]> bbs.cooldavid.org Git - net-next-2.6.git/blame - fs/ext3/super.c
git://bbs.cooldavid.org / net-next-2.6.git / blame
? search:
summary | shortlog | log | commit | commitdiff | tree
blob | blame (incremental) | history | HEAD
SLUB: simplify IRQ off handling
[net-next-2.6.git] / fs / ext3 / super.c
CommitLineData
1da177e4
LT		 1/*
2 * linux/fs/ext3/super.c
3 *
4 * Copyright (C) 1992, 1993, 1994, 1995
5 * Remy Card (card@masi.ibp.fr)
6 * Laboratoire MASI - Institut Blaise Pascal
7 * Universite Pierre et Marie Curie (Paris VI)
8 *
9 * from
10 *
11 * linux/fs/minix/inode.c
12 *
13 * Copyright (C) 1991, 1992 Linus Torvalds
14 *
15 * Big-endian to little-endian byte-swapping/bitmaps by
16 * David S. Miller (davem@caip.rutgers.edu), 1995
17 */
18
1da177e4
LT		 19#include <linux/module.h>
20#include <linux/string.h>
21#include <linux/fs.h>
22#include <linux/time.h>
23#include <linux/jbd.h>
24#include <linux/ext3_fs.h>
25#include <linux/ext3_jbd.h>
26#include <linux/slab.h>
27#include <linux/init.h>
28#include <linux/blkdev.h>
29#include <linux/parser.h>
30#include <linux/smp_lock.h>
31#include <linux/buffer_head.h>
a5694255 32#include <linux/exportfs.h>
1da177e4
LT		 33#include <linux/vfs.h>
34#include <linux/random.h>
35#include <linux/mount.h>
36#include <linux/namei.h>
37#include <linux/quotaops.h>
8fc2751b 38#include <linux/seq_file.h>
3fc74269 39#include <linux/log2.h>
381be254 40
1da177e4 41#include <asm/uaccess.h>
381be254 42
1da177e4
LT		 43#include "xattr.h"
44#include "acl.h"
381be254 45#include "namei.h"
1da177e4 46
71b96257
JL		 47static int ext3_load_journal(struct super_block *, struct ext3_super_block *,
48 unsigned long journal_devnum);
1da177e4 49static int ext3_create_journal(struct super_block *, struct ext3_super_block *,
eee194e7 50 unsigned int);
1da177e4
LT		 51static void ext3_commit_super (struct super_block * sb,
52 struct ext3_super_block * es,
53 int sync);
54static void ext3_mark_recovery_complete(struct super_block * sb,
55 struct ext3_super_block * es);
56static void ext3_clear_journal_err(struct super_block * sb,
57 struct ext3_super_block * es);
58static int ext3_sync_fs(struct super_block *sb, int wait);
59static const char *ext3_decode_error(struct super_block * sb, int errno,
60 char nbuf[16]);
61static int ext3_remount (struct super_block * sb, int * flags, char * data);
726c3342 62static int ext3_statfs (struct dentry * dentry, struct kstatfs * buf);
1da177e4
LT		 63static void ext3_unlockfs(struct super_block *sb);
64static void ext3_write_super (struct super_block * sb);
65static void ext3_write_super_lockfs(struct super_block *sb);
66
ae6ddcc5 67/*
1da177e4
LT		 68 * Wrappers for journal_start/end.
69 *
70 * The only special thing we need to do here is to make sure that all
71 * journal_end calls result in the superblock being marked dirty, so
72 * that sync() will call the filesystem's write_super callback if
ae6ddcc5 73 * appropriate.
1da177e4
LT		 74 */
75handle_t *ext3_journal_start_sb(struct super_block *sb, int nblocks)
76{
77 journal_t *journal;
78
79 if (sb->s_flags & MS_RDONLY)
80 return ERR_PTR(-EROFS);
81
82 /* Special case here: if the journal has aborted behind our
83 * backs (eg. EIO in the commit thread), then we still need to
84 * take the FS itself readonly cleanly. */
85 journal = EXT3_SB(sb)->s_journal;
86 if (is_journal_aborted(journal)) {
87 ext3_abort(sb, __FUNCTION__,
88 "Detected aborted journal");
89 return ERR_PTR(-EROFS);
90 }
91
92 return journal_start(journal, nblocks);
93}
94
ae6ddcc5 95/*
1da177e4
LT		 96 * The only special thing we need to do here is to make sure that all
97 * journal_stop calls result in the superblock being marked dirty, so
98 * that sync() will call the filesystem's write_super callback if
ae6ddcc5 99 * appropriate.
1da177e4
LT		 100 */
101int __ext3_journal_stop(const char *where, handle_t *handle)
102{
103 struct super_block *sb;
104 int err;
105 int rc;
106
107 sb = handle->h_transaction->t_journal->j_private;
108 err = handle->h_err;
109 rc = journal_stop(handle);
110
111 if (!err)
112 err = rc;
113 if (err)
114 __ext3_std_error(sb, where, err);
115 return err;
116}
117
118void ext3_journal_abort_handle(const char *caller, const char *err_fn,
119 struct buffer_head *bh, handle_t *handle, int err)
120{
121 char nbuf[16];
122 const char *errstr = ext3_decode_error(NULL, err, nbuf);
123
124 if (bh)
125 BUFFER_TRACE(bh, "abort");
126
127 if (!handle->h_err)
128 handle->h_err = err;
129
130 if (is_handle_aborted(handle))
131 return;
132
133 printk(KERN_ERR "%s: aborting transaction: %s in %s\n",
134 caller, errstr, err_fn);
135
136 journal_abort_handle(handle);
137}
138
139/* Deal with the reporting of failure conditions on a filesystem such as
140 * inconsistencies detected or read IO failures.
141 *
142 * On ext2, we can store the error state of the filesystem in the
143 * superblock. That is not possible on ext3, because we may have other
144 * write ordering constraints on the superblock which prevent us from
145 * writing it out straight away; and given that the journal is about to
146 * be aborted, we can't rely on the current, or future, transactions to
147 * write out the superblock safely.
148 *
149 * We'll just use the journal_abort() error code to record an error in
150 * the journal instead. On recovery, the journal will compain about
151 * that error until we've noted it down and cleared it.
152 */
153
154static void ext3_handle_error(struct super_block *sb)
155{
156 struct ext3_super_block *es = EXT3_SB(sb)->s_es;
157
158 EXT3_SB(sb)->s_mount_state |= EXT3_ERROR_FS;
159 es->s_state |= cpu_to_le16(EXT3_ERROR_FS);
160
161 if (sb->s_flags & MS_RDONLY)
162 return;
163
7543fc7b 164 if (!test_opt (sb, ERRORS_CONT)) {
1da177e4
LT		 165 journal_t *journal = EXT3_SB(sb)->s_journal;
166
167 EXT3_SB(sb)->s_mount_opt |= EXT3_MOUNT_ABORT;
168 if (journal)
169 journal_abort(journal, -EIO);
170 }
7543fc7b
VA		 171 if (test_opt (sb, ERRORS_RO)) {
172 printk (KERN_CRIT "Remounting filesystem read-only\n");
173 sb->s_flags |= MS_RDONLY;
174 }
175 ext3_commit_super(sb, es, 1);
1da177e4
LT		 176 if (test_opt(sb, ERRORS_PANIC))
177 panic("EXT3-fs (device %s): panic forced after error\n",
178 sb->s_id);
1da177e4
LT		 179}
180
181void ext3_error (struct super_block * sb, const char * function,
182 const char * fmt, ...)
183{
184 va_list args;
185
186 va_start(args, fmt);
187 printk(KERN_CRIT "EXT3-fs error (device %s): %s: ",sb->s_id, function);
188 vprintk(fmt, args);
189 printk("\n");
190 va_end(args);
191
192 ext3_handle_error(sb);
193}
194
195static const char *ext3_decode_error(struct super_block * sb, int errno,
196 char nbuf[16])
197{
198 char *errstr = NULL;
199
200 switch (errno) {
201 case -EIO:
202 errstr = "IO failure";
203 break;
204 case -ENOMEM:
205 errstr = "Out of memory";
206 break;
207 case -EROFS:
208 if (!sb || EXT3_SB(sb)->s_journal->j_flags & JFS_ABORT)
209 errstr = "Journal has aborted";
210 else
211 errstr = "Readonly filesystem";
212 break;
213 default:
214 /* If the caller passed in an extra buffer for unknown
215 * errors, textualise them now. Else we just return
216 * NULL. */
217 if (nbuf) {
218 /* Check for truncated error codes... */
219 if (snprintf(nbuf, 16, "error %d", -errno) >= 0)
220 errstr = nbuf;
221 }
222 break;
223 }
224
225 return errstr;
226}
227
228/* __ext3_std_error decodes expected errors from journaling functions
229 * automatically and invokes the appropriate error response. */
230
231void __ext3_std_error (struct super_block * sb, const char * function,
232 int errno)
233{
234 char nbuf[16];
30121624 235 const char *errstr;
1da177e4 236
30121624
ST		 237 /* Special case: if the error is EROFS, and we're not already
238 * inside a transaction, then there's really no point in logging
239 * an error. */
240 if (errno == -EROFS && journal_current_handle() == NULL &&
241 (sb->s_flags & MS_RDONLY))
242 return;
243
244 errstr = ext3_decode_error(sb, errno, nbuf);
1da177e4
LT		 245 printk (KERN_CRIT "EXT3-fs error (device %s) in %s: %s\n",
246 sb->s_id, function, errstr);
247
248 ext3_handle_error(sb);
249}
250
251/*
252 * ext3_abort is a much stronger failure handler than ext3_error. The
253 * abort function may be used to deal with unrecoverable failures such
254 * as journal IO errors or ENOMEM at a critical moment in log management.
255 *
256 * We unconditionally force the filesystem into an ABORT|READONLY state,
257 * unless the error response on the fs has been set to panic in which
258 * case we take the easy way out and panic immediately.
259 */
260
261void ext3_abort (struct super_block * sb, const char * function,
262 const char * fmt, ...)
263{
264 va_list args;
265
266 printk (KERN_CRIT "ext3_abort called.\n");
267
268 va_start(args, fmt);
269 printk(KERN_CRIT "EXT3-fs error (device %s): %s: ",sb->s_id, function);
270 vprintk(fmt, args);
271 printk("\n");
272 va_end(args);
273
274 if (test_opt(sb, ERRORS_PANIC))
275 panic("EXT3-fs panic from previous error\n");
276
277 if (sb->s_flags & MS_RDONLY)
278 return;
279
280 printk(KERN_CRIT "Remounting filesystem read-only\n");
281 EXT3_SB(sb)->s_mount_state |= EXT3_ERROR_FS;
282 sb->s_flags |= MS_RDONLY;
283 EXT3_SB(sb)->s_mount_opt |= EXT3_MOUNT_ABORT;
284 journal_abort(EXT3_SB(sb)->s_journal, -EIO);
285}
286
287void ext3_warning (struct super_block * sb, const char * function,
288 const char * fmt, ...)
289{
290 va_list args;
291
292 va_start(args, fmt);
293 printk(KERN_WARNING "EXT3-fs warning (device %s): %s: ",
294 sb->s_id, function);
295 vprintk(fmt, args);
296 printk("\n");
297 va_end(args);
298}
299
300void ext3_update_dynamic_rev(struct super_block *sb)
301{
302 struct ext3_super_block *es = EXT3_SB(sb)->s_es;
303
304 if (le32_to_cpu(es->s_rev_level) > EXT3_GOOD_OLD_REV)
305 return;
306
307 ext3_warning(sb, __FUNCTION__,
308 "updating to rev %d because of new feature flag, "
309 "running e2fsck is recommended",
310 EXT3_DYNAMIC_REV);
311
312 es->s_first_ino = cpu_to_le32(EXT3_GOOD_OLD_FIRST_INO);
313 es->s_inode_size = cpu_to_le16(EXT3_GOOD_OLD_INODE_SIZE);
314 es->s_rev_level = cpu_to_le32(EXT3_DYNAMIC_REV);
315 /* leave es->s_feature_*compat flags alone */
316 /* es->s_uuid will be set by e2fsck if empty */
317
318 /*
319 * The rest of the superblock fields should be zero, and if not it
320 * means they are likely already in use, so leave them alone. We
321 * can leave it up to e2fsck to clean up any inconsistencies there.
322 */
323}
324
325/*
326 * Open the external journal device
327 */
328static struct block_device *ext3_blkdev_get(dev_t dev)
329{
330 struct block_device *bdev;
331 char b[BDEVNAME_SIZE];
332
333 bdev = open_by_devnum(dev, FMODE_READ|FMODE_WRITE);
334 if (IS_ERR(bdev))
335 goto fail;
336 return bdev;
337
338fail:
339 printk(KERN_ERR "EXT3: failed to open journal device %s: %ld\n",
340 __bdevname(dev, b), PTR_ERR(bdev));
341 return NULL;
342}
343
344/*
345 * Release the journal device
346 */
347static int ext3_blkdev_put(struct block_device *bdev)
348{
349 bd_release(bdev);
350 return blkdev_put(bdev);
351}
352
353static int ext3_blkdev_remove(struct ext3_sb_info *sbi)
354{
355 struct block_device *bdev;
356 int ret = -ENODEV;
357
358 bdev = sbi->journal_bdev;
359 if (bdev) {
360 ret = ext3_blkdev_put(bdev);
361 sbi->journal_bdev = NULL;
362 }
363 return ret;
364}
365
366static inline struct inode *orphan_list_entry(struct list_head *l)
367{
368 return &list_entry(l, struct ext3_inode_info, i_orphan)->vfs_inode;
369}
370
371static void dump_orphan_list(struct super_block *sb, struct ext3_sb_info *sbi)
372{
373 struct list_head *l;
374
ae6ddcc5 375 printk(KERN_ERR "sb orphan head is %d\n",
1da177e4
LT		 376 le32_to_cpu(sbi->s_es->s_last_orphan));
377
378 printk(KERN_ERR "sb_info orphan list:\n");
379 list_for_each(l, &sbi->s_orphan) {
380 struct inode *inode = orphan_list_entry(l);
381 printk(KERN_ERR " "
eee194e7 382 "inode %s:%lu at %p: mode %o, nlink %d, next %d\n",
1da177e4 383 inode->i_sb->s_id, inode->i_ino, inode,
ae6ddcc5 384 inode->i_mode, inode->i_nlink,
1da177e4
LT		 385 NEXT_ORPHAN(inode));
386 }
387}
388
389static void ext3_put_super (struct super_block * sb)
390{
391 struct ext3_sb_info *sbi = EXT3_SB(sb);
392 struct ext3_super_block *es = sbi->s_es;
393 int i;
394
395 ext3_xattr_put_super(sb);
396 journal_destroy(sbi->s_journal);
397 if (!(sb->s_flags & MS_RDONLY)) {
398 EXT3_CLEAR_INCOMPAT_FEATURE(sb, EXT3_FEATURE_INCOMPAT_RECOVER);
399 es->s_state = cpu_to_le16(sbi->s_mount_state);
400 BUFFER_TRACE(sbi->s_sbh, "marking dirty");
401 mark_buffer_dirty(sbi->s_sbh);
402 ext3_commit_super(sb, es, 1);
403 }
404
405 for (i = 0; i < sbi->s_gdb_count; i++)
406 brelse(sbi->s_group_desc[i]);
407 kfree(sbi->s_group_desc);
408 percpu_counter_destroy(&sbi->s_freeblocks_counter);
409 percpu_counter_destroy(&sbi->s_freeinodes_counter);
410 percpu_counter_destroy(&sbi->s_dirs_counter);
411 brelse(sbi->s_sbh);
412#ifdef CONFIG_QUOTA
413 for (i = 0; i < MAXQUOTAS; i++)
414 kfree(sbi->s_qf_names[i]);
415#endif
416
417 /* Debugging code just in case the in-memory inode orphan list
418 * isn't empty. The on-disk one can be non-empty if we've
419 * detected an error and taken the fs readonly, but the
420 * in-memory list had better be clean by this point. */
421 if (!list_empty(&sbi->s_orphan))
422 dump_orphan_list(sb, sbi);
423 J_ASSERT(list_empty(&sbi->s_orphan));
424
f98393a6 425 invalidate_bdev(sb->s_bdev);
1da177e4
LT		 426 if (sbi->journal_bdev && sbi->journal_bdev != sb->s_bdev) {
427 /*
428 * Invalidate the journal device's buffers. We don't want them
429 * floating about in memory - the physical journal device may
430 * hotswapped, and it breaks the `ro-after' testing code.
