?  search:
summary | log | graphiclog1 | graphiclog2 | commit | commitdiff | tree | refs | edit | fork
blame | history | raw | HEAD
Merge git://git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux-2.6
[wrt350n-kernel.git] / fs / ext3 / super.c
blobd11115401f8da4c979490372d7bbccd91494f298
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
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>
32 #include <linux/exportfs.h>
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>
38 #include <linux/seq_file.h>
39 #include <linux/log2.h>
40
41 #include <asm/uaccess.h>
42
43 #include "xattr.h"
44 #include "acl.h"
45 #include "namei.h"
46
47 static int ext3_load_journal(struct super_block *, struct ext3_super_block *,
48 unsigned long journal_devnum);
49 static int ext3_create_journal(struct super_block *, struct ext3_super_block *,
50 unsigned int);
51 static void ext3_commit_super (struct super_block * sb,
52 struct ext3_super_block * es,
53 int sync);
54 static void ext3_mark_recovery_complete(struct super_block * sb,
55 struct ext3_super_block * es);
56 static void ext3_clear_journal_err(struct super_block * sb,
57 struct ext3_super_block * es);
58 static int ext3_sync_fs(struct super_block *sb, int wait);
59 static const char *ext3_decode_error(struct super_block * sb, int errno,
60 char nbuf[16]);
61 static int ext3_remount (struct super_block * sb, int * flags, char * data);
62 static int ext3_statfs (struct dentry * dentry, struct kstatfs * buf);
63 static void ext3_unlockfs(struct super_block *sb);
64 static void ext3_write_super (struct super_block * sb);
65 static void ext3_write_super_lockfs(struct super_block *sb);
66
67 /*
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
73 * appropriate.
74 */
75 handle_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
95 /*
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
99 * appropriate.
100 */
101 int __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
118 void 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
154 static 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
164 if (!test_opt (sb, ERRORS_CONT)) {
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 }
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);
176 if (test_opt(sb, ERRORS_PANIC))
177 panic("EXT3-fs (device %s): panic forced after error\n",
178 sb->s_id);
179 }
180
181 void 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
195 static 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
231 void __ext3_std_error (struct super_block * sb, const char * function,
232 int errno)
233 {
234 char nbuf[16];
235 const char *errstr;
236
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);
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
261 void 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
287 void 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
300 void 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 */
328 static 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
338 fail:
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 */
347 static int ext3_blkdev_put(struct block_device *bdev)
348 {
349 bd_release(bdev);
350 return blkdev_put(bdev);
351 }
352
353 static 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
366 static 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
371 static void dump_orphan_list(struct super_block *sb, struct ext3_sb_info *sbi)
372 {
373 struct list_head *l;
374
375 printk(KERN_ERR "sb orphan head is %d\n",
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 " "
382 "inode %s:%lu at %p: mode %o, nlink %d, next %d\n",
383 inode->i_sb->s_id, inode->i_ino, inode,
384 inode->i_mode, inode->i_nlink,
385 NEXT_ORPHAN(inode));
386 }
387 }
388
389 static 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
425 invalidate_bdev(sb->s_bdev);
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);
433 invalidate_bdev(sbi->journal_bdev);
434 ext3_blkdev_remove(sbi);
435 }
436 sb->s_fs_info = NULL;
437 kfree(sbi);
438 return;
439 }
440
441 static struct kmem_cache *ext3_inode_cachep;
442
443 /*
444 * Called inside transaction, so use GFP_NOFS
445 */
446 static struct inode *ext3_alloc_inode(struct super_block *sb)
447 {
448 struct ext3_inode_info *ei;
449
450 ei = kmem_cache_alloc(ext3_inode_cachep, GFP_NOFS);
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
462 static void ext3_destroy_inode(struct inode *inode)
463 {
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 }
472 kmem_cache_free(ext3_inode_cachep, EXT3_I(inode));
473 }
474
475 static void init_once(struct kmem_cache * cachep, void *foo)
476 {
477 struct ext3_inode_info *ei = (struct ext3_inode_info *) foo;
478
479 INIT_LIST_HEAD(&ei->i_orphan);
480 #ifdef CONFIG_EXT3_FS_XATTR
481 init_rwsem(&ei->xattr_sem);
482 #endif
483 mutex_init(&ei->truncate_mutex);
484 inode_init_once(&ei->vfs_inode);
485 }
486
487 static int init_inodecache(void)
488 {
489 ext3_inode_cachep = kmem_cache_create("ext3_inode_cache",
490 sizeof(struct ext3_inode_info),
491 0, (SLAB_RECLAIM_ACCOUNT|
492 SLAB_MEM_SPREAD),
493 init_once);
494 if (ext3_inode_cachep == NULL)
495 return -ENOMEM;
496 return 0;
497 }
498
499 static void destroy_inodecache(void)
500 {
501 kmem_cache_destroy(ext3_inode_cachep);
502 }
503
504 static 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
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 }
518 #endif
519 ext3_discard_reservation(inode);
520 EXT3_I(inode)->i_block_alloc_info = NULL;
521 if (unlikely(rsv))
522 kfree(rsv);
523 }
524
525 static inline void ext3_show_quota_options(struct seq_file *seq, struct super_block *sb)
526 {
527 #if defined(CONFIG_QUOTA)
528 struct ext3_sb_info *sbi = EXT3_SB(sb);
529
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
546 }
547
548 /*
549 * Show an option if
550 * - it's set to a non-default value OR
551 * - if the per-sb default is different from the global default
552 */
553 static int ext3_show_options(struct seq_file *seq, struct vfsmount *vfs)
554 {
555 struct super_block *sb = vfs->mnt_sb;
556 struct ext3_sb_info *sbi = EXT3_SB(sb);
557 struct ext3_super_block *es = sbi->s_es;
558 unsigned long def_mount_opts;
559
560 def_mount_opts = le32_to_cpu(es->s_default_mount_opts);
561
562 if (sbi->s_sb_block != 1)
563 seq_printf(seq, ",sb=%lu", sbi->s_sb_block);
564 if (test_opt(sb, MINIX_DF))
565 seq_puts(seq, ",minixdf");
566 if (test_opt(sb, GRPID))
567 seq_puts(seq, ",grpid");
568 if (!test_opt(sb, GRPID) && (def_mount_opts & EXT3_DEFM_BSDGROUPS))
569 seq_puts(seq, ",nogrpid");
570 if (sbi->s_resuid != EXT3_DEF_RESUID ||
571 le16_to_cpu(es->s_def_resuid) != EXT3_DEF_RESUID) {
572 seq_printf(seq, ",resuid=%u", sbi->s_resuid);
573 }
574 if (sbi->s_resgid != EXT3_DEF_RESGID ||
575 le16_to_cpu(es->s_def_resgid) != EXT3_DEF_RESGID) {
576 seq_printf(seq, ",resgid=%u", sbi->s_resgid);
577 }
578 if (test_opt(sb, ERRORS_RO)) {
579 int def_errors = le16_to_cpu(es->s_errors);
580
581 if (def_errors == EXT3_ERRORS_PANIC ||
582 def_errors == EXT3_ERRORS_CONTINUE) {
583 seq_puts(seq, ",errors=remount-ro");
584 }
585 }
586 if (test_opt(sb, ERRORS_CONT))
587 seq_puts(seq, ",errors=continue");
588 if (test_opt(sb, ERRORS_PANIC))
589 seq_puts(seq, ",errors=panic");
590 if (test_opt(sb, NO_UID32))
591 seq_puts(seq, ",nouid32");
592 if (test_opt(sb, DEBUG))
593 seq_puts(seq, ",debug");
594 if (test_opt(sb, OLDALLOC))
595 seq_puts(seq, ",oldalloc");
596 #ifdef CONFIG_EXT3_FS_XATTR
597 if (test_opt(sb, XATTR_USER))
598 seq_puts(seq, ",user_xattr");
599 if (!test_opt(sb, XATTR_USER) &&
600 (def_mount_opts & EXT3_DEFM_XATTR_USER)) {
601 seq_puts(seq, ",nouser_xattr");
602 }
603 #endif
604 #ifdef CONFIG_EXT3_FS_POSIX_ACL
605 if (test_opt(sb, POSIX_ACL))
606 seq_puts(seq, ",acl");
607 if (!test_opt(sb, POSIX_ACL) && (def_mount_opts & EXT3_DEFM_ACL))
608 seq_puts(seq, ",noacl");
609 #endif
610 if (!test_opt(sb, RESERVATION))
611 seq_puts(seq, ",noreservation");
612 if (sbi->s_commit_interval) {
613 seq_printf(seq, ",commit=%u",
614 (unsigned) (sbi->s_commit_interval / HZ));
615 }
616 if (test_opt(sb, BARRIER))
617 seq_puts(seq, ",barrier=1");
618 if (test_opt(sb, NOBH))
619 seq_puts(seq, ",nobh");
620
621 if (test_opt(sb, DATA_FLAGS) == EXT3_MOUNT_JOURNAL_DATA)
622 seq_puts(seq, ",data=journal");
623 else if (test_opt(sb, DATA_FLAGS) == EXT3_MOUNT_ORDERED_DATA)
624 seq_puts(seq, ",data=ordered");
625 else if (test_opt(sb, DATA_FLAGS) == EXT3_MOUNT_WRITEBACK_DATA)
626 seq_puts(seq, ",data=writeback");
627
628 ext3_show_quota_options(seq, sb);
629
630 return 0;
631 }
632
633
634 static struct inode *ext3_nfs_get_inode(struct super_block *sb,
635 u64 ino, u32 generation)
636 {
637 struct inode *inode;
638
639 if (ino < EXT3_FIRST_INO(sb) && ino != EXT3_ROOT_INO)
640 return ERR_PTR(-ESTALE);
641 if (ino > le32_to_cpu(EXT3_SB(sb)->s_es->s_inodes_count))
642 return ERR_PTR(-ESTALE);
643
644 /* iget isn't really right if the inode is currently unallocated!!
645 *
646 * ext3_read_inode will return a bad_inode if the inode had been
647 * deleted, so we should be safe.
