PCI ASPM: cleanup pcie_aspm_get_cap_device
[linux-2.6/linux-2.6-openrd.git] / fs / ext3 / super.c
blob599dbfe504c39ef8370dd28a4c65aa05346499b4
1 /*
2 * linux/fs/ext3/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/jbd.h>
24 #include <linux/ext3_fs.h>
25 #include <linux/ext3_jbd.h>
26 #include <linux/slab.h>
27 #include <linux/init.h>
28 #include <linux/blkdev.h>
29 #include <linux/parser.h>
30 #include <linux/smp_lock.h>
31 #include <linux/buffer_head.h>
32 #include <linux/exportfs.h>
33 #include <linux/vfs.h>
34 #include <linux/random.h>
35 #include <linux/mount.h>
36 #include <linux/namei.h>
37 #include <linux/quotaops.h>
38 #include <linux/seq_file.h>
39 #include <linux/log2.h>
41 #include <asm/uaccess.h>
43 #include "xattr.h"
44 #include "acl.h"
45 #include "namei.h"
47 #ifdef CONFIG_EXT3_DEFAULTS_TO_ORDERED
48 #define EXT3_MOUNT_DEFAULT_DATA_MODE EXT3_MOUNT_ORDERED_DATA
49 #else
50 #define EXT3_MOUNT_DEFAULT_DATA_MODE EXT3_MOUNT_WRITEBACK_DATA
51 #endif
53 static int ext3_load_journal(struct super_block *, struct ext3_super_block *,
54 unsigned long journal_devnum);
55 static int ext3_create_journal(struct super_block *, struct ext3_super_block *,
56 unsigned int);
57 static int ext3_commit_super(struct super_block *sb,
58 struct ext3_super_block *es,
59 int sync);
60 static void ext3_mark_recovery_complete(struct super_block * sb,
61 struct ext3_super_block * es);
62 static void ext3_clear_journal_err(struct super_block * sb,
63 struct ext3_super_block * es);
64 static int ext3_sync_fs(struct super_block *sb, int wait);
65 static const char *ext3_decode_error(struct super_block * sb, int errno,
66 char nbuf[16]);
67 static int ext3_remount (struct super_block * sb, int * flags, char * data);
68 static int ext3_statfs (struct dentry * dentry, struct kstatfs * buf);
69 static int ext3_unfreeze(struct super_block *sb);
70 static void ext3_write_super (struct super_block * sb);
71 static int ext3_freeze(struct super_block *sb);
74 * Wrappers for journal_start/end.
76 * The only special thing we need to do here is to make sure that all
77 * journal_end calls result in the superblock being marked dirty, so
78 * that sync() will call the filesystem's write_super callback if
79 * appropriate.
81 handle_t *ext3_journal_start_sb(struct super_block *sb, int nblocks)
83 journal_t *journal;
85 if (sb->s_flags & MS_RDONLY)
86 return ERR_PTR(-EROFS);
88 /* Special case here: if the journal has aborted behind our
89 * backs (eg. EIO in the commit thread), then we still need to
90 * take the FS itself readonly cleanly. */
91 journal = EXT3_SB(sb)->s_journal;
92 if (is_journal_aborted(journal)) {
93 ext3_abort(sb, __func__,
94 "Detected aborted journal");
95 return ERR_PTR(-EROFS);
98 return journal_start(journal, nblocks);
102 * The only special thing we need to do here is to make sure that all
103 * journal_stop calls result in the superblock being marked dirty, so
104 * that sync() will call the filesystem's write_super callback if
105 * appropriate.
107 int __ext3_journal_stop(const char *where, handle_t *handle)
109 struct super_block *sb;
110 int err;
111 int rc;
113 sb = handle->h_transaction->t_journal->j_private;
114 err = handle->h_err;
115 rc = journal_stop(handle);
117 if (!err)
118 err = rc;
119 if (err)
120 __ext3_std_error(sb, where, err);
121 return err;
124 void ext3_journal_abort_handle(const char *caller, const char *err_fn,
125 struct buffer_head *bh, handle_t *handle, int err)
127 char nbuf[16];
128 const char *errstr = ext3_decode_error(NULL, err, nbuf);
130 if (bh)
131 BUFFER_TRACE(bh, "abort");
133 if (!handle->h_err)
134 handle->h_err = err;
136 if (is_handle_aborted(handle))
137 return;
139 printk(KERN_ERR "%s: aborting transaction: %s in %s\n",
140 caller, errstr, err_fn);
142 journal_abort_handle(handle);
145 /* Deal with the reporting of failure conditions on a filesystem such as
146 * inconsistencies detected or read IO failures.
148 * On ext2, we can store the error state of the filesystem in the
149 * superblock. That is not possible on ext3, because we may have other
150 * write ordering constraints on the superblock which prevent us from
151 * writing it out straight away; and given that the journal is about to
152 * be aborted, we can't rely on the current, or future, transactions to
153 * write out the superblock safely.
155 * We'll just use the journal_abort() error code to record an error in
156 * the journal instead. On recovery, the journal will compain about
157 * that error until we've noted it down and cleared it.
160 static void ext3_handle_error(struct super_block *sb)
162 struct ext3_super_block *es = EXT3_SB(sb)->s_es;
164 EXT3_SB(sb)->s_mount_state |= EXT3_ERROR_FS;
165 es->s_state |= cpu_to_le16(EXT3_ERROR_FS);
167 if (sb->s_flags & MS_RDONLY)
168 return;
170 if (!test_opt (sb, ERRORS_CONT)) {
171 journal_t *journal = EXT3_SB(sb)->s_journal;
173 EXT3_SB(sb)->s_mount_opt |= EXT3_MOUNT_ABORT;
174 if (journal)
175 journal_abort(journal, -EIO);
177 if (test_opt (sb, ERRORS_RO)) {
178 printk (KERN_CRIT "Remounting filesystem read-only\n");
179 sb->s_flags |= MS_RDONLY;
181 ext3_commit_super(sb, es, 1);
182 if (test_opt(sb, ERRORS_PANIC))
183 panic("EXT3-fs (device %s): panic forced after error\n",
184 sb->s_id);
187 void ext3_error (struct super_block * sb, const char * function,
188 const char * fmt, ...)
190 va_list args;
192 va_start(args, fmt);
193 printk(KERN_CRIT "EXT3-fs error (device %s): %s: ",sb->s_id, function);
194 vprintk(fmt, args);
195 printk("\n");
196 va_end(args);
198 ext3_handle_error(sb);
201 static const char *ext3_decode_error(struct super_block * sb, int errno,
202 char nbuf[16])
204 char *errstr = NULL;
206 switch (errno) {
207 case -EIO:
208 errstr = "IO failure";
209 break;
210 case -ENOMEM:
211 errstr = "Out of memory";
212 break;
213 case -EROFS:
214 if (!sb || EXT3_SB(sb)->s_journal->j_flags & JFS_ABORT)
215 errstr = "Journal has aborted";
216 else
217 errstr = "Readonly filesystem";
218 break;
219 default:
220 /* If the caller passed in an extra buffer for unknown
221 * errors, textualise them now. Else we just return
222 * NULL. */
223 if (nbuf) {
224 /* Check for truncated error codes... */
225 if (snprintf(nbuf, 16, "error %d", -errno) >= 0)
226 errstr = nbuf;
228 break;
231 return errstr;
234 /* __ext3_std_error decodes expected errors from journaling functions
235 * automatically and invokes the appropriate error response. */
237 void __ext3_std_error (struct super_block * sb, const char * function,
238 int errno)
240 char nbuf[16];
241 const char *errstr;
243 /* Special case: if the error is EROFS, and we're not already
244 * inside a transaction, then there's really no point in logging
245 * an error. */
246 if (errno == -EROFS && journal_current_handle() == NULL &&
247 (sb->s_flags & MS_RDONLY))
248 return;
250 errstr = ext3_decode_error(sb, errno, nbuf);
251 printk (KERN_CRIT "EXT3-fs error (device %s) in %s: %s\n",
252 sb->s_id, function, errstr);
254 ext3_handle_error(sb);
258 * ext3_abort is a much stronger failure handler than ext3_error. The
259 * abort function may be used to deal with unrecoverable failures such
260 * as journal IO errors or ENOMEM at a critical moment in log management.
262 * We unconditionally force the filesystem into an ABORT|READONLY state,
263 * unless the error response on the fs has been set to panic in which
264 * case we take the easy way out and panic immediately.
267 void ext3_abort (struct super_block * sb, const char * function,
268 const char * fmt, ...)
270 va_list args;
272 printk (KERN_CRIT "ext3_abort called.\n");
274 va_start(args, fmt);
275 printk(KERN_CRIT "EXT3-fs error (device %s): %s: ",sb->s_id, function);
276 vprintk(fmt, args);
277 printk("\n");
278 va_end(args);
280 if (test_opt(sb, ERRORS_PANIC))
281 panic("EXT3-fs panic from previous error\n");
283 if (sb->s_flags & MS_RDONLY)
284 return;
286 printk(KERN_CRIT "Remounting filesystem read-only\n");
287 EXT3_SB(sb)->s_mount_state |= EXT3_ERROR_FS;
288 sb->s_flags |= MS_RDONLY;
289 EXT3_SB(sb)->s_mount_opt |= EXT3_MOUNT_ABORT;
290 if (EXT3_SB(sb)->s_journal)
291 journal_abort(EXT3_SB(sb)->s_journal, -EIO);
294 void ext3_warning (struct super_block * sb, const char * function,
295 const char * fmt, ...)
297 va_list args;
299 va_start(args, fmt);
300 printk(KERN_WARNING "EXT3-fs warning (device %s): %s: ",
301 sb->s_id, function);
302 vprintk(fmt, args);
303 printk("\n");
304 va_end(args);
307 void ext3_update_dynamic_rev(struct super_block *sb)
309 struct ext3_super_block *es = EXT3_SB(sb)->s_es;
311 if (le32_to_cpu(es->s_rev_level) > EXT3_GOOD_OLD_REV)
312 return;
314 ext3_warning(sb, __func__,
315 "updating to rev %d because of new feature flag, "
316 "running e2fsck is recommended",
317 EXT3_DYNAMIC_REV);
319 es->s_first_ino = cpu_to_le32(EXT3_GOOD_OLD_FIRST_INO);
320 es->s_inode_size = cpu_to_le16(EXT3_GOOD_OLD_INODE_SIZE);
321 es->s_rev_level = cpu_to_le32(EXT3_DYNAMIC_REV);
322 /* leave es->s_feature_*compat flags alone */
323 /* es->s_uuid will be set by e2fsck if empty */
326 * The rest of the superblock fields should be zero, and if not it
327 * means they are likely already in use, so leave them alone. We
328 * can leave it up to e2fsck to clean up any inconsistencies there.
333 * Open the external journal device
335 static struct block_device *ext3_blkdev_get(dev_t dev)
337 struct block_device *bdev;
338 char b[BDEVNAME_SIZE];
340 bdev = open_by_devnum(dev, FMODE_READ|FMODE_WRITE);
341 if (IS_ERR(bdev))
342 goto fail;
343 return bdev;
345 fail:
346 printk(KERN_ERR "EXT3: failed to open journal device %s: %ld\n",
347 __bdevname(dev, b), PTR_ERR(bdev));
348 return NULL;
352 * Release the journal device
354 static int ext3_blkdev_put(struct block_device *bdev)
356 bd_release(bdev);
357 return blkdev_put(bdev, FMODE_READ|FMODE_WRITE);
360 static int ext3_blkdev_remove(struct ext3_sb_info *sbi)
362 struct block_device *bdev;
363 int ret = -ENODEV;
365 bdev = sbi->journal_bdev;
366 if (bdev) {
367 ret = ext3_blkdev_put(bdev);
368 sbi->journal_bdev = NULL;
370 return ret;
373 static inline struct inode *orphan_list_entry(struct list_head *l)
375 return &list_entry(l, struct ext3_inode_info, i_orphan)->vfs_inode;
378 static void dump_orphan_list(struct super_block *sb, struct ext3_sb_info *sbi)
380 struct list_head *l;
382 printk(KERN_ERR "sb orphan head is %d\n",
383 le32_to_cpu(sbi->s_es->s_last_orphan));
385 printk(KERN_ERR "sb_info orphan list:\n");
386 list_for_each(l, &sbi->s_orphan) {
387 struct inode *inode = orphan_list_entry(l);
388 printk(KERN_ERR " "
389 "inode %s:%lu at %p: mode %o, nlink %d, next %d\n",
390 inode->i_sb->s_id, inode->i_ino, inode,
391 inode->i_mode, inode->i_nlink,
392 NEXT_ORPHAN(inode));
396 static void ext3_put_super (struct super_block * sb)
398 struct ext3_sb_info *sbi = EXT3_SB(sb);
399 struct ext3_super_block *es = sbi->s_es;
400 int i, err;
402 ext3_xattr_put_super(sb);
403 err = journal_destroy(sbi->s_journal);
404 sbi->s_journal = NULL;
405 if (err < 0)
406 ext3_abort(sb, __func__, "Couldn't clean up the journal");
408 if (!(sb->s_flags & MS_RDONLY)) {
409 EXT3_CLEAR_INCOMPAT_FEATURE(sb, EXT3_FEATURE_INCOMPAT_RECOVER);
410 es->s_state = cpu_to_le16(sbi->s_mount_state);
411 BUFFER_TRACE(sbi->s_sbh, "marking dirty");
412 mark_buffer_dirty(sbi->s_sbh);
413 ext3_commit_super(sb, es, 1);
416 for (i = 0; i < sbi->s_gdb_count; i++)
417 brelse(sbi->s_group_desc[i]);
418 kfree(sbi->s_group_desc);
419 percpu_counter_destroy(&sbi->s_freeblocks_counter);
420 percpu_counter_destroy(&sbi->s_freeinodes_counter);
421 percpu_counter_destroy(&sbi->s_dirs_counter);
422 brelse(sbi->s_sbh);
423 #ifdef CONFIG_QUOTA
424 for (i = 0; i < MAXQUOTAS; i++)
425 kfree(sbi->s_qf_names[i]);
426 #endif
428 /* Debugging code just in case the in-memory inode orphan list
429 * isn't empty. The on-disk one can be non-empty if we've
430 * detected an error and taken the fs readonly, but the
431 * in-memory list had better be clean by this point. */
432 if (!list_empty(&sbi->s_orphan))
433 dump_orphan_list(sb, sbi);
434 J_ASSERT(list_empty(&sbi->s_orphan));
436 invalidate_bdev(sb->s_bdev);
437 if (sbi->journal_bdev && sbi->journal_bdev != sb->s_bdev) {
439 * Invalidate the journal device's buffers. We don't want them
440 * floating about in memory - the physical journal device may
441 * hotswapped, and it breaks the `ro-after' testing code.
