ext4: Get rid of EXTEND_DISKSIZE flag of ext4_get_blocks_handle()
[linux-2.6/linux-acpi-2.6/ibm-acpi-2.6.git] / fs / ext4 / super.c
blob0a97b1ad3e19b09b5699b9c77aba18dcee12a47b
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/jbd2.h>
25 #include <linux/slab.h>
26 #include <linux/init.h>
27 #include <linux/blkdev.h>
28 #include <linux/parser.h>
29 #include <linux/smp_lock.h>
30 #include <linux/buffer_head.h>
31 #include <linux/exportfs.h>
32 #include <linux/vfs.h>
33 #include <linux/random.h>
34 #include <linux/mount.h>
35 #include <linux/namei.h>
36 #include <linux/quotaops.h>
37 #include <linux/seq_file.h>
38 #include <linux/proc_fs.h>
39 #include <linux/ctype.h>
40 #include <linux/marker.h>
41 #include <linux/log2.h>
42 #include <linux/crc16.h>
43 #include <asm/uaccess.h>
45 #include "ext4.h"
46 #include "ext4_jbd2.h"
47 #include "xattr.h"
48 #include "acl.h"
50 static int default_mb_history_length = 1000;
52 module_param_named(default_mb_history_length, default_mb_history_length,
53 int, 0644);
54 MODULE_PARM_DESC(default_mb_history_length,
55 "Default number of entries saved for mb_history");
57 struct proc_dir_entry *ext4_proc_root;
58 static struct kset *ext4_kset;
60 static int ext4_load_journal(struct super_block *, struct ext4_super_block *,
61 unsigned long journal_devnum);
62 static int ext4_commit_super(struct super_block *sb, int sync);
63 static void ext4_mark_recovery_complete(struct super_block *sb,
64 struct ext4_super_block *es);
65 static void ext4_clear_journal_err(struct super_block *sb,
66 struct ext4_super_block *es);
67 static int ext4_sync_fs(struct super_block *sb, int wait);
68 static const char *ext4_decode_error(struct super_block *sb, int errno,
69 char nbuf[16]);
70 static int ext4_remount(struct super_block *sb, int *flags, char *data);
71 static int ext4_statfs(struct dentry *dentry, struct kstatfs *buf);
72 static int ext4_unfreeze(struct super_block *sb);
73 static void ext4_write_super(struct super_block *sb);
74 static int ext4_freeze(struct super_block *sb);
77 ext4_fsblk_t ext4_block_bitmap(struct super_block *sb,
78 struct ext4_group_desc *bg)
80 return le32_to_cpu(bg->bg_block_bitmap_lo) |
81 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
82 (ext4_fsblk_t)le32_to_cpu(bg->bg_block_bitmap_hi) << 32 : 0);
85 ext4_fsblk_t ext4_inode_bitmap(struct super_block *sb,
86 struct ext4_group_desc *bg)
88 return le32_to_cpu(bg->bg_inode_bitmap_lo) |
89 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
90 (ext4_fsblk_t)le32_to_cpu(bg->bg_inode_bitmap_hi) << 32 : 0);
93 ext4_fsblk_t ext4_inode_table(struct super_block *sb,
94 struct ext4_group_desc *bg)
96 return le32_to_cpu(bg->bg_inode_table_lo) |
97 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
98 (ext4_fsblk_t)le32_to_cpu(bg->bg_inode_table_hi) << 32 : 0);
101 __u32 ext4_free_blks_count(struct super_block *sb,
102 struct ext4_group_desc *bg)
104 return le16_to_cpu(bg->bg_free_blocks_count_lo) |
105 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
106 (__u32)le16_to_cpu(bg->bg_free_blocks_count_hi) << 16 : 0);
109 __u32 ext4_free_inodes_count(struct super_block *sb,
110 struct ext4_group_desc *bg)
112 return le16_to_cpu(bg->bg_free_inodes_count_lo) |
113 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
114 (__u32)le16_to_cpu(bg->bg_free_inodes_count_hi) << 16 : 0);
117 __u32 ext4_used_dirs_count(struct super_block *sb,
118 struct ext4_group_desc *bg)
120 return le16_to_cpu(bg->bg_used_dirs_count_lo) |
121 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
122 (__u32)le16_to_cpu(bg->bg_used_dirs_count_hi) << 16 : 0);
125 __u32 ext4_itable_unused_count(struct super_block *sb,
126 struct ext4_group_desc *bg)
128 return le16_to_cpu(bg->bg_itable_unused_lo) |
129 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
130 (__u32)le16_to_cpu(bg->bg_itable_unused_hi) << 16 : 0);
133 void ext4_block_bitmap_set(struct super_block *sb,
134 struct ext4_group_desc *bg, ext4_fsblk_t blk)
136 bg->bg_block_bitmap_lo = cpu_to_le32((u32)blk);
137 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
138 bg->bg_block_bitmap_hi = cpu_to_le32(blk >> 32);
141 void ext4_inode_bitmap_set(struct super_block *sb,
142 struct ext4_group_desc *bg, ext4_fsblk_t blk)
144 bg->bg_inode_bitmap_lo = cpu_to_le32((u32)blk);
145 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
146 bg->bg_inode_bitmap_hi = cpu_to_le32(blk >> 32);
149 void ext4_inode_table_set(struct super_block *sb,
150 struct ext4_group_desc *bg, ext4_fsblk_t blk)
152 bg->bg_inode_table_lo = cpu_to_le32((u32)blk);
153 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
154 bg->bg_inode_table_hi = cpu_to_le32(blk >> 32);
157 void ext4_free_blks_set(struct super_block *sb,
158 struct ext4_group_desc *bg, __u32 count)
160 bg->bg_free_blocks_count_lo = cpu_to_le16((__u16)count);
161 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
162 bg->bg_free_blocks_count_hi = cpu_to_le16(count >> 16);
165 void ext4_free_inodes_set(struct super_block *sb,
166 struct ext4_group_desc *bg, __u32 count)
168 bg->bg_free_inodes_count_lo = cpu_to_le16((__u16)count);
169 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
170 bg->bg_free_inodes_count_hi = cpu_to_le16(count >> 16);
173 void ext4_used_dirs_set(struct super_block *sb,
174 struct ext4_group_desc *bg, __u32 count)
176 bg->bg_used_dirs_count_lo = cpu_to_le16((__u16)count);
177 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
178 bg->bg_used_dirs_count_hi = cpu_to_le16(count >> 16);
181 void ext4_itable_unused_set(struct super_block *sb,
182 struct ext4_group_desc *bg, __u32 count)
184 bg->bg_itable_unused_lo = cpu_to_le16((__u16)count);
185 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
186 bg->bg_itable_unused_hi = cpu_to_le16(count >> 16);
190 * Wrappers for jbd2_journal_start/end.
192 * The only special thing we need to do here is to make sure that all
193 * journal_end calls result in the superblock being marked dirty, so
194 * that sync() will call the filesystem's write_super callback if
195 * appropriate.
197 handle_t *ext4_journal_start_sb(struct super_block *sb, int nblocks)
199 journal_t *journal;
201 if (sb->s_flags & MS_RDONLY)
202 return ERR_PTR(-EROFS);
204 /* Special case here: if the journal has aborted behind our
205 * backs (eg. EIO in the commit thread), then we still need to
206 * take the FS itself readonly cleanly. */
207 journal = EXT4_SB(sb)->s_journal;
208 if (journal) {
209 if (is_journal_aborted(journal)) {
210 ext4_abort(sb, __func__, "Detected aborted journal");
211 return ERR_PTR(-EROFS);
213 return jbd2_journal_start(journal, nblocks);
216 * We're not journaling, return the appropriate indication.
218 current->journal_info = EXT4_NOJOURNAL_HANDLE;
219 return current->journal_info;
223 * The only special thing we need to do here is to make sure that all
224 * jbd2_journal_stop calls result in the superblock being marked dirty, so
225 * that sync() will call the filesystem's write_super callback if
226 * appropriate.
228 int __ext4_journal_stop(const char *where, handle_t *handle)
230 struct super_block *sb;
231 int err;
232 int rc;
234 if (!ext4_handle_valid(handle)) {
236 * Do this here since we don't call jbd2_journal_stop() in
237 * no-journal mode.
239 current->journal_info = NULL;
240 return 0;
242 sb = handle->h_transaction->t_journal->j_private;
243 err = handle->h_err;
244 rc = jbd2_journal_stop(handle);
246 if (!err)
247 err = rc;
248 if (err)
249 __ext4_std_error(sb, where, err);
250 return err;
253 void ext4_journal_abort_handle(const char *caller, const char *err_fn,
254 struct buffer_head *bh, handle_t *handle, int err)
256 char nbuf[16];
257 const char *errstr = ext4_decode_error(NULL, err, nbuf);
259 BUG_ON(!ext4_handle_valid(handle));
261 if (bh)
262 BUFFER_TRACE(bh, "abort");
264 if (!handle->h_err)
265 handle->h_err = err;
267 if (is_handle_aborted(handle))
268 return;
270 printk(KERN_ERR "%s: aborting transaction: %s in %s\n",
271 caller, errstr, err_fn);
273 jbd2_journal_abort_handle(handle);
276 /* Deal with the reporting of failure conditions on a filesystem such as
277 * inconsistencies detected or read IO failures.
279 * On ext2, we can store the error state of the filesystem in the
280 * superblock. That is not possible on ext4, because we may have other
281 * write ordering constraints on the superblock which prevent us from
282 * writing it out straight away; and given that the journal is about to
283 * be aborted, we can't rely on the current, or future, transactions to
284 * write out the superblock safely.
286 * We'll just use the jbd2_journal_abort() error code to record an error in
287 * the journal instead. On recovery, the journal will compain about
288 * that error until we've noted it down and cleared it.
291 static void ext4_handle_error(struct super_block *sb)
293 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
295 EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS;
296 es->s_state |= cpu_to_le16(EXT4_ERROR_FS);
298 if (sb->s_flags & MS_RDONLY)
299 return;
301 if (!test_opt(sb, ERRORS_CONT)) {
302 journal_t *journal = EXT4_SB(sb)->s_journal;
304 EXT4_SB(sb)->s_mount_opt |= EXT4_MOUNT_ABORT;
305 if (journal)
306 jbd2_journal_abort(journal, -EIO);
308 if (test_opt(sb, ERRORS_RO)) {
309 printk(KERN_CRIT "Remounting filesystem read-only\n");
310 sb->s_flags |= MS_RDONLY;
312 ext4_commit_super(sb, 1);
313 if (test_opt(sb, ERRORS_PANIC))
314 panic("EXT4-fs (device %s): panic forced after error\n",
315 sb->s_id);
318 void ext4_error(struct super_block *sb, const char *function,
319 const char *fmt, ...)
321 va_list args;
323 va_start(args, fmt);
324 printk(KERN_CRIT "EXT4-fs error (device %s): %s: ", sb->s_id, function);
325 vprintk(fmt, args);
326 printk("\n");
327 va_end(args);
329 ext4_handle_error(sb);
332 static const char *ext4_decode_error(struct super_block *sb, int errno,
333 char nbuf[16])
335 char *errstr = NULL;
337 switch (errno) {
338 case -EIO:
339 errstr = "IO failure";
340 break;
341 case -ENOMEM:
342 errstr = "Out of memory";
343 break;
344 case -EROFS:
345 if (!sb || EXT4_SB(sb)->s_journal->j_flags & JBD2_ABORT)
346 errstr = "Journal has aborted";
347 else
348 errstr = "Readonly filesystem";
349 break;
350 default:
351 /* If the caller passed in an extra buffer for unknown
352 * errors, textualise them now. Else we just return
353 * NULL. */
354 if (nbuf) {
355 /* Check for truncated error codes... */
356 if (snprintf(nbuf, 16, "error %d", -errno) >= 0)
357 errstr = nbuf;
359 break;
362 return errstr;
365 /* __ext4_std_error decodes expected errors from journaling functions
366 * automatically and invokes the appropriate error response. */
368 void __ext4_std_error(struct super_block *sb, const char *function, int errno)
370 char nbuf[16];
371 const char *errstr;
373 /* Special case: if the error is EROFS, and we're not already
374 * inside a transaction, then there's really no point in logging
375 * an error. */
376 if (errno == -EROFS && journal_current_handle() == NULL &&
377 (sb->s_flags & MS_RDONLY))
378 return;
380 errstr = ext4_decode_error(sb, errno, nbuf);
381 printk(KERN_CRIT "EXT4-fs error (device %s) in %s: %s\n",
382 sb->s_id, function, errstr);
384 ext4_handle_error(sb);
388 * ext4_abort is a much stronger failure handler than ext4_error. The
389 * abort function may be used to deal with unrecoverable failures such
390 * as journal IO errors or ENOMEM at a critical moment in log management.
392 * We unconditionally force the filesystem into an ABORT|READONLY state,
393 * unless the error response on the fs has been set to panic in which
394 * case we take the easy way out and panic immediately.
397 void ext4_abort(struct super_block *sb, const char *function,
398 const char *fmt, ...)
400 va_list args;
402 printk(KERN_CRIT "ext4_abort called.\n");
404 va_start(args, fmt);
405 printk(KERN_CRIT "EXT4-fs error (device %s): %s: ", sb->s_id, function);
406 vprintk(fmt, args);
407 printk("\n");
408 va_end(args);
410 if (test_opt(sb, ERRORS_PANIC))
411 panic("EXT4-fs panic from previous error\n");
413 if (sb->s_flags & MS_RDONLY)
414 return;
416 printk(KERN_CRIT "Remounting filesystem read-only\n");
417 EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS;
418 sb->s_flags |= MS_RDONLY;
419 EXT4_SB(sb)->s_mount_opt |= EXT4_MOUNT_ABORT;
420 if (EXT4_SB(sb)->s_journal)
421 jbd2_journal_abort(EXT4_SB(sb)->s_journal, -EIO);
424 void ext4_warning(struct super_block *sb, const char *function,
425 const char *fmt, ...)
427 va_list args;
429 va_start(args, fmt);
430 printk(KERN_WARNING "EXT4-fs warning (device %s): %s: ",
431 sb->s_id, function);
432 vprintk(fmt, args);
433 printk("\n");
434 va_end(args);
437 void ext4_grp_locked_error(struct super_block *sb, ext4_group_t grp,
438 const char *function, const char *fmt, ...)
439 __releases(bitlock)
440 __acquires(bitlock)
442 va_list args;
443 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
445 va_start(args, fmt);
446 printk(KERN_CRIT "EXT4-fs error (device %s): %s: ", sb->s_id, function);
447 vprintk(fmt, args);
448 printk("\n");
449 va_end(args);
451 if (test_opt(sb, ERRORS_CONT)) {
452 EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS;
453 es->s_state |= cpu_to_le16(EXT4_ERROR_FS);
454 ext4_commit_super(sb, 0);
455 return;
457 ext4_unlock_group(sb, grp);
458 ext4_handle_error(sb);
460 * We only get here in the ERRORS_RO case; relocking the group
461 * may be dangerous, but nothing bad will happen since the
462 * filesystem will have already been marked read/only and the
463 * journal has been aborted. We return 1 as a hint to callers
464 * who might what to use the return value from
465 * ext4_grp_locked_error() to distinguish beween the
466 * ERRORS_CONT and ERRORS_RO case, and perhaps return more
467 * aggressively from the ext4 function in question, with a
468 * more appropriate error code.
470 ext4_lock_group(sb, grp);
471 return;
474 void ext4_update_dynamic_rev(struct super_block *sb)
476 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
478 if (le32_to_cpu(es->s_rev_level) > EXT4_GOOD_OLD_REV)
479 return;
481 ext4_warning(sb, __func__,
482 "updating to rev %d because of new feature flag, "
483 "running e2fsck is recommended",
484 EXT4_DYNAMIC_REV);
486 es->s_first_ino = cpu_to_le32(EXT4_GOOD_OLD_FIRST_INO);
487 es->s_inode_size = cpu_to_le16(EXT4_GOOD_OLD_INODE_SIZE);
488 es->s_rev_level = cpu_to_le32(EXT4_DYNAMIC_REV);
489 /* leave es->s_feature_*compat flags alone */
490 /* es->s_uuid will be set by e2fsck if empty */
493 * The rest of the superblock fields should be zero, and if not it
494 * means they are likely already in use, so leave them alone. We
495 * can leave it up to e2fsck to clean up any inconsistencies there.
500 * Open the external journal device
502 static struct block_device *ext4_blkdev_get(dev_t dev)
504 struct block_device *bdev;
505 char b[BDEVNAME_SIZE];
507 bdev = open_by_devnum(dev, FMODE_READ|FMODE_WRITE);
508 if (IS_ERR(bdev))
509 goto fail;
510 return bdev;
512 fail:
513 printk(KERN_ERR "EXT4-fs: failed to open journal device %s: %ld\n",
514 __bdevname(dev, b), PTR_ERR(bdev));
515 return NULL;
519 * Release the journal device
521 static int ext4_blkdev_put(struct block_device *bdev)
523 bd_release(bdev);
524 return blkdev_put(bdev, FMODE_READ|FMODE_WRITE);
527 static int ext4_blkdev_remove(struct ext4_sb_info *sbi)
529 struct block_device *bdev;
530 int ret = -ENODEV;
532 bdev = sbi->journal_bdev;
533 if (bdev) {
534 ret = ext4_blkdev_put(bdev);
535 sbi->journal_bdev = NULL;
537 return ret;
540 static inline struct inode *orphan_list_entry(struct list_head *l)
542 return &list_entry(l, struct ext4_inode_info, i_orphan)->vfs_inode;
545 static void dump_orphan_list(struct super_block *sb, struct ext4_sb_info *sbi)
547 struct list_head *l;
549 printk(KERN_ERR "sb orphan head is %d\n",
550 le32_to_cpu(sbi->s_es->s_last_orphan));
552 printk(KERN_ERR "sb_info orphan list:\n");
553 list_for_each(l, &sbi->s_orphan) {
554 struct inode *inode = orphan_list_entry(l);
555 printk(KERN_ERR " "
556 "inode %s:%lu at %p: mode %o, nlink %d, next %d\n",
557 inode->i_sb->s_id, inode->i_ino, inode,
558 inode->i_mode, inode->i_nlink,
559 NEXT_ORPHAN(inode));
563 static void ext4_put_super(struct super_block *sb)
565 struct ext4_sb_info *sbi = EXT4_SB(sb);
566 struct ext4_super_block *es = sbi->s_es;
567 int i, err;
569 ext4_release_system_zone(sb);
570 ext4_mb_release(sb);
571 ext4_ext_release(sb);
572 ext4_xattr_put_super(sb);
573 if (sbi->s_journal) {
574 err = jbd2_journal_destroy(sbi->s_journal);
575 sbi->s_journal = NULL;
576 if (err < 0)
577 ext4_abort(sb, __func__,
578 "Couldn't clean up the journal");
580 if (!(sb->s_flags & MS_RDONLY)) {
581 EXT4_CLEAR_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
582 es->s_state = cpu_to_le16(sbi->s_mount_state);
583 ext4_commit_super(sb, 1);
585 if (sbi->s_proc) {
586 remove_proc_entry(sb->s_id, ext4_proc_root);
588 kobject_del(&sbi->s_kobj);
590 for (i = 0; i < sbi->s_gdb_count; i++)
591 brelse(sbi->s_group_desc[i]);
592 kfree(sbi->s_group_desc);
593 if (is_vmalloc_addr(sbi->s_flex_groups))
594 vfree(sbi->s_flex_groups);
595 else
596 kfree(sbi->s_flex_groups);
597 percpu_counter_destroy(&sbi->s_freeblocks_counter);
598 percpu_counter_destroy(&sbi->s_freeinodes_counter);
599 percpu_counter_destroy(&sbi->s_dirs_counter);
600 percpu_counter_destroy(&sbi->s_dirtyblocks_counter);
601 brelse(sbi->s_sbh);
602 #ifdef CONFIG_QUOTA
603 for (i = 0; i < MAXQUOTAS; i++)
604 kfree(sbi->s_qf_names[i]);
605 #endif
607 /* Debugging code just in case the in-memory inode orphan list
608 * isn't empty. The on-disk one can be non-empty if we've
609 * detected an error and taken the fs readonly, but the
610 * in-memory list had better be clean by this point. */
611 if (!list_empty(&sbi->s_orphan))
612 dump_orphan_list(sb, sbi);
613 J_ASSERT(list_empty(&sbi->s_orphan));
615 invalidate_bdev(sb->s_bdev);
616 if (sbi->journal_bdev && sbi->journal_bdev != sb->s_bdev) {
618 * Invalidate the journal device's buffers. We don't want them
619 * floating about in memory - the physical journal device may
620 * hotswapped, and it breaks the `ro-after' testing code.
