projects / deliverable / linux.git / blame
? search:
summary | shortlog | log | commit | commitdiff | tree
blob | blame (incremental) | history | HEAD
configfs: use capped length for ->store_attribute()
[deliverable/linux.git] / fs / ext3 / super.c
CommitLineData
1da177e4
LT		 1/*
2 * linux/fs/ext3/super.c
3 *
4 * Copyright (C) 1992, 1993, 1994, 1995
5 * Remy Card (card@masi.ibp.fr)
6 * Laboratoire MASI - Institut Blaise Pascal
7 * Universite Pierre et Marie Curie (Paris VI)
8 *
9 * from
10 *
11 * linux/fs/minix/inode.c
12 *
13 * Copyright (C) 1991, 1992 Linus Torvalds
14 *
15 * Big-endian to little-endian byte-swapping/bitmaps by
16 * David S. Miller (davem@caip.rutgers.edu), 1995
17 */
18
1da177e4 19#include <linux/module.h>
1da177e4
LT		 20#include <linux/blkdev.h>
21#include <linux/parser.h>
a5694255 22#include <linux/exportfs.h>
4613ad18 23#include <linux/statfs.h>
1da177e4
LT		 24#include <linux/random.h>
25#include <linux/mount.h>
1da177e4 26#include <linux/quotaops.h>
8fc2751b 27#include <linux/seq_file.h>
3fc74269 28#include <linux/log2.h>
d71bc6db 29#include <linux/cleancache.h>
381be254 30
1da177e4 31#include <asm/uaccess.h>
381be254 32
4613ad18
AV		 33#define CREATE_TRACE_POINTS
34
35#include "ext3.h"
1da177e4
LT		 36#include "xattr.h"
37#include "acl.h"
381be254 38#include "namei.h"
1da177e4 39
bbae8bcc
LT		 40#ifdef CONFIG_EXT3_DEFAULTS_TO_ORDERED
41 #define EXT3_MOUNT_DEFAULT_DATA_MODE EXT3_MOUNT_ORDERED_DATA
42#else
43 #define EXT3_MOUNT_DEFAULT_DATA_MODE EXT3_MOUNT_WRITEBACK_DATA
44#endif
45
71b96257
JL		 46static int ext3_load_journal(struct super_block *, struct ext3_super_block *,
47 unsigned long journal_devnum);
1da177e4 48static int ext3_create_journal(struct super_block *, struct ext3_super_block *,
eee194e7 49 unsigned int);
c4be0c1d
TS		 50static int ext3_commit_super(struct super_block *sb,
51 struct ext3_super_block *es,
1da177e4
LT		 52 int sync);
53static void ext3_mark_recovery_complete(struct super_block * sb,
54 struct ext3_super_block * es);
55static void ext3_clear_journal_err(struct super_block * sb,
56 struct ext3_super_block * es);
57static int ext3_sync_fs(struct super_block *sb, int wait);
58static const char *ext3_decode_error(struct super_block * sb, int errno,
59 char nbuf[16]);
60static int ext3_remount (struct super_block * sb, int * flags, char * data);
726c3342 61static int ext3_statfs (struct dentry * dentry, struct kstatfs * buf);
c4be0c1d 62static int ext3_unfreeze(struct super_block *sb);
c4be0c1d 63static int ext3_freeze(struct super_block *sb);
1da177e4 64
ae6ddcc5 65/*
1da177e4 66 * Wrappers for journal_start/end.
1da177e4
LT		 67 */
68handle_t *ext3_journal_start_sb(struct super_block *sb, int nblocks)
69{
70 journal_t *journal;
71
72 if (sb->s_flags & MS_RDONLY)
73 return ERR_PTR(-EROFS);
74
75 /* Special case here: if the journal has aborted behind our
76 * backs (eg. EIO in the commit thread), then we still need to
77 * take the FS itself readonly cleanly. */
78 journal = EXT3_SB(sb)->s_journal;
79 if (is_journal_aborted(journal)) {
e05b6b52 80 ext3_abort(sb, __func__,
1da177e4
LT		 81 "Detected aborted journal");
82 return ERR_PTR(-EROFS);
83 }
84
85 return journal_start(journal, nblocks);
86}
87
1da177e4
LT		 88int __ext3_journal_stop(const char *where, handle_t *handle)
89{
90 struct super_block *sb;
91 int err;
92 int rc;
93
94 sb = handle->h_transaction->t_journal->j_private;
95 err = handle->h_err;
96 rc = journal_stop(handle);
97
98 if (!err)
99 err = rc;
100 if (err)
101 __ext3_std_error(sb, where, err);
102 return err;
103}
104
105void ext3_journal_abort_handle(const char *caller, const char *err_fn,
106 struct buffer_head *bh, handle_t *handle, int err)
107{
108 char nbuf[16];
109 const char *errstr = ext3_decode_error(NULL, err, nbuf);
110
111 if (bh)
112 BUFFER_TRACE(bh, "abort");
113
114 if (!handle->h_err)
115 handle->h_err = err;
116
117 if (is_handle_aborted(handle))
118 return;
119
4cf46b67
AF		 120 printk(KERN_ERR "EXT3-fs: %s: aborting transaction: %s in %s\n",
121 caller, errstr, err_fn);
1da177e4
LT		 122
123 journal_abort_handle(handle);
124}
125
4cf46b67
AF		 126void ext3_msg(struct super_block *sb, const char *prefix,
127 const char *fmt, ...)
128{
99fbb1e2 129 struct va_format vaf;
4cf46b67
AF		 130 va_list args;
131
132 va_start(args, fmt);
99fbb1e2
JP		 133
134 vaf.fmt = fmt;
135 vaf.va = &args;
136
137 printk("%sEXT3-fs (%s): %pV\n", prefix, sb->s_id, &vaf);
138
4cf46b67
AF		 139 va_end(args);
140}
141
1da177e4
LT		 142/* Deal with the reporting of failure conditions on a filesystem such as
143 * inconsistencies detected or read IO failures.
144 *
145 * On ext2, we can store the error state of the filesystem in the
146 * superblock. That is not possible on ext3, because we may have other
147 * write ordering constraints on the superblock which prevent us from
148 * writing it out straight away; and given that the journal is about to
149 * be aborted, we can't rely on the current, or future, transactions to
150 * write out the superblock safely.
151 *
152 * We'll just use the journal_abort() error code to record an error in
d6b198bc 153 * the journal instead. On recovery, the journal will complain about
1da177e4
LT		 154 * that error until we've noted it down and cleared it.
155 */
156
157static void ext3_handle_error(struct super_block *sb)
158{
159 struct ext3_super_block *es = EXT3_SB(sb)->s_es;
160
161 EXT3_SB(sb)->s_mount_state |= EXT3_ERROR_FS;
162 es->s_state |= cpu_to_le16(EXT3_ERROR_FS);
163
164 if (sb->s_flags & MS_RDONLY)
165 return;
166
7543fc7b 167 if (!test_opt (sb, ERRORS_CONT)) {
1da177e4
LT		 168 journal_t *journal = EXT3_SB(sb)->s_journal;
169
e3c96435 170 set_opt(EXT3_SB(sb)->s_mount_opt, ABORT);
1da177e4
LT		 171 if (journal)
172 journal_abort(journal, -EIO);
173 }
7543fc7b 174 if (test_opt (sb, ERRORS_RO)) {
4cf46b67
AF		 175 ext3_msg(sb, KERN_CRIT,
176 "error: remounting filesystem read-only");
7543fc7b
VA		 177 sb->s_flags |= MS_RDONLY;
178 }
179 ext3_commit_super(sb, es, 1);
1da177e4 180 if (test_opt(sb, ERRORS_PANIC))
4cf46b67 181 panic("EXT3-fs (%s): panic forced after error\n",
1da177e4 182 sb->s_id);
1da177e4
LT		 183}
184
99fbb1e2
JP		 185void ext3_error(struct super_block *sb, const char *function,
186 const char *fmt, ...)
1da177e4 187{
99fbb1e2 188 struct va_format vaf;
1da177e4
LT		 189 va_list args;
190
191 va_start(args, fmt);
99fbb1e2
JP		 192
193 vaf.fmt = fmt;
194 vaf.va = &args;
195
196 printk(KERN_CRIT "EXT3-fs error (device %s): %s: %pV\n",
197 sb->s_id, function, &vaf);
198
1da177e4
LT		 199 va_end(args);
200
201 ext3_handle_error(sb);
202}
203
204static const char *ext3_decode_error(struct super_block * sb, int errno,
205 char nbuf[16])
206{
207 char *errstr = NULL;
208
209 switch (errno) {
210 case -EIO:
211 errstr = "IO failure";
212 break;
213 case -ENOMEM:
214 errstr = "Out of memory";
215 break;
216 case -EROFS:
217 if (!sb || EXT3_SB(sb)->s_journal->j_flags & JFS_ABORT)
218 errstr = "Journal has aborted";
219 else
220 errstr = "Readonly filesystem";
221 break;
222 default:
223 /* If the caller passed in an extra buffer for unknown
224 * errors, textualise them now. Else we just return
225 * NULL. */
226 if (nbuf) {
227 /* Check for truncated error codes... */
228 if (snprintf(nbuf, 16, "error %d", -errno) >= 0)
229 errstr = nbuf;
230 }
231 break;
232 }
233
234 return errstr;
235}
236
237/* __ext3_std_error decodes expected errors from journaling functions
238 * automatically and invokes the appropriate error response. */
239
240void __ext3_std_error (struct super_block * sb, const char * function,
241 int errno)
242{
243 char nbuf[16];
30121624 244 const char *errstr;
1da177e4 245
30121624
ST		 246 /* Special case: if the error is EROFS, and we're not already
247 * inside a transaction, then there's really no point in logging
248 * an error. */
249 if (errno == -EROFS && journal_current_handle() == NULL &&
250 (sb->s_flags & MS_RDONLY))
251 return;
252
253 errstr = ext3_decode_error(sb, errno, nbuf);
4cf46b67 254 ext3_msg(sb, KERN_CRIT, "error in %s: %s", function, errstr);
1da177e4
LT		 255
256 ext3_handle_error(sb);
257}
258
259/*
260 * ext3_abort is a much stronger failure handler than ext3_error. The
261 * abort function may be used to deal with unrecoverable failures such
262 * as journal IO errors or ENOMEM at a critical moment in log management.
263 *
264 * We unconditionally force the filesystem into an ABORT|READONLY state,
265 * unless the error response on the fs has been set to panic in which
266 * case we take the easy way out and panic immediately.
267 */
268
99fbb1e2
JP		 269void ext3_abort(struct super_block *sb, const char *function,
270 const char *fmt, ...)
1da177e4 271{
99fbb1e2 272 struct va_format vaf;
1da177e4
LT		 273 va_list args;
274
1da177e4 275 va_start(args, fmt);
99fbb1e2
JP		 276
277 vaf.fmt = fmt;
278 vaf.va = &args;
279
280 printk(KERN_CRIT "EXT3-fs (%s): error: %s: %pV\n",
281 sb->s_id, function, &vaf);
282
1da177e4
LT		 283 va_end(args);
284
285 if (test_opt(sb, ERRORS_PANIC))
4cf46b67 286 panic("EXT3-fs: panic from previous error\n");
1da177e4
LT		 287
288 if (sb->s_flags & MS_RDONLY)
289 return;
290
4cf46b67
AF		 291 ext3_msg(sb, KERN_CRIT,
292 "error: remounting filesystem read-only");
1da177e4
LT		 293 EXT3_SB(sb)->s_mount_state |= EXT3_ERROR_FS;
294 sb->s_flags |= MS_RDONLY;
e3c96435 295 set_opt(EXT3_SB(sb)->s_mount_opt, ABORT);
44d6f787
HK		 296 if (EXT3_SB(sb)->s_journal)
297 journal_abort(EXT3_SB(sb)->s_journal, -EIO);
1da177e4
LT		 298}
299
99fbb1e2
JP		 300void ext3_warning(struct super_block *sb, const char *function,
301 const char *fmt, ...)
1da177e4 302{
99fbb1e2 303 struct va_format vaf;
1da177e4
LT		 304 va_list args;
305
306 va_start(args, fmt);
99fbb1e2
JP		 307
308 vaf.fmt = fmt;
309 vaf.va = &args;
310
311 printk(KERN_WARNING "EXT3-fs (%s): warning: %s: %pV\n",
312 sb->s_id, function, &vaf);
313
1da177e4
LT		 314 va_end(args);
315}
316
317void ext3_update_dynamic_rev(struct super_block *sb)
318{
319 struct ext3_super_block *es = EXT3_SB(sb)->s_es;
320
321 if (le32_to_cpu(es->s_rev_level) > EXT3_GOOD_OLD_REV)
322 return;
323
4cf46b67
AF		 324 ext3_msg(sb, KERN_WARNING,
325 "warning: updating to rev %d because of "
326 "new feature flag, running e2fsck is recommended",
327 EXT3_DYNAMIC_REV);
1da177e4
LT		 328
329 es->s_first_ino = cpu_to_le32(EXT3_GOOD_OLD_FIRST_INO);
330 es->s_inode_size = cpu_to_le16(EXT3_GOOD_OLD_INODE_SIZE);
331 es->s_rev_level = cpu_to_le32(EXT3_DYNAMIC_REV);
332 /* leave es->s_feature_*compat flags alone */
333 /* es->s_uuid will be set by e2fsck if empty */
334
335 /*
336 * The rest of the superblock fields should be zero, and if not it
337 * means they are likely already in use, so leave them alone. We
338 * can leave it up to e2fsck to clean up any inconsistencies there.
339 */
340}
341
342/*
343 * Open the external journal device
344 */
4cf46b67 345static struct block_device *ext3_blkdev_get(dev_t dev, struct super_block *sb)
1da177e4
LT		 346{
347 struct block_device *bdev;
348 char b[BDEVNAME_SIZE];
349
d4d77629 350 bdev = blkdev_get_by_dev(dev, FMODE_READ|FMODE_WRITE|FMODE_EXCL, sb);
1da177e4
LT		 351 if (IS_ERR(bdev))
352 goto fail;
353 return bdev;
354
355fail:
8d0c2d10 356 ext3_msg(sb, KERN_ERR, "error: failed to open journal device %s: %ld",
4cf46b67
AF		 357 __bdevname(dev, b), PTR_ERR(bdev));
358
1da177e4
LT		 359 return NULL;
360}
361
362/*
363 * Release the journal device
364 */
4385bab1 365static void ext3_blkdev_put(struct block_device *bdev)
1da177e4 366{
4385bab1 367 blkdev_put(bdev, FMODE_READ|FMODE_WRITE|FMODE_EXCL);
1da177e4
LT		 368}
369
4385bab1 370static void ext3_blkdev_remove(struct ext3_sb_info *sbi)
1da177e4
LT		 371{
372 struct block_device *bdev;
1da177e4
LT		 373 bdev = sbi->journal_bdev;
374 if (bdev) {
4385bab1 375 ext3_blkdev_put(bdev);
1da177e4
LT		 376 sbi->journal_bdev = NULL;
377 }
1da177e4
LT		 378}
379
380static inline struct inode *orphan_list_entry(struct list_head *l)
381{
382 return &list_entry(l, struct ext3_inode_info, i_orphan)->vfs_inode;
383}
384
385static void dump_orphan_list(struct super_block *sb, struct ext3_sb_info *sbi)
386{
387 struct list_head *l;
388
4cf46b67 389 ext3_msg(sb, KERN_ERR, "error: sb orphan head is %d",
1da177e4
LT		 390 le32_to_cpu(sbi->s_es->s_last_orphan));
391
4cf46b67 392 ext3_msg(sb, KERN_ERR, "sb_info orphan list:");
1da177e4
LT		 393 list_for_each(l, &sbi->s_orphan) {
394 struct inode *inode = orphan_list_entry(l);
4cf46b67 395 ext3_msg(sb, KERN_ERR, " "
eee194e7 396 "inode %s:%lu at %p: mode %o, nlink %d, next %d\n",
1da177e4 397 inode->i_sb->s_id, inode->i_ino, inode,
ae6ddcc5 398 inode->i_mode, inode->i_nlink,
1da177e4
LT		 399 NEXT_ORPHAN(inode));
400 }
401}
402
403static void ext3_put_super (struct super_block * sb)
404{
405 struct ext3_sb_info *sbi = EXT3_SB(sb);
406 struct ext3_super_block *es = sbi->s_es;
44d6f787 407 int i, err;
1da177e4 408
e0ccfd95 409 dquot_disable(sb, -1, DQUOT_USAGE_ENABLED | DQUOT_LIMITS_ENABLED);
1da177e4 410 ext3_xattr_put_super(sb);
44d6f787
HK		 411 err = journal_destroy(sbi->s_journal);
412 sbi->s_journal = NULL;
413 if (err < 0)
2d7c820e 414 ext3_abort(sb, __func__, "Couldn't clean up the journal");
44d6f787 415
1da177e4
LT		 416 if (!(sb->s_flags & MS_RDONLY)) {
417 EXT3_CLEAR_INCOMPAT_FEATURE(sb, EXT3_FEATURE_INCOMPAT_RECOVER);
418 es->s_state = cpu_to_le16(sbi->s_mount_state);
419 BUFFER_TRACE(sbi->s_sbh, "marking dirty");
420 mark_buffer_dirty(sbi->s_sbh);
421 ext3_commit_super(sb, es, 1);
422 }
423
424 for (i = 0; i < sbi->s_gdb_count; i++)
425 brelse(sbi->s_group_desc[i]);
426 kfree(sbi->s_group_desc);
427 percpu_counter_destroy(&sbi->s_freeblocks_counter);
428 percpu_counter_destroy(&sbi->s_freeinodes_counter);
429 percpu_counter_destroy(&sbi->s_dirs_counter);
430 brelse(sbi->s_sbh);
431#ifdef CONFIG_QUOTA
432 for (i = 0; i < MAXQUOTAS; i++)
433 kfree(sbi->s_qf_names[i]);
434#endif
435
436 /* Debugging code just in case the in-memory inode orphan list
437 * isn't empty. The on-disk one can be non-empty if we've
438 * detected an error and taken the fs readonly, but the
439 * in-memory list had better be clean by this point. */
440 if (!list_empty(&sbi->s_orphan))
441 dump_orphan_list(sb, sbi);
442 J_ASSERT(list_empty(&sbi->s_orphan));
443
f98393a6 444 invalidate_bdev(sb->s_bdev);
1da177e4
LT		 445 if (sbi->journal_bdev && sbi->journal_bdev != sb->s_bdev) {
446 /*
447 * Invalidate the journal device's buffers. We don't want them
448 * floating about in memory - the physical journal device may
449 * hotswapped, and it breaks the `ro-after' testing code.