648 *
649 * Currently we don't know the generation for parent directory, so
650 * a generation of 0 means "accept any"
651 */
652 inode = ext3_iget(sb, ino);
653 if (IS_ERR(inode))
654 return ERR_CAST(inode);
655 if (generation && inode->i_generation != generation) {
656 iput(inode);
657 return ERR_PTR(-ESTALE);
658 }
659
660 return inode;
661 }
662
663 static struct dentry *ext3_fh_to_dentry(struct super_block *sb, struct fid *fid,
664 int fh_len, int fh_type)
665 {
666 return generic_fh_to_dentry(sb, fid, fh_len, fh_type,
667 ext3_nfs_get_inode);
668 }
669
670 static struct dentry *ext3_fh_to_parent(struct super_block *sb, struct fid *fid,
671 int fh_len, int fh_type)
672 {
673 return generic_fh_to_parent(sb, fid, fh_len, fh_type,
674 ext3_nfs_get_inode);
675 }
676
677 #ifdef CONFIG_QUOTA
678 #define QTYPE2NAME(t) ((t)==USRQUOTA?"user":"group")
679 #define QTYPE2MOPT(on, t) ((t)==USRQUOTA?((on)##USRJQUOTA):((on)##GRPJQUOTA))
680
681 static int ext3_dquot_initialize(struct inode *inode, int type);
682 static int ext3_dquot_drop(struct inode *inode);
683 static int ext3_write_dquot(struct dquot *dquot);
684 static int ext3_acquire_dquot(struct dquot *dquot);
685 static int ext3_release_dquot(struct dquot *dquot);
686 static int ext3_mark_dquot_dirty(struct dquot *dquot);
687 static int ext3_write_info(struct super_block *sb, int type);
688 static int ext3_quota_on(struct super_block *sb, int type, int format_id, char *path);
689 static int ext3_quota_on_mount(struct super_block *sb, int type);
690 static ssize_t ext3_quota_read(struct super_block *sb, int type, char *data,
691 size_t len, loff_t off);
692 static ssize_t ext3_quota_write(struct super_block *sb, int type,
693 const char *data, size_t len, loff_t off);
694
695 static struct dquot_operations ext3_quota_operations = {
696 .initialize = ext3_dquot_initialize,
697 .drop = ext3_dquot_drop,
698 .alloc_space = dquot_alloc_space,
699 .alloc_inode = dquot_alloc_inode,
700 .free_space = dquot_free_space,
701 .free_inode = dquot_free_inode,
702 .transfer = dquot_transfer,
703 .write_dquot = ext3_write_dquot,
704 .acquire_dquot = ext3_acquire_dquot,
705 .release_dquot = ext3_release_dquot,
706 .mark_dirty = ext3_mark_dquot_dirty,
707 .write_info = ext3_write_info
708 };
709
710 static struct quotactl_ops ext3_qctl_operations = {
711 .quota_on = ext3_quota_on,
712 .quota_off = vfs_quota_off,
713 .quota_sync = vfs_quota_sync,
714 .get_info = vfs_get_dqinfo,
715 .set_info = vfs_set_dqinfo,
716 .get_dqblk = vfs_get_dqblk,
717 .set_dqblk = vfs_set_dqblk
718 };
719 #endif
720
721 static const struct super_operations ext3_sops = {
722 .alloc_inode = ext3_alloc_inode,
723 .destroy_inode = ext3_destroy_inode,
724 .write_inode = ext3_write_inode,
725 .dirty_inode = ext3_dirty_inode,
726 .delete_inode = ext3_delete_inode,
727 .put_super = ext3_put_super,
728 .write_super = ext3_write_super,
729 .sync_fs = ext3_sync_fs,
730 .write_super_lockfs = ext3_write_super_lockfs,
731 .unlockfs = ext3_unlockfs,
732 .statfs = ext3_statfs,
733 .remount_fs = ext3_remount,
734 .clear_inode = ext3_clear_inode,
735 .show_options = ext3_show_options,
736 #ifdef CONFIG_QUOTA
737 .quota_read = ext3_quota_read,
738 .quota_write = ext3_quota_write,
739 #endif
740 };
741
742 static const struct export_operations ext3_export_ops = {
743 .fh_to_dentry = ext3_fh_to_dentry,
744 .fh_to_parent = ext3_fh_to_parent,
745 .get_parent = ext3_get_parent,
746 };
747
748 enum {
749 Opt_bsd_df, Opt_minix_df, Opt_grpid, Opt_nogrpid,
750 Opt_resgid, Opt_resuid, Opt_sb, Opt_err_cont, Opt_err_panic, Opt_err_ro,
751 Opt_nouid32, Opt_nocheck, Opt_debug, Opt_oldalloc, Opt_orlov,
752 Opt_user_xattr, Opt_nouser_xattr, Opt_acl, Opt_noacl,
753 Opt_reservation, Opt_noreservation, Opt_noload, Opt_nobh, Opt_bh,
754 Opt_commit, Opt_journal_update, Opt_journal_inum, Opt_journal_dev,
755 Opt_abort, Opt_data_journal, Opt_data_ordered, Opt_data_writeback,
756 Opt_usrjquota, Opt_grpjquota, Opt_offusrjquota, Opt_offgrpjquota,
757 Opt_jqfmt_vfsold, Opt_jqfmt_vfsv0, Opt_quota, Opt_noquota,
758 Opt_ignore, Opt_barrier, Opt_err, Opt_resize, Opt_usrquota,
759 Opt_grpquota
760 };
761
762 static match_table_t tokens = {
763 {Opt_bsd_df, "bsddf"},
764 {Opt_minix_df, "minixdf"},
765 {Opt_grpid, "grpid"},
766 {Opt_grpid, "bsdgroups"},
767 {Opt_nogrpid, "nogrpid"},
768 {Opt_nogrpid, "sysvgroups"},
769 {Opt_resgid, "resgid=%u"},
770 {Opt_resuid, "resuid=%u"},
771 {Opt_sb, "sb=%u"},
772 {Opt_err_cont, "errors=continue"},
773 {Opt_err_panic, "errors=panic"},
774 {Opt_err_ro, "errors=remount-ro"},
775 {Opt_nouid32, "nouid32"},
776 {Opt_nocheck, "nocheck"},
777 {Opt_nocheck, "check=none"},
778 {Opt_debug, "debug"},
779 {Opt_oldalloc, "oldalloc"},
780 {Opt_orlov, "orlov"},
781 {Opt_user_xattr, "user_xattr"},
782 {Opt_nouser_xattr, "nouser_xattr"},
783 {Opt_acl, "acl"},
784 {Opt_noacl, "noacl"},
785 {Opt_reservation, "reservation"},
786 {Opt_noreservation, "noreservation"},
787 {Opt_noload, "noload"},
788 {Opt_nobh, "nobh"},
789 {Opt_bh, "bh"},
790 {Opt_commit, "commit=%u"},
791 {Opt_journal_update, "journal=update"},
792 {Opt_journal_inum, "journal=%u"},
793 {Opt_journal_dev, "journal_dev=%u"},
794 {Opt_abort, "abort"},
795 {Opt_data_journal, "data=journal"},
796 {Opt_data_ordered, "data=ordered"},
797 {Opt_data_writeback, "data=writeback"},
798 {Opt_offusrjquota, "usrjquota="},
799 {Opt_usrjquota, "usrjquota=%s"},
800 {Opt_offgrpjquota, "grpjquota="},
801 {Opt_grpjquota, "grpjquota=%s"},
802 {Opt_jqfmt_vfsold, "jqfmt=vfsold"},
803 {Opt_jqfmt_vfsv0, "jqfmt=vfsv0"},
804 {Opt_grpquota, "grpquota"},
805 {Opt_noquota, "noquota"},
806 {Opt_quota, "quota"},
807 {Opt_usrquota, "usrquota"},
808 {Opt_barrier, "barrier=%u"},
809 <<<<<<< HEAD:fs/ext3/super.c
810 {Opt_err, NULL},
811 =======
812 >>>>>>> 264e3e889d86e552b4191d69bb60f4f3b383135a:fs/ext3/super.c
813 {Opt_resize, "resize"},
814 <<<<<<< HEAD:fs/ext3/super.c
815 =======
816 {Opt_err, NULL},
817 >>>>>>> 264e3e889d86e552b4191d69bb60f4f3b383135a:fs/ext3/super.c
818 };
819
820 static ext3_fsblk_t get_sb_block(void **data)
821 {
822 ext3_fsblk_t sb_block;
823 char *options = (char *) *data;
824
825 if (!options || strncmp(options, "sb=", 3) != 0)
826 return 1; /* Default location */
827 options += 3;
828 /*todo: use simple_strtoll with >32bit ext3 */
829 sb_block = simple_strtoul(options, &options, 0);
830 if (*options && *options != ',') {
831 printk("EXT3-fs: Invalid sb specification: %s\n",
832 (char *) *data);
833 return 1;
834 }
835 if (*options == ',')
836 options++;
837 *data = (void *) options;
838 return sb_block;
839 }
840
841 static int parse_options (char *options, struct super_block *sb,
842 unsigned int *inum, unsigned long *journal_devnum,
843 ext3_fsblk_t *n_blocks_count, int is_remount)
844 {
845 struct ext3_sb_info *sbi = EXT3_SB(sb);
846 char * p;
847 substring_t args[MAX_OPT_ARGS];
848 int data_opt = 0;
849 int option;
850 #ifdef CONFIG_QUOTA
851 int qtype;
852 char *qname;
853 #endif
854
855 if (!options)
856 return 1;
857
858 while ((p = strsep (&options, ",")) != NULL) {
859 int token;
860 if (!*p)
861 continue;
862
863 token = match_token(p, tokens, args);
864 switch (token) {
865 case Opt_bsd_df:
866 clear_opt (sbi->s_mount_opt, MINIX_DF);
867 break;
868 case Opt_minix_df:
869 set_opt (sbi->s_mount_opt, MINIX_DF);
870 break;
871 case Opt_grpid:
872 set_opt (sbi->s_mount_opt, GRPID);
873 break;
874 case Opt_nogrpid:
875 clear_opt (sbi->s_mount_opt, GRPID);
876 break;
877 case Opt_resuid:
878 if (match_int(&args[0], &option))
879 return 0;
880 sbi->s_resuid = option;
881 break;
882 case Opt_resgid:
883 if (match_int(&args[0], &option))
884 return 0;
885 sbi->s_resgid = option;
886 break;
887 case Opt_sb:
888 /* handled by get_sb_block() instead of here */
889 /* *sb_block = match_int(&args[0]); */
890 break;
891 case Opt_err_panic:
892 clear_opt (sbi->s_mount_opt, ERRORS_CONT);
893 clear_opt (sbi->s_mount_opt, ERRORS_RO);
894 set_opt (sbi->s_mount_opt, ERRORS_PANIC);
895 break;
896 case Opt_err_ro:
897 clear_opt (sbi->s_mount_opt, ERRORS_CONT);
898 clear_opt (sbi->s_mount_opt, ERRORS_PANIC);
899 set_opt (sbi->s_mount_opt, ERRORS_RO);
900 break;
901 case Opt_err_cont:
902 clear_opt (sbi->s_mount_opt, ERRORS_RO);
903 clear_opt (sbi->s_mount_opt, ERRORS_PANIC);
904 set_opt (sbi->s_mount_opt, ERRORS_CONT);
905 break;
906 case Opt_nouid32:
907 set_opt (sbi->s_mount_opt, NO_UID32);
908 break;
909 case Opt_nocheck:
910 clear_opt (sbi->s_mount_opt, CHECK);
911 break;
912 case Opt_debug:
913 set_opt (sbi->s_mount_opt, DEBUG);
914 break;
915 case Opt_oldalloc:
916 set_opt (sbi->s_mount_opt, OLDALLOC);
917 break;
918 case Opt_orlov:
919 clear_opt (sbi->s_mount_opt, OLDALLOC);
920 break;
921 #ifdef CONFIG_EXT3_FS_XATTR
922 case Opt_user_xattr:
923 set_opt (sbi->s_mount_opt, XATTR_USER);
924 break;
925 case Opt_nouser_xattr:
926 clear_opt (sbi->s_mount_opt, XATTR_USER);
927 break;
928 #else
929 case Opt_user_xattr:
930 case Opt_nouser_xattr:
931 printk("EXT3 (no)user_xattr options not supported\n");
932 break;
933 #endif
934 #ifdef CONFIG_EXT3_FS_POSIX_ACL
935 case Opt_acl:
936 set_opt(sbi->s_mount_opt, POSIX_ACL);
937 break;
938 case Opt_noacl:
939 clear_opt(sbi->s_mount_opt, POSIX_ACL);
940 break;
941 #else
942 case Opt_acl:
943 case Opt_noacl:
944 printk("EXT3 (no)acl options not supported\n");
945 break;
946 #endif
947 case Opt_reservation:
948 set_opt(sbi->s_mount_opt, RESERVATION);
949 break;
950 case Opt_noreservation:
951 clear_opt(sbi->s_mount_opt, RESERVATION);
952 break;
953 case Opt_journal_update:
954 /* @@@ FIXME */
955 /* Eventually we will want to be able to create
956 a journal file here. For now, only allow the
957 user to specify an existing inode to be the
958 journal file. */
959 if (is_remount) {
960 printk(KERN_ERR "EXT3-fs: cannot specify "
961 "journal on remount\n");
962 return 0;
963 }
964 set_opt (sbi->s_mount_opt, UPDATE_JOURNAL);
965 break;
966 case Opt_journal_inum:
967 if (is_remount) {
968 printk(KERN_ERR "EXT3-fs: cannot specify "
969 "journal on remount\n");
970 return 0;
971 }
972 if (match_int(&args[0], &option))
973 return 0;
974 *inum = option;
975 break;
976 case Opt_journal_dev:
977 if (is_remount) {
978 printk(KERN_ERR "EXT3-fs: cannot specify "
979 "journal on remount\n");
980 return 0;
981 }
982 if (match_int(&args[0], &option))
983 return 0;
984 *journal_devnum = option;
985 break;
986 case Opt_noload:
987 set_opt (sbi->s_mount_opt, NOLOAD);
988 break;
989 case Opt_commit:
990 if (match_int(&args[0], &option))
991 return 0;
992 if (option < 0)
993 return 0;
994 if (option == 0)
995 option = JBD_DEFAULT_MAX_COMMIT_AGE;