622 sync_blockdev(sbi->journal_bdev);
623 invalidate_bdev(sbi->journal_bdev);
624 ext4_blkdev_remove(sbi);
626 sb->s_fs_info = NULL;
628 * Now that we are completely done shutting down the
629 * superblock, we need to actually destroy the kobject.
631 unlock_kernel();
632 unlock_super(sb);
633 kobject_put(&sbi->s_kobj);
634 wait_for_completion(&sbi->s_kobj_unregister);
635 lock_super(sb);
636 lock_kernel();
637 kfree(sbi->s_blockgroup_lock);
638 kfree(sbi);
641 static struct kmem_cache *ext4_inode_cachep;
644 * Called inside transaction, so use GFP_NOFS
646 static struct inode *ext4_alloc_inode(struct super_block *sb)
648 struct ext4_inode_info *ei;
650 ei = kmem_cache_alloc(ext4_inode_cachep, GFP_NOFS);
651 if (!ei)
652 return NULL;
654 #ifdef CONFIG_EXT4_FS_POSIX_ACL
655 ei->i_acl = EXT4_ACL_NOT_CACHED;
656 ei->i_default_acl = EXT4_ACL_NOT_CACHED;
657 #endif
658 ei->vfs_inode.i_version = 1;
659 ei->vfs_inode.i_data.writeback_index = 0;
660 memset(&ei->i_cached_extent, 0, sizeof(struct ext4_ext_cache));
661 INIT_LIST_HEAD(&ei->i_prealloc_list);
662 spin_lock_init(&ei->i_prealloc_lock);
664 * Note: We can be called before EXT4_SB(sb)->s_journal is set,
665 * therefore it can be null here. Don't check it, just initialize
666 * jinode.
668 jbd2_journal_init_jbd_inode(&ei->jinode, &ei->vfs_inode);
669 ei->i_reserved_data_blocks = 0;
670 ei->i_reserved_meta_blocks = 0;
671 ei->i_allocated_meta_blocks = 0;
672 ei->i_delalloc_reserved_flag = 0;
673 spin_lock_init(&(ei->i_block_reservation_lock));
675 return &ei->vfs_inode;
678 static void ext4_destroy_inode(struct inode *inode)
680 if (!list_empty(&(EXT4_I(inode)->i_orphan))) {
681 printk("EXT4 Inode %p: orphan list check failed!\n",
682 EXT4_I(inode));
683 print_hex_dump(KERN_INFO, "", DUMP_PREFIX_ADDRESS, 16, 4,
684 EXT4_I(inode), sizeof(struct ext4_inode_info),
685 true);
686 dump_stack();
688 kmem_cache_free(ext4_inode_cachep, EXT4_I(inode));
691 static void init_once(void *foo)
693 struct ext4_inode_info *ei = (struct ext4_inode_info *) foo;
695 INIT_LIST_HEAD(&ei->i_orphan);
696 #ifdef CONFIG_EXT4_FS_XATTR
697 init_rwsem(&ei->xattr_sem);
698 #endif
699 init_rwsem(&ei->i_data_sem);
700 inode_init_once(&ei->vfs_inode);
703 static int init_inodecache(void)
705 ext4_inode_cachep = kmem_cache_create("ext4_inode_cache",
706 sizeof(struct ext4_inode_info),
707 0, (SLAB_RECLAIM_ACCOUNT|
708 SLAB_MEM_SPREAD),
709 init_once);
710 if (ext4_inode_cachep == NULL)
711 return -ENOMEM;
712 return 0;
715 static void destroy_inodecache(void)
717 kmem_cache_destroy(ext4_inode_cachep);
720 static void ext4_clear_inode(struct inode *inode)
722 #ifdef CONFIG_EXT4_FS_POSIX_ACL
723 if (EXT4_I(inode)->i_acl &&
724 EXT4_I(inode)->i_acl != EXT4_ACL_NOT_CACHED) {
725 posix_acl_release(EXT4_I(inode)->i_acl);
726 EXT4_I(inode)->i_acl = EXT4_ACL_NOT_CACHED;
728 if (EXT4_I(inode)->i_default_acl &&
729 EXT4_I(inode)->i_default_acl != EXT4_ACL_NOT_CACHED) {
730 posix_acl_release(EXT4_I(inode)->i_default_acl);
731 EXT4_I(inode)->i_default_acl = EXT4_ACL_NOT_CACHED;
733 #endif
734 ext4_discard_preallocations(inode);
735 if (EXT4_JOURNAL(inode))
736 jbd2_journal_release_jbd_inode(EXT4_SB(inode->i_sb)->s_journal,
737 &EXT4_I(inode)->jinode);
740 static inline void ext4_show_quota_options(struct seq_file *seq,
741 struct super_block *sb)
743 #if defined(CONFIG_QUOTA)
744 struct ext4_sb_info *sbi = EXT4_SB(sb);
746 if (sbi->s_jquota_fmt)
747 seq_printf(seq, ",jqfmt=%s",
748 (sbi->s_jquota_fmt == QFMT_VFS_OLD) ? "vfsold" : "vfsv0");
750 if (sbi->s_qf_names[USRQUOTA])
751 seq_printf(seq, ",usrjquota=%s", sbi->s_qf_names[USRQUOTA]);
753 if (sbi->s_qf_names[GRPQUOTA])
754 seq_printf(seq, ",grpjquota=%s", sbi->s_qf_names[GRPQUOTA]);
756 if (sbi->s_mount_opt & EXT4_MOUNT_USRQUOTA)
757 seq_puts(seq, ",usrquota");
759 if (sbi->s_mount_opt & EXT4_MOUNT_GRPQUOTA)
760 seq_puts(seq, ",grpquota");
761 #endif
765 * Show an option if
766 * - it's set to a non-default value OR
767 * - if the per-sb default is different from the global default
769 static int ext4_show_options(struct seq_file *seq, struct vfsmount *vfs)
771 int def_errors;
772 unsigned long def_mount_opts;
773 struct super_block *sb = vfs->mnt_sb;
774 struct ext4_sb_info *sbi = EXT4_SB(sb);
775 struct ext4_super_block *es = sbi->s_es;
777 def_mount_opts = le32_to_cpu(es->s_default_mount_opts);
778 def_errors = le16_to_cpu(es->s_errors);
780 if (sbi->s_sb_block != 1)
781 seq_printf(seq, ",sb=%llu", sbi->s_sb_block);
782 if (test_opt(sb, MINIX_DF))
783 seq_puts(seq, ",minixdf");
784 if (test_opt(sb, GRPID) && !(def_mount_opts & EXT4_DEFM_BSDGROUPS))
785 seq_puts(seq, ",grpid");
786 if (!test_opt(sb, GRPID) && (def_mount_opts & EXT4_DEFM_BSDGROUPS))
787 seq_puts(seq, ",nogrpid");
788 if (sbi->s_resuid != EXT4_DEF_RESUID ||
789 le16_to_cpu(es->s_def_resuid) != EXT4_DEF_RESUID) {
790 seq_printf(seq, ",resuid=%u", sbi->s_resuid);
792 if (sbi->s_resgid != EXT4_DEF_RESGID ||
793 le16_to_cpu(es->s_def_resgid) != EXT4_DEF_RESGID) {
794 seq_printf(seq, ",resgid=%u", sbi->s_resgid);
796 if (test_opt(sb, ERRORS_RO)) {
797 if (def_errors == EXT4_ERRORS_PANIC ||
798 def_errors == EXT4_ERRORS_CONTINUE) {
799 seq_puts(seq, ",errors=remount-ro");
802 if (test_opt(sb, ERRORS_CONT) && def_errors != EXT4_ERRORS_CONTINUE)
803 seq_puts(seq, ",errors=continue");
804 if (test_opt(sb, ERRORS_PANIC) && def_errors != EXT4_ERRORS_PANIC)
805 seq_puts(seq, ",errors=panic");
806 if (test_opt(sb, NO_UID32) && !(def_mount_opts & EXT4_DEFM_UID16))
807 seq_puts(seq, ",nouid32");
808 if (test_opt(sb, DEBUG) && !(def_mount_opts & EXT4_DEFM_DEBUG))
809 seq_puts(seq, ",debug");
810 if (test_opt(sb, OLDALLOC))
811 seq_puts(seq, ",oldalloc");
812 #ifdef CONFIG_EXT4_FS_XATTR
813 if (test_opt(sb, XATTR_USER) &&
814 !(def_mount_opts & EXT4_DEFM_XATTR_USER))
815 seq_puts(seq, ",user_xattr");
816 if (!test_opt(sb, XATTR_USER) &&
817 (def_mount_opts & EXT4_DEFM_XATTR_USER)) {
818 seq_puts(seq, ",nouser_xattr");
820 #endif
821 #ifdef CONFIG_EXT4_FS_POSIX_ACL
822 if (test_opt(sb, POSIX_ACL) && !(def_mount_opts & EXT4_DEFM_ACL))
823 seq_puts(seq, ",acl");
824 if (!test_opt(sb, POSIX_ACL) && (def_mount_opts & EXT4_DEFM_ACL))
825 seq_puts(seq, ",noacl");
826 #endif
827 if (sbi->s_commit_interval != JBD2_DEFAULT_MAX_COMMIT_AGE*HZ) {
828 seq_printf(seq, ",commit=%u",
829 (unsigned) (sbi->s_commit_interval / HZ));
831 if (sbi->s_min_batch_time != EXT4_DEF_MIN_BATCH_TIME) {
832 seq_printf(seq, ",min_batch_time=%u",
833 (unsigned) sbi->s_min_batch_time);
835 if (sbi->s_max_batch_time != EXT4_DEF_MAX_BATCH_TIME) {
836 seq_printf(seq, ",max_batch_time=%u",
837 (unsigned) sbi->s_min_batch_time);
841 * We're changing the default of barrier mount option, so
842 * let's always display its mount state so it's clear what its
843 * status is.
845 seq_puts(seq, ",barrier=");
846 seq_puts(seq, test_opt(sb, BARRIER) ? "1" : "0");
847 if (test_opt(sb, JOURNAL_ASYNC_COMMIT))
848 seq_puts(seq, ",journal_async_commit");
849 if (test_opt(sb, NOBH))
850 seq_puts(seq, ",nobh");
851 if (test_opt(sb, I_VERSION))
852 seq_puts(seq, ",i_version");
853 if (!test_opt(sb, DELALLOC))
854 seq_puts(seq, ",nodelalloc");
857 if (sbi->s_stripe)
858 seq_printf(seq, ",stripe=%lu", sbi->s_stripe);
860 * journal mode get enabled in different ways
861 * So just print the value even if we didn't specify it
863 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA)
864 seq_puts(seq, ",data=journal");
865 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA)
866 seq_puts(seq, ",data=ordered");
867 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_WRITEBACK_DATA)
868 seq_puts(seq, ",data=writeback");
870 if (sbi->s_inode_readahead_blks != EXT4_DEF_INODE_READAHEAD_BLKS)
871 seq_printf(seq, ",inode_readahead_blks=%u",
872 sbi->s_inode_readahead_blks);
874 if (test_opt(sb, DATA_ERR_ABORT))
875 seq_puts(seq, ",data_err=abort");
877 if (test_opt(sb, NO_AUTO_DA_ALLOC))
878 seq_puts(seq, ",noauto_da_alloc");
880 ext4_show_quota_options(seq, sb);
882 return 0;
885 static struct inode *ext4_nfs_get_inode(struct super_block *sb,
886 u64 ino, u32 generation)
888 struct inode *inode;
890 if (ino < EXT4_FIRST_INO(sb) && ino != EXT4_ROOT_INO)
891 return ERR_PTR(-ESTALE);
892 if (ino > le32_to_cpu(EXT4_SB(sb)->s_es->s_inodes_count))
893 return ERR_PTR(-ESTALE);
895 /* iget isn't really right if the inode is currently unallocated!!
897 * ext4_read_inode will return a bad_inode if the inode had been
898 * deleted, so we should be safe.
900 * Currently we don't know the generation for parent directory, so
901 * a generation of 0 means "accept any"
903 inode = ext4_iget(sb, ino);
904 if (IS_ERR(inode))
905 return ERR_CAST(inode);
906 if (generation && inode->i_generation != generation) {
907 iput(inode);
908 return ERR_PTR(-ESTALE);
911 return inode;
914 static struct dentry *ext4_fh_to_dentry(struct super_block *sb, struct fid *fid,
915 int fh_len, int fh_type)
917 return generic_fh_to_dentry(sb, fid, fh_len, fh_type,
918 ext4_nfs_get_inode);
921 static struct dentry *ext4_fh_to_parent(struct super_block *sb, struct fid *fid,
922 int fh_len, int fh_type)
924 return generic_fh_to_parent(sb, fid, fh_len, fh_type,
925 ext4_nfs_get_inode);
929 * Try to release metadata pages (indirect blocks, directories) which are
930 * mapped via the block device. Since these pages could have journal heads
931 * which would prevent try_to_free_buffers() from freeing them, we must use
932 * jbd2 layer's try_to_free_buffers() function to release them.
934 static int bdev_try_to_free_page(struct super_block *sb, struct page *page,
935 gfp_t wait)
937 journal_t *journal = EXT4_SB(sb)->s_journal;
939 WARN_ON(PageChecked(page));
940 if (!page_has_buffers(page))
941 return 0;
942 if (journal)
943 return jbd2_journal_try_to_free_buffers(journal, page,
944 wait & ~__GFP_WAIT);
945 return try_to_free_buffers(page);
948 #ifdef CONFIG_QUOTA
949 #define QTYPE2NAME(t) ((t) == USRQUOTA ? "user" : "group")
950 #define QTYPE2MOPT(on, t) ((t) == USRQUOTA?((on)##USRJQUOTA):((on)##GRPJQUOTA))
952 static int ext4_write_dquot(struct dquot *dquot);
953 static int ext4_acquire_dquot(struct dquot *dquot);
954 static int ext4_release_dquot(struct dquot *dquot);
955 static int ext4_mark_dquot_dirty(struct dquot *dquot);
956 static int ext4_write_info(struct super_block *sb, int type);
957 static int ext4_quota_on(struct super_block *sb, int type, int format_id,
958 char *path, int remount);
959 static int ext4_quota_on_mount(struct super_block *sb, int type);
960 static ssize_t ext4_quota_read(struct super_block *sb, int type, char *data,
961 size_t len, loff_t off);
962 static ssize_t ext4_quota_write(struct super_block *sb, int type,
963 const char *data, size_t len, loff_t off);
965 static struct dquot_operations ext4_quota_operations = {
966 .initialize = dquot_initialize,
967 .drop = dquot_drop,
968 .alloc_space = dquot_alloc_space,
969 .reserve_space = dquot_reserve_space,
970 .claim_space = dquot_claim_space,
971 .release_rsv = dquot_release_reserved_space,
972 .get_reserved_space = ext4_get_reserved_space,
973 .alloc_inode = dquot_alloc_inode,
974 .free_space = dquot_free_space,
975 .free_inode = dquot_free_inode,
976 .transfer = dquot_transfer,
977 .write_dquot = ext4_write_dquot,
978 .acquire_dquot = ext4_acquire_dquot,
979 .release_dquot = ext4_release_dquot,
980 .mark_dirty = ext4_mark_dquot_dirty,
981 .write_info = ext4_write_info,
982 .alloc_dquot = dquot_alloc,
983 .destroy_dquot = dquot_destroy,
986 static struct quotactl_ops ext4_qctl_operations = {
987 .quota_on = ext4_quota_on,
988 .quota_off = vfs_quota_off,
989 .quota_sync = vfs_quota_sync,
990 .get_info = vfs_get_dqinfo,
991 .set_info = vfs_set_dqinfo,
992 .get_dqblk = vfs_get_dqblk,
993 .set_dqblk = vfs_set_dqblk
995 #endif
997 static const struct super_operations ext4_sops = {
998 .alloc_inode = ext4_alloc_inode,
999 .destroy_inode = ext4_destroy_inode,
1000 .write_inode = ext4_write_inode,
1001 .dirty_inode = ext4_dirty_inode,
1002 .delete_inode = ext4_delete_inode,
1003 .put_super = ext4_put_super,
1004 .sync_fs = ext4_sync_fs,
1005 .freeze_fs = ext4_freeze,
1006 .unfreeze_fs = ext4_unfreeze,
1007 .statfs = ext4_statfs,
1008 .remount_fs = ext4_remount,
1009 .clear_inode = ext4_clear_inode,
1010 .show_options = ext4_show_options,
1011 #ifdef CONFIG_QUOTA
1012 .quota_read = ext4_quota_read,
1013 .quota_write = ext4_quota_write,
1014 #endif
1015 .bdev_try_to_free_page = bdev_try_to_free_page,
1018 static const struct super_operations ext4_nojournal_sops = {
1019 .alloc_inode = ext4_alloc_inode,
1020 .destroy_inode = ext4_destroy_inode,
1021 .write_inode = ext4_write_inode,
1022 .dirty_inode = ext4_dirty_inode,
1023 .delete_inode = ext4_delete_inode,
1024 .write_super = ext4_write_super,
1025 .put_super = ext4_put_super,
1026 .statfs = ext4_statfs,
1027 .remount_fs = ext4_remount,
1028 .clear_inode = ext4_clear_inode,
1029 .show_options = ext4_show_options,
1030 #ifdef CONFIG_QUOTA
1031 .quota_read = ext4_quota_read,
1032 .quota_write = ext4_quota_write,
1033 #endif
1034 .bdev_try_to_free_page = bdev_try_to_free_page,
1037 static const struct export_operations ext4_export_ops = {
1038 .fh_to_dentry = ext4_fh_to_dentry,
1039 .fh_to_parent = ext4_fh_to_parent,
1040 .get_parent = ext4_get_parent,
1043 enum {
1044 Opt_bsd_df, Opt_minix_df, Opt_grpid, Opt_nogrpid,
1045 Opt_resgid, Opt_resuid, Opt_sb, Opt_err_cont, Opt_err_panic, Opt_err_ro,
1046 Opt_nouid32, Opt_debug, Opt_oldalloc, Opt_orlov,
1047 Opt_user_xattr, Opt_nouser_xattr, Opt_acl, Opt_noacl,
1048 Opt_auto_da_alloc, Opt_noauto_da_alloc, Opt_noload, Opt_nobh, Opt_bh,
1049 Opt_commit, Opt_min_batch_time, Opt_max_batch_time,
1050 Opt_journal_update, Opt_journal_dev,
1051 Opt_journal_checksum, Opt_journal_async_commit,
1052 Opt_abort, Opt_data_journal, Opt_data_ordered, Opt_data_writeback,
1053 Opt_data_err_abort, Opt_data_err_ignore, Opt_mb_history_length,
1054 Opt_usrjquota, Opt_grpjquota, Opt_offusrjquota, Opt_offgrpjquota,
1055 Opt_jqfmt_vfsold, Opt_jqfmt_vfsv0, Opt_quota, Opt_noquota,