450 */
451 sync_blockdev(sbi->journal_bdev);
f98393a6 452 invalidate_bdev(sbi->journal_bdev);
1da177e4
LT		 453 ext3_blkdev_remove(sbi);
454 }
455 sb->s_fs_info = NULL;
5df096d6 456 kfree(sbi->s_blockgroup_lock);
1da177e4 457 kfree(sbi);
1da177e4
LT		 458}
459
e18b890b 460static struct kmem_cache *ext3_inode_cachep;
1da177e4
LT		 461
462/*
463 * Called inside transaction, so use GFP_NOFS
464 */
465static struct inode *ext3_alloc_inode(struct super_block *sb)
466{
467 struct ext3_inode_info *ei;
468
e6b4f8da 469 ei = kmem_cache_alloc(ext3_inode_cachep, GFP_NOFS);
1da177e4
LT		 470 if (!ei)
471 return NULL;
1da177e4
LT		 472 ei->i_block_alloc_info = NULL;
473 ei->vfs_inode.i_version = 1;
fe8bc91c
JK		 474 atomic_set(&ei->i_datasync_tid, 0);
475 atomic_set(&ei->i_sync_tid, 0);
1da177e4
LT		 476 return &ei->vfs_inode;
477}
478
785c4bcc
LC		 479static int ext3_drop_inode(struct inode *inode)
480{
481 int drop = generic_drop_inode(inode);
482
483 trace_ext3_drop_inode(inode, drop);
484 return drop;
485}
486
fa0d7e3d
NP		 487static void ext3_i_callback(struct rcu_head *head)
488{
489 struct inode *inode = container_of(head, struct inode, i_rcu);
fa0d7e3d
NP		 490 kmem_cache_free(ext3_inode_cachep, EXT3_I(inode));
491}
492
1da177e4
LT		 493static void ext3_destroy_inode(struct inode *inode)
494{
9f7dd93d
VA		 495 if (!list_empty(&(EXT3_I(inode)->i_orphan))) {
496 printk("EXT3 Inode %p: orphan list check failed!\n",
497 EXT3_I(inode));
498 print_hex_dump(KERN_INFO, "", DUMP_PREFIX_ADDRESS, 16, 4,
499 EXT3_I(inode), sizeof(struct ext3_inode_info),
500 false);
501 dump_stack();
502 }
fa0d7e3d 503 call_rcu(&inode->i_rcu, ext3_i_callback);
1da177e4
LT		 504}
505
51cc5068 506static void init_once(void *foo)
1da177e4
LT		 507{
508 struct ext3_inode_info *ei = (struct ext3_inode_info *) foo;
509
a35afb83 510 INIT_LIST_HEAD(&ei->i_orphan);
1da177e4 511#ifdef CONFIG_EXT3_FS_XATTR
a35afb83 512 init_rwsem(&ei->xattr_sem);
1da177e4 513#endif
a35afb83
CL		 514 mutex_init(&ei->truncate_mutex);
515 inode_init_once(&ei->vfs_inode);
1da177e4 516}
ae6ddcc5 517
1da177e4
LT		 518static int init_inodecache(void)
519{
520 ext3_inode_cachep = kmem_cache_create("ext3_inode_cache",
521 sizeof(struct ext3_inode_info),
fffb60f9
PJ		 522 0, (SLAB_RECLAIM_ACCOUNT|
523 SLAB_MEM_SPREAD),
20c2df83 524 init_once);
1da177e4
LT		 525 if (ext3_inode_cachep == NULL)
526 return -ENOMEM;
527 return 0;
528}
529
530static void destroy_inodecache(void)
531{
8c0a8537
KS		 532 /*
533 * Make sure all delayed rcu free inodes are flushed before we
534 * destroy cache.
535 */
536 rcu_barrier();
1a1d92c1 537 kmem_cache_destroy(ext3_inode_cachep);
1da177e4
LT		 538}
539
6cd37cda 540static inline void ext3_show_quota_options(struct seq_file *seq, struct super_block *sb)
8fc2751b 541{
6cd37cda 542#if defined(CONFIG_QUOTA)
275abf5b 543 struct ext3_sb_info *sbi = EXT3_SB(sb);
8fc2751b 544
1aeec434
JK		 545 if (sbi->s_jquota_fmt) {
546 char *fmtname = "";
547
548 switch (sbi->s_jquota_fmt) {
549 case QFMT_VFS_OLD:
550 fmtname = "vfsold";
551 break;
552 case QFMT_VFS_V0:
553 fmtname = "vfsv0";
554 break;
555 case QFMT_VFS_V1:
556 fmtname = "vfsv1";
557 break;
558 }
559 seq_printf(seq, ",jqfmt=%s", fmtname);
560 }
8fc2751b
MB		 561
562 if (sbi->s_qf_names[USRQUOTA])
563 seq_printf(seq, ",usrjquota=%s", sbi->s_qf_names[USRQUOTA]);
564
565 if (sbi->s_qf_names[GRPQUOTA])
566 seq_printf(seq, ",grpjquota=%s", sbi->s_qf_names[GRPQUOTA]);
567
e3c96435 568 if (test_opt(sb, USRQUOTA))
8fc2751b
MB		 569 seq_puts(seq, ",usrquota");
570
e3c96435 571 if (test_opt(sb, GRPQUOTA))
8fc2751b
MB		 572 seq_puts(seq, ",grpquota");
573#endif
6cd37cda
PO		 574}
575
3c4cec65
JK		 576static char *data_mode_string(unsigned long mode)
577{
578 switch (mode) {
579 case EXT3_MOUNT_JOURNAL_DATA:
580 return "journal";
581 case EXT3_MOUNT_ORDERED_DATA:
582 return "ordered";
583 case EXT3_MOUNT_WRITEBACK_DATA:
584 return "writeback";
585 }
586 return "unknown";
587}
588
571beed8
MS		 589/*
590 * Show an option if
591 * - it's set to a non-default value OR
592 * - if the per-sb default is different from the global default
593 */
34c80b1d 594static int ext3_show_options(struct seq_file *seq, struct dentry *root)
6cd37cda 595{
34c80b1d 596 struct super_block *sb = root->d_sb;
571beed8
MS		 597 struct ext3_sb_info *sbi = EXT3_SB(sb);
598 struct ext3_super_block *es = sbi->s_es;
599 unsigned long def_mount_opts;
600
601 def_mount_opts = le32_to_cpu(es->s_default_mount_opts);
602
603 if (sbi->s_sb_block != 1)
604 seq_printf(seq, ",sb=%lu", sbi->s_sb_block);
605 if (test_opt(sb, MINIX_DF))
606 seq_puts(seq, ",minixdf");
607 if (test_opt(sb, GRPID))
608 seq_puts(seq, ",grpid");
609 if (!test_opt(sb, GRPID) && (def_mount_opts & EXT3_DEFM_BSDGROUPS))
610 seq_puts(seq, ",nogrpid");
1523299d 611 if (!uid_eq(sbi->s_resuid, make_kuid(&init_user_ns, EXT3_DEF_RESUID)) ||
571beed8 612 le16_to_cpu(es->s_def_resuid) != EXT3_DEF_RESUID) {
1523299d
EB		 613 seq_printf(seq, ",resuid=%u",
614 from_kuid_munged(&init_user_ns, sbi->s_resuid));
571beed8 615 }
1523299d 616 if (!gid_eq(sbi->s_resgid, make_kgid(&init_user_ns, EXT3_DEF_RESGID)) ||
571beed8 617 le16_to_cpu(es->s_def_resgid) != EXT3_DEF_RESGID) {
1523299d
EB		 618 seq_printf(seq, ",resgid=%u",
619 from_kgid_munged(&init_user_ns, sbi->s_resgid));
571beed8 620 }
1eca93f9 621 if (test_opt(sb, ERRORS_RO)) {
571beed8
MS		 622 int def_errors = le16_to_cpu(es->s_errors);
623
624 if (def_errors == EXT3_ERRORS_PANIC ||
1eca93f9
AK		 625 def_errors == EXT3_ERRORS_CONTINUE) {
626 seq_puts(seq, ",errors=remount-ro");
571beed8
MS		 627 }
628 }
1eca93f9
AK		 629 if (test_opt(sb, ERRORS_CONT))
630 seq_puts(seq, ",errors=continue");
571beed8
MS		 631 if (test_opt(sb, ERRORS_PANIC))
632 seq_puts(seq, ",errors=panic");
633 if (test_opt(sb, NO_UID32))
634 seq_puts(seq, ",nouid32");
635 if (test_opt(sb, DEBUG))
636 seq_puts(seq, ",debug");
571beed8
MS		 637#ifdef CONFIG_EXT3_FS_XATTR
638 if (test_opt(sb, XATTR_USER))
639 seq_puts(seq, ",user_xattr");
640 if (!test_opt(sb, XATTR_USER) &&
641 (def_mount_opts & EXT3_DEFM_XATTR_USER)) {
642 seq_puts(seq, ",nouser_xattr");
643 }
644#endif
645#ifdef CONFIG_EXT3_FS_POSIX_ACL
646 if (test_opt(sb, POSIX_ACL))
647 seq_puts(seq, ",acl");
648 if (!test_opt(sb, POSIX_ACL) && (def_mount_opts & EXT3_DEFM_ACL))
649 seq_puts(seq, ",noacl");
650#endif
651 if (!test_opt(sb, RESERVATION))
652 seq_puts(seq, ",noreservation");
653 if (sbi->s_commit_interval) {
654 seq_printf(seq, ",commit=%u",
655 (unsigned) (sbi->s_commit_interval / HZ));
656 }
0636c73e
ES		 657
658 /*
659 * Always display barrier state so it's clear what the status is.
660 */
661 seq_puts(seq, ",barrier=");
662 seq_puts(seq, test_opt(sb, BARRIER) ? "1" : "0");
e3c96435 663 seq_printf(seq, ",data=%s", data_mode_string(test_opt(sb, DATA_FLAGS)));
0e4fb5e2
HK		 664 if (test_opt(sb, DATA_ERR_ABORT))
665 seq_puts(seq, ",data_err=abort");
666
dee1d3b6
ES		 667 if (test_opt(sb, NOLOAD))
668 seq_puts(seq, ",norecovery");
669
6cd37cda 670 ext3_show_quota_options(seq, sb);
8fc2751b
MB		 671
672 return 0;
673}
1da177e4 674
fdb36673 675
74af0baa
CH		 676static struct inode *ext3_nfs_get_inode(struct super_block *sb,
677 u64 ino, u32 generation)
fdb36673 678{
fdb36673 679 struct inode *inode;
fdb36673
N		 680
681 if (ino < EXT3_FIRST_INO(sb) && ino != EXT3_ROOT_INO)
682 return ERR_PTR(-ESTALE);
683 if (ino > le32_to_cpu(EXT3_SB(sb)->s_es->s_inodes_count))
684 return ERR_PTR(-ESTALE);
685
686 /* iget isn't really right if the inode is currently unallocated!!
687 *
688 * ext3_read_inode will return a bad_inode if the inode had been
689 * deleted, so we should be safe.
690 *
691 * Currently we don't know the generation for parent directory, so
692 * a generation of 0 means "accept any"
693 */
473043dc
DH		 694 inode = ext3_iget(sb, ino);
695 if (IS_ERR(inode))
696 return ERR_CAST(inode);
697 if (generation && inode->i_generation != generation) {
fdb36673
N		 698 iput(inode);
699 return ERR_PTR(-ESTALE);
700 }
74af0baa
CH		 701
702 return inode;
703}
704
705static struct dentry *ext3_fh_to_dentry(struct super_block *sb, struct fid *fid,
706 int fh_len, int fh_type)
707{
708 return generic_fh_to_dentry(sb, fid, fh_len, fh_type,
709 ext3_nfs_get_inode);
710}
711
712static struct dentry *ext3_fh_to_parent(struct super_block *sb, struct fid *fid,
713 int fh_len, int fh_type)
714{
715 return generic_fh_to_parent(sb, fid, fh_len, fh_type,
716 ext3_nfs_get_inode);
fdb36673
N		 717}
718
6b082b53
TO		 719/*
720 * Try to release metadata pages (indirect blocks, directories) which are
721 * mapped via the block device. Since these pages could have journal heads
722 * which would prevent try_to_free_buffers() from freeing them, we must use
723 * jbd layer's try_to_free_buffers() function to release them.
