]> bbs.cooldavid.org Git - net-next-2.6.git/blame - fs/ext3/super.c
git://bbs.cooldavid.org / net-next-2.6.git / blame
? search:
summary | shortlog | log | commit | commitdiff | tree
blob | blame (incremental) | history | HEAD
ext3: init statistics after journal recovery v2
[net-next-2.6.git] / fs / ext3 / super.c
CommitLineData
1da177e4
LT		 1/*
2 * linux/fs/ext3/super.c
3 *
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)
8 *
9 * from
10 *
11 * linux/fs/minix/inode.c
12 *
13 * Copyright (C) 1991, 1992 Linus Torvalds
14 *
15 * Big-endian to little-endian byte-swapping/bitmaps by
16 * David S. Miller (davem@caip.rutgers.edu), 1995
17 */
18
1da177e4
LT		 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>
a5694255 32#include <linux/exportfs.h>
1da177e4
LT		 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>
8fc2751b 38#include <linux/seq_file.h>
3fc74269 39#include <linux/log2.h>
381be254 40
1da177e4 41#include <asm/uaccess.h>
381be254 42
1da177e4
LT		 43#include "xattr.h"
44#include "acl.h"
381be254 45#include "namei.h"
1da177e4 46
bbae8bcc
LT		 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
52
71b96257
JL		 53static int ext3_load_journal(struct super_block *, struct ext3_super_block *,
54 unsigned long journal_devnum);
1da177e4 55static int ext3_create_journal(struct super_block *, struct ext3_super_block *,
eee194e7 56 unsigned int);
c4be0c1d
TS		 57static int ext3_commit_super(struct super_block *sb,
58 struct ext3_super_block *es,
1da177e4
LT		 59 int sync);
60static void ext3_mark_recovery_complete(struct super_block * sb,
61 struct ext3_super_block * es);
62static void ext3_clear_journal_err(struct super_block * sb,
63 struct ext3_super_block * es);
64static int ext3_sync_fs(struct super_block *sb, int wait);
65static const char *ext3_decode_error(struct super_block * sb, int errno,
66 char nbuf[16]);
67static int ext3_remount (struct super_block * sb, int * flags, char * data);
726c3342 68static int ext3_statfs (struct dentry * dentry, struct kstatfs * buf);
c4be0c1d 69static int ext3_unfreeze(struct super_block *sb);
c4be0c1d 70static int ext3_freeze(struct super_block *sb);
1da177e4 71
ae6ddcc5 72/*
1da177e4
LT		 73 * Wrappers for journal_start/end.
74 *
75 * The only special thing we need to do here is to make sure that all
76 * journal_end calls result in the superblock being marked dirty, so
77 * that sync() will call the filesystem's write_super callback if
ae6ddcc5 78 * appropriate.
1da177e4
LT		 79 */
80handle_t *ext3_journal_start_sb(struct super_block *sb, int nblocks)
81{
82 journal_t *journal;
83
84 if (sb->s_flags & MS_RDONLY)
85 return ERR_PTR(-EROFS);
86
87 /* Special case here: if the journal has aborted behind our
88 * backs (eg. EIO in the commit thread), then we still need to
89 * take the FS itself readonly cleanly. */
90 journal = EXT3_SB(sb)->s_journal;
91 if (is_journal_aborted(journal)) {
e05b6b52 92 ext3_abort(sb, __func__,
1da177e4
LT		 93 "Detected aborted journal");
94 return ERR_PTR(-EROFS);
95 }
96
97 return journal_start(journal, nblocks);
98}
99
ae6ddcc5 100/*
1da177e4
LT		 101 * The only special thing we need to do here is to make sure that all
102 * journal_stop calls result in the superblock being marked dirty, so
103 * that sync() will call the filesystem's write_super callback if
ae6ddcc5 104 * appropriate.
1da177e4
LT		 105 */
106int __ext3_journal_stop(const char *where, handle_t *handle)
107{
108 struct super_block *sb;
109 int err;
110 int rc;
111
112 sb = handle->h_transaction->t_journal->j_private;
113 err = handle->h_err;
114 rc = journal_stop(handle);
115
116 if (!err)
117 err = rc;
118 if (err)
119 __ext3_std_error(sb, where, err);
120 return err;
121}
122
123void ext3_journal_abort_handle(const char *caller, const char *err_fn,
124 struct buffer_head *bh, handle_t *handle, int err)
125{
126 char nbuf[16];
127 const char *errstr = ext3_decode_error(NULL, err, nbuf);
128
129 if (bh)
130 BUFFER_TRACE(bh, "abort");
131
132 if (!handle->h_err)
133 handle->h_err = err;
134
135 if (is_handle_aborted(handle))
136 return;
137
4cf46b67
AF		 138 printk(KERN_ERR "EXT3-fs: %s: aborting transaction: %s in %s\n",
139 caller, errstr, err_fn);
1da177e4
LT		 140
141 journal_abort_handle(handle);
142}
143
4cf46b67
AF		 144void ext3_msg(struct super_block *sb, const char *prefix,
145 const char *fmt, ...)
146{
147 va_list args;
148
149 va_start(args, fmt);
150 printk("%sEXT3-fs (%s): ", prefix, sb->s_id);
151 vprintk(fmt, args);
152 printk("\n");
153 va_end(args);
154}
155
1da177e4
LT		 156/* Deal with the reporting of failure conditions on a filesystem such as
157 * inconsistencies detected or read IO failures.
158 *
159 * On ext2, we can store the error state of the filesystem in the
160 * superblock. That is not possible on ext3, because we may have other
161 * write ordering constraints on the superblock which prevent us from
162 * writing it out straight away; and given that the journal is about to
163 * be aborted, we can't rely on the current, or future, transactions to
164 * write out the superblock safely.
165 *
166 * We'll just use the journal_abort() error code to record an error in
d6b198bc 167 * the journal instead. On recovery, the journal will complain about
1da177e4
LT		 168 * that error until we've noted it down and cleared it.
169 */
170
171static void ext3_handle_error(struct super_block *sb)
172{
173 struct ext3_super_block *es = EXT3_SB(sb)->s_es;
174
175 EXT3_SB(sb)->s_mount_state |= EXT3_ERROR_FS;
176 es->s_state |= cpu_to_le16(EXT3_ERROR_FS);
177
178 if (sb->s_flags & MS_RDONLY)
179 return;
180
7543fc7b 181 if (!test_opt (sb, ERRORS_CONT)) {
1da177e4
LT		 182 journal_t *journal = EXT3_SB(sb)->s_journal;
183
e3c96435 184 set_opt(EXT3_SB(sb)->s_mount_opt, ABORT);
1da177e4
LT		 185 if (journal)
186 journal_abort(journal, -EIO);
187 }
7543fc7b 188 if (test_opt (sb, ERRORS_RO)) {
4cf46b67
AF		 189 ext3_msg(sb, KERN_CRIT,
190 "error: remounting filesystem read-only");
7543fc7b
VA		 191 sb->s_flags |= MS_RDONLY;
192 }
193 ext3_commit_super(sb, es, 1);
1da177e4 194 if (test_opt(sb, ERRORS_PANIC))
4cf46b67 195 panic("EXT3-fs (%s): panic forced after error\n",
1da177e4 196 sb->s_id);
1da177e4
LT		 197}
198
199void ext3_error (struct super_block * sb, const char * function,
200 const char * fmt, ...)
201{
202 va_list args;
203
204 va_start(args, fmt);
205 printk(KERN_CRIT "EXT3-fs error (device %s): %s: ",sb->s_id, function);
206 vprintk(fmt, args);
207 printk("\n");
208 va_end(args);
209
210 ext3_handle_error(sb);
211}
212
213static const char *ext3_decode_error(struct super_block * sb, int errno,
214 char nbuf[16])
215{
216 char *errstr = NULL;
217
218 switch (errno) {
219 case -EIO:
220 errstr = "IO failure";
221 break;
222 case -ENOMEM:
223 errstr = "Out of memory";
224 break;
225 case -EROFS:
226 if (!sb || EXT3_SB(sb)->s_journal->j_flags & JFS_ABORT)
227 errstr = "Journal has aborted";
228 else
229 errstr = "Readonly filesystem";
230 break;
231 default:
232 /* If the caller passed in an extra buffer for unknown
233 * errors, textualise them now. Else we just return
234 * NULL. */
235 if (nbuf) {
236 /* Check for truncated error codes... */
237 if (snprintf(nbuf, 16, "error %d", -errno) >= 0)
238 errstr = nbuf;
239 }
240 break;
241 }
242
243 return errstr;
244}
245
246/* __ext3_std_error decodes expected errors from journaling functions
247 * automatically and invokes the appropriate error response. */
248
249void __ext3_std_error (struct super_block * sb, const char * function,
250 int errno)
251{
252 char nbuf[16];
30121624 253 const char *errstr;
1da177e4 254
30121624
ST		 255 /* Special case: if the error is EROFS, and we're not already
256 * inside a transaction, then there's really no point in logging
257 * an error. */
258 if (errno == -EROFS && journal_current_handle() == NULL &&
259 (sb->s_flags & MS_RDONLY))
260 return;
261
262 errstr = ext3_decode_error(sb, errno, nbuf);
4cf46b67 263 ext3_msg(sb, KERN_CRIT, "error in %s: %s", function, errstr);
1da177e4
LT		 264
265 ext3_handle_error(sb);
266}
267
268/*
269 * ext3_abort is a much stronger failure handler than ext3_error. The
270 * abort function may be used to deal with unrecoverable failures such
271 * as journal IO errors or ENOMEM at a critical moment in log management.
272 *
273 * We unconditionally force the filesystem into an ABORT|READONLY state,
274 * unless the error response on the fs has been set to panic in which
275 * case we take the easy way out and panic immediately.
276 */
277
278void ext3_abort (struct super_block * sb, const char * function,
279 const char * fmt, ...)
280{
281 va_list args;
282
1da177e4 283 va_start(args, fmt);
4cf46b67 284 printk(KERN_CRIT "EXT3-fs (%s): error: %s: ", sb->s_id, function);
1da177e4
LT		 285 vprintk(fmt, args);
286 printk("\n");
287 va_end(args);
288
289 if (test_opt(sb, ERRORS_PANIC))
4cf46b67 290 panic("EXT3-fs: panic from previous error\n");
1da177e4
LT		 291
292 if (sb->s_flags & MS_RDONLY)
293 return;
294
4cf46b67
AF		 295 ext3_msg(sb, KERN_CRIT,
296 "error: remounting filesystem read-only");
1da177e4
LT		 297 EXT3_SB(sb)->s_mount_state |= EXT3_ERROR_FS;
298 sb->s_flags |= MS_RDONLY;
e3c96435 299 set_opt(EXT3_SB(sb)->s_mount_opt, ABORT);
44d6f787
HK		 300 if (EXT3_SB(sb)->s_journal)
301 journal_abort(EXT3_SB(sb)->s_journal, -EIO);
1da177e4
LT		 302}
303
304void ext3_warning (struct super_block * sb, const char * function,
305 const char * fmt, ...)
306{
307 va_list args;
308
309 va_start(args, fmt);
4cf46b67 310 printk(KERN_WARNING "EXT3-fs (%s): warning: %s: ",
1da177e4
LT		 311 sb->s_id, function);
312 vprintk(fmt, args);
313 printk("\n");
314 va_end(args);
315}
316
317void ext3_update_dynamic_rev(struct super_block *sb)
318{
319 struct ext3_super_block *es = EXT3_SB(sb)->s_es;
320
321 if (le32_to_cpu(es->s_rev_level) > EXT3_GOOD_OLD_REV)
322 return;
323
4cf46b67
AF		 324 ext3_msg(sb, KERN_WARNING,
325 "warning: updating to rev %d because of "
326 "new feature flag, running e2fsck is recommended",
327 EXT3_DYNAMIC_REV);
1da177e4
LT		 328
329 es->s_first_ino = cpu_to_le32(EXT3_GOOD_OLD_FIRST_INO);
330 es->s_inode_size = cpu_to_le16(EXT3_GOOD_OLD_INODE_SIZE);
331 es->s_rev_level = cpu_to_le32(EXT3_DYNAMIC_REV);
332 /* leave es->s_feature_*compat flags alone */
333 /* es->s_uuid will be set by e2fsck if empty */
334
335 /*
336 * The rest of the superblock fields should be zero, and if not it
337 * means they are likely already in use, so leave them alone. We
338 * can leave it up to e2fsck to clean up any inconsistencies there.
339 */
340}
341
342/*
343 * Open the external journal device
344 */
4cf46b67 345static struct block_device *ext3_blkdev_get(dev_t dev, struct super_block *sb)
1da177e4
LT		 346{
347 struct block_device *bdev;
348 char b[BDEVNAME_SIZE];
349
350 bdev = open_by_devnum(dev, FMODE_READ|FMODE_WRITE);
351 if (IS_ERR(bdev))
352 goto fail;
353 return bdev;
354
355fail:
4cf46b67
AF		 356 ext3_msg(sb, "error: failed to open journal device %s: %ld",
357 __bdevname(dev, b), PTR_ERR(bdev));
358
1da177e4
LT		 359 return NULL;
360}
361
362/*
363 * Release the journal device
364 */
365static int ext3_blkdev_put(struct block_device *bdev)
366{
367 bd_release(bdev);
9a1c3542 368 return blkdev_put(bdev, FMODE_READ|FMODE_WRITE);
1da177e4
LT		 369}
370
371static int ext3_blkdev_remove(struct ext3_sb_info *sbi)
372{
373 struct block_device *bdev;
374 int ret = -ENODEV;
375
376 bdev = sbi->journal_bdev;
377 if (bdev) {
378 ret = ext3_blkdev_put(bdev);
379 sbi->journal_bdev = NULL;
380 }
381 return ret;
382}
383
384static inline struct inode *orphan_list_entry(struct list_head *l)
385{
386 return &list_entry(l, struct ext3_inode_info, i_orphan)->vfs_inode;
387}
388
389static void dump_orphan_list(struct super_block *sb, struct ext3_sb_info *sbi)
390{
391 struct list_head *l;
392
4cf46b67 393 ext3_msg(sb, KERN_ERR, "error: sb orphan head is %d",
1da177e4
LT		 394 le32_to_cpu(sbi->s_es->s_last_orphan));
395
4cf46b67 396 ext3_msg(sb, KERN_ERR, "sb_info orphan list:");
1da177e4
LT		 397 list_for_each(l, &sbi->s_orphan) {
398 struct inode *inode = orphan_list_entry(l);
4cf46b67 399 ext3_msg(sb, KERN_ERR, " "
eee194e7 400 "inode %s:%lu at %p: mode %o, nlink %d, next %d\n",
1da177e4 401 inode->i_sb->s_id, inode->i_ino, inode,
ae6ddcc5 402 inode->i_mode, inode->i_nlink,
1da177e4
LT		 403 NEXT_ORPHAN(inode));
404 }
405}
406
407static void ext3_put_super (struct super_block * sb)
408{
409 struct ext3_sb_info *sbi = EXT3_SB(sb);
410 struct ext3_super_block *es = sbi->s_es;
44d6f787 411 int i, err;
1da177e4 412
6cfd0148
CH		 413 lock_kernel();
414
1da177e4 415 ext3_xattr_put_super(sb);
44d6f787
HK		 416 err = journal_destroy(sbi->s_journal);
417 sbi->s_journal = NULL;
418 if (err < 0)
2d7c820e 419 ext3_abort(sb, __func__, "Couldn't clean up the journal");
44d6f787 420
1da177e4
LT		 421 if (!(sb->s_flags & MS_RDONLY)) {
422 EXT3_CLEAR_INCOMPAT_FEATURE(sb, EXT3_FEATURE_INCOMPAT_RECOVER);
423 es->s_state = cpu_to_le16(sbi->s_mount_state);
424 BUFFER_TRACE(sbi->s_sbh, "marking dirty");
425 mark_buffer_dirty(sbi->s_sbh);
426 ext3_commit_super(sb, es, 1);
427 }
428
429 for (i = 0; i < sbi->s_gdb_count; i++)
430 brelse(sbi->s_group_desc[i]);
431 kfree(sbi->s_group_desc);
432 percpu_counter_destroy(&sbi->s_freeblocks_counter);
433 percpu_counter_destroy(&sbi->s_freeinodes_counter);
434 percpu_counter_destroy(&sbi->s_dirs_counter);
435 brelse(sbi->s_sbh);
436#ifdef CONFIG_QUOTA
437 for (i = 0; i < MAXQUOTAS; i++)
438 kfree(sbi->s_qf_names[i]);
439#endif
440
441 /* Debugging code just in case the in-memory inode orphan list
442 * isn't empty. The on-disk one can be non-empty if we've
443 * detected an error and taken the fs readonly, but the
444 * in-memory list had better be clean by this point. */
445 if (!list_empty(&sbi->s_orphan))
446 dump_orphan_list(sb, sbi);
447 J_ASSERT(list_empty(&sbi->s_orphan));
448
f98393a6 449 invalidate_bdev(sb->s_bdev);
1da177e4
LT		 450 if (sbi->journal_bdev && sbi->journal_bdev != sb->s_bdev) {
451 /*
452 * Invalidate the journal device's buffers. We don't want them
453 * floating about in memory - the physical journal device may
454 * hotswapped, and it breaks the `ro-after' testing code.
