]> 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
quota: kill the vfs_dq_off and vfs_dq_quota_on_remount wrappers
[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 }
0636c73e
ES		 656
657 /*
658 * Always display barrier state so it's clear what the status is.
659 */
660 seq_puts(seq, ",barrier=");
661 seq_puts(seq, test_opt(sb, BARRIER) ? "1" : "0");
571beed8
MS		 662 if (test_opt(sb, NOBH))
663 seq_puts(seq, ",nobh");
6cd37cda 664
e3c96435 665 seq_printf(seq, ",data=%s", data_mode_string(test_opt(sb, DATA_FLAGS)));
0e4fb5e2
HK		 666 if (test_opt(sb, DATA_ERR_ABORT))
667 seq_puts(seq, ",data_err=abort");
668
dee1d3b6
ES		 669 if (test_opt(sb, NOLOAD))
670 seq_puts(seq, ",norecovery");
671
6cd37cda 672 ext3_show_quota_options(seq, sb);
8fc2751b
MB		 673
674 return 0;
675}
1da177e4 676
fdb36673 677
74af0baa
CH		 678static struct inode *ext3_nfs_get_inode(struct super_block *sb,
679 u64 ino, u32 generation)
fdb36673 680{
fdb36673 681 struct inode *inode;
fdb36673
N		 682
683 if (ino < EXT3_FIRST_INO(sb) && ino != EXT3_ROOT_INO)
684 return ERR_PTR(-ESTALE);
685 if (ino > le32_to_cpu(EXT3_SB(sb)->s_es->s_inodes_count))
686 return ERR_PTR(-ESTALE);
687
688 /* iget isn't really right if the inode is currently unallocated!!
689 *
690 * ext3_read_inode will return a bad_inode if the inode had been
691 * deleted, so we should be safe.
692 *
693 * Currently we don't know the generation for parent directory, so
694 * a generation of 0 means "accept any"
695 */
473043dc
DH		 696 inode = ext3_iget(sb, ino);
697 if (IS_ERR(inode))
698 return ERR_CAST(inode);
699 if (generation && inode->i_generation != generation) {
fdb36673
N		 700 iput(inode);
701 return ERR_PTR(-ESTALE);
702 }
74af0baa
CH		 703
704 return inode;
705}
706
707static struct dentry *ext3_fh_to_dentry(struct super_block *sb, struct fid *fid,
708 int fh_len, int fh_type)
709{
710 return generic_fh_to_dentry(sb, fid, fh_len, fh_type,
711 ext3_nfs_get_inode);
712}
713
714static struct dentry *ext3_fh_to_parent(struct super_block *sb, struct fid *fid,
715 int fh_len, int fh_type)
716{
717 return generic_fh_to_parent(sb, fid, fh_len, fh_type,
718 ext3_nfs_get_inode);
fdb36673
N		 719}
720
6b082b53
TO		 721/*
722 * Try to release metadata pages (indirect blocks, directories) which are
723 * mapped via the block device. Since these pages could have journal heads
724 * which would prevent try_to_free_buffers() from freeing them, we must use
725 * jbd layer's try_to_free_buffers() function to release them.
726 */
727static int bdev_try_to_free_page(struct super_block *sb, struct page *page,
728 gfp_t wait)
729{
730 journal_t *journal = EXT3_SB(sb)->s_journal;
731
732 WARN_ON(PageChecked(page));
733 if (!page_has_buffers(page))
734 return 0;
735 if (journal)
736 return journal_try_to_free_buffers(journal, page,
737 wait & ~__GFP_WAIT);
738 return try_to_free_buffers(page);
739}
740
8fc2751b 741#ifdef CONFIG_QUOTA
1da177e4
LT		 742#define QTYPE2NAME(t) ((t)==USRQUOTA?"user":"group")
743#define QTYPE2MOPT(on, t) ((t)==USRQUOTA?((on)##USRJQUOTA):((on)##GRPJQUOTA))
744
1da177e4
LT		 745static int ext3_write_dquot(struct dquot *dquot);
746static int ext3_acquire_dquot(struct dquot *dquot);
747static int ext3_release_dquot(struct dquot *dquot);
748static int ext3_mark_dquot_dirty(struct dquot *dquot);
749static int ext3_write_info(struct super_block *sb, int type);
2fd83a4f
JK		 750static int ext3_quota_on(struct super_block *sb, int type, int format_id,
751 char *path, int remount);
1da177e4
LT		 752static int ext3_quota_on_mount(struct super_block *sb, int type);
753static ssize_t ext3_quota_read(struct super_block *sb, int type, char *data,
754 size_t len, loff_t off);
755static ssize_t ext3_quota_write(struct super_block *sb, int type,
756 const char *data, size_t len, loff_t off);
757
61e225dc 758static const struct dquot_operations ext3_quota_operations = {
1da177e4
LT		 759 .write_dquot = ext3_write_dquot,
760 .acquire_dquot = ext3_acquire_dquot,
761 .release_dquot = ext3_release_dquot,
762 .mark_dirty = ext3_mark_dquot_dirty,
157091a2
JK		 763 .write_info = ext3_write_info,
764 .alloc_dquot = dquot_alloc,
765 .destroy_dquot = dquot_destroy,
1da177e4
LT		 766};
767
0d54b217 768static const struct quotactl_ops ext3_qctl_operations = {
1da177e4
LT		 769 .quota_on = ext3_quota_on,
770 .quota_off = vfs_quota_off,
771 .quota_sync = vfs_quota_sync,
772 .get_info = vfs_get_dqinfo,
773 .set_info = vfs_set_dqinfo,
774 .get_dqblk = vfs_get_dqblk,
775 .set_dqblk = vfs_set_dqblk
776};
777#endif
778
ee9b6d61 779static const struct super_operations ext3_sops = {
1da177e4
LT		 780 .alloc_inode = ext3_alloc_inode,
781 .destroy_inode = ext3_destroy_inode,
1da177e4
LT		 782 .write_inode = ext3_write_inode,
783 .dirty_inode = ext3_dirty_inode,
784 .delete_inode = ext3_delete_inode,
785 .put_super = ext3_put_super,
1da177e4 786 .sync_fs = ext3_sync_fs,
c4be0c1d
TS		 787 .freeze_fs = ext3_freeze,
788 .unfreeze_fs = ext3_unfreeze,
1da177e4
LT		 789 .statfs = ext3_statfs,
790 .remount_fs = ext3_remount,
791 .clear_inode = ext3_clear_inode,
8fc2751b 792 .show_options = ext3_show_options,
1da177e4
LT		 793#ifdef CONFIG_QUOTA
794 .quota_read = ext3_quota_read,
795 .quota_write = ext3_quota_write,
796#endif
6b082b53 797 .bdev_try_to_free_page = bdev_try_to_free_page,
1da177e4
LT		 798};
799
39655164 800static const struct export_operations ext3_export_ops = {
74af0baa
CH		 801 .fh_to_dentry = ext3_fh_to_dentry,
802 .fh_to_parent = ext3_fh_to_parent,
1da177e4
LT		 803 .get_parent = ext3_get_parent,
804};
805
806enum {
807 Opt_bsd_df, Opt_minix_df, Opt_grpid, Opt_nogrpid,
808 Opt_resgid, Opt_resuid, Opt_sb, Opt_err_cont, Opt_err_panic, Opt_err_ro,
2860b733 809 Opt_nouid32, Opt_nocheck, Opt_debug, Opt_oldalloc, Opt_orlov,
1da177e4 810 Opt_user_xattr, Opt_nouser_xattr, Opt_acl, Opt_noacl,
ade1a29e 811 Opt_reservation, Opt_noreservation, Opt_noload, Opt_nobh, Opt_bh,
71b96257 812 Opt_commit, Opt_journal_update, Opt_journal_inum, Opt_journal_dev,
1da177e4 813 Opt_abort, Opt_data_journal, Opt_data_ordered, Opt_data_writeback,
0e4fb5e2 814 Opt_data_err_abort, Opt_data_err_ignore,
1da177e4 815 Opt_usrjquota, Opt_grpjquota, Opt_offusrjquota, Opt_offgrpjquota,
1aeec434 816 Opt_jqfmt_vfsold, Opt_jqfmt_vfsv0, Opt_jqfmt_vfsv1, Opt_quota,
0636c73e
ES		 817 Opt_noquota, Opt_ignore, Opt_barrier, Opt_nobarrier, Opt_err,
818 Opt_resize, Opt_usrquota, Opt_grpquota
1da177e4
LT		 819};
820
a447c093 821static const match_table_t tokens = {
1da177e4
LT		 822 {Opt_bsd_df, "bsddf"},
823 {Opt_minix_df, "minixdf"},
824 {Opt_grpid, "grpid"},
825 {Opt_grpid, "bsdgroups"},
826 {Opt_nogrpid, "nogrpid"},
827 {Opt_nogrpid, "sysvgroups"},
828 {Opt_resgid, "resgid=%u"},
829 {Opt_resuid, "resuid=%u"},
830 {Opt_sb, "sb=%u"},
831 {Opt_err_cont, "errors=continue"},
832 {Opt_err_panic, "errors=panic"},
833 {Opt_err_ro, "errors=remount-ro"},
834 {Opt_nouid32, "nouid32"},
835 {Opt_nocheck, "nocheck"},
836 {Opt_nocheck, "check=none"},
1da177e4
LT		 837 {Opt_debug, "debug"},
838 {Opt_oldalloc, "oldalloc"},
839 {Opt_orlov, "orlov"},
840 {Opt_user_xattr, "user_xattr"},
841 {Opt_nouser_xattr, "nouser_xattr"},
842 {Opt_acl, "acl"},
843 {Opt_noacl, "noacl"},
844 {Opt_reservation, "reservation"},
845 {Opt_noreservation, "noreservation"},
846 {Opt_noload, "noload"},
dee1d3b6 847 {Opt_noload, "norecovery"},
1da177e4 848 {Opt_nobh, "nobh"},
ade1a29e 849 {Opt_bh, "bh"},
1da177e4
LT		 850 {Opt_commit, "commit=%u"},
851 {Opt_journal_update, "journal=update"},
852 {Opt_journal_inum, "journal=%u"},
71b96257 853 {Opt_journal_dev, "journal_dev=%u"},
1da177e4
LT		 854 {Opt_abort, "abort"},
855 {Opt_data_journal, "data=journal"},
856 {Opt_data_ordered, "data=ordered"},
857 {Opt_data_writeback, "data=writeback"},
0e4fb5e2
HK		 858 {Opt_data_err_abort, "data_err=abort"},
859 {Opt_data_err_ignore, "data_err=ignore"},
1da177e4
LT		 860 {Opt_offusrjquota, "usrjquota="},
861 {Opt_usrjquota, "usrjquota=%s"},
862 {Opt_offgrpjquota, "grpjquota="},
863 {Opt_grpjquota, "grpjquota=%s"},
864 {Opt_jqfmt_vfsold, "jqfmt=vfsold"},
865 {Opt_jqfmt_vfsv0, "jqfmt=vfsv0"},
1aeec434 866 {Opt_jqfmt_vfsv1, "jqfmt=vfsv1"},
8fc2751b 867 {Opt_grpquota, "grpquota"},
1f54587b
JK		 868 {Opt_noquota, "noquota"},
869 {Opt_quota, "quota"},
8fc2751b 870 {Opt_usrquota, "usrquota"},
1da177e4 871 {Opt_barrier, "barrier=%u"},
0636c73e
ES		 872 {Opt_barrier, "barrier"},
873 {Opt_nobarrier, "nobarrier"},
1da177e4 874 {Opt_resize, "resize"},
92587216 875 {Opt_err, NULL},
1da177e4
LT		 876};
877
4cf46b67 878static ext3_fsblk_t get_sb_block(void **data, struct super_block *sb)
1da177e4 879{
e9ad5620
DK		 880 ext3_fsblk_t sb_block;
881 char *options = (char *) *data;
1da177e4
LT		 882
883 if (!options || strncmp(options, "sb=", 3) != 0)
884 return 1; /* Default location */
885 options += 3;
43d23f90 886 /*todo: use simple_strtoll with >32bit ext3 */
1da177e4
LT		 887 sb_block = simple_strtoul(options, &options, 0);
888 if (*options && *options != ',') {
4cf46b67 889 ext3_msg(sb, "error: invalid sb specification: %s",
1da177e4
LT		 890 (char *) *data);
891 return 1;
892 }
893 if (*options == ',')
894 options++;
895 *data = (void *) options;
896 return sb_block;
897}
898
e1f5c67a
DM		 899#ifdef CONFIG_QUOTA
900static int set_qf_name(struct super_block *sb, int qtype, substring_t *args)
901{
902 struct ext3_sb_info *sbi = EXT3_SB(sb);
903 char *qname;
904
905 if (sb_any_quota_loaded(sb) &&
906 !sbi->s_qf_names[qtype]) {
907 ext3_msg(sb, KERN_ERR,
908 "Cannot change journaled "
909 "quota options when quota turned on");
910 return 0;
911 }
912 qname = match_strdup(args);
913 if (!qname) {
914 ext3_msg(sb, KERN_ERR,
915 "Not enough memory for storing quotafile name");
916 return 0;
917 }
918 if (sbi->s_qf_names[qtype] &&
919 strcmp(sbi->s_qf_names[qtype], qname)) {
920 ext3_msg(sb, KERN_ERR,
921 "%s quota file already specified", QTYPE2NAME(qtype));
922 kfree(qname);
923 return 0;
924 }
925 sbi->s_qf_names[qtype] = qname;
926 if (strchr(sbi->s_qf_names[qtype], '/')) {
927 ext3_msg(sb, KERN_ERR,
928 "quotafile must be on filesystem root");
929 kfree(sbi->s_qf_names[qtype]);
930 sbi->s_qf_names[qtype] = NULL;
931 return 0;
932 }
933 set_opt(sbi->s_mount_opt, QUOTA);
934 return 1;
935}
936
937static int clear_qf_name(struct super_block *sb, int qtype) {
938
939 struct ext3_sb_info *sbi = EXT3_SB(sb);
940
941 if (sb_any_quota_loaded(sb) &&
942 sbi->s_qf_names[qtype]) {
943 ext3_msg(sb, KERN_ERR, "Cannot change journaled quota options"
944 " when quota turned on");
945 return 0;
946 }
947 /*
948 * The space will be released later when all options are confirmed
949 * to be correct
950 */
951 sbi->s_qf_names[qtype] = NULL;
952 return 1;
953}
954#endif
955
71b96257 956static int parse_options (char *options, struct super_block *sb,
eee194e7 957 unsigned int *inum, unsigned long *journal_devnum,
43d23f90 958 ext3_fsblk_t *n_blocks_count, int is_remount)
1da177e4
LT		 959{
960 struct ext3_sb_info *sbi = EXT3_SB(sb);
961 char * p;
962 substring_t args[MAX_OPT_ARGS];
963 int data_opt = 0;
964 int option;
965#ifdef CONFIG_QUOTA
e1f5c67a 966 int qfmt;
1da177e4
LT		 967#endif
968
969 if (!options)
970 return 1;
971
972 while ((p = strsep (&options, ",")) != NULL) {
973 int token;
974 if (!*p)
975 continue;
0636c73e
ES		 976 /*
977 * Initialize args struct so we know whether arg was
978 * found; some options take optional arguments.
