
999 call_rcu(&inode->i_rcu, ext4_i_callback);
1002 static void init_once(void *foo)
1004 struct ext4_inode_info *ei = (struct ext4_inode_info *) foo;
1006 INIT_LIST_HEAD(&ei->i_orphan);
1007 init_rwsem(&ei->xattr_sem);
1008 init_rwsem(&ei->i_data_sem);
1009 init_rwsem(&ei->i_mmap_sem);
1010 inode_init_once(&ei->vfs_inode);
1013 static int __init init_inodecache(void)
1015 ext4_inode_cachep = kmem_cache_create("ext4_inode_cache",
1016 sizeof(struct ext4_inode_info),
1017 0, (SLAB_RECLAIM_ACCOUNT|
1018 SLAB_MEM_SPREAD|SLAB_ACCOUNT),
1019 init_once);
1020 if (ext4_inode_cachep == NULL)
1021 return -ENOMEM;
1022 return 0;
1025 static void destroy_inodecache(void)
1028 * Make sure all delayed rcu free inodes are flushed before we
1029 * destroy cache.
1031 rcu_barrier();
1032 kmem_cache_destroy(ext4_inode_cachep);
1035 void ext4_clear_inode(struct inode *inode)
1037 invalidate_inode_buffers(inode);
1038 clear_inode(inode);
1039 dquot_drop(inode);
1040 ext4_discard_preallocations(inode);
1041 ext4_es_remove_extent(inode, 0, EXT_MAX_BLOCKS);
1042 if (EXT4_I(inode)->jinode) {
1043 jbd2_journal_release_jbd_inode(EXT4_JOURNAL(inode),
1044 EXT4_I(inode)->jinode);
1045 jbd2_free_inode(EXT4_I(inode)->jinode);
1046 EXT4_I(inode)->jinode = NULL;
1048 #ifdef CONFIG_EXT4_FS_ENCRYPTION
1049 fscrypt_put_encryption_info(inode, NULL);
1050 #endif
1053 static struct inode *ext4_nfs_get_inode(struct super_block *sb,
1054 u64 ino, u32 generation)
1056 struct inode *inode;
1058 if (ino < EXT4_FIRST_INO(sb) && ino != EXT4_ROOT_INO)
1059 return ERR_PTR(-ESTALE);
1060 if (ino > le32_to_cpu(EXT4_SB(sb)->s_es->s_inodes_count))
1061 return ERR_PTR(-ESTALE);
1063 /* iget isn't really right if the inode is currently unallocated!!
1065 * ext4_read_inode will return a bad_inode if the inode had been
1066 * deleted, so we should be safe.
1068 * Currently we don't know the generation for parent directory, so
1069 * a generation of 0 means "accept any"
1071 inode = ext4_iget_normal(sb, ino);
1072 if (IS_ERR(inode))
1073 return ERR_CAST(inode);
1074 if (generation && inode->i_generation != generation) {
1075 iput(inode);
1076 return ERR_PTR(-ESTALE);
1079 return inode;
1082 static struct dentry *ext4_fh_to_dentry(struct super_block *sb, struct fid *fid,
1083 int fh_len, int fh_type)
1085 return generic_fh_to_dentry(sb, fid, fh_len, fh_type,
1086 ext4_nfs_get_inode);
1089 static struct dentry *ext4_fh_to_parent(struct super_block *sb, struct fid *fid,
1090 int fh_len, int fh_type)
1092 return generic_fh_to_parent(sb, fid, fh_len, fh_type,
1093 ext4_nfs_get_inode);
1097 * Try to release metadata pages (indirect blocks, directories) which are
1098 * mapped via the block device. Since these pages could have journal heads
1099 * which would prevent try_to_free_buffers() from freeing them, we must use
1100 * jbd2 layer's try_to_free_buffers() function to release them.
1102 static int bdev_try_to_free_page(struct super_block *sb, struct page *page,
1103 gfp_t wait)
1105 journal_t *journal = EXT4_SB(sb)->s_journal;
1107 WARN_ON(PageChecked(page));
1108 if (!page_has_buffers(page))
1109 return 0;
1110 if (journal)
1111 return jbd2_journal_try_to_free_buffers(journal, page,
1112 wait & ~__GFP_DIRECT_RECLAIM);
1113 return try_to_free_buffers(page);
1116 #ifdef CONFIG_EXT4_FS_ENCRYPTION
1117 static int ext4_get_context(struct inode *inode, void *ctx, size_t len)
1119 return ext4_xattr_get(inode, EXT4_XATTR_INDEX_ENCRYPTION,
1120 EXT4_XATTR_NAME_ENCRYPTION_CONTEXT, ctx, len);
1123 static int ext4_set_context(struct inode *inode, const void *ctx, size_t len,
1124 void *fs_data)
1126 handle_t *handle = fs_data;
1127 int res, res2, retries = 0;
1129 res = ext4_convert_inline_data(inode);
1130 if (res)
1131 return res;
1134 * If a journal handle was specified, then the encryption context is
1135 * being set on a new inode via inheritance and is part of a larger
1136 * transaction to create the inode. Otherwise the encryption context is
1137 * being set on an existing inode in its own transaction. Only in the
1138 * latter case should the "retry on ENOSPC" logic be used.
1141 if (handle) {
1142 res = ext4_xattr_set_handle(handle, inode,
1143 EXT4_XATTR_INDEX_ENCRYPTION,
1144 EXT4_XATTR_NAME_ENCRYPTION_CONTEXT,
1145 ctx, len, 0);
1146 if (!res) {
1147 ext4_set_inode_flag(inode, EXT4_INODE_ENCRYPT);
1148 ext4_clear_inode_state(inode,
1149 EXT4_STATE_MAY_INLINE_DATA);
1151 * Update inode->i_flags - e.g. S_DAX may get disabled
1153 ext4_set_inode_flags(inode);
1155 return res;
1158 retry:
1159 handle = ext4_journal_start(inode, EXT4_HT_MISC,
1160 ext4_jbd2_credits_xattr(inode));
1161 if (IS_ERR(handle))
1162 return PTR_ERR(handle);
1164 res = ext4_xattr_set_handle(handle, inode, EXT4_XATTR_INDEX_ENCRYPTION,
1165 EXT4_XATTR_NAME_ENCRYPTION_CONTEXT,
1166 ctx, len, 0);
1167 if (!res) {
1168 ext4_set_inode_flag(inode, EXT4_INODE_ENCRYPT);
1169 /* Update inode->i_flags - e.g. S_DAX may get disabled */
1170 ext4_set_inode_flags(inode);
1171 res = ext4_mark_inode_dirty(handle, inode);
1172 if (res)
1173 EXT4_ERROR_INODE(inode, "Failed to mark inode dirty");
1175 res2 = ext4_journal_stop(handle);
1177 if (res == -ENOSPC && ext4_should_retry_alloc(inode->i_sb, &retries))
1178 goto retry;
1179 if (!res)
1180 res = res2;
1181 return res;
1184 static int ext4_dummy_context(struct inode *inode)
1186 return DUMMY_ENCRYPTION_ENABLED(EXT4_SB(inode->i_sb));
1189 static unsigned ext4_max_namelen(struct inode *inode)
1191 return S_ISLNK(inode->i_mode) ? inode->i_sb->s_blocksize :
1192 EXT4_NAME_LEN;
1195 static const struct fscrypt_operations ext4_cryptops = {
1196 .key_prefix = "ext4:",
1197 .get_context = ext4_get_context,
1198 .set_context = ext4_set_context,
1199 .dummy_context = ext4_dummy_context,
1200 .is_encrypted = ext4_encrypted_inode,
1201 .empty_dir = ext4_empty_dir,
1202 .max_namelen = ext4_max_namelen,
1204 #else
1205 static const struct fscrypt_operations ext4_cryptops = {
1206 .is_encrypted = ext4_encrypted_inode,
1208 #endif
1210 #ifdef CONFIG_QUOTA
1211 static char *quotatypes[] = INITQFNAMES;
1212 #define QTYPE2NAME(t) (quotatypes[t])
1214 static int ext4_write_dquot(struct dquot *dquot);
1215 static int ext4_acquire_dquot(struct dquot *dquot);
1216 static int ext4_release_dquot(struct dquot *dquot);
1217 static int ext4_mark_dquot_dirty(struct dquot *dquot);
1218 static int ext4_write_info(struct super_block *sb, int type);
1219 static int ext4_quota_on(struct super_block *sb, int type, int format_id,
1220 const struct path *path);
1221 static int ext4_quota_off(struct super_block *sb, int type);
1222 static int ext4_quota_on_mount(struct super_block *sb, int type);
1223 static ssize_t ext4_quota_read(struct super_block *sb, int type, char *data,
1224 size_t len, loff_t off);
1225 static ssize_t ext4_quota_write(struct super_block *sb, int type,
1226 const char *data, size_t len, loff_t off);
1227 static int ext4_quota_enable(struct super_block *sb, int type, int format_id,
1228 unsigned int flags);
1229 static int ext4_enable_quotas(struct super_block *sb);
1230 static int ext4_get_next_id(struct super_block *sb, struct kqid *qid);
1232 static struct dquot **ext4_get_dquots(struct inode *inode)
1234 return EXT4_I(inode)->i_dquot;
1237 static const struct dquot_operations ext4_quota_operations = {
1238 .get_reserved_space = ext4_get_reserved_space,
1239 .write_dquot = ext4_write_dquot,
1240 .acquire_dquot = ext4_acquire_dquot,
1241 .release_dquot = ext4_release_dquot,
1242 .mark_dirty = ext4_mark_dquot_dirty,
1243 .write_info = ext4_write_info,
1244 .alloc_dquot = dquot_alloc,
1245 .destroy_dquot = dquot_destroy,
1246 .get_projid = ext4_get_projid,
1247 .get_next_id = ext4_get_next_id,
1250 static const struct quotactl_ops ext4_qctl_operations = {
1251 .quota_on = ext4_quota_on,
1252 .quota_off = ext4_quota_off,
1253 .quota_sync = dquot_quota_sync,
1254 .get_state = dquot_get_state,
1255 .set_info = dquot_set_dqinfo,
1256 .get_dqblk = dquot_get_dqblk,
1257 .set_dqblk = dquot_set_dqblk,
1258 .get_nextdqblk = dquot_get_next_dqblk,
1260 #endif
1262 static const struct super_operations ext4_sops = {
1263 .alloc_inode = ext4_alloc_inode,
1264 .destroy_inode = ext4_destroy_inode,
1265 .write_inode = ext4_write_inode,
1266 .dirty_inode = ext4_dirty_inode,
1267 .drop_inode = ext4_drop_inode,
1268 .evict_inode = ext4_evict_inode,
1269 .put_super = ext4_put_super,
1270 .sync_fs = ext4_sync_fs,
1271 .freeze_fs = ext4_freeze,
1272 .unfreeze_fs = ext4_unfreeze,
1273 .statfs = ext4_statfs,
1274 .remount_fs = ext4_remount,
1275 .show_options = ext4_show_options,
1276 #ifdef CONFIG_QUOTA
1277 .quota_read = ext4_quota_read,
1278 .quota_write = ext4_quota_write,
1279 .get_dquots = ext4_get_dquots,
1280 #endif
1281 .bdev_try_to_free_page = bdev_try_to_free_page,
1284 static const struct export_operations ext4_export_ops = {
1285 .fh_to_dentry = ext4_fh_to_dentry,
1286 .fh_to_parent = ext4_fh_to_parent,
1287 .get_parent = ext4_get_parent,
1290 enum {
1291 Opt_bsd_df, Opt_minix_df, Opt_grpid, Opt_nogrpid,
1292 Opt_resgid, Opt_resuid, Opt_sb, Opt_err_cont, Opt_err_panic, Opt_err_ro,
1293 Opt_nouid32, Opt_debug, Opt_removed,
1294 Opt_user_xattr, Opt_nouser_xattr, Opt_acl, Opt_noacl,
1295 Opt_auto_da_alloc, Opt_noauto_da_alloc, Opt_noload,
1296 Opt_commit, Opt_min_batch_time, Opt_max_batch_time, Opt_journal_dev,
1297 Opt_journal_path, Opt_journal_checksum, Opt_journal_async_commit,
1298 Opt_abort, Opt_data_journal, Opt_data_ordered, Opt_data_writeback,
1299 Opt_data_err_abort, Opt_data_err_ignore, Opt_test_dummy_encryption,
1300 Opt_usrjquota, Opt_grpjquota, Opt_offusrjquota, Opt_offgrpjquota,
1301 Opt_jqfmt_vfsold, Opt_jqfmt_vfsv0, Opt_jqfmt_vfsv1, Opt_quota,
1302 Opt_noquota, Opt_barrier, Opt_nobarrier, Opt_err,
1303 Opt_usrquota, Opt_grpquota, Opt_prjquota, Opt_i_version, Opt_dax,
1304 Opt_stripe, Opt_delalloc, Opt_nodelalloc, Opt_mblk_io_submit,
1305 Opt_lazytime, Opt_nolazytime, Opt_debug_want_extra_isize,
1306 Opt_nomblk_io_submit, Opt_block_validity, Opt_noblock_validity,
1307 Opt_inode_readahead_blks, Opt_journal_ioprio,
1308 Opt_dioread_nolock, Opt_dioread_lock,
1309 Opt_discard, Opt_nodiscard, Opt_init_itable, Opt_noinit_itable,
1310 Opt_max_dir_size_kb, Opt_nojournal_checksum,
1313 static const match_table_t tokens = {
1314 {Opt_bsd_df, "bsddf"},
1315 {Opt_minix_df, "minixdf"},
1316 {Opt_grpid, "grpid"},
1317 {Opt_grpid, "bsdgroups"},
1318 {Opt_nogrpid, "nogrpid"},
1319 {Opt_nogrpid, "sysvgroups"},
1320 {Opt_resgid, "resgid=%u"},
1321 {Opt_resuid, "resuid=%u"},
1322 {Opt_sb, "sb=%u"},
1323 {Opt_err_cont, "errors=continue"},
1324 {Opt_err_panic, "errors=panic"},
1325 {Opt_err_ro, "errors=remount-ro"},
1326 {Opt_nouid32, "nouid32"},
1327 {Opt_debug, "debug"},
1328 {Opt_removed, "oldalloc"},
1329 {Opt_removed, "orlov"},
1330 {Opt_user_xattr, "user_xattr"},
1331 {Opt_nouser_xattr, "nouser_xattr"},
1332 {Opt_acl, "acl"},
1333 {Opt_noacl, "noacl"},
1334 {Opt_noload, "norecovery"},
1335 {Opt_noload, "noload"},
1336 {Opt_removed, "nobh"},
1337 {Opt_removed, "bh"},
1338 {Opt_commit, "commit=%u"},
1339 {Opt_min_batch_time, "min_batch_time=%u"},
1340 {Opt_max_batch_time, "max_batch_time=%u"},
1341 {Opt_journal_dev, "journal_dev=%u"},
1342 {Opt_journal_path, "journal_path=%s"},
1343 {Opt_journal_checksum, "journal_checksum"},
1344 {Opt_nojournal_checksum, "nojournal_checksum"},
1345 {Opt_journal_async_commit, "journal_async_commit"},
1346 {Opt_abort, "abort"},
1347 {Opt_data_journal, "data=journal"},
1348 {Opt_data_ordered, "data=ordered"},
1349 {Opt_data_writeback, "data=writeback"},
1350 {Opt_data_err_abort, "data_err=abort"},
1351 {Opt_data_err_ignore, "data_err=ignore"},
1352 {Opt_offusrjquota, "usrjquota="},
1353 {Opt_usrjquota, "usrjquota=%s"},
1354 {Opt_offgrpjquota, "grpjquota="},
1355 {Opt_grpjquota, "grpjquota=%s"},
1356 {Opt_jqfmt_vfsold, "jqfmt=vfsold"},
1357 {Opt_jqfmt_vfsv0, "jqfmt=vfsv0"},
1358 {Opt_jqfmt_vfsv1, "jqfmt=vfsv1"},
1359 {Opt_grpquota, "grpquota"},
1360 {Opt_noquota, "noquota"},
1361 {Opt_quota, "quota"},
1362 {Opt_usrquota, "usrquota"},
1363 {Opt_prjquota, "prjquota"},
1364 {Opt_barrier, "barrier=%u"},
1365 {Opt_barrier, "barrier"},
1366 {Opt_nobarrier, "nobarrier"},
1367 {Opt_i_version, "i_version"},
1368 {Opt_dax, "dax"},
1369 {Opt_stripe, "stripe=%u"},
1370 {Opt_delalloc, "delalloc"},
1371 {Opt_lazytime, "lazytime"},
1372 {Opt_nolazytime, "nolazytime"},
1373 {Opt_debug_want_extra_isize, "debug_want_extra_isize=%u"},
1374 {Opt_nodelalloc, "nodelalloc"},
1375 {Opt_removed, "mblk_io_submit"},
1376 {Opt_removed, "nomblk_io_submit"},
1377 {Opt_block_validity, "block_validity"},
1378 {Opt_noblock_validity, "noblock_validity"},
1379 {Opt_inode_readahead_blks, "inode_readahead_blks=%u"},
1380 {Opt_journal_ioprio, "journal_ioprio=%u"},
1381 {Opt_auto_da_alloc, "auto_da_alloc=%u"},
1382 {Opt_auto_da_alloc, "auto_da_alloc"},
1383 {Opt_noauto_da_alloc, "noauto_da_alloc"},
1384 {Opt_dioread_nolock, "dioread_nolock"},
1385 {Opt_dioread_lock, "dioread_lock"},
1386 {Opt_discard, "discard"},
1387 {Opt_nodiscard, "nodiscard"},
1388 {Opt_init_itable, "init_itable=%u"},
1389 {Opt_init_itable, "init_itable"},
1390 {Opt_noinit_itable, "noinit_itable"},
1391 {Opt_max_dir_size_kb, "max_dir_size_kb=%u"},
1392 {Opt_test_dummy_encryption, "test_dummy_encryption"},
1393 {Opt_removed, "check=none"}, /* mount option from ext2/3 */
1394 {Opt_removed, "nocheck"}, /* mount option from ext2/3 */
1395 {Opt_removed, "reservation"}, /* mount option from ext2/3 */
1396 {Opt_removed, "noreservation"}, /* mount option from ext2/3 */
1397 {Opt_removed, "journal=%u"}, /* mount option from ext2/3 */
1398 {Opt_err, NULL},
1401 static ext4_fsblk_t get_sb_block(void **data)
1403 ext4_fsblk_t sb_block;
1404 char *options = (char *) *data;
1406 if (!options || strncmp(options, "sb=", 3) != 0)
1407 return 1; /* Default location */
1409 options += 3;
1410 /* TODO: use simple_strtoll with >32bit ext4 */
1411 sb_block = simple_strtoul(options, &options, 0);
1412 if (*options && *options != ',') {
1413 printk(KERN_ERR "EXT4-fs: Invalid sb specification: %s\n",
1414 (char *) *data);
1415 return 1;
1417 if (*options == ',')
1418 options++;
1419 *data = (void *) options;
1421 return sb_block;
1424 #define DEFAULT_JOURNAL_IOPRIO (IOPRIO_PRIO_VALUE(IOPRIO_CLASS_BE, 3))
1425 static char deprecated_msg[] = "Mount option \"%s\" will be removed by %s\n"
1426 "Contact linux-ext4@vger.kernel.org if you think we should keep it.\n";
1428 #ifdef CONFIG_QUOTA
1429 static int set_qf_name(struct super_block *sb, int qtype, substring_t *args)