1056 Opt_ignore, Opt_barrier, Opt_nobarrier, Opt_err, Opt_resize,
1057 Opt_usrquota, Opt_grpquota, Opt_i_version,
1058 Opt_stripe, Opt_delalloc, Opt_nodelalloc,
1059 Opt_block_validity, Opt_noblock_validity,
1060 Opt_inode_readahead_blks, Opt_journal_ioprio
1063 static const match_table_t tokens = {
1064 {Opt_bsd_df, "bsddf"},
1065 {Opt_minix_df, "minixdf"},
1066 {Opt_grpid, "grpid"},
1067 {Opt_grpid, "bsdgroups"},
1068 {Opt_nogrpid, "nogrpid"},
1069 {Opt_nogrpid, "sysvgroups"},
1070 {Opt_resgid, "resgid=%u"},
1071 {Opt_resuid, "resuid=%u"},
1072 {Opt_sb, "sb=%u"},
1073 {Opt_err_cont, "errors=continue"},
1074 {Opt_err_panic, "errors=panic"},
1075 {Opt_err_ro, "errors=remount-ro"},
1076 {Opt_nouid32, "nouid32"},
1077 {Opt_debug, "debug"},
1078 {Opt_oldalloc, "oldalloc"},
1079 {Opt_orlov, "orlov"},
1080 {Opt_user_xattr, "user_xattr"},
1081 {Opt_nouser_xattr, "nouser_xattr"},
1082 {Opt_acl, "acl"},
1083 {Opt_noacl, "noacl"},
1084 {Opt_noload, "noload"},
1085 {Opt_nobh, "nobh"},
1086 {Opt_bh, "bh"},
1087 {Opt_commit, "commit=%u"},
1088 {Opt_min_batch_time, "min_batch_time=%u"},
1089 {Opt_max_batch_time, "max_batch_time=%u"},
1090 {Opt_journal_update, "journal=update"},
1091 {Opt_journal_dev, "journal_dev=%u"},
1092 {Opt_journal_checksum, "journal_checksum"},
1093 {Opt_journal_async_commit, "journal_async_commit"},
1094 {Opt_abort, "abort"},
1095 {Opt_data_journal, "data=journal"},
1096 {Opt_data_ordered, "data=ordered"},
1097 {Opt_data_writeback, "data=writeback"},
1098 {Opt_data_err_abort, "data_err=abort"},
1099 {Opt_data_err_ignore, "data_err=ignore"},
1100 {Opt_mb_history_length, "mb_history_length=%u"},
1101 {Opt_offusrjquota, "usrjquota="},
1102 {Opt_usrjquota, "usrjquota=%s"},
1103 {Opt_offgrpjquota, "grpjquota="},
1104 {Opt_grpjquota, "grpjquota=%s"},
1105 {Opt_jqfmt_vfsold, "jqfmt=vfsold"},
1106 {Opt_jqfmt_vfsv0, "jqfmt=vfsv0"},
1107 {Opt_grpquota, "grpquota"},
1108 {Opt_noquota, "noquota"},
1109 {Opt_quota, "quota"},
1110 {Opt_usrquota, "usrquota"},
1111 {Opt_barrier, "barrier=%u"},
1112 {Opt_barrier, "barrier"},
1113 {Opt_nobarrier, "nobarrier"},
1114 {Opt_i_version, "i_version"},
1115 {Opt_stripe, "stripe=%u"},
1116 {Opt_resize, "resize"},
1117 {Opt_delalloc, "delalloc"},
1118 {Opt_nodelalloc, "nodelalloc"},
1119 {Opt_block_validity, "block_validity"},
1120 {Opt_noblock_validity, "noblock_validity"},
1121 {Opt_inode_readahead_blks, "inode_readahead_blks=%u"},
1122 {Opt_journal_ioprio, "journal_ioprio=%u"},
1123 {Opt_auto_da_alloc, "auto_da_alloc=%u"},
1124 {Opt_auto_da_alloc, "auto_da_alloc"},
1125 {Opt_noauto_da_alloc, "noauto_da_alloc"},
1126 {Opt_err, NULL},
1129 static ext4_fsblk_t get_sb_block(void **data)
1131 ext4_fsblk_t sb_block;
1132 char *options = (char *) *data;
1134 if (!options || strncmp(options, "sb=", 3) != 0)
1135 return 1; /* Default location */
1137 options += 3;
1138 /* TODO: use simple_strtoll with >32bit ext4 */
1139 sb_block = simple_strtoul(options, &options, 0);
1140 if (*options && *options != ',') {
1141 printk(KERN_ERR "EXT4-fs: Invalid sb specification: %s\n",
1142 (char *) *data);
1143 return 1;
1145 if (*options == ',')
1146 options++;
1147 *data = (void *) options;
1149 return sb_block;
1152 #define DEFAULT_JOURNAL_IOPRIO (IOPRIO_PRIO_VALUE(IOPRIO_CLASS_BE, 3))
1154 static int parse_options(char *options, struct super_block *sb,
1155 unsigned long *journal_devnum,
1156 unsigned int *journal_ioprio,
1157 ext4_fsblk_t *n_blocks_count, int is_remount)
1159 struct ext4_sb_info *sbi = EXT4_SB(sb);
1160 char *p;
1161 substring_t args[MAX_OPT_ARGS];
1162 int data_opt = 0;
1163 int option;
1164 #ifdef CONFIG_QUOTA
1165 int qtype, qfmt;
1166 char *qname;
1167 #endif
1169 if (!options)
1170 return 1;
1172 while ((p = strsep(&options, ",")) != NULL) {
1173 int token;
1174 if (!*p)
1175 continue;
1177 token = match_token(p, tokens, args);
1178 switch (token) {
1179 case Opt_bsd_df:
1180 clear_opt(sbi->s_mount_opt, MINIX_DF);
1181 break;
1182 case Opt_minix_df:
1183 set_opt(sbi->s_mount_opt, MINIX_DF);
1184 break;
1185 case Opt_grpid:
1186 set_opt(sbi->s_mount_opt, GRPID);
1187 break;
1188 case Opt_nogrpid:
1189 clear_opt(sbi->s_mount_opt, GRPID);
1190 break;
1191 case Opt_resuid:
1192 if (match_int(&args[0], &option))
1193 return 0;
1194 sbi->s_resuid = option;
1195 break;
1196 case Opt_resgid:
1197 if (match_int(&args[0], &option))
1198 return 0;
1199 sbi->s_resgid = option;
1200 break;
1201 case Opt_sb:
1202 /* handled by get_sb_block() instead of here */
1203 /* *sb_block = match_int(&args[0]); */
1204 break;
1205 case Opt_err_panic:
1206 clear_opt(sbi->s_mount_opt, ERRORS_CONT);
1207 clear_opt(sbi->s_mount_opt, ERRORS_RO);
1208 set_opt(sbi->s_mount_opt, ERRORS_PANIC);
1209 break;
1210 case Opt_err_ro:
1211 clear_opt(sbi->s_mount_opt, ERRORS_CONT);
1212 clear_opt(sbi->s_mount_opt, ERRORS_PANIC);
1213 set_opt(sbi->s_mount_opt, ERRORS_RO);
1214 break;
1215 case Opt_err_cont:
1216 clear_opt(sbi->s_mount_opt, ERRORS_RO);
1217 clear_opt(sbi->s_mount_opt, ERRORS_PANIC);
1218 set_opt(sbi->s_mount_opt, ERRORS_CONT);
1219 break;
1220 case Opt_nouid32:
1221 set_opt(sbi->s_mount_opt, NO_UID32);
1222 break;
1223 case Opt_debug:
1224 set_opt(sbi->s_mount_opt, DEBUG);
1225 break;
1226 case Opt_oldalloc:
1227 set_opt(sbi->s_mount_opt, OLDALLOC);
1228 break;
1229 case Opt_orlov:
1230 clear_opt(sbi->s_mount_opt, OLDALLOC);
1231 break;
1232 #ifdef CONFIG_EXT4_FS_XATTR
1233 case Opt_user_xattr:
1234 set_opt(sbi->s_mount_opt, XATTR_USER);
1235 break;
1236 case Opt_nouser_xattr:
1237 clear_opt(sbi->s_mount_opt, XATTR_USER);
1238 break;
1239 #else
1240 case Opt_user_xattr:
1241 case Opt_nouser_xattr:
1242 printk(KERN_ERR "EXT4 (no)user_xattr options "
1243 "not supported\n");
1244 break;
1245 #endif
1246 #ifdef CONFIG_EXT4_FS_POSIX_ACL
1247 case Opt_acl:
1248 set_opt(sbi->s_mount_opt, POSIX_ACL);
1249 break;
1250 case Opt_noacl:
1251 clear_opt(sbi->s_mount_opt, POSIX_ACL);
1252 break;
1253 #else
1254 case Opt_acl:
1255 case Opt_noacl:
1256 printk(KERN_ERR "EXT4 (no)acl options "
1257 "not supported\n");
1258 break;
1259 #endif
1260 case Opt_journal_update:
1261 /* @@@ FIXME */
1262 /* Eventually we will want to be able to create
1263 a journal file here. For now, only allow the
1264 user to specify an existing inode to be the
1265 journal file. */
1266 if (is_remount) {
1267 printk(KERN_ERR "EXT4-fs: cannot specify "
1268 "journal on remount\n");
1269 return 0;
1271 set_opt(sbi->s_mount_opt, UPDATE_JOURNAL);
1272 break;
1273 case Opt_journal_dev:
1274 if (is_remount) {
1275 printk(KERN_ERR "EXT4-fs: cannot specify "
1276 "journal on remount\n");
1277 return 0;
1279 if (match_int(&args[0], &option))
1280 return 0;
1281 *journal_devnum = option;
1282 break;
1283 case Opt_journal_checksum:
1284 set_opt(sbi->s_mount_opt, JOURNAL_CHECKSUM);
1285 break;
1286 case Opt_journal_async_commit:
1287 set_opt(sbi->s_mount_opt, JOURNAL_ASYNC_COMMIT);
1288 set_opt(sbi->s_mount_opt, JOURNAL_CHECKSUM);
1289 break;
1290 case Opt_noload:
1291 set_opt(sbi->s_mount_opt, NOLOAD);
1292 break;
1293 case Opt_commit:
1294 if (match_int(&args[0], &option))
1295 return 0;
1296 if (option < 0)
1297 return 0;
1298 if (option == 0)
1299 option = JBD2_DEFAULT_MAX_COMMIT_AGE;
1300 sbi->s_commit_interval = HZ * option;
1301 break;
1302 case Opt_max_batch_time:
1303 if (match_int(&args[0], &option))
1304 return 0;
1305 if (option < 0)
1306 return 0;
1307 if (option == 0)
1308 option = EXT4_DEF_MAX_BATCH_TIME;
1309 sbi->s_max_batch_time = option;
1310 break;
1311 case Opt_min_batch_time:
1312 if (match_int(&args[0], &option))
1313 return 0;
1314 if (option < 0)
1315 return 0;
1316 sbi->s_min_batch_time = option;
1317 break;
1318 case Opt_data_journal:
1319 data_opt = EXT4_MOUNT_JOURNAL_DATA;
1320 goto datacheck;
1321 case Opt_data_ordered:
1322 data_opt = EXT4_MOUNT_ORDERED_DATA;
1323 goto datacheck;
1324 case Opt_data_writeback:
1325 data_opt = EXT4_MOUNT_WRITEBACK_DATA;
1326 datacheck:
1327 if (is_remount) {
1328 if ((sbi->s_mount_opt & EXT4_MOUNT_DATA_FLAGS)
1329 != data_opt) {
1330 printk(KERN_ERR
1331 "EXT4-fs: cannot change data "
1332 "mode on remount\n");
1333 return 0;
1335 } else {
1336 sbi->s_mount_opt &= ~EXT4_MOUNT_DATA_FLAGS;
1337 sbi->s_mount_opt |= data_opt;
1339 break;
1340 case Opt_data_err_abort:
1341 set_opt(sbi->s_mount_opt, DATA_ERR_ABORT);
1342 break;
1343 case Opt_data_err_ignore:
1344 clear_opt(sbi->s_mount_opt, DATA_ERR_ABORT);
1345 break;
1346 case Opt_mb_history_length:
1347 if (match_int(&args[0], &option))
1348 return 0;
1349 if (option < 0)
1350 return 0;
1351 sbi->s_mb_history_max = option;
1352 break;
1353 #ifdef CONFIG_QUOTA
1354 case Opt_usrjquota:
1355 qtype = USRQUOTA;
1356 goto set_qf_name;
1357 case Opt_grpjquota:
1358 qtype = GRPQUOTA;
1359 set_qf_name:
1360 if (sb_any_quota_loaded(sb) &&
1361 !sbi->s_qf_names[qtype]) {
1362 printk(KERN_ERR
1363 "EXT4-fs: Cannot change journaled "
1364 "quota options when quota turned on.\n");
1365 return 0;
1367 qname = match_strdup(&args[0]);
1368 if (!qname) {
1369 printk(KERN_ERR
1370 "EXT4-fs: not enough memory for "
1371 "storing quotafile name.\n");
1372 return 0;
1374 if (sbi->s_qf_names[qtype] &&
1375 strcmp(sbi->s_qf_names[qtype], qname)) {
1376 printk(KERN_ERR
1377 "EXT4-fs: %s quota file already "
1378 "specified.\n", QTYPE2NAME(qtype));
1379 kfree(qname);
1380 return 0;
1382 sbi->s_qf_names[qtype] = qname;
1383 if (strchr(sbi->s_qf_names[qtype], '/')) {
1384 printk(KERN_ERR
1385 "EXT4-fs: quotafile must be on "
1386 "filesystem root.\n");
1387 kfree(sbi->s_qf_names[qtype]);
1388 sbi->s_qf_names[qtype] = NULL;
1389 return 0;
1391 set_opt(sbi->s_mount_opt, QUOTA);
1392 break;
1393 case Opt_offusrjquota:
1394 qtype = USRQUOTA;
1395 goto clear_qf_name;
1396 case Opt_offgrpjquota:
1397 qtype = GRPQUOTA;
1398 clear_qf_name:
1399 if (sb_any_quota_loaded(sb) &&
1400 sbi->s_qf_names[qtype]) {
1401 printk(KERN_ERR "EXT4-fs: Cannot change "
1402 "journaled quota options when "
1403 "quota turned on.\n");
1404 return 0;
1407 * The space will be released later when all options
1408 * are confirmed to be correct
1410 sbi->s_qf_names[qtype] = NULL;
1411 break;
1412 case Opt_jqfmt_vfsold:
1413 qfmt = QFMT_VFS_OLD;
1414 goto set_qf_format;
1415 case Opt_jqfmt_vfsv0:
1416 qfmt = QFMT_VFS_V0;
1417 set_qf_format:
1418 if (sb_any_quota_loaded(sb) &&
1419 sbi->s_jquota_fmt != qfmt) {
1420 printk(KERN_ERR "EXT4-fs: Cannot change "
1421 "journaled quota options when "
1422 "quota turned on.\n");
1423 return 0;
1425 sbi->s_jquota_fmt = qfmt;
1426 break;
1427 case Opt_quota:
1428 case Opt_usrquota:
1429 set_opt(sbi->s_mount_opt, QUOTA);
1430 set_opt(sbi->s_mount_opt, USRQUOTA);
1431 break;
1432 case Opt_grpquota:
1433 set_opt(sbi->s_mount_opt, QUOTA);
1434 set_opt(sbi->s_mount_opt, GRPQUOTA);
1435 break;
1436 case Opt_noquota:
1437 if (sb_any_quota_loaded(sb)) {
1438 printk(KERN_ERR "EXT4-fs: Cannot change quota "
1439 "options when quota turned on.\n");
1440 return 0;
1442 clear_opt(sbi->s_mount_opt, QUOTA);
1443 clear_opt(sbi->s_mount_opt, USRQUOTA);
1444 clear_opt(sbi->s_mount_opt, GRPQUOTA);
1445 break;
1446 #else
1447 case Opt_quota:
1448 case Opt_usrquota:
1449 case Opt_grpquota:
1450 printk(KERN_ERR
1451 "EXT4-fs: quota options not supported.\n");
1452 break;
1453 case Opt_usrjquota:
1454 case Opt_grpjquota:
1455 case Opt_offusrjquota:
1456 case Opt_offgrpjquota:
1457 case Opt_jqfmt_vfsold:
1458 case Opt_jqfmt_vfsv0:
1459 printk(KERN_ERR
1460 "EXT4-fs: journaled quota options not "
1461 "supported.\n");
1462 break;
1463 case Opt_noquota:
1464 break;
1465 #endif
1466 case Opt_abort:
1467 set_opt(sbi->s_mount_opt, ABORT);
1468 break;
1469 case Opt_nobarrier:
1470 clear_opt(sbi->s_mount_opt, BARRIER);
1471 break;
1472 case Opt_barrier:
1473 if (match_int(&args[0], &option)) {
1474 set_opt(sbi->s_mount_opt, BARRIER);
1475 break;
1477 if (option)
1478 set_opt(sbi->s_mount_opt, BARRIER);
1479 else
1480 clear_opt(sbi->s_mount_opt, BARRIER);
1481 break;
1482 case Opt_ignore:
1483 break;
1484 case Opt_resize:
1485 if (!is_remount) {
1486 printk("EXT4-fs: resize option only available "
1487 "for remount\n");
1488 return 0;
1490 if (match_int(&args[0], &option) != 0)
1491 return 0;
1492 *n_blocks_count = option;
1493 break;
1494 case Opt_nobh:
1495 set_opt(sbi->s_mount_opt, NOBH);
1496 break;
1497 case Opt_bh:
1498 clear_opt(sbi->s_mount_opt, NOBH);
1499 break;
1500 case Opt_i_version:
1501 set_opt(sbi->s_mount_opt, I_VERSION);
1502 sb->s_flags |= MS_I_VERSION;
1503 break;
1504 case Opt_nodelalloc:
1505 clear_opt(sbi->s_mount_opt, DELALLOC);
1506 break;
1507 case Opt_stripe:
1508 if (match_int(&args[0], &option))
1509 return 0;
1510 if (option < 0)
1511 return 0;
1512 sbi->s_stripe = option;
1513 break;
1514 case Opt_delalloc:
1515 set_opt(sbi->s_mount_opt, DELALLOC);
1516 break;
1517 case Opt_block_validity:
1518 set_opt(sbi->s_mount_opt, BLOCK_VALIDITY);
1519 break;
1520 case Opt_noblock_validity:
1521 clear_opt(sbi->s_mount_opt, BLOCK_VALIDITY);
1522 break;
1523 case Opt_inode_readahead_blks:
1524 if (match_int(&args[0], &option))
1525 return 0;
1526 if (option < 0 || option > (1 << 30))
1527 return 0;
1528 if (!is_power_of_2(option)) {
1529 printk(KERN_ERR "EXT4-fs: inode_readahead_blks"
1530 " must be a power of 2\n");
1531 return 0;
1533 sbi->s_inode_readahead_blks = option;
1534 break;
1535 case Opt_journal_ioprio:
1536 if (match_int(&args[0], &option))
1537 return 0;
1538 if (option < 0 || option > 7)
1539 break;
1540 *journal_ioprio = IOPRIO_PRIO_VALUE(IOPRIO_CLASS_BE,
1541 option);
1542 break;
1543 case Opt_noauto_da_alloc:
1544 set_opt(sbi->s_mount_opt,NO_AUTO_DA_ALLOC);
1545 break;
1546 case Opt_auto_da_alloc:
1547 if (match_int(&args[0], &option)) {
1548 clear_opt(sbi->s_mount_opt, NO_AUTO_DA_ALLOC);
1549 break;
1551 if (option)
1552 clear_opt(sbi->s_mount_opt, NO_AUTO_DA_ALLOC);
1553 else
1554 set_opt(sbi->s_mount_opt,NO_AUTO_DA_ALLOC);
1555 break;
1556 default:
1557 printk(KERN_ERR
1558 "EXT4-fs: Unrecognized mount option \"%s\" "
1559 "or missing value\n", p);
1560 return 0;
1563 #ifdef CONFIG_QUOTA
