projects / deliverable / linux.git / blame
? search:
summary | shortlog | log | commit | commitdiff | tree
blob | blame (incremental) | history | HEAD
dquot: cleanup dquot drop routine
[deliverable/linux.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
167 * the journal instead. On recovery, the journal will compain about
168 * that error until we've noted it down and cleared it.
169 */
170
171static void ext3_handle_error(struct super_block *sb)
172{
173 struct ext3_super_block *es = EXT3_SB(sb)->s_es;
174
175 EXT3_SB(sb)->s_mount_state |= EXT3_ERROR_FS;
176 es->s_state |= cpu_to_le16(EXT3_ERROR_FS);
177
178 if (sb->s_flags & MS_RDONLY)
179 return;
180
7543fc7b 181 if (!test_opt (sb, ERRORS_CONT)) {
1da177e4
LT		 182 journal_t *journal = EXT3_SB(sb)->s_journal;
183
e3c96435 184 set_opt(EXT3_SB(sb)->s_mount_opt, ABORT);
1da177e4
LT		 185 if (journal)
186 journal_abort(journal, -EIO);
187 }
7543fc7b 188 if (test_opt (sb, ERRORS_RO)) {
4cf46b67
AF		 189 ext3_msg(sb, KERN_CRIT,
190 "error: remounting filesystem read-only");
7543fc7b
VA		 191 sb->s_flags |= MS_RDONLY;
192 }
193 ext3_commit_super(sb, es, 1);
1da177e4 194 if (test_opt(sb, ERRORS_PANIC))
4cf46b67 195 panic("EXT3-fs (%s): panic forced after error\n",
1da177e4 196 sb->s_id);
1da177e4
LT		 197}
198
199void ext3_error (struct super_block * sb, const char * function,
200 const char * fmt, ...)
201{
202 va_list args;
203
204 va_start(args, fmt);
205 printk(KERN_CRIT "EXT3-fs error (device %s): %s: ",sb->s_id, function);
206 vprintk(fmt, args);
207 printk("\n");
208 va_end(args);
209
210 ext3_handle_error(sb);
211}
212
213static const char *ext3_decode_error(struct super_block * sb, int errno,
214 char nbuf[16])
215{
216 char *errstr = NULL;
217
218 switch (errno) {
219 case -EIO:
220 errstr = "IO failure";
221 break;
222 case -ENOMEM:
223 errstr = "Out of memory";
224 break;
225 case -EROFS:
226 if (!sb || EXT3_SB(sb)->s_journal->j_flags & JFS_ABORT)
227 errstr = "Journal has aborted";
228 else
229 errstr = "Readonly filesystem";
230 break;
231 default:
232 /* If the caller passed in an extra buffer for unknown
233 * errors, textualise them now. Else we just return
234 * NULL. */
235 if (nbuf) {
236 /* Check for truncated error codes... */
237 if (snprintf(nbuf, 16, "error %d", -errno) >= 0)
238 errstr = nbuf;
239 }
240 break;
241 }
242
243 return errstr;
244}
245
246/* __ext3_std_error decodes expected errors from journaling functions
247 * automatically and invokes the appropriate error response. */
248
249void __ext3_std_error (struct super_block * sb, const char * function,
250 int errno)
251{
252 char nbuf[16];
30121624 253 const char *errstr;
1da177e4 254
30121624
ST		 255 /* Special case: if the error is EROFS, and we're not already
256 * inside a transaction, then there's really no point in logging
257 * an error. */
258 if (errno == -EROFS && journal_current_handle() == NULL &&
259 (sb->s_flags & MS_RDONLY))
260 return;
261
262 errstr = ext3_decode_error(sb, errno, nbuf);
4cf46b67 263 ext3_msg(sb, KERN_CRIT, "error in %s: %s", function, errstr);
1da177e4
LT		 264
265 ext3_handle_error(sb);
266}
267
268/*
269 * ext3_abort is a much stronger failure handler than ext3_error. The
270 * abort function may be used to deal with unrecoverable failures such
271 * as journal IO errors or ENOMEM at a critical moment in log management.
272 *
273 * We unconditionally force the filesystem into an ABORT|READONLY state,
274 * unless the error response on the fs has been set to panic in which
275 * case we take the easy way out and panic immediately.
276 */
277
278void ext3_abort (struct super_block * sb, const char * function,
279 const char * fmt, ...)
280{
281 va_list args;
282
1da177e4 283 va_start(args, fmt);
4cf46b67 284 printk(KERN_CRIT "EXT3-fs (%s): error: %s: ", sb->s_id, function);
1da177e4
LT		 285 vprintk(fmt, args);
286 printk("\n");
287 va_end(args);
288
289 if (test_opt(sb, ERRORS_PANIC))
4cf46b67 290 panic("EXT3-fs: panic from previous error\n");
1da177e4
LT		 291
292 if (sb->s_flags & MS_RDONLY)
293 return;
294
4cf46b67
AF		 295 ext3_msg(sb, KERN_CRIT,
296 "error: remounting filesystem read-only");
1da177e4
LT		 297 EXT3_SB(sb)->s_mount_state |= EXT3_ERROR_FS;
298 sb->s_flags |= MS_RDONLY;
e3c96435 299 set_opt(EXT3_SB(sb)->s_mount_opt, ABORT);
44d6f787
HK		 300 if (EXT3_SB(sb)->s_journal)
301 journal_abort(EXT3_SB(sb)->s_journal, -EIO);
1da177e4
LT		 302}
303
304void ext3_warning (struct super_block * sb, const char * function,
305 const char * fmt, ...)
306{
307 va_list args;
308
309 va_start(args, fmt);
4cf46b67 310 printk(KERN_WARNING "EXT3-fs (%s): warning: %s: ",
1da177e4
LT		 311 sb->s_id, function);
312 vprintk(fmt, args);
313 printk("\n");
314 va_end(args);
315}
316
317void ext3_update_dynamic_rev(struct super_block *sb)
318{
319 struct ext3_super_block *es = EXT3_SB(sb)->s_es;
320
321 if (le32_to_cpu(es->s_rev_level) > EXT3_GOOD_OLD_REV)
322 return;
323
4cf46b67
AF		 324 ext3_msg(sb, KERN_WARNING,
325 "warning: updating to rev %d because of "
326 "new feature flag, running e2fsck is recommended",
327 EXT3_DYNAMIC_REV);
1da177e4
LT		 328
329 es->s_first_ino = cpu_to_le32(EXT3_GOOD_OLD_FIRST_INO);
330 es->s_inode_size = cpu_to_le16(EXT3_GOOD_OLD_INODE_SIZE);
331 es->s_rev_level = cpu_to_le32(EXT3_DYNAMIC_REV);
332 /* leave es->s_feature_*compat flags alone */
333 /* es->s_uuid will be set by e2fsck if empty */
334
335 /*
336 * The rest of the superblock fields should be zero, and if not it
337 * means they are likely already in use, so leave them alone. We
338 * can leave it up to e2fsck to clean up any inconsistencies there.
339 */
340}
341
342/*
343 * Open the external journal device
344 */
4cf46b67 345static struct block_device *ext3_blkdev_get(dev_t dev, struct super_block *sb)
1da177e4
LT		 346{
347 struct block_device *bdev;
348 char b[BDEVNAME_SIZE];
349
350 bdev = open_by_devnum(dev, FMODE_READ|FMODE_WRITE);
351 if (IS_ERR(bdev))
352 goto fail;
353 return bdev;
354
355fail:
4cf46b67
AF		 356 ext3_msg(sb, "error: failed to open journal device %s: %ld",
357 __bdevname(dev, b), PTR_ERR(bdev));
358
1da177e4
LT		 359 return NULL;
360}
361
362/*
363 * Release the journal device
364 */
365static int ext3_blkdev_put(struct block_device *bdev)
366{
367 bd_release(bdev);
9a1c3542 368 return blkdev_put(bdev, FMODE_READ|FMODE_WRITE);
1da177e4
LT		 369}
370
371static int ext3_blkdev_remove(struct ext3_sb_info *sbi)
372{
373 struct block_device *bdev;
374 int ret = -ENODEV;
375
376 bdev = sbi->journal_bdev;
377 if (bdev) {
378 ret = ext3_blkdev_put(bdev);
379 sbi->journal_bdev = NULL;
380 }
381 return ret;
382}
383
384static inline struct inode *orphan_list_entry(struct list_head *l)
385{
386 return &list_entry(l, struct ext3_inode_info, i_orphan)->vfs_inode;
387}
388
389static void dump_orphan_list(struct super_block *sb, struct ext3_sb_info *sbi)
390{
391 struct list_head *l;
392
4cf46b67 393 ext3_msg(sb, KERN_ERR, "error: sb orphan head is %d",
1da177e4
LT		 394 le32_to_cpu(sbi->s_es->s_last_orphan));
395
4cf46b67 396 ext3_msg(sb, KERN_ERR, "sb_info orphan list:");
1da177e4
LT		 397 list_for_each(l, &sbi->s_orphan) {
398 struct inode *inode = orphan_list_entry(l);
4cf46b67 399 ext3_msg(sb, KERN_ERR, " "
eee194e7 400 "inode %s:%lu at %p: mode %o, nlink %d, next %d\n",
1da177e4 401 inode->i_sb->s_id, inode->i_ino, inode,
ae6ddcc5 402 inode->i_mode, inode->i_nlink,
1da177e4
LT		 403 NEXT_ORPHAN(inode));
404 }
405}
406
407static void ext3_put_super (struct super_block * sb)
408{
409 struct ext3_sb_info *sbi = EXT3_SB(sb);
410 struct ext3_super_block *es = sbi->s_es;
44d6f787 411 int i, err;
1da177e4 412
6cfd0148
CH		 413 lock_kernel();
414
1da177e4 415 ext3_xattr_put_super(sb);
44d6f787
HK		 416 err = journal_destroy(sbi->s_journal);
417 sbi->s_journal = NULL;
418 if (err < 0)
2d7c820e 419 ext3_abort(sb, __func__, "Couldn't clean up the journal");
44d6f787 420
1da177e4
LT		 421 if (!(sb->s_flags & MS_RDONLY)) {
422 EXT3_CLEAR_INCOMPAT_FEATURE(sb, EXT3_FEATURE_INCOMPAT_RECOVER);
423 es->s_state = cpu_to_le16(sbi->s_mount_state);
424 BUFFER_TRACE(sbi->s_sbh, "marking dirty");
425 mark_buffer_dirty(sbi->s_sbh);
426 ext3_commit_super(sb, es, 1);
427 }
428
429 for (i = 0; i < sbi->s_gdb_count; i++)
430 brelse(sbi->s_group_desc[i]);
431 kfree(sbi->s_group_desc);
432 percpu_counter_destroy(&sbi->s_freeblocks_counter);
433 percpu_counter_destroy(&sbi->s_freeinodes_counter);
434 percpu_counter_destroy(&sbi->s_dirs_counter);
435 brelse(sbi->s_sbh);
436#ifdef CONFIG_QUOTA
437 for (i = 0; i < MAXQUOTAS; i++)
438 kfree(sbi->s_qf_names[i]);
439#endif
440
441 /* Debugging code just in case the in-memory inode orphan list
442 * isn't empty. The on-disk one can be non-empty if we've
443 * detected an error and taken the fs readonly, but the
444 * in-memory list had better be clean by this point. */
445 if (!list_empty(&sbi->s_orphan))
446 dump_orphan_list(sb, sbi);
447 J_ASSERT(list_empty(&sbi->s_orphan));
448
f98393a6 449 invalidate_bdev(sb->s_bdev);
1da177e4
LT		 450 if (sbi->journal_bdev && sbi->journal_bdev != sb->s_bdev) {
451 /*
452 * Invalidate the journal device's buffers. We don't want them
453 * floating about in memory - the physical journal device may
454 * hotswapped, and it breaks the `ro-after' testing code.
