?  search:
summary | log | graphiclog1 | graphiclog2 | commit | commitdiff | tree | refs | edit | fork
blame | history | raw | HEAD
PNPACPI: compute Address Space length rather than using _LEN
[linux-2.6/x86.git] / fs / ext3 / super.c
blob1bee604cc6cd4996c3dd36480c054bd07ee4dfbe
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 #ifdef CONFIG_EXT3_DEFAULTS_TO_ORDERED
48 #define EXT3_MOUNT_DEFAULT_DATA_MODE EXT3_MOUNT_ORDERED_DATA
49 #else
50 #define EXT3_MOUNT_DEFAULT_DATA_MODE EXT3_MOUNT_WRITEBACK_DATA
51 #endif
52
53 static int ext3_load_journal(struct super_block *, struct ext3_super_block *,
54 unsigned long journal_devnum);
55 static int ext3_create_journal(struct super_block *, struct ext3_super_block *,
56 unsigned int);
57 static int ext3_commit_super(struct super_block *sb,
58 struct ext3_super_block *es,
59 int sync);
60 static void ext3_mark_recovery_complete(struct super_block * sb,
61 struct ext3_super_block * es);
62 static void ext3_clear_journal_err(struct super_block * sb,
63 struct ext3_super_block * es);
64 static int ext3_sync_fs(struct super_block *sb, int wait);
65 static const char *ext3_decode_error(struct super_block * sb, int errno,
66 char nbuf[16]);
67 static int ext3_remount (struct super_block * sb, int * flags, char * data);
68 static int ext3_statfs (struct dentry * dentry, struct kstatfs * buf);
69 static int ext3_unfreeze(struct super_block *sb);
70 static int ext3_freeze(struct super_block *sb);
71
72 /*
73 * Wrappers for journal_start/end.
74 *
75 * The only special thing we need to do here is to make sure that all
76 * journal_end calls result in the superblock being marked dirty, so
77 * that sync() will call the filesystem's write_super callback if
78 * appropriate.
79 */
80 handle_t *ext3_journal_start_sb(struct super_block *sb, int nblocks)
81 {
82 journal_t *journal;
83
84 if (sb->s_flags & MS_RDONLY)
85 return ERR_PTR(-EROFS);
86
87 /* Special case here: if the journal has aborted behind our
88 * backs (eg. EIO in the commit thread), then we still need to
89 * take the FS itself readonly cleanly. */
90 journal = EXT3_SB(sb)->s_journal;
91 if (is_journal_aborted(journal)) {
92 ext3_abort(sb, __func__,
93 "Detected aborted journal");
94 return ERR_PTR(-EROFS);
95 }
96
97 return journal_start(journal, nblocks);
98 }
99
100 /*
101 * The only special thing we need to do here is to make sure that all
102 * journal_stop calls result in the superblock being marked dirty, so
103 * that sync() will call the filesystem's write_super callback if
104 * appropriate.
105 */
106 int __ext3_journal_stop(const char *where, handle_t *handle)
107 {
108 struct super_block *sb;
109 int err;
110 int rc;
111
112 sb = handle->h_transaction->t_journal->j_private;
113 err = handle->h_err;
114 rc = journal_stop(handle);
115
116 if (!err)
117 err = rc;
118 if (err)
119 __ext3_std_error(sb, where, err);
120 return err;
121 }
122
123 void ext3_journal_abort_handle(const char *caller, const char *err_fn,
124 struct buffer_head *bh, handle_t *handle, int err)
125 {
126 char nbuf[16];
127 const char *errstr = ext3_decode_error(NULL, err, nbuf);
128
129 if (bh)
130 BUFFER_TRACE(bh, "abort");
131
132 if (!handle->h_err)
133 handle->h_err = err;
134
135 if (is_handle_aborted(handle))
136 return;
137
138 printk(KERN_ERR "EXT3-fs: %s: aborting transaction: %s in %s\n",
139 caller, errstr, err_fn);
140
141 journal_abort_handle(handle);
142 }
143
144 void ext3_msg(struct super_block *sb, const char *prefix,
145 const char *fmt, ...)
146 {
147 va_list args;
148
149 va_start(args, fmt);
150 printk("%sEXT3-fs (%s): ", prefix, sb->s_id);
151 vprintk(fmt, args);
152 printk("\n");
153 va_end(args);
154 }
155
156 /* Deal with the reporting of failure conditions on a filesystem such as
157 * inconsistencies detected or read IO failures.
158 *
159 * On ext2, we can store the error state of the filesystem in the
160 * superblock. That is not possible on ext3, because we may have other
161 * write ordering constraints on the superblock which prevent us from
162 * writing it out straight away; and given that the journal is about to
163 * be aborted, we can't rely on the current, or future, transactions to
164 * write out the superblock safely.
165 *
166 * We'll just use the journal_abort() error code to record an error in
167 * the journal instead. On recovery, the journal will complain about
168 * that error until we've noted it down and cleared it.
169 */
170
171 static void ext3_handle_error(struct super_block *sb)
172 {
173 struct ext3_super_block *es = EXT3_SB(sb)->s_es;
174
175 EXT3_SB(sb)->s_mount_state |= EXT3_ERROR_FS;
176 es->s_state |= cpu_to_le16(EXT3_ERROR_FS);
177
178 if (sb->s_flags & MS_RDONLY)
179 return;
180
181 if (!test_opt (sb, ERRORS_CONT)) {
182 journal_t *journal = EXT3_SB(sb)->s_journal;
183
184 set_opt(EXT3_SB(sb)->s_mount_opt, ABORT);
185 if (journal)
186 journal_abort(journal, -EIO);
187 }
188 if (test_opt (sb, ERRORS_RO)) {
189 ext3_msg(sb, KERN_CRIT,
190 "error: remounting filesystem read-only");
191 sb->s_flags |= MS_RDONLY;
192 }
193 ext3_commit_super(sb, es, 1);
194 if (test_opt(sb, ERRORS_PANIC))
195 panic("EXT3-fs (%s): panic forced after error\n",
196 sb->s_id);
197 }
198
199 void ext3_error (struct super_block * sb, const char * function,
200 const char * fmt, ...)
201 {
202 va_list args;
203
204 va_start(args, fmt);
205 printk(KERN_CRIT "EXT3-fs error (device %s): %s: ",sb->s_id, function);
206 vprintk(fmt, args);
207 printk("\n");
208 va_end(args);
209
210 ext3_handle_error(sb);
211 }
212
213 static const char *ext3_decode_error(struct super_block * sb, int errno,
214 char nbuf[16])
215 {
216 char *errstr = NULL;
217
218 switch (errno) {
219 case -EIO:
220 errstr = "IO failure";
221 break;
222 case -ENOMEM:
223 errstr = "Out of memory";
224 break;
225 case -EROFS:
226 if (!sb || EXT3_SB(sb)->s_journal->j_flags & JFS_ABORT)
227 errstr = "Journal has aborted";
228 else
229 errstr = "Readonly filesystem";
230 break;
231 default:
232 /* If the caller passed in an extra buffer for unknown
233 * errors, textualise them now. Else we just return
234 * NULL. */
235 if (nbuf) {
236 /* Check for truncated error codes... */
237 if (snprintf(nbuf, 16, "error %d", -errno) >= 0)
238 errstr = nbuf;
239 }
240 break;
241 }
242
243 return errstr;
244 }
245
246 /* __ext3_std_error decodes expected errors from journaling functions
247 * automatically and invokes the appropriate error response. */
248
249 void __ext3_std_error (struct super_block * sb, const char * function,
250 int errno)
251 {
252 char nbuf[16];
253 const char *errstr;
254
255 /* Special case: if the error is EROFS, and we're not already
256 * inside a transaction, then there's really no point in logging
257 * an error. */
258 if (errno == -EROFS && journal_current_handle() == NULL &&
259 (sb->s_flags & MS_RDONLY))
260 return;
261
262 errstr = ext3_decode_error(sb, errno, nbuf);
263 ext3_msg(sb, KERN_CRIT, "error in %s: %s", function, errstr);
264
265 ext3_handle_error(sb);
266 }
267
268 /*
269 * ext3_abort is a much stronger failure handler than ext3_error. The
270 * abort function may be used to deal with unrecoverable failures such
271 * as journal IO errors or ENOMEM at a critical moment in log management.
272 *
273 * We unconditionally force the filesystem into an ABORT|READONLY state,
274 * unless the error response on the fs has been set to panic in which
275 * case we take the easy way out and panic immediately.
276 */
277
278 void ext3_abort (struct super_block * sb, const char * function,
279 const char * fmt, ...)
280 {
281 va_list args;
282
283 va_start(args, fmt);
284 printk(KERN_CRIT "EXT3-fs (%s): error: %s: ", sb->s_id, function);
285 vprintk(fmt, args);
286 printk("\n");
287 va_end(args);
288
289 if (test_opt(sb, ERRORS_PANIC))
290 panic("EXT3-fs: panic from previous error\n");
291
292 if (sb->s_flags & MS_RDONLY)
293 return;
294
295 ext3_msg(sb, KERN_CRIT,
296 "error: remounting filesystem read-only");
297 EXT3_SB(sb)->s_mount_state |= EXT3_ERROR_FS;
298 sb->s_flags |= MS_RDONLY;
299 set_opt(EXT3_SB(sb)->s_mount_opt, ABORT);
300 if (EXT3_SB(sb)->s_journal)
301 journal_abort(EXT3_SB(sb)->s_journal, -EIO);
302 }
303
304 void ext3_warning (struct super_block * sb, const char * function,
305 const char * fmt, ...)
306 {
307 va_list args;
308
309 va_start(args, fmt);
310 printk(KERN_WARNING "EXT3-fs (%s): warning: %s: ",
311 sb->s_id, function);
312 vprintk(fmt, args);
313 printk("\n");
314 va_end(args);
315 }
316
317 void ext3_update_dynamic_rev(struct super_block *sb)
318 {
319 struct ext3_super_block *es = EXT3_SB(sb)->s_es;
320
321 if (le32_to_cpu(es->s_rev_level) > EXT3_GOOD_OLD_REV)
322 return;
323
324 ext3_msg(sb, KERN_WARNING,
325 "warning: updating to rev %d because of "
326 "new feature flag, running e2fsck is recommended",
327 EXT3_DYNAMIC_REV);
328
329 es->s_first_ino = cpu_to_le32(EXT3_GOOD_OLD_FIRST_INO);
330 es->s_inode_size = cpu_to_le16(EXT3_GOOD_OLD_INODE_SIZE);
331 es->s_rev_level = cpu_to_le32(EXT3_DYNAMIC_REV);
332 /* leave es->s_feature_*compat flags alone */
333 /* es->s_uuid will be set by e2fsck if empty */
334
335 /*
336 * The rest of the superblock fields should be zero, and if not it
337 * means they are likely already in use, so leave them alone. We
338 * can leave it up to e2fsck to clean up any inconsistencies there.
339 */
340 }
341
342 /*
343 * Open the external journal device
344 */
345 static struct block_device *ext3_blkdev_get(dev_t dev, struct super_block *sb)
346 {
347 struct block_device *bdev;
348 char b[BDEVNAME_SIZE];
349
350 bdev = open_by_devnum(dev, FMODE_READ|FMODE_WRITE);
351 if (IS_ERR(bdev))
352 goto fail;
353 return bdev;
354
355 fail:
356 ext3_msg(sb, "error: failed to open journal device %s: %ld",
357 __bdevname(dev, b), PTR_ERR(bdev));
358
359 return NULL;
360 }
361
362 /*
363 * Release the journal device
364 */
365 static int ext3_blkdev_put(struct block_device *bdev)
366 {
367 bd_release(bdev);
368 return blkdev_put(bdev, FMODE_READ|FMODE_WRITE);
369 }
370
371 static int ext3_blkdev_remove(struct ext3_sb_info *sbi)
372 {
373 struct block_device *bdev;
374 int ret = -ENODEV;
375
376 bdev = sbi->journal_bdev;
377 if (bdev) {
378 ret = ext3_blkdev_put(bdev);
379 sbi->journal_bdev = NULL;
380 }
381 return ret;
382 }
383
384 static inline struct inode *orphan_list_entry(struct list_head *l)
385 {
386 return &list_entry(l, struct ext3_inode_info, i_orphan)->vfs_inode;
387 }
388
389 static void dump_orphan_list(struct super_block *sb, struct ext3_sb_info *sbi)
390 {
391 struct list_head *l;
392
393 ext3_msg(sb, KERN_ERR, "error: sb orphan head is %d",
394 le32_to_cpu(sbi->s_es->s_last_orphan));
395
396 ext3_msg(sb, KERN_ERR, "sb_info orphan list:");
397 list_for_each(l, &sbi->s_orphan) {
398 struct inode *inode = orphan_list_entry(l);
399 ext3_msg(sb, KERN_ERR, " "
400 "inode %s:%lu at %p: mode %o, nlink %d, next %d\n",
401 inode->i_sb->s_id, inode->i_ino, inode,
402 inode->i_mode, inode->i_nlink,
403 NEXT_ORPHAN(inode));
404 }
405 }
406
407 static void ext3_put_super (struct super_block * sb)
408 {
409 struct ext3_sb_info *sbi = EXT3_SB(sb);
410 struct ext3_super_block *es = sbi->s_es;
411 int i, err;
412
413 lock_kernel();
414
415 ext3_xattr_put_super(sb);
416 err = journal_destroy(sbi->s_journal);
417 sbi->s_journal = NULL;
418 if (err < 0)
419 ext3_abort(sb, __func__, "Couldn't clean up the journal");
420
421 if (!(sb->s_flags & MS_RDONLY)) {
422 EXT3_CLEAR_INCOMPAT_FEATURE(sb, EXT3_FEATURE_INCOMPAT_RECOVER);
423 es->s_state = cpu_to_le16(sbi->s_mount_state);
424 BUFFER_TRACE(sbi->s_sbh, "marking dirty");
425 mark_buffer_dirty(sbi->s_sbh);
426 ext3_commit_super(sb, es, 1);
427 }
428
429 for (i = 0; i < sbi->s_gdb_count; i++)
430 brelse(sbi->s_group_desc[i]);
431 kfree(sbi->s_group_desc);
432 percpu_counter_destroy(&sbi->s_freeblocks_counter);
433 percpu_counter_destroy(&sbi->s_freeinodes_counter);
434 percpu_counter_destroy(&sbi->s_dirs_counter);
435 brelse(sbi->s_sbh);
436 #ifdef CONFIG_QUOTA
437 for (i = 0; i < MAXQUOTAS; i++)
438 kfree(sbi->s_qf_names[i]);
439 #endif
440
441 /* Debugging code just in case the in-memory inode orphan list
442 * isn't empty. The on-disk one can be non-empty if we've
443 * detected an error and taken the fs readonly, but the
444 * in-memory list had better be clean by this point. */
445 if (!list_empty(&sbi->s_orphan))
446 dump_orphan_list(sb, sbi);
447 J_ASSERT(list_empty(&sbi->s_orphan));
448
449 invalidate_bdev(sb->s_bdev);
450 if (sbi->journal_bdev && sbi->journal_bdev != sb->s_bdev) {
451 /*
452 * Invalidate the journal device's buffers. We don't want them
453 * floating about in memory - the physical journal device may
454 * hotswapped, and it breaks the `ro-after' testing code.