455 */
456 sync_blockdev(sbi->journal_bdev);
f98393a6 457 invalidate_bdev(sbi->journal_bdev);
1da177e4
LT		 458 ext3_blkdev_remove(sbi);
459 }
460 sb->s_fs_info = NULL;
5df096d6 461 kfree(sbi->s_blockgroup_lock);
1da177e4 462 kfree(sbi);
6cfd0148
CH		 463
464 unlock_kernel();
1da177e4
LT		 465}
466
e18b890b 467static struct kmem_cache *ext3_inode_cachep;
1da177e4
LT		 468
469/*
470 * Called inside transaction, so use GFP_NOFS
471 */
472static struct inode *ext3_alloc_inode(struct super_block *sb)
473{
474 struct ext3_inode_info *ei;
475
e6b4f8da 476 ei = kmem_cache_alloc(ext3_inode_cachep, GFP_NOFS);
1da177e4
LT		 477 if (!ei)
478 return NULL;
1da177e4
LT		 479 ei->i_block_alloc_info = NULL;
480 ei->vfs_inode.i_version = 1;
fe8bc91c
JK		 481 atomic_set(&ei->i_datasync_tid, 0);
482 atomic_set(&ei->i_sync_tid, 0);
1da177e4
LT		 483 return &ei->vfs_inode;
484}
485
486static void ext3_destroy_inode(struct inode *inode)
487{
9f7dd93d
VA		 488 if (!list_empty(&(EXT3_I(inode)->i_orphan))) {
489 printk("EXT3 Inode %p: orphan list check failed!\n",
490 EXT3_I(inode));
491 print_hex_dump(KERN_INFO, "", DUMP_PREFIX_ADDRESS, 16, 4,
492 EXT3_I(inode), sizeof(struct ext3_inode_info),
493 false);
494 dump_stack();
495 }
1da177e4
LT		 496 kmem_cache_free(ext3_inode_cachep, EXT3_I(inode));
497}
498
51cc5068 499static void init_once(void *foo)
1da177e4
LT		 500{
501 struct ext3_inode_info *ei = (struct ext3_inode_info *) foo;
502
a35afb83 503 INIT_LIST_HEAD(&ei->i_orphan);
1da177e4 504#ifdef CONFIG_EXT3_FS_XATTR
a35afb83 505 init_rwsem(&ei->xattr_sem);
1da177e4 506#endif
a35afb83
CL		 507 mutex_init(&ei->truncate_mutex);
508 inode_init_once(&ei->vfs_inode);
1da177e4 509}
ae6ddcc5 510
1da177e4
LT		 511static int init_inodecache(void)
512{
513 ext3_inode_cachep = kmem_cache_create("ext3_inode_cache",
514 sizeof(struct ext3_inode_info),
fffb60f9
PJ		 515 0, (SLAB_RECLAIM_ACCOUNT|
516 SLAB_MEM_SPREAD),
20c2df83 517 init_once);
1da177e4
LT		 518 if (ext3_inode_cachep == NULL)
519 return -ENOMEM;
520 return 0;
521}
522
523static void destroy_inodecache(void)
524{
1a1d92c1 525 kmem_cache_destroy(ext3_inode_cachep);
1da177e4
LT		 526}
527
528static void ext3_clear_inode(struct inode *inode)
529{
530 struct ext3_block_alloc_info *rsv = EXT3_I(inode)->i_block_alloc_info;
257ba15c 531
9f754758 532 dquot_drop(inode);
1da177e4
LT		 533 ext3_discard_reservation(inode);
534 EXT3_I(inode)->i_block_alloc_info = NULL;
e6022603
AM		 535 if (unlikely(rsv))
536 kfree(rsv);
1da177e4
LT		 537}
538
6cd37cda 539static inline void ext3_show_quota_options(struct seq_file *seq, struct super_block *sb)
8fc2751b 540{
6cd37cda 541#if defined(CONFIG_QUOTA)
275abf5b 542 struct ext3_sb_info *sbi = EXT3_SB(sb);
8fc2751b 543
1aeec434
JK		 544 if (sbi->s_jquota_fmt) {
545 char *fmtname = "";
546
547 switch (sbi->s_jquota_fmt) {
548 case QFMT_VFS_OLD:
549 fmtname = "vfsold";
550 break;
551 case QFMT_VFS_V0:
552 fmtname = "vfsv0";
553 break;
554 case QFMT_VFS_V1:
555 fmtname = "vfsv1";
556 break;
557 }
558 seq_printf(seq, ",jqfmt=%s", fmtname);
559 }
8fc2751b
MB		 560
561 if (sbi->s_qf_names[USRQUOTA])
562 seq_printf(seq, ",usrjquota=%s", sbi->s_qf_names[USRQUOTA]);
563
564 if (sbi->s_qf_names[GRPQUOTA])
565 seq_printf(seq, ",grpjquota=%s", sbi->s_qf_names[GRPQUOTA]);
566
e3c96435 567 if (test_opt(sb, USRQUOTA))
8fc2751b
MB		 568 seq_puts(seq, ",usrquota");
569
e3c96435 570 if (test_opt(sb, GRPQUOTA))
8fc2751b
MB		 571 seq_puts(seq, ",grpquota");
572#endif
6cd37cda
PO		 573}
574
3c4cec65
JK		 575static char *data_mode_string(unsigned long mode)
576{
577 switch (mode) {
578 case EXT3_MOUNT_JOURNAL_DATA:
579 return "journal";
580 case EXT3_MOUNT_ORDERED_DATA:
581 return "ordered";
582 case EXT3_MOUNT_WRITEBACK_DATA:
583 return "writeback";
584 }
585 return "unknown";
586}
587
571beed8
MS		 588/*
589 * Show an option if
590 * - it's set to a non-default value OR
591 * - if the per-sb default is different from the global default
592 */
6cd37cda
PO		 593static int ext3_show_options(struct seq_file *seq, struct vfsmount *vfs)
594{
595 struct super_block *sb = vfs->mnt_sb;
571beed8
MS		 596 struct ext3_sb_info *sbi = EXT3_SB(sb);
597 struct ext3_super_block *es = sbi->s_es;
598 unsigned long def_mount_opts;
599
600 def_mount_opts = le32_to_cpu(es->s_default_mount_opts);
601
602 if (sbi->s_sb_block != 1)
603 seq_printf(seq, ",sb=%lu", sbi->s_sb_block);
604 if (test_opt(sb, MINIX_DF))
605 seq_puts(seq, ",minixdf");
606 if (test_opt(sb, GRPID))
607 seq_puts(seq, ",grpid");
608 if (!test_opt(sb, GRPID) && (def_mount_opts & EXT3_DEFM_BSDGROUPS))
609 seq_puts(seq, ",nogrpid");
610 if (sbi->s_resuid != EXT3_DEF_RESUID ||
611 le16_to_cpu(es->s_def_resuid) != EXT3_DEF_RESUID) {
612 seq_printf(seq, ",resuid=%u", sbi->s_resuid);
613 }
614 if (sbi->s_resgid != EXT3_DEF_RESGID ||
615 le16_to_cpu(es->s_def_resgid) != EXT3_DEF_RESGID) {
616 seq_printf(seq, ",resgid=%u", sbi->s_resgid);
617 }
1eca93f9 618 if (test_opt(sb, ERRORS_RO)) {
571beed8
MS		 619 int def_errors = le16_to_cpu(es->s_errors);
620
621 if (def_errors == EXT3_ERRORS_PANIC ||
1eca93f9
AK		 622 def_errors == EXT3_ERRORS_CONTINUE) {
623 seq_puts(seq, ",errors=remount-ro");
571beed8
MS		 624 }
625 }
1eca93f9
AK		 626 if (test_opt(sb, ERRORS_CONT))
627 seq_puts(seq, ",errors=continue");
571beed8
MS		 628 if (test_opt(sb, ERRORS_PANIC))
629 seq_puts(seq, ",errors=panic");
630 if (test_opt(sb, NO_UID32))
631 seq_puts(seq, ",nouid32");
632 if (test_opt(sb, DEBUG))
633 seq_puts(seq, ",debug");
634 if (test_opt(sb, OLDALLOC))
635 seq_puts(seq, ",oldalloc");
636#ifdef CONFIG_EXT3_FS_XATTR
637 if (test_opt(sb, XATTR_USER))
638 seq_puts(seq, ",user_xattr");
639 if (!test_opt(sb, XATTR_USER) &&
640 (def_mount_opts & EXT3_DEFM_XATTR_USER)) {
641 seq_puts(seq, ",nouser_xattr");
642 }
643#endif
644#ifdef CONFIG_EXT3_FS_POSIX_ACL
645 if (test_opt(sb, POSIX_ACL))
646 seq_puts(seq, ",acl");
647 if (!test_opt(sb, POSIX_ACL) && (def_mount_opts & EXT3_DEFM_ACL))
648 seq_puts(seq, ",noacl");
649#endif
650 if (!test_opt(sb, RESERVATION))
651 seq_puts(seq, ",noreservation");
652 if (sbi->s_commit_interval) {
653 seq_printf(seq, ",commit=%u",
654 (unsigned) (sbi->s_commit_interval / HZ));
655 }
656 if (test_opt(sb, BARRIER))
657 seq_puts(seq, ",barrier=1");
658 if (test_opt(sb, NOBH))
659 seq_puts(seq, ",nobh");
6cd37cda 660
e3c96435 661 seq_printf(seq, ",data=%s", data_mode_string(test_opt(sb, DATA_FLAGS)));
0e4fb5e2
HK		 662 if (test_opt(sb, DATA_ERR_ABORT))
663 seq_puts(seq, ",data_err=abort");
664
dee1d3b6
ES		 665 if (test_opt(sb, NOLOAD))
666 seq_puts(seq, ",norecovery");
667
6cd37cda 668 ext3_show_quota_options(seq, sb);
8fc2751b
MB		 669
670 return 0;
671}
1da177e4 672
fdb36673 673
74af0baa
CH		 674static struct inode *ext3_nfs_get_inode(struct super_block *sb,
675 u64 ino, u32 generation)
fdb36673 676{
fdb36673 677 struct inode *inode;
fdb36673
N		 678
679 if (ino < EXT3_FIRST_INO(sb) && ino != EXT3_ROOT_INO)
680 return ERR_PTR(-ESTALE);
681 if (ino > le32_to_cpu(EXT3_SB(sb)->s_es->s_inodes_count))
682 return ERR_PTR(-ESTALE);
683
684 /* iget isn't really right if the inode is currently unallocated!!
685 *
686 * ext3_read_inode will return a bad_inode if the inode had been
687 * deleted, so we should be safe.
688 *
689 * Currently we don't know the generation for parent directory, so
690 * a generation of 0 means "accept any"
691 */
473043dc
DH		 692 inode = ext3_iget(sb, ino);
693 if (IS_ERR(inode))
694 return ERR_CAST(inode);
695 if (generation && inode->i_generation != generation) {
fdb36673
N		 696 iput(inode);
697 return ERR_PTR(-ESTALE);
698 }
74af0baa
CH		 699
700 return inode;
701}
702
703static struct dentry *ext3_fh_to_dentry(struct super_block *sb, struct fid *fid,
704 int fh_len, int fh_type)
705{
706 return generic_fh_to_dentry(sb, fid, fh_len, fh_type,
707 ext3_nfs_get_inode);
708}
709
710static struct dentry *ext3_fh_to_parent(struct super_block *sb, struct fid *fid,
711 int fh_len, int fh_type)
712{
713 return generic_fh_to_parent(sb, fid, fh_len, fh_type,
714 ext3_nfs_get_inode);
fdb36673
N		 715}
716
6b082b53
TO		 717/*
718 * Try to release metadata pages (indirect blocks, directories) which are
719 * mapped via the block device. Since these pages could have journal heads
720 * which would prevent try_to_free_buffers() from freeing them, we must use
721 * jbd layer's try_to_free_buffers() function to release them.
722 */
723static int bdev_try_to_free_page(struct super_block *sb, struct page *page,
724 gfp_t wait)
725{
726 journal_t *journal = EXT3_SB(sb)->s_journal;
727
728 WARN_ON(PageChecked(page));
729 if (!page_has_buffers(page))
730 return 0;
731 if (journal)
732 return journal_try_to_free_buffers(journal, page,
733 wait & ~__GFP_WAIT);
734 return try_to_free_buffers(page);
735}
736
8fc2751b 737#ifdef CONFIG_QUOTA
1da177e4
LT		 738#define QTYPE2NAME(t) ((t)==USRQUOTA?"user":"group")
739#define QTYPE2MOPT(on, t) ((t)==USRQUOTA?((on)##USRJQUOTA):((on)##GRPJQUOTA))
740
1da177e4
LT		 741static int ext3_write_dquot(struct dquot *dquot);
742static int ext3_acquire_dquot(struct dquot *dquot);
743static int ext3_release_dquot(struct dquot *dquot);
744static int ext3_mark_dquot_dirty(struct dquot *dquot);
745static int ext3_write_info(struct super_block *sb, int type);
2fd83a4f
JK		 746static int ext3_quota_on(struct super_block *sb, int type, int format_id,
747 char *path, int remount);
1da177e4
LT		 748static int ext3_quota_on_mount(struct super_block *sb, int type);
749static ssize_t ext3_quota_read(struct super_block *sb, int type, char *data,
750 size_t len, loff_t off);
751static ssize_t ext3_quota_write(struct super_block *sb, int type,
752 const char *data, size_t len, loff_t off);
753
61e225dc 754static const struct dquot_operations ext3_quota_operations = {
1da177e4
LT		 755 .write_dquot = ext3_write_dquot,
756 .acquire_dquot = ext3_acquire_dquot,
757 .release_dquot = ext3_release_dquot,
758 .mark_dirty = ext3_mark_dquot_dirty,
157091a2
JK		 759 .write_info = ext3_write_info,
760 .alloc_dquot = dquot_alloc,
761 .destroy_dquot = dquot_destroy,
1da177e4
LT		 762};
763
0d54b217 764static const struct quotactl_ops ext3_qctl_operations = {
1da177e4
LT		 765 .quota_on = ext3_quota_on,
766 .quota_off = vfs_quota_off,
767 .quota_sync = vfs_quota_sync,
768 .get_info = vfs_get_dqinfo,
769 .set_info = vfs_set_dqinfo,
770 .get_dqblk = vfs_get_dqblk,
771 .set_dqblk = vfs_set_dqblk
772};
773#endif
774
ee9b6d61 775static const struct super_operations ext3_sops = {
1da177e4
LT		 776 .alloc_inode = ext3_alloc_inode,
777 .destroy_inode = ext3_destroy_inode,
1da177e4
LT		 778 .write_inode = ext3_write_inode,
779 .dirty_inode = ext3_dirty_inode,
780 .delete_inode = ext3_delete_inode,
781 .put_super = ext3_put_super,
1da177e4 782 .sync_fs = ext3_sync_fs,
c4be0c1d
TS		 783 .freeze_fs = ext3_freeze,
784 .unfreeze_fs = ext3_unfreeze,
1da177e4
LT		 785 .statfs = ext3_statfs,
786 .remount_fs = ext3_remount,
787 .clear_inode = ext3_clear_inode,
8fc2751b 788 .show_options = ext3_show_options,
1da177e4
LT		 789#ifdef CONFIG_QUOTA
790 .quota_read = ext3_quota_read,
791 .quota_write = ext3_quota_write,
792#endif
6b082b53 793 .bdev_try_to_free_page = bdev_try_to_free_page,
1da177e4
LT		 794};
795
39655164 796static const struct export_operations ext3_export_ops = {
74af0baa
CH		 797 .fh_to_dentry = ext3_fh_to_dentry,
798 .fh_to_parent = ext3_fh_to_parent,
1da177e4
LT		 799 .get_parent = ext3_get_parent,
800};
801
802enum {
803 Opt_bsd_df, Opt_minix_df, Opt_grpid, Opt_nogrpid,
804 Opt_resgid, Opt_resuid, Opt_sb, Opt_err_cont, Opt_err_panic, Opt_err_ro,
2860b733 805 Opt_nouid32, Opt_nocheck, Opt_debug, Opt_oldalloc, Opt_orlov,
1da177e4 806 Opt_user_xattr, Opt_nouser_xattr, Opt_acl, Opt_noacl,
ade1a29e 807 Opt_reservation, Opt_noreservation, Opt_noload, Opt_nobh, Opt_bh,
71b96257 808 Opt_commit, Opt_journal_update, Opt_journal_inum, Opt_journal_dev,
1da177e4 809 Opt_abort, Opt_data_journal, Opt_data_ordered, Opt_data_writeback,
0e4fb5e2 810 Opt_data_err_abort, Opt_data_err_ignore,
1da177e4 811 Opt_usrjquota, Opt_grpjquota, Opt_offusrjquota, Opt_offgrpjquota,
1aeec434
JK		 812 Opt_jqfmt_vfsold, Opt_jqfmt_vfsv0, Opt_jqfmt_vfsv1, Opt_quota,
813 Opt_noquota, Opt_ignore, Opt_barrier, Opt_err, Opt_resize,
814 Opt_usrquota, Opt_grpquota
1da177e4
LT		 815};
816
a447c093 817static const match_table_t tokens = {
1da177e4
LT		 818 {Opt_bsd_df, "bsddf"},
819 {Opt_minix_df, "minixdf"},
820 {Opt_grpid, "grpid"},
821 {Opt_grpid, "bsdgroups"},
822 {Opt_nogrpid, "nogrpid"},
823 {Opt_nogrpid, "sysvgroups"},
824 {Opt_resgid, "resgid=%u"},
825 {Opt_resuid, "resuid=%u"},
826 {Opt_sb, "sb=%u"},
827 {Opt_err_cont, "errors=continue"},
828 {Opt_err_panic, "errors=panic"},
829 {Opt_err_ro, "errors=remount-ro"},
830 {Opt_nouid32, "nouid32"},
831 {Opt_nocheck, "nocheck"},
832 {Opt_nocheck, "check=none"},
1da177e4
LT		 833 {Opt_debug, "debug"},
834 {Opt_oldalloc, "oldalloc"},
835 {Opt_orlov, "orlov"},
836 {Opt_user_xattr, "user_xattr"},
837 {Opt_nouser_xattr, "nouser_xattr"},
838 {Opt_acl, "acl"},
839 {Opt_noacl, "noacl"},