455 */
456 sync_blockdev(sbi->journal_bdev);
f98393a6 457 invalidate_bdev(sbi->journal_bdev);
1da177e4
LT		 458 ext3_blkdev_remove(sbi);
459 }
460 sb->s_fs_info = NULL;
5df096d6 461 kfree(sbi->s_blockgroup_lock);
1da177e4 462 kfree(sbi);
6cfd0148
CH		 463
464 unlock_kernel();
1da177e4
LT		 465}
466
e18b890b 467static struct kmem_cache *ext3_inode_cachep;
1da177e4
LT		 468
469/*
470 * Called inside transaction, so use GFP_NOFS
471 */
472static struct inode *ext3_alloc_inode(struct super_block *sb)
473{
474 struct ext3_inode_info *ei;
475
e6b4f8da 476 ei = kmem_cache_alloc(ext3_inode_cachep, GFP_NOFS);
1da177e4
LT		 477 if (!ei)
478 return NULL;
1da177e4
LT		 479 ei->i_block_alloc_info = NULL;
480 ei->vfs_inode.i_version = 1;
fe8bc91c
JK		 481 atomic_set(&ei->i_datasync_tid, 0);
482 atomic_set(&ei->i_sync_tid, 0);
1da177e4
LT		 483 return &ei->vfs_inode;
484}
485
486static void ext3_destroy_inode(struct inode *inode)
487{
9f7dd93d
VA		 488 if (!list_empty(&(EXT3_I(inode)->i_orphan))) {
489 printk("EXT3 Inode %p: orphan list check failed!\n",
490 EXT3_I(inode));
491 print_hex_dump(KERN_INFO, "", DUMP_PREFIX_ADDRESS, 16, 4,
492 EXT3_I(inode), sizeof(struct ext3_inode_info),
493 false);
494 dump_stack();
495 }
1da177e4
LT		 496 kmem_cache_free(ext3_inode_cachep, EXT3_I(inode));
497}
498
51cc5068 499static void init_once(void *foo)
1da177e4
LT		 500{
501 struct ext3_inode_info *ei = (struct ext3_inode_info *) foo;
502
a35afb83 503 INIT_LIST_HEAD(&ei->i_orphan);
1da177e4 504#ifdef CONFIG_EXT3_FS_XATTR
a35afb83 505 init_rwsem(&ei->xattr_sem);
1da177e4 506#endif
a35afb83
CL		 507 mutex_init(&ei->truncate_mutex);
508 inode_init_once(&ei->vfs_inode);
1da177e4 509}
ae6ddcc5 510
1da177e4
LT		 511static int init_inodecache(void)
512{
513 ext3_inode_cachep = kmem_cache_create("ext3_inode_cache",
514 sizeof(struct ext3_inode_info),
fffb60f9
PJ		 515 0, (SLAB_RECLAIM_ACCOUNT|
516 SLAB_MEM_SPREAD),
20c2df83 517 init_once);
1da177e4
LT		 518 if (ext3_inode_cachep == NULL)
519 return -ENOMEM;
520 return 0;
521}
522
523static void destroy_inodecache(void)
524{
1a1d92c1 525 kmem_cache_destroy(ext3_inode_cachep);
1da177e4
LT		 526}
527
528static void ext3_clear_inode(struct inode *inode)
529{
530 struct ext3_block_alloc_info *rsv = EXT3_I(inode)->i_block_alloc_info;
257ba15c 531
9f754758 532 dquot_drop(inode);
1da177e4
LT		 533 ext3_discard_reservation(inode);
534 EXT3_I(inode)->i_block_alloc_info = NULL;
e6022603
AM		 535 if (unlikely(rsv))
536 kfree(rsv);
1da177e4
LT		 537}
538
6cd37cda 539static inline void ext3_show_quota_options(struct seq_file *seq, struct super_block *sb)
8fc2751b 540{
6cd37cda 541#if defined(CONFIG_QUOTA)
275abf5b 542 struct ext3_sb_info *sbi = EXT3_SB(sb);
8fc2751b 543
1aeec434
JK		 544 if (sbi->s_jquota_fmt) {
545 char *fmtname = "";
546
547 switch (sbi->s_jquota_fmt) {
548 case QFMT_VFS_OLD:
549 fmtname = "vfsold";
550 break;
551 case QFMT_VFS_V0:
552 fmtname = "vfsv0";
553 break;
554 case QFMT_VFS_V1:
555 fmtname = "vfsv1";
556 break;
557 }
558 seq_printf(seq, ",jqfmt=%s", fmtname);
559 }
8fc2751b
MB		 560
561 if (sbi->s_qf_names[USRQUOTA])
562 seq_printf(seq, ",usrjquota=%s", sbi->s_qf_names[USRQUOTA]);
563
564 if (sbi->s_qf_names[GRPQUOTA])
565 seq_printf(seq, ",grpjquota=%s", sbi->s_qf_names[GRPQUOTA]);
566
e3c96435 567 if (test_opt(sb, USRQUOTA))
8fc2751b
MB		 568 seq_puts(seq, ",usrquota");
569
e3c96435 570 if (test_opt(sb, GRPQUOTA))
8fc2751b
MB		 571 seq_puts(seq, ",grpquota");
572#endif
6cd37cda
PO		 573}
574
3c4cec65
JK		 575static char *data_mode_string(unsigned long mode)
576{
577 switch (mode) {
578 case EXT3_MOUNT_JOURNAL_DATA:
579 return "journal";
580 case EXT3_MOUNT_ORDERED_DATA:
581 return "ordered";
582 case EXT3_MOUNT_WRITEBACK_DATA:
583 return "writeback";
584 }
585 return "unknown";
586}
587
571beed8
MS		 588/*
589 * Show an option if
590 * - it's set to a non-default value OR
591 * - if the per-sb default is different from the global default
592 */
6cd37cda
PO		 593static int ext3_show_options(struct seq_file *seq, struct vfsmount *vfs)
594{
595 struct super_block *sb = vfs->mnt_sb;
571beed8
MS		 596 struct ext3_sb_info *sbi = EXT3_SB(sb);
597 struct ext3_super_block *es = sbi->s_es;
598 unsigned long def_mount_opts;
599
600 def_mount_opts = le32_to_cpu(es->s_default_mount_opts);
601
602 if (sbi->s_sb_block != 1)
603 seq_printf(seq, ",sb=%lu", sbi->s_sb_block);
604 if (test_opt(sb, MINIX_DF))
605 seq_puts(seq, ",minixdf");
606 if (test_opt(sb, GRPID))
607 seq_puts(seq, ",grpid");
608 if (!test_opt(sb, GRPID) && (def_mount_opts & EXT3_DEFM_BSDGROUPS))
609 seq_puts(seq, ",nogrpid");
610 if (sbi->s_resuid != EXT3_DEF_RESUID ||
611 le16_to_cpu(es->s_def_resuid) != EXT3_DEF_RESUID) {
612 seq_printf(seq, ",resuid=%u", sbi->s_resuid);
613 }
614 if (sbi->s_resgid != EXT3_DEF_RESGID ||
615 le16_to_cpu(es->s_def_resgid) != EXT3_DEF_RESGID) {
616 seq_printf(seq, ",resgid=%u", sbi->s_resgid);
617 }
1eca93f9 618 if (test_opt(sb, ERRORS_RO)) {
571beed8
MS		 619 int def_errors = le16_to_cpu(es->s_errors);
620
621 if (def_errors == EXT3_ERRORS_PANIC ||
1eca93f9
AK		 622 def_errors == EXT3_ERRORS_CONTINUE) {
623 seq_puts(seq, ",errors=remount-ro");
571beed8
MS		 624 }
625 }
1eca93f9
AK		 626 if (test_opt(sb, ERRORS_CONT))
627 seq_puts(seq, ",errors=continue");
571beed8
MS		 628 if (test_opt(sb, ERRORS_PANIC))
629 seq_puts(seq, ",errors=panic");
630 if (test_opt(sb, NO_UID32))
631 seq_puts(seq, ",nouid32");
632 if (test_opt(sb, DEBUG))
633 seq_puts(seq, ",debug");
634 if (test_opt(sb, OLDALLOC))
635 seq_puts(seq, ",oldalloc");
636#ifdef CONFIG_EXT3_FS_XATTR
637 if (test_opt(sb, XATTR_USER))
638 seq_puts(seq, ",user_xattr");
639 if (!test_opt(sb, XATTR_USER) &&
640 (def_mount_opts & EXT3_DEFM_XATTR_USER)) {
641 seq_puts(seq, ",nouser_xattr");
642 }
643#endif
644#ifdef CONFIG_EXT3_FS_POSIX_ACL
645 if (test_opt(sb, POSIX_ACL))
646 seq_puts(seq, ",acl");
647 if (!test_opt(sb, POSIX_ACL) && (def_mount_opts & EXT3_DEFM_ACL))
648 seq_puts(seq, ",noacl");
649#endif
650 if (!test_opt(sb, RESERVATION))
651 seq_puts(seq, ",noreservation");
652 if (sbi->s_commit_interval) {
653 seq_printf(seq, ",commit=%u",
654 (unsigned) (sbi->s_commit_interval / HZ));
655 }
656 if (test_opt(sb, BARRIER))
657 seq_puts(seq, ",barrier=1");
658 if (test_opt(sb, NOBH))
659 seq_puts(seq, ",nobh");
6cd37cda 660
e3c96435 661 seq_printf(seq, ",data=%s", data_mode_string(test_opt(sb, DATA_FLAGS)));
0e4fb5e2
HK		 662 if (test_opt(sb, DATA_ERR_ABORT))
663 seq_puts(seq, ",data_err=abort");
664
dee1d3b6
ES		 665 if (test_opt(sb, NOLOAD))
666 seq_puts(seq, ",norecovery");
667
6cd37cda 668 ext3_show_quota_options(seq, sb);
8fc2751b
MB		 669
670 return 0;
671}
1da177e4 672
fdb36673 673
74af0baa
CH		 674static struct inode *ext3_nfs_get_inode(struct super_block *sb,
675 u64 ino, u32 generation)
fdb36673 676{
fdb36673 677 struct inode *inode;
fdb36673
N		 678
679 if (ino < EXT3_FIRST_INO(sb) && ino != EXT3_ROOT_INO)
680 return ERR_PTR(-ESTALE);
681 if (ino > le32_to_cpu(EXT3_SB(sb)->s_es->s_inodes_count))
682 return ERR_PTR(-ESTALE);
683
684 /* iget isn't really right if the inode is currently unallocated!!
685 *
686 * ext3_read_inode will return a bad_inode if the inode had been
687 * deleted, so we should be safe.
688 *
689 * Currently we don't know the generation for parent directory, so
690 * a generation of 0 means "accept any"
691 */
473043dc
DH		 692 inode = ext3_iget(sb, ino);
693 if (IS_ERR(inode))
694 return ERR_CAST(inode);
695 if (generation && inode->i_generation != generation) {
fdb36673
N		 696 iput(inode);
697 return ERR_PTR(-ESTALE);
698 }
74af0baa
CH		 699
700 return inode;
701}
702
703static struct dentry *ext3_fh_to_dentry(struct super_block *sb, struct fid *fid,
704 int fh_len, int fh_type)
705{
706 return generic_fh_to_dentry(sb, fid, fh_len, fh_type,
707 ext3_nfs_get_inode);
708}
709
710static struct dentry *ext3_fh_to_parent(struct super_block *sb, struct fid *fid,
711 int fh_len, int fh_type)
712{
713 return generic_fh_to_parent(sb, fid, fh_len, fh_type,
714 ext3_nfs_get_inode);
fdb36673
N		 715}
716
6b082b53
TO		 717/*
718 * Try to release metadata pages (indirect blocks, directories) which are
719 * mapped via the block device. Since these pages could have journal heads
720 * which would prevent try_to_free_buffers() from freeing them, we must use
721 * jbd layer's try_to_free_buffers() function to release them.
722 */
723static int bdev_try_to_free_page(struct super_block *sb, struct page *page,
724 gfp_t wait)
725{
726 journal_t *journal = EXT3_SB(sb)->s_journal;
727
728 WARN_ON(PageChecked(page));
729 if (!page_has_buffers(page))
730 return 0;
731 if (journal)
732 return journal_try_to_free_buffers(journal, page,
733 wait & ~__GFP_WAIT);
734 return try_to_free_buffers(page);
735}
736
8fc2751b 737#ifdef CONFIG_QUOTA
1da177e4
LT		 738#define QTYPE2NAME(t) ((t)==USRQUOTA?"user":"group")
739#define QTYPE2MOPT(on, t) ((t)==USRQUOTA?((on)##USRJQUOTA):((on)##GRPJQUOTA))
740
1da177e4
LT		 741static int ext3_write_dquot(struct dquot *dquot);
742static int ext3_acquire_dquot(struct dquot *dquot);
743static int ext3_release_dquot(struct dquot *dquot);
744static int ext3_mark_dquot_dirty(struct dquot *dquot);
745static int ext3_write_info(struct super_block *sb, int type);
2fd83a4f
JK		 746static int ext3_quota_on(struct super_block *sb, int type, int format_id,
747 char *path, int remount);
1da177e4
LT		 748static int ext3_quota_on_mount(struct super_block *sb, int type);
749static ssize_t ext3_quota_read(struct super_block *sb, int type, char *data,
750 size_t len, loff_t off);
751static ssize_t ext3_quota_write(struct super_block *sb, int type,
752 const char *data, size_t len, loff_t off);
753
61e225dc 754static const struct dquot_operations ext3_quota_operations = {
a219ce37 755 .initialize = dquot_initialize,
1da177e4
LT		 756 .write_dquot = ext3_write_dquot,
757 .acquire_dquot = ext3_acquire_dquot,
758 .release_dquot = ext3_release_dquot,
759 .mark_dirty = ext3_mark_dquot_dirty,
157091a2
JK		 760 .write_info = ext3_write_info,
761 .alloc_dquot = dquot_alloc,
762 .destroy_dquot = dquot_destroy,
1da177e4
LT		 763};
764
0d54b217 765static const struct quotactl_ops ext3_qctl_operations = {
1da177e4
LT		 766 .quota_on = ext3_quota_on,
767 .quota_off = vfs_quota_off,
768 .quota_sync = vfs_quota_sync,
769 .get_info = vfs_get_dqinfo,
770 .set_info = vfs_set_dqinfo,
771 .get_dqblk = vfs_get_dqblk,
772 .set_dqblk = vfs_set_dqblk
773};
774#endif
775
ee9b6d61 776static const struct super_operations ext3_sops = {
1da177e4
LT		 777 .alloc_inode = ext3_alloc_inode,
778 .destroy_inode = ext3_destroy_inode,
1da177e4
LT		 779 .write_inode = ext3_write_inode,
780 .dirty_inode = ext3_dirty_inode,
781 .delete_inode = ext3_delete_inode,
782 .put_super = ext3_put_super,
1da177e4 783 .sync_fs = ext3_sync_fs,
c4be0c1d
TS		 784 .freeze_fs = ext3_freeze,
785 .unfreeze_fs = ext3_unfreeze,
1da177e4
LT		 786 .statfs = ext3_statfs,
787 .remount_fs = ext3_remount,
788 .clear_inode = ext3_clear_inode,
8fc2751b 789 .show_options = ext3_show_options,
1da177e4
LT		 790#ifdef CONFIG_QUOTA
791 .quota_read = ext3_quota_read,
792 .quota_write = ext3_quota_write,
793#endif
6b082b53 794 .bdev_try_to_free_page = bdev_try_to_free_page,
1da177e4
LT		 795};
796
39655164 797static const struct export_operations ext3_export_ops = {
74af0baa
CH		 798 .fh_to_dentry = ext3_fh_to_dentry,
799 .fh_to_parent = ext3_fh_to_parent,
1da177e4
LT		 800 .get_parent = ext3_get_parent,
801};
802
803enum {
804 Opt_bsd_df, Opt_minix_df, Opt_grpid, Opt_nogrpid,
805 Opt_resgid, Opt_resuid, Opt_sb, Opt_err_cont, Opt_err_panic, Opt_err_ro,
2860b733 806 Opt_nouid32, Opt_nocheck, Opt_debug, Opt_oldalloc, Opt_orlov,
1da177e4 807 Opt_user_xattr, Opt_nouser_xattr, Opt_acl, Opt_noacl,
ade1a29e 808 Opt_reservation, Opt_noreservation, Opt_noload, Opt_nobh, Opt_bh,
71b96257 809 Opt_commit, Opt_journal_update, Opt_journal_inum, Opt_journal_dev,
1da177e4 810 Opt_abort, Opt_data_journal, Opt_data_ordered, Opt_data_writeback,
0e4fb5e2 811 Opt_data_err_abort, Opt_data_err_ignore,
1da177e4 812 Opt_usrjquota, Opt_grpjquota, Opt_offusrjquota, Opt_offgrpjquota,
1aeec434
JK		 813 Opt_jqfmt_vfsold, Opt_jqfmt_vfsv0, Opt_jqfmt_vfsv1, Opt_quota,
814 Opt_noquota, Opt_ignore, Opt_barrier, Opt_err, Opt_resize,
815 Opt_usrquota, Opt_grpquota
1da177e4
LT		 816};
817
a447c093 818static const match_table_t tokens = {
1da177e4
LT		 819 {Opt_bsd_df, "bsddf"},
820 {Opt_minix_df, "minixdf"},
821 {Opt_grpid, "grpid"},
822 {Opt_grpid, "bsdgroups"},
823 {Opt_nogrpid, "nogrpid"},
824 {Opt_nogrpid, "sysvgroups"},
825 {Opt_resgid, "resgid=%u"},
826 {Opt_resuid, "resuid=%u"},
827 {Opt_sb, "sb=%u"},
828 {Opt_err_cont, "errors=continue"},
829 {Opt_err_panic, "errors=panic"},
830 {Opt_err_ro, "errors=remount-ro"},
831 {Opt_nouid32, "nouid32"},
832 {Opt_nocheck, "nocheck"},
833 {Opt_nocheck, "check=none"},
1da177e4
LT		 834 {Opt_debug, "debug"},
835 {Opt_oldalloc, "oldalloc"},
836 {Opt_orlov, "orlov"},
837 {Opt_user_xattr, "user_xattr"},
838 {Opt_nouser_xattr, "nouser_xattr"},
839 {Opt_acl, "acl"},
