projects / linux-2.6-block.git / blame
? search:
summary | shortlog | log | commit | commitdiff | tree
blob | blame (incremental) | history | HEAD
Merge branch 'master' of git://git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux...
[linux-2.6-block.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
71b96257
JL		 47static int ext3_load_journal(struct super_block *, struct ext3_super_block *,
48 unsigned long journal_devnum);
1da177e4 49static int ext3_create_journal(struct super_block *, struct ext3_super_block *,
eee194e7 50 unsigned int);
c4be0c1d
TS		 51static int ext3_commit_super(struct super_block *sb,
52 struct ext3_super_block *es,
1da177e4
LT		 53 int sync);
54static void ext3_mark_recovery_complete(struct super_block * sb,
55 struct ext3_super_block * es);
56static void ext3_clear_journal_err(struct super_block * sb,
57 struct ext3_super_block * es);
58static int ext3_sync_fs(struct super_block *sb, int wait);
59static const char *ext3_decode_error(struct super_block * sb, int errno,
60 char nbuf[16]);
61static int ext3_remount (struct super_block * sb, int * flags, char * data);
726c3342 62static int ext3_statfs (struct dentry * dentry, struct kstatfs * buf);
c4be0c1d 63static int ext3_unfreeze(struct super_block *sb);
1da177e4 64static void ext3_write_super (struct super_block * sb);
c4be0c1d 65static int ext3_freeze(struct super_block *sb);
1da177e4 66
ae6ddcc5 67/*
1da177e4
LT		 68 * Wrappers for journal_start/end.
69 *
70 * The only special thing we need to do here is to make sure that all
71 * journal_end calls result in the superblock being marked dirty, so
72 * that sync() will call the filesystem's write_super callback if
ae6ddcc5 73 * appropriate.
1da177e4
LT		 74 */
75handle_t *ext3_journal_start_sb(struct super_block *sb, int nblocks)
76{
77 journal_t *journal;
78
79 if (sb->s_flags & MS_RDONLY)
80 return ERR_PTR(-EROFS);
81
82 /* Special case here: if the journal has aborted behind our
83 * backs (eg. EIO in the commit thread), then we still need to
84 * take the FS itself readonly cleanly. */
85 journal = EXT3_SB(sb)->s_journal;
86 if (is_journal_aborted(journal)) {
e05b6b52 87 ext3_abort(sb, __func__,
1da177e4
LT		 88 "Detected aborted journal");
89 return ERR_PTR(-EROFS);
90 }
91
92 return journal_start(journal, nblocks);
93}
94
ae6ddcc5 95/*
1da177e4
LT		 96 * The only special thing we need to do here is to make sure that all
97 * journal_stop calls result in the superblock being marked dirty, so
98 * that sync() will call the filesystem's write_super callback if
ae6ddcc5 99 * appropriate.
1da177e4
LT		 100 */
101int __ext3_journal_stop(const char *where, handle_t *handle)
102{
103 struct super_block *sb;
104 int err;
105 int rc;
106
107 sb = handle->h_transaction->t_journal->j_private;
108 err = handle->h_err;
109 rc = journal_stop(handle);
110
111 if (!err)
112 err = rc;
113 if (err)
114 __ext3_std_error(sb, where, err);
115 return err;
116}
117
118void ext3_journal_abort_handle(const char *caller, const char *err_fn,
119 struct buffer_head *bh, handle_t *handle, int err)
120{
121 char nbuf[16];
122 const char *errstr = ext3_decode_error(NULL, err, nbuf);
123
124 if (bh)
125 BUFFER_TRACE(bh, "abort");
126
127 if (!handle->h_err)
128 handle->h_err = err;
129
130 if (is_handle_aborted(handle))
131 return;
132
133 printk(KERN_ERR "%s: aborting transaction: %s in %s\n",
134 caller, errstr, err_fn);
135
136 journal_abort_handle(handle);
137}
138
139/* Deal with the reporting of failure conditions on a filesystem such as
140 * inconsistencies detected or read IO failures.
141 *
142 * On ext2, we can store the error state of the filesystem in the
143 * superblock. That is not possible on ext3, because we may have other
144 * write ordering constraints on the superblock which prevent us from
145 * writing it out straight away; and given that the journal is about to
146 * be aborted, we can't rely on the current, or future, transactions to
147 * write out the superblock safely.
148 *
149 * We'll just use the journal_abort() error code to record an error in
150 * the journal instead. On recovery, the journal will compain about
151 * that error until we've noted it down and cleared it.
152 */
153
154static void ext3_handle_error(struct super_block *sb)
155{
156 struct ext3_super_block *es = EXT3_SB(sb)->s_es;
157
158 EXT3_SB(sb)->s_mount_state |= EXT3_ERROR_FS;
159 es->s_state |= cpu_to_le16(EXT3_ERROR_FS);
160
161 if (sb->s_flags & MS_RDONLY)
162 return;
163
7543fc7b 164 if (!test_opt (sb, ERRORS_CONT)) {
1da177e4
LT		 165 journal_t *journal = EXT3_SB(sb)->s_journal;
166
167 EXT3_SB(sb)->s_mount_opt |= EXT3_MOUNT_ABORT;
168 if (journal)
169 journal_abort(journal, -EIO);
170 }
7543fc7b
VA		 171 if (test_opt (sb, ERRORS_RO)) {
172 printk (KERN_CRIT "Remounting filesystem read-only\n");
173 sb->s_flags |= MS_RDONLY;
174 }
175 ext3_commit_super(sb, es, 1);
1da177e4
LT		 176 if (test_opt(sb, ERRORS_PANIC))
177 panic("EXT3-fs (device %s): panic forced after error\n",
178 sb->s_id);
1da177e4
LT		 179}
180
181void ext3_error (struct super_block * sb, const char * function,
182 const char * fmt, ...)
183{
184 va_list args;
185
186 va_start(args, fmt);
187 printk(KERN_CRIT "EXT3-fs error (device %s): %s: ",sb->s_id, function);
188 vprintk(fmt, args);
189 printk("\n");
190 va_end(args);
191
192 ext3_handle_error(sb);
193}
194
195static const char *ext3_decode_error(struct super_block * sb, int errno,
196 char nbuf[16])
197{
198 char *errstr = NULL;
199
200 switch (errno) {
201 case -EIO:
202 errstr = "IO failure";
203 break;
204 case -ENOMEM:
205 errstr = "Out of memory";
206 break;
207 case -EROFS:
208 if (!sb || EXT3_SB(sb)->s_journal->j_flags & JFS_ABORT)
209 errstr = "Journal has aborted";
210 else
211 errstr = "Readonly filesystem";
212 break;
213 default:
214 /* If the caller passed in an extra buffer for unknown
215 * errors, textualise them now. Else we just return
216 * NULL. */
217 if (nbuf) {
218 /* Check for truncated error codes... */
219 if (snprintf(nbuf, 16, "error %d", -errno) >= 0)
220 errstr = nbuf;
221 }
222 break;
223 }
224
225 return errstr;
226}
227
228/* __ext3_std_error decodes expected errors from journaling functions
229 * automatically and invokes the appropriate error response. */
230
231void __ext3_std_error (struct super_block * sb, const char * function,
232 int errno)
233{
234 char nbuf[16];
30121624 235 const char *errstr;
1da177e4 236
30121624
ST		 237 /* Special case: if the error is EROFS, and we're not already
238 * inside a transaction, then there's really no point in logging
239 * an error. */
240 if (errno == -EROFS && journal_current_handle() == NULL &&
241 (sb->s_flags & MS_RDONLY))
242 return;
243
244 errstr = ext3_decode_error(sb, errno, nbuf);
1da177e4
LT		 245 printk (KERN_CRIT "EXT3-fs error (device %s) in %s: %s\n",
246 sb->s_id, function, errstr);
247
248 ext3_handle_error(sb);
249}
250
251/*
252 * ext3_abort is a much stronger failure handler than ext3_error. The
253 * abort function may be used to deal with unrecoverable failures such
254 * as journal IO errors or ENOMEM at a critical moment in log management.
255 *
256 * We unconditionally force the filesystem into an ABORT|READONLY state,
257 * unless the error response on the fs has been set to panic in which
258 * case we take the easy way out and panic immediately.
259 */
260
261void ext3_abort (struct super_block * sb, const char * function,
262 const char * fmt, ...)
263{
264 va_list args;
265
266 printk (KERN_CRIT "ext3_abort called.\n");
267
268 va_start(args, fmt);
269 printk(KERN_CRIT "EXT3-fs error (device %s): %s: ",sb->s_id, function);
270 vprintk(fmt, args);
271 printk("\n");
272 va_end(args);
273
274 if (test_opt(sb, ERRORS_PANIC))
275 panic("EXT3-fs panic from previous error\n");
276
277 if (sb->s_flags & MS_RDONLY)
278 return;
279
280 printk(KERN_CRIT "Remounting filesystem read-only\n");
281 EXT3_SB(sb)->s_mount_state |= EXT3_ERROR_FS;
282 sb->s_flags |= MS_RDONLY;
283 EXT3_SB(sb)->s_mount_opt |= EXT3_MOUNT_ABORT;
44d6f787
HK		 284 if (EXT3_SB(sb)->s_journal)
285 journal_abort(EXT3_SB(sb)->s_journal, -EIO);
1da177e4
LT		 286}
287
288void ext3_warning (struct super_block * sb, const char * function,
289 const char * fmt, ...)
290{
291 va_list args;
292
293 va_start(args, fmt);
294 printk(KERN_WARNING "EXT3-fs warning (device %s): %s: ",
295 sb->s_id, function);
296 vprintk(fmt, args);
297 printk("\n");
298 va_end(args);
299}
300
301void ext3_update_dynamic_rev(struct super_block *sb)
302{
303 struct ext3_super_block *es = EXT3_SB(sb)->s_es;
304
305 if (le32_to_cpu(es->s_rev_level) > EXT3_GOOD_OLD_REV)
306 return;
307
e05b6b52 308 ext3_warning(sb, __func__,
1da177e4
LT		 309 "updating to rev %d because of new feature flag, "
310 "running e2fsck is recommended",
311 EXT3_DYNAMIC_REV);
312
313 es->s_first_ino = cpu_to_le32(EXT3_GOOD_OLD_FIRST_INO);
314 es->s_inode_size = cpu_to_le16(EXT3_GOOD_OLD_INODE_SIZE);
315 es->s_rev_level = cpu_to_le32(EXT3_DYNAMIC_REV);
316 /* leave es->s_feature_*compat flags alone */
317 /* es->s_uuid will be set by e2fsck if empty */
318
319 /*
320 * The rest of the superblock fields should be zero, and if not it
321 * means they are likely already in use, so leave them alone. We
322 * can leave it up to e2fsck to clean up any inconsistencies there.
323 */
324}
325
326/*
327 * Open the external journal device
328 */
329static struct block_device *ext3_blkdev_get(dev_t dev)
330{
331 struct block_device *bdev;
332 char b[BDEVNAME_SIZE];
333
334 bdev = open_by_devnum(dev, FMODE_READ|FMODE_WRITE);
335 if (IS_ERR(bdev))
336 goto fail;
337 return bdev;
338
339fail:
340 printk(KERN_ERR "EXT3: failed to open journal device %s: %ld\n",
341 __bdevname(dev, b), PTR_ERR(bdev));
342 return NULL;
343}
344
345/*
346 * Release the journal device
347 */
348static int ext3_blkdev_put(struct block_device *bdev)
349{
350 bd_release(bdev);
9a1c3542 351 return blkdev_put(bdev, FMODE_READ|FMODE_WRITE);
1da177e4
LT		 352}
353
354static int ext3_blkdev_remove(struct ext3_sb_info *sbi)
355{
356 struct block_device *bdev;
357 int ret = -ENODEV;
358
359 bdev = sbi->journal_bdev;
360 if (bdev) {
361 ret = ext3_blkdev_put(bdev);
362 sbi->journal_bdev = NULL;
363 }
364 return ret;
365}
366
367static inline struct inode *orphan_list_entry(struct list_head *l)
368{
369 return &list_entry(l, struct ext3_inode_info, i_orphan)->vfs_inode;
370}
371
372static void dump_orphan_list(struct super_block *sb, struct ext3_sb_info *sbi)
373{
374 struct list_head *l;
375
ae6ddcc5 376 printk(KERN_ERR "sb orphan head is %d\n",
1da177e4
LT		 377 le32_to_cpu(sbi->s_es->s_last_orphan));
378
379 printk(KERN_ERR "sb_info orphan list:\n");
380 list_for_each(l, &sbi->s_orphan) {
381 struct inode *inode = orphan_list_entry(l);
382 printk(KERN_ERR " "
eee194e7 383 "inode %s:%lu at %p: mode %o, nlink %d, next %d\n",
1da177e4 384 inode->i_sb->s_id, inode->i_ino, inode,
ae6ddcc5 385 inode->i_mode, inode->i_nlink,
1da177e4
LT		 386 NEXT_ORPHAN(inode));
387 }
388}
389
390static void ext3_put_super (struct super_block * sb)
391{
392 struct ext3_sb_info *sbi = EXT3_SB(sb);
393 struct ext3_super_block *es = sbi->s_es;
44d6f787 394 int i, err;
1da177e4
LT		 395
396 ext3_xattr_put_super(sb);
44d6f787
HK		 397 err = journal_destroy(sbi->s_journal);
398 sbi->s_journal = NULL;
399 if (err < 0)
2d7c820e 400 ext3_abort(sb, __func__, "Couldn't clean up the journal");
44d6f787 401
1da177e4
LT		 402 if (!(sb->s_flags & MS_RDONLY)) {
403 EXT3_CLEAR_INCOMPAT_FEATURE(sb, EXT3_FEATURE_INCOMPAT_RECOVER);
404 es->s_state = cpu_to_le16(sbi->s_mount_state);
405 BUFFER_TRACE(sbi->s_sbh, "marking dirty");
406 mark_buffer_dirty(sbi->s_sbh);
407 ext3_commit_super(sb, es, 1);
408 }
409
410 for (i = 0; i < sbi->s_gdb_count; i++)
411 brelse(sbi->s_group_desc[i]);
412 kfree(sbi->s_group_desc);
413 percpu_counter_destroy(&sbi->s_freeblocks_counter);
414 percpu_counter_destroy(&sbi->s_freeinodes_counter);
415 percpu_counter_destroy(&sbi->s_dirs_counter);
416 brelse(sbi->s_sbh);
417#ifdef CONFIG_QUOTA
418 for (i = 0; i < MAXQUOTAS; i++)
419 kfree(sbi->s_qf_names[i]);
420#endif
421
422 /* Debugging code just in case the in-memory inode orphan list
423 * isn't empty. The on-disk one can be non-empty if we've
424 * detected an error and taken the fs readonly, but the
425 * in-memory list had better be clean by this point. */
426 if (!list_empty(&sbi->s_orphan))
427 dump_orphan_list(sb, sbi);
428 J_ASSERT(list_empty(&sbi->s_orphan));
429
f98393a6 430 invalidate_bdev(sb->s_bdev);
1da177e4
LT		 431 if (sbi->journal_bdev && sbi->journal_bdev != sb->s_bdev) {
432 /*
433 * Invalidate the journal device's buffers. We don't want them
434 * floating about in memory - the physical journal device may
435 * hotswapped, and it breaks the `ro-after' testing code.