1564 if (sbi->s_qf_names[USRQUOTA] || sbi->s_qf_names[GRPQUOTA]) {
1565 if ((sbi->s_mount_opt & EXT4_MOUNT_USRQUOTA) &&
1566 sbi->s_qf_names[USRQUOTA])
1567 clear_opt(sbi->s_mount_opt, USRQUOTA);
1569 if ((sbi->s_mount_opt & EXT4_MOUNT_GRPQUOTA) &&
1570 sbi->s_qf_names[GRPQUOTA])
1571 clear_opt(sbi->s_mount_opt, GRPQUOTA);
1573 if ((sbi->s_qf_names[USRQUOTA] &&
1574 (sbi->s_mount_opt & EXT4_MOUNT_GRPQUOTA)) ||
1575 (sbi->s_qf_names[GRPQUOTA] &&
1576 (sbi->s_mount_opt & EXT4_MOUNT_USRQUOTA))) {
1577 printk(KERN_ERR "EXT4-fs: old and new quota "
1578 "format mixing.\n");
1579 return 0;
1582 if (!sbi->s_jquota_fmt) {
1583 printk(KERN_ERR "EXT4-fs: journaled quota format "
1584 "not specified.\n");
1585 return 0;
1587 } else {
1588 if (sbi->s_jquota_fmt) {
1589 printk(KERN_ERR "EXT4-fs: journaled quota format "
1590 "specified with no journaling "
1591 "enabled.\n");
1592 return 0;
1595 #endif
1596 return 1;
1599 static int ext4_setup_super(struct super_block *sb, struct ext4_super_block *es,
1600 int read_only)
1602 struct ext4_sb_info *sbi = EXT4_SB(sb);
1603 int res = 0;
1605 if (le32_to_cpu(es->s_rev_level) > EXT4_MAX_SUPP_REV) {
1606 printk(KERN_ERR "EXT4-fs warning: revision level too high, "
1607 "forcing read-only mode\n");
1608 res = MS_RDONLY;
1610 if (read_only)
1611 return res;
1612 if (!(sbi->s_mount_state & EXT4_VALID_FS))
1613 printk(KERN_WARNING "EXT4-fs warning: mounting unchecked fs, "
1614 "running e2fsck is recommended\n");
1615 else if ((sbi->s_mount_state & EXT4_ERROR_FS))
1616 printk(KERN_WARNING
1617 "EXT4-fs warning: mounting fs with errors, "
1618 "running e2fsck is recommended\n");
1619 else if ((__s16) le16_to_cpu(es->s_max_mnt_count) >= 0 &&
1620 le16_to_cpu(es->s_mnt_count) >=
1621 (unsigned short) (__s16) le16_to_cpu(es->s_max_mnt_count))
1622 printk(KERN_WARNING
1623 "EXT4-fs warning: maximal mount count reached, "
1624 "running e2fsck is recommended\n");
1625 else if (le32_to_cpu(es->s_checkinterval) &&
1626 (le32_to_cpu(es->s_lastcheck) +
1627 le32_to_cpu(es->s_checkinterval) <= get_seconds()))
1628 printk(KERN_WARNING
1629 "EXT4-fs warning: checktime reached, "
1630 "running e2fsck is recommended\n");
1631 if (!sbi->s_journal)
1632 es->s_state &= cpu_to_le16(~EXT4_VALID_FS);
1633 if (!(__s16) le16_to_cpu(es->s_max_mnt_count))
1634 es->s_max_mnt_count = cpu_to_le16(EXT4_DFL_MAX_MNT_COUNT);
1635 le16_add_cpu(&es->s_mnt_count, 1);
1636 es->s_mtime = cpu_to_le32(get_seconds());
1637 ext4_update_dynamic_rev(sb);
1638 if (sbi->s_journal)
1639 EXT4_SET_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
1641 ext4_commit_super(sb, 1);
1642 if (test_opt(sb, DEBUG))
1643 printk(KERN_INFO "[EXT4 FS bs=%lu, gc=%u, "
1644 "bpg=%lu, ipg=%lu, mo=%04lx]\n",
1645 sb->s_blocksize,
1646 sbi->s_groups_count,
1647 EXT4_BLOCKS_PER_GROUP(sb),
1648 EXT4_INODES_PER_GROUP(sb),
1649 sbi->s_mount_opt);
1651 if (EXT4_SB(sb)->s_journal) {
1652 printk(KERN_INFO "EXT4 FS on %s, %s journal on %s\n",
1653 sb->s_id, EXT4_SB(sb)->s_journal->j_inode ? "internal" :
1654 "external", EXT4_SB(sb)->s_journal->j_devname);
1655 } else {
1656 printk(KERN_INFO "EXT4 FS on %s, no journal\n", sb->s_id);
1658 return res;
1661 static int ext4_fill_flex_info(struct super_block *sb)
1663 struct ext4_sb_info *sbi = EXT4_SB(sb);
1664 struct ext4_group_desc *gdp = NULL;
1665 ext4_group_t flex_group_count;
1666 ext4_group_t flex_group;
1667 int groups_per_flex = 0;
1668 size_t size;
1669 int i;
1671 if (!sbi->s_es->s_log_groups_per_flex) {
1672 sbi->s_log_groups_per_flex = 0;
1673 return 1;
1676 sbi->s_log_groups_per_flex = sbi->s_es->s_log_groups_per_flex;
1677 groups_per_flex = 1 << sbi->s_log_groups_per_flex;
1679 /* We allocate both existing and potentially added groups */
1680 flex_group_count = ((sbi->s_groups_count + groups_per_flex - 1) +
1681 ((le16_to_cpu(sbi->s_es->s_reserved_gdt_blocks) + 1) <<
1682 EXT4_DESC_PER_BLOCK_BITS(sb))) / groups_per_flex;
1683 size = flex_group_count * sizeof(struct flex_groups);
1684 sbi->s_flex_groups = kzalloc(size, GFP_KERNEL);
1685 if (sbi->s_flex_groups == NULL) {
1686 sbi->s_flex_groups = vmalloc(size);
1687 if (sbi->s_flex_groups)
1688 memset(sbi->s_flex_groups, 0, size);
1690 if (sbi->s_flex_groups == NULL) {
1691 printk(KERN_ERR "EXT4-fs: not enough memory for "
1692 "%u flex groups\n", flex_group_count);
1693 goto failed;
1696 for (i = 0; i < sbi->s_groups_count; i++) {
1697 gdp = ext4_get_group_desc(sb, i, NULL);
1699 flex_group = ext4_flex_group(sbi, i);
1700 atomic_set(&sbi->s_flex_groups[flex_group].free_inodes,
1701 ext4_free_inodes_count(sb, gdp));
1702 atomic_set(&sbi->s_flex_groups[flex_group].free_blocks,
1703 ext4_free_blks_count(sb, gdp));
1704 atomic_set(&sbi->s_flex_groups[flex_group].used_dirs,
1705 ext4_used_dirs_count(sb, gdp));
1708 return 1;
1709 failed:
1710 return 0;
1713 __le16 ext4_group_desc_csum(struct ext4_sb_info *sbi, __u32 block_group,
1714 struct ext4_group_desc *gdp)
1716 __u16 crc = 0;
1718 if (sbi->s_es->s_feature_ro_compat &
1719 cpu_to_le32(EXT4_FEATURE_RO_COMPAT_GDT_CSUM)) {
1720 int offset = offsetof(struct ext4_group_desc, bg_checksum);
1721 __le32 le_group = cpu_to_le32(block_group);
1723 crc = crc16(~0, sbi->s_es->s_uuid, sizeof(sbi->s_es->s_uuid));
1724 crc = crc16(crc, (__u8 *)&le_group, sizeof(le_group));
1725 crc = crc16(crc, (__u8 *)gdp, offset);
1726 offset += sizeof(gdp->bg_checksum); /* skip checksum */
1727 /* for checksum of struct ext4_group_desc do the rest...*/
1728 if ((sbi->s_es->s_feature_incompat &
1729 cpu_to_le32(EXT4_FEATURE_INCOMPAT_64BIT)) &&
1730 offset < le16_to_cpu(sbi->s_es->s_desc_size))
1731 crc = crc16(crc, (__u8 *)gdp + offset,
1732 le16_to_cpu(sbi->s_es->s_desc_size) -
1733 offset);
1736 return cpu_to_le16(crc);
1739 int ext4_group_desc_csum_verify(struct ext4_sb_info *sbi, __u32 block_group,
1740 struct ext4_group_desc *gdp)
1742 if ((sbi->s_es->s_feature_ro_compat &
1743 cpu_to_le32(EXT4_FEATURE_RO_COMPAT_GDT_CSUM)) &&
1744 (gdp->bg_checksum != ext4_group_desc_csum(sbi, block_group, gdp)))
1745 return 0;
1747 return 1;
1750 /* Called at mount-time, super-block is locked */
1751 static int ext4_check_descriptors(struct super_block *sb)
1753 struct ext4_sb_info *sbi = EXT4_SB(sb);
1754 ext4_fsblk_t first_block = le32_to_cpu(sbi->s_es->s_first_data_block);
1755 ext4_fsblk_t last_block;
1756 ext4_fsblk_t block_bitmap;
1757 ext4_fsblk_t inode_bitmap;
1758 ext4_fsblk_t inode_table;
1759 int flexbg_flag = 0;
1760 ext4_group_t i;
1762 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_FLEX_BG))
1763 flexbg_flag = 1;
1765 ext4_debug("Checking group descriptors");
1767 for (i = 0; i < sbi->s_groups_count; i++) {
1768 struct ext4_group_desc *gdp = ext4_get_group_desc(sb, i, NULL);
1770 if (i == sbi->s_groups_count - 1 || flexbg_flag)
1771 last_block = ext4_blocks_count(sbi->s_es) - 1;
1772 else
1773 last_block = first_block +
1774 (EXT4_BLOCKS_PER_GROUP(sb) - 1);
1776 block_bitmap = ext4_block_bitmap(sb, gdp);
1777 if (block_bitmap < first_block || block_bitmap > last_block) {
1778 printk(KERN_ERR "EXT4-fs: ext4_check_descriptors: "
1779 "Block bitmap for group %u not in group "
1780 "(block %llu)!\n", i, block_bitmap);
1781 return 0;
1783 inode_bitmap = ext4_inode_bitmap(sb, gdp);
1784 if (inode_bitmap < first_block || inode_bitmap > last_block) {
1785 printk(KERN_ERR "EXT4-fs: ext4_check_descriptors: "
1786 "Inode bitmap for group %u not in group "
1787 "(block %llu)!\n", i, inode_bitmap);
1788 return 0;
1790 inode_table = ext4_inode_table(sb, gdp);
1791 if (inode_table < first_block ||
1792 inode_table + sbi->s_itb_per_group - 1 > last_block) {
1793 printk(KERN_ERR "EXT4-fs: ext4_check_descriptors: "
1794 "Inode table for group %u not in group "
1795 "(block %llu)!\n", i, inode_table);
1796 return 0;
1798 ext4_lock_group(sb, i);
1799 if (!ext4_group_desc_csum_verify(sbi, i, gdp)) {
1800 printk(KERN_ERR "EXT4-fs: ext4_check_descriptors: "
1801 "Checksum for group %u failed (%u!=%u)\n",
1802 i, le16_to_cpu(ext4_group_desc_csum(sbi, i,
1803 gdp)), le16_to_cpu(gdp->bg_checksum));
1804 if (!(sb->s_flags & MS_RDONLY)) {
1805 ext4_unlock_group(sb, i);
1806 return 0;
1809 ext4_unlock_group(sb, i);
1810 if (!flexbg_flag)
1811 first_block += EXT4_BLOCKS_PER_GROUP(sb);
1814 ext4_free_blocks_count_set(sbi->s_es, ext4_count_free_blocks(sb));
1815 sbi->s_es->s_free_inodes_count =cpu_to_le32(ext4_count_free_inodes(sb));
1816 return 1;
1819 /* ext4_orphan_cleanup() walks a singly-linked list of inodes (starting at
1820 * the superblock) which were deleted from all directories, but held open by
1821 * a process at the time of a crash. We walk the list and try to delete these
1822 * inodes at recovery time (only with a read-write filesystem).
1824 * In order to keep the orphan inode chain consistent during traversal (in
1825 * case of crash during recovery), we link each inode into the superblock
1826 * orphan list_head and handle it the same way as an inode deletion during
1827 * normal operation (which journals the operations for us).
1829 * We only do an iget() and an iput() on each inode, which is very safe if we
1830 * accidentally point at an in-use or already deleted inode. The worst that
1831 * can happen in this case is that we get a "bit already cleared" message from
1832 * ext4_free_inode(). The only reason we would point at a wrong inode is if
1833 * e2fsck was run on this filesystem, and it must have already done the orphan
1834 * inode cleanup for us, so we can safely abort without any further action.
1836 static void ext4_orphan_cleanup(struct super_block *sb,
1837 struct ext4_super_block *es)
1839 unsigned int s_flags = sb->s_flags;
1840 int nr_orphans = 0, nr_truncates = 0;
1841 #ifdef CONFIG_QUOTA
1842 int i;
1843 #endif
1844 if (!es->s_last_orphan) {
1845 jbd_debug(4, "no orphan inodes to clean up\n");
1846 return;
1849 if (bdev_read_only(sb->s_bdev)) {
1850 printk(KERN_ERR "EXT4-fs: write access "
1851 "unavailable, skipping orphan cleanup.\n");
1852 return;
1855 if (EXT4_SB(sb)->s_mount_state & EXT4_ERROR_FS) {
1856 if (es->s_last_orphan)
1857 jbd_debug(1, "Errors on filesystem, "
1858 "clearing orphan list.\n");
1859 es->s_last_orphan = 0;
1860 jbd_debug(1, "Skipping orphan recovery on fs with errors.\n");
1861 return;
1864 if (s_flags & MS_RDONLY) {
1865 printk(KERN_INFO "EXT4-fs: %s: orphan cleanup on readonly fs\n",
1866 sb->s_id);
1867 sb->s_flags &= ~MS_RDONLY;
1869 #ifdef CONFIG_QUOTA
1870 /* Needed for iput() to work correctly and not trash data */
1871 sb->s_flags |= MS_ACTIVE;
1872 /* Turn on quotas so that they are updated correctly */
1873 for (i = 0; i < MAXQUOTAS; i++) {
1874 if (EXT4_SB(sb)->s_qf_names[i]) {
1875 int ret = ext4_quota_on_mount(sb, i);
1876 if (ret < 0)
1877 printk(KERN_ERR
1878 "EXT4-fs: Cannot turn on journaled "
1879 "quota: error %d\n", ret);
1882 #endif
1884 while (es->s_last_orphan) {
1885 struct inode *inode;
1887 inode = ext4_orphan_get(sb, le32_to_cpu(es->s_last_orphan));
1888 if (IS_ERR(inode)) {
1889 es->s_last_orphan = 0;
1890 break;
1893 list_add(&EXT4_I(inode)->i_orphan, &EXT4_SB(sb)->s_orphan);
1894 vfs_dq_init(inode);
1895 if (inode->i_nlink) {
1896 printk(KERN_DEBUG
1897 "%s: truncating inode %lu to %lld bytes\n",
1898 __func__, inode->i_ino, inode->i_size);
1899 jbd_debug(2, "truncating inode %lu to %lld bytes\n",
1900 inode->i_ino, inode->i_size);
1901 ext4_truncate(inode);
1902 nr_truncates++;
1903 } else {
1904 printk(KERN_DEBUG
1905 "%s: deleting unreferenced inode %lu\n",
1906 __func__, inode->i_ino);
1907 jbd_debug(2, "deleting unreferenced inode %lu\n",
1908 inode->i_ino);
1909 nr_orphans++;
1911 iput(inode); /* The delete magic happens here! */
1914 #define PLURAL(x) (x), ((x) == 1) ? "" : "s"
1916 if (nr_orphans)
1917 printk(KERN_INFO "EXT4-fs: %s: %d orphan inode%s deleted\n",
1918 sb->s_id, PLURAL(nr_orphans));
1919 if (nr_truncates)
1920 printk(KERN_INFO "EXT4-fs: %s: %d truncate%s cleaned up\n",
1921 sb->s_id, PLURAL(nr_truncates));
1922 #ifdef CONFIG_QUOTA
1923 /* Turn quotas off */
1924 for (i = 0; i < MAXQUOTAS; i++) {
1925 if (sb_dqopt(sb)->files[i])
1926 vfs_quota_off(sb, i, 0);
1928 #endif
1929 sb->s_flags = s_flags; /* Restore MS_RDONLY status */
1933 * Maximal extent format file size.
1934 * Resulting logical blkno at s_maxbytes must fit in our on-disk
1935 * extent format containers, within a sector_t, and within i_blocks
1936 * in the vfs. ext4 inode has 48 bits of i_block in fsblock units,
1937 * so that won't be a limiting factor.
1939 * Note, this does *not* consider any metadata overhead for vfs i_blocks.
1941 static loff_t ext4_max_size(int blkbits, int has_huge_files)
1943 loff_t res;
1944 loff_t upper_limit = MAX_LFS_FILESIZE;
1946 /* small i_blocks in vfs inode? */
1947 if (!has_huge_files || sizeof(blkcnt_t) < sizeof(u64)) {
1949 * CONFIG_LBD is not enabled implies the inode
1950 * i_block represent total blocks in 512 bytes
1951 * 32 == size of vfs inode i_blocks * 8
1953 upper_limit = (1LL << 32) - 1;
1955 /* total blocks in file system block size */
1956 upper_limit >>= (blkbits - 9);
1957 upper_limit <<= blkbits;
1960 /* 32-bit extent-start container, ee_block */
1961 res = 1LL << 32;
1962 res <<= blkbits;
1963 res -= 1;
1965 /* Sanity check against vm- & vfs- imposed limits */
1966 if (res > upper_limit)
1967 res = upper_limit;
1969 return res;
1973 * Maximal bitmap file size. There is a direct, and {,double-,triple-}indirect
1974 * block limit, and also a limit of (2^48 - 1) 512-byte sectors in i_blocks.
1975 * We need to be 1 filesystem block less than the 2^48 sector limit.
1977 static loff_t ext4_max_bitmap_size(int bits, int has_huge_files)
1979 loff_t res = EXT4_NDIR_BLOCKS;
1980 int meta_blocks;
1981 loff_t upper_limit;
1982 /* This is calculated to be the largest file size for a dense, block
1983 * mapped file such that the file's total number of 512-byte sectors,
1984 * including data and all indirect blocks, does not exceed (2^48 - 1).
1986 * __u32 i_blocks_lo and _u16 i_blocks_high represent the total
1987 * number of 512-byte sectors of the file.