724 */
725static int bdev_try_to_free_page(struct super_block *sb, struct page *page,
726 gfp_t wait)
727{
728 journal_t *journal = EXT3_SB(sb)->s_journal;
729
730 WARN_ON(PageChecked(page));
731 if (!page_has_buffers(page))
732 return 0;
733 if (journal)
734 return journal_try_to_free_buffers(journal, page,
735 wait & ~__GFP_WAIT);
736 return try_to_free_buffers(page);
737}
738
8fc2751b 739#ifdef CONFIG_QUOTA
1da177e4
LT		 740#define QTYPE2NAME(t) ((t)==USRQUOTA?"user":"group")
741#define QTYPE2MOPT(on, t) ((t)==USRQUOTA?((on)##USRJQUOTA):((on)##GRPJQUOTA))
742
1da177e4
LT		 743static int ext3_write_dquot(struct dquot *dquot);
744static int ext3_acquire_dquot(struct dquot *dquot);
745static int ext3_release_dquot(struct dquot *dquot);
746static int ext3_mark_dquot_dirty(struct dquot *dquot);
747static int ext3_write_info(struct super_block *sb, int type);
2fd83a4f 748static int ext3_quota_on(struct super_block *sb, int type, int format_id,
f00c9e44 749 struct path *path);
1da177e4
LT		 750static int ext3_quota_on_mount(struct super_block *sb, int type);
751static ssize_t ext3_quota_read(struct super_block *sb, int type, char *data,
752 size_t len, loff_t off);
753static ssize_t ext3_quota_write(struct super_block *sb, int type,
754 const char *data, size_t len, loff_t off);
755
61e225dc 756static const struct dquot_operations ext3_quota_operations = {
1da177e4
LT		 757 .write_dquot = ext3_write_dquot,
758 .acquire_dquot = ext3_acquire_dquot,
759 .release_dquot = ext3_release_dquot,
760 .mark_dirty = ext3_mark_dquot_dirty,
157091a2
JK		 761 .write_info = ext3_write_info,
762 .alloc_dquot = dquot_alloc,
763 .destroy_dquot = dquot_destroy,
1da177e4
LT		 764};
765
0d54b217 766static const struct quotactl_ops ext3_qctl_operations = {
1da177e4 767 .quota_on = ext3_quota_on,
287a8095
CH		 768 .quota_off = dquot_quota_off,
769 .quota_sync = dquot_quota_sync,
770 .get_info = dquot_get_dqinfo,
771 .set_info = dquot_set_dqinfo,
772 .get_dqblk = dquot_get_dqblk,
773 .set_dqblk = dquot_set_dqblk
1da177e4
LT		 774};
775#endif
776
ee9b6d61 777static const struct super_operations ext3_sops = {
1da177e4
LT		 778 .alloc_inode = ext3_alloc_inode,
779 .destroy_inode = ext3_destroy_inode,
1da177e4
LT		 780 .write_inode = ext3_write_inode,
781 .dirty_inode = ext3_dirty_inode,
785c4bcc 782 .drop_inode = ext3_drop_inode,
ac14a95b 783 .evict_inode = ext3_evict_inode,
1da177e4 784 .put_super = ext3_put_super,
1da177e4 785 .sync_fs = ext3_sync_fs,
c4be0c1d
TS		 786 .freeze_fs = ext3_freeze,
787 .unfreeze_fs = ext3_unfreeze,
1da177e4
LT		 788 .statfs = ext3_statfs,
789 .remount_fs = ext3_remount,
8fc2751b 790 .show_options = ext3_show_options,
1da177e4
LT		 791#ifdef CONFIG_QUOTA
792 .quota_read = ext3_quota_read,
793 .quota_write = ext3_quota_write,
794#endif
6b082b53 795 .bdev_try_to_free_page = bdev_try_to_free_page,
1da177e4
LT		 796};
797
39655164 798static const struct export_operations ext3_export_ops = {
74af0baa
CH		 799 .fh_to_dentry = ext3_fh_to_dentry,
800 .fh_to_parent = ext3_fh_to_parent,
1da177e4
LT		 801 .get_parent = ext3_get_parent,
802};
803
804enum {
805 Opt_bsd_df, Opt_minix_df, Opt_grpid, Opt_nogrpid,
806 Opt_resgid, Opt_resuid, Opt_sb, Opt_err_cont, Opt_err_panic, Opt_err_ro,
2860b733 807 Opt_nouid32, Opt_nocheck, Opt_debug, Opt_oldalloc, Opt_orlov,
1da177e4 808 Opt_user_xattr, Opt_nouser_xattr, Opt_acl, Opt_noacl,
ade1a29e 809 Opt_reservation, Opt_noreservation, Opt_noload, Opt_nobh, Opt_bh,
71b96257 810 Opt_commit, Opt_journal_update, Opt_journal_inum, Opt_journal_dev,
1da177e4 811 Opt_abort, Opt_data_journal, Opt_data_ordered, Opt_data_writeback,
0e4fb5e2 812 Opt_data_err_abort, Opt_data_err_ignore,
1da177e4 813 Opt_usrjquota, Opt_grpjquota, Opt_offusrjquota, Opt_offgrpjquota,
1aeec434 814 Opt_jqfmt_vfsold, Opt_jqfmt_vfsv0, Opt_jqfmt_vfsv1, Opt_quota,
0636c73e
ES		 815 Opt_noquota, Opt_ignore, Opt_barrier, Opt_nobarrier, Opt_err,
816 Opt_resize, Opt_usrquota, Opt_grpquota
1da177e4
LT		 817};
818
a447c093 819static const match_table_t tokens = {
1da177e4
LT		 820 {Opt_bsd_df, "bsddf"},
821 {Opt_minix_df, "minixdf"},
822 {Opt_grpid, "grpid"},
823 {Opt_grpid, "bsdgroups"},
824 {Opt_nogrpid, "nogrpid"},
825 {Opt_nogrpid, "sysvgroups"},
826 {Opt_resgid, "resgid=%u"},
827 {Opt_resuid, "resuid=%u"},
828 {Opt_sb, "sb=%u"},
829 {Opt_err_cont, "errors=continue"},
830 {Opt_err_panic, "errors=panic"},
831 {Opt_err_ro, "errors=remount-ro"},
832 {Opt_nouid32, "nouid32"},
833 {Opt_nocheck, "nocheck"},
834 {Opt_nocheck, "check=none"},
1da177e4
LT		 835 {Opt_debug, "debug"},
836 {Opt_oldalloc, "oldalloc"},
837 {Opt_orlov, "orlov"},
838 {Opt_user_xattr, "user_xattr"},
839 {Opt_nouser_xattr, "nouser_xattr"},
840 {Opt_acl, "acl"},
841 {Opt_noacl, "noacl"},
842 {Opt_reservation, "reservation"},
843 {Opt_noreservation, "noreservation"},
844 {Opt_noload, "noload"},
dee1d3b6 845 {Opt_noload, "norecovery"},
1da177e4 846 {Opt_nobh, "nobh"},
ade1a29e 847 {Opt_bh, "bh"},
1da177e4
LT		 848 {Opt_commit, "commit=%u"},
849 {Opt_journal_update, "journal=update"},
850 {Opt_journal_inum, "journal=%u"},
71b96257 851 {Opt_journal_dev, "journal_dev=%u"},
1da177e4
LT		 852 {Opt_abort, "abort"},
853 {Opt_data_journal, "data=journal"},
854 {Opt_data_ordered, "data=ordered"},
855 {Opt_data_writeback, "data=writeback"},
0e4fb5e2
HK		 856 {Opt_data_err_abort, "data_err=abort"},
857 {Opt_data_err_ignore, "data_err=ignore"},
1da177e4
LT		 858 {Opt_offusrjquota, "usrjquota="},
859 {Opt_usrjquota, "usrjquota=%s"},
860 {Opt_offgrpjquota, "grpjquota="},
861 {Opt_grpjquota, "grpjquota=%s"},
862 {Opt_jqfmt_vfsold, "jqfmt=vfsold"},
863 {Opt_jqfmt_vfsv0, "jqfmt=vfsv0"},
1aeec434 864 {Opt_jqfmt_vfsv1, "jqfmt=vfsv1"},
8fc2751b 865 {Opt_grpquota, "grpquota"},
1f54587b
JK		 866 {Opt_noquota, "noquota"},
867 {Opt_quota, "quota"},
8fc2751b 868 {Opt_usrquota, "usrquota"},
1da177e4 869 {Opt_barrier, "barrier=%u"},
0636c73e
ES		 870 {Opt_barrier, "barrier"},
871 {Opt_nobarrier, "nobarrier"},
1da177e4 872 {Opt_resize, "resize"},
92587216 873 {Opt_err, NULL},
1da177e4
LT		 874};
875
4cf46b67 876static ext3_fsblk_t get_sb_block(void **data, struct super_block *sb)
1da177e4 877{
e9ad5620
DK		 878 ext3_fsblk_t sb_block;
879 char *options = (char *) *data;
1da177e4
LT		 880
881 if (!options || strncmp(options, "sb=", 3) != 0)
882 return 1; /* Default location */
883 options += 3;
43d23f90 884 /*todo: use simple_strtoll with >32bit ext3 */
1da177e4
LT		 885 sb_block = simple_strtoul(options, &options, 0);
886 if (*options && *options != ',') {
8d0c2d10 887 ext3_msg(sb, KERN_ERR, "error: invalid sb specification: %s",
1da177e4
LT		 888 (char *) *data);
889 return 1;
890 }
891 if (*options == ',')
892 options++;
893 *data = (void *) options;
894 return sb_block;
895}
896
e1f5c67a
DM		 897#ifdef CONFIG_QUOTA
898static int set_qf_name(struct super_block *sb, int qtype, substring_t *args)
899{
900 struct ext3_sb_info *sbi = EXT3_SB(sb);
901 char *qname;
902
903 if (sb_any_quota_loaded(sb) &&
904 !sbi->s_qf_names[qtype]) {
905 ext3_msg(sb, KERN_ERR,
906 "Cannot change journaled "
907 "quota options when quota turned on");
908 return 0;
909 }
910 qname = match_strdup(args);
911 if (!qname) {
912 ext3_msg(sb, KERN_ERR,
913 "Not enough memory for storing quotafile name");
914 return 0;
915 }
f56426ae
JK		 916 if (sbi->s_qf_names[qtype]) {
917 int same = !strcmp(sbi->s_qf_names[qtype], qname);
918
e1f5c67a 919 kfree(qname);
f56426ae
JK		 920 if (!same) {
921 ext3_msg(sb, KERN_ERR,
922 "%s quota file already specified",
923 QTYPE2NAME(qtype));
924 }
925 return same;
e1f5c67a 926 }
f56426ae 927 if (strchr(qname, '/')) {
e1f5c67a
DM		 928 ext3_msg(sb, KERN_ERR,
929 "quotafile must be on filesystem root");
f56426ae 930 kfree(qname);
e1f5c67a
DM		 931 return 0;
932 }
f56426ae 933 sbi->s_qf_names[qtype] = qname;
e1f5c67a
DM		 934 set_opt(sbi->s_mount_opt, QUOTA);
935 return 1;
936}
937
938static int clear_qf_name(struct super_block *sb, int qtype) {
939
940 struct ext3_sb_info *sbi = EXT3_SB(sb);
941
942 if (sb_any_quota_loaded(sb) &&
943 sbi->s_qf_names[qtype]) {
944 ext3_msg(sb, KERN_ERR, "Cannot change journaled quota options"
945 " when quota turned on");
946 return 0;
947 }
f56426ae
JK		 948 if (sbi->s_qf_names[qtype]) {
949 kfree(sbi->s_qf_names[qtype]);
950 sbi->s_qf_names[qtype] = NULL;
951 }
e1f5c67a
DM		 952 return 1;
953}
954#endif
955
71b96257 956static int parse_options (char *options, struct super_block *sb,
eee194e7 957 unsigned int *inum, unsigned long *journal_devnum,
43d23f90 958 ext3_fsblk_t *n_blocks_count, int is_remount)
1da177e4
LT		 959{
960 struct ext3_sb_info *sbi = EXT3_SB(sb);
961 char * p;
962 substring_t args[MAX_OPT_ARGS];
963 int data_opt = 0;
964 int option;
1523299d
EB		 965 kuid_t uid;
966 kgid_t gid;
1da177e4 967#ifdef CONFIG_QUOTA
e1f5c67a 968 int qfmt;
1da177e4
LT		 969#endif
970
971 if (!options)
972 return 1;
973
974 while ((p = strsep (&options, ",")) != NULL) {
975 int token;
976 if (!*p)
977 continue;
0636c73e
ES		 978 /*
979 * Initialize args struct so we know whether arg was
980 * found; some options take optional arguments.
981 */
76f59b3b 982 args[0].to = args[0].from = NULL;
1da177e4
LT		 983 token = match_token(p, tokens, args);
984 switch (token) {
985 case Opt_bsd_df:
986 clear_opt (sbi->s_mount_opt, MINIX_DF);
987 break;
988 case Opt_minix_df:
989 set_opt (sbi->s_mount_opt, MINIX_DF);
990 break;
991 case Opt_grpid:
992 set_opt (sbi->s_mount_opt, GRPID);
993 break;
994 case Opt_nogrpid:
995 clear_opt (sbi->s_mount_opt, GRPID);
996 break;
997 case Opt_resuid:
998 if (match_int(&args[0], &option))
999 return 0;
1523299d
EB		 1000 uid = make_kuid(current_user_ns(), option);
1001 if (!uid_valid(uid)) {
1002 ext3_msg(sb, KERN_ERR, "Invalid uid value %d", option);
66415f6b 1003 return 0;
1523299d
EB		 1004
1005 }
1006 sbi->s_resuid = uid;
1da177e4
LT		 1007 break;
1008 case Opt_resgid:
1009 if (match_int(&args[0], &option))
1010 return 0;
1523299d
EB		 1011 gid = make_kgid(current_user_ns(), option);
1012 if (!gid_valid(gid)) {
1013 ext3_msg(sb, KERN_ERR, "Invalid gid value %d", option);
66415f6b 1014 return 0;
1523299d
EB		 1015 }
1016 sbi->s_resgid = gid;
1da177e4
LT		 1017 break;
1018 case Opt_sb:
1019 /* handled by get_sb_block() instead of here */
1020 /* *sb_block = match_int(&args[0]); */
1021 break;
1022 case Opt_err_panic:
1023 clear_opt (sbi->s_mount_opt, ERRORS_CONT);
1024 clear_opt (sbi->s_mount_opt, ERRORS_RO);
1025 set_opt (sbi->s_mount_opt, ERRORS_PANIC);
1026 break;
1027 case Opt_err_ro:
1028 clear_opt (sbi->s_mount_opt, ERRORS_CONT);
1029 clear_opt (sbi->s_mount_opt, ERRORS_PANIC);
1030 set_opt (sbi->s_mount_opt, ERRORS_RO);
1031 break;
1032 case Opt_err_cont:
1033 clear_opt (sbi->s_mount_opt, ERRORS_RO);
1034 clear_opt (sbi->s_mount_opt, ERRORS_PANIC);
1035 set_opt (sbi->s_mount_opt, ERRORS_CONT);
1036 break;
1037 case Opt_nouid32:
1038 set_opt (sbi->s_mount_opt, NO_UID32);
1039 break;
1da177e4
LT		 1040 case Opt_nocheck:
1041 clear_opt (sbi->s_mount_opt, CHECK);
1042 break;
1043 case Opt_debug:
1044 set_opt (sbi->s_mount_opt, DEBUG);
1045 break;
1046 case Opt_oldalloc:
fbc85402
LC		 1047 ext3_msg(sb, KERN_WARNING,
1048 "Ignoring deprecated oldalloc option");
1da177e4
LT		 1049 break;
1050 case Opt_orlov:
fbc85402
LC		 1051 ext3_msg(sb, KERN_WARNING,
1052 "Ignoring deprecated orlov option");
1da177e4
LT		 1053 break;
1054#ifdef CONFIG_EXT3_FS_XATTR
1055 case Opt_user_xattr:
1056 set_opt (sbi->s_mount_opt, XATTR_USER);
1057 break;
1058 case Opt_nouser_xattr:
1059 clear_opt (sbi->s_mount_opt, XATTR_USER);
1060 break;
1061#else
1062 case Opt_user_xattr:
1063 case Opt_nouser_xattr:
4cf46b67
AF		 1064 ext3_msg(sb, KERN_INFO,
1065 "(no)user_xattr options not supported");
1da177e4
LT		 1066 break;
1067#endif
1068#ifdef CONFIG_EXT3_FS_POSIX_ACL
1069 case Opt_acl:
1070 set_opt(sbi->s_mount_opt, POSIX_ACL);
1071 break;
1072 case Opt_noacl:
1073 clear_opt(sbi->s_mount_opt, POSIX_ACL);
1074 break;
1075#else
1076 case Opt_acl:
1077 case Opt_noacl:
4cf46b67
AF		 1078 ext3_msg(sb, KERN_INFO,
1079 "(no)acl options not supported");
1da177e4
LT		 1080 break;
1081#endif
1082 case Opt_reservation:
1083 set_opt(sbi->s_mount_opt, RESERVATION);
1084 break;
1085 case Opt_noreservation:
1086 clear_opt(sbi->s_mount_opt, RESERVATION);
1087 break;
1088 case Opt_journal_update:
1089 /* @@@ FIXME */
1090 /* Eventually we will want to be able to create
1091 a journal file here. For now, only allow the
1092 user to specify an existing inode to be the
1093 journal file. */
1094 if (is_remount) {
4cf46b67
AF		 1095 ext3_msg(sb, KERN_ERR, "error: cannot specify "
1096 "journal on remount");
1da177e4
LT		 1097 return 0;
1098 }
1099 set_opt (sbi->s_mount_opt, UPDATE_JOURNAL);
1100 break;
1101 case Opt_journal_inum:
1102 if (is_remount) {
4cf46b67
AF		 1103 ext3_msg(sb, KERN_ERR, "error: cannot specify "
1104 "journal on remount");
1da177e4
LT		 1105 return 0;
1106 }
1107 if (match_int(&args[0], &option))
1108 return 0;
1109 *inum = option;
1110 break;
71b96257
JL		 1111 case Opt_journal_dev:
1112 if (is_remount) {
4cf46b67
AF		 1113 ext3_msg(sb, KERN_ERR, "error: cannot specify "
1114 "journal on remount");
71b96257
JL		 1115 return 0;
1116 }
1117 if (match_int(&args[0], &option))
1118 return 0;
1119 *journal_devnum = option;
1120 break;
1da177e4
LT		 1121 case Opt_noload:
1122 set_opt (sbi->s_mount_opt, NOLOAD);
1123 break;
1124 case Opt_commit:
1125 if (match_int(&args[0], &option))
1126 return 0;
1127 if (option < 0)
1128 return 0;
1129 if (option == 0)
1130 option = JBD_DEFAULT_MAX_COMMIT_AGE;
1131 sbi->s_commit_interval = HZ * option;
1132 break;
1133 case Opt_data_journal:
1134 data_opt = EXT3_MOUNT_JOURNAL_DATA;
1135 goto datacheck;
1136 case Opt_data_ordered:
1137 data_opt = EXT3_MOUNT_ORDERED_DATA;
1138 goto datacheck;
1139 case Opt_data_writeback:
1140 data_opt = EXT3_MOUNT_WRITEBACK_DATA;
1141 datacheck:
1142 if (is_remount) {
e3c96435 1143 if (test_opt(sb, DATA_FLAGS) == data_opt)
3c4cec65 1144 break;
4cf46b67
AF		 1145 ext3_msg(sb, KERN_ERR,
1146 "error: cannot change "
3c4cec65
JK		 1147 "data mode on remount. The filesystem "
1148 "is mounted in data=%s mode and you "
4cf46b67 1149 "try to remount it in data=%s mode.",
e3c96435
DM		 1150 data_mode_string(test_opt(sb,
1151 DATA_FLAGS)),
3c4cec65
JK		 1152 data_mode_string(data_opt));
1153 return 0;
1da177e4 1154 } else {
e3c96435 1155 clear_opt(sbi->s_mount_opt, DATA_FLAGS);
1da177e4
LT		 1156 sbi->s_mount_opt |= data_opt;
1157 }
1158 break;
0e4fb5e2
HK		 1159 case Opt_data_err_abort:
1160 set_opt(sbi->s_mount_opt, DATA_ERR_ABORT);