840 {Opt_reservation, "reservation"},
841 {Opt_noreservation, "noreservation"},
842 {Opt_noload, "noload"},
dee1d3b6 843 {Opt_noload, "norecovery"},
1da177e4 844 {Opt_nobh, "nobh"},
ade1a29e 845 {Opt_bh, "bh"},
1da177e4
LT		 846 {Opt_commit, "commit=%u"},
847 {Opt_journal_update, "journal=update"},
848 {Opt_journal_inum, "journal=%u"},
71b96257 849 {Opt_journal_dev, "journal_dev=%u"},
1da177e4
LT		 850 {Opt_abort, "abort"},
851 {Opt_data_journal, "data=journal"},
852 {Opt_data_ordered, "data=ordered"},
853 {Opt_data_writeback, "data=writeback"},
0e4fb5e2
HK		 854 {Opt_data_err_abort, "data_err=abort"},
855 {Opt_data_err_ignore, "data_err=ignore"},
1da177e4
LT		 856 {Opt_offusrjquota, "usrjquota="},
857 {Opt_usrjquota, "usrjquota=%s"},
858 {Opt_offgrpjquota, "grpjquota="},
859 {Opt_grpjquota, "grpjquota=%s"},
860 {Opt_jqfmt_vfsold, "jqfmt=vfsold"},
861 {Opt_jqfmt_vfsv0, "jqfmt=vfsv0"},
1aeec434 862 {Opt_jqfmt_vfsv1, "jqfmt=vfsv1"},
8fc2751b 863 {Opt_grpquota, "grpquota"},
1f54587b
JK		 864 {Opt_noquota, "noquota"},
865 {Opt_quota, "quota"},
8fc2751b 866 {Opt_usrquota, "usrquota"},
1da177e4 867 {Opt_barrier, "barrier=%u"},
1da177e4 868 {Opt_resize, "resize"},
92587216 869 {Opt_err, NULL},
1da177e4
LT		 870};
871
4cf46b67 872static ext3_fsblk_t get_sb_block(void **data, struct super_block *sb)
1da177e4 873{
e9ad5620
DK		 874 ext3_fsblk_t sb_block;
875 char *options = (char *) *data;
1da177e4
LT		 876
877 if (!options || strncmp(options, "sb=", 3) != 0)
878 return 1; /* Default location */
879 options += 3;
43d23f90 880 /*todo: use simple_strtoll with >32bit ext3 */
1da177e4
LT		 881 sb_block = simple_strtoul(options, &options, 0);
882 if (*options && *options != ',') {
4cf46b67 883 ext3_msg(sb, "error: invalid sb specification: %s",
1da177e4
LT		 884 (char *) *data);
885 return 1;
886 }
887 if (*options == ',')
888 options++;
889 *data = (void *) options;
890 return sb_block;
891}
892
e1f5c67a
DM		 893#ifdef CONFIG_QUOTA
894static int set_qf_name(struct super_block *sb, int qtype, substring_t *args)
895{
896 struct ext3_sb_info *sbi = EXT3_SB(sb);
897 char *qname;
898
899 if (sb_any_quota_loaded(sb) &&
900 !sbi->s_qf_names[qtype]) {
901 ext3_msg(sb, KERN_ERR,
902 "Cannot change journaled "
903 "quota options when quota turned on");
904 return 0;
905 }
906 qname = match_strdup(args);
907 if (!qname) {
908 ext3_msg(sb, KERN_ERR,
909 "Not enough memory for storing quotafile name");
910 return 0;
911 }
912 if (sbi->s_qf_names[qtype] &&
913 strcmp(sbi->s_qf_names[qtype], qname)) {
914 ext3_msg(sb, KERN_ERR,
915 "%s quota file already specified", QTYPE2NAME(qtype));
916 kfree(qname);
917 return 0;
918 }
919 sbi->s_qf_names[qtype] = qname;
920 if (strchr(sbi->s_qf_names[qtype], '/')) {
921 ext3_msg(sb, KERN_ERR,
922 "quotafile must be on filesystem root");
923 kfree(sbi->s_qf_names[qtype]);
924 sbi->s_qf_names[qtype] = NULL;
925 return 0;
926 }
927 set_opt(sbi->s_mount_opt, QUOTA);
928 return 1;
929}
930
931static int clear_qf_name(struct super_block *sb, int qtype) {
932
933 struct ext3_sb_info *sbi = EXT3_SB(sb);
934
935 if (sb_any_quota_loaded(sb) &&
936 sbi->s_qf_names[qtype]) {
937 ext3_msg(sb, KERN_ERR, "Cannot change journaled quota options"
938 " when quota turned on");
939 return 0;
940 }
941 /*
942 * The space will be released later when all options are confirmed
943 * to be correct
944 */
945 sbi->s_qf_names[qtype] = NULL;
946 return 1;
947}
948#endif
949
71b96257 950static int parse_options (char *options, struct super_block *sb,
eee194e7 951 unsigned int *inum, unsigned long *journal_devnum,
43d23f90 952 ext3_fsblk_t *n_blocks_count, int is_remount)
1da177e4
LT		 953{
954 struct ext3_sb_info *sbi = EXT3_SB(sb);
955 char * p;
956 substring_t args[MAX_OPT_ARGS];
957 int data_opt = 0;
958 int option;
959#ifdef CONFIG_QUOTA
e1f5c67a 960 int qfmt;
1da177e4
LT		 961#endif
962
963 if (!options)
964 return 1;
965
966 while ((p = strsep (&options, ",")) != NULL) {
967 int token;
968 if (!*p)
969 continue;
970
971 token = match_token(p, tokens, args);
972 switch (token) {
973 case Opt_bsd_df:
974 clear_opt (sbi->s_mount_opt, MINIX_DF);
975 break;
976 case Opt_minix_df:
977 set_opt (sbi->s_mount_opt, MINIX_DF);
978 break;
979 case Opt_grpid:
980 set_opt (sbi->s_mount_opt, GRPID);
981 break;
982 case Opt_nogrpid:
983 clear_opt (sbi->s_mount_opt, GRPID);
984 break;
985 case Opt_resuid:
986 if (match_int(&args[0], &option))
987 return 0;
988 sbi->s_resuid = option;
989 break;
990 case Opt_resgid:
991 if (match_int(&args[0], &option))
992 return 0;
993 sbi->s_resgid = option;
994 break;
995 case Opt_sb:
996 /* handled by get_sb_block() instead of here */
997 /* *sb_block = match_int(&args[0]); */
998 break;
999 case Opt_err_panic:
1000 clear_opt (sbi->s_mount_opt, ERRORS_CONT);
1001 clear_opt (sbi->s_mount_opt, ERRORS_RO);
1002 set_opt (sbi->s_mount_opt, ERRORS_PANIC);
1003 break;
1004 case Opt_err_ro:
1005 clear_opt (sbi->s_mount_opt, ERRORS_CONT);
1006 clear_opt (sbi->s_mount_opt, ERRORS_PANIC);
1007 set_opt (sbi->s_mount_opt, ERRORS_RO);
1008 break;
1009 case Opt_err_cont:
1010 clear_opt (sbi->s_mount_opt, ERRORS_RO);
1011 clear_opt (sbi->s_mount_opt, ERRORS_PANIC);
1012 set_opt (sbi->s_mount_opt, ERRORS_CONT);
1013 break;
1014 case Opt_nouid32:
1015 set_opt (sbi->s_mount_opt, NO_UID32);
1016 break;
1da177e4
LT		 1017 case Opt_nocheck:
1018 clear_opt (sbi->s_mount_opt, CHECK);
1019 break;
1020 case Opt_debug:
1021 set_opt (sbi->s_mount_opt, DEBUG);
1022 break;
1023 case Opt_oldalloc:
1024 set_opt (sbi->s_mount_opt, OLDALLOC);
1025 break;
1026 case Opt_orlov:
1027 clear_opt (sbi->s_mount_opt, OLDALLOC);
1028 break;
1029#ifdef CONFIG_EXT3_FS_XATTR
1030 case Opt_user_xattr:
1031 set_opt (sbi->s_mount_opt, XATTR_USER);
1032 break;
1033 case Opt_nouser_xattr:
1034 clear_opt (sbi->s_mount_opt, XATTR_USER);
1035 break;
1036#else
1037 case Opt_user_xattr:
1038 case Opt_nouser_xattr:
4cf46b67
AF		 1039 ext3_msg(sb, KERN_INFO,
1040 "(no)user_xattr options not supported");
1da177e4
LT		 1041 break;
1042#endif
1043#ifdef CONFIG_EXT3_FS_POSIX_ACL
1044 case Opt_acl:
1045 set_opt(sbi->s_mount_opt, POSIX_ACL);
1046 break;
1047 case Opt_noacl:
1048 clear_opt(sbi->s_mount_opt, POSIX_ACL);
1049 break;
1050#else
1051 case Opt_acl:
1052 case Opt_noacl:
4cf46b67
AF		 1053 ext3_msg(sb, KERN_INFO,
1054 "(no)acl options not supported");
1da177e4
LT		 1055 break;
1056#endif
1057 case Opt_reservation:
1058 set_opt(sbi->s_mount_opt, RESERVATION);
1059 break;
1060 case Opt_noreservation:
1061 clear_opt(sbi->s_mount_opt, RESERVATION);
1062 break;
1063 case Opt_journal_update:
1064 /* @@@ FIXME */
1065 /* Eventually we will want to be able to create
1066 a journal file here. For now, only allow the
1067 user to specify an existing inode to be the
1068 journal file. */
1069 if (is_remount) {
4cf46b67
AF		 1070 ext3_msg(sb, KERN_ERR, "error: cannot specify "
1071 "journal on remount");
1da177e4
LT		 1072 return 0;
1073 }
1074 set_opt (sbi->s_mount_opt, UPDATE_JOURNAL);
1075 break;
1076 case Opt_journal_inum:
1077 if (is_remount) {
4cf46b67
AF		 1078 ext3_msg(sb, KERN_ERR, "error: cannot specify "
1079 "journal on remount");
1da177e4
LT		 1080 return 0;
1081 }
1082 if (match_int(&args[0], &option))
1083 return 0;
1084 *inum = option;
1085 break;
71b96257
JL		 1086 case Opt_journal_dev:
1087 if (is_remount) {
4cf46b67
AF		 1088 ext3_msg(sb, KERN_ERR, "error: cannot specify "
1089 "journal on remount");
71b96257
JL		 1090 return 0;
1091 }
1092 if (match_int(&args[0], &option))
1093 return 0;
1094 *journal_devnum = option;
1095 break;
1da177e4
LT		 1096 case Opt_noload:
1097 set_opt (sbi->s_mount_opt, NOLOAD);
1098 break;
1099 case Opt_commit:
1100 if (match_int(&args[0], &option))
1101 return 0;
1102 if (option < 0)
1103 return 0;
1104 if (option == 0)
1105 option = JBD_DEFAULT_MAX_COMMIT_AGE;
1106 sbi->s_commit_interval = HZ * option;
1107 break;
1108 case Opt_data_journal:
1109 data_opt = EXT3_MOUNT_JOURNAL_DATA;
1110 goto datacheck;
1111 case Opt_data_ordered:
1112 data_opt = EXT3_MOUNT_ORDERED_DATA;
1113 goto datacheck;
1114 case Opt_data_writeback:
1115 data_opt = EXT3_MOUNT_WRITEBACK_DATA;
1116 datacheck:
1117 if (is_remount) {
e3c96435 1118 if (test_opt(sb, DATA_FLAGS) == data_opt)
3c4cec65 1119 break;
4cf46b67
AF		 1120 ext3_msg(sb, KERN_ERR,
1121 "error: cannot change "
3c4cec65
JK		 1122 "data mode on remount. The filesystem "
1123 "is mounted in data=%s mode and you "
4cf46b67 1124 "try to remount it in data=%s mode.",
e3c96435
DM		 1125 data_mode_string(test_opt(sb,
1126 DATA_FLAGS)),
3c4cec65
JK		 1127 data_mode_string(data_opt));
1128 return 0;
1da177e4 1129 } else {
e3c96435 1130 clear_opt(sbi->s_mount_opt, DATA_FLAGS);
1da177e4
LT		 1131 sbi->s_mount_opt |= data_opt;
1132 }
1133 break;
0e4fb5e2
HK		 1134 case Opt_data_err_abort:
1135 set_opt(sbi->s_mount_opt, DATA_ERR_ABORT);
1136 break;
1137 case Opt_data_err_ignore:
1138 clear_opt(sbi->s_mount_opt, DATA_ERR_ABORT);
1139 break;
1da177e4
LT		 1140#ifdef CONFIG_QUOTA
1141 case Opt_usrjquota:
e1f5c67a 1142 if (!set_qf_name(sb, USRQUOTA, &args[0]))
1da177e4 1143 return 0;
e1f5c67a
DM		 1144 break;
1145 case Opt_grpjquota:
1146 if (!set_qf_name(sb, GRPQUOTA, &args[0]))
1da177e4 1147 return 0;
1da177e4
LT		 1148 break;
1149 case Opt_offusrjquota:
e1f5c67a
DM		 1150 if (!clear_qf_name(sb, USRQUOTA))
1151 return 0;
1152 break;
1da177e4 1153 case Opt_offgrpjquota:
e1f5c67a 1154 if (!clear_qf_name(sb, GRPQUOTA))
1da177e4 1155 return 0;
1da177e4
LT		 1156 break;
1157 case Opt_jqfmt_vfsold:
d06bf1d2
JK		 1158 qfmt = QFMT_VFS_OLD;
1159 goto set_qf_format;
1da177e4 1160 case Opt_jqfmt_vfsv0:
d06bf1d2 1161 qfmt = QFMT_VFS_V0;
1aeec434
JK		 1162 goto set_qf_format;
1163 case Opt_jqfmt_vfsv1:
1164 qfmt = QFMT_VFS_V1;
d06bf1d2 1165set_qf_format:
ee0d5ffe 1166 if (sb_any_quota_loaded(sb) &&
d06bf1d2 1167 sbi->s_jquota_fmt != qfmt) {
4cf46b67 1168 ext3_msg(sb, KERN_ERR, "error: cannot change "
d06bf1d2 1169 "journaled quota options when "
4cf46b67 1170 "quota turned on.");
d06bf1d2
JK		 1171 return 0;
1172 }
1173 sbi->s_jquota_fmt = qfmt;
1da177e4 1174 break;
1f54587b 1175 case Opt_quota:
8fc2751b
MB		 1176 case Opt_usrquota:
1177 set_opt(sbi->s_mount_opt, QUOTA);
1178 set_opt(sbi->s_mount_opt, USRQUOTA);
1179 break;
1180 case Opt_grpquota:
1f54587b 1181 set_opt(sbi->s_mount_opt, QUOTA);
8fc2751b 1182 set_opt(sbi->s_mount_opt, GRPQUOTA);
1f54587b
JK		 1183 break;
1184 case Opt_noquota:
ee0d5ffe 1185 if (sb_any_quota_loaded(sb)) {
4cf46b67
AF		 1186 ext3_msg(sb, KERN_ERR, "error: cannot change "
1187 "quota options when quota turned on.");
1f54587b
JK		 1188 return 0;
1189 }
1190 clear_opt(sbi->s_mount_opt, QUOTA);
8fc2751b
MB		 1191 clear_opt(sbi->s_mount_opt, USRQUOTA);
1192 clear_opt(sbi->s_mount_opt, GRPQUOTA);
1f54587b 1193 break;
1da177e4 1194#else
8fc2751b
MB		 1195 case Opt_quota:
1196 case Opt_usrquota:
1197 case Opt_grpquota:
4cf46b67
AF		 1198 ext3_msg(sb, KERN_ERR,
1199 "error: quota options not supported.");
fa1ff1e0 1200 break;
1da177e4
LT		 1201 case Opt_usrjquota:
1202 case Opt_grpjquota:
1203 case Opt_offusrjquota:
1204 case Opt_offgrpjquota:
1205 case Opt_jqfmt_vfsold:
1206 case Opt_jqfmt_vfsv0:
1aeec434 1207 case Opt_jqfmt_vfsv1:
4cf46b67
AF		 1208 ext3_msg(sb, KERN_ERR,
1209 "error: journaled quota options not "
1210 "supported.");
1da177e4 1211 break;
1f54587b
JK		 1212 case Opt_noquota:
1213 break;
1da177e4
LT		 1214#endif
1215 case Opt_abort:
1216 set_opt(sbi->s_mount_opt, ABORT);
1217 break;
1218 case Opt_barrier:
1219 if (match_int(&args[0], &option))
1220 return 0;
1221 if (option)
1222 set_opt(sbi->s_mount_opt, BARRIER);
1223 else
1224 clear_opt(sbi->s_mount_opt, BARRIER);
1225 break;
1226 case Opt_ignore:
1227 break;
1228 case Opt_resize:
08c6a96f 1229 if (!is_remount) {
4cf46b67
AF		 1230 ext3_msg(sb, KERN_ERR,
1231 "error: resize option only available "
1232 "for remount");
1da177e4
LT		 1233 return 0;
1234 }
c7f1721e
KZ		 1235 if (match_int(&args[0], &option) != 0)
1236 return 0;
1da177e4
LT		 1237 *n_blocks_count = option;
1238 break;
1239 case Opt_nobh:
1240 set_opt(sbi->s_mount_opt, NOBH);
1241 break;
ade1a29e
BP		 1242 case Opt_bh:
1243 clear_opt(sbi->s_mount_opt, NOBH);
1244 break;
1da177e4 1245 default:
4cf46b67
AF		 1246 ext3_msg(sb, KERN_ERR,
1247 "error: unrecognized mount option \"%s\" "
1248 "or missing value", p);
1da177e4
LT		 1249 return 0;
1250 }
1251 }
1252#ifdef CONFIG_QUOTA
8fc2751b 1253 if (sbi->s_qf_names[USRQUOTA] || sbi->s_qf_names[GRPQUOTA]) {
e3c96435 1254 if (test_opt(sb, USRQUOTA) && sbi->s_qf_names[USRQUOTA])
8fc2751b 1255 clear_opt(sbi->s_mount_opt, USRQUOTA);
e3c96435 1256 if (test_opt(sb, GRPQUOTA) && sbi->s_qf_names[GRPQUOTA])
8fc2751b
MB		 1257 clear_opt(sbi->s_mount_opt, GRPQUOTA);
1258
e1f5c67a 1259 if (test_opt(sb, GRPQUOTA) || test_opt(sb, USRQUOTA)) {
4cf46b67
AF		 1260 ext3_msg(sb, KERN_ERR, "error: old and new quota "
1261 "format mixing.");
8fc2751b
MB		 1262 return 0;
1263 }
1264
1265 if (!sbi->s_jquota_fmt) {
4cf46b67
AF		 1266 ext3_msg(sb, KERN_ERR, "error: journaled quota format "
1267 "not specified.");