979 */
980 args[0].to = args[0].from = 0;
1da177e4
LT		 981 token = match_token(p, tokens, args);
982 switch (token) {
983 case Opt_bsd_df:
984 clear_opt (sbi->s_mount_opt, MINIX_DF);
985 break;
986 case Opt_minix_df:
987 set_opt (sbi->s_mount_opt, MINIX_DF);
988 break;
989 case Opt_grpid:
990 set_opt (sbi->s_mount_opt, GRPID);
991 break;
992 case Opt_nogrpid:
993 clear_opt (sbi->s_mount_opt, GRPID);
994 break;
995 case Opt_resuid:
996 if (match_int(&args[0], &option))
997 return 0;
998 sbi->s_resuid = option;
999 break;
1000 case Opt_resgid:
1001 if (match_int(&args[0], &option))
1002 return 0;
1003 sbi->s_resgid = option;
1004 break;
1005 case Opt_sb:
1006 /* handled by get_sb_block() instead of here */
1007 /* *sb_block = match_int(&args[0]); */
1008 break;
1009 case Opt_err_panic:
1010 clear_opt (sbi->s_mount_opt, ERRORS_CONT);
1011 clear_opt (sbi->s_mount_opt, ERRORS_RO);
1012 set_opt (sbi->s_mount_opt, ERRORS_PANIC);
1013 break;
1014 case Opt_err_ro:
1015 clear_opt (sbi->s_mount_opt, ERRORS_CONT);
1016 clear_opt (sbi->s_mount_opt, ERRORS_PANIC);
1017 set_opt (sbi->s_mount_opt, ERRORS_RO);
1018 break;
1019 case Opt_err_cont:
1020 clear_opt (sbi->s_mount_opt, ERRORS_RO);
1021 clear_opt (sbi->s_mount_opt, ERRORS_PANIC);
1022 set_opt (sbi->s_mount_opt, ERRORS_CONT);
1023 break;
1024 case Opt_nouid32:
1025 set_opt (sbi->s_mount_opt, NO_UID32);
1026 break;
1da177e4
LT		 1027 case Opt_nocheck:
1028 clear_opt (sbi->s_mount_opt, CHECK);
1029 break;
1030 case Opt_debug:
1031 set_opt (sbi->s_mount_opt, DEBUG);
1032 break;
1033 case Opt_oldalloc:
1034 set_opt (sbi->s_mount_opt, OLDALLOC);
1035 break;
1036 case Opt_orlov:
1037 clear_opt (sbi->s_mount_opt, OLDALLOC);
1038 break;
1039#ifdef CONFIG_EXT3_FS_XATTR
1040 case Opt_user_xattr:
1041 set_opt (sbi->s_mount_opt, XATTR_USER);
1042 break;
1043 case Opt_nouser_xattr:
1044 clear_opt (sbi->s_mount_opt, XATTR_USER);
1045 break;
1046#else
1047 case Opt_user_xattr:
1048 case Opt_nouser_xattr:
4cf46b67
AF		 1049 ext3_msg(sb, KERN_INFO,
1050 "(no)user_xattr options not supported");
1da177e4
LT		 1051 break;
1052#endif
1053#ifdef CONFIG_EXT3_FS_POSIX_ACL
1054 case Opt_acl:
1055 set_opt(sbi->s_mount_opt, POSIX_ACL);
1056 break;
1057 case Opt_noacl:
1058 clear_opt(sbi->s_mount_opt, POSIX_ACL);
1059 break;
1060#else
1061 case Opt_acl:
1062 case Opt_noacl:
4cf46b67
AF		 1063 ext3_msg(sb, KERN_INFO,
1064 "(no)acl options not supported");
1da177e4
LT		 1065 break;
1066#endif
1067 case Opt_reservation:
1068 set_opt(sbi->s_mount_opt, RESERVATION);
1069 break;
1070 case Opt_noreservation:
1071 clear_opt(sbi->s_mount_opt, RESERVATION);
1072 break;
1073 case Opt_journal_update:
1074 /* @@@ FIXME */
1075 /* Eventually we will want to be able to create
1076 a journal file here. For now, only allow the
1077 user to specify an existing inode to be the
1078 journal file. */
1079 if (is_remount) {
4cf46b67
AF		 1080 ext3_msg(sb, KERN_ERR, "error: cannot specify "
1081 "journal on remount");
1da177e4
LT		 1082 return 0;
1083 }
1084 set_opt (sbi->s_mount_opt, UPDATE_JOURNAL);
1085 break;
1086 case Opt_journal_inum:
1087 if (is_remount) {
4cf46b67
AF		 1088 ext3_msg(sb, KERN_ERR, "error: cannot specify "
1089 "journal on remount");
1da177e4
LT		 1090 return 0;
1091 }
1092 if (match_int(&args[0], &option))
1093 return 0;
1094 *inum = option;
1095 break;
71b96257
JL		 1096 case Opt_journal_dev:
1097 if (is_remount) {
4cf46b67
AF		 1098 ext3_msg(sb, KERN_ERR, "error: cannot specify "
1099 "journal on remount");
71b96257
JL		 1100 return 0;
1101 }
1102 if (match_int(&args[0], &option))
1103 return 0;
1104 *journal_devnum = option;
1105 break;
1da177e4
LT		 1106 case Opt_noload:
1107 set_opt (sbi->s_mount_opt, NOLOAD);
1108 break;
1109 case Opt_commit:
1110 if (match_int(&args[0], &option))
1111 return 0;
1112 if (option < 0)
1113 return 0;
1114 if (option == 0)
1115 option = JBD_DEFAULT_MAX_COMMIT_AGE;
1116 sbi->s_commit_interval = HZ * option;
1117 break;
1118 case Opt_data_journal:
1119 data_opt = EXT3_MOUNT_JOURNAL_DATA;
1120 goto datacheck;
1121 case Opt_data_ordered:
1122 data_opt = EXT3_MOUNT_ORDERED_DATA;
1123 goto datacheck;
1124 case Opt_data_writeback:
1125 data_opt = EXT3_MOUNT_WRITEBACK_DATA;
1126 datacheck:
1127 if (is_remount) {
e3c96435 1128 if (test_opt(sb, DATA_FLAGS) == data_opt)
3c4cec65 1129 break;
4cf46b67
AF		 1130 ext3_msg(sb, KERN_ERR,
1131 "error: cannot change "
3c4cec65
JK		 1132 "data mode on remount. The filesystem "
1133 "is mounted in data=%s mode and you "
4cf46b67 1134 "try to remount it in data=%s mode.",
e3c96435
DM		 1135 data_mode_string(test_opt(sb,
1136 DATA_FLAGS)),
3c4cec65
JK		 1137 data_mode_string(data_opt));
1138 return 0;
1da177e4 1139 } else {
e3c96435 1140 clear_opt(sbi->s_mount_opt, DATA_FLAGS);
1da177e4
LT		 1141 sbi->s_mount_opt |= data_opt;
1142 }
1143 break;
0e4fb5e2
HK		 1144 case Opt_data_err_abort:
1145 set_opt(sbi->s_mount_opt, DATA_ERR_ABORT);
1146 break;
1147 case Opt_data_err_ignore:
1148 clear_opt(sbi->s_mount_opt, DATA_ERR_ABORT);
1149 break;
1da177e4
LT		 1150#ifdef CONFIG_QUOTA
1151 case Opt_usrjquota:
e1f5c67a 1152 if (!set_qf_name(sb, USRQUOTA, &args[0]))
1da177e4 1153 return 0;
e1f5c67a
DM		 1154 break;
1155 case Opt_grpjquota:
1156 if (!set_qf_name(sb, GRPQUOTA, &args[0]))
1da177e4 1157 return 0;
1da177e4
LT		 1158 break;
1159 case Opt_offusrjquota:
e1f5c67a
DM		 1160 if (!clear_qf_name(sb, USRQUOTA))
1161 return 0;
1162 break;
1da177e4 1163 case Opt_offgrpjquota:
e1f5c67a 1164 if (!clear_qf_name(sb, GRPQUOTA))
1da177e4 1165 return 0;
1da177e4
LT		 1166 break;
1167 case Opt_jqfmt_vfsold:
d06bf1d2
JK		 1168 qfmt = QFMT_VFS_OLD;
1169 goto set_qf_format;
1da177e4 1170 case Opt_jqfmt_vfsv0:
d06bf1d2 1171 qfmt = QFMT_VFS_V0;
1aeec434
JK		 1172 goto set_qf_format;
1173 case Opt_jqfmt_vfsv1:
1174 qfmt = QFMT_VFS_V1;
d06bf1d2 1175set_qf_format:
ee0d5ffe 1176 if (sb_any_quota_loaded(sb) &&
d06bf1d2 1177 sbi->s_jquota_fmt != qfmt) {
4cf46b67 1178 ext3_msg(sb, KERN_ERR, "error: cannot change "
d06bf1d2 1179 "journaled quota options when "
4cf46b67 1180 "quota turned on.");
d06bf1d2
JK		 1181 return 0;
1182 }
1183 sbi->s_jquota_fmt = qfmt;
1da177e4 1184 break;
1f54587b 1185 case Opt_quota:
8fc2751b
MB		 1186 case Opt_usrquota:
1187 set_opt(sbi->s_mount_opt, QUOTA);
1188 set_opt(sbi->s_mount_opt, USRQUOTA);
1189 break;
1190 case Opt_grpquota:
1f54587b 1191 set_opt(sbi->s_mount_opt, QUOTA);
8fc2751b 1192 set_opt(sbi->s_mount_opt, GRPQUOTA);
1f54587b
JK		 1193 break;
1194 case Opt_noquota:
ee0d5ffe 1195 if (sb_any_quota_loaded(sb)) {
4cf46b67
AF		 1196 ext3_msg(sb, KERN_ERR, "error: cannot change "
1197 "quota options when quota turned on.");
1f54587b
JK		 1198 return 0;
1199 }
1200 clear_opt(sbi->s_mount_opt, QUOTA);
8fc2751b
MB		 1201 clear_opt(sbi->s_mount_opt, USRQUOTA);
1202 clear_opt(sbi->s_mount_opt, GRPQUOTA);
1f54587b 1203 break;
1da177e4 1204#else
8fc2751b
MB		 1205 case Opt_quota:
1206 case Opt_usrquota:
1207 case Opt_grpquota:
4cf46b67
AF		 1208 ext3_msg(sb, KERN_ERR,
1209 "error: quota options not supported.");
fa1ff1e0 1210 break;
1da177e4
LT		 1211 case Opt_usrjquota:
1212 case Opt_grpjquota:
1213 case Opt_offusrjquota:
1214 case Opt_offgrpjquota:
1215 case Opt_jqfmt_vfsold:
1216 case Opt_jqfmt_vfsv0:
1aeec434 1217 case Opt_jqfmt_vfsv1:
4cf46b67
AF		 1218 ext3_msg(sb, KERN_ERR,
1219 "error: journaled quota options not "
1220 "supported.");
1da177e4 1221 break;
1f54587b
JK		 1222 case Opt_noquota:
1223 break;
1da177e4
LT		 1224#endif
1225 case Opt_abort:
1226 set_opt(sbi->s_mount_opt, ABORT);
1227 break;
0636c73e
ES		 1228 case Opt_nobarrier:
1229 clear_opt(sbi->s_mount_opt, BARRIER);
1230 break;
1da177e4 1231 case Opt_barrier:
0636c73e
ES		 1232 if (args[0].from) {
1233 if (match_int(&args[0], &option))
1234 return 0;
1235 } else
1236 option = 1; /* No argument, default to 1 */
1da177e4
LT		 1237 if (option)
1238 set_opt(sbi->s_mount_opt, BARRIER);
1239 else
1240 clear_opt(sbi->s_mount_opt, BARRIER);
1241 break;
1242 case Opt_ignore:
1243 break;
1244 case Opt_resize:
08c6a96f 1245 if (!is_remount) {
4cf46b67
AF		 1246 ext3_msg(sb, KERN_ERR,
1247 "error: resize option only available "
1248 "for remount");
1da177e4
LT		 1249 return 0;
1250 }
c7f1721e
KZ		 1251 if (match_int(&args[0], &option) != 0)
1252 return 0;
1da177e4
LT		 1253 *n_blocks_count = option;
1254 break;
1255 case Opt_nobh:
1256 set_opt(sbi->s_mount_opt, NOBH);
1257 break;
ade1a29e
BP		 1258 case Opt_bh:
1259 clear_opt(sbi->s_mount_opt, NOBH);
1260 break;
1da177e4 1261 default:
4cf46b67
AF		 1262 ext3_msg(sb, KERN_ERR,
1263 "error: unrecognized mount option \"%s\" "
1264 "or missing value", p);
1da177e4
LT		 1265 return 0;
1266 }
1267 }
1268#ifdef CONFIG_QUOTA
8fc2751b 1269 if (sbi->s_qf_names[USRQUOTA] || sbi->s_qf_names[GRPQUOTA]) {
e3c96435 1270 if (test_opt(sb, USRQUOTA) && sbi->s_qf_names[USRQUOTA])
8fc2751b 1271 clear_opt(sbi->s_mount_opt, USRQUOTA);
e3c96435 1272 if (test_opt(sb, GRPQUOTA) && sbi->s_qf_names[GRPQUOTA])
8fc2751b
MB		 1273 clear_opt(sbi->s_mount_opt, GRPQUOTA);
1274
e1f5c67a 1275 if (test_opt(sb, GRPQUOTA) || test_opt(sb, USRQUOTA)) {
4cf46b67
AF		 1276 ext3_msg(sb, KERN_ERR, "error: old and new quota "
1277 "format mixing.");
8fc2751b
MB		 1278 return 0;
1279 }
1280
1281 if (!sbi->s_jquota_fmt) {
4cf46b67
AF		 1282 ext3_msg(sb, KERN_ERR, "error: journaled quota format "
1283 "not specified.");
8fc2751b
MB		 1284 return 0;
1285 }
1286 } else {
1287 if (sbi->s_jquota_fmt) {