996 sbi->s_commit_interval = HZ * option;
997 break;
998 case Opt_data_journal:
999 data_opt = EXT3_MOUNT_JOURNAL_DATA;
1000 goto datacheck;
1001 case Opt_data_ordered:
1002 data_opt = EXT3_MOUNT_ORDERED_DATA;
1003 goto datacheck;
1004 case Opt_data_writeback:
1005 data_opt = EXT3_MOUNT_WRITEBACK_DATA;
1006 datacheck:
1007 if (is_remount) {
1008 if ((sbi->s_mount_opt & EXT3_MOUNT_DATA_FLAGS)
1009 != data_opt) {
1010 printk(KERN_ERR
1011 "EXT3-fs: cannot change data "
1012 "mode on remount\n");
1013 return 0;
1014 }
1015 } else {
1016 sbi->s_mount_opt &= ~EXT3_MOUNT_DATA_FLAGS;
1017 sbi->s_mount_opt |= data_opt;
1018 }
1019 break;
1020 #ifdef CONFIG_QUOTA
1021 case Opt_usrjquota:
1022 qtype = USRQUOTA;
1023 goto set_qf_name;
1024 case Opt_grpjquota:
1025 qtype = GRPQUOTA;
1026 set_qf_name:
1027 if (sb_any_quota_enabled(sb)) {
1028 printk(KERN_ERR
1029 "EXT3-fs: Cannot change journalled "
1030 "quota options when quota turned on.\n");
1031 return 0;
1032 }
1033 qname = match_strdup(&args[0]);
1034 if (!qname) {
1035 printk(KERN_ERR
1036 "EXT3-fs: not enough memory for "
1037 "storing quotafile name.\n");
1038 return 0;
1039 }
1040 if (sbi->s_qf_names[qtype] &&
1041 strcmp(sbi->s_qf_names[qtype], qname)) {
1042 printk(KERN_ERR
1043 "EXT3-fs: %s quota file already "
1044 "specified.\n", QTYPE2NAME(qtype));
1045 kfree(qname);
1046 return 0;
1047 }
1048 sbi->s_qf_names[qtype] = qname;
1049 if (strchr(sbi->s_qf_names[qtype], '/')) {
1050 printk(KERN_ERR
1051 "EXT3-fs: quotafile must be on "
1052 "filesystem root.\n");
1053 kfree(sbi->s_qf_names[qtype]);
1054 sbi->s_qf_names[qtype] = NULL;
1055 return 0;
1056 }
1057 set_opt(sbi->s_mount_opt, QUOTA);
1058 break;
1059 case Opt_offusrjquota:
1060 qtype = USRQUOTA;
1061 goto clear_qf_name;
1062 case Opt_offgrpjquota:
1063 qtype = GRPQUOTA;
1064 clear_qf_name:
1065 if (sb_any_quota_enabled(sb)) {
1066 printk(KERN_ERR "EXT3-fs: Cannot change "
1067 "journalled quota options when "
1068 "quota turned on.\n");
1069 return 0;
1070 }
1071 /*
1072 * The space will be released later when all options
1073 * are confirmed to be correct
1074 */
1075 sbi->s_qf_names[qtype] = NULL;
1076 break;
1077 case Opt_jqfmt_vfsold:
1078 sbi->s_jquota_fmt = QFMT_VFS_OLD;
1079 break;
1080 case Opt_jqfmt_vfsv0:
1081 sbi->s_jquota_fmt = QFMT_VFS_V0;
1082 break;
1083 case Opt_quota:
1084 case Opt_usrquota:
1085 set_opt(sbi->s_mount_opt, QUOTA);
1086 set_opt(sbi->s_mount_opt, USRQUOTA);
1087 break;
1088 case Opt_grpquota:
1089 set_opt(sbi->s_mount_opt, QUOTA);
1090 set_opt(sbi->s_mount_opt, GRPQUOTA);
1091 break;
1092 case Opt_noquota:
1093 if (sb_any_quota_enabled(sb)) {
1094 printk(KERN_ERR "EXT3-fs: Cannot change quota "
1095 "options when quota turned on.\n");
1096 return 0;
1097 }
1098 clear_opt(sbi->s_mount_opt, QUOTA);
1099 clear_opt(sbi->s_mount_opt, USRQUOTA);
1100 clear_opt(sbi->s_mount_opt, GRPQUOTA);
1101 break;
1102 #else
1103 case Opt_quota:
1104 case Opt_usrquota:
1105 case Opt_grpquota:
1106 case Opt_usrjquota:
1107 case Opt_grpjquota:
1108 case Opt_offusrjquota:
1109 case Opt_offgrpjquota:
1110 case Opt_jqfmt_vfsold:
1111 case Opt_jqfmt_vfsv0:
1112 printk(KERN_ERR
1113 "EXT3-fs: journalled quota options not "
1114 "supported.\n");
1115 break;
1116 case Opt_noquota:
1117 break;
1118 #endif
1119 case Opt_abort:
1120 set_opt(sbi->s_mount_opt, ABORT);
1121 break;
1122 case Opt_barrier:
1123 if (match_int(&args[0], &option))
1124 return 0;
1125 if (option)
1126 set_opt(sbi->s_mount_opt, BARRIER);
1127 else
1128 clear_opt(sbi->s_mount_opt, BARRIER);
1129 break;
1130 case Opt_ignore:
1131 break;
1132 case Opt_resize:
1133 if (!is_remount) {
1134 printk("EXT3-fs: resize option only available "
1135 "for remount\n");
1136 return 0;
1137 }
1138 if (match_int(&args[0], &option) != 0)
1139 return 0;
1140 *n_blocks_count = option;
1141 break;
1142 case Opt_nobh:
1143 set_opt(sbi->s_mount_opt, NOBH);
1144 break;
1145 case Opt_bh:
1146 clear_opt(sbi->s_mount_opt, NOBH);
1147 break;
1148 default:
1149 printk (KERN_ERR
1150 "EXT3-fs: Unrecognized mount option \"%s\" "
1151 "or missing value\n", p);
1152 return 0;
1153 }
1154 }
1155 #ifdef CONFIG_QUOTA
1156 if (sbi->s_qf_names[USRQUOTA] || sbi->s_qf_names[GRPQUOTA]) {
1157 if ((sbi->s_mount_opt & EXT3_MOUNT_USRQUOTA) &&
1158 sbi->s_qf_names[USRQUOTA])
1159 clear_opt(sbi->s_mount_opt, USRQUOTA);
1160
1161 if ((sbi->s_mount_opt & EXT3_MOUNT_GRPQUOTA) &&
1162 sbi->s_qf_names[GRPQUOTA])
1163 clear_opt(sbi->s_mount_opt, GRPQUOTA);
1164
1165 if ((sbi->s_qf_names[USRQUOTA] &&
1166 (sbi->s_mount_opt & EXT3_MOUNT_GRPQUOTA)) ||
1167 (sbi->s_qf_names[GRPQUOTA] &&
1168 (sbi->s_mount_opt & EXT3_MOUNT_USRQUOTA))) {
1169 printk(KERN_ERR "EXT3-fs: old and new quota "
1170 "format mixing.\n");
1171 return 0;
1172 }
1173
1174 if (!sbi->s_jquota_fmt) {
1175 printk(KERN_ERR "EXT3-fs: journalled quota format "
1176 "not specified.\n");
1177 return 0;
1178 }
1179 } else {
1180 if (sbi->s_jquota_fmt) {
1181 printk(KERN_ERR "EXT3-fs: journalled quota format "
1182 "specified with no journalling "
1183 "enabled.\n");
1184 return 0;
1185 }
1186 }
1187 #endif
1188 return 1;
1189 }
1190
1191 static int ext3_setup_super(struct super_block *sb, struct ext3_super_block *es,
1192 int read_only)
1193 {
1194 struct ext3_sb_info *sbi = EXT3_SB(sb);
1195 int res = 0;
1196
1197 if (le32_to_cpu(es->s_rev_level) > EXT3_MAX_SUPP_REV) {
1198 printk (KERN_ERR "EXT3-fs warning: revision level too high, "
1199 "forcing read-only mode\n");
1200 res = MS_RDONLY;
1201 }
1202 if (read_only)
1203 return res;
1204 if (!(sbi->s_mount_state & EXT3_VALID_FS))
1205 printk (KERN_WARNING "EXT3-fs warning: mounting unchecked fs, "
1206 "running e2fsck is recommended\n");
1207 else if ((sbi->s_mount_state & EXT3_ERROR_FS))
1208 printk (KERN_WARNING
1209 "EXT3-fs warning: mounting fs with errors, "
1210 "running e2fsck is recommended\n");
1211 else if ((__s16) le16_to_cpu(es->s_max_mnt_count) >= 0 &&
1212 le16_to_cpu(es->s_mnt_count) >=
1213 (unsigned short) (__s16) le16_to_cpu(es->s_max_mnt_count))
1214 printk (KERN_WARNING
1215 "EXT3-fs warning: maximal mount count reached, "
1216 "running e2fsck is recommended\n");
1217 else if (le32_to_cpu(es->s_checkinterval) &&
1218 (le32_to_cpu(es->s_lastcheck) +
1219 le32_to_cpu(es->s_checkinterval) <= get_seconds()))
1220 printk (KERN_WARNING
1221 "EXT3-fs warning: checktime reached, "
1222 "running e2fsck is recommended\n");
1223 #if 0
1224 /* @@@ We _will_ want to clear the valid bit if we find
1225 inconsistencies, to force a fsck at reboot. But for
1226 a plain journaled filesystem we can keep it set as
1227 valid forever! :) */
1228 es->s_state = cpu_to_le16(le16_to_cpu(es->s_state) & ~EXT3_VALID_FS);
1229 #endif
1230 if (!(__s16) le16_to_cpu(es->s_max_mnt_count))
1231 es->s_max_mnt_count = cpu_to_le16(EXT3_DFL_MAX_MNT_COUNT);
1232 le16_add_cpu(&es->s_mnt_count, 1);
1233 es->s_mtime = cpu_to_le32(get_seconds());
1234 ext3_update_dynamic_rev(sb);
1235 EXT3_SET_INCOMPAT_FEATURE(sb, EXT3_FEATURE_INCOMPAT_RECOVER);
1236
1237 ext3_commit_super(sb, es, 1);
1238 if (test_opt(sb, DEBUG))
1239 printk(KERN_INFO "[EXT3 FS bs=%lu, gc=%lu, "
1240 "bpg=%lu, ipg=%lu, mo=%04lx]\n",
1241 sb->s_blocksize,
1242 sbi->s_groups_count,
1243 EXT3_BLOCKS_PER_GROUP(sb),
1244 EXT3_INODES_PER_GROUP(sb),
1245 sbi->s_mount_opt);
1246
1247 printk(KERN_INFO "EXT3 FS on %s, ", sb->s_id);
1248 if (EXT3_SB(sb)->s_journal->j_inode == NULL) {
1249 char b[BDEVNAME_SIZE];
1250
1251 printk("external journal on %s\n",
1252 bdevname(EXT3_SB(sb)->s_journal->j_dev, b));
1253 } else {
1254 printk("internal journal\n");
1255 }
1256 return res;
1257 }
1258
1259 /* Called at mount-time, super-block is locked */
1260 static int ext3_check_descriptors(struct super_block *sb)
1261 {
1262 struct ext3_sb_info *sbi = EXT3_SB(sb);
1263 ext3_fsblk_t first_block = le32_to_cpu(sbi->s_es->s_first_data_block);
1264 ext3_fsblk_t last_block;
1265 int i;
1266
1267 ext3_debug ("Checking group descriptors");
1268
1269 for (i = 0; i < sbi->s_groups_count; i++) {
1270 struct ext3_group_desc *gdp = ext3_get_group_desc(sb, i, NULL);
1271
1272 if (i == sbi->s_groups_count - 1)
1273 last_block = le32_to_cpu(sbi->s_es->s_blocks_count) - 1;
1274 else
1275 last_block = first_block +
1276 (EXT3_BLOCKS_PER_GROUP(sb) - 1);
1277
1278 if (le32_to_cpu(gdp->bg_block_bitmap) < first_block ||
1279 le32_to_cpu(gdp->bg_block_bitmap) > last_block)
1280 {
1281 ext3_error (sb, "ext3_check_descriptors",
1282 "Block bitmap for group %d"
1283 " not in group (block %lu)!",
1284 i, (unsigned long)
1285 le32_to_cpu(gdp->bg_block_bitmap));
1286 return 0;
1287 }
1288 if (le32_to_cpu(gdp->bg_inode_bitmap) < first_block ||
1289 le32_to_cpu(gdp->bg_inode_bitmap) > last_block)
1290 {
1291 ext3_error (sb, "ext3_check_descriptors",
1292 "Inode bitmap for group %d"
1293 " not in group (block %lu)!",
1294 i, (unsigned long)
1295 le32_to_cpu(gdp->bg_inode_bitmap));
1296 return 0;
1297 }
1298 if (le32_to_cpu(gdp->bg_inode_table) < first_block ||
1299 le32_to_cpu(gdp->bg_inode_table) + sbi->s_itb_per_group - 1 >
1300 last_block)
1301 {
1302 ext3_error (sb, "ext3_check_descriptors",
1303 "Inode table for group %d"
1304 " not in group (block %lu)!",
1305 i, (unsigned long)
1306 le32_to_cpu(gdp->bg_inode_table));
1307 return 0;
1308 }
1309 first_block += EXT3_BLOCKS_PER_GROUP(sb);
1310 }
1311
1312 sbi->s_es->s_free_blocks_count=cpu_to_le32(ext3_count_free_blocks(sb));
1313 sbi->s_es->s_free_inodes_count=cpu_to_le32(ext3_count_free_inodes(sb));
1314 return 1;
1315 }
1316
1317
1318 /* ext3_orphan_cleanup() walks a singly-linked list of inodes (starting at
1319 * the superblock) which were deleted from all directories, but held open by
1320 * a process at the time of a crash. We walk the list and try to delete these
1321 * inodes at recovery time (only with a read-write filesystem).
1322 *
1323 * In order to keep the orphan inode chain consistent during traversal (in
1324 * case of crash during recovery), we link each inode into the superblock
1325 * orphan list_head and handle it the same way as an inode deletion during
1326 * normal operation (which journals the operations for us).
1327 *
1328 * We only do an iget() and an iput() on each inode, which is very safe if we
1329 * accidentally point at an in-use or already deleted inode. The worst that
1330 * can happen in this case is that we get a "bit already cleared" message from
1331 * ext3_free_inode(). The only reason we would point at a wrong inode is if
1332 * e2fsck was run on this filesystem, and it must have already done the orphan
1333 * inode cleanup for us, so we can safely abort without any further action.