443 sync_blockdev(sbi->journal_bdev);
444 invalidate_bdev(sbi->journal_bdev);
445 ext3_blkdev_remove(sbi);
447 sb->s_fs_info = NULL;
448 kfree(sbi->s_blockgroup_lock);
449 kfree(sbi);
450 return;
453 static struct kmem_cache *ext3_inode_cachep;
456 * Called inside transaction, so use GFP_NOFS
458 static struct inode *ext3_alloc_inode(struct super_block *sb)
460 struct ext3_inode_info *ei;
462 ei = kmem_cache_alloc(ext3_inode_cachep, GFP_NOFS);
463 if (!ei)
464 return NULL;
465 #ifdef CONFIG_EXT3_FS_POSIX_ACL
466 ei->i_acl = EXT3_ACL_NOT_CACHED;
467 ei->i_default_acl = EXT3_ACL_NOT_CACHED;
468 #endif
469 ei->i_block_alloc_info = NULL;
470 ei->vfs_inode.i_version = 1;
471 return &ei->vfs_inode;
474 static void ext3_destroy_inode(struct inode *inode)
476 if (!list_empty(&(EXT3_I(inode)->i_orphan))) {
477 printk("EXT3 Inode %p: orphan list check failed!\n",
478 EXT3_I(inode));
479 print_hex_dump(KERN_INFO, "", DUMP_PREFIX_ADDRESS, 16, 4,
480 EXT3_I(inode), sizeof(struct ext3_inode_info),
481 false);
482 dump_stack();
484 kmem_cache_free(ext3_inode_cachep, EXT3_I(inode));
487 static void init_once(void *foo)
489 struct ext3_inode_info *ei = (struct ext3_inode_info *) foo;
491 INIT_LIST_HEAD(&ei->i_orphan);
492 #ifdef CONFIG_EXT3_FS_XATTR
493 init_rwsem(&ei->xattr_sem);
494 #endif
495 mutex_init(&ei->truncate_mutex);
496 inode_init_once(&ei->vfs_inode);
499 static int init_inodecache(void)
501 ext3_inode_cachep = kmem_cache_create("ext3_inode_cache",
502 sizeof(struct ext3_inode_info),
503 0, (SLAB_RECLAIM_ACCOUNT|
504 SLAB_MEM_SPREAD),
505 init_once);
506 if (ext3_inode_cachep == NULL)
507 return -ENOMEM;
508 return 0;
511 static void destroy_inodecache(void)
513 kmem_cache_destroy(ext3_inode_cachep);
516 static void ext3_clear_inode(struct inode *inode)
518 struct ext3_block_alloc_info *rsv = EXT3_I(inode)->i_block_alloc_info;
519 #ifdef CONFIG_EXT3_FS_POSIX_ACL
520 if (EXT3_I(inode)->i_acl &&
521 EXT3_I(inode)->i_acl != EXT3_ACL_NOT_CACHED) {
522 posix_acl_release(EXT3_I(inode)->i_acl);
523 EXT3_I(inode)->i_acl = EXT3_ACL_NOT_CACHED;
525 if (EXT3_I(inode)->i_default_acl &&
526 EXT3_I(inode)->i_default_acl != EXT3_ACL_NOT_CACHED) {
527 posix_acl_release(EXT3_I(inode)->i_default_acl);
528 EXT3_I(inode)->i_default_acl = EXT3_ACL_NOT_CACHED;
530 #endif
531 ext3_discard_reservation(inode);
532 EXT3_I(inode)->i_block_alloc_info = NULL;
533 if (unlikely(rsv))
534 kfree(rsv);
537 static inline void ext3_show_quota_options(struct seq_file *seq, struct super_block *sb)
539 #if defined(CONFIG_QUOTA)
540 struct ext3_sb_info *sbi = EXT3_SB(sb);
542 if (sbi->s_jquota_fmt)
543 seq_printf(seq, ",jqfmt=%s",
544 (sbi->s_jquota_fmt == QFMT_VFS_OLD) ? "vfsold": "vfsv0");
546 if (sbi->s_qf_names[USRQUOTA])
547 seq_printf(seq, ",usrjquota=%s", sbi->s_qf_names[USRQUOTA]);
549 if (sbi->s_qf_names[GRPQUOTA])
550 seq_printf(seq, ",grpjquota=%s", sbi->s_qf_names[GRPQUOTA]);
552 if (sbi->s_mount_opt & EXT3_MOUNT_USRQUOTA)
553 seq_puts(seq, ",usrquota");
555 if (sbi->s_mount_opt & EXT3_MOUNT_GRPQUOTA)
556 seq_puts(seq, ",grpquota");
557 #endif
561 * Show an option if
562 * - it's set to a non-default value OR
563 * - if the per-sb default is different from the global default
565 static int ext3_show_options(struct seq_file *seq, struct vfsmount *vfs)
567 struct super_block *sb = vfs->mnt_sb;
568 struct ext3_sb_info *sbi = EXT3_SB(sb);
569 struct ext3_super_block *es = sbi->s_es;
570 unsigned long def_mount_opts;
572 def_mount_opts = le32_to_cpu(es->s_default_mount_opts);
574 if (sbi->s_sb_block != 1)
575 seq_printf(seq, ",sb=%lu", sbi->s_sb_block);
576 if (test_opt(sb, MINIX_DF))
577 seq_puts(seq, ",minixdf");
578 if (test_opt(sb, GRPID))
579 seq_puts(seq, ",grpid");
580 if (!test_opt(sb, GRPID) && (def_mount_opts & EXT3_DEFM_BSDGROUPS))
581 seq_puts(seq, ",nogrpid");
582 if (sbi->s_resuid != EXT3_DEF_RESUID ||
583 le16_to_cpu(es->s_def_resuid) != EXT3_DEF_RESUID) {
584 seq_printf(seq, ",resuid=%u", sbi->s_resuid);
586 if (sbi->s_resgid != EXT3_DEF_RESGID ||
587 le16_to_cpu(es->s_def_resgid) != EXT3_DEF_RESGID) {
588 seq_printf(seq, ",resgid=%u", sbi->s_resgid);
590 if (test_opt(sb, ERRORS_RO)) {
591 int def_errors = le16_to_cpu(es->s_errors);
593 if (def_errors == EXT3_ERRORS_PANIC ||
594 def_errors == EXT3_ERRORS_CONTINUE) {
595 seq_puts(seq, ",errors=remount-ro");
598 if (test_opt(sb, ERRORS_CONT))
599 seq_puts(seq, ",errors=continue");
600 if (test_opt(sb, ERRORS_PANIC))
601 seq_puts(seq, ",errors=panic");
602 if (test_opt(sb, NO_UID32))
603 seq_puts(seq, ",nouid32");
604 if (test_opt(sb, DEBUG))
605 seq_puts(seq, ",debug");
606 if (test_opt(sb, OLDALLOC))
607 seq_puts(seq, ",oldalloc");
608 #ifdef CONFIG_EXT3_FS_XATTR
609 if (test_opt(sb, XATTR_USER))
610 seq_puts(seq, ",user_xattr");
611 if (!test_opt(sb, XATTR_USER) &&
612 (def_mount_opts & EXT3_DEFM_XATTR_USER)) {
613 seq_puts(seq, ",nouser_xattr");
615 #endif
616 #ifdef CONFIG_EXT3_FS_POSIX_ACL
617 if (test_opt(sb, POSIX_ACL))
618 seq_puts(seq, ",acl");
619 if (!test_opt(sb, POSIX_ACL) && (def_mount_opts & EXT3_DEFM_ACL))
620 seq_puts(seq, ",noacl");
621 #endif
622 if (!test_opt(sb, RESERVATION))
623 seq_puts(seq, ",noreservation");
624 if (sbi->s_commit_interval) {
625 seq_printf(seq, ",commit=%u",
626 (unsigned) (sbi->s_commit_interval / HZ));
628 if (test_opt(sb, BARRIER))
629 seq_puts(seq, ",barrier=1");
630 if (test_opt(sb, NOBH))
631 seq_puts(seq, ",nobh");
633 if (test_opt(sb, DATA_FLAGS) == EXT3_MOUNT_JOURNAL_DATA)
634 seq_puts(seq, ",data=journal");
635 else if (test_opt(sb, DATA_FLAGS) == EXT3_MOUNT_ORDERED_DATA)
636 seq_puts(seq, ",data=ordered");
637 else if (test_opt(sb, DATA_FLAGS) == EXT3_MOUNT_WRITEBACK_DATA)
638 seq_puts(seq, ",data=writeback");
640 if (test_opt(sb, DATA_ERR_ABORT))
641 seq_puts(seq, ",data_err=abort");
643 ext3_show_quota_options(seq, sb);
645 return 0;
649 static struct inode *ext3_nfs_get_inode(struct super_block *sb,
650 u64 ino, u32 generation)
652 struct inode *inode;
654 if (ino < EXT3_FIRST_INO(sb) && ino != EXT3_ROOT_INO)
655 return ERR_PTR(-ESTALE);
656 if (ino > le32_to_cpu(EXT3_SB(sb)->s_es->s_inodes_count))
657 return ERR_PTR(-ESTALE);
659 /* iget isn't really right if the inode is currently unallocated!!
661 * ext3_read_inode will return a bad_inode if the inode had been
662 * deleted, so we should be safe.
664 * Currently we don't know the generation for parent directory, so
665 * a generation of 0 means "accept any"
667 inode = ext3_iget(sb, ino);
668 if (IS_ERR(inode))
669 return ERR_CAST(inode);
670 if (generation && inode->i_generation != generation) {
671 iput(inode);
672 return ERR_PTR(-ESTALE);
675 return inode;
678 static struct dentry *ext3_fh_to_dentry(struct super_block *sb, struct fid *fid,
679 int fh_len, int fh_type)
681 return generic_fh_to_dentry(sb, fid, fh_len, fh_type,
682 ext3_nfs_get_inode);
685 static struct dentry *ext3_fh_to_parent(struct super_block *sb, struct fid *fid,
686 int fh_len, int fh_type)
688 return generic_fh_to_parent(sb, fid, fh_len, fh_type,
689 ext3_nfs_get_inode);
693 * Try to release metadata pages (indirect blocks, directories) which are
694 * mapped via the block device. Since these pages could have journal heads
695 * which would prevent try_to_free_buffers() from freeing them, we must use
696 * jbd layer's try_to_free_buffers() function to release them.
698 static int bdev_try_to_free_page(struct super_block *sb, struct page *page,
699 gfp_t wait)
701 journal_t *journal = EXT3_SB(sb)->s_journal;
703 WARN_ON(PageChecked(page));
704 if (!page_has_buffers(page))
705 return 0;
706 if (journal)
707 return journal_try_to_free_buffers(journal, page,
708 wait & ~__GFP_WAIT);
709 return try_to_free_buffers(page);
712 #ifdef CONFIG_QUOTA
713 #define QTYPE2NAME(t) ((t)==USRQUOTA?"user":"group")
714 #define QTYPE2MOPT(on, t) ((t)==USRQUOTA?((on)##USRJQUOTA):((on)##GRPJQUOTA))
716 static int ext3_write_dquot(struct dquot *dquot);
717 static int ext3_acquire_dquot(struct dquot *dquot);
718 static int ext3_release_dquot(struct dquot *dquot);
719 static int ext3_mark_dquot_dirty(struct dquot *dquot);
720 static int ext3_write_info(struct super_block *sb, int type);
721 static int ext3_quota_on(struct super_block *sb, int type, int format_id,
722 char *path, int remount);
723 static int ext3_quota_on_mount(struct super_block *sb, int type);
724 static ssize_t ext3_quota_read(struct super_block *sb, int type, char *data,
725 size_t len, loff_t off);
726 static ssize_t ext3_quota_write(struct super_block *sb, int type,
727 const char *data, size_t len, loff_t off);
729 static struct dquot_operations ext3_quota_operations = {
730 .initialize = dquot_initialize,
731 .drop = dquot_drop,
732 .alloc_space = dquot_alloc_space,
733 .alloc_inode = dquot_alloc_inode,
734 .free_space = dquot_free_space,
735 .free_inode = dquot_free_inode,
736 .transfer = dquot_transfer,
737 .write_dquot = ext3_write_dquot,
738 .acquire_dquot = ext3_acquire_dquot,
739 .release_dquot = ext3_release_dquot,
740 .mark_dirty = ext3_mark_dquot_dirty,
741 .write_info = ext3_write_info,
742 .alloc_dquot = dquot_alloc,
743 .destroy_dquot = dquot_destroy,
746 static struct quotactl_ops ext3_qctl_operations = {
747 .quota_on = ext3_quota_on,
748 .quota_off = vfs_quota_off,
749 .quota_sync = vfs_quota_sync,
750 .get_info = vfs_get_dqinfo,
751 .set_info = vfs_set_dqinfo,
752 .get_dqblk = vfs_get_dqblk,
753 .set_dqblk = vfs_set_dqblk
755 #endif
757 static const struct super_operations ext3_sops = {
758 .alloc_inode = ext3_alloc_inode,
759 .destroy_inode = ext3_destroy_inode,
760 .write_inode = ext3_write_inode,
761 .dirty_inode = ext3_dirty_inode,
762 .delete_inode = ext3_delete_inode,
763 .put_super = ext3_put_super,
764 .write_super = ext3_write_super,
765 .sync_fs = ext3_sync_fs,
766 .freeze_fs = ext3_freeze,
767 .unfreeze_fs = ext3_unfreeze,
768 .statfs = ext3_statfs,
769 .remount_fs = ext3_remount,
770 .clear_inode = ext3_clear_inode,
771 .show_options = ext3_show_options,
772 #ifdef CONFIG_QUOTA
773 .quota_read = ext3_quota_read,
774 .quota_write = ext3_quota_write,
775 #endif
776 .bdev_try_to_free_page = bdev_try_to_free_page,
779 static const struct export_operations ext3_export_ops = {
780 .fh_to_dentry = ext3_fh_to_dentry,
781 .fh_to_parent = ext3_fh_to_parent,
782 .get_parent = ext3_get_parent,
785 enum {
786 Opt_bsd_df, Opt_minix_df, Opt_grpid, Opt_nogrpid,