431 */
432 sync_blockdev(sbi->journal_bdev);
f98393a6 433 invalidate_bdev(sbi->journal_bdev);
1da177e4
LT		 434 ext3_blkdev_remove(sbi);
435 }
436 sb->s_fs_info = NULL;
437 kfree(sbi);
438 return;
439}
440
e18b890b 441static struct kmem_cache *ext3_inode_cachep;
1da177e4
LT		 442
443/*
444 * Called inside transaction, so use GFP_NOFS
445 */
446static struct inode *ext3_alloc_inode(struct super_block *sb)
447{
448 struct ext3_inode_info *ei;
449
e6b4f8da 450 ei = kmem_cache_alloc(ext3_inode_cachep, GFP_NOFS);
1da177e4
LT		 451 if (!ei)
452 return NULL;
453#ifdef CONFIG_EXT3_FS_POSIX_ACL
454 ei->i_acl = EXT3_ACL_NOT_CACHED;
455 ei->i_default_acl = EXT3_ACL_NOT_CACHED;
456#endif
457 ei->i_block_alloc_info = NULL;
458 ei->vfs_inode.i_version = 1;
459 return &ei->vfs_inode;
460}
461
462static void ext3_destroy_inode(struct inode *inode)
463{
9f7dd93d
VA		 464 if (!list_empty(&(EXT3_I(inode)->i_orphan))) {
465 printk("EXT3 Inode %p: orphan list check failed!\n",
466 EXT3_I(inode));
467 print_hex_dump(KERN_INFO, "", DUMP_PREFIX_ADDRESS, 16, 4,
468 EXT3_I(inode), sizeof(struct ext3_inode_info),
469 false);
470 dump_stack();
471 }
1da177e4
LT		 472 kmem_cache_free(ext3_inode_cachep, EXT3_I(inode));
473}
474
e18b890b 475static void init_once(void * foo, struct kmem_cache * cachep, unsigned long flags)
1da177e4
LT		 476{
477 struct ext3_inode_info *ei = (struct ext3_inode_info *) foo;
478
a35afb83 479 INIT_LIST_HEAD(&ei->i_orphan);
1da177e4 480#ifdef CONFIG_EXT3_FS_XATTR
a35afb83 481 init_rwsem(&ei->xattr_sem);
1da177e4 482#endif
a35afb83
CL		 483 mutex_init(&ei->truncate_mutex);
484 inode_init_once(&ei->vfs_inode);
1da177e4 485}
ae6ddcc5 486
1da177e4
LT		 487static int init_inodecache(void)
488{
489 ext3_inode_cachep = kmem_cache_create("ext3_inode_cache",
490 sizeof(struct ext3_inode_info),
fffb60f9
PJ		 491 0, (SLAB_RECLAIM_ACCOUNT|
492 SLAB_MEM_SPREAD),
20c2df83 493 init_once);
1da177e4
LT		 494 if (ext3_inode_cachep == NULL)
495 return -ENOMEM;
496 return 0;
497}
498
499static void destroy_inodecache(void)
500{
1a1d92c1 501 kmem_cache_destroy(ext3_inode_cachep);
1da177e4
LT		 502}
503
504static void ext3_clear_inode(struct inode *inode)
505{
506 struct ext3_block_alloc_info *rsv = EXT3_I(inode)->i_block_alloc_info;
507#ifdef CONFIG_EXT3_FS_POSIX_ACL
e6022603
AM		 508 if (EXT3_I(inode)->i_acl &&
509 EXT3_I(inode)->i_acl != EXT3_ACL_NOT_CACHED) {
510 posix_acl_release(EXT3_I(inode)->i_acl);
511 EXT3_I(inode)->i_acl = EXT3_ACL_NOT_CACHED;
512 }
513 if (EXT3_I(inode)->i_default_acl &&
514 EXT3_I(inode)->i_default_acl != EXT3_ACL_NOT_CACHED) {
515 posix_acl_release(EXT3_I(inode)->i_default_acl);
516 EXT3_I(inode)->i_default_acl = EXT3_ACL_NOT_CACHED;
517 }
1da177e4
LT		 518#endif
519 ext3_discard_reservation(inode);
520 EXT3_I(inode)->i_block_alloc_info = NULL;
e6022603
AM		 521 if (unlikely(rsv))
522 kfree(rsv);
1da177e4
LT		 523}
524
6cd37cda 525static inline void ext3_show_quota_options(struct seq_file *seq, struct super_block *sb)
8fc2751b 526{
6cd37cda 527#if defined(CONFIG_QUOTA)
275abf5b 528 struct ext3_sb_info *sbi = EXT3_SB(sb);
8fc2751b 529
8fc2751b
MB		 530 if (sbi->s_jquota_fmt)
531 seq_printf(seq, ",jqfmt=%s",
532 (sbi->s_jquota_fmt == QFMT_VFS_OLD) ? "vfsold": "vfsv0");
533
534 if (sbi->s_qf_names[USRQUOTA])
535 seq_printf(seq, ",usrjquota=%s", sbi->s_qf_names[USRQUOTA]);
536
537 if (sbi->s_qf_names[GRPQUOTA])
538 seq_printf(seq, ",grpjquota=%s", sbi->s_qf_names[GRPQUOTA]);
539
540 if (sbi->s_mount_opt & EXT3_MOUNT_USRQUOTA)
541 seq_puts(seq, ",usrquota");
542
543 if (sbi->s_mount_opt & EXT3_MOUNT_GRPQUOTA)
544 seq_puts(seq, ",grpquota");
545#endif
6cd37cda
PO		 546}
547
548static int ext3_show_options(struct seq_file *seq, struct vfsmount *vfs)
549{
550 struct super_block *sb = vfs->mnt_sb;
551
552 if (test_opt(sb, DATA_FLAGS) == EXT3_MOUNT_JOURNAL_DATA)
553 seq_puts(seq, ",data=journal");
554 else if (test_opt(sb, DATA_FLAGS) == EXT3_MOUNT_ORDERED_DATA)
555 seq_puts(seq, ",data=ordered");
556 else if (test_opt(sb, DATA_FLAGS) == EXT3_MOUNT_WRITEBACK_DATA)
557 seq_puts(seq, ",data=writeback");
558
559 ext3_show_quota_options(seq, sb);
8fc2751b
MB		 560
561 return 0;
562}
1da177e4 563
fdb36673
N		 564
565static struct dentry *ext3_get_dentry(struct super_block *sb, void *vobjp)
566{
567 __u32 *objp = vobjp;
568 unsigned long ino = objp[0];
569 __u32 generation = objp[1];
570 struct inode *inode;
571 struct dentry *result;
572
573 if (ino < EXT3_FIRST_INO(sb) && ino != EXT3_ROOT_INO)
574 return ERR_PTR(-ESTALE);
575 if (ino > le32_to_cpu(EXT3_SB(sb)->s_es->s_inodes_count))
576 return ERR_PTR(-ESTALE);
577
578 /* iget isn't really right if the inode is currently unallocated!!
579 *
580 * ext3_read_inode will return a bad_inode if the inode had been
581 * deleted, so we should be safe.
582 *
583 * Currently we don't know the generation for parent directory, so
584 * a generation of 0 means "accept any"
585 */
586 inode = iget(sb, ino);
587 if (inode == NULL)
588 return ERR_PTR(-ENOMEM);
589 if (is_bad_inode(inode) ||
590 (generation && inode->i_generation != generation)) {
591 iput(inode);
592 return ERR_PTR(-ESTALE);
593 }
594 /* now to find a dentry.
595 * If possible, get a well-connected one
596 */
597 result = d_alloc_anon(inode);
598 if (!result) {
599 iput(inode);
600 return ERR_PTR(-ENOMEM);
601 }
602 return result;
603}
604
8fc2751b 605#ifdef CONFIG_QUOTA
1da177e4
LT		 606#define QTYPE2NAME(t) ((t)==USRQUOTA?"user":"group")
607#define QTYPE2MOPT(on, t) ((t)==USRQUOTA?((on)##USRJQUOTA):((on)##GRPJQUOTA))
608
609static int ext3_dquot_initialize(struct inode *inode, int type);
610static int ext3_dquot_drop(struct inode *inode);
611static int ext3_write_dquot(struct dquot *dquot);
612static int ext3_acquire_dquot(struct dquot *dquot);
613static int ext3_release_dquot(struct dquot *dquot);
614static int ext3_mark_dquot_dirty(struct dquot *dquot);
615static int ext3_write_info(struct super_block *sb, int type);
616static int ext3_quota_on(struct super_block *sb, int type, int format_id, char *path);
617static int ext3_quota_on_mount(struct super_block *sb, int type);
618static ssize_t ext3_quota_read(struct super_block *sb, int type, char *data,
619 size_t len, loff_t off);
620static ssize_t ext3_quota_write(struct super_block *sb, int type,
621 const char *data, size_t len, loff_t off);
622
623static struct dquot_operations ext3_quota_operations = {
624 .initialize = ext3_dquot_initialize,
625 .drop = ext3_dquot_drop,
626 .alloc_space = dquot_alloc_space,
627 .alloc_inode = dquot_alloc_inode,
628 .free_space = dquot_free_space,
629 .free_inode = dquot_free_inode,
630 .transfer = dquot_transfer,
631 .write_dquot = ext3_write_dquot,
632 .acquire_dquot = ext3_acquire_dquot,
633 .release_dquot = ext3_release_dquot,
634 .mark_dirty = ext3_mark_dquot_dirty,
635 .write_info = ext3_write_info
636};
637
638static struct quotactl_ops ext3_qctl_operations = {
639 .quota_on = ext3_quota_on,
640 .quota_off = vfs_quota_off,
641 .quota_sync = vfs_quota_sync,
642 .get_info = vfs_get_dqinfo,
643 .set_info = vfs_set_dqinfo,
644 .get_dqblk = vfs_get_dqblk,
645 .set_dqblk = vfs_set_dqblk
646};
647#endif
648
ee9b6d61 649static const struct super_operations ext3_sops = {
1da177e4
LT		 650 .alloc_inode = ext3_alloc_inode,
651 .destroy_inode = ext3_destroy_inode,
652 .read_inode = ext3_read_inode,
653 .write_inode = ext3_write_inode,
654 .dirty_inode = ext3_dirty_inode,
655 .delete_inode = ext3_delete_inode,
656 .put_super = ext3_put_super,
657 .write_super = ext3_write_super,
658 .sync_fs = ext3_sync_fs,
659 .write_super_lockfs = ext3_write_super_lockfs,
660 .unlockfs = ext3_unlockfs,
661 .statfs = ext3_statfs,
662 .remount_fs = ext3_remount,
663 .clear_inode = ext3_clear_inode,
8fc2751b 664 .show_options = ext3_show_options,
1da177e4
LT		 665#ifdef CONFIG_QUOTA
666 .quota_read = ext3_quota_read,
667 .quota_write = ext3_quota_write,
668#endif
669};
670
1da177e4
LT		 671static struct export_operations ext3_export_ops = {
672 .get_parent = ext3_get_parent,
fdb36673 673 .get_dentry = ext3_get_dentry,
1da177e4
LT		 674};
675
676enum {
677 Opt_bsd_df, Opt_minix_df, Opt_grpid, Opt_nogrpid,
678 Opt_resgid, Opt_resuid, Opt_sb, Opt_err_cont, Opt_err_panic, Opt_err_ro,
2860b733 679 Opt_nouid32, Opt_nocheck, Opt_debug, Opt_oldalloc, Opt_orlov,
1da177e4 680 Opt_user_xattr, Opt_nouser_xattr, Opt_acl, Opt_noacl,
ade1a29e 681 Opt_reservation, Opt_noreservation, Opt_noload, Opt_nobh, Opt_bh,
71b96257 682 Opt_commit, Opt_journal_update, Opt_journal_inum, Opt_journal_dev,
1da177e4
LT		 683 Opt_abort, Opt_data_journal, Opt_data_ordered, Opt_data_writeback,
684 Opt_usrjquota, Opt_grpjquota, Opt_offusrjquota, Opt_offgrpjquota,
1f54587b 685 Opt_jqfmt_vfsold, Opt_jqfmt_vfsv0, Opt_quota, Opt_noquota,
8fc2751b
MB		 686 Opt_ignore, Opt_barrier, Opt_err, Opt_resize, Opt_usrquota,
687 Opt_grpquota
1da177e4
LT		 688};
689
690static match_table_t tokens = {
691 {Opt_bsd_df, "bsddf"},
692 {Opt_minix_df, "minixdf"},
693 {Opt_grpid, "grpid"},
694 {Opt_grpid, "bsdgroups"},
695 {Opt_nogrpid, "nogrpid"},
696 {Opt_nogrpid, "sysvgroups"},
697 {Opt_resgid, "resgid=%u"},
698 {Opt_resuid, "resuid=%u"},
699 {Opt_sb, "sb=%u"},
700 {Opt_err_cont, "errors=continue"},
701 {Opt_err_panic, "errors=panic"},
702 {Opt_err_ro, "errors=remount-ro"},
703 {Opt_nouid32, "nouid32"},
704 {Opt_nocheck, "nocheck"},
705 {Opt_nocheck, "check=none"},
1da177e4
LT		 706 {Opt_debug, "debug"},
707 {Opt_oldalloc, "oldalloc"},
708 {Opt_orlov, "orlov"},
709 {Opt_user_xattr, "user_xattr"},
710 {Opt_nouser_xattr, "nouser_xattr"},
711 {Opt_acl, "acl"},
712 {Opt_noacl, "noacl"},
713 {Opt_reservation, "reservation"},
714 {Opt_noreservation, "noreservation"},
715 {Opt_noload, "noload"},
716 {Opt_nobh, "nobh"},
ade1a29e 717 {Opt_bh, "bh"},
1da177e4
LT		 718 {Opt_commit, "commit=%u"},
719 {Opt_journal_update, "journal=update"},
720 {Opt_journal_inum, "journal=%u"},
71b96257 721 {Opt_journal_dev, "journal_dev=%u"},
1da177e4
LT		 722 {Opt_abort, "abort"},
723 {Opt_data_journal, "data=journal"},
724 {Opt_data_ordered, "data=ordered"},
725 {Opt_data_writeback, "data=writeback"},
726 {Opt_offusrjquota, "usrjquota="},
727 {Opt_usrjquota, "usrjquota=%s"},
728 {Opt_offgrpjquota, "grpjquota="},
729 {Opt_grpjquota, "grpjquota=%s"},
730 {Opt_jqfmt_vfsold, "jqfmt=vfsold"},
731 {Opt_jqfmt_vfsv0, "jqfmt=vfsv0"},
8fc2751b 732 {Opt_grpquota, "grpquota"},
1f54587b
JK		 733 {Opt_noquota, "noquota"},
734 {Opt_quota, "quota"},
8fc2751b 735 {Opt_usrquota, "usrquota"},
1da177e4
LT		 736 {Opt_barrier, "barrier=%u"},
737 {Opt_err, NULL},
738 {Opt_resize, "resize"},
739};
740
43d23f90 741static ext3_fsblk_t get_sb_block(void **data)
1da177e4 742{
e9ad5620
DK		 743 ext3_fsblk_t sb_block;
744 char *options = (char *) *data;
1da177e4