1431 struct ext4_sb_info *sbi = EXT4_SB(sb);
1432 char *qname;
1433 int ret = -1;
1435 if (sb_any_quota_loaded(sb) &&
1436 !sbi->s_qf_names[qtype]) {
1437 ext4_msg(sb, KERN_ERR,
1438 "Cannot change journaled "
1439 "quota options when quota turned on");
1440 return -1;
1442 if (ext4_has_feature_quota(sb)) {
1443 ext4_msg(sb, KERN_INFO, "Journaled quota options "
1444 "ignored when QUOTA feature is enabled");
1445 return 1;
1447 qname = match_strdup(args);
1448 if (!qname) {
1449 ext4_msg(sb, KERN_ERR,
1450 "Not enough memory for storing quotafile name");
1451 return -1;
1453 if (sbi->s_qf_names[qtype]) {
1454 if (strcmp(sbi->s_qf_names[qtype], qname) == 0)
1455 ret = 1;
1456 else
1457 ext4_msg(sb, KERN_ERR,
1458 "%s quota file already specified",
1459 QTYPE2NAME(qtype));
1460 goto errout;
1462 if (strchr(qname, '/')) {
1463 ext4_msg(sb, KERN_ERR,
1464 "quotafile must be on filesystem root");
1465 goto errout;
1467 sbi->s_qf_names[qtype] = qname;
1468 set_opt(sb, QUOTA);
1469 return 1;
1470 errout:
1471 kfree(qname);
1472 return ret;
1475 static int clear_qf_name(struct super_block *sb, int qtype)
1478 struct ext4_sb_info *sbi = EXT4_SB(sb);
1480 if (sb_any_quota_loaded(sb) &&
1481 sbi->s_qf_names[qtype]) {
1482 ext4_msg(sb, KERN_ERR, "Cannot change journaled quota options"
1483 " when quota turned on");
1484 return -1;
1486 kfree(sbi->s_qf_names[qtype]);
1487 sbi->s_qf_names[qtype] = NULL;
1488 return 1;
1490 #endif
1492 #define MOPT_SET 0x0001
1493 #define MOPT_CLEAR 0x0002
1494 #define MOPT_NOSUPPORT 0x0004
1495 #define MOPT_EXPLICIT 0x0008
1496 #define MOPT_CLEAR_ERR 0x0010
1497 #define MOPT_GTE0 0x0020
1498 #ifdef CONFIG_QUOTA
1499 #define MOPT_Q 0
1500 #define MOPT_QFMT 0x0040
1501 #else
1502 #define MOPT_Q MOPT_NOSUPPORT
1503 #define MOPT_QFMT MOPT_NOSUPPORT
1504 #endif
1505 #define MOPT_DATAJ 0x0080
1506 #define MOPT_NO_EXT2 0x0100
1507 #define MOPT_NO_EXT3 0x0200
1508 #define MOPT_EXT4_ONLY (MOPT_NO_EXT2 | MOPT_NO_EXT3)
1509 #define MOPT_STRING 0x0400
1511 static const struct mount_opts {
1512 int token;
1513 int mount_opt;
1514 int flags;
1515 } ext4_mount_opts[] = {
1516 {Opt_minix_df, EXT4_MOUNT_MINIX_DF, MOPT_SET},
1517 {Opt_bsd_df, EXT4_MOUNT_MINIX_DF, MOPT_CLEAR},
1518 {Opt_grpid, EXT4_MOUNT_GRPID, MOPT_SET},
1519 {Opt_nogrpid, EXT4_MOUNT_GRPID, MOPT_CLEAR},
1520 {Opt_block_validity, EXT4_MOUNT_BLOCK_VALIDITY, MOPT_SET},
1521 {Opt_noblock_validity, EXT4_MOUNT_BLOCK_VALIDITY, MOPT_CLEAR},
1522 {Opt_dioread_nolock, EXT4_MOUNT_DIOREAD_NOLOCK,
1523 MOPT_EXT4_ONLY | MOPT_SET},
1524 {Opt_dioread_lock, EXT4_MOUNT_DIOREAD_NOLOCK,
1525 MOPT_EXT4_ONLY | MOPT_CLEAR},
1526 {Opt_discard, EXT4_MOUNT_DISCARD, MOPT_SET},
1527 {Opt_nodiscard, EXT4_MOUNT_DISCARD, MOPT_CLEAR},
1528 {Opt_delalloc, EXT4_MOUNT_DELALLOC,
1529 MOPT_EXT4_ONLY | MOPT_SET | MOPT_EXPLICIT},
1530 {Opt_nodelalloc, EXT4_MOUNT_DELALLOC,
1531 MOPT_EXT4_ONLY | MOPT_CLEAR},
1532 {Opt_nojournal_checksum, EXT4_MOUNT_JOURNAL_CHECKSUM,
1533 MOPT_EXT4_ONLY | MOPT_CLEAR},
1534 {Opt_journal_checksum, EXT4_MOUNT_JOURNAL_CHECKSUM,
1535 MOPT_EXT4_ONLY | MOPT_SET | MOPT_EXPLICIT},
1536 {Opt_journal_async_commit, (EXT4_MOUNT_JOURNAL_ASYNC_COMMIT |
1537 EXT4_MOUNT_JOURNAL_CHECKSUM),
1538 MOPT_EXT4_ONLY | MOPT_SET | MOPT_EXPLICIT},
1539 {Opt_noload, EXT4_MOUNT_NOLOAD, MOPT_NO_EXT2 | MOPT_SET},
1540 {Opt_err_panic, EXT4_MOUNT_ERRORS_PANIC, MOPT_SET | MOPT_CLEAR_ERR},
1541 {Opt_err_ro, EXT4_MOUNT_ERRORS_RO, MOPT_SET | MOPT_CLEAR_ERR},
1542 {Opt_err_cont, EXT4_MOUNT_ERRORS_CONT, MOPT_SET | MOPT_CLEAR_ERR},
1543 {Opt_data_err_abort, EXT4_MOUNT_DATA_ERR_ABORT,
1544 MOPT_NO_EXT2},
1545 {Opt_data_err_ignore, EXT4_MOUNT_DATA_ERR_ABORT,
1546 MOPT_NO_EXT2},
1547 {Opt_barrier, EXT4_MOUNT_BARRIER, MOPT_SET},
1548 {Opt_nobarrier, EXT4_MOUNT_BARRIER, MOPT_CLEAR},
1549 {Opt_noauto_da_alloc, EXT4_MOUNT_NO_AUTO_DA_ALLOC, MOPT_SET},
1550 {Opt_auto_da_alloc, EXT4_MOUNT_NO_AUTO_DA_ALLOC, MOPT_CLEAR},
1551 {Opt_noinit_itable, EXT4_MOUNT_INIT_INODE_TABLE, MOPT_CLEAR},
1552 {Opt_commit, 0, MOPT_GTE0},
1553 {Opt_max_batch_time, 0, MOPT_GTE0},
1554 {Opt_min_batch_time, 0, MOPT_GTE0},
1555 {Opt_inode_readahead_blks, 0, MOPT_GTE0},
1556 {Opt_init_itable, 0, MOPT_GTE0},
1557 {Opt_dax, EXT4_MOUNT_DAX, MOPT_SET},
1558 {Opt_stripe, 0, MOPT_GTE0},
1559 {Opt_resuid, 0, MOPT_GTE0},
1560 {Opt_resgid, 0, MOPT_GTE0},
1561 {Opt_journal_dev, 0, MOPT_NO_EXT2 | MOPT_GTE0},
1562 {Opt_journal_path, 0, MOPT_NO_EXT2 | MOPT_STRING},
1563 {Opt_journal_ioprio, 0, MOPT_NO_EXT2 | MOPT_GTE0},
1564 {Opt_data_journal, EXT4_MOUNT_JOURNAL_DATA, MOPT_NO_EXT2 | MOPT_DATAJ},
1565 {Opt_data_ordered, EXT4_MOUNT_ORDERED_DATA, MOPT_NO_EXT2 | MOPT_DATAJ},
1566 {Opt_data_writeback, EXT4_MOUNT_WRITEBACK_DATA,
1567 MOPT_NO_EXT2 | MOPT_DATAJ},
1568 {Opt_user_xattr, EXT4_MOUNT_XATTR_USER, MOPT_SET},
1569 {Opt_nouser_xattr, EXT4_MOUNT_XATTR_USER, MOPT_CLEAR},
1570 #ifdef CONFIG_EXT4_FS_POSIX_ACL
1571 {Opt_acl, EXT4_MOUNT_POSIX_ACL, MOPT_SET},
1572 {Opt_noacl, EXT4_MOUNT_POSIX_ACL, MOPT_CLEAR},
1573 #else
1574 {Opt_acl, 0, MOPT_NOSUPPORT},
1575 {Opt_noacl, 0, MOPT_NOSUPPORT},
1576 #endif
1577 {Opt_nouid32, EXT4_MOUNT_NO_UID32, MOPT_SET},
1578 {Opt_debug, EXT4_MOUNT_DEBUG, MOPT_SET},
1579 {Opt_debug_want_extra_isize, 0, MOPT_GTE0},
1580 {Opt_quota, EXT4_MOUNT_QUOTA | EXT4_MOUNT_USRQUOTA, MOPT_SET | MOPT_Q},
1581 {Opt_usrquota, EXT4_MOUNT_QUOTA | EXT4_MOUNT_USRQUOTA,
1582 MOPT_SET | MOPT_Q},
1583 {Opt_grpquota, EXT4_MOUNT_QUOTA | EXT4_MOUNT_GRPQUOTA,
1584 MOPT_SET | MOPT_Q},
1585 {Opt_prjquota, EXT4_MOUNT_QUOTA | EXT4_MOUNT_PRJQUOTA,
1586 MOPT_SET | MOPT_Q},
1587 {Opt_noquota, (EXT4_MOUNT_QUOTA | EXT4_MOUNT_USRQUOTA |
1588 EXT4_MOUNT_GRPQUOTA | EXT4_MOUNT_PRJQUOTA),
1589 MOPT_CLEAR | MOPT_Q},
1590 {Opt_usrjquota, 0, MOPT_Q},
1591 {Opt_grpjquota, 0, MOPT_Q},
1592 {Opt_offusrjquota, 0, MOPT_Q},
1593 {Opt_offgrpjquota, 0, MOPT_Q},
1594 {Opt_jqfmt_vfsold, QFMT_VFS_OLD, MOPT_QFMT},
1595 {Opt_jqfmt_vfsv0, QFMT_VFS_V0, MOPT_QFMT},
1596 {Opt_jqfmt_vfsv1, QFMT_VFS_V1, MOPT_QFMT},
1597 {Opt_max_dir_size_kb, 0, MOPT_GTE0},
1598 {Opt_test_dummy_encryption, 0, MOPT_GTE0},
1599 {Opt_err, 0, 0}
1602 static int handle_mount_opt(struct super_block *sb, char *opt, int token,
1603 substring_t *args, unsigned long *journal_devnum,
1604 unsigned int *journal_ioprio, int is_remount)
1606 struct ext4_sb_info *sbi = EXT4_SB(sb);
1607 const struct mount_opts *m;
1608 kuid_t uid;
1609 kgid_t gid;
1610 int arg = 0;
1612 #ifdef CONFIG_QUOTA
1613 if (token == Opt_usrjquota)
1614 return set_qf_name(sb, USRQUOTA, &args[0]);
1615 else if (token == Opt_grpjquota)
1616 return set_qf_name(sb, GRPQUOTA, &args[0]);
1617 else if (token == Opt_offusrjquota)
1618 return clear_qf_name(sb, USRQUOTA);
1619 else if (token == Opt_offgrpjquota)
1620 return clear_qf_name(sb, GRPQUOTA);
1621 #endif
1622 switch (token) {
1623 case Opt_noacl:
1624 case Opt_nouser_xattr:
1625 ext4_msg(sb, KERN_WARNING, deprecated_msg, opt, "3.5");
1626 break;
1627 case Opt_sb:
1628 return 1; /* handled by get_sb_block() */
1629 case Opt_removed:
1630 ext4_msg(sb, KERN_WARNING, "Ignoring removed %s option", opt);
1631 return 1;
1632 case Opt_abort:
1633 sbi->s_mount_flags |= EXT4_MF_FS_ABORTED;
1634 return 1;
1635 case Opt_i_version:
1636 sb->s_flags |= MS_I_VERSION;
1637 return 1;
1638 case Opt_lazytime:
1639 sb->s_flags |= MS_LAZYTIME;
1640 return 1;
1641 case Opt_nolazytime:
1642 sb->s_flags &= ~MS_LAZYTIME;
1643 return 1;
1646 for (m = ext4_mount_opts; m->token != Opt_err; m++)
1647 if (token == m->token)
1648 break;
1650 if (m->token == Opt_err) {
1651 ext4_msg(sb, KERN_ERR, "Unrecognized mount option \"%s\" "
1652 "or missing value", opt);
1653 return -1;
1656 if ((m->flags & MOPT_NO_EXT2) && IS_EXT2_SB(sb)) {
1657 ext4_msg(sb, KERN_ERR,
1658 "Mount option \"%s\" incompatible with ext2", opt);
1659 return -1;
1661 if ((m->flags & MOPT_NO_EXT3) && IS_EXT3_SB(sb)) {
1662 ext4_msg(sb, KERN_ERR,
1663 "Mount option \"%s\" incompatible with ext3", opt);
1664 return -1;
1667 if (args->from && !(m->flags & MOPT_STRING) && match_int(args, &arg))
1668 return -1;
1669 if (args->from && (m->flags & MOPT_GTE0) && (arg < 0))
1670 return -1;
1671 if (m->flags & MOPT_EXPLICIT) {
1672 if (m->mount_opt & EXT4_MOUNT_DELALLOC) {
1673 set_opt2(sb, EXPLICIT_DELALLOC);
1674 } else if (m->mount_opt & EXT4_MOUNT_JOURNAL_CHECKSUM) {
1675 set_opt2(sb, EXPLICIT_JOURNAL_CHECKSUM);
1676 } else
1677 return -1;
1679 if (m->flags & MOPT_CLEAR_ERR)
1680 clear_opt(sb, ERRORS_MASK);
1681 if (token == Opt_noquota && sb_any_quota_loaded(sb)) {
1682 ext4_msg(sb, KERN_ERR, "Cannot change quota "
1683 "options when quota turned on");
1684 return -1;
1687 if (m->flags & MOPT_NOSUPPORT) {
1688 ext4_msg(sb, KERN_ERR, "%s option not supported", opt);
1689 } else if (token == Opt_commit) {
1690 if (arg == 0)
1691 arg = JBD2_DEFAULT_MAX_COMMIT_AGE;
1692 sbi->s_commit_interval = HZ * arg;
1693 } else if (token == Opt_debug_want_extra_isize) {
1694 sbi->s_want_extra_isize = arg;
1695 } else if (token == Opt_max_batch_time) {
1696 sbi->s_max_batch_time = arg;
1697 } else if (token == Opt_min_batch_time) {
1698 sbi->s_min_batch_time = arg;
1699 } else if (token == Opt_inode_readahead_blks) {
1700 if (arg && (arg > (1 << 30) || !is_power_of_2(arg))) {
1701 ext4_msg(sb, KERN_ERR,
1702 "EXT4-fs: inode_readahead_blks must be "
1703 "0 or a power of 2 smaller than 2^31");
1704 return -1;
1706 sbi->s_inode_readahead_blks = arg;
1707 } else if (token == Opt_init_itable) {
1708 set_opt(sb, INIT_INODE_TABLE);
1709 if (!args->from)
1710 arg = EXT4_DEF_LI_WAIT_MULT;
1711 sbi->s_li_wait_mult = arg;
1712 } else if (token == Opt_max_dir_size_kb) {
1713 sbi->s_max_dir_size_kb = arg;
1714 } else if (token == Opt_stripe) {
1715 sbi->s_stripe = arg;
1716 } else if (token == Opt_resuid) {
1717 uid = make_kuid(current_user_ns(), arg);
1718 if (!uid_valid(uid)) {
1719 ext4_msg(sb, KERN_ERR, "Invalid uid value %d", arg);
1720 return -1;
1722 sbi->s_resuid = uid;
1723 } else if (token == Opt_resgid) {
1724 gid = make_kgid(current_user_ns(), arg);
1725 if (!gid_valid(gid)) {
1726 ext4_msg(sb, KERN_ERR, "Invalid gid value %d", arg);
1727 return -1;
1729 sbi->s_resgid = gid;
1730 } else if (token == Opt_journal_dev) {
1731 if (is_remount) {
1732 ext4_msg(sb, KERN_ERR,
1733 "Cannot specify journal on remount");
1734 return -1;
1736 *journal_devnum = arg;
1737 } else if (token == Opt_journal_path) {
1738 char *journal_path;
1739 struct inode *journal_inode;
1740 struct path path;
1741 int error;
1743 if (is_remount) {
1744 ext4_msg(sb, KERN_ERR,
1745 "Cannot specify journal on remount");
1746 return -1;
1748 journal_path = match_strdup(&args[0]);
1749 if (!journal_path) {
1750 ext4_msg(sb, KERN_ERR, "error: could not dup "
1751 "journal device string");
1752 return -1;
1755 error = kern_path(journal_path, LOOKUP_FOLLOW, &path);
1756 if (error) {
1757 ext4_msg(sb, KERN_ERR, "error: could not find "
1758 "journal device path: error %d", error);
1759 kfree(journal_path);
1760 return -1;
1763 journal_inode = d_inode(path.dentry);
1764 if (!S_ISBLK(journal_inode->i_mode)) {
1765 ext4_msg(sb, KERN_ERR, "error: journal path %s "
1766 "is not a block device", journal_path);
1767 path_put(&path);
1768 kfree(journal_path);
1769 return -1;
1772 *journal_devnum = new_encode_dev(journal_inode->i_rdev);
1773 path_put(&path);
1774 kfree(journal_path);
1775 } else if (token == Opt_journal_ioprio) {
1776 if (arg > 7) {
1777 ext4_msg(sb, KERN_ERR, "Invalid journal IO priority"
1778 " (must be 0-7)");
1779 return -1;
1781 *journal_ioprio =
1782 IOPRIO_PRIO_VALUE(IOPRIO_CLASS_BE, arg);
1783 } else if (token == Opt_test_dummy_encryption) {
1784 #ifdef CONFIG_EXT4_FS_ENCRYPTION
1785 sbi->s_mount_flags |= EXT4_MF_TEST_DUMMY_ENCRYPTION;
1786 ext4_msg(sb, KERN_WARNING,
1787 "Test dummy encryption mode enabled");
1788 #else
1789 ext4_msg(sb, KERN_WARNING,
1790 "Test dummy encryption mount option ignored");
1791 #endif
1792 } else if (m->flags & MOPT_DATAJ) {
1793 if (is_remount) {
1794 if (!sbi->s_journal)
1795 ext4_msg(sb, KERN_WARNING, "Remounting file system with no journal so ignoring journalled data option");
1796 else if (test_opt(sb, DATA_FLAGS) != m->mount_opt) {
1797 ext4_msg(sb, KERN_ERR,
1798 "Cannot change data mode on remount");
1799 return -1;
1801 } else {
1802 clear_opt(sb, DATA_FLAGS);
1803 sbi->s_mount_opt |= m->mount_opt;
1805 #ifdef CONFIG_QUOTA
1806 } else if (m->flags & MOPT_QFMT) {
1807 if (sb_any_quota_loaded(sb) &&
1808 sbi->s_jquota_fmt != m->mount_opt) {
1809 ext4_msg(sb, KERN_ERR, "Cannot change journaled "
1810 "quota options when quota turned on");
1811 return -1;
1813 if (ext4_has_feature_quota(sb)) {
1814 ext4_msg(sb, KERN_INFO,
1815 "Quota format mount options ignored "
1816 "when QUOTA feature is enabled");
1817 return 1;
1819 sbi->s_jquota_fmt = m->mount_opt;
1820 #endif
1821 } else if (token == Opt_dax) {
1822 #ifdef CONFIG_FS_DAX
1823 ext4_msg(sb, KERN_WARNING,
1824 "DAX enabled. Warning: EXPERIMENTAL, use at your own risk");
1825 sbi->s_mount_opt |= m->mount_opt;
1826 #else
1827 ext4_msg(sb, KERN_INFO, "dax option not supported");
1828 return -1;
1829 #endif
1830 } else if (token == Opt_data_err_abort) {
1831 sbi->s_mount_opt |= m->mount_opt;
1832 } else if (token == Opt_data_err_ignore) {
1833 sbi->s_mount_opt &= ~m->mount_opt;
1834 } else {
1835 if (!args->from)
1836 arg = 1;
1837 if (m->flags & MOPT_CLEAR)
1838 arg = !arg;
1839 else if (unlikely(!(m->flags & MOPT_SET))) {
1840 ext4_msg(sb, KERN_WARNING,
1841 "buggy handling of option %s", opt);
1842 WARN_ON(1);
1843 return -1;
1845 if (arg != 0)
1846 sbi->s_mount_opt |= m->mount_opt;
1847 else
1848 sbi->s_mount_opt &= ~m->mount_opt;
1850 return 1;
1853 static int parse_options(char *options, struct super_block *sb,
1854 unsigned long *journal_devnum,
1855 unsigned int *journal_ioprio,
1856 int is_remount)
1858 struct ext4_sb_info *sbi = EXT4_SB(sb);
1859 char *p;
1860 substring_t args[MAX_OPT_ARGS];
1861 int token;
1863 if (!options)
1864 return 1;
1866 while ((p = strsep(&options, ",")) != NULL) {
1867 if (!*p)
1868 continue;
1870 * Initialize args struct so we know whether arg was
1871 * found; some options take optional arguments.
1873 args[0].to = args[0].from = NULL;
1874 token = match_token(p, tokens, args);
1875 if (handle_mount_opt(sb, p, token, args, journal_devnum,
1876 journal_ioprio, is_remount) < 0)
1877 return 0;
1879 #ifdef CONFIG_QUOTA
1881 * We do the test below only for project quotas. 'usrquota' and
1882 * 'grpquota' mount options are allowed even without quota feature
1883 * to support legacy quotas in quota files.
