BKL: remove extraneous #include <smp_lock.h>
[linux-2.6/linux-acpi-2.6/ibm-acpi-2.6.git] / fs / ext3 / super.c
blobacf8695fa8f035935390dd6bc0c8e44e8d3e5961
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/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/log2.h>
40 #include <asm/uaccess.h>
42 #include "xattr.h"
43 #include "acl.h"
44 #include "namei.h"
46 #ifdef CONFIG_EXT3_DEFAULTS_TO_ORDERED
47 #define EXT3_MOUNT_DEFAULT_DATA_MODE EXT3_MOUNT_ORDERED_DATA
48 #else
49 #define EXT3_MOUNT_DEFAULT_DATA_MODE EXT3_MOUNT_WRITEBACK_DATA
50 #endif
52 static int ext3_load_journal(struct super_block *, struct ext3_super_block *,
53 unsigned long journal_devnum);
54 static int ext3_create_journal(struct super_block *, struct ext3_super_block *,
55 unsigned int);
56 static int ext3_commit_super(struct super_block *sb,
57 struct ext3_super_block *es,
58 int sync);
59 static void ext3_mark_recovery_complete(struct super_block * sb,
60 struct ext3_super_block * es);
61 static void ext3_clear_journal_err(struct super_block * sb,
62 struct ext3_super_block * es);
63 static int ext3_sync_fs(struct super_block *sb, int wait);
64 static const char *ext3_decode_error(struct super_block * sb, int errno,
65 char nbuf[16]);
66 static int ext3_remount (struct super_block * sb, int * flags, char * data);
67 static int ext3_statfs (struct dentry * dentry, struct kstatfs * buf);
68 static int ext3_unfreeze(struct super_block *sb);
69 static int ext3_freeze(struct super_block *sb);
72 * Wrappers for journal_start/end.
74 * The only special thing we need to do here is to make sure that all
75 * journal_end calls result in the superblock being marked dirty, so
76 * that sync() will call the filesystem's write_super callback if
77 * appropriate.
79 handle_t *ext3_journal_start_sb(struct super_block *sb, int nblocks)
81 journal_t *journal;
83 if (sb->s_flags & MS_RDONLY)
84 return ERR_PTR(-EROFS);
86 /* Special case here: if the journal has aborted behind our
87 * backs (eg. EIO in the commit thread), then we still need to
88 * take the FS itself readonly cleanly. */
89 journal = EXT3_SB(sb)->s_journal;
90 if (is_journal_aborted(journal)) {
91 ext3_abort(sb, __func__,
92 "Detected aborted journal");
93 return ERR_PTR(-EROFS);
96 return journal_start(journal, nblocks);
100 * The only special thing we need to do here is to make sure that all
101 * journal_stop calls result in the superblock being marked dirty, so
102 * that sync() will call the filesystem's write_super callback if
103 * appropriate.
105 int __ext3_journal_stop(const char *where, handle_t *handle)
107 struct super_block *sb;
108 int err;
109 int rc;
111 sb = handle->h_transaction->t_journal->j_private;
112 err = handle->h_err;
113 rc = journal_stop(handle);
115 if (!err)
116 err = rc;
117 if (err)
118 __ext3_std_error(sb, where, err);
119 return err;
122 void ext3_journal_abort_handle(const char *caller, const char *err_fn,
123 struct buffer_head *bh, handle_t *handle, int err)
125 char nbuf[16];
126 const char *errstr = ext3_decode_error(NULL, err, nbuf);
128 if (bh)
129 BUFFER_TRACE(bh, "abort");
131 if (!handle->h_err)
132 handle->h_err = err;
134 if (is_handle_aborted(handle))
135 return;
137 printk(KERN_ERR "EXT3-fs: %s: aborting transaction: %s in %s\n",
138 caller, errstr, err_fn);
140 journal_abort_handle(handle);
143 void ext3_msg(struct super_block *sb, const char *prefix,
144 const char *fmt, ...)
146 va_list args;
148 va_start(args, fmt);
149 printk("%sEXT3-fs (%s): ", prefix, sb->s_id);
150 vprintk(fmt, args);
151 printk("\n");
152 va_end(args);
155 /* Deal with the reporting of failure conditions on a filesystem such as
156 * inconsistencies detected or read IO failures.
158 * On ext2, we can store the error state of the filesystem in the
159 * superblock. That is not possible on ext3, because we may have other
160 * write ordering constraints on the superblock which prevent us from
161 * writing it out straight away; and given that the journal is about to
162 * be aborted, we can't rely on the current, or future, transactions to
163 * write out the superblock safely.
165 * We'll just use the journal_abort() error code to record an error in
166 * the journal instead. On recovery, the journal will complain about
167 * that error until we've noted it down and cleared it.
170 static void ext3_handle_error(struct super_block *sb)
172 struct ext3_super_block *es = EXT3_SB(sb)->s_es;
174 EXT3_SB(sb)->s_mount_state |= EXT3_ERROR_FS;
175 es->s_state |= cpu_to_le16(EXT3_ERROR_FS);
177 if (sb->s_flags & MS_RDONLY)
178 return;
180 if (!test_opt (sb, ERRORS_CONT)) {
181 journal_t *journal = EXT3_SB(sb)->s_journal;
183 set_opt(EXT3_SB(sb)->s_mount_opt, ABORT);
184 if (journal)
185 journal_abort(journal, -EIO);
187 if (test_opt (sb, ERRORS_RO)) {
188 ext3_msg(sb, KERN_CRIT,
189 "error: remounting filesystem read-only");
190 sb->s_flags |= MS_RDONLY;
192 ext3_commit_super(sb, es, 1);
193 if (test_opt(sb, ERRORS_PANIC))
194 panic("EXT3-fs (%s): panic forced after error\n",
195 sb->s_id);
198 void ext3_error (struct super_block * sb, const char * function,
199 const char * fmt, ...)
201 va_list args;
203 va_start(args, fmt);
204 printk(KERN_CRIT "EXT3-fs error (device %s): %s: ",sb->s_id, function);
205 vprintk(fmt, args);
206 printk("\n");
207 va_end(args);
209 ext3_handle_error(sb);
212 static const char *ext3_decode_error(struct super_block * sb, int errno,
213 char nbuf[16])
215 char *errstr = NULL;
217 switch (errno) {
218 case -EIO:
219 errstr = "IO failure";
220 break;
221 case -ENOMEM:
222 errstr = "Out of memory";
223 break;
224 case -EROFS:
225 if (!sb || EXT3_SB(sb)->s_journal->j_flags & JFS_ABORT)
226 errstr = "Journal has aborted";
227 else
228 errstr = "Readonly filesystem";
229 break;
230 default:
231 /* If the caller passed in an extra buffer for unknown
232 * errors, textualise them now. Else we just return
233 * NULL. */
234 if (nbuf) {
235 /* Check for truncated error codes... */
236 if (snprintf(nbuf, 16, "error %d", -errno) >= 0)
237 errstr = nbuf;
239 break;
242 return errstr;
245 /* __ext3_std_error decodes expected errors from journaling functions
246 * automatically and invokes the appropriate error response. */
248 void __ext3_std_error (struct super_block * sb, const char * function,
249 int errno)
251 char nbuf[16];
252 const char *errstr;
254 /* Special case: if the error is EROFS, and we're not already
255 * inside a transaction, then there's really no point in logging
256 * an error. */
257 if (errno == -EROFS && journal_current_handle() == NULL &&
258 (sb->s_flags & MS_RDONLY))
259 return;
261 errstr = ext3_decode_error(sb, errno, nbuf);
262 ext3_msg(sb, KERN_CRIT, "error in %s: %s", function, errstr);
264 ext3_handle_error(sb);
268 * ext3_abort is a much stronger failure handler than ext3_error. The
269 * abort function may be used to deal with unrecoverable failures such
270 * as journal IO errors or ENOMEM at a critical moment in log management.
272 * We unconditionally force the filesystem into an ABORT|READONLY state,
273 * unless the error response on the fs has been set to panic in which
274 * case we take the easy way out and panic immediately.
277 void ext3_abort (struct super_block * sb, const char * function,
278 const char * fmt, ...)
280 va_list args;
282 va_start(args, fmt);
283 printk(KERN_CRIT "EXT3-fs (%s): error: %s: ", sb->s_id, function);
284 vprintk(fmt, args);
285 printk("\n");
286 va_end(args);
288 if (test_opt(sb, ERRORS_PANIC))
289 panic("EXT3-fs: panic from previous error\n");
291 if (sb->s_flags & MS_RDONLY)
292 return;
294 ext3_msg(sb, KERN_CRIT,
295 "error: remounting filesystem read-only");
296 EXT3_SB(sb)->s_mount_state |= EXT3_ERROR_FS;
297 sb->s_flags |= MS_RDONLY;
298 set_opt(EXT3_SB(sb)->s_mount_opt, ABORT);
299 if (EXT3_SB(sb)->s_journal)
300 journal_abort(EXT3_SB(sb)->s_journal, -EIO);
303 void ext3_warning (struct super_block * sb, const char * function,
304 const char * fmt, ...)
306 va_list args;
308 va_start(args, fmt);
309 printk(KERN_WARNING "EXT3-fs (%s): warning: %s: ",
310 sb->s_id, function);
311 vprintk(fmt, args);
312 printk("\n");
313 va_end(args);
316 void ext3_update_dynamic_rev(struct super_block *sb)
318 struct ext3_super_block *es = EXT3_SB(sb)->s_es;
320 if (le32_to_cpu(es->s_rev_level) > EXT3_GOOD_OLD_REV)
321 return;
323 ext3_msg(sb, KERN_WARNING,
324 "warning: updating to rev %d because of "
325 "new feature flag, running e2fsck is recommended",
326 EXT3_DYNAMIC_REV);
328 es->s_first_ino = cpu_to_le32(EXT3_GOOD_OLD_FIRST_INO);
329 es->s_inode_size = cpu_to_le16(EXT3_GOOD_OLD_INODE_SIZE);
330 es->s_rev_level = cpu_to_le32(EXT3_DYNAMIC_REV);
331 /* leave es->s_feature_*compat flags alone */
332 /* es->s_uuid will be set by e2fsck if empty */
335 * The rest of the superblock fields should be zero, and if not it
336 * means they are likely already in use, so leave them alone. We
337 * can leave it up to e2fsck to clean up any inconsistencies there.
342 * Open the external journal device
344 static struct block_device *ext3_blkdev_get(dev_t dev, struct super_block *sb)
346 struct block_device *bdev;
347 char b[BDEVNAME_SIZE];
349 bdev = open_by_devnum(dev, FMODE_READ|FMODE_WRITE);
350 if (IS_ERR(bdev))
351 goto fail;
352 return bdev;
354 fail:
355 ext3_msg(sb, "error: failed to open journal device %s: %ld",
356 __bdevname(dev, b), PTR_ERR(bdev));
358 return NULL;
362 * Release the journal device
364 static int ext3_blkdev_put(struct block_device *bdev)
366 bd_release(bdev);
367 return blkdev_put(bdev, FMODE_READ|FMODE_WRITE);
370 static int ext3_blkdev_remove(struct ext3_sb_info *sbi)
372 struct block_device *bdev;
373 int ret = -ENODEV;
375 bdev = sbi->journal_bdev;
376 if (bdev) {
377 ret = ext3_blkdev_put(bdev);
378 sbi->journal_bdev = NULL;
380 return ret;
383 static inline struct inode *orphan_list_entry(struct list_head *l)
385 return &list_entry(l, struct ext3_inode_info, i_orphan)->vfs_inode;
388 static void dump_orphan_list(struct super_block *sb, struct ext3_sb_info *sbi)
390 struct list_head *l;
392 ext3_msg(sb, KERN_ERR, "error: sb orphan head is %d",
393 le32_to_cpu(sbi->s_es->s_last_orphan));
395 ext3_msg(sb, KERN_ERR, "sb_info orphan list:");
396 list_for_each(l, &sbi->s_orphan) {
397 struct inode *inode = orphan_list_entry(l);
398 ext3_msg(sb, KERN_ERR, " "
399 "inode %s:%lu at %p: mode %o, nlink %d, next %d\n",
400 inode->i_sb->s_id, inode->i_ino, inode,
401 inode->i_mode, inode->i_nlink,
402 NEXT_ORPHAN(inode));
406 static void ext3_put_super (struct super_block * sb)
408 struct ext3_sb_info *sbi = EXT3_SB(sb);
409 struct ext3_super_block *es = sbi->s_es;
410 int i, err;
412 dquot_disable(sb, -1, DQUOT_USAGE_ENABLED | DQUOT_LIMITS_ENABLED);
413 ext3_xattr_put_super(sb);
414 err = journal_destroy(sbi->s_journal);
415 sbi->s_journal = NULL;
416 if (err < 0)
417 ext3_abort(sb, __func__, "Couldn't clean up the journal");
419 if (!(sb->s_flags & MS_RDONLY)) {
420 EXT3_CLEAR_INCOMPAT_FEATURE(sb, EXT3_FEATURE_INCOMPAT_RECOVER);
421 es->s_state = cpu_to_le16(sbi->s_mount_state);
422 BUFFER_TRACE(sbi->s_sbh, "marking dirty");
423 mark_buffer_dirty(sbi->s_sbh);
424 ext3_commit_super(sb, es, 1);
427 for (i = 0; i < sbi->s_gdb_count; i++)
428 brelse(sbi->s_group_desc[i]);
429 kfree(sbi->s_group_desc);
430 percpu_counter_destroy(&sbi->s_freeblocks_counter);
431 percpu_counter_destroy(&sbi->s_freeinodes_counter);
432 percpu_counter_destroy(&sbi->s_dirs_counter);
433 brelse(sbi->s_sbh);
434 #ifdef CONFIG_QUOTA
435 for (i = 0; i < MAXQUOTAS; i++)
436 kfree(sbi->s_qf_names[i]);
437 #endif
439 /* Debugging code just in case the in-memory inode orphan list
440 * isn't empty. The on-disk one can be non-empty if we've
441 * detected an error and taken the fs readonly, but the
442 * in-memory list had better be clean by this point. */
443 if (!list_empty(&sbi->s_orphan))
444 dump_orphan_list(sb, sbi);
445 J_ASSERT(list_empty(&sbi->s_orphan));
447 invalidate_bdev(sb->s_bdev);
448 if (sbi->journal_bdev && sbi->journal_bdev != sb->s_bdev) {
450 * Invalidate the journal device's buffers. We don't want them
451 * floating about in memory - the physical journal device may
452 * hotswapped, and it breaks the `ro-after' testing code.