787 Opt_resgid, Opt_resuid, Opt_sb, Opt_err_cont, Opt_err_panic, Opt_err_ro,
788 Opt_nouid32, Opt_nocheck, Opt_debug, Opt_oldalloc, Opt_orlov,
789 Opt_user_xattr, Opt_nouser_xattr, Opt_acl, Opt_noacl,
790 Opt_reservation, Opt_noreservation, Opt_noload, Opt_nobh, Opt_bh,
791 Opt_commit, Opt_journal_update, Opt_journal_inum, Opt_journal_dev,
792 Opt_abort, Opt_data_journal, Opt_data_ordered, Opt_data_writeback,
793 Opt_data_err_abort, Opt_data_err_ignore,
794 Opt_usrjquota, Opt_grpjquota, Opt_offusrjquota, Opt_offgrpjquota,
795 Opt_jqfmt_vfsold, Opt_jqfmt_vfsv0, Opt_quota, Opt_noquota,
796 Opt_ignore, Opt_barrier, Opt_err, Opt_resize, Opt_usrquota,
797 Opt_grpquota
800 static const match_table_t tokens = {
801 {Opt_bsd_df, "bsddf"},
802 {Opt_minix_df, "minixdf"},
803 {Opt_grpid, "grpid"},
804 {Opt_grpid, "bsdgroups"},
805 {Opt_nogrpid, "nogrpid"},
806 {Opt_nogrpid, "sysvgroups"},
807 {Opt_resgid, "resgid=%u"},
808 {Opt_resuid, "resuid=%u"},
809 {Opt_sb, "sb=%u"},
810 {Opt_err_cont, "errors=continue"},
811 {Opt_err_panic, "errors=panic"},
812 {Opt_err_ro, "errors=remount-ro"},
813 {Opt_nouid32, "nouid32"},
814 {Opt_nocheck, "nocheck"},
815 {Opt_nocheck, "check=none"},
816 {Opt_debug, "debug"},
817 {Opt_oldalloc, "oldalloc"},
818 {Opt_orlov, "orlov"},
819 {Opt_user_xattr, "user_xattr"},
820 {Opt_nouser_xattr, "nouser_xattr"},
821 {Opt_acl, "acl"},
822 {Opt_noacl, "noacl"},
823 {Opt_reservation, "reservation"},
824 {Opt_noreservation, "noreservation"},
825 {Opt_noload, "noload"},
826 {Opt_nobh, "nobh"},
827 {Opt_bh, "bh"},
828 {Opt_commit, "commit=%u"},
829 {Opt_journal_update, "journal=update"},
830 {Opt_journal_inum, "journal=%u"},
831 {Opt_journal_dev, "journal_dev=%u"},
832 {Opt_abort, "abort"},
833 {Opt_data_journal, "data=journal"},
834 {Opt_data_ordered, "data=ordered"},
835 {Opt_data_writeback, "data=writeback"},
836 {Opt_data_err_abort, "data_err=abort"},
837 {Opt_data_err_ignore, "data_err=ignore"},
838 {Opt_offusrjquota, "usrjquota="},
839 {Opt_usrjquota, "usrjquota=%s"},
840 {Opt_offgrpjquota, "grpjquota="},
841 {Opt_grpjquota, "grpjquota=%s"},
842 {Opt_jqfmt_vfsold, "jqfmt=vfsold"},
843 {Opt_jqfmt_vfsv0, "jqfmt=vfsv0"},
844 {Opt_grpquota, "grpquota"},
845 {Opt_noquota, "noquota"},
846 {Opt_quota, "quota"},
847 {Opt_usrquota, "usrquota"},
848 {Opt_barrier, "barrier=%u"},
849 {Opt_resize, "resize"},
850 {Opt_err, NULL},
853 static ext3_fsblk_t get_sb_block(void **data)
855 ext3_fsblk_t sb_block;
856 char *options = (char *) *data;
858 if (!options || strncmp(options, "sb=", 3) != 0)
859 return 1; /* Default location */
860 options += 3;
861 /*todo: use simple_strtoll with >32bit ext3 */
862 sb_block = simple_strtoul(options, &options, 0);
863 if (*options && *options != ',') {
864 printk("EXT3-fs: Invalid sb specification: %s\n",
865 (char *) *data);
866 return 1;
868 if (*options == ',')
869 options++;
870 *data = (void *) options;
871 return sb_block;
874 static int parse_options (char *options, struct super_block *sb,
875 unsigned int *inum, unsigned long *journal_devnum,
876 ext3_fsblk_t *n_blocks_count, int is_remount)
878 struct ext3_sb_info *sbi = EXT3_SB(sb);
879 char * p;
880 substring_t args[MAX_OPT_ARGS];
881 int data_opt = 0;
882 int option;
883 #ifdef CONFIG_QUOTA
884 int qtype, qfmt;
885 char *qname;
886 #endif
888 if (!options)
889 return 1;
891 while ((p = strsep (&options, ",")) != NULL) {
892 int token;
893 if (!*p)
894 continue;
896 token = match_token(p, tokens, args);
897 switch (token) {
898 case Opt_bsd_df:
899 clear_opt (sbi->s_mount_opt, MINIX_DF);
900 break;
901 case Opt_minix_df:
902 set_opt (sbi->s_mount_opt, MINIX_DF);
903 break;
904 case Opt_grpid:
905 set_opt (sbi->s_mount_opt, GRPID);
906 break;
907 case Opt_nogrpid:
908 clear_opt (sbi->s_mount_opt, GRPID);
909 break;
910 case Opt_resuid:
911 if (match_int(&args[0], &option))
912 return 0;
913 sbi->s_resuid = option;
914 break;
915 case Opt_resgid:
916 if (match_int(&args[0], &option))
917 return 0;
918 sbi->s_resgid = option;
919 break;
920 case Opt_sb:
921 /* handled by get_sb_block() instead of here */
922 /* *sb_block = match_int(&args[0]); */
923 break;
924 case Opt_err_panic:
925 clear_opt (sbi->s_mount_opt, ERRORS_CONT);
926 clear_opt (sbi->s_mount_opt, ERRORS_RO);
927 set_opt (sbi->s_mount_opt, ERRORS_PANIC);
928 break;
929 case Opt_err_ro:
930 clear_opt (sbi->s_mount_opt, ERRORS_CONT);
931 clear_opt (sbi->s_mount_opt, ERRORS_PANIC);
932 set_opt (sbi->s_mount_opt, ERRORS_RO);
933 break;
934 case Opt_err_cont:
935 clear_opt (sbi->s_mount_opt, ERRORS_RO);
936 clear_opt (sbi->s_mount_opt, ERRORS_PANIC);
937 set_opt (sbi->s_mount_opt, ERRORS_CONT);
938 break;
939 case Opt_nouid32:
940 set_opt (sbi->s_mount_opt, NO_UID32);
941 break;
942 case Opt_nocheck:
943 clear_opt (sbi->s_mount_opt, CHECK);
944 break;
945 case Opt_debug:
946 set_opt (sbi->s_mount_opt, DEBUG);
947 break;
948 case Opt_oldalloc:
949 set_opt (sbi->s_mount_opt, OLDALLOC);
950 break;
951 case Opt_orlov:
952 clear_opt (sbi->s_mount_opt, OLDALLOC);
953 break;
954 #ifdef CONFIG_EXT3_FS_XATTR
955 case Opt_user_xattr:
956 set_opt (sbi->s_mount_opt, XATTR_USER);
957 break;
958 case Opt_nouser_xattr:
959 clear_opt (sbi->s_mount_opt, XATTR_USER);
960 break;
961 #else
962 case Opt_user_xattr:
963 case Opt_nouser_xattr:
964 printk("EXT3 (no)user_xattr options not supported\n");
965 break;
966 #endif
967 #ifdef CONFIG_EXT3_FS_POSIX_ACL
968 case Opt_acl:
969 set_opt(sbi->s_mount_opt, POSIX_ACL);
970 break;
971 case Opt_noacl:
972 clear_opt(sbi->s_mount_opt, POSIX_ACL);
973 break;
974 #else
975 case Opt_acl:
976 case Opt_noacl:
977 printk("EXT3 (no)acl options not supported\n");
978 break;
979 #endif
980 case Opt_reservation:
981 set_opt(sbi->s_mount_opt, RESERVATION);
982 break;
983 case Opt_noreservation:
984 clear_opt(sbi->s_mount_opt, RESERVATION);
985 break;
986 case Opt_journal_update:
987 /* @@@ FIXME */
988 /* Eventually we will want to be able to create
989 a journal file here. For now, only allow the
990 user to specify an existing inode to be the
991 journal file. */
992 if (is_remount) {
993 printk(KERN_ERR "EXT3-fs: cannot specify "
994 "journal on remount\n");
995 return 0;
997 set_opt (sbi->s_mount_opt, UPDATE_JOURNAL);
998 break;
999 case Opt_journal_inum:
1000 if (is_remount) {
1001 printk(KERN_ERR "EXT3-fs: cannot specify "
1002 "journal on remount\n");
1003 return 0;
1005 if (match_int(&args[0], &option))
1006 return 0;
1007 *inum = option;
1008 break;
1009 case Opt_journal_dev:
1010 if (is_remount) {
1011 printk(KERN_ERR "EXT3-fs: cannot specify "
1012 "journal on remount\n");
1013 return 0;
1015 if (match_int(&args[0], &option))
1016 return 0;
1017 *journal_devnum = option;
1018 break;
1019 case Opt_noload:
1020 set_opt (sbi->s_mount_opt, NOLOAD);
1021 break;
1022 case Opt_commit:
1023 if (match_int(&args[0], &option))
1024 return 0;
1025 if (option < 0)
1026 return 0;
1027 if (option == 0)
1028 option = JBD_DEFAULT_MAX_COMMIT_AGE;
1029 sbi->s_commit_interval = HZ * option;
1030 break;
1031 case Opt_data_journal:
1032 data_opt = EXT3_MOUNT_JOURNAL_DATA;
1033 goto datacheck;
1034 case Opt_data_ordered:
1035 data_opt = EXT3_MOUNT_ORDERED_DATA;
1036 goto datacheck;
1037 case Opt_data_writeback:
1038 data_opt = EXT3_MOUNT_WRITEBACK_DATA;
1039 datacheck:
1040 if (is_remount) {
1041 if ((sbi->s_mount_opt & EXT3_MOUNT_DATA_FLAGS)
1042 != data_opt) {
1043 printk(KERN_ERR
1044 "EXT3-fs: cannot change data "
1045 "mode on remount\n");
1046 return 0;
1048 } else {
1049 sbi->s_mount_opt &= ~EXT3_MOUNT_DATA_FLAGS;
1050 sbi->s_mount_opt |= data_opt;
1052 break;
1053 case Opt_data_err_abort:
1054 set_opt(sbi->s_mount_opt, DATA_ERR_ABORT);
1055 break;
1056 case Opt_data_err_ignore:
1057 clear_opt(sbi->s_mount_opt, DATA_ERR_ABORT);
1058 break;
1059 #ifdef CONFIG_QUOTA
1060 case Opt_usrjquota:
1061 qtype = USRQUOTA;
1062 goto set_qf_name;
1063 case Opt_grpjquota:
1064 qtype = GRPQUOTA;
1065 set_qf_name:
1066 if (sb_any_quota_loaded(sb) &&
1067 !sbi->s_qf_names[qtype]) {
1068 printk(KERN_ERR
1069 "EXT3-fs: Cannot change journaled "
1070 "quota options when quota turned on.\n");
1071 return 0;
1073 qname = match_strdup(&args[0]);
1074 if (!qname) {
1075 printk(KERN_ERR
1076 "EXT3-fs: not enough memory for "
1077 "storing quotafile name.\n");
1078 return 0;
1080 if (sbi->s_qf_names[qtype] &&
1081 strcmp(sbi->s_qf_names[qtype], qname)) {
1082 printk(KERN_ERR
1083 "EXT3-fs: %s quota file already "
1084 "specified.\n", QTYPE2NAME(qtype));
1085 kfree(qname);
1086 return 0;
1088 sbi->s_qf_names[qtype] = qname;
1089 if (strchr(sbi->s_qf_names[qtype], '/')) {
1090 printk(KERN_ERR
1091 "EXT3-fs: quotafile must be on "
1092 "filesystem root.\n");
1093 kfree(sbi->s_qf_names[qtype]);
1094 sbi->s_qf_names[qtype] = NULL;
1095 return 0;
1097 set_opt(sbi->s_mount_opt, QUOTA);
1098 break;
1099 case Opt_offusrjquota:
1100 qtype = USRQUOTA;
1101 goto clear_qf_name;
1102 case Opt_offgrpjquota:
1103 qtype = GRPQUOTA;
1104 clear_qf_name:
1105 if (sb_any_quota_loaded(sb) &&
1106 sbi->s_qf_names[qtype]) {
1107 printk(KERN_ERR "EXT3-fs: Cannot change "
1108 "journaled quota options when "
1109 "quota turned on.\n");
1110 return 0;
1113 * The space will be released later when all options
1114 * are confirmed to be correct
1116 sbi->s_qf_names[qtype] = NULL;
1117 break;
1118 case Opt_jqfmt_vfsold:
1119 qfmt = QFMT_VFS_OLD;
1120 goto set_qf_format;
1121 case Opt_jqfmt_vfsv0:
1122 qfmt = QFMT_VFS_V0;
1123 set_qf_format:
1124 if (sb_any_quota_loaded(sb) &&
1125 sbi->s_jquota_fmt != qfmt) {
1126 printk(KERN_ERR "EXT3-fs: Cannot change "
1127 "journaled quota options when "
1128 "quota turned on.\n");
1129 return 0;
1131 sbi->s_jquota_fmt = qfmt;
1132 break;
1133 case Opt_quota:
1134 case Opt_usrquota:
1135 set_opt(sbi->s_mount_opt, QUOTA);
1136 set_opt(sbi->s_mount_opt, USRQUOTA);
1137 break;
1138 case Opt_grpquota:
1139 set_opt(sbi->s_mount_opt, QUOTA);
1140 set_opt(sbi->s_mount_opt, GRPQUOTA);
1141 break;
1142 case Opt_noquota:
1143 if (sb_any_quota_loaded(sb)) {
1144 printk(KERN_ERR "EXT3-fs: Cannot change quota "
1145 "options when quota turned on.\n");
1146 return 0;
1148 clear_opt(sbi->s_mount_opt, QUOTA);
1149 clear_opt(sbi->s_mount_opt, USRQUOTA);
1150 clear_opt(sbi->s_mount_opt, GRPQUOTA);
1151 break;
1152 #else
1153 case Opt_quota:
1154 case Opt_usrquota:
1155 case Opt_grpquota:
1156 printk(KERN_ERR
1157 "EXT3-fs: quota options not supported.\n");
1158 break;
1159 case Opt_usrjquota:
1160 case Opt_grpjquota:
1161 case Opt_offusrjquota:
1162 case Opt_offgrpjquota:
1163 case Opt_jqfmt_vfsold:
1164 case Opt_jqfmt_vfsv0:
1165 printk(KERN_ERR
1166 "EXT3-fs: journaled quota options not "
1167 "supported.\n");
1168 break;
1169 case Opt_noquota:
1170 break;
1171 #endif
1172 case Opt_abort:
1173 set_opt(sbi->s_mount_opt, ABORT);
1174 break;
1175 case Opt_barrier:
1176 if (match_int(&args[0], &option))
1177 return 0;
1178 if (option)
1179 set_opt(sbi->s_mount_opt, BARRIER);
1180 else
1181 clear_opt(sbi->s_mount_opt, BARRIER);
1182 break;
1183 case Opt_ignore:
1184 break;
1185 case Opt_resize:
1186 if (!is_remount) {
1187 printk("EXT3-fs: resize option only available "
1188 "for remount\n");
1189 return 0;
1191 if (match_int(&args[0], &option) != 0)
1192 return 0;
1193 *n_blocks_count = option;
1194 break;
1195 case Opt_nobh:
1196 set_opt(sbi->s_mount_opt, NOBH);
1197 break;
1198 case Opt_bh:
1199 clear_opt(sbi->s_mount_opt, NOBH);
1200 break;
1201 default:
1202 printk (KERN_ERR
1203 "EXT3-fs: Unrecognized mount option \"%s\" "
1204 "or missing value\n", p);
1205 return 0;
1208 #ifdef CONFIG_QUOTA
1209 if (sbi->s_qf_names[USRQUOTA] || sbi->s_qf_names[GRPQUOTA]) {
1210 if ((sbi->s_mount_opt & EXT3_MOUNT_USRQUOTA) &&
1211 sbi->s_qf_names[USRQUOTA])
1212 clear_opt(sbi->s_mount_opt, USRQUOTA);
1214 if ((sbi->s_mount_opt & EXT3_MOUNT_GRPQUOTA) &&
1215 sbi->s_qf_names[GRPQUOTA])
1216 clear_opt(sbi->s_mount_opt, GRPQUOTA);
1218 if ((sbi->s_qf_names[USRQUOTA] &&
1219 (sbi->s_mount_opt & EXT3_MOUNT_GRPQUOTA)) ||
1220 (sbi->s_qf_names[GRPQUOTA] &&
1221 (sbi->s_mount_opt & EXT3_MOUNT_USRQUOTA))) {
1222 printk(KERN_ERR "EXT3-fs: old and new quota "
1223 "format mixing.\n");
1224 return 0;
1227 if (!sbi->s_jquota_fmt) {
1228 printk(KERN_ERR "EXT3-fs: journaled quota format "
1229 "not specified.\n");
1230 return 0;
1232 } else {
1233 if (sbi->s_jquota_fmt) {
1234 printk(KERN_ERR "EXT3-fs: journaled quota format "
1235 "specified with no journaling "
1236 "enabled.\n");
1237 return 0;
1240 #endif
1241 return 1;
1244 static int ext3_setup_super(struct super_block *sb, struct ext3_super_block *es,
1245 int read_only)
1247 struct ext3_sb_info *sbi = EXT3_SB(sb);
1248 int res = 0;
1250 if (le32_to_cpu(es->s_rev_level) > EXT3_MAX_SUPP_REV) {
1251 printk (KERN_ERR "EXT3-fs warning: revision level too high, "
1252 "forcing read-only mode\n");
1253 res = MS_RDONLY;
1255 if (read_only)
1256 return res;
1257 if (!(sbi->s_mount_state & EXT3_VALID_FS))
1258 printk (KERN_WARNING "EXT3-fs warning: mounting unchecked fs, "
1259 "running e2fsck is recommended\n");
1260 else if ((sbi->s_mount_state & EXT3_ERROR_FS))
1261 printk (KERN_WARNING
1262 "EXT3-fs warning: mounting fs with errors, "
1263 "running e2fsck is recommended\n");
1264 else if ((__s16) le16_to_cpu(es->s_max_mnt_count) >= 0 &&
1265 le16_to_cpu(es->s_mnt_count) >=
1266 (unsigned short) (__s16) le16_to_cpu(es->s_max_mnt_count))
1267 printk (KERN_WARNING
1268 "EXT3-fs warning: maximal mount count reached, "
1269 "running e2fsck is recommended\n");
1270 else if (le32_to_cpu(es->s_checkinterval) &&
1271 (le32_to_cpu(es->s_lastcheck) +
1272 le32_to_cpu(es->s_checkinterval) <= get_seconds()))
1273 printk (KERN_WARNING
1274 "EXT3-fs warning: checktime reached, "
1275 "running e2fsck is recommended\n");
1276 #if 0
1277 /* @@@ We _will_ want to clear the valid bit if we find
1278 inconsistencies, to force a fsck at reboot. But for
1279 a plain journaled filesystem we can keep it set as
1280 valid forever! :) */
1281 es->s_state &= cpu_to_le16(~EXT3_VALID_FS);
1282 #endif
1283 if (!(__s16) le16_to_cpu(es->s_max_mnt_count))
1284 es->s_max_mnt_count = cpu_to_le16(EXT3_DFL_MAX_MNT_COUNT);
1285 le16_add_cpu(&es->s_mnt_count, 1);
1286 es->s_mtime = cpu_to_le32(get_seconds());
1287 ext3_update_dynamic_rev(sb);
1288 EXT3_SET_INCOMPAT_FEATURE(sb, EXT3_FEATURE_INCOMPAT_RECOVER);
1290 ext3_commit_super(sb, es, 1);
1291 if (test_opt(sb, DEBUG))
1292 printk(KERN_INFO "[EXT3 FS bs=%lu, gc=%lu, "
1293 "bpg=%lu, ipg=%lu, mo=%04lx]\n",
1294 sb->s_blocksize,
1295 sbi->s_groups_count,
1296 EXT3_BLOCKS_PER_GROUP(sb),
1297 EXT3_INODES_PER_GROUP(sb),
1298 sbi->s_mount_opt);
1300 printk(KERN_INFO "EXT3 FS on %s, ", sb->s_id);
1301 if (EXT3_SB(sb)->s_journal->j_inode == NULL) {
1302 char b[BDEVNAME_SIZE];
1304 printk("external journal on %s\n",
1305 bdevname(EXT3_SB(sb)->s_journal->j_dev, b));
1306 } else {
1307 printk("internal journal\n");
1309 return res;
1312 /* Called at mount-time, super-block is locked */
1313 static int ext3_check_descriptors(struct super_block *sb)
1315 struct ext3_sb_info *sbi = EXT3_SB(sb);
1316 int i;
1318 ext3_debug ("Checking group descriptors");
1320 for (i = 0; i < sbi->s_groups_count; i++) {
1321 struct ext3_group_desc *gdp = ext3_get_group_desc(sb, i, NULL);
1322 ext3_fsblk_t first_block = ext3_group_first_block_no(sb, i);
1323 ext3_fsblk_t last_block;
1325 if (i == sbi->s_groups_count - 1)
1326 last_block = le32_to_cpu(sbi->s_es->s_blocks_count) - 1;
1327 else
1328 last_block = first_block +
1329 (EXT3_BLOCKS_PER_GROUP(sb) - 1);
1331 if (le32_to_cpu(gdp->bg_block_bitmap) < first_block ||
1332 le32_to_cpu(gdp->bg_block_bitmap) > last_block)
1334 ext3_error (sb, "ext3_check_descriptors",
1335 "Block bitmap for group %d"
1336 " not in group (block %lu)!",
1337 i, (unsigned long)
1338 le32_to_cpu(gdp->bg_block_bitmap));
1339 return 0;
1341 if (le32_to_cpu(gdp->bg_inode_bitmap) < first_block ||
1342 le32_to_cpu(gdp->bg_inode_bitmap) > last_block)
1344 ext3_error (sb, "ext3_check_descriptors",
1345 "Inode bitmap for group %d"
1346 " not in group (block %lu)!",
1347 i, (unsigned long)
1348 le32_to_cpu(gdp->bg_inode_bitmap));
1349 return 0;
1351 if (le32_to_cpu(gdp->bg_inode_table) < first_block ||
1352 le32_to_cpu(gdp->bg_inode_table) + sbi->s_itb_per_group - 1 >
1353 last_block)
1355 ext3_error (sb, "ext3_check_descriptors",
1356 "Inode table for group %d"
1357 " not in group (block %lu)!",
1358 i, (unsigned long)
1359 le32_to_cpu(gdp->bg_inode_table));
1360 return 0;
1364 sbi->s_es->s_free_blocks_count=cpu_to_le32(ext3_count_free_blocks(sb));
1365 sbi->s_es->s_free_inodes_count=cpu_to_le32(ext3_count_free_inodes(sb));
1366 return 1;
1370 /* ext3_orphan_cleanup() walks a singly-linked list of inodes (starting at
1371 * the superblock) which were deleted from all directories, but held open by
1372 * a process at the time of a crash. We walk the list and try to delete these
1373 * inodes at recovery time (only with a read-write filesystem).
1375 * In order to keep the orphan inode chain consistent during traversal (in
1376 * case of crash during recovery), we link each inode into the superblock
1377 * orphan list_head and handle it the same way as an inode deletion during
1378 * normal operation (which journals the operations for us).
1380 * We only do an iget() and an iput() on each inode, which is very safe if we
1381 * accidentally point at an in-use or already deleted inode. The worst that
1382 * can happen in this case is that we get a "bit already cleared" message from
1383 * ext3_free_inode(). The only reason we would point at a wrong inode is if
1384 * e2fsck was run on this filesystem, and it must have already done the orphan
1385 * inode cleanup for us, so we can safely abort without any further action.
1387 static void ext3_orphan_cleanup (struct super_block * sb,
1388 struct ext3_super_block * es)
1390 unsigned int s_flags = sb->s_flags;
1391 int nr_orphans = 0, nr_truncates = 0;
1392 #ifdef CONFIG_QUOTA
1393 int i;
1394 #endif
1395 if (!es->s_last_orphan) {
1396 jbd_debug(4, "no orphan inodes to clean up\n");
1397 return;
1400 if (bdev_read_only(sb->s_bdev)) {
1401 printk(KERN_ERR "EXT3-fs: write access "
1402 "unavailable, skipping orphan cleanup.\n");
1403 return;
1406 if (EXT3_SB(sb)->s_mount_state & EXT3_ERROR_FS) {
1407 if (es->s_last_orphan)
1408 jbd_debug(1, "Errors on filesystem, "
1409 "clearing orphan list.\n");
1410 es->s_last_orphan = 0;
1411 jbd_debug(1, "Skipping orphan recovery on fs with errors.\n");
1412 return;
1415 if (s_flags & MS_RDONLY) {
1416 printk(KERN_INFO "EXT3-fs: %s: orphan cleanup on readonly fs\n",
1417 sb->s_id);
1418 sb->s_flags &= ~MS_RDONLY;
1420 #ifdef CONFIG_QUOTA
1421 /* Needed for iput() to work correctly and not trash data */
1422 sb->s_flags |= MS_ACTIVE;
1423 /* Turn on quotas so that they are updated correctly */
1424 for (i = 0; i < MAXQUOTAS; i++) {
1425 if (EXT3_SB(sb)->s_qf_names[i]) {
1426 int ret = ext3_quota_on_mount(sb, i);
1427 if (ret < 0)
1428 printk(KERN_ERR
1429 "EXT3-fs: Cannot turn on journaled "
1430 "quota: error %d\n", ret);
1433 #endif
1435 while (es->s_last_orphan) {
1436 struct inode *inode;
1438 inode = ext3_orphan_get(sb, le32_to_cpu(es->s_last_orphan));
1439 if (IS_ERR(inode)) {
1440 es->s_last_orphan = 0;
1441 break;
1444 list_add(&EXT3_I(inode)->i_orphan, &EXT3_SB(sb)->s_orphan);
1445 vfs_dq_init(inode);
1446 if (inode->i_nlink) {
1447 printk(KERN_DEBUG
1448 "%s: truncating inode %lu to %Ld bytes\n",
1449 __func__, inode->i_ino, inode->i_size);
1450 jbd_debug(2, "truncating inode %lu to %Ld bytes\n",
1451 inode->i_ino, inode->i_size);
1452 ext3_truncate(inode);
1453 nr_truncates++;
1454 } else {
1455 printk(KERN_DEBUG
1456 "%s: deleting unreferenced inode %lu\n",
1457 __func__, inode->i_ino);
1458 jbd_debug(2, "deleting unreferenced inode %lu\n",
1459 inode->i_ino);
1460 nr_orphans++;
1462 iput(inode); /* The delete magic happens here! */
1465 #define PLURAL(x) (x), ((x)==1) ? "" : "s"
1467 if (nr_orphans)
1468 printk(KERN_INFO "EXT3-fs: %s: %d orphan inode%s deleted\n",
1469 sb->s_id, PLURAL(nr_orphans));
1470 if (nr_truncates)
1471 printk(KERN_INFO "EXT3-fs: %s: %d truncate%s cleaned up\n",
1472 sb->s_id, PLURAL(nr_truncates));
1473 #ifdef CONFIG_QUOTA
1474 /* Turn quotas off */
1475 for (i = 0; i < MAXQUOTAS; i++) {
1476 if (sb_dqopt(sb)->files[i])
1477 vfs_quota_off(sb, i, 0);
1479 #endif
1480 sb->s_flags = s_flags; /* Restore MS_RDONLY status */
1484 * Maximal file size. There is a direct, and {,double-,triple-}indirect
1485 * block limit, and also a limit of (2^32 - 1) 512-byte sectors in i_blocks.
1486 * We need to be 1 filesystem block less than the 2^32 sector limit.