1885 if (test_opt(sb, PRJQUOTA) && !ext4_has_feature_project(sb)) {
1886 ext4_msg(sb, KERN_ERR, "Project quota feature not enabled. "
1887 "Cannot enable project quota enforcement.");
1888 return 0;
1890 if (sbi->s_qf_names[USRQUOTA] || sbi->s_qf_names[GRPQUOTA]) {
1891 if (test_opt(sb, USRQUOTA) && sbi->s_qf_names[USRQUOTA])
1892 clear_opt(sb, USRQUOTA);
1894 if (test_opt(sb, GRPQUOTA) && sbi->s_qf_names[GRPQUOTA])
1895 clear_opt(sb, GRPQUOTA);
1897 if (test_opt(sb, GRPQUOTA) || test_opt(sb, USRQUOTA)) {
1898 ext4_msg(sb, KERN_ERR, "old and new quota "
1899 "format mixing");
1900 return 0;
1903 if (!sbi->s_jquota_fmt) {
1904 ext4_msg(sb, KERN_ERR, "journaled quota format "
1905 "not specified");
1906 return 0;
1909 #endif
1910 if (test_opt(sb, DIOREAD_NOLOCK)) {
1911 int blocksize =
1912 BLOCK_SIZE << le32_to_cpu(sbi->s_es->s_log_block_size);
1914 if (blocksize < PAGE_SIZE) {
1915 ext4_msg(sb, KERN_ERR, "can't mount with "
1916 "dioread_nolock if block size != PAGE_SIZE");
1917 return 0;
1920 return 1;
1923 static inline void ext4_show_quota_options(struct seq_file *seq,
1924 struct super_block *sb)
1926 #if defined(CONFIG_QUOTA)
1927 struct ext4_sb_info *sbi = EXT4_SB(sb);
1929 if (sbi->s_jquota_fmt) {
1930 char *fmtname = "";
1932 switch (sbi->s_jquota_fmt) {
1933 case QFMT_VFS_OLD:
1934 fmtname = "vfsold";
1935 break;
1936 case QFMT_VFS_V0:
1937 fmtname = "vfsv0";
1938 break;
1939 case QFMT_VFS_V1:
1940 fmtname = "vfsv1";
1941 break;
1943 seq_printf(seq, ",jqfmt=%s", fmtname);
1946 if (sbi->s_qf_names[USRQUOTA])
1947 seq_show_option(seq, "usrjquota", sbi->s_qf_names[USRQUOTA]);
1949 if (sbi->s_qf_names[GRPQUOTA])
1950 seq_show_option(seq, "grpjquota", sbi->s_qf_names[GRPQUOTA]);
1951 #endif
1954 static const char *token2str(int token)
1956 const struct match_token *t;
1958 for (t = tokens; t->token != Opt_err; t++)
1959 if (t->token == token && !strchr(t->pattern, '='))
1960 break;
1961 return t->pattern;
1965 * Show an option if
1966 * - it's set to a non-default value OR
1967 * - if the per-sb default is different from the global default
1969 static int _ext4_show_options(struct seq_file *seq, struct super_block *sb,
1970 int nodefs)
1972 struct ext4_sb_info *sbi = EXT4_SB(sb);
1973 struct ext4_super_block *es = sbi->s_es;
1974 int def_errors, def_mount_opt = nodefs ? 0 : sbi->s_def_mount_opt;
1975 const struct mount_opts *m;
1976 char sep = nodefs ? '\n' : ',';
1978 #define SEQ_OPTS_PUTS(str) seq_printf(seq, "%c" str, sep)
1979 #define SEQ_OPTS_PRINT(str, arg) seq_printf(seq, "%c" str, sep, arg)
1981 if (sbi->s_sb_block != 1)
1982 SEQ_OPTS_PRINT("sb=%llu", sbi->s_sb_block);
1984 for (m = ext4_mount_opts; m->token != Opt_err; m++) {
1985 int want_set = m->flags & MOPT_SET;
1986 if (((m->flags & (MOPT_SET|MOPT_CLEAR)) == 0) ||
1987 (m->flags & MOPT_CLEAR_ERR))
1988 continue;
1989 if (!(m->mount_opt & (sbi->s_mount_opt ^ def_mount_opt)))
1990 continue; /* skip if same as the default */
1991 if ((want_set &&
1992 (sbi->s_mount_opt & m->mount_opt) != m->mount_opt) ||
1993 (!want_set && (sbi->s_mount_opt & m->mount_opt)))
1994 continue; /* select Opt_noFoo vs Opt_Foo */
1995 SEQ_OPTS_PRINT("%s", token2str(m->token));
1998 if (nodefs || !uid_eq(sbi->s_resuid, make_kuid(&init_user_ns, EXT4_DEF_RESUID)) ||
1999 le16_to_cpu(es->s_def_resuid) != EXT4_DEF_RESUID)
2000 SEQ_OPTS_PRINT("resuid=%u",
2001 from_kuid_munged(&init_user_ns, sbi->s_resuid));
2002 if (nodefs || !gid_eq(sbi->s_resgid, make_kgid(&init_user_ns, EXT4_DEF_RESGID)) ||
2003 le16_to_cpu(es->s_def_resgid) != EXT4_DEF_RESGID)
2004 SEQ_OPTS_PRINT("resgid=%u",
2005 from_kgid_munged(&init_user_ns, sbi->s_resgid));
2006 def_errors = nodefs ? -1 : le16_to_cpu(es->s_errors);
2007 if (test_opt(sb, ERRORS_RO) && def_errors != EXT4_ERRORS_RO)
2008 SEQ_OPTS_PUTS("errors=remount-ro");
2009 if (test_opt(sb, ERRORS_CONT) && def_errors != EXT4_ERRORS_CONTINUE)
2010 SEQ_OPTS_PUTS("errors=continue");
2011 if (test_opt(sb, ERRORS_PANIC) && def_errors != EXT4_ERRORS_PANIC)
2012 SEQ_OPTS_PUTS("errors=panic");
2013 if (nodefs || sbi->s_commit_interval != JBD2_DEFAULT_MAX_COMMIT_AGE*HZ)
2014 SEQ_OPTS_PRINT("commit=%lu", sbi->s_commit_interval / HZ);
2015 if (nodefs || sbi->s_min_batch_time != EXT4_DEF_MIN_BATCH_TIME)
2016 SEQ_OPTS_PRINT("min_batch_time=%u", sbi->s_min_batch_time);
2017 if (nodefs || sbi->s_max_batch_time != EXT4_DEF_MAX_BATCH_TIME)
2018 SEQ_OPTS_PRINT("max_batch_time=%u", sbi->s_max_batch_time);
2019 if (sb->s_flags & MS_I_VERSION)
2020 SEQ_OPTS_PUTS("i_version");
2021 if (nodefs || sbi->s_stripe)
2022 SEQ_OPTS_PRINT("stripe=%lu", sbi->s_stripe);
2023 if (EXT4_MOUNT_DATA_FLAGS & (sbi->s_mount_opt ^ def_mount_opt)) {
2024 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA)
2025 SEQ_OPTS_PUTS("data=journal");
2026 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA)
2027 SEQ_OPTS_PUTS("data=ordered");
2028 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_WRITEBACK_DATA)
2029 SEQ_OPTS_PUTS("data=writeback");
2031 if (nodefs ||
2032 sbi->s_inode_readahead_blks != EXT4_DEF_INODE_READAHEAD_BLKS)
2033 SEQ_OPTS_PRINT("inode_readahead_blks=%u",
2034 sbi->s_inode_readahead_blks);
2036 if (nodefs || (test_opt(sb, INIT_INODE_TABLE) &&
2037 (sbi->s_li_wait_mult != EXT4_DEF_LI_WAIT_MULT)))
2038 SEQ_OPTS_PRINT("init_itable=%u", sbi->s_li_wait_mult);
2039 if (nodefs || sbi->s_max_dir_size_kb)
2040 SEQ_OPTS_PRINT("max_dir_size_kb=%u", sbi->s_max_dir_size_kb);
2041 if (test_opt(sb, DATA_ERR_ABORT))
2042 SEQ_OPTS_PUTS("data_err=abort");
2044 ext4_show_quota_options(seq, sb);
2045 return 0;
2048 static int ext4_show_options(struct seq_file *seq, struct dentry *root)
2050 return _ext4_show_options(seq, root->d_sb, 0);
2053 int ext4_seq_options_show(struct seq_file *seq, void *offset)
2055 struct super_block *sb = seq->private;
2056 int rc;
2058 seq_puts(seq, (sb->s_flags & MS_RDONLY) ? "ro" : "rw");
2059 rc = _ext4_show_options(seq, sb, 1);
2060 seq_puts(seq, "\n");
2061 return rc;
2064 static int ext4_setup_super(struct super_block *sb, struct ext4_super_block *es,
2065 int read_only)
2067 struct ext4_sb_info *sbi = EXT4_SB(sb);
2068 int res = 0;
2070 if (le32_to_cpu(es->s_rev_level) > EXT4_MAX_SUPP_REV) {
2071 ext4_msg(sb, KERN_ERR, "revision level too high, "
2072 "forcing read-only mode");
2073 res = MS_RDONLY;
2075 if (read_only)
2076 goto done;
2077 if (!(sbi->s_mount_state & EXT4_VALID_FS))
2078 ext4_msg(sb, KERN_WARNING, "warning: mounting unchecked fs, "
2079 "running e2fsck is recommended");
2080 else if (sbi->s_mount_state & EXT4_ERROR_FS)
2081 ext4_msg(sb, KERN_WARNING,
2082 "warning: mounting fs with errors, "
2083 "running e2fsck is recommended");
2084 else if ((__s16) le16_to_cpu(es->s_max_mnt_count) > 0 &&
2085 le16_to_cpu(es->s_mnt_count) >=
2086 (unsigned short) (__s16) le16_to_cpu(es->s_max_mnt_count))
2087 ext4_msg(sb, KERN_WARNING,
2088 "warning: maximal mount count reached, "
2089 "running e2fsck is recommended");
2090 else if (le32_to_cpu(es->s_checkinterval) &&
2091 (le32_to_cpu(es->s_lastcheck) +
2092 le32_to_cpu(es->s_checkinterval) <= get_seconds()))
2093 ext4_msg(sb, KERN_WARNING,
2094 "warning: checktime reached, "
2095 "running e2fsck is recommended");
2096 if (!sbi->s_journal)
2097 es->s_state &= cpu_to_le16(~EXT4_VALID_FS);
2098 if (!(__s16) le16_to_cpu(es->s_max_mnt_count))
2099 es->s_max_mnt_count = cpu_to_le16(EXT4_DFL_MAX_MNT_COUNT);
2100 le16_add_cpu(&es->s_mnt_count, 1);
2101 es->s_mtime = cpu_to_le32(get_seconds());
2102 ext4_update_dynamic_rev(sb);
2103 if (sbi->s_journal)
2104 ext4_set_feature_journal_needs_recovery(sb);
2106 ext4_commit_super(sb, 1);
2107 done:
2108 if (test_opt(sb, DEBUG))
2109 printk(KERN_INFO "[EXT4 FS bs=%lu, gc=%u, "
2110 "bpg=%lu, ipg=%lu, mo=%04x, mo2=%04x]\n",
2111 sb->s_blocksize,
2112 sbi->s_groups_count,
2113 EXT4_BLOCKS_PER_GROUP(sb),
2114 EXT4_INODES_PER_GROUP(sb),
2115 sbi->s_mount_opt, sbi->s_mount_opt2);
2117 cleancache_init_fs(sb);
2118 return res;
2121 int ext4_alloc_flex_bg_array(struct super_block *sb, ext4_group_t ngroup)
2123 struct ext4_sb_info *sbi = EXT4_SB(sb);
2124 struct flex_groups *new_groups;
2125 int size;
2127 if (!sbi->s_log_groups_per_flex)
2128 return 0;
2130 size = ext4_flex_group(sbi, ngroup - 1) + 1;
2131 if (size <= sbi->s_flex_groups_allocated)
2132 return 0;
2134 size = roundup_pow_of_two(size * sizeof(struct flex_groups));
2135 new_groups = ext4_kvzalloc(size, GFP_KERNEL);
2136 if (!new_groups) {
2137 ext4_msg(sb, KERN_ERR, "not enough memory for %d flex groups",
2138 size / (int) sizeof(struct flex_groups));
2139 return -ENOMEM;
2142 if (sbi->s_flex_groups) {
2143 memcpy(new_groups, sbi->s_flex_groups,
2144 (sbi->s_flex_groups_allocated *
2145 sizeof(struct flex_groups)));
2146 kvfree(sbi->s_flex_groups);
2148 sbi->s_flex_groups = new_groups;
2149 sbi->s_flex_groups_allocated = size / sizeof(struct flex_groups);
2150 return 0;
2153 static int ext4_fill_flex_info(struct super_block *sb)
2155 struct ext4_sb_info *sbi = EXT4_SB(sb);
2156 struct ext4_group_desc *gdp = NULL;
2157 ext4_group_t flex_group;
2158 int i, err;
2160 sbi->s_log_groups_per_flex = sbi->s_es->s_log_groups_per_flex;
2161 if (sbi->s_log_groups_per_flex < 1 || sbi->s_log_groups_per_flex > 31) {
2162 sbi->s_log_groups_per_flex = 0;
2163 return 1;
2166 err = ext4_alloc_flex_bg_array(sb, sbi->s_groups_count);
2167 if (err)
2168 goto failed;
2170 for (i = 0; i < sbi->s_groups_count; i++) {
2171 gdp = ext4_get_group_desc(sb, i, NULL);
2173 flex_group = ext4_flex_group(sbi, i);
2174 atomic_add(ext4_free_inodes_count(sb, gdp),
2175 &sbi->s_flex_groups[flex_group].free_inodes);
2176 atomic64_add(ext4_free_group_clusters(sb, gdp),
2177 &sbi->s_flex_groups[flex_group].free_clusters);
2178 atomic_add(ext4_used_dirs_count(sb, gdp),
2179 &sbi->s_flex_groups[flex_group].used_dirs);
2182 return 1;
2183 failed:
2184 return 0;
2187 static __le16 ext4_group_desc_csum(struct super_block *sb, __u32 block_group,
2188 struct ext4_group_desc *gdp)
2190 int offset = offsetof(struct ext4_group_desc, bg_checksum);
2191 __u16 crc = 0;
2192 __le32 le_group = cpu_to_le32(block_group);
2193 struct ext4_sb_info *sbi = EXT4_SB(sb);
2195 if (ext4_has_metadata_csum(sbi->s_sb)) {
2196 /* Use new metadata_csum algorithm */
2197 __u32 csum32;
2198 __u16 dummy_csum = 0;
2200 csum32 = ext4_chksum(sbi, sbi->s_csum_seed, (__u8 *)&le_group,
2201 sizeof(le_group));
2202 csum32 = ext4_chksum(sbi, csum32, (__u8 *)gdp, offset);
2203 csum32 = ext4_chksum(sbi, csum32, (__u8 *)&dummy_csum,
2204 sizeof(dummy_csum));
2205 offset += sizeof(dummy_csum);
2206 if (offset < sbi->s_desc_size)
2207 csum32 = ext4_chksum(sbi, csum32, (__u8 *)gdp + offset,
2208 sbi->s_desc_size - offset);
2210 crc = csum32 & 0xFFFF;
2211 goto out;
2214 /* old crc16 code */
2215 if (!ext4_has_feature_gdt_csum(sb))
2216 return 0;
2218 crc = crc16(~0, sbi->s_es->s_uuid, sizeof(sbi->s_es->s_uuid));
2219 crc = crc16(crc, (__u8 *)&le_group, sizeof(le_group));
2220 crc = crc16(crc, (__u8 *)gdp, offset);
2221 offset += sizeof(gdp->bg_checksum); /* skip checksum */
2222 /* for checksum of struct ext4_group_desc do the rest...*/
2223 if (ext4_has_feature_64bit(sb) &&
2224 offset < le16_to_cpu(sbi->s_es->s_desc_size))
2225 crc = crc16(crc, (__u8 *)gdp + offset,
2226 le16_to_cpu(sbi->s_es->s_desc_size) -
2227 offset);
2229 out:
2230 return cpu_to_le16(crc);
2233 int ext4_group_desc_csum_verify(struct super_block *sb, __u32 block_group,
2234 struct ext4_group_desc *gdp)
2236 if (ext4_has_group_desc_csum(sb) &&
2237 (gdp->bg_checksum != ext4_group_desc_csum(sb, block_group, gdp)))
2238 return 0;
2240 return 1;
2243 void ext4_group_desc_csum_set(struct super_block *sb, __u32 block_group,
2244 struct ext4_group_desc *gdp)
2246 if (!ext4_has_group_desc_csum(sb))
2247 return;
2248 gdp->bg_checksum = ext4_group_desc_csum(sb, block_group, gdp);
2251 /* Called at mount-time, super-block is locked */
2252 static int ext4_check_descriptors(struct super_block *sb,
2253 ext4_fsblk_t sb_block,
2254 ext4_group_t *first_not_zeroed)
2256 struct ext4_sb_info *sbi = EXT4_SB(sb);
2257 ext4_fsblk_t first_block = le32_to_cpu(sbi->s_es->s_first_data_block);
2258 ext4_fsblk_t last_block;
2259 ext4_fsblk_t block_bitmap;
2260 ext4_fsblk_t inode_bitmap;
2261 ext4_fsblk_t inode_table;
2262 int flexbg_flag = 0;
2263 ext4_group_t i, grp = sbi->s_groups_count;
2265 if (ext4_has_feature_flex_bg(sb))
2266 flexbg_flag = 1;
2268 ext4_debug("Checking group descriptors");
2270 for (i = 0; i < sbi->s_groups_count; i++) {
2271 struct ext4_group_desc *gdp = ext4_get_group_desc(sb, i, NULL);
2273 if (i == sbi->s_groups_count - 1 || flexbg_flag)
2274 last_block = ext4_blocks_count(sbi->s_es) - 1;
2275 else
2276 last_block = first_block +
2277 (EXT4_BLOCKS_PER_GROUP(sb) - 1);
2279 if ((grp == sbi->s_groups_count) &&
2280 !(gdp->bg_flags & cpu_to_le16(EXT4_BG_INODE_ZEROED)))
2281 grp = i;
2283 block_bitmap = ext4_block_bitmap(sb, gdp);
2284 if (block_bitmap == sb_block) {
2285 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2286 "Block bitmap for group %u overlaps "
2287 "superblock", i);
2289 if (block_bitmap < first_block || block_bitmap > last_block) {
2290 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2291 "Block bitmap for group %u not in group "
2292 "(block %llu)!", i, block_bitmap);
2293 return 0;
2295 inode_bitmap = ext4_inode_bitmap(sb, gdp);
2296 if (inode_bitmap == sb_block) {
2297 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2298 "Inode bitmap for group %u overlaps "
2299 "superblock", i);
2301 if (inode_bitmap < first_block || inode_bitmap > last_block) {
2302 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2303 "Inode bitmap for group %u not in group "
2304 "(block %llu)!", i, inode_bitmap);
2305 return 0;
2307 inode_table = ext4_inode_table(sb, gdp);
2308 if (inode_table == sb_block) {
2309 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2310 "Inode table for group %u overlaps "
2311 "superblock", i);
2313 if (inode_table < first_block ||
2314 inode_table + sbi->s_itb_per_group - 1 > last_block) {
2315 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2316 "Inode table for group %u not in group "
2317 "(block %llu)!", i, inode_table);
2318 return 0;
2320 ext4_lock_group(sb, i);
2321 if (!ext4_group_desc_csum_verify(sb, i, gdp)) {
2322 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2323 "Checksum for group %u failed (%u!=%u)",
2324 i, le16_to_cpu(ext4_group_desc_csum(sb, i,
2325 gdp)), le16_to_cpu(gdp->bg_checksum));
2326 if (!(sb->s_flags & MS_RDONLY)) {
2327 ext4_unlock_group(sb, i);
2328 return 0;
2331 ext4_unlock_group(sb, i);
2332 if (!flexbg_flag)
2333 first_block += EXT4_BLOCKS_PER_GROUP(sb);
2335 if (NULL != first_not_zeroed)
2336 *first_not_zeroed = grp;
2337 return 1;
2340 /* ext4_orphan_cleanup() walks a singly-linked list of inodes (starting at
2341 * the superblock) which were deleted from all directories, but held open by
2342 * a process at the time of a crash. We walk the list and try to delete these
2343 * inodes at recovery time (only with a read-write filesystem).
2345 * In order to keep the orphan inode chain consistent during traversal (in
2346 * case of crash during recovery), we link each inode into the superblock
2347 * orphan list_head and handle it the same way as an inode deletion during
2348 * normal operation (which journals the operations for us).
2350 * We only do an iget() and an iput() on each inode, which is very safe if we
2351 * accidentally point at an in-use or already deleted inode. The worst that
2352 * can happen in this case is that we get a "bit already cleared" message from
2353 * ext4_free_inode(). The only reason we would point at a wrong inode is if
2354 * e2fsck was run on this filesystem, and it must have already done the orphan
2355 * inode cleanup for us, so we can safely abort without any further action.
2357 static void ext4_orphan_cleanup(struct super_block *sb,
2358 struct ext4_super_block *es)
2360 unsigned int s_flags = sb->s_flags;
2361 int ret, nr_orphans = 0, nr_truncates = 0;
2362 #ifdef CONFIG_QUOTA
2363 int i;
2364 #endif
2365 if (!es->s_last_orphan) {
2366 jbd_debug(4, "no orphan inodes to clean up\n");
2367 return;
2370 if (bdev_read_only(sb->s_bdev)) {
2371 ext4_msg(sb, KERN_ERR, "write access "
2372 "unavailable, skipping orphan cleanup");
2373 return;
2376 /* Check if feature set would not allow a r/w mount */
2377 if (!ext4_feature_set_ok(sb, 0)) {
2378 ext4_msg(sb, KERN_INFO, "Skipping orphan cleanup due to "
2379 "unknown ROCOMPAT features");
2380 return;
2383 if (EXT4_SB(sb)->s_mount_state & EXT4_ERROR_FS) {
2384 /* don't clear list on RO mount w/ errors */
2385 if (es->s_last_orphan && !(s_flags & MS_RDONLY)) {
2386 ext4_msg(sb, KERN_INFO, "Errors on filesystem, "
2387 "clearing orphan list.\n");
2388 es->s_last_orphan = 0;
2390 jbd_debug(1, "Skipping orphan recovery on fs with errors.\n");
2391 return;
2394 if (s_flags & MS_RDONLY) {
2395 ext4_msg(sb, KERN_INFO, "orphan cleanup on readonly fs");
2396 sb->s_flags &= ~MS_RDONLY;
2398 #ifdef CONFIG_QUOTA
2399 /* Needed for iput() to work correctly and not trash data */
2400 sb->s_flags |= MS_ACTIVE;
2401 /* Turn on quotas so that they are updated correctly */
2402 for (i = 0; i < EXT4_MAXQUOTAS; i++) {
2403 if (EXT4_SB(sb)->s_qf_names[i]) {
2404 int ret = ext4_quota_on_mount(sb, i);
2405 if (ret < 0)
2406 ext4_msg(sb, KERN_ERR,
2407 "Cannot turn on journaled "
2408 "quota: error %d", ret);
2411 #endif
2413 while (es->s_last_orphan) {
2414 struct inode *inode;
2417 * We may have encountered an error during cleanup; if
2418 * so, skip the rest.
2420 if (EXT4_SB(sb)->s_mount_state & EXT4_ERROR_FS) {
2421 jbd_debug(1, "Skipping orphan recovery on fs with errors.\n");
2422 es->s_last_orphan = 0;
2423 break;
2426 inode = ext4_orphan_get(sb, le32_to_cpu(es->s_last_orphan));
2427 if (IS_ERR(inode)) {
2428 es->s_last_orphan = 0;
2429 break;
2432 list_add(&EXT4_I(inode)->i_orphan, &EXT4_SB(sb)->s_orphan);
2433 dquot_initialize(inode);
2434 if (inode->i_nlink) {
2435 if (test_opt(sb, DEBUG))
2436 ext4_msg(sb, KERN_DEBUG,
2437 "%s: truncating inode %lu to %lld bytes",
2438 __func__, inode->i_ino, inode->i_size);
2439 jbd_debug(2, "truncating inode %lu to %lld bytes\n",
2440 inode->i_ino, inode->i_size);
2441 inode_lock(inode);
2442 truncate_inode_pages(inode->i_mapping, inode->i_size);
2443 ret = ext4_truncate(inode);
2444 if (ret)
2445 ext4_std_error(inode->i_sb, ret);
2446 inode_unlock(inode);
2447 nr_truncates++;
2448 } else {
2449 if (test_opt(sb, DEBUG))
2450 ext4_msg(sb, KERN_DEBUG,
2451 "%s: deleting unreferenced inode %lu",
2452 __func__, inode->i_ino);
2453 jbd_debug(2, "deleting unreferenced inode %lu\n",
2454 inode->i_ino);
2455 nr_orphans++;
2457 iput(inode); /* The delete magic happens here! */
2460 #define PLURAL(x) (x), ((x) == 1) ? "" : "s"
2462 if (nr_orphans)
2463 ext4_msg(sb, KERN_INFO, "%d orphan inode%s deleted",
2464 PLURAL(nr_orphans));
2465 if (nr_truncates)
2466 ext4_msg(sb, KERN_INFO, "%d truncate%s cleaned up",
2467 PLURAL(nr_truncates));
2468 #ifdef CONFIG_QUOTA
2469 /* Turn quotas off */
2470 for (i = 0; i < EXT4_MAXQUOTAS; i++) {
2471 if (sb_dqopt(sb)->files[i])
2472 dquot_quota_off(sb, i);
2474 #endif
2475 sb->s_flags = s_flags; /* Restore MS_RDONLY status */
2479 * Maximal extent format file size.
2480 * Resulting logical blkno at s_maxbytes must fit in our on-disk
2481 * extent format containers, within a sector_t, and within i_blocks
2482 * in the vfs. ext4 inode has 48 bits of i_block in fsblock units,
2483 * so that won't be a limiting factor.
2485 * However there is other limiting factor. We do store extents in the form
2486 * of starting block and length, hence the resulting length of the extent
2487 * covering maximum file size must fit into on-disk format containers as
2488 * well. Given that length is always by 1 unit bigger than max unit (because
2489 * we count 0 as well) we have to lower the s_maxbytes by one fs block.
2491 * Note, this does *not* consider any metadata overhead for vfs i_blocks.
2493 static loff_t ext4_max_size(int blkbits, int has_huge_files)
2495 loff_t res;
2496 loff_t upper_limit = MAX_LFS_FILESIZE;
2498 /* small i_blocks in vfs inode? */
2499 if (!has_huge_files || sizeof(blkcnt_t) < sizeof(u64)) {
2501 * CONFIG_LBDAF is not enabled implies the inode
2502 * i_block represent total blocks in 512 bytes
2503 * 32 == size of vfs inode i_blocks * 8
2505 upper_limit = (1LL << 32) - 1;
2507 /* total blocks in file system block size */
2508 upper_limit >>= (blkbits - 9);
2509 upper_limit <<= blkbits;
2513 * 32-bit extent-start container, ee_block. We lower the maxbytes
2514 * by one fs block, so ee_len can cover the extent of maximum file
2515 * size
2517 res = (1LL << 32) - 1;
2518 res <<= blkbits;
2520 /* Sanity check against vm- & vfs- imposed limits */
2521 if (res > upper_limit)
2522 res = upper_limit;
2524 return res;
2528 * Maximal bitmap file size. There is a direct, and {,double-,triple-}indirect
2529 * block limit, and also a limit of (2^48 - 1) 512-byte sectors in i_blocks.
2530 * We need to be 1 filesystem block less than the 2^48 sector limit.
2532 static loff_t ext4_max_bitmap_size(int bits, int has_huge_files)
2534 loff_t res = EXT4_NDIR_BLOCKS;
2535 int meta_blocks;
2536 loff_t upper_limit;
2537 /* This is calculated to be the largest file size for a dense, block
2538 * mapped file such that the file's total number of 512-byte sectors,
2539 * including data and all indirect blocks, does not exceed (2^48 - 1).
2541 * __u32 i_blocks_lo and _u16 i_blocks_high represent the total
2542 * number of 512-byte sectors of the file.
2545 if (!has_huge_files || sizeof(blkcnt_t) < sizeof(u64)) {
2547 * !has_huge_files or CONFIG_LBDAF not enabled implies that
2548 * the inode i_block field represents total file blocks in
2549 * 2^32 512-byte sectors == size of vfs inode i_blocks * 8
2551 upper_limit = (1LL << 32) - 1;
2553 /* total blocks in file system block size */
2554 upper_limit >>= (bits - 9);
2556 } else {
2558 * We use 48 bit ext4_inode i_blocks
2559 * With EXT4_HUGE_FILE_FL set the i_blocks
2560 * represent total number of blocks in
2561 * file system block size
2563 upper_limit = (1LL << 48) - 1;
2567 /* indirect blocks */
2568 meta_blocks = 1;
2569 /* double indirect blocks */
2570 meta_blocks += 1 + (1LL << (bits-2));
2571 /* tripple indirect blocks */
2572 meta_blocks += 1 + (1LL << (bits-2)) + (1LL << (2*(bits-2)));
2574 upper_limit -= meta_blocks;
2575 upper_limit <<= bits;
2577 res += 1LL << (bits-2);
2578 res += 1LL << (2*(bits-2));
2579 res += 1LL << (3*(bits-2));
2580 res <<= bits;
2581 if (res > upper_limit)
2582 res = upper_limit;
2584 if (res > MAX_LFS_FILESIZE)
2585 res = MAX_LFS_FILESIZE;
2587 return res;
2590 static ext4_fsblk_t descriptor_loc(struct super_block *sb,
2591 ext4_fsblk_t logical_sb_block, int nr)
2593 struct ext4_sb_info *sbi = EXT4_SB(sb);
2594 ext4_group_t bg, first_meta_bg;
2595 int has_super = 0;
2597 first_meta_bg = le32_to_cpu(sbi->s_es->s_first_meta_bg);
2599 if (!ext4_has_feature_meta_bg(sb) || nr < first_meta_bg)
2600 return logical_sb_block + nr + 1;
2601 bg = sbi->s_desc_per_block * nr;
2602 if (ext4_bg_has_super(sb, bg))
2603 has_super = 1;
2606 * If we have a meta_bg fs with 1k blocks, group 0's GDT is at
2607 * block 2, not 1. If s_first_data_block == 0 (bigalloc is enabled
2608 * on modern mke2fs or blksize > 1k on older mke2fs) then we must
2609 * compensate.
2611 if (sb->s_blocksize == 1024 && nr == 0 &&
2612 le32_to_cpu(EXT4_SB(sb)->s_es->s_first_data_block) == 0)
2613 has_super++;
2615 return (has_super + ext4_group_first_block_no(sb, bg));
2619 * ext4_get_stripe_size: Get the stripe size.