454 sync_blockdev(sbi->journal_bdev);
455 invalidate_bdev(sbi->journal_bdev);
456 ext3_blkdev_remove(sbi);
458 sb->s_fs_info = NULL;
459 kfree(sbi->s_blockgroup_lock);
460 kfree(sbi);
463 static struct kmem_cache *ext3_inode_cachep;
466 * Called inside transaction, so use GFP_NOFS
468 static struct inode *ext3_alloc_inode(struct super_block *sb)
470 struct ext3_inode_info *ei;
472 ei = kmem_cache_alloc(ext3_inode_cachep, GFP_NOFS);
473 if (!ei)
474 return NULL;
475 ei->i_block_alloc_info = NULL;
476 ei->vfs_inode.i_version = 1;
477 atomic_set(&ei->i_datasync_tid, 0);
478 atomic_set(&ei->i_sync_tid, 0);
479 return &ei->vfs_inode;
482 static void ext3_destroy_inode(struct inode *inode)
484 if (!list_empty(&(EXT3_I(inode)->i_orphan))) {
485 printk("EXT3 Inode %p: orphan list check failed!\n",
486 EXT3_I(inode));
487 print_hex_dump(KERN_INFO, "", DUMP_PREFIX_ADDRESS, 16, 4,
488 EXT3_I(inode), sizeof(struct ext3_inode_info),
489 false);
490 dump_stack();
492 kmem_cache_free(ext3_inode_cachep, EXT3_I(inode));
495 static void init_once(void *foo)
497 struct ext3_inode_info *ei = (struct ext3_inode_info *) foo;
499 INIT_LIST_HEAD(&ei->i_orphan);
500 #ifdef CONFIG_EXT3_FS_XATTR
501 init_rwsem(&ei->xattr_sem);
502 #endif
503 mutex_init(&ei->truncate_mutex);
504 inode_init_once(&ei->vfs_inode);
507 static int init_inodecache(void)
509 ext3_inode_cachep = kmem_cache_create("ext3_inode_cache",
510 sizeof(struct ext3_inode_info),
511 0, (SLAB_RECLAIM_ACCOUNT|
512 SLAB_MEM_SPREAD),
513 init_once);
514 if (ext3_inode_cachep == NULL)
515 return -ENOMEM;
516 return 0;
519 static void destroy_inodecache(void)
521 kmem_cache_destroy(ext3_inode_cachep);
524 static inline void ext3_show_quota_options(struct seq_file *seq, struct super_block *sb)
526 #if defined(CONFIG_QUOTA)
527 struct ext3_sb_info *sbi = EXT3_SB(sb);
529 if (sbi->s_jquota_fmt) {
530 char *fmtname = "";
532 switch (sbi->s_jquota_fmt) {
533 case QFMT_VFS_OLD:
534 fmtname = "vfsold";
535 break;
536 case QFMT_VFS_V0:
537 fmtname = "vfsv0";
538 break;
539 case QFMT_VFS_V1:
540 fmtname = "vfsv1";
541 break;
543 seq_printf(seq, ",jqfmt=%s", fmtname);
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 (test_opt(sb, USRQUOTA))
553 seq_puts(seq, ",usrquota");
555 if (test_opt(sb, GRPQUOTA))
556 seq_puts(seq, ",grpquota");
557 #endif
560 static char *data_mode_string(unsigned long mode)
562 switch (mode) {
563 case EXT3_MOUNT_JOURNAL_DATA:
564 return "journal";
565 case EXT3_MOUNT_ORDERED_DATA:
566 return "ordered";
567 case EXT3_MOUNT_WRITEBACK_DATA:
568 return "writeback";
570 return "unknown";
574 * Show an option if
575 * - it's set to a non-default value OR
576 * - if the per-sb default is different from the global default
578 static int ext3_show_options(struct seq_file *seq, struct vfsmount *vfs)
580 struct super_block *sb = vfs->mnt_sb;
581 struct ext3_sb_info *sbi = EXT3_SB(sb);
582 struct ext3_super_block *es = sbi->s_es;
583 unsigned long def_mount_opts;
585 def_mount_opts = le32_to_cpu(es->s_default_mount_opts);
587 if (sbi->s_sb_block != 1)
588 seq_printf(seq, ",sb=%lu", sbi->s_sb_block);
589 if (test_opt(sb, MINIX_DF))
590 seq_puts(seq, ",minixdf");
591 if (test_opt(sb, GRPID))
592 seq_puts(seq, ",grpid");
593 if (!test_opt(sb, GRPID) && (def_mount_opts & EXT3_DEFM_BSDGROUPS))
594 seq_puts(seq, ",nogrpid");
595 if (sbi->s_resuid != EXT3_DEF_RESUID ||
596 le16_to_cpu(es->s_def_resuid) != EXT3_DEF_RESUID) {
597 seq_printf(seq, ",resuid=%u", sbi->s_resuid);
599 if (sbi->s_resgid != EXT3_DEF_RESGID ||
600 le16_to_cpu(es->s_def_resgid) != EXT3_DEF_RESGID) {
601 seq_printf(seq, ",resgid=%u", sbi->s_resgid);
603 if (test_opt(sb, ERRORS_RO)) {
604 int def_errors = le16_to_cpu(es->s_errors);
606 if (def_errors == EXT3_ERRORS_PANIC ||
607 def_errors == EXT3_ERRORS_CONTINUE) {
608 seq_puts(seq, ",errors=remount-ro");
611 if (test_opt(sb, ERRORS_CONT))
612 seq_puts(seq, ",errors=continue");
613 if (test_opt(sb, ERRORS_PANIC))
614 seq_puts(seq, ",errors=panic");
615 if (test_opt(sb, NO_UID32))
616 seq_puts(seq, ",nouid32");
617 if (test_opt(sb, DEBUG))
618 seq_puts(seq, ",debug");
619 if (test_opt(sb, OLDALLOC))
620 seq_puts(seq, ",oldalloc");
621 #ifdef CONFIG_EXT3_FS_XATTR
622 if (test_opt(sb, XATTR_USER))
623 seq_puts(seq, ",user_xattr");
624 if (!test_opt(sb, XATTR_USER) &&
625 (def_mount_opts & EXT3_DEFM_XATTR_USER)) {
626 seq_puts(seq, ",nouser_xattr");
628 #endif
629 #ifdef CONFIG_EXT3_FS_POSIX_ACL
630 if (test_opt(sb, POSIX_ACL))
631 seq_puts(seq, ",acl");
632 if (!test_opt(sb, POSIX_ACL) && (def_mount_opts & EXT3_DEFM_ACL))
633 seq_puts(seq, ",noacl");
634 #endif
635 if (!test_opt(sb, RESERVATION))
636 seq_puts(seq, ",noreservation");
637 if (sbi->s_commit_interval) {
638 seq_printf(seq, ",commit=%u",
639 (unsigned) (sbi->s_commit_interval / HZ));
643 * Always display barrier state so it's clear what the status is.
645 seq_puts(seq, ",barrier=");
646 seq_puts(seq, test_opt(sb, BARRIER) ? "1" : "0");
647 seq_printf(seq, ",data=%s", data_mode_string(test_opt(sb, DATA_FLAGS)));
648 if (test_opt(sb, DATA_ERR_ABORT))
649 seq_puts(seq, ",data_err=abort");
651 if (test_opt(sb, NOLOAD))
652 seq_puts(seq, ",norecovery");
654 ext3_show_quota_options(seq, sb);
656 return 0;
660 static struct inode *ext3_nfs_get_inode(struct super_block *sb,
661 u64 ino, u32 generation)
663 struct inode *inode;
665 if (ino < EXT3_FIRST_INO(sb) && ino != EXT3_ROOT_INO)
666 return ERR_PTR(-ESTALE);
667 if (ino > le32_to_cpu(EXT3_SB(sb)->s_es->s_inodes_count))
668 return ERR_PTR(-ESTALE);
670 /* iget isn't really right if the inode is currently unallocated!!
672 * ext3_read_inode will return a bad_inode if the inode had been
673 * deleted, so we should be safe.
675 * Currently we don't know the generation for parent directory, so
676 * a generation of 0 means "accept any"
678 inode = ext3_iget(sb, ino);
679 if (IS_ERR(inode))
680 return ERR_CAST(inode);
681 if (generation && inode->i_generation != generation) {
682 iput(inode);
683 return ERR_PTR(-ESTALE);
686 return inode;
689 static struct dentry *ext3_fh_to_dentry(struct super_block *sb, struct fid *fid,
690 int fh_len, int fh_type)
692 return generic_fh_to_dentry(sb, fid, fh_len, fh_type,
693 ext3_nfs_get_inode);
696 static struct dentry *ext3_fh_to_parent(struct super_block *sb, struct fid *fid,
697 int fh_len, int fh_type)
699 return generic_fh_to_parent(sb, fid, fh_len, fh_type,
700 ext3_nfs_get_inode);
704 * Try to release metadata pages (indirect blocks, directories) which are
705 * mapped via the block device. Since these pages could have journal heads
706 * which would prevent try_to_free_buffers() from freeing them, we must use
707 * jbd layer's try_to_free_buffers() function to release them.
709 static int bdev_try_to_free_page(struct super_block *sb, struct page *page,
710 gfp_t wait)
712 journal_t *journal = EXT3_SB(sb)->s_journal;
714 WARN_ON(PageChecked(page));
715 if (!page_has_buffers(page))
716 return 0;
717 if (journal)
718 return journal_try_to_free_buffers(journal, page,
719 wait & ~__GFP_WAIT);
720 return try_to_free_buffers(page);
723 #ifdef CONFIG_QUOTA
724 #define QTYPE2NAME(t) ((t)==USRQUOTA?"user":"group")
725 #define QTYPE2MOPT(on, t) ((t)==USRQUOTA?((on)##USRJQUOTA):((on)##GRPJQUOTA))
727 static int ext3_write_dquot(struct dquot *dquot);
728 static int ext3_acquire_dquot(struct dquot *dquot);
729 static int ext3_release_dquot(struct dquot *dquot);
730 static int ext3_mark_dquot_dirty(struct dquot *dquot);
731 static int ext3_write_info(struct super_block *sb, int type);
732 static int ext3_quota_on(struct super_block *sb, int type, int format_id,
733 char *path);
734 static int ext3_quota_on_mount(struct super_block *sb, int type);
735 static ssize_t ext3_quota_read(struct super_block *sb, int type, char *data,
736 size_t len, loff_t off);
737 static ssize_t ext3_quota_write(struct super_block *sb, int type,
738 const char *data, size_t len, loff_t off);
740 static const struct dquot_operations ext3_quota_operations = {
741 .write_dquot = ext3_write_dquot,
742 .acquire_dquot = ext3_acquire_dquot,
743 .release_dquot = ext3_release_dquot,
744 .mark_dirty = ext3_mark_dquot_dirty,
745 .write_info = ext3_write_info,
746 .alloc_dquot = dquot_alloc,
747 .destroy_dquot = dquot_destroy,
750 static const struct quotactl_ops ext3_qctl_operations = {
751 .quota_on = ext3_quota_on,
752 .quota_off = dquot_quota_off,
753 .quota_sync = dquot_quota_sync,
754 .get_info = dquot_get_dqinfo,
755 .set_info = dquot_set_dqinfo,
756 .get_dqblk = dquot_get_dqblk,
757 .set_dqblk = dquot_set_dqblk
759 #endif
761 static const struct super_operations ext3_sops = {
762 .alloc_inode = ext3_alloc_inode,
763 .destroy_inode = ext3_destroy_inode,
764 .write_inode = ext3_write_inode,
765 .dirty_inode = ext3_dirty_inode,
766 .evict_inode = ext3_evict_inode,
767 .put_super = ext3_put_super,
768 .sync_fs = ext3_sync_fs,
769 .freeze_fs = ext3_freeze,
770 .unfreeze_fs = ext3_unfreeze,
771 .statfs = ext3_statfs,
772 .remount_fs = ext3_remount,
773 .show_options = ext3_show_options,
774 #ifdef CONFIG_QUOTA
775 .quota_read = ext3_quota_read,
776 .quota_write = ext3_quota_write,
777 #endif
778 .bdev_try_to_free_page = bdev_try_to_free_page,
781 static const struct export_operations ext3_export_ops = {
782 .fh_to_dentry = ext3_fh_to_dentry,
783 .fh_to_parent = ext3_fh_to_parent,
784 .get_parent = ext3_get_parent,
787 enum {
788 Opt_bsd_df, Opt_minix_df, Opt_grpid, Opt_nogrpid,
789 Opt_resgid, Opt_resuid, Opt_sb, Opt_err_cont, Opt_err_panic, Opt_err_ro,
790 Opt_nouid32, Opt_nocheck, Opt_debug, Opt_oldalloc, Opt_orlov,
791 Opt_user_xattr, Opt_nouser_xattr, Opt_acl, Opt_noacl,
792 Opt_reservation, Opt_noreservation, Opt_noload, Opt_nobh, Opt_bh,
793 Opt_commit, Opt_journal_update, Opt_journal_inum, Opt_journal_dev,
794 Opt_abort, Opt_data_journal, Opt_data_ordered, Opt_data_writeback,
795 Opt_data_err_abort, Opt_data_err_ignore,
796 Opt_usrjquota, Opt_grpjquota, Opt_offusrjquota, Opt_offgrpjquota,
797 Opt_jqfmt_vfsold, Opt_jqfmt_vfsv0, Opt_jqfmt_vfsv1, Opt_quota,
798 Opt_noquota, Opt_ignore, Opt_barrier, Opt_nobarrier, Opt_err,
799 Opt_resize, Opt_usrquota, Opt_grpquota
802 static const match_table_t tokens = {
803 {Opt_bsd_df, "bsddf"},
804 {Opt_minix_df, "minixdf"},
805 {Opt_grpid, "grpid"},
806 {Opt_grpid, "bsdgroups"},
807 {Opt_nogrpid, "nogrpid"},
808 {Opt_nogrpid, "sysvgroups"},