436 */
437 sync_blockdev(sbi->journal_bdev);
f98393a6 438 invalidate_bdev(sbi->journal_bdev);
1da177e4
LT		 439 ext3_blkdev_remove(sbi);
440 }
441 sb->s_fs_info = NULL;
5df096d6 442 kfree(sbi->s_blockgroup_lock);
1da177e4
LT		 443 kfree(sbi);
444 return;
445}
446
e18b890b 447static struct kmem_cache *ext3_inode_cachep;
1da177e4
LT		 448
449/*
450 * Called inside transaction, so use GFP_NOFS
451 */
452static struct inode *ext3_alloc_inode(struct super_block *sb)
453{
454 struct ext3_inode_info *ei;
455
e6b4f8da 456 ei = kmem_cache_alloc(ext3_inode_cachep, GFP_NOFS);
1da177e4
LT		 457 if (!ei)
458 return NULL;
459#ifdef CONFIG_EXT3_FS_POSIX_ACL
460 ei->i_acl = EXT3_ACL_NOT_CACHED;
461 ei->i_default_acl = EXT3_ACL_NOT_CACHED;
462#endif
463 ei->i_block_alloc_info = NULL;
464 ei->vfs_inode.i_version = 1;
465 return &ei->vfs_inode;
466}
467
468static void ext3_destroy_inode(struct inode *inode)
469{
9f7dd93d
VA		 470 if (!list_empty(&(EXT3_I(inode)->i_orphan))) {
471 printk("EXT3 Inode %p: orphan list check failed!\n",
472 EXT3_I(inode));
473 print_hex_dump(KERN_INFO, "", DUMP_PREFIX_ADDRESS, 16, 4,
474 EXT3_I(inode), sizeof(struct ext3_inode_info),
475 false);
476 dump_stack();
477 }
1da177e4
LT		 478 kmem_cache_free(ext3_inode_cachep, EXT3_I(inode));
479}
480
51cc5068 481static void init_once(void *foo)
1da177e4
LT		 482{
483 struct ext3_inode_info *ei = (struct ext3_inode_info *) foo;
484
a35afb83 485 INIT_LIST_HEAD(&ei->i_orphan);
1da177e4 486#ifdef CONFIG_EXT3_FS_XATTR
a35afb83 487 init_rwsem(&ei->xattr_sem);
1da177e4 488#endif
a35afb83
CL		 489 mutex_init(&ei->truncate_mutex);
490 inode_init_once(&ei->vfs_inode);
1da177e4 491}
ae6ddcc5 492
1da177e4
LT		 493static int init_inodecache(void)
494{
495 ext3_inode_cachep = kmem_cache_create("ext3_inode_cache",
496 sizeof(struct ext3_inode_info),
fffb60f9
PJ		 497 0, (SLAB_RECLAIM_ACCOUNT|
498 SLAB_MEM_SPREAD),
20c2df83 499 init_once);
1da177e4
LT		 500 if (ext3_inode_cachep == NULL)
501 return -ENOMEM;
502 return 0;
503}
504
505static void destroy_inodecache(void)
506{
1a1d92c1 507 kmem_cache_destroy(ext3_inode_cachep);
1da177e4
LT		 508}
509
510static void ext3_clear_inode(struct inode *inode)
511{
512 struct ext3_block_alloc_info *rsv = EXT3_I(inode)->i_block_alloc_info;
513#ifdef CONFIG_EXT3_FS_POSIX_ACL
e6022603
AM		 514 if (EXT3_I(inode)->i_acl &&
515 EXT3_I(inode)->i_acl != EXT3_ACL_NOT_CACHED) {
516 posix_acl_release(EXT3_I(inode)->i_acl);
517 EXT3_I(inode)->i_acl = EXT3_ACL_NOT_CACHED;
518 }
519 if (EXT3_I(inode)->i_default_acl &&
520 EXT3_I(inode)->i_default_acl != EXT3_ACL_NOT_CACHED) {
521 posix_acl_release(EXT3_I(inode)->i_default_acl);
522 EXT3_I(inode)->i_default_acl = EXT3_ACL_NOT_CACHED;
523 }
1da177e4
LT		 524#endif
525 ext3_discard_reservation(inode);
526 EXT3_I(inode)->i_block_alloc_info = NULL;
e6022603
AM		 527 if (unlikely(rsv))
528 kfree(rsv);
1da177e4
LT		 529}
530
6cd37cda 531static inline void ext3_show_quota_options(struct seq_file *seq, struct super_block *sb)
8fc2751b 532{
6cd37cda 533#if defined(CONFIG_QUOTA)
275abf5b 534 struct ext3_sb_info *sbi = EXT3_SB(sb);
8fc2751b 535
8fc2751b
MB		 536 if (sbi->s_jquota_fmt)
537 seq_printf(seq, ",jqfmt=%s",
538 (sbi->s_jquota_fmt == QFMT_VFS_OLD) ? "vfsold": "vfsv0");
539
540 if (sbi->s_qf_names[USRQUOTA])
541 seq_printf(seq, ",usrjquota=%s", sbi->s_qf_names[USRQUOTA]);
542
543 if (sbi->s_qf_names[GRPQUOTA])
544 seq_printf(seq, ",grpjquota=%s", sbi->s_qf_names[GRPQUOTA]);
545
546 if (sbi->s_mount_opt & EXT3_MOUNT_USRQUOTA)
547 seq_puts(seq, ",usrquota");
548
549 if (sbi->s_mount_opt & EXT3_MOUNT_GRPQUOTA)
550 seq_puts(seq, ",grpquota");
551#endif
6cd37cda
PO		 552}
553
571beed8
MS		 554/*
555 * Show an option if
556 * - it's set to a non-default value OR
557 * - if the per-sb default is different from the global default
558 */
6cd37cda
PO		 559static int ext3_show_options(struct seq_file *seq, struct vfsmount *vfs)
560{
561 struct super_block *sb = vfs->mnt_sb;
571beed8
MS		 562 struct ext3_sb_info *sbi = EXT3_SB(sb);
563 struct ext3_super_block *es = sbi->s_es;
564 unsigned long def_mount_opts;
565
566 def_mount_opts = le32_to_cpu(es->s_default_mount_opts);
567
568 if (sbi->s_sb_block != 1)
569 seq_printf(seq, ",sb=%lu", sbi->s_sb_block);
570 if (test_opt(sb, MINIX_DF))
571 seq_puts(seq, ",minixdf");
572 if (test_opt(sb, GRPID))
573 seq_puts(seq, ",grpid");
574 if (!test_opt(sb, GRPID) && (def_mount_opts & EXT3_DEFM_BSDGROUPS))
575 seq_puts(seq, ",nogrpid");
576 if (sbi->s_resuid != EXT3_DEF_RESUID ||
577 le16_to_cpu(es->s_def_resuid) != EXT3_DEF_RESUID) {
578 seq_printf(seq, ",resuid=%u", sbi->s_resuid);
579 }
580 if (sbi->s_resgid != EXT3_DEF_RESGID ||
581 le16_to_cpu(es->s_def_resgid) != EXT3_DEF_RESGID) {
582 seq_printf(seq, ",resgid=%u", sbi->s_resgid);
583 }
1eca93f9 584 if (test_opt(sb, ERRORS_RO)) {
571beed8
MS		 585 int def_errors = le16_to_cpu(es->s_errors);
586
587 if (def_errors == EXT3_ERRORS_PANIC ||
1eca93f9
AK		 588 def_errors == EXT3_ERRORS_CONTINUE) {
589 seq_puts(seq, ",errors=remount-ro");
571beed8
MS		 590 }
591 }
1eca93f9
AK		 592 if (test_opt(sb, ERRORS_CONT))
593 seq_puts(seq, ",errors=continue");
571beed8
MS		 594 if (test_opt(sb, ERRORS_PANIC))
595 seq_puts(seq, ",errors=panic");
596 if (test_opt(sb, NO_UID32))
597 seq_puts(seq, ",nouid32");
598 if (test_opt(sb, DEBUG))
599 seq_puts(seq, ",debug");
600 if (test_opt(sb, OLDALLOC))
601 seq_puts(seq, ",oldalloc");
602#ifdef CONFIG_EXT3_FS_XATTR
603 if (test_opt(sb, XATTR_USER))
604 seq_puts(seq, ",user_xattr");
605 if (!test_opt(sb, XATTR_USER) &&
606 (def_mount_opts & EXT3_DEFM_XATTR_USER)) {
607 seq_puts(seq, ",nouser_xattr");
608 }
609#endif
610#ifdef CONFIG_EXT3_FS_POSIX_ACL
611 if (test_opt(sb, POSIX_ACL))
612 seq_puts(seq, ",acl");
613 if (!test_opt(sb, POSIX_ACL) && (def_mount_opts & EXT3_DEFM_ACL))
614 seq_puts(seq, ",noacl");
615#endif
616 if (!test_opt(sb, RESERVATION))
617 seq_puts(seq, ",noreservation");
618 if (sbi->s_commit_interval) {
619 seq_printf(seq, ",commit=%u",
620 (unsigned) (sbi->s_commit_interval / HZ));
621 }
622 if (test_opt(sb, BARRIER))
623 seq_puts(seq, ",barrier=1");
624 if (test_opt(sb, NOBH))
625 seq_puts(seq, ",nobh");
6cd37cda
PO		 626
627 if (test_opt(sb, DATA_FLAGS) == EXT3_MOUNT_JOURNAL_DATA)
628 seq_puts(seq, ",data=journal");
629 else if (test_opt(sb, DATA_FLAGS) == EXT3_MOUNT_ORDERED_DATA)
630 seq_puts(seq, ",data=ordered");
631 else if (test_opt(sb, DATA_FLAGS) == EXT3_MOUNT_WRITEBACK_DATA)
632 seq_puts(seq, ",data=writeback");
633
0e4fb5e2
HK		 634 if (test_opt(sb, DATA_ERR_ABORT))
635 seq_puts(seq, ",data_err=abort");
636
6cd37cda 637 ext3_show_quota_options(seq, sb);
8fc2751b
MB		 638
639 return 0;
640}
1da177e4 641
fdb36673 642
74af0baa
CH		 643static struct inode *ext3_nfs_get_inode(struct super_block *sb,
644 u64 ino, u32 generation)
fdb36673 645{
fdb36673 646 struct inode *inode;
fdb36673
N		 647
648 if (ino < EXT3_FIRST_INO(sb) && ino != EXT3_ROOT_INO)
649 return ERR_PTR(-ESTALE);
650 if (ino > le32_to_cpu(EXT3_SB(sb)->s_es->s_inodes_count))
651 return ERR_PTR(-ESTALE);
652
653 /* iget isn't really right if the inode is currently unallocated!!
654 *
655 * ext3_read_inode will return a bad_inode if the inode had been
656 * deleted, so we should be safe.
657 *
658 * Currently we don't know the generation for parent directory, so
659 * a generation of 0 means "accept any"
660 */
473043dc
DH		 661 inode = ext3_iget(sb, ino);
662 if (IS_ERR(inode))
663 return ERR_CAST(inode);
664 if (generation && inode->i_generation != generation) {
fdb36673
N		 665 iput(inode);
666 return ERR_PTR(-ESTALE);
667 }
74af0baa
CH		 668
669 return inode;
670}
671
672static struct dentry *ext3_fh_to_dentry(struct super_block *sb, struct fid *fid,
673 int fh_len, int fh_type)
674{
675 return generic_fh_to_dentry(sb, fid, fh_len, fh_type,
676 ext3_nfs_get_inode);
677}
678
679static struct dentry *ext3_fh_to_parent(struct super_block *sb, struct fid *fid,
680 int fh_len, int fh_type)
681{
682 return generic_fh_to_parent(sb, fid, fh_len, fh_type,
683 ext3_nfs_get_inode);
fdb36673
N		 684}
685
6b082b53
TO		 686/*
687 * Try to release metadata pages (indirect blocks, directories) which are
688 * mapped via the block device. Since these pages could have journal heads
689 * which would prevent try_to_free_buffers() from freeing them, we must use
690 * jbd layer's try_to_free_buffers() function to release them.
691 */
692static int bdev_try_to_free_page(struct super_block *sb, struct page *page,
693 gfp_t wait)
694{
695 journal_t *journal = EXT3_SB(sb)->s_journal;
696
697 WARN_ON(PageChecked(page));
698 if (!page_has_buffers(page))
699 return 0;
700 if (journal)
701 return journal_try_to_free_buffers(journal, page,
702 wait & ~__GFP_WAIT);
703 return try_to_free_buffers(page);
704}
705
8fc2751b 706#ifdef CONFIG_QUOTA
1da177e4
LT		 707#define QTYPE2NAME(t) ((t)==USRQUOTA?"user":"group")
708#define QTYPE2MOPT(on, t) ((t)==USRQUOTA?((on)##USRJQUOTA):((on)##GRPJQUOTA))
709
710static int ext3_dquot_initialize(struct inode *inode, int type);
711static int ext3_dquot_drop(struct inode *inode);
712static int ext3_write_dquot(struct dquot *dquot);
713static int ext3_acquire_dquot(struct dquot *dquot);
714static int ext3_release_dquot(struct dquot *dquot);
715static int ext3_mark_dquot_dirty(struct dquot *dquot);
716static int ext3_write_info(struct super_block *sb, int type);
2fd83a4f
JK		 717static int ext3_quota_on(struct super_block *sb, int type, int format_id,
718 char *path, int remount);
1da177e4
LT		 719static int ext3_quota_on_mount(struct super_block *sb, int type);
720static ssize_t ext3_quota_read(struct super_block *sb, int type, char *data,
721 size_t len, loff_t off);
722static ssize_t ext3_quota_write(struct super_block *sb, int type,
723 const char *data, size_t len, loff_t off);
724
725static struct dquot_operations ext3_quota_operations = {
726 .initialize = ext3_dquot_initialize,
727 .drop = ext3_dquot_drop,
728 .alloc_space = dquot_alloc_space,
729 .alloc_inode = dquot_alloc_inode,
730 .free_space = dquot_free_space,
731 .free_inode = dquot_free_inode,
732 .transfer = dquot_transfer,
733 .write_dquot = ext3_write_dquot,
734 .acquire_dquot = ext3_acquire_dquot,
735 .release_dquot = ext3_release_dquot,
736 .mark_dirty = ext3_mark_dquot_dirty,
157091a2
JK		 737 .write_info = ext3_write_info,
738 .alloc_dquot = dquot_alloc,
739 .destroy_dquot = dquot_destroy,
1da177e4
LT		 740};
741
742static struct quotactl_ops ext3_qctl_operations = {
743 .quota_on = ext3_quota_on,
744 .quota_off = vfs_quota_off,
745 .quota_sync = vfs_quota_sync,
746 .get_info = vfs_get_dqinfo,
747 .set_info = vfs_set_dqinfo,
748 .get_dqblk = vfs_get_dqblk,
749 .set_dqblk = vfs_set_dqblk
750};
751#endif
752
ee9b6d61 753static const struct super_operations ext3_sops = {
1da177e4
LT		 754 .alloc_inode = ext3_alloc_inode,
755 .destroy_inode = ext3_destroy_inode,
1da177e4
LT		 756 .write_inode = ext3_write_inode,
757 .dirty_inode = ext3_dirty_inode,
758 .delete_inode = ext3_delete_inode,
759 .put_super = ext3_put_super,
760 .write_super = ext3_write_super,
761 .sync_fs = ext3_sync_fs,
c4be0c1d
TS		 762 .freeze_fs = ext3_freeze,
763 .unfreeze_fs = ext3_unfreeze,
1da177e4
LT		 764 .statfs = ext3_statfs,
765 .remount_fs = ext3_remount,
766 .clear_inode = ext3_clear_inode,
8fc2751b 767 .show_options = ext3_show_options,
1da177e4
LT		 768#ifdef CONFIG_QUOTA
769 .quota_read = ext3_quota_read,
770 .quota_write = ext3_quota_write,
771#endif
6b082b53 772 .bdev_try_to_free_page = bdev_try_to_free_page,
1da177e4
LT		 773};
774
39655164 775static const struct export_operations ext3_export_ops = {
74af0baa
CH		 776 .fh_to_dentry = ext3_fh_to_dentry,
777 .fh_to_parent = ext3_fh_to_parent,
1da177e4
LT		 778 .get_parent = ext3_get_parent,
779};
780
781enum {
782 Opt_bsd_df, Opt_minix_df, Opt_grpid, Opt_nogrpid,
783 Opt_resgid, Opt_resuid, Opt_sb, Opt_err_cont, Opt_err_panic, Opt_err_ro,
2860b733 784 Opt_nouid32, Opt_nocheck, Opt_debug, Opt_oldalloc, Opt_orlov,
1da177e4 785 Opt_user_xattr, Opt_nouser_xattr, Opt_acl, Opt_noacl,
ade1a29e 786 Opt_reservation, Opt_noreservation, Opt_noload, Opt_nobh, Opt_bh,
71b96257 787 Opt_commit, Opt_journal_update, Opt_journal_inum, Opt_journal_dev,
1da177e4 788 Opt_abort, Opt_data_journal, Opt_data_ordered, Opt_data_writeback,
0e4fb5e2 789 Opt_data_err_abort, Opt_data_err_ignore,
1da177e4 790 Opt_usrjquota, Opt_grpjquota, Opt_offusrjquota, Opt_offgrpjquota,
1f54587b 791 Opt_jqfmt_vfsold, Opt_jqfmt_vfsv0, Opt_quota, Opt_noquota,
8fc2751b
MB		 792 Opt_ignore, Opt_barrier, Opt_err, Opt_resize, Opt_usrquota,
793 Opt_grpquota
1da177e4
LT		 794};
795
a447c093 796static const match_table_t tokens = {
1da177e4