1334 */
1335 static void ext3_orphan_cleanup (struct super_block * sb,
1336 struct ext3_super_block * es)
1337 {
1338 unsigned int s_flags = sb->s_flags;
1339 int nr_orphans = 0, nr_truncates = 0;
1340 #ifdef CONFIG_QUOTA
1341 int i;
1342 #endif
1343 if (!es->s_last_orphan) {
1344 jbd_debug(4, "no orphan inodes to clean up\n");
1345 return;
1346 }
1347
1348 if (bdev_read_only(sb->s_bdev)) {
1349 printk(KERN_ERR "EXT3-fs: write access "
1350 "unavailable, skipping orphan cleanup.\n");
1351 return;
1352 }
1353
1354 if (EXT3_SB(sb)->s_mount_state & EXT3_ERROR_FS) {
1355 if (es->s_last_orphan)
1356 jbd_debug(1, "Errors on filesystem, "
1357 "clearing orphan list.\n");
1358 es->s_last_orphan = 0;
1359 jbd_debug(1, "Skipping orphan recovery on fs with errors.\n");
1360 return;
1361 }
1362
1363 if (s_flags & MS_RDONLY) {
1364 printk(KERN_INFO "EXT3-fs: %s: orphan cleanup on readonly fs\n",
1365 sb->s_id);
1366 sb->s_flags &= ~MS_RDONLY;
1367 }
1368 #ifdef CONFIG_QUOTA
1369 /* Needed for iput() to work correctly and not trash data */
1370 sb->s_flags |= MS_ACTIVE;
1371 /* Turn on quotas so that they are updated correctly */
1372 for (i = 0; i < MAXQUOTAS; i++) {
1373 if (EXT3_SB(sb)->s_qf_names[i]) {
1374 int ret = ext3_quota_on_mount(sb, i);
1375 if (ret < 0)
1376 printk(KERN_ERR
1377 "EXT3-fs: Cannot turn on journalled "
1378 "quota: error %d\n", ret);
1379 }
1380 }
1381 #endif
1382
1383 while (es->s_last_orphan) {
1384 struct inode *inode;
1385
1386 inode = ext3_orphan_get(sb, le32_to_cpu(es->s_last_orphan));
1387 if (IS_ERR(inode)) {
1388 es->s_last_orphan = 0;
1389 break;
1390 }
1391
1392 list_add(&EXT3_I(inode)->i_orphan, &EXT3_SB(sb)->s_orphan);
1393 DQUOT_INIT(inode);
1394 if (inode->i_nlink) {
1395 printk(KERN_DEBUG
1396 "%s: truncating inode %lu to %Ld bytes\n",
1397 __FUNCTION__, inode->i_ino, inode->i_size);
1398 jbd_debug(2, "truncating inode %lu to %Ld bytes\n",
1399 inode->i_ino, inode->i_size);
1400 ext3_truncate(inode);
1401 nr_truncates++;
1402 } else {
1403 printk(KERN_DEBUG
1404 "%s: deleting unreferenced inode %lu\n",
1405 __FUNCTION__, inode->i_ino);
1406 jbd_debug(2, "deleting unreferenced inode %lu\n",
1407 inode->i_ino);
1408 nr_orphans++;
1409 }
1410 iput(inode); /* The delete magic happens here! */
1411 }
1412
1413 #define PLURAL(x) (x), ((x)==1) ? "" : "s"
1414
1415 if (nr_orphans)
1416 printk(KERN_INFO "EXT3-fs: %s: %d orphan inode%s deleted\n",
1417 sb->s_id, PLURAL(nr_orphans));
1418 if (nr_truncates)
1419 printk(KERN_INFO "EXT3-fs: %s: %d truncate%s cleaned up\n",
1420 sb->s_id, PLURAL(nr_truncates));
1421 #ifdef CONFIG_QUOTA
1422 /* Turn quotas off */
1423 for (i = 0; i < MAXQUOTAS; i++) {
1424 if (sb_dqopt(sb)->files[i])
1425 vfs_quota_off(sb, i);
1426 }
1427 #endif
1428 sb->s_flags = s_flags; /* Restore MS_RDONLY status */
1429 }
1430
1431 /*
1432 * Maximal file size. There is a direct, and {,double-,triple-}indirect
1433 * block limit, and also a limit of (2^32 - 1) 512-byte sectors in i_blocks.
1434 * We need to be 1 filesystem block less than the 2^32 sector limit.
1435 */
1436 static loff_t ext3_max_size(int bits)
1437 {
1438 loff_t res = EXT3_NDIR_BLOCKS;
1439 int meta_blocks;
1440 loff_t upper_limit;
1441
1442 /* This is calculated to be the largest file size for a
1443 * dense, file such that the total number of
1444 * sectors in the file, including data and all indirect blocks,
1445 * does not exceed 2^32 -1
1446 * __u32 i_blocks representing the total number of
1447 * 512 bytes blocks of the file
1448 */
1449 upper_limit = (1LL << 32) - 1;
1450
1451 /* total blocks in file system block size */
1452 upper_limit >>= (bits - 9);
1453
1454
1455 /* indirect blocks */
1456 meta_blocks = 1;
1457 /* double indirect blocks */
1458 meta_blocks += 1 + (1LL << (bits-2));
1459 /* tripple indirect blocks */
1460 meta_blocks += 1 + (1LL << (bits-2)) + (1LL << (2*(bits-2)));
1461
1462 upper_limit -= meta_blocks;
1463 upper_limit <<= bits;
1464
1465 res += 1LL << (bits-2);
1466 res += 1LL << (2*(bits-2));
1467 res += 1LL << (3*(bits-2));
1468 res <<= bits;
1469 if (res > upper_limit)
1470 res = upper_limit;
1471
1472 if (res > MAX_LFS_FILESIZE)
1473 res = MAX_LFS_FILESIZE;
1474
1475 return res;
1476 }
1477
1478 static ext3_fsblk_t descriptor_loc(struct super_block *sb,
1479 ext3_fsblk_t logic_sb_block,
1480 int nr)
1481 {
1482 struct ext3_sb_info *sbi = EXT3_SB(sb);
1483 unsigned long bg, first_meta_bg;
1484 int has_super = 0;
1485
1486 first_meta_bg = le32_to_cpu(sbi->s_es->s_first_meta_bg);
1487
1488 if (!EXT3_HAS_INCOMPAT_FEATURE(sb, EXT3_FEATURE_INCOMPAT_META_BG) ||
1489 nr < first_meta_bg)
1490 return (logic_sb_block + nr + 1);
1491 bg = sbi->s_desc_per_block * nr;
1492 if (ext3_bg_has_super(sb, bg))
1493 has_super = 1;
1494 return (has_super + ext3_group_first_block_no(sb, bg));
1495 }
1496
1497
1498 static int ext3_fill_super (struct super_block *sb, void *data, int silent)
1499 {
1500 struct buffer_head * bh;
1501 struct ext3_super_block *es = NULL;
1502 struct ext3_sb_info *sbi;
1503 ext3_fsblk_t block;
1504 ext3_fsblk_t sb_block = get_sb_block(&data);
1505 ext3_fsblk_t logic_sb_block;
1506 unsigned long offset = 0;
1507 unsigned int journal_inum = 0;
1508 unsigned long journal_devnum = 0;
1509 unsigned long def_mount_opts;
1510 struct inode *root;
1511 int blocksize;
1512 int hblock;
1513 int db_count;
1514 int i;
1515 int needs_recovery;
1516 int ret = -EINVAL;
1517 __le32 features;
1518 int err;
1519
1520 sbi = kzalloc(sizeof(*sbi), GFP_KERNEL);
1521 if (!sbi)
1522 return -ENOMEM;
1523 sb->s_fs_info = sbi;
1524 sbi->s_mount_opt = 0;
1525 sbi->s_resuid = EXT3_DEF_RESUID;
1526 sbi->s_resgid = EXT3_DEF_RESGID;
1527 sbi->s_sb_block = sb_block;
1528
1529 unlock_kernel();
1530
1531 blocksize = sb_min_blocksize(sb, EXT3_MIN_BLOCK_SIZE);
1532 if (!blocksize) {
1533 printk(KERN_ERR "EXT3-fs: unable to set blocksize\n");
1534 goto out_fail;
1535 }
1536
1537 /*
1538 * The ext3 superblock will not be buffer aligned for other than 1kB
1539 * block sizes. We need to calculate the offset from buffer start.
1540 */
1541 if (blocksize != EXT3_MIN_BLOCK_SIZE) {
1542 logic_sb_block = (sb_block * EXT3_MIN_BLOCK_SIZE) / blocksize;
1543 offset = (sb_block * EXT3_MIN_BLOCK_SIZE) % blocksize;
1544 } else {
1545 logic_sb_block = sb_block;
1546 }
1547
1548 if (!(bh = sb_bread(sb, logic_sb_block))) {
1549 printk (KERN_ERR "EXT3-fs: unable to read superblock\n");
1550 goto out_fail;
1551 }
1552 /*
1553 * Note: s_es must be initialized as soon as possible because
1554 * some ext3 macro-instructions depend on its value
1555 */
1556 es = (struct ext3_super_block *) (((char *)bh->b_data) + offset);
1557 sbi->s_es = es;
1558 sb->s_magic = le16_to_cpu(es->s_magic);
1559 if (sb->s_magic != EXT3_SUPER_MAGIC)
1560 goto cantfind_ext3;
1561
1562 /* Set defaults before we parse the mount options */
1563 def_mount_opts = le32_to_cpu(es->s_default_mount_opts);
1564 if (def_mount_opts & EXT3_DEFM_DEBUG)
1565 set_opt(sbi->s_mount_opt, DEBUG);
1566 if (def_mount_opts & EXT3_DEFM_BSDGROUPS)
1567 set_opt(sbi->s_mount_opt, GRPID);
1568 if (def_mount_opts & EXT3_DEFM_UID16)
1569 set_opt(sbi->s_mount_opt, NO_UID32);
1570 #ifdef CONFIG_EXT3_FS_XATTR
1571 if (def_mount_opts & EXT3_DEFM_XATTR_USER)
1572 set_opt(sbi->s_mount_opt, XATTR_USER);
1573 #endif
1574 #ifdef CONFIG_EXT3_FS_POSIX_ACL
1575 if (def_mount_opts & EXT3_DEFM_ACL)
1576 set_opt(sbi->s_mount_opt, POSIX_ACL);
1577 #endif
1578 if ((def_mount_opts & EXT3_DEFM_JMODE) == EXT3_DEFM_JMODE_DATA)
1579 sbi->s_mount_opt |= EXT3_MOUNT_JOURNAL_DATA;
1580 else if ((def_mount_opts & EXT3_DEFM_JMODE) == EXT3_DEFM_JMODE_ORDERED)
1581 sbi->s_mount_opt |= EXT3_MOUNT_ORDERED_DATA;
1582 else if ((def_mount_opts & EXT3_DEFM_JMODE) == EXT3_DEFM_JMODE_WBACK)
1583 sbi->s_mount_opt |= EXT3_MOUNT_WRITEBACK_DATA;
1584
1585 if (le16_to_cpu(sbi->s_es->s_errors) == EXT3_ERRORS_PANIC)
1586 set_opt(sbi->s_mount_opt, ERRORS_PANIC);
1587 else if (le16_to_cpu(sbi->s_es->s_errors) == EXT3_ERRORS_CONTINUE)
1588 set_opt(sbi->s_mount_opt, ERRORS_CONT);
1589 else
1590 set_opt(sbi->s_mount_opt, ERRORS_RO);
1591
1592 sbi->s_resuid = le16_to_cpu(es->s_def_resuid);
1593 sbi->s_resgid = le16_to_cpu(es->s_def_resgid);
1594
1595 set_opt(sbi->s_mount_opt, RESERVATION);
1596
1597 if (!parse_options ((char *) data, sb, &journal_inum, &journal_devnum,
1598 NULL, 0))
1599 goto failed_mount;
1600
1601 sb->s_flags = (sb->s_flags & ~MS_POSIXACL) |
1602 ((sbi->s_mount_opt & EXT3_MOUNT_POSIX_ACL) ? MS_POSIXACL : 0);
1603
1604 if (le32_to_cpu(es->s_rev_level) == EXT3_GOOD_OLD_REV &&
1605 (EXT3_HAS_COMPAT_FEATURE(sb, ~0U) ||
1606 EXT3_HAS_RO_COMPAT_FEATURE(sb, ~0U) ||
1607 EXT3_HAS_INCOMPAT_FEATURE(sb, ~0U)))
1608 printk(KERN_WARNING
1609 "EXT3-fs warning: feature flags set on rev 0 fs, "
1610 "running e2fsck is recommended\n");
1611 /*
1612 * Check feature flags regardless of the revision level, since we
1613 * previously didn't change the revision level when setting the flags,
1614 * so there is a chance incompat flags are set on a rev 0 filesystem.
1615 */
1616 features = EXT3_HAS_INCOMPAT_FEATURE(sb, ~EXT3_FEATURE_INCOMPAT_SUPP);
1617 if (features) {
1618 printk(KERN_ERR "EXT3-fs: %s: couldn't mount because of "
1619 "unsupported optional features (%x).\n",
1620 sb->s_id, le32_to_cpu(features));
1621 goto failed_mount;
1622 }
1623 features = EXT3_HAS_RO_COMPAT_FEATURE(sb, ~EXT3_FEATURE_RO_COMPAT_SUPP);
1624 if (!(sb->s_flags & MS_RDONLY) && features) {
1625 printk(KERN_ERR "EXT3-fs: %s: couldn't mount RDWR because of "
1626 "unsupported optional features (%x).\n",
1627 sb->s_id, le32_to_cpu(features));
1628 goto failed_mount;
1629 }
1630 blocksize = BLOCK_SIZE << le32_to_cpu(es->s_log_block_size);
1631
1632 if (blocksize < EXT3_MIN_BLOCK_SIZE ||
1633 blocksize > EXT3_MAX_BLOCK_SIZE) {
1634 printk(KERN_ERR
1635 "EXT3-fs: Unsupported filesystem blocksize %d on %s.\n",
1636 blocksize, sb->s_id);
1637 goto failed_mount;
1638 }
1639
1640 hblock = bdev_hardsect_size(sb->s_bdev);
1641 if (sb->s_blocksize != blocksize) {
1642 /*
1643 * Make sure the blocksize for the filesystem is larger
1644 * than the hardware sectorsize for the machine.