1990 if (!has_huge_files || sizeof(blkcnt_t) < sizeof(u64)) {
1992 * !has_huge_files or CONFIG_LBD not enabled implies that
1993 * the inode i_block field represents total file blocks in
1994 * 2^32 512-byte sectors == size of vfs inode i_blocks * 8
1996 upper_limit = (1LL << 32) - 1;
1998 /* total blocks in file system block size */
1999 upper_limit >>= (bits - 9);
2001 } else {
2003 * We use 48 bit ext4_inode i_blocks
2004 * With EXT4_HUGE_FILE_FL set the i_blocks
2005 * represent total number of blocks in
2006 * file system block size
2008 upper_limit = (1LL << 48) - 1;
2012 /* indirect blocks */
2013 meta_blocks = 1;
2014 /* double indirect blocks */
2015 meta_blocks += 1 + (1LL << (bits-2));
2016 /* tripple indirect blocks */
2017 meta_blocks += 1 + (1LL << (bits-2)) + (1LL << (2*(bits-2)));
2019 upper_limit -= meta_blocks;
2020 upper_limit <<= bits;
2022 res += 1LL << (bits-2);
2023 res += 1LL << (2*(bits-2));
2024 res += 1LL << (3*(bits-2));
2025 res <<= bits;
2026 if (res > upper_limit)
2027 res = upper_limit;
2029 if (res > MAX_LFS_FILESIZE)
2030 res = MAX_LFS_FILESIZE;
2032 return res;
2035 static ext4_fsblk_t descriptor_loc(struct super_block *sb,
2036 ext4_fsblk_t logical_sb_block, int nr)
2038 struct ext4_sb_info *sbi = EXT4_SB(sb);
2039 ext4_group_t bg, first_meta_bg;
2040 int has_super = 0;
2042 first_meta_bg = le32_to_cpu(sbi->s_es->s_first_meta_bg);
2044 if (!EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_META_BG) ||
2045 nr < first_meta_bg)
2046 return logical_sb_block + nr + 1;
2047 bg = sbi->s_desc_per_block * nr;
2048 if (ext4_bg_has_super(sb, bg))
2049 has_super = 1;
2051 return (has_super + ext4_group_first_block_no(sb, bg));
2055 * ext4_get_stripe_size: Get the stripe size.
2056 * @sbi: In memory super block info
2058 * If we have specified it via mount option, then
2059 * use the mount option value. If the value specified at mount time is
2060 * greater than the blocks per group use the super block value.
2061 * If the super block value is greater than blocks per group return 0.
2062 * Allocator needs it be less than blocks per group.
2065 static unsigned long ext4_get_stripe_size(struct ext4_sb_info *sbi)
2067 unsigned long stride = le16_to_cpu(sbi->s_es->s_raid_stride);
2068 unsigned long stripe_width =
2069 le32_to_cpu(sbi->s_es->s_raid_stripe_width);
2071 if (sbi->s_stripe && sbi->s_stripe <= sbi->s_blocks_per_group)
2072 return sbi->s_stripe;
2074 if (stripe_width <= sbi->s_blocks_per_group)
2075 return stripe_width;
2077 if (stride <= sbi->s_blocks_per_group)
2078 return stride;
2080 return 0;
2083 /* sysfs supprt */
2085 struct ext4_attr {
2086 struct attribute attr;
2087 ssize_t (*show)(struct ext4_attr *, struct ext4_sb_info *, char *);
2088 ssize_t (*store)(struct ext4_attr *, struct ext4_sb_info *,
2089 const char *, size_t);
2090 int offset;
2093 static int parse_strtoul(const char *buf,
2094 unsigned long max, unsigned long *value)
2096 char *endp;
2098 while (*buf && isspace(*buf))
2099 buf++;
2100 *value = simple_strtoul(buf, &endp, 0);
2101 while (*endp && isspace(*endp))
2102 endp++;
2103 if (*endp || *value > max)
2104 return -EINVAL;
2106 return 0;
2109 static ssize_t delayed_allocation_blocks_show(struct ext4_attr *a,
2110 struct ext4_sb_info *sbi,
2111 char *buf)
2113 return snprintf(buf, PAGE_SIZE, "%llu\n",
2114 (s64) percpu_counter_sum(&sbi->s_dirtyblocks_counter));
2117 static ssize_t session_write_kbytes_show(struct ext4_attr *a,
2118 struct ext4_sb_info *sbi, char *buf)
2120 struct super_block *sb = sbi->s_buddy_cache->i_sb;
2122 return snprintf(buf, PAGE_SIZE, "%lu\n",
2123 (part_stat_read(sb->s_bdev->bd_part, sectors[1]) -
2124 sbi->s_sectors_written_start) >> 1);
2127 static ssize_t lifetime_write_kbytes_show(struct ext4_attr *a,
2128 struct ext4_sb_info *sbi, char *buf)
2130 struct super_block *sb = sbi->s_buddy_cache->i_sb;
2132 return snprintf(buf, PAGE_SIZE, "%llu\n",
2133 sbi->s_kbytes_written +
2134 ((part_stat_read(sb->s_bdev->bd_part, sectors[1]) -
2135 EXT4_SB(sb)->s_sectors_written_start) >> 1));
2138 static ssize_t inode_readahead_blks_store(struct ext4_attr *a,
2139 struct ext4_sb_info *sbi,
2140 const char *buf, size_t count)
2142 unsigned long t;
2144 if (parse_strtoul(buf, 0x40000000, &t))
2145 return -EINVAL;
2147 if (!is_power_of_2(t))
2148 return -EINVAL;
2150 sbi->s_inode_readahead_blks = t;
2151 return count;
2154 static ssize_t sbi_ui_show(struct ext4_attr *a,
2155 struct ext4_sb_info *sbi, char *buf)
2157 unsigned int *ui = (unsigned int *) (((char *) sbi) + a->offset);
2159 return snprintf(buf, PAGE_SIZE, "%u\n", *ui);
2162 static ssize_t sbi_ui_store(struct ext4_attr *a,
2163 struct ext4_sb_info *sbi,
2164 const char *buf, size_t count)
2166 unsigned int *ui = (unsigned int *) (((char *) sbi) + a->offset);
2167 unsigned long t;
2169 if (parse_strtoul(buf, 0xffffffff, &t))
2170 return -EINVAL;
2171 *ui = t;
2172 return count;
2175 #define EXT4_ATTR_OFFSET(_name,_mode,_show,_store,_elname) \
2176 static struct ext4_attr ext4_attr_##_name = { \
2177 .attr = {.name = __stringify(_name), .mode = _mode }, \
2178 .show = _show, \
2179 .store = _store, \
2180 .offset = offsetof(struct ext4_sb_info, _elname), \
2182 #define EXT4_ATTR(name, mode, show, store) \
2183 static struct ext4_attr ext4_attr_##name = __ATTR(name, mode, show, store)
2185 #define EXT4_RO_ATTR(name) EXT4_ATTR(name, 0444, name##_show, NULL)
2186 #define EXT4_RW_ATTR(name) EXT4_ATTR(name, 0644, name##_show, name##_store)
2187 #define EXT4_RW_ATTR_SBI_UI(name, elname) \
2188 EXT4_ATTR_OFFSET(name, 0644, sbi_ui_show, sbi_ui_store, elname)
2189 #define ATTR_LIST(name) &ext4_attr_##name.attr
2191 EXT4_RO_ATTR(delayed_allocation_blocks);
2192 EXT4_RO_ATTR(session_write_kbytes);
2193 EXT4_RO_ATTR(lifetime_write_kbytes);
2194 EXT4_ATTR_OFFSET(inode_readahead_blks, 0644, sbi_ui_show,
2195 inode_readahead_blks_store, s_inode_readahead_blks);
2196 EXT4_RW_ATTR_SBI_UI(mb_stats, s_mb_stats);
2197 EXT4_RW_ATTR_SBI_UI(mb_max_to_scan, s_mb_max_to_scan);
2198 EXT4_RW_ATTR_SBI_UI(mb_min_to_scan, s_mb_min_to_scan);
2199 EXT4_RW_ATTR_SBI_UI(mb_order2_req, s_mb_order2_reqs);
2200 EXT4_RW_ATTR_SBI_UI(mb_stream_req, s_mb_stream_request);
2201 EXT4_RW_ATTR_SBI_UI(mb_group_prealloc, s_mb_group_prealloc);
2203 static struct attribute *ext4_attrs[] = {
2204 ATTR_LIST(delayed_allocation_blocks),
2205 ATTR_LIST(session_write_kbytes),
2206 ATTR_LIST(lifetime_write_kbytes),
2207 ATTR_LIST(inode_readahead_blks),
2208 ATTR_LIST(mb_stats),
2209 ATTR_LIST(mb_max_to_scan),
2210 ATTR_LIST(mb_min_to_scan),
2211 ATTR_LIST(mb_order2_req),
2212 ATTR_LIST(mb_stream_req),
2213 ATTR_LIST(mb_group_prealloc),
2214 NULL,
2217 static ssize_t ext4_attr_show(struct kobject *kobj,
2218 struct attribute *attr, char *buf)
2220 struct ext4_sb_info *sbi = container_of(kobj, struct ext4_sb_info,
2221 s_kobj);
2222 struct ext4_attr *a = container_of(attr, struct ext4_attr, attr);
2224 return a->show ? a->show(a, sbi, buf) : 0;
2227 static ssize_t ext4_attr_store(struct kobject *kobj,
2228 struct attribute *attr,
2229 const char *buf, size_t len)
2231 struct ext4_sb_info *sbi = container_of(kobj, struct ext4_sb_info,
2232 s_kobj);
2233 struct ext4_attr *a = container_of(attr, struct ext4_attr, attr);
2235 return a->store ? a->store(a, sbi, buf, len) : 0;
2238 static void ext4_sb_release(struct kobject *kobj)
2240 struct ext4_sb_info *sbi = container_of(kobj, struct ext4_sb_info,
2241 s_kobj);
2242 complete(&sbi->s_kobj_unregister);
2246 static struct sysfs_ops ext4_attr_ops = {
2247 .show = ext4_attr_show,
2248 .store = ext4_attr_store,
2251 static struct kobj_type ext4_ktype = {
2252 .default_attrs = ext4_attrs,
2253 .sysfs_ops = &ext4_attr_ops,
2254 .release = ext4_sb_release,
2257 static int ext4_fill_super(struct super_block *sb, void *data, int silent)
2258 __releases(kernel_lock)
2259 __acquires(kernel_lock)
2261 struct buffer_head *bh;
2262 struct ext4_super_block *es = NULL;
2263 struct ext4_sb_info *sbi;
2264 ext4_fsblk_t block;
2265 ext4_fsblk_t sb_block = get_sb_block(&data);
2266 ext4_fsblk_t logical_sb_block;
2267 unsigned long offset = 0;
2268 unsigned long journal_devnum = 0;
2269 unsigned long def_mount_opts;
2270 struct inode *root;
2271 char *cp;
2272 const char *descr;
2273 int ret = -EINVAL;
2274 int blocksize;
2275 unsigned int db_count;
2276 unsigned int i;
2277 int needs_recovery, has_huge_files;
2278 int features;
2279 __u64 blocks_count;
2280 int err;
2281 unsigned int journal_ioprio = DEFAULT_JOURNAL_IOPRIO;
2283 sbi = kzalloc(sizeof(*sbi), GFP_KERNEL);
2284 if (!sbi)
2285 return -ENOMEM;
2287 sbi->s_blockgroup_lock =
2288 kzalloc(sizeof(struct blockgroup_lock), GFP_KERNEL);
2289 if (!sbi->s_blockgroup_lock) {
2290 kfree(sbi);
2291 return -ENOMEM;
2293 sb->s_fs_info = sbi;
2294 sbi->s_mount_opt = 0;
2295 sbi->s_resuid = EXT4_DEF_RESUID;
2296 sbi->s_resgid = EXT4_DEF_RESGID;
2297 sbi->s_inode_readahead_blks = EXT4_DEF_INODE_READAHEAD_BLKS;
2298 sbi->s_sb_block = sb_block;
2299 sbi->s_sectors_written_start = part_stat_read(sb->s_bdev->bd_part,
2300 sectors[1]);
2302 unlock_kernel();
2304 /* Cleanup superblock name */
2305 for (cp = sb->s_id; (cp = strchr(cp, '/'));)
2306 *cp = '!';
2308 blocksize = sb_min_blocksize(sb, EXT4_MIN_BLOCK_SIZE);
2309 if (!blocksize) {
2310 printk(KERN_ERR "EXT4-fs: unable to set blocksize\n");
2311 goto out_fail;
2315 * The ext4 superblock will not be buffer aligned for other than 1kB
2316 * block sizes. We need to calculate the offset from buffer start.
2318 if (blocksize != EXT4_MIN_BLOCK_SIZE) {
2319 logical_sb_block = sb_block * EXT4_MIN_BLOCK_SIZE;
2320 offset = do_div(logical_sb_block, blocksize);
2321 } else {
2322 logical_sb_block = sb_block;
2325 if (!(bh = sb_bread(sb, logical_sb_block))) {
2326 printk(KERN_ERR "EXT4-fs: unable to read superblock\n");
2327 goto out_fail;
2330 * Note: s_es must be initialized as soon as possible because
2331 * some ext4 macro-instructions depend on its value
2333 es = (struct ext4_super_block *) (((char *)bh->b_data) + offset);
2334 sbi->s_es = es;
2335 sb->s_magic = le16_to_cpu(es->s_magic);
2336 if (sb->s_magic != EXT4_SUPER_MAGIC)
2337 goto cantfind_ext4;
2338 sbi->s_kbytes_written = le64_to_cpu(es->s_kbytes_written);
2340 /* Set defaults before we parse the mount options */
2341 def_mount_opts = le32_to_cpu(es->s_default_mount_opts);
2342 if (def_mount_opts & EXT4_DEFM_DEBUG)
2343 set_opt(sbi->s_mount_opt, DEBUG);
2344 if (def_mount_opts & EXT4_DEFM_BSDGROUPS)
2345 set_opt(sbi->s_mount_opt, GRPID);
2346 if (def_mount_opts & EXT4_DEFM_UID16)
2347 set_opt(sbi->s_mount_opt, NO_UID32);
2348 #ifdef CONFIG_EXT4_FS_XATTR
2349 if (def_mount_opts & EXT4_DEFM_XATTR_USER)
2350 set_opt(sbi->s_mount_opt, XATTR_USER);
2351 #endif
2352 #ifdef CONFIG_EXT4_FS_POSIX_ACL
2353 if (def_mount_opts & EXT4_DEFM_ACL)
2354 set_opt(sbi->s_mount_opt, POSIX_ACL);
2355 #endif
2356 if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_DATA)
2357 sbi->s_mount_opt |= EXT4_MOUNT_JOURNAL_DATA;
2358 else if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_ORDERED)
2359 sbi->s_mount_opt |= EXT4_MOUNT_ORDERED_DATA;
2360 else if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_WBACK)
2361 sbi->s_mount_opt |= EXT4_MOUNT_WRITEBACK_DATA;
2363 if (le16_to_cpu(sbi->s_es->s_errors) == EXT4_ERRORS_PANIC)
2364 set_opt(sbi->s_mount_opt, ERRORS_PANIC);
2365 else if (le16_to_cpu(sbi->s_es->s_errors) == EXT4_ERRORS_CONTINUE)
2366 set_opt(sbi->s_mount_opt, ERRORS_CONT);
2367 else
2368 set_opt(sbi->s_mount_opt, ERRORS_RO);
2370 sbi->s_resuid = le16_to_cpu(es->s_def_resuid);
2371 sbi->s_resgid = le16_to_cpu(es->s_def_resgid);
2372 sbi->s_commit_interval = JBD2_DEFAULT_MAX_COMMIT_AGE * HZ;
2373 sbi->s_min_batch_time = EXT4_DEF_MIN_BATCH_TIME;
2374 sbi->s_max_batch_time = EXT4_DEF_MAX_BATCH_TIME;
2375 sbi->s_mb_history_max = default_mb_history_length;
2377 set_opt(sbi->s_mount_opt, BARRIER);
2380 * enable delayed allocation by default
2381 * Use -o nodelalloc to turn it off
2383 set_opt(sbi->s_mount_opt, DELALLOC);
2385 if (!parse_options((char *) data, sb, &journal_devnum,
2386 &journal_ioprio, NULL, 0))
2387 goto failed_mount;
2389 sb->s_flags = (sb->s_flags & ~MS_POSIXACL) |
2390 ((sbi->s_mount_opt & EXT4_MOUNT_POSIX_ACL) ? MS_POSIXACL : 0);
2392 if (le32_to_cpu(es->s_rev_level) == EXT4_GOOD_OLD_REV &&
2393 (EXT4_HAS_COMPAT_FEATURE(sb, ~0U) ||
2394 EXT4_HAS_RO_COMPAT_FEATURE(sb, ~0U) ||
2395 EXT4_HAS_INCOMPAT_FEATURE(sb, ~0U)))
2396 printk(KERN_WARNING
2397 "EXT4-fs warning: feature flags set on rev 0 fs, "
2398 "running e2fsck is recommended\n");
2401 * Check feature flags regardless of the revision level, since we
2402 * previously didn't change the revision level when setting the flags,
2403 * so there is a chance incompat flags are set on a rev 0 filesystem.