2620 * @sbi: In memory super block info
2622 * If we have specified it via mount option, then
2623 * use the mount option value. If the value specified at mount time is
2624 * greater than the blocks per group use the super block value.
2625 * If the super block value is greater than blocks per group return 0.
2626 * Allocator needs it be less than blocks per group.
2629 static unsigned long ext4_get_stripe_size(struct ext4_sb_info *sbi)
2631 unsigned long stride = le16_to_cpu(sbi->s_es->s_raid_stride);
2632 unsigned long stripe_width =
2633 le32_to_cpu(sbi->s_es->s_raid_stripe_width);
2634 int ret;
2636 if (sbi->s_stripe && sbi->s_stripe <= sbi->s_blocks_per_group)
2637 ret = sbi->s_stripe;
2638 else if (stripe_width && stripe_width <= sbi->s_blocks_per_group)
2639 ret = stripe_width;
2640 else if (stride && stride <= sbi->s_blocks_per_group)
2641 ret = stride;
2642 else
2643 ret = 0;
2646 * If the stripe width is 1, this makes no sense and
2647 * we set it to 0 to turn off stripe handling code.
2649 if (ret <= 1)
2650 ret = 0;
2652 return ret;
2656 * Check whether this filesystem can be mounted based on
2657 * the features present and the RDONLY/RDWR mount requested.
2658 * Returns 1 if this filesystem can be mounted as requested,
2659 * 0 if it cannot be.
2661 static int ext4_feature_set_ok(struct super_block *sb, int readonly)
2663 if (ext4_has_unknown_ext4_incompat_features(sb)) {
2664 ext4_msg(sb, KERN_ERR,
2665 "Couldn't mount because of "
2666 "unsupported optional features (%x)",
2667 (le32_to_cpu(EXT4_SB(sb)->s_es->s_feature_incompat) &
2668 ~EXT4_FEATURE_INCOMPAT_SUPP));
2669 return 0;
2672 if (readonly)
2673 return 1;
2675 if (ext4_has_feature_readonly(sb)) {
2676 ext4_msg(sb, KERN_INFO, "filesystem is read-only");
2677 sb->s_flags |= MS_RDONLY;
2678 return 1;
2681 /* Check that feature set is OK for a read-write mount */
2682 if (ext4_has_unknown_ext4_ro_compat_features(sb)) {
2683 ext4_msg(sb, KERN_ERR, "couldn't mount RDWR because of "
2684 "unsupported optional features (%x)",
2685 (le32_to_cpu(EXT4_SB(sb)->s_es->s_feature_ro_compat) &
2686 ~EXT4_FEATURE_RO_COMPAT_SUPP));
2687 return 0;
2690 * Large file size enabled file system can only be mounted
2691 * read-write on 32-bit systems if kernel is built with CONFIG_LBDAF
2693 if (ext4_has_feature_huge_file(sb)) {
2694 if (sizeof(blkcnt_t) < sizeof(u64)) {
2695 ext4_msg(sb, KERN_ERR, "Filesystem with huge files "
2696 "cannot be mounted RDWR without "
2697 "CONFIG_LBDAF");
2698 return 0;
2701 if (ext4_has_feature_bigalloc(sb) && !ext4_has_feature_extents(sb)) {
2702 ext4_msg(sb, KERN_ERR,
2703 "Can't support bigalloc feature without "
2704 "extents feature\n");
2705 return 0;
2708 #ifndef CONFIG_QUOTA
2709 if (ext4_has_feature_quota(sb) && !readonly) {
2710 ext4_msg(sb, KERN_ERR,
2711 "Filesystem with quota feature cannot be mounted RDWR "
2712 "without CONFIG_QUOTA");
2713 return 0;
2715 if (ext4_has_feature_project(sb) && !readonly) {
2716 ext4_msg(sb, KERN_ERR,
2717 "Filesystem with project quota feature cannot be mounted RDWR "
2718 "without CONFIG_QUOTA");
2719 return 0;
2721 #endif /* CONFIG_QUOTA */
2722 return 1;
2726 * This function is called once a day if we have errors logged
2727 * on the file system
2729 static void print_daily_error_info(unsigned long arg)
2731 struct super_block *sb = (struct super_block *) arg;
2732 struct ext4_sb_info *sbi;
2733 struct ext4_super_block *es;
2735 sbi = EXT4_SB(sb);
2736 es = sbi->s_es;
2738 if (es->s_error_count)
2739 /* fsck newer than v1.41.13 is needed to clean this condition. */
2740 ext4_msg(sb, KERN_NOTICE, "error count since last fsck: %u",
2741 le32_to_cpu(es->s_error_count));
2742 if (es->s_first_error_time) {
2743 printk(KERN_NOTICE "EXT4-fs (%s): initial error at time %u: %.*s:%d",
2744 sb->s_id, le32_to_cpu(es->s_first_error_time),
2745 (int) sizeof(es->s_first_error_func),
2746 es->s_first_error_func,
2747 le32_to_cpu(es->s_first_error_line));
2748 if (es->s_first_error_ino)
2749 printk(KERN_CONT ": inode %u",
2750 le32_to_cpu(es->s_first_error_ino));
2751 if (es->s_first_error_block)
2752 printk(KERN_CONT ": block %llu", (unsigned long long)
2753 le64_to_cpu(es->s_first_error_block));
2754 printk(KERN_CONT "\n");
2756 if (es->s_last_error_time) {
2757 printk(KERN_NOTICE "EXT4-fs (%s): last error at time %u: %.*s:%d",
2758 sb->s_id, le32_to_cpu(es->s_last_error_time),
2759 (int) sizeof(es->s_last_error_func),
2760 es->s_last_error_func,
2761 le32_to_cpu(es->s_last_error_line));
2762 if (es->s_last_error_ino)
2763 printk(KERN_CONT ": inode %u",
2764 le32_to_cpu(es->s_last_error_ino));
2765 if (es->s_last_error_block)
2766 printk(KERN_CONT ": block %llu", (unsigned long long)
2767 le64_to_cpu(es->s_last_error_block));
2768 printk(KERN_CONT "\n");
2770 mod_timer(&sbi->s_err_report, jiffies + 24*60*60*HZ); /* Once a day */
2773 /* Find next suitable group and run ext4_init_inode_table */
2774 static int ext4_run_li_request(struct ext4_li_request *elr)
2776 struct ext4_group_desc *gdp = NULL;
2777 ext4_group_t group, ngroups;
2778 struct super_block *sb;
2779 unsigned long timeout = 0;
2780 int ret = 0;
2782 sb = elr->lr_super;
2783 ngroups = EXT4_SB(sb)->s_groups_count;
2785 for (group = elr->lr_next_group; group < ngroups; group++) {
2786 gdp = ext4_get_group_desc(sb, group, NULL);
2787 if (!gdp) {
2788 ret = 1;
2789 break;
2792 if (!(gdp->bg_flags & cpu_to_le16(EXT4_BG_INODE_ZEROED)))
2793 break;
2796 if (group >= ngroups)
2797 ret = 1;
2799 if (!ret) {
2800 timeout = jiffies;
2801 ret = ext4_init_inode_table(sb, group,
2802 elr->lr_timeout ? 0 : 1);
2803 if (elr->lr_timeout == 0) {
2804 timeout = (jiffies - timeout) *
2805 elr->lr_sbi->s_li_wait_mult;
2806 elr->lr_timeout = timeout;
2808 elr->lr_next_sched = jiffies + elr->lr_timeout;
2809 elr->lr_next_group = group + 1;
2811 return ret;
2815 * Remove lr_request from the list_request and free the
2816 * request structure. Should be called with li_list_mtx held
2818 static void ext4_remove_li_request(struct ext4_li_request *elr)
2820 struct ext4_sb_info *sbi;
2822 if (!elr)
2823 return;
2825 sbi = elr->lr_sbi;
2827 list_del(&elr->lr_request);
2828 sbi->s_li_request = NULL;
2829 kfree(elr);
2832 static void ext4_unregister_li_request(struct super_block *sb)
2834 mutex_lock(&ext4_li_mtx);
2835 if (!ext4_li_info) {
2836 mutex_unlock(&ext4_li_mtx);
2837 return;
2840 mutex_lock(&ext4_li_info->li_list_mtx);
2841 ext4_remove_li_request(EXT4_SB(sb)->s_li_request);
2842 mutex_unlock(&ext4_li_info->li_list_mtx);
2843 mutex_unlock(&ext4_li_mtx);
2846 static struct task_struct *ext4_lazyinit_task;
2849 * This is the function where ext4lazyinit thread lives. It walks
2850 * through the request list searching for next scheduled filesystem.
2851 * When such a fs is found, run the lazy initialization request
2852 * (ext4_rn_li_request) and keep track of the time spend in this
2853 * function. Based on that time we compute next schedule time of
2854 * the request. When walking through the list is complete, compute
2855 * next waking time and put itself into sleep.
2857 static int ext4_lazyinit_thread(void *arg)
2859 struct ext4_lazy_init *eli = (struct ext4_lazy_init *)arg;
2860 struct list_head *pos, *n;
2861 struct ext4_li_request *elr;
2862 unsigned long next_wakeup, cur;
2864 BUG_ON(NULL == eli);
2866 cont_thread:
2867 while (true) {
2868 next_wakeup = MAX_JIFFY_OFFSET;
2870 mutex_lock(&eli->li_list_mtx);
2871 if (list_empty(&eli->li_request_list)) {
2872 mutex_unlock(&eli->li_list_mtx);
2873 goto exit_thread;
2875 list_for_each_safe(pos, n, &eli->li_request_list) {
2876 int err = 0;
2877 int progress = 0;
2878 elr = list_entry(pos, struct ext4_li_request,
2879 lr_request);
2881 if (time_before(jiffies, elr->lr_next_sched)) {
2882 if (time_before(elr->lr_next_sched, next_wakeup))
2883 next_wakeup = elr->lr_next_sched;
2884 continue;
2886 if (down_read_trylock(&elr->lr_super->s_umount)) {
2887 if (sb_start_write_trylock(elr->lr_super)) {
2888 progress = 1;
2890 * We hold sb->s_umount, sb can not
2891 * be removed from the list, it is
2892 * now safe to drop li_list_mtx
2894 mutex_unlock(&eli->li_list_mtx);
2895 err = ext4_run_li_request(elr);
2896 sb_end_write(elr->lr_super);
2897 mutex_lock(&eli->li_list_mtx);
2898 n = pos->next;
2900 up_read((&elr->lr_super->s_umount));
2902 /* error, remove the lazy_init job */
2903 if (err) {
2904 ext4_remove_li_request(elr);
2905 continue;
2907 if (!progress) {
2908 elr->lr_next_sched = jiffies +
2909 (prandom_u32()
2910 % (EXT4_DEF_LI_MAX_START_DELAY * HZ));
2912 if (time_before(elr->lr_next_sched, next_wakeup))
2913 next_wakeup = elr->lr_next_sched;
2915 mutex_unlock(&eli->li_list_mtx);
2917 try_to_freeze();
2919 cur = jiffies;
2920 if ((time_after_eq(cur, next_wakeup)) ||
2921 (MAX_JIFFY_OFFSET == next_wakeup)) {
2922 cond_resched();
2923 continue;
2926 schedule_timeout_interruptible(next_wakeup - cur);
2928 if (kthread_should_stop()) {
2929 ext4_clear_request_list();
2930 goto exit_thread;
2934 exit_thread:
2936 * It looks like the request list is empty, but we need
2937 * to check it under the li_list_mtx lock, to prevent any
2938 * additions into it, and of course we should lock ext4_li_mtx
2939 * to atomically free the list and ext4_li_info, because at
2940 * this point another ext4 filesystem could be registering
2941 * new one.
2943 mutex_lock(&ext4_li_mtx);
2944 mutex_lock(&eli->li_list_mtx);
2945 if (!list_empty(&eli->li_request_list)) {
2946 mutex_unlock(&eli->li_list_mtx);
2947 mutex_unlock(&ext4_li_mtx);
2948 goto cont_thread;
2950 mutex_unlock(&eli->li_list_mtx);
2951 kfree(ext4_li_info);
2952 ext4_li_info = NULL;
2953 mutex_unlock(&ext4_li_mtx);
2955 return 0;
2958 static void ext4_clear_request_list(void)
2960 struct list_head *pos, *n;
2961 struct ext4_li_request *elr;
2963 mutex_lock(&ext4_li_info->li_list_mtx);
2964 list_for_each_safe(pos, n, &ext4_li_info->li_request_list) {
2965 elr = list_entry(pos, struct ext4_li_request,
2966 lr_request);
2967 ext4_remove_li_request(elr);
2969 mutex_unlock(&ext4_li_info->li_list_mtx);
2972 static int ext4_run_lazyinit_thread(void)
2974 ext4_lazyinit_task = kthread_run(ext4_lazyinit_thread,
2975 ext4_li_info, "ext4lazyinit");
2976 if (IS_ERR(ext4_lazyinit_task)) {
2977 int err = PTR_ERR(ext4_lazyinit_task);
2978 ext4_clear_request_list();
2979 kfree(ext4_li_info);
2980 ext4_li_info = NULL;
2981 printk(KERN_CRIT "EXT4-fs: error %d creating inode table "
2982 "initialization thread\n",
2983 err);
2984 return err;
2986 ext4_li_info->li_state |= EXT4_LAZYINIT_RUNNING;
2987 return 0;
2991 * Check whether it make sense to run itable init. thread or not.
2992 * If there is at least one uninitialized inode table, return
2993 * corresponding group number, else the loop goes through all
2994 * groups and return total number of groups.
2996 static ext4_group_t ext4_has_uninit_itable(struct super_block *sb)
2998 ext4_group_t group, ngroups = EXT4_SB(sb)->s_groups_count;
2999 struct ext4_group_desc *gdp = NULL;
3001 for (group = 0; group < ngroups; group++) {
3002 gdp = ext4_get_group_desc(sb, group, NULL);
3003 if (!gdp)
3004 continue;
3006 if (!(gdp->bg_flags & cpu_to_le16(EXT4_BG_INODE_ZEROED)))
3007 break;
3010 return group;
3013 static int ext4_li_info_new(void)
3015 struct ext4_lazy_init *eli = NULL;
3017 eli = kzalloc(sizeof(*eli), GFP_KERNEL);
3018 if (!eli)
3019 return -ENOMEM;
3021 INIT_LIST_HEAD(&eli->li_request_list);
3022 mutex_init(&eli->li_list_mtx);
3024 eli->li_state |= EXT4_LAZYINIT_QUIT;
3026 ext4_li_info = eli;
3028 return 0;
3031 static struct ext4_li_request *ext4_li_request_new(struct super_block *sb,
3032 ext4_group_t start)
3034 struct ext4_sb_info *sbi = EXT4_SB(sb);
3035 struct ext4_li_request *elr;
3037 elr = kzalloc(sizeof(*elr), GFP_KERNEL);
3038 if (!elr)
3039 return NULL;
3041 elr->lr_super = sb;
3042 elr->lr_sbi = sbi;
3043 elr->lr_next_group = start;
3046 * Randomize first schedule time of the request to
3047 * spread the inode table initialization requests
3048 * better.
3050 elr->lr_next_sched = jiffies + (prandom_u32() %
3051 (EXT4_DEF_LI_MAX_START_DELAY * HZ));
3052 return elr;
3055 int ext4_register_li_request(struct super_block *sb,
3056 ext4_group_t first_not_zeroed)
3058 struct ext4_sb_info *sbi = EXT4_SB(sb);
3059 struct ext4_li_request *elr = NULL;
3060 ext4_group_t ngroups = EXT4_SB(sb)->s_groups_count;
3061 int ret = 0;
3063 mutex_lock(&ext4_li_mtx);
3064 if (sbi->s_li_request != NULL) {
3066 * Reset timeout so it can be computed again, because
3067 * s_li_wait_mult might have changed.
3069 sbi->s_li_request->lr_timeout = 0;
3070 goto out;
3073 if (first_not_zeroed == ngroups ||
3074 (sb->s_flags & MS_RDONLY) ||
3075 !test_opt(sb, INIT_INODE_TABLE))
3076 goto out;
3078 elr = ext4_li_request_new(sb, first_not_zeroed);
3079 if (!elr) {
3080 ret = -ENOMEM;
3081 goto out;
3084 if (NULL == ext4_li_info) {
3085 ret = ext4_li_info_new();
3086 if (ret)
3087 goto out;
3090 mutex_lock(&ext4_li_info->li_list_mtx);
3091 list_add(&elr->lr_request, &ext4_li_info->li_request_list);
3092 mutex_unlock(&ext4_li_info->li_list_mtx);
3094 sbi->s_li_request = elr;
3096 * set elr to NULL here since it has been inserted to
3097 * the request_list and the removal and free of it is
3098 * handled by ext4_clear_request_list from now on.
3100 elr = NULL;
3102 if (!(ext4_li_info->li_state & EXT4_LAZYINIT_RUNNING)) {
3103 ret = ext4_run_lazyinit_thread();
3104 if (ret)
3105 goto out;
3107 out:
3108 mutex_unlock(&ext4_li_mtx);
3109 if (ret)
3110 kfree(elr);
3111 return ret;
3115 * We do not need to lock anything since this is called on
3116 * module unload.
3118 static void ext4_destroy_lazyinit_thread(void)
3121 * If thread exited earlier
3122 * there's nothing to be done.
3124 if (!ext4_li_info || !ext4_lazyinit_task)
3125 return;
3127 kthread_stop(ext4_lazyinit_task);
3130 static int set_journal_csum_feature_set(struct super_block *sb)
3132 int ret = 1;
3133 int compat, incompat;
3134 struct ext4_sb_info *sbi = EXT4_SB(sb);
3136 if (ext4_has_metadata_csum(sb)) {
3137 /* journal checksum v3 */
3138 compat = 0;
3139 incompat = JBD2_FEATURE_INCOMPAT_CSUM_V3;
3140 } else {
3141 /* journal checksum v1 */
3142 compat = JBD2_FEATURE_COMPAT_CHECKSUM;
3143 incompat = 0;
3146 jbd2_journal_clear_features(sbi->s_journal,
3147 JBD2_FEATURE_COMPAT_CHECKSUM, 0,
3148 JBD2_FEATURE_INCOMPAT_CSUM_V3 |
3149 JBD2_FEATURE_INCOMPAT_CSUM_V2);
3150 if (test_opt(sb, JOURNAL_ASYNC_COMMIT)) {
3151 ret = jbd2_journal_set_features(sbi->s_journal,
3152 compat, 0,
3153 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT |
3154 incompat);
3155 } else if (test_opt(sb, JOURNAL_CHECKSUM)) {
3156 ret = jbd2_journal_set_features(sbi->s_journal,
3157 compat, 0,
3158 incompat);
3159 jbd2_journal_clear_features(sbi->s_journal, 0, 0,
3160 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT);
3161 } else {
3162 jbd2_journal_clear_features(sbi->s_journal, 0, 0,
3163 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT);
3166 return ret;
3170 * Note: calculating the overhead so we can be compatible with
3171 * historical BSD practice is quite difficult in the face of
3172 * clusters/bigalloc. This is because multiple metadata blocks from
3173 * different block group can end up in the same allocation cluster.
3174 * Calculating the exact overhead in the face of clustered allocation
3175 * requires either O(all block bitmaps) in memory or O(number of block
3176 * groups**2) in time. We will still calculate the superblock for
3177 * older file systems --- and if we come across with a bigalloc file
3178 * system with zero in s_overhead_clusters the estimate will be close to
3179 * correct especially for very large cluster sizes --- but for newer
3180 * file systems, it's better to calculate this figure once at mkfs
3181 * time, and store it in the superblock. If the superblock value is
3182 * present (even for non-bigalloc file systems), we will use it.
3184 static int count_overhead(struct super_block *sb, ext4_group_t grp,
3185 char *buf)
3187 struct ext4_sb_info *sbi = EXT4_SB(sb);
3188 struct ext4_group_desc *gdp;
3189 ext4_fsblk_t first_block, last_block, b;
3190 ext4_group_t i, ngroups = ext4_get_groups_count(sb);
3191 int s, j, count = 0;
3193 if (!ext4_has_feature_bigalloc(sb))
3194 return (ext4_bg_has_super(sb, grp) + ext4_bg_num_gdb(sb, grp) +
3195 sbi->s_itb_per_group + 2);
3197 first_block = le32_to_cpu(sbi->s_es->s_first_data_block) +
3198 (grp * EXT4_BLOCKS_PER_GROUP(sb));
3199 last_block = first_block + EXT4_BLOCKS_PER_GROUP(sb) - 1;
3200 for (i = 0; i < ngroups; i++) {
3201 gdp = ext4_get_group_desc(sb, i, NULL);
3202 b = ext4_block_bitmap(sb, gdp);
3203 if (b >= first_block && b <= last_block) {
3204 ext4_set_bit(EXT4_B2C(sbi, b - first_block), buf);
3205 count++;
3207 b = ext4_inode_bitmap(sb, gdp);
3208 if (b >= first_block && b <= last_block) {
3209 ext4_set_bit(EXT4_B2C(sbi, b - first_block), buf);
3210 count++;
3212 b = ext4_inode_table(sb, gdp);
3213 if (b >= first_block && b + sbi->s_itb_per_group <= last_block)
3214 for (j = 0; j < sbi->s_itb_per_group; j++, b++) {
3215 int c = EXT4_B2C(sbi, b - first_block);
3216 ext4_set_bit(c, buf);
3217 count++;
3219 if (i != grp)
3220 continue;
3221 s = 0;
3222 if (ext4_bg_has_super(sb, grp)) {
3223 ext4_set_bit(s++, buf);
3224 count++;
3226 j = ext4_bg_num_gdb(sb, grp);
3227 if (s + j > EXT4_BLOCKS_PER_GROUP(sb)) {
3228 ext4_error(sb, "Invalid number of block group "
3229 "descriptor blocks: %d", j);
3230 j = EXT4_BLOCKS_PER_GROUP(sb) - s;
3232 count += j;
3233 for (; j > 0; j--)
3234 ext4_set_bit(EXT4_B2C(sbi, s++), buf);
3236 if (!count)
3237 return 0;
3238 return EXT4_CLUSTERS_PER_GROUP(sb) -
3239 ext4_count_free(buf, EXT4_CLUSTERS_PER_GROUP(sb) / 8);
3243 * Compute the overhead and stash it in sbi->s_overhead
3245 int ext4_calculate_overhead(struct super_block *sb)
3247 struct ext4_sb_info *sbi = EXT4_SB(sb);
3248 struct ext4_super_block *es = sbi->s_es;
3249 struct inode *j_inode;
3250 unsigned int j_blocks, j_inum = le32_to_cpu(es->s_journal_inum);
3251 ext4_group_t i, ngroups = ext4_get_groups_count(sb);
3252 ext4_fsblk_t overhead = 0;
3253 char *buf = (char *) get_zeroed_page(GFP_NOFS);
3255 if (!buf)
3256 return -ENOMEM;
3259 * Compute the overhead (FS structures). This is constant
3260 * for a given filesystem unless the number of block groups
3261 * changes so we cache the previous value until it does.
3265 * All of the blocks before first_data_block are overhead
3267 overhead = EXT4_B2C(sbi, le32_to_cpu(es->s_first_data_block));
3270 * Add the overhead found in each block group
3272 for (i = 0; i < ngroups; i++) {
3273 int blks;
3275 blks = count_overhead(sb, i, buf);
3276 overhead += blks;
3277 if (blks)
3278 memset(buf, 0, PAGE_SIZE);
3279 cond_resched();
3283 * Add the internal journal blocks whether the journal has been
3284 * loaded or not
3286 if (sbi->s_journal && !sbi->journal_bdev)
3287 overhead += EXT4_NUM_B2C(sbi, sbi->s_journal->j_maxlen);
3288 else if (ext4_has_feature_journal(sb) && !sbi->s_journal) {
3289 j_inode = ext4_get_journal_inode(sb, j_inum);
3290 if (j_inode) {
3291 j_blocks = j_inode->i_size >> sb->s_blocksize_bits;
3292 overhead += EXT4_NUM_B2C(sbi, j_blocks);
3293 iput(j_inode);
3294 } else {
3295 ext4_msg(sb, KERN_ERR, "can't get journal size");
3298 sbi->s_overhead = overhead;
3299 smp_wmb();
3300 free_page((unsigned long) buf);
3301 return 0;
3304 static void ext4_set_resv_clusters(struct super_block *sb)
3306 ext4_fsblk_t resv_clusters;
3307 struct ext4_sb_info *sbi = EXT4_SB(sb);
3310 * There's no need to reserve anything when we aren't using extents.