LT		 797 {Opt_bsd_df, "bsddf"},
798 {Opt_minix_df, "minixdf"},
799 {Opt_grpid, "grpid"},
800 {Opt_grpid, "bsdgroups"},
801 {Opt_nogrpid, "nogrpid"},
802 {Opt_nogrpid, "sysvgroups"},
803 {Opt_resgid, "resgid=%u"},
804 {Opt_resuid, "resuid=%u"},
805 {Opt_sb, "sb=%u"},
806 {Opt_err_cont, "errors=continue"},
807 {Opt_err_panic, "errors=panic"},
808 {Opt_err_ro, "errors=remount-ro"},
809 {Opt_nouid32, "nouid32"},
810 {Opt_nocheck, "nocheck"},
811 {Opt_nocheck, "check=none"},
1da177e4
LT		 812 {Opt_debug, "debug"},
813 {Opt_oldalloc, "oldalloc"},
814 {Opt_orlov, "orlov"},
815 {Opt_user_xattr, "user_xattr"},
816 {Opt_nouser_xattr, "nouser_xattr"},
817 {Opt_acl, "acl"},
818 {Opt_noacl, "noacl"},
819 {Opt_reservation, "reservation"},
820 {Opt_noreservation, "noreservation"},
821 {Opt_noload, "noload"},
822 {Opt_nobh, "nobh"},
ade1a29e 823 {Opt_bh, "bh"},
1da177e4
LT		 824 {Opt_commit, "commit=%u"},
825 {Opt_journal_update, "journal=update"},
826 {Opt_journal_inum, "journal=%u"},
71b96257 827 {Opt_journal_dev, "journal_dev=%u"},
1da177e4
LT		 828 {Opt_abort, "abort"},
829 {Opt_data_journal, "data=journal"},
830 {Opt_data_ordered, "data=ordered"},
831 {Opt_data_writeback, "data=writeback"},
0e4fb5e2
HK		 832 {Opt_data_err_abort, "data_err=abort"},
833 {Opt_data_err_ignore, "data_err=ignore"},
1da177e4
LT		 834 {Opt_offusrjquota, "usrjquota="},
835 {Opt_usrjquota, "usrjquota=%s"},
836 {Opt_offgrpjquota, "grpjquota="},
837 {Opt_grpjquota, "grpjquota=%s"},
838 {Opt_jqfmt_vfsold, "jqfmt=vfsold"},
839 {Opt_jqfmt_vfsv0, "jqfmt=vfsv0"},
8fc2751b 840 {Opt_grpquota, "grpquota"},
1f54587b
JK		 841 {Opt_noquota, "noquota"},
842 {Opt_quota, "quota"},
8fc2751b 843 {Opt_usrquota, "usrquota"},
1da177e4 844 {Opt_barrier, "barrier=%u"},
1da177e4 845 {Opt_resize, "resize"},
92587216 846 {Opt_err, NULL},
1da177e4
LT		 847};
848
43d23f90 849static ext3_fsblk_t get_sb_block(void **data)
1da177e4 850{
e9ad5620
DK		 851 ext3_fsblk_t sb_block;
852 char *options = (char *) *data;
1da177e4
LT		 853
854 if (!options || strncmp(options, "sb=", 3) != 0)
855 return 1; /* Default location */
856 options += 3;
43d23f90 857 /*todo: use simple_strtoll with >32bit ext3 */
1da177e4
LT		 858 sb_block = simple_strtoul(options, &options, 0);
859 if (*options && *options != ',') {
860 printk("EXT3-fs: Invalid sb specification: %s\n",
861 (char *) *data);
862 return 1;
863 }
864 if (*options == ',')
865 options++;
866 *data = (void *) options;
867 return sb_block;
868}
869
71b96257 870static int parse_options (char *options, struct super_block *sb,
eee194e7 871 unsigned int *inum, unsigned long *journal_devnum,
43d23f90 872 ext3_fsblk_t *n_blocks_count, int is_remount)
1da177e4
LT		 873{
874 struct ext3_sb_info *sbi = EXT3_SB(sb);
875 char * p;
876 substring_t args[MAX_OPT_ARGS];
877 int data_opt = 0;
878 int option;
879#ifdef CONFIG_QUOTA
d06bf1d2 880 int qtype, qfmt;
1da177e4
LT		 881 char *qname;
882#endif
883
884 if (!options)
885 return 1;
886
887 while ((p = strsep (&options, ",")) != NULL) {
888 int token;
889 if (!*p)
890 continue;
891
892 token = match_token(p, tokens, args);
893 switch (token) {
894 case Opt_bsd_df:
895 clear_opt (sbi->s_mount_opt, MINIX_DF);
896 break;
897 case Opt_minix_df:
898 set_opt (sbi->s_mount_opt, MINIX_DF);
899 break;
900 case Opt_grpid:
901 set_opt (sbi->s_mount_opt, GRPID);
902 break;
903 case Opt_nogrpid:
904 clear_opt (sbi->s_mount_opt, GRPID);
905 break;
906 case Opt_resuid:
907 if (match_int(&args[0], &option))
908 return 0;
909 sbi->s_resuid = option;
910 break;
911 case Opt_resgid:
912 if (match_int(&args[0], &option))
913 return 0;
914 sbi->s_resgid = option;
915 break;
916 case Opt_sb:
917 /* handled by get_sb_block() instead of here */
918 /* *sb_block = match_int(&args[0]); */
919 break;
920 case Opt_err_panic:
921 clear_opt (sbi->s_mount_opt, ERRORS_CONT);
922 clear_opt (sbi->s_mount_opt, ERRORS_RO);
923 set_opt (sbi->s_mount_opt, ERRORS_PANIC);
924 break;
925 case Opt_err_ro:
926 clear_opt (sbi->s_mount_opt, ERRORS_CONT);
927 clear_opt (sbi->s_mount_opt, ERRORS_PANIC);
928 set_opt (sbi->s_mount_opt, ERRORS_RO);
929 break;
930 case Opt_err_cont:
931 clear_opt (sbi->s_mount_opt, ERRORS_RO);
932 clear_opt (sbi->s_mount_opt, ERRORS_PANIC);
933 set_opt (sbi->s_mount_opt, ERRORS_CONT);
934 break;
935 case Opt_nouid32:
936 set_opt (sbi->s_mount_opt, NO_UID32);
937 break;
1da177e4
LT		 938 case Opt_nocheck:
939 clear_opt (sbi->s_mount_opt, CHECK);
940 break;
941 case Opt_debug:
942 set_opt (sbi->s_mount_opt, DEBUG);
943 break;
944 case Opt_oldalloc:
945 set_opt (sbi->s_mount_opt, OLDALLOC);
946 break;
947 case Opt_orlov:
948 clear_opt (sbi->s_mount_opt, OLDALLOC);
949 break;
950#ifdef CONFIG_EXT3_FS_XATTR
951 case Opt_user_xattr:
952 set_opt (sbi->s_mount_opt, XATTR_USER);
953 break;
954 case Opt_nouser_xattr:
955 clear_opt (sbi->s_mount_opt, XATTR_USER);
956 break;
957#else
958 case Opt_user_xattr:
959 case Opt_nouser_xattr:
960 printk("EXT3 (no)user_xattr options not supported\n");
961 break;
962#endif
963#ifdef CONFIG_EXT3_FS_POSIX_ACL
964 case Opt_acl:
965 set_opt(sbi->s_mount_opt, POSIX_ACL);
966 break;
967 case Opt_noacl:
968 clear_opt(sbi->s_mount_opt, POSIX_ACL);
969 break;
970#else
971 case Opt_acl:
972 case Opt_noacl:
973 printk("EXT3 (no)acl options not supported\n");
974 break;
975#endif
976 case Opt_reservation:
977 set_opt(sbi->s_mount_opt, RESERVATION);
978 break;
979 case Opt_noreservation:
980 clear_opt(sbi->s_mount_opt, RESERVATION);
981 break;
982 case Opt_journal_update:
983 /* @@@ FIXME */
984 /* Eventually we will want to be able to create
985 a journal file here. For now, only allow the
986 user to specify an existing inode to be the
987 journal file. */
988 if (is_remount) {
989 printk(KERN_ERR "EXT3-fs: cannot specify "
990 "journal on remount\n");
991 return 0;
992 }
993 set_opt (sbi->s_mount_opt, UPDATE_JOURNAL);
994 break;
995 case Opt_journal_inum:
996 if (is_remount) {
997 printk(KERN_ERR "EXT3-fs: cannot specify "
998 "journal on remount\n");
999 return 0;
1000 }
1001 if (match_int(&args[0], &option))
1002 return 0;
1003 *inum = option;
1004 break;
71b96257
JL		 1005 case Opt_journal_dev:
1006 if (is_remount) {
1007 printk(KERN_ERR "EXT3-fs: cannot specify "
1008 "journal on remount\n");
1009 return 0;
1010 }
1011 if (match_int(&args[0], &option))
1012 return 0;
1013 *journal_devnum = option;
1014 break;
1da177e4
LT		 1015 case Opt_noload:
1016 set_opt (sbi->s_mount_opt, NOLOAD);
1017 break;
1018 case Opt_commit:
1019 if (match_int(&args[0], &option))
1020 return 0;
1021 if (option < 0)
1022 return 0;
1023 if (option == 0)
1024 option = JBD_DEFAULT_MAX_COMMIT_AGE;
1025 sbi->s_commit_interval = HZ * option;
1026 break;
1027 case Opt_data_journal:
1028 data_opt = EXT3_MOUNT_JOURNAL_DATA;
1029 goto datacheck;
1030 case Opt_data_ordered:
1031 data_opt = EXT3_MOUNT_ORDERED_DATA;
1032 goto datacheck;
1033 case Opt_data_writeback:
1034 data_opt = EXT3_MOUNT_WRITEBACK_DATA;
1035 datacheck:
1036 if (is_remount) {
1037 if ((sbi->s_mount_opt & EXT3_MOUNT_DATA_FLAGS)
1038 != data_opt) {
1039 printk(KERN_ERR
1040 "EXT3-fs: cannot change data "
1041 "mode on remount\n");
1042 return 0;
1043 }
1044 } else {
1045 sbi->s_mount_opt &= ~EXT3_MOUNT_DATA_FLAGS;
1046 sbi->s_mount_opt |= data_opt;
1047 }
1048 break;
0e4fb5e2
HK		 1049 case Opt_data_err_abort:
1050 set_opt(sbi->s_mount_opt, DATA_ERR_ABORT);
1051 break;
1052 case Opt_data_err_ignore:
1053 clear_opt(sbi->s_mount_opt, DATA_ERR_ABORT);
1054 break;
1da177e4
LT		 1055#ifdef CONFIG_QUOTA
1056 case Opt_usrjquota:
1057 qtype = USRQUOTA;
1058 goto set_qf_name;
1059 case Opt_grpjquota:
1060 qtype = GRPQUOTA;
1061set_qf_name:
ee0d5ffe 1062 if (sb_any_quota_loaded(sb) &&
d06bf1d2 1063 !sbi->s_qf_names[qtype]) {
1da177e4 1064 printk(KERN_ERR
99aeaf63 1065 "EXT3-fs: Cannot change journaled "
1da177e4
LT		 1066 "quota options when quota turned on.\n");
1067 return 0;
1068 }
1069 qname = match_strdup(&args[0]);
1070 if (!qname) {
1071 printk(KERN_ERR
1072 "EXT3-fs: not enough memory for "
1073 "storing quotafile name.\n");
1074 return 0;
1075 }
1076 if (sbi->s_qf_names[qtype] &&
1077 strcmp(sbi->s_qf_names[qtype], qname)) {
1078 printk(KERN_ERR
1079 "EXT3-fs: %s quota file already "
1080 "specified.\n", QTYPE2NAME(qtype));
1081 kfree(qname);
1082 return 0;
1083 }
1084 sbi->s_qf_names[qtype] = qname;
1085 if (strchr(sbi->s_qf_names[qtype], '/')) {
1086 printk(KERN_ERR
1087 "EXT3-fs: quotafile must be on "
1088 "filesystem root.\n");
1089 kfree(sbi->s_qf_names[qtype]);
1090 sbi->s_qf_names[qtype] = NULL;
1091 return 0;
1092 }
1f54587b 1093 set_opt(sbi->s_mount_opt, QUOTA);
1da177e4
LT		 1094 break;
1095 case Opt_offusrjquota:
1096 qtype = USRQUOTA;
1097 goto clear_qf_name;
1098 case Opt_offgrpjquota:
1099 qtype = GRPQUOTA;
1100clear_qf_name:
ee0d5ffe 1101 if (sb_any_quota_loaded(sb) &&
d06bf1d2 1102 sbi->s_qf_names[qtype]) {
1da177e4 1103 printk(KERN_ERR "EXT3-fs: Cannot change "
99aeaf63 1104 "journaled quota options when "
1da177e4
LT		 1105 "quota turned on.\n");
1106 return 0;
1107 }
08c6a96f
JK		 1108 /*
1109 * The space will be released later when all options
1110 * are confirmed to be correct
1111 */
1da177e4
LT		 1112 sbi->s_qf_names[qtype] = NULL;
1113 break;
1114 case Opt_jqfmt_vfsold:
d06bf1d2
JK		 1115 qfmt = QFMT_VFS_OLD;
1116 goto set_qf_format;
1da177e4 1117 case Opt_jqfmt_vfsv0:
d06bf1d2
JK		 1118 qfmt = QFMT_VFS_V0;
1119set_qf_format:
ee0d5ffe 1120 if (sb_any_quota_loaded(sb) &&
d06bf1d2
JK		 1121 sbi->s_jquota_fmt != qfmt) {
1122 printk(KERN_ERR "EXT3-fs: Cannot change "
1123 "journaled quota options when "
1124 "quota turned on.\n");
1125 return 0;
1126 }
1127 sbi->s_jquota_fmt = qfmt;
1da177e4 1128 break;
1f54587b 1129 case Opt_quota:
8fc2751b
MB		 1130 case Opt_usrquota:
1131 set_opt(sbi->s_mount_opt, QUOTA);
1132 set_opt(sbi->s_mount_opt, USRQUOTA);
1133 break;
1134 case Opt_grpquota:
1f54587b 1135 set_opt(sbi->s_mount_opt, QUOTA);
8fc2751b 1136 set_opt(sbi->s_mount_opt, GRPQUOTA);
1f54587b
JK		 1137 break;
1138 case Opt_noquota:
ee0d5ffe 1139 if (sb_any_quota_loaded(sb)) {
1f54587b
JK		 1140 printk(KERN_ERR "EXT3-fs: Cannot change quota "
1141 "options when quota turned on.\n");
1142 return 0;
1143 }
1144 clear_opt(sbi->s_mount_opt, QUOTA);
8fc2751b
MB		 1145 clear_opt(sbi->s_mount_opt, USRQUOTA);
1146 clear_opt(sbi->s_mount_opt, GRPQUOTA);
1f54587b 1147 break;
1da177e4 1148#else
8fc2751b
MB		 1149 case Opt_quota:
1150 case Opt_usrquota:
1151 case Opt_grpquota:
fa1ff1e0
JK		 1152 printk(KERN_ERR
1153 "EXT3-fs: quota options not supported.\n");
1154 break;
1da177e4
LT		 1155 case Opt_usrjquota:
1156 case Opt_grpjquota:
1157 case Opt_offusrjquota:
1158 case Opt_offgrpjquota:
1159 case Opt_jqfmt_vfsold:
1160 case Opt_jqfmt_vfsv0:
1161 printk(KERN_ERR
fa1ff1e0 1162 "EXT3-fs: journaled quota options not "
1da177e4
LT		 1163 "supported.\n");
1164 break;
1f54587b
JK		 1165 case Opt_noquota:
1166 break;
1da177e4
LT		 1167#endif
1168 case Opt_abort:
1169 set_opt(sbi->s_mount_opt, ABORT);
1170 break;
1171 case Opt_barrier:
1172 if (match_int(&args[0], &option))
1173 return 0;
1174 if (option)
1175 set_opt(sbi->s_mount_opt, BARRIER);
1176 else
1177 clear_opt(sbi->s_mount_opt, BARRIER);
1178 break;
1179 case Opt_ignore:
1180 break;
1181 case Opt_resize:
08c6a96f 1182 if (!is_remount) {
1da177e4
LT		 1183 printk("EXT3-fs: resize option only available "
1184 "for remount\n");
1185 return 0;
1186 }
c7f1721e
KZ		 1187 if (match_int(&args[0], &option) != 0)
1188 return 0;
1da177e4
LT		 1189 *n_blocks_count = option;
1190 break;
1191 case Opt_nobh:
1192 set_opt(sbi->s_mount_opt, NOBH);
1193 break;
ade1a29e
BP		 1194 case Opt_bh:
1195 clear_opt(sbi->s_mount_opt, NOBH);
1196 break;
1da177e4
LT		 1197 default:
1198 printk (KERN_ERR
1199 "EXT3-fs: Unrecognized mount option \"%s\" "
1200 "or missing value\n", p);
1201 return 0;
1202 }
1203 }
1204#ifdef CONFIG_QUOTA
8fc2751b
MB		 1205 if (sbi->s_qf_names[USRQUOTA] || sbi->s_qf_names[GRPQUOTA]) {
1206 if ((sbi->s_mount_opt & EXT3_MOUNT_USRQUOTA) &&
1207 sbi->s_qf_names[USRQUOTA])
1208 clear_opt(sbi->s_mount_opt, USRQUOTA);
1209
1210 if ((sbi->s_mount_opt & EXT3_MOUNT_GRPQUOTA) &&
1211 sbi->s_qf_names[GRPQUOTA])
1212 clear_opt(sbi->s_mount_opt, GRPQUOTA);
1213
1214 if ((sbi->s_qf_names[USRQUOTA] &&
1215 (sbi->s_mount_opt & EXT3_MOUNT_GRPQUOTA)) ||
1216 (sbi->s_qf_names[GRPQUOTA] &&
1217 (sbi->s_mount_opt & EXT3_MOUNT_USRQUOTA))) {
1218 printk(KERN_ERR "EXT3-fs: old and new quota "
1219 "format mixing.\n");
1220 return 0;
1221 }
1222
1223 if (!sbi->s_jquota_fmt) {
99aeaf63 1224 printk(KERN_ERR "EXT3-fs: journaled quota format "
8fc2751b
MB		 1225 "not specified.\n");
1226 return 0;
1227 }
1228 } else {