809 {Opt_resgid, "resgid=%u"},
810 {Opt_resuid, "resuid=%u"},
811 {Opt_sb, "sb=%u"},
812 {Opt_err_cont, "errors=continue"},
813 {Opt_err_panic, "errors=panic"},
814 {Opt_err_ro, "errors=remount-ro"},
815 {Opt_nouid32, "nouid32"},
816 {Opt_nocheck, "nocheck"},
817 {Opt_nocheck, "check=none"},
818 {Opt_debug, "debug"},
819 {Opt_oldalloc, "oldalloc"},
820 {Opt_orlov, "orlov"},
821 {Opt_user_xattr, "user_xattr"},
822 {Opt_nouser_xattr, "nouser_xattr"},
823 {Opt_acl, "acl"},
824 {Opt_noacl, "noacl"},
825 {Opt_reservation, "reservation"},
826 {Opt_noreservation, "noreservation"},
827 {Opt_noload, "noload"},
828 {Opt_noload, "norecovery"},
829 {Opt_nobh, "nobh"},
830 {Opt_bh, "bh"},
831 {Opt_commit, "commit=%u"},
832 {Opt_journal_update, "journal=update"},
833 {Opt_journal_inum, "journal=%u"},
834 {Opt_journal_dev, "journal_dev=%u"},
835 {Opt_abort, "abort"},
836 {Opt_data_journal, "data=journal"},
837 {Opt_data_ordered, "data=ordered"},
838 {Opt_data_writeback, "data=writeback"},
839 {Opt_data_err_abort, "data_err=abort"},
840 {Opt_data_err_ignore, "data_err=ignore"},
841 {Opt_offusrjquota, "usrjquota="},
842 {Opt_usrjquota, "usrjquota=%s"},
843 {Opt_offgrpjquota, "grpjquota="},
844 {Opt_grpjquota, "grpjquota=%s"},
845 {Opt_jqfmt_vfsold, "jqfmt=vfsold"},
846 {Opt_jqfmt_vfsv0, "jqfmt=vfsv0"},
847 {Opt_jqfmt_vfsv1, "jqfmt=vfsv1"},
848 {Opt_grpquota, "grpquota"},
849 {Opt_noquota, "noquota"},
850 {Opt_quota, "quota"},
851 {Opt_usrquota, "usrquota"},
852 {Opt_barrier, "barrier=%u"},
853 {Opt_barrier, "barrier"},
854 {Opt_nobarrier, "nobarrier"},
855 {Opt_resize, "resize"},
856 {Opt_err, NULL},
859 static ext3_fsblk_t get_sb_block(void **data, struct super_block *sb)
861 ext3_fsblk_t sb_block;
862 char *options = (char *) *data;
864 if (!options || strncmp(options, "sb=", 3) != 0)
865 return 1; /* Default location */
866 options += 3;
867 /*todo: use simple_strtoll with >32bit ext3 */
868 sb_block = simple_strtoul(options, &options, 0);
869 if (*options && *options != ',') {
870 ext3_msg(sb, "error: invalid sb specification: %s",
871 (char *) *data);
872 return 1;
874 if (*options == ',')
875 options++;
876 *data = (void *) options;
877 return sb_block;
880 #ifdef CONFIG_QUOTA
881 static int set_qf_name(struct super_block *sb, int qtype, substring_t *args)
883 struct ext3_sb_info *sbi = EXT3_SB(sb);
884 char *qname;
886 if (sb_any_quota_loaded(sb) &&
887 !sbi->s_qf_names[qtype]) {
888 ext3_msg(sb, KERN_ERR,
889 "Cannot change journaled "
890 "quota options when quota turned on");
891 return 0;
893 qname = match_strdup(args);
894 if (!qname) {
895 ext3_msg(sb, KERN_ERR,
896 "Not enough memory for storing quotafile name");
897 return 0;
899 if (sbi->s_qf_names[qtype] &&
900 strcmp(sbi->s_qf_names[qtype], qname)) {
901 ext3_msg(sb, KERN_ERR,
902 "%s quota file already specified", QTYPE2NAME(qtype));
903 kfree(qname);
904 return 0;
906 sbi->s_qf_names[qtype] = qname;
907 if (strchr(sbi->s_qf_names[qtype], '/')) {
908 ext3_msg(sb, KERN_ERR,
909 "quotafile must be on filesystem root");
910 kfree(sbi->s_qf_names[qtype]);
911 sbi->s_qf_names[qtype] = NULL;
912 return 0;
914 set_opt(sbi->s_mount_opt, QUOTA);
915 return 1;
918 static int clear_qf_name(struct super_block *sb, int qtype) {
920 struct ext3_sb_info *sbi = EXT3_SB(sb);
922 if (sb_any_quota_loaded(sb) &&
923 sbi->s_qf_names[qtype]) {
924 ext3_msg(sb, KERN_ERR, "Cannot change journaled quota options"
925 " when quota turned on");
926 return 0;
929 * The space will be released later when all options are confirmed
930 * to be correct
932 sbi->s_qf_names[qtype] = NULL;
933 return 1;
935 #endif
937 static int parse_options (char *options, struct super_block *sb,
938 unsigned int *inum, unsigned long *journal_devnum,
939 ext3_fsblk_t *n_blocks_count, int is_remount)
941 struct ext3_sb_info *sbi = EXT3_SB(sb);
942 char * p;
943 substring_t args[MAX_OPT_ARGS];
944 int data_opt = 0;
945 int option;
946 #ifdef CONFIG_QUOTA
947 int qfmt;
948 #endif
950 if (!options)
951 return 1;
953 while ((p = strsep (&options, ",")) != NULL) {
954 int token;
955 if (!*p)
956 continue;
958 * Initialize args struct so we know whether arg was
959 * found; some options take optional arguments.
961 args[0].to = args[0].from = 0;
962 token = match_token(p, tokens, args);
963 switch (token) {
964 case Opt_bsd_df:
965 clear_opt (sbi->s_mount_opt, MINIX_DF);
966 break;
967 case Opt_minix_df:
968 set_opt (sbi->s_mount_opt, MINIX_DF);
969 break;
970 case Opt_grpid:
971 set_opt (sbi->s_mount_opt, GRPID);
972 break;
973 case Opt_nogrpid:
974 clear_opt (sbi->s_mount_opt, GRPID);
975 break;
976 case Opt_resuid:
977 if (match_int(&args[0], &option))
978 return 0;
979 sbi->s_resuid = option;
980 break;
981 case Opt_resgid:
982 if (match_int(&args[0], &option))
983 return 0;
984 sbi->s_resgid = option;
985 break;
986 case Opt_sb:
987 /* handled by get_sb_block() instead of here */
988 /* *sb_block = match_int(&args[0]); */
989 break;
990 case Opt_err_panic:
991 clear_opt (sbi->s_mount_opt, ERRORS_CONT);
992 clear_opt (sbi->s_mount_opt, ERRORS_RO);
993 set_opt (sbi->s_mount_opt, ERRORS_PANIC);
994 break;
995 case Opt_err_ro:
996 clear_opt (sbi->s_mount_opt, ERRORS_CONT);
997 clear_opt (sbi->s_mount_opt, ERRORS_PANIC);
998 set_opt (sbi->s_mount_opt, ERRORS_RO);
999 break;
1000 case Opt_err_cont:
1001 clear_opt (sbi->s_mount_opt, ERRORS_RO);
1002 clear_opt (sbi->s_mount_opt, ERRORS_PANIC);
1003 set_opt (sbi->s_mount_opt, ERRORS_CONT);
1004 break;
1005 case Opt_nouid32:
1006 set_opt (sbi->s_mount_opt, NO_UID32);
1007 break;
1008 case Opt_nocheck:
1009 clear_opt (sbi->s_mount_opt, CHECK);
1010 break;
1011 case Opt_debug:
1012 set_opt (sbi->s_mount_opt, DEBUG);
1013 break;
1014 case Opt_oldalloc:
1015 set_opt (sbi->s_mount_opt, OLDALLOC);
1016 break;
1017 case Opt_orlov:
1018 clear_opt (sbi->s_mount_opt, OLDALLOC);
1019 break;
1020 #ifdef CONFIG_EXT3_FS_XATTR
1021 case Opt_user_xattr:
1022 set_opt (sbi->s_mount_opt, XATTR_USER);
1023 break;
1024 case Opt_nouser_xattr:
1025 clear_opt (sbi->s_mount_opt, XATTR_USER);
1026 break;
1027 #else
1028 case Opt_user_xattr:
1029 case Opt_nouser_xattr:
1030 ext3_msg(sb, KERN_INFO,
1031 "(no)user_xattr options not supported");
1032 break;
1033 #endif
1034 #ifdef CONFIG_EXT3_FS_POSIX_ACL
1035 case Opt_acl:
1036 set_opt(sbi->s_mount_opt, POSIX_ACL);
1037 break;
1038 case Opt_noacl:
1039 clear_opt(sbi->s_mount_opt, POSIX_ACL);
1040 break;
1041 #else
1042 case Opt_acl:
1043 case Opt_noacl:
1044 ext3_msg(sb, KERN_INFO,
1045 "(no)acl options not supported");
1046 break;
1047 #endif
1048 case Opt_reservation:
1049 set_opt(sbi->s_mount_opt, RESERVATION);
1050 break;
1051 case Opt_noreservation:
1052 clear_opt(sbi->s_mount_opt, RESERVATION);
1053 break;
1054 case Opt_journal_update:
1055 /* @@@ FIXME */
1056 /* Eventually we will want to be able to create
1057 a journal file here. For now, only allow the
1058 user to specify an existing inode to be the
1059 journal file. */
1060 if (is_remount) {
1061 ext3_msg(sb, KERN_ERR, "error: cannot specify "
1062 "journal on remount");
1063 return 0;
1065 set_opt (sbi->s_mount_opt, UPDATE_JOURNAL);
1066 break;
1067 case Opt_journal_inum:
1068 if (is_remount) {
1069 ext3_msg(sb, KERN_ERR, "error: cannot specify "
1070 "journal on remount");
1071 return 0;
1073 if (match_int(&args[0], &option))
1074 return 0;
1075 *inum = option;
1076 break;
1077 case Opt_journal_dev:
1078 if (is_remount) {
1079 ext3_msg(sb, KERN_ERR, "error: cannot specify "
1080 "journal on remount");
1081 return 0;
1083 if (match_int(&args[0], &option))
1084 return 0;
1085 *journal_devnum = option;
1086 break;
1087 case Opt_noload:
1088 set_opt (sbi->s_mount_opt, NOLOAD);
1089 break;
1090 case Opt_commit:
1091 if (match_int(&args[0], &option))
1092 return 0;
1093 if (option < 0)
1094 return 0;
1095 if (option == 0)
1096 option = JBD_DEFAULT_MAX_COMMIT_AGE;
1097 sbi->s_commit_interval = HZ * option;
1098 break;
1099 case Opt_data_journal:
1100 data_opt = EXT3_MOUNT_JOURNAL_DATA;
1101 goto datacheck;
1102 case Opt_data_ordered:
1103 data_opt = EXT3_MOUNT_ORDERED_DATA;
1104 goto datacheck;
1105 case Opt_data_writeback:
1106 data_opt = EXT3_MOUNT_WRITEBACK_DATA;
1107 datacheck:
1108 if (is_remount) {
1109 if (test_opt(sb, DATA_FLAGS) == data_opt)
1110 break;
1111 ext3_msg(sb, KERN_ERR,
1112 "error: cannot change "
1113 "data mode on remount. The filesystem "
1114 "is mounted in data=%s mode and you "
1115 "try to remount it in data=%s mode.",
1116 data_mode_string(test_opt(sb,
1117 DATA_FLAGS)),
1118 data_mode_string(data_opt));
1119 return 0;
1120 } else {
1121 clear_opt(sbi->s_mount_opt, DATA_FLAGS);
1122 sbi->s_mount_opt |= data_opt;
1124 break;
1125 case Opt_data_err_abort:
1126 set_opt(sbi->s_mount_opt, DATA_ERR_ABORT);
1127 break;
1128 case Opt_data_err_ignore:
1129 clear_opt(sbi->s_mount_opt, DATA_ERR_ABORT);
1130 break;
1131 #ifdef CONFIG_QUOTA
1132 case Opt_usrjquota:
1133 if (!set_qf_name(sb, USRQUOTA, &args[0]))
1134 return 0;
1135 break;
1136 case Opt_grpjquota:
1137 if (!set_qf_name(sb, GRPQUOTA, &args[0]))
1138 return 0;
1139 break;
1140 case Opt_offusrjquota:
1141 if (!clear_qf_name(sb, USRQUOTA))
1142 return 0;
1143 break;
1144 case Opt_offgrpjquota:
1145 if (!clear_qf_name(sb, GRPQUOTA))
1146 return 0;
1147 break;
1148 case Opt_jqfmt_vfsold:
1149 qfmt = QFMT_VFS_OLD;
1150 goto set_qf_format;
1151 case Opt_jqfmt_vfsv0:
1152 qfmt = QFMT_VFS_V0;
1153 goto set_qf_format;
1154 case Opt_jqfmt_vfsv1:
1155 qfmt = QFMT_VFS_V1;
1156 set_qf_format:
1157 if (sb_any_quota_loaded(sb) &&
1158 sbi->s_jquota_fmt != qfmt) {
1159 ext3_msg(sb, KERN_ERR, "error: cannot change "
1160 "journaled quota options when "
1161 "quota turned on.");
1162 return 0;
1164 sbi->s_jquota_fmt = qfmt;
1165 break;
1166 case Opt_quota:
1167 case Opt_usrquota:
1168 set_opt(sbi->s_mount_opt, QUOTA);
1169 set_opt(sbi->s_mount_opt, USRQUOTA);
1170 break;
1171 case Opt_grpquota:
1172 set_opt(sbi->s_mount_opt, QUOTA);
1173 set_opt(sbi->s_mount_opt, GRPQUOTA);
1174 break;
1175 case Opt_noquota:
1176 if (sb_any_quota_loaded(sb)) {
1177 ext3_msg(sb, KERN_ERR, "error: cannot change "
1178 "quota options when quota turned on.");
1179 return 0;
1181 clear_opt(sbi->s_mount_opt, QUOTA);
1182 clear_opt(sbi->s_mount_opt, USRQUOTA);
1183 clear_opt(sbi->s_mount_opt, GRPQUOTA);
1184 break;
1185 #else
1186 case Opt_quota:
1187 case Opt_usrquota:
1188 case Opt_grpquota:
1189 ext3_msg(sb, KERN_ERR,
1190 "error: quota options not supported.");
1191 break;
1192 case Opt_usrjquota:
1193 case Opt_grpjquota:
1194 case Opt_offusrjquota:
1195 case Opt_offgrpjquota:
1196 case Opt_jqfmt_vfsold:
1197 case Opt_jqfmt_vfsv0:
1198 case Opt_jqfmt_vfsv1:
1199 ext3_msg(sb, KERN_ERR,
1200 "error: journaled quota options not "
1201 "supported.");
1202 break;
1203 case Opt_noquota:
1204 break;
1205 #endif
1206 case Opt_abort:
1207 set_opt(sbi->s_mount_opt, ABORT);
1208 break;
1209 case Opt_nobarrier:
1210 clear_opt(sbi->s_mount_opt, BARRIER);
1211 break;
1212 case Opt_barrier:
1213 if (args[0].from) {
1214 if (match_int(&args[0], &option))
1215 return 0;
1216 } else
1217 option = 1; /* No argument, default to 1 */
1218 if (option)
1219 set_opt(sbi->s_mount_opt, BARRIER);
1220 else
1221 clear_opt(sbi->s_mount_opt, BARRIER);
1222 break;
1223 case Opt_ignore:
1224 break;
1225 case Opt_resize:
1226 if (!is_remount) {
1227 ext3_msg(sb, KERN_ERR,
1228 "error: resize option only available "
1229 "for remount");
1230 return 0;
1232 if (match_int(&args[0], &option) != 0)
1233 return 0;
1234 *n_blocks_count = option;
1235 break;
1236 case Opt_nobh:
1237 ext3_msg(sb, KERN_WARNING,
1238 "warning: ignoring deprecated nobh option");
1239 break;
1240 case Opt_bh:
1241 ext3_msg(sb, KERN_WARNING,
1242 "warning: ignoring deprecated bh option");
1243 break;
1244 default:
1245 ext3_msg(sb, KERN_ERR,
1246 "error: unrecognized mount option \"%s\" "
1247 "or missing value", p);
1248 return 0;
1251 #ifdef CONFIG_QUOTA
1252 if (sbi->s_qf_names[USRQUOTA] || sbi->s_qf_names[GRPQUOTA]) {
1253 if (test_opt(sb, USRQUOTA) && sbi->s_qf_names[USRQUOTA])
1254 clear_opt(sbi->s_mount_opt, USRQUOTA);
1255 if (test_opt(sb, GRPQUOTA) && sbi->s_qf_names[GRPQUOTA])
1256 clear_opt(sbi->s_mount_opt, GRPQUOTA);
1258 if (test_opt(sb, GRPQUOTA) || test_opt(sb, USRQUOTA)) {
1259 ext3_msg(sb, KERN_ERR, "error: old and new quota "
1260 "format mixing.");
1261 return 0;
1264 if (!sbi->s_jquota_fmt) {
1265 ext3_msg(sb, KERN_ERR, "error: journaled quota format "
1266 "not specified.");
1267 return 0;
1269 } else {
1270 if (sbi->s_jquota_fmt) {
1271 ext3_msg(sb, KERN_ERR, "error: journaled quota format "
1272 "specified with no journaling "
1273 "enabled.");
1274 return 0;
1277 #endif
1278 return 1;
1281 static int ext3_setup_super(struct super_block *sb, struct ext3_super_block *es,
1282 int read_only)
1284 struct ext3_sb_info *sbi = EXT3_SB(sb);
1285 int res = 0;
1287 if (le32_to_cpu(es->s_rev_level) > EXT3_MAX_SUPP_REV) {
1288 ext3_msg(sb, KERN_ERR,
1289 "error: revision level too high, "
1290 "forcing read-only mode");
1291 res = MS_RDONLY;
1293 if (read_only)
1294 return res;
1295 if (!(sbi->s_mount_state & EXT3_VALID_FS))
1296 ext3_msg(sb, KERN_WARNING,
1297 "warning: mounting unchecked fs, "
1298 "running e2fsck is recommended");
1299 else if ((sbi->s_mount_state & EXT3_ERROR_FS))
1300 ext3_msg(sb, KERN_WARNING,
1301 "warning: mounting fs with errors, "
1302 "running e2fsck is recommended");
1303 else if ((__s16) le16_to_cpu(es->s_max_mnt_count) > 0 &&
1304 le16_to_cpu(es->s_mnt_count) >=
1305 le16_to_cpu(es->s_max_mnt_count))
1306 ext3_msg(sb, KERN_WARNING,
1307 "warning: maximal mount count reached, "
1308 "running e2fsck is recommended");
1309 else if (le32_to_cpu(es->s_checkinterval) &&
1310 (le32_to_cpu(es->s_lastcheck) +
1311 le32_to_cpu(es->s_checkinterval) <= get_seconds()))
1312 ext3_msg(sb, KERN_WARNING,
1313 "warning: checktime reached, "
1314 "running e2fsck is recommended");
1315 #if 0
1316 /* @@@ We _will_ want to clear the valid bit if we find
1317 inconsistencies, to force a fsck at reboot. But for
1318 a plain journaled filesystem we can keep it set as
1319 valid forever! :) */
1320 es->s_state &= cpu_to_le16(~EXT3_VALID_FS);
1321 #endif
1322 if (!le16_to_cpu(es->s_max_mnt_count))
1323 es->s_max_mnt_count = cpu_to_le16(EXT3_DFL_MAX_MNT_COUNT);
1324 le16_add_cpu(&es->s_mnt_count, 1);
1325 es->s_mtime = cpu_to_le32(get_seconds());
1326 ext3_update_dynamic_rev(sb);
1327 EXT3_SET_INCOMPAT_FEATURE(sb, EXT3_FEATURE_INCOMPAT_RECOVER);
1329 ext3_commit_super(sb, es, 1);
1330 if (test_opt(sb, DEBUG))
1331 ext3_msg(sb, KERN_INFO, "[bs=%lu, gc=%lu, "
1332 "bpg=%lu, ipg=%lu, mo=%04lx]",
1333 sb->s_blocksize,
1334 sbi->s_groups_count,
1335 EXT3_BLOCKS_PER_GROUP(sb),
1336 EXT3_INODES_PER_GROUP(sb),
1337 sbi->s_mount_opt);
1339 if (EXT3_SB(sb)->s_journal->j_inode == NULL) {
1340 char b[BDEVNAME_SIZE];
1341 ext3_msg(sb, KERN_INFO, "using external journal on %s",
1342 bdevname(EXT3_SB(sb)->s_journal->j_dev, b));
1343 } else {
1344 ext3_msg(sb, KERN_INFO, "using internal journal");
1346 return res;
1349 /* Called at mount-time, super-block is locked */
1350 static int ext3_check_descriptors(struct super_block *sb)
1352 struct ext3_sb_info *sbi = EXT3_SB(sb);
1353 int i;
1355 ext3_debug ("Checking group descriptors");
1357 for (i = 0; i < sbi->s_groups_count; i++) {
1358 struct ext3_group_desc *gdp = ext3_get_group_desc(sb, i, NULL);
1359 ext3_fsblk_t first_block = ext3_group_first_block_no(sb, i);
1360 ext3_fsblk_t last_block;
1362 if (i == sbi->s_groups_count - 1)
1363 last_block = le32_to_cpu(sbi->s_es->s_blocks_count) - 1;
1364 else
1365 last_block = first_block +
1366 (EXT3_BLOCKS_PER_GROUP(sb) - 1);
1368 if (le32_to_cpu(gdp->bg_block_bitmap) < first_block ||
1369 le32_to_cpu(gdp->bg_block_bitmap) > last_block)
1371 ext3_error (sb, "ext3_check_descriptors",
1372 "Block bitmap for group %d"
1373 " not in group (block %lu)!",
1374 i, (unsigned long)
1375 le32_to_cpu(gdp->bg_block_bitmap));
1376 return 0;
1378 if (le32_to_cpu(gdp->bg_inode_bitmap) < first_block ||
1379 le32_to_cpu(gdp->bg_inode_bitmap) > last_block)
1381 ext3_error (sb, "ext3_check_descriptors",
1382 "Inode bitmap for group %d"
1383 " not in group (block %lu)!",
1384 i, (unsigned long)
1385 le32_to_cpu(gdp->bg_inode_bitmap));
1386 return 0;
1388 if (le32_to_cpu(gdp->bg_inode_table) < first_block ||
1389 le32_to_cpu(gdp->bg_inode_table) + sbi->s_itb_per_group - 1 >
1390 last_block)
1392 ext3_error (sb, "ext3_check_descriptors",
1393 "Inode table for group %d"
1394 " not in group (block %lu)!",
1395 i, (unsigned long)
1396 le32_to_cpu(gdp->bg_inode_table));
1397 return 0;
1401 sbi->s_es->s_free_blocks_count=cpu_to_le32(ext3_count_free_blocks(sb));
1402 sbi->s_es->s_free_inodes_count=cpu_to_le32(ext3_count_free_inodes(sb));
1403 return 1;
1407 /* ext3_orphan_cleanup() walks a singly-linked list of inodes (starting at
1408 * the superblock) which were deleted from all directories, but held open by
1409 * a process at the time of a crash. We walk the list and try to delete these
1410 * inodes at recovery time (only with a read-write filesystem).
1412 * In order to keep the orphan inode chain consistent during traversal (in
1413 * case of crash during recovery), we link each inode into the superblock
1414 * orphan list_head and handle it the same way as an inode deletion during
1415 * normal operation (which journals the operations for us).
1417 * We only do an iget() and an iput() on each inode, which is very safe if we
1418 * accidentally point at an in-use or already deleted inode. The worst that
1419 * can happen in this case is that we get a "bit already cleared" message from
1420 * ext3_free_inode(). The only reason we would point at a wrong inode is if
1421 * e2fsck was run on this filesystem, and it must have already done the orphan
1422 * inode cleanup for us, so we can safely abort without any further action.
1424 static void ext3_orphan_cleanup (struct super_block * sb,
1425 struct ext3_super_block * es)
1427 unsigned int s_flags = sb->s_flags;
1428 int nr_orphans = 0, nr_truncates = 0;
1429 #ifdef CONFIG_QUOTA
1430 int i;
1431 #endif
1432 if (!es->s_last_orphan) {
1433 jbd_debug(4, "no orphan inodes to clean up\n");
1434 return;
1437 if (bdev_read_only(sb->s_bdev)) {
1438 ext3_msg(sb, KERN_ERR, "error: write access "
1439 "unavailable, skipping orphan cleanup.");
1440 return;
1443 if (EXT3_SB(sb)->s_mount_state & EXT3_ERROR_FS) {
1444 if (es->s_last_orphan)
1445 jbd_debug(1, "Errors on filesystem, "
1446 "clearing orphan list.\n");
1447 es->s_last_orphan = 0;
1448 jbd_debug(1, "Skipping orphan recovery on fs with errors.\n");
1449 return;
1452 if (s_flags & MS_RDONLY) {
1453 ext3_msg(sb, KERN_INFO, "orphan cleanup on readonly fs");
1454 sb->s_flags &= ~MS_RDONLY;
1456 #ifdef CONFIG_QUOTA
1457 /* Needed for iput() to work correctly and not trash data */
1458 sb->s_flags |= MS_ACTIVE;
1459 /* Turn on quotas so that they are updated correctly */
1460 for (i = 0; i < MAXQUOTAS; i++) {
1461 if (EXT3_SB(sb)->s_qf_names[i]) {
1462 int ret = ext3_quota_on_mount(sb, i);
1463 if (ret < 0)
1464 ext3_msg(sb, KERN_ERR,
1465 "error: cannot turn on journaled "
1466 "quota: %d", ret);
1469 #endif
1471 while (es->s_last_orphan) {
1472 struct inode *inode;
1474 inode = ext3_orphan_get(sb, le32_to_cpu(es->s_last_orphan));
1475 if (IS_ERR(inode)) {
1476 es->s_last_orphan = 0;
1477 break;
1480 list_add(&EXT3_I(inode)->i_orphan, &EXT3_SB(sb)->s_orphan);
1481 dquot_initialize(inode);
1482 if (inode->i_nlink) {
1483 printk(KERN_DEBUG
1484 "%s: truncating inode %lu to %Ld bytes\n",
1485 __func__, inode->i_ino, inode->i_size);
1486 jbd_debug(2, "truncating inode %lu to %Ld bytes\n",
1487 inode->i_ino, inode->i_size);
1488 ext3_truncate(inode);
1489 nr_truncates++;
1490 } else {
1491 printk(KERN_DEBUG
1492 "%s: deleting unreferenced inode %lu\n",
1493 __func__, inode->i_ino);
1494 jbd_debug(2, "deleting unreferenced inode %lu\n",
1495 inode->i_ino);
1496 nr_orphans++;
1498 iput(inode); /* The delete magic happens here! */
1501 #define PLURAL(x) (x), ((x)==1) ? "" : "s"
1503 if (nr_orphans)
1504 ext3_msg(sb, KERN_INFO, "%d orphan inode%s deleted",
1505 PLURAL(nr_orphans));
1506 if (nr_truncates)
1507 ext3_msg(sb, KERN_INFO, "%d truncate%s cleaned up",
1508 PLURAL(nr_truncates));
1509 #ifdef CONFIG_QUOTA
1510 /* Turn quotas off */
1511 for (i = 0; i < MAXQUOTAS; i++) {
1512 if (sb_dqopt(sb)->files[i])
1513 dquot_quota_off(sb, i);
1515 #endif
1516 sb->s_flags = s_flags; /* Restore MS_RDONLY status */
1520 * Maximal file size. There is a direct, and {,double-,triple-}indirect
1521 * block limit, and also a limit of (2^32 - 1) 512-byte sectors in i_blocks.
1522 * We need to be 1 filesystem block less than the 2^32 sector limit.