1645 */
1646 if (blocksize < hblock) {
1647 printk(KERN_ERR "EXT3-fs: blocksize %d too small for "
1648 "device blocksize %d.\n", blocksize, hblock);
1649 goto failed_mount;
1650 }
1651
1652 brelse (bh);
1653 if (!sb_set_blocksize(sb, blocksize)) {
1654 printk(KERN_ERR "EXT3-fs: bad blocksize %d.\n",
1655 blocksize);
1656 goto out_fail;
1657 }
1658 logic_sb_block = (sb_block * EXT3_MIN_BLOCK_SIZE) / blocksize;
1659 offset = (sb_block * EXT3_MIN_BLOCK_SIZE) % blocksize;
1660 bh = sb_bread(sb, logic_sb_block);
1661 if (!bh) {
1662 printk(KERN_ERR
1663 "EXT3-fs: Can't read superblock on 2nd try.\n");
1664 goto failed_mount;
1665 }
1666 es = (struct ext3_super_block *)(((char *)bh->b_data) + offset);
1667 sbi->s_es = es;
1668 if (es->s_magic != cpu_to_le16(EXT3_SUPER_MAGIC)) {
1669 printk (KERN_ERR
1670 "EXT3-fs: Magic mismatch, very weird !\n");
1671 goto failed_mount;
1672 }
1673 }
1674
1675 sb->s_maxbytes = ext3_max_size(sb->s_blocksize_bits);
1676
1677 if (le32_to_cpu(es->s_rev_level) == EXT3_GOOD_OLD_REV) {
1678 sbi->s_inode_size = EXT3_GOOD_OLD_INODE_SIZE;
1679 sbi->s_first_ino = EXT3_GOOD_OLD_FIRST_INO;
1680 } else {
1681 sbi->s_inode_size = le16_to_cpu(es->s_inode_size);
1682 sbi->s_first_ino = le32_to_cpu(es->s_first_ino);
1683 if ((sbi->s_inode_size < EXT3_GOOD_OLD_INODE_SIZE) ||
1684 (!is_power_of_2(sbi->s_inode_size)) ||
1685 (sbi->s_inode_size > blocksize)) {
1686 printk (KERN_ERR
1687 "EXT3-fs: unsupported inode size: %d\n",
1688 sbi->s_inode_size);
1689 goto failed_mount;
1690 }
1691 }
1692 sbi->s_frag_size = EXT3_MIN_FRAG_SIZE <<
1693 le32_to_cpu(es->s_log_frag_size);
1694 if (blocksize != sbi->s_frag_size) {
1695 printk(KERN_ERR
1696 "EXT3-fs: fragsize %lu != blocksize %u (unsupported)\n",
1697 sbi->s_frag_size, blocksize);
1698 goto failed_mount;
1699 }
1700 sbi->s_frags_per_block = 1;
1701 sbi->s_blocks_per_group = le32_to_cpu(es->s_blocks_per_group);
1702 sbi->s_frags_per_group = le32_to_cpu(es->s_frags_per_group);
1703 sbi->s_inodes_per_group = le32_to_cpu(es->s_inodes_per_group);
1704 if (EXT3_INODE_SIZE(sb) == 0 || EXT3_INODES_PER_GROUP(sb) == 0)
1705 goto cantfind_ext3;
1706 sbi->s_inodes_per_block = blocksize / EXT3_INODE_SIZE(sb);
1707 if (sbi->s_inodes_per_block == 0)
1708 goto cantfind_ext3;
1709 sbi->s_itb_per_group = sbi->s_inodes_per_group /
1710 sbi->s_inodes_per_block;
1711 sbi->s_desc_per_block = blocksize / sizeof(struct ext3_group_desc);
1712 sbi->s_sbh = bh;
1713 sbi->s_mount_state = le16_to_cpu(es->s_state);
1714 sbi->s_addr_per_block_bits = ilog2(EXT3_ADDR_PER_BLOCK(sb));
1715 sbi->s_desc_per_block_bits = ilog2(EXT3_DESC_PER_BLOCK(sb));
1716 for (i=0; i < 4; i++)
1717 sbi->s_hash_seed[i] = le32_to_cpu(es->s_hash_seed[i]);
1718 sbi->s_def_hash_version = es->s_def_hash_version;
1719
1720 if (sbi->s_blocks_per_group > blocksize * 8) {
1721 printk (KERN_ERR
1722 "EXT3-fs: #blocks per group too big: %lu\n",
1723 sbi->s_blocks_per_group);
1724 goto failed_mount;
1725 }
1726 if (sbi->s_frags_per_group > blocksize * 8) {
1727 printk (KERN_ERR
1728 "EXT3-fs: #fragments per group too big: %lu\n",
1729 sbi->s_frags_per_group);
1730 goto failed_mount;
1731 }
1732 if (sbi->s_inodes_per_group > blocksize * 8) {
1733 printk (KERN_ERR
1734 "EXT3-fs: #inodes per group too big: %lu\n",
1735 sbi->s_inodes_per_group);
1736 goto failed_mount;
1737 }
1738
1739 if (le32_to_cpu(es->s_blocks_count) >
1740 (sector_t)(~0ULL) >> (sb->s_blocksize_bits - 9)) {
1741 printk(KERN_ERR "EXT3-fs: filesystem on %s:"
1742 " too large to mount safely\n", sb->s_id);
1743 if (sizeof(sector_t) < 8)
1744 printk(KERN_WARNING "EXT3-fs: CONFIG_LBD not "
1745 "enabled\n");
1746 goto failed_mount;
1747 }
1748
1749 if (EXT3_BLOCKS_PER_GROUP(sb) == 0)
1750 goto cantfind_ext3;
1751 sbi->s_groups_count = ((le32_to_cpu(es->s_blocks_count) -
1752 le32_to_cpu(es->s_first_data_block) - 1)
1753 / EXT3_BLOCKS_PER_GROUP(sb)) + 1;
1754 db_count = (sbi->s_groups_count + EXT3_DESC_PER_BLOCK(sb) - 1) /
1755 EXT3_DESC_PER_BLOCK(sb);
1756 sbi->s_group_desc = kmalloc(db_count * sizeof (struct buffer_head *),
1757 GFP_KERNEL);
1758 if (sbi->s_group_desc == NULL) {
1759 printk (KERN_ERR "EXT3-fs: not enough memory\n");
1760 goto failed_mount;
1761 }
1762
1763 bgl_lock_init(&sbi->s_blockgroup_lock);
1764
1765 for (i = 0; i < db_count; i++) {
1766 block = descriptor_loc(sb, logic_sb_block, i);
1767 sbi->s_group_desc[i] = sb_bread(sb, block);
1768 if (!sbi->s_group_desc[i]) {
1769 printk (KERN_ERR "EXT3-fs: "
1770 "can't read group descriptor %d\n", i);
1771 db_count = i;
1772 goto failed_mount2;
1773 }
1774 }
1775 if (!ext3_check_descriptors (sb)) {
1776 printk(KERN_ERR "EXT3-fs: group descriptors corrupted!\n");
1777 goto failed_mount2;
1778 }
1779 sbi->s_gdb_count = db_count;
1780 get_random_bytes(&sbi->s_next_generation, sizeof(u32));
1781 spin_lock_init(&sbi->s_next_gen_lock);
1782
1783 err = percpu_counter_init(&sbi->s_freeblocks_counter,
1784 ext3_count_free_blocks(sb));
1785 if (!err) {
1786 err = percpu_counter_init(&sbi->s_freeinodes_counter,
1787 ext3_count_free_inodes(sb));
1788 }
1789 if (!err) {
1790 err = percpu_counter_init(&sbi->s_dirs_counter,
1791 ext3_count_dirs(sb));
1792 }
1793 if (err) {
1794 printk(KERN_ERR "EXT3-fs: insufficient memory\n");
1795 goto failed_mount3;
1796 }
1797
1798 /* per fileystem reservation list head & lock */
1799 spin_lock_init(&sbi->s_rsv_window_lock);
1800 sbi->s_rsv_window_root = RB_ROOT;
1801 /* Add a single, static dummy reservation to the start of the
1802 * reservation window list --- it gives us a placeholder for
1803 * append-at-start-of-list which makes the allocation logic
1804 * _much_ simpler. */
1805 sbi->s_rsv_window_head.rsv_start = EXT3_RESERVE_WINDOW_NOT_ALLOCATED;
1806 sbi->s_rsv_window_head.rsv_end = EXT3_RESERVE_WINDOW_NOT_ALLOCATED;
1807 sbi->s_rsv_window_head.rsv_alloc_hit = 0;
1808 sbi->s_rsv_window_head.rsv_goal_size = 0;
1809 ext3_rsv_window_add(sb, &sbi->s_rsv_window_head);
1810
1811 /*
1812 * set up enough so that it can read an inode
1813 */
1814 sb->s_op = &ext3_sops;
1815 sb->s_export_op = &ext3_export_ops;
1816 sb->s_xattr = ext3_xattr_handlers;
1817 #ifdef CONFIG_QUOTA
1818 sb->s_qcop = &ext3_qctl_operations;
1819 sb->dq_op = &ext3_quota_operations;
1820 #endif
1821 INIT_LIST_HEAD(&sbi->s_orphan); /* unlinked but open files */
1822
1823 sb->s_root = NULL;
1824
1825 needs_recovery = (es->s_last_orphan != 0 ||
1826 EXT3_HAS_INCOMPAT_FEATURE(sb,
1827 EXT3_FEATURE_INCOMPAT_RECOVER));
1828
1829 /*
1830 * The first inode we look at is the journal inode. Don't try
1831 * root first: it may be modified in the journal!
1832 */
1833 if (!test_opt(sb, NOLOAD) &&
1834 EXT3_HAS_COMPAT_FEATURE(sb, EXT3_FEATURE_COMPAT_HAS_JOURNAL)) {
1835 if (ext3_load_journal(sb, es, journal_devnum))
1836 goto failed_mount3;
1837 } else if (journal_inum) {
1838 if (ext3_create_journal(sb, es, journal_inum))
1839 goto failed_mount3;
1840 } else {
1841 if (!silent)
1842 printk (KERN_ERR
1843 "ext3: No journal on filesystem on %s\n",
1844 sb->s_id);
1845 goto failed_mount3;
1846 }
1847
1848 /* We have now updated the journal if required, so we can
1849 * validate the data journaling mode. */
1850 switch (test_opt(sb, DATA_FLAGS)) {
1851 case 0:
1852 /* No mode set, assume a default based on the journal
1853 capabilities: ORDERED_DATA if the journal can
1854 cope, else JOURNAL_DATA */
1855 if (journal_check_available_features
1856 (sbi->s_journal, 0, 0, JFS_FEATURE_INCOMPAT_REVOKE))
1857 set_opt(sbi->s_mount_opt, ORDERED_DATA);
1858 else
1859 set_opt(sbi->s_mount_opt, JOURNAL_DATA);
1860 break;
1861
1862 case EXT3_MOUNT_ORDERED_DATA:
1863 case EXT3_MOUNT_WRITEBACK_DATA:
1864 if (!journal_check_available_features
1865 (sbi->s_journal, 0, 0, JFS_FEATURE_INCOMPAT_REVOKE)) {
1866 printk(KERN_ERR "EXT3-fs: Journal does not support "
1867 "requested data journaling mode\n");
1868 goto failed_mount4;
1869 }
1870 default:
1871 break;
1872 }
1873
1874 if (test_opt(sb, NOBH)) {
1875 if (!(test_opt(sb, DATA_FLAGS) == EXT3_MOUNT_WRITEBACK_DATA)) {
1876 printk(KERN_WARNING "EXT3-fs: Ignoring nobh option - "
1877 "its supported only with writeback mode\n");
1878 clear_opt(sbi->s_mount_opt, NOBH);
1879 }
1880 }
1881 /*
1882 * The journal_load will have done any necessary log recovery,
1883 * so we can safely mount the rest of the filesystem now.
1884 */
1885
1886 root = ext3_iget(sb, EXT3_ROOT_INO);
1887 if (IS_ERR(root)) {
1888 printk(KERN_ERR "EXT3-fs: get root inode failed\n");
1889 ret = PTR_ERR(root);
1890 goto failed_mount4;
1891 }
1892 if (!S_ISDIR(root->i_mode) || !root->i_blocks || !root->i_size) {
1893 iput(root);
1894 printk(KERN_ERR "EXT3-fs: corrupt root inode, run e2fsck\n");
1895 goto failed_mount4;
1896 }
1897 sb->s_root = d_alloc_root(root);
1898 if (!sb->s_root) {
1899 printk(KERN_ERR "EXT3-fs: get root dentry failed\n");
1900 iput(root);
1901 ret = -ENOMEM;
1902 goto failed_mount4;
1903 }
1904
1905 ext3_setup_super (sb, es, sb->s_flags & MS_RDONLY);
1906 /*
1907 * akpm: core read_super() calls in here with the superblock locked.
1908 * That deadlocks, because orphan cleanup needs to lock the superblock
1909 * in numerous places. Here we just pop the lock - it's relatively
1910 * harmless, because we are now ready to accept write_super() requests,
1911 * and aviro says that's the only reason for hanging onto the
1912 * superblock lock.