455 */
456 sync_blockdev(sbi->journal_bdev);
457 invalidate_bdev(sbi->journal_bdev);
458 ext3_blkdev_remove(sbi);
459 }
460 sb->s_fs_info = NULL;
461 kfree(sbi->s_blockgroup_lock);
462 kfree(sbi);
463
464 unlock_kernel();
465 }
466
467 static struct kmem_cache *ext3_inode_cachep;
468
469 /*
470 * Called inside transaction, so use GFP_NOFS
471 */
472 static struct inode *ext3_alloc_inode(struct super_block *sb)
473 {
474 struct ext3_inode_info *ei;
475
476 ei = kmem_cache_alloc(ext3_inode_cachep, GFP_NOFS);
477 if (!ei)
478 return NULL;
479 ei->i_block_alloc_info = NULL;
480 ei->vfs_inode.i_version = 1;
481 atomic_set(&ei->i_datasync_tid, 0);
482 atomic_set(&ei->i_sync_tid, 0);
483 return &ei->vfs_inode;
484 }
485
486 static void ext3_destroy_inode(struct inode *inode)
487 {
488 if (!list_empty(&(EXT3_I(inode)->i_orphan))) {
489 printk("EXT3 Inode %p: orphan list check failed!\n",
490 EXT3_I(inode));
491 print_hex_dump(KERN_INFO, "", DUMP_PREFIX_ADDRESS, 16, 4,
492 EXT3_I(inode), sizeof(struct ext3_inode_info),
493 false);
494 dump_stack();
495 }
496 kmem_cache_free(ext3_inode_cachep, EXT3_I(inode));
497 }
498
499 static void init_once(void *foo)
500 {
501 struct ext3_inode_info *ei = (struct ext3_inode_info *) foo;
502
503 INIT_LIST_HEAD(&ei->i_orphan);
504 #ifdef CONFIG_EXT3_FS_XATTR
505 init_rwsem(&ei->xattr_sem);
506 #endif
507 mutex_init(&ei->truncate_mutex);
508 inode_init_once(&ei->vfs_inode);
509 }
510
511 static int init_inodecache(void)
512 {
513 ext3_inode_cachep = kmem_cache_create("ext3_inode_cache",
514 sizeof(struct ext3_inode_info),
515 0, (SLAB_RECLAIM_ACCOUNT|
516 SLAB_MEM_SPREAD),
517 init_once);
518 if (ext3_inode_cachep == NULL)
519 return -ENOMEM;
520 return 0;
521 }
522
523 static void destroy_inodecache(void)
524 {
525 kmem_cache_destroy(ext3_inode_cachep);
526 }
527
528 static void ext3_clear_inode(struct inode *inode)
529 {
530 struct ext3_block_alloc_info *rsv = EXT3_I(inode)->i_block_alloc_info;
531
532 dquot_drop(inode);
533 ext3_discard_reservation(inode);
534 EXT3_I(inode)->i_block_alloc_info = NULL;
535 if (unlikely(rsv))
536 kfree(rsv);
537 }
538
539 static inline void ext3_show_quota_options(struct seq_file *seq, struct super_block *sb)
540 {
541 #if defined(CONFIG_QUOTA)
542 struct ext3_sb_info *sbi = EXT3_SB(sb);
543
544 if (sbi->s_jquota_fmt) {
545 char *fmtname = "";
546
547 switch (sbi->s_jquota_fmt) {
548 case QFMT_VFS_OLD:
549 fmtname = "vfsold";
550 break;
551 case QFMT_VFS_V0:
552 fmtname = "vfsv0";
553 break;
554 case QFMT_VFS_V1:
555 fmtname = "vfsv1";
556 break;
557 }
558 seq_printf(seq, ",jqfmt=%s", fmtname);
559 }
560
561 if (sbi->s_qf_names[USRQUOTA])
562 seq_printf(seq, ",usrjquota=%s", sbi->s_qf_names[USRQUOTA]);
563
564 if (sbi->s_qf_names[GRPQUOTA])
565 seq_printf(seq, ",grpjquota=%s", sbi->s_qf_names[GRPQUOTA]);
566
567 if (test_opt(sb, USRQUOTA))
568 seq_puts(seq, ",usrquota");
569
570 if (test_opt(sb, GRPQUOTA))
571 seq_puts(seq, ",grpquota");
572 #endif
573 }
574
575 static char *data_mode_string(unsigned long mode)
576 {
577 switch (mode) {
578 case EXT3_MOUNT_JOURNAL_DATA:
579 return "journal";
580 case EXT3_MOUNT_ORDERED_DATA:
581 return "ordered";
582 case EXT3_MOUNT_WRITEBACK_DATA:
583 return "writeback";
584 }
585 return "unknown";
586 }
587
588 /*
589 * Show an option if
590 * - it's set to a non-default value OR
591 * - if the per-sb default is different from the global default
592 */
593 static int ext3_show_options(struct seq_file *seq, struct vfsmount *vfs)
594 {
595 struct super_block *sb = vfs->mnt_sb;
596 struct ext3_sb_info *sbi = EXT3_SB(sb);
597 struct ext3_super_block *es = sbi->s_es;
598 unsigned long def_mount_opts;
599
600 def_mount_opts = le32_to_cpu(es->s_default_mount_opts);
601
602 if (sbi->s_sb_block != 1)
603 seq_printf(seq, ",sb=%lu", sbi->s_sb_block);
604 if (test_opt(sb, MINIX_DF))
605 seq_puts(seq, ",minixdf");
606 if (test_opt(sb, GRPID))
607 seq_puts(seq, ",grpid");
608 if (!test_opt(sb, GRPID) && (def_mount_opts & EXT3_DEFM_BSDGROUPS))
609 seq_puts(seq, ",nogrpid");
610 if (sbi->s_resuid != EXT3_DEF_RESUID ||
611 le16_to_cpu(es->s_def_resuid) != EXT3_DEF_RESUID) {
612 seq_printf(seq, ",resuid=%u", sbi->s_resuid);
613 }
614 if (sbi->s_resgid != EXT3_DEF_RESGID ||
615 le16_to_cpu(es->s_def_resgid) != EXT3_DEF_RESGID) {
616 seq_printf(seq, ",resgid=%u", sbi->s_resgid);
617 }
618 if (test_opt(sb, ERRORS_RO)) {
619 int def_errors = le16_to_cpu(es->s_errors);
620
621 if (def_errors == EXT3_ERRORS_PANIC ||
622 def_errors == EXT3_ERRORS_CONTINUE) {
623 seq_puts(seq, ",errors=remount-ro");
624 }
625 }
626 if (test_opt(sb, ERRORS_CONT))
627 seq_puts(seq, ",errors=continue");
628 if (test_opt(sb, ERRORS_PANIC))
629 seq_puts(seq, ",errors=panic");
630 if (test_opt(sb, NO_UID32))
631 seq_puts(seq, ",nouid32");
632 if (test_opt(sb, DEBUG))
633 seq_puts(seq, ",debug");
634 if (test_opt(sb, OLDALLOC))
635 seq_puts(seq, ",oldalloc");
636 #ifdef CONFIG_EXT3_FS_XATTR
637 if (test_opt(sb, XATTR_USER))
638 seq_puts(seq, ",user_xattr");
639 if (!test_opt(sb, XATTR_USER) &&
640 (def_mount_opts & EXT3_DEFM_XATTR_USER)) {
641 seq_puts(seq, ",nouser_xattr");
642 }
643 #endif
644 #ifdef CONFIG_EXT3_FS_POSIX_ACL
645 if (test_opt(sb, POSIX_ACL))
646 seq_puts(seq, ",acl");
647 if (!test_opt(sb, POSIX_ACL) && (def_mount_opts & EXT3_DEFM_ACL))
648 seq_puts(seq, ",noacl");
649 #endif
650 if (!test_opt(sb, RESERVATION))
651 seq_puts(seq, ",noreservation");
652 if (sbi->s_commit_interval) {
653 seq_printf(seq, ",commit=%u",
654 (unsigned) (sbi->s_commit_interval / HZ));
655 }
656 if (test_opt(sb, BARRIER))
657 seq_puts(seq, ",barrier=1");
658 if (test_opt(sb, NOBH))
659 seq_puts(seq, ",nobh");
660
661 seq_printf(seq, ",data=%s", data_mode_string(test_opt(sb, DATA_FLAGS)));
662 if (test_opt(sb, DATA_ERR_ABORT))
663 seq_puts(seq, ",data_err=abort");
664
665 if (test_opt(sb, NOLOAD))
666 seq_puts(seq, ",norecovery");
667
668 ext3_show_quota_options(seq, sb);
669
670 return 0;
671 }
672
673
674 static struct inode *ext3_nfs_get_inode(struct super_block *sb,
675 u64 ino, u32 generation)
676 {
677 struct inode *inode;
678
679 if (ino < EXT3_FIRST_INO(sb) && ino != EXT3_ROOT_INO)
680 return ERR_PTR(-ESTALE);
681 if (ino > le32_to_cpu(EXT3_SB(sb)->s_es->s_inodes_count))
682 return ERR_PTR(-ESTALE);
683
684 /* iget isn't really right if the inode is currently unallocated!!
685 *
686 * ext3_read_inode will return a bad_inode if the inode had been
687 * deleted, so we should be safe.
688 *
689 * Currently we don't know the generation for parent directory, so
690 * a generation of 0 means "accept any"
691 */
692 inode = ext3_iget(sb, ino);
693 if (IS_ERR(inode))
694 return ERR_CAST(inode);
695 if (generation && inode->i_generation != generation) {
696 iput(inode);
697 return ERR_PTR(-ESTALE);
698 }
699
700 return inode;
701 }
702
703 static struct dentry *ext3_fh_to_dentry(struct super_block *sb, struct fid *fid,
704 int fh_len, int fh_type)
705 {
706 return generic_fh_to_dentry(sb, fid, fh_len, fh_type,
707 ext3_nfs_get_inode);
708 }
709
710 static struct dentry *ext3_fh_to_parent(struct super_block *sb, struct fid *fid,
711 int fh_len, int fh_type)
712 {
713 return generic_fh_to_parent(sb, fid, fh_len, fh_type,
714 ext3_nfs_get_inode);
715 }
716
717 /*
718 * Try to release metadata pages (indirect blocks, directories) which are
719 * mapped via the block device. Since these pages could have journal heads
720 * which would prevent try_to_free_buffers() from freeing them, we must use
721 * jbd layer's try_to_free_buffers() function to release them.
722 */
723 static int bdev_try_to_free_page(struct super_block *sb, struct page *page,
724 gfp_t wait)
725 {
726 journal_t *journal = EXT3_SB(sb)->s_journal;
727
728 WARN_ON(PageChecked(page));
729 if (!page_has_buffers(page))
730 return 0;
731 if (journal)
732 return journal_try_to_free_buffers(journal, page,
733 wait & ~__GFP_WAIT);
734 return try_to_free_buffers(page);
735 }
736
737 #ifdef CONFIG_QUOTA
738 #define QTYPE2NAME(t) ((t)==USRQUOTA?"user":"group")
739 #define QTYPE2MOPT(on, t) ((t)==USRQUOTA?((on)##USRJQUOTA):((on)##GRPJQUOTA))
740
741 static int ext3_write_dquot(struct dquot *dquot);
742 static int ext3_acquire_dquot(struct dquot *dquot);
743 static int ext3_release_dquot(struct dquot *dquot);
744 static int ext3_mark_dquot_dirty(struct dquot *dquot);
745 static int ext3_write_info(struct super_block *sb, int type);
746 static int ext3_quota_on(struct super_block *sb, int type, int format_id,
747 char *path, int remount);
748 static int ext3_quota_on_mount(struct super_block *sb, int type);
749 static ssize_t ext3_quota_read(struct super_block *sb, int type, char *data,
750 size_t len, loff_t off);
751 static ssize_t ext3_quota_write(struct super_block *sb, int type,
752 const char *data, size_t len, loff_t off);
753
754 static const struct dquot_operations ext3_quota_operations = {
755 .write_dquot = ext3_write_dquot,
756 .acquire_dquot = ext3_acquire_dquot,
757 .release_dquot = ext3_release_dquot,
758 .mark_dirty = ext3_mark_dquot_dirty,
759 .write_info = ext3_write_info,
760 .alloc_dquot = dquot_alloc,
761 .destroy_dquot = dquot_destroy,
762 };
763
764 static const struct quotactl_ops ext3_qctl_operations = {
765 .quota_on = ext3_quota_on,
766 .quota_off = vfs_quota_off,
767 .quota_sync = vfs_quota_sync,
768 .get_info = vfs_get_dqinfo,
769 .set_info = vfs_set_dqinfo,
770 .get_dqblk = vfs_get_dqblk,
771 .set_dqblk = vfs_set_dqblk
772 };
773 #endif
774
775 static const struct super_operations ext3_sops = {
776 .alloc_inode = ext3_alloc_inode,
777 .destroy_inode = ext3_destroy_inode,
778 .write_inode = ext3_write_inode,
779 .dirty_inode = ext3_dirty_inode,
780 .delete_inode = ext3_delete_inode,
781 .put_super = ext3_put_super,
782 .sync_fs = ext3_sync_fs,
783 .freeze_fs = ext3_freeze,
784 .unfreeze_fs = ext3_unfreeze,
785 .statfs = ext3_statfs,
786 .remount_fs = ext3_remount,
787 .clear_inode = ext3_clear_inode,
788 .show_options = ext3_show_options,
789 #ifdef CONFIG_QUOTA
790 .quota_read = ext3_quota_read,
791 .quota_write = ext3_quota_write,
792 #endif
793 .bdev_try_to_free_page = bdev_try_to_free_page,
794 };
795
796 static const struct export_operations ext3_export_ops = {
797 .fh_to_dentry = ext3_fh_to_dentry,
798 .fh_to_parent = ext3_fh_to_parent,
799 .get_parent = ext3_get_parent,
800 };
801
802 enum {
803 Opt_bsd_df, Opt_minix_df, Opt_grpid, Opt_nogrpid,
804 Opt_resgid, Opt_resuid, Opt_sb, Opt_err_cont, Opt_err_panic, Opt_err_ro,
805 Opt_nouid32, Opt_nocheck, Opt_debug, Opt_oldalloc, Opt_orlov,
806 Opt_user_xattr, Opt_nouser_xattr, Opt_acl, Opt_noacl,
807 Opt_reservation, Opt_noreservation, Opt_noload, Opt_nobh, Opt_bh,
808 Opt_commit, Opt_journal_update, Opt_journal_inum, Opt_journal_dev,
809 Opt_abort, Opt_data_journal, Opt_data_ordered, Opt_data_writeback,
810 Opt_data_err_abort, Opt_data_err_ignore,
811 Opt_usrjquota, Opt_grpjquota, Opt_offusrjquota, Opt_offgrpjquota,
812 Opt_jqfmt_vfsold, Opt_jqfmt_vfsv0, Opt_jqfmt_vfsv1, Opt_quota,
813 Opt_noquota, Opt_ignore, Opt_barrier, Opt_err, Opt_resize,
814 Opt_usrquota, Opt_grpquota
815 };
816
817 static const match_table_t tokens = {
818 {Opt_bsd_df, "bsddf"},
819 {Opt_minix_df, "minixdf"},
820 {Opt_grpid, "grpid"},
821 {Opt_grpid, "bsdgroups"},
822 {Opt_nogrpid, "nogrpid"},
823 {Opt_nogrpid, "sysvgroups"},
824 {Opt_resgid, "resgid=%u"},
825 {Opt_resuid, "resuid=%u"},
826 {Opt_sb, "sb=%u"},
827 {Opt_err_cont, "errors=continue"},
828 {Opt_err_panic, "errors=panic"},
829 {Opt_err_ro, "errors=remount-ro"},
830 {Opt_nouid32, "nouid32"},
831 {Opt_nocheck, "nocheck"},
832 {Opt_nocheck, "check=none"},
833 {Opt_debug, "debug"},
834 {Opt_oldalloc, "oldalloc"},
835 {Opt_orlov, "orlov"},
836 {Opt_user_xattr, "user_xattr"},
837 {Opt_nouser_xattr, "nouser_xattr"},
838 {Opt_acl, "acl"},