1161 break;
1162 case Opt_data_err_ignore:
1163 clear_opt(sbi->s_mount_opt, DATA_ERR_ABORT);
1164 break;
1da177e4
LT		 1165#ifdef CONFIG_QUOTA
1166 case Opt_usrjquota:
e1f5c67a 1167 if (!set_qf_name(sb, USRQUOTA, &args[0]))
1da177e4 1168 return 0;
e1f5c67a
DM		 1169 break;
1170 case Opt_grpjquota:
1171 if (!set_qf_name(sb, GRPQUOTA, &args[0]))
1da177e4 1172 return 0;
1da177e4
LT		 1173 break;
1174 case Opt_offusrjquota:
e1f5c67a
DM		 1175 if (!clear_qf_name(sb, USRQUOTA))
1176 return 0;
1177 break;
1da177e4 1178 case Opt_offgrpjquota:
e1f5c67a 1179 if (!clear_qf_name(sb, GRPQUOTA))
1da177e4 1180 return 0;
1da177e4
LT		 1181 break;
1182 case Opt_jqfmt_vfsold:
d06bf1d2
JK		 1183 qfmt = QFMT_VFS_OLD;
1184 goto set_qf_format;
1da177e4 1185 case Opt_jqfmt_vfsv0:
d06bf1d2 1186 qfmt = QFMT_VFS_V0;
1aeec434
JK		 1187 goto set_qf_format;
1188 case Opt_jqfmt_vfsv1:
1189 qfmt = QFMT_VFS_V1;
d06bf1d2 1190set_qf_format:
ee0d5ffe 1191 if (sb_any_quota_loaded(sb) &&
d06bf1d2 1192 sbi->s_jquota_fmt != qfmt) {
4cf46b67 1193 ext3_msg(sb, KERN_ERR, "error: cannot change "
d06bf1d2 1194 "journaled quota options when "
4cf46b67 1195 "quota turned on.");
d06bf1d2
JK		 1196 return 0;
1197 }
1198 sbi->s_jquota_fmt = qfmt;
1da177e4 1199 break;
1f54587b 1200 case Opt_quota:
8fc2751b
MB		 1201 case Opt_usrquota:
1202 set_opt(sbi->s_mount_opt, QUOTA);
1203 set_opt(sbi->s_mount_opt, USRQUOTA);
1204 break;
1205 case Opt_grpquota:
1f54587b 1206 set_opt(sbi->s_mount_opt, QUOTA);
8fc2751b 1207 set_opt(sbi->s_mount_opt, GRPQUOTA);
1f54587b
JK		 1208 break;
1209 case Opt_noquota:
ee0d5ffe 1210 if (sb_any_quota_loaded(sb)) {
4cf46b67
AF		 1211 ext3_msg(sb, KERN_ERR, "error: cannot change "
1212 "quota options when quota turned on.");
1f54587b
JK		 1213 return 0;
1214 }
1215 clear_opt(sbi->s_mount_opt, QUOTA);
8fc2751b
MB		 1216 clear_opt(sbi->s_mount_opt, USRQUOTA);
1217 clear_opt(sbi->s_mount_opt, GRPQUOTA);
1f54587b 1218 break;
1da177e4 1219#else
8fc2751b
MB		 1220 case Opt_quota:
1221 case Opt_usrquota:
1222 case Opt_grpquota:
4cf46b67
AF		 1223 ext3_msg(sb, KERN_ERR,
1224 "error: quota options not supported.");
fa1ff1e0 1225 break;
1da177e4
LT		 1226 case Opt_usrjquota:
1227 case Opt_grpjquota:
1228 case Opt_offusrjquota:
1229 case Opt_offgrpjquota:
1230 case Opt_jqfmt_vfsold:
1231 case Opt_jqfmt_vfsv0:
1aeec434 1232 case Opt_jqfmt_vfsv1:
4cf46b67
AF		 1233 ext3_msg(sb, KERN_ERR,
1234 "error: journaled quota options not "
1235 "supported.");
1da177e4 1236 break;
1f54587b
JK		 1237 case Opt_noquota:
1238 break;
1da177e4
LT		 1239#endif
1240 case Opt_abort:
1241 set_opt(sbi->s_mount_opt, ABORT);
1242 break;
0636c73e
ES		 1243 case Opt_nobarrier:
1244 clear_opt(sbi->s_mount_opt, BARRIER);
1245 break;
1da177e4 1246 case Opt_barrier:
0636c73e
ES		 1247 if (args[0].from) {
1248 if (match_int(&args[0], &option))
1249 return 0;
1250 } else
1251 option = 1; /* No argument, default to 1 */
1da177e4
LT		 1252 if (option)
1253 set_opt(sbi->s_mount_opt, BARRIER);
1254 else
1255 clear_opt(sbi->s_mount_opt, BARRIER);
1256 break;
1257 case Opt_ignore:
1258 break;
1259 case Opt_resize:
08c6a96f 1260 if (!is_remount) {
4cf46b67
AF		 1261 ext3_msg(sb, KERN_ERR,
1262 "error: resize option only available "
1263 "for remount");
1da177e4
LT		 1264 return 0;
1265 }
c7f1721e
KZ		 1266 if (match_int(&args[0], &option) != 0)
1267 return 0;
1da177e4
LT		 1268 *n_blocks_count = option;
1269 break;
1270 case Opt_nobh:
4c4d3901
CH		 1271 ext3_msg(sb, KERN_WARNING,
1272 "warning: ignoring deprecated nobh option");
1da177e4 1273 break;
ade1a29e 1274 case Opt_bh:
4c4d3901
CH		 1275 ext3_msg(sb, KERN_WARNING,
1276 "warning: ignoring deprecated bh option");
ade1a29e 1277 break;
1da177e4 1278 default:
4cf46b67
AF		 1279 ext3_msg(sb, KERN_ERR,
1280 "error: unrecognized mount option \"%s\" "
1281 "or missing value", p);
1da177e4
LT		 1282 return 0;
1283 }
1284 }
1285#ifdef CONFIG_QUOTA
8fc2751b 1286 if (sbi->s_qf_names[USRQUOTA] || sbi->s_qf_names[GRPQUOTA]) {
e3c96435 1287 if (test_opt(sb, USRQUOTA) && sbi->s_qf_names[USRQUOTA])
8fc2751b 1288 clear_opt(sbi->s_mount_opt, USRQUOTA);
e3c96435 1289 if (test_opt(sb, GRPQUOTA) && sbi->s_qf_names[GRPQUOTA])
8fc2751b
MB		 1290 clear_opt(sbi->s_mount_opt, GRPQUOTA);
1291
e1f5c67a 1292 if (test_opt(sb, GRPQUOTA) || test_opt(sb, USRQUOTA)) {
4cf46b67
AF		 1293 ext3_msg(sb, KERN_ERR, "error: old and new quota "
1294 "format mixing.");
8fc2751b
MB		 1295 return 0;
1296 }
1297
1298 if (!sbi->s_jquota_fmt) {
4cf46b67
AF		 1299 ext3_msg(sb, KERN_ERR, "error: journaled quota format "
1300 "not specified.");
8fc2751b
MB		 1301 return 0;
1302 }
1303 } else {
1304 if (sbi->s_jquota_fmt) {
4cf46b67 1305 ext3_msg(sb, KERN_ERR, "error: journaled quota format "
99aeaf63 1306 "specified with no journaling "
4cf46b67 1307 "enabled.");
8fc2751b
MB		 1308 return 0;
1309 }
1da177e4
LT		 1310 }
1311#endif
1da177e4
LT		 1312 return 1;
1313}
1314
1315static int ext3_setup_super(struct super_block *sb, struct ext3_super_block *es,
1316 int read_only)
1317{
1318 struct ext3_sb_info *sbi = EXT3_SB(sb);
1319 int res = 0;
1320
1321 if (le32_to_cpu(es->s_rev_level) > EXT3_MAX_SUPP_REV) {
4cf46b67
AF		 1322 ext3_msg(sb, KERN_ERR,
1323 "error: revision level too high, "
1324 "forcing read-only mode");
1da177e4
LT		 1325 res = MS_RDONLY;
1326 }
1327 if (read_only)
1328 return res;
1329 if (!(sbi->s_mount_state & EXT3_VALID_FS))
4cf46b67
AF		 1330 ext3_msg(sb, KERN_WARNING,
1331 "warning: mounting unchecked fs, "
1332 "running e2fsck is recommended");
1da177e4 1333 else if ((sbi->s_mount_state & EXT3_ERROR_FS))
4cf46b67
AF		 1334 ext3_msg(sb, KERN_WARNING,
1335 "warning: mounting fs with errors, "
1336 "running e2fsck is recommended");
df0d6b8f 1337 else if ((__s16) le16_to_cpu(es->s_max_mnt_count) > 0 &&
1da177e4 1338 le16_to_cpu(es->s_mnt_count) >=
df0d6b8f 1339 le16_to_cpu(es->s_max_mnt_count))
4cf46b67
AF		 1340 ext3_msg(sb, KERN_WARNING,
1341 "warning: maximal mount count reached, "
1342 "running e2fsck is recommended");
1da177e4
LT		 1343 else if (le32_to_cpu(es->s_checkinterval) &&
1344 (le32_to_cpu(es->s_lastcheck) +
1345 le32_to_cpu(es->s_checkinterval) <= get_seconds()))
4cf46b67
AF		 1346 ext3_msg(sb, KERN_WARNING,
1347 "warning: checktime reached, "
1348 "running e2fsck is recommended");
1da177e4
LT		 1349#if 0
1350 /* @@@ We _will_ want to clear the valid bit if we find
1351 inconsistencies, to force a fsck at reboot. But for
1352 a plain journaled filesystem we can keep it set as
1353 valid forever! :) */
e7f23ebd 1354 es->s_state &= cpu_to_le16(~EXT3_VALID_FS);
1da177e4 1355#endif
df0d6b8f 1356 if (!le16_to_cpu(es->s_max_mnt_count))
1da177e4 1357 es->s_max_mnt_count = cpu_to_le16(EXT3_DFL_MAX_MNT_COUNT);
50e8a289 1358 le16_add_cpu(&es->s_mnt_count, 1);
1da177e4
LT		 1359 es->s_mtime = cpu_to_le32(get_seconds());
1360 ext3_update_dynamic_rev(sb);
1361 EXT3_SET_INCOMPAT_FEATURE(sb, EXT3_FEATURE_INCOMPAT_RECOVER);
1362
1363 ext3_commit_super(sb, es, 1);
1364 if (test_opt(sb, DEBUG))
4cf46b67
AF		 1365 ext3_msg(sb, KERN_INFO, "[bs=%lu, gc=%lu, "
1366 "bpg=%lu, ipg=%lu, mo=%04lx]",
1da177e4
LT		 1367 sb->s_blocksize,
1368 sbi->s_groups_count,
1369 EXT3_BLOCKS_PER_GROUP(sb),
1370 EXT3_INODES_PER_GROUP(sb),
1371 sbi->s_mount_opt);
1372
1da177e4
LT		 1373 if (EXT3_SB(sb)->s_journal->j_inode == NULL) {
1374 char b[BDEVNAME_SIZE];
4cf46b67 1375 ext3_msg(sb, KERN_INFO, "using external journal on %s",
1da177e4
LT		 1376 bdevname(EXT3_SB(sb)->s_journal->j_dev, b));
1377 } else {
4cf46b67 1378 ext3_msg(sb, KERN_INFO, "using internal journal");
1da177e4 1379 }
d71bc6db 1380 cleancache_init_fs(sb);
1da177e4
LT		 1381 return res;
1382}
1383
1384/* Called at mount-time, super-block is locked */
197cd65a 1385static int ext3_check_descriptors(struct super_block *sb)
1da177e4
LT		 1386{
1387 struct ext3_sb_info *sbi = EXT3_SB(sb);
1da177e4
LT		 1388 int i;
1389
1390 ext3_debug ("Checking group descriptors");
1391
197cd65a
AM		 1392 for (i = 0; i < sbi->s_groups_count; i++) {
1393 struct ext3_group_desc *gdp = ext3_get_group_desc(sb, i, NULL);
1eaafeae
AM		 1394 ext3_fsblk_t first_block = ext3_group_first_block_no(sb, i);
1395 ext3_fsblk_t last_block;
197cd65a 1396
855565e8
ES		 1397 if (i == sbi->s_groups_count - 1)
1398 last_block = le32_to_cpu(sbi->s_es->s_blocks_count) - 1;
1399 else
1400 last_block = first_block +
1401 (EXT3_BLOCKS_PER_GROUP(sb) - 1);
1402
855565e8
ES		 1403 if (le32_to_cpu(gdp->bg_block_bitmap) < first_block ||
1404 le32_to_cpu(gdp->bg_block_bitmap) > last_block)
1da177e4
LT		 1405 {
1406 ext3_error (sb, "ext3_check_descriptors",
1407 "Block bitmap for group %d"
1408 " not in group (block %lu)!",
1409 i, (unsigned long)
1410 le32_to_cpu(gdp->bg_block_bitmap));
1411 return 0;
1412 }
855565e8
ES		 1413 if (le32_to_cpu(gdp->bg_inode_bitmap) < first_block ||
1414 le32_to_cpu(gdp->bg_inode_bitmap) > last_block)
1da177e4
LT		 1415 {
1416 ext3_error (sb, "ext3_check_descriptors",
1417 "Inode bitmap for group %d"
1418 " not in group (block %lu)!",
1419 i, (unsigned long)
1420 le32_to_cpu(gdp->bg_inode_bitmap));
1421 return 0;
1422 }
855565e8 1423 if (le32_to_cpu(gdp->bg_inode_table) < first_block ||
780dcdb2 1424 le32_to_cpu(gdp->bg_inode_table) + sbi->s_itb_per_group - 1 >
855565e8 1425 last_block)
1da177e4
LT		 1426 {
1427 ext3_error (sb, "ext3_check_descriptors",
1428 "Inode table for group %d"
1429 " not in group (block %lu)!",
1430 i, (unsigned long)
1431 le32_to_cpu(gdp->bg_inode_table));
1432 return 0;
1433 }
1da177e4
LT		 1434 }
1435
1436 sbi->s_es->s_free_blocks_count=cpu_to_le32(ext3_count_free_blocks(sb));
1437 sbi->s_es->s_free_inodes_count=cpu_to_le32(ext3_count_free_inodes(sb));
1438 return 1;
1439}
1440
1441
1442/* ext3_orphan_cleanup() walks a singly-linked list of inodes (starting at
1443 * the superblock) which were deleted from all directories, but held open by
1444 * a process at the time of a crash. We walk the list and try to delete these
1445 * inodes at recovery time (only with a read-write filesystem).
1446 *
1447 * In order to keep the orphan inode chain consistent during traversal (in
1448 * case of crash during recovery), we link each inode into the superblock
1449 * orphan list_head and handle it the same way as an inode deletion during
1450 * normal operation (which journals the operations for us).
1451 *
1452 * We only do an iget() and an iput() on each inode, which is very safe if we
1453 * accidentally point at an in-use or already deleted inode. The worst that
1454 * can happen in this case is that we get a "bit already cleared" message from
1455 * ext3_free_inode(). The only reason we would point at a wrong inode is if
1456 * e2fsck was run on this filesystem, and it must have already done the orphan
1457 * inode cleanup for us, so we can safely abort without any further action.
1458 */
1459static void ext3_orphan_cleanup (struct super_block * sb,
1460 struct ext3_super_block * es)
1461{
1462 unsigned int s_flags = sb->s_flags;
1463 int nr_orphans = 0, nr_truncates = 0;
1464#ifdef CONFIG_QUOTA
1465 int i;
1466#endif
1467 if (!es->s_last_orphan) {
1468 jbd_debug(4, "no orphan inodes to clean up\n");
1469 return;
1470 }
1471
a8f48a95 1472 if (bdev_read_only(sb->s_bdev)) {
4cf46b67
AF		 1473 ext3_msg(sb, KERN_ERR, "error: write access "
1474 "unavailable, skipping orphan cleanup.");
a8f48a95
ES		 1475 return;
1476 }
1477
ce654b37
AG		 1478 /* Check if feature set allows readwrite operations */
1479 if (EXT3_HAS_RO_COMPAT_FEATURE(sb, ~EXT3_FEATURE_RO_COMPAT_SUPP)) {
1480 ext3_msg(sb, KERN_INFO, "Skipping orphan cleanup due to "
1481 "unknown ROCOMPAT features");
1482 return;
a8f48a95
ES		 1483 }
1484
1da177e4 1485 if (EXT3_SB(sb)->s_mount_state & EXT3_ERROR_FS) {
e4230108
ES		 1486 /* don't clear list on RO mount w/ errors */
1487 if (es->s_last_orphan && !(s_flags & MS_RDONLY)) {
1da177e4
LT		 1488 jbd_debug(1, "Errors on filesystem, "
1489 "clearing orphan list.\n");
e4230108
ES		 1490 es->s_last_orphan = 0;
1491 }
1da177e4
LT		 1492 jbd_debug(1, "Skipping orphan recovery on fs with errors.\n");
1493 return;
1494 }
1495
1496 if (s_flags & MS_RDONLY) {
4cf46b67 1497 ext3_msg(sb, KERN_INFO, "orphan cleanup on readonly fs");
1da177e4
LT		 1498 sb->s_flags &= ~MS_RDONLY;
1499 }
1500#ifdef CONFIG_QUOTA
1501 /* Needed for iput() to work correctly and not trash data */
1502 sb->s_flags |= MS_ACTIVE;
1503 /* Turn on quotas so that they are updated correctly */
1504 for (i = 0; i < MAXQUOTAS; i++) {
1505 if (EXT3_SB(sb)->s_qf_names[i]) {
1506 int ret = ext3_quota_on_mount(sb, i);
1507 if (ret < 0)
4cf46b67
AF		 1508 ext3_msg(sb, KERN_ERR,
1509 "error: cannot turn on journaled "
1510 "quota: %d", ret);
1da177e4
LT		 1511 }
1512 }
1513#endif
1514
1515 while (es->s_last_orphan) {
1516 struct inode *inode;
1517
473043dc
DH		 1518 inode = ext3_orphan_get(sb, le32_to_cpu(es->s_last_orphan));
1519 if (IS_ERR(inode)) {
1da177e4
LT		 1520 es->s_last_orphan = 0;
1521 break;
1522 }
1523
1524 list_add(&EXT3_I(inode)->i_orphan, &EXT3_SB(sb)->s_orphan);
871a2931 1525 dquot_initialize(inode);
1da177e4
LT		 1526 if (inode->i_nlink) {
1527 printk(KERN_DEBUG
eee194e7 1528 "%s: truncating inode %lu to %Ld bytes\n",
e05b6b52 1529 __func__, inode->i_ino, inode->i_size);
eee194e7 1530 jbd_debug(2, "truncating inode %lu to %Ld bytes\n",
1da177e4
LT		 1531 inode->i_ino, inode->i_size);
1532 ext3_truncate(inode);
1533 nr_truncates++;
1534 } else {
1535 printk(KERN_DEBUG
eee194e7 1536 "%s: deleting unreferenced inode %lu\n",
e05b6b52 1537 __func__, inode->i_ino);
eee194e7 1538 jbd_debug(2, "deleting unreferenced inode %lu\n",
1da177e4
LT		 1539 inode->i_ino);
1540 nr_orphans++;
1541 }
1542 iput(inode); /* The delete magic happens here! */
1543 }
1544
1545#define PLURAL(x) (x), ((x)==1) ? "" : "s"
1546
1547 if (nr_orphans)
4cf46b67
AF		 1548 ext3_msg(sb, KERN_INFO, "%d orphan inode%s deleted",
1549 PLURAL(nr_orphans));
1da177e4 1550 if (nr_truncates)
4cf46b67
AF		 1551 ext3_msg(sb, KERN_INFO, "%d truncate%s cleaned up",
1552 PLURAL(nr_truncates));
1da177e4
LT		 1553#ifdef CONFIG_QUOTA
1554 /* Turn quotas off */
1555 for (i = 0; i < MAXQUOTAS; i++) {
1556 if (sb_dqopt(sb)->files[i])
287a8095 1557 dquot_quota_off(sb, i);
1da177e4
LT		 1558 }
1559#endif
1560 sb->s_flags = s_flags; /* Restore MS_RDONLY status */
1561}
1562
1da177e4
LT		 1563/*
1564 * Maximal file size. There is a direct, and {,double-,triple-}indirect
1565 * block limit, and also a limit of (2^32 - 1) 512-byte sectors in i_blocks.
1566 * We need to be 1 filesystem block less than the 2^32 sector limit.