1488 static loff_t ext3_max_size(int bits)
1490 loff_t res = EXT3_NDIR_BLOCKS;
1491 int meta_blocks;
1492 loff_t upper_limit;
1494 /* This is calculated to be the largest file size for a
1495 * dense, file such that the total number of
1496 * sectors in the file, including data and all indirect blocks,
1497 * does not exceed 2^32 -1
1498 * __u32 i_blocks representing the total number of
1499 * 512 bytes blocks of the file
1501 upper_limit = (1LL << 32) - 1;
1503 /* total blocks in file system block size */
1504 upper_limit >>= (bits - 9);
1507 /* indirect blocks */
1508 meta_blocks = 1;
1509 /* double indirect blocks */
1510 meta_blocks += 1 + (1LL << (bits-2));
1511 /* tripple indirect blocks */
1512 meta_blocks += 1 + (1LL << (bits-2)) + (1LL << (2*(bits-2)));
1514 upper_limit -= meta_blocks;
1515 upper_limit <<= bits;
1517 res += 1LL << (bits-2);
1518 res += 1LL << (2*(bits-2));
1519 res += 1LL << (3*(bits-2));
1520 res <<= bits;
1521 if (res > upper_limit)
1522 res = upper_limit;
1524 if (res > MAX_LFS_FILESIZE)
1525 res = MAX_LFS_FILESIZE;
1527 return res;
1530 static ext3_fsblk_t descriptor_loc(struct super_block *sb,
1531 ext3_fsblk_t logic_sb_block,
1532 int nr)
1534 struct ext3_sb_info *sbi = EXT3_SB(sb);
1535 unsigned long bg, first_meta_bg;
1536 int has_super = 0;
1538 first_meta_bg = le32_to_cpu(sbi->s_es->s_first_meta_bg);
1540 if (!EXT3_HAS_INCOMPAT_FEATURE(sb, EXT3_FEATURE_INCOMPAT_META_BG) ||
1541 nr < first_meta_bg)
1542 return (logic_sb_block + nr + 1);
1543 bg = sbi->s_desc_per_block * nr;
1544 if (ext3_bg_has_super(sb, bg))
1545 has_super = 1;
1546 return (has_super + ext3_group_first_block_no(sb, bg));
1550 static int ext3_fill_super (struct super_block *sb, void *data, int silent)
1552 struct buffer_head * bh;
1553 struct ext3_super_block *es = NULL;
1554 struct ext3_sb_info *sbi;
1555 ext3_fsblk_t block;
1556 ext3_fsblk_t sb_block = get_sb_block(&data);
1557 ext3_fsblk_t logic_sb_block;
1558 unsigned long offset = 0;
1559 unsigned int journal_inum = 0;
1560 unsigned long journal_devnum = 0;
1561 unsigned long def_mount_opts;
1562 struct inode *root;
1563 int blocksize;
1564 int hblock;
1565 int db_count;
1566 int i;
1567 int needs_recovery;
1568 int ret = -EINVAL;
1569 __le32 features;
1570 int err;
1572 sbi = kzalloc(sizeof(*sbi), GFP_KERNEL);
1573 if (!sbi)
1574 return -ENOMEM;
1576 sbi->s_blockgroup_lock =
1577 kzalloc(sizeof(struct blockgroup_lock), GFP_KERNEL);
1578 if (!sbi->s_blockgroup_lock) {
1579 kfree(sbi);
1580 return -ENOMEM;
1582 sb->s_fs_info = sbi;
1583 sbi->s_mount_opt = 0;
1584 sbi->s_resuid = EXT3_DEF_RESUID;
1585 sbi->s_resgid = EXT3_DEF_RESGID;
1586 sbi->s_sb_block = sb_block;
1588 unlock_kernel();
1590 blocksize = sb_min_blocksize(sb, EXT3_MIN_BLOCK_SIZE);
1591 if (!blocksize) {
1592 printk(KERN_ERR "EXT3-fs: unable to set blocksize\n");
1593 goto out_fail;
1597 * The ext3 superblock will not be buffer aligned for other than 1kB
1598 * block sizes. We need to calculate the offset from buffer start.
1600 if (blocksize != EXT3_MIN_BLOCK_SIZE) {
1601 logic_sb_block = (sb_block * EXT3_MIN_BLOCK_SIZE) / blocksize;
1602 offset = (sb_block * EXT3_MIN_BLOCK_SIZE) % blocksize;
1603 } else {
1604 logic_sb_block = sb_block;
1607 if (!(bh = sb_bread(sb, logic_sb_block))) {
1608 printk (KERN_ERR "EXT3-fs: unable to read superblock\n");
1609 goto out_fail;
1612 * Note: s_es must be initialized as soon as possible because
1613 * some ext3 macro-instructions depend on its value
1615 es = (struct ext3_super_block *) (((char *)bh->b_data) + offset);
1616 sbi->s_es = es;
1617 sb->s_magic = le16_to_cpu(es->s_magic);
1618 if (sb->s_magic != EXT3_SUPER_MAGIC)
1619 goto cantfind_ext3;
1621 /* Set defaults before we parse the mount options */
1622 def_mount_opts = le32_to_cpu(es->s_default_mount_opts);
1623 if (def_mount_opts & EXT3_DEFM_DEBUG)
1624 set_opt(sbi->s_mount_opt, DEBUG);
1625 if (def_mount_opts & EXT3_DEFM_BSDGROUPS)
1626 set_opt(sbi->s_mount_opt, GRPID);
1627 if (def_mount_opts & EXT3_DEFM_UID16)
1628 set_opt(sbi->s_mount_opt, NO_UID32);
1629 #ifdef CONFIG_EXT3_FS_XATTR
1630 if (def_mount_opts & EXT3_DEFM_XATTR_USER)
1631 set_opt(sbi->s_mount_opt, XATTR_USER);
1632 #endif
1633 #ifdef CONFIG_EXT3_FS_POSIX_ACL
1634 if (def_mount_opts & EXT3_DEFM_ACL)
1635 set_opt(sbi->s_mount_opt, POSIX_ACL);
1636 #endif
1637 if ((def_mount_opts & EXT3_DEFM_JMODE) == EXT3_DEFM_JMODE_DATA)
1638 sbi->s_mount_opt |= EXT3_MOUNT_JOURNAL_DATA;
1639 else if ((def_mount_opts & EXT3_DEFM_JMODE) == EXT3_DEFM_JMODE_ORDERED)
1640 sbi->s_mount_opt |= EXT3_MOUNT_ORDERED_DATA;
1641 else if ((def_mount_opts & EXT3_DEFM_JMODE) == EXT3_DEFM_JMODE_WBACK)
1642 sbi->s_mount_opt |= EXT3_MOUNT_WRITEBACK_DATA;
1644 if (le16_to_cpu(sbi->s_es->s_errors) == EXT3_ERRORS_PANIC)
1645 set_opt(sbi->s_mount_opt, ERRORS_PANIC);
1646 else if (le16_to_cpu(sbi->s_es->s_errors) == EXT3_ERRORS_CONTINUE)
1647 set_opt(sbi->s_mount_opt, ERRORS_CONT);
1648 else
1649 set_opt(sbi->s_mount_opt, ERRORS_RO);
1651 sbi->s_resuid = le16_to_cpu(es->s_def_resuid);
1652 sbi->s_resgid = le16_to_cpu(es->s_def_resgid);
1654 set_opt(sbi->s_mount_opt, RESERVATION);
1656 if (!parse_options ((char *) data, sb, &journal_inum, &journal_devnum,
1657 NULL, 0))
1658 goto failed_mount;
1660 sb->s_flags = (sb->s_flags & ~MS_POSIXACL) |
1661 ((sbi->s_mount_opt & EXT3_MOUNT_POSIX_ACL) ? MS_POSIXACL : 0);
1663 if (le32_to_cpu(es->s_rev_level) == EXT3_GOOD_OLD_REV &&
1664 (EXT3_HAS_COMPAT_FEATURE(sb, ~0U) ||
1665 EXT3_HAS_RO_COMPAT_FEATURE(sb, ~0U) ||
1666 EXT3_HAS_INCOMPAT_FEATURE(sb, ~0U)))
1667 printk(KERN_WARNING
1668 "EXT3-fs warning: feature flags set on rev 0 fs, "
1669 "running e2fsck is recommended\n");
1671 * Check feature flags regardless of the revision level, since we
1672 * previously didn't change the revision level when setting the flags,
1673 * so there is a chance incompat flags are set on a rev 0 filesystem.
1675 features = EXT3_HAS_INCOMPAT_FEATURE(sb, ~EXT3_FEATURE_INCOMPAT_SUPP);
1676 if (features) {
1677 printk(KERN_ERR "EXT3-fs: %s: couldn't mount because of "
1678 "unsupported optional features (%x).\n",
1679 sb->s_id, le32_to_cpu(features));
1680 goto failed_mount;
1682 features = EXT3_HAS_RO_COMPAT_FEATURE(sb, ~EXT3_FEATURE_RO_COMPAT_SUPP);
1683 if (!(sb->s_flags & MS_RDONLY) && features) {
1684 printk(KERN_ERR "EXT3-fs: %s: couldn't mount RDWR because of "
1685 "unsupported optional features (%x).\n",
1686 sb->s_id, le32_to_cpu(features));
1687 goto failed_mount;
1689 blocksize = BLOCK_SIZE << le32_to_cpu(es->s_log_block_size);
1691 if (blocksize < EXT3_MIN_BLOCK_SIZE ||
1692 blocksize > EXT3_MAX_BLOCK_SIZE) {
1693 printk(KERN_ERR
1694 "EXT3-fs: Unsupported filesystem blocksize %d on %s.\n",
1695 blocksize, sb->s_id);
1696 goto failed_mount;
1699 hblock = bdev_hardsect_size(sb->s_bdev);
1700 if (sb->s_blocksize != blocksize) {
1702 * Make sure the blocksize for the filesystem is larger
1703 * than the hardware sectorsize for the machine.
1705 if (blocksize < hblock) {
1706 printk(KERN_ERR "EXT3-fs: blocksize %d too small for "
1707 "device blocksize %d.\n", blocksize, hblock);
1708 goto failed_mount;
1711 brelse (bh);
1712 if (!sb_set_blocksize(sb, blocksize)) {
1713 printk(KERN_ERR "EXT3-fs: bad blocksize %d.\n",
1714 blocksize);
1715 goto out_fail;
1717 logic_sb_block = (sb_block * EXT3_MIN_BLOCK_SIZE) / blocksize;
1718 offset = (sb_block * EXT3_MIN_BLOCK_SIZE) % blocksize;
1719 bh = sb_bread(sb, logic_sb_block);
1720 if (!bh) {
1721 printk(KERN_ERR
1722 "EXT3-fs: Can't read superblock on 2nd try.\n");
1723 goto failed_mount;
1725 es = (struct ext3_super_block *)(((char *)bh->b_data) + offset);
1726 sbi->s_es = es;
1727 if (es->s_magic != cpu_to_le16(EXT3_SUPER_MAGIC)) {
1728 printk (KERN_ERR
1729 "EXT3-fs: Magic mismatch, very weird !\n");
1730 goto failed_mount;
1734 sb->s_maxbytes = ext3_max_size(sb->s_blocksize_bits);
1736 if (le32_to_cpu(es->s_rev_level) == EXT3_GOOD_OLD_REV) {
1737 sbi->s_inode_size = EXT3_GOOD_OLD_INODE_SIZE;
1738 sbi->s_first_ino = EXT3_GOOD_OLD_FIRST_INO;
1739 } else {
1740 sbi->s_inode_size = le16_to_cpu(es->s_inode_size);
1741 sbi->s_first_ino = le32_to_cpu(es->s_first_ino);
1742 if ((sbi->s_inode_size < EXT3_GOOD_OLD_INODE_SIZE) ||
1743 (!is_power_of_2(sbi->s_inode_size)) ||
1744 (sbi->s_inode_size > blocksize)) {
1745 printk (KERN_ERR
1746 "EXT3-fs: unsupported inode size: %d\n",
1747 sbi->s_inode_size);
1748 goto failed_mount;
1751 sbi->s_frag_size = EXT3_MIN_FRAG_SIZE <<
1752 le32_to_cpu(es->s_log_frag_size);
1753 if (blocksize != sbi->s_frag_size) {
1754 printk(KERN_ERR
1755 "EXT3-fs: fragsize %lu != blocksize %u (unsupported)\n",
1756 sbi->s_frag_size, blocksize);
1757 goto failed_mount;
1759 sbi->s_frags_per_block = 1;
1760 sbi->s_blocks_per_group = le32_to_cpu(es->s_blocks_per_group);
1761 sbi->s_frags_per_group = le32_to_cpu(es->s_frags_per_group);
1762 sbi->s_inodes_per_group = le32_to_cpu(es->s_inodes_per_group);
1763 if (EXT3_INODE_SIZE(sb) == 0 || EXT3_INODES_PER_GROUP(sb) == 0)
1764 goto cantfind_ext3;
1765 sbi->s_inodes_per_block = blocksize / EXT3_INODE_SIZE(sb);
1766 if (sbi->s_inodes_per_block == 0)
1767 goto cantfind_ext3;
1768 sbi->s_itb_per_group = sbi->s_inodes_per_group /
1769 sbi->s_inodes_per_block;
1770 sbi->s_desc_per_block = blocksize / sizeof(struct ext3_group_desc);
1771 sbi->s_sbh = bh;
1772 sbi->s_mount_state = le16_to_cpu(es->s_state);
1773 sbi->s_addr_per_block_bits = ilog2(EXT3_ADDR_PER_BLOCK(sb));
1774 sbi->s_desc_per_block_bits = ilog2(EXT3_DESC_PER_BLOCK(sb));
1775 for (i=0; i < 4; i++)
1776 sbi->s_hash_seed[i] = le32_to_cpu(es->s_hash_seed[i]);
1777 sbi->s_def_hash_version = es->s_def_hash_version;
1778 i = le32_to_cpu(es->s_flags);
1779 if (i & EXT2_FLAGS_UNSIGNED_HASH)
1780 sbi->s_hash_unsigned = 3;
1781 else if ((i & EXT2_FLAGS_SIGNED_HASH) == 0) {
1782 #ifdef __CHAR_UNSIGNED__
1783 es->s_flags |= cpu_to_le32(EXT2_FLAGS_UNSIGNED_HASH);
1784 sbi->s_hash_unsigned = 3;
1785 #else
1786 es->s_flags |= cpu_to_le32(EXT2_FLAGS_SIGNED_HASH);
1787 #endif
1788 sb->s_dirt = 1;
1791 if (sbi->s_blocks_per_group > blocksize * 8) {
1792 printk (KERN_ERR
1793 "EXT3-fs: #blocks per group too big: %lu\n",
1794 sbi->s_blocks_per_group);
1795 goto failed_mount;
1797 if (sbi->s_frags_per_group > blocksize * 8) {
1798 printk (KERN_ERR
1799 "EXT3-fs: #fragments per group too big: %lu\n",
1800 sbi->s_frags_per_group);
1801 goto failed_mount;
1803 if (sbi->s_inodes_per_group > blocksize * 8) {
1804 printk (KERN_ERR
1805 "EXT3-fs: #inodes per group too big: %lu\n",
1806 sbi->s_inodes_per_group);
1807 goto failed_mount;
1810 if (le32_to_cpu(es->s_blocks_count) >
1811 (sector_t)(~0ULL) >> (sb->s_blocksize_bits - 9)) {
1812 printk(KERN_ERR "EXT3-fs: filesystem on %s:"
1813 " too large to mount safely\n", sb->s_id);
1814 if (sizeof(sector_t) < 8)
1815 printk(KERN_WARNING "EXT3-fs: CONFIG_LBD not "
1816 "enabled\n");
1817 goto failed_mount;
1820 if (EXT3_BLOCKS_PER_GROUP(sb) == 0)
1821 goto cantfind_ext3;
1822 sbi->s_groups_count = ((le32_to_cpu(es->s_blocks_count) -
1823 le32_to_cpu(es->s_first_data_block) - 1)
1824 / EXT3_BLOCKS_PER_GROUP(sb)) + 1;
1825 db_count = (sbi->s_groups_count + EXT3_DESC_PER_BLOCK(sb) - 1) /
1826 EXT3_DESC_PER_BLOCK(sb);
1827 sbi->s_group_desc = kmalloc(db_count * sizeof (struct buffer_head *),
1828 GFP_KERNEL);
1829 if (sbi->s_group_desc == NULL) {
1830 printk (KERN_ERR "EXT3-fs: not enough memory\n");
1831 goto failed_mount;
1834 bgl_lock_init(sbi->s_blockgroup_lock);
1836 for (i = 0; i < db_count; i++) {
1837 block = descriptor_loc(sb, logic_sb_block, i);
1838 sbi->s_group_desc[i] = sb_bread(sb, block);
1839 if (!sbi->s_group_desc[i]) {
1840 printk (KERN_ERR "EXT3-fs: "
1841 "can't read group descriptor %d\n", i);
1842 db_count = i;
1843 goto failed_mount2;
1846 if (!ext3_check_descriptors (sb)) {
1847 printk(KERN_ERR "EXT3-fs: group descriptors corrupted!\n");
1848 goto failed_mount2;
1850 sbi->s_gdb_count = db_count;
1851 get_random_bytes(&sbi->s_next_generation, sizeof(u32));
1852 spin_lock_init(&sbi->s_next_gen_lock);
1854 err = percpu_counter_init(&sbi->s_freeblocks_counter,
1855 ext3_count_free_blocks(sb));
1856 if (!err) {
1857 err = percpu_counter_init(&sbi->s_freeinodes_counter,
1858 ext3_count_free_inodes(sb));
1860 if (!err) {
1861 err = percpu_counter_init(&sbi->s_dirs_counter,
1862 ext3_count_dirs(sb));
1864 if (err) {
1865 printk(KERN_ERR "EXT3-fs: insufficient memory\n");
1866 goto failed_mount3;
1869 /* per fileystem reservation list head & lock */
1870 spin_lock_init(&sbi->s_rsv_window_lock);
1871 sbi->s_rsv_window_root = RB_ROOT;
1872 /* Add a single, static dummy reservation to the start of the
1873 * reservation window list --- it gives us a placeholder for
1874 * append-at-start-of-list which makes the allocation logic
1875 * _much_ simpler. */
1876 sbi->s_rsv_window_head.rsv_start = EXT3_RESERVE_WINDOW_NOT_ALLOCATED;
1877 sbi->s_rsv_window_head.rsv_end = EXT3_RESERVE_WINDOW_NOT_ALLOCATED;
1878 sbi->s_rsv_window_head.rsv_alloc_hit = 0;
1879 sbi->s_rsv_window_head.rsv_goal_size = 0;
1880 ext3_rsv_window_add(sb, &sbi->s_rsv_window_head);
1883 * set up enough so that it can read an inode
1885 sb->s_op = &ext3_sops;
1886 sb->s_export_op = &ext3_export_ops;
1887 sb->s_xattr = ext3_xattr_handlers;
1888 #ifdef CONFIG_QUOTA
1889 sb->s_qcop = &ext3_qctl_operations;
1890 sb->dq_op = &ext3_quota_operations;
1891 #endif
1892 INIT_LIST_HEAD(&sbi->s_orphan); /* unlinked but open files */
1894 sb->s_root = NULL;
1896 needs_recovery = (es->s_last_orphan != 0 ||
1897 EXT3_HAS_INCOMPAT_FEATURE(sb,
1898 EXT3_FEATURE_INCOMPAT_RECOVER));
1901 * The first inode we look at is the journal inode. Don't try
1902 * root first: it may be modified in the journal!