8fc2751b
MB		 1268 return 0;
1269 }
1270 } else {
1271 if (sbi->s_jquota_fmt) {
4cf46b67 1272 ext3_msg(sb, KERN_ERR, "error: journaled quota format "
99aeaf63 1273 "specified with no journaling "
4cf46b67 1274 "enabled.");
8fc2751b
MB		 1275 return 0;
1276 }
1da177e4
LT		 1277 }
1278#endif
1da177e4
LT		 1279 return 1;
1280}
1281
1282static int ext3_setup_super(struct super_block *sb, struct ext3_super_block *es,
1283 int read_only)
1284{
1285 struct ext3_sb_info *sbi = EXT3_SB(sb);
1286 int res = 0;
1287
1288 if (le32_to_cpu(es->s_rev_level) > EXT3_MAX_SUPP_REV) {
4cf46b67
AF		 1289 ext3_msg(sb, KERN_ERR,
1290 "error: revision level too high, "
1291 "forcing read-only mode");
1da177e4
LT		 1292 res = MS_RDONLY;
1293 }
1294 if (read_only)
1295 return res;
1296 if (!(sbi->s_mount_state & EXT3_VALID_FS))
4cf46b67
AF		 1297 ext3_msg(sb, KERN_WARNING,
1298 "warning: mounting unchecked fs, "
1299 "running e2fsck is recommended");
1da177e4 1300 else if ((sbi->s_mount_state & EXT3_ERROR_FS))
4cf46b67
AF		 1301 ext3_msg(sb, KERN_WARNING,
1302 "warning: mounting fs with errors, "
1303 "running e2fsck is recommended");
1da177e4
LT		 1304 else if ((__s16) le16_to_cpu(es->s_max_mnt_count) >= 0 &&
1305 le16_to_cpu(es->s_mnt_count) >=
1306 (unsigned short) (__s16) le16_to_cpu(es->s_max_mnt_count))
4cf46b67
AF		 1307 ext3_msg(sb, KERN_WARNING,
1308 "warning: maximal mount count reached, "
1309 "running e2fsck is recommended");
1da177e4
LT		 1310 else if (le32_to_cpu(es->s_checkinterval) &&
1311 (le32_to_cpu(es->s_lastcheck) +
1312 le32_to_cpu(es->s_checkinterval) <= get_seconds()))
4cf46b67
AF		 1313 ext3_msg(sb, KERN_WARNING,
1314 "warning: checktime reached, "
1315 "running e2fsck is recommended");
1da177e4
LT		 1316#if 0
1317 /* @@@ We _will_ want to clear the valid bit if we find
1318 inconsistencies, to force a fsck at reboot. But for
1319 a plain journaled filesystem we can keep it set as
1320 valid forever! :) */
e7f23ebd 1321 es->s_state &= cpu_to_le16(~EXT3_VALID_FS);
1da177e4
LT		 1322#endif
1323 if (!(__s16) le16_to_cpu(es->s_max_mnt_count))
1324 es->s_max_mnt_count = cpu_to_le16(EXT3_DFL_MAX_MNT_COUNT);
50e8a289 1325 le16_add_cpu(&es->s_mnt_count, 1);
1da177e4
LT		 1326 es->s_mtime = cpu_to_le32(get_seconds());
1327 ext3_update_dynamic_rev(sb);
1328 EXT3_SET_INCOMPAT_FEATURE(sb, EXT3_FEATURE_INCOMPAT_RECOVER);
1329
1330 ext3_commit_super(sb, es, 1);
1331 if (test_opt(sb, DEBUG))
4cf46b67
AF		 1332 ext3_msg(sb, KERN_INFO, "[bs=%lu, gc=%lu, "
1333 "bpg=%lu, ipg=%lu, mo=%04lx]",
1da177e4
LT		 1334 sb->s_blocksize,
1335 sbi->s_groups_count,
1336 EXT3_BLOCKS_PER_GROUP(sb),
1337 EXT3_INODES_PER_GROUP(sb),
1338 sbi->s_mount_opt);
1339
1da177e4
LT		 1340 if (EXT3_SB(sb)->s_journal->j_inode == NULL) {
1341 char b[BDEVNAME_SIZE];
4cf46b67 1342 ext3_msg(sb, KERN_INFO, "using external journal on %s",
1da177e4
LT		 1343 bdevname(EXT3_SB(sb)->s_journal->j_dev, b));
1344 } else {
4cf46b67 1345 ext3_msg(sb, KERN_INFO, "using internal journal");
1da177e4 1346 }
1da177e4
LT		 1347 return res;
1348}
1349
1350/* Called at mount-time, super-block is locked */
197cd65a 1351static int ext3_check_descriptors(struct super_block *sb)
1da177e4
LT		 1352{
1353 struct ext3_sb_info *sbi = EXT3_SB(sb);
1da177e4
LT		 1354 int i;
1355
1356 ext3_debug ("Checking group descriptors");
1357
197cd65a
AM		 1358 for (i = 0; i < sbi->s_groups_count; i++) {
1359 struct ext3_group_desc *gdp = ext3_get_group_desc(sb, i, NULL);
1eaafeae
AM		 1360 ext3_fsblk_t first_block = ext3_group_first_block_no(sb, i);
1361 ext3_fsblk_t last_block;
197cd65a 1362
855565e8
ES		 1363 if (i == sbi->s_groups_count - 1)
1364 last_block = le32_to_cpu(sbi->s_es->s_blocks_count) - 1;
1365 else
1366 last_block = first_block +
1367 (EXT3_BLOCKS_PER_GROUP(sb) - 1);
1368
855565e8
ES		 1369 if (le32_to_cpu(gdp->bg_block_bitmap) < first_block ||
1370 le32_to_cpu(gdp->bg_block_bitmap) > last_block)
1da177e4
LT		 1371 {
1372 ext3_error (sb, "ext3_check_descriptors",
1373 "Block bitmap for group %d"
1374 " not in group (block %lu)!",
1375 i, (unsigned long)
1376 le32_to_cpu(gdp->bg_block_bitmap));
1377 return 0;
1378 }
855565e8
ES		 1379 if (le32_to_cpu(gdp->bg_inode_bitmap) < first_block ||
1380 le32_to_cpu(gdp->bg_inode_bitmap) > last_block)
1da177e4
LT		 1381 {
1382 ext3_error (sb, "ext3_check_descriptors",
1383 "Inode bitmap for group %d"
1384 " not in group (block %lu)!",
1385 i, (unsigned long)
1386 le32_to_cpu(gdp->bg_inode_bitmap));
1387 return 0;
1388 }
855565e8 1389 if (le32_to_cpu(gdp->bg_inode_table) < first_block ||
780dcdb2 1390 le32_to_cpu(gdp->bg_inode_table) + sbi->s_itb_per_group - 1 >
855565e8 1391 last_block)
1da177e4
LT		 1392 {
1393 ext3_error (sb, "ext3_check_descriptors",
1394 "Inode table for group %d"
1395 " not in group (block %lu)!",
1396 i, (unsigned long)
1397 le32_to_cpu(gdp->bg_inode_table));
1398 return 0;
1399 }
1da177e4
LT		 1400 }
1401
1402 sbi->s_es->s_free_blocks_count=cpu_to_le32(ext3_count_free_blocks(sb));
1403 sbi->s_es->s_free_inodes_count=cpu_to_le32(ext3_count_free_inodes(sb));
1404 return 1;
1405}
1406
1407
1408/* ext3_orphan_cleanup() walks a singly-linked list of inodes (starting at
1409 * the superblock) which were deleted from all directories, but held open by
1410 * a process at the time of a crash. We walk the list and try to delete these
1411 * inodes at recovery time (only with a read-write filesystem).
1412 *
1413 * In order to keep the orphan inode chain consistent during traversal (in
1414 * case of crash during recovery), we link each inode into the superblock
1415 * orphan list_head and handle it the same way as an inode deletion during
1416 * normal operation (which journals the operations for us).
1417 *
1418 * We only do an iget() and an iput() on each inode, which is very safe if we
1419 * accidentally point at an in-use or already deleted inode. The worst that
1420 * can happen in this case is that we get a "bit already cleared" message from
1421 * ext3_free_inode(). The only reason we would point at a wrong inode is if
1422 * e2fsck was run on this filesystem, and it must have already done the orphan
1423 * inode cleanup for us, so we can safely abort without any further action.
1424 */
1425static void ext3_orphan_cleanup (struct super_block * sb,
1426 struct ext3_super_block * es)
1427{
1428 unsigned int s_flags = sb->s_flags;
1429 int nr_orphans = 0, nr_truncates = 0;
1430#ifdef CONFIG_QUOTA
1431 int i;
1432#endif
1433 if (!es->s_last_orphan) {
1434 jbd_debug(4, "no orphan inodes to clean up\n");
1435 return;
1436 }
1437
a8f48a95 1438 if (bdev_read_only(sb->s_bdev)) {
4cf46b67
AF		 1439 ext3_msg(sb, KERN_ERR, "error: write access "
1440 "unavailable, skipping orphan cleanup.");
a8f48a95
ES		 1441 return;
1442 }
1443
1da177e4
LT		 1444 if (EXT3_SB(sb)->s_mount_state & EXT3_ERROR_FS) {
1445 if (es->s_last_orphan)
1446 jbd_debug(1, "Errors on filesystem, "
1447 "clearing orphan list.\n");
1448 es->s_last_orphan = 0;
1449 jbd_debug(1, "Skipping orphan recovery on fs with errors.\n");
1450 return;
1451 }
1452
1453 if (s_flags & MS_RDONLY) {
4cf46b67 1454 ext3_msg(sb, KERN_INFO, "orphan cleanup on readonly fs");
1da177e4
LT		 1455 sb->s_flags &= ~MS_RDONLY;
1456 }
1457#ifdef CONFIG_QUOTA
1458 /* Needed for iput() to work correctly and not trash data */
1459 sb->s_flags |= MS_ACTIVE;
1460 /* Turn on quotas so that they are updated correctly */
1461 for (i = 0; i < MAXQUOTAS; i++) {
1462 if (EXT3_SB(sb)->s_qf_names[i]) {
1463 int ret = ext3_quota_on_mount(sb, i);
1464 if (ret < 0)
4cf46b67
AF		 1465 ext3_msg(sb, KERN_ERR,
1466 "error: cannot turn on journaled "
1467 "quota: %d", ret);
1da177e4
LT		 1468 }
1469 }
1470#endif
1471
1472 while (es->s_last_orphan) {
1473 struct inode *inode;
1474
473043dc
DH		 1475 inode = ext3_orphan_get(sb, le32_to_cpu(es->s_last_orphan));
1476 if (IS_ERR(inode)) {
1da177e4
LT		 1477 es->s_last_orphan = 0;
1478 break;
1479 }
1480
1481 list_add(&EXT3_I(inode)->i_orphan, &EXT3_SB(sb)->s_orphan);
871a2931 1482 dquot_initialize(inode);
1da177e4
LT		 1483 if (inode->i_nlink) {
1484 printk(KERN_DEBUG
eee194e7 1485 "%s: truncating inode %lu to %Ld bytes\n",
e05b6b52 1486 __func__, inode->i_ino, inode->i_size);
eee194e7 1487 jbd_debug(2, "truncating inode %lu to %Ld bytes\n",
1da177e4
LT		 1488 inode->i_ino, inode->i_size);
1489 ext3_truncate(inode);
1490 nr_truncates++;
1491 } else {
1492 printk(KERN_DEBUG
eee194e7 1493 "%s: deleting unreferenced inode %lu\n",
e05b6b52 1494 __func__, inode->i_ino);
eee194e7 1495 jbd_debug(2, "deleting unreferenced inode %lu\n",
1da177e4
LT		 1496 inode->i_ino);
1497 nr_orphans++;
1498 }
1499 iput(inode); /* The delete magic happens here! */
1500 }
1501
1502#define PLURAL(x) (x), ((x)==1) ? "" : "s"
1503
1504 if (nr_orphans)
4cf46b67
AF		 1505 ext3_msg(sb, KERN_INFO, "%d orphan inode%s deleted",
1506 PLURAL(nr_orphans));
1da177e4 1507 if (nr_truncates)
4cf46b67
AF		 1508 ext3_msg(sb, KERN_INFO, "%d truncate%s cleaned up",
1509 PLURAL(nr_truncates));
1da177e4
LT		 1510#ifdef CONFIG_QUOTA
1511 /* Turn quotas off */
1512 for (i = 0; i < MAXQUOTAS; i++) {
1513 if (sb_dqopt(sb)->files[i])
2fd83a4f 1514 vfs_quota_off(sb, i, 0);
1da177e4
LT		 1515 }
1516#endif
1517 sb->s_flags = s_flags; /* Restore MS_RDONLY status */
1518}
1519
1da177e4
LT		 1520/*
1521 * Maximal file size. There is a direct, and {,double-,triple-}indirect
1522 * block limit, and also a limit of (2^32 - 1) 512-byte sectors in i_blocks.
1523 * We need to be 1 filesystem block less than the 2^32 sector limit.
1524 */
1525static loff_t ext3_max_size(int bits)
1526{
1527 loff_t res = EXT3_NDIR_BLOCKS;
fe7fdc37
AK		 1528 int meta_blocks;
1529 loff_t upper_limit;
1530
1531 /* This is calculated to be the largest file size for a
1532 * dense, file such that the total number of
1da177e4 1533 * sectors in the file, including data and all indirect blocks,
fe7fdc37
AK		 1534 * does not exceed 2^32 -1
1535 * __u32 i_blocks representing the total number of
1536 * 512 bytes blocks of the file
1537 */
1538 upper_limit = (1LL << 32) - 1;
1539
1540 /* total blocks in file system block size */
1541 upper_limit >>= (bits - 9);
1542
1543
1544 /* indirect blocks */
1545 meta_blocks = 1;
1546 /* double indirect blocks */
1547 meta_blocks += 1 + (1LL << (bits-2));
1548 /* tripple indirect blocks */
1549 meta_blocks += 1 + (1LL << (bits-2)) + (1LL << (2*(bits-2)));
1550
1551 upper_limit -= meta_blocks;
1552 upper_limit <<= bits;
1da177e4
LT		 1553
1554 res += 1LL << (bits-2);
1555 res += 1LL << (2*(bits-2));
1556 res += 1LL << (3*(bits-2));
1557 res <<= bits;
1558 if (res > upper_limit)
1559 res = upper_limit;
fe7fdc37
AK		 1560
1561 if (res > MAX_LFS_FILESIZE)
1562 res = MAX_LFS_FILESIZE;
1563
1da177e4
LT		 1564 return res;
1565}
1566
43d23f90
MC		 1567static ext3_fsblk_t descriptor_loc(struct super_block *sb,
1568 ext3_fsblk_t logic_sb_block,
1da177e4
LT		 1569 int nr)
1570{
1571 struct ext3_sb_info *sbi = EXT3_SB(sb);
43d23f90 1572 unsigned long bg, first_meta_bg;
1da177e4
LT		 1573 int has_super = 0;
1574
1da177e4
LT		 1575 first_meta_bg = le32_to_cpu(sbi->s_es->s_first_meta_bg);
1576
1577 if (!EXT3_HAS_INCOMPAT_FEATURE(sb, EXT3_FEATURE_INCOMPAT_META_BG) ||
1578 nr < first_meta_bg)
1579 return (logic_sb_block + nr + 1);
1580 bg = sbi->s_desc_per_block * nr;
1581 if (ext3_bg_has_super(sb, bg))
1582 has_super = 1;
43d23f90 1583 return (has_super + ext3_group_first_block_no(sb, bg));
1da177e4
LT		 1584}
1585
1586
1587static int ext3_fill_super (struct super_block *sb, void *data, int silent)
1588{
1589 struct buffer_head * bh;
1590 struct ext3_super_block *es = NULL;
1591 struct ext3_sb_info *sbi;
43d23f90 1592 ext3_fsblk_t block;
4cf46b67 1593 ext3_fsblk_t sb_block = get_sb_block(&data, sb);
43d23f90 1594 ext3_fsblk_t logic_sb_block;
1da177e4 1595 unsigned long offset = 0;
eee194e7 1596 unsigned int journal_inum = 0;
71b96257 1597 unsigned long journal_devnum = 0;
1da177e4
LT		 1598 unsigned long def_mount_opts;
1599 struct inode *root;
1600 int blocksize;
1601 int hblock;
1602 int db_count;
1603 int i;
1604 int needs_recovery;
473043dc 1605 int ret = -EINVAL;
1da177e4 1606 __le32 features;
833f4077 1607 int err;
1da177e4 1608
f8314dc6 1609 sbi = kzalloc(sizeof(*sbi), GFP_KERNEL);
1da177e4
LT		 1610 if (!sbi)
1611 return -ENOMEM;
5df096d6
PE		 1612
1613 sbi->s_blockgroup_lock =
1614 kzalloc(sizeof(struct blockgroup_lock), GFP_KERNEL);
1615 if (!sbi->s_blockgroup_lock) {
1616 kfree(sbi);
1617 return -ENOMEM;
1618 }
1da177e4 1619 sb->s_fs_info = sbi;
1da177e4
LT		 1620 sbi->s_mount_opt = 0;
1621 sbi->s_resuid = EXT3_DEF_RESUID;
1622 sbi->s_resgid = EXT3_DEF_RESGID;
571beed8 1623 sbi->s_sb_block = sb_block;
1da177e4
LT		 1624
1625 unlock_kernel();
1626
1627 blocksize = sb_min_blocksize(sb, EXT3_MIN_BLOCK_SIZE);
1628 if (!blocksize) {
4cf46b67 1629 ext3_msg(sb, KERN_ERR, "error: unable to set blocksize");
1da177e4
LT		 1630 goto out_fail;
1631 }
1632
1633 /*
1634 * The ext3 superblock will not be buffer aligned for other than 1kB
1635 * block sizes. We need to calculate the offset from buffer start.