840 {Opt_noacl, "noacl"},
841 {Opt_reservation, "reservation"},
842 {Opt_noreservation, "noreservation"},
843 {Opt_noload, "noload"},
dee1d3b6 844 {Opt_noload, "norecovery"},
1da177e4 845 {Opt_nobh, "nobh"},
ade1a29e 846 {Opt_bh, "bh"},
1da177e4
LT		 847 {Opt_commit, "commit=%u"},
848 {Opt_journal_update, "journal=update"},
849 {Opt_journal_inum, "journal=%u"},
71b96257 850 {Opt_journal_dev, "journal_dev=%u"},
1da177e4
LT		 851 {Opt_abort, "abort"},
852 {Opt_data_journal, "data=journal"},
853 {Opt_data_ordered, "data=ordered"},
854 {Opt_data_writeback, "data=writeback"},
0e4fb5e2
HK		 855 {Opt_data_err_abort, "data_err=abort"},
856 {Opt_data_err_ignore, "data_err=ignore"},
1da177e4
LT		 857 {Opt_offusrjquota, "usrjquota="},
858 {Opt_usrjquota, "usrjquota=%s"},
859 {Opt_offgrpjquota, "grpjquota="},
860 {Opt_grpjquota, "grpjquota=%s"},
861 {Opt_jqfmt_vfsold, "jqfmt=vfsold"},
862 {Opt_jqfmt_vfsv0, "jqfmt=vfsv0"},
1aeec434 863 {Opt_jqfmt_vfsv1, "jqfmt=vfsv1"},
8fc2751b 864 {Opt_grpquota, "grpquota"},
1f54587b
JK		 865 {Opt_noquota, "noquota"},
866 {Opt_quota, "quota"},
8fc2751b 867 {Opt_usrquota, "usrquota"},
1da177e4 868 {Opt_barrier, "barrier=%u"},
1da177e4 869 {Opt_resize, "resize"},
92587216 870 {Opt_err, NULL},
1da177e4
LT		 871};
872
4cf46b67 873static ext3_fsblk_t get_sb_block(void **data, struct super_block *sb)
1da177e4 874{
e9ad5620
DK		 875 ext3_fsblk_t sb_block;
876 char *options = (char *) *data;
1da177e4
LT		 877
878 if (!options || strncmp(options, "sb=", 3) != 0)
879 return 1; /* Default location */
880 options += 3;
43d23f90 881 /*todo: use simple_strtoll with >32bit ext3 */
1da177e4
LT		 882 sb_block = simple_strtoul(options, &options, 0);
883 if (*options && *options != ',') {
4cf46b67 884 ext3_msg(sb, "error: invalid sb specification: %s",
1da177e4
LT		 885 (char *) *data);
886 return 1;
887 }
888 if (*options == ',')
889 options++;
890 *data = (void *) options;
891 return sb_block;
892}
893
e1f5c67a
DM		 894#ifdef CONFIG_QUOTA
895static int set_qf_name(struct super_block *sb, int qtype, substring_t *args)
896{
897 struct ext3_sb_info *sbi = EXT3_SB(sb);
898 char *qname;
899
900 if (sb_any_quota_loaded(sb) &&
901 !sbi->s_qf_names[qtype]) {
902 ext3_msg(sb, KERN_ERR,
903 "Cannot change journaled "
904 "quota options when quota turned on");
905 return 0;
906 }
907 qname = match_strdup(args);
908 if (!qname) {
909 ext3_msg(sb, KERN_ERR,
910 "Not enough memory for storing quotafile name");
911 return 0;
912 }
913 if (sbi->s_qf_names[qtype] &&
914 strcmp(sbi->s_qf_names[qtype], qname)) {
915 ext3_msg(sb, KERN_ERR,
916 "%s quota file already specified", QTYPE2NAME(qtype));
917 kfree(qname);
918 return 0;
919 }
920 sbi->s_qf_names[qtype] = qname;
921 if (strchr(sbi->s_qf_names[qtype], '/')) {
922 ext3_msg(sb, KERN_ERR,
923 "quotafile must be on filesystem root");
924 kfree(sbi->s_qf_names[qtype]);
925 sbi->s_qf_names[qtype] = NULL;
926 return 0;
927 }
928 set_opt(sbi->s_mount_opt, QUOTA);
929 return 1;
930}
931
932static int clear_qf_name(struct super_block *sb, int qtype) {
933
934 struct ext3_sb_info *sbi = EXT3_SB(sb);
935
936 if (sb_any_quota_loaded(sb) &&
937 sbi->s_qf_names[qtype]) {
938 ext3_msg(sb, KERN_ERR, "Cannot change journaled quota options"
939 " when quota turned on");
940 return 0;
941 }
942 /*
943 * The space will be released later when all options are confirmed
944 * to be correct
945 */
946 sbi->s_qf_names[qtype] = NULL;
947 return 1;
948}
949#endif
950
71b96257 951static int parse_options (char *options, struct super_block *sb,
eee194e7 952 unsigned int *inum, unsigned long *journal_devnum,
43d23f90 953 ext3_fsblk_t *n_blocks_count, int is_remount)
1da177e4
LT		 954{
955 struct ext3_sb_info *sbi = EXT3_SB(sb);
956 char * p;
957 substring_t args[MAX_OPT_ARGS];
958 int data_opt = 0;
959 int option;
960#ifdef CONFIG_QUOTA
e1f5c67a 961 int qfmt;
1da177e4
LT		 962#endif
963
964 if (!options)
965 return 1;
966
967 while ((p = strsep (&options, ",")) != NULL) {
968 int token;
969 if (!*p)
970 continue;
971
972 token = match_token(p, tokens, args);
973 switch (token) {
974 case Opt_bsd_df:
975 clear_opt (sbi->s_mount_opt, MINIX_DF);
976 break;
977 case Opt_minix_df:
978 set_opt (sbi->s_mount_opt, MINIX_DF);
979 break;
980 case Opt_grpid:
981 set_opt (sbi->s_mount_opt, GRPID);
982 break;
983 case Opt_nogrpid:
984 clear_opt (sbi->s_mount_opt, GRPID);
985 break;
986 case Opt_resuid:
987 if (match_int(&args[0], &option))
988 return 0;
989 sbi->s_resuid = option;
990 break;
991 case Opt_resgid:
992 if (match_int(&args[0], &option))
993 return 0;
994 sbi->s_resgid = option;
995 break;
996 case Opt_sb:
997 /* handled by get_sb_block() instead of here */
998 /* *sb_block = match_int(&args[0]); */
999 break;
1000 case Opt_err_panic:
1001 clear_opt (sbi->s_mount_opt, ERRORS_CONT);
1002 clear_opt (sbi->s_mount_opt, ERRORS_RO);
1003 set_opt (sbi->s_mount_opt, ERRORS_PANIC);
1004 break;
1005 case Opt_err_ro:
1006 clear_opt (sbi->s_mount_opt, ERRORS_CONT);
1007 clear_opt (sbi->s_mount_opt, ERRORS_PANIC);
1008 set_opt (sbi->s_mount_opt, ERRORS_RO);
1009 break;
1010 case Opt_err_cont:
1011 clear_opt (sbi->s_mount_opt, ERRORS_RO);
1012 clear_opt (sbi->s_mount_opt, ERRORS_PANIC);
1013 set_opt (sbi->s_mount_opt, ERRORS_CONT);
1014 break;
1015 case Opt_nouid32:
1016 set_opt (sbi->s_mount_opt, NO_UID32);
1017 break;
1da177e4
LT		 1018 case Opt_nocheck:
1019 clear_opt (sbi->s_mount_opt, CHECK);
1020 break;
1021 case Opt_debug:
1022 set_opt (sbi->s_mount_opt, DEBUG);
1023 break;
1024 case Opt_oldalloc:
1025 set_opt (sbi->s_mount_opt, OLDALLOC);
1026 break;
1027 case Opt_orlov:
1028 clear_opt (sbi->s_mount_opt, OLDALLOC);
1029 break;
1030#ifdef CONFIG_EXT3_FS_XATTR
1031 case Opt_user_xattr:
1032 set_opt (sbi->s_mount_opt, XATTR_USER);
1033 break;
1034 case Opt_nouser_xattr:
1035 clear_opt (sbi->s_mount_opt, XATTR_USER);
1036 break;
1037#else
1038 case Opt_user_xattr:
1039 case Opt_nouser_xattr:
4cf46b67
AF		 1040 ext3_msg(sb, KERN_INFO,
1041 "(no)user_xattr options not supported");
1da177e4
LT		 1042 break;
1043#endif
1044#ifdef CONFIG_EXT3_FS_POSIX_ACL
1045 case Opt_acl:
1046 set_opt(sbi->s_mount_opt, POSIX_ACL);
1047 break;
1048 case Opt_noacl:
1049 clear_opt(sbi->s_mount_opt, POSIX_ACL);
1050 break;
1051#else
1052 case Opt_acl:
1053 case Opt_noacl:
4cf46b67
AF		 1054 ext3_msg(sb, KERN_INFO,
1055 "(no)acl options not supported");
1da177e4
LT		 1056 break;
1057#endif
1058 case Opt_reservation:
1059 set_opt(sbi->s_mount_opt, RESERVATION);
1060 break;
1061 case Opt_noreservation:
1062 clear_opt(sbi->s_mount_opt, RESERVATION);
1063 break;
1064 case Opt_journal_update:
1065 /* @@@ FIXME */
1066 /* Eventually we will want to be able to create
1067 a journal file here. For now, only allow the
1068 user to specify an existing inode to be the
1069 journal file. */
1070 if (is_remount) {
4cf46b67
AF		 1071 ext3_msg(sb, KERN_ERR, "error: cannot specify "
1072 "journal on remount");
1da177e4
LT		 1073 return 0;
1074 }
1075 set_opt (sbi->s_mount_opt, UPDATE_JOURNAL);
1076 break;
1077 case Opt_journal_inum:
1078 if (is_remount) {
4cf46b67
AF		 1079 ext3_msg(sb, KERN_ERR, "error: cannot specify "
1080 "journal on remount");
1da177e4
LT		 1081 return 0;
1082 }
1083 if (match_int(&args[0], &option))
1084 return 0;
1085 *inum = option;
1086 break;
71b96257
JL		 1087 case Opt_journal_dev:
1088 if (is_remount) {
4cf46b67
AF		 1089 ext3_msg(sb, KERN_ERR, "error: cannot specify "
1090 "journal on remount");
71b96257
JL		 1091 return 0;
1092 }
1093 if (match_int(&args[0], &option))
1094 return 0;
1095 *journal_devnum = option;
1096 break;
1da177e4
LT		 1097 case Opt_noload:
1098 set_opt (sbi->s_mount_opt, NOLOAD);
1099 break;
1100 case Opt_commit:
1101 if (match_int(&args[0], &option))
1102 return 0;
1103 if (option < 0)
1104 return 0;
1105 if (option == 0)
1106 option = JBD_DEFAULT_MAX_COMMIT_AGE;
1107 sbi->s_commit_interval = HZ * option;
1108 break;
1109 case Opt_data_journal:
1110 data_opt = EXT3_MOUNT_JOURNAL_DATA;
1111 goto datacheck;
1112 case Opt_data_ordered:
1113 data_opt = EXT3_MOUNT_ORDERED_DATA;
1114 goto datacheck;
1115 case Opt_data_writeback:
1116 data_opt = EXT3_MOUNT_WRITEBACK_DATA;
1117 datacheck:
1118 if (is_remount) {
e3c96435 1119 if (test_opt(sb, DATA_FLAGS) == data_opt)
3c4cec65 1120 break;
4cf46b67
AF		 1121 ext3_msg(sb, KERN_ERR,
1122 "error: cannot change "
3c4cec65
JK		 1123 "data mode on remount. The filesystem "
1124 "is mounted in data=%s mode and you "
4cf46b67 1125 "try to remount it in data=%s mode.",
e3c96435
DM		 1126 data_mode_string(test_opt(sb,
1127 DATA_FLAGS)),
3c4cec65
JK		 1128 data_mode_string(data_opt));
1129 return 0;
1da177e4 1130 } else {
e3c96435 1131 clear_opt(sbi->s_mount_opt, DATA_FLAGS);
1da177e4
LT		 1132 sbi->s_mount_opt |= data_opt;
1133 }
1134 break;
0e4fb5e2
HK		 1135 case Opt_data_err_abort:
1136 set_opt(sbi->s_mount_opt, DATA_ERR_ABORT);
1137 break;
1138 case Opt_data_err_ignore:
1139 clear_opt(sbi->s_mount_opt, DATA_ERR_ABORT);
1140 break;
1da177e4
LT		 1141#ifdef CONFIG_QUOTA
1142 case Opt_usrjquota:
e1f5c67a 1143 if (!set_qf_name(sb, USRQUOTA, &args[0]))
1da177e4 1144 return 0;
e1f5c67a
DM		 1145 break;
1146 case Opt_grpjquota:
1147 if (!set_qf_name(sb, GRPQUOTA, &args[0]))
1da177e4 1148 return 0;
1da177e4
LT		 1149 break;
1150 case Opt_offusrjquota:
e1f5c67a
DM		 1151 if (!clear_qf_name(sb, USRQUOTA))
1152 return 0;
1153 break;
1da177e4 1154 case Opt_offgrpjquota:
e1f5c67a 1155 if (!clear_qf_name(sb, GRPQUOTA))
1da177e4 1156 return 0;
1da177e4
LT		 1157 break;
1158 case Opt_jqfmt_vfsold:
d06bf1d2
JK		 1159 qfmt = QFMT_VFS_OLD;
1160 goto set_qf_format;
1da177e4 1161 case Opt_jqfmt_vfsv0:
d06bf1d2 1162 qfmt = QFMT_VFS_V0;
1aeec434
JK		 1163 goto set_qf_format;
1164 case Opt_jqfmt_vfsv1:
1165 qfmt = QFMT_VFS_V1;
d06bf1d2 1166set_qf_format:
ee0d5ffe 1167 if (sb_any_quota_loaded(sb) &&
d06bf1d2 1168 sbi->s_jquota_fmt != qfmt) {
4cf46b67 1169 ext3_msg(sb, KERN_ERR, "error: cannot change "
d06bf1d2 1170 "journaled quota options when "
4cf46b67 1171 "quota turned on.");
d06bf1d2
JK		 1172 return 0;
1173 }
1174 sbi->s_jquota_fmt = qfmt;
1da177e4 1175 break;
1f54587b 1176 case Opt_quota:
8fc2751b
MB		 1177 case Opt_usrquota:
1178 set_opt(sbi->s_mount_opt, QUOTA);
1179 set_opt(sbi->s_mount_opt, USRQUOTA);
1180 break;
1181 case Opt_grpquota:
1f54587b 1182 set_opt(sbi->s_mount_opt, QUOTA);
8fc2751b 1183 set_opt(sbi->s_mount_opt, GRPQUOTA);
1f54587b
JK		 1184 break;
1185 case Opt_noquota:
ee0d5ffe 1186 if (sb_any_quota_loaded(sb)) {
4cf46b67
AF		 1187 ext3_msg(sb, KERN_ERR, "error: cannot change "
1188 "quota options when quota turned on.");
1f54587b
JK		 1189 return 0;
1190 }
1191 clear_opt(sbi->s_mount_opt, QUOTA);
8fc2751b
MB		 1192 clear_opt(sbi->s_mount_opt, USRQUOTA);
1193 clear_opt(sbi->s_mount_opt, GRPQUOTA);
1f54587b 1194 break;
1da177e4 1195#else
8fc2751b
MB		 1196 case Opt_quota:
1197 case Opt_usrquota:
1198 case Opt_grpquota:
4cf46b67
AF		 1199 ext3_msg(sb, KERN_ERR,
1200 "error: quota options not supported.");
fa1ff1e0 1201 break;
1da177e4
LT		 1202 case Opt_usrjquota:
1203 case Opt_grpjquota:
1204 case Opt_offusrjquota:
1205 case Opt_offgrpjquota:
1206 case Opt_jqfmt_vfsold:
1207 case Opt_jqfmt_vfsv0:
1aeec434 1208 case Opt_jqfmt_vfsv1:
4cf46b67
AF		 1209 ext3_msg(sb, KERN_ERR,
1210 "error: journaled quota options not "
1211 "supported.");
1da177e4 1212 break;
1f54587b
JK		 1213 case Opt_noquota:
1214 break;
1da177e4
LT		 1215#endif
1216 case Opt_abort:
1217 set_opt(sbi->s_mount_opt, ABORT);
1218 break;
1219 case Opt_barrier:
1220 if (match_int(&args[0], &option))
1221 return 0;
1222 if (option)
1223 set_opt(sbi->s_mount_opt, BARRIER);
1224 else
1225 clear_opt(sbi->s_mount_opt, BARRIER);
1226 break;
1227 case Opt_ignore:
1228 break;
1229 case Opt_resize:
08c6a96f 1230 if (!is_remount) {
4cf46b67
AF		 1231 ext3_msg(sb, KERN_ERR,
1232 "error: resize option only available "
1233 "for remount");
1da177e4
LT		 1234 return 0;
1235 }
c7f1721e
KZ		 1236 if (match_int(&args[0], &option) != 0)
1237 return 0;
1da177e4
LT		 1238 *n_blocks_count = option;
1239 break;
1240 case Opt_nobh:
1241 set_opt(sbi->s_mount_opt, NOBH);
1242 break;
ade1a29e
BP		 1243 case Opt_bh:
1244 clear_opt(sbi->s_mount_opt, NOBH);
1245 break;
1da177e4 1246 default:
4cf46b67
AF		 1247 ext3_msg(sb, KERN_ERR,
1248 "error: unrecognized mount option \"%s\" "
1249 "or missing value", p);
1da177e4
LT		 1250 return 0;
1251 }
1252 }
1253#ifdef CONFIG_QUOTA
8fc2751b 1254 if (sbi->s_qf_names[USRQUOTA] || sbi->s_qf_names[GRPQUOTA]) {
e3c96435 1255 if (test_opt(sb, USRQUOTA) && sbi->s_qf_names[USRQUOTA])
8fc2751b 1256 clear_opt(sbi->s_mount_opt, USRQUOTA);
e3c96435 1257 if (test_opt(sb, GRPQUOTA) && sbi->s_qf_names[GRPQUOTA])
8fc2751b
MB		 1258 clear_opt(sbi->s_mount_opt, GRPQUOTA);
1259
e1f5c67a 1260 if (test_opt(sb, GRPQUOTA) || test_opt(sb, USRQUOTA)) {
4cf46b67
AF		 1261 ext3_msg(sb, KERN_ERR, "error: old and new quota "
1262 "format mixing.");
8fc2751b
MB		 1263 return 0;
1264 }
1265
1266 if (!sbi->s_jquota_fmt) {
4cf46b67
AF		 1267 ext3_msg(sb, KERN_ERR, "error: journaled quota format "