1904 if (!test_opt(sb, NOLOAD) &&
1905 EXT3_HAS_COMPAT_FEATURE(sb, EXT3_FEATURE_COMPAT_HAS_JOURNAL)) {
1906 if (ext3_load_journal(sb, es, journal_devnum))
1907 goto failed_mount3;
1908 } else if (journal_inum) {
1909 if (ext3_create_journal(sb, es, journal_inum))
1910 goto failed_mount3;
1911 } else {
1912 if (!silent)
1913 printk (KERN_ERR
1914 "ext3: No journal on filesystem on %s\n",
1915 sb->s_id);
1916 goto failed_mount3;
1919 /* We have now updated the journal if required, so we can
1920 * validate the data journaling mode. */
1921 switch (test_opt(sb, DATA_FLAGS)) {
1922 case 0:
1923 /* No mode set, assume a default based on the journal
1924 capabilities: ORDERED_DATA if the journal can
1925 cope, else JOURNAL_DATA */
1926 if (journal_check_available_features
1927 (sbi->s_journal, 0, 0, JFS_FEATURE_INCOMPAT_REVOKE))
1928 set_opt(sbi->s_mount_opt, DEFAULT_DATA_MODE);
1929 else
1930 set_opt(sbi->s_mount_opt, JOURNAL_DATA);
1931 break;
1933 case EXT3_MOUNT_ORDERED_DATA:
1934 case EXT3_MOUNT_WRITEBACK_DATA:
1935 if (!journal_check_available_features
1936 (sbi->s_journal, 0, 0, JFS_FEATURE_INCOMPAT_REVOKE)) {
1937 printk(KERN_ERR "EXT3-fs: Journal does not support "
1938 "requested data journaling mode\n");
1939 goto failed_mount4;
1941 default:
1942 break;
1945 if (test_opt(sb, NOBH)) {
1946 if (!(test_opt(sb, DATA_FLAGS) == EXT3_MOUNT_WRITEBACK_DATA)) {
1947 printk(KERN_WARNING "EXT3-fs: Ignoring nobh option - "
1948 "its supported only with writeback mode\n");
1949 clear_opt(sbi->s_mount_opt, NOBH);
1953 * The journal_load will have done any necessary log recovery,
1954 * so we can safely mount the rest of the filesystem now.
1957 root = ext3_iget(sb, EXT3_ROOT_INO);
1958 if (IS_ERR(root)) {
1959 printk(KERN_ERR "EXT3-fs: get root inode failed\n");
1960 ret = PTR_ERR(root);
1961 goto failed_mount4;
1963 if (!S_ISDIR(root->i_mode) || !root->i_blocks || !root->i_size) {
1964 iput(root);
1965 printk(KERN_ERR "EXT3-fs: corrupt root inode, run e2fsck\n");
1966 goto failed_mount4;
1968 sb->s_root = d_alloc_root(root);
1969 if (!sb->s_root) {
1970 printk(KERN_ERR "EXT3-fs: get root dentry failed\n");
1971 iput(root);
1972 ret = -ENOMEM;
1973 goto failed_mount4;
1976 ext3_setup_super (sb, es, sb->s_flags & MS_RDONLY);
1978 * akpm: core read_super() calls in here with the superblock locked.
1979 * That deadlocks, because orphan cleanup needs to lock the superblock
1980 * in numerous places. Here we just pop the lock - it's relatively
1981 * harmless, because we are now ready to accept write_super() requests,
1982 * and aviro says that's the only reason for hanging onto the
1983 * superblock lock.
1985 EXT3_SB(sb)->s_mount_state |= EXT3_ORPHAN_FS;
1986 ext3_orphan_cleanup(sb, es);
1987 EXT3_SB(sb)->s_mount_state &= ~EXT3_ORPHAN_FS;
1988 if (needs_recovery)
1989 printk (KERN_INFO "EXT3-fs: recovery complete.\n");
1990 ext3_mark_recovery_complete(sb, es);
1991 printk (KERN_INFO "EXT3-fs: mounted filesystem with %s data mode.\n",
1992 test_opt(sb,DATA_FLAGS) == EXT3_MOUNT_JOURNAL_DATA ? "journal":
1993 test_opt(sb,DATA_FLAGS) == EXT3_MOUNT_ORDERED_DATA ? "ordered":
1994 "writeback");
1996 lock_kernel();
1997 return 0;
1999 cantfind_ext3:
2000 if (!silent)
2001 printk(KERN_ERR "VFS: Can't find ext3 filesystem on dev %s.\n",
2002 sb->s_id);
2003 goto failed_mount;
2005 failed_mount4:
2006 journal_destroy(sbi->s_journal);
2007 failed_mount3:
2008 percpu_counter_destroy(&sbi->s_freeblocks_counter);
2009 percpu_counter_destroy(&sbi->s_freeinodes_counter);
2010 percpu_counter_destroy(&sbi->s_dirs_counter);
2011 failed_mount2:
2012 for (i = 0; i < db_count; i++)
2013 brelse(sbi->s_group_desc[i]);
2014 kfree(sbi->s_group_desc);
2015 failed_mount:
2016 #ifdef CONFIG_QUOTA
2017 for (i = 0; i < MAXQUOTAS; i++)
2018 kfree(sbi->s_qf_names[i]);
2019 #endif
2020 ext3_blkdev_remove(sbi);
2021 brelse(bh);
2022 out_fail:
2023 sb->s_fs_info = NULL;
2024 kfree(sbi);
2025 lock_kernel();
2026 return ret;
2030 * Setup any per-fs journal parameters now. We'll do this both on
2031 * initial mount, once the journal has been initialised but before we've
2032 * done any recovery; and again on any subsequent remount.
2034 static void ext3_init_journal_params(struct super_block *sb, journal_t *journal)
2036 struct ext3_sb_info *sbi = EXT3_SB(sb);
2038 if (sbi->s_commit_interval)
2039 journal->j_commit_interval = sbi->s_commit_interval;
2040 /* We could also set up an ext3-specific default for the commit
2041 * interval here, but for now we'll just fall back to the jbd
2042 * default. */
2044 spin_lock(&journal->j_state_lock);
2045 if (test_opt(sb, BARRIER))
2046 journal->j_flags |= JFS_BARRIER;
2047 else
2048 journal->j_flags &= ~JFS_BARRIER;
2049 if (test_opt(sb, DATA_ERR_ABORT))
2050 journal->j_flags |= JFS_ABORT_ON_SYNCDATA_ERR;
2051 else
2052 journal->j_flags &= ~JFS_ABORT_ON_SYNCDATA_ERR;
2053 spin_unlock(&journal->j_state_lock);
2056 static journal_t *ext3_get_journal(struct super_block *sb,
2057 unsigned int journal_inum)
2059 struct inode *journal_inode;
2060 journal_t *journal;
2062 /* First, test for the existence of a valid inode on disk. Bad
2063 * things happen if we iget() an unused inode, as the subsequent
2064 * iput() will try to delete it. */
2066 journal_inode = ext3_iget(sb, journal_inum);
2067 if (IS_ERR(journal_inode)) {
2068 printk(KERN_ERR "EXT3-fs: no journal found.\n");
2069 return NULL;
2071 if (!journal_inode->i_nlink) {
2072 make_bad_inode(journal_inode);
2073 iput(journal_inode);
2074 printk(KERN_ERR "EXT3-fs: journal inode is deleted.\n");
2075 return NULL;
2078 jbd_debug(2, "Journal inode found at %p: %Ld bytes\n",
2079 journal_inode, journal_inode->i_size);
2080 if (!S_ISREG(journal_inode->i_mode)) {
2081 printk(KERN_ERR "EXT3-fs: invalid journal inode.\n");
2082 iput(journal_inode);
2083 return NULL;
2086 journal = journal_init_inode(journal_inode);
2087 if (!journal) {
2088 printk(KERN_ERR "EXT3-fs: Could not load journal inode\n");
2089 iput(journal_inode);
2090 return NULL;
2092 journal->j_private = sb;
2093 ext3_init_journal_params(sb, journal);
2094 return journal;
2097 static journal_t *ext3_get_dev_journal(struct super_block *sb,
2098 dev_t j_dev)
2100 struct buffer_head * bh;
2101 journal_t *journal;
2102 ext3_fsblk_t start;
2103 ext3_fsblk_t len;
2104 int hblock, blocksize;
2105 ext3_fsblk_t sb_block;
2106 unsigned long offset;
2107 struct ext3_super_block * es;
2108 struct block_device *bdev;
2110 bdev = ext3_blkdev_get(j_dev);
2111 if (bdev == NULL)
2112 return NULL;
2114 if (bd_claim(bdev, sb)) {
2115 printk(KERN_ERR
2116 "EXT3: failed to claim external journal device.\n");
2117 blkdev_put(bdev, FMODE_READ|FMODE_WRITE);
2118 return NULL;
2121 blocksize = sb->s_blocksize;
2122 hblock = bdev_hardsect_size(bdev);
2123 if (blocksize < hblock) {
2124 printk(KERN_ERR
2125 "EXT3-fs: blocksize too small for journal device.\n");
2126 goto out_bdev;
2129 sb_block = EXT3_MIN_BLOCK_SIZE / blocksize;
2130 offset = EXT3_MIN_BLOCK_SIZE % blocksize;
2131 set_blocksize(bdev, blocksize);
2132 if (!(bh = __bread(bdev, sb_block, blocksize))) {
2133 printk(KERN_ERR "EXT3-fs: couldn't read superblock of "
2134 "external journal\n");
2135 goto out_bdev;
2138 es = (struct ext3_super_block *) (((char *)bh->b_data) + offset);
2139 if ((le16_to_cpu(es->s_magic) != EXT3_SUPER_MAGIC) ||
2140 !(le32_to_cpu(es->s_feature_incompat) &
2141 EXT3_FEATURE_INCOMPAT_JOURNAL_DEV)) {
2142 printk(KERN_ERR "EXT3-fs: external journal has "
2143 "bad superblock\n");
2144 brelse(bh);
2145 goto out_bdev;
2148 if (memcmp(EXT3_SB(sb)->s_es->s_journal_uuid, es->s_uuid, 16)) {
2149 printk(KERN_ERR "EXT3-fs: journal UUID does not match\n");
2150 brelse(bh);
2151 goto out_bdev;
2154 len = le32_to_cpu(es->s_blocks_count);
2155 start = sb_block + 1;
2156 brelse(bh); /* we're done with the superblock */
2158 journal = journal_init_dev(bdev, sb->s_bdev,
2159 start, len, blocksize);
2160 if (!journal) {
2161 printk(KERN_ERR "EXT3-fs: failed to create device journal\n");
2162 goto out_bdev;
2164 journal->j_private = sb;
2165 ll_rw_block(READ, 1, &journal->j_sb_buffer);
2166 wait_on_buffer(journal->j_sb_buffer);
2167 if (!buffer_uptodate(journal->j_sb_buffer)) {
2168 printk(KERN_ERR "EXT3-fs: I/O error on journal device\n");
2169 goto out_journal;
2171 if (be32_to_cpu(journal->j_superblock->s_nr_users) != 1) {
2172 printk(KERN_ERR "EXT3-fs: External journal has more than one "
2173 "user (unsupported) - %d\n",
2174 be32_to_cpu(journal->j_superblock->s_nr_users));
2175 goto out_journal;
2177 EXT3_SB(sb)->journal_bdev = bdev;
2178 ext3_init_journal_params(sb, journal);
2179 return journal;
2180 out_journal:
2181 journal_destroy(journal);
2182 out_bdev:
2183 ext3_blkdev_put(bdev);
2184 return NULL;
2187 static int ext3_load_journal(struct super_block *sb,
2188 struct ext3_super_block *es,
2189 unsigned long journal_devnum)
2191 journal_t *journal;
2192 unsigned int journal_inum = le32_to_cpu(es->s_journal_inum);
2193 dev_t journal_dev;
2194 int err = 0;
2195 int really_read_only;
2197 if (journal_devnum &&
2198 journal_devnum != le32_to_cpu(es->s_journal_dev)) {
2199 printk(KERN_INFO "EXT3-fs: external journal device major/minor "
2200 "numbers have changed\n");
2201 journal_dev = new_decode_dev(journal_devnum);
2202 } else
2203 journal_dev = new_decode_dev(le32_to_cpu(es->s_journal_dev));
2205 really_read_only = bdev_read_only(sb->s_bdev);
2208 * Are we loading a blank journal or performing recovery after a
2209 * crash? For recovery, we need to check in advance whether we
2210 * can get read-write access to the device.