839 {Opt_noacl, "noacl"},
840 {Opt_reservation, "reservation"},
841 {Opt_noreservation, "noreservation"},
842 {Opt_noload, "noload"},
843 {Opt_noload, "norecovery"},
844 {Opt_nobh, "nobh"},
845 {Opt_bh, "bh"},
846 {Opt_commit, "commit=%u"},
847 {Opt_journal_update, "journal=update"},
848 {Opt_journal_inum, "journal=%u"},
849 {Opt_journal_dev, "journal_dev=%u"},
850 {Opt_abort, "abort"},
851 {Opt_data_journal, "data=journal"},
852 {Opt_data_ordered, "data=ordered"},
853 {Opt_data_writeback, "data=writeback"},
854 {Opt_data_err_abort, "data_err=abort"},
855 {Opt_data_err_ignore, "data_err=ignore"},
856 {Opt_offusrjquota, "usrjquota="},
857 {Opt_usrjquota, "usrjquota=%s"},
858 {Opt_offgrpjquota, "grpjquota="},
859 {Opt_grpjquota, "grpjquota=%s"},
860 {Opt_jqfmt_vfsold, "jqfmt=vfsold"},
861 {Opt_jqfmt_vfsv0, "jqfmt=vfsv0"},
862 {Opt_jqfmt_vfsv1, "jqfmt=vfsv1"},
863 {Opt_grpquota, "grpquota"},
864 {Opt_noquota, "noquota"},
865 {Opt_quota, "quota"},
866 {Opt_usrquota, "usrquota"},
867 {Opt_barrier, "barrier=%u"},
868 {Opt_resize, "resize"},
869 {Opt_err, NULL},
870 };
871
872 static ext3_fsblk_t get_sb_block(void **data, struct super_block *sb)
873 {
874 ext3_fsblk_t sb_block;
875 char *options = (char *) *data;
876
877 if (!options || strncmp(options, "sb=", 3) != 0)
878 return 1; /* Default location */
879 options += 3;
880 /*todo: use simple_strtoll with >32bit ext3 */
881 sb_block = simple_strtoul(options, &options, 0);
882 if (*options && *options != ',') {
883 ext3_msg(sb, "error: invalid sb specification: %s",
884 (char *) *data);
885 return 1;
886 }
887 if (*options == ',')
888 options++;
889 *data = (void *) options;
890 return sb_block;
891 }
892
893 #ifdef CONFIG_QUOTA
894 static int set_qf_name(struct super_block *sb, int qtype, substring_t *args)
895 {
896 struct ext3_sb_info *sbi = EXT3_SB(sb);
897 char *qname;
898
899 if (sb_any_quota_loaded(sb) &&
900 !sbi->s_qf_names[qtype]) {
901 ext3_msg(sb, KERN_ERR,
902 "Cannot change journaled "
903 "quota options when quota turned on");
904 return 0;
905 }
906 qname = match_strdup(args);
907 if (!qname) {
908 ext3_msg(sb, KERN_ERR,
909 "Not enough memory for storing quotafile name");
910 return 0;
911 }
912 if (sbi->s_qf_names[qtype] &&
913 strcmp(sbi->s_qf_names[qtype], qname)) {
914 ext3_msg(sb, KERN_ERR,
915 "%s quota file already specified", QTYPE2NAME(qtype));
916 kfree(qname);
917 return 0;
918 }
919 sbi->s_qf_names[qtype] = qname;
920 if (strchr(sbi->s_qf_names[qtype], '/')) {
921 ext3_msg(sb, KERN_ERR,
922 "quotafile must be on filesystem root");
923 kfree(sbi->s_qf_names[qtype]);
924 sbi->s_qf_names[qtype] = NULL;
925 return 0;
926 }
927 set_opt(sbi->s_mount_opt, QUOTA);
928 return 1;
929 }
930
931 static int clear_qf_name(struct super_block *sb, int qtype) {
932
933 struct ext3_sb_info *sbi = EXT3_SB(sb);
934
935 if (sb_any_quota_loaded(sb) &&
936 sbi->s_qf_names[qtype]) {
937 ext3_msg(sb, KERN_ERR, "Cannot change journaled quota options"
938 " when quota turned on");
939 return 0;
940 }
941 /*
942 * The space will be released later when all options are confirmed
943 * to be correct
944 */
945 sbi->s_qf_names[qtype] = NULL;
946 return 1;
947 }
948 #endif
949
950 static int parse_options (char *options, struct super_block *sb,
951 unsigned int *inum, unsigned long *journal_devnum,
952 ext3_fsblk_t *n_blocks_count, int is_remount)
953 {
954 struct ext3_sb_info *sbi = EXT3_SB(sb);
955 char * p;
956 substring_t args[MAX_OPT_ARGS];
957 int data_opt = 0;
958 int option;
959 #ifdef CONFIG_QUOTA
960 int qfmt;
961 #endif
962
963 if (!options)
964 return 1;
965
966 while ((p = strsep (&options, ",")) != NULL) {
967 int token;
968 if (!*p)
969 continue;
970
971 token = match_token(p, tokens, args);
972 switch (token) {
973 case Opt_bsd_df:
974 clear_opt (sbi->s_mount_opt, MINIX_DF);
975 break;
976 case Opt_minix_df:
977 set_opt (sbi->s_mount_opt, MINIX_DF);
978 break;
979 case Opt_grpid:
980 set_opt (sbi->s_mount_opt, GRPID);
981 break;
982 case Opt_nogrpid:
983 clear_opt (sbi->s_mount_opt, GRPID);
984 break;
985 case Opt_resuid:
986 if (match_int(&args[0], &option))
987 return 0;
988 sbi->s_resuid = option;
989 break;
990 case Opt_resgid:
991 if (match_int(&args[0], &option))
992 return 0;
993 sbi->s_resgid = option;
994 break;
995 case Opt_sb:
996 /* handled by get_sb_block() instead of here */
997 /* *sb_block = match_int(&args[0]); */
998 break;
999 case Opt_err_panic:
1000 clear_opt (sbi->s_mount_opt, ERRORS_CONT);
1001 clear_opt (sbi->s_mount_opt, ERRORS_RO);
1002 set_opt (sbi->s_mount_opt, ERRORS_PANIC);
1003 break;
1004 case Opt_err_ro:
1005 clear_opt (sbi->s_mount_opt, ERRORS_CONT);
1006 clear_opt (sbi->s_mount_opt, ERRORS_PANIC);
1007 set_opt (sbi->s_mount_opt, ERRORS_RO);
1008 break;
1009 case Opt_err_cont:
1010 clear_opt (sbi->s_mount_opt, ERRORS_RO);
1011 clear_opt (sbi->s_mount_opt, ERRORS_PANIC);
1012 set_opt (sbi->s_mount_opt, ERRORS_CONT);
1013 break;
1014 case Opt_nouid32:
1015 set_opt (sbi->s_mount_opt, NO_UID32);
1016 break;
1017 case Opt_nocheck:
1018 clear_opt (sbi->s_mount_opt, CHECK);
1019 break;
1020 case Opt_debug:
1021 set_opt (sbi->s_mount_opt, DEBUG);
1022 break;
1023 case Opt_oldalloc:
1024 set_opt (sbi->s_mount_opt, OLDALLOC);
1025 break;
1026 case Opt_orlov:
1027 clear_opt (sbi->s_mount_opt, OLDALLOC);
1028 break;
1029 #ifdef CONFIG_EXT3_FS_XATTR
1030 case Opt_user_xattr:
1031 set_opt (sbi->s_mount_opt, XATTR_USER);
1032 break;
1033 case Opt_nouser_xattr:
1034 clear_opt (sbi->s_mount_opt, XATTR_USER);
1035 break;
1036 #else
1037 case Opt_user_xattr:
1038 case Opt_nouser_xattr:
1039 ext3_msg(sb, KERN_INFO,
1040 "(no)user_xattr options not supported");
1041 break;
1042 #endif
1043 #ifdef CONFIG_EXT3_FS_POSIX_ACL
1044 case Opt_acl:
1045 set_opt(sbi->s_mount_opt, POSIX_ACL);
1046 break;
1047 case Opt_noacl:
1048 clear_opt(sbi->s_mount_opt, POSIX_ACL);
1049 break;
1050 #else
1051 case Opt_acl:
1052 case Opt_noacl:
1053 ext3_msg(sb, KERN_INFO,
1054 "(no)acl options not supported");
1055 break;
1056 #endif
1057 case Opt_reservation:
1058 set_opt(sbi->s_mount_opt, RESERVATION);
1059 break;
1060 case Opt_noreservation:
1061 clear_opt(sbi->s_mount_opt, RESERVATION);
1062 break;
1063 case Opt_journal_update:
1064 /* @@@ FIXME */
1065 /* Eventually we will want to be able to create
1066 a journal file here. For now, only allow the
1067 user to specify an existing inode to be the
1068 journal file. */
1069 if (is_remount) {
1070 ext3_msg(sb, KERN_ERR, "error: cannot specify "
1071 "journal on remount");
1072 return 0;
1073 }
1074 set_opt (sbi->s_mount_opt, UPDATE_JOURNAL);
1075 break;
1076 case Opt_journal_inum:
1077 if (is_remount) {
1078 ext3_msg(sb, KERN_ERR, "error: cannot specify "
1079 "journal on remount");
1080 return 0;
1081 }
1082 if (match_int(&args[0], &option))
1083 return 0;
1084 *inum = option;
1085 break;
1086 case Opt_journal_dev:
1087 if (is_remount) {
1088 ext3_msg(sb, KERN_ERR, "error: cannot specify "
1089 "journal on remount");
1090 return 0;
1091 }
1092 if (match_int(&args[0], &option))
1093 return 0;
1094 *journal_devnum = option;
1095 break;
1096 case Opt_noload:
1097 set_opt (sbi->s_mount_opt, NOLOAD);
1098 break;
1099 case Opt_commit:
1100 if (match_int(&args[0], &option))
1101 return 0;
1102 if (option < 0)
1103 return 0;
1104 if (option == 0)
1105 option = JBD_DEFAULT_MAX_COMMIT_AGE;
1106 sbi->s_commit_interval = HZ * option;
1107 break;
1108 case Opt_data_journal:
1109 data_opt = EXT3_MOUNT_JOURNAL_DATA;
1110 goto datacheck;
1111 case Opt_data_ordered:
1112 data_opt = EXT3_MOUNT_ORDERED_DATA;
1113 goto datacheck;
1114 case Opt_data_writeback:
1115 data_opt = EXT3_MOUNT_WRITEBACK_DATA;
1116 datacheck:
1117 if (is_remount) {
1118 if (test_opt(sb, DATA_FLAGS) == data_opt)
1119 break;
1120 ext3_msg(sb, KERN_ERR,
1121 "error: cannot change "
1122 "data mode on remount. The filesystem "
1123 "is mounted in data=%s mode and you "
1124 "try to remount it in data=%s mode.",
1125 data_mode_string(test_opt(sb,
1126 DATA_FLAGS)),
1127 data_mode_string(data_opt));
1128 return 0;
1129 } else {
1130 clear_opt(sbi->s_mount_opt, DATA_FLAGS);
1131 sbi->s_mount_opt |= data_opt;
1132 }
1133 break;
1134 case Opt_data_err_abort:
1135 set_opt(sbi->s_mount_opt, DATA_ERR_ABORT);
1136 break;
1137 case Opt_data_err_ignore:
1138 clear_opt(sbi->s_mount_opt, DATA_ERR_ABORT);
1139 break;
1140 #ifdef CONFIG_QUOTA
1141 case Opt_usrjquota:
1142 if (!set_qf_name(sb, USRQUOTA, &args[0]))
1143 return 0;
1144 break;
1145 case Opt_grpjquota:
1146 if (!set_qf_name(sb, GRPQUOTA, &args[0]))
1147 return 0;
1148 break;
1149 case Opt_offusrjquota:
1150 if (!clear_qf_name(sb, USRQUOTA))
1151 return 0;
1152 break;
1153 case Opt_offgrpjquota:
1154 if (!clear_qf_name(sb, GRPQUOTA))
1155 return 0;
1156 break;
1157 case Opt_jqfmt_vfsold:
1158 qfmt = QFMT_VFS_OLD;
1159 goto set_qf_format;
1160 case Opt_jqfmt_vfsv0:
1161 qfmt = QFMT_VFS_V0;
1162 goto set_qf_format;
1163 case Opt_jqfmt_vfsv1:
1164 qfmt = QFMT_VFS_V1;
1165 set_qf_format:
1166 if (sb_any_quota_loaded(sb) &&
1167 sbi->s_jquota_fmt != qfmt) {
1168 ext3_msg(sb, KERN_ERR, "error: cannot change "
1169 "journaled quota options when "
1170 "quota turned on.");
1171 return 0;
1172 }
1173 sbi->s_jquota_fmt = qfmt;
1174 break;
1175 case Opt_quota:
1176 case Opt_usrquota:
1177 set_opt(sbi->s_mount_opt, QUOTA);
1178 set_opt(sbi->s_mount_opt, USRQUOTA);
1179 break;
1180 case Opt_grpquota:
1181 set_opt(sbi->s_mount_opt, QUOTA);
1182 set_opt(sbi->s_mount_opt, GRPQUOTA);
1183 break;
1184 case Opt_noquota:
1185 if (sb_any_quota_loaded(sb)) {
1186 ext3_msg(sb, KERN_ERR, "error: cannot change "
1187 "quota options when quota turned on.");
1188 return 0;
1189 }
1190 clear_opt(sbi->s_mount_opt, QUOTA);
1191 clear_opt(sbi->s_mount_opt, USRQUOTA);
1192 clear_opt(sbi->s_mount_opt, GRPQUOTA);
1193 break;
1194 #else
1195 case Opt_quota:
1196 case Opt_usrquota:
1197 case Opt_grpquota:
1198 ext3_msg(sb, KERN_ERR,
1199 "error: quota options not supported.");
1200 break;
1201 case Opt_usrjquota:
1202 case Opt_grpjquota:
1203 case Opt_offusrjquota:
1204 case Opt_offgrpjquota:
1205 case Opt_jqfmt_vfsold:
1206 case Opt_jqfmt_vfsv0:
1207 case Opt_jqfmt_vfsv1:
1208 ext3_msg(sb, KERN_ERR,
1209 "error: journaled quota options not "
1210 "supported.");
1211 break;
1212 case Opt_noquota:
1213 break;
1214 #endif
1215 case Opt_abort:
1216 set_opt(sbi->s_mount_opt, ABORT);
1217 break;
1218 case Opt_barrier:
1219 if (match_int(&args[0], &option))
1220 return 0;
1221 if (option)
1222 set_opt(sbi->s_mount_opt, BARRIER);
1223 else
1224 clear_opt(sbi->s_mount_opt, BARRIER);
1225 break;
1226 case Opt_ignore:
1227 break;
1228 case Opt_resize:
1229 if (!is_remount) {
1230 ext3_msg(sb, KERN_ERR,
1231 "error: resize option only available "
1232 "for remount");
1233 return 0;
1234 }
1235 if (match_int(&args[0], &option) != 0)
1236 return 0;
1237 *n_blocks_count = option;
1238 break;
1239 case Opt_nobh:
1240 set_opt(sbi->s_mount_opt, NOBH);
1241 break;
1242 case Opt_bh:
1243 clear_opt(sbi->s_mount_opt, NOBH);
1244 break;
1245 default:
1246 ext3_msg(sb, KERN_ERR,
1247 "error: unrecognized mount option \"%s\" "
1248 "or missing value", p);
1249 return 0;
1250 }
1251 }
1252 #ifdef CONFIG_QUOTA
1253 if (sbi->s_qf_names[USRQUOTA] || sbi->s_qf_names[GRPQUOTA]) {
1254 if (test_opt(sb, USRQUOTA) && sbi->s_qf_names[USRQUOTA])
1255 clear_opt(sbi->s_mount_opt, USRQUOTA);
1256 if (test_opt(sb, GRPQUOTA) && sbi->s_qf_names[GRPQUOTA])
1257 clear_opt(sbi->s_mount_opt, GRPQUOTA);
1258
1259 if (test_opt(sb, GRPQUOTA) || test_opt(sb, USRQUOTA)) {
1260 ext3_msg(sb, KERN_ERR, "error: old and new quota "
1261 "format mixing.");
1262 return 0;
1263 }
1264
1265 if (!sbi->s_jquota_fmt) {
1266 ext3_msg(sb, KERN_ERR, "error: journaled quota format "
1267 "not specified.");
1268 return 0;
1269 }
1270 } else {