3311 * The space estimates are exact, there are no unwritten extents,
3312 * hole punching doesn't need new metadata... This is needed especially
3313 * to keep ext2/3 backward compatibility.
3315 if (!ext4_has_feature_extents(sb))
3316 return;
3318 * By default we reserve 2% or 4096 clusters, whichever is smaller.
3319 * This should cover the situations where we can not afford to run
3320 * out of space like for example punch hole, or converting
3321 * unwritten extents in delalloc path. In most cases such
3322 * allocation would require 1, or 2 blocks, higher numbers are
3323 * very rare.
3325 resv_clusters = (ext4_blocks_count(sbi->s_es) >>
3326 sbi->s_cluster_bits);
3328 do_div(resv_clusters, 50);
3329 resv_clusters = min_t(ext4_fsblk_t, resv_clusters, 4096);
3331 atomic64_set(&sbi->s_resv_clusters, resv_clusters);
3334 static int ext4_fill_super(struct super_block *sb, void *data, int silent)
3336 char *orig_data = kstrdup(data, GFP_KERNEL);
3337 struct buffer_head *bh;
3338 struct ext4_super_block *es = NULL;
3339 struct ext4_sb_info *sbi = kzalloc(sizeof(*sbi), GFP_KERNEL);
3340 ext4_fsblk_t block;
3341 ext4_fsblk_t sb_block = get_sb_block(&data);
3342 ext4_fsblk_t logical_sb_block;
3343 unsigned long offset = 0;
3344 unsigned long journal_devnum = 0;
3345 unsigned long def_mount_opts;
3346 struct inode *root;
3347 const char *descr;
3348 int ret = -ENOMEM;
3349 int blocksize, clustersize;
3350 unsigned int db_count;
3351 unsigned int i;
3352 int needs_recovery, has_huge_files, has_bigalloc;
3353 __u64 blocks_count;
3354 int err = 0;
3355 unsigned int journal_ioprio = DEFAULT_JOURNAL_IOPRIO;
3356 ext4_group_t first_not_zeroed;
3358 if ((data && !orig_data) || !sbi)
3359 goto out_free_base;
3361 sbi->s_blockgroup_lock =
3362 kzalloc(sizeof(struct blockgroup_lock), GFP_KERNEL);
3363 if (!sbi->s_blockgroup_lock)
3364 goto out_free_base;
3366 sb->s_fs_info = sbi;
3367 sbi->s_sb = sb;
3368 sbi->s_inode_readahead_blks = EXT4_DEF_INODE_READAHEAD_BLKS;
3369 sbi->s_sb_block = sb_block;
3370 if (sb->s_bdev->bd_part)
3371 sbi->s_sectors_written_start =
3372 part_stat_read(sb->s_bdev->bd_part, sectors[1]);
3374 /* Cleanup superblock name */
3375 strreplace(sb->s_id, '/', '!');
3377 /* -EINVAL is default */
3378 ret = -EINVAL;
3379 blocksize = sb_min_blocksize(sb, EXT4_MIN_BLOCK_SIZE);
3380 if (!blocksize) {
3381 ext4_msg(sb, KERN_ERR, "unable to set blocksize");
3382 goto out_fail;
3386 * The ext4 superblock will not be buffer aligned for other than 1kB
3387 * block sizes. We need to calculate the offset from buffer start.
3389 if (blocksize != EXT4_MIN_BLOCK_SIZE) {
3390 logical_sb_block = sb_block * EXT4_MIN_BLOCK_SIZE;
3391 offset = do_div(logical_sb_block, blocksize);
3392 } else {
3393 logical_sb_block = sb_block;
3396 if (!(bh = sb_bread_unmovable(sb, logical_sb_block))) {
3397 ext4_msg(sb, KERN_ERR, "unable to read superblock");
3398 goto out_fail;
3401 * Note: s_es must be initialized as soon as possible because
3402 * some ext4 macro-instructions depend on its value
3404 es = (struct ext4_super_block *) (bh->b_data + offset);
3405 sbi->s_es = es;
3406 sb->s_magic = le16_to_cpu(es->s_magic);
3407 if (sb->s_magic != EXT4_SUPER_MAGIC)
3408 goto cantfind_ext4;
3409 sbi->s_kbytes_written = le64_to_cpu(es->s_kbytes_written);
3411 /* Warn if metadata_csum and gdt_csum are both set. */
3412 if (ext4_has_feature_metadata_csum(sb) &&
3413 ext4_has_feature_gdt_csum(sb))
3414 ext4_warning(sb, "metadata_csum and uninit_bg are "
3415 "redundant flags; please run fsck.");
3417 /* Check for a known checksum algorithm */
3418 if (!ext4_verify_csum_type(sb, es)) {
3419 ext4_msg(sb, KERN_ERR, "VFS: Found ext4 filesystem with "
3420 "unknown checksum algorithm.");
3421 silent = 1;
3422 goto cantfind_ext4;
3425 /* Load the checksum driver */
3426 if (ext4_has_feature_metadata_csum(sb)) {
3427 sbi->s_chksum_driver = crypto_alloc_shash("crc32c", 0, 0);
3428 if (IS_ERR(sbi->s_chksum_driver)) {
3429 ext4_msg(sb, KERN_ERR, "Cannot load crc32c driver.");
3430 ret = PTR_ERR(sbi->s_chksum_driver);
3431 sbi->s_chksum_driver = NULL;
3432 goto failed_mount;
3436 /* Check superblock checksum */
3437 if (!ext4_superblock_csum_verify(sb, es)) {
3438 ext4_msg(sb, KERN_ERR, "VFS: Found ext4 filesystem with "
3439 "invalid superblock checksum. Run e2fsck?");
3440 silent = 1;
3441 ret = -EFSBADCRC;
3442 goto cantfind_ext4;
3445 /* Precompute checksum seed for all metadata */
3446 if (ext4_has_feature_csum_seed(sb))
3447 sbi->s_csum_seed = le32_to_cpu(es->s_checksum_seed);
3448 else if (ext4_has_metadata_csum(sb))
3449 sbi->s_csum_seed = ext4_chksum(sbi, ~0, es->s_uuid,
3450 sizeof(es->s_uuid));
3452 /* Set defaults before we parse the mount options */
3453 def_mount_opts = le32_to_cpu(es->s_default_mount_opts);
3454 set_opt(sb, INIT_INODE_TABLE);
3455 if (def_mount_opts & EXT4_DEFM_DEBUG)
3456 set_opt(sb, DEBUG);
3457 if (def_mount_opts & EXT4_DEFM_BSDGROUPS)
3458 set_opt(sb, GRPID);
3459 if (def_mount_opts & EXT4_DEFM_UID16)
3460 set_opt(sb, NO_UID32);
3461 /* xattr user namespace & acls are now defaulted on */
3462 set_opt(sb, XATTR_USER);
3463 #ifdef CONFIG_EXT4_FS_POSIX_ACL
3464 set_opt(sb, POSIX_ACL);
3465 #endif
3466 /* don't forget to enable journal_csum when metadata_csum is enabled. */
3467 if (ext4_has_metadata_csum(sb))
3468 set_opt(sb, JOURNAL_CHECKSUM);
3470 if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_DATA)
3471 set_opt(sb, JOURNAL_DATA);
3472 else if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_ORDERED)
3473 set_opt(sb, ORDERED_DATA);
3474 else if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_WBACK)
3475 set_opt(sb, WRITEBACK_DATA);
3477 if (le16_to_cpu(sbi->s_es->s_errors) == EXT4_ERRORS_PANIC)
3478 set_opt(sb, ERRORS_PANIC);
3479 else if (le16_to_cpu(sbi->s_es->s_errors) == EXT4_ERRORS_CONTINUE)
3480 set_opt(sb, ERRORS_CONT);
3481 else
3482 set_opt(sb, ERRORS_RO);
3483 /* block_validity enabled by default; disable with noblock_validity */
3484 set_opt(sb, BLOCK_VALIDITY);
3485 if (def_mount_opts & EXT4_DEFM_DISCARD)
3486 set_opt(sb, DISCARD);
3488 sbi->s_resuid = make_kuid(&init_user_ns, le16_to_cpu(es->s_def_resuid));
3489 sbi->s_resgid = make_kgid(&init_user_ns, le16_to_cpu(es->s_def_resgid));
3490 sbi->s_commit_interval = JBD2_DEFAULT_MAX_COMMIT_AGE * HZ;
3491 sbi->s_min_batch_time = EXT4_DEF_MIN_BATCH_TIME;
3492 sbi->s_max_batch_time = EXT4_DEF_MAX_BATCH_TIME;
3494 if ((def_mount_opts & EXT4_DEFM_NOBARRIER) == 0)
3495 set_opt(sb, BARRIER);
3498 * enable delayed allocation by default
3499 * Use -o nodelalloc to turn it off
3501 if (!IS_EXT3_SB(sb) && !IS_EXT2_SB(sb) &&
3502 ((def_mount_opts & EXT4_DEFM_NODELALLOC) == 0))
3503 set_opt(sb, DELALLOC);
3506 * set default s_li_wait_mult for lazyinit, for the case there is
3507 * no mount option specified.
3509 sbi->s_li_wait_mult = EXT4_DEF_LI_WAIT_MULT;
3511 if (sbi->s_es->s_mount_opts[0]) {
3512 char *s_mount_opts = kstrndup(sbi->s_es->s_mount_opts,
3513 sizeof(sbi->s_es->s_mount_opts),
3514 GFP_KERNEL);
3515 if (!s_mount_opts)
3516 goto failed_mount;
3517 if (!parse_options(s_mount_opts, sb, &journal_devnum,
3518 &journal_ioprio, 0)) {
3519 ext4_msg(sb, KERN_WARNING,
3520 "failed to parse options in superblock: %s",
3521 s_mount_opts);
3523 kfree(s_mount_opts);
3525 sbi->s_def_mount_opt = sbi->s_mount_opt;
3526 if (!parse_options((char *) data, sb, &journal_devnum,
3527 &journal_ioprio, 0))
3528 goto failed_mount;
3530 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA) {
3531 printk_once(KERN_WARNING "EXT4-fs: Warning: mounting "
3532 "with data=journal disables delayed "
3533 "allocation and O_DIRECT support!\n");
3534 if (test_opt2(sb, EXPLICIT_DELALLOC)) {
3535 ext4_msg(sb, KERN_ERR, "can't mount with "
3536 "both data=journal and delalloc");
3537 goto failed_mount;
3539 if (test_opt(sb, DIOREAD_NOLOCK)) {
3540 ext4_msg(sb, KERN_ERR, "can't mount with "
3541 "both data=journal and dioread_nolock");
3542 goto failed_mount;
3544 if (test_opt(sb, DAX)) {
3545 ext4_msg(sb, KERN_ERR, "can't mount with "
3546 "both data=journal and dax");
3547 goto failed_mount;
3549 if (ext4_has_feature_encrypt(sb)) {
3550 ext4_msg(sb, KERN_WARNING,
3551 "encrypted files will use data=ordered "
3552 "instead of data journaling mode");
3554 if (test_opt(sb, DELALLOC))
3555 clear_opt(sb, DELALLOC);
3556 } else {
3557 sb->s_iflags |= SB_I_CGROUPWB;
3560 sb->s_flags = (sb->s_flags & ~MS_POSIXACL) |
3561 (test_opt(sb, POSIX_ACL) ? MS_POSIXACL : 0);
3563 if (le32_to_cpu(es->s_rev_level) == EXT4_GOOD_OLD_REV &&
3564 (ext4_has_compat_features(sb) ||
3565 ext4_has_ro_compat_features(sb) ||
3566 ext4_has_incompat_features(sb)))
3567 ext4_msg(sb, KERN_WARNING,
3568 "feature flags set on rev 0 fs, "
3569 "running e2fsck is recommended");
3571 if (es->s_creator_os == cpu_to_le32(EXT4_OS_HURD)) {
3572 set_opt2(sb, HURD_COMPAT);
3573 if (ext4_has_feature_64bit(sb)) {
3574 ext4_msg(sb, KERN_ERR,
3575 "The Hurd can't support 64-bit file systems");
3576 goto failed_mount;
3580 if (IS_EXT2_SB(sb)) {
3581 if (ext2_feature_set_ok(sb))
3582 ext4_msg(sb, KERN_INFO, "mounting ext2 file system "
3583 "using the ext4 subsystem");
3584 else {
3585 ext4_msg(sb, KERN_ERR, "couldn't mount as ext2 due "
3586 "to feature incompatibilities");
3587 goto failed_mount;
3591 if (IS_EXT3_SB(sb)) {
3592 if (ext3_feature_set_ok(sb))
3593 ext4_msg(sb, KERN_INFO, "mounting ext3 file system "
3594 "using the ext4 subsystem");
3595 else {
3596 ext4_msg(sb, KERN_ERR, "couldn't mount as ext3 due "
3597 "to feature incompatibilities");
3598 goto failed_mount;
3603 * Check feature flags regardless of the revision level, since we
3604 * previously didn't change the revision level when setting the flags,
3605 * so there is a chance incompat flags are set on a rev 0 filesystem.
3607 if (!ext4_feature_set_ok(sb, (sb->s_flags & MS_RDONLY)))
3608 goto failed_mount;
3610 blocksize = BLOCK_SIZE << le32_to_cpu(es->s_log_block_size);
3611 if (blocksize < EXT4_MIN_BLOCK_SIZE ||
3612 blocksize > EXT4_MAX_BLOCK_SIZE) {
3613 ext4_msg(sb, KERN_ERR,
3614 "Unsupported filesystem blocksize %d (%d log_block_size)",
3615 blocksize, le32_to_cpu(es->s_log_block_size));
3616 goto failed_mount;
3618 if (le32_to_cpu(es->s_log_block_size) >
3619 (EXT4_MAX_BLOCK_LOG_SIZE - EXT4_MIN_BLOCK_LOG_SIZE)) {
3620 ext4_msg(sb, KERN_ERR,
3621 "Invalid log block size: %u",
3622 le32_to_cpu(es->s_log_block_size));
3623 goto failed_mount;
3626 if (le16_to_cpu(sbi->s_es->s_reserved_gdt_blocks) > (blocksize / 4)) {
3627 ext4_msg(sb, KERN_ERR,
3628 "Number of reserved GDT blocks insanely large: %d",
3629 le16_to_cpu(sbi->s_es->s_reserved_gdt_blocks));
3630 goto failed_mount;
3633 if (sbi->s_mount_opt & EXT4_MOUNT_DAX) {
3634 err = bdev_dax_supported(sb, blocksize);
3635 if (err)
3636 goto failed_mount;
3639 if (ext4_has_feature_encrypt(sb) && es->s_encryption_level) {
3640 ext4_msg(sb, KERN_ERR, "Unsupported encryption level %d",
3641 es->s_encryption_level);
3642 goto failed_mount;
3645 if (sb->s_blocksize != blocksize) {
3646 /* Validate the filesystem blocksize */
3647 if (!sb_set_blocksize(sb, blocksize)) {
3648 ext4_msg(sb, KERN_ERR, "bad block size %d",
3649 blocksize);
3650 goto failed_mount;
3653 brelse(bh);
3654 logical_sb_block = sb_block * EXT4_MIN_BLOCK_SIZE;
3655 offset = do_div(logical_sb_block, blocksize);
3656 bh = sb_bread_unmovable(sb, logical_sb_block);
3657 if (!bh) {
3658 ext4_msg(sb, KERN_ERR,
3659 "Can't read superblock on 2nd try");
3660 goto failed_mount;
3662 es = (struct ext4_super_block *)(bh->b_data + offset);
3663 sbi->s_es = es;
3664 if (es->s_magic != cpu_to_le16(EXT4_SUPER_MAGIC)) {
3665 ext4_msg(sb, KERN_ERR,
3666 "Magic mismatch, very weird!");
3667 goto failed_mount;
3671 has_huge_files = ext4_has_feature_huge_file(sb);
3672 sbi->s_bitmap_maxbytes = ext4_max_bitmap_size(sb->s_blocksize_bits,
3673 has_huge_files);
3674 sb->s_maxbytes = ext4_max_size(sb->s_blocksize_bits, has_huge_files);
3676 if (le32_to_cpu(es->s_rev_level) == EXT4_GOOD_OLD_REV) {
3677 sbi->s_inode_size = EXT4_GOOD_OLD_INODE_SIZE;
3678 sbi->s_first_ino = EXT4_GOOD_OLD_FIRST_INO;
3679 } else {
3680 sbi->s_inode_size = le16_to_cpu(es->s_inode_size);
3681 sbi->s_first_ino = le32_to_cpu(es->s_first_ino);
3682 if ((sbi->s_inode_size < EXT4_GOOD_OLD_INODE_SIZE) ||
3683 (!is_power_of_2(sbi->s_inode_size)) ||
3684 (sbi->s_inode_size > blocksize)) {
3685 ext4_msg(sb, KERN_ERR,
3686 "unsupported inode size: %d",
3687 sbi->s_inode_size);
3688 goto failed_mount;
3690 if (sbi->s_inode_size > EXT4_GOOD_OLD_INODE_SIZE)
3691 sb->s_time_gran = 1 << (EXT4_EPOCH_BITS - 2);
3694 sbi->s_desc_size = le16_to_cpu(es->s_desc_size);
3695 if (ext4_has_feature_64bit(sb)) {
3696 if (sbi->s_desc_size < EXT4_MIN_DESC_SIZE_64BIT ||
3697 sbi->s_desc_size > EXT4_MAX_DESC_SIZE ||
3698 !is_power_of_2(sbi->s_desc_size)) {
3699 ext4_msg(sb, KERN_ERR,
3700 "unsupported descriptor size %lu",
3701 sbi->s_desc_size);
3702 goto failed_mount;
3704 } else
3705 sbi->s_desc_size = EXT4_MIN_DESC_SIZE;
3707 sbi->s_blocks_per_group = le32_to_cpu(es->s_blocks_per_group);
3708 sbi->s_inodes_per_group = le32_to_cpu(es->s_inodes_per_group);
3710 sbi->s_inodes_per_block = blocksize / EXT4_INODE_SIZE(sb);
3711 if (sbi->s_inodes_per_block == 0)
3712 goto cantfind_ext4;
3713 if (sbi->s_inodes_per_group < sbi->s_inodes_per_block ||
3714 sbi->s_inodes_per_group > blocksize * 8) {
3715 ext4_msg(sb, KERN_ERR, "invalid inodes per group: %lu\n",
3716 sbi->s_blocks_per_group);
3717 goto failed_mount;
3719 sbi->s_itb_per_group = sbi->s_inodes_per_group /
3720 sbi->s_inodes_per_block;
3721 sbi->s_desc_per_block = blocksize / EXT4_DESC_SIZE(sb);
3722 sbi->s_sbh = bh;
3723 sbi->s_mount_state = le16_to_cpu(es->s_state);
3724 sbi->s_addr_per_block_bits = ilog2(EXT4_ADDR_PER_BLOCK(sb));
3725 sbi->s_desc_per_block_bits = ilog2(EXT4_DESC_PER_BLOCK(sb));
3727 for (i = 0; i < 4; i++)
3728 sbi->s_hash_seed[i] = le32_to_cpu(es->s_hash_seed[i]);
3729 sbi->s_def_hash_version = es->s_def_hash_version;
3730 if (ext4_has_feature_dir_index(sb)) {
3731 i = le32_to_cpu(es->s_flags);
3732 if (i & EXT2_FLAGS_UNSIGNED_HASH)
3733 sbi->s_hash_unsigned = 3;
3734 else if ((i & EXT2_FLAGS_SIGNED_HASH) == 0) {
3735 #ifdef __CHAR_UNSIGNED__
3736 if (!(sb->s_flags & MS_RDONLY))
3737 es->s_flags |=
3738 cpu_to_le32(EXT2_FLAGS_UNSIGNED_HASH);
3739 sbi->s_hash_unsigned = 3;
3740 #else
3741 if (!(sb->s_flags & MS_RDONLY))
3742 es->s_flags |=
3743 cpu_to_le32(EXT2_FLAGS_SIGNED_HASH);
3744 #endif
3748 /* Handle clustersize */
3749 clustersize = BLOCK_SIZE << le32_to_cpu(es->s_log_cluster_size);
3750 has_bigalloc = ext4_has_feature_bigalloc(sb);
3751 if (has_bigalloc) {
3752 if (clustersize < blocksize) {
3753 ext4_msg(sb, KERN_ERR,
3754 "cluster size (%d) smaller than "
3755 "block size (%d)", clustersize, blocksize);
3756 goto failed_mount;
3758 if (le32_to_cpu(es->s_log_cluster_size) >
3759 (EXT4_MAX_CLUSTER_LOG_SIZE - EXT4_MIN_BLOCK_LOG_SIZE)) {
3760 ext4_msg(sb, KERN_ERR,
3761 "Invalid log cluster size: %u",
3762 le32_to_cpu(es->s_log_cluster_size));
3763 goto failed_mount;
3765 sbi->s_cluster_bits = le32_to_cpu(es->s_log_cluster_size) -
3766 le32_to_cpu(es->s_log_block_size);
3767 sbi->s_clusters_per_group =
3768 le32_to_cpu(es->s_clusters_per_group);
3769 if (sbi->s_clusters_per_group > blocksize * 8) {
3770 ext4_msg(sb, KERN_ERR,
3771 "#clusters per group too big: %lu",
3772 sbi->s_clusters_per_group);
3773 goto failed_mount;
3775 if (sbi->s_blocks_per_group !=
3776 (sbi->s_clusters_per_group * (clustersize / blocksize))) {
3777 ext4_msg(sb, KERN_ERR, "blocks per group (%lu) and "
3778 "clusters per group (%lu) inconsistent",
3779 sbi->s_blocks_per_group,
3780 sbi->s_clusters_per_group);
3781 goto failed_mount;
3783 } else {
3784 if (clustersize != blocksize) {
3785 ext4_warning(sb, "fragment/cluster size (%d) != "
3786 "block size (%d)", clustersize,
3787 blocksize);
3788 clustersize = blocksize;
3790 if (sbi->s_blocks_per_group > blocksize * 8) {
3791 ext4_msg(sb, KERN_ERR,
3792 "#blocks per group too big: %lu",
3793 sbi->s_blocks_per_group);
3794 goto failed_mount;
3796 sbi->s_clusters_per_group = sbi->s_blocks_per_group;
3797 sbi->s_cluster_bits = 0;
3799 sbi->s_cluster_ratio = clustersize / blocksize;
3801 /* Do we have standard group size of clustersize * 8 blocks ? */
3802 if (sbi->s_blocks_per_group == clustersize << 3)
3803 set_opt2(sb, STD_GROUP_SIZE);
3806 * Test whether we have more sectors than will fit in sector_t,
3807 * and whether the max offset is addressable by the page cache.
3809 err = generic_check_addressable(sb->s_blocksize_bits,
3810 ext4_blocks_count(es));
3811 if (err) {
3812 ext4_msg(sb, KERN_ERR, "filesystem"
3813 " too large to mount safely on this system");
3814 if (sizeof(sector_t) < 8)
3815 ext4_msg(sb, KERN_WARNING, "CONFIG_LBDAF not enabled");
3816 goto failed_mount;
3819 if (EXT4_BLOCKS_PER_GROUP(sb) == 0)
3820 goto cantfind_ext4;
3822 /* check blocks count against device size */
3823 blocks_count = sb->s_bdev->bd_inode->i_size >> sb->s_blocksize_bits;
3824 if (blocks_count && ext4_blocks_count(es) > blocks_count) {
3825 ext4_msg(sb, KERN_WARNING, "bad geometry: block count %llu "
3826 "exceeds size of device (%llu blocks)",
3827 ext4_blocks_count(es), blocks_count);
3828 goto failed_mount;
3832 * It makes no sense for the first data block to be beyond the end
3833 * of the filesystem.