1229 if (sbi->s_jquota_fmt) {
99aeaf63
JK		 1230 printk(KERN_ERR "EXT3-fs: journaled quota format "
1231 "specified with no journaling "
8fc2751b
MB		 1232 "enabled.\n");
1233 return 0;
1234 }
1da177e4
LT		 1235 }
1236#endif
1da177e4
LT		 1237 return 1;
1238}
1239
1240static int ext3_setup_super(struct super_block *sb, struct ext3_super_block *es,
1241 int read_only)
1242{
1243 struct ext3_sb_info *sbi = EXT3_SB(sb);
1244 int res = 0;
1245
1246 if (le32_to_cpu(es->s_rev_level) > EXT3_MAX_SUPP_REV) {
1247 printk (KERN_ERR "EXT3-fs warning: revision level too high, "
1248 "forcing read-only mode\n");
1249 res = MS_RDONLY;
1250 }
1251 if (read_only)
1252 return res;
1253 if (!(sbi->s_mount_state & EXT3_VALID_FS))
1254 printk (KERN_WARNING "EXT3-fs warning: mounting unchecked fs, "
1255 "running e2fsck is recommended\n");
1256 else if ((sbi->s_mount_state & EXT3_ERROR_FS))
1257 printk (KERN_WARNING
1258 "EXT3-fs warning: mounting fs with errors, "
1259 "running e2fsck is recommended\n");
1260 else if ((__s16) le16_to_cpu(es->s_max_mnt_count) >= 0 &&
1261 le16_to_cpu(es->s_mnt_count) >=
1262 (unsigned short) (__s16) le16_to_cpu(es->s_max_mnt_count))
1263 printk (KERN_WARNING
1264 "EXT3-fs warning: maximal mount count reached, "
1265 "running e2fsck is recommended\n");
1266 else if (le32_to_cpu(es->s_checkinterval) &&
1267 (le32_to_cpu(es->s_lastcheck) +
1268 le32_to_cpu(es->s_checkinterval) <= get_seconds()))
1269 printk (KERN_WARNING
1270 "EXT3-fs warning: checktime reached, "
1271 "running e2fsck is recommended\n");
1272#if 0
1273 /* @@@ We _will_ want to clear the valid bit if we find
1274 inconsistencies, to force a fsck at reboot. But for
1275 a plain journaled filesystem we can keep it set as
1276 valid forever! :) */
e7f23ebd 1277 es->s_state &= cpu_to_le16(~EXT3_VALID_FS);
1da177e4
LT		 1278#endif
1279 if (!(__s16) le16_to_cpu(es->s_max_mnt_count))
1280 es->s_max_mnt_count = cpu_to_le16(EXT3_DFL_MAX_MNT_COUNT);
50e8a289 1281 le16_add_cpu(&es->s_mnt_count, 1);
1da177e4
LT		 1282 es->s_mtime = cpu_to_le32(get_seconds());
1283 ext3_update_dynamic_rev(sb);
1284 EXT3_SET_INCOMPAT_FEATURE(sb, EXT3_FEATURE_INCOMPAT_RECOVER);
1285
1286 ext3_commit_super(sb, es, 1);
1287 if (test_opt(sb, DEBUG))
1288 printk(KERN_INFO "[EXT3 FS bs=%lu, gc=%lu, "
1289 "bpg=%lu, ipg=%lu, mo=%04lx]\n",
1290 sb->s_blocksize,
1291 sbi->s_groups_count,
1292 EXT3_BLOCKS_PER_GROUP(sb),
1293 EXT3_INODES_PER_GROUP(sb),
1294 sbi->s_mount_opt);
1295
1296 printk(KERN_INFO "EXT3 FS on %s, ", sb->s_id);
1297 if (EXT3_SB(sb)->s_journal->j_inode == NULL) {
1298 char b[BDEVNAME_SIZE];
1299
1300 printk("external journal on %s\n",
1301 bdevname(EXT3_SB(sb)->s_journal->j_dev, b));
1302 } else {
1303 printk("internal journal\n");
1304 }
1da177e4
LT		 1305 return res;
1306}
1307
1308/* Called at mount-time, super-block is locked */
197cd65a 1309static int ext3_check_descriptors(struct super_block *sb)
1da177e4
LT		 1310{
1311 struct ext3_sb_info *sbi = EXT3_SB(sb);
1da177e4
LT		 1312 int i;
1313
1314 ext3_debug ("Checking group descriptors");
1315
197cd65a
AM		 1316 for (i = 0; i < sbi->s_groups_count; i++) {
1317 struct ext3_group_desc *gdp = ext3_get_group_desc(sb, i, NULL);
1eaafeae
AM		 1318 ext3_fsblk_t first_block = ext3_group_first_block_no(sb, i);
1319 ext3_fsblk_t last_block;
197cd65a 1320
855565e8
ES		 1321 if (i == sbi->s_groups_count - 1)
1322 last_block = le32_to_cpu(sbi->s_es->s_blocks_count) - 1;
1323 else
1324 last_block = first_block +
1325 (EXT3_BLOCKS_PER_GROUP(sb) - 1);
1326
855565e8
ES		 1327 if (le32_to_cpu(gdp->bg_block_bitmap) < first_block ||
1328 le32_to_cpu(gdp->bg_block_bitmap) > last_block)
1da177e4
LT		 1329 {
1330 ext3_error (sb, "ext3_check_descriptors",
1331 "Block bitmap for group %d"
1332 " not in group (block %lu)!",
1333 i, (unsigned long)
1334 le32_to_cpu(gdp->bg_block_bitmap));
1335 return 0;
1336 }
855565e8
ES		 1337 if (le32_to_cpu(gdp->bg_inode_bitmap) < first_block ||
1338 le32_to_cpu(gdp->bg_inode_bitmap) > last_block)
1da177e4
LT		 1339 {
1340 ext3_error (sb, "ext3_check_descriptors",
1341 "Inode bitmap for group %d"
1342 " not in group (block %lu)!",
1343 i, (unsigned long)
1344 le32_to_cpu(gdp->bg_inode_bitmap));
1345 return 0;
1346 }
855565e8 1347 if (le32_to_cpu(gdp->bg_inode_table) < first_block ||
780dcdb2 1348 le32_to_cpu(gdp->bg_inode_table) + sbi->s_itb_per_group - 1 >
855565e8 1349 last_block)
1da177e4
LT		 1350 {
1351 ext3_error (sb, "ext3_check_descriptors",
1352 "Inode table for group %d"
1353 " not in group (block %lu)!",
1354 i, (unsigned long)
1355 le32_to_cpu(gdp->bg_inode_table));
1356 return 0;
1357 }
1da177e4
LT		 1358 }
1359
1360 sbi->s_es->s_free_blocks_count=cpu_to_le32(ext3_count_free_blocks(sb));
1361 sbi->s_es->s_free_inodes_count=cpu_to_le32(ext3_count_free_inodes(sb));
1362 return 1;
1363}
1364
1365
1366/* ext3_orphan_cleanup() walks a singly-linked list of inodes (starting at
1367 * the superblock) which were deleted from all directories, but held open by
1368 * a process at the time of a crash. We walk the list and try to delete these
1369 * inodes at recovery time (only with a read-write filesystem).
1370 *
1371 * In order to keep the orphan inode chain consistent during traversal (in
1372 * case of crash during recovery), we link each inode into the superblock
1373 * orphan list_head and handle it the same way as an inode deletion during
1374 * normal operation (which journals the operations for us).
1375 *
1376 * We only do an iget() and an iput() on each inode, which is very safe if we
1377 * accidentally point at an in-use or already deleted inode. The worst that
1378 * can happen in this case is that we get a "bit already cleared" message from
1379 * ext3_free_inode(). The only reason we would point at a wrong inode is if
1380 * e2fsck was run on this filesystem, and it must have already done the orphan
1381 * inode cleanup for us, so we can safely abort without any further action.
1382 */
1383static void ext3_orphan_cleanup (struct super_block * sb,
1384 struct ext3_super_block * es)
1385{
1386 unsigned int s_flags = sb->s_flags;
1387 int nr_orphans = 0, nr_truncates = 0;
1388#ifdef CONFIG_QUOTA
1389 int i;
1390#endif
1391 if (!es->s_last_orphan) {
1392 jbd_debug(4, "no orphan inodes to clean up\n");
1393 return;
1394 }
1395
a8f48a95
ES		 1396 if (bdev_read_only(sb->s_bdev)) {
1397 printk(KERN_ERR "EXT3-fs: write access "
1398 "unavailable, skipping orphan cleanup.\n");
1399 return;
1400 }
1401
1da177e4
LT		 1402 if (EXT3_SB(sb)->s_mount_state & EXT3_ERROR_FS) {
1403 if (es->s_last_orphan)
1404 jbd_debug(1, "Errors on filesystem, "
1405 "clearing orphan list.\n");
1406 es->s_last_orphan = 0;
1407 jbd_debug(1, "Skipping orphan recovery on fs with errors.\n");
1408 return;
1409 }
1410
1411 if (s_flags & MS_RDONLY) {
1412 printk(KERN_INFO "EXT3-fs: %s: orphan cleanup on readonly fs\n",
1413 sb->s_id);
1414 sb->s_flags &= ~MS_RDONLY;
1415 }
1416#ifdef CONFIG_QUOTA
1417 /* Needed for iput() to work correctly and not trash data */
1418 sb->s_flags |= MS_ACTIVE;
1419 /* Turn on quotas so that they are updated correctly */
1420 for (i = 0; i < MAXQUOTAS; i++) {
1421 if (EXT3_SB(sb)->s_qf_names[i]) {
1422 int ret = ext3_quota_on_mount(sb, i);
1423 if (ret < 0)
1424 printk(KERN_ERR
99aeaf63 1425 "EXT3-fs: Cannot turn on journaled "
1da177e4
LT		 1426 "quota: error %d\n", ret);
1427 }
1428 }
1429#endif
1430
1431 while (es->s_last_orphan) {
1432 struct inode *inode;
1433
473043dc
DH		 1434 inode = ext3_orphan_get(sb, le32_to_cpu(es->s_last_orphan));
1435 if (IS_ERR(inode)) {
1da177e4
LT		 1436 es->s_last_orphan = 0;
1437 break;
1438 }
1439
1440 list_add(&EXT3_I(inode)->i_orphan, &EXT3_SB(sb)->s_orphan);
1441 DQUOT_INIT(inode);
1442 if (inode->i_nlink) {
1443 printk(KERN_DEBUG
eee194e7 1444 "%s: truncating inode %lu to %Ld bytes\n",
e05b6b52 1445 __func__, inode->i_ino, inode->i_size);
eee194e7 1446 jbd_debug(2, "truncating inode %lu to %Ld bytes\n",
1da177e4
LT		 1447 inode->i_ino, inode->i_size);
1448 ext3_truncate(inode);
1449 nr_truncates++;
1450 } else {
1451 printk(KERN_DEBUG
eee194e7 1452 "%s: deleting unreferenced inode %lu\n",
e05b6b52 1453 __func__, inode->i_ino);
eee194e7 1454 jbd_debug(2, "deleting unreferenced inode %lu\n",
1da177e4
LT		 1455 inode->i_ino);
1456 nr_orphans++;
1457 }
1458 iput(inode); /* The delete magic happens here! */
1459 }
1460
1461#define PLURAL(x) (x), ((x)==1) ? "" : "s"
1462
1463 if (nr_orphans)
1464 printk(KERN_INFO "EXT3-fs: %s: %d orphan inode%s deleted\n",
1465 sb->s_id, PLURAL(nr_orphans));
1466 if (nr_truncates)
1467 printk(KERN_INFO "EXT3-fs: %s: %d truncate%s cleaned up\n",
1468 sb->s_id, PLURAL(nr_truncates));
1469#ifdef CONFIG_QUOTA
1470 /* Turn quotas off */
1471 for (i = 0; i < MAXQUOTAS; i++) {
1472 if (sb_dqopt(sb)->files[i])
2fd83a4f 1473 vfs_quota_off(sb, i, 0);
1da177e4
LT		 1474 }
1475#endif
1476 sb->s_flags = s_flags; /* Restore MS_RDONLY status */
1477}
1478
1da177e4
LT		 1479/*
1480 * Maximal file size. There is a direct, and {,double-,triple-}indirect
1481 * block limit, and also a limit of (2^32 - 1) 512-byte sectors in i_blocks.
1482 * We need to be 1 filesystem block less than the 2^32 sector limit.
1483 */
1484static loff_t ext3_max_size(int bits)
1485{
1486 loff_t res = EXT3_NDIR_BLOCKS;
fe7fdc37
AK		 1487 int meta_blocks;
1488 loff_t upper_limit;
1489
1490 /* This is calculated to be the largest file size for a
1491 * dense, file such that the total number of
1da177e4 1492 * sectors in the file, including data and all indirect blocks,
fe7fdc37
AK		 1493 * does not exceed 2^32 -1
1494 * __u32 i_blocks representing the total number of
1495 * 512 bytes blocks of the file
1496 */
1497 upper_limit = (1LL << 32) - 1;
1498
1499 /* total blocks in file system block size */
1500 upper_limit >>= (bits - 9);
1501
1502
1503 /* indirect blocks */
1504 meta_blocks = 1;
1505 /* double indirect blocks */
1506 meta_blocks += 1 + (1LL << (bits-2));
1507 /* tripple indirect blocks */
1508 meta_blocks += 1 + (1LL << (bits-2)) + (1LL << (2*(bits-2)));
1509
1510 upper_limit -= meta_blocks;
1511 upper_limit <<= bits;
1da177e4
LT		 1512
1513 res += 1LL << (bits-2);
1514 res += 1LL << (2*(bits-2));
1515 res += 1LL << (3*(bits-2));
1516 res <<= bits;
1517 if (res > upper_limit)
1518 res = upper_limit;
fe7fdc37
AK		 1519
1520 if (res > MAX_LFS_FILESIZE)
1521 res = MAX_LFS_FILESIZE;
1522
1da177e4
LT		 1523 return res;
1524}
1525
43d23f90
MC		 1526static ext3_fsblk_t descriptor_loc(struct super_block *sb,
1527 ext3_fsblk_t logic_sb_block,
1da177e4
LT		 1528 int nr)
1529{
1530 struct ext3_sb_info *sbi = EXT3_SB(sb);
43d23f90 1531 unsigned long bg, first_meta_bg;
1da177e4
LT		 1532 int has_super = 0;
1533
1da177e4
LT		 1534 first_meta_bg = le32_to_cpu(sbi->s_es->s_first_meta_bg);
1535
1536 if (!EXT3_HAS_INCOMPAT_FEATURE(sb, EXT3_FEATURE_INCOMPAT_META_BG) ||
1537 nr < first_meta_bg)
1538 return (logic_sb_block + nr + 1);
1539 bg = sbi->s_desc_per_block * nr;
1540 if (ext3_bg_has_super(sb, bg))
1541 has_super = 1;
43d23f90 1542 return (has_super + ext3_group_first_block_no(sb, bg));
1da177e4
LT		 1543}
1544
1545
1546static int ext3_fill_super (struct super_block *sb, void *data, int silent)
1547{
1548 struct buffer_head * bh;
1549 struct ext3_super_block *es = NULL;
1550 struct ext3_sb_info *sbi;
43d23f90
MC		 1551 ext3_fsblk_t block;
1552 ext3_fsblk_t sb_block = get_sb_block(&data);
1553 ext3_fsblk_t logic_sb_block;
1da177e4 1554 unsigned long offset = 0;
eee194e7 1555 unsigned int journal_inum = 0;
71b96257 1556 unsigned long journal_devnum = 0;
1da177e4
LT		 1557 unsigned long def_mount_opts;
1558 struct inode *root;
1559 int blocksize;
1560 int hblock;
1561 int db_count;
1562 int i;
1563 int needs_recovery;
473043dc 1564 int ret = -EINVAL;
1da177e4 1565 __le32 features;
833f4077 1566 int err;
1da177e4 1567
f8314dc6 1568 sbi = kzalloc(sizeof(*sbi), GFP_KERNEL);
1da177e4
LT		 1569 if (!sbi)
1570 return -ENOMEM;
5df096d6
PE		 1571
1572 sbi->s_blockgroup_lock =
1573 kzalloc(sizeof(struct blockgroup_lock), GFP_KERNEL);
1574 if (!sbi->s_blockgroup_lock) {
1575 kfree(sbi);
1576 return -ENOMEM;
1577 }
1da177e4 1578 sb->s_fs_info = sbi;
1da177e4
LT		 1579 sbi->s_mount_opt = 0;
1580 sbi->s_resuid = EXT3_DEF_RESUID;
1581 sbi->s_resgid = EXT3_DEF_RESGID;
571beed8 1582 sbi->s_sb_block = sb_block;
1da177e4
LT		 1583
1584 unlock_kernel();
1585
1586 blocksize = sb_min_blocksize(sb, EXT3_MIN_BLOCK_SIZE);
1587 if (!blocksize) {
1588 printk(KERN_ERR "EXT3-fs: unable to set blocksize\n");
1589 goto out_fail;
1590 }
1591
1592 /*
1593 * The ext3 superblock will not be buffer aligned for other than 1kB
1594 * block sizes. We need to calculate the offset from buffer start.