1567 */
1568static loff_t ext3_max_size(int bits)
1569{
1570 loff_t res = EXT3_NDIR_BLOCKS;
fe7fdc37
AK		 1571 int meta_blocks;
1572 loff_t upper_limit;
1573
1574 /* This is calculated to be the largest file size for a
1575 * dense, file such that the total number of
1da177e4 1576 * sectors in the file, including data and all indirect blocks,
fe7fdc37
AK		 1577 * does not exceed 2^32 -1
1578 * __u32 i_blocks representing the total number of
1579 * 512 bytes blocks of the file
1580 */
1581 upper_limit = (1LL << 32) - 1;
1582
1583 /* total blocks in file system block size */
1584 upper_limit >>= (bits - 9);
1585
1586
1587 /* indirect blocks */
1588 meta_blocks = 1;
1589 /* double indirect blocks */
1590 meta_blocks += 1 + (1LL << (bits-2));
1591 /* tripple indirect blocks */
1592 meta_blocks += 1 + (1LL << (bits-2)) + (1LL << (2*(bits-2)));
1593
1594 upper_limit -= meta_blocks;
1595 upper_limit <<= bits;
1da177e4
LT		 1596
1597 res += 1LL << (bits-2);
1598 res += 1LL << (2*(bits-2));
1599 res += 1LL << (3*(bits-2));
1600 res <<= bits;
1601 if (res > upper_limit)
1602 res = upper_limit;
fe7fdc37
AK		 1603
1604 if (res > MAX_LFS_FILESIZE)
1605 res = MAX_LFS_FILESIZE;
1606
1da177e4
LT		 1607 return res;
1608}
1609
43d23f90
MC		 1610static ext3_fsblk_t descriptor_loc(struct super_block *sb,
1611 ext3_fsblk_t logic_sb_block,
1da177e4
LT		 1612 int nr)
1613{
1614 struct ext3_sb_info *sbi = EXT3_SB(sb);
43d23f90 1615 unsigned long bg, first_meta_bg;
1da177e4
LT		 1616 int has_super = 0;
1617
1da177e4
LT		 1618 first_meta_bg = le32_to_cpu(sbi->s_es->s_first_meta_bg);
1619
1620 if (!EXT3_HAS_INCOMPAT_FEATURE(sb, EXT3_FEATURE_INCOMPAT_META_BG) ||
1621 nr < first_meta_bg)
1622 return (logic_sb_block + nr + 1);
1623 bg = sbi->s_desc_per_block * nr;
1624 if (ext3_bg_has_super(sb, bg))
1625 has_super = 1;
43d23f90 1626 return (has_super + ext3_group_first_block_no(sb, bg));
1da177e4
LT		 1627}
1628
1629
1630static int ext3_fill_super (struct super_block *sb, void *data, int silent)
1631{
1632 struct buffer_head * bh;
1633 struct ext3_super_block *es = NULL;
1634 struct ext3_sb_info *sbi;
43d23f90 1635 ext3_fsblk_t block;
4cf46b67 1636 ext3_fsblk_t sb_block = get_sb_block(&data, sb);
43d23f90 1637 ext3_fsblk_t logic_sb_block;
1da177e4 1638 unsigned long offset = 0;
eee194e7 1639 unsigned int journal_inum = 0;
71b96257 1640 unsigned long journal_devnum = 0;
1da177e4
LT		 1641 unsigned long def_mount_opts;
1642 struct inode *root;
1643 int blocksize;
1644 int hblock;
1645 int db_count;
1646 int i;
1647 int needs_recovery;
473043dc 1648 int ret = -EINVAL;
1da177e4 1649 __le32 features;
833f4077 1650 int err;
1da177e4 1651
f8314dc6 1652 sbi = kzalloc(sizeof(*sbi), GFP_KERNEL);
1da177e4
LT		 1653 if (!sbi)
1654 return -ENOMEM;
5df096d6
PE		 1655
1656 sbi->s_blockgroup_lock =
1657 kzalloc(sizeof(struct blockgroup_lock), GFP_KERNEL);
1658 if (!sbi->s_blockgroup_lock) {
1659 kfree(sbi);
1660 return -ENOMEM;
1661 }
1da177e4 1662 sb->s_fs_info = sbi;
571beed8 1663 sbi->s_sb_block = sb_block;
1da177e4 1664
1da177e4
LT		 1665 blocksize = sb_min_blocksize(sb, EXT3_MIN_BLOCK_SIZE);
1666 if (!blocksize) {
4cf46b67 1667 ext3_msg(sb, KERN_ERR, "error: unable to set blocksize");
1da177e4
LT		 1668 goto out_fail;
1669 }
1670
1671 /*
1672 * The ext3 superblock will not be buffer aligned for other than 1kB
1673 * block sizes. We need to calculate the offset from buffer start.
1674 */
1675 if (blocksize != EXT3_MIN_BLOCK_SIZE) {
1676 logic_sb_block = (sb_block * EXT3_MIN_BLOCK_SIZE) / blocksize;
1677 offset = (sb_block * EXT3_MIN_BLOCK_SIZE) % blocksize;
1678 } else {
1679 logic_sb_block = sb_block;
1680 }
1681
1682 if (!(bh = sb_bread(sb, logic_sb_block))) {
4cf46b67 1683 ext3_msg(sb, KERN_ERR, "error: unable to read superblock");
1da177e4
LT		 1684 goto out_fail;
1685 }
1686 /*
1687 * Note: s_es must be initialized as soon as possible because
1688 * some ext3 macro-instructions depend on its value
1689 */
57e94d86 1690 es = (struct ext3_super_block *) (bh->b_data + offset);
1da177e4
LT		 1691 sbi->s_es = es;
1692 sb->s_magic = le16_to_cpu(es->s_magic);
1693 if (sb->s_magic != EXT3_SUPER_MAGIC)
1694 goto cantfind_ext3;
1695
1696 /* Set defaults before we parse the mount options */
1697 def_mount_opts = le32_to_cpu(es->s_default_mount_opts);
1698 if (def_mount_opts & EXT3_DEFM_DEBUG)
1699 set_opt(sbi->s_mount_opt, DEBUG);
1700 if (def_mount_opts & EXT3_DEFM_BSDGROUPS)
1701 set_opt(sbi->s_mount_opt, GRPID);
1702 if (def_mount_opts & EXT3_DEFM_UID16)
1703 set_opt(sbi->s_mount_opt, NO_UID32);
2e7842b8 1704#ifdef CONFIG_EXT3_FS_XATTR
1da177e4
LT		 1705 if (def_mount_opts & EXT3_DEFM_XATTR_USER)
1706 set_opt(sbi->s_mount_opt, XATTR_USER);
2e7842b8
HD		 1707#endif
1708#ifdef CONFIG_EXT3_FS_POSIX_ACL
1da177e4
LT		 1709 if (def_mount_opts & EXT3_DEFM_ACL)
1710 set_opt(sbi->s_mount_opt, POSIX_ACL);
2e7842b8 1711#endif
1da177e4 1712 if ((def_mount_opts & EXT3_DEFM_JMODE) == EXT3_DEFM_JMODE_DATA)
e3c96435 1713 set_opt(sbi->s_mount_opt, JOURNAL_DATA);
1da177e4 1714 else if ((def_mount_opts & EXT3_DEFM_JMODE) == EXT3_DEFM_JMODE_ORDERED)
e3c96435 1715 set_opt(sbi->s_mount_opt, ORDERED_DATA);
1da177e4 1716 else if ((def_mount_opts & EXT3_DEFM_JMODE) == EXT3_DEFM_JMODE_WBACK)
e3c96435 1717 set_opt(sbi->s_mount_opt, WRITEBACK_DATA);
1da177e4
LT		 1718
1719 if (le16_to_cpu(sbi->s_es->s_errors) == EXT3_ERRORS_PANIC)
1720 set_opt(sbi->s_mount_opt, ERRORS_PANIC);
1eca93f9 1721 else if (le16_to_cpu(sbi->s_es->s_errors) == EXT3_ERRORS_CONTINUE)
2245d7c2 1722 set_opt(sbi->s_mount_opt, ERRORS_CONT);
1eca93f9
AK		 1723 else
1724 set_opt(sbi->s_mount_opt, ERRORS_RO);
1da177e4 1725
1523299d
EB		 1726 sbi->s_resuid = make_kuid(&init_user_ns, le16_to_cpu(es->s_def_resuid));
1727 sbi->s_resgid = make_kgid(&init_user_ns, le16_to_cpu(es->s_def_resgid));
1da177e4 1728
00eacd66
CH		 1729 /* enable barriers by default */
1730 set_opt(sbi->s_mount_opt, BARRIER);
1da177e4
LT		 1731 set_opt(sbi->s_mount_opt, RESERVATION);
1732
71b96257
JL		 1733 if (!parse_options ((char *) data, sb, &journal_inum, &journal_devnum,
1734 NULL, 0))
1da177e4
LT		 1735 goto failed_mount;
1736
1737 sb->s_flags = (sb->s_flags & ~MS_POSIXACL) |
e3c96435 1738 (test_opt(sb, POSIX_ACL) ? MS_POSIXACL : 0);
1da177e4
LT		 1739
1740 if (le32_to_cpu(es->s_rev_level) == EXT3_GOOD_OLD_REV &&
1741 (EXT3_HAS_COMPAT_FEATURE(sb, ~0U) ||
1742 EXT3_HAS_RO_COMPAT_FEATURE(sb, ~0U) ||
1743 EXT3_HAS_INCOMPAT_FEATURE(sb, ~0U)))
4cf46b67
AF		 1744 ext3_msg(sb, KERN_WARNING,
1745 "warning: feature flags set on rev 0 fs, "
1746 "running e2fsck is recommended");
1da177e4
LT		 1747 /*
1748 * Check feature flags regardless of the revision level, since we
1749 * previously didn't change the revision level when setting the flags,
1750 * so there is a chance incompat flags are set on a rev 0 filesystem.
1751 */
1752 features = EXT3_HAS_INCOMPAT_FEATURE(sb, ~EXT3_FEATURE_INCOMPAT_SUPP);
1753 if (features) {
4cf46b67
AF		 1754 ext3_msg(sb, KERN_ERR,
1755 "error: couldn't mount because of unsupported "
1756 "optional features (%x)", le32_to_cpu(features));
1da177e4
LT		 1757 goto failed_mount;
1758 }
1759 features = EXT3_HAS_RO_COMPAT_FEATURE(sb, ~EXT3_FEATURE_RO_COMPAT_SUPP);
1760 if (!(sb->s_flags & MS_RDONLY) && features) {
4cf46b67
AF		 1761 ext3_msg(sb, KERN_ERR,
1762 "error: couldn't mount RDWR because of unsupported "
1763 "optional features (%x)", le32_to_cpu(features));
1da177e4
LT		 1764 goto failed_mount;
1765 }
1766 blocksize = BLOCK_SIZE << le32_to_cpu(es->s_log_block_size);
1767
1768 if (blocksize < EXT3_MIN_BLOCK_SIZE ||
1769 blocksize > EXT3_MAX_BLOCK_SIZE) {
4cf46b67
AF		 1770 ext3_msg(sb, KERN_ERR,
1771 "error: couldn't mount because of unsupported "
1772 "filesystem blocksize %d", blocksize);
1da177e4
LT		 1773 goto failed_mount;
1774 }
1775
e1defc4f 1776 hblock = bdev_logical_block_size(sb->s_bdev);
1da177e4
LT		 1777 if (sb->s_blocksize != blocksize) {
1778 /*
1779 * Make sure the blocksize for the filesystem is larger
1780 * than the hardware sectorsize for the machine.
1781 */
1782 if (blocksize < hblock) {
4cf46b67
AF		 1783 ext3_msg(sb, KERN_ERR,
1784 "error: fsblocksize %d too small for "
1785 "hardware sectorsize %d", blocksize, hblock);
1da177e4
LT		 1786 goto failed_mount;
1787 }
1788
1789 brelse (bh);
0f0a89eb 1790 if (!sb_set_blocksize(sb, blocksize)) {
4cf46b67
AF		 1791 ext3_msg(sb, KERN_ERR,
1792 "error: bad blocksize %d", blocksize);
0f0a89eb
TS		 1793 goto out_fail;
1794 }
1da177e4
LT		 1795 logic_sb_block = (sb_block * EXT3_MIN_BLOCK_SIZE) / blocksize;
1796 offset = (sb_block * EXT3_MIN_BLOCK_SIZE) % blocksize;
1797 bh = sb_bread(sb, logic_sb_block);
1798 if (!bh) {
4cf46b67
AF		 1799 ext3_msg(sb, KERN_ERR,
1800 "error: can't read superblock on 2nd try");
1da177e4
LT		 1801 goto failed_mount;
1802 }
57e94d86 1803 es = (struct ext3_super_block *)(bh->b_data + offset);
1da177e4
LT		 1804 sbi->s_es = es;
1805 if (es->s_magic != cpu_to_le16(EXT3_SUPER_MAGIC)) {
4cf46b67
AF		 1806 ext3_msg(sb, KERN_ERR,
1807 "error: magic mismatch");
1da177e4
LT		 1808 goto failed_mount;
1809 }
1810 }
1811
1812 sb->s_maxbytes = ext3_max_size(sb->s_blocksize_bits);
1813
1814 if (le32_to_cpu(es->s_rev_level) == EXT3_GOOD_OLD_REV) {
1815 sbi->s_inode_size = EXT3_GOOD_OLD_INODE_SIZE;
1816 sbi->s_first_ino = EXT3_GOOD_OLD_FIRST_INO;
1817 } else {
1818 sbi->s_inode_size = le16_to_cpu(es->s_inode_size);
1819 sbi->s_first_ino = le32_to_cpu(es->s_first_ino);
1820 if ((sbi->s_inode_size < EXT3_GOOD_OLD_INODE_SIZE) ||
3fc74269 1821 (!is_power_of_2(sbi->s_inode_size)) ||
1da177e4 1822 (sbi->s_inode_size > blocksize)) {
4cf46b67
AF		 1823 ext3_msg(sb, KERN_ERR,
1824 "error: unsupported inode size: %d",
1da177e4
LT		 1825 sbi->s_inode_size);
1826 goto failed_mount;
1827 }
1828 }
1829 sbi->s_frag_size = EXT3_MIN_FRAG_SIZE <<
1830 le32_to_cpu(es->s_log_frag_size);
1831 if (blocksize != sbi->s_frag_size) {
4cf46b67
AF		 1832 ext3_msg(sb, KERN_ERR,
1833 "error: fragsize %lu != blocksize %u (unsupported)",
1da177e4
LT		 1834 sbi->s_frag_size, blocksize);
1835 goto failed_mount;
1836 }
1837 sbi->s_frags_per_block = 1;
1838 sbi->s_blocks_per_group = le32_to_cpu(es->s_blocks_per_group);
1839 sbi->s_frags_per_group = le32_to_cpu(es->s_frags_per_group);
1840 sbi->s_inodes_per_group = le32_to_cpu(es->s_inodes_per_group);
b47b6f38 1841 if (EXT3_INODE_SIZE(sb) == 0 || EXT3_INODES_PER_GROUP(sb) == 0)
1da177e4
LT		 1842 goto cantfind_ext3;
1843 sbi->s_inodes_per_block = blocksize / EXT3_INODE_SIZE(sb);
1844 if (sbi->s_inodes_per_block == 0)
1845 goto cantfind_ext3;
1846 sbi->s_itb_per_group = sbi->s_inodes_per_group /
1847 sbi->s_inodes_per_block;
1848 sbi->s_desc_per_block = blocksize / sizeof(struct ext3_group_desc);
1849 sbi->s_sbh = bh;
1850 sbi->s_mount_state = le16_to_cpu(es->s_state);
f0d1b0b3
DH		 1851 sbi->s_addr_per_block_bits = ilog2(EXT3_ADDR_PER_BLOCK(sb));
1852 sbi->s_desc_per_block_bits = ilog2(EXT3_DESC_PER_BLOCK(sb));
1da177e4
LT		 1853 for (i=0; i < 4; i++)
1854 sbi->s_hash_seed[i] = le32_to_cpu(es->s_hash_seed[i]);
1855 sbi->s_def_hash_version = es->s_def_hash_version;
5e1f8c9e
TT		 1856 i = le32_to_cpu(es->s_flags);
1857 if (i & EXT2_FLAGS_UNSIGNED_HASH)
1858 sbi->s_hash_unsigned = 3;
1859 else if ((i & EXT2_FLAGS_SIGNED_HASH) == 0) {
1860#ifdef __CHAR_UNSIGNED__
1861 es->s_flags |= cpu_to_le32(EXT2_FLAGS_UNSIGNED_HASH);
1862 sbi->s_hash_unsigned = 3;
1863#else
1864 es->s_flags |= cpu_to_le32(EXT2_FLAGS_SIGNED_HASH);
1865#endif
5e1f8c9e 1866 }
1da177e4
LT		 1867
1868 if (sbi->s_blocks_per_group > blocksize * 8) {
4cf46b67
AF		 1869 ext3_msg(sb, KERN_ERR,
1870 "#blocks per group too big: %lu",
1da177e4
LT		 1871 sbi->s_blocks_per_group);
1872 goto failed_mount;
1873 }
1874 if (sbi->s_frags_per_group > blocksize * 8) {
4cf46b67
AF		 1875 ext3_msg(sb, KERN_ERR,
1876 "error: #fragments per group too big: %lu",
1da177e4
LT		 1877 sbi->s_frags_per_group);
1878 goto failed_mount;
1879 }
1880 if (sbi->s_inodes_per_group > blocksize * 8) {
4cf46b67
AF		 1881 ext3_msg(sb, KERN_ERR,
1882 "error: #inodes per group too big: %lu",
1da177e4
LT		 1883 sbi->s_inodes_per_group);
1884 goto failed_mount;
1885 }
1886
ad692bf3
DW		 1887 err = generic_check_addressable(sb->s_blocksize_bits,
1888 le32_to_cpu(es->s_blocks_count));
1889 if (err) {
4cf46b67
AF		 1890 ext3_msg(sb, KERN_ERR,
1891 "error: filesystem is too large to mount safely");
fcd5df35 1892 if (sizeof(sector_t) < 8)
4cf46b67
AF		 1893 ext3_msg(sb, KERN_ERR,
1894 "error: CONFIG_LBDAF not enabled");
ad692bf3 1895 ret = err;
fcd5df35
MC		 1896 goto failed_mount;
1897 }
1898
1da177e4
LT		 1899 if (EXT3_BLOCKS_PER_GROUP(sb) == 0)
1900 goto cantfind_ext3;
855565e8
ES		 1901 sbi->s_groups_count = ((le32_to_cpu(es->s_blocks_count) -
1902 le32_to_cpu(es->s_first_data_block) - 1)
1903 / EXT3_BLOCKS_PER_GROUP(sb)) + 1;
81a4e320 1904 db_count = DIV_ROUND_UP(sbi->s_groups_count, EXT3_DESC_PER_BLOCK(sb));
1da177e4
LT		 1905 sbi->s_group_desc = kmalloc(db_count * sizeof (struct buffer_head *),
1906 GFP_KERNEL);
1907 if (sbi->s_group_desc == NULL) {
4cf46b67
AF		 1908 ext3_msg(sb, KERN_ERR,
1909 "error: not enough memory");
4569cd1b 1910 ret = -ENOMEM;
1da177e4
LT		 1911 goto failed_mount;
1912 }
1913
5df096d6 1914 bgl_lock_init(sbi->s_blockgroup_lock);
1da177e4
LT		 1915
1916 for (i = 0; i < db_count; i++) {
1917 block = descriptor_loc(sb, logic_sb_block, i);
1918 sbi->s_group_desc[i] = sb_bread(sb, block);
1919 if (!sbi->s_group_desc[i]) {
4cf46b67
AF		 1920 ext3_msg(sb, KERN_ERR,
1921 "error: can't read group descriptor %d", i);
1da177e4
LT		 1922 db_count = i;
1923 goto failed_mount2;
1924 }
1925 }
1926 if (!ext3_check_descriptors (sb)) {
4cf46b67
AF		 1927 ext3_msg(sb, KERN_ERR,
1928 "error: group descriptors corrupted");
1da177e4
LT		 1929 goto failed_mount2;
1930 }
1931 sbi->s_gdb_count = db_count;
1932 get_random_bytes(&sbi->s_next_generation, sizeof(u32));
1933 spin_lock_init(&sbi->s_next_gen_lock);
0216bfcf 1934
1da177e4
LT		 1935 /* per fileystem reservation list head & lock */
1936 spin_lock_init(&sbi->s_rsv_window_lock);
1937 sbi->s_rsv_window_root = RB_ROOT;
1938 /* Add a single, static dummy reservation to the start of the
1939 * reservation window list --- it gives us a placeholder for
1940 * append-at-start-of-list which makes the allocation logic
1941 * _much_ simpler. */
1942 sbi->s_rsv_window_head.rsv_start = EXT3_RESERVE_WINDOW_NOT_ALLOCATED;
1943 sbi->s_rsv_window_head.rsv_end = EXT3_RESERVE_WINDOW_NOT_ALLOCATED;
1944 sbi->s_rsv_window_head.rsv_alloc_hit = 0;
1945 sbi->s_rsv_window_head.rsv_goal_size = 0;
1946 ext3_rsv_window_add(sb, &sbi->s_rsv_window_head);
1947
1948 /*
1949 * set up enough so that it can read an inode
1950 */
1951 sb->s_op = &ext3_sops;
1952 sb->s_export_op = &ext3_export_ops;
1953 sb->s_xattr = ext3_xattr_handlers;
1954#ifdef CONFIG_QUOTA
1955 sb->s_qcop = &ext3_qctl_operations;
1956 sb->dq_op = &ext3_quota_operations;
1957#endif
03cb5f03 1958 memcpy(sb->s_uuid, es->s_uuid, sizeof(es->s_uuid));
1da177e4 1959 INIT_LIST_HEAD(&sbi->s_orphan); /* unlinked but open files */
b8a052d0 1960 mutex_init(&sbi->s_orphan_lock);
96d2a495 1961 mutex_init(&sbi->s_resize_lock);
1da177e4
LT		 1962
1963 sb->s_root = NULL;
1964
1965 needs_recovery = (es->s_last_orphan != 0 ||
1966 EXT3_HAS_INCOMPAT_FEATURE(sb,
1967 EXT3_FEATURE_INCOMPAT_RECOVER));
1968
1969 /*
1970 * The first inode we look at is the journal inode. Don't try
1971 * root first: it may be modified in the journal!