1271 if (sbi->s_jquota_fmt) {
1272 ext3_msg(sb, KERN_ERR, "error: journaled quota format "
1273 "specified with no journaling "
1274 "enabled.");
1275 return 0;
1276 }
1277 }
1278 #endif
1279 return 1;
1280 }
1281
1282 static int ext3_setup_super(struct super_block *sb, struct ext3_super_block *es,
1283 int read_only)
1284 {
1285 struct ext3_sb_info *sbi = EXT3_SB(sb);
1286 int res = 0;
1287
1288 if (le32_to_cpu(es->s_rev_level) > EXT3_MAX_SUPP_REV) {
1289 ext3_msg(sb, KERN_ERR,
1290 "error: revision level too high, "
1291 "forcing read-only mode");
1292 res = MS_RDONLY;
1293 }
1294 if (read_only)
1295 return res;
1296 if (!(sbi->s_mount_state & EXT3_VALID_FS))
1297 ext3_msg(sb, KERN_WARNING,
1298 "warning: mounting unchecked fs, "
1299 "running e2fsck is recommended");
1300 else if ((sbi->s_mount_state & EXT3_ERROR_FS))
1301 ext3_msg(sb, KERN_WARNING,
1302 "warning: mounting fs with errors, "
1303 "running e2fsck is recommended");
1304 else if ((__s16) le16_to_cpu(es->s_max_mnt_count) >= 0 &&
1305 le16_to_cpu(es->s_mnt_count) >=
1306 (unsigned short) (__s16) le16_to_cpu(es->s_max_mnt_count))
1307 ext3_msg(sb, KERN_WARNING,
1308 "warning: maximal mount count reached, "
1309 "running e2fsck is recommended");
1310 else if (le32_to_cpu(es->s_checkinterval) &&
1311 (le32_to_cpu(es->s_lastcheck) +
1312 le32_to_cpu(es->s_checkinterval) <= get_seconds()))
1313 ext3_msg(sb, KERN_WARNING,
1314 "warning: checktime reached, "
1315 "running e2fsck is recommended");
1316 #if 0
1317 /* @@@ We _will_ want to clear the valid bit if we find
1318 inconsistencies, to force a fsck at reboot. But for
1319 a plain journaled filesystem we can keep it set as
1320 valid forever! :) */
1321 es->s_state &= cpu_to_le16(~EXT3_VALID_FS);
1322 #endif
1323 if (!(__s16) le16_to_cpu(es->s_max_mnt_count))
1324 es->s_max_mnt_count = cpu_to_le16(EXT3_DFL_MAX_MNT_COUNT);
1325 le16_add_cpu(&es->s_mnt_count, 1);
1326 es->s_mtime = cpu_to_le32(get_seconds());
1327 ext3_update_dynamic_rev(sb);
1328 EXT3_SET_INCOMPAT_FEATURE(sb, EXT3_FEATURE_INCOMPAT_RECOVER);
1329
1330 ext3_commit_super(sb, es, 1);
1331 if (test_opt(sb, DEBUG))
1332 ext3_msg(sb, KERN_INFO, "[bs=%lu, gc=%lu, "
1333 "bpg=%lu, ipg=%lu, mo=%04lx]",
1334 sb->s_blocksize,
1335 sbi->s_groups_count,
1336 EXT3_BLOCKS_PER_GROUP(sb),
1337 EXT3_INODES_PER_GROUP(sb),
1338 sbi->s_mount_opt);
1339
1340 if (EXT3_SB(sb)->s_journal->j_inode == NULL) {
1341 char b[BDEVNAME_SIZE];
1342 ext3_msg(sb, KERN_INFO, "using external journal on %s",
1343 bdevname(EXT3_SB(sb)->s_journal->j_dev, b));
1344 } else {
1345 ext3_msg(sb, KERN_INFO, "using internal journal");
1346 }
1347 return res;
1348 }
1349
1350 /* Called at mount-time, super-block is locked */
1351 static int ext3_check_descriptors(struct super_block *sb)
1352 {
1353 struct ext3_sb_info *sbi = EXT3_SB(sb);
1354 int i;
1355
1356 ext3_debug ("Checking group descriptors");
1357
1358 for (i = 0; i < sbi->s_groups_count; i++) {
1359 struct ext3_group_desc *gdp = ext3_get_group_desc(sb, i, NULL);
1360 ext3_fsblk_t first_block = ext3_group_first_block_no(sb, i);
1361 ext3_fsblk_t last_block;
1362
1363 if (i == sbi->s_groups_count - 1)
1364 last_block = le32_to_cpu(sbi->s_es->s_blocks_count) - 1;
1365 else
1366 last_block = first_block +
1367 (EXT3_BLOCKS_PER_GROUP(sb) - 1);
1368
1369 if (le32_to_cpu(gdp->bg_block_bitmap) < first_block ||
1370 le32_to_cpu(gdp->bg_block_bitmap) > last_block)
1371 {
1372 ext3_error (sb, "ext3_check_descriptors",
1373 "Block bitmap for group %d"
1374 " not in group (block %lu)!",
1375 i, (unsigned long)
1376 le32_to_cpu(gdp->bg_block_bitmap));
1377 return 0;
1378 }
1379 if (le32_to_cpu(gdp->bg_inode_bitmap) < first_block ||
1380 le32_to_cpu(gdp->bg_inode_bitmap) > last_block)
1381 {
1382 ext3_error (sb, "ext3_check_descriptors",
1383 "Inode bitmap for group %d"
1384 " not in group (block %lu)!",
1385 i, (unsigned long)
1386 le32_to_cpu(gdp->bg_inode_bitmap));
1387 return 0;
1388 }
1389 if (le32_to_cpu(gdp->bg_inode_table) < first_block ||
1390 le32_to_cpu(gdp->bg_inode_table) + sbi->s_itb_per_group - 1 >
1391 last_block)
1392 {
1393 ext3_error (sb, "ext3_check_descriptors",
1394 "Inode table for group %d"
1395 " not in group (block %lu)!",
1396 i, (unsigned long)
1397 le32_to_cpu(gdp->bg_inode_table));
1398 return 0;
1399 }
1400 }
1401
1402 sbi->s_es->s_free_blocks_count=cpu_to_le32(ext3_count_free_blocks(sb));
1403 sbi->s_es->s_free_inodes_count=cpu_to_le32(ext3_count_free_inodes(sb));
1404 return 1;
1405 }
1406
1407
1408 /* ext3_orphan_cleanup() walks a singly-linked list of inodes (starting at
1409 * the superblock) which were deleted from all directories, but held open by
1410 * a process at the time of a crash. We walk the list and try to delete these
1411 * inodes at recovery time (only with a read-write filesystem).
1412 *
1413 * In order to keep the orphan inode chain consistent during traversal (in
1414 * case of crash during recovery), we link each inode into the superblock
1415 * orphan list_head and handle it the same way as an inode deletion during
1416 * normal operation (which journals the operations for us).
1417 *
1418 * We only do an iget() and an iput() on each inode, which is very safe if we
1419 * accidentally point at an in-use or already deleted inode. The worst that
1420 * can happen in this case is that we get a "bit already cleared" message from
1421 * ext3_free_inode(). The only reason we would point at a wrong inode is if
1422 * e2fsck was run on this filesystem, and it must have already done the orphan
1423 * inode cleanup for us, so we can safely abort without any further action.
1424 */
1425 static void ext3_orphan_cleanup (struct super_block * sb,
1426 struct ext3_super_block * es)
1427 {
1428 unsigned int s_flags = sb->s_flags;
1429 int nr_orphans = 0, nr_truncates = 0;
1430 #ifdef CONFIG_QUOTA
1431 int i;
1432 #endif
1433 if (!es->s_last_orphan) {
1434 jbd_debug(4, "no orphan inodes to clean up\n");
1435 return;
1436 }
1437
1438 if (bdev_read_only(sb->s_bdev)) {
1439 ext3_msg(sb, KERN_ERR, "error: write access "
1440 "unavailable, skipping orphan cleanup.");
1441 return;
1442 }
1443
1444 if (EXT3_SB(sb)->s_mount_state & EXT3_ERROR_FS) {
1445 if (es->s_last_orphan)
1446 jbd_debug(1, "Errors on filesystem, "
1447 "clearing orphan list.\n");
1448 es->s_last_orphan = 0;
1449 jbd_debug(1, "Skipping orphan recovery on fs with errors.\n");
1450 return;
1451 }
1452
1453 if (s_flags & MS_RDONLY) {
1454 ext3_msg(sb, KERN_INFO, "orphan cleanup on readonly fs");
1455 sb->s_flags &= ~MS_RDONLY;
1456 }
1457 #ifdef CONFIG_QUOTA
1458 /* Needed for iput() to work correctly and not trash data */
1459 sb->s_flags |= MS_ACTIVE;
1460 /* Turn on quotas so that they are updated correctly */
1461 for (i = 0; i < MAXQUOTAS; i++) {
1462 if (EXT3_SB(sb)->s_qf_names[i]) {
1463 int ret = ext3_quota_on_mount(sb, i);
1464 if (ret < 0)
1465 ext3_msg(sb, KERN_ERR,
1466 "error: cannot turn on journaled "
1467 "quota: %d", ret);
1468 }
1469 }
1470 #endif
1471
1472 while (es->s_last_orphan) {
1473 struct inode *inode;
1474
1475 inode = ext3_orphan_get(sb, le32_to_cpu(es->s_last_orphan));
1476 if (IS_ERR(inode)) {
1477 es->s_last_orphan = 0;
1478 break;
1479 }
1480
1481 list_add(&EXT3_I(inode)->i_orphan, &EXT3_SB(sb)->s_orphan);
1482 dquot_initialize(inode);
1483 if (inode->i_nlink) {
1484 printk(KERN_DEBUG
1485 "%s: truncating inode %lu to %Ld bytes\n",
1486 __func__, inode->i_ino, inode->i_size);
1487 jbd_debug(2, "truncating inode %lu to %Ld bytes\n",
1488 inode->i_ino, inode->i_size);
1489 ext3_truncate(inode);
1490 nr_truncates++;
1491 } else {
1492 printk(KERN_DEBUG
1493 "%s: deleting unreferenced inode %lu\n",
1494 __func__, inode->i_ino);
1495 jbd_debug(2, "deleting unreferenced inode %lu\n",
1496 inode->i_ino);
1497 nr_orphans++;
1498 }
1499 iput(inode); /* The delete magic happens here! */
1500 }
1501
1502 #define PLURAL(x) (x), ((x)==1) ? "" : "s"
1503
1504 if (nr_orphans)
1505 ext3_msg(sb, KERN_INFO, "%d orphan inode%s deleted",
1506 PLURAL(nr_orphans));
1507 if (nr_truncates)
1508 ext3_msg(sb, KERN_INFO, "%d truncate%s cleaned up",
1509 PLURAL(nr_truncates));
1510 #ifdef CONFIG_QUOTA
1511 /* Turn quotas off */
1512 for (i = 0; i < MAXQUOTAS; i++) {
1513 if (sb_dqopt(sb)->files[i])
1514 vfs_quota_off(sb, i, 0);
1515 }
1516 #endif
1517 sb->s_flags = s_flags; /* Restore MS_RDONLY status */
1518 }
1519
1520 /*
1521 * Maximal file size. There is a direct, and {,double-,triple-}indirect
1522 * block limit, and also a limit of (2^32 - 1) 512-byte sectors in i_blocks.
1523 * We need to be 1 filesystem block less than the 2^32 sector limit.
1524 */
1525 static loff_t ext3_max_size(int bits)
1526 {
1527 loff_t res = EXT3_NDIR_BLOCKS;
1528 int meta_blocks;
1529 loff_t upper_limit;
1530
1531 /* This is calculated to be the largest file size for a
1532 * dense, file such that the total number of
1533 * sectors in the file, including data and all indirect blocks,
1534 * does not exceed 2^32 -1
1535 * __u32 i_blocks representing the total number of
1536 * 512 bytes blocks of the file
1537 */
1538 upper_limit = (1LL << 32) - 1;
1539
1540 /* total blocks in file system block size */
1541 upper_limit >>= (bits - 9);
1542
1543
1544 /* indirect blocks */
1545 meta_blocks = 1;
1546 /* double indirect blocks */
1547 meta_blocks += 1 + (1LL << (bits-2));
1548 /* tripple indirect blocks */
1549 meta_blocks += 1 + (1LL << (bits-2)) + (1LL << (2*(bits-2)));
1550
1551 upper_limit -= meta_blocks;
1552 upper_limit <<= bits;
1553
1554 res += 1LL << (bits-2);
1555 res += 1LL << (2*(bits-2));
1556 res += 1LL << (3*(bits-2));
1557 res <<= bits;
1558 if (res > upper_limit)
1559 res = upper_limit;
1560
1561 if (res > MAX_LFS_FILESIZE)
1562 res = MAX_LFS_FILESIZE;
1563
1564 return res;
1565 }
1566
1567 static ext3_fsblk_t descriptor_loc(struct super_block *sb,
1568 ext3_fsblk_t logic_sb_block,
1569 int nr)
1570 {
1571 struct ext3_sb_info *sbi = EXT3_SB(sb);
1572 unsigned long bg, first_meta_bg;
1573 int has_super = 0;
1574
1575 first_meta_bg = le32_to_cpu(sbi->s_es->s_first_meta_bg);
1576
1577 if (!EXT3_HAS_INCOMPAT_FEATURE(sb, EXT3_FEATURE_INCOMPAT_META_BG) ||
1578 nr < first_meta_bg)
1579 return (logic_sb_block + nr + 1);
1580 bg = sbi->s_desc_per_block * nr;
1581 if (ext3_bg_has_super(sb, bg))
1582 has_super = 1;
1583 return (has_super + ext3_group_first_block_no(sb, bg));
1584 }
1585
1586
1587 static int ext3_fill_super (struct super_block *sb, void *data, int silent)
1588 {
1589 struct buffer_head * bh;
1590 struct ext3_super_block *es = NULL;
1591 struct ext3_sb_info *sbi;
1592 ext3_fsblk_t block;
1593 ext3_fsblk_t sb_block = get_sb_block(&data, sb);
1594 ext3_fsblk_t logic_sb_block;
1595 unsigned long offset = 0;
1596 unsigned int journal_inum = 0;
1597 unsigned long journal_devnum = 0;
1598 unsigned long def_mount_opts;
1599 struct inode *root;
1600 int blocksize;
1601 int hblock;
1602 int db_count;
1603 int i;
1604 int needs_recovery;
1605 int ret = -EINVAL;
1606 __le32 features;
1607 int err;
1608
1609 sbi = kzalloc(sizeof(*sbi), GFP_KERNEL);
1610 if (!sbi)
1611 return -ENOMEM;
1612
1613 sbi->s_blockgroup_lock =
1614 kzalloc(sizeof(struct blockgroup_lock), GFP_KERNEL);
1615 if (!sbi->s_blockgroup_lock) {
1616 kfree(sbi);
1617 return -ENOMEM;
1618 }
1619 sb->s_fs_info = sbi;
1620 sbi->s_mount_opt = 0;
1621 sbi->s_resuid = EXT3_DEF_RESUID;
1622 sbi->s_resgid = EXT3_DEF_RESGID;
1623 sbi->s_sb_block = sb_block;
1624
1625 unlock_kernel();
1626
1627 blocksize = sb_min_blocksize(sb, EXT3_MIN_BLOCK_SIZE);
1628 if (!blocksize) {
1629 ext3_msg(sb, KERN_ERR, "error: unable to set blocksize");
1630 goto out_fail;
1631 }
1632
1633 /*
1634 * The ext3 superblock will not be buffer aligned for other than 1kB
1635 * block sizes. We need to calculate the offset from buffer start.