4cf46b67 1288 ext3_msg(sb, KERN_ERR, "error: journaled quota format "
99aeaf63 1289 "specified with no journaling "
4cf46b67 1290 "enabled.");
8fc2751b
MB		 1291 return 0;
1292 }
1da177e4
LT		 1293 }
1294#endif
1da177e4
LT		 1295 return 1;
1296}
1297
1298static int ext3_setup_super(struct super_block *sb, struct ext3_super_block *es,
1299 int read_only)
1300{
1301 struct ext3_sb_info *sbi = EXT3_SB(sb);
1302 int res = 0;
1303
1304 if (le32_to_cpu(es->s_rev_level) > EXT3_MAX_SUPP_REV) {
4cf46b67
AF		 1305 ext3_msg(sb, KERN_ERR,
1306 "error: revision level too high, "
1307 "forcing read-only mode");
1da177e4
LT		 1308 res = MS_RDONLY;
1309 }
1310 if (read_only)
1311 return res;
1312 if (!(sbi->s_mount_state & EXT3_VALID_FS))
4cf46b67
AF		 1313 ext3_msg(sb, KERN_WARNING,
1314 "warning: mounting unchecked fs, "
1315 "running e2fsck is recommended");
1da177e4 1316 else if ((sbi->s_mount_state & EXT3_ERROR_FS))
4cf46b67
AF		 1317 ext3_msg(sb, KERN_WARNING,
1318 "warning: mounting fs with errors, "
1319 "running e2fsck is recommended");
1da177e4
LT		 1320 else if ((__s16) le16_to_cpu(es->s_max_mnt_count) >= 0 &&
1321 le16_to_cpu(es->s_mnt_count) >=
1322 (unsigned short) (__s16) le16_to_cpu(es->s_max_mnt_count))
4cf46b67
AF		 1323 ext3_msg(sb, KERN_WARNING,
1324 "warning: maximal mount count reached, "
1325 "running e2fsck is recommended");
1da177e4
LT		 1326 else if (le32_to_cpu(es->s_checkinterval) &&
1327 (le32_to_cpu(es->s_lastcheck) +
1328 le32_to_cpu(es->s_checkinterval) <= get_seconds()))
4cf46b67
AF		 1329 ext3_msg(sb, KERN_WARNING,
1330 "warning: checktime reached, "
1331 "running e2fsck is recommended");
1da177e4
LT		 1332#if 0
1333 /* @@@ We _will_ want to clear the valid bit if we find
1334 inconsistencies, to force a fsck at reboot. But for
1335 a plain journaled filesystem we can keep it set as
1336 valid forever! :) */
e7f23ebd 1337 es->s_state &= cpu_to_le16(~EXT3_VALID_FS);
1da177e4
LT		 1338#endif
1339 if (!(__s16) le16_to_cpu(es->s_max_mnt_count))
1340 es->s_max_mnt_count = cpu_to_le16(EXT3_DFL_MAX_MNT_COUNT);
50e8a289 1341 le16_add_cpu(&es->s_mnt_count, 1);
1da177e4
LT		 1342 es->s_mtime = cpu_to_le32(get_seconds());
1343 ext3_update_dynamic_rev(sb);
1344 EXT3_SET_INCOMPAT_FEATURE(sb, EXT3_FEATURE_INCOMPAT_RECOVER);
1345
1346 ext3_commit_super(sb, es, 1);
1347 if (test_opt(sb, DEBUG))
4cf46b67
AF		 1348 ext3_msg(sb, KERN_INFO, "[bs=%lu, gc=%lu, "
1349 "bpg=%lu, ipg=%lu, mo=%04lx]",
1da177e4
LT		 1350 sb->s_blocksize,
1351 sbi->s_groups_count,
1352 EXT3_BLOCKS_PER_GROUP(sb),
1353 EXT3_INODES_PER_GROUP(sb),
1354 sbi->s_mount_opt);
1355
1da177e4
LT		 1356 if (EXT3_SB(sb)->s_journal->j_inode == NULL) {
1357 char b[BDEVNAME_SIZE];
4cf46b67 1358 ext3_msg(sb, KERN_INFO, "using external journal on %s",
1da177e4
LT		 1359 bdevname(EXT3_SB(sb)->s_journal->j_dev, b));
1360 } else {
4cf46b67 1361 ext3_msg(sb, KERN_INFO, "using internal journal");
1da177e4 1362 }
1da177e4
LT		 1363 return res;
1364}
1365
1366/* Called at mount-time, super-block is locked */
197cd65a 1367static int ext3_check_descriptors(struct super_block *sb)
1da177e4
LT		 1368{
1369 struct ext3_sb_info *sbi = EXT3_SB(sb);
1da177e4
LT		 1370 int i;
1371
1372 ext3_debug ("Checking group descriptors");
1373
197cd65a
AM		 1374 for (i = 0; i < sbi->s_groups_count; i++) {
1375 struct ext3_group_desc *gdp = ext3_get_group_desc(sb, i, NULL);
1eaafeae
AM		 1376 ext3_fsblk_t first_block = ext3_group_first_block_no(sb, i);
1377 ext3_fsblk_t last_block;
197cd65a 1378
855565e8
ES		 1379 if (i == sbi->s_groups_count - 1)
1380 last_block = le32_to_cpu(sbi->s_es->s_blocks_count) - 1;
1381 else
1382 last_block = first_block +
1383 (EXT3_BLOCKS_PER_GROUP(sb) - 1);
1384
855565e8
ES		 1385 if (le32_to_cpu(gdp->bg_block_bitmap) < first_block ||
1386 le32_to_cpu(gdp->bg_block_bitmap) > last_block)
1da177e4
LT		 1387 {
1388 ext3_error (sb, "ext3_check_descriptors",
1389 "Block bitmap for group %d"
1390 " not in group (block %lu)!",
1391 i, (unsigned long)
1392 le32_to_cpu(gdp->bg_block_bitmap));
1393 return 0;
1394 }
855565e8
ES		 1395 if (le32_to_cpu(gdp->bg_inode_bitmap) < first_block ||
1396 le32_to_cpu(gdp->bg_inode_bitmap) > last_block)
1da177e4
LT		 1397 {
1398 ext3_error (sb, "ext3_check_descriptors",
1399 "Inode bitmap for group %d"
1400 " not in group (block %lu)!",
1401 i, (unsigned long)
1402 le32_to_cpu(gdp->bg_inode_bitmap));
1403 return 0;
1404 }
855565e8 1405 if (le32_to_cpu(gdp->bg_inode_table) < first_block ||
780dcdb2 1406 le32_to_cpu(gdp->bg_inode_table) + sbi->s_itb_per_group - 1 >
855565e8 1407 last_block)
1da177e4
LT		 1408 {
1409 ext3_error (sb, "ext3_check_descriptors",
1410 "Inode table for group %d"
1411 " not in group (block %lu)!",
1412 i, (unsigned long)
1413 le32_to_cpu(gdp->bg_inode_table));
1414 return 0;
1415 }
1da177e4
LT		 1416 }
1417
1418 sbi->s_es->s_free_blocks_count=cpu_to_le32(ext3_count_free_blocks(sb));
1419 sbi->s_es->s_free_inodes_count=cpu_to_le32(ext3_count_free_inodes(sb));
1420 return 1;
1421}
1422
1423
1424/* ext3_orphan_cleanup() walks a singly-linked list of inodes (starting at
1425 * the superblock) which were deleted from all directories, but held open by
1426 * a process at the time of a crash. We walk the list and try to delete these
1427 * inodes at recovery time (only with a read-write filesystem).
1428 *
1429 * In order to keep the orphan inode chain consistent during traversal (in
1430 * case of crash during recovery), we link each inode into the superblock
1431 * orphan list_head and handle it the same way as an inode deletion during
1432 * normal operation (which journals the operations for us).
1433 *
1434 * We only do an iget() and an iput() on each inode, which is very safe if we
1435 * accidentally point at an in-use or already deleted inode. The worst that
1436 * can happen in this case is that we get a "bit already cleared" message from
1437 * ext3_free_inode(). The only reason we would point at a wrong inode is if
1438 * e2fsck was run on this filesystem, and it must have already done the orphan
1439 * inode cleanup for us, so we can safely abort without any further action.
1440 */
1441static void ext3_orphan_cleanup (struct super_block * sb,
1442 struct ext3_super_block * es)
1443{
1444 unsigned int s_flags = sb->s_flags;
1445 int nr_orphans = 0, nr_truncates = 0;
1446#ifdef CONFIG_QUOTA
1447 int i;
1448#endif
1449 if (!es->s_last_orphan) {
1450 jbd_debug(4, "no orphan inodes to clean up\n");
1451 return;
1452 }
1453
a8f48a95 1454 if (bdev_read_only(sb->s_bdev)) {
4cf46b67
AF		 1455 ext3_msg(sb, KERN_ERR, "error: write access "
1456 "unavailable, skipping orphan cleanup.");
a8f48a95
ES		 1457 return;
1458 }
1459
1da177e4
LT		 1460 if (EXT3_SB(sb)->s_mount_state & EXT3_ERROR_FS) {
1461 if (es->s_last_orphan)
1462 jbd_debug(1, "Errors on filesystem, "
1463 "clearing orphan list.\n");
1464 es->s_last_orphan = 0;
1465 jbd_debug(1, "Skipping orphan recovery on fs with errors.\n");
1466 return;
1467 }
1468
1469 if (s_flags & MS_RDONLY) {
4cf46b67 1470 ext3_msg(sb, KERN_INFO, "orphan cleanup on readonly fs");
1da177e4
LT		 1471 sb->s_flags &= ~MS_RDONLY;
1472 }
1473#ifdef CONFIG_QUOTA
1474 /* Needed for iput() to work correctly and not trash data */
1475 sb->s_flags |= MS_ACTIVE;
1476 /* Turn on quotas so that they are updated correctly */
1477 for (i = 0; i < MAXQUOTAS; i++) {
1478 if (EXT3_SB(sb)->s_qf_names[i]) {
1479 int ret = ext3_quota_on_mount(sb, i);
1480 if (ret < 0)
4cf46b67
AF		 1481 ext3_msg(sb, KERN_ERR,
1482 "error: cannot turn on journaled "
1483 "quota: %d", ret);
1da177e4
LT		 1484 }
1485 }
1486#endif
1487
1488 while (es->s_last_orphan) {
1489 struct inode *inode;
1490
473043dc
DH		 1491 inode = ext3_orphan_get(sb, le32_to_cpu(es->s_last_orphan));
1492 if (IS_ERR(inode)) {
1da177e4
LT		 1493 es->s_last_orphan = 0;
1494 break;
1495 }
1496
1497 list_add(&EXT3_I(inode)->i_orphan, &EXT3_SB(sb)->s_orphan);
871a2931 1498 dquot_initialize(inode);
1da177e4
LT		 1499 if (inode->i_nlink) {
1500 printk(KERN_DEBUG
eee194e7 1501 "%s: truncating inode %lu to %Ld bytes\n",
e05b6b52 1502 __func__, inode->i_ino, inode->i_size);
eee194e7 1503 jbd_debug(2, "truncating inode %lu to %Ld bytes\n",
1da177e4
LT		 1504 inode->i_ino, inode->i_size);
1505 ext3_truncate(inode);
1506 nr_truncates++;
1507 } else {
1508 printk(KERN_DEBUG
eee194e7 1509 "%s: deleting unreferenced inode %lu\n",
e05b6b52 1510 __func__, inode->i_ino);
eee194e7 1511 jbd_debug(2, "deleting unreferenced inode %lu\n",
1da177e4
LT		 1512 inode->i_ino);
1513 nr_orphans++;
1514 }
1515 iput(inode); /* The delete magic happens here! */
1516 }
1517
1518#define PLURAL(x) (x), ((x)==1) ? "" : "s"
1519
1520 if (nr_orphans)
4cf46b67
AF		 1521 ext3_msg(sb, KERN_INFO, "%d orphan inode%s deleted",
1522 PLURAL(nr_orphans));
1da177e4 1523 if (nr_truncates)
4cf46b67
AF		 1524 ext3_msg(sb, KERN_INFO, "%d truncate%s cleaned up",
1525 PLURAL(nr_truncates));
1da177e4
LT		 1526#ifdef CONFIG_QUOTA
1527 /* Turn quotas off */
1528 for (i = 0; i < MAXQUOTAS; i++) {
1529 if (sb_dqopt(sb)->files[i])
2fd83a4f 1530 vfs_quota_off(sb, i, 0);
1da177e4
LT		 1531 }
1532#endif
1533 sb->s_flags = s_flags; /* Restore MS_RDONLY status */
1534}
1535
1da177e4
LT		 1536/*
1537 * Maximal file size. There is a direct, and {,double-,triple-}indirect
1538 * block limit, and also a limit of (2^32 - 1) 512-byte sectors in i_blocks.
1539 * We need to be 1 filesystem block less than the 2^32 sector limit.