1636 */
1637 if (blocksize != EXT3_MIN_BLOCK_SIZE) {
1638 logic_sb_block = (sb_block * EXT3_MIN_BLOCK_SIZE) / blocksize;
1639 offset = (sb_block * EXT3_MIN_BLOCK_SIZE) % blocksize;
1640 } else {
1641 logic_sb_block = sb_block;
1642 }
1643
1644 if (!(bh = sb_bread(sb, logic_sb_block))) {
4cf46b67 1645 ext3_msg(sb, KERN_ERR, "error: unable to read superblock");
1da177e4
LT		 1646 goto out_fail;
1647 }
1648 /*
1649 * Note: s_es must be initialized as soon as possible because
1650 * some ext3 macro-instructions depend on its value
1651 */
1652 es = (struct ext3_super_block *) (((char *)bh->b_data) + offset);
1653 sbi->s_es = es;
1654 sb->s_magic = le16_to_cpu(es->s_magic);
1655 if (sb->s_magic != EXT3_SUPER_MAGIC)
1656 goto cantfind_ext3;
1657
1658 /* Set defaults before we parse the mount options */
1659 def_mount_opts = le32_to_cpu(es->s_default_mount_opts);
1660 if (def_mount_opts & EXT3_DEFM_DEBUG)
1661 set_opt(sbi->s_mount_opt, DEBUG);
1662 if (def_mount_opts & EXT3_DEFM_BSDGROUPS)
1663 set_opt(sbi->s_mount_opt, GRPID);
1664 if (def_mount_opts & EXT3_DEFM_UID16)
1665 set_opt(sbi->s_mount_opt, NO_UID32);
2e7842b8 1666#ifdef CONFIG_EXT3_FS_XATTR
1da177e4
LT		 1667 if (def_mount_opts & EXT3_DEFM_XATTR_USER)
1668 set_opt(sbi->s_mount_opt, XATTR_USER);
2e7842b8
HD		 1669#endif
1670#ifdef CONFIG_EXT3_FS_POSIX_ACL
1da177e4
LT		 1671 if (def_mount_opts & EXT3_DEFM_ACL)
1672 set_opt(sbi->s_mount_opt, POSIX_ACL);
2e7842b8 1673#endif
1da177e4 1674 if ((def_mount_opts & EXT3_DEFM_JMODE) == EXT3_DEFM_JMODE_DATA)
e3c96435 1675 set_opt(sbi->s_mount_opt, JOURNAL_DATA);
1da177e4 1676 else if ((def_mount_opts & EXT3_DEFM_JMODE) == EXT3_DEFM_JMODE_ORDERED)
e3c96435 1677 set_opt(sbi->s_mount_opt, ORDERED_DATA);
1da177e4 1678 else if ((def_mount_opts & EXT3_DEFM_JMODE) == EXT3_DEFM_JMODE_WBACK)
e3c96435 1679 set_opt(sbi->s_mount_opt, WRITEBACK_DATA);
1da177e4
LT		 1680
1681 if (le16_to_cpu(sbi->s_es->s_errors) == EXT3_ERRORS_PANIC)
1682 set_opt(sbi->s_mount_opt, ERRORS_PANIC);
1eca93f9 1683 else if (le16_to_cpu(sbi->s_es->s_errors) == EXT3_ERRORS_CONTINUE)
2245d7c2 1684 set_opt(sbi->s_mount_opt, ERRORS_CONT);
1eca93f9
AK		 1685 else
1686 set_opt(sbi->s_mount_opt, ERRORS_RO);
1da177e4
LT		 1687
1688 sbi->s_resuid = le16_to_cpu(es->s_def_resuid);
1689 sbi->s_resgid = le16_to_cpu(es->s_def_resgid);
1690
1691 set_opt(sbi->s_mount_opt, RESERVATION);
1692
71b96257
JL		 1693 if (!parse_options ((char *) data, sb, &journal_inum, &journal_devnum,
1694 NULL, 0))
1da177e4
LT		 1695 goto failed_mount;
1696
1697 sb->s_flags = (sb->s_flags & ~MS_POSIXACL) |
e3c96435 1698 (test_opt(sb, POSIX_ACL) ? MS_POSIXACL : 0);
1da177e4
LT		 1699
1700 if (le32_to_cpu(es->s_rev_level) == EXT3_GOOD_OLD_REV &&
1701 (EXT3_HAS_COMPAT_FEATURE(sb, ~0U) ||
1702 EXT3_HAS_RO_COMPAT_FEATURE(sb, ~0U) ||
1703 EXT3_HAS_INCOMPAT_FEATURE(sb, ~0U)))
4cf46b67
AF		 1704 ext3_msg(sb, KERN_WARNING,
1705 "warning: feature flags set on rev 0 fs, "
1706 "running e2fsck is recommended");
1da177e4
LT		 1707 /*
1708 * Check feature flags regardless of the revision level, since we
1709 * previously didn't change the revision level when setting the flags,
1710 * so there is a chance incompat flags are set on a rev 0 filesystem.
1711 */
1712 features = EXT3_HAS_INCOMPAT_FEATURE(sb, ~EXT3_FEATURE_INCOMPAT_SUPP);
1713 if (features) {
4cf46b67
AF		 1714 ext3_msg(sb, KERN_ERR,
1715 "error: couldn't mount because of unsupported "
1716 "optional features (%x)", le32_to_cpu(features));
1da177e4
LT		 1717 goto failed_mount;
1718 }
1719 features = EXT3_HAS_RO_COMPAT_FEATURE(sb, ~EXT3_FEATURE_RO_COMPAT_SUPP);
1720 if (!(sb->s_flags & MS_RDONLY) && features) {
4cf46b67
AF		 1721 ext3_msg(sb, KERN_ERR,
1722 "error: couldn't mount RDWR because of unsupported "
1723 "optional features (%x)", le32_to_cpu(features));
1da177e4
LT		 1724 goto failed_mount;
1725 }
1726 blocksize = BLOCK_SIZE << le32_to_cpu(es->s_log_block_size);
1727
1728 if (blocksize < EXT3_MIN_BLOCK_SIZE ||
1729 blocksize > EXT3_MAX_BLOCK_SIZE) {
4cf46b67
AF		 1730 ext3_msg(sb, KERN_ERR,
1731 "error: couldn't mount because of unsupported "
1732 "filesystem blocksize %d", blocksize);
1da177e4
LT		 1733 goto failed_mount;
1734 }
1735
e1defc4f 1736 hblock = bdev_logical_block_size(sb->s_bdev);
1da177e4
LT		 1737 if (sb->s_blocksize != blocksize) {
1738 /*
1739 * Make sure the blocksize for the filesystem is larger
1740 * than the hardware sectorsize for the machine.
1741 */
1742 if (blocksize < hblock) {
4cf46b67
AF		 1743 ext3_msg(sb, KERN_ERR,
1744 "error: fsblocksize %d too small for "
1745 "hardware sectorsize %d", blocksize, hblock);
1da177e4
LT		 1746 goto failed_mount;
1747 }
1748
1749 brelse (bh);
0f0a89eb 1750 if (!sb_set_blocksize(sb, blocksize)) {
4cf46b67
AF		 1751 ext3_msg(sb, KERN_ERR,
1752 "error: bad blocksize %d", blocksize);
0f0a89eb
TS		 1753 goto out_fail;
1754 }
1da177e4
LT		 1755 logic_sb_block = (sb_block * EXT3_MIN_BLOCK_SIZE) / blocksize;
1756 offset = (sb_block * EXT3_MIN_BLOCK_SIZE) % blocksize;
1757 bh = sb_bread(sb, logic_sb_block);
1758 if (!bh) {
4cf46b67
AF		 1759 ext3_msg(sb, KERN_ERR,
1760 "error: can't read superblock on 2nd try");
1da177e4
LT		 1761 goto failed_mount;
1762 }
1763 es = (struct ext3_super_block *)(((char *)bh->b_data) + offset);
1764 sbi->s_es = es;
1765 if (es->s_magic != cpu_to_le16(EXT3_SUPER_MAGIC)) {
4cf46b67
AF		 1766 ext3_msg(sb, KERN_ERR,
1767 "error: magic mismatch");
1da177e4
LT		 1768 goto failed_mount;
1769 }
1770 }
1771
1772 sb->s_maxbytes = ext3_max_size(sb->s_blocksize_bits);
1773
1774 if (le32_to_cpu(es->s_rev_level) == EXT3_GOOD_OLD_REV) {
1775 sbi->s_inode_size = EXT3_GOOD_OLD_INODE_SIZE;
1776 sbi->s_first_ino = EXT3_GOOD_OLD_FIRST_INO;
1777 } else {
1778 sbi->s_inode_size = le16_to_cpu(es->s_inode_size);
1779 sbi->s_first_ino = le32_to_cpu(es->s_first_ino);
1780 if ((sbi->s_inode_size < EXT3_GOOD_OLD_INODE_SIZE) ||
3fc74269 1781 (!is_power_of_2(sbi->s_inode_size)) ||
1da177e4 1782 (sbi->s_inode_size > blocksize)) {
4cf46b67
AF		 1783 ext3_msg(sb, KERN_ERR,
1784 "error: unsupported inode size: %d",
1da177e4
LT		 1785 sbi->s_inode_size);
1786 goto failed_mount;
1787 }
1788 }
1789 sbi->s_frag_size = EXT3_MIN_FRAG_SIZE <<
1790 le32_to_cpu(es->s_log_frag_size);
1791 if (blocksize != sbi->s_frag_size) {
4cf46b67
AF		 1792 ext3_msg(sb, KERN_ERR,
1793 "error: fragsize %lu != blocksize %u (unsupported)",
1da177e4
LT		 1794 sbi->s_frag_size, blocksize);
1795 goto failed_mount;
1796 }
1797 sbi->s_frags_per_block = 1;
1798 sbi->s_blocks_per_group = le32_to_cpu(es->s_blocks_per_group);
1799 sbi->s_frags_per_group = le32_to_cpu(es->s_frags_per_group);
1800 sbi->s_inodes_per_group = le32_to_cpu(es->s_inodes_per_group);
b47b6f38 1801 if (EXT3_INODE_SIZE(sb) == 0 || EXT3_INODES_PER_GROUP(sb) == 0)
1da177e4
LT		 1802 goto cantfind_ext3;
1803 sbi->s_inodes_per_block = blocksize / EXT3_INODE_SIZE(sb);
1804 if (sbi->s_inodes_per_block == 0)
1805 goto cantfind_ext3;
1806 sbi->s_itb_per_group = sbi->s_inodes_per_group /
1807 sbi->s_inodes_per_block;
1808 sbi->s_desc_per_block = blocksize / sizeof(struct ext3_group_desc);
1809 sbi->s_sbh = bh;
1810 sbi->s_mount_state = le16_to_cpu(es->s_state);
f0d1b0b3
DH		 1811 sbi->s_addr_per_block_bits = ilog2(EXT3_ADDR_PER_BLOCK(sb));
1812 sbi->s_desc_per_block_bits = ilog2(EXT3_DESC_PER_BLOCK(sb));
1da177e4
LT		 1813 for (i=0; i < 4; i++)
1814 sbi->s_hash_seed[i] = le32_to_cpu(es->s_hash_seed[i]);
1815 sbi->s_def_hash_version = es->s_def_hash_version;
5e1f8c9e
TT		 1816 i = le32_to_cpu(es->s_flags);
1817 if (i & EXT2_FLAGS_UNSIGNED_HASH)
1818 sbi->s_hash_unsigned = 3;
1819 else if ((i & EXT2_FLAGS_SIGNED_HASH) == 0) {
1820#ifdef __CHAR_UNSIGNED__
1821 es->s_flags |= cpu_to_le32(EXT2_FLAGS_UNSIGNED_HASH);
1822 sbi->s_hash_unsigned = 3;
1823#else
1824 es->s_flags |= cpu_to_le32(EXT2_FLAGS_SIGNED_HASH);
1825#endif
5e1f8c9e 1826 }
1da177e4
LT		 1827
1828 if (sbi->s_blocks_per_group > blocksize * 8) {
4cf46b67
AF		 1829 ext3_msg(sb, KERN_ERR,
1830 "#blocks per group too big: %lu",
1da177e4
LT		 1831 sbi->s_blocks_per_group);
1832 goto failed_mount;
1833 }
1834 if (sbi->s_frags_per_group > blocksize * 8) {
4cf46b67
AF		 1835 ext3_msg(sb, KERN_ERR,
1836 "error: #fragments per group too big: %lu",
1da177e4
LT		 1837 sbi->s_frags_per_group);
1838 goto failed_mount;
1839 }
1840 if (sbi->s_inodes_per_group > blocksize * 8) {
4cf46b67
AF		 1841 ext3_msg(sb, KERN_ERR,
1842 "error: #inodes per group too big: %lu",
1da177e4
LT		 1843 sbi->s_inodes_per_group);
1844 goto failed_mount;
1845 }
1846
fcd5df35
MC		 1847 if (le32_to_cpu(es->s_blocks_count) >
1848 (sector_t)(~0ULL) >> (sb->s_blocksize_bits - 9)) {
4cf46b67
AF		 1849 ext3_msg(sb, KERN_ERR,
1850 "error: filesystem is too large to mount safely");
fcd5df35 1851 if (sizeof(sector_t) < 8)
4cf46b67
AF		 1852 ext3_msg(sb, KERN_ERR,
1853 "error: CONFIG_LBDAF not enabled");
fcd5df35
MC		 1854 goto failed_mount;
1855 }
1856
1da177e4
LT		 1857 if (EXT3_BLOCKS_PER_GROUP(sb) == 0)
1858 goto cantfind_ext3;
855565e8
ES		 1859 sbi->s_groups_count = ((le32_to_cpu(es->s_blocks_count) -
1860 le32_to_cpu(es->s_first_data_block) - 1)
1861 / EXT3_BLOCKS_PER_GROUP(sb)) + 1;
1da177e4
LT		 1862 db_count = (sbi->s_groups_count + EXT3_DESC_PER_BLOCK(sb) - 1) /
1863 EXT3_DESC_PER_BLOCK(sb);
1864 sbi->s_group_desc = kmalloc(db_count * sizeof (struct buffer_head *),
1865 GFP_KERNEL);
1866 if (sbi->s_group_desc == NULL) {
4cf46b67
AF		 1867 ext3_msg(sb, KERN_ERR,
1868 "error: not enough memory");
1da177e4
LT		 1869 goto failed_mount;
1870 }
1871
5df096d6 1872 bgl_lock_init(sbi->s_blockgroup_lock);
1da177e4
LT		 1873
1874 for (i = 0; i < db_count; i++) {
1875 block = descriptor_loc(sb, logic_sb_block, i);
1876 sbi->s_group_desc[i] = sb_bread(sb, block);
1877 if (!sbi->s_group_desc[i]) {
4cf46b67
AF		 1878 ext3_msg(sb, KERN_ERR,
1879 "error: can't read group descriptor %d", i);
1da177e4
LT		 1880 db_count = i;
1881 goto failed_mount2;
1882 }
1883 }
1884 if (!ext3_check_descriptors (sb)) {
4cf46b67
AF		 1885 ext3_msg(sb, KERN_ERR,
1886 "error: group descriptors corrupted");
1da177e4
LT		 1887 goto failed_mount2;
1888 }
1889 sbi->s_gdb_count = db_count;
1890 get_random_bytes(&sbi->s_next_generation, sizeof(u32));
1891 spin_lock_init(&sbi->s_next_gen_lock);
0216bfcf 1892
1da177e4
LT		 1893 /* per fileystem reservation list head & lock */
1894 spin_lock_init(&sbi->s_rsv_window_lock);
1895 sbi->s_rsv_window_root = RB_ROOT;
1896 /* Add a single, static dummy reservation to the start of the
1897 * reservation window list --- it gives us a placeholder for
1898 * append-at-start-of-list which makes the allocation logic
1899 * _much_ simpler. */
1900 sbi->s_rsv_window_head.rsv_start = EXT3_RESERVE_WINDOW_NOT_ALLOCATED;
1901 sbi->s_rsv_window_head.rsv_end = EXT3_RESERVE_WINDOW_NOT_ALLOCATED;
1902 sbi->s_rsv_window_head.rsv_alloc_hit = 0;
1903 sbi->s_rsv_window_head.rsv_goal_size = 0;
1904 ext3_rsv_window_add(sb, &sbi->s_rsv_window_head);
1905
1906 /*
1907 * set up enough so that it can read an inode
1908 */
1909 sb->s_op = &ext3_sops;
1910 sb->s_export_op = &ext3_export_ops;
1911 sb->s_xattr = ext3_xattr_handlers;
1912#ifdef CONFIG_QUOTA
1913 sb->s_qcop = &ext3_qctl_operations;
1914 sb->dq_op = &ext3_quota_operations;
1915#endif
1916 INIT_LIST_HEAD(&sbi->s_orphan); /* unlinked but open files */
b8a052d0 1917 mutex_init(&sbi->s_orphan_lock);
96d2a495 1918 mutex_init(&sbi->s_resize_lock);
1da177e4
LT		 1919
1920 sb->s_root = NULL;
1921
1922 needs_recovery = (es->s_last_orphan != 0 ||
1923 EXT3_HAS_INCOMPAT_FEATURE(sb,
1924 EXT3_FEATURE_INCOMPAT_RECOVER));
1925
1926 /*
1927 * The first inode we look at is the journal inode. Don't try
1928 * root first: it may be modified in the journal!