1636 */
1637 if (blocksize != EXT3_MIN_BLOCK_SIZE) {
1638 logic_sb_block = (sb_block * EXT3_MIN_BLOCK_SIZE) / blocksize;
1639 offset = (sb_block * EXT3_MIN_BLOCK_SIZE) % blocksize;
1640 } else {
1641 logic_sb_block = sb_block;
1642 }
1643
1644 if (!(bh = sb_bread(sb, logic_sb_block))) {
1645 ext3_msg(sb, KERN_ERR, "error: unable to read superblock");
1646 goto out_fail;
1647 }
1648 /*
1649 * Note: s_es must be initialized as soon as possible because
1650 * some ext3 macro-instructions depend on its value
1651 */
1652 es = (struct ext3_super_block *) (((char *)bh->b_data) + offset);
1653 sbi->s_es = es;
1654 sb->s_magic = le16_to_cpu(es->s_magic);
1655 if (sb->s_magic != EXT3_SUPER_MAGIC)
1656 goto cantfind_ext3;
1657
1658 /* Set defaults before we parse the mount options */
1659 def_mount_opts = le32_to_cpu(es->s_default_mount_opts);
1660 if (def_mount_opts & EXT3_DEFM_DEBUG)
1661 set_opt(sbi->s_mount_opt, DEBUG);
1662 if (def_mount_opts & EXT3_DEFM_BSDGROUPS)
1663 set_opt(sbi->s_mount_opt, GRPID);
1664 if (def_mount_opts & EXT3_DEFM_UID16)
1665 set_opt(sbi->s_mount_opt, NO_UID32);
1666 #ifdef CONFIG_EXT3_FS_XATTR
1667 if (def_mount_opts & EXT3_DEFM_XATTR_USER)
1668 set_opt(sbi->s_mount_opt, XATTR_USER);
1669 #endif
1670 #ifdef CONFIG_EXT3_FS_POSIX_ACL
1671 if (def_mount_opts & EXT3_DEFM_ACL)
1672 set_opt(sbi->s_mount_opt, POSIX_ACL);
1673 #endif
1674 if ((def_mount_opts & EXT3_DEFM_JMODE) == EXT3_DEFM_JMODE_DATA)
1675 set_opt(sbi->s_mount_opt, JOURNAL_DATA);
1676 else if ((def_mount_opts & EXT3_DEFM_JMODE) == EXT3_DEFM_JMODE_ORDERED)
1677 set_opt(sbi->s_mount_opt, ORDERED_DATA);
1678 else if ((def_mount_opts & EXT3_DEFM_JMODE) == EXT3_DEFM_JMODE_WBACK)
1679 set_opt(sbi->s_mount_opt, WRITEBACK_DATA);
1680
1681 if (le16_to_cpu(sbi->s_es->s_errors) == EXT3_ERRORS_PANIC)
1682 set_opt(sbi->s_mount_opt, ERRORS_PANIC);
1683 else if (le16_to_cpu(sbi->s_es->s_errors) == EXT3_ERRORS_CONTINUE)
1684 set_opt(sbi->s_mount_opt, ERRORS_CONT);
1685 else
1686 set_opt(sbi->s_mount_opt, ERRORS_RO);
1687
1688 sbi->s_resuid = le16_to_cpu(es->s_def_resuid);
1689 sbi->s_resgid = le16_to_cpu(es->s_def_resgid);
1690
1691 set_opt(sbi->s_mount_opt, RESERVATION);
1692
1693 if (!parse_options ((char *) data, sb, &journal_inum, &journal_devnum,
1694 NULL, 0))
1695 goto failed_mount;
1696
1697 sb->s_flags = (sb->s_flags & ~MS_POSIXACL) |
1698 (test_opt(sb, POSIX_ACL) ? MS_POSIXACL : 0);
1699
1700 if (le32_to_cpu(es->s_rev_level) == EXT3_GOOD_OLD_REV &&
1701 (EXT3_HAS_COMPAT_FEATURE(sb, ~0U) ||
1702 EXT3_HAS_RO_COMPAT_FEATURE(sb, ~0U) ||
1703 EXT3_HAS_INCOMPAT_FEATURE(sb, ~0U)))
1704 ext3_msg(sb, KERN_WARNING,
1705 "warning: feature flags set on rev 0 fs, "
1706 "running e2fsck is recommended");
1707 /*
1708 * Check feature flags regardless of the revision level, since we
1709 * previously didn't change the revision level when setting the flags,
1710 * so there is a chance incompat flags are set on a rev 0 filesystem.
1711 */
1712 features = EXT3_HAS_INCOMPAT_FEATURE(sb, ~EXT3_FEATURE_INCOMPAT_SUPP);
1713 if (features) {
1714 ext3_msg(sb, KERN_ERR,
1715 "error: couldn't mount because of unsupported "
1716 "optional features (%x)", le32_to_cpu(features));
1717 goto failed_mount;
1718 }
1719 features = EXT3_HAS_RO_COMPAT_FEATURE(sb, ~EXT3_FEATURE_RO_COMPAT_SUPP);
1720 if (!(sb->s_flags & MS_RDONLY) && features) {
1721 ext3_msg(sb, KERN_ERR,
1722 "error: couldn't mount RDWR because of unsupported "
1723 "optional features (%x)", le32_to_cpu(features));
1724 goto failed_mount;
1725 }
1726 blocksize = BLOCK_SIZE << le32_to_cpu(es->s_log_block_size);
1727
1728 if (blocksize < EXT3_MIN_BLOCK_SIZE ||
1729 blocksize > EXT3_MAX_BLOCK_SIZE) {
1730 ext3_msg(sb, KERN_ERR,
1731 "error: couldn't mount because of unsupported "
1732 "filesystem blocksize %d", blocksize);
1733 goto failed_mount;
1734 }
1735
1736 hblock = bdev_logical_block_size(sb->s_bdev);
1737 if (sb->s_blocksize != blocksize) {
1738 /*
1739 * Make sure the blocksize for the filesystem is larger
1740 * than the hardware sectorsize for the machine.
1741 */
1742 if (blocksize < hblock) {
1743 ext3_msg(sb, KERN_ERR,
1744 "error: fsblocksize %d too small for "
1745 "hardware sectorsize %d", blocksize, hblock);
1746 goto failed_mount;
1747 }
1748
1749 brelse (bh);
1750 if (!sb_set_blocksize(sb, blocksize)) {
1751 ext3_msg(sb, KERN_ERR,
1752 "error: bad blocksize %d", blocksize);
1753 goto out_fail;
1754 }
1755 logic_sb_block = (sb_block * EXT3_MIN_BLOCK_SIZE) / blocksize;
1756 offset = (sb_block * EXT3_MIN_BLOCK_SIZE) % blocksize;
1757 bh = sb_bread(sb, logic_sb_block);
1758 if (!bh) {
1759 ext3_msg(sb, KERN_ERR,
1760 "error: can't read superblock on 2nd try");
1761 goto failed_mount;
1762 }
1763 es = (struct ext3_super_block *)(((char *)bh->b_data) + offset);
1764 sbi->s_es = es;
1765 if (es->s_magic != cpu_to_le16(EXT3_SUPER_MAGIC)) {
1766 ext3_msg(sb, KERN_ERR,
1767 "error: magic mismatch");
1768 goto failed_mount;
1769 }
1770 }
1771
1772 sb->s_maxbytes = ext3_max_size(sb->s_blocksize_bits);
1773
1774 if (le32_to_cpu(es->s_rev_level) == EXT3_GOOD_OLD_REV) {
1775 sbi->s_inode_size = EXT3_GOOD_OLD_INODE_SIZE;
1776 sbi->s_first_ino = EXT3_GOOD_OLD_FIRST_INO;
1777 } else {
1778 sbi->s_inode_size = le16_to_cpu(es->s_inode_size);
1779 sbi->s_first_ino = le32_to_cpu(es->s_first_ino);
1780 if ((sbi->s_inode_size < EXT3_GOOD_OLD_INODE_SIZE) ||
1781 (!is_power_of_2(sbi->s_inode_size)) ||
1782 (sbi->s_inode_size > blocksize)) {
1783 ext3_msg(sb, KERN_ERR,
1784 "error: unsupported inode size: %d",
1785 sbi->s_inode_size);
1786 goto failed_mount;
1787 }
1788 }
1789 sbi->s_frag_size = EXT3_MIN_FRAG_SIZE <<
1790 le32_to_cpu(es->s_log_frag_size);
1791 if (blocksize != sbi->s_frag_size) {
1792 ext3_msg(sb, KERN_ERR,
1793 "error: fragsize %lu != blocksize %u (unsupported)",
1794 sbi->s_frag_size, blocksize);
1795 goto failed_mount;
1796 }
1797 sbi->s_frags_per_block = 1;
1798 sbi->s_blocks_per_group = le32_to_cpu(es->s_blocks_per_group);
1799 sbi->s_frags_per_group = le32_to_cpu(es->s_frags_per_group);
1800 sbi->s_inodes_per_group = le32_to_cpu(es->s_inodes_per_group);
1801 if (EXT3_INODE_SIZE(sb) == 0 || EXT3_INODES_PER_GROUP(sb) == 0)
1802 goto cantfind_ext3;
1803 sbi->s_inodes_per_block = blocksize / EXT3_INODE_SIZE(sb);
1804 if (sbi->s_inodes_per_block == 0)
1805 goto cantfind_ext3;
1806 sbi->s_itb_per_group = sbi->s_inodes_per_group /
1807 sbi->s_inodes_per_block;
1808 sbi->s_desc_per_block = blocksize / sizeof(struct ext3_group_desc);
1809 sbi->s_sbh = bh;
1810 sbi->s_mount_state = le16_to_cpu(es->s_state);
1811 sbi->s_addr_per_block_bits = ilog2(EXT3_ADDR_PER_BLOCK(sb));
1812 sbi->s_desc_per_block_bits = ilog2(EXT3_DESC_PER_BLOCK(sb));
1813 for (i=0; i < 4; i++)
1814 sbi->s_hash_seed[i] = le32_to_cpu(es->s_hash_seed[i]);
1815 sbi->s_def_hash_version = es->s_def_hash_version;
1816 i = le32_to_cpu(es->s_flags);
1817 if (i & EXT2_FLAGS_UNSIGNED_HASH)
1818 sbi->s_hash_unsigned = 3;
1819 else if ((i & EXT2_FLAGS_SIGNED_HASH) == 0) {
1820 #ifdef __CHAR_UNSIGNED__
1821 es->s_flags |= cpu_to_le32(EXT2_FLAGS_UNSIGNED_HASH);
1822 sbi->s_hash_unsigned = 3;
1823 #else
1824 es->s_flags |= cpu_to_le32(EXT2_FLAGS_SIGNED_HASH);
1825 #endif
1826 }
1827
1828 if (sbi->s_blocks_per_group > blocksize * 8) {
1829 ext3_msg(sb, KERN_ERR,
1830 "#blocks per group too big: %lu",
1831 sbi->s_blocks_per_group);
1832 goto failed_mount;
1833 }
1834 if (sbi->s_frags_per_group > blocksize * 8) {
1835 ext3_msg(sb, KERN_ERR,
1836 "error: #fragments per group too big: %lu",
1837 sbi->s_frags_per_group);
1838 goto failed_mount;
1839 }
1840 if (sbi->s_inodes_per_group > blocksize * 8) {
1841 ext3_msg(sb, KERN_ERR,
1842 "error: #inodes per group too big: %lu",
1843 sbi->s_inodes_per_group);
1844 goto failed_mount;
1845 }
1846
1847 if (le32_to_cpu(es->s_blocks_count) >
1848 (sector_t)(~0ULL) >> (sb->s_blocksize_bits - 9)) {
1849 ext3_msg(sb, KERN_ERR,
1850 "error: filesystem is too large to mount safely");
1851 if (sizeof(sector_t) < 8)
1852 ext3_msg(sb, KERN_ERR,
1853 "error: CONFIG_LBDAF not enabled");
1854 goto failed_mount;
1855 }
1856
1857 if (EXT3_BLOCKS_PER_GROUP(sb) == 0)
1858 goto cantfind_ext3;
1859 sbi->s_groups_count = ((le32_to_cpu(es->s_blocks_count) -
1860 le32_to_cpu(es->s_first_data_block) - 1)
1861 / EXT3_BLOCKS_PER_GROUP(sb)) + 1;
1862 db_count = (sbi->s_groups_count + EXT3_DESC_PER_BLOCK(sb) - 1) /
1863 EXT3_DESC_PER_BLOCK(sb);
1864 sbi->s_group_desc = kmalloc(db_count * sizeof (struct buffer_head *),
1865 GFP_KERNEL);
1866 if (sbi->s_group_desc == NULL) {
1867 ext3_msg(sb, KERN_ERR,
1868 "error: not enough memory");
1869 goto failed_mount;
1870 }
1871
1872 bgl_lock_init(sbi->s_blockgroup_lock);
1873
1874 for (i = 0; i < db_count; i++) {
1875 block = descriptor_loc(sb, logic_sb_block, i);
1876 sbi->s_group_desc[i] = sb_bread(sb, block);
1877 if (!sbi->s_group_desc[i]) {
1878 ext3_msg(sb, KERN_ERR,
1879 "error: can't read group descriptor %d", i);
1880 db_count = i;
1881 goto failed_mount2;
1882 }
1883 }
1884 if (!ext3_check_descriptors (sb)) {
1885 ext3_msg(sb, KERN_ERR,
1886 "error: group descriptors corrupted");
1887 goto failed_mount2;
1888 }
1889 sbi->s_gdb_count = db_count;
1890 get_random_bytes(&sbi->s_next_generation, sizeof(u32));
1891 spin_lock_init(&sbi->s_next_gen_lock);
1892
1893 err = percpu_counter_init(&sbi->s_freeblocks_counter,
1894 ext3_count_free_blocks(sb));
1895 if (!err) {
1896 err = percpu_counter_init(&sbi->s_freeinodes_counter,
1897 ext3_count_free_inodes(sb));
1898 }
1899 if (!err) {
1900 err = percpu_counter_init(&sbi->s_dirs_counter,
1901 ext3_count_dirs(sb));
1902 }
1903 if (err) {
1904 ext3_msg(sb, KERN_ERR, "error: insufficient memory");
1905 goto failed_mount3;
1906 }
1907
1908 /* per fileystem reservation list head & lock */
1909 spin_lock_init(&sbi->s_rsv_window_lock);
1910 sbi->s_rsv_window_root = RB_ROOT;
1911 /* Add a single, static dummy reservation to the start of the
1912 * reservation window list --- it gives us a placeholder for
1913 * append-at-start-of-list which makes the allocation logic
1914 * _much_ simpler. */
1915 sbi->s_rsv_window_head.rsv_start = EXT3_RESERVE_WINDOW_NOT_ALLOCATED;
1916 sbi->s_rsv_window_head.rsv_end = EXT3_RESERVE_WINDOW_NOT_ALLOCATED;
1917 sbi->s_rsv_window_head.rsv_alloc_hit = 0;
1918 sbi->s_rsv_window_head.rsv_goal_size = 0;
1919 ext3_rsv_window_add(sb, &sbi->s_rsv_window_head);
1920
1921 /*
1922 * set up enough so that it can read an inode
1923 */
1924 sb->s_op = &ext3_sops;
1925 sb->s_export_op = &ext3_export_ops;
1926 sb->s_xattr = ext3_xattr_handlers;
1927 #ifdef CONFIG_QUOTA
1928 sb->s_qcop = &ext3_qctl_operations;
1929 sb->dq_op = &ext3_quota_operations;
1930 #endif
1931 INIT_LIST_HEAD(&sbi->s_orphan); /* unlinked but open files */
1932 mutex_init(&sbi->s_orphan_lock);
1933 mutex_init(&sbi->s_resize_lock);
1934
1935 sb->s_root = NULL;
1936
1937 needs_recovery = (es->s_last_orphan != 0 ||
1938 EXT3_HAS_INCOMPAT_FEATURE(sb,
1939 EXT3_FEATURE_INCOMPAT_RECOVER));
1940
1941 /*
1942 * The first inode we look at is the journal inode. Don't try
1943 * root first: it may be modified in the journal!