1268 "not specified.");
8fc2751b
MB		 1269 return 0;
1270 }
1271 } else {
1272 if (sbi->s_jquota_fmt) {
4cf46b67 1273 ext3_msg(sb, KERN_ERR, "error: journaled quota format "
99aeaf63 1274 "specified with no journaling "
4cf46b67 1275 "enabled.");
8fc2751b
MB		 1276 return 0;
1277 }
1da177e4
LT		 1278 }
1279#endif
1da177e4
LT		 1280 return 1;
1281}
1282
1283static int ext3_setup_super(struct super_block *sb, struct ext3_super_block *es,
1284 int read_only)
1285{
1286 struct ext3_sb_info *sbi = EXT3_SB(sb);
1287 int res = 0;
1288
1289 if (le32_to_cpu(es->s_rev_level) > EXT3_MAX_SUPP_REV) {
4cf46b67
AF		 1290 ext3_msg(sb, KERN_ERR,
1291 "error: revision level too high, "
1292 "forcing read-only mode");
1da177e4
LT		 1293 res = MS_RDONLY;
1294 }
1295 if (read_only)
1296 return res;
1297 if (!(sbi->s_mount_state & EXT3_VALID_FS))
4cf46b67
AF		 1298 ext3_msg(sb, KERN_WARNING,
1299 "warning: mounting unchecked fs, "
1300 "running e2fsck is recommended");
1da177e4 1301 else if ((sbi->s_mount_state & EXT3_ERROR_FS))
4cf46b67
AF		 1302 ext3_msg(sb, KERN_WARNING,
1303 "warning: mounting fs with errors, "
1304 "running e2fsck is recommended");
1da177e4
LT		 1305 else if ((__s16) le16_to_cpu(es->s_max_mnt_count) >= 0 &&
1306 le16_to_cpu(es->s_mnt_count) >=
1307 (unsigned short) (__s16) le16_to_cpu(es->s_max_mnt_count))
4cf46b67
AF		 1308 ext3_msg(sb, KERN_WARNING,
1309 "warning: maximal mount count reached, "
1310 "running e2fsck is recommended");
1da177e4
LT		 1311 else if (le32_to_cpu(es->s_checkinterval) &&
1312 (le32_to_cpu(es->s_lastcheck) +
1313 le32_to_cpu(es->s_checkinterval) <= get_seconds()))
4cf46b67
AF		 1314 ext3_msg(sb, KERN_WARNING,
1315 "warning: checktime reached, "
1316 "running e2fsck is recommended");
1da177e4
LT		 1317#if 0
1318 /* @@@ We _will_ want to clear the valid bit if we find
1319 inconsistencies, to force a fsck at reboot. But for
1320 a plain journaled filesystem we can keep it set as
1321 valid forever! :) */
e7f23ebd 1322 es->s_state &= cpu_to_le16(~EXT3_VALID_FS);
1da177e4
LT		 1323#endif
1324 if (!(__s16) le16_to_cpu(es->s_max_mnt_count))
1325 es->s_max_mnt_count = cpu_to_le16(EXT3_DFL_MAX_MNT_COUNT);
50e8a289 1326 le16_add_cpu(&es->s_mnt_count, 1);
1da177e4
LT		 1327 es->s_mtime = cpu_to_le32(get_seconds());
1328 ext3_update_dynamic_rev(sb);
1329 EXT3_SET_INCOMPAT_FEATURE(sb, EXT3_FEATURE_INCOMPAT_RECOVER);
1330
1331 ext3_commit_super(sb, es, 1);
1332 if (test_opt(sb, DEBUG))
4cf46b67
AF		 1333 ext3_msg(sb, KERN_INFO, "[bs=%lu, gc=%lu, "
1334 "bpg=%lu, ipg=%lu, mo=%04lx]",
1da177e4
LT		 1335 sb->s_blocksize,
1336 sbi->s_groups_count,
1337 EXT3_BLOCKS_PER_GROUP(sb),
1338 EXT3_INODES_PER_GROUP(sb),
1339 sbi->s_mount_opt);
1340
1da177e4
LT		 1341 if (EXT3_SB(sb)->s_journal->j_inode == NULL) {
1342 char b[BDEVNAME_SIZE];
4cf46b67 1343 ext3_msg(sb, KERN_INFO, "using external journal on %s",
1da177e4
LT		 1344 bdevname(EXT3_SB(sb)->s_journal->j_dev, b));
1345 } else {
4cf46b67 1346 ext3_msg(sb, KERN_INFO, "using internal journal");
1da177e4 1347 }
1da177e4
LT		 1348 return res;
1349}
1350
1351/* Called at mount-time, super-block is locked */
197cd65a 1352static int ext3_check_descriptors(struct super_block *sb)
1da177e4
LT		 1353{
1354 struct ext3_sb_info *sbi = EXT3_SB(sb);
1da177e4
LT		 1355 int i;
1356
1357 ext3_debug ("Checking group descriptors");
1358
197cd65a
AM		 1359 for (i = 0; i < sbi->s_groups_count; i++) {
1360 struct ext3_group_desc *gdp = ext3_get_group_desc(sb, i, NULL);
1eaafeae
AM		 1361 ext3_fsblk_t first_block = ext3_group_first_block_no(sb, i);
1362 ext3_fsblk_t last_block;
197cd65a 1363
855565e8
ES		 1364 if (i == sbi->s_groups_count - 1)
1365 last_block = le32_to_cpu(sbi->s_es->s_blocks_count) - 1;
1366 else
1367 last_block = first_block +
1368 (EXT3_BLOCKS_PER_GROUP(sb) - 1);
1369
855565e8
ES		 1370 if (le32_to_cpu(gdp->bg_block_bitmap) < first_block ||
1371 le32_to_cpu(gdp->bg_block_bitmap) > last_block)
1da177e4
LT		 1372 {
1373 ext3_error (sb, "ext3_check_descriptors",
1374 "Block bitmap for group %d"
1375 " not in group (block %lu)!",
1376 i, (unsigned long)
1377 le32_to_cpu(gdp->bg_block_bitmap));
1378 return 0;
1379 }
855565e8
ES		 1380 if (le32_to_cpu(gdp->bg_inode_bitmap) < first_block ||
1381 le32_to_cpu(gdp->bg_inode_bitmap) > last_block)
1da177e4
LT		 1382 {
1383 ext3_error (sb, "ext3_check_descriptors",
1384 "Inode bitmap for group %d"
1385 " not in group (block %lu)!",
1386 i, (unsigned long)
1387 le32_to_cpu(gdp->bg_inode_bitmap));
1388 return 0;
1389 }
855565e8 1390 if (le32_to_cpu(gdp->bg_inode_table) < first_block ||
780dcdb2 1391 le32_to_cpu(gdp->bg_inode_table) + sbi->s_itb_per_group - 1 >
855565e8 1392 last_block)
1da177e4
LT		 1393 {
1394 ext3_error (sb, "ext3_check_descriptors",
1395 "Inode table for group %d"
1396 " not in group (block %lu)!",
1397 i, (unsigned long)
1398 le32_to_cpu(gdp->bg_inode_table));
1399 return 0;
1400 }
1da177e4
LT		 1401 }
1402
1403 sbi->s_es->s_free_blocks_count=cpu_to_le32(ext3_count_free_blocks(sb));
1404 sbi->s_es->s_free_inodes_count=cpu_to_le32(ext3_count_free_inodes(sb));
1405 return 1;
1406}
1407
1408
1409/* ext3_orphan_cleanup() walks a singly-linked list of inodes (starting at
1410 * the superblock) which were deleted from all directories, but held open by
1411 * a process at the time of a crash. We walk the list and try to delete these
1412 * inodes at recovery time (only with a read-write filesystem).
1413 *
1414 * In order to keep the orphan inode chain consistent during traversal (in
1415 * case of crash during recovery), we link each inode into the superblock
1416 * orphan list_head and handle it the same way as an inode deletion during
1417 * normal operation (which journals the operations for us).
1418 *
1419 * We only do an iget() and an iput() on each inode, which is very safe if we
1420 * accidentally point at an in-use or already deleted inode. The worst that
1421 * can happen in this case is that we get a "bit already cleared" message from
1422 * ext3_free_inode(). The only reason we would point at a wrong inode is if
1423 * e2fsck was run on this filesystem, and it must have already done the orphan
1424 * inode cleanup for us, so we can safely abort without any further action.
1425 */
1426static void ext3_orphan_cleanup (struct super_block * sb,
1427 struct ext3_super_block * es)
1428{
1429 unsigned int s_flags = sb->s_flags;
1430 int nr_orphans = 0, nr_truncates = 0;
1431#ifdef CONFIG_QUOTA
1432 int i;
1433#endif
1434 if (!es->s_last_orphan) {
1435 jbd_debug(4, "no orphan inodes to clean up\n");
1436 return;
1437 }
1438
a8f48a95 1439 if (bdev_read_only(sb->s_bdev)) {
4cf46b67
AF		 1440 ext3_msg(sb, KERN_ERR, "error: write access "
1441 "unavailable, skipping orphan cleanup.");
a8f48a95
ES		 1442 return;
1443 }
1444
1da177e4
LT		 1445 if (EXT3_SB(sb)->s_mount_state & EXT3_ERROR_FS) {
1446 if (es->s_last_orphan)
1447 jbd_debug(1, "Errors on filesystem, "
1448 "clearing orphan list.\n");
1449 es->s_last_orphan = 0;
1450 jbd_debug(1, "Skipping orphan recovery on fs with errors.\n");
1451 return;
1452 }
1453
1454 if (s_flags & MS_RDONLY) {
4cf46b67 1455 ext3_msg(sb, KERN_INFO, "orphan cleanup on readonly fs");
1da177e4
LT		 1456 sb->s_flags &= ~MS_RDONLY;
1457 }
1458#ifdef CONFIG_QUOTA
1459 /* Needed for iput() to work correctly and not trash data */
1460 sb->s_flags |= MS_ACTIVE;
1461 /* Turn on quotas so that they are updated correctly */
1462 for (i = 0; i < MAXQUOTAS; i++) {
1463 if (EXT3_SB(sb)->s_qf_names[i]) {
1464 int ret = ext3_quota_on_mount(sb, i);
1465 if (ret < 0)
4cf46b67
AF		 1466 ext3_msg(sb, KERN_ERR,
1467 "error: cannot turn on journaled "
1468 "quota: %d", ret);
1da177e4
LT		 1469 }
1470 }
1471#endif
1472
1473 while (es->s_last_orphan) {
1474 struct inode *inode;
1475
473043dc
DH		 1476 inode = ext3_orphan_get(sb, le32_to_cpu(es->s_last_orphan));
1477 if (IS_ERR(inode)) {
1da177e4
LT		 1478 es->s_last_orphan = 0;
1479 break;
1480 }
1481
1482 list_add(&EXT3_I(inode)->i_orphan, &EXT3_SB(sb)->s_orphan);
81a05227 1483 vfs_dq_init(inode);
1da177e4
LT		 1484 if (inode->i_nlink) {
1485 printk(KERN_DEBUG
eee194e7 1486 "%s: truncating inode %lu to %Ld bytes\n",
e05b6b52 1487 __func__, inode->i_ino, inode->i_size);
eee194e7 1488 jbd_debug(2, "truncating inode %lu to %Ld bytes\n",
1da177e4
LT		 1489 inode->i_ino, inode->i_size);
1490 ext3_truncate(inode);
1491 nr_truncates++;
1492 } else {
1493 printk(KERN_DEBUG
eee194e7 1494 "%s: deleting unreferenced inode %lu\n",
e05b6b52 1495 __func__, inode->i_ino);
eee194e7 1496 jbd_debug(2, "deleting unreferenced inode %lu\n",
1da177e4
LT		 1497 inode->i_ino);
1498 nr_orphans++;
1499 }
1500 iput(inode); /* The delete magic happens here! */
1501 }
1502
1503#define PLURAL(x) (x), ((x)==1) ? "" : "s"
1504
1505 if (nr_orphans)
4cf46b67
AF		 1506 ext3_msg(sb, KERN_INFO, "%d orphan inode%s deleted",
1507 PLURAL(nr_orphans));
1da177e4 1508 if (nr_truncates)
4cf46b67
AF		 1509 ext3_msg(sb, KERN_INFO, "%d truncate%s cleaned up",
1510 PLURAL(nr_truncates));
1da177e4
LT		 1511#ifdef CONFIG_QUOTA
1512 /* Turn quotas off */
1513 for (i = 0; i < MAXQUOTAS; i++) {
1514 if (sb_dqopt(sb)->files[i])
2fd83a4f 1515 vfs_quota_off(sb, i, 0);
1da177e4
LT		 1516 }
1517#endif
1518 sb->s_flags = s_flags; /* Restore MS_RDONLY status */
1519}
1520
1da177e4
LT		 1521/*
1522 * Maximal file size. There is a direct, and {,double-,triple-}indirect
1523 * block limit, and also a limit of (2^32 - 1) 512-byte sectors in i_blocks.
1524 * We need to be 1 filesystem block less than the 2^32 sector limit.
1525 */
1526static loff_t ext3_max_size(int bits)
1527{
1528 loff_t res = EXT3_NDIR_BLOCKS;
fe7fdc37
AK		 1529 int meta_blocks;
1530 loff_t upper_limit;
1531
1532 /* This is calculated to be the largest file size for a
1533 * dense, file such that the total number of
1da177e4 1534 * sectors in the file, including data and all indirect blocks,
fe7fdc37
AK		 1535 * does not exceed 2^32 -1
1536 * __u32 i_blocks representing the total number of
1537 * 512 bytes blocks of the file
1538 */
1539 upper_limit = (1LL << 32) - 1;
1540
1541 /* total blocks in file system block size */
1542 upper_limit >>= (bits - 9);
1543
1544
1545 /* indirect blocks */
1546 meta_blocks = 1;
1547 /* double indirect blocks */
1548 meta_blocks += 1 + (1LL << (bits-2));
1549 /* tripple indirect blocks */
1550 meta_blocks += 1 + (1LL << (bits-2)) + (1LL << (2*(bits-2)));
1551
1552 upper_limit -= meta_blocks;
1553 upper_limit <<= bits;
1da177e4
LT		 1554
1555 res += 1LL << (bits-2);
1556 res += 1LL << (2*(bits-2));
1557 res += 1LL << (3*(bits-2));
1558 res <<= bits;
1559 if (res > upper_limit)
1560 res = upper_limit;
fe7fdc37
AK		 1561
1562 if (res > MAX_LFS_FILESIZE)
1563 res = MAX_LFS_FILESIZE;
1564
1da177e4
LT		 1565 return res;
1566}
1567
43d23f90
MC		 1568static ext3_fsblk_t descriptor_loc(struct super_block *sb,
1569 ext3_fsblk_t logic_sb_block,
1da177e4
LT		 1570 int nr)
1571{
1572 struct ext3_sb_info *sbi = EXT3_SB(sb);
43d23f90 1573 unsigned long bg, first_meta_bg;
1da177e4
LT		 1574 int has_super = 0;
1575
1da177e4
LT		 1576 first_meta_bg = le32_to_cpu(sbi->s_es->s_first_meta_bg);
1577
1578 if (!EXT3_HAS_INCOMPAT_FEATURE(sb, EXT3_FEATURE_INCOMPAT_META_BG) ||
1579 nr < first_meta_bg)
1580 return (logic_sb_block + nr + 1);
1581 bg = sbi->s_desc_per_block * nr;
1582 if (ext3_bg_has_super(sb, bg))
1583 has_super = 1;
43d23f90 1584 return (has_super + ext3_group_first_block_no(sb, bg));
1da177e4
LT		 1585}
1586
1587
1588static int ext3_fill_super (struct super_block *sb, void *data, int silent)
1589{
1590 struct buffer_head * bh;
1591 struct ext3_super_block *es = NULL;
1592 struct ext3_sb_info *sbi;
43d23f90 1593 ext3_fsblk_t block;
4cf46b67 1594 ext3_fsblk_t sb_block = get_sb_block(&data, sb);
43d23f90 1595 ext3_fsblk_t logic_sb_block;
1da177e4 1596 unsigned long offset = 0;
eee194e7 1597 unsigned int journal_inum = 0;
71b96257 1598 unsigned long journal_devnum = 0;
1da177e4
LT		 1599 unsigned long def_mount_opts;
1600 struct inode *root;
1601 int blocksize;
1602 int hblock;
1603 int db_count;
1604 int i;
1605 int needs_recovery;
473043dc 1606 int ret = -EINVAL;
1da177e4 1607 __le32 features;
833f4077 1608 int err;
1da177e4 1609
f8314dc6 1610 sbi = kzalloc(sizeof(*sbi), GFP_KERNEL);
1da177e4
LT		 1611 if (!sbi)
1612 return -ENOMEM;
5df096d6
PE		 1613
1614 sbi->s_blockgroup_lock =
1615 kzalloc(sizeof(struct blockgroup_lock), GFP_KERNEL);
1616 if (!sbi->s_blockgroup_lock) {
1617 kfree(sbi);
1618 return -ENOMEM;
1619 }
1da177e4 1620 sb->s_fs_info = sbi;
1da177e4
LT		 1621 sbi->s_mount_opt = 0;
1622 sbi->s_resuid = EXT3_DEF_RESUID;
1623 sbi->s_resgid = EXT3_DEF_RESGID;
571beed8 1624 sbi->s_sb_block = sb_block;
1da177e4
LT		 1625
1626 unlock_kernel();
1627
1628 blocksize = sb_min_blocksize(sb, EXT3_MIN_BLOCK_SIZE);
1629 if (!blocksize) {
4cf46b67 1630 ext3_msg(sb, KERN_ERR, "error: unable to set blocksize");
1da177e4
LT		 1631 goto out_fail;
1632 }
1633
1634 /*
1635 * The ext3 superblock will not be buffer aligned for other than 1kB
1636 * block sizes. We need to calculate the offset from buffer start.