LT		 745
746 if (!options || strncmp(options, "sb=", 3) != 0)
747 return 1; /* Default location */
748 options += 3;
43d23f90 749 /*todo: use simple_strtoll with >32bit ext3 */
1da177e4
LT		 750 sb_block = simple_strtoul(options, &options, 0);
751 if (*options && *options != ',') {
752 printk("EXT3-fs: Invalid sb specification: %s\n",
753 (char *) *data);
754 return 1;
755 }
756 if (*options == ',')
757 options++;
758 *data = (void *) options;
759 return sb_block;
760}
761
71b96257 762static int parse_options (char *options, struct super_block *sb,
eee194e7 763 unsigned int *inum, unsigned long *journal_devnum,
43d23f90 764 ext3_fsblk_t *n_blocks_count, int is_remount)
1da177e4
LT		 765{
766 struct ext3_sb_info *sbi = EXT3_SB(sb);
767 char * p;
768 substring_t args[MAX_OPT_ARGS];
769 int data_opt = 0;
770 int option;
771#ifdef CONFIG_QUOTA
772 int qtype;
773 char *qname;
774#endif
775
776 if (!options)
777 return 1;
778
779 while ((p = strsep (&options, ",")) != NULL) {
780 int token;
781 if (!*p)
782 continue;
783
784 token = match_token(p, tokens, args);
785 switch (token) {
786 case Opt_bsd_df:
787 clear_opt (sbi->s_mount_opt, MINIX_DF);
788 break;
789 case Opt_minix_df:
790 set_opt (sbi->s_mount_opt, MINIX_DF);
791 break;
792 case Opt_grpid:
793 set_opt (sbi->s_mount_opt, GRPID);
794 break;
795 case Opt_nogrpid:
796 clear_opt (sbi->s_mount_opt, GRPID);
797 break;
798 case Opt_resuid:
799 if (match_int(&args[0], &option))
800 return 0;
801 sbi->s_resuid = option;
802 break;
803 case Opt_resgid:
804 if (match_int(&args[0], &option))
805 return 0;
806 sbi->s_resgid = option;
807 break;
808 case Opt_sb:
809 /* handled by get_sb_block() instead of here */
810 /* *sb_block = match_int(&args[0]); */
811 break;
812 case Opt_err_panic:
813 clear_opt (sbi->s_mount_opt, ERRORS_CONT);
814 clear_opt (sbi->s_mount_opt, ERRORS_RO);
815 set_opt (sbi->s_mount_opt, ERRORS_PANIC);
816 break;
817 case Opt_err_ro:
818 clear_opt (sbi->s_mount_opt, ERRORS_CONT);
819 clear_opt (sbi->s_mount_opt, ERRORS_PANIC);
820 set_opt (sbi->s_mount_opt, ERRORS_RO);
821 break;
822 case Opt_err_cont:
823 clear_opt (sbi->s_mount_opt, ERRORS_RO);
824 clear_opt (sbi->s_mount_opt, ERRORS_PANIC);
825 set_opt (sbi->s_mount_opt, ERRORS_CONT);
826 break;
827 case Opt_nouid32:
828 set_opt (sbi->s_mount_opt, NO_UID32);
829 break;
1da177e4
LT		 830 case Opt_nocheck:
831 clear_opt (sbi->s_mount_opt, CHECK);
832 break;
833 case Opt_debug:
834 set_opt (sbi->s_mount_opt, DEBUG);
835 break;
836 case Opt_oldalloc:
837 set_opt (sbi->s_mount_opt, OLDALLOC);
838 break;
839 case Opt_orlov:
840 clear_opt (sbi->s_mount_opt, OLDALLOC);
841 break;
842#ifdef CONFIG_EXT3_FS_XATTR
843 case Opt_user_xattr:
844 set_opt (sbi->s_mount_opt, XATTR_USER);
845 break;
846 case Opt_nouser_xattr:
847 clear_opt (sbi->s_mount_opt, XATTR_USER);
848 break;
849#else
850 case Opt_user_xattr:
851 case Opt_nouser_xattr:
852 printk("EXT3 (no)user_xattr options not supported\n");
853 break;
854#endif
855#ifdef CONFIG_EXT3_FS_POSIX_ACL
856 case Opt_acl:
857 set_opt(sbi->s_mount_opt, POSIX_ACL);
858 break;
859 case Opt_noacl:
860 clear_opt(sbi->s_mount_opt, POSIX_ACL);
861 break;
862#else
863 case Opt_acl:
864 case Opt_noacl:
865 printk("EXT3 (no)acl options not supported\n");
866 break;
867#endif
868 case Opt_reservation:
869 set_opt(sbi->s_mount_opt, RESERVATION);
870 break;
871 case Opt_noreservation:
872 clear_opt(sbi->s_mount_opt, RESERVATION);
873 break;
874 case Opt_journal_update:
875 /* @@@ FIXME */
876 /* Eventually we will want to be able to create
877 a journal file here. For now, only allow the
878 user to specify an existing inode to be the
879 journal file. */
880 if (is_remount) {
881 printk(KERN_ERR "EXT3-fs: cannot specify "
882 "journal on remount\n");
883 return 0;
884 }
885 set_opt (sbi->s_mount_opt, UPDATE_JOURNAL);
886 break;
887 case Opt_journal_inum:
888 if (is_remount) {
889 printk(KERN_ERR "EXT3-fs: cannot specify "
890 "journal on remount\n");
891 return 0;
892 }
893 if (match_int(&args[0], &option))
894 return 0;
895 *inum = option;
896 break;
71b96257
JL		 897 case Opt_journal_dev:
898 if (is_remount) {
899 printk(KERN_ERR "EXT3-fs: cannot specify "
900 "journal on remount\n");
901 return 0;
902 }
903 if (match_int(&args[0], &option))
904 return 0;
905 *journal_devnum = option;
906 break;
1da177e4
LT		 907 case Opt_noload:
908 set_opt (sbi->s_mount_opt, NOLOAD);
909 break;
910 case Opt_commit:
911 if (match_int(&args[0], &option))
912 return 0;
913 if (option < 0)
914 return 0;
915 if (option == 0)
916 option = JBD_DEFAULT_MAX_COMMIT_AGE;
917 sbi->s_commit_interval = HZ * option;
918 break;
919 case Opt_data_journal:
920 data_opt = EXT3_MOUNT_JOURNAL_DATA;
921 goto datacheck;
922 case Opt_data_ordered:
923 data_opt = EXT3_MOUNT_ORDERED_DATA;
924 goto datacheck;
925 case Opt_data_writeback:
926 data_opt = EXT3_MOUNT_WRITEBACK_DATA;
927 datacheck:
928 if (is_remount) {
929 if ((sbi->s_mount_opt & EXT3_MOUNT_DATA_FLAGS)
930 != data_opt) {
931 printk(KERN_ERR
932 "EXT3-fs: cannot change data "
933 "mode on remount\n");
934 return 0;
935 }
936 } else {
937 sbi->s_mount_opt &= ~EXT3_MOUNT_DATA_FLAGS;
938 sbi->s_mount_opt |= data_opt;
939 }
940 break;
941#ifdef CONFIG_QUOTA
942 case Opt_usrjquota:
943 qtype = USRQUOTA;
944 goto set_qf_name;
945 case Opt_grpjquota:
946 qtype = GRPQUOTA;
947set_qf_name:
948 if (sb_any_quota_enabled(sb)) {
949 printk(KERN_ERR
950 "EXT3-fs: Cannot change journalled "
951 "quota options when quota turned on.\n");
952 return 0;
953 }
954 qname = match_strdup(&args[0]);
955 if (!qname) {
956 printk(KERN_ERR
957 "EXT3-fs: not enough memory for "
958 "storing quotafile name.\n");
959 return 0;
960 }
961 if (sbi->s_qf_names[qtype] &&
962 strcmp(sbi->s_qf_names[qtype], qname)) {
963 printk(KERN_ERR
964 "EXT3-fs: %s quota file already "
965 "specified.\n", QTYPE2NAME(qtype));
966 kfree(qname);
967 return 0;
968 }
969 sbi->s_qf_names[qtype] = qname;
970 if (strchr(sbi->s_qf_names[qtype], '/')) {
971 printk(KERN_ERR
972 "EXT3-fs: quotafile must be on "
973 "filesystem root.\n");
974 kfree(sbi->s_qf_names[qtype]);
975 sbi->s_qf_names[qtype] = NULL;
976 return 0;
977 }
1f54587b 978 set_opt(sbi->s_mount_opt, QUOTA);
1da177e4
LT		 979 break;
980 case Opt_offusrjquota:
981 qtype = USRQUOTA;
982 goto clear_qf_name;
983 case Opt_offgrpjquota:
984 qtype = GRPQUOTA;
985clear_qf_name:
986 if (sb_any_quota_enabled(sb)) {
987 printk(KERN_ERR "EXT3-fs: Cannot change "
988 "journalled quota options when "
989 "quota turned on.\n");
990 return 0;
991 }
08c6a96f
JK		 992 /*
993 * The space will be released later when all options
994 * are confirmed to be correct
995 */
1da177e4
LT		 996 sbi->s_qf_names[qtype] = NULL;
997 break;
998 case Opt_jqfmt_vfsold:
999 sbi->s_jquota_fmt = QFMT_VFS_OLD;
1000 break;
1001 case Opt_jqfmt_vfsv0:
1002 sbi->s_jquota_fmt = QFMT_VFS_V0;
1003 break;
1f54587b 1004 case Opt_quota:
8fc2751b
MB		 1005 case Opt_usrquota:
1006 set_opt(sbi->s_mount_opt, QUOTA);
1007 set_opt(sbi->s_mount_opt, USRQUOTA);
1008 break;
1009 case Opt_grpquota:
1f54587b 1010 set_opt(sbi->s_mount_opt, QUOTA);
8fc2751b 1011 set_opt(sbi->s_mount_opt, GRPQUOTA);
1f54587b
JK		 1012 break;
1013 case Opt_noquota:
1014 if (sb_any_quota_enabled(sb)) {
1015 printk(KERN_ERR "EXT3-fs: Cannot change quota "
1016 "options when quota turned on.\n");
1017 return 0;
1018 }
1019 clear_opt(sbi->s_mount_opt, QUOTA);
8fc2751b
MB		 1020 clear_opt(sbi->s_mount_opt, USRQUOTA);
1021 clear_opt(sbi->s_mount_opt, GRPQUOTA);
1f54587b 1022 break;
1da177e4 1023#else
8fc2751b
MB		 1024 case Opt_quota:
1025 case Opt_usrquota:
1026 case Opt_grpquota:
1da177e4
LT		 1027 case Opt_usrjquota:
1028 case Opt_grpjquota:
1029 case Opt_offusrjquota:
1030 case Opt_offgrpjquota:
1031 case Opt_jqfmt_vfsold:
1032 case Opt_jqfmt_vfsv0:
1033 printk(KERN_ERR
1034 "EXT3-fs: journalled quota options not "
1035 "supported.\n");
1036 break;
1f54587b
JK		 1037 case Opt_noquota:
1038 break;
1da177e4
LT		 1039#endif
1040 case Opt_abort:
1041 set_opt(sbi->s_mount_opt, ABORT);
1042 break;
1043 case Opt_barrier:
1044 if (match_int(&args[0], &option))
1045 return 0;
1046 if (option)
1047 set_opt(sbi->s_mount_opt, BARRIER);
1048 else
1049 clear_opt(sbi->s_mount_opt, BARRIER);
1050 break;
1051 case Opt_ignore:
1052 break;
1053 case Opt_resize:
08c6a96f 1054 if (!is_remount) {
1da177e4
LT		 1055 printk("EXT3-fs: resize option only available "
1056 "for remount\n");
1057 return 0;
1058 }
c7f1721e
KZ		 1059 if (match_int(&args[0], &option) != 0)
1060 return 0;
1da177e4
LT		 1061 *n_blocks_count = option;
1062 break;
1063 case Opt_nobh:
1064 set_opt(sbi->s_mount_opt, NOBH);
1065 break;
ade1a29e
BP		 1066 case Opt_bh:
1067 clear_opt(sbi->s_mount_opt, NOBH);
1068 break;
1da177e4
LT		 1069 default:
1070 printk (KERN_ERR
1071 "EXT3-fs: Unrecognized mount option \"%s\" "
1072 "or missing value\n", p);
1073 return 0;
1074 }
1075 }
1076#ifdef CONFIG_QUOTA
8fc2751b
MB		 1077 if (sbi->s_qf_names[USRQUOTA] || sbi->s_qf_names[GRPQUOTA]) {
1078 if ((sbi->s_mount_opt & EXT3_MOUNT_USRQUOTA) &&
1079 sbi->s_qf_names[USRQUOTA])
1080 clear_opt(sbi->s_mount_opt, USRQUOTA);
1081
1082 if ((sbi->s_mount_opt & EXT3_MOUNT_GRPQUOTA) &&
1083 sbi->s_qf_names[GRPQUOTA])
1084 clear_opt(sbi->s_mount_opt, GRPQUOTA);
1085
1086 if ((sbi->s_qf_names[USRQUOTA] &&
1087 (sbi->s_mount_opt & EXT3_MOUNT_GRPQUOTA)) ||
1088 (sbi->s_qf_names[GRPQUOTA] &&
1089 (sbi->s_mount_opt & EXT3_MOUNT_USRQUOTA))) {
1090 printk(KERN_ERR "EXT3-fs: old and new quota "
1091 "format mixing.\n");
1092 return 0;
1093 }
1094
1095 if (!sbi->s_jquota_fmt) {
1096 printk(KERN_ERR "EXT3-fs: journalled quota format "
1097 "not specified.\n");
1098 return 0;
1099 }
1100 } else {
1101 if (sbi->s_jquota_fmt) {
1102 printk(KERN_ERR "EXT3-fs: journalled quota format "
1103 "specified with no journalling "
1104 "enabled.\n");
1105 return 0;
1106 }
1da177e4
LT		 1107 }
1108#endif
1da177e4
LT		 1109 return 1;
1110}
1111
1112static int ext3_setup_super(struct super_block *sb, struct ext3_super_block *es,
1113 int read_only)
1114{
1115 struct ext3_sb_info *sbi = EXT3_SB(sb);
1116 int res = 0;
1117
1118 if (le32_to_cpu(es->s_rev_level) > EXT3_MAX_SUPP_REV) {
1119 printk (KERN_ERR "EXT3-fs warning: revision level too high, "
1120 "forcing read-only mode\n");
1121 res = MS_RDONLY;
1122 }
1123 if (read_only)
1124 return res;
1125 if (!(sbi->s_mount_state & EXT3_VALID_FS))
1126 printk (KERN_WARNING "EXT3-fs warning: mounting unchecked fs, "
1127 "running e2fsck is recommended\n");
1128 else if ((sbi->s_mount_state & EXT3_ERROR_FS))
1129 printk (KERN_WARNING
1130 "EXT3-fs warning: mounting fs with errors, "
1131 "running e2fsck is recommended\n");
1132 else if ((__s16) le16_to_cpu(es->s_max_mnt_count) >= 0 &&
1133 le16_to_cpu(es->s_mnt_count) >=