1540 */
1541static loff_t ext3_max_size(int bits)
1542{
1543 loff_t res = EXT3_NDIR_BLOCKS;
fe7fdc37
AK		 1544 int meta_blocks;
1545 loff_t upper_limit;
1546
1547 /* This is calculated to be the largest file size for a
1548 * dense, file such that the total number of
1da177e4 1549 * sectors in the file, including data and all indirect blocks,
fe7fdc37
AK		 1550 * does not exceed 2^32 -1
1551 * __u32 i_blocks representing the total number of
1552 * 512 bytes blocks of the file
1553 */
1554 upper_limit = (1LL << 32) - 1;
1555
1556 /* total blocks in file system block size */
1557 upper_limit >>= (bits - 9);
1558
1559
1560 /* indirect blocks */
1561 meta_blocks = 1;
1562 /* double indirect blocks */
1563 meta_blocks += 1 + (1LL << (bits-2));
1564 /* tripple indirect blocks */
1565 meta_blocks += 1 + (1LL << (bits-2)) + (1LL << (2*(bits-2)));
1566
1567 upper_limit -= meta_blocks;
1568 upper_limit <<= bits;
1da177e4
LT		 1569
1570 res += 1LL << (bits-2);
1571 res += 1LL << (2*(bits-2));
1572 res += 1LL << (3*(bits-2));
1573 res <<= bits;
1574 if (res > upper_limit)
1575 res = upper_limit;
fe7fdc37
AK		 1576
1577 if (res > MAX_LFS_FILESIZE)
1578 res = MAX_LFS_FILESIZE;
1579
1da177e4
LT		 1580 return res;
1581}
1582
43d23f90
MC		 1583static ext3_fsblk_t descriptor_loc(struct super_block *sb,
1584 ext3_fsblk_t logic_sb_block,
1da177e4
LT		 1585 int nr)
1586{
1587 struct ext3_sb_info *sbi = EXT3_SB(sb);
43d23f90 1588 unsigned long bg, first_meta_bg;
1da177e4
LT		 1589 int has_super = 0;
1590
1da177e4
LT		 1591 first_meta_bg = le32_to_cpu(sbi->s_es->s_first_meta_bg);
1592
1593 if (!EXT3_HAS_INCOMPAT_FEATURE(sb, EXT3_FEATURE_INCOMPAT_META_BG) ||
1594 nr < first_meta_bg)
1595 return (logic_sb_block + nr + 1);
1596 bg = sbi->s_desc_per_block * nr;
1597 if (ext3_bg_has_super(sb, bg))
1598 has_super = 1;
43d23f90 1599 return (has_super + ext3_group_first_block_no(sb, bg));
1da177e4
LT		 1600}
1601
1602
1603static int ext3_fill_super (struct super_block *sb, void *data, int silent)
1604{
1605 struct buffer_head * bh;
1606 struct ext3_super_block *es = NULL;
1607 struct ext3_sb_info *sbi;
43d23f90 1608 ext3_fsblk_t block;
4cf46b67 1609 ext3_fsblk_t sb_block = get_sb_block(&data, sb);
43d23f90 1610 ext3_fsblk_t logic_sb_block;
1da177e4 1611 unsigned long offset = 0;
eee194e7 1612 unsigned int journal_inum = 0;
71b96257 1613 unsigned long journal_devnum = 0;
1da177e4
LT		 1614 unsigned long def_mount_opts;
1615 struct inode *root;
1616 int blocksize;
1617 int hblock;
1618 int db_count;
1619 int i;
1620 int needs_recovery;
473043dc 1621 int ret = -EINVAL;
1da177e4 1622 __le32 features;
833f4077 1623 int err;
1da177e4 1624
f8314dc6 1625 sbi = kzalloc(sizeof(*sbi), GFP_KERNEL);
1da177e4
LT		 1626 if (!sbi)
1627 return -ENOMEM;
5df096d6
PE		 1628
1629 sbi->s_blockgroup_lock =
1630 kzalloc(sizeof(struct blockgroup_lock), GFP_KERNEL);
1631 if (!sbi->s_blockgroup_lock) {
1632 kfree(sbi);
1633 return -ENOMEM;
1634 }
1da177e4 1635 sb->s_fs_info = sbi;
1da177e4
LT		 1636 sbi->s_mount_opt = 0;
1637 sbi->s_resuid = EXT3_DEF_RESUID;
1638 sbi->s_resgid = EXT3_DEF_RESGID;
571beed8 1639 sbi->s_sb_block = sb_block;
1da177e4
LT		 1640
1641 unlock_kernel();
1642
1643 blocksize = sb_min_blocksize(sb, EXT3_MIN_BLOCK_SIZE);
1644 if (!blocksize) {
4cf46b67 1645 ext3_msg(sb, KERN_ERR, "error: unable to set blocksize");
1da177e4
LT		 1646 goto out_fail;
1647 }
1648
1649 /*
1650 * The ext3 superblock will not be buffer aligned for other than 1kB
1651 * block sizes. We need to calculate the offset from buffer start.
1652 */
1653 if (blocksize != EXT3_MIN_BLOCK_SIZE) {
1654 logic_sb_block = (sb_block * EXT3_MIN_BLOCK_SIZE) / blocksize;
1655 offset = (sb_block * EXT3_MIN_BLOCK_SIZE) % blocksize;
1656 } else {
1657 logic_sb_block = sb_block;
1658 }
1659
1660 if (!(bh = sb_bread(sb, logic_sb_block))) {
4cf46b67 1661 ext3_msg(sb, KERN_ERR, "error: unable to read superblock");
1da177e4
LT		 1662 goto out_fail;
1663 }
1664 /*
1665 * Note: s_es must be initialized as soon as possible because
1666 * some ext3 macro-instructions depend on its value
1667 */
1668 es = (struct ext3_super_block *) (((char *)bh->b_data) + offset);
1669 sbi->s_es = es;
1670 sb->s_magic = le16_to_cpu(es->s_magic);
1671 if (sb->s_magic != EXT3_SUPER_MAGIC)
1672 goto cantfind_ext3;
1673
1674 /* Set defaults before we parse the mount options */
1675 def_mount_opts = le32_to_cpu(es->s_default_mount_opts);
1676 if (def_mount_opts & EXT3_DEFM_DEBUG)
1677 set_opt(sbi->s_mount_opt, DEBUG);
1678 if (def_mount_opts & EXT3_DEFM_BSDGROUPS)
1679 set_opt(sbi->s_mount_opt, GRPID);
1680 if (def_mount_opts & EXT3_DEFM_UID16)
1681 set_opt(sbi->s_mount_opt, NO_UID32);
2e7842b8 1682#ifdef CONFIG_EXT3_FS_XATTR
1da177e4
LT		 1683 if (def_mount_opts & EXT3_DEFM_XATTR_USER)
1684 set_opt(sbi->s_mount_opt, XATTR_USER);
2e7842b8
HD		 1685#endif
1686#ifdef CONFIG_EXT3_FS_POSIX_ACL
1da177e4
LT		 1687 if (def_mount_opts & EXT3_DEFM_ACL)
1688 set_opt(sbi->s_mount_opt, POSIX_ACL);
2e7842b8 1689#endif
1da177e4 1690 if ((def_mount_opts & EXT3_DEFM_JMODE) == EXT3_DEFM_JMODE_DATA)
e3c96435 1691 set_opt(sbi->s_mount_opt, JOURNAL_DATA);
1da177e4 1692 else if ((def_mount_opts & EXT3_DEFM_JMODE) == EXT3_DEFM_JMODE_ORDERED)
e3c96435 1693 set_opt(sbi->s_mount_opt, ORDERED_DATA);
1da177e4 1694 else if ((def_mount_opts & EXT3_DEFM_JMODE) == EXT3_DEFM_JMODE_WBACK)
e3c96435 1695 set_opt(sbi->s_mount_opt, WRITEBACK_DATA);
1da177e4
LT		 1696
1697 if (le16_to_cpu(sbi->s_es->s_errors) == EXT3_ERRORS_PANIC)
1698 set_opt(sbi->s_mount_opt, ERRORS_PANIC);
1eca93f9 1699 else if (le16_to_cpu(sbi->s_es->s_errors) == EXT3_ERRORS_CONTINUE)
2245d7c2 1700 set_opt(sbi->s_mount_opt, ERRORS_CONT);
1eca93f9
AK		 1701 else
1702 set_opt(sbi->s_mount_opt, ERRORS_RO);
1da177e4
LT		 1703
1704 sbi->s_resuid = le16_to_cpu(es->s_def_resuid);
1705 sbi->s_resgid = le16_to_cpu(es->s_def_resgid);
1706
1707 set_opt(sbi->s_mount_opt, RESERVATION);
1708
71b96257
JL		 1709 if (!parse_options ((char *) data, sb, &journal_inum, &journal_devnum,
1710 NULL, 0))
1da177e4
LT		 1711 goto failed_mount;
1712
1713 sb->s_flags = (sb->s_flags & ~MS_POSIXACL) |
e3c96435 1714 (test_opt(sb, POSIX_ACL) ? MS_POSIXACL : 0);
1da177e4
LT		 1715
1716 if (le32_to_cpu(es->s_rev_level) == EXT3_GOOD_OLD_REV &&
1717 (EXT3_HAS_COMPAT_FEATURE(sb, ~0U) ||
1718 EXT3_HAS_RO_COMPAT_FEATURE(sb, ~0U) ||
1719 EXT3_HAS_INCOMPAT_FEATURE(sb, ~0U)))
4cf46b67
AF		 1720 ext3_msg(sb, KERN_WARNING,
1721 "warning: feature flags set on rev 0 fs, "
1722 "running e2fsck is recommended");
1da177e4
LT		 1723 /*
1724 * Check feature flags regardless of the revision level, since we
1725 * previously didn't change the revision level when setting the flags,
1726 * so there is a chance incompat flags are set on a rev 0 filesystem.
1727 */
1728 features = EXT3_HAS_INCOMPAT_FEATURE(sb, ~EXT3_FEATURE_INCOMPAT_SUPP);
1729 if (features) {
4cf46b67
AF		 1730 ext3_msg(sb, KERN_ERR,
1731 "error: couldn't mount because of unsupported "
1732 "optional features (%x)", le32_to_cpu(features));
1da177e4
LT		 1733 goto failed_mount;
1734 }
1735 features = EXT3_HAS_RO_COMPAT_FEATURE(sb, ~EXT3_FEATURE_RO_COMPAT_SUPP);
1736 if (!(sb->s_flags & MS_RDONLY) && features) {
4cf46b67
AF		 1737 ext3_msg(sb, KERN_ERR,
1738 "error: couldn't mount RDWR because of unsupported "
1739 "optional features (%x)", le32_to_cpu(features));
1da177e4
LT		 1740 goto failed_mount;
1741 }
1742 blocksize = BLOCK_SIZE << le32_to_cpu(es->s_log_block_size);
1743
1744 if (blocksize < EXT3_MIN_BLOCK_SIZE ||
1745 blocksize > EXT3_MAX_BLOCK_SIZE) {
4cf46b67
AF		 1746 ext3_msg(sb, KERN_ERR,
1747 "error: couldn't mount because of unsupported "
1748 "filesystem blocksize %d", blocksize);
1da177e4
LT		 1749 goto failed_mount;
1750 }
1751
e1defc4f 1752 hblock = bdev_logical_block_size(sb->s_bdev);
1da177e4
LT		 1753 if (sb->s_blocksize != blocksize) {
1754 /*
1755 * Make sure the blocksize for the filesystem is larger
1756 * than the hardware sectorsize for the machine.