1637 */
1638 if (blocksize != EXT3_MIN_BLOCK_SIZE) {
1639 logic_sb_block = (sb_block * EXT3_MIN_BLOCK_SIZE) / blocksize;
1640 offset = (sb_block * EXT3_MIN_BLOCK_SIZE) % blocksize;
1641 } else {
1642 logic_sb_block = sb_block;
1643 }
1644
1645 if (!(bh = sb_bread(sb, logic_sb_block))) {
4cf46b67 1646 ext3_msg(sb, KERN_ERR, "error: unable to read superblock");
1da177e4
LT		 1647 goto out_fail;
1648 }
1649 /*
1650 * Note: s_es must be initialized as soon as possible because
1651 * some ext3 macro-instructions depend on its value
1652 */
1653 es = (struct ext3_super_block *) (((char *)bh->b_data) + offset);
1654 sbi->s_es = es;
1655 sb->s_magic = le16_to_cpu(es->s_magic);
1656 if (sb->s_magic != EXT3_SUPER_MAGIC)
1657 goto cantfind_ext3;
1658
1659 /* Set defaults before we parse the mount options */
1660 def_mount_opts = le32_to_cpu(es->s_default_mount_opts);
1661 if (def_mount_opts & EXT3_DEFM_DEBUG)
1662 set_opt(sbi->s_mount_opt, DEBUG);
1663 if (def_mount_opts & EXT3_DEFM_BSDGROUPS)
1664 set_opt(sbi->s_mount_opt, GRPID);
1665 if (def_mount_opts & EXT3_DEFM_UID16)
1666 set_opt(sbi->s_mount_opt, NO_UID32);
2e7842b8 1667#ifdef CONFIG_EXT3_FS_XATTR
1da177e4
LT		 1668 if (def_mount_opts & EXT3_DEFM_XATTR_USER)
1669 set_opt(sbi->s_mount_opt, XATTR_USER);
2e7842b8
HD		 1670#endif
1671#ifdef CONFIG_EXT3_FS_POSIX_ACL
1da177e4
LT		 1672 if (def_mount_opts & EXT3_DEFM_ACL)
1673 set_opt(sbi->s_mount_opt, POSIX_ACL);
2e7842b8 1674#endif
1da177e4 1675 if ((def_mount_opts & EXT3_DEFM_JMODE) == EXT3_DEFM_JMODE_DATA)
e3c96435 1676 set_opt(sbi->s_mount_opt, JOURNAL_DATA);
1da177e4 1677 else if ((def_mount_opts & EXT3_DEFM_JMODE) == EXT3_DEFM_JMODE_ORDERED)
e3c96435 1678 set_opt(sbi->s_mount_opt, ORDERED_DATA);
1da177e4 1679 else if ((def_mount_opts & EXT3_DEFM_JMODE) == EXT3_DEFM_JMODE_WBACK)
e3c96435 1680 set_opt(sbi->s_mount_opt, WRITEBACK_DATA);
1da177e4
LT		 1681
1682 if (le16_to_cpu(sbi->s_es->s_errors) == EXT3_ERRORS_PANIC)
1683 set_opt(sbi->s_mount_opt, ERRORS_PANIC);
1eca93f9 1684 else if (le16_to_cpu(sbi->s_es->s_errors) == EXT3_ERRORS_CONTINUE)
2245d7c2 1685 set_opt(sbi->s_mount_opt, ERRORS_CONT);
1eca93f9
AK		 1686 else
1687 set_opt(sbi->s_mount_opt, ERRORS_RO);
1da177e4
LT		 1688
1689 sbi->s_resuid = le16_to_cpu(es->s_def_resuid);
1690 sbi->s_resgid = le16_to_cpu(es->s_def_resgid);
1691
1692 set_opt(sbi->s_mount_opt, RESERVATION);
1693
71b96257
JL		 1694 if (!parse_options ((char *) data, sb, &journal_inum, &journal_devnum,
1695 NULL, 0))
1da177e4
LT		 1696 goto failed_mount;
1697
1698 sb->s_flags = (sb->s_flags & ~MS_POSIXACL) |
e3c96435 1699 (test_opt(sb, POSIX_ACL) ? MS_POSIXACL : 0);
1da177e4
LT		 1700
1701 if (le32_to_cpu(es->s_rev_level) == EXT3_GOOD_OLD_REV &&
1702 (EXT3_HAS_COMPAT_FEATURE(sb, ~0U) ||
1703 EXT3_HAS_RO_COMPAT_FEATURE(sb, ~0U) ||
1704 EXT3_HAS_INCOMPAT_FEATURE(sb, ~0U)))
4cf46b67
AF		 1705 ext3_msg(sb, KERN_WARNING,
1706 "warning: feature flags set on rev 0 fs, "
1707 "running e2fsck is recommended");
1da177e4
LT		 1708 /*
1709 * Check feature flags regardless of the revision level, since we
1710 * previously didn't change the revision level when setting the flags,
1711 * so there is a chance incompat flags are set on a rev 0 filesystem.
1712 */
1713 features = EXT3_HAS_INCOMPAT_FEATURE(sb, ~EXT3_FEATURE_INCOMPAT_SUPP);
1714 if (features) {
4cf46b67
AF		 1715 ext3_msg(sb, KERN_ERR,
1716 "error: couldn't mount because of unsupported "
1717 "optional features (%x)", le32_to_cpu(features));
1da177e4
LT		 1718 goto failed_mount;
1719 }
1720 features = EXT3_HAS_RO_COMPAT_FEATURE(sb, ~EXT3_FEATURE_RO_COMPAT_SUPP);
1721 if (!(sb->s_flags & MS_RDONLY) && features) {
4cf46b67
AF		 1722 ext3_msg(sb, KERN_ERR,
1723 "error: couldn't mount RDWR because of unsupported "
1724 "optional features (%x)", le32_to_cpu(features));
1da177e4
LT		 1725 goto failed_mount;
1726 }
1727 blocksize = BLOCK_SIZE << le32_to_cpu(es->s_log_block_size);
1728
1729 if (blocksize < EXT3_MIN_BLOCK_SIZE ||
1730 blocksize > EXT3_MAX_BLOCK_SIZE) {
4cf46b67
AF		 1731 ext3_msg(sb, KERN_ERR,
1732 "error: couldn't mount because of unsupported "
1733 "filesystem blocksize %d", blocksize);
1da177e4
LT		 1734 goto failed_mount;
1735 }
1736
e1defc4f 1737 hblock = bdev_logical_block_size(sb->s_bdev);
1da177e4
LT		 1738 if (sb->s_blocksize != blocksize) {
1739 /*
1740 * Make sure the blocksize for the filesystem is larger
1741 * than the hardware sectorsize for the machine.
1742 */
1743 if (blocksize < hblock) {
4cf46b67
AF		 1744 ext3_msg(sb, KERN_ERR,
1745 "error: fsblocksize %d too small for "
1746 "hardware sectorsize %d", blocksize, hblock);
1da177e4
LT		 1747 goto failed_mount;
1748 }
1749
1750 brelse (bh);
0f0a89eb 1751 if (!sb_set_blocksize(sb, blocksize)) {
4cf46b67
AF		 1752 ext3_msg(sb, KERN_ERR,
1753 "error: bad blocksize %d", blocksize);
0f0a89eb
TS		 1754 goto out_fail;
1755 }
1da177e4
LT		 1756 logic_sb_block = (sb_block * EXT3_MIN_BLOCK_SIZE) / blocksize;
1757 offset = (sb_block * EXT3_MIN_BLOCK_SIZE) % blocksize;
1758 bh = sb_bread(sb, logic_sb_block);
1759 if (!bh) {
4cf46b67
AF		 1760 ext3_msg(sb, KERN_ERR,
1761 "error: can't read superblock on 2nd try");
1da177e4
LT		 1762 goto failed_mount;
1763 }
1764 es = (struct ext3_super_block *)(((char *)bh->b_data) + offset);
1765 sbi->s_es = es;
1766 if (es->s_magic != cpu_to_le16(EXT3_SUPER_MAGIC)) {
4cf46b67
AF		 1767 ext3_msg(sb, KERN_ERR,
1768 "error: magic mismatch");
1da177e4
LT		 1769 goto failed_mount;
1770 }
1771 }
1772
1773 sb->s_maxbytes = ext3_max_size(sb->s_blocksize_bits);
1774
1775 if (le32_to_cpu(es->s_rev_level) == EXT3_GOOD_OLD_REV) {
1776 sbi->s_inode_size = EXT3_GOOD_OLD_INODE_SIZE;
1777 sbi->s_first_ino = EXT3_GOOD_OLD_FIRST_INO;
1778 } else {
1779 sbi->s_inode_size = le16_to_cpu(es->s_inode_size);
1780 sbi->s_first_ino = le32_to_cpu(es->s_first_ino);
1781 if ((sbi->s_inode_size < EXT3_GOOD_OLD_INODE_SIZE) ||
3fc74269 1782 (!is_power_of_2(sbi->s_inode_size)) ||
1da177e4 1783 (sbi->s_inode_size > blocksize)) {
4cf46b67
AF		 1784 ext3_msg(sb, KERN_ERR,
1785 "error: unsupported inode size: %d",
1da177e4
LT		 1786 sbi->s_inode_size);
1787 goto failed_mount;
1788 }
1789 }
1790 sbi->s_frag_size = EXT3_MIN_FRAG_SIZE <<
1791 le32_to_cpu(es->s_log_frag_size);
1792 if (blocksize != sbi->s_frag_size) {
4cf46b67
AF		 1793 ext3_msg(sb, KERN_ERR,
1794 "error: fragsize %lu != blocksize %u (unsupported)",
1da177e4
LT		 1795 sbi->s_frag_size, blocksize);
1796 goto failed_mount;
1797 }
1798 sbi->s_frags_per_block = 1;
1799 sbi->s_blocks_per_group = le32_to_cpu(es->s_blocks_per_group);
1800 sbi->s_frags_per_group = le32_to_cpu(es->s_frags_per_group);
1801 sbi->s_inodes_per_group = le32_to_cpu(es->s_inodes_per_group);
b47b6f38 1802 if (EXT3_INODE_SIZE(sb) == 0 || EXT3_INODES_PER_GROUP(sb) == 0)
1da177e4
LT		 1803 goto cantfind_ext3;
1804 sbi->s_inodes_per_block = blocksize / EXT3_INODE_SIZE(sb);
1805 if (sbi->s_inodes_per_block == 0)
1806 goto cantfind_ext3;
1807 sbi->s_itb_per_group = sbi->s_inodes_per_group /
1808 sbi->s_inodes_per_block;
1809 sbi->s_desc_per_block = blocksize / sizeof(struct ext3_group_desc);
1810 sbi->s_sbh = bh;
1811 sbi->s_mount_state = le16_to_cpu(es->s_state);
f0d1b0b3
DH		 1812 sbi->s_addr_per_block_bits = ilog2(EXT3_ADDR_PER_BLOCK(sb));
1813 sbi->s_desc_per_block_bits = ilog2(EXT3_DESC_PER_BLOCK(sb));
1da177e4
LT		 1814 for (i=0; i < 4; i++)
1815 sbi->s_hash_seed[i] = le32_to_cpu(es->s_hash_seed[i]);
1816 sbi->s_def_hash_version = es->s_def_hash_version;
5e1f8c9e
TT		 1817 i = le32_to_cpu(es->s_flags);
1818 if (i & EXT2_FLAGS_UNSIGNED_HASH)
1819 sbi->s_hash_unsigned = 3;
1820 else if ((i & EXT2_FLAGS_SIGNED_HASH) == 0) {
1821#ifdef __CHAR_UNSIGNED__
1822 es->s_flags |= cpu_to_le32(EXT2_FLAGS_UNSIGNED_HASH);
1823 sbi->s_hash_unsigned = 3;
1824#else
1825 es->s_flags |= cpu_to_le32(EXT2_FLAGS_SIGNED_HASH);
1826#endif
5e1f8c9e 1827 }
1da177e4
LT		 1828
1829 if (sbi->s_blocks_per_group > blocksize * 8) {
4cf46b67
AF		 1830 ext3_msg(sb, KERN_ERR,
1831 "#blocks per group too big: %lu",
1da177e4
LT		 1832 sbi->s_blocks_per_group);
1833 goto failed_mount;
1834 }
1835 if (sbi->s_frags_per_group > blocksize * 8) {
4cf46b67
AF		 1836 ext3_msg(sb, KERN_ERR,
1837 "error: #fragments per group too big: %lu",
1da177e4
LT		 1838 sbi->s_frags_per_group);
1839 goto failed_mount;
1840 }
1841 if (sbi->s_inodes_per_group > blocksize * 8) {
4cf46b67
AF		 1842 ext3_msg(sb, KERN_ERR,
1843 "error: #inodes per group too big: %lu",
1da177e4
LT		 1844 sbi->s_inodes_per_group);
1845 goto failed_mount;
1846 }
1847
fcd5df35
MC		 1848 if (le32_to_cpu(es->s_blocks_count) >
1849 (sector_t)(~0ULL) >> (sb->s_blocksize_bits - 9)) {
4cf46b67
AF		 1850 ext3_msg(sb, KERN_ERR,
1851 "error: filesystem is too large to mount safely");
fcd5df35 1852 if (sizeof(sector_t) < 8)
4cf46b67
AF		 1853 ext3_msg(sb, KERN_ERR,
1854 "error: CONFIG_LBDAF not enabled");
fcd5df35
MC		 1855 goto failed_mount;
1856 }
1857
1da177e4
LT		 1858 if (EXT3_BLOCKS_PER_GROUP(sb) == 0)
1859 goto cantfind_ext3;
855565e8
ES		 1860 sbi->s_groups_count = ((le32_to_cpu(es->s_blocks_count) -
1861 le32_to_cpu(es->s_first_data_block) - 1)
1862 / EXT3_BLOCKS_PER_GROUP(sb)) + 1;
1da177e4
LT		 1863 db_count = (sbi->s_groups_count + EXT3_DESC_PER_BLOCK(sb) - 1) /
1864 EXT3_DESC_PER_BLOCK(sb);
1865 sbi->s_group_desc = kmalloc(db_count * sizeof (struct buffer_head *),
1866 GFP_KERNEL);
1867 if (sbi->s_group_desc == NULL) {
4cf46b67
AF		 1868 ext3_msg(sb, KERN_ERR,
1869 "error: not enough memory");
1da177e4
LT		 1870 goto failed_mount;
1871 }
1872
5df096d6 1873 bgl_lock_init(sbi->s_blockgroup_lock);
1da177e4
LT		 1874
1875 for (i = 0; i < db_count; i++) {
1876 block = descriptor_loc(sb, logic_sb_block, i);
1877 sbi->s_group_desc[i] = sb_bread(sb, block);
1878 if (!sbi->s_group_desc[i]) {
4cf46b67
AF		 1879 ext3_msg(sb, KERN_ERR,
1880 "error: can't read group descriptor %d", i);
1da177e4
LT		 1881 db_count = i;
1882 goto failed_mount2;
1883 }
1884 }
1885 if (!ext3_check_descriptors (sb)) {
4cf46b67
AF		 1886 ext3_msg(sb, KERN_ERR,
1887 "error: group descriptors corrupted");
1da177e4
LT		 1888 goto failed_mount2;
1889 }
1890 sbi->s_gdb_count = db_count;
1891 get_random_bytes(&sbi->s_next_generation, sizeof(u32));
1892 spin_lock_init(&sbi->s_next_gen_lock);
0216bfcf 1893
833f4077
PZ		 1894 err = percpu_counter_init(&sbi->s_freeblocks_counter,
1895 ext3_count_free_blocks(sb));
1896 if (!err) {
1897 err = percpu_counter_init(&sbi->s_freeinodes_counter,
1898 ext3_count_free_inodes(sb));
1899 }
1900 if (!err) {
1901 err = percpu_counter_init(&sbi->s_dirs_counter,
1902 ext3_count_dirs(sb));
1903 }
1904 if (err) {
4cf46b67 1905 ext3_msg(sb, KERN_ERR, "error: insufficient memory");
833f4077
PZ		 1906 goto failed_mount3;
1907 }
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))
0216bfcf 1949 goto failed_mount3;
1da177e4
LT		 1950 } else if (journal_inum) {
1951 if (ext3_create_journal(sb, es, journal_inum))
0216bfcf 1952 goto failed_mount3;
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?");
0216bfcf 1958 goto failed_mount3;
1da177e4
LT		 1959 }
1960
1961 /* We have now updated the journal if required, so we can
1962 * validate the data journaling mode. */
1963 switch (test_opt(sb, DATA_FLAGS)) {
1964 case 0:
1965 /* No mode set, assume a default based on the journal
1966 capabilities: ORDERED_DATA if the journal can
1967 cope, else JOURNAL_DATA */
1968 if (journal_check_available_features
1969 (sbi->s_journal, 0, 0, JFS_FEATURE_INCOMPAT_REVOKE))
bbae8bcc 1970 set_opt(sbi->s_mount_opt, DEFAULT_DATA_MODE);
1da177e4
LT		 1971 else
1972 set_opt(sbi->s_mount_opt, JOURNAL_DATA);
1973 break;
1974
1975 case EXT3_MOUNT_ORDERED_DATA:
1976 case EXT3_MOUNT_WRITEBACK_DATA:
1977 if (!journal_check_available_features
1978 (sbi->s_journal, 0, 0, JFS_FEATURE_INCOMPAT_REVOKE)) {
4cf46b67
AF		 1979 ext3_msg(sb, KERN_ERR,
1980 "error: journal does not support "
1981 "requested data journaling mode");
0216bfcf 1982 goto failed_mount4;
1da177e4
LT		 1983 }
1984 default:
1985 break;
1986 }
1987
1988 if (test_opt(sb, NOBH)) {
1da177e4 1989 if (!(test_opt(sb, DATA_FLAGS) == EXT3_MOUNT_WRITEBACK_DATA)) {
4cf46b67
AF		 1990 ext3_msg(sb, KERN_WARNING,
1991 "warning: ignoring nobh option - "
1992 "it is supported only with writeback mode");
1da177e4
LT		 1993 clear_opt(sbi->s_mount_opt, NOBH);
1994 }
1995 }
1996 /*
1997 * The journal_load will have done any necessary log recovery,
1998 * so we can safely mount the rest of the filesystem now.