1595 */
1596 if (blocksize != EXT3_MIN_BLOCK_SIZE) {
1597 logic_sb_block = (sb_block * EXT3_MIN_BLOCK_SIZE) / blocksize;
1598 offset = (sb_block * EXT3_MIN_BLOCK_SIZE) % blocksize;
1599 } else {
1600 logic_sb_block = sb_block;
1601 }
1602
1603 if (!(bh = sb_bread(sb, logic_sb_block))) {
1604 printk (KERN_ERR "EXT3-fs: unable to read superblock\n");
1605 goto out_fail;
1606 }
1607 /*
1608 * Note: s_es must be initialized as soon as possible because
1609 * some ext3 macro-instructions depend on its value
1610 */
1611 es = (struct ext3_super_block *) (((char *)bh->b_data) + offset);
1612 sbi->s_es = es;
1613 sb->s_magic = le16_to_cpu(es->s_magic);
1614 if (sb->s_magic != EXT3_SUPER_MAGIC)
1615 goto cantfind_ext3;
1616
1617 /* Set defaults before we parse the mount options */
1618 def_mount_opts = le32_to_cpu(es->s_default_mount_opts);
1619 if (def_mount_opts & EXT3_DEFM_DEBUG)
1620 set_opt(sbi->s_mount_opt, DEBUG);
1621 if (def_mount_opts & EXT3_DEFM_BSDGROUPS)
1622 set_opt(sbi->s_mount_opt, GRPID);
1623 if (def_mount_opts & EXT3_DEFM_UID16)
1624 set_opt(sbi->s_mount_opt, NO_UID32);
2e7842b8 1625#ifdef CONFIG_EXT3_FS_XATTR
1da177e4
LT		 1626 if (def_mount_opts & EXT3_DEFM_XATTR_USER)
1627 set_opt(sbi->s_mount_opt, XATTR_USER);
2e7842b8
HD		 1628#endif
1629#ifdef CONFIG_EXT3_FS_POSIX_ACL
1da177e4
LT		 1630 if (def_mount_opts & EXT3_DEFM_ACL)
1631 set_opt(sbi->s_mount_opt, POSIX_ACL);
2e7842b8 1632#endif
1da177e4
LT		 1633 if ((def_mount_opts & EXT3_DEFM_JMODE) == EXT3_DEFM_JMODE_DATA)
1634 sbi->s_mount_opt |= EXT3_MOUNT_JOURNAL_DATA;
1635 else if ((def_mount_opts & EXT3_DEFM_JMODE) == EXT3_DEFM_JMODE_ORDERED)
1636 sbi->s_mount_opt |= EXT3_MOUNT_ORDERED_DATA;
1637 else if ((def_mount_opts & EXT3_DEFM_JMODE) == EXT3_DEFM_JMODE_WBACK)
1638 sbi->s_mount_opt |= EXT3_MOUNT_WRITEBACK_DATA;
1639
1640 if (le16_to_cpu(sbi->s_es->s_errors) == EXT3_ERRORS_PANIC)
1641 set_opt(sbi->s_mount_opt, ERRORS_PANIC);
1eca93f9 1642 else if (le16_to_cpu(sbi->s_es->s_errors) == EXT3_ERRORS_CONTINUE)
2245d7c2 1643 set_opt(sbi->s_mount_opt, ERRORS_CONT);
1eca93f9
AK		 1644 else
1645 set_opt(sbi->s_mount_opt, ERRORS_RO);
1da177e4
LT		 1646
1647 sbi->s_resuid = le16_to_cpu(es->s_def_resuid);
1648 sbi->s_resgid = le16_to_cpu(es->s_def_resgid);
1649
1650 set_opt(sbi->s_mount_opt, RESERVATION);
1651
71b96257
JL		 1652 if (!parse_options ((char *) data, sb, &journal_inum, &journal_devnum,
1653 NULL, 0))
1da177e4
LT		 1654 goto failed_mount;
1655
1656 sb->s_flags = (sb->s_flags & ~MS_POSIXACL) |
1657 ((sbi->s_mount_opt & EXT3_MOUNT_POSIX_ACL) ? MS_POSIXACL : 0);
1658
1659 if (le32_to_cpu(es->s_rev_level) == EXT3_GOOD_OLD_REV &&
1660 (EXT3_HAS_COMPAT_FEATURE(sb, ~0U) ||
1661 EXT3_HAS_RO_COMPAT_FEATURE(sb, ~0U) ||
1662 EXT3_HAS_INCOMPAT_FEATURE(sb, ~0U)))
ae6ddcc5 1663 printk(KERN_WARNING
1da177e4
LT		 1664 "EXT3-fs warning: feature flags set on rev 0 fs, "
1665 "running e2fsck is recommended\n");
1666 /*
1667 * Check feature flags regardless of the revision level, since we
1668 * previously didn't change the revision level when setting the flags,
1669 * so there is a chance incompat flags are set on a rev 0 filesystem.
1670 */
1671 features = EXT3_HAS_INCOMPAT_FEATURE(sb, ~EXT3_FEATURE_INCOMPAT_SUPP);
1672 if (features) {
1673 printk(KERN_ERR "EXT3-fs: %s: couldn't mount because of "
1674 "unsupported optional features (%x).\n",
1675 sb->s_id, le32_to_cpu(features));
1676 goto failed_mount;
1677 }
1678 features = EXT3_HAS_RO_COMPAT_FEATURE(sb, ~EXT3_FEATURE_RO_COMPAT_SUPP);
1679 if (!(sb->s_flags & MS_RDONLY) && features) {
1680 printk(KERN_ERR "EXT3-fs: %s: couldn't mount RDWR because of "
1681 "unsupported optional features (%x).\n",
1682 sb->s_id, le32_to_cpu(features));
1683 goto failed_mount;
1684 }
1685 blocksize = BLOCK_SIZE << le32_to_cpu(es->s_log_block_size);
1686
1687 if (blocksize < EXT3_MIN_BLOCK_SIZE ||
1688 blocksize > EXT3_MAX_BLOCK_SIZE) {
ae6ddcc5 1689 printk(KERN_ERR
1da177e4
LT		 1690 "EXT3-fs: Unsupported filesystem blocksize %d on %s.\n",
1691 blocksize, sb->s_id);
1692 goto failed_mount;
1693 }
1694
1695 hblock = bdev_hardsect_size(sb->s_bdev);
1696 if (sb->s_blocksize != blocksize) {
1697 /*
1698 * Make sure the blocksize for the filesystem is larger
1699 * than the hardware sectorsize for the machine.
1700 */
1701 if (blocksize < hblock) {
1702 printk(KERN_ERR "EXT3-fs: blocksize %d too small for "
1703 "device blocksize %d.\n", blocksize, hblock);
1704 goto failed_mount;
1705 }
1706
1707 brelse (bh);
0f0a89eb
TS		 1708 if (!sb_set_blocksize(sb, blocksize)) {
1709 printk(KERN_ERR "EXT3-fs: bad blocksize %d.\n",
1710 blocksize);
1711 goto out_fail;
1712 }
1da177e4
LT		 1713 logic_sb_block = (sb_block * EXT3_MIN_BLOCK_SIZE) / blocksize;
1714 offset = (sb_block * EXT3_MIN_BLOCK_SIZE) % blocksize;
1715 bh = sb_bread(sb, logic_sb_block);
1716 if (!bh) {
ae6ddcc5 1717 printk(KERN_ERR
1da177e4
LT		 1718 "EXT3-fs: Can't read superblock on 2nd try.\n");
1719 goto failed_mount;
1720 }
1721 es = (struct ext3_super_block *)(((char *)bh->b_data) + offset);
1722 sbi->s_es = es;
1723 if (es->s_magic != cpu_to_le16(EXT3_SUPER_MAGIC)) {
ae6ddcc5 1724 printk (KERN_ERR
1da177e4
LT		 1725 "EXT3-fs: Magic mismatch, very weird !\n");
1726 goto failed_mount;
1727 }
1728 }
1729
1730 sb->s_maxbytes = ext3_max_size(sb->s_blocksize_bits);
1731
1732 if (le32_to_cpu(es->s_rev_level) == EXT3_GOOD_OLD_REV) {
1733 sbi->s_inode_size = EXT3_GOOD_OLD_INODE_SIZE;
1734 sbi->s_first_ino = EXT3_GOOD_OLD_FIRST_INO;
1735 } else {
1736 sbi->s_inode_size = le16_to_cpu(es->s_inode_size);
1737 sbi->s_first_ino = le32_to_cpu(es->s_first_ino);
1738 if ((sbi->s_inode_size < EXT3_GOOD_OLD_INODE_SIZE) ||
3fc74269 1739 (!is_power_of_2(sbi->s_inode_size)) ||
1da177e4
LT		 1740 (sbi->s_inode_size > blocksize)) {
1741 printk (KERN_ERR
1742 "EXT3-fs: unsupported inode size: %d\n",
1743 sbi->s_inode_size);
1744 goto failed_mount;
1745 }
1746 }
1747 sbi->s_frag_size = EXT3_MIN_FRAG_SIZE <<
1748 le32_to_cpu(es->s_log_frag_size);
1749 if (blocksize != sbi->s_frag_size) {
1750 printk(KERN_ERR
1751 "EXT3-fs: fragsize %lu != blocksize %u (unsupported)\n",
1752 sbi->s_frag_size, blocksize);
1753 goto failed_mount;
1754 }
1755 sbi->s_frags_per_block = 1;
1756 sbi->s_blocks_per_group = le32_to_cpu(es->s_blocks_per_group);
1757 sbi->s_frags_per_group = le32_to_cpu(es->s_frags_per_group);
1758 sbi->s_inodes_per_group = le32_to_cpu(es->s_inodes_per_group);
b47b6f38 1759 if (EXT3_INODE_SIZE(sb) == 0 || EXT3_INODES_PER_GROUP(sb) == 0)
1da177e4
LT		 1760 goto cantfind_ext3;
1761 sbi->s_inodes_per_block = blocksize / EXT3_INODE_SIZE(sb);
1762 if (sbi->s_inodes_per_block == 0)
1763 goto cantfind_ext3;
1764 sbi->s_itb_per_group = sbi->s_inodes_per_group /
1765 sbi->s_inodes_per_block;
1766 sbi->s_desc_per_block = blocksize / sizeof(struct ext3_group_desc);
1767 sbi->s_sbh = bh;
1768 sbi->s_mount_state = le16_to_cpu(es->s_state);
f0d1b0b3
DH		 1769 sbi->s_addr_per_block_bits = ilog2(EXT3_ADDR_PER_BLOCK(sb));
1770 sbi->s_desc_per_block_bits = ilog2(EXT3_DESC_PER_BLOCK(sb));
1da177e4
LT		 1771 for (i=0; i < 4; i++)
1772 sbi->s_hash_seed[i] = le32_to_cpu(es->s_hash_seed[i]);
1773 sbi->s_def_hash_version = es->s_def_hash_version;
5e1f8c9e
TT		 1774 i = le32_to_cpu(es->s_flags);
1775 if (i & EXT2_FLAGS_UNSIGNED_HASH)
1776 sbi->s_hash_unsigned = 3;
1777 else if ((i & EXT2_FLAGS_SIGNED_HASH) == 0) {
1778#ifdef __CHAR_UNSIGNED__
1779 es->s_flags |= cpu_to_le32(EXT2_FLAGS_UNSIGNED_HASH);
1780 sbi->s_hash_unsigned = 3;
1781#else
1782 es->s_flags |= cpu_to_le32(EXT2_FLAGS_SIGNED_HASH);
1783#endif
1784 sb->s_dirt = 1;
1785 }
1da177e4
LT		 1786
1787 if (sbi->s_blocks_per_group > blocksize * 8) {
1788 printk (KERN_ERR
1789 "EXT3-fs: #blocks per group too big: %lu\n",
1790 sbi->s_blocks_per_group);
1791 goto failed_mount;
1792 }
1793 if (sbi->s_frags_per_group > blocksize * 8) {
1794 printk (KERN_ERR
1795 "EXT3-fs: #fragments per group too big: %lu\n",
1796 sbi->s_frags_per_group);
1797 goto failed_mount;
1798 }
1799 if (sbi->s_inodes_per_group > blocksize * 8) {
1800 printk (KERN_ERR
1801 "EXT3-fs: #inodes per group too big: %lu\n",
1802 sbi->s_inodes_per_group);
1803 goto failed_mount;
1804 }
1805
fcd5df35
MC		 1806 if (le32_to_cpu(es->s_blocks_count) >
1807 (sector_t)(~0ULL) >> (sb->s_blocksize_bits - 9)) {
1808 printk(KERN_ERR "EXT3-fs: filesystem on %s:"
1809 " too large to mount safely\n", sb->s_id);
1810 if (sizeof(sector_t) < 8)
1811 printk(KERN_WARNING "EXT3-fs: CONFIG_LBD not "
1812 "enabled\n");
1813 goto failed_mount;
1814 }
1815
1da177e4
LT		 1816 if (EXT3_BLOCKS_PER_GROUP(sb) == 0)
1817 goto cantfind_ext3;
855565e8
ES		 1818 sbi->s_groups_count = ((le32_to_cpu(es->s_blocks_count) -
1819 le32_to_cpu(es->s_first_data_block) - 1)
1820 / EXT3_BLOCKS_PER_GROUP(sb)) + 1;
1da177e4
LT		 1821 db_count = (sbi->s_groups_count + EXT3_DESC_PER_BLOCK(sb) - 1) /
1822 EXT3_DESC_PER_BLOCK(sb);
1823 sbi->s_group_desc = kmalloc(db_count * sizeof (struct buffer_head *),
1824 GFP_KERNEL);
1825 if (sbi->s_group_desc == NULL) {
1826 printk (KERN_ERR "EXT3-fs: not enough memory\n");
1827 goto failed_mount;
1828 }
1829
5df096d6 1830 bgl_lock_init(sbi->s_blockgroup_lock);
1da177e4
LT		 1831
1832 for (i = 0; i < db_count; i++) {
1833 block = descriptor_loc(sb, logic_sb_block, i);
1834 sbi->s_group_desc[i] = sb_bread(sb, block);
1835 if (!sbi->s_group_desc[i]) {
1836 printk (KERN_ERR "EXT3-fs: "
1837 "can't read group descriptor %d\n", i);
1838 db_count = i;
1839 goto failed_mount2;
1840 }
1841 }
1842 if (!ext3_check_descriptors (sb)) {
d2e5b13c 1843 printk(KERN_ERR "EXT3-fs: group descriptors corrupted!\n");
1da177e4
LT		 1844 goto failed_mount2;
1845 }
1846 sbi->s_gdb_count = db_count;
1847 get_random_bytes(&sbi->s_next_generation, sizeof(u32));
1848 spin_lock_init(&sbi->s_next_gen_lock);
0216bfcf 1849
833f4077
PZ		 1850 err = percpu_counter_init(&sbi->s_freeblocks_counter,
1851 ext3_count_free_blocks(sb));
1852 if (!err) {
1853 err = percpu_counter_init(&sbi->s_freeinodes_counter,
1854 ext3_count_free_inodes(sb));
1855 }
1856 if (!err) {
1857 err = percpu_counter_init(&sbi->s_dirs_counter,
1858 ext3_count_dirs(sb));
1859 }
1860 if (err) {
1861 printk(KERN_ERR "EXT3-fs: insufficient memory\n");
1862 goto failed_mount3;
1863 }
0216bfcf 1864
1da177e4
LT		 1865 /* per fileystem reservation list head & lock */
1866 spin_lock_init(&sbi->s_rsv_window_lock);
1867 sbi->s_rsv_window_root = RB_ROOT;
1868 /* Add a single, static dummy reservation to the start of the
1869 * reservation window list --- it gives us a placeholder for
1870 * append-at-start-of-list which makes the allocation logic
1871 * _much_ simpler. */
1872 sbi->s_rsv_window_head.rsv_start = EXT3_RESERVE_WINDOW_NOT_ALLOCATED;
1873 sbi->s_rsv_window_head.rsv_end = EXT3_RESERVE_WINDOW_NOT_ALLOCATED;
1874 sbi->s_rsv_window_head.rsv_alloc_hit = 0;
1875 sbi->s_rsv_window_head.rsv_goal_size = 0;
1876 ext3_rsv_window_add(sb, &sbi->s_rsv_window_head);
1877
1878 /*
1879 * set up enough so that it can read an inode
1880 */
1881 sb->s_op = &ext3_sops;
1882 sb->s_export_op = &ext3_export_ops;
1883 sb->s_xattr = ext3_xattr_handlers;
1884#ifdef CONFIG_QUOTA
1885 sb->s_qcop = &ext3_qctl_operations;
1886 sb->dq_op = &ext3_quota_operations;
1887#endif
1888 INIT_LIST_HEAD(&sbi->s_orphan); /* unlinked but open files */
1889
1890 sb->s_root = NULL;
1891
1892 needs_recovery = (es->s_last_orphan != 0 ||
1893 EXT3_HAS_INCOMPAT_FEATURE(sb,
1894 EXT3_FEATURE_INCOMPAT_RECOVER));
1895
1896 /*
1897 * The first inode we look at is the journal inode. Don't try
1898 * root first: it may be modified in the journal!