1757 */
1758 if (blocksize < hblock) {
4cf46b67
AF		 1759 ext3_msg(sb, KERN_ERR,
1760 "error: fsblocksize %d too small for "
1761 "hardware sectorsize %d", blocksize, hblock);
1da177e4
LT		 1762 goto failed_mount;
1763 }
1764
1765 brelse (bh);
0f0a89eb 1766 if (!sb_set_blocksize(sb, blocksize)) {
4cf46b67
AF		 1767 ext3_msg(sb, KERN_ERR,
1768 "error: bad blocksize %d", blocksize);
0f0a89eb
TS		 1769 goto out_fail;
1770 }
1da177e4
LT		 1771 logic_sb_block = (sb_block * EXT3_MIN_BLOCK_SIZE) / blocksize;
1772 offset = (sb_block * EXT3_MIN_BLOCK_SIZE) % blocksize;
1773 bh = sb_bread(sb, logic_sb_block);
1774 if (!bh) {
4cf46b67
AF		 1775 ext3_msg(sb, KERN_ERR,
1776 "error: can't read superblock on 2nd try");
1da177e4
LT		 1777 goto failed_mount;
1778 }
1779 es = (struct ext3_super_block *)(((char *)bh->b_data) + offset);
1780 sbi->s_es = es;
1781 if (es->s_magic != cpu_to_le16(EXT3_SUPER_MAGIC)) {
4cf46b67
AF		 1782 ext3_msg(sb, KERN_ERR,
1783 "error: magic mismatch");
1da177e4
LT		 1784 goto failed_mount;
1785 }
1786 }
1787
1788 sb->s_maxbytes = ext3_max_size(sb->s_blocksize_bits);
1789
1790 if (le32_to_cpu(es->s_rev_level) == EXT3_GOOD_OLD_REV) {
1791 sbi->s_inode_size = EXT3_GOOD_OLD_INODE_SIZE;
1792 sbi->s_first_ino = EXT3_GOOD_OLD_FIRST_INO;
1793 } else {
1794 sbi->s_inode_size = le16_to_cpu(es->s_inode_size);
1795 sbi->s_first_ino = le32_to_cpu(es->s_first_ino);
1796 if ((sbi->s_inode_size < EXT3_GOOD_OLD_INODE_SIZE) ||
3fc74269 1797 (!is_power_of_2(sbi->s_inode_size)) ||
1da177e4 1798 (sbi->s_inode_size > blocksize)) {
4cf46b67
AF		 1799 ext3_msg(sb, KERN_ERR,
1800 "error: unsupported inode size: %d",
1da177e4
LT		 1801 sbi->s_inode_size);
1802 goto failed_mount;
1803 }
1804 }
1805 sbi->s_frag_size = EXT3_MIN_FRAG_SIZE <<
1806 le32_to_cpu(es->s_log_frag_size);
1807 if (blocksize != sbi->s_frag_size) {
4cf46b67
AF		 1808 ext3_msg(sb, KERN_ERR,
1809 "error: fragsize %lu != blocksize %u (unsupported)",
1da177e4
LT		 1810 sbi->s_frag_size, blocksize);
1811 goto failed_mount;
1812 }
1813 sbi->s_frags_per_block = 1;
1814 sbi->s_blocks_per_group = le32_to_cpu(es->s_blocks_per_group);
1815 sbi->s_frags_per_group = le32_to_cpu(es->s_frags_per_group);
1816 sbi->s_inodes_per_group = le32_to_cpu(es->s_inodes_per_group);
b47b6f38 1817 if (EXT3_INODE_SIZE(sb) == 0 || EXT3_INODES_PER_GROUP(sb) == 0)
1da177e4
LT		 1818 goto cantfind_ext3;
1819 sbi->s_inodes_per_block = blocksize / EXT3_INODE_SIZE(sb);
1820 if (sbi->s_inodes_per_block == 0)
1821 goto cantfind_ext3;
1822 sbi->s_itb_per_group = sbi->s_inodes_per_group /
1823 sbi->s_inodes_per_block;
1824 sbi->s_desc_per_block = blocksize / sizeof(struct ext3_group_desc);
1825 sbi->s_sbh = bh;
1826 sbi->s_mount_state = le16_to_cpu(es->s_state);
f0d1b0b3
DH		 1827 sbi->s_addr_per_block_bits = ilog2(EXT3_ADDR_PER_BLOCK(sb));
1828 sbi->s_desc_per_block_bits = ilog2(EXT3_DESC_PER_BLOCK(sb));
1da177e4
LT		 1829 for (i=0; i < 4; i++)
1830 sbi->s_hash_seed[i] = le32_to_cpu(es->s_hash_seed[i]);
1831 sbi->s_def_hash_version = es->s_def_hash_version;
5e1f8c9e
TT		 1832 i = le32_to_cpu(es->s_flags);
1833 if (i & EXT2_FLAGS_UNSIGNED_HASH)
1834 sbi->s_hash_unsigned = 3;
1835 else if ((i & EXT2_FLAGS_SIGNED_HASH) == 0) {
1836#ifdef __CHAR_UNSIGNED__
1837 es->s_flags |= cpu_to_le32(EXT2_FLAGS_UNSIGNED_HASH);
1838 sbi->s_hash_unsigned = 3;
1839#else
1840 es->s_flags |= cpu_to_le32(EXT2_FLAGS_SIGNED_HASH);
1841#endif
5e1f8c9e 1842 }
1da177e4
LT		 1843
1844 if (sbi->s_blocks_per_group > blocksize * 8) {
4cf46b67
AF		 1845 ext3_msg(sb, KERN_ERR,
1846 "#blocks per group too big: %lu",
1da177e4
LT		 1847 sbi->s_blocks_per_group);
1848 goto failed_mount;
1849 }
1850 if (sbi->s_frags_per_group > blocksize * 8) {
4cf46b67
AF		 1851 ext3_msg(sb, KERN_ERR,
1852 "error: #fragments per group too big: %lu",
1da177e4
LT		 1853 sbi->s_frags_per_group);
1854 goto failed_mount;
1855 }
1856 if (sbi->s_inodes_per_group > blocksize * 8) {
4cf46b67
AF		 1857 ext3_msg(sb, KERN_ERR,
1858 "error: #inodes per group too big: %lu",
1da177e4
LT		 1859 sbi->s_inodes_per_group);
1860 goto failed_mount;
1861 }
1862
fcd5df35
MC		 1863 if (le32_to_cpu(es->s_blocks_count) >
1864 (sector_t)(~0ULL) >> (sb->s_blocksize_bits - 9)) {
4cf46b67
AF		 1865 ext3_msg(sb, KERN_ERR,
1866 "error: filesystem is too large to mount safely");
fcd5df35 1867 if (sizeof(sector_t) < 8)
4cf46b67
AF		 1868 ext3_msg(sb, KERN_ERR,
1869 "error: CONFIG_LBDAF not enabled");
fcd5df35
MC		 1870 goto failed_mount;
1871 }
1872
1da177e4
LT		 1873 if (EXT3_BLOCKS_PER_GROUP(sb) == 0)
1874 goto cantfind_ext3;
855565e8
ES		 1875 sbi->s_groups_count = ((le32_to_cpu(es->s_blocks_count) -
1876 le32_to_cpu(es->s_first_data_block) - 1)
1877 / EXT3_BLOCKS_PER_GROUP(sb)) + 1;
1da177e4
LT		 1878 db_count = (sbi->s_groups_count + EXT3_DESC_PER_BLOCK(sb) - 1) /
1879 EXT3_DESC_PER_BLOCK(sb);
1880 sbi->s_group_desc = kmalloc(db_count * sizeof (struct buffer_head *),
1881 GFP_KERNEL);
1882 if (sbi->s_group_desc == NULL) {
4cf46b67
AF		 1883 ext3_msg(sb, KERN_ERR,
1884 "error: not enough memory");
1da177e4
LT		 1885 goto failed_mount;
1886 }
1887
5df096d6 1888 bgl_lock_init(sbi->s_blockgroup_lock);
1da177e4
LT		 1889
1890 for (i = 0; i < db_count; i++) {
1891 block = descriptor_loc(sb, logic_sb_block, i);
1892 sbi->s_group_desc[i] = sb_bread(sb, block);
1893 if (!sbi->s_group_desc[i]) {
4cf46b67
AF		 1894 ext3_msg(sb, KERN_ERR,
1895 "error: can't read group descriptor %d", i);
1da177e4
LT		 1896 db_count = i;
1897 goto failed_mount2;
1898 }
1899 }
1900 if (!ext3_check_descriptors (sb)) {
4cf46b67
AF		 1901 ext3_msg(sb, KERN_ERR,
1902 "error: group descriptors corrupted");
1da177e4
LT		 1903 goto failed_mount2;
1904 }
1905 sbi->s_gdb_count = db_count;
1906 get_random_bytes(&sbi->s_next_generation, sizeof(u32));
1907 spin_lock_init(&sbi->s_next_gen_lock);
0216bfcf 1908
1da177e4
LT		 1909 /* per fileystem reservation list head & lock */
1910 spin_lock_init(&sbi->s_rsv_window_lock);
1911 sbi->s_rsv_window_root = RB_ROOT;
1912 /* Add a single, static dummy reservation to the start of the
1913 * reservation window list --- it gives us a placeholder for
1914 * append-at-start-of-list which makes the allocation logic
1915 * _much_ simpler. */
1916 sbi->s_rsv_window_head.rsv_start = EXT3_RESERVE_WINDOW_NOT_ALLOCATED;
1917 sbi->s_rsv_window_head.rsv_end = EXT3_RESERVE_WINDOW_NOT_ALLOCATED;
1918 sbi->s_rsv_window_head.rsv_alloc_hit = 0;
1919 sbi->s_rsv_window_head.rsv_goal_size = 0;
1920 ext3_rsv_window_add(sb, &sbi->s_rsv_window_head);
1921
1922 /*
1923 * set up enough so that it can read an inode
1924 */
1925 sb->s_op = &ext3_sops;
1926 sb->s_export_op = &ext3_export_ops;
1927 sb->s_xattr = ext3_xattr_handlers;
1928#ifdef CONFIG_QUOTA
1929 sb->s_qcop = &ext3_qctl_operations;
1930 sb->dq_op = &ext3_quota_operations;
1931#endif
1932 INIT_LIST_HEAD(&sbi->s_orphan); /* unlinked but open files */
b8a052d0 1933 mutex_init(&sbi->s_orphan_lock);
96d2a495 1934 mutex_init(&sbi->s_resize_lock);
1da177e4
LT		 1935
1936 sb->s_root = NULL;
1937
1938 needs_recovery = (es->s_last_orphan != 0 ||
1939 EXT3_HAS_INCOMPAT_FEATURE(sb,
1940 EXT3_FEATURE_INCOMPAT_RECOVER));
1941
1942 /*
1943 * The first inode we look at is the journal inode. Don't try
1944 * root first: it may be modified in the journal!
1945 */
1946 if (!test_opt(sb, NOLOAD) &&
1947 EXT3_HAS_COMPAT_FEATURE(sb, EXT3_FEATURE_COMPAT_HAS_JOURNAL)) {
71b96257 1948 if (ext3_load_journal(sb, es, journal_devnum))
41d1a636 1949 goto failed_mount2;
1da177e4
LT		 1950 } else if (journal_inum) {
1951 if (ext3_create_journal(sb, es, journal_inum))
41d1a636 1952 goto failed_mount2;
1da177e4
LT		 1953 } else {
1954 if (!silent)
4cf46b67
AF		 1955 ext3_msg(sb, KERN_ERR,
1956 "error: no journal found. "
1957 "mounting ext3 over ext2?");
41d1a636
DM		 1958 goto failed_mount2;
1959 }
1960 err = percpu_counter_init(&sbi->s_freeblocks_counter,
1961 ext3_count_free_blocks(sb));
1962 if (!err) {
1963 err = percpu_counter_init(&sbi->s_freeinodes_counter,
1964 ext3_count_free_inodes(sb));
1965 }
1966 if (!err) {
1967 err = percpu_counter_init(&sbi->s_dirs_counter,
1968 ext3_count_dirs(sb));
1969 }
1970 if (err) {
1971 ext3_msg(sb, KERN_ERR, "error: insufficient memory");
0216bfcf 1972 goto failed_mount3;
1da177e4
LT		 1973 }
1974
1975 /* We have now updated the journal if required, so we can
1976 * validate the data journaling mode. */
1977 switch (test_opt(sb, DATA_FLAGS)) {
1978 case 0:
1979 /* No mode set, assume a default based on the journal
1980 capabilities: ORDERED_DATA if the journal can
1981 cope, else JOURNAL_DATA */
1982 if (journal_check_available_features
1983 (sbi->s_journal, 0, 0, JFS_FEATURE_INCOMPAT_REVOKE))
bbae8bcc 1984 set_opt(sbi->s_mount_opt, DEFAULT_DATA_MODE);
1da177e4
LT		 1985 else
1986 set_opt(sbi->s_mount_opt, JOURNAL_DATA);
1987 break;
1988
1989 case EXT3_MOUNT_ORDERED_DATA:
1990 case EXT3_MOUNT_WRITEBACK_DATA:
1991 if (!journal_check_available_features
1992 (sbi->s_journal, 0, 0, JFS_FEATURE_INCOMPAT_REVOKE)) {
4cf46b67
AF		 1993 ext3_msg(sb, KERN_ERR,
1994 "error: journal does not support "
1995 "requested data journaling mode");
41d1a636 1996 goto failed_mount3;
1da177e4
LT		 1997 }
1998 default:
1999 break;
2000 }
2001
2002 if (test_opt(sb, NOBH)) {
1da177e4 2003 if (!(test_opt(sb, DATA_FLAGS) == EXT3_MOUNT_WRITEBACK_DATA)) {
4cf46b67
AF		 2004 ext3_msg(sb, KERN_WARNING,
2005 "warning: ignoring nobh option - "
2006 "it is supported only with writeback mode");
1da177e4
LT		 2007 clear_opt(sbi->s_mount_opt, NOBH);
2008 }
2009 }
2010 /*
2011 * The journal_load will have done any necessary log recovery,
2012 * so we can safely mount the rest of the filesystem now.
2013 */
2014
473043dc
DH		 2015 root = ext3_iget(sb, EXT3_ROOT_INO);
2016 if (IS_ERR(root)) {
4cf46b67 2017 ext3_msg(sb, KERN_ERR, "error: get root inode failed");
473043dc 2018 ret = PTR_ERR(root);
41d1a636 2019 goto failed_mount3;
1da177e4
LT		 2020 }
2021 if (!S_ISDIR(root->i_mode) || !root->i_blocks || !root->i_size) {
473043dc 2022 iput(root);
4cf46b67 2023 ext3_msg(sb, KERN_ERR, "error: corrupt root inode, run e2fsck");
41d1a636 2024 goto failed_mount3;
1da177e4 2025 }
473043dc
DH		 2026 sb->s_root = d_alloc_root(root);
2027 if (!sb->s_root) {
4cf46b67 2028 ext3_msg(sb, KERN_ERR, "error: get root dentry failed");
473043dc
DH		 2029 iput(root);
2030 ret = -ENOMEM;
41d1a636 2031 goto failed_mount3;
473043dc 2032 }
1da177e4
LT		 2033
2034 ext3_setup_super (sb, es, sb->s_flags & MS_RDONLY);
ed505ee4 2035
1da177e4
LT		 2036 EXT3_SB(sb)->s_mount_state |= EXT3_ORPHAN_FS;
2037 ext3_orphan_cleanup(sb, es);
2038 EXT3_SB(sb)->s_mount_state &= ~EXT3_ORPHAN_FS;
2039 if (needs_recovery)
4cf46b67 2040 ext3_msg(sb, KERN_INFO, "recovery complete");
1da177e4 2041 ext3_mark_recovery_complete(sb, es);
4cf46b67 2042 ext3_msg(sb, KERN_INFO, "mounted filesystem with %s data mode",
1da177e4
LT		 2043 test_opt(sb,DATA_FLAGS) == EXT3_MOUNT_JOURNAL_DATA ? "journal":
2044 test_opt(sb,DATA_FLAGS) == EXT3_MOUNT_ORDERED_DATA ? "ordered":
2045 "writeback");
2046
1da177e4
LT		 2047 lock_kernel();
2048 return 0;
2049
2050cantfind_ext3:
2051 if (!silent)
4cf46b67
AF		 2052 ext3_msg(sb, KERN_INFO,
2053 "error: can't find ext3 filesystem on dev %s.",
1da177e4
LT		 2054 sb->s_id);
2055 goto failed_mount;
2056
0216bfcf
MC		 2057failed_mount3:
2058 percpu_counter_destroy(&sbi->s_freeblocks_counter);
2059 percpu_counter_destroy(&sbi->s_freeinodes_counter);
2060 percpu_counter_destroy(&sbi->s_dirs_counter);
41d1a636 2061 journal_destroy(sbi->s_journal);
1da177e4
LT		 2062failed_mount2:
2063 for (i = 0; i < db_count; i++)
2064 brelse(sbi->s_group_desc[i]);
2065 kfree(sbi->s_group_desc);
2066failed_mount:
2067#ifdef CONFIG_QUOTA
2068 for (i = 0; i < MAXQUOTAS; i++)
2069 kfree(sbi->s_qf_names[i]);
2070#endif
2071 ext3_blkdev_remove(sbi);
2072 brelse(bh);
2073out_fail:
2074 sb->s_fs_info = NULL;
de5ce037 2075 kfree(sbi->s_blockgroup_lock);
1da177e4
LT		 2076 kfree(sbi);
2077 lock_kernel();
473043dc 2078 return ret;
1da177e4
LT		 2079}
2080
2081/*
2082 * Setup any per-fs journal parameters now. We'll do this both on
2083 * initial mount, once the journal has been initialised but before we've
ae6ddcc5 2084 * done any recovery; and again on any subsequent remount.