1999 */
2000
473043dc
DH		 2001 root = ext3_iget(sb, EXT3_ROOT_INO);
2002 if (IS_ERR(root)) {
4cf46b67 2003 ext3_msg(sb, KERN_ERR, "error: get root inode failed");
473043dc 2004 ret = PTR_ERR(root);
0216bfcf 2005 goto failed_mount4;
1da177e4
LT		 2006 }
2007 if (!S_ISDIR(root->i_mode) || !root->i_blocks || !root->i_size) {
473043dc 2008 iput(root);
4cf46b67 2009 ext3_msg(sb, KERN_ERR, "error: corrupt root inode, run e2fsck");
0216bfcf 2010 goto failed_mount4;
1da177e4 2011 }
473043dc
DH		 2012 sb->s_root = d_alloc_root(root);
2013 if (!sb->s_root) {
4cf46b67 2014 ext3_msg(sb, KERN_ERR, "error: get root dentry failed");
473043dc
DH		 2015 iput(root);
2016 ret = -ENOMEM;
2017 goto failed_mount4;
2018 }
1da177e4
LT		 2019
2020 ext3_setup_super (sb, es, sb->s_flags & MS_RDONLY);
ed505ee4 2021
1da177e4
LT		 2022 EXT3_SB(sb)->s_mount_state |= EXT3_ORPHAN_FS;
2023 ext3_orphan_cleanup(sb, es);
2024 EXT3_SB(sb)->s_mount_state &= ~EXT3_ORPHAN_FS;
2025 if (needs_recovery)
4cf46b67 2026 ext3_msg(sb, KERN_INFO, "recovery complete");
1da177e4 2027 ext3_mark_recovery_complete(sb, es);
4cf46b67 2028 ext3_msg(sb, KERN_INFO, "mounted filesystem with %s data mode",
1da177e4
LT		 2029 test_opt(sb,DATA_FLAGS) == EXT3_MOUNT_JOURNAL_DATA ? "journal":
2030 test_opt(sb,DATA_FLAGS) == EXT3_MOUNT_ORDERED_DATA ? "ordered":
2031 "writeback");
2032
1da177e4
LT		 2033 lock_kernel();
2034 return 0;
2035
2036cantfind_ext3:
2037 if (!silent)
4cf46b67
AF		 2038 ext3_msg(sb, KERN_INFO,
2039 "error: can't find ext3 filesystem on dev %s.",
1da177e4
LT		 2040 sb->s_id);
2041 goto failed_mount;
2042
0216bfcf 2043failed_mount4:
1da177e4 2044 journal_destroy(sbi->s_journal);
0216bfcf
MC		 2045failed_mount3:
2046 percpu_counter_destroy(&sbi->s_freeblocks_counter);
2047 percpu_counter_destroy(&sbi->s_freeinodes_counter);
2048 percpu_counter_destroy(&sbi->s_dirs_counter);
1da177e4
LT		 2049failed_mount2:
2050 for (i = 0; i < db_count; i++)
2051 brelse(sbi->s_group_desc[i]);
2052 kfree(sbi->s_group_desc);
2053failed_mount:
2054#ifdef CONFIG_QUOTA
2055 for (i = 0; i < MAXQUOTAS; i++)
2056 kfree(sbi->s_qf_names[i]);
2057#endif
2058 ext3_blkdev_remove(sbi);
2059 brelse(bh);
2060out_fail:
2061 sb->s_fs_info = NULL;
de5ce037 2062 kfree(sbi->s_blockgroup_lock);
1da177e4
LT		 2063 kfree(sbi);
2064 lock_kernel();
473043dc 2065 return ret;
1da177e4
LT		 2066}
2067
2068/*
2069 * Setup any per-fs journal parameters now. We'll do this both on
2070 * initial mount, once the journal has been initialised but before we've
ae6ddcc5 2071 * done any recovery; and again on any subsequent remount.
1da177e4
LT		 2072 */
2073static void ext3_init_journal_params(struct super_block *sb, journal_t *journal)
2074{
2075 struct ext3_sb_info *sbi = EXT3_SB(sb);
2076
2077 if (sbi->s_commit_interval)
2078 journal->j_commit_interval = sbi->s_commit_interval;
2079 /* We could also set up an ext3-specific default for the commit
2080 * interval here, but for now we'll just fall back to the jbd
2081 * default. */
2082
2083 spin_lock(&journal->j_state_lock);
2084 if (test_opt(sb, BARRIER))
2085 journal->j_flags |= JFS_BARRIER;
2086 else
2087 journal->j_flags &= ~JFS_BARRIER;
0e4fb5e2
HK		 2088 if (test_opt(sb, DATA_ERR_ABORT))
2089 journal->j_flags |= JFS_ABORT_ON_SYNCDATA_ERR;
2090 else
2091 journal->j_flags &= ~JFS_ABORT_ON_SYNCDATA_ERR;
1da177e4
LT		 2092 spin_unlock(&journal->j_state_lock);
2093}
2094
eee194e7
ES		 2095static journal_t *ext3_get_journal(struct super_block *sb,
2096 unsigned int journal_inum)
1da177e4
LT		 2097{
2098 struct inode *journal_inode;
2099 journal_t *journal;
2100
2101 /* First, test for the existence of a valid inode on disk. Bad
2102 * things happen if we iget() an unused inode, as the subsequent
2103 * iput() will try to delete it. */
2104
473043dc
DH		 2105 journal_inode = ext3_iget(sb, journal_inum);
2106 if (IS_ERR(journal_inode)) {
4cf46b67 2107 ext3_msg(sb, KERN_ERR, "error: no journal found");
1da177e4
LT		 2108 return NULL;
2109 }
2110 if (!journal_inode->i_nlink) {
2111 make_bad_inode(journal_inode);
2112 iput(journal_inode);
4cf46b67 2113 ext3_msg(sb, KERN_ERR, "error: journal inode is deleted");
1da177e4
LT		 2114 return NULL;
2115 }
2116
2117 jbd_debug(2, "Journal inode found at %p: %Ld bytes\n",
2118 journal_inode, journal_inode->i_size);
473043dc 2119 if (!S_ISREG(journal_inode->i_mode)) {
4cf46b67 2120 ext3_msg(sb, KERN_ERR, "error: invalid journal inode");
1da177e4
LT		 2121 iput(journal_inode);
2122 return NULL;
2123 }
2124
2125 journal = journal_init_inode(journal_inode);
2126 if (!journal) {
4cf46b67 2127 ext3_msg(sb, KERN_ERR, "error: could not load journal inode");
1da177e4
LT		 2128 iput(journal_inode);
2129 return NULL;
2130 }
2131 journal->j_private = sb;
2132 ext3_init_journal_params(sb, journal);
2133 return journal;
2134}
2135
2136static journal_t *ext3_get_dev_journal(struct super_block *sb,
2137 dev_t j_dev)
2138{
2139 struct buffer_head * bh;
2140 journal_t *journal;
43d23f90 2141 ext3_fsblk_t start;
1c2bf374 2142 ext3_fsblk_t len;
1da177e4 2143 int hblock, blocksize;
43d23f90 2144 ext3_fsblk_t sb_block;
1da177e4
LT		 2145 unsigned long offset;
2146 struct ext3_super_block * es;
2147 struct block_device *bdev;
2148
4cf46b67 2149 bdev = ext3_blkdev_get(j_dev, sb);
1da177e4
LT		 2150 if (bdev == NULL)
2151 return NULL;
2152
2153 if (bd_claim(bdev, sb)) {
4cf46b67
AF		 2154 ext3_msg(sb, KERN_ERR,
2155 "error: failed to claim external journal device");
9a1c3542 2156 blkdev_put(bdev, FMODE_READ|FMODE_WRITE);
1da177e4
LT		 2157 return NULL;
2158 }
2159
2160 blocksize = sb->s_blocksize;
e1defc4f 2161 hblock = bdev_logical_block_size(bdev);
1da177e4 2162 if (blocksize < hblock) {
4cf46b67
AF		 2163 ext3_msg(sb, KERN_ERR,
2164 "error: blocksize too small for journal device");
1da177e4
LT		 2165 goto out_bdev;
2166 }
2167
2168 sb_block = EXT3_MIN_BLOCK_SIZE / blocksize;
2169 offset = EXT3_MIN_BLOCK_SIZE % blocksize;
2170 set_blocksize(bdev, blocksize);
2171 if (!(bh = __bread(bdev, sb_block, blocksize))) {
4cf46b67
AF		 2172 ext3_msg(sb, KERN_ERR, "error: couldn't read superblock of "
2173 "external journal");
1da177e4
LT		 2174 goto out_bdev;
2175 }
2176
2177 es = (struct ext3_super_block *) (((char *)bh->b_data) + offset);
2178 if ((le16_to_cpu(es->s_magic) != EXT3_SUPER_MAGIC) ||
2179 !(le32_to_cpu(es->s_feature_incompat) &
2180 EXT3_FEATURE_INCOMPAT_JOURNAL_DEV)) {
4cf46b67
AF		 2181 ext3_msg(sb, KERN_ERR, "error: external journal has "
2182 "bad superblock");
1da177e4
LT		 2183 brelse(bh);
2184 goto out_bdev;
2185 }
2186
2187 if (memcmp(EXT3_SB(sb)->s_es->s_journal_uuid, es->s_uuid, 16)) {
4cf46b67 2188 ext3_msg(sb, KERN_ERR, "error: journal UUID does not match");
1da177e4
LT		 2189 brelse(bh);
2190 goto out_bdev;
2191 }
2192
2193 len = le32_to_cpu(es->s_blocks_count);
2194 start = sb_block + 1;
2195 brelse(bh); /* we're done with the superblock */
2196
2197 journal = journal_init_dev(bdev, sb->s_bdev,
2198 start, len, blocksize);
2199 if (!journal) {
4cf46b67
AF		 2200 ext3_msg(sb, KERN_ERR,
2201 "error: failed to create device journal");
1da177e4
LT		 2202 goto out_bdev;
2203 }
2204 journal->j_private = sb;
2205 ll_rw_block(READ, 1, &journal->j_sb_buffer);
2206 wait_on_buffer(journal->j_sb_buffer);
2207 if (!buffer_uptodate(journal->j_sb_buffer)) {
4cf46b67 2208 ext3_msg(sb, KERN_ERR, "I/O error on journal device");
1da177e4
LT		 2209 goto out_journal;
2210 }
2211 if (be32_to_cpu(journal->j_superblock->s_nr_users) != 1) {
4cf46b67
AF		 2212 ext3_msg(sb, KERN_ERR,
2213 "error: external journal has more than one "
2214 "user (unsupported) - %d",
1da177e4
LT		 2215 be32_to_cpu(journal->j_superblock->s_nr_users));
2216 goto out_journal;
2217 }
2218 EXT3_SB(sb)->journal_bdev = bdev;
2219 ext3_init_journal_params(sb, journal);
2220 return journal;
2221out_journal:
2222 journal_destroy(journal);
2223out_bdev:
2224 ext3_blkdev_put(bdev);
2225 return NULL;
2226}
2227
71b96257
JL		 2228static int ext3_load_journal(struct super_block *sb,
2229 struct ext3_super_block *es,
2230 unsigned long journal_devnum)
1da177e4
LT		 2231{
2232 journal_t *journal;
eee194e7 2233 unsigned int journal_inum = le32_to_cpu(es->s_journal_inum);
71b96257 2234 dev_t journal_dev;
1da177e4
LT		 2235 int err = 0;
2236 int really_read_only;
2237
71b96257
JL		 2238 if (journal_devnum &&
2239 journal_devnum != le32_to_cpu(es->s_journal_dev)) {
4cf46b67
AF		 2240 ext3_msg(sb, KERN_INFO, "external journal device major/minor "
2241 "numbers have changed");
71b96257
JL		 2242 journal_dev = new_decode_dev(journal_devnum);
2243 } else
2244 journal_dev = new_decode_dev(le32_to_cpu(es->s_journal_dev));
2245
1da177e4
LT		 2246 really_read_only = bdev_read_only(sb->s_bdev);
2247
2248 /*
2249 * Are we loading a blank journal or performing recovery after a
2250 * crash? For recovery, we need to check in advance whether we
2251 * can get read-write access to the device.