1524 static loff_t ext3_max_size(int bits)
1526 loff_t res = EXT3_NDIR_BLOCKS;
1527 int meta_blocks;
1528 loff_t upper_limit;
1530 /* This is calculated to be the largest file size for a
1531 * dense, file such that the total number of
1532 * sectors in the file, including data and all indirect blocks,
1533 * does not exceed 2^32 -1
1534 * __u32 i_blocks representing the total number of
1535 * 512 bytes blocks of the file
1537 upper_limit = (1LL << 32) - 1;
1539 /* total blocks in file system block size */
1540 upper_limit >>= (bits - 9);
1543 /* indirect blocks */
1544 meta_blocks = 1;
1545 /* double indirect blocks */
1546 meta_blocks += 1 + (1LL << (bits-2));
1547 /* tripple indirect blocks */
1548 meta_blocks += 1 + (1LL << (bits-2)) + (1LL << (2*(bits-2)));
1550 upper_limit -= meta_blocks;
1551 upper_limit <<= bits;
1553 res += 1LL << (bits-2);
1554 res += 1LL << (2*(bits-2));
1555 res += 1LL << (3*(bits-2));
1556 res <<= bits;
1557 if (res > upper_limit)
1558 res = upper_limit;
1560 if (res > MAX_LFS_FILESIZE)
1561 res = MAX_LFS_FILESIZE;
1563 return res;
1566 static ext3_fsblk_t descriptor_loc(struct super_block *sb,
1567 ext3_fsblk_t logic_sb_block,
1568 int nr)
1570 struct ext3_sb_info *sbi = EXT3_SB(sb);
1571 unsigned long bg, first_meta_bg;
1572 int has_super = 0;
1574 first_meta_bg = le32_to_cpu(sbi->s_es->s_first_meta_bg);
1576 if (!EXT3_HAS_INCOMPAT_FEATURE(sb, EXT3_FEATURE_INCOMPAT_META_BG) ||
1577 nr < first_meta_bg)
1578 return (logic_sb_block + nr + 1);
1579 bg = sbi->s_desc_per_block * nr;
1580 if (ext3_bg_has_super(sb, bg))
1581 has_super = 1;
1582 return (has_super + ext3_group_first_block_no(sb, bg));
1586 static int ext3_fill_super (struct super_block *sb, void *data, int silent)
1588 struct buffer_head * bh;
1589 struct ext3_super_block *es = NULL;
1590 struct ext3_sb_info *sbi;
1591 ext3_fsblk_t block;
1592 ext3_fsblk_t sb_block = get_sb_block(&data, sb);
1593 ext3_fsblk_t logic_sb_block;
1594 unsigned long offset = 0;
1595 unsigned int journal_inum = 0;
1596 unsigned long journal_devnum = 0;
1597 unsigned long def_mount_opts;
1598 struct inode *root;
1599 int blocksize;
1600 int hblock;
1601 int db_count;
1602 int i;
1603 int needs_recovery;
1604 int ret = -EINVAL;
1605 __le32 features;
1606 int err;
1608 sbi = kzalloc(sizeof(*sbi), GFP_KERNEL);
1609 if (!sbi)
1610 return -ENOMEM;
1612 sbi->s_blockgroup_lock =
1613 kzalloc(sizeof(struct blockgroup_lock), GFP_KERNEL);
1614 if (!sbi->s_blockgroup_lock) {
1615 kfree(sbi);
1616 return -ENOMEM;
1618 sb->s_fs_info = sbi;
1619 sbi->s_mount_opt = 0;
1620 sbi->s_resuid = EXT3_DEF_RESUID;
1621 sbi->s_resgid = EXT3_DEF_RESGID;
1622 sbi->s_sb_block = sb_block;
1624 blocksize = sb_min_blocksize(sb, EXT3_MIN_BLOCK_SIZE);
1625 if (!blocksize) {
1626 ext3_msg(sb, KERN_ERR, "error: unable to set blocksize");
1627 goto out_fail;
1631 * The ext3 superblock will not be buffer aligned for other than 1kB
1632 * block sizes. We need to calculate the offset from buffer start.
1634 if (blocksize != EXT3_MIN_BLOCK_SIZE) {
1635 logic_sb_block = (sb_block * EXT3_MIN_BLOCK_SIZE) / blocksize;
1636 offset = (sb_block * EXT3_MIN_BLOCK_SIZE) % blocksize;
1637 } else {
1638 logic_sb_block = sb_block;
1641 if (!(bh = sb_bread(sb, logic_sb_block))) {
1642 ext3_msg(sb, KERN_ERR, "error: unable to read superblock");
1643 goto out_fail;
1646 * Note: s_es must be initialized as soon as possible because
1647 * some ext3 macro-instructions depend on its value
1649 es = (struct ext3_super_block *) (bh->b_data + offset);
1650 sbi->s_es = es;
1651 sb->s_magic = le16_to_cpu(es->s_magic);
1652 if (sb->s_magic != EXT3_SUPER_MAGIC)
1653 goto cantfind_ext3;
1655 /* Set defaults before we parse the mount options */
1656 def_mount_opts = le32_to_cpu(es->s_default_mount_opts);
1657 if (def_mount_opts & EXT3_DEFM_DEBUG)
1658 set_opt(sbi->s_mount_opt, DEBUG);
1659 if (def_mount_opts & EXT3_DEFM_BSDGROUPS)
1660 set_opt(sbi->s_mount_opt, GRPID);
1661 if (def_mount_opts & EXT3_DEFM_UID16)
1662 set_opt(sbi->s_mount_opt, NO_UID32);
1663 #ifdef CONFIG_EXT3_FS_XATTR
1664 if (def_mount_opts & EXT3_DEFM_XATTR_USER)
1665 set_opt(sbi->s_mount_opt, XATTR_USER);
1666 #endif
1667 #ifdef CONFIG_EXT3_FS_POSIX_ACL
1668 if (def_mount_opts & EXT3_DEFM_ACL)
1669 set_opt(sbi->s_mount_opt, POSIX_ACL);
1670 #endif
1671 if ((def_mount_opts & EXT3_DEFM_JMODE) == EXT3_DEFM_JMODE_DATA)
1672 set_opt(sbi->s_mount_opt, JOURNAL_DATA);
1673 else if ((def_mount_opts & EXT3_DEFM_JMODE) == EXT3_DEFM_JMODE_ORDERED)
1674 set_opt(sbi->s_mount_opt, ORDERED_DATA);
1675 else if ((def_mount_opts & EXT3_DEFM_JMODE) == EXT3_DEFM_JMODE_WBACK)
1676 set_opt(sbi->s_mount_opt, WRITEBACK_DATA);
1678 if (le16_to_cpu(sbi->s_es->s_errors) == EXT3_ERRORS_PANIC)
1679 set_opt(sbi->s_mount_opt, ERRORS_PANIC);
1680 else if (le16_to_cpu(sbi->s_es->s_errors) == EXT3_ERRORS_CONTINUE)
1681 set_opt(sbi->s_mount_opt, ERRORS_CONT);
1682 else
1683 set_opt(sbi->s_mount_opt, ERRORS_RO);
1685 sbi->s_resuid = le16_to_cpu(es->s_def_resuid);
1686 sbi->s_resgid = le16_to_cpu(es->s_def_resgid);
1688 set_opt(sbi->s_mount_opt, RESERVATION);
1690 if (!parse_options ((char *) data, sb, &journal_inum, &journal_devnum,
1691 NULL, 0))
1692 goto failed_mount;
1694 sb->s_flags = (sb->s_flags & ~MS_POSIXACL) |
1695 (test_opt(sb, POSIX_ACL) ? MS_POSIXACL : 0);
1697 if (le32_to_cpu(es->s_rev_level) == EXT3_GOOD_OLD_REV &&
1698 (EXT3_HAS_COMPAT_FEATURE(sb, ~0U) ||
1699 EXT3_HAS_RO_COMPAT_FEATURE(sb, ~0U) ||
1700 EXT3_HAS_INCOMPAT_FEATURE(sb, ~0U)))
1701 ext3_msg(sb, KERN_WARNING,
1702 "warning: feature flags set on rev 0 fs, "
1703 "running e2fsck is recommended");
1705 * Check feature flags regardless of the revision level, since we
1706 * previously didn't change the revision level when setting the flags,
1707 * so there is a chance incompat flags are set on a rev 0 filesystem.
1709 features = EXT3_HAS_INCOMPAT_FEATURE(sb, ~EXT3_FEATURE_INCOMPAT_SUPP);
1710 if (features) {
1711 ext3_msg(sb, KERN_ERR,
1712 "error: couldn't mount because of unsupported "
1713 "optional features (%x)", le32_to_cpu(features));
1714 goto failed_mount;
1716 features = EXT3_HAS_RO_COMPAT_FEATURE(sb, ~EXT3_FEATURE_RO_COMPAT_SUPP);
1717 if (!(sb->s_flags & MS_RDONLY) && features) {
1718 ext3_msg(sb, KERN_ERR,
1719 "error: couldn't mount RDWR because of unsupported "
1720 "optional features (%x)", le32_to_cpu(features));
1721 goto failed_mount;
1723 blocksize = BLOCK_SIZE << le32_to_cpu(es->s_log_block_size);
1725 if (blocksize < EXT3_MIN_BLOCK_SIZE ||
1726 blocksize > EXT3_MAX_BLOCK_SIZE) {
1727 ext3_msg(sb, KERN_ERR,
1728 "error: couldn't mount because of unsupported "
1729 "filesystem blocksize %d", blocksize);
1730 goto failed_mount;
1733 hblock = bdev_logical_block_size(sb->s_bdev);
1734 if (sb->s_blocksize != blocksize) {
1736 * Make sure the blocksize for the filesystem is larger
1737 * than the hardware sectorsize for the machine.
1739 if (blocksize < hblock) {
1740 ext3_msg(sb, KERN_ERR,
1741 "error: fsblocksize %d too small for "
1742 "hardware sectorsize %d", blocksize, hblock);
1743 goto failed_mount;
1746 brelse (bh);
1747 if (!sb_set_blocksize(sb, blocksize)) {
1748 ext3_msg(sb, KERN_ERR,
1749 "error: bad blocksize %d", blocksize);
1750 goto out_fail;
1752 logic_sb_block = (sb_block * EXT3_MIN_BLOCK_SIZE) / blocksize;
1753 offset = (sb_block * EXT3_MIN_BLOCK_SIZE) % blocksize;
1754 bh = sb_bread(sb, logic_sb_block);
1755 if (!bh) {
1756 ext3_msg(sb, KERN_ERR,
1757 "error: can't read superblock on 2nd try");
1758 goto failed_mount;
1760 es = (struct ext3_super_block *)(bh->b_data + offset);
1761 sbi->s_es = es;
1762 if (es->s_magic != cpu_to_le16(EXT3_SUPER_MAGIC)) {
1763 ext3_msg(sb, KERN_ERR,
1764 "error: magic mismatch");
1765 goto failed_mount;
1769 sb->s_maxbytes = ext3_max_size(sb->s_blocksize_bits);
1771 if (le32_to_cpu(es->s_rev_level) == EXT3_GOOD_OLD_REV) {
1772 sbi->s_inode_size = EXT3_GOOD_OLD_INODE_SIZE;
1773 sbi->s_first_ino = EXT3_GOOD_OLD_FIRST_INO;
1774 } else {
1775 sbi->s_inode_size = le16_to_cpu(es->s_inode_size);
1776 sbi->s_first_ino = le32_to_cpu(es->s_first_ino);
1777 if ((sbi->s_inode_size < EXT3_GOOD_OLD_INODE_SIZE) ||
1778 (!is_power_of_2(sbi->s_inode_size)) ||
1779 (sbi->s_inode_size > blocksize)) {
1780 ext3_msg(sb, KERN_ERR,
1781 "error: unsupported inode size: %d",
1782 sbi->s_inode_size);
1783 goto failed_mount;
1786 sbi->s_frag_size = EXT3_MIN_FRAG_SIZE <<
1787 le32_to_cpu(es->s_log_frag_size);
1788 if (blocksize != sbi->s_frag_size) {
1789 ext3_msg(sb, KERN_ERR,
1790 "error: fragsize %lu != blocksize %u (unsupported)",
1791 sbi->s_frag_size, blocksize);
1792 goto failed_mount;
1794 sbi->s_frags_per_block = 1;
1795 sbi->s_blocks_per_group = le32_to_cpu(es->s_blocks_per_group);
1796 sbi->s_frags_per_group = le32_to_cpu(es->s_frags_per_group);
1797 sbi->s_inodes_per_group = le32_to_cpu(es->s_inodes_per_group);
1798 if (EXT3_INODE_SIZE(sb) == 0 || EXT3_INODES_PER_GROUP(sb) == 0)
1799 goto cantfind_ext3;
1800 sbi->s_inodes_per_block = blocksize / EXT3_INODE_SIZE(sb);
1801 if (sbi->s_inodes_per_block == 0)
1802 goto cantfind_ext3;
1803 sbi->s_itb_per_group = sbi->s_inodes_per_group /
1804 sbi->s_inodes_per_block;
1805 sbi->s_desc_per_block = blocksize / sizeof(struct ext3_group_desc);
1806 sbi->s_sbh = bh;
1807 sbi->s_mount_state = le16_to_cpu(es->s_state);
1808 sbi->s_addr_per_block_bits = ilog2(EXT3_ADDR_PER_BLOCK(sb));
1809 sbi->s_desc_per_block_bits = ilog2(EXT3_DESC_PER_BLOCK(sb));
1810 for (i=0; i < 4; i++)
1811 sbi->s_hash_seed[i] = le32_to_cpu(es->s_hash_seed[i]);
1812 sbi->s_def_hash_version = es->s_def_hash_version;
1813 i = le32_to_cpu(es->s_flags);
1814 if (i & EXT2_FLAGS_UNSIGNED_HASH)
1815 sbi->s_hash_unsigned = 3;
1816 else if ((i & EXT2_FLAGS_SIGNED_HASH) == 0) {
1817 #ifdef __CHAR_UNSIGNED__
1818 es->s_flags |= cpu_to_le32(EXT2_FLAGS_UNSIGNED_HASH);
1819 sbi->s_hash_unsigned = 3;
1820 #else
1821 es->s_flags |= cpu_to_le32(EXT2_FLAGS_SIGNED_HASH);
1822 #endif
1825 if (sbi->s_blocks_per_group > blocksize * 8) {
1826 ext3_msg(sb, KERN_ERR,
1827 "#blocks per group too big: %lu",
1828 sbi->s_blocks_per_group);
1829 goto failed_mount;
1831 if (sbi->s_frags_per_group > blocksize * 8) {
1832 ext3_msg(sb, KERN_ERR,
1833 "error: #fragments per group too big: %lu",
1834 sbi->s_frags_per_group);
1835 goto failed_mount;
1837 if (sbi->s_inodes_per_group > blocksize * 8) {
1838 ext3_msg(sb, KERN_ERR,
1839 "error: #inodes per group too big: %lu",
1840 sbi->s_inodes_per_group);
1841 goto failed_mount;
1844 if (generic_check_addressable(sb->s_blocksize_bits,
1845 le32_to_cpu(es->s_blocks_count))) {
1846 ext3_msg(sb, KERN_ERR,
1847 "error: filesystem is too large to mount safely");
1848 if (sizeof(sector_t) < 8)
1849 ext3_msg(sb, KERN_ERR,
1850 "error: CONFIG_LBDAF not enabled");
1851 goto failed_mount;
1854 if (EXT3_BLOCKS_PER_GROUP(sb) == 0)
1855 goto cantfind_ext3;
1856 sbi->s_groups_count = ((le32_to_cpu(es->s_blocks_count) -
1857 le32_to_cpu(es->s_first_data_block) - 1)
1858 / EXT3_BLOCKS_PER_GROUP(sb)) + 1;
1859 db_count = DIV_ROUND_UP(sbi->s_groups_count, EXT3_DESC_PER_BLOCK(sb));
1860 sbi->s_group_desc = kmalloc(db_count * sizeof (struct buffer_head *),
1861 GFP_KERNEL);
1862 if (sbi->s_group_desc == NULL) {
1863 ext3_msg(sb, KERN_ERR,
1864 "error: not enough memory");
1865 ret = -ENOMEM;
1866 goto failed_mount;
1869 bgl_lock_init(sbi->s_blockgroup_lock);
1871 for (i = 0; i < db_count; i++) {
1872 block = descriptor_loc(sb, logic_sb_block, i);
1873 sbi->s_group_desc[i] = sb_bread(sb, block);
1874 if (!sbi->s_group_desc[i]) {
1875 ext3_msg(sb, KERN_ERR,
1876 "error: can't read group descriptor %d", i);
1877 db_count = i;
1878 goto failed_mount2;
1881 if (!ext3_check_descriptors (sb)) {
1882 ext3_msg(sb, KERN_ERR,
1883 "error: group descriptors corrupted");
1884 goto failed_mount2;
1886 sbi->s_gdb_count = db_count;
1887 get_random_bytes(&sbi->s_next_generation, sizeof(u32));
1888 spin_lock_init(&sbi->s_next_gen_lock);
1890 /* per fileystem reservation list head & lock */
1891 spin_lock_init(&sbi->s_rsv_window_lock);
1892 sbi->s_rsv_window_root = RB_ROOT;
1893 /* Add a single, static dummy reservation to the start of the
1894 * reservation window list --- it gives us a placeholder for
1895 * append-at-start-of-list which makes the allocation logic
1896 * _much_ simpler. */
1897 sbi->s_rsv_window_head.rsv_start = EXT3_RESERVE_WINDOW_NOT_ALLOCATED;
1898 sbi->s_rsv_window_head.rsv_end = EXT3_RESERVE_WINDOW_NOT_ALLOCATED;
1899 sbi->s_rsv_window_head.rsv_alloc_hit = 0;
1900 sbi->s_rsv_window_head.rsv_goal_size = 0;
1901 ext3_rsv_window_add(sb, &sbi->s_rsv_window_head);
1904 * set up enough so that it can read an inode
1906 sb->s_op = &ext3_sops;
1907 sb->s_export_op = &ext3_export_ops;
1908 sb->s_xattr = ext3_xattr_handlers;
1909 #ifdef CONFIG_QUOTA
1910 sb->s_qcop = &ext3_qctl_operations;
1911 sb->dq_op = &ext3_quota_operations;
1912 #endif
1913 INIT_LIST_HEAD(&sbi->s_orphan); /* unlinked but open files */
1914 mutex_init(&sbi->s_orphan_lock);
1915 mutex_init(&sbi->s_resize_lock);
1917 sb->s_root = NULL;
1919 needs_recovery = (es->s_last_orphan != 0 ||
1920 EXT3_HAS_INCOMPAT_FEATURE(sb,
1921 EXT3_FEATURE_INCOMPAT_RECOVER));
1924 * The first inode we look at is the journal inode. Don't try
1925 * root first: it may be modified in the journal!