2405 features = EXT4_HAS_INCOMPAT_FEATURE(sb, ~EXT4_FEATURE_INCOMPAT_SUPP);
2406 if (features) {
2407 printk(KERN_ERR "EXT4-fs: %s: couldn't mount because of "
2408 "unsupported optional features (%x).\n", sb->s_id,
2409 (le32_to_cpu(EXT4_SB(sb)->s_es->s_feature_incompat) &
2410 ~EXT4_FEATURE_INCOMPAT_SUPP));
2411 goto failed_mount;
2413 features = EXT4_HAS_RO_COMPAT_FEATURE(sb, ~EXT4_FEATURE_RO_COMPAT_SUPP);
2414 if (!(sb->s_flags & MS_RDONLY) && features) {
2415 printk(KERN_ERR "EXT4-fs: %s: couldn't mount RDWR because of "
2416 "unsupported optional features (%x).\n", sb->s_id,
2417 (le32_to_cpu(EXT4_SB(sb)->s_es->s_feature_ro_compat) &
2418 ~EXT4_FEATURE_RO_COMPAT_SUPP));
2419 goto failed_mount;
2421 has_huge_files = EXT4_HAS_RO_COMPAT_FEATURE(sb,
2422 EXT4_FEATURE_RO_COMPAT_HUGE_FILE);
2423 if (has_huge_files) {
2425 * Large file size enabled file system can only be
2426 * mount if kernel is build with CONFIG_LBD
2428 if (sizeof(root->i_blocks) < sizeof(u64) &&
2429 !(sb->s_flags & MS_RDONLY)) {
2430 printk(KERN_ERR "EXT4-fs: %s: Filesystem with huge "
2431 "files cannot be mounted read-write "
2432 "without CONFIG_LBD.\n", sb->s_id);
2433 goto failed_mount;
2436 blocksize = BLOCK_SIZE << le32_to_cpu(es->s_log_block_size);
2438 if (blocksize < EXT4_MIN_BLOCK_SIZE ||
2439 blocksize > EXT4_MAX_BLOCK_SIZE) {
2440 printk(KERN_ERR
2441 "EXT4-fs: Unsupported filesystem blocksize %d on %s.\n",
2442 blocksize, sb->s_id);
2443 goto failed_mount;
2446 if (sb->s_blocksize != blocksize) {
2447 /* Validate the filesystem blocksize */
2448 if (!sb_set_blocksize(sb, blocksize)) {
2449 printk(KERN_ERR "EXT4-fs: bad block size %d.\n",
2450 blocksize);
2451 goto failed_mount;
2454 brelse(bh);
2455 logical_sb_block = sb_block * EXT4_MIN_BLOCK_SIZE;
2456 offset = do_div(logical_sb_block, blocksize);
2457 bh = sb_bread(sb, logical_sb_block);
2458 if (!bh) {
2459 printk(KERN_ERR
2460 "EXT4-fs: Can't read superblock on 2nd try.\n");
2461 goto failed_mount;
2463 es = (struct ext4_super_block *)(((char *)bh->b_data) + offset);
2464 sbi->s_es = es;
2465 if (es->s_magic != cpu_to_le16(EXT4_SUPER_MAGIC)) {
2466 printk(KERN_ERR
2467 "EXT4-fs: Magic mismatch, very weird !\n");
2468 goto failed_mount;
2472 sbi->s_bitmap_maxbytes = ext4_max_bitmap_size(sb->s_blocksize_bits,
2473 has_huge_files);
2474 sb->s_maxbytes = ext4_max_size(sb->s_blocksize_bits, has_huge_files);
2476 if (le32_to_cpu(es->s_rev_level) == EXT4_GOOD_OLD_REV) {
2477 sbi->s_inode_size = EXT4_GOOD_OLD_INODE_SIZE;
2478 sbi->s_first_ino = EXT4_GOOD_OLD_FIRST_INO;
2479 } else {
2480 sbi->s_inode_size = le16_to_cpu(es->s_inode_size);
2481 sbi->s_first_ino = le32_to_cpu(es->s_first_ino);
2482 if ((sbi->s_inode_size < EXT4_GOOD_OLD_INODE_SIZE) ||
2483 (!is_power_of_2(sbi->s_inode_size)) ||
2484 (sbi->s_inode_size > blocksize)) {
2485 printk(KERN_ERR
2486 "EXT4-fs: unsupported inode size: %d\n",
2487 sbi->s_inode_size);
2488 goto failed_mount;
2490 if (sbi->s_inode_size > EXT4_GOOD_OLD_INODE_SIZE)
2491 sb->s_time_gran = 1 << (EXT4_EPOCH_BITS - 2);
2494 sbi->s_desc_size = le16_to_cpu(es->s_desc_size);
2495 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_64BIT)) {
2496 if (sbi->s_desc_size < EXT4_MIN_DESC_SIZE_64BIT ||
2497 sbi->s_desc_size > EXT4_MAX_DESC_SIZE ||
2498 !is_power_of_2(sbi->s_desc_size)) {
2499 printk(KERN_ERR
2500 "EXT4-fs: unsupported descriptor size %lu\n",
2501 sbi->s_desc_size);
2502 goto failed_mount;
2504 } else
2505 sbi->s_desc_size = EXT4_MIN_DESC_SIZE;
2507 sbi->s_blocks_per_group = le32_to_cpu(es->s_blocks_per_group);
2508 sbi->s_inodes_per_group = le32_to_cpu(es->s_inodes_per_group);
2509 if (EXT4_INODE_SIZE(sb) == 0 || EXT4_INODES_PER_GROUP(sb) == 0)
2510 goto cantfind_ext4;
2512 sbi->s_inodes_per_block = blocksize / EXT4_INODE_SIZE(sb);
2513 if (sbi->s_inodes_per_block == 0)
2514 goto cantfind_ext4;
2515 sbi->s_itb_per_group = sbi->s_inodes_per_group /
2516 sbi->s_inodes_per_block;
2517 sbi->s_desc_per_block = blocksize / EXT4_DESC_SIZE(sb);
2518 sbi->s_sbh = bh;
2519 sbi->s_mount_state = le16_to_cpu(es->s_state);
2520 sbi->s_addr_per_block_bits = ilog2(EXT4_ADDR_PER_BLOCK(sb));
2521 sbi->s_desc_per_block_bits = ilog2(EXT4_DESC_PER_BLOCK(sb));
2523 for (i = 0; i < 4; i++)
2524 sbi->s_hash_seed[i] = le32_to_cpu(es->s_hash_seed[i]);
2525 sbi->s_def_hash_version = es->s_def_hash_version;
2526 i = le32_to_cpu(es->s_flags);
2527 if (i & EXT2_FLAGS_UNSIGNED_HASH)
2528 sbi->s_hash_unsigned = 3;
2529 else if ((i & EXT2_FLAGS_SIGNED_HASH) == 0) {
2530 #ifdef __CHAR_UNSIGNED__
2531 es->s_flags |= cpu_to_le32(EXT2_FLAGS_UNSIGNED_HASH);
2532 sbi->s_hash_unsigned = 3;
2533 #else
2534 es->s_flags |= cpu_to_le32(EXT2_FLAGS_SIGNED_HASH);
2535 #endif
2536 sb->s_dirt = 1;
2539 if (sbi->s_blocks_per_group > blocksize * 8) {
2540 printk(KERN_ERR
2541 "EXT4-fs: #blocks per group too big: %lu\n",
2542 sbi->s_blocks_per_group);
2543 goto failed_mount;
2545 if (sbi->s_inodes_per_group > blocksize * 8) {
2546 printk(KERN_ERR
2547 "EXT4-fs: #inodes per group too big: %lu\n",
2548 sbi->s_inodes_per_group);
2549 goto failed_mount;
2552 if (ext4_blocks_count(es) >
2553 (sector_t)(~0ULL) >> (sb->s_blocksize_bits - 9)) {
2554 printk(KERN_ERR "EXT4-fs: filesystem on %s:"
2555 " too large to mount safely\n", sb->s_id);
2556 if (sizeof(sector_t) < 8)
2557 printk(KERN_WARNING "EXT4-fs: CONFIG_LBD not "
2558 "enabled\n");
2559 goto failed_mount;
2562 if (EXT4_BLOCKS_PER_GROUP(sb) == 0)
2563 goto cantfind_ext4;
2565 /* check blocks count against device size */
2566 blocks_count = sb->s_bdev->bd_inode->i_size >> sb->s_blocksize_bits;
2567 if (blocks_count && ext4_blocks_count(es) > blocks_count) {
2568 printk(KERN_WARNING "EXT4-fs: bad geometry: block count %llu "
2569 "exceeds size of device (%llu blocks)\n",
2570 ext4_blocks_count(es), blocks_count);
2571 goto failed_mount;
2575 * It makes no sense for the first data block to be beyond the end
2576 * of the filesystem.
2578 if (le32_to_cpu(es->s_first_data_block) >= ext4_blocks_count(es)) {
2579 printk(KERN_WARNING "EXT4-fs: bad geometry: first data"
2580 "block %u is beyond end of filesystem (%llu)\n",
2581 le32_to_cpu(es->s_first_data_block),
2582 ext4_blocks_count(es));
2583 goto failed_mount;
2585 blocks_count = (ext4_blocks_count(es) -
2586 le32_to_cpu(es->s_first_data_block) +
2587 EXT4_BLOCKS_PER_GROUP(sb) - 1);
2588 do_div(blocks_count, EXT4_BLOCKS_PER_GROUP(sb));
2589 if (blocks_count > ((uint64_t)1<<32) - EXT4_DESC_PER_BLOCK(sb)) {
2590 printk(KERN_WARNING "EXT4-fs: groups count too large: %u "
2591 "(block count %llu, first data block %u, "
2592 "blocks per group %lu)\n", sbi->s_groups_count,
2593 ext4_blocks_count(es),
2594 le32_to_cpu(es->s_first_data_block),
2595 EXT4_BLOCKS_PER_GROUP(sb));
2596 goto failed_mount;
2598 sbi->s_groups_count = blocks_count;
2599 db_count = (sbi->s_groups_count + EXT4_DESC_PER_BLOCK(sb) - 1) /
2600 EXT4_DESC_PER_BLOCK(sb);
2601 sbi->s_group_desc = kmalloc(db_count * sizeof(struct buffer_head *),
2602 GFP_KERNEL);
2603 if (sbi->s_group_desc == NULL) {
2604 printk(KERN_ERR "EXT4-fs: not enough memory\n");
2605 goto failed_mount;
2608 #ifdef CONFIG_PROC_FS
2609 if (ext4_proc_root)
2610 sbi->s_proc = proc_mkdir(sb->s_id, ext4_proc_root);
2611 #endif
2613 bgl_lock_init(sbi->s_blockgroup_lock);
2615 for (i = 0; i < db_count; i++) {
2616 block = descriptor_loc(sb, logical_sb_block, i);
2617 sbi->s_group_desc[i] = sb_bread(sb, block);
2618 if (!sbi->s_group_desc[i]) {
2619 printk(KERN_ERR "EXT4-fs: "
2620 "can't read group descriptor %d\n", i);
2621 db_count = i;
2622 goto failed_mount2;
2625 if (!ext4_check_descriptors(sb)) {
2626 printk(KERN_ERR "EXT4-fs: group descriptors corrupted!\n");
2627 goto failed_mount2;
2629 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_FLEX_BG))
2630 if (!ext4_fill_flex_info(sb)) {
2631 printk(KERN_ERR
2632 "EXT4-fs: unable to initialize "
2633 "flex_bg meta info!\n");
2634 goto failed_mount2;
2637 sbi->s_gdb_count = db_count;
2638 get_random_bytes(&sbi->s_next_generation, sizeof(u32));
2639 spin_lock_init(&sbi->s_next_gen_lock);
2641 err = percpu_counter_init(&sbi->s_freeblocks_counter,
2642 ext4_count_free_blocks(sb));
2643 if (!err) {
2644 err = percpu_counter_init(&sbi->s_freeinodes_counter,
2645 ext4_count_free_inodes(sb));
2647 if (!err) {
2648 err = percpu_counter_init(&sbi->s_dirs_counter,
2649 ext4_count_dirs(sb));
2651 if (!err) {
2652 err = percpu_counter_init(&sbi->s_dirtyblocks_counter, 0);
2654 if (err) {
2655 printk(KERN_ERR "EXT4-fs: insufficient memory\n");
2656 goto failed_mount3;
2659 sbi->s_stripe = ext4_get_stripe_size(sbi);
2662 * set up enough so that it can read an inode
2664 if (!test_opt(sb, NOLOAD) &&
2665 EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL))
2666 sb->s_op = &ext4_sops;
2667 else
2668 sb->s_op = &ext4_nojournal_sops;
2669 sb->s_export_op = &ext4_export_ops;
2670 sb->s_xattr = ext4_xattr_handlers;
2671 #ifdef CONFIG_QUOTA
2672 sb->s_qcop = &ext4_qctl_operations;
2673 sb->dq_op = &ext4_quota_operations;
2674 #endif
2675 INIT_LIST_HEAD(&sbi->s_orphan); /* unlinked but open files */
2676 mutex_init(&sbi->s_orphan_lock);
2677 mutex_init(&sbi->s_resize_lock);
2679 sb->s_root = NULL;
2681 needs_recovery = (es->s_last_orphan != 0 ||
2682 EXT4_HAS_INCOMPAT_FEATURE(sb,
2683 EXT4_FEATURE_INCOMPAT_RECOVER));
2686 * The first inode we look at is the journal inode. Don't try
2687 * root first: it may be modified in the journal!
2689 if (!test_opt(sb, NOLOAD) &&
2690 EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL)) {
2691 if (ext4_load_journal(sb, es, journal_devnum))
2692 goto failed_mount3;
2693 if (!(sb->s_flags & MS_RDONLY) &&
2694 EXT4_SB(sb)->s_journal->j_failed_commit) {
2695 printk(KERN_CRIT "EXT4-fs error (device %s): "
2696 "ext4_fill_super: Journal transaction "
2697 "%u is corrupt\n", sb->s_id,
2698 EXT4_SB(sb)->s_journal->j_failed_commit);
2699 if (test_opt(sb, ERRORS_RO)) {
2700 printk(KERN_CRIT
2701 "Mounting filesystem read-only\n");
2702 sb->s_flags |= MS_RDONLY;
2703 EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS;
2704 es->s_state |= cpu_to_le16(EXT4_ERROR_FS);
2706 if (test_opt(sb, ERRORS_PANIC)) {
2707 EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS;
2708 es->s_state |= cpu_to_le16(EXT4_ERROR_FS);
2709 ext4_commit_super(sb, 1);
2710 goto failed_mount4;
2713 } else if (test_opt(sb, NOLOAD) && !(sb->s_flags & MS_RDONLY) &&
2714 EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER)) {
2715 printk(KERN_ERR "EXT4-fs: required journal recovery "
2716 "suppressed and not mounted read-only\n");
2717 goto failed_mount4;
2718 } else {
2719 clear_opt(sbi->s_mount_opt, DATA_FLAGS);
2720 set_opt(sbi->s_mount_opt, WRITEBACK_DATA);
2721 sbi->s_journal = NULL;
2722 needs_recovery = 0;
2723 goto no_journal;
2726 if (ext4_blocks_count(es) > 0xffffffffULL &&
2727 !jbd2_journal_set_features(EXT4_SB(sb)->s_journal, 0, 0,
2728 JBD2_FEATURE_INCOMPAT_64BIT)) {
2729 printk(KERN_ERR "EXT4-fs: Failed to set 64-bit journal feature\n");
2730 goto failed_mount4;
2733 if (test_opt(sb, JOURNAL_ASYNC_COMMIT)) {
2734 jbd2_journal_set_features(sbi->s_journal,
2735 JBD2_FEATURE_COMPAT_CHECKSUM, 0,
2736 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT);
2737 } else if (test_opt(sb, JOURNAL_CHECKSUM)) {
2738 jbd2_journal_set_features(sbi->s_journal,
2739 JBD2_FEATURE_COMPAT_CHECKSUM, 0, 0);
2740 jbd2_journal_clear_features(sbi->s_journal, 0, 0,
2741 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT);
2742 } else {
2743 jbd2_journal_clear_features(sbi->s_journal,
2744 JBD2_FEATURE_COMPAT_CHECKSUM, 0,
2745 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT);
2748 /* We have now updated the journal if required, so we can
2749 * validate the data journaling mode. */
2750 switch (test_opt(sb, DATA_FLAGS)) {
2751 case 0:
2752 /* No mode set, assume a default based on the journal
2753 * capabilities: ORDERED_DATA if the journal can
2754 * cope, else JOURNAL_DATA
2756 if (jbd2_journal_check_available_features
2757 (sbi->s_journal, 0, 0, JBD2_FEATURE_INCOMPAT_REVOKE))
2758 set_opt(sbi->s_mount_opt, ORDERED_DATA);
2759 else
2760 set_opt(sbi->s_mount_opt, JOURNAL_DATA);
2761 break;
2763 case EXT4_MOUNT_ORDERED_DATA:
2764 case EXT4_MOUNT_WRITEBACK_DATA:
2765 if (!jbd2_journal_check_available_features
2766 (sbi->s_journal, 0, 0, JBD2_FEATURE_INCOMPAT_REVOKE)) {
2767 printk(KERN_ERR "EXT4-fs: Journal does not support "
2768 "requested data journaling mode\n");
2769 goto failed_mount4;
2771 default:
2772 break;
2774 set_task_ioprio(sbi->s_journal->j_task, journal_ioprio);
2776 no_journal:
2778 if (test_opt(sb, NOBH)) {
2779 if (!(test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_WRITEBACK_DATA)) {
2780 printk(KERN_WARNING "EXT4-fs: Ignoring nobh option - "
2781 "its supported only with writeback mode\n");
2782 clear_opt(sbi->s_mount_opt, NOBH);
2786 * The jbd2_journal_load will have done any necessary log recovery,
2787 * so we can safely mount the rest of the filesystem now.
2790 root = ext4_iget(sb, EXT4_ROOT_INO);
2791 if (IS_ERR(root)) {
2792 printk(KERN_ERR "EXT4-fs: get root inode failed\n");
2793 ret = PTR_ERR(root);
2794 goto failed_mount4;
2796 if (!S_ISDIR(root->i_mode) || !root->i_blocks || !root->i_size) {
2797 iput(root);
2798 printk(KERN_ERR "EXT4-fs: corrupt root inode, run e2fsck\n");
2799 goto failed_mount4;
2801 sb->s_root = d_alloc_root(root);
2802 if (!sb->s_root) {
2803 printk(KERN_ERR "EXT4-fs: get root dentry failed\n");
2804 iput(root);
2805 ret = -ENOMEM;
2806 goto failed_mount4;
2809 ext4_setup_super(sb, es, sb->s_flags & MS_RDONLY);
2811 /* determine the minimum size of new large inodes, if present */
2812 if (sbi->s_inode_size > EXT4_GOOD_OLD_INODE_SIZE) {
2813 sbi->s_want_extra_isize = sizeof(struct ext4_inode) -
2814 EXT4_GOOD_OLD_INODE_SIZE;
2815 if (EXT4_HAS_RO_COMPAT_FEATURE(sb,
2816 EXT4_FEATURE_RO_COMPAT_EXTRA_ISIZE)) {
2817 if (sbi->s_want_extra_isize <
2818 le16_to_cpu(es->s_want_extra_isize))
2819 sbi->s_want_extra_isize =
2820 le16_to_cpu(es->s_want_extra_isize);
2821 if (sbi->s_want_extra_isize <
2822 le16_to_cpu(es->s_min_extra_isize))
2823 sbi->s_want_extra_isize =
2824 le16_to_cpu(es->s_min_extra_isize);
2827 /* Check if enough inode space is available */
2828 if (EXT4_GOOD_OLD_INODE_SIZE + sbi->s_want_extra_isize >
2829 sbi->s_inode_size) {
2830 sbi->s_want_extra_isize = sizeof(struct ext4_inode) -
2831 EXT4_GOOD_OLD_INODE_SIZE;
2832 printk(KERN_INFO "EXT4-fs: required extra inode space not"
2833 "available.\n");
2836 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA) {
2837 printk(KERN_WARNING "EXT4-fs: Ignoring delalloc option - "
2838 "requested data journaling mode\n");
2839 clear_opt(sbi->s_mount_opt, DELALLOC);
2840 } else if (test_opt(sb, DELALLOC))
2841 printk(KERN_INFO "EXT4-fs: delayed allocation enabled\n");
2843 err = ext4_setup_system_zone(sb);
2844 if (err) {
2845 printk(KERN_ERR "EXT4-fs: failed to initialize system "
2846 "zone (%d)\n", err);
2847 goto failed_mount4;
2850 ext4_ext_init(sb);
2851 err = ext4_mb_init(sb, needs_recovery);
2852 if (err) {
2853 printk(KERN_ERR "EXT4-fs: failed to initalize mballoc (%d)\n",
2854 err);
2855 goto failed_mount4;
2858 sbi->s_kobj.kset = ext4_kset;
2859 init_completion(&sbi->s_kobj_unregister);
2860 err = kobject_init_and_add(&sbi->s_kobj, &ext4_ktype, NULL,
2861 "%s", sb->s_id);
2862 if (err) {
2863 ext4_mb_release(sb);
2864 ext4_ext_release(sb);
2865 goto failed_mount4;
2868 EXT4_SB(sb)->s_mount_state |= EXT4_ORPHAN_FS;
2869 ext4_orphan_cleanup(sb, es);
2870 EXT4_SB(sb)->s_mount_state &= ~EXT4_ORPHAN_FS;
2871 if (needs_recovery) {
2872 printk(KERN_INFO "EXT4-fs: recovery complete.\n");
2873 ext4_mark_recovery_complete(sb, es);
2875 if (EXT4_SB(sb)->s_journal) {
2876 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA)
2877 descr = " journalled data mode";
2878 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA)
2879 descr = " ordered data mode";
2880 else
2881 descr = " writeback data mode";
2882 } else
2883 descr = "out journal";
2885 printk(KERN_INFO "EXT4-fs: mounted filesystem %s with%s\n",
2886 sb->s_id, descr);
2888 lock_kernel();
2889 return 0;
2891 cantfind_ext4:
2892 if (!silent)
2893 printk(KERN_ERR "VFS: Can't find ext4 filesystem on dev %s.\n",
2894 sb->s_id);
2895 goto failed_mount;
2897 failed_mount4:
2898 printk(KERN_ERR "EXT4-fs (device %s): mount failed\n", sb->s_id);
2899 ext4_release_system_zone(sb);
2900 if (sbi->s_journal) {
2901 jbd2_journal_destroy(sbi->s_journal);
2902 sbi->s_journal = NULL;
2904 failed_mount3:
2905 if (sbi->s_flex_groups) {
2906 if (is_vmalloc_addr(sbi->s_flex_groups))
2907 vfree(sbi->s_flex_groups);
2908 else
2909 kfree(sbi->s_flex_groups);
2911 percpu_counter_destroy(&sbi->s_freeblocks_counter);
2912 percpu_counter_destroy(&sbi->s_freeinodes_counter);
2913 percpu_counter_destroy(&sbi->s_dirs_counter);
2914 percpu_counter_destroy(&sbi->s_dirtyblocks_counter);
2915 failed_mount2:
2916 for (i = 0; i < db_count; i++)
2917 brelse(sbi->s_group_desc[i]);
2918 kfree(sbi->s_group_desc);
2919 failed_mount:
2920 if (sbi->s_proc) {
2921 remove_proc_entry(sb->s_id, ext4_proc_root);
2923 #ifdef CONFIG_QUOTA
2924 for (i = 0; i < MAXQUOTAS; i++)
2925 kfree(sbi->s_qf_names[i]);
2926 #endif
2927 ext4_blkdev_remove(sbi);
2928 brelse(bh);
2929 out_fail:
2930 sb->s_fs_info = NULL;
2931 kfree(sbi->s_blockgroup_lock);
2932 kfree(sbi);
2933 lock_kernel();
2934 return ret;
2938 * Setup any per-fs journal parameters now. We'll do this both on
2939 * initial mount, once the journal has been initialised but before we've
2940 * done any recovery; and again on any subsequent remount.