2252 */
2253
2254 if (EXT3_HAS_INCOMPAT_FEATURE(sb, EXT3_FEATURE_INCOMPAT_RECOVER)) {
2255 if (sb->s_flags & MS_RDONLY) {
4cf46b67
AF		 2256 ext3_msg(sb, KERN_INFO,
2257 "recovery required on readonly filesystem");
1da177e4 2258 if (really_read_only) {
4cf46b67
AF		 2259 ext3_msg(sb, KERN_ERR, "error: write access "
2260 "unavailable, cannot proceed");
1da177e4
LT		 2261 return -EROFS;
2262 }
4cf46b67
AF		 2263 ext3_msg(sb, KERN_INFO,
2264 "write access will be enabled during recovery");
1da177e4
LT		 2265 }
2266 }
2267
2268 if (journal_inum && journal_dev) {
4cf46b67
AF		 2269 ext3_msg(sb, KERN_ERR, "error: filesystem has both journal "
2270 "and inode journals");
1da177e4
LT		 2271 return -EINVAL;
2272 }
2273
2274 if (journal_inum) {
2275 if (!(journal = ext3_get_journal(sb, journal_inum)))
2276 return -EINVAL;
2277 } else {
2278 if (!(journal = ext3_get_dev_journal(sb, journal_dev)))
2279 return -EINVAL;
2280 }
2281
2282 if (!really_read_only && test_opt(sb, UPDATE_JOURNAL)) {
2283 err = journal_update_format(journal);
2284 if (err) {
4cf46b67 2285 ext3_msg(sb, KERN_ERR, "error updating journal");
1da177e4
LT		 2286 journal_destroy(journal);
2287 return err;
2288 }
2289 }
2290
2291 if (!EXT3_HAS_INCOMPAT_FEATURE(sb, EXT3_FEATURE_INCOMPAT_RECOVER))
2292 err = journal_wipe(journal, !really_read_only);
2293 if (!err)
2294 err = journal_load(journal);
2295
2296 if (err) {
4cf46b67 2297 ext3_msg(sb, KERN_ERR, "error loading journal");
1da177e4
LT		 2298 journal_destroy(journal);
2299 return err;
2300 }
2301
2302 EXT3_SB(sb)->s_journal = journal;
2303 ext3_clear_journal_err(sb, es);
71b96257
JL		 2304
2305 if (journal_devnum &&
2306 journal_devnum != le32_to_cpu(es->s_journal_dev)) {
2307 es->s_journal_dev = cpu_to_le32(journal_devnum);
71b96257
JL		 2308
2309 /* Make sure we flush the recovery flag to disk. */
2310 ext3_commit_super(sb, es, 1);
2311 }
2312
1da177e4
LT		 2313 return 0;
2314}
2315
4cf46b67
AF		 2316static int ext3_create_journal(struct super_block *sb,
2317 struct ext3_super_block *es,
eee194e7 2318 unsigned int journal_inum)
1da177e4
LT		 2319{
2320 journal_t *journal;
952d9de1 2321 int err;
1da177e4
LT		 2322
2323 if (sb->s_flags & MS_RDONLY) {
4cf46b67
AF		 2324 ext3_msg(sb, KERN_ERR,
2325 "error: readonly filesystem when trying to "
2326 "create journal");
1da177e4
LT		 2327 return -EROFS;
2328 }
2329
952d9de1
BP		 2330 journal = ext3_get_journal(sb, journal_inum);
2331 if (!journal)
1da177e4
LT		 2332 return -EINVAL;
2333
4cf46b67 2334 ext3_msg(sb, KERN_INFO, "creating new journal on inode %u",
1da177e4
LT		 2335 journal_inum);
2336
952d9de1
BP		 2337 err = journal_create(journal);
2338 if (err) {
4cf46b67 2339 ext3_msg(sb, KERN_ERR, "error creating journal");
1da177e4
LT		 2340 journal_destroy(journal);
2341 return -EIO;
2342 }
2343
2344 EXT3_SB(sb)->s_journal = journal;
2345
2346 ext3_update_dynamic_rev(sb);
2347 EXT3_SET_INCOMPAT_FEATURE(sb, EXT3_FEATURE_INCOMPAT_RECOVER);
2348 EXT3_SET_COMPAT_FEATURE(sb, EXT3_FEATURE_COMPAT_HAS_JOURNAL);
2349
2350 es->s_journal_inum = cpu_to_le32(journal_inum);
1da177e4
LT		 2351
2352 /* Make sure we flush the recovery flag to disk. */
2353 ext3_commit_super(sb, es, 1);
2354
2355 return 0;
2356}
2357
c4be0c1d
TS		 2358static int ext3_commit_super(struct super_block *sb,
2359 struct ext3_super_block *es,
1da177e4
LT		 2360 int sync)
2361{
2362 struct buffer_head *sbh = EXT3_SB(sb)->s_sbh;
c4be0c1d 2363 int error = 0;
1da177e4
LT		 2364
2365 if (!sbh)
c4be0c1d 2366 return error;
96ec2e0a
TT		 2367 /*
2368 * If the file system is mounted read-only, don't update the
2369 * superblock write time. This avoids updating the superblock
2370 * write time when we are mounting the root file system
2371 * read/only but we need to replay the journal; at that point,
2372 * for people who are east of GMT and who make their clock
2373 * tick in localtime for Windows bug-for-bug compatibility,
2374 * the clock is set in the future, and this will cause e2fsck
2375 * to complain and force a full file system check.
2376 */
2377 if (!(sb->s_flags & MS_RDONLY))
2378 es->s_wtime = cpu_to_le32(get_seconds());
1da177e4
LT		 2379 es->s_free_blocks_count = cpu_to_le32(ext3_count_free_blocks(sb));
2380 es->s_free_inodes_count = cpu_to_le32(ext3_count_free_inodes(sb));
2381 BUFFER_TRACE(sbh, "marking dirty");
2382 mark_buffer_dirty(sbh);
2383 if (sync)
c4be0c1d
TS		 2384 error = sync_dirty_buffer(sbh);
2385 return error;
1da177e4
LT		 2386}
2387
2388
2389/*
2390 * Have we just finished recovery? If so, and if we are mounting (or
2391 * remounting) the filesystem readonly, then we will end up with a
2392 * consistent fs on disk. Record that fact.
2393 */
2394static void ext3_mark_recovery_complete(struct super_block * sb,
2395 struct ext3_super_block * es)
2396{
2397 journal_t *journal = EXT3_SB(sb)->s_journal;
2398
2399 journal_lock_updates(journal);
2d7c820e
HK		 2400 if (journal_flush(journal) < 0)
2401 goto out;
2402
1da177e4
LT		 2403 if (EXT3_HAS_INCOMPAT_FEATURE(sb, EXT3_FEATURE_INCOMPAT_RECOVER) &&
2404 sb->s_flags & MS_RDONLY) {
2405 EXT3_CLEAR_INCOMPAT_FEATURE(sb, EXT3_FEATURE_INCOMPAT_RECOVER);
1da177e4
LT		 2406 ext3_commit_super(sb, es, 1);
2407 }
2d7c820e
HK		 2408
2409out:
1da177e4
LT		 2410 journal_unlock_updates(journal);
2411}
2412
2413/*
2414 * If we are mounting (or read-write remounting) a filesystem whose journal
2415 * has recorded an error from a previous lifetime, move that error to the
2416 * main filesystem now.
2417 */
4cf46b67
AF		 2418static void ext3_clear_journal_err(struct super_block *sb,
2419 struct ext3_super_block *es)
1da177e4
LT		 2420{
2421 journal_t *journal;
2422 int j_errno;
2423 const char *errstr;
2424
2425 journal = EXT3_SB(sb)->s_journal;
2426
2427 /*
2428 * Now check for any error status which may have been recorded in the
2429 * journal by a prior ext3_error() or ext3_abort()
2430 */
2431
2432 j_errno = journal_errno(journal);
2433 if (j_errno) {
2434 char nbuf[16];
2435
2436 errstr = ext3_decode_error(sb, j_errno, nbuf);
e05b6b52 2437 ext3_warning(sb, __func__, "Filesystem error recorded "
1da177e4 2438 "from previous mount: %s", errstr);
e05b6b52 2439 ext3_warning(sb, __func__, "Marking fs in need of "
1da177e4
LT		 2440 "filesystem check.");
2441
2442 EXT3_SB(sb)->s_mount_state |= EXT3_ERROR_FS;
2443 es->s_state |= cpu_to_le16(EXT3_ERROR_FS);
2444 ext3_commit_super (sb, es, 1);
2445
2446 journal_clear_err(journal);
2447 }
2448}
2449
2450/*
2451 * Force the running and committing transactions to commit,
2452 * and wait on the commit.
2453 */
2454int ext3_force_commit(struct super_block *sb)
2455{
2456 journal_t *journal;
2457 int ret;
2458
2459 if (sb->s_flags & MS_RDONLY)
2460 return 0;
2461
2462 journal = EXT3_SB(sb)->s_journal;
1da177e4
LT		 2463 ret = ext3_journal_force_commit(journal);
2464 return ret;
2465}
2466
1da177e4
LT		 2467static int ext3_sync_fs(struct super_block *sb, int wait)
2468{
02ac597c 2469 tid_t target;
c87591b7 2470
02ac597c
JK		 2471 if (journal_start_commit(EXT3_SB(sb)->s_journal, &target)) {
2472 if (wait)
2473 log_wait_commit(EXT3_SB(sb)->s_journal, target);
2474 }
1da177e4
LT		 2475 return 0;
2476}
2477
2478/*
2479 * LVM calls this function before a (read-only) snapshot is created. This
2480 * gives us a chance to flush the journal completely and mark the fs clean.
2481 */
c4be0c1d 2482static int ext3_freeze(struct super_block *sb)
1da177e4 2483{
c4be0c1d
TS		 2484 int error = 0;
2485 journal_t *journal;
1da177e4
LT		 2486
2487 if (!(sb->s_flags & MS_RDONLY)) {
c4be0c1d 2488 journal = EXT3_SB(sb)->s_journal;
1da177e4
LT		 2489
2490 /* Now we set up the journal barrier. */
2491 journal_lock_updates(journal);
2d7c820e
HK		 2492
2493 /*
2494 * We don't want to clear needs_recovery flag when we failed
2495 * to flush the journal.
2496 */
c4be0c1d
TS		 2497 error = journal_flush(journal);
2498 if (error < 0)
2499 goto out;
1da177e4
LT		 2500
2501 /* Journal blocked and flushed, clear needs_recovery flag. */
2502 EXT3_CLEAR_INCOMPAT_FEATURE(sb, EXT3_FEATURE_INCOMPAT_RECOVER);
c4be0c1d
TS		 2503 error = ext3_commit_super(sb, EXT3_SB(sb)->s_es, 1);
2504 if (error)
2505 goto out;
1da177e4 2506 }
c4be0c1d
TS		 2507 return 0;
2508
2509out:
2510 journal_unlock_updates(journal);
2511 return error;
1da177e4
LT		 2512}
2513
2514/*
2515 * Called by LVM after the snapshot is done. We need to reset the RECOVER
2516 * flag here, even though the filesystem is not technically dirty yet.
2517 */
c4be0c1d 2518static int ext3_unfreeze(struct super_block *sb)
1da177e4
LT		 2519{
2520 if (!(sb->s_flags & MS_RDONLY)) {
2521 lock_super(sb);
2522 /* Reser the needs_recovery flag before the fs is unlocked. */
2523 EXT3_SET_INCOMPAT_FEATURE(sb, EXT3_FEATURE_INCOMPAT_RECOVER);
2524 ext3_commit_super(sb, EXT3_SB(sb)->s_es, 1);
2525 unlock_super(sb);
2526 journal_unlock_updates(EXT3_SB(sb)->s_journal);
2527 }
c4be0c1d 2528 return 0;
1da177e4
LT		 2529}
2530
2531static int ext3_remount (struct super_block * sb, int * flags, char * data)
2532{
2533 struct ext3_super_block * es;
2534 struct ext3_sb_info *sbi = EXT3_SB(sb);
43d23f90 2535 ext3_fsblk_t n_blocks_count = 0;
08c6a96f
JK		 2536 unsigned long old_sb_flags;
2537 struct ext3_mount_options old_opts;
2538 int err;
2539#ifdef CONFIG_QUOTA
2540 int i;
2541#endif
2542
337eb00a
AIB		 2543 lock_kernel();
2544
08c6a96f 2545 /* Store the original options */
bbd6851a 2546 lock_super(sb);
08c6a96f
JK		 2547 old_sb_flags = sb->s_flags;
2548 old_opts.s_mount_opt = sbi->s_mount_opt;
2549 old_opts.s_resuid = sbi->s_resuid;
2550 old_opts.s_resgid = sbi->s_resgid;
2551 old_opts.s_commit_interval = sbi->s_commit_interval;
2552#ifdef CONFIG_QUOTA
2553 old_opts.s_jquota_fmt = sbi->s_jquota_fmt;
2554 for (i = 0; i < MAXQUOTAS; i++)
2555 old_opts.s_qf_names[i] = sbi->s_qf_names[i];
2556#endif
1da177e4
LT		 2557
2558 /*
2559 * Allow the "check" option to be passed as a remount option.
2560 */
71b96257 2561 if (!parse_options(data, sb, NULL, NULL, &n_blocks_count, 1)) {
08c6a96f
JK		 2562 err = -EINVAL;
2563 goto restore_opts;
2564 }
1da177e4 2565
e3c96435 2566 if (test_opt(sb, ABORT))
e05b6b52 2567 ext3_abort(sb, __func__, "Abort forced by user");
1da177e4
LT		 2568
2569 sb->s_flags = (sb->s_flags & ~MS_POSIXACL) |
e3c96435 2570 (test_opt(sb, POSIX_ACL) ? MS_POSIXACL : 0);
1da177e4
LT		 2571
2572 es = sbi->s_es;
2573
2574 ext3_init_journal_params(sb, sbi->s_journal);
2575
2576 if ((*flags & MS_RDONLY) != (sb->s_flags & MS_RDONLY) ||
2577 n_blocks_count > le32_to_cpu(es->s_blocks_count)) {
e3c96435 2578 if (test_opt(sb, ABORT)) {
08c6a96f
JK		 2579 err = -EROFS;
2580 goto restore_opts;
2581 }
1da177e4
LT		 2582
2583 if (*flags & MS_RDONLY) {
2584 /*
2585 * First of all, the unconditional stuff we have to do
2586 * to disable replay of the journal when we next remount
2587 */
2588 sb->s_flags |= MS_RDONLY;
2589
2590 /*
2591 * OK, test if we are remounting a valid rw partition
2592 * readonly, and if so set the rdonly flag and then
2593 * mark the partition as valid again.
2594 */
2595 if (!(es->s_state & cpu_to_le16(EXT3_VALID_FS)) &&
2596 (sbi->s_mount_state & EXT3_VALID_FS))
2597 es->s_state = cpu_to_le16(sbi->s_mount_state);
2598
2599 ext3_mark_recovery_complete(sb, es);
2600 } else {
2601 __le32 ret;
2602 if ((ret = EXT3_HAS_RO_COMPAT_FEATURE(sb,
2603 ~EXT3_FEATURE_RO_COMPAT_SUPP))) {
4cf46b67
AF		 2604 ext3_msg(sb, KERN_WARNING,
2605 "warning: couldn't remount RDWR "
2606 "because of unsupported optional "
2607 "features (%x)", le32_to_cpu(ret));
08c6a96f
JK		 2608 err = -EROFS;
2609 goto restore_opts;
1da177e4 2610 }
ea9a05a1
ES		 2611
2612 /*
2613 * If we have an unprocessed orphan list hanging
2614 * around from a previously readonly bdev mount,
2615 * require a full umount/remount for now.
2616 */
2617 if (es->s_last_orphan) {
4cf46b67 2618 ext3_msg(sb, KERN_WARNING, "warning: couldn't "
ea9a05a1
ES		 2619 "remount RDWR because of unprocessed "
2620 "orphan inode list. Please "
4cf46b67 2621 "umount/remount instead.");
ea9a05a1
ES		 2622 err = -EINVAL;
2623 goto restore_opts;
2624 }
2625
1da177e4
LT		 2626 /*
2627 * Mounting a RDONLY partition read-write, so reread
2628 * and store the current valid flag. (It may have
2629 * been changed by e2fsck since we originally mounted
2630 * the partition.)
2631 */
2632 ext3_clear_journal_err(sb, es);
2633 sbi->s_mount_state = le16_to_cpu(es->s_state);
a4e4de36 2634 if ((err = ext3_group_extend(sb, es, n_blocks_count)))
08c6a96f 2635 goto restore_opts;
1da177e4
LT		 2636 if (!ext3_setup_super (sb, es, 0))
2637 sb->s_flags &= ~MS_RDONLY;
2638 }
2639 }
08c6a96f
JK		 2640#ifdef CONFIG_QUOTA
2641 /* Release old quota file names */
2642 for (i = 0; i < MAXQUOTAS; i++)
2643 if (old_opts.s_qf_names[i] &&
2644 old_opts.s_qf_names[i] != sbi->s_qf_names[i])
2645 kfree(old_opts.s_qf_names[i]);
2646#endif
bbd6851a 2647 unlock_super(sb);
337eb00a 2648 unlock_kernel();
1da177e4 2649 return 0;
08c6a96f
JK		 2650restore_opts:
2651 sb->s_flags = old_sb_flags;
2652 sbi->s_mount_opt = old_opts.s_mount_opt;
2653 sbi->s_resuid = old_opts.s_resuid;
2654 sbi->s_resgid = old_opts.s_resgid;
2655 sbi->s_commit_interval = old_opts.s_commit_interval;
2656#ifdef CONFIG_QUOTA
2657 sbi->s_jquota_fmt = old_opts.s_jquota_fmt;
2658 for (i = 0; i < MAXQUOTAS; i++) {
2659 if (sbi->s_qf_names[i] &&
2660 old_opts.s_qf_names[i] != sbi->s_qf_names[i])
2661 kfree(sbi->s_qf_names[i]);
2662 sbi->s_qf_names[i] = old_opts.s_qf_names[i];
2663 }
2664#endif
bbd6851a 2665 unlock_super(sb);
337eb00a 2666 unlock_kernel();
08c6a96f 2667 return err;
1da177e4
LT		 2668}
2669
726c3342 2670static int ext3_statfs (struct dentry * dentry, struct kstatfs * buf)
1da177e4 2671{
726c3342 2672 struct super_block *sb = dentry->d_sb;
09fe316a
AT		 2673 struct ext3_sb_info *sbi = EXT3_SB(sb);
2674 struct ext3_super_block *es = sbi->s_es;
50ee0a32 2675 u64 fsid;
1da177e4 2676
a71ce8c6
BP		 2677 if (test_opt(sb, MINIX_DF)) {
2678 sbi->s_overhead_last = 0;
2679 } else if (sbi->s_blocks_last != le32_to_cpu(es->s_blocks_count)) {
2680 unsigned long ngroups = sbi->s_groups_count, i;
2681 ext3_fsblk_t overhead = 0;
1da177e4
LT		 2682 smp_rmb();
2683
2684 /*
a71ce8c6
BP		 2685 * Compute the overhead (FS structures). This is constant
2686 * for a given filesystem unless the number of block groups
2687 * changes so we cache the previous value until it does.