1927 if (!test_opt(sb, NOLOAD) &&
1928 EXT3_HAS_COMPAT_FEATURE(sb, EXT3_FEATURE_COMPAT_HAS_JOURNAL)) {
1929 if (ext3_load_journal(sb, es, journal_devnum))
1930 goto failed_mount2;
1931 } else if (journal_inum) {
1932 if (ext3_create_journal(sb, es, journal_inum))
1933 goto failed_mount2;
1934 } else {
1935 if (!silent)
1936 ext3_msg(sb, KERN_ERR,
1937 "error: no journal found. "
1938 "mounting ext3 over ext2?");
1939 goto failed_mount2;
1941 err = percpu_counter_init(&sbi->s_freeblocks_counter,
1942 ext3_count_free_blocks(sb));
1943 if (!err) {
1944 err = percpu_counter_init(&sbi->s_freeinodes_counter,
1945 ext3_count_free_inodes(sb));
1947 if (!err) {
1948 err = percpu_counter_init(&sbi->s_dirs_counter,
1949 ext3_count_dirs(sb));
1951 if (err) {
1952 ext3_msg(sb, KERN_ERR, "error: insufficient memory");
1953 ret = err;
1954 goto failed_mount3;
1957 /* We have now updated the journal if required, so we can
1958 * validate the data journaling mode. */
1959 switch (test_opt(sb, DATA_FLAGS)) {
1960 case 0:
1961 /* No mode set, assume a default based on the journal
1962 capabilities: ORDERED_DATA if the journal can
1963 cope, else JOURNAL_DATA */
1964 if (journal_check_available_features
1965 (sbi->s_journal, 0, 0, JFS_FEATURE_INCOMPAT_REVOKE))
1966 set_opt(sbi->s_mount_opt, DEFAULT_DATA_MODE);
1967 else
1968 set_opt(sbi->s_mount_opt, JOURNAL_DATA);
1969 break;
1971 case EXT3_MOUNT_ORDERED_DATA:
1972 case EXT3_MOUNT_WRITEBACK_DATA:
1973 if (!journal_check_available_features
1974 (sbi->s_journal, 0, 0, JFS_FEATURE_INCOMPAT_REVOKE)) {
1975 ext3_msg(sb, KERN_ERR,
1976 "error: journal does not support "
1977 "requested data journaling mode");
1978 goto failed_mount3;
1980 default:
1981 break;
1985 * The journal_load will have done any necessary log recovery,
1986 * so we can safely mount the rest of the filesystem now.
1989 root = ext3_iget(sb, EXT3_ROOT_INO);
1990 if (IS_ERR(root)) {
1991 ext3_msg(sb, KERN_ERR, "error: get root inode failed");
1992 ret = PTR_ERR(root);
1993 goto failed_mount3;
1995 if (!S_ISDIR(root->i_mode) || !root->i_blocks || !root->i_size) {
1996 iput(root);
1997 ext3_msg(sb, KERN_ERR, "error: corrupt root inode, run e2fsck");
1998 goto failed_mount3;
2000 sb->s_root = d_alloc_root(root);
2001 if (!sb->s_root) {
2002 ext3_msg(sb, KERN_ERR, "error: get root dentry failed");
2003 iput(root);
2004 ret = -ENOMEM;
2005 goto failed_mount3;
2008 ext3_setup_super (sb, es, sb->s_flags & MS_RDONLY);
2010 EXT3_SB(sb)->s_mount_state |= EXT3_ORPHAN_FS;
2011 ext3_orphan_cleanup(sb, es);
2012 EXT3_SB(sb)->s_mount_state &= ~EXT3_ORPHAN_FS;
2013 if (needs_recovery)
2014 ext3_msg(sb, KERN_INFO, "recovery complete");
2015 ext3_mark_recovery_complete(sb, es);
2016 ext3_msg(sb, KERN_INFO, "mounted filesystem with %s data mode",
2017 test_opt(sb,DATA_FLAGS) == EXT3_MOUNT_JOURNAL_DATA ? "journal":
2018 test_opt(sb,DATA_FLAGS) == EXT3_MOUNT_ORDERED_DATA ? "ordered":
2019 "writeback");
2021 return 0;
2023 cantfind_ext3:
2024 if (!silent)
2025 ext3_msg(sb, KERN_INFO,
2026 "error: can't find ext3 filesystem on dev %s.",
2027 sb->s_id);
2028 goto failed_mount;
2030 failed_mount3:
2031 percpu_counter_destroy(&sbi->s_freeblocks_counter);
2032 percpu_counter_destroy(&sbi->s_freeinodes_counter);
2033 percpu_counter_destroy(&sbi->s_dirs_counter);
2034 journal_destroy(sbi->s_journal);
2035 failed_mount2:
2036 for (i = 0; i < db_count; i++)
2037 brelse(sbi->s_group_desc[i]);
2038 kfree(sbi->s_group_desc);
2039 failed_mount:
2040 #ifdef CONFIG_QUOTA
2041 for (i = 0; i < MAXQUOTAS; i++)
2042 kfree(sbi->s_qf_names[i]);
2043 #endif
2044 ext3_blkdev_remove(sbi);
2045 brelse(bh);
2046 out_fail:
2047 sb->s_fs_info = NULL;
2048 kfree(sbi->s_blockgroup_lock);
2049 kfree(sbi);
2050 return ret;
2054 * Setup any per-fs journal parameters now. We'll do this both on
2055 * initial mount, once the journal has been initialised but before we've
2056 * done any recovery; and again on any subsequent remount.
2058 static void ext3_init_journal_params(struct super_block *sb, journal_t *journal)
2060 struct ext3_sb_info *sbi = EXT3_SB(sb);
2062 if (sbi->s_commit_interval)
2063 journal->j_commit_interval = sbi->s_commit_interval;
2064 /* We could also set up an ext3-specific default for the commit
2065 * interval here, but for now we'll just fall back to the jbd
2066 * default. */
2068 spin_lock(&journal->j_state_lock);
2069 if (test_opt(sb, BARRIER))
2070 journal->j_flags |= JFS_BARRIER;
2071 else
2072 journal->j_flags &= ~JFS_BARRIER;
2073 if (test_opt(sb, DATA_ERR_ABORT))
2074 journal->j_flags |= JFS_ABORT_ON_SYNCDATA_ERR;
2075 else
2076 journal->j_flags &= ~JFS_ABORT_ON_SYNCDATA_ERR;
2077 spin_unlock(&journal->j_state_lock);
2080 static journal_t *ext3_get_journal(struct super_block *sb,
2081 unsigned int journal_inum)
2083 struct inode *journal_inode;
2084 journal_t *journal;
2086 /* First, test for the existence of a valid inode on disk. Bad
2087 * things happen if we iget() an unused inode, as the subsequent
2088 * iput() will try to delete it. */
2090 journal_inode = ext3_iget(sb, journal_inum);
2091 if (IS_ERR(journal_inode)) {
2092 ext3_msg(sb, KERN_ERR, "error: no journal found");
2093 return NULL;
2095 if (!journal_inode->i_nlink) {
2096 make_bad_inode(journal_inode);
2097 iput(journal_inode);
2098 ext3_msg(sb, KERN_ERR, "error: journal inode is deleted");
2099 return NULL;
2102 jbd_debug(2, "Journal inode found at %p: %Ld bytes\n",
2103 journal_inode, journal_inode->i_size);
2104 if (!S_ISREG(journal_inode->i_mode)) {
2105 ext3_msg(sb, KERN_ERR, "error: invalid journal inode");
2106 iput(journal_inode);
2107 return NULL;
2110 journal = journal_init_inode(journal_inode);
2111 if (!journal) {
2112 ext3_msg(sb, KERN_ERR, "error: could not load journal inode");
2113 iput(journal_inode);
2114 return NULL;
2116 journal->j_private = sb;
2117 ext3_init_journal_params(sb, journal);
2118 return journal;
2121 static journal_t *ext3_get_dev_journal(struct super_block *sb,
2122 dev_t j_dev)
2124 struct buffer_head * bh;
2125 journal_t *journal;
2126 ext3_fsblk_t start;
2127 ext3_fsblk_t len;
2128 int hblock, blocksize;
2129 ext3_fsblk_t sb_block;
2130 unsigned long offset;
2131 struct ext3_super_block * es;
2132 struct block_device *bdev;
2134 bdev = ext3_blkdev_get(j_dev, sb);
2135 if (bdev == NULL)
2136 return NULL;
2138 if (bd_claim(bdev, sb)) {
2139 ext3_msg(sb, KERN_ERR,
2140 "error: failed to claim external journal device");
2141 blkdev_put(bdev, FMODE_READ|FMODE_WRITE);
2142 return NULL;
2145 blocksize = sb->s_blocksize;
2146 hblock = bdev_logical_block_size(bdev);
2147 if (blocksize < hblock) {
2148 ext3_msg(sb, KERN_ERR,
2149 "error: blocksize too small for journal device");
2150 goto out_bdev;
2153 sb_block = EXT3_MIN_BLOCK_SIZE / blocksize;
2154 offset = EXT3_MIN_BLOCK_SIZE % blocksize;
2155 set_blocksize(bdev, blocksize);
2156 if (!(bh = __bread(bdev, sb_block, blocksize))) {
2157 ext3_msg(sb, KERN_ERR, "error: couldn't read superblock of "
2158 "external journal");
2159 goto out_bdev;
2162 es = (struct ext3_super_block *) (bh->b_data + offset);
2163 if ((le16_to_cpu(es->s_magic) != EXT3_SUPER_MAGIC) ||
2164 !(le32_to_cpu(es->s_feature_incompat) &
2165 EXT3_FEATURE_INCOMPAT_JOURNAL_DEV)) {
2166 ext3_msg(sb, KERN_ERR, "error: external journal has "
2167 "bad superblock");
2168 brelse(bh);
2169 goto out_bdev;
2172 if (memcmp(EXT3_SB(sb)->s_es->s_journal_uuid, es->s_uuid, 16)) {
2173 ext3_msg(sb, KERN_ERR, "error: journal UUID does not match");
2174 brelse(bh);
2175 goto out_bdev;
2178 len = le32_to_cpu(es->s_blocks_count);
2179 start = sb_block + 1;
2180 brelse(bh); /* we're done with the superblock */
2182 journal = journal_init_dev(bdev, sb->s_bdev,
2183 start, len, blocksize);
2184 if (!journal) {
2185 ext3_msg(sb, KERN_ERR,
2186 "error: failed to create device journal");
2187 goto out_bdev;
2189 journal->j_private = sb;
2190 ll_rw_block(READ, 1, &journal->j_sb_buffer);
2191 wait_on_buffer(journal->j_sb_buffer);
2192 if (!buffer_uptodate(journal->j_sb_buffer)) {
2193 ext3_msg(sb, KERN_ERR, "I/O error on journal device");
2194 goto out_journal;
2196 if (be32_to_cpu(journal->j_superblock->s_nr_users) != 1) {
2197 ext3_msg(sb, KERN_ERR,
2198 "error: external journal has more than one "
2199 "user (unsupported) - %d",
2200 be32_to_cpu(journal->j_superblock->s_nr_users));
2201 goto out_journal;
2203 EXT3_SB(sb)->journal_bdev = bdev;
2204 ext3_init_journal_params(sb, journal);
2205 return journal;
2206 out_journal:
2207 journal_destroy(journal);
2208 out_bdev:
2209 ext3_blkdev_put(bdev);
2210 return NULL;
2213 static int ext3_load_journal(struct super_block *sb,
2214 struct ext3_super_block *es,
2215 unsigned long journal_devnum)
2217 journal_t *journal;
2218 unsigned int journal_inum = le32_to_cpu(es->s_journal_inum);
2219 dev_t journal_dev;
2220 int err = 0;
2221 int really_read_only;
2223 if (journal_devnum &&
2224 journal_devnum != le32_to_cpu(es->s_journal_dev)) {
2225 ext3_msg(sb, KERN_INFO, "external journal device major/minor "
2226 "numbers have changed");
2227 journal_dev = new_decode_dev(journal_devnum);
2228 } else
2229 journal_dev = new_decode_dev(le32_to_cpu(es->s_journal_dev));
2231 really_read_only = bdev_read_only(sb->s_bdev);
2234 * Are we loading a blank journal or performing recovery after a
2235 * crash? For recovery, we need to check in advance whether we
2236 * can get read-write access to the device.