3835 if (le32_to_cpu(es->s_first_data_block) >= ext4_blocks_count(es)) {
3836 ext4_msg(sb, KERN_WARNING, "bad geometry: first data "
3837 "block %u is beyond end of filesystem (%llu)",
3838 le32_to_cpu(es->s_first_data_block),
3839 ext4_blocks_count(es));
3840 goto failed_mount;
3842 blocks_count = (ext4_blocks_count(es) -
3843 le32_to_cpu(es->s_first_data_block) +
3844 EXT4_BLOCKS_PER_GROUP(sb) - 1);
3845 do_div(blocks_count, EXT4_BLOCKS_PER_GROUP(sb));
3846 if (blocks_count > ((uint64_t)1<<32) - EXT4_DESC_PER_BLOCK(sb)) {
3847 ext4_msg(sb, KERN_WARNING, "groups count too large: %u "
3848 "(block count %llu, first data block %u, "
3849 "blocks per group %lu)", sbi->s_groups_count,
3850 ext4_blocks_count(es),
3851 le32_to_cpu(es->s_first_data_block),
3852 EXT4_BLOCKS_PER_GROUP(sb));
3853 goto failed_mount;
3855 sbi->s_groups_count = blocks_count;
3856 sbi->s_blockfile_groups = min_t(ext4_group_t, sbi->s_groups_count,
3857 (EXT4_MAX_BLOCK_FILE_PHYS / EXT4_BLOCKS_PER_GROUP(sb)));
3858 db_count = (sbi->s_groups_count + EXT4_DESC_PER_BLOCK(sb) - 1) /
3859 EXT4_DESC_PER_BLOCK(sb);
3860 if (ext4_has_feature_meta_bg(sb)) {
3861 if (le32_to_cpu(es->s_first_meta_bg) > db_count) {
3862 ext4_msg(sb, KERN_WARNING,
3863 "first meta block group too large: %u "
3864 "(group descriptor block count %u)",
3865 le32_to_cpu(es->s_first_meta_bg), db_count);
3866 goto failed_mount;
3869 sbi->s_group_desc = ext4_kvmalloc(db_count *
3870 sizeof(struct buffer_head *),
3871 GFP_KERNEL);
3872 if (sbi->s_group_desc == NULL) {
3873 ext4_msg(sb, KERN_ERR, "not enough memory");
3874 ret = -ENOMEM;
3875 goto failed_mount;
3878 bgl_lock_init(sbi->s_blockgroup_lock);
3880 for (i = 0; i < db_count; i++) {
3881 block = descriptor_loc(sb, logical_sb_block, i);
3882 sbi->s_group_desc[i] = sb_bread_unmovable(sb, block);
3883 if (!sbi->s_group_desc[i]) {
3884 ext4_msg(sb, KERN_ERR,
3885 "can't read group descriptor %d", i);
3886 db_count = i;
3887 goto failed_mount2;
3890 if (!ext4_check_descriptors(sb, logical_sb_block, &first_not_zeroed)) {
3891 ext4_msg(sb, KERN_ERR, "group descriptors corrupted!");
3892 ret = -EFSCORRUPTED;
3893 goto failed_mount2;
3896 sbi->s_gdb_count = db_count;
3897 get_random_bytes(&sbi->s_next_generation, sizeof(u32));
3898 spin_lock_init(&sbi->s_next_gen_lock);
3900 setup_timer(&sbi->s_err_report, print_daily_error_info,
3901 (unsigned long) sb);
3903 /* Register extent status tree shrinker */
3904 if (ext4_es_register_shrinker(sbi))
3905 goto failed_mount3;
3907 sbi->s_stripe = ext4_get_stripe_size(sbi);
3908 sbi->s_extent_max_zeroout_kb = 32;
3911 * set up enough so that it can read an inode
3913 sb->s_op = &ext4_sops;
3914 sb->s_export_op = &ext4_export_ops;
3915 sb->s_xattr = ext4_xattr_handlers;
3916 sb->s_cop = &ext4_cryptops;
3917 #ifdef CONFIG_QUOTA
3918 sb->dq_op = &ext4_quota_operations;
3919 if (ext4_has_feature_quota(sb))
3920 sb->s_qcop = &dquot_quotactl_sysfile_ops;
3921 else
3922 sb->s_qcop = &ext4_qctl_operations;
3923 sb->s_quota_types = QTYPE_MASK_USR | QTYPE_MASK_GRP | QTYPE_MASK_PRJ;
3924 #endif
3925 memcpy(sb->s_uuid, es->s_uuid, sizeof(es->s_uuid));
3927 INIT_LIST_HEAD(&sbi->s_orphan); /* unlinked but open files */
3928 mutex_init(&sbi->s_orphan_lock);
3930 sb->s_root = NULL;
3932 needs_recovery = (es->s_last_orphan != 0 ||
3933 ext4_has_feature_journal_needs_recovery(sb));
3935 if (ext4_has_feature_mmp(sb) && !(sb->s_flags & MS_RDONLY))
3936 if (ext4_multi_mount_protect(sb, le64_to_cpu(es->s_mmp_block)))
3937 goto failed_mount3a;
3940 * The first inode we look at is the journal inode. Don't try
3941 * root first: it may be modified in the journal!
3943 if (!test_opt(sb, NOLOAD) && ext4_has_feature_journal(sb)) {
3944 err = ext4_load_journal(sb, es, journal_devnum);
3945 if (err)
3946 goto failed_mount3a;
3947 } else if (test_opt(sb, NOLOAD) && !(sb->s_flags & MS_RDONLY) &&
3948 ext4_has_feature_journal_needs_recovery(sb)) {
3949 ext4_msg(sb, KERN_ERR, "required journal recovery "
3950 "suppressed and not mounted read-only");
3951 goto failed_mount_wq;
3952 } else {
3953 /* Nojournal mode, all journal mount options are illegal */
3954 if (test_opt2(sb, EXPLICIT_JOURNAL_CHECKSUM)) {
3955 ext4_msg(sb, KERN_ERR, "can't mount with "
3956 "journal_checksum, fs mounted w/o journal");
3957 goto failed_mount_wq;
3959 if (test_opt(sb, JOURNAL_ASYNC_COMMIT)) {
3960 ext4_msg(sb, KERN_ERR, "can't mount with "
3961 "journal_async_commit, fs mounted w/o journal");
3962 goto failed_mount_wq;
3964 if (sbi->s_commit_interval != JBD2_DEFAULT_MAX_COMMIT_AGE*HZ) {
3965 ext4_msg(sb, KERN_ERR, "can't mount with "
3966 "commit=%lu, fs mounted w/o journal",
3967 sbi->s_commit_interval / HZ);
3968 goto failed_mount_wq;
3970 if (EXT4_MOUNT_DATA_FLAGS &
3971 (sbi->s_mount_opt ^ sbi->s_def_mount_opt)) {
3972 ext4_msg(sb, KERN_ERR, "can't mount with "
3973 "data=, fs mounted w/o journal");
3974 goto failed_mount_wq;
3976 sbi->s_def_mount_opt &= EXT4_MOUNT_JOURNAL_CHECKSUM;
3977 clear_opt(sb, JOURNAL_CHECKSUM);
3978 clear_opt(sb, DATA_FLAGS);
3979 sbi->s_journal = NULL;
3980 needs_recovery = 0;
3981 goto no_journal;
3984 if (ext4_has_feature_64bit(sb) &&
3985 !jbd2_journal_set_features(EXT4_SB(sb)->s_journal, 0, 0,
3986 JBD2_FEATURE_INCOMPAT_64BIT)) {
3987 ext4_msg(sb, KERN_ERR, "Failed to set 64-bit journal feature");
3988 goto failed_mount_wq;
3991 if (!set_journal_csum_feature_set(sb)) {
3992 ext4_msg(sb, KERN_ERR, "Failed to set journal checksum "
3993 "feature set");
3994 goto failed_mount_wq;
3997 /* We have now updated the journal if required, so we can
3998 * validate the data journaling mode. */
3999 switch (test_opt(sb, DATA_FLAGS)) {
4000 case 0:
4001 /* No mode set, assume a default based on the journal
4002 * capabilities: ORDERED_DATA if the journal can
4003 * cope, else JOURNAL_DATA
4005 if (jbd2_journal_check_available_features
4006 (sbi->s_journal, 0, 0, JBD2_FEATURE_INCOMPAT_REVOKE))
4007 set_opt(sb, ORDERED_DATA);
4008 else
4009 set_opt(sb, JOURNAL_DATA);
4010 break;
4012 case EXT4_MOUNT_ORDERED_DATA:
4013 case EXT4_MOUNT_WRITEBACK_DATA:
4014 if (!jbd2_journal_check_available_features
4015 (sbi->s_journal, 0, 0, JBD2_FEATURE_INCOMPAT_REVOKE)) {
4016 ext4_msg(sb, KERN_ERR, "Journal does not support "
4017 "requested data journaling mode");
4018 goto failed_mount_wq;
4020 default:
4021 break;
4024 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA &&
4025 test_opt(sb, JOURNAL_ASYNC_COMMIT)) {
4026 ext4_msg(sb, KERN_ERR, "can't mount with "
4027 "journal_async_commit in data=ordered mode");
4028 goto failed_mount_wq;
4031 set_task_ioprio(sbi->s_journal->j_task, journal_ioprio);
4033 sbi->s_journal->j_commit_callback = ext4_journal_commit_callback;
4035 no_journal:
4036 sbi->s_mb_cache = ext4_xattr_create_cache();
4037 if (!sbi->s_mb_cache) {
4038 ext4_msg(sb, KERN_ERR, "Failed to create an mb_cache");
4039 goto failed_mount_wq;
4042 if ((DUMMY_ENCRYPTION_ENABLED(sbi) || ext4_has_feature_encrypt(sb)) &&
4043 (blocksize != PAGE_SIZE)) {
4044 ext4_msg(sb, KERN_ERR,
4045 "Unsupported blocksize for fs encryption");
4046 goto failed_mount_wq;
4049 if (DUMMY_ENCRYPTION_ENABLED(sbi) && !(sb->s_flags & MS_RDONLY) &&
4050 !ext4_has_feature_encrypt(sb)) {
4051 ext4_set_feature_encrypt(sb);
4052 ext4_commit_super(sb, 1);
4056 * Get the # of file system overhead blocks from the
4057 * superblock if present.
4059 if (es->s_overhead_clusters)
4060 sbi->s_overhead = le32_to_cpu(es->s_overhead_clusters);
4061 else {
4062 err = ext4_calculate_overhead(sb);
4063 if (err)
4064 goto failed_mount_wq;
4068 * The maximum number of concurrent works can be high and
4069 * concurrency isn't really necessary. Limit it to 1.
4071 EXT4_SB(sb)->rsv_conversion_wq =
4072 alloc_workqueue("ext4-rsv-conversion", WQ_MEM_RECLAIM | WQ_UNBOUND, 1);
4073 if (!EXT4_SB(sb)->rsv_conversion_wq) {
4074 printk(KERN_ERR "EXT4-fs: failed to create workqueue\n");
4075 ret = -ENOMEM;
4076 goto failed_mount4;
4080 * The jbd2_journal_load will have done any necessary log recovery,
4081 * so we can safely mount the rest of the filesystem now.
4084 root = ext4_iget(sb, EXT4_ROOT_INO);
4085 if (IS_ERR(root)) {
4086 ext4_msg(sb, KERN_ERR, "get root inode failed");
4087 ret = PTR_ERR(root);
4088 root = NULL;
4089 goto failed_mount4;
4091 if (!S_ISDIR(root->i_mode) || !root->i_blocks || !root->i_size) {
4092 ext4_msg(sb, KERN_ERR, "corrupt root inode, run e2fsck");
4093 iput(root);
4094 goto failed_mount4;
4096 sb->s_root = d_make_root(root);
4097 if (!sb->s_root) {
4098 ext4_msg(sb, KERN_ERR, "get root dentry failed");
4099 ret = -ENOMEM;
4100 goto failed_mount4;
4103 if (ext4_setup_super(sb, es, sb->s_flags & MS_RDONLY))
4104 sb->s_flags |= MS_RDONLY;
4106 /* determine the minimum size of new large inodes, if present */
4107 if (sbi->s_inode_size > EXT4_GOOD_OLD_INODE_SIZE &&
4108 sbi->s_want_extra_isize == 0) {
4109 sbi->s_want_extra_isize = sizeof(struct ext4_inode) -
4110 EXT4_GOOD_OLD_INODE_SIZE;
4111 if (ext4_has_feature_extra_isize(sb)) {
4112 if (sbi->s_want_extra_isize <
4113 le16_to_cpu(es->s_want_extra_isize))
4114 sbi->s_want_extra_isize =
4115 le16_to_cpu(es->s_want_extra_isize);
4116 if (sbi->s_want_extra_isize <
4117 le16_to_cpu(es->s_min_extra_isize))
4118 sbi->s_want_extra_isize =
4119 le16_to_cpu(es->s_min_extra_isize);
4122 /* Check if enough inode space is available */
4123 if (EXT4_GOOD_OLD_INODE_SIZE + sbi->s_want_extra_isize >
4124 sbi->s_inode_size) {
4125 sbi->s_want_extra_isize = sizeof(struct ext4_inode) -
4126 EXT4_GOOD_OLD_INODE_SIZE;
4127 ext4_msg(sb, KERN_INFO, "required extra inode space not"
4128 "available");
4131 ext4_set_resv_clusters(sb);
4133 err = ext4_setup_system_zone(sb);
4134 if (err) {
4135 ext4_msg(sb, KERN_ERR, "failed to initialize system "
4136 "zone (%d)", err);
4137 goto failed_mount4a;
4140 ext4_ext_init(sb);
4141 err = ext4_mb_init(sb);
4142 if (err) {
4143 ext4_msg(sb, KERN_ERR, "failed to initialize mballoc (%d)",
4144 err);
4145 goto failed_mount5;
4148 block = ext4_count_free_clusters(sb);
4149 ext4_free_blocks_count_set(sbi->s_es,
4150 EXT4_C2B(sbi, block));
4151 err = percpu_counter_init(&sbi->s_freeclusters_counter, block,
4152 GFP_KERNEL);
4153 if (!err) {
4154 unsigned long freei = ext4_count_free_inodes(sb);
4155 sbi->s_es->s_free_inodes_count = cpu_to_le32(freei);
4156 err = percpu_counter_init(&sbi->s_freeinodes_counter, freei,
4157 GFP_KERNEL);
4159 if (!err)
4160 err = percpu_counter_init(&sbi->s_dirs_counter,
4161 ext4_count_dirs(sb), GFP_KERNEL);
4162 if (!err)
4163 err = percpu_counter_init(&sbi->s_dirtyclusters_counter, 0,
4164 GFP_KERNEL);
4165 if (!err)
4166 err = percpu_init_rwsem(&sbi->s_journal_flag_rwsem);
4168 if (err) {
4169 ext4_msg(sb, KERN_ERR, "insufficient memory");
4170 goto failed_mount6;
4173 if (ext4_has_feature_flex_bg(sb))
4174 if (!ext4_fill_flex_info(sb)) {
4175 ext4_msg(sb, KERN_ERR,
4176 "unable to initialize "
4177 "flex_bg meta info!");
4178 goto failed_mount6;
4181 err = ext4_register_li_request(sb, first_not_zeroed);
4182 if (err)
4183 goto failed_mount6;
4185 err = ext4_register_sysfs(sb);
4186 if (err)
4187 goto failed_mount7;
4189 #ifdef CONFIG_QUOTA
4190 /* Enable quota usage during mount. */
4191 if (ext4_has_feature_quota(sb) && !(sb->s_flags & MS_RDONLY)) {
4192 err = ext4_enable_quotas(sb);
4193 if (err)
4194 goto failed_mount8;
4196 #endif /* CONFIG_QUOTA */
4198 EXT4_SB(sb)->s_mount_state |= EXT4_ORPHAN_FS;
4199 ext4_orphan_cleanup(sb, es);
4200 EXT4_SB(sb)->s_mount_state &= ~EXT4_ORPHAN_FS;
4201 if (needs_recovery) {
4202 ext4_msg(sb, KERN_INFO, "recovery complete");
4203 ext4_mark_recovery_complete(sb, es);
4205 if (EXT4_SB(sb)->s_journal) {
4206 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA)
4207 descr = " journalled data mode";
4208 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA)
4209 descr = " ordered data mode";
4210 else
4211 descr = " writeback data mode";
4212 } else
4213 descr = "out journal";
4215 if (test_opt(sb, DISCARD)) {
4216 struct request_queue *q = bdev_get_queue(sb->s_bdev);
4217 if (!blk_queue_discard(q))
4218 ext4_msg(sb, KERN_WARNING,
4219 "mounting with \"discard\" option, but "
4220 "the device does not support discard");
4223 if (___ratelimit(&ext4_mount_msg_ratelimit, "EXT4-fs mount"))
4224 ext4_msg(sb, KERN_INFO, "mounted filesystem with%s. "
4225 "Opts: %.*s%s%s", descr,
4226 (int) sizeof(sbi->s_es->s_mount_opts),
4227 sbi->s_es->s_mount_opts,
4228 *sbi->s_es->s_mount_opts ? "; " : "", orig_data);
4230 if (es->s_error_count)
4231 mod_timer(&sbi->s_err_report, jiffies + 300*HZ); /* 5 minutes */
4233 /* Enable message ratelimiting. Default is 10 messages per 5 secs. */
4234 ratelimit_state_init(&sbi->s_err_ratelimit_state, 5 * HZ, 10);
4235 ratelimit_state_init(&sbi->s_warning_ratelimit_state, 5 * HZ, 10);
4236 ratelimit_state_init(&sbi->s_msg_ratelimit_state, 5 * HZ, 10);
4238 kfree(orig_data);
4239 return 0;
4241 cantfind_ext4:
4242 if (!silent)
4243 ext4_msg(sb, KERN_ERR, "VFS: Can't find ext4 filesystem");
4244 goto failed_mount;
4246 #ifdef CONFIG_QUOTA
4247 failed_mount8:
4248 ext4_unregister_sysfs(sb);
4249 #endif
4250 failed_mount7:
4251 ext4_unregister_li_request(sb);
4252 failed_mount6:
4253 ext4_mb_release(sb);
4254 if (sbi->s_flex_groups)
4255 kvfree(sbi->s_flex_groups);
4256 percpu_counter_destroy(&sbi->s_freeclusters_counter);
4257 percpu_counter_destroy(&sbi->s_freeinodes_counter);
4258 percpu_counter_destroy(&sbi->s_dirs_counter);
4259 percpu_counter_destroy(&sbi->s_dirtyclusters_counter);
4260 failed_mount5:
4261 ext4_ext_release(sb);
4262 ext4_release_system_zone(sb);
4263 failed_mount4a:
4264 dput(sb->s_root);
4265 sb->s_root = NULL;
4266 failed_mount4:
4267 ext4_msg(sb, KERN_ERR, "mount failed");
4268 if (EXT4_SB(sb)->rsv_conversion_wq)
4269 destroy_workqueue(EXT4_SB(sb)->rsv_conversion_wq);
4270 failed_mount_wq:
4271 if (sbi->s_mb_cache) {
4272 ext4_xattr_destroy_cache(sbi->s_mb_cache);
4273 sbi->s_mb_cache = NULL;
4275 if (sbi->s_journal) {
4276 jbd2_journal_destroy(sbi->s_journal);
4277 sbi->s_journal = NULL;
4279 failed_mount3a:
4280 ext4_es_unregister_shrinker(sbi);
4281 failed_mount3:
4282 del_timer_sync(&sbi->s_err_report);
4283 if (sbi->s_mmp_tsk)
4284 kthread_stop(sbi->s_mmp_tsk);
4285 failed_mount2:
4286 for (i = 0; i < db_count; i++)
4287 brelse(sbi->s_group_desc[i]);
4288 kvfree(sbi->s_group_desc);
4289 failed_mount:
4290 if (sbi->s_chksum_driver)
4291 crypto_free_shash(sbi->s_chksum_driver);
4292 #ifdef CONFIG_QUOTA
4293 for (i = 0; i < EXT4_MAXQUOTAS; i++)
4294 kfree(sbi->s_qf_names[i]);
4295 #endif
4296 ext4_blkdev_remove(sbi);
4297 brelse(bh);
4298 out_fail:
4299 sb->s_fs_info = NULL;
4300 kfree(sbi->s_blockgroup_lock);
4301 out_free_base:
4302 kfree(sbi);
4303 kfree(orig_data);
4304 return err ? err : ret;
4308 * Setup any per-fs journal parameters now. We'll do this both on
4309 * initial mount, once the journal has been initialised but before we've
4310 * done any recovery; and again on any subsequent remount.
4312 static void ext4_init_journal_params(struct super_block *sb, journal_t *journal)
4314 struct ext4_sb_info *sbi = EXT4_SB(sb);
4316 journal->j_commit_interval = sbi->s_commit_interval;
4317 journal->j_min_batch_time = sbi->s_min_batch_time;
4318 journal->j_max_batch_time = sbi->s_max_batch_time;
4320 write_lock(&journal->j_state_lock);
4321 if (test_opt(sb, BARRIER))
4322 journal->j_flags |= JBD2_BARRIER;
4323 else
4324 journal->j_flags &= ~JBD2_BARRIER;
4325 if (test_opt(sb, DATA_ERR_ABORT))
4326 journal->j_flags |= JBD2_ABORT_ON_SYNCDATA_ERR;
4327 else
4328 journal->j_flags &= ~JBD2_ABORT_ON_SYNCDATA_ERR;
4329 write_unlock(&journal->j_state_lock);
4332 static struct inode *ext4_get_journal_inode(struct super_block *sb,
4333 unsigned int journal_inum)
4335 struct inode *journal_inode;
4338 * Test for the existence of a valid inode on disk. Bad things
4339 * happen if we iget() an unused inode, as the subsequent iput()
4340 * will try to delete it.