1944 */
1945 if (!test_opt(sb, NOLOAD) &&
1946 EXT3_HAS_COMPAT_FEATURE(sb, EXT3_FEATURE_COMPAT_HAS_JOURNAL)) {
1947 if (ext3_load_journal(sb, es, journal_devnum))
1948 goto failed_mount3;
1949 } else if (journal_inum) {
1950 if (ext3_create_journal(sb, es, journal_inum))
1951 goto failed_mount3;
1952 } else {
1953 if (!silent)
1954 ext3_msg(sb, KERN_ERR,
1955 "error: no journal found. "
1956 "mounting ext3 over ext2?");
1957 goto failed_mount3;
1958 }
1959
1960 /* We have now updated the journal if required, so we can
1961 * validate the data journaling mode. */
1962 switch (test_opt(sb, DATA_FLAGS)) {
1963 case 0:
1964 /* No mode set, assume a default based on the journal
1965 capabilities: ORDERED_DATA if the journal can
1966 cope, else JOURNAL_DATA */
1967 if (journal_check_available_features
1968 (sbi->s_journal, 0, 0, JFS_FEATURE_INCOMPAT_REVOKE))
1969 set_opt(sbi->s_mount_opt, DEFAULT_DATA_MODE);
1970 else
1971 set_opt(sbi->s_mount_opt, JOURNAL_DATA);
1972 break;
1973
1974 case EXT3_MOUNT_ORDERED_DATA:
1975 case EXT3_MOUNT_WRITEBACK_DATA:
1976 if (!journal_check_available_features
1977 (sbi->s_journal, 0, 0, JFS_FEATURE_INCOMPAT_REVOKE)) {
1978 ext3_msg(sb, KERN_ERR,
1979 "error: journal does not support "
1980 "requested data journaling mode");
1981 goto failed_mount4;
1982 }
1983 default:
1984 break;
1985 }
1986
1987 if (test_opt(sb, NOBH)) {
1988 if (!(test_opt(sb, DATA_FLAGS) == EXT3_MOUNT_WRITEBACK_DATA)) {
1989 ext3_msg(sb, KERN_WARNING,
1990 "warning: ignoring nobh option - "
1991 "it is supported only with writeback mode");
1992 clear_opt(sbi->s_mount_opt, NOBH);
1993 }
1994 }
1995 /*
1996 * The journal_load will have done any necessary log recovery,
1997 * so we can safely mount the rest of the filesystem now.
1998 */
1999
2000 root = ext3_iget(sb, EXT3_ROOT_INO);
2001 if (IS_ERR(root)) {
2002 ext3_msg(sb, KERN_ERR, "error: get root inode failed");
2003 ret = PTR_ERR(root);
2004 goto failed_mount4;
2005 }
2006 if (!S_ISDIR(root->i_mode) || !root->i_blocks || !root->i_size) {
2007 iput(root);
2008 ext3_msg(sb, KERN_ERR, "error: corrupt root inode, run e2fsck");
2009 goto failed_mount4;
2010 }
2011 sb->s_root = d_alloc_root(root);
2012 if (!sb->s_root) {
2013 ext3_msg(sb, KERN_ERR, "error: get root dentry failed");
2014 iput(root);
2015 ret = -ENOMEM;
2016 goto failed_mount4;
2017 }
2018
2019 ext3_setup_super (sb, es, sb->s_flags & MS_RDONLY);
2020
2021 EXT3_SB(sb)->s_mount_state |= EXT3_ORPHAN_FS;
2022 ext3_orphan_cleanup(sb, es);
2023 EXT3_SB(sb)->s_mount_state &= ~EXT3_ORPHAN_FS;
2024 if (needs_recovery)
2025 ext3_msg(sb, KERN_INFO, "recovery complete");
2026 ext3_mark_recovery_complete(sb, es);
2027 ext3_msg(sb, KERN_INFO, "mounted filesystem with %s data mode",
2028 test_opt(sb,DATA_FLAGS) == EXT3_MOUNT_JOURNAL_DATA ? "journal":
2029 test_opt(sb,DATA_FLAGS) == EXT3_MOUNT_ORDERED_DATA ? "ordered":
2030 "writeback");
2031
2032 lock_kernel();
2033 return 0;
2034
2035 cantfind_ext3:
2036 if (!silent)
2037 ext3_msg(sb, KERN_INFO,
2038 "error: can't find ext3 filesystem on dev %s.",
2039 sb->s_id);
2040 goto failed_mount;
2041
2042 failed_mount4:
2043 journal_destroy(sbi->s_journal);
2044 failed_mount3:
2045 percpu_counter_destroy(&sbi->s_freeblocks_counter);
2046 percpu_counter_destroy(&sbi->s_freeinodes_counter);
2047 percpu_counter_destroy(&sbi->s_dirs_counter);
2048 failed_mount2:
2049 for (i = 0; i < db_count; i++)
2050 brelse(sbi->s_group_desc[i]);
2051 kfree(sbi->s_group_desc);
2052 failed_mount:
2053 #ifdef CONFIG_QUOTA
2054 for (i = 0; i < MAXQUOTAS; i++)
2055 kfree(sbi->s_qf_names[i]);
2056 #endif
2057 ext3_blkdev_remove(sbi);
2058 brelse(bh);
2059 out_fail:
2060 sb->s_fs_info = NULL;
2061 kfree(sbi->s_blockgroup_lock);
2062 kfree(sbi);
2063 lock_kernel();
2064 return ret;
2065 }
2066
2067 /*
2068 * Setup any per-fs journal parameters now. We'll do this both on
2069 * initial mount, once the journal has been initialised but before we've
2070 * done any recovery; and again on any subsequent remount.
2071 */
2072 static void ext3_init_journal_params(struct super_block *sb, journal_t *journal)
2073 {
2074 struct ext3_sb_info *sbi = EXT3_SB(sb);
2075
2076 if (sbi->s_commit_interval)
2077 journal->j_commit_interval = sbi->s_commit_interval;
2078 /* We could also set up an ext3-specific default for the commit
2079 * interval here, but for now we'll just fall back to the jbd
2080 * default. */
2081
2082 spin_lock(&journal->j_state_lock);
2083 if (test_opt(sb, BARRIER))
2084 journal->j_flags |= JFS_BARRIER;
2085 else
2086 journal->j_flags &= ~JFS_BARRIER;
2087 if (test_opt(sb, DATA_ERR_ABORT))
2088 journal->j_flags |= JFS_ABORT_ON_SYNCDATA_ERR;
2089 else
2090 journal->j_flags &= ~JFS_ABORT_ON_SYNCDATA_ERR;
2091 spin_unlock(&journal->j_state_lock);
2092 }
2093
2094 static journal_t *ext3_get_journal(struct super_block *sb,
2095 unsigned int journal_inum)
2096 {
2097 struct inode *journal_inode;
2098 journal_t *journal;
2099
2100 /* First, test for the existence of a valid inode on disk. Bad
2101 * things happen if we iget() an unused inode, as the subsequent
2102 * iput() will try to delete it. */
2103
2104 journal_inode = ext3_iget(sb, journal_inum);
2105 if (IS_ERR(journal_inode)) {
2106 ext3_msg(sb, KERN_ERR, "error: no journal found");
2107 return NULL;
2108 }
2109 if (!journal_inode->i_nlink) {
2110 make_bad_inode(journal_inode);
2111 iput(journal_inode);
2112 ext3_msg(sb, KERN_ERR, "error: journal inode is deleted");
2113 return NULL;
2114 }
2115
2116 jbd_debug(2, "Journal inode found at %p: %Ld bytes\n",
2117 journal_inode, journal_inode->i_size);
2118 if (!S_ISREG(journal_inode->i_mode)) {
2119 ext3_msg(sb, KERN_ERR, "error: invalid journal inode");
2120 iput(journal_inode);
2121 return NULL;
2122 }
2123
2124 journal = journal_init_inode(journal_inode);
2125 if (!journal) {
2126 ext3_msg(sb, KERN_ERR, "error: could not load journal inode");
2127 iput(journal_inode);
2128 return NULL;
2129 }
2130 journal->j_private = sb;
2131 ext3_init_journal_params(sb, journal);
2132 return journal;
2133 }
2134
2135 static journal_t *ext3_get_dev_journal(struct super_block *sb,
2136 dev_t j_dev)
2137 {
2138 struct buffer_head * bh;
2139 journal_t *journal;
2140 ext3_fsblk_t start;
2141 ext3_fsblk_t len;
2142 int hblock, blocksize;
2143 ext3_fsblk_t sb_block;
2144 unsigned long offset;
2145 struct ext3_super_block * es;
2146 struct block_device *bdev;
2147
2148 bdev = ext3_blkdev_get(j_dev, sb);
2149 if (bdev == NULL)
2150 return NULL;
2151
2152 if (bd_claim(bdev, sb)) {
2153 ext3_msg(sb, KERN_ERR,
2154 "error: failed to claim external journal device");
2155 blkdev_put(bdev, FMODE_READ|FMODE_WRITE);
2156 return NULL;
2157 }
2158
2159 blocksize = sb->s_blocksize;
2160 hblock = bdev_logical_block_size(bdev);
2161 if (blocksize < hblock) {
2162 ext3_msg(sb, KERN_ERR,
2163 "error: blocksize too small for journal device");
2164 goto out_bdev;
2165 }
2166
2167 sb_block = EXT3_MIN_BLOCK_SIZE / blocksize;
2168 offset = EXT3_MIN_BLOCK_SIZE % blocksize;
2169 set_blocksize(bdev, blocksize);
2170 if (!(bh = __bread(bdev, sb_block, blocksize))) {
2171 ext3_msg(sb, KERN_ERR, "error: couldn't read superblock of "
2172 "external journal");
2173 goto out_bdev;
2174 }
2175
2176 es = (struct ext3_super_block *) (((char *)bh->b_data) + offset);
2177 if ((le16_to_cpu(es->s_magic) != EXT3_SUPER_MAGIC) ||
2178 !(le32_to_cpu(es->s_feature_incompat) &
2179 EXT3_FEATURE_INCOMPAT_JOURNAL_DEV)) {
2180 ext3_msg(sb, KERN_ERR, "error: external journal has "
2181 "bad superblock");
2182 brelse(bh);
2183 goto out_bdev;
2184 }
2185
2186 if (memcmp(EXT3_SB(sb)->s_es->s_journal_uuid, es->s_uuid, 16)) {
2187 ext3_msg(sb, KERN_ERR, "error: journal UUID does not match");
2188 brelse(bh);
2189 goto out_bdev;
2190 }
2191
2192 len = le32_to_cpu(es->s_blocks_count);
2193 start = sb_block + 1;
2194 brelse(bh); /* we're done with the superblock */
2195
2196 journal = journal_init_dev(bdev, sb->s_bdev,
2197 start, len, blocksize);
2198 if (!journal) {
2199 ext3_msg(sb, KERN_ERR,
2200 "error: failed to create device journal");
2201 goto out_bdev;
2202 }
2203 journal->j_private = sb;
2204 ll_rw_block(READ, 1, &journal->j_sb_buffer);
2205 wait_on_buffer(journal->j_sb_buffer);
2206 if (!buffer_uptodate(journal->j_sb_buffer)) {
2207 ext3_msg(sb, KERN_ERR, "I/O error on journal device");
2208 goto out_journal;
2209 }
2210 if (be32_to_cpu(journal->j_superblock->s_nr_users) != 1) {
2211 ext3_msg(sb, KERN_ERR,
2212 "error: external journal has more than one "
2213 "user (unsupported) - %d",
2214 be32_to_cpu(journal->j_superblock->s_nr_users));
2215 goto out_journal;
2216 }
2217 EXT3_SB(sb)->journal_bdev = bdev;
2218 ext3_init_journal_params(sb, journal);
2219 return journal;
2220 out_journal:
2221 journal_destroy(journal);
2222 out_bdev:
2223 ext3_blkdev_put(bdev);
2224 return NULL;
2225 }
2226
2227 static int ext3_load_journal(struct super_block *sb,
2228 struct ext3_super_block *es,
2229 unsigned long journal_devnum)
2230 {
2231 journal_t *journal;
2232 unsigned int journal_inum = le32_to_cpu(es->s_journal_inum);
2233 dev_t journal_dev;
2234 int err = 0;
2235 int really_read_only;
2236
2237 if (journal_devnum &&
2238 journal_devnum != le32_to_cpu(es->s_journal_dev)) {
2239 ext3_msg(sb, KERN_INFO, "external journal device major/minor "
2240 "numbers have changed");
2241 journal_dev = new_decode_dev(journal_devnum);
2242 } else
2243 journal_dev = new_decode_dev(le32_to_cpu(es->s_journal_dev));
2244
2245 really_read_only = bdev_read_only(sb->s_bdev);
2246
2247 /*
2248 * Are we loading a blank journal or performing recovery after a
2249 * crash? For recovery, we need to check in advance whether we
2250 * can get read-write access to the device.