1899 */
1900 if (!test_opt(sb, NOLOAD) &&
1901 EXT3_HAS_COMPAT_FEATURE(sb, EXT3_FEATURE_COMPAT_HAS_JOURNAL)) {
71b96257 1902 if (ext3_load_journal(sb, es, journal_devnum))
0216bfcf 1903 goto failed_mount3;
1da177e4
LT		 1904 } else if (journal_inum) {
1905 if (ext3_create_journal(sb, es, journal_inum))
0216bfcf 1906 goto failed_mount3;
1da177e4
LT		 1907 } else {
1908 if (!silent)
1909 printk (KERN_ERR
1910 "ext3: No journal on filesystem on %s\n",
1911 sb->s_id);
0216bfcf 1912 goto failed_mount3;
1da177e4
LT		 1913 }
1914
1915 /* We have now updated the journal if required, so we can
1916 * validate the data journaling mode. */
1917 switch (test_opt(sb, DATA_FLAGS)) {
1918 case 0:
1919 /* No mode set, assume a default based on the journal
1920 capabilities: ORDERED_DATA if the journal can
1921 cope, else JOURNAL_DATA */
1922 if (journal_check_available_features
1923 (sbi->s_journal, 0, 0, JFS_FEATURE_INCOMPAT_REVOKE))
1924 set_opt(sbi->s_mount_opt, ORDERED_DATA);
1925 else
1926 set_opt(sbi->s_mount_opt, JOURNAL_DATA);
1927 break;
1928
1929 case EXT3_MOUNT_ORDERED_DATA:
1930 case EXT3_MOUNT_WRITEBACK_DATA:
1931 if (!journal_check_available_features
1932 (sbi->s_journal, 0, 0, JFS_FEATURE_INCOMPAT_REVOKE)) {
1933 printk(KERN_ERR "EXT3-fs: Journal does not support "
1934 "requested data journaling mode\n");
0216bfcf 1935 goto failed_mount4;
1da177e4
LT		 1936 }
1937 default:
1938 break;
1939 }
1940
1941 if (test_opt(sb, NOBH)) {
1da177e4
LT		 1942 if (!(test_opt(sb, DATA_FLAGS) == EXT3_MOUNT_WRITEBACK_DATA)) {
1943 printk(KERN_WARNING "EXT3-fs: Ignoring nobh option - "
1944 "its supported only with writeback mode\n");
1945 clear_opt(sbi->s_mount_opt, NOBH);
1946 }
1947 }
1948 /*
1949 * The journal_load will have done any necessary log recovery,
1950 * so we can safely mount the rest of the filesystem now.
1951 */
1952
473043dc
DH		 1953 root = ext3_iget(sb, EXT3_ROOT_INO);
1954 if (IS_ERR(root)) {
1da177e4 1955 printk(KERN_ERR "EXT3-fs: get root inode failed\n");
473043dc 1956 ret = PTR_ERR(root);
0216bfcf 1957 goto failed_mount4;
1da177e4
LT		 1958 }
1959 if (!S_ISDIR(root->i_mode) || !root->i_blocks || !root->i_size) {
473043dc 1960 iput(root);
1da177e4 1961 printk(KERN_ERR "EXT3-fs: corrupt root inode, run e2fsck\n");
0216bfcf 1962 goto failed_mount4;
1da177e4 1963 }
473043dc
DH		 1964 sb->s_root = d_alloc_root(root);
1965 if (!sb->s_root) {
1966 printk(KERN_ERR "EXT3-fs: get root dentry failed\n");
1967 iput(root);
1968 ret = -ENOMEM;
1969 goto failed_mount4;
1970 }
1da177e4
LT		 1971
1972 ext3_setup_super (sb, es, sb->s_flags & MS_RDONLY);
1973 /*
1974 * akpm: core read_super() calls in here with the superblock locked.
1975 * That deadlocks, because orphan cleanup needs to lock the superblock
1976 * in numerous places. Here we just pop the lock - it's relatively
1977 * harmless, because we are now ready to accept write_super() requests,
1978 * and aviro says that's the only reason for hanging onto the
1979 * superblock lock.
1980 */
1981 EXT3_SB(sb)->s_mount_state |= EXT3_ORPHAN_FS;
1982 ext3_orphan_cleanup(sb, es);
1983 EXT3_SB(sb)->s_mount_state &= ~EXT3_ORPHAN_FS;
1984 if (needs_recovery)
1985 printk (KERN_INFO "EXT3-fs: recovery complete.\n");
1986 ext3_mark_recovery_complete(sb, es);
1987 printk (KERN_INFO "EXT3-fs: mounted filesystem with %s data mode.\n",
1988 test_opt(sb,DATA_FLAGS) == EXT3_MOUNT_JOURNAL_DATA ? "journal":
1989 test_opt(sb,DATA_FLAGS) == EXT3_MOUNT_ORDERED_DATA ? "ordered":
1990 "writeback");
1991
1da177e4
LT		 1992 lock_kernel();
1993 return 0;
1994
1995cantfind_ext3:
1996 if (!silent)
1997 printk(KERN_ERR "VFS: Can't find ext3 filesystem on dev %s.\n",
1998 sb->s_id);
1999 goto failed_mount;
2000
0216bfcf 2001failed_mount4:
1da177e4 2002 journal_destroy(sbi->s_journal);
0216bfcf
MC		 2003failed_mount3:
2004 percpu_counter_destroy(&sbi->s_freeblocks_counter);
2005 percpu_counter_destroy(&sbi->s_freeinodes_counter);
2006 percpu_counter_destroy(&sbi->s_dirs_counter);
1da177e4
LT		 2007failed_mount2:
2008 for (i = 0; i < db_count; i++)
2009 brelse(sbi->s_group_desc[i]);
2010 kfree(sbi->s_group_desc);
2011failed_mount:
2012#ifdef CONFIG_QUOTA
2013 for (i = 0; i < MAXQUOTAS; i++)
2014 kfree(sbi->s_qf_names[i]);
2015#endif
2016 ext3_blkdev_remove(sbi);
2017 brelse(bh);
2018out_fail:
2019 sb->s_fs_info = NULL;
2020 kfree(sbi);
2021 lock_kernel();
473043dc 2022 return ret;
1da177e4
LT		 2023}
2024
2025/*
2026 * Setup any per-fs journal parameters now. We'll do this both on
2027 * initial mount, once the journal has been initialised but before we've
ae6ddcc5 2028 * done any recovery; and again on any subsequent remount.
1da177e4
LT		 2029 */
2030static void ext3_init_journal_params(struct super_block *sb, journal_t *journal)
2031{
2032 struct ext3_sb_info *sbi = EXT3_SB(sb);
2033
2034 if (sbi->s_commit_interval)
2035 journal->j_commit_interval = sbi->s_commit_interval;
2036 /* We could also set up an ext3-specific default for the commit
2037 * interval here, but for now we'll just fall back to the jbd
2038 * default. */
2039
2040 spin_lock(&journal->j_state_lock);
2041 if (test_opt(sb, BARRIER))
2042 journal->j_flags |= JFS_BARRIER;
2043 else
2044 journal->j_flags &= ~JFS_BARRIER;
0e4fb5e2
HK		 2045 if (test_opt(sb, DATA_ERR_ABORT))
2046 journal->j_flags |= JFS_ABORT_ON_SYNCDATA_ERR;
2047 else
2048 journal->j_flags &= ~JFS_ABORT_ON_SYNCDATA_ERR;
1da177e4
LT		 2049 spin_unlock(&journal->j_state_lock);
2050}
2051
eee194e7
ES		 2052static journal_t *ext3_get_journal(struct super_block *sb,
2053 unsigned int journal_inum)
1da177e4
LT		 2054{
2055 struct inode *journal_inode;
2056 journal_t *journal;
2057
2058 /* First, test for the existence of a valid inode on disk. Bad
2059 * things happen if we iget() an unused inode, as the subsequent
2060 * iput() will try to delete it. */
2061
473043dc
DH		 2062 journal_inode = ext3_iget(sb, journal_inum);
2063 if (IS_ERR(journal_inode)) {
1da177e4
LT		 2064 printk(KERN_ERR "EXT3-fs: no journal found.\n");
2065 return NULL;
2066 }
2067 if (!journal_inode->i_nlink) {
2068 make_bad_inode(journal_inode);
2069 iput(journal_inode);
2070 printk(KERN_ERR "EXT3-fs: journal inode is deleted.\n");
2071 return NULL;
2072 }
2073
2074 jbd_debug(2, "Journal inode found at %p: %Ld bytes\n",
2075 journal_inode, journal_inode->i_size);
473043dc 2076 if (!S_ISREG(journal_inode->i_mode)) {
1da177e4
LT		 2077 printk(KERN_ERR "EXT3-fs: invalid journal inode.\n");
2078 iput(journal_inode);
2079 return NULL;
2080 }
2081
2082 journal = journal_init_inode(journal_inode);
2083 if (!journal) {
2084 printk(KERN_ERR "EXT3-fs: Could not load journal inode\n");
2085 iput(journal_inode);
2086 return NULL;
2087 }
2088 journal->j_private = sb;
2089 ext3_init_journal_params(sb, journal);
2090 return journal;
2091}
2092
2093static journal_t *ext3_get_dev_journal(struct super_block *sb,
2094 dev_t j_dev)
2095{
2096 struct buffer_head * bh;
2097 journal_t *journal;
43d23f90 2098 ext3_fsblk_t start;
1c2bf374 2099 ext3_fsblk_t len;
1da177e4 2100 int hblock, blocksize;
43d23f90 2101 ext3_fsblk_t sb_block;
1da177e4
LT		 2102 unsigned long offset;
2103 struct ext3_super_block * es;
2104 struct block_device *bdev;
2105
2106 bdev = ext3_blkdev_get(j_dev);
2107 if (bdev == NULL)
2108 return NULL;
2109
2110 if (bd_claim(bdev, sb)) {
2111 printk(KERN_ERR
2112 "EXT3: failed to claim external journal device.\n");
9a1c3542 2113 blkdev_put(bdev, FMODE_READ|FMODE_WRITE);
1da177e4
LT		 2114 return NULL;
2115 }
2116
2117 blocksize = sb->s_blocksize;
2118 hblock = bdev_hardsect_size(bdev);
2119 if (blocksize < hblock) {
2120 printk(KERN_ERR
2121 "EXT3-fs: blocksize too small for journal device.\n");
2122 goto out_bdev;
2123 }
2124
2125 sb_block = EXT3_MIN_BLOCK_SIZE / blocksize;
2126 offset = EXT3_MIN_BLOCK_SIZE % blocksize;
2127 set_blocksize(bdev, blocksize);
2128 if (!(bh = __bread(bdev, sb_block, blocksize))) {
2129 printk(KERN_ERR "EXT3-fs: couldn't read superblock of "
2130 "external journal\n");
2131 goto out_bdev;
2132 }
2133
2134 es = (struct ext3_super_block *) (((char *)bh->b_data) + offset);
2135 if ((le16_to_cpu(es->s_magic) != EXT3_SUPER_MAGIC) ||
2136 !(le32_to_cpu(es->s_feature_incompat) &
2137 EXT3_FEATURE_INCOMPAT_JOURNAL_DEV)) {
2138 printk(KERN_ERR "EXT3-fs: external journal has "
2139 "bad superblock\n");
2140 brelse(bh);
2141 goto out_bdev;
2142 }
2143
2144 if (memcmp(EXT3_SB(sb)->s_es->s_journal_uuid, es->s_uuid, 16)) {
2145 printk(KERN_ERR "EXT3-fs: journal UUID does not match\n");
2146 brelse(bh);
2147 goto out_bdev;
2148 }
2149
2150 len = le32_to_cpu(es->s_blocks_count);
2151 start = sb_block + 1;
2152 brelse(bh); /* we're done with the superblock */
2153
2154 journal = journal_init_dev(bdev, sb->s_bdev,
2155 start, len, blocksize);
2156 if (!journal) {
2157 printk(KERN_ERR "EXT3-fs: failed to create device journal\n");
2158 goto out_bdev;
2159 }
2160 journal->j_private = sb;
2161 ll_rw_block(READ, 1, &journal->j_sb_buffer);
2162 wait_on_buffer(journal->j_sb_buffer);
2163 if (!buffer_uptodate(journal->j_sb_buffer)) {
2164 printk(KERN_ERR "EXT3-fs: I/O error on journal device\n");
2165 goto out_journal;
2166 }
2167 if (be32_to_cpu(journal->j_superblock->s_nr_users) != 1) {
2168 printk(KERN_ERR "EXT3-fs: External journal has more than one "
2169 "user (unsupported) - %d\n",
2170 be32_to_cpu(journal->j_superblock->s_nr_users));
2171 goto out_journal;
2172 }
2173 EXT3_SB(sb)->journal_bdev = bdev;
2174 ext3_init_journal_params(sb, journal);
2175 return journal;
2176out_journal:
2177 journal_destroy(journal);
2178out_bdev:
2179 ext3_blkdev_put(bdev);
2180 return NULL;
2181}
2182
71b96257
JL		 2183static int ext3_load_journal(struct super_block *sb,
2184 struct ext3_super_block *es,
2185 unsigned long journal_devnum)
1da177e4
LT		 2186{
2187 journal_t *journal;
eee194e7 2188 unsigned int journal_inum = le32_to_cpu(es->s_journal_inum);
71b96257 2189 dev_t journal_dev;
1da177e4
LT		 2190 int err = 0;
2191 int really_read_only;
2192
71b96257
JL		 2193 if (journal_devnum &&
2194 journal_devnum != le32_to_cpu(es->s_journal_dev)) {
2195 printk(KERN_INFO "EXT3-fs: external journal device major/minor "
2196 "numbers have changed\n");
2197 journal_dev = new_decode_dev(journal_devnum);
2198 } else
2199 journal_dev = new_decode_dev(le32_to_cpu(es->s_journal_dev));
2200
1da177e4
LT		 2201 really_read_only = bdev_read_only(sb->s_bdev);
2202
2203 /*
2204 * Are we loading a blank journal or performing recovery after a
2205 * crash? For recovery, we need to check in advance whether we
2206 * can get read-write access to the device.