4342 journal_inode = ext4_iget(sb, journal_inum);
4343 if (IS_ERR(journal_inode)) {
4344 ext4_msg(sb, KERN_ERR, "no journal found");
4345 return NULL;
4347 if (!journal_inode->i_nlink) {
4348 make_bad_inode(journal_inode);
4349 iput(journal_inode);
4350 ext4_msg(sb, KERN_ERR, "journal inode is deleted");
4351 return NULL;
4354 jbd_debug(2, "Journal inode found at %p: %lld bytes\n",
4355 journal_inode, journal_inode->i_size);
4356 if (!S_ISREG(journal_inode->i_mode)) {
4357 ext4_msg(sb, KERN_ERR, "invalid journal inode");
4358 iput(journal_inode);
4359 return NULL;
4361 return journal_inode;
4364 static journal_t *ext4_get_journal(struct super_block *sb,
4365 unsigned int journal_inum)
4367 struct inode *journal_inode;
4368 journal_t *journal;
4370 BUG_ON(!ext4_has_feature_journal(sb));
4372 journal_inode = ext4_get_journal_inode(sb, journal_inum);
4373 if (!journal_inode)
4374 return NULL;
4376 journal = jbd2_journal_init_inode(journal_inode);
4377 if (!journal) {
4378 ext4_msg(sb, KERN_ERR, "Could not load journal inode");
4379 iput(journal_inode);
4380 return NULL;
4382 journal->j_private = sb;
4383 ext4_init_journal_params(sb, journal);
4384 return journal;
4387 static journal_t *ext4_get_dev_journal(struct super_block *sb,
4388 dev_t j_dev)
4390 struct buffer_head *bh;
4391 journal_t *journal;
4392 ext4_fsblk_t start;
4393 ext4_fsblk_t len;
4394 int hblock, blocksize;
4395 ext4_fsblk_t sb_block;
4396 unsigned long offset;
4397 struct ext4_super_block *es;
4398 struct block_device *bdev;
4400 BUG_ON(!ext4_has_feature_journal(sb));
4402 bdev = ext4_blkdev_get(j_dev, sb);
4403 if (bdev == NULL)
4404 return NULL;
4406 blocksize = sb->s_blocksize;
4407 hblock = bdev_logical_block_size(bdev);
4408 if (blocksize < hblock) {
4409 ext4_msg(sb, KERN_ERR,
4410 "blocksize too small for journal device");
4411 goto out_bdev;
4414 sb_block = EXT4_MIN_BLOCK_SIZE / blocksize;
4415 offset = EXT4_MIN_BLOCK_SIZE % blocksize;
4416 set_blocksize(bdev, blocksize);
4417 if (!(bh = __bread(bdev, sb_block, blocksize))) {
4418 ext4_msg(sb, KERN_ERR, "couldn't read superblock of "
4419 "external journal");
4420 goto out_bdev;
4423 es = (struct ext4_super_block *) (bh->b_data + offset);
4424 if ((le16_to_cpu(es->s_magic) != EXT4_SUPER_MAGIC) ||
4425 !(le32_to_cpu(es->s_feature_incompat) &
4426 EXT4_FEATURE_INCOMPAT_JOURNAL_DEV)) {
4427 ext4_msg(sb, KERN_ERR, "external journal has "
4428 "bad superblock");
4429 brelse(bh);
4430 goto out_bdev;
4433 if ((le32_to_cpu(es->s_feature_ro_compat) &
4434 EXT4_FEATURE_RO_COMPAT_METADATA_CSUM) &&
4435 es->s_checksum != ext4_superblock_csum(sb, es)) {
4436 ext4_msg(sb, KERN_ERR, "external journal has "
4437 "corrupt superblock");
4438 brelse(bh);
4439 goto out_bdev;
4442 if (memcmp(EXT4_SB(sb)->s_es->s_journal_uuid, es->s_uuid, 16)) {
4443 ext4_msg(sb, KERN_ERR, "journal UUID does not match");
4444 brelse(bh);
4445 goto out_bdev;
4448 len = ext4_blocks_count(es);
4449 start = sb_block + 1;
4450 brelse(bh); /* we're done with the superblock */
4452 journal = jbd2_journal_init_dev(bdev, sb->s_bdev,
4453 start, len, blocksize);
4454 if (!journal) {
4455 ext4_msg(sb, KERN_ERR, "failed to create device journal");
4456 goto out_bdev;
4458 journal->j_private = sb;
4459 ll_rw_block(REQ_OP_READ, REQ_META | REQ_PRIO, 1, &journal->j_sb_buffer);
4460 wait_on_buffer(journal->j_sb_buffer);
4461 if (!buffer_uptodate(journal->j_sb_buffer)) {
4462 ext4_msg(sb, KERN_ERR, "I/O error on journal device");
4463 goto out_journal;
4465 if (be32_to_cpu(journal->j_superblock->s_nr_users) != 1) {
4466 ext4_msg(sb, KERN_ERR, "External journal has more than one "
4467 "user (unsupported) - %d",
4468 be32_to_cpu(journal->j_superblock->s_nr_users));
4469 goto out_journal;
4471 EXT4_SB(sb)->journal_bdev = bdev;
4472 ext4_init_journal_params(sb, journal);
4473 return journal;
4475 out_journal:
4476 jbd2_journal_destroy(journal);
4477 out_bdev:
4478 ext4_blkdev_put(bdev);
4479 return NULL;
4482 static int ext4_load_journal(struct super_block *sb,
4483 struct ext4_super_block *es,
4484 unsigned long journal_devnum)
4486 journal_t *journal;
4487 unsigned int journal_inum = le32_to_cpu(es->s_journal_inum);
4488 dev_t journal_dev;
4489 int err = 0;
4490 int really_read_only;
4492 BUG_ON(!ext4_has_feature_journal(sb));
4494 if (journal_devnum &&
4495 journal_devnum != le32_to_cpu(es->s_journal_dev)) {
4496 ext4_msg(sb, KERN_INFO, "external journal device major/minor "
4497 "numbers have changed");
4498 journal_dev = new_decode_dev(journal_devnum);
4499 } else
4500 journal_dev = new_decode_dev(le32_to_cpu(es->s_journal_dev));
4502 really_read_only = bdev_read_only(sb->s_bdev);
4505 * Are we loading a blank journal or performing recovery after a
4506 * crash? For recovery, we need to check in advance whether we
4507 * can get read-write access to the device.
4509 if (ext4_has_feature_journal_needs_recovery(sb)) {
4510 if (sb->s_flags & MS_RDONLY) {
4511 ext4_msg(sb, KERN_INFO, "INFO: recovery "
4512 "required on readonly filesystem");
4513 if (really_read_only) {
4514 ext4_msg(sb, KERN_ERR, "write access "
4515 "unavailable, cannot proceed");
4516 return -EROFS;
4518 ext4_msg(sb, KERN_INFO, "write access will "
4519 "be enabled during recovery");
4523 if (journal_inum && journal_dev) {
4524 ext4_msg(sb, KERN_ERR, "filesystem has both journal "
4525 "and inode journals!");
4526 return -EINVAL;
4529 if (journal_inum) {
4530 if (!(journal = ext4_get_journal(sb, journal_inum)))
4531 return -EINVAL;
4532 } else {
4533 if (!(journal = ext4_get_dev_journal(sb, journal_dev)))
4534 return -EINVAL;
4537 if (!(journal->j_flags & JBD2_BARRIER))
4538 ext4_msg(sb, KERN_INFO, "barriers disabled");
4540 if (!ext4_has_feature_journal_needs_recovery(sb))
4541 err = jbd2_journal_wipe(journal, !really_read_only);
4542 if (!err) {
4543 char *save = kmalloc(EXT4_S_ERR_LEN, GFP_KERNEL);
4544 if (save)
4545 memcpy(save, ((char *) es) +
4546 EXT4_S_ERR_START, EXT4_S_ERR_LEN);
4547 err = jbd2_journal_load(journal);
4548 if (save)
4549 memcpy(((char *) es) + EXT4_S_ERR_START,
4550 save, EXT4_S_ERR_LEN);
4551 kfree(save);
4554 if (err) {
4555 ext4_msg(sb, KERN_ERR, "error loading journal");
4556 jbd2_journal_destroy(journal);
4557 return err;
4560 EXT4_SB(sb)->s_journal = journal;
4561 ext4_clear_journal_err(sb, es);
4563 if (!really_read_only && journal_devnum &&
4564 journal_devnum != le32_to_cpu(es->s_journal_dev)) {
4565 es->s_journal_dev = cpu_to_le32(journal_devnum);
4567 /* Make sure we flush the recovery flag to disk. */
4568 ext4_commit_super(sb, 1);
4571 return 0;
4574 static int ext4_commit_super(struct super_block *sb, int sync)
4576 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
4577 struct buffer_head *sbh = EXT4_SB(sb)->s_sbh;
4578 int error = 0;
4580 if (!sbh || block_device_ejected(sb))
4581 return error;
4583 * If the file system is mounted read-only, don't update the
4584 * superblock write time. This avoids updating the superblock
4585 * write time when we are mounting the root file system
4586 * read/only but we need to replay the journal; at that point,
4587 * for people who are east of GMT and who make their clock
4588 * tick in localtime for Windows bug-for-bug compatibility,
4589 * the clock is set in the future, and this will cause e2fsck
4590 * to complain and force a full file system check.
4592 if (!(sb->s_flags & MS_RDONLY))
4593 es->s_wtime = cpu_to_le32(get_seconds());
4594 if (sb->s_bdev->bd_part)
4595 es->s_kbytes_written =
4596 cpu_to_le64(EXT4_SB(sb)->s_kbytes_written +
4597 ((part_stat_read(sb->s_bdev->bd_part, sectors[1]) -
4598 EXT4_SB(sb)->s_sectors_written_start) >> 1));
4599 else
4600 es->s_kbytes_written =
4601 cpu_to_le64(EXT4_SB(sb)->s_kbytes_written);
4602 if (percpu_counter_initialized(&EXT4_SB(sb)->s_freeclusters_counter))
4603 ext4_free_blocks_count_set(es,
4604 EXT4_C2B(EXT4_SB(sb), percpu_counter_sum_positive(
4605 &EXT4_SB(sb)->s_freeclusters_counter)));
4606 if (percpu_counter_initialized(&EXT4_SB(sb)->s_freeinodes_counter))
4607 es->s_free_inodes_count =
4608 cpu_to_le32(percpu_counter_sum_positive(
4609 &EXT4_SB(sb)->s_freeinodes_counter));
4610 BUFFER_TRACE(sbh, "marking dirty");
4611 ext4_superblock_csum_set(sb);
4612 if (sync)
4613 lock_buffer(sbh);
4614 if (buffer_write_io_error(sbh)) {
4616 * Oh, dear. A previous attempt to write the
4617 * superblock failed. This could happen because the
4618 * USB device was yanked out. Or it could happen to
4619 * be a transient write error and maybe the block will
4620 * be remapped. Nothing we can do but to retry the
4621 * write and hope for the best.
4623 ext4_msg(sb, KERN_ERR, "previous I/O error to "
4624 "superblock detected");
4625 clear_buffer_write_io_error(sbh);
4626 set_buffer_uptodate(sbh);
4628 mark_buffer_dirty(sbh);
4629 if (sync) {
4630 unlock_buffer(sbh);
4631 error = __sync_dirty_buffer(sbh,
4632 test_opt(sb, BARRIER) ? REQ_FUA : REQ_SYNC);
4633 if (error)
4634 return error;
4636 error = buffer_write_io_error(sbh);
4637 if (error) {
4638 ext4_msg(sb, KERN_ERR, "I/O error while writing "
4639 "superblock");
4640 clear_buffer_write_io_error(sbh);
4641 set_buffer_uptodate(sbh);
4644 return error;
4648 * Have we just finished recovery? If so, and if we are mounting (or
4649 * remounting) the filesystem readonly, then we will end up with a
4650 * consistent fs on disk. Record that fact.
4652 static void ext4_mark_recovery_complete(struct super_block *sb,
4653 struct ext4_super_block *es)
4655 journal_t *journal = EXT4_SB(sb)->s_journal;
4657 if (!ext4_has_feature_journal(sb)) {
4658 BUG_ON(journal != NULL);
4659 return;
4661 jbd2_journal_lock_updates(journal);
4662 if (jbd2_journal_flush(journal) < 0)
4663 goto out;
4665 if (ext4_has_feature_journal_needs_recovery(sb) &&
4666 sb->s_flags & MS_RDONLY) {
4667 ext4_clear_feature_journal_needs_recovery(sb);
4668 ext4_commit_super(sb, 1);
4671 out:
4672 jbd2_journal_unlock_updates(journal);
4676 * If we are mounting (or read-write remounting) a filesystem whose journal
4677 * has recorded an error from a previous lifetime, move that error to the
4678 * main filesystem now.
4680 static void ext4_clear_journal_err(struct super_block *sb,
4681 struct ext4_super_block *es)
4683 journal_t *journal;
4684 int j_errno;
4685 const char *errstr;
4687 BUG_ON(!ext4_has_feature_journal(sb));
4689 journal = EXT4_SB(sb)->s_journal;
4692 * Now check for any error status which may have been recorded in the
4693 * journal by a prior ext4_error() or ext4_abort()
4696 j_errno = jbd2_journal_errno(journal);
4697 if (j_errno) {
4698 char nbuf[16];
4700 errstr = ext4_decode_error(sb, j_errno, nbuf);
4701 ext4_warning(sb, "Filesystem error recorded "
4702 "from previous mount: %s", errstr);
4703 ext4_warning(sb, "Marking fs in need of filesystem check.");
4705 EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS;
4706 es->s_state |= cpu_to_le16(EXT4_ERROR_FS);
4707 ext4_commit_super(sb, 1);
4709 jbd2_journal_clear_err(journal);
4710 jbd2_journal_update_sb_errno(journal);
4715 * Force the running and committing transactions to commit,
4716 * and wait on the commit.
4718 int ext4_force_commit(struct super_block *sb)
4720 journal_t *journal;
4722 if (sb->s_flags & MS_RDONLY)
4723 return 0;
4725 journal = EXT4_SB(sb)->s_journal;
4726 return ext4_journal_force_commit(journal);
4729 static int ext4_sync_fs(struct super_block *sb, int wait)
4731 int ret = 0;
4732 tid_t target;
4733 bool needs_barrier = false;
4734 struct ext4_sb_info *sbi = EXT4_SB(sb);
4736 if (unlikely(ext4_forced_shutdown(EXT4_SB(sb))))
4737 return 0;
4739 trace_ext4_sync_fs(sb, wait);
4740 flush_workqueue(sbi->rsv_conversion_wq);
4742 * Writeback quota in non-journalled quota case - journalled quota has
4743 * no dirty dquots
4745 dquot_writeback_dquots(sb, -1);
4747 * Data writeback is possible w/o journal transaction, so barrier must
4748 * being sent at the end of the function. But we can skip it if
4749 * transaction_commit will do it for us.
4751 if (sbi->s_journal) {
4752 target = jbd2_get_latest_transaction(sbi->s_journal);
4753 if (wait && sbi->s_journal->j_flags & JBD2_BARRIER &&
4754 !jbd2_trans_will_send_data_barrier(sbi->s_journal, target))
4755 needs_barrier = true;
4757 if (jbd2_journal_start_commit(sbi->s_journal, &target)) {
4758 if (wait)
4759 ret = jbd2_log_wait_commit(sbi->s_journal,
4760 target);
4762 } else if (wait && test_opt(sb, BARRIER))
4763 needs_barrier = true;
4764 if (needs_barrier) {
4765 int err;
4766 err = blkdev_issue_flush(sb->s_bdev, GFP_KERNEL, NULL);
4767 if (!ret)
4768 ret = err;
4771 return ret;
4775 * LVM calls this function before a (read-only) snapshot is created. This
4776 * gives us a chance to flush the journal completely and mark the fs clean.
4778 * Note that only this function cannot bring a filesystem to be in a clean
4779 * state independently. It relies on upper layer to stop all data & metadata
4780 * modifications.
4782 static int ext4_freeze(struct super_block *sb)
4784 int error = 0;
4785 journal_t *journal;
4787 if (sb->s_flags & MS_RDONLY)
4788 return 0;
4790 journal = EXT4_SB(sb)->s_journal;
4792 if (journal) {
4793 /* Now we set up the journal barrier. */
4794 jbd2_journal_lock_updates(journal);
4797 * Don't clear the needs_recovery flag if we failed to
4798 * flush the journal.
4800 error = jbd2_journal_flush(journal);
4801 if (error < 0)
4802 goto out;
4804 /* Journal blocked and flushed, clear needs_recovery flag. */
4805 ext4_clear_feature_journal_needs_recovery(sb);
4808 error = ext4_commit_super(sb, 1);
4809 out:
4810 if (journal)
4811 /* we rely on upper layer to stop further updates */
4812 jbd2_journal_unlock_updates(journal);
4813 return error;
4817 * Called by LVM after the snapshot is done. We need to reset the RECOVER
4818 * flag here, even though the filesystem is not technically dirty yet.
4820 static int ext4_unfreeze(struct super_block *sb)
4822 if ((sb->s_flags & MS_RDONLY) || ext4_forced_shutdown(EXT4_SB(sb)))
4823 return 0;
4825 if (EXT4_SB(sb)->s_journal) {
4826 /* Reset the needs_recovery flag before the fs is unlocked. */
4827 ext4_set_feature_journal_needs_recovery(sb);
4830 ext4_commit_super(sb, 1);
4831 return 0;
4835 * Structure to save mount options for ext4_remount's benefit
4837 struct ext4_mount_options {
4838 unsigned long s_mount_opt;
4839 unsigned long s_mount_opt2;
4840 kuid_t s_resuid;
4841 kgid_t s_resgid;
4842 unsigned long s_commit_interval;
4843 u32 s_min_batch_time, s_max_batch_time;
4844 #ifdef CONFIG_QUOTA
4845 int s_jquota_fmt;
4846 char *s_qf_names[EXT4_MAXQUOTAS];
4847 #endif
4850 static int ext4_remount(struct super_block *sb, int *flags, char *data)
4852 struct ext4_super_block *es;
4853 struct ext4_sb_info *sbi = EXT4_SB(sb);
4854 unsigned long old_sb_flags;
4855 struct ext4_mount_options old_opts;
4856 int enable_quota = 0;
4857 ext4_group_t g;
4858 unsigned int journal_ioprio = DEFAULT_JOURNAL_IOPRIO;
4859 int err = 0;
4860 #ifdef CONFIG_QUOTA
4861 int i, j;
4862 #endif
4863 char *orig_data = kstrdup(data, GFP_KERNEL);
4865 /* Store the original options */
4866 old_sb_flags = sb->s_flags;
4867 old_opts.s_mount_opt = sbi->s_mount_opt;
4868 old_opts.s_mount_opt2 = sbi->s_mount_opt2;
4869 old_opts.s_resuid = sbi->s_resuid;
4870 old_opts.s_resgid = sbi->s_resgid;
4871 old_opts.s_commit_interval = sbi->s_commit_interval;
4872 old_opts.s_min_batch_time = sbi->s_min_batch_time;
4873 old_opts.s_max_batch_time = sbi->s_max_batch_time;
4874 #ifdef CONFIG_QUOTA
4875 old_opts.s_jquota_fmt = sbi->s_jquota_fmt;
4876 for (i = 0; i < EXT4_MAXQUOTAS; i++)
4877 if (sbi->s_qf_names[i]) {
4878 old_opts.s_qf_names[i] = kstrdup(sbi->s_qf_names[i],
4879 GFP_KERNEL);
4880 if (!old_opts.s_qf_names[i]) {
4881 for (j = 0; j < i; j++)
4882 kfree(old_opts.s_qf_names[j]);
4883 kfree(orig_data);
4884 return -ENOMEM;
4886 } else
4887 old_opts.s_qf_names[i] = NULL;
4888 #endif
4889 if (sbi->s_journal && sbi->s_journal->j_task->io_context)
4890 journal_ioprio = sbi->s_journal->j_task->io_context->ioprio;
4892 if (!parse_options(data, sb, NULL, &journal_ioprio, 1)) {
4893 err = -EINVAL;
4894 goto restore_opts;
4897 if ((old_opts.s_mount_opt & EXT4_MOUNT_JOURNAL_CHECKSUM) ^
4898 test_opt(sb, JOURNAL_CHECKSUM)) {
4899 ext4_msg(sb, KERN_ERR, "changing journal_checksum "
4900 "during remount not supported; ignoring");
4901 sbi->s_mount_opt ^= EXT4_MOUNT_JOURNAL_CHECKSUM;
4904 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA) {
4905 if (test_opt2(sb, EXPLICIT_DELALLOC)) {
4906 ext4_msg(sb, KERN_ERR, "can't mount with "
4907 "both data=journal and delalloc");
4908 err = -EINVAL;
4909 goto restore_opts;
4911 if (test_opt(sb, DIOREAD_NOLOCK)) {
4912 ext4_msg(sb, KERN_ERR, "can't mount with "
4913 "both data=journal and dioread_nolock");
4914 err = -EINVAL;
4915 goto restore_opts;
4917 if (test_opt(sb, DAX)) {
4918 ext4_msg(sb, KERN_ERR, "can't mount with "
4919 "both data=journal and dax");
4920 err = -EINVAL;
4921 goto restore_opts;
4923 } else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA) {
4924 if (test_opt(sb, JOURNAL_ASYNC_COMMIT)) {
4925 ext4_msg(sb, KERN_ERR, "can't mount with "
4926 "journal_async_commit in data=ordered mode");
4927 err = -EINVAL;
4928 goto restore_opts;
4932 if ((sbi->s_mount_opt ^ old_opts.s_mount_opt) & EXT4_MOUNT_DAX) {
4933 ext4_msg(sb, KERN_WARNING, "warning: refusing change of "
4934 "dax flag with busy inodes while remounting");
4935 sbi->s_mount_opt ^= EXT4_MOUNT_DAX;
4938 if (sbi->s_mount_flags & EXT4_MF_FS_ABORTED)
4939 ext4_abort(sb, "Abort forced by user");
4941 sb->s_flags = (sb->s_flags & ~MS_POSIXACL) |
4942 (test_opt(sb, POSIX_ACL) ? MS_POSIXACL : 0);
4944 es = sbi->s_es;
4946 if (sbi->s_journal) {
4947 ext4_init_journal_params(sb, sbi->s_journal);
4948 set_task_ioprio(sbi->s_journal->j_task, journal_ioprio);
4951 if (*flags & MS_LAZYTIME)
4952 sb->s_flags |= MS_LAZYTIME;
4954 if ((*flags & MS_RDONLY) != (sb->s_flags & MS_RDONLY)) {
4955 if (sbi->s_mount_flags & EXT4_MF_FS_ABORTED) {
4956 err = -EROFS;
4957 goto restore_opts;
4960 if (*flags & MS_RDONLY) {
4961 err = sync_filesystem(sb);
4962 if (err < 0)
4963 goto restore_opts;
4964 err = dquot_suspend(sb, -1);
4965 if (err < 0)
4966 goto restore_opts;
4969 * First of all, the unconditional stuff we have to do
4970 * to disable replay of the journal when we next remount
4972 sb->s_flags |= MS_RDONLY;
4975 * OK, test if we are remounting a valid rw partition
4976 * readonly, and if so set the rdonly flag and then
4977 * mark the partition as valid again.
4979 if (!(es->s_state & cpu_to_le16(EXT4_VALID_FS)) &&
4980 (sbi->s_mount_state & EXT4_VALID_FS))
4981 es->s_state = cpu_to_le16(sbi->s_mount_state);
4983 if (sbi->s_journal)
4984 ext4_mark_recovery_complete(sb, es);
4985 } else {
4986 /* Make sure we can mount this feature set readwrite */
4987 if (ext4_has_feature_readonly(sb) ||
4988 !ext4_feature_set_ok(sb, 0)) {
4989 err = -EROFS;
4990 goto restore_opts;
4993 * Make sure the group descriptor checksums
4994 * are sane. If they aren't, refuse to remount r/w.
4996 for (g = 0; g < sbi->s_groups_count; g++) {
4997 struct ext4_group_desc *gdp =
4998 ext4_get_group_desc(sb, g, NULL);
5000 if (!ext4_group_desc_csum_verify(sb, g, gdp)) {
5001 ext4_msg(sb, KERN_ERR,
5002 "ext4_remount: Checksum for group %u failed (%u!=%u)",
5003 g, le16_to_cpu(ext4_group_desc_csum(sb, g, gdp)),
5004 le16_to_cpu(gdp->bg_checksum));
5005 err = -EFSBADCRC;
5006 goto restore_opts;
5011 * If we have an unprocessed orphan list hanging
5012 * around from a previously readonly bdev mount,
5013 * require a full umount/remount for now.
5015 if (es->s_last_orphan) {
5016 ext4_msg(sb, KERN_WARNING, "Couldn't "
5017 "remount RDWR because of unprocessed "
5018 "orphan inode list. Please "
5019 "umount/remount instead");
5020 err = -EINVAL;
5021 goto restore_opts;
5025 * Mounting a RDONLY partition read-write, so reread
5026 * and store the current valid flag. (It may have
5027 * been changed by e2fsck since we originally mounted
5028 * the partition.)