1da177e4
LT		 2688 */
2689
2690 /*
2691 * All of the blocks before first_data_block are
2692 * overhead
2693 */
2694 overhead = le32_to_cpu(es->s_first_data_block);
2695
2696 /*
2697 * Add the overhead attributed to the superblock and
2698 * block group descriptors. If the sparse superblocks
2699 * feature is turned on, then not all groups have this.
2700 */
2701 for (i = 0; i < ngroups; i++) {
2702 overhead += ext3_bg_has_super(sb, i) +
2703 ext3_bg_num_gdb(sb, i);
2704 cond_resched();
2705 }
2706
2707 /*
2708 * Every block group has an inode bitmap, a block
2709 * bitmap, and an inode table.
2710 */
a71ce8c6
BP		 2711 overhead += ngroups * (2 + sbi->s_itb_per_group);
2712 sbi->s_overhead_last = overhead;
2713 smp_wmb();
2714 sbi->s_blocks_last = le32_to_cpu(es->s_blocks_count);
1da177e4
LT		 2715 }
2716
2717 buf->f_type = EXT3_SUPER_MAGIC;
2718 buf->f_bsize = sb->s_blocksize;
a71ce8c6 2719 buf->f_blocks = le32_to_cpu(es->s_blocks_count) - sbi->s_overhead_last;
52d9f3b4 2720 buf->f_bfree = percpu_counter_sum_positive(&sbi->s_freeblocks_counter);
1da177e4
LT		 2721 buf->f_bavail = buf->f_bfree - le32_to_cpu(es->s_r_blocks_count);
2722 if (buf->f_bfree < le32_to_cpu(es->s_r_blocks_count))
2723 buf->f_bavail = 0;
2724 buf->f_files = le32_to_cpu(es->s_inodes_count);
52d9f3b4 2725 buf->f_ffree = percpu_counter_sum_positive(&sbi->s_freeinodes_counter);
1da177e4 2726 buf->f_namelen = EXT3_NAME_LEN;
50ee0a32
PE		 2727 fsid = le64_to_cpup((void *)es->s_uuid) ^
2728 le64_to_cpup((void *)es->s_uuid + sizeof(u64));
2729 buf->f_fsid.val[0] = fsid & 0xFFFFFFFFUL;
2730 buf->f_fsid.val[1] = (fsid >> 32) & 0xFFFFFFFFUL;
1da177e4
LT		 2731 return 0;
2732}
2733
2734/* Helper function for writing quotas on sync - we need to start transaction before quota file
2735 * is locked for write. Otherwise the are possible deadlocks:
2736 * Process 1 Process 2
2737 * ext3_create() quota_sync()
2738 * journal_start() write_dquot()
81a05227 2739 * vfs_dq_init() down(dqio_mutex)
d3be915f 2740 * down(dqio_mutex) journal_start()
1da177e4
LT		 2741 *
2742 */
2743
2744#ifdef CONFIG_QUOTA
2745
2746static inline struct inode *dquot_to_inode(struct dquot *dquot)
2747{
2748 return sb_dqopt(dquot->dq_sb)->files[dquot->dq_type];
2749}
2750
1da177e4
LT		 2751static int ext3_write_dquot(struct dquot *dquot)
2752{
2753 int ret, err;
2754 handle_t *handle;
2755 struct inode *inode;
2756
2757 inode = dquot_to_inode(dquot);
2758 handle = ext3_journal_start(inode,
1f54587b 2759 EXT3_QUOTA_TRANS_BLOCKS(dquot->dq_sb));
1da177e4
LT		 2760 if (IS_ERR(handle))
2761 return PTR_ERR(handle);
2762 ret = dquot_commit(dquot);
2763 err = ext3_journal_stop(handle);
2764 if (!ret)
2765 ret = err;
2766 return ret;
2767}
2768
2769static int ext3_acquire_dquot(struct dquot *dquot)
2770{
2771 int ret, err;
2772 handle_t *handle;
2773
2774 handle = ext3_journal_start(dquot_to_inode(dquot),
1f54587b 2775 EXT3_QUOTA_INIT_BLOCKS(dquot->dq_sb));
1da177e4
LT		 2776 if (IS_ERR(handle))
2777 return PTR_ERR(handle);
2778 ret = dquot_acquire(dquot);
2779 err = ext3_journal_stop(handle);
2780 if (!ret)
2781 ret = err;
2782 return ret;
2783}
2784
2785static int ext3_release_dquot(struct dquot *dquot)
2786{
2787 int ret, err;
2788 handle_t *handle;
2789
2790 handle = ext3_journal_start(dquot_to_inode(dquot),
1f54587b 2791 EXT3_QUOTA_DEL_BLOCKS(dquot->dq_sb));
9c3013e9
JK		 2792 if (IS_ERR(handle)) {
2793 /* Release dquot anyway to avoid endless cycle in dqput() */
2794 dquot_release(dquot);
1da177e4 2795 return PTR_ERR(handle);
9c3013e9 2796 }
1da177e4
LT		 2797 ret = dquot_release(dquot);
2798 err = ext3_journal_stop(handle);
2799 if (!ret)
2800 ret = err;
2801 return ret;
2802}
2803
2804static int ext3_mark_dquot_dirty(struct dquot *dquot)
2805{
99aeaf63 2806 /* Are we journaling quotas? */
1da177e4
LT		 2807 if (EXT3_SB(dquot->dq_sb)->s_qf_names[USRQUOTA] ||
2808 EXT3_SB(dquot->dq_sb)->s_qf_names[GRPQUOTA]) {
2809 dquot_mark_dquot_dirty(dquot);
2810 return ext3_write_dquot(dquot);
2811 } else {
2812 return dquot_mark_dquot_dirty(dquot);
2813 }
2814}
2815
2816static int ext3_write_info(struct super_block *sb, int type)
2817{
2818 int ret, err;
2819 handle_t *handle;
2820
2821 /* Data block + inode block */
2822 handle = ext3_journal_start(sb->s_root->d_inode, 2);
2823 if (IS_ERR(handle))
2824 return PTR_ERR(handle);
2825 ret = dquot_commit_info(sb, type);
2826 err = ext3_journal_stop(handle);
2827 if (!ret)
2828 ret = err;
2829 return ret;
2830}
2831
2832/*
2833 * Turn on quotas during mount time - we need to find
2834 * the quota file and such...
2835 */
2836static int ext3_quota_on_mount(struct super_block *sb, int type)
2837{
84de856e
CH		 2838 return vfs_quota_on_mount(sb, EXT3_SB(sb)->s_qf_names[type],
2839 EXT3_SB(sb)->s_jquota_fmt, type);
1da177e4
LT		 2840}
2841
2842/*
2843 * Standard function to be called on quota_on
2844 */
2845static int ext3_quota_on(struct super_block *sb, int type, int format_id,
8264613d 2846 char *name, int remount)
1da177e4
LT		 2847{
2848 int err;
8264613d 2849 struct path path;
1da177e4 2850
1f54587b
JK		 2851 if (!test_opt(sb, QUOTA))
2852 return -EINVAL;
8264613d 2853 /* When remounting, no checks are needed and in fact, name is NULL */
9cfe7b90 2854 if (remount)
8264613d 2855 return vfs_quota_on(sb, type, format_id, name, remount);
9cfe7b90 2856
8264613d 2857 err = kern_path(name, LOOKUP_FOLLOW, &path);
1da177e4
LT		 2858 if (err)
2859 return err;
9cfe7b90 2860
1da177e4 2861 /* Quotafile not on the same filesystem? */
8264613d
AV		 2862 if (path.mnt->mnt_sb != sb) {
2863 path_put(&path);
1da177e4
LT		 2864 return -EXDEV;
2865 }
9cfe7b90
JK		 2866 /* Journaling quota? */
2867 if (EXT3_SB(sb)->s_qf_names[type]) {
2868 /* Quotafile not of fs root? */
8264613d 2869 if (path.dentry->d_parent != sb->s_root)
4cf46b67
AF		 2870 ext3_msg(sb, KERN_WARNING,
2871 "warning: Quota file not on filesystem root. "
2872 "Journaled quota will not work.");
9cfe7b90
JK		 2873 }
2874
2875 /*
2876 * When we journal data on quota file, we have to flush journal to see
2877 * all updates to the file when we bypass pagecache...
2878 */
8264613d 2879 if (ext3_should_journal_data(path.dentry->d_inode)) {
9cfe7b90
JK		 2880 /*
2881 * We don't need to lock updates but journal_flush() could
2882 * otherwise be livelocked...
2883 */
2884 journal_lock_updates(EXT3_SB(sb)->s_journal);
2d7c820e 2885 err = journal_flush(EXT3_SB(sb)->s_journal);
9cfe7b90 2886 journal_unlock_updates(EXT3_SB(sb)->s_journal);
2d7c820e 2887 if (err) {
12e1ec9f 2888 path_put(&path);
2d7c820e
HK		 2889 return err;
2890 }
9cfe7b90
JK		 2891 }
2892
8264613d
AV		 2893 err = vfs_quota_on_path(sb, type, format_id, &path);
2894 path_put(&path);
77e69dac 2895 return err;
1da177e4
LT		 2896}
2897
2898/* Read data from quotafile - avoid pagecache and such because we cannot afford
2899 * acquiring the locks... As quota files are never truncated and quota code
2900 * itself serializes the operations (and noone else should touch the files)
2901 * we don't have to be afraid of races */
2902static ssize_t ext3_quota_read(struct super_block *sb, int type, char *data,
2903 size_t len, loff_t off)
2904{
2905 struct inode *inode = sb_dqopt(sb)->files[type];
2906 sector_t blk = off >> EXT3_BLOCK_SIZE_BITS(sb);
2907 int err = 0;
2908 int offset = off & (sb->s_blocksize - 1);
2909 int tocopy;
2910 size_t toread;
2911 struct buffer_head *bh;
2912 loff_t i_size = i_size_read(inode);
2913
2914 if (off > i_size)
2915 return 0;
2916 if (off+len > i_size)
2917 len = i_size-off;
2918 toread = len;
2919 while (toread > 0) {
2920 tocopy = sb->s_blocksize - offset < toread ?
2921 sb->s_blocksize - offset : toread;
2922 bh = ext3_bread(NULL, inode, blk, 0, &err);
2923 if (err)
2924 return err;
2925 if (!bh) /* A hole? */
2926 memset(data, 0, tocopy);
2927 else
2928 memcpy(data, bh->b_data+offset, tocopy);
2929 brelse(bh);
2930 offset = 0;
2931 toread -= tocopy;
2932 data += tocopy;
2933 blk++;
2934 }
2935 return len;
2936}
2937
2938/* Write to quotafile (we know the transaction is already started and has
2939 * enough credits) */
2940static ssize_t ext3_quota_write(struct super_block *sb, int type,
2941 const char *data, size_t len, loff_t off)
2942{
2943 struct inode *inode = sb_dqopt(sb)->files[type];
2944 sector_t blk = off >> EXT3_BLOCK_SIZE_BITS(sb);
2945 int err = 0;
2946 int offset = off & (sb->s_blocksize - 1);
1da177e4 2947 int journal_quota = EXT3_SB(sb)->s_qf_names[type] != NULL;
1da177e4
LT		 2948 struct buffer_head *bh;
2949 handle_t *handle = journal_current_handle();
2950
9c3013e9 2951 if (!handle) {
4cf46b67
AF		 2952 ext3_msg(sb, KERN_WARNING,
2953 "warning: quota write (off=%llu, len=%llu)"
2954 " cancelled because transaction is not started.",
9c3013e9
JK		 2955 (unsigned long long)off, (unsigned long long)len);
2956 return -EIO;
2957 }
e5472147
DM		 2958
2959 /*
2960 * Since we account only one data block in transaction credits,
2961 * then it is impossible to cross a block boundary.
2962 */
2963 if (sb->s_blocksize - offset < len) {
2964 ext3_msg(sb, KERN_WARNING, "Quota write (off=%llu, len=%llu)"
2965 " cancelled because not block aligned",
2966 (unsigned long long)off, (unsigned long long)len);
2967 return -EIO;
2968 }
5c81a419 2969 mutex_lock_nested(&inode->i_mutex, I_MUTEX_QUOTA);
e5472147
DM		 2970 bh = ext3_bread(handle, inode, blk, 1, &err);
2971 if (!bh)
2972 goto out;
2973 if (journal_quota) {
2974 err = ext3_journal_get_write_access(handle, bh);
2975 if (err) {
2976 brelse(bh);
1da177e4 2977 goto out;
1da177e4 2978 }
1da177e4 2979 }
e5472147
DM		 2980 lock_buffer(bh);
2981 memcpy(bh->b_data+offset, data, len);
2982 flush_dcache_page(bh->b_page);
2983 unlock_buffer(bh);
2984 if (journal_quota)
2985 err = ext3_journal_dirty_metadata(handle, bh);
2986 else {
2987 /* Always do at least ordered writes for quotas */
2988 err = ext3_journal_dirty_data(handle, bh);
2989 mark_buffer_dirty(bh);
2990 }
2991 brelse(bh);
1da177e4 2992out:
e5472147 2993 if (err) {
f5c8f7da 2994 mutex_unlock(&inode->i_mutex);
1da177e4 2995 return err;
f5c8f7da 2996 }
e5472147
DM		 2997 if (inode->i_size < off + len) {
2998 i_size_write(inode, off + len);
1da177e4
LT		 2999 EXT3_I(inode)->i_disksize = inode->i_size;
3000 }
3001 inode->i_version++;
3002 inode->i_mtime = inode->i_ctime = CURRENT_TIME;
3003 ext3_mark_inode_dirty(handle, inode);
1b1dcc1b 3004 mutex_unlock(&inode->i_mutex);
e5472147 3005 return len;
1da177e4
LT		 3006}
3007
3008#endif
3009
454e2398
DH		 3010static int ext3_get_sb(struct file_system_type *fs_type,
3011 int flags, const char *dev_name, void *data, struct vfsmount *mnt)
1da177e4 3012{
454e2398 3013 return get_sb_bdev(fs_type, flags, dev_name, data, ext3_fill_super, mnt);
1da177e4
LT		 3014}
3015
3016static struct file_system_type ext3_fs_type = {
3017 .owner = THIS_MODULE,
3018 .name = "ext3",
3019 .get_sb = ext3_get_sb,
3020 .kill_sb = kill_block_super,
3021 .fs_flags = FS_REQUIRES_DEV,
3022};
3023
3024static int __init init_ext3_fs(void)
3025{
3026 int err = init_ext3_xattr();
3027 if (err)
3028 return err;
3029 err = init_inodecache();
3030 if (err)
3031 goto out1;
3032 err = register_filesystem(&ext3_fs_type);
3033 if (err)
3034 goto out;
3035 return 0;
3036out:
3037 destroy_inodecache();
3038out1:
e9ad5620 3039 exit_ext3_xattr();
1da177e4
LT		 3040 return err;
3041}
3042
3043static void __exit exit_ext3_fs(void)
3044{
3045 unregister_filesystem(&ext3_fs_type);
3046 destroy_inodecache();
3047 exit_ext3_xattr();
3048}
3049
3050MODULE_AUTHOR("Remy Card, Stephen Tweedie, Andrew Morton, Andreas Dilger, Theodore Ts'o and others");
3051MODULE_DESCRIPTION("Second Extended Filesystem with journaling extensions");
3052MODULE_LICENSE("GPL");
3053module_init(init_ext3_fs)
3054module_exit(exit_ext3_fs)
Linux kernel - Deliverables
This page took 0.716457 seconds and 5 git commands to generate.