2942 static void ext4_init_journal_params(struct super_block *sb, journal_t *journal)
2944 struct ext4_sb_info *sbi = EXT4_SB(sb);
2946 journal->j_commit_interval = sbi->s_commit_interval;
2947 journal->j_min_batch_time = sbi->s_min_batch_time;
2948 journal->j_max_batch_time = sbi->s_max_batch_time;
2950 spin_lock(&journal->j_state_lock);
2951 if (test_opt(sb, BARRIER))
2952 journal->j_flags |= JBD2_BARRIER;
2953 else
2954 journal->j_flags &= ~JBD2_BARRIER;
2955 if (test_opt(sb, DATA_ERR_ABORT))
2956 journal->j_flags |= JBD2_ABORT_ON_SYNCDATA_ERR;
2957 else
2958 journal->j_flags &= ~JBD2_ABORT_ON_SYNCDATA_ERR;
2959 spin_unlock(&journal->j_state_lock);
2962 static journal_t *ext4_get_journal(struct super_block *sb,
2963 unsigned int journal_inum)
2965 struct inode *journal_inode;
2966 journal_t *journal;
2968 BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL));
2970 /* First, test for the existence of a valid inode on disk. Bad
2971 * things happen if we iget() an unused inode, as the subsequent
2972 * iput() will try to delete it. */
2974 journal_inode = ext4_iget(sb, journal_inum);
2975 if (IS_ERR(journal_inode)) {
2976 printk(KERN_ERR "EXT4-fs: no journal found.\n");
2977 return NULL;
2979 if (!journal_inode->i_nlink) {
2980 make_bad_inode(journal_inode);
2981 iput(journal_inode);
2982 printk(KERN_ERR "EXT4-fs: journal inode is deleted.\n");
2983 return NULL;
2986 jbd_debug(2, "Journal inode found at %p: %lld bytes\n",
2987 journal_inode, journal_inode->i_size);
2988 if (!S_ISREG(journal_inode->i_mode)) {
2989 printk(KERN_ERR "EXT4-fs: invalid journal inode.\n");
2990 iput(journal_inode);
2991 return NULL;
2994 journal = jbd2_journal_init_inode(journal_inode);
2995 if (!journal) {
2996 printk(KERN_ERR "EXT4-fs: Could not load journal inode\n");
2997 iput(journal_inode);
2998 return NULL;
3000 journal->j_private = sb;
3001 ext4_init_journal_params(sb, journal);
3002 return journal;
3005 static journal_t *ext4_get_dev_journal(struct super_block *sb,
3006 dev_t j_dev)
3008 struct buffer_head *bh;
3009 journal_t *journal;
3010 ext4_fsblk_t start;
3011 ext4_fsblk_t len;
3012 int hblock, blocksize;
3013 ext4_fsblk_t sb_block;
3014 unsigned long offset;
3015 struct ext4_super_block *es;
3016 struct block_device *bdev;
3018 BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL));
3020 bdev = ext4_blkdev_get(j_dev);
3021 if (bdev == NULL)
3022 return NULL;
3024 if (bd_claim(bdev, sb)) {
3025 printk(KERN_ERR
3026 "EXT4-fs: failed to claim external journal device.\n");
3027 blkdev_put(bdev, FMODE_READ|FMODE_WRITE);
3028 return NULL;
3031 blocksize = sb->s_blocksize;
3032 hblock = bdev_hardsect_size(bdev);
3033 if (blocksize < hblock) {
3034 printk(KERN_ERR
3035 "EXT4-fs: blocksize too small for journal device.\n");
3036 goto out_bdev;
3039 sb_block = EXT4_MIN_BLOCK_SIZE / blocksize;
3040 offset = EXT4_MIN_BLOCK_SIZE % blocksize;
3041 set_blocksize(bdev, blocksize);
3042 if (!(bh = __bread(bdev, sb_block, blocksize))) {
3043 printk(KERN_ERR "EXT4-fs: couldn't read superblock of "
3044 "external journal\n");
3045 goto out_bdev;
3048 es = (struct ext4_super_block *) (((char *)bh->b_data) + offset);
3049 if ((le16_to_cpu(es->s_magic) != EXT4_SUPER_MAGIC) ||
3050 !(le32_to_cpu(es->s_feature_incompat) &
3051 EXT4_FEATURE_INCOMPAT_JOURNAL_DEV)) {
3052 printk(KERN_ERR "EXT4-fs: external journal has "
3053 "bad superblock\n");
3054 brelse(bh);
3055 goto out_bdev;
3058 if (memcmp(EXT4_SB(sb)->s_es->s_journal_uuid, es->s_uuid, 16)) {
3059 printk(KERN_ERR "EXT4-fs: journal UUID does not match\n");
3060 brelse(bh);
3061 goto out_bdev;
3064 len = ext4_blocks_count(es);
3065 start = sb_block + 1;
3066 brelse(bh); /* we're done with the superblock */
3068 journal = jbd2_journal_init_dev(bdev, sb->s_bdev,
3069 start, len, blocksize);
3070 if (!journal) {
3071 printk(KERN_ERR "EXT4-fs: failed to create device journal\n");
3072 goto out_bdev;
3074 journal->j_private = sb;
3075 ll_rw_block(READ, 1, &journal->j_sb_buffer);
3076 wait_on_buffer(journal->j_sb_buffer);
3077 if (!buffer_uptodate(journal->j_sb_buffer)) {
3078 printk(KERN_ERR "EXT4-fs: I/O error on journal device\n");
3079 goto out_journal;
3081 if (be32_to_cpu(journal->j_superblock->s_nr_users) != 1) {
3082 printk(KERN_ERR "EXT4-fs: External journal has more than one "
3083 "user (unsupported) - %d\n",
3084 be32_to_cpu(journal->j_superblock->s_nr_users));
3085 goto out_journal;
3087 EXT4_SB(sb)->journal_bdev = bdev;
3088 ext4_init_journal_params(sb, journal);
3089 return journal;
3091 out_journal:
3092 jbd2_journal_destroy(journal);
3093 out_bdev:
3094 ext4_blkdev_put(bdev);
3095 return NULL;
3098 static int ext4_load_journal(struct super_block *sb,
3099 struct ext4_super_block *es,
3100 unsigned long journal_devnum)
3102 journal_t *journal;
3103 unsigned int journal_inum = le32_to_cpu(es->s_journal_inum);
3104 dev_t journal_dev;
3105 int err = 0;
3106 int really_read_only;
3108 BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL));
3110 if (journal_devnum &&
3111 journal_devnum != le32_to_cpu(es->s_journal_dev)) {
3112 printk(KERN_INFO "EXT4-fs: external journal device major/minor "
3113 "numbers have changed\n");
3114 journal_dev = new_decode_dev(journal_devnum);
3115 } else
3116 journal_dev = new_decode_dev(le32_to_cpu(es->s_journal_dev));
3118 really_read_only = bdev_read_only(sb->s_bdev);
3121 * Are we loading a blank journal or performing recovery after a
3122 * crash? For recovery, we need to check in advance whether we
3123 * can get read-write access to the device.
3125 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER)) {
3126 if (sb->s_flags & MS_RDONLY) {
3127 printk(KERN_INFO "EXT4-fs: INFO: recovery "
3128 "required on readonly filesystem.\n");
3129 if (really_read_only) {
3130 printk(KERN_ERR "EXT4-fs: write access "
3131 "unavailable, cannot proceed.\n");
3132 return -EROFS;
3134 printk(KERN_INFO "EXT4-fs: write access will "
3135 "be enabled during recovery.\n");
3139 if (journal_inum && journal_dev) {
3140 printk(KERN_ERR "EXT4-fs: filesystem has both journal "
3141 "and inode journals!\n");
3142 return -EINVAL;
3145 if (journal_inum) {
3146 if (!(journal = ext4_get_journal(sb, journal_inum)))
3147 return -EINVAL;
3148 } else {
3149 if (!(journal = ext4_get_dev_journal(sb, journal_dev)))
3150 return -EINVAL;
3153 if (journal->j_flags & JBD2_BARRIER)
3154 printk(KERN_INFO "EXT4-fs: barriers enabled\n");
3155 else
3156 printk(KERN_INFO "EXT4-fs: barriers disabled\n");
3158 if (!really_read_only && test_opt(sb, UPDATE_JOURNAL)) {
3159 err = jbd2_journal_update_format(journal);
3160 if (err) {
3161 printk(KERN_ERR "EXT4-fs: error updating journal.\n");
3162 jbd2_journal_destroy(journal);
3163 return err;
3167 if (!EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER))
3168 err = jbd2_journal_wipe(journal, !really_read_only);
3169 if (!err)
3170 err = jbd2_journal_load(journal);
3172 if (err) {
3173 printk(KERN_ERR "EXT4-fs: error loading journal.\n");
3174 jbd2_journal_destroy(journal);
3175 return err;
3178 EXT4_SB(sb)->s_journal = journal;
3179 ext4_clear_journal_err(sb, es);
3181 if (journal_devnum &&
3182 journal_devnum != le32_to_cpu(es->s_journal_dev)) {
3183 es->s_journal_dev = cpu_to_le32(journal_devnum);
3185 /* Make sure we flush the recovery flag to disk. */
3186 ext4_commit_super(sb, 1);
3189 return 0;
3192 static int ext4_commit_super(struct super_block *sb, int sync)
3194 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
3195 struct buffer_head *sbh = EXT4_SB(sb)->s_sbh;
3196 int error = 0;
3198 if (!sbh)
3199 return error;
3200 if (buffer_write_io_error(sbh)) {
3202 * Oh, dear. A previous attempt to write the
3203 * superblock failed. This could happen because the
3204 * USB device was yanked out. Or it could happen to
3205 * be a transient write error and maybe the block will
3206 * be remapped. Nothing we can do but to retry the
3207 * write and hope for the best.
3209 printk(KERN_ERR "EXT4-fs: previous I/O error to "
3210 "superblock detected for %s.\n", sb->s_id);
3211 clear_buffer_write_io_error(sbh);
3212 set_buffer_uptodate(sbh);
3214 es->s_wtime = cpu_to_le32(get_seconds());
3215 es->s_kbytes_written =
3216 cpu_to_le64(EXT4_SB(sb)->s_kbytes_written +
3217 ((part_stat_read(sb->s_bdev->bd_part, sectors[1]) -
3218 EXT4_SB(sb)->s_sectors_written_start) >> 1));
3219 ext4_free_blocks_count_set(es, percpu_counter_sum_positive(
3220 &EXT4_SB(sb)->s_freeblocks_counter));
3221 es->s_free_inodes_count = cpu_to_le32(percpu_counter_sum_positive(
3222 &EXT4_SB(sb)->s_freeinodes_counter));
3223 sb->s_dirt = 0;
3224 BUFFER_TRACE(sbh, "marking dirty");
3225 mark_buffer_dirty(sbh);
3226 if (sync) {
3227 error = sync_dirty_buffer(sbh);
3228 if (error)
3229 return error;
3231 error = buffer_write_io_error(sbh);
3232 if (error) {
3233 printk(KERN_ERR "EXT4-fs: I/O error while writing "
3234 "superblock for %s.\n", sb->s_id);
3235 clear_buffer_write_io_error(sbh);
3236 set_buffer_uptodate(sbh);
3239 return error;
3243 * Have we just finished recovery? If so, and if we are mounting (or
3244 * remounting) the filesystem readonly, then we will end up with a
3245 * consistent fs on disk. Record that fact.
3247 static void ext4_mark_recovery_complete(struct super_block *sb,
3248 struct ext4_super_block *es)
3250 journal_t *journal = EXT4_SB(sb)->s_journal;
3252 if (!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL)) {
3253 BUG_ON(journal != NULL);
3254 return;
3256 jbd2_journal_lock_updates(journal);
3257 if (jbd2_journal_flush(journal) < 0)
3258 goto out;
3260 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER) &&
3261 sb->s_flags & MS_RDONLY) {
3262 EXT4_CLEAR_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
3263 ext4_commit_super(sb, 1);
3266 out:
3267 jbd2_journal_unlock_updates(journal);
3271 * If we are mounting (or read-write remounting) a filesystem whose journal
3272 * has recorded an error from a previous lifetime, move that error to the
3273 * main filesystem now.
3275 static void ext4_clear_journal_err(struct super_block *sb,
3276 struct ext4_super_block *es)
3278 journal_t *journal;
3279 int j_errno;
3280 const char *errstr;
3282 BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL));
3284 journal = EXT4_SB(sb)->s_journal;
3287 * Now check for any error status which may have been recorded in the
3288 * journal by a prior ext4_error() or ext4_abort()
3291 j_errno = jbd2_journal_errno(journal);
3292 if (j_errno) {
3293 char nbuf[16];
3295 errstr = ext4_decode_error(sb, j_errno, nbuf);
3296 ext4_warning(sb, __func__, "Filesystem error recorded "
3297 "from previous mount: %s", errstr);
3298 ext4_warning(sb, __func__, "Marking fs in need of "
3299 "filesystem check.");
3301 EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS;
3302 es->s_state |= cpu_to_le16(EXT4_ERROR_FS);
3303 ext4_commit_super(sb, 1);
3305 jbd2_journal_clear_err(journal);
3310 * Force the running and committing transactions to commit,
3311 * and wait on the commit.
3313 int ext4_force_commit(struct super_block *sb)
3315 journal_t *journal;
3316 int ret = 0;
3318 if (sb->s_flags & MS_RDONLY)
3319 return 0;
3321 journal = EXT4_SB(sb)->s_journal;
3322 if (journal)
3323 ret = ext4_journal_force_commit(journal);
3325 return ret;
3328 static void ext4_write_super(struct super_block *sb)
3330 ext4_commit_super(sb, 1);
3333 static int ext4_sync_fs(struct super_block *sb, int wait)
3335 int ret = 0;
3336 tid_t target;
3338 trace_mark(ext4_sync_fs, "dev %s wait %d", sb->s_id, wait);
3339 if (jbd2_journal_start_commit(EXT4_SB(sb)->s_journal, &target)) {
3340 if (wait)
3341 jbd2_log_wait_commit(EXT4_SB(sb)->s_journal, target);
3343 return ret;
3347 * LVM calls this function before a (read-only) snapshot is created. This
3348 * gives us a chance to flush the journal completely and mark the fs clean.
3350 static int ext4_freeze(struct super_block *sb)
3352 int error = 0;
3353 journal_t *journal;
3355 if (sb->s_flags & MS_RDONLY)
3356 return 0;
3358 journal = EXT4_SB(sb)->s_journal;
3360 /* Now we set up the journal barrier. */
3361 jbd2_journal_lock_updates(journal);
3364 * Don't clear the needs_recovery flag if we failed to flush
3365 * the journal.
3367 error = jbd2_journal_flush(journal);
3368 if (error < 0) {
3369 out:
3370 jbd2_journal_unlock_updates(journal);
3371 return error;
3374 /* Journal blocked and flushed, clear needs_recovery flag. */
3375 EXT4_CLEAR_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
3376 error = ext4_commit_super(sb, 1);
3377 if (error)
3378 goto out;
3379 return 0;
3383 * Called by LVM after the snapshot is done. We need to reset the RECOVER
3384 * flag here, even though the filesystem is not technically dirty yet.
3386 static int ext4_unfreeze(struct super_block *sb)
3388 if (sb->s_flags & MS_RDONLY)
3389 return 0;
3391 lock_super(sb);
3392 /* Reset the needs_recovery flag before the fs is unlocked. */
3393 EXT4_SET_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
3394 ext4_commit_super(sb, 1);
3395 unlock_super(sb);
3396 jbd2_journal_unlock_updates(EXT4_SB(sb)->s_journal);
3397 return 0;
3400 static int ext4_remount(struct super_block *sb, int *flags, char *data)
3402 struct ext4_super_block *es;
3403 struct ext4_sb_info *sbi = EXT4_SB(sb);
3404 ext4_fsblk_t n_blocks_count = 0;
3405 unsigned long old_sb_flags;
3406 struct ext4_mount_options old_opts;
3407 ext4_group_t g;
3408 unsigned int journal_ioprio = DEFAULT_JOURNAL_IOPRIO;
3409 int err;
3410 #ifdef CONFIG_QUOTA
3411 int i;
3412 #endif
3414 /* Store the original options */
3415 old_sb_flags = sb->s_flags;
3416 old_opts.s_mount_opt = sbi->s_mount_opt;
3417 old_opts.s_resuid = sbi->s_resuid;
3418 old_opts.s_resgid = sbi->s_resgid;
3419 old_opts.s_commit_interval = sbi->s_commit_interval;
3420 old_opts.s_min_batch_time = sbi->s_min_batch_time;
3421 old_opts.s_max_batch_time = sbi->s_max_batch_time;
3422 #ifdef CONFIG_QUOTA
3423 old_opts.s_jquota_fmt = sbi->s_jquota_fmt;
3424 for (i = 0; i < MAXQUOTAS; i++)
3425 old_opts.s_qf_names[i] = sbi->s_qf_names[i];
3426 #endif
3427 if (sbi->s_journal && sbi->s_journal->j_task->io_context)
3428 journal_ioprio = sbi->s_journal->j_task->io_context->ioprio;
3431 * Allow the "check" option to be passed as a remount option.
3433 if (!parse_options(data, sb, NULL, &journal_ioprio,
3434 &n_blocks_count, 1)) {
3435 err = -EINVAL;
3436 goto restore_opts;
3439 if (sbi->s_mount_opt & EXT4_MOUNT_ABORT)
3440 ext4_abort(sb, __func__, "Abort forced by user");
3442 sb->s_flags = (sb->s_flags & ~MS_POSIXACL) |
3443 ((sbi->s_mount_opt & EXT4_MOUNT_POSIX_ACL) ? MS_POSIXACL : 0);
3445 es = sbi->s_es;
3447 if (sbi->s_journal) {
3448 ext4_init_journal_params(sb, sbi->s_journal);
3449 set_task_ioprio(sbi->s_journal->j_task, journal_ioprio);
3452 if ((*flags & MS_RDONLY) != (sb->s_flags & MS_RDONLY) ||
3453 n_blocks_count > ext4_blocks_count(es)) {
3454 if (sbi->s_mount_opt & EXT4_MOUNT_ABORT) {
3455 err = -EROFS;
3456 goto restore_opts;
3459 if (*flags & MS_RDONLY) {
3461 * First of all, the unconditional stuff we have to do
3462 * to disable replay of the journal when we next remount
3464 sb->s_flags |= MS_RDONLY;
3467 * OK, test if we are remounting a valid rw partition
3468 * readonly, and if so set the rdonly flag and then
3469 * mark the partition as valid again.