5030 if (sbi->s_journal)
5031 ext4_clear_journal_err(sb, es);
5032 sbi->s_mount_state = le16_to_cpu(es->s_state);
5033 if (!ext4_setup_super(sb, es, 0))
5034 sb->s_flags &= ~MS_RDONLY;
5035 if (ext4_has_feature_mmp(sb))
5036 if (ext4_multi_mount_protect(sb,
5037 le64_to_cpu(es->s_mmp_block))) {
5038 err = -EROFS;
5039 goto restore_opts;
5041 enable_quota = 1;
5046 * Reinitialize lazy itable initialization thread based on
5047 * current settings
5049 if ((sb->s_flags & MS_RDONLY) || !test_opt(sb, INIT_INODE_TABLE))
5050 ext4_unregister_li_request(sb);
5051 else {
5052 ext4_group_t first_not_zeroed;
5053 first_not_zeroed = ext4_has_uninit_itable(sb);
5054 ext4_register_li_request(sb, first_not_zeroed);
5057 ext4_setup_system_zone(sb);
5058 if (sbi->s_journal == NULL && !(old_sb_flags & MS_RDONLY))
5059 ext4_commit_super(sb, 1);
5061 #ifdef CONFIG_QUOTA
5062 /* Release old quota file names */
5063 for (i = 0; i < EXT4_MAXQUOTAS; i++)
5064 kfree(old_opts.s_qf_names[i]);
5065 if (enable_quota) {
5066 if (sb_any_quota_suspended(sb))
5067 dquot_resume(sb, -1);
5068 else if (ext4_has_feature_quota(sb)) {
5069 err = ext4_enable_quotas(sb);
5070 if (err)
5071 goto restore_opts;
5074 #endif
5076 *flags = (*flags & ~MS_LAZYTIME) | (sb->s_flags & MS_LAZYTIME);
5077 ext4_msg(sb, KERN_INFO, "re-mounted. Opts: %s", orig_data);
5078 kfree(orig_data);
5079 return 0;
5081 restore_opts:
5082 sb->s_flags = old_sb_flags;
5083 sbi->s_mount_opt = old_opts.s_mount_opt;
5084 sbi->s_mount_opt2 = old_opts.s_mount_opt2;
5085 sbi->s_resuid = old_opts.s_resuid;
5086 sbi->s_resgid = old_opts.s_resgid;
5087 sbi->s_commit_interval = old_opts.s_commit_interval;
5088 sbi->s_min_batch_time = old_opts.s_min_batch_time;
5089 sbi->s_max_batch_time = old_opts.s_max_batch_time;
5090 #ifdef CONFIG_QUOTA
5091 sbi->s_jquota_fmt = old_opts.s_jquota_fmt;
5092 for (i = 0; i < EXT4_MAXQUOTAS; i++) {
5093 kfree(sbi->s_qf_names[i]);
5094 sbi->s_qf_names[i] = old_opts.s_qf_names[i];
5096 #endif
5097 kfree(orig_data);
5098 return err;
5101 #ifdef CONFIG_QUOTA
5102 static int ext4_statfs_project(struct super_block *sb,
5103 kprojid_t projid, struct kstatfs *buf)
5105 struct kqid qid;
5106 struct dquot *dquot;
5107 u64 limit;
5108 u64 curblock;
5110 qid = make_kqid_projid(projid);
5111 dquot = dqget(sb, qid);
5112 if (IS_ERR(dquot))
5113 return PTR_ERR(dquot);
5114 spin_lock(&dq_data_lock);
5116 limit = (dquot->dq_dqb.dqb_bsoftlimit ?
5117 dquot->dq_dqb.dqb_bsoftlimit :
5118 dquot->dq_dqb.dqb_bhardlimit) >> sb->s_blocksize_bits;
5119 if (limit && buf->f_blocks > limit) {
5120 curblock = dquot->dq_dqb.dqb_curspace >> sb->s_blocksize_bits;
5121 buf->f_blocks = limit;
5122 buf->f_bfree = buf->f_bavail =
5123 (buf->f_blocks > curblock) ?
5124 (buf->f_blocks - curblock) : 0;
5127 limit = dquot->dq_dqb.dqb_isoftlimit ?
5128 dquot->dq_dqb.dqb_isoftlimit :
5129 dquot->dq_dqb.dqb_ihardlimit;
5130 if (limit && buf->f_files > limit) {
5131 buf->f_files = limit;
5132 buf->f_ffree =
5133 (buf->f_files > dquot->dq_dqb.dqb_curinodes) ?
5134 (buf->f_files - dquot->dq_dqb.dqb_curinodes) : 0;
5137 spin_unlock(&dq_data_lock);
5138 dqput(dquot);
5139 return 0;
5141 #endif
5143 static int ext4_statfs(struct dentry *dentry, struct kstatfs *buf)
5145 struct super_block *sb = dentry->d_sb;
5146 struct ext4_sb_info *sbi = EXT4_SB(sb);
5147 struct ext4_super_block *es = sbi->s_es;
5148 ext4_fsblk_t overhead = 0, resv_blocks;
5149 u64 fsid;
5150 s64 bfree;
5151 resv_blocks = EXT4_C2B(sbi, atomic64_read(&sbi->s_resv_clusters));
5153 if (!test_opt(sb, MINIX_DF))
5154 overhead = sbi->s_overhead;
5156 buf->f_type = EXT4_SUPER_MAGIC;
5157 buf->f_bsize = sb->s_blocksize;
5158 buf->f_blocks = ext4_blocks_count(es) - EXT4_C2B(sbi, overhead);
5159 bfree = percpu_counter_sum_positive(&sbi->s_freeclusters_counter) -
5160 percpu_counter_sum_positive(&sbi->s_dirtyclusters_counter);
5161 /* prevent underflow in case that few free space is available */
5162 buf->f_bfree = EXT4_C2B(sbi, max_t(s64, bfree, 0));
5163 buf->f_bavail = buf->f_bfree -
5164 (ext4_r_blocks_count(es) + resv_blocks);
5165 if (buf->f_bfree < (ext4_r_blocks_count(es) + resv_blocks))
5166 buf->f_bavail = 0;
5167 buf->f_files = le32_to_cpu(es->s_inodes_count);
5168 buf->f_ffree = percpu_counter_sum_positive(&sbi->s_freeinodes_counter);
5169 buf->f_namelen = EXT4_NAME_LEN;
5170 fsid = le64_to_cpup((void *)es->s_uuid) ^
5171 le64_to_cpup((void *)es->s_uuid + sizeof(u64));
5172 buf->f_fsid.val[0] = fsid & 0xFFFFFFFFUL;
5173 buf->f_fsid.val[1] = (fsid >> 32) & 0xFFFFFFFFUL;
5175 #ifdef CONFIG_QUOTA
5176 if (ext4_test_inode_flag(dentry->d_inode, EXT4_INODE_PROJINHERIT) &&
5177 sb_has_quota_limits_enabled(sb, PRJQUOTA))
5178 ext4_statfs_project(sb, EXT4_I(dentry->d_inode)->i_projid, buf);
5179 #endif
5180 return 0;
5183 /* Helper function for writing quotas on sync - we need to start transaction
5184 * before quota file is locked for write. Otherwise the are possible deadlocks:
5185 * Process 1 Process 2
5186 * ext4_create() quota_sync()
5187 * jbd2_journal_start() write_dquot()
5188 * dquot_initialize() down(dqio_mutex)
5189 * down(dqio_mutex) jbd2_journal_start()
5193 #ifdef CONFIG_QUOTA
5195 static inline struct inode *dquot_to_inode(struct dquot *dquot)
5197 return sb_dqopt(dquot->dq_sb)->files[dquot->dq_id.type];
5200 static int ext4_write_dquot(struct dquot *dquot)
5202 int ret, err;
5203 handle_t *handle;
5204 struct inode *inode;
5206 inode = dquot_to_inode(dquot);
5207 handle = ext4_journal_start(inode, EXT4_HT_QUOTA,
5208 EXT4_QUOTA_TRANS_BLOCKS(dquot->dq_sb));
5209 if (IS_ERR(handle))
5210 return PTR_ERR(handle);
5211 ret = dquot_commit(dquot);
5212 err = ext4_journal_stop(handle);
5213 if (!ret)
5214 ret = err;
5215 return ret;
5218 static int ext4_acquire_dquot(struct dquot *dquot)
5220 int ret, err;
5221 handle_t *handle;
5223 handle = ext4_journal_start(dquot_to_inode(dquot), EXT4_HT_QUOTA,
5224 EXT4_QUOTA_INIT_BLOCKS(dquot->dq_sb));
5225 if (IS_ERR(handle))
5226 return PTR_ERR(handle);
5227 ret = dquot_acquire(dquot);
5228 err = ext4_journal_stop(handle);
5229 if (!ret)
5230 ret = err;
5231 return ret;
5234 static int ext4_release_dquot(struct dquot *dquot)
5236 int ret, err;
5237 handle_t *handle;
5239 handle = ext4_journal_start(dquot_to_inode(dquot), EXT4_HT_QUOTA,
5240 EXT4_QUOTA_DEL_BLOCKS(dquot->dq_sb));
5241 if (IS_ERR(handle)) {
5242 /* Release dquot anyway to avoid endless cycle in dqput() */
5243 dquot_release(dquot);
5244 return PTR_ERR(handle);
5246 ret = dquot_release(dquot);
5247 err = ext4_journal_stop(handle);
5248 if (!ret)
5249 ret = err;
5250 return ret;
5253 static int ext4_mark_dquot_dirty(struct dquot *dquot)
5255 struct super_block *sb = dquot->dq_sb;
5256 struct ext4_sb_info *sbi = EXT4_SB(sb);
5258 /* Are we journaling quotas? */
5259 if (ext4_has_feature_quota(sb) ||
5260 sbi->s_qf_names[USRQUOTA] || sbi->s_qf_names[GRPQUOTA]) {
5261 dquot_mark_dquot_dirty(dquot);
5262 return ext4_write_dquot(dquot);
5263 } else {
5264 return dquot_mark_dquot_dirty(dquot);
5268 static int ext4_write_info(struct super_block *sb, int type)
5270 int ret, err;
5271 handle_t *handle;
5273 /* Data block + inode block */
5274 handle = ext4_journal_start(d_inode(sb->s_root), EXT4_HT_QUOTA, 2);
5275 if (IS_ERR(handle))
5276 return PTR_ERR(handle);
5277 ret = dquot_commit_info(sb, type);
5278 err = ext4_journal_stop(handle);
5279 if (!ret)
5280 ret = err;
5281 return ret;
5285 * Turn on quotas during mount time - we need to find
5286 * the quota file and such...
5288 static int ext4_quota_on_mount(struct super_block *sb, int type)
5290 return dquot_quota_on_mount(sb, EXT4_SB(sb)->s_qf_names[type],
5291 EXT4_SB(sb)->s_jquota_fmt, type);
5294 static void lockdep_set_quota_inode(struct inode *inode, int subclass)
5296 struct ext4_inode_info *ei = EXT4_I(inode);
5298 /* The first argument of lockdep_set_subclass has to be
5299 * *exactly* the same as the argument to init_rwsem() --- in
5300 * this case, in init_once() --- or lockdep gets unhappy
5301 * because the name of the lock is set using the
5302 * stringification of the argument to init_rwsem().
5304 (void) ei; /* shut up clang warning if !CONFIG_LOCKDEP */
5305 lockdep_set_subclass(&ei->i_data_sem, subclass);
5309 * Standard function to be called on quota_on
5311 static int ext4_quota_on(struct super_block *sb, int type, int format_id,
5312 const struct path *path)
5314 int err;
5316 if (!test_opt(sb, QUOTA))
5317 return -EINVAL;
5319 /* Quotafile not on the same filesystem? */
5320 if (path->dentry->d_sb != sb)
5321 return -EXDEV;
5322 /* Journaling quota? */
5323 if (EXT4_SB(sb)->s_qf_names[type]) {
5324 /* Quotafile not in fs root? */
5325 if (path->dentry->d_parent != sb->s_root)
5326 ext4_msg(sb, KERN_WARNING,
5327 "Quota file not on filesystem root. "
5328 "Journaled quota will not work");
5332 * When we journal data on quota file, we have to flush journal to see
5333 * all updates to the file when we bypass pagecache...
5335 if (EXT4_SB(sb)->s_journal &&
5336 ext4_should_journal_data(d_inode(path->dentry))) {
5338 * We don't need to lock updates but journal_flush() could
5339 * otherwise be livelocked...
5341 jbd2_journal_lock_updates(EXT4_SB(sb)->s_journal);
5342 err = jbd2_journal_flush(EXT4_SB(sb)->s_journal);
5343 jbd2_journal_unlock_updates(EXT4_SB(sb)->s_journal);
5344 if (err)
5345 return err;
5347 lockdep_set_quota_inode(path->dentry->d_inode, I_DATA_SEM_QUOTA);
5348 err = dquot_quota_on(sb, type, format_id, path);
5349 if (err)
5350 lockdep_set_quota_inode(path->dentry->d_inode,
5351 I_DATA_SEM_NORMAL);
5352 return err;
5355 static int ext4_quota_enable(struct super_block *sb, int type, int format_id,
5356 unsigned int flags)
5358 int err;
5359 struct inode *qf_inode;
5360 unsigned long qf_inums[EXT4_MAXQUOTAS] = {
5361 le32_to_cpu(EXT4_SB(sb)->s_es->s_usr_quota_inum),
5362 le32_to_cpu(EXT4_SB(sb)->s_es->s_grp_quota_inum),
5363 le32_to_cpu(EXT4_SB(sb)->s_es->s_prj_quota_inum)
5366 BUG_ON(!ext4_has_feature_quota(sb));
5368 if (!qf_inums[type])
5369 return -EPERM;
5371 qf_inode = ext4_iget(sb, qf_inums[type]);
5372 if (IS_ERR(qf_inode)) {
5373 ext4_error(sb, "Bad quota inode # %lu", qf_inums[type]);
5374 return PTR_ERR(qf_inode);
5377 /* Don't account quota for quota files to avoid recursion */
5378 qf_inode->i_flags |= S_NOQUOTA;
5379 lockdep_set_quota_inode(qf_inode, I_DATA_SEM_QUOTA);
5380 err = dquot_enable(qf_inode, type, format_id, flags);
5381 iput(qf_inode);
5382 if (err)
5383 lockdep_set_quota_inode(qf_inode, I_DATA_SEM_NORMAL);
5385 return err;
5388 /* Enable usage tracking for all quota types. */
5389 static int ext4_enable_quotas(struct super_block *sb)
5391 int type, err = 0;
5392 unsigned long qf_inums[EXT4_MAXQUOTAS] = {
5393 le32_to_cpu(EXT4_SB(sb)->s_es->s_usr_quota_inum),
5394 le32_to_cpu(EXT4_SB(sb)->s_es->s_grp_quota_inum),
5395 le32_to_cpu(EXT4_SB(sb)->s_es->s_prj_quota_inum)
5397 bool quota_mopt[EXT4_MAXQUOTAS] = {
5398 test_opt(sb, USRQUOTA),
5399 test_opt(sb, GRPQUOTA),
5400 test_opt(sb, PRJQUOTA),
5403 sb_dqopt(sb)->flags |= DQUOT_QUOTA_SYS_FILE;
5404 for (type = 0; type < EXT4_MAXQUOTAS; type++) {
5405 if (qf_inums[type]) {
5406 err = ext4_quota_enable(sb, type, QFMT_VFS_V1,
5407 DQUOT_USAGE_ENABLED |
5408 (quota_mopt[type] ? DQUOT_LIMITS_ENABLED : 0));
5409 if (err) {
5410 ext4_warning(sb,
5411 "Failed to enable quota tracking "
5412 "(type=%d, err=%d). Please run "
5413 "e2fsck to fix.", type, err);
5414 return err;
5418 return 0;
5421 static int ext4_quota_off(struct super_block *sb, int type)
5423 struct inode *inode = sb_dqopt(sb)->files[type];
5424 handle_t *handle;
5426 /* Force all delayed allocation blocks to be allocated.
5427 * Caller already holds s_umount sem */
5428 if (test_opt(sb, DELALLOC))
5429 sync_filesystem(sb);
5431 if (!inode)
5432 goto out;
5434 /* Update modification times of quota files when userspace can
5435 * start looking at them */
5436 handle = ext4_journal_start(inode, EXT4_HT_QUOTA, 1);
5437 if (IS_ERR(handle))
5438 goto out;
5439 inode->i_mtime = inode->i_ctime = current_time(inode);
5440 ext4_mark_inode_dirty(handle, inode);
5441 ext4_journal_stop(handle);
5443 out:
5444 return dquot_quota_off(sb, type);
5447 /* Read data from quotafile - avoid pagecache and such because we cannot afford
5448 * acquiring the locks... As quota files are never truncated and quota code
5449 * itself serializes the operations (and no one else should touch the files)
5450 * we don't have to be afraid of races */
5451 static ssize_t ext4_quota_read(struct super_block *sb, int type, char *data,
5452 size_t len, loff_t off)
5454 struct inode *inode = sb_dqopt(sb)->files[type];
5455 ext4_lblk_t blk = off >> EXT4_BLOCK_SIZE_BITS(sb);
5456 int offset = off & (sb->s_blocksize - 1);
5457 int tocopy;
5458 size_t toread;
5459 struct buffer_head *bh;
5460 loff_t i_size = i_size_read(inode);
5462 if (off > i_size)
5463 return 0;
5464 if (off+len > i_size)
5465 len = i_size-off;
5466 toread = len;
5467 while (toread > 0) {
5468 tocopy = sb->s_blocksize - offset < toread ?
5469 sb->s_blocksize - offset : toread;
5470 bh = ext4_bread(NULL, inode, blk, 0);
5471 if (IS_ERR(bh))
5472 return PTR_ERR(bh);
5473 if (!bh) /* A hole? */
5474 memset(data, 0, tocopy);
5475 else
5476 memcpy(data, bh->b_data+offset, tocopy);
5477 brelse(bh);
5478 offset = 0;
5479 toread -= tocopy;
5480 data += tocopy;
5481 blk++;
5483 return len;
5486 /* Write to quotafile (we know the transaction is already started and has
5487 * enough credits) */
5488 static ssize_t ext4_quota_write(struct super_block *sb, int type,
5489 const char *data, size_t len, loff_t off)
5491 struct inode *inode = sb_dqopt(sb)->files[type];
5492 ext4_lblk_t blk = off >> EXT4_BLOCK_SIZE_BITS(sb);
5493 int err, offset = off & (sb->s_blocksize - 1);
5494 int retries = 0;
5495 struct buffer_head *bh;
5496 handle_t *handle = journal_current_handle();
5498 if (EXT4_SB(sb)->s_journal && !handle) {
5499 ext4_msg(sb, KERN_WARNING, "Quota write (off=%llu, len=%llu)"
5500 " cancelled because transaction is not started",
5501 (unsigned long long)off, (unsigned long long)len);
5502 return -EIO;
5505 * Since we account only one data block in transaction credits,
5506 * then it is impossible to cross a block boundary.
5508 if (sb->s_blocksize - offset < len) {
5509 ext4_msg(sb, KERN_WARNING, "Quota write (off=%llu, len=%llu)"
5510 " cancelled because not block aligned",
5511 (unsigned long long)off, (unsigned long long)len);
5512 return -EIO;
5515 do {
5516 bh = ext4_bread(handle, inode, blk,
5517 EXT4_GET_BLOCKS_CREATE |
5518 EXT4_GET_BLOCKS_METADATA_NOFAIL);
5519 } while (IS_ERR(bh) && (PTR_ERR(bh) == -ENOSPC) &&
5520 ext4_should_retry_alloc(inode->i_sb, &retries));
5521 if (IS_ERR(bh))
5522 return PTR_ERR(bh);
5523 if (!bh)
5524 goto out;
5525 BUFFER_TRACE(bh, "get write access");
5526 err = ext4_journal_get_write_access(handle, bh);
5527 if (err) {
5528 brelse(bh);
5529 return err;
5531 lock_buffer(bh);
5532 memcpy(bh->b_data+offset, data, len);
5533 flush_dcache_page(bh->b_page);
5534 unlock_buffer(bh);
5535 err = ext4_handle_dirty_metadata(handle, NULL, bh);
5536 brelse(bh);
5537 out:
5538 if (inode->i_size < off + len) {
5539 i_size_write(inode, off + len);
5540 EXT4_I(inode)->i_disksize = inode->i_size;
5541 ext4_mark_inode_dirty(handle, inode);
5543 return len;
5546 static int ext4_get_next_id(struct super_block *sb, struct kqid *qid)
5548 const struct quota_format_ops *ops;
5550 if (!sb_has_quota_loaded(sb, qid->type))
5551 return -ESRCH;
5552 ops = sb_dqopt(sb)->ops[qid->type];
5553 if (!ops || !ops->get_next_id)
5554 return -ENOSYS;
5555 return dquot_get_next_id(sb, qid);
5557 #endif
5559 static struct dentry *ext4_mount(struct file_system_type *fs_type, int flags,
5560 const char *dev_name, void *data)
5562 return mount_bdev(fs_type, flags, dev_name, data, ext4_fill_super);
5565 #if !defined(CONFIG_EXT2_FS) && !defined(CONFIG_EXT2_FS_MODULE) && defined(CONFIG_EXT4_USE_FOR_EXT2)
5566 static inline void register_as_ext2(void)
5568 int err = register_filesystem(&ext2_fs_type);
5569 if (err)
5570 printk(KERN_WARNING
5571 "EXT4-fs: Unable to register as ext2 (%d)\n", err);
5574 static inline void unregister_as_ext2(void)
5576 unregister_filesystem(&ext2_fs_type);
5579 static inline int ext2_feature_set_ok(struct super_block *sb)
5581 if (ext4_has_unknown_ext2_incompat_features(sb))
5582 return 0;
5583 if (sb->s_flags & MS_RDONLY)
5584 return 1;
5585 if (ext4_has_unknown_ext2_ro_compat_features(sb))
5586 return 0;
5587 return 1;
5589 #else
5590 static inline void register_as_ext2(void) { }
5591 static inline void unregister_as_ext2(void) { }
5592 static inline int ext2_feature_set_ok(struct super_block *sb) { return 0; }
5593 #endif
5595 static inline void register_as_ext3(void)
5597 int err = register_filesystem(&ext3_fs_type);
5598 if (err)
5599 printk(KERN_WARNING
5600 "EXT4-fs: Unable to register as ext3 (%d)\n", err);
5603 static inline void unregister_as_ext3(void)
5605 unregister_filesystem(&ext3_fs_type);
5608 static inline int ext3_feature_set_ok(struct super_block *sb)
5610 if (ext4_has_unknown_ext3_incompat_features(sb))
5611 return 0;
5612 if (!ext4_has_feature_journal(sb))
5613 return 0;
5614 if (sb->s_flags & MS_RDONLY)
5615 return 1;
5616 if (ext4_has_unknown_ext3_ro_compat_features(sb))
5617 return 0;
5618 return 1;
5621 static struct file_system_type ext4_fs_type = {
5622 .owner = THIS_MODULE,
5623 .name = "ext4",
5624 .mount = ext4_mount,
5625 .kill_sb = kill_block_super,
5626 .fs_flags = FS_REQUIRES_DEV,
5628 MODULE_ALIAS_FS("ext4");
5630 /* Shared across all ext4 file systems */
5631 wait_queue_head_t ext4__ioend_wq[EXT4_WQ_HASH_SZ];
5633 static int __init ext4_init_fs(void)
5635 int i, err;
5637 ratelimit_state_init(&ext4_mount_msg_ratelimit, 30 * HZ, 64);
5638 ext4_li_info = NULL;
5639 mutex_init(&ext4_li_mtx);
5641 /* Build-time check for flags consistency */
5642 ext4_check_flag_values();
5644 for (i = 0; i < EXT4_WQ_HASH_SZ; i++)
5645 init_waitqueue_head(&ext4__ioend_wq[i]);
5647 err = ext4_init_es();
5648 if (err)
5649 return err;
5651 err = ext4_init_pageio();
5652 if (err)
5653 goto out5;
5655 err = ext4_init_system_zone();
5656 if (err)
5657 goto out4;
5659 err = ext4_init_sysfs();
5660 if (err)
5661 goto out3;
5663 err = ext4_init_mballoc();
5664 if (err)
5665 goto out2;
5666 err = init_inodecache();
5667 if (err)
5668 goto out1;
5669 register_as_ext3();
5670 register_as_ext2();
5671 err = register_filesystem(&ext4_fs_type);
5672 if (err)
5673 goto out;
5675 return 0;
5676 out:
5677 unregister_as_ext2();
5678 unregister_as_ext3();
5679 destroy_inodecache();
5680 out1:
5681 ext4_exit_mballoc();
5682 out2:
5683 ext4_exit_sysfs();
5684 out3:
5685 ext4_exit_system_zone();
5686 out4:
5687 ext4_exit_pageio();
5688 out5:
5689 ext4_exit_es();
5691 return err;
5694 static void __exit ext4_exit_fs(void)
5696 ext4_destroy_lazyinit_thread();
5697 unregister_as_ext2();
5698 unregister_as_ext3();
5699 unregister_filesystem(&ext4_fs_type);
5700 destroy_inodecache();
5701 ext4_exit_mballoc();
5702 ext4_exit_sysfs();
5703 ext4_exit_system_zone();
5704 ext4_exit_pageio();
5705 ext4_exit_es();
5708 MODULE_AUTHOR("Remy Card, Stephen Tweedie, Andrew Morton, Andreas Dilger, Theodore Ts'o and others");
5709 MODULE_DESCRIPTION("Fourth Extended Filesystem");
5710 MODULE_LICENSE("GPL");
5711 module_init(ext4_init_fs)
5712 module_exit(ext4_exit_fs)