2213 if (EXT3_HAS_INCOMPAT_FEATURE(sb, EXT3_FEATURE_INCOMPAT_RECOVER)) {
2214 if (sb->s_flags & MS_RDONLY) {
2215 printk(KERN_INFO "EXT3-fs: INFO: recovery "
2216 "required on readonly filesystem.\n");
2217 if (really_read_only) {
2218 printk(KERN_ERR "EXT3-fs: write access "
2219 "unavailable, cannot proceed.\n");
2220 return -EROFS;
2222 printk (KERN_INFO "EXT3-fs: write access will "
2223 "be enabled during recovery.\n");
2227 if (journal_inum && journal_dev) {
2228 printk(KERN_ERR "EXT3-fs: filesystem has both journal "
2229 "and inode journals!\n");
2230 return -EINVAL;
2233 if (journal_inum) {
2234 if (!(journal = ext3_get_journal(sb, journal_inum)))
2235 return -EINVAL;
2236 } else {
2237 if (!(journal = ext3_get_dev_journal(sb, journal_dev)))
2238 return -EINVAL;
2241 if (!really_read_only && test_opt(sb, UPDATE_JOURNAL)) {
2242 err = journal_update_format(journal);
2243 if (err) {
2244 printk(KERN_ERR "EXT3-fs: error updating journal.\n");
2245 journal_destroy(journal);
2246 return err;
2250 if (!EXT3_HAS_INCOMPAT_FEATURE(sb, EXT3_FEATURE_INCOMPAT_RECOVER))
2251 err = journal_wipe(journal, !really_read_only);
2252 if (!err)
2253 err = journal_load(journal);
2255 if (err) {
2256 printk(KERN_ERR "EXT3-fs: error loading journal.\n");
2257 journal_destroy(journal);
2258 return err;
2261 EXT3_SB(sb)->s_journal = journal;
2262 ext3_clear_journal_err(sb, es);
2264 if (journal_devnum &&
2265 journal_devnum != le32_to_cpu(es->s_journal_dev)) {
2266 es->s_journal_dev = cpu_to_le32(journal_devnum);
2267 sb->s_dirt = 1;
2269 /* Make sure we flush the recovery flag to disk. */
2270 ext3_commit_super(sb, es, 1);
2273 return 0;
2276 static int ext3_create_journal(struct super_block * sb,
2277 struct ext3_super_block * es,
2278 unsigned int journal_inum)
2280 journal_t *journal;
2281 int err;
2283 if (sb->s_flags & MS_RDONLY) {
2284 printk(KERN_ERR "EXT3-fs: readonly filesystem when trying to "
2285 "create journal.\n");
2286 return -EROFS;
2289 journal = ext3_get_journal(sb, journal_inum);
2290 if (!journal)
2291 return -EINVAL;
2293 printk(KERN_INFO "EXT3-fs: creating new journal on inode %u\n",
2294 journal_inum);
2296 err = journal_create(journal);
2297 if (err) {
2298 printk(KERN_ERR "EXT3-fs: error creating journal.\n");
2299 journal_destroy(journal);
2300 return -EIO;
2303 EXT3_SB(sb)->s_journal = journal;
2305 ext3_update_dynamic_rev(sb);
2306 EXT3_SET_INCOMPAT_FEATURE(sb, EXT3_FEATURE_INCOMPAT_RECOVER);
2307 EXT3_SET_COMPAT_FEATURE(sb, EXT3_FEATURE_COMPAT_HAS_JOURNAL);
2309 es->s_journal_inum = cpu_to_le32(journal_inum);
2310 sb->s_dirt = 1;
2312 /* Make sure we flush the recovery flag to disk. */
2313 ext3_commit_super(sb, es, 1);
2315 return 0;
2318 static int ext3_commit_super(struct super_block *sb,
2319 struct ext3_super_block *es,
2320 int sync)
2322 struct buffer_head *sbh = EXT3_SB(sb)->s_sbh;
2323 int error = 0;
2325 if (!sbh)
2326 return error;
2327 es->s_wtime = cpu_to_le32(get_seconds());
2328 es->s_free_blocks_count = cpu_to_le32(ext3_count_free_blocks(sb));
2329 es->s_free_inodes_count = cpu_to_le32(ext3_count_free_inodes(sb));
2330 BUFFER_TRACE(sbh, "marking dirty");
2331 mark_buffer_dirty(sbh);
2332 if (sync)
2333 error = sync_dirty_buffer(sbh);
2334 return error;
2339 * Have we just finished recovery? If so, and if we are mounting (or
2340 * remounting) the filesystem readonly, then we will end up with a
2341 * consistent fs on disk. Record that fact.
2343 static void ext3_mark_recovery_complete(struct super_block * sb,
2344 struct ext3_super_block * es)
2346 journal_t *journal = EXT3_SB(sb)->s_journal;
2348 journal_lock_updates(journal);
2349 if (journal_flush(journal) < 0)
2350 goto out;
2352 lock_super(sb);
2353 if (EXT3_HAS_INCOMPAT_FEATURE(sb, EXT3_FEATURE_INCOMPAT_RECOVER) &&
2354 sb->s_flags & MS_RDONLY) {
2355 EXT3_CLEAR_INCOMPAT_FEATURE(sb, EXT3_FEATURE_INCOMPAT_RECOVER);
2356 sb->s_dirt = 0;
2357 ext3_commit_super(sb, es, 1);
2359 unlock_super(sb);
2361 out:
2362 journal_unlock_updates(journal);
2366 * If we are mounting (or read-write remounting) a filesystem whose journal
2367 * has recorded an error from a previous lifetime, move that error to the
2368 * main filesystem now.
2370 static void ext3_clear_journal_err(struct super_block * sb,
2371 struct ext3_super_block * es)
2373 journal_t *journal;
2374 int j_errno;
2375 const char *errstr;
2377 journal = EXT3_SB(sb)->s_journal;
2380 * Now check for any error status which may have been recorded in the
2381 * journal by a prior ext3_error() or ext3_abort()
2384 j_errno = journal_errno(journal);
2385 if (j_errno) {
2386 char nbuf[16];
2388 errstr = ext3_decode_error(sb, j_errno, nbuf);
2389 ext3_warning(sb, __func__, "Filesystem error recorded "
2390 "from previous mount: %s", errstr);
2391 ext3_warning(sb, __func__, "Marking fs in need of "
2392 "filesystem check.");
2394 EXT3_SB(sb)->s_mount_state |= EXT3_ERROR_FS;
2395 es->s_state |= cpu_to_le16(EXT3_ERROR_FS);
2396 ext3_commit_super (sb, es, 1);
2398 journal_clear_err(journal);
2403 * Force the running and committing transactions to commit,
2404 * and wait on the commit.
2406 int ext3_force_commit(struct super_block *sb)
2408 journal_t *journal;
2409 int ret;
2411 if (sb->s_flags & MS_RDONLY)
2412 return 0;
2414 journal = EXT3_SB(sb)->s_journal;
2415 sb->s_dirt = 0;
2416 ret = ext3_journal_force_commit(journal);
2417 return ret;
2421 * Ext3 always journals updates to the superblock itself, so we don't
2422 * have to propagate any other updates to the superblock on disk at this
2423 * point. (We can probably nuke this function altogether, and remove
2424 * any mention to sb->s_dirt in all of fs/ext3; eventual cleanup...)
2426 static void ext3_write_super (struct super_block * sb)
2428 if (mutex_trylock(&sb->s_lock) != 0)
2429 BUG();
2430 sb->s_dirt = 0;
2433 static int ext3_sync_fs(struct super_block *sb, int wait)
2435 tid_t target;
2437 sb->s_dirt = 0;
2438 if (journal_start_commit(EXT3_SB(sb)->s_journal, &target)) {
2439 if (wait)
2440 log_wait_commit(EXT3_SB(sb)->s_journal, target);
2442 return 0;
2446 * LVM calls this function before a (read-only) snapshot is created. This
2447 * gives us a chance to flush the journal completely and mark the fs clean.
2449 static int ext3_freeze(struct super_block *sb)
2451 int error = 0;
2452 journal_t *journal;
2453 sb->s_dirt = 0;
2455 if (!(sb->s_flags & MS_RDONLY)) {
2456 journal = EXT3_SB(sb)->s_journal;
2458 /* Now we set up the journal barrier. */
2459 journal_lock_updates(journal);
2462 * We don't want to clear needs_recovery flag when we failed
2463 * to flush the journal.
2465 error = journal_flush(journal);
2466 if (error < 0)
2467 goto out;
2469 /* Journal blocked and flushed, clear needs_recovery flag. */
2470 EXT3_CLEAR_INCOMPAT_FEATURE(sb, EXT3_FEATURE_INCOMPAT_RECOVER);
2471 error = ext3_commit_super(sb, EXT3_SB(sb)->s_es, 1);
2472 if (error)
2473 goto out;
2475 return 0;
2477 out:
2478 journal_unlock_updates(journal);
2479 return error;
2483 * Called by LVM after the snapshot is done. We need to reset the RECOVER
2484 * flag here, even though the filesystem is not technically dirty yet.
2486 static int ext3_unfreeze(struct super_block *sb)
2488 if (!(sb->s_flags & MS_RDONLY)) {
2489 lock_super(sb);
2490 /* Reser the needs_recovery flag before the fs is unlocked. */
2491 EXT3_SET_INCOMPAT_FEATURE(sb, EXT3_FEATURE_INCOMPAT_RECOVER);
2492 ext3_commit_super(sb, EXT3_SB(sb)->s_es, 1);
2493 unlock_super(sb);
2494 journal_unlock_updates(EXT3_SB(sb)->s_journal);
2496 return 0;
2499 static int ext3_remount (struct super_block * sb, int * flags, char * data)
2501 struct ext3_super_block * es;
2502 struct ext3_sb_info *sbi = EXT3_SB(sb);
2503 ext3_fsblk_t n_blocks_count = 0;
2504 unsigned long old_sb_flags;
2505 struct ext3_mount_options old_opts;
2506 int err;
2507 #ifdef CONFIG_QUOTA
2508 int i;
2509 #endif
2511 /* Store the original options */
2512 old_sb_flags = sb->s_flags;
2513 old_opts.s_mount_opt = sbi->s_mount_opt;
2514 old_opts.s_resuid = sbi->s_resuid;
2515 old_opts.s_resgid = sbi->s_resgid;
2516 old_opts.s_commit_interval = sbi->s_commit_interval;
2517 #ifdef CONFIG_QUOTA
2518 old_opts.s_jquota_fmt = sbi->s_jquota_fmt;
2519 for (i = 0; i < MAXQUOTAS; i++)
2520 old_opts.s_qf_names[i] = sbi->s_qf_names[i];
2521 #endif
2524 * Allow the "check" option to be passed as a remount option.
2526 if (!parse_options(data, sb, NULL, NULL, &n_blocks_count, 1)) {
2527 err = -EINVAL;
2528 goto restore_opts;
2531 if (sbi->s_mount_opt & EXT3_MOUNT_ABORT)
2532 ext3_abort(sb, __func__, "Abort forced by user");
2534 sb->s_flags = (sb->s_flags & ~MS_POSIXACL) |
2535 ((sbi->s_mount_opt & EXT3_MOUNT_POSIX_ACL) ? MS_POSIXACL : 0);
2537 es = sbi->s_es;
2539 ext3_init_journal_params(sb, sbi->s_journal);
2541 if ((*flags & MS_RDONLY) != (sb->s_flags & MS_RDONLY) ||
2542 n_blocks_count > le32_to_cpu(es->s_blocks_count)) {
2543 if (sbi->s_mount_opt & EXT3_MOUNT_ABORT) {
2544 err = -EROFS;
2545 goto restore_opts;
2548 if (*flags & MS_RDONLY) {
2550 * First of all, the unconditional stuff we have to do
2551 * to disable replay of the journal when we next remount
2553 sb->s_flags |= MS_RDONLY;
2556 * OK, test if we are remounting a valid rw partition
2557 * readonly, and if so set the rdonly flag and then
2558 * mark the partition as valid again.
2560 if (!(es->s_state & cpu_to_le16(EXT3_VALID_FS)) &&
2561 (sbi->s_mount_state & EXT3_VALID_FS))
2562 es->s_state = cpu_to_le16(sbi->s_mount_state);
2565 * We have to unlock super so that we can wait for
2566 * transactions.
2568 unlock_super(sb);
2569 ext3_mark_recovery_complete(sb, es);
2570 lock_super(sb);
2571 } else {
2572 __le32 ret;
2573 if ((ret = EXT3_HAS_RO_COMPAT_FEATURE(sb,
2574 ~EXT3_FEATURE_RO_COMPAT_SUPP))) {
2575 printk(KERN_WARNING "EXT3-fs: %s: couldn't "
2576 "remount RDWR because of unsupported "
2577 "optional features (%x).\n",
2578 sb->s_id, le32_to_cpu(ret));
2579 err = -EROFS;
2580 goto restore_opts;
2584 * If we have an unprocessed orphan list hanging
2585 * around from a previously readonly bdev mount,
2586 * require a full umount/remount for now.
2588 if (es->s_last_orphan) {
2589 printk(KERN_WARNING "EXT3-fs: %s: couldn't "
2590 "remount RDWR because of unprocessed "
2591 "orphan inode list. Please "
2592 "umount/remount instead.\n",
2593 sb->s_id);
2594 err = -EINVAL;
2595 goto restore_opts;
2599 * Mounting a RDONLY partition read-write, so reread
2600 * and store the current valid flag. (It may have
2601 * been changed by e2fsck since we originally mounted
2602 * the partition.)