1da177e4
LT		 2085 */
2086static void ext3_init_journal_params(struct super_block *sb, journal_t *journal)
2087{
2088 struct ext3_sb_info *sbi = EXT3_SB(sb);
2089
2090 if (sbi->s_commit_interval)
2091 journal->j_commit_interval = sbi->s_commit_interval;
2092 /* We could also set up an ext3-specific default for the commit
2093 * interval here, but for now we'll just fall back to the jbd
2094 * default. */
2095
2096 spin_lock(&journal->j_state_lock);
2097 if (test_opt(sb, BARRIER))
2098 journal->j_flags |= JFS_BARRIER;
2099 else
2100 journal->j_flags &= ~JFS_BARRIER;
0e4fb5e2
HK		 2101 if (test_opt(sb, DATA_ERR_ABORT))
2102 journal->j_flags |= JFS_ABORT_ON_SYNCDATA_ERR;
2103 else
2104 journal->j_flags &= ~JFS_ABORT_ON_SYNCDATA_ERR;
1da177e4
LT		 2105 spin_unlock(&journal->j_state_lock);
2106}
2107
eee194e7
ES		 2108static journal_t *ext3_get_journal(struct super_block *sb,
2109 unsigned int journal_inum)
1da177e4
LT		 2110{
2111 struct inode *journal_inode;
2112 journal_t *journal;
2113
2114 /* First, test for the existence of a valid inode on disk. Bad
2115 * things happen if we iget() an unused inode, as the subsequent
2116 * iput() will try to delete it. */
2117
473043dc
DH		 2118 journal_inode = ext3_iget(sb, journal_inum);
2119 if (IS_ERR(journal_inode)) {
4cf46b67 2120 ext3_msg(sb, KERN_ERR, "error: no journal found");
1da177e4
LT		 2121 return NULL;
2122 }
2123 if (!journal_inode->i_nlink) {
2124 make_bad_inode(journal_inode);
2125 iput(journal_inode);
4cf46b67 2126 ext3_msg(sb, KERN_ERR, "error: journal inode is deleted");
1da177e4
LT		 2127 return NULL;
2128 }
2129
2130 jbd_debug(2, "Journal inode found at %p: %Ld bytes\n",
2131 journal_inode, journal_inode->i_size);
473043dc 2132 if (!S_ISREG(journal_inode->i_mode)) {
4cf46b67 2133 ext3_msg(sb, KERN_ERR, "error: invalid journal inode");
1da177e4
LT		 2134 iput(journal_inode);
2135 return NULL;
2136 }
2137
2138 journal = journal_init_inode(journal_inode);
2139 if (!journal) {
4cf46b67 2140 ext3_msg(sb, KERN_ERR, "error: could not load journal inode");
1da177e4
LT		 2141 iput(journal_inode);
2142 return NULL;
2143 }
2144 journal->j_private = sb;
2145 ext3_init_journal_params(sb, journal);
2146 return journal;
2147}
2148
2149static journal_t *ext3_get_dev_journal(struct super_block *sb,
2150 dev_t j_dev)
2151{
2152 struct buffer_head * bh;
2153 journal_t *journal;
43d23f90 2154 ext3_fsblk_t start;
1c2bf374 2155 ext3_fsblk_t len;
1da177e4 2156 int hblock, blocksize;
43d23f90 2157 ext3_fsblk_t sb_block;
1da177e4
LT		 2158 unsigned long offset;
2159 struct ext3_super_block * es;
2160 struct block_device *bdev;
2161
4cf46b67 2162 bdev = ext3_blkdev_get(j_dev, sb);
1da177e4
LT		 2163 if (bdev == NULL)
2164 return NULL;
2165
2166 if (bd_claim(bdev, sb)) {
4cf46b67
AF		 2167 ext3_msg(sb, KERN_ERR,
2168 "error: failed to claim external journal device");
9a1c3542 2169 blkdev_put(bdev, FMODE_READ|FMODE_WRITE);
1da177e4
LT		 2170 return NULL;
2171 }
2172
2173 blocksize = sb->s_blocksize;
e1defc4f 2174 hblock = bdev_logical_block_size(bdev);
1da177e4 2175 if (blocksize < hblock) {
4cf46b67
AF		 2176 ext3_msg(sb, KERN_ERR,
2177 "error: blocksize too small for journal device");
1da177e4
LT		 2178 goto out_bdev;
2179 }
2180
2181 sb_block = EXT3_MIN_BLOCK_SIZE / blocksize;
2182 offset = EXT3_MIN_BLOCK_SIZE % blocksize;
2183 set_blocksize(bdev, blocksize);
2184 if (!(bh = __bread(bdev, sb_block, blocksize))) {
4cf46b67
AF		 2185 ext3_msg(sb, KERN_ERR, "error: couldn't read superblock of "
2186 "external journal");
1da177e4
LT		 2187 goto out_bdev;
2188 }
2189
2190 es = (struct ext3_super_block *) (((char *)bh->b_data) + offset);
2191 if ((le16_to_cpu(es->s_magic) != EXT3_SUPER_MAGIC) ||
2192 !(le32_to_cpu(es->s_feature_incompat) &
2193 EXT3_FEATURE_INCOMPAT_JOURNAL_DEV)) {
4cf46b67
AF		 2194 ext3_msg(sb, KERN_ERR, "error: external journal has "
2195 "bad superblock");
1da177e4
LT		 2196 brelse(bh);
2197 goto out_bdev;
2198 }
2199
2200 if (memcmp(EXT3_SB(sb)->s_es->s_journal_uuid, es->s_uuid, 16)) {
4cf46b67 2201 ext3_msg(sb, KERN_ERR, "error: journal UUID does not match");
1da177e4
LT		 2202 brelse(bh);
2203 goto out_bdev;
2204 }
2205
2206 len = le32_to_cpu(es->s_blocks_count);
2207 start = sb_block + 1;
2208 brelse(bh); /* we're done with the superblock */
2209
2210 journal = journal_init_dev(bdev, sb->s_bdev,
2211 start, len, blocksize);
2212 if (!journal) {
4cf46b67
AF		 2213 ext3_msg(sb, KERN_ERR,
2214 "error: failed to create device journal");
1da177e4
LT		 2215 goto out_bdev;
2216 }
2217 journal->j_private = sb;
2218 ll_rw_block(READ, 1, &journal->j_sb_buffer);
2219 wait_on_buffer(journal->j_sb_buffer);
2220 if (!buffer_uptodate(journal->j_sb_buffer)) {
4cf46b67 2221 ext3_msg(sb, KERN_ERR, "I/O error on journal device");
1da177e4
LT		 2222 goto out_journal;
2223 }
2224 if (be32_to_cpu(journal->j_superblock->s_nr_users) != 1) {
4cf46b67
AF		 2225 ext3_msg(sb, KERN_ERR,
2226 "error: external journal has more than one "
2227 "user (unsupported) - %d",
1da177e4
LT		 2228 be32_to_cpu(journal->j_superblock->s_nr_users));
2229 goto out_journal;
2230 }
2231 EXT3_SB(sb)->journal_bdev = bdev;
2232 ext3_init_journal_params(sb, journal);
2233 return journal;
2234out_journal:
2235 journal_destroy(journal);
2236out_bdev:
2237 ext3_blkdev_put(bdev);
2238 return NULL;
2239}
2240
71b96257
JL		 2241static int ext3_load_journal(struct super_block *sb,
2242 struct ext3_super_block *es,
2243 unsigned long journal_devnum)
1da177e4
LT		 2244{
2245 journal_t *journal;
eee194e7 2246 unsigned int journal_inum = le32_to_cpu(es->s_journal_inum);
71b96257 2247 dev_t journal_dev;
1da177e4
LT		 2248 int err = 0;
2249 int really_read_only;
2250
71b96257
JL		 2251 if (journal_devnum &&
2252 journal_devnum != le32_to_cpu(es->s_journal_dev)) {
4cf46b67
AF		 2253 ext3_msg(sb, KERN_INFO, "external journal device major/minor "
2254 "numbers have changed");
71b96257
JL		 2255 journal_dev = new_decode_dev(journal_devnum);
2256 } else
2257 journal_dev = new_decode_dev(le32_to_cpu(es->s_journal_dev));
2258
1da177e4
LT		 2259 really_read_only = bdev_read_only(sb->s_bdev);
2260
2261 /*
2262 * Are we loading a blank journal or performing recovery after a
2263 * crash? For recovery, we need to check in advance whether we
2264 * can get read-write access to the device.
2265 */
2266
2267 if (EXT3_HAS_INCOMPAT_FEATURE(sb, EXT3_FEATURE_INCOMPAT_RECOVER)) {
2268 if (sb->s_flags & MS_RDONLY) {
4cf46b67
AF		 2269 ext3_msg(sb, KERN_INFO,
2270 "recovery required on readonly filesystem");
1da177e4 2271 if (really_read_only) {
4cf46b67
AF		 2272 ext3_msg(sb, KERN_ERR, "error: write access "
2273 "unavailable, cannot proceed");
1da177e4
LT		 2274 return -EROFS;
2275 }
4cf46b67
AF		 2276 ext3_msg(sb, KERN_INFO,
2277 "write access will be enabled during recovery");
1da177e4
LT		 2278 }
2279 }
2280
2281 if (journal_inum && journal_dev) {
4cf46b67
AF		 2282 ext3_msg(sb, KERN_ERR, "error: filesystem has both journal "
2283 "and inode journals");
1da177e4
LT		 2284 return -EINVAL;
2285 }
2286
2287 if (journal_inum) {
2288 if (!(journal = ext3_get_journal(sb, journal_inum)))
2289 return -EINVAL;
2290 } else {
2291 if (!(journal = ext3_get_dev_journal(sb, journal_dev)))
2292 return -EINVAL;
2293 }
2294
0636c73e
ES		 2295 if (!(journal->j_flags & JFS_BARRIER))
2296 printk(KERN_INFO "EXT3-fs: barriers not enabled\n");
2297
1da177e4
LT		 2298 if (!really_read_only && test_opt(sb, UPDATE_JOURNAL)) {
2299 err = journal_update_format(journal);
2300 if (err) {
4cf46b67 2301 ext3_msg(sb, KERN_ERR, "error updating journal");
1da177e4
LT		 2302 journal_destroy(journal);
2303 return err;
2304 }
2305 }
2306
2307 if (!EXT3_HAS_INCOMPAT_FEATURE(sb, EXT3_FEATURE_INCOMPAT_RECOVER))
2308 err = journal_wipe(journal, !really_read_only);
2309 if (!err)
2310 err = journal_load(journal);
2311
2312 if (err) {
4cf46b67 2313 ext3_msg(sb, KERN_ERR, "error loading journal");
1da177e4
LT		 2314 journal_destroy(journal);
2315 return err;
2316 }
2317
2318 EXT3_SB(sb)->s_journal = journal;
2319 ext3_clear_journal_err(sb, es);
71b96257
JL		 2320
2321 if (journal_devnum &&
2322 journal_devnum != le32_to_cpu(es->s_journal_dev)) {
2323 es->s_journal_dev = cpu_to_le32(journal_devnum);
71b96257
JL		 2324
2325 /* Make sure we flush the recovery flag to disk. */
2326 ext3_commit_super(sb, es, 1);
2327 }
2328
1da177e4
LT		 2329 return 0;
2330}
2331
4cf46b67
AF		 2332static int ext3_create_journal(struct super_block *sb,
2333 struct ext3_super_block *es,
eee194e7 2334 unsigned int journal_inum)
1da177e4
LT		 2335{
2336 journal_t *journal;
952d9de1 2337 int err;
1da177e4
LT		 2338
2339 if (sb->s_flags & MS_RDONLY) {
4cf46b67
AF		 2340 ext3_msg(sb, KERN_ERR,
2341 "error: readonly filesystem when trying to "
2342 "create journal");
1da177e4
LT		 2343 return -EROFS;
2344 }
2345
952d9de1
BP		 2346 journal = ext3_get_journal(sb, journal_inum);
2347 if (!journal)
1da177e4
LT		 2348 return -EINVAL;
2349
4cf46b67 2350 ext3_msg(sb, KERN_INFO, "creating new journal on inode %u",
1da177e4
LT		 2351 journal_inum);
2352
952d9de1
BP		 2353 err = journal_create(journal);
2354 if (err) {
4cf46b67 2355 ext3_msg(sb, KERN_ERR, "error creating journal");
1da177e4
LT		 2356 journal_destroy(journal);
2357 return -EIO;
2358 }
2359
2360 EXT3_SB(sb)->s_journal = journal;
2361
2362 ext3_update_dynamic_rev(sb);
2363 EXT3_SET_INCOMPAT_FEATURE(sb, EXT3_FEATURE_INCOMPAT_RECOVER);
2364 EXT3_SET_COMPAT_FEATURE(sb, EXT3_FEATURE_COMPAT_HAS_JOURNAL);
2365
2366 es->s_journal_inum = cpu_to_le32(journal_inum);
1da177e4
LT		 2367
2368 /* Make sure we flush the recovery flag to disk. */
2369 ext3_commit_super(sb, es, 1);
2370
2371 return 0;
2372}
2373
c4be0c1d
TS		 2374static int ext3_commit_super(struct super_block *sb,
2375 struct ext3_super_block *es,
1da177e4
LT		 2376 int sync)
2377{
2378 struct buffer_head *sbh = EXT3_SB(sb)->s_sbh;
c4be0c1d 2379 int error = 0;
1da177e4
LT		 2380
2381 if (!sbh)
c4be0c1d 2382 return error;
96ec2e0a
TT		 2383 /*
2384 * If the file system is mounted read-only, don't update the
2385 * superblock write time. This avoids updating the superblock
2386 * write time when we are mounting the root file system
2387 * read/only but we need to replay the journal; at that point,
2388 * for people who are east of GMT and who make their clock
2389 * tick in localtime for Windows bug-for-bug compatibility,
2390 * the clock is set in the future, and this will cause e2fsck
2391 * to complain and force a full file system check.
2392 */
2393 if (!(sb->s_flags & MS_RDONLY))
2394 es->s_wtime = cpu_to_le32(get_seconds());
1da177e4
LT		 2395 es->s_free_blocks_count = cpu_to_le32(ext3_count_free_blocks(sb));
2396 es->s_free_inodes_count = cpu_to_le32(ext3_count_free_inodes(sb));
2397 BUFFER_TRACE(sbh, "marking dirty");
2398 mark_buffer_dirty(sbh);
2399 if (sync)
c4be0c1d
TS		 2400 error = sync_dirty_buffer(sbh);
2401 return error;
1da177e4
LT		 2402}
2403
2404
2405/*
2406 * Have we just finished recovery? If so, and if we are mounting (or
2407 * remounting) the filesystem readonly, then we will end up with a
2408 * consistent fs on disk. Record that fact.