2207 */
2208
2209 if (EXT3_HAS_INCOMPAT_FEATURE(sb, EXT3_FEATURE_INCOMPAT_RECOVER)) {
2210 if (sb->s_flags & MS_RDONLY) {
2211 printk(KERN_INFO "EXT3-fs: INFO: recovery "
2212 "required on readonly filesystem.\n");
2213 if (really_read_only) {
2214 printk(KERN_ERR "EXT3-fs: write access "
2215 "unavailable, cannot proceed.\n");
2216 return -EROFS;
2217 }
2218 printk (KERN_INFO "EXT3-fs: write access will "
2219 "be enabled during recovery.\n");
2220 }
2221 }
2222
2223 if (journal_inum && journal_dev) {
2224 printk(KERN_ERR "EXT3-fs: filesystem has both journal "
2225 "and inode journals!\n");
2226 return -EINVAL;
2227 }
2228
2229 if (journal_inum) {
2230 if (!(journal = ext3_get_journal(sb, journal_inum)))
2231 return -EINVAL;
2232 } else {
2233 if (!(journal = ext3_get_dev_journal(sb, journal_dev)))
2234 return -EINVAL;
2235 }
2236
2237 if (!really_read_only && test_opt(sb, UPDATE_JOURNAL)) {
2238 err = journal_update_format(journal);
2239 if (err) {
2240 printk(KERN_ERR "EXT3-fs: error updating journal.\n");
2241 journal_destroy(journal);
2242 return err;
2243 }
2244 }
2245
2246 if (!EXT3_HAS_INCOMPAT_FEATURE(sb, EXT3_FEATURE_INCOMPAT_RECOVER))
2247 err = journal_wipe(journal, !really_read_only);
2248 if (!err)
2249 err = journal_load(journal);
2250
2251 if (err) {
2252 printk(KERN_ERR "EXT3-fs: error loading journal.\n");
2253 journal_destroy(journal);
2254 return err;
2255 }
2256
2257 EXT3_SB(sb)->s_journal = journal;
2258 ext3_clear_journal_err(sb, es);
71b96257
JL		 2259
2260 if (journal_devnum &&
2261 journal_devnum != le32_to_cpu(es->s_journal_dev)) {
2262 es->s_journal_dev = cpu_to_le32(journal_devnum);
2263 sb->s_dirt = 1;
2264
2265 /* Make sure we flush the recovery flag to disk. */
2266 ext3_commit_super(sb, es, 1);
2267 }
2268
1da177e4
LT		 2269 return 0;
2270}
2271
2272static int ext3_create_journal(struct super_block * sb,
2273 struct ext3_super_block * es,
eee194e7 2274 unsigned int journal_inum)
1da177e4
LT		 2275{
2276 journal_t *journal;
952d9de1 2277 int err;
1da177e4
LT		 2278
2279 if (sb->s_flags & MS_RDONLY) {
2280 printk(KERN_ERR "EXT3-fs: readonly filesystem when trying to "
2281 "create journal.\n");
2282 return -EROFS;
2283 }
2284
952d9de1
BP		 2285 journal = ext3_get_journal(sb, journal_inum);
2286 if (!journal)
1da177e4
LT		 2287 return -EINVAL;
2288
eee194e7 2289 printk(KERN_INFO "EXT3-fs: creating new journal on inode %u\n",
1da177e4
LT		 2290 journal_inum);
2291
952d9de1
BP		 2292 err = journal_create(journal);
2293 if (err) {
1da177e4
LT		 2294 printk(KERN_ERR "EXT3-fs: error creating journal.\n");
2295 journal_destroy(journal);
2296 return -EIO;
2297 }
2298
2299 EXT3_SB(sb)->s_journal = journal;
2300
2301 ext3_update_dynamic_rev(sb);
2302 EXT3_SET_INCOMPAT_FEATURE(sb, EXT3_FEATURE_INCOMPAT_RECOVER);
2303 EXT3_SET_COMPAT_FEATURE(sb, EXT3_FEATURE_COMPAT_HAS_JOURNAL);
2304
2305 es->s_journal_inum = cpu_to_le32(journal_inum);
2306 sb->s_dirt = 1;
2307
2308 /* Make sure we flush the recovery flag to disk. */
2309 ext3_commit_super(sb, es, 1);
2310
2311 return 0;
2312}
2313
c4be0c1d
TS		 2314static int ext3_commit_super(struct super_block *sb,
2315 struct ext3_super_block *es,
1da177e4
LT		 2316 int sync)
2317{
2318 struct buffer_head *sbh = EXT3_SB(sb)->s_sbh;
c4be0c1d 2319 int error = 0;
1da177e4
LT		 2320
2321 if (!sbh)
c4be0c1d 2322 return error;
1da177e4
LT		 2323 es->s_wtime = cpu_to_le32(get_seconds());
2324 es->s_free_blocks_count = cpu_to_le32(ext3_count_free_blocks(sb));
2325 es->s_free_inodes_count = cpu_to_le32(ext3_count_free_inodes(sb));
2326 BUFFER_TRACE(sbh, "marking dirty");
2327 mark_buffer_dirty(sbh);
2328 if (sync)
c4be0c1d
TS		 2329 error = sync_dirty_buffer(sbh);
2330 return error;
1da177e4
LT		 2331}
2332
2333
2334/*
2335 * Have we just finished recovery? If so, and if we are mounting (or
2336 * remounting) the filesystem readonly, then we will end up with a
2337 * consistent fs on disk. Record that fact.
2338 */
2339static void ext3_mark_recovery_complete(struct super_block * sb,
2340 struct ext3_super_block * es)
2341{
2342 journal_t *journal = EXT3_SB(sb)->s_journal;
2343
2344 journal_lock_updates(journal);
2d7c820e
HK		 2345 if (journal_flush(journal) < 0)
2346 goto out;
2347
030703e4 2348 lock_super(sb);
1da177e4
LT		 2349 if (EXT3_HAS_INCOMPAT_FEATURE(sb, EXT3_FEATURE_INCOMPAT_RECOVER) &&
2350 sb->s_flags & MS_RDONLY) {
2351 EXT3_CLEAR_INCOMPAT_FEATURE(sb, EXT3_FEATURE_INCOMPAT_RECOVER);
2352 sb->s_dirt = 0;
2353 ext3_commit_super(sb, es, 1);
2354 }
030703e4 2355 unlock_super(sb);
2d7c820e
HK		 2356
2357out:
1da177e4
LT		 2358 journal_unlock_updates(journal);
2359}
2360
2361/*
2362 * If we are mounting (or read-write remounting) a filesystem whose journal
2363 * has recorded an error from a previous lifetime, move that error to the
2364 * main filesystem now.
2365 */
2366static void ext3_clear_journal_err(struct super_block * sb,
2367 struct ext3_super_block * es)
2368{
2369 journal_t *journal;
2370 int j_errno;
2371 const char *errstr;
2372
2373 journal = EXT3_SB(sb)->s_journal;
2374
2375 /*
2376 * Now check for any error status which may have been recorded in the
2377 * journal by a prior ext3_error() or ext3_abort()
2378 */
2379
2380 j_errno = journal_errno(journal);
2381 if (j_errno) {
2382 char nbuf[16];
2383
2384 errstr = ext3_decode_error(sb, j_errno, nbuf);
e05b6b52 2385 ext3_warning(sb, __func__, "Filesystem error recorded "
1da177e4 2386 "from previous mount: %s", errstr);
e05b6b52 2387 ext3_warning(sb, __func__, "Marking fs in need of "
1da177e4
LT		 2388 "filesystem check.");
2389
2390 EXT3_SB(sb)->s_mount_state |= EXT3_ERROR_FS;
2391 es->s_state |= cpu_to_le16(EXT3_ERROR_FS);
2392 ext3_commit_super (sb, es, 1);
2393
2394 journal_clear_err(journal);
2395 }
2396}
2397
2398/*
2399 * Force the running and committing transactions to commit,
2400 * and wait on the commit.
2401 */
2402int ext3_force_commit(struct super_block *sb)
2403{
2404 journal_t *journal;
2405 int ret;
2406
2407 if (sb->s_flags & MS_RDONLY)
2408 return 0;
2409
2410 journal = EXT3_SB(sb)->s_journal;
2411 sb->s_dirt = 0;
2412 ret = ext3_journal_force_commit(journal);
2413 return ret;
2414}
2415
2416/*
2417 * Ext3 always journals updates to the superblock itself, so we don't
2418 * have to propagate any other updates to the superblock on disk at this
6cdfcc27
TT		 2419 * point. (We can probably nuke this function altogether, and remove
2420 * any mention to sb->s_dirt in all of fs/ext3; eventual cleanup...)
1da177e4 2421 */
1da177e4
LT		 2422static void ext3_write_super (struct super_block * sb)
2423{
7892f2f4 2424 if (mutex_trylock(&sb->s_lock) != 0)
1da177e4
LT		 2425 BUG();
2426 sb->s_dirt = 0;
2427}
2428
2429static int ext3_sync_fs(struct super_block *sb, int wait)
2430{
1da177e4 2431 sb->s_dirt = 0;
c87591b7
AJ		 2432 if (wait)
2433 ext3_force_commit(sb);
2434 else
2435 journal_start_commit(EXT3_SB(sb)->s_journal, NULL);
2436
1da177e4
LT		 2437 return 0;
2438}
2439
2440/*
2441 * LVM calls this function before a (read-only) snapshot is created. This
2442 * gives us a chance to flush the journal completely and mark the fs clean.
2443 */
c4be0c1d 2444static int ext3_freeze(struct super_block *sb)
1da177e4 2445{
c4be0c1d
TS		 2446 int error = 0;
2447 journal_t *journal;
1da177e4
LT		 2448 sb->s_dirt = 0;
2449
2450 if (!(sb->s_flags & MS_RDONLY)) {
c4be0c1d 2451 journal = EXT3_SB(sb)->s_journal;
1da177e4
LT		 2452
2453 /* Now we set up the journal barrier. */
2454 journal_lock_updates(journal);
2d7c820e
HK		 2455
2456 /*
2457 * We don't want to clear needs_recovery flag when we failed
2458 * to flush the journal.
2459 */
c4be0c1d
TS		 2460 error = journal_flush(journal);
2461 if (error < 0)
2462 goto out;
1da177e4
LT		 2463
2464 /* Journal blocked and flushed, clear needs_recovery flag. */
2465 EXT3_CLEAR_INCOMPAT_FEATURE(sb, EXT3_FEATURE_INCOMPAT_RECOVER);
c4be0c1d
TS		 2466 error = ext3_commit_super(sb, EXT3_SB(sb)->s_es, 1);
2467 if (error)
2468 goto out;
1da177e4 2469 }
c4be0c1d
TS		 2470 return 0;
2471
2472out:
2473 journal_unlock_updates(journal);
2474 return error;
1da177e4
LT		 2475}
2476
2477/*
2478 * Called by LVM after the snapshot is done. We need to reset the RECOVER
2479 * flag here, even though the filesystem is not technically dirty yet.
2480 */
c4be0c1d 2481static int ext3_unfreeze(struct super_block *sb)
1da177e4
LT		 2482{
2483 if (!(sb->s_flags & MS_RDONLY)) {
2484 lock_super(sb);
2485 /* Reser the needs_recovery flag before the fs is unlocked. */
2486 EXT3_SET_INCOMPAT_FEATURE(sb, EXT3_FEATURE_INCOMPAT_RECOVER);
2487 ext3_commit_super(sb, EXT3_SB(sb)->s_es, 1);
2488 unlock_super(sb);
2489 journal_unlock_updates(EXT3_SB(sb)->s_journal);
2490 }
c4be0c1d 2491 return 0;
1da177e4
LT		 2492}
2493
2494static int ext3_remount (struct super_block * sb, int * flags, char * data)
2495{
2496 struct ext3_super_block * es;
2497 struct ext3_sb_info *sbi = EXT3_SB(sb);
43d23f90 2498 ext3_fsblk_t n_blocks_count = 0;
08c6a96f
JK		 2499 unsigned long old_sb_flags;
2500 struct ext3_mount_options old_opts;
2501 int err;
2502#ifdef CONFIG_QUOTA
2503 int i;
2504#endif
2505
2506 /* Store the original options */
2507 old_sb_flags = sb->s_flags;
2508 old_opts.s_mount_opt = sbi->s_mount_opt;
2509 old_opts.s_resuid = sbi->s_resuid;
2510 old_opts.s_resgid = sbi->s_resgid;
2511 old_opts.s_commit_interval = sbi->s_commit_interval;
2512#ifdef CONFIG_QUOTA
2513 old_opts.s_jquota_fmt = sbi->s_jquota_fmt;
2514 for (i = 0; i < MAXQUOTAS; i++)
2515 old_opts.s_qf_names[i] = sbi->s_qf_names[i];
2516#endif
1da177e4
LT		 2517
2518 /*
2519 * Allow the "check" option to be passed as a remount option.
2520 */
71b96257 2521 if (!parse_options(data, sb, NULL, NULL, &n_blocks_count, 1)) {
08c6a96f
JK		 2522 err = -EINVAL;
2523 goto restore_opts;
2524 }
1da177e4
LT		 2525
2526 if (sbi->s_mount_opt & EXT3_MOUNT_ABORT)
e05b6b52 2527 ext3_abort(sb, __func__, "Abort forced by user");
1da177e4
LT		 2528
2529 sb->s_flags = (sb->s_flags & ~MS_POSIXACL) |
2530 ((sbi->s_mount_opt & EXT3_MOUNT_POSIX_ACL) ? MS_POSIXACL : 0);
2531
2532 es = sbi->s_es;
2533
2534 ext3_init_journal_params(sb, sbi->s_journal);
2535
2536 if ((*flags & MS_RDONLY) != (sb->s_flags & MS_RDONLY) ||
2537 n_blocks_count > le32_to_cpu(es->s_blocks_count)) {
08c6a96f
JK		 2538 if (sbi->s_mount_opt & EXT3_MOUNT_ABORT) {
2539 err = -EROFS;
2540 goto restore_opts;
2541 }
1da177e4
LT		 2542
2543 if (*flags & MS_RDONLY) {
2544 /*
2545 * First of all, the unconditional stuff we have to do
2546 * to disable replay of the journal when we next remount
2547 */
2548 sb->s_flags |= MS_RDONLY;
2549
2550 /*
2551 * OK, test if we are remounting a valid rw partition
2552 * readonly, and if so set the rdonly flag and then
2553 * mark the partition as valid again.
2554 */
2555 if (!(es->s_state & cpu_to_le16(EXT3_VALID_FS)) &&
2556 (sbi->s_mount_state & EXT3_VALID_FS))
2557 es->s_state = cpu_to_le16(sbi->s_mount_state);
2558
030703e4
JK		 2559 /*
2560 * We have to unlock super so that we can wait for
2561 * transactions.
2562 */
2563 unlock_super(sb);
1da177e4 2564 ext3_mark_recovery_complete(sb, es);
030703e4 2565 lock_super(sb);
1da177e4
LT		 2566 } else {
2567 __le32 ret;
2568 if ((ret = EXT3_HAS_RO_COMPAT_FEATURE(sb,
2569 ~EXT3_FEATURE_RO_COMPAT_SUPP))) {
2570 printk(KERN_WARNING "EXT3-fs: %s: couldn't "
2571 "remount RDWR because of unsupported "
2572 "optional features (%x).\n",
2573 sb->s_id, le32_to_cpu(ret));
08c6a96f
JK		 2574 err = -EROFS;
2575 goto restore_opts;
1da177e4 2576 }
ea9a05a1
ES		 2577
2578 /*
2579 * If we have an unprocessed orphan list hanging
2580 * around from a previously readonly bdev mount,
2581 * require a full umount/remount for now.
2582 */
2583 if (es->s_last_orphan) {
2584 printk(KERN_WARNING "EXT3-fs: %s: couldn't "
2585 "remount RDWR because of unprocessed "
2586 "orphan inode list. Please "
2587 "umount/remount instead.\n",
2588 sb->s_id);
2589 err = -EINVAL;
2590 goto restore_opts;
2591 }
2592
1da177e4
LT		 2593 /*
2594 * Mounting a RDONLY partition read-write, so reread
2595 * and store the current valid flag. (It may have
2596 * been changed by e2fsck since we originally mounted
2597 * the partition.)
2598 */
2599 ext3_clear_journal_err(sb, es);
2600 sbi->s_mount_state = le16_to_cpu(es->s_state);
a4e4de36 2601 if ((err = ext3_group_extend(sb, es, n_blocks_count)))
08c6a96f 2602 goto restore_opts;
1da177e4
LT		 2603 if (!ext3_setup_super (sb, es, 0))
2604 sb->s_flags &= ~MS_RDONLY;
2605 }
2606 }
08c6a96f
JK		 2607#ifdef CONFIG_QUOTA
2608 /* Release old quota file names */
2609 for (i = 0; i < MAXQUOTAS; i++)
2610 if (old_opts.s_qf_names[i] &&
2611 old_opts.s_qf_names[i] != sbi->s_qf_names[i])
2612 kfree(old_opts.s_qf_names[i]);
2613#endif
1da177e4 2614 return 0;
08c6a96f
JK		 2615restore_opts:
2616 sb->s_flags = old_sb_flags;
2617 sbi->s_mount_opt = old_opts.s_mount_opt;
2618 sbi->s_resuid = old_opts.s_resuid;
2619 sbi->s_resgid = old_opts.s_resgid;
2620 sbi->s_commit_interval = old_opts.s_commit_interval;
2621#ifdef CONFIG_QUOTA
2622 sbi->s_jquota_fmt = old_opts.s_jquota_fmt;
2623 for (i = 0; i < MAXQUOTAS; i++) {
2624 if (sbi->s_qf_names[i] &&
2625 old_opts.s_qf_names[i] != sbi->s_qf_names[i])
2626 kfree(sbi->s_qf_names[i]);
2627 sbi->s_qf_names[i] = old_opts.s_qf_names[i];
2628 }
2629#endif
2630 return err;
1da177e4
LT		 2631}
2632
726c3342 2633static int ext3_statfs (struct dentry * dentry, struct kstatfs * buf)
1da177e4 2634{
726c3342 2635 struct super_block *sb = dentry->d_sb;
09fe316a
AT		 2636 struct ext3_sb_info *sbi = EXT3_SB(sb);
2637 struct ext3_super_block *es = sbi->s_es;
50ee0a32 2638 u64 fsid;
1da177e4 2639
a71ce8c6
BP		 2640 if (test_opt(sb, MINIX_DF)) {
2641 sbi->s_overhead_last = 0;
2642 } else if (sbi->s_blocks_last != le32_to_cpu(es->s_blocks_count)) {
2643 unsigned long ngroups = sbi->s_groups_count, i;
2644 ext3_fsblk_t overhead = 0;
1da177e4
LT		 2645 smp_rmb();
2646
2647 /*
a71ce8c6
BP		 2648 * Compute the overhead (FS structures). This is constant
2649 * for a given filesystem unless the number of block groups
2650 * changes so we cache the previous value until it does.