2239 if (EXT3_HAS_INCOMPAT_FEATURE(sb, EXT3_FEATURE_INCOMPAT_RECOVER)) {
2240 if (sb->s_flags & MS_RDONLY) {
2241 ext3_msg(sb, KERN_INFO,
2242 "recovery required on readonly filesystem");
2243 if (really_read_only) {
2244 ext3_msg(sb, KERN_ERR, "error: write access "
2245 "unavailable, cannot proceed");
2246 return -EROFS;
2248 ext3_msg(sb, KERN_INFO,
2249 "write access will be enabled during recovery");
2253 if (journal_inum && journal_dev) {
2254 ext3_msg(sb, KERN_ERR, "error: filesystem has both journal "
2255 "and inode journals");
2256 return -EINVAL;
2259 if (journal_inum) {
2260 if (!(journal = ext3_get_journal(sb, journal_inum)))
2261 return -EINVAL;
2262 } else {
2263 if (!(journal = ext3_get_dev_journal(sb, journal_dev)))
2264 return -EINVAL;
2267 if (!(journal->j_flags & JFS_BARRIER))
2268 printk(KERN_INFO "EXT3-fs: barriers not enabled\n");
2270 if (!really_read_only && test_opt(sb, UPDATE_JOURNAL)) {
2271 err = journal_update_format(journal);
2272 if (err) {
2273 ext3_msg(sb, KERN_ERR, "error updating journal");
2274 journal_destroy(journal);
2275 return err;
2279 if (!EXT3_HAS_INCOMPAT_FEATURE(sb, EXT3_FEATURE_INCOMPAT_RECOVER))
2280 err = journal_wipe(journal, !really_read_only);
2281 if (!err)
2282 err = journal_load(journal);
2284 if (err) {
2285 ext3_msg(sb, KERN_ERR, "error loading journal");
2286 journal_destroy(journal);
2287 return err;
2290 EXT3_SB(sb)->s_journal = journal;
2291 ext3_clear_journal_err(sb, es);
2293 if (journal_devnum &&
2294 journal_devnum != le32_to_cpu(es->s_journal_dev)) {
2295 es->s_journal_dev = cpu_to_le32(journal_devnum);
2297 /* Make sure we flush the recovery flag to disk. */
2298 ext3_commit_super(sb, es, 1);
2301 return 0;
2304 static int ext3_create_journal(struct super_block *sb,
2305 struct ext3_super_block *es,
2306 unsigned int journal_inum)
2308 journal_t *journal;
2309 int err;
2311 if (sb->s_flags & MS_RDONLY) {
2312 ext3_msg(sb, KERN_ERR,
2313 "error: readonly filesystem when trying to "
2314 "create journal");
2315 return -EROFS;
2318 journal = ext3_get_journal(sb, journal_inum);
2319 if (!journal)
2320 return -EINVAL;
2322 ext3_msg(sb, KERN_INFO, "creating new journal on inode %u",
2323 journal_inum);
2325 err = journal_create(journal);
2326 if (err) {
2327 ext3_msg(sb, KERN_ERR, "error creating journal");
2328 journal_destroy(journal);
2329 return -EIO;
2332 EXT3_SB(sb)->s_journal = journal;
2334 ext3_update_dynamic_rev(sb);
2335 EXT3_SET_INCOMPAT_FEATURE(sb, EXT3_FEATURE_INCOMPAT_RECOVER);
2336 EXT3_SET_COMPAT_FEATURE(sb, EXT3_FEATURE_COMPAT_HAS_JOURNAL);
2338 es->s_journal_inum = cpu_to_le32(journal_inum);
2340 /* Make sure we flush the recovery flag to disk. */
2341 ext3_commit_super(sb, es, 1);
2343 return 0;
2346 static int ext3_commit_super(struct super_block *sb,
2347 struct ext3_super_block *es,
2348 int sync)
2350 struct buffer_head *sbh = EXT3_SB(sb)->s_sbh;
2351 int error = 0;
2353 if (!sbh)
2354 return error;
2356 if (buffer_write_io_error(sbh)) {
2358 * Oh, dear. A previous attempt to write the
2359 * superblock failed. This could happen because the
2360 * USB device was yanked out. Or it could happen to
2361 * be a transient write error and maybe the block will
2362 * be remapped. Nothing we can do but to retry the
2363 * write and hope for the best.
2365 ext3_msg(sb, KERN_ERR, "previous I/O error to "
2366 "superblock detected");
2367 clear_buffer_write_io_error(sbh);
2368 set_buffer_uptodate(sbh);
2371 * If the file system is mounted read-only, don't update the
2372 * superblock write time. This avoids updating the superblock
2373 * write time when we are mounting the root file system
2374 * read/only but we need to replay the journal; at that point,
2375 * for people who are east of GMT and who make their clock
2376 * tick in localtime for Windows bug-for-bug compatibility,
2377 * the clock is set in the future, and this will cause e2fsck
2378 * to complain and force a full file system check.
2380 if (!(sb->s_flags & MS_RDONLY))
2381 es->s_wtime = cpu_to_le32(get_seconds());
2382 es->s_free_blocks_count = cpu_to_le32(ext3_count_free_blocks(sb));
2383 es->s_free_inodes_count = cpu_to_le32(ext3_count_free_inodes(sb));
2384 BUFFER_TRACE(sbh, "marking dirty");
2385 mark_buffer_dirty(sbh);
2386 if (sync) {
2387 error = sync_dirty_buffer(sbh);
2388 if (buffer_write_io_error(sbh)) {
2389 ext3_msg(sb, KERN_ERR, "I/O error while writing "
2390 "superblock");
2391 clear_buffer_write_io_error(sbh);
2392 set_buffer_uptodate(sbh);
2395 return error;
2400 * Have we just finished recovery? If so, and if we are mounting (or
2401 * remounting) the filesystem readonly, then we will end up with a
2402 * consistent fs on disk. Record that fact.
2404 static void ext3_mark_recovery_complete(struct super_block * sb,
2405 struct ext3_super_block * es)
2407 journal_t *journal = EXT3_SB(sb)->s_journal;
2409 journal_lock_updates(journal);
2410 if (journal_flush(journal) < 0)
2411 goto out;
2413 if (EXT3_HAS_INCOMPAT_FEATURE(sb, EXT3_FEATURE_INCOMPAT_RECOVER) &&
2414 sb->s_flags & MS_RDONLY) {
2415 EXT3_CLEAR_INCOMPAT_FEATURE(sb, EXT3_FEATURE_INCOMPAT_RECOVER);
2416 ext3_commit_super(sb, es, 1);
2419 out:
2420 journal_unlock_updates(journal);
2424 * If we are mounting (or read-write remounting) a filesystem whose journal
2425 * has recorded an error from a previous lifetime, move that error to the
2426 * main filesystem now.
2428 static void ext3_clear_journal_err(struct super_block *sb,
2429 struct ext3_super_block *es)
2431 journal_t *journal;
2432 int j_errno;
2433 const char *errstr;
2435 journal = EXT3_SB(sb)->s_journal;
2438 * Now check for any error status which may have been recorded in the
2439 * journal by a prior ext3_error() or ext3_abort()
2442 j_errno = journal_errno(journal);
2443 if (j_errno) {
2444 char nbuf[16];
2446 errstr = ext3_decode_error(sb, j_errno, nbuf);
2447 ext3_warning(sb, __func__, "Filesystem error recorded "
2448 "from previous mount: %s", errstr);
2449 ext3_warning(sb, __func__, "Marking fs in need of "
2450 "filesystem check.");
2452 EXT3_SB(sb)->s_mount_state |= EXT3_ERROR_FS;
2453 es->s_state |= cpu_to_le16(EXT3_ERROR_FS);
2454 ext3_commit_super (sb, es, 1);
2456 journal_clear_err(journal);
2461 * Force the running and committing transactions to commit,
2462 * and wait on the commit.
2464 int ext3_force_commit(struct super_block *sb)
2466 journal_t *journal;
2467 int ret;
2469 if (sb->s_flags & MS_RDONLY)
2470 return 0;
2472 journal = EXT3_SB(sb)->s_journal;
2473 ret = ext3_journal_force_commit(journal);
2474 return ret;
2477 static int ext3_sync_fs(struct super_block *sb, int wait)
2479 tid_t target;
2481 if (journal_start_commit(EXT3_SB(sb)->s_journal, &target)) {
2482 if (wait)
2483 log_wait_commit(EXT3_SB(sb)->s_journal, target);
2485 return 0;
2489 * LVM calls this function before a (read-only) snapshot is created. This
2490 * gives us a chance to flush the journal completely and mark the fs clean.
2492 static int ext3_freeze(struct super_block *sb)
2494 int error = 0;
2495 journal_t *journal;
2497 if (!(sb->s_flags & MS_RDONLY)) {
2498 journal = EXT3_SB(sb)->s_journal;
2500 /* Now we set up the journal barrier. */
2501 journal_lock_updates(journal);
2504 * We don't want to clear needs_recovery flag when we failed
2505 * to flush the journal.
2507 error = journal_flush(journal);
2508 if (error < 0)
2509 goto out;
2511 /* Journal blocked and flushed, clear needs_recovery flag. */
2512 EXT3_CLEAR_INCOMPAT_FEATURE(sb, EXT3_FEATURE_INCOMPAT_RECOVER);
2513 error = ext3_commit_super(sb, EXT3_SB(sb)->s_es, 1);
2514 if (error)
2515 goto out;
2517 return 0;
2519 out:
2520 journal_unlock_updates(journal);
2521 return error;
2525 * Called by LVM after the snapshot is done. We need to reset the RECOVER
2526 * flag here, even though the filesystem is not technically dirty yet.
2528 static int ext3_unfreeze(struct super_block *sb)
2530 if (!(sb->s_flags & MS_RDONLY)) {
2531 lock_super(sb);
2532 /* Reser the needs_recovery flag before the fs is unlocked. */
2533 EXT3_SET_INCOMPAT_FEATURE(sb, EXT3_FEATURE_INCOMPAT_RECOVER);
2534 ext3_commit_super(sb, EXT3_SB(sb)->s_es, 1);
2535 unlock_super(sb);
2536 journal_unlock_updates(EXT3_SB(sb)->s_journal);
2538 return 0;
2541 static int ext3_remount (struct super_block * sb, int * flags, char * data)
2543 struct ext3_super_block * es;
2544 struct ext3_sb_info *sbi = EXT3_SB(sb);
2545 ext3_fsblk_t n_blocks_count = 0;
2546 unsigned long old_sb_flags;
2547 struct ext3_mount_options old_opts;
2548 int enable_quota = 0;
2549 int err;
2550 #ifdef CONFIG_QUOTA
2551 int i;
2552 #endif
2554 /* Store the original options */
2555 lock_super(sb);
2556 old_sb_flags = sb->s_flags;
2557 old_opts.s_mount_opt = sbi->s_mount_opt;
2558 old_opts.s_resuid = sbi->s_resuid;
2559 old_opts.s_resgid = sbi->s_resgid;
2560 old_opts.s_commit_interval = sbi->s_commit_interval;
2561 #ifdef CONFIG_QUOTA
2562 old_opts.s_jquota_fmt = sbi->s_jquota_fmt;
2563 for (i = 0; i < MAXQUOTAS; i++)
2564 old_opts.s_qf_names[i] = sbi->s_qf_names[i];
2565 #endif
2568 * Allow the "check" option to be passed as a remount option.
2570 if (!parse_options(data, sb, NULL, NULL, &n_blocks_count, 1)) {
2571 err = -EINVAL;
2572 goto restore_opts;
2575 if (test_opt(sb, ABORT))
2576 ext3_abort(sb, __func__, "Abort forced by user");
2578 sb->s_flags = (sb->s_flags & ~MS_POSIXACL) |
2579 (test_opt(sb, POSIX_ACL) ? MS_POSIXACL : 0);
2581 es = sbi->s_es;
2583 ext3_init_journal_params(sb, sbi->s_journal);
2585 if ((*flags & MS_RDONLY) != (sb->s_flags & MS_RDONLY) ||
2586 n_blocks_count > le32_to_cpu(es->s_blocks_count)) {
2587 if (test_opt(sb, ABORT)) {
2588 err = -EROFS;
2589 goto restore_opts;
2592 if (*flags & MS_RDONLY) {
2593 err = dquot_suspend(sb, -1);
2594 if (err < 0)
2595 goto restore_opts;
2598 * First of all, the unconditional stuff we have to do
2599 * to disable replay of the journal when we next remount
2601 sb->s_flags |= MS_RDONLY;
2604 * OK, test if we are remounting a valid rw partition
2605 * readonly, and if so set the rdonly flag and then
2606 * mark the partition as valid again.
2608 if (!(es->s_state & cpu_to_le16(EXT3_VALID_FS)) &&
2609 (sbi->s_mount_state & EXT3_VALID_FS))
2610 es->s_state = cpu_to_le16(sbi->s_mount_state);
2612 ext3_mark_recovery_complete(sb, es);
2613 } else {
2614 __le32 ret;
2615 if ((ret = EXT3_HAS_RO_COMPAT_FEATURE(sb,
2616 ~EXT3_FEATURE_RO_COMPAT_SUPP))) {
2617 ext3_msg(sb, KERN_WARNING,
2618 "warning: couldn't remount RDWR "
2619 "because of unsupported optional "
2620 "features (%x)", le32_to_cpu(ret));
2621 err = -EROFS;
2622 goto restore_opts;
2626 * If we have an unprocessed orphan list hanging
2627 * around from a previously readonly bdev mount,
2628 * require a full umount/remount for now.
2630 if (es->s_last_orphan) {
2631 ext3_msg(sb, KERN_WARNING, "warning: couldn't "
2632 "remount RDWR because of unprocessed "
2633 "orphan inode list. Please "
2634 "umount/remount instead.");
2635 err = -EINVAL;
2636 goto restore_opts;
2640 * Mounting a RDONLY partition read-write, so reread
2641 * and store the current valid flag. (It may have
2642 * been changed by e2fsck since we originally mounted
2643 * the partition.)