1929 */
1930 if (!test_opt(sb, NOLOAD) &&
1931 EXT3_HAS_COMPAT_FEATURE(sb, EXT3_FEATURE_COMPAT_HAS_JOURNAL)) {
71b96257 1932 if (ext3_load_journal(sb, es, journal_devnum))
41d1a636 1933 goto failed_mount2;
1da177e4
LT		 1934 } else if (journal_inum) {
1935 if (ext3_create_journal(sb, es, journal_inum))
41d1a636 1936 goto failed_mount2;
1da177e4
LT		 1937 } else {
1938 if (!silent)
4cf46b67
AF		 1939 ext3_msg(sb, KERN_ERR,
1940 "error: no journal found. "
1941 "mounting ext3 over ext2?");
41d1a636
DM		 1942 goto failed_mount2;
1943 }
1944 err = percpu_counter_init(&sbi->s_freeblocks_counter,
1945 ext3_count_free_blocks(sb));
1946 if (!err) {
1947 err = percpu_counter_init(&sbi->s_freeinodes_counter,
1948 ext3_count_free_inodes(sb));
1949 }
1950 if (!err) {
1951 err = percpu_counter_init(&sbi->s_dirs_counter,
1952 ext3_count_dirs(sb));
1953 }
1954 if (err) {
1955 ext3_msg(sb, KERN_ERR, "error: insufficient memory");
0216bfcf 1956 goto failed_mount3;
1da177e4
LT		 1957 }
1958
1959 /* We have now updated the journal if required, so we can
1960 * validate the data journaling mode. */
1961 switch (test_opt(sb, DATA_FLAGS)) {
1962 case 0:
1963 /* No mode set, assume a default based on the journal
1964 capabilities: ORDERED_DATA if the journal can
1965 cope, else JOURNAL_DATA */
1966 if (journal_check_available_features
1967 (sbi->s_journal, 0, 0, JFS_FEATURE_INCOMPAT_REVOKE))
bbae8bcc 1968 set_opt(sbi->s_mount_opt, DEFAULT_DATA_MODE);
1da177e4
LT		 1969 else
1970 set_opt(sbi->s_mount_opt, JOURNAL_DATA);
1971 break;
1972
1973 case EXT3_MOUNT_ORDERED_DATA:
1974 case EXT3_MOUNT_WRITEBACK_DATA:
1975 if (!journal_check_available_features
1976 (sbi->s_journal, 0, 0, JFS_FEATURE_INCOMPAT_REVOKE)) {
4cf46b67
AF		 1977 ext3_msg(sb, KERN_ERR,
1978 "error: journal does not support "
1979 "requested data journaling mode");
41d1a636 1980 goto failed_mount3;
1da177e4
LT		 1981 }
1982 default:
1983 break;
1984 }
1985
1986 if (test_opt(sb, NOBH)) {
1da177e4 1987 if (!(test_opt(sb, DATA_FLAGS) == EXT3_MOUNT_WRITEBACK_DATA)) {
4cf46b67
AF		 1988 ext3_msg(sb, KERN_WARNING,
1989 "warning: ignoring nobh option - "
1990 "it is supported only with writeback mode");
1da177e4
LT		 1991 clear_opt(sbi->s_mount_opt, NOBH);
1992 }
1993 }
1994 /*
1995 * The journal_load will have done any necessary log recovery,
1996 * so we can safely mount the rest of the filesystem now.
1997 */
1998
473043dc
DH		 1999 root = ext3_iget(sb, EXT3_ROOT_INO);
2000 if (IS_ERR(root)) {
4cf46b67 2001 ext3_msg(sb, KERN_ERR, "error: get root inode failed");
473043dc 2002 ret = PTR_ERR(root);
41d1a636 2003 goto failed_mount3;
1da177e4
LT		 2004 }
2005 if (!S_ISDIR(root->i_mode) || !root->i_blocks || !root->i_size) {
473043dc 2006 iput(root);
4cf46b67 2007 ext3_msg(sb, KERN_ERR, "error: corrupt root inode, run e2fsck");
41d1a636 2008 goto failed_mount3;
1da177e4 2009 }
473043dc
DH		 2010 sb->s_root = d_alloc_root(root);
2011 if (!sb->s_root) {
4cf46b67 2012 ext3_msg(sb, KERN_ERR, "error: get root dentry failed");
473043dc
DH		 2013 iput(root);
2014 ret = -ENOMEM;
41d1a636 2015 goto failed_mount3;
473043dc 2016 }
1da177e4
LT		 2017
2018 ext3_setup_super (sb, es, sb->s_flags & MS_RDONLY);
ed505ee4 2019
1da177e4
LT		 2020 EXT3_SB(sb)->s_mount_state |= EXT3_ORPHAN_FS;
2021 ext3_orphan_cleanup(sb, es);
2022 EXT3_SB(sb)->s_mount_state &= ~EXT3_ORPHAN_FS;
2023 if (needs_recovery)
4cf46b67 2024 ext3_msg(sb, KERN_INFO, "recovery complete");
1da177e4 2025 ext3_mark_recovery_complete(sb, es);
4cf46b67 2026 ext3_msg(sb, KERN_INFO, "mounted filesystem with %s data mode",
1da177e4
LT		 2027 test_opt(sb,DATA_FLAGS) == EXT3_MOUNT_JOURNAL_DATA ? "journal":
2028 test_opt(sb,DATA_FLAGS) == EXT3_MOUNT_ORDERED_DATA ? "ordered":
2029 "writeback");
2030
1da177e4
LT		 2031 lock_kernel();
2032 return 0;
2033
2034cantfind_ext3:
2035 if (!silent)
4cf46b67
AF		 2036 ext3_msg(sb, KERN_INFO,
2037 "error: can't find ext3 filesystem on dev %s.",
1da177e4
LT		 2038 sb->s_id);
2039 goto failed_mount;
2040
0216bfcf
MC		 2041failed_mount3:
2042 percpu_counter_destroy(&sbi->s_freeblocks_counter);
2043 percpu_counter_destroy(&sbi->s_freeinodes_counter);
2044 percpu_counter_destroy(&sbi->s_dirs_counter);
41d1a636 2045 journal_destroy(sbi->s_journal);
1da177e4
LT		 2046failed_mount2:
2047 for (i = 0; i < db_count; i++)
2048 brelse(sbi->s_group_desc[i]);
2049 kfree(sbi->s_group_desc);
2050failed_mount:
2051#ifdef CONFIG_QUOTA
2052 for (i = 0; i < MAXQUOTAS; i++)
2053 kfree(sbi->s_qf_names[i]);
2054#endif
2055 ext3_blkdev_remove(sbi);
2056 brelse(bh);
2057out_fail:
2058 sb->s_fs_info = NULL;
de5ce037 2059 kfree(sbi->s_blockgroup_lock);
1da177e4
LT		 2060 kfree(sbi);
2061 lock_kernel();
473043dc 2062 return ret;
1da177e4
LT		 2063}
2064
2065/*
2066 * Setup any per-fs journal parameters now. We'll do this both on
2067 * initial mount, once the journal has been initialised but before we've
ae6ddcc5 2068 * done any recovery; and again on any subsequent remount.
1da177e4
LT		 2069 */
2070static void ext3_init_journal_params(struct super_block *sb, journal_t *journal)
2071{
2072 struct ext3_sb_info *sbi = EXT3_SB(sb);
2073
2074 if (sbi->s_commit_interval)
2075 journal->j_commit_interval = sbi->s_commit_interval;
2076 /* We could also set up an ext3-specific default for the commit
2077 * interval here, but for now we'll just fall back to the jbd
2078 * default. */
2079
2080 spin_lock(&journal->j_state_lock);
2081 if (test_opt(sb, BARRIER))
2082 journal->j_flags |= JFS_BARRIER;
2083 else
2084 journal->j_flags &= ~JFS_BARRIER;
0e4fb5e2
HK		 2085 if (test_opt(sb, DATA_ERR_ABORT))
2086 journal->j_flags |= JFS_ABORT_ON_SYNCDATA_ERR;
2087 else
2088 journal->j_flags &= ~JFS_ABORT_ON_SYNCDATA_ERR;
1da177e4
LT		 2089 spin_unlock(&journal->j_state_lock);
2090}
2091
eee194e7
ES		 2092static journal_t *ext3_get_journal(struct super_block *sb,
2093 unsigned int journal_inum)
1da177e4
LT		 2094{
2095 struct inode *journal_inode;
2096 journal_t *journal;
2097
2098 /* First, test for the existence of a valid inode on disk. Bad
2099 * things happen if we iget() an unused inode, as the subsequent
2100 * iput() will try to delete it. */
2101
473043dc
DH		 2102 journal_inode = ext3_iget(sb, journal_inum);
2103 if (IS_ERR(journal_inode)) {
4cf46b67 2104 ext3_msg(sb, KERN_ERR, "error: no journal found");
1da177e4
LT		 2105 return NULL;
2106 }
2107 if (!journal_inode->i_nlink) {
2108 make_bad_inode(journal_inode);
2109 iput(journal_inode);
4cf46b67 2110 ext3_msg(sb, KERN_ERR, "error: journal inode is deleted");
1da177e4
LT		 2111 return NULL;
2112 }
2113
2114 jbd_debug(2, "Journal inode found at %p: %Ld bytes\n",
2115 journal_inode, journal_inode->i_size);
473043dc 2116 if (!S_ISREG(journal_inode->i_mode)) {
4cf46b67 2117 ext3_msg(sb, KERN_ERR, "error: invalid journal inode");
1da177e4
LT		 2118 iput(journal_inode);
2119 return NULL;
2120 }
2121
2122 journal = journal_init_inode(journal_inode);
2123 if (!journal) {
4cf46b67 2124 ext3_msg(sb, KERN_ERR, "error: could not load journal inode");
1da177e4
LT		 2125 iput(journal_inode);
2126 return NULL;
2127 }
2128 journal->j_private = sb;
2129 ext3_init_journal_params(sb, journal);
2130 return journal;
2131}
2132
2133static journal_t *ext3_get_dev_journal(struct super_block *sb,
2134 dev_t j_dev)
2135{
2136 struct buffer_head * bh;
2137 journal_t *journal;
43d23f90 2138 ext3_fsblk_t start;
1c2bf374 2139 ext3_fsblk_t len;
1da177e4 2140 int hblock, blocksize;
43d23f90 2141 ext3_fsblk_t sb_block;
1da177e4
LT		 2142 unsigned long offset;
2143 struct ext3_super_block * es;
2144 struct block_device *bdev;
2145
4cf46b67 2146 bdev = ext3_blkdev_get(j_dev, sb);
1da177e4
LT		 2147 if (bdev == NULL)
2148 return NULL;
2149
2150 if (bd_claim(bdev, sb)) {
4cf46b67
AF		 2151 ext3_msg(sb, KERN_ERR,
2152 "error: failed to claim external journal device");
9a1c3542 2153 blkdev_put(bdev, FMODE_READ|FMODE_WRITE);
1da177e4
LT		 2154 return NULL;
2155 }
2156
2157 blocksize = sb->s_blocksize;
e1defc4f 2158 hblock = bdev_logical_block_size(bdev);
1da177e4 2159 if (blocksize < hblock) {
4cf46b67
AF		 2160 ext3_msg(sb, KERN_ERR,
2161 "error: blocksize too small for journal device");
1da177e4
LT		 2162 goto out_bdev;
2163 }
2164
2165 sb_block = EXT3_MIN_BLOCK_SIZE / blocksize;
2166 offset = EXT3_MIN_BLOCK_SIZE % blocksize;
2167 set_blocksize(bdev, blocksize);
2168 if (!(bh = __bread(bdev, sb_block, blocksize))) {
4cf46b67
AF		 2169 ext3_msg(sb, KERN_ERR, "error: couldn't read superblock of "
2170 "external journal");
1da177e4
LT		 2171 goto out_bdev;
2172 }
2173
2174 es = (struct ext3_super_block *) (((char *)bh->b_data) + offset);
2175 if ((le16_to_cpu(es->s_magic) != EXT3_SUPER_MAGIC) ||
2176 !(le32_to_cpu(es->s_feature_incompat) &
2177 EXT3_FEATURE_INCOMPAT_JOURNAL_DEV)) {
4cf46b67
AF		 2178 ext3_msg(sb, KERN_ERR, "error: external journal has "
2179 "bad superblock");
1da177e4
LT		 2180 brelse(bh);
2181 goto out_bdev;
2182 }
2183
2184 if (memcmp(EXT3_SB(sb)->s_es->s_journal_uuid, es->s_uuid, 16)) {
4cf46b67 2185 ext3_msg(sb, KERN_ERR, "error: journal UUID does not match");
1da177e4
LT		 2186 brelse(bh);
2187 goto out_bdev;
2188 }
2189
2190 len = le32_to_cpu(es->s_blocks_count);
2191 start = sb_block + 1;
2192 brelse(bh); /* we're done with the superblock */
2193
2194 journal = journal_init_dev(bdev, sb->s_bdev,
2195 start, len, blocksize);
2196 if (!journal) {
4cf46b67
AF		 2197 ext3_msg(sb, KERN_ERR,
2198 "error: failed to create device journal");
1da177e4
LT		 2199 goto out_bdev;
2200 }
2201 journal->j_private = sb;
2202 ll_rw_block(READ, 1, &journal->j_sb_buffer);
2203 wait_on_buffer(journal->j_sb_buffer);
2204 if (!buffer_uptodate(journal->j_sb_buffer)) {
4cf46b67 2205 ext3_msg(sb, KERN_ERR, "I/O error on journal device");
1da177e4
LT		 2206 goto out_journal;
2207 }
2208 if (be32_to_cpu(journal->j_superblock->s_nr_users) != 1) {
4cf46b67
AF		 2209 ext3_msg(sb, KERN_ERR,
2210 "error: external journal has more than one "
2211 "user (unsupported) - %d",
1da177e4
LT		 2212 be32_to_cpu(journal->j_superblock->s_nr_users));
2213 goto out_journal;
2214 }
2215 EXT3_SB(sb)->journal_bdev = bdev;
2216 ext3_init_journal_params(sb, journal);
2217 return journal;
2218out_journal:
2219 journal_destroy(journal);
2220out_bdev:
2221 ext3_blkdev_put(bdev);
2222 return NULL;
2223}
2224
71b96257
JL		 2225static int ext3_load_journal(struct super_block *sb,
2226 struct ext3_super_block *es,
2227 unsigned long journal_devnum)
1da177e4
LT		 2228{
2229 journal_t *journal;
eee194e7 2230 unsigned int journal_inum = le32_to_cpu(es->s_journal_inum);
71b96257 2231 dev_t journal_dev;
1da177e4
LT		 2232 int err = 0;
2233 int really_read_only;
2234
71b96257
JL		 2235 if (journal_devnum &&
2236 journal_devnum != le32_to_cpu(es->s_journal_dev)) {
4cf46b67
AF		 2237 ext3_msg(sb, KERN_INFO, "external journal device major/minor "
2238 "numbers have changed");
71b96257
JL		 2239 journal_dev = new_decode_dev(journal_devnum);
2240 } else
2241 journal_dev = new_decode_dev(le32_to_cpu(es->s_journal_dev));
2242
1da177e4
LT		 2243 really_read_only = bdev_read_only(sb->s_bdev);
2244
2245 /*
2246 * Are we loading a blank journal or performing recovery after a
2247 * crash? For recovery, we need to check in advance whether we
2248 * can get read-write access to the device.