1913 */
1914 EXT3_SB(sb)->s_mount_state |= EXT3_ORPHAN_FS;
1915 ext3_orphan_cleanup(sb, es);
1916 EXT3_SB(sb)->s_mount_state &= ~EXT3_ORPHAN_FS;
1917 if (needs_recovery)
1918 printk (KERN_INFO "EXT3-fs: recovery complete.\n");
1919 ext3_mark_recovery_complete(sb, es);
1920 printk (KERN_INFO "EXT3-fs: mounted filesystem with %s data mode.\n",
1921 test_opt(sb,DATA_FLAGS) == EXT3_MOUNT_JOURNAL_DATA ? "journal":
1922 test_opt(sb,DATA_FLAGS) == EXT3_MOUNT_ORDERED_DATA ? "ordered":
1923 "writeback");
1924
1925 lock_kernel();
1926 return 0;
1927
1928 cantfind_ext3:
1929 if (!silent)
1930 printk(KERN_ERR "VFS: Can't find ext3 filesystem on dev %s.\n",
1931 sb->s_id);
1932 goto failed_mount;
1933
1934 failed_mount4:
1935 journal_destroy(sbi->s_journal);
1936 failed_mount3:
1937 percpu_counter_destroy(&sbi->s_freeblocks_counter);
1938 percpu_counter_destroy(&sbi->s_freeinodes_counter);
1939 percpu_counter_destroy(&sbi->s_dirs_counter);
1940 failed_mount2:
1941 for (i = 0; i < db_count; i++)
1942 brelse(sbi->s_group_desc[i]);
1943 kfree(sbi->s_group_desc);
1944 failed_mount:
1945 #ifdef CONFIG_QUOTA
1946 for (i = 0; i < MAXQUOTAS; i++)
1947 kfree(sbi->s_qf_names[i]);
1948 #endif
1949 ext3_blkdev_remove(sbi);
1950 brelse(bh);
1951 out_fail:
1952 sb->s_fs_info = NULL;
1953 kfree(sbi);
1954 lock_kernel();
1955 return ret;
1956 }
1957
1958 /*
1959 * Setup any per-fs journal parameters now. We'll do this both on
1960 * initial mount, once the journal has been initialised but before we've
1961 * done any recovery; and again on any subsequent remount.
1962 */
1963 static void ext3_init_journal_params(struct super_block *sb, journal_t *journal)
1964 {
1965 struct ext3_sb_info *sbi = EXT3_SB(sb);
1966
1967 if (sbi->s_commit_interval)
1968 journal->j_commit_interval = sbi->s_commit_interval;
1969 /* We could also set up an ext3-specific default for the commit
1970 * interval here, but for now we'll just fall back to the jbd
1971 * default. */
1972
1973 spin_lock(&journal->j_state_lock);
1974 if (test_opt(sb, BARRIER))
1975 journal->j_flags |= JFS_BARRIER;
1976 else
1977 journal->j_flags &= ~JFS_BARRIER;
1978 spin_unlock(&journal->j_state_lock);
1979 }
1980
1981 static journal_t *ext3_get_journal(struct super_block *sb,
1982 unsigned int journal_inum)
1983 {
1984 struct inode *journal_inode;
1985 journal_t *journal;
1986
1987 /* First, test for the existence of a valid inode on disk. Bad
1988 * things happen if we iget() an unused inode, as the subsequent
1989 * iput() will try to delete it. */
1990
1991 journal_inode = ext3_iget(sb, journal_inum);
1992 if (IS_ERR(journal_inode)) {
1993 printk(KERN_ERR "EXT3-fs: no journal found.\n");
1994 return NULL;
1995 }
1996 if (!journal_inode->i_nlink) {
1997 make_bad_inode(journal_inode);
1998 iput(journal_inode);
1999 printk(KERN_ERR "EXT3-fs: journal inode is deleted.\n");
2000 return NULL;
2001 }
2002
2003 jbd_debug(2, "Journal inode found at %p: %Ld bytes\n",
2004 journal_inode, journal_inode->i_size);
2005 if (!S_ISREG(journal_inode->i_mode)) {
2006 printk(KERN_ERR "EXT3-fs: invalid journal inode.\n");
2007 iput(journal_inode);
2008 return NULL;
2009 }
2010
2011 journal = journal_init_inode(journal_inode);
2012 if (!journal) {
2013 printk(KERN_ERR "EXT3-fs: Could not load journal inode\n");
2014 iput(journal_inode);
2015 return NULL;
2016 }
2017 journal->j_private = sb;
2018 ext3_init_journal_params(sb, journal);
2019 return journal;
2020 }
2021
2022 static journal_t *ext3_get_dev_journal(struct super_block *sb,
2023 dev_t j_dev)
2024 {
2025 struct buffer_head * bh;
2026 journal_t *journal;
2027 ext3_fsblk_t start;
2028 ext3_fsblk_t len;
2029 int hblock, blocksize;
2030 ext3_fsblk_t sb_block;
2031 unsigned long offset;
2032 struct ext3_super_block * es;
2033 struct block_device *bdev;
2034
2035 bdev = ext3_blkdev_get(j_dev);
2036 if (bdev == NULL)
2037 return NULL;
2038
2039 if (bd_claim(bdev, sb)) {
2040 printk(KERN_ERR
2041 "EXT3: failed to claim external journal device.\n");
2042 blkdev_put(bdev);
2043 return NULL;
2044 }
2045
2046 blocksize = sb->s_blocksize;
2047 hblock = bdev_hardsect_size(bdev);
2048 if (blocksize < hblock) {
2049 printk(KERN_ERR
2050 "EXT3-fs: blocksize too small for journal device.\n");
2051 goto out_bdev;
2052 }
2053
2054 sb_block = EXT3_MIN_BLOCK_SIZE / blocksize;
2055 offset = EXT3_MIN_BLOCK_SIZE % blocksize;
2056 set_blocksize(bdev, blocksize);
2057 if (!(bh = __bread(bdev, sb_block, blocksize))) {
2058 printk(KERN_ERR "EXT3-fs: couldn't read superblock of "
2059 "external journal\n");
2060 goto out_bdev;
2061 }
2062
2063 es = (struct ext3_super_block *) (((char *)bh->b_data) + offset);
2064 if ((le16_to_cpu(es->s_magic) != EXT3_SUPER_MAGIC) ||
2065 !(le32_to_cpu(es->s_feature_incompat) &
2066 EXT3_FEATURE_INCOMPAT_JOURNAL_DEV)) {
2067 printk(KERN_ERR "EXT3-fs: external journal has "
2068 "bad superblock\n");
2069 brelse(bh);
2070 goto out_bdev;
2071 }
2072
2073 if (memcmp(EXT3_SB(sb)->s_es->s_journal_uuid, es->s_uuid, 16)) {
2074 printk(KERN_ERR "EXT3-fs: journal UUID does not match\n");
2075 brelse(bh);
2076 goto out_bdev;
2077 }
2078
2079 len = le32_to_cpu(es->s_blocks_count);
2080 start = sb_block + 1;
2081 brelse(bh); /* we're done with the superblock */
2082
2083 journal = journal_init_dev(bdev, sb->s_bdev,
2084 start, len, blocksize);
2085 if (!journal) {
2086 printk(KERN_ERR "EXT3-fs: failed to create device journal\n");
2087 goto out_bdev;
2088 }
2089 journal->j_private = sb;
2090 ll_rw_block(READ, 1, &journal->j_sb_buffer);
2091 wait_on_buffer(journal->j_sb_buffer);
2092 if (!buffer_uptodate(journal->j_sb_buffer)) {
2093 printk(KERN_ERR "EXT3-fs: I/O error on journal device\n");
2094 goto out_journal;
2095 }
2096 if (be32_to_cpu(journal->j_superblock->s_nr_users) != 1) {
2097 printk(KERN_ERR "EXT3-fs: External journal has more than one "
2098 "user (unsupported) - %d\n",
2099 be32_to_cpu(journal->j_superblock->s_nr_users));
2100 goto out_journal;
2101 }
2102 EXT3_SB(sb)->journal_bdev = bdev;
2103 ext3_init_journal_params(sb, journal);
2104 return journal;
2105 out_journal:
2106 journal_destroy(journal);
2107 out_bdev:
2108 ext3_blkdev_put(bdev);
2109 return NULL;
2110 }
2111
2112 static int ext3_load_journal(struct super_block *sb,
2113 struct ext3_super_block *es,
2114 unsigned long journal_devnum)
2115 {
2116 journal_t *journal;
2117 unsigned int journal_inum = le32_to_cpu(es->s_journal_inum);
2118 dev_t journal_dev;
2119 int err = 0;
2120 int really_read_only;
2121
2122 if (journal_devnum &&
2123 journal_devnum != le32_to_cpu(es->s_journal_dev)) {
2124 printk(KERN_INFO "EXT3-fs: external journal device major/minor "
2125 "numbers have changed\n");
2126 journal_dev = new_decode_dev(journal_devnum);
2127 } else
2128 journal_dev = new_decode_dev(le32_to_cpu(es->s_journal_dev));
2129
2130 really_read_only = bdev_read_only(sb->s_bdev);
2131
2132 /*
2133 * Are we loading a blank journal or performing recovery after a
2134 * crash? For recovery, we need to check in advance whether we
2135 * can get read-write access to the device.
2136 */
2137
2138 if (EXT3_HAS_INCOMPAT_FEATURE(sb, EXT3_FEATURE_INCOMPAT_RECOVER)) {
2139 if (sb->s_flags & MS_RDONLY) {
2140 printk(KERN_INFO "EXT3-fs: INFO: recovery "
2141 "required on readonly filesystem.\n");
2142 if (really_read_only) {
2143 printk(KERN_ERR "EXT3-fs: write access "
2144 "unavailable, cannot proceed.\n");
2145 return -EROFS;
2146 }
2147 printk (KERN_INFO "EXT3-fs: write access will "
2148 "be enabled during recovery.\n");
2149 }
2150 }
2151
2152 if (journal_inum && journal_dev) {
2153 printk(KERN_ERR "EXT3-fs: filesystem has both journal "
2154 "and inode journals!\n");
2155 return -EINVAL;
2156 }
2157
2158 if (journal_inum) {
2159 if (!(journal = ext3_get_journal(sb, journal_inum)))
2160 return -EINVAL;
2161 } else {
2162 if (!(journal = ext3_get_dev_journal(sb, journal_dev)))
2163 return -EINVAL;
2164 }
2165
2166 if (!really_read_only && test_opt(sb, UPDATE_JOURNAL)) {
2167 err = journal_update_format(journal);
2168 if (err) {
2169 printk(KERN_ERR "EXT3-fs: error updating journal.\n");
2170 journal_destroy(journal);
2171 return err;
2172 }
2173 }
2174
2175 if (!EXT3_HAS_INCOMPAT_FEATURE(sb, EXT3_FEATURE_INCOMPAT_RECOVER))
2176 err = journal_wipe(journal, !really_read_only);
2177 if (!err)
2178 err = journal_load(journal);
2179
2180 if (err) {
2181 printk(KERN_ERR "EXT3-fs: error loading journal.\n");
2182 journal_destroy(journal);
2183 return err;
2184 }
2185
2186 EXT3_SB(sb)->s_journal = journal;
2187 ext3_clear_journal_err(sb, es);
2188
2189 if (journal_devnum &&
2190 journal_devnum != le32_to_cpu(es->s_journal_dev)) {
2191 es->s_journal_dev = cpu_to_le32(journal_devnum);
2192 sb->s_dirt = 1;
2193
2194 /* Make sure we flush the recovery flag to disk. */
2195 ext3_commit_super(sb, es, 1);
2196 }
2197
2198 return 0;
2199 }
2200
2201 static int ext3_create_journal(struct super_block * sb,
2202 struct ext3_super_block * es,
2203 unsigned int journal_inum)
2204 {
2205 journal_t *journal;
2206 int err;
2207
2208 if (sb->s_flags & MS_RDONLY) {
2209 printk(KERN_ERR "EXT3-fs: readonly filesystem when trying to "
2210 "create journal.\n");
2211 return -EROFS;
2212 }
2213
2214 journal = ext3_get_journal(sb, journal_inum);
2215 if (!journal)
2216 return -EINVAL;
2217
2218 printk(KERN_INFO "EXT3-fs: creating new journal on inode %u\n",
2219 journal_inum);
2220
2221 err = journal_create(journal);
2222 if (err) {
2223 printk(KERN_ERR "EXT3-fs: error creating journal.\n");
2224 journal_destroy(journal);
2225 return -EIO;
2226 }
2227
2228 EXT3_SB(sb)->s_journal = journal;
2229
2230 ext3_update_dynamic_rev(sb);
2231 EXT3_SET_INCOMPAT_FEATURE(sb, EXT3_FEATURE_INCOMPAT_RECOVER);
2232 EXT3_SET_COMPAT_FEATURE(sb, EXT3_FEATURE_COMPAT_HAS_JOURNAL);
2233
2234 es->s_journal_inum = cpu_to_le32(journal_inum);
2235 sb->s_dirt = 1;
2236
2237 /* Make sure we flush the recovery flag to disk. */
2238 ext3_commit_super(sb, es, 1);
2239
2240 return 0;
2241 }
2242
2243 static void ext3_commit_super (struct super_block * sb,
2244 struct ext3_super_block * es,
2245 int sync)
2246 {
2247 struct buffer_head *sbh = EXT3_SB(sb)->s_sbh;
2248
2249 if (!sbh)
2250 return;
2251 es->s_wtime = cpu_to_le32(get_seconds());
2252 es->s_free_blocks_count = cpu_to_le32(ext3_count_free_blocks(sb));
2253 es->s_free_inodes_count = cpu_to_le32(ext3_count_free_inodes(sb));
2254 BUFFER_TRACE(sbh, "marking dirty");
2255 mark_buffer_dirty(sbh);
2256 if (sync)
2257 sync_dirty_buffer(sbh);
2258 }
2259
2260
2261 /*
2262 * Have we just finished recovery? If so, and if we are mounting (or
2263 * remounting) the filesystem readonly, then we will end up with a
2264 * consistent fs on disk. Record that fact.