1134 (unsigned short) (__s16) le16_to_cpu(es->s_max_mnt_count))
1135 printk (KERN_WARNING
1136 "EXT3-fs warning: maximal mount count reached, "
1137 "running e2fsck is recommended\n");
1138 else if (le32_to_cpu(es->s_checkinterval) &&
1139 (le32_to_cpu(es->s_lastcheck) +
1140 le32_to_cpu(es->s_checkinterval) <= get_seconds()))
1141 printk (KERN_WARNING
1142 "EXT3-fs warning: checktime reached, "
1143 "running e2fsck is recommended\n");
1144#if 0
1145 /* @@@ We _will_ want to clear the valid bit if we find
1146 inconsistencies, to force a fsck at reboot. But for
1147 a plain journaled filesystem we can keep it set as
1148 valid forever! :) */
1149 es->s_state = cpu_to_le16(le16_to_cpu(es->s_state) & ~EXT3_VALID_FS);
1150#endif
1151 if (!(__s16) le16_to_cpu(es->s_max_mnt_count))
1152 es->s_max_mnt_count = cpu_to_le16(EXT3_DFL_MAX_MNT_COUNT);
1153 es->s_mnt_count=cpu_to_le16(le16_to_cpu(es->s_mnt_count) + 1);
1154 es->s_mtime = cpu_to_le32(get_seconds());
1155 ext3_update_dynamic_rev(sb);
1156 EXT3_SET_INCOMPAT_FEATURE(sb, EXT3_FEATURE_INCOMPAT_RECOVER);
1157
1158 ext3_commit_super(sb, es, 1);
1159 if (test_opt(sb, DEBUG))
1160 printk(KERN_INFO "[EXT3 FS bs=%lu, gc=%lu, "
1161 "bpg=%lu, ipg=%lu, mo=%04lx]\n",
1162 sb->s_blocksize,
1163 sbi->s_groups_count,
1164 EXT3_BLOCKS_PER_GROUP(sb),
1165 EXT3_INODES_PER_GROUP(sb),
1166 sbi->s_mount_opt);
1167
1168 printk(KERN_INFO "EXT3 FS on %s, ", sb->s_id);
1169 if (EXT3_SB(sb)->s_journal->j_inode == NULL) {
1170 char b[BDEVNAME_SIZE];
1171
1172 printk("external journal on %s\n",
1173 bdevname(EXT3_SB(sb)->s_journal->j_dev, b));
1174 } else {
1175 printk("internal journal\n");
1176 }
1da177e4
LT		 1177 return res;
1178}
1179
1180/* Called at mount-time, super-block is locked */
1181static int ext3_check_descriptors (struct super_block * sb)
1182{
1183 struct ext3_sb_info *sbi = EXT3_SB(sb);
855565e8
ES		 1184 ext3_fsblk_t first_block = le32_to_cpu(sbi->s_es->s_first_data_block);
1185 ext3_fsblk_t last_block;
1da177e4
LT		 1186 struct ext3_group_desc * gdp = NULL;
1187 int desc_block = 0;
1188 int i;
1189
1190 ext3_debug ("Checking group descriptors");
1191
1192 for (i = 0; i < sbi->s_groups_count; i++)
1193 {
855565e8
ES		 1194 if (i == sbi->s_groups_count - 1)
1195 last_block = le32_to_cpu(sbi->s_es->s_blocks_count) - 1;
1196 else
1197 last_block = first_block +
1198 (EXT3_BLOCKS_PER_GROUP(sb) - 1);
1199
1da177e4
LT		 1200 if ((i % EXT3_DESC_PER_BLOCK(sb)) == 0)
1201 gdp = (struct ext3_group_desc *)
1202 sbi->s_group_desc[desc_block++]->b_data;
855565e8
ES		 1203 if (le32_to_cpu(gdp->bg_block_bitmap) < first_block ||
1204 le32_to_cpu(gdp->bg_block_bitmap) > last_block)
1da177e4
LT		 1205 {
1206 ext3_error (sb, "ext3_check_descriptors",
1207 "Block bitmap for group %d"
1208 " not in group (block %lu)!",
1209 i, (unsigned long)
1210 le32_to_cpu(gdp->bg_block_bitmap));
1211 return 0;
1212 }
855565e8
ES		 1213 if (le32_to_cpu(gdp->bg_inode_bitmap) < first_block ||
1214 le32_to_cpu(gdp->bg_inode_bitmap) > last_block)
1da177e4
LT		 1215 {
1216 ext3_error (sb, "ext3_check_descriptors",
1217 "Inode bitmap for group %d"
1218 " not in group (block %lu)!",
1219 i, (unsigned long)
1220 le32_to_cpu(gdp->bg_inode_bitmap));
1221 return 0;
1222 }
855565e8 1223 if (le32_to_cpu(gdp->bg_inode_table) < first_block ||
780dcdb2 1224 le32_to_cpu(gdp->bg_inode_table) + sbi->s_itb_per_group - 1 >
855565e8 1225 last_block)
1da177e4
LT		 1226 {
1227 ext3_error (sb, "ext3_check_descriptors",
1228 "Inode table for group %d"
1229 " not in group (block %lu)!",
1230 i, (unsigned long)
1231 le32_to_cpu(gdp->bg_inode_table));
1232 return 0;
1233 }
855565e8 1234 first_block += EXT3_BLOCKS_PER_GROUP(sb);
1da177e4
LT		 1235 gdp++;
1236 }
1237
1238 sbi->s_es->s_free_blocks_count=cpu_to_le32(ext3_count_free_blocks(sb));
1239 sbi->s_es->s_free_inodes_count=cpu_to_le32(ext3_count_free_inodes(sb));
1240 return 1;
1241}
1242
1243
1244/* ext3_orphan_cleanup() walks a singly-linked list of inodes (starting at
1245 * the superblock) which were deleted from all directories, but held open by
1246 * a process at the time of a crash. We walk the list and try to delete these
1247 * inodes at recovery time (only with a read-write filesystem).
1248 *
1249 * In order to keep the orphan inode chain consistent during traversal (in
1250 * case of crash during recovery), we link each inode into the superblock
1251 * orphan list_head and handle it the same way as an inode deletion during
1252 * normal operation (which journals the operations for us).
1253 *
1254 * We only do an iget() and an iput() on each inode, which is very safe if we
1255 * accidentally point at an in-use or already deleted inode. The worst that
1256 * can happen in this case is that we get a "bit already cleared" message from
1257 * ext3_free_inode(). The only reason we would point at a wrong inode is if
1258 * e2fsck was run on this filesystem, and it must have already done the orphan
1259 * inode cleanup for us, so we can safely abort without any further action.
1260 */
1261static void ext3_orphan_cleanup (struct super_block * sb,
1262 struct ext3_super_block * es)
1263{
1264 unsigned int s_flags = sb->s_flags;
1265 int nr_orphans = 0, nr_truncates = 0;
1266#ifdef CONFIG_QUOTA
1267 int i;
1268#endif
1269 if (!es->s_last_orphan) {
1270 jbd_debug(4, "no orphan inodes to clean up\n");
1271 return;
1272 }
1273
a8f48a95
ES		 1274 if (bdev_read_only(sb->s_bdev)) {
1275 printk(KERN_ERR "EXT3-fs: write access "
1276 "unavailable, skipping orphan cleanup.\n");
1277 return;
1278 }
1279
1da177e4
LT		 1280 if (EXT3_SB(sb)->s_mount_state & EXT3_ERROR_FS) {
1281 if (es->s_last_orphan)
1282 jbd_debug(1, "Errors on filesystem, "
1283 "clearing orphan list.\n");
1284 es->s_last_orphan = 0;
1285 jbd_debug(1, "Skipping orphan recovery on fs with errors.\n");
1286 return;
1287 }
1288
1289 if (s_flags & MS_RDONLY) {
1290 printk(KERN_INFO "EXT3-fs: %s: orphan cleanup on readonly fs\n",
1291 sb->s_id);
1292 sb->s_flags &= ~MS_RDONLY;
1293 }
1294#ifdef CONFIG_QUOTA
1295 /* Needed for iput() to work correctly and not trash data */
1296 sb->s_flags |= MS_ACTIVE;
1297 /* Turn on quotas so that they are updated correctly */
1298 for (i = 0; i < MAXQUOTAS; i++) {
1299 if (EXT3_SB(sb)->s_qf_names[i]) {
1300 int ret = ext3_quota_on_mount(sb, i);
1301 if (ret < 0)
1302 printk(KERN_ERR
1303 "EXT3-fs: Cannot turn on journalled "
1304 "quota: error %d\n", ret);
1305 }
1306 }
1307#endif
1308
1309 while (es->s_last_orphan) {
1310 struct inode *inode;
1311
1312 if (!(inode =
1313 ext3_orphan_get(sb, le32_to_cpu(es->s_last_orphan)))) {
1314 es->s_last_orphan = 0;
1315 break;
1316 }
1317
1318 list_add(&EXT3_I(inode)->i_orphan, &EXT3_SB(sb)->s_orphan);
1319 DQUOT_INIT(inode);
1320 if (inode->i_nlink) {
1321 printk(KERN_DEBUG
eee194e7 1322 "%s: truncating inode %lu to %Ld bytes\n",
1da177e4 1323 __FUNCTION__, inode->i_ino, inode->i_size);
eee194e7 1324 jbd_debug(2, "truncating inode %lu to %Ld bytes\n",
1da177e4
LT		 1325 inode->i_ino, inode->i_size);
1326 ext3_truncate(inode);
1327 nr_truncates++;
1328 } else {
1329 printk(KERN_DEBUG
eee194e7 1330 "%s: deleting unreferenced inode %lu\n",
1da177e4 1331 __FUNCTION__, inode->i_ino);
eee194e7 1332 jbd_debug(2, "deleting unreferenced inode %lu\n",
1da177e4
LT		 1333 inode->i_ino);
1334 nr_orphans++;
1335 }
1336 iput(inode); /* The delete magic happens here! */
1337 }
1338
1339#define PLURAL(x) (x), ((x)==1) ? "" : "s"
1340
1341 if (nr_orphans)
1342 printk(KERN_INFO "EXT3-fs: %s: %d orphan inode%s deleted\n",
1343 sb->s_id, PLURAL(nr_orphans));
1344 if (nr_truncates)
1345 printk(KERN_INFO "EXT3-fs: %s: %d truncate%s cleaned up\n",
1346 sb->s_id, PLURAL(nr_truncates));
1347#ifdef CONFIG_QUOTA
1348 /* Turn quotas off */
1349 for (i = 0; i < MAXQUOTAS; i++) {
1350 if (sb_dqopt(sb)->files[i])
1351 vfs_quota_off(sb, i);
1352 }
1353#endif
1354 sb->s_flags = s_flags; /* Restore MS_RDONLY status */
1355}
1356
1da177e4
LT		 1357/*
1358 * Maximal file size. There is a direct, and {,double-,triple-}indirect
1359 * block limit, and also a limit of (2^32 - 1) 512-byte sectors in i_blocks.
1360 * We need to be 1 filesystem block less than the 2^32 sector limit.
1361 */
1362static loff_t ext3_max_size(int bits)
1363{
1364 loff_t res = EXT3_NDIR_BLOCKS;
1365 /* This constant is calculated to be the largest file size for a
1366 * dense, 4k-blocksize file such that the total number of
1367 * sectors in the file, including data and all indirect blocks,
1368 * does not exceed 2^32. */
1369 const loff_t upper_limit = 0x1ff7fffd000LL;
1370
1371 res += 1LL << (bits-2);
1372 res += 1LL << (2*(bits-2));
1373 res += 1LL << (3*(bits-2));
1374 res <<= bits;
1375 if (res > upper_limit)
1376 res = upper_limit;
1377 return res;
1378}
1379
43d23f90
MC		 1380static ext3_fsblk_t descriptor_loc(struct super_block *sb,
1381 ext3_fsblk_t logic_sb_block,
1da177e4
LT		 1382 int nr)
1383{
1384 struct ext3_sb_info *sbi = EXT3_SB(sb);
43d23f90 1385 unsigned long bg, first_meta_bg;
1da177e4
LT		 1386 int has_super = 0;
1387
1da177e4
LT		 1388 first_meta_bg = le32_to_cpu(sbi->s_es->s_first_meta_bg);
1389
1390 if (!EXT3_HAS_INCOMPAT_FEATURE(sb, EXT3_FEATURE_INCOMPAT_META_BG) ||
1391 nr < first_meta_bg)
1392 return (logic_sb_block + nr + 1);
1393 bg = sbi->s_desc_per_block * nr;
1394 if (ext3_bg_has_super(sb, bg))
1395 has_super = 1;
43d23f90 1396 return (has_super + ext3_group_first_block_no(sb, bg));
1da177e4
LT		 1397}
1398
1399
1400static int ext3_fill_super (struct super_block *sb, void *data, int silent)
1401{
1402 struct buffer_head * bh;
1403 struct ext3_super_block *es = NULL;
1404 struct ext3_sb_info *sbi;
43d23f90
MC		 1405 ext3_fsblk_t block;
1406 ext3_fsblk_t sb_block = get_sb_block(&data);
1407 ext3_fsblk_t logic_sb_block;
1da177e4 1408 unsigned long offset = 0;
eee194e7 1409 unsigned int journal_inum = 0;
71b96257 1410 unsigned long journal_devnum = 0;
1da177e4
LT		 1411 unsigned long def_mount_opts;
1412 struct inode *root;
1413 int blocksize;
1414 int hblock;
1415 int db_count;
1416 int i;
1417 int needs_recovery;
1418 __le32 features;
833f4077 1419 int err;
1da177e4 1420
f8314dc6 1421 sbi = kzalloc(sizeof(*sbi), GFP_KERNEL);
1da177e4
LT		 1422 if (!sbi)
1423 return -ENOMEM;
1424 sb->s_fs_info = sbi;
1da177e4
LT		 1425 sbi->s_mount_opt = 0;
1426 sbi->s_resuid = EXT3_DEF_RESUID;
1427 sbi->s_resgid = EXT3_DEF_RESGID;
1428
1429 unlock_kernel();
1430
1431 blocksize = sb_min_blocksize(sb, EXT3_MIN_BLOCK_SIZE);
1432 if (!blocksize) {
1433 printk(KERN_ERR "EXT3-fs: unable to set blocksize\n");
1434 goto out_fail;
1435 }
1436
1437 /*
1438 * The ext3 superblock will not be buffer aligned for other than 1kB
1439 * block sizes. We need to calculate the offset from buffer start.