2645 ext3_clear_journal_err(sb, es);
2646 sbi->s_mount_state = le16_to_cpu(es->s_state);
2647 if ((err = ext3_group_extend(sb, es, n_blocks_count)))
2648 goto restore_opts;
2649 if (!ext3_setup_super (sb, es, 0))
2650 sb->s_flags &= ~MS_RDONLY;
2651 enable_quota = 1;
2654 #ifdef CONFIG_QUOTA
2655 /* Release old quota file names */
2656 for (i = 0; i < MAXQUOTAS; i++)
2657 if (old_opts.s_qf_names[i] &&
2658 old_opts.s_qf_names[i] != sbi->s_qf_names[i])
2659 kfree(old_opts.s_qf_names[i]);
2660 #endif
2661 unlock_super(sb);
2663 if (enable_quota)
2664 dquot_resume(sb, -1);
2665 return 0;
2666 restore_opts:
2667 sb->s_flags = old_sb_flags;
2668 sbi->s_mount_opt = old_opts.s_mount_opt;
2669 sbi->s_resuid = old_opts.s_resuid;
2670 sbi->s_resgid = old_opts.s_resgid;
2671 sbi->s_commit_interval = old_opts.s_commit_interval;
2672 #ifdef CONFIG_QUOTA
2673 sbi->s_jquota_fmt = old_opts.s_jquota_fmt;
2674 for (i = 0; i < MAXQUOTAS; i++) {
2675 if (sbi->s_qf_names[i] &&
2676 old_opts.s_qf_names[i] != sbi->s_qf_names[i])
2677 kfree(sbi->s_qf_names[i]);
2678 sbi->s_qf_names[i] = old_opts.s_qf_names[i];
2680 #endif
2681 unlock_super(sb);
2682 return err;
2685 static int ext3_statfs (struct dentry * dentry, struct kstatfs * buf)
2687 struct super_block *sb = dentry->d_sb;
2688 struct ext3_sb_info *sbi = EXT3_SB(sb);
2689 struct ext3_super_block *es = sbi->s_es;
2690 u64 fsid;
2692 if (test_opt(sb, MINIX_DF)) {
2693 sbi->s_overhead_last = 0;
2694 } else if (sbi->s_blocks_last != le32_to_cpu(es->s_blocks_count)) {
2695 unsigned long ngroups = sbi->s_groups_count, i;
2696 ext3_fsblk_t overhead = 0;
2697 smp_rmb();
2700 * Compute the overhead (FS structures). This is constant
2701 * for a given filesystem unless the number of block groups
2702 * changes so we cache the previous value until it does.
2706 * All of the blocks before first_data_block are
2707 * overhead
2709 overhead = le32_to_cpu(es->s_first_data_block);
2712 * Add the overhead attributed to the superblock and
2713 * block group descriptors. If the sparse superblocks
2714 * feature is turned on, then not all groups have this.
2716 for (i = 0; i < ngroups; i++) {
2717 overhead += ext3_bg_has_super(sb, i) +
2718 ext3_bg_num_gdb(sb, i);
2719 cond_resched();
2723 * Every block group has an inode bitmap, a block
2724 * bitmap, and an inode table.
2726 overhead += ngroups * (2 + sbi->s_itb_per_group);
2727 sbi->s_overhead_last = overhead;
2728 smp_wmb();
2729 sbi->s_blocks_last = le32_to_cpu(es->s_blocks_count);
2732 buf->f_type = EXT3_SUPER_MAGIC;
2733 buf->f_bsize = sb->s_blocksize;
2734 buf->f_blocks = le32_to_cpu(es->s_blocks_count) - sbi->s_overhead_last;
2735 buf->f_bfree = percpu_counter_sum_positive(&sbi->s_freeblocks_counter);
2736 buf->f_bavail = buf->f_bfree - le32_to_cpu(es->s_r_blocks_count);
2737 if (buf->f_bfree < le32_to_cpu(es->s_r_blocks_count))
2738 buf->f_bavail = 0;
2739 buf->f_files = le32_to_cpu(es->s_inodes_count);
2740 buf->f_ffree = percpu_counter_sum_positive(&sbi->s_freeinodes_counter);
2741 buf->f_namelen = EXT3_NAME_LEN;
2742 fsid = le64_to_cpup((void *)es->s_uuid) ^
2743 le64_to_cpup((void *)es->s_uuid + sizeof(u64));
2744 buf->f_fsid.val[0] = fsid & 0xFFFFFFFFUL;
2745 buf->f_fsid.val[1] = (fsid >> 32) & 0xFFFFFFFFUL;
2746 return 0;
2749 /* Helper function for writing quotas on sync - we need to start transaction before quota file
2750 * is locked for write. Otherwise the are possible deadlocks:
2751 * Process 1 Process 2
2752 * ext3_create() quota_sync()
2753 * journal_start() write_dquot()
2754 * dquot_initialize() down(dqio_mutex)
2755 * down(dqio_mutex) journal_start()
2759 #ifdef CONFIG_QUOTA
2761 static inline struct inode *dquot_to_inode(struct dquot *dquot)
2763 return sb_dqopt(dquot->dq_sb)->files[dquot->dq_type];
2766 static int ext3_write_dquot(struct dquot *dquot)
2768 int ret, err;
2769 handle_t *handle;
2770 struct inode *inode;
2772 inode = dquot_to_inode(dquot);
2773 handle = ext3_journal_start(inode,
2774 EXT3_QUOTA_TRANS_BLOCKS(dquot->dq_sb));
2775 if (IS_ERR(handle))
2776 return PTR_ERR(handle);
2777 ret = dquot_commit(dquot);
2778 err = ext3_journal_stop(handle);
2779 if (!ret)
2780 ret = err;
2781 return ret;
2784 static int ext3_acquire_dquot(struct dquot *dquot)
2786 int ret, err;
2787 handle_t *handle;
2789 handle = ext3_journal_start(dquot_to_inode(dquot),
2790 EXT3_QUOTA_INIT_BLOCKS(dquot->dq_sb));
2791 if (IS_ERR(handle))
2792 return PTR_ERR(handle);
2793 ret = dquot_acquire(dquot);
2794 err = ext3_journal_stop(handle);
2795 if (!ret)
2796 ret = err;
2797 return ret;
2800 static int ext3_release_dquot(struct dquot *dquot)
2802 int ret, err;
2803 handle_t *handle;
2805 handle = ext3_journal_start(dquot_to_inode(dquot),
2806 EXT3_QUOTA_DEL_BLOCKS(dquot->dq_sb));
2807 if (IS_ERR(handle)) {
2808 /* Release dquot anyway to avoid endless cycle in dqput() */
2809 dquot_release(dquot);
2810 return PTR_ERR(handle);
2812 ret = dquot_release(dquot);
2813 err = ext3_journal_stop(handle);
2814 if (!ret)
2815 ret = err;
2816 return ret;
2819 static int ext3_mark_dquot_dirty(struct dquot *dquot)
2821 /* Are we journaling quotas? */
2822 if (EXT3_SB(dquot->dq_sb)->s_qf_names[USRQUOTA] ||
2823 EXT3_SB(dquot->dq_sb)->s_qf_names[GRPQUOTA]) {
2824 dquot_mark_dquot_dirty(dquot);
2825 return ext3_write_dquot(dquot);
2826 } else {
2827 return dquot_mark_dquot_dirty(dquot);
2831 static int ext3_write_info(struct super_block *sb, int type)
2833 int ret, err;
2834 handle_t *handle;
2836 /* Data block + inode block */
2837 handle = ext3_journal_start(sb->s_root->d_inode, 2);
2838 if (IS_ERR(handle))
2839 return PTR_ERR(handle);
2840 ret = dquot_commit_info(sb, type);
2841 err = ext3_journal_stop(handle);
2842 if (!ret)
2843 ret = err;
2844 return ret;
2848 * Turn on quotas during mount time - we need to find
2849 * the quota file and such...
2851 static int ext3_quota_on_mount(struct super_block *sb, int type)
2853 return dquot_quota_on_mount(sb, EXT3_SB(sb)->s_qf_names[type],
2854 EXT3_SB(sb)->s_jquota_fmt, type);
2858 * Standard function to be called on quota_on
2860 static int ext3_quota_on(struct super_block *sb, int type, int format_id,
2861 char *name)
2863 int err;
2864 struct path path;
2866 if (!test_opt(sb, QUOTA))
2867 return -EINVAL;
2869 err = kern_path(name, LOOKUP_FOLLOW, &path);
2870 if (err)
2871 return err;
2873 /* Quotafile not on the same filesystem? */
2874 if (path.mnt->mnt_sb != sb) {
2875 path_put(&path);
2876 return -EXDEV;
2878 /* Journaling quota? */
2879 if (EXT3_SB(sb)->s_qf_names[type]) {
2880 /* Quotafile not of fs root? */
2881 if (path.dentry->d_parent != sb->s_root)
2882 ext3_msg(sb, KERN_WARNING,
2883 "warning: Quota file not on filesystem root. "
2884 "Journaled quota will not work.");
2888 * When we journal data on quota file, we have to flush journal to see
2889 * all updates to the file when we bypass pagecache...
2891 if (ext3_should_journal_data(path.dentry->d_inode)) {
2893 * We don't need to lock updates but journal_flush() could
2894 * otherwise be livelocked...
2896 journal_lock_updates(EXT3_SB(sb)->s_journal);
2897 err = journal_flush(EXT3_SB(sb)->s_journal);
2898 journal_unlock_updates(EXT3_SB(sb)->s_journal);
2899 if (err) {
2900 path_put(&path);
2901 return err;
2905 err = dquot_quota_on_path(sb, type, format_id, &path);
2906 path_put(&path);
2907 return err;
2910 /* Read data from quotafile - avoid pagecache and such because we cannot afford
2911 * acquiring the locks... As quota files are never truncated and quota code
2912 * itself serializes the operations (and noone else should touch the files)
2913 * we don't have to be afraid of races */
2914 static ssize_t ext3_quota_read(struct super_block *sb, int type, char *data,
2915 size_t len, loff_t off)
2917 struct inode *inode = sb_dqopt(sb)->files[type];
2918 sector_t blk = off >> EXT3_BLOCK_SIZE_BITS(sb);
2919 int err = 0;
2920 int offset = off & (sb->s_blocksize - 1);
2921 int tocopy;
2922 size_t toread;
2923 struct buffer_head *bh;
2924 loff_t i_size = i_size_read(inode);
2926 if (off > i_size)
2927 return 0;
2928 if (off+len > i_size)
2929 len = i_size-off;
2930 toread = len;
2931 while (toread > 0) {
2932 tocopy = sb->s_blocksize - offset < toread ?
2933 sb->s_blocksize - offset : toread;
2934 bh = ext3_bread(NULL, inode, blk, 0, &err);
2935 if (err)
2936 return err;
2937 if (!bh) /* A hole? */
2938 memset(data, 0, tocopy);
2939 else
2940 memcpy(data, bh->b_data+offset, tocopy);
2941 brelse(bh);
2942 offset = 0;
2943 toread -= tocopy;
2944 data += tocopy;
2945 blk++;
2947 return len;
2950 /* Write to quotafile (we know the transaction is already started and has
2951 * enough credits) */
2952 static ssize_t ext3_quota_write(struct super_block *sb, int type,
2953 const char *data, size_t len, loff_t off)
2955 struct inode *inode = sb_dqopt(sb)->files[type];
2956 sector_t blk = off >> EXT3_BLOCK_SIZE_BITS(sb);
2957 int err = 0;
2958 int offset = off & (sb->s_blocksize - 1);
2959 int journal_quota = EXT3_SB(sb)->s_qf_names[type] != NULL;
2960 struct buffer_head *bh;
2961 handle_t *handle = journal_current_handle();
2963 if (!handle) {
2964 ext3_msg(sb, KERN_WARNING,
2965 "warning: quota write (off=%llu, len=%llu)"
2966 " cancelled because transaction is not started.",
2967 (unsigned long long)off, (unsigned long long)len);
2968 return -EIO;
2972 * Since we account only one data block in transaction credits,
2973 * then it is impossible to cross a block boundary.
2975 if (sb->s_blocksize - offset < len) {
2976 ext3_msg(sb, KERN_WARNING, "Quota write (off=%llu, len=%llu)"
2977 " cancelled because not block aligned",
2978 (unsigned long long)off, (unsigned long long)len);
2979 return -EIO;
2981 mutex_lock_nested(&inode->i_mutex, I_MUTEX_QUOTA);
2982 bh = ext3_bread(handle, inode, blk, 1, &err);
2983 if (!bh)
2984 goto out;
2985 if (journal_quota) {
2986 err = ext3_journal_get_write_access(handle, bh);
2987 if (err) {
2988 brelse(bh);
2989 goto out;
2992 lock_buffer(bh);
2993 memcpy(bh->b_data+offset, data, len);
2994 flush_dcache_page(bh->b_page);
2995 unlock_buffer(bh);
2996 if (journal_quota)
2997 err = ext3_journal_dirty_metadata(handle, bh);
2998 else {
2999 /* Always do at least ordered writes for quotas */
3000 err = ext3_journal_dirty_data(handle, bh);
3001 mark_buffer_dirty(bh);
3003 brelse(bh);
3004 out:
3005 if (err) {
3006 mutex_unlock(&inode->i_mutex);
3007 return err;
3009 if (inode->i_size < off + len) {
3010 i_size_write(inode, off + len);
3011 EXT3_I(inode)->i_disksize = inode->i_size;
3013 inode->i_version++;
3014 inode->i_mtime = inode->i_ctime = CURRENT_TIME;
3015 ext3_mark_inode_dirty(handle, inode);
3016 mutex_unlock(&inode->i_mutex);
3017 return len;
3020 #endif
3022 static struct dentry *ext3_mount(struct file_system_type *fs_type,
3023 int flags, const char *dev_name, void *data)
3025 return mount_bdev(fs_type, flags, dev_name, data, ext3_fill_super);
3028 static struct file_system_type ext3_fs_type = {
3029 .owner = THIS_MODULE,
3030 .name = "ext3",
3031 .mount = ext3_mount,
3032 .kill_sb = kill_block_super,
3033 .fs_flags = FS_REQUIRES_DEV,
3036 static int __init init_ext3_fs(void)
3038 int err = init_ext3_xattr();
3039 if (err)
3040 return err;
3041 err = init_inodecache();
3042 if (err)
3043 goto out1;
3044 err = register_filesystem(&ext3_fs_type);
3045 if (err)
3046 goto out;
3047 return 0;
3048 out:
3049 destroy_inodecache();
3050 out1:
3051 exit_ext3_xattr();
3052 return err;
3055 static void __exit exit_ext3_fs(void)
3057 unregister_filesystem(&ext3_fs_type);
3058 destroy_inodecache();
3059 exit_ext3_xattr();
3062 MODULE_AUTHOR("Remy Card, Stephen Tweedie, Andrew Morton, Andreas Dilger, Theodore Ts'o and others");
3063 MODULE_DESCRIPTION("Second Extended Filesystem with journaling extensions");
3064 MODULE_LICENSE("GPL");
3065 module_init(init_ext3_fs)
3066 module_exit(exit_ext3_fs)