2409 */
2410static void ext3_mark_recovery_complete(struct super_block * sb,
2411 struct ext3_super_block * es)
2412{
2413 journal_t *journal = EXT3_SB(sb)->s_journal;
2414
2415 journal_lock_updates(journal);
2d7c820e
HK		 2416 if (journal_flush(journal) < 0)
2417 goto out;
2418
1da177e4
LT		 2419 if (EXT3_HAS_INCOMPAT_FEATURE(sb, EXT3_FEATURE_INCOMPAT_RECOVER) &&
2420 sb->s_flags & MS_RDONLY) {
2421 EXT3_CLEAR_INCOMPAT_FEATURE(sb, EXT3_FEATURE_INCOMPAT_RECOVER);
1da177e4
LT		 2422 ext3_commit_super(sb, es, 1);
2423 }
2d7c820e
HK		 2424
2425out:
1da177e4
LT		 2426 journal_unlock_updates(journal);
2427}
2428
2429/*
2430 * If we are mounting (or read-write remounting) a filesystem whose journal
2431 * has recorded an error from a previous lifetime, move that error to the
2432 * main filesystem now.
2433 */
4cf46b67
AF		 2434static void ext3_clear_journal_err(struct super_block *sb,
2435 struct ext3_super_block *es)
1da177e4
LT		 2436{
2437 journal_t *journal;
2438 int j_errno;
2439 const char *errstr;
2440
2441 journal = EXT3_SB(sb)->s_journal;
2442
2443 /*
2444 * Now check for any error status which may have been recorded in the
2445 * journal by a prior ext3_error() or ext3_abort()
2446 */
2447
2448 j_errno = journal_errno(journal);
2449 if (j_errno) {
2450 char nbuf[16];
2451
2452 errstr = ext3_decode_error(sb, j_errno, nbuf);
e05b6b52 2453 ext3_warning(sb, __func__, "Filesystem error recorded "
1da177e4 2454 "from previous mount: %s", errstr);
e05b6b52 2455 ext3_warning(sb, __func__, "Marking fs in need of "
1da177e4
LT		 2456 "filesystem check.");
2457
2458 EXT3_SB(sb)->s_mount_state |= EXT3_ERROR_FS;
2459 es->s_state |= cpu_to_le16(EXT3_ERROR_FS);
2460 ext3_commit_super (sb, es, 1);
2461
2462 journal_clear_err(journal);
2463 }
2464}
2465
2466/*
2467 * Force the running and committing transactions to commit,
2468 * and wait on the commit.
2469 */
2470int ext3_force_commit(struct super_block *sb)
2471{
2472 journal_t *journal;
2473 int ret;
2474
2475 if (sb->s_flags & MS_RDONLY)
2476 return 0;
2477
2478 journal = EXT3_SB(sb)->s_journal;
1da177e4
LT		 2479 ret = ext3_journal_force_commit(journal);
2480 return ret;
2481}
2482
1da177e4
LT		 2483static int ext3_sync_fs(struct super_block *sb, int wait)
2484{
02ac597c 2485 tid_t target;
c87591b7 2486
02ac597c
JK		 2487 if (journal_start_commit(EXT3_SB(sb)->s_journal, &target)) {
2488 if (wait)
2489 log_wait_commit(EXT3_SB(sb)->s_journal, target);
2490 }
1da177e4
LT		 2491 return 0;
2492}
2493
2494/*
2495 * LVM calls this function before a (read-only) snapshot is created. This
2496 * gives us a chance to flush the journal completely and mark the fs clean.
2497 */
c4be0c1d 2498static int ext3_freeze(struct super_block *sb)
1da177e4 2499{
c4be0c1d
TS		 2500 int error = 0;
2501 journal_t *journal;
1da177e4
LT		 2502
2503 if (!(sb->s_flags & MS_RDONLY)) {
c4be0c1d 2504 journal = EXT3_SB(sb)->s_journal;
1da177e4
LT		 2505
2506 /* Now we set up the journal barrier. */
2507 journal_lock_updates(journal);
2d7c820e
HK		 2508
2509 /*
2510 * We don't want to clear needs_recovery flag when we failed
2511 * to flush the journal.
2512 */
c4be0c1d
TS		 2513 error = journal_flush(journal);
2514 if (error < 0)
2515 goto out;
1da177e4
LT		 2516
2517 /* Journal blocked and flushed, clear needs_recovery flag. */
2518 EXT3_CLEAR_INCOMPAT_FEATURE(sb, EXT3_FEATURE_INCOMPAT_RECOVER);
c4be0c1d
TS		 2519 error = ext3_commit_super(sb, EXT3_SB(sb)->s_es, 1);
2520 if (error)
2521 goto out;
1da177e4 2522 }
c4be0c1d
TS		 2523 return 0;
2524
2525out:
2526 journal_unlock_updates(journal);
2527 return error;
1da177e4
LT		 2528}
2529
2530/*
2531 * Called by LVM after the snapshot is done. We need to reset the RECOVER
2532 * flag here, even though the filesystem is not technically dirty yet.
2533 */
c4be0c1d 2534static int ext3_unfreeze(struct super_block *sb)
1da177e4
LT		 2535{
2536 if (!(sb->s_flags & MS_RDONLY)) {
2537 lock_super(sb);
2538 /* Reser the needs_recovery flag before the fs is unlocked. */
2539 EXT3_SET_INCOMPAT_FEATURE(sb, EXT3_FEATURE_INCOMPAT_RECOVER);
2540 ext3_commit_super(sb, EXT3_SB(sb)->s_es, 1);
2541 unlock_super(sb);
2542 journal_unlock_updates(EXT3_SB(sb)->s_journal);
2543 }
c4be0c1d 2544 return 0;
1da177e4
LT		 2545}
2546
2547static int ext3_remount (struct super_block * sb, int * flags, char * data)
2548{
2549 struct ext3_super_block * es;
2550 struct ext3_sb_info *sbi = EXT3_SB(sb);
43d23f90 2551 ext3_fsblk_t n_blocks_count = 0;
08c6a96f
JK		 2552 unsigned long old_sb_flags;
2553 struct ext3_mount_options old_opts;
c79d967d 2554 int enable_quota = 0;
08c6a96f
JK		 2555 int err;
2556#ifdef CONFIG_QUOTA
2557 int i;
2558#endif
2559
337eb00a
AIB		 2560 lock_kernel();
2561
08c6a96f 2562 /* Store the original options */
bbd6851a 2563 lock_super(sb);
08c6a96f
JK		 2564 old_sb_flags = sb->s_flags;
2565 old_opts.s_mount_opt = sbi->s_mount_opt;
2566 old_opts.s_resuid = sbi->s_resuid;
2567 old_opts.s_resgid = sbi->s_resgid;
2568 old_opts.s_commit_interval = sbi->s_commit_interval;
2569#ifdef CONFIG_QUOTA
2570 old_opts.s_jquota_fmt = sbi->s_jquota_fmt;
2571 for (i = 0; i < MAXQUOTAS; i++)
2572 old_opts.s_qf_names[i] = sbi->s_qf_names[i];
2573#endif
1da177e4
LT		 2574
2575 /*
2576 * Allow the "check" option to be passed as a remount option.
2577 */
71b96257 2578 if (!parse_options(data, sb, NULL, NULL, &n_blocks_count, 1)) {
08c6a96f
JK		 2579 err = -EINVAL;
2580 goto restore_opts;
2581 }
1da177e4 2582
e3c96435 2583 if (test_opt(sb, ABORT))
e05b6b52 2584 ext3_abort(sb, __func__, "Abort forced by user");
1da177e4
LT		 2585
2586 sb->s_flags = (sb->s_flags & ~MS_POSIXACL) |
e3c96435 2587 (test_opt(sb, POSIX_ACL) ? MS_POSIXACL : 0);
1da177e4
LT		 2588
2589 es = sbi->s_es;
2590
2591 ext3_init_journal_params(sb, sbi->s_journal);
2592
2593 if ((*flags & MS_RDONLY) != (sb->s_flags & MS_RDONLY) ||
2594 n_blocks_count > le32_to_cpu(es->s_blocks_count)) {
e3c96435 2595 if (test_opt(sb, ABORT)) {
08c6a96f
JK		 2596 err = -EROFS;
2597 goto restore_opts;
2598 }
1da177e4
LT		 2599
2600 if (*flags & MS_RDONLY) {
0f0dd62f
CH		 2601 err = dquot_suspend(sb, -1);
2602 if (err < 0)
c79d967d 2603 goto restore_opts;
c79d967d 2604
1da177e4
LT		 2605 /*
2606 * First of all, the unconditional stuff we have to do
2607 * to disable replay of the journal when we next remount
2608 */
2609 sb->s_flags |= MS_RDONLY;
2610
2611 /*
2612 * OK, test if we are remounting a valid rw partition
2613 * readonly, and if so set the rdonly flag and then
2614 * mark the partition as valid again.
2615 */
2616 if (!(es->s_state & cpu_to_le16(EXT3_VALID_FS)) &&
2617 (sbi->s_mount_state & EXT3_VALID_FS))
2618 es->s_state = cpu_to_le16(sbi->s_mount_state);
2619
2620 ext3_mark_recovery_complete(sb, es);
2621 } else {
2622 __le32 ret;
2623 if ((ret = EXT3_HAS_RO_COMPAT_FEATURE(sb,
2624 ~EXT3_FEATURE_RO_COMPAT_SUPP))) {
4cf46b67
AF		 2625 ext3_msg(sb, KERN_WARNING,
2626 "warning: couldn't remount RDWR "
2627 "because of unsupported optional "
2628 "features (%x)", le32_to_cpu(ret));
08c6a96f
JK		 2629 err = -EROFS;
2630 goto restore_opts;
1da177e4 2631 }
ea9a05a1
ES		 2632
2633 /*
2634 * If we have an unprocessed orphan list hanging
2635 * around from a previously readonly bdev mount,
2636 * require a full umount/remount for now.
2637 */
2638 if (es->s_last_orphan) {
4cf46b67 2639 ext3_msg(sb, KERN_WARNING, "warning: couldn't "
ea9a05a1
ES		 2640 "remount RDWR because of unprocessed "
2641 "orphan inode list. Please "
4cf46b67 2642 "umount/remount instead.");
ea9a05a1
ES		 2643 err = -EINVAL;
2644 goto restore_opts;
2645 }
2646
1da177e4
LT		 2647 /*
2648 * Mounting a RDONLY partition read-write, so reread
2649 * and store the current valid flag. (It may have
2650 * been changed by e2fsck since we originally mounted
2651 * the partition.)
2652 */
2653 ext3_clear_journal_err(sb, es);
2654 sbi->s_mount_state = le16_to_cpu(es->s_state);
a4e4de36 2655 if ((err = ext3_group_extend(sb, es, n_blocks_count)))
08c6a96f 2656 goto restore_opts;
1da177e4
LT		 2657 if (!ext3_setup_super (sb, es, 0))
2658 sb->s_flags &= ~MS_RDONLY;
c79d967d 2659 enable_quota = 1;
1da177e4
LT		 2660 }
2661 }
08c6a96f
JK		 2662#ifdef CONFIG_QUOTA
2663 /* Release old quota file names */
2664 for (i = 0; i < MAXQUOTAS; i++)
2665 if (old_opts.s_qf_names[i] &&
2666 old_opts.s_qf_names[i] != sbi->s_qf_names[i])
2667 kfree(old_opts.s_qf_names[i]);
2668#endif
bbd6851a 2669 unlock_super(sb);
337eb00a 2670 unlock_kernel();
c79d967d
CH		 2671
2672 if (enable_quota)
0f0dd62f 2673 dquot_resume(sb, -1);
1da177e4 2674 return 0;
08c6a96f
JK		 2675restore_opts:
2676 sb->s_flags = old_sb_flags;
2677 sbi->s_mount_opt = old_opts.s_mount_opt;
2678 sbi->s_resuid = old_opts.s_resuid;
2679 sbi->s_resgid = old_opts.s_resgid;
2680 sbi->s_commit_interval = old_opts.s_commit_interval;
2681#ifdef CONFIG_QUOTA
2682 sbi->s_jquota_fmt = old_opts.s_jquota_fmt;
2683 for (i = 0; i < MAXQUOTAS; i++) {
2684 if (sbi->s_qf_names[i] &&
2685 old_opts.s_qf_names[i] != sbi->s_qf_names[i])
2686 kfree(sbi->s_qf_names[i]);
2687 sbi->s_qf_names[i] = old_opts.s_qf_names[i];
2688 }
2689#endif
bbd6851a 2690 unlock_super(sb);
337eb00a 2691 unlock_kernel();
08c6a96f 2692 return err;
1da177e4
LT		 2693}
2694
726c3342 2695static int ext3_statfs (struct dentry * dentry, struct kstatfs * buf)
1da177e4 2696{
726c3342 2697 struct super_block *sb = dentry->d_sb;
09fe316a
AT		 2698 struct ext3_sb_info *sbi = EXT3_SB(sb);
2699 struct ext3_super_block *es = sbi->s_es;
50ee0a32 2700 u64 fsid;
1da177e4 2701
a71ce8c6
BP		 2702 if (test_opt(sb, MINIX_DF)) {
2703 sbi->s_overhead_last = 0;
2704 } else if (sbi->s_blocks_last != le32_to_cpu(es->s_blocks_count)) {
2705 unsigned long ngroups = sbi->s_groups_count, i;
2706 ext3_fsblk_t overhead = 0;
1da177e4
LT		 2707 smp_rmb();
2708
2709 /*
a71ce8c6
BP		 2710 * Compute the overhead (FS structures). This is constant
2711 * for a given filesystem unless the number of block groups
2712 * changes so we cache the previous value until it does.
1da177e4
LT		 2713 */
2714
2715 /*
2716 * All of the blocks before first_data_block are
2717 * overhead
2718 */
2719 overhead = le32_to_cpu(es->s_first_data_block);
2720
2721 /*
2722 * Add the overhead attributed to the superblock and
2723 * block group descriptors. If the sparse superblocks
2724 * feature is turned on, then not all groups have this.
2725 */
2726 for (i = 0; i < ngroups; i++) {
2727 overhead += ext3_bg_has_super(sb, i) +
2728 ext3_bg_num_gdb(sb, i);
2729 cond_resched();
2730 }
2731
2732 /*
2733 * Every block group has an inode bitmap, a block
2734 * bitmap, and an inode table.