1440 */
1441 if (blocksize != EXT3_MIN_BLOCK_SIZE) {
1442 logic_sb_block = (sb_block * EXT3_MIN_BLOCK_SIZE) / blocksize;
1443 offset = (sb_block * EXT3_MIN_BLOCK_SIZE) % blocksize;
1444 } else {
1445 logic_sb_block = sb_block;
1446 }
1447
1448 if (!(bh = sb_bread(sb, logic_sb_block))) {
1449 printk (KERN_ERR "EXT3-fs: unable to read superblock\n");
1450 goto out_fail;
1451 }
1452 /*
1453 * Note: s_es must be initialized as soon as possible because
1454 * some ext3 macro-instructions depend on its value
1455 */
1456 es = (struct ext3_super_block *) (((char *)bh->b_data) + offset);
1457 sbi->s_es = es;
1458 sb->s_magic = le16_to_cpu(es->s_magic);
1459 if (sb->s_magic != EXT3_SUPER_MAGIC)
1460 goto cantfind_ext3;
1461
1462 /* Set defaults before we parse the mount options */
1463 def_mount_opts = le32_to_cpu(es->s_default_mount_opts);
1464 if (def_mount_opts & EXT3_DEFM_DEBUG)
1465 set_opt(sbi->s_mount_opt, DEBUG);
1466 if (def_mount_opts & EXT3_DEFM_BSDGROUPS)
1467 set_opt(sbi->s_mount_opt, GRPID);
1468 if (def_mount_opts & EXT3_DEFM_UID16)
1469 set_opt(sbi->s_mount_opt, NO_UID32);
2e7842b8 1470#ifdef CONFIG_EXT3_FS_XATTR
1da177e4
LT		 1471 if (def_mount_opts & EXT3_DEFM_XATTR_USER)
1472 set_opt(sbi->s_mount_opt, XATTR_USER);
2e7842b8
HD		 1473#endif
1474#ifdef CONFIG_EXT3_FS_POSIX_ACL
1da177e4
LT		 1475 if (def_mount_opts & EXT3_DEFM_ACL)
1476 set_opt(sbi->s_mount_opt, POSIX_ACL);
2e7842b8 1477#endif
1da177e4
LT		 1478 if ((def_mount_opts & EXT3_DEFM_JMODE) == EXT3_DEFM_JMODE_DATA)
1479 sbi->s_mount_opt |= EXT3_MOUNT_JOURNAL_DATA;
1480 else if ((def_mount_opts & EXT3_DEFM_JMODE) == EXT3_DEFM_JMODE_ORDERED)
1481 sbi->s_mount_opt |= EXT3_MOUNT_ORDERED_DATA;
1482 else if ((def_mount_opts & EXT3_DEFM_JMODE) == EXT3_DEFM_JMODE_WBACK)
1483 sbi->s_mount_opt |= EXT3_MOUNT_WRITEBACK_DATA;
1484
1485 if (le16_to_cpu(sbi->s_es->s_errors) == EXT3_ERRORS_PANIC)
1486 set_opt(sbi->s_mount_opt, ERRORS_PANIC);
1487 else if (le16_to_cpu(sbi->s_es->s_errors) == EXT3_ERRORS_RO)
1488 set_opt(sbi->s_mount_opt, ERRORS_RO);
2245d7c2
DM		 1489 else
1490 set_opt(sbi->s_mount_opt, ERRORS_CONT);
1da177e4
LT		 1491
1492 sbi->s_resuid = le16_to_cpu(es->s_def_resuid);
1493 sbi->s_resgid = le16_to_cpu(es->s_def_resgid);
1494
1495 set_opt(sbi->s_mount_opt, RESERVATION);
1496
71b96257
JL		 1497 if (!parse_options ((char *) data, sb, &journal_inum, &journal_devnum,
1498 NULL, 0))
1da177e4
LT		 1499 goto failed_mount;
1500
1501 sb->s_flags = (sb->s_flags & ~MS_POSIXACL) |
1502 ((sbi->s_mount_opt & EXT3_MOUNT_POSIX_ACL) ? MS_POSIXACL : 0);
1503
1504 if (le32_to_cpu(es->s_rev_level) == EXT3_GOOD_OLD_REV &&
1505 (EXT3_HAS_COMPAT_FEATURE(sb, ~0U) ||
1506 EXT3_HAS_RO_COMPAT_FEATURE(sb, ~0U) ||
1507 EXT3_HAS_INCOMPAT_FEATURE(sb, ~0U)))
ae6ddcc5 1508 printk(KERN_WARNING
1da177e4
LT		 1509 "EXT3-fs warning: feature flags set on rev 0 fs, "
1510 "running e2fsck is recommended\n");
1511 /*
1512 * Check feature flags regardless of the revision level, since we
1513 * previously didn't change the revision level when setting the flags,
1514 * so there is a chance incompat flags are set on a rev 0 filesystem.
1515 */
1516 features = EXT3_HAS_INCOMPAT_FEATURE(sb, ~EXT3_FEATURE_INCOMPAT_SUPP);
1517 if (features) {
1518 printk(KERN_ERR "EXT3-fs: %s: couldn't mount because of "
1519 "unsupported optional features (%x).\n",
1520 sb->s_id, le32_to_cpu(features));
1521 goto failed_mount;
1522 }
1523 features = EXT3_HAS_RO_COMPAT_FEATURE(sb, ~EXT3_FEATURE_RO_COMPAT_SUPP);
1524 if (!(sb->s_flags & MS_RDONLY) && features) {
1525 printk(KERN_ERR "EXT3-fs: %s: couldn't mount RDWR because of "
1526 "unsupported optional features (%x).\n",
1527 sb->s_id, le32_to_cpu(features));
1528 goto failed_mount;
1529 }
1530 blocksize = BLOCK_SIZE << le32_to_cpu(es->s_log_block_size);
1531
1532 if (blocksize < EXT3_MIN_BLOCK_SIZE ||
1533 blocksize > EXT3_MAX_BLOCK_SIZE) {
ae6ddcc5 1534 printk(KERN_ERR
1da177e4
LT		 1535 "EXT3-fs: Unsupported filesystem blocksize %d on %s.\n",
1536 blocksize, sb->s_id);
1537 goto failed_mount;
1538 }
1539
1540 hblock = bdev_hardsect_size(sb->s_bdev);
1541 if (sb->s_blocksize != blocksize) {
1542 /*
1543 * Make sure the blocksize for the filesystem is larger
1544 * than the hardware sectorsize for the machine.
1545 */
1546 if (blocksize < hblock) {
1547 printk(KERN_ERR "EXT3-fs: blocksize %d too small for "
1548 "device blocksize %d.\n", blocksize, hblock);
1549 goto failed_mount;
1550 }
1551
1552 brelse (bh);
1553 sb_set_blocksize(sb, blocksize);
1554 logic_sb_block = (sb_block * EXT3_MIN_BLOCK_SIZE) / blocksize;
1555 offset = (sb_block * EXT3_MIN_BLOCK_SIZE) % blocksize;
1556 bh = sb_bread(sb, logic_sb_block);
1557 if (!bh) {
ae6ddcc5 1558 printk(KERN_ERR
1da177e4
LT		 1559 "EXT3-fs: Can't read superblock on 2nd try.\n");
1560 goto failed_mount;
1561 }
1562 es = (struct ext3_super_block *)(((char *)bh->b_data) + offset);
1563 sbi->s_es = es;
1564 if (es->s_magic != cpu_to_le16(EXT3_SUPER_MAGIC)) {
ae6ddcc5 1565 printk (KERN_ERR
1da177e4
LT		 1566 "EXT3-fs: Magic mismatch, very weird !\n");
1567 goto failed_mount;
1568 }
1569 }
1570
1571 sb->s_maxbytes = ext3_max_size(sb->s_blocksize_bits);
1572
1573 if (le32_to_cpu(es->s_rev_level) == EXT3_GOOD_OLD_REV) {
1574 sbi->s_inode_size = EXT3_GOOD_OLD_INODE_SIZE;
1575 sbi->s_first_ino = EXT3_GOOD_OLD_FIRST_INO;
1576 } else {
1577 sbi->s_inode_size = le16_to_cpu(es->s_inode_size);
1578 sbi->s_first_ino = le32_to_cpu(es->s_first_ino);
1579 if ((sbi->s_inode_size < EXT3_GOOD_OLD_INODE_SIZE) ||
3fc74269 1580 (!is_power_of_2(sbi->s_inode_size)) ||
1da177e4
LT		 1581 (sbi->s_inode_size > blocksize)) {
1582 printk (KERN_ERR
1583 "EXT3-fs: unsupported inode size: %d\n",
1584 sbi->s_inode_size);
1585 goto failed_mount;
1586 }
1587 }
1588 sbi->s_frag_size = EXT3_MIN_FRAG_SIZE <<
1589 le32_to_cpu(es->s_log_frag_size);
1590 if (blocksize != sbi->s_frag_size) {
1591 printk(KERN_ERR
1592 "EXT3-fs: fragsize %lu != blocksize %u (unsupported)\n",
1593 sbi->s_frag_size, blocksize);
1594 goto failed_mount;
1595 }
1596 sbi->s_frags_per_block = 1;
1597 sbi->s_blocks_per_group = le32_to_cpu(es->s_blocks_per_group);
1598 sbi->s_frags_per_group = le32_to_cpu(es->s_frags_per_group);
1599 sbi->s_inodes_per_group = le32_to_cpu(es->s_inodes_per_group);
1600 if (EXT3_INODE_SIZE(sb) == 0)
1601 goto cantfind_ext3;
1602 sbi->s_inodes_per_block = blocksize / EXT3_INODE_SIZE(sb);
1603 if (sbi->s_inodes_per_block == 0)
1604 goto cantfind_ext3;
1605 sbi->s_itb_per_group = sbi->s_inodes_per_group /
1606 sbi->s_inodes_per_block;
1607 sbi->s_desc_per_block = blocksize / sizeof(struct ext3_group_desc);
1608 sbi->s_sbh = bh;
1609 sbi->s_mount_state = le16_to_cpu(es->s_state);
f0d1b0b3
DH		 1610 sbi->s_addr_per_block_bits = ilog2(EXT3_ADDR_PER_BLOCK(sb));
1611 sbi->s_desc_per_block_bits = ilog2(EXT3_DESC_PER_BLOCK(sb));
1da177e4
LT		 1612 for (i=0; i < 4; i++)
1613 sbi->s_hash_seed[i] = le32_to_cpu(es->s_hash_seed[i]);
1614 sbi->s_def_hash_version = es->s_def_hash_version;
1615
1616 if (sbi->s_blocks_per_group > blocksize * 8) {
1617 printk (KERN_ERR
1618 "EXT3-fs: #blocks per group too big: %lu\n",
1619 sbi->s_blocks_per_group);
1620 goto failed_mount;
1621 }
1622 if (sbi->s_frags_per_group > blocksize * 8) {
1623 printk (KERN_ERR
1624 "EXT3-fs: #fragments per group too big: %lu\n",
1625 sbi->s_frags_per_group);
1626 goto failed_mount;
1627 }
1628 if (sbi->s_inodes_per_group > blocksize * 8) {
1629 printk (KERN_ERR
1630 "EXT3-fs: #inodes per group too big: %lu\n",
1631 sbi->s_inodes_per_group);
1632 goto failed_mount;
1633 }
1634
fcd5df35
MC		 1635 if (le32_to_cpu(es->s_blocks_count) >
1636 (sector_t)(~0ULL) >> (sb->s_blocksize_bits - 9)) {
1637 printk(KERN_ERR "EXT3-fs: filesystem on %s:"
1638 " too large to mount safely\n", sb->s_id);
1639 if (sizeof(sector_t) < 8)
1640 printk(KERN_WARNING "EXT3-fs: CONFIG_LBD not "
1641 "enabled\n");
1642 goto failed_mount;
1643 }
1644
1da177e4
LT		 1645 if (EXT3_BLOCKS_PER_GROUP(sb) == 0)
1646 goto cantfind_ext3;
855565e8
ES		 1647 sbi->s_groups_count = ((le32_to_cpu(es->s_blocks_count) -
1648 le32_to_cpu(es->s_first_data_block) - 1)
1649 / EXT3_BLOCKS_PER_GROUP(sb)) + 1;
1da177e4
LT		 1650 db_count = (sbi->s_groups_count + EXT3_DESC_PER_BLOCK(sb) - 1) /
1651 EXT3_DESC_PER_BLOCK(sb);
1652 sbi->s_group_desc = kmalloc(db_count * sizeof (struct buffer_head *),
1653 GFP_KERNEL);
1654 if (sbi->s_group_desc == NULL) {
1655 printk (KERN_ERR "EXT3-fs: not enough memory\n");
1656 goto failed_mount;
1657 }
1658
1da177e4
LT		 1659 bgl_lock_init(&sbi->s_blockgroup_lock);
1660
1661 for (i = 0; i < db_count; i++) {
1662 block = descriptor_loc(sb, logic_sb_block, i);
1663 sbi->s_group_desc[i] = sb_bread(sb, block);
1664 if (!sbi->s_group_desc[i]) {
1665 printk (KERN_ERR "EXT3-fs: "
1666 "can't read group descriptor %d\n", i);
1667 db_count = i;
1668 goto failed_mount2;
1669 }
1670 }
1671 if (!ext3_check_descriptors (sb)) {
d2e5b13c 1672 printk(KERN_ERR "EXT3-fs: group descriptors corrupted!\n");
1da177e4
LT		 1673 goto failed_mount2;
1674 }
1675 sbi->s_gdb_count = db_count;
1676 get_random_bytes(&sbi->s_next_generation, sizeof(u32));
1677 spin_lock_init(&sbi->s_next_gen_lock);
0216bfcf 1678
833f4077
PZ		 1679 err = percpu_counter_init(&sbi->s_freeblocks_counter,
1680 ext3_count_free_blocks(sb));
1681 if (!err) {
1682 err = percpu_counter_init(&sbi->s_freeinodes_counter,
1683 ext3_count_free_inodes(sb));
1684 }
1685 if (!err) {
1686 err = percpu_counter_init(&sbi->s_dirs_counter,
1687 ext3_count_dirs(sb));
1688 }
1689 if (err) {
1690 printk(KERN_ERR "EXT3-fs: insufficient memory\n");
1691 goto failed_mount3;
1692 }
0216bfcf 1693
1da177e4
LT		 1694 /* per fileystem reservation list head & lock */
1695 spin_lock_init(&sbi->s_rsv_window_lock);
1696 sbi->s_rsv_window_root = RB_ROOT;
1697 /* Add a single, static dummy reservation to the start of the
1698 * reservation window list --- it gives us a placeholder for
1699 * append-at-start-of-list which makes the allocation logic
1700 * _much_ simpler. */
1701 sbi->s_rsv_window_head.rsv_start = EXT3_RESERVE_WINDOW_NOT_ALLOCATED;
1702 sbi->s_rsv_window_head.rsv_end = EXT3_RESERVE_WINDOW_NOT_ALLOCATED;
1703 sbi->s_rsv_window_head.rsv_alloc_hit = 0;
1704 sbi->s_rsv_window_head.rsv_goal_size = 0;
1705 ext3_rsv_window_add(sb, &sbi->s_rsv_window_head);
1706
1707 /*
1708 * set up enough so that it can read an inode
1709 */
1710 sb->s_op = &ext3_sops;
1711 sb->s_export_op = &ext3_export_ops;
1712 sb->s_xattr = ext3_xattr_handlers;
1713#ifdef CONFIG_QUOTA
1714 sb->s_qcop = &ext3_qctl_operations;
1715 sb->dq_op = &ext3_quota_operations;
1716#endif
1717 INIT_LIST_HEAD(&sbi->s_orphan); /* unlinked but open files */
1718
1719 sb->s_root = NULL;
1720
1721 needs_recovery = (es->s_last_orphan != 0 ||
1722 EXT3_HAS_INCOMPAT_FEATURE(sb,
1723 EXT3_FEATURE_INCOMPAT_RECOVER));
1724
1725 /*
1726 * The first inode we look at is the journal inode. Don't try
1727 * root first: it may be modified in the journal!