2249 */
2250
2251 if (EXT3_HAS_INCOMPAT_FEATURE(sb, EXT3_FEATURE_INCOMPAT_RECOVER)) {
2252 if (sb->s_flags & MS_RDONLY) {
4cf46b67
AF		 2253 ext3_msg(sb, KERN_INFO,
2254 "recovery required on readonly filesystem");
1da177e4 2255 if (really_read_only) {
4cf46b67
AF		 2256 ext3_msg(sb, KERN_ERR, "error: write access "
2257 "unavailable, cannot proceed");
1da177e4
LT		 2258 return -EROFS;
2259 }
4cf46b67
AF		 2260 ext3_msg(sb, KERN_INFO,
2261 "write access will be enabled during recovery");
1da177e4
LT		 2262 }
2263 }
2264
2265 if (journal_inum && journal_dev) {
4cf46b67
AF		 2266 ext3_msg(sb, KERN_ERR, "error: filesystem has both journal "
2267 "and inode journals");
1da177e4
LT		 2268 return -EINVAL;
2269 }
2270
2271 if (journal_inum) {
2272 if (!(journal = ext3_get_journal(sb, journal_inum)))
2273 return -EINVAL;
2274 } else {
2275 if (!(journal = ext3_get_dev_journal(sb, journal_dev)))
2276 return -EINVAL;
2277 }
2278
2279 if (!really_read_only && test_opt(sb, UPDATE_JOURNAL)) {
2280 err = journal_update_format(journal);
2281 if (err) {
4cf46b67 2282 ext3_msg(sb, KERN_ERR, "error updating journal");
1da177e4
LT		 2283 journal_destroy(journal);
2284 return err;
2285 }
2286 }
2287
2288 if (!EXT3_HAS_INCOMPAT_FEATURE(sb, EXT3_FEATURE_INCOMPAT_RECOVER))
2289 err = journal_wipe(journal, !really_read_only);
2290 if (!err)
2291 err = journal_load(journal);
2292
2293 if (err) {
4cf46b67 2294 ext3_msg(sb, KERN_ERR, "error loading journal");
1da177e4
LT		 2295 journal_destroy(journal);
2296 return err;
2297 }
2298
2299 EXT3_SB(sb)->s_journal = journal;
2300 ext3_clear_journal_err(sb, es);
71b96257
JL		 2301
2302 if (journal_devnum &&
2303 journal_devnum != le32_to_cpu(es->s_journal_dev)) {
2304 es->s_journal_dev = cpu_to_le32(journal_devnum);
71b96257
JL		 2305
2306 /* Make sure we flush the recovery flag to disk. */
2307 ext3_commit_super(sb, es, 1);
2308 }
2309
1da177e4
LT		 2310 return 0;
2311}
2312
4cf46b67
AF		 2313static int ext3_create_journal(struct super_block *sb,
2314 struct ext3_super_block *es,
eee194e7 2315 unsigned int journal_inum)
1da177e4
LT		 2316{
2317 journal_t *journal;
952d9de1 2318 int err;
1da177e4
LT		 2319
2320 if (sb->s_flags & MS_RDONLY) {
4cf46b67
AF		 2321 ext3_msg(sb, KERN_ERR,
2322 "error: readonly filesystem when trying to "
2323 "create journal");
1da177e4
LT		 2324 return -EROFS;
2325 }
2326
952d9de1
BP		 2327 journal = ext3_get_journal(sb, journal_inum);
2328 if (!journal)
1da177e4
LT		 2329 return -EINVAL;
2330
4cf46b67 2331 ext3_msg(sb, KERN_INFO, "creating new journal on inode %u",
1da177e4
LT		 2332 journal_inum);
2333
952d9de1
BP		 2334 err = journal_create(journal);
2335 if (err) {
4cf46b67 2336 ext3_msg(sb, KERN_ERR, "error creating journal");
1da177e4
LT		 2337 journal_destroy(journal);
2338 return -EIO;
2339 }
2340
2341 EXT3_SB(sb)->s_journal = journal;
2342
2343 ext3_update_dynamic_rev(sb);
2344 EXT3_SET_INCOMPAT_FEATURE(sb, EXT3_FEATURE_INCOMPAT_RECOVER);
2345 EXT3_SET_COMPAT_FEATURE(sb, EXT3_FEATURE_COMPAT_HAS_JOURNAL);
2346
2347 es->s_journal_inum = cpu_to_le32(journal_inum);
1da177e4
LT		 2348
2349 /* Make sure we flush the recovery flag to disk. */
2350 ext3_commit_super(sb, es, 1);
2351
2352 return 0;
2353}
2354
c4be0c1d
TS		 2355static int ext3_commit_super(struct super_block *sb,
2356 struct ext3_super_block *es,
1da177e4
LT		 2357 int sync)
2358{
2359 struct buffer_head *sbh = EXT3_SB(sb)->s_sbh;
c4be0c1d 2360 int error = 0;
1da177e4
LT		 2361
2362 if (!sbh)
c4be0c1d 2363 return error;
96ec2e0a
TT		 2364 /*
2365 * If the file system is mounted read-only, don't update the
2366 * superblock write time. This avoids updating the superblock
2367 * write time when we are mounting the root file system
2368 * read/only but we need to replay the journal; at that point,
2369 * for people who are east of GMT and who make their clock
2370 * tick in localtime for Windows bug-for-bug compatibility,
2371 * the clock is set in the future, and this will cause e2fsck
2372 * to complain and force a full file system check.
2373 */
2374 if (!(sb->s_flags & MS_RDONLY))
2375 es->s_wtime = cpu_to_le32(get_seconds());
1da177e4
LT		 2376 es->s_free_blocks_count = cpu_to_le32(ext3_count_free_blocks(sb));
2377 es->s_free_inodes_count = cpu_to_le32(ext3_count_free_inodes(sb));
2378 BUFFER_TRACE(sbh, "marking dirty");
2379 mark_buffer_dirty(sbh);
2380 if (sync)
c4be0c1d
TS		 2381 error = sync_dirty_buffer(sbh);
2382 return error;
1da177e4
LT		 2383}
2384
2385
2386/*
2387 * Have we just finished recovery? If so, and if we are mounting (or
2388 * remounting) the filesystem readonly, then we will end up with a
2389 * consistent fs on disk. Record that fact.
2390 */
2391static void ext3_mark_recovery_complete(struct super_block * sb,
2392 struct ext3_super_block * es)
2393{
2394 journal_t *journal = EXT3_SB(sb)->s_journal;
2395
2396 journal_lock_updates(journal);
2d7c820e
HK		 2397 if (journal_flush(journal) < 0)
2398 goto out;
2399
1da177e4
LT		 2400 if (EXT3_HAS_INCOMPAT_FEATURE(sb, EXT3_FEATURE_INCOMPAT_RECOVER) &&
2401 sb->s_flags & MS_RDONLY) {
2402 EXT3_CLEAR_INCOMPAT_FEATURE(sb, EXT3_FEATURE_INCOMPAT_RECOVER);
1da177e4
LT		 2403 ext3_commit_super(sb, es, 1);
2404 }
2d7c820e
HK		 2405
2406out:
1da177e4
LT		 2407 journal_unlock_updates(journal);
2408}
2409
2410/*
2411 * If we are mounting (or read-write remounting) a filesystem whose journal
2412 * has recorded an error from a previous lifetime, move that error to the
2413 * main filesystem now.
2414 */
4cf46b67
AF		 2415static void ext3_clear_journal_err(struct super_block *sb,
2416 struct ext3_super_block *es)
1da177e4
LT		 2417{
2418 journal_t *journal;
2419 int j_errno;
2420 const char *errstr;
2421
2422 journal = EXT3_SB(sb)->s_journal;
2423
2424 /*
2425 * Now check for any error status which may have been recorded in the
2426 * journal by a prior ext3_error() or ext3_abort()
2427 */
2428
2429 j_errno = journal_errno(journal);
2430 if (j_errno) {
2431 char nbuf[16];
2432
2433 errstr = ext3_decode_error(sb, j_errno, nbuf);
e05b6b52 2434 ext3_warning(sb, __func__, "Filesystem error recorded "
1da177e4 2435 "from previous mount: %s", errstr);
e05b6b52 2436 ext3_warning(sb, __func__, "Marking fs in need of "
1da177e4
LT		 2437 "filesystem check.");
2438
2439 EXT3_SB(sb)->s_mount_state |= EXT3_ERROR_FS;
2440 es->s_state |= cpu_to_le16(EXT3_ERROR_FS);
2441 ext3_commit_super (sb, es, 1);
2442
2443 journal_clear_err(journal);
2444 }
2445}
2446
2447/*
2448 * Force the running and committing transactions to commit,
2449 * and wait on the commit.
2450 */
2451int ext3_force_commit(struct super_block *sb)
2452{
2453 journal_t *journal;
2454 int ret;
2455
2456 if (sb->s_flags & MS_RDONLY)
2457 return 0;
2458
2459 journal = EXT3_SB(sb)->s_journal;
1da177e4
LT		 2460 ret = ext3_journal_force_commit(journal);
2461 return ret;
2462}
2463
1da177e4
LT		 2464static int ext3_sync_fs(struct super_block *sb, int wait)
2465{
02ac597c 2466 tid_t target;
c87591b7 2467
02ac597c
JK		 2468 if (journal_start_commit(EXT3_SB(sb)->s_journal, &target)) {
2469 if (wait)
2470 log_wait_commit(EXT3_SB(sb)->s_journal, target);
2471 }
1da177e4
LT		 2472 return 0;
2473}
2474
2475/*
2476 * LVM calls this function before a (read-only) snapshot is created. This
2477 * gives us a chance to flush the journal completely and mark the fs clean.
2478 */
c4be0c1d 2479static int ext3_freeze(struct super_block *sb)
1da177e4 2480{
c4be0c1d
TS		 2481 int error = 0;
2482 journal_t *journal;
1da177e4
LT		 2483
2484 if (!(sb->s_flags & MS_RDONLY)) {
c4be0c1d 2485 journal = EXT3_SB(sb)->s_journal;
1da177e4
LT		 2486
2487 /* Now we set up the journal barrier. */
2488 journal_lock_updates(journal);
2d7c820e
HK		 2489
2490 /*
2491 * We don't want to clear needs_recovery flag when we failed
2492 * to flush the journal.
2493 */
c4be0c1d
TS		 2494 error = journal_flush(journal);
2495 if (error < 0)
2496 goto out;
1da177e4
LT		 2497
2498 /* Journal blocked and flushed, clear needs_recovery flag. */
2499 EXT3_CLEAR_INCOMPAT_FEATURE(sb, EXT3_FEATURE_INCOMPAT_RECOVER);
c4be0c1d
TS		 2500 error = ext3_commit_super(sb, EXT3_SB(sb)->s_es, 1);
2501 if (error)
2502 goto out;
1da177e4 2503 }
c4be0c1d
TS		 2504 return 0;
2505
2506out:
2507 journal_unlock_updates(journal);
2508 return error;
1da177e4
LT		 2509}
2510
2511/*
2512 * Called by LVM after the snapshot is done. We need to reset the RECOVER
2513 * flag here, even though the filesystem is not technically dirty yet.
2514 */
c4be0c1d 2515static int ext3_unfreeze(struct super_block *sb)
1da177e4
LT		 2516{
2517 if (!(sb->s_flags & MS_RDONLY)) {
2518 lock_super(sb);
2519 /* Reser the needs_recovery flag before the fs is unlocked. */
2520 EXT3_SET_INCOMPAT_FEATURE(sb, EXT3_FEATURE_INCOMPAT_RECOVER);
2521 ext3_commit_super(sb, EXT3_SB(sb)->s_es, 1);
2522 unlock_super(sb);
2523 journal_unlock_updates(EXT3_SB(sb)->s_journal);
2524 }
c4be0c1d 2525 return 0;
1da177e4
LT		 2526}
2527
2528static int ext3_remount (struct super_block * sb, int * flags, char * data)
2529{
2530 struct ext3_super_block * es;
2531 struct ext3_sb_info *sbi = EXT3_SB(sb);
43d23f90 2532 ext3_fsblk_t n_blocks_count = 0;
08c6a96f
JK		 2533 unsigned long old_sb_flags;
2534 struct ext3_mount_options old_opts;
2535 int err;
2536#ifdef CONFIG_QUOTA
2537 int i;
2538#endif
2539
337eb00a
AIB		 2540 lock_kernel();
2541
08c6a96f 2542 /* Store the original options */
bbd6851a 2543 lock_super(sb);
08c6a96f
JK		 2544 old_sb_flags = sb->s_flags;
2545 old_opts.s_mount_opt = sbi->s_mount_opt;
2546 old_opts.s_resuid = sbi->s_resuid;
2547 old_opts.s_resgid = sbi->s_resgid;
2548 old_opts.s_commit_interval = sbi->s_commit_interval;
2549#ifdef CONFIG_QUOTA
2550 old_opts.s_jquota_fmt = sbi->s_jquota_fmt;
2551 for (i = 0; i < MAXQUOTAS; i++)
2552 old_opts.s_qf_names[i] = sbi->s_qf_names[i];
2553#endif
1da177e4
LT		 2554
2555 /*
2556 * Allow the "check" option to be passed as a remount option.
2557 */
71b96257 2558 if (!parse_options(data, sb, NULL, NULL, &n_blocks_count, 1)) {
08c6a96f
JK		 2559 err = -EINVAL;
2560 goto restore_opts;
2561 }
1da177e4 2562
e3c96435 2563 if (test_opt(sb, ABORT))
e05b6b52 2564 ext3_abort(sb, __func__, "Abort forced by user");
1da177e4
LT		 2565
2566 sb->s_flags = (sb->s_flags & ~MS_POSIXACL) |
e3c96435 2567 (test_opt(sb, POSIX_ACL) ? MS_POSIXACL : 0);
1da177e4
LT		 2568
2569 es = sbi->s_es;
2570
2571 ext3_init_journal_params(sb, sbi->s_journal);
2572
2573 if ((*flags & MS_RDONLY) != (sb->s_flags & MS_RDONLY) ||
2574 n_blocks_count > le32_to_cpu(es->s_blocks_count)) {
e3c96435 2575 if (test_opt(sb, ABORT)) {
08c6a96f
JK		 2576 err = -EROFS;
2577 goto restore_opts;
2578 }
1da177e4
LT		 2579
2580 if (*flags & MS_RDONLY) {
2581 /*
2582 * First of all, the unconditional stuff we have to do
2583 * to disable replay of the journal when we next remount
2584 */
2585 sb->s_flags |= MS_RDONLY;
2586
2587 /*
2588 * OK, test if we are remounting a valid rw partition
2589 * readonly, and if so set the rdonly flag and then
2590 * mark the partition as valid again.
2591 */
2592 if (!(es->s_state & cpu_to_le16(EXT3_VALID_FS)) &&
2593 (sbi->s_mount_state & EXT3_VALID_FS))
2594 es->s_state = cpu_to_le16(sbi->s_mount_state);
2595
2596 ext3_mark_recovery_complete(sb, es);
2597 } else {
2598 __le32 ret;
2599 if ((ret = EXT3_HAS_RO_COMPAT_FEATURE(sb,
2600 ~EXT3_FEATURE_RO_COMPAT_SUPP))) {
4cf46b67
AF		 2601 ext3_msg(sb, KERN_WARNING,
2602 "warning: couldn't remount RDWR "
2603 "because of unsupported optional "
2604 "features (%x)", le32_to_cpu(ret));
08c6a96f
JK		 2605 err = -EROFS;
2606 goto restore_opts;
1da177e4 2607 }
ea9a05a1
ES		 2608
2609 /*
2610 * If we have an unprocessed orphan list hanging
2611 * around from a previously readonly bdev mount,
2612 * require a full umount/remount for now.
2613 */
2614 if (es->s_last_orphan) {
4cf46b67 2615 ext3_msg(sb, KERN_WARNING, "warning: couldn't "
ea9a05a1
ES		 2616 "remount RDWR because of unprocessed "
2617 "orphan inode list. Please "
4cf46b67 2618 "umount/remount instead.");
ea9a05a1
ES		 2619 err = -EINVAL;
2620 goto restore_opts;
2621 }
2622
1da177e4
LT		 2623 /*
2624 * Mounting a RDONLY partition read-write, so reread
2625 * and store the current valid flag. (It may have
2626 * been changed by e2fsck since we originally mounted
2627 * the partition.)