2735 */
a71ce8c6
BP		 2736 overhead += ngroups * (2 + sbi->s_itb_per_group);
2737 sbi->s_overhead_last = overhead;
2738 smp_wmb();
2739 sbi->s_blocks_last = le32_to_cpu(es->s_blocks_count);
1da177e4
LT		 2740 }
2741
2742 buf->f_type = EXT3_SUPER_MAGIC;
2743 buf->f_bsize = sb->s_blocksize;
a71ce8c6 2744 buf->f_blocks = le32_to_cpu(es->s_blocks_count) - sbi->s_overhead_last;
52d9f3b4 2745 buf->f_bfree = percpu_counter_sum_positive(&sbi->s_freeblocks_counter);
1da177e4
LT		 2746 buf->f_bavail = buf->f_bfree - le32_to_cpu(es->s_r_blocks_count);
2747 if (buf->f_bfree < le32_to_cpu(es->s_r_blocks_count))
2748 buf->f_bavail = 0;
2749 buf->f_files = le32_to_cpu(es->s_inodes_count);
52d9f3b4 2750 buf->f_ffree = percpu_counter_sum_positive(&sbi->s_freeinodes_counter);
1da177e4 2751 buf->f_namelen = EXT3_NAME_LEN;
50ee0a32
PE		 2752 fsid = le64_to_cpup((void *)es->s_uuid) ^
2753 le64_to_cpup((void *)es->s_uuid + sizeof(u64));
2754 buf->f_fsid.val[0] = fsid & 0xFFFFFFFFUL;
2755 buf->f_fsid.val[1] = (fsid >> 32) & 0xFFFFFFFFUL;
1da177e4
LT		 2756 return 0;
2757}
2758
2759/* Helper function for writing quotas on sync - we need to start transaction before quota file
2760 * is locked for write. Otherwise the are possible deadlocks:
2761 * Process 1 Process 2
2762 * ext3_create() quota_sync()
2763 * journal_start() write_dquot()
871a2931 2764 * dquot_initialize() down(dqio_mutex)
d3be915f 2765 * down(dqio_mutex) journal_start()
1da177e4
LT		 2766 *
2767 */
2768
2769#ifdef CONFIG_QUOTA
2770
2771static inline struct inode *dquot_to_inode(struct dquot *dquot)
2772{
2773 return sb_dqopt(dquot->dq_sb)->files[dquot->dq_type];
2774}
2775
1da177e4
LT		 2776static int ext3_write_dquot(struct dquot *dquot)
2777{
2778 int ret, err;
2779 handle_t *handle;
2780 struct inode *inode;
2781
2782 inode = dquot_to_inode(dquot);
2783 handle = ext3_journal_start(inode,
1f54587b 2784 EXT3_QUOTA_TRANS_BLOCKS(dquot->dq_sb));
1da177e4
LT		 2785 if (IS_ERR(handle))
2786 return PTR_ERR(handle);
2787 ret = dquot_commit(dquot);
2788 err = ext3_journal_stop(handle);
2789 if (!ret)
2790 ret = err;
2791 return ret;
2792}
2793
2794static int ext3_acquire_dquot(struct dquot *dquot)
2795{
2796 int ret, err;
2797 handle_t *handle;
2798
2799 handle = ext3_journal_start(dquot_to_inode(dquot),
1f54587b 2800 EXT3_QUOTA_INIT_BLOCKS(dquot->dq_sb));
1da177e4
LT		 2801 if (IS_ERR(handle))
2802 return PTR_ERR(handle);
2803 ret = dquot_acquire(dquot);
2804 err = ext3_journal_stop(handle);
2805 if (!ret)
2806 ret = err;
2807 return ret;
2808}
2809
2810static int ext3_release_dquot(struct dquot *dquot)
2811{
2812 int ret, err;
2813 handle_t *handle;
2814
2815 handle = ext3_journal_start(dquot_to_inode(dquot),
1f54587b 2816 EXT3_QUOTA_DEL_BLOCKS(dquot->dq_sb));
9c3013e9
JK		 2817 if (IS_ERR(handle)) {
2818 /* Release dquot anyway to avoid endless cycle in dqput() */
2819 dquot_release(dquot);
1da177e4 2820 return PTR_ERR(handle);
9c3013e9 2821 }
1da177e4
LT		 2822 ret = dquot_release(dquot);
2823 err = ext3_journal_stop(handle);
2824 if (!ret)
2825 ret = err;
2826 return ret;
2827}
2828
2829static int ext3_mark_dquot_dirty(struct dquot *dquot)
2830{
99aeaf63 2831 /* Are we journaling quotas? */
1da177e4
LT		 2832 if (EXT3_SB(dquot->dq_sb)->s_qf_names[USRQUOTA] ||
2833 EXT3_SB(dquot->dq_sb)->s_qf_names[GRPQUOTA]) {
2834 dquot_mark_dquot_dirty(dquot);
2835 return ext3_write_dquot(dquot);
2836 } else {
2837 return dquot_mark_dquot_dirty(dquot);
2838 }
2839}
2840
2841static int ext3_write_info(struct super_block *sb, int type)
2842{
2843 int ret, err;
2844 handle_t *handle;
2845
2846 /* Data block + inode block */
2847 handle = ext3_journal_start(sb->s_root->d_inode, 2);
2848 if (IS_ERR(handle))
2849 return PTR_ERR(handle);
2850 ret = dquot_commit_info(sb, type);
2851 err = ext3_journal_stop(handle);
2852 if (!ret)
2853 ret = err;
2854 return ret;
2855}
2856
2857/*
2858 * Turn on quotas during mount time - we need to find
2859 * the quota file and such...
2860 */
2861static int ext3_quota_on_mount(struct super_block *sb, int type)
2862{
84de856e
CH		 2863 return vfs_quota_on_mount(sb, EXT3_SB(sb)->s_qf_names[type],
2864 EXT3_SB(sb)->s_jquota_fmt, type);
1da177e4
LT		 2865}
2866
2867/*
2868 * Standard function to be called on quota_on
2869 */
2870static int ext3_quota_on(struct super_block *sb, int type, int format_id,
8264613d 2871 char *name, int remount)
1da177e4
LT		 2872{
2873 int err;
8264613d 2874 struct path path;
1da177e4 2875
1f54587b
JK		 2876 if (!test_opt(sb, QUOTA))
2877 return -EINVAL;
8264613d 2878 /* When remounting, no checks are needed and in fact, name is NULL */
9cfe7b90 2879 if (remount)
8264613d 2880 return vfs_quota_on(sb, type, format_id, name, remount);
9cfe7b90 2881
8264613d 2882 err = kern_path(name, LOOKUP_FOLLOW, &path);
1da177e4
LT		 2883 if (err)
2884 return err;
9cfe7b90 2885
1da177e4 2886 /* Quotafile not on the same filesystem? */
8264613d
AV		 2887 if (path.mnt->mnt_sb != sb) {
2888 path_put(&path);
1da177e4
LT		 2889 return -EXDEV;
2890 }
9cfe7b90
JK		 2891 /* Journaling quota? */
2892 if (EXT3_SB(sb)->s_qf_names[type]) {
2893 /* Quotafile not of fs root? */
8264613d 2894 if (path.dentry->d_parent != sb->s_root)
4cf46b67
AF		 2895 ext3_msg(sb, KERN_WARNING,
2896 "warning: Quota file not on filesystem root. "
2897 "Journaled quota will not work.");
9cfe7b90
JK		 2898 }
2899
2900 /*
2901 * When we journal data on quota file, we have to flush journal to see
2902 * all updates to the file when we bypass pagecache...
2903 */
8264613d 2904 if (ext3_should_journal_data(path.dentry->d_inode)) {
9cfe7b90
JK		 2905 /*
2906 * We don't need to lock updates but journal_flush() could
2907 * otherwise be livelocked...
2908 */
2909 journal_lock_updates(EXT3_SB(sb)->s_journal);
2d7c820e 2910 err = journal_flush(EXT3_SB(sb)->s_journal);
9cfe7b90 2911 journal_unlock_updates(EXT3_SB(sb)->s_journal);
2d7c820e 2912 if (err) {
12e1ec9f 2913 path_put(&path);
2d7c820e
HK		 2914 return err;
2915 }
9cfe7b90
JK		 2916 }
2917
8264613d
AV		 2918 err = vfs_quota_on_path(sb, type, format_id, &path);
2919 path_put(&path);
77e69dac 2920 return err;
1da177e4
LT		 2921}
2922
2923/* Read data from quotafile - avoid pagecache and such because we cannot afford
2924 * acquiring the locks... As quota files are never truncated and quota code
2925 * itself serializes the operations (and noone else should touch the files)
2926 * we don't have to be afraid of races */
2927static ssize_t ext3_quota_read(struct super_block *sb, int type, char *data,
2928 size_t len, loff_t off)
2929{
2930 struct inode *inode = sb_dqopt(sb)->files[type];
2931 sector_t blk = off >> EXT3_BLOCK_SIZE_BITS(sb);
2932 int err = 0;
2933 int offset = off & (sb->s_blocksize - 1);
2934 int tocopy;
2935 size_t toread;
2936 struct buffer_head *bh;
2937 loff_t i_size = i_size_read(inode);
2938
2939 if (off > i_size)
2940 return 0;
2941 if (off+len > i_size)
2942 len = i_size-off;
2943 toread = len;
2944 while (toread > 0) {
2945 tocopy = sb->s_blocksize - offset < toread ?
2946 sb->s_blocksize - offset : toread;
2947 bh = ext3_bread(NULL, inode, blk, 0, &err);
2948 if (err)
2949 return err;
2950 if (!bh) /* A hole? */
2951 memset(data, 0, tocopy);
2952 else
2953 memcpy(data, bh->b_data+offset, tocopy);
2954 brelse(bh);
2955 offset = 0;
2956 toread -= tocopy;
2957 data += tocopy;
2958 blk++;
2959 }
2960 return len;
2961}
2962
2963/* Write to quotafile (we know the transaction is already started and has
2964 * enough credits) */
2965static ssize_t ext3_quota_write(struct super_block *sb, int type,
2966 const char *data, size_t len, loff_t off)
2967{
2968 struct inode *inode = sb_dqopt(sb)->files[type];
2969 sector_t blk = off >> EXT3_BLOCK_SIZE_BITS(sb);
2970 int err = 0;
2971 int offset = off & (sb->s_blocksize - 1);
1da177e4 2972 int journal_quota = EXT3_SB(sb)->s_qf_names[type] != NULL;
1da177e4
LT		 2973 struct buffer_head *bh;
2974 handle_t *handle = journal_current_handle();
2975
9c3013e9 2976 if (!handle) {
4cf46b67
AF		 2977 ext3_msg(sb, KERN_WARNING,
2978 "warning: quota write (off=%llu, len=%llu)"
2979 " cancelled because transaction is not started.",
9c3013e9
JK		 2980 (unsigned long long)off, (unsigned long long)len);
2981 return -EIO;
2982 }
e5472147
DM		 2983
2984 /*
2985 * Since we account only one data block in transaction credits,
2986 * then it is impossible to cross a block boundary.
2987 */
2988 if (sb->s_blocksize - offset < len) {
2989 ext3_msg(sb, KERN_WARNING, "Quota write (off=%llu, len=%llu)"
2990 " cancelled because not block aligned",
2991 (unsigned long long)off, (unsigned long long)len);
2992 return -EIO;
2993 }
5c81a419 2994 mutex_lock_nested(&inode->i_mutex, I_MUTEX_QUOTA);
e5472147
DM		 2995 bh = ext3_bread(handle, inode, blk, 1, &err);
2996 if (!bh)
2997 goto out;
2998 if (journal_quota) {
2999 err = ext3_journal_get_write_access(handle, bh);
3000 if (err) {
3001 brelse(bh);
1da177e4 3002 goto out;
1da177e4 3003 }
1da177e4 3004 }
e5472147
DM		 3005 lock_buffer(bh);
3006 memcpy(bh->b_data+offset, data, len);
3007 flush_dcache_page(bh->b_page);
3008 unlock_buffer(bh);
3009 if (journal_quota)
3010 err = ext3_journal_dirty_metadata(handle, bh);
3011 else {
3012 /* Always do at least ordered writes for quotas */
3013 err = ext3_journal_dirty_data(handle, bh);
3014 mark_buffer_dirty(bh);
3015 }
3016 brelse(bh);
1da177e4 3017out:
e5472147 3018 if (err) {
f5c8f7da 3019 mutex_unlock(&inode->i_mutex);
1da177e4 3020 return err;
f5c8f7da 3021 }
e5472147
DM		 3022 if (inode->i_size < off + len) {
3023 i_size_write(inode, off + len);
1da177e4
LT		 3024 EXT3_I(inode)->i_disksize = inode->i_size;
3025 }
3026 inode->i_version++;
3027 inode->i_mtime = inode->i_ctime = CURRENT_TIME;
3028 ext3_mark_inode_dirty(handle, inode);
1b1dcc1b 3029 mutex_unlock(&inode->i_mutex);
e5472147 3030 return len;
1da177e4
LT		 3031}
3032
3033#endif
3034
454e2398
DH		 3035static int ext3_get_sb(struct file_system_type *fs_type,
3036 int flags, const char *dev_name, void *data, struct vfsmount *mnt)
1da177e4 3037{
454e2398 3038 return get_sb_bdev(fs_type, flags, dev_name, data, ext3_fill_super, mnt);
1da177e4
LT		 3039}
3040
3041static struct file_system_type ext3_fs_type = {
3042 .owner = THIS_MODULE,
3043 .name = "ext3",
3044 .get_sb = ext3_get_sb,
3045 .kill_sb = kill_block_super,
3046 .fs_flags = FS_REQUIRES_DEV,
3047};
3048
3049static int __init init_ext3_fs(void)
3050{
3051 int err = init_ext3_xattr();
3052 if (err)
3053 return err;
3054 err = init_inodecache();
3055 if (err)
3056 goto out1;
3057 err = register_filesystem(&ext3_fs_type);
3058 if (err)
3059 goto out;
3060 return 0;
3061out:
3062 destroy_inodecache();
3063out1:
e9ad5620 3064 exit_ext3_xattr();
1da177e4
LT		 3065 return err;
3066}
3067
3068static void __exit exit_ext3_fs(void)
3069{
3070 unregister_filesystem(&ext3_fs_type);
3071 destroy_inodecache();
3072 exit_ext3_xattr();
3073}
3074
3075MODULE_AUTHOR("Remy Card, Stephen Tweedie, Andrew Morton, Andreas Dilger, Theodore Ts'o and others");
3076MODULE_DESCRIPTION("Second Extended Filesystem with journaling extensions");
3077MODULE_LICENSE("GPL");
3078module_init(init_ext3_fs)
3079module_exit(exit_ext3_fs)
Clone of davem's net-next-2.6 tree