1728 */
1729 if (!test_opt(sb, NOLOAD) &&
1730 EXT3_HAS_COMPAT_FEATURE(sb, EXT3_FEATURE_COMPAT_HAS_JOURNAL)) {
71b96257 1731 if (ext3_load_journal(sb, es, journal_devnum))
0216bfcf 1732 goto failed_mount3;
1da177e4
LT		 1733 } else if (journal_inum) {
1734 if (ext3_create_journal(sb, es, journal_inum))
0216bfcf 1735 goto failed_mount3;
1da177e4
LT		 1736 } else {
1737 if (!silent)
1738 printk (KERN_ERR
1739 "ext3: No journal on filesystem on %s\n",
1740 sb->s_id);
0216bfcf 1741 goto failed_mount3;
1da177e4
LT		 1742 }
1743
1744 /* We have now updated the journal if required, so we can
1745 * validate the data journaling mode. */
1746 switch (test_opt(sb, DATA_FLAGS)) {
1747 case 0:
1748 /* No mode set, assume a default based on the journal
1749 capabilities: ORDERED_DATA if the journal can
1750 cope, else JOURNAL_DATA */
1751 if (journal_check_available_features
1752 (sbi->s_journal, 0, 0, JFS_FEATURE_INCOMPAT_REVOKE))
1753 set_opt(sbi->s_mount_opt, ORDERED_DATA);
1754 else
1755 set_opt(sbi->s_mount_opt, JOURNAL_DATA);
1756 break;
1757
1758 case EXT3_MOUNT_ORDERED_DATA:
1759 case EXT3_MOUNT_WRITEBACK_DATA:
1760 if (!journal_check_available_features
1761 (sbi->s_journal, 0, 0, JFS_FEATURE_INCOMPAT_REVOKE)) {
1762 printk(KERN_ERR "EXT3-fs: Journal does not support "
1763 "requested data journaling mode\n");
0216bfcf 1764 goto failed_mount4;
1da177e4
LT		 1765 }
1766 default:
1767 break;
1768 }
1769
1770 if (test_opt(sb, NOBH)) {
1da177e4
LT		 1771 if (!(test_opt(sb, DATA_FLAGS) == EXT3_MOUNT_WRITEBACK_DATA)) {
1772 printk(KERN_WARNING "EXT3-fs: Ignoring nobh option - "
1773 "its supported only with writeback mode\n");
1774 clear_opt(sbi->s_mount_opt, NOBH);
1775 }
1776 }
1777 /*
1778 * The journal_load will have done any necessary log recovery,
1779 * so we can safely mount the rest of the filesystem now.
1780 */
1781
1782 root = iget(sb, EXT3_ROOT_INO);
1783 sb->s_root = d_alloc_root(root);
1784 if (!sb->s_root) {
1785 printk(KERN_ERR "EXT3-fs: get root inode failed\n");
1786 iput(root);
0216bfcf 1787 goto failed_mount4;
1da177e4
LT		 1788 }
1789 if (!S_ISDIR(root->i_mode) || !root->i_blocks || !root->i_size) {
1790 dput(sb->s_root);
1791 sb->s_root = NULL;
1792 printk(KERN_ERR "EXT3-fs: corrupt root inode, run e2fsck\n");
0216bfcf 1793 goto failed_mount4;
1da177e4
LT		 1794 }
1795
1796 ext3_setup_super (sb, es, sb->s_flags & MS_RDONLY);
1797 /*
1798 * akpm: core read_super() calls in here with the superblock locked.
1799 * That deadlocks, because orphan cleanup needs to lock the superblock
1800 * in numerous places. Here we just pop the lock - it's relatively
1801 * harmless, because we are now ready to accept write_super() requests,
1802 * and aviro says that's the only reason for hanging onto the
1803 * superblock lock.
1804 */
1805 EXT3_SB(sb)->s_mount_state |= EXT3_ORPHAN_FS;
1806 ext3_orphan_cleanup(sb, es);
1807 EXT3_SB(sb)->s_mount_state &= ~EXT3_ORPHAN_FS;
1808 if (needs_recovery)
1809 printk (KERN_INFO "EXT3-fs: recovery complete.\n");
1810 ext3_mark_recovery_complete(sb, es);
1811 printk (KERN_INFO "EXT3-fs: mounted filesystem with %s data mode.\n",
1812 test_opt(sb,DATA_FLAGS) == EXT3_MOUNT_JOURNAL_DATA ? "journal":
1813 test_opt(sb,DATA_FLAGS) == EXT3_MOUNT_ORDERED_DATA ? "ordered":
1814 "writeback");
1815
1da177e4
LT		 1816 lock_kernel();
1817 return 0;
1818
1819cantfind_ext3:
1820 if (!silent)
1821 printk(KERN_ERR "VFS: Can't find ext3 filesystem on dev %s.\n",
1822 sb->s_id);
1823 goto failed_mount;
1824
0216bfcf 1825failed_mount4:
1da177e4 1826 journal_destroy(sbi->s_journal);
0216bfcf
MC		 1827failed_mount3:
1828 percpu_counter_destroy(&sbi->s_freeblocks_counter);
1829 percpu_counter_destroy(&sbi->s_freeinodes_counter);
1830 percpu_counter_destroy(&sbi->s_dirs_counter);
1da177e4
LT		 1831failed_mount2:
1832 for (i = 0; i < db_count; i++)
1833 brelse(sbi->s_group_desc[i]);
1834 kfree(sbi->s_group_desc);
1835failed_mount:
1836#ifdef CONFIG_QUOTA
1837 for (i = 0; i < MAXQUOTAS; i++)
1838 kfree(sbi->s_qf_names[i]);
1839#endif
1840 ext3_blkdev_remove(sbi);
1841 brelse(bh);
1842out_fail:
1843 sb->s_fs_info = NULL;
1844 kfree(sbi);
1845 lock_kernel();
1846 return -EINVAL;
1847}
1848
1849/*
1850 * Setup any per-fs journal parameters now. We'll do this both on
1851 * initial mount, once the journal has been initialised but before we've
ae6ddcc5 1852 * done any recovery; and again on any subsequent remount.
1da177e4
LT		 1853 */
1854static void ext3_init_journal_params(struct super_block *sb, journal_t *journal)
1855{
1856 struct ext3_sb_info *sbi = EXT3_SB(sb);
1857
1858 if (sbi->s_commit_interval)
1859 journal->j_commit_interval = sbi->s_commit_interval;
1860 /* We could also set up an ext3-specific default for the commit
1861 * interval here, but for now we'll just fall back to the jbd
1862 * default. */
1863
1864 spin_lock(&journal->j_state_lock);
1865 if (test_opt(sb, BARRIER))
1866 journal->j_flags |= JFS_BARRIER;
1867 else
1868 journal->j_flags &= ~JFS_BARRIER;
1869 spin_unlock(&journal->j_state_lock);
1870}
1871
eee194e7
ES		 1872static journal_t *ext3_get_journal(struct super_block *sb,
1873 unsigned int journal_inum)
1da177e4
LT		 1874{
1875 struct inode *journal_inode;
1876 journal_t *journal;
1877
1878 /* First, test for the existence of a valid inode on disk. Bad
1879 * things happen if we iget() an unused inode, as the subsequent
1880 * iput() will try to delete it. */
1881
1882 journal_inode = iget(sb, journal_inum);
1883 if (!journal_inode) {
1884 printk(KERN_ERR "EXT3-fs: no journal found.\n");
1885 return NULL;
1886 }
1887 if (!journal_inode->i_nlink) {
1888 make_bad_inode(journal_inode);
1889 iput(journal_inode);
1890 printk(KERN_ERR "EXT3-fs: journal inode is deleted.\n");
1891 return NULL;
1892 }
1893
1894 jbd_debug(2, "Journal inode found at %p: %Ld bytes\n",
1895 journal_inode, journal_inode->i_size);
1896 if (is_bad_inode(journal_inode) || !S_ISREG(journal_inode->i_mode)) {
1897 printk(KERN_ERR "EXT3-fs: invalid journal inode.\n");
1898 iput(journal_inode);
1899 return NULL;
1900 }
1901
1902 journal = journal_init_inode(journal_inode);
1903 if (!journal) {
1904 printk(KERN_ERR "EXT3-fs: Could not load journal inode\n");
1905 iput(journal_inode);
1906 return NULL;
1907 }
1908 journal->j_private = sb;
1909 ext3_init_journal_params(sb, journal);
1910 return journal;
1911}
1912
1913static journal_t *ext3_get_dev_journal(struct super_block *sb,
1914 dev_t j_dev)
1915{
1916 struct buffer_head * bh;
1917 journal_t *journal;
43d23f90 1918 ext3_fsblk_t start;
1c2bf374 1919 ext3_fsblk_t len;
1da177e4 1920 int hblock, blocksize;
43d23f90 1921 ext3_fsblk_t sb_block;
1da177e4
LT		 1922 unsigned long offset;
1923 struct ext3_super_block * es;
1924 struct block_device *bdev;
1925
1926 bdev = ext3_blkdev_get(j_dev);
1927 if (bdev == NULL)
1928 return NULL;
1929
1930 if (bd_claim(bdev, sb)) {
1931 printk(KERN_ERR
1932 "EXT3: failed to claim external journal device.\n");
1933 blkdev_put(bdev);
1934 return NULL;
1935 }
1936
1937 blocksize = sb->s_blocksize;
1938 hblock = bdev_hardsect_size(bdev);
1939 if (blocksize < hblock) {
1940 printk(KERN_ERR
1941 "EXT3-fs: blocksize too small for journal device.\n");
1942 goto out_bdev;
1943 }
1944
1945 sb_block = EXT3_MIN_BLOCK_SIZE / blocksize;
1946 offset = EXT3_MIN_BLOCK_SIZE % blocksize;
1947 set_blocksize(bdev, blocksize);
1948 if (!(bh = __bread(bdev, sb_block, blocksize))) {
1949 printk(KERN_ERR "EXT3-fs: couldn't read superblock of "
1950 "external journal\n");
1951 goto out_bdev;
1952 }
1953
1954 es = (struct ext3_super_block *) (((char *)bh->b_data) + offset);
1955 if ((le16_to_cpu(es->s_magic) != EXT3_SUPER_MAGIC) ||
1956 !(le32_to_cpu(es->s_feature_incompat) &
1957 EXT3_FEATURE_INCOMPAT_JOURNAL_DEV)) {
1958 printk(KERN_ERR "EXT3-fs: external journal has "
1959 "bad superblock\n");
1960 brelse(bh);
1961 goto out_bdev;
1962 }
1963
1964 if (memcmp(EXT3_SB(sb)->s_es->s_journal_uuid, es->s_uuid, 16)) {
1965 printk(KERN_ERR "EXT3-fs: journal UUID does not match\n");
1966 brelse(bh);
1967 goto out_bdev;
1968 }
1969
1970 len = le32_to_cpu(es->s_blocks_count);
1971 start = sb_block + 1;
1972 brelse(bh); /* we're done with the superblock */
1973
1974 journal = journal_init_dev(bdev, sb->s_bdev,
1975 start, len, blocksize);
1976 if (!journal) {
1977 printk(KERN_ERR "EXT3-fs: failed to create device journal\n");
1978 goto out_bdev;
1979 }
1980 journal->j_private = sb;
1981 ll_rw_block(READ, 1, &journal->j_sb_buffer);
1982 wait_on_buffer(journal->j_sb_buffer);
1983 if (!buffer_uptodate(journal->j_sb_buffer)) {
1984 printk(KERN_ERR "EXT3-fs: I/O error on journal device\n");
1985 goto out_journal;
1986 }
1987 if (be32_to_cpu(journal->j_superblock->s_nr_users) != 1) {
1988 printk(KERN_ERR "EXT3-fs: External journal has more than one "
1989 "user (unsupported) - %d\n",
1990 be32_to_cpu(journal->j_superblock->s_nr_users));
1991 goto out_journal;
1992 }
1993 EXT3_SB(sb)->journal_bdev = bdev;
1994 ext3_init_journal_params(sb, journal);
1995 return journal;
1996out_journal:
1997 journal_destroy(journal);
1998out_bdev:
1999 ext3_blkdev_put(bdev);
2000 return NULL;
2001}
2002
71b96257
JL		 2003static int ext3_load_journal(struct super_block *sb,
2004 struct ext3_super_block *es,
2005 unsigned long journal_devnum)
1da177e4
LT		 2006{
2007 journal_t *journal;
eee194e7 2008 unsigned int journal_inum = le32_to_cpu(es->s_journal_inum);
71b96257 2009 dev_t journal_dev;
1da177e4
LT		 2010 int err = 0;
2011 int really_read_only;
2012
71b96257
JL		 2013 if (journal_devnum &&
2014 journal_devnum != le32_to_cpu(es->s_journal_dev)) {
2015 printk(KERN_INFO "EXT3-fs: external journal device major/minor "
2016 "numbers have changed\n");
2017 journal_dev = new_decode_dev(journal_devnum);
2018 } else
2019 journal_dev = new_decode_dev(le32_to_cpu(es->s_journal_dev));
2020
1da177e4
LT		 2021 really_read_only = bdev_read_only(sb->s_bdev);
2022
2023 /*
2024 * Are we loading a blank journal or performing recovery after a
2025 * crash? For recovery, we need to check in advance whether we
2026 * can get read-write access to the device.