2251 */
2252
2253 if (EXT3_HAS_INCOMPAT_FEATURE(sb, EXT3_FEATURE_INCOMPAT_RECOVER)) {
2254 if (sb->s_flags & MS_RDONLY) {
2255 ext3_msg(sb, KERN_INFO,
2256 "recovery required on readonly filesystem");
2257 if (really_read_only) {
2258 ext3_msg(sb, KERN_ERR, "error: write access "
2259 "unavailable, cannot proceed");
2260 return -EROFS;
2261 }
2262 ext3_msg(sb, KERN_INFO,
2263 "write access will be enabled during recovery");
2264 }
2265 }
2266
2267 if (journal_inum && journal_dev) {
2268 ext3_msg(sb, KERN_ERR, "error: filesystem has both journal "
2269 "and inode journals");
2270 return -EINVAL;
2271 }
2272
2273 if (journal_inum) {
2274 if (!(journal = ext3_get_journal(sb, journal_inum)))
2275 return -EINVAL;
2276 } else {
2277 if (!(journal = ext3_get_dev_journal(sb, journal_dev)))
2278 return -EINVAL;
2279 }
2280
2281 if (!really_read_only && test_opt(sb, UPDATE_JOURNAL)) {
2282 err = journal_update_format(journal);
2283 if (err) {
2284 ext3_msg(sb, KERN_ERR, "error updating journal");
2285 journal_destroy(journal);
2286 return err;
2287 }
2288 }
2289
2290 if (!EXT3_HAS_INCOMPAT_FEATURE(sb, EXT3_FEATURE_INCOMPAT_RECOVER))
2291 err = journal_wipe(journal, !really_read_only);
2292 if (!err)
2293 err = journal_load(journal);
2294
2295 if (err) {
2296 ext3_msg(sb, KERN_ERR, "error loading journal");
2297 journal_destroy(journal);
2298 return err;
2299 }
2300
2301 EXT3_SB(sb)->s_journal = journal;
2302 ext3_clear_journal_err(sb, es);
2303
2304 if (journal_devnum &&
2305 journal_devnum != le32_to_cpu(es->s_journal_dev)) {
2306 es->s_journal_dev = cpu_to_le32(journal_devnum);
2307
2308 /* Make sure we flush the recovery flag to disk. */
2309 ext3_commit_super(sb, es, 1);
2310 }
2311
2312 return 0;
2313 }
2314
2315 static int ext3_create_journal(struct super_block *sb,
2316 struct ext3_super_block *es,
2317 unsigned int journal_inum)
2318 {
2319 journal_t *journal;
2320 int err;
2321
2322 if (sb->s_flags & MS_RDONLY) {
2323 ext3_msg(sb, KERN_ERR,
2324 "error: readonly filesystem when trying to "
2325 "create journal");
2326 return -EROFS;
2327 }
2328
2329 journal = ext3_get_journal(sb, journal_inum);
2330 if (!journal)
2331 return -EINVAL;
2332
2333 ext3_msg(sb, KERN_INFO, "creating new journal on inode %u",
2334 journal_inum);
2335
2336 err = journal_create(journal);
2337 if (err) {
2338 ext3_msg(sb, KERN_ERR, "error creating journal");
2339 journal_destroy(journal);
2340 return -EIO;
2341 }
2342
2343 EXT3_SB(sb)->s_journal = journal;
2344
2345 ext3_update_dynamic_rev(sb);
2346 EXT3_SET_INCOMPAT_FEATURE(sb, EXT3_FEATURE_INCOMPAT_RECOVER);
2347 EXT3_SET_COMPAT_FEATURE(sb, EXT3_FEATURE_COMPAT_HAS_JOURNAL);
2348
2349 es->s_journal_inum = cpu_to_le32(journal_inum);
2350
2351 /* Make sure we flush the recovery flag to disk. */
2352 ext3_commit_super(sb, es, 1);
2353
2354 return 0;
2355 }
2356
2357 static int ext3_commit_super(struct super_block *sb,
2358 struct ext3_super_block *es,
2359 int sync)
2360 {
2361 struct buffer_head *sbh = EXT3_SB(sb)->s_sbh;
2362 int error = 0;
2363
2364 if (!sbh)
2365 return error;
2366 /*
2367 * If the file system is mounted read-only, don't update the
2368 * superblock write time. This avoids updating the superblock
2369 * write time when we are mounting the root file system
2370 * read/only but we need to replay the journal; at that point,
2371 * for people who are east of GMT and who make their clock
2372 * tick in localtime for Windows bug-for-bug compatibility,
2373 * the clock is set in the future, and this will cause e2fsck
2374 * to complain and force a full file system check.
2375 */
2376 if (!(sb->s_flags & MS_RDONLY))
2377 es->s_wtime = cpu_to_le32(get_seconds());
2378 es->s_free_blocks_count = cpu_to_le32(ext3_count_free_blocks(sb));
2379 es->s_free_inodes_count = cpu_to_le32(ext3_count_free_inodes(sb));
2380 BUFFER_TRACE(sbh, "marking dirty");
2381 mark_buffer_dirty(sbh);
2382 if (sync)
2383 error = sync_dirty_buffer(sbh);
2384 return error;
2385 }
2386
2387
2388 /*
2389 * Have we just finished recovery? If so, and if we are mounting (or
2390 * remounting) the filesystem readonly, then we will end up with a
2391 * consistent fs on disk. Record that fact.
2392 */
2393 static void ext3_mark_recovery_complete(struct super_block * sb,
2394 struct ext3_super_block * es)
2395 {
2396 journal_t *journal = EXT3_SB(sb)->s_journal;
2397
2398 journal_lock_updates(journal);
2399 if (journal_flush(journal) < 0)
2400 goto out;
2401
2402 if (EXT3_HAS_INCOMPAT_FEATURE(sb, EXT3_FEATURE_INCOMPAT_RECOVER) &&
2403 sb->s_flags & MS_RDONLY) {
2404 EXT3_CLEAR_INCOMPAT_FEATURE(sb, EXT3_FEATURE_INCOMPAT_RECOVER);
2405 ext3_commit_super(sb, es, 1);
2406 }
2407
2408 out:
2409 journal_unlock_updates(journal);
2410 }
2411
2412 /*
2413 * If we are mounting (or read-write remounting) a filesystem whose journal
2414 * has recorded an error from a previous lifetime, move that error to the
2415 * main filesystem now.
2416 */
2417 static void ext3_clear_journal_err(struct super_block *sb,
2418 struct ext3_super_block *es)
2419 {
2420 journal_t *journal;
2421 int j_errno;
2422 const char *errstr;
2423
2424 journal = EXT3_SB(sb)->s_journal;
2425
2426 /*
2427 * Now check for any error status which may have been recorded in the
2428 * journal by a prior ext3_error() or ext3_abort()
2429 */
2430
2431 j_errno = journal_errno(journal);
2432 if (j_errno) {
2433 char nbuf[16];
2434
2435 errstr = ext3_decode_error(sb, j_errno, nbuf);
2436 ext3_warning(sb, __func__, "Filesystem error recorded "
2437 "from previous mount: %s", errstr);
2438 ext3_warning(sb, __func__, "Marking fs in need of "
2439 "filesystem check.");
2440
2441 EXT3_SB(sb)->s_mount_state |= EXT3_ERROR_FS;
2442 es->s_state |= cpu_to_le16(EXT3_ERROR_FS);
2443 ext3_commit_super (sb, es, 1);
2444
2445 journal_clear_err(journal);
2446 }
2447 }
2448
2449 /*
2450 * Force the running and committing transactions to commit,
2451 * and wait on the commit.
2452 */
2453 int ext3_force_commit(struct super_block *sb)
2454 {
2455 journal_t *journal;
2456 int ret;
2457
2458 if (sb->s_flags & MS_RDONLY)
2459 return 0;
2460
2461 journal = EXT3_SB(sb)->s_journal;
2462 ret = ext3_journal_force_commit(journal);
2463 return ret;
2464 }
2465
2466 static int ext3_sync_fs(struct super_block *sb, int wait)
2467 {
2468 tid_t target;
2469
2470 if (journal_start_commit(EXT3_SB(sb)->s_journal, &target)) {
2471 if (wait)
2472 log_wait_commit(EXT3_SB(sb)->s_journal, target);
2473 }
2474 return 0;
2475 }
2476
2477 /*
2478 * LVM calls this function before a (read-only) snapshot is created. This
2479 * gives us a chance to flush the journal completely and mark the fs clean.
2480 */
2481 static int ext3_freeze(struct super_block *sb)
2482 {
2483 int error = 0;
2484 journal_t *journal;
2485
2486 if (!(sb->s_flags & MS_RDONLY)) {
2487 journal = EXT3_SB(sb)->s_journal;
2488
2489 /* Now we set up the journal barrier. */
2490 journal_lock_updates(journal);
2491
2492 /*
2493 * We don't want to clear needs_recovery flag when we failed
2494 * to flush the journal.
2495 */
2496 error = journal_flush(journal);
2497 if (error < 0)
2498 goto out;
2499
2500 /* Journal blocked and flushed, clear needs_recovery flag. */
2501 EXT3_CLEAR_INCOMPAT_FEATURE(sb, EXT3_FEATURE_INCOMPAT_RECOVER);
2502 error = ext3_commit_super(sb, EXT3_SB(sb)->s_es, 1);
2503 if (error)
2504 goto out;
2505 }
2506 return 0;
2507
2508 out:
2509 journal_unlock_updates(journal);
2510 return error;
2511 }
2512
2513 /*
2514 * Called by LVM after the snapshot is done. We need to reset the RECOVER
2515 * flag here, even though the filesystem is not technically dirty yet.
2516 */
2517 static int ext3_unfreeze(struct super_block *sb)
2518 {
2519 if (!(sb->s_flags & MS_RDONLY)) {
2520 lock_super(sb);
2521 /* Reser the needs_recovery flag before the fs is unlocked. */
2522 EXT3_SET_INCOMPAT_FEATURE(sb, EXT3_FEATURE_INCOMPAT_RECOVER);
2523 ext3_commit_super(sb, EXT3_SB(sb)->s_es, 1);
2524 unlock_super(sb);
2525 journal_unlock_updates(EXT3_SB(sb)->s_journal);
2526 }
2527 return 0;
2528 }
2529
2530 static int ext3_remount (struct super_block * sb, int * flags, char * data)
2531 {
2532 struct ext3_super_block * es;
2533 struct ext3_sb_info *sbi = EXT3_SB(sb);
2534 ext3_fsblk_t n_blocks_count = 0;
2535 unsigned long old_sb_flags;
2536 struct ext3_mount_options old_opts;
2537 int err;
2538 #ifdef CONFIG_QUOTA
2539 int i;
2540 #endif
2541
2542 lock_kernel();
2543
2544 /* Store the original options */
2545 lock_super(sb);
2546 old_sb_flags = sb->s_flags;
2547 old_opts.s_mount_opt = sbi->s_mount_opt;
2548 old_opts.s_resuid = sbi->s_resuid;
2549 old_opts.s_resgid = sbi->s_resgid;
2550 old_opts.s_commit_interval = sbi->s_commit_interval;
2551 #ifdef CONFIG_QUOTA
2552 old_opts.s_jquota_fmt = sbi->s_jquota_fmt;
2553 for (i = 0; i < MAXQUOTAS; i++)
2554 old_opts.s_qf_names[i] = sbi->s_qf_names[i];
2555 #endif
2556
2557 /*
2558 * Allow the "check" option to be passed as a remount option.
2559 */
2560 if (!parse_options(data, sb, NULL, NULL, &n_blocks_count, 1)) {
2561 err = -EINVAL;
2562 goto restore_opts;
2563 }
2564
2565 if (test_opt(sb, ABORT))
2566 ext3_abort(sb, __func__, "Abort forced by user");
2567
2568 sb->s_flags = (sb->s_flags & ~MS_POSIXACL) |
2569 (test_opt(sb, POSIX_ACL) ? MS_POSIXACL : 0);
2570
2571 es = sbi->s_es;
2572
2573 ext3_init_journal_params(sb, sbi->s_journal);
2574
2575 if ((*flags & MS_RDONLY) != (sb->s_flags & MS_RDONLY) ||
2576 n_blocks_count > le32_to_cpu(es->s_blocks_count)) {
2577 if (test_opt(sb, ABORT)) {
2578 err = -EROFS;
2579 goto restore_opts;
2580 }
2581
2582 if (*flags & MS_RDONLY) {
2583 /*
2584 * First of all, the unconditional stuff we have to do
2585 * to disable replay of the journal when we next remount
2586 */
2587 sb->s_flags |= MS_RDONLY;
2588
2589 /*
2590 * OK, test if we are remounting a valid rw partition
2591 * readonly, and if so set the rdonly flag and then
2592 * mark the partition as valid again.
2593 */
2594 if (!(es->s_state & cpu_to_le16(EXT3_VALID_FS)) &&
2595 (sbi->s_mount_state & EXT3_VALID_FS))
2596 es->s_state = cpu_to_le16(sbi->s_mount_state);
2597
2598 ext3_mark_recovery_complete(sb, es);
2599 } else {
2600 __le32 ret;
2601 if ((ret = EXT3_HAS_RO_COMPAT_FEATURE(sb,
2602 ~EXT3_FEATURE_RO_COMPAT_SUPP))) {
2603 ext3_msg(sb, KERN_WARNING,
2604 "warning: couldn't remount RDWR "
2605 "because of unsupported optional "
2606 "features (%x)", le32_to_cpu(ret));
2607 err = -EROFS;
2608 goto restore_opts;
2609 }
2610
2611 /*
2612 * If we have an unprocessed orphan list hanging
2613 * around from a previously readonly bdev mount,
2614 * require a full umount/remount for now.
2615 */
2616 if (es->s_last_orphan) {
2617 ext3_msg(sb, KERN_WARNING, "warning: couldn't "
2618 "remount RDWR because of unprocessed "
2619 "orphan inode list. Please "
2620 "umount/remount instead.");
2621 err = -EINVAL;
2622 goto restore_opts;
2623 }
2624
2625 /*
2626 * Mounting a RDONLY partition read-write, so reread
2627 * and store the current valid flag. (It may have
2628 * been changed by e2fsck since we originally mounted
2629 * the partition.)