2265 */
2266 static void ext3_mark_recovery_complete(struct super_block * sb,
2267 struct ext3_super_block * es)
2268 {
2269 journal_t *journal = EXT3_SB(sb)->s_journal;
2270
2271 journal_lock_updates(journal);
2272 journal_flush(journal);
2273 lock_super(sb);
2274 if (EXT3_HAS_INCOMPAT_FEATURE(sb, EXT3_FEATURE_INCOMPAT_RECOVER) &&
2275 sb->s_flags & MS_RDONLY) {
2276 EXT3_CLEAR_INCOMPAT_FEATURE(sb, EXT3_FEATURE_INCOMPAT_RECOVER);
2277 sb->s_dirt = 0;
2278 ext3_commit_super(sb, es, 1);
2279 }
2280 unlock_super(sb);
2281 journal_unlock_updates(journal);
2282 }
2283
2284 /*
2285 * If we are mounting (or read-write remounting) a filesystem whose journal
2286 * has recorded an error from a previous lifetime, move that error to the
2287 * main filesystem now.
2288 */
2289 static void ext3_clear_journal_err(struct super_block * sb,
2290 struct ext3_super_block * es)
2291 {
2292 journal_t *journal;
2293 int j_errno;
2294 const char *errstr;
2295
2296 journal = EXT3_SB(sb)->s_journal;
2297
2298 /*
2299 * Now check for any error status which may have been recorded in the
2300 * journal by a prior ext3_error() or ext3_abort()
2301 */
2302
2303 j_errno = journal_errno(journal);
2304 if (j_errno) {
2305 char nbuf[16];
2306
2307 errstr = ext3_decode_error(sb, j_errno, nbuf);
2308 ext3_warning(sb, __FUNCTION__, "Filesystem error recorded "
2309 "from previous mount: %s", errstr);
2310 ext3_warning(sb, __FUNCTION__, "Marking fs in need of "
2311 "filesystem check.");
2312
2313 EXT3_SB(sb)->s_mount_state |= EXT3_ERROR_FS;
2314 es->s_state |= cpu_to_le16(EXT3_ERROR_FS);
2315 ext3_commit_super (sb, es, 1);
2316
2317 journal_clear_err(journal);
2318 }
2319 }
2320
2321 /*
2322 * Force the running and committing transactions to commit,
2323 * and wait on the commit.
2324 */
2325 int ext3_force_commit(struct super_block *sb)
2326 {
2327 journal_t *journal;
2328 int ret;
2329
2330 if (sb->s_flags & MS_RDONLY)
2331 return 0;
2332
2333 journal = EXT3_SB(sb)->s_journal;
2334 sb->s_dirt = 0;
2335 ret = ext3_journal_force_commit(journal);
2336 return ret;
2337 }
2338
2339 /*
2340 * Ext3 always journals updates to the superblock itself, so we don't
2341 * have to propagate any other updates to the superblock on disk at this
2342 * point. Just start an async writeback to get the buffers on their way
2343 * to the disk.
2344 *
2345 * This implicitly triggers the writebehind on sync().
2346 */
2347
2348 static void ext3_write_super (struct super_block * sb)
2349 {
2350 if (mutex_trylock(&sb->s_lock) != 0)
2351 BUG();
2352 sb->s_dirt = 0;
2353 }
2354
2355 static int ext3_sync_fs(struct super_block *sb, int wait)
2356 {
2357 tid_t target;
2358
2359 sb->s_dirt = 0;
2360 if (journal_start_commit(EXT3_SB(sb)->s_journal, &target)) {
2361 if (wait)
2362 log_wait_commit(EXT3_SB(sb)->s_journal, target);
2363 }
2364 return 0;
2365 }
2366
2367 /*
2368 * LVM calls this function before a (read-only) snapshot is created. This
2369 * gives us a chance to flush the journal completely and mark the fs clean.
2370 */
2371 static void ext3_write_super_lockfs(struct super_block *sb)
2372 {
2373 sb->s_dirt = 0;
2374
2375 if (!(sb->s_flags & MS_RDONLY)) {
2376 journal_t *journal = EXT3_SB(sb)->s_journal;
2377
2378 /* Now we set up the journal barrier. */
2379 journal_lock_updates(journal);
2380 journal_flush(journal);
2381
2382 /* Journal blocked and flushed, clear needs_recovery flag. */
2383 EXT3_CLEAR_INCOMPAT_FEATURE(sb, EXT3_FEATURE_INCOMPAT_RECOVER);
2384 ext3_commit_super(sb, EXT3_SB(sb)->s_es, 1);
2385 }
2386 }
2387
2388 /*
2389 * Called by LVM after the snapshot is done. We need to reset the RECOVER
2390 * flag here, even though the filesystem is not technically dirty yet.
2391 */
2392 static void ext3_unlockfs(struct super_block *sb)
2393 {
2394 if (!(sb->s_flags & MS_RDONLY)) {
2395 lock_super(sb);
2396 /* Reser the needs_recovery flag before the fs is unlocked. */
2397 EXT3_SET_INCOMPAT_FEATURE(sb, EXT3_FEATURE_INCOMPAT_RECOVER);
2398 ext3_commit_super(sb, EXT3_SB(sb)->s_es, 1);
2399 unlock_super(sb);
2400 journal_unlock_updates(EXT3_SB(sb)->s_journal);
2401 }
2402 }
2403
2404 static int ext3_remount (struct super_block * sb, int * flags, char * data)
2405 {
2406 struct ext3_super_block * es;
2407 struct ext3_sb_info *sbi = EXT3_SB(sb);
2408 ext3_fsblk_t n_blocks_count = 0;
2409 unsigned long old_sb_flags;
2410 struct ext3_mount_options old_opts;
2411 int err;
2412 #ifdef CONFIG_QUOTA
2413 int i;
2414 #endif
2415
2416 /* Store the original options */
2417 old_sb_flags = sb->s_flags;
2418 old_opts.s_mount_opt = sbi->s_mount_opt;
2419 old_opts.s_resuid = sbi->s_resuid;
2420 old_opts.s_resgid = sbi->s_resgid;
2421 old_opts.s_commit_interval = sbi->s_commit_interval;
2422 #ifdef CONFIG_QUOTA
2423 old_opts.s_jquota_fmt = sbi->s_jquota_fmt;
2424 for (i = 0; i < MAXQUOTAS; i++)
2425 old_opts.s_qf_names[i] = sbi->s_qf_names[i];
2426 #endif
2427
2428 /*
2429 * Allow the "check" option to be passed as a remount option.
2430 */
2431 if (!parse_options(data, sb, NULL, NULL, &n_blocks_count, 1)) {
2432 err = -EINVAL;
2433 goto restore_opts;
2434 }
2435
2436 if (sbi->s_mount_opt & EXT3_MOUNT_ABORT)
2437 ext3_abort(sb, __FUNCTION__, "Abort forced by user");
2438
2439 sb->s_flags = (sb->s_flags & ~MS_POSIXACL) |
2440 ((sbi->s_mount_opt & EXT3_MOUNT_POSIX_ACL) ? MS_POSIXACL : 0);
2441
2442 es = sbi->s_es;
2443
2444 ext3_init_journal_params(sb, sbi->s_journal);
2445
2446 if ((*flags & MS_RDONLY) != (sb->s_flags & MS_RDONLY) ||
2447 n_blocks_count > le32_to_cpu(es->s_blocks_count)) {
2448 if (sbi->s_mount_opt & EXT3_MOUNT_ABORT) {
2449 err = -EROFS;
2450 goto restore_opts;
2451 }
2452
2453 if (*flags & MS_RDONLY) {
2454 /*
2455 * First of all, the unconditional stuff we have to do
2456 * to disable replay of the journal when we next remount
2457 */
2458 sb->s_flags |= MS_RDONLY;
2459
2460 /*
2461 * OK, test if we are remounting a valid rw partition
2462 * readonly, and if so set the rdonly flag and then
2463 * mark the partition as valid again.
2464 */
2465 if (!(es->s_state & cpu_to_le16(EXT3_VALID_FS)) &&
2466 (sbi->s_mount_state & EXT3_VALID_FS))
2467 es->s_state = cpu_to_le16(sbi->s_mount_state);
2468
2469 /*
2470 * We have to unlock super so that we can wait for
2471 * transactions.
2472 */
2473 unlock_super(sb);
2474 ext3_mark_recovery_complete(sb, es);
2475 lock_super(sb);
2476 } else {
2477 __le32 ret;
2478 if ((ret = EXT3_HAS_RO_COMPAT_FEATURE(sb,
2479 ~EXT3_FEATURE_RO_COMPAT_SUPP))) {
2480 printk(KERN_WARNING "EXT3-fs: %s: couldn't "
2481 "remount RDWR because of unsupported "
2482 "optional features (%x).\n",
2483 sb->s_id, le32_to_cpu(ret));
2484 err = -EROFS;
2485 goto restore_opts;
2486 }
2487
2488 /*
2489 * If we have an unprocessed orphan list hanging
2490 * around from a previously readonly bdev mount,
2491 * require a full umount/remount for now.
2492 */
2493 if (es->s_last_orphan) {
2494 printk(KERN_WARNING "EXT3-fs: %s: couldn't "
2495 "remount RDWR because of unprocessed "
2496 "orphan inode list. Please "
2497 "umount/remount instead.\n",
2498 sb->s_id);
2499 err = -EINVAL;
2500 goto restore_opts;
2501 }
2502
2503 /*
2504 * Mounting a RDONLY partition read-write, so reread
2505 * and store the current valid flag. (It may have
2506 * been changed by e2fsck since we originally mounted
2507 * the partition.)
2508 */
2509 ext3_clear_journal_err(sb, es);
2510 sbi->s_mount_state = le16_to_cpu(es->s_state);
2511 if ((err = ext3_group_extend(sb, es, n_blocks_count)))
2512 goto restore_opts;
2513 if (!ext3_setup_super (sb, es, 0))
2514 sb->s_flags &= ~MS_RDONLY;
2515 }
2516 }
2517 #ifdef CONFIG_QUOTA
2518 /* Release old quota file names */
2519 for (i = 0; i < MAXQUOTAS; i++)
2520 if (old_opts.s_qf_names[i] &&
2521 old_opts.s_qf_names[i] != sbi->s_qf_names[i])
2522 kfree(old_opts.s_qf_names[i]);
2523 #endif
2524 return 0;
2525 restore_opts:
2526 sb->s_flags = old_sb_flags;
2527 sbi->s_mount_opt = old_opts.s_mount_opt;
2528 sbi->s_resuid = old_opts.s_resuid;
2529 sbi->s_resgid = old_opts.s_resgid;
2530 sbi->s_commit_interval = old_opts.s_commit_interval;
2531 #ifdef CONFIG_QUOTA
2532 sbi->s_jquota_fmt = old_opts.s_jquota_fmt;
2533 for (i = 0; i < MAXQUOTAS; i++) {
2534 if (sbi->s_qf_names[i] &&
2535 old_opts.s_qf_names[i] != sbi->s_qf_names[i])
2536 kfree(sbi->s_qf_names[i]);
2537 sbi->s_qf_names[i] = old_opts.s_qf_names[i];
2538 }
2539 #endif
2540 return err;
2541 }
2542
2543 static int ext3_statfs (struct dentry * dentry, struct kstatfs * buf)
2544 {
2545 struct super_block *sb = dentry->d_sb;
2546 struct ext3_sb_info *sbi = EXT3_SB(sb);
2547 struct ext3_super_block *es = sbi->s_es;
2548 u64 fsid;
2549
2550 if (test_opt(sb, MINIX_DF)) {
2551 sbi->s_overhead_last = 0;
2552 } else if (sbi->s_blocks_last != le32_to_cpu(es->s_blocks_count)) {
2553 unsigned long ngroups = sbi->s_groups_count, i;
2554 ext3_fsblk_t overhead = 0;
2555 smp_rmb();
2556
2557 /*
2558 * Compute the overhead (FS structures). This is constant
2559 * for a given filesystem unless the number of block groups
2560 * changes so we cache the previous value until it does.
2561 */
2562
2563 /*
2564 * All of the blocks before first_data_block are
2565 * overhead
2566 */
2567 overhead = le32_to_cpu(es->s_first_data_block);
2568
2569 /*
2570 * Add the overhead attributed to the superblock and
2571 * block group descriptors. If the sparse superblocks
2572 * feature is turned on, then not all groups have this.
2573 */
2574 for (i = 0; i < ngroups; i++) {
2575 overhead += ext3_bg_has_super(sb, i) +
2576 ext3_bg_num_gdb(sb, i);
2577 cond_resched();
2578 }
2579
2580 /*
2581 * Every block group has an inode bitmap, a block
2582 * bitmap, and an inode table.