2027 */
2028
2029 if (EXT3_HAS_INCOMPAT_FEATURE(sb, EXT3_FEATURE_INCOMPAT_RECOVER)) {
2030 if (sb->s_flags & MS_RDONLY) {
2031 printk(KERN_INFO "EXT3-fs: INFO: recovery "
2032 "required on readonly filesystem.\n");
2033 if (really_read_only) {
2034 printk(KERN_ERR "EXT3-fs: write access "
2035 "unavailable, cannot proceed.\n");
2036 return -EROFS;
2037 }
2038 printk (KERN_INFO "EXT3-fs: write access will "
2039 "be enabled during recovery.\n");
2040 }
2041 }
2042
2043 if (journal_inum && journal_dev) {
2044 printk(KERN_ERR "EXT3-fs: filesystem has both journal "
2045 "and inode journals!\n");
2046 return -EINVAL;
2047 }
2048
2049 if (journal_inum) {
2050 if (!(journal = ext3_get_journal(sb, journal_inum)))
2051 return -EINVAL;
2052 } else {
2053 if (!(journal = ext3_get_dev_journal(sb, journal_dev)))
2054 return -EINVAL;
2055 }
2056
2057 if (!really_read_only && test_opt(sb, UPDATE_JOURNAL)) {
2058 err = journal_update_format(journal);
2059 if (err) {
2060 printk(KERN_ERR "EXT3-fs: error updating journal.\n");
2061 journal_destroy(journal);
2062 return err;
2063 }
2064 }
2065
2066 if (!EXT3_HAS_INCOMPAT_FEATURE(sb, EXT3_FEATURE_INCOMPAT_RECOVER))
2067 err = journal_wipe(journal, !really_read_only);
2068 if (!err)
2069 err = journal_load(journal);
2070
2071 if (err) {
2072 printk(KERN_ERR "EXT3-fs: error loading journal.\n");
2073 journal_destroy(journal);
2074 return err;
2075 }
2076
2077 EXT3_SB(sb)->s_journal = journal;
2078 ext3_clear_journal_err(sb, es);
71b96257
JL		 2079
2080 if (journal_devnum &&
2081 journal_devnum != le32_to_cpu(es->s_journal_dev)) {
2082 es->s_journal_dev = cpu_to_le32(journal_devnum);
2083 sb->s_dirt = 1;
2084
2085 /* Make sure we flush the recovery flag to disk. */
2086 ext3_commit_super(sb, es, 1);
2087 }
2088
1da177e4
LT		 2089 return 0;
2090}
2091
2092static int ext3_create_journal(struct super_block * sb,
2093 struct ext3_super_block * es,
eee194e7 2094 unsigned int journal_inum)
1da177e4
LT		 2095{
2096 journal_t *journal;
952d9de1 2097 int err;
1da177e4
LT		 2098
2099 if (sb->s_flags & MS_RDONLY) {
2100 printk(KERN_ERR "EXT3-fs: readonly filesystem when trying to "
2101 "create journal.\n");
2102 return -EROFS;
2103 }
2104
952d9de1
BP		 2105 journal = ext3_get_journal(sb, journal_inum);
2106 if (!journal)
1da177e4
LT		 2107 return -EINVAL;
2108
eee194e7 2109 printk(KERN_INFO "EXT3-fs: creating new journal on inode %u\n",
1da177e4
LT		 2110 journal_inum);
2111
952d9de1
BP		 2112 err = journal_create(journal);
2113 if (err) {
1da177e4
LT		 2114 printk(KERN_ERR "EXT3-fs: error creating journal.\n");
2115 journal_destroy(journal);
2116 return -EIO;
2117 }
2118
2119 EXT3_SB(sb)->s_journal = journal;
2120
2121 ext3_update_dynamic_rev(sb);
2122 EXT3_SET_INCOMPAT_FEATURE(sb, EXT3_FEATURE_INCOMPAT_RECOVER);
2123 EXT3_SET_COMPAT_FEATURE(sb, EXT3_FEATURE_COMPAT_HAS_JOURNAL);
2124
2125 es->s_journal_inum = cpu_to_le32(journal_inum);
2126 sb->s_dirt = 1;
2127
2128 /* Make sure we flush the recovery flag to disk. */
2129 ext3_commit_super(sb, es, 1);
2130
2131 return 0;
2132}
2133
2134static void ext3_commit_super (struct super_block * sb,
2135 struct ext3_super_block * es,
2136 int sync)
2137{
2138 struct buffer_head *sbh = EXT3_SB(sb)->s_sbh;
2139
2140 if (!sbh)
2141 return;
2142 es->s_wtime = cpu_to_le32(get_seconds());
2143 es->s_free_blocks_count = cpu_to_le32(ext3_count_free_blocks(sb));
2144 es->s_free_inodes_count = cpu_to_le32(ext3_count_free_inodes(sb));
2145 BUFFER_TRACE(sbh, "marking dirty");
2146 mark_buffer_dirty(sbh);
2147 if (sync)
2148 sync_dirty_buffer(sbh);
2149}
2150
2151
2152/*
2153 * Have we just finished recovery? If so, and if we are mounting (or
2154 * remounting) the filesystem readonly, then we will end up with a
2155 * consistent fs on disk. Record that fact.
2156 */
2157static void ext3_mark_recovery_complete(struct super_block * sb,
2158 struct ext3_super_block * es)
2159{
2160 journal_t *journal = EXT3_SB(sb)->s_journal;
2161
2162 journal_lock_updates(journal);
2163 journal_flush(journal);
030703e4 2164 lock_super(sb);
1da177e4
LT		 2165 if (EXT3_HAS_INCOMPAT_FEATURE(sb, EXT3_FEATURE_INCOMPAT_RECOVER) &&
2166 sb->s_flags & MS_RDONLY) {
2167 EXT3_CLEAR_INCOMPAT_FEATURE(sb, EXT3_FEATURE_INCOMPAT_RECOVER);
2168 sb->s_dirt = 0;
2169 ext3_commit_super(sb, es, 1);
2170 }
030703e4 2171 unlock_super(sb);
1da177e4
LT		 2172 journal_unlock_updates(journal);
2173}
2174
2175/*
2176 * If we are mounting (or read-write remounting) a filesystem whose journal
2177 * has recorded an error from a previous lifetime, move that error to the
2178 * main filesystem now.
2179 */
2180static void ext3_clear_journal_err(struct super_block * sb,
2181 struct ext3_super_block * es)
2182{
2183 journal_t *journal;
2184 int j_errno;
2185 const char *errstr;
2186
2187 journal = EXT3_SB(sb)->s_journal;
2188
2189 /*
2190 * Now check for any error status which may have been recorded in the
2191 * journal by a prior ext3_error() or ext3_abort()
2192 */
2193
2194 j_errno = journal_errno(journal);
2195 if (j_errno) {
2196 char nbuf[16];
2197
2198 errstr = ext3_decode_error(sb, j_errno, nbuf);
2199 ext3_warning(sb, __FUNCTION__, "Filesystem error recorded "
2200 "from previous mount: %s", errstr);
2201 ext3_warning(sb, __FUNCTION__, "Marking fs in need of "
2202 "filesystem check.");
2203
2204 EXT3_SB(sb)->s_mount_state |= EXT3_ERROR_FS;
2205 es->s_state |= cpu_to_le16(EXT3_ERROR_FS);
2206 ext3_commit_super (sb, es, 1);
2207
2208 journal_clear_err(journal);
2209 }
2210}
2211
2212/*
2213 * Force the running and committing transactions to commit,
2214 * and wait on the commit.
2215 */
2216int ext3_force_commit(struct super_block *sb)
2217{
2218 journal_t *journal;
2219 int ret;
2220
2221 if (sb->s_flags & MS_RDONLY)
2222 return 0;
2223
2224 journal = EXT3_SB(sb)->s_journal;
2225 sb->s_dirt = 0;
2226 ret = ext3_journal_force_commit(journal);
2227 return ret;
2228}
2229
2230/*
2231 * Ext3 always journals updates to the superblock itself, so we don't
2232 * have to propagate any other updates to the superblock on disk at this
2233 * point. Just start an async writeback to get the buffers on their way
2234 * to the disk.
2235 *
2236 * This implicitly triggers the writebehind on sync().
2237 */
2238
2239static void ext3_write_super (struct super_block * sb)
2240{
7892f2f4 2241 if (mutex_trylock(&sb->s_lock) != 0)
1da177e4
LT		 2242 BUG();
2243 sb->s_dirt = 0;
2244}
2245
2246static int ext3_sync_fs(struct super_block *sb, int wait)
2247{
2248 tid_t target;
2249
2250 sb->s_dirt = 0;
2251 if (journal_start_commit(EXT3_SB(sb)->s_journal, &target)) {
2252 if (wait)
2253 log_wait_commit(EXT3_SB(sb)->s_journal, target);
2254 }
2255 return 0;
2256}
2257
2258/*
2259 * LVM calls this function before a (read-only) snapshot is created. This
2260 * gives us a chance to flush the journal completely and mark the fs clean.
2261 */
2262static void ext3_write_super_lockfs(struct super_block *sb)
2263{
2264 sb->s_dirt = 0;
2265
2266 if (!(sb->s_flags & MS_RDONLY)) {
2267 journal_t *journal = EXT3_SB(sb)->s_journal;
2268
2269 /* Now we set up the journal barrier. */
2270 journal_lock_updates(journal);
2271 journal_flush(journal);
2272
2273 /* Journal blocked and flushed, clear needs_recovery flag. */
2274 EXT3_CLEAR_INCOMPAT_FEATURE(sb, EXT3_FEATURE_INCOMPAT_RECOVER);
2275 ext3_commit_super(sb, EXT3_SB(sb)->s_es, 1);
2276 }
2277}
2278
2279/*
2280 * Called by LVM after the snapshot is done. We need to reset the RECOVER
2281 * flag here, even though the filesystem is not technically dirty yet.
2282 */
2283static void ext3_unlockfs(struct super_block *sb)
2284{
2285 if (!(sb->s_flags & MS_RDONLY)) {
2286 lock_super(sb);
2287 /* Reser the needs_recovery flag before the fs is unlocked. */
2288 EXT3_SET_INCOMPAT_FEATURE(sb, EXT3_FEATURE_INCOMPAT_RECOVER);
2289 ext3_commit_super(sb, EXT3_SB(sb)->s_es, 1);
2290 unlock_super(sb);
2291 journal_unlock_updates(EXT3_SB(sb)->s_journal);
2292 }
2293}
2294
2295static int ext3_remount (struct super_block * sb, int * flags, char * data)
2296{
2297 struct ext3_super_block * es;
2298 struct ext3_sb_info *sbi = EXT3_SB(sb);
43d23f90 2299 ext3_fsblk_t n_blocks_count = 0;
08c6a96f
JK		 2300 unsigned long old_sb_flags;
2301 struct ext3_mount_options old_opts;
2302 int err;
2303#ifdef CONFIG_QUOTA
2304 int i;
2305#endif
2306
2307 /* Store the original options */
2308 old_sb_flags = sb->s_flags;
2309 old_opts.s_mount_opt = sbi->s_mount_opt;
2310 old_opts.s_resuid = sbi->s_resuid;
2311 old_opts.s_resgid = sbi->s_resgid;
2312 old_opts.s_commit_interval = sbi->s_commit_interval;
2313#ifdef CONFIG_QUOTA
2314 old_opts.s_jquota_fmt = sbi->s_jquota_fmt;
2315 for (i = 0; i < MAXQUOTAS; i++)
2316 old_opts.s_qf_names[i] = sbi->s_qf_names[i];
2317#endif
1da177e4
LT		 2318
2319 /*
2320 * Allow the "check" option to be passed as a remount option.
2321 */
71b96257 2322 if (!parse_options(data, sb, NULL, NULL, &n_blocks_count, 1)) {
08c6a96f
JK		 2323 err = -EINVAL;
2324 goto restore_opts;
2325 }
1da177e4
LT		 2326
2327 if (sbi->s_mount_opt & EXT3_MOUNT_ABORT)
2328 ext3_abort(sb, __FUNCTION__, "Abort forced by user");
2329
2330 sb->s_flags = (sb->s_flags & ~MS_POSIXACL) |
2331 ((sbi->s_mount_opt & EXT3_MOUNT_POSIX_ACL) ? MS_POSIXACL : 0);
2332
2333 es = sbi->s_es;
2334
2335 ext3_init_journal_params(sb, sbi->s_journal);
2336
2337 if ((*flags & MS_RDONLY) != (sb->s_flags & MS_RDONLY) ||
2338 n_blocks_count > le32_to_cpu(es->s_blocks_count)) {
08c6a96f
JK		 2339 if (sbi->s_mount_opt & EXT3_MOUNT_ABORT) {
2340 err = -EROFS;
2341 goto restore_opts;
2342 }
1da177e4
LT		 2343
2344 if (*flags & MS_RDONLY) {
2345 /*
2346 * First of all, the unconditional stuff we have to do
2347 * to disable replay of the journal when we next remount
2348 */
2349 sb->s_flags |= MS_RDONLY;
2350
2351 /*
2352 * OK, test if we are remounting a valid rw partition
2353 * readonly, and if so set the rdonly flag and then
2354 * mark the partition as valid again.
2355 */
2356 if (!(es->s_state & cpu_to_le16(EXT3_VALID_FS)) &&
2357 (sbi->s_mount_state & EXT3_VALID_FS))
2358 es->s_state = cpu_to_le16(sbi->s_mount_state);
2359
030703e4
JK		 2360 /*
2361 * We have to unlock super so that we can wait for
2362 * transactions.
2363 */
2364 unlock_super(sb);
1da177e4 2365 ext3_mark_recovery_complete(sb, es);
030703e4 2366 lock_super(sb);
1da177e4
LT		 2367 } else {
2368 __le32 ret;
2369 if ((ret = EXT3_HAS_RO_COMPAT_FEATURE(sb,
2370 ~EXT3_FEATURE_RO_COMPAT_SUPP))) {
2371 printk(KERN_WARNING "EXT3-fs: %s: couldn't "
2372 "remount RDWR because of unsupported "
2373 "optional features (%x).\n",
2374 sb->s_id, le32_to_cpu(ret));
08c6a96f
JK		 2375 err = -EROFS;
2376 goto restore_opts;
1da177e4 2377 }
ea9a05a1
ES		 2378
2379 /*
2380 * If we have an unprocessed orphan list hanging
2381 * around from a previously readonly bdev mount,
2382 * require a full umount/remount for now.
2383 */
2384 if (es->s_last_orphan) {
2385 printk(KERN_WARNING "EXT3-fs: %s: couldn't "
2386 "remount RDWR because of unprocessed "
2387 "orphan inode list. Please "
2388 "umount/remount instead.\n",
2389 sb->s_id);
2390 err = -EINVAL;
2391 goto restore_opts;
2392 }
2393
1da177e4
LT		 2394 /*
2395 * Mounting a RDONLY partition read-write, so reread
2396 * and store the current valid flag. (It may have
2397 * been changed by e2fsck since we originally mounted
2398 * the partition.)
2399 */
2400 ext3_clear_journal_err(sb, es);
2401 sbi->s_mount_state = le16_to_cpu(es->s_state);
a4e4de36 2402 if ((err = ext3_group_extend(sb, es, n_blocks_count)))
08c6a96f 2403 goto restore_opts;
1da177e4
LT		 2404 if (!ext3_setup_super (sb, es, 0))
2405 sb->s_flags &= ~MS_RDONLY;
2406 }
2407 }
08c6a96f
JK		 2408#ifdef CONFIG_QUOTA
2409 /* Release old quota file names */
2410 for (i = 0; i < MAXQUOTAS; i++)
2411 if (old_opts.s_qf_names[i] &&
2412 old_opts.s_qf_names[i] != sbi->s_qf_names[i])
2413 kfree(old_opts.s_qf_names[i]);
2414#endif
1da177e4 2415 return 0;
08c6a96f
JK		 2416restore_opts:
2417 sb->s_flags = old_sb_flags;
2418 sbi->s_mount_opt = old_opts.s_mount_opt;
2419 sbi->s_resuid = old_opts.s_resuid;
2420 sbi->s_resgid = old_opts.s_resgid;
2421 sbi->s_commit_interval = old_opts.s_commit_interval;
2422#ifdef CONFIG_QUOTA
2423 sbi->s_jquota_fmt = old_opts.s_jquota_fmt;
2424 for (i = 0; i < MAXQUOTAS; i++) {
2425 if (sbi->s_qf_names[i] &&
2426 old_opts.s_qf_names[i] != sbi->s_qf_names[i])
2427 kfree(sbi->s_qf_names[i]);
2428 sbi->s_qf_names[i] = old_opts.s_qf_names[i];
2429 }
2430#endif
2431 return err;
1da177e4
LT		 2432}
2433
726c3342 2434static int ext3_statfs (struct dentry * dentry, struct kstatfs * buf)
1da177e4 2435{
726c3342 2436 struct super_block *sb = dentry->d_sb;
09fe316a
AT		 2437 struct ext3_sb_info *sbi = EXT3_SB(sb);
2438 struct ext3_super_block *es = sbi->s_es;
50ee0a32 2439 u64 fsid;
1da177e4 2440
a71ce8c6
BP		 2441 if (test_opt(sb, MINIX_DF)) {
2442 sbi->s_overhead_last = 0;
2443 } else if (sbi->s_blocks_last != le32_to_cpu(es->s_blocks_count)) {
2444 unsigned long ngroups = sbi->s_groups_count, i;
2445 ext3_fsblk_t overhead = 0;
1da177e4
LT		 2446 smp_rmb();
2447
2448 /*
a71ce8c6
BP		 2449 * Compute the overhead (FS structures). This is constant
2450 * for a given filesystem unless the number of block groups
2451 * changes so we cache the previous value until it does.