3471 if (!(es->s_state & cpu_to_le16(EXT4_VALID_FS)) &&
3472 (sbi->s_mount_state & EXT4_VALID_FS))
3473 es->s_state = cpu_to_le16(sbi->s_mount_state);
3475 if (sbi->s_journal)
3476 ext4_mark_recovery_complete(sb, es);
3477 } else {
3478 int ret;
3479 if ((ret = EXT4_HAS_RO_COMPAT_FEATURE(sb,
3480 ~EXT4_FEATURE_RO_COMPAT_SUPP))) {
3481 printk(KERN_WARNING "EXT4-fs: %s: couldn't "
3482 "remount RDWR because of unsupported "
3483 "optional features (%x).\n", sb->s_id,
3484 (le32_to_cpu(sbi->s_es->s_feature_ro_compat) &
3485 ~EXT4_FEATURE_RO_COMPAT_SUPP));
3486 err = -EROFS;
3487 goto restore_opts;
3491 * Make sure the group descriptor checksums
3492 * are sane. If they aren't, refuse to remount r/w.
3494 for (g = 0; g < sbi->s_groups_count; g++) {
3495 struct ext4_group_desc *gdp =
3496 ext4_get_group_desc(sb, g, NULL);
3498 if (!ext4_group_desc_csum_verify(sbi, g, gdp)) {
3499 printk(KERN_ERR
3500 "EXT4-fs: ext4_remount: "
3501 "Checksum for group %u failed (%u!=%u)\n",
3502 g, le16_to_cpu(ext4_group_desc_csum(sbi, g, gdp)),
3503 le16_to_cpu(gdp->bg_checksum));
3504 err = -EINVAL;
3505 goto restore_opts;
3510 * If we have an unprocessed orphan list hanging
3511 * around from a previously readonly bdev mount,
3512 * require a full umount/remount for now.
3514 if (es->s_last_orphan) {
3515 printk(KERN_WARNING "EXT4-fs: %s: couldn't "
3516 "remount RDWR because of unprocessed "
3517 "orphan inode list. Please "
3518 "umount/remount instead.\n",
3519 sb->s_id);
3520 err = -EINVAL;
3521 goto restore_opts;
3525 * Mounting a RDONLY partition read-write, so reread
3526 * and store the current valid flag. (It may have
3527 * been changed by e2fsck since we originally mounted
3528 * the partition.)
3530 if (sbi->s_journal)
3531 ext4_clear_journal_err(sb, es);
3532 sbi->s_mount_state = le16_to_cpu(es->s_state);
3533 if ((err = ext4_group_extend(sb, es, n_blocks_count)))
3534 goto restore_opts;
3535 if (!ext4_setup_super(sb, es, 0))
3536 sb->s_flags &= ~MS_RDONLY;
3539 ext4_setup_system_zone(sb);
3540 if (sbi->s_journal == NULL)
3541 ext4_commit_super(sb, 1);
3543 #ifdef CONFIG_QUOTA
3544 /* Release old quota file names */
3545 for (i = 0; i < MAXQUOTAS; i++)
3546 if (old_opts.s_qf_names[i] &&
3547 old_opts.s_qf_names[i] != sbi->s_qf_names[i])
3548 kfree(old_opts.s_qf_names[i]);
3549 #endif
3550 return 0;
3552 restore_opts:
3553 sb->s_flags = old_sb_flags;
3554 sbi->s_mount_opt = old_opts.s_mount_opt;
3555 sbi->s_resuid = old_opts.s_resuid;
3556 sbi->s_resgid = old_opts.s_resgid;
3557 sbi->s_commit_interval = old_opts.s_commit_interval;
3558 sbi->s_min_batch_time = old_opts.s_min_batch_time;
3559 sbi->s_max_batch_time = old_opts.s_max_batch_time;
3560 #ifdef CONFIG_QUOTA
3561 sbi->s_jquota_fmt = old_opts.s_jquota_fmt;
3562 for (i = 0; i < MAXQUOTAS; i++) {
3563 if (sbi->s_qf_names[i] &&
3564 old_opts.s_qf_names[i] != sbi->s_qf_names[i])
3565 kfree(sbi->s_qf_names[i]);
3566 sbi->s_qf_names[i] = old_opts.s_qf_names[i];
3568 #endif
3569 return err;
3572 static int ext4_statfs(struct dentry *dentry, struct kstatfs *buf)
3574 struct super_block *sb = dentry->d_sb;
3575 struct ext4_sb_info *sbi = EXT4_SB(sb);
3576 struct ext4_super_block *es = sbi->s_es;
3577 u64 fsid;
3579 if (test_opt(sb, MINIX_DF)) {
3580 sbi->s_overhead_last = 0;
3581 } else if (sbi->s_blocks_last != ext4_blocks_count(es)) {
3582 ext4_group_t i, ngroups = ext4_get_groups_count(sb);
3583 ext4_fsblk_t overhead = 0;
3586 * Compute the overhead (FS structures). This is constant
3587 * for a given filesystem unless the number of block groups
3588 * changes so we cache the previous value until it does.
3592 * All of the blocks before first_data_block are
3593 * overhead
3595 overhead = le32_to_cpu(es->s_first_data_block);
3598 * Add the overhead attributed to the superblock and
3599 * block group descriptors. If the sparse superblocks
3600 * feature is turned on, then not all groups have this.
3602 for (i = 0; i < ngroups; i++) {
3603 overhead += ext4_bg_has_super(sb, i) +
3604 ext4_bg_num_gdb(sb, i);
3605 cond_resched();
3609 * Every block group has an inode bitmap, a block
3610 * bitmap, and an inode table.
3612 overhead += ngroups * (2 + sbi->s_itb_per_group);
3613 sbi->s_overhead_last = overhead;
3614 smp_wmb();
3615 sbi->s_blocks_last = ext4_blocks_count(es);
3618 buf->f_type = EXT4_SUPER_MAGIC;
3619 buf->f_bsize = sb->s_blocksize;
3620 buf->f_blocks = ext4_blocks_count(es) - sbi->s_overhead_last;
3621 buf->f_bfree = percpu_counter_sum_positive(&sbi->s_freeblocks_counter) -
3622 percpu_counter_sum_positive(&sbi->s_dirtyblocks_counter);
3623 ext4_free_blocks_count_set(es, buf->f_bfree);
3624 buf->f_bavail = buf->f_bfree - ext4_r_blocks_count(es);
3625 if (buf->f_bfree < ext4_r_blocks_count(es))
3626 buf->f_bavail = 0;
3627 buf->f_files = le32_to_cpu(es->s_inodes_count);
3628 buf->f_ffree = percpu_counter_sum_positive(&sbi->s_freeinodes_counter);
3629 es->s_free_inodes_count = cpu_to_le32(buf->f_ffree);
3630 buf->f_namelen = EXT4_NAME_LEN;
3631 fsid = le64_to_cpup((void *)es->s_uuid) ^
3632 le64_to_cpup((void *)es->s_uuid + sizeof(u64));
3633 buf->f_fsid.val[0] = fsid & 0xFFFFFFFFUL;
3634 buf->f_fsid.val[1] = (fsid >> 32) & 0xFFFFFFFFUL;
3636 return 0;
3639 /* Helper function for writing quotas on sync - we need to start transaction
3640 * before quota file is locked for write. Otherwise the are possible deadlocks:
3641 * Process 1 Process 2
3642 * ext4_create() quota_sync()
3643 * jbd2_journal_start() write_dquot()
3644 * vfs_dq_init() down(dqio_mutex)
3645 * down(dqio_mutex) jbd2_journal_start()
3649 #ifdef CONFIG_QUOTA
3651 static inline struct inode *dquot_to_inode(struct dquot *dquot)
3653 return sb_dqopt(dquot->dq_sb)->files[dquot->dq_type];
3656 static int ext4_write_dquot(struct dquot *dquot)
3658 int ret, err;
3659 handle_t *handle;
3660 struct inode *inode;
3662 inode = dquot_to_inode(dquot);
3663 handle = ext4_journal_start(inode,
3664 EXT4_QUOTA_TRANS_BLOCKS(dquot->dq_sb));
3665 if (IS_ERR(handle))
3666 return PTR_ERR(handle);
3667 ret = dquot_commit(dquot);
3668 err = ext4_journal_stop(handle);
3669 if (!ret)
3670 ret = err;
3671 return ret;
3674 static int ext4_acquire_dquot(struct dquot *dquot)
3676 int ret, err;
3677 handle_t *handle;
3679 handle = ext4_journal_start(dquot_to_inode(dquot),
3680 EXT4_QUOTA_INIT_BLOCKS(dquot->dq_sb));
3681 if (IS_ERR(handle))
3682 return PTR_ERR(handle);
3683 ret = dquot_acquire(dquot);
3684 err = ext4_journal_stop(handle);
3685 if (!ret)
3686 ret = err;
3687 return ret;
3690 static int ext4_release_dquot(struct dquot *dquot)
3692 int ret, err;
3693 handle_t *handle;
3695 handle = ext4_journal_start(dquot_to_inode(dquot),
3696 EXT4_QUOTA_DEL_BLOCKS(dquot->dq_sb));
3697 if (IS_ERR(handle)) {
3698 /* Release dquot anyway to avoid endless cycle in dqput() */
3699 dquot_release(dquot);
3700 return PTR_ERR(handle);
3702 ret = dquot_release(dquot);
3703 err = ext4_journal_stop(handle);
3704 if (!ret)
3705 ret = err;
3706 return ret;
3709 static int ext4_mark_dquot_dirty(struct dquot *dquot)
3711 /* Are we journaling quotas? */
3712 if (EXT4_SB(dquot->dq_sb)->s_qf_names[USRQUOTA] ||
3713 EXT4_SB(dquot->dq_sb)->s_qf_names[GRPQUOTA]) {
3714 dquot_mark_dquot_dirty(dquot);
3715 return ext4_write_dquot(dquot);
3716 } else {
3717 return dquot_mark_dquot_dirty(dquot);
3721 static int ext4_write_info(struct super_block *sb, int type)
3723 int ret, err;
3724 handle_t *handle;
3726 /* Data block + inode block */
3727 handle = ext4_journal_start(sb->s_root->d_inode, 2);
3728 if (IS_ERR(handle))
3729 return PTR_ERR(handle);
3730 ret = dquot_commit_info(sb, type);
3731 err = ext4_journal_stop(handle);
3732 if (!ret)
3733 ret = err;
3734 return ret;
3738 * Turn on quotas during mount time - we need to find
3739 * the quota file and such...
3741 static int ext4_quota_on_mount(struct super_block *sb, int type)
3743 return vfs_quota_on_mount(sb, EXT4_SB(sb)->s_qf_names[type],
3744 EXT4_SB(sb)->s_jquota_fmt, type);
3748 * Standard function to be called on quota_on
3750 static int ext4_quota_on(struct super_block *sb, int type, int format_id,
3751 char *name, int remount)
3753 int err;
3754 struct path path;
3756 if (!test_opt(sb, QUOTA))
3757 return -EINVAL;
3758 /* When remounting, no checks are needed and in fact, name is NULL */
3759 if (remount)
3760 return vfs_quota_on(sb, type, format_id, name, remount);
3762 err = kern_path(name, LOOKUP_FOLLOW, &path);
3763 if (err)
3764 return err;
3766 /* Quotafile not on the same filesystem? */
3767 if (path.mnt->mnt_sb != sb) {
3768 path_put(&path);
3769 return -EXDEV;
3771 /* Journaling quota? */
3772 if (EXT4_SB(sb)->s_qf_names[type]) {
3773 /* Quotafile not in fs root? */
3774 if (path.dentry->d_parent != sb->s_root)
3775 printk(KERN_WARNING
3776 "EXT4-fs: Quota file not on filesystem root. "
3777 "Journaled quota will not work.\n");
3781 * When we journal data on quota file, we have to flush journal to see
3782 * all updates to the file when we bypass pagecache...
3784 if (EXT4_SB(sb)->s_journal &&
3785 ext4_should_journal_data(path.dentry->d_inode)) {
3787 * We don't need to lock updates but journal_flush() could
3788 * otherwise be livelocked...
3790 jbd2_journal_lock_updates(EXT4_SB(sb)->s_journal);
3791 err = jbd2_journal_flush(EXT4_SB(sb)->s_journal);
3792 jbd2_journal_unlock_updates(EXT4_SB(sb)->s_journal);
3793 if (err) {
3794 path_put(&path);
3795 return err;
3799 err = vfs_quota_on_path(sb, type, format_id, &path);
3800 path_put(&path);
3801 return err;
3804 /* Read data from quotafile - avoid pagecache and such because we cannot afford
3805 * acquiring the locks... As quota files are never truncated and quota code
3806 * itself serializes the operations (and noone else should touch the files)
3807 * we don't have to be afraid of races */
3808 static ssize_t ext4_quota_read(struct super_block *sb, int type, char *data,
3809 size_t len, loff_t off)
3811 struct inode *inode = sb_dqopt(sb)->files[type];
3812 ext4_lblk_t blk = off >> EXT4_BLOCK_SIZE_BITS(sb);
3813 int err = 0;
3814 int offset = off & (sb->s_blocksize - 1);
3815 int tocopy;
3816 size_t toread;
3817 struct buffer_head *bh;
3818 loff_t i_size = i_size_read(inode);
3820 if (off > i_size)
3821 return 0;
3822 if (off+len > i_size)
3823 len = i_size-off;
3824 toread = len;
3825 while (toread > 0) {
3826 tocopy = sb->s_blocksize - offset < toread ?
3827 sb->s_blocksize - offset : toread;
3828 bh = ext4_bread(NULL, inode, blk, 0, &err);
3829 if (err)
3830 return err;
3831 if (!bh) /* A hole? */
3832 memset(data, 0, tocopy);
3833 else
3834 memcpy(data, bh->b_data+offset, tocopy);
3835 brelse(bh);
3836 offset = 0;
3837 toread -= tocopy;
3838 data += tocopy;
3839 blk++;
3841 return len;
3844 /* Write to quotafile (we know the transaction is already started and has
3845 * enough credits) */
3846 static ssize_t ext4_quota_write(struct super_block *sb, int type,
3847 const char *data, size_t len, loff_t off)
3849 struct inode *inode = sb_dqopt(sb)->files[type];
3850 ext4_lblk_t blk = off >> EXT4_BLOCK_SIZE_BITS(sb);
3851 int err = 0;
3852 int offset = off & (sb->s_blocksize - 1);
3853 int tocopy;
3854 int journal_quota = EXT4_SB(sb)->s_qf_names[type] != NULL;
3855 size_t towrite = len;
3856 struct buffer_head *bh;
3857 handle_t *handle = journal_current_handle();
3859 if (EXT4_SB(sb)->s_journal && !handle) {
3860 printk(KERN_WARNING "EXT4-fs: Quota write (off=%llu, len=%llu)"
3861 " cancelled because transaction is not started.\n",
3862 (unsigned long long)off, (unsigned long long)len);
3863 return -EIO;
3865 mutex_lock_nested(&inode->i_mutex, I_MUTEX_QUOTA);
3866 while (towrite > 0) {
3867 tocopy = sb->s_blocksize - offset < towrite ?
3868 sb->s_blocksize - offset : towrite;
3869 bh = ext4_bread(handle, inode, blk, 1, &err);
3870 if (!bh)
3871 goto out;
3872 if (journal_quota) {
3873 err = ext4_journal_get_write_access(handle, bh);
3874 if (err) {
3875 brelse(bh);
3876 goto out;
3879 lock_buffer(bh);
3880 memcpy(bh->b_data+offset, data, tocopy);
3881 flush_dcache_page(bh->b_page);
3882 unlock_buffer(bh);
3883 if (journal_quota)
3884 err = ext4_handle_dirty_metadata(handle, NULL, bh);
3885 else {
3886 /* Always do at least ordered writes for quotas */
3887 err = ext4_jbd2_file_inode(handle, inode);
3888 mark_buffer_dirty(bh);
3890 brelse(bh);
3891 if (err)
3892 goto out;
3893 offset = 0;
3894 towrite -= tocopy;
3895 data += tocopy;
3896 blk++;
3898 out:
3899 if (len == towrite) {
3900 mutex_unlock(&inode->i_mutex);
3901 return err;
3903 if (inode->i_size < off+len-towrite) {
3904 i_size_write(inode, off+len-towrite);
3905 EXT4_I(inode)->i_disksize = inode->i_size;
3907 inode->i_mtime = inode->i_ctime = CURRENT_TIME;
3908 ext4_mark_inode_dirty(handle, inode);
3909 mutex_unlock(&inode->i_mutex);
3910 return len - towrite;
3913 #endif
3915 static int ext4_get_sb(struct file_system_type *fs_type, int flags,
3916 const char *dev_name, void *data, struct vfsmount *mnt)
3918 return get_sb_bdev(fs_type, flags, dev_name, data, ext4_fill_super,mnt);
3921 static struct file_system_type ext4_fs_type = {
3922 .owner = THIS_MODULE,
3923 .name = "ext4",
3924 .get_sb = ext4_get_sb,
3925 .kill_sb = kill_block_super,
3926 .fs_flags = FS_REQUIRES_DEV,
3929 #ifdef CONFIG_EXT4DEV_COMPAT
3930 static int ext4dev_get_sb(struct file_system_type *fs_type, int flags,
3931 const char *dev_name, void *data,struct vfsmount *mnt)
3933 printk(KERN_WARNING "EXT4-fs: Update your userspace programs "
3934 "to mount using ext4\n");
3935 printk(KERN_WARNING "EXT4-fs: ext4dev backwards compatibility "
3936 "will go away by 2.6.31\n");
3937 return get_sb_bdev(fs_type, flags, dev_name, data, ext4_fill_super,mnt);
3940 static struct file_system_type ext4dev_fs_type = {
3941 .owner = THIS_MODULE,
3942 .name = "ext4dev",
3943 .get_sb = ext4dev_get_sb,
3944 .kill_sb = kill_block_super,
3945 .fs_flags = FS_REQUIRES_DEV,
3947 MODULE_ALIAS("ext4dev");
3948 #endif
3950 static int __init init_ext4_fs(void)
3952 int err;
3954 err = init_ext4_system_zone();
3955 if (err)
3956 return err;
3957 ext4_kset = kset_create_and_add("ext4", NULL, fs_kobj);
3958 if (!ext4_kset)
3959 goto out4;
3960 ext4_proc_root = proc_mkdir("fs/ext4", NULL);
3961 err = init_ext4_mballoc();
3962 if (err)
3963 goto out3;
3965 err = init_ext4_xattr();
3966 if (err)
3967 goto out2;
3968 err = init_inodecache();
3969 if (err)
3970 goto out1;
3971 err = register_filesystem(&ext4_fs_type);
3972 if (err)
3973 goto out;
3974 #ifdef CONFIG_EXT4DEV_COMPAT
3975 err = register_filesystem(&ext4dev_fs_type);
3976 if (err) {
3977 unregister_filesystem(&ext4_fs_type);
3978 goto out;
3980 #endif
3981 return 0;
3982 out:
3983 destroy_inodecache();
3984 out1:
3985 exit_ext4_xattr();
3986 out2:
3987 exit_ext4_mballoc();
3988 out3:
3989 remove_proc_entry("fs/ext4", NULL);
3990 kset_unregister(ext4_kset);
3991 out4:
3992 exit_ext4_system_zone();
3993 return err;
3996 static void __exit exit_ext4_fs(void)
3998 unregister_filesystem(&ext4_fs_type);
3999 #ifdef CONFIG_EXT4DEV_COMPAT
4000 unregister_filesystem(&ext4dev_fs_type);
4001 #endif
4002 destroy_inodecache();
4003 exit_ext4_xattr();
4004 exit_ext4_mballoc();
4005 remove_proc_entry("fs/ext4", NULL);
4006 kset_unregister(ext4_kset);
4007 exit_ext4_system_zone();
4010 MODULE_AUTHOR("Remy Card, Stephen Tweedie, Andrew Morton, Andreas Dilger, Theodore Ts'o and others");
4011 MODULE_DESCRIPTION("Fourth Extended Filesystem");
4012 MODULE_LICENSE("GPL");
4013 module_init(init_ext4_fs)
4014 module_exit(exit_ext4_fs)