2583 */
2584 overhead += ngroups * (2 + sbi->s_itb_per_group);
2585 sbi->s_overhead_last = overhead;
2586 smp_wmb();
2587 sbi->s_blocks_last = le32_to_cpu(es->s_blocks_count);
2588 }
2589
2590 buf->f_type = EXT3_SUPER_MAGIC;
2591 buf->f_bsize = sb->s_blocksize;
2592 buf->f_blocks = le32_to_cpu(es->s_blocks_count) - sbi->s_overhead_last;
2593 buf->f_bfree = percpu_counter_sum_positive(&sbi->s_freeblocks_counter);
2594 es->s_free_blocks_count = cpu_to_le32(buf->f_bfree);
2595 buf->f_bavail = buf->f_bfree - le32_to_cpu(es->s_r_blocks_count);
2596 if (buf->f_bfree < le32_to_cpu(es->s_r_blocks_count))
2597 buf->f_bavail = 0;
2598 buf->f_files = le32_to_cpu(es->s_inodes_count);
2599 buf->f_ffree = percpu_counter_sum_positive(&sbi->s_freeinodes_counter);
2600 es->s_free_inodes_count = cpu_to_le32(buf->f_ffree);
2601 buf->f_namelen = EXT3_NAME_LEN;
2602 fsid = le64_to_cpup((void *)es->s_uuid) ^
2603 le64_to_cpup((void *)es->s_uuid + sizeof(u64));
2604 buf->f_fsid.val[0] = fsid & 0xFFFFFFFFUL;
2605 buf->f_fsid.val[1] = (fsid >> 32) & 0xFFFFFFFFUL;
2606 return 0;
2607 }
2608
2609 /* Helper function for writing quotas on sync - we need to start transaction before quota file
2610 * is locked for write. Otherwise the are possible deadlocks:
2611 * Process 1 Process 2
2612 * ext3_create() quota_sync()
2613 * journal_start() write_dquot()
2614 * DQUOT_INIT() down(dqio_mutex)
2615 * down(dqio_mutex) journal_start()
2616 *
2617 */
2618
2619 #ifdef CONFIG_QUOTA
2620
2621 static inline struct inode *dquot_to_inode(struct dquot *dquot)
2622 {
2623 return sb_dqopt(dquot->dq_sb)->files[dquot->dq_type];
2624 }
2625
2626 static int ext3_dquot_initialize(struct inode *inode, int type)
2627 {
2628 handle_t *handle;
2629 int ret, err;
2630
2631 /* We may create quota structure so we need to reserve enough blocks */
2632 handle = ext3_journal_start(inode, 2*EXT3_QUOTA_INIT_BLOCKS(inode->i_sb));
2633 if (IS_ERR(handle))
2634 return PTR_ERR(handle);
2635 ret = dquot_initialize(inode, type);
2636 err = ext3_journal_stop(handle);
2637 if (!ret)
2638 ret = err;
2639 return ret;
2640 }
2641
2642 static int ext3_dquot_drop(struct inode *inode)
2643 {
2644 handle_t *handle;
2645 int ret, err;
2646
2647 /* We may delete quota structure so we need to reserve enough blocks */
2648 handle = ext3_journal_start(inode, 2*EXT3_QUOTA_DEL_BLOCKS(inode->i_sb));
2649 if (IS_ERR(handle))
2650 return PTR_ERR(handle);
2651 ret = dquot_drop(inode);
2652 err = ext3_journal_stop(handle);
2653 if (!ret)
2654 ret = err;
2655 return ret;
2656 }
2657
2658 static int ext3_write_dquot(struct dquot *dquot)
2659 {
2660 int ret, err;
2661 handle_t *handle;
2662 struct inode *inode;
2663
2664 inode = dquot_to_inode(dquot);
2665 handle = ext3_journal_start(inode,
2666 EXT3_QUOTA_TRANS_BLOCKS(dquot->dq_sb));
2667 if (IS_ERR(handle))
2668 return PTR_ERR(handle);
2669 ret = dquot_commit(dquot);
2670 err = ext3_journal_stop(handle);
2671 if (!ret)
2672 ret = err;
2673 return ret;
2674 }
2675
2676 static int ext3_acquire_dquot(struct dquot *dquot)
2677 {
2678 int ret, err;
2679 handle_t *handle;
2680
2681 handle = ext3_journal_start(dquot_to_inode(dquot),
2682 EXT3_QUOTA_INIT_BLOCKS(dquot->dq_sb));
2683 if (IS_ERR(handle))
2684 return PTR_ERR(handle);
2685 ret = dquot_acquire(dquot);
2686 err = ext3_journal_stop(handle);
2687 if (!ret)
2688 ret = err;
2689 return ret;
2690 }
2691
2692 static int ext3_release_dquot(struct dquot *dquot)
2693 {
2694 int ret, err;
2695 handle_t *handle;
2696
2697 handle = ext3_journal_start(dquot_to_inode(dquot),
2698 EXT3_QUOTA_DEL_BLOCKS(dquot->dq_sb));
2699 if (IS_ERR(handle)) {
2700 /* Release dquot anyway to avoid endless cycle in dqput() */
2701 dquot_release(dquot);
2702 return PTR_ERR(handle);
2703 }
2704 ret = dquot_release(dquot);
2705 err = ext3_journal_stop(handle);
2706 if (!ret)
2707 ret = err;
2708 return ret;
2709 }
2710
2711 static int ext3_mark_dquot_dirty(struct dquot *dquot)
2712 {
2713 /* Are we journalling quotas? */
2714 if (EXT3_SB(dquot->dq_sb)->s_qf_names[USRQUOTA] ||
2715 EXT3_SB(dquot->dq_sb)->s_qf_names[GRPQUOTA]) {
2716 dquot_mark_dquot_dirty(dquot);
2717 return ext3_write_dquot(dquot);
2718 } else {
2719 return dquot_mark_dquot_dirty(dquot);
2720 }
2721 }
2722
2723 static int ext3_write_info(struct super_block *sb, int type)
2724 {
2725 int ret, err;
2726 handle_t *handle;
2727
2728 /* Data block + inode block */
2729 handle = ext3_journal_start(sb->s_root->d_inode, 2);
2730 if (IS_ERR(handle))
2731 return PTR_ERR(handle);
2732 ret = dquot_commit_info(sb, type);
2733 err = ext3_journal_stop(handle);
2734 if (!ret)
2735 ret = err;
2736 return ret;
2737 }
2738
2739 /*
2740 * Turn on quotas during mount time - we need to find
2741 * the quota file and such...
2742 */
2743 static int ext3_quota_on_mount(struct super_block *sb, int type)
2744 {
2745 return vfs_quota_on_mount(sb, EXT3_SB(sb)->s_qf_names[type],
2746 EXT3_SB(sb)->s_jquota_fmt, type);
2747 }
2748
2749 /*
2750 * Standard function to be called on quota_on
2751 */
2752 static int ext3_quota_on(struct super_block *sb, int type, int format_id,
2753 char *path)
2754 {
2755 int err;
2756 struct nameidata nd;
2757
2758 if (!test_opt(sb, QUOTA))
2759 return -EINVAL;
2760 /* Not journalling quota? */
2761 if (!EXT3_SB(sb)->s_qf_names[USRQUOTA] &&
2762 !EXT3_SB(sb)->s_qf_names[GRPQUOTA])
2763 return vfs_quota_on(sb, type, format_id, path);
2764 err = path_lookup(path, LOOKUP_FOLLOW, &nd);
2765 if (err)
2766 return err;
2767 /* Quotafile not on the same filesystem? */
2768 if (nd.path.mnt->mnt_sb != sb) {
2769 path_put(&nd.path);
2770 return -EXDEV;
2771 }
2772 /* Quotafile not of fs root? */
2773 if (nd.path.dentry->d_parent->d_inode != sb->s_root->d_inode)
2774 printk(KERN_WARNING
2775 "EXT3-fs: Quota file not on filesystem root. "
2776 "Journalled quota will not work.\n");
2777 path_put(&nd.path);
2778 return vfs_quota_on(sb, type, format_id, path);
2779 }
2780
2781 /* Read data from quotafile - avoid pagecache and such because we cannot afford
2782 * acquiring the locks... As quota files are never truncated and quota code
2783 * itself serializes the operations (and noone else should touch the files)
2784 * we don't have to be afraid of races */
2785 static ssize_t ext3_quota_read(struct super_block *sb, int type, char *data,
2786 size_t len, loff_t off)
2787 {
2788 struct inode *inode = sb_dqopt(sb)->files[type];
2789 sector_t blk = off >> EXT3_BLOCK_SIZE_BITS(sb);
2790 int err = 0;
2791 int offset = off & (sb->s_blocksize - 1);
2792 int tocopy;
2793 size_t toread;
2794 struct buffer_head *bh;
2795 loff_t i_size = i_size_read(inode);
2796
2797 if (off > i_size)
2798 return 0;
2799 if (off+len > i_size)
2800 len = i_size-off;
2801 toread = len;
2802 while (toread > 0) {
2803 tocopy = sb->s_blocksize - offset < toread ?
2804 sb->s_blocksize - offset : toread;
2805 bh = ext3_bread(NULL, inode, blk, 0, &err);
2806 if (err)
2807 return err;
2808 if (!bh) /* A hole? */
2809 memset(data, 0, tocopy);
2810 else
2811 memcpy(data, bh->b_data+offset, tocopy);
2812 brelse(bh);
2813 offset = 0;
2814 toread -= tocopy;
2815 data += tocopy;
2816 blk++;
2817 }
2818 return len;
2819 }
2820
2821 /* Write to quotafile (we know the transaction is already started and has
2822 * enough credits) */
2823 static ssize_t ext3_quota_write(struct super_block *sb, int type,
2824 const char *data, size_t len, loff_t off)
2825 {
2826 struct inode *inode = sb_dqopt(sb)->files[type];
2827 sector_t blk = off >> EXT3_BLOCK_SIZE_BITS(sb);
2828 int err = 0;
2829 int offset = off & (sb->s_blocksize - 1);
2830 int tocopy;
2831 int journal_quota = EXT3_SB(sb)->s_qf_names[type] != NULL;
2832 size_t towrite = len;
2833 struct buffer_head *bh;
2834 handle_t *handle = journal_current_handle();
2835
2836 if (!handle) {
2837 printk(KERN_WARNING "EXT3-fs: Quota write (off=%Lu, len=%Lu)"
2838 " cancelled because transaction is not started.\n",
2839 (unsigned long long)off, (unsigned long long)len);
2840 return -EIO;
2841 }
2842 mutex_lock_nested(&inode->i_mutex, I_MUTEX_QUOTA);
2843 while (towrite > 0) {
2844 tocopy = sb->s_blocksize - offset < towrite ?
2845 sb->s_blocksize - offset : towrite;
2846 bh = ext3_bread(handle, inode, blk, 1, &err);
2847 if (!bh)
2848 goto out;
2849 if (journal_quota) {
2850 err = ext3_journal_get_write_access(handle, bh);
2851 if (err) {
2852 brelse(bh);
2853 goto out;
2854 }
2855 }
2856 lock_buffer(bh);
2857 memcpy(bh->b_data+offset, data, tocopy);
2858 flush_dcache_page(bh->b_page);
2859 unlock_buffer(bh);
2860 if (journal_quota)
2861 err = ext3_journal_dirty_metadata(handle, bh);
2862 else {
2863 /* Always do at least ordered writes for quotas */
2864 err = ext3_journal_dirty_data(handle, bh);
2865 mark_buffer_dirty(bh);
2866 }
2867 brelse(bh);
2868 if (err)
2869 goto out;
2870 offset = 0;
2871 towrite -= tocopy;
2872 data += tocopy;
2873 blk++;
2874 }
2875 out:
2876 if (len == towrite)
2877 return err;
2878 if (inode->i_size < off+len-towrite) {
2879 i_size_write(inode, off+len-towrite);
2880 EXT3_I(inode)->i_disksize = inode->i_size;
2881 }
2882 inode->i_version++;
2883 inode->i_mtime = inode->i_ctime = CURRENT_TIME;
2884 ext3_mark_inode_dirty(handle, inode);
2885 mutex_unlock(&inode->i_mutex);
2886 return len - towrite;
2887 }
2888
2889 #endif
2890
2891 static int ext3_get_sb(struct file_system_type *fs_type,
2892 int flags, const char *dev_name, void *data, struct vfsmount *mnt)
2893 {
2894 return get_sb_bdev(fs_type, flags, dev_name, data, ext3_fill_super, mnt);
2895 }
2896
2897 static struct file_system_type ext3_fs_type = {
2898 .owner = THIS_MODULE,
2899 .name = "ext3",
2900 .get_sb = ext3_get_sb,
2901 .kill_sb = kill_block_super,
2902 .fs_flags = FS_REQUIRES_DEV,
2903 };
2904
2905 static int __init init_ext3_fs(void)
2906 {
2907 int err = init_ext3_xattr();
2908 if (err)
2909 return err;
2910 err = init_inodecache();
2911 if (err)
2912 goto out1;
2913 err = register_filesystem(&ext3_fs_type);
2914 if (err)
2915 goto out;
2916 return 0;
2917 out:
2918 destroy_inodecache();
2919 out1:
2920 exit_ext3_xattr();
2921 return err;
2922 }
2923
2924 static void __exit exit_ext3_fs(void)
2925 {
2926 unregister_filesystem(&ext3_fs_type);
2927 destroy_inodecache();
2928 exit_ext3_xattr();
2929 }
2930
2931 MODULE_AUTHOR("Remy Card, Stephen Tweedie, Andrew Morton, Andreas Dilger, Theodore Ts'o and others");
2932 MODULE_DESCRIPTION("Second Extended Filesystem with journaling extensions");
2933 MODULE_LICENSE("GPL");
2934 module_init(init_ext3_fs)
2935 module_exit(exit_ext3_fs)
WRT350N Kernel Patches