1da177e4
LT		 2651 */
2652
2653 /*
2654 * All of the blocks before first_data_block are
2655 * overhead
2656 */
2657 overhead = le32_to_cpu(es->s_first_data_block);
2658
2659 /*
2660 * Add the overhead attributed to the superblock and
2661 * block group descriptors. If the sparse superblocks
2662 * feature is turned on, then not all groups have this.
2663 */
2664 for (i = 0; i < ngroups; i++) {
2665 overhead += ext3_bg_has_super(sb, i) +
2666 ext3_bg_num_gdb(sb, i);
2667 cond_resched();
2668 }
2669
2670 /*
2671 * Every block group has an inode bitmap, a block
2672 * bitmap, and an inode table.
2673 */
a71ce8c6
BP		 2674 overhead += ngroups * (2 + sbi->s_itb_per_group);
2675 sbi->s_overhead_last = overhead;
2676 smp_wmb();
2677 sbi->s_blocks_last = le32_to_cpu(es->s_blocks_count);
1da177e4
LT		 2678 }
2679
2680 buf->f_type = EXT3_SUPER_MAGIC;
2681 buf->f_bsize = sb->s_blocksize;
a71ce8c6 2682 buf->f_blocks = le32_to_cpu(es->s_blocks_count) - sbi->s_overhead_last;
52d9f3b4 2683 buf->f_bfree = percpu_counter_sum_positive(&sbi->s_freeblocks_counter);
a71ce8c6 2684 es->s_free_blocks_count = cpu_to_le32(buf->f_bfree);
1da177e4
LT		 2685 buf->f_bavail = buf->f_bfree - le32_to_cpu(es->s_r_blocks_count);
2686 if (buf->f_bfree < le32_to_cpu(es->s_r_blocks_count))
2687 buf->f_bavail = 0;
2688 buf->f_files = le32_to_cpu(es->s_inodes_count);
52d9f3b4 2689 buf->f_ffree = percpu_counter_sum_positive(&sbi->s_freeinodes_counter);
a71ce8c6 2690 es->s_free_inodes_count = cpu_to_le32(buf->f_ffree);
1da177e4 2691 buf->f_namelen = EXT3_NAME_LEN;
50ee0a32
PE		 2692 fsid = le64_to_cpup((void *)es->s_uuid) ^
2693 le64_to_cpup((void *)es->s_uuid + sizeof(u64));
2694 buf->f_fsid.val[0] = fsid & 0xFFFFFFFFUL;
2695 buf->f_fsid.val[1] = (fsid >> 32) & 0xFFFFFFFFUL;
1da177e4
LT		 2696 return 0;
2697}
2698
2699/* Helper function for writing quotas on sync - we need to start transaction before quota file
2700 * is locked for write. Otherwise the are possible deadlocks:
2701 * Process 1 Process 2
2702 * ext3_create() quota_sync()
2703 * journal_start() write_dquot()
d3be915f
IM		 2704 * DQUOT_INIT() down(dqio_mutex)
2705 * down(dqio_mutex) journal_start()
1da177e4
LT		 2706 *
2707 */
2708
2709#ifdef CONFIG_QUOTA
2710
2711static inline struct inode *dquot_to_inode(struct dquot *dquot)
2712{
2713 return sb_dqopt(dquot->dq_sb)->files[dquot->dq_type];
2714}
2715
2716static int ext3_dquot_initialize(struct inode *inode, int type)
2717{
2718 handle_t *handle;
2719 int ret, err;
2720
2721 /* We may create quota structure so we need to reserve enough blocks */
1f54587b 2722 handle = ext3_journal_start(inode, 2*EXT3_QUOTA_INIT_BLOCKS(inode->i_sb));
1da177e4
LT		 2723 if (IS_ERR(handle))
2724 return PTR_ERR(handle);
2725 ret = dquot_initialize(inode, type);
2726 err = ext3_journal_stop(handle);
2727 if (!ret)
2728 ret = err;
2729 return ret;
2730}
2731
2732static int ext3_dquot_drop(struct inode *inode)
2733{
2734 handle_t *handle;
2735 int ret, err;
2736
2737 /* We may delete quota structure so we need to reserve enough blocks */
1f54587b 2738 handle = ext3_journal_start(inode, 2*EXT3_QUOTA_DEL_BLOCKS(inode->i_sb));
07c9938a
JK		 2739 if (IS_ERR(handle)) {
2740 /*
2741 * We call dquot_drop() anyway to at least release references
2742 * to quota structures so that umount does not hang.
2743 */
2744 dquot_drop(inode);
1da177e4 2745 return PTR_ERR(handle);
07c9938a 2746 }
1da177e4
LT		 2747 ret = dquot_drop(inode);
2748 err = ext3_journal_stop(handle);
2749 if (!ret)
2750 ret = err;
2751 return ret;
2752}
2753
2754static int ext3_write_dquot(struct dquot *dquot)
2755{
2756 int ret, err;
2757 handle_t *handle;
2758 struct inode *inode;
2759
2760 inode = dquot_to_inode(dquot);
2761 handle = ext3_journal_start(inode,
1f54587b 2762 EXT3_QUOTA_TRANS_BLOCKS(dquot->dq_sb));
1da177e4
LT		 2763 if (IS_ERR(handle))
2764 return PTR_ERR(handle);
2765 ret = dquot_commit(dquot);
2766 err = ext3_journal_stop(handle);
2767 if (!ret)
2768 ret = err;
2769 return ret;
2770}
2771
2772static int ext3_acquire_dquot(struct dquot *dquot)
2773{
2774 int ret, err;
2775 handle_t *handle;
2776
2777 handle = ext3_journal_start(dquot_to_inode(dquot),
1f54587b 2778 EXT3_QUOTA_INIT_BLOCKS(dquot->dq_sb));
1da177e4
LT		 2779 if (IS_ERR(handle))
2780 return PTR_ERR(handle);
2781 ret = dquot_acquire(dquot);
2782 err = ext3_journal_stop(handle);
2783 if (!ret)
2784 ret = err;
2785 return ret;
2786}
2787
2788static int ext3_release_dquot(struct dquot *dquot)
2789{
2790 int ret, err;
2791 handle_t *handle;
2792
2793 handle = ext3_journal_start(dquot_to_inode(dquot),
1f54587b 2794 EXT3_QUOTA_DEL_BLOCKS(dquot->dq_sb));
9c3013e9
JK		 2795 if (IS_ERR(handle)) {
2796 /* Release dquot anyway to avoid endless cycle in dqput() */
2797 dquot_release(dquot);
1da177e4 2798 return PTR_ERR(handle);
9c3013e9 2799 }
1da177e4
LT		 2800 ret = dquot_release(dquot);
2801 err = ext3_journal_stop(handle);
2802 if (!ret)
2803 ret = err;
2804 return ret;
2805}
2806
2807static int ext3_mark_dquot_dirty(struct dquot *dquot)
2808{
99aeaf63 2809 /* Are we journaling quotas? */
1da177e4
LT		 2810 if (EXT3_SB(dquot->dq_sb)->s_qf_names[USRQUOTA] ||
2811 EXT3_SB(dquot->dq_sb)->s_qf_names[GRPQUOTA]) {
2812 dquot_mark_dquot_dirty(dquot);
2813 return ext3_write_dquot(dquot);
2814 } else {
2815 return dquot_mark_dquot_dirty(dquot);
2816 }
2817}
2818
2819static int ext3_write_info(struct super_block *sb, int type)
2820{
2821 int ret, err;
2822 handle_t *handle;
2823
2824 /* Data block + inode block */
2825 handle = ext3_journal_start(sb->s_root->d_inode, 2);
2826 if (IS_ERR(handle))
2827 return PTR_ERR(handle);
2828 ret = dquot_commit_info(sb, type);
2829 err = ext3_journal_stop(handle);
2830 if (!ret)
2831 ret = err;
2832 return ret;
2833}
2834
2835/*
2836 * Turn on quotas during mount time - we need to find
2837 * the quota file and such...
2838 */
2839static int ext3_quota_on_mount(struct super_block *sb, int type)
2840{
84de856e
CH		 2841 return vfs_quota_on_mount(sb, EXT3_SB(sb)->s_qf_names[type],
2842 EXT3_SB(sb)->s_jquota_fmt, type);
1da177e4
LT		 2843}
2844
2845/*
2846 * Standard function to be called on quota_on
2847 */
2848static int ext3_quota_on(struct super_block *sb, int type, int format_id,
8264613d 2849 char *name, int remount)
1da177e4
LT		 2850{
2851 int err;
8264613d 2852 struct path path;
1da177e4 2853
1f54587b
JK		 2854 if (!test_opt(sb, QUOTA))
2855 return -EINVAL;
8264613d 2856 /* When remounting, no checks are needed and in fact, name is NULL */
9cfe7b90 2857 if (remount)
8264613d 2858 return vfs_quota_on(sb, type, format_id, name, remount);
9cfe7b90 2859
8264613d 2860 err = kern_path(name, LOOKUP_FOLLOW, &path);
1da177e4
LT		 2861 if (err)
2862 return err;
9cfe7b90 2863
1da177e4 2864 /* Quotafile not on the same filesystem? */
8264613d
AV		 2865 if (path.mnt->mnt_sb != sb) {
2866 path_put(&path);
1da177e4
LT		 2867 return -EXDEV;
2868 }
9cfe7b90
JK		 2869 /* Journaling quota? */
2870 if (EXT3_SB(sb)->s_qf_names[type]) {
2871 /* Quotafile not of fs root? */
8264613d 2872 if (path.dentry->d_parent != sb->s_root)
9cfe7b90
JK		 2873 printk(KERN_WARNING
2874 "EXT3-fs: Quota file not on filesystem root. "
2875 "Journaled quota will not work.\n");
2876 }
2877
2878 /*
2879 * When we journal data on quota file, we have to flush journal to see
2880 * all updates to the file when we bypass pagecache...
2881 */
8264613d 2882 if (ext3_should_journal_data(path.dentry->d_inode)) {
9cfe7b90
JK		 2883 /*
2884 * We don't need to lock updates but journal_flush() could
2885 * otherwise be livelocked...
2886 */
2887 journal_lock_updates(EXT3_SB(sb)->s_journal);
2d7c820e 2888 err = journal_flush(EXT3_SB(sb)->s_journal);
9cfe7b90 2889 journal_unlock_updates(EXT3_SB(sb)->s_journal);
2d7c820e 2890 if (err) {
12e1ec9f 2891 path_put(&path);
2d7c820e
HK		 2892 return err;
2893 }
9cfe7b90
JK		 2894 }
2895
8264613d
AV		 2896 err = vfs_quota_on_path(sb, type, format_id, &path);
2897 path_put(&path);
77e69dac 2898 return err;
1da177e4
LT		 2899}
2900
2901/* Read data from quotafile - avoid pagecache and such because we cannot afford
2902 * acquiring the locks... As quota files are never truncated and quota code
2903 * itself serializes the operations (and noone else should touch the files)
2904 * we don't have to be afraid of races */
2905static ssize_t ext3_quota_read(struct super_block *sb, int type, char *data,
2906 size_t len, loff_t off)
2907{
2908 struct inode *inode = sb_dqopt(sb)->files[type];
2909 sector_t blk = off >> EXT3_BLOCK_SIZE_BITS(sb);
2910 int err = 0;
2911 int offset = off & (sb->s_blocksize - 1);
2912 int tocopy;
2913 size_t toread;
2914 struct buffer_head *bh;
2915 loff_t i_size = i_size_read(inode);
2916
2917 if (off > i_size)
2918 return 0;
2919 if (off+len > i_size)
2920 len = i_size-off;
2921 toread = len;
2922 while (toread > 0) {
2923 tocopy = sb->s_blocksize - offset < toread ?
2924 sb->s_blocksize - offset : toread;
2925 bh = ext3_bread(NULL, inode, blk, 0, &err);
2926 if (err)
2927 return err;
2928 if (!bh) /* A hole? */
2929 memset(data, 0, tocopy);
2930 else
2931 memcpy(data, bh->b_data+offset, tocopy);
2932 brelse(bh);
2933 offset = 0;
2934 toread -= tocopy;
2935 data += tocopy;
2936 blk++;
2937 }
2938 return len;
2939}
2940
2941/* Write to quotafile (we know the transaction is already started and has
2942 * enough credits) */
2943static ssize_t ext3_quota_write(struct super_block *sb, int type,
2944 const char *data, size_t len, loff_t off)
2945{
2946 struct inode *inode = sb_dqopt(sb)->files[type];
2947 sector_t blk = off >> EXT3_BLOCK_SIZE_BITS(sb);
2948 int err = 0;
2949 int offset = off & (sb->s_blocksize - 1);
2950 int tocopy;
2951 int journal_quota = EXT3_SB(sb)->s_qf_names[type] != NULL;
2952 size_t towrite = len;
2953 struct buffer_head *bh;
2954 handle_t *handle = journal_current_handle();
2955
9c3013e9
JK		 2956 if (!handle) {
2957 printk(KERN_WARNING "EXT3-fs: Quota write (off=%Lu, len=%Lu)"
2958 " cancelled because transaction is not started.\n",
2959 (unsigned long long)off, (unsigned long long)len);
2960 return -EIO;
2961 }
5c81a419 2962 mutex_lock_nested(&inode->i_mutex, I_MUTEX_QUOTA);
1da177e4
LT		 2963 while (towrite > 0) {
2964 tocopy = sb->s_blocksize - offset < towrite ?
2965 sb->s_blocksize - offset : towrite;
2966 bh = ext3_bread(handle, inode, blk, 1, &err);
2967 if (!bh)
2968 goto out;
2969 if (journal_quota) {
2970 err = ext3_journal_get_write_access(handle, bh);
2971 if (err) {
2972 brelse(bh);
2973 goto out;
2974 }
2975 }
2976 lock_buffer(bh);
2977 memcpy(bh->b_data+offset, data, tocopy);
2978 flush_dcache_page(bh->b_page);
2979 unlock_buffer(bh);
2980 if (journal_quota)
2981 err = ext3_journal_dirty_metadata(handle, bh);
2982 else {
2983 /* Always do at least ordered writes for quotas */
2984 err = ext3_journal_dirty_data(handle, bh);
2985 mark_buffer_dirty(bh);
2986 }
2987 brelse(bh);
2988 if (err)
2989 goto out;
2990 offset = 0;
2991 towrite -= tocopy;
2992 data += tocopy;
2993 blk++;
2994 }
2995out:
f5c8f7da
JK		 2996 if (len == towrite) {
2997 mutex_unlock(&inode->i_mutex);
1da177e4 2998 return err;
f5c8f7da 2999 }
1da177e4
LT		 3000 if (inode->i_size < off+len-towrite) {
3001 i_size_write(inode, off+len-towrite);
3002 EXT3_I(inode)->i_disksize = inode->i_size;
3003 }
3004 inode->i_version++;
3005 inode->i_mtime = inode->i_ctime = CURRENT_TIME;
3006 ext3_mark_inode_dirty(handle, inode);
1b1dcc1b 3007 mutex_unlock(&inode->i_mutex);
1da177e4
LT		 3008 return len - towrite;
3009}
3010
3011#endif
3012
454e2398
DH		 3013static int ext3_get_sb(struct file_system_type *fs_type,
3014 int flags, const char *dev_name, void *data, struct vfsmount *mnt)
1da177e4 3015{
454e2398 3016 return get_sb_bdev(fs_type, flags, dev_name, data, ext3_fill_super, mnt);
1da177e4
LT		 3017}
3018
3019static struct file_system_type ext3_fs_type = {
3020 .owner = THIS_MODULE,
3021 .name = "ext3",
3022 .get_sb = ext3_get_sb,
3023 .kill_sb = kill_block_super,
3024 .fs_flags = FS_REQUIRES_DEV,
3025};
3026
3027static int __init init_ext3_fs(void)
3028{
3029 int err = init_ext3_xattr();
3030 if (err)
3031 return err;
3032 err = init_inodecache();
3033 if (err)
3034 goto out1;
3035 err = register_filesystem(&ext3_fs_type);
3036 if (err)
3037 goto out;
3038 return 0;
3039out:
3040 destroy_inodecache();
3041out1:
e9ad5620 3042 exit_ext3_xattr();
1da177e4
LT		 3043 return err;
3044}
3045
3046static void __exit exit_ext3_fs(void)
3047{
3048 unregister_filesystem(&ext3_fs_type);
3049 destroy_inodecache();
3050 exit_ext3_xattr();
3051}
3052
3053MODULE_AUTHOR("Remy Card, Stephen Tweedie, Andrew Morton, Andreas Dilger, Theodore Ts'o and others");
3054MODULE_DESCRIPTION("Second Extended Filesystem with journaling extensions");
3055MODULE_LICENSE("GPL");
3056module_init(init_ext3_fs)
3057module_exit(exit_ext3_fs)
Linux 6.x block layer and io_uring tree(s)