1972 */
1973 if (!test_opt(sb, NOLOAD) &&
1974 EXT3_HAS_COMPAT_FEATURE(sb, EXT3_FEATURE_COMPAT_HAS_JOURNAL)) {
71b96257 1975 if (ext3_load_journal(sb, es, journal_devnum))
41d1a636 1976 goto failed_mount2;
1da177e4
LT		 1977 } else if (journal_inum) {
1978 if (ext3_create_journal(sb, es, journal_inum))
41d1a636 1979 goto failed_mount2;
1da177e4
LT		 1980 } else {
1981 if (!silent)
4cf46b67
AF		 1982 ext3_msg(sb, KERN_ERR,
1983 "error: no journal found. "
1984 "mounting ext3 over ext2?");
41d1a636
DM		 1985 goto failed_mount2;
1986 }
1987 err = percpu_counter_init(&sbi->s_freeblocks_counter,
1988 ext3_count_free_blocks(sb));
1989 if (!err) {
1990 err = percpu_counter_init(&sbi->s_freeinodes_counter,
1991 ext3_count_free_inodes(sb));
1992 }
1993 if (!err) {
1994 err = percpu_counter_init(&sbi->s_dirs_counter,
1995 ext3_count_dirs(sb));
1996 }
1997 if (err) {
1998 ext3_msg(sb, KERN_ERR, "error: insufficient memory");
4569cd1b 1999 ret = err;
0216bfcf 2000 goto failed_mount3;
1da177e4
LT		 2001 }
2002
2003 /* We have now updated the journal if required, so we can
2004 * validate the data journaling mode. */
2005 switch (test_opt(sb, DATA_FLAGS)) {
2006 case 0:
2007 /* No mode set, assume a default based on the journal
2008 capabilities: ORDERED_DATA if the journal can
2009 cope, else JOURNAL_DATA */
2010 if (journal_check_available_features
2011 (sbi->s_journal, 0, 0, JFS_FEATURE_INCOMPAT_REVOKE))
bbae8bcc 2012 set_opt(sbi->s_mount_opt, DEFAULT_DATA_MODE);
1da177e4
LT		 2013 else
2014 set_opt(sbi->s_mount_opt, JOURNAL_DATA);
2015 break;
2016
2017 case EXT3_MOUNT_ORDERED_DATA:
2018 case EXT3_MOUNT_WRITEBACK_DATA:
2019 if (!journal_check_available_features
2020 (sbi->s_journal, 0, 0, JFS_FEATURE_INCOMPAT_REVOKE)) {
4cf46b67
AF		 2021 ext3_msg(sb, KERN_ERR,
2022 "error: journal does not support "
2023 "requested data journaling mode");
41d1a636 2024 goto failed_mount3;
1da177e4
LT		 2025 }
2026 default:
2027 break;
2028 }
2029
1da177e4
LT		 2030 /*
2031 * The journal_load will have done any necessary log recovery,
2032 * so we can safely mount the rest of the filesystem now.
2033 */
2034
473043dc
DH		 2035 root = ext3_iget(sb, EXT3_ROOT_INO);
2036 if (IS_ERR(root)) {
4cf46b67 2037 ext3_msg(sb, KERN_ERR, "error: get root inode failed");
473043dc 2038 ret = PTR_ERR(root);
41d1a636 2039 goto failed_mount3;
1da177e4
LT		 2040 }
2041 if (!S_ISDIR(root->i_mode) || !root->i_blocks || !root->i_size) {
473043dc 2042 iput(root);
4cf46b67 2043 ext3_msg(sb, KERN_ERR, "error: corrupt root inode, run e2fsck");
41d1a636 2044 goto failed_mount3;
1da177e4 2045 }
48fde701 2046 sb->s_root = d_make_root(root);
473043dc 2047 if (!sb->s_root) {
4cf46b67 2048 ext3_msg(sb, KERN_ERR, "error: get root dentry failed");
473043dc 2049 ret = -ENOMEM;
41d1a636 2050 goto failed_mount3;
473043dc 2051 }
1da177e4 2052
f007dbf8
ES		 2053 if (ext3_setup_super(sb, es, sb->s_flags & MS_RDONLY))
2054 sb->s_flags |= MS_RDONLY;
ed505ee4 2055
1da177e4
LT		 2056 EXT3_SB(sb)->s_mount_state |= EXT3_ORPHAN_FS;
2057 ext3_orphan_cleanup(sb, es);
2058 EXT3_SB(sb)->s_mount_state &= ~EXT3_ORPHAN_FS;
63894ab9
EG		 2059 if (needs_recovery) {
2060 ext3_mark_recovery_complete(sb, es);
4cf46b67 2061 ext3_msg(sb, KERN_INFO, "recovery complete");
63894ab9 2062 }
4cf46b67 2063 ext3_msg(sb, KERN_INFO, "mounted filesystem with %s data mode",
1da177e4
LT		 2064 test_opt(sb,DATA_FLAGS) == EXT3_MOUNT_JOURNAL_DATA ? "journal":
2065 test_opt(sb,DATA_FLAGS) == EXT3_MOUNT_ORDERED_DATA ? "ordered":
2066 "writeback");
2067
1da177e4
LT		 2068 return 0;
2069
2070cantfind_ext3:
2071 if (!silent)
4cf46b67
AF		 2072 ext3_msg(sb, KERN_INFO,
2073 "error: can't find ext3 filesystem on dev %s.",
1da177e4
LT		 2074 sb->s_id);
2075 goto failed_mount;
2076
0216bfcf
MC		 2077failed_mount3:
2078 percpu_counter_destroy(&sbi->s_freeblocks_counter);
2079 percpu_counter_destroy(&sbi->s_freeinodes_counter);
2080 percpu_counter_destroy(&sbi->s_dirs_counter);
41d1a636 2081 journal_destroy(sbi->s_journal);
1da177e4
LT		 2082failed_mount2:
2083 for (i = 0; i < db_count; i++)
2084 brelse(sbi->s_group_desc[i]);
2085 kfree(sbi->s_group_desc);
2086failed_mount:
2087#ifdef CONFIG_QUOTA
2088 for (i = 0; i < MAXQUOTAS; i++)
2089 kfree(sbi->s_qf_names[i]);
2090#endif
2091 ext3_blkdev_remove(sbi);
2092 brelse(bh);
2093out_fail:
2094 sb->s_fs_info = NULL;
de5ce037 2095 kfree(sbi->s_blockgroup_lock);
1da177e4 2096 kfree(sbi);
473043dc 2097 return ret;
1da177e4
LT		 2098}
2099
2100/*
2101 * Setup any per-fs journal parameters now. We'll do this both on
2102 * initial mount, once the journal has been initialised but before we've
ae6ddcc5 2103 * done any recovery; and again on any subsequent remount.
1da177e4
LT		 2104 */
2105static void ext3_init_journal_params(struct super_block *sb, journal_t *journal)
2106{
2107 struct ext3_sb_info *sbi = EXT3_SB(sb);
2108
2109 if (sbi->s_commit_interval)
2110 journal->j_commit_interval = sbi->s_commit_interval;
2111 /* We could also set up an ext3-specific default for the commit
2112 * interval here, but for now we'll just fall back to the jbd
2113 * default. */
2114
2115 spin_lock(&journal->j_state_lock);
2116 if (test_opt(sb, BARRIER))
2117 journal->j_flags |= JFS_BARRIER;
2118 else
2119 journal->j_flags &= ~JFS_BARRIER;
0e4fb5e2
HK		 2120 if (test_opt(sb, DATA_ERR_ABORT))
2121 journal->j_flags |= JFS_ABORT_ON_SYNCDATA_ERR;
2122 else
2123 journal->j_flags &= ~JFS_ABORT_ON_SYNCDATA_ERR;
1da177e4
LT		 2124 spin_unlock(&journal->j_state_lock);
2125}
2126
eee194e7
ES		 2127static journal_t *ext3_get_journal(struct super_block *sb,
2128 unsigned int journal_inum)
1da177e4
LT		 2129{
2130 struct inode *journal_inode;
2131 journal_t *journal;
2132
2133 /* First, test for the existence of a valid inode on disk. Bad
2134 * things happen if we iget() an unused inode, as the subsequent
2135 * iput() will try to delete it. */
2136
473043dc
DH		 2137 journal_inode = ext3_iget(sb, journal_inum);
2138 if (IS_ERR(journal_inode)) {
4cf46b67 2139 ext3_msg(sb, KERN_ERR, "error: no journal found");
1da177e4
LT		 2140 return NULL;
2141 }
2142 if (!journal_inode->i_nlink) {
2143 make_bad_inode(journal_inode);
2144 iput(journal_inode);
4cf46b67 2145 ext3_msg(sb, KERN_ERR, "error: journal inode is deleted");
1da177e4
LT		 2146 return NULL;
2147 }
2148
2149 jbd_debug(2, "Journal inode found at %p: %Ld bytes\n",
2150 journal_inode, journal_inode->i_size);
473043dc 2151 if (!S_ISREG(journal_inode->i_mode)) {
4cf46b67 2152 ext3_msg(sb, KERN_ERR, "error: invalid journal inode");
1da177e4
LT		 2153 iput(journal_inode);
2154 return NULL;
2155 }
2156
2157 journal = journal_init_inode(journal_inode);
2158 if (!journal) {
4cf46b67 2159 ext3_msg(sb, KERN_ERR, "error: could not load journal inode");
1da177e4
LT		 2160 iput(journal_inode);
2161 return NULL;
2162 }
2163 journal->j_private = sb;
2164 ext3_init_journal_params(sb, journal);
2165 return journal;
2166}
2167
2168static journal_t *ext3_get_dev_journal(struct super_block *sb,
2169 dev_t j_dev)
2170{
2171 struct buffer_head * bh;
2172 journal_t *journal;
43d23f90 2173 ext3_fsblk_t start;
1c2bf374 2174 ext3_fsblk_t len;
1da177e4 2175 int hblock, blocksize;
43d23f90 2176 ext3_fsblk_t sb_block;
1da177e4
LT		 2177 unsigned long offset;
2178 struct ext3_super_block * es;
2179 struct block_device *bdev;
2180
4cf46b67 2181 bdev = ext3_blkdev_get(j_dev, sb);
1da177e4
LT		 2182 if (bdev == NULL)
2183 return NULL;
2184
1da177e4 2185 blocksize = sb->s_blocksize;
e1defc4f 2186 hblock = bdev_logical_block_size(bdev);
1da177e4 2187 if (blocksize < hblock) {
4cf46b67
AF		 2188 ext3_msg(sb, KERN_ERR,
2189 "error: blocksize too small for journal device");
1da177e4
LT		 2190 goto out_bdev;
2191 }
2192
2193 sb_block = EXT3_MIN_BLOCK_SIZE / blocksize;
2194 offset = EXT3_MIN_BLOCK_SIZE % blocksize;
2195 set_blocksize(bdev, blocksize);
2196 if (!(bh = __bread(bdev, sb_block, blocksize))) {
4cf46b67
AF		 2197 ext3_msg(sb, KERN_ERR, "error: couldn't read superblock of "
2198 "external journal");
1da177e4
LT		 2199 goto out_bdev;
2200 }
2201
57e94d86 2202 es = (struct ext3_super_block *) (bh->b_data + offset);
1da177e4
LT		 2203 if ((le16_to_cpu(es->s_magic) != EXT3_SUPER_MAGIC) ||
2204 !(le32_to_cpu(es->s_feature_incompat) &
2205 EXT3_FEATURE_INCOMPAT_JOURNAL_DEV)) {
4cf46b67
AF		 2206 ext3_msg(sb, KERN_ERR, "error: external journal has "
2207 "bad superblock");
1da177e4
LT		 2208 brelse(bh);
2209 goto out_bdev;
2210 }
2211
2212 if (memcmp(EXT3_SB(sb)->s_es->s_journal_uuid, es->s_uuid, 16)) {
4cf46b67 2213 ext3_msg(sb, KERN_ERR, "error: journal UUID does not match");
1da177e4
LT		 2214 brelse(bh);
2215 goto out_bdev;
2216 }
2217
2218 len = le32_to_cpu(es->s_blocks_count);
2219 start = sb_block + 1;
2220 brelse(bh); /* we're done with the superblock */
2221
2222 journal = journal_init_dev(bdev, sb->s_bdev,
2223 start, len, blocksize);
2224 if (!journal) {
4cf46b67
AF		 2225 ext3_msg(sb, KERN_ERR,
2226 "error: failed to create device journal");
1da177e4
LT		 2227 goto out_bdev;
2228 }
2229 journal->j_private = sb;
d03e1292
ZL		 2230 if (!bh_uptodate_or_lock(journal->j_sb_buffer)) {
2231 if (bh_submit_read(journal->j_sb_buffer)) {
2232 ext3_msg(sb, KERN_ERR, "I/O error on journal device");
2233 goto out_journal;
2234 }
1da177e4
LT		 2235 }
2236 if (be32_to_cpu(journal->j_superblock->s_nr_users) != 1) {
4cf46b67
AF		 2237 ext3_msg(sb, KERN_ERR,
2238 "error: external journal has more than one "
2239 "user (unsupported) - %d",
1da177e4
LT		 2240 be32_to_cpu(journal->j_superblock->s_nr_users));
2241 goto out_journal;
2242 }
2243 EXT3_SB(sb)->journal_bdev = bdev;
2244 ext3_init_journal_params(sb, journal);
2245 return journal;
2246out_journal:
2247 journal_destroy(journal);
2248out_bdev:
2249 ext3_blkdev_put(bdev);
2250 return NULL;
2251}
2252
71b96257
JL		 2253static int ext3_load_journal(struct super_block *sb,
2254 struct ext3_super_block *es,
2255 unsigned long journal_devnum)
1da177e4
LT		 2256{
2257 journal_t *journal;
eee194e7 2258 unsigned int journal_inum = le32_to_cpu(es->s_journal_inum);
71b96257 2259 dev_t journal_dev;
1da177e4
LT		 2260 int err = 0;
2261 int really_read_only;
2262
71b96257
JL		 2263 if (journal_devnum &&
2264 journal_devnum != le32_to_cpu(es->s_journal_dev)) {
4cf46b67
AF		 2265 ext3_msg(sb, KERN_INFO, "external journal device major/minor "
2266 "numbers have changed");
71b96257
JL		 2267 journal_dev = new_decode_dev(journal_devnum);
2268 } else
2269 journal_dev = new_decode_dev(le32_to_cpu(es->s_journal_dev));
2270
1da177e4
LT		 2271 really_read_only = bdev_read_only(sb->s_bdev);
2272
2273 /*
2274 * Are we loading a blank journal or performing recovery after a
2275 * crash? For recovery, we need to check in advance whether we
2276 * can get read-write access to the device.