1da177e4
LT		 2452 */
2453
2454 /*
2455 * All of the blocks before first_data_block are
2456 * overhead
2457 */
2458 overhead = le32_to_cpu(es->s_first_data_block);
2459
2460 /*
2461 * Add the overhead attributed to the superblock and
2462 * block group descriptors. If the sparse superblocks
2463 * feature is turned on, then not all groups have this.
2464 */
2465 for (i = 0; i < ngroups; i++) {
2466 overhead += ext3_bg_has_super(sb, i) +
2467 ext3_bg_num_gdb(sb, i);
2468 cond_resched();
2469 }
2470
2471 /*
2472 * Every block group has an inode bitmap, a block
2473 * bitmap, and an inode table.
2474 */
a71ce8c6
BP		 2475 overhead += ngroups * (2 + sbi->s_itb_per_group);
2476 sbi->s_overhead_last = overhead;
2477 smp_wmb();
2478 sbi->s_blocks_last = le32_to_cpu(es->s_blocks_count);
1da177e4
LT		 2479 }
2480
2481 buf->f_type = EXT3_SUPER_MAGIC;
2482 buf->f_bsize = sb->s_blocksize;
a71ce8c6 2483 buf->f_blocks = le32_to_cpu(es->s_blocks_count) - sbi->s_overhead_last;
52d9f3b4 2484 buf->f_bfree = percpu_counter_sum_positive(&sbi->s_freeblocks_counter);
a71ce8c6 2485 es->s_free_blocks_count = cpu_to_le32(buf->f_bfree);
1da177e4
LT		 2486 buf->f_bavail = buf->f_bfree - le32_to_cpu(es->s_r_blocks_count);
2487 if (buf->f_bfree < le32_to_cpu(es->s_r_blocks_count))
2488 buf->f_bavail = 0;
2489 buf->f_files = le32_to_cpu(es->s_inodes_count);
52d9f3b4 2490 buf->f_ffree = percpu_counter_sum_positive(&sbi->s_freeinodes_counter);
a71ce8c6 2491 es->s_free_inodes_count = cpu_to_le32(buf->f_ffree);
1da177e4 2492 buf->f_namelen = EXT3_NAME_LEN;
50ee0a32
PE		 2493 fsid = le64_to_cpup((void *)es->s_uuid) ^
2494 le64_to_cpup((void *)es->s_uuid + sizeof(u64));
2495 buf->f_fsid.val[0] = fsid & 0xFFFFFFFFUL;
2496 buf->f_fsid.val[1] = (fsid >> 32) & 0xFFFFFFFFUL;
1da177e4
LT		 2497 return 0;
2498}
2499
2500/* Helper function for writing quotas on sync - we need to start transaction before quota file
2501 * is locked for write. Otherwise the are possible deadlocks:
2502 * Process 1 Process 2
2503 * ext3_create() quota_sync()
2504 * journal_start() write_dquot()
d3be915f
IM		 2505 * DQUOT_INIT() down(dqio_mutex)
2506 * down(dqio_mutex) journal_start()
1da177e4
LT		 2507 *
2508 */
2509
2510#ifdef CONFIG_QUOTA
2511
2512static inline struct inode *dquot_to_inode(struct dquot *dquot)
2513{
2514 return sb_dqopt(dquot->dq_sb)->files[dquot->dq_type];
2515}
2516
2517static int ext3_dquot_initialize(struct inode *inode, int type)
2518{
2519 handle_t *handle;
2520 int ret, err;
2521
2522 /* We may create quota structure so we need to reserve enough blocks */
1f54587b 2523 handle = ext3_journal_start(inode, 2*EXT3_QUOTA_INIT_BLOCKS(inode->i_sb));
1da177e4
LT		 2524 if (IS_ERR(handle))
2525 return PTR_ERR(handle);
2526 ret = dquot_initialize(inode, type);
2527 err = ext3_journal_stop(handle);
2528 if (!ret)
2529 ret = err;
2530 return ret;
2531}
2532
2533static int ext3_dquot_drop(struct inode *inode)
2534{
2535 handle_t *handle;
2536 int ret, err;
2537
2538 /* We may delete quota structure so we need to reserve enough blocks */
1f54587b 2539 handle = ext3_journal_start(inode, 2*EXT3_QUOTA_DEL_BLOCKS(inode->i_sb));
1da177e4
LT		 2540 if (IS_ERR(handle))
2541 return PTR_ERR(handle);
2542 ret = dquot_drop(inode);
2543 err = ext3_journal_stop(handle);
2544 if (!ret)
2545 ret = err;
2546 return ret;
2547}
2548
2549static int ext3_write_dquot(struct dquot *dquot)
2550{
2551 int ret, err;
2552 handle_t *handle;
2553 struct inode *inode;
2554
2555 inode = dquot_to_inode(dquot);
2556 handle = ext3_journal_start(inode,
1f54587b 2557 EXT3_QUOTA_TRANS_BLOCKS(dquot->dq_sb));
1da177e4
LT		 2558 if (IS_ERR(handle))
2559 return PTR_ERR(handle);
2560 ret = dquot_commit(dquot);
2561 err = ext3_journal_stop(handle);
2562 if (!ret)
2563 ret = err;
2564 return ret;
2565}
2566
2567static int ext3_acquire_dquot(struct dquot *dquot)
2568{
2569 int ret, err;
2570 handle_t *handle;
2571
2572 handle = ext3_journal_start(dquot_to_inode(dquot),
1f54587b 2573 EXT3_QUOTA_INIT_BLOCKS(dquot->dq_sb));
1da177e4
LT		 2574 if (IS_ERR(handle))
2575 return PTR_ERR(handle);
2576 ret = dquot_acquire(dquot);
2577 err = ext3_journal_stop(handle);
2578 if (!ret)
2579 ret = err;
2580 return ret;
2581}
2582
2583static int ext3_release_dquot(struct dquot *dquot)
2584{
2585 int ret, err;
2586 handle_t *handle;
2587
2588 handle = ext3_journal_start(dquot_to_inode(dquot),
1f54587b 2589 EXT3_QUOTA_DEL_BLOCKS(dquot->dq_sb));
9c3013e9
JK		 2590 if (IS_ERR(handle)) {
2591 /* Release dquot anyway to avoid endless cycle in dqput() */
2592 dquot_release(dquot);
1da177e4 2593 return PTR_ERR(handle);
9c3013e9 2594 }
1da177e4
LT		 2595 ret = dquot_release(dquot);
2596 err = ext3_journal_stop(handle);
2597 if (!ret)
2598 ret = err;
2599 return ret;
2600}
2601
2602static int ext3_mark_dquot_dirty(struct dquot *dquot)
2603{
2604 /* Are we journalling quotas? */
2605 if (EXT3_SB(dquot->dq_sb)->s_qf_names[USRQUOTA] ||
2606 EXT3_SB(dquot->dq_sb)->s_qf_names[GRPQUOTA]) {
2607 dquot_mark_dquot_dirty(dquot);
2608 return ext3_write_dquot(dquot);
2609 } else {
2610 return dquot_mark_dquot_dirty(dquot);
2611 }
2612}
2613
2614static int ext3_write_info(struct super_block *sb, int type)
2615{
2616 int ret, err;
2617 handle_t *handle;
2618
2619 /* Data block + inode block */
2620 handle = ext3_journal_start(sb->s_root->d_inode, 2);
2621 if (IS_ERR(handle))
2622 return PTR_ERR(handle);
2623 ret = dquot_commit_info(sb, type);
2624 err = ext3_journal_stop(handle);
2625 if (!ret)
2626 ret = err;
2627 return ret;
2628}
2629
2630/*
2631 * Turn on quotas during mount time - we need to find
2632 * the quota file and such...
2633 */
2634static int ext3_quota_on_mount(struct super_block *sb, int type)
2635{
84de856e
CH		 2636 return vfs_quota_on_mount(sb, EXT3_SB(sb)->s_qf_names[type],
2637 EXT3_SB(sb)->s_jquota_fmt, type);
1da177e4
LT		 2638}
2639
2640/*
2641 * Standard function to be called on quota_on
2642 */
2643static int ext3_quota_on(struct super_block *sb, int type, int format_id,
2644 char *path)
2645{
2646 int err;
2647 struct nameidata nd;
2648
1f54587b
JK		 2649 if (!test_opt(sb, QUOTA))
2650 return -EINVAL;
1da177e4
LT		 2651 /* Not journalling quota? */
2652 if (!EXT3_SB(sb)->s_qf_names[USRQUOTA] &&
2653 !EXT3_SB(sb)->s_qf_names[GRPQUOTA])
2654 return vfs_quota_on(sb, type, format_id, path);
2655 err = path_lookup(path, LOOKUP_FOLLOW, &nd);
2656 if (err)
2657 return err;
2658 /* Quotafile not on the same filesystem? */
2659 if (nd.mnt->mnt_sb != sb) {
2660 path_release(&nd);
2661 return -EXDEV;
2662 }
2663 /* Quotafile not of fs root? */
2664 if (nd.dentry->d_parent->d_inode != sb->s_root->d_inode)
2665 printk(KERN_WARNING
2666 "EXT3-fs: Quota file not on filesystem root. "
2667 "Journalled quota will not work.\n");
2668 path_release(&nd);
2669 return vfs_quota_on(sb, type, format_id, path);
2670}
2671
2672/* Read data from quotafile - avoid pagecache and such because we cannot afford
2673 * acquiring the locks... As quota files are never truncated and quota code
2674 * itself serializes the operations (and noone else should touch the files)
2675 * we don't have to be afraid of races */
2676static ssize_t ext3_quota_read(struct super_block *sb, int type, char *data,
2677 size_t len, loff_t off)
2678{
2679 struct inode *inode = sb_dqopt(sb)->files[type];
2680 sector_t blk = off >> EXT3_BLOCK_SIZE_BITS(sb);
2681 int err = 0;
2682 int offset = off & (sb->s_blocksize - 1);
2683 int tocopy;
2684 size_t toread;
2685 struct buffer_head *bh;
2686 loff_t i_size = i_size_read(inode);
2687
2688 if (off > i_size)
2689 return 0;
2690 if (off+len > i_size)
2691 len = i_size-off;
2692 toread = len;
2693 while (toread > 0) {
2694 tocopy = sb->s_blocksize - offset < toread ?
2695 sb->s_blocksize - offset : toread;
2696 bh = ext3_bread(NULL, inode, blk, 0, &err);
2697 if (err)
2698 return err;
2699 if (!bh) /* A hole? */
2700 memset(data, 0, tocopy);
2701 else
2702 memcpy(data, bh->b_data+offset, tocopy);
2703 brelse(bh);
2704 offset = 0;
2705 toread -= tocopy;
2706 data += tocopy;
2707 blk++;
2708 }
2709 return len;
2710}
2711
2712/* Write to quotafile (we know the transaction is already started and has
2713 * enough credits) */
2714static ssize_t ext3_quota_write(struct super_block *sb, int type,
2715 const char *data, size_t len, loff_t off)
2716{
2717 struct inode *inode = sb_dqopt(sb)->files[type];
2718 sector_t blk = off >> EXT3_BLOCK_SIZE_BITS(sb);
2719 int err = 0;
2720 int offset = off & (sb->s_blocksize - 1);
2721 int tocopy;
2722 int journal_quota = EXT3_SB(sb)->s_qf_names[type] != NULL;
2723 size_t towrite = len;
2724 struct buffer_head *bh;
2725 handle_t *handle = journal_current_handle();
2726
9c3013e9
JK		 2727 if (!handle) {
2728 printk(KERN_WARNING "EXT3-fs: Quota write (off=%Lu, len=%Lu)"
2729 " cancelled because transaction is not started.\n",
2730 (unsigned long long)off, (unsigned long long)len);
2731 return -EIO;
2732 }
5c81a419 2733 mutex_lock_nested(&inode->i_mutex, I_MUTEX_QUOTA);
1da177e4
LT		 2734 while (towrite > 0) {
2735 tocopy = sb->s_blocksize - offset < towrite ?
2736 sb->s_blocksize - offset : towrite;
2737 bh = ext3_bread(handle, inode, blk, 1, &err);
2738 if (!bh)
2739 goto out;
2740 if (journal_quota) {
2741 err = ext3_journal_get_write_access(handle, bh);
2742 if (err) {
2743 brelse(bh);
2744 goto out;
2745 }
2746 }
2747 lock_buffer(bh);
2748 memcpy(bh->b_data+offset, data, tocopy);
2749 flush_dcache_page(bh->b_page);
2750 unlock_buffer(bh);
2751 if (journal_quota)
2752 err = ext3_journal_dirty_metadata(handle, bh);
2753 else {
2754 /* Always do at least ordered writes for quotas */
2755 err = ext3_journal_dirty_data(handle, bh);
2756 mark_buffer_dirty(bh);
2757 }
2758 brelse(bh);
2759 if (err)
2760 goto out;
2761 offset = 0;
2762 towrite -= tocopy;
2763 data += tocopy;
2764 blk++;
2765 }
2766out:
2767 if (len == towrite)
2768 return err;
2769 if (inode->i_size < off+len-towrite) {
2770 i_size_write(inode, off+len-towrite);
2771 EXT3_I(inode)->i_disksize = inode->i_size;
2772 }
2773 inode->i_version++;
2774 inode->i_mtime = inode->i_ctime = CURRENT_TIME;
2775 ext3_mark_inode_dirty(handle, inode);
1b1dcc1b 2776 mutex_unlock(&inode->i_mutex);
1da177e4
LT		 2777 return len - towrite;
2778}
2779
2780#endif
2781
454e2398
DH		 2782static int ext3_get_sb(struct file_system_type *fs_type,
2783 int flags, const char *dev_name, void *data, struct vfsmount *mnt)
1da177e4 2784{
454e2398 2785 return get_sb_bdev(fs_type, flags, dev_name, data, ext3_fill_super, mnt);
1da177e4
LT		 2786}
2787
2788static struct file_system_type ext3_fs_type = {
2789 .owner = THIS_MODULE,
2790 .name = "ext3",
2791 .get_sb = ext3_get_sb,
2792 .kill_sb = kill_block_super,
2793 .fs_flags = FS_REQUIRES_DEV,
2794};
2795
2796static int __init init_ext3_fs(void)
2797{
2798 int err = init_ext3_xattr();
2799 if (err)
2800 return err;
2801 err = init_inodecache();
2802 if (err)
2803 goto out1;
2804 err = register_filesystem(&ext3_fs_type);
2805 if (err)
2806 goto out;
2807 return 0;
2808out:
2809 destroy_inodecache();
2810out1:
e9ad5620 2811 exit_ext3_xattr();
1da177e4
LT		 2812 return err;
2813}
2814
2815static void __exit exit_ext3_fs(void)
2816{
2817 unregister_filesystem(&ext3_fs_type);
2818 destroy_inodecache();
2819 exit_ext3_xattr();
2820}
2821
2822MODULE_AUTHOR("Remy Card, Stephen Tweedie, Andrew Morton, Andreas Dilger, Theodore Ts'o and others");
2823MODULE_DESCRIPTION("Second Extended Filesystem with journaling extensions");
2824MODULE_LICENSE("GPL");
2825module_init(init_ext3_fs)
2826module_exit(exit_ext3_fs)
Clone of davem's net-next-2.6 tree