2604 ext3_clear_journal_err(sb, es);
2605 sbi->s_mount_state = le16_to_cpu(es->s_state);
2606 if ((err = ext3_group_extend(sb, es, n_blocks_count)))
2607 goto restore_opts;
2608 if (!ext3_setup_super (sb, es, 0))
2609 sb->s_flags &= ~MS_RDONLY;
2612 #ifdef CONFIG_QUOTA
2613 /* Release old quota file names */
2614 for (i = 0; i < MAXQUOTAS; i++)
2615 if (old_opts.s_qf_names[i] &&
2616 old_opts.s_qf_names[i] != sbi->s_qf_names[i])
2617 kfree(old_opts.s_qf_names[i]);
2618 #endif
2619 return 0;
2620 restore_opts:
2621 sb->s_flags = old_sb_flags;
2622 sbi->s_mount_opt = old_opts.s_mount_opt;
2623 sbi->s_resuid = old_opts.s_resuid;
2624 sbi->s_resgid = old_opts.s_resgid;
2625 sbi->s_commit_interval = old_opts.s_commit_interval;
2626 #ifdef CONFIG_QUOTA
2627 sbi->s_jquota_fmt = old_opts.s_jquota_fmt;
2628 for (i = 0; i < MAXQUOTAS; i++) {
2629 if (sbi->s_qf_names[i] &&
2630 old_opts.s_qf_names[i] != sbi->s_qf_names[i])
2631 kfree(sbi->s_qf_names[i]);
2632 sbi->s_qf_names[i] = old_opts.s_qf_names[i];
2634 #endif
2635 return err;
2638 static int ext3_statfs (struct dentry * dentry, struct kstatfs * buf)
2640 struct super_block *sb = dentry->d_sb;
2641 struct ext3_sb_info *sbi = EXT3_SB(sb);
2642 struct ext3_super_block *es = sbi->s_es;
2643 u64 fsid;
2645 if (test_opt(sb, MINIX_DF)) {
2646 sbi->s_overhead_last = 0;
2647 } else if (sbi->s_blocks_last != le32_to_cpu(es->s_blocks_count)) {
2648 unsigned long ngroups = sbi->s_groups_count, i;
2649 ext3_fsblk_t overhead = 0;
2650 smp_rmb();
2653 * Compute the overhead (FS structures). This is constant
2654 * for a given filesystem unless the number of block groups
2655 * changes so we cache the previous value until it does.
2659 * All of the blocks before first_data_block are
2660 * overhead
2662 overhead = le32_to_cpu(es->s_first_data_block);
2665 * Add the overhead attributed to the superblock and
2666 * block group descriptors. If the sparse superblocks
2667 * feature is turned on, then not all groups have this.
2669 for (i = 0; i < ngroups; i++) {
2670 overhead += ext3_bg_has_super(sb, i) +
2671 ext3_bg_num_gdb(sb, i);
2672 cond_resched();
2676 * Every block group has an inode bitmap, a block
2677 * bitmap, and an inode table.
2679 overhead += ngroups * (2 + sbi->s_itb_per_group);
2680 sbi->s_overhead_last = overhead;
2681 smp_wmb();
2682 sbi->s_blocks_last = le32_to_cpu(es->s_blocks_count);
2685 buf->f_type = EXT3_SUPER_MAGIC;
2686 buf->f_bsize = sb->s_blocksize;
2687 buf->f_blocks = le32_to_cpu(es->s_blocks_count) - sbi->s_overhead_last;
2688 buf->f_bfree = percpu_counter_sum_positive(&sbi->s_freeblocks_counter);
2689 es->s_free_blocks_count = cpu_to_le32(buf->f_bfree);
2690 buf->f_bavail = buf->f_bfree - le32_to_cpu(es->s_r_blocks_count);
2691 if (buf->f_bfree < le32_to_cpu(es->s_r_blocks_count))
2692 buf->f_bavail = 0;
2693 buf->f_files = le32_to_cpu(es->s_inodes_count);
2694 buf->f_ffree = percpu_counter_sum_positive(&sbi->s_freeinodes_counter);
2695 es->s_free_inodes_count = cpu_to_le32(buf->f_ffree);
2696 buf->f_namelen = EXT3_NAME_LEN;
2697 fsid = le64_to_cpup((void *)es->s_uuid) ^
2698 le64_to_cpup((void *)es->s_uuid + sizeof(u64));
2699 buf->f_fsid.val[0] = fsid & 0xFFFFFFFFUL;
2700 buf->f_fsid.val[1] = (fsid >> 32) & 0xFFFFFFFFUL;
2701 return 0;
2704 /* Helper function for writing quotas on sync - we need to start transaction before quota file
2705 * is locked for write. Otherwise the are possible deadlocks:
2706 * Process 1 Process 2
2707 * ext3_create() quota_sync()
2708 * journal_start() write_dquot()
2709 * vfs_dq_init() down(dqio_mutex)
2710 * down(dqio_mutex) journal_start()
2714 #ifdef CONFIG_QUOTA
2716 static inline struct inode *dquot_to_inode(struct dquot *dquot)
2718 return sb_dqopt(dquot->dq_sb)->files[dquot->dq_type];
2721 static int ext3_write_dquot(struct dquot *dquot)
2723 int ret, err;
2724 handle_t *handle;
2725 struct inode *inode;
2727 inode = dquot_to_inode(dquot);
2728 handle = ext3_journal_start(inode,
2729 EXT3_QUOTA_TRANS_BLOCKS(dquot->dq_sb));
2730 if (IS_ERR(handle))
2731 return PTR_ERR(handle);
2732 ret = dquot_commit(dquot);
2733 err = ext3_journal_stop(handle);
2734 if (!ret)
2735 ret = err;
2736 return ret;
2739 static int ext3_acquire_dquot(struct dquot *dquot)
2741 int ret, err;
2742 handle_t *handle;
2744 handle = ext3_journal_start(dquot_to_inode(dquot),
2745 EXT3_QUOTA_INIT_BLOCKS(dquot->dq_sb));
2746 if (IS_ERR(handle))
2747 return PTR_ERR(handle);
2748 ret = dquot_acquire(dquot);
2749 err = ext3_journal_stop(handle);
2750 if (!ret)
2751 ret = err;
2752 return ret;
2755 static int ext3_release_dquot(struct dquot *dquot)
2757 int ret, err;
2758 handle_t *handle;
2760 handle = ext3_journal_start(dquot_to_inode(dquot),
2761 EXT3_QUOTA_DEL_BLOCKS(dquot->dq_sb));
2762 if (IS_ERR(handle)) {
2763 /* Release dquot anyway to avoid endless cycle in dqput() */
2764 dquot_release(dquot);
2765 return PTR_ERR(handle);
2767 ret = dquot_release(dquot);
2768 err = ext3_journal_stop(handle);
2769 if (!ret)
2770 ret = err;
2771 return ret;
2774 static int ext3_mark_dquot_dirty(struct dquot *dquot)
2776 /* Are we journaling quotas? */
2777 if (EXT3_SB(dquot->dq_sb)->s_qf_names[USRQUOTA] ||
2778 EXT3_SB(dquot->dq_sb)->s_qf_names[GRPQUOTA]) {
2779 dquot_mark_dquot_dirty(dquot);
2780 return ext3_write_dquot(dquot);
2781 } else {
2782 return dquot_mark_dquot_dirty(dquot);
2786 static int ext3_write_info(struct super_block *sb, int type)
2788 int ret, err;
2789 handle_t *handle;
2791 /* Data block + inode block */
2792 handle = ext3_journal_start(sb->s_root->d_inode, 2);
2793 if (IS_ERR(handle))
2794 return PTR_ERR(handle);
2795 ret = dquot_commit_info(sb, type);
2796 err = ext3_journal_stop(handle);
2797 if (!ret)
2798 ret = err;
2799 return ret;
2803 * Turn on quotas during mount time - we need to find
2804 * the quota file and such...
2806 static int ext3_quota_on_mount(struct super_block *sb, int type)
2808 return vfs_quota_on_mount(sb, EXT3_SB(sb)->s_qf_names[type],
2809 EXT3_SB(sb)->s_jquota_fmt, type);
2813 * Standard function to be called on quota_on
2815 static int ext3_quota_on(struct super_block *sb, int type, int format_id,
2816 char *name, int remount)
2818 int err;
2819 struct path path;
2821 if (!test_opt(sb, QUOTA))
2822 return -EINVAL;
2823 /* When remounting, no checks are needed and in fact, name is NULL */
2824 if (remount)
2825 return vfs_quota_on(sb, type, format_id, name, remount);
2827 err = kern_path(name, LOOKUP_FOLLOW, &path);
2828 if (err)
2829 return err;
2831 /* Quotafile not on the same filesystem? */
2832 if (path.mnt->mnt_sb != sb) {
2833 path_put(&path);
2834 return -EXDEV;
2836 /* Journaling quota? */
2837 if (EXT3_SB(sb)->s_qf_names[type]) {
2838 /* Quotafile not of fs root? */
2839 if (path.dentry->d_parent != sb->s_root)
2840 printk(KERN_WARNING
2841 "EXT3-fs: Quota file not on filesystem root. "
2842 "Journaled quota will not work.\n");
2846 * When we journal data on quota file, we have to flush journal to see
2847 * all updates to the file when we bypass pagecache...
2849 if (ext3_should_journal_data(path.dentry->d_inode)) {
2851 * We don't need to lock updates but journal_flush() could
2852 * otherwise be livelocked...
2854 journal_lock_updates(EXT3_SB(sb)->s_journal);
2855 err = journal_flush(EXT3_SB(sb)->s_journal);
2856 journal_unlock_updates(EXT3_SB(sb)->s_journal);
2857 if (err) {
2858 path_put(&path);
2859 return err;
2863 err = vfs_quota_on_path(sb, type, format_id, &path);
2864 path_put(&path);
2865 return err;
2868 /* Read data from quotafile - avoid pagecache and such because we cannot afford
2869 * acquiring the locks... As quota files are never truncated and quota code
2870 * itself serializes the operations (and noone else should touch the files)
2871 * we don't have to be afraid of races */
2872 static ssize_t ext3_quota_read(struct super_block *sb, int type, char *data,
2873 size_t len, loff_t off)
2875 struct inode *inode = sb_dqopt(sb)->files[type];
2876 sector_t blk = off >> EXT3_BLOCK_SIZE_BITS(sb);
2877 int err = 0;
2878 int offset = off & (sb->s_blocksize - 1);
2879 int tocopy;
2880 size_t toread;
2881 struct buffer_head *bh;
2882 loff_t i_size = i_size_read(inode);
2884 if (off > i_size)
2885 return 0;
2886 if (off+len > i_size)
2887 len = i_size-off;
2888 toread = len;
2889 while (toread > 0) {
2890 tocopy = sb->s_blocksize - offset < toread ?
2891 sb->s_blocksize - offset : toread;
2892 bh = ext3_bread(NULL, inode, blk, 0, &err);
2893 if (err)
2894 return err;
2895 if (!bh) /* A hole? */
2896 memset(data, 0, tocopy);
2897 else
2898 memcpy(data, bh->b_data+offset, tocopy);
2899 brelse(bh);
2900 offset = 0;
2901 toread -= tocopy;
2902 data += tocopy;
2903 blk++;
2905 return len;
2908 /* Write to quotafile (we know the transaction is already started and has
2909 * enough credits) */
2910 static ssize_t ext3_quota_write(struct super_block *sb, int type,
2911 const char *data, size_t len, loff_t off)
2913 struct inode *inode = sb_dqopt(sb)->files[type];
2914 sector_t blk = off >> EXT3_BLOCK_SIZE_BITS(sb);
2915 int err = 0;
2916 int offset = off & (sb->s_blocksize - 1);
2917 int tocopy;
2918 int journal_quota = EXT3_SB(sb)->s_qf_names[type] != NULL;
2919 size_t towrite = len;
2920 struct buffer_head *bh;
2921 handle_t *handle = journal_current_handle();
2923 if (!handle) {
2924 printk(KERN_WARNING "EXT3-fs: Quota write (off=%Lu, len=%Lu)"
2925 " cancelled because transaction is not started.\n",
2926 (unsigned long long)off, (unsigned long long)len);
2927 return -EIO;
2929 mutex_lock_nested(&inode->i_mutex, I_MUTEX_QUOTA);
2930 while (towrite > 0) {
2931 tocopy = sb->s_blocksize - offset < towrite ?
2932 sb->s_blocksize - offset : towrite;
2933 bh = ext3_bread(handle, inode, blk, 1, &err);
2934 if (!bh)
2935 goto out;
2936 if (journal_quota) {
2937 err = ext3_journal_get_write_access(handle, bh);
2938 if (err) {
2939 brelse(bh);
2940 goto out;
2943 lock_buffer(bh);
2944 memcpy(bh->b_data+offset, data, tocopy);
2945 flush_dcache_page(bh->b_page);
2946 unlock_buffer(bh);
2947 if (journal_quota)
2948 err = ext3_journal_dirty_metadata(handle, bh);
2949 else {
2950 /* Always do at least ordered writes for quotas */
2951 err = ext3_journal_dirty_data(handle, bh);
2952 mark_buffer_dirty(bh);
2954 brelse(bh);
2955 if (err)
2956 goto out;
2957 offset = 0;
2958 towrite -= tocopy;
2959 data += tocopy;
2960 blk++;
2962 out:
2963 if (len == towrite) {
2964 mutex_unlock(&inode->i_mutex);
2965 return err;
2967 if (inode->i_size < off+len-towrite) {
2968 i_size_write(inode, off+len-towrite);
2969 EXT3_I(inode)->i_disksize = inode->i_size;
2971 inode->i_version++;
2972 inode->i_mtime = inode->i_ctime = CURRENT_TIME;
2973 ext3_mark_inode_dirty(handle, inode);
2974 mutex_unlock(&inode->i_mutex);
2975 return len - towrite;
2978 #endif
2980 static int ext3_get_sb(struct file_system_type *fs_type,
2981 int flags, const char *dev_name, void *data, struct vfsmount *mnt)
2983 return get_sb_bdev(fs_type, flags, dev_name, data, ext3_fill_super, mnt);
2986 static struct file_system_type ext3_fs_type = {
2987 .owner = THIS_MODULE,
2988 .name = "ext3",
2989 .get_sb = ext3_get_sb,
2990 .kill_sb = kill_block_super,
2991 .fs_flags = FS_REQUIRES_DEV,
2994 static int __init init_ext3_fs(void)
2996 int err = init_ext3_xattr();
2997 if (err)
2998 return err;
2999 err = init_inodecache();
3000 if (err)
3001 goto out1;
3002 err = register_filesystem(&ext3_fs_type);
3003 if (err)
3004 goto out;
3005 return 0;
3006 out:
3007 destroy_inodecache();
3008 out1:
3009 exit_ext3_xattr();
3010 return err;
3013 static void __exit exit_ext3_fs(void)
3015 unregister_filesystem(&ext3_fs_type);
3016 destroy_inodecache();
3017 exit_ext3_xattr();
3020 MODULE_AUTHOR("Remy Card, Stephen Tweedie, Andrew Morton, Andreas Dilger, Theodore Ts'o and others");
3021 MODULE_DESCRIPTION("Second Extended Filesystem with journaling extensions");
3022 MODULE_LICENSE("GPL");
3023 module_init(init_ext3_fs)
3024 module_exit(exit_ext3_fs)