2277 */
2278
2279 if (EXT3_HAS_INCOMPAT_FEATURE(sb, EXT3_FEATURE_INCOMPAT_RECOVER)) {
2280 if (sb->s_flags & MS_RDONLY) {
4cf46b67
AF		 2281 ext3_msg(sb, KERN_INFO,
2282 "recovery required on readonly filesystem");
1da177e4 2283 if (really_read_only) {
4cf46b67
AF		 2284 ext3_msg(sb, KERN_ERR, "error: write access "
2285 "unavailable, cannot proceed");
1da177e4
LT		 2286 return -EROFS;
2287 }
4cf46b67
AF		 2288 ext3_msg(sb, KERN_INFO,
2289 "write access will be enabled during recovery");
1da177e4
LT		 2290 }
2291 }
2292
2293 if (journal_inum && journal_dev) {
4cf46b67
AF		 2294 ext3_msg(sb, KERN_ERR, "error: filesystem has both journal "
2295 "and inode journals");
1da177e4
LT		 2296 return -EINVAL;
2297 }
2298
2299 if (journal_inum) {
2300 if (!(journal = ext3_get_journal(sb, journal_inum)))
2301 return -EINVAL;
2302 } else {
2303 if (!(journal = ext3_get_dev_journal(sb, journal_dev)))
2304 return -EINVAL;
2305 }
2306
0636c73e
ES		 2307 if (!(journal->j_flags & JFS_BARRIER))
2308 printk(KERN_INFO "EXT3-fs: barriers not enabled\n");
2309
1da177e4
LT		 2310 if (!really_read_only && test_opt(sb, UPDATE_JOURNAL)) {
2311 err = journal_update_format(journal);
2312 if (err) {
4cf46b67 2313 ext3_msg(sb, KERN_ERR, "error updating journal");
1da177e4
LT		 2314 journal_destroy(journal);
2315 return err;
2316 }
2317 }
2318
2319 if (!EXT3_HAS_INCOMPAT_FEATURE(sb, EXT3_FEATURE_INCOMPAT_RECOVER))
2320 err = journal_wipe(journal, !really_read_only);
2321 if (!err)
2322 err = journal_load(journal);
2323
2324 if (err) {
4cf46b67 2325 ext3_msg(sb, KERN_ERR, "error loading journal");
1da177e4
LT		 2326 journal_destroy(journal);
2327 return err;
2328 }
2329
2330 EXT3_SB(sb)->s_journal = journal;
2331 ext3_clear_journal_err(sb, es);
71b96257 2332
31d710a7 2333 if (!really_read_only && journal_devnum &&
71b96257
JL		 2334 journal_devnum != le32_to_cpu(es->s_journal_dev)) {
2335 es->s_journal_dev = cpu_to_le32(journal_devnum);
71b96257
JL		 2336
2337 /* Make sure we flush the recovery flag to disk. */
2338 ext3_commit_super(sb, es, 1);
2339 }
2340
1da177e4
LT		 2341 return 0;
2342}
2343
4cf46b67
AF		 2344static int ext3_create_journal(struct super_block *sb,
2345 struct ext3_super_block *es,
eee194e7 2346 unsigned int journal_inum)
1da177e4
LT		 2347{
2348 journal_t *journal;
952d9de1 2349 int err;
1da177e4
LT		 2350
2351 if (sb->s_flags & MS_RDONLY) {
4cf46b67
AF		 2352 ext3_msg(sb, KERN_ERR,
2353 "error: readonly filesystem when trying to "
2354 "create journal");
1da177e4
LT		 2355 return -EROFS;
2356 }
2357
952d9de1
BP		 2358 journal = ext3_get_journal(sb, journal_inum);
2359 if (!journal)
1da177e4
LT		 2360 return -EINVAL;
2361
4cf46b67 2362 ext3_msg(sb, KERN_INFO, "creating new journal on inode %u",
1da177e4
LT		 2363 journal_inum);
2364
952d9de1
BP		 2365 err = journal_create(journal);
2366 if (err) {
4cf46b67 2367 ext3_msg(sb, KERN_ERR, "error creating journal");
1da177e4
LT		 2368 journal_destroy(journal);
2369 return -EIO;
2370 }
2371
2372 EXT3_SB(sb)->s_journal = journal;
2373
2374 ext3_update_dynamic_rev(sb);
2375 EXT3_SET_INCOMPAT_FEATURE(sb, EXT3_FEATURE_INCOMPAT_RECOVER);
2376 EXT3_SET_COMPAT_FEATURE(sb, EXT3_FEATURE_COMPAT_HAS_JOURNAL);
2377
2378 es->s_journal_inum = cpu_to_le32(journal_inum);
1da177e4
LT		 2379
2380 /* Make sure we flush the recovery flag to disk. */
2381 ext3_commit_super(sb, es, 1);
2382
2383 return 0;
2384}
2385
c4be0c1d
TS		 2386static int ext3_commit_super(struct super_block *sb,
2387 struct ext3_super_block *es,
1da177e4
LT		 2388 int sync)
2389{
2390 struct buffer_head *sbh = EXT3_SB(sb)->s_sbh;
c4be0c1d 2391 int error = 0;
1da177e4
LT		 2392
2393 if (!sbh)
c4be0c1d 2394 return error;
dff6825e
DW		 2395
2396 if (buffer_write_io_error(sbh)) {
2397 /*
2398 * Oh, dear. A previous attempt to write the
2399 * superblock failed. This could happen because the
2400 * USB device was yanked out. Or it could happen to
2401 * be a transient write error and maybe the block will
2402 * be remapped. Nothing we can do but to retry the
2403 * write and hope for the best.
2404 */
2405 ext3_msg(sb, KERN_ERR, "previous I/O error to "
2406 "superblock detected");
2407 clear_buffer_write_io_error(sbh);
2408 set_buffer_uptodate(sbh);
2409 }
96ec2e0a
TT		 2410 /*
2411 * If the file system is mounted read-only, don't update the
2412 * superblock write time. This avoids updating the superblock
2413 * write time when we are mounting the root file system
2414 * read/only but we need to replay the journal; at that point,
2415 * for people who are east of GMT and who make their clock
2416 * tick in localtime for Windows bug-for-bug compatibility,
2417 * the clock is set in the future, and this will cause e2fsck
2418 * to complain and force a full file system check.
2419 */
2420 if (!(sb->s_flags & MS_RDONLY))
2421 es->s_wtime = cpu_to_le32(get_seconds());
1da177e4
LT		 2422 es->s_free_blocks_count = cpu_to_le32(ext3_count_free_blocks(sb));
2423 es->s_free_inodes_count = cpu_to_le32(ext3_count_free_inodes(sb));
2424 BUFFER_TRACE(sbh, "marking dirty");
2425 mark_buffer_dirty(sbh);
dff6825e 2426 if (sync) {
c4be0c1d 2427 error = sync_dirty_buffer(sbh);
dff6825e
DW		 2428 if (buffer_write_io_error(sbh)) {
2429 ext3_msg(sb, KERN_ERR, "I/O error while writing "
2430 "superblock");
2431 clear_buffer_write_io_error(sbh);
2432 set_buffer_uptodate(sbh);
2433 }
2434 }
c4be0c1d 2435 return error;
1da177e4
LT		 2436}
2437
2438
2439/*
2440 * Have we just finished recovery? If so, and if we are mounting (or
2441 * remounting) the filesystem readonly, then we will end up with a
2442 * consistent fs on disk. Record that fact.
2443 */
2444static void ext3_mark_recovery_complete(struct super_block * sb,
2445 struct ext3_super_block * es)
2446{
2447 journal_t *journal = EXT3_SB(sb)->s_journal;
2448
2449 journal_lock_updates(journal);
2d7c820e
HK		 2450 if (journal_flush(journal) < 0)
2451 goto out;
2452
1da177e4
LT		 2453 if (EXT3_HAS_INCOMPAT_FEATURE(sb, EXT3_FEATURE_INCOMPAT_RECOVER) &&
2454 sb->s_flags & MS_RDONLY) {
2455 EXT3_CLEAR_INCOMPAT_FEATURE(sb, EXT3_FEATURE_INCOMPAT_RECOVER);
1da177e4
LT		 2456 ext3_commit_super(sb, es, 1);
2457 }
2d7c820e
HK		 2458
2459out:
1da177e4
LT		 2460 journal_unlock_updates(journal);
2461}
2462
2463/*
2464 * If we are mounting (or read-write remounting) a filesystem whose journal
2465 * has recorded an error from a previous lifetime, move that error to the
2466 * main filesystem now.
2467 */
4cf46b67
AF		 2468static void ext3_clear_journal_err(struct super_block *sb,
2469 struct ext3_super_block *es)
1da177e4
LT		 2470{
2471 journal_t *journal;
2472 int j_errno;
2473 const char *errstr;
2474
2475 journal = EXT3_SB(sb)->s_journal;
2476
2477 /*
2478 * Now check for any error status which may have been recorded in the
2479 * journal by a prior ext3_error() or ext3_abort()
2480 */
2481
2482 j_errno = journal_errno(journal);
2483 if (j_errno) {
2484 char nbuf[16];
2485
2486 errstr = ext3_decode_error(sb, j_errno, nbuf);
e05b6b52 2487 ext3_warning(sb, __func__, "Filesystem error recorded "
1da177e4 2488 "from previous mount: %s", errstr);
e05b6b52 2489 ext3_warning(sb, __func__, "Marking fs in need of "
1da177e4
LT		 2490 "filesystem check.");
2491
2492 EXT3_SB(sb)->s_mount_state |= EXT3_ERROR_FS;
2493 es->s_state |= cpu_to_le16(EXT3_ERROR_FS);
2494 ext3_commit_super (sb, es, 1);
2495
2496 journal_clear_err(journal);
2497 }
2498}
2499
2500/*
2501 * Force the running and committing transactions to commit,
2502 * and wait on the commit.
2503 */
2504int ext3_force_commit(struct super_block *sb)
2505{
2506 journal_t *journal;
2507 int ret;
2508
2509 if (sb->s_flags & MS_RDONLY)
2510 return 0;
2511
2512 journal = EXT3_SB(sb)->s_journal;
1da177e4
LT		 2513 ret = ext3_journal_force_commit(journal);
2514 return ret;
2515}
2516
1da177e4
LT		 2517static int ext3_sync_fs(struct super_block *sb, int wait)
2518{
02ac597c 2519 tid_t target;
c87591b7 2520
785c4bcc 2521 trace_ext3_sync_fs(sb, wait);
a1177825
JK		 2522 /*
2523 * Writeback quota in non-journalled quota case - journalled quota has
2524 * no dirty dquots
2525 */
2526 dquot_writeback_dquots(sb, -1);
02ac597c
JK		 2527 if (journal_start_commit(EXT3_SB(sb)->s_journal, &target)) {
2528 if (wait)
2529 log_wait_commit(EXT3_SB(sb)->s_journal, target);
2530 }
1da177e4
LT		 2531 return 0;
2532}
2533
2534/*
2535 * LVM calls this function before a (read-only) snapshot is created. This
2536 * gives us a chance to flush the journal completely and mark the fs clean.
2537 */
c4be0c1d 2538static int ext3_freeze(struct super_block *sb)
1da177e4 2539{
c4be0c1d
TS		 2540 int error = 0;
2541 journal_t *journal;
1da177e4
LT		 2542
2543 if (!(sb->s_flags & MS_RDONLY)) {
c4be0c1d 2544 journal = EXT3_SB(sb)->s_journal;
1da177e4
LT		 2545
2546 /* Now we set up the journal barrier. */
2547 journal_lock_updates(journal);
2d7c820e
HK		 2548
2549 /*
2550 * We don't want to clear needs_recovery flag when we failed
2551 * to flush the journal.
2552 */
c4be0c1d
TS		 2553 error = journal_flush(journal);
2554 if (error < 0)
2555 goto out;
1da177e4
LT		 2556
2557 /* Journal blocked and flushed, clear needs_recovery flag. */
2558 EXT3_CLEAR_INCOMPAT_FEATURE(sb, EXT3_FEATURE_INCOMPAT_RECOVER);
c4be0c1d
TS		 2559 error = ext3_commit_super(sb, EXT3_SB(sb)->s_es, 1);
2560 if (error)
2561 goto out;
1da177e4 2562 }
c4be0c1d
TS		 2563 return 0;
2564
2565out:
2566 journal_unlock_updates(journal);
2567 return error;
1da177e4
LT		 2568}
2569
2570/*
2571 * Called by LVM after the snapshot is done. We need to reset the RECOVER
2572 * flag here, even though the filesystem is not technically dirty yet.
2573 */
c4be0c1d 2574static int ext3_unfreeze(struct super_block *sb)
1da177e4
LT		 2575{
2576 if (!(sb->s_flags & MS_RDONLY)) {
1da177e4
LT		 2577 /* Reser the needs_recovery flag before the fs is unlocked. */
2578 EXT3_SET_INCOMPAT_FEATURE(sb, EXT3_FEATURE_INCOMPAT_RECOVER);
2579 ext3_commit_super(sb, EXT3_SB(sb)->s_es, 1);
1da177e4
LT		 2580 journal_unlock_updates(EXT3_SB(sb)->s_journal);
2581 }
c4be0c1d 2582 return 0;
1da177e4
LT		 2583}
2584
2585static int ext3_remount (struct super_block * sb, int * flags, char * data)
2586{
2587 struct ext3_super_block * es;
2588 struct ext3_sb_info *sbi = EXT3_SB(sb);
43d23f90 2589 ext3_fsblk_t n_blocks_count = 0;
08c6a96f
JK		 2590 unsigned long old_sb_flags;
2591 struct ext3_mount_options old_opts;
c79d967d 2592 int enable_quota = 0;
08c6a96f
JK		 2593 int err;
2594#ifdef CONFIG_QUOTA
2595 int i;
2596#endif
2597
2598 /* Store the original options */
2599 old_sb_flags = sb->s_flags;
2600 old_opts.s_mount_opt = sbi->s_mount_opt;
2601 old_opts.s_resuid = sbi->s_resuid;
2602 old_opts.s_resgid = sbi->s_resgid;
2603 old_opts.s_commit_interval = sbi->s_commit_interval;
2604#ifdef CONFIG_QUOTA
2605 old_opts.s_jquota_fmt = sbi->s_jquota_fmt;
2606 for (i = 0; i < MAXQUOTAS; i++)
f56426ae
JK		 2607 if (sbi->s_qf_names[i]) {
2608 old_opts.s_qf_names[i] = kstrdup(sbi->s_qf_names[i],
2609 GFP_KERNEL);
2610 if (!old_opts.s_qf_names[i]) {
2611 int j;
2612
2613 for (j = 0; j < i; j++)
2614 kfree(old_opts.s_qf_names[j]);
2615 return -ENOMEM;
2616 }
2617 } else
2618 old_opts.s_qf_names[i] = NULL;
08c6a96f 2619#endif
1da177e4
LT		 2620
2621 /*
2622 * Allow the "check" option to be passed as a remount option.
2623 */
71b96257 2624 if (!parse_options(data, sb, NULL, NULL, &n_blocks_count, 1)) {
08c6a96f
JK		 2625 err = -EINVAL;
2626 goto restore_opts;
2627 }
1da177e4 2628
e3c96435 2629 if (test_opt(sb, ABORT))
e05b6b52 2630 ext3_abort(sb, __func__, "Abort forced by user");
1da177e4
LT		 2631
2632 sb->s_flags = (sb->s_flags & ~MS_POSIXACL) |
e3c96435 2633 (test_opt(sb, POSIX_ACL) ? MS_POSIXACL : 0);
1da177e4
LT		 2634
2635 es = sbi->s_es;
2636
2637 ext3_init_journal_params(sb, sbi->s_journal);
2638
2639 if ((*flags & MS_RDONLY) != (sb->s_flags & MS_RDONLY) ||
2640 n_blocks_count > le32_to_cpu(es->s_blocks_count)) {
e3c96435 2641 if (test_opt(sb, ABORT)) {
08c6a96f
JK		 2642 err = -EROFS;
2643 goto restore_opts;
2644 }
1da177e4
LT		 2645
2646 if (*flags & MS_RDONLY) {
0f0dd62f
CH		 2647 err = dquot_suspend(sb, -1);
2648 if (err < 0)
c79d967d 2649 goto restore_opts;
c79d967d 2650
1da177e4
LT		 2651 /*
2652 * First of all, the unconditional stuff we have to do
2653 * to disable replay of the journal when we next remount
2654 */
2655 sb->s_flags |= MS_RDONLY;
2656
2657 /*
2658 * OK, test if we are remounting a valid rw partition
2659 * readonly, and if so set the rdonly flag and then
2660 * mark the partition as valid again.
2661 */
2662 if (!(es->s_state & cpu_to_le16(EXT3_VALID_FS)) &&
2663 (sbi->s_mount_state & EXT3_VALID_FS))
2664 es->s_state = cpu_to_le16(sbi->s_mount_state);
2665
2666 ext3_mark_recovery_complete(sb, es);
2667 } else {
2668 __le32 ret;
2669 if ((ret = EXT3_HAS_RO_COMPAT_FEATURE(sb,
2670 ~EXT3_FEATURE_RO_COMPAT_SUPP))) {
4cf46b67
AF		 2671 ext3_msg(sb, KERN_WARNING,
2672 "warning: couldn't remount RDWR "
2673 "because of unsupported optional "
2674 "features (%x)", le32_to_cpu(ret));
08c6a96f
JK		 2675 err = -EROFS;
2676 goto restore_opts;
1da177e4 2677 }
ea9a05a1
ES		 2678
2679 /*
2680 * If we have an unprocessed orphan list hanging
2681 * around from a previously readonly bdev mount,
1cde201d 2682 * require a full umount & mount for now.
ea9a05a1
ES		 2683 */
2684 if (es->s_last_orphan) {
4cf46b67 2685 ext3_msg(sb, KERN_WARNING, "warning: couldn't "
ea9a05a1
ES		 2686 "remount RDWR because of unprocessed "
2687 "orphan inode list. Please "
1cde201d 2688 "umount & mount instead.");
ea9a05a1
ES		 2689 err = -EINVAL;
2690 goto restore_opts;
2691 }
2692
1da177e4
LT		 2693 /*
2694 * Mounting a RDONLY partition read-write, so reread
2695 * and store the current valid flag. (It may have
2696 * been changed by e2fsck since we originally mounted
2697 * the partition.)