2630 */
2631 ext3_clear_journal_err(sb, es);
2632 sbi->s_mount_state = le16_to_cpu(es->s_state);
2633 if ((err = ext3_group_extend(sb, es, n_blocks_count)))
2634 goto restore_opts;
2635 if (!ext3_setup_super (sb, es, 0))
2636 sb->s_flags &= ~MS_RDONLY;
2637 }
2638 }
2639 #ifdef CONFIG_QUOTA
2640 /* Release old quota file names */
2641 for (i = 0; i < MAXQUOTAS; i++)
2642 if (old_opts.s_qf_names[i] &&
2643 old_opts.s_qf_names[i] != sbi->s_qf_names[i])
2644 kfree(old_opts.s_qf_names[i]);
2645 #endif
2646 unlock_super(sb);
2647 unlock_kernel();
2648 return 0;
2649 restore_opts:
2650 sb->s_flags = old_sb_flags;
2651 sbi->s_mount_opt = old_opts.s_mount_opt;
2652 sbi->s_resuid = old_opts.s_resuid;
2653 sbi->s_resgid = old_opts.s_resgid;
2654 sbi->s_commit_interval = old_opts.s_commit_interval;
2655 #ifdef CONFIG_QUOTA
2656 sbi->s_jquota_fmt = old_opts.s_jquota_fmt;
2657 for (i = 0; i < MAXQUOTAS; i++) {
2658 if (sbi->s_qf_names[i] &&
2659 old_opts.s_qf_names[i] != sbi->s_qf_names[i])
2660 kfree(sbi->s_qf_names[i]);
2661 sbi->s_qf_names[i] = old_opts.s_qf_names[i];
2662 }
2663 #endif
2664 unlock_super(sb);
2665 unlock_kernel();
2666 return err;
2667 }
2668
2669 static int ext3_statfs (struct dentry * dentry, struct kstatfs * buf)
2670 {
2671 struct super_block *sb = dentry->d_sb;
2672 struct ext3_sb_info *sbi = EXT3_SB(sb);
2673 struct ext3_super_block *es = sbi->s_es;
2674 u64 fsid;
2675
2676 if (test_opt(sb, MINIX_DF)) {
2677 sbi->s_overhead_last = 0;
2678 } else if (sbi->s_blocks_last != le32_to_cpu(es->s_blocks_count)) {
2679 unsigned long ngroups = sbi->s_groups_count, i;
2680 ext3_fsblk_t overhead = 0;
2681 smp_rmb();
2682
2683 /*
2684 * Compute the overhead (FS structures). This is constant
2685 * for a given filesystem unless the number of block groups
2686 * changes so we cache the previous value until it does.
2687 */
2688
2689 /*
2690 * All of the blocks before first_data_block are
2691 * overhead
2692 */
2693 overhead = le32_to_cpu(es->s_first_data_block);
2694
2695 /*
2696 * Add the overhead attributed to the superblock and
2697 * block group descriptors. If the sparse superblocks
2698 * feature is turned on, then not all groups have this.
2699 */
2700 for (i = 0; i < ngroups; i++) {
2701 overhead += ext3_bg_has_super(sb, i) +
2702 ext3_bg_num_gdb(sb, i);
2703 cond_resched();
2704 }
2705
2706 /*
2707 * Every block group has an inode bitmap, a block
2708 * bitmap, and an inode table.
2709 */
2710 overhead += ngroups * (2 + sbi->s_itb_per_group);
2711 sbi->s_overhead_last = overhead;
2712 smp_wmb();
2713 sbi->s_blocks_last = le32_to_cpu(es->s_blocks_count);
2714 }
2715
2716 buf->f_type = EXT3_SUPER_MAGIC;
2717 buf->f_bsize = sb->s_blocksize;
2718 buf->f_blocks = le32_to_cpu(es->s_blocks_count) - sbi->s_overhead_last;
2719 buf->f_bfree = percpu_counter_sum_positive(&sbi->s_freeblocks_counter);
2720 buf->f_bavail = buf->f_bfree - le32_to_cpu(es->s_r_blocks_count);
2721 if (buf->f_bfree < le32_to_cpu(es->s_r_blocks_count))
2722 buf->f_bavail = 0;
2723 buf->f_files = le32_to_cpu(es->s_inodes_count);
2724 buf->f_ffree = percpu_counter_sum_positive(&sbi->s_freeinodes_counter);
2725 buf->f_namelen = EXT3_NAME_LEN;
2726 fsid = le64_to_cpup((void *)es->s_uuid) ^
2727 le64_to_cpup((void *)es->s_uuid + sizeof(u64));
2728 buf->f_fsid.val[0] = fsid & 0xFFFFFFFFUL;
2729 buf->f_fsid.val[1] = (fsid >> 32) & 0xFFFFFFFFUL;
2730 return 0;
2731 }
2732
2733 /* Helper function for writing quotas on sync - we need to start transaction before quota file
2734 * is locked for write. Otherwise the are possible deadlocks:
2735 * Process 1 Process 2
2736 * ext3_create() quota_sync()
2737 * journal_start() write_dquot()
2738 * dquot_initialize() down(dqio_mutex)
2739 * down(dqio_mutex) journal_start()
2740 *
2741 */
2742
2743 #ifdef CONFIG_QUOTA
2744
2745 static inline struct inode *dquot_to_inode(struct dquot *dquot)
2746 {
2747 return sb_dqopt(dquot->dq_sb)->files[dquot->dq_type];
2748 }
2749
2750 static int ext3_write_dquot(struct dquot *dquot)
2751 {
2752 int ret, err;
2753 handle_t *handle;
2754 struct inode *inode;
2755
2756 inode = dquot_to_inode(dquot);
2757 handle = ext3_journal_start(inode,
2758 EXT3_QUOTA_TRANS_BLOCKS(dquot->dq_sb));
2759 if (IS_ERR(handle))
2760 return PTR_ERR(handle);
2761 ret = dquot_commit(dquot);
2762 err = ext3_journal_stop(handle);
2763 if (!ret)
2764 ret = err;
2765 return ret;
2766 }
2767
2768 static int ext3_acquire_dquot(struct dquot *dquot)
2769 {
2770 int ret, err;
2771 handle_t *handle;
2772
2773 handle = ext3_journal_start(dquot_to_inode(dquot),
2774 EXT3_QUOTA_INIT_BLOCKS(dquot->dq_sb));
2775 if (IS_ERR(handle))
2776 return PTR_ERR(handle);
2777 ret = dquot_acquire(dquot);
2778 err = ext3_journal_stop(handle);
2779 if (!ret)
2780 ret = err;
2781 return ret;
2782 }
2783
2784 static int ext3_release_dquot(struct dquot *dquot)
2785 {
2786 int ret, err;
2787 handle_t *handle;
2788
2789 handle = ext3_journal_start(dquot_to_inode(dquot),
2790 EXT3_QUOTA_DEL_BLOCKS(dquot->dq_sb));
2791 if (IS_ERR(handle)) {
2792 /* Release dquot anyway to avoid endless cycle in dqput() */
2793 dquot_release(dquot);
2794 return PTR_ERR(handle);
2795 }
2796 ret = dquot_release(dquot);
2797 err = ext3_journal_stop(handle);
2798 if (!ret)
2799 ret = err;
2800 return ret;
2801 }
2802
2803 static int ext3_mark_dquot_dirty(struct dquot *dquot)
2804 {
2805 /* Are we journaling quotas? */
2806 if (EXT3_SB(dquot->dq_sb)->s_qf_names[USRQUOTA] ||
2807 EXT3_SB(dquot->dq_sb)->s_qf_names[GRPQUOTA]) {
2808 dquot_mark_dquot_dirty(dquot);
2809 return ext3_write_dquot(dquot);
2810 } else {
2811 return dquot_mark_dquot_dirty(dquot);
2812 }
2813 }
2814
2815 static int ext3_write_info(struct super_block *sb, int type)
2816 {
2817 int ret, err;
2818 handle_t *handle;
2819
2820 /* Data block + inode block */
2821 handle = ext3_journal_start(sb->s_root->d_inode, 2);
2822 if (IS_ERR(handle))
2823 return PTR_ERR(handle);
2824 ret = dquot_commit_info(sb, type);
2825 err = ext3_journal_stop(handle);
2826 if (!ret)
2827 ret = err;
2828 return ret;
2829 }
2830
2831 /*
2832 * Turn on quotas during mount time - we need to find
2833 * the quota file and such...
2834 */
2835 static int ext3_quota_on_mount(struct super_block *sb, int type)
2836 {
2837 return vfs_quota_on_mount(sb, EXT3_SB(sb)->s_qf_names[type],
2838 EXT3_SB(sb)->s_jquota_fmt, type);
2839 }
2840
2841 /*
2842 * Standard function to be called on quota_on
2843 */
2844 static int ext3_quota_on(struct super_block *sb, int type, int format_id,
2845 char *name, int remount)
2846 {
2847 int err;
2848 struct path path;
2849
2850 if (!test_opt(sb, QUOTA))
2851 return -EINVAL;
2852 /* When remounting, no checks are needed and in fact, name is NULL */
2853 if (remount)
2854 return vfs_quota_on(sb, type, format_id, name, remount);
2855
2856 err = kern_path(name, LOOKUP_FOLLOW, &path);
2857 if (err)
2858 return err;
2859
2860 /* Quotafile not on the same filesystem? */
2861 if (path.mnt->mnt_sb != sb) {
2862 path_put(&path);
2863 return -EXDEV;
2864 }
2865 /* Journaling quota? */
2866 if (EXT3_SB(sb)->s_qf_names[type]) {
2867 /* Quotafile not of fs root? */
2868 if (path.dentry->d_parent != sb->s_root)
2869 ext3_msg(sb, KERN_WARNING,
2870 "warning: Quota file not on filesystem root. "
2871 "Journaled quota will not work.");
2872 }
2873
2874 /*
2875 * When we journal data on quota file, we have to flush journal to see
2876 * all updates to the file when we bypass pagecache...
2877 */
2878 if (ext3_should_journal_data(path.dentry->d_inode)) {
2879 /*
2880 * We don't need to lock updates but journal_flush() could
2881 * otherwise be livelocked...
2882 */
2883 journal_lock_updates(EXT3_SB(sb)->s_journal);
2884 err = journal_flush(EXT3_SB(sb)->s_journal);
2885 journal_unlock_updates(EXT3_SB(sb)->s_journal);
2886 if (err) {
2887 path_put(&path);
2888 return err;
2889 }
2890 }
2891
2892 err = vfs_quota_on_path(sb, type, format_id, &path);
2893 path_put(&path);
2894 return err;
2895 }
2896
2897 /* Read data from quotafile - avoid pagecache and such because we cannot afford
2898 * acquiring the locks... As quota files are never truncated and quota code
2899 * itself serializes the operations (and noone else should touch the files)
2900 * we don't have to be afraid of races */
2901 static ssize_t ext3_quota_read(struct super_block *sb, int type, char *data,
2902 size_t len, loff_t off)
2903 {
2904 struct inode *inode = sb_dqopt(sb)->files[type];
2905 sector_t blk = off >> EXT3_BLOCK_SIZE_BITS(sb);
2906 int err = 0;
2907 int offset = off & (sb->s_blocksize - 1);
2908 int tocopy;
2909 size_t toread;
2910 struct buffer_head *bh;
2911 loff_t i_size = i_size_read(inode);
2912
2913 if (off > i_size)
2914 return 0;
2915 if (off+len > i_size)
2916 len = i_size-off;
2917 toread = len;
2918 while (toread > 0) {
2919 tocopy = sb->s_blocksize - offset < toread ?
2920 sb->s_blocksize - offset : toread;
2921 bh = ext3_bread(NULL, inode, blk, 0, &err);
2922 if (err)
2923 return err;
2924 if (!bh) /* A hole? */
2925 memset(data, 0, tocopy);
2926 else
2927 memcpy(data, bh->b_data+offset, tocopy);
2928 brelse(bh);
2929 offset = 0;
2930 toread -= tocopy;
2931 data += tocopy;
2932 blk++;
2933 }
2934 return len;
2935 }
2936
2937 /* Write to quotafile (we know the transaction is already started and has
2938 * enough credits) */
2939 static ssize_t ext3_quota_write(struct super_block *sb, int type,
2940 const char *data, size_t len, loff_t off)
2941 {
2942 struct inode *inode = sb_dqopt(sb)->files[type];
2943 sector_t blk = off >> EXT3_BLOCK_SIZE_BITS(sb);
2944 int err = 0;
2945 int offset = off & (sb->s_blocksize - 1);
2946 int journal_quota = EXT3_SB(sb)->s_qf_names[type] != NULL;
2947 struct buffer_head *bh;
2948 handle_t *handle = journal_current_handle();
2949
2950 if (!handle) {
2951 ext3_msg(sb, KERN_WARNING,
2952 "warning: quota write (off=%llu, len=%llu)"
2953 " cancelled because transaction is not started.",
2954 (unsigned long long)off, (unsigned long long)len);
2955 return -EIO;
2956 }
2957
2958 /*
2959 * Since we account only one data block in transaction credits,
2960 * then it is impossible to cross a block boundary.
2961 */
2962 if (sb->s_blocksize - offset < len) {
2963 ext3_msg(sb, KERN_WARNING, "Quota write (off=%llu, len=%llu)"
2964 " cancelled because not block aligned",
2965 (unsigned long long)off, (unsigned long long)len);
2966 return -EIO;
2967 }
2968 mutex_lock_nested(&inode->i_mutex, I_MUTEX_QUOTA);
2969 bh = ext3_bread(handle, inode, blk, 1, &err);
2970 if (!bh)
2971 goto out;
2972 if (journal_quota) {
2973 err = ext3_journal_get_write_access(handle, bh);
2974 if (err) {
2975 brelse(bh);
2976 goto out;
2977 }
2978 }
2979 lock_buffer(bh);
2980 memcpy(bh->b_data+offset, data, len);
2981 flush_dcache_page(bh->b_page);
2982 unlock_buffer(bh);
2983 if (journal_quota)
2984 err = ext3_journal_dirty_metadata(handle, bh);
2985 else {
2986 /* Always do at least ordered writes for quotas */
2987 err = ext3_journal_dirty_data(handle, bh);
2988 mark_buffer_dirty(bh);
2989 }
2990 brelse(bh);
2991 out:
2992 if (err) {
2993 mutex_unlock(&inode->i_mutex);
2994 return err;
2995 }
2996 if (inode->i_size < off + len) {
2997 i_size_write(inode, off + len);
2998 EXT3_I(inode)->i_disksize = inode->i_size;
2999 }
3000 inode->i_version++;
3001 inode->i_mtime = inode->i_ctime = CURRENT_TIME;
3002 ext3_mark_inode_dirty(handle, inode);
3003 mutex_unlock(&inode->i_mutex);
3004 return len;
3005 }
3006
3007 #endif
3008
3009 static int ext3_get_sb(struct file_system_type *fs_type,
3010 int flags, const char *dev_name, void *data, struct vfsmount *mnt)
3011 {
3012 return get_sb_bdev(fs_type, flags, dev_name, data, ext3_fill_super, mnt);
3013 }
3014
3015 static struct file_system_type ext3_fs_type = {
3016 .owner = THIS_MODULE,
3017 .name = "ext3",
3018 .get_sb = ext3_get_sb,
3019 .kill_sb = kill_block_super,
3020 .fs_flags = FS_REQUIRES_DEV,
3021 };
3022
3023 static int __init init_ext3_fs(void)
3024 {
3025 int err = init_ext3_xattr();
3026 if (err)
3027 return err;
3028 err = init_inodecache();
3029 if (err)
3030 goto out1;
3031 err = register_filesystem(&ext3_fs_type);
3032 if (err)
3033 goto out;
3034 return 0;
3035 out:
3036 destroy_inodecache();
3037 out1:
3038 exit_ext3_xattr();
3039 return err;
3040 }
3041
3042 static void __exit exit_ext3_fs(void)
3043 {
3044 unregister_filesystem(&ext3_fs_type);
3045 destroy_inodecache();
3046 exit_ext3_xattr();
3047 }
3048
3049 MODULE_AUTHOR("Remy Card, Stephen Tweedie, Andrew Morton, Andreas Dilger, Theodore Ts'o and others");
3050 MODULE_DESCRIPTION("Second Extended Filesystem with journaling extensions");
3051 MODULE_LICENSE("GPL");
3052 module_init(init_ext3_fs)
3053 module_exit(exit_ext3_fs)
x86 "subsystem" repository