2628 */
2629 ext3_clear_journal_err(sb, es);
2630 sbi->s_mount_state = le16_to_cpu(es->s_state);
a4e4de36 2631 if ((err = ext3_group_extend(sb, es, n_blocks_count)))
08c6a96f 2632 goto restore_opts;
1da177e4
LT		 2633 if (!ext3_setup_super (sb, es, 0))
2634 sb->s_flags &= ~MS_RDONLY;
2635 }
2636 }
08c6a96f
JK		 2637#ifdef CONFIG_QUOTA
2638 /* Release old quota file names */
2639 for (i = 0; i < MAXQUOTAS; i++)
2640 if (old_opts.s_qf_names[i] &&
2641 old_opts.s_qf_names[i] != sbi->s_qf_names[i])
2642 kfree(old_opts.s_qf_names[i]);
2643#endif
bbd6851a 2644 unlock_super(sb);
337eb00a 2645 unlock_kernel();
1da177e4 2646 return 0;
08c6a96f
JK		 2647restore_opts:
2648 sb->s_flags = old_sb_flags;
2649 sbi->s_mount_opt = old_opts.s_mount_opt;
2650 sbi->s_resuid = old_opts.s_resuid;
2651 sbi->s_resgid = old_opts.s_resgid;
2652 sbi->s_commit_interval = old_opts.s_commit_interval;
2653#ifdef CONFIG_QUOTA
2654 sbi->s_jquota_fmt = old_opts.s_jquota_fmt;
2655 for (i = 0; i < MAXQUOTAS; i++) {
2656 if (sbi->s_qf_names[i] &&
2657 old_opts.s_qf_names[i] != sbi->s_qf_names[i])
2658 kfree(sbi->s_qf_names[i]);
2659 sbi->s_qf_names[i] = old_opts.s_qf_names[i];
2660 }
2661#endif
bbd6851a 2662 unlock_super(sb);
337eb00a 2663 unlock_kernel();
08c6a96f 2664 return err;
1da177e4
LT		 2665}
2666
726c3342 2667static int ext3_statfs (struct dentry * dentry, struct kstatfs * buf)
1da177e4 2668{
726c3342 2669 struct super_block *sb = dentry->d_sb;
09fe316a
AT		 2670 struct ext3_sb_info *sbi = EXT3_SB(sb);
2671 struct ext3_super_block *es = sbi->s_es;
50ee0a32 2672 u64 fsid;
1da177e4 2673
a71ce8c6
BP		 2674 if (test_opt(sb, MINIX_DF)) {
2675 sbi->s_overhead_last = 0;
2676 } else if (sbi->s_blocks_last != le32_to_cpu(es->s_blocks_count)) {
2677 unsigned long ngroups = sbi->s_groups_count, i;
2678 ext3_fsblk_t overhead = 0;
1da177e4
LT		 2679 smp_rmb();
2680
2681 /*
a71ce8c6
BP		 2682 * Compute the overhead (FS structures). This is constant
2683 * for a given filesystem unless the number of block groups
2684 * changes so we cache the previous value until it does.
1da177e4
LT		 2685 */
2686
2687 /*
2688 * All of the blocks before first_data_block are
2689 * overhead
2690 */
2691 overhead = le32_to_cpu(es->s_first_data_block);
2692
2693 /*
2694 * Add the overhead attributed to the superblock and
2695 * block group descriptors. If the sparse superblocks
2696 * feature is turned on, then not all groups have this.
2697 */
2698 for (i = 0; i < ngroups; i++) {
2699 overhead += ext3_bg_has_super(sb, i) +
2700 ext3_bg_num_gdb(sb, i);
2701 cond_resched();
2702 }
2703
2704 /*
2705 * Every block group has an inode bitmap, a block
2706 * bitmap, and an inode table.
2707 */
a71ce8c6
BP		 2708 overhead += ngroups * (2 + sbi->s_itb_per_group);
2709 sbi->s_overhead_last = overhead;
2710 smp_wmb();
2711 sbi->s_blocks_last = le32_to_cpu(es->s_blocks_count);
1da177e4
LT		 2712 }
2713
2714 buf->f_type = EXT3_SUPER_MAGIC;
2715 buf->f_bsize = sb->s_blocksize;
a71ce8c6 2716 buf->f_blocks = le32_to_cpu(es->s_blocks_count) - sbi->s_overhead_last;
52d9f3b4 2717 buf->f_bfree = percpu_counter_sum_positive(&sbi->s_freeblocks_counter);
1da177e4
LT		 2718 buf->f_bavail = buf->f_bfree - le32_to_cpu(es->s_r_blocks_count);
2719 if (buf->f_bfree < le32_to_cpu(es->s_r_blocks_count))
2720 buf->f_bavail = 0;
2721 buf->f_files = le32_to_cpu(es->s_inodes_count);
52d9f3b4 2722 buf->f_ffree = percpu_counter_sum_positive(&sbi->s_freeinodes_counter);
1da177e4 2723 buf->f_namelen = EXT3_NAME_LEN;
50ee0a32
PE		 2724 fsid = le64_to_cpup((void *)es->s_uuid) ^
2725 le64_to_cpup((void *)es->s_uuid + sizeof(u64));
2726 buf->f_fsid.val[0] = fsid & 0xFFFFFFFFUL;
2727 buf->f_fsid.val[1] = (fsid >> 32) & 0xFFFFFFFFUL;
1da177e4
LT		 2728 return 0;
2729}
2730
2731/* Helper function for writing quotas on sync - we need to start transaction before quota file
2732 * is locked for write. Otherwise the are possible deadlocks:
2733 * Process 1 Process 2
2734 * ext3_create() quota_sync()
2735 * journal_start() write_dquot()
871a2931 2736 * dquot_initialize() down(dqio_mutex)
d3be915f 2737 * down(dqio_mutex) journal_start()
1da177e4
LT		 2738 *
2739 */
2740
2741#ifdef CONFIG_QUOTA
2742
2743static inline struct inode *dquot_to_inode(struct dquot *dquot)
2744{
2745 return sb_dqopt(dquot->dq_sb)->files[dquot->dq_type];
2746}
2747
1da177e4
LT		 2748static int ext3_write_dquot(struct dquot *dquot)
2749{
2750 int ret, err;
2751 handle_t *handle;
2752 struct inode *inode;
2753
2754 inode = dquot_to_inode(dquot);
2755 handle = ext3_journal_start(inode,
1f54587b 2756 EXT3_QUOTA_TRANS_BLOCKS(dquot->dq_sb));
1da177e4
LT		 2757 if (IS_ERR(handle))
2758 return PTR_ERR(handle);
2759 ret = dquot_commit(dquot);
2760 err = ext3_journal_stop(handle);
2761 if (!ret)
2762 ret = err;
2763 return ret;
2764}
2765
2766static int ext3_acquire_dquot(struct dquot *dquot)
2767{
2768 int ret, err;
2769 handle_t *handle;
2770
2771 handle = ext3_journal_start(dquot_to_inode(dquot),
1f54587b 2772 EXT3_QUOTA_INIT_BLOCKS(dquot->dq_sb));
1da177e4
LT		 2773 if (IS_ERR(handle))
2774 return PTR_ERR(handle);
2775 ret = dquot_acquire(dquot);
2776 err = ext3_journal_stop(handle);
2777 if (!ret)
2778 ret = err;
2779 return ret;
2780}
2781
2782static int ext3_release_dquot(struct dquot *dquot)
2783{
2784 int ret, err;
2785 handle_t *handle;
2786
2787 handle = ext3_journal_start(dquot_to_inode(dquot),
1f54587b 2788 EXT3_QUOTA_DEL_BLOCKS(dquot->dq_sb));
9c3013e9
JK		 2789 if (IS_ERR(handle)) {
2790 /* Release dquot anyway to avoid endless cycle in dqput() */
2791 dquot_release(dquot);
1da177e4 2792 return PTR_ERR(handle);
9c3013e9 2793 }
1da177e4
LT		 2794 ret = dquot_release(dquot);
2795 err = ext3_journal_stop(handle);
2796 if (!ret)
2797 ret = err;
2798 return ret;
2799}
2800
2801static int ext3_mark_dquot_dirty(struct dquot *dquot)
2802{
99aeaf63 2803 /* Are we journaling quotas? */
1da177e4
LT		 2804 if (EXT3_SB(dquot->dq_sb)->s_qf_names[USRQUOTA] ||
2805 EXT3_SB(dquot->dq_sb)->s_qf_names[GRPQUOTA]) {
2806 dquot_mark_dquot_dirty(dquot);
2807 return ext3_write_dquot(dquot);
2808 } else {
2809 return dquot_mark_dquot_dirty(dquot);
2810 }
2811}
2812
2813static int ext3_write_info(struct super_block *sb, int type)
2814{
2815 int ret, err;
2816 handle_t *handle;
2817
2818 /* Data block + inode block */
2819 handle = ext3_journal_start(sb->s_root->d_inode, 2);
2820 if (IS_ERR(handle))
2821 return PTR_ERR(handle);
2822 ret = dquot_commit_info(sb, type);
2823 err = ext3_journal_stop(handle);
2824 if (!ret)
2825 ret = err;
2826 return ret;
2827}
2828
2829/*
2830 * Turn on quotas during mount time - we need to find
2831 * the quota file and such...
2832 */
2833static int ext3_quota_on_mount(struct super_block *sb, int type)
2834{
84de856e
CH		 2835 return vfs_quota_on_mount(sb, EXT3_SB(sb)->s_qf_names[type],
2836 EXT3_SB(sb)->s_jquota_fmt, type);
1da177e4
LT		 2837}
2838
2839/*
2840 * Standard function to be called on quota_on
2841 */
2842static int ext3_quota_on(struct super_block *sb, int type, int format_id,
8264613d 2843 char *name, int remount)
1da177e4
LT		 2844{
2845 int err;
8264613d 2846 struct path path;
1da177e4 2847
1f54587b
JK		 2848 if (!test_opt(sb, QUOTA))
2849 return -EINVAL;
8264613d 2850 /* When remounting, no checks are needed and in fact, name is NULL */
9cfe7b90 2851 if (remount)
8264613d 2852 return vfs_quota_on(sb, type, format_id, name, remount);
9cfe7b90 2853
8264613d 2854 err = kern_path(name, LOOKUP_FOLLOW, &path);
1da177e4
LT		 2855 if (err)
2856 return err;
9cfe7b90 2857
1da177e4 2858 /* Quotafile not on the same filesystem? */
8264613d
AV		 2859 if (path.mnt->mnt_sb != sb) {
2860 path_put(&path);
1da177e4
LT		 2861 return -EXDEV;
2862 }
9cfe7b90
JK		 2863 /* Journaling quota? */
2864 if (EXT3_SB(sb)->s_qf_names[type]) {
2865 /* Quotafile not of fs root? */
8264613d 2866 if (path.dentry->d_parent != sb->s_root)
4cf46b67
AF		 2867 ext3_msg(sb, KERN_WARNING,
2868 "warning: Quota file not on filesystem root. "
2869 "Journaled quota will not work.");
9cfe7b90
JK		 2870 }
2871
2872 /*
2873 * When we journal data on quota file, we have to flush journal to see
2874 * all updates to the file when we bypass pagecache...
2875 */
8264613d 2876 if (ext3_should_journal_data(path.dentry->d_inode)) {
9cfe7b90
JK		 2877 /*
2878 * We don't need to lock updates but journal_flush() could
2879 * otherwise be livelocked...
2880 */
2881 journal_lock_updates(EXT3_SB(sb)->s_journal);
2d7c820e 2882 err = journal_flush(EXT3_SB(sb)->s_journal);
9cfe7b90 2883 journal_unlock_updates(EXT3_SB(sb)->s_journal);
2d7c820e 2884 if (err) {
12e1ec9f 2885 path_put(&path);
2d7c820e
HK		 2886 return err;
2887 }
9cfe7b90
JK		 2888 }
2889
8264613d
AV		 2890 err = vfs_quota_on_path(sb, type, format_id, &path);
2891 path_put(&path);
77e69dac 2892 return err;
1da177e4
LT		 2893}
2894
2895/* Read data from quotafile - avoid pagecache and such because we cannot afford
2896 * acquiring the locks... As quota files are never truncated and quota code
2897 * itself serializes the operations (and noone else should touch the files)
2898 * we don't have to be afraid of races */
2899static ssize_t ext3_quota_read(struct super_block *sb, int type, char *data,
2900 size_t len, loff_t off)
2901{
2902 struct inode *inode = sb_dqopt(sb)->files[type];
2903 sector_t blk = off >> EXT3_BLOCK_SIZE_BITS(sb);
2904 int err = 0;
2905 int offset = off & (sb->s_blocksize - 1);
2906 int tocopy;
2907 size_t toread;
2908 struct buffer_head *bh;
2909 loff_t i_size = i_size_read(inode);
2910
2911 if (off > i_size)
2912 return 0;
2913 if (off+len > i_size)
2914 len = i_size-off;
2915 toread = len;
2916 while (toread > 0) {
2917 tocopy = sb->s_blocksize - offset < toread ?
2918 sb->s_blocksize - offset : toread;
2919 bh = ext3_bread(NULL, inode, blk, 0, &err);
2920 if (err)
2921 return err;
2922 if (!bh) /* A hole? */
2923 memset(data, 0, tocopy);
2924 else
2925 memcpy(data, bh->b_data+offset, tocopy);
2926 brelse(bh);
2927 offset = 0;
2928 toread -= tocopy;
2929 data += tocopy;
2930 blk++;
2931 }
2932 return len;
2933}
2934
2935/* Write to quotafile (we know the transaction is already started and has
2936 * enough credits) */
2937static ssize_t ext3_quota_write(struct super_block *sb, int type,
2938 const char *data, size_t len, loff_t off)
2939{
2940 struct inode *inode = sb_dqopt(sb)->files[type];
2941 sector_t blk = off >> EXT3_BLOCK_SIZE_BITS(sb);
2942 int err = 0;
2943 int offset = off & (sb->s_blocksize - 1);
1da177e4 2944 int journal_quota = EXT3_SB(sb)->s_qf_names[type] != NULL;
1da177e4
LT		 2945 struct buffer_head *bh;
2946 handle_t *handle = journal_current_handle();
2947
9c3013e9 2948 if (!handle) {
4cf46b67
AF		 2949 ext3_msg(sb, KERN_WARNING,
2950 "warning: quota write (off=%llu, len=%llu)"
2951 " cancelled because transaction is not started.",
9c3013e9
JK		 2952 (unsigned long long)off, (unsigned long long)len);
2953 return -EIO;
2954 }
e5472147
DM		 2955
2956 /*
2957 * Since we account only one data block in transaction credits,
2958 * then it is impossible to cross a block boundary.
2959 */
2960 if (sb->s_blocksize - offset < len) {
2961 ext3_msg(sb, KERN_WARNING, "Quota write (off=%llu, len=%llu)"
2962 " cancelled because not block aligned",
2963 (unsigned long long)off, (unsigned long long)len);
2964 return -EIO;
2965 }
5c81a419 2966 mutex_lock_nested(&inode->i_mutex, I_MUTEX_QUOTA);
e5472147
DM		 2967 bh = ext3_bread(handle, inode, blk, 1, &err);
2968 if (!bh)
2969 goto out;
2970 if (journal_quota) {
2971 err = ext3_journal_get_write_access(handle, bh);
2972 if (err) {
2973 brelse(bh);
1da177e4 2974 goto out;
1da177e4 2975 }
1da177e4 2976 }
e5472147
DM		 2977 lock_buffer(bh);
2978 memcpy(bh->b_data+offset, data, len);
2979 flush_dcache_page(bh->b_page);
2980 unlock_buffer(bh);
2981 if (journal_quota)
2982 err = ext3_journal_dirty_metadata(handle, bh);
2983 else {
2984 /* Always do at least ordered writes for quotas */
2985 err = ext3_journal_dirty_data(handle, bh);
2986 mark_buffer_dirty(bh);
2987 }
2988 brelse(bh);
1da177e4 2989out:
e5472147 2990 if (err) {
f5c8f7da 2991 mutex_unlock(&inode->i_mutex);
1da177e4 2992 return err;
f5c8f7da 2993 }
e5472147
DM		 2994 if (inode->i_size < off + len) {
2995 i_size_write(inode, off + len);
1da177e4
LT		 2996 EXT3_I(inode)->i_disksize = inode->i_size;
2997 }
2998 inode->i_version++;
2999 inode->i_mtime = inode->i_ctime = CURRENT_TIME;
3000 ext3_mark_inode_dirty(handle, inode);
1b1dcc1b 3001 mutex_unlock(&inode->i_mutex);
e5472147 3002 return len;
1da177e4
LT		 3003}
3004
3005#endif
3006
454e2398
DH		 3007static int ext3_get_sb(struct file_system_type *fs_type,
3008 int flags, const char *dev_name, void *data, struct vfsmount *mnt)
1da177e4 3009{
454e2398 3010 return get_sb_bdev(fs_type, flags, dev_name, data, ext3_fill_super, mnt);
1da177e4
LT		 3011}
3012
3013static struct file_system_type ext3_fs_type = {
3014 .owner = THIS_MODULE,
3015 .name = "ext3",
3016 .get_sb = ext3_get_sb,
3017 .kill_sb = kill_block_super,
3018 .fs_flags = FS_REQUIRES_DEV,
3019};
3020
3021static int __init init_ext3_fs(void)
3022{
3023 int err = init_ext3_xattr();
3024 if (err)
3025 return err;
3026 err = init_inodecache();
3027 if (err)
3028 goto out1;
3029 err = register_filesystem(&ext3_fs_type);
3030 if (err)
3031 goto out;
3032 return 0;
3033out:
3034 destroy_inodecache();
3035out1:
e9ad5620 3036 exit_ext3_xattr();
1da177e4
LT		 3037 return err;
3038}
3039
3040static void __exit exit_ext3_fs(void)
3041{
3042 unregister_filesystem(&ext3_fs_type);
3043 destroy_inodecache();
3044 exit_ext3_xattr();
3045}
3046
3047MODULE_AUTHOR("Remy Card, Stephen Tweedie, Andrew Morton, Andreas Dilger, Theodore Ts'o and others");
3048MODULE_DESCRIPTION("Second Extended Filesystem with journaling extensions");
3049MODULE_LICENSE("GPL");
3050module_init(init_ext3_fs)
3051module_exit(exit_ext3_fs)
Clone of davem's net-next-2.6 tree