2698 */
2699 ext3_clear_journal_err(sb, es);
2700 sbi->s_mount_state = le16_to_cpu(es->s_state);
a4e4de36 2701 if ((err = ext3_group_extend(sb, es, n_blocks_count)))
08c6a96f 2702 goto restore_opts;
1da177e4
LT		 2703 if (!ext3_setup_super (sb, es, 0))
2704 sb->s_flags &= ~MS_RDONLY;
c79d967d 2705 enable_quota = 1;
1da177e4
LT		 2706 }
2707 }
08c6a96f
JK		 2708#ifdef CONFIG_QUOTA
2709 /* Release old quota file names */
2710 for (i = 0; i < MAXQUOTAS; i++)
f56426ae 2711 kfree(old_opts.s_qf_names[i]);
08c6a96f 2712#endif
c79d967d 2713 if (enable_quota)
0f0dd62f 2714 dquot_resume(sb, -1);
1da177e4 2715 return 0;
08c6a96f
JK		 2716restore_opts:
2717 sb->s_flags = old_sb_flags;
2718 sbi->s_mount_opt = old_opts.s_mount_opt;
2719 sbi->s_resuid = old_opts.s_resuid;
2720 sbi->s_resgid = old_opts.s_resgid;
2721 sbi->s_commit_interval = old_opts.s_commit_interval;
2722#ifdef CONFIG_QUOTA
2723 sbi->s_jquota_fmt = old_opts.s_jquota_fmt;
2724 for (i = 0; i < MAXQUOTAS; i++) {
f56426ae 2725 kfree(sbi->s_qf_names[i]);
08c6a96f
JK		 2726 sbi->s_qf_names[i] = old_opts.s_qf_names[i];
2727 }
2728#endif
2729 return err;
1da177e4
LT		 2730}
2731
726c3342 2732static int ext3_statfs (struct dentry * dentry, struct kstatfs * buf)
1da177e4 2733{
726c3342 2734 struct super_block *sb = dentry->d_sb;
09fe316a
AT		 2735 struct ext3_sb_info *sbi = EXT3_SB(sb);
2736 struct ext3_super_block *es = sbi->s_es;
50ee0a32 2737 u64 fsid;
1da177e4 2738
a71ce8c6
BP		 2739 if (test_opt(sb, MINIX_DF)) {
2740 sbi->s_overhead_last = 0;
2741 } else if (sbi->s_blocks_last != le32_to_cpu(es->s_blocks_count)) {
2742 unsigned long ngroups = sbi->s_groups_count, i;
2743 ext3_fsblk_t overhead = 0;
1da177e4
LT		 2744 smp_rmb();
2745
2746 /*
a71ce8c6
BP		 2747 * Compute the overhead (FS structures). This is constant
2748 * for a given filesystem unless the number of block groups
2749 * changes so we cache the previous value until it does.
1da177e4
LT		 2750 */
2751
2752 /*
2753 * All of the blocks before first_data_block are
2754 * overhead
2755 */
2756 overhead = le32_to_cpu(es->s_first_data_block);
2757
2758 /*
2759 * Add the overhead attributed to the superblock and
2760 * block group descriptors. If the sparse superblocks
2761 * feature is turned on, then not all groups have this.
2762 */
2763 for (i = 0; i < ngroups; i++) {
2764 overhead += ext3_bg_has_super(sb, i) +
2765 ext3_bg_num_gdb(sb, i);
2766 cond_resched();
2767 }
2768
2769 /*
2770 * Every block group has an inode bitmap, a block
2771 * bitmap, and an inode table.
2772 */
a71ce8c6
BP		 2773 overhead += ngroups * (2 + sbi->s_itb_per_group);
2774 sbi->s_overhead_last = overhead;
2775 smp_wmb();
2776 sbi->s_blocks_last = le32_to_cpu(es->s_blocks_count);
1da177e4
LT		 2777 }
2778
2779 buf->f_type = EXT3_SUPER_MAGIC;
2780 buf->f_bsize = sb->s_blocksize;
a71ce8c6 2781 buf->f_blocks = le32_to_cpu(es->s_blocks_count) - sbi->s_overhead_last;
52d9f3b4 2782 buf->f_bfree = percpu_counter_sum_positive(&sbi->s_freeblocks_counter);
1da177e4
LT		 2783 buf->f_bavail = buf->f_bfree - le32_to_cpu(es->s_r_blocks_count);
2784 if (buf->f_bfree < le32_to_cpu(es->s_r_blocks_count))
2785 buf->f_bavail = 0;
2786 buf->f_files = le32_to_cpu(es->s_inodes_count);
52d9f3b4 2787 buf->f_ffree = percpu_counter_sum_positive(&sbi->s_freeinodes_counter);
1da177e4 2788 buf->f_namelen = EXT3_NAME_LEN;
50ee0a32
PE		 2789 fsid = le64_to_cpup((void *)es->s_uuid) ^
2790 le64_to_cpup((void *)es->s_uuid + sizeof(u64));
2791 buf->f_fsid.val[0] = fsid & 0xFFFFFFFFUL;
2792 buf->f_fsid.val[1] = (fsid >> 32) & 0xFFFFFFFFUL;
1da177e4
LT		 2793 return 0;
2794}
2795
2796/* Helper function for writing quotas on sync - we need to start transaction before quota file
2797 * is locked for write. Otherwise the are possible deadlocks:
2798 * Process 1 Process 2
2799 * ext3_create() quota_sync()
2800 * journal_start() write_dquot()
871a2931 2801 * dquot_initialize() down(dqio_mutex)
d3be915f 2802 * down(dqio_mutex) journal_start()
1da177e4
LT		 2803 *
2804 */
2805
2806#ifdef CONFIG_QUOTA
2807
2808static inline struct inode *dquot_to_inode(struct dquot *dquot)
2809{
4c376dca 2810 return sb_dqopt(dquot->dq_sb)->files[dquot->dq_id.type];
1da177e4
LT		 2811}
2812
1da177e4
LT		 2813static int ext3_write_dquot(struct dquot *dquot)
2814{
2815 int ret, err;
2816 handle_t *handle;
2817 struct inode *inode;
2818
2819 inode = dquot_to_inode(dquot);
2820 handle = ext3_journal_start(inode,
1f54587b 2821 EXT3_QUOTA_TRANS_BLOCKS(dquot->dq_sb));
1da177e4
LT		 2822 if (IS_ERR(handle))
2823 return PTR_ERR(handle);
2824 ret = dquot_commit(dquot);
2825 err = ext3_journal_stop(handle);
2826 if (!ret)
2827 ret = err;
2828 return ret;
2829}
2830
2831static int ext3_acquire_dquot(struct dquot *dquot)
2832{
2833 int ret, err;
2834 handle_t *handle;
2835
2836 handle = ext3_journal_start(dquot_to_inode(dquot),
1f54587b 2837 EXT3_QUOTA_INIT_BLOCKS(dquot->dq_sb));
1da177e4
LT		 2838 if (IS_ERR(handle))
2839 return PTR_ERR(handle);
2840 ret = dquot_acquire(dquot);
2841 err = ext3_journal_stop(handle);
2842 if (!ret)
2843 ret = err;
2844 return ret;
2845}
2846
2847static int ext3_release_dquot(struct dquot *dquot)
2848{
2849 int ret, err;
2850 handle_t *handle;
2851
2852 handle = ext3_journal_start(dquot_to_inode(dquot),
1f54587b 2853 EXT3_QUOTA_DEL_BLOCKS(dquot->dq_sb));
9c3013e9
JK		 2854 if (IS_ERR(handle)) {
2855 /* Release dquot anyway to avoid endless cycle in dqput() */
2856 dquot_release(dquot);
1da177e4 2857 return PTR_ERR(handle);
9c3013e9 2858 }
1da177e4
LT		 2859 ret = dquot_release(dquot);
2860 err = ext3_journal_stop(handle);
2861 if (!ret)
2862 ret = err;
2863 return ret;
2864}
2865
2866static int ext3_mark_dquot_dirty(struct dquot *dquot)
2867{
99aeaf63 2868 /* Are we journaling quotas? */
1da177e4
LT		 2869 if (EXT3_SB(dquot->dq_sb)->s_qf_names[USRQUOTA] ||
2870 EXT3_SB(dquot->dq_sb)->s_qf_names[GRPQUOTA]) {
2871 dquot_mark_dquot_dirty(dquot);
2872 return ext3_write_dquot(dquot);
2873 } else {
2874 return dquot_mark_dquot_dirty(dquot);
2875 }
2876}
2877
2878static int ext3_write_info(struct super_block *sb, int type)
2879{
2880 int ret, err;
2881 handle_t *handle;
2882
2883 /* Data block + inode block */
2884 handle = ext3_journal_start(sb->s_root->d_inode, 2);
2885 if (IS_ERR(handle))
2886 return PTR_ERR(handle);
2887 ret = dquot_commit_info(sb, type);
2888 err = ext3_journal_stop(handle);
2889 if (!ret)
2890 ret = err;
2891 return ret;
2892}
2893
2894/*
2895 * Turn on quotas during mount time - we need to find
2896 * the quota file and such...
2897 */
2898static int ext3_quota_on_mount(struct super_block *sb, int type)
2899{
287a8095
CH		 2900 return dquot_quota_on_mount(sb, EXT3_SB(sb)->s_qf_names[type],
2901 EXT3_SB(sb)->s_jquota_fmt, type);
1da177e4
LT		 2902}
2903
2904/*
2905 * Standard function to be called on quota_on
2906 */
2907static int ext3_quota_on(struct super_block *sb, int type, int format_id,
f00c9e44 2908 struct path *path)
1da177e4
LT		 2909{
2910 int err;
1da177e4 2911
1f54587b
JK		 2912 if (!test_opt(sb, QUOTA))
2913 return -EINVAL;
9cfe7b90 2914
1da177e4 2915 /* Quotafile not on the same filesystem? */
d8c9584e 2916 if (path->dentry->d_sb != sb)
1da177e4 2917 return -EXDEV;
9cfe7b90
JK		 2918 /* Journaling quota? */
2919 if (EXT3_SB(sb)->s_qf_names[type]) {
2920 /* Quotafile not of fs root? */
f00c9e44 2921 if (path->dentry->d_parent != sb->s_root)
4cf46b67
AF		 2922 ext3_msg(sb, KERN_WARNING,
2923 "warning: Quota file not on filesystem root. "
2924 "Journaled quota will not work.");
9cfe7b90
JK		 2925 }
2926
2927 /*
2928 * When we journal data on quota file, we have to flush journal to see
2929 * all updates to the file when we bypass pagecache...
2930 */
f00c9e44 2931 if (ext3_should_journal_data(path->dentry->d_inode)) {
9cfe7b90
JK		 2932 /*
2933 * We don't need to lock updates but journal_flush() could
2934 * otherwise be livelocked...
2935 */
2936 journal_lock_updates(EXT3_SB(sb)->s_journal);
2d7c820e 2937 err = journal_flush(EXT3_SB(sb)->s_journal);
9cfe7b90 2938 journal_unlock_updates(EXT3_SB(sb)->s_journal);
f00c9e44 2939 if (err)
2d7c820e 2940 return err;
9cfe7b90
JK		 2941 }
2942
f00c9e44 2943 return dquot_quota_on(sb, type, format_id, path);
1da177e4
LT		 2944}
2945
2946/* Read data from quotafile - avoid pagecache and such because we cannot afford
2947 * acquiring the locks... As quota files are never truncated and quota code
25985edc 2948 * itself serializes the operations (and no one else should touch the files)
1da177e4
LT		 2949 * we don't have to be afraid of races */
2950static ssize_t ext3_quota_read(struct super_block *sb, int type, char *data,
2951 size_t len, loff_t off)
2952{
2953 struct inode *inode = sb_dqopt(sb)->files[type];
2954 sector_t blk = off >> EXT3_BLOCK_SIZE_BITS(sb);
2955 int err = 0;
2956 int offset = off & (sb->s_blocksize - 1);
2957 int tocopy;
2958 size_t toread;
2959 struct buffer_head *bh;
2960 loff_t i_size = i_size_read(inode);
2961
2962 if (off > i_size)
2963 return 0;
2964 if (off+len > i_size)
2965 len = i_size-off;
2966 toread = len;
2967 while (toread > 0) {
2968 tocopy = sb->s_blocksize - offset < toread ?
2969 sb->s_blocksize - offset : toread;
2970 bh = ext3_bread(NULL, inode, blk, 0, &err);
2971 if (err)
2972 return err;
2973 if (!bh) /* A hole? */
2974 memset(data, 0, tocopy);
2975 else
2976 memcpy(data, bh->b_data+offset, tocopy);
2977 brelse(bh);
2978 offset = 0;
2979 toread -= tocopy;
2980 data += tocopy;
2981 blk++;
2982 }
2983 return len;
2984}
2985
2986/* Write to quotafile (we know the transaction is already started and has
2987 * enough credits) */
2988static ssize_t ext3_quota_write(struct super_block *sb, int type,
2989 const char *data, size_t len, loff_t off)
2990{
2991 struct inode *inode = sb_dqopt(sb)->files[type];
2992 sector_t blk = off >> EXT3_BLOCK_SIZE_BITS(sb);
2993 int err = 0;
2994 int offset = off & (sb->s_blocksize - 1);
1da177e4 2995 int journal_quota = EXT3_SB(sb)->s_qf_names[type] != NULL;
1da177e4
LT		 2996 struct buffer_head *bh;
2997 handle_t *handle = journal_current_handle();
2998
9c3013e9 2999 if (!handle) {
4cf46b67
AF		 3000 ext3_msg(sb, KERN_WARNING,
3001 "warning: quota write (off=%llu, len=%llu)"
3002 " cancelled because transaction is not started.",
9c3013e9
JK		 3003 (unsigned long long)off, (unsigned long long)len);
3004 return -EIO;
3005 }
e5472147
DM		 3006
3007 /*
3008 * Since we account only one data block in transaction credits,
3009 * then it is impossible to cross a block boundary.
3010 */
3011 if (sb->s_blocksize - offset < len) {
3012 ext3_msg(sb, KERN_WARNING, "Quota write (off=%llu, len=%llu)"
3013 " cancelled because not block aligned",
3014 (unsigned long long)off, (unsigned long long)len);
3015 return -EIO;
3016 }
e5472147
DM		 3017 bh = ext3_bread(handle, inode, blk, 1, &err);
3018 if (!bh)
3019 goto out;
3020 if (journal_quota) {
3021 err = ext3_journal_get_write_access(handle, bh);
3022 if (err) {
3023 brelse(bh);
1da177e4 3024 goto out;
1da177e4 3025 }
1da177e4 3026 }
e5472147
DM		 3027 lock_buffer(bh);
3028 memcpy(bh->b_data+offset, data, len);
3029 flush_dcache_page(bh->b_page);
3030 unlock_buffer(bh);
3031 if (journal_quota)
3032 err = ext3_journal_dirty_metadata(handle, bh);
3033 else {
3034 /* Always do at least ordered writes for quotas */
3035 err = ext3_journal_dirty_data(handle, bh);
3036 mark_buffer_dirty(bh);
3037 }
3038 brelse(bh);
1da177e4 3039out:
905c3937 3040 if (err)
1da177e4 3041 return err;
e5472147
DM		 3042 if (inode->i_size < off + len) {
3043 i_size_write(inode, off + len);
1da177e4
LT		 3044 EXT3_I(inode)->i_disksize = inode->i_size;
3045 }
3046 inode->i_version++;
3047 inode->i_mtime = inode->i_ctime = CURRENT_TIME;
3048 ext3_mark_inode_dirty(handle, inode);
e5472147 3049 return len;
1da177e4
LT		 3050}
3051
3052#endif
3053
152a0836
AV		 3054static struct dentry *ext3_mount(struct file_system_type *fs_type,
3055 int flags, const char *dev_name, void *data)
1da177e4 3056{
152a0836 3057 return mount_bdev(fs_type, flags, dev_name, data, ext3_fill_super);
1da177e4
LT		 3058}
3059
3060static struct file_system_type ext3_fs_type = {
3061 .owner = THIS_MODULE,
3062 .name = "ext3",
152a0836 3063 .mount = ext3_mount,
1da177e4
LT		 3064 .kill_sb = kill_block_super,
3065 .fs_flags = FS_REQUIRES_DEV,
3066};
7f78e035 3067MODULE_ALIAS_FS("ext3");
1da177e4
LT		 3068
3069static int __init init_ext3_fs(void)
3070{
3071 int err = init_ext3_xattr();
3072 if (err)
3073 return err;
3074 err = init_inodecache();
3075 if (err)
3076 goto out1;
3077 err = register_filesystem(&ext3_fs_type);
3078 if (err)
3079 goto out;
3080 return 0;
3081out:
3082 destroy_inodecache();
3083out1:
e9ad5620 3084 exit_ext3_xattr();
1da177e4
LT		 3085 return err;
3086}
3087
3088static void __exit exit_ext3_fs(void)
3089{
3090 unregister_filesystem(&ext3_fs_type);
3091 destroy_inodecache();
3092 exit_ext3_xattr();
3093}
3094
3095MODULE_AUTHOR("Remy Card, Stephen Tweedie, Andrew Morton, Andreas Dilger, Theodore Ts'o and others");
3096MODULE_DESCRIPTION("Second Extended Filesystem with journaling extensions");
3097MODULE_LICENSE("GPL");
3098module_init(init_ext3_fs)
3099module_exit(exit_ext3_fs)
Linux kernel - Deliverables
This page took 1.256257 seconds and 5 git commands to generate.